III. RESULTS AND DISCUSSION

The crystalline structure of the MA–Sn–I systems was investigated by powder X-ray diffraction (XRD) analysis. The diffraction patterns of the “as-prepared” MASn _x I₃ systems (Fig. 1) revealed CH₃NH₃SnI₃, with a tetragonal crystalline structure (space group I4cm), and monoclinic SnI₂ [indicated with “o” in Fig. 1(a)] as the major crystalline phases. By observing the XRD patterns, there are some important points to note. First, the crystalline reflections of the MASnI₃ tetragonal phase are shifted to higher diffraction angles compared to those reported in the literature. This can be explained by the presence and interactions of water molecules with the crystalline structure of the MASnI₃ phase through hydrogen bonds with iodine ions coordinated to the Sn-ions. The shift of the crystalline reflections has been recently reported by Dang ^{Reference Dang, Zhou, Liu, Ju, Xia, Xia and Tao18} during the investigation of MASnI₃ single crystals after exposure to air for a month and supported by the elemental analysis (CHN-Analysis) performed on the as-prepared CH₃NH₃Sn _x I₃ powders for x = 0.9, x = 1.0, and x = 1.2, respectively, (summarized in Table I). A description of the CHN-analysis and procedure is reported in the Supplementary Material. Indeed, the experimental values of carbon, nitrogen, and hydrogen (in wt%) would confirm the presence of the methylammonium group in the as-prepared materials independent of the Sn-content in the starting materials. The higher carbon and hydrogen contents can be explained by the absorption from species in the environment (moisture and carbon dioxide). Second, the intensity of the crystalline reflections of the MASnI₃ phase increases at increasing the Sn-content (till x = 1.2); at higher Sn-content (x = 1.4), a broadening of the crystalline peaks has been detected. This can be explained by the higher Sn-content into the material. Tin sometimes plays the role of a glass former and sometimes that of a network modifier. ^{Reference Bartenev, Suzdalev and Tsyganov19} In noncrystalline oxide materials (or “glasses”), constituents can be either network formers or intermediates or network modifiers. The network formers (e.g., Si, B, and Ge) form a highly cross-linked network of chemical bonds. The intermediates (e.g., Ti, Al, Zr, Be, Mg, and Zn) can act as both network formers and modifiers, according to the glass composition. The network modifiers (e.g., Ca, Pb, Li, Na, K, and P₂O₅) alter the network structure; they are usually present as ions, compensated by nearby nonbridging oxygen atoms, bound by one covalent bond to the glass network and holding one negative charge to compensate for the positive ion nearby. On the other hand, no effect of the Sn-content on the morphology of the as-prepared materials has been observed: the crystallites show an irregular polyhedral morphology, with a length ranging between 50 and 400 μm.

FIG. 1.

XRD patterns of CH₃NH₃Sn _x I₃ systems for different values of x, (a) as-prepared, (b) after annealing in nitrogen at T = 150 °C, t = 8 h. The monoclinic SnI₂ phase has been indicated with “o”.

TABLE I.

Elemental analysis of CH₃NH₃Sn _x I₃ systems for different values of “x”. The experimental values were obtained as the average of two in-a-row measurements.

^a Ref. Reference Dang, Zhou, Liu, Ju, Xia, Xia and Tao18.

The thermal properties of the as-prepared CH₃NH₃Sn _x I₃ systems were investigated by dynamic differential thermal analysis (DTA) and dynamic thermogravimetry (TG) in the range between 30 and 600 °C (heating rate of 10 K/min) under flowing nitrogen and synthetic air. A detailed discussion on the results of the thermal analysis has been previously reported. ^{Reference Dimesso, Fasel, Lakus-Wollny, Mayer and Jaegermann20} We found out that under nitrogen, the thermal event(s) above T = 200 °C are related to the decomposition of the perovskitic systems to solid Sn(II) iodide and gaseous methylamine and hydrogen iodide followed by the melting of the solid phase according to the processes

