F I J : A COLUMBUS Compatible program for finding
confluences
• Andy Young• At the Yarkony Group• JHU
USES OF FIJ AND POLYHES
• Polyhes -> minimization of arbitrary states• Polyhes -> minimization of the crossing
seam• FIJ -> Fitting potential surfaces• FIJ -> Mapping Loci of Confluences
Definition of a Confluence
2 seams with different # of dimensions meet; Confluence embedded in seam of higher
symmetry
Some molecules known to have confluences
• BH2
• O3
• C(3P) + H2
• HNCO• Possibly adding the CH2CHO radical soon
Why is a confluence important?
• Suppression of geometric phase effect• Polar angle in the gh plane is normally
arbitrary • Possible effect on dynamics
Criteria for the Existence of a Confluence
• Will first consider a general 2 by 2 symmetric Hamiltonian with G and –G as diagonal components and W as off diagonal.
• Will assume that W is a product of potentials associated with the two separate symmetries. This will be used to define a quantity that can identify a confluence.
Generalized Hamiltonian
( ) ( ) ( ) ( ) ( )
( ) ( ) ( ) ( ) ( )
S R G R W R G R W R
W R S R G R W R G R
⎛ ⎞ ⎛ ⎞− −→⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎜ ⎟+⎝ ⎠ ⎝ ⎠
ur ur ur ur ur
ur ur ur ur ur
Allow
1 2( ) ( ) ( )W R V R V R=ur ur ur
then
1 2
1 2
( ) ( ) ( )( ) ( )
( ) ( ) ( )( ) ( )
G R V R V RG R W R
V R V R G RW R G R
⎛ ⎞⎛ ⎞ −−→ ⎜ ⎟⎜ ⎟⎜ ⎟ ⎜ ⎟
⎝ ⎠ ⎝ ⎠
ur ur urur ur
ur ur urur ur
Defining the g-h plane for regular Hamiltonian
( ) ( )
( ) ( )
g R G R
h R W R
t g h
= ∇
= ∇
= ×
ur ur ur
r ur ur
r ur r
Recall also that• If we approximate
then G0=S • We removed common part
of diagonal components• If we are at a degeneracy,
then the zeroth part of an expansion of W is also zero.
• So all components of our generalized Hamiltonian starts with linear terms.
0 1( ) ( ) ( )G R G R G R≈ +ur ur ur
In terms of electronic wavefunctions Psi(I,J)
( )12 I I J Jg H H= Ψ ∇ Ψ − Ψ ∇ Ψ
ur
I Jh H= Ψ ∇ Ψr
With our ‘new’ Hamiltonian, define
( )1,21 2( ) ( ) ( )t G R V R V R= ∇ ×∇
uur ur ur ur
( ) ( )1 22 1( ) ( )V R g h V R g h= × + ×ur ur r ur ur r
0=
Hyper-cylindrical coordinates
{ }1 2, , , ,... mz z zρ θ
int 2m N= −
The seam coordinates are
{ }kz
Some “intermediate” or reduced quantities
( ) ( )2 2 2 2cos sinq g hθ θ= +
( )( )
sinarctan
coshg
θλ
θ⎛ ⎞
= ⎜ ⎟⎜ ⎟⎝ ⎠
Uses of the Polar Coordinates:First Order Energy in terms of the Tuning and Coupling Coordinates and the Tilt Parameters
( )
( ) ( )
( )
( )( )( ) )()sin()cos(
)(sin)(cos)sin()cos(
1
1
)(sin)(cos)sin(1)cos(121
21
2222
2222
θρθθρ
θθρθθρ
θθρθθρ
qssE
hgssE
hsh
s
gsg
s
hghsh
gsg
E
EHHs
yx
yx
IJIJy
IJIJx
IJIJIJIJ
IJJJIIIJ
±+=
+±+=
⋅=
⋅=
+±⎟⎟⎠
⎞⎜⎜⎝
⎛⋅+⋅=
∆=Ψ∇Ψ+Ψ∇Ψ=
r
r
rr
r
Decomposition of the Derivative Coupling
1i
mIJ IJ IJ IJ
zi
f f f fρ θ=
= + +∑
Expressions in terms of Hamiltonian elements expansion polynomials M
( ) ( )( )
( )( )
12
zIJ M Mf z
q q
ρ
θ
θ θλ θ ρ
θ θ θ θ θ⎛ ⎞∂ ∂ ∂
= + +⎜ ⎟⎜ ⎟∂ ∂ ∂⎝ ⎠
0z =
( ) ( )( )
1 12
IJ Mf
q
ρ
θ
θλ θ
ρ ρ θ θ θ⎛ ⎞⎛ ⎞ ∂ ∂
= + ⎜ ⎟⎜ ⎟ ⎜ ⎟∂ ∂⎝ ⎠ ⎝ ⎠
Maintaining the Position of the Seam Singularity Through Coordinate Transformation Requires
( ) ( ) ( )( )1, sin ,cosIJ IJ IJx yf f fθ θ θ
ρ⎛ ⎞
= −⎜ ⎟⎝ ⎠
The last equation on the previous slide shows that singularity can be isolated to a single term, and the remaining parts of the derivative coupling can be used to optimize seam coordinates, search for confluences, and characterize PES’s in the vicinity of the intersection.
Derivative coupling asymmetry in gh plane near a confluence
( ) ( ) ( ) ( )2 2 2 2 2
1 12 2 2 cos sin
gh ghq g h
λ θρ θ θ θ θ
∂= =
∂ +
0θ =for12hg
=
2πθ =for
12gh
=
~0 for h/g < 0.001
Search Method
1
m
kk kk
Z zχ ξ=
=∑ur r
( )2,k hk
ha
ξ = −
( ){ }2,hka
11
m
kk
χ=
=∑
Analytic values for low order expansion coefficients
( ) ( ) ( ) ( ) ( ) ( ) ( )1, 2,1 sin 2 cos cos sin2 k
h hIJIJ z k k
k
E q f a az
λ θ λ θ θρ⎛ ⎞∂
− ∆ + = +⎜ ⎟∂⎝ ⎠
Unit First-Order Cone = Aqua; Quadratic Cone = Red
Cs 12A’-22A’’
Nearby MEX’s of different symmetries; Cs
Nearby MEX’s of different symmetries; C1
References
• Yarkony, Matsika. J.Phys.Chem.A. vol.106, pg. 2580, (2002.)
• Yarkony. J.Phys.Chem.A. vol. 105, pg. 2642. (2001.)• Ivanic, Atchity,Ruedenberg. JCP vol. 107, pg. 4307.
(1997).• Yarkony, JCP vol. 112, pg. 2111. (2000.)• Yarkony, JCP, in press.• Young, Yarkony, JCP, in press.
• I would like to thank my advisor, Dr. Yarkony.
• I would also like to thank all of the COLUMBUS developers who have enabled us to produce results more efficiently.
Things that we think would be worthwhile with regards to
COLUMBUS• Reduce code size with fortran 90 modules.• Use select_kind values and allocated memory;
remove “work arrays” in favor of allocatable types or actual pointers (now officially supported in fortran language.)
• Change parallel CI from GA based to MPI based.• Fix parallel CI to run with cidrtfl from cidrtms.x
and not just from cidrt.x.• Calculate interstate density matrix without using
sum rules (to improve efficiency.)