Composition and atomic ordering of Ge/Si(001) wetting layers

Thin Solid Films 515 (2007) 5587 – 5592 www.elsevier.com/locate/tsf Composition and atomic ordering of Ge/Si(001) wetting layers A. Malachias a,b,⁎, T.H. Metzger a , M. Stoffel b , O.G. Schmidt b , V. Holy c a European Synchrotron Radiation Facility, 6 Rue Jules Horowitz, 38043 Grenoble, France Max-Planck Institute für Festkörperforschung, Heisenbergstrasse 1, D-70569 Stuttgart, Germany Charles University, Faculty of Mathematics and Physics, Department of Electronic Structures, Ke Karlovu 5, 121 16 Praha, Czech Republic b c Available online 25 January 2007 Abstract A combination of X-ray diffraction with anomalous X-ray scattering at the Ge K edge and specular reflectivity measurements is used to reveal both composition and atomic ordering in Ge:Si wetting layers. By comparing the intensity distribution close to the (400) and (200) surface reflections we show that the Ge wetting layer is composed of a SiGe alloy which exhibits atomic ordering. Due to the Si interdiffusion the wetting layer thickness is larger than the nominal 3 ML Ge deposition. The chemical depth distribution is obtained from X-ray reflectivity measurements and confirms the enhanced Ge interdiffusion. These phenomena evidence the crucial interplay between surface kinetics and intermixing in SiGe thin films and nanostructures on Si(001) substrates. © 2006 Elsevier B.V. All rights reserved. Keywords: Atomic ordering; Wetting layer; Grazing incidence diffraction; Anomalous X-ray scattering Self-assembled semiconductor nanostructures have been extensively investigated during the past decades. The attention was mainly focused on Stranski–Krastanow (S–K) islands since their electronic and optical properties are of particular interest for applications [1,2]. Deposition methods such as molecular beam epitaxy (MBE) or chemical vapor deposition (CVD) allow a fine tuning of island shape and size distribution. A general description of S–K growth mode can be underlined as follows. In the beginning of the deposition process the heteroepitaxial growth proceeds in a layer-by-layer mode up to a critical coverage. Then, for thicker layers the strain energy stored in the two-dimensional wetting-layer (WL) is partially released by the formation of three-dimensional islands [3]. Recent structural studies have been focused on the interdiffusion of atoms from the substrate to S–K islands. Such effect was independently observed in different material systems by techniques such as chemical etching [4,5], transmission electron microscopy (TEM) [6,7], anomalous X-ray scattering (AXS) [8,9] and photoluminescence measurements [10]. Although the island formation, strain, composition and morphological evolution were systematically addressed by ⁎ Corresponding author. Max-Planck Institute für Festkörperforschung, Heisenbergstrasse 1, D-70569 Stuttgart, Germany. E-mail address: [email protected] (A. Malachias). 0040-6090/$ - see front matter © 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.tsf.2006.12.021 several authors there is a lack of chemical and structural information concerning the WL. Additionally, there is an open debate on the origin of interdiffusion in S–K systems. For instance, surface diffusion on the WL and trench formation around the island base were observed and proposed as paths for strain relaxation [6]. However, the validity of these models could only be tested on a mesoscopic scale. Nevertheless, it is crucial to understand how the incorporation of atoms from the substrate proceeds on the atomic level. A direct evidence of the surface kinetic role in interdiffusion was recently described for Ge/Si(001) islands. Atomically ordered domains were found inside self-assembled Ge domes and at the WL [11]. The formation of such domains is related to the existence of a stepflow kinetic mechanism that is responsible for interdiffusion during the Ge deposition on Si(001) surfaces. In this work we further explore the composition and atomic ordering of the WL. Our