Influence of twinned structure on the morphology of CdTe(1 1 1) layers grown by MOCVD on GaAs(1 0 0) substrates

Influence of twinned structure on the morphology of CdTe(1 1 1) layers grown by MOCVD on GaAs(1 0 0) substrates

ARTICLE IN PRESS Journal of Crystal Growth 257 (2003) 60–68 Influence of twinned structure on the morphology of CdTe(1 1 1) layers grown by MOCVD on ...

407KB Sizes 0 Downloads 6 Views

ARTICLE IN PRESS

Journal of Crystal Growth 257 (2003) 60–68

Influence of twinned structure on the morphology of CdTe(1 1 1) layers grown by MOCVD on GaAs(1 0 0) substrates I. Mora-Sero! a, C. Polopb, C. Ocalb, M. Aguilo! c, V. Mun˜oz-Sanjose! a,* a

Department F!ısica Aplicada, Universitat de Val!encia, C/Dr. Moliner 50, Burjassot (Val"encia) 46100, Spain b Instituto de Ciencia de Materiales de Madrid, C.S.I.C., Cantoblanco, Madrid 28049, Spain c Laboratori de F!ısica Aplicada i Cristal  lografia, Universitat Rovira i Virgili, Tarragona 43005, Spain Received 2 September 2002; accepted 15 May 2003 Communicated by J.B. Mullin

Abstract The morphology and structure of CdTe(1 1 1) layers grown on GaAs(1 0 0) by MOCVD have been studied by atomic force microscopy (AFM) and X-ray texture analysis. Growth conditions have been chosen so that mirror-like CdTe layers are obtained. Layers whose growth times vary between 10 s and 2 h have been investigated. The X-ray texture analysis shows that the CdTe layers grown on GaAs substrates that were thermally treated at 580 C for 30 min in a H2 atmosphere exhibit a (1 1 1) preferential orientation and are twinned. This twinned structure of the (1 1 1)CdTe layer which is observed as 60 rotated triangular crystallites in the AFM images strongly influences the surface morphology. The AFM results have been interpreted using a dynamic scaling theory. The occurrence of a 2D–3D growth transition has been detected after periods of growth in the range of 100–300 s. r 2003 Elsevier B.V. All rights reserved. PACS: 68.55.Jk; 81.05.Dz; 81.15.Gh Keywords: A1. Atomic force microscopy; A1. Growth models; A1. X-ray texture analysis; A3. Metalorganic chemical vapour deposition; B1. CdTe

1. Introduction Nowadays it is well established that CdTe is a material of great interest for the fabrication of electronic devices as solar cells [1,2], optical waveguides [3] and gamma ray and X-ray detectors [4]. The mixture of CdTe with HgTe, forming the alloy Hg1xCdxTe (MCT), has *Corresponding author. Fax: +34-96-3983146. E-mail address: [email protected] (V. Mun˜oz-Sanjos!e).

important technological applications in IR detection [5,6]. Several methods have been developed for the crystal growth of CdTe; among them, metalorganic chemical vapour deposition (MOCVD) has been largely applied [7,8]. MOCVD has become an important industrial technique for manufacturing thin solid layers for use in optoelectronic devices [9]. The success of this technique stems primarily from its flexibility and the high compositional control that it offers.

0022-0248/03/$ - see front matter r 2003 Elsevier B.V. All rights reserved. doi:10.1016/S0022-0248(03)01410-6

ARTICLE IN PRESS I. Mora-Sero! et al. / Journal of Crystal Growth 257 (2003) 60–68

