Intrinsic Formamidinium Tin Iodide Nanocrystals by Suppressing the Sn(IV) Impurities

The long search for nontoxic alternatives to lead halide perovskites (LHPs) has shown that some compelling properties of LHPs, such as low effective masses of carriers, can only be attained in their closest Sn(II) and Ge(II) analogues, despite their tendency toward oxidation. Judicious choice of chemistry allowed formamidinium tin iodide (FASnI3) to reach a power conversion efficiency of 14.81% in photovoltaic devices. This progress motivated us to develop a synthesis of colloidal FASnI3 NCs with a concentration of Sn(IV) reduced to an insignificant level and to probe their intrinsic structural and optical properties. Intrinsic FASnI3 NCs exhibit unusually low absorption coefficients of 4 × 103 cm–1 at the first excitonic transition, a 190 meV increase of the band gap as compared to the bulk material, and a lack of excitonic resonances. These features are attributed to a highly disordered lattice, distinct from the bulk FASnI3 as supported by structural characterizations and first-principles calculations.

I n a decade since the landmark works of the Snaith and Graẗzel groups, 1,2 lead halide perovskites (LHPs) have paved their way aside from photovoltaics toward many eminently different applications ranging from backlit displays and light-emitting diodes to hard radiation and neutron detection. 3−15 Such a broad utility of LHPs is rooted in a rare combination of their properties, including low density of carrier trap states (10 9 −10 10 cm −3 ) despite a large density of point defects, 5,16,17 long electron−hole diffusion lengths (2− 175 μm), 5,18,19 high carrier mobilities (2.5−1000 cm 2 V −1 s −1 ), 5,17,20−22 long charge carrier lifetimes (0.08−450 μs), 5,[18][19][20][21]23 small carrier effective masses (0.069−0.25 m 0 ), 24 and high optical absorption coefficients at the absorption edge (2−7 × 10 4 cm −1 ). 17 Although LHP-based devices may meet RoHS compliancy in some cases, 25,26 the anticipated scale of the LHP market calls for lead-free alternatives. 27 Compelling electronic characteristics of LHPs largely arise from the pronounced tolerance to intrinsic defects 28,29 and lattice softness allowing large polarons and efficient screening of the carriers. 30−33 Tin(II)-and germanium(II)-based AMX 3 halide perovskites [A = Cs, formamidinium (FA), or methylammonium (MA)] have always been envisioned as the closest alternatives to LHPs. The metal−halide−metal (M−X−M) angle of about 180°ensures the highest σ overlap of the orbitals and the efficient dispersion of the resulting electronic bands. Tilting the octahedra or alternating cations in other halide perovskitoids reduces orbital overlap, opens the band gap, and leads to heavier carriers. 34,35 Metal halides with edge-or face-sharing octahedra, or even with fully isolated ones (0D-halides), exhibit properties that are notably different from the prototypic LHPs. Although these properties are promising for some applications, 36−47 such materials are unlikely to deliver the photovoltaic performance or narrowband excitonic luminescence similar to LHPs.
The main factor limiting the performance of Sn and Ge halide perovskites is their low stability toward oxidation, which by a number of pathways can lead to degenerate p-type conductivity. 48 Nevertheless, considerable progress has been achieved in FASnI 3 -based photovoltaics over the past seven years with a 7-fold increase of the power conversion efficiency (PCE), with a benchmark now of 14.81% ( Figure S1). 49,50 Synthetic approaches can be grouped as follows: (i) reduction of the present Sn(IV) impurities by comproportionation with metallic Sn(0) 51 In contrast, the synthesis of tin halide perovskite nanocrystals (NCs) remains scarce in the literature and is mainly limited to CsSnX 3 or mixed APb y Sn 1−y X 3 compositions. 61−69 L. Dai et al. have recently reported the protocol allowing the synthesis of FASnI 3 NCs with good morphological quality. 70 However, their work focuses on the hot-carrier relaxation processes in FASnI 3 NCs and does not discuss the effect of the present impurities on the optical properties of NCs, which we find to be crucial.
