Journal of Physics and Chemistry of Solids 64 (2003) 1499–1504 www.elsevier.com/locate/jpcs

Diffusion of indium and gallium in Cu(In,Ga)Se2 thin film solar cells O. Lundberga,*, J. Lua, A. Rockettb, M. Edoffa, L. Stolta a˚

Angstro¨m Solar Center, Uppsala University, P.O. Box 534, SE-751 21 Uppsala, Sweden b University of Illinois, 1-107 ESB, 1101 W. Springfiled Ave., Urbana, IL 61801, USA

Abstract The diffusion of indium and gallium in polycrystalline thin film Cu(In,Ga)Se2 layers has been investigated. Bilayer structures of CuInSe2 on top of CuGaSe2 and vice versa have been fabricated in both a Cu-rich and Cu-poor process (in relation to the ideal stoichiometry). In each process molybdenum coated soda-lime glass with and without a sodium barrier was used. These bilayers were analyzed with secondary ion mass spectrometry, X-ray diffraction, scanning electron microscope and transmission electron microscope equipped with energy dispersive X-ray spectroscopy. It was found that the grain boundary diffusion was not significantly higher than the diffusion inside the grains, also for Cu-rich layers. The diffusion is suggested to mainly proceed via vacant metal sites in the lattice structure. In sodium free films a higher diffusion into the bottom layers, compared to films with sodium, was seen in all cases. This observation was explained with a larger number of vacancies, that facilitates indium and gallium diffusion, in the sodium free films. The difference in diffusion between indium in CGS layers and gallium in CIS layers, in both Cu-rich and Cu-poor processes, was small for layers with sodium. q 2003 Elsevier Ltd. All rights reserved. Keywords: D. Diffusion

1. Introduction Cu(In,Ga)Se2 (CIGS) thin films have attracted great interest owing to the high conversion efficiency of solar cells made from this material [1]. One of the special qualities of the CIGS material is its variable band gap. By increasing the Ga/(Ga þ In) ratio from 0 to 1 the band gap is increased from 1.02 to 1.66 eV. This quality can be used, not only to adjust the band gap to the solar spectrum, but also to vary the band gap as a function of the depth in the CIGS film and thereby further improve the solar cell performance. In many of the different CIGS fabrication techniques, an in depth variation of the Ga/(Ga þ In) ratio is introduced, intentionally or unintentionally. For example in the selenisation processes, the metals are first deposited and thereafter selenised to form CIGS, a higher Ga/(Ga þ In) ratio is spontaneously obtained towards the back contact [2]. In coevaporation processes a higher Ga/(Ga þ In) ratio towards the back contact is sometimes introduced intentionally by * Corresponding author. Fax: þ 46-18-555095. E-mail address: [email protected] (O. Lundberg).

starting the deposition by evaporating pure CuGaSe2, or CIGS with a very high Ga content and then continue by evaporating CIGS with lower Ga content. This is commonly referred to as Ga-grading and has been shown to increase the efficiency of the resulting solar cells [3,4]. In the processes mentioned above, the diffusion of gallium and indium will affect the resulting depth profile of the Ga/(Ga þ In) ratio and thus the resulting performance of the solar cell device. An understanding of the properties affecting this diffusion as well as the diffusion mechanisms itself is therefore important. The diffusion of indium and gallium has been investigated in a number of earlier studies [5 – 10]. A common belief in the CIGS community is that the diffusion of gallium and indium is significantly increased if the CIGS film is grown under Cu-rich conditions. In our Ga-grading experiments we have, however, not seen this large difference in the diffusion between CIGS films grown under Cu-rich and Cu-poor conditions [4,11]. The paper mostly referred to concerning the high diffusion in Cu-rich grown CIGS is Walter and Schock [5]. The experiments made in this study were, however, performed using

0022-3697/03/$ - see front matter q 2003 Elsevier Ltd. All rights reserved. doi:10.1016/S0022-3697(03)00127-6

1500

O. Lundberg et al. / Journal of Physics and Chemistry of Solids 64 (2003) 1499–1504

borosilicate glass (Corning 7059) substrates, resulting in essentially sodium-free CIGS films. The commonly used substrate material for device quality CIGS is soda-lime glass, which results in a significant sodium concentration in the CIGS films, clearly correlated with a different microstructure [12]. In the work presented here we investigate how the diffusion of gallium and indium is affected by different growth conditions, Cu-rich and Cupoor as well as with and without sodium, in order to obtain additional information about the diffusion mechanisms.