measurements show that the final WL thickness is larger than expected from the nominal deposited amount of Ge. This result is attributed to strong kinetic interdiffusion. A kinetic mechanism of SiGe ordering on Si(001) surfaces was proposed by Jesson et. al. [12]. Since ordering is associated with dimer formation, a step-flow atomistic model was suggested to explain how SiGe order arises naturally at monolayer step edges during coherently 2D island growth without the need of atomic rearrangement after the deposition of a complete 5588 A. Malachias et al. / Thin Solid Films 515 (2007) 5587–5592 Fig. 1. (a) SiGe atomic ordering mechanism steps as proposed by Ref [Jesson91]. Si-rich sites are represented by: Si-γ (dark, white cross) and Si-δ (dark, white spot). Ge rich sites are denoted by Ge-α (white) and Ge-β (white, dark cross). The explanation of the order mechanism can be found in the text. (b) Schematic representation of the Si/Ge atomic ordering in-plane arrangement with an anti-phase boundary indicated by the dashed line. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article). bilayer. Fig. 1a shows a schematic view of a Si(001) surface. In the represented [1 − 1 0] projection the dimers are oriented perpendicular to the picture plane, i.e. each point in Fig. 1(a) corresponds to a Si or Ge dimer in the [1 1 0] direction. The growth will proceed in the [1 − 1 0] direction initially replacing the Si rebounded dimer of (I) by a Ge dimer in the perpendicular direction. The second Ge dimer in this structure will be deposited between the two Si atoms of step (II). Despite of being kinetically frozen at low temperatures, the configuration shown in (III) is energetically unfavorable due to the high stress on the Si marked dimer and the presence of unsaturated bounds at the step edge on the right. The next Si atom (on the right, just after the second Ge dimer) is located in a suitable site for a possible adatom-substrate exchange mechanism driven by totalenergy minimization. The exchange of Si and Ge atoms at this step edge (see arrow in step III) replaces a Si dangling bond by a less energetic Ge one. A driving force of the order of 0.5 eV was estimated for this site exchange [12], leading to the configuration shown in (IV) that is more stable than (III). Since this site exchange has a probability to happen, two different types of Gerich (α, β) and Si-rich (γ, δ) sites will arise. The first kind of sites – Ge-α and Si-γ – are related to originally preferential sites for substrate Si and deposited Ge dimers. Ge-β and Si-δ sites correspond to dimers that have been rearranged by the exchange mechanism. The growth front continues along the [1 −1 0] direction, alternating Si and Ge atoms by repeating steps I–IV until the first ordered layer is complete. The next layers will grow following the b111N ordered direction which was proven to be less energetic over an arbitrary vertical ordering [13]. This model explains very well the beginning of the ordering process but may not be applied to the Si incorporation in thick layers. A schematic representation of the final ordered structure is shown in Fig. 1(b). Ge atoms deposited on a (2 × 1) reconstructed Si(001) surface select specific sites and produce rows with the same atomic species along the [1 1 0] or [1 − 1 0] direction. Anti-phase boundaries are formed when they are shifted by one atomic distance in the direction perpendicular to these rows. At this intersection an anti-phase boundary in the [010] direction can be created, as represented by the red dashed line. This drawing contains only one type of Si and Ge atoms since each layer has only one atomic site for each atomic species. A detailed X-ray investigation about possible Si0.5Ge0.5 ordered structures in thin films was performed by Tischler et al. [14]. The crystallographic measurements of the superstructure reflection intensities lead to a modified RS2 ordering model (called RS3), with two different structures: the main b111N ordered structure and a secondary structure