Insight into the Photodynamics of Photostabilizer Molecules

Solar exposure of avobenzone, one of the most widely used commercial UVA filters on the market, is known to cause significant degradation. This finding has fueled research into developing photostabilizer molecules. In an effort to provide insight into their stand-alone photoprotection properties, the excited state dynamics of the photostabilizer, 3-(3,4,5-trimethoxybenzylidene) pentane-2,4-dione (TMBP), and its phenolic derivative, 3-(4-hydroxy-3,5-dimethoxybenzylidene) pentane-2,4-dione (DMBP), were studied with ultrafast transient absorption spectroscopy. Solutions of TMPB and DMBP in ethanol and in an industry-standard emollient, as well as TMBP and DMBP deposited on synthetic skin mimic, were investigated. These experiments were allied with computational methods to aid interpretation of the experimental data. Upon photoexcitation, these photostabilizers repopulate the electronic ground state via nonradiative decay within a few picoseconds involving a twisted intramolecular charge transfer configuration in the excited state, followed by internal conversion and subsequent vibrational cooling in the ground state. This finding implies that, aside from acting as a photostabilizer to certain UV filters, TMBP and DMBP may offer additional photoprotection in a sunscreen formulation as a stand-alone UV filter. Finally, TMBP and DMBP could also find applications as molecular photon-to-heat converters.


C) Computational results
The optimized S0 and S1 geometry for both TMBP and DMBP are shown in Figure S8. Following the S0 geometry optimization, TD-DFT calculations were performed at the optimized ground state geometries to obtain their vertical excitation energies as well as the absorption wavelength presented in Table S1. The orbitals for the three lowest excitation energies are shown in Figure S9 for both TMBP and DMBP. Following this, the geometries of the molecules were optimized in their respective excited state to gain insight into any geometry change that facilitate their relaxation mechanism. The characterization of the excited state of TMBP and DMBP are shown in Figure S10. Figure S8. Optimized geometries of TMBP and DMBP in the S0 and S1 state calculated at DFT and TD-DFT PBE0/6-311++G** using implicit ethanol. The S1 geometry showed increased bond length and a twisted geometry at the C3-C2 allylic double bond.  Figure S9. Molecular orbitals (MOs) for the three lowest excitations of TMBP (left) and DMBP (right) calculated at PBE0/6-311++G** using implicit ethanol. For both molecules the excitation at their respective UV absorption λmax corresponds to a HOMO -LUMO transition, i.e. S2 and S1 in TMBP and DMBP respectively.

S10
The calculated wavenumbers and the associated vibrational modes for the optimised S0 geometry of TMBP and DMBP are shown in Table S2. These results guided the assignment of vibrational modes to the experimentally observed FTIR bands. A scaling factor was calculated for each predicted spectrum by using the most intense experimental peak as a reference. This scaling was then applied to the calculated S0 wavenumbers so that the calculated wavenumber accurately matched the reference experimental peak. This method has previously been employed for similar systems. 1,2 Table S2. Computed S0 vibrational wavenumbers and transition intensities for TMBP and DMBP in implicit ethanol using PBE0/6-311++G**. Description of the associated vibrational modes are also described. Atom labels can be found in Figure S8.

TEAS
In this work, a global fitting procedure is employed to determine a set of optimized parameters for the TEA spectra using the Glotaran software package. 3 The various experimental measurables were accounted for by the several components of the algorithm utilized. The TEA spectra obtained from our measurements and presented in the current work are inherently chirped, i.e. the position of t = 0 varies with each probe wavelength (λpr) as a result of group velocity dispersion (GVD) artefacts, which is accounted for by including a third order of polynomial in the fitting algorithm. Assuming → D…) kinetic model, the global fitting algorithm in Glotaran models the data for each λpr and t with a superposition Ѱ of n components l with the function below: The residuals between the raw data and the fit, determined by the global fitting procedure (and given in Figure S14) demonstrate the quality of the fit.

S11 ii) TVAS
For the ground state bleach recovery of the features associated with the TVAS data, an exponential decay function is used to fit the recovery kinetics without the convolution of the IRF using the function: A(λ) is the amplitude of the i th exponential decay component with lifetime i, t -t0 is the pumpprobe time delay, t0 is the offset recovery time relative to t = 0, I0 is the baseline signal offset, and n is the number of exponential decay functions required to model the data accurately.

E) TEAS additional data.
The TEAS data presented as a false colour heatmap in the main manuscript is presented as a line plots of mΔOD vs probe wavelength at selected pump-probe delay times in Figure S11.
The TEAS data measurement for both TMBP and DMBP in ethanol at a higher concentration of 30 mM as a comparison to the 1 mM measurement are shown in Figure S12. Also shown in Figure S13 are the TEAS data obtained following deposition of the CCT bulk solution of both TMBP and DMBP on the surface of a synthetic skin mimic. These data are at best the same as those obtained in the bulk solution, indicating minimum environment effects on the observed dynamics.

