Cu(In,Ga)Se2 (CIGS) solar cells are sensitive to heat-light soaking (HLS) treatments performed at elevated temperatures. Such treatments often result in improved open-circuit voltage at the expense of short-circuit current and/or fill factor, and may be reversible. In this contribution, we analyze the changes brought in the devices by heat and light treatments to CIGS, to deliver hints regarding the fundamental reasons behind the observed changes and to ensure the long-term stability of the performance improvement. We apply HLS treatments to improve the performance of low-temperature grown CIGS solar cells on flexible polyimide and on glass substrates. We first investigate the treatment conditions favorable for device performances. A gain in open-circuit voltage is obtained, with no clear statistical impact on short-circuit current and fill factor. The changes are stable over periods of several months. Then we combine material analysis techniques to optoelectronic and electrical characterization on samples subject to a unique HLS treatment at the optimal conditions. The voltage gain arises from a large increase in the doping density in the absorber layer, correlated with increased concentrations of alkali elements Na and Rb. This improvement is partly offset by a reduction in minority charge carrier lifetime. The microscopic origin for the observed changes is discussed. At low temperatures, we also measure device enhancement due to the front interface behavior, however with unclear impact on standard operating conditions. The HLS treatment yielded a CIGS solar cell with certified record 21.4% power conversion efficiency on a flexible substrate.
Article
Heat-light soaking treatments for high-performance CIGS solar cells on flexible substrates
https://doi.org/10.21203/rs.3.rs-2116168/v1
This work is licensed under a CC BY 4.0 License
You are reading this latest preprint version
Thin film photovoltaics
Solar cells
CIGS
Cu(In
Ga)Se2
Heat-light soaking
HLS
The highest reported efficiencies for the thin film photovoltaic (PV) technology Cu(In,Ga)Se2 (CIGS) are to date 23.35% on rigid 1, and 20.8% on flexible substrate 2. This technology has proven long-term performance stability, yet, some device properties can be affected by exposition to light and/or heat in a metastable manner. There exist large differences in the reported magnitude of the effects, and the impact of light and/or heat treatment tends to be relatively minor for most good-efficiency devices. It is also worth noting that the effects summarized below appear to be similar for CIGS layers made by co-evaporation 3, 2-step method 4,5 and even by epitaxy on GaAs substrates 6.
CIGS devices react to light illumination or so-called light-soaking (LS), because of modifications of the absorber material properties. An increase in VOC is generally reported in association with an increase in net doping density, while the JSC and long-wavelength spectral response may suffer from the related narrowing of the space charge region 7,8. In line with increased doping, CIGS layers exhibit the so-called persistent photoconductivity 9. These changes are only transients, the effects disappearing after seconds, hours, days or even weeks. Experimental time-resolved photoluminescence (TRPL) lifetimes tend to decrease, which in turn may limit the diffusion lengths and exacerbate the JSC degradation for devices suffering from collection issues 10. The magnitude of the LS effects appears to depend on extrinsic elements such as Na 11–13 and other heavier alkali 3,14. The doping increase tends to be most effective when the device is maintained in open-circuit condition during treatment, and to a lower extent in short-circuit 12,14. Also, changes similar to LS can be triggered by forward voltage biasing 12,15, suggesting that the position of the Fermi level plays an important role.
CIGS devices also react to heat exposure, which is often applied in combination with light in the so-called heat-light soaking treatment (HLS). Effects appear overall similar to LS, with an increase in VOC and net doping density, a reduction in TRPL lifetimes and an influence of Na and heavy alkalis elements 16. This similarity suggests a common or closely related origin for the observable changes to the absorbers. HLS treatments are often applied for a few hours at temperatures of about 80 to 120°C. However, in addition to the effects of the simpler LS treatments, a long-term HLS treatment (400h at 90°) was reported to improve the device FF through the diode parameters 17. Further, the combination of HLS and heat-soaking was reported to be helpful, the later heat-soaking improving the performances of devices limited by minority carrier lifetime and current collection issues 18.
