You are on page 1of 10

Quantized Hamilton dynamics

Cite as: J. Chem. Phys. 113, 6557 (2000); https://doi.org/10.1063/1.1290288


Submitted: 18 May 2000 • Accepted: 11 July 2000 • Published Online: 12 October 2000

Oleg V. Prezhdo and Yu. V. Pereverzev

ARTICLES YOU MAY BE INTERESTED IN

Quantized Hamilton dynamics for a general potential


The Journal of Chemical Physics 116, 4450 (2002); https://doi.org/10.1063/1.1451060

Extension of quantized Hamilton dynamics to higher orders


The Journal of Chemical Physics 116, 8704 (2002); https://doi.org/10.1063/1.1474585

Semiclassical Moyal dynamics


The Journal of Chemical Physics 149, 244113 (2018); https://doi.org/10.1063/1.5067005

J. Chem. Phys. 113, 6557 (2000); https://doi.org/10.1063/1.1290288 113, 6557

© 2000 American Institute of Physics.


JOURNAL OF CHEMICAL PHYSICS VOLUME 113, NUMBER 16 22 OCTOBER 2000

Quantized Hamilton dynamics


Oleg V. Prezhdo and Yu. V. Pereverzev
Department of Chemistry, University of Washington, Seattle, Washington 98195-1700
共Received 18 May 2000; accepted 11 July 2000兲
The Hamilton approach to classical dynamics is extended to incorporate quantum effects.
Quantization of the Hamilton equations of motion results in a hierarchy of equations that are
equivalent to quantum mechanics in the Heisenberg form. Closure of the hierarchy gives
approximations to the exact quantum dynamics. A specific dynamics algorithm is presented and
tested against model applications that exhibit tunneling and zero point motion effects. The quantized
Hamilton approach is found accurate, consistent, flexible, and computationally very efficient.
© 2000 American Institute of Physics. 关S0021-9606共00兲00638-3兴

I. INTRODUCTION Another type of the quantum-classical approximation is


needed when the same degree of freedom is neither classical,
Nonrelativistic quantum mechanics provides a rigorous nor fully quantum-mechanical. For example, the same mode
framework for investigation of most problems encountered can switch between quantum and classical regimes depend-
in chemistry. Nevertheless, more often than not it is much ing on the available energy. Situations of this sort are, per-
easier to treat problems classically, rather than quantum me- haps, more common than the purely classical/purely quan-
chanically. Each quantum mechanical degree of freedom re- tum combinations. Proton motion from one water molecule
quires an infinitely dimensional Hilbert space. Classically, to the next involves quantum tunneling. At the same time, a
the Hilbert space is mapped onto a two-dimensional phase proton attached to the one molecule is well described by
space. Low dimensionality and computational simplicity of classical mechanics. A wide amplitude torsional motion
classical mechanics provides a strong driving force for clas- around a C–C bond in a molecule such as ethane CH3 –CH3
sical and mixed quantum-classical approaches. A separation is very well described classically, except near the top of the
of a system into quantum and classical parts can be based on rotation barrier where tunneling through the H•••H repulsion
relative masses or energies associated with different degrees becomes important. The borderline between quantum and
of freedom. For example, molecules consisting of electrons classical behaviors of the same particle depends on particle’s
and nuclei are customarily split into quantum and classical energy. The CH3 –CH3 torsional motion is classical at high
subsystems and are viewed as a collection of quantum- energies. When the total energy is close to the energy of the
mechanical electrons moving in the field of classical nuclei. rotational barrier, tunneling comes into play. As the energy is
This description gives the Born–Oppenheimer approxima- lowered further, the dynamics once again becomes classical.
tion that is justified by the large ratio of the nuclear and At very low energies, the torsional motion proceeds around
electronic masses. High frequency vibrational modes of the the minimum of the potential, and the quantum zero point
nuclei can be also treated quantum mechanically, while the energy 共ZPE兲 effect dominates. In a diatomic molecule, ZPE
low frequency modes are described classically. The high/low plays its role at low vibrational energy, but is negligible near
frequency separation of the vibrational modes is justified by the dissociation limit. Examples of this kind can be found in
the number of available quantum states that depends on the all states of matter and systems of any size.
total energy relative to the quantum of vibrational energy. In Several approaches that deal with tunneling, ZPE
both cases, quantum and classical mechanics are applied to and other quantum corrections to classical mechanics
different degree of freedom. The quantum and classical de- have been proposed. Various semiclassical
grees of freedom are then coupled to each other. The cou- methods5,14–17,20–22,41–46 link quantum and classical mechan-
pling scheme is not unique. Various quantum-classical meth- ics. Tunneling is typically treated via WKB approximations,
ods that differ in the coupling procedure have been where the tunneling amplitude is calculated along a trajec-
developed.1–30 Classical p and q can be coupled to the cor- tory that connects the point of entrance into the barrier with
responding quantum expectation values based on the Ehren- the exit point.41,42,45,46 Straightforward in one dimension,
fest theorem.31 The result is the mean-field semiclassical description of tunneling becomes more compli-
approach.2–4,12,13,20,24,26 Extensions of the mean-field method cated in many dimensions.45,46 While the entrance into the
to include higher order quantum observables lead to multi- barrier is specified by the point where the classical trajectory
configuration mean-field1,16,26,28 and surface hits the barrier, the exit point is not unique. All possible exit
2,3,6–13,25,27
hopping methods, where several classical trajecto- points should be sampled and imaginary time trajectories
ries are coupled to a single quantum mechanical evolution. should be propagated that connect the entry and exit
Application of these methods range from gas phase scatter- points.14,15,42 The multiple spawning approach to tunneling47
ing reactions9,14,15,22,32–34 to surface,11,35 solution,7,8,10,36–38 is a variant of this procedure, where the tunneling probability
and biological chemistry.39,40 is determined by the overlap of Gaussian wavepackets rep-

