Bright Blue and Green Luminescence of Sb(III) in Double Perovskite Cs2MInCl6 (M = Na, K) Matrices.

The vast structural and compositional space of metal halides has recently become a major research focus for designing inexpensive and versatile light sources; in particular, for applications in displays, solid-state lighting, lasing, etc. Compounds with isolated ns2-metal halide centers often exhibit bright broadband emission that stems from self-trapped excitons (STEs). The Sb(III) halides are attractive STE emitters due to their low toxicity and oxidative stability; however, coupling these features with an appropriately robust, fully inorganic material containing Sb3+ in an octahedral halide environment has proven to be a challenge. Here, we investigate Sb3+ as a dopant in a solution-grown metal halide double perovskite (DP) matrix, namely Cs2MInCl6:xSb (M = Na, K, x = 0-100%). Cs2KInCl6 is found to crystallize in the tetragonal DP phase, unlike Cs2NaInCl6 that adopts the traditional cubic DP structure. This structural difference results in distinct emission colors, as Cs2NaInCl6:xSb and Cs2KInCl6:xSb compounds exhibit broadband blue and green emissions, respectively, with photoluminescence quantum yields (PLQYs) of up to 93%. Spectroscopic and computational investigations confirm that this efficient emission originates from Sb(III)-hosted STEs. These fully inorganic DP compounds demonstrate that Sb(III) can be incorporated as a bright emissive center for stable lighting applications.

Light-emitting ns2-metal halide-based materials have been the subject of continued interest over several decades, owing to the facile synthesis by wet and solid-state methods as well as their exceptional electronic and optical properties. The latter include narrowband and broadband emission for display technologies,[1−4] lasing,[5−8] quantum light sources,[9] solid-state lighting,[10] scintillators,[11] remote thermography,[12] and luminescent solar light converters and concentrators.[13,14]

The vast variety of luminescence characteristics arises from structural tunability. Metal halide polyhedral units as building blocks can form extended structures (e.g., APbX3, A2AgBiX6, where A = monovalent cation, X = Cl, Br, I) with predominantly Wannier–Mott-type excitons,[15−17] layered structures with both Wannier–Mott and Frenkel excitons,[18,19] and fully isolated 0D metal halides exhibiting characteristic broadband emission from self-trapped excitons (STEs) of Frenkel-type.[20] For generating STE emission in the visible region, the most suitable ns2-metals include Ge(II), Sn(II), Sb(III), and Te(IV), whereas Pb(II) and Bi(III) systematically display higher energy emission, often in the UV region.[21−23] In this work, we have concentrated our attention on the broad STE emission from Sb(III) due to its oxidative stability, in great contrast to the unstable oxidation states of Ge(II) and Sn(II). Organic–inorganic Sb(III) halide hybrids with various organic cations exhibit broadband and strongly Stokes-shifted emission in the yellow-red region with near-unity photoluminescence quantum yield (PLQY) at room temperature (RT).[24−27] In contrast, the known fully inorganic Sb(III) compounds (e.g., Cs3Sb2X9, Cs2NaSbX6)[28,29] are nonemissive at RT, although they are much more desirable due to their higher thermal stability. Therefore, to achieve an efficient RT STE of Sb3+, we adopt an alternative strategy, namely, doping ns2-metal ions into a suitable inorganic halide matrix.[20,30] In this case, the doping level serves as a useful tuning knob for maximizing the emission intensity or preventing self-quenching at high concentrations. For example, the Sb3+ ions in Cs2NaSbCl6 fully quench each other at RT,[29] requiring the use of diluted Sb3+ systems for RT emission. The most suitable fully inorganic matrix has to provide an Oh halide environment for Sb3+ and at the same time possess a wide optical band gap to prevent self-absorption. Doping on an isovalent host site is also advantageous, although examples of aliovalent Sb3+ doping are known.[31] The matrix satisfying these requirements can be found among the halide elpasolites, also known as halide double perovskites (DPs).[32] The structure of cubic elpasolites can be viewed as derived from a cubic perovskite structure AMIIX3 by the transmutation of a pair of MII sites into distinct MI and MIII sites, resulting in an A2MIMIIIX6 stoichiometry. Although elpasolites are 3D structurally, in some compositions (where MI is an alkali halide metal), the orbital overlap between the nearest [MIIICl6] octahedra is broken by the adjacent [MICl6] octahedra, resulting in reduced electronic dimensionality.[33] The emission of Sb3+ dopant in some bulk halide DP (Cs2NaMX6, where M = Y, Sc, La and X = Cl, Br) was reported by Oomen et al. in the late 80s.[34,35] These compounds exhibited emission centered in the range from 440 to 492 nm, depending on the size of the space available for the Sb3+ ion. In this study, we have selected In3+ as a trivalent cation due to its oxidative and photostability, as well as the large optical band gap of the resulting DP. In addition to Na+, the larger K+ was selected to fine-tune the Sb3+ emission. During the preparation of this manuscript, Zeng et al. also reported on Sb-doped Cs2NaInCl6.[36]

