Photoinitiated Single-Crystal to Single-Crystal Redox Transformations of Titanium-Oxo Clusters

Titanium-oxo clusters can undergo photochemical reactions under UV light, resulting in the reduction of the titanium-oxo core and oxidation of surface ligands. This is an important step in photocatalytic processes in light-absorbing Ti/O-based clusters, metal–organic frameworks, and (nano)material surfaces; however, studying the direct outcome of this photochemical process is challenging due to the fragility of the immediate photoproducts. In this report, titanium-oxo clusters [TiO(OiPr)(L)]n (n = 4, L = O2PPh2, or n = 6, L = O2CCH2tBu) undergo a two-electron photoredox reaction in the single-crystal state via an irreversible single-crystal to single-crystal (SC-SC) transformation initiated by a UV laser. The process is monitored by single crystal X-ray diffraction revealing the photoreduction of the cluster with coproduction of an (oxidized) acetone ligand, which is retained in the structure as a ligand to Ti(3+). The results demonstrate that photochemistry of inorganic molecules can be studied in the single crystal phase, allowing characterization of photoproducts which are unstable in the solution phase.

X-ray diffraction of photoirradiated crystals experiments: Measurements were carried-out at Diamond Light Source on beamline I19-EH2 using a wavelength of 0.48590 Å and a four-circle Newport diffractometer equipped with a Dectris Eiger CdTe X 4M detector.Crystals were taken from a flask under a flow of nitrogen and mounted onto MiTeGen UV-Vis™ loops under fomblin-Y oil swiftly in air before being placed on the goniometer under a 100K nitrogen stream.Crystals were photoirradiated in situ using the PORTO variable wavelength pulse laser (305 nm, 800 mW/cm 2 ).During irradiation the crystals were rotated.The laser and X-ray spot were overlapped on the sample position.X-ray diffraction measurements were carried out in between laser irradiation intervals.

Notes on SCXRD crystal refinement methodology
For all data sets spotfinding and integration were done using the DIALS automated pipeline, 4 unirradiated crystal structures (time = 0 s) were solved using SUPERFLIP 5 and refined using CRYSTALS 6 implementation of SHELXS. 7Later timepoints used the time = 0 s model as an initial solution and were refined in the same way.
During the photoirradiation experiment, a complete 360° omega scan was performed at each timepoint, which gave a good compromise between completeness, redundancy and collection time and beam dose for the samples with the P 2/m Laue class.

