Effect of substrate miscut on low-temperature homoepitaxial growth on Si(111) mediated by overlayers of Au: Evidence of step flow

Observations of homoepitaxial growth on low-angle miscut (cid:126)(cid:59) 0.1° (cid:33) Si (cid:126) 111 (cid:33) substrates through an overlayer of Au, together with earlier results on highly miscut Si (cid:126) 111 (cid:33) surfaces, indicate that growth in this system occurs by step ﬂow. The growth temperatures were between 375 and 500 °C. In the optimum range of Au coverage (cid:126) 0.6–1.0 ML (cid:33) , ion channeling measurements yield at best (cid:120) min (cid:53) 5.0%, and cross-sectional transmission electron microscopy reveals stacking faults on (cid:126) 111 (cid:33) planes. Films produced under similar conditions on bare Si (cid:126) 111 (cid:33) substrates are much more defective. On the other hand, the defect density in the present ﬁlms is higher than that in ﬁlms grown on substrates with a higher miscut angle. The improvement in ﬁlm quality resulting from the Au overlayers is attributed to an increase in the diffusion length of the Si adatoms, caused by Au passivation of the Si terraces. It is suggested that Au is more efﬁcient than other overlayers in promoting step ﬂow because Au passivates the Si (cid:126) 111 (cid:33) terraces without passivating the step edges. © 1997 American Institute of Physics. (cid:64) S0003-6951 (cid:126) 97 (cid:33) 02113-X (cid:35)

Observations of homoepitaxial growth on low-angle miscut ͑ϳ0.1°͒ Si͑111͒ substrates through an overlayer of Au, together with earlier results on highly miscut Si͑111͒ surfaces, indicate that growth in this system occurs by step flow. The growth temperatures were between 375 and 500°C. In the optimum range of Au coverage ͑0.6-1.0 ML͒, ion channeling measurements yield at best min ϭ5.0%, and cross-sectional transmission electron microscopy reveals stacking faults on ͑111͒ planes. Films produced under similar conditions on bare Si͑111͒ substrates are much more defective. On the other hand, the defect density in the present films is higher than that in films grown on substrates with a higher miscut angle. The improvement in film quality resulting from the Au overlayers is attributed to an increase in the diffusion length of the Si adatoms, caused by Au passivation of the Si terraces. It is suggested that Au is more efficient than other overlayers in promoting step flow because Au passivates the Si͑111͒ terraces without passivating the step edges. © 1997 American Institute of Physics. ͓S0003-6951͑97͒02113-X͔ Epitaxial growth of thin films using mediating overlayers 1 is of interest because this technique can produce high quality films at temperatures considerably below those required for growth on bare substrates. Another advantage of using mediating overlayers has been shown for homoepitaxial and heteroepitaxial growth where the overlayer can actually alter the growth mode of the process. In particular, heteroepitaxial growth of Ge on Si has been shown to occur by two-dimensional ͑2D͒ island growth rather than by the thermodynamically preferred three-dimensional island growth. [2][3][4][5][6][7][8] For the case of homoepitaxy, growth on Si͑111͒ is inherently difficult at low temperatures. The substrate must typically be held above 700°C 9 during deposition to obtain high quality films of unlimited thickness. At lower temperatures, Weir et al. 10 demonstrated homoepitaxial growth on bare Si͑111͒ for limited thicknesses at growth temperatures of 380-450°C. The films were about 350 Å thick, and the epitaxial quality deteriorated rapidly with distance from the initial interface. Buffer layers predeposited at 700°C were necessary to obtain clean substrate surfaces.
Several studies have shown 5,11-17 that mediating overlayers can be used to improve homoepitaxial growth on Si͑111͒. Most of these studies, however, have been on very thin epitaxial layers ͓1-50 ML of Si, where 1 MLϭ7.84 ϫ10 14 atoms/cm 2 on Si͑111͔͒, and therefore offer only limited insight into development of the microstructure. For Sn and Au overlayers on Si͑111͒, Iwanari et al. 11,12 and Minoda et al., 16 respectively, observed indications of step flow by monitoring growth of the first few monolayers with reflection high-energy electron diffraction and reflection electron microscopy. The development of the microstructure in thicker layers grown through Au on Si͑111͒ was reported in our earlier letter, 17 where we showed that defect-free homoepitaxial films as thick as 5000 Å ͑with no indication that this was an upper limit͒ could be grown at temperatures as low as 450°C by using a 0.6-1 ML Au mediating layer.
