Competing Nonadiabatic Relaxation Pathways for Near-UV Excited ortho-Nitrophenol in Aqueous Solution

Nitrophenols are atmospheric pollutants found in brown carbon aerosols produced by biomass burning. Absorption of solar radiation by these nitrophenols contributes to atmospheric radiative forcing, but quantifying this climate impact requires better understanding of their photochemical pathways. Here, the photochemistry of near-UV (λ = 350 nm) excited ortho-nitrophenol in aqueous solution is investigated using transient absorption spectroscopy and time-resolved infrared spectroscopy over the fs to μs time scale to characterize the excited states, intermediates, and photoproducts. Interpretation of the transient spectroscopy data is supported by quantum chemical calculations using linear-response time-dependent density functional theory (LR-TDDFT). Our results indicate efficient nonradiative decay via an S1(ππ*)/S0 conical intersection leading to hot ground state ortho-nitrophenol which vibrationally cools in solution. A previously unreported minor pathway involves intersystem crossing near an S1(nπ*) minimum, with decay of the resulting triplet ortho-nitrophenol facilitated by deprotonation. These efficient relaxation pathways account for the low quantum yields of photodegradation.


S1. Further Computational Results
Figure S1a shows transitions from the gas-phase ground state minimum, calculated at the ωB97X-D3/ZORA-def2-TZVP level of theory in the gas phase using ORCA. 1,2 he two peaks in the UV/vis spectrum of aqueous oNP at 350 nm and 278 nm correspond to transitions labelled S 2 and S 4 respectively.LR-TDDFT also identifies two transitions with zero oscillator strength to states S 1 and S 3 .There is an offset of around 0.7 eV between the theory and experiment, which can be partly explained by the absence of a description of solvation in the computed results, and partly by a tendency of the functional to overestimate excitation energies. 3gure S1b-d also show transitions from the gas phase ground state minimum, calculated using LR-TDDFT at the ωB97X-D3/ZORA-def2-TZVP level of theory with a water CPCM (Figure S1b), and calculated by ADC(2)/aug-cc-pVDZ in the gas phase (Figure S1c) and with a water CPCM (Figure S1d).The ordering of the states S 1 and S 2 is particularly sensitive to solvation modelling.It should also be noted that when calculated by ωB97XD/aug-cc-pVDZ in the gas phase using Gaussian, 4 the bright ππ* state is formally S 1 at the FC geometry.ADC(2) gives a better match to the experimental results, particularly when solvation is modelled.Figure S2 shows how the energies of excited singlet and triplet states change, along the LIIC calculated in Figure 3 in the main manuscript, when vertical transitions are calculated with solvation modelled by a water CPCM.The same key and intermediate geometries were used.The recalculated LIIC using a CPCM shows a small energy barrier along the LIIC towards the S 1 (π*)/S 0 intersection seam (IS).There are two competing arguments for whether the associated excited-state intramolecular proton transfer (ESIPT) is a barrierless process in solution or not.On the one hand, inclusion of an explicit protic solvent may further increase the energy barrier along the LIIC.On the other hand, the given LIIC involves concerted proton transfer and NO 2 torsion through a series of intermediate geometries; there may exist a lower energy pathway deviating from the LIIC, such as a stepwise process, which may avoid this energy barrier.Figure S4 shows natural transition orbitals describing the character of S 1 and S 2 at key geometries.
These allow the characterization of these states as nπ* or ππ*.
From the LIICs and NTOs (Figure S3 and Figure S4) it is also apparent that the S 2 (nπ*) surface may itself consist of more than one diabatic surface.Near the FC region the non-bonding orbital is primarily located on the oxygen atoms of the nitro group, and nearer the intersection seam the nonbonding orbital has greater density on the oxygen of the phenol group.The change in energies of these diabatic surfaces along the LIIC can partly be explained by the proton transfer between the phenol and nitro groups.

S2. Equations
The SOC magnitudes reported in this work were calculated for each molecular structure using the following equation to account for all the triplet sublevels.

S3. Further Transient Absorption Spectroscopy Data
Transient absorption spectroscopy data obtained over extended time delays using electro-optical delay methods (Figure S6) were used to determine the time constants of deprotonation of oNP to form oNP -, and subsequent re-protonation.Time constants are give in Table 1 of the main manuscript.TRIR spectra of aqueous (in D 2 O) solutions of oNP obtained using only optical delays and a probe in the range 1450 cm -1 to 1650 cm -1 show the effect of added acid on the TRIR spectrum.There is evidence of the appearance of a peak at 1500 cm -1 (A) in the late time in the solution without added acid, which could be attributed to the anion.The spectra are otherwise alike.S7).Bounds of integrals were determined by edges of positive or negative features at 100 fs, excepting f' where the dead pixel at ~1587 cm -1 was avoided.Data are fitted to bi-or monoexponential decays, over the range 1-100 ps, except for f', which is fitted in the range 5 -100 ps.Time constants are given in Table S1.A simple fit of integration regions covering the two largest bleach and HGSA features in the early time data (Figure S8, Table S1) gives time constants in the order of ~3 ps for the decay of both the GSB and HGSA features which are directly comparable to the HGSA decay seen in the TA spectra.
The growth of HGSA d' is comparable to the 400-fs time constant seen in the TA and attributed to decay from the first bright state.Figure S8 also shows that the majority of the GSB recovers within the first 20 ps.
In the first 2-3 ps a feature overlaps GSB f and its corresponding HGSA f'.For GSB f this has been fitted with a time constant of 1.7 ps.Only the decay of f' has been fitted, once the contribution from this overlapping signal has reduced.The overlapping signal could be attributed to the bleach e, time zero artifact, ESA from the excited bright state, or distribution of vibrational energy across modes.