When suitably small amounts of Sb3+ are doped into Cs2NaInCl6, the products retain the original structure, as confirmed by powder XRD (Figure S1). The peak positions shift linearly toward lower angles with increasing Sb3+ content, indicating a larger unit cell in agreement with Vegard’s law (Figure S1b,c). This increase is the opposite of what is expected from the stated ionic radii of Sb3+ (76 pm), which is smaller than In3+ (80 pm),[37] but is consistent with the larger lattice constant of Cs2NaSbCl6 (10.7780 Å) relative to that of Cs2NaInCl6.[38] This disparity can be ascribed to the bond-destabilizing effect of the ns2 lone pair, as described for Bi3+ by Shannon.[37] To maintain an average cubic symmetry in Cs2NaSbCl6, the stereoactivity of the lone pair must lead to a highly dynamic distortion of the SbCl63– octahedra.[29] The bond-length increase also confirms that Sb3+ substitutes into the In3+ position of Cs2NaInCl6, as any substitution of Na+ by the much smaller Sb3+ ions would decrease the size of the unit cell. When 60% or more of Sb3+ is added during synthesis, weak peaks characteristic of the Cs3Sb2Cl9 impurity phase start to appear.

Contrary to the sodium-containing DP, the Cs2KInCl6:xSb (x = 0, 1, 5, 10, 15, 20, and 40%) compounds could be prepared only as fine powders, which were not suitable for single-crystal XRD. Powder XRD patterns of Cs2KInCl6 show doubled peaks near the strong reflections of the Fm3̅m cubic DP structure, indicating a structure of lower symmetry (Figure a,c). LeBail pattern matching using GSAS-II[39] (Figure S2b) reveals that these compounds possess tetragonal symmetry, with the space group I4/m and unit cell parameters a = b = 16.9732(2) Å and c = 10.9937(1) Å. This corresponds to a distorted DP structure (Figure S2a) due to the large size of the K+ cation relative to the B-site cavity, as recently observed for the large cations In+ and Tl+ in the mixed-valent compounds CsInCl3 and CsTlCl3.[28,40] Rietveld refinement of the structure was attempted with the CsInCl3 structure as a starting point but did not converge due to the large unit cell of this supercell structure. The tetragonal Cs2KInCl6 structure is reported in this study for the first time. In theoretical studies, Cs2KInCl6 is usually assumed to adopt a cubic structure; however, this has been not experimentally demonstrated.[33] Previously, only the monoclinic phase of Cs2KInCl6, isostructural to Cs2KBiCl6, was reported.[41,42]

Figure 1

(a) Comparison of Cs2NaInCl6 and Cs2KInCl6 crystal structures, experimental and simulated XRD patterns for (b) Cs2NaInCl6 and (c) Cs2KInCl6. (d) Images of Cs2NaInCl6:1%Sb crystals under visible light (Vis) and 365 nm UV excitation.

