Analysis of Perovskite Solar Cell Degradation over Time Using NIR Spectroscopy—A Novel Approach

Abstract

In recent years, there has been a dynamic development of photovoltaic materials based on perovskite structures. Solar cells based on perovskite materials are characterised by a relatively high price/performance ratio. Achieving stability at elevated temperatures has remained one of the greatest challenges in the perovskite solar cell research community. However, significant progress in this field has been made by utilising different compositional engineering routes for the fabrication of perovskite semiconductors such as triple cation-based perovskite structures. In this work, a new approach for the rapid analysis of the changes occurring in time in perovskite structures was developed. We implemented a quick and inexpensive method of estimating the ageing of perovskite structures based on an express diagnosis of light reflection in the near-infrared region. The possibility of using optical reflectance in the NIR range (900–1700 nm) to observe the ageing of perovskite structures over time was investigated, and changes in optical reflectance spectra of original perovskite solar cell structures during one month after PSC production were monitored. The ratio of characteristic pikes in the reflection spectra was determined, and statistical analysis by the two-dimensional correlation spectroscopy (2D-COS) method was performed. This method allowed correctly detecting critical points in thermal ageing over time.

1. Introduction

The development of photovoltaics is still generally costly. This also applies to the production of photovoltaic systems and the installation of such systems for the production of electricity. According to the latest report of the Fraunhofer Institute of Solar Energy Systems ISE [1] dated 24 February 2022, the total PV Market growth during one year, from the end of 2020 to the end of 2021, accounted for approximately 27% (from 144 GW to 183 GW), with the total installed capacity at the end of 2020 exceeding 700 GW, and the total global energy production (PV Power Generation) amounting to 855.7 TWh. Compared to other thin film technologies, organic–inorganic hybrid halide-based perovskites have received widespread attention from researchers due to the rapid improvement in their power conversion efficiencies (PCE), which rose from 3.8% in 2009 to 25.5% in 2021 [2,3,4,5,6]. In this general progress of PV development, simultaneously with the traditional technologies based on monocrystalline and multicrystalline silicon and other typical thin-film technologies such as CIGS and CdTe, other technologies have progressed, based on the use of materials that, according to the NREL terminology [2], are referred to as emerging materials and include perovskite-based photovoltaic materials and structures. The highest efficiency of perovskite solar cells achieved in the last seven years increased from 14% (EPFL) in 2014 to 26.1% (INST) [7] in 2022 and even to 29.6% (NZB) for tandem perovskite solar cells [8]. These values are comparable with those of the most efficient silicon-based cells [2]. The success associated with achieving very high-performance indicators for the best perovskite cells and the low cost of producing cells with average performance indicators comparable to those of commercial silicon or thin-film cells provide the basis for forecasting a significant share of perovskite cells and systems in photovoltaic systems in the near future. Despite being a highly promising class of semiconductors, perovskite materials suffer from various types of mechanical and thermal-based degradation, which has hindered their commercialisation. Elevated temperatures and hygroscopic conditions can lead to the evolution of organic volatile chemical species leading to the formation of lead-based iodides [9,10,11,12]. In addition, perovskite materials display low fractal energy, and perovskite-based solar cell architectures consist of multiple charge transport layers with highly mismatched coefficients of thermal expansion (CTE), making them easily prone to delamination [13,14,15]. However, thanks to new achievements of composition engineering, especially robust encapsulation strategies, perovskite solar cells are capable of achieving over 1000 h of stability under thermal, humidity and light-based stresses [14,16].

In connection with the above, the rapid and effective analysis of changes in the time domain in perovskite structures is essential for the entire technological spectrum of perovskite solar cells.

There are several effective methods for controlling the quality and performance of PV structures at different stages of their production, which are also used for the analysis of perovskite cells. These methods are also used in the analysis of the degradation of structures and links in the time domain. Among these methods, it is worth highlighting photo and electroluminescence methods [17,18,19], X-ray methods, especially X-ray diffraction (GIWAXD) and high-resolution X-ray photoelectron spectroscopy (HR-XPS) [18,19,20], and optical spectroscopic methods VS/NIR, in particular the FTIR method [18,19,21].

