Amorphization of elemental and compound semiconductors upon ion implantation

Transcription

1 Amorphization of elemental and compound semiconductors upon ion implantation K. S. Jones and C. J. Santana 214 Rhines Hall, Department of Materials Science and Engineering, University of Florida, Gainesville, Florida (Received 20 December 1989; accepted 11 January 1991) Cross-sectional TEM studies of ion implantation induced amorphization in a large number of semiconductors have been performed. Samples of Si, AlAs, GaAs, GaP, GaSb, InP, InAs, and ZnSe were simultaneously implanted at 77 K with 20 kev Si + at doses between 1 x /cm 2 and 1 x /cm 2. A dose of 1 x /cm 2 minimized the ion beam induced epitaxial crystallization and sputtering effects. The depth of the amorphous layer at this dose was compared with Monte Carlo damage density distribution calculations (TRIM'90). The threshold damage density (TDD) necessary for amorphization was determined for each compound. The values of the threshold damage density vary from as low as 2.4 x kev/cm 3 for InAs up to 7.3 x kev/cm 3 for AlAs. ZnSe never became amorphous and GaSb exhibited an unusual disordering after the highest dose. The values of the threshold damage density for the various compositions were compared with known thermochemical data and several bond energy estimates. No single calculation explained all of the trends observed. I. INTRODUCTION It is well documented that if sufficient energy is deposited into elastic (nuclear) collisions during ion implantation, the material will undergo a phase transformation from the crystalline to the amorphous state. Amorphization of single crystal Si has been studied extensively for the past 20 years because of its application to semiconductor doping. 1 " 7 Amorphization is known to reduce the random channeling tails of light ion implants and to increase dopant activation. In addition, amorphization followed by solid phase epitaxial regrowth can result in a significant decrease in the concentration and stability of implantation related dislocation loops. 8 Few studies of amorphization of compound semiconductors have been reported. 910 During solid phase epitaxial regrowth of GaAs, extensive category III 8 (regrowth related) defects form. Upon annealing, these defect microstructures can evolve from high densities of microtwins and stacking faults into dislocation networks. 11 " 15 The presence of these defects, which also lowers dopant activation, suggests that implantation induced amorphization should be avoided. This has been accomplished by using light ions (i.e., Be, Si), low doses (<1 x /cm 2 ), and elevated implantation temperatures (>100 C). 16 Because of this success, little work on the fundamental nature of the crystalline to amorphous transformation in compound semiconductors has been performed. 2 ' 9 In order to predict better the conditions necessary to avoid amorphization in compound semiconductors, there is a need to understand the threshold damage density of different alloy systems for a variety of different species and implant conditions. This requires both a fundamental understanding of what solid state properties dominate the amorphization process and accurate experimental data. In addition to predicting how to avoid amorphization, knowledge of the threshold damage density may also allow us to use the defects associated with the solid phase epitaxial regrowth process for other purposes. Ion implantation and subsequent annealing of compound semiconductor superlattices have been found to enhance AlAs/GaAs superlattice disordering, which is useful for fabricating optical waveguides. It is known that the morphology of the defects arising from the implantation process is related to the formation of an amorphous layer and that the implantation induced defects influence the superlattice intermixing rates Thus optimization of superlattice intermixing also necessitates an increased understanding of the amorphization threshold for different compositions. II. EXPERIMENTAL A. Procedure Eight different semiconductor materials were chosen for ion implantation: Si, AlAs, GaP, GaAs, GaSb, InP, InAs, and ZnSe. A piece of each of the (100) oriented semiconductors was mounted on three 3" silicon wafers and then implanted at 7 off axis. This mounting procedure circumvented the problem of the sensitivity of the threshold damage density to the implant conditions 19 ' 20 and allowed for a direct comparison among 1048 J. Mater. Res., Vol. 6, No. 5, May Materials Research Society