Refinement and disorder model of 1.py.
Two separate crystal series were collected for 1.py on two different crystals.As the diffraction data were similar from each crystal and show concordant changes on irradiation the time series includes data from both crystals.The time series on crystal one is t = 0 s, 60 s, 360 s, 1560 s, 6160 s, whereas for crystal two is t =30 s, t= 90 s, t = 150 s, t = 300 s and t= 900 s.Note that comparison of the initial data collections of each crystal, crystal 1 (no irradiation) and crystal 2 (t = 30 s), is helpful for differentiating photochemical conversion from any sequential X-ray induced beam damage (see supporting note 2).
In general, to allow for greater comparisons between structures after different irradiation times the same disorder model was used across all structures in the series except for the initial t = 0 s structures (which do not exhibit disorder).When the disorder model is forced on to the initial structure occupancy of disordered fragments tends towards zero (or negative) and the resulting thermal ellipsoids become non positive definite.This is symptomatic of the disorder model not being appropriate for the initial structure as the photogenerated disorder is not present at time = 0 s.
Because of the chemical similarity of each isopropoxide unit it is reasonable to propose that the resulting acetone and isopropanol (and remaining isopropoxide) could be disordered in each of the four sites.Notably, each site evolves differently under irradiation and so a disorder model was used that reflected this.Ti4 -O81 -C82 [Ti4 -O4 -C4 in main text] is modelled as one fragment with complete occupancy, no restraints or constraints on geometry are applied.All other titanium isopropyl fragments are split at their carbons and modelled as two parts sharing a common unsplit oxygen.It is necessary to apply geometry restraints to the resulting fragments to have a convergent structure and these were chosen with great care.Within each part of the OC 3 ligand fragment the geometry of each half of the fragment is restrained as similar using SAME, SIMU and DELU restraints, as a reflection plane is chemically sensible for any of acetone/isopropoxide/isopropanol.However, no geometry restraints are applied between each part, this means each part is not restrained to being similar to any other part and can have differing geometry.This was done to allow for the possibility of an acetone fragment and isopropoxide/isopropanol fragment being co-located in the average structure.Further disorder modelling was necessary for two phenyl rings which are disordered in the photoproduct structure.These are modelled as two split parts and restrained with equivalent geometry as there is no reason to expect chemical difference.Finally, restraints were placed on the co-crystallised pyridine solvent molecule which shows significant disorder as irradiation time increases (noting that it is wellresolved at t=0).
In the analysis of bond lengths within this paper the model without disorder is used for before irradiation (t=0) and all other datasets are treated with the disorder model.
Refinement and disorder model of 3.
3 was modelled in a similar method to 1.py.In the irradiated structures two isopropoxide/isopropanol/acetone fragments on the Ti1 and Ti3 positions were split, whilst the isopropoxide at Ti2 remains unsplit.
Each component of the disordered isopropoxide/isopropanol/acetone was restrained to have two-fold symmetry within the molecule but with no restraints between each part.Additionally, one tert-butyl group was modelled over two sites with restraints including similarity between parts.The occupancy of the acetone component (at Ti1) grows to ~0.4 after 60 s irradiation, before rising to ~0.5 after 150 s and settling at ~0.6 at 300 s or longer.Due to the symmetry of the molecule, the acetone occupancy is expected to reach 0.5 on completion of reaction.Forcing the acetone fragment to an occupancy of 0.5 has little effect on the structure but resulted in less satisfactory atomic displacement parameters, and, therefore, free refinement of occupancy was used.
In the analysis of bond lengths within this paper the model without disorder is used for before irradiation (t=0) and all other datasets are treated with the disorder model.site should give a value of ~4 using Ti(4+) parameters, and a Ti(3+) gives ~3 using Ti(3+) parameters. 9n a delocalised system e.g. with a formal oxidation state of Ti(3.5+),Ti(4+) parameters will give a value of ~3.6 whilst Ti(3+) parameters will return a value of ~3.4. 2,8 ond valence sum calculation of 3 for each Ti environment over the total irradiation timeline using expected geometry for Ti(IV), left, or Ti(III), right.Indicating a selective drop in oxidation state with three distinct regions assigned as Ti 4+ (red shaded), Ti 3.5+ (purple shaded), and Ti 3+ (blue shaded) based on reported bond valence sum values from previously reported well-resolved related structures.

Supporting Note 1. Exploration of co-crystallised solvent molecules
The photoreactivity of several crystalline forms of compound 1 containing different co-solvent molecules were explored as part of this project.These crystals were prepared by recrystallisation of 1.tol from different solvents as described below.
1.DMSO 1.DMSO has been reported previously from the synthesis of 1 in DMSO. 13ternatively 1.DMSO can be prepared from 1.tol in DMSO.30 mg (0.021 mmol) of 1.tol was dissolved in minimal hot DMSO (75 °C) and upon cooling slowly to room temperature crystals of 1.DMSO were formed.

i PrOH
1. i PrOH has been reported previously from the synthesis of 1 in i PrOH. 14ernatively, 1. i PrOH can be prepared by suspending 1.tol in i PrOH in a sealed Teflon lined autoclave.The Flask was heated to 120°C and then allowed to cool slowly to yield crystals.

1.THF
30 mg (0.021 mmol) of 1.tol was dissolved in minimal THF, and this solution was layered with pentane to yield crystals of 1.THF.

1.2MeCN
30 mg (0.021 mmol) of 1.tol was dissolved in minimal MeCN.After leaving to stand crystals of 1.2MeCN formed.

