Probing Highly Efficient Photoisomerization of a Bridged Azobenzene

ARTICLE pubs.acs.org/JPCA

Probing Highly Efficient Photoisomerization of a Bridged Azobenzene by a Combination of CASPT2//CASSCF Calculation with Semiclassical Dynamics Simulation Lihong Liu, Shuai Yuan, and Wei-Hai Fang* College of Chemistry, Beijing Normal University, Beijing 100875, China

Yong Zhang* Department of Chemistry, Chemical Biology, Biomedical Engineering, Stevens Institute of Technology, Castle Point on the Hudson, Hoboken, New Jersey 07030, United States

bS Supporting Information ABSTRACT: Mechanism of phototriggered isomerization of azobenzene and its derivatives is of broad interest. In this paper, the S0 and S1 potential energy surfaces of the ethylene-bridged azobenzene (1) that was recently reported to have highly efficient photoisomerization were determined by ab initio electronic structure calculations at different levels and further investigated by a semiclassical dynamics simulation. Unlike azobenzene, the cis isomer of 1 was found to be more stable than the trans isomer, consistent with the experimental observation. The thermal isomerization between cis and trans isomers proceeds via an inversion mechanism with a high barrier. Interestingly, only one minimum-energy conical intersection was determined between the S0 and S1 states (CI) for both cis f trans and trans f cis photoisomerization processes and confirmed to act as the S1 f S0 decay funnel. The S1 state lifetime is ∼30 fs for the trans isomer, while that for the cis isomer is much longer, due to a redistribution of the initial excitation energies. The S1 relaxation dynamics investigated here provides a good account for the higher efficiency observed experimentally for the trans f cis photoisomerization than the reverse process. Once the system decays to the S0 state via CI, formation of the trans product occurs as the downhill motion on the S0 surface, while formation of the cis isomer needs to overcome small barriers on the pathways of the azo-moiety isomerization and rotation of the phenyl ring. These features support the larger experimental quantum yield for the cis f trans photoisomerization than the trans f cis process.

’ INTRODUCTION Photoinduced cis
trans isomerization of the NdN double bond is a ubiquitous photochemical process and represents one of the simplest pathways for converting light energy into mechanical motion at a molecular level, which forms a fundamental step in many photochemical and photobiological systems.1
3 The photoisomerization has many potential applications in the context of optical memories, opto-electronic switching, molecular motors, and light-driven molecular shuttle.4
8 As a representative, azobenzene (AB) and its derivatives have fascinated chemists for many decades due to their remarkable structural changes between their elongated (E) and more compact (Z) forms, the reversibility of their transformations, and the high photostability guaranteeing large numbers of switching cycles.4,5 For these reasons, they have been extensively investigated to understand the isomerization processes and to explore their applications in various fields.9
14 The S1(1nπ*) electronic absorption bands lie in the same wavelength range of 430
450 nm for both E and Z isomers of r 2011 American Chemical Society

azobenzene, making it difficult to selectively excite the E or Z isomer to the S1 state. To switch AB back and forth, one of the isomers requires to be excited to the S2(1ππ*) state by the UV light, which may result in photodamage of other moieties attached to the AB unit. In addition, the isomerization yield is smaller for excitation to the 1ππ* state than to the 1nπ* state. Recently, an ethylene-bridged azobenzene (1) was reported to have a greatly enhanced conversion efficiency and quantum yield.15 Experimentally it has been established that the photoinduced isomerization between cis (1Z) and trans (1E) proceeds with high efficiency and a quantum yield was measured to be 72% for 1Z f 1E by the blue light at 370
400 nm and 50% for 1E f 1Z by the green light at 480
550 nm.15 Hence, the bridged azobenzene is of high potential as a molecular photoswitch. To understand the photochromic properties of the bridged Received: April 20, 2011 Revised: July 25, 2011 Published: July 25, 2011 10027

dx.doi.org/10.1021/jp203704x | J. Phys. Chem. A 2011, 115, 10027–10034

The Journal of Physical Chemistry A azobenzene, Bockmann and co-workers investigated the nonadiabatic dynamics processes of the S1 state on the basis of ab initio MD simulations employing the PBE functional and plane waves with dual-space pseudopotential.16 Dynamics simulations demonstrated that the photoinduced 1E f 1Z isomerization is an ultrafast process with a time constant of about 40 fs. A total of 30 trajectories were calculated and roughly half of them result in successful 1E f 1Z isomerization,16 which is consistent with a quantum yield of ∼50% measured experimentally. Following earlier observations for hindered-azobenzene systems, the ultrafast photoisomerization was ascribed to the favorable preorientation of the phenyl rings and to a pedal motion of the NdN moiety. The photochemical behavior of the ethylene-bridged azobenzene was also studied with classical-mechanical on-thefly dynamics, employing a floating-occupation configurationinteraction approach at the semiempirical AM1 level (FOCIAM1),17 which provides a different insight into the 1E f 1Z photoisomerization mechanism. The calculated trajectories reveal that the 1E f 1Z dynamics proceeds through a conical intersection surrounded by a rather flat potential energy landscape and then encounters a sizable barrier in the electronic ground state that markedly reduced the quantum yield of the 1E f 1Z isomerization process.17 It is well-known that features of the potential energy surface, such as barrier heights and stationary and conical intersections, can help with understanding a chemical reaction. However, as far as we know, there is no report of ab initio electronic structure calculations of the S0 and S1(1nπ*) potential energy surfaces for the ethylene-bridged azobenzene, although a few dynamics simulations have been conducted for exploring its photochemical behavior. In the present study, multiconfiguration second-order perturbation theory (CASPT2) has been used to calculate the S0 and S1 potential energy surfaces, which was followed by a semiclassical dynamics simulation.18
20 The relaxation dynamics from the S1 Franck
Condon (FC) structure of 1Z or 1E to CI was found to control the efficiency of 1Z f 1E or 1E f 1Z isomerization processes, while their quantum yields are determined to a large extent by the shape of the S0 potential energy surface.

