Добавил:
Опубликованный материал нарушает ваши авторские права? Сообщите нам.
Вуз: Предмет: Файл:

Baer M., Billing G.D. (eds.) - The role of degenerate states in chemistry (Adv.Chem.Phys. special issue, Wiley, 2002)

.pdf
Скачиваний:
17
Добавлен:
15.08.2013
Размер:
4.29 Mб
Скачать

conical intersections and the spin–orbit interaction

565

additional conditions. Geometrical constraints can also be used to map out the seam of conical intersection in its full dimensionality. This constrained minimization can be accomplished by minimizing the following Lagrangian [7]:

5

Ncon

 

X

X

ð15aÞ

LijðR; n; kÞ ¼ EieðRÞ þ ViðRÞxi þ

KiðRÞli

i¼1

i¼1

 

where n and k are Lagrange multipliers. Expanding Lij through second order yields the Newton–Raphson equations [7]:

2

vðRÞy

ð0

Þ

ð0

Þ

32 dn

 

QijðR; n; kÞ

v R

k R

 

dR

4

kðRÞy

0

 

0

 

54 dk

32

giðRÞ þ vyn þ kyk 5 ¼ 4 VðRÞ

KðRÞ

3

5ð15bÞ

where rVi vi and rKi ki and Qij rrLij. The gradients v are more costly to evaluate than their nonrelativistic counterparts. For this reason, it is useful to search along the direction corresponding to dR while Eije decreases. This simple extension of an idea from conjugate gradient theory can significantly reduce the computational effort needed to solve Eq. (15b). The performance of Eq. (15b) is discussed in Section III.

F.Conical Intersections: Description

In this section, notions used to describe nonrelativistic conical intersections are extended to the present case. For simplicity, unless otherwise specified we consider the Z ¼ 3 case. The analogous treatment for Z ¼ 5 will be reported in [17].

1.Orthogonal Intersection Adapted Coordinates

At a point of conical intersection, Rx;ij, the four degenerate wave functions are

defined up to a rotation U, consistent with time-reversal symmetry. As a consequence of this arbitrariness the gijðRÞ; hr;ijðRÞ; hi;ijðRÞ; hr;iTjðRÞ; hi;iTjðRÞ need not be orthogonal. Orthogonality of gijðRÞ; hr;ijðRÞ; hi;ijðRÞ; hr;iTjðRÞ;

hi;iTjðRÞ greatly simplifies the analysis of Eq. (13). In the nonrelativistic, Z ¼ 2, case, the degenerate 0K; K ¼ I; J are defined only up to a one parameter rotation. We used this flexibility to require orthogonality of gIJ and hIJ : Below we demonstrate how this orthogonality requirement can be extended to the Z ¼ 3 case.

For Z ¼ 3 define the rotated states by

 

 

~

~

TdÞU

ð16aÞ

ðd

TdÞ ¼ ðd

566

 

spiridoula matsika and david r. yarkony

 

 

 

where

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

dy ¼ ðdiy

djy Þ

 

 

u

0

 

ð16bÞ

 

 

 

 

 

 

U ¼ 0

u

and

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

u

¼

eiðaþgÞ=2 cos b=2

eið aþgÞ=2 sin b=2

ð

16c

Þ

 

 

e ið aþgÞ=2 sin b=2

e iðaþgÞ=2 cos b=2 !