2. Experimental In order to study the diffusion of indium and gallium in CIGS films, bilayers of CuInSe2 (CIS) on top of CuGaSe2 (CGS) and vice versa were fabricated. The bilayer structures were fabricated with co-evaporation both in a Cu-rich ([Cu] . ([Ga] þ [In]) and in a Cu-poor process ([Cu] , ([Ga] þ [In]) with the evaporation rates according to Fig. 1. The evaporation time was 30 min, resulting in a total film thickness of around 1 mm. Also single CGS and CIS layers were fabricated under the same conditions. These are not described in detail in this paper, but they are discussed in Bodega˚rd et al. [13]. A mass spectrometer controlled the rates of the metals, evaporated from open boat sources, while the Se source was heated to a constant value, resulting in a rate 3– 5 times higher than the rate required for stoichiometric CIS, respectively, CGS formation. The substrate temperature was held constant at 510 8C during the deposition. In each CIS/CGS and CGS/CIS run, molybdenum coated soda-lime glass substrates, with and without a sodium barrier of Al2O3, was used. Consequently, the two bilayer structures, CIS on CGS and CGS on CIS, were fabricated in four different conditions—Cu-rich with and without sodium and Cu-poor with and without sodium. CIGS bulk composition values and compositional depth profiles were determined with energy dispersive X-ray spectroscopy (EDS) and secondary ion mass spectrometry (SIMS), respectively. Bulk structural analysis was made by X-ray diffraction (XRD). In order to estimate the grain size, scanning electron microscope (SEM) was used and

Fig. 1. Evaporation profiles for CuInSe2 on CuGaSe2 bilayers for both a Cu-rich and a Cu-poor process.

a transmission electron microscope (TEM) equipped with EDS was used for local (micro) compositional determination of the bilayers.

3. Results In Fig. 2 results from the eight bilayers are presented as follows: 1. In the first column the order of the CGS and CIS layers and under which condition they were grown is specified. The Cu/(Ga þ In) ratio from EDS measurements is also given. Since the sodium barrier, Al2O3, does not completely prevent sodium diffusion, ‘No Na’ in reality means sodium concentrations in the order of 1000 times lower than the sodium concentration in CIGS layers grown on soda-lime glass without a sodium barrier. 2. The second column shows SEM cross sectional images. The bar corresponds to 1 mm. 3. In the third column the Ga/(Ga þ In) and the In/ (In þ Ga) ratios are presented, calculated from normalized SIMS data (0 , In,Ga , 1), as a function of sputter time, i.e. depth. The signal from the molybdenum layer is also shown in order to illustrate when the back contact is reached. 4. The fourth column shows XRD spectra of the (112) peak profile. The dotted lines indicate the peak values of the single layers, CGS at 27.658 and CIS at 26.658. The deviation of the current peak position, from these dotted lines, can be used to estimate the intermixing between the CGS and CIS layers. From the Cu/(Ga þ In) ratio we can see that all layers are relatively close to stoichiometry (Cu/(Ga þ In) ¼ 1). By comparing the SEM pictures, we see that the layers grown under Cu-rich conditions in general have substantially larger grains than the layers grown under Cu-poor conditions. For bilayers grown under the same conditions, the ones with CIS as a bottom layer have larger grains than the ones with CGS as a bottom layer. (The respective part of the bilayer will be referred to as the CIS or CGS layer, even if they intermix after, it would not remain pure CIS or CGS anymore. ‘Bottom’ refers to the layer that is grown directly on the back contact). In most of the CIGS films a two-layer structure can be seen, where the CGS layer has smaller grains than the CIS layer. In the Cu-rich sodium free film with a CGS bottom layer, this two-layer structure is, however, less obvious. For the corresponding film, with sodium, there is an apparent small-grained layer at the bottom, but it is thinner than the expected 0.5 mm. The SIMS depth profiles and the XRD data together give a good picture of how much the bilayers have intermixed at the different growth conditions. In order to study the diffusion of gallium and indium in these bilayers the bottom layers are more suitable than the top layers, since

