4 : . • .. . A by . . TOFI ORNLP 2515 * To will fi A : . . * 1 + . .1.. . .. men * EEEEEEEE .. + MICROCOPY RESOLUTION TEST CHART NATIONAL BUREAU OF STANDARDS -1963 odoo-p-2015 F 66-93 NOV 2 9 1966 Conf.660923.3 HC $7.00; 150 NOTE: This is a manuscript of a lecture to be presented at the Cairo Solid State Conference "Interaction of Radia- tion with Solids," Cairo, Egypt, September 3-7, 1966, and will be published in a book titled "Interaction of Radiation with Solids. - MASTER - . c. ' - . - ' RADIATION EFFECTS IN SEMICONDUCTORS James H. Crawford, Jr. LEGAL NOTICE 11216- i This report was prepared as an account of Government sponsored work. Neither the United States, aor the Commission, nor way person acting on behalf of the Commission: A. Makes any warranty or representation, expressed or implied, with respect to the accu- racy, completeness, or usefulness of the information contained in this report, or that the use of any loformation, appuratus, method, or process disclosed in this report may not infringe privately owned rights; or B. Assumes any liabillues with respect to the use of, or for damages resulting from the use of any information, apparatus, method, or process disc!osed in this report. As used in the above, "persoa acting on biball of the Commission" Includes any em- ployee or contractor of the Commission, or employee of such contractor, to the extent that such omployse or contractor of the Commission, or employee of such contractor prepares, disseminates, or provides access to, any lolormation pursuant to his employmeat or contract with the Commission, or his employment with such contractor, . . -+ + * - . F . - . 2 - - SOLID STATE DIVISION OAK RIDGE NATIONAL LABORATORY Oak Ridge, Tennessee, U.S.A. Operated by UNION CARBIDE CORPORATION for the U.S. Atomic Energy Commission RELEASED FOR ANNOUNCEMENT : IN NUCLEAR SCIENCE ABSTRACTS m wird op het striptis. Sunt proble . - .' * . - - . " """ " " * * * -* - . - : . . RADIATION EFFECTS IN SEMICONDUCTORS“ James H. Crawford, Jr. Solid State Division, Oak Ridge National Laboratory Oak Ridge, Tennessee, U.S.A. - - Introduction - - --- - Nearly all of the properties that make semiconductors interesting to study and valuable for applications in solid-state electronics are strongly dependent upon the presence of imperfections in the crystalline structure. Therefore, they are excellent materials in which to investi- ..BA gate damage to the crystal lattice produced by high-energy radiation. This modulation of electronic behavior, which we call "Radiation Effects", is a clue to the type of damage introduced and, if we are clever enough .. A - os K to connect specific defect structures with the various qualitative and .ee,+Sa quantitative changes in properties, we should be able to construct a complete picture of the lattice damage introduced by various types of energetic radiation. Studies of radiation effects in semiconductors have an additional advantage: From them we learn much about radiation sensitivity of properties vital to the operation of semiconductor de- vices, as well as various factors which affect this sensitivity. Now many potential applications of semiconductor devices may involve radia- tion environment, space vehicle communication gear, for example. There- fore, knowledge pertaining to radiation sensitivity will be invaluable in developing devices with longer service lives in outer space or when V used near a nuclear reactor power source. *Research sponsored by the U. S. Atomic Energy Commission under contract with Union Carbide Corporation. As a field of study, radiation effects in semiconductors is not new. The first work in this field was done on germanium and silicon nearly 20 years ago at Purdue University and Oak Ridge National Labora- tory, and the late Karl Lark-Horovitzt was an outstanding leader in these investigations. But although the field is not new, its present complexion is vastly different from what it was six or seven years ago. Before that time the main emphasis was colored by what turned out to be a vain attempt to test quantitatively in terms of defect yield the sim- ple theory of radiation damage, and by a valiant but equally vain attempt to account for the multitude of defect energy levels in terms of simple interstitials and vacancies. Beginning in 1959, several developments occurred which revealed without doubt that in the room-temperature range both the yield and nature of radiation defects in silicon and ger- manium are strongly influenced by the type and amount of impurity present. Since this discovery much effort has been devoted to the determination of the complex structures of defects, involving both impurities and simple vacancies, which are responsible for changes in electronic behavior. These studies have employed properties such as optical absorption of polarized light and electron spin resonance to reveal the symmetry and composition of the photon and microwave absorption centers. In addition, annealing studies have given some insight into the defect-complex struc- ture through the kinetics of the various relaxation processes, leading to both their formation and decomposition. In this brief summary, I would like to discuss the types of de- fects that result from irradiation with various types of energetic particles and photons; to indicate briefly the way in which imperfections " ..!!!..." 1 XY, liniste , . . -3- influence semiconducting behavior; to present evidence for the various defect-impurity interaction products that ace produced as a result of . . the migration of simple defects; and, finally, to mention recent inter- esting developments in the area of compound semiconductors. - - - - - . - - - . - The Nature of Radiation Defects . -.3 . In elemental systems such as germanium and silicon, which will be our primary concern, defects are not created in the perfect lattice by ionization through radiochemical processes as they are in the alkali. - - - halides. Instead, elastic collisions between the incoming particles and lattice atom are required. Moreover, the recoil energy imparted to the lattice atom must be greater than some minimum value Ea, called the displacement energy, which is required to create a stable interstitial- vacancy pair. In the early, naive damage theory, the displacement prob- ability was assumed