I - T ' • • Proceedings Second International Conference on the DURABILITY OF BUILDING MATERIALS AND COMPONENTS September 14-16, 1981 National Bureau of Standards Gaithersburg, Maryland, U.S.A. SPONSORED BY CIB, International Council for Building Research Studies and Documentation RILEM, International Union of Testing and Research Laboratories for Materials and Structures American Society for Testing and Materials National Research Council of Canada National Bureau of Standards, U.S. Department of Commerce STEERING COMMITTEE Chairman: P.J. Sereda, Division of Building Research National Research Council of Canada, Ottawa Ontario, Canada (NRC representative) Members: G.J.C. Frohnsdorff, Center for Building Technology, National Bureau of Standards, Washington, DC, USA (NBS representative) R.A. Jones, Small Homes Research Council, University of Illinois, Champaign, IL, USA (ASTM representative) T. Sneck, Division of Building Technology and Community Development, Technical Research Centre of Finland, Finland (RILEM representative) E.J. Gibson, Princes Risborough Laboratory, Building Research Establishment, Aylesbury, United Kingdom (CIB representative) K.G. Martin, Division of Building Research CSIRO, Australia (Member-at-large) ORGANIZING COMMITTEE !iembers: \rm^MMm&MM$;\Z, LYRASIS'M^fffbefs and Sloan Foundation J.F. Collins, Alternative Materials Utilization Branch, Department of Energy, Washington, DC E. Kennett, American Institute of Architects Research Corporation, Washington, DC L.W. Masters, Center for Building Technology National Bureau of Standards, Washington, DC F. Matzke, National Institute of Building Sciences, Washington, DC E. Passaglia, Center for Materials Science, National Bureau of Standards, Washington, DC W.R. Prindle, Corning Glass Works, Corning, NY B. Horner (Secretary ) , Center for Building Technology, National Bureau of Standards, Washington, DC G. Pignone, (Local Arrangements), Public Information Division, National Bureau of Standards, Washington, DC http://www.archive.org/details/secondintematioOOinte . ; Second International Conference on DURABILITY OF BUILDING MATERIALS AND COMPONENTS September 14-16, 1981 PREPRINTS Conference sponsored by: International Council for Building Research Studies and Documentation, CIB International Union of Testing and Research Laboratories and Materials and Structures, RILEM American Society for Testing and Materials, ASTM National Research Council of Canada, NRC National Bureau of Standards, NBS Preprints compiled by: Geoffrey Frohnsdorff and Barbara Horner National Bureau of Standards U.S. Department of Commerce National Bureau of Standards Washington, DC 20234 FOREWORD We are very pleased that the Second International Conference on the Durability of Building Materials and Components is being held at the National Bureau of Standards. The National Bureau of Standards co- sponsored the successful first International Conferenced hosted by the National Research Council of Canada in Ottawa in 1978 and it is gratify- ing to see the series of conferences planned at that time becoming a reality. Since its founding in 1901, the National Bureau of Standards has made many contributions to the understanding of the performance of building materials. In spite of 80 years work behind us, we recognize that much has still to be done to develop the methodologies and the data bases which are needed by the construction industry if it is to make the best use of its material resources. We must bring fundamental knowledge and systematic approaches such as those which have been so successful in the aerospace and electronics industries, to bear on the problems of durabil- ity and reliability of building materials. We believe that this series of conferences will play an important role in promoting contact between researchers and users of durability knowledge and helping insure that the knowledge is used effectively in the materials selection process. We are grateful to the national and international organizations which have joined with us in sponsoring this conference. The problems we face in considering the durability of building materials have no national boundaries and progress in solving them must be expected to be made more quickly as a result of this international cooperation. Richard N. Wright, Director Center for Building Technology National Engineering Laboratory National Bureau of Standards John B. Wachtman, Director Center for Materials Science National Measurement Laboratory National Bureau of Standards September 14, 1981 ii ACKNOWLEDGMENTS We wish to acknowledge a number of people who have made special contributions to the preparation of this volume. First, we wish to thank Mr. M. Fickelson, Secretary General of RILEM, for the French translations of the abstracts. We also wish to thank the staff of the Special Activi- ties Division of NBS, particularly JoAnne Lorden, Greta Pignone, and Sara Torrence for their advice and help at all stages of the work. Next, we wish to thank the members of the Word Processing Center of the NBS Center for Building Technology, particularly, Charity Starr, Ulesia Gray, Tracey Kistler, Mary Ramsburg, Brenda Thompson, and Rebecca Meyer for their efforts which ensured that the printing deadline could be met. Last, but not least, we thank our many colleagues in the Building Materials Division who have assisted by reviewing the papers. Geoffrey Frohnsdorff Barbara Horner iii TABLE OF CONTENTS Page FOREWORD ii ACKNOWLEDGEMENTS iii ECONOMICS AND INSTITUTIONAL FACTORS ECONOMIC CONTEXT FOR THE COST-BENEFIT EVALUATION OF ALTERNATIVE LEVELS OF DURABILITY IN BUILDING MATERIALS 1 E.B. Berman, Edward Berman Associates. INSTITUTIONAL FACTORS AFFECTING DURABILITY PERFORMANCE OF BUILDING MATERIALS 8 R. Dillon, National Institute of Building Sciences. ECONOMIC ASPECTS OF THE DURABILITY OF BUILDING MATERIALS: AN EXPLORATORY ANALYSIS 13 A.S. Rakhra, National Research Council, Canada. PROBLEMS IN THE IMPLEMENTATION OF DURABILITY ASPECTS 22 T. Sneck, Technical Research Centre of Finland. ENVIRONMENTAL FACTORS ENVIRONMENT, MICROCLIMATE AND THE DURABILITY OF BUILDING MATERIALS 28 H. Ashton and P.J. Sereda, National Research Council of Canada. QUANTITATIVE CONSIDERATIONS OF MOISTURE AS A CLIMATIC FACTOR IN WEATHERING 32 K. Martin and R.E. Price, Division of Building Research, CSIRO, Australia. IV Page ESTIMATION OF TOTAL SURFACE AREA AND SPATIAL DISTRIBUTION OF EXPOSED MATERIALS FROM COMMONLY AVAILABLE INFORMATION FOR U.S. METROPOLITAN AREAS 39 M.D. Koontz and J .E . McFadden, GEOMET Technologies, and F.H. Haynie, U.S. Environmental Protection Agency. PREDICTION OF SERVICE LIFE METHODOLOGIES AND TEST METHODS FOR SERVICE LIFE PREDICTION 63 T. Sneck, Technical Research Centre of Finland. THE DEGRADATION AND PREDICTION OF SERVICE LIFE OF BUILDING COMPONENTS 69 Yoshitaka Ishizuka, Ministry of Construction, Japan. PREDICTION OF THE SERVICE LIFE OF STRUCTURES EXPOSED TO FROST ACTION 76 G. Fagerlund, Cementa AB, Sweden. TIME TRANSFORMATION FUNCTIONS COMMONLY USED IN LIFE TESTING ANALYSIS 77 J. Martin, National Bureau of Standards. BUILDING STONE AND TERRA COTTA S0 2 and N0 2 ATTACK ON MARBLE 88 K. L. Gauri and J. Gwinn, University of Louisville. MEASUREMENTS OF WET AND DRY DEPOSITION ON MARBLE 93 R. Livingston, M. Kantz, P. Brown and J. Gorsheimer, U.S. Environmental Protection Agency. EROSION OF MARBLE 95 F. Haynie, Environmental Research Center, U.S. Environmental Protection Agency. ARCHITECTURAL TERRA COTTA 106 T. Prudon, The Ehrenkrantz Group PC and J. Stockbridge, Wiss, Janney, Elstner, Associates, Inc. v Page DEGRADATION AND REHABILITATION OF TERRA COTTA 108 S. Thomasen, Wiss, Janney, Elstner, and Associates, Inc. GYPSUM PRODUCTS DURABILITY OF GYPSUM BOARD 115 L. Eby and B. Veschurof, United States Gypsum Company. CONCRETE AND OTHER CEMENTITIOUS MATERIALS AGGREGATE QUALITY FROM MULTIVARIATE STATISTICAL ANALYSIS OF AGGREGATE PROPERTIES 126 P. Hudec, University of Windsor, Canada. AIR CONTENT OF PLASTIC AND HARDENED CEMENT 136 D. Reidenouer and R. Howe, Pennsylvania Department of Transportation . ABSORPTIVITY, A MEASURE OF CURING QUALITY AS RELATED TO DURABILITY OF CONCRETE SURFACES 153 E. Senbetta and C. Scholer, Purdue University. ABNORMAL FREEZING OF WATER IN HARDENED CEMENT PASTE 160 M. Setzer, Technical University, Munich, West Germany. DEVELOPMENT AND INTERACTION OF A CONCRETE ADDITIVE FOR IMPROVED PERFORMANCE AND DURABILITY 170 S. Syal, University of Illinois, Urbana, Illinois. S. Kataria, Feeders India Consultants. THE DURABILITY OF PREFABRICATED REINFORCED CONCRETE EXTERNAL WALLS AND CLADDINGS IN BUILDINGS 188 J. Kami, Israel Institute of Technology. TESTING AND ESTIMATION OF DURABILITY OF CONCRETE AND RADWASTE CONCRETE FOR LONG-TERM STORAGE 194 S. Pihlajavaara, Technical Research Centre of Finland. vx Page CORROSION OF REINFORCEMENT IN CONCRETE BRIDGES AT DIFFERENT AGES DUE TO CARBONATION AND CHLORIDE PENETRATION 199 A. Volkwein and R. Springenschmid, Technical University, Munich, West Germany. CORROSION OF STEEL IN REINFORCED CONCRETE IN MARINE AND OTHER CHLORIDE ENVIRONMENTS 210 R. Browne, Taylor Woodrow Research Laboratories, England. THE DEVELOPMENT OF A METHOD OF PREDICTING THE WEATHERING BEHAVIOR OF GLASS REINFORCED CEMENT COMPOSITES 225 D. Oakley, D. Litherland and B. Procter, Pilkington Brothers, Ltd., England. DURABILITY CRITERIA FOR CELLULAR CONCRETE OUTER WALL 232 S. Nakano and S. Tada, Misawa Homes Institute of Research and Development , Japan . DURABILITY PERFORMANCE OF POLYMER-MODIFIED MORTARS 242 Y. Ohama, College of Engineering, Nihon University, Japan. GLASS PRACTICAL CONSIDERATIONS IN GLASS FRACTURE 249 T. Schwartz, Simpson, Gumpertz, and Heger, Inc. INTERACTION OF CURTAIN WALLS AND METAL EDGE BANDED INSULATING GLASS UNITS 264 T. Schwartz and M. Zarghamee, Simpson, Gumpertz, and Heger, Inc. THERMAL INSULATION MATHEMATICAL MODELS FOR CORROSION IN BUILDING INSULATION SYSTEMS 274 J. Pommersheim and John Lobo, Bucknell University, and J. Clifton, National Bureau of Standards. VII Page WOOD THE PHOTODEGRADATION OF WOOD DURING SOLAR IRRADIATION 279 E. Miller and H. Derbyshire, Building Research Establishment, England . INFLUENCE OF CLIMATE UPON HAZARDS FOR WOOD DECAY 288 R. DeGroot, Forest Products Laboratory, Forest Service, U.S. Department of Agriculture. PAINTS INITIAL DEGRADATION OF CORROSION PROTECTION BY ORGANIC COATINGS ... 297 D. Thomas, Lehigh University. MATHEMATICAL MODELS FOR THE CORROSION PROTECTIVE PERFORMANCE OF ORGANIC COATINGS 305 L. Masters, National Bureau of Standards and J. Pommersheim, Bucknell University OUTDOOR DURABILITY AND PERFORMANCE OF PREPAINTED, COLD ROLLED, GALVANIZED AND ALUMINIZED STEELS 313 C. Christ, Armco, Inc. TENSILE ADHESION TESTING FOR MEASURING COATING DURABILITY 327 H. Bleile, S. Rodgers and David W. Taylor Naval Ship R&D Center. DEVELOPMENT OF A PERFORMANCE DATA BASE FOR MARINE COATINGS 336 H. Bleile, T. Radakovich, M. Magner, and S. Rodgers David W. Taylor Naval Ship R&D Center. PLASTICS PREDICTION OF THE SERVICE LIFE OF SYNTHETIC POLYMERS 347 L. Hawkins, Bell Laboratories, Retired. viii Page DEGRADATION PROCESSES OF BUILDING MATERIALS AND COMPONENTS: A SHORT REVIEW OF SOME PROPOSALS FOR RESEARCH 354 P. Eurin, Centre Scientifique et Technique du Batiment, France. DETERIORATION PROCESS OF POLYMER MATERIALS AND ITS CORRELATIONS WITH DEPTH FROM SURFACE 367 T. Fukushima, Ministry of Construction, Japan. AGING OF POLYMERS IN BUILDINGS UNDER SIMULTANEOUS ACTION OF ARTIFICIAL WEATHERING AND MECHANICAL STRESS 378 R. Kwasny, H. Sasse and I. Schrage, Institute for Bauf orschung , West Germany. WEATHERING OF SEVERAL POLYMERS AFTER FOUR YEARS OF EXPOSURE 387 R. Cope and G. Revirand, Centre Scientifique et Technique du Batiment, France. DURABILITY OF PVC WINDOW FRAMES 410 P. Svane, Teknologisk Institut, Denmark. DURABILITY OF GLASS REINFORCED POLYESTER USED IN CLADDINGS FOR BUILDINGS 412 A. Brookes, University of Liverpool, England. EARLY DETECTION OF POLYMERIC DEGRADATION USING ELECTRON MICROSCOPY 424 M. McKnight and W. Byrd, National Bureau of Standards. APPARATUS FOR ACCELERATED WEATHERING OF BUILDING MATERIALS AND COMPONENTS 435 T. Gjelsvik, Norwegian Building Research Institute. ROOFING A PRELIMINARY EVALUATION OF THE TENSILE AND ELONGATION PROPERTIES OF SINGLE-PLY SHEET ROOFING MEMBRANE MATERIALS 442 R. Mathey and W. Rossiter, National Bureau of Standards. IX Page METHODS FOR THE STUDY OF AGING OF BITUMEN-POLYMER SBS MATERIALS ... 452 J. Marechal, Centre Scientif ique et Technique du Batiment, France. FATIGUE BEHAVIOR OF ASPHALTIC ROOFING MEMBRANES 463 P. Nelson, Simpson, Gumpertz, and Heger, Inc. ROOFING HISTORY SURVEY 464 D. Van Court, Western Electric Company. AUTHOR INDEX AND CROSS-REFERENCE TO CONFERENCE PROGRAM 469 x AN ECONOMIC CONTEXT FOR THE COST BENEFIT EVALUATION OF ALTERNATIVE LEVELS OF DURABILITY IN BUILDING MATERIALS Edward B. Berman Edward B. Berman Associates, Inc. Marblehead, Massachusetts 01945 Abstract: An economic structure is developed below for the cost benefit evaluation of alternative levels of durability in building materials. The structure has the following characteristics: 1. The evaluation of building material alternatives is imbedded into a life-cycle cost benefit analysis of the structures constructed out of them. 2. The alternatives entered into the evaluation are assumed to be discrete and fully defined in materials, design, building procedure, and operations and maintenance procedure . 3. The useful life of the structure is developed from overlapping, interactive distributions of physical life and economic life, either of which can be affected by durability. 4. A simple replacement model, designed earlier for the National Bureau of Standards, is assigned the generation of distributions of economic life and the estimation of discounted life-cycle system cost for specified sequences of replacements. 5. Finally, a simple Monte Carlo model is suggested for the evaluation of the three-way tradeoff of initial cost, operating cost, and useful life. Keywords: Building materials; cost benefit evaluation; durability; economics; economic life; life-cycle costs; models. Imbedding : Imbedding develops from a simple principle of systems analysis which states that the analysis should be pursued at the lowest system level at which: 1. Substantially all of the differentiation among the alternatives is captured (thus, all other things do in fact remain equal). 2. It is possible to obtain the necessary data and pursue the analysis. Durability alternatives are likely to force the analysis up to the level of the whole building since durability is likely to affect the useful life of the building, either by affecting the physical life, and/or by affecting annual maintenance costs, and hence the optimal economic life. Discrete Alternatives ; The analysis will deal with discrete (i.e., not continuous) alternatives that are fully defined in: materials specifications design of structure building procedure initial cost operations and maintenance concept annual costs for operations and maintenance distribution of physical life time pattern in initial cost for successor buildings time pattern in operations and maintenance costs for building and successors . The purpose of the evaluative structure is to compare two or more such discrete alternatives. Distributions of Physical and Economic Life : Figure 1 shows conditional probability distributions of system discarding for physical and economic reasons. These probabilities are the probabilities of failure or discard in a year from each cause on condition that the system survives to the beginning of the year . Figure 1 shows a shift of the physical failure distribution towards the right (from P^(t) to P2(t)) representing an increase in physical durabil- ity; but durability could also shift the economic discard distribution to the right by lowering annual maintenance costs. The conditional probabilities of failure for physical and economic reasons each tend to rise monotonically over time. The sum of P(t) and E(t) will tend to approach 1.0 asymptotically. An unconditional probability distribution of system discard in year t, f(t), is generated by multiplying the probability of survival to year t by the sum of conditional probabilities of discard in year t: t-1 f(t) - [ TT (1 - Pi - E ± ) ] (P t + E t ) (1) i=l Figure 2 shows f(t) as a unimodal distribution. Thus, in early years, f(t) is small because the conditional probability of discard in the year is small. In the late years, f(t) is small because the probability of survival to year t is small. In between, where both probabilities are substantial, f(t) reaches its peak. The f(t) distribution represents useful life. An increase in durability could shift either P(t) or E(t) towards the right; either effect would shift f(t) towards the right. An average of the f(t) distribution is shown as point a in figure 2. This average constitutes the expected useful life of the building. The P(t) and E(t) functions are not independent. The shift to the right in P(t) would decrease the probability of physical failure early in any year, and thus increase the probability of economic discard, to the extent that such discard was obviated by earlier physical failure. Thus, the shift to the right in P(t) would cause a smaller shift to the left in E(t). This interdependence does not exist in the Monte Carlo approach described below. The Discount Rate ; In estimating life-cycle system costs, a discount rate is used on future costs and benefits to reduce them to equivalent present values, in which form the costs and benefits of different years may properly be aggregated to obtain total cost and total benefit for the system in present value. Discounting reflects the fact that a dollar of present cost is more costly than a dollar of cost next year because, if not needed this year, the dollar could be invested in something useful until needed next year. Similarly, a benefit available earlier could be invested over the intervening period. The formula: is used to discount future values to values in year 1, where C^ is the cost (or benefit) in present value, C t is the cost (or benefit) in year t value, and d is the discount rate. Economic Life and Life-Cycle Cost : The firm of Edward B. Berman Associates, Inc. has recently submitted a report, Methodology for Cost- Benefit Evaluation of Materials Degradation Programs (April 1981), to the National Bureau of Standards, Center for Materials Sciences. The report includes a design for a computer model of economic replacement which would adapt well to the needs of a three-way analysis of initial cost, annual cost of operations and maintenance, and useful life for durability alternatives. The model will estimate a life-cycle system cost for an existing item of capital at alternative times of first replacement, with an optimized use- ful life for the successor capital. The relevant capabilities for our present purposes in evaluating durability alternatives are: 1. The generation of optimal useful lives for successor capital for each alternative future year of purchase. 2. The ability to handle empirical functions for initial cost and costs of operations and maintenance, controlled by: the year of purchase the years since purchase. 3. A costing module which will generate a discounted life-cycle system cost for any specified sequence of replacements. The Three-Way Tradeoff Analysis of Durability Alternatives ; As has been noted above, a modification in the durability of a building material can affect (1) the annual cost of operations and maintenance, and (2) either or both of the distribution of physical life and the distribution of eco- nomic life. The distributions of physical life and economic life in turn determine the useful life of the building. Thus, in order to evaluate durability alternatives for building materials, we must deal with the three-way tradeoff among: initial cost annual cost of operations and maintenance useful life. Before proceeding to a description of a structure for evaluating the three-way tradeoff, it would be appropriate first to illustrate the mechanism of interaction between the two-way tradeoff of initial cost and annual cost of operations and maintenance on the one hand, and use- ful life on the other. In order to illustrate the point, let us consider what would seem to be a two-way tradeoff between initial cost and the annual cost of operations and maintenance . An alternative which, as compared to another alternative, offers higher initial cost against lower annual operations and maintenance cost (a typ- ical situation with increased durability) with an equal discounted pre- sent value over the optimal useful life, would appear to leave the buyer indifferent between the alternatives. That alternative (with increased durability) will however change the optimal useful life (specifically, it will tend to increase it) and will thus be a preferred alternative. The reason for the impact on useful life is that the higher initial cost, being a sunk cost when it becomes time to consider replacement, will not enter into the optimal replacement decision, whereas the lower annual cost of operations and maintenance will very much be involved with the replacement decision, and will tend to make that alternative more resis- tant to replacement. Thus, the apparent two-way tradeoff becomes a three-way tradeoff. The structure for evaluating the three-way tradeoff must do the following 1. It must treat the interactive effects among initial cost, annual operations and maintenance cost, and optimal economic replacement. 2. It must treat the interactive effects between maximum physical life and optimal economic life. 3. It must evaluate a sequence of successive replacements, convert them into a time stream of initial costs and annual costs of operations and maintenance, and then discount and sum those costs into a single present value. A simple Monte Carlo model is suggested below which, working in combination with the Center for Materials Sciences Replacement Model described above, would be capable of evaluating the three-way tradeoff of initial cost, annual cost of operations and maintenance, and useful life. A Monte Carlo model is one which evaluates a stochastic situation involv- ing probability distributions (here, the distribution of physical life) by: 1. Selecting a random number each time a distribution is encountered. 2. Interpreting those random numbers into specific values from the distributions . 3. Running many iterations and averaging the outcomes to derive an expected outcome. The use of the random number to select a value from a probability distribution is illustrated in figure 3 for the case of a distribution of maximum physical life. Note in figure 3 that the cumulative form of the distribution is used. If a 3-digit random number is to be used, the cumu- lative frequency is normalized to sum to one thousand, as in figure 3. The 3-digit random number is used to enter the probability distribution on the vertical axis, and a specific maximum physical life is selected on the horizontal axis as that maximum physical life which is consistent with the random number interpreted as a cumulative frequency. The procedure for using the two models, the Monte Carlo model and the Center for Materials Sciences Replacement Model, would be as follows: 1. For any one of the discrete durability alternatives which is to be evaluated, a single run of the Replacement Model would be used to calculate a time series showing the optimal economic useful life for each alternative future year of purchase. This is a normal output for the Replacement Model . 2. A distribution of maximum physical life, f p (t), would be calculated: h M = [TT (1 - p i } 1 (p t> (3) i=i and the cumulative form would then be used as the basis for the Monte Carlo determination of specific maximum useful lives, as needed for the simulation. 3. The Monte Carlo model would then be run for a number of iterations (specifically, enough iterations to obtain a stabilized result). Each iteration would follow the procedure: a. Each iteration would cover a number of years T sufficient to incorporate several replacements. 5 b. An economic useful life for the first purchase would be drawn from the output of the Replacement Model by selecting the optimal useful life for purchase in year 1. c. A maximum physical life would be selected for the first purchase by taking a random draw, and interpreting the number into a maximum life. d. The lesser of maximum physical life and optimal economic life would be used as the life of the first purchase, L]_. e. The process would be repeated for the year Lj_ + 1 to determine a life 1,2 for the second purchase. f. The process would be repeated until the sum of L^ over i was equal to or greater than T. g. The Replacement model would then be entered to generate a discounted life-cycle cost for the specified sequence of replacements. 4. After completion of N iterations, an average of the costs for the N iterations would be calculated. This average cost would represent the life-cycle system cost for the discrete durability alternative. 5. By repeating steps 1 through 4 above for each durability alternative to be evaluated, a set of fully comparable life-cycle system costs will be generated which would provide a basis for a recommendation, from an economic point of view, of a preferred alternative. The models which are incorporated into the procedure are simple and straightforward. Both the Monte Carlo model and the Replacement Model have the advantage of being able to accept empirical functions represent- ing costs and probabilities. Although neither model has been implemented as a computer program as yet, neither represents a difficult challenge. FIGURE 1 CONDITIONAL PROBABILITY OF DISCARD IN YEAR t Pl(t) P 2 (t) YEAR t CONDITIONAL PROBABILITY DISTRIBUTION OF DISCARD REPRESENTING BOTH MAXIMUM PHYSICAL LIFE AND OPTIMAL ECONOMIC LIFE WITH INTERACTION UNCONDITIONAL PROBABILITY OF DISCARD IN YEAR t FIGURE 2 YEAR t USEFUL LIFE DISTRIBUTION FIGURE 3 1000 CUMULATIVE FREQUENCY RANDOM NUMBER A y i SPECIFIC LIFE MAXIMUM PHYSICAL LIFE USING A RANDOM NUMBER TO SELECT A MAXIMUM PHYSICAL LIFE INSTITUTIONAL FACTORS AFFECTING THE DURABILITY PERFORMANCE OF BUILDING MATERIALS AND COMPONENTS Robert M. Dillon, AIA, ASCE National Institute of Building Sciences Washington, D.C. 20005 Abstract: Quite apart from the availability, selection, and application of materials to achieve durability, is the value placed on durability by various elements of society that affect and are effected by building and buildings. First cost, functional life and replacement cost, all play a part in the selection and use of materials, as does the intensity of use and abuse. The financier is concerned about the security of the investment in rela- tion to the term of that investment. The insuror is concerned about the risk in relation to the ability to pay. The owner/user is concerned about first cost, maintenance cost, and replacement cost in relation to his ability to pay and the period of intended ownership or use--and the nature of that use. Durability of materials and components is a dubious virtue if first cost makes the investment impossible or shaky. A useful life that extends far beyond functional life is not a prudent investment. Conversely, lack of adequate durability can shorten functional life and destroy an investment. Each nation—and indeed, each segment of society and often each locale- has its own perceptions of the value of durability and its own institu- tions for coping. It is important, however, to separate health and safety issues from what are essentially livability and marketability issues. The two tend to become entwined and institutions, public and private, battle mightily over who should set the value structure and over how it is to be articulated. Key Words: Optimum performance life; cost-benefit-risk; life cycle costing; functional obsolescence; market valuation of durability and performance; institutional positions on conservation of land and materials. It was just about 25 years ago that similar debates on the optimum life of building materials, products, components, and the whole of houses and buildings were in full bloom. The immediate post-war period had been one of lively building activity, population mobility, innovation and attempts to transfer wartime technologies to building markets, and especially a gathering of the building disciplines. It was an exciting period and an important one in the history of building science and technology. As each innovation was offered for approval and as each failure was analyzed for cause, the inevitable questions were: How well and for how long should the given element perform? How should performance be defined, and how can it be predicted? As answers to these questions were sought, the performance concept and approach emerged, and the beginnings of life-cycle costing and cost- benefit-risk analysis as well. The issue of "optimum performance life" springs from deep, intertwined, and shifting factual and philosophical roots. As a general rule, the type of facts involved are not immutable laws of nature; rather, those that are supportable only at a point in time, subject to change with new knowledge or circumstance. And, philosophically, optimum performance life certainly is a relative term. There are those who believe that gregarious living is either the natural desire or the lot of man; there are those who don't. There are those who believe we are consuming land and resources for building at too fast a rate; there are those who don't. There are those who feel we should preserve our historical sites, not just isolated structures of note but whole areas; there are those who don't. These are all yery serious matters in the minds of many and their debate is necessary if for no other reason than to increase the range and depth of knowledge on the issues involved—and to produce reasonable consensus so that society may function. Rarely are positions on these issues universally held--that is, most such positions are shades between the extremes or selective, and, not infrequently, individuals may hold views that are inconsistent with each another. For example, we need more housing, more recreational facilities, more jobs, but don't build them here. The great clash in streams of thought rose in the 1960s. There was a major drive to increase housing production and everything that goes with it, and a major drive to preserve the environment and existing life styles—growth and no-growth at the same time; increased consumption and decreased consumption at the same time. New institutions were created to promote and/or to defend each position, and even coalitions of insti- tutions. Existing institutions reeled under the onslaught and are con- tinuing to do so. In effect, we have been suffering from a tyranny of minority views. The result has been--at least in building--the worst housing production crisis in 40 years and the highest prices in history. It cannot all be blamed on the high cost of oil. Most of our basic institutions were founded on the principal of a free market, with government playing a key role in preventing "too much" concentration of wealth and in curbing excesses. The system has now run amok and even our best minds seem quite unable to describe what has happened let alone what is likely to happen. Nevertheless, there clearly is a mood to unravel the knot. How long this will last and what the consequences will be is anyone's guess. All this may seem to be a far cry from the issue of durability of build- ing materials and components but it is not. In housing and building technology, we are caught in this cross-current of institutional pres- sures. The basic parameters of optimum performance life have not changed all that much—functional obsolescence, destruction through use and abuse, and first cost versus operation, maintenance and replacement cost. Functional obsolescence applies to the use of land as well as to struc- tures and their constituent parts. We can and have improved our ability to facilitate structure change to accommodate functional change, but we have accomplished much less in our ability to predict and articulate desired settlement patterns. It is difficult in the extreme to cope with the dynamics of human settlements as community planners well know. If mistakes are made, people and businesses will vote with their feet. Du- rability of structures in such instances can be a deterrant to needed changes in land use, or at least wasteful of resources where permanence is achieved at higher cost. It is unlikely that this dilemma will be readily resolved, although greater attention to community planning can certainly help. We also have learned a great deal about the use and abuse of structures and their constituent parts. Large owner-users of housing and buildings that have a long-term investment interest in properties have developed systems to track operation, maintenance and repair costs. The data collected, however, are not shared as much as they should be. One insti- tutional change that can and should be made is a legal uncoupling of per- formance experience data sources and liability. The U.S. building com- munity has always suffered from a lack of performance feedback on a large enough scale to enable all participants to better guage optimum performance life. Further, particularly in public and rental housing, there has seemingly been an unwillingness of our governmental institu- tions to support the provider against the user — specifically the abusing user. It is difficult to create livable yet indestructible environments. Learning to match durability and use is necessary, but having to cope with willful destruction of property is quite another matter—it simply should not be necessary. Finally, the issue of first cost versus life cycle cost provides perhaps the greatest dilemma of all. Our financial institutions simply have not found a way to alter the relationship between cost and the ability to pay, nor cost and the willingness to pay. Life cycle costing has meaning for the investor only for the ownership life he has in mind. When the payment is beyond that life, the concern shifts to marketability. The financier or insurer should have a concern 10 that extends well beyond that of the mobile individual or speculative investor, but this tends to be true only when and where the market is or is expected to be competitive. When there is a scarcity, there is little incentive to be concerned about marketability. This is certainly the case today in the area of housing. Institutional practices undoubtedly could be changed to decrease specula- tive opportunities and thus increase the value of durability, but doing so would have far-reaching implications. We do know that past efforts to artificially give increased valuation to qualities such as durability have not been supported by free market valuation. Where increased dura- bility translates into increased first cost, the choice may quickly come down to a go or no go decision. In housing this can mean fewer buyers can be qualified. More often than not when the decision is no housing or housing of lesser durability, the choice will be lesser durability if it is recognized or considered at all. Faced with the like decision of durability that will be adequate for the period of intended use and a higher first cost, the same is likely to be true. Obviously, informed owner-users in the building market would prefer the best of both worlds — adequate durability and low cost. But forcing the first without achieving the other — which has so often been the case with actions taken by our regulatory institutions—does little more than put the product out of reach. The building technologist can be of greatest help by doing all possible to provide decision-makers— institutions and individuals—with the know- ledge and tools to make sound decisions. The institutional dialogue on the best national posture regarding re- source utilization in relation to building and development is likely to become more rational only as the base of knowledge as to the available options becomes more fully developed and articulated. The same may be said of the market-place decisions that must support any national policy that may emerge. Needed is clear information on the predictable performance life of materials, products, components, and systems under given performance conditions and at given first cost and operation and maintenance costs. Efforts to extend performance life and reliability generally will be a plus. However, equal effort needs to be expended to be sure that per- formance capabilities are not excessive and thus wasteful. Needed also, is a much better system for determining optimum performance life of buildings and their constituent parts given a range of use and user assumptions. In a free-market economy it must be assumed that individuals and entities, given adequate information and analytical tools, will, in the aggregate, make better decisions than central planners. It must be assumed also, 11 that suppliers will seek to and will satisfy market needs if they are free to do so and the market is or can be defined. Durability obviously is a significant factor in such decisions and will be affected by national policies as to the preservation or utilization of land and other resources However, durability must be regarded as a relative attribute and one that will doubtless change over time. The need is to provide the grist for progressively more substantive debate and the basis for progressively more informed decision-making. 12 ECONOMIC ASPECTS OF THE DURABILITY OF BUILDING MATERIALS: AN EXPLORATORY ANALYSIS A.S. Rakhra Division of Building Research National Research Council of Canada Ottawa, Ontario K1A 0R6 Canada Abstract: This paper discusses briefly the place and role of building materials in economic activity in Canada. It explores the "economics" of the durability of building materials, especially in relation to life-cycle costs, social costs, manufacturing options, and energy consumption. The paper concludes that the activities of the building industry have a significant influence on the levels of output, income and employment in the Canadian economy. In some circumstances, economic factors favour the use of more durable materials; in others they favour less durable ones. Key Words: Building materials; construction in Canada; durability of building materials; life-cycle costing; social costs. MATERIALS AND CONSTRUCTION ACTIVITY From the economic standpoint construction materials may be classified under three main headings: basic or primary materials such as sand and clay, which require little or no processing before use; semi-processed materials such as cement and timber, which require a limited amount of preparation; and fully-processed materials such as fabricated steel and plastics, which require manufacturing. This breakdown is significant because the economic resources used in the form of capital and labour are usually proportional to the amount of processing required. Between 1961 and 1974, the latest year for which input-output data are available, the value of materials used in building and engineering construction in Canada was roughly 35 per cent of the value of the completed construction. The figures in Table I, show that consumption of materials of all kinds in the economy as a whole was lower than the figure for construction by between 5 and 8 percentage points. Consumption of construction materials, however, grew by only 4.4 per cent on an annual compound basis, from $3,741 million to $6,318 million. At the same time the growth rate for all materials was 5.9 per cent. Figure 1 shows the growth in the value of major groups of construction materials between 1961 and 1974, again in terms of constant (1971) dollars. For most of this period the fastest-growing group was metal- fabricated materials, followed by non-metallic minerals and plastics. 13 In constant (1971) dollars the value of construction work performed rose to $19,258 million in 1977 and marginally further to $19,521 million in 1979. In current (1979) dollars the latter figure was $42,371 million, or 16.3 per cent of the Gross National Product (GNP) . Of this, $24,437 million was spent on buildings and, if the 35 per cent portion was still valid, $14,830 million was spent on materials for building and engineering work. The available data do not permit calculation of a precise value for the annual consumption of building materials in Canada, but assuming that the building sector accounted for 60 per cent of all construction activity in 1979, it would contribute approximately 9.8 per cent to the GNP. Building materials would therefore contribute between 3 and 4 per cent of GNP. CONCEPTS OF DURABILITY These concepts have been identified in a number of ways. For example, Carruthers 2 has written that "durability" is one aspect of the "performance with time" of a building product, component, or element and of the building itself. The durability of building materials may be defined in terms of resistance to changes in state or property with time or to deterioration and decay. It may also be defined in terms of performance in relation to specifications, in terms of use, operation and maintenance, and in terms of behaviour in relation to other components and materials or to the physical and chemical environments in which each operates. The durability of building materials may be associated with service life. But neither durability nor service life are necessarily synonymous with economic life. Materials may, for example, remain in service long after they have ceased to be economically attractive to use or operate. A material's economic life may also be influenced by technical obsolescence, changing tastes, building codes or other standards, and by the arrival on the market of a technically equivalent substitute with lower initial and in-service costs. In economic terms, durability is a relative rather than an absolute concept. It involves life-cycle costing. It requires that the materials competing for use in a particular application be ranked using initial, operating, and repair costs, while meeting the given objectives of the building. If two or more of these materials are ranked as economic equals, the final choice must be made on non- economic grounds. To sum up, the durability of a building material can be defined, in economic terms, as the time period over which it continues to meet the objectives of the building in which it is used at the minimum life-cycle cost. This matching of time period with objectives and costs should lead to the proper allocation of the physical resources involved, which is the principal objective of economic analysis. 14 AN ECONOMIC FRAMEWORK The demand for building materials has three principal components. The requirements for materials in new buildings, those for retrofitting, rehabilitation and other repair activities, and those recycled or rescued materials from demolished buildings that are fit for use in new construction or repairs. Materials in the third classification will effectively reduce over-all demand in the other two. Mathematically, this demand can be expressed as: D* = D t + D* - rD t_1 (1) n e m where D = total demand for materials, for period t; D = demand for materials for new buildings, for period t; D = demand for materials for repairs, etc., for period t; D = materials made available by demolition of buildings in period t-1; r = proportion of materials recycled from buildings demolished in period t-1, for period t. As stated earlier, for every $1 spent on building and engineering, roughly $0.35 is spent on materials. This same proportion has been assumed by the building sector itself, and corresponds to D 1 - in Eq. (1). There are, however, no statistics available that allow this figure to be broken down into three parts. In practical terms, rD^ - -"- would be especially difficult to measure. Nevertheless, the equation does show that if rD^ - ! and D| are influenced by one factor or another, then D^ will change, assuming D 1 - remains unchanged. One such factor is energy conservation. While additional insulating materials are required for new buildings, the demand for these materials for retrofitting and repairs will also increase, leading to a larger total demand -- D* -- if rD t_ l remains unchanged during the period in question. Another factor is the aging of the building stock. As buildings approach the end of their useful lives and face demolition, less and less will be spent on repairs. If more buildings are at this stage than new ones being built, D will fall from its previous levels. This decline will be accelerated if recycled materials from buildings already demolished are used increasingly for repairs. DURABILITY AND LIFE-CYCLE COSTS It is becoming common to consider not only the initial costs of a new building, but also the net present value of the costs likely to be incurred for maintenance and repair, renovation, etc., during its expected lifetime. Using life-cycle cost criteria, the building having the lowest initial cost might not necessarily be built. 15 The durability of the materials used in a building will influence its life-cycle costs. This is illustrated in Figure 2, in which Dq and Dj represent repair cost curves for less- and more-durable materials. As in the case of buildings, the costs of repairing materials declines as they approach the end of their lifetimes. The difference between Tq and Ti represents the extended lifetime possible using D^, and the area under the Dj curve, representing the lifetime repair costs, will be at least smaller than the area under the Dq curve, and at worst equal to it. Another way of representing the economic consequences of material durability is shown in Figure 3. In this case, two materials, Y and Y', begin their service lives with similar physical properties, represented by Yq. Y 1 deteriorates at rate r', which is slower than r. The curves Yt = YQe" rt and Y£ = YQe~ r t represent the process of deterioration, or loss of durability. Clearly Y* should have a longer service life than Y. The area between the two curves indicates the potential savings that will accrue from using Y' rather than Y during the period from tg to t2« Initially, Y 1 may cost more per unit to produce than Y, offsetting at least some of the potential savings over the years and reducing the life-cycle costs. The potential savings to be derived from the use of more durable materials vis-a-vis the use of less durable materials are shown in Table II. It is clear from the table that life-cycle costs of painting one ft 2 of wall area by more durable paint A (latex) are much smaller than those of painting the same area with less durable paint B (calcimine) . DURABILITY: SOME ECONOMIC EFFECTS This section illustrates some of the linkages that exist between economic principles and the property of durability in building materials, linkages that have not yet been examined in detail in terms of their impact on the Canadian economy. Two economic terms in frequent use are "multiplier" and "accelerator." Multiplier effects occur when an increase (or decrease) in a given level of investment in building results in a series of increases (or decreases) in the levels of income of contractors, suppliers and others associated with the business. Accelerator (or decelerator) effects occur when such increases (decreases) in income levels result in higher (lower) demand for buildings, the construction of which will induce (or discourage) additional investment. These terms usually apply to a short period of a year or two rather than to a longer. The effects of a multiplier or an accelerator in building are also felt in the resources, manufacturing and service sectors of the economy. In the short run, also, building materials manufacturers will tend to prefer the less durable varieties because their prices may be lower and more attractive. On the other hand, in the long run more durable materials will be more attractive to the purchaser. The problem for the manufacturer, therefore, is to adjust the balance of his production 16 between more and less durable materials in order to avoid the possible consequences of declining output, income, and employment levels in the firms manufacturing less durable goods. Such adjustments are also needed in the long run to avoid wasting material resources which, in economic theory, are considered to be relatively scarce and to have alternative uses. For example, the demand for steel as a building material is in competition with the demand for it for machinery, automobiles and other end products. It could also be that a slack in the demand for iron in the building industry will be picked up by increase in demand by the auto industry. At the present time energy used to produce building materials is another area where waste should be avoided. There are a number of ways in which energy, economics, and durability are linked. For example, durability should help reduce the consumption of both energy and materials, as well as construction expense, when materials can be recycled for use in new or rehabilitated buildings. On the other hand, the degree of durability in a material may be linked directly to the consumption of energy in its production, and the price per unit of the material to the price per unit of energy. This may discourage technical innovation in the development of more durable building materials, as may uncertainties in the future supply of suitable forms of energy for materials production. The production of building materials with given sets of physical properties may also have social costs, paid by society as a whole. One example of such costs is the environmental pollution resulting from production processes, another is the damage caused when waste materials degrade over a period of time in disposal sites. Metallic corrosion, and a corresponding lack of durability, during the service life of materials add to the costs of building owners in the long run and may also incur social costs in the form of environmental pollution. The use of more durable goods may, thus, reduce social costs over the long run. CONCLUSION The activities of the building industry have a significant influence on the levels of output, income and employment in the Canadian economy. About one-third of this influence is the result of the production and use of building materials. The degree of durability of various building materials can also have some impact on the economy and, more particularly, on the costs of constructing and maintaining individual buildings. In some circumstan- ces, economic factors may favour the use of more durable materials; in others they may favour less durable ones. In practice, the selection between more or less durable materials depends on the availability of service life, maintenance and other cost -related information. At this point in time this information may not always be available. 17 ACKNOWLEDGEMENT The author wishes to acknowledge the assistance of A.H. Wilson in the preparation of this paper. This paper is a contribution from the Division of Building Research, National Research Council of Canada, and is published with the approval of the Director of the Division. REFERENCES 1 Construction in Canada 1978-1980, Statistics Canada, Cat. No. 64-201 (Ottawa, 1980). 2 Carruthers, J.F.S. The Performance with Time of Components. Proceedings, First International Conference on Durability of Building Materials and Components, (P.J. Sereda and G.G. Litvan, Eds.), Ottawa, 21-23 August 1978. 18 § CJ h- LO CO CO hJ 2 < O \— 1 i— i a! J UJ _J H 1— 1 S 2 J i— i J < *< CO *» cC CO < ►J hJ < hJ 1 — 1 o M a£ a UJ UJ E- r-H CT) < i— 1 H 2 O H i— i ^ U H 3 CO as 2 H O CO u 2 O •s u "* r-^ tu C7> o r-H ■ 2 r-H O \D i— i Oi H <—\ a, § CO 2 O CJ <4H O a> r- 1 aj > \D r-H P X 0j P O 3 ^ & +-> P O O o •H t-H P --I Ph aj trt <4H c o O CJ (/) r— I 03 •H fH CD P Kj CNI ■I" II m C o l+H •H O r-H P aj o CD P 3 3 O P r— I H P aj C C o u o (/) MH ■H H O P aj O •H CD 3 fH 3 P aj > o u S o\° o\° LO o\° CTlLn OOCTlHOgKlrJ-iON tONH^tOOH^O Or- it^t-o^vor^oo HOCTlCM-NfMl^OO H;NHfOOOMOOH NCDl/)COOMH(0 OICT^CTiCTlO^CTiOO e aj P in fi o u e •H X B o c o o UJ G aj •H 13 aj (=! aj CJ cu rG P lp o fH 3 • P O en 3 r-- P cr> P r-H CO P 3 aj Dh 2 P aj 5 P o P i O P 3 QhUJ c CTi r-H O l-O CD 1 X \Q H r-H , O aj S. -d aj • C P aj aj CJ CJ »v o ^t •H r^ P CT> C/) r-H •H 1 P r-H aj \D P cr> CO r-H *s aj u fH 3 o co 19 TABLE II COMPARATIVE LIFE -CYCLE COSTS* OF WALL PAINTING CONSTANT 1978 DOLLARS Initial Cost per ft Present Value of All Repainting per ft 2 Total Life-Cycle Cost per ft 2 Savings Using Paint A per ft 2 Savings for Total Wall Area (3436 ft 2 ) Paint A Paint B (latex) (calcimine) ~r- ** 25* 20* 56. 16* 85.93* 81.16* 105.93* 24, .77* $851 * Based on the following assumptions: The economic life of the building is 40 years; the discount rate is 11 per cent per year; the frequency of repainting the walls with latex is 5 years, with calcimine 3 years; the initial cost of painting the walls with two coats of latex is 25* per ft 2 ; the cost of repainting (with one coat) after 5 years is 30* per ft , thereafter increasing at an annual rate of 3 per cent up to the 35th year, after which no further repainting is done; the additional cost of specially preparing the walls for repainting with latex after 15 years is 20* per ft , and after 30 years is 30* per ft ; the corresponding initial cost for calcimine (two coats) is 20* per ft , the cost of repainting (one coat) after 3 years is 17* per ft , thereafter increasing at an annual rate of 3 per cent up to the 36th year, after which no further repainting is done; the additional cost of specially preparing the walls for repaint- ing with calcimine after 9 years is 20* per ft , 25* after 18 years, 30* after 27 years, and 35* after 36 years; all of the above costs include labour and materials; no costs are incurred for painting during the periods between complete repainting. ** Based on Landsdowne's Construction Cost Handbook, London, Ontario, 1978. 20 00 16 ae < i 14 o o 1—4 1? — r- o* •-* 5 10 ~ < i— i/i 8 — z o — o 6 a=> z o — 1 4 "—_ 8 l___ — -I — 4 METAL FABRICATED NON -METALLIC MINERALS (CEMENT AND OTHER NON -METALLIC) LUMBER SAWMILLS AND OTHER WOOD PRODUCTS r ELECTRICAL AND ' COMMUNICATION PRODUCTS \- PRIMARY METAL PRODUCTS (IRON, STEEL, ALUMINUM ETC) 1 f ■ ._ PLASTIC FABRICATED PRODUCTS 1961 1963 1965 1967 1969 1971 1973 1974 FIGURE 1 CONSUMPTION OF MAJOR CONSTRUCTION MATERIALS, 1961-1974, IN CONSTANT (1971) DOLLARS CANADA FIGURE 2 AGE OF BUILDING MATERIALS AND REPAIR AND REPLACEMENT COSTS FIGURE 3 RELATION BETWEEN DURABILITY AND DETERIORATION OF BUILDING MATERIALS 01 PROBLEMS IN THE IMPLEMENTATION OF DURABILITY ASPECTS TENHO SNECK TECHNICAL RESEARCH CENTRE OF FINLAND SF-02150, ESPOO 15, FINLAND Abstract: The implementation of both theoretical and practical aspects of durability is often impeded by the attitudes of parties concerned. As the importance of durability is growing, there is a need of more information on the fact that the service life of materials is limited and dependent on their conditions of use. The concept of durability needs clarification, an attempt to create and use a common vocabulary and systematics is necessary in order to get a concentrated attack on the problems. Product development and design for durability, aspects related to regulations, standards and the evaluation of the durability performance belong to the systemati zat ion field. The conservation of natural resources calls for immediate actions concerning improved durability but some of the measures taken may result in severe dura- bility problems in the case this phenomenon is forgotten. The import- ance of the resource conservation is of such an order that there is a need to express this by changing the present definition of materials science and engineering. The redefinition might serve as an indicator of the impact the time factor and the limitations in materials re- sources and supply should have on the research and technology related to materials. The economic implications of durability need more consideration. All the time it is important to keep in mind that it may be dangerous to single out a certain factor from its context if this context is then neglected. State-of-the-art reports would be highly welcome in order to give guidance to researchers. Key words: Building design; conservation of resources; durability; economic implications; evaluation of durability; implementation of durability; materials science and engineering; performance over time; servi ce 1 i fe. 1 Background The growing interest in durability is due to many reasons and trends in the world of today. Conservation of natural resources is the most important factor. This may be achieved in several ways: using less materials, making the materials last longer, substituting materials and recycling by using waste materials and byproducts. The implications of maintenance, the rehabilitation of old buildings and all economic aspects of durability become more evident. Another aspect of availa- bility of materials has to be stressed as there are readily available materials but knowledge of their durability or of procedures to improve it are lacking. 22 Construction for extreme conditions is a challenge for the building of today. Marine construction, especially off-shore structures, buildings for arctic or tropical climate, or for other special or aggressive conditions are examples. For instance, energy saving problems may be solved by new design solutions even in such cases when their perform- ance over time is unknown. The construction in developing countries presents vast problems, also related to durability. The roofs, walls and building frame are the critical components of the low-cost houses (l). The improvement of their durability is of importance and it should be accomplished avoiding the use of expensive materials. Under- standing the performance requirements of the different parts of the houses, and the recognition of the quite different durabilities of the same materials in different locations will help in the search of cost- reducing solutions. Step-by-step improvements were recommended by a seminar on building regulations (2). A meeting for RILEM committee chairmen and interested participants of the RILEM General Council meeting took place in Dubendorf in September, 1980 with the aim to exchange information, identify needs for co- ordination and for new activities (3)- The desirability of using a common vocabulary and methodology was stressed. This paper has been written as a consequence of the concerns expressed during this meeting. It displays a personal view on the matters. 2 The concept of durability There are trends towards a growing international cooperation but there are at the same time some matters which seem to cause difficulties. One concerns the expression "durability" itself. This was quite evident at the RILEM meeting. Durability is presently expressed and measured both as the resistance against the effects of certain agents or as the life time in years. There is also a discussion going on whether durability is a property of a material, a building element, or both. If one takes a look at the different existing definitions of durability, all opinions are as well justified. As it seems to be unrewarding to try to attack the problem starting with definitions (3), the lexicographer approach used by Archer (k) to analyze design research might give something. There is a possibility to attempt to discover and record the meaning of words and phrases on the bases of the ways in which they are used and meant by the community concerned. As the opinions differ, a choice has to be made. It is mostly expected that a newly constructed building will last long, a rather important service life is required. The introduction of the time element in the study of the behaviour of buildings and their parts has led to the expression "performance over time" which according to Carruthers (5) means the ability of a building product to maintain its performance to an acceptable extent. The term "durability" is used quite synonymously and both are measured in years. For others, 23 "durability" is a materials property, clearly connected to deterioration and mostly expressed as the resistance to an agent or a combination of agents. A "durability" of this kind is one aspect of the "performance over time." The dualistic "durability" approach is displayed by the excellent IC-IB performance guide (6) which makes use of both ways to express i t. In the study of building elements or other structures it is important to investigate the durability of materials in structures. The durability of a material in a structure depends upon the other materials present and also upon the design. Joints and interfaces may be critical for the durability. The deterioration of a material does not necessarily destroy it. The destruction of an air proofing course may affect the tightness and energy saving performances without any visible change. The table below shows the ways the author thinks the two durability systems are applied to buildings, building elements and materials. The term may be expressed in Y=years or as R=resi stance to agents. Building Element Materia House Wall Brick Durability Y Y R Performance y - Y (R) ove r t i me Summarizing, the author prefers the term "performance over time" as a general expression of all changes caused by environmental, use and accidental factors. "Durability" is applied to materials as such and to materials incorporated in structures. It forms a part of "perform- ance over time." As the table shows does the use of "durability" as an over all expression not impede with the work to be done. However, the different expressions result in frustrations in people who do not know about the different ways to express things. At present it might be to some aid to tell which system is used. 3 Need of systematics The solving of problems connected with some important factor, as energy today, may result in durability problems and a concentration on dura- bility may, vice versa, lead to some other difficulties. Rzevski (7) has given an interesting picture on design methodologies. Starting with the ideas of Popper that the results of a problem solving activity inevitably create a new problem, he goes on by defining as a criteria for a good design methodology that it does not create more problems than it solves. It is important to try to keep the "whole" in mind. Ostlund (8) feels that so much knowledge on durability now exists that 24 it would be possible to study combined durability and safety problems with probabilistic methods. Fagerlund (9) argues that functional re- quirements can often be expressed indirectly as a minimum or maximum value of a single property or of a complex of fundamental properties. Destructive processes can affect the object in such a way that it does not any more fulfill the requirements. This way of looking at the dif- ferent requirements and durability in combination is a very important development. In order to make knowledgeable people from different disciplines able to participate in the solving of durability problems dealing with the use and maintenance, product development and design, it would be neces- sary to have some kind of common systematics. It would be preferable to accept some documents which have been accepted by a consensus approach. Durability is included in the documents of ISO/TC 59/SC 3 Functional/ user requirements and performance of building construction. DP 6241 Guidance for the preparation of performance standards in building consists of four tables of interest: Basic user requirements, General building types and functions, Sub-systems of the building fabric and a Master List of the agents which may influence the performance of the building. The ASTM Standard E632-78 on Recommended practice for deve- loping short-term accelerated tests for prediction of the service life of building components and materials is very important for durability studies. The new CIB Master List for structuring documents relating to buildings, building elements, components, materials and services is in preparation. This is an important list of properties. The use of the ISO, ASTM and CIB documents together seems to be a useful way to get a common basis for systematics - with the exception of durability de f i n i t i on s . The opinions of Cohen (10) are of a very great interest. If we study the interrelation between materials structure/property/performance, and try to predict the properties from structure, and the performance from properties, it is more likely that we fail than succeed. The great success of the materials structure/property/performance concept arises according to Cohen from the countercurrent flow of knowledge between principles and experience, between theory and empirism, and between scientific quest and human need. This process only operates if the different disciplines involved cooperate. Everything should be done to get this complicated procedure working. It seems to be difficult to grasp how heavily the durability of a mate- rial really depends upon the environmental agents. The environment has to be known. Very little can be accomplished if the way the material degrades is not known. The need to develop methods for the measurement of processes and properties is quite central and should be encouraged. It is claimed that durability testing is costly and the results worth- less. This is not necessarily always true but improvements are needed. Methodologies are needed starting with consideration of the require- ments put on the object in use, and ending with the judgment of the resul ts. 25 k Economy and resources Webb (3) gave a comprehensive picture of the economic implications of durability. He analyzed the direct and indirect costs of durability, the negative aspects of excessive durability and concluded that we need a more integrated approach at the design stage. We have to aim at a building where all materials have about the same life or a house with very durable basic fabric and relatively easily replaceable parts with a shorter life. The questions which Webb asked were highly relevant. They concerned the role the different types of deterioration are play- ing in planning work on national and international levels, the need to develop methods of quantifying the economic, technical and social costs of poor durability, and the need to disseminate knowledge re- garding deterioration and its consequences. The treatment of the dura- bility problem in this way needs careful attention as enough policy forming information is not given. Durability is strongly related to saving of natural resources. From this poi nt of view the opinions of Huggins (11) on the impact of mate- rials availability are highly relevant. He claims that the present definition of materials science and engineering focuses on properties. Huggins feels that a strong shift to resource-related activities is ne- cessary. It is tempting to take up this idea by Huggins and redefine the present "paradigm" which has the following shape: "Materials science and engineering are concerned with the generation and application of knowledge relating the composition, structure, and processing of materials to their properties and uses" by adding "taking into account their availability." This way of thinking should have a heavy impact on the applications. It would strongly emphasize the importance of durability, influence the scientific community and education and promote the implementation of durab i 1 i ty. 5 References 1 UNIDO, "Appropriate industrial technology for construction and building materials," Monographs on Appropriate Industrial Technology No 12, United States, New York, 1980, p. 8 2 UNCHS and SCBR, "UN Seminar of experts on building codes and regulations in developing countries," Swedish Council for Building Research, Stockholm, 1 980 3 Sneck, T. , "RILEM and durability," Materials and Structures , in print k Archer, B., "View of the nature of the design research, " De s i gn : Science: Method, Westbury House, Guildford, 1981, pp. 31~^7 26 5 Carruthers, J., "The performance with time of components," Du rab jjjtyof Bui 1 d I ng _ Mate r i a 1 s and Components , ASTM STP 691, American Society for Testing and Materials, Philadelphia, 1980, pp. 98-105 6 IC-IB, "Guide des performances du batiment," Syndicat d'etudes IC-IB, Brussels, 1979 7 Rzevski, G., "On the design of a design methodology," ref. 4, pp. 6-17 8 Ostlund, L. , "Durability and safety," Studies on Concrete Technolog y, Swedish Cement and Concrete Research Institute, Stockholm, 1978, pp. 277-286 9 Fagerlund, G. , "Service life of structures," RILEM Symposium on Quality Control of Concrete Structures , Swedish Cement and Concrete Research Institute, Stockholm, 1979, pp. 199-215 10 Cohen, M. , "Unknowables in the essence of materials science and engineering," Materials Scienc e _ and Engineering , Vol. 25i 1976, pp. 3~4 11 Huggins, R. A., "Basic research in materials," Science, Vol. 191, No. 4228, 1976, pp. 647-191 27 ENVIRONMENT, MICROENVIRONMENT, AND DURABILIITY OF BUILDING MATERIALS H. E. Ashton and P. J. Sereda Division of Building Research National Research Council of Canada Ottawa, Ontario K1A 0R6, Canada Abstract: This paper tries to show that microenvironment defines the precise conditions of solid materials and the immediate layers of liquid or gas prevailing at the site where chemical or physical processes of deterioration are taking place. It argues that the microenvironment depends on many factors, including weather, design, and material charac- teristics, and that it is not likely that this system will yield easily to analysis and prediction. Thus, it emphasizes the importance of mea- suring the conditions of the microenvironment as a guide to the direc- tion of future studies and investigations. It argues also that the use of individual weather elements may be limited because the sequence in which events occur, including simultaneous occurrences of various elements, is important. This paper suggests that the following param- eters of the microenvironment should be measured or monitored: surface temperature; surface moisture as time-of-wetness; occurrence of near- saturated state of porous materials during freezing conditions; total deposition of pollutants on exposed surfaces; and UV radiation dose. Keywords: Building materials; durability; environment; microenvironment; moisture; pollution; solar radiation; sulfur dioxide; sensors; time-of-wetness; ultraviolet radiation. Microenvironment defines the boundary conditions at the surface and even inside the pores of a material and is determined by the interaction of the material with its microclimate. Materials deteriorate due to a number of physical, chemical, and biologi- cal processes fostered by certain combinations of microenvironment factors . This paper is concerned with the need to monitor microenvironmental factors of temperature, moisture, solar radiation, and pollution as they are related to the processes of deterioration of materials in service. The temperature of a given material or component of a structure is a continuously variable factor determined by the diurnal atmospheric temper- ature change, solar radiation, thermal characteristics of the material, wind speed, and the heat gain or loss from the ground and the occupied space . 28 Temperature can affect the performance of materials in at least two ways, causing movements of building elements or components that result in stress and deformation and causing internal stresses in composite materials when components of the system have different coefficients of thermal expansion. Review of literature, including performance of sealants and concrete, suggests that it is necessary to measure the temperature of a material in its microenvironment in order to determine the effect on performance and durability. Most processes of deterioration and corrosion involve moisture - as a direct agent of the process, as a medium of the reaction, or as a con- stituent of the microenvironment fostering the process. Clearly, moisture is a most important factor of microenvironment. In certain processes, such as the corrosion of metals, only trace amounts as invisible films on the surface, corresponding to equilibrium conditions in the surrounding atmos- phere of about 85% RH, are enough to enable the process to continue at a high rate. Other processes, such as destruction of a porous material by freezing and thawing, require moisture content conditions near satura- tion, temperature below freezing (from 0°C to as low as -13°C) and a rate of cooling above a certain value. To understand, monitor and control the various processes of deterioration it is vital to measure the moisture conditions of the environment, and particularly of the microenvironment in which the materials must serve. The atmosphere is the primary source and receptor of moisture for the materials of a building, but soil also can serve as a primary source for materials located below ground level. A method employing miniature sensors was developed for the in situ measure- ment of time-of -wetness on surfaces. Because the sensor is very small, of the size of a postage stamp and bonded to the surface, its temperature is similar to that of the surface so that when it detects moisture it means that the surface is wet. Preliminary work indicates that plastic specimens exposed horizontally experience a time-of-wetness 1 .7 times greater than do similar specimens exposed vertically, facing south. Porous building materials such as concrete, masonry, and wood require significant quantities of water to increase their moisture content to a level where processes such as frost action and rot can take place. A principal source of such quantities of water is rain or melting snow. An indication of the potential for wetting of exposed materials can be obtained from the product of the normal annual rainfall and the normal annual mean wind speed, called the driving-rain index. There is evidence to show that high levels of moisture occur during late fall and winter and that near saturation can coincide with rapid change of temperature to below and above freezing, causing a destructive freezing-thawing cycle. At present there are no available data to indi- cate the actual frequency of occurrence of such freezing-thawing cycles in materials under normal service conditions. All evidence points to 29 the need to measure the moisture content of porous materials in service on buildings (just as the need exists to measure the surface moisture of non-porous materials). Ultraviolet radiation (UV) is the dominant climatic factor in the weathering of plastics and paints. Generally, only the total solar radiation intensity is measured at various meteorological stations throughout the world. Simple methods of monitoring the solar UV radia- tion dose employ certain UV-sensitive materials such as poly(phenylene oxide) film (PPO) and poly(vinyl chloride) film. More recently, a joint program undertaken by NRC and NBS involved the evaluation of two light- sensitive materials: NBS Standard Reference Material 702 (transparent yellow plastic chip), and ICI red chip. Results of this investigation are not yet published. There seems to be no information on the distribution of UV irradiance on surfaces of walls of buildings, information that is required to relate the conditions of the microenvironment to those at exposure sites and meteorological stations. Failures of paints and plastics occur more frequently on southern than on northern exposures in the northern hemis- phere, clearly illustrating the influence of photo-oxidation. Because other processes are also involved, it is not possible to account for the failures until the actual dose of UV can be measured at different expo- sures . Although pollution in the air, in the soil, and in water affects durability of building materials, this paper will confine itself pri- marily to air pollution. Only the effects of sulphur oxides and particulates such as soot and sea salts, however, have been studied throughout the world in relation to durability of materials. Monitoring of the concentration of pollutants in the atmosphere and in rain is done by various government agencies in many countries, but little effort is directed to the study of their effect on building materials, particularly that of acidic components in rainwater . The effect of SO2 on atmospheric corrosion of metals has been established reasonably well . The most useful method of measuring SO2 with respect to its effect on corrosion and deterioration should be that involving the rate of deposition on lead peroxide. This method is cheap and simple, providing a cumulative measure of SO2 over a given period, and should reflect the effect of air flow (wind) and relative humidity. It is clear that present data on pollution by air of SO2 do not indicate adequately the microenvironmental conditions that prevail for materials in service on buildings subjected to sources of local pollution. There is need to measure the distribution of concentrations of SO2 with regard to height and orientation of buildings. There is also a need to study the concentrations of acids on surfaces subjected to different climatic conditions . 30 Sodium chloride is considered to be the aggressive constituent of sea salts, contributing to very high rates of metal corrosion in the zone near the ocean where sea spray is experienced. Corrosion of ordinary steel by sea water is another example of the effect of microenvironraent . The rate of corrosion of a steel piling in the splash zone increases by a factor of 3 over the rate for the section in the sea water. SUMMARY This paper tries to show that microenvironment defines the precise conditions of solid materials and the immediate layers of liquid or gas prevailing at the site where chemical or physical processes of deteriora- tion are taking place. It argues that the microenvironment depends on many factors, including weather, design, and material characteristics, and that it is not likely that this system will yield easily to analysis and prediction. Thus, it emphasizes the importance of measuring the conditions of the microenviron- ment as a guide to the direction of future studies and investigations. It argues also that the use of individual weather elements may be limited because the sequence in which events occur, including simultaneous occur- rences of various elements, is important. This paper suggests that the following parameters of the microenvironment should be measured or monitored: surface temperature; surface moisture as time-of-wetness; occurrence of near-saturated state of porous mate- rials during freezing conditions; total deposition of pollutants on exposed surfaces; and UV radiation dose. 31 QUANTITATIVE CONSIDERATIONS OF MOISTURE AS A CLIMATIC FACTOR IN WEATHERING K. G. Martin Division of Building Research Commonwealth Scientific and Industrial Research Organization Melbourne, Victoria, Australia R. E. Price Department of Physics Western Australia Institute of Technology Perth, Western Australia, Australia Abstract: Following the concept of the solar weathering index which is a climatic parameter for physical damage functions that quantify photo-oxidation of organic building materials, functions that relate to moisture associated degradation of building materials have been considered. These have involved studies where some quantitative information has been given on both exposure and effect, such as atmospheric corrosion of metals, chalking of paints, photo-oxidation of polyester resins and strength loss in timber composites. The pertinent moisture parameters have been discussed and where possible evaluated for Australian sites from meteorological data. Several new climatic parameters have been suggested and further work on validating these as factors in physical damage functions suggested. Key words: Physical damage functions; atmospheric corrosion of metals; chalking of paint; photo-oxidation of polyester resins; loss in strength of timber composites; time of wetness, coincident air temperature; moisture content of air, water vapour pressure percentiles, Useful predictive evaluations of the weathering of building materials and components require the development of physical damage functions. These functions may be defined as quantitative relationships between the decay of a critical property of a material or product and the climatic factor responsible for the decay. Development of these functions requires the climatic factors to be described in quantitative terms or units that relate to the degradation process. Since weathering processes are complex simplifying decisions are needed to define the operative degradation mechanism and to quantify the extent of exposure as a climatic factor. Because climatic factors are variable and because materials often deteriorate in a non-linear manner, clock time itself is not a valid basis for damage functions. 