Fluoride Chemistry in Tin Halide Perovskites

Abstract Tin is the frontrunner for substituting toxic lead in perovskite solar cells. However, tin suffers the detrimental oxidation of SnII to SnIV. Most of reported strategies employ SnF2 in the perovskite precursor solution to prevent SnIV formation. Nevertheless, the working mechanism of this additive remains debated. To further elucidate it, we investigate the fluoride chemistry in tin halide perovskites by complementary analytical tools. NMR analysis of the precursor solution discloses a strong preferential affinity of fluoride anions for SnIV over SnII, selectively complexing it as SnF4. Hard X‐ray photoelectron spectroscopy on films shows the lower tendency of SnF4 than SnI4 to get included in the perovskite structure, hence preventing the inclusion of SnIV in the film. Finally, small‐angle X‐ray scattering reveals the strong influence of fluoride on the colloidal chemistry of precursor dispersions, directly affecting perovskite crystallization.


Introduction
Metal halide perovskite materials have shown enormous potential for the processing of efficient and stable photovoltaics. [1] However,t he dominant type of perovskite solar cells (PSCs) are based on lead, am etal whose toxicity and environmental hazard can hinder its commercial application in numerous fields. [2] Thel ead threat pushed the scientific community to develop lead-free perovskite materials to maintain excellent photovoltaic performance while avoiding environmental risks.I nt his sense,t in halide perovskites are the best candidate to replace the dominant lead-based counterparts. [3,4] Nevertheless,t hese materials face some difficulties related to their inherent physicochemical characteristics. [5] Themost important one is the ease with which Sn II oxidizes into Sn IV species,leading to the substantial decline in the performance through the undesirable formation of electron traps and p-doping of the material. [6] Previous studies have reported many origins of this oxidation, such as the solvent, [7,8] thep rocessing conditions [9] or even spontaneously through disproportionation in tin-poor environments. [10] Stopping this oxidation is one of the requirements to achieve efficient and stable tin halide PSCs.For this reason, several trials have been made tackling the oxidation of Sn II . These include the use of new solvent systems to avoid the oxidation by dimethyl sulfoxide (DMSO), [11] employing reducing agents to eliminate the content of Sn IV ,s uch as metallic Sn powder [12] or hypophosphorous acid [13] or introducing additives for alleviating the formation of Sn IV ,like the ever-present SnF 2 . [6,14] SnF 2 has achieved remarkable success as an additive in the tin halide perovskite field. Since its first use in PSCs by Kumar and co-workers, [15] it has been proven over time as an imperative to achieve good results (Figure 1a). There is barely any good cell performance report without SnF 2 ; exceptional cases use SnCl 2 , [16,17] which may behave similarly to SnF 2 ,o r2Dm aterials,w hich are another popular strategy for processing tin-based perovskites. [18][19][20] Thea ppeal of SnF 2 in the community is such that the number of studies not using it (nor SnCl 2 )quickly stagnated over the years,being in 2020 below 10 %ofthe total publications on tin halide perovskites in that year ( Figure S1). Its success lies mainly in the impossibility to obtain photovoltaic behaviour in the solar cells fabricated without it. In Figure 1b,w ec ollected data from studies in which solar cells were made with and without SnF 2 . [15,[21][22][23][24][25][26][27] Theimprovement in efficiency for both inorganic and hybrid tin halide perovskites is enormous,with negligible efficiency for the SnF 2 -free cases.O ne of the most reported improvements is the better substrate coverage and film morphology obtained with SnF 2 , [22,25,28] which implies that SnF 2 affects the film crystallization, af actor that remains unexplored. Nevertheless,its addition needs to be controlled, as at oo-high content of SnF 2 is reported to induce phase separation. [22,25,28] Theother most explored effect is the ability of SnF 2 to reduce the formation of Sn IV and its related defects, with the reported benefits usually being reduced recombination [24] and ablue shift of the absorption onset. [24,29,30] Besides, Savill and co-workers found out that even low amounts of SnF 2 are sufficient to positively impact mitigating Sn IV formation in tin/lead perovskites. [29] This effect on oxidation suppression could originate from introducing aS n-rich environment, reducing Sn vacancies. [10] This interpretation has been proposed already in the first use of this additive by Kumar et al. [15] However,t he question of what fluoride is doing and why we do not provide the Sn-rich environment simply with ah igher SnI 2 ratio in respect to FAIr emains unanswered. Related to this,o ther Sn II species that could provide the same beneficial effect were already discussed by Yokoyama et al. [31] Using SnI 2 excess also led to good results in one of the first studies on inorganic tin halide perovskites. [32] Overall, reports in literature consistently lead to the same results:SnF 2 has acritical positive influence on the formation of high-quality ASnX 3 (where A = methylammonium (MA + ), formamidinium (FA + )and Cs + ,and X = Cl À ,Br À and I À )films and holds ap articular role in the stability of these materials against their oxidation to Sn IV .T hese two possibly related aspects are the key to SnF 2 being the predominant, most robust additive in the tin-based perovskite field. While there has been extensive exploration of the impact and functioning of SnF 2 in these thin films,t he chemical mechanism and influence in the processing are entirely unknown. Introducing an exact Scheme of its working procedure remains amust in the field. This move would open the door to optimized application and help identify new additives in the future.
