The formation of a continuous amorphous layer by room.. temperature implantation of boron into silicon

Size: px

Start display at page:

Download "The formation of a continuous amorphous layer by room.. temperature implantation of boron into silicon"

Jeremy Watson
6 years ago
Views:

1 The formation of a continuous amorphous layer by room.. temperature implantation of boron into silicon K. S. Jones Department a/materials Science and Engineering, University of Florida, Gainesville, Florida D. K. Sadana IBM T. J. Watson Research Center, Yorktown Heights, New York S. Prussin TR W, Inc., Redondo Beach, California J. Washburn and E. R. Weber Department of Materials Science and Mineral Engineering, University of California, Berkeley, California W. J. Hamilton Jet Propulsion Laboratory, California Institute aftechnology, Pasadena, California (Received B May 1987; accepted for publication 27 October 1987). Ion implantation of 60 ke V boron into {100} silicon at medium beam currents (150 tta) was performed at K over the dose range from 1 to 8 X /cm 2 Diffraction contrast and high-resolution phase contrast transmission electron microscopy (TEM) were used on planview and 90 cross-section samples to study the formation of a continuous amorphous layer as a function of increasing dose. Our TEM results show that, unlike implantation of with heavier ions where amorphization initially occurs at or around the projected range, the amorphization by high dose ( > 5 X 1016/cm2) B-+- -implanted first occurs at and/or near the surface. It is proposed that the buildup of a high concentration of vacancies which inevitably occurs near the surface during high-dose n-' implantation is primarily responsible for the observed near-surface amorphization. Based on the results of this investigation and those available in the published literature, it appears that low temperature (slow recombination rate for point defects) and high beam current (high generation rate for point defects) implantation may result in the optimum conditions for amorphous layer formation with boron. I. INTRODUCTION The formation of a continuous amorphous layer during ion implantation of silicon is desirable because significant activation of the implanted dopant upon solid-phase-epitaxial crystallizaion results even at low temperatures (550 C).1.2 Room-temperature implantation of silicon with boron has not generally been successful in producing a continuous amorphous layer 3 except at very high beam currents (2 ma).4.5 nce heavier ions do result in amorphous layer formation, it is of interest to understand the differences between light- and heavy-ion amorphization. The crystalline-to-amorphous conversion has been shown 4 6,7 to be a first-order phase transformation, not just a gradual accumulation of point defects. Among the evidence supporting this is the abrupt amorphous/crystalline interface observed with high-resolution transmission electron microscopy (TEM). When compared with computer-simulated images, this result indicates that the crystal relaxes to an amorphous state when the point defect concentration reaches a critical value. 3,6,7 The experimental results which follow will confirm the first-order phase transformation nature oftlle crystalline-toamorphous conversion for boron implanted at room temperature. However, the location of the amorphous layer is completely different from any previously reported amorphous layers produced by ion implantation. Thus, changes in the models describing this system 5 6 must be made with respect to the source of the amorphous layer nuclei as well as the point defects primarily responsible for amorphous layer formation. I!. EXPERIMENT Czrochralski-grown (loo) silicon n-type 3-in. wafers, 5-10 n cm, were implanted with 60 key llb+ at doses between 1 and 8X 1016/cm2. The implantations were done using a Varian medium current ion implanter model CF at a constant beam current of 150 pa (beam density of 0.1 W/cm2). A Waycoo1 8 endstation was used in which circulating room-temperature freon stabilizes the wafer temperature and prevents excessive heating during implantation. Eight wafers were first implanted with a 1 X 1016/cm2 dose of boron, after which one was removed for the TEM sample preparation and the rest were implanted with another dose of 1 X 1016/cm2 fohowed by another wafer being removed, and so forth, until a cumulative dose of 8 X / cm 2 was achieved. This further reduced the amount of any heating to that of a 1 X /cm 2 dose. Plan-view and cross-sectional TEM specimens of the unannealed specimens were prepared. The plan-view samples were chemically thinned from the backside in the standard manner, while the cross-sectional samples were thinned by ion milling with argon at 5 kv J. Appl. Phys. 63 (5), 1 March i 1988 American Institute of Physics 1414

$endstation), cross-sectional TEM, bright field, Ibo The microscopy was done on either a Philips 301 (diffraction contrast results) or a modified Philips 400 (high resolution, phase contrast results)$