In the fabrication of Hg1xCdxTe IR detecting diodes, the device performance demands a very high quality in terms of both electrical properties and structural characteristics. For the growth of Hg1xCdxTe epitaxial layers difficulties with an easy and low cost availability of large area CdTe and (Cd, Zn)Te substrates have undoubtedly stimulated a considerable interest in the use of alternative substrates, especially GaAs. The choice of GaAs as substrate is attractive because of its relatively low-cost in large areas and the availability of good quality wafers. Nevertheless, it has an important drawback due to the lattice mismatch between substrate and layer which is as large as 14.7% for Hg0.8Cd0.2Te. To solve this problem and to avoid auto-doping contamination of the Hg1xCdxTe layer from the substrate (Ga is a rapid diffuser in Hg1xCdxTe) and also to obtain a smooth surface morphology, it is necessary to use a buffer layer, normally of CdTe, [10,11]. However, the growth and characteristics of this buffer layer strongly influences the morphology of the Hg1xCdxTe subsequently grown [12,13]. It is possible to grow (1 1 1) or (1 0 0)CdTe on (1 0 0)GaAs depending on substrate processing before growth due to the large lattice mismatch between CdTe and GaAs [14–21]. The (1 1 1) orientation has a smoother surface morphology but is twinned [18,22] (rotation of 60 around the [1 1 1] direction normal to the surface). The (1 0 0) orientation shows a better crystalline quality but it presents morphological defects, namely hillocks [23]. A significant effort has been devoted to find out the origin of this growth behaviour. Nishino et al. [24], and Kim et al. [25] have investigated the mechanism for hillocks formation on the (1 0 0)CdTe/(1 0 0)GaAs. The twinned nature of (1 1 1)CdTe layers is well known [26–28]. One way to avoid the formation of microtwins is to increase the tilt angle of the growth surface from the (1 1 1) B orientation [29], or to employ vicinal substrates as (2 1 1) A, (3 1 1) B, (5 1 1) B [30–32]. Despite of the interest and intense work devoted to the subject the growth mechanisms that rule the growth process of (1 1 1)CdTe on (1 0 0)GaAs substrates are not completely understood. In the past few years continuous evolution involving

61

characterisation techniques together with the development of growth theories have opened new ways of dealing with this topic. In this sense, X-ray texture analysis and atomic force microscopy (AFM) are two powerful characterisation techniques that provide a new approach to analyse the growth processes. X-ray texture analysis can provide a structural characterisation of the layers, thus improving the value of standard X-ray diffraction. AFM can provide 3D images of the sample surface that show morphological detail that can be used to follow the evolution of surface roughness. In spite of the development of several techniques for in situ monitoring of the MOCVD growth process, they are not widely used [33] because of difficulties associated with their implementation in conventional reactors. AFM is a good alternative for systematic ex situ studies that permit the use of dynamic scaling theory [34]. In this theory two characteristic lengths are usually employed to study the growth dynamics: the surface roughness (s) and the lateral correlation length (x). It is assumed that both, s and x; change with growth time (t) as sBtb and xBt1=Z : The growth exponent (b) and the coarsening exponent (Z), depend on the growth mechanisms operating during deposition. The scaling exponent (a), that depends on the growth mechanism [34] can also be introduced. For self-affine surfaces the three exponents are related, Z ¼ a=b: Thus, in some cases the experimental determination of these exponents and the comparison with the predicted values by different growth models can be used to identify the main growth mechanisms involved in the layer deposition. Following these ideas, we have studied (1 1 1)CdTe growth by MOCVD on (1 0 0)GaAs substrates, in order to best understand the growth process and the resulting correlations. The evolution of CdTe layers during the growth process has been determined by X-ray texture analysis and AFM measurements. To investigate the evolution of surface morphology we have systematically analysed the layers deposited using different growth times (i.e. different thickness). Finally, the results are compiled and analysed in the frame of the dynamic scaling theory.

ARTICLE IN PRESS 62

I. Mora-Sero! et al. / Journal of Crystal Growth 257 (2003) 60–68

2. Experimental procedure CdTe layers were grown at atmospheric pressure in a horizontal vent-run type MOCVD reactor (Quantax 226 refurbished by EMF Ltd). For this purpose, electronic grade di-iso-propyltelluride (DIPTe) and dimethylcadmium (DMCd) were used as Te and Cd precursors, respectively. H2 was the carrier gas, and epiready GaAs(1 0 0)70.5 wafers were employed as substrates. As previously mentioned different authors have pointed out the difficulties of controlling the CdTe orientation on GaAs and several procedures have been proposed. Nevertheless such a diversity of methods [14–16,18–21,35] highlights the difficulties and complexity of the process that some times needs to be adapted to each particular situation. In our current experimental work we have obtained (1 1 1)CdTe layers on GaAs substrates by an in situ thermal treatment on nonchemically etched GaAs substrates (i.e. inside the reactor just before the growth process) at 580 C for 30 min under a 3.0 slm total H2 flow. This temperature was also used by Cheng et al. [15] to growth CdTe on 2 -off (1 0 0)GaAs. Their samples were grown using a ‘‘two-step’’ process. The initial layers were grown at lower temperatures (280–320 C). The deposit at lower temperatures was polycrystalline with both (1 1 1) and (1 0 0)-oriented grains. When this initial layer was heated to the ‘‘bulk’’ layer growth temperature, epitaxial regrowth occurs giving a single (1 0 0) crystal film [15]. We have used a growth temperature of 363 C in the overall process where the results reported in the following sections show only a (1 1 1) orientation due to the different growth protocol. In addition the other experimental conditions were the following: The total H2 flow rate through the reactor chamber was kept constant at 3.55 slm. In all cases the molar flows of DMCd and DIPTe through the chamber were 10 and 20 mmol/min, respectively. The substrates were positioned in the chamber of the MOCVD reactor with the [0 1 1] direction parallel to the H2 flow and consequently with the ½0 1 1%  perpendicular to it. In order to study the evolution of surface morphology, CdTe layers were prepared using