2 the different compounds. Materials were implanted with 20 kev 29 Si + at a dose of 1 x 10 14, 1 x 10 15, or 1 x /cm 2. The current density was =0.04 ^,A/cm 2 for the 1 x /cm 2 dose and 0.63 fia/cm 2 for the two higher doses. The implantation temperature was 77 K in order to limit the mobility of point defects and to increase the likelihood of heterogeneous amorphization. (110) cross-sectional TEM (XTEM) specimens of each of the semiconductors were prepared. All sample preparation procedures were adjusted to minimize specimen heating. A dicing saw was used to cut the materials into thin (200 /xm) strips which were then epoxied together at a temperature of 80 C for 30 min. The strips were lapped to a thickness of approximately pum and placed in an ion mill for their final thinning at 77 K. The microscopy was done on a JEOL 200CX scanning transmission electron microscope (STEM). Micrographs were taken using either a multi-beam (seven beams in the objective aperture) bright-field on axis (110) condition or a bright-field, g 2 2o and g M0, 2-beam condition. Transmission electron diffraction was used to confirm amorphous phase formation in the subsurface layers. It has been shown that by comparing calculated plots of the damage density distribution from Brice 21 (Gaussian), Christel et al. 22 (Boltzmann transport equation), or Ziegler and Biersack 23 (Monte Carlo simulations) with the actual depth of the amorphous layer, it is possible to determine the threshold damage density (TDD) for amorphization. 24 This TDD~value is~simply the amount of energy deposited into elastic collisions necessary for amorphization of the material. The damage density distributions for each of the semiconductor materials were calculated in this study using the Monte Carlo computer program, TRIM'90. Simulations for each material were run for an average of ions at a simulation depth of 1000 A. This depth was used to ensure that all scattering events were included. The damage density distribution was calculated in a graphing program by multiplying the total vacancies per ion by the implantation dose. This value was then multiplied by the semiconductor binding energy (15 ev) and the appropriate conversion factors to yield the units of kev/cm 3. An optical microscope was used to measure the amorphous depth of each material from the XTEM negative. This depth was correlated with the calculated damage density distribution tables to determine the corresponding threshold damage density values. The measured depth varied along the TEM negative to give a range of TDD values. The TDD values were plotted with error bars of 30%, which represented the largest calculated error. Sputtering calculations were conducted to determine if sputtering had an effect on the TDD of the materials, TRIM'90 was used to run the profiles at a simulation depth of 1000 A and at an average of 3400 ions. The total sputtering depth was calculated by using the sputtering yield computed by TRIM'90. For the compound semiconductor materials the sputtering yield for each element was added together to get a total sputtering yield. The total sputtering yield was then multiplied by the implantation dose and divided by the atom concentration in the semiconductor unit cell to yield the sputtering depth. This value was multiplied by the appropriate conversion factors to yield the units of angstroms. The sputtering depth was added to the amorphous depth and the sputter TDD located on the calculated data tables. B. Results Figure 1 shows cross-sectional TEM micrographs of GaP, AlAs, GaAs, and ZnSe in the as-implanted condition. After a 1 x /cm 2 implant, InP was the only material showing the formation of an amorphous layer. All samples did, however, contain a layer of small dislocation loops. The depths of these loops were within a standard deviation of the projected Si + ion range for all cases except the ZnSe. The defect layer in the ZnSe was much deeper than expected. The depth of this defect layer increased with dose. There were fewer loops observed for AlAs than GaAs, which is consistent with its higher TDD values and is also consistent with RIE studies on the effect of increasing Al content in AlGaAs. 