In this work, we explain the origin of the beneficial effects of SnF 2 in the processing and stability to oxidation of tin halide perovskites by studying the chemistry of fluoride in these solutions.U sing ac ombination of complementary solution and film characterization techniques,w ep ropose that the role of SnF 2 is not limited to the resulting thin film but also affects the precursor solution properties critically and hence their processing.The study of the solution chemistry of fluoride in formamidinium (FA)-based FASnI 3 precursor solutions by 119 Sn-and 1 H-NMR revealed astrongly predominant affinity of the fluoride anion for Sn IV over Sn II .With the help of hard X-ray photoelectron spectroscopy (HAXPES) analysis,w es how how SnF 2 increases the Sn II content in perovskite samples,a ni ndication that Sn IV is partially prevented from being incorporated in the perovskite film. Meanwhile,s mall-angle X-ray scattering (SAXS) enables an understanding of how fluoride anion modifies perovskite subunits interaction in solution, generating improved homogeneous crystal growth conditions.Furthermore,experiments with other fluoride species and SnCl 2 prove that these effects are not exclusive to SnF 2 .Thus,the chemistry of ahard Lewis base like fluoride,c ombined with the Sn-rich environment, make SnF 2 ahighly suitable additive for processing tin halide perovskites.

Results and Discussion
Previous works concluded that SnF 2 could reduce the Sn IV content in solutions and films. [14,30] However,t he redox activity cannot explain the multiple effects of SnF 2 in ASnX 3 perovskites entirely.Therefore,wepostulated that SnF 2 must be involved in ad ifferent type of chemical reaction. In the early stages of tin halide perovskites development, it was thought that ay ellow color for the solution implied the elimination of Sn IV through its reduction by SnF 2 . [28] Using  [15,[21][22][23][24][25][26][27] NMR, we uncover that Sn IV and SnF 2 do not undergo aredox reaction, but as imple ligand exchange reaction, producing colorless SnF 4 in solution. In this regard, we prepared FASnI 3 precursors solutions with and without SnF 2 and Sn IV to investigate their signature chemistry. 119 Sn-NMR is sensitive to Sn nuclei in different electronic environments,allowing to identify of the Sn species existing in the solution, including other oxidation states of the same nucleus. Figure 2a depicts the change in color from orange to the pale yellow of a1MF ASnI 3 solution in DMSO after the addition of SnF 2 .F ollowing the same method, we added SnF 2 to aS n IV -containing solution after being aged by heating at 100 8 8Cf or 3h.S aidaminov et al. and we recently described that this thermal treatment promotes Sn II oxidation in DMSO solutions. [7,8] Thec olor of the aged solution goes from an intense red to apale yellow to that of the fresh sample.Unlike the fresh solution, we know that certain content of Sn IV is present for the aged one.S till, the color after SnF 2 for the aged solution is the same as for the fresh solution, suggesting that the characteristics and species in the solution are changing, no matter the Sn IV content. 