growth times between 10 and 7200 s. Samples with thicknesses >1 mm were cleaved and their thickness measured from the corresponding crosssections using a scanning electron microscope. In this case the growth rate obtained for the above conditions was B1.8 mm/h. In order to determine the thickness for thinner samples corresponding to 100 s of growth time we used a Veeco Wyko NT1100 optical profilometer. AFM measurements were performed, under ambient conditions in the contact mode with a home-made microscope head [36], combined with a SPM100 control unit and software from Nanotect. The cantilevers were V-shaped Si3N4 sharpened with nominal force constants of K ¼ 0:1  0:5 N/m (Park Scientific Instruments). Force versus distance curves were systematically obtained to check the tip condition through measurement of the adhesion force. Vertical and horizontal scales were calibrated using monoa( and the surface lattice tomic steps (2.35 A) ( periodicity (2.88 A) of an Au(1 1 1) single crystal, respectively. The AFM software provides the s and x parameters calculated from the surface morphology hð~ r ; tÞ: These parameters are necessary in order to analyse the results in the theoretical framework of the dynamic scaling theory. The s parameter is defined as s ¼ /½hð~ r ; tÞ  /hS1=2 S where / S means the spatial average over the measured area. Roughness is given in terms of the saturation root mean square, i.e. the rms calculated from images of areas large enough (lateral sizes of 7–9 mm). The lateral correlation length, x; is related to the lateral size of the surface features. In this work, x is obtained from the crossover of the 2D power spectral density (PSD) of the AFM images [34] provided by the AFM software. We note that a system with more than one correlation length should show slope changes at the PSD curves. The structural X-ray texture characterisation was carried out in a Siemens D-5000 diffractometer equipped with a Schulz goniometer. Cu Ka radiation was used for diffraction experiments. The pole figures of several samples were determined at a fixed detector position 2yhkl (2y ¼ 23:758 ; 56.820 , 45.348 for (1 1 1)CdTe, (4 0 0)CdTe, and (2 2 0)GaAs poles, respectively)

ARTICLE IN PRESS I. Mora-Sero! et al. / Journal of Crystal Growth 257 (2003) 60–68

by rotating the sample around a normal axis (j angle) and a parallel one (w angle). The covered angular range was 0 pjp360 ; and 0 pwp81 using a step size of Dj ¼ Dw ¼ 3 and a time step size of 3 s.

3. Results and discussion X-ray texture analysis was carried out on samples of different thickness. Results are summarised in Fig. 1. The (2 2 0) pole figure for the GaAs substrate is shown in Fig. 1a, where the [0 1 0] and [0 0 1] directions, parallel to the surface (1 0 0) plane, are represented. A typical pole figure for a (4 0 0)CdTe is depicted in Fig. 1b. Instead of the three-fold symmetry indicative of the [1 1 1] direction that can be expected from the CdTe space group ðF4% 3mÞ; we observe a six-fold symmetry associated with the {1 0 0} family thus revealing the existence of microtwins This six-fold symmetry consists of two axis systems correlated by a rotation of 60 around the [1 1 1] direction. The six peaks of this pole figure present similar intensities and no central peak appears, indicating that there are not CdTe crystallites with a (1 0 0)

Fig. 1. (a) (2 2 0)GaAs substrate pole figure. (b) (4 0 0)CdTe pole figure of a layer grown for 3000 s. (c) (1 1 1)CdTe pole figure of a layer grown for 500 s. (d) (1 1 1)CdTe pole figure of a layer grown for 7200 s.