Extended exposure of Cs2KInCl6:xSb to the acidic environment results in the incorporation of water molecules and the presence of the impurity phase Cs2InCl5:xSb·H2O (Figure S2c). This complicates the interpretation of the powder XRD, as there is no clear trend in the lattice parameters with increasing Sb3+ content. This phase conversion also alters the optical properties as the maximum of the emission shifts toward longer wavelengths, while the maximum of the excitation band remains unchanged (Figure S2d,e). Cs2NaInCl6:xSb samples exhibited no degradation upon storage in air for a few months, whereas with Cs2KInCl6:xSb, yellow-luminescent grains started to appear after a few weeks in air, indicating that the tetragonal structure may be metastable (Figure S3). Similar instability was observed for isostructural CsInCl3.[28]

Sb-doped Cs2NaInCl6 and Cs2KInCl6 samples exhibit broadband blue and green PL (Figure a,b), respectively, while the undoped Cs2NaInCl6 and Cs2KInCl6 compounds exhibited no visible PL at RT (Figure S4). This is in contrast to the study of Zeng et al.,[36] who reported a PL band with a maximum at 445 nm even for undoped structures. Sb-doped compounds also exhibit new bands in the absorption spectra corresponding to Sb states within the band gap of the Cs2MInCl6 matrix (Figure S5). As the PL intensities are low for the samples with high Sb3+ content due to self-quenching, we mainly analyzed the samples with Sb3+ concentration below 20%. For ns2 ions such as Sb3+, the ground state is denoted by the 1S0 atomic term, whereas the excited state (sp) splits into four energy levels, namely, 1P1, 3P0, 3P1, and 3P2. The 1S0–1P1 transition is allowed, and the 1S0–3P1 transition is partially allowed due to spin–orbit coupling for heavy atoms, while the 1S0–3P2 and 1S0–3P0 transitions are totally forbidden at the electric dipole transition level. The characteristic excitation band (commonly denoted as the A-band) with two maxima at 320 and 335 nm can be assigned to be the 1S0–3P1 transition of Sb3+ (PLE maps are presented in Figure S6a,b). The split in the PLE band has been ascribed to a dynamic Jahn–Teller distortion in the excited state.[34] The PL emission situated in the 380–700 nm range (with a maximum near 445 nm for Cs2NaInCl6:5%Sb and 495 nm for Cs2KInCl6:5%Sb) can be ascribed to the 3P1–1S0 transition, similar to the case of Cs2NaMCl6 (M = Sc, Y, La) matrices.[34] The Stokes shift of Sb3+ luminescence, defined as the energy difference between the maximum of the excitation and emission bands, is equal to 110 and 160 nm for Cs2NaInCl6:5%Sb and Cs2KInCl6:5%Sb, respectively. The shape and maxima of the PL excitation (PLE) and PL spectra remain the same for different amounts of Sb3+ ions in the crystal structure (Figure S7a,b). The only exception is in the PL spectrum of Cs2KInCl6:1%Sb, where a shoulder was observed at low energy, which seems to originate from the Cs2InCl5·H2O:1%Sb impurity.

Figure 2

Typical PL (under 320 nm UV excitation) and PLE of (a) Cs2NaInCl6:5%Sb and (b) Cs2KInCl6:5%Sb. (c) PLQY dependence on the concentration of the Sb3+ dopant.

As demonstrated in Figure c, the amount of Sb3+ dopant strongly influences PLQY. The highest PLQYs were obtained for the Cs2NaInCl6:1%Sb and Cs2KInCl6:5%Sb structures (λexc = 320 nm) and were equal to 82 and 93%, respectively. A further increase of the doping level caused the self-quenching effect. Figure a shows the time-resolved PL decay curves of Cs2NaInCl6:5%Sb and Cs2KInCl6:5%Sb measured at 445 and 495 nm, respectively, at RT (λexc—320 nm). The curves at all doping levels up to 20% can be fitted biexponentially (see Table S6 and Figure S7c). For Cs2NaInCl6:1%Sb, we obtained a short-lived PL lifetime of 96 ns with a percentage of 12% and a long-lived PL lifetime of 1029 ns with a percentage of 88%. As the concentration increases, we can observe a gradual decrease of both short- and long-lived PL lifetimes, along with a higher proportion of the faster component (Table S6). According to Reisfeld et al.,[43] two recombination routes are possible for the excited state in Sb3+:3P0–1S0 recombination with a faster decay rate and 3P1–1S0 with a slower decay rate. Similar behavior is observed for Cs2KInCl6:xSb samples (see Table S6 and Figure S7d).