’ COMPUTATIONAL METHODS Structures of 1Z, 1E, and the transition state (TS) of the 1Z
1E isomerization in the S0 state are optimized by using the complete active space self-consistent field method (CASSCF) with an active space of 10 electrons in eight orbitals, referred to as CAS(10,8) hereafter. The optimized structures are confirmed to be the minimum-energy points or first-order saddle point by harmonic frequency calculations. The conical intersection between the S0 and S1 states is determined by the state-averaged CASSCF optimization. The π, π*, and two nonbonding orbitals in the NdN group were included in the active space for the CASSCF calculations, which provides a reasonable account for the 1Z
1E isomerization in the S0 and S1 states. The rest of the orbitals and electrons in the active space come from π and π* orbitals of the aromatic rings. For comparison, the density functional theory at the B3LYP level is used to investigate the 1Z
1E isomerization in the S0 state. Several basis sets were used for the B3LYP calculations to estimate the effect of the basis set size. All CASSCF calculations have been carried out with the 6-31G* basis set by using the Gaussian 03 program.21 To account for dynamical electron correlation, the single-point energy is

ARTICLE

recalculated at the CASPT2 level using a three-state-averaged CAS(14,12) zeroth-order wave functions (referred to as CASPT2(14,12) hereafter), which was performed by using the Molcas program.22 Because it is practically impossible to calculate the potential energy surface as a function of all internal degrees of freedom for the system studied here, the S0 and S1 potential energy surfaces were determined by the CASPT2(14,12)/6-31G* calculations on the selected grid, which is based on the B3LYP/6-31G* and CAS(6,4)/6-31G* optimized S0 and S1 structures, respectively. A semiclassical dynamics method has been employed to explore the photoisomerization mechanism between 1Z and 1E. This method has been described in detail in the previous studies18
20 and only a brief description is given here. The state of the valence electrons is calculated by the time-dependent Schr€odinger equation in a nonorthogonal basis, ip

∂Ψj ¼ S
1 3 H 3 Ψj ∂t

where S is the overlap matrix of the atomic orbitals. Electronic structure calculations were performed at the DFTB level20 that has essentially the same strengths and limitations as the timedependent density functional theory. The bonding is well described by the DFTB method, but the excited-state energies are typically too low. For this reason, we matched the effective central photon energy of the laser pulse to the relevant densityfunctional excitation energy. The photon energy selected in the present work is based on the energy gap between the highest occupied molecular orbital (HOMO) and the lowest unoccupied molecular orbital (LUMO), which is 2.72 eV for 1Z and 2.20 eV for 1E, respectively. The laser pulse is characterized by a vector potential, which is coupled to the electronic Hamiltonian through the timedependent Peierls substitution   iq 0 AðtÞ 3 ðX
X0 Þ ðX
X 0 Þ exp Hab ðX
X 0 Þ ¼ Hab pc where Hab(X
X0 ) is the Hamiltonian matrix element for basis functions a and b on atoms at X and X0 , respectively. A laser pulse is applied to the present molecule with full-width at halfmaximum duration (fwhm) of 85 fs. The nuclear motion numerically integrated with the velocity Verlet algorithm is solved by the Ehrenfest equation of motion   d2 XlR 1 ∂H ∂S ∂ Ψþ
ip Ψj Ml 2 ¼
2 j j 3 ∂XlR ∂XlR 3 ∂t 3 dt

∑

þ hc


∂Urep ∂XlR

which is obtained by neglecting the terms of second and higher order in the quantum fluctuations in the exact Ehrenfest theorem. This basic method has been successfully used in previous studies of the photoisomerization of butadiene, stilbene, and azobenzene, as well as other photoinduced processes.20,23
34

’ RESULTS AND DISCUSSION Stationary Structures and Their Relative Energies. The equilibrium structures of 1Z and 1E in the ground state (S0) were optimized at the B3LYP and CAS(10,8) levels with several different basis sets. The optimized structures are depicted in Figure 1 along with the key bond parameters from the CAS(10,8)/6-31G* 10028

dx.doi.org/10.1021/jp203704x |J. Phys. Chem. A 2011, 115, 10027–10034

The Journal of Physical Chemistry A

Figure 1. Schematic Structures of 1Z, 1E, TS, and CI, along with the selected atom numbering and the CAS(10,8)/6-31G* bond parameters.

