Nonadiabatic Photodissociation of the Hydroxymethyl Radical from the

May 29, 2015 - Department of Chemistry, Johns Hopkins University, Baltimore, Maryland 21218, United States. J. Phys ... The Hd is based exclusively on...
0 downloads 8 Views 5MB Size
Article pubs.acs.org/JPCA

Nonadiabatic Photodissociation of the Hydroxymethyl Radical from the 22A State. Surface Hopping Simulations Based on a Full NineDimensional Representation of the 1,2,32A Potential Energy Surfaces Coupled by Conical Intersections Christopher L. Malbon* and David R. Yarkony* Department of Chemistry, Johns Hopkins University, Baltimore, Maryland 21218, United States W Web-Enhanced Feature *

ABSTRACT: The nonadiabatic photodissociation CH2OH(12A) + hv → CH2OH(2,32A) → CH2O + H or HCOH(cis or trans) + H is addressed using trajectory surface hopping dynamics on a quasi-diabatic representation, Hd, of the 1,2,32A coupled, adiabatic potential energy surfaces. We focus on dynamics originating on the 22A potential energy surface. The Hd is based exclusively on electronic structure data obtained from a multireference configuration interaction single and double excitation expansion, composed of over 67 million configuration state functions, and treats all nine internal degrees of freedom in an even-handed manner. Each simulation is based on bundles of 10000 trajectories randomly selected from harmonic Wigner distributions and propagated for up to 1 ps. The bimodal distribution in the kinetic energy release spectrum is explained in terms of direct versus quasistatistical dissociation.

1. INTRODUCTION The hydroxymethyl radical and its photodissociation have been the subject of a considerable number of experimental1−6 and computational7−14 studies. When high overtone spectroscopy is used to study the ground electronic state of CH2OH, photodissociation to CH2O + H and isomerization to methoxy are observed.15,16 Bowman, Reisler, Krylov, and co-workers denoted BRK belowconstructed a full, nine-dimensional representation of the ground electronic state using CCSD(T) wave functions in a combined experimental−computational study of the aforementioned processes. That work provides high-quality benchmarks for the ground-state results in the present study. When excited beyond its electronic ground state to its lowlying 22A state (nominally a 3s Rydberg state) or the 32A state (nominally a 3px Rydberg state), hydroxymethyl decays into one of three products: formaldehyde, cis-hydroxymethylene, or trans-hydroxymethylene

The above photodissociation processes have been studied extensively by the Reisler group,3−6,15−18 including the recent time-sliced, velocity-mapped imaging experiments of RZR. The computational work in refs 10 and 12, addressed the hydroxymethyl radical’s photodissociation from a static, or electronic structure, perspective and provided considerable insights into the nonadiabatic mechanisms of photodissociation. However, there have as yet been no nuclear dynamics simulations of these reactions to confirm these inferences. Addressing this deficiency, we have constructed a full, ninedimensional, quasi-diabatic representation, Hd, of the 1,2,32A coupled adiabatic states that govern the photodissociation. In this work, surface hopping trajectory19 simulations originating on the 22A state and based on the Hd described in section 2 are used to address several issues raised by the experimental studies of RZR. Laser energies insufficient to access the hydroxymethylene channels yield bimodal kinetic energy release (KER) spectra. The origin of this bimodal distribution will be explained. The limited production of cis- and

CH 2OH + hν → CH 2OH(22A, 32 A)→ D0 = 10160 (7694) cm−1

H + CH 2O H + trans‐HCOH H + cis‐HCOH

(1a)

D0 = 28420 (27112) cm

−1

(1b)

D0 = 29970 (28734) cm

−1

(1c)

Special Issue: 100 Years of Combustion Kinetics at Argonne: A Festschrift for Lawrence B. Harding, Joe V. Michael, and Albert F. Wagner

Dissociation energies are taken from ref 17, which will be denoted RZR below. The present ab initio determined quantities are given in parentheses. © 2015 American Chemical Society

