Excited-state intramolecular proton transfer to carbon atoms: nonadiabatic surface-hopping dynamics simulations

Excited-state intramolecular proton transfer (ESIPT) between two highly electronegative atoms, for example, oxygen and nitrogen, has been intensely studied experimentally and computationally, whereas there has been much less theoretical work on ESIPT to other atoms such as carbon. We have employed CASSCF, MS-CASPT2, RI-ADC(2), OM2/MRCI, DFT, and TDDFT methods to study the mechanistic photochemistry of 2-phenylphenol, for which such an ESIPT has been observed experimentally. According to static electronic structure calculations, irradiation of 2-phenylphenol populates the bright S 1 state, which has a rather flat potential in the Franck–Condon region (with a shallow enol minimum at the CASSCF level) and may undergo an essentially barrierless ESIPT to the more stable S 1 keto species. There are two S 1 /S 0 conical intersections that mediate relaxation to the ground state, one in the enol region and one in the keto region, with the latter one substantially lower in energy. After S 1 - S 0 internal conversion, the transient keto species can return back to the S 0 enol structure via reverse ground-state hydrogen transfer in a facile tautomerization. This mechanistic scenario is verified by OM2/MRCI-based fewest-switches surface-hopping simulations that provide detailed dynamic information. In these trajectories, ESIPT is complete within 118 fs; the corresponding S 1 excited-state lifetime is computed to be 373 fs in vacuum. Most of the trajectories decay to the ground state via the S 1 /S 0 conical intersection in the keto region (67%), and the remaining ones via the enol region (33%). The combination of static electronic structure computations and nonadiabatic dynamics simulations is expected to be generally useful for understanding the mechanistic photophysics and photochemistry of molecules with intramolecular hydrogen bonds.


Introduction
Excited-state intramolecular or intermolecular proton transfers are elementary processes occurring in many molecular and biochemical systems [1][2][3][4][5] and electronic devices, [6][7][8][9] for example, in natural and artificial photosynthesis, 10,11 water-splitting photocatalysis, 12 green fluorescent proteins, 13,14 and photoswitches. 15Understanding these excited-state proton transfer processes is important both from fundamental and technological points of view.7][18][19][20][21][22][23][24][25][26][27][28][29][30] Most of these previous studies focused on excited-state proton transfer processes between two highly electronegative atoms, e.g.27,[31][32][33][34][35][36][37] What about excited-state intramolecular proton transfer (ESIPT) in molecules without strong hydrogen bonds, for example, in alcohols or phenols?Such ESIPT processes were first investigated in the 1980s, 38,39 with proton transfer to aromatic carbon atoms being first addressed at the beginning of this century. 402][43][44][45][46][47][48] They first studied photochemical deuterium incorporation at the ortho and para positions of 2-phenylphenol in various solvent mixtures 49 and found that the predominant exchange at the ortho position is independent of water and methanol contents, implying an intramolecular process.They also investigated the photochemistry of o-hydroxybiaryls, which features not only an efficient excited-state proton transfer to the ortho carbon atom of the naphthyl ring, but also a novel ring-closing reaction. 50Flegel et al. 51 studied the photoaddition of water and alcohols to the 9-and 10-positions of the anthracene moiety of 9-(2-hydroxyphenyl)anthracene in acetonitrile and methanol mixtures and proposed a mechanism involving water-mediated excited-state proton transfer from the phenolic OH group to the anthracene fragment.Basaric ánd Wan 52 investigated the potential excited-state proton transfer in four derivatives of 9-(2-hydroxyphenyl)anthracene.Nayak and Wan 53 explored photochemical deuterium incorporation in extended orthosubstituted biaryl systems and reported the longest solvent-assisted proton-relay chain.They proposed direct and water-assisted proton transfer mechanisms to explain photohydration at the ortho and distal positions, respectively.In these experimental studies, it was generally believed that excited-state intramolecular proton transfer to ortho positions is efficient in phenols.
The underlying photophysical and photochemical mechanisms in these systems have not yet been elucidated in detail, e.g., with regard to the relevant structures, proton transfer paths, excited-state potential energy surfaces, lifetimes, and decay channels.We are aware of only one recent theoretical study in this context, 54 which employed the single-reference second-order coupled cluster (RI-CC2) method to explore direct and water-assisted excited-state proton transfer in 2-phenyl-1-naphthol.Given this situation, we decided to perform high-level multi-reference electronic structure computations and trajectory-based surface-hopping dynamics simulations to study the mechanistic photochemistry of the prototypical 2-phenylphenol molecule, with emphasis on the ESIPT process to the ortho carbon atom and the deactivation channels leading back to the ground state.