(from Et 2 O)
1.tol (30 mg, 0.021 mmol) was partially dissolved in hot diethyl ether.The resulting solution was decanted from the remaining solid and from that solution grew crystals of 1 without co-solvent.
1.tol, 2 1.py, 1.DMSO, 13 1.i PrOH, 14 1.THF (tol = toluene, py = pyridine) adopt a similar crystal structure (space group = P2 1 /c).In contrast, 1.2MeCN, with two solvent molecules, adopts a different structure (space group = P2 1 /m).If crystals of 1 are prepared from a solution of Et 2 O then a crystalline form is found without a solvent molecule (space group = P-1).No major changes are noted in the structure of cluster 1 in the differing crystal forms.These compounds were ground to fine powders and exposed to UV irradiation under inert atmosphere, using a Analytik Jena UVLM-26 EL series UV lamp (output 302 nm, approx.3 mW/cm2 @ 1 cm, 6 W power). 1.py undergoes a notable white to purple colour change under the UV lamp, whilst the other powders show varying degrees of colour change (Table S1).Note that coordination of pyridine to photoreduced Ti-oxo clusters has been reported to result in a strong purple colour due to metal to (pyridine) ligand charge transfer processes, therefore, the stronger colour found from 1.py does not necessarily signify a greater photoconversion. 2,8 n the single crystal phase 1.py changes colour from colourless to blue under UV light (at 100 K), with evidence that co-crystallised pyridine molecules do not coordinate to the reduced Ti-oxo cluster.In the higher surface area powder phase, under room temperature conditions, the purple colouration may indicate that some ligand exchange to allow pyridine coordination is possible.
In most cases the photoconversion of powders is low (see Table S1) after several hours under the low-power UV lamp, with most photoreactivity expected to occur on the surface of the powders.Without a more careful analysis of the surface areas of these different powders, detailed analysis of the effect of co-solvents on the rate of photoreactivity in the solid-state is unclear.It is possible that the influence of co-crystallised solvent may enact topochemical changes which may cause changes in reaction site selectivity, however, detailed study was beyond the scope of this study. 15radiated samples (room temperature) of powdered 1.py were analysed by FTIR spectroscopy within a N 2 filled glovebox.No clear signal for acetone could be detected in this case, we anticipate that the small percentages of photochemical conversion, coupled with the volatility of acetone during room temperature irradiation are most likely to account for this.N.B. acetone formation was confirmed by NMR spectroscopic analysis of the irradiated powder.

Supporting Note 2. Evaluation of X-ray beam damage
Intense X-ray irradiation can induce beam damage, leading to reduction of diffraction quality or even chemical reactivity within a crystal.Experiments show that the unirradiated dataset of 1.py ('crystal 1', t = 0 s) exhibits different structural metrics to the first data collection of 1.py 'crystal 2' which was collected after 30s UV irradiation, Figs.S13-17.Different structural metrics are also found for subsequent datasets which have received the same dose of X-rays but different time periods under the laser.This confirms that photochemical transformation under laser UV light is the dominant chemical process occurring in these crystals over the timelines recorded.However, we are currently examining whether X-ray induced redox processes may occur over longer periods of X-ray irradiation in the absence of UV light, and we hope to communicate these findings in the future.formula given is for the asymmetric unit.The crystallographic models of the irradiated structures are deficient by half an H atom.This H 0.5 is associated with an isopropanol (OH) which is not located in the structure.Formula deviations from H 53.5 are due to occupation of the acetone fragment refining to slightly greater than 0.5.

Figure S2 .
Figure S2.Diffuse reflectance UV-Vis absorption plot of 3 as a powder (top) and Tauc plot analysis giving an absorption onset of 367 nm, 3.38 eV (bottom).Note that the baseline is shifted upwards relative to the reference sample due to some light leakage associated with the use of an airtight sample holder in the diffuse reflectance experiment.

Figure S4 .
Figure S4.UV-Vis absorption plot of 1.py as a powder (top) and Tauc plot analysis giving an absorption onset of 365 nm, 3.40 eV (bottom).Note that the baseline is shifted upwards relative to the reference sample due to some light leakage associated with the use of an airtight sample holder in the diffuse reflectance experiment.

Figure S5 .
Figure S5.UV-Vis absorption plot of 1.py as a powder as synthesised (black), after irradiation with 302 nm lamp for 1 hour (purple) and after subsequent oxidation with air (yellow).Inset in the same sequence shows images of the powders of 1.py in the quartz cells.Note that the baseline is shifted upwards relative to the reference sample due to some light leakage associated with the use of an airtight sample holder in the diffuse reflectance experiment.

