Strongly coloured thiocyanate frameworks with perovskite-analogue structures

We report the first examples of thiocyanate-based analogues of the cyanide Prussian blue compounds, MIII[Bi(SCN)6], M = Fe, Cr, Sc. These compounds adopt the primitive cubic pcu topology and show strict cation order. Optical absorption measurements show these compounds have band gaps within the visible and near IR region, suggesting that they may be useful for applications where light harvesting is key, such as photocatalysis. We also show that Cr[Bi(SCN)6] can reversibly uptake water into its framework structure pointing towards the possibility of using these frameworks for host/guest chemistry.

Molecular framework materials, where metal centres are connected by molecular ligands into three dimensional networks, bridge the gap between the inorganic and organic solid state. Perhaps the most widely studied inorganic framework structure-type is the ABX 3 perovskite, which is formed from a three dimensional, topologically cubic, network of corner sharing BX 3 octahedra, charge balanced by A-site guests occupying the network cavities. Perovskites are important because of the wide range of physical properties that they can support. The relatively open structure permits ferroelectricity via the correlated displacement of cations and the facile (de)intercalation of guests (as in WO 3 electrochromic devices 1 ), and the 3D cubic BX 3 connectivity facilitates correlated behaviour, including multiferroicity, 2 octahedral-tilt driven ferroelasticity 3 and the extended electronic states found in heavy metal halide solar cell absorbers. 4 Molecular framework perovskites are of particular interest as the large 'X' ligands permit both unusual exibility impossible in inorganic perovskites (e.g. tilt-driven ferroelectricity 5,6 ) and a more extensive range of guest chemistry. 7 The properties of these materials are already of great interest, notably the Prussian blue metal cyanides, which are useful as both catalysts 8 and electrochemical components, 9 and the alkylammonium metal formates, which can possess simultaneous dipolar and magnetic order. 10 Despite the promise of molecular perovskites, the range of chemistry is limited relative to that of their 'atomic' counterparts: most work is conned to studies of the rst-row transition metals and their complexes with a small number of ligands. 7 Exploring the chemistries of underexploited metal cations and molecular ligands thus offers the opportunity to produce new materials with properties very different from the current exemplars. Thiocyanate (SCN À ) is a particularly interesting candidate small ligand for the assembly of molecular frameworks, as its chemical soness allows easy access to a wider range of coordination chemistry. In addition, thiocyanate based materials oen have strong optical absorption, for example the terpyridine Ru(SCN) 3 complexes used in dye-sensitised solar cells, 11 and can contain reasonably strong magnetic interactions. 12,13 Despite the potential of this ligand, there are relatively few reported structures of three dimensional thiocyanate based frameworks, of which only two are perovskites, CsCd(SCN) 3 and the double perovskite (NH 4 ) 2 CdNi(SCN) 6 . [14][15][16][17][18] In addition, there are no known thiocyanate analogues of the Prussian blue structured (i.e. A-site vacancy double perovskite) cyanides.
One reason why there are fewer thiocyanate-based framework materials than for other small molecular ligands such as formate, azide or cyanide, are its stringent bonding requirements. Unlike other widely investigated small ligands, the coordinative preferences of the two termini of the SCN À ligand are quite distinct: the S-terminus is so and binds to Class B metals and the N-terminus is harder and binds to Class A metals. As forming a framework requires both ends of the ligand to coordinate to a metal, this means a simple homometallic framework will have to have intermediate hardness.
There are very few metals that have been shown to form homoleptic complexes with both the S and N-termini of the SCN À ligand and of these, only Cd 2+ has been investigated for its framework forming ability [ Fig. 1(c)]. 17,18 Not only does this limit the range of possible compounds, as a d 10 cation, Cd 2+ is inherently unpromising for optical and magnetic applications. Double perovskite derived structures, where two different metal cations full the differing bonding requirements of the two termini, offer one route for expanding the compositional space of thiocyanate.
Out of the metals which preferentially coordinate to S-terminus of NCS À Bi 3+ is perhaps the most attractive option, as it is non-toxic and inexpensive. Previous work has established that the homoleptic octahedral [Bi(SCN) 6 6 ]. This solution was evaporated to dryness, and then extracted using 100 mL of dry tetrahydrofuran. Aer removal of the solvent in vacuo and drying at 50 C for 2 h, a very dark green microcrystalline powder of Fe[Bi(SCN) 6 ] was produced (1.02 g, 67%).  The route used to synthesise microcrystalline Fe[Bi(SCN) 6 ] was adapted for the synthesis of Sc[Bi(SCN) 6 ]. KSCN (0.6 mmol, 0.0583 g) was added to a stirred suspension of Bi(NO 3 ) 3 $5H 2 O (0.1 mmol 0.0485 g) in 10 mL of butanone producing a yellow-orange solution and a white precipitate of KNO 3 . Sc III (NO 3 ) 3 $xH 2 O (0.1 mmol, 0.023 g), was then added, producing a deep orange solution. The reaction mixture was stirred for overnight and then ltered to remove KNO 3 , yielding an orange solution of Sc[Bi(SCN) 6 ]. This solution was evaporated to dryness from which diffraction quality single crystals of Sc[Bi(SCN) 6 ] could be obtained.

