Polymorphic Control of Solution-Processed Cu2SnS3 Films with Thiol-Amine Ink Formulation.

There is increasing demand for tailored molecular inks that produce phase-pure solution-processed semiconductor films. Within the Cu-Sn-S phase space, Cu2SnS3 belongs to the I2-IV-VI3 class of semiconductors that crystallizes in several different polymorphs. We report the ability of thiol-amine solvent mixtures to dissolve inexpensive bulk Cu2S and SnO precursors to generate free-flowing molecular inks. Upon mild annealing, polymorphic control over phase-pure tetragonal (I4̅2m) and orthorhombic (Cmc21) Cu2SnS3 films was realized simply by switching the identity of the thiol (i.e., 1,2-ethanedithiol vs 2-mercaptoethanol, respectively). Polymorph control is dictated by differences in the resulting molecular metal-thiolate complexes and their subsequent decomposition profiles, which likely seed distinct Cu2-x S phases that template the ternary sulfide sublattice. The p-type tetragonal and orthorhombic Cu2SnS3 films possess similar experimental direct optical band gaps of 0.94 and 0.88 eV, respectively, and strong photoelectrochemical current responses. Understanding how ink formulation dictates polymorph choice should inform the development of other thiol-amine inks for solution-processed films.

Introduction

Ternary Cu2SnS3 is a potentially more earth-abundant replacement for current direct band gap semiconductors, such as CdTe and Cu(In,Ga)(S,Se)2 (CIGS), while still possessing favorable optoelectronic properties.[1−3] Cu2SnS3 belongs to the class of I2–IV–VI3 semiconductors that possess high optical absorption coefficients (>105 cm–1), p-type conductivity, and tunable direct band gaps from 0.9 to 1.8 eV.[3−6] Several different Cu2SnS3 syntheses for applications ranging from solar cells,[7−11] photocatalysts,[12−14] electrocatalysts,[15,16] and thermoelectrics[17,18] have been reported. The I2–IV–VI3 stoichiometry of Cu2SnS3 exists in three different crystal structures on the bulk Cu–Sn–S phase diagram, including monoclinic (Cc), tetragonal (I4̅2m), and cubic (F4̅3m).[19,20] The monoclinic structure is isomorphic to the Cu2SiS3 structure type.[20] Among the disordered structure types, the cubic zinc blende polymorph has F4̅3m symmetry and the tetragonal stannite polymorph (supercell of zinc blende) has I4̅2m symmetry.[20] Cu2SnS3 has also been reported to form a hexagonal wurtzite (63) phase in colloidal nanocrystals, but this purported polymorph does not exist on the bulk Cu–Sn–S bulk phase diagram.[1,3,14,19,21]

The structure of a material directly affects its electronic and optical properties.[22,23] For example, metastable wurtzite-like Cu2ZnSn(S1–Se)4 nanocrystals have wider band gap tunability compared to the thermodynamic kesterite phase in the same compositional range.[24] While colloidal nanocrystal syntheses allow for the formation of metastable phases on bulk phase diagrams, or the isolation of phases previously unknown in bulk,[25] the utilization of these nanocrystals in functional devices is generally hampered by the insulating nature of the organic surface ligands. Methods used to remove ligands for devices involve complex ligand exchanges or thermally decomposing the ligands, which can leave large carbonaceous impurities in the semiconductor layer and at interfaces.[26,27] No direct solution deposition of metastable Cu2SnS3 thin films using molecular inks has been previously reported.

In 2013, our group developed a simple and versatile “alkahest” solvent system consisting of a short chain thiol (e.g., 1,2-ethanedithiol, mercaptoethanol, etc.) and an amine (e.g., 1,2-ethylenediamine) that dissolves over 100 bulk materials, including oxides, chalcogenides, and zero-valent metals. This dissolution process yields molecular inks amenable to solution processing; phase-pure metal chalcogenide thin films can be recovered upon solution processing and mild heating.[28,29] The alkahest solvent system has been leveraged to solution process high-efficiency solar cells,[30,31] electrocatalysts,[32,33] thermoelectrics,[34,35] and a few reports of nanostructured devices, such as tremella-like SnS2 and Sb2Se3 nanowires.[36,37] While compositional control in multinary chalcogenide semiconductors can be achieved by tuning the solute formulation of alkahest-derived inks,[38,39] polymorphic phase control has yet to be demonstrated. Herein, we report phase control over two different Cu2SnS3 structural polymorphs simply by tuning the ink formulation (i.e., thiol choice of 1,2-ethanedithiol or mercaptoethanol). We propose an alternate orthorhombic structure (Cmc21) for the metastable Cu2SnS3 phase, with a hexagonally close-packed S2– sublattice that is isostructural with orthorhombic Ag2GeS3.[40] To the best of our knowledge, this is the first example where the metastable Cu2SnS3 polymorph has been observed outside the context of colloidal nanocrystals. The preparation of the two phase-pure Cu2SnS3 polymorphs using alkahest inks, and their characterization, are discussed in detail. Density functional theory (DFT) calculations were carried out within the PBE & HSE06 level of theory to compare the optoelectronic properties of the thermodynamically stable monoclinic structure to the proposed metastable orthorhombic polymorph of Cu2SnS3.