We thus sought to develop a colloidal synthesis of FASnI 3 NCs with high morphological quality and purity and study their intrinsic optical properties and structure, as reported herein. We devote special care to the purity of all of the involved precursors to exclude possible doping during the synthesis. We show that pure FASnI 3 NCs exhibit a disorder of the I-sites that reduces the Sn−I−Sn angle to 167°and opens the band gap by ∼190 meV. Our findings are supported by ab initio calculations for the average cubic structure of FASnI 3 showing that the octahedra exhibit a significant tilting that profoundly affects the electronic structure. Such distortion also allows for unusually pronounced photoinduced states. Based on these findings, we propose an additional pathway for tuning the optical properties of FASnI 3 NCs via altering their lattice with bifunctional organic cations.
In the footstep of our earlier reports on the synthesis of FAPbI 3 NCs, the first edition of FASnI 3 NCs involved 1octadecene (ODE) as a solvent ( Figure S2). Briefly, oleylamine (OAm), oleic acid (OA), and formamidinium oleate are sequentially injected into a hot SnI 2 solution in trioctylphosphine (TOP) and ODE. However, we have found two rarely considered important synthetic parameters to be crucial for the synthesis of Sn(IV)-free FASnI 3 NCs (see Supporting Information Note S1 for details). First, all tested commercially available sources of SnI 2 are of >99% metal basis purity but contain significant amounts of SnO, SnO 2 , and SnI 4 that notably affect the synthesis. Therefore, we have developed a two-step purification procedure that results in SnI 2 with an insignificant amount of Sn(IV) species, as evidenced by X-ray photoelectron spectroscopy (XPS) and X-ray diffraction (XRD). Second, reactions performed at 80°C in ODE nearly instantly lead to the oxidation of iodide anions to polyiodide ones, apparent from their characteristic red-brown color. Note that photoluminescence (PL) spectra of NCs synthesized in ODE do not correlate with their size, and the PL maximum fluctuates in a broad range from 700 to 850 nm ( Figure S3). Therefore, we opt for a synthesis in aromatic solvents (toluene, mesitylene, or Dowtherm A) that proceeds without color change until formamidinium oleate solution is injected and NCs start to grow ( Figures S4 and S5).
The optimized synthesis proceeds by solubilizing SnI 2 in Dowtherm A in the presence of oleylamine and TOP, followed by the hot injection of formamidinium oleate dissolved in Dowtherm A (see the SI for the details). The stability of asprepared NCs is sufficient to wash them by precipitation with acetonitrile, if washing is performed air-free. This protocol allows the removal of all byproducts and impurities of layered perovskites with PL around 700 nm, which is always present in the crude solution (Figures 1a−c).