Figure 3

(a) Time-resolved PL decay curves of Cs2NaInCl6:5%Sb and Cs2KInCl6:5%Sb measured at 445 and 495 nm, respectively, at RT (λexc—320 nm) and the corresponding time constants of the biexponential fit. (b) Experimental dependence of the PL peak broadening with temperature for Cs2NaInCl6:1%Sb and Cs2KInCl6:5%Sb. The experimental data are fitted with a model (solid line), from which the electron–phonon coupling strength (γLO) and optical phonon energies (ELO) are obtained.

To better understand the influence of exciton–phonon coupling, we have analyzed the PL line broadening with temperature. This has been done by first extracting the PL linewidth by fitting the peak with exponentially modified Gaussian, and then plotting the obtained full width at half maximum against temperature. The resulting data can be fit with the Rudin model[44]The Γ0 is a temperature-independent inhomogeneous broadening, which is related to scattering due to disorder and imperfections. The second and third terms represent homogeneous broadening, resulting from scattering from acoustic and optical phonons. The last term accounts for the scattering of ionized impurities and does not contribute in the case of Cs2MInCl6:xSb. Electron–phonon coupling is proportional to the occupation of the phonon energy states and can be described by Bose–Einstein distributionwhere the constant γLO represents the exciton–optical phonon coupling strength and ELO is the energy of the optical phonon involved in homogeneous PL linewidth broadening. For the acoustic phonons, the dependence is linear: Γac = γacT. For the fitting, we have taken into account only the first three terms. Figure S8 shows the evolution of the PL line shape with temperature. Figure b summarizes the results obtained from fitting of PL and lists fit values related to the optical phonon coupling strength and energy. In the case of Cs2NaInCl6:xSb, the energies of phonons involved in PL broadening are lower (17.7 ± 6 meV) than those for Cs2KInCl6:xSb (23.8 ± 7 meV). The same trend is observed in the coupling constant values (69 ± 1 vs 92 ± 2 meV). Interestingly, the 17.7 meV phonon energy for Cs2NaInCl6:xSb falls in the region of the phonon band structure with states originating mainly from the Na–Cl bonds. The inhomogeneous broadening is also higher in the case of Cs2KInCl6:xSb. This analysis reflects the difference in the crystal structure, where the lower symmetry of the potassium analogue influences the optical properties of the impurity ion Sb3+.

DFT calculations were utilized to analyze the electronic structure of Cs2MInCl6:xSb and rationalize the STE-derived optical properties. The electronic density of states (DOS) and projected molecular orbitals have been calculated with the CP2K package at the DFT/PBE level of theory using the model system comprising a Cs2MInCl6 2 × 2 × 2 supercell with one Sb atom per supercell (about 3 mol %). It has been previously discussed that Cs2MInCl6 possesses a direct but parity-forbidden band gap (2.73 eV) at the Γ-point.[33] This results in a very low oscillator strength of the edge-to-edge transition, rendering undoped Cs2MInCl6 weakly absorbing and nonluminescent. In Cs2MInCl6, the highest occupied molecular orbital (HOMO) and the lowest unoccupied molecular orbital (LUMO) comprise Cl-p states and In-s + Cl-p states, respectively. The electronic structure is not well dispersed and electronic bands are flat (see the Supporting Information for computational details). In a recent study by Zeng et al., STE emission from Sb-doped Cs2NaInCl6 was explored with DFT and rationalized as an effect of breaking the parity-forbidden transition and modulation of the density of states due to Sb3+ doping.[36]

The analysis of the DOS of the Sb-doped Cs2NaInCl6 with a cubic structure shows that the SbCl6 HOMO appears above the host valence band while the SbCl6 LUMO appears in a gap deep in the host conduction band (Figures a and S9). The oscillator strength of the SbCl6 HOMO–LUMO transition is much higher than those of the matrix, and therefore both optical excitation and emission are occurring between SbCl6 states. The SbCl6 HOMO has Sb-s and Cl-p character, whereas the LUMO consists of Sb-p and Cl-p orbitals. This can also be confirmed by calculating the SbCl6 HOMO–LUMO difference, 3.92 eV, which is close to the experimental PLE peak at 12 K of 3.75 eV (Figure S10). For Sb-doped Cs2KInCl6, which has a tetragonal structure, the predicted SbCl6 HOMO–LUMO gap is lower (3.73 eV) and better agrees with the experiment (Figure c). The calculations confirm that the observed luminescence mechanism does not involve states from the DP host, but both HOMO and LUMO are localized on the dopant sites, namely, SbCl6 octahedra.