Experimental Section

General Considerations

All materials were used as received. 1,2-Ethylenediamine (en, 99.5%), 2-mercaptoethanol (merc, 99%), and copper(I) sulfide (Cu2S, 99.99%) were purchased from Sigma-Aldrich. 1,2-Ethanedithiol (EDT, 98+%) and tin(II) oxide (SnO, 99%) were purchased from Alfa Aesar.

Ink Formulations and Processing

To generate the ink for tetragonal Cu2SnS3, 9.95 mg (0.063 mmol) of Cu2S and 8.42 mg (0.063 mmol) of SnO were added to 0.2 mL of EDT and 0.8 mL of en and allowed to stir for 2 h at 30 °C. Full dissolution is observed to occur within minutes. If the EDT/en (1:4 vol/vol) solvent mixture solidifies, gentle heating will resolubilize the solution. To generate the ink for orthorhombic Cu2SnS3, 9.95 mg (0.063 mmol) of Cu2S and 8.42 mg (0.063 mmol) of SnO were added separately to 0.1 mL of merc and 0.4 mL of en and allowed to dissolve for up to 2 h at 30 °C with stirring before finally being combined after full dissolution right before deposition and annealing.

Inks to produce the tetragonal and orthorhombic polymorphs of Cu2SnS3 were drop-cast onto fluorine-doped tin oxide (FTO), Si, or borosilicate glass substrates that were cleaned by sequential sonication in methanol, acetone, and isopropyl alcohol for 15 min each before being blown dry using pressurized nitrogen gas. Before deposition, the substrates were ozone cleaned for 15 min. Finally, 20 μL of ink was deposited onto 1 × 1 cm2 precut substrates, and the tip of the pipette was used to spread the ink across the full substrate without allowing the tip of the syringe to touch the surface of the substrate. Films were annealed identically under flowing nitrogen to 330 °C (ca. 15 °C min–1 ramp rate) and held for 10 min, and then allowed to cool naturally to room temperature.

Organic Content Determination

Thermal gravimetric analysis (TGA) was performed on a TA Instruments TGA Q50 instrument, and samples were run in an alumina crucible under a flowing nitrogen atmosphere with a heating rate of 5 °C min–1. The TGA samples were prepared by drying the ink in an alumina crucible to 100 °C under a flowing nitrogen atmosphere in an aluminum annealing chamber prior to TGA analysis to avoid excessive corrosion of the thermocouple in the TGA. Fourier transform infrared spectroscopy (FT-IR) spectra were collected on an Agilent Cary 630 spectrometer by diamond attenuated total reflection (ATR). The samples were prepared by drop casting the inks onto glass substrates and drying to 100 °C before annealing to 330 °C under a flowing stream of nitrogen and transferring the Cu2SnS3 to the crystal.

Structural and Optical Characterization

Powder X-ray diffraction (XRD) patterns were collected using a Rigaku Ultima IV diffractometer operated at 44 mA and 40 kV, in the 2θ range of 10–70° using Cu Kα radiation (λ = 1.5406 Å). For powder diffraction studies, inks were drop-cast on a glass substrate and dried to 350 °C in an aluminum annealing chamber under flowing nitrogen. The powders were removed from the glass substrate and ground in an agate mortar. For structural refinements, the step size and collection time were 0.01° and 3 s step–1. All patterns were recorded under ambient conditions. Rietveld refinements were carried out using the General Structure Analysis System II (GSAS-2) software package. The following parameters were refined: (1) scale factor, (2) background (modeled using a shifted Chebyshev polynomial function), (3) peak shape, (4) lattice constants, (5) fractional atomic coordinates of the Cu, Sn, and S atoms constrained by the site symmetry, (6) preferred orientation using a spherical harmonic model, and (7) isotropic thermal parameters for each chemical species. The Rwp and χ2 indicators were employed to assess the quality of the refined structural models. Diffuse reflectance UV–vis–NIR spectroscopy was performed on a PerkinElmer Lambda 950 equipped with a 150 mm integrating sphere; 12 mg of powdered Cu2SnS3 sample (tetragonal or orthorhombic polymorph) was mixed with 350 mg of BaO in a mortar and pestle and placed in a solid sample holder. Transmittance UV–vis–NIR spectroscopy was performed by placing a drop-cast film in front of the detector and recording a transmittance spectrum from 300 to 900 nm. Scanning electron microscopy/energy dispersive X-ray spectroscopy (SEM–EDS) was performed using an FEI Helios G4 P-FIB at 20 kV. Top surface micrographs were acquired via SEM using a beam current of 0.8 nA and an accelerating voltage of 5 kV. Raman spectra were conducted on samples deposited on Si substrates annealed to 330 °C. Spectra were recorded for 1 min using an average of three scans using a Horiba XploRA confocal Raman microscope with 532 nm excitation. The Raman microscope was covered with a black tarp to reduce ambient light exposure. X-ray photoelectron spectroscopy (XPS) was performed using a Kratos Axis Ultra X-ray photoelectron spectrometer with a monochromatic aluminum anode (1486.6 eV). An operating current of 5 mA and voltage of 12 kV with a step size of 0.1 eV and a pass energy of 20 eV was used to acquire 20 high-resolution scans for each element. An operating current of 5 mA and voltage of 12 kV with a step size of 1 eV and pass energy of 80 eV was used to acquire five survey scans for each sample. Pressure in the analysis chamber was <1 × 10–9 Torr. XPS was performed on Cu2SnS3 thin films deposited on Si substrates. Inductively coupled plasma–mass spectrometry (ICP-MS) was conducted with a third-party service (Galbraith Laboratories, Knoxville, TN).

