Bridging two worlds: (DABCO-H)CuKI₃ a hybrid copper iodide phosphor with a perovskite structure

Synthesis and crystal structure

The ligand (DABCO-H) was prepared as its bromide salt in almost quantitative yield by protonation of DABCO with HBr (molar ratio 1:1) in ethanol. Compound 1 was then isolated as single crystals by reactants slow diffusion technique.

Compound 1 crystallizes in the non-centrosymmetric chiral orthorhombic space group P2₁2₁2₁. Table S1 (Supplementary Material) contains the main crystal data and structure refinement details. The experimental and simulated powder X-ray diffraction patterns of 1 are in good agreement (Fig. S1, Supplementary Material), proving the phase purity of the as-prepared product. The asymmetric unit, Fig. 1a, contains one copper, three iodine, one monoprotonated (DABCO-H), and one potassium cation. A stereochemical analysis by means of continuous shape measures [8] (Tables S2–3, Supplementary Material) shows that the Cu centre is tetrahedrally coordinated by three iodide ions and one nitrogen from DABCO-H, while the K⁺ ion has an octahedral distorted geometry due to six surrounding iodide ions. The Cu1–N1 distance is 2.1180(4) Å, while Cu–I distances are longer: 2.6678(5) Å, 2.6828(7) Å and 2.6612(7) Å when considering I1, I2 and I3, respectively. The six distances K–I are in the range 3.496–3.577 Å, Fig. 1b. Each iodide ion adopts a μ₃-I bridging mode (among one Cu⁺ and two K⁺ ions) and the system forms a 3D network. A closer inspection evidences that each KI₆ octahedron shares the corners leading to a perovskite structure. As shown in Fig. 1c, eight K⁺ ions lay on the vertices of a distorted cube, where each edge contains one iodide ion, and the ligand DABCO-H is inside the cubo-octahedral cavity. According to the ABX₃ perovskite general formula, the systems can be described as (DABCO-H)CuKI₃, where the potassium cation occupies the B site and (DABCO-H) the A site. DABCO derivatives have already been successfully employed as A site to prepare metal-free perovskites [9], [10], [11], [12], [13], [14] and other lead-free HOIPs [15], [16], [17], [18]. Alternatively, the structure can be seen considering the cube formed by eight DABCO-H units with a KI₆ octahedron in the middle, Fig. 1d. This view allows to easily evidence that the tetrahedron of the CuI₃N unit is sharing a face with the KI₆ octahedron (K···Cu distance, 3.4397(13) Å) and that it lays in one of the eight interstitial tetrahedral sites. Both the K···K and (DABCO-H)···(DABCO-H) distances along the cube edges of Fig. 1c and d are 7 Å circa (Supplementary Material, Figs. S2a, b). Potassium as B site is not very common and it has been found, for instance, using nitrate [12], iodate [9] and halides [16, 19, 20] anions. Recently, a closely related HOIP, (R-3AQ)KI₃ where R-3AQ²⁺ is (R)-(+)-3-aminoquinuclidin, has been reported [20]. In this case, a pseudo-cubic perovskite framework composed of KI₆ octahedra with slight distortions hosts the doubly protonated R-3AQ²⁺ cation in the cubo-octahedral cavity to achieve the charge balance. In the case of compound 1, the charge neutrality is ensured by the presence of the monoprotonated DABCO-H coordinated to a Cu(I) site. The presence of the Cu–N coordination bond directly connects the inorganic framework to the organic part, and this also implies that the cationic ligand is more constrained with a precise orientation inside the cavity even at room temperature. On the contrary, often HOIPs single crystals at room temperature give highly disordered organic cations that lead to an ordered structure only lowering the temperature.