to be a step function: 0 if T< Eq, and unity or greater if T > Eq. Using this simple concert, E, is usually measured by exposing a crystal to a prescribed dose of Van de Graaff electrons and increasing the energy until a change in a radiation sensitive property is first detected. E, is related to this threshold electron energy by the relativistic expression for the maxil in energy imparted in an elas- tic collision, namely, Ez = 2(E + 2 me?) E/Me? (1) where E and m are the energy and rest mass of the electron, and M is the VW mass of the target atom. L ' .. . . ve . . The naive theory does not account for the ordered arrangemert of atoms. Obviously, if the lattice structure is considered, the ease of displacement will depend upon the direction of the impulse: In open directions E, will be much smaller than in the direction of a neighbor- ing ion. Also, in determining E, care must be taken that the defects created are stable enough for experimental detection. Such a complica- tion is encountered in germanium. Low-temperature electron bombardment experiments (1 to 4.5 MeV electrons at 4°K) of MacKay and Klontzº have shown that conduction electrons are required to stabilize close interstitial-vacancy pairs. These defects, which act as acceptors and hence remove conduction electrons, are therefore produced in higher yield the higher the concentration of conduction electrons. In fact, in p-type specimens where the current is carried by positive holes in- stead of electrons the yield is almost below detection. From an experi- mental point of view then, the electronic state of the semiconductor (whether an electron or hole conductor) would appear to have a strong influence upon the apparent value of E, for low-temperature (<10°K) irradiations with electrons since even at liquid helium temperature these close-pair defects are unstable in the absence of conduction electrons. Another complicatior, which may also be encountered in other solids as well but for which there is direct evidence in silicon is the creation of divacancies by electron bombardment. This occurs when the impulse is along the appropriate crystallographic direction and the re- coil energy high enough. Using electron spin resonance techniques, Corbett and Watkins have found that divacancies are formed in silicon . . IN 2017 ..11 . WO SNL .. : WoW .. during bombardment with 0.7 to 1.5 MeV electrons at 20°K, and that the yield is highest for the bombarding beam parallel to the Si-Si bond di- rection. These results indicate that the divacancy is formed by a single collision event and that divacancies are a primary radiation defect. As the energy of the primary recoil increases to many times Eg, the recoil acts as a hombarding particle and secondary, and even tertiary, displacements are created as it collides with other lattice atoms. This results in what is called a displacement cascade in which the defects are closely spaced and have a high local concentration. Damage produced by fast neutron bombardment in a nuclear reactor consists predominantly of regions of high defect density (disordered regions) of this type. Typi- cally, these regions are roughly spherical with a diameter of ~100 Å and these have been resolved in germanium by electron micrography.° These disordered regions have a rather unusual effect on the electrical behavior of n-type germanium. Since the dominant effect of radiation defects in germanium is acceptor action, the disordered regions are p-type islands in an n-type matrix and as such they must be isolated from the matrix by a potential barrier and an electrical double layer, such as shown in Fig. 1. The radial extent of the space-charge zone in the junc- tion surrounding the disordered region is roughly given by the Debye- Hückel length. (2) ' . " . . . . . where q is the electronic charge, e is the dielectric constant, and N j.s the donor density. For reasonable purity n-type germanium (donor concentration of 2014 to 10+5), this distance is between one and two orders of magnitude larger than the radius of the disordered region itself. The potential gradient extending over this region denudes it of mobile charge so that the entire space-charge zone acts as an insu- - - - lator blocking current flow. The situation is shown schematically in -- - - - + Fig. 2. Because oỉ this double-layer, disordered regions markedly re- duce the conductivity and for non-overlapping space-charge zonesº the conductivity is approximately 0~0 (1 - 1)/(1 + ) (3) where o, is the true bulk conductivity and I is the fraction of the volume occupied by non-overlapping insulating zones, which is given approximately by f = 7 prº obat, (4) where T.. is the average volume of the region affected by space-charge zones, nº is the density of lattice sites, o. is the scattering cross S section for high-energy electrons, and Øt is the integrated flux of neutrons with energies (>0.7 MeV) sufficient to create a disordered region large enough to support a space-charge zone. Solution of Poisson's equation for the spherical case yields "sc = vo e rz'a Na 15) . - " 7112 -7- where yo is the difference in electrochemical potential between the dis- p ordered region and the unperturbed matrix, r, is the radius of the dis- - - ordered re e net donor concentration in the matrix. - This model of disordered regions and associated space-charge zones has been verified for neutron-bonbarded n-type germanium by several inves- tigators in terms of electrical properties alone'>7910 ard tie disordered - - - -- regions are found to have a radius of 50 to 100 Å. The electrical results - ca - have been confirmed by means of thermal conductivity measurements recently made by Van Dong. -- He found the mean disordered region radius to be 70%. 