Received: January 24, 2015 Revised: April 5, 2015 Published: May 29, 2015 7498

DOI: 10.1021/acs.jpca.5b00758 J. Phys. Chem. A 2015, 119, 7498−7509

Article

The Journal of Physical Chemistry A

where Bu,v is a 3 × 3 symmetric matrix with a 1 in (u,v) and (v,u) elements and the remaining elements are 0. P̅κ is a standard group theoretical projection operator for the complete nuclear permutation inversion (CNPI) group.20 The Vl are the Nc linear coefficients of combination determined by the fitting procedure. The g(l)(R) are the monomial basis functions constructed from products of symmetry unique functions constructed from the scalar internuclear distances ri,j = ∥ri,j∥ and the internuclear vectors ri,j = Ri − Rj, where Rj locates the jth atom and 1 ≤ i < j ≤ Natom. Each g(l) is a product of the 21 functions in Table 2. The atomic numbering used in Table 2 and throughout this work is

trans-hydroxymethylene at energies below that of the 32A state, as found experimentally,17 will be confirmed and also explained. The KER spectra change markedly and abruptly as laser energies become sufficient to reach the 32A state.17 Hydroxymethylene channels are accessed via new pathways now energetically available. Barriers on the 32A potential energy surface and 32A−22A and 22A−12A conical intersection seams characterize these pathways. Radiationless decay through avoided intersections of the 32A and 22A states is also possible. Analysis of these alternatives necessitates the separation of dynamics originating on the 22A and 32A states. Here, we present computational results for the 22A contribution. Section 2 describes Hd and the electronic structure representation upon which it is based. This Hd represents the first full nine-dimensional representation of the 12A, 22A coupled adiabatic potential energy surfaces. Section 3 presents a surface hopping trajectory analysis of the questions raised above. Section 4 summarizes and concludes.

Table 2. Elementary Functions Comprising g(i)

2. THEORETICAL DESCRIPTION In this section, an overview of Hd and the electronic structure data from which it is constructed is reported. 2.A. Electronic Structure Treatment. The electronic structure treatment uses cc-p-VTZ atomic orbital bases on oxygen, carbon, and hydrogen augmented with s, p, and d Rydberg functions on oxygen [s(0.032), p(0.028), d(0.015)] and carbon [s(0.023), p(0.021), d(0.015)], where the Rydberg exponents are given parentheses. The multireference single and double excitation configuration interaction wave functions, comprised of 67 482 877 configuration state functions (CSFs), are built from single and double excitations out of a reference space that includes a two-orbital core, C(1s) and O(1s), a fourorbital doubly occupied space and a seven-orbital, five-electron active space. Four sets of doubly occupied orbitals were found to be relevant to the construction of the global Hd. The doubly occupied orbitals and the corresponding active space orbital are summarized in Table 1. The set of orbitals yielding the lowest

#

type

atomsc

function

note

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21

stretch stretch stretch stretch stretch stretch stretch OOP OOP OOP stretch stretch stretch stretch bend bend stretch stretch stretch stretch OOP

1,2 2,3 1,5 2,5 3,4 1,4 4,5 2,3,4,1 1,5,2,4 1,5,2,4 2,4 2,4 2,4 1,5 3,2,4 5,1,2 1,5 2,3 3,4 3,5 2,3,4,5

expb expb expb expb expb expb expb 4Rd 4Re 4Rf tanhh tanhh tanhh tanhh cos(∠324)g cos(∠512)g Yui Yui Yui Yui 4Rd

C−O stretch C−H stretch O−Ha stretch C−Ha stretch H−H stretch O−H stretch H−Ha stretch C−H−H−O umbrella C−Ha−H−O OOP C−Ha−H−O OOP C−H stretch partitioning C−H stretch partitioning C−H stretch partitioning O−H stretch partitioning H−C−H angle bend Ha−O−C angle bend O−Ha close range repulsive C−H close range repulsive H−H close range repulsive H−Ha close range repulsive O−H−H−H umbrella OOP

