Atomistics of Ge deposition on Si(100) by atomic layer epitaxy

(1)

Atomistics of Ge Deposition on Si(100) by Atomic Layer Epitaxy

D.-S. Lin,1,* J.-L. Wu,1_{S.-Y. Pan,}1_{and T.-C. Chiang}2

1_{Institute of Physics, National Chiao-Tung University, 1001 Ta-Hsueh Road, Hsinchu 300, Taiwan, Republic of China} 2_{Department of Physics, University of Illinois at Urbana-Champaign, 1110 West Green Street, Urbana, Illinois 61801-3080}

and Frederick Seitz Materials Research Laboratory, University of Illinois at Urbana-Champaign, 104 South Goodwin Avenue, Urbana, Illinois 61801-2902

(Received 7 October 2002; published 29 January 2003)

Chlorine termination of mixed Ge=Si100 surfaces substantially enhances the contrast between Ge and Si sites in scanning tunneling microscopy observations. This finding enables a detailed inves-tigation of the spatial distribution of Ge atoms deposited on Si(100) by atomic layer epitaxy. The results are corroborated by photoemission measurements aided by an unusually large chemical shift between Cl adsorbed on Si and Ge. Adsorbate-substrate atomic exchange during growth is shown to be important. The resulting interface is thus graded, but characterized by a very short length scale of about one monolayer.

DOI: 10.1103/PhysRevLett.90.046102 PACS numbers: 68.35.Dv, 68.35.Fx, 71.15.Nc, 79.60.–i

A detailed understanding of the atomic processes of deposition is essential for the synthesis and manufac-turing of nanoscale materials and devices. A system that has attracted much attention in the past decade is the growth of Ge on Si(100) [1– 4]. It is of technological interest because of its potential for optoelectronics, wire-less and broadband communication, and computing ap-plications [5]. There is also much scientific interest because it is a simple model system suitable for funda-mental investigations, both experifunda-mental and theoretical [6,7]. Indeed, this system has been examined in detail by numerous methods, including scanning tunneling microscopy (STM) and photoemission spectroscopy [1,2,4,8–10]. These two techniques form a powerful com-bination of surface probes and are the methods chosen for the present study. STM provides a direct view of the surface atomic structure, and past investigations have already yielded a wealth of information. However, it has proven very difficult to distinguish Si and Ge atoms on the surface due to their similarity in electronic struc-ture, and the problem is compounded by nonuniform apparent buckling of dimers for both the clean and the mixed surfaces [9]. Thus, critical growth issues in regard to atomic spatial distribution, atomic exchange, and in-termixing remain largely unexplored or unanswered. Photoemission studies of the core levels, likewise, are hampered by a lack of strong spectroscopic contrast be-tween atoms in different environments. The intrinsic sur-face shifts of core levels can be useful indications of surface chemistry and composition, but for Ge=Si, the shifts are mostly quite small and not fully resolved, and the assignments of some of the components remain un-certain [2,4,11].

In the present Letter, we report a new approach that overcomes these difficulties and is based on an idea analogous to the chemical staining method widely em-ployed in optical microscopy of biological samples. Trials

of various ‘‘staining agents’’ have indicated Cl as a suit-able choice. Terminating the Ge=Si surface with Cl leads to a large STM contrast between Si and Ge surface sites, thus permitting a detailed counting and characterization of the spatial distributions of the different surface atoms. Photoemission measurements also reveal an unusually large chemical shift between Cl adsorbed on Si and Ge [12,13], thus permitting an independent determination of the surface composition. Measurements of the Si and Ge chemically shifted core levels provide additional infor-mation regarding the surface and subsurface composition. This approach is applied to a study of the growth of Ge on Si(100) by atomic layer epitaxy (ALE) with digermane as the source gas. ALE offers an important advantage over the standard molecular beam epitaxy method in that the amount of deposition is quantized (0.4 atomic layer per cycle for the present case) [8]. Therefore, process control is easily achieved under wide ranges of experimental conditions. Because of the similarity between Ge and Si, one might expect the ALE growth mode at the initial stage to be similar to that of Si on Si(100) —layer buildup via step flow over an otherwise intact substrate. The lower surface free energy of Ge also favors an abrupt interface [3]. Our results show that, instead, the deposited Ge atoms replace Si atoms at random positions on the sur-face, resulting in a growing surface composed of a Si=Ge surface alloy. The interface is not abrupt, but the inter-mixed region is characterized by a very short length scale of a mere 1 monolayer (ML).

In our experiment, samples of Si(100) were cut from commercial p-type wafers and cleaned by direct joule heating to 1450 K for a few seconds. Deposition of Ge was performed by ALE. Each deposition cycle in-volved saturating the surface with 15 L (1 L 106torr sec) of digermane exposure at 350 K. This was fol-lowed by thermal annealing at 950 K to desorb the hydro-gen, leaving behind a net deposition of 0.4 ML of Ge.

