Control of Nonadiabatic Passage through a Conical Intersection by a

Letter pubs.acs.org/JPCL

Control of Nonadiabatic Passage through a Conical Intersection by a Dynamic Resonance Michael Epshtein, Yair Yifrach, Alexander Portnov, and Ilana Bar* Department of Physics, Ben-Gurion University of the Negev, Beer-Sheva 84105, Israel S Supporting Information *

ABSTRACT: Nonadiabatic processes, dominated by dynamic passage of reactive fluxes through conical intersections (CIs), are considered to be appealing means for manipulating reaction paths, particularly via initial vibrational preparation. Nevertheless, obtaining direct experimental evidence of whether specific-mode excitation affects the passage at the CI is challenging, requiring well-resolved time- or frequency-domain experiments. Here promotion of methylamine-d2 (CH3ND2) molecules to spectralresolved rovibronic states on the excited S1 potential energy surface, coupled to sensitive D photofragment probing, allowed us to follow the N−D bond fission dynamics. The branching ratios between slow and fast D photofragments and the internal energies of the CH3ND(X̃ ) photofragments confirm correlated anomalies for predissociation initiated from specific rovibronic states. These anomalies reflect the existence of a dynamic resonance that strongly depends on the energy of the initially excited rovibronic states, the evolving vibrational mode on the repulsive S1 part during N−D bond elongation, and the manipulated passage through the CI that leads to CH3ND radicals excited with C−N−D bending. This resonance plays an important role in the bifurcation dynamics at the CI and can be foreseen to exist in other photoinitiated processes and to control their outcome.

E

state NH2, while a single quantum of the antisymmetric stretching mode yielded exclusively NH2 (2A1). Control of nonadiabatic passage through the CI has also been reported in the case of vibrational excitation of a specific mode in the electronically excited state of thioanisole (C6H5SCH3),21,22 thioanisole-d3,23 substituted thioanisoles,24 and in thiophenol-d1.25 For example, in thioanisole it was shown that vibrational excitation of a specific mode (S-CH3 stretch) can select a particular product channel and that this excitation remains localized when passing through a CI. Therefore, it was suggested that the nonadiabatic dynamics is extremely sensitive to the nuclear configuration of the initially prepared reactive flux, eventually determining the trajectory of the reaction path in terms of its proximity to the CI seam. Of particular interest are the studies on methylamine isotopologues that investigated the N−H(D) fragmentation, starting from well-characterized quantum states on the S1 potential energy surface (PES), corresponding to the first electronic state.26,27 In a pioneering study by Kim and coworkers,26 S1 vibronic states were excited and the H(D) photoproducts were detected by velocity map imaging (VMI). The excited states in CH3NH2 and CD3ND2 included the zeropoint level and the fundamental and first overtone of the amino wagging, ν9, and in the latter also the second overtone, 3ν9. This study revealed a bimodal appearance of high and low

lectronically nonadiabatic processes or non-Born−Oppenheimer (BO) processes take place in cases where the BO approximation, which assumes that the fast electronic motion is separable from the slow nuclear motion, breaks down.1−9 These processes occur following an electronic excitation via nuclear motions that induce coupling between two (or more) closelying adiabatic surfaces and cross at least a 2D coordinate to generate a conical intersection (CI). These CIs are points of degeneracy between electronic states that act as dynamic funnels for radiationless deactivation of the excited state, playing crucial roles in a variety of physical, chemical, and biological processes.10−12 They have implications on processes ranging from the simplest bimolecular reactions, that is, the hydrogen-exchange reaction,13 to the photostability of biomolecules,11 to the light-induced isomerization of retinal14−16 and were even predicted to promote nonradiative recombination in oxygen-containing defects on the surface of silicon nanocrystals.17 Therefore, processes that involve CIs became topics of intense interest for theoretical and experimental studies, seeking to understand and control them. Indeed, experiments using different schemes were performed in an attempt to control and understand the dynamic passage of the reactive flux through the CI in a variety of small-tomedium-sized gas-phase molecules. For instance, by using vibrationally mediated photodissociation (VMP) of ammonia18,19 and phenol,20 the nonadiabatic transition probability in N−H or O−H bond cleavage, respectively, could be manipulated. Essentially, for ammonia it was found that the major pathway for umbrella, bending, and symmetric stretching modes resulted in nonadiabatic dissociation to produce ground© 2016 American Chemical Society