Molecular Solute Identification

Electrospray ionization–mass spectrometry (ESI-MS) samples were prepared from fully dissolved solutions of Cu2S (20 mg mL–1, 1:4 (vol/vol) EDT/en or merc/en) diluted with DMSO to ca. 1 μM. Solutions were mixed right before injection into the MS instrument to avoid possible decomposition products. Samples were injected through the main nebulizer using a syringe pump fitted with a 500 μL Hamilton syringe (1750RN) at 5 μL min–1 flow rate. Mass spectra were measured using Agilent 6545 qToF instrument equipped with a dual AJS electrospray ionization source operating in negative ion mode with the following ionization parameters: capillary voltage 3.5 kV, nozzle voltage 0.0 kV, nitrogen was applied as a nebulizer gas 35 psi, sheath gas 12 L min–1, 275 °C, dry gas 10 L min–1, 300 °C, and collision gas. For external calibration and tuning, a low-concentration tuning mix solution by Agilent Technologies was utilized at 10:1 for further dilution. Spectra were recorded in m/z 50–2000 range. All of the mass spectra were recorded at 1 Hz.

Photoelectrochemical Measurements

Photoelectrochemical measurements were performed on FTO substrates (Sigma-Aldrich) with a conductivity of ca. 7 Ω sq–1 that were cut into 0.5 × 2.5 cm2 pieces and cleaned as described above. Kapton tape was used to mask off a portion of the substrate for electrode contact. Photoelectrochemical responses for both polymorphs of Cu2SnS3 were performed using a BASi Epsilon-EC potentiostat. A 3-neck flask was used with a Pt-wire counter electrode and a Pt-wire pseudoreference electrode. The working electrodes were tetragonal and orthorhombic Cu2SnS3 films drop-cast on FTO-coated glass and annealed to 330 °C. The tetragonal Cu2SnS3 electrode was further annealed at 550 °C to improve the rigidity of the working electrode and prevent delamination during measurements. An aqueous 0.1 M Na2S/0.01 M sulfur electrolyte was made from nitrogen-sparged deionized water. For photoelectrochemical experiments, a standard laboratory white light placed ca. 15 cm from the samples was used to illuminate the working electrode. The total illumination areas of the tetragonal and orthorhombic Cu2SnS3 working electrodes were ∼0.75 cm2.

Computational Methodology

Density functional theory (DFT)[41−43] with the projector-augmented-wave (PAW) potentials[44,45] and the Perdew–Burke–Ernzehof (PBE) functional under the generalized gradient approximation[46,47] was employed in geometry optimization calculations. The optimized structures can be seen in Figure S1. Due to the known tendency of the PBE functional to strongly underestimate the band gap because of the self-interaction error, the more accurate and computationally expensive Heyd–Scuseria–Ernzehof (HSE06)[48−51] hybrid functional was used to further evaluate the total energy and electronic structure. All calculations were performed within the Vienna Ab-initio Simulation Package (VASP).[52−55] The PAW PBE versions included in the POTCAR files for each species were PAW_PBE Cu 22Jun2005, PAW_PBE S 06Sep2000, and PAW_PBE Sn 08Apr2002. During the geometry optimization with PBE, we utilized a large plane wave basis energy cutoff (ENCUT) of 520 eV. An 8 × 8 × 8 Γ centered k-point mesh was used to sample the Brillouin zone. Due to the high computation expense of the HSE06 calculations, the ENCUT was lowered to 400 eV. Further, to accurately probe the electronic structure with the HSE06 functional, we set the k-point mesh along its high-symmetry k-paths: Z–G–Y–A–B–D–E–C and Γ–Z–T–Y–S–X–U–R, for the monoclinic and orthorhombic polymorphs, respectively.