25 Increasing the dose to 1 x /cm 2 resulted in the formation of a continuous amorphous layer for all the semiconductors studied except ZnSe. Very little solid phase regrowth was observed in any sample at this dose, except GaSb (Fig. 2). In order to determine the as-implanted amorphous layer depth for GaSb, the extended defect structure was studied more closely. Formation of stacking faults during SPE of compound semiconductors is well known. 12 For GaSb the SPE related stacking faults originate at a well-defined depth (660 A). This depth was chosen to define the depth of the as-implanted amorphous layer. For the remaining samples, the amorphous/crystalline interface depth measured on the micrograph was used. Table I shows the measured amorphous layer depths after a 1 x /cm 2 dose, as well as the TDD values determined from TRIM calculations. For example, the depth of the amorphous layer in GaAs after a 1 x /cm 2 Si + dose was A, which corresponds to a threshold damage density of =3.3 x kev/cm 3. The lower TDD value for GaP (9.1 x kev/cm 3 ), relative to AlAs (7.3 x kev/cm 3 ), can be noted visually in Fig. 1 where, despite comparable densities and average atomic number (and, accordingly, comparable damage density distributions), it is evident that GaP has a lower threshold dam- J. Mater. Res., Vol. 6, No. 5, May

3 a) b) c)»- O.ljjm -«GaP. : -., GaAs AlAs * - ZnSe FIG. 1. Cross-sectional TEM micrographs of several semiconductors after ion implantation with 20 kev Si + at 77 K. The Si + doses were (a) 1 x /cm 2, (b) 1 x /cm 2, and (c) 1 x /cm 2. Two beam bright-field g O 4o TEM imaging conditions of the (110) cross sections were used J. Mater. Res., Vol. 6, No. 5, May 1991

4 a) b) c) 0.1pm -^Surface FIG. 2. Cross-sectional TEM micrographs of GaSb after ion implantation with 20 kev Si + at 77 K. The Si + doses were (a) 1 x /cm 2, (b) 1 x /cm 2, and (c) 1 x /cm 2. Two beam bright-field g 0 4o TEM imaging conditions of the (110) cross sections were used. age density (deeper amorphous layer) than AlAs at both 1 x /cm 2 and 1 x /cm 2 amorphizing doses. Table I shows the results of the sputtering calculations. These results indicate that the amount of sputtering, after a 1 x /ctn 2 dose, was minimal (a maximum of only 12.7 A for ZnSe). Thus, the lack of any SPE regrowth and the prediction of minimal sputtering indicate that the comparison of the effect of composition to the threshold damage density for amorphization after a 1 x /cm 2 dose is relatively accurate. This is plotted in Fig. 3. Figure 3 also shows the effect of sputtering on the TDD values. At this dose, sputtering appears to have little effect on the TDD values. A discussion of the significance of these trends is presented in the next section. Increasing the dose to 1 x /cm 2 resulted in a number of problems that made determination of the relative TDD values less certain. For ZnSe, the dislocation loop concentration continued to increase and no amorphization was ever observed. Figure 2 shows that the GaSb surface appears to have completely decomposed after implantation with a 1 x /cm 2 dose. This phenomenon, which prevented any TDD analysis, may be related to "anomalous disordering". 24 ' 26 The thickness of the disordered layer (-1500 A) appears to be much larger than the ion range (Rp = 275 A). Other researchers 27 have noted difficulty in characterizing high dose implants into antimony based compound semiconductors (GaSb or InSb). Further study of this phenomenon is in progress. For the other semiconductors, implantation at the higher doses resulted in an increased amount of SPE regrowth, resulting in some error in the TDD calculation. SPE was identified by the formation of a large number of stacking faults and a deeper damage layer. Significant SPE was observed in AlAs (-90 A SPE), GaAs (-150 A SPE), InAs (-620 A SPE), and InP (polycrystalline). Little (<30 A) or no SPE was observed for Si and GaP. In addition to increased SPE regrowth upon higher dose implantation, a second error in the TDD calculation may arise from surface sputtering, TRIM calculations suggest that for certain compounds that become amorphous (i.e., GaAs), the sputtered thickness can be as