 

Then, using Eqs. (16), in Eqs. (13b)–(13d) we deduce

 

 

 

 

~ji

¼ ð g

ji

cos b

þ h

r; ji

 

sin b cos g þ h

i;ij

sin b sin gÞ

ð17aÞ

g

 

 

 

 

 

 

and

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

~ ji

 

 

 

~r; ji

~i; ji

 

 

ð17bÞ

 

 

 

 

 

 

 

h

¼ h

þ ih

 

 

where

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

~r; ji

¼ g

ji

sin b cos a þ h

r; ji

ðcos b cos g cos a sin g sin aÞ

 

 

 

h

 

 

 

 

 

 

~i; ji

 

þ hi;ijðcos b sin g cos a þ cos g sin aÞ

 

ð17cÞ

¼ g

ji

sin b sin a h

r; ji

ðcos b cos g sin a sin g cos aÞ

 

 

 

h

 

 

 

 

 

 

 

 

hi;ijðcos b cos g sin a cos g cos aÞ

 

ð17dÞ

Then the three requirements

n1

¼ g

ji

 

~r; ji

¼ 0

 

h

 

~

 

 

 

 

 

n2

¼ g

ji

 

~i; ji

¼ 0

 

h

 

~

 

 

 

 

 

n3

~r; ji

~i; ji

¼ 0

¼ h

 

 

h

 

define a; b; g. For example, Eqs. (17a) and (17c) with nonrelativistic limit for Eq. (18a)

ð18aÞ

ð18bÞ

ð18cÞ

a ¼ g ¼ 0 give the

n1 ¼ 0 ¼ ð gji cos b þ hr; ji sin bÞ ðgji cos b þ hr; ji sin bÞ

 

 

ð19aÞ

The solution to Eq. (19a) is

 

 

 

 

 

 

 

 

 

 

 

 

 

 

2gji hr; jihr; ji hr; ji

gji gjiÞ ¼ tan 2b

 

 

 

 

ð19bÞ

as obtained previously [18]. Further,

choosing

b

¼

p=2; g

¼

0; a

¼

0 and

 

 

ij

 

r;ij

 

 

i;

 

 

 

b ¼ p=2; g ¼ 0; a ¼ p=2

shows that g , h

 

, and h

 

ij

are interchangeable at a

 

 

 

conical intersection. The

solution to the

nonlinear

Eqs. (18a)–(18c)

can be

conical intersections and the spin–orbit interaction

567

obtained numerically using the following Newton–Raphson procedure

 

niðxn þ dxÞ ¼ 0 ¼ niðxnÞ þ rniðxnÞ dx

i ¼ 1--3

ð20aÞ

so that

 

 

xnþ1 ¼ xn FðxnÞ 1nðxnÞ

 

ð20bÞ

where Fji ðq=qxjÞni is computed by divided difference and x ¼ ða; b; gÞ. The properties of the solutions of Eq. (20) will be discussed in Section III.

The orthogonal gij

r;ij

 

 

i;ij

define a set of Cartesian axes z

¼

gij

gij

 

^x ¼ h

r;ij

=h

r;ij

; ^y ¼ h

i;ij, h i;ij, and

h

 

ij

¼ jjg

ij

jj; h

r;ij

¼ jjh

r;ij

jj, and h

i^;ij

 

=i;ij

 

;

 

 

 

=h

; where g

 

 

 

 

 

¼ jjh

jj

that span the branching space. The associated coordinates (x; y; z) are referred to as orthogonal intersection adapted coordinates.

2.A Transformational Invariant

In the nonrelativistic case, at a given Rx;IJ , the quantity gIJ hIJ was shown to be invariant under the transformation in Eq. (16), for a ¼ g ¼ 0. This invariant, whose value depends on Rx;IJ , was used to systematically locate confluences, [18–21], intersection points at which two distinct branches of the conical

intersection seam intersect. Here, we show that the scalar triple product, gij hr;ij hi;ij is the invariant for Z ¼ 3. Since the gij; hr;ij, and hi;ij cannot be

assumed orthogonal the scalar triple product has the following form (suppressing the ij superscripts)

g hr hi ¼ ðgxi þ gy j þ gzkÞ ðhrxi þ hry j þ hrzkÞ ðhixi þ hiy j þ hizkÞ

¼ ðgxhyr gyhxr Þhzi þ ðgyhzr gzhyr Þhxi þ ðgzhxr gxhzr Þhyi

ð21Þ

To demonstrate the invariance insert, Eqs. (17a), (17c), and (17d) into Eq. (21) giving