32 In a previous paper (1) the degradation of polymers in plastics and paints when exposed to sunlight was assumed to be primarily a photo- oxidation mechanism. Laboratory and field studies (2,3,4) showed that the rate of photo-oxidation of a polymer could be expressed as: R = 32fl n (1) where 3 = proportion of ultraviolet radiation in global radiation f = frequency of occurrence of global irradiance of level I, and n = an empirically-determined index characteristic of a particular polymer. The function was termed the solar weathering index for the site-polymer combination with values of n ranging from 1 to 2 and data reported to describe sites about Australia. The above concept has a theoretical basis where the primary process of generation of free radicals by absorption of solar energy is the rate determining step in degradation. Where secondary reactions such as those related to moisture become rate determining the physical damage function becomes much more complex. This paper considers degradation mechanisms involving moisture and the processing of climatic data to relate to these mechanisms. MOISTURE DEGRADATION PROCESSES (a) Atmospheric corrosion of metals Summaries of recent developments relating climatic conditions and extent of corrosion have been given by both Boyd (5) and Haynie (6) largely based on the work of Guttman and Sereda (7) and Sereda (8) . A physical damage function was shown by Haynie to represent the observations from five field studies and one laboratory study by suitable choice of two regression coefficients. The form of the function was: C = AT + BT (SOJ (2) z w w 2 where C was the extent of corrosion, w T the time of wetness, expressed as a percentage of total time, (SO J the mean concentration of SO , and A and B the regression coefficients. T was measured at specific sites with a bimetallic cell but can be estimated from frequency distributions of RH based upon the assumption that moisture deposits on a corroding plate when the RH rises above some critical level. For atmospheric corrosion of zinc plate the critical RH was empirically established as 86.5% (7) and although some doubt was expressed as to whether this critical value applied to all metals (8) it was proposed that for general corrosivity studies T w 33 be taken as the time the RH exceeds 85%. (b) Chalking of paint Much less information is available for chalking of paint because most studies have been done under laboratory cyclic conditions where it has not been possible to separate the influence of climatic variables. Hoffman (9), however, determined chalking rates against time under constant conditions in an exposure chamber. When ultraviolet irradiance exceeded a relatively low level, RH was found to be the most important climatic factor. Exposure trials in the tropics (Lae, Papua New Guinea) and in an urban Mediterranean type climate (Melbourne) gave some indication that time of wetness based upon a critical RH of 60% could be used as the significant climatic variable. However, a physical damage function was not established. (c) Yellowing of polyester resins Limited studies in an exposure chamber under controlled conditions (10) have indicated that the rate of yellowing increases steadily with the amount of moisture in the atmosphere. The data also indicated an effect of temperature separate to that of changing the amount of moisture in the atmosphere. Thus for polyester resins the climatic factors that discriminate localities are solar weathering index, surface temperature and water vapour pressure, but a general physical damage function has not been established. (d) Strength decay in timber composites The loss in strength of timber composites such as particleboard has been discussed by several authors recently in relation to the role of moisture (11-14). Two mechanisms appear to be important; firstly, the hydrolysis of the adhesive and secondly, internal bond failure caused by stresses at the glue line as the timber swells and shrinks with changes in moisture content. Field information on plywood has been broadly related to laboratory tests in British Standards by Knight (11) who indicates that moisture and temperature are the climatic factors determining extent of degradation. In more recent work Millett et at (12) found that the temperature dependence of the loss of strength fitted an Arrhenius-type equation for both wet and dry conditions, and were thus able to predict life times wet and dry at 20 C by extra- polation. These dependencies of extent of degradation upon temperature suggest that climatic factors to characterize different sites should be in terms of mean annual air temperatures under dry and wet conditions. Suitable definitions would need to be made of dry and wet. Some attention has also been given to laboratory cyclic tests. Okuma et at (13) developed a degradation coefficient based on a specimen's residual stress after a specified number of wet and dry cycles. Sandoe et a'L (14) cycled specimens between 85 and 35% RH and showed an 34 approximately linear relation between decrease in flexure induced tensile stress and number of cycles. Wet dry cycling was also previously found to relate to both dimensional change and loss in strength in exposed bituminous roofing felts (15,16) although quantit- ative functions were not developed. The suggestion is that the critical parameter is the number of cycles through a significant change in RH rather than the magnitude of the change itself. Evaluation of such a parameter again requires defining wet and dry. One interesting approach reported by Palmer and Stashevski (17) is to cycle between the annual limits of equilibrium moisture content reached by the timber species concerned. This approach simulates a one cycle a year sheltered field situation and characterizes sites by the magnitude of the change. For non-sheltered situations shorter term wetting and drying cycles would be important but there is insufficient information to define a cycle and further research on this matter is warranted. PROCESSING OF AUSTRALIAN CLIMATIC DATA (a) Time of wetness Hourly readings of % RH for eight sites have been processed in terms of percentage of time that the % RH exceeded certain levels to give cumulative frequency plots for each site in the same way as reported for Canada (8). The results indicate that Port Moresby is not signif- icantly different from the representative plot for Canadian coastal centres while Darwin to Melbourne is similar to the Canadian inland. For similar levels of pollutant, corrosion would thus be similar in Canada to that in the tropics and similar in Darwin, Sydney and Melbourne. Specific corrosion data (18), however, indicate that Sydney is about twice as corrosive as Melbourne. Examination of general climatic data relating to moisture for the two cities shows that a major difference is in the contribution to wet surfaces by frosts. This suggests that a coincident temperature/% RH factor may be needed to characterize the sites. Data from the Australian Bureau of Meteorology were therefore analysed to average the air temperatures for periods when the RH exceeded 85%. It was then possible to discriminate between the sites and explain the Sydney Versus Melbourne corrosion data. (b) Ambient Moisture Content The amount of moisture in the atmosphere is widely available from meteorological observations of air temperature and % RH and reference to psychometric charts. For long term weathering studies mean annual data would characterize a site and provide a basis for establishing physical damage functions by combining with the solar weathering index for cases where photo-oxidation followed by secondary hydrolysis is the 35 predominant degradation mechanism. (c) Wet- dry cycles (i) Annual cycle extremes The simplest way to characterize a sheltered site in relation to wet- dry cycling is to consider only the annual seasonal cycle and charact- erize sites by the extremities of the conditions reached. Examination of meteorological data suggests that the 1 percent percentiles of temperature for each range of RH experienced be taken as the extremes. These values for Melbourne together with the water vapour pressure calculated from them indicate that the low temperature extremes correspond to water vapour pressures from 6.5 to 7.5 kPa with a mean of 7.1 kPa while the high temperature extremes correspond to 13.9 to 22.0 kPa with a mean of 19.8 kPa. Water vapour pressures for extreme temperatures are thus relatively constant and can be used to character- ize Melbourne in terms of the magnitude of the annual wet dry cycle. (ii) Number of cycles between extremes Sheltered sites may also be characterized by the number of cycles per annum between arbitrary extremes. For example, based on the annual Melbourne data this could be the number of times the water vapour pressure changed by 10 kPa and back again, i.e. from 8 to 18 to 8 kPa. This information requires processing of data in time sequence that has not been possible to date. (iii) Number of cycles for non-sheltered situations Where wetting by rain would be a significant factor, there is little useful data available. The number of rainy days is generally recorded at meteorological sites but is considered a poor proxy for the number of wet dry cycles experienced by an exposed surface. Further research on this matter is needed. CONCLUSIONS There are a number of different ways that moisture may enter degradation reactions with building materials but only a few of these processes have been quantified in terms of physical damage functions which would indicate how climatic data need to be processed to relate to the degrad- ation and to thus quantitatively characterize the severity of different sites. For atmospheric corrosion time of wetness has been accepted as the climatic parameter of prime importance. Chalking of paint appears to have similar dependencies upon moisture as does atmospheric corrosion but the degradation process is less 36 completely understood. For photochemical reactions where moisture may participate, limited data indicate that the degradation rate is related to the ambient moisture content. This parameter is readily evaluated from meteorolog- ical observations and mean ambient moisture contents may be tabulated to characterize different sites. Loss of strength of hygroscopic materials such as timber composites is known to be related to moisture cycling although quantitative functions have not been developed. One approach to characterizing sites in relation to sheltered exposures is to consider only an annual cycle. It is suggested that the 1 percent percentile values of frequency of occurrence of coincident temperatures for the annual range of relative humidities experienced provides a reasonable basis for the parameter. For the external exposure case where rain and drying by sunshine are important there is no data available upon which a site could be characterized and research on this matter is needed. Such climatic data parameters are reasonable suggestions for the devel- opment of physical damage functions but require to be validated in field studies before use as predictive indicators. REFERENCES 1. Martin, K. G., "Appraisal of durability of organic materials and components - The Australian view," American Society for Testing and Materials, Special Technical Publication 691, 1980, p. 106. 2. Martin, K. G. and Tilley, R. I., J. Applied Chemistry Vol. 19, 1969, p. 235. 3. Martin, K. G. and Tilley, R. I., British Polymer J. Vol. 1, 1969, p. 213. 4. Martin, K. G., "Solar weathering indices for Australian sites CSIRO Division of Building Research, Technical Paper No. 18, 1977. 5. Boyd, D. W. , "Weather and the deterioration of building materials," American Society for Testing and Materials, Special Technical Publication 691, 1980, p. 145. 6. Haynie, F. H. , "Theoretical air pollution and climate effects on materials confirmed by zinc corrosion data," ibid p. 157. 7. Guttman, H. and Sereda, P. J., "Measurement of atmospheric factors affecting the corrosion of metals," American Society for Testing and Materials, Special Technical Publication 435, 1968, p. 326. 37 8. Sereda, P. J., "Weather factors affecting corrosion of metals," American Society for Testing and Materials, Special Technical Publication 558, 1974, p. 7. 9. Hoffman, E., "Chalking of paints," Journal of the Oil and Colour Chemists Association , Vol. 54, 1971, p. 450. 10. Martin, K. G., "Thermosetting polyester resins, influence of environmental conditions upon photodegradation of unreinforced resins", CSIRO Division of Building Research, Report No. 28, 1974. 11. Knight, R. A. G. , "The efficiency of adhesives for wood," UK Ministry of Technology Forest Products Research Bulletin N-. 38, HMSO, 1968. 12. Millett, M. A., Gillespie, R. H. , and Baker, A. J., "Precision of rate-process method for predicting durability of adhesive bonds," American Society for Testing and Materials, Special Technical Publication 691, 1980, p. 913. 13. Okuma, M. , Lin, T. H. and Omiki, S., "Durability of structural particleboard evaluated by repetitive loading tests," ibid, p. 935. 14. Sando, M. D. , Keenan, F. J., Beall, F. C. and Fox, S. P., "Effect of accelerated aging on tensile perpendicular-to-glueline strength of glued laminated beams," ibid p. 924. 15. Martin, K. G., "Changes in bituminous roofing felts with changes in moisture content," CSIRO Division of Building Research, Technical Paper 18, 1959. 16. Martin, K. G., "Deterioration of bituminous roofing fabrics," CSIRO Division of Building Research, Technical Paper 11, 1961. 17. Palmer, R. E and Stashevski, A. M. , "The effect of humidity cycling on commercial particleboards," CSIRO Division of Building Research, Report 1979. 18. Martin, K. G and King, G. A., "Corrosivity measurements at some Australian cities," J. Australasian Corrosion Association (accepted for publication). 38 ESTIMATION OF TOTAL SURFACE AREA AND SPATIAL DISTRIBUTION OF EXPOSED BUILDING MATERIALS FROM COMMONLY AVAILABLE INFORMATION FOR U.S. METROPOLITAN AREAS Michael D. Koontz James E. McFadden GEOMET Technologies, Inc. Rockville, Maryland 20850 Fred H. Haynie Environmental Sciences Research Laboratory U.S. Environmental Protection Agency Research Triangle Park, North Carolina 27711 Abstract: One difficulty in translating physical damage functions into economic costs for assessing pollution control stragegies is in esti- mating how much susceptible material is exposed to a particular level of pollution. The authors sought to estimate a model relating exposed material quantities and densities to proxy variables measurable at the census tract level for two U.S. cities -- Baltimore and St. Louis. Atlases developed by the Sanborn Map Company were the primary source of information to determine quantities and types of surface materials in randomly selected census tracts for each city. Visual surveys were utilized to verify and update the estimated proportions of different material types. Potential predictor variables were chosen from census data, land use data, and establishment data available from Federal Agencies and from planning offices in each city. It was found that a sizable portion of intracity variations in material quantities and den- sities could be explained by available proxy variables. However, re- sults generally were inconsistent between the two cities. With the possible exception of residential structures, tract-by-tract predictive models are not expected to be universally applicable. Per capita esti- mates of total and specific material quantities are provided for each city. Key words: Building materials; Sanborn atlases; land uses; regression models; census tracts; pollution effects. INTRODUCTION The purpose of this study was to develop and validate a stochastic model to predict the quantities and distributions of exposed building mat- erials across subdivisions of major metropolitan centers. Such a model should provide estimates of total surface area and geographic distri- bution for community- used building materials exposed to varying ambient concentrations of atmospheric pollutants. To accomplish this objective, data on the distribution of building material types and quantities, as well as potential predictor variables, were collected for the cities of Baltimore, Maryland and St. Louis, Missouri. 39 MATERIAL DISTRIBUTION SURVEY The material distribution survey was applied to city subdivisions de- lineated by the U.S. Census Bureau and termed census tracts. Two types of surveys were used to obtain the necessary information: (1) a map survey for characterizing building size and material composition, and (2) a visual survey for validating and correcting information from the map survey. The map survey used atlases developed for St. Louis during 1955-56 and for Baltimore during 1971-72 by the Sanborn Map Company. These atlases provide scaled drawings that outline all buildings, struc- tures, and roadways, as illustrated in Figure 1. Information provided for individual structures includes building height in feet above ground, structure use information (such as residence, warehouse, office), and color-coded information on building materials and composition (Table 1). These maps were the primary source of information on the quantities and distribution of major building materials. To validate and correct data obtained from the Sanborn Atlas survey, a visual survey of building ma- terials was performed. This survey, which yielded percentage material composition estimates, was most useful in making allowances for glass and window areas not identified in the Sanborn Atlases and in identi- fying areas where significant redevelopment had occurred. A stratified random sampling scheme was developed to select a represent- ative set of census tracts for the survey. This sampling scheme was based on the assumption that building materials would differ greatly in types and quantities used for residential, commercial, and indus- trial purposes. The universe of census tracts in each city was strat- ified according to the percentage of existing land use acreage in each of three primary zoning classifications: residential, industrial, and commercial. Tracts that were not classified as one of the three pri- mary land use categories were included in a fourth stratum identified as mixed usage. Since it was assumed that the largest variation in the quantities and types of materials would occur in the industrial and commercial sectors and since a relatively small number of census tracts were assigned to these strata, a 100 percent sample was taken in each. For the residential and mixed land use strata, 25 percent samples were taken. A simple random sampling procedure was employed to select a subset of census blocks within each census tract chosen for the map survey. All buildings and structures were surveyed in each census block selected for the sample. The tracts used in the visual survey constituted a random subsample of those selected for the Sanborn survey sample. From each of the four land use strata, a minimum of 25 percent of the Sanborn survey tracts were selected for the visual survey, as shown in Table 2. For the tracts in the visual survey, the same blocks as those randomly selected for the Sanborn survey were chosen. CALCULATION OF MATERIAL QUANTITIES The material data from the surveys were coded and processed by computer to obtain exposed surface area quantities. The formulas for calculat- 40 ing wall and roof areas were based on the physical configuration of the individual structures surveyed, as shown in Figure 2. Because no in- formation was available regarding roof pitch, all building roofs were treated as horizontal for purposes of roof area calculations. For sampled census blocks, material types and quantities were coded on a structure-by-structure basis. Exposed surface areas were aggregated across structures on a block-by-block basis. Residential ,commerical and industrial components, segregated on the basis of structure use codes, were maintained throughout the aggregation process. The ratio of the total number of census blocks within each selected tract to the number of blocks actually surveyed was used as a weighting factor in estimating tract material quantities from the block sample. For each census tract included in the survey, the exposed area was summarized separately for roof and wall surfaces for each category of building material types. Additionally, each material category and wall/ roof totals were summarized by the three building use classifications. SELECTION OF PREDICTOR VARIABLES The predictor variables for the modeling effort were selected on the basis of their consistent availability across major metropolitan areas in the United States and, more particularly, across the two cities selected for this study. Parameters were further restricted to those that could be calculated at the census tract level. Population and housing characteristics were obtained from U.S. Census Bureau pub- lications. Land use and establishment characteristics were obtained from planning agencies in each city. For St. Louis, establishment data were readily available only for firms with 100 or more employees. Only parameters considered to have logical relationships to material quantities were utilized in the modeling effort. Residential zoning acreage and population/housing characteristics were expected to in- fluence material quantities for residential structures. Commercial and industrial zoning acreage and establishment characteristics were expected to influence material quantities for corresponding structures. Commercial and industrial establishments were segregated on the basis of Standard Industrial Classification (SIC) codes developed by the U.S. Office of Management and Budget. Selected parameters were com- bined to generate additional predictor variables, such as population, housing and establishment density and the proportion of total tract acreage designated for residential, commercial, and industrial use. MODEL SPECIFICATION AND ESTIMATION Two types of models were specified and estimated: a material quantity (total square footage) model, and a material density (total square footage per acre) model. In keeping with the segregations used in the material estimation process, separate models were specified for the six combinations of wall versus roof areas, and residential versus commercial versus industrial structures. 41 Specific variables used as candidate predictors for the material quantity models are listed in Table 3. The land use acreage corre- sponding to each structure type (residential, commercial, industrial) was used as the land use predictor for each material quantity model. Housing/population predictors were housing units per residential acre and three other variables expected to influence the nature and size of housing units. Establishment characteristics consisted of size and density of commercial and industrial establishments. The specification of the material density models was simpler in terms of the number of candidate predictors. The proportion of tract acreage in the approp- riate land use category was used along with population per acre and total establishments per acre. Attempts to predict quantities of specific types of materials yielded results which generally were inconsistent between cities and difficult to interpret. These results are not considered reliable and are not presented as part of this paper. The optimal estimating equations were determined by applying the least- squares regression technique known as stepwise inclusion of candidate predictor variables. Only variables whose regression coefficients were different from zero at the five percent level of significance were retained for the final equations. MODELING RESULTS The next two tables list the means across all sampled census tracts for the variables used in the material quantity and density models for each city. Table 4 shows that residential and commercial wall and roof areas are approximately twice as large, on the average, for St. Louis as for Baltimore tracts, with industrial wall and roof areas much closer in quantity for the two cities. In both cities, residential structures account for the greatest quantity of exposed surface area. Residential and industrial acreage are both substantially higher in St. Louis than in Baltimore, with the reverse true for commercial acreage. Housing density per residential acre and the proportion of units with one person per room or less are higher in Baltimore, whereas St. Louis has a larger proportion of older homes and housing units in multiunit structures. Commercial establishments in Baltimore are greater in density and smaller in size than industrial establishments. Corre- sponding information for St. Louis, based only on large establishments, is not as meaningful . The figures in Table 5 reflect a higher material density for Baltimore than for St. Louis in all six categories. Population density is nearly three times as high in Baltimore as in St. Louis. These differences in density may be explained partially by the fact that St. Louis figures are 20 years earlier than those for Baltimore. In Table 6 the results of estimating residential material quantity models for each city are shown. The values adjacent to each variable 42 name are regression coefficients and t-values (in parentheses) which represent the ratio of each coefficient to its standard error. The relative size of each t-value reflects the relative explanatory power of the corresponding variable. All t-values exceed 2.0, the critical value for assessing whether coefficients are different from zero at the 5 percent significance level. The constant term marks the height at whicl^the regression surface intercepts the dependent variable axis. The R value reflects the proportion of variance in the dependent variable explained by the listed variables. For residential wall area, the most influential predictor in both cities was residential acreage. The proportion of older housing units was also a significant predictor for both cities. The residential roof area results were quite similar, and the proportion of variance explained exceeded 50 percent in all four cases. As Table 7 illustrates, the commercial wall and roof area results differ between the two cities. The only common predictor for commercial wall area was housing density. Commercial acreage was a dominant predictor for St. Louis but not for Baltimore. Although the roof area results were again similar to the wall area results, there were no common predictors between the two cities. The proportion of variance explained exceeded 50 percent in three of the four cases. The industrial wall and roof area estimates were even more disparate between cities, with no common predictors selected (Table 8). Indus- trial acreage was the chief predictor for the Baltimore models, whereas housing density was the only consistent predictor for the St. Louis models. The proportion of variance explained again exceeded 50 per- cent in three of the four "cases. The results of the material density models were generally more consis- tent between the two cities, especially those for exposed wall areas. As Table 9 shows, population per acre was the most important predictor in both cities for residential walls, establishments per acre was the leading predictor in both cities for commerical walls, and proportion of total acreage designated industrial was the best predictor in both cities for industrial walls, followed closely by total establishments per acre. The proportion of variance explained for these models ranged from 34 to 93 percent. The material density results for exposed roof areas (Table ? 10) were not as consistent between the two cities, al- though the R values are about the same as those for exposed wall areas. DISCUSSION The previous results indicate that it is possible to explain a sizable portion of intracity variations in material quantities and densities by using available proxy variables. However, it is also apparent that the predictors of choice as well as the relationships between predictors and material quantities or densities are generally unique to each city. As noted previously, the material density model results were generally more consistent between cities than the material quantity model results. 43 Patterns of land use are controlled by local governments and can be expected to vary considerably among cities. Census tracts are gen- erally drawn based on predominant land use. Thus, tract-by-tract pre- dictive models based on land use designation are not expected to be universally applicable. Only for residential surface areas, which correlate highly with size of the population or housing stock, can reasonably accurate tract-to-tract predictions be made. With some know- ledge of the construction patterns within a city, intracity distributions of residential surface quantities by type of material might also be estimated. The amounts of commerical and industrial surfaces as well as specific material types for such structures can probably be estimated only at the city level on the basis of population figures or the number and types of establishments. Table 11 lists per capita amounts of the six surface area categories used throughout the modeling effort. For comparison purposes, the figures in parentheses are regression coefficients and standard errors obtained from regressions through the origin of each material category on pop- ulation size, using the sampled tracts in each city. The figures in parentheses indicate that the "tightest" predictions (as indicated by the size of each coefficient relative to its standard error) are ob- tained for residential surfaces, as expected. With the exception of residential surfaces, differences in per capita estimates between the two cities are not remarkable. The substantially lower per capita estimates in Baltimore for residential wall surfaces can be explained by the fact that row houses sharing common walls on both sides are predominant in Baltimore but not in St. Louis; Figures 3 and 4 illus- trate this difference. The latter figure also illustrates the fact that the per capita roof values, which do not take account of roof pitch, are underestimates, particularly for residential structures. The per capita amounts are based on center city data for two older, well -developed cities. Thus, their application should be restricted to similar areas. With the exception of residential values, these estimates should be applied at the city rather than census tract level. It is expected that per capita amounts of material surfaces would be higher in suburbs and younger developing cities where population densities are lower. Although regression equations with significant coefficients could also be obtained for most specific material types in each city, these results were even less consistent. Therefore, a universally applicable tract-by- tract prediction model cannot be developed at this more detailed level. The material compositions for each of the six surface classifications. as estimated from the Sanborn survey, are shown in Table 12. For materials expected to be affected by air pollution, values are reasonably consistent between cities. There is a higher proportional use of brick in St. Louis walls as opposed to cinderblock, wood and iron in Baltimore walls, particularly for commerical and industrial structures. Baltimore has a higher proportion of noncombus- tible materials and lower proportion of composition materials in its roofs than St. Louis, particularly for residential structures. 44 Table 13 lists per capita values of specific types of materials grouped according to the expected effects of air pollution. There are differences between cities, some of which can be readily explained. For example, because residential surface area per person is lower in Baltimore than in St. Louis, the amount of brick should also be lower in Baltimore. As another example, the 15-year difference in Sanborn survey times for the two cities should affect the amount of sheet metal available for use. Because the market for sheet metal roofing and siding increased at a rate of approximately 7 percent per year during this period, it is expected that the cumulative amount in the early 1970's would be about three times that of the mid-1950's, other things being equal. Similar care should be taken in applying these estimates to other cities, since there are certain to be regional differences in the use of specific material types because of factors such as climate and local availability of materials. 45 _»ts_aiL >- JL-L i I 1 1 I i ' i I I I l~ l Figure 1. Sample Sanborn Atlas. 46 Detached units WA = 2R (HW + DH) H RA = RWD | Attached units: WA = 2RHW + 2DH RA = RWD r H s^ s *— w - -w - «a — -w — - D Circular structures where WA = RirWH RA = Rir (0.5W)' WA = wall area in square feet RA = roof area in square feet R = replication factor based on the number of identical structures H = building height in feet above ground level W = building width in feet D = building depth in feet. Figure 2. Exposed Surface Area Formulas 47 k p : .,J II HI'l !■•*«:«-'' '.!:■ Figure 3. Baltimore Residential Clock Face, 48 1 1 Figure 4. St. Louis Residential Block Face, 49 WALL MATERIAL TYPES Steel /metal Cinder block Brick Painted wood Concrete Stone • Marble • Iron • Asbestos • Slate • Tile ROOF MATERIAL TYPES • No roof • Composition • Noncombustible • Wood shingles t Asbestos shingles • Concrete • Steel/corrugated metal § Slate • Iron STRUCTURE TYPES § Detached unit • Row unit t Garden apartment • High rise • Parking facility • Garage service building • Storage building t Manufacturing facility • Storage tanks • Vacant lots Table 1. Material and Structure Types. 50 Baltimore Land Use Total Census Tracts Tracts in Sanborn Survey Tracts in Visual Survey Commercial Industrial Residential Mixed 5 13 142 41 5 12* 35 10 3 4 10 3 Total 201 62 20 * One tract was a government installation and could not be surveyed. St. Louis Land Use Total Census Tracts Tracts in Sanborn Survey Tracts in Visual Survey Commercial Industrial Residential Mi xed 5 15 89 17 5 15 23 5 3 5 6 2 Total 126 48 16 Table 2. Sampling Fractions for Sanborn and Visual Surveys 51 Land Use Characteristics t Residential acreage t Commercial acreage • Industrial acreage Housing/ Population Characteristics t Housing units per t Proportion of units residential acre in 5+- unit structures Proportion of units t Proportion of units aged 30+ years with < 1 person/room Establishment Characteristics • Commercial establishments • Industrial establishments per commercial acre per industrial acre • Average commercial t Average industrial establishment size* establishment size* * Number of employees Table 3. Candidate Predictors for Material Quantity Models 52 Dependent Variables Baltimore St. Louis Residential wall square footage 1,336,637.14 2,844,910.83 Residential roof square footage 758,663.95 1,272,275.98 Commercial wall square footage 571,576.90 1,110,115.50 Commercial roof square footage 480,052.71 670,306.23 Industrial wall square footage 579,198.64 624,099.72 Industrial roof square footage 901,867.05 695,907.98 Land Use Characteristics Residential acreage 62.15 102.75 Commercial acreage 48.16 27.17 Industrial acreage 35.42 127.17 Housing/Population Characteristics Housing units per residential acre 54.15 34.56 Proportion aged 30+ years 0.64 0.78 Proportion in 5+- unit structures 0.13 0.21 Proportion with <_ 1 person/room 0.84 0.75 Establishment Characteristics Commercial establishments per commercial acre 14.75 0.08 Average commercial establishment size 29.67 160.81 Industrial establishments per industrial acre 5.54 0.08 Average industrial establishment size 47.37 369.04 Table 4. Means of Dependent and Candidate Predictor Variables for Material Quantity Models. 53 Dependent Variables Baltimore St. Louis Residential wall square footage per acre 20,258.03 10,639.42 Residential roof square footage per acre 10,975.65 4,772.88 Commercial wall square footage per acre 10,049.08 4,394.98 Commercial roof square footage per acre 7,777.21 2,632.01 Industrial wall square footage per acre 2,651.48 2,123.35 Industrial roof square footage per acre 3,196.37 2,251.95 Land Use Characteristics Proportion of acreage designated residential 0.53 0.37 Proportion of acreage designated commercial 0.20 0.10 Proportion of acreage designated industrial 0.13 0.40 Housing/Population Characteristics Population per acre 67.22 25.41 Establishment Characteristics Establishments per acre 3.36 0.04 Table 5. Means of Dependent and Candidate Predictor Variables for Material Density Models. 54 Residential Wall Area Predictors Baltimore St. Louis Residential acreage Proportion of housing units aged 30+ years Average commercial establishment size Constant term n 2 10977.4 (8.37) 1051462 (3.78) Residential acreage Proportion of housing units aged 30+ years Average commercial establishment size Constant term n2 50099.0 0.53 19532.3 (8.15) 2393043 (3.16) - 2355.0 (3.31) 7755.3 -1405400 0.54 0.64 Residential Roof Area Baltimore St. Louis 5117.8 (8.04) 8381.3 (8.54) ;s" 606812.8 (4.50) 798616.5 (2.58) 775.5 (2.66) •335855.2 0.66 Table 6. Material Quantity Modeling Results for Residential Structures, 55 Commercial Wall Area Predictors Commercial acreage Housing units per residential acre Proportion of housing units in 5+- unit structures Commercial establishments per commercial acre Constant term R 2 Baltimore 4965.5 (2.89) 3745254 (4.97) -195653.3 0.50 St. Louis 51196.6 (19.52) 5150.5 ( 3.60) 2334543 ( 3.37) -636322.9 0.92 Commercial Roof Area Commercial acreage Housing units per residential acre Proportion of housing units in 5+- unit structures Proportion of housing units with <_ 1 person/room Commercial establishments per commercial acre Constant term n2 Baltimore 3484.0 (2.61) 2558228 (4.38) -49009.4 0.44 St. Louis 20957.9 (13.19) -1241317 (-3.40) 1356446 ( 3.31) 934765.6 0.87 Table 7. Material Quantity Modeling Results for Commercial Structures. 56 Industrial Wall Area Predictors Industrial acreage Housing units per residential acre Proportion of housing units with <_ 1 person/room Average industrial establishment size Constant term n2 Baltimore 13338.7 (10.27) -2600178 (-3.17) 2285467 0.66 St. Louis 8436.9 (4.57) 841.8 (4.21) 21844.0 0.56 Industrial Roof Area Industrial acreage Housing units per residential acre Constant term n2 Baltimore 25788.6 (12.23) -11633.0 0.71 St. Louis 11678.4 (4.83) 292299.0 0.34 Table 8. Material Quantity Modeling Results for Industrial Structures 57 Residential Wall Density Predictors Baltimore St. Louis Residential proportion of total acreage - 10997.09 (5.57) Population per acre 328.76 (10.69) 233.13 (7.38) Total establishments per acre -383.77 (-3.51) - Constant term -548.14 602.95 R 2 0.66 0.77 Commercial Mall Density Commercial proportion of total acreage Population per acre Total establishments per acre Constant term n2 Baltimore 27503.57 ( 3.33) 1281.23 (15.00) 9.40 0.82 St. Louis 73.82 (4.27) 89802.80 (25.17) -688.73 0.93 Industrial Wall Density Industrial proportion of total acreage Total establishments per acre Constant term n2 Baltimore 14522.34 ( 4.56) 93.18 ( 3.10) 462.63 0.38 St. Louis 4583.47 (3.76) 11804.64 (2.87) -136.98 0.34 Table 9. Wall Material Density Modeling Results 58 Residential Roof Density Predictors Baltimore St. Louis Residential proportion of total acreage 5592.59 (7.32) Population per acre 160.39 (12.76) 97.45 (6.72) Total establishments per acre -204.97 (-4. 59) - Constant term 884.93 205.29 R 2 0.73 0.80 Commercial Roof Density Baltimore St. Louis Commercial proportion of total acreage Population per acre Total establishments per acre Constant term r>2 18903.52 ( 3.10) 989.28 (15.66) 510.67 0.83 23626.21 (15.31) 23.61 -406.21 (2.01) 0.85 Industrial Roof Density Industrial proportion of total acreage Constant term n 2 Baltimore 20645.55 ( 5.19) 530.91 0.31 St. Louis 5116.26 (4.00) 199.54 0.26 Table 10. Roof Material Density Modeling Results 59 Baltimore St. Louis Residential wall area 30.3 (27.1 +1.8) 46.9 (37.8+2.1) Commercial wall area 8.0 ( 8.4+3.1) 10.1 (12.2+3.0) Industrial wall area 5.6 ( 8.8+3.5) 5.6 ( 5.8+1.7) Residential roof area 16.6(14.9+0.8) 21.1 (16.5+1.0) Commercial roof area 7.3 ( 7.5 + 2.3) 6.8 ( 8.0 + 1.4) Industrial roof area 8.6 (14.6 + 6.1) 7.3 ( 7.0 + 1.9) Table 11. Per Capita Exposed Material Estimates (m /person) 60 Baltimore St. Louis Wall Materials Residential Commercial Industrial Residential Commercial Industrial Steel/metal 0.02 1.75 8.02 0.06 0.27 5.03 Cinder block 2.87 8.25 9.73 0.35 2.59 2.38 Brick 56.72 50.15 32.87 60.70 66.83 64.67 Painted wood 17.55 5.12 11.26 17.67 2.15 6.45 Concrete 0.005 5.15 5.98 0.61 1.44 4.64 Stone 0.59 2.16 2.61 0.21 1.04 1.93 Marble - - - 0.11 - - Iron 1.66 1.64 15.16 0.10 0.33 4.01 Asbestos 0.44 0.74 4.24 0.02 0.17 0.88 Slate - - - 0.16 0.03 0.02 Tile 0.16 0.05 0.13 0.004 0.16 - Glass* 20.00 25.00 Baltimore 10.00 20.00 25.00 St. Louis 10.00 Roof Materials Residential Commercial Industrial Residential Commercial Industrial Composition 52.50 Noncombustible 47.13 Wood shingles 0.34 Asbestos shingles Concrete 0.03 Steel/corrugated metal Slate Iron 69.02 17.29 0.26 0.01 11.45 0.56 0.89 0.51 65.70 10.35 0.75 0.91 13.07 4.74 4.48 89.19 9.40 0.04 0.005 1.33 0.004 76.38 6.90 2.52 0.29 13.78 0.13 87.77 5.09 0.08 0.12 6.53 0.04 0.36 Set at a constant percentage by structure type on basis of visual survey Table 12. Material Composition Estimates (Percentage of Total Exposed Surface Area). 61 Materials Pollution Efft ;cts Baltimore St. Louis Steel /metal Corrosion 1.04 0.34 Cinder block/concrete/ stone/marble/slate Erosion 5.35 3.39 Brick/ tile Soiling 23.10 38.86 Wood/asbestos Soiling/erosion 7.01 9.16 Iron Corrosion 1.91 0.34 Glass Soiling 8.62 12.47 Other None 29.38 33.24 Table 13. Per Capita Estimates (m /person) for Selected Material Types Grouped by Expected Pollution Effects. 62 METHODOLOGIES AND TEST METHODS FOR SERVICE LIFE PREDICTION TENHO SNECK TECHNICAL RESEARCH CENTRE OF FINLAND SF-02150, ESPOO 15, FINLAND Abstract: Buildings, other structures, their parts and materials must perform over time in such a way that their properties exceed the minim- um accepted values for a certain time when properly maintained. This time is called the service life. Durability is an important aspect in this connection as it describes the ability of an object to withstand the effects of the many degrading factors originating from the environ- ment. The durability of a building element (e.g., roof, wall) is here studied as the durability of materials as a part of the structure. Development of methodologies and methods and the collection of data is needed to be able to predict the service life more precisely than before. Feed-back information from in-service conditions, practical, semi practical and field tests are performed in conjunction with laboratory tests with large or small test specimens. In all cases it is necessary to know or to collect information on the agents responsible for degradation, the degradation mechanism and the basic properties of the object. The data have to be processed. The increasing utilization of models, need of basic knowledge and application of statistical methods has to be noticed. Examples are given and some problem areas identi f ied. Key words: corrosion of reinforcement, degradation mechanisms, dura- bility, expert judgment, frost resistance, in-service tests, know how, laboratory tests, performance over time, plastics, practical tests, prediction of service life, probabilistic methods, service life, weathering. 1 Introduction The prediction of service life is a challenging task which is all the time increasing in importance. One difficulty lies in the definition of the object to be studied. A very careful analysis has to be made in order to be able to ask the right questions. For instance, considering the service lifes of building elements (e.g., walls, roofs) it is easier to study the behaviour of general products (bricks, lumber) or materials (burnt clay, wood) in the structure and to draw conclusions on the service life of the element from this. A model is needed which should take the "whole" into account. For instance, transport, erection and the application of new building methods may lead to new problems. The identification of the factor which limits the service life is important. In a Belgian study on durability, the limit of use was 63 considered and a summary is given by Ammar and Longuet (1). The task is then to estimate the time elapsing to reach this point. The real challenge lies in the development of methodologies in order to serve product development and regulatory purposes. 2 Col lect i ng data Long-term practical experience can be obtained by studying existing buildings. This kind of feed-back information, experimental buildings, exposing test specimens in buildings and other types of tests in practical circumstances are quite necessary. Also, field tests with large or small specimens can be used. The purposes of feed-back studies are twofold. They give a picture about what is really happening and they can also be used to check models and degradation mechanisms. As this kind of studies are lacking, a systematic collection of data would be important. Both a general methodology for such a practical feed-back system and methods for special cases are needed. Collaboration in this field would be recom- mendable and the work on the development of methods for inspecting the state of existing building structures should be accelerated. Building pathodology and failure investigations may serve the purpose of service life studies but this is not necessarily true as most building failures are caused by errors in some procedure. However, they may give valu- able information on degradation mechanisms and also on the way materials and components end their lives. In the laboratory the testing takes place in controlled conditions and it is often possible to follow the conducting of the test effectively. The test specimen may be small or large and even full scale tests are possible, as one in some laboratories may test small houses. Equipment exists for testing large specimens, for instance, apparatus for one- sided freezing, large weatherometers, roof weatherometers, apparatus for driving rain tests, testing of bathrooms, etc. As the apparatus for large specimen tests is expensive, a critical appraisal is quite in order, and it would be recommendable to list existing units. In order to succeed with the principle to set up service life require- ments for the building or its parts, there is an absolute necessity to develop predictive tests. It is quite evident that there have to be tests of different duration but there are no other alternatives than short time tests to assess the service life of new materials or old materials in new applications. The ASTM standard E 632-78 (2) is applicable to all components and materials used in building. A predictive service life test consists of an ageing test and a property measurement test. In the former, the material is exposed to the agents which induce changes in its properties in practice. To determine the effects of ageing properties are mea- sured before and after the ageing tests. The standard is divided in 64 four parts: Problem definition, Pre-Testing, Testing, and Interpret- ation and reporting of data. The standard contains a list of degrad- ation factors (agents), a matrix for identifying observable changes in the material and an informative flow chart of the procedure as a whole. It is worth while to mention that the matrix also contains the inter- faces in the structure. In order to succeed with predictive tests, basic knowledge about pro- perties and degradation mechanisms is needed. A combination of prac- tical and laboratory tests is mostly necessary. An imaginative approach should be adopted to all advanced equipment for measurements of materi- als properties. Also the measurement of the environmental agents causing degradation is of importance. 3 Processing data It is evident that both the collecting and processing of data must take place within the same framework. The acceptance of some systematics is necessary. The systematics (3) presented by I SO/TC 59/SC 3 for the application of the performance concept should be accepted. Especially, the lists of requirements and agents are relevant to durability studies. The ASTM E632-78 (2) list of degradation factors is based on the same way of reasoning. This standard should be adopted as a basis for the development of the prediction of the service life. According to these documents and chapter 2 following information will be processed when the service life prediction is written: Performance requirements of the object Acceptable performance Critical performance characteristics Agents affecting the object Degradation mechanisms Accelerated tests Other laboratory tests Short and long term practical tests Property measurements Suitability of methods used Model s Factors of design have to be included in the "agents" as the behaviour of a product may be drastically different in a structure. Models are quite essential for describing the relation between exposure time and property value and the relation between time to failure and the agent causing the failure. Models allow even prediction of failure before it occurs in the opinion of Clark (h) . A statistical treatment of the data is needed. The results obtained from the different tests can be used to check and doublecheck each other, to verify assumptions and to give evidence of probable shortcomings and errors which then have to be reported very carefully. - The importance of practical experience and know-how cannot be overestimated. 65 Durability problems are mostly of such a nature that judgment by experts is inevitable in the testing procedure. In this case the judgment should be as objective as possible by the application of strict rules. The decision making process, the procedure used and the factors taken into account have to be documented in such a way that the process can be reconstructed (5). k Examples Frohnsdorff, Masters and Martin (6) report on the application of the ASTM standard (2). The examples concern the resistance of concrete structures to freezing and thawing, statistical analysis of the fatigue life of deep grooved ball bearings and development of tests for corrosion protective coating systems for steel structures. A striking example of a prediction methodology is I EC 216 on the thermal endurance of electrical insulating materials (7). It is composed of four parts: Part 1 tells about the basic matters and mechanisms and gives advice how to conduct the test. Part 2 gives a selection of proposed test methods for determining certain vital properties before and after the endurance test. Part 3 and k give additional information on the application of statistical methods and the calculation of thermal endurance profiles. The procedure is based on the Arrhenius law of reaction rates. This is not valid if the ageing process consists of several reactions and the activation energies of these reactions differ. Paloniemi and Lind- strom (8) work with an equalized ageing process (EAP) where they after measuring the reaction rates of the important ageing reactions, try to equalize the acceleration factor by choosing appropriate gas concent- rations. - This is an example which shows the possibilities to improve a methodology once it has been put to work. Tuutti (9) reports on the breaking down of the corrosion process of steel embedded in concrete into sub-processes: an initiation period and a propagation period. By analyzing the action and effects of the more important agents, as carbon dioxide and chlorides in breaking down the passivity of the reinforcement and then estimating the corrosion rate, Tuutti is able to draw conclusions on the service life of the concrete in question. The frost resistance of porous building materials is a very important property in the cold regions. There is a possibility to choose among a multitude of test methods and theories. Fagerlund (10) tells about the prediction of the service life of concrete exposed to frost action. The method is built up from three theories and ideas dealing with the theory of a maximum allowable mean free spacing between air voids, the theory of the critical degree of saturation and the idea to deal with the critical degree of saturation as a materials property and separate- ly with the actual degree of saturation in practical situations. 66 5 Concluding remarks The examples show once more that prediction will only be possible if the degradation mechanism or mechanisms can be identified. It is clear that the clarification of this matter is difficult, especially in cases where a combination of different agents are acting upon a material. Test methods need careful attention. They are needed for many purposes and it is of vital importance to have methods which fit into the metho- dology of prediction in question. Papers on the state-of-the-art of practically applied prediction methodologies would be helpful. The whole problem is of such a magnitude that the need for international cooperation has to be stressed. 6 References 1 Ammar, C. and Longuet, M. , "Belgian requirements about buildings service life," Durabi 1 ity of bui lding materials and components , ASTM STP 691 » American Society for Testing and Materials, Philadelphia, 1980, pp. 77"90 2 ASTM, "Standard recommended practice for developing short-term accelerated test for prediction of the service life of building components and materials," ASTM E 632-78, Annual Book of ASTM Standards , Philadelphia, 1978 3 ISO/TC 59/SC 3, "DP 6241 Guidance for the preparation of perform- ance standards in building," London, 1979 h Clark, J.E., "Weathering," Encyklopedia of Polymer Science and Technolog y, Vol. 14, Interscience Publishers, New York, 1971. pp. 780-795 5 Nordtest, "Nordtest guidelines for the acceptance of test results," Doc Gen 012, Nordtest, Stockholm, 1979, p. k 6 Frohnsdorff, G. , Masters, L. and Martin, J., "An approach to improved durability tests for building materials and components," NBS Technical Note 1120, U.S. Department of Commerce, National Bureau of Standards, Washington, DC, 1980 7 IEC, "Guide for the determination of thermal endurance properties of electrical insulating materials," IEC Publication 216, Part 1: "General procedures for the determination of thermal endurance properties, temperature indices and thermal endurance profiles," Geneve, 197*+ Part 2 Part 3 Part h List of materials and available tests," Geneve, 197*+ "Statistical methods," Geneve, 1980 "Instructions for calculating the thermal endurance prof i le," Geneve, 1980 67 8 "Theory of equalization of thermal ageing processes of electrical insulating materials in thermal endurance tests," I E E E T r a n s a c t i on s on Electrical Insulation , Vol. EI-16 No. 1, 1981 Paloniemi, P., I: "Review of theoretical basis of test methods and chemical and physical aspects of ageing," pp. 1-6 Paloniemi, P., II: "The theory with practical approximations and application principles," pp. 7~17 Paloniemi, P. and Lindstrom, P., Ill: "Tests results on an enamelled wire, a polyester glass laminate and an epoxy casting resin," pp. 1 8- 30 9 Tuutti, K. , "Service life of structures with regard to corrosion of embedded steel," RILEM Qual ity control of concrete structures , Vol. 1, Swedish Cement and Concrete Research Institute, Stockholm, 1979, pp. 293-301 10 Fagerlund, G. , "Prediction of the service life of concrete exposed to frost action," Studies on Concrete Technology , Swedish Cement and Concrete Research Institute, Stockholm, 1 979 1 PP' 249-276 68 THE DEGRADATION AND PREDICTION OF SERVICE LIFE OF BUILDING COMPONENT Yoshitaka Ishizuka Government Buildings Department Ministry of Construction Kasumigaseki , Tokyo 100 Abstract: The Government Building Department of the Ministry of Construction in Japan makes a field survey on the state of non-wooden buildings of competent facilities of government offices every five years. This survey plays an objective indication from the technical point when re- building, enlargement, modification and so on are requested, The survey of the governmental facilities in each ministry is intended to get "Degraded Index" of structure, built com- ponent and materials individually, and finally the residual value of whole buildings will be objectively calculated. The conditions of climate and site influence the degrada- tion of buildings and therefore service life varies. According to the survey over the whole country, the data of "Degraded Index" of one of members which have same lapse of time show extensive distribution. Generally it shows the normal distribution. Mean of this normal distribution can be regarded as its typical value. A locus of means of "De- graded Index" data draws a line of degradation complying with lapse of time. The above-mentioned method predicts both service life of structure, built component and materi- als respectively and difference due to the conditions of climate and site. Key words: Field survey; Degraded Index; Accumulation; Degraded performance Activities of Construction of government buildings in Ministry of Construction are planning, designing, construc- ting, indicating and rebuilding competent facilities of government offices. 1. Survey on the Degraded Index of the Government Build- ings The Ministry of Construction makes a survey on the state of non-wood buildings of competent facilities of government offices every five years since 1969. The purpose of this survey is to obtain the present situation of degradation of whole facilities. This survey plays a role of objective indication from the technical point of view when rebuilding, 69 enlargement, modification and so on are requested to me. The original data sheet of the survey is entitled "Survey on the state of non-wood buildings" (Table-1 shows.) I can obtain the Degraded Index which shows the total degraded value of the buildings concerned by calculating each de- graded index of structures, major components and facilities and equipments. Scale of the degraded index " D" is defined as follows. D: 1 - 0.9 Not degraded or slightly degraded D: 0.9 - 0.7 Degraded parts can be observed, however, there is no problems about performance. Or minor improvement may be needed. D: 0.7 - 0.5 There are some degraded parts and partial improvement is needed. D: 0.5 - 0.3 There are many degraded parts and lowering of performance is remarkable. However it can be used if the improvement will be done. D: 0.3 - 0.1 Replacement should be done. In total 3,9 87 buildings were surveyed and the number of the buildings was classified by the used years. (Table-2 shows) I can get the estimated service life of the components of buildings by using the each degraded index :D, considering the conditions of location, climate and so on. 2. Estimation of the service life by the accumulated dis- tribution method The Degraded Index of the components is distributed from 1 to 0. When the value of ten times of degraded index is K, K is distributed from 10 to and the degradation of build- ing components is proceeded according to the decreasing of the number. When K is 8, degradation is slight and im- provement is not needed. When K is 7 , minor improvement is needed. When the intersection of the borderline between K7 and K8 and the line of accumulated distribution is sup- posed Al , all of the components are needed to be improved or replaced in the cases more than Al . (Figure-1 shows.) Because the minor improvement or partial replacement is sometimes done, accumulated distribution of K is uneven. Figure-2 shows the case of roof with asphalt water proof and concrete for protection of the asphalt. When the intersection of the extended border line between K7 and K8 and line of accumulated distribution is A2 , A2 is re- garded as the service life of roof with asphalt water proof and concrete for protection of the asphalt. These lines varies according to the conditions of location, climate and so on. 1) Estimation of the service life of roof with asphalt water proof and concrete for protection of asphalt All facilities 25.5 years Cold areas 21 years Areas except cold areas 27 years 70 Coastal areas 18 years Areas except coastal areas 28 years T: years Y(T): function of degraded performance The function of degraded performance shows the level of performance re- mained at certain year. T>5 Y(T)=-4.9T + 125 2) Estimation of the service life of roof with asphalt water proof (referred to figure-4) Function of degraded performance T>4 Y(T)=-4.8T + 115 3) Estimation of service life of aluminum outdoor fittings such as windows and doors (referred to figure-5) All facilities 40 years Cold areas 37 years Areas except cold areas 40 years Coastal areas 23 years Areas except coastal areas 40 years Aluminum outdoor fittings are generally durable in acid arid weak in alkali. Service life of fittings in coastal areas is almost half of that in the areas except coastal areas. T>4 Y(T)=-2.75T + 110 3. Estimation of the service life by the performance limit method The Degraded Index of the building components is distrib- uted from 1 to . When the value of 100 times of degraded index is P, P is distributed from 100 to and shows the normal distribution. Median of the normal distribution is representing value of P. Therefore the locus of the median of P in time shows the degradation line. As showed in figure-5, in the period between the initial performance level and performance- lost-level, when the performance limit reaches certain permitted level, improvement or partial replacement is done in order to recover the performance. When the initial performance level is 100, P50 is regarded as the lower limit permitted for improvement or partial replacement because lowering of performance is remarkable in case of under P50 . When the point, at which the locus of median of P in time intersects the line of P50, is B, B shows the service life of components. Figure-6 and 7 show the Degraded Index of the tile of out- door wall. The intersection of the degradation line of the tile of outdoor wall and the line of P50 is regarded as the service life of the tile. Urban area G(T)=-1.67T + 100 Coastal areas G(T)=-1.92T + 100 Areas between mountains G(T)=-1.79T + 100 Cold and snow areas G(T)=-2.17T + 100 71 4. Conclusion The history of the improvement of components of each build- ing varies and the locus of degraded index is uneven. The Degraded Index is effective to analize the service life of building components because it can obtain the tendency of their degradation. 72 T O.80 (ex) 0.140 (ex) o.iSO (ex) a*80 (ex) 0. 80 ■finish KOO-f dspkd/t W4.t«t- p^eo-f; 4sp^«.R w*ter pt-oo-f if d concrete outdoH~ w*JI C0.O23) tilt; »«o»-t 1 0.76 ecuip lu*ii*4ire eleciv-Jc wi^e to. 015; ttlep^one (.O.Oi) (,o.of5) plu-r*bi 1 0.00s) ait- Kawdli^i t pdt-tj4.l -to£.S " t ■ used yea^S -5 iio -15 *~* ^15 -JO ** *-ibu"tio'»i K «L«d co<*»c»-et« -fov- pc-o"Uc- iio-h of *.*ph«.lt (%) uu w VA \ v\ k I \ rf N 1 V 10 \ t~ lL_ 10 io a J ° *o u£«d yed>£ Figure- 3 AspKiK vw4-tet- P»-oof Fi K>0 •O iO A io 40 74 pe»-f ok- •»<.<* ce li*»it Td.ble-3 .Sehvice life of fcmiMi*! co-npo^iie'Mt 4**$ Fifuke-6 Tile o-f outdoor ( Conditio* o-f site) r- service life Asp^K: wii«t- P»-oe>f (/•AV-J) 4«4 e»*icre-te. • 4II "ficilitief 25 • co|ot &\rtd.$ 21 ' Ar«dt *xce»t cold • COdrt&l 4>«4i • 4t-««A$ except coaj-ta.| Ahe,£ Ffjut-e-7 Tile of outdoor ( CCM o/f fiCJ4l of clj indie) Ajpki-K wd-tev- pt-oo-f Til« of outdoor . eLt-ed-S bei*««*i • Co(ol (Utl 5*0* S£i Wcoet -f ifctj-wjf of i«idoo>- 24 19 30 28 26 23 iO 30 '■j x If Ai»eB.i ej ecept cold dv-edj 10 20 30 40 SO Ufec< yed»» 75 PREDICTION OF THE SERVICE LIFE OF BUILDING STRUCTURES EXPOSED TO FROST ACTION Goran Fagerlund Cementa AB DANDERYD, Sweden Abstract: The paper presents a method for calculation of the service life of building structures that are exposed to frost action. The method can be regarded as a coalescence between theories and ideas that have previously been presented by Powers, Warris and the author; viz. the theory of a critical -or maximum allowable- mean free spacing between air voids, the theory of the critical air void absorption or critical degree of saturation which will be a function of the critical spacing and of the air void distribution and the idea of separating a frost resistance problem into one part that is only a function of materials properties - the critical degree of saturation - and one part that is also a function of environmental characteristics - the actual degree of saturation. The report contains: (i) a method for calculating the effect of air void absorption on the air void spacing; (ii) relations between mate- rials properties and the critical spacing factor; (iii) methods for determination of the critical air void absorption and the critical degree of saturation; (iv) a discussion of the concept "service life" showing that the main problem lies in estimation of the time process of air void absorption; (v) a theoretical treatment of the long-term absorption process in air void systems; (vi) an attempt to analytically link the service life to materials properties; (vii) a presentation of a purely empirical method for estimation of the service life. 76 TIME TRANSFORMATION FUNCTIONS COMMONLY USED IN LIFE TESTING ANALYSIS Jonathan W. Martin Materials Research Engineer National Bureau of Standards Abstract: The accelerated aging problem and the literature are reviewed within the framework of reliability theory and life testing analysis. In accomplishing this objective the assumptions and the generalized equation for accelerated aging are derived having the form F S (t) = F S (p(t;S,0)) I where F5 (t) is the life distribution at the accelerated stress level; Fc is the life distribution at the in-service stress level; and p(t,S,0) is the time transformation function. The time transformation function is of particular interest because it relates material or system perfor- mance to the stress environment. For demonstrative purposes, an extreme value distribution having wide acceptance is reparameteritized in terms of commonly published time transformation functions. These transformation functions provide a chemical/physical basis for modeling the time-dependent degradation process of a wide-range of polymeric materials. Examples of applica- tions from the literature and research completed or in process at the National Bureau of Standards are given. Keywords: Arrhenius model; Eyring model; inverse power law; reliability; superposition principle; time transformation function. 1. INTRODUCTION Durability properties are established for a material in one of two ways — outdoor exposure tests and accelerated aging tests. Accelerated aging tests are experiments in which one or more of the stress variables typically encountered in-service are elevated to levels higher than nor- mal, thus increasing the rate of degradation. Accelerated aging tests are preferred to outdoor exposure tests because outdoor exposure tests are typically (1) time consuming, (2) expensive, (3) represent durability results only for a given test location, and (4) make material durability improvements difficult, since It is often impossible to quantify the stress history to which a specimen was exposed. Accelerated aging tests correct these deficiencies, but require that a methodology exists for extrapolating performance from accelerated stress levels to in-service conditions and that new failure mechanisms are not introduced at the accelerated stress levels. Models relating material degradation response 77 to the imposed stresses are referred to as time transformation functions. Whenever possible, these time transformation functions should model the time-dependent chemical/physical degradation. Models having a chemical/ physical basis lend validity to extrapolations to in-service stress levels. Such functions are also useful in determining if new failure mechanisms are introduced. The objectives of this paper are 1) to mathematically describe the accelerated aging problem in terms of reliability theory and life testing analysis; 2) to identify from the literature time transformation func- tions having practical application and obeying the proposed reliability format; and 3) to present National Bureau of Standards research results for completed and on-going research following the format of the proposed analysis procedure. 2. MATHEMATICAL PRESENTATION OF THE ACCELERATED AGING PROBLEM Mathematically, the accelerated aging process can be stated as follows. Let Z denote a vector of stresses to which the material is typically expected to be subjected to in-service. Included in Z are mechanical, thermal, electrical, diluent, and irradiance stresses. Let S be any combinations of these stresses such that: S < S X < ... < Si < S 1+1 < ... < S m ; (1) where Sq is the stress level at normal conditions; S m is the stress level at the most severe condition; and **i ^ ^i+1 indicates that the i stress combination is less severe than the (i+l) tn combination. Assuming the above, it is the mission of the accelerated aging program to predict the performance of the material at stress level Sq from its performance at stress levels greater than Sq. The experimental proce- dure for determining performance at the in-service stress level is as follows: (1) randomly select n^ specimens (1=1, 1, 2 ..., m) from a nominal population of material for each of m different stress levels; (2) for each stress level, subject the specimens to its predesignated stress combination and observe times to failure: and (3) terminate the test for the i tn stress combination when the r? failure (r^ < n.) is observed. Since all n^ specimens are exposed to stress combination, S^, simultaneously the observed times to failures are ordered; that is tli < t2i < •*• < tji < •*• < tri ( 2 ) where t.j£ is the j time to failure for specimens exposed to stress combination S^; and r i < n i- These ordered times to failure form an empirical cumulative distribution function. To this empirical cumulative distribution function a "theore- tical" cumulative distribution is fit such that the probability that a specimen will fail at or before time t is denoted 78 V t! S i ) = P Si (T * ° (3) where F5 (t) is the "theoretical" cumulative distribution function * (henceforth called the life distribution) of the random variable T for specimens subjected to stress S^; P ( • ) is the probability of the event in parentheses; 0^ is a parameter vector for Fg (t); and t is a time. * Assuming Fe (t; 0-i ) is continuous, then, the probability density i function, fg (t), exists. Denoting the reliability of the material by 1-F S (t^. 0i) = Rg (t), the hazard rate, Xg (t), is defined by i * i i f S,(t) A s (t) = * - for R S (t) > b i R q (t) b i (4) b i and for t > 0. The hazard rate is the limiting value of the probability that a device will fail in the next small time interval given that it has survived up to the start of the interval, divided by the interval length. An acceleration of the degradation process has occurred if F S (t; Gi) < Fg (t; i+1 ) for all t > (5) and i = , 1 , . . . , m-1 . An alternative and a stronger condition for acceleration (Proschan, Singpurwalla 1979; Allen 1959) is defined as \M < x s i+1 Ct) (6) for all t>0 and i=0, 1, ..., m-1. Eq. 6 forms the basis for some of non-parametric approaches. Given eq. 5, a function p(t; S, 0) exists relating the life distribution at any elevated stress level I to the life distribution at normal condi- tions as follows: F S (t;0) = Fg (p(t; S, 0)) (7) where p(t; S, 0) is the time transformation function relating the parameter vector to applied stress S; hence, the time transformation function relates performance of the material to its operating and ser- vice environment. Commonly, the acceleration function is assumed to be independent of exposure time; that is p(t,S!0) = Y (S,0) • t. (8) Assuming the inverse function F~l(t;0) exists, then p(t;S,0) can be determined by p(t;S,0) = Fg 1 (F s (t;0)). (9) I 79 Implicit in eq. 9 is that three conditions can be satisfied: (1) the life distribution F(t; 0) belongs to a known or derivable parametric family; (2) the form of F(t; 0) does not change from stress level to stress level; the value of the parameters, however, may change; and (3) the form of the time transformation function is known or can be hypothe- sized. Nonparametric procedures are used when conditions (1), (2), and/ or (3) can not be made. 3. TIME TRANSFORMATION FUNCTIONS In proposing mathematical models for predicting building material service life, the following considerations were taken into account: • Since flaw size and locations are known to be random variables, the model must be statistical. • Since fracture is thought to be predicated on the maximum flaw size, the choice of a distributional model, F(t), should be taken from the limiting class of distributions having minimum times to failures. The Weibull distribution belongs to this class and is currently the most popular distribution used in modeling failure times (Mann, Schafer, Singpurwalla 1975). For demonstrative purposes, therefore, it is used here. A material justification for this choice is given by Epstein (1948), Bolotin (1969), and Cohen (1974). • The last consideration is that in reparameterizing, an attempt must be made to functionally relate the Weibull distribution parameters to temperature, mechanical stress, and the presence of swelling agents. This is to reflect the strong dependence viscoelastic processes have on the intensity of these factors. These conditions are considered in the arguments which follow. The cumulative distribution function for the Weibull distribution is given by F(t;a, 3) = 1 - exp(-(t) a ) (10) 3 for a, 3, and t > 0. In eq. 10, a is a shape parameter and 3 is scale parameter. The shape parameter is often thought to be a material con- stant independent of the operating and stress environment. (Halpin, Kopf 1970; Halpin, Polly 1967; Whittaker, Besumer 1969; Saunders 1976). In some cases, however, this assumption does not appear to be valid (Martin, Saunders 1981; Chiao, Moore 1971) in that empirically the shape parameter appears to be functionally dependent on the imposed stress. The scale parameter, 3, on the other hand, is definitely a function of the stress environment. Halpin and Polley (1967) represent this dependence by 3 - 2 (11) a a a T a d a I where is a generalized crack growth vector; 80 a a an acceleration factor for mechanical stress; a are empirical constants. R Q (T) = In J< = In A - B/ T Under certain assumptions it can be shown that the acceleration factor for the Arrhenius model (Thomas 1963; Singpurwalla 1973) is given by a T = exp {-B(l/T - 1/T )} (17) and for the Eyring model (Thomas, 1963; Singpurwalla, 1973) a„ = T_ exp {-B(l/T - 1/T )} (18) T o where T is the reference temperature; and T is the temperature of interest. Practical applications of the Arrhenius and Eyring models abound. Examples of applications combining mechanical and thermal stress include: (1) the failure of epoxy adhered tensile lap shear specimens (McAbee, Tanner, Levi 1970); (2) the failure of sode-lime glass (Charles 1958) stressed in bending; (3) the time to failure for nylon-yarn stressed in tension (Coleman 1956); (4) the failure of rocket propellant material (McAbee, and Levi 1967); and (5) the time to failure of PMMA and Nylon 6 yarn subjected to mechanical and thermal stresses (Zhurkov, Tomashovsky 1966). Research in the Center for Building Technology, National Bureau of Standards indicates that the Arrehenius model adequately describes change in transmittance for many solar cover plate materials as a function of thermal environment. Examples are given in figures 3 and 4. 3.2.2 Thermal Acceleration Factor Equation at Temperature Above the Glass Temperature In the temperature range T g - 20 °C < T < Tg + 100 °C (where Tg is the glass transition temperature) time-temperature viscoelastic, and diele- ctric properties can be quantitatively predicted using the time- temperature superposition principle proposed by Williams, Landel, and 83 Ferry (W-L-F equation). The acceleration factor for the time-temperature superposition principle is given by 1„ a T - =°1 - ( -«o>. (») C 2 + (T-T ) where T is the temperature of interest T Q is the reference temperature, and Cl» C2 are empirical constants equal to 17.44 and 51.6 (Williams, Landel, Ferry 1955). Many applications were found using the time-temperature superposition principle. Doyle (1957) used it in modeling the effect of different aging temperatures on the short-term strength of adhesive laminates. Fujita and Kishimoto (1958) modeled changes in the relaxation modulus of polymethyl acrylate and polyvinyl acetate in water and in methanol at different temperatures. They concluded that temperature and moisture acceleration factors are multiplicative; hence, this provides an empiri- cal validation of eq. 11. Brunt (1962) repeated Fujita' s and Kishimoto' s experiment using a modified-oil alkyd paint. Brunt agreed with Fujita and Kishimoto 's conclusion that the acceleration factors due to moisture and temperature are multiplicative. Kwei (1966) also studied the time to failure in tension of highly crosslinked epoxy resins at different temperatures, and successfully modeled failure times using the superposi- tion principle. Other researchers using the superposition principle include: Hefty (1966) on the effect of temperature on short-term proper- ties of cast epoxies; Halpin and Polley (1967) for predicting time to failure of tensile loaded styrene-butadiene copolymer (SBR) coupons; and Kaelble (1965) in studying changes in the dynamic tensile deformation and ultimate strength properties of epoxy resins. Research is currently in progress in the Center for Building Technology, National Bureau of Standards on organic coatings for structural steel and adhesive systems for composite panels used in military tactical shelters subjected to the combined effect of diluents and temperatures greater than the glass transition temperature. It is felt that eq. 19 might adequately describe the behavior of these materials. 3.3 ACCELERATION FACTOR FOR DILUENTS Diluents, like temperature, have the effect of increasing the free volume of the polymer (Brydson 1972); hence, diluents have the same effect on the time-scale as increases in temperature; that is, a superposition principle for diluents holds. Based on only a few empirical studies, an equation of the form (Halpin and Polley 1967; Halpin 1968) in " C l (1 - c Q /n) (20) In aj = — — 7- d C 2 + 1 - c Q /n seems to adequately model changes in diluent concentration; where 84 c is the concentration of diluent; n is the density of the polymer at temperature T; and Cl> ^2 are empirical constants. As mentioned in section 3.2.2, Brunt (1962) and Fujita and Kishimoto (1958) demonstrated superposition for diluents. Halpin (1968) cited similar success with DP 600 and DP 2060. 3.4 ACCELERATION FACTOR FOR PHOTODEGRADATION Polymeric building materials (coatings, sealants, plastics, roofing material, etc.) exposed to near ultraviolet radiaton (290 nm < X < 400 nm) photolytically degrade over time (Schnabel, Kiwi 1978; Koike, Tanaka 1980). Degradation results from high energy photon absorption by the polymer chain, chromophores, or impurities within the materials. Absorp- tion of high energy photons can result in chain scissions, increased oxidation, the formation of crosslinks and conjugated double bonds. Macrostructural changes resulting from photooxidation manifest in the form of discoloration, cracking, embrittlement , shrinkage, crazing, etc. The acceleration factor for photolytic degradation, aj, therefore, has the form ai = (f) < 21 > o where A is a constant; aj is the acceleration factor for photolytic degradation; I is the reference relative intensity; and I is the relative intensity of interest. For transparent polymeric materials, the value of the constant A is approximately one (Maxim, Kuist 1963; MacCallum, Schoff 1971). For opaque material, the value of the constant A has been reported to be less than one (Koike, Tanaka 1980). Research is currently in progress in the Center for Building Technology, National Bureau of Standards to functionally relate the synergistic effects of temperature and irradiance on the change in molecular weight of commercial PMMA. Results to date indicate that the value of A in eq. 22 is equal to one for PMMA specimens exposed at 115°C. This was empirically demonstrated by exposing specimens at three known irradiance levels and stopping the tests at times such that I l t l = I 2 t 2 = I 3 t 3* Peak molecular weights as a function of exposure at the three irradiance levels are diagrammed in figure 5. 85 4. SUMMARY AND CONCLUSION The accelerated aging problem and the literature are reviewed within the framework of reliability theory and life testing analysis. In accom- plishing this objective the assumptions and the generalized equation for accelerated aging is derived having the form F s (t) = F S (p(t;S,0)) I where Fs (t) is the life distribution at the accelerated stress level; Fg is the life distribution at the in-service stress level; and p(t,S,0) is the time transformation function. The assumptions involved in deriving this expression are detailed. The time transformation function is of particular interest because it relates material or system performance to the stress environment. For demonstrative purposes, an extreme value distribution having wide acceptance is reparameteritized in terms of commonly published time trans- formation functions. These transformation functions provide a chemical/ physical basis for modeling the time-dependent degradation process of a wide-range of polymeric materials. Examples of applications are given. 5. REFERENCES [1] Allen, W. R. Inferences from tests with continuously increasing stress. J. Oper. Res. Soc. 303-312: 1959. [2] Andrews, E. H. Fracture. in Polymer Science Vol. I; Jenkins, A. D., editor, Am. Elsevier Publ. Comp: New York: 579-619; 1972. [3] Bolotin, V. W., Statistical methods in structural mechanics. San Francisco, Holden-Day; 1969. [4] Brunt, N. A., The influence of swelling by water vapor on the mechanical properties of a polymer. Kolloid-Zeitschrift 185:119; 1962. [5] Brydson, J. A., The glass transition, melting point and structure in polymer science, in Polymer Science, A. D. Jenkins, ed. New York, North Holland; 1972. [6] Bueche, F., Tensile strength of plastics above the glass temperature. J. Appl. Phys. 26(9): 1133-1140; 1955. [7] Bueche, F., Tensile strength of plastics below glass temperature. J. Appl. Phys. 28(7): 784-787; 1957. [8] Bueche, F., Tensile strength of rubbers. J. Polym. Sci. 26: 189-200; 1957. [9] Bueche, F. Tensile strength of plastics: effects of flaws and chain relaxation. J. Appl. Phys. 29(8): 1231-1234; 1958. [10] Charles, R. J., Static fatigue of glass II. J. Appl. Phys. 29(11); 1554-1560; 1958. [11] Chiao, T. T. ; Moore, R. L., Stress-rupture of S-Glass/epoxy multifilament strands. J. Comp. Mat. 5: 2-11; 1971. [12] Cohen, J. W., Some ideas and models in reliability theory. Statistical Neerlandica 28: 1-10; 1974. 86 [13] Coleman, B. D., Application of the theory of absolute reaction rates to the creep failure of polymeric filament. J. Poly. Sci. 20: 447-A55; 1956. [14] Coleman, B. D., Time dependence of mechanical breakdown phenomena. J. Appl. Phys. 27: 862-866; 1956. [15] Coleman, B. D.; Knox, A. G., The interpretation of creep failure in textile fiberglass as a rate process. Textiles Res. J. 27: 393; 1957. [16] Doyle, C. D. Application of the superposition principle to data on heat-aging of plastics. Modern Plastics 34(11): 141; 1957. [17] Endicott, H. S.; Hatch, B. D.; Schmer, R. G. Application of Eyring model to capacitor aging data. IEEE Trans. Comp. Parts CP-12: 34-41; 1965. [18] Epstein, B., Statistical aspects of fracture problems. J. Appl. Phys. 19: 541-549; 1948. [19] Fallow, B.; Burguierre, C; Murel, J. F., First approach on multiple stress accelerated life testing of electrical insulation. In 1979 Annual Report Conf. on Elec. Insulation and Dielectric Phenomena. National Academy of Science, Washington, D.C.; 1979. [20] Fujita, H.; Kishimoto, A., Diffusion - controlled stress relaxation in polymers: stress relaxation in swollen polymers. J. Polym. Sci. 28: 547-567; 1958. [21] Hagan, R. S.; Thomas, J. R., Plastics part properties and their relationship to end-use performance. 67th Annual Meeting ASTM June 21-26; 1964. [22] Halpin, J. C, Fracture of amorphous polymer solids: time to break. J. Appl. Phys. 35: 3133; 1964. [23] Halpin, J. C, Introduction to viscoelasticity . in Composite Materials Workshop, S. W. Tsai, J. C. Halpin, N. J. Pagano, eds., Technomic, Stamford, Conn.; 1968. [24] Halpin, J. C.J Polley, H. W., Observations on the fracture of viscoelastic bodies. J. Comp. Mat. 1: 64-81; 1967. [25] Halpin, J. C. ; Kopf, J. R. ; Goldberg, W., Time dependent static strength and reliability for composites. J. Comp. Mat. 4: 462; 1970. [26] Hefty, R. W., Prediction of long-term effects of humidity on cast epoxies. Modern Plastics: 163-168; 1966. [27] Hsiano, C. C. ; Sauer, J. A., On crazing of linear high polymers. J. Appl. Phys. 21 (11): 1071-1083; 1950. [28] Kaelble, D. H. , Dynamic and tensile properties of epoxy resins. J. Appl. Polym. Sci. 9: 1213-1225; 1965. [29] Koike, M.; Tanaka, K. , The quantitative estimation of degradation of rubber and plastic sheets and films in outdoor exposure. Rept . Res. Lab. Eng. Mat. Tokyo Inst. Tech. 5: 143-152; 1980. [30] Kwei, T. K., Strength of epoxy polymers I. Effect of chemical structure and environmental conditions. J. Appl. Polym. Sci. 10: 1647-1655; 1966. [31] MacCallum, J. R. ; Schoff, C. K. Photolytic decomposition of polymethacrylates. Trans. Faraday Soc. 67: 2372-2382; 1971. 87 [32] Mann, N. R. ; Schafer, R. E.; Singpurwalla, N. D., Methods for statistical analysis of reliability and life data. New York, Wiley. [33] Martin, J. W., The analysis of life data for wood in the bending mode. Wood Science and Technology 14: 187-206; 1980. [34] Martin, J. W. ; Saunders, S. C, The relationship between life test analysis and tests at constant stress ratio in structural use of wood in Adverse Environments. R.V. Meyers, R. M. Kellogg, eds.; New York: Van Nostrand - Rhinehold, 1981. [35] Maxim, L. D. ; Kuist, C. H. U. V. Degradation of polymers: I scissoring polymers. Am Chem Soc. Div. Polymer Chem Preprints 4(2):352-359; 1963. [36] McAbee, E.; Levi, D. W., Treatment of propellant mechanical property data by reaction rate analysis. J. Appl. Poly. Sci. 11: 2067-2069; 1967. [37] McAbee, E.; Tanner, W. C; Levi, D. W., Prediction of failure times for some adhesive bonded joints. J. Adhesion 2: 106; 1970. [38] Nelson, W. B.; Hahn, G. J., Linear estimation of a regression relationship from censored data. Technometrics 14: 247-249; 1972. [39] Proschan, F.; Singpurwalla, N.D., A new approach to inference from accelerated life tests. In Optimizating Methods in Statistics, J. S. Rustagi, ed . , Academic Press; 1979. [40] Rusch, K. C, Time-temperature superposition and relaxation behavior in polymeric glasses. J. Macromol. Sci.-Phys. B2(2): 179-204; 1968. [41] Schnabel, W.; Kiwi, J. Photodegradation. in Aspects of Degradation and Stabilization of Polymers, Jellinek, H. H. G [ed], Elsevier Scientific Publ.: New York; 1978. [42] Singpurwalla, N. D., Inference from accelerated life tests using Arrhenius type re-parameterizations. Technometrics 15(2): 289-299; 1973. [43] Thomas, R. E., When is a life test truly accelerated? Electronic Design 64-70; 1964. [44] Thomas, R. E.; Gorton, H. C., Research toward a physics of aging of electronic component parts. In Proc. Sec. Ann. Symp. on Phys. of Frac. in Elec; 1963. [45] whitman, L. C.; Doigan, P. Calculation of life characteristics of insulation. Elec. Eng. 73: 541; 1954. [46] Whittaker, I. C.; Besumer, P. M. , A reliability analysis approach to fatigue life variability of aircraft structures. Boeing Comp. Tech. Rept. AFML-TR-64-65. [47] Wu, E.M.; Ruhmann, D.C., Stress rupture of glass-epoxy composites: environmental and stress effects, in Composite Reliability, ASTM- STP 560: 263-287. [48] Zhurkov, S. N. ; Tomashevsky, E. E. In Physical Basis of Yield and Fracture. Inst, of Physics 200; 1966. 87A (Figures unavailable for printing) 87B SO AND NO ATTACK ON MARBLE K. Lai Gauri and Joel A. Gwinn Department of Geology and Department of Physics University of Louisville Belknap Campus Louisville, Kentucky 40292 Abstract: The reaction of marble (CaC03) with SO2 in a controlled at- mosphere produced calcium sulfite hemihydrate, 2CaS0o.H.20. A 10 ppm concentration produced nearly 13 mole percent sulfite in 24 hrs. At this concentration a steady state had been reached. Some of the reacted samples were placed on shelves of a closed container with water under- neath so that moisture was able to condense as droplets at the sample surfaces. In a few days the 2CaS03«H.20 was converted to CaSO^^^O. Some other reacted sulf ite-bearing samples were stored away in 1974 in a room which experienced normal indoor temperature and humidity con- ditions of 65-75°F and 60-80% R.H. These samples, reexamined in Feb- ruary, 1981, had exactly the same quantity of 2CaS0n.H20 as they had nearly 7 years ago. Also some samples with sulfite were immersed in deionized water for a few hours to determine the ionic composition. The analyses revealed the presence of SO^only. These observations reveal that moisture in the liquid phase is essential for oxidizing calcium sulfite into gypsum. Marble samples were also exposed to water-satu- rated atmospheres enriched in NO2 ; the NO2 was obtained from permeation tubes in the same fashion as SO2 . In none of the exposed specimens were traces of nitrate found. Nonetheless, marbles weathered in a natural environment do contain NO3 . Since the ratio of nitrates to sulfate has been increasing in the atmosphere, this ratio in the stone may serve as a fingerprint of the history of acidity of precipitation. Key words: Acid precipitation; Air pollution; Deterioration of marble monuments; NO2 - Calcite reactions; SO2 - Calcite reactions; Weather- ing. The burning of fossil fuels, especially of coal, produces gases and particulates which attack marble. Among the gases, SO2 and NO2 are most deleterious. These gases and certain particulates dissolve in water and descend as acid rain. In the northeastern United States, pH values below 4 are now quite common for rain even in areas which received nearly neutral precipitation in the fifties [1]. The emanations re- sponsible for this, generated by the coal plants in the vicinity of the Ohio Valley, contributed 92% of emissions east of the Mississippi River [2]. Another point in consideration of the acidity of precipitation with relation to marble is the relative proportion of CO2, SO2 , and NO2 in the atmosphere. In the fifties and earlier, when the precipitation had 88 started to become significantly acidic, CO2 was the main cause of mar- ble decay. Then SO2 became the main source of acidity, since CO2 in the presence of even very small amount of SO2 is no longer functional as a source of acidity. Presently the NO2 concentrations have increased considerably, so that the present proportions of H2SO4 and HNO3 are 65- 70% and 30-35% respectively [1]. The following treatment of the mechanism of S02~calcite reaction, the mechanism of oxidation of S02> and of the SO4 /NO3 ratios in stone is based primarily on our laboratory studies. Further, this article pre- sents the relationship between the quantities of sulfur in the environ- ment and the degree of deterioration of marble. Marble specimens were reacted with artificial atmospheres enriched with 5-12 ppm SO [3, 4]. These atmospheres were generated (Figure 1) by passing humid air over the permeation tube at 20°C. In the reaction chamber the specimens were suspended above water, the SO2 concentration of which was in equilibrium with the SO2 partial-pressure of the given atmosphere. The air flow insured that condensation of moisture did not occur on the specimen surfaces. These experiments revealed that S02~calcite reaction must have proceeded in the following fashion CaCO. S0 2 +H 2 CaSO .1/2H + CO Figure 2 indicates that reaction equilibrium is achieved within 24 hours suggesting the formation of a surface layer of sulfite imperme- able to SO2. Finally, mass absorption calculations, based on the sul- fite/ calcite diffraction intensity ratio [4, p. 118], lead to a film thickness of less than 10 microns. N0 2 - C0 2 PERMEATION TUBE SCRUBBER MARBLE SLABS REACTION CHAMBER FIG. I DYNAMIC N0 2 OR C0 2 ATMOSPHERE FIG. 2 REACTION RATE OF SOg WITH MARBLE 89 The x-ray dif fractometer was the only analytical tool which we employed for determination of the reaction rates. This has limitations imposed by the orientation of the growing sulfite crystals as well as the pre- ferred orientation of calcite in marble. We plan to repeat these ex- periments using x-ray fluorescence to acquire improved data for the cal- culation of the reaction rates. These experiments have also provided information on the mechanism of oxidation of SC^. In a plume, the SO2 is considered to oxidize to SO3 which then presumably forms H2SO4. Because the conversion of S0„ into SO3 [6] is, thermodynamically, a rather difficult process, the occur- rence in the plume of ammonia, metallic ions, and photosynthesis, etc. are considered to catalyze the reaction. In our experiments, some of the reacted sulf ite-bearing samples were wrapped in paper towels and stored away in 1974 in a room which had ex- perienced normal indoor temperature and humidity conditions of 65-75°F and 60-80% R.H. Reexamined in February 1981, these samples had exactly the same quantity of 2CaS03-H20 as they had nearly seven years ago. Also, in 1974, some of the reacted sulfite bearing samples were placed on shelves in a closed container with some water underneath so that moisture was able to condense as droplets on the sample surfaces. In a few days 2CaS0 3 .H 2 had changed into CaSO^I^O. Further, in 1981 some of these reacted samples were immersed in water for the determ- ination of the ionic composition of the reaction product; only SO4 and no SO^ - were detected. The conclusion drawn by us is that in nature, as in the laboratory, the initial product of S02~calcite reaction is the sulfite which then changes into sulfate in the presence of liquid moisture. This view is in variance to that commonly held that it is sulfuric acid which forms the calcium sulfate. Our conclusion is in harmony with the preponder- ance of sulfite particles in a plume near its source where the tempera- ture is high and condensation of moisture does not occur. In view of this, it is necessary that reaction rates be calculated on the basis of conversion of calcite into calcium sulfate via calcium sulfite rather than from calcite directly to calcium sulfate by conversion of SO2 into SO3 and its subsequent solution in water to form H2SO4. In further experiments, marble samples were exposed to water saturated atmospheres enriched in NO2 (6 ppm and 12 ppm) in the same fashion as in the SO2 reactions described above. In none of the exposed specimens were traces of nitrate found. Nonetheless, marbles weathered in a nat- ural environment do contain 3 . Since the ratio of nitrates to sul- fate has been increasing in the atmosphere [1], this ratio in the stone may serve as a fingerprint of the history of acid precipitation. How- ever, this would have to be corroborated with detailed field and lab- oratory studies. Finally, even small quantities of SO2 in the air can cause serious dam- age to the stone. For instance, the SO2 concentration of Louisville air does not normally rise above 0.05 ppm. Yet gypsum crust on marble 90 surfaces is common [7]. Essentially, SO2, by chemical reaction, ini- tiates the deterioration phenomenon. The formation of the sulfite crusts changes the physical properties of the marble in such a fashion that much of the subsequent decay is due to the impermeability of the crust [4] rather than due to SO2 reactions. Figure 3 illustrates the more severe damage from exfoliation of crusts. ■'M / FIG. 3. Severe Damage from Exfoliation of Crusts (Cave Hill Cemetery, Louisville, KY) 91 REFERENCES 1. United States Environmental Protection Agency, "Acid Rain," E.P.A.- 600/9-79/036, July 1980, 36 p. For copy write to: Publications Center for Environmental Research Information, USEPA Cincinnati, Ohio 45268, Phone (513) 684-7562. 2. Likens, G. E., "Acid Precipitation," Chemical and Engineering News , 54, 1976, p. 29. 3. Gauri, K. L. and Rao, M.V.A. , "Certain Epoxies, Fluorocarbon- acrylics, and Silicons as Stone Preservatives, " Geol. Soc. Am. Eng. Geol. Case Hist . 11, 1978, p. 73-79. 4. Gauri, K. L., Gwinn, J. A. and Popli, R. K., "Performance Criteria for Stone Treatment, " Proceedings 2nd International Symposium on the Deterioration of Building Stones, National Technical Univ., Athens, Greece, Sept. 27-0ct. 1, 1976, p. 143-152. 5. Doderer, G. C, "X-ray Diffraction and Fluorescence in Research on Protecting Marble Against Attack by Sulfur Dioxide," M.S. (Chem- istry) thesis, University of Louisville, 1973, 157 p. 6. Urone, P., "Atmospheric Chemistry of Sulfur-containing Pollutants" in Sulfur in the Environment: Nriagu, T. 0., Ed., Part 1, 1978, p. 297-321. 7. Gauri, K. L., "Pollutant Effects on Stone Monuments," Env. Sci. & Technology, v. 15, April 1981, p. 386-390. 92 MEASUREMENTS OF WET AND DRY DEPOSITION ON MARBLE Richard Livingston Office of International Activities U.S. Environmental Protection Agency A- 106 401 M Street, S.W. Washington, D. C. 20460 Marcus Kantz Surveillance and Analysis, Region II U.S. Environmental Protection Agency Woodbridge Avenue Edison, New Jersey 08817 Paul Brown Surveillance and Analysis, Region II U.S. Environmental Protection Agency Jill Dorsheimer Surveillance and Analysis, Region II U. S. Environmental Protection Agency Introduction Photographic records of calcareous stone sculptures in a number of sites around the world suggest that the rate of weathering of stone has accelerated significantly during the last 100 years. Several air pol- lutants including acid rain have been suggested as agents for this accelerated loss. However, no quantitative estimates have been avail- able regarding the rate of loss or the relative contributions of various air pollutants to this rate. Therefore, an experiment was set up involving a marble statue on Bowling Green Customs House in New York City to measure the material carried off in rain water runoff. Method The statue chosen was one of several on the upper story of the Customs House. The Tennessee Marble statue, which was sculpted in 1905 and put in place in 1907, is of a standing viking maiden, approx- imately one and one-half times lifesize. The statue is one of twelve located on the north facade of the Customs House at the sixth floor level. This statue was chosen over the others at the site because it was free standing and did not have any metal fittings. 93 To collect the runoff the statue was fitted around the base with plastic guttering which was connected by plastic tubing to two 19-liter plastic containers. The rainfall itself was collected in 5 gallon plastic containers fitted with funnels. The two sets of samples were analyzed for pH, calcium, sodium, sulfate, nitrate, chlorides and ammonia. This enabled a rough mass balance to be made concerning the amount of calcium dissolved from the marble during rainstorms. A series of washdowns of the statue was also performed using distilled water as a control, in order to estimate the effect of dry deposition. Several successive washings with distilled water were carried out in order to investigate the leaching rate of calcium sulfate from the statue's surface. Results Significantly more sulfates and nitrates, as well as all other ions, were measured in the runoff over the incoming rain, which suggests that current wet deposition is relatively a minor contribution to the material loss rate. In combination with data from sulfation plates and acid gas plates and rainfall information, this runoff data indicates that dry deposition contributes most of the chemical deterioration of the stone. The rain serves primarily to wash the reacted materials from the surface. The role of nitrates in the material loss appears to be minor compared to the sulfates. 94 EROSION OF MARBLE Fred H. Haynie Environmental Sciences Research Laboratory, EPA Research Triangle Park, North Carolina 27711 ABSTRACT : The results of 30 months of environmental exposure of marble samples at nine monitoring sites in St. Louis, MO., have been evaluated. Climatic and air pollution data were subjected to a rigorous evaluation to eliminate extreme recording errors and to estimate missing values. The marble erosion data were evaluated with respect to the possible effects of time-of-wetness, wind speed, temperature, and levels of total gaseous sulfur compounds and total gaseous oxides of nitrogen. The only factors found to be significant were time-of-wetness and the flux to the surface of gaseous sulfur compounds. It was concluded that gaseous sulfur compounds stoichiometrically react with the marble during periods of wetness. An empirical equation was developed that accounts for 98.7% of the non-error variability. KEY WORDS : Marble erosion; air pollution effects, time-of-wetness. INTRODUCTION : From the fall of 1974 to the spring of 1977, EPA conducted an air pollution modeling study in St. Louis, MO. A sophisticated-, air monitoring system was established to provide data for this study* 1 / Data from continuous monitors at 25 sites were checked and recorded by a central computer. Nine of the 25 sites were selected for a three year materials exposure study(2). Among the materials were specimens of white Georgia marble. This paper describes the procedures used to analyze the environmental data and to relate marble erosion to those results. MEASUREMENT OF MARBLE EROSION : Sets of triplicate specimens were exposed for various periods at the nine sites. Exposures were initiated on 10/15/74. With a few exceptions, sets of triplicate specimens were re- moved on 1/15/75, 4/15/75, 10/15/75, 10/15/76, and 4/15/77. Table 1 gives the weight loss as thickness loss from one side in micrometers. An analysis of variance of these data is given in Table 2. The error vari- ance on the means is calculated from the standard deviations. Both the time and site effects reflect differences in environmental exposure. ANALYSIS OF ENVIRONMENTAL DATA : Four measured climate factors and two pollutant factors are considered to have a possible effect on the erosion of marble. The factors are temperature, dew point, wind speed, wind direction, total gaseous sulfur compounds, and total gaseous oxides of nitrogen. Temperature : Some data were missing and values were not obtained when the system was not operating. Two procedures were used to calculate missing values. A system hourly average of available data and individual differences between site hourly averages and the system hourly average were calculated. The average of these differences for each site was obtained for each three months period. This difference was applied to 95 to system hourly averages to get missing site values. During periods when the system was not operating, differences were applied to Lambert Field values rather than system hourly averages. Wind Speed : Procedures similar to those used to obtain missing tempera- ture were used to get wind speed values, however, ratios rather than differences were used. Dew Point : Dew point was the most important but least accurate measure- ment. A system hourly average was calculated excluding the highest and three lowest values. A best estimate of the hourly average dew point for a specific site was calculated using the relationship: DP E = DP A + (T - T A )/3 (1) where DP E and DP. are the site estimated and system average dew points respectively, T and T« are site and system average temperatures respec- tively. During periods when the system was not operating, DP„ and T« are replaced by Lambert Field values. The estimated value was used when values were missing or when the difference between the estimated and recorded value exceeded 2°C. Hourly average relative humidities were calculated from the data using: RH = 100 exp -6.033(T - DP)/(74.73 + T) (2) Where RH is % relative humidity and T and DP are temperature and dew point in °C respectively. Wind Direction : All pairs of hourly average wind direction were checked for differences of less than 40 . If so, a system average and differences from system averages for each site were calculated. When a site differed by more than 40 from five or more sites, the value was checked against previous and following values at the same site. If it differed by more than 40°from the previous hour but not from the following hour, a front- al passage may have occured and the value was assumed to be correct. However, it was not included in calculating the system average. When it differed too much from both values, it was assumed to be erroneous and was treated as missing data. Vaues for missing data were obtained by applying site average differences for a quarter to hourly system averages (or Lambert Field data when the system was not operating). Gaseous Pollutant Data : The recorded data contained many errors and missing values, especially for 1975 prior to an audit (3) .Regression model were developed for each site from 1976 data. Data were divided into three month periods and quadrants of wind direction. The logarithm of hourly average pollution level exceeding 0.005 ppm was step-wise regressed against wind speed, wind direction and wind direction squared, temperature, and a sine function of the time of day. Only statistically significant coefficients were included in the resulting models. The reasonableness of the models is demonstrated in Figure 1 which shows S0o concentration as a function of wind direction. Also shown are the locations of major sources. The lobes on the plots tend to point toward these sources. The models were used to calculate missing hourly averages and replace values that differed from the model values by more than a factor of four. 96 ENVIRONMENTAL DATA REDUCTION : It was necessary to reduce the more than 150,00 hourly averages to a more usable form before determining which factors produce significant effects. These values were reduced to time- of wetness, a temperature factor, and the stoichiometric effects of pollutant flux accumulations. Time-of-wetness : For this study, the surface was considered to be wet when the relative humidity was 90% or greater and the temperature was greater than - 3°C , or during rain periods. The number of hours these criteria were met at each site were totaled for each exposure period. These data for ninety quarters were regressed against quartly average values for relative humidity and temperature to yield: f tw = exp - 0.0196 - 0.856/(T + 3) - 3.65(100 - RH)/RH (3) where f. is the fraction of time when wet,T is the average temperature in C, and RH is the average relative humidity. Since averages are more readily available than distributions, time-of-wetness was also calculated using equation (3). Results from both techniques are given in Table 3. Temperature : To be consistant with a possible Arrhenius type effect, an inverse of the average absolute temperature factor was calculated for each period of exposure. Table 4 gives these values for when the surface was wet. Pollutants : The effect of pollutants on marble is expected to be a function of the amount reaching the surface. The amount deposited per unit time is by definition the product of the ambient concentration and deposition velocity. From theory (4) the deposition velocity can be estimated from wind speed. Hourly deposition velocities were calculated for each site. Hourly pollutant accumulations were calculated and totaled for each exposure period. Estimates of accumulated amounts were also calculated from average wind speeds and concentrations for the exposure periods. These values were converted to stoichiometrically equivalent amounts of marble surface lost. The results are presented in Tables 5 and 6. It is possible for NOo to react with marble when the amount of N0p is stoichiometrically in excess of the amount of SO?. A comparison of Tables 5 and 6 indicates this possibility for some sites and exposure periods. RELATING MARBLE EROSION TO ENVIRONMENTAL FACTORS : Table 7 gives simple correlation coefficients for marble erosion with different factors. The best fit is with time-of-wetness calculated from exposure period averages of temperature and relative humidity. A multiple regression of the logarithm of marble erosion on the logarithm of the time-of-wetness, the temperature factor, and the stoichiometric pollutant effects was used to determine their statistical significance. The results indicated that neither temperature nor N0 2 were significant factors. Non-linearity with time-of-wetness and S0~ , however, were significant. As a result the data were least squares fitted to the following equation: E m = A + Bt * 89 + C S0 9 (4) m w 2 v ' where E m is marble erosion in micrometers, tw is time-of-wetness in years 97 calculated from temperature and relative humidity averages for the exposure periods, SCL is the stoichiometric marble damage in micrometers calculated from average wind speed and S0 2 concentration for the exposure period, and A, B, and C are the resulting coefficients. The results are given in Table 8. Figure 2 compares the regression results with the observed values. More than a dozen other regression models using a com- bination of two or more environmental factors were evaluated, but none produced as good a fit as equation (4). The probability that the lack of fit is caused by other than random error is only 0.8. ESTIMATING MARBLE EROSION RATES IN PARTICULAR ENVIRONMENTS : The erosion rate of this type of marble can be expressed as: R E = (22.1 + 0.20 u*S0 2 ) f t (5) where Rr is marble erosion rate in micrometers per year, u is deposition velocity in cm/sec, S0 2 is total gaseous sulfur concentration expressed as SOo in micrograms per cubic meter, and f. is fraction of time when wet calculated using equation (3). Standard deviations on the coeffici- entsare + 1.4 and +0.03 respectively. Deposition velocity is a function of surface geometry as well as wind speed. At long term average speeds it should be around 0.7 cm/sec for large flat surfaces. Figure 3 demonstrates the synergistic effect of relative humidity with S0 2 concentration as well as the direct effect of S0 2 at an average temperature of 10°C with a range of plus or minus one standard deviation. At any level of relative humidity the erosion rate of marble exposed to 100 >jg/m of S0 2 is nearly double that in a relatively clean environment. The magnitude of this effect, however, is much greater at high relative humidities. The fact that these results tend to confirm the theoritical expec- tations that S0 2 stoichiometrically reacts with marble when the surface is wet, suggests that damage from acid rain can be estimated from theoretical considerations. Assuming all the acidity in rain is caused by sulfur compounds, the following equation can be used: R = 171 r exp - 2.3 pH (6) where R is the rate of marble erosion injum/year, r is annual rainfall in cm, and pH is the annual average pH of rain. A decrease of one pH unit increases the erosion rate by an order of magnitude. CONCLUSIONS : 1) Marble stoichiometrically reacts with gaseous sulfur compounds when the surface is wet; 2)the effects of relative humidity and S0 2 are synergistic with the greatest effect occuring at high relative humidities; and 3) Theoretical expectations are confirmed, thus, they can be extended with some degree of confidence to estimate the damaging effects of acid rain. 98 REFERENCES 1. Hern, D. H. and Taterka, M. H., "Refional Air Monitoring System Flow and Procedures Manual", Report for EPA Contract 68-02-2093 by Rock- well International, Creve Coeur, Missouri, August 1977. 2. Mansfeld, F., "Regional Air Pollution Study - Effects of Airborne Sulfur Pollutiants of Materials", EPA-600/4-80-007, January 1980. 3. Smith, F., "Audit and Study of the RAMS/RAPS Programs and Recommend- ations for a Quality Assurance Plan for RAPS," Report for EPA Con- tract 68-02-1772 , by Research. Triangle. Institute:,. Research: Triangle Park, N. C, June 1976. 4. Haynie, F. H., "Theoretical Air Pollution and Climate Effects on Materials Confirmed by Zinc Corrosion Data," Durability of Building Materials and Components, ASTM STP 691, American Society for Testing and Materials, Philadelphia, 1980, pp. 157-175. 99 Table 1. Marble Erosion as a Function of Exposure Time at Nine Sites in the St. Louis, Missouri Area * Mean Erosion and Estimated Standard Deviation - urn Site Number 0.252 years 0.499 years 1.0 years 2.003 years 2.501 years 103 105 106 108 112 115 118 120 122 2.686 4 .279 3.492 ± .669 3.126 4 .205 3.614 ± .215 3.328 ± .351 3.780 ± .052 3.205 4 .613 3.199 2 4 .013 4.629 4 .289 4.574 ± .439 5.571 ± .407 5.279 i .105 6.065 4 1.002 5.383 4 .511 6.528 4 .671 6.167 4 .690 3.164 4 .178 6.473 4 .535 9.757 14.863 2 4 .512 16.638 4 .683 10.239 ± .679 9.426 4 .794 11.770 4 .397. 15.967 4 2.106 14.032 2 4 1.752 9.951 2 4 .222 15.266 4 .278 10.421 4 .596 10.242 4 .507 15.774 16.315 4 1.784 9.912 2 4 .650 15.232 2 4 1.779 12.133 4 .245 17.038 2 4 .954 17.032 4 1.298 18.230 4 1.462 16.082 4 4 1.169 20.110 4 .354 17.024 4 4 2.160 18.581 4 .936 18.818 4 4 1.698 18.449 4 4 .679 19.932 4 2.053 * Subscript on mean indicates the number of specimens In the sample 1f other than three. Table 2. Analysis of Variance of Means of Marble Erosion Variable Sum of Squares Degrees of Freedom Mean Square Time 1423.1189 Site 24.4841 Time x Site 9.1068 Total 1456.7098 Error 14.0164 4 8 32 44 43 355.7797 3.0605 0.2846 0.3260 1091.483 9.389' 0.873 •Greater than 0.9995 probability that variable effect is statistically significant. 100 Table 3. Time-of-Wetness for Five Exposure Periods at Nine Sites in the St. Louis, Missouri Area. * Time-of-wetness - years Site No. 0.252 years 0.499 years 1.0 years 2.003 years 2.501 years 103 0.0822 (0.0824) 0.1463 (0.1412) 0.2913 (0.2916) 0.4683 (0.4547) 0.5350 (0.5152) 105 0.0733 (0.0696) 0.1317 (0.1238) 0.2369 (0.2482) 0.3656 (0.3826) 0.4175 (0.4252) 106 0.0733 (0.0716) 0.1323 (0.1252) 0.2467 (0.2548) 0.3879 (0.3946) 0.4513 (0.4527) 108 0.0882 (0.0885) 0.1564 (0.1497) 0.3235 (0.3154) 0.5286 (0.4947) 0.6030 (0.5702) 112 0.0766 (0.0753) 0.1346 (0.1282) 0.2563 (0.2614) 0.4114 (0.4146) 0.4688 (0.4627) 115 0.0879 (0.0890) 0.1621 (0.1557) 0.3145 (0.3120) 0.4992 (0.4847) 0.5688 (0.5652) 118 0.0866 (0.0847) 0.1531 (0.1442) 0.3056 (0.3006) 0.4961 (0.4767) 0. 5664 (0.5477) 120 0.0847 (0.0852) 0.1454 (0.1447) 0.2863 (0.2908) 0.4662 (0.4587) 0.5333 (0.5277) 122 0.0955 (0.0975). 0.1696 (0.1662) 0.3515 (0.3487) 0.5807 (0.5608) 0.6501 (0.6503) ♦Values in ( ) were calculated from t< anperature and relative hurrn dity averages for periods. Others are the accumulated hours meeting time-of-wetness criteria. Table 4. Temperature Factor As a Function of Site and Exposure Period When Wet Temperature Factor - • 1000/ OK Site 0.252 years 0.499 years 1.0 years 2 003 years 2.501 years 103 3.587 3.601 3.515 3.517 3.575 105 3.581 3.596 3.513 3.511 3.518 106 3.581 3.596 3.515 3.515 3.524 108 3.591 3.604 3.520 3.519 3.528 112 3.582 3.596 3.513 3.510 3.518 115 3.589 3.604 3.521 3.519 3.528 118 3.589 3.603 3.516 3.515 3.524 120 3.589 3.604 3.517 3.518 3.527 122 3.600 3.614 3.526 3.529 3.536 101 Table 5. Marble Erosion Calculated from Stiochiometric Reaction with Deposited SO* at Nine Sites in the St. Louis, Missouri Area Marble * Erosion - pm Site No. 0.252 years 0.499 years 1.0 years 2.003 years 2.501 years 103 1.765 (1.875) 3.413 (3.703) 5.712 (6.077) 12.143 (12.811) 16.067 (17.088) 105 1.837 (2.025) 4.018 (4.387) 6.510 (7.157) 13.292 (14.560) 17.527 (19.210) 106 1.920 (2.176) 4.013 (4.496) 6.538 (7.295) 13.470 (14.781) 17.455 (19.404) 108 2.658 (2.780) 4.423 (4.737) 6.627 (6.990) 15.296 (15.772) 21.057 (21.774) 112 1.775 (1.873) 3.887 (4.173) 6.838 (7.208) 14.437 (15.103) 18.096 (19.180) 115 1.593 (1.670) 4.573 (4.693) 7.097 (8.284) 14.730 (15.319) 19.264 (19.948) 118 1.365 (1.485) 2.542 (2.783) 4.323 (4.621) 8.528 (9.099) 11.304 (12.183) 120 2.214 (2.314) 3.435 (3.753) 5.372 (5.521) 11.089 (11.613) 15.157 (15.954) 122 2.498 (2.466) 4.612 (4.660) 8.203 (8.401) 16.284 (16.652) 19.819 (20.538) ♦Values in ( ) were calculated using average wind speed and S02 concentration for the exposure period. The remaining values are from summations of hourly amounts deposited. Table 6. Marble Erosion Calculated from Stiochiometric Reaction with Deposited NO2 at Nine Sites in the St. Louis, Missouri Area * Marble Erosion - ym Site No. 0.252 years 0.499 years 1.0 years 2.003 years 3.501 years 103 2.505 4.596 7.836 16.306 21 .463 105 2.794 5.513 8.818 17.999 23.796 106 2.683 4.212 7.472 15.966 21.454 108 1.915 3.444 5.497 11.606 15.382 112 2.231 4.373 7.409 15.583 20.033 115 1.034 1.683 2.724 5.592 7.693 118 0.727 1.591 2.520 5.089 6.832 120 1.767 3.235 5.174 5.507 9.075 122 0.645 1.134 2.044 4.497 5.663 102 Table 7. Correlation Coefficients for Marble Erosion with Different Environmental Factors Factor Correlation Coefficient Total time Accumulated time-of -wetness Averages calculated time-of -wetness Accumulated stioc hi oneiric SO2 damage Averages calculated SO2 damage Accumulated stiochiometric NO? damage Accumulated Excess stiochiometric S0 o and N0« damage 0.9799 0.9795 0.9845 0.9447 0.9554 0.6833 0.9150 Table 8. Regression Analysis of Marble Erosion Data Coefficient Value Standard Deviation Partial residual F with respect to replicate error A 0.7838 t .2214 12.53 16.88 B 22.098 i 1.405 247.37 333.27 C 0.2776 ± 0.0398 48.65 65.54 Residual ms Error ms 0.4392 0.3260 correlation coefficient for regression 0.9936 F for lack of fit 1.347 103 o s CD I o <_> S UJ cc X 3 C ID o > ««- u m at 01 Ol ■— o XI c t. 0) g* ft o > u k ft »« «/> in -o en o ^1 o « o o """ - wusoaa 3iaavw 03AU3sao 01 •.- o> k IO 3 k O Ol IS) > «« «/l T- 0> 3 £3 CM O 01 «/> .C *» c c a ■«- k 01 i- c ■Oi- c ?♦> t- ID * to <♦- r- O o> •4J u c ft f- H- <»- u» ft c o i!s 1- • u «-» c • ft -2 «l o k o 0» CM s ■r- O k U. IS) « 104 20 40 60 80 100 120 140 160 180 200 TOTAL GASEOUS SULFUR AS S0 2 - yg/m 3 Figure 3 . Effect of average SO2 concentration and average relative humidity on marble erosion rate at an average temperature of 10°G with a one standard deviation from the mean range 105 ARCHITECTURAL TERRA COTTA AND CERAMIC VENEER: PROBLEMS IN DURABILITY AND REPAIR Theodore H. M. Prudon and Jerry Stockbridge The Ehrenkrantz Group New York, New York Abstract: The use of architectural terra cotta and ceramic veneer in the United States is based on a long history of its application in Europe. Major nineteenth century architects such as James Renwick or Richard Upjohn were influential in the introduction of the material in this country. Handmade architectural terra cotta units, glazed and unglazed, enjoyed considerable popularity from 1880 until World War II. Beginning in the 1930' s, and particularly after World War II, the cost of manufacturing handmade terra cotta units became prohibitive, and machine-made ceramic veneer became more common. These machine-made units, which are less ornamental and less detailed than the handmade units, remained in use, although somewhat limited, until the 1960's. Architectural terra cotta and ceramic veneer have now reached a critical point in their life spans. Because very little scientific research on performance standards or manufacturing control was taken before World War II, many of the manufacturing processes, details, and design toler- ances were developed through experimentation. As a result, failures have developed in terra cotta or ceramic veneer-clad buildings; intrin- sic flaws in the material itself, inadequate design and detailing, maintenance and repair, have contributed to this deterioration. Since many of these terra cotta buildings are visually quite attractive, and more importantly because they are located in major urban areas where concern for public safety has increased in the last couple of years, along with the demand for the adaptive use of existing building stock, several major buildings have formed the basis for extensive study and repair work. To determine where failures have occurred in architectural terra cotta or ceramic veneer-clad structures, a thorough inspection and evaluation of existing conditions is often required. In examining, it is to be determined whether the failures are caused by the material itself or as a result of the design and installation. Often failures of terra cotta- clad structures are caused or accelerated by improper detailing. Struc- tural integrity of the units, glaze fit, and condition of the glaze must also be evaluated. It is crucial that one determine whether a substantial stress has built up in the cladding. Surprisingly, it was discovered in the examination of older structures that substantial pressures far exceeding what was to be considered normal, or safe, for that matter, had built up. Exten- sive strain relief testing was used to determine its magnitude. This experience appeared to be applicable to a great many structures built 106 in the first quarter of this century. In addition, frequently no provisions were made for thermal expansion and contractions occurred. The research and information available to date is limited and remains to be expanded. However, what has started as an evaluation of early terra cotta — a substantial problem in itself, considering its widespread use — has proved to provide an insight into the performance of other clay materials such as brick cladding. More research in Canada, and more recently in England, had indicated that some of this experience is applicable to more contemporary construction. More importantly, it may assist in a better understanding of clay materials and help to improve detailing. 107 DEGRADATION AND REHABILITATION OF TERRA COTTA Sven E. Thomasen Wiss, Janney, Elstner and Associates, Inc. 1900 Powell Street Emeryville, CA 94608 Abstract: Architectural terra cotta cladding was used extensively be- fore the 1930 f s. With change in building technology, the usage declined drastically, but a large number of outstanding examples of ornate terra cotta structures are still in service. Terra cotta has been used in these buildings sometimes as a load-bearing masonry, but mostly as cladding over a framing system. Generally, no provisions were made for movement, either absolute nor differential, in the backup framing or in the cladding. Moisture expansion of the clay body and thermal fluctua- tions caused cracks to develop, allowing moisture to enter the wall, thereby accelerating the weathering and deterioration. This study deals with the distress, deformations and the strain accumulation in the terra cotta cladding caused by weathering and by contraction and expan- sion of the material. Field investigations of terra cotta failure and laboratory tests are discussed and examples of degradation of the material and the different modes of failure are shown. The cause of deterioration is analyzed and recommendations for rehabilitation introduced. Key words: Building cladding deterioration; field investigation; laboratory analysis of terra cotta; stress measurements; terra cotta distress; terra cotta rehabilitation. Introduction: Terra cotta is generally durable and permanent because of the excellent weathering properties and the hard surface of the glaze. But terra cotta cladding material on many highrise buildings has deteriorated, mostly because of disregard and lack of understanding of the behavior of the material. Most terra cotta failures are interrelated in that deterioration in one area will cause failure in another area of the facade. The main causes for failure are the results of material expansion, water intrusion and inappropriate design and installation. Terra cotta details: The typical terra coLi:a for building cladding is a clay block with web stiffeners in the back and with glazing at the front surface. The single firing was to cone 4 to 5 (about 2060°F) , resulting in good hardness of the glaze and high strength of the clay 108 body. The cladding was erected with mortar joints using a portland cement mortar with lime added. Joints were narrow, often about 3/16 inch. Very few terra cotta claddings had expansion joints or internal flashings and weepholes. The vertical support of plain ashlar units were by shelf angles at the floor levels, while horizontal support was provided by Z-shaped steel straps anchored into a slot in each terra cotta block. The back-up material was masonry or concrete walls. Ornamental units generally had multiple anchors. Terra cotta failures: Many different kinds of distress can often be observed at the same building facade (Fig. 1). The more common failures are: Glazing failures - Environmental exposure and thermal expansion and contraction can result in glaze crazing if the temperature or the co- efficient of expansion of the glaze and the clay body is incompatible. Crazing allows water to enter the clay body, causing pinhole spalling of glaze when the water in the clay pores freezes. More water can now enter the block resulting in general glaze splitting and the loss of the entire glazed surface. Expansion failures - Long-term shrinkage can occur in concrete frames, but most displacement in terra cotta cladding comes from moisture and temperature expansion of the clay body. The thermal and the wet/drying cycles are often associated with permanent lengthening of the terra cotta blocks, and this plus the absence of expansion joints in the facade, creates compressive stresses. Failure can be in the form of buckling of block units (Fig. 2), or crushing at the base of the facade where expansion stresses are combined with compressive gravity loads (Fig. 3) . Horizontal expansion and vertical stress also combine to cause vertical cracks at corners and at wall intersections. Moisture related failures - Deterioration of the terra cotta glaze and cracking caused by expansion allows water to enter the facade. This results in further distress from freezing/ thawing and from corrosion of metal anchoring. Rust scale expansion at the toe of the shelf angle will push the block out and create failure at the critical support (Fig. 4). Field investigations: The extent and the seriousness of the deteriora- tion can best be evaluated at the site. The first step is a visual examination of the terra cotta surface. This can be by binoculars from the ground or adjacent buildings, or it can be from suspended scaffolds. Tapping with a wooden mallet can detect internal cracking. A terra cotta block with internal cracks sounds different from a sound block when hit with a mallet. Pachometers can be used to detect embedded steel members and wall ties. The metal detector can verify the location of shelf angles and struc- tural steel supports. 109 Inspection openings can be cut in the wall. This is the best method to verify the location and the condition of the wall supports, and especially to detect corrosion of the embedded metal anchors. Soniscopes have been used to locate cracks in the terra cotta, but the many voids in blocks often create strange patterns, making soniscopes a less useful tool. Stress measurements are one of the best tools to measure distress in the cladding. Electrical resistance strain gages are attached to the terra cotta surface and the gages are read. Then the terra cotta block, with the gages attached, is cut loose from the wall and the gages are read again. The locked-in stress in the terra cotta can now be found by multiplying the measured strain difference by the terra cotta modulus of elasticity. Strain measurements at several points of the facade will establish a stress map for the structure (Fig. 5). Strain measurements can also evaluate if cutting of expansion joints into the facade will relieve the built-up terra cotta stresses. Laboratory analysis: Once it has been determined a material has de- cayed or lost some of its original qualities, it is often helpful to identify the material properties by laboratory analysis. The commonly performed tests on terra cotta are: Petrographic - The consistency of the glaze and the clay body is evalu- ated through a stereomicroscopic examination. The density, the degree of deterioration, and the composition of the materials can be estab- lished by an experienced petrographer. Compressive strength - Cubes of terra cotta are cut from the facade blocks and tested in compression. The compressive strength of terra cotta is generally higher than the strength of modern clay tiles, re- flecting that better clay and higher firing temperature were used for the terra cotta. Absorption - Ideally, a glazed specimen should produce a zero absorp- tion rate if its glaze is intact, sound, and craze-free. The actual rate of absorption of both glaze and clay body is tested by submerging dry specimens in water for 24 hours. The weight gain of the specimens is a measure of the materials absorption characteristic. Thermal coefficients - Tests are performed on the complete terra cotta block, and on separate pieces of glaze or clay body only. Strain gages are mounted on selected samples which are then subjected to cyclic temperature changes, say from 30°F to 150°F. The temperature should represent the normal wall exposure since the thermal coefficients tend to be non-linear outside the temperature range. The tests establish the normal thermal coefficient as well as any permanent distortion at the conclusion of the thermal cycles. Clay materials sometime expand more than they contract, resulting in permanent elongation. The separate thermal coefficients for the glaze and the clay body are eval- uated for non-compatibility which could result in crazing and distress 110 in the glaze. Terra cotta rehabilitation: The restoration of defective terra cotta is rarely an inexpensive process. The damaged pieces will have to be re- placed and the sources of distress eliminated. The facade will have to be made watertight, deteriorated anchors and support replaced, and the built-in stresses in the terra cotta will have to be relieved by the cutting of expansion joints in the facade. Replacement of damaged pieces with new terra cotta is possible, but often expensive. Plain ashlar blocks are available, but more decorative pieces require field forming and custom casting. Precast concrete, sometimes reinforced with fiberglass, has been used as a less expensive replacement material. It can be cast to reproduce the terra cotta details, the concrete color can be matched to the existing material and surface coatings will duplicate the gloss of the terra cotta glazing. Spalling of ornamental terra cotta can be patched with special compounds such as polymer concrete. The patch can be coated to match the surrounding glazing. Ill CRAZED AND SPALLED GLAGE BUCKLED BLOCKS SPALLING AT SHELF ANGLE DIAGONAL CRACKS AT CORNER BUCKLED BLOCKS AT SHELF ANGLE ATLANTA CITY HALL - FRONT ELEVATION Fig. 1 - Typical Types of Terra-Cotta Failures. 112 •High compressive stress caused by material expansion Back-up block Buckled terra-cotta facade , ■ Section through wall Fig. 2 - Buckled Terra-Cotta Block, High Compressive Stress, Gravity load and high compressive stress in column Cracking in terra-cotta Expansion and cracking of window sill Fig. 3 - Damaged Sills at Base of Tower Pier Rust scale ex- pansion causes spalling of terra-cotta facade -Concrete floor slab Shelf angle -Rust scale from shelf angle Fig. 4 - Spalling of Terra-Cotta Facade from Shelf Angle Toe, 113 .1 • . • ■ "i.I:- : -a'.fl • Floor 12 Floor 11 Floor 10 Floor 9 Floor 8 Floor 7 Floor 6 Floor 5 Floor 4 STRAIN RELIEF TEST DATA Identl- tltitlon Data Verdeal Horizontal Strain u aUro-la./lB. Sttaae pal Strain u ■lcro-ls./in. Screaa pal 1 2 3 A 5 + 245 + 300 +64 5 +235 620 747 1606 585 -30 ♦10 +60 -75 25 149 Scraaa la baaed on Hodulua of ElaaclcltT E-2.49 > 10* pal. Vertical C.je HorliontdU C.»g« V^V" SI iliii Fig. 5: Strain Relief Test 114 DURABILITY OF GYPSUM BOARD William C. Veschuroff and Lawrence T. Eby United States Gypsum Company Des Plaines, Illinois 60016 Abstract : Gypsum board as a commercial product is a lami- nate of hydrated calcium sulfate with paper faces. It is a stable and durable material when properly protected from moisture and undue stresses as is evidenced by over fifty years of service in many structures. It is the limitations within which gypsum board must be used to assure its durability that requires careful considera- tion. It is excluded from exterior use and exposure to moisture which may affect the bond of the paper to the gypsum core. Moisture may affect the core if it is under stress; a common example is sag of ceiling boards under high humidity conditions if insufficiently supported. Hygrometric movements are isometric and low compared to wood products. Any long duration of wetness on the surface of the board can cause deterioration of the bond between the paper and the core. Another enemy of gypsum board is exposure to high heat. Under normal conditions this is no problem, but any lengthy exposure to temperatures in excess of 125 °F can cause cal- cining of the gypsum at the bond line and loss of bond be- tween the paper and the core. There are few places for the interior use of gypsum board where this is a problem. On the other hand, gypsum board can withstand excessive heat without burning, providing fire resistance to walls while absorbing heat of dehydration and vaporization of the water of crystallization in the core. Fire resistance is proportional to the thickness of the gypsum board if held in place during a fire. Key Words : Bond; building panel; calcium sulfate; core; durability; dynamic performance; exterior; fire resistance; gypsum; gypsum board; impact; interior; moisture; paper; sag; shear. 115 Gypsum board is a specialized composite building panel which can provide excellent long term durability, providing certain inherent limitations are recognized and observed. The basis of these limitations is susceptibility to sus- tained moisture, sustained load and sustained high tempera- ture. Examples of these performance properties will be ex- amined in order to understand under what conditions gypsum board should be used to provide long term satisfaction. Gypsum board is an envelope of paper filled with the dihy- drate of calcium sulfate. The paper must have special properties of strength to act as a stressed skin or mem- brane, and have the porosity to allow moisture to escape rapidly while drying the board and to allow the gypsum crystals to grow into the fiber matrix for good bond. Ad- ditives may be in the gypsum core for various reasons: foaming agents to make it lighter, glass and paper fibers for reinforcement in a fire, boric acid to prevent sag, asphalt and wax to resist water absorption, starch for im- proved bond to paper, etc [1,2]. It is not intended to discuss the technology of making gypsum board here but only to recognize that there are many minor variations which affect the properties of the generic product and that many types of gypsum board are produced. In all cases, however, the paper envelope provides the stress skin or tension member and the gypsum core provides inert mass with good compression strength. The paper surface must also provide the base for a decorative finish. In some cases a hard thin-coat plaster is applied before painting for improved surface durability. Interior Durability : The normal use of gypsum board is as a non-load bearing building panel which is supported on a frame with nails or screws, with or without adhesive attachment. Adhesive attachment improves shear strength by distributing the shear load over the bond area of the adhesive as compared to point stress concentration at the nail or screw. Al- though gypsum board is not used to sustain structural loads due to long term creep properties [3], it can easily handle normal dynamic loads and exhibits resistance to shear and impact. Ultimate shear strength data are shown in Table I for a typical gypsum wallboard used in mobile homes. Gypsum board may also be mounted in a grid vertically or horizontally, the former by kerfing or in channels and the latter by laying on a ceiling grid. Shear properties are not expected under these conditions; however, sag can be a problem with a lay-in ceiling panel. Unrestrained with no fasteners, sag is controlled by the dimensions of 116 the panel, the weight on the panel (including added insula- tion) and the composition of the core. Even in restrained situations where ceiling board is nailed or screw attached, similar considerations must be made for sag. For example, it is possible that more sag will occur if the board is placed parallel to the joist rather than perpendicular to the joist because the paper skin of the gypsum board is weaker across the board than it is in the machine direction. Recommended application is shown in Tables II, III and IV. Additives, such as, boric acid, can modify the core to make it more resistant to sag. Sag is exaggerated in high moisture conditions. Although hygrometric changes in gypsum board are small, they do occur with slight hysteresis. Both the paper and the core con- tribute to these effects of moisture. Paper fibers swell with moisture and gypsum crystals can rearrange under high moisture conditions. The linear hygrometric coefficient for conditioned gypsum board is 7 X 10" 6 in/in/% r.h. and is reversible in going from 5% to 90% relative humidity. Board is conditioned once it has been exposed to high humidity and original hys- teresis or set has been established. The hygrometric co- efficient may be twice that value for unconditioned gypsum board for the first exposure to high humidity. The linear coefficient of thermal expansion tested between 44 and 87°F is 9 X 10 6 in/in/°F and is taken into consideration also for any movement of the board. Its thermal expansion is close to other materials with which it is used, such as, steel with a thermal coefficient of 7 X 10" 6 in/in/°F, and concrete at 6 X 10"* 6 in/in/°F. Gypsum board is stable under both high and low humidity con- ditions as long as the dew point is not encountered with condensation of water in the board. Where this could occur, special precautions must be taken. Paint with low perme- ability should be used on gypsum board in kitchens, bath- rooms, laundries, etc. where high humidity and cold walls exist. Special protection is given to both the paper and the gypsum core to produce a more water-resistant board to use in these areas, as well as, sheathing applications. A good barrier to water vapor is, of course, aluminum foil which may be laminated to the back or face of the board to prevent migration of water vapor through the board. Under- standing the destructive effects of condensed water in gypsum board and materials associated with it, such as, some insulation products, provides the basis for installa- tion specifications which provide an indefinitely long life. The most vulnerable part of gypsum board to deterioration 117 by moisture is the paper- to-core bond. This can lead to blistering, cracking and removal of the paper. Furthermore, the core may be softened, lose its integrity, and in time erode with an excess of water since gypsum has a solubility of about 0.2%. The strength of the core and its bond to the paper is dependent upon the matrix of gypsum crystals which in itself may be modified or destroyed in the presence of water. This can be observed in micrographic studies. It was mentioned earlier that gypsum board has resistance to dynamic loads. As a composite of a flexible stress skin on a massive core of high compressive strength, it withstands impact to a remarkable degree. This is necessary for any wall panel. Experimentally, this is demonstrated with a standard wall impact test such as, ASTM E-72, which uses a 60-lb bag drop on the wall panel. In most instances, fail- ure was due to the framing member which occurred at 300 ft-lb impact with 3/8~inch gypsum board nailed on both sides to 2" X 4" wood studs and 250 ft-lb for 2" X 3" studs, both 16-inch and 24-inch spacings gave similar results. Failure can occur at half these values with a knot in the stud, showing the dependence of framing strength in this test. In considering the durability of gypsum board, it should be noted that impact or gouging can be easily repaired and the surface refinished without removal of the panel unless, of course, the damage is too extensive. Further demonstration of the dynamic properties of gypsum wallboard is its performance under seismic conditions. Both laboratory experiments and examination of structures that have come through earthquakes show the durability of gypsum board walls and ceilings compared to other building materials. This ability to resist dynamic loading has contributed to the outstanding performance of gypsum board in shaft walls. The gypsum board elevator shaft walls in the New York City World Trade Building were designed after dynamic cycling tests proved the ability of gypsum board to withstand the pressures and movement involved in this type of service. The pressure used for dynamic testing was 7.5 lb/ft 2 (36.6 kg/m 2 ) cycled over 1,000,000 times with maximum deflection in the shaft wall of ± 0.2 inch (5 mm), the 2-ft wide section being composed of a 1-inch liner-board and two 0.5-inch face layers. Exterior Durability : Numerous attempts have been made to use various types of gypsum board on the exterior of buildings with discouraging and disasterous results. On the other hand, there are a 118 few selected applications where it is practical and success- ful. The most successful has been exterior ceiling and soffit board made to certain specifications and properly in- stalled. The board has a special core and special paper. It is also protected from driving rain, snow and any form of water infiltration. Probably without exception, the greatest enemy of gypsum board in exterior walls is liquid water from rain, snow or condensation and the failure mechanism is destruction of the bond of paper to gypsum core. There are many examples of this where trial installations have been made. Many of these appear entirely satisfactory for several years followed by rapid deterioration. This occurs even with various types of exterior coatings to protect the paper. There are such wide ranges of temperature and humidity con- ditions encountered in exterior exposure, that there appears to be hardly any geographical area that would guarantee trouble-free performance. Gypsum board performs well as a sheathing product because of relatively good racking resistance as well as fire re- sistance. This may be classed as semi-exterior board since it involves wide temperature changes; however, there is no exterior exposure of its surface and its performance is not dependent upon the integrity of the paper bond to core to the same extent as a true exterior board. It is possible to surpass the HUD-FHA requirements for sheathing both dry and wet with 0.4-inch special gypsum board adhe- sively attached; however, the commercial gypsum sheathing is usually 0.5-inch board nailed without adhesive which meets about 95% of the HUD-FHA criteria for maximum load per ASTM E-72. Fire Resistance ; Not only does gypsum board present no hazard from starting or propagating a fire but it also acts as a barrier to fire by virtue of its chemical composition. Even though it has a combustible surface, it takes a great amount of heat or fire-load to burn the paper off the gypsum core. The core itself is endothermic, requiring heat to liberate the water of hydration from the gypsum. Consequently, its thermal barrier properties are proportional to the thickness of the gypsum board. This is illustrated in Figure 1. Under fire condition, the total thermal transmission through the board is primarily dependent upon calcination or removal of the water of crystallization, which requires much more heat than the normal heat capacity of the board, before a com- bustible temperature is reached, on the opposite side of the wall. 119 Fire resistance is measured by standardized tests, such as, ASTM E-119. Tests of this type must be used to predict actual performance because other factors are also important. All gypsum boards do not perform in the same way when heat- ed through the calcination stage. As the dihydrate calcines to the hemihydrate, shrinkage and cracking takes place in the core. Additives must be present to prevent the board from falling apart while the calcination is taking place. Two types of additives are used, glass and paper fibers for reinforcement and expandable materials like unexpanded vermiculite to fill the voids as the gypsum shrinks. Bal- anced formulations for the core are essential to produce maximum fire resistance [2], Conclusion ; Gypsum board is a proven durable building panel providing the proper product is used according to recommended practices. Liquid water is its greatest enemy and can destroy its paper to gypsum bond. Special board can be made for exterior ceilings and soffits but other exterior applications are prone to failure. Enhanced performance in fire-rated walls and ceilings is accomplished by special additives in the core. Creep properties prevent it from use as a structural material for long term sustained loads but its performance in dynamic shear, impact and com- pression is adequate. Various types of gypsum board are made to meet different types of environment and should be specified accordingly for long term durability. REFERENCES 1. Summerfield, J. M. , "Gypsum", Chemical and Process Technology Encyclopedia , McGraw-Hill Book Company, 1974, pp. 566-570. 2. Willey, Grant S., "Fire Resistant Plaster Product", U. S. Patent 3,454,456, July 8, 1969. 3. Sattler, H., "Elastic and Plastic Deformations of Plaster Units Under Compressive Stress", Materiaux Et Constructions, Vol. 7, No. 39, 1974, pp. 159-168. 120 Table I Shear Resistance 5/16" Gypsum Panels Attached to Wood Framing Application - Mobile Homes Ceilings - Point attachment 2" x 2" trusses-rosettes and staples Walls - Panels one side 2" x 3" studs-white glue and staples 2" x 3" studs-construction adhesive ** Walls - Panels both sides studs-white glue and staples studs-construction adhesive studs-white glue and staples studs-construction adhesive studs-white glue one side and construction adhesive other side 1" x 3" 1" X 3" 2" X 3" 2" X 3" 2" X 3" Ultimate Shear Strength lb/ft 433 kg/m 644 619 921 619 921 810 1205 735 1094 880 1310 880 1310 880 1310 * Polyvinyl acetate emulsion adhesive ** Solvent-based rubber adhesive 121 Table II Maximum Frame Spacing for Gypsum Panels Application Application to Framing Perpendicular Gypsum Board Thickness inch (mm) Maximum ing of inches 16 o.c. Spac- Framing, (mm) Ceilings 3/8 (9.5) (406) Ceilings Perpendicular 3/8 (9.5) 16 (406) Ceilings Perpendicular 1/2 (12.7) 16 (406) Ceilings Parallel or Perpendicular 5/8 (15.9) 24 (610) Sidewalls Parallel or Perpendicular 3/8 (9.5) 16 (406) Sidewalls Parallel or Perpendicular 1/2 (12.7) 24 (610) Sidewalls Parallel or Perpendicular 5/8 (15.9) 24 (610) 122 Table III Maximum Frame Spacing for Texturing Gypsum Ceiling Panels Gypsum Board Maximum o.c. Spacing Application Thickness of Framing, to Framing inch (mm) inches (mm) Not recommended 3/8 (9.5) Perpendicular only 1/2 (12.7) 16 (406) Perpendicular only 5/8 (15.9) 24 (610) * Heavy textures contain as much as 20 gallons of water per 100 pounds of texture. 123 Table IV Maximum Weight on Back of Gypsum Board Ceiling Panels to Prevent Noticeable Sag (L/100) in 45 days at 90°F (32°C) and 90% Relative Humidity Gypsum Board Thickness inch (mm) Frame o.c. inches Spacing, (mm) Maximum Weight lb/ft Z (kg/m) 2 1/2 (12.7) 24 (406) 1.3 (6.3) 1/2 (12.7) 16 (610) 2.2 (10.7) 5/8 (15.9) 24 (610) 2.2 (10.7) 124 Figure 1 Fire Resistance Finish Rating of Gypsum Board Type X Finish Rating is the time required for a wood framing mem- ber to reach to a temperature of 250°F. average or 325°F at any one location above the ambient temperature on the surface nearest the fire when the test sample is subjected to ASTM E-119 time- temperature conditions. 0) •p C •H a c -H ■M to •H c ■H 60 - 50 - 40 - 30 • 20 - 10 " 0.25 0.50 0.75 1.00 1.25 Gypsum Board Thickness, inches 125 AGGREGATE QUALITY FROM MULTIVARIATE STATISTICAL ANALYSIS OF AGGREGATE PROPERTIES. Peter P. Hudec Department of Geological Engineering University of Windsor Windsor Ont. N9B 3PU Abstract: The basic properties of aggregates are adsorption / absorption / degree of saturation, density, and thermal and isothermal expansion. The derived properties are the resistance to abrasion, to frost action, and chemical inertness. The approach to-date has been to determine the derived properties by simulation tests. The principal tests in use are the sulphate (magnesium or sodium) to measure frost resistance, and Los Angeles abrasion to measure abrasion resistance. Experience has shown that the tests perform well to separate the very good from the very poor aggregate, but discriminate very poorly between aggregates of the intermediate category. The sulphate test especially has low reproducibility in the critical loss range. Fifteen different tests, measuring basic properties and derived properties were done on one hundred aggregates. The aggregates represented all durability classes and rock types. Analysis of the results has shown that the basic properties listed above can be used to predict the durability of aggregate with much better precission and reproducibility than the standard tests. A statistical multivariate technique of stepwise regression was used to derive equations of durability using the basic properties as variables. Any standard test result can be calculated by use of properly structured equation. The calculated durability parameters achieve much better discrimination in aggregate quality than any standard test, as shown by comparing the service records of aggregates tested. The statistical techniques allow derivation of durability equations for a particular rock type or aggregate type, for particular region, or climatic severity. The accuracy of aggregate durability forecast improves continually as more data is added on which equations can be re-calculated and ref i ned. 126 Key Words: durability; aggregates; tests; multivariate statistics; frost action; abrasion; strength; wear INTRODUCTION The current specification tests measure some basic property of the aggregates, such as absorption, or a simulated response to tests such as magnesium sulphate. Passing limits have been assigned to each test, largely determined by experience with the aggregate material of that geographic locality and the climatic conditions. The aggregate must pass each of the specified tests in order to be accepted. The current standard tests are often vestiges of the past, used because no better tests are available/ but used also because of natural resistance to change from established procedures. The limits developed for the tests are based on the principal aggregate type used/ and therefore do not necessarily apply to other petrographic types. They are designed to separate the very good from the very bad aggregate/ but have difficulty in distinguishing between the not-so-good and the not-so-bad material. When the marginal aggregate or the uncommon petrographic type fails one test but passes others/ the test is often waived. There is no test or property of the aggregate that does not depend on some other property or combination of properties. Thus a test result is a cumulative response of several properties to the test conditions. A simple correlation of, for instance, absorption to freezing and thawing loss has been known for many years. But not until the availability of large computers and the development of the multivariate analysis has it been possible to investigate the multiple relationships of properties and test results. This research has applied modern multivariate statistical techniques to a variety of tests in order to investigate the interrelationships that may exist among them. Once such relationships are established/ a more meanigfull testing program can be devised. The work described in this paper has been supported by the Ontario Ministry of Transportation and Communication (MTC) who provided the samples and performed some of the standard tests. 127 TESTS AND MEASUREMENTS Most of the standard tests and several new tests and measurements that have been devised were applied to one hundred samples of aggregate material from operating quarries and pits in Ontario. The aggregate material was selected to provide a wide range of petrographic types and durability. Field performance of frost resistance / wear, and strength as estimated by the MTC staff was used as the principal dependent variable for statistical comparisons described later. The tests carried out are listed in Table 1. Fifteen distinct types of tests were done, as well as several sub-sets of tests, resulting in a thirty five data items for each sample - a 100 x 35 data matrix. A general description of the new or less common tests is given below. Freeze-Thaw test (in water and in solutions of alcohol and salt) was done on submerged and on saturated, drained samples for five cycles of alternate freezing and thawing. Water adsorbed in seventy two hours (three days) on a dry sample in a 95% relative humidity ,22 deg. C conditions was determi ned. Wet-Dry test consisted of alternately wetting and drying the sample for five cycles. Temperature cycling from room to 85 deg. C and back of samples submerged in salt and alcohol solutions was done, and deterioration measured after five cycles. Petrographic Number determination is a test unique to Ontario, used as one of the principal methods of aggregate evaluation. A durability value of 1, 3, 6 and 10 (l=best, 10=worst) is assigned to certain petrographic types. The sum of the products of percent by weight of each petrographic type and its factor gives the petrographic number. The larger the sum, the worse is the aggregate qual i ty. Bulk Expansion is the volume change of the aggregate between dry and wet state. It is measured in a steel cylinder by a sensitive gauge which detects volumetric expansion in terms of linear expansion. Abrasion of aggregate in dry and saturated state was done in a modified Franklin Slaker, an apparatus in which the sample is rotated in a small, perforated drum along with a charge of steel balls. It is a 1 ow- intensi ty abrasion test with minimal impact (as opposed to Los Angeles Abrasion test, which is mainly an impact test). Impact Test on dry and wet aggregate was done using the British Impact tester. A hammer is dropped on the aggregate from a specified height, and the deterioration is by impact crushing. 128 All other tests are more or less standard, and will not be discussed. Any one individual test is not a good measure of the material's durability. However, certain tests taken together define the durability property sought with a good degree of precission. STATATISTICAL EVALUATION Both simple and multivariate statistics were used to evaluate the data; lack of space permits discussion of only the stepwise regression statistic. In this procedure, a dependent varable is picked against which a set of independent variables is compared. The procedure first picks the independent variable that correlates best with the dependent variable. The second independent variable is then picked which improves this correlation, and so on until a preselected number of variables are picked. The 'best' stepwise regression procedure allows substitution of already picked variables with others than improve the correlation. An equation is thus built up that allows the calculation of the dependent variable from the independent variables. The dependent variables chosen were the standard tests of Magnesium Sulphate, Los Angeles Abrasion, and Petrographic Number (Ontario); in addition, the service records of Frost Resistance, Wear Resistance and Strength were also picked. The equations developed for these dependent variables are given in Table 2. Taking the Magnesium Sulphate test as an example, the test results for Magnesium sulphate can be calculated from the test results of Freeze-Thaw, salt, drained, Absorption, and Wet Abrasion. Figure 1 shows the plot of calculated Magnesium Sulphate against the actual Magnesium Sulphate test results. When the frost resistance record is taken into account, it is seen that the calculated results differentiate the frost resistance better than the actual test results. The intent of the above example is not to suggest the use of calculated magnesium sulphate tests, but to illustrate that the the magnesium sulphate test relates not only to the frost resistance, but also to absorption and abrasion. Similarly, Los Angeles abrasion relates best to impact test, log of adsorption, and vacuum absorption. Perhaps the most most interesting results come from comparing service records to test results. The service records were categorized as 1, 2, and 3, where l=good, 3=poor, and were estimated for frost resistance, wear, and strength. Table 2 gives the tests which best correlate 129 with each of the service record. The equations derived can be used to calculate an index of frost, wear, and strength. It is interesting to note that magnesium sulphate does not enter the frost resistance index, although it is the premiere test used by many to evaluate frost resistance. Another point to note that L.A. Abrasion does not enter the wear index equation, even though it is designed to measure wear. Figures 2 and 3 show the relationship of the frost index to strength and wear respectively. When actual service records of frost resistance are plotted, it is possible to divide the figures into fields of excellent, good, fair, poor, and deleterious aggregate types. The author submits that this is a far more superior method of aggregate evalution, since it is based on actual service records, and has contained in it the results of several tests (see equations of Table 2). CONCLUSIONS 1. Multivariate stepwise regression technique successful in picking out those tests that best correlate to the desired property of the aggregate. 2. Using a statistical approach and sufficiently broad test and data base, it is possible to arrive at indexes of durability that are much better than any current methods of aggregate evaluation. 3. If sufficient data is available, the technique can be refined and made specific for each major aggregate type, such as carbonate, crushed gravel, trap rocks, etc. k. Similarly, the technique can be adapted to regional (geologic) conditions and specific climatic (weathering) envi ronments. POSTSCRIPT The above paper represents an excerpt from a much larger research volume which will be published by the Ministry of Transportation and Communications (Ontario) in the near future. 130 Table 1 - List of Tests Conducted Freeze-Thaw (five cycles): Water-saturated, in air 5% NaCl solution, saturated, in air 5% NaCl solution, saturated, in solution 5% Alcohol solution, saturated, in air 5% Alcohol solution, saturated, in solution Adsorption (72 hr, 95% rel . humidity, 22 deg. C) Normal, untreated samples 5% NaCl solution-treated samples Absorption, saturated surface dried Vacuum absorption, saturated surface dried Wet-Dry cycl ing, 5 cycles High - Low Temperataure cycling (room temp, to 85 deg. C, 5 cycles) 5% NaCl solution 5% Alcohol solution Bulk Expansion on saturation: After 5 minutes After 30 minutes Magnesium Sulphate: Gradation B Gradation C Petrographic Number: Concrete PN Granular "A" PN Los Angeles Abrasion 131 Table 1, Contd. Low Intensity Abrasion Dry - loss through drum - loss through sieve Wet - loss through drum - loss through sieve Impact Test (British Impact Testing Machine) Saturated Sample Dry Sample Densi ty Apparent 1 Apparent 2 Bulk Calculated Pore Parameters Adsorbed water volume Bulk water volume Ratio adsorbed to bulk water Unfilled void volume Service Records Frost Resistance Wear Resistance Strength 132 Table 2. STEPWISE REGRESSION EQUATIONS Magnesium Sulphate = - 12.6 + .88 (Freeze-Thaw, salt, in air) + 6.5 (Absorbtion) + 6.5 (Dry Abrasion Drum) Los Angeles Abrasion s -1.6 ♦ 1.09 (Dry Impact) - 4.25 (log Adsorption) + 1.7 (Vac. Absorb. ) Petrographic Number - 1127 - 379 (Density) + .75 (Bulk Expansion) ♦ 38.2 (Wet Abrasion drum) Frost Resistance Index » .578 + .65 (log Freeze-Thaw, salt, in air) + 1.29 (Absorption) - .84 (Vacuum Absorption) Strength Index - 3.05 + 1.21 (Adsorption) + .28 (Wet Abrasion drum) - .81 (Density) Wear Index ■ 1.19 ♦ .80 (log Vacuum Absorption) + .0021 (Petrographic Number) + .0067 (Mangesium Sulphate) 133 90 o < 113 Z o < 60 30 I.TI o -I -30|- < POOR 3 2 2 2,11 A I 32 3 2 2lj|ii«Bl II GOOD Fig 1 DELETERIOUS 3 3 I, 2, 3 : FROST RESISTANCE RECORD J L J ' 1 ' 1 1 1 1 1 1 1 -J I 11.11. 20 40 60 80 100 MAGNESIUM SULPHATE LOSS %, (B) 3.5 - X UJ Q »- O QC U. 2.5 1.5 0.5 - Fig 2 FAIR - GOOD 2 2 j I22 I I ' 2 ll I IV I EXCELLENT ,' i I, i J I L POOR -i FAIR 2 I I , * — a- ' .,, i ill i i i GOOD DELETERIOUS 3 3 POOR 1, 2, 3 : FROST RESISTANCE RECORO FAIR J I I I I L J I L 0.5 1.5 2.5 STRENGTH INDEX 3.5 134 3.5 Fig 3 1, 2, 3 : FROST RESISTANCE RECORD WEAR INDEX Fig. 1 Comparison of actual and calculated Magnesium Sulphate test results. Fig. 2 Classification of aggregates according to Frost and Strength indexes. Fig. 3 Classification of aggregates according to Frost and Wear indexes. 135 AIR CONTENT OF PLASTIC AND HARDENED CONCRETE D. R. Reidenouer Pennsylvania Department of Transportation Harrisburg, Pennsylvania 17120 R. H. Howe Pennsylvania Department of Transportation Harrisburg, Pennsylvania 17120 Abstract: Air meter tests performed by different operators on the same batch of plastic concrete have a 95% probability of being within 0.80% of the mean. Analysis of variance of linear traverse data revealed that this test has good precision, as long as the air content does not exceed 7%; both wi thin-operator and between-operator variances were non-significant below this level of air. When air meter results are compared to the entrained air content of hardened concrete, as deter- mined by the linear traverse test, there is a 95% probability that the maximum variation between the means of the two will not exceed 1.65%. There is essentially no change in the entrained air content between plastic and hardened concrete . whether vibrated 3 seconds or 30 seconds. Compressive strength is directly related to the total air content; for each percent increase in total air, the compressive strength decreases about ^00 psi. Key Words: Plastic and hardened concrete; entrained and entrapped air voids; linear traverse analysis; analysis of variance. There is frequently poor correlation between lab (linear traverse) results and field (air meter) results. It is particularly difficult to relate the air content determined on an individual core to the field test on a specific batch of concrete. Consequently, there is some doubt among various individuals in the concrete industry that the linear traverse analysis shows accurately the amount of air that was in the plastic concrete. In light of this situation, the relationship between the air content in plastic and hardened concrete needed to be convincingly demonstrated. This study was conducted with this goal in mind, and it had the following objectives: 1. To estimate between-operator variability for air content tests on plastic concrete and the within batch variability of air contents by linear traverse analysis. 