OM2/MRCI method
[79][80] During geometry optimizations, all required energies, gradients, and nonadiabatic coupling elements were computed analytically.Conical intersections were optimized using the Lagrange-Newton approach. 81,82n OM2/MRCI calculations, the restricted open-shell HF formalism was applied in the self-consistent field (SCF) treatment (i.e., the orbitals were optimized for the leading configuration of the S 1 state with two singly occupied orbitals).The active space in the MRCI calculations included 12 electrons in 10 orbitals (see ESI †).In terms of the SCF configuration, it comprised the five highest doubly occupied orbitals, two singly occupied orbitals, and the three lowest unoccupied orbitals.For the MRCI treatment, three configuration state functions were chosen as references, namely the SCF configuration and the two closed-shell configurations derived therefrom (i.e., all singlet configurations that can be generated from the HOMO and the LUMO of the closed-shell ground state).The MRCI wavefunction was built by allowing all single and double excitations from these three references.
The nonadiabatic dynamics was studied by performing 1 ps OM2/MRCI trajectory surface-hopping simulations.The initial atomic coordinates and velocities were randomly selected from 5 ps trajectories of ground-state molecular dynamics.The number of excited-state dynamics runs was then chosen according to the computed S 0 -S 1 transition probability.A total of 193 surface-hopping trajectories were run, with all relevant energies, gradients, and nonadiabatic coupling vectors being computed on-the-fly as needed.For points with an S 1 -S 0 energy gap of less than 10 kcal mol À1 , the fewest-switches criterion was applied to decide whether to hop.The time step was chosen to be 0.1 fs for nuclear motion and 0.0005 fs for electronic propagation.The unitary propagator evaluated at a mid-point was used to propagate the electronic motion.The translational and rotational motions were removed in each step.The empirical decoherence correction (0.1 a.u.) proposed by Granucci et al. was employed. 83The final evaluations were done for the 148 trajectories that finished successfully and satisfied our energy continuity criterion (no changes greater than 30 kcal mol À1 between any two consecutive MD steps).9][84][85][86]

Ground-state properties and vertical excitation energies
Apart from the most stable ground-state structure of 2-phenylphenol (S0-ENOL), there is also a minimum for the keto tautomer (S0-KETO), see Fig. 1 and Table 1.For each of the two minima, OM2/MRCI and CASSCF yield similar geometries.S0-ENOL is more stable than S0-KETO by 34.7 (33.7) kcal mol À1 at the OM2/MRCI (MS-CASPT2) level.
The computed vertical excitation energies for the first excited singlet state (S 1 ) are collected in Table 2 obtained from laser flash photolysis of 2-phenylphenol in solutions. 49A slightly lower experimental value of 4.28 eV has been reported for 2-phenyl-1-naphthol featuring more extensive conjugation. 54The S 1 state at the Franck-Condon point is spectroscopically bright; its oscillator strength is computed to be 0.155 at the TD-CAM-B3LYP level.Molecular orbital analysis shows that the S 0 -S 1 electronic transition mainly originates from the HOMO-LUMO single excitation (Fig. 2).The HOMO is mainly localized in the phenolic part, whereas the LUMO is localized in the phenyl group.Hence, the S 1 state is of charge-transfer character, which sets the stage for the subsequent excited-state proton transfer.In fact, this kind of electronic structure change has been found in many similar intramolecularly hydrogen-bonded systems. 18,20,22,26,27,30cited-state minima At the CASSCF level, there is a shallow S 1 minimum in the Franck-Condon region of 2-phenylphenol (S1-ENOL), which is computed to lie 103.0 kcal mol À1 above the S0-ENOL minimum in single-point MS-CASPT2 calculations.At the OM2/MRCI   level, no such S 1 minimum could be located since all minimizations starting from the S0-ENOL equilibrium geometry led directly to the S 1 keto species (S1-KETO), see the left panel of Fig. 3. Likewise, the minimum-energy path for proton transfer computed at the RI-ADC(2)/def2-SVP level indicates an essentially barrierless excited-state enol-keto tautomerization, see the right panel of Fig. 3.This is also verified by OM2/MRCI nonadiabatic dynamics simulations (vide infra).CASSCF optimization yields another S 1 minimum in the keto region (S1-KETO, see Fig. 1).At this geometry, there is no significant charge transfer in the S 1 state; the S 0 -S 1 transition involves mostly the central C3QC5 double-bond region, causing an elongation of this bond from 1.396 to 1.480 Å (S 0 versus S 1 keto minimum, CASSCF values).Other geometric parameters change only slightly (Fig. 1).According to single-point MS-CASPT2 calculations, S1-KETO lies 74.4 kcal mol À1 above S0-ENOL and 28.6 kcal mol À1 below S1-ENOL (Table 1).Thus, the excited-state proton transfer that yields the keto species is highly exothermic; in other words, 2-phenylphenol is a strong photoacid in S 1 .As already mentioned, this proton transfer is computed to be essentially barrierless and is thus expected to be ultrafast.In terms of excited-state topology, our present results are consistent with recent RI-CC2 computations on a similar system, 2-phenyl-1naphthol. 54

