Room-temperature synthesis of lead-free copper(I)-antimony(III)-based double perovskite nanocrystals

Halide double perovskites (HDPs) have been proposed as potential alternatives to lead halide perovskites in the field of photovoltaics and photoluminescence (PL). Using the strategy of cation-transmutation, namely substituting two Pb²⁺ with monovalent(I) and trivalent(III) cations, a three-dimensional (3D) double perovskite structure with the formula A₂B(I)B(III)X₆ (A is a small organic/inorganic cation, B is a metal cation, and X is a halide anion) is formed.^1–3 As photoluminescent materials, the performance of HDPs is on par with that of lead halide perovskites, for example, bismuth-doped Cs₂(Ag_0.6Na_0.4)InCl₆ emitted efficient warm-white light with 86% ± 5% quantum efficiency and showed outstanding stability for over 1000 h.⁴ However, the photovoltaic performance of HDPs is still much inferior compared to lead halide perovskites. One representative HDP is Cs₂AgBiBr₆, which possesses an indirect bandgap of 1.95 eV.⁵ Such a large indirect bandgap greatly limits the efficient optical absorption of Cs₂AgBiBr₆ and thus leads to much lower power conversion efficiency (PCE) (∼6%) than that of lead-based halide perovskites (PCE ∼26%).^6–9 

Besides Ag-based HDPs, MA₂KBiCl₆ (MA = methylammonium)¹⁰ and Cs₂NaBiX₆ (X = Br, Cl)¹¹ were successfully synthesized but show even larger band gaps than Cs₂AgBiBr₆. Notably, the iodine-containing double perovskite Cs₂NaBiI₆ has also been synthesized. It showed a smaller indirect bandgap (1.66 eV) than Cs₂AgBiBr₆, and it has to date achieved a PCE of 0.42%.^12,13 Developing novel HDPs with narrow band gaps is highly desired for better photovoltaic performance. Compared to A₂B(I)B(III)X₆ HDPs [B(I) = K, Na and Ag], Cu(I)-based HDPs are predicted to have narrower band gaps.¹⁴ Moreover, in contrast to the noble metal Ag(I), the high abundance and low price of Cu provides great potential for commercial production. However, Cu(I)-based 3D HDPs bulk crystals have been rarely reported experimentally. To the best of our knowledge, only Cs₂CuSbCl₆ double perovskite nanocrystals (DPNCs) have been successfully synthesized via a modified hot-injection method.¹⁵ Cs₂CuSbCl₆ exhibits an indirect bandgap of 1.66 eV, which is comparable to that of Cs₂NaBiI₆ and much smaller than that of Cs₂AgBiBr₆. Furthermore, substituting Cl with Br or I in Cs₂CuSbCl₆ can potentially further reduce the bandgap and enhance the optical absorption for photovoltaic applications.

The ground-state structure of Cu(I) halides favors three- or four-fold rather than six-fold coordination, which is suggested by theoretical calculations.¹⁶ However, Cu(I)-based perovskite structures containing [CuX₆]⁵⁻ octahedra can be obtained via non-equilibrium methods. Besides the aforementioned example of Cs₂CuSbCl₆ nanocrystals, the double perovskite structure with six-fold coordination of Cu(I) can also be stabilized through dimensional reduction to layered perovskites. Recently, some Cu(I)-based layered double perovskites have been successfully synthesized where Cu(I) displays extremely distorted six-fold coordination but the perovskite structure is maintained, for instance, (C₆H₁₆N₂)₂CuBiI₈,¹⁷ (PPDA)₂CuRuCl₈,¹⁸ (PEA)₄CuInCl₈,^19,20 and (BA)₄CuInCl₈²⁰ (PPDA = para-phenylenediammonium, PEA = phenethylammonium, and BA = butylammonium). It is evident that both the large organic cations and long-chain organic ligands can stabilize the Cu(I)-based double perovskite structure, offering an opportunity to investigate their optoelectronic properties. Therefore, developing new Cu(I)-based 3D HDPs at the nanoscale is a promising route to overcome the issues of kinetic instability and high formation energy of these materials.¹⁵ 

The synthesis of halide perovskite nanocrystals (PNCs) has greatly developed since the first successful synthesis of CsPbX₃ (X = Cl, Br, and I) PNCs via a hot-injection method.²¹ The two most-developed colloidal syntheses are hot-injection and room-temperature ligand-assisted reprecipitation (LARP) methods. Although the hot-injection method can give the best control over the size and morphology of PNCs, it is not easily scalable due to the required high temperatures and protective inert gas environments, leading to high energy consumption and complex operation during production.^22–24 Due to the inherent ionic nature of halide perovskites, it is also possible to synthesize high-quality PNCs by the LARP method under ambient conditions. This relatively simple, energy-efficient, and easily scalable approach relies on the spontaneous crystallization of substances upon reaching a supersaturated state, which can be achieved by lowering the temperature, evaporating solvents, or adding anti-solvents in which the solubility of the substance is low. However, the typical LARP method requires coordinating polar solvents with high precursor solubilities, such as N,N-dimethylformamide (DMF), which is not favored for large-scale production due to its toxicity.^25,26 In addition, they are technically difficult to remove during the purification process, leading to a defective surface and poor stability.²³ Searching for less toxic and easily removable solvents is a major objective for the synthesis of PNCs. Furthermore, the development of the LARP method is far behind that of the hot-injection method regarding control over the size and morphology of the PNCs, especially for DPNCs.

