MOLECULAR PHYSICS, 10 MARCH 2004, VOL. 102, NO. 5, 499–506 Excited state proton transfer and internal conversion in o-hydroxybenzaldehyde: new insights from non-adiabatic ab initio molecular dynamics NIKOS L. DOLTSINIS* Lehrstuhl für Theoretische Chemie, Ruhr-Universität Bochum, 44780 Bochum, Germany (Received 25 November 2003; accepted 16 January 2004) A recently developed non-adiabatic ab initio molecular dynamics method coupling the S1 excited electronic state to the ground state by a surface hopping technique has been applied to study photoexcited o-hydroxybenzaldehyde. Vertical excitation of the enol tautomer to the pp* state leads to spontaneous proton transfer (PT) and thus formation of the keto tautomer. The PT process has been found to be entirely driven by the breathing mode of the H-chelate ring. In fact, there exists no barrier for PT in the S1 state. Non-radiative decay of the pp* excited keto tautomer through internal conversion (IC) is observed on a picosecond time scale. The probability for a non-adiabatic surface hop from the S1 state back to the ground state is seen to be correlated to the chelate ring breathing mode and to the temporal changes in the methoxy OH bond length. There are indications of an increased IC rate associated with the initial PT event. 1. Introduction Excited state proton transfer (PT) and internal conversion (IC) are two of the most fundamental photochemical processes. They play an important role in numerous biological systems such as DNA, where PT causes genetic damage [1] and excited state radiationless decay is thought to be fast [2–6] in order to prevent any photochemical reactions. The smallest aromatic molecule that shows both PT and IC is o-hydroxybenzaldehyde (OHBA) (see figure 1 for the structures of the two tautomers involved); it is therefore the ideal candidate for studying these photochemical processes under well-defined conditions. According to experimental evidence, photoexcitation to the S1 electronic state of the enol tautomer [7] first results in ultrafast tautomerization leading to the keto form by intramolecular PT Figure 1. Schematic view of the OHBA enol and keto tautomers. After photo-excitation of the enol form, hydrogen HT is transferred from the donor oxygen, OD , to the acceptor oxygen, OA . *e-mail: [email protected] in less than 50 fs followed by fast IC on the picosecond time scale [7, 8]. From an experimental point of view it has become increasingly clear that the excited state reaction barrier for PT must be either very small or even non-existent [7, 8] judging by the extremely short time scale measured for PT. Ab initio electronic structure calculations do, on the whole, confirm this picture; there remain some doubts, however, concerning the precise shape of the excited state potential [9–12]. Beyond these purely static aspects, interesting questions have been raised by a recent experimental study [8] as to the dynamical mechanism of PT, in particular with regard to the role of the skeletal vibrational modes of the H-chelate ring. Perhaps even more intriguing is the problem of IC in OHBA. Although this process is well studied experimentally and excited state lifetimes are known to a considerable degree of accuracy [7, 8], a coherent account of the underlying reaction mechanism has yet to emerge. The most promising reaction scheme so far has been put forward by Sobolewski and Domcke [9, 11], who have proposed that de-excitation may proceed via a conical intersection between the ground state and a dissociative n * or p * state. In order to reach the conical intersection the system has to overcome a potential barrier in the pp* state in the direction perpendicular to the PT coordinate, i.e. involving hydrogen detachment. This paradigm has been highly successful in explaining the non-radiative decay of a number of aromatic molecules [11, 13, 14]. Molecular Physics ISSN 0026–8976 print/ISSN 1362–3028 online # 2004 Taylor & Francis