These methods, especially when applied in combination, are very effective and allow for a profound explanation not only of the parameters characterising changes in cell structures over time, but also of the physical mechanisms of this process.

It should be emphasised that, currently, the application of these methods requires a quite complex and expensive equipment. This limits the expressiveness and mobility of measurements of specific solar cell structures necessary for long-term cell analysis.

An attempt to propose such a quick and inexpensive method of estimating the ageing of perovskite structures with time was undertaken in this paper, based on the express diagnosis of the spectrum of light reflection in the near infrared (900–1700 nm) from perovskite solar cells, analogically to a method previously used [22,23]. The applied method uses miniature, very mobile equipment, allows us to conduct and record a six-fold measurement of the entire spectrum in a few seconds and is characterised by exceptional mobility and simplicity. Such approach, having an obvious limitation as to the depth of the analysis of the processes involved, opens a new path for the practical analysis of the temporal stability of perovskite solar cells.

2. Samples and the Method of Measurement

Materials: Unless otherwise stated, all the materials were purchased from Sigma-Aldrich Chemie GmbH (Taufkirchen, Germany) and used as received. Silver pellets were obtained from Kurt Lesker Company Ltd. (Dresden, Germany); PET substrates coated with IZO (sheet resistance of 15 Ωcm⁻¹) were obtained from Eastman Chemical B.V. (Rotterdam, The Netherlands), formamidine iodide was obtained from Ajay North America (Powder Springs, GA, USA), and poly-triarylamine (PTAA) from Ossila BV (Leiden, The Netherlands). Methylammonium bromide was synthesised in house using a reported method [24]. The chemical component [6,6]-phenyl-C61 butyric acid n-hexyl ester (PCB6) was obtained from Nano-C.

Planar heterojunction PSCs were fabricated with the following architecture: ITO/PTAA/Cs_0.04(MA_0.17FA_0.83)_0.96Pb(I_0.83Br_0.17)₃/PCB6/Ag. First, a PTAA solution (1.5 mg/mL in toluene) was spin-coated at ambient conditions at 5000 rpm for 30 s, followed by annealing at 100 °C for 10 min (≈20 nm thick). Subsequently, the samples were transferred into a nitrogen-filled glovebox for perovskite layer deposition. For perovskite precursor preparation, stock solutions of PbI₂ and PbBr₂ (1.5 M) in a dimethylformamide (DMF)/dimethyl sulfoxide (DMSO) mixture [4:1 v/v] were prepared. Then, for the FAPbI₃ and MAPbBr₃ solutions, FAI and MABr powders were weighed into separate vials, followed by the addition of PbI₂ (into FAI) and PbBr₂ (into MABr) solutions. Both lead solutions were added in excess to obtain an over-stoichiometric lead content (FAI/MABr:PbI₂/PbBr₂ equals 1.0:1.09). The final perovskite precursor solution was prepared by mixing the solutions of FAPbI₃ and MAPbBr₃ in a 5:1 v/v ratio. Then, 40 μL of a CsI solution (1.5 M solution in DMSO) was added to 1 mL of the mixture. A perovskite layer (≈580 nm thick) was deposited on top of PTAA with a two-step spin-coating procedure, at 650 rpm for 2 s and 4500 rpm for 32 s. Anhydrous diethyl ether (150 μL) was dispensed on the sample in the last 11 s of the spinning sequence. The sample was dried for two minutes, then annealed for ten minutes at 50 °C and then annealed for 60 min at 100 °C for the final crystallisation. For the electron transport material (PCB6), a PCB6 solution (20 mg/mL in chlorobenzene) was spin-coated at 4000 rpm for 30 s and annealed at 60 °C for 10 min. Finally, 5 nm of BCP buffer layer and 95 nm of Ag electrode were deposited on the top of the devices by thermal evaporation at ≈10⁻⁶ mbar, through a shadow mask.