2. To determine the effect of vibration on air content. 136 3. To compare air content of plastic concrete with air content of hardened concrete. If. To show the effect of variations in air content on compressive strength. Procedure Construction of Test Slabs In order to achieve the objectives of this project, twelve plain concrete slabs, each 8 ft. by 8 ft. by 9 in. thick were placed on 6 in. of subbase on a test site in Harrisburg, Pennsylvania. The concrete mix used in all the slabs contained 6 l/k bags of cement per cubic yard. There were six slabs containing limestone as coarse aggregate and six composed of gravel coarse aggregate. The varying levels of air content and vibration are shown in the field plot design in Figure 1. Normal vibration was achieved by placing a spud type vibrator in the center of each quarter of each slab for three seconds. For excessive vibration, the spud was placed at two foot intervals for 30 seconds. Figure 1 gives the code designation for each slab. Thus, ML-Ify f stands for limestone coarse aggregate, no air entraining agent (about 1% entrained air) and normal vibration (vibration level l) . The slabs were cured under polyethylene sheeting for 28 days. At the end of this time a series of six-inch diameter cores were drilled from the slabs to investigate the properties of the hardened concrete. Tests on Plastic Concrete The participation of construction inspectors in this project was considered important for 'communicating the results to the project level, Consequently, each of the eleven engineering districts of Pennsylvania sent an experienced project inspector to the test site. These inspectors were each equipped with a calibrated Acme Air Meter. Upon delivery of the concrete to the test site, each inspector made one air meter test (ASTM 231-75) on each of the six different batches. Tests on Hardened Concrete Ten six-inch diameter cores were removed from each of the twelve slabs. Of these cores, five were used for compression testing and five for linear traverse analysis. Compression tests were performed according to ASTM C39-72. The linear traverse analysis procedure used to study the air void system in hardened concrete is outlined in Pennsylvania Test Method 623 (par. 5.3). In PTM 623, the selection of one milli- meter as a division between entrained and entrapped air is based on the research work of Larson et al (l). 137 Results and Discussion Air Meter Tests on Plastic Concrete The air content of the plastic concrete was determined for each mix by- eleven different operators using Acme pressure-type air meters. The results of their tests are given in Table 1. The results show that, for most of the mixes, the air meter operator variability is rather small. To get a better picture of this variability, the standard deviation about the mean was determined for each mix. The last line of Table 2 contains this data. In a normal curve, 95% of the distribution lies less than two standard deviations from the mean. Thus, in 95% of air meter tests performed by different operators, the results will vary between 0.986% and 0.316% from the mean. It would be desirable to obtain an overall measure of variance for all the slabs instead of just the range of standard deviation. Since each mix has a different air void content, it is not valid to combine the results to obtain one grand mean and standard deviation. However, if we consider the standard deviations as values normally distributed about an ideal mean deviation of zero, we can determine a standard deviation of the deviations. Taking the average of the seven standard deviations given in Table 1, a mean standard deviation of 0.26 was obtained. This average and the seven standard deviations were then used to calculate a standard deviation of the deviations. This standard deviation was designated by the symbol && and was found to be 0.11. From this value the researchers set up confidence limits as follows: Confidence Limit (C) = o* x a Where: n = number of samples y^- a = factor based on desired significance level If we select a 95% significance level the equation becomes: C = 2lil x 2 -78 = 0.14 This means that there is a 95% probability that the standard deviation of air meter values will be within the limits + O.lU about the mean. In other words, one can designate the precision of the standard deviation to be 0.26 + O.lU. Using the 95% confidence limits, we see that the maximum standard deviation would be 0.26 + 0.1^ = 0.^0. Two standard deviations would be 0.80 and would include 95% of the area under a normal distribution curve. Thus, we can say that in 95% of the air meter tests performed by different operators, the results should not vary more than 0.80% from the mean. 138 Linear Traverse Analysis Five cores from each test slab were analyzed for air content using the linear traverse method. The average percentages of entrained, entrapped, and total air content for each slab were calculated from the test results of each set of five cores. These averages are presented in Table 2. The average percent air obtained by air meter tests for each slab is repeated in the last column of the table for comparison purposes. Effect of Vibration on Air Content - The effect of vibration on entrapped air is clearly shown in Table 2. The entrapped air is always less in the slabs which underwent excessive vibration (those with the subscript 2 after V) . The average entrapped air content of the slabs with limited vibration is 1.88% while the slabs which were vibrated excessively contained only 0.86% entrapped air. Although the difference is not great, it is a significant difference relative to the total amount of entrapped air present. Thus, vibration can significantly reduce the amount of entrapped air in concrete. To check the validity of these visual comparisons, a 't' test was used to compare the means of both the entrained and entrapped air at the two vibration levels. The null hypothesis used for the ' t' test of the entrapped air was, "there is no significant difference between the means of the entrapped air at vibration level one and two." Using the standard methods, a ' t' value of 5.16 was obtained and it was found to be highly significant at the 5% level. In other words, the mean of entrapped air of the slabs receiving limited vibration is signifi- cantly different from the mean of the excessively vibrated slabs. Turning to the entrained air content, we begin with the null hypothesis that "there is no significant difference between the entrained air content of slabs from vibration levels one and two". Because the entrained air was designed at three levels it is not valid to determine a mean for all the values at each vibration level. Therefore, the comparison of entrained air was performed separately for each mix. The results of this statistical study are presented in Table 3- There are no results for mix MG-4 because two different batches were used for the two vibration levels. The 't' test values are not significant for the first three mixes. Therefore, the null hypothesis is accepted, and we can say that in the case of mixes ML-1, ML-^, and ML-7 vibration had no effect on the entrained air content. On the other hand, mixes MG-1, and MG-7 yield significant 't' values and the null hypothesis would be rejected. However, there are some other factors to consider. In this phase of the research we were particularly interested in observing whether excessive vibration causes a reduction in entrained 139 air. Therefore, although mix MG-7 shows a significant difference between the entrained air content of the concrete at two levels of vibration, the means still indicate that the vibration does not reduce entrained air. In mix MG-1 there is a different situation. No entraining agent was used in mix MG-1 so the entrained air present was produced naturally during mixing and placing of the concrete. Accord- ing to Backstrom (2) the film surrounding entrained air voids tends to resist distortion, and larger entrained air voids are less likely to disintegrate into smaller bubbles during vibration than entrapped voids are. Since mix MG-1 contains no entraining agent, the voids present (including those of entrained air size) are more susceptible to dis- tortion and the air content could be reduced by vibration. Therefore, it is the opinion of the researchers that the significance of the 't' test in the case of mix MG-1 is not very strong evidence that vibration may reduce the entrained air content. The general conclusion, from this area of study, is that excessive vibration has little or no effect on the entrained air content of concrete. On the other hand, vibration significatly reduces entrapped air in concrete. Within Batch Variation of Air Content - The percent total air in the hardened concrete varies from 0.38% to 3 •00$. However, in order to fully study this variation, it was necessary first to determine how much of the variation was caused by normal operator variation. It was decided that this could best be accomplished by using the analysis of variance technique. A testing program was designed which consisted of four different operators performing linear traverse analysis on three different samples. The results of these linear traverse tests are presented in the Appendix. After all the data was obtained, an analysis of variance was performed on the test results. The results are presented in Table k. Comparing the variances to the residual variance, it is apparent that at none of the air levels is the within-operator variance significant (the residual variance is greater) . In studying the between-operator variance, one can see that is is not significant at the two lower levels of air, but is at the highest level of air. It appears that between-operator variation can be disregarded at lower levels of air but cannot be ignored when the total air content exceeds 7$» If two operators each make three determinations on a sample, the difference between the two averages that is statistically acceptable is determined by the following formula: Where o is the residual variance r(a) is a factor obtained from a table of significant ranges for a 5% level of a multiple range test (3). 140 This formula and the residual variances of Table k were used to cal- culate the results given in Table 5» The numbers in the last column of the table are a measure of the precision at the various air levels. For instance, at the highest level of air, two results may differ as much as 2.08% and be considered statistically to represent the same air level. Variation Within Slabs - Using the information obtained from the analysis of variance, we can now consider the variation of air content within slabs. The variation introduced by the linear traverse method for a particular air level is the total variation found in Table k. To determine if the between core variation (within slab) was significantly greater than the variation produced by the linear traverse, a standard "F" test was performed. The 5% significance level was chosen for the critical value. The results of the calculation can be found in Table 6. The results show that all but one of the F values of the lowest air levels are significant, but only two of the "F" values of the higher two air levels exceeds the critical value. From the variances given in the second column of Table 5> one can estimate that the air content would vary about 0.5% to 2% within a given batch of concrete. Any variance above 2% should be looked on with suspicion as being • abnormally high. Comparison of Air Content in Plastic and Hardened Concrete Inspection of Table 2 reveals the variation of air content between plastic and hardened concrete. The difference between meter results and total air varied from a high of 2.60% in slab MG-1V, to a low of 0.2*1$ in slab MG-7V1. The average variation between total air content in hardened concrete and the air meter values is 1.29%. This average variation plus the range of variations among slabs would seem to indicate that there is only a fair correlation between linear traverse and air meter results. However, it will be noted that there is considerable entrapped air in some of the slabs. Apparently, the vibration and working of the concrete during placement was not sufficient to decrease the entrapped air as expected. However, in using the air meter one rods each layer 25 times, thus reducing the entrapped air considerably. Consequently, the air content reported from air meter readings is largely entrained air. Therefore, the lack of good correlation among linear traverse (total air) and air meter values may be due largely to differences in entrapped air content. This hypothesis is supported when one compares the entrained air values in Table k with the air meter results of Table 1. The variation between the two ranges from a high of l.k% for MG-7Vl to a low of 0.06% for ML-UV2. The maximum variation is much less in this case and the average variation is only 0.67%. To obtain some idea of the probable deviation of entrained air values from air meter results, the differences were considered to be a sample 141 representing a normal distribution with a true mean of zero. This set of differences has a mean of O.67 and a standard deviation of 0.^9. Two standard deviations would be O.98 and would include 95% of the area under a normal curve. Therefore, there is a 95% probability that the difference between air meter results and entrained air will not deviate more than O.98 from the mean of O.67. Relating this to the actual percentage of air, it can be said that in 95 out of 100 cases the mean entrained air (determined by linear traverse analysis) will not vary more than 1.65% from the mean air meter results of the same concrete. The mean value of either entrained air or air meter results should be based on an average of two or more readings for best correlation. Compressive Strength Five cores from each slab were tested for compressive strength. The average strength for each slab is given in Table 7« Previously, it has been shown that vibration of plastic concrete reduces the entrapped air content of concrete, thus producing more consolidated concrete. One would expect this concrete to have more strength than less consolidated concrete and the results in Table 7 clearly support this. In five of the six pairs of slabs, the compressive strength of the V*2 slabs is higher than the V]_ slabs. Thus, all factors being equal, vibration will result in a significant increase in compressive strength of the concrete by consolidating it more. Compressive Strength Versus Air Content To statistically investigate the apparent relationship between compressive strength and air content, regression and correlation analysis were performed. The average total air content for each slab was compared to the average compressive strength of each slab, respectively. The values are plotted on the graph in Figure 2. The straight line in the figure was obtained from the regression analysis and represents the best fitting line for the points. The correlation factor for this comparison was 0.92. Thus, there is a very good correlation between total air void content and compressive strength in concrete. This conclusion is similar to that reached by other investigators. In addition to the general correlation, Figure 2 can also be used to determine the relative change of compressive strength with increase in total air content. By using the slope of the line, one finds that the compressive strength decreases about ^00 psi for every percent increase in air. This means that a concrete mix that was designed for ^000 psi strength with an air content of 5*00% would have only 2800 psi compressive strength if the air content were increased to 8.00%, all other factors being equal. 142 Limestone C.A*. Gravel C.A. 8' 0% air normal vibration (ML-IV ) 0% normal vibration (MG-IV ) 0% air excessive vibration (ML-IV 2 ) 0% air excessive vibration (MG-IV 2 ) k% air normal vibration k% air normal vibration (MG-^V ) k% air excessive vibration (ML-4V 2 ) k% air excessive vibration (MG-^V 2 ) 7% air normal vibration (ML-^) 7% air normal vibration (MG-TV.^ 7% air excessive vibration (ML-7V 2 ) 7% air excessive vibration (MG-7V 2 ) 8' Figure 1: Design of Field Plot 143 TABLE 1: Air Content of Plastic Concrete from Air Meter Tests (as performed by District Inspectors) Total A ir Content (%) Inspector from Mix Mix Mix Mix Mix Mix Mix District ML-1 ML-1+ ML-7 MG-1 MG-l+V 1 MG-l+V MG-7 1 1.6 3.1* 5.8 1.9 1+.0 ■ 5.8 7.0 2 1.6 3.1+ 5.6 1.7 l+.l 5.6 7.0 3 1.8 3.5 5.9 2.1 l+.l 5.6 7.1+ k 1.1* 1+.2 5.1+ 1.5 3.9 5.2 6.7 5 1.1* 3.0 5.8 1.5 1+.0 5.3 6.9 6 1.5 3.2 6.0 1.7 1+.0 5.2 7.7 8 1.6 3.2 5.8 2.0 1+.3 5.9 7.2 9 2.0 3.6 6.0 2.1 1+.5 5.9 7.1+ 10 1.1+ 3.1+ 5.6 1.7 l+.l 5.7 6.9 11 1.6 3.3 5.9 1.9 l+.l 5.9 6.8 12 3.2 3.2 5.9 1.9 1+.2 5.7 6.8 Mean 1.71+ 3.1+0 5.79 1.82 1+.12 5.62 7.07 Standard Deviation 0.1+93 0.299 O.185 0.201+ O.I58 0.158 0.299 144 TABLE 2: Air Content of Hardened Concrete by- Linear Traverse Analysis (average values per slab) Air Content (%) Average of Five Tests per Slab Slab No. Entrained Air Entrapped Air Total Air Air Content by Air Meter ML-1V O.76 1.62 2.38 1.7*+ ML-1V 2 1.22 1.00 2.22 ML-ML 3.68 1.5* 5.22 3.^0 ML-^V 2 3.^6 0.65 if.ii ML-7V 1 6.7^ iM 8.19 5.79 ML-7V 2 7.19 0.76 7.95 MG-1V 1 1.68 2.73 k.k2 1.82 MG-1V 2 O.67 1.6*1. 2.31 MG-4V- 3.88 2.77 6.65 k. 12 MG-^V 6.21 0.57 6.78 5.62 MG-7V 1 5.66 1.17 6.83 7.07 MG-7V 2 6.78 O.54 7.33 145 TABLE 3: Comparison of Entrained Air at Two Different Levels of Vibration Using the 't' Test Mix No. Mean of Entrained Air (%) 't' Significance Using 8 Degrees of Freedom v i V 2 ML-1 ML-1* ML-7 MG-1 MG-7 O.76 3.68 6.7^ 1.68 5.66 1.22 3.^6 7.19 0.67 6.78 O.85 0.81 0.79 3.52 3.6^ NS* NS NS Significant at 5% Level Significant at 1% Level *NS-No significant difference between the means. 146 TABLE k: Variances at Three Different Air Levels Variance Air Between Within Residual Total Levels Operator Operator Variance Variance Low {0-3%) 0.053 0.004 0.021 0.027 Medium (5-7$) 0.520 0.166 0.562 O.V78 High (756) 5.81* 0.012 0.986 2.125 ^Significant at 5% level. TABLE 5 : Limits for Differences Between Two Means in Linear Traverse Analysis at Three Levels of Air Air Level s 2 (a) r(a) Low 0.021 3.64 Medium 0.562 3.64 High 0.986 3.64 0.30** 1.58** 2.08** ** All values at 95% level of significance, 147 TABLE 6: F-Test Results Obtained from Comparing Within Slabs and Linear Traverse-Induced Variations in Concrete Variations Air Slab Within Slab Induced by "F" Level No. (between core) Linear Traverse Test Value Low MG-1V, MG-1V, ML-1V. ML-1V, .9& .327 .089 .677 .027 .027 .027 .027 20.15* 12.11* 3.50 26.63* Medium MG-UV ] MG-UV, ML-UV. ML-4V, 2.091 1-5^3 .955 .3^0 .V78 .478 ^.37 3.23 2.00 l.kl High MG-7V- MG-7V, ML-7V. ML-7V, .121 .936 .888 1.05^ 2.125 2.125 2.125 2.125 17.56* 2.27 2.39 2.02 ^Significant at 5% level. 148 TABLE 7: Compressive Strength of Cores Slab No. Compressive Strength (psi)* Average of Five Tests Per Slab ML-1V 5922 ML-1V 6^01 ML-^V 1 5577 ML-W 2 5893 ML-7V ^25^ ML-7V ¥+50 MG-1V 1 5863 MG-1V 2 6258 MG-W 1 5121 MG-4V k76k MG-7V 1 3850 MG-7V kYjk * Compression tests were performed approximately ^4-0 days after placement of the slabs. 149 o o o o O o O o O ID m o o o o o o ro O O O CM O O O 0) tf) i» a> > o a fc» h- k. o a> c *Zj E 2 H- ^_ T c Q) jC o ♦» > "c o» £ a c a> d < ty5 h- a> Z "5 .> UJ w ■z. i§ to 0) o C o $ cr JS o O < 5 GO "O _j < CL c o h- £ o CO ,»-*, h- C *<0 •*- >. ^ o a> c or < • • CVJ UJ or 3 CD (!Sd) H19N3&LS 3AISS3ddW0D 150 References 1. Larson, T. D., et al, "Durability of Bridge Deck Concrete," Department of Civil Engineering, The Pennsylvania State University, a series of eight reports, 1965 ~ 1968. 2. Backstrom, et al, "Origin, Evolution and Effects of the Air Void System in Concrete, Parts 1 to k - Influence of Water Cement Ratio and Compaction, " Journal of American Concrete Institute, August 1958, pp. 359 - 372. 3. Duncan, D. B., "Multiple Range and Multiple F Tests," Biometrics, March 1955- k. Gregg, L. E., "Experiments with Air Entrainment in Cement Concrete," Bulletin of the Engineering Experimental Station, University of Kentucky, Lexington, Kentucky, Vol. 2, no. 1, 19^7. 5. Ivey, D. L. and Torrans, P. H. , "Air Void Systems in Ready-Mixed Concrete," Journal of Materials, Vol. 5, no. 2, 1970, pp. ^92-522. 6. Powers, T. C, "Topics in Concrete Technology," Bulletin Yjk, Research and Development Laboratories of the Portland Cement Association, 1964, pp. 35-77. 7. Powers, T. C. and Brownyard, T. L., "Studies of the Physical Properties of Hardened Portland Cement Paste," Journal of the American Concrete Institute, Proceedings, Vol. k3 f 19^7 > pp. 101-992. 151 APPENDIX Results of Linear Traverse Analysis on Three Selected Concrete Blocks for Analysis of Variance Block No. Total Air ») Operator Grand Mean of 12 Tests A B C B MG-LV 2 (5) Ave. MG-4V (11) Ave. MG-7i(lD 2.66 2.1+3 2.30 2.21 2.08 2.30 2.38 2.35 2.1+1 2.1+1 2.66 2.1+1 2.38 7.00 8.96 2.1+6 7.53 6.1+2 6.61 2.20 6.85 7.67 8.31 2.38 6.36 7.69 5.97 2.^9 6.1+2 7.01 7.15 6.85 6.85 8.76 7.61+ 7.61 11.15 9.75 11.95 6.67 8.78 9.11+ 8.1+3 6.86 9.22 8.30 7.57 7.75 10.95 8.78 8.36 152 ABSORPTIVITY, A MEASURE OF CURING QUALITY AS RELATED TO DURABILITY OF CONCRETE SURFACES Ephraim Senbetta Civil Engineering Purdue University West Lafayette, IN 47907 C. F. Scholer Civil Engineering Purdue University West Lafayette, IN 47907 Abstract: It is a well known fact that proper curing of newly placed concrete is essential if desirable qualities and expected durability of the hardened concrete are to be realized. Curing is particularly im- portant for exposed concrete structures such as precast panels, slabs cast on grade such as in tilt up construction, pavements, sidewalks and patios, and in many exposed architectural panels where proper uniform curing helps assure uniform color and pleasing appearance. Although there are specifications for the various curing methods, review of the literature indicates that there is no method by which the effectiveness of a certain curing method and the resulting quality of the concrete can be evaluated. The method that has been developed in this study involves characteristic changes in absorptivity of the concrete in going from the surface into the bulk. The test is quick and easy and reflects changes in the pore structure of the paste depending on the severity of the ex- posure condition and extent of curing. For a given curing method, duration of curing, and atmospheric conditions, a characteristic curve which shows changes in the absorptivity of the concrete at various depths is produced. This is a reflection of changes in the pore struc- ture at the various depths. The better the curing the less porous the paste and the resulting absorptivity is small and vice versa. There is evidence which indicates that changes in the pore structure of the paste from one layer of the concrete to another as determined by absorptivity test, is directly influenced by the curing. By examining the various layers of core samples, an evaluation of the effectiveness of the concrete curing can be made. Key Words: Absorption tests; absorptivity; cement paste; concrete; curing; curing methods; durability; mortar; slabs. Successful use of concrete as a building material requires that pro- visions be made for the hydration of the Portland cement component. 153 This, time and temperature effected chemical reaction between cement and water produces hydration products which form hardened cement paste and create strength for supporting structural loads and the imperme- ability required for many environments which need durable concrete. The process which brings about the desired hydration is generally known as the curing of the concrete. In most situations the conditions of curing are not assured for the length of time which may be necessary for the complete development of the concrete's potential. In so far as the effect on structural capacity, it is not of great consequence since the formwork and volume of the concrete involved will inhibit drying of the great majority of the mass for a time sufficient to allow adequate strength to develop unless it is necessary for this strength to exist on the exposed surface of the concrete. Principal examples of exposed surfaces which are critical with respect to their strength, as well as durability, are flat slabs which are ex- posed to both severe wear and to aggressive exposures. These slabs are common on both pavements and industrial floors. Construction practices are often such that adequate curing may well be missing for the surfaces of many of these slabs. No matter how poor the curing conditions may be, the bottom of a slab is not affected nearly as much as the surface. Therefore, it is reasonable to expect a certain characteristic change in paste properties with depth for a given curing condition and duration of curing. In this study the effectiveness of the curing was evaluated based on the behavior of the entire thickness or at least the affected thickness rather than just the surface. Con- sequently, the tests were performed on samples taken from the various depths of the total thickness of each slab. An observation was also made on how deep into the slab the effect of poor curing had influenced the material . Mortar was used in this investigation rather than concrete because the amount of cement paste (cement and water) in the mix is greater and the effect of aggregate on small test samples reduced when mortar is used, Differences no doubt exist between concrete and mortar but mortar is certainly indicative of the performance of concrete. Absorption tests considered in this study include the British test for determining the initial surface absorption of concrete (1)*, and Figg's method for measuring the air and water permeability of concrete (2). Largely due to the desire to investigate the effect of curing quality at different depths near the surface these methods were not used. Absorptivity test which was devised and used by Powers and Brownyard on hardened cement paste (3), and later by Dolch in a study of the pore structure of Indiana limestone coarse aggregates (4) is a simple test that has proved to be a sensitive means of evaluating curing quality. ♦Numbers in parentheses refer to the list of references at the end of this paper. 154 It is known that a sample of cement paste maintains the same external volume during the hydration process, but within the sample itself the volume of solids increases. This in turn is the cause for the reduction of the porosity of the paste (5). Therefore, if concrete is subjected to good curing for a given length of time its porosity can be expected to be less than that of a similar concrete sample that was not cured properly during the specified curing period. Absorptivity depends on porosity, size of pores and microcracking, and the coefficient of absorptivity, K a is defined by ($" K a t where j- = amount of water per unit area absorbed in elapsed time t # K a = coefficient of absorptivity The test was done in the following manner. A one inch diameter core which was cored from a 3.5" thick slab was taken out of a bath of methanol where it was stored, since the end of the curing period and the top 1 mm of the core was cut off and discarded. Then, the rest of the core was sliced into several 1 cm thick disks. The cutting was done using an Isomet low speed saw equipped with a low concentration diamond blade. In the cutting process, ethanol was used as a coolant, and the core was never in contact with water after it was cored. When the core was sliced, each disk was marked on its side to indicate its top cross section which was tested after drying. From a three and half inch long core, eight disks were obtained, and on the back side of each disk a number was written to indicate from what depth the disk was obtained. The cut surfaces had a virtually polished texture. As soon as the slicing was done, the disks were put in a desiccator and the desiccator was connected to a vacuum pump for approximately 48 hours This was the drying process. Other methods of drying were tried but none were better. Oven drying at temperatures as low as 60°C was found to alter the samples by introducing microcracks. Also, the effect of heat was not the same on all disks. For a disk obtained from near the top of a poorly cured sample, the effect of heat was minimal, but for disks from deeper zones the effect of heat was pronounced. Therefore, drying under vacuum at room temperature was the best method to have the capillary pores at least partially dry. The duration of drying was arbitrarily chosen but it is believed that drying for a shorter period of time would not empty the pores adequately and drying for longer periods of time resulted in higher absorptivity values. Part of the results of this investigation are shown in Figures 1 and 2. The two figures are typical plots of the results of the absorptivity tests done on mortar samples made with graded silica sand from Ottawa, Illinois. The mix proportions for the mortar consisted of a water- 155 cement ratio of 0.5, and a sand to cement ratio of 2.74. Other mortars with the same mix proportions were made with natural sand and they pro- duced similar results. The absorptivity values for the wery top of each core are ignored because the very surface region was affected not only by the curing but also by the formation of water channels from bleeding. Figure 1 shows a decrease of absorptivity with age in a similar fashion as increase in strength of concrete with age. Also, for curing with wet burlap at ages three and five days there was not much difference in the absorptivity of the surface region compared to the absorptivity values for regions deeper into the slab. This is indicative of good curing. Figure 2 shows absorptivity test results for slabs cured under different curing conditions but for the same length of time. The plots show that the slabs that were cured with wet burlap, with plastic cover, and with no cover in a windy environment all had essentially the same absorptivity throughout their entire thickness. On the other hand, the slabs that were cured with curing compound, and with no cover in a calm environment displayed significantly higher absorptivity values at their surfaces in comparison to the absorptivity values at deeper zones. The poor performance of the curing compound was due to, as it was later discovered, the poor quality of the material used. The result for the sample cured with no cover in a windy environment may be surprising because that type of a curing condition is thought to produce a poor surface. As shown on Figure 2, although the sample cured with no cover in a windy environment was not quite as good as the samples cured with wet burlap and plastic cover, it had an approximately uniform ab- sorptivity value throughout its entire thickness. The reason for this is thought to be due to the rapid rate of evaporation of mixing water while the mortar was still plastic thereby reducing the effective water-cement ratio at the surface of the mortar (6). Also evaporation of water up until the time when the mortar ceases to be plastic contributes to surface densifi cation as a result of forces due to capillary action and surface tension (7, 8). This of course, does not imply that the more the evaporation the better, because high rate of evaporation accompanied by low bleed rate and slow setting can cause plastic shrinkage cracking. The effect of the curing condition or method of curing interacting with time and humidity is in itself of interest and practical value. How- ever, a major reason for this investigation was to develop a means of evaluating the quality of curing on an existing hardened concrete (9). No method, exists today by which one may judge whether or not poor concrete performance is caused by it having been poorly cured. The existence of the pronounced increase in absorptivity near the surface with, poor methods of curing indicated that similar tests on concrete may well be used as indicators of its past treatment. This paper is based on the Ph.D. Thesis of E. Senbetta, (Purdue University, 1981) and represents initial results of a research project supported by the Indiana State Highway Commission. 156 References: 1. "Test for Determining the Initial Surface Absorption of Concrete," British Standard 1881, Part 5, British Standards House, London, 1970. 2. Figg, J. W., "Methods of Measuring the Air and Water Permeability of Concrete," Magazine of Concrete Research , Vol. 25, No. 85, December 1973, p. 213. 3. Powers, T. C, and Brownyard, T. L., "Studies of the Physical Properties of Hardened Portland Cement Paste," Research Department Bulletin No. 22, Portland Cement Association, 1948, p. 873. 4. Dolch, W. L., Permeability and Absorptivity of Indiana Limestone Coarse Aggregates , Ph.D. Thesis, Purdue University, 1956. 5. Popovics, S., Concrete-Making Materials , McGraw-Hill, 1979, p. 86. 6. Burnett, G. E., and Spindler, M. R., "Effect of Time of Application of Sealing Compound on the Quality of Concrete," ACI Journal , November 1952 p. 193. 7. Shalon, R., and Ravina, D., "Studies in Concreting in Hot Countries," RILEM Symposium on Concrete and Reinforced Concrete in Hot Countries, Haifa, July 1960. 8. Swayze, M. A., "Finishing and Curing: A Key to Durable Concrete Surfaces," ACI Journal , December 1950, p. 321. 9. Senbetta, E., Development of Techniques to Quantify Concrete Curing Quality , Ph.D. Thesis, Purdue University, 1981. 157 CM (0 i © m I E o o X (0 26 24 2 2 20 A A 1 DAY 10 nm) behaves similarly to bulk water. Freezing sets in between -5°C and -12°C depending on nucleation (supercooling) and continues to -40°C depending on pore radius (depression of freez- ing point). 2. Water in pores between 3 and 10 nm in diameter seems to be structured. Its freezing point is lowered to about -40°C to -60°C, with a maximum at -43 °C. The heat of fusion is drastically diminished. 3. Below -60°C the only unfrozen water is that which interacts most strongly with the solid surfaces. It consists either of liquid films a few molecular layers thick or of water in very small "pores". This water "freezes" gradually between -60°C and -130°C. The abnormal freezing is reflected in abnormal mechanical behaviour. In thermal dilation we not only find expansion, as can be expected, but also severe contraction. Either effect can be dominant depending on conditions. Elastic modulus and damping are strongly affected by ice formation. The effects can be understood in terms of "Extended Munich Model." Key words: Abnormal freezing; concrete; differential thermal analysis; dispersed systems; hardened cement paste, solid-water inter- actions; sorption; thermodynamics; x-ray. 1. INTRODUCTION Water which is physically bound in hardened cement paste (hep) interacts strongly with the solid particles. Therefore the phase transition from liquid water to ice in this system is quite different from bulk behav- iour. Frost deterioration of hardened cement paste and concrete cannot be explained simply by the expansion of the water as it freezes. Sev- eral other superimposed mechanisms must be considered. Powers [1,2] was the first to emphasize that the freezing point of water in hep is the lower the smaller the pores of the hep. Helmuth [3] improved these findings. Litvan [4] states that the equilibrium between unfrozen water 160 in the small pores of hep and the ice formed outside is disturbed, which leads to transport of water from the pores to the ice. This mechanism is equivalent to the water transport in soils studied by Everett and Haynes [5]. According to Litvan, water in the pores should only solid- ify to a glassy, amorphous state. However, as mentioned below, we found ice Ih in pores. The interactions of water and ice with the solid gel particles of hep can only be understood if we take into account the non- macroscopic behaviour of water in this highly-dispersed system; this is quite different from bulk behaviour. In an effort to fill this gap, we developed the "Extended Munich Model." Its basic thermodynamic consid- erations have been published [6]. Subsequently we proved and improved the model by the experiments outlined below [7-13]. Of course, macroscopic aspects of frost action must be kept in mind, e.g. Harnik, Meier and Rosli [14,15] hint at the inhomogeneity of temperature, of salt concentration, and of the structure of the concrete near the sur- face of the specimen. Podvalny [16] calculated the stress around an aggregate particle in concrete. 2 . THEORY The theory is developed from thermodynamic considerations which pay special attention to the interaction between the highly-dispersed gel particles of hep and the physically-bound water and ice in the system. (The theory and its validity is discussed in detail in [6]). To a fair approximation, the following relation between freezing point and pore radius can be used: v.( d> . - ) T AT = ! 8* 1 §» w ° (1) h .R AT = Change in freezing point with respect to bulk water T = Freezing point of bulk water g ^; (j>g w = Specific surface energies of the gel-ice (g,i) and the gel water (g,w) interfaces respectively R = Hydraulic radius of the pore (fraction of pore volume/pore surface) h = Heat of fusion of pore water v^ = Molar volume of ice In this context it must be noted that the heat of fusion h is also dependent on the pore radius and that the surface energy of the gel-ice interface changes with temperature, since there exists an unfrozen film several molecular layers in thickness between pore ice and gel surface, whose thickness diminishes as the temperature is lowered. The freezing point is lowered by the interactions of the pore water with the internal surface of the gel. In equation 1, the surface interac- tions and the changes upon freezing are represented by ^g i> g w an( * n represents the lowered heat of fusion in smaller pores. 161 Thermodynamically, surface interaction lowers the chemical potential (i.e. the energy content of one mole) of water near the surfaces. Under equilibrium conditions, if particle exchange is allowed, the chemical potentials at the boundary of two phases must be equal. In pores, the chemical potentials of condensed water and ice are balanced by pressure terras, Jvdp, in which v is the raolar volume and p the pressure. As a rule, this leads to a negative pressure. Since, below its freezing point, bulk ice has a lower chemical potential than bulk water, after freezing an additional negative pressure is generated between pore ice and unfrozen pore water. To establish this negative pressure in a rigid system only small amounts of water must be transported from smaller to bigger pores. It should be noted that the unfrozen water is not super- cooled, which would imply a metastable or unstable state, even though its freezing point is lowered. There is a real depression of the freez- ing point. It can be shown that, even after freezing, a pressure dif- ference between pores of different size is produced: p - * g)1 - * g „ li--i-l ■ w \k'k\ Again this leads to a negative pressure in the small pores. This could explaLn a contraction. Finally, one must recognLze that it is only one of the necessary condi- tions that thermodynamic rules be fulfilled. In addition, statistical and kinetic effects must be considered, e.g. heterogeneous and homogen- eous nucleation which leads to supercooling [12]. 3. RESULTS AND DISCUSSION As a first step, the chemical potential of sorbed water films on surfaces has been measured in the observed condensation [10]. The change of chemical potential with thickness for the sorbed film follows an exponential law. It has been shown in this article too, that the van der Waals interaction plays a less important role than structural inter- action. By a superposition of two of these chemical potentials, the chemical potential of water in a slit-shaped pore has been calculated (fig. 1). It can be seen that the chemical potential is lower the smaller the pore. In a second step [7,8], differential thermal analysis (DTA) has been used to evaluate the heat of fusion and the depression of freezing point in pores of different size (fig. 2). The heat of fusion is plotted in fig. 3 [13]. From the DTA results, three kinds of water can be distinguished : - Water in pores with a radius bigger than 10 nra is nearly comparable with bulk water. Indeed, in pores bigger than 100 nm in diameter no significant difference is observed. These pores are only filled in water-saturated specimens. 162 In pores above 10 nm in diameter, water condenses in a relative humidity range between 90 percent and 100 per- cent. Its freezing point is already remarkably reduced in the smaller pores to as low as -30 °C. However, only about 30 percent of the water in a water-saturated sample is of this kind which freezes at temperatures down to -30°C. - In pores between 3 ran and 10 ran in diameter, water con- denses between 50 percent and 90 percent r.h. The struc- ture of this water is somewhat different from that of bulk water. It undergoes a phase transition around -43 °C and has a reduced heat of fusion. - Finally, there is the water near the internal surfaces which does not freeze above -60°C. In the DTA results and in dielectric constant measurements, a phase trans- ition of this water could not be observed. However, the results of Mo'ssbauer spectroscopy and of dynamic modulus of elasticity measurements show that this water "freezes" between -60°C and -130°C [17,11]. From the results of x-ray diffraction measurements [12], we can state that water mainly freezes in the crystalline structure of ice Ih. The x-ray studies also provided a measure of the relative amount of frozen water and showed a distinct hysteresis. If freezing sets in in a sample of hep which is almost saturated with water, we normally observe a rapid contraction [9,13] in our thermal expansion experiments (fig. 4). This can be explained by the above- mentioned negative pressure in smaller pores. An expansion of the spe- cimen, which should occur due to the volume expansion of the water-ice transition, has only been observed in samples with a large number of the bigger pores, e.g. samples with a high water-cement ratio. Whereas diffusion of small amounts of water above -30 °C seems to be possible, diffusion at -40°C is extremely difficult. This is reasonable since transport at this temperature is only possible in the very small pores containing unfrozen water and in the unfrozen liquid- like film near the surface, which is only a few molecular layers thick. Therefore, at -40°C, the effect of the expansion of the ice is observable. Finally, the measurement of the dynamic modulus of elasticity and its evaluation by an appropriate model allows calculation of the relative amount of ice in hep as a function of temperature [11]. Using eq. 1, the pore size distribution of ice-filled pores can be calculated from these data (fig. 5). 163 4. CONCLUSIONS Water in hardened cement paste has a structure quite different from the bulk water phases. This results in the superposition of other phenomena on the pure thermal expansions of bulk water, bulk ice, and dry hep: - The freezing point depends on the pore radius of the hep The changes in the interactions between the gel particle surfaces and pore water and pore ice lead to a contrac- tion of hardened cement paste and to some redistribution of the water. - Only about 1/3 of the total water in water-saturated sam- ples freezes above -20°C. This water is also the first to evaporate. Therefore, the water content is a critical factor in determining the durability of concrete against frost action. For this reason, storage conditions used in durability tests must be well-defined and correlated with the actual exposures of concrete structures. 164 5 . REFERENCES 1. Powers, T. C, "A Working Hypthesis for Further Studies of Frost Resistance of Concrete," J. A. C. I., Proc . , Vol. 41, 1945, p. 245. 2. Powers, T. C, Helmuth, R. A., "Theory of Volume Changes in Hard- ened Cement Paste," Proc. Highw. Res. Board, Vol. 32, 1953, p. 285. 3. Helmuth, R. A., "National Bureau of Standards Monograph 43, Pro- ceedings 4th International Conference on the Chemistry of Cement, Washington, D.C., Vol. 2, 1960, p. 855. 4. Litvan, G. C, "Freeze- Thaw Durability of Porous Building Materi- als," Durability of Building Materials and Components, ASTM STP 691, P. J. Sereda and G. G. Litvan eds., 1980, p. 455. 5. Everett, D. H., Haynes, J. M. , "Capillary Properties of Some Model Pore Systems with Special Reference to Frost Damage," RILEM Bull. New Ser., Vol. 27, 1965, p. 31. 6. Setzer, M. J., "Einflu3 des Wassergehalts auf die Eigenschaf ten des Erharteten Betons," ("Influence of Water Content on the Properties of Concrete"), Schriftenreihe Deutscher Ausschu3 f. Stahlbeton, Vol. 280, 1977, p. 44. 7. Stockhausen, N. , Dorner, H., Zech, B., Setzer, M. J., "Unter- suchungen von Gefriervorgangen mit hilfe der DTA" ("Investigation of Frost Phenomena by DTA"), Cem. Concr. Res., Vol. 9, 1979 p. 783. 8. Dorner, H. W., Setzer, M. J., "Tief teraperatur-DTA-Untersuchungen des Zementsteingefiiges bei unterschiedlichem Hydratationsgrad, " ("Low Temperature Studies of the Structure of Hardened Cement Paste with Different Degree of Hydration"), Cem. Concr. Res., Vol. 10, 1980, p. 403. 9. Stockhausen, N., Setzer, M. J., "Anomalien der thermischen Ausdehnung und Gef riervorgange in Zementstein" , ("Anomalous Thermal Length Change and Frost Phenomena in Hardened Cement Paste"), Tonindustriezeitung Vol. 104, 1980, p. 83. 10. Badmann, R., Stockhausen, N., Setzer, M. J., "The Statistical Thickness and Chemical Potential of Adsorbed Water Films," J. Coll. Interface Sci., accepted for publication. 165 11. Zech, B., "Zura Gef rierverhalten des Wassers im Beton," ("A Contribution to Freezing of Water in Concrete"), Thesis, Technische Universitat Miinchen, 1981. 12. Badmann, R., "Das physikalisch gebundene Wasser des Zementsteins in der Nahe des Gef rierpunktes, " (Physically Bound Water in Hardened Cement Paste Studied near the Freezing Point"), Thesis, Technische Universitat Miinchen, 1981. 13. Stockhausen, N., "Die Dilatation hochporoser Festkorper bei Wasseraufnahme und Eisbildung," (The Dilatation of Highly Porous Solid Materials due to Water Sorption and Ice Formation"), Thesis, Technische Universitat Miinchen, 1981. 14. Rosli, A., Harnik, A. B., "Improving the Durability of Concrete to Freezing and Deicing Salts," Durability of Building Materials and Components, ASTM STP 691, P. J. Sereda and G. G. Litvan Eds., 1980, p. 464. 15. Harnik, A. B., Meier, U., Rosli, A., "Combined Influence of Freezing and Deicing Salt on Concrete - Physical Aspects," Durabil- ity of Building Materials and Components, ASTM STP 691, P. J. Sereda and G. G. Litvan Eds., 1980, p. 474. 16. Podvalny, A. M. , "Phenomenological Aspect of Concrete Durability Theory," Mater. Constr., Vol. 9, 1976, p. 151. 17. Ubelhack, H. J., "Einflu3 der Kopplung der kolloidalen Hydrate auf die mechanischen Eigenschaf ten des Zementsteins - Eine Mo 3 bauer- effekt Studie," ("Influence of Coupling of Colloidal Hydrates on the Mechanical Properties of Hardened Cement Paste - A Mossbauer- Effect Study"), Thesis, Technische Universitat Miinchen, 1976. 166 Fig.1: Calculated chemical potential uof water in a slit pore. (/13/) 167 Fig. 2 DTA signal (97%, 91%, [111) of hep stored 82%, 75%,) as at different rel. humidities a function of temperature 168 Ho W, " J_] mol-gj. 5 ■• 1 ■■ -$j- «30[j/g]^/ avv H o 2 / I H 10 I I I I I I I I I I 20 AW/W [%] Fig. 3: Heat of fusion of water in hardened cement paste as a function of rel . water content AW/W (/13/) 169 ' o' 00 I "... '•••-.. - o • r. E •••-... • o • CD QJ •( — o • • • " CM 1 -M cr CD -o i •-.. o o E (0 o • • V 1 u i- QJ • QJ i— +-> • • -t-J S- i- <\- . o CO 1 03 +-> +-> -O QJ CO r O o T III o o o o o QJ CO in m o in o m 1 V T ^ cvi 5-S CD *3" bored at di f sample a t 1 1 CO » "l° 1 Q- S- • (j -a • • _c • h- .C QJ • • M- -C • S-. O +-> • O ^ \^j C M- •
E • h*. 03 O • • +J -r- o • • • • • . O CM 1 «3" +-> C 03 +-> i— 03 •«- -l-> -a o3 • QJ r™ - •*-» • • 4-> 03 T3 i (0 • • . O o o E S- QJ o • • ■ 5- QJ SZ -C 1— r™ • I QJ 1—^ • • -l-> T3 5|- o (0 1 IS 03 ■1- J5. ' 1 1 ■■ r i i i o o o o o o o u_ m o m m o tf> T~ 1 ^ 1 T- T" 170 • »••,! o c QJ +-> C o o s. cu n3 4-> cu cu rO -a cu i. o +-> D. o T3 +-> u 03 (L) Q. E 03 I/) >> $- ■a O c o 03 1 1 CT> 4-> rtS Q 1 1 4-> «3 •i— <*^_*^ 1- • CO "D CU o ^5 ,_ •r— 4-> E 03 •r— E -a en cu •<— E o CO 1— 3 CM 1 -t-> c QJ 4-> -a • CD o o 1 «3- cu -t-> CT> 171 N° N° vO ,0 ^O o N ° N o^ o^ ^N ^ CD CD ID 00 IT) C^ IT) ^J I O o t CD O LO OO CD ^ LO IT) \; \? ,0 v o O^ O^ C> C> O C^» O LO O CD CD O^ CD & o^ 6 s - CD IT) E c 10 CO o- H yp/ 3 DP to • CD cr S- CD 3 to +-> "O CD fO c i- S- 03 O CD Ql Q. #\ E >> "O a; +-> o> ■+-> •1— 1— 1 — -t-> •r- •1 — c +-> 4- cd S- 03 00 c CO 13 1 — •1 — cd 13 +-> -a 13 •1 — O JCI F •!■■ 4- S~ O +-> •1— to +j E •r-" c 03 -a =3 c >> CD E -O N 03 •1 — CD to a> .£ .c +-> S- 4- O F O Q. O S- -(-> t— 4- Z3 03 O •1— ■a +-> CD TD C 4-> CD CD (13 4-> S~ 1 — 03 CD 13 13 4- 1 — 4- r— 03 •1 — 03 > Q CD • • ID • CD 172 DEVELOPMENT AND INTRACTION OF A CONCRETE ADDITIVE FOR IMPROVED PERFORMANCE AND DURABILITY S. K. Syal University of Illinois Urbana, Illinois 61801 S. S. Kataria Feeders India Consultants New Delhi , India Abstract: Two percent of the additive SLC, developed by mixing equal amounts of alkali -carbonate and black liquor sulphate/sulphite lye (i.e. lignosulphonate-based material) from paper and pulp works in India, can replace the use of gypsum and produce a cement with improved strength, lower water requirements (about 30 percent less) for same consistency, and better durability. When used as simple additive to the normal Portland cement (clinker and gypsum system), it produced in- creased strength and greater durability towards aggressive reagents and decreased the water requirement for the same consistency. The decrease in water requirements is greater for clinker with no gypsum than it is for cement. Porosity, microstructural , and morphological studies suggest that the lower porosity and increased strengths observed in the presence of the additive are due to the formation of Type-III C-S-H gel in the early stages of hydration instead of Type- 1 or Type- 1 1 generally observed in the normal systems. Another important factor is that very small and disorfented portlandite crystals, occur as secondary hydration products only in the presence of the additive, while large and well -oriented patterns of Ca(0H) 2 crystals are observed at the primary as well as secondary hydration stages in systems without additive. In presence of the additive, the mechanism of hydration does not allow the formation of hollow-shells or Hadley grains in the hardened matrix of cement paste, thus the preferential fracture sites are eliminated. The atomistic mechanism explains that the increased durability in concretes is due to denser, more uniform cement matrix and stronger matrix-aggregate bond. Key words: Additive SLC; atomistic mechanism; cement matrix; durability; Hadley grains; matrix-aggregate bond; microstructure; morphology; porosity; portlandite crystals. Introduction: Durability is considered as the resistance of cement to either internal or external physical, physico-chemical, or chemical actions (1,2). In traditional concrete, the matrix is more deformable than the aggregates. The failure surfaces of concrete mostly follow 173 the contour of aggregates rather than the bulk of the matrix. Recently, engineers and physico-chemists have felt the need of joint efforts to probe in the bond between the matrix and the aggregates to arrive at the real basic cause, rather than giving suitable explanations for the experimental observations made at macro level, though it is necessary thatthe model suggested should be in agreement with the experimental observations made. As the bonds stressed directly in tension are only stressed indirectly in compression, the higher strength in compression than that in tension is very well explained by considering that bond plays a major role in the crack propagation and the failure of concrete. The porosity has an important influence on the strength (3,4) and durability of hydrated pastes. All these relationships may be even more complex due to the fact that total porosity, average size of pores, pore-size distribution, and kinetics of pore formation in cement pastes influencing their permeability and penetrability depend to some extent on the chemical and mineral ical composition of cement. The interde- pendence between cement composition, cement fineness and pore structure of the paste is even more complex due to the influence of other variable factors on the paste itself and on concrete, such as w/c ratio, concrete composition, concrete curing, etc. An optimum amount of micropores . seems to be necessary for an adequate resistance of concrete (5). It seems to be clear that the nature and structure of the cement paste, as far as its porosity and chemical composition are concerned, determine its permeability and also the penetrability of salts and different ions in it (6,7). It is agreed that the aggregates are surrounded by a particular matrix, varying in structure, composition and cohesion, between the surface of the aggregate and the matrix of hydrated set-cement, which is called the transitional ring (8). The transitional ring happens to be the most highly stressed part, and the weakest region of the concrete. It is in this region that the first- irreversibil ities will occur; by packing, resulting from microruptures and closing of pores, if it is compressed; and by cracking when subjected to tensile or shear stresses. The cracking, once initiated, then propagates from ring to ring. Not only is a better understanding of the composition and structure of this region necessary, but it is also desirable to improve the structure of this region to have improved performance and more durable concrete system. Many researchers (9-14) have studied the applications of lignin compounds through patended formulations but mostly were limited to the matrix part only as the data was mainly confined to the studies of cement pastes. We have made an attempt to develope from Indian waste byproducts a similar formulation as patented abroad and study the performance of not only clinker pastes but also OPC pastes, mortars and concretes prepared with OPC (15,16,17). Diamond et. al . (18) tried to understand the interfacial region between the matrix and the aggregates in case of Portland cement, and a further effort was made by Maso (8) towards a better understanding of the composition and structure of the transitional ring. As extension of our earlier work (19), in the 174 present paper, we have tried the use of lignin superplasticizer (SLC) aiming at decreasing the ring porosity, diminishing the size and orientation of the crystals developing which would probably have very favorable consequences on the behavior of motars and concretes under mechanical, physical or chemical actions. Materials and Experimental Techniques: A commercial OPC (ordinary Portland cement) clinker was ground to fineness of 3000, 4500 and 5000 cm 2 / g and a portion of these was blended with 6 percent gyp.sum to produce OPC cements of respective fineness and the composition as given in Table- I. Additive SLC was prepared by mixing equal amounts of SL and alkali carbonate (20,21). The physical testing of samples was performed according to IS4031 , while for paste strength studies 2 inch cubes were cast. The consistency of pastes, the flow of mortars and the slump of concretes in presence and absence of the additive, (2 percent by weight of cement/clinker content of which was used in each case), was same in the respective system. The chemical attack studies were performed on the motar samples that were cured for 28 days in water before being immersed in 5 percent sulphuric acid solution. The porosity measurements were made on a GE Mercury porosimeter using the mercury intrusion method, up to a pressure of 60,000 PSI, [thus suitable for pores of diameter 0.003 ym (30A°)], by using the mortar samples in which the hydration was stopped at the selected periods by dipping the samples in acetone held in an ice bath and then vacuum drying of the sample at room temperature (25°C). Results and Discussion: The setting time, compressive strength, porosity and pore size distribution, and effects of acid attack are given in Tables II, III and Figures 1-4. There is a water reduction for same consistency, or flow, of around 30 percent in pure pastes and in mortars using cement: sand as 1:3; and about 20 percent in concretes with cement: sand: aggregates as 1:2:4. (i) Strength and Durability: An increase in the compressive strength is observed in all the systems when the superplasticizer is added. The pastes from cement + SLC systems showed an increase of around 10 percent over the control while in clinker + SLC system paste, an increase of 80 percent is attained at the stage of 7 days of hydration. In mortars, increase in strength lies in the range of 20-25 percent at 7 days of hydration, while in case of concretes, at 28 days of hydration, the increase in strength is marginal in case of cement + SLC system but it is quite prominent, around 50 percent, in case of clinker + SLC system. The porosity of mortars from cement, cement + SLC and clinker + SLC, have shown an appreciable decrease in total porosity. At 7 days of hydration, the reduction in pore-volume being around 35 percent in case of clinker + SLC system, while it is around 20 percent in case of cement + SLC system, as compared to the respective control systems. The durability studies under aggressive conditions of the mortars from cement and cement + SLC systems have shown the latter to have more resistance. The loss in weight of the mortar cubes after being kept 175 immersed in 5 percent sulphuric acid solution for 28 days is 5 percent less, while the compressive strength is still much higher, being 42.6 percent of that of initial as compared to only 25.7 percent in case of control . A 2 percent of SLC is quite optimum as shown in figure 1 as further addition does not show any detectable increase in strength of concrete. Figure 2 shows that the effect of additive SLC is not merely due to its water reducing property, because even the maximum of the compressive strength values for normal mortars (obtained at w/c=.35, and flow is not standard) are lower, and the values corresponding to w/c=.28, are extremely low when compared to strengths attained in the presence of additive at w/c=0.28. The experimental values attained in presence of 2 percent SLC are quite comparable to the theoretical maxima possible in the normal mortar system, if somehow the hydration could be attained at w/c=0.28. Thus, SLCgivesthe higher strengths while retaining the same basic character of the hydrated system. (ii) Chemistry of SLC intraction: An understanding of the complex interactions of lignosulphonate carbonate additive with cement and clinker minerals, and the possible effects of these interactions on the hydration reactions of cement, helps in explaining the increased strength of the pastes of the system, which acts as the hardened cement matrix in case of mortars and concretes. Our views are in conformity with the possibilities expressed by Skalny et. al . (9). SLC super- plasticizers seem to behave like a highly anionic complex involving complex (lignosulphonate-carbonate)~ x ions (where x may be 3 or more) and is a better dispersant than lignosulphonate alone, thus resulting in a very high liguosulphonate concentration in the liquid phase of the hydrating cement just after mixing of water to it. The hydration products of free lime and surface hydrolysis of C 3 A, mainly Ca 2+ , OH" and positively charged hydrated C 3 A, are simultaneously attacked or chemisorbed by the already present ions from SLC solution, namely Na , OH*, CO 2 ", (liguosulphonate-carbonate complex)~ x ion. The gradual formation of CaC0 3 decreases the CA 2+ and C0|~ concentrations and at the same time the adsorption of lignosulphonate complex on C 3 A inhibits its hydration, thus decreasing the probability of the formation of portlandite and ettringite at this very .early (primary) hydration stage of the clinker. In case of cement, gypsum dissolves increasing both Ca +2 and S0£" concentrations. But the gradual formation of CaC0 3 decreases the Ca 2+ and CO 2 " concentrations, the SO 2 " concentration continues to increase as ettringite cannot form due to lack of C 3 A hydration, until the (lignosulphonate-carbonate)' x complex is disrupted by CO 2- removal and hydrated C 3 A- lignosulphonate subsequently reacts with SO 2 " in solution to form calcium lignosulphonate and ettringite. In further processes, nuclei of CSH and Ca(0H) 2 may be formed as discussed by Skalny et. al . (9). The above sequence of the reactions clearly explains that the formation of portlandite and ettringite is delayed and do not form as the foremost hydrated products in the very beginning of the hydration reactions as in the case of normal cements (in the absence of the SLC superplasticizer). Rather these are formed as subsequent products only, the primary products being CaC0 3 , hydrated 176 C 3 A-lignosulphonate and calcium lignosulphonate. Further, CSH formed is of Type-III (Figure 5) instead of Type-I or Type-II usually observed in normal portland cement hydration. The increase in the compressive strength of the cement paste in presence of SLC seems to be due to the decrease in porosity of the hardened cement matrix due to lesser and smaller crystals of ettringite and portlandite, which may have decreased orientation also (this needs further studies), thus decreasing the porosity and increasing the durability of the hardened matrix. Based on paste studies made by Young and Lawrence (22) and Berger (23), we can derive that during the early stage of hydration, areas between Ca(0H) 2 crystals which consist of partially hydrated particles bonded together by calcium silicate hydrate (CSH) represent the high-porosity-portion of the paste. Because of the presence of Ca(0H) 2 , fracture in the hardened paste during this period propagates preferentially through the areas bonded by the CSH phase and around the Ca(0H) 2 crystals. Almost all the fractured grains observed by Barnes et. al . (24) seem to be of the hollow-shell (Hadley grains) variety. In presence of SLC, Ca(0H) 2 crystals are smaller and closer, giving a close packing, eliminating the possibility of space for unhydrated particles or Hadley grains, thus eliminating the preferential fracture sites. Further, the CSH bonding gel is of Type-III in place of Type-I or Type-II, which is stronger and more durable due to its higher polymeric form of silicates as compared to the polymeric forms of silicates in Type-I or Type-II CSH (25). (iii). Matrix-Aggregate Interface: The increase in the strength of concrete in presence of SLC superplasticizer must lead to changes in the "transitional ring", which is possible if porosity of the ring is decreased, which will automatically increase the durability of concrete towards mechanical stresses and chemical attack also. Considering the hypothesis concerning the formation mechanism of the transitional ring (8), let us discuss the increased strength of concrete system in the presence of SLC superplasticizers. When the mixing of cement, aggregate and water takes place, the aggregates become coated by a film of water several ions thick. The thickness of the water film surrounding the aggregates at the time of mixing plays an important role in the orienta- tion of the first generation crystals and, as a consequence, of those of the second generation. The hydration process, for the formation of transitional ring occurs according to Lechatelier's theory and not as in the body of the hardened cement matrix, by formation of a crystal i zed coagulated network. The transitional ring is probably an area of mini- mal concentration and it is in this part of the ring that the largest crystals and the highest porosity will be found. In the water surround- ing the aggregates the first nuclei i to form are those of the hydrated components corresponding to the most mobile ions; ettringite and portlandite in the case of normal portland cement. These happen to be the main products formed during the hydration of pure cement, and since the development of crystals is not hindered, they reach considerable size, larger than in the body of the paste where supersaturation is much higher and space more limited. They form a looser network, the pores of which are progressively saturated with less mobile ions, such 177 as silicon and aluminum ions. A second generation of crystals then develops in the voids left by those of the first generation, and in case of normal portland cement, there is CSH, and the ettringite and postland- ite in crystals of smaller size. The first generation crystals do not develop in continuous layers in direct contact with the aggregate. Diamond et. al. (18) observed the existence, in close vicinity of the aggregates, of a twin layer which they called "duplex film". This consisted of portlandite crystals, the axes of which are perpendicular to the aggregates, and layers of CSH (of Type-I or Type-II). Beyond this film about one micron thick, they observed a highly porous zone containing portlandite crystals, the axes of which run parallel to the aggregates, which are well-formed and rather large in size. Large spaces between the crystals are later partially filled with smaller, secondary portlandite crystals, with their axes perpendicular to the aggregate surface, and with CSH (of Type-I or Type-II) and Hadley grains When SLC superplasticizer is being used in the mix along with the water, the formation of portlandite and ettringite as first generation . crystals is absent, as discussed in the case of hydration of cement/ clinker in presence of SLC. Further, if small amounts are formed, the development of crystals is hindered by highly anionic (lignosulphonate- carbonate)' x ions. So, the large well oriented crystals of portlandite and ettringite forming the porous zone are no longer present, and in their place only small, poorly-oriented crystals of portlandite and ettringite may be formed, as shown by SEM studies (16), which are bonded by CSH of Type-Ill and the possibility of Hadley grains is al- together absent. Secondly, just after mixing the concrete in normal cases, the aggregates are coated by a thick film of water (H + -0H _ ) because of the aggregate's surface charges involving Si 4+ , while the water film is much thinner in the presence of SLC because the anionic complex (lignosulphonate-carbonate)~ x having the maximum mobility, adsorbs on the aggregate surface, thus releasing some of the water. In the water surrounding the aggregate, the first nuclei i to form are those of the hydrated components formed from the most mobile ions which are (1 ignosulphonate-carbonate) -x ; viz. calcium 1 ignosulphonate and calcium carbonate are formed. In this area there is no more preferred orientation of the crystals and hence the porosity is much lesser to that in the normal portland cement in absence of SLC, which leads to the stronger bonds between the set hardened cement matrix and the aggregates. The SEM studies of the matrix in presence of SLC have shown only smaller and more disoriented crystals of portlandite and ettringite rather than the large crystals seen in the absence of SLC (17). The detailed microstructural studies of the transitional rings in the systems using SLC is still in progress and there is need for the confirmation of above hypothesis. Conclusion: At the time of mixing of cement-SLC-aggregates-water, the aggregates become surrounded by a water-SLC film and early hydration mechanism is modified. Around the aggregates, the transitional ring formed is different in its composition and its microstructure from that 178 of the body of the hardened cement matrix; it is no longer very porous and lower in cohesion, as expected in the normal Portland cement in the absence of SLC. So the increased durability in concretes is due to denser, more uniform cement matrix and stronger matrix-aggregate bond. Acknowledgements: The authors express their gratitude to Professor J. F. Young, Civil Engineering and Ceramics Departments, University of Illinois at Urbana-Champaign for his valuable suggestions. Dr. Syal acknowledges special grants from National Science Foundation. References: 1. Calleja, J., "Durability", Proc. VII ICCC (Paris), Vol. I, VII-2 (1980). 2. Helmut, ''Durability" , Cem. Res. Progr. (1977) p. 195. 3. Popovics, S. , "Strength Development of Portland Cement Paste", Prov. VI ICCC (Moscow), Sect. II (1974). 4. Roy, D. M. and Goude, F. R., "Optimization of Strength in Cement Paste", Prov. VI ICCC (Moscow), Sect. II (1974). 5. Richartz, W. and Locher, F. W. , "Cement Research in Germany", Cem. Res. Progr. (1977) p. 303. 6. Ushiyama, S. and Goto, S. , "Diffusion of Various Ions in Hardened Portland Cement Paste", Proc. VI ICCC (Moscow), Sect. II (1974). 7. Mehta, P. K. and Manmohan, D. , "Pore-size distribution and permeability of hardened cement pastes", Proc. 7th ICCC (Paris), Theme VII (1980). 8. Maso, J. C. , "The bond between Aggregates and Hydrated cement paste", Proc. VII ICCC (Paris), Vol. I, VII (1980) 1. 9. Skalny, J., Klemm, W. A. and Javed, I., "Hydration of Cement - Lignosulfonate - Alkali Carbonate system", J. Am. Cer. Soc. 62 (1979) 461. 10. Skvara, F. et. al., "Cement paste and mortar with low water to cement ratio", Cem. and Cone. Res. 10 (1980) p. 253. 11. Odler, I. et. al., "On the combined effect of water soluble ligno- sulfonates and carbonates on the portland cement and clinker pastes", Cem. and Cone. Res. 8 (1978) p. 469, p. 525. 12. Odler, I., Skalny, J. and Brunauer, S., "Properties of the system clinker - Lignosulfonate - Carbonate", Proc. VI ICCC (Moscow), Sec. II (1974) 6. 179 13. Ciach, T. D. and Swenson, C. G. , "Morphology and Microstructure of hydrating portland cement and its constituents", Cem. and Cone. Res. 1 (1971) 515. 14. Diamond, S. and Toledo, C. G. , "The microstructure of low porosity Portland cement paste", II. Cemento 75 (1978) p. 189. 15. Syal , S. K. , "High-Strength Syalex cement and concrete", (i) A poster summary, 7th ICCC (Paris) 1980. (ii) Indian Patent (1979), (Applied for). 16. Syal, S. K. et. al., "Improvements in the Performance and Durability of Concrete", A project report (to be published), 1981. 17. Syal, S. K. et al . , "Development of a Concrete Additive Formulation for Improved Performance and Durability", Summary, All India Seminar on Cement Manufacture, New Delhi (1981). 18. Barnes, B. D., Diamond, S. and Dolch, W. L., "Micromorphology of the Interfacial zone around aggregates in portland cement", J. Am. Cer. Soc. 62 (1979) p. 21. 19. Syal, S. K. and Kataria, S. S. , "Role of Lignin Superplasticizer in Construction Industry", 2nd Int. Conf. on Superplasticizers in Concrete, Canada (1981). 20. Syal, S. K. et. al . , "Evaluation of Various Techniques for Moisture Reduction in Wet Process Cement Plants for Achieving Fuel Economy", A project report, CRI (1979). 21. Syal, S. K. and Kataria, S. S., "Energy Economy and Productivity Enhancement by use of Additives in Cement Production", Proc. Annual Conv. of Chemists, IND-28 (1980). 22. Young, J. F. and Lawrence, F. V., "SEM Studies of the Fracture Surface of Bulk Ca 3 Si0 5 Paste", 73rd annual meeting of the American Ceramic Soc, Chicago, II. (1971). 23. Berger, R. L., "Calcium Hydroxide: Its role in the Fracture of Tricalcium Silicate Paste", Science 175 (1972) p. 626. 24. Barnes, B. D. , Diamond, S., and Dolch, W. L. , "Hollow-shell Hydration of Cement Particles in Bulk Cement Paste", Cem. and Cone. Res. 8 (1978) p. 263. 25. Syal, S. K. et. al., "Examination of Silicate-polymerization in CSH", (in process). 180 TABLE I: COMPOSITION OF ORDINARY PORTLAND CEMENT USED Oxide Ana ilysis (pe rcent) Phase Analysis (pe rcent) CaO 60.4 C 3 S 55.0 Si0 2 26.5 C 2 S 20.0 A1 2 3 4.8 C 3 A 8.8 Fe 2 3 2.0 C 4 AF 9.0 SO 3 2.2 MgO 2.0 MgO 2.0 CS 5.2 Alkalies <1.5 Others <1.0 181 TABLE I I Compressive Strength of Systems with and without SLC (A) Pure pastes having constant consistency 5-7 mm . Sample Code Description Additi' /e w/c ratio Compressive Kg/crr 1 Streng , 2 th Setti Time, ing , Mts 1 day 3 days 7 days 28 days Ini tial Final NP (Control ) Cement ? (3300 cnT/g) Nil 0.28 68 194 304 473 145 300 SNP ii 2 0.20 99 223 335 532 210 235 SCP Clinker „ (3300 ci//g) 2 0.19 181 260 549 800 160 180 SCP Clinker „ (5000 cr//g) 2 0.175 660 780 840 880 105 155 (Kg/cm 2 ) (CC/gm) (Vol. %) 0.40 1 day 247 3 days 367 7 days 400 .70 .060 .049 17.32 13.84 12.59 (B) Mortars (Cement:Sand =" 1 !3)-having constant flow 100+5 mm. Sample Description Additive W/C Compressive Vol. of Porosity Code % ratio Strength Pores N Cement „ Nil (Control) (4500 cnT/g) R Clinker 2 (4500 cnT/g) NM Cement „ Nil (Control) (3300 cnT/g) SNM Cement 2 (3300 cmVg) NM-DS Cement „ Nil (3300 cnT/g) N.P Cement ? Nil (Control) (3300 cm /g( |C) Concrete (Cement :Sand .-Aggregate = 1 :2:4 )-havi ng constant compaction factor 0.80. 0.25 0.40 0.28 0.40 0.28 1 day 3 days 7 days 445 477 502 .037 .035 .029 9.29 9.34 7.34 1 day 3 days 7 days ( 82 184 238 .079 .062 21.28 16.28 1 day 3 days 7 days 126 267 316 .074 .050 19.22 13.13 1 day 3 days 7 days 82 100 168 .069 17.87 1 day 3 days 7 days 68 194 304 .116 30.03 Sample Description Code Additive W/C Compressive Strength 7o Ratio Kg /cm 2 7 days 28 days NC (Control SNC SC Cement 2. (3300 cm /g CI inker 2 Nil 2 2 0.60 0.50 0.50 101 120 181 153 157 228 (3300 cmVg ^L82 TABLE III: Durability Studies of Mortars Against 5 Percent Sulphuric Acid Solution. Mortar Initial Initial Duration Weight Loss Compressive Code wt. of Comp. St. of attack (Percent of St . (Percent Cubes (Kg/cm 2 ) (days) Initial) of Initial ) (ams) NM 834 380 7 5.53 64.7 (Control ) 14 21 16.94 30.40 46.3 33.1 850 384 28 40.50 25.7 SNM 7 4.07 67.1 14 11.63 60.5 21 24.34 44.7 856 348 28 35.37 42.6 NM-DS 7 2.57 81.6 14 8.52 68.4 21 13.70 50.5 28 17.40 48.8 183 o O O O o o o o O CO C\J 00 oo LlJ \ >- cc zs CD CC - <=X. oo oo I— >- O O Q i — . 2! I— ro < -2L a: < U_ * > o >- <: 3: q • i— C\J i 4 i||5uaj|s aAjssajduuoQ o o If) ro t\J CNJ «fr LU LU tvj o: i— ^~ LU Z c Li. O Q> Ll_ C_> — . E Q 1— CVJ a> T*~ O CC LlJ o s: o Li_ LU O a> 1— Q — ^— LU Z £ z: d- >» o s: LO CD l_3 33 1 B* LL. LO O Q 3: lu ** 1— x CM > CO i— ' ZC Li. LU t- a; o: (— o ■o OO LL. CX> X> 1 1 1 * LU (.) 1— < > «4— 00 i—i o oo i— LU i—i c o cc a CO q. a 2: <: O ^_ o o <_J LL. o c a> =t oo ro Q ;3 1 o (J o 00 — ' c CNJ t— o CC Z 33 O \84 6/ DO NOIlVai3N3d UJ en 6/33 N0llVdl3N3d 185 r«« i- L — o o * o o h- s "T" s T (f/33) 4- twrnoA BM4 T o 186 cc — i-t UJ < a h- < O x 3: 3 ^ cc ae a o X X in m -a -a c c (0 (0 uj a x 2 < <=> UJ 2 in < ur> u_ 3 •- *- cn m -a xj c c < u N 187 THE DURABILITY OF PREFABRICATED REINFORCED CONCRETE EXTERNAL WALLS AND CLADDING IN BUILDINGS J . Kami Professor, Matwei Gunsbourg Chair in Civil Engineering Technion, Israel Institute of Technology Haifa 32000, Israel. Abstract: In 1964, an investigation was initiated at the Building Re- search Station, Garston, U.K. to assess the durability of prefabricated r.c. external wall panels used in several industrialized building sys- tems throughout the country. The systems and buildings covered by this survey were selected with the object of laying the foundation for a durability study of dwelling houses subjected to atmospheric weather- ing - the most common form of exposure. As damage does not usually be- come visible within a short period, a long-term study was decided upon, including periodic visits to building sites for thorough and systematic inspection. To compare the degree of exposure, the Index of Exposure to Driving Rain was used. This index is based on the annual rainfall and mean wind speed. Accordingly, the country was divided into three zones: Sheltered, Moderate and Severe, with each zone further sub- divided into inland influence and coastal influence. The chosen panels were subjected to non-destructive testing at the factory and both to non-destructive and destructive testing at the B.R.S. Samples of the aggregates and cements used were also tested. The detailed information about the quality of the concrete and elements, reliable results of the tests carried out, and the information acquired in fundamental laboratory research - will, it is assumed, help to find a laboratory method for assessing the durability of concrete. With such a tool to hand, combined with available information on the exposure con- ditions at the locality in question - it is hoped that an indicator can be worked out for the durability of reinforced concrete buildings. Key words: Atmospheric weathering; concrete in pre-fabrication; driving rain; exposure conditions; joints between wall panels; long-term dura- bility; method for assessing the durability of concrete and of r.c. buildings; quality of concrete; visual inspection. The severe world-wide housing shortage, and the lack of skilled construc- tion labour, gave rise to the idea of factory-made houses and thus led to industrialization of the building industry. By this means, with mecha- nization applied to fabrication as well as to assembly and erection, several advantages are gained. 188 Concrete - the most common material in the building industry - occupies the same position in pre-f abrication. Future trends inevitably involve still more industrialization and development of larger individual units [1]. In these circumstances, the need was felt for more informa- tion on the long-term durability of concrete buildings constructed by these techniques. Accordingly the present survey was undertaken at the Building Research Station, Garston, U.K., in 1964. Durability is undoubtedly the most important property of buildings and structures. To quote Prof. Campus L^ J; "Durability should be taken as much into consideration as strength calculations in order to ensure the efficacy, safety, economy and good appearance of a building." Yet, while the strength of a concrete element or structure can be predicted, this is not yet the case with durability. In fact, the fundamental re- search work in this field, currently carried out at numerous institu- tions, and the information provided by small-scale field trials - can- not possibly be conclusive. Only full-scale tests on buildings, where all influencing factors act simultaneously, are capable of yielding the complete picture. The durability of a reinforced-concrete building depends in general on the properties of the concrete, on the depth of cover provided for the reinforcement, on the quality of the design and workmanship, and fin- ally, on the conditions of exposure to which the building is subjected. The durability of a prefabricated building is determined also by the quality of the joints between wall panels or cladding panels and col- umns, etc. which must be watertight and weatherproof. Degradation or deterioration mostly sets in under combined action of internal and ex- ternal factors, characterized by Lea and Davey[3], As the survey was concerned inter alia with the effects of atmospheric weathering, the main factors to be considered were air, moisture, wind and rain - especially "driving rain", as well as sea-water salt-spray driven by wind in coastal areas, and air pollution (acid gases) - aggravated by rain or humidity - in industrial areas. The main steps to be taken for inhibiting the deterioration of concrete and corrosion of embedded steel are: The concrete should be made from sound ingredients in correctly designed mix proportions; additives should entail no danger of corrosion and be used judiciously. The con- crete should be uniform in composition, and possess sufficient workabi- lity to ensure full compaction on placing. The concrete should have adequate compressive and flexural strength, and sufficiently low ab- sorption. Sufficient cover should be provided for the reinforcement. Good workmanship should be ensured in placing and compacting. Concrete elements should be well cured and mature on delivery. Surface finishes applied at the factory should not require periodic renewal. This survey of concrete construction in Britain was undertaken as the first stage of a long-term durability study of dwelling houses subjec- 189 ted to the most common mode of exposure - atmospheric weathering. The survey was confined to prefabricated concrete external walls and cladd- ing and the systems and buildings covered by it were chosen accordingly. The emphasis on prefabricated elements was motivated by the following considerations: Quality control is better in an industrial plant, en- suring higher uniformity in both the production process and the product. Information on the history of the product is also available and samples are easier to obtain. Buildings made from prefabricated units are often erected in groups, thus providing a larger "population" and more reli- able information through site inspection. Factory-made products are de- livered to different localities, so that the influence of different ex- posure conditions can be ascertained. For the purposes of the survey, the U.K. was divided into zones accord- ing to level of exposure to driving rain, namely: Sheltered - where the driving rain indext^J is 3m 2 /sec. or less. Moderate - where the index is between 3 and 7m2/sec. Severe - where the index is 7m2/sec. or more. Each zone was further subdivided into sub-zones: Inland - beyond 8 km of the sea or a large estuary. Coastal - within 8 km of the sea or a large estuary. The survey was planned to include visits to prefabri- cated-element factories and to building sites, as well as tests at B.R. S. laboratories on elements and materials obtained from the factor- ies. The information thus gathered was to be embodied in technical records. Four systems were included in the survey: Systems "A" and "B" (multi-storey buildings) - "sandwich" panels. System "C" (low-rise "normal" housing, 1 to 5 storeys) - "hollow" panels. System "D" (multi- storey buildings) - "waffle" panels. The information collected at each factory concerned the following: System; Manufacturer; Contract; Type of exposed surface, dimensions and weight of panels; Production of pan- els: batching and mixing, mix proportions, water /cement ratio, slump, admixtures, reinforcement, moulds, placing and compaction, curing and finishing; Quality control. From each contract, three panels were sel- ected for full examination at the B.R.S., and three more for non-destruc- tive testing at the factory. Samples of the different aggregates, brick slips, cements and ready-mixed lime/sand for pointing were taken for examination at the B.R.S. The chosen panels were subjected to non- destructive testing at the factories: Relative strength (Schmidt hammer) [5J. Moisture content (micro-wave moisture meter). Depth of cover (covermeter) t"J . At the Building Research Station the panels were sub- jected to non-destructive testing (Schmidt hammer, flatness, accuracy of dimensions). Afterwards, they were cut into segments and covermeter readings were taken. The pieces thus obtained yielded a large number of cores, beams and prisms. The specimens were used for the destructive tests: the cores - strength L?] and absorption, the beams - strength, the prisms - shrinkage. One intact segment of each type was placed on the natural exposure site and one reserved for moisture-content calibration to]. Site visits took place approximately 2 years after completion of the buildings and after approximately 8 years after completion of the buildings. 