Various post-treatments for PNCs can assist to transform their morphology, composition, and structure.²⁷ The soft lattice of PNCs allows for tuning their chemical composition via anion-exchange by introducing halide salts to a solution of PNCs, which offers a facile route to access novel materials with different optoelectronic properties.^28–30 For DPNCs, we note that Cs₂AgBiI₆ is impossible to synthesize through direct synthesis; however, its DPNCs have been obtained by the post-treatment of Cs₂AgBiBr₆ DPNCs via anion exchange.³⁰ Notably, although Cs₂CuSbCl₆ DPNCs have been obtained via hot-injection,¹⁵ the bromide and iodide congeners have not yet been synthesized. Given that the anion-exchange reactivity of nanocrystals can provide access to metastable phases that may be difficult or impossible to synthesize directly,^30–32 we hypothesized that Cs₂CuSbBr₆ and Cs₂CuSbI₆ DPNCs may be obtained by using this method. Furthermore, the poor solubility of precursors inhibits the fabrication of high-quality HDP films via a solution-processed method.³³ In contrast, DPNCs show favorable solution processability, offering a facile and effective way to obtain high-quality DPNC films.

In our work, we demonstrate for the first time a facile and fast synthesis method for Cs₂CuSbCl₆ DPNCs via a modified LARP method under ambient conditions. We choose methanol as a solvent for precursor salts as it is less toxic and easily removed and systematically study the effects of types of anti-solvents and the concentration of ligands on the size and morphology of Cs₂CuSbCl₆ DPNCs. Our computational structure search shows that the double perovskite Cs₂CuSbCl₆, containing [CuCl₆] and [SbCl₆] moieties, is metastable (30 meV/atom) compared to the ground state with [CuCl₃] and [SbCl₆] polyhedra but that it can be stabilized through the solution-based nanocrystal synthesis. Using the anion-exchange method, Cs₂CuSbBr₆ DPNCs are obtained for the first time, featuring a narrow bandgap of 0.9 eV. Finally, taking advantage of the solution processability of DPNCs, smooth and dense Cs₂CuSbCl₆ and Cs₂CuSbBr₆ DPNC films are successfully fabricated.

Cesium acetate (CsAc, 99.9%), copper chloride (99.999%), antimony trichloride (≥99.95%), methanol (anhydrous, 99.8%), isopropanol (anhydrous, 99.5%), chloroform (anhydrous, ≥99%), o-xylene (anhydrous, 97%), toluene (anhydrous, 99.8%), m-xylene (anhydrous, 99%), benzene (anhydrous, 99.8%), hexane (anhydrous, 95%), octane (anhydrous, ≥99%), ethyl acetate (anhydrous, 99.8%), methyl acetate (anhydrous, 99.5%), oleic acid (90%), oleyamine (70%), bromotrimethylsilane (>97%), and iodotrimethylsilane (97%) were used (as purchased from Sigma-Aldrich) without further purification.

Synthesis of Cs₂CuSbCl₆ DPNCs. The Cs₂CuSbCl₆ DPNCs were synthesized by a modified LARP method under ambient conditions. In a typical synthesis, 383.8 mg (2 mmol) of cesium acetate (CsAc) was dissolved in 1 ml of methanol as the source of Cs. To prepare the copper/antimony source, 268.9 mg (2 mmol) of copper(II) chloride (CuCl₂) and 912.5 mg (4 mmol) of antimony trichloride (SbCl₃) were dissolved in 1 ml of methanol. The CsAc precursor (150 μl) was added directly into the anti-solvent (9 ml) with different amounts of oleic acid (OA) and oleyamine (OAm) to form a transparent solution, termed the host solution. The solution of CuCl₂ and SbCl₃ (150 μl), termed the guest solution, was added into the host solution dropwise under vigorous stirring. The color of the mixture changed from blue to green to brown and finally to black as all the guest solution was mixed with the host solution. The reaction was terminated after 5 min. For purification of PNCs, the crude black DPNC solution was transferred to a 50 ml centrifuge tube with the addition of ethyl acetate (EA) at a volume ratio of 9:1 (crude DPNC solution to EA) to enhance precipitation. The mixture was centrifuged at 7830 revolutions per minute (rpm) [7197 relative centrifugal force (rcf)] for 10 min. In the case of hexane as the anti-solvent, no extra EA was required because the crude solution could be separated easily. The supernatant was then decanted, and the residual pellet was re-dispersed in hexane (5 ml). This was then centrifuged a second time at 4500 rpm (2377 rcf) for 5 min. The supernatant was collected for further characterization, post-treatments, and film deposition.

Anion exchange reaction. Cs₂CuSbCl₆ DPNCs synthesized by using 9 ml of toluene with 400 μl of OA and 175 μl of OAm were chosen as the starting materials. In a typical anion exchange reaction, the Cs₂CuSbCl₆ DPNC solution obtained via the method described above was used directly with a concentration of ∼0.02 M, which was determined by inductively coupled plasma optical emission spectrometry (ICP-OES). Meanwhile, a 0.1 M solution of bromotrimethylsilane (TMSBr) in hexane was prepared. The reaction was performed in a N₂-filled glovebox. Different amounts of TMSBr solution (0.2, 0.4, 0.8, 1.5, 2, 3 ml) were added to 1 ml of Cs₂CuSbCl₆ DPNC solution (containing 0.02 mmol of Cs₂CuSbCl₆) to achieve different levels of conversion. After the addition of TMSBr solution, the mixture was stirred for 5 min and then dried under vacuum to remove excess reagent and by-products. The final samples were dissolved in hexane or octane for further characterization, post-treatments, and film deposition. The attempt to exchange Br to I was performed analogously but with iodotrimethylsilane (TMSI) instead.

Post-purification and thin-film synthesis of DPNCs. To prepare the DPNC solution for film deposition, the DPNCs were further purified by washing with EA to remove excess ligands and then redispersed in octane at a concentration of 70 mg/ml. A two-step spin-coating process was used to fabricate films. Specifically, 30 µl of the as-prepared octane solution of DPNCs was spin-coated on the substrate at 1000 rpm for 20 s and 2000 rpm for 45 s. Then, methyl acetate (MA, 150 µl) was dropped on the DPNC layers to remove the long-chain ligands during the second spin-coating step. Films with different thicknesses were obtained by repeating this process. The films were annealed at different temperatures to remove the residual solvents for further characterization and application.

