Cu(In,Ga)Se2 (CIGS)-based solar cells have the highest potential of all thin-film solar cells with record efficiencies exceeding 20%.1 CIGS layers are conventionally deposited in high-vacuum chambers using either thermal co-evaporation of the elements, including selenium,2 or sequential sputtering of Cu, In, and Ga with subsequent selenization.3, 4 Using sequential sputtering, the highest efficiencies of more than 16% could be demonstrated by Johanna Solar for a rather complex and costly pentanary system Cu(In,Ga)(Se,S)2 from sputtered bilayer structures of CuGa and In, treated first in H2Se and subsequently in H2S.5 A similar process was originally developed by Siemens and improved by Showa Shell and is called the sulfurization after selenization method.6
Although these processes are very promising, a further reduction of process complexity and production costs is desirable. Up to now, most sputter-based fabrication methods work with more than one target and thus require the control of several cathodes which makes the process quite complex. A single-target approach avoids this challenge. Recent publications deal with the development of such a Cu-In-Ga (CIG) alloy target,7, 8 whereas first results could be demonstrated using alloy targets leading to cell efficiencies exceeding 6%.9, 10 We demonstrate a simple fabrication method for CIGS solar cells employing such an alloy target. The cell efficiencies well over 13% indicate a high potential for cost reduction. The selenization is made with a nonvacuum setup to avoid expensive vacuum equipment and sophisticated selenization reactors. Furthermore, it employs elemental selenium instead of highly toxic H2Se.
CIG precursor layers are sputter-deposited on Mo-covered glass substrates from one single target without any grading within the layer. The target is made from microalloyed powder from Heraeus with 48.5 at.% copper, 38.5 at.% indium, and 13 at.% gallium.8 Standard parameters of 0.5 W/cm2 dc power, an Ar flow of 30 sccm, and no additional heating of the substrate were applied. Deposition time and process gas pressure was varied from 100 to 2000 s and 1 to 15 μbar, respectively.
The samples were selenized in a simple tubular furnace in nitrogen-diluted selenium vapor slightly below atmospheric pressure (for safety reasons), similar to the setups described in Refs. 11 and 12. A continuous flux of nitrogen carries the selenium vapor along the sample reaction zone. The intrinsic temperature gradient within the furnace is used to establish a sample temperature of about 550°C while the crucible with the elemental selenium is placed at the edge region of the furnace at about 400°C to 500°C. The process time was varied between 2 and 90 min.
Absorber layers are completed using conventional methods by a CdS buffer layer, a ZnO window layer, and a ZnO:Al layer to obtain a complete solar cell, as described in Ref. 13.
The film morphology of precursor layers as well as of selenized layers was studied with a scanning electron microscope (SEM) (XL30 SFEG Sirion from FEI Company); energy-dispersive x-ray spectroscopy (EDX) measurements were carried out with a Noran detector in the same instrument. The precursor surface was measured by a three-dimensional (3D) laser scanning microscope VK9700 from Keyence. Secondary neutral mass spectrometry (SNMS) measurements were made with a SSM 200 from Leybold. The crystallinity is monitored by x-ray diffraction (XRD), performed with a Siemens D5000 (copper target) at 40 kV and 40 mA tube voltage and current, respectively. X-ray fluorescence analysis (XRF) is performed at 10−1 mbar in an EAGLE XXL system, equipped with an energy-dispersive Si (Li) detector and a 50 kV Rh x-ray source. Current-voltage curves were measured using a Keithley 238 source-measure unit under simulated AM 1.5 global solar irradiation with a ORIEL sun simulator at 100 mW/cm2 to extract the basic solar cell characteristics from devices of approximately 0.24 cm2 in size.
Results and Discussion
SEM cross-sections and 3D microscope measurements of precursor layer surfaces are shown in Fig. 1. We observe a phase-segregated morphology consisting of a closed, compact underlayer and a rough, irregular top layer. For increasing deposition time, the closed underlayer increases linearly, while it decreases for increasing sputter gas pressure. As expected, the overall film thickness increases linearly with sputter time and decreases slightly with sputter pressure (determined via XRF, data not shown).