190 The mixes used for the panel concretes proved to be fairly lean. Whe- ther or not the grade of concrete produced will meet the particular re- quirements of durability, will only be seen from the long-term perform- ance of the panels. The Schmidt hammer test yielded fairly consistent results on similar panels. As is well known, this is a surface hardness test and, although concrete shows a certain degree of correlation be- tween crushing strength and impact hardness !■" J , this relationship is not linear. The test did indicate differences in surface hardness from one type of panel to another. The covermeter test similarly yielded consis- tent readings on similar panels and indicated differences from one type to another. The thinnest cover recorded was on the sandwich panels. Noting that the outer skin of the sandwich panels is non-loadbearing, the findings of the survey compare favourably with the depth of cover recommended for other types of precast reinforced units L-LOJ. The dimen- sional measurements on the panels revealed small random deviations from specified dimensions, and the question here is whether or not such de- viations could be tolerated. In practice, all systems under considera- tion resort to storey-to-storey adjustment and the ranges of tolerance are much wider than the measured deviations. The shrinkage and moisture migration of the concretes proved to be normal for fairly lean mixes of this type. The absorption figures compare favourably with those given in BS 368:1956 (precast concrete flags) and BS 340:1963 (precast conc- rete kerbs, etc.). The chemical analyses of the cements and the vibra- ted mortar test indicated that all the cements met the requirements of BS 12:1958 (Portland cement - ordinary and rapid-hardening), except that the rapid-hardening cements in two cases slightly exceeded the permis- sible SO3 content. The strength results on cores and beams showed variability from one panel to another, even on similar panels. Indivi- dual strength values on specimens from the same panel also showed wide scatter. This scatter is largely the consequence of non-uniform com- paction, especially in the waffle and hollow panels. In the case of the sandwich panels, the picture is obscured by cracking of the seg- ments as they were levered off. In general, the compressive strength values are of the order to be expected from the results of the vibrated mortar test on the cements. The scatter in the strength and absorption results indicates the need for a larger "population" of specimens, but any enlargement of the programme would have required considerable ex- pansion of the facilities available at the BRS for handling and cutting large panels. The sieve analyses of the aggregates showed them to con- form to the requirements of BS 882:1965 (Aggregates from natural sources for concrete) except for one or two instances of minor deviation. Visual inspection of the marked panels and others during the site visits showed that, except for cases of water penetration through joints, the panels themselves were in good condition after approximately 8 years of service. The potentially serious trouble experienced with the joints may be solved by providing heavier perimeter reinforcement at vulner- able points. Since failure of these joints is serious and may create conditions unfavourable to the long-term durability of the panels, it would appear advisable to include testing of the various jointing sys- tems in any future work. To find out if, and to what extent, the spe- 191 cific conditions of a locality affect the durability of panels - the site visits should be continued for many years, as damage does not usually become visible within a short period. In general , it may be said that sound materials were used in the manu- facture of the panels and that the quality of the concrete was as could be expected from the mix proportions employed. The depth of cover also seemed to be adequate, but some shifting of the mesh was evident. The site visits showed the durability of the exposed-aggregate finishes after approximately 8 years of natural exposure to be satisfactory, al- though the brick-slip tiles on some of the panels appeared to be losing the surf ace- applied sanded finish. The pointing material used for these tiles seemed rather suspect, and the possible consequences of the frost action could be serious if water penetrated behind the tiles. To conclude : The detailed information about the quality of the concrete and elements, reliable results of the tests carried out, and the infor- mation acquired in fundamental laboratory research - will, it is assumed, help to find a laboratory method for assessing the durability of con- crete . With such a tool to hand, combined with available information on the exposure conditions at the locality in question - it is hoped that an indicator can be worked out for the durability of reinforced concrete buildings . Acknowledgement is due to the Director, Building Research Station, Garston, Watford, U.K., where the investigation was carried out in 1964 during author's Sabbatical leave from the Technion. The author wishes also to acknowledge Mr. R.W. Sharpe's valuable contribution as co- ordinator of the results. 192 REFERENCES 1. Lea, F.M., Development in Building Materials, Prospect for the Fut- ure. Chemistry and Industry, Oct. 27, 1962. 2. Campus, F., Pathology of Buildings. Corrosion of Concrete and Rein- forcement. B.R.S. Translation from the French, April, 1958. 3. Lea, F.M. and Davey, N. , The Deterioration of Concrete in Structures, D.S.I.R., B.R.S. , 1949. 4. Lacy, R.E. and Shellard, N.C., An Index of Driving Rain, Meteorologi- cal Magazine, Vol. 91, 1962, and An Index of Exposure to Driving Rain B.R.S. Digest (second series) 23, 1962. 5. Orchard, D.F. , "Concrete Technology", Contractors Record Ltd., London Volume 2, pp. 182-183, 1962. 6. Orchard, D.F., "Concrete Technology", Contractors Record Ltd., London Volume 2, pp. 451-452, 1962. and Halstead, P.E., "The Covermeter - apparatus for measuring the depth of reinforcement below the surface of hardened concrete", Cement and Concrete Association Research Note No. Rp.5. 7. Wright, P.J.F., "Comments on an indirect tensile test on concrete cylinders", C. and C.A. , Magazine of Concrete Research, Volume 7, No. 20, pp. 87-96, July, 1955. 8. "The Micro-wave method for determining the water content of walls" , BRS Note No. C.657, May 1959. 9. Gaede, K. , "Non-destructive Testing of Concrete by the Steel Ball Impact Method", C. and C.A., Library Abstract, Number 3, 1953. 10. "The structural use of pre-cast concrete", British Standard Code of Practice, C.P. 116:1965. 11-17. Internal notes of the B.R.S. 193 TESTING AND ESTIMATION OF DURABILITY OF CONCRETE AND RADWASTE CONCRETE FOR LONG-TERM STORAGE S.E.Pihlajavaara Technical Research Centre of Finland Laboratory of Concrete and Silicate Technology SF - 02150 Espoo 15, Finland Abstract: Radioactive waste barriers and other problem-waste barriers are systems including concrete (cement paste), reinforced concrete structures and other building materials. These systems, which have very high requirements especially for long-term durability, are and will be a challenge to the building industry. This research on a special concrete system with demands on service life prediction may produce fruitful ideas for other building applications. The first step in the process of disposal leading to final storage of many harmful wastes, including especially the radioactive wet wastes of nuclear power stations, is to solidify the wastes with portland cement, for example. In the paper, the solidification of low and medium radio- active nuclear sludges or wet wastes^like ion-exchange resins and evaporator concentrates , with portland cement into a product called radwaste concrete has oeen studied. The radwaste concrete will be cast in a container to be designed for that particular purpose. To satisfy long life requirements a reinforced concrete cylinder with special linings seems to be a suitable container. The solidification of the nuclear power station wastes with cement (apart from fuel waste) and the durability of the solidified product called radwaste concrete and the disposal system leading to final storage has been studied at the Technical Research Centre of Finland in 1976 - 1981. Simulants, i.e. nonradioactive wastes, have been used in the experimental studies. A brief review of the disposal and testing principles, of test methods and results, and of life prediction of the system radwaste concrete plus container with a proposal for the applicable multi -barrier system, i.e. radwaste concrete proper in a reinforced concrete container with special linings, will be presented. Dropping tests of the containers, permeability and leaching studies of radwaste concrete, among others, will be carried out later in 1981. Key words: Concrete; durability; life prediction; long-term storage; radioactive wastes; radwaste concrete; testing; waste barriers. 1. General disposal system for wastes of nuclear power stations; Interrelation between waste content, strength and durability of rad- 194 waste concrete, durability of container, transportation risks, safety of storage, and radiation risks tolerated by the environment: Table 1 indicates an international system for the safe disposal of the cement-solidified low and medium radioactive wastes (wet wastes such as spent ionexchange resins or evaporator concentrates). The cements, e.g. portland cement, are usually negatively susceptible to the wastes added and, therefore, the greater the amount of the waste is the smaller the strength and durability of the solidified product or rad- waste concrete is. Poor radwaste concrete needs a strong container and a reliable storage depending on the safety requirements. 2. Testing and evaluation system for the durability of radwaste con- crete: The objective is to determine the long-term strength, dura- bility and safety, i.e. physical and chemical bonds within the concrete and their behaviour under the anticipated degradation conditions ranging from modest to severe. These types of test results, together with other relevant technical factors as well as with social criteria, will be needed in the establishment of the final choice of radwaste concretes, their possible containers and their storage, in which the long-term safety is of primary importance. Table 2 shows briefly the two-phase testing system employed in our research. 3. Conclusions from experimental results and general experience; Applicable disposal system of radwaste concrete in a reinforced con- crete container with linings: Only the radwaste concretes with too small uneconomical amounts of wastes passed the phase I tests (Table 2). Due to this and other considerations a simple system for technical application has been introduced in Table 3. The linear corrosion model has been adopted as a basis for service life calculations. For example, if the corrosion or deterioration rate starting from the outside in the 100 mm wall of the first-class concrete container is 0.1 mm/year and it is still serviceable after the deterioration of the outside layer of 50 mm, the service life of this part of the container is 500 years. LITERATURE: 1. Pihlajavaara, S.E. & Aittola, Jussi-Pekka, Radwaste concrete: solidification of nuclear wastes with portland cement. A progress report of study May 1976 - March 1978. Espoo 1978. Nordiska kon- taktorganet for atomenergifragor, NKA report AQ (79)3. 34 p. (In English). 2. Pihlajavaara, S.E., An analysis of materials durability and its testing with reference to radwaste concrete. Espoo 1979. Nordiska kontaktorganet for atomenergifragor, NKA report A0 (79)2, 9 p. (In English). 195 UJ I- LU OT U Z o o UJ I- < a < UJ CO > y> *• -I < UJ I/) _J ffi o < 1- Q UJ U. < 10 UJ 10 UJ > (X UJ UJ z o tO UJ 10 >- > 10 o Z UJ UJ t- Ul H i- > > z UJ Ul < < X UJ 0£ X K or (J or u o X o UJ 3 a UJ X z z o a ar o o o o o i- o o ar or W or Q Z < V _i UJ Z 0_ or. o UJ z 3 »- UJ o a < a or o UJ -1 _l _i -i 1- UJ o a: UJ z 10 i/> 17) to 10 ~~ < t— to UJ z o Z h- I H to < H a UJ z V- z to or Z o t/> _l < 1- u z o or or o RESISTANT METAL, INFORCED CONCRE o a. to tc UJ z < 1- z i- z UJ z UJ or a X X -I < 1- Ul z o or or o o o z o or UJ z 1- z < 1- to to o Q UJ U or o u. z u o z o or i- to > or is z z _l X 1- = z < ar i- i- z UJ or o UJ i- UJ Ul UJ UJ o or to ar or > i u. u. z UJ a o o I 1- to z a UJ i- < u UJ > < »- < 1- z UJ UJ t- UJ a: o z o X 1- i z 1- z UJ z UJ or -J -J < z -j < z or o UJ < or UJ a o z Ul i- < or UJ a o z 1- < UJ ar (9 o Ul to UJ ffi t- -I a Z o V a z UJ < o z -I in z < or or o a. a: < UJ a to UJ U) J- > i— 10 x UJ t/> X UJ (/> OT 10 J- 1- < U UJ o o m o z to a: z or «- in in o 1- a. Ul UJ E A A A A UJ o z z or t- — • UJ or u z o u UJ o i- u. z UJ 1- 10 < z UJ or 3 o t/> < z a. o UJ o < ~ O Q a u. $ z £ z ar < UJ o UJ < < < or 3 o (K 0T UJ > X a X — 10 o X WATER THERMA FREEZE LEACHIN IMPACT RADIATH OTHER or UJ i- < -1 < Z or UJ UJ N UJ UJ or or UJ i- < (SEALED 28 d UJ z Ul -J ffi < 1- z to t/> £ i- u. (/) U) to < _l •-« » H n 3 UJ o or a: UJ x o i i- a y- LU UJ 7 O i- =3 O < z z O III j < CO > o 1- ar UJ 1- z 3 z > to z or •~ UJ UJ 10 *: z UJ < 1 i- < ( > UJ CD z o > o -I o $ -1 III < i I JL l~ to 1- l h- u. UJ z o N UJ z Ul or to UJ a Ul Ul CC u. 3 a to UJ o UJ i or i- UJ >- _j Ul < _i a. o z < to CO ^ z < < ar UJ a: to 3 _l UJ UJ a z z o z < lO UJ > or UJ ar 1— »- I UJ < z 1- h- t- o < i a o z to o < cc 7 z o -1 cc U. o UJ I Ul I- -I O 1- Ul T u. < Q X o o _l Z I— 1- 10 < o 1- 1— o o < T z UJ z o 1- 1- o > UJ tr Ul z UJ o -1 o UJ U) a. or UJ >- Q Ul 5 o ar UJ T d a UJ z 1- UJ z > z < ar o ar > < z a u. i- z V- UJ »- < _i or in < X ID X CD Ul 1- O 1- i- < or z Q III Ul 3 o Ul rr _l Q z i < ffi UJ _) -" UJ to to UJ 1- u> 10 o a. I 1- Z -I < ar UJ UJ or a. III to < > u> X 1- h- X UJ to t- -^ — — — . X X X X X X 196 TEST PHASE I SIMPLE PRELIMARY TESTS DURABILITY INDICATORS - STRENGTH -CRACKS -DISINTEGRATION { ) 1. SEALED CONDITION -BASIC HYDRATION (HARDENING) CONDITION OF CEMENT -BASIC VIRGIN CONDITION BEFORE TESTS PROPER W -NO ENVIRONMENT EFFECTS (NO MOISTURE EXCHANGE. NO C0 2 EFFECT) -SIMULATES THE INNER PART OF A MASSIVE CON- CRETE BLOCK II 2. WATER TEST: UNDER WATER AT 20*C -SIMULATES SIMPLY THE WET AND MOIST ENVIRONMENT TO WHICH THE CONCRETE MAY BE SUBJECTED DURING ITS LIFETIME -LOWER WET STRENGTH INDICATES W AT E R- 1 NDUCED CHANGES, LESS CHEMICAL BONDS. TENDENCY TO SOLUBILITY, LESS FROST RESISTANCE. ETC. 4 3. THERMAL TESTS IN HEATING CHAMBER 105*C -SIMULATES SIMPLY THE ELEVATED TEMPERATURE AND SUDDEN TEMPERATURE CHANGES INCLUDING DRYIN G ( TE MP E RATU RE AND MOISTURE GRADIENTS) TO WHICH THE CONCRETE MAY BE SUB- JECTED DURING ITS LIFE TIME -SIMULATES THE ELEVATED TEMPERATURE (ABOUT 100*C) CRE- ATED BY THE HYDRATION OF CEMENT IN MASSIVE BLOCKS WITH CEMENT-RICH MIXES. m SATISTACTORY DURABILITY NO SIGNIFICANT STRENGTH REDUCTION. CRACKS AND/OR DIS- INTEGRATION IN THE TESTS ABOVE TEST PHASE II MORE SOPHISTICATED DURABILITY TESTS (VARIOUS TYPES, MANY CYCLES) AND DETERMINATION OF PROPERTIES REL- EVAN7 - TO THE PERFORMANCE OF THE CONCRETE ( E.G. LEACHING) TABLE 2. PRINCIPLE OF DURABILITY ASSESMENT OF RADWASTE CONCRETE PROPER 197 Z 7 > < o o _l z tO < 1- a z X _> < a. a: _! z UJ z X o LL II < or O z < o UJ UJ o to o < 01 z QC to h- LLJ X Z z \- o III UJ UJ UJ h- £ t£ h- (/) UJ CJ V or Z t- 10 o < 3 o IU a a T UJ UJ UJ 1- or t- 7 10 Z >- < < i- =t 1- UJ Q 111 o U. < O CO < cc z 01 IS) u. o 1 3: o H- JO £C < C^ UJ ^. O a> 1- o *— 1 1- < -J >- CC o z o < O o II h- _l _J < < to or UJ IIJ to o X 1- < o CO 1- 1U _l 1- < rr o u _l 10 o ft < 1- z o a. cc UJ 1— < UJ UJ x o o z o UJ X t- o _l cc UJ UJ to 1- to z < < U. a Q o a < z z 1- 111 to < z < x CO o or X > UJ o o -I 1- o < ai UJ UJ z or X cc < o 1- cc o O o CO o -J A o z UJ cc >- to UJ > 00 00 UJ cc a. X o o >- o UJ h- LU cc o z o o Q UJ o or o UJ z I o H — >■ i- i- < CD a: < o cc Q o UJ n _i to < u. cc o Ul UJ a < UJ CD Z UJ cc X o o UJ w CC _| t- < to z o z iC fM _ °c _l UJ — I- ffi < < £ cc to to < cc UJ o 1- rs| 00 Q 1- _l < z cc ui — z a to o > _J I/) cc _J o Ui < cc 10 5 cc a IT a UJ lit u H 7- or < < X < 1- UJ 1- Z z. W o _J IUUI o CK O LL < Q UJ o z z o o UJ cc a < cc a z UJ _l 1C o < cc o o V cc < z < — o X - cc < UJ >- X X c or o UJ I- < cc z O o cc cc o o to or < o u. cc UJ < > X X o © < X UJ to o a c Ml O < S § UJ X UJ u, si UJ Q > O < UJ to or S a. to or "~ ,, v X 1— 5C Al z to > < Q to o to z 1- fN UJ < to cc > z 1— LU to Q cc >- III Q a < () cc a z UJ < a a UJ r* z _) Ul < < cc UJ 1- Ul UJ ti- to to H ll! 1 < cc UJ o Ul ^ UJ < -J a. z CO cc x => UJ < o o X > < X UJ z cc <) II cc 3 m o u. o r* 198 CORROSION OF REINFORCEMENT IN CONCRETE BRIDGES AT DIFFERENT AGE DUE TO CARBONATION AND CHLORIDE PENETRATION A. Volkwein and R. Springenschmid Building Material Institute, Techn. Univ of Munich, Fed. Rep. of Germany Abstract: More than 20 reinforced and prestressed (post- tensioned) concrete bridges at an age of 12 to 80 years have been investigated. While carbonation of concrete (be- tween 1 and 60 mm, at cracks sometimes more than 100 mm deep) did not cause severe corrosion, if rain or splashing and seeping water had been kept away from the construction, the influence of sucking water is mainly dangerous espe- cially if solutions of deicing salts could penetrate into the concrete. Diffusion tests showed that carbonated con- crete can have larger pores than concrete of high alkalini- ty. - Chloride penetration (up to 60 mm deep) depends on the kind of water attack and the exposure situation to the driveway. In concretes contaminated by chlorides and dete- riorated, accumulations of ettringite were observed partic- ularly in cracks and pores or around aggregates and rebars. As sulphate content had not been increased it was concluded, that chlorides could cause some chemical reactions in addi- tion to the formation of Fr.iedel's salt or similar and that transportations of soluble cement constituents are possible if the concrete is frequently wetted by salt solutions. In these cases corrosion of reinforcement was usually more severe while corrosion in sound concrete was not always ob- served although chloride contents around rebars did exceed more than Q.k'% of cementweight . Key words: Carbonation of concrete; chloride penetration; concrete bridges; corrosion of reinforcement; ettringite in chloride contaminated concrete. Some reinforced and prestressed (posttensioned) concrete bridges at an age of 12 to 80 years had to be demolished and replaced, altered or repaired, generally because of traffic necessities rarely because of deterioration or cor- rosion. More than 20 of these bridges have been investiga- ted on carbonation, chloride penetration, deterioration"' of concrete and corrosion of reinforcement (1). 199 The carbonation depth was measured by phenolphthalein to in- dicate the pH-value of the concrete pore water (above 9 resp. below 8). Below surfaces which had always been wetted by rain well consolidated concrete was carbonated to a maxi- mum depth of only 1, rarely 20 mm. Usually there was no cor- rosion if the carbonation did not reach the reinforcement. But in several cases, especially at vertical rebars in more porous concrete, carbonation was found (less than 1 mm) just around the rebar while the main border of surface car- bonation had not reached it. In these cases the steel was often heavily corroded probably because water could seep a- long the rebar dissolving calcium hydroxide out of the sur- rounding concrete and reducing alkalinity. As expected, at surfaces which had been permanently dry ex- cept for the influence of relative humidity carbonation was generally deeper, sometimes more than 60 mm. From Fig. 1 one can conclude that the measured carbonation depths do not clearly depend on age or compression strength as a pro- portion to concrete porosity. In place of that it seemed that besides the actual moisture content and mode of curing after placement of concrete the kind of cement that had been used influenced the carbonation progress, too 3 e. g. an 80 year old concrete made with Roman cement showed a maxi- mum carbonation depth of only 1 mm although this concrete had been poorly consolidated. In other cases 30 to 50 year old concretes were carbonated up to 100 and 200 mm at cracks or at regions of poorly consolidated construction joints. On the other side also bending cracks with crack withs up to 0.3 mm have been found showing no deeper carbonation than the uncracked concrete. In these cases of dry and carbonated concrete usually no severe corrosion of reinforcement was found. On the other hand steel had been corroded severely if seeping water could wet the carbonated concrete from time to time, e. g. beneath a leaking expansion joint of the bridge deck. If the concrete cover had been cracked - approximately 20 years ago - and had not fallen off for a long time crevice corrosion occured with corrosion losses up to H or 6 mm although no chlorides did support the corrosion process. This is a higher corrosion rate than in case of unprotected steel in rural or industrial atmosphere. To study the pore structure of carbonated concrete in com- parison with non-carbonated one some diffusion tests have been done. From a drilled core 30 mm thick slices were cut and tested on diffusion of water vapor through a gradient from 98 to 65 % r. h. It was confirmed that carbonated con- crete was conducting vapor about twice as well. Another 200 test showed that this difference can hardly be caused by in- sufficient curing. A 10 mm thick concrete slice was cut rec- tangularly to the surface. The sample was tested on capilla- ry water absorption parallely to the surface. It could be seen that the water - coloured by potassium permanganate - did rise about twice as fast in the carbonated concrete, Fig. 2. Just below the border of carbonation one can see a smaller capillary time rate probably because of concentra- tion of calcium hydroxide, as the theory of P. Schiefil (2) postulates if a final carbonation depth has been reached. From the twice as high capillary rate of carbonated con- crete one can conclude that the pore sizes had become larger as a result of carbonation. This observation is also con- firmed by investigations by a scanning electron microscope, however it appears to be in opposition to the mass increase as a result of binding carbondioxide . - The larger pore sizes of carbonated concrete can be reasonable for lower equilibrium water contents and low corrosion rates of re- bars in case of dry concrete. Otherwise water can infiltra- te carbonated concrete more easily increasing corrosion rates . In order to maintain roads in winter certain amounts of de- icing salts - usually sodium chloride - have to be applied to the road. In relation to the amount of rain water or melting snow and ice chloride solutions of different con- centrations have had an effect on the components of bridge structures for the last 20 years, Fig. 3 (3). Chloride pen- etration into the top of bridge decks was negligible if the water proofing has been well in function. In cases of un- derflow of the sealing penetration depths up to 50 mm were measured in sound concretes of high compression strength, more than 90 mm in partially frost deteriorated concretes. Large depths of 30 to 60 mm were also measured at columns, abutments and curbs close to the road in the zones of splashing water without concrete having been visually dam- aged, Fig. 4. The underside view of superstructures had been contaminated by chlorides as a result of sprayed fog thrown up by vehi- cles (penetration depths up to 30 mm, sometimes similar to carbonation depths, see Fig. 5) • Pitting or severe corro- sion of reinforcement in these areas was not so often ob- served, although chloride content around the rebar some- times did exceed more than 0.4 % of cementweight . On the other hand weak pitting corrosion sometimes was found in non carbonated concrete with chloride contents less than 0.4 %. Repeatedly these rust formations were situated at larger pores in formerly well consolidated concrete. 201 Severe corrosion (pitting as well as uniform corrosion) was found if the concrete had been wetted frequently by rain or splashing water or from time to time by seeping water, par- ticularly in porous or poorly consolidated concretes with chloride contents of more than 0.2 %. It was not at all possible to define a certain threshold of chloride content causing corrosion. Corrosion rather more depends on the influence of moisture and imperfections of concrete quality. In concretes highly contaminated by chlorides and already deteriorated remarkable accumulations of needleshaped crys- tals were found particularly in cracks, pores and around aggregates, Fig. 6. These crystals were identified as com- pounds of mainly ettringite and less thaumasite (4). As sulphate content had not been increased it could be con- cluded that chlorides did cause the formation of ettringite by the natural sulphate of cement. In tendon ducts containing chloride contaminated injection mortar, ettringite was also observed around surfaces of ten- dons and metal sheaths . By these changes of concrete structure and hardened cement paste one can see that some chemical reactions - additional to the formation of Friedel's salt or similar - and espe- cially transportations of soluble cement constituents are possible, if the concrete or mortar is frequently wetted by salt solutions. In these cases corrosion of reinforcement usually was severe. It can not be said whether these ettringite crystals had caused the deterioration of con- crete or had grown up in already opened cracks . But it has to be expected that these compounds containing sulphate were supporting the corrosion of steel besides the influence of chlorides . 202 REFERENCES 1. Springenschmid, R., "tiber die Dauerhaf tigkeit von Bau- werken aus Beton und Stahlbeton", Zement und Beton, Heft 5, 1976 2. Schieftl, P., "Zur Frage der zulassigen Riftbreite und der erf orderlichen Betondeckung im Stahlbetonbau unter beson- derer Beriicksichtigung der Karbonatisierung des Betons", Deutsch. Ausschufi f. Stahlbeton, Heft 235, 1976 3. Volkwein, A., "Eindringen von Chloridionen in den Beton von Straftenbrucken" , Strafie und Autobahn, 31. Jahrg., Heft 4, 1980 4. Volkwein, A., "Ettringit-ahnliche Phasen in altem Zement- stein und Beton", Tonindustrie-Zeitung (TIZ) 103. Jahrg., Heft 9, 1979 203 E ~ 80 £ 60 o csi C o o c o jQ o o 40 20 o CO o I I I I s; 12 22 m A fl ! II i i I I I _ • o T LO I I I 'I I I ! max. mean values | min. o CO I o 39 50 67 80 age ( years) (square root scale) Fig. 1 — Carbonation depths of always dry concrete in rela- tion to the age of different bridges (the numbers in brackets indicate the compressive strength in N/mm2) . 204 iv—" tmii V W**tV*~tlr***ir d ' t Fig. 2--Different capillary rise (dark coloured) of carbon- ated (right part) and non-carbonated concrete (left part of the sample) after 7 days capillary water ab- sorption test. (Right edge of sample = concrete sur- face; lower edge = beginning of capillary rise). 205 superstructure column or. abutment surface water seeping water splashing water sprayed fog Fig. 3 — Chloride solutions acting on a bridge structure. 206 C71 QJ c_ LJ C o LJ >% >5 A 0.4- 0.3- 0.2- i L L 0.1- 006 compr. strength (N/mrn2)" age (years) consolidation 45 35 poor 70 21 good 100 n a 'O^. ( same motorway) 50 100 penetration depth (mm ) Fig. 4--Chloride penetration into concrete close to the road as a result of splashing water. 207 C o "S-C c±E -V J O d. .a ai 1 \ ■ | \ • c ' 1 ' F a _ • i - £ J— on UJ >\ v_»» / • •R /n ^ ■ • i M 1 \ • E " E •o © 1 • 1 • ■ r : J 1 1 • lo 1 r • i ig_ • T 1 1 • / / • 1 i — » ■ y • OO O LO d o 1 -i. ... * ft above ad / • — • • QJ a •4— o z> e £ \_ j 3 ^— N !\ <— / • \ • JC \ • \ • 1 - r^ \ , -4— \ Q. f ©j y 1 • OJ -D C O oo 1 1 ■ LO c D UJ o O O I CD u a CO cd cu ■P o Ph -P CO U CD CD TJ •H CO CD c P P cd c • O hO •H O -P Cm Cd m T3 P CD (D >s C Cd CD U CX CX CO CD TJ ^ •H O M O +3 rH H JG 3 O CO I m w •H 208 Fig. 6--Format ions of ettringite around an aggregate grain (magnification 500x). 209 CORROSION OF STEEL IN REINFORCED CONCRETE IN MARINE AND OTHER CHLORIDE ENVIRONMENTS R.D. Browne Taylor Woodrow Research Laboratories 345 Ruislip Road Southall, UB1 2QX, England Abstract: Concrete structures in marine environments can exhibit damage due to reinforcement corrosion after 15 years in the U.K. and after as little as 5 years in hotter climates. Chlorides in concrete from contaminated aggregate, or as calcium chloride for strength acceleration, or from de-icing salts for roads, have also caused widespread damage. Corrosion generally occurs when the electrical resistivity of the concrete cover to steel is low (5-10,000 ohm cm), when atmospheric oxygen can diffuse through the cover layer, and when the chloride level of the steel surface is 0.4% by wt. of cement. The diffusion rate of the chlorides from the environment through the concrete cover to the steel enables the time of onset of corrosion (tj) to be predicted. This timescale is used as the basis for predicting the durability life of the structure and a nomogram is presented which allows the designer to select his cover thickness and his concrete to give a required maintenance-free life. The time from steel activation to the occurrence of damage (t2), depends on atmospheric oxygen diffusion to the steel and the bursting forces from corrosion products. The above considerations provide a basis for design, repair and inspection procedures and additional protection of structures. Key words: Chloride attack; concrete spalling; cracking; durability; inspection techniques; lamination; life prediction; reinforcement corrosion; repair techniques. INTRODUCTION Concrete made from natural stone aggregate and sand with a cement/water binder, is by far the most utilised artificial material made by man. Since the 19th Century it has been extensively used throughout the world for many forms of construction: for dams, bridges, roads, storage tanks, to the structural frame and cladding of buildings, down to small components such as fencing posts, street furniture and railway sleepers. It has a long record of success, but constant vigilance by engineers and builders is necessary to ensure it is made of the right materials and the correct mixes for the particular application and environment. 210 Mild steel and high tensile steel is cast within the material to enhance its low tensile and shear strength in the form of reinforcing bars or tensioned steel wires or bars. The alkali cement environment within the material is sufficient generally to protect the embedded steel without further coating of the concrete external surface or the steel itself; in fact, reinforcement is allowed to rust before casting the concrete around it, to enhance the bond strength at the steel/concrete interface by mechanical friction. Reinforcement is nowdays generally ribbed or indented to increase the mechanical bond even further. By providing an adequate concrete layer between the embedded steel and the concrete surface, corrosion of the untreated steel by an aggressive environment is prevented. Cover thickness is generally designed to range from an upper value of 75mm for extreme exposure conditions, down to 15mm, for mild conditions. Traditionally the engineer has, mainly from experience augmented by exposure trials, selected from his codes^' both the minimum thickness of cover and the lowest quality of the concrete mix to meet the degree of exposure likely in his structure. Such an approach has lacked a detailed scientific understanding of the importance of the environment, the way it can penetrate to the embedded steel and the influence of the hardened cement/aggregate matrix on its effectiveness to protect the steel. The result has been that a number of major problems have occurred in structures and buildings when, in particular, chloride ions present in the concrete or in a marine atmosphere, and carbon dioxide from the atmosphere, have resulted in cracking, spalling or lamination of the cover layer (Fig. 1. a,b,c,d), exposing the corroding steel and allowing it to lose its cross-section to an extent that the structure is weakened. Rust staining of facades of buildings from corroding reinforcement can also degrade their appearance. Resulting from our detailed investigations on the durability of concrete^) f or gravity platforms for oil production and storage in the North Sea (Fig. 2), we have researched since the early 1970's, the effect of the marine environment on the performance of embedded steel in concrete structures. Existing marine structures have been surveyed both above and below water; laboratory tests have been initiated to measure the permeability of concrete to seawater under hydrostatic pressure, to examine the protection of reinforcement by the concrete against corrosion fatigue, and to develop further suitable NDT inspection techniques'^). From this work, a simple environmental penetration and damage theory has evolved to aid design for durability and to provide a better basis for inspection and repair.'^) The results have been applied to a range of structures both in the U.K. climate and overseas in the hot climates, not only to marine structures but also to land-based construction; for road bridges, where de-icing salt from the deck penetrates through joints down to the structure beneath causing cracking and spalling; in buildings, where calcium chloride has been added to concrete to accelerate its strength again, with the same type of damage; and 211 where aggregates containing salt have been allowed to be used in Middle East buildings, with similar symptoms. The paper therefore briefly describes the deterioration theory, proposes a method of design for durability, as well as, giving a basis for further protection, inspection and repair procedure to enhance the life of a structure. It is hoped that the concepts presented for steel in concrete, will help advance the developing science of deterioration processes in structures and design methods for low maintenance, long life performance which for other materials is being discussed at this important Conference. Without a stronger basis for the design of the life of a structure, the already rapidly increasing effort and cost of remedial work and replacement of existing structures, will not be halted, thus preventing society concentrating on new construction. In the U.K. alone, in 1979 over 36% of construction was devoted to replacement or restoration work, and this is increasing year by year. The design life for a bridge is 120 years, whilst most other buildings and other structures are expected to last for at least 50 years, and preferably longer. Many may not reach this timescale without major repairs. THE MECHANISM OF CORROSION DAMAGE OF STEEL IN CONCRETE The alkalis present in hardened cement passivate the embedded steel. As shown in Fig. 3, the process of corrosion is induced by three main factors: i) Chlorides - which depassivate the steel, allowing corrosion to initiate. ii) Low electrical resistivity of the cover concrete - permitting electrolytic corrosion cells to develop. iii) Penetration of atmospheric oxygen to the steel through the concrete cover - producing corrosion products with a volume over 4 times the reacted steel, thereby disrupting the concrete cover. Only certain combinations of chloride level at the steel surface, concrete resistivity and oxygen availability, result in damage: a) the chloride level has to exceed about 0.4% CI" by weight of cement, although this may vary with cement type. b) the resistivity of the concrete has to drop from a normal value of greater than 20,000 ohm.cm. to 5-100,000 ohm.cm. c) oxygen in the air has to diffuse through the concrete layer sufficiently to produce corrosion damage in the lifetime of the structure. 212 Where any one of the three factors is absent, damage does not occur. The resistivity of concrete drops rapidly with increasing moisture content and is lowered still further by the presence of chlorides (Fig. 4). But Oxygen diffusion is also dramatically decreased by higher moisture contents in the concrete cover (Fig. 5). Optimum corrosion conditions can occur when the cover is partially dry due to the interaction of these two factors. For a marine structure in the underwater zone, the lack of oxygen (5-10 p.p.m. in sea water) very effectively stifles corrosion even though the seawater (containing 1.9% chloride) under hydrostatic pressure can penetrate to the steel, and the resistivity of the concrete is low. In the marine atmospheric zone where the chlorides have penetrated to the steel to the critical level, atmospheric oxygen can flow through the cover to the steel if the oxygen diffusivity of the concrete is high enough. The high humidity saline environment results in not only salt deposition on the structure's surface but also prevents drying out of moisture in the concrete, preventing the concrete resistivity increasing above the critical levels. Damage therefore concentrates in this zone. In the tidal zone, the repetitive wetting of the concrete cover restricts oxygen diffusion except where large tidal ranges permit the upper zone to be wetted only infrequently. The latter region can show similar damage to the atmospheric zone. In a dry dock in hot climates, the building dock is not flooded frequently for many months, which can permit the previously immersed walls to dry out sufficiently for oxygen to diffuse to the steel. In one case, this has resulted in extensive delamination of the walls. For land based structures where calcium chloride is present in concrete buildings or where deicing salts have been regularly applied to bridge decks, the three factors can all be present: chloride ions at the steel surface, the concrete resistivity being low and atmospheric oxygen diffusing through the partially dry concrete. In the extremely dry Middle Eastern environment, one building 25 years old established serious cracking and corroded reinforcement only in certain areas. It was discovered that although high chloride levels existed throughout the concrete, the high oxygen diffusion for the dry concrete was negated by the exceptionally high resistivity of the concrete, in excess of 150,000 ohm. cm. However, the latter dropped to 5,000 ohm.cm. when wetted. Corrosion could only occur after the yearly rainfall period in wetted parts of the structure, including the external facade columns and roof slabs, and internally, where basement flooding had occurred. PROPOSED BASIS OF DESIGN For concrete of a low resistivity, consider (tj) to be the time for chloride ions to diffuse to the steel, and (t2), the time thereafter for oxygen diffusion to result in a level of steel corrosion (d) that damage occurs (Fig. 6): 213 d = tj + t 2 The value of tj equals the time for the chloride ions to diffuse from the surface through the cover and reach the critical level at the steel. Samples at different distances from the concrete surface, obtained by drillings taken from structures, have given chloride distribution curves that suggests that Fick's diffusion law can be applied: dC D dt c " dx Where C = the chloride ion concentration at a distance, x, after time, t. D c = the chloride diffusion coefficient Fig. 7 shows a typical chloride distribution for a marine structure. A "best fit" diffusion curved through the results enables the diffusion coefficient (Dc) to be obtained and the intercept with the concrete surface (Cc) to be extrapolated. The latter chloride content is an artificial surface chloride value since in practice salt depositions on the surface have never been measured directly. Using the resulting values of Dc and Cc, curves of chloride variation with depth from the surface can be drawn for different exposure times. Table 1 shows the range of values taken for a number of structures for a range of chloride exposure conditions. DESIGN APPROACH It seems reasonable to base the design method on the assumption that chloride level should not exceed the critical value at the steel surface during the life of the structure, i.e. the design life = tj The time for significant damage to occur (t2), is less predictable due to the many parameters involved in the two main factors: (a) The rate of oxygen diffusion (Do) through the cover. (b) The time for bursting forces to disrupt the concrete. t 2 depends on the moisture content of the concrete, its quality, its tensile strength, the size and concentration of the steel, the cover thickness, and the shape of the surface of the structure (corners are very vulnerable). Values of t 2 observed in practice appear to range from 6 months to 5 years where critical chloride levels exist in concretes. In hot climates, the 214 higher temperature and the more rapid drying out of the concretes can also give short term t2 values. DESIGN NOMOGRAM A family of curves of chloride concentration with distance from the surface with time for different surface chloride levels (C c ) and chloride diffusion coefficients (D c ) can be presented as shown in the nomogram in Fig. 8. The variable scale ordinate enables the effect of different C c values to be evaluated. By selecting a C c and D c value, the design life for the steel at different cover thicknesses can be examined; or conversely, the D c value and cover thickness can be selected for a particular life. For example, for a 'life' of 10 years, a D c value of 1 cm^/sec and a critical chloride level at the steel of 0.4% by wt. of cement, a cover of 50mm is required when a very high surface chloride is taken, i.e. 10% by wt. of cement. If the D c value was 10 times greater, then the cover would have to be absurdly high, 500mm. For corners, the life may be approximately halved, or conversely, the cover has to be doubled to give the same life due to the two-directional diffusion path from the two sides. The criticality of D c can readily be seen from the chart. Values obtained so far indicate that for normal concretes, exceedance of the mix water/cement ratio (w/c) above 0A by wt., permits continuity of the internal pore structure of the hardened cement resulting in high D c values. Low w/c ratio mixes give high compressive strength concretes (+50 N/mm') as are now in more common usage. One such structure we have surveyed in the North Sea, had a chloride diffusion coefficent of 0.3 x 10"^ cm^/sec giving a life of 75 years, well in excess of the required structural design life of 20-30 years. Many conventional structures however have been built with w/c ratios exceeding 0.5, resulting in spalling in the U.K. marine climate, particularly at corners after 15 years or so. Table 1 illustrates these points for a number of structures recording the D c and w/c values and the resulting deterioration. Evidence is emerging that the use of cement replacement materials (e.g. pulverised fuel ash and blast furnace slag) can reduce the D c values by significantly reducing the size of the interconnected pores in the hardened cement. Some structural lightweight aggregate concretes appear also to give lower D c values, possibly due to the more extensive cement hydration from removal of water stored in the lightweight aggregate's pores. Using the above method, the designer for the first time can select his cover and concrete to meet his particular life requirements. If he is worried about the weight of thick cover, he can specify a better concrete to give his long life. More work is required, however, on grading concretes for their D c values. 215 If required he can check the D c value for a concrete by testing of samples with chloride saturated surfaces and observing the rate of chloride penetration with time. Care is needed to avoid artificially high values by storing the samples in water, if in practice this is unlikely to occur. ADDITIONAL PROTECTION From the above design approach, it can be seen that the designer can examine critical exposure conditions of his structure and either provide additional cover and better concrete, or he could use protective coatings on the concrete surface which restrict chloride ingress, or even seal the actual steel surface against chlorides. We have found by screening different surface coatings and materials which soak into the concrete surface, that chloride penetration can be reduced to an insignificant level. Also, reinforcement coated with powder epoxies and other resin-based products can isolate the steel from chloride attack. With such products the designer may deal with chloride environments more effectively in the future. INSPECTION When structures are subject to chloride attack, techniques we employ'^' °' enable us to check the extent of chloride penetration by simple drillings; checking whether the steel has become active by use of copper/copper sulphate j cell, surface electrical potential, measurements linked to the reinforcement; by 4-probe electrical resistivity measurements of the cover concrete; and even by oxygen diffusion measurements of the cover layer from 100mm cores extracted from the structure. Visual checking for cracking, spalling and lamination together with the use of the rebound impact hammer for checking voidage beneath the surface, all complement these measurements. REPAIR A number of states of penetration are considered where chloride damaged structures need repair: (a) Where chloride is penetrating to the steel but has not yet reached the critical level at the steel surface, surface coatings can enhance their life considerable. (b) Where chlorides are above the critical level at the steel, then coatings which we are investigating to restrict oxygen diffusion, can be applied to extend the life to its design value. It may be better to apply less effective, but cheaper, coatings periodically throughout a damaged structures life to maintain its durability, rather than carrying initially the high cost of treating large areas of the structure with more expensive higher quality coatings. The economics of using different forms of further protection depend on the area requiring protection. 216 ATMOSPHERE CO ? Penetration of CO2 as carbonic acid in the free water in concrete, tends to move as a front more slowly than chlorides towards the steel. In marine structures, even after 20-35 years, carbonation only to a depth of 10mm has been frequently recorded. Inside buildings where central heating and people can generate a moist, warm environment, the low covers frequently employed (down to 15mm or less) could result in reinforcement corrosion in ceiling slabs and beams particularly at exposed corners. Plaster appears to provide little protection and even emulsion paints may be easily penetrated by CO2. Further measurement of the extent of penetration on such structures is, therefore, necessary to more clearly define cover and concrete requirements for long life design. CONCLUSIONS The paper has outlined a method for the design of concrete structures against chloride attack on embedded steel based on the diffusion coefficient of concrete for chloride ions. It is believed that such a basis will provide a more rational approach for the designer to select his cover and his concrete, or to specify additional protection measures. This should enable the structure to require little maintenance for the whole of its design life. Our approach to inspection and repair of existing structures is also outlined. Further work on the basic concept, however, is still needed to cover the range of environments and materials employed in structures around the world. ACKNOWLEDGEMENTS The author wishes to thank the Directors of Taylor Woodrow Construction for permission to publish the paper based on work undertaken in the Research Laboratory and, in addition, Mr. M.P. Geoghegan, Dr. A.F. Baker, Mr. F. Papworth and other laboratory staff for their efforts in undertaking survey measurements on actual structures and for surviving endless discussions on the results obtained, from which the contents of this paper have been derived. 217 REFERENCES 1. British Standard Code of Practice. "The Structural Use of Concrete" CP 110 Part 1 "Design, Materials and Workmanship". The British Standards Institution, London, 1972. 2. Browne, R.D. and Domone, P.L.J. "The Long Term Performance of Concrete in the Marine Environment". Paper No.5, Session C. Research and Development for Offshore Structures. Institution of Civil Engineers, London, 1975. 3. Taylor Woodrow Research Laboratories. "Marine Durability Survey of the Tongue Sands Tower". Concrete in the Oceans Technical Report No.5. The Cement and Concrete Association, Wexham Springs, Slough SL3 6PL, England, 1980. 4. Browne, R.D. "Mechanisms of Corrosion of Steel in Concrete in Relation to Design, Inspection and Repair of Offshore and Coastal Structures". Performance of Concrete in the Marine Environment American Concrete Institute SP-65, 1980. 5. Browne, R.D., Domone, P.L.J, and Geoghegan, M.P. "Inspection and Monitoring of Concrete Structures for Steel Corrosion" Paper No. 2802. Offshore Technology Conference, Houston, Texas, 1977. 6. Browne, R.D., Doyle, V.J. and Papworth, P. "Inspection and Repair of Offshore Concrete Structures". Paper No. 220, European Offshore Conference, London, October 1980. 7. Tuutti, K. "Chloride in Pore Water" Steel in Concrete Electrochemistry and Corrosion Newsletter No. 5/6. Korrosionscentralen ATV, Denmark, January 1980. 218 o U C n3 XI 0) y ^ Q > 4-1 L- 3 to in CD to -J CD l* td 3 > u X) 3 r i~ (fl +-> CO CD DO (d 4-1 o £ 4-" CO CI) a M-t cd O > cd CD 00 cd O 4-1 UJ CQ < 00 c .-H • (D fO • DO CX c +-■ u 3 CO CD in X> c CO c o • in cd CD Li (0 c o .— * •M u CD CO c E ?3 ■M CL CJ If) CD £ 2 CD -t-> fO c E J0 CD > o U CD CD > CO 4-1 o CO o • — » JD i— CD > CD to JO X « CO * CO 4H .=5 CO E CD C l_ O tO o U 13 10 CD ■(-« CD CD .-j to TO (0 a o Z CO CD x c > XI O '3 on 3 o h -^ c ^.9 +-> CD cd flj »-5 in (N ir\ ^ 2: hh j- •3- ■a- u-n r^ VO Cone Speci • o • O V • o • O V • o • o V <1> TD • -H 4-> c »^ >— " *— ' >■» ' "— " *»^ >-^ ^^» v£> J- rr\ — * O0 -* v£) O u _ 42 SJ • ■ • • iA • • +-> 3 CD ^ CO _> c o /— v '33 u»? •— i OS O 2 a 4-H C^- o • O • • • • ON (N<* 4-H ^H >"~s, ^H \0 •a- — H — 4 — * ON 3 c X • x: o E CD u^ o\ Ch o 00 00 UU ^u • • o • • • • CN O "!/? CD fO 00 ^ lA o o •3- CO ro •* r^ (N CD "c0 +-< m CO O DO c X '3 y o a 1- CD X 3 do y ^ X ^ X (D C + o N CD c JZ o N CO ^-t L- a CO < CN cj- c o to N 4- 1 CO CO to X d) •-H DO c +J C o J- • -h N (D u a. 1 • — 1 a CD 0) c o to -♦-» "8 O i_ 0) o o < 0) D o o o o o o O w O O O ■*: in id oo D O + X • + o X OX xo t? o CD O CM ir> o < o o CO o CM (DdlN) dOW s229 a FIG. 2 TEMPERATURE DEPENDENCE OF SIC STRENGTH r 10.000 -1000 I/) >% o ■o (N 6 E o o c CD o CO o o Q) -100 0) E -10 CemFIL AR Fibre o Development Fibre 2-8 30 __L 4 /T x 10" 3 (°K" 1 ) 230 3*2 3-/. LU LU 3 O cr o Q LU q: LU I LU Q O Q. C o o o O ir> o o o o o- in - Csl- o 0) < I- o LU co < Z> X LU _) LL CD (_ _g Ll in in in en O LL o o CO o CM o - (DdW) UOW 331 DURABILITY CRITERIA FOR CELLULAR CONCRETE OUTER WALL S. Nakano and S. Tada Misawa Homes Instutute of Research and Development 2—4—5 Takaido-Higashi Suginami-ku, Tokyo, Japan. 168 Abstract: Analysis was made on the durability of structural material, especially the mono- lithic cellular concrete outer wall. Assuming that the free water content only exhibits change over time, the durability criteria can be remarkably simplified. Changes in average moisture content were inspected, and an approximation of the change curve based on a comparison with theoretical analysis was made. At the same time, the relation of moisture content and physical properties was tested in the laboratory, and the performance over time was expressed as a function of time. The method of anticipating performance change with the changes in average moisture content as a medium should be applied to insulation performance only. As for change in structural performance over time, the highest moisture region in the wall is subject to frost damage and corrosion of reinforcement. Thus, changes in moisture content distribution should be studied carefully. This report discusses thermal and acoustical performance. The results are used for showing the user the scope of guarantee indicating the change, when performance requirements have been promised, rather than for anticipating the overall service life of dwelling. Key words: Building material; cellular concrete; durability criteria; methodology; moisture content; performance over time. 1. INTRODUCTION If the bulk density and thickness vary, the cellular concrete can produce a balanced value in regard to the performance requirements of the outer wall. Especially in Japan, monolithic use is fully possible. This is advantageous when considering the durability criteria. Outer walls consisting of single material have less factors affecting perfomance over time, compared with outer walls made of composite materials, and it can be expected that the method of assessing durability is simplified. This report describes autoclaved cellular concrete with a bulk density of 500kg/m 3 and a thickness of about 0.12m, which is most widely used. This material consists of more than 75% of pores, moisture distributed among them, and the substance. Among these com- ponents, it can be assumed that only moisture shows changes over time. 232 2. CHANGES IN MOISTURE CONTENT Changes in moisture content are change in capillary potential, and the corresponding property change are dominated by thermodynamical principles. Moisture content is em- ployed here as the quantity to be measured phenomenally. 2.1 Average Moisture Content The changes in moisture content of autoclaved cellular concrete have been measured often by researches, because of the wide usage for a long time, and special attention to the moisture problem. Ottoson, Tviet, Kunzel, Elmroth, and the author conducted in-situ measurements (Table 1), and the results showed good agreement with the theoretical analysis by Hanson and Sandberg (Fig. 1). The result showed that under ordinary climatic conditions, the initial moisture content of about 200 to 150kg/m 3 will become the equi- librium moisture content of 20kg/m 3 , and no significant change has been observed there- after. Matsumoto analyzed the annual variations beginning with the hygroscopic initial moisture content by the equation of simultaneous heat and moisture transfer based on the non-equilibrium thermodynamics, and the result was consistent with the above conclusion. Figure 2 shows an approximation with a straight line of the average moisture content, for the purpose of showing the performance changee as a function of time. # = 160-46.7t (3 > t > 0) (1) # = 20 (t > 3) (2) where ^ : moisture content (kg/m 3 ), and t : time (year).* This approximation of the average moisture content includes a sufficient margin for danger, while the application should be limited to thermal and acoustical performance. As for the changes in structural performance, it is necessary to study the moisture distribution in the wall, and the highest moisture region should be considered. 2.2 Moisture Content Distribution When especially high moisture content is observed in a certain portion in the wall, it may be mainly due to the intrusion of rain from defective exterior treatment, or internal condensation. While the latter case can be discussed based on the data of extended study, the basic data for discussion the former case is still short. An internal condensation experiment using the vapor tight layer on the cold side was conducted by Larsson, Hanson, Matsumoto, and the author. Assuming the boundary con- dition is constant, the increase of moisture content due to condensation starting from the air-dry condition does not continue infinitely. Unless the wall is extremely thin, it may be very rare that the moisture content of 200kg/m 3 or more, corresponding the maximum value of water diffusivity is observed. Sandberg and Matsumoto conducted numerical experiments on moisture content distribution by cyclic boundary conditions, while the author performed in-situ measurements. The contribution ratios of the following elements for increased moisture content due to internal condensation have been analyzed already: Vapor and water diffusivity, out door temperature, sol-air temperature, daily change, 233 vapour conductivity of exterior waterproof layer, interior relative humidity, and orienta- tions of the wall. From the above observation, the general features of the change in moisture content due to internal condensation can be grasped, and it is understood that there is no possi- bility of exceeding 350kg/m 3 , which is dangerous related to frost damage. In fact, when inspecting frost damage in Japan, it is often the case that molten snow is constantly supplied to the part of defective exterior treatment, or condensed water on interior cold bridge is absorbed by the base cellular concrete, to exceeding the critical degree of saturation. There- fore, even though it can be concluded that the change in structural performance, such as frost damage and corrosion of reinforcement, are affected by moisture in the wall, anticipa- tion of moisture intruding from the exterior is necessary for exact prediction. 3. PERFORMANCE CHANGE 3.1 Thermal Insulation In japan, there is no standard which controls thermal insulation of the outer wall; only the Hokkaido region has a related regulation. There is a grading of requirements for in- dustrial dwellings for the benefit of the users to judge performance (Table 3). As an index of the insulation performance, coefficient of thermal transmittance K is used: K = 1 (3) 1 +d + 1 ai A ao where ai, ao : surface coefficients of heat transfer, being 9.3 (W/m 2 k) and 2.3 (W/m 2 k), re- spectively, d : the thickness of the wall, being 0.12 (m), and X : coefficient of heat transfer (W/mk). There are abundant records of measurement on the relation between the coefficient of heat transfer X and moisture content ^, and the dependence of X on V is lineary correlated, when the moisture content is 200 kg/m 3 or less. Figure 4 shows the measurement by steady-state heat flow method. When approximating by the least squares method, the following can be obtained: X = 1.07 x 10~ 3 * + 0.089 (4) Furthermore, using eq. (1), X = 0.26 - 0.05t (0 < + < 3) (5) where X : coefficient of heat transfer (W/mk), and t : time (year). Therefore, from eq. (3) and eq. (5), thermal insulation performance K within the range of t = ~ 3 can be expressed as follows: K = (0.26 - 0.05t) / (0.159 - 0.075t) (6) 234 By Table 3, it is set that the requirements should be below K = 0.89. Based on this, Figure 5 shows how the insulation performance changes over time when using eq. (6). It is shown that thermal insulation performance increases over time, and clears the requirements after 2.7 years, and becomes constant after three years. 3.2 Sound Insulation Similar to thermal insulation, performance change of sound insulation as grade 2 of Table 3 is examined. Transmission loss TL is used as the index of sound insulation; it is given as the function of frequency and surface density with the monolithic wall under question. TL= 10 log* (|^) 2 - 10 log 10 [log e (l + (|^) 2 )] (7) where C pa f m sound speed in the air (m/sec), air density (kg/m 3 ), frequency (Hz) and surface density (kg/m 2 ). By substituting the specific impedance of air ( =414 kg/m 2 sec) for paC, 500 (Hz) for f, and 0.12 500 + (160 - 46.7t) 'for m, the following can be obtained: _. in . r (39600 - 2802t) 2 " L= 10 log 10 l 46 |ogio ( 396 00-2802t) -9.775 J " 4Zb (y) (0£t <3) where t is time (year) and the equation (8) shows change in sound insulation over time. With the ordinary cellular concrete reinforced elements, coincidence effects generates near 500 Hz, and there is a case that TL decreases from the calculated value based on the mass law, to become about 30 dB or less. This calculation is applicable for large cellular concrete panels of room size. Transmission loss of actual building using these panels has been already measured (Fig. 7). By the measurement, the coincidence effect is observed not with 500 Hz, but about 2000 Hz. Sound insulation value decreases over time, corresponding to the changes of surface density accompanying drying of the wall, forming a straight line on the chart, and ultimately becomes constant. 4. CONCLUSION Change in performance of monolithic cellular concrete is expressed as a function of time, with the changes in average moisture content as the medium. It is necessary to obtain how the requirements promised to the user changes over time, when selling the house as merchandise. The house of cellular concrete reported here is a single family house develop- ed for high quality, low price, and high productivity. Supply of this house has begun from January, 1981 , and the above values will be proved in the future. 235 Acknowledgement This report was compiled largely from the encouragement of Mr. Nireki of the Building Research Institute of Japan, who is the theoretical leader of "durability criteria based on performance concept" in Japan. This study forms a part of the technical development of the national project "HOUSE 55", sponsored by the Mistry of International Trade and In- dustry and The Ministry of Construction. 5. REFERENCES 1. Ottoson, G. "Dehydration and Contraction in Lightweight Concrete", Moisture Pro- blem in Building , Symp. of Rl LEM/CIB, Helsinki. 1965, 2-30. 2. Kunzel, H. "Das Warme-und Feuchtigkeitsverhalten von Casbeton im Hochbau", Beton, Vol. , No 3, 1971, pp 101-104. 3. Elmroth, A. Hoglund, I. "Influence of Moisture on the Thermal Resistence of External Walls of Cellucar Concrete", Moisture Problem in Building, Symp. of R I LEM/CIB, Helsinki, 1965,4-10. 4. TViet, A. "Heat Transmission through Test Walls of Autoclaved Cellular Concrete", Lightweight Concrete , Symp. of RILEM, Goteborg. 1960, pp 381-403. 5. Hanson, R. "Moisture in Lightweight Concrete Roof", Lightweight Concrete, Symp of RILEM, Goteborg. 1960, pp 405-431. 6. Sandberg, P. I. "Byggnadsdelars Fuktbalans i Naturigt Klimat" Division of Building Technology, Lund Technical University, Lund, Sweden. Report 43, 1973. 7. Matsumoto, M. "Simultaneous Heat and Moisture Transfer in Porous Wall and Anylasis of Internal Condensation", Energy Conservation in Heating and Ventilating Buildings , Int. Seminar Heat and Mass Transfer, Dubrovnik, 1977. 8. Idem "Analysis of Annual Variation of Moisture Distribution in Cellular Concrete Wall" Trans. Architectural Institute of Japan , 1980, pp 609-610. 9. Larsson, L. E. "Migration of Moisture in Siporex Roofs" Lightweight Concrete , Symp. of RILEM, Goteborg. 1960, pp 331-351. 10. Nakano, S., Tada, S. "Properties of Moist Cellular Concrete — Pore Structure and Water Movement", Trans. A.I.J., 1980, pp 87-88. 236 TABLE 1 — Conditions of measurements Symbol Author Interior condition Location •Duration Dimension (kg/m 3 ) (m) • east west Kunzel S 60~70% 13- 17°C W 40-50% 20° C Horzkirchen 1961 ~ 1966 590 0.15 © Ottoson Stockholm 1959- 1963 555. 521 C Elmroth 22° C Stockholm 1963- 1965 400 0.25 A Sand berg 22° C 50% Lund Calculation 550 0.20 x east Bwest Tviet 20~22°C 40 ~ 50% Trondbeim 1957 ~ 1960 550 0.25 ■ Hanson 55% Stockholm Calculation 530 0.15 O Tada 12~26°C 60 ~ 75% Tokyo 1978-1981 500 0.12 TABLE 2 — The increase of moisture content in cellular concrete wall due to forced internal condensation Author Conditi interior ons exterior Duration (day) I ncrease (kg/m 3 ) Dimension (kg/m 3 ) (m) Larsson 18~22°C 50 ~ 70% -14°C 45 220 500 0.118 Hanson 45°C 45% -40°C 540 140 530 0.120 Matsumoto 20°C 75% 12.5°C 55 170 500 0.020 Tada 20° C 60% -10°C 60 82 498 0.120 TABLE 3 — Grading of performance requirements for industrialized dwellings in Japan Grade 1 2 3 4 R , m 2 hr °Ci 1 kcal ' >1.7 1.7~ 1.3 1.3~ 1.0 1.0-0.7 K (W/mK) <0.68 0.68 ~ 0.89 0.89- 1.16 1.16- 1.66 TL (dB) >30 30-25 25-20 20-15 237 200 E CT5 c CD +-■ c o U CD > < 150 •|2 100 CD CXI CO FIG. 1 12 3 4 5 Time (year) Variations of average moisture content in cellular concrete 200 J| 150 h c CD +- 1 c o o CD i_ =3 +-< ^? o E 100 - 50 - .*-' — o— • -O.— • -••Or— * — O — .— -o-. --0- — ■•- CO CO ■o c 3 _c ■D C 3 O 00 12 3 4 Time (year) FIG. 6 — Changing performance of sound insulation 240 10L-1 125 FIG. 7 250 2000 500 1000 Frequency (Hz) In-situ measurement of transmission loss (2 years old) 4000 241 DURABILITY PERFORMANCE OF POLYMER-MODIFIED MORTARS Y. Ohama College of Engineering Nihon University Koriyama, Fukushima, Japan Abstract: Recently polymer-modified mortars have been widely used in the field of construction work in the world, because of their advantages such as high strengths, excellent adhesion, waterproofness and chemical resistance. The durability performance of the polymer-modified mortars has hardly been studied until now. The polymer-modified mortars using various polymer dispersions were prepared with different polymer-cement ratios, and tested for weatherability , freezing- thawing resistance, heat resistance and chemical resistance. The test results demonstrate that most polymer-modified mortars have good durability performance over conventional cement mortar. Key words: Chemical resistance; durability performance; freeze-thaw durability; heat resistance; polymer dispersions; polymer-modified mortars; weatherability. Introduction- Polymer-modified mortars are characterized by containing large amounts of polymers which supplement the portland cement as a binder. The presence of polymer phases confers various desirable properties on the mortars. On the other hand, there is some fear that their weatherability may be limited because of undurable polymers contained. In increasing demand for the polymer-modified mortars in recent years, it is strongly expected to clarify their durability performance for the practical uses. The durability such as weatherabil- ity, freeze-thaw resistance, heat resistance and chemical resistance of the polymer-modified mortars is tested and discussed in this paper. Weatherability - SBR (styrene-butadiene rubber)-, PAE (polyacrylic ester) - and PVAC (polyvinyl acetate)-modified mortars were prepared with cement : sand ratio of 1 : 3 (by weight) and polymer-cement ratios of 0, 20 %, and their weatherability was investigated through 8-year outdoor exposure at Building Research Institute Outdoor Exposure Site in Ibaraki Prefecture, Japan. Mortar specimens 40 x 40 x 160 mm before and after outdoor exposure were tested for flexural and compressive strengths according to JIS R 5201 (Physical Testing Methods of Cement). The weatherability of the polymer-modified mortars is shown in Fig. 1. Except PVAC-modified mortars, the flexural and compressive strengths of most polymer-modified mortars under outdoor exposure tend to become 24 2 Outdoor Exposure Period (years) CM e o s: 4-> C d) U > ■H ra ra CD a E O o 5001- 400 - 300 200 -, e 100 - 0.51 Outdoor Exposure Period (years) Fig. 1. Weatherability of Polymer-Modified Mortars. Note; SBR - styrene-butadiene rubber, PAE - polyacrylic ester, PVAC - polyvinyl acetate. nearly constant at the age of 1 year or more, and they give the weather- ability over or like unmodified mortar. It appears that the maximum compressive strength at the age of 6 months is due to the progress of cement hydration during outdoor exposure. The adhesion durability of the polymer-modified mortars to ordinary cement mortar is also studied through 10-year outdoor exposure in Tokyo and Sapporo, Japan. In the outdoor exposure test, mortar specimens 40 x 40 x 160 mm were prepared by joining the SBR- and PAE-modified mortars with cement : sand ratio of 1 : 2 (by weight) and polymer-cement ratios of 0, 20 % to the 40 x 40 mm cross-sections of the cement mortar half-prisms as adherents, and before and after outdoor exposure tested for flexural strength accord- ing to JIS R 5201. Their adhesion to the ordinary cement mortar was determined as a flexural strength from the results. The adhesion durability of the polymer-modified mortars is demonstrated in Figs. 2 and 3. The adhesion retention shown in Fig. 3 is calculated by the following equation: Adhesion retention (%) Adhesion after outdoor exposure (kg/cm^) Adhesion before outdoor exposure (kg/cm2) x 100 The types of failure in these adhesion tests were adhesive for unmodified mortars, and cohesive for most polymer-modified raortars. The adhesion of the polymer-modified mortars tends to decrease with additional outdoor exposure period and to become nearly constant at the age of about 60 months (5 years), regardless of the mortar types and the 243 c * o — •H U CO CO > u as c •H '— T3 C\J u e o o o cm -p j^ o -— •H £h CO CD CD -P T3 O < 2 100 Unbonded Specimens a— a d • b i«^;-_^ailed^tl20 H"o„ ths , a S8R ~ b NUnraodif ied d. Failed at 6 Months 03 612 36 60 120 Outdoor Exposure Period in Sapporo (months) Fig. 2. Adhesion of Polymer-Modified Mortars (to Ordinary Mortar) vs. Outdoor Exposure Period in Tokyo or Sapporo, Japan. Note; *Flexural strength in the case of unbonded specimens. Exposure Type of Site Mortar Adhesion Retention* (%) 50 100 20°C, 50% R.H. (Control) Tokyo Sapporo S B R P A E - 1 P A E - 2 Unmodified tUnbonded S B R P A E - 1 P A E - 2 Unmodified Unbonded rSBR P A E - 1 P A E - 2 Unmodified L Unbonded 150 — i — Failed at 12 Months Failed at 120 Months Failed at 6 Months Fig. 3. Adhesion Retention of Polymer-Modified Mortars after 120-Month (10-Year) Outdoor Exposure. Note;*:Flexural strength retention in the case of unbonded specimens. 244 exposure sites. In contrast to the unmodified mortar-bonded specimens, which failed within 1- year outdoor exposure in Sapporo and Tokyo, most polymer-modified mortar-bonded specimens had the adhesion satisfactory in practical use after 10-year outdoor exposure. From these results, it is considered to be able to deny a general view that their weatherability may be limited because of undurable polymers contained. Freeze-Thaw Durability - SBR-modified mortar, which is the most popular polymer-modified mortar in Japan, was selected for freeze-thaw durability test, and was prepared with cement : sand ratio of 1 : 3 (by weight) and polymer-cement ratios of to 25 %. Specimens 40 x 40 x 160 mm made of it were tested for freeze-thaw durability in accordance with ASTM C 666 (Standard Test Method for Resistance of Concrete to Rapid Freezing and Thawing). Fig. 4 shows the freeze-thaw durability of the SBR-modified mortar. In comparison with unmodified mortar which failed at 70 cycles, the freeze-thaw durability of the SBR-modified mortar is remarkably improved because of reduction of water-cement ratio, partial filling of pores and air-entrainment due to the effects of polymer contained. Ch o CO a H D -n '~"> o S*i S "■ — C ) >s •H -P F •H crs O C •H >, +J Q Ul ro (1) H > W •H -P CC Q) K 160 - 140 - 120 - 100 \ Failed at 70 Cycles ^ I 50 100 Freeze-Thaw Cycles Polymer-Cement Ratio 10% 25% 20% 15% 5% 150 Fig. 4. Freeze-Thaw Durability of SBR-Modified Mortar. 245 i S/r - r- O o o o ( iuo/§>{) q^guaa^s T Bjnxa Ti o ID ( r iiJO/g>i) qq.Su3J^s T Bjnxs TJ ra >> 03 TD TD •H k a> a. M c •H 4-' ca 0) X p o o CM en M CD -P U s T3 CD •H ^-^ Cm (fl •H >> T3 a O T3 S ■o m tu •H U CD e >> Q, H o M C a. •H Cm ro O a; X CD p O c o CO IT) -P .H CO •H CO (D CX P nJ (D K to • >> LO ro TJ - ' ' taO T! •H O fc (h i) q^Suaj^s TBanxaxj 246 CO u CO -P (h O T3 CD •H Cm •H T3 O s I u CD e >> i— i o a, Ch o CD o c 05 -P CO ■H co cd « CO o ■H E CD 43 O 4Q (0