Results and Discussion

Ink Formulation and Conversion

To formulate a typical ink to yield tetragonal Cu2SnS3, bulk powders of Cu2S and SnO were mixed in a 1:1 (mol/mol) stoichiometric ratio in EDT/en (1:4 vol/vol) with an overall concentration of ca. 20 mg mL–1. The bulk precursors may be dissolved together or separately to return to the tetragonal phase. Bulk Cu2S and SnO powders both have overall solubility limits of 10–15 wt % in EDT/en (1:4 vol/vol) mixtures under ambient conditions (1 atm, 25 °C). The inks were stirred at 30 °C for 2 h to yield a free-flowing, optically clear orange-brown ink free of scattering. The ink was stable for multiple days under inert atmosphere conditions, as indicated by the lack of color change and solid precipitates. To formulate a typical ink to yield orthorhombic Cu2SnS3, bulk powders of Cu2S and SnO were separately dissolved in a 1:1 (mol/mol) ratio in a mixture of merc/en (1:4 vol/vol) with an overall concentration of ca. 20 mg mL–1 in each ink. The bulk Cu2S and SnO powders have solubility limits of 10–15 and 5–10 wt %, respectively, in merc/en (1:4 vol/vol) mixtures under ambient conditions (1 atm, 25 °C). The SnO ink in merc/en was stirred at 30 °C for 2 h to yield a free-flowing, optically clear, and colorless ink free of scattering. The Cu2S ink in merc/en was a faint light brown color, optically clear, and free of scattering after dissolution under the same conditions. As described above, both Cu2S and SnO inks were stable for multiple days under inert atmosphere conditions. Right before annealing, the two Cu2S and SnO inks in merc/en were mixed to form an almost colorless and optically clear ink. When the bulk precursors were codissolved in merc/en, mixed phase products were recovered upon annealing. While the bulk Cu2S and SnO precursors returned phase-pure tetragonal and orthorhombic phases (vide infra), using different combinations of oxides and sulfides (i.e., CuO, CuS, Cu2O, SnS, SnS2) also returned the tetragonal and orthorhombic polymorphs when using EDT/en and merc/en solvents, respectively. To test the viability of thiol–amine solutions to produce metal chalcogenides at scale, inks made at 10× volumetric scale also produced the same phase-pure tetragonal and orthorhombic phases of Cu2SnS3 when annealed at 330 °C for 10 min.

Thermogravimetric analysis (TGA) was employed to determine the endpoint of organic volatilization and molecular solute decomposition for dried inks from both EDT/en and merc/en solvent combinations (Figure S2). Organic mass loss begins ca. 120 °C and mass loss ends at temperatures <350 °C for both inks. The strongest IR bands of the dried inks belong to the thiol (∼675 cm–1 ν(C–S) stretch, ∼1430 cm–1 δ(CH2) bend);[29,56,57] however, upon annealing the dried inks to 330 °C, a complete loss of IR bands is observed that corroborates the decomposition temperature measured by TGA (Figure S3).

Structural Characterization

The resulting dark gray materials after solution deposition and annealing to 330 °C were confirmed to be phase-pure tetragonal and orthorhombic Cu2SnS3 by powder X-ray diffraction from the EDT/en and merc/en inks, respectively. Rietveld refinements were employed to determine the proper space group for each phase. For the Cu2SnS3 resulting from the EDT/en ink, three phases are possible candidates (i.e., monoclinic, tetragonal, cubic), and their diffraction patterns only differ slightly by minor reflections. Representations of the crystal structures used for refinements are supplied in the Supporting Information. To assess the phase of Cu2SnS3 resulting from the EDT/en ink, powder XRD data sets were refined against these monoclinic (Cc), tetragonal (I4̅2m), and cubic (F4̅3m) crystal structures. In the first refinement, the monoclinic phase failed to converge. In the second refinement, the disordered tetragonal phase led to a refinement with a reduced χ2 of 1.69 and a wR of 2.29% (Figure a,b, Table S1). The Rietveld refinement to the tetragonal structure returned lattice constants of a = 5.4267(6) Å and c = 10.6869(3) Å with a unit cell volume of V = 314.72(8) Å3, which match well to previously reported values for bulk tetragonal Cu2SnS3 (a = 5.41 Å, c = 10.82 Å, V = 316.68 Å).[58] In the third refinement, we simulated a disordered cubic zinc blende structure in which the Cu and Sn atoms are randomly distributed over the Zn site as there are no published zinc blende Cu2SnS3 structures in the ICSD; this refinement returned a slightly worse reduced χ2 of 3.56 and a wR of 3.32% (Figure S4).

