An experimental and computational study of the e ﬀ ect of aqueous solution on the multiphoton ionisation photoelectron spectrum of phenol †

We revisit the photoelectron spectroscopy of aqueous phenol in an e ﬀ ort to improve our understanding of the impact of inhomogeneous broadening and inelastic scattering on solution-phase photoelectron spectra. Following resonance-enhanced multiphoton ionisation via the 1 1 pp * and 1 1 ps * states of phenol, we observe 1 1 pp * – D 0 /D 1 ionisation and competing direct S 0 – D 0 /D 1 ionisation. Following resonance-enhanced multiphoton ionisation via the 2 1 pp * state, we observe the signature of solvated electrons. By comparing the photoelectron spectra of aqueous phenol with those of gas-phase phenol, we ﬁ nd that inelastic scattering results in peak shifts with similar values to those that have been observed in photoelectron spectra of solvated electrons, highlighting the need for a robust way of deconvoluting the e ﬀ ect of inelastic scattering from liquid-phase photoelectron spectra. We also present a computational strategy for calculating vertical ionisation energies using a quantum-mechanics/e ﬀ ective fragmentation potential (QM/EFP) approach, in which we ﬁ nd that optimising the con ﬁ gurations obtained from molecular dynamics simulations and using the [phenol $ (H 2 O) 5 ] QM [(H 2 O) n $ 250 ] EFP (B3LYP/aug-cc-pvdz) method gives good agreement with experiment.


Introduction
Ionisation is the most fundamental photophysical process accompanying the interaction of ionising radiation with biologically important molecules and plays a central role in radiation chemistry and biology. The formation of an electron and a radical is the rst step in a chain of chemical reactions that results in DNA damage. The most direct way to probe ionisation experimentally is to use photoelectron spectroscopy (PES), which measures the electron kinetic energies (eKEs) of electrons emitted following ionisation. The eKE distribution encodes the role of each vibrational mode of the radical in its subsequent structural relaxation and, in the case of solution-phase photoelectron spectra, also contains information about solvent relaxation. However, the interpretation of solutionphase photoelectron spectra is complicated by the inhomogeneous environment of the solution causing spectral broadening and inelastic scattering of photoelectrons in the solution before emission causing the measured electron kinetic energies to be lower than their true values. 1,2 Disentangling the various contributions to solution-phase photoelectron spectra requires gas-phase PES as an essential reference and synergistic contributions from experiment and theory. Here, we revisit the PES of aqueous phenol in an effort to improve our understanding of the impact of inhomogeneous broadening and inelastic scattering on the photoelectron spectra.
Phenol is a ubiquitous molecular motif in many biologically relevant chromophores. It is the chromophore in the amino acid tyrosine, which plays an important role in photosynthesis, 3 and it is a building block of the chromophore in green uorescent protein, the most widely used uorescent probe for in vivo monitoring of biological and biochemical processes. 4,5 The UV absorption spectrum of phenol is dominated by two bands centered around 270 nm (4.6 eV) and 210 nm (5.9 eV), corresponding to transitions from the electronic ground state, S 0 , to the rst two 1 pp* states, labelled 1 1 pp* and 2 1 pp* (Fig. 1). Between these two 1 pp* states lies a 1 ps* state, labelled 1 1 ps*. The 1 1 ps* state is composed of Ocentered p 3s and ps* congurations and is dissociative along the O-H stretch coordinate. The 1 1 ps* state forms conical intersections (CIs) with the 1 1 pp* and S 0 states at modest O-H bond lengths and, therefore, plays an important role in the photostability of chromophores containing the phenol motif. 6,7 There have been numerous experimental and computational studies of the photochemistry and photophysics of isolated phenol molecules in vacuo and in solution. 6, Gas-phase studies have revealed that following photoexcitation above the 1 1 pp*/1 1 ps* CI, the dissociative 1 1 ps* potential energy surface is accessed directly and O-H bond ssion occurs, forming phenoxyl radical and hydrogen atom products (PhOc + H) on a femtosecond timescale. 20 Following photoexcitation just below the 1 1 pp*/1 1 ps* CI, the dissociative 1 1 ps* potential energy surface is accessed by tunnelling through the barrier under the CI, forming PhOc + H on a nanosecond timescale. 14,22,23 In hexane, an aprotic solvent, the initial bond ssion processes and timescales have been found to be very similar to those in the gas phase. 35 However, in aqueous solution, new relaxation pathways are possible. Following photoexcitation of the 2 1 pp* state at 200 nm and the 1 1 pp* state below the 1 1 pp*/1 1 ps* CI, solvated electrons and PhOc radicals were observed to be formed on timescales of 200 fs and 2 ns, respectively, using transient absorption spectroscopy. 28 In both cases, autoionisation was proposed as a mechanism for the formation of solvated electrons. A recent liquid-microjet PES study by our group found that following photoexcitation of the 1 1 ps* state at 235 nm, just above the 1 1 pp*/1 1 ps* CI, IC to the 1 1 pp* state occurred on a 150 fs timescale. 31 It was also suggested that solvated electrons were formed on the same ultrafast timescale by a sequential mechanism, involving O-H bond ssion to form PhOc + H followed by proton-coupled electron transfer (PCET).