Figure S6 .
Figure S6.UV-Vis absorption plot of powders of 3 as synthesised (black), after irradiation with 302 nm lamp for 1 hour (purple) and after subsequent oxidation with air (yellow).Inset in the same sequence images of the powders of 3 in the quartz cells.Note that the baseline is shifted upwards relative to the reference sample due to some light leakage associated with the use of an airtight sample holder in the diffuse reflectance experiment.

Figure S7. 1 H
Figure S7.1 H NMR spectra of 1.py (bottom) and 1.py after irradiation with 302 nm for 6 h in the solid state (top), both dissolved in d 8 -toluene.Analysis of the spectra suggest ~6% of the sample has undergone photoredox reactivity in the solid-state under these conditions with equal amounts of acetone (1.56 ppm) and i PrOH (0.94, 3.63) produced in the process.The small quantity of photoreduced Ti-oxo cluster is not observed, consistent with previous reports and the likelihood of a paramagnetic species. 2 Associated 31 P { 1 H} NMR spectra remained unchanged, with a single peak observed at 32.47 ppm.

Figure S8. 1 H 8 Figure S9. 1 H 8 Figure S10 .
Figure S8.1 H NMR spectra of 2 (bottom) and 2 after irradiation with 302 nm for 6 h in the solid state (top), both dissolved in d 8 -toluene.Analysis of the spectra suggest ~9% of the sample has undergone photoredox reactivity in the solid-state under these conditions with acetone (1.56 ppm) produced in the process.The small quantity of photoreduced Ti-oxo cluster is not observed.The photoreduced cluster is expected to coordinate any produced i PrOH (1 equiv.expected) under these noncoordinating solvent conditions, and therefore free i PrOH is also not observed.It is noteworthy that the reported spectra of the photoproduct of 2 in the presence of excess i PrOH, [Ti 6 O 6 (O i Pr) 4 ( i PrOH) 4 ], is similar to 2, with overlap of signals expected.8

Figure S13 .
Figure S13.Bond valence sum calculations provide oxidation state information by using the local bond parameters of an atom.The calculation relies on deviation from expected tabulated values e.g. a Ti(4+) site should give a value of ~4 using Ti(4+) parameters, and a Ti(3+) gives ~3 using Ti(3+) parameters.9In a delocalised system e.g. with a formal oxidation state of Ti(3.5+),Ti(4+) parameters will give a value of ~3.6 whilst Ti(3+) parameters will return a value of ~3.4.2,8Bond valence sum calculation of 3 for each Ti environment over the total irradiation timeline using expected geometry for Ti(IV), left, or Ti(III), right.Indicating a selective drop in oxidation state with three distinct regions assigned as Ti 4+ (red shaded), Ti 3.5+ (purple shaded), and Ti 3+ (blue shaded) based on reported bond valence sum values from previously reported well-resolved related structures.2,8

Fig S14 .
Fig S14.Time series of crystal structures of 1.py before and after laser irradiation (labelled by duration of laser irradiation).Titanium = lilac, carbon = grey, oxygen = red, phosphorus = purple, displacement ellipsoids drawn at 50%, all disorder shown, H atoms removed for clarity.

Figure S16 .
Figure S16.O-C bond lengths in acetone-like fragment that forms at Ti1 on 1.py.Graph showing structural parameters calculated from crystal structure models of 1.py for each Ti environment over the total irradiation timeline.Error bars drawn at ± 3σ.

Figure S17 . 8 Figure S18 .
Figure S17.Bond valence sum calculations provide oxidation state information by using the local bond parameters of an atom.The calculation relies on deviation from expected tabulated values e.g. a Ti(4+) site should give a value of ~4 using Ti(4+) parameters, and a Ti(3+) gives ~3 using Ti(3+) parameters.9In a delocalised system e.g. with a formal oxidation state of Ti(3.5+),Ti(4+) parameters will give a value of ~3.6 whilst Ti(3+) parameters will return a value of ~3.4.2,8Bond valence sum calculation of 1.py for each Ti environment over the total irradiation timeline using expected geometry for Ti(IV), left, or Ti(III), right.Indicating a selective drop in oxidation state with two distinct regions assigned as Ti 4+ (red shaded), Ti 3.5+ (purple shaded) based on reported bond valence sum values from previously reported well-resolved related structures.2,8