O. Lundberg et al. / Journal of Physics and Chemistry of Solids 64 (2003) 1499–1504

1501

Fig. 2. All bilayer presented with their Cu/(In þ Ga) ratio (column 1), a SEM picture (column 2), a SIMS profile with the sputter time on the yaxis and the In/(In þ Ga) ratio or Ga/(Ga þ In) ratio on the x-axis (column 3) and an XRD plot of the (112) peak profile. The dotted line in the SIMS profile represents the measurement signal from the molybdenum back contact and the dotted lines in the XRD plot indicate where the single CGS and CIS layers had their peak values.

1502

O. Lundberg et al. / Journal of Physics and Chemistry of Solids 64 (2003) 1499–1504

intermixing occurs primarily via diffusion in an already grown layer. In the top-layer, however, the intermixing is more complex, since it proceeds during the growth of the film. Hence, from the lower half of the SIMS depth profiles and the shift and shape of the XRD peak originating from the bottom layer, we can make the following observations concerning diffusion of indium and gallium in polycrystalline CIGS films: † The diffusion of both indium and gallium is higher in layers grown with no access of sodium. The largest diffusion is seen in the sodium free layers grown under Cu-rich conditions. † The diffusion of indium into the CGS bottom layers is similar to diffusion of gallium into the CIS bottom layers, under the same conditions. † The diffusion in layers grown under Cu-rich and Cu-poor conditions, with access of sodium, is similar. A slightly larger amount of indium has diffused into the upper part of the CGS bottom layer in the Cu-rich case as compared to the Cu-poor case. This is seen from the broad XRD peak and the relatively high In/(In þ Ga) ratio at this point. Concerning the top layers, there is a significant difference in the degree of intermixing between the CGS and CIS top layers. In the bilayer structure with a CGS layer grown on top of a CIS layer, much less indium is intermixed in the CGS top layer than the amounts of gallium intermixed in the CIS top layer in the reverse structure. This is seen from the ’S-shaped’ SIMS depth profiles and the less shifted, thinner XRD peaks for these top layers. In order to investigate how important the grain boundary diffusion is in relation to the intra-grain diffusion, we analyzed the distribution of In and Ga within a large grain. The size of the grain needs to be comparable to the distance over which we observe a compositional difference due to diffusion. The bilayers with CIS bottom layers, grown under Cu-rich conditions, are most suitable since they have the largest grains. In Fig. 3, a TEM picture of one grain in a cross-sectional sample from the Cu-rich, sodium-free,

bilayer with CIS in the bottom part is shown. As can be seen, one vertical and one horizontal scan through the grain have been performed. In the graph to the right the compositional variations along these lines are presented. There is only a vertical compositional gradient through the grain, horizontally the composition is uniform.