The origin of the LS and HLS effects has been widely discussed. The most generally-accepted explanation relates to the amphoteric VCu-VSe divacancy defect complex, first investigated theoretically by Lany and Zunger 19,20. In addition to a change from shallow donor to shallow acceptor, electron capture modifies the local conformation of the defect. The return to the initial conformation requires overcoming a potential barrier responsible for the metastable behavior, over a timescale predicted to be longer than seconds and dependent on the doping level. Carrier lifetimes may reduce as a consequence of the apparition of the donor defect with energy near the conduction band edge. While the VCu-VSe model is often well received 21, the authors argued that it does not by itself explain all experimental observations 13,22, such as the impact of alkali elements. Other open questions are the reasons for the large variations in reported relaxation timescales, or the impact of heat in HLS treatments as in the proposed mechanisms. Only the relaxation process is significantly impacted by thermal activation and not the excitation. Other origins for LS effects were also proposed in the literature as summarized for example by Eisenbarth et al. 23: DX centers reduce compensation thus increasing net doping, Cu migration, or formation of a p+ surface layer. In addition to changes in the bulk, other layers and interfaces were also shown to play a role in the change of PV parameters upon LS and HLS, for example following light-induced changes in the buffer layer composition or (photo)doping 4,24 or in the Schottky barrier at the back contact 5.
In this contribution, we explore the effects of heat-light soaking treatments on CIGS solar cells, deposited by a low-temperature process on flexible polyimide (PI) and on glass substrates. We first explore the treatment conditions able to induce long-lasting benefits to the PV performances. Then we investigate in detail the effects of the HLS treatments on VOC and other device properties by means of optical, electrical and material analytic methods and speculate on the mechanisms responsible for the observed behaviors. Finally, we present a record CIGS solar cell on flexible substrate.
Effect of HLS
To explore the effects of HLS on our solar cells and identify the best treatment conditions, we subjected CIGS solar cells on PI substrates to subsequent HLS treatments with progressively increased temperatures. The sample containing 15 cells was illuminated for 24h at 80°C in N2 ambient in open-circuit condition, after which JV characteristics were monitored 2 hours after the end of treatment, as well as several times during the following days. Then, the sample was subjected to a new HLS treatment at a more elevated temperature (90°C) and characterized similarly as previously. The procedure was repeated until reaching the final 120°C temperature. Figure 1 shows the evolution of VOC with time. Starting from about 725 mV, the VOC increases up to 740 mV shortly (2 hours) after the first annealing, before stabilization slightly above the initial value (730 mV). Subsequent annealings at progressively more elevated temperatures also evidence this dual behavior: an initial VOC increase metastable on the timescale of hours, followed by relaxation at an improved VOC value. The best performance was achieved after annealing at 110°C for 24h. JSC values are unchanged upon annealing. The FF may slightly vary by a few percent, however without a clear trend as different samples did exhibit opposite behaviors (not shown). The overall stability of JSC and FF upon annealing is in contrast with part of existing reports, as JSC degradation especially is often observed 7,8,16. Finally, treatment at 120°C leads to performance degradation due to saturation of VOC gain and a slight FF decrease (see Supplementary Information Figure SI.1).
Further, we conducted comparable experiments to probe various other parameters. A very similar behavior was observed for cells on polymer PI and soda-lime glass (SLG) substrates. The presence or absence of the MgF2 anti-reflective coating appears irrelevant, suggesting a mechanism internal to the device. Repeated identical HLS treatment does not bring further benefits compared to a single treatment. Finally, cells maintained in open circuit conditions during HLS treatment improve comparatively better as cells driven in MPP conditions.
Changes in VOC
In the next step, we investigated in-depth specific samples to reveal the nature of the changes induced during HLS treatment. Solar cells were processed from the same CIGS deposition run performed on both PI and SLG substrates. The cells were characterized before and after a unique 24 h HLS treatment at 110°C, identified previously as the optimum treatment condition. In the following, we show data of the sample on PI substrate whenever possible. For temperature-dependent measurements, we present the data of the SLG sample, due to technical reasons. The cells on PI and SLG substrates show very similar efficiencies and PV parameters, and almost identical behaviors for all investigated techniques.
The VOC evolution with time is shown in Figure 2(a). The VOC was stable for 2 months prior to HLS treatment, and increased by about 15 mV upon HLS, after which it remained stable on a timescale of months. JSC and FF only underwent weak changes upon HLS (see Supplementary Information Figure SI.1). External quantum efficiency (EQE) measurements exclude changes in absorber bandgap as well as in sub-bandgap absorption tail (see SI Figure SI.2(c,f)): the VOC improvement arise from higher external radiative efficiency and quasi Fermi level splitting (QFLS).
The internal QFLS potential energy limits the device VOC. In order to discriminate the contributions of both electrons and holes to the QFLS and device voltage, we performed photoluminescence continuous wave (PL-CW) and TRPL measurements on sample fragments after removal of the front electrode by acetic acid wet etching. Figure 2(b) shows power dependencies of PL-CW under around 1-sun equivalent excitation power. The PL intensity increases slightly less than a factor of 2 upon treatment, as reported in Figure 2(d), with a corresponding QFLS increase by about 20 meV, in good agreement with terminal VOC.