17 crdered region and the space-charge on electron micrographs of replicas of etched surfaces of germanium after bombardment with 14 MeV neutrons. Figure 3a shows a micrograph on which the space-charge zone (1000 to 1500 Å radius) is clearly visible. Figure 35 shows an enlargement of one of the etch figures, and at the center is detectable a crater of f radius 150 to 200 Å which is presumably the disordered region. Inter- estingly enough, presumably because of its smaller atomic mass and dif- ferent defect energy level structure, neutron-irradiated silicon shows no definite evidence of disordered regions.13 Not all the energy of primary recoils is consumed in the displace- ment cascade. It was mentioned earlier that ionization is not effective in creating lattice defects in elemental solids. In fact, energy lost through "inelastic" ionization and excitation processes is not available for the displacement of lattice atoms. The higher the velocity and the larger the charge of the primary recoil, the greater the amount of energy RE . .t lost in these inelastic processes. Another mechanism of fruitless 'W YA JU " -8- * * - - * - dissipation of recoil energy is through "focusing" or focused colli- . . . . . . sions+* down a close-packed row in a crystal. The succession of energy . transfers in a focused chain interaction in effect drains energy away from the region of primary impact or the cascade region. An additional mode of energy loss for energetic recoils or incident ions is called "channeling".+? This occurs when the high energy particles find them- 11 selves in an open channel or crystallographic direction. Small energy losses occur through ionization or very small angle collisions with the rows of atoms defining the channel. The most open direction and hence the most probable channel in the diamond structure is along any of the <110> directions. This is clearly shown in Fig. 4. In concrast, views along the [100] direction in Fig. 5, and the (111) direction in Fig. 6 show very little open space. That this is indeed the channel direction for the diamond lattice is shown by the dependence of the range of 40 keV xenon ions in silicon in Fig. 7. These measurements by Davies and coworkers' show quite dramatically that ions penetrate much more deeply along this direction than any other low index direction. Also shown for comparison is the range in sing.le-crystal aluminum which has a face- centered cubic structure with a (110) channel direction. Another very effective demonstration of channeling is afforded by the response of silicon p-n junction fission detectors. One problem that arises in using junction detectors in fission-fragment spectroscopy is the failure of the pulse-height absorbed energy calibration for high energy The smaller than expected pulse of current indicates that some other process besides ionization is available for dissipating a portion of the . -9- S .- . energy. Recently, Moak and coworkers' have investigated the influence . ... . . - - of crystal orientation upon pulse-height response of a detector to "'I ions with energies in the range of 10 to 100 MeV. When the detector was aligned such that the incident beam was parallel to a <111> direction, the pulse-height defect for the channel ions virtually disappeared. -- - - - --- Therefore, one is lead to the conclusion that the pulse-height defect is - caused by energy losses in elastic collisions which displace atoms. When the ions are channeled the probability of displacement collisions is much reduced and nearly all of the particle energy appears as ionization cur- rent in the p-n junction. Influence of Defects on Semiconducting Behavior Defects, like impurity atoms, are capable of introducing local energy states into the band structure of semiconductors. Depending upon their nature and the electronic characteristics of the host lattice, de- fect centers may act as (a) donors, yielding electrons on thermal excita- tion in n-type material or capturing holes in p-type; (b) acceptors, in which capacity they may thermally ionize to produce holes in p-type conductors and capture electrons in n-type; or (c) as amphoteric centers capable of capturing electrons in n-type and holes in p-type. Moreover, it is common for radiation defects to possess two or more energy levels. In consequence, introduction of defects by irradiation can produce pro- found and often complicated changes in the electronic behavior. Not only is the carrier concentration altered even to the extent of inducing 5. conversion from n- to p-type (or conversely), but carrier scattering is affected by the introduction of charged centers and the influence upon 4 dh 13 2. . -10. non-equilibrium minority-carrier processes such as trapping and recom- bination by the deep level defects which act both as efficient recombin- ation centers and traps is very large. In short, all semiconducting properties are influenced in some degree by exposure to energetic par- ticles or photons capable of displacing lattice atoms. In studies of radiation effects to use is made of the influence of defects on properties to gain quentitative information about the position and relative numbers of energy levels produced by a given exposure to various types arıd energies of incident particles. The most common in- vestigation involves the temperature dependence of the carrier concen- tration and the mobility which are obtained from measurements of Hall coefficient and mobility. This approach yields the position of the thermally accessible energy levels and information on their charge state. Other electronic properties that have been useful in locating enero levels of radiation defects not readily detected by the above approach include excess carrier lifetime, trapping of minority carriers, and the spectral dependence of photoconductivity. Both electron spin resonance and optical absorption have been very valuable tools in gaining detailed information about defect structure. Additional information on defect stability and the tendency for various defects to interact with other imperfections (point defects, im- purities, and dislocations) is obtained from the thermal annealing of radiation defects. By employing a radiation sensitive property as an index of the defect content, the kinetics governing the restoration of initial conditions through some annealing cycle (isochronal or isothermal) may be examined to yield values of defect motion energies or the binding energies for defects in certain complexes.to """" - -... "O m .. " coons. Wh -li. The Structure of Radiation Defects As we have already indicated above, delects remaining in germanium or silicon irradiated with electrons in the 0.5 to 5 MeV range, or even more massive particles such as protons or neutrons in the room-temperature - range, are not simple interstitials and vacancies but rather are invari- - -- - - ably complexes formed