calculation. It was found that the B3LYP/6-31G* optimized 1Z and 1E structures are nearly the same as those from B3LYP optimizations with the larger 6-311++G** and cc-pVTZ basis sets. So, the size of basis set has little influence on the optimized structures of the 1Z and 1E isomers. Thus, the 6-31G* basis set is used for the CASSCF and CASPT2 calculations. The C
N
N
C dihedral angle and N
N
C2 angle are predicted to be 6.0° and 121.1° in the 1Z isomer and 147.5° and 111.1° in the 1E isomer respectively, at the CAS(10,8)/6-31G* level, which are in good agreement with the available data from the single crystal X-ray diffraction structure.15 The 1Z and 1E isomers have nonplanar structures in the ground state. Especially, the 1E isomer is significantly deviated from the planar structure due to the severe constraints by the ethylene bridge. The 1Z isomer is more stable than 1E by 6.7
7.4 kcal/mol at the B3LYP level with a number of basis sets, which again indicates that the size of basis set has no significant effect on the calculated relative energies. This energy difference was predicted to be 7.6 kcal/mol at the CASPT2(14,12)/6-31G* level. This observation is consistent with the experimental fact that 1Z is thermodynamically more stable.15 In addition to the 1Z and 1E isomers, a local minimum was predicted to exist in the ground state by the FOCI-AM1 electronic structure calculation and the FOCI-AM1-based dynamics simulation,17 which is more stable than 1E by about 5.0 kcal/mol. With the FOCI-AM1 optimized structure for this local minimum as the initial guess, all attempts to search for a minimum-energy structure in the ground state lead to formation of the 1Z isomer at the B3LYP, CAS(10,8), and MP2 levels, which was further supported by the CASPT2(14,12)/6-31G* calculated potential energy surface for the ground state. Therefore, except for 1Z and 1E, no other S0 minimum was found in the present study. Previous investigations show that in the ground state, the E-form of azobenzene is more stable than the Z-form by ∼12.0 kcal/mol,35 which is different from the case of the ethylene-bridged azobenzene studied here. The E-form has a planar structure for azobenzene and the conjugated interaction between the aromatic rings and the NdN π bond makes the Eform of azobenzene more stable, in contrast with the nonplanar 1E structure of the ethylene-bridged azobenzene. Such a change has a considerable effect on the spectroscopic and photoswitch character of the ethylene-bridged azobenzene.

ARTICLE

A transition state (TS) was determined to connect 1Z and 1E in the S0 state by the B3LYP and CAS(10,8) calculations with different basis sets. As discussed above, the size of basis set has little influence on the optimized TS structure and its relative energy. However, the TS position in the S0 surface exhibits a dependence on the computational method. For example, the C
N
N
C dihedral angle and N
N
C2 angle in the TS structure are 108.9 and 168.0° at the B3LYP/6-31G* level and become 94.1 and 163.9° at the CAS(10,8)/6-31G* level, respectively. The CAS(10,8)/6-31G* optimized TS structure for the isomerization in the ground state is validated by the CASPT2(14,12)/6-31G* calculated potential energy surface, discussed below. A comparison of the 1Z, 1E, and TS structures (Figure 1) clearly shows that the NdN bond length is basically unchanged in the 1E f 1Z or 1Z f 1E processes, retaining a double bond character. The most striking change is associated with the N
N
C2 angle that is increased from 121.1° in 1Z to 163.9° in the TS structure. As seen from the TS geometry, the 1Z
1E isomerization in the ground state proceeds via an inversion mechanism. The calculated barrier heights for the 1Z f 1E and 1E f 1Z processes are 36.7 and 29.1 kcal/mol, respectively, at the CASPT2(14,12)/6-31G* level. The laser excitation at 370
400 nm moves the 1Z isomer into the S1 electronic state, which mainly involves one-electron transition from HOMO to LUMO orbital. HOMO is the nonbonding orbital localized on the N atoms and LUMO is the NdN π* antibonding orbital. The S1 state exhibits a clear 1 nπ* character. No minimum-energy structures were found for 1E or 1Z in the S1 state by the CAS(10,8)/6-31G* optimization. The S0 f S1 vertical excitation energies for 1Z and 1E were calculated to be 64.9 and 54.0 kcal/mol, respectively, at the CASPT2(14,12)/6-31G* level of theory, which are in reasonable agreement with the S0 f S1 electronic absorption bands at 404 nm (70.7 kcal/mol) for the 1Z isomer and at 490 nm (58.3 kcal/mol) for the 1E isomer. Potential Energy Surfaces of the S0 and S1 States. The S0 potential energy surface was scanned at the B3LYP/6-31G* level on a regular grid over the C
N
N
C dihedral angle and the N
N
C2 angle with all other bond parameters optimized fully. The C
N
N
C dihedral angle is varied from 10.0 to 150.0° with an interval of 20.0°, while the grid on the N
N
C angle is in the range from 120 to 160° with a step-size of 10.0°. In addition, 15 structures between the regular grid points were calculated for some regions where energies are changed sharply with the C
N
N
C dihedral angle and the N
N
C2 angle. On the basis of the B3LYP/6-31G* optimized structures on the grid, the S0 potential energy surface was determined by the CASPT2(14,12)/6-31G* single-point calculation. The S0 stationary structures (1Z, 1E, and TS) optimized at the CAS(10,8)/ 6-31G* level were well reproduced by the CASPT2(14,12)/ 6-31G* calculations. Except for the 1Z and 1E isomers, no other minimum-energy structures were found on the CASPT2(14,12)/6-31G* calculated potential energy surface for the ethylene-bridged azobenzene in the ground state. Similarly, the S1 potential energy surface has been determined at the CASPT2(14,12)/6-31G* level of theory on the basis of CAS(6,4)/6-31G* optimized structures for the S1 state. The regular grid for the S1 state is the same as that for the S0 state, with additional 12 structures in between the regular grid points. Figure 2 shows contour plots of the S0 and S1 potential energy surfaces as a function of the C
N
N
C and N
N
C2 angles. A minimum-energy conical intersection (CI) between the S0 and 10029