Figure 2(c) shows the TRPL decays at various excitation power levels. The absence of fast initial transient in the few first nanoseconds suggests a weak band bending. We observe two qualitative changes in the decays upon HLS: a strong increase in initial intensity (ΣA), and faster decays (τeff). As reported in Figure 2(d), the time-integrated intensity increases by about 20 meV, in very good agreement with both PL-CW and terminal VOC. Therefore, the data qualifies for further interpretation, based on 2-exponential fits (values in Supplementary Information Table SI.1). The pulse with fluence reported in Figure 2(d) generates about 1e11 – 2e12 cm-2 charges, yielding about 1e14 – 2e15 cm-3 charge density when redistributed within the about 1 µm deep bandgap minimum region (GGI gradient in Supplementary Information Figure SI.3), whereas the doping level is about 1e17 cm-3 (see C-V below). Therefore, TRPL measurements satisfy the low-injection condition, except for the highest power levels in "Before" state (with lower doping) for which we may approach high injection condition briefly after the pulse. In low injection condition, the radiative recombination rate is proportional to doping 25: we conclude that doping density increases by half an order of magnitude upon HLS. Absolute values for the doping density are presented below from C-V profiling. Upon HLS treatment, τeff decreases from about 250 down to about 120 ns. The QFLS increase due to doping density (about +40 meV) is partly counterbalanced by an increased recombination rate (about -20 meV).
The origin of τeff degradation deserves further discussion. The measurement conditions remain far from the radiative limit (external radiative efficiency is < 1%, see Table 1). With reasonable assumptions of deep traps, low injection conditions and not too dissimilar electron and hole lifetimes, Shockley-Read-Hall recombination is also independent of doping level. The deep defect assumption represents a worst-case scenario, because as will be presented below, the only identified signature possibly related to a defect becomes shallower upon HLS, thus becoming less detrimental if anything. Therefore, the decreased τeff value should not be a direct consequence of the increased doping level; but it could be caused by an increase in concentration of a defect also responsible for the doping increase. We cannot rule out a possible degradation in the top interface recombination rate, as plausible values in the range of S=1000 cm/s could limit the TRPL decays to the observed values 25.
The optical diode ideality factor extracted from the PL-CW power dependency is about 1.3, unchanged upon HLS. This similarity is in line with the almost unchanged device FF. The optical ideality factor is however significantly lower than the one extracted from J-V curves or suns-VOC data (which are also unchanged upon HLS at about 1.5-1.6). Literature hints that the optical ideality factor is limited to values above 1 by metastable defects in the absorber bulk 26,27, possibly the amphoteric VSe-VCu defect complex. We note that as the optical diode factor is unchanged upon HLS, the defect possibly responsible for the optical diode factor is unlikely to be also responsible for the decreased TRPL decay times after HLS treatment.
The doping level was estimated from C-V profiling, at room temperature (cell on PI substrate) and as a function of temperature (cell on SLG). Figure 3(a) displays the apparent doping density as a function of apparent depth, with symbols highlighting the 0 V value. The room-temperature behavior is very similar for both cells on SLG and PI substrates. Upon HLS, the apparent doping density increases dramatically by more than one order of magnitude, and the space-charge region width shrinks. However, the corresponding increase in QFLS (>60 meV) is not consistent with the electrical and optical data presented above. Figure 3(b) compares the built-in voltage Vbi, and device VOC (see Mott-Schottky data evaluation in Supplementary Information Figure SI.4). The voltage values agree reasonably well in the "After" state, especially near room temperature, but strongly differ in the "Before" state. We believe that before HLS treatment, the front interface induces a capacitance response that differs from the textbook behavior (e.g. via traps, dipole, etc.), resulting in unreliable apparent doping density (in addition to being hard to interpret due to charge injection and SCR thickness, see 28,29). By contrast, the Vbi value after HLS treatment is reasonable and therefore we accept the apparent doping density value of about 1e17 cm-3. Combined with our earlier TRPL analysis suggesting an increase in doping density by about half an order of magnitude, we conclude that the HLS treatment increases the effective doping density in the CIGS layer from about 3e16 up to 1e17 cm-3.