by the interaction of simple point defects with other imperfections, notably impurity atoms and vacancy pairs as well as more complex clusters of defects. The first clue to this state of affairs came in 1959 when Bemskity and Watkins and coworkers observed that irradiation of oxygen contain- ing natype silicon gave rise to an imperfection with both a spin-resonance spectrum and an ionization energy that are quite different from that ob- served in specimens containing little oxygen. It was concluded and sub- sequently confirmed by optical absorption of polarized light that the defect in the oxygen-containing material is a silicon vacancy associated with an interstitial oxygen atom. The model for this defect is shown in Fig. 8, and it is distinguished by the fact that it acts as an acceptor center with an energy level located 0.17 eV below the conduction band. Conclusive identification of this center with an oxygen-vacancy com- plex was accomplished via an optical absorption band at 12u, which is produced by electron irradiation under the same conditions that the de- fect resonance is produced. This band is made birefringent by uniaxial pressure in the same way that certain equivalent orientations of the pa ramagnetic center become more heavily populated by the application of pressure and, significantly, the thermal annealing of both the birefrin- gence and the imbalance in the spin system follows the same path. The set) . .. -12- final point of evidence that the center responsible for the 12u band con- tains oxygen is the observation that specimens doped with oxygen enriched in +90 exhibit second band shifted to lower frequencies by the amount ex- pected for the increase in isotopic mass to influence an antisymmetric stretching vibration. Since this early discovery, a number of defect complexes have been identified in both n- and p-type silicon after irradiation with electrons. Figure 9 shows the mode122 of the vacancy-phosphorous complex which in n-type silicon is responsible for at least one acceptor level located 0.43 eV below the conduction band. This center can be identified by ESR alone since the hyperfine interaction of the unbalanced spin with the inovement of the phosphorous demonstrates conclusively that phosphorous impurity is a part of the defect structure. In Fig. 10 we see the model of the divacancy,' the production of which as a primary defect by electron bombardment along a <111> direction we have already discussed. This rather complex-looking center is respon- sible for three energy levels: An acceptor state ~0.4 eV below the con- duction band; an acceptor state quite near the center of the forbidden energy gap (~0.5 eV above the valence band); and a donor level quite near the top of the valence band. More recent studies on p-type silicon have revealed the existence of a vacancy-acceptor complex (specifically a vacancy attached to an alumi- num atom), interstitial acceptor impurity (both gallium and aluminum), and complexes one component of which is an interstitial acceptor atom. An example of the latter is a Gas molecular complex, with one of the gallium atoms in a substitutional site and the other occupying an interstitial = -- - - -13- - . - - position. The energy levels that have been attributed to these defects are summarized in Fig. 11. We must now ask by what mechanism are these complexes formed? The divacancies’ have already been discussed and it was shown that they are primary defects formed directly by a single collision. However, the same cannot be said for the impurity-containing complexes because the im- purity concentration is so small (<10*4 mole fraction) that the probability that these can be created directly by the irradiation for the exposures employed is vanishingly small. Therefore, it must be concluded that these are formed by the interaction of mobile simple defects (intersti- tiels and vacancies) with these imperfections. A study of the ESR spec- tra of silicon irradiated at low temperature shows that this is indeed the case. Watkins,64 has observed the ESR spectrum for the isolated vacancy in p-type silicon after electron bombardment at both 49 and 20°K. However, the vacar.cy resonance disappears as the crystal is warmed to the vicinity of 170°K, presumably because the vacancy migrates to and forms complexes with other imperfections or annihilates at vacancy sinks. Pulse - - - . - - - annealing studies reveal that the activation energy for motion is 0.33 $ 0.03 ev. In n-type silicon the vacancy is not detectable by an ESR signal presumably because it contains only paired electrons when the - - - - - - .- Fermi level is in the upper half of the forbidden energy gap. However, the motion and trapping of vacancies can be detected through the forma- tion of phosphorous-vacancy pairs which are formed during annealing of n-type silicon after low-temperature irradiation. It is noteworthy that - 8 : the vacancies move with a much smaller activation energy in n-type material, - - - confirming the hypothesis put forth many years ago that the charge state - - - -7 - = - - - - - -14- of defects should influence their ease of motion. From consideration of the energy level structure of the vacancy, it is concluded that the neu- tral vacancy is the form that moves in p-typc Si with the 0.33 eV motion energy while both the vº and v“ forms expected to exist in n-type Si (Fig. 9) have appreciably smaller activation energies of migration." All attempts at detecting isolated interstitials by ESR methods have failed. Although one might conclude from this that they may always be present in a nonmagnetic form and therefore escape detection, careful experiments with p-type silicon irradiated at 4°K indicate quite another cause for their absence. Watkins") has identified a resonance associated with interstitial aluminum ions (A1**) and measurements of the relative intensities of microwave absorption show that their concentration is vir- tually the same as the isolated vacancy concentration. The se interstitial impurities are tightly lodged in the lattice and begin to move only upon heating to the vicinity of 200°C when they also interact with other im- perfections to form complexes. In gallium-doped material post-irradiation anneals near 200°C there appear gallium pairs, one member of which is in- terstitial adjacent to the other member in a substitutional site. As a result of these observations, it is concluded that the inter- stitial silicon moves rapidly through the lattice by some type of inter- stitialcy mechanism (the interstitial propagates by interchange with substitutional atoms) and at temperatures even as low as 4°K when impuri- ties such as Al, Ga, or perhaps even donor atoms are encountered, these are ejected into interstitial sites, thus breaking the interstitialcy chain process. The interstitial impurities can now migrate only by a normal interstitial diffusion and a much higher activation energy. (Note WAN " 7 LAW 1 W . OW." . " " W . 