Texture manipulation of CuInSe2 thin films

Texture manipulation of CuInSe2 thin films

Thin Solid Films 361±362 (2000) 167±171 www.elsevier.com/locate/tsf Texture manipulation of CuInSe2 thin ®lms Miguel A. Contreras a,*, Brian Egaas a,...

141KB Sizes 0 Downloads 28 Views

Thin Solid Films 361±362 (2000) 167±171 www.elsevier.com/locate/tsf

Texture manipulation of CuInSe2 thin ®lms Miguel A. Contreras a,*, Brian Egaas a, David King a, Amy Swartzlander a, Thorsten Dullweber b a


National Renewable Energy Laboratory, 1617 Cole Boulevard, Golden, CO 80401, USA Institut fuÈr Physikalische Elektronik, UniversitaÈt Stuttgart, Pfaffenwaldring 47, 70569 Stuttgart, Germany

Abstract We present a growth method that allows tailoring of preferred orientation in CuInSe2 thin-®lms grown on Mo-coated soda-lime glass substrates. In particular, ®lms exhibiting a (204) preferred orientation are demonstrated, in addition to already reported (112) and randomly oriented ®lms. Effects of substrate composition, growth temperature, and ®nal ®lm composition on texture phenomena are presented. In general, we ®nd that texture is highly dependent on growth temperature, substrate material and, in the case of Mo-coated soda-lime glass substrates, the structural properties of the Mo layer. We provide evidence for the attainment of such structural orientation, and we present a growth model to explain the mechanism allowing such modi®cations. q 2000 Elsevier Science S.A. All rights reserved. Keywords: Texture; Preferred orientation; CIS; CIGS; CuInSe2

1. Introduction In general, physical vapor deposition methods [1±3] operating in a co-deposition mode in CuInSe2 (CIS) thin-®lm growth have traditionally yielded randomly oriented ®lms and/or, in some cases, ®lms with a (112) preferred orientation [4,5]. By deviating from the reaction pathway associated with co-deposition, we have been able to tailor the structural properties of CIS ®lms grown on Mo-coated sodalime glass (SLG) substrates. Our approach involves a sequential process where initially an indium selenide precursor layer is fabricated and into which Cu and Se are subsequently incorporated at a given substrate temperature. We have found that within a certain range of substrate temperature, such a reaction pathway allows us to change the preferred orientation of CIS ®lms. This approach to CIS compound formation is actually the basis for the `threestage' process reported elsewhere [6]. In general, we can represent this two-step reaction pathway as the unbalanced equations In…g† 1 Se…g† ! Inx Sey …s†


Inx Sey . …s† 1 Cu…g† 1 Se…g† ! CuInSe2 …s†


The subscripts x and y in Eqs. (1) and (2) are meant to be generic for now; nevertheless, in a later section, we will qualify the crystallographic characteristics of such precur* Corresponding author. Tel.: 11-303-384-6478; fax: 11-303-3846430. E-mail address: [email protected] (M.A. Contreras)

sors because their structural properties are the key to modifying the structure of CIS.

2. Experimental and results Ê /s) onto the InxSey precursor layer Cu evaporation (,5 A has been done at different substrate temperatures while Ê /s). In general, for maintaining a Se overpressure (,25 A some Mo/SLG substrates, we ®nd that when Cu is incorporated using a substrate temperature (Ts) of ,400±5008C, the resulting CIS ®lms consistently present a (220/204) preferred orientation. On the other hand, when Cu deposition is done at T s . 5008C, the ®lms begin to show random orientation or somewhat of a (112) texture with increased substrate temperature. Fig. 1 shows the X-ray u /2u scans for CIS ®lms obtained in this fashion on selected substrates. We note that the observation of (220/204) preferred orientation, from the u /2u scans, has been veri®ed by pole-®gure analysis (PFA). Such PFA (not shown) proves that the higher intensities of the (220/204) re¯ections in the u /2u scans are not due to ordering or other phenomena that can lead to similar behavior of X-ray diffraction (XRD) data. We point out that attainment of an orientation de®nition function (ODF) for these ®lms is hindered by the overlapping qualities of the most intense re¯ections and the relatively low intensities of the non-overlapping re¯ections for this material system. Therefore, a de®nite orientation, either (220) or (204), must be deduced by other means. We discuss this issue in a later section of this paper.