SIMS depth profiling was performed before and after HLS treatment to link the changes in optoelectrical properties to chemical composition. As visible in Figure 3(c), the apparent concentration inside the absorber of both Na and Rb increases by a factor of 2 to 3 upon HLS. The alkali concentrations at front and rear interfaces are within expectations. The increase in the bulk upon HLS is not related to the substrate as the same behavior is observed for samples on alkali-free PI (Figure 3(c)) and SLG (not shown) substrates. It is possible that alkali elements, initially located inside the Mo back contact acting as a reservoir, are released upon treatment into the CIGS layer. Alkali elements incorporate preferentially at the grain boundaries, with little direct impact on doping. However, optimized PDT conditions typically lead to saturation of the absorber bulk and ion exchange 30. We speculate that the HLS condition favors alkali incorporation along extended crystallographic defects as well as within grain interiors presumably for Na, where the impact on doping is enhanced. An alternative explanation is related to instrumentation artifacts, as modification in local alkali implantation, e.g. bulk versus grain boundaries or substitutional versus interstitial may affect matrix effects and therefore the detected signals. In summary, we believe the increase in doping density evidenced by TRPL and C-V methods originates from an increase in Na and Rb alkali concentrations in the bulk.
Other changes
We observe further changes upon treatment, delivering additional insights into the effect of HLS, although with unclear connection to the room-temperature PV performances. Figure 4(a) shows the temperature-dependent VOC of a cell on SLG substrate, for different illumination intensities. Extrapolation of the data points to T=0 K can be interpreted as the transition energy of the dominant recombination channel 31–33. In our case, the high-temperature data extrapolates to about 1.21 eV, very similarly both before and after treatment. This value roughly corresponds to the CIGS bandgap near the front interface, which is slightly higher than the bandgap minimum extracted from EQE (1.15 eV by derivative method; see Supplementary Information Figure SI.2(c)): near room temperature, we assume the dominant recombination mechanism to be a band-to-band transition. The low-temperature data extrapolate to lower values, especially so before HLS, suggesting a change in the dominant recombination mechanism to being defect-mediated. The energy difference of the intercepts for high and low temperature ranges, denoted EA in Figure 4(a), is interpreted as the defect energy to the nearest band. Its energy decreases from about 175 meV down to 115 meV upon HLS treatment, which in principle is favorable to the device performance.
C-f admittance spectroscopy measurements shown in Figure 4(b) reveal a similar signature. The spectra show a change in capacitance values in line with the C-V profiling discussed earlier, as well as a clear main step at low temperatures. Assuming a barrier/trap without temperature dependency, its characteristic energy decreases from about 219 meV down to 118 meV upon HLS. We note the similarities in values as compared to the activation energies derived from VOC-T data. Both before and after HLS, the step characteristics (activation energy & thermal prefactor) perfectly matches the so-called N1 signature as parametrized in Ref. 34, and therefore originate from the same feature. The interpretation of the N1 defect is notoriously ambiguous. We assume that the same feature is responsible for both the capacitance step and VOC-T behavior, and its energy is decreased upon HLS treatment. If assuming this feature is a defect, the SRH theory predicts that shallower defects have a lesser impact on recombination rate and device performances as compared to deep defects: HLS should improve the TRPL decay times, however, the opposite is observed experimentally. Therefore, and considering the insights of built-in voltage, we postulate that the admittance step and the VOC-T behaviors actually originate from a change in band alignment and a more favorable interface barrier after HLS, with unclear localization within the device.
Last, temperature-dependent dark and illuminated J-V measurements are shown in Figure 4(c), for three different temperatures. The slight VOC improvement upon HLS is visible. Decreasing temperature evidences a blocking behavior affecting dark and illuminated curves in a similar manner, suggesting a barrier for injection current. The behavior is slightly improved after HLS treatment (compare both 214K curves), hinting at an improvement in the band alignment upon HLS.
We now speculate on the possible role of the amphoteric metastable defect complex VCu-VSe, which is often invoked to explain the effects of LS and HLS. As a first hypothesis, we assume that the defect complex is responsible for the long-term increase in VOC and doping level, with relaxation time extending over several months. The electronic state b moving within the bandgap explains the reduced TRPL decays. The nature of the initial transient on the timescale of hours is unclear: Lany and Zunger indeed predicted a dependency of relaxation time on the doping level 19, but its value is predicted to vary only by one order of magnitude assuming the observed 3-fold change in doping level. The predicted value for relaxation time is also much shorter (see below). This first hypothesis also does not explain the higher long-term effectiveness of the treatment performed at higher temperatures. As a second line of thought, we hypothesize that the VCu-VSe defect complex is responsible for the initial VOC transient on the timescale of hours. This is in reasonable agreement with the different treatments shown in Figure 1, as a value of 100 s with 2.5e16 cm-3 net doping density predicted by Lany and Zunger 19. As observed, switching the amphoteric defect has a stronger effect on VOC when the base doping level is low (see Figure 1, experiments at 80°C or 90°C in comparison to 100°C or 110°C), because of the logarithmic dependency of VOC on net doping. The long-term increases in VOC and doping could be ascribed to the increase in Na and Rb alkali concentrations. The origin of the decreased TRPL lifetime upon HLS is less clear, and we tentatively linked it to the same mechanism responsible for the increased doping level. Considering the discussion above, we conclude that the VCu-VSe complex is probably responsible for the VOC transient behavior with a timescale of hours, while the long-term VOC increase arises from an alkali-induced increase in doping level. In the samples investigated, this mechanism involving alkali required exposure to temperatures of 100°C or above to be observable without ambiguity.
