Logo: to the web site of Uppsala University

The band gap of Zn(O,S) and (Zn,Mg)O buffer layers are varied with the objective of changing the conduction band alignment at the buffer layer/CuGaSe2 interface. To achieve this, alternative buffer layers are deposited using atomic layer deposition. The optimal compositions for CuGaSe2 solar cells are found to be close to the same for (Zn,Mg)O and the same for Zn(O,S) as in the CuIn0.7Ga0.3Se2 solar cell case. At the optimal compositions the solar cell conversion efficiency for (Zn,Mg)O buffer layers is 6.2% and for Zn(O,S) buffer layers it is 3.9% compared to the CdS reference cells which have 5-8% efficiency.

In this paper, lightsoaking and temperature-dependent current-voltage (JVT) measurements on Cu(In,Ga)Se2 solar cells with atomic layer deposited Zn1-xMgxO buffer layers are presented. A range of Mg concentrations are used, from pure ZnO (x=0) to 26% Mg (x=0·26). Since this kind of solar cells exhibit strong metastable behaviour, lightsoaking is needed prior to the JVT-measurements to enable fitting of these to the one-diode model. The most prominent effect of lightsoaking cells with Mg-rich buffer layers is an increased fill factor, while the effect on cells with pure ZnO buffer is mainly to increase Voc·. The activation energy is extracted from JVT-measurement data by applying three different methods and the ideality factors are fitted to two different models of temperature-dependence. A buffer layer consisting either of ZnO or Zn1-xMgxO with a minor Mg content gives solar cells dominated by interface recombination, which probably can be related to a negative conduction band offset. A relatively high Mg content in the buffer layer (x=0·21) leads to solar cells dominated by recombination in the space charge region. The recombination is interpreted as being tunnelling-enhanced. The situation in between these Mg concentrations is less clear. Before lightsoaking, the sample with x=0·12 has the highest efficiency of 15·3%, while after lightsoaking the sample with x=0·21 holds the best efficiency, 16·1%, exceeding the value for the CdS reference. The Jsc values of the Zn1-xMgxO cells surpass that of the reference due to the larger bandgap of Zn1-xMgxO compared to CdS.

Resistivity and Hall measurements are conducted on atomic layer deposited Zn_1−xMg_xO thin films of different thicknesses and compositions. It is found that the films exhibit persistent photoconductivity after UV-light exposure. The effect is more pronounced for thinner films with higher magnesium content. These are also the films with the highest resistivity. Light-induced excess conductivity is still present in some of the films after weeks of dark storage. Conductivity relaxation is faster at higher temperatures. From Hall measurements, it is observed that conductivity changes are a combined effect of changes in the mobility and concentration of free carriers.

A baseline parameter set for electrical modelling of Cu(In,Ga)Se₂ solar cells with compositionally graded absorber and CdS buffer layer is established. The cases with and without Fermi level pinning as well as withand without a surface defect layer are considered. Simulations with a defect layer are observed to give the best correspondence to measurements. Zn_1−xMg_xO buffer layers are introduced and initial modelling of the lightsoaking behaviour is performed. Simulation results are compared with experimental data.

Surface modifications of 3-stage co-evaporated Cu(In,Ga)Se₂ (CIGS) thin films are investigated by finishing the evaporation with gallium-free (CuInSe₂, CIS) stages of various lengths. We find substantial interdiffusion of indium and gallium, smearing out the Ga/(Ga+In) profile so that the addition of a CIS layer merely lowers the gallium content at the surface. For the thinnest top layer, equivalent to 20 nm of pure CIS, we cannot detect any compositional difference compared to the reference device. The modification are evaluated both by electrical characterization of actual solar-cell devices and by electrical modelling, using the latest version of SCAPS-1D. The best solar-cell device from this series is obtained for the 20 nm top layer, with an efficiency of 16.3 % after anti­reflective coating. However, we observe a trend of decreasing open-circuit voltage for increasing top-layer thicknesses, and we do not find direct evidence that the lowering of the gallium concentration at the CIGS surface should generally be expected to improve the device performance. A simulated device with reduced bulk and interface defect levels achieves 20 % efficiency, but the trends concerning the CIS top layer remain the same.

Electrical modeling of Cu(In,Ga)Se₂ solar cells with Zn_1-xMg_xO buffer layers is performed. A number of  different  device  models  are  implemented  and  tested  by  comparing  simulation  results  and measurement data. Room temperature light-soaking and dark-light cross-over behavior as well aslow-temperature characteristics of these cells are studied. The light-soaking improvements in the solarcell  parameters  are  attributed  to  an  increase  in  buffer  donor  density,  due  to  persistent  photoconductivity, that counteracts charged acceptors in the absorber-buffer region. Dark-light JV-curvecross-over is explained by deep acceptor defects with small electron capture cross-section, in thebuffer. Best correspondence to measurements on ZnO and Zn_0.83Mg_0.17O cells is obtained with models including absorber-buffer interface acceptor states. No wideband-gap surface defect layer is needed to reproduce measurement data.

The purpose of this work is to investigate the interplay between the absorber layer of Cu(In,Ga)Se₂ solar cells and the contacts, including the buffer layer, of these devices. With this in mind Cu(In,Ga)Se₂ devices with gallium-graded absorber layers of different thicknesses and  different  types  of  buffer  layers are fabricated. Absorbers are co-evaporated in-line with the substrate speed determining the layer thickness. Absorber layers and finished devices are characterized. V_oc and FF optima are found for cells with 0.8 µm to 0.9 µm thick absorber layers but the highest efficiencies are found for standard devices with 1.6 µm absorbers due to a high J_sc. Cu(In,Ga)Se₂ cells with Zn(O,S) buffer layers are found to be more efficient than CdS reference devices for the same absorber thickness due to a higher J_sc caused mainly by less light absorption in the alternative buffer layer. For cells with absorber layers thinner than normal, a better QE was also observed at longer wavelengths. Electrical simulations are used to reproduce the behaviour of the devices. It is found that recombination at the back contact limits the thinner devices with CdS buffer layers while the thin Zn(O,S) devices also have a problem with interface recombination. These recombination paths are over-shadowed in the standard thick devices by recombination in the CIGS layer.