as reported by Mitzi ^{Reference Mitzi21} and recently confirmed by Brunetti ^{Reference Brunetti, Cavallo, Ciccioli, Gigli and Latini22} for the Pb-containing analogues. In our case, the processes (1) and (2) are confirmed by the absence of a “plateau” in the TG-curve. On the other hand, we observed a shift of the thermal events and consequently of the processes (1) and (2) toward lower temperatures in air. A possible explanation is related to the different heat transfer and thermal conductivities of the analyzed materials in air (air is a worse thermal insulator than nitrogen), and the formation of the Sn²⁺–O and Sn⁴⁺–O species according to the reaction 2SnO(s) + O₂(g) → 2SnO₂(s) was also considered. Interestingly, the thermal profiles in nitrogen and air showed an endothermic event peaking at T = 140 °C, which is not related to any weight loss in the thermogravimetric profiles. To provide more information on the reactions associated to this small thermal event, the as-prepared powders were thermally treated at T = 150 °C for t = 8 h under flowing nitrogen and flowing synthetic air. The diffraction patterns of the MASn _x I₃ systems annealed in nitrogen are shown in Fig. 1(b). At room temperature, the hybrid compounds CH₃NH₃Sn _x I₃ adopted a nearly cubic perovskite structure (space group P4mm) ^{Reference Stoumpos, Malliakas and Kanatzidis23} ; this structure is constructed of a network of corner sharing SnX₆ octahedra that encompass an organic monovalent cation, [CH₃NH₃]⁺. The organic monovalent cation is highly disordered and free to rotate at high temperature because the thermal energy surpasses the energy barriers between different stable configurations, ^{Reference Noel, Stranks, Abate, Wehrenfennig, Guarnera, Haghighirad, Sadhanala, Eperon, Pathak, Johnston, Petrozza, Herza and Snaith24} which can indicate that the organic cation does not form a strong bond with the halide atoms. After annealing in air, the XRD patterns revealed the presence of very broad reflections (Fig. S1, x = 0.9 and x = 1.0, respectively), which can be explained by the growth of noncrystalline phases related to a possible degradation reaction of the perovskites. Indeed, the reaction of oxygen atoms with the Sn-ions of the perovskite through defects and dislocations in the crystalline structure of the MASnI₃ phase can lead to the formation of Sn(II)–O [and possibly Sn(IV)–O] binds ^{Reference Dimesso, Das, Stoehr, Mayer and Jaegermann15} and consequently to the formation of noncrystalline phases due to the nature of Sn-oxides as network modifiers.

Typical SEM images of the MASn _x I₃ systems are shown in Figs. 2(a) and 2(b). For x = 0.9, the crystallites show irregular polyhedral morphology, with a length ranging between 2 and 4 μm [Fig. 2(a)], at higher Sn-content (x = 1.2), a “melt-like” morphology with irregularly shaped crystallites can be observed. The morphological analysis supports the XRD and DTA data on a phase transformation without heat exchange during annealing before decomposition of the material. The phase transformation occurs independent of the initial Sn-content in the material which can indicate an “active contribution” of the SnI₂ species in the mechanism. A similar behavior has been observed by Ogomi ^{Reference Conings, Drijkoningen, Gauquelin, Babayigit, D’Haen, D’Olieslaeger, Ethirajan, Verbeeck, Manca, Mosconi, De Angelis and Boyen13} and more recently by Ju ^{Reference Ju, Sun, Zhaob and Liang25} during the investigation of MAPb _x Sn_1−x I₃ systems, particularly for x ≈ 0.5 by computational calculations. The calculated results reveal that the band gaps of MAPb _x Sn_1−x I₃ systems are sensitive to the Sn/Pb ratio and are proportional to the x component, consistent with the previous reports. The key role of I-containing species is supported by our results. The CHN-analysis of the annealed samples (reported in Table I) revealed experimental values of carbon, nitrogen, and hydrogen very close to the theoretical ones which would indicate that the change of the crystalline structure of the MASnI₃ phase (at T ≈ 140 °C) took place without any lost of methylammonium ions. The oxidation state of the elements was determined by XPS investigations on the MASI systems after annealing under nitrogen. The XPS survey spectra (not reported in this work) detected the presence of carbon, nitrogen, tin, iodine, and oxygen. The presence of oxygen can be explained by the adsorption of impurities in the atmosphere due to grinding of the material as well as due to the ex situ XPS measurements. The fit of the Sn3d and I3d core level spectra was performed using the Thermo Scientific™ Avantage Software (Thermo Fisher Scientific, Inc., Waltham, Massachusetts) for surface analysis. The background was corrected using “smart background fitting”, and the peaks were fitted by a combination of Gaussian and Lorentzian functions. The Sn3d _5/2 core level XPS profile showed a main peak position at 487.5 eV, attributed to the formation of the Sn–I bond, and a small peak at lower binding energy (485 eV), possibly related to the formation of metallic Sn(0). ^{Reference Liu, Fan, Li, Zhang and Mai14}

FIG. 2.

(a) Typical SEM-images of cubic CH₃NH₃Sn _x I₃ systems for x = 0.9 and x = 1.2, respectively; (b) Sn3d and I3d photoelectron emission spectra of the cubic-MASn_0.9I₃ system.