X-ray diffraction (XRD) measurements. DPNCs without purification and thin-film XRD data were recorded using a Bruker D8 Discover diffractometer with Ni-filtered Cu Kα radiation and a LynxEye position-sensitive detector in Bragg−Brentano geometry. Powder XRD data of DPNCs resulting from the post-purification process were recorded using a STOE STADI P diffractometer with Ge-filtered Cu Kα radiation and a DECTRIS solid-state strip detector MYTHEN 1K in Debye−Scherrer geometry. For the preparation of samples without post-purification, the DPNCs were dried under vacuum to remove the solvents and redispersed in a small amount of hexane. The corresponding solutions with a high mass concentration were dropped on glass and dried under vacuum to remove the residual hexane. To prepare the samples resulting from post-purification for powder XRD measurements, the pellets obtained after purification were dried under vacuum and then ground into a fine powder, which was not sticky anymore due to the low ligand density.

Absorption measurements. The optical absorption spectra of both DPNC solutions and films were recorded on a Perkin–Elmer Lambda 1050 spectrometer equipped with a 150 mm integrating sphere.

Steady-state photoluminescence (PL) and time-correlated single-photon counting (TCSPC). A home-built confocal laser scanning microscope (CLSM) setup was used for characterizing the photoluminescence of DPNCs. Confocal PL images were obtained by placing the Cs₂CuSbCl₆ DPNC film sample (synthesized in toluene with 400 μl of OA and 175 μl of OAm) on an x, y piezo scan stage at room temperature. A subpicosecond pulsed fiber laser with a repetition rate of 40 MHz (Toptica iChrome TVIS) with a central wavelength of 476 nm was employed as the illumination source. An air objective [Nikon Chrome-Free Infinity-Corrected (CFI) Plan Apochromat 60×] with a numerical aperture of 0.95× and 60× magnification was used for focusing the laser light and for collecting the sample PL. The fluence of the excitation light in the focus was set to 1.2 mJ cm⁻². Residual scattered laser light was suppressed using a long-pass filter (LP) with an edge at 490 nm. For the measurement of the PL spectra, the emitted light was sent to a grating spectrometer (Andor Shamrock SR303i) equipped with a charge coupled device (CCD) camera (Andor Newton Du-920-OE). For time-resolved measurements, the PL light was directed to an avalanche photodiode (APD, MPD PDM Series) combined with a TCSPC-Card (Becker & Hickl).

Scanning electron microscopy (SEM) and energy dispersive x-ray spectroscopy (EDX). The DPNC samples were prepared by dropping diluted DPNC solutions (in hexane) on aluminum foil and dried under vacuum. The DPNC film samples were prepared on the ITO substrates following the method mentioned above. SEM and EDX measurements were performed on an FEI Helios G3 UC instrument equipped with an additional concentric backscattered electron detector.

High-resolution transmission electron microscopy (HR-TEM). HR-TEM images were obtained on a Titan Themis instrument at 120 kV accelerating voltage. HR-TEM specimens were prepared by dropping 10 μl of dilute DPNC solutions (in hexane) on carbon-coated Cu grids.

The details of x-ray photoelectron spectroscopy (XPS) and ultraviolet photoelectron spectroscopy (UPS) measurements can be found in the supplementary material.

Crystal structure prediction for Cs₂CuSbCl₆ was performed using the Ab Initio Random Structure Searching (AIRSS) method.^34,35 The search was constrained to the number of atoms in the reduced formula of 10 and generated and relaxed 829 random structures. The plane wave density functional theory (DFT) code CASTEP with core-corrected ultrasoft pseudopotentials generated as defined in the built-in QC5 library was used for efficient structure sampling.³⁶ A plane wave cutoff energy was set to 400 eV. The reciprocal space sampling was performed by using Monkhorst–Pack grids with a spacing of 0.05 × 2π Å⁻¹. For further relaxations and property calculations for low-energy structures, the Vienna ab initio simulation package (VASP) was used.^37,38 The projector augmented wave with exchange correction potential of Perdew–Burke–Ernzerhof (PBE) was used.³⁹ An energy cutoff of 400 eV and a k-point density of 0.03 × 2π Å⁻¹ were employed. The convergence criteria for the energy and force were set to 10⁻⁵ eV and 0.01 eV/Å, respectively. For electronic structure calculations, a hybrid functional with a 25% nonlocal Fock exchange (HSE06) was employed.⁴⁰ The spin–orbit coupling (SOC) effect was included as well. The calculations of the energy above the convex hull (E_hull) were performed via pymatgen.^41,42 The theoretical conversion efficiency [spectroscopic limited maximum efficiency (SLME)] of the solar cell absorbers was calculated based on the method proposed by Yu et al. and a python code (SL3ME).^43,44

Theoretically, Cu(I)- and Ag(I)-based 3D HDPs have been predicted to be potential substitutions for lead-based halide perovskites as less-toxic photovoltaic absorbers.^14,45,46 Ag(I)-based 3D HDPs have been successfully synthesized, for example, Cs₂AgBiX₆ (X = Cl, Br),⁴⁷ Cs₂AgSbCl₆,⁴⁸ and Cs₂AgInCl₆.⁴⁸ However, Cu(I)-based 3D HDPs have been rarely studied experimentally. The relative instability of Cu(I)-based 3D HDPs is attributed to the preference of Cu(I) for three- or four-fold coordination rather than six-fold coordination, based on the comparison of the DFT calculated total energy for simple binary halides CuCl in different crystal structures and unsuccessful experimental trials for synthesizing Cu(I)-based double perovskites [Cs₂CuInX₆ (X = Cl, Br)] via solid-state methods.⁴⁹ Notably, several Cu(I)-based layered double perovskites [(C₆H₁₆N₂)₂CuBiI₈,¹⁷ (PPDA)₂CuRuCl₈,¹⁸ (PEA)₄CuInCl₈,^19,20 and (BA)₄CuInCl₈²⁰] have been successfully synthesized where the Cu(I) atoms have distorted six-fold coordination, and the perovskite structure is maintained within a two-dimensional plane.⁴⁹ 