XPS combined with inert-gas transfer was used to collect the Sn 3d core level spectra. 71 The acquired spectra for Sn 3d 5/2 and 3d 3/2 do not show apparent asymmetry and can be wellfitted with a single component, indicating Sn is mainly present in one oxidation state (Figure 1d). FASnI 3 samples made from purified and nonpurified SnI 2 showed great similarity ( Figure  S6). However, due to the high proximity of peaks corresponding to Sn 2+ and Sn 4+ and inaccuracies related to the referencing of the binding energy scale as well as charging effects, the spectrum analysis can sometimes be ambiguous, leading to possibly wrong assumptions about the oxidation state. This is especially the case with low Sn 2+ or Sn 4+ contents. The modified Auger parameter, α′ (AP), 72 allows one to cancel out shifts related to charging and band bending effects and was shown to be useful for resolving the oxidation state and chemical environment of Sn in perovskite-type structures, including polycrystalline FASnI 3 films. 71 The value of α′ corresponds to the sum of the binding energy (BE) of the Sn 3d 5 / 2 photoelectron line and the kinetic energy (KE) of the corresponding Sn M 4 N 4,5 N 4,5 Auger line. Any shift in α′ can be directly related to a change in the local electronic polarizability, 72 which is extremely sensitive to changes in the local chemical state of the atom in the compound. 73,74 Shifts in α′ can be visualized in a so-called chemical-state (or Wagner) plot, where constant α′ values lie on diagonal lines ( Figure 1e). Strikingly, the α′ value for nonpurified SnI 2 (919.1 eV) is highly similar to that reported for SnO 2 , indicating its presence and potentially that of other Sn(IV) compounds on the powder surface. The purified SnI 2 , in contrast, exhibits a higher α′ value (922.2 eV), similar to other Sn(II) compounds. Likewise, the α′ values of the FASnI 3 NC films made from purified and nonpurified SnI 2 significantly differ (922.3 eV vs 921.6 eV), with the first value matching precisely the one reported for polycrystalline FASnI 3 films. 71 These data demonstrate the enhanced sensitivity of the AP analysis for detecting minuscule differences in the local chemical state of Sn. In agreement with XPS data, we have not observed any signs of the oxidized FA 2 SnI 6 form in 119 Sn solid-state NMR, which is expected at −4818 ppm. 75 We also did not observe the effect of a number of reducing agents on the optical properties of FASnI 3 NCs, as summarized in Supporting Information Note S2. These findings suggest that the amount of Sn(IV) impurities in the synthesized FASnI 3 NCs is The core region of the FASnI 3 NCs can be more effectively probed by hard X-ray photoelectron spectroscopy (HAXPES) analysis using a hard Cr Kα X-ray source, resulting in an approximate increase of the information depth by a factor of 3. 76 This comparison helps us to study the distribution of N in the NC core and on the surface in the form of formamidinium (N FA + ) and oleylammonium (N OAm + ), respectively ( Figure 1f). HAXPES reveals that the total N OAm + /N FA + ratio (1:1) is lowered compared to that of the XPS analysis (3.6:1), which probes the surface of NCs preferentially. This finding indicates that NCs in solution have a significant amount of unbound OAmI ligands, in agreement with the analogous CsPbBr 3 NCs and ICP-MS data showing a strong excess of I over Sn. 77 Pure colloidal FASnI 3 NCs exhibit absorption edge and PL in the range 770−830 nm with PL quantum yield (QY) of 0.1%. (Figure 2a). Time-resolved emission spectra (TRES) are found to be wavelength-independent, confirming the uniformity of the NCs (Figures 2b, Figure S7). PL decay is biexponential with 0.3 ns (90%) and 1.9 ns (10%) lifetimes, in agreement with reported values for bulk FASnI 3 . 78 The excitonic absorption peak of pure FASnI 3 NCs is unresolved, similar to the previously reported spectra of other tin halide perovskite NCs. 62,65 We note that the absorption tail lasting to ∼800 nm belongs to NCs and not to the scattering, as ensured by filtering the solution through a 0.2 μm PTFE filter and evidenced by PL upon excitation at 635 nm ( Figure S8).
Absorption edge and PL of the synthesized FASnI 3 NCs are notably (by 190 meV) shifted from the band gap of bulk FASnI 3 (Figure 2a). Although the increase of the band gap is expected for colloidal semiconductor NCs, such a big shift can hardly be explained solely by quantum confinement. For example, the confinement energy in FAPbI 3 and CsPbI 3 NCs of similar sizes is 89−93 meV. 17,79,80 These NCs are about 10 nm large, whereas the exciton Bohr radius in FASnI 3 should be 3.5−4.4 nm, depending on the set of effective masses reported for bulk material, 81,82 indicating that these NCs should be in a weak confinement regime.