Although the electronic relaxation dynamics of photoexcited neutral phenol molecules has been studied extensively, there has been less interest in the photoionisation of phenol. PES [36][37][38][39][40] and multiphoton ionisation (MPI) mass-spectrometry 41 measurements of gas-phase phenol have determined the rst two adiabatic ionisation energies (AIEs) to be 8.508 eV (ref. 41) and 9.36 eV, 36 for the ground and rst electronically excited states of the radical ion, D 0 and D 1 , respectively. A resonance-enhanced MPI (REMPI) PES study of gas-phase phenol revealed the vertical ionisation energy (VIE) from the 1 1 pp* state to D 0 to be around 0.3 eV higher than the AIE. 42 Recent quantum dynamics calculations have identied the key vibrational modes contributing to the subsequent electronic relaxation of the radical ion following photoionisation. 33 In aqueous solution, an X-ray PES study of aqueous phenol using a liquid-microjet revealed the VIEs for D 0 and D 1 to be lowered to 7.8 AE 0.1 eV and 8.6 AE 0.1 eV, respectively. 43 44 Despite both studies yielding VIEs within experimental error of the X-ray PES data, the two experiments did not yield values that were in good agreement with each other. The photoelectron spectra recorded by us were analysed by tting the data to a single photoionisation process from 1 1 pp*, to D 0 . 31 The photoelectron spectra recorded by Roy et al. were analysed by tting the data to two ionisation processes from 1 1 pp*, to D 0 and D 1 , and by including a shi to account for inelastic scattering, estimated from photoelectron spectra of solvated electrons in aqueous solution. 44 Calculations of phenol$(H 2 O) 4 clusters, in which the VIEs were determined using the equation-of-motion coupled-cluster method with single and double excitations for ionisation potentials (EOM-IP-CCSD) 45 method for phenol perturbed by the electrostatic eld of a 35Å spherical box of water molecules modelled using the effective fragment potential (EFP) method, have given VIEs to D 0 and D 1 of 7.9 eV and 8.6 eV, respectively, 43 in agreement with the experimental measurements.
The different approaches to the analysis of liquid-microjet MPI photoelectron spectra 31,44 motivated us to revisit the MPI PES of aqueous phenol. In this paper, we compare the results of new liquid-microjet MPI PES experiments with gasphase MPI PES measurements 31 and liquid-microjet X-ray PES measurements. 43 We also compare photoionisation calculations of phenol in the gas phase with those in aqueous solution using density functional theory (DFT) and EOM-IP-CCSD methods for phenol$(H 2 O) 5 clusters perturbed by the electrostatic eld of water molecules modelled using the EFP method.