F) Residuals for the fits of TMBP and DMBP in their different solvent environments
The residuals from the sequential global fitting with respect to the raw TEA spectra data (i.e., the difference between the fit and the raw data at each data point) are shown in Figure S14. The smallsignal intensities of the residual compared to the raw TEA spectra demonstrate the quality of the fits.

G) Solvent alone instrument response
The TEAS measurements of the time zero solvent-only scan were recorded to obtain the instrument response function (IRF), which determines the limiting temporal resolution of the present experiments. The data are reported in Figure S15. CCT shows a very weak time-zero response following excitation at 348 nm at the pump-pulse power employed for the experiment. Furthermore, the instrument response function in CCT photoexcited at 348 nm ( Figure S15 (d)) has a strong contribution of cross-phase modulation between pump and probe pulses. However, this should not affect the conclusions of the manuscript given the timescales we are investigating in the solute, especially when DMBP is photoexcited in CCT at 348 nm. For this reason, we chose to follow the approach of Kovalenko et al. 4 , in which we use a frequency-dependent cross-correlation function to S14 model the IRF in Figure S15. The value obtained for the temporal resolution of the solvent-only time zero response is shown in each panel in Figure S15, and ranges from 70 to 100 fs.   Figure S17 shows the emission lifetime measurements for both TMBP and DMBP. The absorption and emission spectra for TMBP and DMBP are further reported in Figure S18. To add, we have also carried out emission lifetime measurements after flushing nitrogen through the samples; the data return similar lifetimes to those reported in Figure S17. Taken together (the short emission lifetimes and small Stokes shifts), we propose that the emission we observe is from the singlet state and that the long-lived component in our TEA spectra is trapped population in the S1 state. To note, we cannot rule out triplet state emission due to limitations to our experimental setup. Figure S17. Fluorescence lifetime measurements for (a) TMBP, and (b) DMBP. In both cases, data are shown as black circles, with the red line being a kinetic fit from which fluorescence lifetime is obtained. The instrument response for these measurements is also shown as blue squares.

K) Steady-state photostability of TMBP and DMBP in CCT.
Additional photostability data obtained for TMBP and DMBP in CCT are reported in Figure S20.
These data revealed that TMBP maintained its high photostability with ~2%; in contrast, a significant reduction (~30%) in the absorbance of DMBP is observed. Possible sources of the reduction in absorption could include the formation of molecular photoproduct that absorbs at a different wavelength. However, the 2 ns transient absorption profile does not show any significant difference when compared to that of TMBP ( Figure S16). Hence, it is unlikely that any photoproducts different from those formed in TMBP dissolved in CCT following irradiation are formed in DMBP. Another plausible explanation could be that upon photoexcitation of CCT, it becomes a photoacid which could then react with DMBP resulting in the decrease in the absorption spectrum ( Figure S20b). L) UV/Visible difference spectrum. Figure S21 shows the difference spectra obtained by subtracting the pre-irradiated absorption spectrum from post-irradiated (with a solar simulator) absorption spectrum for TMBP and DMBP.
For both samples in ethanol, there is a decrease in absorption at ~320 and ~340 nm for TMBP and DMBP respectively, in both solvent environments which corresponds to reduction of the absorption peak. As described in the main manuscript, the difference spectra do not closely match the absorption in the 2 ns TAS reported in Figure S16 (for TMBP and DMBP), lending to the conclusion that trapped excited state population in the singlet state is the main source of incomplete GSB recovery. Furthermore, the additional negative peak observed at 450 nm in the difference spectrum of DMBP in ethanol ( Figure S21c) Figure S21. Difference spectra between 2 hr solar like irradiation and before irradiation for TMBP in (a) ethanol, (b) CCT; and for DMBP in (c) ethanol, and (d) CCT.

M) 1 H NMR of TMBP and DMBP before and after solar irradiation
In an attempt to investigate the possible formation of any long-lived photoproducts upon UV excitation of both TMBP and DMBP, 1 H NMR spectra of separate samples prepared to 0.5 M were recorded in deuterated ethanol (ethanol-d6) before and after 5 hours of continuous irradiation under a solar simulator with irradiance equivalence to 7 suns (7000 W/m 2 ). Irradiation was achieved in a 1 cm cuvette. As shown in Figure S22, the data did not reveal any obseveable new peaks after sample irradiation, suggesting little or no photoproduct formation. However, we note that the integration of the peak assigned to hydroxy proton (-OH, labelled 9) in DMBP varies between before (1.46) and after (4.63) irradiation. Integration of all other peaks however remained the same before and after irradiation. This could result from a number of reasons, including (i) effect of water molecules in the solvent due to the hygroscopic nature of the solvent, (ii) effect of hydrogen bonding between ethanol and the phenolic group, or (iii) formation of potential photoproduct.

Before Irradiation
After Irradiation Before Irradiation After Irradiation