5 * : - > * -- . * - - - - -15- - - - - P - - - - - - that the activation energy of interstitial motion for Li', the smallest - ion studied, is 0.66 ev in silicon.29) . .- * - * It should be mentioned here that these defect structures account Por only a fraction of the resonances that have been isolated from the complex spectra of electron-irradiated silicon. In addition, other resonances including those for S = 1 states have been obtained on fast neutron-irradiated silicon and an oxygen containing state in p-type silicon28 has been reported. As a result of this outstanding series of ESR investigations, we are now in a position to understand the full complexity of the problem of identifying, the imperfections responsible for the various acceptor and donor states not only in irradiated silicon but in germanium and other semiconductors as well. Rather than attempting to force experi- mental observations into a conceptual framework in which only the exist- ence of simple defects is admitted, we have learned to accept the prin- ciple that the products of irradiation will include nearly every conceiv- able complex between vacancies and impurities that are present as well as numerous interstitial-impurity complexes. This new point of view has led to quite rapid progress in understanding striking differences in behavior in specimens with slightly different impurity content. For example, we show in Fig. 12 the effect of oxygen impurity on the radiation response of the excess carrier lifetime T in n-type silicon. Although the two specimens have nearly the same initial lifetime and conductivity, the oxygen-free material is much more strongly affected than crystals which contain 10+T to 10+ oxygen atoms per cm. The difference can be readily iz explained: The deep acceptor state associated with the phosphorous- vacancy complex, which is the dominant radiation defect in the oxygen-free -16- crystal, iis a much more efficient recombination center than the shallow state associated with the oxygen-vacancy complex. Interestingly, the higher purity specimen is more sensitive to radiation with respect to this particular property than the less pure one, and since excess car- rier lifetime is so important to the operation of most devices, it would appear that, other things being eque.l, a diode or transistor fabricated from quartz-crucible grown silicon would have a better radiation resis- tance than the less highly oxygen contaminated floating-zone grown crystals. Let us now see what the results obtained on silicon can tell us about germanium. Unfortunately, defects in germanium are not so amenable to study by ESR techniques and one is therefore forced for the most part to infer the nature of the processes associated with electrical property changes resulting from irradiation and subsequent annealing, in terms of defect structures analogous to those found in silicon. The possibility of defect-impurity complexes in irradiated germanium was first pointed out by Brown and coworkers who found that radiation defects were ther- mally more stable in n-type specimens doped with small donor atoms (As and P) than those doped with antimony. The recovery is much more rapid in the antimony-doped specimens than in the others. A nairo, interpreta- tion of this observation would be that the smaller donor atoms, i.e. arsenic or phosphorous, trap and stabilize a simple defect. Subsequent workS+,32 has shown this view to be untenable since the rate of recovery in this temperature range increases as the antínony concentration in- creases, indicating that the impurity must either be directly involved in the unstable defects or participate in the annealing stages. The -17- influence of antimony concentration on the recovery of excess carrier lifetime 18 shown in Fig. 13. Arsenic-doped material shows only a very slight influence of concentration. - - Until recently there was no direct evidence that point defects ... - - - - - . - -: :- . might exhibit a similar behavior and form similar complexes with impuri- ties in germanium as they do in silicon. However, Whan has presented conclusive evidence that vacancies begin to move in germanium near 650K and form a series of oxygen complexes which can be detected through their vibronic infrared absorption spectra. Armed with this vital piece of evi- dence, analysis of annealing data in terms of vacancy-impurity complexes and perhaps even interstitial-impurity complexes can be carried out with reasonable confidence. Indeed such defects have been invoked in inter- pretins the isochronal recovery of electron concentration and reciprocal mobility in n-type germanium which had been exposed to °°co y rays at 78°K. Evidently, even at this temperature vacancies are able to migrate anä form donor-vacancy complexes, and the mutual annihilation of intersti- tial donors and vacancies. It should be pointed out that Whan's results as well as those of Watkins' on the ease of vacancy motion confront us with a serious para- dox concerning the mechanism underlying self-diffusion in these inaterials. The self-diffusion energy for germanium as measured by Letaw and coworkers34 is 3 eV while recent studies on silicon35,30 yield 5.1 ev. If the diffusion is assumed to proceed through the motion of vacancies -- - this energy is the sum of the energy of formation and the energy of mo- tion of the vacancies. Now these component energy values can be separately determined by quenching and subsequent annealing experiments which have " -1.8. been carried out in the case of germanium. The temperature dependence of the quenched-in vacancy concentration as reflected by changes in Hall coefficient yields a formation