2 FIG key lib+, medium current (150 pa), room temperature (Waycoo! endstation), cross-sectional TEM, bright field, Ibo The microscopy was done on either a Philips 301 (diffraction contrast results) or a modified Philips 400 (high resolution, phase contrast results) electron microscope. All samples were examined in the as-implanted condition, and no post-implant annealing was done. m. RESULTS AND DISCUSSION Figure 1 shows cross-sectional TEM micrographs of the near-surface region as a function of increasing IIB+ dose. For doses of 5 X 1016/ em 2 and less, the only con tin uous amorphous layer which forms is a very thin layer ( 200 A) at the surface. Above 5 X 1O!6/cm2 a continuous amorphous layer forms which is 1100 A thick. The thin amorphous layer which forms on the surface (>4X 1016/cm2) may arise from oxygen knock-on from the native oxide on the surface. Although boron is not as efficient as heavier ions at producing oxygen recoils due to its low mass, it may still occur due to the high doses being studied. The diffraction pattern ofthe thick amorphous material (>6 X /cm 2 ) shows the usual Debye-Scherr ring pattern characteristic of amorphous ma- terials. The lower-dose sample (5 X 1016/cm2) shows large regions of varying contrast below the thin surface amorphous layer which form a continuous amorphous layer after the next implantation. Diffuse dark-field TEM of this sample indicates that the regions of varying contrast contain pockets of amorphous material. The amorphous layer which forms at the higher dose (6X /cm 2 ) is shallower than the peak of the Brice 9 damage density distribution curve, D p' This is quite different from amorphization by heavier ions which is observed to form first at Dp and to subsequently grow deeper and toward the surface with increasing dose. High-resolution TEM micrographs of the regions indicated in Fig. 1 appear in Fig.2. The upper amorphous/crystalline (alc) interface region in Fig.2(a) confirms the diffuse dark-field results, showing the regions of varying contrast to be large amorphous pockets within the crystalline material. High-resolution TEM of the lower amorphous/crystalline interface (6X /cm 2 ) is shown in Fig. 2(b). Again, the interface is atomically abrupt when observed edge-on, confirming that the phase change from crystalline to amorphous material is a first-order transforma- ale IntE.'rfac:e ale Interfdce Cryst.al] Stacking Fault Pocket CI'ystell ine 3i (oj (bi FIG. 2. High resolution TEM, multibeam condition, 60 key!lb'!; A: 5X lo'6(cm 2, B: 6X IO'6(cm J. Appl. Phys., Vol. 63, No.5. 1 March 1988 Jones etal. 1415

3 key i I I j 1Il 'i: '2 '2.S!.2 c( Il.i: a..i:; 8 :> ;; I :> >- >- l >- l- S l- S l- S iii en iii z z z w w w c w W III 0 C! C! 04 J <t <t ::.; :;: :;: <: :1 <: :1 "" C I c C DEPTH (10:3 A) Medium Curren! Low Cllrreni (150 gal (2 300K 77K High Currenl (2 rna) 3MK FIG. 3. Damage density distribution and corresponding amorphous/crystalline interface depth for dill:'erent 11 B + implant conditions. tion. In the region beneath the a/c interface an extrinsic stacking fault can be seen. In Fig. 1, it was shown that increasing the dose from 6 to 8X IO l6 /cm 2 resulted only in a slight increase in the thickness of the amorphous region ( < 100 A). This may be due to overlap of amorphous pockets within this distance of the a/c interface [Fig. 2(b) 1 which still existed after the production of a continuous amorphous layer. The region well below ( A) the a/c interface for the higher-dose samples (;;.;.6X /cm 2 ) contains a large number of extrinsic stacking faults and {lo} defects. Figure 3 shows a comparison between the damage density distribution curves for the three conditions (one from this investigation and the other two from Shih) where a continuous amorphous layer has been produced upon B implantation of. Figure 3 (a) shows the damage density distribution curve (Brice) for our sample which was at 60 key, The a/c depth reflects the medium beam current (265 j.ta), room-temperature implant of dose 6 X 016/cm 2 in Fig.!. Figure 3(b) is the Brice curve for a l00kevimplant where a low beam current (2 pa), liquid-nitrogen implant of dose 1 X 1016/cm 2 resulted in an amorphous layer extending to the depth indicated. Figure 3(c) is the Brice curve for a 35 ke V boron implant. The high beam current ( - 2 rna) roorntemperature implant of dose 5 X cm 2 resulted in the ole depth shown. It was reported that the maximum temperature ofthe wafer during this implant was 35 C. 4 The integration of the area under the curve down to the a/c depth yields a value of the total energy deposited into elastic collisions which is necessary to amorphize the silicon. Figure 4 is a schematic graph of the total energy (kev I cm 2 ) deposited into elastic collisions necessary to create a continuous amorphous layer for the three conditions discussed in Fig. 3. They axis of Fig. 3 is in ev IA ion, thus the integrated value must be multiplied by the dose (previously mentioned) to yield Fig. 4. The value calculated for the medium current room-temperature implantation (our results) is the minimum amount of energy which can be deposited into the target and still form a continuous amorphous layer as the formation of an amorphous layer under these conditions was studied in depth. It is also the highest value of the three conditions compared. The value of 2 X 1017keV Icm 2 for low current implantation at 77 K is sufficient to create a continuous amorphous layer under these implant conditions. 4 nce these conditions were not studied by us, a lower dose and subsequently lower-energy density for amorphization may be sufficient. The energy density value for amorphization by high current, room-temperature implantation may also be less. These results imply that a high current, 77 K S.2, ;;> 4.0 -'"" - S2 " 3.0 >, 2!' Jl "" 2,0 <5 0- "0 l:! M,g 10 mi!limum Medium Current Low Current High Current High Current 150f(A 2mA 17K 300K 77K 300K Implantation Conditions FIG. 4. Integrated deposited energy values for conditions of observed amorphous layer formation J. Appl. Phys., Vol. 63, No.5, 1 March 1988 Jones etal. 1416

