A‑Band Absorption Spectrum of the ClSO Radical: Electronic Structure of the Sulfinyl Group

‑

Band

Absorption

Spectrum

of the ClSO Radical:

Electronic

Structure

of the Sulfinyl

Group

Published

as part of The Journal

of Physical

Chemistry

virtual

special

issue “Marsha

I. Lester

Festschrift”.

Wen

Chao,

Gregory

H. Jones,

Mitchio

Okumura,

Carl J. Percival,

and Frank

A. F. Winiberg

Cite This:

J. Phys.

Chem.

2023,

127, 8374−8382

Read Online

ACCESS

Metrics

& More

Article

Recommendations

sı

Supporting

Information

ABSTRACT:

Sulfur

oxide species

(RSO

) play a critical

role in many

fields,

ranging

from biology

to atmospheric

chemistry.

Chlorine-containing

sulfur

oxides

may play a key role in sulfate

aerosol

formation

in Venus’

cloud layer by

catalyzing

the oxidation

of SO to SO

via sulfinyl

radicals

(RSO).

We present

results

from the gas-phase

−

vis

transient

absorption

spectroscopy

study of the

simplest

sulfinyl

radical,

ClSO,

generated

from the pulsed-laser

photolysis

thionyl

chloride

at 248 nm (at 40 Torr of N

and 292 K). A weak absorption

spectrum

from 350 to 480 nm with a peak at 385 nm was observed,

with partially

resolved

vibronic

bands

(spacing

= 226 cm

−

), and a peak cross section

(385

nm) = (7.6

1.9)

−

. From

ab initio

calculations

at the EOMEE-

CCSD/ano-pVQZ

level, we assigned

this band to 1

′

←

′′

and 2

′

←

′′

transitions.

The spectrum

was modeled

as a sum of a bound-to-free

transition

to the 1

′

state and a bound-to-bound

transition

to the 2

′

state

with similar

oscillator

strengths;

the prediction

agreed

well with the observed

spectrum.

We attributed

the vibronic

structure

to a

progression

in the bending

vibration

of the 2

′

state. Further

calculations

at the XDW-CASPT2

level predicted

a conical

intersection

between

the excited

′

and 2

′

potential

energy

surfaces

near the Franck

−

Condon

region.

The geometry

of the

minimum-energy

conical

intersection

was similar

to that of the ground-state

geometry.

The lack of structure

at shorter

wavelengths

could be evidence

of a short excited-state

lifetime

arising

from strong

vibronic

coupling.

From simplified

molecular

orbital

analysis,

we attributed

the ClSO spectrum

to transitions

involving

the out-of-plane

orbitals

along the S

−

O bond and the in-plane

orbital

possessing

character

along the S

−

Cl bond. We hypothesize

that these orbitals

are common

to other sulfinyl

radicals,

RSO,

which

would

share a combination

of a strong

and a weak transition

in the UV (near 300 nm) and visible

(400

−

600

nm) regions.

■

INTRODUCTION

The formation

of sulfur

oxide

species

(RSO

), generated

through

reactions

between

oxygen

and sulfinyl

radicals

(RSO),

is important

in many

fields,

including

biology,

1,2

organic

synthesis,

3,4

and atmospheric

chemistry.

For instance,

important

source

of sulfate

aerosols

in the Earth’s

atmosphere

is believed

to be regulated

by RSO

oxidation,

with HSO

and

7,8

as intermediates

in the formation

of sulfates.

However,

in the middle

atmosphere

of Venus,

condensed

sulfuric

acid

clouds

have been observed

despite

extremely

low oxygen

levels.

Sulfate

aerosol

formation

is attributed

to the catalytic

role of chlorine,

involving

in the oxidation

of CO and SO to

form CO

and SO

To comprehend

the mechanism

chlorine

catalysis

of sulfinyl

radical

oxidation,

direct

kinetic

measurements

of sulfur species

and chlorine

in the laboratory

must be conducted

along

with theoretical

calculations

provide

chemical

insights.

The sulfinyl

chloride

radical,

ClSO,

is the smallest

sulfinyl

radical

containing

a chlorine

atom and represents

an ideal

system

for high-level

ab initio

calculations.

Although

the

ClSO radical

can be easily generated

through

the photolysis

thionyl

chloride

(Cl

SO),

13,14

available

spectroscopic

measure-

ments

are limited

to ground-state

properties

such as infrared

(IR),

far-IR,

and electron

paramagnetic

resonance

spectra.

