Effects of substituent position on aminobenzoate relaxation pathways in solution

Temperature-dependent Ultraviolet/Visible (UV/Vis) absorption spectra of both m-MA (methyl-3-aminobenzoate, Alfa Aesar, 98%) and p-MA (methyl-4-aminobenzoate, Aldrich, 98%) dissolved in cyclohexane (CHX, Merck, 99.7%), acetonitrile (ACN, Fisher Scientific, 99.8%), and ethanol (EtOH, VWR, absolute) were obtained with an Agilent Cary 60 UV/Vis spectrophotometer equipped with a temperature control unit (TC 1 Temperature Controller, Quantum NorthWest). The absorption spectra for each solution were taken first at 15°C and then in increasing 10°C steps up to 75°C, allowing 30 minutes in between each scan for temperature equilibration. The spectra obtained following this procedure are shown in Figure S1 for m-MA, and Figure S2 for p-MA.


S2. Fluorescence Quantum Yields
The methodology followed for the determination of fluorescence quantum yields has been described in detail by Würth et al., 1 by comparing the emission intensity of the samples of interest (denoted below) with those of known standards (denoted below). Briefly, all sample and standard solutions, detailed in Table S1, were prepared at 10 -3 M concentration and were further diluted in a 1 cm pathlength quartz cuvette as required to achieve an absorbance of 0.1 at the wavelength of interest (comparable to the pump wavelength, pu, used in ultrafast experiments, see main manuscript). Five scans of the emission spectra were taken for each solution using a Horiba Scientific Fluorolog®-3 spectrofluorometer with a slit width of 5 mm. The average of these five scans for each sample was used to calculate the respective fluorescence quantum yields (Fl) using the following equation: where F and F refer to integral photon flux, taken here to be the area under the emission curves of each sample and standard, respectively. In addition, = 1 − 10 − , where is the absorbance at the wavelength of interest, and is the refractive index of the solvent for each solution at the wavelength of interest. The calculated values of Fl following this method are presented in Table S1.

S3. Fluorescence Lifetimes
The fluorescence lifetimes (Fl) of m-MA were measured in a 1 cm pathlength quartz cuvette for 0.01 mM solutions of m-MA dissolved in ACN and EtOH; the corresponding measurement for m-MA in CHX, following similar methodology, has previously been reported (Fl  1.7 ns). 2 The time-resolved emission spectra were recorded with a Horiba Fluorolog®-3 spectrofluorometer employing a 318 nm NanoLED as the photoexcitation light source. Blank traces were also collected in order to determine the instrument response associated with these measurements. The fluorescence lifetime values were extracted by fitting the time-resolved emission traces with an exponential decay convolved with a Gaussian function to account for the instrument response. The results, shown in Figure S3, yield fluorescence lifetimes of Fl = 10.24 ± 0.02 ns for m-MA in ACN, and Fl = 14.00 ± 0.03 ns for m-MA in EtOH; the quoted errors are the errors associated with the fit.  The time resolution of these measurements is 24 ps/channel, and they were ran at 2 MHz repetition rate.  The time resolution of these measurements is 24 ps/channel, and they were ran at 2 MHz repetition rate. Figure S6: TCSPC decay of p-MA in ethanol: experimental decay (red), IRF (blue), fit result (black), and residuals (green). These measurements were obtained for photoexcitation with 279 nm and by monitoring emission at 340 nm (Δem = 7 nm). The time resolution of these measurements is 24 ps/channel, and they were ran at 2 MHz repetition rate.

S4. Power Dependence Studies
Under laser irradiation conditions, multiphoton absorption is possible, and this may initiate photodynamics that are not possible under solar irradiation, for which photodynamics are almost completely induced by single-photon absorption (multiphoton effects under solar irradiation are negligible). Hence, multiphoton-induced photodynamics are outside of the scope of our work, which aims to explore the photodynamics of sunscreen UV filters. To ensure single-photon initiated photodynamics, power dependence studies are carried out for each sample studied. These studies consist of collecting TEAS data at a selected number of time delays (t) and at different pump laser powers. The logarithm of transient signal with respect to the logarithmic value of power for each time delay (log(signal) vs. log(power)) is then plotted and the resulting data is fitted using a linear function using least squares regression. The gradient of this fit provides physical information regarding the signal-power dependence, i.e. a gradient of 1 (within error) is taken to be indicative of singlephoton dynamics. Figures S7 and S8 show a representative sample of such plots, which lead us to confidently conclude that the presented photodynamics result from single-photon absorption and are, therefore, relevant to the purpose of understanding the ultrafast photodynamics of m-MA and p-MA in the context of sunscreen use.

S5. Fitting procedure
The transient electronic absorption spectroscopy (TEAS) data collected was globally fit using the Glotaran software package. 3 The algorithm employed includes several components to account for different experimental measurables. Firstly, the TEAS data obtained in the presented work and presented in the main manuscript is inherently chirped, i.e. t = 0 is different for each probe wavelength (pr), due to group velocity dispersion (GVD) artefacts, 4 which is accounted for by including a third order polynomial in the fitting algorithm. 3 This chirp effect was corrected in the TAS presented in Figures 4 and 7 in the main manuscript using the KOALA package. 5 Moreover, assuming here a sequential kinetic model → C, as is the assumption throughout the main manuscript) the global fitting algorithm in Glotaran models the data for each pr and each t with a superposition Ψ of n components : Ψ( pr , Δ ) = ∑ EADS (∆ , )EADS ( pr ) =1 In the equation above, EADS is a linear combination of exponential decays convolved with the Gaussian instrument response function (IRF, approximately 70-140 fs, see below) and EADS is the evolution associated difference spectra associated with component . For each set of initial parameters , the model is iterated until convergence is achieved and the quality of the fit is evaluated by analysis of the resulting residuals, which correspond to the difference between the fit and the raw data at each data point. This fitting procedure allows us to extract quantitative dynamical information by yielding time constants, , associated with each evolution associated difference spectra (EADS). These are the time constants reported in Table 2 of the main manuscript.