Figure S19 .Figure S20 .*Figure S22 .*Figure S23 . 11 Figure
Figure S19.Molecular Hirshfeld d norm surfaces of 3 showing the relative Van der Waals contact distance (Blue, maximum +1.63 Å, white ~0 Å, red -0.12 Å).Crystal Explorer 21.5 was used to construct the molecular Hirshfeld surfaces in the crystal structure of 3, red indicates intermolecular contacts shorter than the van der Waals radii, white indicates intermolecular distance near the sum of Van der Waals radii with zero d norm value, and blue indicates contact distances longer than the sum of van der Waals radii.10In 3 most of the isopropoxide ligand environments appear to have long nearest contacts (shown in blue in the plot), suggesting there is accessible space for some molecular rearrangement.Notably the Ti1 isopoproxide fragment has one close contact, which may limit its mobility and therefore reactivity.In contrast, Ti2 and Ti3 isopropoxides have similar cavity space surrounding them.

Figure S25 .
Figure S25.X-band spectrum of two separately prepared samples of 3 at 145 K and 150 K after irradiating the quartz tube with 302 nm UV light at room temperature for 1 hour.Main spectrum shows g ~2 peak and also low-field transition at g ~4, characteristic of the presence of a triplet state.Inset with spectrum centred on main transition.2,12

Figure S26 .
Figure S26.X-band spectra of 1.py at room-temperature (292 K) before (black) and after (red) irradiating the quartz tube with 302 nm UV light for 1 hour in air.The irradiated sample rapidly oxidised under air and was yellow in colour.Spectra collected with the same spectrometer settings (mw power, time constant, conversion time, modulation frequency, and modulation amplitude).After irradiation a new paramagnetic signal is observed at g ⊥ = 2.007 and g ‖ = 2.021 consistent with formation of a superoxide species.2

Figure S27 .
Figure S27.X-band spectrum (black) and simulated (red) spectrum of 1.py at room-temperature (292 K) after (red) irradiating the quartz tube with 302 nm UV light for 1 hour in the presence of air as an oxidant.The irradiated and oxidised sample was a yellow colour in contrast to the purple colour observed without air present.Simulated spectra models an axial spin systems with g ⊥ = 2.007 and g ‖ = 2.021 and gaussian linewidth of 9.1G.

Figure S28 .
Figure S28.X-band spectra of 3 at room-temperature (292 K) before (black) and after (red) irradiating the quartz tube with 302 nm UV light for 16 hours in air.The irradiated sample oxidised under air and was yellow in colour.Spectra collected with the same spectrometer settings (mw power, time constant, conversion time, modulation frequency, and modulation amplitude).After irradiation a new paramagnetic signal is observed at g ⊥ = 2.007 and g ‖ = 2.020 consistent with formation of a superoxide species.2

Figure S29 .
Figure S29.X-band spectrum (black) and simulated (red) spectrum of 3 at room-temperature (292 K) after (red) irradiating the quartz tube with 302 nm UV light for 16 hours in air.Simulated spectra models axial spin systems with g ⊥ = 2.007 and g ‖ = 2.020 and , gaussian linewidth of 7.2G.

Figure S30. 1 H
Figure S30. 1 H (quantitative) and 13 C echo-detected MAS NMR spectra of 1.py before and after illumination recorded at 23 T, 30 kHz MAS and 298 K. † indicates an experimental artefact.

Figure Supporting Note 1 .
Figure Supporting Note 1. Unit cells of 1 when co-crystallised with different solvent molecules.Toluene, THF, isopropanol carbons highlighted in yellow for clarity.

Figure Supporting Note 2 .
Figure Supporting Note 2. Graph showing Ti(1)-O i Pr bond lengths from crystal structure models of 1.py for each Ti environment.Error bars drawn at ± 3σ.Crystal 1 data points in red, crystal 2 data points in black.Triangle = exposed to X-rays for one data collection; square = exposed to X-rays for two data collections; circle = exposed to X-rays for three data collections; star = exposed to X-rays for four data collections; asterisk = exposed to X-rays for five data collections.

Table S1 . Observations during photoirradiation of powdered samples
.61 †irradiated structures of irradiated 1.py should have 73 H atoms, however, due to the disordered acetone/isopropoxide/isopropanol sites it is not always possible to add all H at a sensible geometry, which leads to a deficit of H in the crystallographic model.