0040-6090/00/$ - see front matter q 2000 Elsevier Science S.A. All rights reserved. PII: S 0040-609 0(99)00778-6


M.A. Contreras et al. / Thin Solid Films 361±362 (2000) 167±171

204) CIS ®lm shows considerably less Na in its surface as compared to the other rather randomly oriented CIS ®lm. The above experimentation indicates that, in addition to the growth conditions previously mentioned, the attainment of the (220/204) texture is strongly dependent on the structural properties of the Mo layer. Properties like morphology, grain size, and stress state are variables that undoubtedly affect the CIS nucleation and growth process, in addition to the diffusion of Na from the glass, through the Mo layer, and into the CIS ®lm. To date, it is unclear which variable is more relevant for the attainment of the texture. Fig. 1. XRD scans of CuInSe2 thin ®lms. CIS ®lm grown from Cu diffusion into indium selenide precursor layer (a) on Mo/SLG at T s , 4608C, (b) on Mo/SLG and T s , 6008C, and (c) on bare SLG substrate at T s , 4608C. Data taken at room temperature using the Ka 1 line of Cu.

2.1. CIS on selected substrates For amorphous substrates, such as bare SLG and 7059 Corning glass, the resulting CIS ®lms always present a strong (112) preferred orientation, regardless of the substrate temperature at which Cu is introduced (see Fig. 1). A similar behavior was observed for CIS grown on a highly (100) oriented Mo foil. However, for polycrystalline alumina (plate), the resulting ®lms tended to be randomly oriented under the growth conditions leading to the (220/ 204) texture in Mo/SLG substrates. Randomly oriented CIS was also obtained for Mo-coated (,1 mm) alumina and Mo coated Si (100) substrates. We have found that the (220/204) preferred orientation could only be achieved on speci®c types of Mo/SLG substrates. The Mo layer in such substrates is rather dense, almost free of pinholes, and nearly free of sodium (oxide) in their surface. We note that a correlation between Na and the (112) preferred orientation of CIS has been reported [7]; additionally, we have previously determined that Na, when present at a critical level, can hinder the attainment of the (220/204) preferred orientation [8]. We have analyzed, using XRD, scanning electron microscopy (SEM), Auger electron spectroscopy (AES), and X-ray photoelectron spectroscopy (XPS), two different Mo/SLG substrates that do not incorporate a Na barrier, one that permits attainment of (220/204) texture, and another that does not. The substrate allowing the (220/204) texture was characterized in its as-deposited state by larger grain sizes, a lesser degree of tensile stress (denser ®lm), and ineffectual amounts of Na on its surface. The other as-deposited Mo/ SLG substrate shows about 5 times more Na (see Table 1). After simultaneous CIS growth on each substrate, the (220/

2.2. Cu content and texture Preliminary data on the effect of Cu content and texture of CIS suggests that the (220/204) preferred orientation can be attained for Cu-poor materials (Cu=In , 1). In fact, we succeeded at growing CuIn3Se5 ®lms with (220/204) preferred orientation. However, Cu-rich ®lms under similar growth conditions tend to be randomly oriented. Such preliminary data suggest that copper selenide phases present during growth of Cu-rich CIS may hinder the attainment of the (220/204) preferred orientation. 3. Indium selenide precursor layers In an attempt to structurally identify the InxSey precursors, we have quenched the ®lms to `freeze' the actual structures into which Cu is diffused. In an ideal experimental setup, such characterization should be done in situ; but our capabilities are limited, and such in situ analysis cannot be accomplished. In general, the initial indium selenide precursor layer Ê /s) (,0.5 mm thick) is formed by co-evaporation of In (7 A Ê /s) at a given substrate temperature. We have and Se (,25 A characterized precursor ®lms grown (and quenched) at a substrate temperature ranging from ,460 to ,6008C. For the purpose of quenching, the substrate heater and the Se source are turned off immediately after terminating In evaporation. Of particular interest are the extreme points of this substrate temperature range (values of 4608C and 6008C) because Cu incorporation into the indium selenide precursor layers at these temperatures results in CIS ®lms that are clearly distinct in their structural properties. We ®nd that in both situations, quenching at 4608C and at 6008C, and for all substrates (Mo/SLG and bare SLG), the resulting precursor layers can be described as a hexagonal form of In2Se3. These precursor layers match very well to the g phase of In2Se3 as described in JCPDS card 40-1407

