Perovskites on Ice: An Additive‐Free Approach to Increase the Shelf‐Life of Triple‐Cation Perovskite Precursor Solutions

Abstract The development of stable perovskite precursor solutions is critical if solution‐processable perovskite solar cells (PSCs) are to be practically manufacturable. Ideally, such precursors should combine high solution stability without using chemical additives that might compromise PSC performance. Here, it was shown that the shelf‐life of high‐performing perovskite precursors could be greatly improved by storing solutions at low‐temperature without the need to alter chemical composition. Devices fabricated from solutions stored for 31 days at 4 °C achieved a champion power conversion efficiency (PCE) of 18.6 % (97 % of original PCE). The choice of precursor solvent also impacted solution shelf‐life, with DMSO‐based solutions having enhanced solution stability compared to those including DMF. The compositions of aged precursors were explored using NMR spectroscopy, and films made from these solutions were analysed using X‐ray diffraction. It was concluded that the improvement in precursor solution stability is directly linked to the suppression of an addition‐elimination reaction and the preservation of higher amounts of methylammonium within solution.


Experimental Methods
Note 1 -Device Fabrication.

Note 2 -Arrhenius Relationship
We explored the time-dependent aging of the perovskite precursor solutions at a series of different temperatures. This allowed us to estimate an overall activation energy for the reactions taking place.
Here, we used the Arrhenius relationship [1] which we have used to relate the reduction in device efficiency (caused by the degradation of the precursor solution) to its storage temperature.
ln( ) = − ) * + , -+ ln( ) Here, EA is an activation energy, kB is the Boltzmann constant, T is the solution storage temperature, A is a fitting constant and l is the degradation rate. Here, we define degradation rate using the following equation where PCEInitial is the average initial PCE of devices made from freshly prepared solutions and PCEn is the average PCE for devices that had been prepared from solutions that had been aged for n hours.

Note 3 -NMR Analysis
All analysis were carried out using TopSpin software. MA and FA peaks were identified by their chemical shift and comparison to 1 H reference spectra of individual components. To estimate absolute molarity changes, the area under each peak in the aged precursor solution was integrated and compared with the area under the internal standard. Overlapping peaks were deconvoluted using TopSpin software to extract their individual intensities. The internal standard used was 1,2,4,5-tetrachloro-3-nitrobenzene which was prepared at a concentration of 383 mM in d6-DMSO. This peak can be seen at 8.436 ppm in Fig S14-S23. This calibration peak however overlaps with a broad peak at ~8.4 ppm. We speculate this broad peak is made from several unresolved peaks corresponding to the NH components of FA and MA usually found at 9.0 ppm and 8.7 ppm for FA and 7.8p pm for MA. Here, we suspect that significant proton exchange between molecules may cause these peaks to merge and broaden. To isolate the signal corresponding to the TCNB standard (and thereby also determine the area of the other peaks), we have use the deconvolution feature in the TopSpin software. An example of this deconvolution process is shown in Figure S1a. Here it can be seen that the individual peaks can be successfully isolated and their integrated area determined. We acknowledge that this deconvolution process may lead to an uncertainty in the integration values which we believe to be around ±10%.
We have also performed a 1 H NMR calibration experiment, measuring TC solutions containing 3 different total solids concentrations: 225mg/ml, 774mg/ml and 901mg/ml. For each of these samples, the integrated areas of the FA-CH and MA-CH3 peaks are calibrated with reference to the area of the (known concentration) TCNB standard, with this process utilised throughout this study. We present the results of this experiment in Figure S1b. Figure S1b show that the molarities determined scale in an approximately linear fashion with concentration for both the MA and FA peak; a result that indicates that this methodology can accurately measure concentration within solution. The molarity determined from each peak is detailed in Tables 6-8. The measured intensities have been scaled where necessary to calculate molarity to compensate for the number of hydrogen atoms (or protons) contributing to it, e.g. in each MA CH3 group there are three hydrogen atoms which contribute to one peak. Absolute molarity changes shown in Figure 5d and SW were calculated using = @=@A − 3 , where ΔM is the change between the initial molarity Minit and the molarity at time c. d.

Figure S3 a-d. Average and standard deviation of the performance metrics for devices made from a TC-mixed (DMF:DMSO 4:1) solution aged at various temperatures over a 3 week period. Part e shows a linear plot of log(λ/t) against 1/T (see equation 1) where k = λ/t represents rate of performance reduction per hour and T is temperature in K.
Here a best fit to the data gives the activation energy EA/kB.
e. f.    Figure S5. Here the TC dissolved in DMSO solvent and stored at LT. Devices were then made from these solutions at several time points as indicated. Figure S7. Images of all films prepared for solutions aged for 4 months together with a control film prepared from a freshly prepared solution.

Figure S13. XRD patterns for films made from 115-day-aged TC-DMSO solution stored at RT (orange) and LT (green) compared with a control made from a fresh DMSO solution.
(arb units)

NMR
Peak positions were calibrated with respect to an internal isolated standard of 1,2,4,5-Tetrachloro-3nitrobenzene (TCNB) (peak at 8.4364 ppm) in DMSO-d6 in a capillary tube which was placed inside the sample tube containing the precursor solution being measured. The spectra presented below therefore contain peaks that derive from both the sample and the standard. We highlight peaks associated with the TCNB standard using a blue box.
Note that the standard capillary tube used was washed before use with isopropanol and water. These species inside the capillary can be seen as contamination peaks at 4.3 ppm, 3.8 ppm, 1.0 ppm (for IPA) and 3.3 ppm (for water) in the following 1 H spectra. For the 1 H measurements, we find a chemical shift between the samples that are dissolved in different solvents.
It is important to emphasize that the exact chemical shift of any proton environment is highly dependent on its surroundings and can vary due to the due to concentration, pH or solvent co-ordination. Indeed, in all solutions (even though such solvents are deuterated), we observe some residual non-deuterated solvent (i.e. DMSO or DMF) that is generated through the replacement of deuterium by a proton. As there is a different chemical environment in the TC solution and TCNB solutions, we observe some shift of the peaks associated with the residual solvents; for example we see peaks at 2.2 and 2.5 ppm which are both associated with DMSO. Figure S14. 2 [3]). We tabulate the area of integrated regions as extracted from the spectra in the in Table 6.  Table 6. Integral areas and subsequent concentrations for the integrals shown in Figure S15.   320  1550  14  290  1550  50  40  21  219  1430  90  90  10  28  169  1380  145  155  20  Table 7. Molarity (in mM) of organic molecules within solution estimated from NMR measurements (within ±10%) of aging TC solutions in d7-DMF/d6-DMSO. This was measured using Topspin and referenced to the internal standard TCNB having a molarity of 383mM. Peaks consisting of multiple hydrogen signals have been scaled to reflect their true concentration.  Table 9. Molarity (in mM) of organic molecules within solution estimated from NMR measurements (within ±10%) of aging TC solutions in d7-DMF. This was measured using Topspin and referenced to the internal standard TCNB having a molarity of 383mM. Peaks consisting of multiple hydrogen signals have been scaled to reflect their true concentration. Figure S25. TC-d6-DMSO precursor solution absolute molarity changes calculated from NMR data. Comparing to data shown in Figure 5 of the main paper, there is a much smaller molarity change than occurs in comparably aged DMF/DMSO precursors.