63

texture. On the other hand, pole figures of (1 1 1)CdTe corresponding to various samples were obtained; two representative examples are depicted in Figs. 1c and d. showing an intense central peak (j ¼ 0 ; w ¼ 0 ) characteristic of (1 1 1) oriented samples. This peak is narrower for samples corresponding to a growth time of 7200 s (Fig. 1d) than for the ones corresponding to growth times of 500 s (Fig. 1c) pointing out that the crystalline quality improves with layer thickness as previously reported [12]. The central peak is actually a double peak (wB3 ), indicating a tilt between the (1 1 1)CdTe layer and the (1 0 0)GaAs substrate. This tilt appears in order to accommodate the mismatch between both materials [37] and it is related to a non-completely perpendicular dangling bond of the first Te atoms bounded with the GaAs surface (see below). Since the tilt shifts the diffraction angle, some diffracted peaks can be out of the pole figure region. In fact, instead of the six peaks at w ¼ 70:53 expected for a twinned layer, only four of them are observable due to a 60 rotation around the [1 1 1] axis of the {1 1 1} orientation. However, the twining description is still valid and supported by the appearance of the two missing peaks in the pole figure of some samples when reducing the threshold intensity. The structural information provided by the Xray texture analysis can be complemented by AFM surface morphology measurements of CdTe layers corresponding to different growth times (different thickness). From high magnification images we obtain morphological details as crystallite shape and structure. Large area images (up to 9 mm) are used to obtain the roughness parameters. Fig. 2 shows AFM topographic images of two lateral sizes (500 nm and 3 mm (labelled )). Lighter contrast corresponds to higher regions. The alphabetical order of the images corresponds to increasing growth times. AFM of the as-received (1 0 0)GaAs substrate showed it to exhibit an extremely flat surface (not shown) with a saturation rms of 0.3 nm. Fig. 2a and a show the GaAs substrate surface after thermal treatment. Due to this treatment the surface becomes rougher (saturation rms of 3.3 nm) and presents ridges 40 nm apart, running

ARTICLE IN PRESS 64

I. Mora-Sero! et al. / Journal of Crystal Growth 257 (2003) 60–68

(a)

(a)*

(b)

(b)*

(c)

(c)*

(c)

(d)*

Fig. 2. (a)–(a) GaAs surface after thermal treatment, before growth of CdTe layer. (b)–(b) CdTe surface 100 s of growth time; (c)–(c) CdTe surface 300 s of growth time; (d)–(d) CdTe surface 1000 s of growth time.

along the direction ½0 1 1%  and being formed by small clusters. These ridges come from an asymmetric evaporation of surface atoms, induced by thermal treatment, and can be understood considering the zinc-blende structure. When an atom is evaporated, the neighbouring atoms along the ½0 1 1%  direction would be evaporated easier than those along the [0 1 1] because of the new weaker bonding field they feel. The surface is nonstabilised and the posterior growth of a (1 1 1)

instead of a (1 0 0) oriented CdTe layer can be explained following the model by Cohen-Solal et al. [38]. In the heteroepitaxial growth of CdTe on (1 0 0)GaAs substrates with a non-stabilised surface with missing As atoms, the positions of these atoms are occupied by Te atoms in a selective site of the (0 1 1) plane. Through a small rotation of the dangling bonds, each Te atom binds to one As and two Ga atoms in first neighbouring positions on the plane underneath. This construction leaves one dangling bond, nearly perpendicular to the (1 0 0) surface, per each three-fold bound Te atoms. Subsequently, incoming Cd atoms would tie the Te dangling bonds and initiate the multilayer growth process with a (1 1 1) orientation parallel to the (1 0 0)GaAs in the [1 1 1] direction. The model implies that the initial tetrahedral structure would exhibit a mirror symmetry leading to a six-fold symmetry instead of the expected three-fold symmetry. These results show that striking differences appear to be related to the growth temperature. In the work of Cheng et al. [15] a two step procedure was needed to transform the (1 1 1) and (1 0 0) oriented polycrystalline layer obtained at low temperature into a (1 0 0) film. A rearrangement of the first layer occurred following a second growth step at higher temperature. In order to analyse these results we carried out a set of experiments involving the growth of CdTe on (1 0 0)GaAs layers at growth temperatures in the range of 298–320 C. XRD measurements show that the (1 1 1)–(1 0 0) mixed orientation detected by Cheng et al. remains even for thicker layers if we maintain the same growth temperature for the overall process. However XRD and texture analysis for CdTe grown at 363 C show a (1 1 1) orientation and only in some few cases a residual (1 0 0) diffraction peak was detected but with an intensity four orders lower than the (1 1 1) peak. This fact is clear evidence that not only the surface preparation of the substrate plays and important role on the film orientation but also the growth temperature. Fig. 2b shows the CdTe surface morphology corresponding to a growth time of 100 s. We find a high density (around 2–4 1010 cm2) of closely spaced triangular islands B40 nm wide and