Champion device
Thanks to the benefits of the HLS treatment, we demonstrate a champion CIGS cell with 21.4% power conversion efficiency on flexible substrate, independently certified by Fraunhofer ISE. The JV and normalized EQE curves are shown in Figure 5. Table 1 details the PV parameters of the new champion cell together with the previous best 2. The absorber was grown on Mo-coated PI substrate with a composition of 0.40 GGI and 0.99 CGI, and about 3.3 µm thickness. The window layers consist of around 20 nm CdS, 65 nm ZnO, 110 nm ZnO:Al, and 105 nm MgF2 anti-reflective coating. A unique HLS treatment was applied at 110°C for 24 h. As compared to our best previous mark, the nearly 2-fold increase in external radiative efficiency (ERE) calculated according to Ref. 35 corresponds to an improvement of about 15 mV, confirming the VOC gain obtained with HLS treatment.
Table 1: PV characteristics of the champion cell on flexible PI substrate, as independently certified by Fraunhofer ISE. The previous best cell parameters are reported from reference 2. The external radiative efficiency (ERE) is computed according to reference 35.
|
|
VOC [mV] |
JSC [mA cm-2] |
FF [%] |
Efficiency [%] |
Area [cm2] |
ERE [%] |
|
Best cell on PI |
746.7 ± 3.0 |
37.35 ± 0.60 |
76.69 ± 0.35 |
21.38 ± 0.36 |
0.5039 ± 0.0050 |
0.34 |
|
Previous best cell |
734.4 ± 2.5 |
36.74 ± 0.70 |
77.17 ± 0.50 |
20.82 ± 0.42 |
0.5149 ± 0.0032 |
0.20 |
If the doping level of treated layers (about 1e17 cm-3) is close to the generally assumed optimum value for thin film p-n junction cells, further performance enhancement could be obtained upon reduction of the non-radiative recombination. We indeed observe a two-fold decrease in TRPL decay times, whose origin remains unclear. While our champion device is far from the radiative limit, retrieving the carrier lifetime before treatment would lead to a further 20 mV VOC boost. Further performance gains may also be obtained by improving the relatively low 76.7% FF value. A fit to the JV data of the champion cell yields an electrical ideality factor of about 1.6 (1.5 by JSC vs VOC method), significantly larger than the optical PL-CW 1.3 value. This difference in ideality factor translates into about 3% FF loss, which can be described generically as current transport losses, and may be improved by e.g. thinning the absorber and improving the interfaces. By contrast, the optical ideality factor value of 1.3 is limited by absorber quality, presumably defects 26,27 reducing the FF by another 3% as compared to the perfect 1.0 value.
In summary, we investigated and demonstrated improvement in VOC and PV performances of CIGS solar cells upon HLS treatment. The improvement is driven by a VOC boost of about 15 mV, while JSC remains constant and FF slightly varies inconsistently from sample to sample. The improvement in VOC stems from two separate mechanisms: one is metastable on the timescale of hours tentatively ascribed to the VCu-VSe divacancy defect complex, and one is stable over months, believed to be connected to the redistribution of alkali elements within the absorber. Optimal HLS conditions were identified as 110°C for 24h in N2 ambient. The HLS treatment results in a record CIGS cell on flexible substrate with independently certified 21.4% power conversion efficiency.
Investigating the long-term improvement of our devices, the VOC increase originates from an increase in doping density up to about 1e17 cm-3, which increases the QFLS by about +40 meV. The increased doping density is related to a significant increase in Na and Rb alkali concentration in the absorber bulk. This improvement is partly compensated by a decrease in minority carrier lifetime.
Furthermore, we identify signs of improved interfaces and band alignment. After HLS, Mott-Schottky plots and Vbi values become more consistent with textbook analysis than before treatment. Activation energies of VOC-T extrapolation and of the C-f capacitance step decrease similarly, which is tentatively interpreted as a decreased barrier for injection current in the window layers.