A schematic of the perovskite solar cell stack is shown in Figure 1, and a view of a cross section of a typical solar cell structure was investigated. The chosen parameters of the measured solar cell structures are listed in

Table 1. Table describing the investigated samples.

Sample	Ag Electrode
S1	Yes
S2	Yes
S3	Yes
S4	Yes
S5	No
S6	No
S7	No
S8	No

.

Current–voltage measurements: J–V characterisation and stabilised power output measurements were performed using a Keithley 2461 source measure unit (SMU) under simulated AM1.5G irradiation (100 mW/cm²) using an Sun 3000 AAA-rated solar simulator (Abet Technologies Inc., Milford, CT, USA) calibrated against an RR-208-KG5 silicon reference cell (Abet Technologies Inc.). The mismatch factor for the studied perovskite solar cells was calculated to be 0.968 using external quantum efficiencies (EQEs) of the reference and test cells, lamp’s spectrum and AM1.5G, and this value was used to correct the intensity of the solar simulator lamp to provide one sun illumination [25]. Typically, the PCE efficiency was about 18% under AM1.5.

DLP NIRCAN Measurements: The DLP NIRSCAN NANO device, a compact spectrometer from Texas Instruments, Dallas, TX, USA, operating in the wavelength band from 900 to 1700 nm, was used to record spectra. Its dimensions, 58 mm × 62 mm × 36 mm, and the absence of an external power source make it possible to perform mobile spectra measurements [26]. The device works in reflectance mode. Reflectance (R) is the ratio between the intensity of light reflected from the sample (I) and the intensity of reflected background light or of light reflected from a reference surface (I_r) [22]:

$$R=\frac{I}{I_r}\cdot 100\%$$

(1)

The main advantages of the device are its mobility and capability to perform measurements under various conditions.

The standard software, compatible with a PC running on a Windows operating system, enables registration, acquisition and data visualisation. It is possible to control the device using a mobile application via the Bluetooth standard. The spectrometer can register 228 measurement points in the wavelength band from 900 to 1700 nm within 2.63 s, repeating the measurement six times for every point, then calculating the mean for a given point. The result is presented in a dialog window of the software in the form of a chart or file in DAT or CSV format. The number of scans can be changed, which directly influences the time of measurement.

Figure 2 presents the principle of the device’s operation and how it performs measurements. The sample situated near the spectrometer’s window is illuminated by two light sources in the form of lightbulbs controlled by the device’s controller.

The reflectance spectrum of the lightbulbs from the reference surface is known and saved in the device’s memory. Depending on the tested object, a specific part of the light is absorbed by the sample, a certain amount is scattered, and a part of the light is reflected off the object’s surface into the slit of the device. Further on, the light goes to the optical system in the process, where it is split by a diffraction grating and directed by a lens to the DMD micro-mirror array, which directs the light in a particular sequence to the detector, where it is processed into an analogue signal [23].

3. Results and Discussion

Figure 3 and Figure 4 show the results of the measurement of the reflectance spectrum of PSC structures (perovskite solar cells) made consistently every 24 h under identical conditions using the Express NIR San method from the foil side and the electrode side. The measurements started within about seven days from the production of the structures tested. From end of production to the commencement of the measurements, the samples were stored in hermetic packaging under a nitrogen atmosphere.

From the formal analysis of the obtained data, it was observed that the characteristic reflection spectrum from the active side (foil side) contained two peaks, one in the region of about 1100 nm, and the other in the region of about 1600 nm. Over time, the spectrum transformed slightly, and the relative intensity of the peak’s changed. The objective parameter of these changes could not be described by the peak intensity alone, because the simplified measurement method used did not allow to determine the absolute value of the reflected light with high accuracy. However, the parameter of the relative strength of the peaks could be used as a ratio of the intensity of the reflected light in selected areas of the spectrum.