Thermogravimetric analysis
Thermogravimetric analysis was carried out with a Mettler Toledo TGA/SDTA 851 thermobalance. Powder samples of 20-40 mg were placed in a 100 mL Al 2 O 3 crucible and heated to 650 C (at 10 C min À1 ) under a ow of N 2 .

Optical measurements
Diffuse reectance measurements were carried out on nely ground powdered samples, diluted to 10 wt% with BaSO 4 to remove effects of strong absorption, using a Varian Cary 50 UV-vis spectrometer equipped with a diffuse reectance accessory (DRA) probe (Barrelino, Harrick Scientic) measured over the range 350-1000 nm. The optical absorption measurements were performed using an integration sphere equipped Perki-nElmer lambda 750 UV-vis-NIR commercial setup. The set-up includes light source of tungsten-halogen (for vis-NIR region) and deuterium (for UV region) lamps, a conventional photomultiplier (PMT) detector, a 10 cm integrating sphere module attachment, a monochromator, and a detector slit width of 10 nm. All measurements were performed under standard ambient conditions. Films were prepared by drop-casting the crystals, dissolved/suspended in the appropriate solvent (DMF), on the quartz/glass slide substrate. The transmitted light was corrected using a reference of an empty quartz coverslip of the same type and thickness as the substrate used for the sample. The absorption onsets were determined by extrapolation from a Tauc plot. We used our DFT results to guide our assumed Tauc exponents: for Cr[Bi(SCN) 6 ] and Sc[Bi(SCN) 6 ] we carried out ts for both direct and indirect band gaps and for Fe[Bi(SCN) 6 ] we used the Tauc exponent for a direct band gap. Further details of the optical characterisation can be found in the ESI. †

Powder X-ray diffraction
All microcrystalline samples were initially assessed for their purity via their X-ray powder diffraction patterns, measured using a PANalytical Empyrean diffractometer (Cu Ka radiation, l ¼ 1.541Å). High resolution synchrotron measurements on Cr [Bi(SCN) 6 ] were carried out at beamline 11-BM at the Advanced Photon Source (APS) using a wavelength of 0.414537Å and on Cr [Bi(SCN) 6 6 ] on a PANalytical Empyrean diffractometer using a Anton Parr XRK 900 furnace. Analysis of all powder diffraction data (including indexing, Pawley renement and Rietveld renement) was carried out using the TOPAS Academic 4.1 structure renement soware package. [24][25][26] Full details of the powder X-ray renements can be found in the ESI. †

Single crystal X-ray diffraction
Single-crystal X-ray diffraction data were collected using a Nonius KappaCCD diffractometer, using graphite monochromated Mo Ka radiation (l ¼ 0.7107Å). Structure solution was carried out using SHELXT and renement with SHELXL, within the OLEX2 graphical interface. [27][28][29] All non-hydrogen atoms were rened anisotropically with no additional restraints or constraints. Supplementary crystallographic data for this paper including of all single crystal structures can be found in the ESI † and were deposited in the CCDC 1860296-1860300.

