A rapidly-reversible absorptive and emissive vapochromic Pt(II) pincer-based chemical sensor

Selective, robust and cost-effective chemical sensors for detecting small volatile-organic compounds (VOCs) have widespread applications in industry, healthcare and environmental monitoring. Here we design a Pt(II) pincer-type material with selective absorptive and emissive responses to methanol and water. The yellow anhydrous form converts reversibly on a subsecond timescale to a red hydrate in the presence of parts-per-thousand levels of atmospheric water vapour. Exposure to methanol induces a similarly-rapid and reversible colour change to a blue methanol solvate. Stable smart coatings on glass demonstrate robust switching over 104 cycles, and flexible microporous polymer membranes incorporating microcrystals of the complex show identical vapochromic behaviour. The rapid vapochromic response can be rationalised from the crystal structure, and in combination with quantum-chemical modelling, we provide a complete microscopic picture of the switching mechanism. We discuss how this multiscale design approach can be used to obtain new compounds with tailored VOC selectivity and spectral responses.


Supplementary
Supplementary Table 2 Gas-phase formation energies of the Form-I, Form-II and Form-III complexes, calculated using the Vienna ab initio simulation package [2] with the PBEsol functional [3] and PBEsol with the DFT-D3 dispersion correction with Becke-Johnson damping (PBEsol + D3 (BJ)). [4,5] These complexation energies may be compared to the solid-state formation energies in Supplementary Table   3. For comparison, energies obtained with the dispersion-corrected PBE0 functional [6] (PBE0 + D3 (BJ)) using Gaussian 09 [7] are also given. Supplementary Table 3 Solid-state formation energies of the hydrated Form-I, anhydrous Form-II and methanolic Form-III, calculated relative to the gas-phase components using the PBEsol functional [3] with and without the DFT-D3 dispersion correction with Becke-Johnson damping (PBEsol + D3 (BJ)). [4,5] The second column shows the per formula unit (F.U.) differences in the formation energies of the two solvated forms relative to the anhydrous Form-II. In this note, we discuss in detail the procedure and results from the computational modelling performed to support the experimental characterisation in this study.
All quantum-chemical calculations were performed within the Kohn-Sham density-functional theory formalism, [8,9] with molecular and periodic calculations carried out as described below.
Initial models were build using the Avogadro software, [10] and the geometries were optimised using the Gaussian 09 suite of programs [7] with the PBE0 exchange-correlation functional [6] and the default convergence criteria. A basis of 6-31g(d) quality was used for the main-group atoms, [11] and the Pt, Br and I atoms were treated using the LANL2DZ effective-core pseudopotential [12,13] and corresponding double-zeta basis set. The optimised structures were confirmed to be energetic minima by the absence of negative vibrational modes in the nuclear Hessian matrices. Natural population analyses (NPAs) [14] were performed on the optimised structures to obtain the energies of the ' ( orbitals, which were then compared to the energies of the frontier molecular orbitals ( Fig. 1 in the text).
We also performed a series of calculations to estimate the complexation energies associated with the Pt-pincer dimer (Form-II structure) and the dimer with H-bonded water (Form-I) and methanol (Form-III). Initial models were built by extracting the complexes and the constituent components, viz.
the Pt-pincer complex 1, H 2 O, methanol, and the (Pt-pincer) 2 , (Pt-pincer.H 2 O) 2 and (Pt-pincer.CH 3 OH) 2 dimers from the experimental X-ray structures. These were then optimised with the same parameters as for the initial ligand comparison. This set of calculations was performed with the DFT-D3 dispersion correction [4] with Becke-Johnson damping [5] added to the PBE0 exchange-correlation functional.