Figure 1

(a) Rietveld refinement of the XRD data corresponding to Cu2SnS3 resulting from the EDT/en ink, confirming that the tetragonal I4̅2m unit cell is an appropriate structural model for this polymorph (χ2 1.69, wR 2.29%; a = 5.43 Å, c = 10.69 Å, V = 315.19 Å3). (λ = 1.5406 Å) (b) Structure of disordered tetragonal Cu2SnS3. (c) Rietveld refinement of the XRD data corresponding to Cu2SnS3 resulting from the merc/en ink, confirming that the orthorhombic Cmc21 unit cell is an appropriate structural model for the metastable polymorph (χ2 2.18, wR 4.02%, a = 11.46 Å, b = 6.63 Å, c = 6.32 Å, V = 479.95 Å3). (λ = 1.5406 Å) (d) Structure of ordered orthorhombic Cu2SnS3. Sulfur atoms are yellow, tin atoms are silver, and copper atoms are blue.

The Cu2SnS3 resulting from the merc/en ink yields a distinctly wurtzite-like diffraction pattern. This polymorph has previously only been observed in colloidal nanocrystals and has been assumed to adopt a hexagonal wurtzite (63) structure that does not exist on the Cu–Sn–S bulk phase diagram.[1,3,14,19,21] The disordered wurtzite polymorph derives from the ZnS wurtzite structure, where Cu+ and Sn4+ randomly occupy the Zn site with a 2:1 ratio, respectively. However, there are no known published structures for wurtzite Cu2SnS3. A simulated wurtzite cell with Cu and Sn randomly distributed (2:1 ratio, respectively) on the Zn site, while preserving the sulfur positions, returned a reduced χ2 of 3.05 and a wR of 4.75% (Figure S5). However, tetrahedrally bonded ternary I–III–VI2 sulfides with hexagonal S2– sublattices, such as wurtzite-like CuInS2 and AgInS2, are known to possess cation ordering and not adopt a true disordered wurtzite structure.[59,60] This led us to consider whether there are any known I2–IV–VI3 materials with hexagonal S2– sublattices, which revealed the orthorhombic polymorph (Cmc21) of Ag2GeS3. This structure is like the wurtzite structure type with a hexagonal close-packed S2– sublattice, but with the major distinction being that there is an ordering of the Ag+ and Ge4+ cations. We simulated the orthorhombic structure type by replacing the Ag+ and Ge4+ cations with Cu+ and Sn4+ and allowed the cell and atoms to relax using the GGA PBE exchange-correlation functional with forces converged to 3 meV Å–1. This structure type resulted in an improved refinement with a reduced χ2 of 2.18 and wR of 4.02% (Figure c,d and Table S2). The refinement returned lattice parameters of a = 11.4569(4) Å, b = 6.6268(9) Å, c = 6.3215(7) Å, and V = 479.95(8) Å3, matching quite well to the simulated crystal structure (a = 11.45 Å, b = 6.62 Å, c = 6.34, V = 480.56 Å3). This suggests a potential for cation ordering in our wurtzite-like Cu2SnS3 polymorph due to the fitting of minor reflections and intensity mismatch of the difference curve at 2θ = 47.9°. In our search for an ordered structure, we simulated several other possibilities for an ordered wurtzite-like structure. The simplest ordered cell we considered was the wurtzite ZnS structure expanded to a 1 × 1 × 3 supercell, where three ordering pattern iterations were possible. Expanding the supercell to 1 × 2 × 3 led to 45 ordering iterations. The ordered orthorhombic 1 × 1 × 1 cell led to 77 different ordering iterations with similar minor reflections (only differing in reflection intensities), leading to statistically similar refinements. The resulting orthorhombic Cu2SnS3 films remained in this metastable polymorph for over 4 months at room temperature. “Wurtzite” Cu2SnS3 has been previously reported to undergo a phase transition to the zinc blende polymorph at elevated temperatures (>500 °C).[3] Our orthorhombic polymorph was similarly annealed at 550 °C and experimentally confirmed to relax to either the cubic zinc blende or tetragonal polymorph (Figure S6).

The Rietveld refinements for the tetragonal and orthorhombic Cu2SnS3 polymorphs were not improved with the addition of any binary Cu2–S, SnS, or SnS2 phases, suggesting that both materials are phase pure. Average grain sizes of 11.0 and 68.0 nm for the tetragonal and orthorhombic polymorphs, respectively, were extracted from refinements. Raman spectroscopy data corroborated the phase purity of the solution-processed tetragonal and orthorhombic Cu2SnS3 materials. For the tetragonal polymorph, a Raman active mode at 324 cm–1 matches well with spectra previously reported for tetragonal Cu2SnS3 (Figure S7a).[61] The orthorhombic polymorph has Raman active modes at 291 and 314 cm–1 that match well to previously reported “wurtzite” Cu2SnS3 (Figure S7b).[3] Neither have Raman active modes from potential binary impurities, such as Cu2–S, SnS, or SnS2.[62]