Conical intersections
At the OM2/MRCI level, we were able to locate two S 1 /S 0 minimum-energy conical intersections (S1S0-ENOL and S1S0-KETO).Selected bond lengths and dihedral angles are given in Fig. 1 and in Table 1, respectively.In S1S0-ENOL, the H10 atom is still attached to the O9 atom (phenol species) but the OH group is extruded out of the ring plane, with a C4C3C2O9 dihedral angle of À831 (OM2/MRCI).This strong out-of-plane deformation significantly increases the S 0 energy, thus closing the S 0 -S 1 energy gap and reaching an S 1 /S 0 conical intersection.In S1S0-KETO, the H10 atom is already bonded to the C7 atom (keto species); the two rings are not coplanar with a C2C3C5C7 dihedral angle of 581.Energetically, S1S0-ENOL [S1S0-KETO] is computed to lie 92.1 [73.7] kcal mol À1 above the S0-ENOL ground state, and 21.4 kcal mol À1 [39.8 kcal mol À1 ] below the S 1 energy at the Franck-Condon point (113.5 kcal mol À1 ); hence, these two conical intersections are energetically accessible.Taking into consideration that S1S0-KETO is more stable than S1S0-ENOL by 18.4 kcal mol À1 , the former is expected to play a more vital role in excited-state deactivation.
It is worth stressing that OM2/MRCI and the ab initio methods give similar structures and energies for the two S 1 /S 0 conical intersections (Table 1).Taking S1S0-ENOL as an example, the dihedral angles C2C3C5C7, C4C3C2O9, and C3C1C2O9 are computed to be À281, À831, and 1411 at the OM2/MRCI level, compared with À351, À971, and 1551 at the CASSCF level, respectively, (see Fig. 1).The relative energies from OM2/MRCI and single-point MS-CASPT2 calculations are also reasonably close to each other: the values of S1S0-ENOL [S1S0-KETO] are 92.1 [73.7] kcal mol À1 for OM2/MRCI, and 98.8/100.7 [69.8/73.9]kcal mol À1 for MS-CASPT2.In the latter case, the quoted S 0 and S 1 state energies differ slightly because they come from single-point MS-CASPT2 calculations at CASSCF-optimized geometries.

Excited-state decay paths
The preceding static electronic structure computations suggest the following scenario for the photoinduced processes in 2-phenylphenol.Upon irradiation, the spectroscopically bright S 1 state is populated in the Franck-Condon region, from which the S 1 /S 0 conical intersection with an intact phenol moiety is energetically accessible (with relaxation to the ground state via S1S0-ENOL).A competitive process involves an essentially barrierless excited-state proton transfer yielding an S 1 keto minimum, which can decay to the ground state via the S 1 /S 0 conical intersection in the keto region (S1S0-KETO); back in the S 0 state, the keto species S0-KETO can return to the more stable tautomer S0-ENOL via reverse ground-state hydrogen transfer.
To verify this mechanism and to explore the timescales of the underlying photophysical and photochemical events, we have performed trajectory-based fewest-switches surface-hopping dynamics simulations starting in the S 1 state of 2-phenylphenol.