Furthermore,

relevant

theoretical

studies

are scarce.

In the absence

of suitable

detection

methods,

kinetic

investigations

were limited

to mass spectrometric

detection.

Recently,

we observed

a strong

ultraviolet

absorption

spectrum

of the ClSO

radical

in the gas phase

near 300 nm,

which

we assigned

to the 1

′′

←

′′

transition

of the ClSO

radical.

We used the strong

UV absorption

to monitor

ClSO

and study the kinetics

of the ClSO + Cl

→

SO reaction

pressures

ranging

from 10 to 90 Torr at 292 K.

Received:

July 24, 2023

Revised:

September

14, 2023

Published:

September

29,

2023

Article

pubs.acs.org/JPCA

The Authors.

Published

American

Chemical

Society

8374

https://doi.org/10.1021/acs.jpca.3c04977

J. Phys.

Chem.

2023,

127, 8374

−

8382

This article is licensed under CC-BY 4.0

studies

in cryogenic

matrices

−

have shown

that

a few sulfinyl

radicals

exhibit

both strong

UV absorption

near

300 nm and weak

features

near 400 nm, indicating

that

analogous

weak features

might also exist for ClSO.

The goal of

this work was to search

for a longer-wavelength

spectrum

gas-phase

ClSO

by transient

absorption

spectroscopy.

employed

the same

apparatus

used to detect

the shorter

wavelength

spectrum,

a White-cell-based

transient

−

vis

absorption

spectrometer

coupled

into a flow reactor.

Radicals

were

generated

by pulsed-laser

photolysis,

and transient

absorption

spectra

were recorded

in the 300

−

580

nm window.

Low-lying

excited

electronic

states

were computed

using

Coupled

Cluster

and Complete

Active

Space

Perturbation

Theory

(CASPT2)

methods.

The observed

spectra

were then

assigned

by comparison

with the results

from the

ab initio

calculations.

Additionally,

a molecular

orbital

(MO)-based

analysis

was performed

to provide

chemical

insight

into the

role of the Cl atom and the sulfinyl

electronic

structure

in the

ClSO radical.

■

EXPERIMENTAL

METHODS

The experimental

instrument

and the theoretical

methods

have

been previously

reported.

In short, a small stream

of nitrogen

(evaporated

from liquid

nitrogen)

flowed

through

a bubbler

containing

SO (Sigma-Aldrich

>99%)

held in a temper-

ature-controlled

bath at 292 K (Fisherbrand,

Isotemp

4100).

This mixture

was then introduced

into a flow reactor

whose

pressure

was monitored

by a capacitance

gauge (MKS

127AA-

00100A)

and controlled

by a throttle

valve (MKS

type 153).

The gas refreshing

rate (MKS

GM50A

and 1179A)

was slightly

faster than the repetition

rate of the photolysis

laser (Coherent

COMPex

205F,

KrF).

A broadband

Xe plasma

light source

(LDLS,

Energetiq

EQ-99)

was directed

into a white cell (

140 cm, Acton

Optics,

10 passes,

eff

≈

450 cm). The light was

then projected

into a two-exit

grating

spectrograph

(Acton

SpectraPro

300i),

which

allowed

both an image-intensified

CCD (Princeton

Instruments

PI-MAX4,

1024

256) and a

photomultiplier

tube (Hamamatsu

R928)

to collect

trans-

mitted

light simultaneously.

A long-pass

filter (Semrock

LP02

−

257RU-30

40) was used to separate

the photolysis

beam

from the probed

beam

to reduce

the background

absorbance.

For measurements

longer

than 520 nm, a long-

pass filter (Thorlabs

FEL0450)

was placed

in front of the

entrance

slit of the spectrometer

to remove

second-order

diffraction.

The concentrations

of Cl

SO were in the range of

(0.4

−

and were balanced

with N

for a total

pressure

of 40 Torr at a temperature

of 292 K.

■

THEORETICAL

METHODS

All equation

of motion

coupled

cluster

(EOM-CCSD)

calculations

were performed

using

the CFOUR

program

suite,

with the exception

of the unrestricted

reference

LR-

CCSDT,

which

was performed

using the MRCC

package.

All calculations

made use of the ano-pVQZ

basis set,

which

has coverage

from H

−

Ar,

as the use of atomic

natural

orbital

(ANO)

basis sets have been found

to be exceptionally

effective

for calculating

harmonic

frequencies.