~ ¼ ~ ~r ~i

I g h h

¼hrxhiy hryhixÞðsin b sin aÞ þ ð gxhiy þ gyhixÞðcos a sin g þ cos g cos b sin aÞ

þðgxhry gyhrxÞð cos a cos g þ cos b sin a sin gÞgf gzsin b sin a

hrzðsin g cos a þ cos b sin a cos gÞ hizðcos b sin a sin g cos g cos aÞg

þcyclic permutations

¼hrxhiy þ hryhixÞgz þ ðhryhiz hrzhiyÞgx þ ðhixhrz hizhrxÞgygfðsin2b sin2aÞ

þðcos a sin g þ cos g cos b sin aÞ2 þ ðcos b sin asin g cos g cos aÞ2g

¼ hi hr g ¼ g hr hi ¼ I

ð22Þ

The use of this invariant is discussed in Section III.

568

spiridoula matsika and david r. yarkony

3.Local Topography: Energy

The topography of a conical intersection affects the propensity for a nonadiabatic transition. Here, we focus on the essential linear terms. Higher order effects are described in [10]. The local topography can be determined from Eq. (13). For Z ¼ 3, Eq. (13) becomes, in orthgonal intersection adapted coordinates

~ ½1&

dR ¼ ðs

ij

ij

zrz þ h

r;ij

xrx h

i;ij

yry

ð23Þ

H

 

dRÞI g

 

 

where I is a 2 2 unit matrix, and the r are the Pauli matrices.

To determine the eigenfunctions and eigenvalues of Eq. (23), it is convenient

to introduce spherical polar coordinates,

x ¼ r cos f sin y; y ¼ r sin f sin y,

and z ¼ r cos y and make the definitions

 

 

 

 

 

 

 

 

 

 

 

 

hi;ij sin f ¼ hðfÞ sin zðfÞ

 

 

 

hr;ijcos f ¼ hðfÞ cos zðfÞ

 

ð24aÞ

 

hðfÞ2

¼ ðhi;ij sin fÞ2 þ ðhr;ij cos fÞ2

 

 

 

 

hi;ij

 

 

 

 

 

 

 

 

tan zðfÞ ¼

 

tan f

 

 

ð24bÞ

 

 

hr;ij

 

 

hðfÞ sin y ¼ qðy; fÞ sin lðyÞ

 

 

 

gij cos y ¼ qðy; fÞ cos lðyÞ

ð24cÞ

q

 

;

 

2

 

h

 

 

 

sin

 

2

 

gij cos

 

2

tan

;

 

 

hðfÞ

tan

 

 

 

24d

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

fÞ

 

¼ ð

ðfÞ

yÞ

 

þ ð

yÞ

 

fÞ ¼ gij

y

 

ð

Þ

 

ðy

 

 

 

 

 

 

 

lðy

 

 

 

Then Eq. (23) becomes

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

~

½1&

dR

¼ Iðs

ij

dRÞ þ rqðy; fÞð rz cos lðy; fÞ þ sin lðy; fÞMðe

iz f

ÞÞrxÞ

 

H

 

 

ð

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

ð25Þ

where MijðxÞ ¼ dijðxd1j þ x d2jÞ. This Hamiltonian can be diagonalized by the transformation

Uðy; fÞ ¼

cos lðy; fÞ=2

 

 

 

 

sin lðy; fÞ=2

 

ð26aÞ

e izðfÞ sin lðy; fÞ=2

e izðfÞ cos lðy; fÞ=2

that is

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

~

½1&

 

RU

 

s

ij

 

R

I

q

 

;

 

 

 

27

 

 

UyH

 

d

¼ ð

 

d

ðy

fÞrz

 

ð

Þ

 

 

 

 

 

 

 

Þ

 

r

 