Experimental
Photoelectron spectra were recorded using our recirculating liquid-microjet magnetic-bottle time-of-ight (TOF) photoelectron spectrometer that has been described in detail elsewhere. 46 Briey, a 100 mM aqueous phenol solution, with 30 mM sodium uoride added to minimise charging effects and increase the conductivity, was introduced through a 20 mm diameter quartz nozzle into the spectrometer. The liquid-microjet was intersected with femtosecond laser pulses approximately 1 mm below the nozzle, in the region of laminar ow. The femtosecond laser pulses were generated by frequency upconverting the output of an optical parametric amplier pumped by an amplied Ti:sapphire femtosecond laser system operating at 1 kHz; the electric eld vectors of the laser pulses were parallel to the TOF axis and the 1/e 2 pulse duration of the 235.5 nm pulses was measured to be $150 fs. Photoelectrons were detected at the end of the TOF tube and the photoelectron current was recorded together with the arrival time relative to the trigger of the laser pulse. The photoelectron count-rate was kept at around 500 Hz to avoid space-charge effects and saturation of the detector. eKE spectra were determined by calibrating the TOF against the MPI photoelectron spectrum of NO 47 and multiplying the photoelectron counts by the Jacobian m e s 2 /(t À t 0 ) 3 , where m e is the mass of an electron, t is the TOF and s and t 0 are calibration constants. Photoelectron spectra of Xe were recorded to determine the energy resolution and streaming potential, which were DE/E $ 1% and f str ¼ 0, † respectively.
Compared with our previous work, 31 we have employed a recirculating system instead of a liquid nitrogen cold-trap, which improves the quality of the photoelectron spectra at low eKE. We have also rewritten the data analysis soware and corrected an error in the way the Jacobian was implemented, which results in higher photoelectron counts at low eKE (see ref. 46).

Computational
To benchmark the quantum mechanical (QM) computational methods used to calculate the VIEs and vertical excitation energies (VEEs) of aqueous phenol, calculations of isolated phenol molecules in the gas phase were performed at the same levels of theory. The structure of gas-phase phenol was optimised using the B3LYP 48-51 /6-311++G(3df,3pd) 52-54 method and frequency-calculations were performed to ensure that a minimum on the potential energy surface was reached. The VIE was determined using two methods: B3LYP/aug-cc-pvdz to determine the energy difference between neutral phenol and its corresponding cation, at the minimum energy geometry of neutral phenol, and the EOM-IP-CCSD/6-31+G* method. VEEs were calculated using the equation-of-motion coupled-cluster method with single and double excitations for excitation energies (EOM-EE-CCSD) 45 and the algebraic diagrammatic construction method to second order (ADC(2)), 55,56 both with the 6-31+G* basis set. All gas-phase calculations were performed using the QChem soware package 57 apart from the optimisation and frequency calculations which were performed using Gaussian 09. 58 Several steps were involved in the calculation of VIEs and VEEs of phenol in aqueous solution. First, a classical molecular dynamics (MD) simulation (NAMD, 59 developed by the Theoretical and Computational Biophysics Group in the Beckman Institute for Advanced Science and Technology at the University of Illinois at Urbana-Champaign) was used to sample an ensemble of conformations of phenol in bulk water (Fig. 2). In the MD simulation, phenol was soaked in a sphere of water with radius 50Å (17 877 water molecules) and the CHARMM force eld was used to model the system. [60][61][62] The system was minimised for 2 ps and then allowed to equilibrate at 300 K for a further 20 ps before running a trajectory for 150 ps. Frames from the trajectory were saved every 500 fs (300 frames in total) for subsequent quantum mechanical/EFP (QM/EFP) calculations. Hybrid QM/EFP methods provide a rigorous yet computationally affordable way to include solute-solvent interactions. 63 In the EFP region, solvent molecules are modelled as discrete entities using non-empirical model potentials that perturb the QM region by their electrostatic potentials. The EFP also includes polarisation, dispersion and exchange interaction energies at the QM/EFP interface and between the individual EFP fragments.