Hopping-point distribution
The S 1 -S 0 hopping-point distribution extracted from all surfacehopping trajectories reflects the topology of the conical intersection seam. 84The two types of S 1 /S 0 conical intersections in 2-phenylphenol, S1S0-ENOL and S1S0-KETO, clearly govern our nonadiabatic dynamics simulations.Fig. 4 depicts the distributions of the C7H10 distance and the C4C3C2O9 dihedral angle  at all S 1 -S 0 hopping points.Obviously, there are two main hopping regions, which cluster around two minimum-energy S 1 /S 0 conical intersections S1S0-ENOL and S1S0-KETO.A closer examination of the C7H10 distance distribution at all hopping points in Fig. 4 shows that most of the trajectories (67%) hop to the S 0 state via the keto conical intersection seam.This preference arises from two factors: first, the S 1 proton transfer is essentially barrierless so that the S 1 keto species is generated easily, and second, S1S0-KETO is thermodynamically favored over S1S0-ENOL because its potential energy is lower by 18.4 kcal mol À1 (OM2/MRCI).Hence, it is not surprising that most trajectories decay to the S 0 state via S1S0-KETO in our dynamics simulations.

S 1 lifetime
In our simulations, 118 of 148 (80%) trajectories have reached the S 0 state at the end of the 1 ps nonadiabatic dynamics runs.As shown in the left panel of Fig. 5, most of the S 1 -S 0 hops happen between 100 and 400 fs (only 4 hops after 400 fs).Again, this ultrafast decay is consistent with the excited-state topological features, i.e. an almost barrierless proton transfer and two efficient deactivation channels (vide supra).
The S 1 excited-state deactivation can be viewed as a firstorder elementary reaction.The S 1 state population is thus ruled by the following rate equation: where k is the corresponding rate constant; p 0 is the S 1 population at the end of the run (0.2 in this work); and t 0 is the initial delay time (57 fs).After fitting the time-dependent state population in Fig. 4 to eqn (1), we obtain an S 1 excitedstate lifetime of 373 fs.One should note that the S 1 excited-state lifetime may be expected to increase in the condensed phase, in particular in a rigid environment.

Product distribution
Fig. 6 shows the product distribution at the end of the 1 ps nonadiabatic simulations.Overall, there are four kinds of products, namely S 1 enol (13%) and keto (5%), and S 0 enol (47%) and keto (35%); the enol : keto ratio is estimated to be 3 : 2. The top panel illustrates the distribution of the resulting phenol conformers.Most of the trajectories ending up in the phenol region have the H10 atom bonded to the O9 atom.However, there are also some S 0 phenol products that have the H8 atom bonded to the O9 atom, not the H10 atom (see H8O9 distribution).In these trajectories, the excited-state proton transfer and the reverse ground-state hydrogen transfer involve two different hydrogen atoms (H10 and H8, respectively).The bottom panel depicts the distribution of the keto conformers after the 1 ps simulations.There are ca.90 trajectories with the H8 atom bonded to the C7 atom, and ca.40 trajectories with the H10 atom bonded to the C7 atom.

