Effect of Precursor Stoichiometry on the Performance and Stability of MAPbBr3 Photovoltaic Devices

The wide band gap methylammonium lead bromide perovskite is promising for applications in tandem solar cells and light-emitting diodes. Despite its utility, there is only a limited understanding of its reproducibility and stability. Herein, the dependence of the properties, performance, and shelf storage of thin films and devices on minute changes to the precursor solution stoichiometry is examined in detail. Although photovoltaic cells based on these solution changes exhibit similar initial performance, the shelf-storage depends strongly on the precursor solution stoichiometry. While all devices exhibit some degree of healing, the bromide-deficient films show a remarkable improvement, more than doubling in their photoconversion efficiency. Photoluminescence spectroscopy experiments performed under different atmospheres suggest that this increase is due in part to a trap healing mechanism that occurs upon exposure to the environment. Our results highlight the importance of understanding and manipulating defects in lead halide perovskites to produce long-lasting, stable devices.


Introduction
Hybrid organic-inorganic lead halide perovskites have earned a lot of research attention in the past decade due to their broad-spectrum absorption and efficient photocurrent generation in solar cells, with record performances reaching those of silicon (24.2% photoconversion efficiency, PCE, at the time of writing) [1] . In order to reach commercial potential, however, perovskite solar cells must achieve these high values reliably from device to device, and furthermore, must retain their performance at least long enough to recoup production costs.
Despite predictions of high defect tolerance [2,3] , these two aspects remain elusive, with many factors playing overlapping roles in the device behavior over time. It is known, for example, that the interactions at the device interfaces can contribute to device breakdown [4][5][6][7] . Another key aspect determining device stability is related to the microstructure of the perovskite active layer [8] , with large, uniform grains having been shown to be more stable than small grains upon exposure to both oxygen [9] and humidity [10] . The composition of the active layer also plays an important role, with multi-cation compositions showing an overall higher stability [11][12][13] . Even upon taking these factors into account, literature reports concerning device stability vary greatly even for devices fabricated from the same recipe in the same device architecture [14] . This suggests that stability and reproducibility issues share a link, in which variation in the reproducibility of device performance may also lead to variation in its stability. For example, certain recipes for fabrication of perovskite layers may result in non-homogenous films, which may also serve to increase sample-to-sample variation [15] . Recently, our group uncovered one key factor that adversely impacts the reproducibility and stability of MAPbI3 perovskite solar cells: the exact stoichiometric ratio of the precursor solution. We showed that purposefully adjusting the ratio between the precursor components (methylammonium iodide (MAI) and lead acetate trihydrate (Pb(Ac)2)) in small, almost negligible amounts, results in large variations 3 in the subsequent photovoltaic (PV) performance and stability [16] . These stoichiometric variations are correlated with the photoluminescence (PL) behavior of the MAPbI3 thin films, displaying significant differences in both the initial PL quantum efficiency (PLQE) and its evolution after exposure to light and oxygen [17] .
To date, much of the work on stability and performance has focused on the high achievers of the perovskite PV family: MAPbI3 and the triple cation film composition. While solar cells using methylammonium lead tribromide (MAPbBr3) perovskites show much lower PCE, its wide band gap and corresponding high V OC offer the potential for inclusion into tandem cells, where a MAPbBr3 absorber is combined with a narrow band gap material in order to collect additional photons and achieve higher performance [18][19][20] . Furthermore, its light emission at 545 nm is ideal for application in light-emitting diodes (LEDs), as green is one of the fundamental pixel colors [21][22][23][24][25] . Despite the critical importance of reproducibility and stability for MAPbBr3 films and devices, few reports exist addressing these issues. Apart from early works that suggest MAPbBr3 to be stable upon exposure to light and elevated temperature [26,27] , no systematic studies addressing the stability and reproducibility of such films could be found.
In this report, we carefully examine the effect of precursor solution stoichiometry on the properties, performance and storage stability of MAPbBr3 thin-films and devices. By deliberately and incrementally adjusting the ratio of MABr to Pb(Ac)2 in the precursor solution, similar to our previous work on MAPbI3, we tune the composition of the film from slightly bromide-deficient to bromide-excessive, and examine the response of the active layer in PVs and LEDs. While the initial performance of devices shows little dependence on the stoichiometry of the precursor solution, the evolution of their performance upon storage varies drastically. Remarkably, the slightly understoichiometric bromide-based films show a large degree of defect healing that is evident in both the PLQE and device performance, resulting in a significant increase in its PCE upon storage. Our results underline the strong role that film composition plays in defining the properties, performance, and stability of devices, and further 4 promote the idea that defect engineering may be a viable strategy to produce desirable properties in perovskites.

