Efficiency Improvement of Near‐Stoichiometric CuInSe2 Solar Cells for Application in Tandem Devices

State‐of‐the‐art Cu(In,Ga)Se2 (CIGS) solar cells are grown with considerably substoichiometric Cu concentrations. The resulting defects, as well as potential improvements through increasing the Cu concentration, have been known in the field for many years. However, so far, cells with high Cu concentrations show decreased photovoltaic parameters. In this work, it is shown that RbF postdeposition treatment of CuInSe2 solar cells allows for capturing the benefits from the improved absorber quality with increasing Cu content. A reduced defect density and an increased doping level for cells with high Cu concentrations close to stoichiometry are demonstrated. Implementing a high mobility front transparent conductive oxide (TCO), the improved absorbers with 1.00 eV bandgap yield a solar cell efficiency of 19.2%, and combined with a perovskite top cell a 4‐terminal tandem efficiency of 25.0% are demonstrated, surpassing the record efficiency of both subcell technologies.


Introduction
Structural defects play an important role in semiconductor materials and devices. Point defects, especially, such as vacan cies, impurities, antisites, and interstitials, as well as the defect pairs of those, are known to influence the electronic properties of the materials. The effects can be beneficial, as for the doping of the semiconductor, or detrimental by providing centers of recombination and trapping. [1,2] For highly efficient Cu(In,Ga) Se 2 (CIGS) solar cells, semiempirical optimization of the elemental composition leads to an optimum Cu to groupIII element ratio (CGI) in the range of 0.80-0.90. [3][4][5][6] As a result of this offstoichiometric, Cudeficient composition, a high density of native defects exist within the absorber layer. [7,8] The remark ably low formation energies of some of these defects result in addition to a large gain in V OC due to increased Cu content and RbF PDT, a higher current density is obtained by sub stitution of conventional aluminumdoped zinc oxide (AZO) front contact with highmobility indium-zinc oxide (IZO) to improve the nearinfrared (NIR) response. A maximum solar cell efficiency of 19.2% is achieved. Due to the low bandgap (1.0 eV), such cells are highly suitable for tandem devices in conjunction with perovskite top cells. Such tandems represent one of the most promising technologies to overcome the ther modynamic limitations for single junction solar cells, the so called Shockley Queisser limit. The combination of CIGS and perovskite is especially interesting in this field due to its all thinfilm nature and the tunability of the bandgap of top and bottom cells. This allows us to carefully match the absorption of both partners, which is advantageous for 2 and 4terminal devices. Here we show the efficiency gains when perovskite and CIS cells are used in 4terminal tandem configuration.