Typical trajectories
In our nonadiabatic dynamics simulations, we see three different photocycles that start from S1-ENOL and end up at S0-ENOL: (I) the S 1 state decays directly to the ground state, without excitedstate intramolecular proton transfer (ESIPT); (II) the S 1 state first evolves towards the S 1 keto species via an ultrafast barrierless ESIPT and then decays to the ground state in the keto region followed by a reverse ground-state hydrogen transfer (GSHT) involving the same migrating hydrogen atom; (III) the photocycle is the same as in case (II) except that different hydrogen atoms are involved in ESIPT and GSHT.In the following, we present for each  photocycle pattern a representative trajectory to illustrate the main photophysical and photochemical events.Fig. 7 shows a typical trajectory for case (I) with direct decay via the S1S0-ENOL conical intersection.Within the first 400 fs, the system starts to rotate around its central C3C5 bond (strong changes in the C2C3C5C6 and C2C3C5C7 dihedral angles; only small fluctuations in the C4C3C2O9 and C3C1C2O9 dihedral angles).During this process, the nonadiabatic coupling remains small and the S 1 -S 0 energy gap remains large, so there is no nonadiabatic transition.After about 400 fs, the C4C3C2O9 dihedral angle starts to decrease from 1801 to 401 at ca. 600 fs.The S 1 and S 0 states now become energetically close to each other (within 4 kcal mol À1 ) and there is a large nonadiabatic coupling; thus, a nonadiabatic S 1 -S 0 hop takes place, with relaxation of S 1 to the S 0 state.Thereafter, the C4C3C2O9 and C3C1C2O9 dihedral angles move back towards their original values (from twisted to a more planar arrangement).There is no ESIPT in this  trajectory.We emphasize that this photocycle pattern occurs only rarely in our trajectories.
Fig. 8 depicts a typical trajectory for case (II) with deactivation to the S 0 state via the S1S0-KETO conical intersection.In the initial stage of this trajectory, the O9H10 and C7H10 distances quickly increase and decrease, respectively.At ca.50 fs, the ESIPT is complete and the S 1 keto species S1-KETO is formed, which remains in the S 1 state for another 150 fs (while retaining a rather short O9H10 distance indicative of excited-state hydrogen bonding interactions).Thereafter, it decays to the S 0 state at a point where the S 1 -S 0 nonadiabatic coupling becomes very large (panel c) and the S 1 -S 0 gap is very small (panel d).Interestingly, the generated S 0 keto species does not return back to the enol region immediately; instead, it roams the keto region for additional 500 fs.Then, a reverse groundstate hydrogen transfer takes place, regenerating the S 0 enol conformer and completing the photocycle.The rotation around the C3-C5 bond starts after ca.700 fs (see the C2C3C5C6 and C2C3C5C7 dihedral angles in panel b) while the C4C3C2O9 and C3C1C2O9 dihedral angles do not vary much.
Fig. 9 presents a typical trajectory for case (III).Here, the O9H10 and C7H10 distances fluctuate around their equilibrium positions in the first 100 fs; then, they start to increase and decrease quickly.At about 110 fs, the S 1 keto species S1-KETO is formed, which stays in the S 1 state for ca.50 fs and then decays to the S 0 state at 165 fs, when the keto S 1 /S 0 conical intersection is encountered.The generated keto species roams the keto region in the S 0 state for a longer time (640 fs).After ca.800 fs, the most stable S 0 phenol conformer is regenerated via a reverse ground-state hydrogen transfer.The rotation around the C3-C5 bond starts after ca.900 fs (see the C2C3C5C6 and C2C3C5C7 dihedral angles in panel b).Interestingly, the H10 atom bonded to the O9 atom is transferred to the C7 atom in the ESIPT process, while the H8 atom originally bonded to the C7 atom is transferred to the O9 atom in the final GSHT step.