 

 

From the preceding analysis, it is seen that the coordinate space near Rx;ij can be usefully partitioned into the branching space described in terms of intersection adapted coordinates ðr; y; fÞ or (x; y; z) and its orthogonal complement the seam space spanned by a set of mutually orthonormal set wi; i ¼ 4 Nint. From Eq. (27), spherical radius r is the parameter that lifts the degeneracy linearly in the branching space spanned by ^x; ^y; and ^z.

conical intersections and the spin–orbit interaction

569

These results can be simplified by introducing scaled orthogonal intersection adapted coordinates x0 ¼ xhr;ij; y0 ¼ yhi;ij, and z0 ¼ zgij. In these coordinates,

rhðfÞ ¼ r0h0

ðf0Þ ¼ r0; rqðy; fÞ ¼ r0q0ðy0; f0Þ ¼ r0; zðfÞ ¼ z0ðf0Þ ¼ f0, and

lðy; fÞ ¼ l0

ðy0; f0Þ ¼ y0, where x2 þ y2 ¼ r2. This is the coordinate system

to be used to consider Berry’s geometric phase theorem [22].

4. Local Topography: Conical Parameters

In the nonrelativistic case, the key linear portion of the double cone is

characterized

by four conical

parameters:

a

 

strength

 

parameter

d

H½ð

 

IJ2

Þ þ ðh

IJ2

Þ=2&, an IJ

ij

¼ ð

 

IJ2

Þ ð

 

IJ2

Þ=ð

 

IJ2

Þ þ ð

IJ2¼

g

 

 

g

 

h

 

g

 

Þ

 

 

 

 

asymmetry parameter

 

 

 

 

 

 

 

 

h ,

and two tilt parameters sw =d; w ¼ x; y. If s ¼ 0 the double cone is vertical. The affect of these parameters on nuclear dynamics has been investigated using time dependent wavepackets [23]. Here the situation is similar but more complicated. In this case, the six parameters, sijw w ¼ x; y; z; gij; hr;ij, and hi;ij can be used to define a strength parameter, d ¼ H½ðgij2 Þ þ ðhr;ij2 Þ þ ðhi;ij2 Þ&, three tilt parameters

sijw=d w ¼ x; y; z and two asymmetry parameters, 2ðfÞ ¼ ½ðhðfÞ2 gij2 Þ= ðhðfÞ2 þ gij2 Þ and 1 ¼ ðhr;ij2 Þ ðhi;ij2 Þ=ðhr;ij2 Þ þ ðhi;ij2 Þ&: In future work, the

affect of these and higher order parameters on nuclear dynamics will be considered.

5.Derivative Couplings

By using Eq. (10), the derivative couplings

 

 

 

fijðRÞ ¼ h ieðr; RÞjr jeðr; RÞir

 

 

 

 

ð28aÞ

are given by

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

fijðRÞ ¼

 

xliðRÞrxljðRÞ þ

l2P

NliðRÞrNljðRÞ þ CSFfijðRÞ

 

ð28bÞ

 

l2Q

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

X

 

 

 

 

 

 

 

 

X

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

where the

nonsingular

term

CSF

f

ij

ðRÞ

is largely

 

negligible

near

a

conical

intersection

[20]. As in

the

 

 

 

 

f

ij

is

singular at

a

conical

nonrelativistic

case,

 

intersection. The singularity appears in the lowest order contribution:

 

 

 

 

 

fkl0Þ

¼

 

~ð0Þ;k

ðy

;

f

; w

~ð0Þ;l

ðy

;

f

; w

Þ

 

k; l

2

i; j

 

ð

28c

Þ

 

 

 

xa

 

 

 

Þrxa

 

 

 

 

 

 

 

 

X

a2Q

 

 

 

n

 

ðr ¼

0;

y

;

f

 

 

 

 

To evaluate fkl0Þ the ~ð0Þ;w

 

 

 

 

), are required.