The frozen-core

approximation

was applied

to all calculations

since the ano-

pVXZ

basis sets were not optimized

for core correlation.

The

ground

state was optimized

using CCSD

with an unrestricted

Hartree

−

Fock

(UHF)

reference,

while

the 1

′

and 2

′

excited

states

were optimized

using

EOMEE-CCSD.

explore

the possible

effects

of spin contamination,

we also

employed

restricted

Hartree

−

Fock

(RHF)

references

optimize

the 1

′

and 2

′

excited

states using the EOMEA

(ClSO

cation

reference)

and EOMIP

(ClSO

−

anion

reference)

approaches,

respectively.

The ground

state was

also optimized

by using these two methods

for calculations

the adiabatic

transition

energies.

Although

the combination

of EOM-CCSD

and EOM-

CCSDT

provides

a highly

accurate

treatment

of dynamic

correlation

in the excited

state, it fails to describe

the behavior

of the adiabatic

states in a very small region

about the conical

intersection,

yielding

complex-valued

energies

due to the non-

Hermitian

character

of the effective

Hamiltonian.

characterize

the splitting

of the adiabatic

states in the vicinity

of the conical

intersection,

we turned

to multireference

perturbation

theory,

in particular

XDW-CASPT2.

While

CASPT2

is ultimately

a perturbation

theory

and describes

dynamic

correlation

less accurately

than coupled

cluster-based

methods,

its extended

multistate

variants

(

. XDW-CASPT2

and XMS-CASPT2)

correctly

describe

the topology

of the

conical

intersections.

XDW-CASPT2

is especially

suited

for

this purpose,

as it interpolates

between

state-specific

and state-

averaged

Fock operators

as the adiabatic

states approach

each

other and mix.

The minimum-energy

conical

intersection

(MECI)

between

the 1

′

and 2

′

states

was located

via XDW-CASPT2

calculations

based on a CASSCF(7,5)

reference

averaged

over

the three lowest

electronic

states

as implemented

in Open-

Molcas.

−

The conical

intersection

was located

using the

projected

constrained

optimization

approach

35,36

in which

the

average

energy

of the two states is minimized

subject

to the

constraints

that the adiabatic

energy

difference

is zero and a

dummy

constraint

that prevents

motion

along the analytically

computed

derivative

coupling

vector.

An imaginary

shift of

0.05 eV was applied

for all calculations

to avoid intruder

states.

The weighting

factor

used for the dynamically

weighted

Fock

matrix

was as described

in ref 39, corresponding

to the

OpenMolcas

keywords

of DWType

= 3 and DWMS

= 1.0. The

conical

intersection

was characterized

in the branching

plane

by displaced

geometries

in the plane defined

by the

and

vectors,

which

span the same plane as the

and

vectors

and

are related

to them by rotation.

The displacements

formed

grid in polar coordinates

with a spacing

of 0.05 Å in

and 20

. Parallel

computations

were aided

by the use of GNU

parallel.

The absolute

cross section

as a function

of wavelength

was

simulated

in a rough

approximation

neglecting

coupling

between

the 1

′

and 2

′

states

as the sum of two

independent

spectral

contributions.

Franck

−

Condon

factors

for the 2

′

←

′′

transition

at 0 K including

Duschinsky

rotation

effects

were calculated

using the ezFCF

package,

using vibrational

frequencies

and 0

−

0 transition

energies

taken

from the EOMEE-CCSD/ano-pVQZ

calculations.

The abso-

lute contribution

to the cross section

is given by

(

)

i f

(

)

/ 2

where

is an empirical

broadening

parameter

chosen

correspond

to a full width at half-maximum

of 100 cm

−

. The

′

contribution

to the spectrum

was estimated

using

multidimensional

extension

of the reflection

principle

and is

given by

The Journal

of Physical

Chemistry

pubs.acs.org/JPCA

Article

https://doi.org/10.1021/acs.jpca.3c04977

J. Phys.

Chem.

2023,

127, 8374

−

8382

8375

(

)

exp

i f

1 A

vertical

{

where

−

is the norm of the gradient

of the upper

state in

−

(see ref 42), and the

are the ground-state

vibrational

frequencies

(expressed

in cm

−

). EOMEE-CCSD

was used for

the vertical

excitation

energy,

excited-state

gradient,

and

transition

dipole

moment.