Ltd http://www.tandf.co.uk/journals DOI: 10.1080/00268970410001668408 500 Nikos L. Doltsinis However, in the case of OHBA the absence of any deuteration effects on the measured IC rates hints at the existence of an alternative mechanism. In an attempt to shed light on the dynamics of photo-excited OHBA we have carried out non-adiabatic ab initio molecular dynamics (MD) simulations on coupled S0 and S1 potential energy surfaces (PES). A fulldimensional treatment of a system with many degrees of freedom, such as the present one, calls for an electronic structure method which cheaply and reliably provides excited state energies and nuclear gradients. In order to further reduce the computational effort it is necessary to employ a so-called on the fly MD scheme whereby one avoids calculation of the entire PES prior to the simulation and only solves the electronic structure problem along the classical MD trajectory. These criteria are fulfilled by the density functional restricted openshell Kohn–Sham (ROKS) [15, 16] method as implemented within the framework of Car–Parrinello MD (CP-MD) [17, 18]. In the present work, we use a recent non-adiabatic extension of CP-MD (na-CP-MD) [16, 19, 20] which couples the Kohn–Sham density functional electronic ground state to the ROKS S1 excited state by means of a surface hopping scheme [16, 20–24]. 2. Theoretical approach 2.1. Car–Parrinello surface hopping A detailed description of the Car–Parrinello surface hopping method has been published elsewhere [16, 19, 20]; it shall only be reviewed briefly here. We have adopted a mixed quantum-classical picture treating the atomic nuclei according to classical mechanics and the electrons quantum mechanically. The total electronic wavefunction of the system, C, is taken to be a linear combination of adiabatic state functions Fj , C¼ X Z aj Fj exp i Ej dt= h , ð1Þ where n is half the (even) number of electrons. Separate variational optimization of the two wavefunctions genand ð1Þ erally results in the molecular orbitals ð0Þ l l being different. Hence, for the overlap matrix elements, Sij ¼ hFi jFj i we obtain S01 ¼ S10 S and Sii ¼ 1. Therefore, the non-adiabatic coupling matrix elements, Dij ¼ hFi jð@=@tÞjFj i, are Dii ¼ 0 and Dij ¼ Dji . Substitution of (1) into the TDSE and integration over the electronic coordinates following multiplication by F*i from the left yields the coupled equations of motion for the wavefunction coefficients 1 p1 p1 ia SðE E Þ þ a D a D S 1 0 1 1 01 0 10 S2 1 p0 p0 1 p0 a0 D10 a1 D01 S ia1 S2 ðE0 E1 Þ , ð2Þ a_ 1 ¼ 2 S 1 p1 a_ 0 ¼ R where pj exp ði Ej dt=hÞ and Hii ¼ hFi jHjFi i ¼ Ei , H01 ¼ H10 ¼ E0 S. In the CP formalism, computation of the nonadiabatic coupling elements, Dij , is straightforward and efficient, since the orbital velocities, _l , are available at no additional cost due to the underlying dynamical propagation scheme. If, instead of being dynamically propagated, the wavefunctions are optimized at each point of the trajectory (so-called Born–Oppenheimer mode), the non-adiabatic coupling elements are calculated using a finite difference scheme. Numerical integration of (2) yields the expansion coefficients ai , whose square moduli, ja0 j2 and ja1 j2 , could be interpreted as the occupation numbers of ground and excited states, respectively, if the two adiabatic wavefunctions were orthonormal. However, since the definition of state populations in this basis is ambiguous, we project the wavefunction C onto an orthonormal set of auxiliary wavefunctions, F0i ¼ c0i F0 þ c1i F1 , ci ¼ ðc0i , c1i Þ being a solution of the eigenvalue problem Hci ¼ Sci Ei , j where Ej is the energy expectation value of wavefunction Fj and the time-dependent expansion coefficients aj are to be determined such that C is a solution to the timedependent electronic Schrödinger equation (TDSE), _ . In the present case, our basis functions are HC ¼ i hC the S0 closed-shell KS ground state determinant, ð0Þ ð0Þ ð0Þ F0 ¼ jð0Þ 1 1 n n i C ¼ d0 F00 þ d1 F01 ¼ b0 F0 þ b1 F1 , ð3Þ where bj ¼ aj pj . Transformation to this orthonormal basis leads to the square moduli of the expansion cofficients di jd0 j2 ¼ jb0 j2 þ S2 jb1 j2 þ 2S < ðb*0 b1 Þ, 2 2 2 jd1 j ¼ ð1 S Þjb1 j , ð4Þ ð5Þ and the ROKS S1 excited state wavefunction [15], E E 1 n ð1Þ ð1Þ ð1Þ ð1Þ ð1Þ ð1Þ ð1Þ F1 ¼ 1=2 ð1Þ 1 1 n nþ1 þ 1 1 n nþ1 , 2 which add up to unity and can be regarded as proper state populations. Following Tully’s fewest switches criterion [22] recipe, the non-adiabatic transition Excited state proton transfer and internal conversion in o-hydroxybenzaldehyde ð6Þ where t is the MD time step. A hop from surface i to surface j is carried out when a uniform random number >Pi (and Pi >0), provided that the potential energy Ej is smaller than the total energy of the system. The latter condition rules out any so-called classically forbidden transitions. After each surface jump atomic velocities are rescaled in order to conserve total energy. In the case of a classically forbidden transition, we retain the nuclear velocities, since this procedure has been demonstrated to be more accurate than alternative suggestions [25]. 2.2. Technical details The above surface hopping algorithm has been implemented in the CPMD package [18, 26]. For the example discussed below, the KS and ROKS equations were solved using the BLYP exchange-correlation functional [27, 28] in a plane wave basis truncated at 70 Ry in conjunction with Troullier–Martins pseudopotentials [29]. Cut-off radii of 1.05, 1.23 and 0.5 au were employed for the construction of the 2s, 2p pseudo-wavefunctions of O and C, and the 1s pseudowavefunction of H, respectively. We chose a periodically repeated simple cubic unit cell of length 22.68 au. All MD simulations were carried out in the Born– Oppenheimer mode with a comparatively small time step of t ¼ 1 au, the latter being due to the necessity to compute the wavefunction time derivative from a finite difference scheme. The state amplitudes ai were integrated using a standard fourth-order Runge–Kutta scheme [30] with a time step of 0.04 au. 3. Results and discussion 3.1. Proton transfer At optimized ground state geometry (enol form), our ROKS calculation yields a vertical excitation energy to the S1 (pp* ) state of 3.12 eV. Experimental values range from 3.8 to 3.9 eV [7, 31–34] suggesting that ROKS underestimates the vertical excitation energy by 0.7 to 0.8 eV, in accord with previous observations [15, 16, 35] regarding ROKS energy gaps. It is believed, however, that ROKS excited state potential energy surfaces have qualitatively correct shapes and are merely subjected to a constant energy shift. Support for this argument can be found by comparing our calculated energy gap of 1.65 eV at optimized S1 keto geometry to the experimental fluorescence maximum between 2.4 and 2.5 eV [7, 31–34]. Thus the ROKS energy shift at this point of the PES is 0.75 to 0.85 eV, very similar to the shift for the enol vertical excitation energy, R(OO) [Å] d jdi j2 =jdi j2 , dt R(OH) [Å] Pi ¼ t resulting in a theoretical estimate for the Stokes shift of 1.47 eV in surprisingly good agreement with the experimental value of 1.4–1.5 eV [7, 31–34]. We have carried out adiabatic on the fly molecular dynamics simulations at various temperatures using the ROKS S1 potential in order to probe the hypothetical barrier for PT. For all temperatures spontaneous PT was observed, suggesting the non-existence of such a barrier in accordance with previous CASSCF and TDDFT calculations of mimimum energy profiles [11]. Figure 2 (c) shows the S1 energy profile as a function of the simulation time after vertical excitation of the ground state optimized enol tautomer at a constant temperature of 1 K. The S1 energy is seen to continuously decrease while both donor–acceptor OO distance, ROD OA , and the distance between the transferring hydrogen (HT ) and the accepting oxygen (OA ), ROA