large as =90 A after a 1 x /crn 2 implant. Profilometer studies of implanted and unimplanted regions of the GaAs sample (a high sputtering yield sample) were conducted. The profilometer is sensitive (theoretically) to 20 A step heights. Several runs from masked to unmasked to masked regions were made. These results were inconclusive; however, they did suggest that if sputtering was occurring it was equal to or less than 100 A, which is consistent with the TRIM calculations. For the aforementioned reasons, determining the TDD values at the 1 x /cm 2 dose would be inaccurate except for perhaps Si and GaP. The determination of the relative threshold damage density as a function of composition is much more accurate at a dose of 1 x /cm 2. III. DISCUSSION The TDD value for GaAs was -3.3 x kev/cm 3 after a 1 x /cm 2 Si + dose. This is equivalent to a J. Mater. Res., Vol. 6, No. 5, May

5 TABLE I. Summary of the peak defect positions, the amorphous layer thicknesses, and the threshold damage density as a function of composition for 20 kev Si implants at two doses. Also included is the effect of sputtering on the threshold damage density and a term that is proportional to the activation energy for solid phase epitaxial regrowth. 1 X10 14 atoms/cm 2 1 x atoms/cm 2 Target Peak defect depth (A) Range (A) Straggle (A) Amorphous layer (A) TDD (kev/cm 3 ) Trim depth sputtered (A) TDD with sputtering Term proportional to AEv (ev) Si AlAs GaP GaAs GaSb InP InAs ZnSe No visible defects Amorph. 295 to 305 TDD 1.1 x Damage x x x 10" 3.3 x x 10" 4.2 x 10" 2.4 x >2.0 x x x x 10" 3.0 x x 10" 3.9 x 10" 2.0 x value of 11.7 ev/molecule, which is comparable to previously reported TDD values of 11.2 ev/molecule for Se implantation in GaAs 28 ' 29 and 11.3 ev/molecule for N implantation in GaAs. 19 Opyd et al. 30 reported a value of ev/molecule for Si implants into GaAs at 77 K and a dose of 1 x /cm 2. It is unclear why there is such a large difference between this value and the other three reported values. The SPE observed after a 1 x /cm 2 dose was significant in some cases. SPE was observed in the AlAs, GaAs, and InAs samples implanted with 1 x / cm 2 but not in those implanted with 1 x /cm 2. This indicates that SPE regrowth occurs during the implan- E o, 10 '5> cv a <u CO CO 1 0 Q o Without Sputtering With Sputtering Si AlAs InP InAs GaP GaAs GaSb ZnSe Semiconductor FIG. 3. The threshold damage density for amorphization as a function of composition after 1 x /cm 2 20 kev Si + implantation at 77 K. tation process (ion beam induced epitaxial recrystallization IBIEC), not during the TEM sample preparation or observation processes. This IBIEC in the 1 x / cm 2 samples obviously introduces a large error in the TDD calculations. Therefore, the relative changes in TDD after a 1 x /cm 2 dose are more accurate than after a 1 x /cm 2 dose. One possible source of error in the TDD values would arise if there was a change in density (i.e., swelling) upon amorphization. Because TEM measures the actual amorphous layer thickness, if any of these semiconductors swell upon amorphization the effective threshold damage density would be reduced. Swelling has been previously observed upon amorphization of SiC 38 and to a lesser extent Si. 3 For a /cm 2 dose, SiC swells between 7 and 9% while Si swells 6% (or less). This is comparable to the error in determining the amorphous layer thickness (less than ± 5%). To the best of the authors' knowledge, little or no swelling data exist for the other semiconductors studied, and no density determinations were made in this study. In an attempt to understand better the trends in the threshold damage density values observed after implantation with 1 x /cm 2 Si +, a number of known thermodynamic values were compared as a function of composition. The TDD trends were compared with the melting point, the enthalpy of formation, Pauling's, 31 Van Vechten's, 32 ' 33 and Sanderson's 34 ' 35 bond energy calculations. Swanson et al. 36 have attempted to calculate the amorphization threshold from first principles. A summary of this theory was given by Kelly. 2 A lower limit for the amorphization threshold is defined as being when the fraction of vacancies in the crystal f v times 1052 J. Mater. Res., Vol. 6, No. 5, May 1991