For the QM/EFP calculations, the QM region was selected to be phenol plus the ve water molecules closest to any atom within phenol. Although fewer water molecules have been employed for other calculations of phenol in aqueous solution, 34,43 we used ve because we found this was enough to ensure that all the water molecules that are hydrogen-bonded to phenol (donor-acceptor distance < 3Å, donor-H-acceptor bond angle 180 AE 20 ) were included in the QM region. The EFP region was selected to be all other water molecules within 10Å of any atom in the phenol molecule (250-300 water molecules). A radius of 10Å was selected because it has been shown to work well for similar calculations for the green uorescent protein chromophore in bulk water. 64 The water molecules outside the EFP region were then discarded. The EFP parameters used to represent water were the standard parameters in the Q-Chem 65 or Firey 66 libraries.
The QM and EFP selections were made independently for each of the 300 frames saved from the MD simulation and therefore each frame has a different phenol conformation as well as different congurations of water molecules, in both the QM and EFP regions. The energies of each of the 300 congurations were calculated using QM/EFP at the B3LYP/aug-cc-pvdz level of theory for both S 0 (Fig. S7 †) and D 0 using the QChem soware package. VIEs were then calculated as the difference between these S 0 and D 0 energies. Higher level calculations were then performed using a selection of the 300 frames. To make these calculations computationally affordable, we selected a relatively small set of frames and a smaller basis set. The ten congurations with S 0 energies closest to the mean S 0 energy at the QM/EFP (B3LYP/aug-cc-pvdz) level were selected (Fig. S7 †) for QM/EFP calculations using EOM-IP-CCSD, EOM-EE-CCSD and ADC(2), with the 6-31+G* basis set. These calculations were performed using the QChem soware package.
To investigate whether optimisation with QM/EFP improved the calculations, additional QM/EFP calculations of VIEs and VEEs, using B3LYP/aug-cc-pvdz, EOM-IP-CCSD/6-31+G*, EOM-EE-CCSD/6-31+G* and ADC(2)/6-31+G*, were performed for the same ten frames following optimisation at the PBE0 (ref. 67-69)/ aug-cc-pvdz level of theory. The geometry optimisation was carried out using the Firey quantum chemistry package, 70 which is partially based on the GAMESS (US) source code; 71 it adjusts the positions of all atoms within the QM region at the PBE0/aug-cc-pvdz level and the 250-300 explicit water molecules modelled by EFP are also reorientated (rotational and translational degrees of freedom).

Results and discussion
In Fig. 3, we present 1 + 1 MPI photoelectron spectra of phenol in aqueous solution as a function of one-photon electron binding energy, eBE ¼ hn À eKE where hn is photon energy, together with the equivalent gas-phase photoelectron spectra. 31 The photoelectron spectra of phenol in aqueous solution are shied to lower eBEs by around 0.8 eV compared to the gas phase and are similar to those reported in our earlier work; 31 however, as a result of the improved quality of data obtained using a recirculator compared to a liquid nitrogen cold-trap (Section 2.1) we are able to identify additional features. The 275-249.7 nm photoelectron spectra recorded following resonance-enhanced MPI via the 1 1 pp* state are now best t with two Gaussians, corresponding to 1 1 pp*-D 0 /D 1 ionisation processes. At 275 nm, the area of the peak corresponding to ionisation to D 0 is around three times larger than that of the peak corresponding to ionisation to D 1 , in agreement with our earlier calculations of photoionisation cross-sections from the 1 1 pp* state. 31 This contrasts with the 267 nm MPI photoelectron spectrum reported by Roy et al. in which the area of the peak corresponding to ionisation to D 0 was substantially less than the area of the peak corresponding to ionisation to D 1 . 44 The ratios of the areas of the two peaks corresponding to ionisation from 1 1 pp* to D 0 and D 1 are observed to decrease with increasing photon energy (Fig. 3), unlike the calculations; 31 this could be attributed to increased solute and solvent reorganisation during ionisation to D 1 compared to ionisation to D 0 . The residuals of the ts at the low eBE edges of the 265.5-249.7 nm spectra are plotted as insets in Fig. 3 and can be attributed to non-resonant S 0 -D 0 MPI that competes with resonance-enhanced MPI. The contribution from non-resonant MPI increases with decreasing S 0 -1 1 pp* absorption cross-section, as we would expect. This feature was not observed in either of the previous MPI studies of aqueous phenol but is observed in the MPI gas-phase PES. 31,44 The 235.5 nm (5.26 eV) MPI photoelectron spectrum is very broad and can be t with either three or four Gaussians; however, we believe that tting to three Gaussians is more appropriate (see below). There are four processes contributing to the photoelectron spectrum t to three Gaussians: resonant 1 1 pp*-D 0 /D 1 MPI and non-resonant S 0 -D 0 /D 1 MPI. At this photon energy, the photoelectron spectra corresponding to 1 1 pp*-D 0 and S 0 -D 1 lie on top of one another and it is not possible to distinguish between them by tting an additional Gaussian. The peak centered at 8.0 AE 0.1 eV two-photon eBE corresponds to S 0 -D 0 ionisation and is close to the X-ray PES measurement. 43 At 235.5 nm, the 1 1 pp* state is not populated directly ( Fig. 1) but our observation of 1 1 pp*-D 0 /D 1 ionisation is consistent with photoexcitation of the 1 1 ps* state followed by rapid relaxation to the 1 1 pp*/1 1 ps* CI, aer which some population will undergo IC to the 1 1 pp* state before photoionisation, on the timescale of the measurement ($150 fs).