P H Y S I C A L R E V I E W L E T T E R S week ending

31 JANUARY 2003 VOLUME90, NUMBER4

(2)

Here, 1 ML 5:76 1014_atoms=cm2_{is defined in terms} of the Si(100) surface. After Ge growth, chlorine gas was introduced into the chamber to saturate the sample sur-face at near room temperature. The Cl 2p, Si 2p, and Ge 3d core levels were measured using synchrotron ra-diation at the Synchrotron Rara-diation Research Center in Hsinchu, Taiwan. STM measurements were performed in a variable-temperature STM system.

Both Si(100) and Ge(100) form a (2 1) dimer recon-struction, as observed by STM. The same (2 1) dimer structure is also observed for the first few cycles of ALE growth of Ge on Si(100). All dimers appear to be about the same under a variety of biasing conditions [1,8], making the distinction between Ge and Si highly uncer-tain. One could imagine that the simplest growth mode would be for the deposited Ge atoms to diffuse on the surface to nearby step edges, where they contribute to step flow as in the growth of Si on Si at similar temperatures [14]. A careful STM examination shows the overall step structure after growth to be similar to that of the original surface. Therefore, step flow does occur for the first few cycles of deposition, but there is no evidence for bands or areas of different chemical compositions neighboring step edges [8].

It is known that Cl adsorption on pure Si(100) and Ge(100) surfaces saturates the dimer dangling bonds while preserving the basic (2 1) dimer structure. Analysis of STM images is considerably simplified be-cause the resulting dimers are symmetric with no buck-ling to cause confusion [15,16]. Figure 1(a) shows a STM image of Si(100) after one cycle of ALE deposition (0.4 ML Ge) followed by Cl termination. The two atoms in a dimer are well resolved in this high-resolution image. Clearly, there are two kinds of surface atoms, with one kind being much brighter than the other. The bright atoms are associated with Ge, as the number of such atoms increases as more Ge is deposited. This contrast enhance-ment is most pronounced for STM images of the empty states with a sample bias 2 V. An immediate conclu-sion, based on the spatial distribution of Ge, is that the simple picture of incorporation of diffusing Ge at step edges leading to step flow is incorrect. Rather, the depos-ited Ge atoms replace Si at random sites on the surface, resulting in both Ge-Ge and Ge-Si dimers with no appar-ent ordering or organization. Note that a pair of Ge atoms is deposited for each digermane molecule, and the con-ventional picture for Ge=Si growth involves diffusing dimers as the basic building block for growth [1,17]. Yet in the present case isolated Ge atomic sites are abundantly observed, suggesting that individual Ge adatoms are re-sponsible for the growth.

Figure 1(b) is a much larger-scale scan for a sample deposited with two ALE cycles. The population of the bright (Ge) atoms is now significantly higher. Direct counting of the Ge and Si surface atoms yields a relative measure of the surface Ge and Si coverages. The absolute

coverages are obtained by normalizing the sum of Si and Ge coverages to 1 ML. The results are shown in Fig. 2 as diamonds after 0, 1, 2, and 3 cycles of ALE deposition.

Figure 3 displays the Cl 2p core level spectra for Cl-terminated surfaces of Si(100) and the same after 1, 3, and 5 cycles of Ge deposition for Ge coverages of 0.4, 1.2, and 2.0 ML, respectively. For Cl-terminated Si, the two peaks correspond to the spin-orbit splitting of Cl attached to Si (labeled Cl-Si in the figure). With increas-ing Ge coverages, a new chemically shifted component, labeled Cl-Ge, emerges and becomes increasingly more intense. This component must correspond to Cl adsorbed on Ge. The difference in binding energy between Cl-Si and Cl-Ge, 0.64 eV, is surprisingly large in view of the similar electronic structure of Si and Ge. This large difference is, however, quite welcome in the present

FIG. 1 (color). STM images obtained with a tunneling current of 0.05 nA and a sample bias of 1.9 V. (a) is for Si(100) after one cycle of Ge ALE growth and Cl termination and covers an area of 48 36 A2. The bright and dark atoms correspond to Cl adsorbed on Ge and Si, respectively. (b) is the same but for two cycles of Ge ALE growth and an imaged area of 205 165 A2.

(3)

investigation, as the intensity evolution of the two com-ponents, obtained by a least-squares fitting of the line shapes, yields a direct measure of the surface Si and Ge coverages. The results, shown in Fig. 2 as circles, are in excellent agreement with the surface coverages obtained by STM counting (diamonds).

