Stability Testing and Electronic Characterization of Kesterite Solar Cells for Indoor Applications

The market for low-power devices such as sensors, communication technologies, wristwatches, and more is booming. Indoor photovoltaics (IPV), i.e. solar cells optimized for use indoors, can guarantee their autonomous operation. (Figure 1) 

Indoor lighting encompasses a large variety of spectra and illumination intensities causing uncertainty in the definition of the efficiency of indoor solar cells. Recently, Parsons et al.[1] presented the IEC protocol TS 62607-7-2:2023, which defines the necessary characteristics of the light source for standardized measurement of the performance of indoor PV. According to the standard, the ideal bandgap for an IPV absorber to maximize the harvested energy indoors is about 1.8–1.9 eV.  

While performance is essential for assessing the feasibility of a PV technology, stability is necessary for its commercialization. 

This blog post provides a hands-on introduction to help researchers understand indoor PVs better. Moreover, we present as an example our stability testing and advanced electrical characterization of kesterite solar cells with the all-in-one measurement tool Paios.  

During ageing we observed a sharp decrease in fill factor (FF) which did not recover with rest in the dark, unlike the open-circuit voltage (Voc) and the short-circuit current (Jsc). 

We demonstrate here that it is possible to attribute this to an increase in doping as well as extraction barriers. This analysis was carried out with the drift-diffusion code of the simulation software Setfos.  

Figure 1: (left) Number of annual publications on indoor PV by area of focus. (right) Number of annual publications on indoor PV by class of PV technology. [2] 

Long-term stability testing of kesterite with low light conditions 

With light intensities close to 1-sun or above, ensuring low series resistance in a solar cell device is essential to avoid severe losses in performance. At low light intensities (e.g. lower than 0.1 sun, typical of indoor PV) the photocurrent generated during operation is significantly smaller than that generated under traditional outdoor applications. Leakage current due to defective regions in the device has then a comparably higher impact on power conversion efficiency than series resistance. For this reason, the shunt resistance gains more importance in the field of IPV. 

Unlike perovskite, kesterite solar cells demonstrated to be good candidates for indoor PV thanks to the high shunt resistance even at low light intensities. More specifically, Se-rich compositions have a better performance compared to S-rich compositions, despite the lower bandgap. This is because S-rich kesterite suffers from deep-level defects impacting the maximum Voc at low-light intensities.[3] 

In this study we used a non-encapsulated Li-alloyed Se-rich kesterite (Li:Cu2ZnSn(S,Se)4) with a bandgap of 1.1eV which already proved to be a favorable composition for high-efficiency kesterite PV.[4] mentioned above.  

Our all-in-one characterization instrument Paios is an ideal tool for studying indoor PVs, thanks to the integrated white LED with a spectrum that is close to the one indicated by Parsons et al.  

We used an illumination intensity of about 10000 lux and a temperature of 85°C in air. While typical operation of IPVs would involve close to room temperature conditions, the high temperature chosen here is relevant to situations where devices in proximity to heat sources need to be powered. More in general, the use of high temperature is expected to accelerate the ageing process. The selected light intensity is also a reasonable upper limit to most indoor illumination conditions. 

The stability test was performed under MPP tracking conditions combined with JV characterizations to follow the evolution of the PV parameters over time.  

Aging damages the fill factor 

During aging the MPP had an initial increase reaching a plateau after about 200 hours of aging. The aging was stopped at 425 hours when the MPP reached about T98, defined here as the time when 98% of the maximum MPP is reached during the aging. (Figure 4) 

Figure 4: normalized power output from MPP tracking measurement at 85°C, in air, and illumination with a white LED.  

Figure 5 shows the evolution of the PV parameters as obtained by the JV measurements performed before, during, and after aging. After aging the device was left in the dark and characterized regularly for 4 days to record if the lost performance was recovered.  

At the beginning of the ageing the Voc starts with a voltage that is about 120 mV lower than the fresh device while the Jsc shows a small increase. At the same time, the FF drops by about 10 points. All these changes are attributed to the increase in temperature (from 25 to 85°C) rather than to device aging. 

During aging, the Voc increases and tends to saturate at long time scales, similarly to the MPP tracking curve, with a total increase of about 50 mV. The FF decreases by another 8 percentage points, reaching about 52%, while Jsc loses 0.07mA/cm2. 