DFT calculations
Geometry optimised structures and associated spectral properties were calculated using density-functional theory (DFT), using the structure of Fe[Bi(SCN) 6 ] determined from single crystal diffraction as a starting model. Calculations were performed using the plane wave CASTEP DFT code, 30 and the Perdew-Burke-Ernzerhof (PBE) exchange-correlation functional was used with Vanderbilt ultraso pseudopotentials. 31 A basis set containing plane waves with energies of up to 800 eV and a Monkhorst-Pack (MP) grid corresponding to a Brillouin zone (BZ) sampling grid ner than 2p Â 0.03Å À1 was used. Geometry optimisations were performed with spin polarisation and spectral calculations were additionally performed with DFT+U. To determine the inuence of the value of U on the electronic structure, separate calculations were performed with U ¼ 0 and 2.5-5.5 eV for the Cr d orbitals and U ¼ 0 and 4.0-6.0 eV for the Fe d orbitals in 0.5 eV steps. Changing U values had little impact on band occupancy and the overall structure of the bands, however by changing the energy of the metal d orbitals it was possible to nd values of U that reproduced the experimentally observed band gaps, U ¼ 2.5 eV and U ¼ 4.5 eV were chosen for the Cr and Fe d-orbitals respectively. These values are broadly consistent with U values found in previous studies. 32,33 Density of states for both materials was calculated using OptaDOS 34,35 using adaptive broadening. 36 Additional information on the calculations, and particularly the variation in U can be found in the ESI. †