0021-9606/2000/113(16)/6557/9/$17.00 6557 © 2000 American Institute of Physics


6558 J. Chem. Phys., Vol. 113, No. 16, 22 October 2000 O. V. Prezhdo and Y. V. Pereverzev

resenting classical particles, rather than from the WKB ap- i


proximation. In order to account for ZPE using classical me- 关关 A,B 兴兴 ⫽ 关 A,B 兴 ⫹ 兵 A,B 其 . 共1兲

chanics, classical trajectories are sorted out by certain criteria
that are designed to exclude trajectories without ZPE.43,44,48 The bracket is the sum of the quantum commutator 关쐓, 쐓兴
The difficulty of such approaches lies in the fundamental fact and the classical Poisson bracket 兵쐓, 쐓其. Separation of the
that ‘‘criteria based on the behavior of individual trajectories, total system into quantum and classical subsystems defines a
independent of the behavior of the ensemble, are not state space that is a direct sum of the quantum Hilbert and
valid.’’44 Simulation of ensembles of coupled trajectories can classical phase spaces. Observables, including the Hamil-
be very expensive computationally in many-dimensional sys- tonian, are functions of quantum operators acting in the Hil-
tems. The coupling terms are typically very sensitive to the bert space and classical p and q in the phase space. For
density of trajectories. Trajectories rapidly diverge, and sta- example, if an observable is a product of quantum ␣, ␤ and
tistical ensembles of very large sizes are required for good classical a, b functions, such as in a quantum-classical cou-
convergence. The nonlocal nature of quantum mechanics pling term Qq, the corresponding Lie bracket has the form
complicates development of semiclassical methods based on
关关 ␣ 共 P̂,Q̂ 兲 a 共 p,q 兲 , ␤ 共 P̂,Q̂ 兲 b 共 p,q 兲兴兴
swarms of trajectories, as observed in a number of approxi-
mate schemes.49–55 i
The approach introduced in the present paper extends ⫽ 关 ␣ , ␤ 兴 ab⫹ 兵 a,b 其 共 ␣␤ ⫹ ␤␣ 兲 /2. 共2兲