Last, we identify pathways for further improvement of PV performances of CIGS on flexible PI substrates. Retrieving the carrier lifetime measured before treatment would lead to a further 20 mV VOC boost. We also discriminate contributions of a bulk defect and of transport losses and quantify their impact on FF. Finally, further research should aim at replicating the impact of the HLS treatment, based on time-effective modifications to the fabrication process in view of the industrial implementation of our findings.
The CIGS solar cells were fabricated on Mo-coated flexible PI substrates, or when specified SLG substrates coated by SiOx alkali diffusion barrier and Mo back contact. The CIGS absorbers were deposited by multi-stage coevaporation including alkali treatments, as detailed in 2 (process "D"). The window layers consist of about 25 nm chemical bath deposited CdS, 65 nm ZnO and 110 nm ZnO:Al by RF sputtering. The cells were finalized with evaporated Ni/Al grids and mechanical scribing, into devices about 0.5 cm2 designated illumination area. Unless specified otherwise, a 105 nm MgF2 antireflective coating (ARC) was deposited on top by e-beam evaporation.
The HLS treatments were performed for 24 h in 0.5 bar N2 atmosphere, at 110°C unless specified otherwise. An illumination of about 1-sun intensity was provided by a white-light LED with emission intensity essentially comprised between 425 and 750 nm, and the cells were kept in open-circuit conditions.
JV data were acquired under standard test conditions (25°C, 1000 W/m2) in an ABA-class sun simulator and a Keithley 2400 source meter in a 4-probe configuration. EQE measurements were acquired using a chopped white halogen lamp spectrally filtered with a LOT MSH-300 monochromator. The samples were placed under 0.2 sun halogen light bias, and a calibrated Si cell was used as a reference.
Absorber composition and thickness were evaluated by a home made X-ray fluorescence (XRF) tool. Compositional depth profiles were measured with a ION-TOF GmbH TOF SIMS5 secondary ion mass spectrometer (SIMS) (0.6 pA, 25 keV Bi+ primary beam with 50x50 μm2 raster size, 250 nA, 2 keV O2+ sputtering beam on 300x300 μm2 area). GGI ([Ga]/([Ga]+[In])) depth profiles were obtained by scaling the integrated elemental traces with the integral GGI values obtained from XRF.
C–V depth profiles were obtained with an Agilent E4980A LCR meter at a frequency of 1 kHz in the temperature range 123-323 K, with samples kept in the dark. C-f admittance spectroscopy (0 V bias) and JSC-VOC data were acquired with the same system, the latter using a halogen lamp with sets of neutral density filters.
Data availability statement
The data displayed in the figures are also provided in a compressed archive file as Supplementary information. The datasets generated and analyzed during the current study are available from the corresponding author on reasonable request.
Acknowledgements
This work received funding from the Swiss Federal Office of Energy (SFOE) under ImproCIS project (Contract no.: SI/501614-01), the Swiss State Secretary for Education, Research and Innovation (SERI) under contract number 17.00105 (EMPIR project HyMet) and the European Union’s Horizon 2020 research and innovation programme under grant agreement No 850937 ("PERCISTAND"). X.S. acknowledges funding from the ETH Zurich Postdoctoral Fellowship.
The Laboratory for Nanoscale Material Science at Empa is gratefully acknowledged for granting access to SIMS measurements facilities.
Competing interests
The Authors declare no Competing Financial or Non-Financial Interests.
Author contributions
R. C.: Conceptualization, Methodology, Data Curation, Writing - Original Draft, Writing - Review & Editing, Visualization, Supervision, Funding acquisition.
S. N.: Conceptualization, Methodology, Investigation, Writing - Review & Editing.
S.-C. Y.: Investigation, Writing - Review & Editing.