The ground-state

CCSD

vibrational

frequencies

were used for the

Orbitals

were visualized

with the IBOView

package,

while

vibrational

arrow

diagrams

were generated

with the PyVibMS

plugin

to pymol.

■

RESULTS

Experimental

ClSO A-Band

Spectrum.

Figure

1 shows

representative

spectra

of the Cl

SO/N

photolysis

system

at 40

Torr, recorded

at a delay time of 100

s. We have previously

reported

a strong

absorption

below

320 nm, which

has a

maximum

at 303 nm, and assigned

it to the 1

′′

←

′′

transition,

the B band of the ClSO radical.

We found

another

weak absorption

with a maximum

peak

position

near 385 nm, exhibiting

a series

of vibronic

bands.

The apparent

threshold

of this weak absorption

feature

around

480 nm. The slight variations

in baseline

above 500 nm

show the uncertainty

in baseline

under

high [Cl

SO]. The

precursor

may also have a small absorption

(Figure

S1). A fit

to the partially

resolved

vibrational

progression

observed

in the

experiment

yielded

an average

spacing

of 226 cm

−

. The

progression

was linear

over the range observed;

as shown

Figure

1, fitting

to a quadratic

function

was not a statistically

significant

improvement.

The weak absorption

band centered

near 385 nm has a

decay

behavior

identical

to the absorption

near 303 nm to

within

the experimental

uncertainty

(Figure

S2). The intensity

was found

to be proportional

to the concentration

of Cl

SO in

the reactor

(Figures

S3 and S4) when [Cl

SO] < 1.3

−

. The concentrations

of Cl

SO used in this experiment

were quite high ([Cl

SO]

≈

−

) to ensure

a large

enough

transient

signal.

However,

this high concentration

also

resulted

in a strong

absorption

of the photolysis

laser (

(248

nm) = 7.05

−

= 45 cm,

= 37.7%),

leading

inhomogeneous

radical

formation

throughout

the flow reactor

along

the excimer

path (Figure

S4). As a result

of these

conditions,

we did not attempt

a quantitative

kinetic

analysis

the time-dependent

data to obtain

the rate constants.

Based

the concentration

dependence

and temporal

behavior

distinct

wavelengths,

we tentatively

assigned

the origin

of the

weak band to ClSO radicals.

We calibrated

the absolute

cross section

of the weak band

near 385 nm from the absorbance

of the 1

′′

←

′′

transition

in the 310

−

320

nm range,

which

has a peak cross

section,

(303

nm) = (2.0

0.5)

−

Figure

demonstrates

that the weak band has a maximum

absorption

cross section

(385

nm) = (7.6

1.9)

−

near 385

nm (1 standard

deviation).

EOM-CCSD

Calculations.

Table

1 summarizes

the results

of coupled

cluster

point calculations

for the X

′′

ground

state

and the two lowest

lying

′

excited

states.

The 1

′

state

arises

from the

SOMO

−

LUMO

transition,

and the 2

′

state arises from the

SOMO-1

to SOMO

transition.

All three

states have bound

minima.

The adiabatic

transition

energy

the 2

′

state was found to be 20,463

−

, higher

than that of

the 1

′

state (16,122

−

); thus, there was no change

in the

energy

ordering

of the

′

excited

states.

The minimum

energy

of the 2

′

state is predicted

to be higher

than the bond

dissociation

energy

(Cl

−

SO)

= 19,048

−

using

the

HEAT-345(Q)

method

−

(see Table

S1), while

the

minimum

of the 1

′

state is predicted

to be below

(Cl

−

SO).

The observed

absorption

band

lies well above

dissociation.

The 1

′

excited

state is characterized

by a large ClSO angle

and a longer

ClS bond length

than those of the X

′′

state. The

′

excited

state has an almost

ClSO angle and a longer

SO bond length

and is closer

to the geometry

of the ground

state. Both

′

excited

states have lower bending

frequencies

(214 and 227 cm

−

) than the ground

state (316 cm

−

). The

′

state has a lower

ClS stretching

frequency

(333 cm

−

)

than the ground

state, while 2

′

has a lower

SO stretching

Figure

(Upper

panel)

Recorded

spectra

of the Cl

SO/N

/248 nm

system

at 100

s after pulsed

laser (exposure

time = 117.5

averaged

for 12,288

shots)

at 40 Torr and 292 K. Cyan, blue, and red

lines represent

spectra

at different

grating

angles

(center

wavelengths

of 370, 420, and 520 nm), while the gray lines show the background

noise without

SO. The black simulated

spectra

downward

is the

sum of contributions

from both the 1

′

(orange)

and 2

′

(olive,

stick spectra

convoluted

with a Gaussian

function

with fwhm

= 100

−

) states.