HT (see also figure 1 for nomenclature) shorten (figures 2 (a) and (b)). Simultaneously, the original hydroxyl bond length, ROD HT , steadily grows (figure 2 (b)) and does not, as might have been suspected, oscillate about any potential enol equilibrium value. In the present example, the geometrical transition state, i.e. ROA HT ¼ ROD HT , is reached in roughly T z ¼ 50 fs (figure 2 (b)). As the new methoxy OH bond is formed, we notice a sharp drop in energy, which levels off at about 60 fs (figure 2 (c)). Figure 2 thus demonstrates that the geometrical PT transition state can give a lower bound for the PT time, but observable energetic changes take place with a delay of several femtoseconds. Due to the absence of an excited state enol minimum it is not straightforward to define an enol–keto energy splitting. A purely static estimate may be obtained by 2.6 a) 2.5 1.5 b) 1 E [eV] probability from state i to any other state is 501 3 2.5 c) 0 time [fs] 100 Figure 2. (a) OD OA distance, (b) OD HT (– –) and OA HT (—) distances, and (c) excited state energy relative to ground state minimum as a function of simulation time at a temperature of 1 K. Nikos L. Doltsinis taking the difference between the vertical enol excitation energy at optimized S0 geometry and the keto S1 minimum yielding a value of 0.78 eV. However, we have determined a dynamical ensemble average at 300 K of 0:4 0:1 eV (see end of this section). Experimentally an enol–keto splitting of approximately 0.5 eV has been measured [7], whereas CASPT2 and TDDFT ab initio electronic structure calculations predict 0.52 and 0.29 eV, respectively [11]. A detailed comparison of ROKS S1 data with previous theoretical as well as experimental results is given in table 1. In order to gain insight into the PT process under laboratory conditions, we have calculated 20 nonadiabatic surface hopping trajectories initially propagating the system in the S1 electronic state at a vibrational temperature of 300 K. For each run the starting configuration has been picked at random from a ground state CP-MD simulation at 300 K. We observe spontaneous PT in the space of T z ¼ 25:1 15:4 fs (ensemble average standard deviation), measuring the time between vertical excitation and the moment the geometrical PT transition state, i.e. ROD HT ¼ ROA HT is reached. In order to enable meaningful comparisons with results from photoelectron spectroscopy, however, it is important that we define PT in terms of changes in electronic structure which lead to an experimentally observable decrease in energy once the keto tautomer has been formed. As we have discussed above, these changes take place several femtoseconds after the geometrical PT transition state has been reached. Therefore the time for PT of T z ¼ 25:1 15:4 fs stated here should be seen as a lower bound. As figure 2 demonstrates the excited state energy continues to fall well beyond the geometrical PT transition state. Experimentally measurable PT times should therefore lie above 25 fs (an upper limit may be placed at 50 fs, since the keto minimum is reached at 2T z in figure 2, but here the system is not allowed to gather kinetic energy beyond 1 K). Time-resolved photoelectron spectroscopy places an upper bound for the PT time at 50 fs [7]. Stock et al. [8] report a PT time of 45 fs from transient absorption measurements. Considering the limitations of our theoretical model and the statistical uncertainty the agreement is remarkable. Analysing the changes in the donor–acceptor oxygen– oxygen distance during PT at room temperature (see figure 3), it becomes apparent that shortening of the OO distance by roughly 10% of its enol equilibrium value is a prerequisite for PT to occur. Therefore, we conclude that the breathing mode of the chelate ring should be rate determining, in agreement with previous proposals [8, 36, 37]. A rough estimate of the corresponding vibrational frequency can be obtained by measuring the Table 