Here, a global crystal structure search of the quaternary compound Cs₂CuSbCl₆ was performed to identify the possible crystal structures of Cs₂CuSbCl₆. We found three structure types including three-, four-, and six-fold coordinated Cu(I) atoms, which demonstrate the diversity of the local environment of the Cu(I)-halogen moieties [Fig. 1(a)]. The calculated E_hull shows that the R3 symmetry of Cs₂CuSbCl₆ with planar [CuCl₃] polyhedra is the most energetically favorable, followed by the Cm symmetry containing [CuCl₄] tetrahedra and Fmm2 and $Fm3̄m$ symmetry with [CuCl₆] octahedra. The metastable phase $Fm3̄m$ is 30 meV/atom higher in energy, which is within the energy range of possible entropy stabilization. Indeed, Cs₂CuSbCl₆ DPNCs with alternating [CuX₆] and [SbX₆] octahedra were synthesized by Zhou et al.¹⁵ In this study, we also demonstrate the successful formation of metastable Cs₂CuSbCl₆ in $Fm3̄m$ symmetry with [CuX₆] moieties utilizing the LARP method.

As illustrated in Fig. 1(b), Cs₂CuSbCl₆ DPNCs were synthesized by a modified LARP method under ambient conditions. For the conventional LARP method, all reagents are dissolved in a good solvent (e.g., DMF or dimethyl sulfoxide) and dropped into an anti-solvent (e.g., toluene or hexane) with the assist of organic ligands to form PNCs. However, the poor solubility of cesium chloride makes it difficult to prepare one solution with all the reagents. Thus, we modified this method by using CsAc instead of CsCl as the Cs precursor since the former has a higher solubility than the latter. The Cs and Cu/Sb/Cl precursors were prepared separately. Methanol was chosen as the solvent for the perovskite precursors, which is less toxic compared to the commonly used DMF.²⁶ In addition, methanol has a low boiling point, making it readily removable, and thus avoiding issues such as redissolving nanocrystals or creating crystallographic surface defects. For more details, see Sec. II and supplementary material Video.

For LARP, the polarity of solvents plays an important role in the crystal formation of PNCs.^50–52 Dutta et al. reported that the rate of formation of CsPbBr₃ PNCs can be drastically altered with a small difference in solvent polarity of anti-solvents.⁵⁰ Here, seven hydrophobic solvents with different polarities were chosen as the media for host solutions. They were isopropanol, chloroform, o-xylene, toluene, m-xylene, benzene, and hexane with the dielectric constants 19.92, 4.81, 2.57, 2.38, 2.37, 2.28, and 1.89, respectively. The experimental details can be found in Sec. II and supplementary material. We carried out XRD measurements to confirm the structure and phase purity of the products obtained by using the aforementioned seven anti-solvents before [Fig. 1(c)] and after post-purification [Fig. 1(d)]. As shown in Fig. 1(c), the XRD patterns of as-prepared DPNCs without post-treatment are confirmed as the cubic Cs₂CuSbCl₆ structure with $Fm3̄m$ space group symmetry where corner-sharing octahedra of [CuCl₆] and [SbCl₆] alternate along three directions, and Cs⁺ ions occupy the cuboctahedral cavities. The post-purification of DPNCs can remove excess insulating ligands, potentially benefitting the charge-carrier mobility of the DPNCs in film morphology (see below). However, there is a trade-off between the targeted increase of charge-carrier mobility and the structural stability of the DPNCs, especially for Cs₂CuSbCl₆, which has not been reported in the form of bulk crystals. It is thus crucial to confirm the phase purity of Cs₂CuSbCl₆ DPNCs after post-purification. As shown in Fig. 1(d), the DPNCs display good phase purity after post-treatment, except for the products obtained by using chloroform and isopropanol as anti-solvents. The additional reflections at 2θ = 21.6° and 30.7° from the sample synthesized in chloroform correspond to the (201) and (311) crystal planes of Cs₂CuCl₄ (PDF no. 01-071-0901), respectively. The impurity in the sample synthesized in isopropanol is confirmed as Cs₂CuCl₄. This phenomenon, where residual chloroform and isopropanol, with relatively higher polarities, can affect the phase purity of Cs₂CuSbCl₆ DPNCs during the post-purification process, has also been observed in lead-based PNCs.⁵² 

As mentioned in the Sec. II, CuCl₂ was chosen as the Cu-providing precursor. Interestingly, the formal oxidation state of Cu in CuCl₂ is +2, while that in the desired HDP structure Cs₂CuSbCl_6, it is +1. To confirm the structure of the product, we compared the diffractograms to exclude the 〈111〉-oriented layered double perovskite structure Cs₄CuSb₂Cl₁₂, where the oxidation of Cu would be +2 [Figs. 1(c) and 1(d)]. To further distinguish the oxidation state of Cu, x-ray photoelectron spectroscopy (XPS) measurements were carried out (Fig. S1). The Cu core-level spectrum consists of a 2p doublet exhibiting binding energies of 931.6 and 951.7 eV corresponding to Cu 2p_3/2 and Cu 2p_1/2, respectively. This is consistent with the Cu(I) state.¹⁵ The characteristic core level for Cu(II) is expected at about 3 eV higher binding energy (935 and 955 eV) and is not observed. A similar phenomenon was also observed in the synthesis of Cs₂CuSbCl₆ DPNCs with the hot-injection method.¹⁵ The other elements were also found in their anticipated oxidation states, that is, Cs(I), Sb(III), and Cl(I), respectively.