S5. Comparison of Kinetics in
TA data for oNP in D 2 O indicate that the time constants for all processes differ slightly from those in water.The difference in τ 1 is subtle, but the larger value in D 2 O could indicate a kinetic isotope effect, which may benefit from further study.The slower cooling of the hot ground state in D 2 O can also be explained by the different density of vibrational states of the solvent.Other time constants involving the breaking or formation of O-H/O-D bonds, which occur too slowly to be fitted from this data, will also be affected by the solvent isotope effect.For time-resolved infrared (TRIR) data collected at the University of Bristol, using the apparatus shown schematically in Figure S11, the output of a Ti:sapphire laser (Coherent Astrella -1 kHz, 800 nm, 7 W, 35 fs pulse width) was split equally between two optical parametric amplifiers (OPAs -Coherent OPerA Solo) generating tuneable pulses in the UV/visible and mid-infrared.The UV/visible pulse, used as a pump, was collimated, and passed through a 500 Hz mechanical chopper to allow sequential pump-on, pump-off measurements.The pump was aligned through a 60-cm delay stage allowing optical delays of up to ~4 ns.The tuneable broadband mid-IR output from the OPA was used as a probe, with 20% delivered to a reference 128-element MCT (Mercury Cadmium Telluride) detector mounted in an IR spectrometer.The remaining 80% was overlapped with the optical pump and focussed at the sample position, before being focussed into a matching MCT detector mounted in a separate IR spectrometer.

S6. Experimental -Description of Bristol Laser Setup
For TA measurements, a white light continuum (WLC) probe in the range 300-750 nm was generated by focusing a small portion of the 800-nm fundamental onto a CaF 2 plate.The pump and probe were focussed and overlapped at the sample position, and the probe subsequently focussed into a detector after passing through a low pass filter to remove pump scatter.
In both TA and TRIR measurements, the pump and probe were linearly polarised at the magic-angle geometry.The probe was focussed to a spot approximately half the area of the pump spot.The TA IRF was fitted with a FWHM of ~120 fs.

Figure S1 .
Figure S1.Experimental UV/vis absorption spectra of ortho-nitrophenol in aqueous solution (curves) with theoretical vertical excitations from the gas phase ground state minimum; calculated using LR-TDDFT/TDA/ωB97X-D3/ZORA-def2-TZVP in the gas phase (a) and with a water CPCM (b), and calculated using ADC(2)/aug-cc-pVDZ in the gas phase (c) and with a water CPCM (d).

Figure
Figure S3 benchmarks the ωB97X-D3/ZORA-def2-TZVP calculations against ADC(2), showing similar trends in behaviour for the lowest 3 excited singlet states along the LIIC, although the excited states have higher energy when calculated by ωB97X-D3 compared to ADC(2).

Figure S4 .
Figure S4.NTOs showing the character of the S 1 and S 2 excited singlet states, calculated by LR-TDDFT/TDA/ωB97X-D3/ZORA-def2-TZVP, at the geometries of the ground state minimum (FC -Franck-Condon), the geometry representative of the S 1 (ππ*)/S 0 intersection seam, and the S 1 (nπ*) minimum.Singular values for each pair of NTOs are reported.

Figure S5 .
Figure S5.LIIC from the ground state minimum geometry to the T 1 minimum geometry with 10 intermediate steps with energies of the ground and excited states calculated at the (LR-TD)DFT/TDA/ωB97X-D3/ZORA-def2-TZVP level of theory, and the energy of the lowest triplet state calculated by unrestricted DFT/ωB97X-D3/ZORA-def2-TZVP.

Figure S6 .
Figure S6.Kinetic traces for an integrated region from 400 -475 nm from the transient spectra of oNP in water at its intrinsic pH, collected using electro-optic delays.Plotted points are the mean of pairs of adjacent experimental data points, and the red line shows a fitted biexponential curve.

Figure S8 .
Figure S8.Kinetics of integration regions around bleaches d and f and HGSA features d' and f' of early time TRIR data gathered at Bristol (FigureS7).Bounds of integrals were determined by edges of positive or negative features at 100 fs, excepting f' where the dead pixel at ~1587 cm -1 was avoided.Data are fitted to bi-or monoexponential decays, over the range 1-100 ps, except for f', which is fitted in the range 5 -100 ps.Time constants are given in TableS1.

Figure S9 .
Figure S9.Kinetic traces extracted from the TRIR spectra of oNP in D 2 O collected at RAL using electro-optic delays, fitted with Gaussian peaks.Kinetic traces are fitted to a triexponential function in the range 4 ps -4 μs.

Water and D 2 OFigure S10 .
Figure S10.Kinetic traces for a wavelength-integrated region from 400 -600 nm, from the transient absorption spectra of oNP in D 2 O and water, in both cases at the intrinsic pH (pD) of oNP.The data from the solution in D 2 O are fitted to a modified biexponential function (red) or higher-order exponential (blue).

Figure S11 .
Figure S11.Simplified schematic diagram of the transient absorption (TA) spectroscopy and time-resolved infrared (TRIR) spectroscopy setups at the University of Bristol.

Table S1 .
Time constants derived from exponential fitting of integrated regions as shown in FigureS8.τ 1 represents a growth, τ 2 a decay.Error margins are statistical errors of the fits.

Table S3 .
Time constants derived from wavelength-integrated regions of the TA data.Error margins are statistical errors of the fits.