1. Comparison of ROKS S1 data to CASPT2 and TDDFT results reported previously in the literature [11] as well as experimental numbers [7, 31–34]. ROKS is seen to underestimate the energy gap between ground and excited states. However, the good agreement with experiment in the case of the Stokes shift and the enol–keto splitting (the number in parentheses is the relevant dynamical splitting, see text for details) indicates that the ROKS S1 potential shape is qualitatively reasonable. E CASPT2 TDDFT ROKS vertical 3.74 3.97 3.12 adiabatic fluorescence 3.22 2.41 3.68 3.05 2.34 1.65 Stokes shift 1.33 0.92 1.47 enol–keto 0.52 0.29 0.78 (0.4) Exp. 3.8 3.9 3.4 2.4 2.5 1.5 1.4 0.5 [34], [7, 31–33] [7] [31, 32], [33, 34] [31, 32], [33, 34] [7] 3 distance [Å] 502 2 1 0 100 200 time [fs] 300 400 Figure 3. Typical time evolution of the OD OA (- -), OD HT (– –) and OA HT (—) distances after vertical photoexcitation at time t ¼ 0 for an initial vibrational temperature of 300 K. period of the OO oscillation shown in figure 3. The vibrational energy of 256 cm1 determined this way is indeed close to transmission oscillation frequencies of 275–296 cm1 observed experimentally by Stock et al. [8]. Figure 3 further shows that after PT the newly formed methoxy OH bond is vibrationally excited, but the kinetic energy rather rapidly dissipates into other degrees of freedom. In order to obtain a dynamical estimate for the enol– keto energy splitting, we have calculated ensemble averages of the difference between the S1 energy at t ¼ 0 (vertical excitation) and at t ¼ 100, 200, 300, 400 and 500 fs (see figure 4 for details). Additional time averaging over the interval [100, 500] fs gives a dynami- Excited state proton transfer and internal conversion in o-hydroxybenzaldehyde 0.7 ∆E [eV] 0.5 0.3 0.1 0 100 200 300 400 time [fs] 500 600 Figure 4. S1 energy difference between vertical excitation (t ¼ 0) and t ¼ 100, 200, 300, 400 and 500 fs averaged over the ensemble of trajectories. The error bars represent the respective standard deviations and the horizontal dashed line is the time average. 20 S1 population 15 10 5 0 0 200 400 600 time [fs] 800 1000 Figure 5. S1 population as a function of time from surface hopping trajectories (). The three exponential fits which have been obtained (i) imposing that all molecules occupy the S1 state at t ¼ 0 (—), (ii) without boundary condition (– –) and (iii) including only the data points at t 300 fs (- - -), give relaxation times of 333, 405 and 841 fs, respectively. cal splitting of 0:4 0:1 eV, in good agreement with the experimental estimate of approximately 0.5 eV [7]. 3.2. Internal conversion In an attempt to unravel the mechanism by which radiationless decay of the keto S1 excited state takes place, we have analysed 20 non-adiabatic surface hopping trajectories obtained for different initial 503 conditions sampled from a ground state MD run at 300 K. In order to get an estimate of the IC rate, we have calculated the S1 population as a function of time (see figure 5) starting from the time of vertical excitation (t ¼ 0). Similar to the experimental procedure [7] the time constant for radiationless decay may be obtained from an exponential fit. Imposing the correct initial condition that all molecules are in the excited state at t ¼ 0 thus yields a decay time of 333 fs. However, it is apparent from figure 5 that this fit does not describe our data points too well, the short time decay is too slow whereas the long time decay is too fast. Removing the boundary condition at t ¼ 0 gives a somewhat better fit with a decay time of 405 fs. Experimentally, data points near t ¼ 0 are usually neglected since they are part of the pulse width limited coherent artefact [7]. Following this recipe, we have performed a fit taking into account only the data points at t 300 fs leading to an IC time of 841 fs. It should be stressed here that all our theoretical values stated above are lower bounds, since