The comparison

between

experimental

and theoretical

absolute

cross sections

is shown

in Figure

S5. (Lower

panel)

Linear

and quadratic

fits to the observed

positions

of the vibrational

peaks

within

the marked

region.

The Journal

of Physical

Chemistry

pubs.acs.org/JPCA

Article

https://doi.org/10.1021/acs.jpca.3c04977

J. Phys.

Chem.

2023,

127, 8374

−

8382

8376

frequency

(735 cm

−

); both results

are consistent

with the

respective

bond length

changes.

Further

calculations

utilizing

closed-shell

references

produced

results

that were similar

those

of the EOMEE

calculations,

suggesting

minimal

spin

contamination

effects.

Assignment

of the 385 nm Spectrum.

Both low-lying

′

excited

states

are predicted

to have similar

vertical

excitation

energies

consistent

with the observed

band at 385

nm.

The transition

to the 1

′

state is dissociative,

while the

transition

to 2

′

is bound.

The observed

cross section

within

a factor

of 2 of the theoretical

absolute

cross-section

maximum,

the

(377 nm) = 1.8

−

. The cross section

relative

to that of the strong

Bband

is consistent

with the

oscillator

strengths

that we previously

predicted.

We found

′

) = 3.89

−

′

) = 2.88

−

, and

′′

) =

1.22

−

for transition

to the higher

excited

states.

The

ratio of the predicted

oscillator

strengths

is (

′′

)/(

′

)

′

))=

18.0, consistent

with the ratio of the observed

absolute

cross sections

at the peaks,

(305

nm)/

(385

nm) =

2.0

−

/7.6

−

= 26.3.

The bending

frequencies

predicted

for both

′

states agree

well with the experimentally

observed

vibrational

progression

of 266 cm

−

(see Figure

1). In addition,

Figure

2 shows

doublet

peaks

at 422, 435, and 439 nm, which

may be

attributed

to the difference

between

the SO stretching

frequency

and three times the bending

frequency

of the 2

′

state.

For instance,

the doublet

peak at 435 nm could

correspond

to the (4,2,0)

←

(0,0,0)

and (3,5,0)

←

(0,0,0)

transitions,

leading

to an estimated

frequency

for the SO

stretching

mode

of 755 cm

−

. From

the calculated

Franck

−

Condon

contour,

we extrapolated

the origin to be roughly

near

490 nm (20,400

−

), which

is consistent

with the 0

−

transition

energy

of the 2

′

state shown

in Table

The observed

vibrational

structure

disappears

on the blue

side, while the simulated

spectrum

exhibits

a clear vibrational

progression

of the 2

′

state across

the whole

band. The peak

broadening

on the blue side might

be due to congestion

and

anharmonicity

or decreases

in the lifetime

of the 2

′

excited

state caused

by internal

conversion

(

vide

infra

, see

′′

Influence

of the Conical

Intersection

′′

A few peaks within

the 320

−

360

nm range (Figure

S6) were

also observed,

which

have a small spacing

of approximately

367

−

and a large spacing

of 1060

−

. These

energy

differences

are close to the calculated

ClS stretching

and SO

stretching

frequencies

of the 1

′

state and may warrant

further

theoretical

investigation.

As a first-order

approximation,

we modeled

the spectrum

with the EOMEE-CCSD/ano-pVQZ

results,

assuming

that the

spectrum

was the incoherent

sum of the 1

′

←

′′

and

′

←

′′

transitions

weighted

by the computed

oscillator

strengths

(see Figure

1). We treated

the bound-to-bound

transition

to the 2

′

state as two independent

displaced

harmonic

oscillator

models

based

on the minimum-energy

geometries

of the X

′′

and 2

′

states

adopted

from the

EOMEE-CCSD/ano-pVQZ

calculations

including

Duschinsky

rotation

at 0 K (Figure

S7). We simulated

the 1

′

←

′′

transition

using a multidimensional

extension

of the reflection

principle.

The two states made roughly

equal contributions.