119 Sn-NMR of the fresh solution indicates neither elimination nor formation of Sn IV ,even though its color changes to pale yellow ( Figure 2b). However, we observe the Sn II shielding resulting in ac hemical shift change from À574 ppm to À604 ppm. Regarding the Sn II species (SnI 2 and SnF 2 ), they cannot be differentiated in solution, as they show up in asingle signal belonging to the average electronic environment of Sn II in solution. In this sense,the addition of increasing amounts of SnF 2 shifts the Sn II peak to lower chemical shift values in afairly linear manner,asweshow in Figure S2. In contrast to the fresh solution, the aged FASnI 3 solution showed the expected SnI 4 signal at À2025 ppm. After the addition of SnF 2 ,t his peak disappeared, and an ew quintuplet rose at À770 ppm ( Figure 2c). To identify the newly formed species,w em easured solutions in DMSO of SnI 2 ,S nI 4 ,S nF 2 and SnF 4 by 119 Sn-NMR (Figures 2d), indicating that the species corresponds to SnF 4 .T his result, therefore,i mplies that SnF 2 cannot reduce Sn IV from an oxidized sample.I nstead, it coordinates Sn IV via al igand exchange reaction between the fluorides from Sn II F 2 and the iodides from oxidized Sn IV I 4 [Eq. (1)]: As shown in Figure 2e,acolorless solution comprising SnI 2 and SnF 4 presented both mixed species signals in NMR. However,m ixing SnF 2 and SnI 4 ,t he resulting NMR species observed were SnI 2 and SnF 4 .C onsequently,t he complexation selectivity of fluoride ions towards Sn IV is absolute ( Figure 2f). This can be easily explained by the "hard and soft (Lewis) acids and bases" (i.e.H SAB theory) nature of the different solution species.Fluoride is asmall, non-polarizable, very electronegative anion that shows as tronger affinity for acation of asimilar nature,that is,Sn IV ,which is smaller and more electronegative than its reduced analogue Sn II .T his hard Lewis base character of fluoride anions was already applied in previous works on lead halide perovskites,owing to its ability to passivate vacancies due to their strong bonds with Pb II . [33] In the present case,w ef ind that fluoridesr ole is connected to Sn IV complexation. Simultaneously,t he passivation of undercoordinated Sn II in the thin film should not be excluded.
Theonly appearance so far of this species can be found in Nakamura and co-workers work. Thea uthors use an SnF 2selective reducing agent to effectively generate Sn 0 nanoparticles to scavenge Sn IV from the solution. [34] Even though there is no particular discussion on the formation of SnF 4 and it does not affect their mechanism, the 119 Sn-NMR spectra provided in their work show the signal, also am ultiplet, corresponding to SnF 4 at approximately À750 ppm when adding SnF 2 to aS n IV -containing solution. This different multiplicity that the SnF 4 signal presents in 119 Sn-NMR compared to the rest of the Sn species can be explained by coupling between the Sn and halide nuclei. SnF 4 has four chemically equivalent 19 F-spins (each with spin 1/2) to couple with, which results in aperfect quintuplet as observed. In this sense,wewould expect to observe atriplet for SnF 2 .However, the four coordination sites are not saturated in SnF 2 ,a nd, thus,t here is an exchange with other impurity-related compounds,s uch as water. Birchall and DØnsf ound the same behaviour.T hey claimed that the missing coupling between 19 Fand 119 Sn is due to the exchange between different hydrated species of SnF 2 . [35] Our samples contain water since non-anhydrous [D 6 ]DMSO was used as as olvent for the NMR experiments,t herefore agreeing with the previously reported experiences.F or other species,n os plitting occurs since chloride,bromide,and iodide do not have NMR-active nuclei in significant amounts or with detectable line widths.