ARTICLE IN PRESS I. Mora-Sero! et al. / Journal of Crystal Growth 257 (2003) 60–68

B2 nm thick. These triangular islands present two orientations and they correspond to the twinned structure observed by X-ray texture analysis. The density of CdTe islands is of the same order as the density of clusters (3 1010 cm2) observed on the GaAs substrate after thermal treatment, suggesting that they might act as nucleation centres. TenaZaera [39] has shown that, under the experimental growth conditions of this work, the growth regime is basically governed by mass transport, favouring nucleation at the highest locations of the substrate. Furthermore, the mean size of these triangular islands (B40 nm) coincides with the mean distance between ridges in the substrate. As it is visible in the large (3 mm size) image the influence of the initial ridges, along the ½0 1 1%  direction, is reflected on the CdTe layer topography. This influence is also confirmed by the surface roughness data (Fig. 3) that, for growth times lower than 100 s, is similar to that of the initial GaAs surface. We conclude that during the first seconds of the growth process, the (1 1 1) twinned structure of CdTe is already present and the overall CdTe morphology is governed by this substrate morphology, i.e. the substrate preparation plays a fundamental role in the morphology of the growing layers. For growth times longer than 100 s a clear change of the layer morphology is observed. We find that after 300 s, there are no traces of the

Fig. 3. Evolution of CdTe surface roughness as a function of growth time. The values given for the roughness correspond to the saturation rms values, i.e. calculated for images of areas large enough (lateral sizes of 7–9 mm).

65

initial substrate ridges in the AFM images (Fig. 2c). Even if the saturation rms is still low, the large scale surface morphology is completely different. The triangular islands, visible at high magnification (Fig. 2c), remain being 2 nm thick and appear piled up as small bricks in specific regions. This gives the surface, an overall appearance consisting of nearly circular features of lateral ( size of E200 nm and depth of only few tens of A. A close inspection of these features indicates that, in fact, they have a vague hexagonal shape, reminiscent of small twinned building blocks. Looking at the images, we can see that for longer growth times (1000 s; Figs. 2d and d) the pile up process occurs now at two lateral distances (or lateral correlation lengths) providing, roughly speaking, quasi-hexagonal shaped features of two lateral sizes (E200 and E500 nm in average). The smaller ones correspond to those commented on above, the larger ones start to develop above 500 s and are responsible for the increase in the rms seen in Fig. 3. We note that, although the triangular twins are still present for these long growth times, the high magnification images confirm that the twins grow laterally. Whereas, for thinner layers, the islands are homogeneous in size and are present randomly distributed in two orientations (Fig. 2b). For thicker layers obtained after 5000 s, large areas of single orientation develop and form the border of the structures (Fig. 4). In the following we analyse a possible explanation for the observed morphology. When two triangular islands of equal orientation grow laterally, they merge and form a large single domain (see above). However, lateral merging is forbidden for twinned islands (rotated 60 around the [1 1 1] normal surface) during their growth and subsequently a twin boundary is formed. As long as the growth continues new different surface levels are developed. Depending on whether these levels meet laterally the triangular shape of the twins causes quasi-hexagonal structures to be formed. The lateral size of these structures leads to a given correlation length for the system. This is observed as the different lateral size of the hexagonal shaped features commented on above. Our results can be analysed in the frame of the dynamic scaling theory [34]. The growth exponent

ARTICLE IN PRESS 66

I. Mora-Sero! et al. / Journal of Crystal Growth 257 (2003) 60–68

Fig. 4. Detail of a quasi-hexagonal feature on CdTe surface (5000 s of growth time).