dx.doi.org/10.1021/jp203704x |J. Phys. Chem. A 2011, 115, 10027–10034

The Journal of Physical Chemistry A

ARTICLE

Figure 2. Contour plots of the optimal S0 and S1 potential energy surfaces as a function of the C
N
N
C and N
N
C2 angles, along with the relaxation pathways from CI to 1E and 1Z on the S0 surface, as well as relaxation pathways from Franck
Condon structures of 1E and 1Z to CI on the S1 surface.

S1 states was determined from the S0 and S1 potential energy surfaces, which was confirmed by the state-averaged CAS(10,8)/ 6-31G* full optimization. The C
N
N
C dihedral angle and the N
N
C2 angle are 99.1° and 135.2° respectively in the CAS(10,8)/6-31G* optimized CI structure, matching closely to those determined from the CASPT2(14,12)/6-31G* calculated S0 and S1 potential energy surfaces. With respect to the S0 minimum of 1Z, the CI structure has its relative energy of 43.8 and 41.3 kcal/mol at the CAS(10,8)/6-31G* and CASPT2(14,12)/6-31G* levels, respectively. The relative energies are collected in Table 1 for the critical structures in the S0 and S1 states. After the submission of the present work, the photoisomerization between 1Z and 1E was studied with the combined femtosecond time-resolved spectroscopy and quantum chemical calculations.36 The optimized structures and the calculated relative energies are well consistent with those reported in the present work, which will be discussed in detail below. The FOCI-AM1-based dynamics simulation17 indicates that the two photoisomerization reactions, 1Z f 1E and 1E f 1Z, proceed via two different conical intersections of CI
1E and CI
1Z respectively. The C
N
N
C dihedral angle and the N
N
C2 angle are 95.5 and 132.1° respectively in the CI
1E structure, which is close to those in the CI structure reported in the present work. Based on the CI
1Z nuclear configuration reported in the previous study,17 we attempted to optimize a conical intersection between the S0 and S1 surfaces. But all optimizations lead to the same CI structure, indicating that CI
1Z is not the real minimum-energy conical intersection. In fact, the energy difference between the S0 and S1 states at the CI
1Z structure was calculated to be 9.4 kcal/mol at the CASPT2(14,12) level of theory in the previous study,17 which again suggests that this is not a valid conical intersection. Only one conical intersection was found by the combined fs-spectroscopy and CASPT2//CASSCF study.36 Photoinduced 1E f 1Z Isomerization Dynamics. The photoinduced 1E f 1Z isomerization dynamics in the gas phase has been investigated with different nonadiabatic dynamics methods.16,17,34 For comparison, the 1E f 1Z dynamics process was explored here with a semiclassical dynamics method. After 1500 fs molecular dynamics (MD) simulation at room temperature, starting from the 1E equilibrium geometry in the ground

Table 1. Relative Energies (kcal/mol) for the Critical Structures in the S0 and S1 States, Which Are Calculated at Different Levels of Theory with the 6-31G* Basis Set relative energies structures

B3LYP

CAS(10,8)

CASPT2(14,12)

1E(S0)

6.7a

7.9a

7.6a

1Z(S0)

a

a

a

TS(S0)

0.0

0.0 a

34.2 /27.5

b

10.2a 0.0a

0.0 a

40.4 /32.5 a

b

DFTB

a

b

36.7 /29.1

39.2a/29.0b

41.3

45.0c

FC-S1(1E)

55.5b

54.0b

50.7b

FC-S1(1Z)

67.3a

64.9a

62.7a

CI(S0/S1)

a

43.8

a

Energies relative to1Z(S0). b Energies relative to1E(S0). trajectory-averaged energy relative to 1Z(S0).

c

The

state, 15 snapshots were taken with a time-interval of 100 fs as the initial structures for the subsequent dynamics simulations. A laser pulse with full-width at half-maximum duration of 85 fs is applied to the initial structures and effective photon energy centered at 2.2 eV, on the basis of the HOMO
LUMO gap of 1E. Trajectories were propagated 1000 fs and around half of them were observed to produce the 1Z isomer in the ground state, which is consistent with the 50% yield measured experi mentally.15 Electronic populations and energies of HOMO
1(π), HOMO(n), and LUMO(π*) orbitals are plotted in Figure 3 as a function of time for one representative trajectory, where variations of the selected bond parameters with time are also displayed. It is interesting to note that the laser pulse brings one electron to the LUMO by ∼170 fs and leaves a hole at HOMO, but the laser pulse has little influence on the electronic population of the HOMO
1 orbital. This shows that the 1nπ* state is populated initially by photoexcitation. As can be seen from the electronic population in Figure 3a, the trajectory propagates in the 1nπ* state in the first 30 fs after laser irradiation and then electronic transition takes place from LUMO to HOMO in a very short period. Accordingly, the energy gap between HOMO and LUMO decreases abruptly to zero and the S1 and S0 states become degenerate in energy, which can be seen from Figure 3b. 10030

dx.doi.org/10.1021/jp203704x |J. Phys. Chem. A 2011, 115, 10027–10034

The Journal of Physical Chemistry A

Figure 3. Electronic populations (a), orbital energies (b), and the selected bond parameters (c
g) are plotted as a function of time for the photoinduced 1E f 1Z isomerization process for one representative trajectory.