6 the energy stored in a vacancy E v is greater than the change in energy between the crystalline and amorphous states AE c^a. One difficulty with compound semiconductors is that AE c^a is not known. The energy stored in vacancies in the crystal is also not known but would be expected to be proportional to the bond energy. The effect of the entropy makes accurate estimates even more difficult. If the changes in AE c^a were relatively small in magnitude, one might expect the TDD to vary with bond energy (i.e., f v E v ). None of the bond energy estimates predicted the trends or the magnitude of the changes in TDD with composition. This implies that bond energy estimates alone are insufficient and the additional information on the energy difference between the crystalline and amorphous phases may be necessary before these trends can be predicted from first principles. In addition to the bond energy affecting the threshold damage density, it is possible there is a change in the mechanism of amorphization with composition. It is known that amorphization in Si at low temperatures occurs via heterogeneous amorphization for heavier ions. There are recent RBS and optical results published by Wesch et al. 19 that indicate upon room temperature ion implantation, GaAs amorphization with heavy ions occurs via homogeneous nucleation whereas GaP shows an increase in defect concentration more indicative of amorphization via heterogeneous nucleation. The GaAs samples that were implanted at low temperature (40 K) and measured at low temperature (40 K) indicate that for both heavy (Bi) and light (N) ion implantation, amorphization is occurring via heterogeneous nucleation. If the GaAs amorphization mechanism changes from homogeneous to heterogeneous upon cooling and the GaP is already heterogeneous at room temperature, then one can conclude that amorphization of GaP at 77 K probably occurs via heterogeneous nucleation. Thus, Si, GaP, and GaAs appear to be amorphized via heterogeneous nucleation at 77 K. In order to determine if one would expect a mechanism change for AlAs or ZnSe, we examine the activation energy for migration of a point defect. An unusually low activation energy for migration could result in a significant collapse of the amorphous ion track (cylinder) leading to a homogeneous amorphization. This, in turn, would lead to a higher threshold damage density as is observed for Si implanted at room temperature with lighter ions. 8 Licoppe et al? 1 recently showed that the activation energy for solid phase epitaxial regrowth (i.e., the amorphous to crystalline transformation) was directly proportional to a characteristic migration energy AE m for vacancies in covalent crystals given by AE m = M 2 W 2 Da 0 where M is the average mass of the constituent atoms, W-a is the Debye frequency, and a 0 is the lattice parameter. Table I shows the calculated values of AE m for the semiconductors studied. The observation of significant amounts of IBIEC in AlAs, GaAs, InP, and InAs is consistent with the lower predicted activation energies for SPE. With respect to the question of the amorphization mechanism, the predicted SPE activation energy for AlAs is equivalent to GaAs. Thus one might expect AlAs also to be amorphized at low temperatures via heterogeneous nucleation. The activation energy for ZnSe is even higher, implying a strong tendency toward low defect diffusivities also favoring heterogeneous nucleation. However, the ZnSe is sufficiently ionic such that this estimate for covalent crystals may not be accurate. In ZnSe, it is possible some point defect may have unusually high diffusivities which could lead to amorphization via homogeneous nucleation and thus a much higher value of the threshold damage density. It is also possible that energy released upon electron-hole recombination during implantation in this wide band gap (Eg 2.7 ev) semiconductor is assisting the collapse of any amorphous tracks. IV. CONCLUSIONS The effect of composition on the threshold damage density for amorphization has been studied. After a dose of 1 x /cm 2 none of the substrates was amorphized except for InP. After a dose of 1 x /cm 2 all of the semiconductors were amorphized except ZnSe. A higher dose of 1 x /cm 2 resulted in varying amounts of IBIEC (regrowth) in AlAs, GaAs, InP, and InAs and resulted