Periodic calculations
Periodic calculations were carried out within the pseudopotential plane-wave DFT formalism implemented in the Vienna ab initio simulation package (VASP) [2] and Quantum ESPRESSO (QE) [15] codes.
Initial coordinates for the anhydrous, hydrated and methanolic forms of 1 (i.e. Form-II, Form-I and Form-III) were taken from the X-ray structures and fully optimised in VASP until the magnitude of the forces on the ions fell below 10 -2 eV Å -1 . The PBEsol functional [3] was used to model quantummechanical exchange and correlation, with and without the D3 dispersion correction with Becke-Johnson damping [4,5] (we denote the dispersion-corrected functional by PBEsol + D3 (BJ)). Projector augmented-wave (PAW) pseudopotentials [16,17] treating the C, O and N 2s and 2p and the Pt 6s and 5d states as valence were used to model the ion cores. The electronic structure was expanded in a planewave basis with a 750 eV kinetic-energy cutoff, and the electronic Brillouin zone was sampled using a Γ-centered Monkhorst-Pack k-point mesh [18] with 1×1×3 subdivisions. The precision of the chargedensity grids was chosen automatically to avoid aliasing errors. These convergence criteria were found to be sufficient to converge the absolute values of the total energy and stress tensor to within 1 meV per atom and 1 kbar (0.1 GPa), respectively. During the wavefunction optimisation, the tolerance for the electronic minimisation was set to 10 -8 eV, and the PAW projection was performed in real space.
The optimised lattice parameters were found to be in good agreement with the experimental For consistency with VASP, we used PBEsol with PAW pseudopotentials, 1 generated using the same technique [17] and with the same valence and core electrons as those in the VASP calculations, and applied comparable convergence criteria, viz. a plane-wave cutoff of 55 Ry for the electronic wavefunctions and a tolerance of 10 -9 Ry for the wavefunction minimisation.
A single electronic minimisation on the relaxed (single) unit cells, using the same k-point mesh as in the VASP calculations, yielded electronic gaps between the highest-occupied and lowestunoccupied crystal orbitals (HOCOs/LUCOs) within 1 %, with a maximum offset of 8 meV, providing some assurance that the two codes produce similar electronic structures.
To further assess the correspondence between the VASP and QE calculations, we allowed the PBEsol bandgap of ~50-100 meV (2.5 -9 %) compared to the VASP calculations. While some of these changes are less than satisfactory, given the differences between the software and technical setups -in particular, the inevitable differences in the pseudopotentials, despite their being generated using the same technique -we consider them acceptable. We should also note that the differences may be due in part to the two codes using different break criterion for the cell relaxation, despite our attempting to match convergence parameters as closely as possible.
Based on these calculations, we are confident that using QE to visualise the frontier orbitals of the VASP-optimised structures is reasonable in this case, despite differences between the two codes. 1 We More accurate electronic-structure calculations were carried out by performing single-point calculations on the optimised structures in VASP using PBE0, [6] from which we simulated solid-state absorption spectra by calculating the frequency-dependent dielectric functions with the linear-optics routines in VASP. [19] In these calculations, a more accurate reciprocal-space PAW projection was employed, and the number of electronic bands was increased to roughly triple the default to ensure convergence of the sum over empty Kohn-Sham states. Form-I and blue Form-III are correctly reproduced by these calculations, taking into account the normalisation of the spectra in Fig. 3 (b) in the text.

Formation and complexation energies
To explore the energetics of the transformations between Form-I, Form-II and Form-III, we calculated and compared solid-state formation energies relative to the gas-phase components (Supplementary Table 3 higher. This may be ascribed to the in principle more accurate description of electrostatic interactions afforded by PBE0.

Smart coatings on glass
Thin films of complex 1 on microscope glass slides ( Fig. 3 (a)) were prepared by drop-casting a dichloromethane solution and allowing the solvent to evaporate. Similar films were prepared on highly-ordered pyrolytic graphic (HOPG) substrates by drop-casting a solution of 1 in chloroform. The microstructure of the HOPG films was then analysed by scanning-electron microscopy (SEM) using a JEOL SEM6480LV microscope operated at 10 kV (Fig. 3 (f), text).