Table 1 Surface XPS semi-quantitative compositional analysis for two as-deposited Mo/SLG substrates Element

Mo (at.%)

Na (at.%)

O (at.%)

C (at.%)


Substrate 1 Substrate 2

19.1 19.8

1.2 5.8

42.5 49.1

37.2 25.3

Allows (220/204) texture Does not allow (220/204) texture

M.A. Contreras et al. / Thin Solid Films 361±362 (2000) 167±171

Ê , c ˆ 19:382 A Ê ). However, there is a clear (a ˆ 7:1286 A difference in the preferred orientation between them (see Fig. 2). The In2Se3 precursor layer obtained in Mo/SLG substrates by quenching from ,4608C is highly (100) oriented (a oriented). On the other hand, the precursor obtained on Mo/SLG by quenching from 6008C is (001) oriented (c oriented), but not as highly c oriented as the precursor on SLG substrates obtained from either initial quenching temperature. As before, the preferred orientation arguments for the precursor layers have been veri®ed by PFA (not shown). From this analysis, we deduce that a (220/204) texture of CIS ®lms is obtained when Cu incorporation is carried out onto a highly a oriented g -In2Se3(H) precursor, and the (112) texture of CIS is obtained when Cu incorporates into a highly c oriented g -In2Se3(H) precursor layer. Similarly, the randomly oriented CIS is a result of the lesser degree of preferred orientation from the g -In2Se3(H) precursor layer. We note that no space group (S.G.) has been assigned to the g phase in JCPDS card 40-1407; however, the Pearson's Handbook of Crystallographic Data lists three In2Se3(H) compounds that closely match the g phase reported in the JCPDS database. Two of these compounds are assigned S.G. 169 and the other S.G. 170. Our modeling of such structures indicates that S.G. 170 describes more accurately our precursor layers. 4. Preferred orientation in CIS An accurate determination of orientation, either (220) or (204), is hindered by the fact that the most intense XRD re¯ections involved overlap and that the non-overlapping re¯ections are too weak for pole-®gure analysis using our 2 kW Cu anode diffractometer. However, it is possible to deduce the most likely orientation based on a combination of pole ®gures and high-quality/resolution u /2u scans on speci®c re¯ections. The PFA unequivocally indicates a (220) or (204) preferred orientation, i.e. most of the grains in the ®lm have grown with their {220} or {204} planes parallel to

Fig. 2. XRD scans for g -In2Se3(H) precursor layers (a) on Mo/SLG quenched from 4608C, (b) on Mo/SLG quenched from 6008C, (c) on bare SLG quenched from 4608C, and (d) peak labels corresponding to g -In2Se3 JCPDS card 40-1407.


Fig. 3. u /2u XRD scans taken at 0.018 steps on the (112) and (220/204) re¯ections. Labeled vertical lines represent the peak positions for a powder sample (JCPDS card 40-1487), and the unlabeled vertical lines are the expected positions corrected for compressive stress state and/or ®lm composition.