ordered along the b100N direction. More recently, metastable ordered structures were discovered near the surface [15]. It is remarkable that all works were done in alloy samples and the kinetic origin of ordering has been continuously corroborated [13–16]. In this study we investigate two samples grown by solid source MBE on Si(001) substrates. In the first sample 11 monolayers (MLs) of Ge were deposited at a substrate temperature of 700 °C. Atomic force microscopy measurements showed a highly monodisperse distribution of dome shaped islands [17]. In the second sample 3 MLs of Ge were deposited at the same temperature. This coverage is near to the WL critical thickness. AFM measurements show that no island formation was observed. In this work we refer to these samples as the Dome sample and the WL sample, respectively. A. Malachias et al. / Thin Solid Films 515 (2007) 5587–5592 X-ray grazing incidence diffraction measurements were performed at the ID01 beamline of the European Synchrotron Radiation Facility. The incident angle was fixed at 0.17° for the Dome sample and 0.1° for the WL sample below the critical angle for total external reflection. A position sensitive detector was mounted to acquire the scattering signal in the exit angle range from 0 to 1.5°. The X-ray photon energy was set to 8000 eV for the Dome sample. For the WL sample two X-ray energies were used in order to perform AXS measurements and gain chemical sensitivity: at 11,040 eV and at 11,103 eV (Ge K edge) [8,9]. Longitudinal (radial) ϴ–2ϴ scans were performed by varying the momentum transfer qr = (4π/λ)sin(2ϴ/2). Such scans are sensitive to the local in-plane lattice parameter given by a′ = 2π/qr. Enhanced depth-sensitivity measurements were performed along the direction perpendicular to the surface, where the qz momentum transfer is given by qz = (2π/λ)(sinαi + sinαf), where αi and αf are the incident and exit angles, respectively. The structure factor of diamond holds for Si and Ge crystals since they have the same lattice structure. Allowed (h, k, l) reflections with non-zero structure factor of both materials are observed in two cases: a) h, k and l are all odd and; b) h, k and l are all even and h + k + l = 4n, where n is an integer [18]. Key examples of the first type of reflection are (1 1 1), (3 3 3), (3 1 5). The second type of reflection includes (2 2 0), (4 0 0) and (6 5589 2 0). In these reflections the scattered intensity is proportional to the square of the sum of atomic scattering factors of the elements that constitute the crystal i.e., all atoms contribute to the scattered signal. Alternatively, mixing Si and Ge atoms in an ordered way may give rise to superstructure reflections that are caused by the repetition of a specific atomic arrangement inside the crystal. In such reflections any observed intensity originates exclusively from the regions of the crystal that are ordered. Their position in reciprocal space depends on the periodicity of the superlattice unit cell, which differs from the chemical unit cell. For this reason it is generally possible to find a superstructure reflection in reciprocal space positions where no allowed (fundamental) reflection exists [14,15]. Fig. 2(a) shows an X-ray radial ϴ–2ϴ scan in the vicinity of the Si (400) reciprocal space position for the Dome sample. The diffraction profile reveals a broad scattered intensity distribution spanning from the Si reciprocal space position qr = 4.628 Å− 1 down to qr = 4.480 Å− 1. This is a direct indication that the inplane lattice parameter that is constrained to the Si value at the bottom relaxes for increasing height inside Ge islands. A similar scan performed near the Si (200) reciprocal space position reveals an unexpected result. The reflection shown in Fig. 2(b) is forbidden for pure Si and Ge crystals. The (200) structure factor is equal to zero and the scattered intensity is only Fig. 2. Radial scans along qr in the vicinity of (a) Si (400) reflection, and (b) Si (200) reflection for the dome sample. Angular scans performed at qr = 2π × h / 5.50 Å (indicated by the vertical lines in (a) and (b)) are shown in (c) and (d) for the (400) and (200) reflection, respectively. 