As soon as aging stops and the temperature stabilizes to 25°C, the Voc and Jsc show complete recovery. After 4 days, they exceed the values of the fresh device. On the contrary, the FF drops down to 43% as the aging stops. It recovers to about 47% after 4 days in the dark. 

Overall, between fresh and recovered device: 

• Voc gains 30mV.  

• Jsc variation is negligible with a value on the order of Jsc = 3.75mA/cm2.  

• FF loses about 20%. 

Figure 5: evolution of PV parameters (Voc, FF, and Jsc) of the kesterite solar cell. PV parameters were recorded for a fresh (non-aged) device, during the ageing, and during rest time (room temperature in the dark). 

Advanced Characterization

To identify the main contributions to the drop in FF, the analysis of the JV curves is not sufficient. Indeed, there are multiple degradation mechanisms that could cause similar variations to the JV curves. 

We performed dark impedance (IS), capacitance-voltage (CV), open-circuit voltage decay (OCVD), and transient photocurrent (TPC) measurements (Figure 6) to isolate the factors impacting device performance with aging. 

The characterization was performed on the fresh device, at the end of the MPP tracking and 4 days after storing in the dark. All measurements were performed at 25°C.  

First observation is that the device recovery reduces in magnitude the difference between the aged and the fresh device, but it maintains the type of change caused by the aging. Such recovery has been observed and discussed in previous reports.[5] The comparison will focus on the difference between the fresh and the aged device. 

Table 1. List of observed changes in each technique and list of (some of) the possible causes. 

Table 1 provides an overview of the variation observed in each measurement between the fresh and the aged device and the possible causes that could justify the variation. Each measurement has multiple, potentially concurrent, causes that can explain the variations. There are four main causes that are shared among the measurements:  

• Increase in series resistance: it can originate from the contacts. Impedance spectra highlight an increase from 7 to 14 ohms in series resistance from the fresh to the aged device. This is not sufficient to explain the drop in the fill factor. 

• Increase in doping and shallow traps: it would require a variation in the composition, which would point to changes in stoichiometries or defect formation. This is why it is mentioned as apparent doping. It could also relate to the formation of shallow traps. 

• Low mobility: a decrease in mobility in an inorganic semiconductor is not expected. Although if the degradation process produces traps, transport would be affected.  

All these can be concurring changes in the aged device and it’s difficult to estimate which one is the more impactful. 

For these reasons, we resorted to drift-diffusion simulations to isolate and investigate the impact of each parameter.  

Figure 6: Dark impedance (IS), capacitance-voltage (CV), open-circuit voltage decay (OCVD), and transient photocurrent (TPC) measurements of the kesterite device fresh, aged and after the recovery (rest). 

Simulations 

Figure 7: Drift-diffusion simulation of the kesterite solar cell before ageing. (a) Energy level diagram used in the simulation. (b) Comparison between simulated and experimental data. 

We use the drift-diffusion simulation tool Setfos to investigate the origin of the observed trend and, therefore, highlight possible sources of degradation in our kesterite solar cells. First, we develop a model that can reproduce the quasi-steady-state response of the fresh device.  We obtain a very good fit to the experimental current-voltage (Figure 7b), by using the experimental device parameters and by refining the model starting from input data reported in the literature[6,7] (Figure 7a).  

Figure 8: Investigation of ageing process in the kesterite solar cell. (a) Effect of acceptor doping variations in the active layer. (b) Effect of introducing an electron extraction barrier at the CZTSSe/CdS interface (5×1016 cm-3 acceptor doping in the active layer is considered). 

Next, we explore the ageing effect discussed in the previous section, and we carry out an analysis of how variations in each individual device parameter affect the simulated response. We find that an increase in acceptor doping of the kesterite active layer can explain the increase in open circuit voltage observed during ageing (Figure 8a). However, implementing such an effect fails to reproduce the pronounced drop in fill factor and low forward bias current of the aged device. 