Results and discussion
All three compounds were synthesised by reacting thiocyanatobismuthate solutions, in dilute nitric acid or polar aprotic solvents, with a transition metal source. The key difference between the synthetic routes required for these materials was due to the large difference in solubilities. Cr III [Bi(SCN) 6 6 ] it was possible to grow diffraction quality single crystals, from which we were able to solve their structures. These structures were used as the starting points for Rietveld renement of synchrotron powder diffraction data of Cr III [Bi(SCN) 6 ], conrming that despite the differences in solubility, which we ascribe to the kinetic inertness of Cr III complexes, all three compounds adopt isotypical structures consisting of M III N 6 and Bi III S 6 octahedra linked by thiocyanate ligands into a 3D network with pcu topology [ Fig. 1a and b]. This structure is both analogous to the cyanide Prussian blue structure and can also be described as a double perovskite structure with vacancies occupying the A-site.
Unlike the cyanide-based Prussian blue materials, which crystallise with high symmetry cubic structures and oen have substantial cation and anion disorder, 37 these thiocyanate frameworks are well-ordered and adopt distorted monoclinic structures. The size of the distortion is demonstrated by greatly reduced volume compared to the hypothetical cubic Fm 3m aristotype, which is 54% for Fe[Bi(SCN) 6 ] [ Fig. 2a and b]. This distortion means that the structures are much less porous than might be expected from consideration of the length of the NCS À ligand alone.
Symmetry analysis carried out using the ISODISTORT soware 38 showed that the two most important symmetry-adapted distortion modes which relate this monoclinic structure to the face-centered cubic parent structure are the G 4 + and X 3 + modes.
These two modes are, to a good approximation, tilts of rigid metal octahedra. The X 3 + mode corresponds to an in-phase tilt along c, i.e. the tilts in every plane perpendicular to c are in the same sense. The G 4 + mode corresponds two equal sized out-ofphase tilts along the a and b directions, i.e. the tilts in sequential planes perpendicular to a and b directions are in opposite senses. The octahedral tilts pattern can therefore be summarised by the a À a À c + Glazer notation (conventionally Fig. 2a]. 39 This means that perhaps surprisingly, these very distorted structures adopt the most common tilt system for double perovskites, 40 and that non-standard cooperative perovskite distortions such as anti-phase tilts and columnar shis, although common in other molecular perovskite-type frameworks, do not play a signicant role in these structures. 5,41 The origin of tilting in both inorganic and molecular perovskite structures is ordinarily the size mismatch between the A-site cation and the pore size of the anionic frameworks. 40,42 Prussian blue structured materials, like these M III [Bi(SCN) 6 ] frameworks, have no A-site cation and a neutral framework and so 'tolerance factor' arguments cannot be used. Instead, the size of the tilts is largely driven by the general tendency towards higher density structures and the geometry of  6 ], which has Ni(NCS) 6 octahedra, has tilts which belong to the conventional a + b À b À tilt system whereas CsCd(SCN) 3 , which has more exible Cd(NCS) 6 octahedra with bent Cd-N-C bond angles (as small as 116.5 ), has more complex cooperative distortions, including unconventional tilts. 5 The shapes of the molecular orbitals of the thiocyanate ligand also explain the magnitude of the observed tilts in these frameworks. There is no unique way of decomposing the total tilts into separate rotations around each of the three axes, as the octahedral rotations do not commute. 39 While this effect is small for small (<10 ) tilts, it cannot be neglected for larger tilts. A good approximate decomposition of the octahedral rotations into separate tilt angles can be found in the angles between the axes of the metal octahedron, i.e. the metal-ligand bonds, and the pseudo-cubic unit cell axes, dened by the vectors which connect the metal centres [ Fig. 2d and 3 , which give an average tilt angle of hQi ¼ 52.7 . As the M-N-C bond angle is nearly linear, the sum of the tilts of the BiS 6 and MN 6 octahedra relates directly to the Bi-S-C bond angle (Q + U z 180 À q).
The scale of the octahedral tilts in NCS perovskite-derived structures is therefore set by the bent Bi-SCN bonding, rather than the mismatch between a guest cation and the framework. The tilt-angles are therefore likely to be only weakly dependent on temperature, and so thermal second-order octahedral-tilt driven phase transitions are not expected to be common in this series of compounds.
As anticipated, therefore, variable temperature X-ray diffraction experiments conrmed that both Fe[Bi(SCN) 6 6 ] is carried out in more acidic solution, and also slowly forms from Cr[Bi(SCN) 6 ] at room temperature and ambient humidity over a period of weeks and when Cr [Bi(SCN) 6 ]$H 2 O is heated above 150 C, this guest water is lost, leading to a reduction in cell volume by 19Å 3 and a mass loss of 3% [ Fig. 3(a)]. Further experiments will be required to assess how stable this material is to extended cycling of water uptake/ release. The clearest diffraction signature of the dehydration is the separation of the {110} and (002) reections, which are near coincident in the hydrate, but do not overlap in the anhydrate [ Fig. 3(c)]. As with the other M[Bi(SCN) 6 ] frameworks, this apparent peak overlap does not correspond to any change in framework symmetry, but merely results from the near cubic pseudo-symmetry of the metal substructure.