4. Discussion We start by discussing the influence of grain boundaries on the diffusion in the bottom layers. In Fig. 2 we observed that the diffusion into the CIS and CGS bottom layers, grown under both Cu-rich and Cu-poor conditions and with access of sodium was similar. In the SEM pictures, we could see that the variation in grain size between these layers was large. For example, the CIS bottom layer grown under Curich conditions has up to 10 times larger grains than the CGS bottom layer grown under Cu-poor conditions. Despite this, the diffusion into these layers is almost the same, which indicates that the significantly larger grain boundary area, in the small-grained layers, not has any influence on the diffusion. In other words, the grain boundary diffusion does not seem to be significantly higher than the diffusion in the grains. Another indication of this comes from the compositional gradients in Fig. 3. If the grain boundary diffusion was much faster than the intragranular diffusion, it would be expected that large CIS and CGS grains, positioned in the bottom layer, would have a higher concentration of the diffusing element, gallium or indium, close to the grain boundary, and a decreased concentration towards the middle of the grain. In Fig. 3 no such horizontal gradient is seen, only a vertical compositional gradient through the grain is present. In Schroeder et al. [6] the gallium diffusion in single crystal CIS layers with different Cu/In ratios was investigated. The results were compared with Marudachalam et al. [9], where the diffusion coefficients for indium and gallium in polycrystalline CIGS films grown under Cu-poor conditions were determined, and the diffusion coefficients were similar in single- and polycrystalline CIGS films. These observations together with our results strongly indicate that the grain boundary diffusion is not substantially faster than the diffusion inside the grains, and that this is valid also for layers grown under Cu-rich conditions. 4.1. Vacancy diffusion

Fig. 3. TEM picture of a bilayer, CuGaSe2 on top of CuInSe2, grown in Cu-rich conditions without sodium. The white circles indicate the position of the EDS measurements. In the graph to the right the Ga/(Ga þ In) ratio is presented for both the horizontal and vertical series of measurement points.

Diffusion in a crystal can proceed through three different basic mechanisms: (1) interchange by rotation (2) migration through interstitial sites or (3) diffusion via vacant lattice sites [14]. In Marudachalam et al. [7] CIGS films were fabricated by selenisation, and as mentioned before, in this process a higher Ga/(Ga þ In) ratio towards the back contact is spontaneously obtained. Heating of these films in argon atmosphere resulted in a homogenization of

O. Lundberg et al. / Journal of Physics and Chemistry of Solids 64 (2003) 1499–1504

the Ga/(Ga þ In) ratio. Heating the films in a Se atmosphere, however, resulted in no such homogenization. Marudachalam et al. [7] explain these observations with that in the layers heated in argon atmosphere, Se vacancies ðVSe Þ are formed and in order to maintain local charge neutrality this also induces the formation of metal vacancies, which facilitates gallium and indium diffusion. In Schroeder et al. [6] the gallium diffusion in single crystal CIS as a function of the Cu/In ratio was investigated and a minimum diffusion close to the stoichiometric composition was observed, where also a minimal amount of metal vacancies are expected to be present. The observations both from Marudachalam et al. [7] and from Schroeder et al. [6] suggest that gallium and indium diffusion proceed according to mechanism (3)—diffusion via vacant lattice sites. In the following we will discuss how our observations can be explained from the assumption that the diffusion mechanism is via vacant lattice sites. Our most consistent and significant observation is that the diffusion is higher in layers without sodium, and especially in those grown under Cu-rich conditions. Also in Schroeder et al. [6], a significantly higher gallium diffusion was observed in the single crystal sodium free CIS layers grown under Cu-rich conditions compared to the layers grown in Cu-poor conditions. The higher concentration of the, for indium and gallium diffusion more favorable VGa and VIn ; compared to VCu ; in the Cu-rich samples was suggested to be the explanation for this observation. This could also explain our observations of a substantially higher diffusion in the Curich, sodium free polycrystalline CIS and CGS layers, compared to the Cu-poor sodium free layers. By growing the layers with access of sodium, the diffusion, in both the Curich and Cu-poor CIS and CGS layers, is reduced to almost the same level. We know from many studies that sodium improves the quality of the CIGS material in several different ways. An improved device performance, microstructure, (112) orientation and many other effects have been observed [12,15]. It is therefore conceivable that also the concentration of metal vacancies is reduced with the presence of sodium. In Wei et al. [15] it was suggested, from theoretical calculations, that the presence of sodium inhibits the formation of the defect pair ð2VCu þ InCu Þ: An additional peak in the XRD data for sodium free layers, corresponding to a superstructure reflection, indirectly indicating a higher concentration of the defect pair ð2VCu þ InCu Þ; was observed by Herberholtz et al. [16] in agreement with these calculations. Thus, it is likely that CIGS films grown with access of sodium have lower concentration of vacancies, facilitating gallium and indium diffusion, compared to CIGS films without sodium. That would explain our observations of less diffusion in the sodium containing layers. 4.2. CuxSe influenced diffusion In Walter and Schock [5] it is concluded that in depth variation of the Ga/(Ga þ In) ratio is homogenized if