4 implant may be very efficient at creating a continuous amorphous layer. Previous publications 10 have addressed the concept of a threshold damage density or the damage density necessary for amorphization. Use of Brice's damage density distribution curves along with knowledge of the effective threshold damage density, from cross-sectional TEM or channeled RBS data, allows one to predict the type (buried, surface) and depth of the amorphous layer, if any, which is produced. Unfortunately, the threshold damage density obtained this way has not been very accurate in predicting the critical dose for amorphization for light as well as heavy ions simultaneously. The effective threshold damage density is a strong function of both ion mass and wafer temperature during implantation. The dependence on ion mass is known to be much greater for room-temperature implants than for implants at 77 K.7.11 The threshold damage density value for room-temperature implants increases as the mass of the ion decreases. Thus, for heavier ions (>27 Al +) the amorphous layer is usually observed to be centered around the peak of the damage density distribution, DD' when it first forms. For light-ion amorphization, it previous modei 3 12 suggests that the embryo of amorphous phase formation is a divacancy-di-interstitial pair which forms at the end of the light-ion track. These embryos act as nuclei for amorphous phase growth which can occur when the point defect concentration reaches a critical value. This implies that amorphous layer formation would first occur at the depth where the concentration of interstitials and vacancies is a maximum, i.e., >Dp. The low implant temperature or high-dose rate data is not in direct contradiction with this model since the amorphous layer extends below Dp. The location of the initial amorphous material produced by the other implant conditions was not reported. However, our room-temperature, medium current implants show amorphous layer formation occurs at a depth shallower than Dp [Fig.3(a)]. The amorphous layer is observed to form by the growth of individual amorphous pockets upon increasing dose. The growth of the amorphous pockets is consistent with Model B proposed by Shih et al and stated above. However, the shallower than Dp location of the initial amorphous regions suggests the nuclei cannot simply be divacancy-di -interstitial pairs forming at the end of the light-ion track. We believe that the shallower location of the amorphous layer is related to the high concentration of vacancy complexes which are expected to be present in the near-surface region after the high-dose B + implantation. Vacancy clustering near the surface has been shown to result in the formation of stacking fault tetrahedra during annealing. 13 On the other hand, the presence of interstitials in deeper regions (Fig.2(b)] results in the formation of extrinsic stacking faults. These are believed to act as sinks for excess interstitials preventing amorphous layer formation in much the same manner as occurs for metals during implantation where dislocation loop growth is preferred to amorphization. 14 The mobility of vacancy complexes versus interstitial complexes may be sufficiently lower that they are less likely to form extended defects which can reduce the strain energy associated with the high dose of boron. The initial amorphization of the surface [Fig.2(a)J can arise from oxygen recoils from the native oxide at the surface. No special cleaning was done to the wafers prior to or between the various implantation steps therefore the implantations were done through a thin native oxide on the surface. The TRIM (Monte Carlo-type) calculations (assuming 30 A of native oxide) indi.cate that a high concentration of recoiled oxygen (10 21 _10 22 / cm 3 ) is present in the surface region of after the B + implantation. phase growth is believed to occur when the concentration of point defects around the nuclei reaches a critical value. It is proposed that the reason the damage density necessary for amorphous layer formation is greater for room-temperature, medium current boron implantation (Fig.4), is related to the concentration of point defects at any given time. The concentration of point defects around the nuclei is proportional to the generation rate of point defects minus the recombination rate of point defects. The generation rate (proportional to the beam current) is lower for the medium beam current implants than the high beam current implants. The recombination rate (related to the implant temperature through the dissociation of point defect clusters and the diffusion of point defects) is greater for room-temperature implants than 77 K implants. From Fig.4, it appears that the decrease in recombination rate between 300 and 77 K is greater than the decrease in generation rate between 150 J.l and 2 pa, since the total energy density necessary to create an amorphous layer to the indicated depths is less for the 77 K implant than for our 300 K implant. For our implant conditions, the total concentration of point defects at any given time is lower for room-temperature, medium current implants than it is for the other two cases discussed and the total damage density for amorphous layer formation must be greater. nce the observed amorphization does not follow previously verified models, the question arose: could the observed phenomena be attributed to knock-on of Fe from the slits of the implanter? This is not predicted since boron is much lighter than iron. Detailed SIMS analysis showed that no i.ron could be found in the implanted region. There was also some concern that the ion milling process might playa role in the observed amorphous layer formation. However, examination around the hole of the chemically thinned planview samples revealed a thick amorphous layer on the surface of the specimen. The results illustrated in Figs.3 and 4 highlight an additional point. Under high ftuence conditions the dynamic concentration of interstitials and vacancies is increased and so the amorphization is increased. On the other hand, beam heating of the wafer results in increased mobility of point defects which decreases amorphization. Competition between these two processes results in a wide a/ c transition region associated with the high current room-temperature implantation. As suggested from FigA a high current implant at 77 K may result in the amorphization at a low dose and a relatively narrow ale transition region. This lower value for the effective threshold damage density would result in more of the final boron distribution residing within the amorphous layer which may yield higher electrical activa J. Appl. Phys., Vol. 63, No.5, 1 March 1988 Jones eta/. 1417