This affinity of fluorides towards Sn IV can have several important implications that may reduce Sn IV content in the final film. Forinstance,the strong preference for Sn IV means that fluorides could complex it as soon as it is generated, whether from O 2 in the environment or DMSO-driven oxidation. [7,8,11] In fact, one should think if SnF 2 would be as valuable for other solvents as in DMSO,asfluorides could be critical in sequestrating Sn IV as soon as it is oxidized by this solvent, making it less harmful. Thec onversion of SnI 4 into SnF 4 will also prevent the SnI 4 -driven degradation pathways recently described by Lanzetta et al. [36] Furthermore,t he selective complexation of Sn IV as SnF 4 may hinder its ability to form any perovskite-like complex in solution. It has been widely reported for SnF 2 that this materialse xcess tends to undergo phase separation. [22,25,28] Conclusively,i fS n IV is retained as SnF 4 ,i tw ould be challenging to incorporate this form into the perovskite lattice.Instead, it would be displaced to grain boundaries or even removed from the film. As aresult, the point defects resulting from incorporated Sn IV in the perovskite lattice can be significantly reduced. To prove this,w ec ompared the different Sn IV species ability to coordinate with FAIb ya nalyzing the 1 H-NMR of these solutions.F igure S4 shows how all SnI 2 (FASnI 3 ), SnF 4 and SnI 4 cause the splitting of the FAIa minic protons,p ointing out ac ertain degree of interaction between the species. However,t he signals for all N-and C-attached protons are slightly shielded in the FASnI 3 solution (where the formation of perovskite adducts in solution occurs), whilst for the case of SnF 4, there is no shielding, suggesting that the interaction of SnF 4 species might have al ower affinity towards perovskite precursors.M oreover,t he shift is very pronounced for SnI 4 , which might imply strong coordination with FAIa nd an increased ability to get incorporated in the perovskite, resulting in adverse consequences for the photovoltaic properties of the films.
(SnO 2 )o xide compounds are energetically overlapping with the respective Sn environments expected to be present in the sample set (i.e., FASnI 3 and the different Sn-based additives) and O-related lines were detected in the HAXPES measurements,t he Sn 4d spectra in Figure 3l ikely contain Sn oxide derived spectral features and thus consist of more than two doublet peaks.C omparing the spectra in Figure 3a with the spectra in Figure 3b demonstrates that the high BE Sn 4+related features are more prominent in the 2keV measurements than in the 6keV,w hich indicates an increased prevalence of the Sn 4+ related species (in the form of SnO 2 or SnX 4 )near the surface of samples than deeper within their bulk. However,significant differences in the line shape of the spectra for ag iven excitation set reveal pronounced changes in the Sn chemical environment of the investigated samples concerning the presence/absence and kind of additives during processing. Thes pectra of samples with additives containing Sn 2+ or F À display as ignificant increase in the low BE Sn 2+related signal. This finding seems to be in line with the NMR results described above,t hat capturing Sn IV in the form of fluorinated species prevents its incorporation in the films. However,the observed variability of properties at the surface of the samples associated with handling conditions of the Snbased perovskite samples (as has been already reported, [30] and further discussed in Supporting Information, see As depicted previously in Figure 2a,weattribute the color change of FASnI 3 in DMSO to the change in solution properties and not to the Sn IV content. Nevertheless,i ti s crucial to understand the underlying reasons that led to this change,a swee xpect it to exhibit ad ecisive influence on the crystallization dynamics. Here,weperform transmission small-angle X-ray scattering (SAXS) to reveal the effect of SnF 2 on the perovskite precursors in solution. Thereby,aproposed nucleation mechanism indicates that the use of SnF 2 promotes homogeneously distributed growth, yielding improvements in the overall crystal quality.Using the SAXS instrument at BESSY II at X-ray energies of 8keV and 10 keV (DE/E = 2 10 À4 ), we cover a q-range from 0.05 to 8.5 nm À1 (size range:2 09.4-0.74 nm). Figure 4a compares the SAXS scattering curve of ap lain FASnI 3 solution in DMSO with FASnI 3 containing SnF 2 .Atfirst sight, the comparison does not show significant variations.
Nevertheless,b ya pplying am odel fit using the software SASfit, [41] which offers several different form and structure factors describing various shapes of particles and their interaction, small changes regarding the particle interplay in the high q-region can be observed. Here,h owever,i nterpretation requires particular precaution since we are already in the proximity of interatomic distances.The general behaviour of the initial perovskite precursor stage is highly dependent on the specific solvent environment. Literature shows that solvents with al ower donating number interact weakly, whereas stronger donating solvents interact strongly with the metal of ap erovskite precursor solution. [42] Therefore, stronger donating solvents tend to hinder the iodide coordination of the metal. Since DMSO is known to be strongly donating and hence decelerates the perovskite crystallization process,w eh ere include N,N-dimethylformamide (DMF) with lower donating effect to investigate further the possible influence of SnF 2 on the early stages of crystallization. Figure S6 shows the same effect on the color of SnF 2 in DMF as in DMSO,a sp roof that the same visual transformation occurred.
