Raman and ion channeling analysis of damage in ion-implanted Gab: Dependence on ion dose and dose rate

Transcription

1 Raman and ion channeling analysis of damage in ion-implanted Gab: Dependence on ion dose and dose rate U. V. Desnica Ruder Boskovic Institute, Zagreb, Croatia, Yugoslavia J. Wagner Fraunhofer-Institut fur Angewandte Festkorperphysik, Tullasrasse 72, 7800 Freiburg, Germamy T. E. Haynes and 0. W. Holland Oak Ridge National Laboratory, Oak Ridge, Tennessee (Received 2 August 1991; accepted for publication 12 November 1991) Raman scattering and ion channeling techniques were used to investigate the damage in GaAs implanted at room temperature with loo-kev Si+ ions. The ion-induced damage was analyzed for different ion doses and dose rates (current densities). The development of different damage components was monitored by comparing a Raman signal which is specific to amorphization in GaAs to ion channeling results which are sensitive to smallvolume crystalline defects, as well as to amorphous regions. Raman analysis showed that the rate of growth of the amorphous fraction with implant dose was comparable to the growth rate of the total damage as determined by ion channeling. However, while Raman analysis indicated a weak dependence of damage on dose rate, the ion channeling results showed a substantially stronger dependence. These results demonstrate that the damage morphology in GaAs is dependent upon both dose and dose rate, and that the doserate-dependent component of the total damage consists primarily of crystalline defects. I. INTRODUCTION Room-temperature implantation of Si donors is a common way to prepare conductive, n-type layers in GaAs substrates. However, remaining problems include the inability to obtain high electrical activation of the implanted dopants, especially at high doses, along with an extreme sensitivity to certain processing parameters, which results in poor reproducibility of implanted layer characteristics. Recently, the implantation dose rate (i.e., ion current density) was shown to have a significant effect on the electrical activation of Si-implanted GaAs, such that an increase in activation was obtained by using a low ion current density during implantation. This effect has also been correlated with the dose-rate dependence of ion-induced damage measured by ion channeling in the GaAs lattice.3 4 The dose rate was found to have a strong effect on the total damage produced in GaAs over a wide range of ion dose and dose rate. in this paper, results of Raman and ion channeling studies of dose-rate effects on damage formation in Si + -implanted GaAs are presented. Since these two methods have different sensitivities for various types of lattice damage, this comparative study yields new insight into the evolution of damage morphology during ion implantation in GaAs, and specifically the nature of the defects which are responsible for the reported dose-rate effects. II. EXPERIMENT Undoped, commercial GaAs( 100) wafers were implanted at room temperature (23 C!) with loo-kev Si + ions. During implantation, substrates were tilted 7 with respect to the incident beam to minimize channeling ef- fects. Several samples were implanted with different ion doses (4X X 10 cm2) andor different dose rates, i.e., different ion current densities, J ( PAcm ). The dipole-allowed first-order Raman backscattering spectra from the implanted (100) surface were measured at room temperature. The Raman spectra were excited with the 2.57-eV line from a Kr-ion laser. The probing depths for both crystalline and amorphous GaAs regions are shallower than the thickness of the implanted layer, which is approximately 100 nm, so that the undamaged substrate does not contribute to the Raman spectra. The scattered light was filtered and dispersed in a triple monochromator and detected with an intensified silicon diode array at a spectral resolution of 6 cm -. In ion channeling measurements, the implantation-induced damage was determined from (100) axial channeling spectra using 2-MeV He + ions backscattered at 160. The depth profiles of the damage were extracted from the channeling spectra using a self-consistent iterative procedure to calculate the dechanneling background.6 It is well known that some ion implantation damage in GaAs is unstable and anneals at room tempefature over time periods of the order of days. Consequently, all samples in the present study were kept for more than 1 year at room temperature prior to Raman and ion channeling analysis in order to reach a stable damage level. Therefore, the component of the initial implant damage which is unstable at room temperature is not discussed in this paper. III. RESULTS Figure 1 shows dipole-allowed first-order Raman spectra obtained from samples implanted with 3oSi + at doses ranging from 1 X 1Or4 to 3 X 10 cm -. In the unim J. Appl. Phys. 71 (6), 15 March American institute of Physics 2591