Uðy; fÞ by

 

 

 

 

 

 

 

 

 

 

 

 

 

~ð0Þ;i

ðr ¼

0

~ð0Þ; j

ðr

¼

0

; y; fÞÞ ðn

ð0Þ;i

ð

Rx;ij

Þ;

ðn

 

; y; fÞ; n

 

 

 

These are given in terms of

nð0Þ; jðRx;ijÞÞUðy; fÞ ð29Þ

570 spiridoula matsika and david r. yarkony

Note that since the eigenfunctions are determined only up to an overall phase, the

 

 

 

 

 

~

½1&

dR:

 

 

 

 

 

following transformations also diagonalize H

 

 

 

 

 

 

V

ðy

;

fÞ ¼

eizðfÞ=2 cos lðy; fÞ=2

eizðfÞ=2 sin lðy; fÞ=2

!

ð

26b

Þ

 

 

e izðfÞ=2 sin lðy; fÞ=2

e izðfÞ=2 cos lðy; fÞ=2

 

B

ðy

;

fÞ ¼

cos lðy; fÞ=2

eizðfÞ sin lðy; fÞ=2

!

 

ð

26c

Þ

 

 

e izðfÞ sin lðy; fÞ=2

cos lðy; fÞ=2

 

 

As discussed in detail in [10], equivalent results are not obtained with these three unitary transformations. A principal difference between the U; V, and B results is the phase of the wave function after being transported around a closed loop C, centered on the z axis parallel to but not in the (x; y) plane. The perturbative wave functions obtained from Uðy; fÞ or Bðy; fÞ are, as seen from Eq. (26a) or (26c), single-valued when transported around C that is h ei ðr; R0Þj ei ðr; RnÞi ¼ 1, where R0 ¼ Rn denote the beginning and end of this loop. This is a necessary condition for Berry’s geometric phase theorem [22] to hold. On the other hand, the perturbative wave functions obtained from V(y; f) in Eq. (26b) are not single valued when transported around C.

U, V, and B also yield different fkk. By using Eqs. (24a)–(24d) and (26a) the derivative couplings are

fij0Þ ¼ 1=2ðrlÞ þ iðrzÞð1=2Þ sin l

ð30aÞ

fii0Þ ¼ iðrzÞ sin 2l=2

ð30bÞ

fjj0Þ ¼ iðrzÞcos 2l=2

ð30cÞ

From Eqs. (30a)–(30c), the singularity in fkl, as the conical intersection is approached, is of order 1=r. Only foij; o ¼ y; f are singular [10]. As in the nonrelativistic case, knowledge of the singular part of the derivative coupling can be used to construct a local diabatic representation that removes the singularity

[10].

If Vðy; fÞ had been used in lieu of Uðy; fÞ; jfij0Þj would be unchanged, but fkk0Þ becomes fkk0Þ;V , which is given by

fii0Þ;V ¼ fii0Þ þ irz=2

¼ iðcos lÞ=2rz

ð30dÞ

fjj0Þ;V ¼ fjj0Þ þ irz=2

¼ iðcos lÞ=2rz

ð30eÞ

Finally, had Bðy; fÞ been used, fii0Þ would be unchanged, but

fjj0Þ;B ¼ fjj0Þ þ irz ¼ iðrzÞ sin 2 l=2 ¼ fii0Þ

ð30fÞ

conical intersections and the spin–orbit interaction

571

III.NUMERICAL RESULTS

In this section, the spin–orbit interaction is treated in the Breit–Pauli [13,24–26] approximation and incorporated into the Hamiltonian using quasidegenerate perturbation theory [27]. This approach, which is described in [8], is commonly used in nuclear dynamics and is adequate for molecules containing only atoms with atomic numbers no larger than that of Kr.