a

Permutationally nonequivalent hydrogen in CNPI symmetry. Exponential function defined in section 2.B. cFor bending-type coordinates, the vertex is the middle atom. For four-atom out-of-plane coordinates, where one atom is treated as permutationally unique, that atom is the first atom. In both cases, the permutationally unique atom is shown in bold. dThe 4 distance (4R) scaling of out-of-plane coordinates specifies one atom to be special among the four atoms. This atom will not be allowed to permute with the other three atoms. The three distances launching from this atom are used to perform scaling. eThe 6 distance (6R) scaling of out-of-plane coordinates is achieved by using all six pairs of internuclear distances to scale the scalar triple product coordinates. In this case, all four atoms are permutationally equivalent (up to a sign change of the coordinate). f The bond−bond dot product coordinate is defined as the dot product between two bond vectors. This involves four atoms and effectively describes both the angular motion and out-of-plane torsions. gThe cosine of the bond angle is also scaled by the inverse of the lengths of the two borders to ensure that the coordinate properly vanishes upon dissociation. hThe tanh functions partition the surface based upon H− X bond distances and have a maximum order of 4. iYukawa function defined in section 2.B. b

Table 1. Orbital Spaces for Three-State SA-MCSCF Wave Functionsa DOCC(4 orbitals)

CAS(7 orbital, 5 e−)

1

σCO,σn,πCO,σOH3

π*CO,σ3s,σ3px,,σCH1,σCH2,σ*CH2,σ*CH1

2

σCO,σn,πCO,σCH1

π*CO,σ3s,σ3px,,σOH3,σCH2,σ*CH2,σ*OH3

3

σCO,σn,πCO,σCH2

π*CO,σ3s,σ3px,,σOH3,σCH1,σ*CH1,σ*OH3

4

CO,πCO,σCH1,σCH2

π*CO,σ3s,σ3px,,σn,σOH3,σ*OH3,πCO*b

a

H1 and H2 are attached to the carbon, and H3 is attached to the oxygen. bDistorted.

three-state, state-averaged multiconfiguration self-consistent field (SA-MCSCF) energy compose the reference space at a given geometry. Discontinuities attributable to changes in orbital solutions, endemic to global electronic structure descriptions, are smoothed by the least-squares fit to the ab initio data points, introducing an ostensible fitting error but yielding smooth surfaces, qualitatively better than the ab initio data from which they are constructed. 2.B. Hd: Form and Construction. Hd has the form

given in Figure 1. The selection of the functions and their parameters, origins, and decay rates is of paramount importance. The use of overcomplete sets of functions allows us to keep the order of the polynomials low. Here, the total order of all functions in a g(l) is less than or equal to 4. The g(i)’s, described in detail in Table 2, include exponential functions, exp[−α(r

Nc d

H (R) =

∑ Vl [P ̅ κ[u(l), v(l)]g(l)(R)]Bu(l), v(l) l=1

(2) 7499

DOI: 10.1021/acs.jpca.5b00758 J. Phys. Chem. A 2015, 119, 7498−7509

Article

The Journal of Physical Chemistry A

subspace of nuclear coordinate space for which Hd reliably represents the ab initio data. A procedure for its construction based on classical surface hopping trajectories has been described in ref 22. For the energies employed in this work on the 1,22A states, 99% of trajectories sample geometries exclusively in the domain of definition of the 6256 points noted above. For higher-energy trajectories originating on the 32A potential energy surface, only 95% of trajectories remain within the domain of definition. Thus, the current Hd will be refined to treat 32A state dynamics. 2.C. Electronic Structure Data and Its Representation. We now turn to the quality of the current representation. We address two questions: how well does Hd reproduce the ab initio data, and how accurately does that data describe CH2OH and its photodissociation? −3 2.C.i. Ground-State Potential Energy Surface and Qmin CH2OH . min −1 min −3 Figure 1a and b, respectively, reports QCH and QCH , the 2OH 2OH 2 equilibrium structures of the ground and 3 A excited states, and compares them with literature values. As in ref 10, no local minimum on the 22A potential energy surface was found in this region. Table A.1, in Appendix A, reports and compares with literature values the corresponding harmonic frequencies. The Hd determined frequencies are in good accord with the ab initio values, which are in turn in satisfactory accord with the literature values. Using these frequencies we find T 0 (3 2 A) = 33753(33573)[35053].17 Here, and unless otherwise noted throughout this work, Hd determined values are in bold type face, ab initio determined values are in parentheses and literature values are in square brackets. The electronic energies are min −1 reported in cm−1 relative to E1(QCH ), the energy of the 2OH CH2OH ground-state equilibrium structure. T0 provides a valuable assay of the quality of the representation in the Franck−Condon region.