In our earlier work, the 235.5 nm MPI photoelectron spectrum was t to four Gaussians. The additional Gaussian was attributed to the photoelectron spectrum of the solvated electron that we proposed was formed following relaxation through the 1 1 pp*/1 1 ps* CI aer which, in addition to IC to the 1 1 pp* state, O-H dissociation could occur to form PhOc + H, followed by proton-coupled electron transfer, H(aq) + H 2 O / H 2 O + (aq) + e À (aq). It is possible that there is a contribution from solvated electrons to the peak we have assigned as S 1 -D 1 .
The 199 nm (6.23 eV) MPI photoelectron spectrum, which has not been reported before, is dominated by a peak centred around 5.25 AE 0.1 eV eBE with a long tail at low eBE. The peak centred around 5.25 AE 0.1 eV eBE can be t with a single Gaussian and corresponds to 11.5 AE 0.1 eV two-photon eBE, which we attribute to ionisation from the 1b 1 molecular orbital of water. 72 The residual of this t is plotted as an inset and it can be t with a Gaussian centred around 4.0 AE 0.1 eV (one-photon eBE) which we attribute to the photoelectron spectrum of the solvated electron. This value of eBE lies between the values of 4.5 eV and 3.7 eV obtained from careful measurements of the photoelectron spectra of solvated electrons at 5.8 eV and 13.6 eV, respectively, 2 and its observation is consistent with transient absorption measurements of solvated electrons being formed on a 200 fs timescale following 200 nm excitation. 28 The residual on the low eBE edge of the solvated electron photoelectron spectrum can be attributed to S 0 -D 0 ionisation (Fig. S4 †). Now we consider the effect of inelastic scattering on our photoelectron spectra. Recent careful measurements of UV photoelectron spectra of solvated electrons revealed that the measured eBE gradually increased with photon energy, indicating that the photoelectron energy diminished as a result of electron-solvent molecule inelastic scattering before emission from the surface of the liquid. 1 Subsequent scattering simulations quantied the role of inelastic scattering on the photoelectron spectra. 2 In order to investigate the impact of inelastic scattering on the peak positions and widths of our liquid-microjet photoelectron spectra, we plotted the eKEs of the maxima and full-width half-maxima (FWHM) of the Gaussians tted to the 1 1 pp*-D 0 /D 1 processes, as a function of photon energy, alongside those for the 1 1 pp*-D 0 process for gas-phase phenol (Fig. 4). The peak widths do not seem to vary substantially with photon energy. The overall shapes of the three sets of data are very similar, with the peak eKEs remaining the same for both 275 nm and 265.5 nm spectra but then increasing approximately linearly. The gradient of the line tted to the linearly increasing component of the plot for the gas-phase data is 1.02 AE 0.02, indicating that the propensity for Fig. 4 Plots of fitted Gaussian peak maxima (data points) and full-width half maxima (shaded areas) corresponding to S 1 (1 1 pp*)-D 0 in the gas phase (g) and S 1 (1 1 pp*)-D 0 /D 1 in aqueous solution (aq), as a function of photon energy (bottom axis) and wavelength (top axis). Solid straight lines are fits to the higher photon energy data points, with gradients m indicated. Dashed lines are peak maxima estimated using S 0 -D 0 /D 1 VIEs obtained from Xray PES 43 and the S 0 -S 1 (1 1 pp*) AEE determined from the UV-vis absorption spectrum (Fig. 1, orange dashed line), assuming