b is calculated from the CdTe surface roughness as a function of the growth time (Fig. 3). Two separate growth regimes of layers can be seen. The first one, with a value of s constant (i.e. b ¼ 0) similar to that of the GaAs surface after thermal pre-treatment, indicates a 2D growth. The second regime starts approximately after about 100 s. At this point the substrate roughness does not influence the roughness of the grown layer any more. A linear fit of the data in this regime gives a growth exponent of b ¼ 0:5 corresponding to a 3D growth mode. A transition from 2D–3D growth is characteristic of a Stanski-Krastanov (SK) growth mode, in which a change in the nucleation behaviour typically appears after the deposition of 1–5 monolayers [40]. In our case the samples grown during 100 s have 35 nm thickness, that is, more than 70 monolayers have grown during this growth time. An anomalous 2D growth behaviour, having relatively long times of growth (100 s) with several monolayers, appears in this case. On the other hand in the case of CdTe grown on GaAs substrates no delay in the nucleation, that could justify this 2D growth during long times, has been observed using in situ laser monitoring of the growth process [41]. Another example of anom-

alous behaviour has been observed by K-H. Shim et al. [42], for low-temperature growth of AlN by plasma-assisted molecular beam epitaxy. The possible mechanism that they suggest is the socalled ‘‘magic lattice’’ matching theorem, developed to explain the pseudomorphic 2D growth of heterostructures with a large lattice mismatch. For a ‘‘magic lattice’’ matched system constituted by a layer with a lattice constant af and a substrate with a lattice constant as ; the strain is given by eðm; nÞ ¼ ðmaf  nas Þ=maf ; where m and n are the number of unit lattices of the layer and the substrate, respectively. In the case of the CdTe(1 1 1)/ GaAs(1 0 0) system, the values that minimise the above defined strain correspond to arrangements m ¼ 5; n ¼ 4 and m ¼ 6; n ¼ 5: These arrangements lead to smaller compressive and tensile strains than those that would be expected from a direct comparison between the substrate and layer lattice parameters. This effective strain reduction permits the 2D growth to continue for a longer time. Different growth rates has been also ( observed for films grown during 100 s (3.5 A/s) ( and thicker ones (5 A/s), confirming two different growth mechanisms with two different growth rates. In Fig. 5 the PSD curves, PSDðkÞ; are plotted for different samples with different growth

Fig. 5. PSD curves for different growth time samples. (’) 30 s of growth time; (&) 50 s of growth time; ( ) 100 s of growth time; (J) 300 s of growth time; (m) 500 s of growth time; (m) 1000 s of growth time; (E) 3000 s of growth time; (B) 5000 s of growth time.

ARTICLE IN PRESS I. Mora-Sero! et al. / Journal of Crystal Growth 257 (2003) 60–68

times. Provided by the AFM software, the PSD are calculated as the square of the 2D Fourier transform coefficients of the digitised surface. It can be appreciated the differences between the PSD of layers corresponding to low growth times and layers from larger growth times. First, the values a ¼ b ¼ 0; obtained for low growth times, could indicate that a condensation/evaporation mechanism has taken place. However, for thicker layers, a and b are 0.7 and 0.5, respectively. Second, we observe that depending on growth time, the PSD curves present slope changes related to different correlation lengths. The log(x) vs. log(t) is represented in the low left inset in Fig. 5. From this plot a value of Z ¼ 1:39 can be calculated, that is extremely close to the value obtained from the a and b values and the relation for self-affine surfaces Z ¼ a=b which is 1.4 in our case a result that can be expected from the observed morphology (Fig. 2c and Fig. 2d).

4. Conclusions In order to investigate the surface morphology during the MOCVD CdTe growth process we have systematically studied layers of different thickness using X-ray texture and AFM techniques. X-ray texture analysis shows that the samples are (1 1 1) oriented, with no (1 0 0) textured crystallites. As indicated by the (1 1 1) peak line width, the crystalline quality improves with the layer thickness. Furthermore, the (4 0 0)CdTe pole figure reveals the existence of a six-fold symmetry of the {1 0 0} orientation, indicating the existence of microtwins. The AFM study demonstrates that the surface of the GaAs substrate is clearly affected by the thermal pre-treatment, which produces a rough surface with small clusters forming ridges along the direction ½0 1 1% : This non-stabilised substrate surface is at the origin of the (1 1 1) growth of CdTe [38] and plays a crucial role in the subsequent CdTe layer morphology. For low growth times, the surface layer exhibits triangular islands with a similar density to the substrate cluster density, suggesting that these clusters operate as nucleation centres. The size of the