min−3 Figure 1. Qmin−1 CH2O and QCH2O , the equilibrium structures on the ground 2 1 A and the second excited 32A states. Hd determined values are above ab initio determined values, which are above literature values. Qmin−1 CH2O literature values are from BRK. Qmin−3 literature values are from ref 10. CH2O

−r0)], Yukawa functions, exp[−α(r − r0)]/r, hyperbolic tangents, as well as scaled bending functions. The following discussion serves to illustrate how the basic functions are combined to construct the g(i). Fine details concerning the constituent functions are given in Table 2. When constructing g(l), several restrictions are imposed: (i) Each g(l) is restricted to having no more than one tanh function. (ii) Function 21 describes the out-of-plane motion of four atoms, where three atoms may be treated as permutationally equivalent. Functions 11, 12, and 13 are hyperbolic tangent functions that provide the flexibility necessary to describe C−H dissociation. Function 21 is only necessary near CH3O, and functions 11, 12, and 13 are required only when C−H bonds have been significantly stretched. This allows us to exclude g(l)’s composed of function 21 and functions 11, 12, or 13. (iii) Function 12, a hyperbolic tangent function, has a point of inflection optimally located for the trans-HCOH dissociation channel. Function 13, another hyperbolic tangent, has its inflection point optimally placed for the cis-HCOH dissociation channel. It is not necessary to include g(l)’s composed of products of these two functions. (iv) Assuming, reasonably, that the potential energy surfaces are locally quadratic in the diabatic representation, we may restrict some of our stretching coordinates to second order; thus, the sum of the order of functions 1, 4, 5, 6, 7, 17, 18, 19, and 20 will not exceed 2. Accompanying Hd is an electronic Schrödinger equation [Hd(R) − IE(J m)(R)]d J(R) = 0

(3)

Ea,(m) J

from which the model energies (R), energy gradients, and derivative couplings are obtained. The Vl are determined from electronic structure data at only 6256 nuclear configurations Rn. This is possible because the use of energy gradients and derivative couplings in the defining equations means that each Rn contributes as many as 57 equations. For the current Hd, the 24184 Vl satisfy 244945 equations in a least-squares sense. The choice of points Rn at which the electronic structure data are determined is a key issue. With nine internal degrees of freedom, even a small grid of five points in each degree of freedom would require calculations at almost 2 million data points. Instead, surface hopping trajectories, obtained using the ANT program,21 are used to determine the domain of nuclear configurations for which Hd must accurately represent the electronic structure data if the nonadiabatic process at hand is to be well described.22 We define the domain of definition as the

sad−3 Figure 2. Qsad−1 CH2O+H (a) and QCH3O (b), the saddle point structures on 2 the ground 1 A potential energy surface. Hd determined values are above ab initio determined values, which are above literature values. Literature values are from BRK.

Figure 2a and b, respectively, reports the structures and imaginary frequency modes of two key saddle points, Qsad−1 CH2O+H sad−1 and QCH , leading to the H + CH O asymptote and to CH O 2 3O, 3 respectively. The Hd and ab initio determined structures are seen to be in good accord with each other and the literature values taken from BRK. Table A.2 in Appendix A reports the sad−1 corresponding frequencies and barrier heights. The QCH 2O+H 7500

DOI: 10.1021/acs.jpca.5b00758 J. Phys. Chem. A 2015, 119, 7498−7509

Article

The Journal of Physical Chemistry A imaginary frequency, 1693i(1758i)[1756i]16 cm−1 and the sad−1 QCH imaginary frequency 2165i(2061i)[1934i]16 cm−1 are 3O in satisfactory accord. The barrier height including the zero sad−1 is in good accord with that of BRK. The barrier point for QCH 3O −1 height for Qsad−1 below that reported by BRK, CH2O+H is ∼1500 cm reflecting the underestimation of D0(CH2O+H) discussed below. Barriers to the HCOH + H channels were not pursued as preliminary estimates put them at less than 200 cm−1 above their respective asymptotes. Figure 3 continues this largely ground-state comparison, reporting the asymptotic channel structures: formaldehyde, and