The broadness of the XRD peaks and small grain sizes indicate potential nanostructuring of the resulting solution-processed Cu2SnS3 films. Scanning electron microscopy (SEM) images of drop-cast tetragonal and orthorhombic Cu2SnS3 films on Si substrates confirm the presence of nanostructured grains (Figures S8 and S9). While the metastable phase has only been observed as ligand-stabilized colloidal nanocrystals, the SEM images coupled with the FT-IR spectra prove the solution deposition of ligand-less orthorhombic Cu2SnS3 directly from molecular thiol–amine solutions. ICP-MS was used to assess the average elemental composition of our drop-cast films. Elemental compositions of Cu1.85Sn0.92S3.00 (Cu/(Cu + Sn) = 0.67) for the tetragonal polymorph and Cu1.98Sn1.02S3.00 (Cu/(Cu + Sn) = 0.66) for the orthorhombic polymorph were obtained. These values are within the range of previously reported experimental stoichiometries for tetragonal and “wurtzite” samples, as Cu2SnS3 is known to possess wide compositional variations with the Cu/(Cu + Sn) ratio ranging from 0.26 to 0.72 for the tetragonal polymorph and 0.49–0.81 for the “wurtzite” polymorph.[1]

X-ray photoelectron spectroscopy was used to gain insights into the valence states of the tetragonal and orthorhombic polymorphs. Survey scans of drop-cast tetragonal and orthorhombic Cu2SnS3 on Si substrates that had been exposed to atmospheric oxygen are provided in the Supporting Information (Figures S10 and S11). High-resolution spectra of the Cu 2p, Sn 3d, and S 2p regions of the tetragonal Cu2SnS3 are given in Figure S12, and the corresponding peak fittings are given in Table S3. The Cu 2p region can be fit with a single set of doublets at 952.0 and 932.2 eV with a splitting of 19.8 eV, consistent with an assignment of Cu+ in tetragonal Cu2SnS3. The Sn 3d region can be fit with two sets of doublets, each with a splitting of 8.4 eV. The lower binding energy peaks at 494.8 and 486.4 eV (FWHM = 0.78 eV) match well to Sn4+–S.[61] The smaller, higher binding energy doublet at 495.2 and 486.8 eV (FWHM = 1.75 eV) most likely corresponds to amorphous surface oxide or other nonstoichiometric tin oxides (SnO) due to Sn–O having a higher binding energy than corresponding Sn–S species, as well as amorphous oxides displaying larger FWHM values.[63] This is reasonable to expect as our materials were exposed to atmospheric oxygen several times during processing and travel for measurements, consistent with the presence of an O 1s peak in the XPS survey scan. The S 2p spectrum is fit well by a doublet with a splitting of 1.2 eV and S 2p3/2 peak centered at 161.4 eV. This is consistent with an assignment of an S2– metal sulfide.[39,63] A doublet at higher binding energies corresponds to oxidized surface SO species. Thus, XPS analysis confirms the expected formal valence states of the tetragonal polymorph to be (Cu+)2(Sn4+)(S2–)3. High-resolution spectra of the Cu 2p, Sn 3d, and S 2p regions of the orthorhombic phase are given in Figure S13, and the corresponding peak fittings are given in Table S4. The Cu 2p region can also be fit with a single set of doublets, indicating a single Cu+ environment, matching prior literature.[1,3,39,61,64] The Sn 3d spectrum was fit with two sets of doublets corresponding to Sn4+ in a sulfide environment, but the peaks corresponding to surface oxide are much larger relative to the Sn–S peaks when compared to the tetragonal phase.[1,3] The S 2p region is also fit with two sets of doublets, each with a splitting of 1.2 eV, which matches well with prior literature for an S2– metal sulfide environment in addition to oxidized surface SO species.[1,3,39,61,64]

Electronic Structure Calculations of Cu2SnS3 Polymorphs

DFT calculations indicate that the orthorhombic polymorph of Cu2SnS3 is metastable, with a hull energy of 7.6 meV atom–1 above the lowest-energy thermodynamically stable monoclinic polymorph of Cu2SnS3. In principle, metastable materials with predicted energies of ∼0.1 eV atom–1 above the ground state are synthesizable under appropriate conditions, depending on the chemical composition.[65] The low-lying metastability of our wurtzite-like orthorhombic polymorph is not unprecedented as the median energy above the ground state for metastable ternary polymorphs is 6.9 meV atom–1, irrespective of composition.[66]