The miscut angle of the substrates used in our earlier study was 2.63°. To investigate the importance of step flow, we performed a series of growth experiments on Au-covered Si͑111͒ with low miscut angles, which have larger terrace widths. Samples were obtained from 3-in. n-type wafers, with ϭ0.1-0.4 ⍀ cm and a miscut angle less than 0.1°toward the ͗112͘ direction. The oxide desorption and film depositions were done in an ultrahigh vacuum ͑UHV͒ chamber, with a base pressure of 7.0ϫ10 Ϫ11 Torr. The experimental procedure used has been described in detail elsewhere. 17 Following epitaxial growth, the samples were analyzed by Rutherford backscattering spectrometry ͑RBS͒ using 2 MeV 4 He ϩ , and ion channeling. After RBS, most of the samples were studied by cross-sectional transmission electron microscopy ͑TEM͒.
The substrate temperature and the Au coverage were varied to determine the effect of these parameters on the epitaxial quality of the film. In the previous study, 17 it was found that Si deposition rate had no observable effect on the film quality. Au coverages were in the range of 0.31-1.70 ML. Low-energy electron diffraction ͑LEED͒ and RBS showed the following surface reconstructions before and after Si deposition at 375-500°C: mixed 5ϫ1 and 7ϫ7 for less than 0.4 ML of Au, mixed 5ϫ1 and (ͱ3ϫͱ3)R30°b etween 0.4 and 0.8 ML of Au; only (ͱ3ϫͱ3)R30°͑or 6ϫ6, above 1 ML͒ above 0.8 ML of Au. These observations are in agreement with earlier LEED and scanning tunneling microscopy ͑STM͒ studies. 18 Figure 1 shows four sets of RBS data for deposition at 450°C, at a deposition rate of about 1.0 Å/s. Figure 1͑a͒ is for a control sample with no ͑0.0 ML͒ Au on the surface. The film is highly defective, and, as was seen for the analogous samples with higher miscut angles, 17 the film quality deteriorates rapidly starting at the interface. The present film, however, is much more defective, with min ϭ60%. This poor film quality is confirmed by TEM ͓Fig. 2͑a͔͒, which reveals a high stacking fault density throughout the film. Unlike in the films grown on highly miscut bare Si͑111͒, the a͒ Present address: Central Research Labs, Texas Instruments, Dallas, TX 75265. b͒ Electronic mail: spaepen@deas.harvard.edu stacking faults in the present film do not all begin at the interface; some stacking faults nucleate within the film. Further, several separate crystals can be seen within the film. Even though this sample has the poorest quality of all the ones grown on low-miscut substrates, the film surface is still smooth after 1200 Å of growth. Figure 1͑b͒ is for a 1400 Å film grown with 0.31 ML Au, and resembles that for the film grown with no Au ͓Fig. 1͑a͔͒. The film grown with Au, however, is of higher quality, with min ϭ21%. The aligned spectrum has been multiplied by a factor of 2 to show its similarity to that of the film grown with no Au. The corresponding TEM image ͓Fig. 2͑b͔͒ shows that the film contains stacking faults on ͑111͒ planes. This film has a higher density of stacking faults than a film grown on a higher miscut angle substrate 17 deposited with a lower Au coverage ͑0.15 ML͒ ͑see Fig. 3͒. Figure 1͑c͒ shows the random and aligned spectra for a 1100 Å film grown with 1.0 ML Au. The Au coverage range 0.8-1.0 ML is optimal for both miscut angles. The aligned spectrum shows that the film is somewhat defective, with min ϭ9.2%; the defect density appears constant throughout the film thickness. The aligned spectrum has been multiplied by a factor of 5 to show detail. This quality is considerably lower than that of the corresponding film grown on the highly miscut substrate, which was nearly defect-free ͑see Fig. 3͒. The TEM image of Fig. 2͑c͒ shows a few stacking faults that nucleated at the interface and extend into the film. The surface peak for Au at high energy in Figs. 2͑b͒ and 2͑c͒ shows that in both cases the Au remained at the surface during growth. Figure 1͑d͒ shows the spectra for a 1600 Å film grown with 1.70 ML Au. The high energy portion of the spectra shows that there is significant trapping of Au at the interface and throughout the film. The TEM image ͓Fig. 2͑d͔͒ shows that stacking faults are not present in the Si film, but that Au inclusions cause the large amount of dechanneling in the spectrum of Fig. 1͑d͒, with min ϭ23%. Unlike films grown with 3.0 ML of Au on a more miscut