we restricted the length of the trajectories to a maximum of about 1 ps at which point three molecules had still not undergone a transition to the ground state. Our lower bound of 841 fs should be compared to the experimental rate [7] of 0:62 0:15 ps1 determined at an excitation energy of 4.34 eV, corresponding to a decay time between 1.3 and 2.1 ps. An MD simulation at 300 K corresponds to a vibrational energy of 0.5 eV and therefore, assuming an electronic excitation energy of 3.9 eV, to an experimental measurement at 4.4 eV. Having made sure that our calculated relaxation times are in line with experiment, we shall now discuss in detail the underlying mechanism. As a first step, we have monitored the energy gap between the S0 and S1 electronic states in order to detect any conical intersections or avoided crossings (see figure 6). Perhaps somewhat unexpectedly, the energy gap remains larger than 1 eV over the whole length of the simulation. The dominant energetic changes are seen to occur during the initial PT phase, where the energy separation is reduced by almost 2 eV. Our findings present an alternative to the decay mechanism proposed by Sobolewski and Domcke [10, 11], who have shown that a conical intersection between an n * or p * excited state and the ground state can be reached if a potential barrier in the S1 state along the hydrogen HT dissociation coordinate is overcome. We have verified that the Sobolewski– Domcke scenario is plausible and, more importantly, that our ROKS approach is capable of describing such a case. Figure 7 demonstrates that the ROKS lowest singlet excited state indeed adopts n * character when hydrogen atom HT is moved to a distance of ROD HT ¼ ROA HT ¼ 1:99 A (otherwise assuming ground state geometry). For the sake of completeness, we 504 Nikos L. Doltsinis energy [eV] 3 2 1 0 0 200 400 600 time [fs] Figure 6. Ground (- -) and first excited state (—) energies relative to the initial S0 energy and the corresponding energy gap (—) as a function of time for an initial vibrational temperature of 300 K. Figure 7. ROKS S1 singly occupied molecular orbitals (a) at enol S0 equilibrium geometry (pp* character) and (b) at ROD HT ¼ ROA HT ¼ 1:99 A (n * character). mention that for ROD HT ¼ ROA HT ¼ 1:65 A ROKS still yields the pp* state. These observations are compatible with the CASPT2 calculations on malonaldehyde by Sobolweski and Domcke [10, 11]. However, our dynamical treatment shows that a vibrational temperature of 300 K in the S1 state is not sufficient to cross the dissociative barrier and thus reach the conical intersection. Nevertheless, at higher excitation energies this decay channel may still play a dominant role in the IC process. Our preliminary conclusion is that since the Sobolewski–Domcke decay channel is not accessed in our simulations there must be another efficient decay mechanism. In the following, we shall establish which OHBA vibrational modes in the pp* excited state couple non-adiabatically to the ground state and thus provide an alternative IC mechanism. For this purpose, we compare the time evolution of the non-adiabatic transition probability to that of various geometric variables. In figure 8 the probability for a non-adiabatic surface hop from the excited state to the ground state (see equation (6)) and the donor–acceptor OO distance are plotted as a function of the simulation time. Clearly, the two are correlated: small OO distances coincide with peaks in the transition probability envelope, whereas large OO distances coincide with small transition probabilities. Furthermore, for all our trajectories the decay probability is highest at the moment of PT (see figure 8 for a typical example). Partially, this can be explained by our observations for PT in the previous subsection, where we have seen that PT occurs at short OO distances close to the OO turning point. Thus if the condition for PT is fulfilled the non-adiabatic decay probability