67

triangular twinned features is determined by the distance between the ridges. A clear change in the layer morphology is observed for growth times of about 100 s. At low magnification, the surface appearance consists of slightly depressed features (hundred nm wide and ( deep) which vaguely present only some tens of A an hexagonal shape due to the triangular twins that form their edges. For longer growth times, islands with equal orientation meet and start growing laterally. As a consequence large areas of single orientation (not twinned) are formed and the crystalline quality improves. The dynamic scaling theory has been employed to quantify the results. Two separate regimes of layer growth have been determined. The first one corresponds to a 2D growth with a s similar to that of the GaAs surface after thermal treatment, while the other regime, which appears for a growth time between 100 and 300 s, shows a transition to a 3D growth. During the last regime the correlation length analysis prove the self-affine characteristic of the growing layer surface. Combining a diffraction technique with real space microscopy has been crucial in understanding the influence of the crystallographic characteristics of the growing layers, in our case the twin formation, and on their morphological scaling properties. In this way, AFM and X-ray texture analysis have been shown to be powerful complementary tools in the study of the MOCVD growth process, especially in the case when in situ monitoring is not available.

Acknowledgements This work was partially supported by the ! Generalitat Valenciana (GV01-536), Comision Interministerial para Ciencia y Tecnolog!ıa ! Industrial de Optica (1FD97-0086) and Asociacion (AIDO). We want also acknowledge Francisco Fabregat-Santiano from UJI University (Univer! and Ma Jesus ! Iba! n˜ez and sitat Jaume I de Castello) ! Monica Vicent from ITC (Institut de Tecnologia ! for the Cera" mica de la Universitat de Castello) help with the profilometer measurements.

ARTICLE IN PRESS 68

I. Mora-Sero! et al. / Journal of Crystal Growth 257 (2003) 60–68

References [1] A.W. Brinkman, CdTe-based solar cells, in: P. Capper (Ed.), Properties of Narrow Gap Cadmium-Based Compounds, 1994, pp. 591–597. [2] T. Aramoto, S. Kumazawa, H. Higuchi, T. Arita, S. Shibutani, T. Nishio, J. Nakajima, A. Hanafusa, T. Hibino, K. Omura, H. Ohyama, M. Murozomo, Jpn. J. Appl. Phys. 36 (1997) 6304. [3] J. Rams, N.V. Sochinskii, V. Mun˜oz, J.M. Cabrera, Appl. Phys. A 71 (2000) 277. [4] J.F. Butter, CdTe and CdZnTe as gamma-ray and X-ray detector materials, in: P. Capper (Ed.), Properties of Narrow Gap Cadmium-based Compounds, INSPEC IEE, London, 1994, pp. 587–590. [5] J. Tunnicliffe, S.J.C. Irvine, O.D. Dosser, J.B. Mullin, J. Crystal Growth 68 (1984) 245. [6] N.V. Sochinskii, V. Mun˜oz, S. Bernardi, J.I. Espeso, E. Alves, M.F. da Silva, J.C. Soares, C. Mar!ın, E. Di!eguez, J. Crystal Growth 179 (1997) 585. [7] S.J.C. Irvine, Properties of Narrow Gap Cadmium-based Compounds. in: P. Capper (Ed.), INSPEC IEE, London, 1994. [8] S.J.C. Irvine, A. Stafford, M.U. Ahmed, P. Prete, R. Berrigan, Prog. Crystal Growth Charact. 35 (2–4) (1997) 177. [9] R.L. Moon, J. Crystal Growth 170 (1997) 1. [10] S.J.C. Irvine, J.B. Mullin, J. Giess, J.S. Gough, A. Royle, J. Crystal Growth 93 (1988) 732. [11] H. Nishino, I. Sugiyama, Y. Nishijima, J. Appl. Phys. 80 (6) (1996) 3238. [12] A.M. Keir, A. Graham, S.J. Barnett, J. Giess, M.G. Astles, S.J.C. Irvine, J. Crystal Growth 101 (1990) 572. [13] A. Wasenczuk, A.F.W. Willoughby, P. Mackett, E. O’Keefe, P. Capper, C.D. Maxey, J. Crystal Growth 159 (1996) 1090. [14] Yu.Yu. Loginov, P.D. Brown, C.J. Humphreys, Phys. Solid State 38 (2) (1996) 272. [15] T.T. Cheng, M. Aindow, I.P. Jones, J.E. Hails, D.J. Williams, J. Crystal Growth 154 (1995) 251. [16] T.T. Cheng, M. Aindow, I.P. Jones, J.E. Hails, A. Graham, J. Giess, D.J. Williams, M.G. Astles, Compound semiconductor epitaxy symposium, Mater. Res. Soc. 340 (1994) 581. [17] M. Ekawa, K. Yasuda, S. Sone, Y. Sugiura, M. Saji, A. Tanaka, J. Appl. Phys. 67 (11) (1990) 6865. [18] P.Y. Lu, L.M. Williams, S.N.G. Chu, J. Vac. Sci. Technol. A 4 (1986) 2137.