Fig. 1:

XRD structure. a) Asymmetric unit of compound 1. b) Octahedral coordination of potassium ion and K–I distances (Å). c) View of the perovskite structure with the A site (DABCO-H) inside the cubo-octahedral cavity formed by six KI₆ octahedrons. d) View of the perovskite structure with the B site (K⁺ ion) in the middle of a cube formed by eight A sites (DABCO-H) and highlighting the Cu ion occupying an interstitial tetrahedral site. The thermal ellipsoids are set at a 50 % probability level; in b and c, DABCO-H hydrogen atoms are omitted for clarity and only the protonated N–H is showed; § = ½ − x, 1 − y, −½ + z; # = 1 − x, ½ + y, ½ − z; * = −x, ½ + y, ½ − z.

Goldschmidt, in 1926, introduced the concept of tolerance factor (t) [21], eq. (1), a geometrical parameter to evaluate the structural stability of ideal perovskites with general formula ABX₃ and then widely used to predict the stability of cubic perovskites and which structure-type is formed. The tolerance factor is derived only considering the chemical formula and the ionic radii, r _i , of each ion (A, B, X). For instance: t for cubic structures is 0.9–1, for distorted perovskites is 0.8–0.89, and when t < 0.8 other structures such as ilmenite are formed.

This concept has been later extended also to hybrid perovskites by substituting the original A ionic radius (r _A ) with an effective radius for molecular cations (r _Aeff ). The latter is calculated according to the method proposed by Cheetham et al. [22] as the sum between r _mass , i.e. the distance between the centre of mass of the molecule (DABCO-H) and the atom with the largest distance to it (excluding hydrogen atoms) and r _ion , i.e. the corresponding ionic radius of this atom (in this case N). DABCO-H has a r _Aeff value of 2.728 Å (r _mass  = 1.268 Å and r _ion for N³⁻ is 1.460 Å). The effective ionic radii from Shannon [23] have been used for K⁺, I⁻ and N³⁻. According to eq. (2), the tolerance factor for (DABCO-H)CuKI₃ is of 0.974 supporting the stability of the perovskite structure. As a matter of fact, most of the reported ABX₃ hybrid perovskites show a cubic structure when tolerance factors are between 0.81 and 1.01 [22]. It has to be noted that in the calculation of t for compound 1, Cu⁺ ion was not considered since, occupying an interstitial site, it introduces only a slightly longer distance between K⁺ ion and DABCO-H centroid along the Cu1–N1 bond. The mean value of distances between K⁺ ions and the centroid of the DABCO-H is 6.065 Å, Fig. S2c (Supplementary Material), where the distance along the Cu1–N1 bond direction (6.833 Å) is 11 % longer compared to the average value.

X-ray photoelectron spectroscopy (XPS)

Compound 1 was further characterized by X-ray Photoelectron Spectroscopy (XPS). High resolution scans for C 1s, K 2p, N 1s, I 3d and Cu 2p are shown in Fig. 2. The sample contains C, N, K, Cu and I in the expected ratio according to the formula (DABCO-H)CuKI₃. Cu 2p strongly overlaps with I 3p photoemission peak, hence the fitting of this region must carefully consider the features of both photoemission peaks, as reported in Fig. 2. From the fitting, the binding energy (BE) value of Cu 2p_3/2 is 932.1 eV and the energy separation between Cu 2p_3/2 and Cu 2p_1/2 is 20.1 eV, such values are typical of Cu(I) species. The K 2p_3/2 photoemission peak has a BE of 292.9 eV and I 3d_5/2 a BE of 619.0 eV, values very close to those found in pure KI. In the ligand, two kinds of nitrogen atoms are present: one is the protonated nitrogen and the other one is the nitrogen coordinated to copper. This also leads to two slightly different chemical environments for the carbon atoms. Figure 2 highlights these differences in the chemical environment as C1 and C2 (green and orange, respectively) along with N1 and N2 (blue and azure, respectively). XPS allows to distinguish these contributions: N 1s spectrum clearly shows two peaks (N1 at 399.5 eV, N2 at 401.8 eV), but also C 1s can be fitted considering two contributions (C1 at 285.3 eV, C2 286.2 at eV). Either in the case of N 1s and C 1s, the two synthetic peaks are in a 1:1 ratio as expected considering the compound structure. It has to be noted that even few minutes of exposure to the X-ray beam during XPS analyses induces a rapid degradation of the sample as proved by the appearance of a Cu(II) contribution at higher BE (934.6 eV for Cu 2p_3/2), and by the change of the shape of the I 3d doublets evolving towards asymmetric peaks with a tail at higher binding energies Fig. S3 (Supplementary Material).

