© EDP Sciences 2010

1 Introduction

Up to now, tandem and multi junction solar cells are the only concept exceeding the Shockley-Queisser efficiency limit of 30% under solar illumination without concentration [1]. High transparency of the top cell below its energy gap is – apart from efficient absorption above E_g – a crucial requirement for an efficient tandem cell. The chalcopyrites constitute a system of absorber materials with energy gaps suitable for tandem cells, e.g. CuGaSe₂ with E_g = 1.68 eV and Cu(In,Ga)Se₂ with E_g = 1.1 eV. However, the best CuGaSe₂/Cu(In,Ga)Se₂ tandem efficiency published so far is 7.4% [2]. This relatively low value is fundamentally related to the low efficiency of the top cell together with a top cell transparency of only 60% in its sub-gap range.

Previously we had developed an optical model of the n-ZnO/i-ZnO/CdS/CuGaSe₂/SnO₂:F/glass solar cell that allowed for the description of the optical properties of this top cell [3]. Based on this model in [4] an optimized top cell stack had been derived that showed significant improvement in transparency. In this paper we will give the experimental proof of enhanced top cell transmission and resulting gain in bottom cell performance in the tandem. An improved efficiency of the mechanically stacked CuGaSe₂/Cu(In,Ga)Se₂ tandem will be reported which is however still far from surpassing the Cu(In,Ga)Se₂ single cell efficiency. Our discussion will point out options and requirements for building an efficient chalcopyrite tandem.

2 Experimental and results

Chalcopyrite absorbers were prepared by physical vapor deposition in a three stage process [5]. For the bottom cell the approx. 2 μm thick Cu(In,Ga)Se₂ was deposited onto a molybdenum back contact, whereas for the CuGaSe₂ top cell a transparent back contact is required. SnO₂:F with a thickness of approximately 850 nm was used in the initial configuration. The junction of each cell was formed by chemical bath deposition of CdS and sputtering of intrinsic and Al-doped ZnO. The standard configuration of a Cu(In,Ga)Se₂ bottom cell is ZnO:Al(200 nm)/ i-ZnO(125 nm)/CdS(50 nm)/Cu(In,Ga)Se₂(2000 nm)/ Mo(800 nm)/glass substrate.

The starting structure and the layer thicknesses of the CuGaSe₂ top cell are indicated as initial stack (J) in Table 1. The theoretically optimized top cell stack derived in [4] is set beside. It is characterized by 1) reduced layer thicknesses; 2) careful adaptation of layer thicknesses to anti-reflection behavior; 3) reduced reflection by an anti-reflection coating on top and 4) a substrate simulating monolithic integration. Furthermore, the experimentally realized stacks are given: first the absorber was grown with a thickness approaching 1 μm (opt.stackexp. (A) in Table 1), then the thicknesses of the front ZnO layers were reduced (stack (B)). Note; for the thinner transparent back contact, the SnO₂:F was replaced by ZnO:Al, which has comparable optical properties but was in contrast to SnO₂:F available with arbitrary thickness (stack (C)). Hence, stack (C) implements the theoretically derived modifications 1) and 2). Steps 3) and 4) were not implemented because monolithic integration is not yet feasible and the MgF₂ anti-reflection coating lacks long-term stability. In addition, the exact tuning of the layers to the optimal (anti-reflective) thicknesses is difficult to achieve in the experiment.

Figure 1 shows the transmission spectra measured for the various CuGaSe₂ top cell stacks. The lowest curve gives the measurement of the initial stack (J) which reaches a maximum transparency of 60%. A reduction of the absorber layer thickness for approx. 1/3 leads to an increase in top cell transparency of 8% in the wavelength range from 700 to 1200 nm (stack (A)). Further overall enchancement of the top cell transparency (including the long wavelengths) is obtained by reduction of the thicknesses of the front ZnO layers (stack (B)). The optimized structure (C) finally features an average transparency of 80% and stays at a constant level in the wavelength region defined by the energy gaps of the top CuGaSe₂ (λ_g ≈ 700 nm) and the bottom Cu(In,Ga)Se₂ (λ_g ≈ 1200 nm) absorber. The 80% transmission of the optimized stack means an increase of 20% absolute compared to the initial stack. The observed improved transparency fitted well to the optical model established in Diplot [6], see the dashed curve in Figure 1. The parameters were derived from the modeling of the initial stack (compare [3]) but corrected to include reduced layer thicknesses and related small changes in material properties. The increase in transparency in the long-wavelength regime is due to reduced free charge carrier absorption in the front and back transparent conducting oxide of the top cell. Close to the energy gap, additional reduced defect absorption of the thinner CuGaSe₂  absorber contributes to the enhanced transmission.