2 GaAs: 3oSi+ 100 kev hvl = 2.57eV X(Y Y )K I I 1.25 PAcm* t I t I I I oa RAMAN SHIFT (cm4 1 1 I I I RAMAN SHIFT (cm4 ) FIG. 1. Raman spectra of IOO-keV Si + -implanted GaAs, obtained using 2.57-eV excitation polarized along the (110) direction, without polarization discrimination for the scattered light. Implantation doses are indicated on the figure; the implant current density was The Raman spectrum of unimplanted GaAs is shown for a comparison. planted sample, only a sharp longitudinal optical (LO) phonon line at 292 cm - is present. This line is due to phonon scattering in defect-free, crystalline GaAs. As the implantation dose is increased, the LO peak is broadened and shifted toward lower frequencies. The LO peak is present in the spectra up to an ion dose of 1 X loi5 cm -. In addition to the LO peak, spectra from implanted samples also contain three broadbands which are known to originate from amorphized regions in GaAs8 At a dose of 3X 10 cm-, the LO peak has vanished completely and only the broadbands remain, indicating a com$letely amorphized surface layer. This result is in excellent quantitative agreement with the findings of Wagner and Fritzsche,g who also found that the surface layer implanted with lookev Si ions was fully amorphiied for doses above 1 X 10 cmw2. Dose-rate effects on the Raman spectra are illustrated in Fig. 2. This figure shows spectra from samples implanted with a single ion dose of 2 X 1014 cm - 2, but at different ion current densities. The intensity of the LO peak, relative to that of the broadbands, is reduced for implantation at higher-dose rates. This indicates that higher-dose rates produce greater disorder, consistent with previous reports based on ion channeling.223 To quantify changes in the Raman spectra of ion-implanted GaAs induced by both the variation of the ion dose and dose rate, the procedure proposed by Wagner and Fritzscheg was applied. Accordingly, the ratio, Rd, of the scattering intensity, I(a), at 250 cm - [which is the maximum of the TO phonon band in completely amorphized FIG. 2. Raman spectra from GaAs implanted with a fixed ion dose of 2 X 1014 cm - * at various indicated dose rates. GaAs (Ref. 8)] to the intensity, 1 (LO), at the maximum of the LO phonon peak was taken as a semiquantitative measure of the implantation-induced damage detected by Raman, i.e., of the relative volumes of the crystalline and amorphized regions. A summary of the results for different ion doses as well as for various dose rates (at a fixed dose of 2~ 1014 cm -*) is presented in Fig. 3. The amount of damage detected by Raman increases slowly with the dose until a large fraction of the implanted layer becomes amorphized, after which the ratio Rd increases sharply. In DOSE ( Sicm2) FIG. 3. Intensity ratio of allowed first-order Raman scattering Rd = Qa)N(LO), where I(4) is the scattering intensity at 250 cm-, and I(L0) is the intensity of the LO peak at 283 cm-. Different ion doses and dose rates are depicted with different symbols J. Appl. Phys., Vol. 71, No. 6, 15 March 1992 Desnica et a. 2592

3 DEPTH (nm) I -I- 1 -~ -T kev Si - GaAs(100) I- -2 2ooo 5 2 looo m ENERGY (MeV) FIG. 4. Ion channeling spectra from the same samples as in Fig. 2 as measured just prior to Raman analysis. The spectrum from a randomly oriented crystal and a channeling spectrum from an unimplanted crystal are shown for reference. Spectra were obtained using a 2.0-MeV 4He+ beam detected at a 160 scattering angle. The depth scale is shown for scattering from As. Fig. 3, as in Fig. 2, dose-rate effects are also clearly visible; Rd increases as J increases. The depth-integrated total damage, Ndr was determined from ion channeling spectra (Fig. 4) obtained from the same set of samples. The dependence of Nd on dose and dose rate is presented in Fig. 5. Comparison of these results with the Raman results (in Fig. 3) shows that the damage detected by both techniques has a similar dependence on ion dose. For instance, in ion channeling measurements, as in the Raman results, a very strong increase of the damage is observed for doses of 10 cm - 2 and above. At a dose of 3X 1015 cmu2, ion channeling also indicates that the implanted layer is completely amorphous (the channeling I I I 10'4 10'5 DOSE (Si cm* 1 FIG. 5. Integrated damage yields, Nd, derived from ion channeling spectra, as a function of dose for loo-kev Si + implants into GaAs at various dose rates. Same samples as in Figs. l-3. Symbols denoting specific ion dose and dose rate are the same as in Fig P ; 1.5 I.2 s g 1.0 E H Z~ r,q I RBS DAMAGE IlO cm* ) FIG. 6. Correlation between Raman intensity ratio, R,+ with the total damage measured by ion channeling, Nd, for different ion doses and various dose rates. Symbols correspond to those on Figs. 3 and 4. yield was equal to the yield from a randomly oriented sample), consistent with the Raman results. However, the dose-rate effects are much more pronounced in the ion channeling data than in the Raman results, i.e., Nd increases much more dramatically with dose rate (Fig. 5) than does Rd (Fig. 3). In order to make a quantitative comparison between the Raman scattering and ion channeling results, Rd is plotted against Nd in Fig. 6. This clearly establishes the correlation between the amount of implantation-induced damage detected by these two complementary methods. Two different correlations are observed depending upon whether the ion dose or the dose rate is the principal variable. For a given increase in Ndl the increase in Rd is much greater with dose than with dose rate. The correlation is approximately linear in either case. IV. DISCUSSION This study focuses on a comparison of measurements of the implantation-induced damage as determined by firstorder Raman backscattering and ion channeling. The results for both characterization techniques have shown similar general trends, especially regarding the increase of the damage with the ion dose, but this comparison also re vealed a considerable disagreement in the quantitative evaluation of dose-rate effects. This is due to the different sensitivities of the two methods for various types of lattice damage, as outlined below, leading to some general conclusions regarding the nature of the dose-rate-dependent component of damage, and also the dose-dependent component. The main features of the Raman backscattering spectra shown in Figs. 1 and 2 are well understood. First, the single, sharp line at 292 cm - is due to the longitudinal optical phonon at the I point, i.e., the center of the Brillouin zone. As expected, the first-order Raman backscattering spectrum from the (100) surface of the unimplanted sample shows only this sharp line. However, 2593 J. Appl. Phys., Vol. 71, No. 6, 15 March 1992 Desnica et a. 2593