Theevolution of amaximum in the SAXS scattering curve of FASnI 3 in DMF given in Figure 4bshows aclear difference compared to the scattering curve of pure FASnI 3 in DMSO. Themaximum emerges based on adominant structure factor, which evolves due to particle interaction. Themean spacing d between the mass centers of the individual interacting particles can be calculated as discussed by Raghuwanshi et al. using the magnitude of q at the peak maximum. [43] In the plain FASnI 3 solution in DMF,this results in amean spacing d of approximately 1.5 nm. Adding SnF 2 to the solution leads to as hift of this peak maximum to higher q and, consequently, lower mean d spacing of 1.2 nm. Besides the shift of the maximum, also the slope at lower q-values disappears.T he shallow negative slope for both DMF and DMSO solutions gives rise to the presence of larger structures with abroad size distribution (> 100 nm) in the solution. We propose that the larger sizes represent aggregates consisting of small interacting subunits formed by ion-to-ion attachment. We assign these subunits to particles or clusters in an average dimension of 0.4 nm observed in all scattering curves.Inthe DMF case, we assume that these aggregates form by nearly oriented attachment, as described in the non-classical nucleation theory being pre-ordered arrangements (Figure 4c). [44][45][46] Thewell-pronounced structure factor peak can evidence this, showing the recurring distance d between subunits,r epresenting the average distance between the mass centers of the units and could thus be considered the tin-to-tin distance due to the high electron density of tin. As pecific recurring distance d can also be noticed in the case of SnF 2 addition. However,there is no negative slope at low q-values assigned to larger higher-level structures.T herefore,wec onclude that the total size distribution generally appears to be more homogeneous;t he nearly oriented attachment with the recurring distance d of 1.2 nm might be considerably more extensive than in the plain FASnI 3 solution or even of infinite size.A dditionally,w ep erformed several runs for every sample to prove no damage caused by the beam (Figure S7).
Pre-ordered arrangements of subunits set the starting point for the further crystallization of at hin film on asubstrate.The broad size distribution of comparable smaller aggregates might result in films including unordered pores or pinholes because solvent evaporation leaves holes between the pre-ordered totals.I nstead, the more uniform size distribution due to al arger oriented attachment of the subunits supports homogeneously distributed crystal growth, asuitable substrate coverage and improved film morphology, precisely what is observed by SnF 2 addition in literature. [22,25,28] With the premise of the already advanced aggregation of elements in the DMF solvent, as imilar mechanism during the late stages of enhanced crystallization may be expected for the case of DMSO.Byapplying pressure via spin coating and solvent evaporation, the concentration of FASnI 3 in the solution increases.F ollowing the evolution of SAXS scattering curves for FASnI 3 concentration series in DMF and DMSO,itsuggests that astructure factor maximum is formed at higher concentrations in the case of DMSO comparable to the DMF solution ( Figure S8). In this sense, the observed behaviour for DMF can be extrapolated to amore advanced stage of precursor formation for the DMSO precursor solution. Therefore,SnF 2 as an additive leads to an in total more homogeneous crystallization of the tin halide perovskite thin film, and thus to abetter morphology.
Regarding the change in solution color caused by SnF 2 addition, we speculate that fluoride modifies the coordination level of tin centers by iodide ions,hindering the formation of colorful, highly coordinated [SnI x ] 2Àx units.T he fact that better morphology films are obtained through the pale yellow, SnF 2 -containing solution points out that, with solution color as an indication, the properties of the existing formations in solution critically influence the crystallization dynamics of tin halide perovskites.This feature is currently underexplored for these materials and proves to be much more complex and sensitive than for their lead analogues due to its quite restricted processing conditions.
Other SnX 2 (X = Cl, Br, I) and their influence on perovskite properties are also frequently discussed in the literature. [16,17,47] We further performed SAXS on FASnI 3 precursor solutions according to SnX 2 addition, given in Figure S9, to compare their respective functionality to the SnF 2 addition. Similar to the scattering curves shown in Figure 4a,n os ignificant influence or difference between different Xcan be noted. However, it should not be ruled out that they could have the same behavior difference as SnF 2 in other solvents like DMSO and DMF.T his confirms the need to investigate the strong dependence of additives and compositions used for tin halide perovskites.Finally,ascattering curve for the presence of Sn IV is given in the inset window in Figure S9, for which we measured an aged sample of FASnI 3 .T he effect of temperature-induced degradation of DMSO solutions on its properties seems relevant, confirming that we did not influence unexpected Sn IV content in FASnI 3 with or without SnX 2 .