Table 1. Summary

of the Calculations

of ClSO for Geometries,

Harmonic

Frequencies,

Adiabatic

Transition

Energies

Relative

to the Ground

State, and the 0

−

0 Transition

Energies

SCF reference

UHF/EOMEE

RHF/EOMIP

(anion)

RHF/EOMEA

(cation)

geometry-optimized

′′

′

′′

′

′′

′

geometry

−

O)/Å

1.461

1.672

1.459

1.464

1.643

1.453

1.460

(Cl

−

S)/Å

2.046

2.013

2.276

2.034

2.012

2.035

2.259

∠

(ClSO)/

109.25

91.50

152.69

109.42

90.19

109.57

149.78

frequency

/cm

−

ClSO bending

316.3

226.5

214.3

320.6

221.0

323.1

232.9

ClS stretching

515.5

550.7

333.3

541.8

544.6

533.8

345.1

SO stretching

1199.5

735.4

1197.7

1190.3

802.8

1248.3

1195.4

/cm

−

CCSD/ano-pVQZ

20,697

18,517

21,027

17,558

(20,742)

(16,759)

(20,463)

(16,122)

−

0 transition

energy/cm

−

[nm]

20,482

16,616

20,221

15,956

[488.2]

[601.8]

[494.5]

[626.7]

Optimization

and frequency

at the CCSD/ano-pVQZ

level.

Correction

for the triples

HLC

= CCSDT/ano-pVDZ

−

CCSD/ano-pVDZ

similar

to ref 47.

At the CCSDT/ano-pVQZ

level.

Figure

Closer

look at the structure

of the weak absorption

band in

the 360

−

480

nm range.

A few doublet

peaks were observed,

yielding

an average

spacing

of 76 cm

−

. The stick spectrum

displays

the

intensities

predicted

for excitation

to the 2

′

state, incorporating

Franck

−

Condon

factors

and thermal

population

effects

at 0 K. The

energy

of the computed

transition

origin

has been shifted

to red by

870 cm

−

for alignment

with the vibronic

progression

of the

experimental

spectrum

(green

dotted

lines),

while

the red dashed

lines indicate

predicted

doublets.

Inset is a magnification

of the

spectrum

between

415 and 445 nm.

The Journal

of Physical

Chemistry

pubs.acs.org/JPCA

Article

https://doi.org/10.1021/acs.jpca.3c04977

J. Phys.

Chem.

2023,

127, 8374

−

8382

8377

Prediction

of a Conical

Intersection

between

the 1

′

and 2

′

States.

Our previous

calculations

found

that the

vertical

excitation

energies

of the two

′

states were very close

in energy.

This result

suggested

that a conical

intersection

between

the 1

′

and 2

′

states might

exist near the ground-

state geometry.

To investigate

this phenomenon,

we generated

a cut of the bending

potential

energy

curve

by linearly

connecting

the geometries

of X

′′

, 1

′

, and 2

′

states

(Figure

S8). We employed

the EOMIP

and EOMEA

approaches

to treat 1

′

and 2

′

independently

(Figure

S9). In addition,

the EOMEE/ano-pVTZ

calculation

failed to

converge

at the upper

state (Figure

S10).

To locate

the conical

intersection,

we explored

the lowest

three potential

energy

surfaces

(PES)

of the ClSO radical

using

XDW-CASPT2

31,39

with a seven-electrons

in five-orbitals

CASSCF

reference.

The active

orbitals

in the ground-state

equilibrium

geometry

are pictured

in Figure

5. The resulting

equilibrium

geometries

and relative

energies

are summarized

Table

We found

a sloped-type

conical

intersection

of the 1

′

and

′

states,

as depicted

in Figure

3, with a minimum-energy

conical

intersection

at 26,414

−

, close to the peak of the

absorption.

The geometries

of the MECI

and ground

state are

similar,

with the MECI

SO bond being slightly

longer.

Indeed,

we found

that the change

between

′′

and MECI

in the

branching

space is small (0.06 Å, Table 2) with a larger change

in seam space,

indicating

that the two

′

excited-state

PES do

cross near the Franck

−

Condon

region,

although

the MECI

slightly

far away from the ground-state

geometry.

The corresponding

motions

and

vectors

in the

branching

space as well as the motions

in seam space are also

illustrated

in Figure

4. Motion

along

the

vector

(large

contribution

from the antisymmetric

stretch)

on the lower

surface

lowers

the energy,

presumably

leading

to the 1A

′

well

(contracting

the SO bond

and elongating

the SCl bond).

Motion

along

the

vector

mainly

consisted

of symmetric

stretching.