Figure 4 shows the Si 2p and Ge 3d core level spectra for the Cl-terminated surfaces with various amounts of Ge deposition. The Si 2p spectrum for the Cl-terminated Si(100) shows two components. The Si component, at a higher binding energy, corresponds to Si directly bonded to Cl, while the B component corresponds to subsurface Si. The chemical shift of Sirelative to B, about 0.9 eV, is consistent with a nominal charge state of 1 for the Si directly bonded to Cl [12]. In the same figure, the inten-sity of Si diminishes for increasing amounts of Ge deposited on the surface. At a 2-ML Ge deposition (5 cycles of ALE), there is very little Si left, because the sample surface is now nearly fully covered by Ge under the Cl layer. The measured intensity evolution of Sireflects the Si coverage in the dimer layer. The results are shown in Fig. 2 using squares, and are in good accord with those obtained from STM counting and Cl core level intensity measurements.

Figure 4 shows that the Ge core level line shape simi-larly consists of two components, Ge and B, with the former being Ge in the surface dimer layer in direct bonding with Cl, and the latter being Ge in the subsurface region. At one cycle of ALE deposition, the spectrum is dominated by Ge with a small, but noticeable, B

component. This indicates that most of the Ge atoms reside in the dimer layer, but some are already covered by the growth. The intensities of both components in-crease for more cycles of ALE growth as expected. The increase in B simply means that more Ge atoms become buried under the surface as the deposition increases. The intensity of Geprovides a measure of the Ge coverage in the dimer layer. Results from this analysis are shown in Fig. 2 using squares, which agree well with the results from the other measurements. The somewhat larger dis-crepancy at 3 and 5 cycles of deposition is within the experimental error of up to 10% each for STM counting and core level intensity measurements. The two curves in Fig. 2 represent the average of the three independent measurements, where available: STM counting, Cl core level analysis, and Si=Ge core level analysis. The good agreement among the three measurements confirms our interpretation and assignments. The shapes of these curves are an indication of the degree of intermixing at the interface. For an ideal layer growth and an abrupt interface, the curves should be straight lines going from 1 to 0 for Si and 0 to 1 for Ge at the completion the first ML deposition of Ge. As the STM results show, Ge atoms, upon deposition, replace Si at random surface sites, rather than coalescence and cover up the substrate surface. This growth behavior rules out an abrupt interface and leads to a graded, intermixed region [18]. The range of this

Clean Si

Relative Binding Energy (eV)

97 98 99 100 101 102 103 104 P h o to e m issi o n I n te nsi ty (arb. u n it s) 0.4 Cl-Si x-ML-Ge/Si(100) + Cl₂ Cl 2p Core hν = 240 eV Cl-Ge 1.2 x = 2.0

FIG. 3. Photoemission spectra for the Cl 2p core level. The samples are Cl-terminated Si deposited with 0, 1, 3, and 5 cycles of Ge ALE (corresponding to Ge deposition of 0, 0.4, 1.2, and 2.0 ML, respectively). The two components, Cl-Si and Cl-Ge, correspond to Cl bonded to Si and Ge, respectively.

ALE Deposition Cycles

0 1 2 3 4 5 6

S

i

and Ge Surface Coverages (ML)

0.0 0.2 0.4 0.6 0.8 1.0 Cl Core Si/Ge Core STM Average Ge Si

FIG. 2. Si and Ge surface coverages (within the dimer layer) deduced from three independent measurements: STM count-ing, Cl 2p core level analysis, and Si=Ge core level analysis. The curves represent averages of the results, where available, for both Si and Ge.

(4)

intermixing is, however, very limited. Fitting the two curves in Fig. 2 to exponential decay functions, the decay length is about 1 ML. This is consistent with the surface-layer interchange mechanism mentioned above, and im-plies no appreciable interlayer or bulk diffusion otherwise to further broaden the intermixed region.

This study is made possible by a large STM contrast between Ge and Si using Cl as a contrast-enhancement agent. Also, the difference in Cl 2p chemical shift be-tween Cl adsorbed on Ge and Si (0.64 eV) is unusually large, which facilitates a detailed analysis of the surface composition. The sign of the Cl 2p chemical shift differ-ence suggests that Cl bonded to Ge (Cl-Ge) is more negatively charged, or more ionic, than Cl bonded to Si (Cl-Si), in agreement with a recent calculation [16]. These results imply that there are significant differences in the electronic structure, bond length, and steric interaction [19] between Cl on Ge and Cl on Si, which together account for the enhanced STM contrast.

In summary, distinguishing Si and Ge atoms on mixed Si=Ge surfaces in atomic-resolution images has been a difficult issue for this important prototypical system. In our work, image contrast enhancement is facilitated by Cl termination, which additionally aids spectroscopic mea-surements of surface chemical composition. This ap-proach enables three independent measurements: STM counting, Cl core level analysis, and Si=Ge core level analysis, which together provide a consistent and detailed picture of the atomic processes involved in the growth.