Although SAXS detected no difference for the different SnX 2 additives in DMSO solutions,t hey still caused ac olor change in aclear trend (Figure 5a). Both SnF 2 and SnCl 2 led to asimilar yellow coloration of FASnI 3 solution, while SnBr 2 affected it mildly more.T his trend could mean that the colorful, highly coordinated [SnI x ] 2Àx iodostannates were hindered more strongly as the halide Xi saharder Lewis base.T he fresh solutions were analyzed by 119 Sn-NMR ( Figure S10), although no notable difference was found except for the shielding effect from SnF 2 ,a lready discussed in Figure 2. In this sense,S nF 2 is the only SnX 2 additive shifting the FASnI 3 signal upfield, while SnCl 2 and SnBr 2 go slightly in the opposite direction. This effect correlates well with the chemical shifts in the different SnX 2 -pure solutions in DMSO ( Figure S11), except for the SnI 2 case.Hence,the final Sn II signal position might be an average value of all Sn II species.A fter heating the solutions,w eo bserved that the solutionsdarkening was negligible for chloride-and fluoridecontaining solutions,w hich could be due to both lower Sn II oxidation and more efficient Sn IV complexation by these anions,a se xplained in Figure 2. Theh eated solutions were measured by 119 Sn-NMR (Figure 5b), showing that the content of Sn IV ,all in the form of SnF 4 ,had been significantly reduced in comparison to SnF 2 -free heated solution (Figure 2c). We hypothesize that fluoride could affect DMSO and Sn II environments,maybe through the modulation of [SnI x ] 2Àx adducts,m aking these two species less eager to undergo ar edox reaction. Also,w eo bserved that SnCl 2 addition had the same effect as SnF 2 ,l eading to both reduction of the oxidation and the selective complexation of Sn IV through the formation of SnCl 4 (Figure 5c). These results prove that hard Lewis bases like chloride and fluoride can block the formation of Sn IV in the solution and their introduction into the perovskite film through two different mechanisms:c omplexation of Sn IV and antioxidative character.O ur findings agree with previous reports on reducing Sn IV by the addition of SnF 2 [23,28,30] or SnCl 2 [16,17] and suggest that many of the other additives employed in literature for tin halide perovskites may work in the same fashion.
To confirm that fluoride was responsible for these changes in solution, we prepared FASnI 3 solutions containing other fluoride-based compounds.U nfortunately,o ther common species (i.e.C sF and NaF) had limited solubility in common solvents.T herefore we had to saturate the FASnI 3 solution below a5 %m olar ratio ( Figure S12). Even though the concentration was lower than for SnF 2 ,weobserved the exact change in color from orange to pale yellow,p otentially affecting perovskite subunits in solution in the same fashion. Similarly,t hese solutions experienced no darkening of solutions aged at 100 8 8Cf or 3h in different conditions, proving the complexation of Sn IV in the form of SnF 4 .E ven though these particular additives may not be directly implementable due to the strong influence that Cs + and Na + cations can have in the perovskite solar cells processing and performance,t hese results confirm the universality of the working principle for fluoride-based compounds.F urthermore,t hey suggest that SnF 2 additive may be eventually replaceable by other fluoride-based species if applied in the right conditions.