In order to examine the effects of anti-solvents on the size and morphology of DPNCs, the as-prepared Cs₂CuSbCl₆ DPNCs were characterized by SEM (Fig. 2). The DPNCs for SEM measurements were prepared without post-purification. Interestingly, different morphologies can be achieved by using anti-solvents with diverse dielectric constants [summarized in Fig. 2(h)]. As shown in Fig. 2, Cs₂CuSbCl₆ DPNCs synthesized in hexane show irregular shapes with an average size of about 43 nm. In benzene, the as-synthesized samples show hexagonal plate morphology with an average length of 194 nm and a width of 82 nm. By using m-xylene as an anti-solvent, hexagonally shaped nanocrystals can be obtained with smaller size (45 nm on average) but less regular in morphology, which is quite similar to nanocrystals synthesized in toluene (49 nm on average). However, when o-xylene is used as the anti-solvent, the morphology shows a significant change, forming nanorods with an average length of about 307 nm and a width of 28 nm. Chloroform with relatively higher polarity can assist the formation of a mix of hexagonally shaped nanocrystals and nanocuboids with an average size of 67 nm. Cs₂CuSbCl₆ nanocuboids (mean size 123 nm) and nanowires can be synthesized with isopropanol serving as an anti-solvent. With the highest dielectric constant among these seven solvents, we hypothesize that isopropanol can not only act as a solvent but also a nucleophile to partially remove OAm on the surface of DPNCs, which leads to the formation of relatively larger nanocrystals. The same phenomenon was also observed in the synthesis of CsPbBr₃ nanocrystals.⁵² Size distribution histograms of Cs₂CuSbCl₆ DPNCs synthesized with these seven different anti-solvents can be accessed in the supplementary material (Fig. S2). Based on our observations, it is clear that the type of anti-solvent has a strong impact on the kinetic pathways, which results in different anisotropically grown nanostructures.

In the synthesis of lead-based PNCs and other conventional quantum dots, organic acids and bases are commonly used ligands, with the ability to solvate the precursors effectively, affect the kinetic pathway to control the shape and size of nanomaterials, and avoid the aggregation or agglomeration of the nanomaterials.^23,53,54 OA and OAm are frequently used pairs to synthesize PNCs. The impact of OA and OAm on lead-based PNCs synthesized by the hot-injection method has been already comprehensively studied, but their effect on DPNCs prepared via the LARP method remains poorly understood.^55,56

First, we studied the effect of different amounts of OA on the size and morphology of Cs₂CuSbCl₆ DPNCs synthesized by using three representative anti-solvents: chloroform, toluene, and hexane, where three volumes of OA (300, 400, and 500 µl) were tested, and the volume of OAm was fixed at 175 µl in the synthesis recipe. When chloroform was used as an anti-solvent, the smallest amount of OA (300 µl) leads to irregular morphologies, which are dumbbell-shaped nanocrystals with a length of about 120 nm, rhombus-shaped nanocrystals with a length of about 90 nm, and other nanocrystals with irregular shapes [Fig. 3(a)]. Increasing the amount of OA to 400 µl, smaller nanospheres and nanocuboids can be obtained with an average size of 27 nm [Fig. 3(b)]. Further increasing the amount of OA to 500 µl, the size of nanocrystals increases, resulting in nanocuboids with an average length of 97 nm and a width of 50 nm [Fig. 3(c)]. In the case of toluene as an anti-solvent, a large range of uniform nanorods with an average width of 29 nm and a length of 104 nm can be synthesized when the smallest amount of OA (300 µl) is used [Fig. 3(d)]. Increasing the amount of OA, the morphology can be changed to nanocuboids [Fig. 3(e) with 400 µL of OA, mean size 42 nm] and hexagonally shaped nanocrystals [Fig. 3(f) with 500 µL of OA, mean size 75 nm]. The parameter window for successful synthesis of Cs₂CuSbCl₆ DPNCs is relatively narrow when hexane is chosen as the anti-solvent. The DPNCs could not be obtained when 300 µl of OA was used in the reaction, leading to precipitation of the reagents with no uniform dispersion in hexane after the centrifugation, indicating that the size of particles was too big to be stable in the solution. Additionally, by increasing the amount of OA to 500 µl, big particles with a size of hundreds of nanometers can be produced with irregular morphologies [Fig. 3(i)]. However, when 400 µl of OA was introduced to the reaction, irregular-shaped Cs₂CuSbCl₆ DPNCs with a relatively broad size distribution were obtained [Fig. 3(h)]. In short, toluene is demonstrated to be the best anti-solvent to synthesize Cs₂CuSbCl₆ DPNCs, resulting in well-defined sizes and morphologies.

To study the impact of the amount of OAm on the synthesis of Cs₂CuSbCl₆ DPNCs, toluene was used as the anti-solvent. Four different amounts of OAm in the synthesis recipe were tested, which were 150, 175, 200, and 225 µl. The smallest amount of OAm leads to the formation of nanocuboids with an average length of 98 nm and a width of 66 nm [Fig. 4(a)]. When further increasing the amount of OAm to 175 and 200 µl, the morphology changes to a mixture of nanocuboids and nanospheres with average sizes of 32 nm [Figs. 4(b)] and 43 nm [Fig. 4(c)], respectively. When the amount of OAm reaches 225 µl [OA/OAm ratio of 1.78, Fig. 4(d)], the morphology of nanorods with an average length of 116 nm and a width of 30 nm is obtained, similar to the nanorods synthesized by using 300 µl OA and 175 µl OAm [OA/OAm volume ratio of 1.71, Fig. 3(d)], which have a similar OA/OAm ratio. The same phenomenon can also be observed in the cases of OA/OAm volume ratios of 2.85 [500 µl of OA and 175 µl of OAm, Fig. 3(f)] and 2.67 [400 µl of OA and 150 µl of OAm, Fig. 4(a)], where bigger particles can be obtained. A smaller quantity of OAm provides faster growth kinetics and forms Cs₂CuSbCl₆ DPNCs with relatively bigger sizes.⁵² The above observations indicate that the ratio of OA to OAm plays an important role in controlling the size and morphology of Cs₂CuSbCl₆ DPNCs.