Fig. 2:

XPS spectra and fitting of C 1s, K 2p, N 1s, I 3d and Cu 2p along with calculated and XPS experimental elements ratio for compound 1.

Photoluminescence

The absorption spectrum of sample powders of compound 1 is entirely localized in the ultraviolet region and shows the most intense absorption peak at 290 nm (Fig. 3a). The shape of the spectrum is similar to that of the closely related (R-3AQ)KI₃ HOIP, recently reported [20]. By illuminating powders of compound 1 with monochromatic radiation in the 280–350 nm wavelength range a blue luminescence, clearly visible even to the naked eye, is observed. The emission spectrum, reported in Fig. 3a, appears as a single well-defined band centred at 456 nm. Photoluminescence excitation (PLE) spectrum has an onset at ca. 340 nm indicating that the excitation process encompasses energy levels well-above the band gap that is 3.0 eV as derived from a Tauc plot (Figure S4, Supplementary Material).

Fig. 3:

Absorption and emission spectroscopies. a) UV–vis absorption, emission (λ _exc = 290 nm) and PLE (at 450 nm) spectra. b) Lifetime decay for compound 1.

The lifetime decay, Figure 3b, in the range of microseconds (6.3 μs) at room temperature indicates that compound 1 is phosphorescence emitting. The luminescence of copper iodide hybrid materials [24] generally originates mainly from the so-called cluster-centred (CC) transfer that usually gives lower energy emission in the green-yellow-red, or from a combination of halide-to-ligand charge transfer (XLCT) and metal-to-ligand charge transfer (MLCT) at higher energy with colour emission in the violet-blue. The emission maximum centred at 456 nm and the absence in compound 1 of a copper-based cluster and short Cu···Cu interactions suggest that, in this case, the luminescence is due to ligand-centred mixed XLCT and MLCT transitions.

Density of states and XPS valence band

To further study the origin of luminescence in compound 1, density functional theory calculations (DFT) were used to derive the total density of states (TDOS) and the partial density of states (PDOS). As depicted in Fig. 4a–d, the calculation results of DOS show that the highest occupied levels (corresponding to the valence band maximum region) are mainly composed of atomic states originated from Cu 3d and I 5p.

Fig. 4:

DFT calculations and valence band. a) Plot of calculated total and partial density of states (FWHM = 0.8 eV) with PBE0 functional. b) Overlap of XPS valence band and calculated total density of states (FWHM = 2.4 eV). c) Fitting of the XPS valence band. d) PDOS of the highest occupied levels as narrow lines that contribute to the convoluted Cu and I PDOS along with 3D contour plots of the molecular orbitals in this region. Displayed isosurfaces correspond to ±0.04 e^1/2 Å^−3/2 values.

The composition of the valence band is in agreement with previous reports on copper iodide hybrid phosphors [25, 26], where the inorganic moiety (Cu and I) dominates this region. A more detailed analysis of the composition of the molecular orbitals (MOs) reveals that there is a combination of both Cu 3d and I 5p (Fig. 4d). No MOs localized only on Cu or I are present. On the contrary, the atomic states of the lowest unoccupied levels (corresponding to the conduction band minimum) are mostly due to the organic ligands (C 2p and N 2p) with a smaller contribution from K 4s. The analysis of the composition of the valence and conduction bands shows a clear division: the former is mainly due to the inorganic moiety, the latter mainly to the organic one. This strongly supports the previous hypothesis that the emission originated from both MLCT and XLCT transitions. The calculated difference between the highest occupied molecular orbital (HOMO) and the lowest unoccupied molecular orbital (LUMO) is 2.82 eV, in good agreement with the experimental value of 3.0 eV (Fig. S4, Supplementary Material). The valence band region was also investigated by XPS. Figure 4b shows a comparison between the experimental XPS valence band region and the calculated total density of states with a larger broadening (FWHM = 2.4 eV). The XPS valence band region can also be well fitted (Fig. 4c), considering four peaks that account for the two iodine and the two copper contributions evidenced in Fig. 4a. The fitting was performed introducing constrains for the energy separations and for the relative intensities of the peaks according to the results obtained from PDOS calculations with lower broadening (FWHM = 0.8 eV).