Received: February 24, 2016 Accepted: April 21, 2016 Published: April 21, 2016 1717

DOI: 10.1021/acs.jpclett.6b00425 J. Phys. Chem. Lett. 2016, 7, 1717−1724

Letter

The Journal of Physical Chemistry Letters

Figure 1. (a) Measured two-color reduced-Doppler (TCRD) velocity map images of D photofragments ensuing from CH3ND2 predissociation and the respective (b) Abel inverted images and (c) retrieved total kinetic energy distributions (black circles) fitted by changing color (green, red, blue, and fading magenta) lines, representing the contributions of high and low kinetic energy photofragments (blue lines) as well as (d) D action spectrum with the assignment of the vibronic features, based on ref 28. The changing color and black arrows above the spectrum indicate the rovibronic transitions used for the measurement of TCRD velocity map images of the D photofragments. The arrows in the right corner of the TCRD images in panel indicate the vertical polarization of the photolyzing laser.

states with very narrow line widths,28 relative to those in CH3NH2 and CD3NH2. It was expected that by using the same approach as before we will be able to tune into resonance, if it exists, even more precisely. Although simple arguments led us to presume that the CH3ND2 isotopologue, which contains the ND2 moiety, will behave similarly to CD3ND2 that showed no mode dependence,26 we believed that if a resonance could be uncovered, then it will assist in shedding light on its origin. Indeed, by taking advantage of the possibility to prepare different initial rovibronic states on S1 and of the higher sensitivity of the TCRD imaging, we have been able to show the presence of a dynamic resonance. This resonance is strongly dependent on the excitation energy, the energy channeled into a different vibrationally excited mode of the dissociating parent, and the manipulated passage through the S1/S0 CI that leads to CH3ND (X̃ ) radical with high internal energy content. An overview of the experimental and analyzed results obtained from the predissociation of CH3ND2, initially preexcited to specific rovibronic states in the first electronically excited state, S1, is presented in Figure 1. These results show the measured (a) 2D TCRD velocity map images of the ensuing D photofragments; (b) the corresponding Abel inverted images,29 reflecting central cuts through the 3D KE distributions; (c) the TKEDs (black circles) related to the overall center of mass translational energy distributions of the photofragments,30 fitted by changing color (green, red, blue, and fading magenta) lines, resulting from the contribution of two individual asymmetric components (blue lines); and (d) the D action spectrum, which exhibits the total yield of the D photofragments, eliminated in CH3ND2 predissociation as a function of the excitation energy. The D action spectrum, Figure 1d, exhibits the rovibronic structure of CH3ND2, related to the population of different Franck−Condon (FC) active vibrational levels of the S1 state and to the internal rotation of the top (CH3) relative to the frame (NH2). In addition, it discloses the ensuing dynamics,

H(D) velocity components, leading to higher yield of slow H photoproducts from CH3NH2, initially excited with one quantum of amino wagging; however, no mode dependence was encountered in N−D bond cleavage of CD3ND2. The different behavior of the two molecules was attributed to the relatively long lifetime of the vibronic levels of CD3ND2, τ ≈ 2−10 ps, compared with τ ≈ 0.2 to 0.3 ps in CH3NH2, and to the more effective vibrational redistribution due to the higher density of S1 states, which lead to the loss of the initially prepared states in CD3ND2. Our very recent study on CH3NH2 and CD3NH2, containing the amino moiety,27 provided the opportunity to investigate the dominant N−H bond cleavage in both compounds. This study included preparation of different initial vibrational states in S1 and probing of the ensuing photofragments by frequencyresolved two-color reduced-Doppler (TCRD) ion imaging. The total kinetic energy distributions (TKEDs) for the fast and slow H products, resulting from N−H bond cleavage via two different dissociation pathways, showed anomalous behavior for some vibronic states. In particular, ν9 in CH3NH2 and ν9 and ν7 (CH3 rock) in CD3NH2, indicated the presence of dynamic resonances in the product branching ratio and in the anisotropy parameters, β. These resonances were associated with reduced release of H photofragments of high KE that exhibited higher negative β parameters. This enabled obtaining insight into the predissociation dynamics of these species and moreover proposing that the dynamics at the CI does not necessarily depend on the initially excited nuclear motions but is rather related to the energy difference between the initially prepared vibrational states and the CIs. Nevertheless, the puzzle of what exactly leads to the dynamical resonances and their fundamental aspects on the CI dynamics was not resolved. To overcome this difficulty, we have chosen to study here the predissociation dynamics of the CH3ND2 isotopologue. This is due to the slight blue shift in the band origins, encountered through H → D substitution, which allows access of somewhat different energies on the excited electronic state and rovibronic 1718