The thermal stability of the cubic CH₃NH₃Sn _x I₃ systems was investigated by dynamic DTA-TG measurements in the temperature range between 30 and 500 °C (heating rate of 10 K/min) under flowing nitrogen and synthetic air. The results of the thermal investigation, for x = 1.0, are previously reported. ^{Reference Dimesso, Fasel, Lakus-Wollny, Mayer and Jaegermann20} Under nitrogen, the DTA profile was related to (i) the decomposition of the perovskite system with formation of SnI₂ as solid phase and (ii) the melting of the SnI₂; this fact is supported by the weight losses recorded in the TG profile (and first derivative) in which weight losses of Δw ≈ −0.13% at T = 85 °C and Δw ≈ −0.90% at T = 150 °C were determined. On the other hand, in air, the DTA profile of MASn _x I₃ systems showed a shift toward lower temperatures of the endothermic events related to the decomposition of the MASnI₃ phase and melting of solid SnI₂. A deeper analysis of the thermal features revealed the presence of additional thermal events which were explained by the formation of Sn(IV)–O binds from Sn(IV)–I compound(s) in the vapor phase and the decomposition of methylamine into CO₂, H₂O, and N₂. ^{Reference Li and Oshima26} Indeed, this is supported by the results of the dynamic thermogravimetric profile for x = 1.0 which recorded in air a weight loss of 0.35 wt% at t = 85 °C and 0.51 wt% at T = 150 °C, respectively, and of the elemental analysis (Table I) on the samples annealed in air which revealed a lower amount of carbon and nitrogen compared to the theoretical values, whereas the excess of hydrogen can be explained by the formation of tin oxide hydrates (SnO₂·nH₂O) in contact with moisture of the ambient atmosphere.

The weight loss mechanism and the influence of the starting Sn-content on this effect, MASn _x I₃ systems (for x = 0.9 and x = 1.0), were investigated by static TG-analysis in nitrogen and air. The samples were heated up to T = 85 °C and to T = 150 °C, respectively, and the weight loss was measured in a time range t = 2 h. The weight losses as a function of time at constant temperature are shown in Fig. 3 in nitrogen atmosphere and in Fig. 4 in air, respectively. The loss rates (wt% per minute) have been calculated by linear fit of the experimental data; the corresponding fit parameters are summarized in Table SI. Under nitrogen, weight loss rates of −0.0022 wt%/min (T = 85 °C) and −0.048 wt%/min (T = 150 °C) for the MASn_0.9I₃ phase [Fig. 3(a)] and −0.0039 wt%/min (T = 85 °C) and −0.0693 wt%/min (T = 150 °C) for the MASn_1.0I₃ system [Fig. 3(b)] were estimated, respectively. From the obtained results, we suggest that the weight losses can be related to the iodine (-containing) species and may be related to a change of the stoichiometric ratios among cations and iodine anions in the perovskite active material with the time. This fact would be supported by the CHN-analysis (Table I) on the samples annealed under nitrogen which showed a small difference of the carbon and nitrogen contents compared to the untreated samples (as-prepared). This would indicate that the methylammonium ions are not involved in the weight losses of the Sn-containing perovskite. We speculate that the loss of iodine could “create the conditions” for the disproportion of the Sn(II)-ions, and this would provide an explanation of the poor PCE performance of this system reported in the literature. ^{Reference Conings, Drijkoningen, Gauquelin, Babayigit, D’Haen, D’Olieslaeger, Ethirajan, Verbeeck, Manca, Mosconi, De Angelis and Boyen13}

FIG. 3.

Isothermal curves of the CH₃NH₃Sn _x I₃ systems at T = 85 °C and T = 150 °C, t = 2 h under nitrogen for (a) x = 0.9 and (b) x = 1.0, respectively. The black line is a linear fit of the data originated by the program OriginPro 8. The corresponding fit parameters are reported in Table SI.

FIG. 4.

Isothermal curves of the CH₃NH₃Sn _x I₃ systems at T = 85 °C and T = 150 °C, t = 2 h in air for (a) x = 0.9 and (b) for x = 1.0, respectively. The black line is a linear fit of the data originated by the program OriginPro 8. The corresponding fit parameters are reported in Table SI.