Water vapour sensitivity measurements
To estimate the sensitivity to water vapour, we carried out a standard LiCl test as follows. The temperature of the solution was raised incrementally from 4 to 40 °C, being monitored using the thermocouple, while reflectance spectra were recorded continuously. The film was observed to change from the yellow anhydrous Form-II to the red hydrated Form-I over the range of approx. 29-34 °C, which were converted to water vapour concentrations of ~4,500-6,000 ppmV ( Supplementary   Figures 5 and 6) using the data in Ref. [22]. The reflectance spectra were post-processed using a custom code, written in the Python programming language [23] with the NumPy, [24] SciPy [25] and Matplotlib [26] packages, and were smoothed with 21-point triangle filter and differentiated once using spline interpolation. This code has been made open-source and can be obtained from a public GitHub repository. [27] Under controlled increase of the water vapour concentration, the spectroscopy also provides additional insight into the switching mechanism.
The first-derivative diffuse-reflectance spectra over a range of temperatures (equivalently, water vapour concentrations; Supplementary Figure 7) shows multiple features corresponding to the colour-transition edge of the yellow Form-II, viz. two peaks at ~530 and 560 nm, and a shoulder around 500 nm, along with a single peak corresponding to the transition edge of the red Form-I (~570-610 nm).
As the water vapour concentration is increased over the range where the transition takes place, the features from Form-II diminish in intensity, while the maximum of the peak from Form-I grows and the maximum shifts continuously.
The continuous shift of the Form-I transition edge shows that, at intermediate water vapour concentrations, the crystallites can be "partially filled", i.e. can contain less than one water molecule per Pt-pincer complex. If the solvent molecules can diffuse through open channels in the crystal structure, this would allow the average Pt-Pt distance and/or the offset between adjacent complexes to vary continuously, thereby allowing for a continuous change in the absorption profile, as observed. This lends further support to the diffusive mechanism inferred from the structural analysis in the text.
The persistence of the features from Form-II suggests that, at intermediate water vapour concentrations, the film consists of a mixture of Form-II and Form-I crystallites (layering can be ruled out both from the films being relatively thin, and also based on the intensities of the spectral features [28]). This suggests that formation of the solvent channels requires a certain critical water vapour concentration, presumably dependent on the crystallite size, which is consistent with the conversion from Form-II to Form-I beginning with the solvent adsorbing to the crystal surface and then facilitating the formation of an initial channel structure for subsequent rapid diffusion.

Cycling endurance tests
To test the cycling endurance, a coated glass slide was mounted on a motorised holder spinning After preparation, the film could be removed from the support, yielding a free-standing membrane containing impregnated microcrystallites of 1 (Figs. 3 (g, h)). The microstructure of the supported membranes was studied using optical microscopy (GXML 3230 equipped with a GXCAM USB HiChrome-S digital camera and the GXCapture software; Supplementary Figure 9).

Growth of single crystals for X-ray diffraction studies
Red crystals with a needle morphology, consistent in appearance with the bulk powder of Form-I, were grown by slow evaporation of an acetonitrile solution under ambient conditions (i.e. in the presence of atmospheric water vapour). The structure was characterised using single-crystal X-ray diffraction, and a simulated powder pattern matched that recorded from the bulk solid powder, confirming it to be a representative structure (Supplementary Figure 10).
Blue crystals with a needle morphology, consistent in appearance with Form-III, were obtained from a slow-cooled methanol solution, and were similarly confirmed to be structurally representative of the bulk solid by matching simulated and recorded powder patterns (Supplementary Figure 11).
These crystals could not be kept in ambient air due to conversion to the hydrated Form-I. To collect an X-ray structure, a methanol suspension of the needle crystals was injected into a layer of Fomblin oil, and a single crystal was mounted suspended in a droplet of oil onto a micromount. This was then flash cooled using the cryostream mounted on the diffractometer, effectively avoiding exposure to ambient air during data collection. Slow evaporation of a dilute acetone solution under dry nitrogen yielded a fourth set of yellow crystals. These were found to be solvent free, with a powder pattern consistent with the bulk solid (Supplementary Figure 12). As with the Form-III crystals, careful preparation was required in order to avoid exposure to ambient air. An acetone suspension was injected into a layer of Fomblin oil, as with Form-III, and the sample was frozen on the mount in-situ on the microscope stage using liquid nitrogen before moving it to the diffractometer.

Single-crystal X-ray crystallographic studies
Crystallographic data for the X-ray structures of Form-I, Form-II and Form-III is summarised in Supplementary Table 1 Figure 20 (b)).

X-ray Single Crystal Structure Determination of the ethanol solvate form of complex 1.EtOH
Crystals of the ethanol solvate of complex 1 (1.EtOH) were obtained by slow evaporation from a saturated solution of 1 in ethanol. The single-crystal X-ray structure confirms that one equivalent of ethanol per Pt(II)-pincer molecule is included in the crystal (see Supplementary Figure 21). 1.EtOH crystallises in the same P2 1 /n space group as Forms I to III, but with a significantly larger unit cell volume due to the inclusion of the bulkier ethanol solvent molecule. The unit cell expansion is largely isotropic, and the structure of 1.EtOH can be considered to be an expanded analogue of the other solvated forms of 1 reported in this study. As for the methanolic Form-III, the ethanol molecule is involved in a single, discrete hydrogen bonding interaction to N(1) of the CN-ligand. However, unlike Form-III, the more sterically-demanding ethanol molecule cannot be included in the crystal structure without disrupting the Pt…Pt overlap between adjacent pincer molecules in the stack. Supplementary   Figure 22 shows that there is now no significant ' ( overlap within the stack, which explains the yellow colour of the 1.EtOH crystals.