To elucidate the origin of this discrepancy, we explore other parameters that can negatively affect charge extraction and therefore the fill factor of the solar cell. Our investigation shows that introducing an electron extraction barrier at the CZTSSe/CdS interface produces variations in the current voltage characteristics in line with the observed trend. In Figure 8b, we show that introducing a 3 nm thick layer at such interface with an increased electron extraction barrier (0.1 eV larger than the fresh device) results in a very good fit to the experimental data. The formation of such extraction barrier has been previously discussed in the literature.[8] 

Furthermore, we study the simulated device response obtained using other transient and frequency domain techniques (Figure 9). We use the same increase in acceptor doping combined with the presence of an electron extraction barrier evaluated in the fitting of the simulation to the experimental J-V curve under light to describe the aged device. Such changes in solar cell parameters induce variations in the time- and frequency domain results which are qualitatively in line with the experimental observation, increasing confidence in the proposed interpretation. 

Figure 9: Simulation of fresh and aged kesterite solar cell reproducing experimental trends shown in Figure 4: (a) capacitance-frequency, (b) Mott-Schottky analysis of capacitance-voltage data, (c) Open circuit voltage decay data and (d) transient photocurrent (rise and decay data plotted in the same panel). The black arrows emphasize the effect of ageing on each technique, consistently with Figure 4. The aged data refer to input parameters where the doping in the active layer is increased by a factor of 5 with respect to the fresh device. In addition, the electron extraction barrier at the CZTSe/CdS interface is also increased from 0.1 eV to 0.2 eV. 

Conclusions 

We employed advanced experimental and simulation tools to investigate performance and ageing of kesterite solar cells for indoor application. Measurements performed with Paios allow us to probe the effect of continuous exposure to a white LED light source under relevant intensities for indoor applications and at relatively high temperatures (85 °C). The main observations pointed to a drop in fill factor in J-V characteristics as well as a slight increase in Voc. Both effects could be quantitatively reproduced with the help of drift-diffusion simulations performed using Setfos, whereby we demonstrated that an increase in the acceptor doping level in the kesterite and the appearance of an electron extraction barrier would be consistent with the experimental observations. Correspondence between simulations and experiments concerning other transient and frequency domain techniques support the proposed effect. This study demonstrates the importance of (i) combining different experimental techniques to probe power conversion performance as well as charge dynamics in devices and of (ii) exploring possible device degradation mechanisms using drift-diffusion simulations.   

Bibliography 

[1] D. E. Parsons, G. Koutsourakis, J. C. Blakesley, APL Energy 2024, 2, 016110. 

[2] A. Chakraborty, G. Lucarelli, J. Xu, Z. Skafi, S. Castro-Hermosa, A. B. Kaveramma, R. G. Balakrishna, T. M. Brown, Nano Energy 2024, 128, 109932. 

[3] J. Park, H. Yoo, V. Karade, K. S. Gour, E. Choi, M. Kim, X. Hao, S. J. Shin, J. Kim, H. Shim, D. Kim, J. H. Kim, J. Yun, J. H. Kim, J. Mater. Chem. A 2020, 8, 14538. 

[4] A. Cabas-Vidani, S. G. Haass, C. Andres, R. Caballero, R. Figi, C. Schreiner, J. A. Márquez, C. Hages, T. Unold, D. Bleiner, A. N. Tiwari, Y. E. Romanyuk, Adv. Energy Mater. 2018, 8, 1801191. 

[5] S. Campbell, M. Duchamp, B. Ford, M. Jones, L. L. Nguyen, M. C. Naylor, X. Xu, P. Maiello, G. Zoppi, V. Barrioz, N. S. Beattie, Y. Qu, ACS Appl. Energy Mater. 2022, 5, 5404. 

[6] R. K. Ghosh, S. Mahapatra, IEEE J. Electron Devices Soc. 2013, 1, 175. 

[7] A. Jimenez-Arguijo, A. G. Medaille, A. Navarro-Güell, M. Jimenez-Guerra, K. J. Tiwari, M. Placidi, M. S. Mkehlane, E. Iwuoha, A. Perez-Rodriguez, E. Saucedo, S. Giraldo, Z. Jehl Li-Kao, Solar Energy Materials and Solar Cells 2023, 251, 112109. 

[8] S. Campbell, M. Duchamp, B. Ford, M. Jones, L. L. Nguyen, M. C. Naylor, X. Xu, P. Maiello, G. Zoppi, V. Barrioz, N. S. Beattie, Y. Qu, ACS Appl. Energy Mater. 2022, 5, 5404.