Methylene blue adsorption

During the evaluation of the possible use of compound 1 for environmental remediation applications, we found that it possesses high adsorption properties towards the methylene blue (MB) dye. Hence, its adsorbent behaviour was studied in detail. Figure 5a shows the visible absorption spectra of MB at different time intervals in the presence of compound 1. The experiment was performed keeping a water MB solution in the dark and under stirring. The amount of free MB, monitored through the absorption maximum at 663 nm (Fig. 5b), drastically decreased in the early stages and then changed less dramatically at later times (Fig. 5c). Equation (3) was used to measure the amount of adsorbed dye.

Where q _t (mg g⁻¹) is the adsorption capacity, C _t (mg L⁻¹) and C ₀ (mg L⁻¹) are the dye concentration at time t and zero, respectively, in a solution of volume V (L), containing a known amount of compound 1, m (g). The experimental adsorption capacity at equilibrium, q _e , is 3.1 mg g⁻¹ and was determined as the average of the last four points (Fig. 5c). Figure 5b and c also evidence that the equilibrium is reached in almost 20 min with nearly complete dye adsorption. The equilibrium adsorption capacity can also be derived through the analysis of adsorption kinetics. We considered two possible kinetic models (Fig. 5d and e): a pseudo first order (PFO) and a pseudo second order (PSO), according to their linearized forms, eqs. (4) and (5).

Fig. 5:

MB dye adsorption studies. a) The absorption spectra (initial MB concentration: 3.2 mg L⁻¹) measured at various dye adsorption times in water. b) Absorbance vs. time monitored at 663 nm, inset: picture of the cuvette with the MB water solution at 0 and 26 min. c) MB concentration and adsorption capacity vs. time. d) Experimental data fitted with a PFO model. e) Experimental data fitted with a PSO model. f) Experimental and fitted kinetic profiles using the IPD model.

Table 1 reports the fitting values along with their figures of merit. The PSO model has the best figure of merit (R ² = 0.999) with a kinetic constant k ₂ of 0.20 g mg⁻¹ min⁻¹ and it gives a calculated value for q _e of 3.2 mg g⁻¹ in very good agreement with the experimental one (3.1 mg g⁻¹). Even if the figure of merit for the PFO model is good (R ² = 0.983), with a kinetic constant k ₁ of 0.19 min⁻¹, the derived q _e of 2.7 mg g⁻¹ differs much more from the experimental value.

The dye adsorption is dependent on diffusion and transport processes. The adsorption process can be divided in four main phases: (i) bulk transport of adsorbate molecules from the liquid to the external adsorbent surface, (ii) film transport by propagation through a boundary layer, (iii) intraparticle diffusion of adsorbate molecules from the external surface to the internal pores, (iv) adsorption of solute molecules on the adsorbent. It is well known that the second and third stages occur slowly, and both could be the rate-limiting step. For this reason, the intraparticle diffusion model (IPD, eq. (6)) proposed by Weber and Morris [27] was evaluated.