A.1,22A0 and 12A0 States of H2 þ OH

The nonrelativistic 1,22A0 conical intersection seam in the H2 þ OH supermolecule has been well studied [28–30] because of its role in the non-adiabatic quenching reaction

H

2

þ

OH

A2 þ

Þ !

H

2

þ

OH

X2

Þ

or H

O

H 2S

Þ

 

 

ð

 

 

ð

 

2

 

þ ð

 

The C1v portion

of this

seam is a

2 þ 2

symmetry-allowed

conical

intersection. The character of the seam including spin–orbit coupling can be understood by starting with the degenerate 12 and 12 þ states. Turning on the spin–orbit interaction within the 2 manifold splits the 2 state into a (lower energy) 2 3=2 state and (higher energy) 2 1=2 state (see Fig. 1). Then with the full spin–orbit interaction included the molecule can distort to make either the upper pair, the 2E0 and 3E0 states, or the lower pair, the 1E0 and 2E0 states, degenerate. From Figure 1, for C1v geometries the 2E0 3E0 intersection is a ‘‘same symmetry’’, ¼ 1=2; 1=2, intersection, while the 1E0 2E0 intersection is a different symmetry ¼ 3=2; 1=2 intersection. However, both intersections are conical intersections since both ¼ 3=2 and ¼ 1=2 states decompose into one E0 and one E00 (Kramers’ doublets) when the molecule is distorted to Cs configurations. Here, we consider the more computationally challenging same

Figure 1. Representation of degenerate states from nonrelativistic components. (a) Degenerate zeroth-order states at Rx Rx;IJ . (b) Spin–orbit interaction splits 2 state. (c) With full spin–orbit interaction turned on, degeneracy is restored by changing geometry to Re;x Rx;ij.

572

spiridoula matsika and david r. yarkony

symmetry, 2E0 3E0, intersection to illustrate the ideas developed in Section 2.F.5. This system provides a stringent test of the numerical procedures since the spin–orbit interaction is relatively modest and the energy splitting changes rapidly in the region of interest.

The nonrelativistic states are described at the first-order configuration interaction level using a six orbital, eight electron, active space with the oxygen 1s orbital kept doubly occupied. The molecular orbitals were constructed from a state-averaged multiconfigurational self-consistent field procedure [31] using an extended atomic orbital basis on oxygen and hydrogen. The details of this description can be found in [30].

In the present calculations, the molecule is restricted to Cs symmetry. There are five internal degrees of freedom (the out-of-plane mode is excluded

to preserve Cs symmetry). Nuclear configurations will be denoted R ¼ ðRðH1 OÞ; RðO H2Þ; RðH2 H3Þ; H1OH2; OH2H3Þ corresponding to the

arrangement H1 O H2 H3. It will be convenient to refer to the R by their RðH2 H3Þ value, writing, RðRðH2 H3Þ ¼ bÞ RðbÞ ¼ ðRðH1 OÞ; RðO H2Þ; RðH2 H3Þ ¼ b; H1OH2; OH2H3Þ. For collinear geometries, the two angles will be suppressed. Equation (14) defines only three internal coordinates. Therefore two additional constraints are needed. These are provided by the value of RðH2 H3Þ ¼ b and/or the energy minimization requirement.

B.Convergence of Eq. (15b)

Figure 2 illustrates the efficacy of Eq. (15), considering convergence, in the absense of geometrical constraints, to a local energy minimum on the relativistic seam of conical intersection. Reported are the relativistic energy separation E32e and the nonrelativistic energy separation, E22A0;12A0 . The search was initiated at the structure indicated on the left-hand side of Figure 2, a point slightly displaced from the nonrelativistic seam. At this point, E22A0;12A0 11 cm 1 andE32e 70 cm 1. At the converged structure, achieved after 15 iterations,

pictured on the right-hand side, E32e < 0:2 cm 1 while E22A0;12A0 70 cm 1. The large changes in E32e between iterations 8 and 9, and 12 and 13, reflect, in

part, the use of the ‘‘conjugate gradient’’ extrapolation noted previously. These results strongly support the utility of the present approach. It is worth noting that once an initial point on a seam is found locating additional points is facilitated by the fact that given an Rx;ij corresponding to given K, Eq. (15b) can be used to predict a good starting value for a neighboring Rx;ij0 corresponding to K0 [32].