Purwins14 described the existence of a two-phase regime in the Cu-In-Ga phase diagram, consisting of elemental indium together with the copper-rich Cu16(In1-y,Gay)9 and/or Cu9(In,Ga)4 phases. Indeed, our XRD measurements of a precursor indicate the existence of elemental indium as well as Cu16(In1-y,Gay)9 and/or Cu9(In,Ga)4 within the layer (see Fig. 2). It was not possible to distinguish between the two phases within the XRD pattern. Furthermore, it is not clear if this phase combination evolves during the sputtering or if it is already present in the target.
The atomic composition determined by EDX of the underlayer and the top layer is shown in Table 1. The atomic ratio CGI = Cu/(Ga+In) of ca. 1.8 for the underlayer fits well to the copper-rich Cu16(In1-y,Gay)9 phase. The top layer contains much more In than the underlayer and we hence expect In to be present as a mixed alloy along with the Cu-In-Ga phase(s).
The atomic composition of the underlayer and the top layer of a sputtered precursor, determined by EDX measurement, as well as the resulting atomic ratio CGI = Cu/(Ga+In).
|EDX||Cu / at.%||In / at.%||Ga / at.%||CGI|
The sputtered film generally contains less copper and less gallium than the original target. The graph in Fig. 3 shows the average atomic concentration of precursors which were sputtered at pressures from 1 to 15 μbar. We observe selective sputtering with a decreasing copper and gallium content as well as an increasing indium content for increasing sputter gas pressure. This phenomenon is probably related to differences in the sputter yield for elements with different atomic masses as is the case in an alloy target containing the light Cu and heavy In. It correlates with the decreasing thickness of the copper-rich underlayer with increasing pressure, as shown in the SEM cross-sections in Fig. 1.
Selenization and Solar Cells
XRD measurements of the CIGS layers are carried out for various selenization durations, as shown in Fig. 2. Already after 7 min selenization we find the CIGS phase beginning to evolve. For short selenization durations Cu(In,Ga)3Se5 and copper indium alloys can be seen.
Critical metallic phases are not detectable anymore in the diffraction pattern after 25 min of selenization, although shunting is still sometimes an issue. The shunting behavior could be related to incomplete CIGS transformation with residual (amorphous) metallic phases. Thus, although CIGS conversion is shown by XRD to be complete after about 25 min, a selenization time of 60 min is favorable in order to obtain maximum cell efficiencies.
Alberts et.al investigated the influence of elemental Se or H2Se, respectively, and found that Se-vapor-treated samples had nonuniform surface morphologies correlated with a relatively large variation in the Cu/In atomic ratio and Se concentration through the depth of the samples.6 As shown in the SNMS depth profile [Fig. 4] our samples are more homogeneous. We find a relatively constant CGI [Cu/(Ga+In) ratio] as well as a constant Se/(Cu+In+Ga) ratio throughout the depth of the sample. We only see an increase of Se content at the molybdenum back contact, which is heightened with selenization time [Fig. 4]. We attribute this feature to the formation of molybdenum selenide. Furthermore, we observe a beneficial gallium gradient in the depth profile of the absorber layer.
A layer selenized for 60 min is shown in Fig. 5, indicating a polycrystalline structure. It has an atomic composition of Cu: 22.1 ± 0.5 at.%, In: 20.0 ± 0.9 at.%, Ga: 5.4±0.3 at.%, and Se: 52.6 ± 0.6 at.%, in agreement with the desired ratio for Cu(In,Ga)Se2.
With these devices we can attain very good current-voltage characteristics and quantum efficiencies. Table 2 shows current-voltage characteristics of different cells that have been selenized in different runs and finished without antireflective coating or metallic grids. The accordant curves for the best cell with a conversion efficiency of 13.7% are shown in Fig. 5.
Current-voltage characteristics of different cells (A = 0.24 cm2), without antireflective coating or metallic grids.
We investigated the film growth and selenization reaction of Cu-In-Ga precursor layers manufactured from a simple and cost-effective single-target sputter process and using a nonvacuum selenium vapor selenization reaction. The film growth during sputtering takes place via Cu- and In-rich segregated phase alloys. We could demonstrate highly efficient CIGS solar cells with cell efficiencies well above 13% with this approach.