5590 A. Malachias et al. / Thin Solid Films 515 (2007) 5587–5592 Fig. 3. qr/qa reciprocal space map in the vicinity of the Si (200) reflection. The horizontal dashed line indicates the position of the angular cut of Fig. 2(d). The WL peak is shown by the arrow. In this map the color intensity scale is linear. observed for a partially ordered SiGe alloy. Hence, this measurement gives evidence of long-range atomic ordering in this system. Angular ϴ scans measured at a fixed qr (2ϴ) position essentially probe the Fourier transform of the island region with constant lattice parameter. The angular scan, performed at the (400) reflection (Fig. 2(c)), exhibits a broad peak centered at qa = 0 and subsidiary maxima, denoting the finite size and narrow size distribution of the island ensemble [8]. In contrast to the (400) reflection, an angular scan performed at the (200) reflection yields a very different profile as seen in Fig. 2(d). A pronounced minimum is observed at qa = 0, which cannot be generated by structures that are interfering constructively. The complete qa/qr map in the vicinity of the (200) reflection for the Dome sample is shown in Fig. 3. Spanning from qr = 2.27 Å− 1 to 2.33 Å− 1 two different structures are seen. In the region of the strained alloy (qr b 2.31 Å− 1) the double peak structure is always present. This evidences that SiGe alloy regions inside the islands are mainly ordered with anti-phase boundaries among domains. The weak narrow peak observed exactly at the Si (200) position indicates that the WL is partially ordered but without establishing antiphase boundaries. Although in Fig. 2(b) the peak intensity at qr = 2.314 Å− 1 may suggest that the WL scatters with a considerably high intensity one should notice that most of the scattering from the islands lies beside the radial path (qa = 0). The existence of atomic ordering in antiphase domains inside Ge islands is described in Ref. [11] and will not be discussed any further. Here we further explore the origin of the scattering peak centered at the Si lattice position qr = 2.314 Å− 1 shown in Fig. 2 (b) and indicated by the arrow in the map of Fig. 3. In order to possibly ascribe this scattering to a thin WL, we must rule out some other possibilities. Firstly, since the GID measurements of Fig. 2 were performed on uncapped islands one may argue whether the WL is oxidized or preserved. In fact, the kinetic mechanism that is responsible for ordering in SiGe alloys pull down Ge atoms while Si atoms are pumped to the sample surface [12]. Since the native Si-rich oxide is more stable than a Ge-rich Fig. 4. Anomalous transversal scans at the Si (200) reciprocal space position of the WL sample. The Ge content is extracted from the integrated intensity ratio between measurements at 11,040 eV and 11,103 eV. The inset shows transversal scans performed at both energies in a reference 8 Å-thick Si0.5Ge0.5 film grown at 700 °C. oxide the actual scenario points out to a strong interdiffusion of Ge atoms towards the substrate. Secondly, a narrow peak at the Si (200) position may result from the Si crystal truncation rod (CTR) [19] that spans between the (2 0 2) and the (2 0 −2) reciprocal space positions. This hypothesis can be probed either by following the scattering at the (2 0 l) direction or by chemically sensitive AXS measurements. Finally, the (2 0 0) peak may be ascribed to scattering from imperfect symmetries that arise from the electron clouds of Si atoms. This last consideration can be evaluated by performing anomalous scattering at the Ge K edge. In such condition any intensity contrast measured close to the Ge K edge can be ascribed unambiguously to the presence of Ge atoms. No variation in the scattered intensity is expected from reciprocal space structures originated from pure Si such as multiple scattering and/or CTRs from the substrate. In