On the other hand, static TG measurements performed in air revealed higher weight loss rates (Table SI) which confirm a more negative effect of the oxygen ions on the perovskite material. For x = 0.9, the weight loss rates were −0.0062 wt%/min at T = 85 °C and −0.2463 wt%/min at T = 150 °C, [Fig. 4(a)] which are 3 times and 5 times higher than the corresponding rates in nitrogen. For x = 1.0, loss rates of −0.0098 wt%/min (T = 85 °C) and −0.2885 wt%/min (T = 150 °C) were estimated [Fig. 4(b)] which are 2.5 times, 4 times higher than those observed under nitrogen and higher than the loss rates observed in the MASn_0.9I₃ system. These data can be explained by using oxygen atoms as a substitute for iodine anions in the perovskite structure of the MA–Sn–I systems, which leads to the oxidation fo Sn(II) into Sn(IV) as well as to the formation of Sn(II)–O and Sn(IV)–O binds as observed previously in the Cs- and Cs–Pb-containing systems. ^{Reference Liu, Fan, Li, Zhang and Mai14,Reference Dimesso, Das, Stoehr, Mayer and Jaegermann15}

The differences in the structural and morphological features in dependence on the Sn-content can have a dramatic influence on the optical properties of the prepared materials. In Fig. 5, the optical absorption spectra [Figs. 5(a) and 5(b)] and the Tauc’s plots [Figs. 5(c) and 5(d)] of the cubic-MASn _x I₃ systems are shown. Assuming a bulk-like behavior, the adsorption edges [λ_on-set (nm), the frequency values at which the photons absorption begins] can be obtained by fitting the linear part of the absorbance spectrum [Fig. 5(b)]. ^{Reference Kojima, Teshima, Shirai and Miyasaka2} The values of the adsorption edges are reported in Table SII. The adsorption spectra of the MASn _x I₃ systems showed a dramatic dependence of the optical properties on the starting Sn-content. Indeed, in the 0.9 ≤ x ≤ 1.1 composition range, the optical spectra display a strong, abrupt absorption in the infrared region with absorbance edges at 1107.0 nm for x = 0.9, at 1098.6 nm for x = 1.0 and at 1073.2 nm for x = 1.1, respectively. Furthermore, as shown in Fig. 5(a), the pregap region does not show strong absorption tails indicating the high quality of the prepared materials, whereas at a higher Sn-content (x ≥ 1.2), a broad tail of the absorbance profile can be observed (Fig. S2). As nano-particles in powders tend to agglomerate and consequently develop bulk-like bands, the relationship between the adsorption coefficient (α) near the absorption edge and the optical band gap (E _g) for direct interband transitions obeys a relationship as previously reported [Fig. 5(c)]. ^{Reference Dimesso, Dimamay, Hamburger and Jaegermann5,Reference Dimesso, Kim and Jaegermann6} By extrapolating to α = 0 [Fig. 5(d)], the plots revealed a direct semiconducting behavior with band energy gaps of 1.18–1.19 eV. The optical band gap values are lower than 1.3 eV which is commonly reported in the literature and very close to the bulk silicon band gap (E ₀ ≈ 1.11 eV at 300 K). ^{Reference Kittel27} This behavior can be related with Sn cation vacancies which are the most common defects in the Sn-containing systems ^{Reference Needleman9} and possibly with the iodine stoichiometry in our annealed samples.

FIG. 5.

(a) Normalized absorption spectra and (b) magnification of (a) with fit of the linear part of the absorbance spectra; (c) plots (αhν)² for direct transitions, where α is absorption coefficient and hν is the photon energy, of the cubic-MASn _x I₃ systems; and (d) band gaps E _g are obtained by extrapolation to α = 0. The values are reported in Table SII.

The PL emission properties of the MA–Sn–I systems were investigated at room temperature using an excitation wave length of λ_exc = 500 nm. The PL emission spectra, in the wave length range between 950 and 1100 nm, of the cubic-MASn _x I₃ systems are shown in Fig. 6(a), and the corresponding PL-peak values are reported in Table SII. By observing Fig. 6(a) and the data in Table SII values, some points are to be noted. A dramatic influence of the initial Sn-content on the PL emission spectra of the MASn _x I₃ system has been observed. Indeed, for x = 0.9, a band peaking at ∼1064 nm with a broad shoulder at ∼1078 nm was detected, and an additional band peaking at ∼1014 nm [Fig. 6(a) and Table SII] was observed at a higher Sn-content (for x ≥ 1.0). Moreover, additional PL-emission bands peaking at ∼1057 nm, ∼1084 nm, respectively, with shoulders at ∼1035–1038 and ∼1070 nm, appeared in the 1.0 ≤ x ≤ 1.2 range. Finally, for x = 1.4, further additional bands peaking at ∼1040 and ∼1068 nm, respectively, were observed [Fig. 6(a)]. The wavelengths of the PL-peak maxima are higher than those reported in the literature ^{Reference Noel, Stranks, Abate, Wehrenfennig, Guarnera, Haghighirad, Sadhanala, Eperon, Pathak, Johnston, Petrozza, Herza and Snaith24,Reference Hao, Stoumpos, Cao, Chang and Kanatzidis28} at room temperature. Hao ^{Reference Hao, Stoumpos, Cao, Chang and Kanatzidis28} reported a PL-emission peak at 950 nm for the MASnI₃ phase at room temperature, and Noel ^{Reference Noel, Stranks, Abate, Wehrenfennig, Guarnera, Haghighirad, Sadhanala, Eperon, Pathak, Johnston, Petrozza, Herza and Snaith24} reported a broad adsorption edge at 1000 nm and a PL emission peaked at ∼980 nm with a photoexcitation of 500 nm.