4 changes appear in the spectra following implantation, as defects are introduced into the lattice. These defects destroy the translational symmetry of the lattice, lowering the symmetry from the crystal space group to the point group of the lattice site, thus relaxing the selection rules. The LO phonon line, therefore, is shifted in frequency and broadened as a result of either these ion-induced defects and the associated lattice strains, 2 or the change of the phonon localization length.13 Besides the broadening and shifting of the LO mode to lower frequencies, additional modes also appear in the spectrum, in the form of broadbands at lower frequencies. These bands are assigned to disorder-activated, first-order scattering, with three maxima arising respectively from scattering by transverseacoustic phonons at the L and X points, from IIA phonons near the K point, and from transverse-optical phonons at the l?, X, and L points.819 These bands, all having maxima at relatively low frequencies, are understood to originate from amorphized regions.8 In contrast, crystalline regions, even as small as 5 nm, are known to produce only modes with frequencies larger than 280 cm -l.13 Hence, the change in the dipole-allowed, first-order Raman spectra shown in Figs. 1 and 2 are interpreted primarily to result from increases in the amorphized fraction of the implanted layer at the expense of the crystalline fraction. The measured Raman signal is a superposition of two components, the LO peak and the broad bands, which are essentially independent, and these components contribute to the Raman spectrum in proportion to the total volume of the crystalline and amorphous regions contained within the laser probe depth. On the other hand, ion channeling measurements are sensitive to atoms displaced from lattice sites in directions perpendicular to the direction of the probing beam ( (loo> in the present study). Such displacements are associated with both crystalline defects (i.e., small clusters of point defects, or extended dislocations) and amorphous regions. Therefore, damage profiles determined by ion channeling are generally a good estimate of the total damage, although the sensitivity to a particular type of defect is dependent on the detailed nature of the displacements comprising the defect, as well as on the total number of atoms involved. Therefore, both the first-order Raman scattering and ion channeling results are sensitive to amorphous regions formed during ion implantation, but only the ion channeling data can detect small volume crystalline defects. According to the above interpretation of the Raman spectra, two conclusions follow immediately from the data in Fig. 3. First, the amorphous fraction clearly increases with ion dose, and this increase becomes particularly strong above 1 X 10 cm -. Secondly, the amorphous volume also increases with dose rate, although the net change in the amorphous volume is small over the range of dose rates considered. Two additional conclusions may be derived from Fig. 6. Since the integrated damage yield Nd determined by ion channeling is a measure of the total damage in the implanted layer, the correlation of Rd, which measures the amorphous fraction, implies that the amorphous volume grown uniformly as damage accumu- lates due to either increasing the dose or dose rate, i.e., there appear to be no thresholds or transitions in the mode of amorphization over this range of doses and rates. Finally, the relatively small increase of Rd with Nd for the variable dose-rate series of samples (Fig. 6) implies that the large majority of the dose-rate-dependent component of the damage is due to the formation of crystalline defects (which are not detected by the present Raman measurements) and only a small minority of this damage is due to amorphization. Implantation-induced damage is generally produced either by heterogeneous processes, in which damage nucleates within the volume of the collision cascade during quenching, or by homogeneous processes, in which the damage forms through interactions between uniformly distributed, mobile defects which have escaped from the cascade volume. Either or both of these types of processes may produce dose-dependent damage. However, for typical ion current densities the probability of cascade overlap during quench is negligible and dose-rate-dependent damage is the result solely of homogeneous nucleation, and occurs when the mobile, primary defects are not stable. Based on the present experimental results, some aspects of damage formation in GaAs can be described within this framework Foremost, these results indicate that homogeneous nucleation in GaAs produces mainly drystalline defects, such as small point-defect clusters of dislocation loops. Therefore, dose-rate effects are mainly observed by ion channeling, and only secondarily by Raman scattering. Consequently, over the range of doses and rates used, amorphization in GaAs during room-temperature implantation of Si is primarily a heterogeneous process, which may occur either directly during single ion impacts, or through nonsimultaneous overlap of multiple cascade volumes. l4 V. CONCLUSION The damage produced in GaAs by Si + implantation for different doses and at various dose rates was characterized by Raman scattering and ion channeling techniques. First-order Raman spectra reveal primarily the degree of amorphization, while ion channeling detects point defects and small-volume crystalline defects with greater sensitivity. Raman measurements showed that the amorphous fraction increases uniformly with ion dose until the implanted layer becomes completely amorphized at a dose of 3 X 10 cm -. Raman scattering results also indicated an increase of amorphization with increasing current density at a fixed dose. However, ion channeling measurements detected a much greater relative increase of damage as a function of dose rate. Therefore, the main effect of increased dose rate on damage accumulation during roomtemperature implantation of GaAs with 100-keV Si ionsi is to promote the formation of small-volume, crystalline defects by the coalescence of mobile, primary defects J. Appl. Phys., Vol. 71, No. 6, 15 March 1992 Desnica et al. 2594