substrate, 17 this film contains some stacking faults near the surface of the grown layer. This further supports the trend, illustrated by Fig. 3: the quality of films on low miscut substrates is much lower than that for the corresponding Au coverages on higher miscut substrates at the same temperature. Figure 3 is a summary plot of min as a function of Au coverage and substrate temperatures ranging from 375 to 500°C, at deposition rates of 1.0-3.0 A/s for films deposited on low-and high-miscut substrates. Open symbols represent data from our previous study 17 and filled symbols represent data from the low-angle miscut substrates used in the present study. The largest error in min values was 0.2% for all samples measured. Note that the trend is the same for both types. In both cases, the minimum is found at about 1 ML Au. In this case, min ϭ5.0% for low-miscut substrates, which is significantly worse than that for films grown on higher-miscut substrates, where min ϭ2.2%. For both miscut angles at low Au coverages, the deposited films are defective because of a high stacking fault density. At Au coverages much above 1 ML, however, the films are poor because of a large amount of dechanneling from Au inclusions within the film. Although at 375°C, min ϭ100% for 1.0 ML Au on the low-miscut ͑and high miscut͒ substrate, TEM reveals that the films are still crystalline, albeit highly defective, and not amorphous.
The main difference between the low-and high-miscut substrates is the width of their terraces: ϳ1700 and 70 Å, respectively, for the ones used in our two studies. The higher density of stacking faults found in films grown on lowmiscut substrates under otherwise identical conditions can be explained by the higher probability of layer nucleation in the time it takes the Si adatoms to diffuse to a step edge for incorporation into the crystal. The difference in crystal quality between the two growth modes may therefore be considered as indirect evidence for step flow as a growth mechanism, since no such difference would be expected if the growth were governed by two-dimensional island nucleation. Although others have shown step flow over larger terrace widths, [11][12][13]16 the films in our studies were grown at deposition rates of у1 Å/s, whereas the other studies used deposition rates of at most 1 ML/min͓ϭ0.025 Å/s on Si͑111͔͒. This difference in deposition rates accounts for the difference in observed terrace widths over which step flow occurs.
The most plausible interpretation of the role of the Au layer in the improvement of the crystal quality, for both miscut angles, over that grown on bare Si is that it increases the diffusion length of the Si adatoms by passivating the Si surface and thereby making the attachment of Si on the terrace less likely. Other overlayers that are known to passivate the Si͑111͒ surface 19 such as As or Sb, however, have been observed to induce 2D island growth, 5,15 and are less effective than Au in improving the quality of the homoepitaxial crystal. This difference may be explained using the proposal by Kandel and Kaxiras 20 that all passivating overlayers increase the surface diffusion length, but that some overlayers such as As and Sb also passivate the step edges, whereas others such as Sn and Au do not. In support of this, Kandel and Kaxiras 20 have made specific arguments for the passivation of ͗112͘ steps by Sb atoms on the Si͑111͒ surface, which makes it more likely for a Si adatom to exchange sites with an Sb overlayer atom on a terrace than with an Sb atom at a step, which would lead to 2D island growth despite an increased Si surface diffusion length.
One author ͑G.D.W.͒ acknowledges support by a Graduate Fellowship from the Office of Naval Research. This work has been supported by the National Science Foundation through the Harvard Materials Research Science and Engineering Center under Contract No. DMR-94-00396.
FIG. 3. Plot of min vs Au coverage at several temperatures, for deposition rates of about 1 Å/s. Open symbols represent data points from films grown on highly miscut ͑2.63°͒ Si͑111͒ ͑Ref. 17͒; filled symbols represent data points from films grown on low-miscut ͑0.1°͒ Si͑111͒ in the present study. The data at 375°C for both substrates have min ϭ100% even with 1 ML Au. These films are still fully crystalline, albeit highly defective.