CD o 1 CM 1 00 o O O o o O o 00 LO O CO LO O O 2 CX) 1 00 1 LO 1 LO 1 LO 1 s CD 1 1 o 1 CD CD CM rH 00 rH <>r CD oo iH ^r CO co 43 O pd 6 CD 00 o is "ST 1 00 1 LO o o o o o o o O O O o o o o o t^ oo ^ 'nT LO CD rH CD o O 1 1 i 1 1 1 1 1 rH rH CD E 3 1 1 ce CD CD 00 LO CD o LO 00 CD 00 LO CD CD 9i K CO CM LO IS cn ■g 1 LO 1 CM 00 CD CM o ^J- oo LO CO rH oo ^ CD rH O s 00 00 LO LO LO LO CO o ts LO 00 O IS o CM r-i rH [> rH o oo s IS 00 rH 00 CO LO 00 CD CD LO IS rH rH o rH rH rH o a, 1 1 1 1 1 1 1 1 1 rH 1 r-\ CD / ID ID LO LO CD LO LO O LO O LO o P 1 -P O O P 1 o O 1 ii 1 1 o CO v u rH CM ^r rH CO CO LO rH CO o O LO -p a o/ CD CD CD rH rH w >> s/ • M h y , CO CD CD -P CC S3 ^ / O ^/ CO c iH •H *— / o ! CO "J- 5 / 43 o CO / U •H ^ / T3 T3 CD CO CD E -t- 5 / •H •H CD -P CD CD O c i CD c / O TD CO -P -P C CD CD 43 CD / CO CO T> •H CD Cm CO CO CD T3 <-\ O o / TJ •H X -P rH Cm M &0 •M >> c/ •H T> CJ O T3 o co 3 rH -p o M 43 r-\ / i ' O ■H •H •H CO ■o •H T3 u Cm CO D •H M O -P ■H / / of CO U o M ■H o •H T3 <-{ CO c T3 <-\ CD O M / / / M o CO o M CO O >> D E >> 43 O >-{ CO / / / H o rH 3 3 CO CO 43 CO D E E 43 o U TJ CD CD •H bo // / «H CO •H 43 r-{ O o •H D 3 CD o CD C C o 3 // / O u O •H Cm u •H O E E CO •H •H E E c •-* CD •M CD CQ /// "H D o E O 3 •H u •H D CD c C 3 CD 43 CO H CO >> /// a E Cm M O U Cm +J >> -P •H •H C o o •H •H N CJ CD O O > CD /// a, cd rH T3 u 73 H CD -p o T3 T3 M E E T3 T3 c •H a CO M CO C i/ 3 >> 43 >> 3 O 3 CO O O CO E E O O CD U co CO CD CD CO H O CD X o X 00 < m J CD CD S < < CD CD 0Q H (X O w DC O 247 Heat Resistance - SBR-, PAE- and EVA (ethylene-vinyl acetate )-modified mortars were mixed with cement : sand ratio of 1 : 3 (by weight) and polymer-cement ratios of 0, 10, 20 %, mortar specimens 40 x 40 x 160 mm were molded, and they were heated in 100 °C, 150 °C and 200 °C hot-air ovens for 28 days. The specimens before and after hot-air exposure were tested for flexural strength at 20 °C according to JIS R 5201. The heat resist- ance of polymer-modified mortars is showed in Fig. 5. The heat resistance of the polymer-modified mortars is governed by the nature of polymers used, polymer-cement ratio, heating temperature and period, and ultimately by thermal degradation of the polymers. With additional heating period, the flexural strength of most polymer-modified mortars under 100°C heating tends to reach the maximum at the period of about 3 days, and the large strength reduction is hardly recognized at the period of 7 days or more. This will be due to the strengthening of polymer films with drying. However, the flexural strength of 150 C-and 200 C-heated polymer-modified mortars sharply decreases within short heating period ( 1 to 3 days), because of thermal degradation of the polymers used, and then becomes nearly constant with additional heating period. Especially the thermal degradation of PAE-modified mortar with higher polymer-cement ratio is remarkable under 200 °C heating. As a result, it is considered that the maximum temperature limit for strength retention of most polymer-modified mortars is approximately 100 °C. Chemical Resistance - SBR-, CR (chloroprene rubber)-, NBR (acrylonitrile- butadiene rubber)- and NR (natural rubber )-modified mortars were prepared with cement : sand ratio of 1 : 3 (by weight) and polymer-cement ratios of 0, 20 %, and their chemical resistance was tested through 28-day immersion in various 20 C chemicals. The weight and volume changes of mortar specimens 40 x 40 x 160 mm made of them were determined. Table 1 lists the chemical resistance of polymer-modified mortars. The chemical resistance of the polymer-modified mortars depends on the nature of polymers added, polymer-cement ratio and the nature of chemicals. Most polymer-modified mortars are considerably attacked by inorganic and organic acids, but hardly attacked by alkali and various salts except ammonium sulfate. Their resistance to organic solvents and oils is good except NR-modified mortar. In particular, NBR-modified mortar shows excellent resistance to these chemicals. The polymer-modified mortars generally have good chemical resistance in contrast to ordinary cement mortar (P/C, %) . Conclusions- Through long-term outdoor exposure tests, polymer-modified mortars show unexpectedly excellent durability (including adhesion dura- bility) over conventional cement mortar. Such durability is also supported by the test results of freeze-thaw durability. The maximum temperature limit for strength retention of most polymer-modified mortars is approximately 100 C. The polymer-modified mortars generally possess good chemical resistance in contrast to ordinary cement mortar. 248 PRACTICAL CONSIDERATIONS IN GLASS FRACTURE by Thomas A. Schwartz Simpson Gumpertz & Heger Inc. 1 696 Massachusetts Avenue, Cambridge, MA 02 1 38 ABSTRACT: The high incidence of glass failures in buildings has shown the need for a reassessment of our traditional methods of design, manufacture and installation of glass. An understanding of the conditions that contribute to fracture is a prerequisite to the improvement of glazing technology. This paper presents factors contributing to glass fracture, gives results of laboratory studies, discusses methods for visually determining the source and cause of fractures, and presents recommendations to prevent such fractures. An under- standing of fracture surface markings permits reconstruction of the conditions that culminate in glass failure. With this knowledge, designers can identify the causes of failure and take steps to increase the service reliability of glass in windows and curtain walls. The results of our study of glass fracture are summarized as follows: (i) All edge flaws are potentially serious stress concentrators which can lead to service failures. Some "scored" edge characteristics, considered acceptable by glass manufacturers, weaken the glass as much as other types of edge damage that they consider unacceptable, (ii) Annealed glass can fracture at stress levels as low as 1000 psi, or 10% of the "as-received" strength, when the glass is subjected to aggressive environments and sustained stress in the presence of serious flaws, (iii) The recessed or "cold edge" configuration typical of most glazing designs can lead to failure under thermal stresses. Annealed glass specimens, with edges in "as-received" condition, can fracture at thermal gradients as low as 42 F across a 1/2 in. wide "cold edge". These findings suggest that the reliability of architectural glazing can be improved if (i) designers consider the strength degrading effects of long-term loads, environment, and surface defects as part of their glass selection process, (ii) designers use discretion and impose larger than "normal" factors of safety if significant uncertainties with respect to service loads and glass strength exist, (iii) installers reject glass edges containing shark teeth, (iv) designers consider the thermal characteristics of glazing enclosures in the selection of glass. KEYWORDS: Annealed glass, curtain wall, glazing, hackle, heat strength- ened glass, mirror, mist, residual stress, rib-marks, shark teeth, static fatigue, stress concentration, striations, tempered glass, Wallner lines. I . THE STRENGTH OF GLASS Glass is quite strong in its unflawed state. Laboratory prepared, flaw- free glass fibers have achieved breaking strengths on the order of one million psi. However, the strength of glass is severely diminished by surface flaws, intro- duced by normal handling and abrasion, reducing the actual in-service strength of glass to less than 1% of the virgin fiber strength. I.I Mechanisms for Reducing Glass Strength Surface defects are stress concentrators that reduce glass strength by concentrating tensile stress in a small portion of the glass volume. Stress concentrators magnify the stress level in the bulk material by many orders of 249 magnitude and cause fracture at relatively low nominal stress levels. Physical damage is the most common cause of architectural glass breakage. Surface Damage: Surface damage is most significant since glass breaks from tensile stresses which are maximum at the surface. The magnitude of the stress concentration associated with a particular flaw is a function of the flaw's depth and sharpness at the tip. A small defect with a sharp tip may constitute a more severe stress concentration than a deeper but blunt flaw. Glass units for buildings are cut to the required dimensions by scoring the surface and then breaking by bending the glass across the scored line. The resultant edge varies from smooth, except for the scored groove (Photo I), to ragged (Photo 2). The flaws developed in ragged scored edges are termed "shark teeth". They are common in architectural glass and are considered acceptable by most glass manufacturers, if the depth of penetration of the shark teeth is limited to one-half the thickness of the plate. Our data show that this criterion is inappropriate and that shark teeth constitute significant weakening defects. Environment and Load Duration: The strength of glass is significantly reduced by long-term loads and moisture in the glass environment. This phenomenon is known as environmental stress corrosion. The strength of glass under stress is diminished by moisture vapor which accelerates corrosion of glass at the tips of flaws, sharpens the crack tip, and promotes crack propagation (Ref. I), which in turn produces a more severe stress concentration, and increases the probability of failure. While duration of load and environment are important factors, well documented in the scientific literature, these factors are not adequately addressed by industry standard glass selection charts, which are based on a one- minute uniform load duration. These cannot be compared with other charts or actual tests based on a different load duration unless these data are converted to a common time basis by utilizing the concept of static fatigue (I) to quantify the reduction in glass strength due to sustained loads. 1.2 Experimental Results on Glass Strength Reducing Mechanisms Surface Damage and Environmental Stress Corrosion: We investigated the effect of surface flaws and environmental stress corrosion by bending narrow plates of annealed glass to fracture in air and in water environments. The results, recorded in Table I, show the dramatic dependence of glass strength on surface condition and moisture. The short-term strength of glass is reduced by about 20% when stressed while in contact with liquid water and by about 50% when severe corner chips are introduced. When the effects of severe flaws and moisture are combined, the strength reduction is greater than 60%. 1.3 Glass Strengthening Techniques There are numerous methods for strengthening glass by producing com- pression at the surface of the light. The principal commercial process is known as chill tempering, wherein the molten glass is rapidly cooled by a blast of air. The residual stress distribution developed is as shown in Figure la. Surface compressive stresses in the range of 10,000 to 18,000 psi can be achieved by this method. When the typical bending stress distribution (Figure lb) is applied to tempered glass, some of the residual compressive stress is on one surface overcome (Figure Ic). Fracture will occur only when the entire residual compressive stress in this glass surface is overcome by the tensile component of the bending stress. Fully tempered glass is about four times stronger than annealed glass of the same size and thickness. Heat strengthened glass is tempered in a manner so that the residual stresses produced are lower than those in fully tempered glass. The lower 250 temper eliminates some of the spontaneous fracture problems associated with fully tempered glass, as will be discussed later in this paper. Heat strengthened glass is about twice as strong as annealed glass of the same size and thickness. 2. SOURCES OF STRESS Architectural glass is subject to (i) Structural Stress, (ii) Thermal Stress, (iii) Impact Stress, and (iv) Residual Stress, as described below:. 2. 1 Structural Stress (Bending) Structural stresses develop from deflection of the structure, wind loads, glazing frame distortions, gravity loads, glass-to-frame pressures, and clamping or support conditions. 2.2 Thermal Stress In building applications, thermal stresses that cause fracture generally develop from the temperature differential between the warmer exposed central region of the glass (i.e. vision area) and the cooler edges (i.e. concealed perimeter). As the central region of the glass is warmed by the absorption of solar radiation or by convection, it tends to expand while the cooler perimeter resists this expansion. This temperature gradient across the plate results in "hoop stress" tension along the perimeter. The greater the temperature difference between the central region and the edge, the larger the tension and the greater the likelihood of breakage. Thermal stress breakage is more likely to occur in reflective or tinted heat-absorbing glass than in clear glass. Spandrel glass with an opaque coating is particularly susceptible to thermal breakage. Experimental Results on Thermal Fractures: We produced thermal fractures in 3/16 in. thick annealed glass by placing the glass in the apparatus shown in Figure 2 and heating the exposed central region of the glass with an infrared heat lamp. Our , tests results (Figure 3) show that a temperature difference as low as 42°F across a 1/2 in. recessed edge of clear annealed glass is sufficient to cause fracture in samples with acceptable edge quality. 2.3 Impact Stress Impact stresses result from the loads of very short duration such as those from projectiles or explosions. Most fractures from impact stresses result from vandalism or air-borne missiles, including roofing gravel (4). 2.4 Residual Stress As discussed previously, residual stresses produced by tempering can increase the effective strength of glass. However, these same residual stresses, in the presence of certain impurities, can cause spontaneous fracture in tempered glass. A particularly damaging impurity is nickel sulfide (NiS). This small crystalline inclusion can undergo a thermally activated solid phase transformation with an associated volumetric expansion at some point after the manufacture of the glass. This expansion opens internal flaws and induces tensile stresses around the particle. If the particle lies within the central residual tension zone of the light, the tensile stress associated with particle expansion, coupled with the residual tensile stress already present and the stress concentrating effect of internal flaws, result in spontaneous fracture. It is impossible to predict the exact time of the NiS phase transformation and the subsequent fracture since the transformation time is a function of the total heat energy absorbed by the particle and its chemistry. Some lights break from NiS impurities before they have cooled from the tempering process, but others may break years after manufacture. The probability of NiS impurity fractures in tempered glass can be reduced by subjecting the glass to a "heat soak" process after tempering. In this process, the glass is heated to below its 251 annealing temperature for a prolonged period to impart sufficient thermal energy to trigger the expansion of any NiS that may be present. This process, however, is expensive and not 100% effective. 3. FRACTURE CHARACTERISTICS There are three distinct fracture surface zones adjacent to the point of origin, as illustrated in Figure 4 and Photo 3. The region nearest the origin is a smooth area (mirror), which is bounded by a narrow band of hazy surface (mist), which in turn is bounded by a very coarse (hackled) region. 3.1 Radius of the Mirror Zone — Stress Magnitude The magnitude of the fracture stress can be determined by measurement of the size of the mirror zone surrounding the fracture origin. The relationship between the size of the mirror zone and the stress required to initiate fracture is described by the empirical equation; a f = C/r^, where o* is the modulus of rupture in psi, C and b are experimentally determined constants, and r is the radius of the mirror zone in inches, measured from the point of origin to the mirror/mist interface. For annealed glass used in architectural applications values of 0.5 for b and 1950 for C agree reasonably well with experimental results. Fractures that are initiated at low stress levels due to serious flaws do not show defined mirror, mist, and hackled regions. In this instance the only markings that may be apparent on the fracture surface of the propagating crack are rib-marks, Wallner lines, and occasionally striations, as discussed below. 3.2 Rib-Marks, Wallner Lines and Striations — Stress Distribution and Orien- tation Rib-marks represent the imprint of the crack front at some point during the fracture process. They are formed by a change in the plane of crack propagation (Figure 5). A crack will always remain perpendicular to the direction of the principal tensile stress, so a change in the direction of the principal tensile stress at the crack tip will cause the crack to change its plane or to deflect. Rib-marks are always curved with the concave side toward the origin of fracture. They are used to determine the direction of crack propagation, the location of the crack origin, and the type of stress driving the crack. In bending fractures, rib-marks are asymmetrical to the mid-plane of the glass, with the apex of the curve shifted toward the tension side of the glass (Photo 4). Wallner lines are the result of small undulations in the crack front produced as the crack interacts with elastic stress waves during propagation. Wallner lines, like rib-marks, are always concave toward the origin of fracture, and are used to detect the direction of crack propagation The angle between intersecting Wallner lines can be used to determine the speed of the crack. Striations form as the crack front readjusts to a rotation of the principal tensile stress about an axis parallel to the direction of crack advance. As the crack front attempts to remain perpendicular to the principal tensile stress, it splits into several parallel, non-coplanar fronts (Figure 6). Striations align in the direction of crack propagation, are perpendicular to rib-marks, and often occur where the crack exits the plate, as shown on the bottom of the fracture surface depicted in Photo 10. 3.3 Mechanical Fractures by Bending of Thin Plates Thin plates of glass bent to fracture about their weak axis have the following fracture characteristics. 252 Annealed Glass: The mirror region surrounding the origin of fracture is bounded by regions of mist and hackle along the adjacent surface, but is not bounded on the side toward the middle of the glass (Photos 5 and 6). This configuration is the result of the typical bending stress gradient; i.e., maximum tensile stress along the bottom surface, diminishing to zero at the mid-plane of the glass. Low stress breaks, i.e., fractures at stresses below about 3,000 psi which result from serious flaws, generally do not exhibit mist or hackle on the fracture surface. Wa liner lines, rib-marks and striations are usually apparent and provide ample information to locate the fracture origin and the direction of the bending load. For example, in Photo 4 the rib-mark on the fracture surface of a simply supported beam broken by bending, shows that the tension surface was on the bottom of the plate, therefore the load was applied from the top, and the fracture propagated from left to right. By noting the propagation direction of all crack branches, the fracture can be retraced to the origin. Extremely low stress breaks (i.e., fractures at stress levels below about 1,500 psi, which result from the combination of strength reducing mechanisms such as severe flaws under sustained loads in the presence of high concentrations of moisture), exhibit clear, featureless fracture surfaces. The initial stress concentrating flaw at the origin of fracture can grow slowly to considerable length under these conditions, as shown in Photo 7, until the concentrated stress level reaches a critical value and rapid crack propagation occurs. The static fatigue fracture that produced the fracture surface shown in Photo 7 initiated after 24 hours of exposure to water and 1,000 psi tensile stress. Another indicator of the type and magnitude of fracture initiation stress in bending of annealed glass is the number of crack branches diverging from the origin, and the angle of the crack branching pattern, as shown in Photo 8 and Figure 7. The wider the crack branching angle, the greater the fracture stress. Fractures that initiate at maximum bending stresses below about 6,500 psi usually exhibit no crack branching near the origin. Tempered Glass: In fully tempered glass the fracture surfaces are more complex than described above and somewhat less revealing. Surface defects in tempered glass must penetrate to a depth of about 20 percent of the thickness before the residual tension zone is reached. Once the crack reaches the tension zone, it rapidly propagates through the body of the glass. A great deal of elastic energy is released, a large amount of new surface is produced, and the glass fragments are small, rounded, and relatively harmless compared with the fragments produced by the fracture of annealed or heat strengthened glass. Reasonably flaw-free lights of tempered glass require so much energy to initiate fracture that the glass at the origin is often completely crushed and the fracture mirror area obliterated. The fracture surface of tempered glass broken by bending stress, shows fracture mirrors that fully enclose the origin, and extend only 1/4 to 1/3 into the thickness of the plate (Photo 9). The mirror, mist and hackled regions are symmetrical about the origin. Photo 10 shows some general characteristics of propagating cracks in tempered glass, away from the primary fracture surface. The appearance of mist and hackle along the mid-plane signifies the release of residual elastic energy along the central tension region. Heat Strengthened Glass: The fracture surface characteristics of heat strengthened glass may resemble those of either fully tempered or annealed glass, depending on the magnitude of the fracture stress and the severity of the fracture initiating flaw. 253 The characteristics of high energy breaks (i.e., > 18,000 psi) in heat strengthened glass resemble those of fully tempered glass in that the mirrors have a low profile and are closed and the initial crack plane branches and diverges throughout the body of the glass as the crack front reaches the central tension region. Lower energy breaks (i.e., < 10,000 psi) show characteristics similar to annealed glass fracture, where crack branching occurs, but there is no spontaneous divergence of the fracture throughout the interior tension zone of the glass. Mirrors surrounding surface flaws resemble tempered glass mirrors (Photo I I), but those surrounding corner flaws are similar to annealed glass mirrors (Photo 12). 3.4 Spontaneous Fractures in Tempered Glass When a flaw occurs in the central tension zone of a tempered glass light, the glass may shatter spontaneously without any apparent cause. If the glass fragments remain intact in the opening, the origin of fracture is apparent from the "sun-burst" crack pattern. The two fracture fragments immediately adjacent to the origin are larger than the typical fragments produced and are usually five or six-sided polygons (Photo 13). The fracture surface surrounding the internal flaw is smooth, indicating a low-stress fracture. The smooth surface gives way to mist and hackle along the mid-plane of the light within a distance less than the glass thickness. These same fracture characteristics are apparent where the spontaneous fracture is caused by a nickel sulfide stone. The fracture surface surrounding a NiS stone contains small sharp cracks emanating from the particle (Photo 14) and is smooth, indicating a low stress break. The cracks emanating from the NiS particle resemble static fatigue cracks (Photo 7). The smooth surface disappears close to the origin, and the surface develops the typical tempered glass fracture surface features, as the crack reacts to the residual stress. The NiS particle is a dark sphere, typically about 200 micrometers in diameter, and lies close to the mid-plane of the glass. It can be positively identified by scanning electron microscope analysis, typical results of which are shown in Photos 15 and 16. 3.5 Thermal Fractures The typical thermal crack in architectural glass starts within the recessed perimeter and propagates perpendicular to the edge until it reaches the vision area. At reasonably high fracture stresses in annealed glass (i.e. > 6,000 psi) the fracture surface contains bands of mist and hackle completely enclosing the mirror (Photos 17 and 18). The full enclosure of the mirror reflects the uniformity of the thermal stress field through the thickness of the plate, as opposed to the abrupt stress gradient associated with bending stresses. At high stress, concentric bands of mist and hackle develop, showing a cyclic build-up and subsidence of fracture energy as the fracture propagates. The location of the most violent hackling coincides with the development of a crack branch (Photo 18). Thermal fracture surfaces away from the fracture origin become smooth. The uniformity of the thermal stress field through the thickness of the plate produces rib-marks that are symmetrical about the mid-plane (Photo 19) as opposed to the rib-marks associated with bending stresses (Photo 4). Striations are generally absent on the fracture surfaces of thermal breaks. The number of thermal crack branches increases with the fracture stress but the angle of crack branching is not an indication of stress level as is the case for bending type fractures. Typically, the thermal crack wanders in smooth curves across the plate (Photo 20). When mechanical loads are superimposed, cracks may turn more abruptly (Photo 21). 254 3.6 Impact Fractures High velocity, low mass projectiles generally produce a local fracture at the contact point which flares out to form a cone-shaped spall on the side of the plate opposite from the contact surface. From this cone, radial and ring cracks propagate outward across the plate (Photo 22). With low velocity, high mass projectiles the ring and radial cracks may develop without the local cone fracture at the contact point (Photo 23). The typical crack patterns for high velocity impacts of heat strengthened and tempered plates are shown in Photos 2k and 25, respectively. Heat strengthened plates show radial cracks but few ring cracks. Fully tempered plates fragment completely, but a vague sun-burst pattern emanating from the origin is still apparent. The fracture surface at the origin in heat strengthened and fully tempered glass is often obscured by local fragmenting. 4. CONCLUSIONS The traditional methods of design, manufacturing and installation of glazing are not adequate to meet the demands of current architectural glazing applications. It is incumbent upon the designers, manufacturers and installers of architectural glass to familiarize themselves with all of the factors that determine glass durability, to improve the predictability of glass installations. The findings presented in this paper suggest some of the factors to consider, such as: (i) Designers cannot accurately predict glass behavior without considering the strength degrading effects of long-term loads, environment, and surface defects. The combination of sustained loads, moisture exposure, and severe surface flaws, can reduce glass strength to 10% of its "as-received" strength, (ii) Designers should use discretion and impose larger than normal factors of safety if significant uncertainties with respect to service loads and glass strength exist. The unique aspects of each building and the consequences of breakage should be assessed for each installation. Glass thickness selection charts published by the manufacturers should be employed only as guides, (iii) Scored edges containing shark teeth are seriously weakened. We recommend against glazing lights containing shark teeth, of any length, along their edges. These lights should never be glazed in areas receiving direct solar exposure, (iv) Designers must consider the thermal characteristics of glazing enclosures and interior fixtures, to eliminate significant temperature differences between the exposed area and recessed perimeter of the glass. Heat strengthening should be considered for glass in moderate thermal exposures, (v) When tempered glass is required, designers should consider using the heat soak process to reduce the risk of spontaneous fractures resulting from NiS impurities. A more practical alternative may be to use heat-strengthened glass which is significantly stronger than annealed glass but is not susceptible to NiS fractures, (vi) When fractures do occur, every effort should be made to analyze the fracture, determine the cause of the break, and make this information available to the public. ACKNOWLEDGEMENTS The research for this paper was conducted under the guidance of Professor Donald R. Uhlmann at the Massachusetts Institute of Technology, with funding by the National Science Foundation. Simpson Gumpertz & Heger Inc. also provided financial and editorial support for this paper. The author thanks the above people and organizations for their efforts and support. 255 REFERENCES 1. Charles, R. J., " Static Fatigue of Glass," Journal of Applied Physics , Vol. 29, 1958, pg. 1549. 2. Kingery, W. D., Bowen, H. K., Uhlmann, D. R., Introduction to Ceramics , (2nd edition), John Wiley and Sons, Inc., 1976, pg. 796. 3. Stahn, D., "Thermal Stresses in Heat-Absorbing Building Glass", Thermal Stresses in Severe Environments , edited by Hasselman and Heller, Plenum Press, New York, 1 980, pg. 320. 4. Minor, J. E., and Beason, W. L., "Window Glass Failure During Wind- storms", The Glass Industry , February 1 975, pg. 12-15. 256 Table I — Effect of Surface Flaws and Environmental Stress Corrosion on the Fracture of Annealed Glass Sample Condition In Air In Water Strength Reduction Due To Environment Average Fracture Average Fracture Stress (psi) Stress (psi) As received (no induced flaw) 9,460 (17)* 7,550 (6) 20% Sandblasted 9,170 (12) 6,750 (5) 26% Surface "pin" flaw (produced by sharp tool) 6,660 (12) 5,550 (6) 17% Corner "chip" flaw (produced by blunt tool) 4,350 (12) 3,400 (6) 22% Strength reduction due to induced flaws 54% 55% 64% Number of specimens shown in parentheses. Table 2 — Effect of Shark Teeth on Glass Edge Strength Industry Standard Edge Condition Average Strength of Edge (psi) Good No shark teeth No corner damage Acceptable Shark teeth — penetration about 1/2 width of edge Poor Damaged by chipping edge 10,500 (5)* 4,875 (6) 4,590 (6) Number of specimens in parentheses. 257 COMFHESSION (a) TEMPERED STRESS DISTRIBUTION COMPRESSION « — ( — » TENSION fb) BENDING STRESS DISTRIBUTION (c) SUPERIMPOSED LOADS OF 'ol 8. (fc) FIGURE I: MECHANICAL & RESIDUAL STRESS DISTRIBUTIONS t 3 THERMOCOUPLE TEMPERATURE READ-OUT5 £1 | BAOIANIHtAT 60 70 90 100 FIGURE 2 THERMAL TEST APPARATUS FIGURE 3: O VS. AT, FOR ANNEALED GLASS FIGURE 4 TYPICAL FRACTURE SURFACE CHARACTERISTICS 258 FIGURES : FORMATION OF RIB5 FORMATION OF STRIATIONS BRANCHING ANGLE (DEGREES! FIGURE 7 : ANGLE OF CRACK BRANCHING VS. FRACTURE STRESS (cr ( ) FOP ANNEALED GLASS FRACTURED BY BENDING 259 Photo 1 - "Clean" Scored Edge (4X Magnification) Photo 2 - Scored Edge with Sharks Teeth (4X Mag.) Photo 3 - Typical Fracture Surface (7X Mag.) Photo 4 - Rib-Mark from Crack Propagated by Bending Stress (Tension on Bottom of Plate) (7X Mag.) Photo 5 - Surface Origin Bending Fracture (5X Mag.) Photo 6 - Edge Origin Bending Fracture (5X Mag.) 260 Photo 7 - Static Fatigue Crack Low Stress Break Photo 8 - Crack Branching from Bending Fracture (Actual Size) Photo 9 - Typical Fracture Surface Markings in Tempered Glass (3X Mag.) Photo 10 - Secondary Fracture Surface in Tempered Glass (6X Mag.) Photo 1 1 - Fracture Surface of Heat Strengthened Glass with the Fracture Origin at a Surface Flaw (6X Mag.) Photo 12 - Fracture Surface of Heat Strengthened Glass with the Fracture Origin at a Corner Flaw (4X Mag.) 261 Photo 13 - Plan View of Fracture Origin in Tempered Glass (Actual Size) Photo 14 - Fracture Surface at NiS Particle (Light color of particle due to reflected light) (10X Mag.) Photo 16 - SEM Display Photo 15 - NiS fbrticle (65X Mag.) Photo 17 - Fracture Surface of Ther- mally Broken Annealed Glass -Surface Origin (4X Mag.) Photo 18 - Fracture Surface of Thermally Broken An- nealed Glass-Corner Origin-Shows Coinci- dence of Hackling & Crack Branching (4X Mag) 262 Photo 19 - Typical Rib-marks on Thermal Fracture Surface (7X Mag.) Photo 20 - Typical Thermal Crack Patterns (2 Actual Size) Photo 22 - Typical Impact Fracture Pattern in Annealed Glass fa Actual Size) Photo 21 -Thermal Crack Deflected by Load at ® (2 Actual Size) 1 J I j : Photo 24 - Typical Impact Fracture Pattern in Heat Streng- thened Glass fa Full Size) Photo 23 - Impact Fracture Pattern Produced in Annealed Glass by High Mass/Low Velocity Projectile (i Actual Size) Photo 25 - Typical Impact Fracture Pattern in Tempered Glass (i Actual Size) 263 INTERACTION OF CURTAIN WALLS AND METAL EDGE BANDED INSULATING GLASS UNITS M. S. Zarghamee and T. A. Schwartz Simpson Gumpertz & Heger Inc. Cambridge, Massachusetts 02138 Abstract: This paper presents the results of a study designed to discover causes for the loss of metal edge bands from insulating glass units during service. We have studied the effects of the movements between a dry-sealed pressure glazing curtain wall and an insulated glazing unit resulting from normal changes in environmental conditions. We have shown theoretically and experimentally that at reasonable clamping forces of the curtain wall acting on the edge band of the glazing unit, the interaction of the curtain wall and the glazing unit does not result in the loss of the edge band from the glazing unit. However, if the clamping force is sufficiently small, cyclic movements between the curtain wall and glass will push the metal edge band off the glazing unit. Key Words: Curtain wall; insulating glass units; pressure glazing; edge band; interpane spacer. I. Introduction: Experience with certain glazing units that failed, either by seal failure or glass fracture, shows that although the glazing units were installed initially with edge bands on, the failed units lost their edge bands in the course of service. Furthermore, units that lost their edge bands had damaged edges on the number I and 4 surfaces, i.e. outermost edges (see Fig. I for terminology). The removal of the edge band results in the loss of perimeter compression required to maintain the air tight seal in most insulating glass units. For many glazing units the loss of the edge bands may result in large spacer migrations, seal failure, or glass fracture. Variations of environmental conditions, such as wind, temperature, and baro- metric pressure, affect the behavior of double glazing units in two different ways: (i) the glazing unit responds directly to the changes of environmental conditions, and (ii) the changes in the environmental conditions result in the deformations of the curtain wall system; these deformations apply forces and impose displacements on the glazing unit that may result in the failure of the glass. This paper presents the results of a theoretical and experimental study designed to discover causes for the loss of metal edge bands from insulating glass units glazed in a "dry-seal" pressure glazing curtain wall system. We show that the direct effects of changes in environmental conditions do not result in sufficiently large interpane pressure to push the edge band off the glazing unit by outward movement of the interpane spacer. Considering the cyclic movements of the curtain wall relative to the glass and the clamping force of the curtain wall, we then demonstrate the mechanism by which the edge band can separate from the glazing unit. 264 The theoretical analysis is based on considerations of relative frictional forces developed between the rubber gasket, the metal edge band and the glass. We conducted laboratory experiments on a portion of a curtain wall, using various clamping forces, and found the results to be in good agreement with the theoretical analysis. 2. Direct Environmental Effects on Insulated Glazing Units: The interpane pressure, deflections, and stresses in a sealed double glazing unit, subjected to temperature, wind and barometric pressures different from those at the time of sealing, may be approximated by a model developed by Solvason [I] . Solvason considers only the linear bending theory of glass plates and neglects the sizable nonlinear effects of in-plane stretching resulting from the bending deflections of thin plates. Including the effects of in-plane stretching [2], the variations of interpane pressure with temperature, wind, and barometric pressure are shown in Fig. 2. Note that the interpane pressure for the range of temperatures, wind and the barometric pressures shown, rarely exceeds 0.2 psi. This pressure, although it may cause significant stresses in the glass panes, is not sufficient to force the interpane metal spacer outward so as to push the steel edge band off the glazing unit. We measured the clamping force of the edge band and found it to be about 8 lbs per lineal inch and the coefficient of static friction of the edge band against the glass to be 0.19; therefore, an interpane pressure differential of 6 psi is required to push the interpane spacer outward and the edge band off the glazing unit. 3. Movements of Curtain Wall Relative to Glass: Cyclic movements of curtain wall components relative to the glass may be caused by the wind effects on the structural frame or by the thermal expansion and contraction of the curtain wall due to solar radiation. The response of the structural frame of a building to wind action consists of a static and a dynamic component. The static component is the response of the structure to the mean wind; whereas, the dynamic component is a random lateral vibration of the building caused by the wind gust. The magnitude of the dynamic component of the response is usually a significant portion of the total structural response. Buildings are usually designed for lateral drift ratio of at least 0.002; that is, the sum of the dynamic and static components of the structural response under the action of design wind, with a recurrence interval of 50-100 years, is expected to result in lateral deflections less than 0.2 percent of the height. Lateral displacement of a structural frame is partly a result of column shortening and lengthening that does not produce racking of the curtain wall. However, if the structure is not excessively tall and slender, a major part of the lateral displacement results in racking of the curtain wall system. For a wind having a recurrence period of one year acting on a properly designed structural frame of a building, an amplitude of the dynamic component of the drift per story of 0.01 to 0.05 in. is not unusual. As the curtain wall is exposed to solar radiation, its temperature increases until an equilibrium state is reached at which point the rate of energy absorption is equal to the rate of energy loss to the surroundings through radiation, convec- tion, and conduction. Both calculations and measurements indicate that the exposed parts of the curtain wall system frequently get as much as 60 F warmer than the temperature inside the building. The glass temperature also increases 265 but not to the same extent. If the curtain wall system does not allow for free thermal expansion of its horizontal members, such expansions will result in movements of the vertical mull ions with respect to the glass, the magnitude of which is L (a| AT| - <*2 AT2)/2 where L is the width of glazing unit, a| and a-2 are the coefficients of thermal expansion of the aluminum and the glass, and AT | anc j AT2 are the temperature differentials of the aluminum and the glass. For a 60-inch wide glazing unit, an average relative thermal displacement of the mull ion with respect to the glass of 0.02 - 0.05 in. is considered typical. Therefore, cyclic movement of the curtain wall relative to glass of approxi- mately 0.02 - 0.05 in. may be caused by the changes in the environmental conditions that frequently occur over the life of the structure. 4. Effects of Clamping Pressure on Relative Movements Between Curtain Wall and Glass: In a dry-seal pressure glazing curtain wall system, the glazing unit is held by the clamping force of the curtain wall which may be applied through clamping bolts. If the clamping force is sufficiently low, it is shown below that cyclic movements of the mullions will result in ratchet-like slippage at the edge band/gasket interface which results in the eventual removal of the edge band from the glazing unit. However, if the magnitude of clamping pressure is sufficient and if the gasket is properly positioned, cyclic movement of the mullion relative to the glass results only in a similar cyclic movement of the edge band with no net slippage at the edge band/gasket interface. Since the amplitude of the cyclic movement is not more than 0.05 in. and the edge band has about a 3/8 inch bite, the cyclic movement does not cause the edge band to come off the glazing unit. Theoretical Results: Let P be the compressive force that the curtain wall gasket, at any instance, applies to the glazing unit. P depends not only on the initial clamping force P Q but also on the deformation of the gasket that may increase or decrease the initial clamping force. The gasket is supported partially on the steel edge band and partially on the glass. Let P rs and Prgbe the two parts of P that are applied to the steel edge band and the glass (see r ig. 3); that is, P = P +P (I) rs rg In addition to the clamping force of the curtain wall, the edge band also exerts a clamping force P e to the glass. Therefore, the compressive force at the edge band/glass interface, P S g ' s P =P +P (2) sg rs e If we assume first that during the movement of the mullion relative to the glass, slippage occurs at the edge band/glass interface, which also requires slippage between the gasket and the glass, the frictional force at the edge band/glass interface would be F. =F c +F rrt (3) I sg rg If we assume then that the slippage occurs at the gasket/edge band interface, which also requires slippage between the gasket and the glass, the frictional force at the gasket/edge band interface would be F 2 - F rs + F rg <*> Since slippage occurs at the interface with least resistance to motion, for slippage to occur at the gasket/edge band interface, we must have F 2 then Eq. (6) requires that "rs P rs-% F V%< P rs + P e ) V P p < — sg — e_ - p ,,* rs_ "rs^sg e W where C is a coefficient depending only upon the coefficients of friction and is approximately 0.5*. Therefore, the necessary condition for slippage to occur in the gasket/edge band interface is that the part of the curtain wall compressive force which acts on the edge band should be less than a limiting value P c = C P e . The deformation of the gasket, or its movement relative to the edge band, change the compressive force that the gasket applies to the glazing unit. Such deformations or movements caused by an inward movement of the mull ion relative to the glass and in the plane of glass tend to reduce the compressive force; whereas, those caused by an outward movement of the mullion, as it pushes the gasket in the wedge formed between the mullion and the leg of the edge band, tend to increase the compressive force. If the initial clamping force is high that even after its reduction by the deformation of the gasket caused by an inward movement of the mullion, the compressive force P is greater than P c , then there will be no slippage at the gasket/edge band interface and all movements at the edge band/glass interface will be reversible. If the initial clamping force P — P c or, if after the deformation of the gasket the compressive force P acting on the glazing unit does not exceed P c , then any additional inward movement of the mullion will result in the gasket sliding over the edge band and thus reducing the compressive force of the curtain wall further. When the direction of motion reverses, the compressive force on the edge band starts to increase as the gasket deforms in the opposite direction and slides over the edge band until P exceeds P c . At this point the gasket will stop sliding over the edge band and the edge band will start to slide out along with the mullion. The outward movement of the edge band is also facilitated by an outward force exerted to the leading edge of the edge band as it encounters the gasket sliding over it. Our tests show that y rs = 0.57, Mrg = 0.54, and y S g = 0.19 for static friction and 0.14 for dynamic case. ^267 A combination of these effects in a complete cycle of movement of the mullion results in a net outward migration of the edge band relative to the glass. If this cycle is repeated many times, the edge band will be forced off the unit. Experimental Results: In this section, we present the results of our experiments on a curtain wall/glass mockup. We cut two small sections from the edge of an insulating glass unit (with edge bands in place). We epoxied a threaded rod into the space between the two lights of glass of one section, perpendicular to the edge. The sections were clamped into a typical two-piece aluminum curtain wall mullion, with neoprene gaskets at the curtain wall/glass interface (see Fig. 4). We fixed the assembled mockup in a rigid frame, so that the one glass unit with the threaded rod protrusion could be moved, by mechanical means, slightly within the mullion section to simulate expected relative movements between the glass and curtain wall. The glass was held in the mullion by bolts which, when tightened, applied the compressive force to the glass through the extruded aluminum glass retainer. We clamped the glass with the retainer at three force levels, 22, 4, and 2 in. -lbs, of torque on the bolt which corresponds to average clamping forces of 73, 13, and 7 lbs/in. At each clamping force we displaced the glass test section relative to the mullion gasket by approximately 1/16 in. outward and then 1/16 in. inward, to complete a cycle. During and after each cycle we observed the relative position of gaskets, edge band and glass, and recorded the net movement of these components, if any. At mullion clamping forces associated with mullion bolt torques of 22 and 4 in.- Ibs, the outward 1/16 in. displacement of the glass resulted in slippage of the glass at the glass/gasket and glass/edge band interfaces. There was no slippage at the gasket/edge band interface. On the return glass displacement, the cycle reversed, resulting in no net translation of any component. At the mullion clamping force associated with the 2 in.-lbs mullion bolt torque (i.e. about 7 lbs/in.), the outward glass displacement resulted in the same slippage as described above. However, on the return inward movement, the gasket slipped from the edge band. At the completion of this cycle the edge band was displaced toward the edge of the glass unit. In the next cycle, the edge band was displaced further toward the edge of the unit (see Fig. 4). After a few cycles the edge band came off the unit in a "peeling" mode. As the edge band peeled from the edge of the unit it chipped the edge of the glass. This edge damage may reduce the glass strength significantly [3 J and lead to failure under expected loadings. 5. Conclusion: The above tests confirm the results of the theoretical analysis presented in this paper, and show that at low curtain wall clamping forces, relative cyclic movement between a metal edge banded insulating glass unit and the curtain wall results in gradual removal of the edge band from the unit, edge damage to the glass, and loss of the interpane seal. These results prove that curtain wall gasket clamping pressures on glass of about 10 lb/in. or greater are required to maintain the position of the metal edge band on an insulating glass unit. This requirement conflicts with the recommendations of the major U. S. glass manufacturers who require that clamping pressures on 268 glass be maintained at or below 10 lbs/in. and clamping pressures on metal edge bands be kept at I lb/in. or less [4] . The results emphasize the need to maintain proper curtain wall clamping pressure in curtain walls containing edge banded insulating glass units. REFERENCES 1. Solvason, K. R., "Pressures and Stresses in Sealed Double Glazing Units," National Research Council of Canada, NRCC 14167, Ottawa, August 1974. 2. Zarghamee, M. S., "Environmental Effects on Insulating Glass Units," paper in preparation. 3. Schwartz, T. A., "Practical Considerations in Glass Fracture," to be pre- sented at the Second International Conference on Durability of Building Materials and Components, National Bureau of Standards, Washington, D.C., September 1981. 4. "Recommended Glazing Practices for Reflective Insulating Units over 20 sq ft in Area", PPG Industries, Pittsburgh, PA. 269 Glass 5urface No.—> Edae txxnd r Spacer Glass Figure I - Terminology 270 *0.I0 +0.05 I i I C t c 0.00 •aos -0.10 -o.is ■020 m^scs^z \5PK _a_Qpsf— -"""I A< 5»a Level AtlZSOFt Elmv. SO Ambien-6 Temp. Degress f /SO Figure 2 - Variation Of Interpane Pressure Differential With Temperature, And Wind And Barometric Pressures For A 58 x 67 x I Inch Nominal Insulation Glass Unit With 0233 Inch Thick Light, Sealed At Sea Level At 65> F Temperature. >271 Mul lion S+eel Edge Band Rubber Gatkei Glass Figure 3 - Interface Forces 272 Clamping Bolt Epoxy 77777/ Fixed Cyclic Movement Imposed Figure 4 - Test Set Up And Position Of Edge Bond At Low Clamping Torque After Few Cycles v 273 CORROSION PROCESSES IN BUILDING INSULATION SYSTEMS James M. Pommersheim and John Lobo Department of Chemical Engineering Bucknell University Lewisburg, Pennsylvania 17837 James R. Clifton Materials and Composites Division Center for Building Technology National Bureau of Standards Washington, D.C. 20234 Abstract: Mathematical models have been developed for the corrosion of metal building service elements (such as electrical receptacle boxes and pipes) in contact with thermal insulation. The amount of corrosion and corrosion rate depend on the amount of condensation, the rate of drying and the leaching rate of impurities from the insulation. Condensation on a metal element is more likely when it is colder than the surrounding insulation or when humidities are high. The amount of condensation within an electrical receptacle box increases with increasing flow rate to a maximum after which it decreases. Higher relative humidity and low temperature result in longer drying times. Higher drying rates and higher leaching rates result in increased corrosion rates. Patch corrosion was more sensitive to changes in drying and leaching rates than uniform corrosion, and increases more rapidly with time. Key words: Condensation; corrosion; electrical receptacle boxes; mathematical models; patch corrosion; thermal insulation; relative humidity; uniform corrosion. 1. NATURE OF CORROSION IN INSULATION SYSTEMS Thermal insulation has lower thermal efficiency and dimensional stability when wet than when dry. Condensed water can also cause corrosion. The water can leach out binders, flame retarders, and other formulating agents from the insulation. Such impurities, in the presence of dissolved oxygen can initiate and accelerate corrosion of metal service elements which are in contact with the insulation. This includes water lines such as copper tubing or iron pipe, aluminum and copper wiring and the galvanized steel used in electrical receptacle 274 boxes. Corrosion is the destruction of a metal surface usually by electrochemical processes. The driving force for corrosion is the free energy difference between the corroding metal and the corrosion products. The rate of corrosion is a function of the temperature and the electromotive force as determined by the galvanic series. Corrosion can occur when water, oxygen and an electrolyte are simultan- eously present in the vicinity of the metal surface. The electrolyte can be a foreign ion such as chloride or sulfate which comes from the wetted insulation, or it can be the corrosion products themselves. In some instances the surface of the metal is not perfectly clean and accumulation of impurity ions may help initiate localized corrosion. Crevice corrosion occurs in confined spaces where oxygen has limited access, while galvanic corrosion occurs at places where dissimilar metals join [1]. Corrosion can also occur as a degradation reaction as in the action of sulfate ion on copper to form CuSO, . For iron-based metals the most common corrosion half-cell reactions are: Anodic Reaction : (oxidation) Cathodic Reaction : (reduction) Fe -*■ Fe + 2e H + 1/2 + 2e 2 2 2 (OH) Aluminum can undergo similar reactions, especially in the presence of chloride ion which destroys the passive oxide film that usually protects aluminum from corrosion. Copper will generally corrode by contact of this surface with acidic solutions. The anodic and cathodic reactions occur at different sites on the metal surface. Anodic reactions typically occur in pits or crevices. Dissolution of the metal as reaction proceeds causes pits to enlarge. The electrons released in the anodic reactions travel through the metal to the location of the cathode. This completes the circuit. Driven by the thermodynamic potential, the anode dissolves and releases electrons which generate hydroxyl ions. Electron flow is retarded (or equivalently , circuit resistance is increased) by several mech- anisms. These include the build-up of oxide layers or layers of other solid corrosion products, such as Fe(OH).. or Fe^Oo, the formation of an electrical double layer adjacent to the metal surface [2], and the absorption or formation of gases such as hydrogen or oxygen at the cathode surface (activation polarization) . 2. FACTORS AFFECTING CORROSION IN INSULATION SYSTEMS For a given metal-insulation system corrosion rates will vary with: a) the temperature of the system 275 b. the pH of the electrolyte c) the thickness and permeability of the corrosion products d) the type of corrosion which occurs at the metal surface e) the type of condensation which occurs f) the nature, concentration and distribution of the leachable impurity within the insulation g) drying rates In general higher temperatures give higher corrosion rates because the electrochemical reactions are accelerated and because the diffusion rates of ions are increased. For any given metal and solution in contact with it the pH, electromotive potential and local oxygen concentration will determine the tendency to corrode. Such thermo- dynamic information is conveniently summarized by a Pourbaix diagram [1]. Using this diagram, one can determine whether a metal surface is in a region of immunity where the tendency to corrode is nill, in a region where the tendency to corrode is high, or in a region where the tendency for corrosion may still exist, but where a protective or passive film forms, greatly reducing the rate of corrosion. Thicker and less permeable corrosion products decrease corrosion rates since they slow and impede the passage of diffusing species to the corrosion site. In general the diffusion of water will not be rate limiting since the corrosion occurs in an aqueous medium. On the other hand the diffusion of dissolved oxygen and ions may effect the corrosion rates because their concentrations are often low and they may be consumed in the corrosion reactions. Sweating type condensation can also occur directly on the inside of electrical receptacle boxes under suitable conditions of temperature, pressure and relative humidity. It is well known that a significant amount of air can pass through such a box around the wiring or through other holes in electrical boxes. It is recommended that the face-plates in wall outlets be gasketed in order to prevent convective heat losses. In general building contractors do not seal receptacles from outside or inside air flow. Condensation within the box is likely in the winter when the walls and insulation are colder than building interiors. Warm moisture laden air brought in through the face-plate will condense there. In the summer, warm, humid air brought in from outside can condense within the cooler electrical box located next to an air- conditioned building interior. Whether condensation occurs will depend on the temperatures and humidities of the inside and outside air as well as the direction of the air flow (into or out of the receptacle) . This in turn depends on the pressure distribution within the building. In the second type of condensation (wetted ring type) the moisture condenses in the insulation around the object as well as in the immediate vicinity of the object. The metal again serves as a focus for moisture condensation but the moisture build-up within the insulation 276 is much greater. For a cold metal pipe the condensation can assume the form of a wetted ring centered around the pipe. For insulations having a low porosity such as perlite or polyurethane the drainage of water will be low. The surface tension forces within such insulations are larger than gravitational forces, and the water is held up in the narrow pores against the downward pull of gravity. Some insulations may show intermediate behavior between the wetted-ring and the sweating types of condensation. Fiberboard, for example, which has an intermediate porosity, can hold water in a wetted ring pattern when the condensation occurs in discrete liquid bridges separated by moisture barriers. Such an internal structure is likely to cause the shift of anodic and cathodic sites over the surface with time, especially when the corrosion products become detached from the surface, due, e.g., to differences in density between the reactants and corrosion products. This can permit substantially uniform corrosion to occur. In some cases the corrosion products will build-up uniformly. This type of corrosion is termed "uniform corrosion". As time passes the thick- ening corrosion products provide large resistance to diffusion so that corrosion rates diminish. On the other hand corrosion can occur in separate patches on the surface as corrosion "rings" advance into the uncorroded metal. This is here termed "patch corrosion". In patch corrosion the oxygen lean interface between the corroded and uncorroded metal is the anode while the cathode is distributed across the uncorroded metal surface. Because of the release of hydroxyl ions most of the surface will be basic. In extreme cases of patch corrosion, where there is little surface area for lateral growth, the patches may form an interconnected web. Corrosion growth patterns at later time resemble that found with uniform corrosion. Condensation occurs on or near metal objects in contact with insulation when the metal surface is colder than the dew point temperature in its vicinity. There are basically two types of condensation which can occur around metal objects. In the first (sweating type) condensation occurs at the surface in drops which run off due to the action of gravity. Liquid hold-up is relatively low, although new liquid is always being brought to the surface. The "sweating" of a pipe in the open air is an example of this type of condensation. The same kind of condensation can occur in an insulation system. This type of condensation is more likely to occur with insulation systems that are relatively porous. Examples include matted fiberglass, or granular insulations like cellulosic fluff or vermiculite. Such materials cannot retain the water in their pore structure against the downward pull of gravity. The nature, concentration and distribution of leachable impurity within the insulation will also affect the corrosion process. Different impurities will accelerate the corrosion of different metals 277 at different rates. Higher concentrations of impurity usually lead to larger corrosion rates, as does larger concentrations of impurity located in the insulation adjacent to the metal surface. In addition, since condensation occurs there, the potential for leaching of the impurity from the binder will be greater closer to the metal. Leaching rates will be proportional to the surface area (fiber area) of the insulation to which the condensed water is exposed, as well as the concentration driving force between the suface of the insulation and the adjacent liquid. The concentration of leachable impurity in solution will increase with time with wetted-ring type condensation. If conditions of temperature and relative humidity change, the insulation can dry. The rate of drying will indirectly affect the corrosion since the lowered volumes of solution will further increase the concentration. Less corrosion might be expected for the case of sweating type condensation since the metal is being cleansed somewhat with fresh solution. In this case concentrations at the metal surface would be more likely to remain constant. 3. REFERENCES 1. Uhlig, H.H. , Corrosion and Corrosion Control , 2nd edition, John Wiley & Sons, Inc., New York, 1971. 2. Borkris, ed. Electrical Double Layer . 278 THE PHOTODEGRADATION OF WOOD DURING SOLAR IRRADIATION E. R. Miller Building Research Establishment Princes Risborough Laboratory, Princes Risborough Aylesbury, Buckinghamshire, England, UK H. Derbyshire Building Research Establishment Princes Risborough Laboratory, Princes Risborough Aylesbury, Buckinghamshire, England, UK Abstract: Photodegradation by sunlight is an important factor in the surface breakdown and loss of colour of wood exposed out of doors. It increases maintenance costs, since weathered wood has poor paint-holding properties. Clear varnishes and semi-transparent stains break down prematurely on wood because light passing through the film degrades the substrate. There is a need for increased information on the photo- degradation processes in wood and on methods of preventing them. An investigation is described of the effectiveness of different spectral regions in sunlight in degrading wood. It involved monitoring the loss in tensile strength of thin wood veneer strips exposed to sunlight behind filters, scanning electron microscopy and chemical analysis. It was shown that ultraviolet light is highly active in degrading wood, but that a large contribution to the rate of breakdown results from visible light. The loss in strength is associated with a light-induced depolymerisation of lignin and cell wall constituents, and to the subsequent breakdown of the wood microstructure. The involvement of visible light in the photodegradation of wood is of great significance in imposing a limitation in principle on the performance of natural wood finishes. It points the need for the development of pretreatments which stabilise the wood against photo- degradation. Key Words: Photodegradation; weathering; wood; paint; natural finishes; tensile strength measurements; scanning electron microscopy. 1. INTRODUCTION In recent years the poor durability and consequent high maintenance costs of surface coatings on exterior wood have been a source of general concern. The difficulties stem from a number of causes, including 279 100 r c o to to E c 2 300 350 400 600 800 1000 1400 Wavelength - rnji Figure la Spectral transmission of filters 1800 2200 lOil 300 350 400 600 800 1000 1400 1800 Wavelength - itim 2200 Figure lb Spectral solar power transmitted "by filters 280 design details of some modern buildings and timber components -which give insufficient protection against the weather, the use of unsuitable coatings, and the failure to seal the end-grain of permeable species against water entry. Paint maintenance problems are compounded by the fact that the surface of wood exposed to sunlight becomes degraded photochemically and its paint holding properties are impaired. The durability of clear and semi-transparent finishes on wood is reduced by phot odegradat ion of the substrate caused by light transmitted through the film. Previous work on the surface breakdown of wood during weathering has been reviewed by reference (l). The objectives of the present work were to provide information on the rates of photodegradation of exterior wood, and on the relative effectiveness of different spectral regions in sunlight in causing breakdown. 2. EXPERIMENTAL The work has involved scanning electron microscopy and chemical analysis, but the main criterion adopted as a measure of photodegrada- tion during solar irradiation is the loss in tensile strength of wood strips about 150 urn in thickness. The wood species used were lime {Tilia Vulgaris) and Scots pine {Pinus sylvestvis) . The strips were exposed to sunlight in shallow metal trays; they were shielded from rain by opaque or various transmitting filters but open to air circulation.. The spectral transmittance curves for the filters are shown in Figure 1. The samples were exposed at ^5 to the vertical, facing south at Princes Risborough, England. The poly (phenylene oxide) technique reported in reference (2) was used to monitor continuously the levels of solar radiation incident on the filters. The technique is sensitive primarily to ultraviolet and radiation is expressed in black lamp equivalent time (BLET) units. Measurements showed that there was a linear relationship between BLET and total solar radiation during the exposure period. 1 BLET = 6 x 10^ J m" 2 = IT W/h m~ 2 . It can be seen from Figure 1 that differences in total solar radiation transmitted by the filters are small, and that overall absorption of energy by the filters is low. Manifestations of breakdown were accordingly related to accumulated radiation dose in BLET incident on the filters. Tensile testing was carried out on a Model 1026 Instron using a 500 N load cell, an elongation rate of 0.5 mm/min and a gauge length of 50 mm, 281 500 1000 1500 Exposure - BLET 2000 2500 Figure 2a Loss of tensile strength during weathering, pine 500 1000 1500 2000 Exposure - BLET 2500 Figure 2b Loss of tensile strength during weathering, lime Filter type Control % Opaque • P4 ® P3 P2 O G1 A Fully exposed 282 3. RESULTS AND DISCUSSION Tensile strength values of weathered strips expressed as a percentage of the original value are shown in Figure 2. The graphs show the very rapid loss in tensile strength that occurred, particularly for unshielded strips which after nine weeks exposure (1050 BLET) were reduced to 22 and 10 per cent of their original strength for pine and lime respectively. For strips shielded by filters the greatest strength loss occurred behind Gl, which transmits the widest range of solar wavelengths, and the least behind ~Pk which transmits no ultraviolet wavelengths. The tensile properties of strips exposed behind opaque (BL) filters were essentially constant throughout, indicating that environmental influences were not contributing. The tensile strength data presented in Figure 2 may be adequately represented by a first order decay process with a rate constant propor- tional to the radiation dose, according to a relationship. -kd a = o . exp o r where a is the tensile strength after a dose d, o Q the original value of tensile strength, and k is the rate constant in units of (BLET) - . The value of k calculated from the gradients of log a against d is shown in Table 1. Table 1 First order rate constants for strength loss Filter First order rate constant (k) BLET -1 Pine L: Line Gl 8.1+ -k x 10 9.5 x 10 P2 7.1 -h x 10 7.7 x 10~ P3 k.h x 10~ U.6 -h x 10 pU 3.6 x 10~ U.6 -1+ x 10 BL 0.27 x 10~ 0.23 -h x 10 It will be seen that wavelengths longer than U00 nm transmitted by Filter Pk have made a significant contribution to degradation. Subse- quent work has shown that the range of active wavelengths extends to about 5^+0 nm for pine but to intermediate wavelengths for lime. 283 Figure 3 SEM photomicrographs of typical failure in tension 3a Pine, unweathered, dry-tested 3b Lime, weathered (2500 BLET, filter Gl), dry-tested 3c Pine, weathered (1500 BLET, filter Gl ) , wet -tested 3d Lime, weathered (1500 BLET, filter Gl), wet-tested 284 Changes in wood constituents Scanning electron microscope studies made during this work showed that breakdown of the lignin-rich middle lamella region is an early- consequence of weathering, as reported by other workers (l). In order to assess effects on cellulose viscosity, measurements were made on holocellulose prepared from fresh and weathered wood (Table 2). Table 2 Effect of weathering on cellulose disperse viscosity (radiation dose 1000 BLET; Filter Gl) Reduction in tensile Cellulose disperse Species strength after exposure viscosity (cp) {%) Unweathered Weathered Pine 72.1 63. h 9-0 Lime 82.9 *+3 . h k.Q It is evident that strength loss is associated with a pronounced fall in cellulose viscosity resulting from a depolymerisation of the cellulose constituent. The work has shown that solar wavelengths active in producing breakdown lie in the ultraviolet and visible region, and that infrared radiation on its own is not effective. It is believed that under the exposure conditions thermal degradation and hydrolysis can be ruled out, and that the main mechanism involved in the breakdown of the cellulose is photosensitised degradation. This depends on the presence of compounds which absorb long wave ultraviolet and visible light and which in their excited states can induce the degradation of cellulose. This effect is accepted as an important cause of loss of strength or 'tendering' of cellulosic textiles. Wet strength measurements Measurements were made of the wet tensile strength of strips. The ratios of wet to dry strength are shown in Table 3. Table 3 Ratio of wet : dry strength after solar irradiation (1500 BLET) Species Filter Gl P2 P3 Tk BL Unweathered control Pine 0.57 0.36 0.H6 0.51 O.65 0.57 Lime 0.08 0.18 0.30 0.52 O.78 O.65 285 It can be seen that pine and lime exhibit fundamental differences in their tensile behaviour when wetted. For lime the ratio of wet to dry- tensile strength falls progressively as the cut-off for the filters moves to shorter wavelengths, ultimately reaching values of less than 0.1 for strips behind a Gl filter. This clear progressive fall in wet: dry ratio is absent in pine. Insight into this difference was gained from an examination of the fractured ends of strips under the scanning electron microscope. Dry tests of both species, weathered and unweat her ed, demonstrated the same mode of failure : simple fracture across the cell wall (Figure 3a and 3b). There was however a clear difference in the mode of failure of weathered pine and lime strips when tested wet. For weathered pine, Figure 3c, the failure mechanism remained essentially that of the dry tested strips. In wet -tested lime, failure no longer occurred across the cell wall but in the middle lamella. Figure 3d shows how the fibres have simply pulled out in the axial direction as a result of shear failure in the intercellular material. These results suggest that the reason for the pronounced drop in the ratio of wet to dry strength for weathered lime is a change in the mode of failure from one of breakage of the cellulose chains comprising the fibrils to one of slippage. It has been pointed out in reference (3) that there is an increased tendency to tensile failure by slippage as wood becomes degraded. The higher resistance of pine to slippage when tested wet is not yet explained. k. CONCLUSIONS It has been shown that even a short period of exposure to sunlight results in a pronounced loss of surface integrity in wood, and that tensile measurements on weathered strips yield an accurate and relevant assessment of weathering damage. The work confirmed that breakdown of lignin in the middle lamella is an early result of exposure to sunlight, and has shown that profound effects on surface integrity stem from depolymerisation of the cellulose constituent. The use of filters has enabled the effects of different spectral regions to be distinguished and, in line with energy considerations, the most rapid breakdown has been found to occur in wood exposed to light containing ultraviolet wavelengths. When it is taken into account that only about 5 per cent of total solar energy lies in the ultraviolet regions it is clear that this region of the spectrum is most effective in causing photodegradation. Notwithstanding this, wood exposed only to wavelengths longer than U00 nm will degrade at about half the rate observed for material exposed to the full solar spectrum, and this is explained by the large amount of energy present in visible and infrared wavelengths. The finding that visible light is active in wood photodegradation is of great practical significance 286 because it imposes a limitation in principle on the performance of clear and semi-transparent finishes. It must be concluded that the route to effective natural finishes must lie in the use of pre- treatments which stabilise the wood against phot odegradat ion. 5 . ACKNOWLEDGEMENT This paper summarises an extended report of this work which has been accepted for publication by Holz als Roh-u Werkstoff . 6 . REFERENCES 1. Sell, J. , "Grundsatzliche anforderungen an oberflachen- behandlungen fur Holz im Aussenbau (LiteraturJibersicht ) , Holz als Roh-u Werkstoff , Vol. 33, 1975 pp. 336-3^0. 2. Davis, A., Deane, G. H. W. , Gordon, D., Howell, G. V., Ledbury, K. J., "A world-wide program for the continuous monitoring of solar UV radiation using poly (phenylene oxide) film and a consideration of the results", Journal of Applied Polymer Science , Vol. 20, 1976, pp. H65-H7U. 3. Ifju, G., "Tensile strength behaviour as a function of cellulose in wood", Forest Products Journal , Vol. lU, No. 8, 196U , pp. 366-372. 287 INFLUENCE OF CLIMATE UPON HAZARDS FOR WOOD DECAY Rodney C. DeGroot Forest Products Laboratory Forest Service, U.S. Department of Agriculture Madison, Wisconsin 53705 Abstract: In southeastern United States, where both termites and decay fungi are prevalent, some homeowners experience greater repair costs because of decay than from termite damage. As the trend toward integrated pest control technologies impacts upon the construction and wood products industries, accurate predictions of hazard for wood decay in residential construction will become more important, as will the need to predict comparative performance of alternative protection strategies in given environments. Key words: Climate; preservatives; wood decay; wood protection. Wood-frame construction has been the predominant form of residential housing in the United States for the last century and seems likely to continue as a major building method in the future. Several factors have contributed to the durability of wood frame construction. Supplies of naturally durable woods were often available locally; a slow rate of change in prevailing architectural styles contributed to builder familiarity with construction and to regional selection of designs with proven performance records. But the last decade has witnessed significant changes in the design and construction of the American home (Hans, 1976). Now, regionally oriented designs can quickly gain national acceptance and be used extensively in a variety of environments prior to development of local experiences with the new designs. Also, naturally durable woods are often not locally available or are expensive to provide. If these changes contribute to increased wood deterioration problems in houses, it seems probably that their greatest impact will be in southeastern continental USA. There is no single measure for durability. From a performance concept (Eberhard, 1969), an external wall system should bar insects, rodents, and other forms of animal life, as well as restrict the unauthorized entry of human beings; the roof system should protect the building's occupants and contents from weather; and the floor-ceiling system should provide functional space definition and other design functions for the life of the house. 288 In 1956, the Building Research Advisory Board (BRAB) recommended that control measures for wood used in residential construction be balanced with potentials for biological hazard. In the 1980' s, with increasing concerns about product performance, escalating material costs, and enhanced public environmental awareness, this recommendation has added significance. Standards and Decay Potential Building codes and national standards have been the traditional means for protecting the house buyer, who seldom has opportunity to affect the construction process. Trade or user specifications and standards define degrees of protection required for specific products. The Minimum Property Standards (U.S. HUD 1973) currently address regional variations in anticipated wind loads and seismic risk, but do not reference minimum standards to regional variations in potential for wood decay. Some homeowners experience greater repair costs due to decay than due to damage by termites (DeGroot and Dickerhoof, 1975; Cassens, 1978). In aboveground, exterior wood construction, potential for decay is influenced by climate, degree of exposure to prevailing elements, and by other site-specific factors which subject wood to environmental stresses. The concept that much decay can be prevented by keeping wood dry is well known, but some exterior construction must be exposed to the elements . Defining regional zones of potential biological hazards can be exceedingly important for specifying appropriate material requirements and protective measures. For example, water-repellent treatments, without added preservative, protected experimental window units from decay for 20 years near Madison, Wis. (Feist and Mraz, 1978). In a Panamanian jungle, however, water repellants do not contribute to decay resistance (Verrall, 1959). The gradient of hazard between Wisconsin and Panama is not quantified in measures which can be used to determine the relative degree of protection needed at specific locations. Scheffer's Decay Index Several theoretical models of climatic influence on biologic activity may be considered in the definition of regional zones of wood decay potential. Scheffer (1971) developed a formula, based on precipitation and temperature, that estimated potentials of climate to promote decay in of f-the-ground wood structures. 1 Jan [(T " 35)(D " 3)] Climate Index = ^-r 289 T is the mean monthly temperature (°F) ; D, the mean number of days in Dec the month with 0.01 inch or more of precipitation; and I T is the Jan summation of products for the respective months, January through December. The sum of products is arbitrarily divided by 30 to make the index for the United States fall largely within the range of 0-100. Areas with indexes less than 35 were considered the least favorable for decay; 30 to 65, intermediate; and more than 65, the most conducive to decay (Fig. 1) . In the southern continental United States, three sets of research information provide similar results. Scheffer's (1971) contours of wood decay hazard resemble Visher's (1954) contours, which Livingston developed to define climatic regions in terms of moisture-temperature index of plant growth. Meentemeyer' s (1974) contours for predicting percentage of annual leaf production that can be decomposed within 12 months are similar. Indeed, the contour of Meentemeyer ' s predicted potential of 120 percent decay of annual leaf production appears to combine portions of Scheffer's contours for decay hazard indexes of 70 and 80. Most of the Southeastern United States is in Scheffer's proposed high decay hazard area. The peninsula of Florida has a hazard rating of 100 or more; and coastal cities, such as New Orleans, La., and Mobile, Ala., have decay hazards rating above 90. The decay hazard index for interior cities seems particularly influenced by annual rainfall. Dallas, Texas, for example, has a decay hazard rating of 38 while Birmingham, Ala. is rated at 72. Comparison With Other Measures The Scheffer Decay Hazard Index agrees well with decay rates computed from some test units exposed in different geographic areas (Scheffer, 1971), but differs with others (Table 1). Further comparisons of surveys in Mobile County, Ala., and in Raleigh, N.C., with the Scheffer Index suggests that the Scheffer Index reflects relative decay incidence in older homes, but not in newer houses (Table 2). One factor contributing to the difference between relative decay indexes and relative percentages of houses with decay may be the variable interaction of paint films with decay. The paint film in tight joints apparently is more important in protecting wood from decay in areas of low decay hazard than in high hazard zones. For example, paint blocked rain wetting and prevented representative decay in joints of post rail experimental units exposed 15 years at Madison, Wis., but only prevented * 290 Table 1, . --Severi ty of decay as determine d in different studies Ponde irosa pine Location Decay index- Relative rate of decay XI average percent of failure- post rail units after Frequency of decay in houses built Post rail Floori ng 63-67-/ 60-69- 7 4 yr 7 yr pet pet pet pet Southern Wisconsin 39 1.0 1.0 Southern Oregon 47 1.0 1.0 Southern Mississippi 99 1.9 2.5 7 89 Mobile, Ala 99 24 Raleigh, N.C 66 1/ Scheffer, T. C. , 1971. 2/ Scheffer, T. C, A. F. Verrall, and G. Harvey. 1963 3/ Peterson, M. D. and M. P. Levi. 1975. 4/ De Groot, R. C. and H. E. Dickerhoof. 1975. decay development in some experimental units exposed in the high rain- fall areas of Oregon and Mississippi (Scheffer, Verrall, and Harvey; 1971). When films of oil-based paint are not intact, however, they may accentu- ate decay by retarding moisture evaporation from wetted wood. Thus, the most extensive decay in "sash units" exposed near Ottawa, Ontario, was present in the untreated painted controls rather than in the untreated, unpainted units (Sedziak, Shields, and Krzyzewski; 1970). These effects would be most noticeable in houses less than 20 years old In comparing total incidence of decay with decay indexes for three cities—Raleigh, N.C; Baton Rouge, La.; and Mobile, Ala. --there is better correlation between decay index and observed frequencies of decay with Raleigh and Baton Rouge than with any other combination of cities (Table 3). It seems likely, therefore, that the incidence of wood decay, above ground, in residential construction is not linearly distributed with the Scheffer Index at the upper limits of the index. The frequencies 291 Table 2. --Incidence of decay problems in houses surveyed in Mobile County, Ala.y and in Raleigh , H.C?/ Mobile Co. , Ala. - 99-/ House age Raleigh, N.C. -66^ House age Percent with decay 95% confidence 4/ interval- Percent with decay 95% confidence 4/ interval- Yr Yr 0-3 15 3-27 0-5 3 0-6 4-13 24 17-31 6-10 6 2-10 14-23 34 25-43 11-20 19 14-24 24-33 25 11-37 20+ 19 13-25 1/ De Groot and Dickerhoof, 1975 2/ Peterson and Levi, 1975. 3/ Scheffer Index. 4/ Computed from published data. Table 3. -- Comparison of observed incidence of decay with climate index for three cities in the Southeast USA Climate index Raleigh 66 Baton Rouge 78 Mobile 99 Decay-percent of all homes sampled 13 16 24 of wood decay problems in actual construction undoubtedly are influenced by design, rates of drying in individual components, relative efficacy of moderate protective measures such as use of water repellants, as well as by climate. It cannot be determined whether the apparent increase in incidence of decay in the upper range of the Scheffer Index reflects: Underestimation of climatic potential; a substantial nonlinear decrease in effectiveness of brush, dip, or soak treatments with preser- vatives; or an interaction of these and other factors. 292 As the trend for reduced dependency upon broad-spectrum biocides impacts upon construction and wood products industries, accurate predictions of minimum levels of required protection will become more critical. The precision with which potential for wood decay in aboveground construc- tion can be estimated must be improved before discrete zones of decay potential can be defined. Formulation of building regulations based upon regional variation in hazard potential for wood decay would have a significant impact upon housing costs, materials requirements, and building durability. REFERENCES 1. Building Research Advisory Board, Federal Housing Administration, "A Study of Protection Against Decay and Termites in Residential Construction," Report under Contract No. HA-fh-646, Reprints by Building Research Institute, Washington, D.C., 1956. 2. Cassens, D. , "Importance of Wood Deterioration in Single-family Residences for East Baton Rouge, La.," Forest Products Journal , Vol. 28, No. 8, 1978, pp. 19-24. 3. De Groot, R. C, and Dickerhoof, H. E., "Wood Deterioration Problems in Single-Family Houses in Mobile County, Ala.," Forest Products Journal , Vol. 25, No. 3, 1975, pp. 54-58. 4. Eberhard, J. P., "The Performance Concept: A Study of Application to Housing," Volume One, Institute for Applied Technology, National Bureau of Standards, Washington, DC 20234, 1969, 353 p. 5. Feist, W. C, and Mraz, E. A., "Wood Finishing: Water Repellents and Water-Repellent Preservatives," USDA Forest Service Research Note FPL-0124, Forest Products Laboratory, Madison, Wis. 1978, 8 p. 6. Hans, G. E., "The American Home in Another Perspective," Forest Products Journal , Vol. 26, No. 7, 1976, pp. 14-20. 7. Meentemeyer, V., "Climatic Water Budget Approach to Forest Prob- lems, Part II, The Prediction of Regional Differences in Decompo- sition Rate of Organic Debris," Climatology , Vol. 27, 1974, pp. 35-74. 8. Peterson, M. D. and Levi, M. P., "A Survey of Construction Stand- ards and Biodeterioration Problems in Single-Family Homes in Raleigh, N.C.," Proceedings, 1975 Meeting of American Wood- Preservers Association, Vol. 71, 1975, pp. 87-95. 293 9. Scheffer, T. C, Verrall, A. F. , and Harvey, G. , M On-Site Preserva- tive Treatments: Their Effectiveness for Exterior Millwork of Different Species Used in Various Climates," Forest Products Journal , Vol 13, No. 1, 1963, pp. 7-13. 10. Scheffer, T. C, "A Climate Index for Estimating Potential for Decay in Wood Structures Above Ground," Forest Products Journal , Vol. 21, No. 10, 1971, pp. 25-31. 11. Scheffer, T. C. , Verrall, A. F. , and Harvey, G. , "Fifteen-Year Appraisal of Dip Treating for Protecting Exterior Woodwork: Effectiveness on Different Wood Species and in Various Climates," Material und Organismen , Vol. 6, No. 1, 1971, pp. 27-44. 12. Sedziak, H. P., Shields, J. K. , and Krzyzewski, J., "Effectiveness of Brush and Dip Preservative Treatments for Aboveground Exterior Exposure of Wood," Journal of International Biodeterioration Bulletin , Vol. 6, No. 4, 1970, pp. 149-155. 13. U.S. Department of Housing and Urban Development, "HUD Minimum Property Standards as Revised," Superintendent of Documents, U.S. Government Printing Office, Washington, DC, 20402, 1973. 14. Verrall, A. F. , Preservative Moisture-Repellent Treatments for Wooden Packing Boxes," Forest Products Journal , Vol. 9, No. 1, 1959, pp. 3-24. 15. Visher, S. E., "Climatic Atlas of the United States," Harvard University Press, Cambridge, MA, 1954, 403 p. 294 Figure l.--Scheffer climate index map for estimating potential for decay in wood structures above ground. All of the area in 65 and above zone is considered to be a high hazard area (Scheffer, 1971). 295 o to > 2 S THAN TO 65 ^ 5 1 £ & is ^ 3 ^ * * ^* -J tt Wf/, c cd E • a. en o C r- •r— CD 4-> > 03 CU O -o U <+- 4-> o C CU c s~ o 03 •1 — Cl -)-> co 03 £= +J co CD S- s- cu c--a CD c s- =3 (_> C •r— o +J •1— ro CO E O CD s*. JC <- O o go o CD S- 3 304 MATHEMATICAL MODELS FOR THE CORROSION PROTECTIVE PERFORMANCE OF ORGANIC COATINGS James M. Pommersheim Department of Chemical Engineering Bucknell University Lewisburg, PA 17837 Larry W. Masters Center for Building Technology National Bureau of Standards Washington, DC 20234 ABSTRACT Mathematical models were developed for conceptual models describing the principal phenomena that occur in the loss of corrosion protective per- formance of polymeric coatings. The research focused upon the loss of corrosion protection through the formation and growth of blisters under coatings applied to steel. Models were developed for 1) water and oxygen permeability through polymeric coatings, 2) the formation and growth of blisters beneath coatings, and 3) the polarization occurring at electrode surfaces. The conceptual and mathematical models are described in this paper. Key Words: Blister formation; blister growth; conceptual models, corrosion; mathematical models; permeability; protective coatings. 1 . INTRODUCTION One of the most important mechanisms by which corrosion protective coatings for steel fail stems from the formation and growth of blisters beneath coatings. Once blisters form, corrosion of the substrate (i.e., steel) is likely to proceed rapidly because water, oxygen, and an elec- trolyte are present at the metallic interface. The purpose of the research described in this paper has been to develop conceptual and mathematical models to aid in predicting the loss of effectiveness of corrosion protective coatings. 