To evaluate the suitability of Cs₂CuSbCl₆ DPNCs as a potential absorber for photovoltaic applications, we characterized the optical absorption of the nanocrystals synthesized by using different anti-solvents [Fig. 5(a)]. The band gaps of these nanocrystals were calculated by using the Tauc plot method⁵⁷ (Fig. S3), which reveals that the Cs₂CuSbCl₆ DPNCs synthesized with different anti-solvents show similar indirect band gaps with values between 1.67 and 1.78 eV. The Cs₂CuSbCl₆ DPNCs synthesized in chloroform have the smallest bandgap of 1.67 eV, while the ones synthesized in isopropanol have the biggest bandgap of 1.78 eV. Cs₂CuSbCl₆ DPNCs synthesized by using other anti-solvents showed similar band gaps of around 1.72 eV (Fig. S3). The reason for this observation is not clear, for example, different surface–ligand interactions and surface defect populations may lead to the differences in the value of the bandgap.²⁷ 

The PL of Cs₂CuSbCl₆ DPNCs was observed after subpicosecond pulsed laser excitation at 476 nm [Fig. 5(b)]. The PL spectrum of Cs₂CuSbCl₆ DPNCs shows a broad emission band peaking at 1.8 eV (689 nm) with a full width at half maximum (FWHM) of 0.56 eV (215 nm). Surprisingly, a 10 times higher PL intensity was recorded after irradiation with a 405 nm laser. Such “light soaking” (the PL of the sample increases after being exposed to UV light) also changes the PL spectrum of Cs₂CuSbCl₆ DPNCs. A significant spectral broadening to the blue side with the peak now centering at 1.85 eV (670 nm), and a FHWM of 0.65 eV (235 nm) was revealed when the sample was fully light-soaked, i.e., without further increase in PL intensity upon UV exposure. A TCSPC transient of light-soaked Cs₂CuSbCl₆ DPNCs recorded with 476 nm excitation is shown in Fig. S4. The PL transients of Cs₂CuSbCl₆ DPNCs can be modeled by a bi-exponential decay function with a very short decay time below 10 ps (the temporal resolution of the experiment) and a slower decay time of 97 ps.

We successfully obtained Cs₂CuSbBr₆ DPNCs via an anion-exchange reaction of Cs₂CuSbCl₆ DPNCs. Various anion-exchange agents can be used to drive the conversion; however, the agents should be chosen carefully as a wide range of reagents may cause the conversion of crystal phases or stoichiometries of the desired DPNC phase.⁵⁸ In addition, the degradation of PNCs can also occur during the purification after the anion-exchange reaction aimed at removing by-products and excess reagents. TMSI, TMSBr, and TMSCl have been widely used for anion-exchange reactions of PNCs, as they can be easily removed due to their low boiling points of 107, 79, and 57 °C, respectively. They offer good miscibility with nonpolar alkane and arene solvents like hexane and toluene, which are commonly used for the dispersion of PNCs. The strength of the TMS-X bond (X = I, Br, Cl) increases significantly from the I to Br to Cl, and TMS-X (X = I, Br, Cl) have dissociation energies of 77, 96, and 113 (kcal/mol), respectively, which means that the substitution of TMS-Br bonds for TMS-I bonds or TMS-Cl bonds for TMS-Br is energetically favorable.⁵⁸ Here, TMSBr was adopted as the reagent for the conversion of Cs₂CuSbCl₆ DPNCs to Cs₂CuSbBr₆ DPNCs. To achieve different levels of conversion, Cs₂CuSbCl₆ DPNCs (0.02 mmol) were reacted with different amounts of TMSBr solution (0.1 M). The details of the reactions can be found in Sec. II.

To identify the crystal structure of the products after the anion-exchange reaction, XRD measurements were carried out. An increasing amount of TMSBr leads to a shift to lower angles 2θ when compared to Cs₂CuSbCl₆ DPNCs, indicating the increase of lattice constants, as expected after the substitution of Cl with Br (Fig. S5a). The XRD pattern of the completely converted product matches well with the simulated XRD pattern of Cs₂CuSbBr₆ DPNCs [Fig. 6(a)], and the expansion of lattice constants can also be verified by HRTEM (Fig. S6). The XPS results further confirm the complete removal of Cl in the final sample, as the characteristic peaks of Cl 2p (located at 198.5 and 200.1 eV) are not observed in the XPS spectrum (inset of Fig. S7a). Additionally and importantly, the oxidation state of Cu is maintained as +1. The expected element ratio in the resulting Cs₂CuSbBr₆ DPNCs is reasonably well confirmed as 2:0.7:0.8:6 (Cs:Cu:Sb:Br) by XPS, which is slightly off-stoichiometry. However, within the common uncertainties, the XPS data match the expected element ratios of the DPNCs quite well (which might also slightly differ due to the possible existence of surface defects). HRTEM and SEM characterizations of the products obtained after anion-exchange reactions further illustrate that both the size and the morphology of the DPNCs are maintained (Figs. S6b and S8).

The introduction of Br anions into Cs₂CuSbCl₆ DPNCs by the addition of TMSBr solutions results in a significant red-shift of the absorption edges [Fig. 6(b) and Fig. S5(b)]. The color of the DPNC solution changes from light brown to black after the complete conversion [inset of Fig. 6(b)]. The absorption spectra of DPNC solutions with different conversion levels are shown in Fig. S5b. Figure 6(b) shows the absorption spectra of both Cs₂CuSbCl₆ and Cs₂CuSbBr₆ DPNC solutions (completely converted). The Cs₂CuSbBr₆ DPNCs exhibit a broad absorption with a strongly reduced indirect bandgap of 0.9 eV, which is estimated with the Tauc plot method (Fig. S9). This is, to the best of our knowledge, the narrowest bandgap among the experimentally reported HDP materials reported to date.

While the preparation of Cs₂CuSbBr₆ DPNCs by the anion-exchange reaction was demonstrated to be successful, at this point, attempts to access the conversion of Cs₂CuSbBr₆ to Cs₂CuSbI₆ DPNCs via the same method by using TMSI proved to be unsuccessful. A preceding theoretical study indicates that Cs₂CuSbI₆ is unstable with a negative decomposition enthalpy of −11 meV/atom.¹⁴ As a result, our attempts at synthesizing Cs₂CuSbI₆ DPNCs led to an orange mixture of products that were confirmed to be Cs₃Cu₂I₅, Cs₃Sb₂I_9, and CsI by XRD (Fig. S10).