Upon inspecting nuclear structures where electronic transition takes place for all reactive trajectories, it was found that the nonadiabatic decay from S1 to S0 occurs in the CI vicinity. Therefore, both the electronic structure calculation and dynamics simulation reveal that the CI acts as a nonadiabatic S1 f S0 decay funnel for the photoinduced 1E f 1Z isomerization process. As shown in Figures 3c
g, the C
N
N
C dihedral angle decreases remarkably after laser irradiation of the 1E isomer, while the N
N
C2 angle shows a significant increase and the N
N distance is noticeably elongated due to the S0 f S1 excitation. These results clearly show that the initial photoexcitation energies are mainly distributed in the C
N
N
C torsional, the N
N
C2 bending, and the N
N stretching modes. Since there are large energy gradients along the C
N
N-C torsional, the N
N
C2 bending, and the N
N stretching directions, the relaxation from the S1 FC structure of the 1E isomer to the CI intersection is predicted to be an ultrafast process. The average time needed for the relaxation process is 35 fs. The excited state lifetime is inferred to be less than 50 fs by nonlinear least-squares fits to the observed absorption-time

ARTICLE

profiles.36 The azo-moiety isomerization thus takes place rapidly, and the C
N
N
C dihedral angle decreases to zero once decaying to the ground state. However, it takes ∼500 fs for the N
N
C2
C4 and C3
C1
N
N dihedral angles to reach their equilibrium values in the ground-state 1Z isomer. Structural change of the ethylene-bridged benzene framework is slow in comparison with the azo-moiety isomerization. The complete 1E f 1Z isomerization has a time scale of ∼550 fs. The present and previous16,17,34 dynamics simulations show agreement on the time scale for the S1 relaxation process from 1E to CI. However, different S0 dynamics processes from CI to 1Z were reported in the previous studies16,17,34 after the system passes through CI to the ground state. The nonadiabatic AIMD simulation16 revealed that the 1E f 1Z photoisomerization is an ultrafast process. Trajectories are propagated into the 1Z potential well within 50 fs and lead to successful 1E f 1Z photoisomerization. However, the trajectories were found to be trapped in a local minimum with the C
N
N
C dihedral angle of ∼6.0° by Hartke and coworkers,17 which is separated from the 1Z potential well by a transition state with a barrier of ∼37 kcal/mol. Density functional-based molecular dynamics simulations have been performed by Allen and co-workers,34 which shows that the 1E f 1Z photoisomerization involves two processes: (1) Upon photoexcitation to the S1 state, each molecule moves on its S1 potential energy surface via rotation around the N
N bond to CI, where de-excitation occurs. (2) Subsequently, additional rotation around the N
N bond occurs in the electronic ground state, and the 1Z geometry is achieved at about 450 fs, due to twisting of the phenyl rings around their C
N bonds. As discussed above, the additional local minimum determined at the FOCI-AM1 level17 was not confirmed in the ground state by the present B3LYP, CASPT2(14,12), and MP2 calculations. The transition state at the FOCI-AM1 level is characterized with one imaginary frequency of 35.9 cm
1 in the previous study.17 A similar transition state was obtained by the B3LYP/6-31G* and MP2/6-31G* calculations, but it was confirmed to connect the two equivalent 1Z minima in the ground state and the barrier height is only 13.5 kcal/mol at the MP2/6-31G* level. The 1E f 1Z photoisomerization pathway proposed by Hartke and coworkers17 is thus not supported by the present electronic structure calculations. The C
N
N
C dihedral angle is 99.1° in the CI structure, which is closer to the 1E isomer in view of the rotation around the N
N bond than the 1Z isomer. Trajectories are propagated to the 1E side of the S0 surface via the CI and formation of the 1Z isomer is blocked slightly by the rotation around the N
N bond, as seen from Figure 2. In addition, the C3
C1
N
N and N
N
C2
C4 dihedral angles in the CI structure are different from those in the 1Z isomer and the possibility for formation of the 1Z isomer is further reduced by twisting of the phenyl ring around the C
N bond. These features of the S0 potential energy surface determined at the CASPT2(14,12)/6-31G* level are consistent with the time constant (∼500 fs) estimated by the present and previous34 dynamics simulations for the complete 1E f 1Z photoisomerization, rather than a short time constant of ∼50 fs from the nonadiabatic AIMD simulations.16 The 1E f 1Z photoisomerization was inferred to take place on a time scale of 370 fs (50 fs for the relaxation from the S1 FC geometry to the CI structure and 320 fs for the process from CI to 1Z in the S0 state) in a recent study.36 10031

dx.doi.org/10.1021/jp203704x |J. Phys. Chem. A 2011, 115, 10027–10034

The Journal of Physical Chemistry A

Figure 4. Electronic populations (a), orbital energies (b), and the selected bond parameters (c
g) are plotted as a function of time for the photoinduced 1Z f 1E isomerization process for one representative trajectory.