the following paragraphs we provide evidence of the WL scattering that rule out the possibilities discussed above. Anomalous X-ray scattering measurements were performed on the WL sample to confirm that the (200) peak is generated by Fig. 5. X-ray reflectivity of the WL sample measured at 11,103 eV. Fits using the two-layer and three-layer electron density profiles sketched in the inset are represented by dashed and solid lines, respectively. A. Malachias et al. / Thin Solid Films 515 (2007) 5587–5592 a SiGe alloy. Two X-ray photon energies below the Ge K edge were used: 11,103 eV and 11,040 eV. The Ge scattering factor is reduced by ∼ 30% at the first energy with respect to the second while the Si scattering factor remains constant [20]. Fig. 4 shows transversal scans at the (200) Si position using αi = 0.1° for both energies. The reduction of the scattered intensity at 11,103 eV directly points to a considerably high Ge composition at the ordered part of the WL. Using the ratio between the integrated intensities one obtains directly a Ge content of 0.51 ± 0.04. This is in agreement with the CuAu–I type structure discussed in Refs. [11,15]. Additionally, considering that a Si0.49Ge0.51 film is obtained from 3 ML Ge deposition at this temperature the total estimated WL thickness is about 8.5 Å. The inset of Fig. 4 shows the same measurement performed on a nominally Si0.5Ge0.5 8 Åthick film specially grown at 700 °C as a reference sample. The obtained concentration of 0.52 ± 0.06 indicates essentially the same concentration for the ordered alloy. In order to corroborate the results described above reflectivity measurements were performed on the WL sample. One reflectivity curve measured at 11,103 eV is shown in Fig. 5 together with simulations performed using a two-layer and a three-layer models. Electron density profiles normalized by the Si value ρSi = 0.7 electrons/Å3 are shown in the inset and represented by dashed line (two-layer) and solid line (threelayer) [21]. For 11,103 eV the Ge anomalous atomic scattering factor leads to an electronic density of 1.12 e− /Å3 (ρGe / ρSi = 1.6). Since all measurements were performed in ambient environment some considerations about the ordering and oxidation processes must be done before describing the obtained layer composition. Firstly, since the growth temperature of 700 °C optimizes the Ge–Si ordering mechanism [11,16] we assume that an average Si0.5Ge0.5 composition is found at the whole WL structure that has ∼ 6 ML thickness. Secondly, the existence of a Ge-rich topmost layer reduces the surface energy [22,23] since Ge dangling bonds are less reactive than Si. Finally, the difference between the heat of formation of SiO2 (− 204 kcal/mol) and GeO2 (− 119 kcal/mol) drives the surface oxide composition towards a relative Si enrichment despite the formation of pure GeO2 with depletion of Ge atoms 2to a Ge-rich layer below the oxide/semiconductor interface [24,25]. A qualitative scenario that corresponds to the expected chemical depth profile, described above is obtained by the threelayer electron density model. A thin (4 Å) SixGe1 − xO2 oxide layer is found at the surface with x = 0.30 ± 0.05. A SixGe1 − x subsurface alloy layer with 6 Å thickness and x = 0.08 ± 0.04 is found on top of a 9 Å Si-rich alloy with x = 0.75 ± 0.05 [21]. Despite leading to a similar fit the two-layer model implies a layered structure with the absence of surface oxidation. Since an oxide layer consisting of pure GeO2 may lead to an electron density of ρ′ = 0.94 e − /Å 3 (ρ′ / ρSi = 1.35) such profile is physically inconsistent. In this model the top layer consists of a SixGe1 − x alloy with x = 0.11 ± 0.06 while the second alloy layer has x = 0.65 ± 0.05. Although both models fit the measured data well, the threelayer profile is physically consistent with dry oxidation on SiGe. However, both fits point out to a depth limit of ∼20 Å for Ge 5591 interdiffusion. These results indicate that atomic ordering may take place in the alloy layers and at the