To compare the optoelectronic properties of the monoclinic to the orthorhombic polymorph of Cu2SnS3, we calculated the band structure and density of states (DOS) for both polymorphs in their lowest-energy relaxed geometries using the HSE06 functional (Figures , S14, and 15). Although we did not experimentally refine the monoclinic polymorph, it has previously been determined to be the thermodynamically preferred lowest-energy phase, thus serving as the benchmark to which we compare our proposed orthorhombic polymorph of Cu2SnS3.[20] Additionally, the refined tetragonal polymorph possesses a disordered crystal structure (metal site contains 66.7% Cu + 33.3% Sn) resulting in an extensive and impractical site-occupation search to determine an ordered tetragonal polymorph to calculate due to the atomic position occupations being fractional rather than integer values.[20] To check that the tetragonal polymorph is not significantly more stable than the monoclinic and orthorhombic polymorphs, we have sampled several tetragonal structures and calculated their energies. The sampled structures had energies of 80–90 meV atom–1, >3 kBT for T = 300 K, higher than the monoclinic structure and an order of magnitude higher than the orthorhombic polymorph. The valence band maxima (VBM) are composed primarily of Cu-d and S-p orbitals for both polymorphs. The most considerable contributions to the conduction band minima (CBM) originate from S-s and p and Sn-s for both polymorphs (Figure S15). Interestingly, while both polymorphs have a low DOS near the CBM, the DOS is significantly lower for the orthorhombic polymorph of Cu2SnS3. A dense sampling of the Brillouin zone is required to obtain an accurate band structure along its high-symmetry k-path. We employed the computationally expensive and accurate HSE06 DFT functional for this purpose (Figure ). Both polymorphs feature a steep dispersion of the CBM around the Γ point, which rationalizes the lower DOS near the CBM. The orthorhombic polymorph may be preferable over the monoclinic polymorph in solar absorber applications because it has fewer states and lower DOS near the CBM, as can be seen in the larger-scale DOS shown in Figures S14 and S15. Sparse manifolds of states near the CBM can increase the lifetime of hot electrons.[67] The direct Γ–Γ band gaps of Cu2SnS3, as determined by the DOS and band structure calculations, are 0.64 and 0.49 eV for orthorhombic and monoclinic polymorphs, respectively. The predicted direct band gap nature of orthorhombic Cu2SnS3 from the calculations agrees with the experiment; however, the calculations underestimate these values likely due to the strong electron correlation in these materials that may not be fully captured even by the HSE06 functional.

Figure 2

Electronic band structure and density of states (DOS) calculated with the HSE06 functional for (a) monoclinic and (b) orthorhombic phases of Cu2SnS3. Orbital-resolved DOS are provided in Figure S15.

Property Measurements

The optical band gap of the resulting tetragonal and orthorhombic polymorphs were measured by UV–vis–NIR spectroscopy (Figures a,b, S16, and S17). Similar direct optical band gaps of Eg,dir = 0.94 and 0.88 eV for the tetragonal and orthorhombic polymorphs of Cu2SnS3, respectively, were determined by extrapolating the square of the linear portion of the Kubelka–Munk function-treated diffuse reflectance spectra. These band gaps lie within the range of previously reported experimental values for both tetragonal Cu2SnS3 and “wurtzite” Cu2SnS3 nanocrystals.[1,3,21,61] To demonstrate the photoresponse of these films, transient electrochemical photocurrent response measurements were performed with the two nanostructured polymorphs drop-cast on FTO. The measurements were performed in an electrolyte solution of 0.1 M Na2S/0.01 M sulfur dissolved in deionized water with Pt-wire counter and pseudoreference electrodes and the Cu2SnS3/FTO as the working electrode. With chopped illumination from a standard laboratory solar simulator, controlled potential electrolysis performed with an applied potential of −600 mV yielded stable photocurrents of ∼30 and ∼60 μA cm–2 for the tetragonal and orthorhombic phases, respectively (Figure c,d). The same measurements performed at positive applied potentials did not return a photocurrent, which confirms the p-type nature of these materials.[3−6]

Figure 3

Kubelka–Munk functions of absorption data to estimate direct optical band gaps for (a) tetragonal and (b) orthorhombic polymorphs of Cu2SnS3. Transient photocurrent response of solution-processed (c) tetragonal and (d) orthorhombic films of Cu2SnS3 deposited on FTO substrates in 0.1 M Na2S/0.01 M sulfur (aq) electrolyte under a potential of −600 mV vs Pt pseudoreference electrode using chopped AM1.5 light.