To objectively select these areas, the two-dimensional correlation spectroscopy (2D-COS) method [27] was applied. The 2D-COS approach examines the correlations that may exist between the time-varying reflectance spectra. The intensities of the peaks laying on the main diagonal of the 2D-COS synchronous spectrum (autocorrelation spectrum) are proportional to the relative extent of changes in the original spectra. The wavelengths of the most and the least intense peaks observed on the main diagonal of the 2D-COS synchronous spectrum were chosen for the calculation of the reflectance ratio vs. time:

R_ratio(t) = R(t, λ_max)/R(t, λ_min)

where:

t—time, λ_max, λ_min—wavelengths of the most and the least intense peaks on the main diagonal of the 2D-COS synchronous spectrum.

The raw reflectance data were used for 2D-COS synchronous spectrum calculation according to the equation [28]:

$$\mathsf{\Phi }=\frac{1}{m-1}A{\left(t,\lambda \right)}^{T}A\left(t,\lambda \right)$$

(2)

where: A(t, λ) is the m × n array of reflectance spectra data, m is the number of measurements, n is the number of points in the spectrum.

As a result of such analysis, it was possible to generate a series of 2D charts (the 2D-COS synchronous spectra) in which it was possible to determine the areas corresponding to the ranges on the reflection spectra, the intensity ratio of which was most sensitive to changes in the transition from spectrum to spectrum, in other words in the time domain. These 2D plots are shown in Figure 5 for the selected sample group (samples S2 and S6).

The analysis of the 2D-COS synchronous spectra showed that the spectra most sensitive to changes in the intensity ratio were not the same for the individual analysed samples and also depended on the side of the sample from which the reflection spectrum was measured, from the foil side or the electrode side.

We observed that: first, samples without the sputtering electrode had a similar reflectance spectrum regardless of the side from which the reflection was measured; second, the most substantial peak of the reflected light intensity lay in the region of 900–1000 nm (Figure 5c,d), while the exact location of the file was not stable from sample to sample and for the same sample over time. In diagram Figure 5b, there is a characteristic area of 1050–1150 m (the orange area in the chart). In the remaining areas, the 2D-COS synchronous spectra characteristics are similar to the characteristics presented in Figure 5c,d. The characteristic in Figure 5a obtained during the measurements from the Ag foil side was completely different from the others and did not undergo any noticeable changes in the time domain.

The analysis of the 2D-COS synchronous spectra allowed us to select the optimal spectral regions to determine the R_ratio(t) parameter for a series of measurements for each of the measured samples. These results are shown in Figure 6a, Figure 7a, Figure 8a and Figure 9a in the form of the autocorrelation spectra (main diagonal of the 2D-COS synchronous spectrum). and the dependence of the R_ratio(t) parameter in the time domain over 720 h is shown in Figure 6b, Figure 7b, Figure 8b and Figure 9b. In Figure 6b, Figure 7b, Figure 8b and Figure 9b, the R_ratio(t) parameter was normalised, to compare the results, to a value of 1.0 at the measurement start point (for t = 0).

According to the autocorrelation spectra presented in Figure 7a, the chosen spectral regions for calculating the parameter R_ratio(t) were recorded as follows:

• S1: R_ratio(t) = R(t, 1278 nm)/R(t, 1583 nm)
• S2: R_ratio(t) = R(t, 1321 nm)/R(t, 1583 nm)
• S3: R_ratio(t) = R(t, 1314 nm)/R(t, 1583 nm)
• S4: R_ratio(t) = R(t, 1268 nm)/R(t, 1583 nm)

As can be seen, the measurements of the R_ratio(t) parameter varied over time, for this series of samples lacking a sputtered silver electrode, up to a maximum of 300%. The nature of the changes was very similar to that observed in the two series of measurements presented in Figure 5 and Figure 6, regardless of the side from which the sizes of the reflection spectrum were determined. It can be concluded that the critical areas on the measurement time axis were approximately 4–6 days and 22–24 days. Under these measurement conditions, assuming that the parameters of the transparent film did not change, it can be concluded that the changes in the R_ratio(t) parameter may be directly related to the changes in the optical properties of the perovskite layer and, in a sense, represent its so-called ageing.