the surface of the substrate. Let us now consider the u /2u scans. A high-quality u /2u XRD scan on the (112) and (220/ 204) re¯ections reveals that the CIS ®lm is in a compressive stress state. This observation is founded on the comparison of the unstrained powder diffraction pattern (JCPDS card 40-1487) and the peak position for the experimental data (see Fig. 3). Because the experimental XRD data reveal a systematic shift toward higher values of 2u , a lattice parameter calculation based on it would yield a smaller value than that of an unstrained powder sample. We note that composition is a factor that in¯uences lattice and unit cell size (Cu vacancies), but for the purposes of this discussion, the true reason for the XRD peak shift is not relevant. Nevertheless, the compressive stress is common to CIS ®lms, and it is characteristic of evaporated samples [9]. If we correct the peak position to account for the ®lm compressive stress-state (or composition, whatever the case may be), we ®nd that the highest intensity in the (220/204) re¯ection is better aligned with the (204) re¯ection and not with the (220), see Fig. 3. Let us now consider the relative intensities of those and other re¯ections (see Fig. 4) and the expected intensities (vertical lines in the ®gures) from a randomly oriented powder sample (JCPDS card 40-1487). From Fig. 4, we see that in a randomly oriented CIS ®lm, the intensity of the (211) re¯ection is expected to be higher than that of the (101) and (103) re¯ections; that is, I…211† . I…101† . I…103†. However, this is not the case in the experimental data, and the oriented ®lm shows practically the opposite, i.e., I…211† , I…101† , I…103†. This situation can be accounted for in terms of the angles that

Fig. 4. XRD scan for textured CIS ®lm and expected intensities (vertical lines) from a randomly oriented powder sample (JCPDS card 40-1487).


M.A. Contreras et al. / Thin Solid Films 361±362 (2000) 167±171

those planes ({211}, {101}, and {103}) form with the {204} planes, lending further validity to a (204) texture. Because the experimental data already include the premises of structure factors, multiplicity, and absorption, the discrepancies in peak intensity of the above singlets must be attributed to aspects of preferred orientation. If a priori we assume a (204) preferred orientation, then the (103) re¯ection can be expected to be enhanced as compared to the (101) and (211) re¯ections, because of the smaller angle (11.328) that {103} planes form with {204} planes (see Table 2). Identical arguments can explain the lower intensity of (101) and (211) re¯ections. On the other hand, if the preferred orientation were a priori assumed to be (220), then the expected intensity for the (103) re¯ection should be much diminished because of the high angle (66.838) that {103} planes form with {220} planes. Additionally, the (211) re¯ection should be enhanced because of its smaller angle with {220} planes (22.188). Because that is not the case, the above analysis suggests that the preferred orientation is (204) in nature, not (220). 5. A growth model for the (112) and (204) preferred orientation in CIS Modeling of the precursor layers and CIS ®lms has been performed in real and reciprocal space. The In2Se3(H) Ê, precursor materials are assigned S.G. 170, a ˆ 7:11 A Ê c ˆ 19:3 A, and atomic coordinates as described in [10]. Ê and For CIS, we use S.G. 122 with a ˆ 5:782 A Ê c ˆ 11:619 A and the ideal atomic coordinates for the chalcopyrite structure (CuFeS2). As mentioned before, the CIS ®lms grown on bare SLG always result in a highly (112) preferred orientation, regardless of the growth conditions. We have determined that the In2Se3(H) precursor layers in these cases are highly c oriented. Furthermore, we ®nd through our modeling that there is an equivalent symmetry between the {001} planes, or surfaces, of In2Se3(H) and that of {112} planes of CIS. Similarly, for the {204} plane of CIS ®lms, there is an equivalent symmetry between such a surface, or more

Table 2 Calculated angles (8) between selected crystallographic planes in CIS chalÊ and c ˆ 11:619 A Ê copyrite structure with a ˆ 5:782 A Plane1\plane 2






(204) (220)

35.33 35.14

11.32 66.83

18.41 50.72

39.46 22.18


(204) 59.92

generally {102}, and that of {100} planes in In2Se3(H). This equivalent symmetry can be appreciated in real space, but it is more easily seen in reciprocal space. Fig. 5 shows the modeling of reciprocal-space features of the surfaces involved. The hexagonal nature of the {001} planes in the precursor layer can also be seen in the {112} planes of CIS; similarly, the rectangular character of the {100} planes of the precursor can be seen in the {102} planes of CIS. Considering that (1) the experimental work suggests the texture in CIS is intimately related to the initial texture of the precursor ®lm, and (2) the modeling indicates an equivalent symmetry between the precursor and the resulting CIS ®lm. We propose that, for the growth process previously described, nucleation and growth of CIS follow the preexisting symmetry and characteristics of the precursor layer. Intuitively, the system requires less energy to maintain the existing symmetry than changing it to a different form or crystallographic arrangement. 6. Conclusions We have presented a growth process that allows tailoring the structural properties (preferred orientation) of CIS thin ®lms. The sequential evaporation process involves the growth of an indium selenide precursor layer into which Cu and Se are subsequently incorporated at a given substrate temperature. We demonstrated that randomly oriented ®lms and (112) and (204) oriented thin ®lms can be produced by careful selection of substrate and growth conditions. In particular, ®lms with a (204) preferred orientation are demonstrated on speci®c Mo/SLG substrates. The attainment of (204) preferred orientation is strongly dependent on the structural properties of the Mo layer.