Optoelectronic Properties
To tune the composition of the MAPbBr3 films, we employed the same strategy as was previously used for MAPbI3 [16,17] , based on the one-step lead-acetate method for film fabrication, as depicted schematically in Figure 1a [28] . This recipe has been shown to produce compact films with a highly uniform composition [15,29] , reducing pixel-to-pixel or spot-to-spot variation in the measurements [30] . Here, the lead acetate and methylammonium bromide precursors are weighed and dissolved in DMF such that the molar ratio of MABr:Pb(Ac)2 (denoted as y) is under the "ideal" stoichiometry of 3i.e. y = 2.95. Following the fabrication of the devices at y = 2.95, a small amount of MABr stock solution in DMF is added to the precursor solution such that the stoichiometry is now 2.97 MABr:Pb(Ac)2. By repeating this method, we create a series of films constituting the profile shown in Figure 1a One possible change that might be observed is the broadening or narrowing of the optical gap, which would be indicative of changes to the electronic structure of perovskite. As can be seen in Figure 1b, the UV-Vis absorption of each film, the absorption onset at 550 nm and peak position near 530 nm is independent of solution stoichiometry; therefore, the band gap is tolerant to the range of error in precursor composition introduced by varying y. The normalized photoluminescence spectra (Figure 1c) agree with this observation, demonstrating the same peak position and shape for each film. This is also the case for MAPbI3, where both under-and over-stoichiometric films show an absorption onset of about 780 nm [16] . The impact of the precursor solution stoichiometry on film surface composition is, however, directly observed by X-ray photoelectron spectroscopy (XPS), as shown in Figure 2a. Here, the intensity of the Pb4f7/2, Br3d5/2, and N1s peaks was tracked for seven spots on two films of each y, allowing us to qualitatively assess the surface uniformity in addition to surface chemistry (with the full survey being shown in the supplementary information, Figure S1). As shown by the triangles in Figure 2a, the ratio between the atomic percentages of Br:Pb increases slightly with increasing MABr:Pb(Ac)2 in solution, ranging from 3.75 to a maximum of 4.1.
Similarly, the amount of methylammonium at the surface, deduced by tracking the ratio of the N to Pb, also increases, ranging from 1.55 to a maximum of 1.75 for the overstoichiometric films. These trends agree with those observed for MAPbI3; however, the I:Pb and N:Pb ratios in these films increase more rapidlyrising at a rate of 0.075 per 0.01 change in y, whereas the bromide films only rise at a rate of 0.05 per 0.01 change in y. Comparable error bars indicate that the surface of the films is uniform both across a single sample and between multiple samples. Notably, the ionization potential (Figure 2b) is largely unchanged over the range of stoichiometries presented here. This contrasts the results obtained for MAPbI3 films, where the ionization potential increases monotonically by 0.3 eV over the range Δy = 0.12 [16] .

Microstructure
The microstructure of films often corresponds to properties observed; for example, a high JSC often correlates with a large grain size [31] . Therefore, we evaluated the surface structure of the films via scanning electron microscopy, shown in Figure 3. Unlike its iodide counterparts, where the microstructure is largely unchanged over Δy = 0.1 [16,17] , the microstructure of the

Photovoltaic Performance and Stability
Despite the changes in film microstructure and chemical composition at the surface, the initial performance of photovoltaic cells is highly tolerant to changes in y (Figure 4, with device structure shown in Fig. 1a). The open-circuit voltage (Voc, Fig. 4b) of the understoichiometric films is somewhat lower than that of the overstoichiometric films; however, this difference approaches the sample-to-sample variation. This observation coincides with the lack of change 8 observed in the ionization potential ( Figure 2b) and bandgap (Figure 1b,c). The short-circuit current density (Jsc), fill-factor (FF), and PCE are all approximately constant as y varies. These results contrast the MAPbI3 films, where an increase in the MAI:Pb(Ac)2 ratio (and the I:Pb ratio on the surface of the films) coincides with an increase in the VOC, varying linearly over a range of 0.2 V for Δy = 0.1, with the PCE following suit [16] . Following the initial measurements, the devices were covered and placed on a shelf in the lab, to be remeasured again at 10-and 24-day intervals. Interestingly, they display stark differences due to y when aged over the course of several weeks. As shown in Figure 5a, the change in VOC over time depends on the precise film composition, with the y = 2.95 films continually increasing over the time period measured. The VOC for other y diminishes over time, or diminishes and then heals slightly, with no observable trend. The fill factor remains roughly constant within the pixel-to-pixel variation. The most apparent trend appears when examining the Jsc. For higher y, the current improves slightlyby about 1 mAcm -2 . As y decreases, such that the precursor solution and films contain less Br, the Jsc improvement is much more drastic.

