Experimental observation of nonadiabatic bifurcation dynamics at resonances in the continuum

In the strong vibronic state mixing regime, both Herzberg type-I and type-II predissociations coexist and proceed in a competitive way.


UV-UV depletion spectroscopy of 2-fluorothioanisole
UV-UV depletion spectroscopy was performed to check that vibronic transitions start from the same zero-point level of the ground state of the one isomer. The burning laser was scanned while the wavelength of the probe laser was fixed at the band origin with a time delay of 100 ns. cm -1 and a = 0.1519 Å -2 with a barrier of 40 cm -1 were obtained from the fitting of origin (0 cm -1 ), τ 1 (32 cm -1 ), and τ 2 (82 cm -1 ) bands in R2PI spectrum. (k = 0.0172 mdyne/Å) The corresponding ratio of the Franck-Condon factors for these bands is 1:1.7:0.3. The red solid and dashed lines are symmetrically allowed and forbidden transition levels, respectively. The eigenvalues and eigenfunctions for the potential was obtained using Wavepacket 4.7.3 1 . The time-independent Schrődinger equation was solved by the Fourier grid Hamiltonian method. The parameters of the reduced mass and the force constants were taken from the constrained optimization (see Figure S3). Figure S3. (a) R2PI spectrum and (b) Franck-Condon simulation of 2-fluorothioanisole. The optimized geometry of S 1 state in the inset was obtained by freezing six geometry parameters using TD-B3LYP 2 /6-311++G(d,p). The S-CH 3 dihedral angle is fixed at 12.5°, and all the atoms except for the hydrogens and methyl group are in-plane. The Franck-Condon simulation was performed with this S 1 geometry using FCLabII. 3,4 The two imaginary frequencies (-87 and -146 cm -1 ) were converted to the positive ones for the simulation. b Both 658 and 666 cm -1 S 1 bands have ν s mode characters in the SEVI spectra, and corresponding assignments are tentatively given by considering the spectral pattern of R2PI spectrum. c 824 cm -1 S 1 band has a mode character of 12 and corresponding assignments are tentatively given by considering the spectral pattern of R2PI spectrum.  a The torsion vibration and stretching vibration are denoted by τ and ν, respectively. The bending mode in which the vibrating atoms preserve a well-defined plane is designated by β. The perpendicular bending vibration with respect to such a plane is denoted by γ. The symmetric or asymmetric character of the vibration is indicated by a subscript. Of the bending modes of a CH n group, β s designates the scissoring vibration and β as , γ s , and γ a denote the rocking, wagging, and twisting modes, respectively. Normal modes are labeled according to Ref. [5]. b The experimental values are observed from R2PI spectrum. c The experimental values are observed from SEVI spectra.  shown here are the weighted averaged and the resultant values are plotted in Figure 5(b).

Fitting procedure
The total translational energy distribution was first subtracted by the background for multiphoton dissociative ionization. This background at the lowest translational energy was fitted as a Boltzmann-like function: (Equation S1) where A 1 is the amplitude and C 1 is the width. Above the 1345 cm -1 S 1 internal energy, a Gaussian function was further included for the background at the center of the image: where A 2 , B 2 , and C 2 are the amplitude, center, and width, respectively. The type of the function was determined by a series of imaging experiments at the same excitation wavelength.
It was assumed that the total translational energy distribution at the S 1 origin has only the low translational energy channel (I) so it was fitted as two Gaussian functions (the same form as Equation S2) corresponding to the and Ã states of 2-fluorothiophenoxyl radical. The -Ã ̃ẽ nergy gap was fixed at 8.006 kcal/mol (2800 cm -1 ) in the whole fitting procedure in order to reduce the number of variable. The distribution at the origin was fitted as a Gaussian function centered at 17.2 kcal/mol, covering the majority of the distribution under 24.368 kcal/mol (which is the maximum translational energy of the Ã state). The left (small) distribution was fitted and assigned as the state. The rest of the distribution at the 32 and 82 cm -1 S 1 internal energy after t he subtraction of fits at origin (both the and Ã states) was assigned to the high translational ẽ nergy channel (II) which also has both the and Ã states. The position of state of the low ̃̃1 energy channel overlaps with that of the high energy channel, causing to increase the uncertainty in the /Ã ratio of the low energy channel. We assumed that the /Ã ratio increases with the ̃ĩ ncrease of the excitation energy and the ratio for the low energy channel was restricted to 0.14 referred to the value for thioanisole. The center and the width of all the Gaussian functions increase as the excitation energy increases.
The error range of the fraction (Г) of the high translational energy channel in the total distribution was obtained by the repetition of the imaging experiment and a fitting procedure by adjusting the relative amount of the Ã states of the low and high translational energy channel.
The final error bar was determined as the largest one.