Photoinduced 1Z f 1E Isomerization Dynamics. To start dynamics simulation with different initial structures for the 1Z isomer, a number of snapshots were taken from 2000 fs MD simulation at room temperature. Then a laser pulse is applied to the initial eleven structures with a fwhm of 85 fs and effective photon energy centered at ∼2.7 eV. It was observed that 11 trajectories pass through CI to the S0 state, but eight of them are reactive and lead to formation of the 1E isomer in the ground state, while the left three trajectories return to the 1Z isomer in the ground state. This is in good agreement with the experimentally observed 72% yield of 1Z f 1E.15 Time evolution of electronic populations and energies for HOMO
1, HOMO, and LUMO orbitals are plotted in Figure 4 for one of the representative reactive trajectories, together with variations of the selected bond parameters over time. The S1(1nπ*) state of the 1Z isomer is confirmed to be populated initially by the electronic populations of the frontier orbitals shown in Figure 4a. As seen from Figure 4b, the increase in the HOMO orbital energy coincides with the decrease in LUMO orbital energy soon after the laser pulse is applied. Gradually, the energy gap between HOMO and LUMO goes

ARTICLE

to zero at a time of 1833 fs, which shows that the S1 and S0 states become degenerate in energy. Upon inspection of structures for a few snapshots at a time of ∼1833 fs, it was found that they are close to the CI structure. Similar situation was found for other trajectories. Hence, the combined electronic structure calculation and dynamics simulation performed here support that the CI acts as a nonadiabatic decay funnel for both the photoinduced 1Z f 1E and 1E f 1Z isomerization processes. The NdN π bond is weakened by one-electron excitation from n to π* orbital, which results in an elongation of the N
N distance. In addition, the C
N
N
C dihedral angle increases to ∼60° once 1Z is excited to the S1 state due to a rehybridization of the N atom from sp2 to sp3. These are consistent with the initial stage of time evolution for the N
N distance and the C
N
N
C dihedral angle shown in Figures 4c,d. Before nonadiabatic decay to the ground state, the trajectory propagates in the 1nπ* state for about 1660 fs (1833
170 fs). In this period, the N
N bond length decreases slightly, but the N
C2 single bond is shortened remarkably with a clear double-bond character, which blocks the rotation of the phenyl ring around the N
C2 bond. The change of bond parameters with time shows that the initial energies distributed in the N
N stretching mode are partially transferred to the N
C2 stretching mode. As a result, it takes long time to relax from the S1 FC structure of the 1Z isomer to the CI region, which is 1648 fs on average for the 11 trajectories. In comparison with the lifetime of 70 fs inferred experimentally,36 the present dynamics simulations overestimate the excited-state time constant significantly. The main reason is that the present Ehrenfest method has its weakness that causes averaging over the S1 and S0 states, rather than following the time evolution of the S1 potential energy surface. Once the trajectory is propagated into the CI region, one electron transition takes place from LUMO to HOMO in a short period, leading to the electronic ground state. Then, formation of the 1E isomer occurs as the downhill motion on the S0 surface, as shown in Figure 2. The 1Z f 1E photoisomerization is completed within 1700 fs. These results show that the photoinduced 1E f 1Z isomerization proceeds in a period much shorter than the photoinduced 1Z f 1E isomerization, which is consistent with the higher efficiency for the 1E f 1Z process.15 Upon photoexcitation of the 1E isomer at ∼550 nm, the system at the S1 FC structure has a tiny amount of initial kinetic energy. The direction of the nonzero energy gradient at the S1 FC structure results in the population of the selected vibrational modes and the shape of the S1 surface controls the redistribution of the initially localized vibrational energy among other modes. The CASPT2(14,12)/6-31G* calculated S1 potential energy surface shows that the initial S1 relaxation is dominated by the C
N
N
C torsional and N
N
C2 bending vibrations upon irradiation of the 1E isomer and is the downhill motion on the S1 surface to the CI region. This is consistent with a small time scale of 30
40 fs for the relaxation process from the S1 FC structure of the 1E isomer to the CI region. However, the relaxation process is slow from the S1 FC structure of the 1Z isomer to the CI region. The CAS(10,8)/6-31G* calculations reveal that the nonzero gradients at the S1 FC structure are in the direction of the N
N stretching and the C
N
N
C torsional motions for the 1Z isomer, which result in the initial excitation energies distributed mainly on the N
N stretching and the C
N
N
C torsional modes. Then, the initial energies distributed in the N
N stretching mode are partially transferred to the N
C2 stretching mode, leading to a clear double-bond character for the 10032