2.C.ii. Conical Intersections. We now turn to attributes related to electronic nonadibaticity. Figure 4 reports the

Figure 4. Characterization of CH2OH at minimum-energy conical 2 2 intersections. (a,b) The g and h at Qmex CH2O (1 A−2 A). (c,d) The g and h mex mex 2 2 at Qt‑HCOH (1 A−2 A). (e,f) The g and h at Qc‑HCOH (12A−22A). Hd determined distances and bond angles are above the ab initio determined quantities, which are above literature values. Literature values are from ref 12. The g and h vectors are normalized. Hd and ab initio determined vectors are indistinguishable on the scale of the figure, although their magnitudes may differ; see Figure 5a and b. mex mex geometries Q CH (1 2 A−2 2 A), Q c‑HCOH (1 2 A−2 2 A), and 2O mex (12A−22A), minimum-energy conical intersections Qt‑HCOH leading to the formaldehyde and the cis- and trans-hydroxymethylene channels, respectively. The energies of these conical intersections are 18146.4(18146.5)[19652.8], 32844.0(32872.0)[33268.5], and 30322.0(30596.2) cm−1 for mex 2 2 mex 2 2 QCH (12A−22A), Qmex c‑HCOH(1 A−2 A), and Qt‑HCOH(1 A−2 A), 2O respectively. Here, the literature values are taken from ref 12. 2 2 2 2 mex The energies for Qmex c‑HCOH(1 A−2 A) and QCH2O(1 A−2 A) are seen to be in good accord with the previous results obtained by mex (12A−22A) reflects one of us.12 The larger difference for QCH 2O the under estimation of D0(CH2O+H) noted above. The 2 2 energies for the conical intersections at Qmex c‑HCOH(1 A−2 A) and mex (12A−22A) are well above the energies of the Qt‑HCOH corresponding asymptotes for which De(cis-HCOH+H) = 31069 (31227) cm−1 and De(trans-HCOH+H) = 29385 (29516) cm−1, as reported in Table A.3. The orthogonal g (energy difference gradient) and h (interstate coupling) vectors23, which define the branching plane,24 are also reported. The Hd and ab initio determined structures are seen to be in good accord. Particularly gratifying is

Figure 3. Asymptotic structures of (a) CH2O, (b) trans-HCOH, and (c) cis-HCOH. Hd determined values are above ab initio determined values, which are above literature values. Literature values are from ref 14.

trans- and cis-hydroxymethylene. Again, the Hd, ab initio, and literature structures are in good accord. Table A.3 in Appendix A reports the harmonic frequencies of the asymptotic structures min −1 and the corresponding De relative to E1(QCH ). Reactions 2OH 1a−1c in section 1 report the corresponding D0. Agreement of Hd and ab initio determined frequencies and literature reported values is typical of the results reported in this section. However, D0(CH2O+H) is ∼2500 cm−1 below the experimental value. The impact of this deficiency is mitigated by the fact that each of D0(H+trans-HCOH), D0(H+cis-HCOH), and T0(32A) is approximately 1500 cm−1 below the corresponding experimental value. Also note that the HCOH + H asymptotes are energetically well separated from the CH2O + H asymptote, being approximately 18000−20000 cm−1 above the CH2O + H asymptote. 7501