Figure 4

(a) Electronic structure of Cs2NaInCl6 with one Sb3+ ion at a Γ point in a 2 × 2 × 2 supercell, computed at a DFT/PBE level. (b) Projected atomic HOMO (orange) and LUMO (yellow) of the host material and SbCl6 centers; the HOMO and the LUMO of the host consist of Cl-p and In-s + Cl-p states, respectively, whereas the HOMO and the LUMO of the dopant comprise Sb-s + Cl-p and Sb-p + Cl-p, respectively. (c) Electronic structure of Cs2KInCl6 with one Sb3+ ion at a Γ point in a 1 × 1 × 2 supercell, computed at the DFT/PBE level.

In summary, we have designed and synthesized Cs2MInCl6:xSb (M = Na, K) phosphors with blue and green emission and PLQYs of 82 and 93%, respectively. Experimental and computational studies reveal that the optical properties originate from STEs localized on SbCl6 centers. We demonstrate that the emission maximum of the Sb3+ dopant can be tuned not only by exchanging the MIII cation, but also the MI cation. Replacement of the Na+ cation by the larger K+ cation causes structural changes due to the size of the cation, resulting in a shift from cubic to tetragonal symmetry. This change is reflected in the increase of the Stokes shift and electron–phonon coupling. We anticipate that such fully inorganic, thermally robust, and bright emitters will find applications in solid-state lighting as well as emerging niches such as remote thermography, scintillation, neutron detection, cathodoluminescence, etc.

Experimental Section

The list of chemicals is available in the Supporting Information.

Synthesis of Cs2NaInCl6:xSb Powder

Cs2NaInCl6:xSb was synthesized from stoichiometric quantities of CsCl, NaCl, InAc3 and SbAc3 precursors in HCl. Briefly, 0.6 mmol of CsCl and 0.3 mmol of NaCl powders were dissolved in 4 mL of HCl under stirring in an 8 mL vial at 100 °C. Separately, stoichiometric amounts of SbAc3 and InAc3 were dissolved in 1 mL of HCl and heated at 100 °C. When all of the precursors were dissolved, both solutions were mixed together and immediate precipitation of the colorless product was observed. The as-obtained product was separated by filtration and dried in air at 50 °C for 6 h. Reaction yield: 79%.

Synthesis of Cs2KInCl6:xSb Powder

Cs2KInCl6:xSb was synthesized in the same way as Cs2NaInCl6:xSb, with the only difference being that 0.6 mmol of CsCl and 0.15 mmol of K2CO3 were dissolved in 1 mL of HCl at 100 °C, and stoichiometric amounts of SbAc3 and InAc3 were dissolved in 1 mL of HCl in another vial. When all of the precursors were dissolved, both solutions were mixed together and immediate precipitation of the fine powder was observed. The resulting material was collected and filtered immediately after formation. Leaving the product in the mother solution longer results in the formation of a stable Cs2InCl5x H2O:xSb phase. Reaction yield: 59%.

Synthesis of Cs2NaInCl6:xSb Single Crystals

Cs2NaInCl6:xSb was synthesized from stoichiometric quantities of CsCl, NaCl, InAc3, and SbAc3 precursors in HCl. Briefly, 0.6 mmol of CsCl, 0.3 mmol of NaCl, and stoichiometric amounts of SbAc3 and InAc3 were dissolved in 6 mL of HCl under stirring in a 12 mL vial and heated at 100 °C. When all of the precursors were dissolved, the temperature was decreased to room temperature and small crystals grew overnight. To ensure the total removal of surface-adsorbed ions, the tiny crystals were rinsed with fresh HCl. The crystalline product was dried in the air at 50 °C for 6 h.