dx.doi.org/10.1021/jp203704x |J. Phys. Chem. A 2011, 115, 10027–10034

The Journal of Physical Chemistry A N
C2 bond, which blocks a fast increase of the C
N
N
C dihedral angle to reach the CI region. Therefore, the S1 lifetime is much longer for the 1Z isomer than that for the 1E isomer. Although the S1 lifetime is much shorter for the 1E isomer than that for the 1Z isomer, a 50% quantum yield was observed for the photoinduced 1E f 1Z isomerization at 480
550 nm, which is lower than a quantum yield of 72% for the photoinduced 1Z f 1E isomerization at 370
400 nm.15 It should be noted that the CI acts as a nonadiabatic decay funnel for both the photoinduced 1Z f 1E and 1E f 1Z isomerization processes. Therefore, it is considered as the initial structure for the subsequent processes in the ground state. In view of the azomoiety isomerization, the CI is in a favorable position for formation of the 1E product in the S0 state, as illustrated by the C
N
N
C dihedral angles which are 6.0, 99.1, and 147.5° in the 1Z, CI, and 1E structures, respectively. Additionally, structural change of the ethylene-bridged benzene framework is almost free in the process from CI to 1E, because the 1E formation occurs as the downhill motion on the S0 surface. However, the situation is considerably different for the 1Z formation, where the system needs to overcome small barriers on the pathways of the azo-moiety isomerization and rotation of the phenyl ring around the C
N bond. Consequently, formation of the 1E isomer is always favored over the 1Z formation, regardless of the photoinduced isomerization starting from 1E or 1Z isomer. This indicates that the shape of the S0 potential energy surface in the CI region provides a qualitative account for the smaller quantum yield of the 1E f 1Z process than for the 1Z f 1E process.

’ SUMMARY In the present work, the S0 and S1 potential energy surfaces for the ethylene-bridged azobenzene have been determined by the CASPT2(14,12)/6-31G* calculations on the grid of the C
N
N
C and N
N
C2 internal coordinates. The stationary structures in the S0 state have been characterized at the B3LYP, CAS(10,8), and CASPT2(14,12) levels, which all exhibit little dependence on the size of the basis set. The 1Z isomer was found to be more stable than the 1E isomer in the ground state by 7.6 kcal/mol, consistent with the fact that 1Z was experimentally found to be thermodynamically more favorable. The calculated barrier heights for the 1Z f 1E and 1E f 1Z processes in the ground state are 36.7 and 29.1 kcal/mol, respectively, at the CASPT2(14,12)/6-31G* level. The thermal 1Z f 1E isomerization is associated with a large increase of the N
N
C2 angle from 121.1° in 1Z to 163.9° in the TS and proceeds via an inversion mechanism. No minimum-energy structure was found in the S1 state of the ethylene-bridged azobenzene. But one minimum-energy conical intersection between S0 and S1 states (CI) was determined by the CAS(10,8)/6-31G* optimization and further confirmed by the CASPT2(14,12)/6-31G* calculated S0 and S1 potential energy surfaces. Meanwhile, photoinduced isomerization between 1Z and 1E was explored with a semiclassical dynamics method. In general, both the 1Z f 1E and 1E f 1Z photoisomerization reactions involve three processes: the relaxation from the S1 state of the 1Z or 1E isomer to the CI region, nonadiabatic S1 f S0 decay via the CI, and the relaxation from the CI to the 1Z or 1E product in the ground state. The present electronic structure calculations and nonadiabatic dynamics simulations reveal that the initial relaxation from the S1 state of the 1E isomer is the

ARTICLE

downhill motion to the CI region with a lifetime of ∼30 fs, which is consistent with the ultrafast S1 deactivation reported for the 1E isomer in the previous studies. Upon photoexcitation of the 1Z isomer to the S1 state, the initial energies are partially transferred to the N
C2 stretching mode and the N
C2 bond becomes of more double-bond character, which blocks the structural change to the CI region. Redistribution of the internal energies controls the relaxation dynamics from the S1 FC structure to the CI region for the 1Z isomer. Therefore, the S1 lifetime is much longer for the 1Z isomer than that for the 1E isomer. The S1 relaxation dynamics is responsible for higher efficiency observed for the 1Ef1Z photoisomerization than for the 1Z f 1E process. The CI was confirmed to act as a nonadiabatic S1 f S0 decay funnel for both 1Z f 1E and 1E f 1Z photoisomerization processes. The decay in the CI vicinity takes place in a short period and nonadiabatic effect has little influence on the efficiency of the isomerization processes. Once the system decays to the S0 state via the CI intersection, formation of the 1E product occurs as the downhill motion on the S0 surface, while the 1Z formation needs to overcome small barriers on the pathways of the azo-moiety isomerization and rotation of the phenyl ring around the C
N bond. As a result, formation of the 1E isomer is favored over the 1Z formation, which is consistent with the larger quantum yield measured for the 1Z f 1E photoisomerization than that for the 1E f 1Z process.

’ ASSOCIATED CONTENT

bS

Supporting Information. Internal coordinates and absolute energies of the stationary and intersection structures. This material is available free of charge via the Internet at http://pubs. acs.org.

’ AUTHOR INFORMATION Corresponding Author

*E-mail: [email protected]; [email protected].