DOI: 10.1021/acs.jpca.5b00758 J. Phys. Chem. A 2015, 119, 7498−7509

Article

The Journal of Physical Chemistry A

figure reports points on the seam of conical intersection in the vicinity of the energy-minimized point of conical intersection reported in Figure 4a and b. The points were obtained by constraining the bond distance on the abscissa to the indicated value while optimizing the remaining coordinates to minimize the energy of the conical intersection. This procedure was carried out for both the Hd and ab initio representations. Because these ab initio determined conical intersections are not included in the fit, the Hd determined points are predictions of the representation, which can be validated by the corresponding ab initio determined values. These plots confirm the energyminimum character of the point in Figure 4a. The predictions in 2 2 the vicinity of Qmex CH2O(1 A−2 A) are seen to be excellent. The mex predictions in the vicinity of Qt‑HCOH (12A−22A) are, as expected, less accurate but are still smooth and qualitatively correct. Note too from the ordinate scales in Figure 5b that the absolute errors in Figure 5b are not large. Similar results to those 2 2 2 2 mex for Qmex t‑HCOH(1 A−2 A) are found for Qc‑HCOH(1 A−2 A). The error reduction obtained by adjusting the relevant g(i) functions to improve the Hd determined value of R(C−H(i)), i = 3,4, at the minimum-energy conical intersections noted above is expected to similarly impact the predictions in Figure 5b. 2.C.iii. Diabaticity. Figure 6 addresses the diabatic character of Hd, reporting the Hd determined norm of derivative coupling

the observation that the Hd and ab initio determined g and h vectors are indistinguishable on the scale of the figures presented. Note, however, that the elongated bonds R(C− H(3)) and R(C−H(4)) in Figure 4b and c, respectively, are in error by ∼0.1 Å compared with the 0.005 Å error in the R(O− H) in Figure 4a. This is likely a consequence of the fact that these C−H bonds distances are more than 0.5 Å longer than R(O−H). The error can be reduced by adjusting the relevant g(i) functions in Table 2. This will be the subject of a future work. Figure 5a and b approaches the issue of the description of the conical intersection seam from an alternative perspective. Each

d

∥f1,2,H (R)∥ (solid line) and ab initio determined (open circles) norm of the derivative couplings ∥f1,2,ab(R)∥ together with the d

norm of the residual coupling, ∥f1,2,H (R) − f1,2,ab(R)∥ along a path from the 12A minimum to the CH2O + H dissociation limit

d

d

Figure 6. The upper panel reports ∥f1,2,H ∥, ∥f1,2,ab∥, and ∥f1,2,H , − f1,2,ab∥, the magnitudes of the 12A−22A derivative couplings, f12, along a 2 path from the Qmin CH2OH(1 A) minimum to the CH2O + H dissociation 2 2 limit passing through the ab initio determined Qmex CH2O(1 A−2 A). The lower panel reports E1 and E2 along this path. The 1,2 superscripts are suppressed in the figure.

Figure 5. Conical intersections predicted by Hd (open markers). Comparison with ab initio determined results (filled markers). (a) mex 2 2 2 2 Vicinity of Qmex CH2O(1 A−2 A); (b) vicinity of Qt‑HCOH(1 A−2 A). Note d that the point at R(C−H) ≈ 3.33 a0 in (b) is the H determined mex (12A−22A). none of the ab initio determined point in these Qt‑HCOH figures are included in the fit. 7502

DOI: 10.1021/acs.jpca.5b00758 J. Phys. Chem. A 2015, 119, 7498−7509

Article

The Journal of Physical Chemistry A

Figure 7. KER spectra originating from ν = 0. Each panel heading gives the hν laser energy in cm−1. KER spectra from QCTs (blue line) are compared with measured KER spectra from Figure 3 of RZR (black filled circles) with the same laser energy. Results are from RZR rescaled so that the integral is 1. The dashed (red) line expresses QCT results as a sum of contributions from two Gaussians, g1 and g2. mex passing through the ab initio determined QCH (12A−22A). 2O 2

2 2 and ab initio determined values for Qmex CH2O(1 A−2 A) reported in Figure 4a. The analysis of this section will inform the construction of the more complete Hd currently being developed.