2. CONCEPTUAL MODELS The model for corrosion protective performance of inorganic coatings was conceived in terms of three different conceptual submodels: 305 1. submodel for water and oxygen permeability, 2. submodel for blister formation and growth, and 3. submodel for corrosion phenomena. 2.1 WATER AND OXYGEN PERMEABILITY Figure 1 shows a cross-section of a typical polymeric coating as applied to a steel substrate. The coating consists of a binder (vehicle) and pigment particles. If the coating/steel interface is assumed to be free of water and oxygen, corrosion of the steel cannot occur. While water and oxygen are present in the external environment, the polymeric coating serves as a barrier to the diffusion of these species. If the coating were not present, water and oxygen would have ready access to the steel surface. The presence of the coating helps prevent the initi- ation of corrosion and, in addition, may slow the rate of corrosion once it has begun. Most coatings are relatively permeable to the penetration of water and oxygen, especially to water [1]. To arrive at the steel surface, both water and oxygen must first dissolve (or absorb) into the outside of the coating and then diffuse through the coating towards the metallic corrosion site. Some of the water and oxygen will be absorbed within the coating rather than permeating to the steel surface. In addi- tion, some water and oxygen will be adsorbed by the pigment particles within the coating. For a given type of polymeric coating, metal substrate, and corrosive environment, the overall permeation of the organic coating to oxygen and water will depend on the concentrations of oxygen and water outside the coating, the thickness of the coating, the solubility of the compounds into the film, the diffusivity through the film, the pigment to vehicle ratio, the pigment size and shape, and the temperature. With increasing numbers of pigment particles, it is more difficult for species to dif- fuse. At the same time, more retention of penetrants occurs because of the greater particle density. Thus, raising the pigment to binder ratio will lower the permeation by lowering the diffusivity and increasing the amount of adsorption within the coating. Raising the temperature will raise the diffusivity, lower the solubility, and raise the absorption. For most coatings, the overall permeability of the film increases with increasing temperature [2]. 2.2 BLISTER FORMATION AND GROWTH When water has first penetrated through to the steel surface, it mixes with impurity ions present at the surface and small liquid cells of locally high concentration are formed. The large impurity anions, such as CI and SO^, cannot diffuse out through the coating and the coating acts as a semipermeable membrane. Osmotic pressure builds up and more water diffuses into the cells, gradually enlarging them [3], In regions where the adhesion between the coating and the steel is poor, or where the concentration of impurity is especially high, the coating can disbond from the steel substrate and a micro-blister can form. The blister will 306 continue to enlarge but at a slower and slower rate as the impurity ion is diluted and the osmotic pressure drops. If the enlarging blister exposes additional amounts of surface impurity ions, the rate of blister growth is faster. A similar effect occurs when ionic corrosion products (such as hydroxyl ions) are formed within or adjacent to the blister. After sufficient time the blisters link together. This can lead to complete detachment of the coating from the substrate. 2.3 CORROSION PHENOMENA Corrosion reactions will initiate when oxygen, water, and ions are present locally at a specific site on the steel surfaces, either in a detached film or within a blister. For iron-based metals the most common overall half-cell reactions are: Anodic Reaction ; Fe ■* Fe +++ 2e" (1) (oxidation) Cathodic Reaction : H 2 + - 2 + 2e" + 2 (0H)~ (2) (reduction) 2 The anodic and cathodic reactions occur at different sites on the metal surface. The electrons released in the anodic reactions travel through the metal to the location of the cathode. This completes the circuit. Electron flow is retarded (or equivalently, circuit resistance is increased) by several mechanisms. These include: 1. Build-up of oxide layers or layers of other solid corrosion products, such as Fe(0H)3. 2. Formation of an electrical double layer adjacent to the metal surface. 3. Formation of gases such as hydrogen or oxygen at the cathode surface (or oxygen at the anode surface). 4. Adsorption of hydrogen or oxygen ions resulting in overvoltage because of polarization of the anode or cathode. The anodic and cathodic sites may shift over the surface with time as the coating becomes detached. This can permit substantially uniform corro- sion to occur. The liquid films of corrosion products formed beneath detached coatings may be basic [4]. 3. MATHEMATICAL MODELS Mathematical models were developed and solved for: 1. the permeation of water and oxygen through the coating (absorption, diffusion, and adsorption), 307 2. the growth of blisters on the metal substrate, and 3. polarization of the anode and cathode surfaces. Permeation models were based on the flux laws and conservative laws for the diffusing species. At the outside coating surface, the water and oxygen were assumed to absorb by Henry's Law. Adsorption of water and oxygen on to the pigment surface was assumed to occur at rates propor- tional to their local concentrations and the concentration of pigment particles. A volume balance related the porosity of the coating (vol- ume fraction of vehicle) to the pigment-to-vehicle weight ratio. Dif- fusion was modeled based on Fick's Law. For both oxygen and water the separate rate laws established for adsorption, diffusion , and absorp- tion were incorporated into the overall conservative equations. This formulation resulted in two linear partial differential equations: one for oxygen and one for water. Both equations had the same mathematical form. The dependent variable was the concentation and the independent variables were the depth into the coating and time. Parameters in the model included the pigment to vehicle ratio, the pigment particle size, the diffusivity of the penetrant through the coating (D^), its solu- bility in the coating (C^q), and the adsorption coefficient of the penetrant onto the pigment particles (k^). Since C^ , D^, and k^ have different values for water and oxygen, the rates of absorption, diffu- sion, and adsorption differ. Model solutions for concentration as a function of coating depth and time were expressed in terms of dimensionless variables and parameters. Both short and long time solutions to the mathematical model were presented, as were steady-state profiles and concentrations at the metal-coating surface. The model predicted that the permeation of water and oxygen would be inhibited and, thus, the potential for corrosion lessened, with a suitable combination of: 1. lots of small-sized pigment particles, 2. retarded diffusion through the coating, 3. rapid adsorption onto pigment particles, 4. thick (or multiple) coating(s), 5. low solubility of penetrant into vehicle (water and oxygen repellancy) , 6. short dimensionless contact times. Of all these factors the model was found to be most sensitive to the coating thickness L. Increases in L gave much lower concentrations of water and oxygen at the metal surface as well as lower dimensionless contact times. Thus, the protective function of the coating is greatly enhanced when membrane thickness is increased or multiple coats are used. 308 Two criteria were developed which expressed these factors quantitatively. First, it was shown that, so long as the contact time was less than a certain critical value given by eL /D-^ (where e is the fraction of vehicle), penetration to the surface would be insufficient to initiate blister growth or corrosion. Second, it was established that even if a steady-state was attained, either water or oxygen might not be present at the steel surface in sufficient concentrations to initiate corrosion. Mathematically this was expressed by C^q sech a± < C-£ c , where a is a dimensionless parameter which is a measure of the retention of the pene- trant by the coating, and C^ c is the (critical) value of the concentra- tion of the penetrant which is needed to inititate corrosion, as determined, for example, from a Pourbaix diagram [5]. Blister growth was considered, in this study, to occur either over concentrated spots of impurity located at the point of blister initia- tion (as in Reference 6) or over a surface containing a uniform concen- tration of impurity. The solutions to both models had similar mathe- matical forms: the blister radius, surface area, volume and osmotic pressure were all found to vary directly as fractional powers of the time. Blister volume varied as t®*'* for concentrated spots of impurity, and as t 1 *^ for a uniform distribution of impurity. The relative value of the osmotic pressure, although initially higher, was predicted to fall more rapidly for the concentrated distribution of impurity, due to the rapid dilution of the blister with water. For a uniform distribution of impurity the osmotic pressure does not fall as fast because fresh salt is continually being added at the periphery of the blister. The coupling between blister formation and the permeation of water was discussed in terms of the time and distance scales characteristic of each process. Two mathematical models were developed for the corrosion phenomena occurring in organic coatings. These were the potential difference model and the equivalent circuit model. The models treated the polariza- tion occurring when an electrical potential is dissipated in an electro- chemical cell. In the polarization model, the driving force was the potential difference between the anode and the cathode. Model solutions predicted that the potential difference decreases exponentially with time so that both anode and cathode potentials slowly approach a common final potential. In our second model, corrosion was modeled in terms of an equivalent electrical circuit [7]. Polarization rates were taken to be functions of the electromotive force and the circuit resistance. The potential between the anodic and cathodic sites within the corrosion cell was considered to be dissipated across several series resistances. These included the resistance of the electrolyte and the resistance of the corrosion products. Corrosion product resistance was attributable to the formation of product layers next to the cathode surface. The resistance of the electrolyte was taken as a constant, while the resistance of the corrosion products (R) was modeled using several different submodels. This included a constant value of R, one based on Langmuir adsorption (or film formation) of corrosion products, and one based on uniform accumulation of layers of corrosion products. 309 All of the models predicted that the potential between electrodes, the charge and the current decrease with time. The equivalent circuit sub- model with constant R was shown to be analogous to the potential differ- ence model. For the film formation model the potential and charge decreased together but the current fell faster because of the increase of R with time. 4. CONCLUSIONS The development of conceptual and mathematical models is a significant step forward in predicting the loss of effectiveness of corrosion pro- tective coatings. At the same time, much additional research is needed to refine details in the conceptual models and to validate the models experimentally. Also research is needed to couple the individual models which have been developed to date. The coupling is complex and only the separate processes were modeled in this paper. Future modeling and experimental efforts will be directed at elucidating the nature of the coupling [8] . 5. REFERENCES [1] Perera, D. Y. and Van den Eynde, D., "Alteration in Properties of Organic Coatings for Buildings Due to the Leaching Process," Dura- bility of Building Materials and Components, ASTM STP 691 , 1980, pp. 698-710. [2] Kumms, C. A., "Physical Chemical Methods for Organic Protective Coatings," Journal of Coatings Technology , Vol. 52 (664), 1980, pp. 40-53. [3] van der Meer-Lerk, L. A., and Heertyes, P. M., "Blistering of Varnish Films and Substrates Induced by Salts," J. Oil Col. Chem. Assoc , Vol. 58, 1975, pp. 79-84. [4] Ritter, J. J. and Kruger, J., "Studies on the Subcoating Environment of Coated Iron Using Qualitative Ellipsometric and Electrichemical Techniques," Surface Science, Vol. 96, 1980, pp. [5] Uhlig, H. H., Corrosion and Corrosion Control , 2nd edition, John Wiley & Sons, Inc., New York, 1971. [6] van der Meer-Lerk, L. A. and Heertjes, P. M., "Mathematical Model of Growth of Blisters in Varnish Films on Different Substrates," J. Oil Col. Chem. Assoc , Vol. 62, 1979, pp. 256-263. [7] Thirsk, H. R. and Harrison, J. A., A Guide to the Study of Electrode Kinetics, Academic Press, London, 1972. 310 [8] Thomas, D., "Initial Degradation of Corrosion Protection by Organic Coatings," Second International Conference on Durability of Building Materials and Components, National Bureau of Standards, Gaithersburg, MD, September 14-16, 1981. 311 o \ E o ":* c CD C k» CD CD b4 >» X o o *-- flL a> *-• o o N O HI ° o ^ o o o \ ■ss — < o •S =3 CO CD CD CD *-» CO •O CO 5 CD O O O CD CD en c: 03 O i- cu E >> o Q_ U •r— ra cr> i- o <+- o +-> o CD I oo co O S- o o ra c cu en cu %. Q. O) o; o 03 E 00 XI «v* 13 — U c z , . •r» +-> ra s- ^1 •r— r— n3 O S- fl3 CU sr •1— 1 — c ra w — C3 E3 s ra IS Eq l£l tn m r^ U3 Sea •• •• Q | st> >. _i <3 « « £ s 2 « H QQ 0.- Sis to (O ft. X to u. < Oi 03 Ul a. u. O o a. ct at q > 1 ai ca 10 II) CO o IT) C CO o cm r». r-l y, o o 6 Z CJ r-l O WOU i £3 eo co O O to* co 55 V oi o m o o o o r"j o CT> as l-l r— ! c w o o U C o o r« r-l «—) r-( a; < c o o o U LL o c~» o r-l U a. < m o o as CO r— 1 r^ LL c ■ o ^"* T3 ' c o 60 i U •~1 i .^ 5-i rt CD o J-J 'bO 10 >» •H x: CL. r-t Pn <3 CO ^ O ■& * 8 CO g »' be •*4 w *** ' ^» in* O hi S n B • w • • E OQ Pn o • Pn • w I S 1-1 !B *7** m | i i CO i 0) o J3 •» r-i .-4 i— 1 V r-i r-i r-i 3 Q- fe pa fe fa fe V 1 / c n O N o eio o '"5 >> 3 4J u w cSf cci Q oi *: .. ou £ o 4J l- ca *-* E o CQ 'c 3 >> • • i— i) B • C3 3 +j """ f- s- o o LO k. V) a. rt 6 aj «J CQ s_ u O v. C £ iy0 CO 4J X3 ZJ to o Q. X T3 rt 3 3 O E • » OJ ^•-. o 5: . c •> O r3 < , c > H- TC! rt • • +J t> e • M to to < c o a. o S rt bO i E O Vfe. > c O (.O o ,4 4- CU >> ■♦-• — o D, a CO •o o s- O C 4> •a LU C O X 3 «J .. O ft, D K- ^ >> o E CD 4-> m E > 00 o •r— 4-> 03 E s- o CO c o •I— +■> 03 O Q- CD +-> 03 O o 346 PREDICTION OF THE SERVICE LIFE OF SYNTHETIC POLYMERS W. Lincoln Hawkins Bell Laboratories, Retired Abstract: With the entry of synthetic polymers into the building industry, new questions about durability have arisen. In many applica- tions, polymeric construction materials cannot be protected by surface coatings. The use of polymers in exterior applications such as siding and glazing, and in interior applications such as plumbing, wall tile, and electrical insulation requires acceptable durability without surface protection. Synthetic polymers degrade by a complex sequence of chemical reactions when exposed to air and moisture, particularly under the influence of heat or sunlight. The mechanism by which many synthetic polymers degrade has been established, and stabilizers have been developed which are capable of extending the useful life of these materials. As examples, polyethylene, poly( vinyl chloride), styrene, and other polymers used in construction now have acceptable life expectance as a result of stabilizers added to the polymer. Design of these stabilizers is based on the understanding of how each polymer degrades. Methods suitable for estimating the time to failure of synthetic polymers are necessary for successful application of these materials in construction. Principles of tests, used to estimate the durability of polymers, could be applied to testing of other building materials. Keywords: Addition polymers; antioxidants, condensation polymers; degradation; environment; hydrolysis; oxidation mechanisms; polyethylene; polymers; poly(methyl methacrylate); polystyrene; poly(tetrafluoroethylene); poly( vinyl chloride); rubbers; stabilizers; ultraviolet absorbers; ultraviolet radiation. 347 Synthetic polymers have become an important class of materials in modern construction. Because of the wide variation in structure of polymers, these materials provide a wide range of properties which can meet specifications for a broad range of applications. For example, exterior applications in construction include siding, window frames, glazing, shutters, leaders, and various decorative items. In building interiors, synthetic polymers are used as wall tile, plumbing items, electrical insulation and fittings, room dividers, and a variety of decorative items. Although polymeric materials are not generally used in load- bearing applications, their anticipated service life before replacement or refinishing is an important consideration in construction design. The service life of materials is directly related to the environment to which they are exposed, and synthetic polymers are no exception to this general rule. However, environmental factors responsible for the degradation of polymers are much more diverse and complex than those affecting other materials. They may fail as a result of slow, progres- sive chemical change or from mechanical stress as might occur on impact. Synthetic polymers are organic in composition and therefore undergo complex chemical reactions, typical of organic compounds. Dependent on their structure, they may react readily with oxygen, ozone and water. Atmospheric contaminants may function as catalysts to increase rates of hydrolysis or oxidation. Considerable effort has been directed toward understanding the reactions responsible for polymer degradation. As a result of this extensive research, stabilizers have been developed which can extend significantly the service life of many polymers. In construction, durability of polymeric materials will vary considerably between exterior and interior applications. When used out-of-doors, certain polymers will degrade rapidly as the result of exposure to ultraviolet radiation. Although the ozone layer in the upper atmosphere screens out the ultraviolet radiation below 290 nanometers, the region of the spectrum between 300 and 380 nanometers rapidly degrades many polymers. Unprotected polyethylene, for example, embrittles on outdoor exposure in temperate regions in less than two years. Poly(methyl methacrylate) , sold under the trade names of plexiglas and lucite, on the other hand, is so resistant to ultraviolet radiation that it is used as a glazing material. Whenever a synthetic polymer is to be used out- of-doors, ultraviolet-induced degradation must be taken into account and if necessary, adequate protection must be provided. The infrared component of the solar spectrum increases the temperature of exposed materials and promotes thermal degradation. All polymers are vulnerable to this form of degradation although there are important differences in the rate at which different polymers undergo thermal degradation. Usually outdoor weathering involves both thermal and ultraviolet-induced reactions. In both types of degradation, oxidation is the primary chemical reaction. Though polymers vary in the rate at which they react with oxygen, eventually oxygen attacks and destroys 348 all known polymers. Oxidation may cause scission of polymer molecule thereby reducing their molecular weight of it may promote crosslinking between adjacent molecules resulting in an increase in molecular weight. Certain polymers degrade through sequential elimination of small, molecular fragments along the backbone chain. Combinations of these three, basic mechanisms are also known to occur. Evidence of oxidative degradation may be in the form of mechanical or dielectric failure, or simply in aesthetic quality. Degradation resulting from ultraviolet radiation may be inhibited by compounding into the polymer a few percent of a dark pigment which will function as a screen, limiting the amount of radiation reaching the sample bulk. When clear or light-colored properties are required, addition of a percent or less of an ultraviolet absorber can be quite effective." Ultraviolet absorbers protect by preferential absorption of ultraviolet energy followed by release of the absorbed energy by a mechanism which does not degrade the polymer. In general, light screens are more effective than ultraviolet absorbers. Unprotected polyethylene fails out-of-doors in less than two years, but when two or three percent of a fine carbon black are well-dispersed throughout the material, this polymer can be expected to have a service life of at least fifty years . By comparison, the best ultraviolet absorbers available today will only extend the service life of this polymer to about fifteen years. Since poly(vinyl chloride) (PVC) is so widely used in construction, it is important to understand how this polymer degrades. Under the influence of either heat or ultraviolet radiation, PVC undergoes a sequential elimination of hydrogen chloride from adjacent repeating units along the polymer chain. When sequences of nine or more such events have occurred, the product is an intense chromophore. As degra- dation proceeds, the color of PVC goes from yellow to red and finally to black. Intense discoloration is apparent after loss of only a small fraction of the polymer's total content of hydrogen chloride. Stabi- lizers have been developed, however, which extend the service life of this polymer sufficiently to allow its use in many outdoor applications. Protection against thermal oxidation requires addition of antioxidants. These additives are designed to interupt the reaction sequence which leads to oxidative failure. Several types of antioxidants are available, each designed to inhibit thermal oxidation at a specific step in the reaction sequence. Using polyethylene once more as the example, this polymer fails by embrittlement in just over a year when kept in the dark at 40°C, but only a tenth of a percent of a good antioxidant would extend the service life close to forty years at this same temperature. When a polymer is to be subjected to simultaneous conditions of thermal and photodegradation, stabilizers should be added to inhibit both forms of degradation. When light stabilizers and antioxidants are incor- porated into a polymer, however, an antagonistic effect may result. Correct choice of additives can prevent such effects, and in certain instances a synergystic effect may result from such mixtures. 349, H H H H H H H H n C=C C-C -C-C- c-c H H H H H H H H Degradation of synthetic polymers by reaction with water is the next most common factor limiting the service life. Hydrolytic degradation is catalyzed by traces of acid or base in the aqueous medium. However, reaction with water occurs only in specific types of polymers. Similarly, ozone-induced degradation is limited to a very narrow class of polymers. Many factors must be considered in estimating the service life of a polymer. The effect of exposure environment has been discussed. Basic structure of the polymer is also important. The mechanism and rate of degradation is directly related to the chemical structure. Polymers are synthesized by two general reactions, addition and conden- sation polymerization. In addition polymerization, polymer molecules are formed by the rapid coupling of monomers - the small molecules which are the starting materials for the process. Monomers are coupled to each other through coordinate carbon-carbon bonds, as shown schematically in Reaction 1 . (1) The section of the polymer molecule enclosed in brackets is referred to as the "repeating unit," and in the complete polymer chain there would be "n" of these units. The important point is, however, that only coordinate carbon-carbon bonds form the backbone chain. These bonds are not hydrolyzable and so addition-type polymers, all of which contain this type of bond, do not degrade in the presence of water or even dilute acids. Only strong acids such as nitric acid attack polymers of this type, and even then it is not the bonds along the backbone chain which react. Excellent service life can be predicted for polymers synthesized by addition polymerization if attack by moisture is the only significant type of degradation anticipated. Polyethylene, polypropyl- ene, polystyrene, poly(vinyl chloride), poly(methyl methacrylate) , the ABS* molding resins, and most elastomers are representative of addition polymers . In contrast, polymers which are synthesized by condensation polymeriza- tion are susceptible to degradation by water or high humidity. In these polymers, as shown in Reaction 2, each time monomer units join to form the backbone chain, a byproduct is formed. +H 2 H OH OH HN-(CH 2 ) n -C + HN-(CH 2 ) n -C + — N-(CH 2 ) n -C -N-(CH 2 ) n -C — (2) OH OH H +H 2 * Polybutadiene impact-modified copolymer of acrylonitrile and styrene 350' In this example of the synthesis of a polyamide, water is the byproduct which is split off. Unless this reaction product is removed from the reaction mixture, a reverse reaction will occur in which the amide (-N-C ) groups are hydrolyzed thus breaking the polymer chain. Even H after condensation polymers are removed from the synthesis environment, they are still vulnerable to depolymerization when in contact with moisture. This reaction accounts for the failure in contact with dilute acids of typical condensation polymers such as the nylons, the polyesters and the naturally-occurring polymer, cellulose. It must be recognized, however, that there are wide variations in the resistance to degradation of condensation polymers by water. This variability is related to the chemical groupings on either side of the vulnerable bonds which hold the chain together. For example, nylons differ in their resistance to hydrolysis, dependant on the length of the carbon groups between amide linkages. The larger the number of carbon atoms, the more stable will be the nylon. The basic conclusion to be drawn from the preceding discussion is that the service life of a polymer, synthesized by condensation polymerization is likely to be restricted by its sensitivity to moisture attack. Degradation will proceed much more rapidly in contact with dilute acids, and therefore such polymers are not a good choice in areas where acid rain is a problem. Under these conditions, an addition polymer would be preferred - provided all other specified properties are acceptable. The sequence of reactions leading to degradation has been established for several of the important polymers used in construction. The mechanism of degradation for PVC has already been discussed. Natural rubber and synthetic elastomers are used in construction as gaskets and fillers. Degradation of natural rubber has been studied extensively, and the mechanism for its degradation can be used to understand the chemistry of failure in other hydrocarbon polymers including polystyrene (shutters) and the polyolefins (drainage tile and pipe). Thermal oxidation of polyolefins is known to occur by a free-radical initiated, chain reaction. There are, however, important differences between polyolefins in their resistance to thermal oxidation. For example, the service life of polypropylene is much less than that of polyethylene. This difference is readily accounted for by the established mechanism of degradation. Thermal oxidation of a polyolefin is initiated at a site along the polymer chain from which a hydrogen atom is split off. Though loss of hydrogen is a random process, the strength of bonds between hydrogen atoms and the carbon atoms along the chain is not the same in poly- propylene as in polyethylene. This is indicated in the following representation of the repeating units in these two polymers, 351 H H ■C-C- H H n H H -C-C— H CH 3 n Polyethylene Polypropylene All hydrogen atoms along the backbone chain of polyethylene are equivalent, but in polypropylene one carbon atom in each repeating unit has only a single hydrogen atom attached to it. Strength of the bond between this hydrogen and its associated carbon atom is significantly less than that at carbon atoms to which two hydrogens are attached. Thus initiation of thermal oxidation occurs more readily in polypropylene than in polyethylene. This is reflected in the poorer service life of polypropylene. Adequate stabilization, however, can improve the service life of each of these polymers so that the difference in stability can be reduced. Since it is known that degradation of hydrocarbon polymers is initiated by splitting off of a labile hydrogen from the polymer chain, it should be anticipated that the stronger fluorine-carbon bonds in fluorinated polymers would correspond to an improved service life. This is indeed the case as evidenced by the excellent stability of fluorinated polymers such as polytetrafluoroethylene or teflon. Teflon is very difficult to process, however, and expensive as well so that despite its excellent service life it has no significant role in construction. Newer polymers in which fluorine replaces a limited number of hydrogen atoms in a typical hydrocarbon polymer are of interest as insulating compounds. They are expected to give good service life in addition to their designed resistance to burning. Natural rubber when held under stress is rapidly degraded by ozone. The mechanism for this degradation shows that it is the unsaturated bond between carbon atoms in each repeating unit which is the point of attack. Although antiozonants have been developed which inhibit the attack of ozone on natural rubber, an elastomer which does not have this structural feature and hence would not degrade by the ozone mechanism might be preferred. Copolymers of ethylene and propylene are not as efficient elastomers but these materials are not susceptible to ozone degradation. Knowledge of the mechanism by which polymers degrade can be used as a first approximation of the anticipated service life. This information is very useful in the selection of the best polymeric material for a specific application. When a choice of materials is available, polymers with poor stability as indicated by their structure can be avoided. Service life depends on the mechanism of degradation and this in turn on polymer structure. Such selections should, however, be followed by accelerated testing of the material to more precisely predict the service life. 352 Suggested Supplemental Literature "Stability and Structure Relationships," F.H. Winslow and W.L. Hawkins in Crystalline Olefin Polymers , Raff and Doak, Eds., Interscience, New York, 1965. Atmospheric Oxidation and Antioxidants , G. Scott, Elseevier, Amsterdam, 1965. Applied Polymer Symposium No. 4, M.R. Kamal, Ed., Interscience, New York, 1967. Polymer Stabilization, W.L. Hawkins, Ed., Interscience, New York, 1971. 353? DEGRADATION PROCESSES OF BUILDING MATERIALS AND COMPONENTS A SHORT REVIEW AND SOME PROPOSALS FOR RESEARCH Ph. EURIN Centre Scientifique et Technique du Batiment 24, rue Joseph Fourier 38400 Saint Martin d'Heres - France Abstract - The object of this paper is to present a review of main degradation processes of building materials and components and to point out some problems which should need to be resolved or studied more extensively that it has been made up to now. The problems relating to mineral and metal materials will only be refer- red to briefly underline the main directions that would appear to be of interest for current research. The degradation processes for polymer materials will be presented in more detail . A brief description of the elementary phenomena - thermooxydation, photooxydation, hydrolysis - which are responsible of the weathering of polymeric materials is given through some examples. Key words - building material, degradation process, durability, polyme- ric material, weathering 1. INTRODUCTION If there exists an industrial sector which is a large consumer of mate- rials and for which the in-depth study of durability of materials and structures is vital, it is the building trade. The design and production of buildings with durability appropriate for their purpose is the objective foremost in the minds of designers, architects, engineers and companies. To achieve this aim, a better appre- ciation of the durability of the different components and materials used in construction is necessary. An understanding of the processes of de- terioration of materials is thus necessary, and this paper aims to shortly introduce the different contributions which will be presented in this field. In view of the diversity of materials used, it is impossible to give an exhaustive review of deterioration processes in the main categories of materials used in building : minerals, metals and polymers. 354 The problems relating to mineral and metal materials will only be refer- red to briefly underline the main directions that would appear to be of interest for current research. The deterioration processes for polymer materials will be presented in more detail with the aim of introducing those papers which will be devoted to this theme during the current session. 2. MINERAL AND METAL MATERIALS 2.1. Mineral materials such as stone, concrete, earth, terra cotta are probably the most widely used materials in building. Some of these have been used since ancient times. Numerous studies have dealt with these materials and the criteria deter- mining their durability are now fairly well known. Mineral materials can be affected by different aggressive agents amongst which water is one of the foremost. The effect of water can be induced directly (i.e. by direct contact) by rain or immersion of low parts of walls, but also indirectly by migration of water (either liquid or gas) through the porous structure of the minerals. Condensation inside the material may occur in some thermal conditions. Water extracts soluble components, produces variations in dimensions and hence internal stres- ses. The structure of the material may even be destroyed after freezing cycles. Other chemical agents such as chlorides, sulphates, nitrates, acids, bases, pollution, CO2 may have specific effects on the material itself, or on other incorporated materials (for example corrosion of reinforcing bars in concrete). Mechanical strains are poorly tolerated by mineral materials because of their brittle mechanical behaviour. Seasonal stresses are one of the major sources of deterioration of walls or building structures. Although much work has been carried out on these phenomena, it would appear that a better understanding is still needed in certain areas, e.g. water transfer through porous media (coupling of heat and mass transfer), rheological behaviour of concrete, mechanisms of cracking (initiation of cracks and their propagation) in materials themselves and in composite structures (masonry walls), efficiency of protection against the effects of water, especially for light-weight or cellular concrete. As far as the prevention of corrosion of steel in concrete is concerned, some progress has been made in the evaluation of coating efficiency, but it would appear that there is still the need for methods of in situ measurement of the permeability of concrete. 355 2.2. Metals are widely used in building and in this sector corro- sion produces great economic losses. Corrosion mechanisms are well known in simple, characteristic cases, but the explanation of real corrosion effects is somewhat more difficult because of the great num- ber of parameters involved. Simple methods of diagnosis would be very useful ; this could help in the choice of curing techniques to lower the corrosion rate. Significant testing of the efficiency of protection against corrosion is still needed in this field. The methods should aim to evaluate protective coatings against atmospheric corrosion and the efficiency of inhibitors used for water distribution. The development of heating techniques using solar or geothermal energy raises new corrosion problems : corrosion of solar collectors and of heat exchangers due to the evolution of heat transfer fluids (deterio- ration of antifreezing additives) or to the special conditions existing in these systems (high temperatures, pipelines made with different metals, corrosivity of water from geothermal sources). The use of soil for longterm storage of solar energy is being studied. Corrosion of water tanks, exchangers or pipelines embedded in the soil should be studied with the aim of determining appropriate protection (coating, cathodic protection ...). 3. POLYMER MATERIALS 3.1. Polymer materials constitute a wery diverse category of mate- rials, both by their chemical nature and their physical properties. In the building industry, opaque or transparent rigid materials are used for sidings or glass-work, flexible materials are used for gaskets or waterproofing, foamed polymers are used for insulation and liquid resins for paintwork. The deterioration processes of these materials are very complex, as they result from a series of elementary phenomena themselves dependent on numerous factors : composition (nature of the polymer and of additi- ves), impurities, structural defects produced during manufacturing of the polymer or its transformation. The evolution of a given industrial- ly produced polymer is linked to many factors peculiar to that material and to the thermo-mechanical treatment it has undergone. In the extreme, each example should be considered individually as a special case. However, with an understanding of the principal modes of deterioration, methods can be determined for accelerated deterioration which are signi- ficant of a material's performance, and so manufacturers can make appropriate modifications for better performance over time. The main deteriorating agents to be taken into account for the use in buildings are : oxygen, which when associated with temperature or radia- tion leads to thermo-oxidation and photo-oxidation reactions ; water, which by hydrolysis induces chemical destruction of the molecule, and 356 by pi astifi cation and depl as tifi cation causes modifications in the me- chanical behaviour which can induce stress cracking ; pollutant gases or oxidizing agents (especially ozone), which have a specific effect of accelerating oxide reactions. 3.2. Thermal aging can introduce purely physical phenomena (migra- tion of gases, plasticizers, evolution of morphology), or chemical phenomena, the most important of which is oxidation. In a given atmosphere, the kinetics of thermal aging of a material is a function of the temperature which, within a limited temperature range, follows Arrhenius type laws. So in practice, it can often be observed that the lifetime. At, defined according to a correctly chosen criterion for the end of life, varies as a fonction of temperature T, according to a law (1) of the form : D Log At = A + y, where A and B are constant. Experiments at temperatures T > Tu (Tu = temperature of use) give an approximation for the extent of the lifetime At at Tu. But caution is required in such an approach, as it assumes that there is no transition (or morphological change) for the temperature range under consideration (Tu, T). The chemical processes of heat deterioration of polymers in the presence of oxygen in the air include several . stages (2) : . an initiation phase polymer R-R -* 2R* or RH -> R* + H* (the RH links are often the most sensitive to thermolysis) . a propagation phase R* + 2 + R0 2 * R0 2 * + RH + R0 2 H + R* . a terminating phase consisting of the combining of two radicals to form an inactive agent. By creating new radicals, the decomposition of the hydroperoxides ROoH ■* RO + OH* induces the acceleration of the oxidation process : RH + OH* -* R* + H 2 RH + RO* •* R* + ROH Figure 1 shows the acceleration produced after an induction period corresponding to the accumulation time of the hydroperoxides before they inter- react. 357 3.3. Photochemical aging (3) also involves radical type elementa- ry reactions. However, the onset of deterioration is no longer produced by thermolysis, but by absorption of a photon by a molecule which is consequently brought to an excited level. The excess energy thus obtai- ned either leads to a dissociation of the chemical bonding, or to a pho- tophysical reaction (fluorescence or de-activation by non radiative transition). Polymers \/ery often contain chromophoric sites (impurities, unsaturated groups, hydroperoxide groups) at which photochemical reactions are initiated. For example, photons react on carbonyl groups : O o \J^\ bt \J { - + V\ Norrish type * o o v"v\ iiv'U v\ Norr1sh type 2 The hydroperoxides formed easily break down under the effect of the photons : R-OOH ^ RO* + OH* Next, the radicals formed react with the RH substrate according to the propagation and terminating phases described above. 3.4. Factors in thermal and photochemical aging The principal factors in the environment which determine the extent of the photochemical deterioration in polymers are : the spectrum and inten- sity of the light, the ambient temperature, and the possible presence of oxidizing agents in the atmosphere. Thermal an photochemical aging are significantly accelerated by the presence of ozone, NCL or SC>2 (2). Ozone has a particularly severe specific effect on elastomers whose chains contain double bonds. As far as the material is concerned, photo-degradation depends on the nature of the polymer, on the concentration of defects or chromophoric impurities, and thus on the quantity of light energy absorbeb by the material . This deterioration penetrates the material more or less deeply depending on whether it is opaque or transparent, and more or less permeable to oxygen (fig. 2). 358 The network modification resulting from photochemical or thermal aging are the chains scission, reticulation, or creation of conjugated double bonds. The first two lead to a modification of the mechanical properties: generally embrittlement (fig. 3), whereas the conjugated bonds cause colouring of the resin. 3.5. Thermal and photochemical stabilization A knowledge of the deterioration processes enables efficient action to be taken at different points along the chain of elementary deterioration. The fig. 4 established by J. LEMAIRE (4) shows the principles of the battle against weathering. Two rules are proposed by the same author : . intervene by acting on the most primary phenomenon possible, . pay attention not to intervene by means of an additive which may pro- mote another form of weathering : anti -oxidizing agents must not be chromophoric agents. 3.6. Effect of water on polymers The pratical applications where polymers are not in contact with water are rare, if only because the atmosphere contains a significant propor- tion of water. Whether in gaseous or liquid state, water may involve the following processes : plastification, solubilization of adjuvants, erosion, hydro- lysis. Humid-dry cycles accelerate deterioration by combining the mechanical effects due to dimensional variations and in some cases by extracting soluble adjuvants. Hydrolysis is a purely chemical activity affecting certain polymers, the most sensitive being polyamides and polyesters and, to a lesser extent polyurethane. The combined effect of water and mechanical stress induces stress cra- cking. The accelerating action of water on the weathering of plastic materials has been revealed particularly for the case of GRP compounds (5). Deterioration of the fibre-resin interface by hydrolysis leads to modification of appearance (especially for transparent materials) , and to a loss of mechanical properties. 359 3.7. Lines for research Numerous studies are currently devoted to deterioration processes in polymers. Continuation of these projects, in particular the recognition of elementary processes and determination of the concentration of tran- sitional products (e.g. hydroperoxides) would enable deterioration dia- grams to be established (fig. 5), and this would be of appreciable value for the understanting of the phenomena involved. In this way, the empi- rical procedures adopted over the past few years could progressively be replaced by a more scientific approach, with the following conse- quences : . better definition of accelerated tests significant of longterm beha- viour, . better approach to stabilization problems. With this in mind, research on the stabilization processes is obviously needed, as well as work on the incompatibilities which are revealed (for example between fi reproofing additives and protection against photo-oxidation) . The importance of the initial structure as well as the transformation conditions has already been underlined for the weathering of polymer materials. Knowledge of the relationship between the use, the morpholo- gical structure and the long-term behaviour is still not far enough developed. Research in this direction has an obvious pratical value (6). As far as applied research is concerned, the directions for research can be defined according to the applications demanding a better under- standing of durability. Without claiming to be exhaustive, the follo- wing areas may be mentioned : . polymers used for the capture of solar energy : transparent covers for solar collectors, organic glazing, polymeric absorbers, . energy storage : tanks, pipelines for, . heat insulation : long term behaviour of insulation materials, . surface coverings : paintwork, plastic coverings, . assembly by adhesion. This last area of pratical application, which is being widely developed in other industrial sectors, will probably be the object of growing interest in the building sector. So it would seem important to develop studies of the factors determining the durability of bondings between adhesives and components. More generally, a better knowledge of the phenomena arising at the interface between two materials should allow durability to be studied, not only at the material level, but also at 360 the level of multi layered components or composite structures which constitute the great majority of the problems arising in the building sector. 4. BIBLIOGRAPHY 1. Weatherability of plastic materials, Applied Polymer Symposia n° 4, Editor : Musa R. Kamal , Interscience Publishers 2. H.H.G. JELLINEK, Aspects of degradation and stabilisation of polymers, Elsevier Publishing Company 3. B. RANBY and J.F. RABEK, Photodegradation, Photooxydation and photo- stabilisation of Polymers, Wiley Interscience 4. J. LEMAIRE et al . , Vieillissement des polymeres : empirisme ou science, Caoutchoucs et Plastiques, Aout/Sept. 1979 5. A. BLAGA and R.S. YAMASAKI, Journal of Materials Science 8, 1973, pp. 654-666 6. J. VERDU, Effets induits par la transformation sur le vieillissement des polymeres, Colloque GFP Lyon, Nov. 1980 361 TIME Oxydation of polymers in air y : oxygen absorbed carbonyl groups mesured by IR FIG. 1 \362 1 A ♦ e/i .. o «■ -e/i 1 INCI DENT LIGHT X \y w c*) (b) -e/i Cc) y (a) (b) (c) degree of degradation distance to the center of the sample material transparent to UV wavelengths and permeable to 2 " " " " and not " " " " very little transparent to UV wavelengths Repartition of photochemical degradation versus depth. One face of material is exposed to UV rays. FIG. 2 363 ^ IN PACT STRENGTH 500 -■ TIME Variation of impact strength versus aging time for PVC FIG. 3 364 ~J J < < O H H 21 X Id LU -J r X < O 2: O a: r\ H r» H /N x ^ O * 9 ^ X -*= < 1- v-» a. w a: v (M O < o UJ -J o X < I- z UJ X < z o li. >• z o H < >- X o < o H < (D H U. 365 DETERIORATION PROCESS OF POLYMER MATERIALS AND ITS CORRELATION WITH DEPTHS FROM SURFACE T. Fukushima Building Research Institute Ministry of Construction Japanese Government 1-Tatehara, Oho-machi, Tsukuba-gun, Ibaraki-ken, 30£, Japan Abstract: The deterioration process of polymer materials by the simultaneous action of UV-light and diffusive oxygen is analysed theoretically based on the unsteady state dynamics, taking account of the progress of deterioration into polymer materials. As a result, it could be shown for the first time that the depth of deterioration layer of polymer materials increases in proportion to the square root of exposure time. This /%" law is often observed in the cases of neutralization of concretes or oxidation of metals, but the law is expected to be also the case with deterioration of polymer materials from this theoretical study. The theoretical results are compared with experimental data. Consideration into the reduction in flexural strength of polymer materials with time is done. The main points of supposition of the model are as follows: 1) Only polymer molecules activated by absorbing UV-light near the absorp- tion band characteristic of polymer materials can take part in the deterioration reactions, which are depolymerization of activated polymer molecules and photooxidation by the collision of diffusive oxygens with activated polymer molecules. 2) Deteriorated polymer molecules diffuse neither innerward nor outerward and remain in the sites of the original positions. 3) Deterioration proceeds innerward from surface and the degree of deterioration varies continu- ously with times and depths. 4) Degree of deterioration is reflected as the change of IR absorption spectra of surface layers. 5) The deterioration layers don't contribute to the flexural strength of polymer materials. Key words: Activated polymer molecules; depolymerization; deterioration layer; diffusive oxygen; flexural strength reduction; infrared absorption spectra; parabolic law; poly- mer materials; simultaneous action; ultraviolet light acti- vation; 367 Many polymer materials have been widely used recently in housing as building materials such as adhesives, waterproof- ings, coatings, finishing materials and so forth, but they are known to be subject to deterioration under natural weathering, and deteriorate rapidly under accelerated aging tests by the simultaneous action of UV-light and diffusive oxygen (say ozone) . On the other hand, since the advance- ment of social needs of the preservation and prospective use of national resources and the conservation of energies, the concept of durability as performance over time has been of an increasing importance in many fields, especially in building materials and/or components and elements [1-4], At the same time there have been an increasing demand for judgement methods of the degree of the progress of deterio- ration with time as well as those of deterioration behaviour and mechanisms of building materials, in connection with the evaluation of lifetimes of buildings. In order to evaluate the performance over time, it would be necessary for us to consider into the change of materials properties and the effect of deterioration layer on the performance, aparting from the change of performance requirements. This paper deals with the complex deterioration process of polymer mate- rials by the simultaneous action of UV-light and diffusive oxygen based on the unsteady state dynamical analysis, pay- ing special attention to the progress of deterioration innerward from surface. The aim is to derive the law of the increase of the depth of deterioration layer with time, and to consider into the reduction in flexural strength as the deterioration advances. It is shown that the degree of deterioration varies continuously with times and depths and that the depth of deterioration layer increases approximate- ly in accordance with the parabolic law concerning times [5], as is often observed in the cases of neutralization of con- cretes [6], and oxidation of metals [7]. I. OUTLINE OF THE MODEL FOR DETERIORATION PROCESS: The fundamental suppositions of the model for dynamical analysis are as follows: 1) Only polymer molecules which are activated into excited states by absorbing UV-light (photon flux; nQ ) near the absorption band characteristic of polymer materials can take part in photochemical reactions (reaction efficiency; n,mole concentration; c*) . 2) photo- chemical reactions are i) depolymerization of activated polymer molecules (1-st order reaction ;Ri=kiC*) , ii) photo- oxidation by the collision of diffusive oxygens (mole con- centration; C,, * diffusion flux; N., diffusion coefficient; Da) and activated polymer molecules (2-nd order reaction; Rz=k 2 C*Ca) . 3) Deteriorated polymer molecules (mole concent- ration; Cj>) don't diffuse and remain in the sites of origi- nal positions. 4) The influence of temperatures on the photochemical reactions is involved in material-property 368 constants such as rate constants (k,, k 2 ) and diffusion co- efficient (D^) according to the Arrhenius law. 4) Deter- ioration proceeds innerward from surface and the degree of deterioration varies with exposure times (t) and depths from surface (x) . As the deterioration advances, deteriorated polymer molecules having IR-activative functional groups such as -CO -OH are created, and the degree of deterioration is reflected as the increase of absorption (a) or the de- crease of reflection (R) in IR spectra. 5) The deterior- ation layers result in the reduction in flexural strength of polymer materials with the progress of deterioration. Based on the fundamental suppositions described above, the follow- ing three basic processes are considered. a) Absorption Process of UV-light: Considering the material balance within an inf initestimal plate of x~x + dx shown in Fig. 2, on the assumption that a UV-light (wave length ;A , photon flux;no) incides into the polymer material surface as described in Fig.l, the follow- ing differential equations can be obrained. 3C*/3t= -(n/N )3n/ 9 x - ki C* ' (1) _3n/.9x.= -Cg e^n ... (2) (N ;Avogadro's constant, C ; undeteriorated polymer mole- cules concentration at the initial state, ex ? mole absorp- tion coefficient) From equations (1) and (2) are derived equation (3). 3C*/3t= AC £x ex P^" c o e Xx)-ki C*-k 2 C*CA (3) (AErm /N ) b) Diffusion Process of Oxygen: Consider also the material balance of oxygen within the plate shown in Fig. 2, and the following equation is obtained. 3CA/3t= DA3 2 C /3x 2 -k 2 C*CA (4) c) Process of Photochemical Reactions: Overall reaction rate (R=Ri+R 2 ) is the creation rate of deteriorated polymer molecules, and it is also equal to the rate of increase of absorption coefficient in the absorption band of IR-active functional group (say, -CO group, wave- length; \' ) . Consequently, R=kiC* +k 2 C*C A =3CB/3t= 3a X '/3t (5) Here, kj ^Aiexpf-Ei /kT) , k 2 =A 2 exp(-E2/kT) (6) D A =DAoexp(-E /kT) (7) (T; absolute temperature, k; Boltzman constant, Ei, E2/E3; activated eneraies) II. BASIC DIFFERENTIAL EQUATIONS FOR DYNAMICAL ANALYSIS: For the purpose of simplifing analysis, the following two are presumed, but the results of analysis never fail to have the generality. i) Compared with the thickness of the plate of polymer materials (L) , the deterioration depth (5) can be considered to be small enough for us to regard the thickness of the plate as virtually infinitely large, ii) In photo- xidation reaction (R 2 =k2C*C A ), we can consider this reaction 369 as quasi-1 st-order reaction ( R 2 = kj'C A ) . Taking these considerations into account, basic differential equations for dynamical analysis are summarized as follows, together with the initial conditions and boundary conditions. 9C A /8 1= D A8C A /9x - kiC A ( 8) 3C*/at=AC e A expt -Qe,x )-k 1 C*-k{C A (9) 3C B /3 t =k l C*+^CA A (10) [Initial Conditions]; t^O, x=0: c*=C a =Cd=0 [Boundary Conditions] ; [I] t>0,x=0: C A =C A0 [II] t>0,x+«> :C A =-0 [III] t>0,x=6: 3Cb/8x=0 III. RESULTS OF DYNAMICAL ANALYSIS: Solving the siinultaneous differential equations under the given initial conditions and boundary conditions by Laplace transformation method, analytical solutions (11) -(16) are obtained in the form of non-dimension- nal concentrations. ^A^'^/CAO =(l/2){exp(-x^[/§[ erfc(x/2/DAt " ^T^~ ) A *__, tw +exp(x/ kr/D A )erfc(x/2/L^ + deft" )} (11) f -£ X 'S< C °! ^"^(-kxt)} X{ exp(-C e x x)- ( CA /C*)k^ } (12) ♦B=Cfe(x,t)/C * -[kjt- (l-exp^t)}] X{ exp(-C o e x x)/|r } +{(ki-k 1 )C A 0/C *} XJf^dt +a-exp(-k 1 t)lX (Caq/C*)^ (13)° VE9* B / 9 t={l-exp(-k 1 t)} exp(-C e A x)- (C AO /C *)k;exp(-k 1 t) *.- (14) Here , A Cf = C Ae x =C nn e x A erfc x=l-(2//jf)/o x exp(-^ 2 )d^ (15) Profiles of functions based on calculations of a computer are shown in Fig. 3 8. IV. DISCUSSIONS AND CONCLUSION: "" The degree of deterioration varies with times and depths, and influenced by photochemical reaction constants ( k 1# kj ) and diffusion constant ( D A ) . In Fig. 4* 6, 7, 8 are shown the changes of the space-distribution of each concentration functions with times, and the influence of photo- oxidation reaction and oxygen diffusion on the distribution of oxygen concentrations are shown in Fig. 3, 5, respectively. Deriving the time dependence of surface deterioration , we can obtain the equation (16 ■) . <(> B ( 0,t ) = [ kit- { 1- exp(- k,t) }]/ kj+( C A0 /C* ) [( kj- k x )t +1 - exp(-k 1 t)]= kjt( C A0 /C* )= k 2 C AQ t (16) This result shows that the surface deterioration advances approximately in proportion to exposure time, and that the degree of progress depends upon the surface oxygen concentration and photooxidation constant. The situation that the deterioration proceeds continuously innerward from surface with times can be proved by IR absorption spectra and, in fact, was examined by Watanabe et al[8] about outdoor-exposed polymer plates of polystyren,polyoxymethylen and so on. In Fig. 9a) ,9b) are shown their experimental data with a view to compare them with the theoretical re- sults. On the other hand, using the boundary condition [III] ,i. e. the condition Ste/stl £=0, we can obtain approximately the time dependence 370 of the depth of deterioration layer with the advance of deterioration as fo llows . 6= /jTt~ ? 3= 3 ( D A , k 1; ki) (17) From this result it can be shown that the depth of deterioration layer increases in proportion to the square root of exposure times. This para- bolic law was obtained for the first time for the deterioration of poly- mer materials as the natural devivation based on the unsteady state dy- namical analysis, supposing the simultaneous action of UV-light and dif- fusive oxygen. Recently the author has observed the same law experiment- ally in the deterioration by water of Water-glass Inorganic Polymer Films [9] . The parameter 3 which decides the progress of deterioration is a function of material-property constants ( D&, k lt kj ) ,and the degree of the progress of deterioration could be evaluated approximately by the parabolic law concerning times, if deterioration factors and these mate- ial constants are suitably estimated. Further, since these constants are functions of temperature as shown in equations (6) , (7) ,the effect of temperature on the degree of the progress of deterioration can be con- sidered to be involved implicitly in the parabolic law. Such unsteady state dynamical analysis would be rather applicable also to. other dete- rioration processes. The author has just applied the same analytical method to to the carbonation and neutralization process of concretes [10 ] . Defining a deterioration progress factor ( X ) as a ratio of the depth of deteriorated layer to the plate thickness , and considering that dete- rioration layers don't contribute to the flexural strength of polymer plates, the following law as to the the reduction in flexural strength with the advance of deterioration can be obtained. X(t)= 6(t)/L ; ( ^ X = 1 ) (18) F(t)= A{ L-6(t)} 2 = AL 2 [1- {6(t)/L}] 2 =F {l-X(t)} 2 (19) Here, A is a constant, and F Q is the flexural strength of polymer plate at undeteriorated state. It is supposed, as is often observed^ that the flexural strength is proportional to the second power of the thickness of sample plates. According to Watanabe et al[8] , their experimental re- sults showed that deterioration layers near surface of polystyren plates exposed five years under natural weathering , far from not contributing to the flexural strength of plates, does influence on the undeteriorated layers, resulting in the reduction in specific flexural strength in in- ner layers. This experimental facts might be understood by thinking that the deterioration progresses continuously deep into the polymer plates. As to this point, however, it will be necessary for us to study more in detail, and consider into the influence of deterrioration on mechanical strength such as compressive, tensile and flexural strength. As concluding remarks, the deterioration of polymer materials progresses innerward from surface, and the degree of deterioration varies with times and depths. The depth of deterioration layer increases according to the parabolic law concerning times. The deterioration layer influ- ences the mechanical strength (say, flexural strength) , resulting in the reduction in the total strength of polymer plate . m REFERENCES 1. Nireki, T. , "Performance Evaluation of Vertical External Wall — Evaluation of Surface Fiishing Materials," Proceedings, RILEM/ASTM/ CIB Symposium, Otaniemi, Finland, August, 1977. 2. Nireki, T., "An Approach to Durability Studies on the Basis of Performance Concept," Proceedings, 7th CIB Congress, Edinburgh, United Kingdom, September, 1977. 3. Nireki, T. et al., "Application of Performance Concept in Japan" Proceedings, 7th CIB Congress, Edinburgh, September, 1977. 4. Nireki, T. , "Series of Durability Research, 7 th Report —Accelera- tion Factor of Snshine Carbon Arcs," Transactions of Architectural Institute of Japan, 1975. 5. Fukushima, T., "Deterioration Process of Polvmer Materials and Its Correlation with Depths from Surface — Dynamical Analysis Based on Kinetic Theory on the Complex Deterioration by the Simultaneous Action of UV-light and Diffusive Oxygen," Proceedings, Annual meet- ing of Architectural Society of Japan, Kanto, September, 1979. 6. Hamada, M. , "Carbonation of Concrete," Proceedings, 5th Inter- national Symposium on the Chemistry of Cement, Tokyo, 1968, p. 349. 7. Motto, N. F., Transactions of Faraday Society , Vol. 35, 1939, p. 35 ; Vol. 36, 1940, p. 472; Vol. 43, 1947, p. 429. 8. Watanabe, Y., Kitajima, F., and Hat tori, S., "Erosion Process in the Deterioration of Polymer Materials under Outdoor and Accelerat- ed Exposure," Proceedings, 15th Symposium on Polymer Research Results of Japan Industrial Technology Association, Tokyo, October 19-20,1979, p. 177. 9. Fukushima, T,, "Effective Use of Inorganic Polymers for Building Materials — Hardening Processes of Cements and Water Glasses and the Influence of Hardening Conditions on Their Durability," Annual Report of Building Research Institute ,Tsukuba, 1979. 10. Fukushima, T., "Effective Use of Inorganic Polymers for Building Materials — Dynamical Anlysis Based on Chemical Kinetics on the Neutralization of Concretes and on the Suppressing Effect of Water- glass Inorganic Polymer Films," Annual Report of Building Research Institute, Tsukuba, 1980, in Press. "Consideration Based on Kinetic Theory on the Correletion between Carbonation and Neutralization Processes of Concretes," Proceedings, Annual Meeting of Architec- tural Society of Japan, Kyushu, September, 1981, in submission. 372 Figures zA n x0 =I AO -x/hc "(l-RtfcO) ->x ^y Fig. 1 incident process of a UV-light upon the surface of polymer materials x x + dx nix N A I* dx n|x+dx — > NAl x+ dx Fig. 2 material fluxes within an infini te- nia! polymer plates 373 Fig. 3 influence of photooxidation reaction on the distribution of oxygen concentration j — i — i i i ' j L J 1 ' (1) 1=0.01 (year) (2) '0 09 (3) - 25 (M » 0.49 (5) * 1.00 (6) » /, 00 D A = 0.04 (mni ? /year) kj r 1 (year) T — -i 1 1 1 1 )0 1.5 DEPTH FROM SURFACE x(mrn) Fig. 4 time dependence of the distribution of oxygen concentration 3>A ■ ' ' ■ I I I I 1 1— I 4w »J L ( I) D A s0.01 (mm'/year) (2) •• 0'. (3) > 09 (A) - 0'.9 [1 = 1 (year) (5) '' 1.00 Kr 1 U/year) 0.5 10 DEPTH from SURFACE x(mm) 2 Fig. 5 influence of diffusion on the distribution of oxygen concentration j i i i i i_ j L j i i i_ (1) 1*0.01 (year] (2) i 09 (3) ; 0.25 U) * 0/.9 (5) '/ 0.64 D =0 04 (rrrrfyyear) kf =1 [ )/ ye(ii) k| =5 ( l/year) C,,6 A =I0 (mm 1 ] q*/cJ=o.oi T~~" 1 1 1 1 1 1 1 1 r— t- 2 0.3 0./. 5 0.6 0.7 0.8 0.9 1.0 DEPTH FROM SURFACE x ( mm ) Fig. 6 tijne dependence of the distribution of the concentration of activated polymer molecules 375 D A = 0.04 (mir?/yearj k ( ' = l (1 /year) t w- a a. 8 O 9)DDS dA{P)9^ a 383 Fig. 2: Weathering cell (detail) 384 \„. V \ \ \ \ \ \ 0) (UUUU) J 2 i ^-o O) c a> w. -«—» U) « i ^O ©** IT) £ CO t . u. ^^ o o a ■6 I ft: o CO Ll. 385 T \ T3 1 2 1 1 i I I I 0) o o CO -Q c — a X a u) (wui)j 2 in ll. 386 WEATHERING OF SOME POLYMERS AFTER 4 YEARS' EXPOSURE R. COPE and G. REVIRAND Centre Scientifique et Technique du Batiment 24, rue Joseph Fourier 38400 Saint Martin d'Heres - France Abstract - Because the natural weathering tests are a basic reference for any approach of the durability of polymeric materials, and in order to check either the evolution of different materials as measured of arti- ficial degradation test, or the classification of materials based on short term evolution, a weathering program for periods up to 10 years has been set up. This program includes the study of several transparent or opaque PVC, translucent glass reinforced polyesters or opaque molding compounds, a common polycarbonate, elastomeric materials. Every weathered polymer has been characterised in consideration of their aspect properties and eventually their optical properties. For other functional properties a more specific approach was worked out. Key words - aspect, GRP, mechanical properties, optical properties, polycarbonate, polychloroprene, PVC, weathering For about 10 years the CSTB (materials physics division) has been carrying out a study on the weathering of materials used in buildings ; the main objectives are a better understanding of the processus of deterioration and the definition and development of tests for measuring the effects of weathering on polymeric industrial products. The first stages of this program led to the study of damage resulting either from artificial degradation, or from exposure to short term weathering. In the second phase, and because reference to weathering is fundamental to any approach of the durability of materials a program of exposure tests to climatic agents was carried out at different locations over a period of ten years. The purpose of this contribution is to point out the results of 4 years' weathering. Five geographical sites in France and Africa were chosen for setting up stations for exposure to climatic agents. The three French stations are Nantes (temperate Atlantic climate), Champs sur Marne near Paris (temperate climate) and Grenoble (temperate industrial climate). Two stations were set up in Africa, one at Fez (Morocco) which is characte- rized by a hot, dry sub-desert type climate, the other at Abidjan in the 387 Ivory Coast, characterized by a hot and humid, tropical type climate. The samples are exposed on racks at 45° inclination facing South. This program deals with different PVC's whose behaviour is illustrated by the evolution over time of a transparent sample presenting improved mechanical behaviour ; results relating to translucent glass reinforced polyesters and to molding compound are also presented ; a standard polycarbonate is investigated and lastly, the behaviour of an elastome- ric material, polychloroprene, is studied. 1. PVC Four samples of PVC of differing formulation or mode of transformation were studied. These products are considered to have a life-time which is satisfactory outdoors under normal conditions of mechanical stress. These samples were studied following a procedure enabling evaluation of the principal signs resulting from exposure to natural weathering. More precisely, the experimental procedure includes : . measurement of the residual thermal stability of the material, deter- mined by the dehydrochlorination test (DHC), . measurement of the mechanical behaviour of the material under the effects of impact : the impact-strength test, . measurement of the optical properties and more precisely of the light transmission for. translucent or transparent samples, . finally, an evaluation of the surface deterioration using micrographs from scanning electron microscopy observation or surface profiles with a roughometer. The results presented below are thus with respect to a single product, but constitue a very significant example of the behaviour of PVC material . 1.1. Residual thermal stability (s) (1) This test is carried out according to the experimental procedure for which standardization is in progress (ISO DP 182). In figure 1 it can be seen that s decreases regularly to about 50% of its initial value in the three French stations, then stabilises at this value. This behaviour differs in the two african stations where s decreases continuously, reaching 10% of its initial value after 2 years in Fez, and after 4 years in Abidjan. The parameter s thus reveals a greater deterioration of the material in Fez and Abidjan after 2 and 4 years' exposure. 388 1.2. Impact strength - This test, carried out according to the standardized experimental pro- cedure DIN 53.448 (1) revealed the results for which figure 2 gives a graphic representation. The material does not appear to have undergone any alteration as far as impact strength is concerned in the three French stations after 4 years' exposure ; whereas exposure for the same time in a humid tropical climate (Abidjan) results in an appreciable drop in the mechanical parameter measured (AA/A ^ 45%). 1.3. Optical properties The spectra for the total (direct and diffuse) light transmission T are traced for each sample. Figure 3 shows explicitly the relative variation of T at 550 nm wavelength as a function of the exposure time for the different stations. It can be noted, as above, that only the sample exposed for 4 years in Abidjan presents a significant deterioration in its optical transmission properties. 1.4. Deterioration of the surface Figure 4 and 5 give a qualitative profile of the superficial state of the material exposed in Grenoble and in Abidjan. There is no discerna- ble difference even after 4 years, between the states of the test pieces exposed in Grenoble (fig. 4), whereas the profiles of the test pieces exposed for four years in Abidjan present fairly pronounced disturbances (fig. 5). The same observations can be made after examina- tion of micrographs 1, 2 and 3 made with a scanning electron microsco- pe : only the surface of the sample exposed for four years in Abidjan reveals a significant deterioration in its appearance. 1.5. Comments After four years' exposure to weathering, PVC does not present a signi- ficant state of deterioration, except for samples exposed in a humid, tropical climate. Parameter s, representing the residual thermal stability of the material, appears to be a particularly sensitive indication of the effects of weathering. For the material studied, the three French stations are very similar with respect to the damage incurred. Lastly, the material exposed to radiation in a Weatherometer type climatic unit (6000 W filtered xenon tube ; intensity of radiation at sample ^ 800 W/m^ ; temperature of the air in the unit around 30°C ; temperature of the blackbody *\40°C; 389 relative humidity ^60%) presented no deterioration of its mechanical properties after 3000 hours' exposure, whereas the parameter s had dropped by about 20%. 2 - GLASS REINFORCED POLYESTERS (GRP) Seven types of GRP were studied under this project, each one particu- larly differentiated by the formulation of the polyester resin. Results for three of these seven materials will be presented : a stratified material made with a standard resin, the second with improved fire behaviour, and finally a standard molding compound material. The samples were characterized according to the following experimental procedure : . appreciation of the mechanical behaviour of the material and more precisely of its elastic properties (measurement of the torsional stiffness modulus) , . measurement of the optical properties of transparent or translucent products. 2.1. Mechanical behaviour The elastic properties of the composite materials are measured by means of a torsion pendulum described elsewhere (2), enabling measurement on the one hand of the torsional stiffness modulus (G), and on the other hand of the internal friction coefficient of the material. Figure 6, relating to the molding compound reveals a considerable change in the mechanical properties of this material after two years' exposure in the African stations, whereas the effects due to weathering in the three French stations remain slight ; in contrast the mechanical beha- viour for all the rest of the GRP's studied does not appear to be notably damaged after four years' exposure in each of the stations, even in the humid, tropical climate. 2.2. Optical properties The evolution over time for light transmission is shown in figures 7 and 8. An appreciable change in the optical properties of the standard material can be measured (fig. 7) after two years' weathering, whiche- ver the exposure station, but the deterioration is still greater in Abidjan. Finally, the poor performance of polyesters made with HET acid can be verified (fig. 8), and for these materials the optical properties are changed after the first year of exposure, reaching a drop in light transmission at 550 nm ranging from 20 to 40% depending on the 390 exposure station, with the humid, tropical climate remaining the most aggressive for the material. Without attempting to go into a detailed description of the appearance of the samples, it is nevertheless interesting to underline that these materials present a very clear modification of their appearance, resul- ting from a certain yellowing and a superficial defibering which can be considerable. 2.3. Comments These results verify that the humid, tropical climate in Abidjan is shown to be very aggressive for GRP's whose mechanical properties are nevertheless preserved with the exception of the molding compound samples. It should also be underlined that the deterioration the sur- face of the material may lead fairly quickly to unfit for use covering products. 3 - POLYCARBONATE The material exposed is a standard industrial polycarbonate without superficial anti-UV protection. Only the optical properties and the surface state of the material are presented, as the mechanical proper- ties and in particular the impact resistance of the polycarbonate present no deterioration at all after 4 years' exposure to natural weathering, whatever the climatic conditions studied. 3.1. Optical properties Figure 9 shows the light transmission parameter for the material at 550 nm. The samples exposed in the French stations do not present any change whereas a slight decrease in light transmission is observed after two years' exposure for samples in the humid, tropical climate 05%). 3.2. The surface state Qualitative profiles were produced just as for PVC using a roughometer (fig. 10 and 11) and observations from the scanning electron microscope (photos 4, 5, 6 and 7). The surface profiles of the samples exposed in Grenoble (fig. 10) or in Abidjan (fig. 11) show the first signs of weathering in the second year of exposure ; these profiles deepen considerably after four years. If the sample exposed for two years in Grenoble does not yet show any microcracks under the scanning electron microscope (photo 4), a network of microcracks clearly appears after four years' exposure in the same 391 station (photo 5). These microcracks are widely spread on the surface of the test pieces exposed in Abidjan, and a real polygon type network can be observed (photos 6 and 7). This superficial deterioration is only discernable to the naked eye in the form of a slight superficial dullness. 3.3. comments The polycarbonate studied is wery rapidly subject to irreversible damage in a humid, tropical climate (two years). This damage does not appear to alter the functional properties of the product even if the transparent material becomes progressively translucent due to superfi- cial dullness. 4 - POLYCHLOROPRENE The elastomer studied is an industrial formula treated to prevent oxidation, exposed at rest and under permanent stress corresponding to a 20% relative elongation. The parameters used to monitor the weathering of the polychloroprene are its molecular structure and its mechanical behaviour. The molecular structure is verified using a swelling test, while the mechanical behaviour is evaluated from the results of traction tests and relaxation tests (3). 4.1. The molecular structure Figure 12 describes the variation of the molecular weight between cross-links measured using a swelling test of polychloroprene in CHCK, with respect to the exposure time (with unstrained samples). A slight decrease of the molecular weight is noted in the three French stations, with this variation being greater in Abidjan (^20%). This result thus indicates a certain reticulation of polychloroprene over time. The results of tests carried out using samples exposed under stress present a similar, almost identical, profile of variation to those given above. 4.2. Mechanical behaviour The stress and strain at breaking point are measured in a traction test (fig. 13) in normal humidity and temperature conditions and at a speed of 400 mm/mn. The stress at breaking point presents no significant variation with respect to exposure time, whichever the station, whereas strain at breaking decreases appreciably (^40% after 4 years in Abidjan). 392 As for the molecular weight, the influence of stress on the test pieces during exposure is negligible. Lastly, tests under continuous and non-continuous relaxation do not contribute any new information as far as the mechanical behaviour of polychloroprene is concerned. 4.3. Comments The polychloroprene studied does not present any significant changes in its properties after four years' exposure, either with the test pieces under stress or at rest. Only the beginnings of reticulation are measu- rable, whose influence on the mechanical behaviour is reflected by a slight embrittlement (decrease in straing at breaking). The humid, tropical climate appears yet again to be the most aggressive. 5 - CONCLUSIONS As far as the network of exposure stations is concerned, it is notable that the three French stations at Nantes, Champs sur Marne and Grenoble are very close in the signs of weathering exhibited whatever the mate- rial under consideration, in spite of clearly differentiated climatic trends. In view of operational anomalies, it is difficult to undertake a quali- tative appreciation of the Fez station. On the other hand, it is clear that the humid, tropical climate of the Abidjan station is by far the most aggressive whatever the material studied. As far as the materials are concerned, the exposure of PVC samples does not reveal any products weathering prematuraly and highlights the sensitivity of the DHC test to signs of weathering. The performance of the GRP corresponds to that expected (high level of preservation of mechanical properties - deterioration of the surface of materials) . Weathering of the polycarbonate, shown also by superficial degradation appears fairly quickly in the tropical humid climate. The mechanical behaviour of polychloroprene does not reveal significant damage. 393 6 - BIBLIOGRAPHY 1. "Directives communes pour 1 'evaluation des produits en PVC rigide utilises a 1 'exterieur dans le batiment", Cahiers du CSTB n° 1163, livraison 136, jan./fev. 1973 2. J.L. CHEVALIER, R. COPE, P. PEGUIN, G. ROUX, "La mesure des caracte- ristiques mecaniques des materiaux", Cahiers du CSTB n° 1324, livraison 158, mai 1975 3. J.L. CHEVALIER et al . , "Etude du vieillissement des joints en poly- choroprene et en EPDM", Cahiers du CSTB n° 1616, livraison 205, decembre 1979 394 PVC (X2000) Photo n° 1 - reference - " pr* \ '" f* M& ' .-*. :■ ft ■' 4, r - 1 Photo n° 2 - Grenoble, 4 years Photo n° 3 - Abidjan, 4 years 395 POLYCARBONATE ( x zoo) Photo n° 4 - reference - Pho to n 9 5 - Grenoble, 4 years Pnoto n ° 6 " Abidjan, 2 years Photo n° 7 - Abidjan, 4 years 396 O > Ql O X a 9 ,„_, C J% ® «fl I « 4 fc *> TJ 9 D C .„. N c. jC e jQ $ t£ O z < L. A A A A A 1 1 I 1 1 i ! ! I ««» <\l C? V tf> «-i QL hi CO ^ Kim ^ 397 C guss/ttfs • £>| ^ V 14 CO OS*- <* / ^ N C 1 ! •-> C MM N X r M 9 O 2 < U. A A A A 8 t 1 ( 1 ! 3 1 A «— N W ^ W <§ ® <^> j. \j» © CO to < >- CM G> 398 w CGTu > * 0. KJ hi U Z < h- H £ z < x o o ® 8 & N U Z < L. A A I i I I « n n t w A I I A ! I 8 J €9 T K5 « < kl CSJ CK) 2 < U. A A A A A i 1 ! I 1 1 1 — CM CO ^r LD QJ s ? 8 eg T cn ©o / e - ©a 402 10 C\. > — CO<\l o z H CO z < L @> Q ® €9 o C4 CO ? I i 1 8 «55 CO (0 < <\* © ? cn «a / *j 403 rs > Q. hJ O Z < »- H r z < CO C33 <0 < Id CD © © «•» (VI 1 1 1 ! i cn °i / °i - i 404 04» r\ • llj *** o z c < 1- Jk h- L H £ (0 > Z — < « 0. H w (0 CO ft < >• w MM* L CD 19 » ft •*» C X L> Z C •-> - N < U. A 1 A A 1 ! 