Results
All solar cells in this study are processed with single graded absorbers with 1.00 eV front CIS layer while a small amount of Ga is present in a part of the layer close to the interface with Mo. [19] Fabrication details are reported in the experimental details. Figure 1 shows a combined picture of the changes in V OC for samples with different amount of RbF and different concentra tions of Cu in the absorber. The largest effect is seen in V OC , making it the best metric to compare the different experiments. Device efficiencies follow the V OC , as J SC and FF vary to a lesser extent, following the same tendencies (see Figure S1 in the Supporting Information). The Cu contents in this comparison are 0.87, 0.88, and 0.87 for the low CGI samples (no, low, and high RbF) and 0.93, 0.96, and 0.96 in the high CGI case. The respec tive error of measurement is 0.03 absolute. Concentrations above 0.96 lead to decreased V OC (see Figure S2 in the Supporting Infor mation). The notations of low and high RbF refer to source tem peratures of around 500 and 530 °C, respectively, corresponding to an expected doubling of the amount of evaporated RbF.
For samples without RbF, the V OC is reduced when going to high CGI (ΔV OC ≈ 20 mV), as described in the "Intro duction." The external quantum efficiency (EQE) curve of those samples displays a decreased in maximum EQE value ( Figure S3, Supporting Information), consistent with the interface recombination described in literature. [17] Upon RbF treatment, V OC improves for cells with low Cu content, but the gain is considerably larger for cells with increased Cu con tent. For low CGI samples, a gain of about 25 mV is achieved with RbF PDT, similar to that observed with doublegraded absorbers. [27,28] We have previously shown that this improvement is led by a substantial increase in carrier lifetime from about 100 to 400 ns. [26] The highCucontent samples show a remark ably higher improvement of ≈80 mV with RbF treatment, yielding a V OC of 611 mV (best cell value: 606 ± 4 mV on 18 cells). Figure 2a shows the apparent doping density extracted from capacitance-voltage measurements on samples with a high amount of RbF during PDT. The apparent doping increases by about half an order of magnitude. This increase in doping density is expected to reduce the width of the space charge region (SCR). In the investigated solar cells, the col lection of charge carriers generated deep within the absorber appears unaffected by the reduced SCR width; EQE measure ments even show a small increase in the nearinfrared region ( Figure S4, Supporting Information). This is believed to be Adv. Energy Mater. 2019, 9,1901428  caused by increased absorption for high CGI, similar to what has been previously shown. [29] Improved mobility of charge car riers, as reported for stoichiometric absorbers in literature, [14,15] could also contribute to their efficient collection. The Urbach energy has been extracted from exponential decay fits to the EQE below the bandgap. The Urbach energy is reduced from ≈20 to 16 meV upon the increase in Cu con centration, as shown exemplary in Figure 2b for the high RbF cells. The behavior is similar for cells with and without PDT. Above a CGI of ≈0.95 no further reduction in the value of the Urbach energy is obtained ( Figure S5, Supporting Infor mation). The Urbach energy is a measurement for potential fluctuations due to potential nonuniformities and fluctuating charges at defects in the absorber. Hence, our results indi cate a significant reduction in defect con centration with increased Cu concentra tion, independent of RbF treatment. Sim ilar improvements have also been reported for doublegraded absorbers. [30] Narrowbandgap CIS cells are more sensi tive to free carrier absorption in the transparent conductive oxide (TCO) front contact than CIGS, due to the extended NIR response. By replacing AZO with IZO, the absorption losses in the nearinfrared region can be significantly reduced while maintaining comparable sheet resistance in the TCO, as shown in Figure 3b.
At comparable majority carrier concentration (IZO: 3.6 × 10 20 cm −3 ; AZO: 4.0 × 10 20 cm −3 ), the increased mobility (IZO: 47.3 cm 2 V −1 s −1 ; AZO: 14.7 cm 2 V −1 s −1 ) enables a reduction of the TCO thickness and freecarrier absorption while maintaining the same sheet resistance. The gains in the NIR response are partially compensated by losses below 400 nm due to the reduced bandgap of the IZO, resulting in a total increase of 0.5 mA cm −2 in current density. Using IZO as front contact, a solar cell efficiency of 19.2% is obtained, with photovoltaic parameters given in Table 1.
As their spectral response extends in the NIR wavelength up to 1300 nm, we investigated those cells in 4terminal tandem Adv. Energy Mater. 2019, 9, 1901428 a b c d   (Figure 3c). Reproducing this illumination with neutral density and longpass filters, a bottomcell efficiency of 8.0% (Figure 3d and Table 1) is measured, and a tandem effi ciency of 24.1% is achieved in 4terminal configuration. This is well above the efficiency of both subcells (+4.9% vs CIS, +8.0% vs perovskite), and well in the range of the current record efficien cies of both subcell technologies. [32,33] Recently, after initial manuscript submission, an actual 4ter minal tandem device using these bottom cells was measured.

Conclusions
In this work, we show that heavy alkali (RbF) PDT is effective to overcome the recombination issues in CISbased solar cells with Cu concentration close to stoichiometry. The quality of the absorber is improved with high CGI compositions, as assessed by an increase in apparent doping concentration as well as a reduction of defect density evidenced by decreases in Urbach energy. Using the improved processes overcoming some of the key limitations, we demonstrate a solar cell with a V OC of 609 mV and a power conversion efficiency of 19.2% for an EQE bandgap of 1.00 eV. The increases in V OC and FF are attributed to a decrease of recombination in the bulk as well as at the interface. We propose that the presence of an alkali-indiumselenide phase between buffer and absorber layer, suggested by multiple laboratories, [28,40,41] possibly reduces the front inter face recombination.
Combination of a 19.2% CIS cell with a 16.9% semitrans parent perovskite top cell supplied by Solliance enables a 4terminal tandem device with an efficiency of 25.0%.