1503

the CIGS layer reaches a Cu-rich composition during growth. These layers were fabricated on borosilicate glass (Corning 7059), which has a very low sodium content. Their observed trend of a significantly higher intermixing in Curich layers is thus in agreement with our observations for the sodium free layers. However, in Walter and Schock [5] multilayers of CIS and CGS were fabricated and the intermixing process between these layers is not only ‘pure’ diffusion, as the intermixing may occur in a growing layer. These results can therefore not be generalized to diffusion in sodium containing layers. In Walter and Schock [5], the presence of CuxSe, which at 500 8C is believed to be in a quasi-liquid state with a high solubility of indium and gallium, was suggested to be the explanation of the high diffusion. In our layers, we observed the largest difference in diffusion between layers with the same Cu/(In þ Ga) ratio, but with and without sodium. Assuming that the sodium is not affecting the amount of CuxSe or how it is precipitated in the CIGS structure, this result suggests that the presence of CuxSe is not the main determining factor for diffusion of Ga and In. In our Cu-rich CGS layer with sodium, we observe a slight increase of the diffusion into the upper part of the layer, where we also observed an increased grain size. The sodium free CGS layer grown under Cu-rich conditions also had an increased grain size as compared to the single CGS layer grown under the same conditions. This observation of some kind of re-crystallization or grain-growth could be related to the presence of CuxSe. How re-crystallization influences the diffusion is not clear, but it is reasonable to believe that such a type of reorganization would have an impact also on the distribution of In and Ga [13].

5. Summary and conclusions The diffusion of gallium and indium in polycrystalline co-evaporated CIGS thin films has been investigated by fabrication and analyzing eight bilayers, grown under different conditions. Our conclusions are: † The diffusion takes places inside the grains by vacancy diffusion, i.e. the diffusing atoms are moving via vacant lattice sites in the crystal. The diffusion in the grain boundaries is not significantly higher than inside the grains, not even if the layer are Cu-rich. In has a similar diffusivity in CGS as Ga in CIS. † The diffusion is higher in sodium-free films than in films containing sodium, possibly due to increased concentration of metal vacancies. Sodium free Cu-rich layers show higher diffusion as compared to sodium-free Cupoor layers which can be explained by higher concentrations of the for indium and gallium diffusion more favorable VGa and VIn in the Cu-rich films compared to VCu in the Cu-poor films.

1504

O. Lundberg et al. / Journal of Physics and Chemistry of Solids 64 (2003) 1499–1504

† There is only a slightly increased diffusion in Cu-rich layers with sodium compared to Cu-poor layers. This increase could be related to the presence of CuxSe. However, the effect is small, at least for Cu/(Ga þ In) ratios up to 1.1, which means that it is possible to have an in depth variation of the band gap also in Cu-rich CIGS layers provided they contain sodium.

Acknowledgements This work has financially been supported by the Foundation for Strategic Environmental Research (MISTRA), the Swedish Energy Agency and the European research program Production of Large Area CIS based Modules (PROCIS).