energy of w1.9 eV, and the annealing of the excess vacancy concentration which should reflect the activation energy of motion gives a value of ~1.2 eV. The sum of these values is in good agreement with the self-diffusion energy. However, the motion energy determined thermally is at least a factor of five larger than that observed in irradiated specimens, as inferred from the results of Whan. An attractive solution to the paradox would be to say that the diffusion- current at high temperature is carried by divacancies. If the motion energy for the divacancy in germanium is comparable to its value in silicon determined experimentallyc (1.3 eV), it would certainly be consistent with the observed activation energy for the annealing of quenched-in defects. In fact, Seeger pointed out a number of years ago on the basis of the behavior of deformed germanium that the motion energy of germanium divacancies should be in the neighborhood of 1.5 ev. . Another factor favoring this viewpoint is the dislocation jog structure .... in the diamond lattice. Haasen and Seegers have pointed out that the shortest jog produced by the displacement of an edge dislocation from one glide plane to the next adjacent one is two atoms high. Now in many solids vacancies are known to be created in the bulk thermally by evaporation from dislocation jogs (evaporation of a vacancy causes the jog to advance a unit spacing along the dislocation) and these jogs which also act as sinks for the deposition of defects as the crystal --- .mw..cosmem .- -..wo--.......... wiedern is cooled from high temperature. Consequently, if in the diamond lat- tice the energy to move a jog by removing a pair of vacancies one at a time is much larger than removing the pair as a unit, one would expect -19- them to be produced bound together in pairs. If now the winding energy holding this pair together is large enough to prevent thermal breakup, the predominant thermal defects will be divacancies rather than single vacancies. This is just another way of stating that the formation energy of divacancies is smaller than that of single vacancies. In order for divacancy motion to dominate self-diffusion, the following relation must hold: + to V E (5) II . . - where the double-primes refer to the divacancies and the single primes to single vacancies. In view of the small value of EM (00.1 eV) indicated by the experiments of Watkins23 and Whan33, the self-diffusion energies would then require that E > 3 eV for germanium and 5 eV for silicon. It should be pointed out, however, that theoretical calculations of E-40-42 all yield values for germanium that are consistent with the accepted single vacancy model for self-diffusion, i.e., ~2 ev. Consequently, un- less it can be shown that the excellent agreement between these calcula- - - - - tions and experiment is fortuitous, it appears doubtful that the paradox can be resolved on the basis of divacancies being responsible for self- diffusion. Another possible way of accounting f r the disparity between the motion energies of radiation vacancies and thermal vacancies is the in- fluence of charge state. However, inspection of the energy level struc- ture of vacancies as shown in Fig. ll indicates that this suggestion would hardly be valid since at high temperature the vacancy would have an appreciable probability of being negative (the Fermi level lies in RE 14 -20- the center of the gap near the position of the acceptor level) and this is the charge state of highest mobility.28 Evidently, a careful re- examination of our ideas concerning the mechanism of self-diffusion in the diamond structure is required. Compound Semiconductors Because of the binary nature of III-V or II-IV semiconductors, the complexity of defect structures that might be expected to result from irradiation is considerably greater than that encountered in ele- mental semiconductors. Because of mutual interaction between radiation defects, interaction of two types of interstitials and vacancies with chemical impurity and a new type of defect-the misplaced atom (a Group III atom on a Group v lattice site and conversely)—the number of defect structures in these materials is perhaps nearer r than 2N, where N rep- resents the number of prominent defect structures in the elemental case. In view of this seemingly overwhelming complexity, one might question whether the experimental results could be interpreted in terms of any realistic defect models or yield any physical insights into the nature of radiation effects in these materials. For a review of the variety of results on the III-V compounds, the reader is directed to Aukerman's comprehensive review. Although in many instances complexity is indeed an obstacle, in some cases it works to our advantage and I would like to present briefly two such cases. The first of these concerns a peculiarity of the zinc-blende structure in a material such as indium antimonide. Each In atom is covalently bound to four Sb atoms, each being situated at the corners of a regular tetrahedron and the inverse of course holds for each Sb atom. -21- If these atoms could be transformed into identical atoms without any change in position or binding, the lattice would assume the diamond structure, which indeed is a highly symmetrical structure. But let us now examine the environment of each In atom or each Sb atom situ- ated on a given (111) plane. It is immediately evident that the lattice is unsymmetrical with respect to this plene. For example, one covalent bond extends from each In atom to the left and three to the right, whereas this configuration is inverted from the Sb atoms. This sug- gests that there are two possible {111} surfaces, each containing a single type of atom, and a polarization direction, namely, the <111>. The existence of this unsymmetrical situation has important consequences for many properties of III-V compounds: cleavage, piezoelectric effect, chemical behavior both with respect to oxidation and etching, a variety of electronic transport effects, and even crystal growth. ** As might be expected this polarized structure is also reflected in the ease of defect creation by fast electrons. Eisen has deter- mined the threshold energy and the defect yields in indium with the electron beam in the two senses of <111> direction. In the "minus" direction, examining those atom pairs parallel to the beam, the elec- tron secs first the In atom immediately behind which is the Sb a com and immediately behind that there is an open interstitial site. The "plus" direction is the inverse, i.e., the Sb atom, followed by the In atom, foilowed by the interstitial site. The results indicate that --- - --- not only is the displacement energy E, different for the two directions but that the thermal stability of the defects is quite different. For the beam direction into the (121) face (the plus direction), it is seen Ei iza RUS- -22- from Fig. 14 that there is an open interstitial site immediately behind the In atom; hence, this should be the direction for easy In-aton dis- placement. On the other hand, an Sb atom receiving an impulse in this direction would have to collide with an indium atom before reaching the interstitial site. The situation is, of course, inverted for the beam into the (1ll) face, which corresponds to the direction for easy Sb-atam displacement. An analysis of the effect of irradiation on the Hall co- efficient reveals that the displacement energy E,(In) for Ir. in the easy direction is 6.4 eV and the resulting defect is thermally more stable than the structure created by displacement of Sb atoms for which 8.5 < E.(Sb) < 9.9 ev. Although the precise structure of these defects which decrease the electron concentration by capturing electrons cannot be described, this investigation indicates that a displaced In atom plays a dominant role in the one which is thermally stable up to 128°K, and Sb-atom displacement is involved in the one that is stable only up to 83°K. A second series of experiments by R. O. Chester has employed nuclear characteristics of the atoms comprising the compound semiconduc- tor to throw light upon the defects responsible for acceptor and donor action. She studied the response of cadmium sulfide and cadmium telluride to gamma irradiation employing Zifferent y-ray sources, namely, "Co with E, = 1.17 and 1.33 MeV, and 137cs with E, = 0.66 Mev. Because of the markedly different mass and atomic numbers of the ca and S atoms, it can be derionstrated from simple collision theory that the 0.66 MeV photon from 1-37Cs decay displaces s atoms and the Gºco photons predomi- nantly displace ca atoms. The situation is reversed in cadmium telluride MH -23- for which the smaller energy photon preferentially displaces ca atoms but because of the greater similarity of atom mass the distinction is not as great as in cadmium sulfide. Investigation of the behavior of Hall coefficient of gamma-irradiated n-type cadmium sulfide and both n- and p-type cadmium telluride shows that preferential ca-displacement leads to the introduction of acceptor states while displacement of the electro-negative constituent leads to donor action. Because of the propensity of simple defects for forming complexes with other imper- fections, one cannot infer from these results the precise structure of the defects responsible for the dominant electronic states. Never- theless, this study indicates what are the dominant ingredients of the radiation acceptors and donors in the II-VI compounds. REFERENCES 1. For an early review, see K. Lark-Horovitz, Semiconducting Materials, ed. by H. K. Henisch (Academi.c Press, New York) 1951, p. 47. 2. These developments were first reported at the Gatlinburg Conference on Radiation Effects in Semiconductors in 1959, and the Proceedings were published in J. Appl. Phys. 30, No. 8 (1959). 3. For a review, see D. S. Billington and J. H. Crawford, Radiation Effects in Solids (Academic Press, New York) 1950, Chap. II. 4. J. W. MacKay, E. E. Klontz, and G. W. Gobeli, Phys. Rev. Letters 2, 146 (1959); J. W. MacKay and E. E. Klontz, J. Appl. Phys. 30, 1269 (1959): E. E. Klontz and J. W. MacKay, J. Phys. Soc. Japan 18, Suppl. III, 216 (1963). 5. J. W. Corbett and G. D. Watkins, Phys. Rev. Letters I, 314 (1961). 6. J. R. Parsons, R. W. Balluffi, and J. S. Koehler, Appl. Phys. Letters 1, 57 (1962). 7. B. R. Gossick, J. Appl. Phys. 30, 1214 (1959); J. H. Crawford, Jr., and J. W. Cleland, J. Appl. Phys. 30, 1204 (1959); J. W. Cleland and J. H. Crawford, Jr., Proc. International Conf. on Semiconductor Physics, Prague, 1960 (Academic Press, New York) 1961, p. 299. 8. H. J. Juretschke, R. Landauer, and J. A. Swanson, J. Appl. Phys. 27, 838 (1956). 9. H. J. Stein, J. Appl. Phys. 31, 1309 (1960). 10. W. H. Closser, J. Appl. Phys. 31, 1693 (1960). 11. N. Van Dong, J. Appl. Phys. 36, 3450 (1965). See also, N. Van Dong, P. Ngu Tung, and M. Vanderyrer, Compt. Rend. 236, 1722 (1963). REFERENCES (Con't) 12. M. Bertolotti, T. Papa, D. Setti, V. Grosso, and G. Vitali, Nuovo Cimento 29, 1200 (1963); J. Appl. Phys. 36, 3505 (1965). 13. For a summary of recent developments, see F. L. Vook, Radiation Damage in Semiconductors, ed. by P. Baruch (Academic Press, New York) 1965, p. 51. 14. R. H. Silsbee, J. Appl. Phys. 28, 1246 (1957). 15. M. T. Robinson and 0. S. Oen, Appl. Phys. Letters 2, 30 (1963); Phys. Rev. 132, 2385 (1963). 16. J. A. Davies, G. C. Ball, F. Brown, and B. Domeij, Canadian J. Phys. 42, 1070 (1964). 17. C. D. Moak, J. W. T. Dabbs, and W. W. Walker, Bull. Am. Phys. Soc. 11, 101 (1966); to be published in Rev. Sci. Inst. 18. For a coverage of radiation sensitive properties and thermal anneal- ing of radiation defects in semiconductors, the reader is directed to recent review papers: J. H. Crawford, Jr., Interaction of Radia- tion with Solids, ed. by R. Strumane, J. Nihoul, R. Gevers, and S. Amelinckx (North Holland Publishing Co., Amsterdam) 1964, p. 421; V. S. Vavilov, phys. stat. sol. 11, 447 (1965). An excellent cover- age of recent work is also to be found in Radiation Damage in Solids, ed. by P. Baruch (Academic Press, New York) 1965. 19. G. Bemski, J. App... Phys. 30, 1195 (1959). 20. G. D. Watkins, J. W. Corbett, and R. M. Walker, J. Appl. Phys. 30, 1198 (1959). 21. J. W. Corbett, G. D. Watkins, R. M. Chrenko, and R. S. McDonald, Phys. Rev. 221, 2015 (1961'. - . - . . - . .. TEL . . . ! " . Ny 21, REFERENCES (Con't) 22. G. D. Watkins and J. W. Corbett, Disc. Faraday Soc. 31, 86 (1961). 23. G. D. Watkins, Radiation Dama ge in Semiconductors, ed. by P. Baruch (Academic Press, New York) 1965, p. 97. 24. G. D. Watkins, J. Phys. Soc. Japan 18, Suppl. II, 22 (1963). 