ALE deposition of Ge on Si(100) is characterized by Ge replacement of individual Si surface atoms, resulting in a mixed alloy surface layer that enriches with Ge as it grows. The interface is thus graded, but the intermixing length scale is a mere 1 ML due to the surface nature of the intermixing process.

This work is supported by the National Science Council of Taiwan under Contract No. NSC-90-2112-M-009-035 (D. S. L.), the U.S. Department of Energy (Division of Materials Sciences, Office of Basic Energy Sciences) under Grant No. DEFG02-91ER45439, and the U.S. National Science Foundation under Grant No. DMR-0203003 (T. C. C.). We thank C. M. Wei for helpful dis-cussions. The Synchrotron Radiation Research Center is supported by the National Science Council, Taiwan, Republic of China.

*Electronic address: [email protected]

[1] F. Liu, F. Wu, and M. G. Lagally, Chem. Rev. 97, 1045 (1997), and references therein.

[2] R. Larciprete, P. DePadova, C. Quaresima, C. Ottaviani, M. Peloi, and P. Perfetti, Phys. Rev. B 61, 16 006 (2000). [3] P. C. Kelires and J. Tersoff, Phys. Rev. Lett. 63, 1164

(1989).

[4] L. Patthey, E. L. Bullock, T. Abukawa, S. Kono, and L. S. O. Johansson, Phys. Rev. Lett. 75, 2538 (1995). [5] See, for example, D. J. Paul, Adv. Mater. 11, 191 (1999),

and references therein.

[6] K. N. Tu, J. Mayer, and L. Feldman, Electronic Thin Film

Science for Electrical Engineers and Materials Scientist

(Macmillan Publishing, New York, 1992).

[7] C. Teichert, Phys. Rep. 365, 335 (2002), and references therein.

[8] K.-H. Huang, T.-S. Ku, and D.-S. Lin, Phys. Rev. B 56, 4878 (1997).

[9] X. R. Qin, B. S. Swartzentruber, and M. G. Lagally, Phys. Rev. Lett. 84, 4645 (2000).

[10] H.W. Yeom, M. Sasaki, S. Suzuki, S. Sato, S. Hosoi, M. Iwabuchi, K. Higashiyama, H. Fukutani, M. Nakamura, T. Abukawa, and S. Kono, Surf. Sci.

381, L533 (1997).

[11] R. I. G. Uhrberg, E. Landemark, and Y.-C. Chao, J. Electron Spectrosc. Relat. Phenom. 75, 197 (1995). [12] R. D. Schnell, F. J. Himpsel, A. Bogen, D. Rieger, and

W. Steinmann, Phys. Rev. B 32, 8052 (1985).

[13] T. D. Durbin, D. A. Lapiano-Smith, F. R. McFeely, F. J. Himpsel, and J. A. Yarmoff, Surf. Sci. 330, 147 (1995). [14] P.-H. Wu and D.-S. Lin, Phys. Rev. B 57, 12 421 (1998). [15] I. Lyubinetsky, Z. Dohnalek, W. J. Choyke, and J. T. Yates,

Jr., Phys. Rev. B 58, 7950 (1998).

[16] M. C¸ akmak, S. C. A. Gay, and G. P. Srivastava, Surf. Sci.

454 – 456, 166 (2000).

[17] B. Borovsky, M. Krueger, and E. Ganz, Phys. Rev. Lett.

78, 4229 (1997).

[18] R. M. Tromp, Phys. Rev. B 47, 7125 (1993).

[19] C. F. Herrmann, D. Chen, and J. J. Boland, Phys. Rev. Lett. 89, 096102 (2002).

Relative Binding Energy (eV) -1 0 1 2 3 Photoemission I n tensit y (arb. units ) Si+ x-ML-Ge/Si(100) + Cl₂ Si 2p Core hν = 140 eV

Relative Binding Energy (eV) -71 -70 -69 -68 x = 2.0 Ge+ 0.4 1.2 B x-ML-Ge/Si(100) + Cl₂ Ge 3d Core hν = 140 eV B x = 2.0 0.4 1.2 0

FIG. 4. Photoemission spectra for the Si 2p and Ge 3d core levels. The samples are Cl-terminated Si deposited with 0, 1, 3, and 5 cycles of Ge ALE (corresponding to Ge deposition of 0, 0.4, 1.2, and 2.0 ML, respectively). Two components are seen in each case. The components labeled Siand Gecorrespond to Si and Ge dimer atoms directly bonded to Cl, respectively, while the B components correspond to Si and Ge in the subsur-face region.