References [1] M.A. Contreras, B. Egaas, K. Ramanathan, J. Hiltner, A. Swartzlander, F. Hasoon, R. Noufi, Towards 20% efficiency in Cu(In,Ga)Se2 polycrystalline solar cells, Prog. Photovoltaic Res. Appl. 7 (1999) 311– 316. [2] B.M. Basol, V.K. Kapur, A.H. Halani, C.R. Leidholm, J. Sharp, J.R. Sites, A. Swarzlander, R. Matson, H. Ulla, Cu(In,Ga)Se2 thin films and solar cells prepared by selenization of metallic precursors, J. Vac. Sci. Technol. A 14 (1996) 2251–2256. [3] M.A. Contreras, J.R. Tuttle, A. Gabor, A. Tennant, K. Ramanathan, S. Asher, A. Franz, J. Keane, L. Wang, R. Noufi, High efficiency graded bandgap thin-film polycrystalline Cu(In,Ga)Se2—based solar cells, Solar Energy Mater. Sol. Cells 41/42 (1996) 231 –246. [4] M. Bodega˚rdrd, O. Lundberg, J. Malmstro¨m, L. Stolt, Proceedings of the 28th IEEE Photovoltaic Specialist Conference, Anchorage, 2000, p. 430 [5] T. Walter, H.W. Schock, Crystal growth and diffusion in Cu(In,Ga)Se2 chalcopyrite thin films, Thin Solid Films 224 (1993) 74–81.

[6] D.J. Schroeder, G.D. Berry, A. Rockett, Gallium diffusion and diffusivity in CuInSe2 epitaxial layers, Appl. Phys. Lett. 69 (1996) 4068. [7] M. Marudachalam, R.W. Birkmire, H. Hichri, Phases, morphology, and diffusion in CuInxGa12xSe2 thin films, J. Appl. Phys. 82 (1997) 2869–2905. [8] J.R. Tuttle, D.S. Albin, A. Tennant, A.M. Gabor, M. Contreras, R. Noufi, Characterization of variable band-gap thin-film Cu(In,Ga)Se2: a simple model for the interdiffusion of In and Ga in alloy structures, Photovoltaic Science and Engineering Conference 7, Nagoya, Japan, 1993 [9] M. Marudachalam, H. Hichri, R.W. Birkmire, J.M. Schultz, A.B. Swartzlander, M.M. Al-Jassim, Diffusion of In and Ga in selenized Cu –In and Cu –Ga precursors, IEEE 25th Photovoltaic Specialist Conference, Washington, DC, 1996, p. 805 [10] A. Gabor, J.R. Tuttle, M.H. Bode, M.A. Franz, A.L. Tennant, M.A. Contreras, R. Noufi, D.G. Jensen, A.M. Hermann, Bandgap engineering in Cu(In,Ga)Se2 thin films grown from (In, Ga)2Se3 precursors, Solar Energy Mater. Sol. Cells 41/42 (1996) 247 –260. [11] O. Lundberg, M. Edoff, J. Malmstro¨m, L. Stolt, Influence of the Cu(In,Ga)Se2 thickness and Ga-grading on the solar cell performance. Prog. Photovoltaics 11 (2003) 77–88. [12] M. Bodega˚rd, L. Stolt, J. Hedstro¨m, Proceedings of the 12th European Photovoltaic Solar Energy Conference, 1994, p. 1743 [13] M. Bodega˚rd, O. Lundberg, J. Lu, L. Stolt, Re-crystallisation and interdiffusion in CGS/CIS bilayers, Proceedings EMRS Conference, Strasbourg, 2002 [14] A. Kittel, Introduction to Solid State Physics, Wiley, New York, 1986, pp. 18, Chapter 18. [15] S.H. Wei, S.B. Zhang, A. Zunger, Effects of Na on the electrical and structural properties of CuInSe2, J. Appl. Phys. 85 (1999) 7214–7218. [16] R. Herberholtz, H.W. Schock, U. Rau, J.H. Werner, T. Haalboom, T. Go¨decke, F. Ernst, C. Beilharz, K.W. Benz, D. Cahen, New aspects of phase segregation and junction formation in CuInSe2, 26th Photovoltaic Specialist Conference, Anaheim, 1997, p. 323