Experimental Section
Solar Cell Fabrication: CIGS absorbers were grown in a multistage evaporation process on soda lime glass with a sputtered Mo back electrical contact. The highest substrate temperature during growth was 500 °C, and the deposition was finished with an in situ postdeposition treatment with NaF followed by RbF. A detailed description of the growth conditions can be found in ref. [26].
After the absorber growth, all samples were etched in 5% KCN solution for 1 min prior to deposition of about 35 nm CdS as a buffer layer using a chemical bath. Approximately 70 nm of nonintentionally doped zinc oxide was added by sputtering.
As a transparent electrical contact, either AZO or IZO was applied. AZO was deposited by radio-frequency magnetron sputtering at room temperature, using a ZnO target containing 2 wt% Al 2 O 3 . Power density during sputtering was 2.5 W cm −2 , and the sputtering gas was an argonoxygen mixture with 1% oxygen. For the IZO a sputtering target with a composition of In 2 O 3 :ZnO, 89.3:10.7 wt% was used. Power density during pulsed DC sputter deposition was 0.81 W cm −2 ; substrate temperature was 110 °C; and the oxygen flow was 2.33% of the total gas flow. Both TCOs were deposited in the thickness necessary to reach a sheet resistance of 60 Ω sq −1 , which corresponds to a thickness of ≈200 nm for AZO and ≈110 nm for IZO. A typical cross section of the finished device is shown in Figure 3a. The CIS layer shows smaller grains in the graded part of the absorber, as previously reported. [25] All solar cells were finished with Ni/Al grids and a MgF 2 antireflection coating. Cells of ≈0.57 cm 2 were separated by mechanical scribing, and the exact area of each cell was determined by optical scanning at 4800 dpi.
All cells underwent 24 h heat light soaking at ≈1 sun illumination at 80 °C under nitrogen atmosphere as described before. [26] Absorber and Device Characterization: Absorber composition was analyzed by X-ray fluorescence (XRF) after the growth. Because of re-absorption of the secondary X-ray emission, it was important to compare only absorbers with similar composition grading. Inductively coupled plasma optical emission spectrometry (ICP-OES) was used to calibrate the emission for multiple samples within this series. An experimental error (2σ) of ± 0.03 on the Cu/III ratio remained for the XRF, therefore key concentrations by ICP-OES were measured confirming the results within ±0.01 (Table S1, Supporting Information).
Current-voltage (IV) characteristics were measured in a 4-contact mode at standard test conditions (1000 W m −2 , 25 °C, in air) using a Keithley 2400 sourcemeter. A class ABA solar simulator was used to simulate the AM 1.5G spectrum and the light intensity was calibrated with a silicon reference cell certified by Fraunhofer ISE. EQE measurements were performed using a Stanford Research SR830 DSP lock-in amplifier and a monochromator to filter the halogen light source. A halogen light bias with ≈0.2 sun intensity was applied during the measurement. Certified Si and Ge cells were used for calibration.
Adv. Energy Mater. 2019, 9,1901428  To quantify the tandem performance, the perovskite top cell was measured as described above. No pronounced hysteresis was found, and the reported efficiency was taken after MPP tracking for 60 min. To measure the bottom cell performance, the IV response under filtered (neutral density and long pass (780 nm) filters) illumination was taken. In that case, the illumination intensity was adjusted to reproduce the integrated EQE current obtained by multiplication of the top cell transmission with the bottom cell quantum efficiency.
Admittance was measured on a temperature-controlled vacuum stage in a 4-contact sensing mode. An Agilent E4980A LCR meter was used, measuring at 1 kHz and with a test signal of 30 mV.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.