Characterization

Powder XRD patterns were collected in transmission mode (Debye–Scherrer Geometry) with an STADI P diffractometer (STOE & Cie GmbH) equipped with a curved Ge(111) monochromator (Cu Kα1 = 1.54056 Å) and a silicon strip MYTHEN 1K detector (Fa. DECTRIS). For the measurement, the ground powder was placed between adhesive tape, with the exception of Cs2KInCl6, which was measured in a Mark-tube capillary of 0.1 mm diameter. Single-crystal XRD measurements were conducted on a Bruker Smart Platform diffractometer equipped with an Apex I CCD detector and a molybdenum (Mo Kα = 0.71073 Å) sealed tube as an X-ray source. Crystals were tip-mounted on a micromount using paraffin oil. The data were processed with Oxford Diffraction CrysAlis Pro software, and structure solution and refinement were performed with SHELXS and SHELXL respectively, embedded in the Olex2 package.[45,46] For material composition analysis, we used XRF spectroscopy. The Amptek Complete XRF Experimenter’s Kit was used with an Ag Mini-X tube as a source, X-123 as a detector, PX-5 as an amplifier and digitizer, and XRS-FP2 quantitative analysis software to recalculate the measured XRF peak intensity into the material composition. The multiple standards calibration method was used with reference samples of known composition.

A Fluorolog iHR 320 Horiba Jobin Yvon spectrofluorometer equipped with a PMT detector was used to acquire steady-state photoluminescence (PL) and PL excitation (PLE) spectra from solutions. Absolute PL quantum yield (QY) of solutions were measured with a Quantaurus-QY Absolute PL quantum yield spectrometer from Hamamatsu. Time-resolved photoluminescence (TR-PL) measurements were performed using a time-correlated single-photon counting (TCSPC) setup, equipped with an SPC-130-EM counting module (Becker & Hickl GmbH) and an IDQID-100-20-ULN avalanche photodiode (Quantique) for recording the decay traces.

Computational Details

Calculations were carried out at the density functional theory level as implemented in the cp2k quantum chemistry code. A doubled 2 × 2 × 2 unit cell containing 320 atoms with one Sb atom on an In position for cubic Cs2NaInCl6:xSb and a 1 × 1 × 2 unit cell containing 200 atoms with one Sb atom on an In position for tetragonal Cs2KInCl6:xSb was constructed. A mixed plane-wave and Gaussian basis set approach was used to describe the wave function and electronic density, respectively. The kinetic energy cutoff of the plane-wave basis was set to 400 Ry, while a double-ζ basis set plus polarization functions was employed to describe the molecular orbitals. The density of states (DOS) and emission and excitation energies were calculated using the Perdew–Burke–Ernzerhof (PBE) exchange–correlation functional. Scalar relativistic effects have been accounted for using effective core potential functions in the basis set. Spin–orbit coupling effects were not included. Calculations with lattice relaxation for the ground state geometry optimization were performed to account for statistical disorder in the experimental crystal structures. Unit cell parameters were taken from experimental data and not relaxed, whereas atomic coordinates were optimized until the force reached 0.023 eV/Å.

For the band structure calculations, the Vienna ab initio simulations package (VASP) was used. The projector augmented wave (PAW) potentials for atoms were used.[47,48] For the generalized gradient approximation (GGA), the Perdew–Burke–Ernzerhof exchange–correlation functional (PBE) was used.[49] A single cell comprising 320 atoms with a kinetic energy cutoff of 520 eV and automatic γ-centered mesh of 1 × 1 × 1 k-points were used for geometry optimization. Ionic minimization was performed until all forces on atoms were smaller than 0.001 eV/Å. A mixture of the blocked Davidson iteration scheme and the subsequent residual minimization scheme-direct inversion in the iterative subspace (RMM-DIIS) algorithm was used for electronic optimization. Spin–orbit coupling (SOC) was not taken into account due to the known self-interaction error that increases the valence band maximum.[50] The band structure was calculated along a high-symmetry k-point path according to Bradley and Cracknell.[51]