(a) (b)
9 The most understoichiometric film, y = 2.95, has an initial value of -4.5 mAcm -2 ; at 24 days, this value has nearly doubled at -8.5 mAcm -2 . These strong changes result in the clear trend observed in the PCE, where the initial performance for all y is between 3.5 and 4%, but at 10 and 24 days, the y = 2.95 films have more than doubled, showing a PCE of almost 8%. The highest stoichiometry films (y = 3.03) only increase from 4% to 5% PCE over this time period, with the change in PCE versus y displaying a linear increase with decreasing Br content. For longer shelf storage of over 100 days, the films retain this property of highly increased performance for low y and slightly increased performance for high y, suggesting the possibility for long-lived devices when combined with proper encapsulation strategies. This shelf-storage behavior is very different for MAPbI3 films, where the Voc increases between 10 and 20 days for all but the highest stoichiometry studied (y = 3.075), and the Jsc and PCE both decrease over time for all y (i.e. no such healing is observed) [16] . PLQE than the overstoichiometric films [16] . Such behavior is likely a result of the increased formation probability of deep trap states for halide-rich films [2,32] , increasing the rate of non- show only very small increases in PLQE or no change at all. We note that for all stoichiometries we observe no shift in the PL peak position or change in its spectral shape throughout the experiment ( Figure S2). This suggests that the changes in the PLQE are associated with defect healing, rather than changes in the emission properties of the perovskite layers. This PLQE healing behavior is also seen for MAPbI3 films under exposure to oxygen atmosphere [16,17] . A possible explanation has been attributed to the diffusion of oxygen into iodine vacancy sites, and the subsequent formation of a superoxide species [33,34] . This species is noted to be the appropriate size to fill the vacancy and is thought to be responsible for an initial boost in PLQE under oxygen exposure (and the later decomposition of the film into leadhalide, the organic cation, water, and oxygen). A similar mechanism could be responsible for the increase in PLQE in MAPbBr3 films, with the smaller overall healing as compared to Ibased films resulting from the shorter Br-Pb bond length, subsequent lattice stabilization, and lower prevalence of defect states deep in the band gap [2,3,35,36] .
When comparing the effect of solution stoichiometry on the properties of MAPbBr3 perovskites and those of MAPbI3, two key differences stand out. First, the microstructure of the former is were kept the same, the differences are likely to originate from different active layer microstructures and densities of defect states. The microstructure of the perovskite active layer has been shown to influence its stability with degradation processes been shown to commence at grain boundaries in both O2 [9] and humid environments. [38] In the case of MAPbI3, films of all stoichiometries exhibited the same microstructure and so it is predominantly the different densities of ionic defects that determined the storage stability of these devices. In the case of MAPbBr3, both the microstructure and the density of defects vary, making assignment more complex. However, it is interesting to note that understoichiometric MAPbBr3, which did not exhibit clear grain structure, are also the most stable, in agreement with the previously observed initiation of degradation at the grain boundaries. This suggests that for these samples, defect healing results in increased photovoltaic performance, while the smooth grain boundry-free microstructure delays the onset of degradation. However, given the high degree of overlapping phenomena between the microstructure, PV performance, PLQE, and their behavior over time, further research is needed to clearly elucidate the mechanisms at play in this healing behavior of the films. What our experiment clearly shows, however, is that the understoichiometric films display a high degree of improvement over time, indicating that the manipulation of defect states within the bulk film could be an effective strategy for increased film and device lifetime.

Conclusion
In conclusion, we examined the properties, performance, and shelf stability of MAPbBr3 films and devices as a function of changing precursor solution stoichiometry. While the initial PV performance is constant over changing y, the films with a small bromide deficiency undergo a large increase in performance, with their PCE more than doubling in value from 3.5 % to nearly 8 %. Films with excess bromine also improve with prolonged shelf storage, but only by 1%.
The maximum ELQE shows a similar trend. PLQE measurements under nitrogen and dry air measurements indicate that there is likely a trap healing mechanism at play, though the details 13 of such a process remain to be closely examined. Though the overall trends are different than those of the MAPbI3 films, both sets of results indicate the strong role that compositional variation plays in device longevity and film properties, suggesting that purposeful integration of defects into perovskite films may promote their eventual use.

Experimental Section
Sample and Device Fabrication: MAPbBr3 precursor solutions were prepared following the method introduced in previous works [16,17] . Photoemission Spectroscopy: The MAPbBr3 samples were transferred into an ultrahigh vacuum (UHV) chamber of the PES system (Thermo Scientific ESCALAB 250Xi) for measurements.
The samples were exposed to air only for a short time span of approximately 30 seconds. All measurements were performed in the dark and several spots on each sample were measured in order to ensure enough statistics. Ultraviolet photoelectron spectroscopy measurements were carried out using a double-differentially pumped He discharge lamp (hν = 21.22 eV) with a pass energy of 2 eV and a bias at −10 V. XPS measurements were performed using an XR6 monochromated Al Kα source (hν = 1486.6 eV) and a pass energy of 20 eV.

UV-Vis Spectroscopy:
Optical absorption spectra were measured with a Jasco UV-660 spectrophotometer in the range from 400 to 700 nm. The absorption of the substrate was subtracted as a baseline correction.
Photoluminescence Spectroscopy: PLQE measurements were carried out inside an integrating sphere (LabSphere) with excitation by a 405 nm CW laser (Coherent). The spectra were recorded using a QE65 Pro (Ocean Optics) spectrometer.
Scanning Electron Microscopy: SEM imaging was performed using a JSM-7610F FEG-SEM (Jeol). Samples were mounted on standard SEM holders using conductive Ag paste to avoid sample charging. The images were recorded using the secondary electron detector (LEI) at an acceleration voltage of 1.5 kV and a chamber pressure <10 −6 mbar.