C.The Seam: Locus

Further evidence of the efficacy of the algorithm for locating points of conical intersection is provided in Figure 3, which reports additional points on the 2E0– 3E0 intersection seam, determined by introducing the geometrical constraint,

conical intersections and the spin–orbit interaction

573

Figure 2. EðrelÞ E32e and EðnonrelÞ E22 A0;12 A0 at each iteration of the

solution of

Eq. (15b) for OH þ H2 using multireference configuration interaction wave functions.

 

RðH2 H3Þ ¼ b. Note that C1v symmetry was not imposed. The points located on the 2E0–3E0 seam of conical intersection, which are degenerate to <1 cm 1, all had C1v symmetry. Figure 3 shows, that while the Z ¼ 3 seam is necessarily distinct from the nonrelativistic seam the separation is not large. In future work, it will be interesting to see how this conclusion changes as the magnitude of the spin–orbit interaction increases. Along the nonrelativistic seam the relativistic

energy difference, Ee

ð

Rx:IJ

R

H2 H3

, is

 

70 cm 1

> 50% of the OH(2

)

32

ð

ð

 

ÞÞ

 

 

 

fine structure splitting [33] suggesting that when heavier atoms such a chlorine,

where

Aso is

 

780 cm 1

[34], or

even

bromine or iodine are

involved,

e

ðR

x;IJ

Þ,

 

 

so

that nonadiabatic effects

may be

Eji

 

will be much larger,

significantly reduced at the nonrelativistic seam by the inclusion of spin–orbit coupling.

 

 

The

small

Ee

ð

Rx;ij

 

R

H2 H3

ÞÞ

<1 cm 1 and much larger E

2

 

2

 

ð

 

x;ij

ð

 

ð

 

 

32

 

ð

ð

 

2

 

A0;1

 

A

R

 

 

 

ÞÞe

 

 

 

70 cm 1, provide prima facie evidence for a conical

 

 

 

R

H2 H3

, also

 

intersection of H . However, since numerical degeneracies are never exact, an

574

spiridoula matsika and david r. yarkony

Figure 3. The relativistic seam Rx;ijðRðH2 H3ÞÞ: RðO–H1Þ, (empty squares), RðO H2Þ (empty diamonds) E E2E0 ðRðH2 H3ÞÞ (empty circles) on the 2E 3E seam of conical intersection. Filled markers on 12A0 22A0 nonrelativistic seam of conical intersection. The zero of energy is EðnrÞ E12A0 ðRx;IJðRðH2 H3Þ ¼ 2:336ÞÞ ¼ 76:486688 a.u.

alternative means is required to prove the existence of a conical intersection. The demonstration that g hr hi ¼6 0, that is, that g; hr; hi are linearly independent, serves to confirm the ‘‘conical’’ character near the intersection. It is to the determination of these quantities that we now turn.

D. The Seam: Conical Parameters and the Invariant

The lowest order contributions to the energy are described by the conical parameters g; hr; hi, and sk; k ¼ x; y; z, or by d; i ¼ 1; 2 and sk; k ¼ x; y; z. Here and below the superscript ij is suppressed when no confusion will result. We also will use the nonrelativistic convention gijjj^x; hr;yjj^y and hi;ijjj^z, where jj is real ‘‘is parallel to.’’ These parameters [9] are reported in Figure 4a and b. Their continuity is attributable to the use of orthogonal intersection adapted coordinates. For comparison, Figure 4a and b reports the nonrelativistic quantities gIJ ; hIJ , and sIJ , respectively. While noting that there is no unique correspondence