classical Hamilton mechanics in a conceptually novel way. It
provides a unified scheme for treatment of ZPE, tunneling The evolution of an observable A is given by its quantum-
and other quantum effects. The method is motivated by the classical bracket with the Hamiltonian H
ideas of mixed quantum-classical dynamics. In contrast to ⳵A
the existing quantum-classical schemes that couple purely ⫽ 关关 A,H 兴兴 . 共3兲
⳵t
classical and purely quantum mechanical degrees of free-
dom, the new approach considers coupling of quantum and Detailed and rigorous definitions of expressions 共1兲–共3兲 fol-
classical mechanics for the same degree of freedom. The low from the mixed quantum-classical representation of the
method gives Hamilton equations of motion 共EOM兲 for p product of two Heisenberg groups as discussed in Ref. 1.
and q that are coupled to EOMs for other dynamical vari-
ables arising from quantum mechanics. The Hamilton dy-
namics is quantized by coupling to the quantum variables. B. Quantum-classical mixing for the same degree of
Hence, the method is called quantized Hamilton dynamics freedom
共QHD兲. The theory section presents our motivation for QHD
and develops a QHD algorithm. The method is then applied We would like to use this approach in order to combine
to one and two-dimensional models of tunneling and ZPE the exact classical and approximate quantum solutions for
effects. The concluding section summarizes the features of the same particle. We focus on the EOM first. Since a low
the method and discusses further avenues for consideration order potential can provide a good approximation for the
and improvement. ZPE and tunneling effects, we Taylor expand the potential
energy. The quantum Hamiltonian H 0 will include only the
lower order terms in the expansion

II. THEORY H 共 P̂,Q̂ 兲 ⫽H 0 共 P̂,Q̂ 兲 ⫹V 共 Q̂ 兲 ,


共4兲
Consider a time-dependent problem that can be solved P̂ 2
exactly in classical mechanics, either analytically or numeri- H 0 共 P̂,Q̂ 兲 ⫽ ⫹V 0 共 Q̂ 兲 .
2m
cally, but can not be solved in quantum mechanics. Suppose
that an approximation to the full problem exist that admits a The exact time evolution of the expectation value of a quan-
quantum mechanical solution. For instance, ZPE in a highly tum observable A( P̂,Q̂) is given in the Heisenberg represen-
anharmonic dissociation of a diatomic is well described in tation by
the harmonic approximation. Tunneling of proton from one
⳵ 具 A 共 P̂,Q̂ 兲 典 i
water molecule to the next is well described by the instanton ⫽ 具 关 A 共 P̂,Q̂ 兲 ,H 0 共 P̂,Q̂ 兲 ⫹V 共 Q̂ 兲兴 典 , 共5兲
model56,57 that is based on the inverted harmonic potential. ⳵t ប
The following question arises in both cases: Is it possible to where the angular brackets denote quantum mechanical av-
combine the exact classical and approximate quantum solu- eraging, e.g., 具 A( P̂,Q̂) 典 ⬅ 具 ⌿ 兩 A( P̂,Q̂) 兩 ⌿ 典 . The wave func-
tions? tion 兩 ⌿ 典 is time independent in the Heisenberg picture and is
determined by the initial conditions.
A. Quantum-classical mixing for separate degrees of The quantum-classical approximation is achieved by ap-
freedom
proximation of the quantum commutator between A and the
In situations where some degrees of freedom (p,q) ad- higher order potential energy terms with the Poisson bracket.
mit the classical limit and the remaining degrees ( P,Q) can The kinetic and low order potential energy terms are treated
be treated quantum mechanically, the quantum-classical mix- in the usual quantum mechanical way. The proposed
ing is given by the quantum-classical Lie bracket1,29,30 quantum-classical EOM is
J. Chem. Phys., Vol. 113, No. 16, 22 October 2000 Quantized Hamilton dynamics 6559

⳵ 具 A 共 P̂,Q̂ 典 兲 i Step 2. Quantum-classical Hamiltonian. The total


⫽ 具 关 A 共 P̂,Q̂ 兲 ,H 0 共 P̂,Q̂ 兲兴 典 Hamiltonian is split into the quantum H 0 and classical V
⳵t ប
parts as in Eq. 共4兲, for example, by Taylor expansion of the
⫹ 兵 A 共 p,q 兲 ,V 共 q 兲 其 potential energy.