is automatically large. On the other hand, there are points of the trajectory where in spite of the OO distance being clearly shorter than during PT the transition probability does not reach the same level. Therefore, although the chelate ring breathing mode is certainly chiefly responsible for non-adiabatically coupling the S1 and S0 electronic states, there must be additional modes causing modulations of the decay probability. Figure 9 reveals that there is also correlation, on a somewhat smaller time scale, between the surface hopping probability and the velocity of the OA HT stretching motion. In essence, the faster the hydrogen atom HT moves, the higher the probability for a surface jump. Again, this is certainly another factor contributing to the large peak in the non-adiabatic transition probability in the vicinity of the PT event. Generally speaking, the non-adiabatic coupling elements Dij ¼ hFi jð@=@tÞjFj i become large when at least one of the adiabatic wavefunctions involved changes rapidly as the nuclear configuration changes. This is clearly the case during the PT reaction when the excited state wavefunction has to adjust from the enol to the keto electronic structure. Non-adiabatic coupling should be additionally enhanced by the momentum gathered as the system falls into the keto potential well, in particular as a large portion is initially deposited in the two main coupling modes, i.e. the OO and OH vibrations. We believe that this increased non-adiabatic transition probability caused by the initial tautomerization is responsible for the higher IC rate in the first few hundred femtoseconds compared to the long time decay (cf. figure 5). It is worth pointing out that the proposed main decay mechanism via the chelate ring breathing Excited state proton transfer and internal conversion in o-hydroxybenzaldehyde 3.1 PT 0.001 2.7 R(OO) [Å] P21 2.9 2.5 –0.001 0 200 time [fs] 2.3 400 Figure 8. Time evolution of the probability (equation (6)) for a non-adiabatic surface jump from the S1 to the S0 PES (—) and the OD OA distance (—). The vertical dashed line indicates the moment the geometric PT transition state is reached, i.e. ROD HT ¼ ROA HT . 0.1 0 v(OH) [Å/fs] P21 0.001 –0.001 20 40 60 time [fs] 80 100 –0.1 Figure 9. Time evolution of the probability (equation (6)) for a non-adiabatic surface jump from the S1 to the S0 PES (—) and the time derivative, i.e. velocity, of the OA HT distance (—). vibration should be influenced only little by deuteration. This is in accord with the absence of any isotope effects in photoelectron spectroscopic measurements [7]. 4. Conclusions We have performed non-adiabatic ab initio molecular dynamics simulations of the OHBA photocycle coupling the standard density functional ground state and the ROKS S1 excited state using a surface hopping scheme. Ultrafast proton transfer interconverting the enol and keto tautomers has been observed driven by the H-chelate ring breathing mode, which modulates the donor–acceptor oxygen–oxygen distance. We have determined the time at which the geometrical PT transition state is reached to be 25:1 15:4 fs. During 505 the proton transfer reaction, the oxygen–oxygen distance is seen to shorten by about 10%. Radiationless decay of the S1 excited state has been found to be fast, taking place on a picosecond time scale in the absence of any conical intersections. We have been able to demonstrate that the n * conical intersection decay channel, although being present in the direction orthogonal to the PT mode, is not accessible at moderate excitation energies. Rather, it is primarily the chelate ring breathing mode that is responsible for the non-adiabatic coupling between the pp* excited state and the ground state. The secondary coupling mode is found to be the OH stretching vibration. Moreover, there are indications for an increased decay rate in the first few hundred femtoseconds caused by the enol–keto tautomerization. The