This model assumes that internal diffusion of the adsorbate in the adsorbent pore is the slowest step. In eq. (6), k _i ^IPD is the intraparticle diffusion rate constant and C is a constant proportional to the extent of the boundary layer thickness, i.e. the higher its value, the greater the boundary layer effect. The plot of q _t vs. t ^1/2 can be a single straight line or composed by a multi-segment (with two or three contributions). When the linear plot passes through the origin, the intra-particle diffusion is the rate-controlling step, with a film diffusion negligible since C = 0. If the straight line does not pass through the origin this indicates that the film diffusion is involved simultaneously along with intraparticle diffusion. Figure 5f shows a tri-phasic behaviour highlighted as Sections 1, 2 and 3 with different linear fits required for each section. This indicates that three types of diffusion influence the rate-limiting steps. The three rate constants decrease in the order k ₁ ^IPD  > k ₂ ^IPD  > k ₃ ^IPD (Table 1). As previously observed [28], the higher value for Section 1 represents a faster process where the rate-determining step is governed by intra-particle diffusion. The second slope represents a process where the dye is diffusing slower and adsorbing into the pores. The third section, where the slope is almost flat, can be ascribed to the adsorption process close to the equilibrium: the adsorbate diffuses very slowly into the pores. Since only the fitting of Section 1 passes through the origin, the Weber-Morris model suggests that only in the very first stages the intra-particle transport is the rate determining step and globally both film and intra-particle diffusion control the rate of dye adsorption.

Synthesis of (DABCO-H)Br

DABCO (336.6 mg, 3.0 mmol) was dissolved in EtOH (5 mL), resulting in a colourless solution. A solution of HBr (1.5 mL, 3.00 mmol) was then added and the reaction mixture was stirred at room temperature for 45 min. The solvent was removed under reduced pressure and the compound recrystallized from an ethanol/hexane mixture to give a white solid. Yield: 95.6 %.

Synthesis of 1

CuI (9.6 mg, 0.05 mmol) was first dissolved in a KI saturated aqueous solution (1 mL) in a 4 ml vial. Acetonitrile (1 mL) was layered and then DABCO-H (48.3 mg, 0.25 mmol) in methanol (1.5 mL) was added slowly into the vial. The resulting mixture was colourless. After one day white single crystals formed. The mixture was kept undisturbed for three days then single crystals were collected by filtration and washed with acetonitrile. Multiple vials were set up giving an average yield of 75 % (based on DABCO-H).

Adsorption experiments

To a beaker containing 20 mL of a 10⁻⁵ M solution of methylene blue (MB) 20 mg of 1 were added, in these conditions compound 1 is insoluble in water. The mixture was kept under stirring, in the dark, at room temperature. At regular intervals UV–Vis spectra were recorded.

Single crystal and powder X-ray diffraction

Single crystal XRD data were collected using an Oxford Diffraction Gemini E diffractometer, equipped with a 2K × 2K EOS CCD area detector and sealed-tube Enhance (Mo) and (Cu) X-ray sources. A suitable single crystal has been fastened to the top of a nylon loop and measured at room temperature. Data have been collected by means of the ω-scans technique using graphite-monochromated radiation and Mo K α (λ = 0.71073 Å). The detector distance has been set at 45 mm. The diffraction intensities have been corrected for Lorentz/polarization effects as well as with respect to absorption. Empirical multi-scan absorption corrections using equivalent reflections have been performed with the scaling algorithm SCALE3 ABSPACK. Data reduction, finalization and cell refinement were carried out through the CrysAlisPro software. Accurate unit cell parameters were obtained by least squares refinement of the angular settings of strongest reflections, chosen from the whole experiment. The structures were solved with Olex2 [29] by using ShelXT [30] structure solution program by Intrinsic Phasing and refined with the ShelXL [31] refinement package using least-squares minimization. In the last cycles of refinement, non-hydrogen atoms were refined anisotropically. Hydrogen atoms were included in calculated positions, and a riding model was used for their refinement. The accession number for the crystallographic data reported in this paper is CCDC 2303738. Copies of these data can be obtained free of charge from the Cambridge Crystallographic Data Centre (CCDC) via https://www.ccdc.cam.ac.uk:443/data_request/cif.