Influence of Ink Formulation on Polymorph

It has been well demonstrated that metastable, wurtzite-like polymorphs of multinary copper-containing chalcogenide nanocrystals are often templated by the initial nucleation of binary copper chalcogenide seeds with hexagonal (or nearly hexagonal) S2– or Se2– sublattices.[68−70] Indeed, this has also been observed in colloidal Cu2SnS3 nanocrystals, where “wurtzite” Cu2SnS3 was thought to initially nucleate as Cu2–S with a hexagonal close-packed anion sublattice.[71] To better understand the influence of ink formulation on Cu2SnS3 polymorph determination here, we employed a combination of TGA, negative ion mode electrospray ionization mass spectrometry (ESI-(−)MS), and powder XRD to study the differences between the Cu2S inks dissolved in EDT/en or merc/en. While ESI-(−)MS is an indirect method, it is effective in gaining insights into the identities of possible molecular solutes formed in thiol–amine inks.[29,72,73] TGA has been utilized to monitor decomposition temperatures of metal thiolates, which are known to decompose over a wide temperature range (100–350 °C) depending on the identity of the metal (e.g., Cu, Sn, In, etc.) and thiol used.[74−78] Photographs of the Cu2S inks in EDT/en and merc/en are supplied as insets of Figure a,b, showing major color differences between the two, with the EDT/en ink being dark orange/brown and the merc/en ink being virtually colorless. Direct-injection ESI-(−)MS was conducted to gain insights into the differences in the molecular solutes resulting from Cu2S dissolution in EDT/en and merc/en. The ESI-(−)MS data in Figure a,b reveal distinct differences in the major ions between the Cu2S EDT/en and merc/en inks. In the EDT/en ink, the major ion at m/z 246.8810 was attributed to [Cu(C2H4S2)2]− (or [Cu(EDT*)2]−), with the rest of the major ions attributed to different copper thiolate complexes. In the merc/en ink, the major ion at m/z 776.6852 was attributed to the polynuclear copper cluster [Cu5(C2H5OS)6]− (or [Cu5(merc*)6]−).

Figure 4

ESI-(−)MS of Cu2S dissolved in (a) EDT/en and (b) merc/en. TGA traces and derivative curves of a dried (c) Cu2S EDT/en ink and (d) Cu2S merc/en ink.

Derivative curves of TGA traces of these dried Cu2S inks with EDT/en and merc/en show differences between the high-temperature mass loss events (Figure c,d). The high-temperature mass loss event is larger in the EDT/en ink (27%) and occurs at higher temperatures (305 °C) relative to the merc/en ink (12% mass loss at 264 °C). The lower %mass loss suggests that [Cu5(merc*)6]− decomposition in the merc/en ink begins well below the high-temperature mass loss event at 264 °C, given the identical ink concentrations. EDT is expected to bind more strongly to copper as a bidentate X2-type ligand, whereas merc is expected to bind less strongly in either a bidentate or monodentate fashion as an XL-type or X-type ligand, respectively. Thus, the observed differences in decomposition profiles are consistent with the coordination chemistry of the EDT and merc.

Given the differences in the Cu2S inks in terms of their color, solute identity, and decomposition profile, we next sought to understand if the two inks crystallized in different phases of Cu2–S. The powder XRD patterns of the Cu2S inks dissolved in EDT/en and merc/en and annealed to 235 °C are given in Figure S18. This temperature was chosen due to this being the end of the first major mass loss event for both inks. It is observed that the EDT/en ink results in a cubic Cu2–S phase with a cubic close-packed anion sublattice (Figure S18a), which ultimately leads to tetragonal Cu2SnS3. On the other hand, the merc/en ink results in the formation of a monoclinic Cu2–S phase with a hexagonal close-packed anion sublattice (Figure S18b), with ultimately leads to orthorhombic Cu2SnS3. Taken in concert, we hypothesize that the binding strength of the metal thiolates between the two inks leads to the polymorphic control of our Cu2SnS3 films. The lower-temperature decomposition of the merc/en ink resulting from a less thermally stable [Cu5(merc*)6]− complex is a likely kinetic template for the growth of metastable wurtzite-like Cu2SnS3 through a transient monoclinic Cu2–S phase. The [Cu(EDT*)2]− complexes formed in the EDT/en inks decompose at higher temperatures due to the more tightly bound EDT ligands, forming the tetragonal polymorph of Cu2SnS3 from Cu2–S with a cubic S2– sublattice.

Conclusions

In summary, we have demonstrated the ability of an alkahest solvent system to allow for polymorph control in solution-processed tetragonal and orthorhombic Cu2SnS3 films by simply switching the identity of the thiol. When employing EDT/en as the solvents, the thermodynamically preferred tetragonal phase is recovered upon mild annealing. When the identity of the thiol is switched to merc, a metastable wurtzite-like orthorhombic polymorph is recovered. The solution-processed polymorph control is due to the relative strengths of the resultant metal–thiolate complexes formed in solution, which dictates the decomposition profiles and anionic sublattices of the binary Cu2–S phases that likely template the final Cu2SnS3 polymorphs, as has been described with numerous colloidal nanocrystal-based systems. The p-type tetragonal and orthorhombic films possess experimentally determined direct optical band gaps of 0.94 and 0.88 eV, respectively, and strong photocurrent responses. Understanding how the two ink formulations dictate polymorph crystallization should inform the decomposition and subsequent polymorph control of solution-processed thin films from other thiol–amine inks, such as possible polymorph control in quaternary materials such as Cu2ZnSnS4. DFT calculations reveal that the wurtzite-like orthorhombic polymorph has a favorable electronic structure compared to the thermodynamic monoclinic polymorph due to a lower DOS near the CBM suggesting increased hot carrier lifetimes beneficial for solar absorbers or other optoelectronic applications.