On the basis of DFT computations, the band structures and density of states (DOS) of Cs₂CuSbCl₆ and Cs₂CuSbBr₆ in $Fm3̄m$ were calculated. As displayed in Figs. 6(c) and 6(d), Cs₂CuSbCl₆ and Cs₂CuSbBr₆ exhibit indirect band gaps of 1.68 and 1.18 eV with valence band maximum (VBM) and conduction band minimum (CBM) at X point and L point, respectively. Our HSE06 + SOC calculated bandgap for Cs₂CuSbCl₆ agrees well with the previous HSE06 calculated one (1.70 eV), indicating that the effect of spin–orbit coupling (SOC) on the electronic structure of Cs₂CuSbCl₆ is negligible. In addition, the calculated band gaps of Cs₂CuSbCl₆ and Cs₂CuSbBr₆ are in good agreement with our experimental data of these DPNCs. For Cs₂CuSbX₆ (X = Cl, Br), the valence band edges are dominantly derived from the Cu-3d states and the X-p (X = Cl, Br) orbitals while the conduction band edges are mainly composed of Sb-5p and the X-p states (X = Cl, Br), as shown in Figs. 6(c) and 6(d). Cs₂CuSbBr₆ has a narrower bandgap than Cs₂CuSbCl₆, which could be attributed to the VBM upshifting and CBM downshifting in bromides with respect to those in chlorides. The same trend of band gaps has been observed in comparing Cs₂AgBiBr₆ and Cs₂AgBiCl₆.^59,60 The band dispersion around the band edges for Cs₂CuSbBr₆ and Cs₂CuSbCl₆ is similar, which is like in the case of Cs₂AgBiX₆ (X = Cl, Br).⁶¹ The indirect and direct band gaps of Cs₂CuSbX₆ (X = Cl, Br) are greatly reduced compared to Cs₂AgSbX₆ (X = Cl, Br) (see Table S1).

Furthermore, we computed the spectroscopic limited maximum efficiency (SLME) for Cs₂CuSbX₆ (X = Cl, Br) and Cs₂AgSbBr₆ as a function of the thickness of the absorber layer. The SLME of Cs₂AgSbCl₆ was not calculated since the direct transition bandgap is out of the range of the visible spectrum. As displayed in Fig. S11, the predicted SLME of Cs₂CuSbX₆ (X = Cl, Br) and Cs₂AgSbBr₆ sharply increases to 4.1, 5.6, and 5.1% for a film thickness reaching 0.3 µm. When the film thickness further increases, the predicted SLMEs of all compounds increase slowly and gradually saturate. Among the three compounds, Cs₂CuSbBr₆ exhibits the highest predicted SLME (8.5%) at a thickness of 3 µm. The SLME of Cs₂CuSbBr₆ is 1.3 times higher than that of Cs₂AgSbBr₆ at the same film thickness, illustrating that Cu(I)-based double perovskites have a higher conversion efficiency than Ag(I)-based ones.

When aiming for optoelectronic applications of Cs₂CuSbCl₆ and Cs₂CuSbBr₆ DPNCs, high-quality films are a key requirement. Compared to organic–inorganic hybrid perovskite materials (such as MAPbI₃), it is more challenging to adopt the normally utilized solution-processing method to fabricate high-quality HDP films because of the lower solubility of the all-inorganic precursors and the higher formation temperatures.³³ In contrast, the solution processability of DPNCs offers a facile and effective process to obtain high-quality HDP films. Here, thin films of Cs₂CuSbCl₆ and Cs₂CuSbBr₆ were fabricated via a simple direct deposition of corresponding DPNC solutions by spin-coating [Fig. 7(a)]. The purified DPNCs were redispersed in octane at a concentration of about 70 mg/ml for the film deposition (details in Sec. II). MA and EA were used as washing solvents to initiate solid-state ligand exchange, which can avoid the re-dissolving issue as a second layer was deposited in parallel to enhance the electronic coupling between DPNCs. Thus, different film thicknesses could be obtained by a layer-by-layer deposition process. As shown in Fig. 7(b), the SEM image of a Cs₂CuSbCl₆ DPNC film without being exposed to a solid-state ligand exchange process is quite indistinct due to the large amount of ligands left in the film. However, after washing the Cs₂CuSbCl₆ DPNC film with either MA or EA, distinct DPNC films can be observed, which are very smooth and dense [Figs. 7(c) and 7(d), Fig. S12]. A high-quality Cs₂CuSbBr₆ DPNC film can be obtained by the same process (Fig. S13).

We carried out XPS and UPS measurements for both Cs₂CuSbCl₆ and Cs₂CuSbBr₆ DPNC films before and after the solid-state ligand exchange process. XPS spectra evidence the characteristic Cu(I) state at ∼932 (2p_3/2) and ∼952 eV (2p_1/2) binding energy (Fig. S14). Relative concentrations of the different elements have also been calculated (see Table S2). We find that after ligand removal (i.e., oleic acid and oleylamine), the C and O contents decrease, confirming the success of this procedure. Due to the large amounts of C, N, and O on the surfaces, a reliable derivation of the Cs:Cu:Sb:Cl or Br stoichiometry of the films from XPS data is difficult, but the off-stoichiometric compositions of the films might be due to different reasons, as already observed for the DPNCs (see above). Here, the measured element ratios of Cs₂CuSbCl₆ and Cs₂CuSbBr₆ films are, respectively, 1.5:0.4:0.9:6 and 1.9:0.5:0.7:6 (Cs:Cu:Sb:Cl or Br). The semi-quantitative analyses also show that these compositions remain basically unchanged after ligand removal (e.g., the Cs:Cl, Sb:Cl, and Cu:Cl ratios remain relatively similar—see Table S2).