DOI: 10.1021/acs.jpclett.6b00425 J. Phys. Chem. Lett. 2016, 7, 1717−1724

Letter

The Journal of Physical Chemistry Letters

The measured 2D TCRD images, Figure 1a, and even more so the central slices through the 3D KEDs, Figure 1b, include inner features and sharper outer rings, associated with the two deconvoluted blue components of the respective TKEDs (Figure 1c). These observations confirm the release of D photofragments during N−D bond cleavage in CH3ND2 via two distinct channels, corresponding to D photofragments with high and low KEs. As suggested previously,26,27,35,36 the high KE channel is considered to be related to molecules that directly pass via the CI to the diabatic ground-state asymptote, leading to D + CH3ND(X̃ ) photofragments, while the low one is related to those that experience internal conversion (IC) through the CI by passing from S1 to the vibrationally “hot” S0 state and then decomposing to D + CH3ND(X̃ ) photofragments (see Figure 2), produced following energy redistribution. Most notable is the sudden increase in production of D photofragments with low KE relative to those of high KE, in the second column of Figure 1a−c, compared with those of the first column. Furthermore, as can be seen from the succeeding columns of Figure 1a−c, this production is then associated with a decrease, followed by an increase in the amount of low KE D photofragments, as higher energy rovibronic states are accessed. This is even more obvious from the TKEDs of Figure 3 that

following the excitation to the bound portion of the PES, which depends on tunneling and on the subsequent predissociation process. The rovibronic structure28 is associated with appearance of many bands, including the à ← X̃ (S1 ← S0) origin, that is, the 000 state, and the fundamentals, overtones, and combinations of v9 (NH2 wagging) and v7 (CH3 rock) and their rotational features. The dynamics is mostly reflected by the presence of very sharp rotational lines in the first two vibronic bands that become broader as higher vibronic states are accessed and also by the growing yield of the released D photofragments. These observations could be attributed to the tunneling of the D atom through a reaction barrier in the S1 state,31 Figure 2, which leads to relatively long lifetimes of the

Figure 2. Schematic drawing of the potential energy surfaces along the N−D dissociation coordinate, including the conical intersection between the ground (S0) and the excited (S1) electronic states and the geometry of the CH3ND2 molecule at this point.31,32,38 Figure 3. Overview of the fitted total kinetic energy distributions (TKEDs) of the D + CH3ND photofragments, resulting from predissociation of CH3ND2 molecules, promoted to the rovibronic states (marked by changing color arrows in Figure 1d on S1). The green, dashed red, blue, and fading magenta traces represent the rovibronic state of the origin band, 000, the lower (one of the rovibronic states involved in the resonance) and higher energy rovibronic states of ν9, and the yet higher energy rovibronic states. Additional TKEDs for rovibronic states of 000 and of ν9 are displayed in Figure S2 in theSI.

excited states of CH3ND2 and therefore to rotational fine structure28,32 and to higher probabilities for exiting from the well, leading to more effective D photofragment elimination, as higher vibronic states are accessed.33,34 In addition, the larger widths of the features, corresponding to the higher vibronic states, could possibly be attributed to the extra spectral density that occurs at higher excitation energies. As previously mentioned, the CH3ND2 molecule provides an opportunity to excite specific rovibronic lines, where each of them contains several transitions,28 (see Figure S1 and Table S1 in the Supporting Information (SI)). Indeed, TCRD images were measured for all preexcited rovibronic lines, marked by changing color (green and fading brown, red, blue and fading cyan, black and fading magenta) arrows above the action spectrum, Figure 1d. Nevertheless, only some of them, indicated by the changing color arrows (green, red, blue, and fading magenta in Figure 1d), are presented explicitly in the results shown in Figure 1a−c. These representative results include the columns corresponding to the TCRD images, Abel inverted images, and TKEDs, resulting from the predissociation of the rovibronic states of CH3ND2, roughly appearing above the positions of the changing color arrows (Figure 1d).