The bending

mode

is the primary

motion

along the seam

space (Figure

4),

., the degeneracy

of the conical

intersection

is preserved

as the molecule

bends.

The bending

motion

thus

has the weakest

vibronic

coupling

among

the three vibrational

modes.

As a result,

we expect

the lifetimes

of the bending

vibronic

excited

states

to be longer

than other

modes,

consistent

with the observation

of the bending

motion

as the

most

prominent

structural

feature

in the experimental

spectrum.

■

DISCUSSION

Influence

of the Conical

Intersection.

Our current

time-

independent,

low-resolution

absorption

spectrum

does not

provide

more information

about the interactions

between

both

′

states

near the conical

intersection.

We attributed

the

observed

structure

to the 2

′

←

′′

state, but both the 1

′

and 2

′

states

contribute

to the observed

band.

While

simulated

the spectrum

as an incoherent

excitation

of the two

adiabatic

states,

the true excitation

process

produces

superposition

of the two states

and the coefficients

of each

state vary with energy,

in part due to changing

vibronic

coupling

across

the band. Additionally,

the excited

wave packet

would

undergo

nonadiabatic

dynamics

on the coupled

′

and 2

′

states,

especially

near the conical

intersection,

resulting

in a shorter

lifetime

and account

for the loss of the

vibrational

structure

toward

shorter

wavelengths.

The predominant

Franck

−

Condon

active

mode

in the

vibrational

progression

of the observed

spectrum

is the ClSO

bending

mode.

This is consistent

with the prediction

of the

bending

mode as the motion

along the seam space.

However,

broadening

at shorter

wavelengths

and observation

of the

bending

vibrational

progression

could result from other effects,

as discussed

above,

and therefore

only provide

circumstantial

evidence

for the presence

of a conical

intersection.

Table 2. Summary

of the Geometries

and Energy

Difference

Relative

to the Ground

State,

, Calculated

at the XDW-

CASPT2(7,5)

Level

geometry-optimized

′′

′

MECI

geometry

−

O)/Å

1.446

1.480

1.688

1.593

(Cl

−

S)/Å

2.007

2.259

1.997

2.010

∠

(ClSO)/

111.8

156.1

93.0

119.2

(branching

space)/Å

−

0.0283

−

0.4444

0.2473

(branching

space)/Å

−

0.0549

0.1769

−

0.1508

(seam

space)/Å

0.1611

−

0.4138

0.3517

/cm

−

14,431

23,198

26,414

Figure

Adiabatic

potential

energy

surface

of the ground

state and

two excited

′

states of the ClSO radical

in the 2D branching

space

(

; see Figure

4) computed

at the XDW-CASPT2(7,5)-3SA

level

near the MECI

geometry

(

= 0,

= 0). The meshed

line shows

the

profile

of the ground-state

vibrational

wave function.

The black arrow

shows

the vertical

transition

from the MECI

to the ground-state

PES

(which

is flattened

for clarity;

the color bar to the right gives the

energy

scale of the ground-state

PES contours).

Figure

Corresponding

vibrational

motions

of the seam space (the

set of geometries

where

′

and 2

′

states are degenerate)

and the

branching

space (the complement

of the seam space)

near the MECI

geometry.

The Journal

of Physical

Chemistry

pubs.acs.org/JPCA

Article

https://doi.org/10.1021/acs.jpca.3c04977

J. Phys.

Chem.

2023,

127, 8374

−

8382

8378

The fate of the excited

state of the ClSO radical

is unclear.

The molecules

excited

to the 2

′

state could

relax to the

lower

′

state via the conical

intersection

along

the

coordinate;

subsequent

collisions

could

further

relax the

molecule

below

the dissociation

limit into the bound

′

well, at which

point the molecule

could reach the ground

state

by internal

conversion,

collisional

quenching,

or fluorescence.

To quantitatively

predict

the effects

of the conical

intersection

on the observed

spectrum

is challenging

and

goes beyond

the scope

of this study.

Use of a single-shot

vibronic

coupling

model

based on EOM-CCSD

in the spirit of

Ichino

et al.

is complicated

by the lack of a single,

closed-

shell, noninteracting

reference

state.

Furthermore,

the two

electronic

states

the same irreducible

representation,

leading

to orbital

mixing

at the SCF level resulting

in diabats

that hold little chemical

significance.

Molecular

Orbital

Diagram

of the ClSO Radical.

Figure

5 summarizes

the valence

electronic

structure

of the ClSO

radical.