that vibrational energy is conserved during photoionisation. conserving vibrational energy during the 1 1 pp*-D 0 photoionisation process holds extremely well over this energy range. In contrast, the gradients of the lines t to the peak positions corresponding to 1 1 pp*-D 0 /D 1 processes in aqueous phenol are less than unity. Moreover, these lines are shied from the positions estimated using S 0 -D 0 /D 1 VIEs obtained from X-ray PES 43 and the S 0 -1 1 pp* AEE determined from the UV-vis absorption spectrum (4.46 eV, Fig. 1), assuming that vibrational energy is conserved during photoionisation (dashed lines in Fig. 4). We used VIEs from X-ray PES measurements rather than our own measurement at 235.5 nm because we have not deconvoluted inelastic scattering from our 235.5 nm spectrum. Although it is possible that the propensity for conserving vibrational energy does not hold for aqueous phenol, we believe this is unlikely because it holds so well for gas-phase phenol, the UV-vis spectra of gas-phase and aqueous phenol are remarkably similar (Fig. 1) and the overall trends of the lines plotted in Fig. 4 are similar for aqueous phenol and gas-phase phenol. Thus, we believe the differences between the estimated peak positions and the actual peak positions can be attributed to inelastic electron scattering and note that our peak shis are similar to those reported in ref. 2 over the same energy range.
It is this consideration of inelastic scattering that suggests the 235.5 nm photoelectron spectrum should be t to three Gaussians rather than four. In the t to four Gaussians (Fig. S5 †), the S 1 -D 0 peak is shied to lower eBE than the Xray data whereas in the t to three Gaussians (Fig. 3), it is shied to higher eBE than the X-ray data. Although both tted eBEs can be considered to be equivalent to the X-ray data within the experimental errors of both measurements, inelastic scattering would shi the measured peak to higher eBE, which suggests that the t to three Gaussians is more appropriate. The wavelength dependence of inelastic scattering poses a particular problem for photoelectron spectra that span a wide range of eKEs, such as the 235.5 nm photoelectron spectrum. Although it is reasonable to t Gaussians to a true photoelectron spectrum, Gaussians will be distorted by a wavelength-dependent inelastic scattering shi. Therefore, it is desirable to deconvolute inelastic scattering from a measured photoelectron spectrum to obtain a true photoelectron spectrum before tting Gaussians. Unfortunately, this is not possible without detailed modelling of inelastic scattering across the relevant range of eKEs.
In Table 1, we present our measured S 0 -D 0 /D 1 peak maxima together with our calculated VIEs and experimental and calculated values from the literature. For gas-phase phenol, our B3LYP/aug-cc-pvdz and EOM-IP-CCSD/6-31+G* methods both give VIEs that are within 0.2 eV of the experimental AIEs 36,41 and are as good as other calculated VIEs reported in the literature. 31,33,43  ] EFP (B3LYP/aug-ccpvdz) method, is almost 0.4 eV higher than the value obtained from X-ray PES measurements. 43 The average VIE for the ten probable congurations (Section 2.2) calculated using the same method is slightly higher. Interestingly, using the smaller 6-31+G* basis set does not make much difference. The average VIE calculated using the EOM-IP-CCSD/6-31+G* method is around 0.3 eV higher than that calculated using the B3LYP/6-31+G*.