Absorption spectra of the pure FASnI 3 NCs do not exhibit resolved excitonic transitions despite the narrow size dispersion (8−12%). Furthermore, these NCs exhibit rather low intrinsic absorption of about 4 ± 1.7 × 10 3 cm −1 (determined 100 meV above the band gap by the ICP-MS with a combination of acidic and basic digestions for Sn and I, respectively; see Supporting Information Note S3 for details). That is about 4 times lower than the absorption coefficient reported for bulk FASnI 3 . 83 Figure 2d shows spectrally resolved transient absorption (TA) data of FASnI 3 NCs pumped at 400 nm, well above the band gap. This spectral map reveals a pronounced bleach around 600, ascribed to the VB1-CB2 transition (Figure 2c, see Supporting Information Note S4 for details). 84,85 The net spin relaxation kinetics, evaluated as the difference between co-and cross-polarized TA signals, is fitted using a single-exponential decay and yields a lifetime of 2 ps, on par with the one of LHP

Nano Letters
pubs.acs.org/NanoLett Letter NCs (∼1−3 ps), which is much shorter than in the case of CsSnBr 3 NCs (Figures 2e,f). 85 In addition to the expected bleaches, Figure 5d reveals a pronounced photoinduced absorption (PIA) band below the B2 bleach. Depending on the experimental conditions, this PIA band occurs on different time scales and at slightly different energies ( Figure S9). In the case of resonance pumping, the PIA band is polarization-sensitive, appears at ∼0.2 ps, has a maximum at 703 nm, correlates with depopulation of B2, and can be explained as a biexciton shift. 76,85 In the case of pumping at 400 nm, in contrast, PIA appears before the thermalization of carriers to the B2 state, has a maximum around 670 nm, and is also significantly more intense. This type of PIA is nearly absent with resonance pumping. We suggest that this PIA can arise due to the additional photoinduced states created upon lattice distortion induced by hot carriers. Fast PIA to these states becomes possible until the hot carrier is thermalized, and lattice distortion is released. Note that such strong PIA was not observed for CsPbX 3 . 76 In contrast, electron localization has been recently predicted to be energetically favorable in many tin halide perovskites, with bipolaronic states being the most stable form of self-trapped electrons. 86 We hypothesized that the unusually large band gap, distinct PIA, as well as weak and unresolved absorption transitions in FASnI 3 NCs may be related to the reduced lattice symmetry, as evidenced below by wide-angle X-ray total scattering (WAXTS) and solid-state 119 Sn NMR data.
FASnI 3 NCs exhibit a cubic crystal structure and Pm3̅ m space group symmetry (model parameters are reported in Table S1), as recently reported for FASnI 3 bulk crystal ( Figure  3a). 87,88 Results from both Rietveld refinement 89 (Figure 3b). Such a small local distortion makes the Sn−I−Sn angle bent (by about 13°), while the Pm3̅ m cubic symmetry is retained. We verify this picture using density functional theory (DFT) calculations on the average cubic but disordered structure of FASnI 3 that yield an average Sn−I−Sn bending of 10°(see computational details in the SI). The presence of this structural disorder is also suggested by the very large isotropic Debye−Waller factors (B) of iodide ions (B(I) > 6 Å 2 ). Improved modeling of WAXTS data is achieved by splitting the iodide positions, which led to a better match of peak intensities (as detailed in the SI) and reduced B(I) values of 2.8 Å 2 . The same model has already been used to represent the structural disorder in FAPbI 3 and FAPbBr 3 NCs. 79, 91 Remarkably, although large atomic displacement parameters of iodine ions (6.3 Å 2 ) are reported for bulk FASnI 3 , 87 there was no indication of iodide displacement, while the archetypal cubic structure was favored (Table S2). Despite the considerably high B(Sn) factors (>4 Å 2 ), WAXTS data for these NCs do not support the local disorder previously reported for bulk samples where Sn is off-centered along the [111] crystallographic direction 92 (not observed in single crystal XRD data of ref 87; see Table S2).