in anamolous disordering of the GaSb sample. As a dose of 1 x /cm 2 resulted in a minimal amount of IBIEC of the amorphous layer (except for GaSb) and a minimal amount of sputtering, a comparison of the threshold damage density as a function of composition upon 77 K Si + implantation was done. The TRIM'90 calculations indicate the TDD values varied from 2.4 X kev/cm 3 for InAs to 7.3 x kev/cm 3 for AlAs. Comparisons with bond energy estimates were unsuccessful in modeling the TDD values as a function of composition. Further work is necessary to explain these trends. ACKNOWLEDGMENTS The authors would like to thank Steve Pearton and Dave Kisker of AT&T Bell Laboratories and Tim Anderson of the University of Florida. REFERENCES 'Kou-Wei Wang, William G. Spitzer, Graham K. Hubler, and Devendra K. Sadana, J. Appl. Phys. 58, 4553 (1985). J. Mater. Res., Vol. 6, No. 5, May

7 2 Roger Kelly, Nucl. Instrum. Methods 182/183, 351 (1981). 3 J. S. Poate and J. M. Williams, Ion Implantation and Beam Processing (Academic Press, New York, 1984). 4 R. G. Elliman, J. S. Williams, W. L. Brown, A. Leiberich, D. M. Maher, and R.V. Knoell, Nucl. Instrum. Methods in Physics Research B19/20, 435 (1987). 5 S. Prussin, David I. Margolese, and Richard N. Tanber, J. Appl. Phys. 57, 180 (1985). 6 L. G. Kirichenko and V. F. Petrenko, Sov. Phys. Solid State 28, 275 (1986). 7 W. P. Maszara and G. A. Rozgonyi, J. Appl. Phys. 60, 2310 (1986). 8 K. S. Jones, S. Prussin, and E. R. Weber, Appl. Phys. A 45, 1 (1988). 9 D. K. Sadana, Nucl. Instrum. Methods in Physics Research B^8, 375 (1985). 10 D.K. Sadana, H. Choksi, J. Washburn, P.F. Byrne, and N.W. Cheung, Appl. Phys. Lett. 44, 301 (1984). "S. S. Kular, B. J. Sealy, K. G. Stephens, D. Sadana, and G. R. Booker, Solid-State Electron. 23, 831 (1980). I2 C. Licoppe, Y.I. Nissim, P. Krauz, and P. Henoc, Appl. Phys. Lett. 49, 316 (1986). 13 J. S. Williams and M.W. Austin, Appl. Phys. Lett. 36, 994 (1980). 14 Y.I. Nissim, L.A. Christel, T.W. Sigmon, J.F. Gibbons, T. J. Magee, and R. Ormond, Appl. Phys. Lett. 39, 598 (1981). 15 C. Licoppe, Y. I. Nissim, and C. Meriadec, J. Appl. Phys. 58, 3094 (1985). 16 S. J. Pearton, J. M. Poate, F. Sette, J. M. Gibson, D. C. Jacobson, and J. S. Williams, Nucl. Instrum. Methods in Physics Research B19/20, 369 (1987). 17 E.A. Dobisz, B. Tell, H. G. Craighead, and M.C. Tamargo, J. Appl. Phys. 60, 4150 (1986). 18 S.A. Schwarz, T. Venkatesan, D. M. Hwang, H.W. Yoon, R. Bhat, and Y. Arakawa, Appl. Phys. Lett. 50, 281 (1987). 19 W. Wesch, E. Wendler, and G. Gotz, J. Appl. Phys. 65, 519 (1989). 20 L. Csepregi, E.F. Kennedy, T. J. Gallagher, J.W. Mayer, and T.W. Sigmon, J. Appl. Phys. 49, 3906 (1978). 21 D. K. Brice, Ion Implantation Range and Energy Deposition Distributions (Plenum Press, New York, 1975), p L.A. Christel, J.F. Gibbons, and S. Mylroie, J. Appl. Phys. 51, 6176 (1980). 23 J. F. Ziegler and J. P. Biersack, The Stopping and Range of Solids (Pergamon Press, New York, 1985). 24 L. A. Christel, J. F. Gibbons, and T.W. Sigmon, J. Appl. Phys. 52, 7143 (1981). 25 S. J. Pearton, W. S. Hobson, and K. S. Jones, J. Appl. Phys. 66, 5009 (1989). 26 Y. A. Danilov and V. S. Tulochikov, Sov. Phys. of Semiconductors 14, 117 (1980). 27 R. G. Wilson (personal communications). 28 T. Sands, D. K. Sadana, R. Grousky, and J. Washburn, Appl. Phys. Lett. 44, 874 (1984). 29 J.C. Phillips and J. A. Van Vechten, Phys. Rev. B 2, 2147 (1970). 30 W. G. Opyd, J.F. Gibbons, J. C. Bravman, and M.A. Parker, Appl. Phys. Lett. 49, 974 (1986). 31 L. Pauling, The Nature of the Chemical Bond (Cornell University Press, Ithaca, New York, 1960). 32 J. A. Van Vechten, Phys. Rev. 187, 1007 (1969). 33 J. A. Van Vechten, Phys. Rev. B 7, 1479 (1973). 34 R.T. Sanderson, J. Am. Chem. Soc. 105, 2259 (1983). 35 R.T. Sanderson, Chemical Bonds and Bond Energy (Academic Press, New York, 1971). 36 M. C. Swanson, J. R. Parsons, and C.W. Hoelke, Radiat. Eff. 9, 249 (1971). 37 C. Licoppe, Y.I. Nissim, C. Meriadec, P. Henoc, and C. danterroches, in Beam-Solid Interactions and Transient Processes, edited by M. O: Thompson, S.T. Picraux, and J. S. Williams (Mater. Res. Soc. Symp. Proc. 74, Pittsburgh, PA, 1987), p J. M. Williams, C. J. McHargue, and B. R. Appleton, Nucl. Instrum. Methods 209/210, 317 (1983) J. Mater. Res., Vol. 6, No. 5, May 1991