M. O.: Methodology, Investigation, Writing - Review & Editing.
X. S.: Investigation, Writing - Review & Editing.
T. F.: Methodology, Writing - Review & Editing.
A. N. T.: Conceptualization, Writing - Review & Editing, Funding acquisition.
- Nakamura M, Yamaguchi K, Kimoto Y, Yasaki Y, Kato T and Sugimoto H 2019 Cd-Free Cu(In,Ga)(Se,S) 2 Thin-Film Solar Cell With Record Efficiency of 23.35% IEEE J. Photovoltaics 9 1863–7
- Carron R, Nishiwaki S, Feurer T, Hertwig R, Avancini E, Löckinger J, Yang S, Buecheler S and Tiwari A N 2019 Advanced Alkali Treatments for High‐Efficiency Cu(In,Ga)Se 2 Solar Cells on Flexible Substrates Adv. Energy Mater. 9 1900408
- Ishizuka S, Shibata H, Nishinaga J, Kamikawa Y and Fons P J 2018 Effects of RbF postdeposition treatment and heat-light soaking on the metastable acceptor activation of CuInSe 2 thin film photovoltaic devices Appl. Phys. Lett. 113 063901
- Chantana J, Kato T, Sugimoto H and Minemoto T 2018 Enhancement of photovoltaic performances of Cu (In,Ga)(S,Se) 2 solar cell through combination of heat-light soaking and light soaking processes Prog. Photovoltaics Res. Appl. 26 868–76
- Chantana J, Kato T, Sugimoto H and Minemoto T 2018 20% Efficient Zn 0.9 Mg 0.1 O:Al/Zn 0.8 Mg 0.2 O/Cu(In,Ga)(S,Se) 2 Solar Cell Prepared by All-Dry Process through a Combination of Heat-Light-Soaking and Light-Soaking Processes ACS Appl. Mater. Interfaces 10 11361–8
- Nishinaga J, Nagai T, Sugaya T, Shibata H and Niki S 2018 Single-crystal Cu(In,Ga)Se 2 solar cells grown on GaAs substrates Appl. Phys. Express 11 082302
- Haug F-J, Zogg H and Tiwari A N 2002 11% efficiency on CIGS superstrate solar cells without sodium precursor Conference Record of the Twenty-Ninth IEEE Photovoltaic Specialists Conference, 2002. (IEEE) pp 728–31
- Heise S J, Gerliz V, Hammer M S, Ohland J, Keller J and Hammer-Riedel I 2017 Light-induced changes in the minority carrier diffusion length of Cu(In,Ga)Se2 absorber material Sol. Energy Mater. Sol. Cells 163 270–6
- Rau U, Schmitt M, Parisi J, Riedl W and Karg F 1998 Persistent photoconductivity in Cu(In,Ga)Se2 heterojunctions and thin films prepared by sequential deposition Appl. Phys. Lett. 73 223–5
- Salas J F L, Heise S J, Richter M, Gerliz V, Hammer M S, Ohland J and Hammer-Riedel I 2017 Simulation of metastable changes in time resolved photoluminescence of Cu(In,Ga)Se2 thin film solar cells upon light soaking treatment Thin Solid Films 633 40–4
- Chen X, Sastre J, Rumpel M, Flegler A, Singhania A, Balta Bonner J, Hoffmann P and Romanyuk Y E 2021 Photonic methods for rapid crystallization of LiMn2O4 cathodes for solid-state thin-film batteries J. Power Sources 495 229424
- Matsuura J, Khatri I, Lin T, Sugiyama M and Nakada T 2019 Impact of heat‐light soaking and heat‐bias soaking on NaF‐treated CIGS thin film solar cells Prog. Photovoltaics Res. Appl. pip.3135
- Czudek A, Urbaniak A, Eslam A, Wuerz R and Igalson M 2020 Dependence of the Magnitude of Persistent Photoconductivity on Sodium Content in Cu(In,Ga)Se 2 Solar Cells and Thin Films IEEE J. Photovoltaics 10 1926–30
- Walkons C, Jahandardoost M, Friedlmeier T M, Hempel W, Paetel S, Nardone M, Ursprung B, Barnard E S, Kweon K E, Lordi V and Bansal S 2021 Behavior of Na and RbF‐Treated CdS/Cu(In,Ga)Se 2 Solar Cells with Stress Testing under Heat, Light, and Junction Bias Phys. status solidi – Rapid Res. Lett. 15 2000530
- Khatri I, Matsuura J, Sugiyama M and Nakada T 2019 Effect of heat-bias soaking on cesium fluoride-treated CIGS thin film solar cells Prog. Photovoltaics Res. Appl. 27 22–9
- Tangara H, He Y, Islam M M, Ishizuka S and Sakurai T 2022 Analysis of the combined effect of long-term heat light soaking and KF/NaF post-deposition treatment on the open-circuit voltage loss in CIGS solar cells Jpn. J. Appl. Phys. 61 SC1050