Discussion
Our results are consistent with the experiments available for 2-phenylphenol.Lukeman and Wan 49 argued that singlet reactivity is major for 2-phenylphenol, which is consistent with our computations.The S 1 excited-state proton transfer is nearly barrierless and ultrafast, so it is impossible for the system to efficiently populate triplet states in the Franck-Condon region.In addition, the S 1 -T 1 intersystem crossing in the keto region is not expected to be competitive with the efficient internal conversion from the S 1 keto species to the S 0 state.However, this intersystem crossing could become more probable in a rigid environment because the internal conversion involves a large conformational change that could be impeded by steric interactions with the environment.Furthermore, there is experimental evidence that 2-phenylphenol is a strong photoacid in the S 1 state.This point is supported by the MS-CASPT2 results (see Fig. 10), which confirm that the S 1 excited-state proton transfer is highly exothermic -S1-KETO lies 28.6 kcal mol À1 below S1-ENOL and 40.6 kcal mol À1 below the initially populated S 1 Franck-Condon point.
We emphasize in this context that our present computations are carried out in vacuum and thus only consider the intrinsic photochemistry of 2-phenylphenol, for example, direct ESIPT processes to the ortho-position, without accounting for solventassisted intermolecular proton transfer to remote sites such as para-positions.Previous electronic structure computations on a similar system 54 showed that there exists an efficient S 1 /S 0 conical intersection near the keto region, but without optimizing its structure.In this work we precisely located this kind of minimum-energy conical intersection in 2-phenylphenol, both at the OM2/MRCI and CASSCF levels (S1S0-KETO), and we explored its dynamical role in the S 1 photodynamics of 2-phenylphenol using full-dimensional surface-hopping dynamics simulations.We find that 67% trajectories decay to the S 0 state via this conical intersection in the keto region.In addition, we optimized the S 1 /S 0 conical intersection in the Franck-Condon region (S1S0-ENOL), which also plays an important role in the S 1 deactivation (33%).Thus, both conical intersections need to be considered in order to correctly understand the mechanistic photochemistry of 2-phenylphenol and its variants.
How is the S 0 isomer S0-ENOL-1 (see the bottom of Fig. 10) generated in the photodynamics of 2-phenylphenol?Experimentally, Lukeman and Wan 49 assumed that this species comes from S0-KETO via a concerted reverse hydrogen transfer and 1,5hydrogen shift.Our present dynamics simulations do not support this scenario -we do not see any 1,5-hydrogen shift in any of the trajectories.Instead, S0-ENOL-1 is generated by a simple singlebond rotation, after the hydrogen atom originally bonded to the phenyl ring has been transferred to the oxygen atom (Fig. 10).
The S 0 keto species has not yet been detected spectroscopically when using nanosecond laser flash photolysis. 49,54B3LYP calculations indicate that the tautomerization of S0-KETO to the most stable phenol conformer S0-ENOL has to overcome a small barrier of 4.0 kcal mol À1 .It might thus be possible to observe this predicted transient species using ultrafast timeresolved transient spectroscopy.

Summary
With the use of electronic structure computations and trajectory-based surface-hopping dynamics simulations, we have for the first time explored the mechanistic photochemistry of 2-phenylphenol.We have simulated the S 1 excited-state proton transfer and deactivation as well as the reverse hydrogen transfer in the S 0 state.Mechanistically, some trajectories directly evolve from the Franck-Condon region toward an enol-type S 1 /S 0 conical intersection, followed by an S 1 -S 0 internal conversion to the ground-state minimum.Most of the trajectories proceed from the Franck-Condon region to the S 1 keto species via an essentially barrierless ESIPT; the transient S 1 keto species is then de-excited to the ground state via a second S 1 /S 0 conical intersection in the keto region, followed by a quick relaxation back to the most stable phenol minimum via a reverse GSHT process (barrier of ca. 4 kcal mol À1 at the B3LYP level).The nonadiabatic dynamics simulations predict an average lifetime of 118 fs for the ESIPT process. 40,49In these simulations, 67% of the trajectories decay to the S 0 state via the keto S 1 /S 0 conical intersection, and 33% decay via the S 1 /S 0 conical intersection in the Franck-Condon region.According to the computed time-dependent state populations, the S 1 excitedstate lifetime is estimated to be 373 fs in vacuum.We hope that these computational results and mechanistic insights will stimulate further experimental work on 2-phenylphenol, especially by ultrafast time-resolved transient spectroscopy.

Fig. 3 (
Fig. 3 (left) OM2/MRCI optimization path starting from the enol minimum, which leads directly to a keto species after about 50 steps; (right) RI-ADC(2)/def2-SVP computed minimum-energy reaction path with respect to the C7-H10 proton transfer reaction coordinate.See text for discussion.

Fig. 4
Fig. 4 Distribution of the C7H10 distance and the C4C3C2C9 dihedral angle at all S 1 -S 0 hopping points.See text for detailed discussion.

Fig. 5
Fig. 5 Distribution of the S 1 -S 0 hopping times (left) and time-dependent S 1 and S 0 state populations (right).See text for detailed discussion.

Fig. 6
Fig. 6 Distribution of the C7H10, C7H8, H8O9, and O9H10 bond lengths at the end of 1 ps simulations.See text for detailed discussion.

Table 2
49mputed vertical excitation energies to the first excited singlet state of 2-phenylphenol and the experimental band maximum from laser flash photolysis in solution49 This journal is © the Owner Societies 2015 Phys.Chem.Chem.Phys., 2015, 17, 9687--9697 | 9695