Fig. 1 Measured transparency of the CuGaSe2 top cell in the initial configuration (J) and of the experimentally optimized structures (A), (B) and (C) (see Table 1); comparison to modeling results for the optimized stack.

Fig. 2 Measured j-V characteristics of a Cu(In,Ga)Se2 bottom cell filtered by the CuGaSe2 top cell stacks as specified in Table 1. The curve of the unfiltered device is given as a reference. Current densities and efficiencies of the bottom cell are indicated.

The increased transparency of the top cell above its E_g is crucial for improving the short circuit current density of the bottom cell. The j-V characteristics of a Cu(In,Ga)Se₂ bottom cell filtered by the various CuGaSe₂ top cell stacks are shown in Figure 2. Corresponding short circuit current densities j_SC and efficiencies η of the bottom cell are indicated. The current density of 38.9 mA/cm² for the unfiltered bottom cell decreased to 10.6 mA/cm² under the initial stack (J). When filtering with the optimized top cell stack (C), however, the remaining current density increases to j_SC = 15.7 mA/cm². This corresponds to 82% of the maximum achievable value which is given by the split of the solar spectrum according to the 1.7 / 1.1 eV energy gap pair. Figure 2 shows how the current density improves in step with the top cell improvements and how it approaches the theoretical limit. The measured values are in agreement with theoretical calculations for the cases (A), (B) and (C) within an error of 5% (not shown here). In the experiment, the efficiency of the Cu(In,Ga)Se₂ bottom cell shaded by the CuGaSe₂ top cell increases from 4.3 to 6.3%.

3 Discussion

Our latest results of a mechanically stacked CuGaSe₂/Cu(In,Ga)Se₂ tandem in the initial configuration are presented in Figure 3. The j-V characteristics are shown for the CuGaSe₂top cell (J), the Cu(In,Ga)Se₂ bottom cell filtered by it and the mechanical stack calculated from the top and bottom cell curves by considering series connection of the two single devices. A tandem efficiency of 8.5% was determined. This value surpasses previously reported efficiencies [2]. It is based on a top cell efficiency of 4.3% and a bottom cell efficiency of 4.3%, for detailed electrical data see Table 2. As these data show, the device is bottom cell limited regarding the photo current. The top cell photo current is still higher in the initial configuartion but will be lowered to better match when using the optimized top cell structures.

Fig. 3 Solar cell characteristics of a CuGaSe2/Cu(In,Ga)Se2 tandem solar cell and the related CuGaSe2 top and filtered Cu(In,Ga)Se2 bottom cell; for the electrical parameters see Table 2.

Under the improved transparency of the experimentally optimized CuGaSe₂ top cell stack (C), the Cu(In,Ga)Se₂ bottom cell efficiency reached 6.3%, see Figure 2. Assuming an unchanged top cell performance of 4.3%, the tandem device might reach over 10% efficiency, which is however still lower than the efficiency of the single bottom cell. From the optical point of view, the improvement of the top cell performance has been successfully performed. The improvement of the electrical properties of the CuGaSe₂top cell presents the major task of tandem optimization in the future. The present record efficiency of a CuGaSe₂ solar cell is 9.7% on molybdenum [7] and 4.3% on transparent back contact [2]. If the electrical performance of the top cell becomes comparable to the one of the bottom cell – thus also reaching 20% as a single junction device – a tandem efficiency of 26% can be expected. This is the value predicted from our theoretical calculations, compare [4].

4 Summary

In conclusion, the optical model of the chalcopyrite tandem derived before [3, 4] has found experimental verification in this paper. The accuracy of the model as well as predicted design improvements have been shown. The data will be useful to guide further research concerning top cells with superiour optical and electrical properties, to quantify any progress made in the experiment, and to extrapolate feasible tandem efficiencies.

Acknowledgments

This work was supported by the EC-Project ATHLET, No. 019670. The authors thank C. Kelch, M. Kirsch, K. Kraft and T. Münchenberg for technical assistance and J. Hüpkes from the FZ Jülich for providing substrates.

References

All Tables

All Figures

	Fig. 1 Measured transparency of the CuGaSe2 top cell in the initial configuration (J) and of the experimentally optimized structures (A), (B) and (C) (see Table 1); comparison to modeling results for the optimized stack.
In the text

	Fig. 2 Measured j-V characteristics of a Cu(In,Ga)Se2 bottom cell filtered by the CuGaSe2 top cell stacks as specified in Table 1. The curve of the unfiltered device is given as a reference. Current densities and efficiencies of the bottom cell are indicated.
In the text

	Fig. 3 Solar cell characteristics of a CuGaSe2/Cu(In,Ga)Se2 tandem solar cell and the related CuGaSe2 top and filtered Cu(In,Ga)Se2 bottom cell; for the electrical parameters see Table 2.
In the text