Analogously to WAXTS data, 119 Sn solid-state NMR indicates that the lattice symmetry of FASnI 3 NCs might be reduced compared to bulk FASnI 3 : the chemical shift of FASnI 3 NCs (−116 ppm) is significantly lower compared to bulk FASnI 3 (140 ppm, Figure 3c). Note that, after aging and merging of NCs, the 119 Sn chemical shift (144 ppm) coincides with the bulk material, indicating that the signal at −120 ppm is a signature of individual NCs that are not degraded.
Although the time-averaged structure of FASnI 3 NCs is cubic, the observed type of disorder can increase the band gap. DFT calculations of the electron spectral function of the disordered cubic FASnI 3 (color map) and the band structure of the archetypal cubic FASnI 3 (black dots) show that the electronic structure is modified significantly upon allowing the system to accommodate local distortions (Figure 4). In particular, the band gap increases by ∼223 meV due to the reduced Sn−I−Sn angle. This phenomenon is well-known for metal halide perovskite-like compounds. 94 The DOS around the band edges is consistent with the parabolic band approximation varying with the square root of energy. The peak structures in the conduction band reflect the nearly triply degenerate conduction band minimum at the R high symmetry point. Despite the presence of local distortions, the degeneracy should be maintained since the network reflects, on average, a high-symmetry structure. The DFT calculations also reveal the smearing of the electronic structure, which accounts for the absence of excitonic resonances in the absorption spectrum. Moreover, this smearing induces a difference between the Tauc plots for the "direct" or "indirect band model" (see Figure S10 and discussion therein), while still considering zero-phonon optical transition. These simulations show that the highly disordered average cubic lattice of FASnI 3 may significantly depart from the standard picture of a perfectly ordered directband-gap semiconductor when optoelectronic properties are considered.
The observed lattice disorder and PIA hint that intrinsic FASnI 3 NCs are unlikely to exhibit optical properties resembling those of lead halide perovskite NCs. One way to reduce the remarkable, but in this case undesired, smearing of the electronic structure is to alter the lattice via mild doping. This motivated us to explore the possibility of tuning the optical properties of FASnI 3 NCs by doping them with small A-site or bifunctional cations ( Figure 5, Supporting Information Note S5). Indeed, doping with Cs or ethylenediammonium (up to 5%) leads to the onset of a slightly resolved absorption feature around 700 nm, which is absent in undoped FASnI 3 NCs and smears away at higher levels of doping density of states of FASnI 3 calculated using the disordered structure in a 2 × 2 × 2 supercell and the band structure unfolding technique. 93 Black lines in panel a represent the electron band structure calculated using the archetypal cubic FASnI 3 structure at high-symmetry positions but with the FA molecules in orientations as calculated for the disordered structure. Nano Letters pubs.acs.org/NanoLett Letter (Figure 5b). This trend coincides with a PL peak shift, notable narrowing, and PL QY dependence on the doping level ( Figure  5c). The energy gap between the absorption band edge and PL peak position quickly decreases by 3 times (Figure 5c). Altogether, this indicates that doping may be the right strategy to overcome the observed disorder-induced smearing of the optoelectronic properties in the FASnI 3 NCs.
In conclusion, we have developed a colloidal synthesis of monodisperse and Sn(IV)-free FASnI 3 NCs, which allowed us to probe their intrinsic optical properties. We show that 10 nm large pure colloidal FASnI 3 NCs exhibit an unusually large band gap, which cannot be explained solely by quantum confinement and is instead attributed to the distortion of the lattice through the split of I-sites. Such a split causes a reduction in the symmetry and bending of the Sn−I−Sn bond, reducing the overlap of Sn and I 5p orbitals. Pure FASnI 3 NCs also exhibit a nearly featureless absorption spectrum with an absorption coefficient of only 4 ± 1.7 × 10 3 cm −1 near the band gap, which might be caused by the disorder-induced smearing of the electronic structure. We also show that doping FASnI 3 NCs with bifunctional organic cations, such as ethylenediammonium, might be an efficient way to tune the optical properties of these NCs.