To construct

this MO diagram,

we used the ionization

energies

of the SO molecule

(10.3 eV)

and Cl atom (13.0

eV)

to establish

the relative

energy

positions.

Figure

S11 depicts

the nine molecular

orbitals

resulting

from

the combination

of 6

, 2

, and 7

orbitals

of SO, along

with the 3p orbitals

of the Cl atom under

symmetry.

These

orbitals

can be broadly

classified

as bonding,

antibonding,

and

nonbonding.

The term nonbonding

here is used loosely,

referring

to orbitals

that exhibit

both bonding

and antibonding

characteristics

simultaneously

along distinct

chemical

bonds.

For instance,

the second

occupied

molecular

orbital

(SOMO-1)

is a combination

of an in-plane

orbital

along

the SO bond and a

orbital

along the ClS bond.

This can be

interpreted

as the SO moiety

being

stabilized

by donating

electrons

from a

orbital

into the 3

orbital

of the Cl atom.

We may expect

that the reactivity

of the ClSO radical

does not

change

significantly

with respect

to the SO radical,

supported

by the nearly

identical

SO bond length

compared

to free SO

radicals

(

= 1.48 Å).

Since

the ClSO

+ Cl

→

reaction

has been reported,

we expect

that the association

reaction

of SO + Cl

→

ClSO also occurs.

This MO analysis

of the ClSO radical

provides

insights

into

the electronic

structure

of other

sulfinyl

radicals

as the

ionization

energy

of the Cl atom 3

orbital

is similar

to the

ionization

energies

of H, C, N, and O atoms.

Typically,

radical

species

possesses

seven

electrons,

occupying

three

perpendicular

orbitals

similar

to those

of the Cl atom.

Additionally,

we predict

that transitions

from the SOMO-1

to the SOMO

orbital

are likely

to occur

in other

sulfinyl

radicals,

resembling

the 2

′

←

′′

transition

observed

ClSO.

Furthermore,

a strong

←

transition

will also be

present,

similar

to the ClSO 1

′′

←

′′

transition.

Indeed,

the literature

has reported

the presence

of these two distinct

transitions

in certain

sulfinyl

radicals,

as summarized

in Table

Peroxyl

radicals

(RO

) are approximately

isoelectronic

molecules

of sulfinyl

radicals.

Experiments

have shown

that

two transitions

within

the near-IR

and UV

−

vis

regions

are

commonly

observed

in peroxyl

radicals.

Weisman

and Head-

Gordon

have successfully

explained

the observed

trend in

band positions

using a MO picture.

Building

upon similar

concepts,

we have concluded

that the

←

transition

in sulfinyl

radicals

occurs

at a similar

band

position

due to the minimal

mixing

of the orbital

character

from the substitution

group.

Conversely,

the band

positions

of SOMO

←

SOMO-1

transitions

cover a wide range

of wavelengths

because

the energy

of the SOMO-1

orbital

depends

on the characteristics

of the substitution

groups.

It is noteworthy

that the chemiluminescence

spectrum

HSO has been detected

in the visible

range for the SOMO

←

SOMO-1

transition,

which

suggests

that the ClSO radical

might

fluoresce

after excitation

to the A band and could work

as a better

tool to study the chemical

reactivity

and kinetics.

Figure

Schematic

MO diagram

showing

the active space orbitals

the ClSO radical

(top view) used in the XDW-CASPT2

calculation.

Vertical

and horizontal

arrows

indicate

transitions

to excited

states for

the

and

electrons.

The green lines connect

orbitals

with an out-of-

plane orientation.

Table 3. Summary

of the UV

−

Vis

Absorption

Band Positions

of Distinct

Sulfinyl

Radicals

(RSO)

unit/nm

←

SOMO

←

SOMO-1

sample

phase

190

−

240

gas

−

520

−

960

gas

−

∼

270

300

−

500

Ne-matrix

−

260

−

300

450

−

635

Ar-matrix

−

250

−

300

490

−

610

Ar-matrix

−

260

−

350

410

−

470

Ar-matrix

∼

300

∼

450

solution(C

)

CC(H)

−

240

−

310

350

−

490

-matrix

−

260

−

320

350

−

460

gas

Chemiluminescence

emission

spectrum.

The Journal

of Physical

Chemistry

pubs.acs.org/JPCA

Article

https://doi.org/10.1021/acs.jpca.3c04977

J. Phys.

Chem.

2023,

127, 8374

−

8382

8379