Powder XRD was performed using a Bruker D8 Advance Plus diffractometer, operating in Bragg–Brentano geometry and equipped with Cu Kα X-ray source (λ = (λ = 1.5406 Å, 40 kV, and 40 mA), collecting data in the 5–50° 2Θ range (0.02°/step, 1s/step). Calculated PXRD pattern was obtained with the CCDC Mercury software (calculation parameters: 10–50° 2Θ range, step 0.02°, peak FWHM 0.05°).

X-ray photoelectron spectroscopy

X-ray photoelectron spectroscopy (XPS) measurements were performed with a Thermo Scientific ESCALAB QXi spectrometer employing a monochromatic Al Kα X-ray source (1486.6 eV) operating at 200 W, a concentric hemispherical analyser, and a spot size of 900 μm × 200 μm. The pressure in the analysis chamber was better than 10⁻⁷ mbar. Few mg of sample powder were mounted with a double adhesive carbon tape on a sample holder grounded to the instrument. Charge compensation was applied by beams of combined low energy ion (Ar⁺) and electron beams. The sample degrades due to X-ray exposure. For this reason, short acquisitions were performed for a total experiment time of 5 min without observing any sensible degradation of the compound. Survey scan was measured in a binding energy range of 0–1350 eV with a constant pass energy of 100 eV, at 1.0 eV/step, with a dwell time of 50 ms. High-resolution spectra were recorded using a constant pass energy (25 eV), at 0.05 eV/step, with a dwell time of 50 ms. Valence band was recorded using a constant pass energy (40 eV), at 0.1 eV/step, with a dwell time of 50 ms. High-resolution XPS spectra were used for assessment of the elemental state as well as for quantification using the sensitivity factors provided by the manufacturer by means of the Avantage software after background correction with the smart-background function implemented in the same software. Peak fittings were performed in the framework of the Avantage software after background correction with the smart-background function implemented in the same software using pseudo-Voigt functions for the synthetic peaks. XPS valence band was fitted with pseudo-Voigt functions introducing constrains in the energy separations and relative intensities of the synthetic peaks according to the results obtained from DOS calculations.

Absorption spectroscopy and photoluminescence studies

Absorption spectra of the solutions were recorded on a CARY5000 double-beam spectrophotometer in the 400–800 nm range, with a spectral bandwidth of 1 nm.

Diffuse reflectance spectra were recorded in the range 250–800 nm on the same spectrophotometer equipped with an internal diffuse reflectance accessory consisting of a poly(tetrafluoroethylene)-coated integration sphere. The spectra were acquired and plotted as the Kubelka-Munk function F(R). Luminescence spectra of sample powders were recorded with a Horiba JobinYvon Fluorolog-3 spectrofluorimeter in a front-face acquisition geometry. The instrument was equipped with a double-grating monochromator in both the excitation and emission sides coupled to a R928P Hamamatsu photomultiplier and a 450 W Xe arc lamp as the excitation source. Emission spectra were corrected for the detection and optical spectral response of the spectrofluorimeter supplied by the manufacturer. The luminescence lifetimes were measured with an experimental uncertainty of ±10 %, using a pulsed Horiba Spectraled and elaborated with standard software fitting procedures.

Computational details

Single point calculations on the asymmetric unit were carried out by using the Orca program (version 4.2.1) [32]. Different hybrid (PBE0 and B3LYP) and meta-hybrid (M06) functionals had been used with the all-electron triple-ζ quality Alrichs basis set with one polarization function (def2-TZVP) [33] for all atoms. Coulomb and exchange integrals were approximated by using the Resolution of Identity approximation with the def2/JK auxiliary basis set [34]. Dispersion corrections were included by adopting Grimme’s DFT-D3 (D3BJ) method [35]. The total density of states (TDOS) and the partial density of states (PDOS) were obtained with Multiwfn 3.4.1 program [36]. TDOS and PDOS results are very similar for all three functionals, then only outcomes for one (PBE0) will be shown in the text.