SiGe/SiGeO2 interface, giving raise to Si0.5Ge0.5 domains with large lateral (in-plane) size embedded in a SiGe alloy with slightly different stoichiometry. Hence, no X-ray phase shift that should lead to a transversal scan profile similar to that of Fig. 2(d) is observed [11]. The 50% Ge concentration found from anomalous (200) measurements performed at the WL sample refers only to ordered SiGe regions. Although different Ge contents were obtained for the alloy layers, such a concentration of an ordered alloy does not contradict the reflectivity analysis. In summary we have shown an enhanced Ge interdiffusion towards the Si substrate for wetting layers grown by MBE at 700 °C. Such observation is explained by the optimization of the kinetic incorporation of Ge atoms into Si. The existence of a (200) X-ray peak unambiguously indicates partial atomic ordering at the WL. The analysis of the reflectivity measurement directly shows a chemical depth profile that is in qualitative agreement with the results from AXS. References [1] O.G. Schmidt, K. Eberl, IEEE Tans. Electron Devices 48 (2001) 1175. [2] G.S. Kar, S. Kiravittaya, U. Denker, B.-Y. Nguyen, O.G. Schmidt, Appl. Phys. Lett. 88 (2006) 253108. [3] D.J. Eaglesham, M. Cerullo, Phys. Rev. Lett. 64 (1990) 1943. [4] O.G. Schmidt, U. Denker, S. Christiansen, F. Ernst, Appl. Phys. Lett. 81 (2002) 2614; U. Denker, M. Stoffel, O.G. Schmidt, Phys. Rev. Lett. 90 (2003) 196102. [5] U. Denker, A. Rastelli, M. Stoffel, J. Tersoff, G. Katsaros, G. Costantini, K. Kern, N.Y. Jin-Phillipp, D.E. Jesson, O.G. Schmidt, Phys. Rev. Lett. 94 (2005) 216103. [6] S.A. Chaparro, Jeff Drucker, Y. Zhang, D. Chandrasekhar, M.R. McCartney, D.J. Smith, Phys. Rev. Lett. 83 (1999) 1199. [7] M. Floyd, Y. Zhang, K.P. Driver, Jeff Drucker, P.A. Crozier, D.J. Smith, Appl. Phys. Lett. 82 (2003) 1473. [8] A. Malachias, S. Kycia, G. Medeiros-Ribeiro, R. Magalhães-Paniago, T.I. Kamins, R. Stanley Williams, Phys. Rev. Lett. 91 (2003) 176101. [9] T.U. Schülli, J. Stangl, Z. Zhong, R.T. Lechner, M. Sztucki, T.H. Metzger, G. Bauer, Phys. Rev. Lett. 90 (2003) 066105. [10] O.G. Schmidt, C. Lange, K. Eberl, Appl. Phys. Lett. 75 (1999) 1905. [11] A. Malachias, T.U. Schülli, G. Medeiros-Ribeiro, L.G. Cançado, M. Stoffel, O.G. Schmidt, T.H. Metzger, R. Magalhães-Paniago, Phys. Rev., B 72 (2005) 165315. [12] D.E. Jesson, S.J. Pennycook, J.-M. Baribeau, Phys. Rev. Lett. 66 (1991) 750; D.E. Jesson, S.J. Pennycook, J.-M. Baribeau, D.C. Houghton, Phys. Rev. Lett. 68 (1992) 2062; D.E. Jesson, S.J. Pennycook, J.Z. Tischler, J.D. Budai, J.-M. Baribeau, D.C. Houghton, Phys. Rev. Lett. 70 (1993) 2093. [13] F.K. LeGoues, V.P. Kesan, S.S. Iyer, J. Tersoff, R. Tromp, Phys. Rev. Lett. 64 (1990) 2038. [14] J.Z. Tischler, J.D. Budai, D.E. Jesson, G. Eres, P. Zscack, J.M. Baribeau, D.C. Houghton, Phys. Rev., B 51 (1995) 10947. [15] H. Reichert, S.C. Moss, P. Imperatori, K. Evans-lutterodt, Appl. Phys. Lett. 74 (1999) 531. [16] V.P. Kesan, F.K. LeGoues, S.S. Iyer, Phys. Rev., B 46 (1992) 1576. [17] T.U. Schülli, M. Stoffel, A. Hesse, J. Stangl, R.T. Lechner, E. Wintersberger, M. Sztucki, T.H. Metzger, O.G. Schmidt, G. Bauer, Phys. Rev., B 71 (2005) 035326. [18] B.E. Warren, X-ray Diffraction, Dover Publications, 1969. [19] G. Renaud, P.H. Fuoss, J. Bevk, B.S. Freer, Phys. Rev., B 45 (1992) 9192. [20] R. Magalhães-Paniago, G. Medeiros-Ribeiro, A. Malachias, S. Kycia, T.I. Kamins, R.S. Williams, Phys. Rev., B 66 (2002) 245312. 5592 A. Malachias et al. / Thin Solid Films 515 (2007) 5587–5592 [21] S. Benarjee, M.K. Sanyal, A. Datta, S. Kanakaraju, S. Mohan, Phys. Rev., B 54 (1996) 16377. [22] N.R. Zangenberg, J.L. Hansen, J.F. Pedersen, A.N. Larsen, Phys. Rev. Lett. 87 (2001) 125901. [23] L. Nurminem, F. Tavazza, D.P. Landau, A. Kuronen, K. Kaski, Phys. Rev., B 68 (2003) 085326. [24] H.K. Liou, P. Mei, U. Gennser, E.S. Yang, Appl. Phys. Lett. 59 (1991) 1200. [25] M. Spadafora, G. Privitera, A. Terrasi, S. Scalese, C. Bongiorno, A. Carnera, M. Di Marino, E. Napolitani, Appl. Phys. Lett. 83 (2003) 3713.