[19] P.L. Anderson, J. Vac. Sci. Technol. A 4 (1986) 2162. [20] J. Geiss, J.S. Gough, S.J.C. Irvine, J.B. Mullin, G.W. Blackmore, Mater. Res. Soc. Symp. 90 (1987) 389. [21] R.D. Feldman, D.W. Kisker, R.F. Austin, K.S. Jeffers, P.M. Briderbaugh, J.Vac. Sci. Technol. A 4 (1986) 2234. [22] P.D. Brown, J.E. Hails, C.J. Russell, J. Woods, Appl. Phys. Lett. 50 (1987) 1144. [23] J.B. Mullin, S.J.C. Irvine, Proc. Crystal Growth Charact. 29 (1995) 217. [24] H. Nishino, Y. Nishijima, J. Crystal Growth 173 (1997) 330. [25] Y.K. Kim, J.Y. Lee, H.S. Kim, J.H. Song, S.H. Suh, J. Crystal Growth 192 (1998) 109. [26] J.E. Hails, G.J. Russell, A.W. Brinkman, J. Woods, J. Crystal Growth 79 (1986) 940. [27] M. Oron, A. Raizman, H. Shtrikman, G. Cinader, Appl. Phys. Lett. 52 (13) (1988) 1059. [28] I.V. Sabinina, A.K. Gutakovskii, Y.G. Sidorov, V.D. Kuzmin, Phys. Stat. Sol. A 126 (1991) 181. [29] G. Cinader, A. Raizman, A. Sher, J. Vac. Technol. B 9 (3) (1991) 1634. [30] R. Triboulet, A. Tromson-Carli, D. Lorans, T. Nguyen Duy, J. Electron. Mater. 22 (8) (1993) 827. [31] A. Tromson-Carli, G. Patriarche, R. Druilhe, A. Lusson, Y. Marfaing, R. Triboulet, P.D. Brown, A.W. Brinkman, Mater. Sci. Eng. B 16 (1993) 145. [32] A. Tromson-Carli, A. [email protected], R. Druilhe, C. Grattepain, R. Triboulet, Mater. Lett. 23 (1995) 93. [33] J.B. Mullin, S.J.C. Irvine, Proc. Crystal Growth Charact. 29 (1995) 217–252. [34] A.L. Barabasi, H.E. Stanley, Fractal Concepts in Surface Growth, Cambridge University Press, Cambridge, 1995. [35] T.T. Cheng, M. Aindow, I.P. Jones, J.E. Hails, D.J. Williams, M.G. Astles, J. Crystal Growth 135 (1994) 409. [36] W.F. Kolbe, D.F. Ogletree, M. Salmeron, Ultramicroscopy 369 (1992) 42. [37] H. Nagai, J. Appl. Phys. 45 (1974) 3789. [38] G. Cohel-Solal, F. Bailly, M. Barb!e, Appl. Phys. Lett. 49 (22) (1986) 1519. [39] R. Tena-Zaera, Estudi del proc!es de creixement de CdTe, HgTe i Hg(1x)CdxTe per MOCVD, simulacio! num!erica, Universitat de Val!encia D.E.A. work, 2001. [40] J.A. Venables, J.Vac. Sci. Technol. B 4 (1986) 870. [41] S.J.C. Irvine, J. Bajaj, J. Crystal Growth 145 (1994) 74. [42] K.H. Shim, J. Myoung, O. Gluschenkov, K. Kim, C. Kim, K. Robison, Jpn. J. Appl. Phys. 37 (3B) (1998) L313.