’ ACKNOWLEDGMENT This work was supported by grants from the NSFC (Grant Nos. 20720102038 and 21033002) and from the Major State Basic Research Development Programs (Grant No. 2011CB808503). ’ REFERENCES (1) Feringa, B. L. Molecular Switches; Wiley-VCH: Weinheim, 2001. (2) Balzani, V.; Credi, A.; Venturi, M. Molecular Ddevices and Machines; Wiley-VCH: Weinheim, 2008. (3) Ulysse, L.; Cubillos, J.; Chmielewski, J. J. Am. Chem. Soc. 1995, 117, 8466–8467. (4) Wang, X. G.; Kumar, J.; Tripathy, S. K.; Li, L.; Chen, J. I.; Marturunkakul, S. Macromolecules 1997, 30, 219–225. (5) Wang, X. G.; Chen, J. I.; Marturunkakul, S.; Li, L.; Kumar, J.; Tripathy, S. K. Chem. Mater. 1997, 9, 45–50. (6) Ikeda, T.; Tsutsumi, O. Science 1995, 268, 1873–1875. (7) Yu, Y. L.; Nakano, M.; Ikeda, T. Nature 2003, 425, 145–149. (8) Hugel, T.; Holland, N. B.; Cattani, A.; Moroder, L.; Seitz, M.; Gaub, H. E. Science 2002, 296, 1103–1106. (9) Ootani, Y.; Satoh, K.; Nakayama, A.; Noro, T.; Taketsugu1, T. J. Chem. Phys. 2009, 131, 194306. (10) Tiago, M. L.; Ismail-Beigi, S.; Louie, S. G. J. Chem. Phys. 2005, 122, 094311. 10033

dx.doi.org/10.1021/jp203704x |J. Phys. Chem. A 2011, 115, 10027–10034

The Journal of Physical Chemistry A

ARTICLE

(11) Ishikawa, T.; Noro, T.; Shoda, T. J. Chem. Phys. 2001, 115, 7503–7512. (12) Bandara, H. M. D.; Friss, T. R.; Enriquez, M. M.; Isley, W.; Incarvito, C.; Frank, H. A.; Gascon, J.; Burdette, S. C. J. Org. Chem. 2010, 75, 4817–4827. (13) Basheer, M. C.; Oka, Y.; Mathews, M.; Tamaoki, N. Chem.— Eur. J. 2010, 16, 3489–3496. (14) Haberhauer, G.; Kallweit, C. Angew. Chem., Int. Ed. 2010, 49, 2418–2421. (15) Siewertsen, R.; Neumann, H.; Buchheim-Stehn, B.; Herges, R.; N€ather, C.; Renth, F.; Temps, F. J. Am. Chem. Soc. 2009, 131, 15594– 15595. (16) B€ockmann, M.; Doltsinis, N. L.; Marx, D. Angew. Chem., Int. Ed. 2010, 49, 1–4. (17) Carstensen, O.; Sielk, J.; Sch€onborn, J. B.; Granucci, G.; Hartke, B. J. Chem. Phys. 2010, 133, 124305. (18) Dou, Y.; Torralva, B. R.; Allen, R. E. J. Mod. Opt. 2003, 50, 2615–2643. (19) Dou, Y.; Torralva, B. R.; Allen, R. E. Chem. Phys. Lett. 2004, 392, 352–357. (20) Yuan, S.; Dou, Y.; Wu, W.; Hu, Y.; Zhao, J. J. Phys. Chem. A 2008, 112, 13326–133334. (21) Frisch, M. J.; et al. Gaussian 03, Revision D.02; Gaussian, Inc.: Wallingford, CT, 2004. (22) Andersson, K; et al. Molcas, Version 7.1; University of Lund: Sweden, 2008. (23) Torralva, B. R.; Niehaus, T. A.; Elstner, M.; Suhai, S.; Frauenheim, Th.; Allen, R. E. Phys. Rev. B 2001, 64, 153105. (24) Dou, Y.; Torralva, B. R.; Allen, R. E. J. Phys. Chem. A 2003, 107, 8817. (25) Dou, Y.; Allen, R. E. J. Chem. Phys. 2003, 119, 10658. (26) Dou, Y.; Allen, R. E. J. Mod. Opt. 2004, 51, 2485. (27) Sauer, P.; Allen, R. E. Chem. Phys. Lett. 2007, 434, 260. (28) Sauer, P.; Allen, R. E. Chem. Phys. Lett. 2008, 450, 192. (29) Sauer, P.; Allen, R. E. J. Phys. Chem. A 2008, 112, 11142. (30) Dou, Y.; Lei, Y.; Li, A.; Wen, Z.; Torralva, B. R.; Lo, G. V.; Allen, R. E. J. Phys. Chem. A 2007, 111, 1133–1137. (31) Dou, Y.; Hu, Y.; Yuan, S.; Wu, W.; Tang, H. Mol. Phys. 2009, 107, 181–190. (32) Jiang, C.; Xie, R.; Li, F.; Allen, R. E. Chem. Phys. Lett. 2009, 474, 263. (33) Jiang, C.; Xie, R.; Li, F.; Allen, R. E. Chem. Phys. Lett. 2010, 487, 177. (34) Jiang, C.-W.; Xie, R.-H.; Li, F.-L.; Allen, R. E. J. Phys. Chem. A 2011, 115, 244. (35) Cembran, A.; Bernardi, F.; Garavelli, M.; Gagliardi, L.; Orlandi, G. J. Am. Chem. Soc. 2004, 126, 3234–3243. (36) Siewertsen, R.; Schonborn, J. B.; Hartke, B.; Renth, F.; Temps, F. Phys. Chem. Chem. Phys. 2011, 13, 1054–1063.

10034

dx.doi.org/10.1021/jp203704x |J. Phys. Chem. A 2011, 115, 10027–10034