5 ACKNOWLEDGMENTS Research sponsored in part by the Division of Materials Sciences, U.S. Department of Energy under contract No. DE-AC05-840R21400 with Martin Marietta Energy Systems, Inc. For exampie, R. N. Thomas, H. M. Hobgood, G. W. Eldridge, D. L. Barrett, T. T. Braggins, L. B. Ta, and S. K. Wang, in Semiconductors and Setnitnefals (Academic, New York, 1984), Vol. 20, p. 1. F. G. Moore, H. B. Dietrich, E. A. Dobisz, and 0. W. Holland, Appl. Phys. Lett. 57, 911 (1990). T. E Hayncs and 0. W. Holland, Appl. Phys. Lett. 58, 62 (1991). 4T. E. Haynes and 0. W. Holland, Nucl. Instrum. Methods Sect. B 5960, 1028 (1991). In this work loo-kev Si ions were used exclusively. However, it would be interesting, from both a practical and fundamental perspective, to determine whether rate effects are energy dependent. Some preliminary channeling measurements of the damage growth during implantation of Si at 170 kev have indicated an identical dose rate effect at this higher energy (T. E. Haynes, unpublished data). Therefore, dose-rate effects in GaAs do not appear to be strongly energy dependent over this range, and the conclusions made in this paper will very likely hold for Si implants at ion energies throughout the range of kev. F. H. Eisen, in Channeling, edited by D. V. Morgan (Wiley, New York, 1973), pp G. Carter, M. J. Nobes, and I. S. Tashlykov, Radiat. Eff. Lett. 85, 37 (1984). T. Nakamura and T. Katoda, J. Appl. Phys. 53, 5870 (1982). 9J. Wagner and C. R. Fritzsche, J. Appl. Phys. 64, 808 (1988). One has to be aware that Z(a)Z(LO) ratio does not give directly the ratio of volumes of amorphous and crystalline regions. For quantitative analysis, a precise decomposition of the Raman spectrum into two components should be made. Furthermore, corrections should be made for the scattering strengths of both phases, as well for a number of other factors, including the difference in penetration depth of the probing beam in amorphous and crystalline GaAs, the difference of reflection coefficients, etc. However, the ratio Z(a)Z(LO) is a reasonable parameter for assessing trends during processing of crystals, such as by ion implantation or damage annealing. B Weinstein and M. Cardona, Phys. Rev. B 5, ( 1972). Burns, F. H. Dacol, C. R. Wie, E. Burstein, and M. Cardona, Solid State Commun. 62, 449 (1987). I3 K. K. Tiong, P. M. Amirtharaj, F. H. Pollak, and D. E. Aspens, Appl. Phys. Lett. 44, 122 (1984). 14T, E. Haynes and 0. W. Holland, Appl. Phys. Lett. 59, 452 (1991) J. Appl. Phys., Vol. 71, No. 6, 15 March 1992 Desnica et a. 2595

6