1 1 A A 1 1 1 ( •"• CM CO <* 1ft i- 1 - - © 2 T ? <9 en °jl / °i - i 405 406 407 » w 408 Si V y z u ft I G ff G J Z > J G a. m t fei tt 2 < 1 5 H N d y O Z « k. - M w t 10 e e g < M (2««/m»f) snau (S) llfllt 409 DURABILITY OF PVC WINDOW FRAMES P . Svane Technology Insititute Denmark Abstract: Originating from Western Germany, the use of plastic windows - primarily extruded, hard, modified PVC - has spread to the surrounding countries during the last 2-5 years. Sad experience with fungal attack and high maintenance costs for wooden windows certainly have helped the plastic windows to a markedly increased share of the market. One of the major advantages of this relatively new materials for frame- and sash-production is claimed to be its maintenance-free surface. Surprisingly this claim did not show to be convincingly well founded. Many investigations had been carried out, but mostly on the mechanical properties - e.g. impact resistance - of the material. Studies of surface structure and surface appearance were scarce. With the kind permission from the "Farbwerke Hoechst AG" we have been able to carry out introductory investigations on naturally aged PVC- materials weathered 18 years, exposed southwards in the region of Frank- furt a.M., Western Germany. On the surfaces coloar and gloss were measured, partly on the weatheered surfaces, and partly on "fresh" surfaces obtained by removing the eroded top-layer. Also new PVC-material was investigated in this manner. After this the weathering of the surfaces was continued: a) accelerated in the weather-0 meter. b) by natural exposure 45° to the south in Denmark. Finally the measurements of the color and gloss were repeated. The investigations indicated that changes in appearance of white or light greyish PVC after approx. 20 years of natural exposure in an industrial environment were very moderate and probably would continue to be so for several future years. Scanning Electron Microscopy of the different surfaces showed rather similar degradation patterns on the artificially and the naturally aged surfaces. This confirms the reliability of the investigation to some extent. 410 Eventually, heat stability measurements (HCl-evolution) and ATR-analysis (determination of chemical bonds in the surface) were carried on thin slices of the material. It was thus confirmed that chemical changes did occur to a depth of approx. 200 \sm in the investigated 3 mm thick material. The results are encouraging with respect to the surface-durability of white or light greyish PVC intended for plastic windows. 411 DURABILITY OF GLASS REINFORCED POLYESTER USED IN CLADDINGS FOR BUILDINGS A.J. Brookes, B.Arch. PhD. Liverpool University School of Architecture P.O. Box 1^7 t Liverpool, L69 3BX, England. Abstract: To achieve a quality product it is necessary for GRP panels to be manufactured in the right conditions and to correct specificat- ion,.- It may be necessary for architects to have closer quality control during production than for more established products. The problems and proposed solutions of achieving higher fire resistance in glass reinforced plastic panels are discussed. In particular the process of attaining fire resistance, up to one hour, through the use of special fillers and gel coats creates inherent problems due to increased susceptibility to fading from ultra violet light. Various surface finishes and textures and their weathering characteristics are described. The selection of colour can significantly effect the overall weatherability of the panel with regard to ultra violet degredation and possible heat build up. Condensation on the back of the panel can lead to a breaking down of the laminate due to attack by moisture through capillary action. Problems can be associated with the use of timber or metal stiffeners. Delamination of the laminate from the core can occur using sandwich construction. Under these conditions the bond between the two materials becomes critical. It is necessary to predict the expected surface temperatures of the panels. Careful detailing of all fixings and jointing is essential to their proper performance in use. Key words: Building component failure; quality control; fire resistance; weathering; rib patterning; sandwich construction; delamination; condensation; jointing; fixing. 412 QUALITY CONTROL DURING MANUFACTURE: To understand how G.R.P. behaves in use it is necessary to appreciate the process of its manufacture. The 'gel coat 1 , a layer of resin which forms the outer surface of the G.R.P. provides protection for a matrix of glass fibre reinforcement and a polyester (lay up) resin. Coloured pigments can be added to either the resin or the gel coat and in addition chemical and/or fillers can be added to improve the fire retardancy of the composite. The liquid resin is polymerised or 'cured 1 into a hard solid by the addition of chemical catalysts which are mixed in the resin shortly before the material is laid up in the mould with the glass fibre. Excessive catalyst will cause the mixture to overheat since the chemical reaction produced is exothermic resulting in cracking and crazing whilst inadequate catalyst will produce an inadequate state of curing. Photo 1 Following their removal from the mould, panels are usually 'post cured' in boxes where temperature and humidity are controlled. As the performance and durability of the G.R.P. product depends mainly on the cure of the polyester resin, it is essential that there should be some means of establishing whether or not the laminate is suspect. Weighing the laminate will give a check as to whether a layer of reinforcement has been omitted; whilst the 'Barcol Test' for hardness, although not giving an absolute measurement of the cure, will determine a really bad laminate. Photo 2 It is not outside the designer's field to be able to set out these conditions of manufacture in his specification and he may request record sheets for each unit produced, which contain details of castings, mix and tests carried out. On large projects in U.K. the architects have found it necessary to have direct quality control over all the units during manufacture. Photo 3 The specification for the works at Mondial House telecommunications centre in London, for example, includes a clause that "each panel will be inspected dimensionally and for gel surface defects and signs of poor laminating, by the consultants and marked if approved". The G.R.P. panels for Herman Miller factory at Bath were also inspected individually by the architects, Farrel & Grimshaw, in the factory. On the other hand it is sometimes claimed, by the manufacturers, that architects tend to expect a higher degree of finish with G.R.P. than they would require from more established products. Photo k FIRE RESISTANCE AND WEATHERING: One of the major problems of G.R.P. is that of colour fastness in positions exposed to ultra-violet rays whenever fillers or pigments are added in the gel coat. Thus the 413 inclusion of fire retardant additives to the gel coat to reach some sort of fire resistance tends to weaken the material's capacity to resist weathering. In U.K. a class '0' fire resistant laminate, according to British Standard V76, is required for claddings on buildings above 13 metres high or within 1 metre of the boundary. At present this classification can be achieved in one of two ways:- 1) by using a gel-coat and laminating resin which both contain fire retardant additives. The weathering qualities of the laminate produced in this way are reduced by the presence of a fire retardant in the gel coat. Weathering normally takes the form of loss of gloss, changes of colour or shade, and possible chalking of the surface. Such chalking can sometimes be removed by the application of T-cut polish or a similar product. 2) by proprietary systems formulated by companies specialising in G.R.P. claddings. Most, but not all, of these systems rely on a two pack polyurethane surface coating to protect the gel coat from ultra-violet light. Where a relaxation from the Class '0' requirement is permitted a general purpose isophthalic gel coat can be used, which in itself has no fire retardant characteristics, backed with a laminating resin with a fire retardant additive. The addition of this agent to the resin alone does not effect the weatherability of the completed laminate as the gel coat then forms a protective layer. This system has been widely used in the U K Results of fire tests carried out by Yarsley Laboratories on such panels showed that there was no uncontrolled spread of flame over the surface and even after an hour's exposure the panel still retained its form. Photo 5 COLOUR AND TEXTURE: The selection of the colour can significantly affect the overall weatherability of the panels. Before deciding on the colour it is advisable for the architect to discuss this with the manufacturer and his pigment supplier. In general the stronger colours such as oranges and reds have a tendency to fade under the effect of ultra-violet light. Although lacquers based on U.V. stabilised acrylic resins for protecting the gel coat are available, these are not widely used in U.K. No completely satisfactory method of assessing weathering resistance of plastics materials in service by laboratory simulation has yet been found. Weatherometer testing and Xeron arc testing (Reference I.S.O. R. 879: 1968) are popular methods, but these can, at most, only give an indication of performance in use. After 1,000 hours exposure to the 'Xeron test' any change of colour should be moderate and uniform. One possible method to counter effect the change in colour fastness between adjacent panels is to use the striped effect of colour as at the Olivetti factory designed by James Stirling. Photo 6 414 It is not uncommon, because of the ease of moulding G.R.P. panels, that they are used in conjunction with other materials for solving special corner or flashing details. For example, at the factory at Winnick Quay, Warrington, architects Farrel 8c Grimshaw corner panels were site painted to match the silver grey aluminium sandwich wall panels. This mixing of materials can lead to problems of colour matching and differences in rates of colour change. In general colouring of G.R.P. panels is best done by pigment in the resin or gel coat and not as an applied finish. Photo 7 G.R.P. panels are not maintenance free and self cleansing as many people think. Smooth shining G.R.P. in anything approaching a flat panel may suffer from apparent distortion. Vertical texturing offers a pleasing finish which tones down bright colours and reduces visual distraction of dirt and rain streaking. The incorporation of a mini rib finish at Mondial House has resulted in the panels only now having to be washed down after 7 years since initial installation. Special grit and aggregate finishes can also be used to reduce the effects of weathering and surface defects. Surface texturing of the gel coat can be produced easily and cheaply by incorporating a textured material such as riven slate or vinyl wallpaper on to the mould surface before applying a releasing agent. RIB PATTERNING AND DELAMINATION : Although strong, G.R.P. has a low modulus of elasticity compared with many traditional materials. The material will bend or creep under its own weight unless stiffened by ribs or by designing for stiffness in a geometric form or by using sandwich construction. Ribs are preferably pre-formed in G.R.P. Problems may be associated with the use of timber stiff eners resulting in rot and therefore failure of the rib to perform. Where steel sections are laminated, into the back of the G.R.P. skins it should be noted that this material has a differing co-efficient of thermal expansion to that of the G.R.P. Aluminium sections are better in this respect. Rib patterning or shadowing on the face of the panel has also been known to occur and the ribs should be vented to allow an even balance of temperature over the total face area of the panel, both during curing and, more importantly, on site due to heat build up. Photo 8 In sandwich construction two skins of G.R.P. are separated by an insulating core. The edge of the panel must be carried out by bringing the two laminates together. In order to ensure maximum structural connection between the two skins it is customary to use G.R.P. connectors at intervals when using foamed polyure thane. Problems do occur however in choosing an appropriate core material and serious defects through delamination of the insulative core from the panel itself may occur due to the differential expansion and contraction rates of the two materials. The main difficulty is that 415 the outer skin alters dimensionally, whereas the inner one does not, and thus the edge bond between the laminates and the core becomes critical. The architect should inform the manufacturer of any special conditions of use of the panels. For example, if high external temperatures are expected, such as those which would occur with the use of dark coloured wall panels on south facing elevations, resins with higher heat distortion points may also need to be used. Photo 9 Condensation on the back face of the panel, which in contrast to the outer skin often had no gel coat protection, has led to a premature breakdown of the laminate if the glass fibres are exposed at the surface. Current specifications tend to recommend a nominal flow coat on the back of panels for protection. Finally, specification of detailing, fixing and jointing is just as important as requirements for resin, pigments additives and type of glass to be used. Many problems in the past have been attributed to the properties of the G.R.P. when in essence the problem may often lie elsewhere. As with all panel systems, the component will only perform as well as its jointing and fixing will allow it to. 416 REFERENCES 1. Reid & O'Brien, "Glass Fibre Reinforced Plastics for Buildings", Architects Journal , 21 March 1973* p.699 - 706. *t April 1973, p.817 - 826. 2~May 1973, p.1035 - 10^7. 2. Reid & O'Brien, "Principles of Detailing G.R.P. Claddings", Architects Journal, 18 September 197^, p. 697-701. 2 October, 197^, p. 815 - 817. IS~October 197^, p. 9^5 - 9^6. 30 October 197^, p.1061 - 1063. 6 November 197^, p. 1121 - 1122. 27 November 197^, p. 1289 - 1291. 3« Scott Bader Co. Ltd., "Crystic Polyester Handbook", Lund Humphries, London 1980. *f. Hollaway, L. "The Use of Plastics for Loadbearing and Infil Panels" Manning Ripley Publishing, Ltd. , 1975* 5» Hollaway, L. "Design and Specification of G.R.P. Claddings", Manning Ripley Publishing, Ltd., 1978. 6. Blaga, A. "G.R.P. Composite Materials in Construction", Industrial Forum, Vol. 9 No. 1, 1978, pp. 27 - 32. 417 Photo 1 cracking and crazing of the gel coat. (Photo could not be reproduced) Photo 2 Barcol Test for hardness 418 Photo 3 Mondial House telecommunications centre in London Photo 4 Herman Miller factory at Bath. A19 (Photo could not be reproduced) Photo 5 Fire tests on GRP panels at Yarsley laboratories, 420 Photo 6 Olivetti training centre Haslemere, 421 Photo 7 Winnick Quay, Warrington corner detail Photo 8 Venting of g.r.p. ribs. 422 Photo 9 The Vanessa Redgrave Primary School at Hammersmith Photo 10 Early use of moulded GRP for British Rail relay rooms 423 EARLY DETECTION OF POLYMERIC DEGRADATION USING ELECTRON MICROSCOPY M. E. McKnight and W. E. Byrd National Bureau of Standards Washington, D. C. 20234 Abstract: Assessing the durability of materials in varying environ- ments is increasingly important because of the use of new materials which have little or no service history. One approach to this problem is to find sensitive ways to detect and quantitate degradation early in the service life of materials. Comparisons between accelerated and service life testing can be used to reveal whether degradation mechan- isms are the same and to relate time to failure in accelerated testing to service life. Initial work to detect incipient degradation of poly (ethylene terephthalate) and polycarbonate is discussed. Samples examined were obtained from previous NBS research projects to develop data needed as the technical basis for standards for solar collectors and cover plates used on collectors and included samples exposed both to natural weathering at several locations and to accelerated labora- tory weathering. All samples exposed to natural weathering eventually showed crazing and brittle features. Surface degradation after most accelerated testing was different than after natural weathering. Key Words: Accelerated testing; cover plate materials; degradation; microscopy; poly (ethylene terephthalate); polycarbonate; solar energy collectors 1. Introduction Methods for quickly and relaiable assessing the durability of building materials are becoming increasingly more important as more and more new products become available. The need to use new products which have little or no service history is often mandated by the legislative process (i.e., Clean Air Acts, Hazard Materials Act, etc.) or caused by the changing requirements and desires of society. Knowledge of expect- ed performance is essential to the designers of buildings. For tradi- tional materials, this knowledge is frequently obtained from past ser- vice use of the materials. For new materials, it is often impractical or perhaps impossible to wait for results of long-term testing. Hence, accelerated testing is often used to estimate service performance of building materials. However, there are several problems associated with accelerated testing. These include 1) quantitative estimates of service life are seldom made since most accelerated tests are compara- tive rather than predictive, 2) degradation mechanisms are not usually 424 fully understood so that it is difficult to design meaningful accelera- ted tests, 3) factors affecting service life are numerous and difficult to duplicate and quantitate, and A) quantitative evaluations of service life of degraded materials are often difficult to make because of the deficiencies of accelerated tests and because of lack of long-term exposure data. A systematic approach to durability testing of building materials is detailed in ASTM E632-78, "Recommended Practice for Developing Short- Term Accelerated Tests for Prediction of the Service Life of Building Components and Materials "[1]. This procedure emphasizes the impor- tance of defining in-use performance requirements and criteria for failure, characterizing the degradation mechanisms, establishing ways of measuring changes in performance properties, and quantitating rele- vant environmental stresses. In the research described in this paper, we have concentrated on the measurement step. Techniques have been investigated to detect inci- pient changes in material characteristics which are related to perfor- mance requirements. Sensitive techniques to detect early degradation could modify durability testing in two ways. First, it may be possible to obtain degradation rates of materials exposed in service environ- ments in similar times as in some accelerated tests. Although one would only obtain degradation rate for one environment, the problem of duplicating the environmental stresses in an accelerated test would be eliminated. Second, if accelerated testing were to be done, the time to establish rate of degradation would be reduced thereby reducing the cost of the testing. In addition, since techniques to detect early degradation would undoubtedly involve micro-structural material changes, information on degradation mechanisms would be obtained. Several techniques have been used to detect and monitor small changes in polymeric materials which occur as a result of aging. These include infrared spectroscopy, thermal analysis, liquid and gas chromatography, x-ray diffraction, mechanical measurements, and microscopy. In this report, the use of scanning electron microscopy to examine solar cover plate (glazing) materials after both accelerated aging and natural weathering is discussed. 2. Microscopic Examination 2.1 Material Description Two materials which were exposed as part of previous work done at NBS to obtain data as a basis for standards development work (2,3'J were selected for microscopic examination for early detection of degradation. These materials were used commercially as glazing materials at the time the NBS work was initiated (1975). One is a 0.15 mm (6 mil) film based on poly (ethylene terephthalate) , (PET) resin and the other a sheet material based on a bisphenol-A polycarbonate resin. These materials 425 were chosen since performance properties deteriorated substantially during exterior exposure. The transmittance decreased several percent during the course of the exposure, the materials acquired a cloudy appearance, and PET also became brittle. The PET material was exposed to natural weathering at four locations; Florida (hot, low to moderate uv, humid), Arizona (hot, high uv, dry), Maryland (warm, low to moderate uv, humid) and northern California (warm, dry but with morning dampness, moderate uv) . PET was exposed to natural weathering as part of full scale and miniature solar collectors and it was also exposed in several accelerated laboratory tests [3]. Similar exposures were made of polycarbonate although no full scale collectors were exposed [2], Samples of the exposed materials were analyzed for transmittance in the previous studies prior to their use in this research. 2.2 Procedure and Results The surfaces of all the specimens were examined with the scanning elec- tron microscope. With increasing exposure time, the surfaces of both materials exposed outdoors crazed. Also, some cross sections of PET breaks formed during tensile tests were examined. Results of the PET examination will be discussed first. 2 2 Surface crazing of PET samples after eighty 17,000 kJ/m (1500 BTU/ft ) days as a result of Florida exposure was much better defined than after a similar exposure time in Maryland, California, and Arizona. After one hundred and sixty 17,000 kJ/m days, samples from all four outdoor exposure sites showed some crazing, although crazing from Arizona and California samples could barely be observed. In all locations, the cracks deepen and become better defined as exposure time increases. An example of this is shown for PET in figure 1. For approximately equi- valent exposures in Florida, crazing of PET from a full scale collector is more severe than that from a miniature collector. This is illustra- ted in figure 2. An important difference in surface deterioration occurs between mater- ials exposed in service environments and in accelerated aging (heat, heat and moisture, and xenon-arc exposure). No crazing is seen after 2000 hours xenon-arc exposure but after exposure to humidity and elevated temperature (90°C, 95% RH) surface roughening of the material occurs as is shown in figure 3. In an accelerated outdoor exposure test, the addition of water and more severe temperature cycling as in a F'resnel-reflecting concentrator device with water spray leads to severe surface crazing. This exposure is estimated to be equivalent to two years natural exposure in Arizona. An interesting comparison between samples exposed in the outdoor accelerated test described above and in a miniature collector is shown in figure 4. Although it is diffi- cult to quantitate effective exposure times, the times are thought to be similar. 426 A comparison of changes in the integrated solar transmit tance of PET with surface crazing shows a direct relation between transmittance and crazing. Samples exposed in Arizona and to xenon-arc and elevated temperatures had little change in transmittance. Those exposed in Florida and Maryland showed a decreasing transmittance as function of exposure time. From visual examination, one can determine that the mechanical proper- ties of PET change markedly during the course of the natural weathering as well as in the elevated temperature and relative humidity accelerated test. By examining cross sections of breaks formed in tensile tests at room temperature, changes in the ductility of PET can be observed. The change from ductile to brittle, as determined in this way, occurs „ quickly as a result of natural weathering. After eighty 17,000 kJ/m days of Florida exposure, there is only a hint of brittleness. After one hundred and sixty 17,000 kJ/m days the break is completely brittle. This is shown in figure 5. Results of the surface examination of polycarbonate are similar to those of PET. All surfaces which were exposed in service environments even- tually crazed. The phenomena was less severe as a result of Arizona exposure than Florida or Maryland. Materials exposed at constant tem- perature to xenon-arc or to elevated temperatures did not craze. As for PET, transmittance changes were greater in materials which crazed than in those which did not. 3. Discussion Early detection of surface deterioration has been shown to be useful in detecting degradation and in characterizing mechanisms of degradation. The importance of water in addition to ultra-violet radiation in the degradation process in both PET and polycarbonate is clearly demonstra- ted, for example. It is known that, in PET, hydrolytic degradation of the ester linkage at temperatures higher than those used in these tests depends on the initial concentration of carboxyl end groups [4]. Since one of the products of photoxidation is carboxyl end groups [5] , it seems possible that at lower temperatures there could also be a syner- gistic effect involving oxygen, ultraviolet radiation, and water. This work indicates that this is the case. For polycarbonate, surface cracking occured when the material was subjected to natural weathering stresses but not in the accelerated testing environments used in the glazing project. Others have noted the importance of temperature cycling in surface crazing [6,7]. Insufficient exposures were done to delineate the effects of temperature cycling and water in the degrada- tion of polycarbonate in this project. It is likely that both stresses affect the rate of crazing. 4. Conclusions In conclusion, microscopy can be used to detect incipient degradation and is related to performance properties in some solar glazing mater- 427 ials. It is possible to obtain information on the importance of various stresses by comparing degradation from different testing environments. Furthermore, important information of the mechanisms of degradation can be obtained by studying micro-structural properties. This research demonstrates the feasibility of using incipient degrada- tion as a tool to aid service life prediction although further research is needed to fully utilize these procedures. 5. References 1. ASTM "Recommended Practice for Developing Short-Term Accelerated Tests for Prediction of the Service Life of Building Components and Materials," ASTM E632-78, Annual Book of ASTM Standards, Part 18 Philadelphia, PA, 1980. 2. Clark, E. J., Roberts, W. E. , Grimes, J. W. and Embree E. J., "Solar Energy Systems - Standards for Cover Plates for Flat Plate Solar Collectors," NBSTN 1132, 1980. 3. Waksman, D. , Streed, E. , and Seiler, J., "NBS Solar Collector Dura- bility/Reliability Test Program Plan," NBSTN 1136 1981. A. Zimmerman, H. and Kim, N. T., "Investigations on Thermal and Hydrol- ytic Degradation of Poly (ethylene Terephthalate) ," Polym. Eng. and Sci. , Vol. 20., No. 10., 1980, p. 680. 5. Day, M. and Wiles, D. M. , "Photochemical Degradation of Polyethy- lene Terephthalate). III. Determination of Decomposition Products and Reaction Mechanism, " J. App. Polym. Sci, , Vol. 16, 1972, p. 203. 6. Blaga, A., Yamasaki, R. S., "Surface Microcracking Induces by Weathering of Polycarbonate Sheet," J. Mat. Science , Vol. 11, 1976, p. 1513. 7. Factor A. and Chu, M. L., "The Role of Oxygen in Photo-Aging of Bisphenol-A Polycarbonate," Polym. Degradation and Stability , Vol. 2, 1980, p. 203. 428 Figure la. Figure lb. 429 Figure lc Figure Id. Figure 1. Surface deterioration of PET from miniature collectors after 80 (a) and 160 (b) 17,000 kJ/m days of Florida exposure ? and after 80 (c) and 160 (d) 17,000 kJ/m days of Arizona exposure. 450x. 430 Figure 2a, Figure 2b. Figure 2. Surface deterioration of PET after two hundred eighty-five 17,000 kJ/m days exposure as part of a stagnating solar collector (a) and after two hundred forty 17,000 kJ/m days miniature collector exposure in Florida (b) . 450x. 431 Figure 3. Surface deterioration of PET after 500 hours exposure at 92°C and 95 percent relative humidity. 4A00x. 432 Figure 4a, Figure 4. Surface deterioration after exposure (equivalent to two years natural exposure) in a Frensnel- reflecting concentrator with water spray in Arizona (a) and after four hundred eighty 17,000 kJ/m days Arizona miniature collector exposure (b). 450x. 433 Figure 5a, >*m**"%m'" *" Figure 5b. Figure 5. Cross section of PET breaks formed in tensile at room temperature after eighty (a) and one hundred sixty (b) 17,000 kJ/m days of Florida miniature collector exposure. Top of picture corresponds to exposed surface. 434 APPARATUS FOR ACCELERATED WEATHERING OF BUILDING MATERIALS AND COMPONENTS Tore Gjelsvik Norwegian Building Research Institute Trondheim Branch, N-703 1 ! Trondheim - NTH Abstract: Natural weathering of building materials and components is a result of the combined influence of a number of ageing factors. Climate differ greatly, however, from place to place and from one year to an- other. Durability will consequently also be dependent on time and place. Accelerated artificial weathering is a still more complex matter. When the individual ageing factors act in combination, the result can be an- other than when they act alone. Size and shape of samples are also important. It is of great value to take care to accelerate the individ- ual ageing factors to the same extent, but this should not be overempha- sized as it is the final result that counts. A special apparatus for accelerated weathering of building materials and components has been designed and built at the Trondheim Branch of the Norwegian Building Research Institute. This equipment is unique as it can take samples up to 1 m^; a striking contrast to those commercially available. The results are promising when compared with practical ex- periences in Scandinavia. A modified version for samples up to k m was completed in 1 979 and is now running satisfactorily. Key words: Accelerated weathering; aging factors; building materials; components; natural weathering; testing equipment. There seems to be full agreement about the factors that influence the natural weathering or aging of building materials and components. This is a matter that has been dealt with several times before and where publications are readily available. For this reason I shall not go into details here, but restrict myself to a brief listing of the most important factors. §2i§:^_^§:§i?:ti2^ results in a number of chemical reactions, primary as well as secondary, especially when combined with humidity. 5i§^_^?™E?£§:tHE!i§ are a l so first of all the result of solar radiation. L2^JI?2E?£§:tures are especially occurring in winter nights, and here the number of freezing point transitions is important. 435 Temper ature_changes , from high to low and back, are resulting in dilat- ation between different materials. This is especially the case with composite materials and the more complex components. Changes_in_barometric_pressure can be of importance for porous materials and components with sealed air spaces. Changes_in_relative_humidity can result in dimensional changes. Wind is first of all a mechanical factor, but is also resulting in air infiltration and is an important factor in the rain penetration mechan- isms. Rain is contributing to the degradation of materials by supplying the water needed for secondary results of solar radiation. Driving rain combined with wind can also result in leaks and a number of secondary effects. ?£2^i^S_£§:E^i2i££ are difficult to discuss, but airborne sand and ice crystals should not be overlooked. Chemicals are more easily accepted in a world with increasing pollution, but the importance of this factor is difficult to judge more precisely. The natural weathering or ageing is the result of the combined action of all the aging factors involved. To this comes the fact that climate is something which is not constant. It differs from one part of the earth to another, from place to place within the same part, from one year to another, with the orientation as well as with local factors. Natural aging is consequently not only one single thing but a number of different things. Durability of materials and components is then also dependent on time and place. When it comes to accelerated artificial weathering the picture is still more complex, and this is a point where it is not easy to reach any kind of agreement. Basically, all aging factors should be accelerated to the same extent. The point is, however, that the result can be an- other when the factors act in combination than when each of them act alone. Sometimes really unexpected combination effects can occur. It is also a question of size and shape of the test samples. Small pieces of homogeneous materials can give one result, while more complex and large size components can give another. Due to the combination effects, the endevours to accelerate the individual aging factors to the same extent should not be overemphasized. It is the final result that counts, and the most important point is to compose an aging cycle and to select test samples giving a sensible correlation with practice. It should be pointed out, however, that exact correlations can never be found, and that it only is the question of a rough correlation with the average natural weathering in a certain area over a longer period 436 of time. In the early 1960-s a special apparatus for accelerated weathering of building materials and components was designed and built at the Trondheim Branch of the Norwegian Building Research Institute. One of the main points in this connection was the desire to be able to test much larger samples than what was possible in the commercially avail- able Atlas Weather-Ometers . A second main point was to have a suffi- ciently rapid ageing cycle. The final result of the discussions was the apparatus completed in 1 96 5 - The basic features will appear from Figure 1 . The apparatus consists of a circular central chamber (E) and three fixed boxes for the climatic strains (A, B and C). The central chamber (E) has four test frames of about one by one meter each. On Fig. 1 these are marked 1, 2, 3 and k. Usually the test frames are shaped as niches with a closed bottom isulated behind with 50 mm mineral wool. In the test frames or openings the materials and components to be tested are mounted on brackets with hooks, clips, pins or similar. The central chamber is most of the time fixed in a resting position, but is with certain intervals of time moved a quarter turn (90°). The test openings and the samples are in that way subjected to the following test cycle: A. Radiation from 8 sun light lamps of the type Osram Ultra Vitalux GUR 53 300 W or similar and simultaneous heating to an elevated temperature . B. Wetting with a spray of demineralized water, about 15 liters per square meter and hour at a temperature of 296 t 2 K ( 23 - 2 °C ) . C. Cooling and freezing to a temperature of about 253 K (-20 °C). D. Thawing at room temperature 296 + 2 K (23 ± 2 °C) and kO t 10 % RH, with possibilities for inspecting and changing the samples without stopping the test apparatus. Usually the testing time in each position is set at one hour. This gives a maximum black panel temperature of 3^8 t 5 K ( 75 - 5 °C) in position A, and a total time of four hours for each full cycle. The times given can be changed, as the resting time in each position can be adjusted from ten minutes to six hours, but this is normally not done. The size of the test specimens can be selected freely within a maximum size of about 0,9 x 0,9 m. 437 In addition to the special apparatus for accelerated weathering, it is neccessary to have other types of equipment to check the interesting properties of the materials and components to be tested, depending upon what is under test and which properties are interesting. The weather resistance is found from the detectable changes in the visual character- istics and measurable properties. This has to be described in different ways according to the types of materials and components and their intended functions in the exterior envelope. Usually it is described as the changes occuring during the exposure or lack of changes. The possible changes can be fissures, cracks or other types of visual damage, loss of gloss or colour changes, in weight and dimensions, changes in the mechanical properties etc. Unless otherwise agreed, the manufacturers of the materials or compo- nents supply the quantities neccessary for the tests, together with a sufficiently detailed description of the products and their intended use. The samples should be sufficiently large to be representative for the products. The test samples are prepared as specified and agreed. The details of the preparation has to be considered separately in every single case. The test specimens are mounted in the apparatus as prescribed. The exposure time will depend upon the type of material or component to be tested. Testing periods of 56 and 112 days are normal, but the testing time can also be as short as '\h days or as long as 336 days or more. The test method can basically be used for all types of materials and components in the exterior envelope of a bulding, first of all in exterior walls, but also to some extent in roofs etc. The test apparat- us and the test cycle used is always the same, but the duration of the exposure and details in the test procedure have to be considered sepa- rately in every single case. As it will have appeared from the earlier considerations, all kinds of accelerated aging tests involve a certain degree of uncertainty, and the results have to be considered with care. The test method and apparatus described have, however, been used for more than 15 years, and the results seem promising. The interesting point is of course the acceleration factor, the figure to be used for calculating from accelerated testing to natural aging. This factor will depend upon what kind of material or component that is tested and the actual type of outdoor climate to compare with. Usual- ly the acceleration factor is from 12 to 15 times when the normal cycle as described is used, as compared with average natural exposure in the more densely populated areas in Scandinavia. In any case, the information provided by this test method is considered to be giving very valuable indications. 438 It should be added, that the practical experiences referred to were found by combined field studies and accelerated testing of building sealants. Practical experiences are also available with building gask- ets, other types of plastics and rubber materials used in walls, roof- ing materials and a lot of different types of coatings. The apparatus has also turned out to be excellent in testing the frost resistance of porous materials such as roofing tiles. Based on the good experiences with the first apparatus of the type des- cribed, a second one was completed in 1979- A photo is shown on Figure 2. That Mark 2 is about four times the size of Mark 1 , and can take samples up to 1,5 by 2,5 m or close to k m^. On the photo the apparat- us is filled with glazed windows 0,5 by 0,5 m, 15 windows in every test opening and 60 windows alltogether. This series of tests was run to study different factory-applied paint and stain systems. The new apparatus differ from the old one at one point: On the new one the central chamber (E) is conditioned at 296 ± 2 K ( 23 1 2 °C ) and 60 % RH. In this way the painted windows have been subjected to a normal indoor climate on one side and accelerated outdoor climate on the other side. The resting time in each position was for the first series set at 3 hours. In general the test programme can be set within wide limits, and this opens a range of new posibilities. 439 o z a UJ MOVES FROM 'AND TO O ROOM TEMPERATURE 296 K (23°C±2°C) CONTRa UNIT "71 n n^ fr— H S //////////////// JV7///////////////// ?///////////////// FRONT ELEVATION SECTION Fig. 1 Principle sketch of apparatus for accelerated weathering 440 Fig. 2 View of new (1979) apparatus 441 A PRELIMINARY EVALUATION OF THE TENSILE AND ELONGATION PROPERTIES OF SINGLE-PLY SHEET ROOFING MEMBRANE MATERIALS Robert G. Mathey and Walter J. Rossiter, Jr. Center for Building Technology National Bureau of Standards Washington, DC 20234 Abstract: A summary is presented of the results of a preliminary evaluation of two performance properties, tensile strength, and ultimate elongation, of nineteen single-ply sheet roofing membrane materials. Also reported are the changes in mass and length of the membrane materials caused by exposure to heat. The nineteen materials repre- sented the general categories of single-ply sheet membranes (elastomeric , plastomeric, and modified bitumens) and were typical of those used in the United States. Membrane materials included neoprene, ethylene propylene diene terpolymer, chlorosulphonated polyethylene, polyvinyl chloride, chlorinated polyethylene, and modified bitumens. The membrane materials were tested in tension before and after exposure to heat, and heat followed by ultraviolet radiation from a xenon arc. Control (unexposed) specimens were tested at 70 and 0°F and the exposed specimens were tested at 0°F. Three ASTM tensile test procedures were selected to determine the tensile and elongation properties of the membrane materials. All nineteen materials were tested according to a procedure for rubber. The plastics and modified bitumens were also tested according to procedures applicable to reinforced fabrics and bituminous roofing membranes, respectively. Keywords: Elongation; exposure conditions; membrane properties; roofing membranes; single-ply roofing; tensile strength; test methods. The use of single-ply sheet roofing membranes has increased rapidly since the mid-1 970' s. Many types of single-ply sheet membrane materials are available and they fall within the general categories of elastomeric, plastomeric, and modified bitumens. Some of these sheets are reinforced with glass or synthetic fibers. Others have facings consisting of films such as polyethylene or aluminum foil. Standards are not available in the United States to assist architects, specifiers, and designers in the selection of single-ply membranes. As indicated in an earlier National Bureau of Standards publication [1], test procedures and criteria for the prediction of service lives of these membrane materials having a wide variability in properties are lacking. 442 Without criteria it is not possible to make a quantitative judgement in the selection of a single-ply roofing system. This fact is disconcerting, especially when the number of new systems entering the market is considered. A recent survey [2] estimated that single-ply roofing is expected to comprise 25 percent of the low-sloped roofing market by 1985. The survey forecast that 300 x 10^ ft^* of single-ply roofing will be installed in 1981. Reroofing now accounts for about two-thirds of total single-ply installation according to the survey. New roofing is expected to comprise about one-half of the total single- ply installation by 1985. Physical and mechanical properties of the many single-ply sheet roofing materials cover a wide range of values. No technical basis has been established for choosing one value instead of another as the minimum required for acceptable performance. Durability is an essential property for roofing materials and systems. If durability is to be included in criteria for selecting roofing materials and systems, it must be expressed in quantitative terms. Although many durability tests are described in standards and specifications for building materials, there is seldom any satisfactory way of correlating, with each other or with in-service performance, the results of laboratory tests on different materials [3]. Long experience with traditional materials exposed to different climatic conditions permits the estimation of their service lives. The lack of information about the durability of new materials limits the ability to predict their performance. It is difficult to develop accelerated aging tests for use in predicting long-term in-service performance. The degradation mechanisms of building materials are complex and seldom well understood. Tests are often conducted on configurations different from those in service. Furthermore, there are many external factors that affect performance and they are difficult to quantify [4] . The result is that many existing accelerated test procedures do not include all factors of importance and those included seldom relate quantitatively to in-service exposure. Although accelerated aging tests have shortcomings they have been used in many cases to provide data for comparing materials properties. Objectives of the Study The objectives were: to measure the tensile strength and elongation of different types of single-ply sheet materials using various test methods; to investigate the changes in these properties when tested at different temperatures, and after exposure to heat and ultraviolet radiation; and to measure the mass loss and change in length caused by exposure to heat. * Conversion factors to metric (SI) units are given at the end of the paper. 443 Types of Materials The 19 single-ply sheet roofing materials included in the study are listed in table 1 along with their thicknesses. Since some of the plastic and modified bituminous sheet materials were reinforced, test data were grouped as follows: 1. Rubbers (R) 2. Plastics (P) 3. Reinforced plastics (RP) 4. Bituminous materials (B) 5. Reinforced bituminous materials (RB) Test Methods The 19 single-ply sheet roofing materials were tested by three different test methods to determine their tensile strength and elongation at 70° and 0°F, and at 0°F after being subjected to two different exposure conditions. For one condition, rubber and plastic types of materials were exposed to 212 °F in an oven for 14 days and the bituminous materials were exposed to 167°F for 14 days. For the other exposure condition, another set of materials were heated at temperatures noted above for seven days and then exposed to 1000 hours of xenon arc radiation (ASTM D 2565-76, Type B) . The three tensile and elongation tests included ASTM D 412, "Standard Test Methods for Rubber Properties in Tension"; ASTM D 751, "Standard Methods of Testing Coated Fabrics" (procedure B, cut strip method); and ASTM D 2 523, "Standard Practice for Testing Load-Strain Properties of Roofing Membranes". The procedure for measuring elongation given in ASTM D 412 was not followed in order to conduct the tests at 0°F. It was not possible to hold and read a scale to measure elongation in the environmental chamber. Instead of using the benchmark method given in ASTM D 412, the crosshead and chart speeds of the testing machine were used to determine the elongation. The ASTM D 412 tests were conducted at a rate of elongation of 20 in./ min. The ASTM D 751 and ASTM D 2 523 tests were run at 12 in./min and 0.08 in./min respectively. All the bituminous membranes tested according to ASTM D 2523 were tested at the slowest rate of elongation even though two of the samples had values of ultimate elongation greater than 50 percent when tested at 70°F. The distance between the grips of the testing machine at the start of the tests was 2, 3 and 7 in. for methods D 412, D 751, and D 2523 respectively. Preliminary Test Results and Discussion The changes in length and mass of the sheet roofing samples after heating in an oven for 14 days are presented in table 2. The changes in length of the materials ranged from negligible to about 9 percent. The 444 unreinforced plastic materials exhibited in general a larger decrease in length in the longitudinal or machine direction of the roll than the other materials. The range in mass loss upon heating was to 3.6 percent and no trend was observed in the data. Values of percent elongation for different test methods, temperatures, and exposure conditions for the sheet roofing materials are given in table 3. Data in this table are for longitudinal or machine direction (as manufactured) orientation of the materials. Higher values of percent elongation for the plastics and reinforced plastics were obtained from the ASTM D 751 test method than for ASTM D 412. The unreinforced plastics exhibited greater elongation than the reinforced plastics. As expected, higher values of percent elongation were obtained for the rubber control samples compared to the other types of materials when tested at 70 and 0°F. A large change in percent elonga- tion compared to the control samples occurred for the rubber material samples exposed to heating plus ultraviolet radiation. The largest change in percent elongation was due to temperature (70° to 0°F), parti- cularly for sample 18 which had a ratio of elongation at 0°F compared to 70°F of 0.05 when tested by ASTM D 412. Samples 18 and 19 (modified asphalt) when tested by ASTM D 2523 also exhibited relatively low values of extensibility at 0°F. An important conclusion regarding the tensile test methods was that ASTM D 2523 was not a practical test for some of the modified bituminous materials when the rate of elongation was 0.08 in./min. Some of these tests ran for over 4 hours because of the slow rate of loading and the large extensibility of the materials. It is noted that the thicknesses of the sheet materials varied (table 1), Also, the widths of the specimens were not the same. The width of the test specimens for the D 412 test were 1/4 in. wide and the test specimens used in the D 751 and D 2523 tests were 1 in. wide. Table 4 presents the tensile stress in the longitudinal direction (as manufactured) of the sheet roofing materials for the different test methods, test temperatures, and exposure conditions. Values of stress are reported for the break loads (specimen failure) and for intermediate loads (prior to break load). Stresses were determined from the average value of the loads and the cross-sectional area of the test specimens (sheet material thickness x specimen width). In the D 412 test for corresponding conditions, values of stress at break for the unreinforced plastics were higher than those for the rubber specimens. The bituminous specimens, both reinforced and unreinforced, had the lowest stress values. A wide range of values of stress at break was observed for the reinforced plastic materials. 445 In comparing the results for the D 412 and D 751 tests for the unreinf orced plastics, little difference was found in the stress at break for corresponding test conditions. The results of stress at break measurements were variable for two of the reinforced plastics (nos. 13-RP and 16-RP) between these two test methods; while the other two reinforced plastics exhibited comparable values of stress at break for corresponding test conditions. No trend was observed in comparing values of stress at break for the D 412 and D 2523 tests on the bituminous materials. It is noted that little data were available to make this comparison. As expected, the stress at break generally increased significantly when specimens were tested at 0°F compared to tests at 70°F. In most cases, there was little change in the stress at break after the specimens were exposed to heating and ultraviolet radiation, when comparable tests were conducted at 0°F. Summary Preliminary results were presented regarding the changes in elongation and tensile stress of single-ply sheet membrane materials caused by test temperature and exposure to heat and ultraviolet radiation. Different ASTM tensile test procedures were conducted to determine their applicability to single-ply roofing membranes. Changes in length and mass of the sheet membranes due to exposure to heat were determined. The data in this preliminary paper need further analysis. A final report will include the results of the analysis and additional data from the study not presented here. Acknowledgment This study was conducted under the Tri-Service Building Materials Investigational Program and was sponsored by the Directorate of Engineering and Services, U.S. Air Force; Naval Facilities Engineering Command, U.S. Navy; and the Office of Chief of Engineers, U.S. Army. 446 References [1] Rossiter, Jr., Walter J., and Mathey, Robert G., "Elastomeric Roofing: A Survey," National Bureau of Standards, Technical Note 972, July 1978. [2] Civil Engineering, "Singly-Ply Roofing Market is Growing, According to National Survey," April 1981, p. 16. [3] Frohnsdorff, Geoffrey and Masters, Larry W., "The Meaning of Durability and Durability Prediction," Durability of Building Materials and Components, ASTM STP 691, American Society of Testing and Materials, Philadelphia, PA, 1980, pp. 17-30. [4] ASTM "Standard Recommended Practice for Developing Short-Term Accelerated Tests for Prediction of the Service Life of Building Components and Materials," ANSI/ASTM E 632-78, Annual Book of Standards, Part 18, Philadelphia, PA, 1978. Conversion Factors to Metric (SI) Units In view of the presently-accepted practice in this country in this technological area, common U.S. units of measurement have been used throughout this paper. In recognition of the position of the U.S.A. as a signatory to the General Conference on Weights and Measures, which gave official status to the metric SI system of units in 1960, we assist readers interested in making use of the coherent system of SI units by giving conversion factors applicable to U.S. units in this paper. Physical Quantity To Convert From to Multiply by Area ft 2 m2 9.29 x 10" 2 Force lbf N 4.45 Length ft m 3.05 x 10" 1 Length in. m 2 .54 x 10 -2 Length mil m 2 .54 x 10-5 Length/ time in./min m/s 4.23 x 10" 4 Pound-force/ inch 2 lbf/in. 2 Pa 6.89 x 10 3 Temperature °F °C T c = (T F - 32)/1.8 447 Table 1. Single-ply Sheet Roofing Material Test Samples Thickness b Sample No. a Type of Material mils Chlorosulphonated polyethylene/asbestos backed 34 Chlorosulphonated polyethylene/asbestos backed 34 EPDM (ethylene propylene diene terpolymer) 41 EPDM 64 EPDM 56 Neoprene (black) 63 Neoprene (gray) 62 Neoprene (black) 57 Polyvinyl chloride (PVC) 47 Polyvinyl chloride 48 Polyvinyl chloride 33 Polyvinyl chloride 44 Polyester reinforced chlorinated polyethylene (CPE) 34 Non-woven glass reinforced PVC 47 Non-woven glass reinforced PVC 47 Polyester reinforced PVC 51 Modified asphalt/polyethylene sheet on surface 63 c Modified asphalt/polyethylene sheet embedded 156 19-RB Modified asphalt/reinforced with polyester mat 168 a The letter in the sample number designation indicates that the material is a rubber (R), plastic (P), reinforced plastic (RP), bitumen (B), or reinforced bitumen (RB) . " Thickness based on average value of six measurements on each of four samples . c Paper backing removed before thickness measurements. 1- -R 2- -R 3- -R 4- -R 5- -R 6- -R 7- -R 8- -R 9- -P 10- -P 11- -P 12- -P 13- -RP 14- -RP 15- -RP 16- -RP 17- -B 18- -B 448 Table 2 . Changes in Mass and Length of Sheet Roofing Samples Exposed at Elevated Temp eratur ess for 14 Days Exposure Change in Length b Temperature (°F) Mass Loss a (percent) (percent) Sample No. LC jd 1-R 212 0.0 e 0.2 e 0.2 e 2-R 212 0.1 e 0.4e 0.0 e 3-R 212 1.2® 0.2* 1.2 e 4-R 212 2.3« 0.9e l.Oe 5-R 212 1.4 0.5 0.5 6-R 212 3.6 e 1.8 e 2.8 e 7-R 212 1.2 e 0.8 e 0.6 e 8-R 212 l.l e 0.8 e 0.4 e 9-P 212 1.8 4.3 0.0 10-P 212 1.3 2.3 0.3 11 -P 212 2.9 6.6 -1.1* 12 -P 212 2.4 8.9 -0.1 13-RP 212 0.7 0.4 3.6 14-RP 212 1.5 0.3 0.2 15-RP 212 1.6 0.4 0.2 16-RP 212 2.6 3.1 0.6 17-B 167 0.6 0.7 0.8 18-B 167 0.1 -0.8 f 0.1 19-RB 167 0.2 0.2 0.4 a Unless otherwide indicated, value represents the average of four specimens. " Unless otherwise indicated, value' represents the average of four specimens (three measurements per specimen). c Longitudinal or machine direction of roll. " Transverse or cross machine direction of roll. e Average of two specimens. * (-) sign indicates an increase in length. 449 o ~H w V4-4 o o h .— -v I u. o o " *H o 4-1 u to 4-1 01 c m o U to u. co o o J o a) a a) 3' cab l-l O o J O 4J TJ Cat: M o J O tit) b u to o (coo -J r-. n i> -o u, COO M J r-. oo rs vo ffv © © © d 1 d oooo ,*©©,* O ^ .-I ^* .-I i-l O — i o o o Momenr*yDr^i-ifn «*> -1 ■* -1 -4 -» oooo On ^O co O O -» oooo O O T-l OOOO OOOO N N ^ ^ ft 0\ Ol (ft ST ON *40 M3 en en CN CNI CM CM CM CM -4—4 i i : NlBOOsn cm r» —i en co e» ui CM en _i •» i-4 -J oi ai on *© i I •-4cM©ON4nineM06 en -x) o m fl vO O PI « NNH 14-1 u-i ^ y-i r- . © O © r~ M5 N O h n CM I O -i I ^ -I n * m a M (»i -I in on en ^h m ,-t m> vO m N N rf H SO CO r^ I I I I I I I I ■ ■■■ i i i i N «C N (M en en CM en —i en fi is- n -^ ^ N co CI cm cm -j ^ x: o -i o -i 41 4J H 0) r^rsenOf-ONCOcM OBCOOO'JflN -»«sr-»cni/"ieni-ssr "SIMM CMCMCMCMCMCMCMCM t-li-^-sii-l^^Ji-Si-1 •»-»-»>a->9-»i4-- i-4 <* >^ ON N pi o> n io tn m -sT i* no en .» -a- 0\ 00 un no CM —I en -4 /n. CM CM CM CM i-l — 1 i-4 ,-1 St •* -tT •» o a a q a- a. ex, o- CM CM CM CM CM CM i-t ,_! |H rH i-l — 1 -3- -J sj- -st -st ■» a o o o a a aaaa aaaa 04 On 0* On i i i i en -a- m so en en CM CM in m CM CM co oo I I r- 00 00 c 01 * +J £ 41 CO u i-l c 01 a CM < O 41 01 00 00 a a u u 01 4) > > to a 10 01 01 u X -C 41 i-l 4-1 •H a a U-l 4J 4-4 a c 1) 01 a 1 a to ■H 01 01 u l- 1-4 • u frfr ■a o eg 14-1 kl kl o c 01 01 01 3 3 Jt u i-l iH to eg a) 01 ki > > . k4 O •a 1-1 41 01 co O tt) o 4-1 -H 1-4 4J —i kl CO eg ki 01 U o ^ O 4-1 •H —4 a iH to 11 41 ki a u a •H iH J3 oo 41 01 O tO -rt to a 4J O Ji! •H i-l rt e ki OS CO 41 k4 3 01 11 I-l 01 iH ■H iH li 4J ^4 c e o c to a a o M tx. to -O o T) 41 450 u T3 MOO J o co m m <• O O O O 00 cjn © ■» © © M O I tO o o o o -J O — i -i M © O M O r-l <-l M .-I O «-H in-CT-S'cM-*©'*® 00 00 Os in co \0 *cf rs ao oo -» -» H H H PI f-l CM eo oo co m m Ps m oo oo cm C* CM >* vO ps co co *o 'j m ia n m cp» oo oo o —< m co r--3-cMM n s n m vO o« r. o CO CO -3 -3 1 \D CM CO in On CO II 1 1 1 1 1 Ps ,__| II 1 1 1 1 1 U1 o CM CO oo oo oo so >os co m -3 cmoooo in-3'CM s O,-ltncOM rtrtNNNNNN CO CO -3- ~» oo oo oo oo oo CM M PS sO ST 00 CM rH oo ps oo mo .h SO O CM St "* *-* SO \0 co co \0 t— I —i © os o uooen-3- 0> H O 00 r. m -J N OO CO N GO C*» O* O OV COCO-*-* in CM CO -» o sO tc Cr« o> 00 co o r-< co co m m N (O *0 \C ■—< © rs \© -cy CO 00 r> o\ Pi N O 00 CO OS *CT -CT CO CM OS CM *C O PI CO -J"?-*-* oococouo CO o o -» vO CO u -a OJ 01 e co in cp> 00 MOO CM M 1 1 J o ms I i aocomr-.ino>om <• m r*. ^-i «** CM CM * ■^ ^ H CO CM CM ft CM 00 00 00 00 CO CO rO H N N hi s© m cn CM oo H >0 N O O* «* H H (M OJ •o b. cu CD 4-1 co CM 00 c o O SO on t-l -J ps ~3- CO CMCMCMCMCMCMCMCM QCipCiPCiQCiOerOCiOCiOEi I I I I I I I I CM CM CM CM -a- «» -» -sr a a a a Cu Cu CX- O- CM CM CM CM CM CM lH .-I r-t M — I ^ -3> -» •» -» •» -a- a a a a a a Cm Cu Ctt Cu cc ds Da os pa pa I I I I II ft-} n vo ps oo S O 00 00 CO -4D 00 o ■3 >T CO vO PI N ^ S -a* \£> as oo m\os •» CMCMCMCM in M M CO ! I I I £ I I mmtom m m m m r«» r-* r* r- r^r^r^r*. a Q O Q Q Q o o O- 0-i Oh P- I I I I Ot O <-l CM i i i i co ^3 m so CO CO CM CM m m CM CM pa co I I r~ oo 3 i OO 01 3 c OJ E* % o O 4J is. i! 4-1 OJ » CO c 2 CO • 2 O 4J U M • o •O -H CM OJ og T3 4-1 O M-t « r-t • • -H B 4J 73 a -h o) CO OJ 3 00 a g s cm a •3 — t-i OJ Kh II OJ 4J CO > c c S CM < O H OJ OJ 01) or a o V4 i- OJ OJ > > c OJ c OJ If 01 CO •H OJ CJ O t- tJ t-l & o. 41 T3 CO tw u O i-H C OJ CJ OJ 3 j»: ti M M co CO CO OJ L- > > , O TJ M ■a T) CO o CO OJ 11 O 4-1 M 4J 4J -H tw CO CO t- OJ o o J* 4J xH M CO CO •o T3 OJ u s c C 1-1 a. •H H J2 •o 9 OJ OJ CO ■H CO CO 4J o J-§$ During a creep test under constant load, the stress a(t) varies during testing due to the variation in section which is particularly great for bitumen- polymer binding materials. With the knowledge of the load applied during testing, the apparent compliance Da can be easily determined : Da = £ ^ ' cr ' initial stress a Q o If we suppose that the binding material is practically incompressible ■ which has been experimentally verified, the true compliance can be linked to the apparent compliance by the relation : D(t > " 1 + e(t) " a Q (i+e(t)) The curve D(t) has been drawn for binding material I (fig. 2) and coincides with the function e |r3 traced point by point experimentally by measuring the true section of the bitumen test-piece at each moment. Other binding materials have also been subjected to rheologi- cal studies (2). By analysing a creep curve, it is possible to determine : . an initial quasi-elastic deformation Do, to a greater or lesser extent according to the nature of the binding material, . a delayed elastic deformation from to 48 hours, 453 . a levelling of creep to a gentle slope after 48 hours, characteris- tic of stability which can be measured by the value of the compliance De. After creep testing for a given time t, suppression of the load allows observation of recovery and evaluation of the elastic character of the bitumen- polymer binding material (fig. 3). If e r is the residual deformation after a time t for recovery, and ef is the deformation attained after the same length t of creep time preceding recovery, then the residual deformation can be measured by : R t (*) = 100 ft The expression : T R (*) = 100 (1 - |T) is called the recovery rate. An elastic type binding material gives a value of T R approaching 100% and a plastic type binding material approaches 0%. 3. STUDY OF THE THERMAL AGING PROCESSES This accelerated degradation is performed in a ventilated oven at 70°C. Samples taken after 1 week, 2 weeks, 1, 2, 3, 4 and 6 months have enabled observation of the variations in the elastic properties and their association with changes in internal structure. 1) The chemical composition of the binding material evolves. Chemical analysis of the bitumen shows an increase in the asphaltene content as a function of the duration of weathering. Using GPC chroma- tography, progressive deterioration of the SBS polymer due to splitting of chains is revealed (fig. 4). This evolution is confirmed by the measurement of the intrinsic viscosity (n) on the polymer extracted from the binding material. Moreover, infrared analysis shows that the double bonds of the butadiene molecular chains decrease and that oxided products appear. 2) The rheological properties evolves. The creep curves (fig. 5) show : . a progressive decrease of the slope to the origin (dD/dt = =) signi- fying the hardening of the binding materials ; the viscosity n~ increases, 454 . a decrease in the Do value compared with that measured on the new binding material, to a value of almost zero for the sample undergoing aging for 6 months in the oven, . disappearance of the equilibrium value of the compliance De, as a consequence of the cuts in the SBS chains, . progressive disappearance of the elastic character shown up in the recovery curves (fig. 6) which reveal an increasingly slow kinetic of recovery as the duration of degradation progresses. 4. NATURAL WEATHERING The evolution of the physico-chemical properties of the SBS bitumen is also studied using samples taken from roofing or in the exposure sta- tion of the CSTB in Grenoble. These samples are analysed as previously described by GPC or by creep testing. In certain cases, depending on the duration and the site of exposure, the appearance of smaller molecular weight beside those of the initial polymer is observed. Elasticity measurements also show that recovery decreases with aging more or less quickly according to the quality of the material and the climatic conditions to which it has been subjected. It would appear then that the phenomena of chain scission and conse- quently of loss of elasticity produce an evolution in the material which is similar for natural weathering and for artificial aging. This similarity is necessary to be able to obtain a correlation between the two types of deterioration. 5. CONCLUSIONS The present study is concerned with the identification methods and the follow-up of aging in SBS bitumen binding materials whose properties determine the quality of the roofing sheets. The identification can be made by chemical analysis of binding material, splitting up into asphaltenes, malthenes and SBS, as well as by GPC chromatography. The rheological properties can be evaluated by means of creep testing. It is shown that these binding materials display delayed elastic behaviour when new. During artificial degradation in an oven thermo-oxidation is accele- rated, producing scissions of SBS chains, and slowly destroying the initial network of the polymer in the bitumen. This destruction allied 455 with the hardening of the bitumen involves a progressive change for the binding material, from elastic behaviour towards plastic behaviour. Tests under radiation from the weatherometer showed a very similar evolution of the binder to that observed in oven together with the formation of a superficial deteriorated, hard layer. According to our first results, natural weathering appears to act in the same way on the binding material as aging in a ventilated oven, but with a kinetic rate dependent on the quality of the material and the climatic conditions. 6. BIBLIOGRAPHY 1. J.Ch. MARECHAL, "Extraction et dosage du polymere SBS dans les bitumes", Bulletin Liaison Laboratoire des Ponts et Chaussees, 103, Sept. Oct. 1979, p. Ill 2. J.Ch. MARECHAL, "Les bitumes - elastomeres SBS - Caracterisation et degradation des masses d'enrobage", Cahiers du CSTB 1659, Juillet-AoOt 1980 456 WEIG HT . M I STRAIN TRANSDUCER STEEL WIRE 15/100 m m SAMPLE OF ASPHALT . SBS CREEP APPARATUS FOR DETERMINATION OF ELASTIC RECOVERY Fi} : 1 457 458 < > o 5 LU cr I O < co 459 RI DETECTOR 8X POLYMER SBS GPC ANALYSIS OF BINDER N°3 CONDITIONS : fi STYRAGEL COLUMN SOLVENT THF ELUTION 2.5 ml/mn INJECTION 1000 Mg 10 3 + 10 4 + 10 5 A REFERENCE 2 MONTHS 4 MONTHS 6 MON THS MATERIAL AGED +++++++ AGED AGED Fio : * 460 * A M ** 461 W (J * > LU > O O 462 FATIGUE BEHAVIOR OF ASPHALTIC ROOFING MEMBRANES Peter E. Nelson, P.E. Simpson, Gumperty & Heger, Inc. Cambridge, Massachusetts Abstract: The fatigue behavior of three types of asphaltic roofing membranes are presented from research on static and cyclic fatigue tests. The roofing membranes were constructed with plies of either organic, asbestos or glass fiber roofing felts, oriented in the cross machine direction. The performance of these three roofing membranes was determined over a temperature range from 22 °C (72°F) to -29 °C (20°F) by monitoring the magnitude of the load, the duration of the test, and the strain to failure. A direct comparison is made between the dynamic (cyclic) fatigue and creep rupture strengths (static fatigue) tests. The time to failure for the creep rupture test is approximately the same as the failure time for the dynamic fatigue mode. The performance profile of a roofing mem- brane's failure from cyclic fatigue is governed by the magnitude and duration of the loading and not the number of times the loading is applied. The effect from creep loading is more important in defining behavior of a roof than dynamic fatigue loading. Asbestos 4-ply roofing membranes have substantially less strain capacity than organic or glass fiber roofing membranes. Organic roofing mem- branes have less capacity than the glass fiber roofing membranes tested. Analysis shows that if these 3 roofing membranes were all subjected to a temperature drop from*23°C (73°F) to -34°C (-30°F), the roofing mem- branes would use a different portion of their ultimate strain capacity; asbestos membranes would use 19.3%, organic membranes would use 7.5%, and glass fiber membranes would use only 4.0% of their strain capacity. 463 ROOFING HISTORY SURVEY D. P. Van Court Western Electric Co. 222 Broadway, New York, NY 10038 Abstract: Historical data covering 5,711,000 square meters of original and major maintenance roofing work on 3,263,000 square meters of roof deck were gathered and evaluated subjectively and statistically. The data base covers 74 Western Electric locations throughout the United States with most data from manufacturing and warehousing operations. The data did not support the commonly-assumed superiority of coal tar pitch (CTP) built-up roofing systems (BURs) over asphalt BURs for recent construction although the data did show that CTP BURs constructed prior to about 1962 were definitely more durable than newer CTP BURs. Poured gypsum roof decks with no top insulation and board-insulated, roll-formed metal decks were found to be equally acceptable as substrates for BURs but use of top-insulated cementitious deck systems resulted in more rapid BUR failure. Major maintenance work was found to be necessary sooner than expected. Key words: Asphalt roofing; built-up roofing; coal tar pitch roofing; insulated metal roof decks; poured gypsum roof decks; roof decks. Western Electric now has In excess of 3.7 million square meters of roofing of various types and vintages distributed throughout the United States. There are a number of different built-up roofing systems, structural decks, and maintenance procedures. Over the past 15 years, Corporate Engineering has examined all of the major roofs in Western Electric and prepared maintenance recommendations. The data for each individual location have been studied intensively from the point of view of that location's needs, but data from different locations had not previously been compared. In 1979, a review of much of the work done indicated that the data, taken as a whole, might provide useful information for the future guidance of not only the individual plants but also for design, of new facilities. It was believed that this project would yield useful information on the relative durability of the various types of built-up roofing (BUR) being used, the suitability of the various types of decks being selected by the structural designers, and the durability and relative effectiveness of the various types of major maintenance procedures being employed . 464 In prosecuting this study, there were two primary goals. The first was to assemble all of the available data on any given location at one file point. This information was scattered throughout various files within Corporate Engineering and at various locations. Much of the information existed only in the memory of individuals in location organizations. The second was to organize the data so that it could be used to design maintenance programs applicable to all the locations generally instead of having to design each program separately on a location-by-location basis. The data could also be used to assist designers in selecting the most appropriate type of deck and BUR system for each project as it came up. DISCUSSION Roofing Until about 1960, the preferred built up roofing system (BUR) was a coal tar pitch (CTP)system of four or more shingled plies of felt laid directly on the deck. Until that time, the only asphalt BURs used were smooth systems used on slopes in excess of about 1 In 4. About 1962, the Company changed over to a CTP system which included an asphalt base sheet fastened to the deck and three plies of CTP saturated felts set in and flood coated with additional CTP. About the same time, spot shortages of CTP began to be felt in some areas of the country and, in other areas of the country, a penalty of 4 to 16% of the in-place-cost was being incurred when CTP was specified in preference to asphalt. At that time, asphalt systems of a base sheet plus three plies of organic felt set in and flooded with steep asphalt came into use. Traditionally, Western Electric used only aggregate-coated roofs on slopes of less than 1 in 4 except at those locations where it was felt necessary to have to shovel steam power station fly ash from roofs. In the 1970s, problems experienced with organic felt were counteracted first by specifying an asbestos felt rather than organic felt and later by specifying glass felts. The use of asbestos felts has been discontinued primarily because of the emotional impact of any container bearing the word "asbestos" on the occupants of the location, even though the product itself may be inherently safe. Deck Systems Before the mid-1920' s, Western Electric roofs were either book tile laid on structural T's or wooden decks laid on a structural framework. The book tile system was used wherever flat roofs were needed. In areas where extra strength was needed, book tile was replaced by precast concrete planks. Most of the buildings put up during the late 1920s were constructed with either gypsum or concrete plank in place of book tile. This allowed the supporting T members to be spaced further apart. Wherever these decks required extra insulation or needed to be sloped to facilitate drainage, a lightweight concrete fill was poured 465 directly on top of the deck system. Some of the concrete decks had extra insulation applied on top of the deck in the form of a board insulation. The earliest insulated decks of this type used organic fiber boards but, for most of the 1940s and 1950s, the preferred insulation was cellular glass. In the late 1950s, glass fiber insulation came into use and was used nearly interchangeably with asphalt-impregnated perlite board. In the early 1950s, what is now the most common deck type, a poured gypsum deck came into common use. This consists of a mineral fiber or glass fiber form board laid on lightweight T's and covered over with a layer of poured gypsum which is approximately 50 mm thick. The gypsum has a very light reinforcing mesh in it. In the early 1950s, certain small areas of roof were made with metal decks of either the corrugated or rectangular roll-formed shapes. In the early 1960s, metal decks came into wider use, particularly when it appeared that the roof deck work would be done during the winter. As indicated above, the board insulation over a metal deck was usually either a glass fiber or an asphalt-impregnated perlite type. Within the past few years, WE has begun to use composite insulation consisting of factory-laminated layers of urethane foam and asphalt-impregnated perlite. The earliest job done with this type of insulation soon displayed excessive blistering — Company experience with this type of roof has paralleled that of the entire industry. Current recommendations are that this board should be used wherever the extra R value of its insulation is required but that it be placed with the foam side down whereever fire insurance regulations permit. If this is not feasible, the board should be nailed to the deck and second layer of asphalt impregnated board adhered to it or a dry base sheet nailed to the insulation. In the second case, the insulation is laid loose and the base sheet laid out and nailed to the deck through the insulation. SAMPLE SELECTION In order to simplify data collection, the data base was limited to Western Electric-owned locations having floor areas of over 9,290 gross square meters. Locations with less floor area, leased locations, and locations not currently in use were omitted. The final data base covered some 3,263,000 square meters of roof at 74 different company locations. No. of Roofing Size Range, Construction Building Class Locations 1000s of Square Meters Date Range Factories 21 13 to 216 1955 to 1979 Major Warehouses 7 36 to 82 1973 to 1976 Minor Warehouses and ) Remanufacturing Shops) 37 6 to 66 1921 to 1975 466 No. of Roofing Size Range, Construction Building Class Locations 1000s of Square Meters Date Range Offices, Laboratories, etc. 9 3 to 18 1961 to 1979 It was not possible to use true "unit roofs" as units of data in this study. A "unit roof" is an individual roof area bounded on all four sides by higher or lower walls, flashed expansion joints, parapets or gravel stops. The "data units" used include one or more "unit roofs" at a given location constructed at about the same time having: the same type deck; the same type of built-up roofing; the same repair and maintenance chronologies ; and the same general occupancy class beneath the roof. There are minor inaccuracies in the data because non-matching areas which total less than 20% of the area of the selected data unit have been ignored. The entry for poured gypsum roof decks, for example, includes a number of small penthouse roofs which, in fact, have structural roll-formed metal decks. Where roof galleries constitute a major portion of the roof scape on a building, they have been broken out as a separate data unit. The data units include between one and thirty or more unit roofs. Even with this grouping, the 74 locations included in this study generated 275 data units covering 5,711,000 square meters of original and major maintenance work. CONCLUSIONS Roofing Coal tar pitch roofs constructed after about 1960 are inferior to those constructed before 1960 in terms of life to any form of maintenance. It is valid to compare coal tar pitch BURs made after 1962 with asphalt BURs because there is no statistical difference between the average ages of these two groups. Although more coal tar pitch roofs were recoated than asphalt roofs, recoating was done sooner. Similarly, although fewer new coal tar pitch roofs go bad and the areas are less than those in the asphalt data base, the ones that go bad do so sooner. Thus, in today's roofing market, it appears that coal tar pitch roofs are not worth a premium for new work, even though good ones may last longer than asphalt BURs. This information and maintenance material stocking problems indicate that it is not desirable to replace coal tar pitch roofs in kind if more than 50% of the total area of roof at a location requires replacement, instead use asphalt systems. If less than 50% requires replacement, use a good coal tar pitch system to make sure that local maintenance men use the correct materials in maintaining both old and new roof systems. 467 Deck Avoid designs with board insulation over cementitious decks because of the extremely high rates of failure and the early average age at which they fail. Comparing the two most favored decks, insulated steel and poured gypsum, there are more failures of roofing on insulated metal decks, in terms of number of units, than on poured gypsum decks but the failed area over gypsum is larger. There is no clear superiority for one type of deck over the other. Therefore, if the insulation inherent in a poured gypsum roof deck is adequate without additional insulation on top, a poured gypsum roof deck system is neither superior to nor inferior to a board insulated metal deck system. If extra insulation is required, an insulated metal deck system is superior to any system which puts board insulation on top of gypsum or any other type of cementitious deck. Maintenance and Inspection Planning Exclude insulated hot caps from any maintenance program because of the high incidence of early failure. Use uninsulated hot caps only where deck constructions are structurally adequate to support the additional weight of multiple hot caps and where roof traffic is nil. In addition, blister patching should precede hot capping to insure that internal paths are not created for surface water to reach deck structural members. The age to failure and age to resurfacing data indicate that critical ages for coal tar pitch BURs are seven years and for asphalt BURs are ten years. Programs of comprehensive annual inspections should be formalized so that changes in condition at and after those ages can be detected promptly and corrective work scheduled on an economically sound basis before the existing membranes must be removed and replaced completely at great expense. Long range maintenance budgets should include sufficient money for rehabilitation of ten percent of the unit roofs each year beginning at age ten years for coal tar pitch BURs and age twelve years for asphalt BURs. The actual observations made during annual inspections may indicate that initial major maintenance may be deferred, but the funds should be available if needed. 468 AUTHOR INDEX AND CROSS-REFERENCE TO CONFERENCE PROGRAM The program numbers refer to number of papers in the conference program Name Program No. Page No Ashton, H. 1 28 Berman, E. 6 1 Bleile, H. 32,53 327,336 Brookes, A. 13 412 Brown, P. 42 93 Browne, R. 22 210 Christ, C. 27 313 Clifton, J. 9 274 Cope, R. 17 387 DeGroot, R. 10 288 Derbyshire, H. 28 279 Dillon, R. 51 8 Eby, L. 44 115 Eurin, P. 12 354 Fagerlund, G. 37 76 Fukushima, T. 14 367 Gauri, K. • 41 88 Gjelsvik, T. 30 435 Gorsheimer, J. 42 93 Gwinn, J. 41 88 Hawkins, L. 26 346 Haynie, F. 3,43 39,95 Howe, R. 35 136 Hudec, P. 19 126 Ishizuka, Y. 31 69 Kantz, M. 42 93 Kami, J. 4 188 Kataria, S. 23 173 Koontz, M. 3 39 Kwasny, R. 15 378 Litherland, K. 36 225 469 AUTHOR INDEX AND CROSS-REFERENCE TO CONFERENCE PROGRAM Name Program No. Page No. Livington, R. 42 93 Lobo, J. 9 274 Magner, M. 53 336 Marechal, J 45 452 Martin, J. 16 77 Martin, K. 2 32 Masters, L. 11 305 Mathey, R. 45 442 McFadden, J. 3 39 McKnight, M. 29 424 Miller, E. 28 279 Nakano, S. 20 232 Nelson, P. 48 463 Oakley, D. 36 225 Ohama, Y. 34 242 Pihlajavaara, S. 25 194 Pommersheim, J. 9,11 274,305 Price, R. 2 32 Proctor, B. 36 225 Prudon, T. 39 106 Radakovich, T. 53 336 Rakhra, A. 7 13 Reidenouer, D. 35 136 Revirand, G. 17 387 Rodgers, S. 35,53 327,336 Rossiter, W. 45 442 Sasse, H. 15 378 Scholer, C. 24 153 Schrage, I. 15 378 Schwartz, T. 49,50 249,264 Senbetta, E. 24 153 470 AUTHOR INDEX AND CROSS-REFERENCE TO CONFERENCE PROGRAM Name Program No. Page No. Sereda, P. 11 28 Setzer, M. 33 160 Sneck, T. 8,38 22,63 Springenschmid, R. 21 199 Stockbridge, J. 39 106 Svane, P. 18 410 Syal, S. 23 173 Tada, S. 20 232 Thomas, D, 5 297 Thomasen, S. 40 108 Van Court, D. 47 464 Veschuroff, B. 44 115 Volkwein 21 199 Zarghamee 50 264 1>U.S. GOVERNMENT PRINTING OFF I CE t 1981-340-997/178 2 471 LIBRARIES Aoooc^cnoam.