To complete the study,wewanted to investigate how these changes in perovskite solution properties affect the thin film formation and the corresponding solar cells performance.W e fabricated pristine FASnI 3 films and with 10 %o ft he excess of FAI, SnI 2 ,SnBr 2 ,SnCl 2 and SnF 2 .Excess FAIwas tried to study both stoichiometry sides of FASnI 3 .While there was no notable change in the X-ray diffraction patterns ( Figure S13), the scanning electron microscopy (SEM) results offered some differences among samples ( Figure S14). Thes ample with 10 %S nI 2 excess was the only one showing ah igh density of extensive pinholes.I nc ontrast, the film with 10 %S nF 2 was the most homogeneous one,free of pinholes and other minor irregularities that are present in the rest of the films,agreeing with previous papers that used SnF 2 on its beneficial effect on morphology. [22,25,28] There is an evident change in the resulting grain size with the changing halide element ( Figure S15). Fluoride led to the smallest average grain size (568 nm) compared to chloride (629 nm) and bromide (623 nm). These results are orthogonal to those reported previously,where fluoride [25] and chloride [17] were said to increase the perovskite crystals grain size. However,tin halide perovskites sensitive nature implies that significant changes can be expected from minor modifications in the perovskite composition or processing.T herefore,t he effect of halides introduction can vary from study to study. Also,stoichiometry in pure FASnI 3 perovskite strongly affects the pinhole density and the average grain size.The largest size was found for equimolar FASnI 3 (695 nm), which went down when increasing or decreasing the SnI 2 ratio (647 and 568 nm, respectively). Moreover,S nF 2 addition shows an impact on the size distribution itself,w hich is significantly narrowed to plain FASnI 3 thin film. These observations agree with the results by SAXS,a ssuming that au niformly nearly oriented attachment in solution leads to am ore homogeneous distribution of the grain sizes in the film.
We then used these films for solar cells fabrication (more details in the Supporting Information) to investigate any possible trend between additives and performance.I nt his sense,a dding as mall portion of tin halides to FASnI 3 solutions seems beneficial for the device performance,showing some positive trend when moving to lower size halides ( Figure S16). Though it appears that smaller halides-harder Lewis bases-work better by having amore decisive influence in the processing, chloride was the exception. Even though NMR and SAXS found SnCl 2 to have very similar behavior to SnF 2 ,t he resulting devices yielded no efficiency,s uggesting that chloride brings other factors into play.P revious works point out how chloride can be incorporated in the lattice and its tendencyt of orm massive aggregates, [16,17] making its application not as trivial as SnF 2 and requiring amore careful optimization. We suspect SnCl 2 could mimic SnF 2 to some extent in these solutions if the processing conditions are adjusted accordingly.I ti sa lso worth noting the slight improvement in efficiency produced just by using a1 ,1:1 SnI 2 :FAI stoichiometry (i.e.1 0% SnI 2 excess), despite the content of irregularly sized pores ( Figure S14). This matches well with the results in previous works, [11,32] proving the importance of providing aS n-rich environment in the film.

Conclusion
SnF 2 is aw idely used additive for tin and lead/tin halide perovskites,s ystematically showing the same beneficial effects in all reported studies:p erovskite films with lower Sn IV content and improved morphology.W eu ncovered the different roles of fluoride in SnF 2 on Sn IV complexation and colloidal arrangement in the precursor solution. By studying the fluoride chemistry in perovskite solutions and films with different complementary techniques,w ed emonstrated that the fluoride in SnF 2 has acritical role in reducing Sn IV content in the precursor solution and the final perovskite film. We showed by NMR the selective complexation of Sn IV in the form of SnF 4 ,w hich HAXPES revealed to have al ower tendency to get introduced in the film than SnI 4 .M oreover, we showed how the introduction of SnF 2 in perovskite solutions increases their stability against the oxidation caused by DMSO.T his antioxidative character was also found for SnCl 2 ,m eaning that many other reported additives for tin halide perovskites may also block Sn II oxidation by simple tuning of solution properties.A part from reducing Sn IV content in the thin film, SAXS measurements on the related precursor solutions evidenced that fluoride alters the essential formation of pre-organized perovskite clusters.W eidentified an advanced colloidal arrangement in DMF compared to the DMSO solutions that are notably influenced by the addition of SnF 2 .W ea ssigned this arrangement to an advanced nucleation process in DMF compared to DMSO.F inally, based on our findings,w ep roposed an ucleation mechanism that occurs in solution and is affected by the SnF 2 addition resulting in improved overall crystal quality.Inthis sense,the effect of SnF 2 on the film processing will be strongly determined by the environment in which it is applied (i.e. solvent, perovskite composition). Consequently,t here is an immediate need to fundamentally understand and optimize solution properties,their processing and studying the effect of additives.A sw ea re doing in this study with the example of SnF 2 as pioneering work and impulse for further research. Overall, we presented ac omplete comprehensive picture of the working mechanism of SnF 2 in tin halide perovskites processing and provided the community with the guidelines for finding new additives with specific chemical properties to selectively complex Sn IV species and regulate the crystallization.