sum up the simulated traces of Figure 1c, where the changing color, green, red (dashed), blue, and fading magenta, traces represent the respective rovibronic states (marked in Figure 1d) of the origin band, 000, the lower (involved in the resonance, see below) and higher energy rovibronic states of ν9, and the rovibronic states of ν7 and its combinations with ν9, including a different number of quanta for each of the modes. Hence, it is evident that the quantity of low KE D atoms, relative to the high ones, changes as different states are accessed. In addition, this behavior is accompanied by unique changes in the translational energies of the high KE D photofragments, ensuing from predissociation of CH3ND2 molecules initially excited to different rovibronic states. It can be seen that following predissociation from the lower energy rovibronic 1719

DOI: 10.1021/acs.jpclett.6b00425 J. Phys. Chem. Lett. 2016, 7, 1717−1724

Letter

The Journal of Physical Chemistry Letters state of ν9 (dashed red) there is a shift of the distribution to lower translational energy compared with that stemming from the rovibronic state of the 000 (green). Subsequently, the translational energies of the high KE photofragments, resulting from the higher energy rovibronic state of ν9 (blue) and the additional rovibronic states (fading magenta), correspond to gradual movements toward higher values, although the latter exhibit similar most probable translational energies. On the basis of these findings, it was conjectured that the low/high KE D photofragments ratios and the maximal translational energies were altered by predissociation of CH3ND2 molecules, initially excited to different rovibronic states. It is evident that the change in the high translational energy component for different rovibronic states points to the formation of the CH3ND radical with varying internal excitation. Consequently, the dependence of the branching ratios and internal energies of the CH3ND radical on the energy of the initially excited rovibronic states was verified. The branching ratios were obtained from the ratio of the two components fitting the TKEDs for molecules, initially excited to the different rovibronic states (black and colored arrows in Figure 1d). The minimal internal energy, acquired by the CH3ND radical, was retrieved by taking into account the maximal translational energy for each distribution and by using energy conservation. It is considered that the internal energy of the CH3ND radical, Eint(CH3ND), is given by Eint(CH3ND) = Ehν + Eint(CH3ND2) − D0(N−D) − ET, where Ehν is the energy of the predissociating photon, Eint(CH3ND2) is the internal energy of the parent, and D0(N−D) is the bond dissociation energy and ET is the maximal total translational energy of the D + CH3ND photofragments. Because the contribution of the parent internal energy is ≤10 cm−1 (see Table S1 in the Supporting Information (SI)) it was neglected. The D0(N−D) was experimentally determined from the value of ET at the half height of the falling edge of the TKED corresponding to the first measured rovibronic state in the 000 band (marked by the darkest brown vertical line in Figure S2a of the SI). The retrieved value was found to be 35 150 ± 400 cm−1 and it agrees well with the previously reported value26 of 35 350 ± 400 cm−1 for CD3ND2. Choosing this point for estimation of D0(N−D) is reasonable, considering that the falling edge decreases quite sharply and there is some inaccuracy in determination of the tail of the TKED, depending on background subtraction in the TCRD images. Similarly, the maximal translational energies for the D photofragments resulting from the predissociation of CH3ND2, excited to other vibrational states, were also determined from the ET values at the half height of the falling edge of the corresponding TKEDs, where representative ones are shown in Figure 3 and additional ones in Figure S2a,b in the SI. Uncertainties of ±400 cm−1 were also attached to the ET values due to the arbitrariness of the position of the vertical dashed line, the allowance for the recoil momentum imparted to the D+ ion, and the phosphor screen resolution. This allowed us to retrieve the internal energies of the ensuing CH3ND radical, which are listed in Table S2 in the SI. The resulting branching ratios and internal energies of the CH3ND radical are shown in Figure 4a,b, respectively. While comparable values for both parameters were observed for predissociation of CH3ND2 from different rotational states of the 000 state, a sudden increase and then a fall occurred for the lower and higher energy rovibronic states of ν9 (see Table S1 in

Figure 4. Dependence of the (a) branching ratios between the slow and fast D photofragments and the (b) internal energies of the CH3ND photofragments obtained in predissociation of CH3ND2 molecules as a function of the energy of the initially excited rovibronic states (marked by arrows) in Figure 1d.

the SI), respectively, decreasing to minimal values for ν7 and 2ν9 and then gradually rising for the higher energy states. Actually, a similar behavior of the internal energy of the CH3ND radical was observed even when other points of crossing between the vertical line and the TKEDs, or the maximum of the faster of the two basis functions used in modeling the TKEDs, were used as a measure of ET. This implies that the observed trend does not depend on the chosen point of reference for N−D bond energy and ET determination and that the difference in the internal energy is