Fig. 5. Reciprocal space, not to scale, for (a) In2Se3(H) {001}, (b) CIS {112}, (c) In2Se3(H) {100}, and (d) CIS {102}. An equivalent symmetry is observed between precursor layer and CIS.

M.A. Contreras et al. / Thin Solid Films 361±362 (2000) 167±171

These properties are related to issues such as morphology, grain size, and stress state. Furthermore, the presence of Na above a certain critical level hinders the attainment of the (204) preferred orientation, rendering the resulting ®lms toward a randomly oriented structure or a (112) preferred orientation with increased substrate temperature. Such observations suggest that Na lowers the surface energy of CIS during nucleation and growth, leading to the most stable and favorable (112) preferred orientation. Preliminary data on the effect of Cu content on texture suggest that copper selenide phases present during growth of Cu-rich CIS ®lms hinder the attainment of the (204) preferred orientation. Also, we have presented a growth model to explain how such modi®cations occur and the close relationship between precursor layers and resulting CIS ®lm properties. The growth model stipulates that CIS nucleation and growth, via the method described, follow the existing symmetry and structural properties of the indium selenide precursor layer. Acknowledgements We wish to thank R. Matson for SEM work and R. Nou®, K. Ramanathan, and R. Battacharya for technical discussions and critical review of the manuscript. Thanks also to our technicians J. Dolan and J. Keane for work related to


vacuum equipment and thin-®lm deposition. This work was supported by the U.S. Department of Energy under Contract No. DE-AC36-98-GO10337.

References [1] W.S. Chen, J.M. Stewart, B.J. Stanbery, W.E. Devaney, R.A. Mickelsen, Proc. 19th IEEE Photovoltaic Specialists' Conf., IEEE, New York, 1987, p. 1445. [2] J. HedstroÈm, H. Ohlsen, M. BodegaÊrd, A. Kylner, Proc. 23rd IEEE Photovoltaic Specialists' Conf., Louisville, KY, IEEE, New York, 1993, p. 364. [3] T. Nakada, K. Migita, S. Niki, A. Kunioka, Jpn. J. Appl. Phys. 34 (1995) 4715. [4] W.N. Shafarman, R.W. Birkmire, M. Marudachalam, B.E. McCandless, J.M. Schultz, AIP Conf. Proc. 394, NREL/SNL PV Program Review, Lakewood, 1996, p. 123. [5] H. Dittrich, H.W. Schock, in: J.H. Werner, H.P. Strunk (Eds.), Polycrystalline Semiconductors II, Springer, Berlin, 1991, p. 432. [6] M.A. Contreras, J.R. Tuttle, A. Gabor, et al., Record 24th IEEE Photovoltaics Specialists' Conf., Waikoloa, Hawaii, IEEE, New York, 1994, p. 68. [7] M. BodegaÊrd, L. Stolt, J. HedstroÈm, Proc. 12th European Communities Photovoltaic Solar Energy Conf. and Exhib., 1994, p. 1743. [8] M.A. Contreras, B. Egaas, P. Dippo, et al., Proc. 26th IEEE Photovoltaic Specialists' Conf., Anaheim, CA, IEEE, New York, 1997, p. 359. [9] R. Chakrabarti, A.B. Maity, R. Pal, D. Bhattacharyya, S. Chaudhuri, A.K. Pal, Phys. Status Solidi A 161 (1997) 67. [10] S.A. Semilitov, Sov. Phys.-Crystall. 5 (1961) 673.