XPS and UPS data can also be combined to position the main energy levels of the semiconducting Cs₂CuSbCl₆ and Cs₂CuSbBr₆ films, that is, the Fermi level, the VBM, and the CBM. The energy cut-off (E_cut−off) has been deduced from UPS data (Fig. S15). It leads to the work function of the sample, i.e., the energy difference between the vacuum level and the Fermi level of the sample. Unfortunately, clear valence band edges of the sample cannot be distinguished by UPS, as a significant density of states is only observed above 2.0 eV. UPS is a highly surface sensitive technique, and the presence of organic ligands, even after partial removal, may hinder the analysis of the DPNC itself. To overcome this issue, the valence band edge has also been acquired by XPS. The kinetic energy of the valence band photoelectrons is higher in XPS than in UPS and results in a deeper depth of analysis. The valence band edge of Cs₂CuSbCl₆ was estimated at 1.2 eV below the 0.0 eV binding energy reference Fermi level. Similarly, this energy difference has been estimated to be 0.7 eV for Cs₂CuSbBr₆, in agreement with a lower indirect bandgap discussed above. This 0.5 eV offset can be explained by both the substitution of Cl 3p by Br 4p that have higher energy states and by a change of the doping. If the doping level of the sample is modified, by, for instance, anion-exchange, the Fermi level of the sample will shift, and if this reference level shifts, all the spectra should shift accordingly. The Cs 3d_5/2 and the Cu 2p_3/2 states are shifted by 0.2 and 0.1 eV to lower energy after the ion exchange. This indicates that the Fermi level of Cs₂CuSbCl₆ has been shifted closer to the valence band by roughly 0.2 eV after its conversion to Cs₂CuSbBr₆. The remaining 0.3 eV (difference) is due to the substitution of the Cl 3p orbitals by Br 4p orbitals. Looking at a theoretical intrinsic position of the Fermi level (E_F,in), we find that the two materials are n-doped with an energy difference of 0.35 and 0.25 eV (=E_F-E_F,in) for Cs₂CuSbCl₆, and Cs₂CuSbBr₆, respectively. Using the valence band edge obtained by XPS, the work function obtained by UPS, and the optical bandgap obtained by UV–Vis absorption, the resulting electronic structure diagram can be drawn (Fig. S16). Interestingly, we note that the partial ligand removal has shifted the Fermi level of the DPNCs. For instance, the E_F − E_F,in energy difference of Cs₂CuSbCl₆ is reduced from 0.35 eV to 0.05 with this procedure. Hence, the X-ion exchange from Cl to Br leads to more intrinsic materials.

In terms of ambient stability, Cs₂CuSbCl₆ DPNCs covered with a full ligand shell can be uniformly dispersed in hexane and octane as solutions after 520 days (Fig. S17). However, the purified Cs₂CuSbCl₆ DPNCs (solid-state) started to decompose after 8 days in ambient conditions (Fig. S18a). Cs₂CuSbCl₆ DPNC thin films washed with MA show better stability compared to the ones washed with isopropanol, which apparently has a better ligand-removing ability due to its higher dielectric constant (Fig. S19). The Cs₂CuSbCl₆ DPNC films annealed at 140 °C for 5 min both in ambient conditions and in a N₂-filled glovebox show evidence of decomposition (Fig. S19). More stability test details can be found in the supplementary material.

In summary, we report a facile and fast synthesis method for Cs₂CuSbCl₆ DPNCs by using a modified LARP method at ambient conditions. A less toxic and easily removable solvent, methanol, was chosen for solvating precursor salts. We demonstrate that the polarity of anti-solvents and the ratio of OA/OAm ligands have a strong effect on the size and morphology of the resulting DPNCs. Cs₂CuSbBr₆ DPNCs with a narrow bandgap of 0.9 eV were successfully developed for the first time by employing an anion-exchange reaction in solution. Both [CuCl₆]⁵⁻ and [CuBr₆]⁵⁻ octahedra can be stabilized in these double perovskites at the nanoscale with the assist of organic ligands, which we expect to enable the development of other predicted-metastable (as bulk) Cu(I)-based double perovskite materials to enrich the double perovskite family and to benefit the screening of high performance lead-free HDPs materials for optoelectronic applications. Taking advantage of the solution processability of DPNCs, smooth and dense Cs₂CuSbCl₆ and Cs₂CuSbBr₆ DPNC films were successfully fabricated. Optimization of the ligand density and developing new types of ligands, for example, conductive ligands, for these “unstable” DPNCs to balance the charge carrier mobility and phase stability need further effort and are expected to open new vistas in the development of optoelectronic applications of lead-free double perovskite materials.

The supplementary material is available free of charge at Links.

The authors thank Dr. Steffen Schmidt and Dr. Markus Döblinger for performing the SEM and TEM investigation, respectively. The authors acknowledge the Bavarian research network Solar Technologies go Hybrid (SolTech), the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) Excellence Cluster e-conversion (EXC 2089/1–390776260), and the DFG focus program SPP 2196 for funding of Projects Nos. 423746744 (C.M., T.M.) and 424707803 (S.W., R.H., T.B.). D.H. and H.E. gratefully acknowledge the Gauss Center for Supercomputing e.V. (www.gauss-centre.eu) for funding this project by providing computing time on the GCS Supercomputer SuperMUC-NG at Leibniz Supercomputing Center (www.lrz.de). S.W. acknowledges support from the China Scholarship Council.

The authors have no conflicts to disclose.

Shizhe Wang: Investigation (lead); Methodology (lead); Writing – original draft (lead). Dan Han: Investigation (equal); Methodology (equal); Writing – review & editing (equal). Clément Maheu: Investigation (equal); Writing – original draft (equal). Zehua Xu: Investigation (equal); Writing – original draft (equal). Alexander Biewald: Investigation (equal). Hannah Illner: Investigation (equal). Rik Hooijer: Investigation (equal). Thomas Mayer: Writing – review & editing (equal). Achim Hartschuh: Writing – review & editing (equal). Hubert Ebert: Writing – review & editing (equal). Thomas Bein: Supervision (equal); Writing – review & editing (equal).

The data that support the findings of this study are available within the article and its supplementary material.