- Nishinaga J, Koida T, Ishizuka S, Kamikawa Y, Takahashi H, Iioka M, Higuchi H, Ueno Y, Shibata H and Niki S 2017 Effects of long-term heat-light soaking on Cu(In,Ga)Se 2 solar cells with KF postdeposition treatment Appl. Phys. Express 10 092301
- Matsuura J, Shudo K, Khatri I, Sugiyama M and Nakada T 2018 Tunable Carrier Concentration of NaF-Treated CIGS Solar Cells Using Heat-Light Soaking and Subsequent Heating Phys. status solidi - Rapid Res. Lett. 12 1800053
- Lany S and Zunger A 2006 Light- and bias-induced metastabilities in Cu(In,Ga)Se[sub 2] based solar cells caused by the (V[sub Se]-V[sub Cu]) vacancy complex J. Appl. Phys. 100 113725
- Lany S and Zunger A 2005 Anion vacancies as a source of persistent photoconductivity in II-VI and chalcopyrite semiconductors Phys. Rev. B 72 035215
- Maciaszek M and Zabierowski P 2016 Influence of relaxation processes on the evaluation of the metastable defect density in Cu(In,Ga)Se 2 J. Appl. Phys. 119 215103
- Eisenbarth T, Unold T, Caballero R, Kaufmann C A and Schock H W 2010 Interpretation of admittance, capacitance-voltage, and current-voltage signatures in Cu(In,Ga)Se2 thin film solar cells J. Appl. Phys. 107 034509
- Eisenbarth T, Caballero R, Nichterwitz M, Kaufmann C A, Schock H-W and Unold T 2011 Characterization of metastabilities in Cu(In,Ga)Se 2 thin-film solar cells by capacitance and current-voltage spectroscopy J. Appl. Phys. 110 094506
- Kobayashi T, Kumazawa T, Jehl Li Kao Z and Nakada T 2014 Post-treatment effects on ZnS(O,OH)/Cu(In,Ga)Se2 solar cells deposited using thioacetamide-ammonia based solution Sol. Energy Mater. Sol. Cells 123 197–202
- Kirchartz T, Márquez J A, Stolterfoht M and Unold T 2020 Photoluminescence‐Based Characterization of Halide Perovskites for Photovoltaics Adv. Energy Mater. 10 1904134
- Weiss T P, Ehre F, Serrano-Escalante V, Wang T and Siebentritt S 2021 Understanding Performance Limitations of Cu(In,Ga)Se2 Solar Cells due to Metastable Defects – A Route towards Higher Efficiencies Sol. RRL 2100063
- Siebentritt S, Weiss T P, Sood M, Wolter M H, Lomuscio A and Ramirez O 2021 How photoluminescence can predict the efficiency of solar cells J. Phys. Mater. 4 042010
- Ravishankar S, Unold T and Kirchartz T 2021 Comment on “Resolving spatial and energetic distributions of trap states in metal halide perovskite solar cells” Science (80-. ). 371 7672–80
- Ravishankar S, Liu Z, Rau U and Kirchartz T 2022 Multilayer Capacitances: How Selective Contacts Affect Capacitance Measurements of Perovskite Solar Cells Phys. Rev. Appl. 10 13003
- Chirilǎ A, Reinhard P, Pianezzi F, Bloesch P, Uhl A R, Fella C, Kranz L, Keller D, Gretener C, Hagendorfer H, Jaeger D, Erni R, Nishiwaki S, Buecheler S and Tiwari A N 2013 Potassium-induced surface modification of Cu(In,Ga)Se2 thin films for high-efficiency solar cells Nat. Mater. 12 1107–11
- Ott T, Schönberger F, Walter T, Hariskos D, Kiowski O, Salomon O and Schäffler R 2015 Verification of phototransistor model for Cu(In,Ga)Se2solar cells Thin Solid Films vol 582 (Elsevier) pp 392–6
- Zahedi-Azad S, Maiberg M and Scheer R 2020 Effect of Na-PDT and KF-PDT on the photovoltaic performance of wide bandgap Cu (In,Ga)Se2 solar cells Prog. Photovoltaics Res. Appl. 28 1146–57
- Kodalle T, Yetkin H A, Tovar A V, Bertram T, Klenk R, Schlatmann R and Kaufmann C A 2021 A Device Model for Rb-Conditioned Chalcopyrite Solar Cells IEEE J. Photovoltaics 11 232–40
- Werner F, Wolter M H, Siebentritt S, Sozzi G, Di Napoli S, Menozzi R, Jackson P, Witte W, Carron R, Avancini E, Weiss T P and Buecheler S 2018 Alkali treatments of Cu(In,Ga)Se 2 thin-film absorbers and their impact on transport barriers Prog. Photovoltaics Res. Appl. 26 911–23
- Green M A 2012 Radiative efficiency of state-of-the-art photovoltaic cells Prog. Photovoltaics Res. Appl. 20 472–6
(Not answered)
- SIv4.docx
Supplementary information
- FiguresSidata.zip
Data Set 1