2

Also reported are the energies of the 1 A and 2 A states. The small norm of the residual coupling indicates the quasi-diabatic character of the representation. (The residual coupling cannot vanish completely, true diabaticity, because Hd does not describe, and in any case could not eliminate, the nonremovable part25 of the derivative coupling that is part of the ab initio determined values). The large difference in the ab initio and Hd mex determined ∥f1,2∥ at QCH (12A−22A) does not indicate a 2O

3. NONADIABATIC DYNAMICS AND KER SPECTRAL SIMULATIONS The Hd studying states of obtained

significant issue. Rather, it reflects the small difference in the Hd 7503

described in section 2 provides a general tool for nonadiabatic processes involving the 12A and 22A CH2OH. Here, it is used to simulate KER spectra from the excitation of the ν = 0 ground state of DOI: 10.1021/acs.jpca.5b00758 J. Phys. Chem. A 2015, 119, 7498−7509

Article

The Journal of Physical Chemistry A

Figure 8. continued

7504

DOI: 10.1021/acs.jpca.5b00758 J. Phys. Chem. A 2015, 119, 7498−7509

Article

The Journal of Physical Chemistry A

Figure 8. continued

7505

DOI: 10.1021/acs.jpca.5b00758 J. Phys. Chem. A 2015, 119, 7498−7509

Article

The Journal of Physical Chemistry A

Figure 8. Trajectories: (a,b) H (fast) + CH2O, (c,d) H(slow) + CH2O; (e,f) hν = 33898 cm−1, H(slow) + trans-HCOH. hν and KER are in cm−1. Subpanels numbered from the upper right clockwise show the below-described attributes at the time marked by the solid dot in the electronic energy plot: (i) electronic energies of states 1−3; (ii) R(O−H) in au for (a−d) and R(C−H) in au for (e,f); (iii) additional internal coordinates: for (a−d), angles ∠H(3)CH(4) and ∠H(3)CO and distances R(C−H(4)) and R(C−H(3)); for (e,f) ∠COH(5), ∠H(4)CO, R(O−H(5)), and R(C−H(4)); (iv) the molecule. The vertical line in subpanels (ii) and (iii) indicates a surface hop. Enhancements (movies) are available in the online version for panels (a), (b), (c), (d), (e), and (f).

CH2OH to the 22A state, reported by RZR. These simulations, based on Tully’s fewest switches surface hopping dynamics,19 will emphasize the mechanism of the underlying nonadiabatic processes.

3.A. Initial Conditions and Statistics. In these simulations, each experimental condition corresponds to a fixed laser energy, hν. For each set of randomized initial conditions used to describe an experimental condition, the nuclear coordinates and momenta are chosen from an harmonic Wigner distribution26 7506

DOI: 10.1021/acs.jpca.5b00758 J. Phys. Chem. A 2015, 119, 7498−7509

Article

The Journal of Physical Chemistry A for the 12A vibrational levels and then transferred to the 22A electronic state. A particular set of coordinates and momenta are accepted provided that the total energy ET on the 22A potential energy surface is within a small window, less than ±100 cm−1, of hν + EZPE, where EZPE is the vibrational zero-point energy of CH2OH in its 12A ground state. For each set of experimental conditions, 10000 trajectories are run to ensure good statistics. Energy conservation is achieved to within 1 × 10−5eV. 3.B. KER Spectra from the 22A State. Figure 7a−c reports KER spectra with three distinct laser energies (hν = 28169, 29850, and 33898 cm−1) chosen to match laser energies reported in Figure 3 of RZR. In each figure, the simulated spectrum is reported as a solid blue line. The corresponding KER spectra from RZR are included as black dots. The experimental data have been uniformly scaled so that their integral is 1, as is the case for the simulated spectra. The breadth of the spectra are well reproduced. However, the vibrational structure evident in the experimental data for high KER is, not unexpectedly, not captured by the quasi-classical trajectory (QCT) dynamics. Comparing Figure 7a−c, it is seen that increases in laser energy have a similar, broadening, impact on both the measured and simulated KER spectra. The experimental spectra clearly exhibit bimodal character. The simulation is also bimodal. This is shown by expressing the simulated spectrum as the sum (dashed red lines) of two Gaussian distributions g1 and g2 (solid red lines). The origin of the bimodal distribution is an essential issue in understanding the KER spectra. One might suspect that the bimodal character is related to the possibility of distinct outcomes, the production of HCOH and CH2O. However, the present theory and previous experiments (see RZR) agree that from the 22A state, the HCOH channels contribute minimally,