FIG. 6.

(a) PL-emission spectra (RT, λ_exc = 500 nm) in the 950–1100 nm range of the cubic- MASn _x I₃ systems. The corresponding peak values are reported in Table SII; (b) PL-emission spectra (RT, λ_exc = 500 nm) in the range 650–1100 nm for x = 0.9, x = 1.2, and x = 1.4, respectively.

To avoid any misunderstanding and misleading information, we performed a PL-emission survey (at RT and λ_exc = 500 nm) in the 650–1100 nm of the cubic-MASn _x I₃ phase for different values of “x” [Fig. 6(b)]. For x = 0.9, the PL-spectrum consists of bands peaked at ∼675 nm and above 1000 nm. On the other hand, at a higher Sn-content (x = 1.2 and x = 1.4), a shift of the PL-emission peak to ∼681 nm was detected. This emission was observed in SnI₂-doped CdI₂ crystals ^{Reference Novosad and Kovalyuk29} and was attributed to electron–hole recombination on halogen vacancies, followed by energy transfer, via an Auger process, to emission centers involving interstitial Cd. The origin and presence of more bands by increasing the Sn-content can be explained by supposing that the defect formation energy of the Sn-vacancies generating the hole is lower than that of any other defects. This would indicate that a major p-type carrier source is Sn-vacancies, and the black compound is a nearly intrinsic p-type semiconductor. The facile generation of Sn vacancies would be even better activated by the heat energy, resulting in continuously raising the number of allowed spectroscopic transitions. ^{Reference Liu, Fan, Li, Zhang and Mai14} Furthermore, monoclinic SnI₂ in single crystals contains two types of sites ^{Reference Howie, Moser and Trevena30} : octahedral (as in SnTe and in cubic CsSnBr₃) and hexagonal (as in PbCl₂). Since the SnI₂ structure contains these two inequivalent types of cation sites, we can reasonably suppose that there are at least two types of luminescence centers (i.e., defects and impurities or self-trapped excitons) in the MA–Sn–I systems, which can be attributed to these different coordination sites.

The next step of our investigation was the preparation of the layers using the prepared powders as precursors. Detailed descriptions of the device fabrication and characterization methods are reported in the Supplementary Material. From our previous experiments on the preparation of solar cells by dissolution of the prepared powders in an appropriate solvent, ^{Reference Dimesso, Kim and Jaegermann6} we concluded that this process favors the recrystallization of the light harvesting materials on the substrates. After dissolution and recrystallization of the MASnI₃ powders as the light harvesting material, hole-transporting layers (spiro-OMeTAD) were deposited by spin-coating. Finally gold electrodes were thermally evaporated under vacuum. Typical current–voltage (J–V) characteristics measured under simulated sunlight (AM1.5, 100 mW/cm²) illumination, by using spiro-OMeTAD as the hole-conductor are shown in Fig. 7. The J–V curves under illumination (black curves) show a semiconducting behavior which confirms the PL profile observed during our measurements. These data can be explained by supposing a “quenching” process which takes place due to the loss of iodine (-containing) species inducing the formation of “excited dimers” (or “excimers”).

FIG. 7.

Example of J–V characteristics measured under simulated sunlight (AM1.5, 100 mW/cm²) illumination for MASn _x I₃ solar cells. The black curves are obtained under illumination of the solar cell whereas the red curves are obtained without illumination of the device.

Although the low PCEs value the Sn-containing hybrid systems, this study opens up opportunities to discover visible and IR light-emitting materials at room temperature, which is highly desirable in achieving breakthrough applications in photovoltaics, various radiation detectors, and light-emitting diodes because the perovskite structure can incorporate a very broad range of elements. However, the poor stability of the Sn-containing system still represents a challenging factor for large scale applications, and more efforts should be dedicated to the stabilization of this system to be commercially desirable.