Due to the relatively small contribution of the guest water to the total electron density (3%) and peak broadening resulting from the small crystallite sizes, it was not possible to localise the guest water from the fourier difference map generated from Rietveld renement of Cr[Bi(SCN) 6 ]$H 2 O. Calculation of the geometric voids in the empty framework structure did however reveal a single void capable of containing a water molecule [ Fig. 3(b)]. Rietveld renement of the framework structure including an oxygen atom situated in that void with its occupancy xed to the value derived from TGA measurements  Table 1]. 43 The rigidifying effect of guests has been noted in both inorganic and molecular framework materials, where even weakly interacting guests can reduce the material's exibility. 44,45 The maximum thermal expansion seen in Cr[Bi(SCN) 6 ], of 1.7% over 250 C range, is smaller than both the volume change on water uptake (2.3%) and particularly the volume change relative to the cubic structure aristotype (54%). This suggests that octahedral tilt transitions to higher symmetry structures are unlikely to occur in this family on heating, even if framework decomposition can be avoided.
All three materials, Sc[Bi(SCN) 6 ], Cr[Bi(SCN) 6 ] and Fe [Bi(SCN) 6 ] are strongly coloured (respectively, orange-red, brick red and black). Diffuse reectance spectroscopy and absorption spectroscopy on powder samples conrmed that these materials possess band gaps determined from a Tauc plot within the visible and near infra-red region [ Table 2, Fig. 5]. The variation in the observed absorption spectra shows that both bismuththiocyanate and transition metal-thiocyanate moieties play  6 ] over this region (1.7%) is significantly smaller than the volume difference between the hypothetical aristotype and the room temperature structure (54%). As the changes in sin (q) are two orders of magnitude smaller than the strains along the unit cell axis directions the strain eigenvalues are to a good approximation equal to the unit cell axis strains. moieties, but rather must be inuenced by its structure in the solid-state framework. This difference between solution and solid-state spectra has also been reported for alkyammonium salts of the hexakisisothiocyanato iron(III) complexes, suggesting that coordination to bismuth in particular is not responsible for the observed red-shi. 46,47 To understand better the optoelectronic properties of these materials we therefore carried out density-functional theory (DFT) calculations on Fe[Bi(SCN) 6 ] and Cr[Bi(SCN) 6 ]. We calculated the electronic band structures using spin-polarised GGA as implemented by CASTEP on geometry-optimised structures, and then used OptaDOS to determine from these results the optical band gap and the electronic density of states, including the atom and angular momentum projected density of states. The calculated band gaps were signicantly smaller than those observed experimentally. This is commonly observed in DFT calculations due to the self-interaction error, which is particularly pronounced for the contracted d orbitals. We therefore repeated the calculations of the electronic structure, but now also including a Hubbard U in the Hamiltonian, which we systematically varied. We found that introducing this additional factor did not perturb qualitative features of the electronic structure aside from the energy of the transition metal d states. These calculations allowed us to determine that for both Fe[Bi(SCN) 6 Table 2]. Examination of the projected density of states suggests that the transition is primarily LMCT in character, as the states near the valence band maximum are primarily due to thiocyanate, and the conduction band minima states are mainly Fe for Fe [Bi(SCN) 6 ] and primarily Cr and Bi for Cr[Bi(SCN) 6 ] [ Fig. 6a and b]. The narrow width of the calculated conduction and valence bands for both compounds indicates that the near band-edge states are highly localised, and so the effective masses of holes and electrons will be very large [ Fig. 6a and b]. This is also borne out by examination of the Kohn-Sham orbitals at the valence band maximum, which are localised and conned to the ligands [ Fig. 6(d)]. The lack of extensive delocalisation suggests that these frameworks would be better suited to applications where longrange electron transport is not essential, such as photocatalysis than as bulk semiconductors. Photoluminescence measurements on both single crystal and powder samples showed negligible uorescence, suggesting that there is a high concentration of optical defects, borne out by the signicant band-tails observed in both absorption and reectance measurements. The slightly indirect nature of the band gap for these frameworks, particularly for Cr[Bi(SCN) 6 ], could also be a contributory factor to the absence of observed uorescence.

Conclusions
In this paper we report the synthesis, structure and optical properties of the rst examples of a new family of moleculebased perovskite derived materials, transition metal hexathiocyanatobismuthmates M[Bi(SCN) 6 ]. They are the rst example of thiocyanate analogues of the cyanide Prussian blues, and to the best of our knowledge are the rst molecular perovskite derived frameworks which incorporate heavy p-block cations. These compounds are vividly coloured, due to the presence of LMCT bands, and the higher optical absorption seen in these compounds suggests that thiocyanate-based frameworks may well be interesting materials to investigate for photocatalysis. We have also demonstrated the reversible uptake of water into Cr[Bi(SCN) 6 ], which suggests it may be possible to incorporate other guests into this family of compounds. The chemical space opened up by these compounds, by substituting on the B site and introducing A site cations, and their unusual optical properties, suggests that thiocyanate perovskites could be an important family of molecular framework materials.

Conflicts of interest
There are no conicts of interest to declare.