S6. Instrument Response Function
The temporal cross-section of our pump and probe laser pulses, as well as any potential solvent-only photodynamics, determine the fastest time constant () that can be extracted from our experiments. The combination of these quantities is encompassed in the instrument response function (IRF). The IRF for our TEAS experiments is determined by collecting solvent-only responses and fitting them with a Gaussian function of the form where 0 is a signal baseline offset, and are the intensity and the centre of the Gaussian curve, respectively, and is the width of the Gaussian function. The width, , of the most intense Gaussian curve required to fit the data is taken to be the IRF at the experimental conditions corresponding to the given solvent-response data; Figure S9 shows a number of such plots for a number of combinations of solvent, pump wavelengths (those used for the TEAS experiments reported on in the main manuscript) and probe wavelengths. Due to varying group velocity mismatch between the pump and the spectral components of the probe, 6,7 the IRF is probe dependent. Hence, we take the worst-case scenario and do not quote any lifetimes that are shortest than the widest IRF found which is, in the case of these experiments, approximately 140 fs.

S7. Fit Residuals
As mentioned in section S5, the quality of the fits obtained from global fitting with Glotaran is evaluated from analysis of the residuals produced as a result of the fitting procedure. These residuals correspond to the difference between the value of the fit and the value of the raw data for each data point, hence the perfect fit would yield a value of zero across all data points. Figures S10 and S11 show the residuals obtained from fitting the data presented in the main manuscript following the procedure detailed in S5. In addition, it is useful to evaluate the quality of the fit for specific probe wavelengths; these lineouts are shown in Figures S12 and S13.

S8. Supplementary Computational Results
In an attempt to locate a charge-transfer (CT) energy minimum that would justify the large, solvent-dependent Stokes shifts observed for m-MA (see main manuscript), we probed two reaction coordinates for this molecule in implicitly modelled ethanol, the solvent for which the largest Stokes shift is observed. The most obvious molecular motions that could lead to a CT energy minimum away from the Franck-Condon region in m-MA are the rotations of the amino group, akin to what was suggested and modelled for p-MA (see main manuscript), or of the ester group. As also detailed in the Computational Methods section in the main manuscript, these reaction coordinates of interest were probed in implicit ethanol along the static rotation of the bond through equidistant steps along the coordinate with vertical excitations calculated at every step at the PBE0/cc-pVTZ level of theory.
The results of the computational procedure just described reveal that there is a significant barrier to the rotation of the amino group of m-MA in its S1 state, as shown in Figure S14. This observation confirms that, like p-MA, m-MA is not a twist intramolecular charge transfer (TICT) system (see main manuscript for further discussion). In addition, the aforementioned computational procedure also revealed a significant barrier to the rotation of the ester substituent group of m-MA, as shown in Figure S15. For completeness, the equivalent rotation of the ester group in p-MA was also performed, with the results being shown in Figure S16.
As such, we were unable to find a CT energy minimum that may justify the experimental observations in m-MA following these two most obvious reaction coordinates, in particular the large, solvent-dependent Stokes shifts reported on in the main manuscript. Our excited state geometry relaxation calculations also optimized to what is likely a local minimum in the vicinity of the Franck-Condon region, and as such provide no evidence for a lower energy CT energy minimum that could account for our experimental observations. While a full search of the S1 surface of m-MA in implicitly modelled solvent could potentially locate this CT minimum, this demanding computational study is beyond the scope of this work, and it is nevertheless likely that explicit solvent models may be required to appropriately model the spectroscopic behaviour of both m-MA and p-MA. Figure S14: A plot of the calculated energies of the ground (S0, black) and first four excited states of m-MA: S1 (blue), S2 (red), S3 (pink), and S4 (green). The arrows illustrate the dihedral angle that has been rotated in order to probe for local minima, i.e. the rotation of the amino group (NH2). The origin, at 0 degrees, denotes the Franck-Condon region, with each subsequent angle being the rotation along the respective dihedral angle. These potential energy curves were produced at the PBE0/cc-pVTZ level of theory, and all calculations were conducted using implicitly modelled ethanol. Figure S15: A plot of the calculated energies of the ground (S0, black) and first four excited states of m-MA: S1 (blue), S2 (red), S3 (pink), and S4 (green). The arrows illustrate the dihedral angle that has been rotated in order to probe for local minima, i.e. the rotation of the ester group. The origin, at 0 degrees, denotes the Franck-Condon region, with each subsequent angle being the rotation along the respective dihedral angle. These potential energy curves were produced at the PBE0/cc-pVTZ level of theory, and all calculations were conducted using implicitly modelled ethanol. Figure S16: A plot of the calculated energies of the ground (S0, black) and first four excited states of p-MA: S1 (blue), S2 (red), S3 (pink), and S4 (green). The arrows illustrate the dihedral angle that has been rotated in order to probe for local minima, i.e. the rotation of the ester group. The origin, at 0 degrees, denotes the Franck-Condon region, with each subsequent angle being the rotation along the respective dihedral angle. These potential energy curves were produced at the PBE0/cc-pVTZ level of theory, and all calculations were conducted using implicitly modelled ethanol.