author is indebted to D. Marx for his ongoing support. W. Domcke and S. Lochbrunner are kindly acknowledged for informative discussions. Computational resources were provided by the computing centres of RUB and Rechenverbund NRW. This project is partially funded by the Deutsche Forschungsgemeinschaft. References [1] FRIEDBERG, E. C., 2003, Nature, 421, 436. [2] KANG, H., JUNG, B., and KIM, S. K., 2003, J. chem. Phys., 118, 6717. [3] PECOURT, J. M. L., PEON, J., and KOHLER, B., 2001, J. Am. chem. Soc., 123, 10370. [4] REUTHER, A., IGLEV, H., LAENEN, R., and LAUBEREAU, A., 2000, Chem. Phys. Lett., 325, 360. [5] GUSTAVSSON, T., SHARONOV, A., and MARKOVITSI, D., 2002, Chem. Phys. Lett., 351, 195. [6] GUSTAVSSON, T., SHARONOV, A., and MARKOVITSI, D., 2002, Chem. Phys. Lett., 356, 49. [7] LOCHBRUNNER, S., SCHULTZ, T., SCHMITT, M., SHAFFER, J. P., ZGIERSKI, M. Z., and STOLOW, A., 2001, J. chem. Phys., 114, 2519. [8] STOCK, K., BIZJAK, T., and LOCHBRUNNER, S., 2002, Chem. Phys. Lett., 354, 409. [9] SOBOLEWSKI, A. L., and DOMCKE, W., 1994, Chem. Phys., 184, 115. [10] SOBOLEWSKI, A. L., and DOMCKE, W., 1999, J. phys. Chem. A, 103, 4494. [11] SOBOLEWSKI, A. L., and DOMCKE, W., 1999, Phys. Chem. chem. Phys., 1, 3065. [12] SCHEINER, S., 2000, J. phys. Chem., 104, 5898. [13] SOBOLEWSKI, A. L., and DOMCKE, W., 2001, J. phys. Chem. A, 105, 9275. [14] SOBOLEWSKI, A. L., DOMCKE, W., DEDONDER-LARDEUX, C., and JOUVET, C., 2002, Phys. Chem. chem. Phys., 4, 1093. [15] FRANK, I., HUTTER, J., MARX, D., and PARRINELLO, M., 1998, J. chem. Phys., 108, 4060. [16] DOLTSINIS, N. L., and MARX, D., 2002, J. Theor. Comp. Chem., 1, 319. 506 Nikos L. Doltsinis [17] CAR, R., and PARRINELLO, M., 1985, Phys. Rev. Lett., 55, 2471. [18] MARX, D., and HUTTER, J., 2000, Modern Methods and Algorithms of Quantum Chemistry, edited by J. Grotendorst (Jülich: NIC); for downloads see www.theochem.ruhr-uni-bochum.de/go/cprev.html. [19] DOLTSINIS, N. L., and MARX, D., 2002, Phys. Rev. Lett., 88, 166402. [20] DOLTSINIS, N. L., 2002, Quantum Simulations of Complex Many-Body Systems: From Theory to Algorithms, edited by J. Grotendorst, D. Marx and A. Muramatsu (Jülich: NIC); for downloads see www.fz-juelich.de/nic-series/volume10/doltsinis.pdf. [21] TULLY, J. C., and PRESTON, R. K., 1971, J. chem. Phys., 55, 562. [22] TULLY, J. C., 1990, J. chem. Phys., 93, 1061. [23] TULLY, J. C., 1998, Classical and Quantum Dynamics in Condensed Phase Simulations, edited by B. J. Berne, G. Ciccotti and D. F. Coker (Singapore: World Scientific). [24] TULLY, J. C., 1998, Modern Methods for Multidimensional Dynamics Computations in Chemistry, edited by D. L. Thompson (Singapore: World Scientific). [25] MÜLLER, U., and STOCK, G., 1997, J. chem. Phys., 107, 6230. [26] HUTTER, J., BALLONE, P., BERNASCONI, M., FOCHER, P., FOIS, E., GOEDECKER, S., MARX, D., PARRINELLO, M., [27] [28] [29] [30] [31] [32] [33] [34] [35] [36] [37] and TUCKERMAN, M., CPMD’3.4, 2001 (Stuttgart: MPI für Festkörperforschung; Zurich: IBM Zurich Research Laboratory). BECKE, A. D., 1988, Phys. Rev. A, 38, 3098. LEE, C., YANG, W., and PARR, R. C., 1988, Phys. Rev. B, 37, 785. TROULLIER, N., and MARTINS, J. L., 1991, Phys. Rev. B, 43, 1993. PRESS, W. H., TEUKOLSKY, S. A., VETTERLING, W. T., and FLANNERY, B. P., 1999, Numerical Recipes in Fortran 77, Vol. 1 (Cambridge: Cambridge University Press). CATAÑLAN, J., TORIBIO, F., and ACUÑA, A. U., 1982, J. phys. Chem., 86, 303. NAGAOKA, S., and NAGASHIMA, U., 1989, Chem. Phys., 136, 153. NAGAOKA, S., HIROTA, N., SUMITANI, M., YOSHIHARA, K., LIPCZYNSKA-KOCHANY, E., and IWAMURA, H., 1984, J. Am. chem. Soc., 106, 6913. MORGAN, M. A., ORTON, E., and PIMENTEL, G. C., 1990, J. phys. Chem., 94, 7927. LANGER, H., and DOLTSINIS, N. L., 2003, J. chem. Phys., 118, 5400. HEREK, J. L., PEDERSEN, S., BAÑARES, L., and ZEWAIL, A. H., 1992, J. chem. Phys., 97, 9046. LOCHBRUNNER, S., WURZER, A. J., and RIEDLE, E., 2000, J. chem. Phys., 112, 10699.
© Copyright 2026 Paperzz