We found that optimisation of the ten probable [phenol$(H 2 O) 5 ] QM [(H 2 O) n$250 ] EFP congurations lowered the calculated VIE values by around 0.1-0.3 eV. The average VIE calculated for the ten probable congurations following optimisation and using the B3LYP/aug-cc-pvdz method was 7.93 eV, which is in good agreement with the Xray PES measurement. Again, using the smaller 6-31+G* basis set made little difference. The average VIEs calculated for the optimised ten probable congurations using the EOM-IP-CCSD/6-31+G* method are around 0.4 eV higher than the experimental values, although the difference between the S 0 -D 0 and S 0 -D 1 values (0.9 eV) is reasonably close to the difference measured using X-ray PES. These calculations suggest that optimisation of congurations obtained from MD simulations could be important for determining accurate VIEs and that the simple DFT approach seems to work particularly well. It will be interesting to test this procedure for calculating VIEs on other molecules in aqueous solution to see if it is general rather than specic to phenol. Curiously, our EOM-IP-CCSD/EFP method, which includes water molecules that are hydrogen-bonded to phenol in the QM region, does not agree as well with the experimental measurements as the EOM-IP-CCSD/EFP method employed by Ghosh et al., which only included phenol in the QM region. 43 However, it is worth noting that Ghosh et al. applied a correction to account for the effect of increasing basis set from 6-31+G(d) to cc-pVTZ.  Table 1 Calculated VIEs and measured IEs (or peak maxima) from S 0 to D 0 and D 1 in eV FWHM of the S 0 -D 0 photoelectron spectrum obtained from the t to the 235.5 nm spectrum (Fig. 3) is signicantly larger (1.34 eV, similar to the X-ray PES FWHM 43 ). The difference could be attributed to solute and solvent reorganisation, which is not accounted for in the simulation. The eKE dependence of inelastic scattering can also affect the widths of measured photoelectron spectra but to account for this properly requires detailed modelling of the electron scattering process. 2

Conclusion
We have reported new MPI PES measurements of phenol in aqueous solution recorded using our recirculating liquid-microjet apparatus. Following resonant MPI via the 1 1 pp* state, the improved quality of these photoelectron spectra of phenol compared to those reported previously has allowed us to identify 1 1 pp*-D 0 and 1 1 pp*-D 1 ionisation processes and competing direct S 0 -D 0 ionisation. Following resonant MPI via the 2 1 pp* state, we have observed the signature of solvated electrons. Following resonant MPI via the 1 1 ps* state, we observed 1 1 pp*-D 0 /D 1 and S 0 -D 0 /D 1 processes and, although we no longer nd evidence for the formation of solvated electrons, we cannot rule out the possibility that solvated electrons are formed. Time-resolved PES measurements will be able to identify whether or not solvated electrons are formed following photoexcitation of the 1 1 ps* state and such measurements are planned in our laboratory. The VIEs of photoexcited states of biologically relevant molecules in aqueous solution underpin ionisation and charge transfer processes and are thus important in radiation chemistry and biology. Solvated electrons, or more precisely presolvated electrons, are also known to play a role in inducing damage to DNA in aqueous solution. By comparing the MPI photoelectron spectra of aqueous phenol and gas-phase phenol, we found that inelastic scattering resulted in peak shis similar to those reported for photoelectron spectra of the solvated electron. 1,2 The wavelength dependence of inelastic scattering poses a particular problem for interpreting broad photoelectron spectra and highlights a need for a robust way of deconvoluting the effect of inelastic scattering from liquid-phase photoelectron spectra. Quantifying the inelastic scattering of low energy electrons in aqueous solution is also crucially important for improving our understanding of the role of (pre-) solvated electrons in inducing damage in DNA in aqueous solution.
We have developed a QM/EFP protocol for calculating the VIEs of aqueous phenol. We found that DFT with a reasonably large basis set performed well. We also found that optimising the congurations obtained from MD simulations improved the value for the VIE. It will be interesting to investigate whether optimising greater numbers of congurations improves the agreement with experiment and to test the protocol on other molecules. The FWHM of our simulated S 0 -D 0 photoelectron spectrum was less than that obtained from our t to experimental data, which we attribute to solute and solvent reorganisation. Calculating solute and solvent reorganisation energies during ionisation is challenging, but we believe that high quality liquid-microjet photoelectron spectra together with analogous measurements in the gas-phase provide ideal benchmarks for theory.

Conflicts of interest
There are no conicts to declare.