25. H. Y. Fan and K. Lark-Horovitz, Semiconductors and Phosphors, ed. by M. Schon and H. Welker (Interscience, New York) 1958, p. 113. 26. J. Maita, J. Phys. Chem. Solids 4, 68 (1958). 27. Mun Jung and G. S. Newell, Phys. Rev. 132, 648 (1963). 28. N. Almeleh and B. Goldstein, Bull. Am. Phys. Soc. 11, 16 (1966). 29. K. Matsuura and Y. Inuishi, J. Phys. Soc. Japan 16, 339 (1961); Y. Inuishi and K. Matsuura, J. Phys. Soc. Japan 18, Suppl. III, 240 (1963). 30. W. L. Brown, W. M. Augustyniak, and T. R. Waite, J. Appl. Phys. 30, 1258 (1959). 31. 0. L. Curtis, Jr., and J. H. Crawford, Jr., Phys. Rev. 124, 173 (1961); 126, 1342 (1962). 32. J. C. Pigg and J. H. Crawiord, Jr., Phys. Rev. 135, A1141 (1964). 33. R. E. Whan, Appl. Phys. Letters 6, 221. (1965); Phys. Rev. 140, A690 (1965). 34. H. Letaw, W. M. Portnoy, and L. Slifkin, Phys. Rev. 102, 636 (1956). W. R. Wilcox and T. J. Lachapelle, J. Appl. Phys. 35, 240 (1964). 36. B. J. Masters and J. M. Fairfield, Appl. Phys. Letters 8, 380 (1966). 37. For a recent assessment of this matter, see A. Hiraki, J. Phys. Soc. H. Let W. M. Portnoy, and L. Slifkin lys . Rev. Japan 21, 34 (1966). REFERENCES (Con't) 38. A. Seeger, Solid State Physics in Electronics and Telecommunications, ed. by M. Desirant and J. L. Michiels (Academic Press, New York) 1960, Vol. I, p. 61. 39. P. Haasen and A. Seeger, Halbleiterprobleme, ed. by W. Schottky (Vieweg and Solm, Braunschweig) 1958, vol. IV, p. 68. 40. R. A. Swalin, J. Phys. Chem. Solids 18, 290 (1961). 41. A. Scholz and A. Seeger, phys. stat. sol. 3, 1480 (1963); Radiation Damage in Solids, ed. by P. Baruch (Academic Press, New York) 1965, p. 315. 42. K. Bennemann, Phys. Rev. 137, 1497 (1965). 43. L. Aukerman, Semiconductors and Semimetals, ed. by R. K. Wiliardson and A. C. Beer (Academic Press, New York) Vol. 4, Physics of III-V Compounds (to be published). 44. J. W. Faust, Semiconducting Compounds, ed. by R. K. Willardson and H. L. Goering (Reinhold Publishing Corp., New York) 1962, p. 445. 45. F. H. Eisen, Phys. Rev. 135, A1394 (1964). 46. R. O. Chester, to be published. See also, Ph.D. thesis from the University of Tennessee, ORNL-3767 (unpublished). ( : . 3 11 FIGURE CAPTIONS Fig. 1 The structure of the electrical double layers (space-charge zone) associated with a disordered region in germanium. Fig. 2 Schematic drawing of the disordered region and its associated space-charge zone. Part (a) refers to a disordered region large enough to support a double layer of maximum radial extent. Fart (b) shows the relatively minor perturbation of the band edges by a small disordered region. Fig. 3 Replica micrographs of etched structures on fast-neutron bombarded germani.um. (a) Low magnification showing the space- charge zone; (b) high mignification showing the discrdered region evident as a small pit in the center of the larger etch structure. (After Bertolotti et al., ref. 12.) Fig. 4 A vie along a <110> direction of the diamond structure. Fig. 5 A view along a <100> direction of the diamond structure. Fig. 6 A view along a direction of the diamond structure. Fig. 7 Penetration of 40 keV 125xe ions into single crystals of silicon and aluminum. (After Davies et al, ref. 16.) Fig. 8 Model for oxygen-vacancy center in electron-irradiated silicon. (After Watkins and Corbett, ref. 20.) Fig. 9 Model for the phosphorous-vacancy center in electron-irradiated silicon. (After Watkins and Corbett, ref. 22.) Fig. 10 Model of the divacancy in electron-irradiated silicon. (After - :.- --. . Corbett and Watkins, ref. 5.) . - -- * .* 1 5'S "I FIGURE CAPTIONS (Con't) Fig. ll Position in the band gap of energy levels associateå with defect complexes as identified by ESR in electron-irradiated silicon. D-V - donor-vacancy complex 0-V - oxygen-vacancy complex va - divacancy Al, - interstitial aluminum Fig. 12 Change in reciprocal of lifetime of excess carriers in n-type silicon exposed to y rays, showing the difference in response of oxygen-free (vacuum floating zone )and oxygen containing (crucible grown) material. (After Matsuura and Inuishi, ref. 29.) The influence of antimony concentration on the recovery of Fig. 13 excess carrier lifetime in y-irradiated n-type germanium. (After Curtis and Crawford, ref. 31.) A projection of the zinc-blende lattice onto a {110} plane. Fig. 14 ORNL-LR-DWG 45249R | CONDUCTION BAND - - - - - - - ENERGY +- TFERMI LEVEL VALENCE BAND - - --- --- - - - - - P. TYPE N-TYPE lal -- -* CHARGE DENSITY -9N, L- (0) Fig. 1 - - - - ORNL -LR-DWG 37886 CONDUCTION BAND -- -- î --- - ----_946 - 2- - -- - VALENCE BANDS 11 los lo > 150 Å NA - 5x108 cm-3 ☺ < 150 Å NA - 5x10"cm-3 Fig. 2 R * 1.141 . . . que *. A . 1 ' og + . Fig. 3 - . * i -- -- --- * ; T etirdi . -weverancie m ini .- - 5 .- . - .- .-. .- -- -- .. . me magere... a -. . .- tore . - . - . - . - . -. .-- - - . , waitint 2 EA Ai e . * AN Fig. 4 17 9. WE, . I . AVA : 41 PH 7 i? ... . p . 25. ... ! . WA Fig. 5 ' " : 19 . th *** . Fig. 6 25 . . d I APA . M NU ....4 -. -.. .. . - - - *.. *. 2 = A . . . * t 1 : 9 : * ... -- - -. -- VR 4 - - - - - - - - - - - - - -- : - - - - - ORNL-DWG 94-9578 110 IN SILICON ---IN ALUMINUM RESIDUAL ACTIVITY (%) 50 200 250 100 150 DEPTH (ug/cm2) Fig. 7 ORNL-LR-DWG 54380 A CENTER Fig. 8 . . :-. Fig. 9 E CENTER ORNL-LR-DWG 51384 : -- -- - - -- .-Z-V79 - : .. - .- - -. VE s - set . i - ZEEL 18 - - NUN - - - - - * - - - - - Fig. 10 į ---- - -- -: : - 2 - -2- .. ORNL-DWG 66-7636 v D-V O-V 1 V2 Ali Ec - -0.4 – -0.2 - ov AVO v l?) TUITIO. DV I Ο -0.4 - OVO MIT MOHITETTI Ali I Ο -V------ + Ο TILTUMITIT +0.4 - + Ο vo OVO TIHIITIT +0.3 - +0.2 - TIL. +0.1 – out(?) Vet A1* TIILIITITIIL Ev — -V+ - - - ORNL-DWG 63-700 . . - . VACUUM FLOATING-ZONE Si AV (sec) LU + CRUCIBLE GROWN Si 0 1 2 7 8 (x40-7 3 4 5 6 Co60 y-RAY EXPOSURE (r) Fig. 12 ORNL-DWG 63 - 702 ISOCHRONAL ANNEAL'OF LIFETIME Sb-DOPED Ge AFTER CO60 y-IRRADIATION pllcm) [sb] y's/cm? to 15 9.7x1014 9x1016 | 96.2 2.7x1094 01.4 1.2x1046 5.2x1096) FA 0.4 4.2x1015 3x1095 5x1016) FRACTION NOT ANNEALED (A%/%) 70 90 910 230 250 930 950 970 190 210 ANNEALING TEMPERATURE (°C) Fig. 13 ORNL-DWG 66-7501 (111) SURFACE (117) SURFACE "MINUS" DIRECTION [111] "PLUS" DIRECTION Fig. 14 . .. -.-. - - - + - i - END tie AN --- DATE FILMED 12/ 21 / 66 . . ubos na adidas Klem wa more hand h a ikaran *** i ww w . ." . . . . ** NYUM . W * 2 1