5 tion upon solid-phase-epitaxial growth of the amorphous layer. Also, at the lower doses, the likelihood of precipitate formation upon annealing is reduced, IV. CONCLUSIONS Medium current room-temperature implantation of {loo} silicon with 60keV 11 B+ results in the the formation of a continuous amorphous layer after a dose of 6X 1016/cm 2 The initiation of amorphization in the near-surface region rather than at D p does not conform to any previous amorphization models. It is proposed that vacancy complexes near the surface are primarily responsible for the observed shallow amorphization. The formation of a continuous layer is preceded by the growth of individual amorphous pockets and the sudden crystalline-to-amorphous transformation of the remaining crystalline region with increasing dose. The fact that the damage density necessary for amorphous layer formation is greater for room-temperature, medium beam current implant conditions than for high current or lowtemperature implants is explained by the competition between the generation and recombination rates of point defects which exist at any given time. It is proposed that high beam current implantation at 77 K may be the most efficient means of forming a continuous amorphous layer by implantation of boron. ACKNOWLEDGMENTS This project was supported by TRW Inc. and the versity of California M.LC.R.O. Program. '"9. L. Crowder and F. F. Morehead. Jr. Appl. Phys. Lett. 14, 313 (l969). <D. K. Sadana, J. Washburn, and C. W. Magee, J. Appt Phys. 54, 3479 (1983). 3y. Shih, J. Washburn, E. R. Weber, and R. Gronsky, Mater. Res. Soc. Symp. Proe. 45, 65 (1985), 4y' Shih, Ph.D. dissertation (University of California, 1985). 5J. F. Gibbons, in Lectures on Ion Implantation and Proton Enhanced Diffusion, University of Tokyo, Tokyo, Japan (Japan Society of Applied Phys., Tokyo, 1977). 6J. Washburn, C. S, Murty, D. K. Sadana, P. Byrne, R. Gronsky, N. Cheung, and R. Killaas, Nue!. lnstrum. Methods 209,345 (1983). 7J. Narayan, D. Fathy, O. S. Den, and O. W. Holland, Mater. Lett. 2, 211 (1984). 'Varian/Extrion Div., Gloucester, Ma D. K. Brice, in Ion Implantation Range and Energy Deposition Distribution (Plenum, New York, 1975), Chap.!. los. Prussin, D. 1. Margolese, and R. N. Tauber, J.AppI.Phys. 57, 180 (1985). It.F. F. Morehead and B. L. Crowder, Radiat.Eft'. 6, 27 (1970), 12y' Shih, J. Washburn, R. Gronsky, and E. R. Weber, Lawrence Berkeley Laboratory Rep. No (1986), "'D. K. Sadana, J. Washburn, and G. R. Booker, Philos.Mag. B 46, 611 (1982). gg. Carter and W. A. Grant, Ion Implantation of SemIconductors (Halstead, New York, 1976), p J. Appl. Phys., Vol, 63, No.5, 1 March 1988 Jones eta!. 1418

Jr data for 3 x 1013 cm - 2

Jr data for 3 x 1013 cm - 2 Jr 3 Progress on two publications is described. Problem Areas A great deal of time is being required to get RBS and XTEM samples prepared and measurements made on busy equipment. Therefore results are