According to the autocorrelation spectra presented in Figure 8a, the chosen spectral regions for the calculation the parameter R_ratio(t) were the following:

• S1: R_ratio(t) = R(t, 1059 nm)/R(t, 1332 nm)
• S2: R_ratio(t) = R(t, 1108 nm)/R(t, 1397 nm)
• S3: R_ratio(t) = R(t, 1176 nm)/R(t, 1397 nm)
• S4: R_ratio(t) = R(t, 1128 nm)/R(t, 1397 nm)

As can be seen from Figure 9b, which presents the measurements from the Ag foil side, significant changes in the R_ratio(t) parameter were not observed over time, i.e., the parameter fluctuations did not exceed 2%.

The results presented above indicate that in the reflectance spectra measured with the use of the express-measurement method using the DLP NIRSCAN NANO spectrometer, the parameter R_ratio(t) = R(t, λ₁ nm)/R(t, λ₂ nm) can be determined, which can characterise changes in the optical characteristics of light reflection from the perovskite solar cell structure over time and be a specific ageing parameter of the cell. The obtained data show that this process is not monotonic. In the applied tracking area, the most characteristic changes were observed after 4–5 and 21–22 days from the start of the measurements (12–12 and 28–29 days from the production date). Obviously, introducing this type of measurement into the production cycle and starting the measurements immediately after the production of cell structures may provide a result that can be directly applied and considered in the technological process.

4. Conclusions

The spectra of optical reflectance in the NIR range for perovskite cell structures were analysed. Due to the possible changes in the absolute values of the spectra, the ratio of the reflectance for the wavelength for which the most significant changes were observed over time to the reflectance for the wavelength for which the most minor changes were observed was selected as an indicator for the evaluation of cell ageing. The 2D-COS method was used to determine the wavelengths. Two groups of samples (without the sputtered silver electrode and with the sputtered electrode) were characterised by the different shapes of the spectra and the different shapes of the reflectance ratio as a function of time. The reflectance ratio characteristics were similar for samples without a sputtered silver electrode, regardless of the surface (the side with the foil and the side with the active layer) on which the reflectance was measured. The greatest changes in reflectance were observed in the region between 1000 nm and 1350 nm, and the smallest changes in the region around 1600 nm. Characteristic differences in the reflectance ratio over time were observed after 4–5 days and 22–24 days. The most significant changes in the reflectance ratio after 30 days amounted to 300%. Much more minor changes in the reflectance ratio were observed for the samples with the sputtered silver electrode. For measurements from the foil side, the ratio changes amounted to a maximum of 25% after 30 days of measures and to a maximum of 3% for sizes from the silver foil side. Hence the conclusion is that the silver electrode protects the active layer against degradation to some extent. The presented method allows finding the moment at which changes in the physicochemical properties of the cell’s active layer appear in a non-destructive and contactless way at any stage of the cell production.

In relation to the physical mechanism of the changes in the spectra, the most common degradation pathway is the generation of iodine vapour, I—which has low defect formation energies [29]. When under storage at ambient conditions, the iodine vapour could escape the perovskite lattice causing further degradation [30].

Metal-based counter electrodes could prevent this to a certain extent [31]. However, under illuminative or electrical stresses, the reaction of metal electrodes with iodine is a cause of defect formation due to the exhaustion of iodine from the perovskite lattice under the influence of an electric field [32]. We still have to carry out future experiments to quantify the chemical species observed by NIR reflection spectroscopy.

The applied method uses miniature, very mobile equipment. It allows us to conduct and record a six-fold measurement of the entire spectrum in a few seconds and is characterised by exceptional mobility and simplicity. Such approach, having an obvious limitation as to the depth of the analysis of the processes involved, opens a new path for the practical study of the temporal stability of perovskite solar cells.