where p⬅ 具 ⌿ 兩 P̂ 兩 ⌿ 典 , q⬅ 具 ⌿ 兩 Q̂ 兩 ⌿ 典 . P̂ 2
H 共 P̂,Q̂ 兲 ⫽ ⫹V 0 共 Q̂ 兲 ⫹V 共 具 Q̂ 典 兲 ,
2m
共6兲
共10兲
It remains to be seen whether the right-hand side of Eq. 共6兲 is 具 Q̂ 典 ⬅ 具 ⌿ 兩 Q̂ 兩 ⌿ 典 .
a true Lie bracket that obeys the same properties as the clas- The kinetic and low order potential V 0 energy term remains
sical Poisson bracket and quantum commutator. A group the- an operator valued function, while the higher order potential
oretical analysis similar to that of Ref. 1 for the quantum- energy term V is a number valued function of the average
classical bracket 共1兲 may provide a rigorous definition of Eq. quantum position.
共6兲. As suggested by Kisil58 this could be achieved by the G 2 Step 3. Equation of motion. Each dynamical variable
Lie group with five elements 兵 P,Q,q,iប,iប̃ 其 satisfying the 具 P̂ Q̂ 典 s from the set 共9兲 evolves according to the quantum-
m n

following commutation relationships: 关 P,Q 兴 ⫽iប, 关 P,q 兴 classical equation of motion 共6兲
⫽iប̃, all other commutators being equal zero.59 The limit ⳵ 具 P̂ m Q̂ n 典 s i
ប̃→0 could provide the desired quantum-classical represen- ⫽ 具 关共 P̂ m Q̂ n 兲 s ,H 0 共 P̂,Q̂ 兲兴 典
⳵t ប
tation if the total Hamiltonian 共4兲 is viewed as a function of
P, Q, and q: H( P̂,Q̂,q)⫽H 0 ( P̂,Q̂)⫹V(q). The issue de- ⫹ 兵 具 P̂ 典 m 具 Q̂ 典 n ,V 共 具 Q̂ 典 兲 其 . 共11兲
serves a separate analysis. We note that if the classical limit The quantum mechanical average of the commutator pro-
as in Eq. 共6兲 is taken in Eq. 共5兲 for both H 0 and V, the duces a new dynamical variable which may or may not be-
classical EOMs are obtained long to the set 共9兲. If the new variable does not belong to the
dp ⳵ 共 V 0 共 q 兲 ⫹V 共 q 兲兲 dq set 共9兲, the closure procedure is applied.
⫽⫺ , ⫽p. 共7兲 Step 4. Closure scheme. For nontrivial H 0 , the commu-
dt ⳵q dt
tator in Eq. 共11兲 increases the power of ( P̂ m Q̂ n ) s introducing
We consider EOM 共6兲 for a set of the quantum mechani-
higher order dynamical variables 具 P̂ m Q̂ n 典 s . For example, a
cal average values of momentum, position operators and
cubic term in the potential energy couples the linear variable
their products 具 P̂ 典 , 具 Q̂ 典 , 具 P̂ 2 典 , 具 Q̂ 2 典 , 具 P̂Q̂ 典 ,... . The EOMs for 具 P̂ 典 to the second order variable 具 Q̂ 2 典 . At some point, a
the average values are coupled and, in general, form an infi-
newly generated variable 具 P̂ m Q̂ n 典 s will be outside the set of
nite hierarchy. The hierarchy is terminated by the approxi-
the selected dynamical variables, Eq. 共9兲. The extra variable
mation of the higher order averages via products of the lower
order averages. The closure scheme is presented below and
具 P̂ m Q̂ n 典 s is approximately decomposed into a product of the
lower order variables from the set 共9兲, terminating the EOM
leads to the QHD algorithm.
hierarchy. The closure procedure is similar to that in
statistical61–63 and quantum64,65 many-body theory. The av-
C. The QHD algorithm erage 具 P̂ m Q̂ n 典 s is decomposed into the irreducible parts and
the highest (m⫹n)-order irreducible term is represented by
We work in the Heisenberg representation, where opera- the sum of all possible products of the lower order irreduc-
tors evolve in time, while the time-independent wave func- ible terms. Decompositions for the third and fourth order
tion 兩 ⌿ 典 specifies initial conditions. terms are given by all possible pairings of the individual
Step 1. Variables. A set A of quantum mechanical ob- operators, as in Wick’s theorem.64 Thus, the following clo-
servables that are functions of the position and momentum sure is obtained for a general third order term
operators is defined
具 ABC 典 ⫽ 具 共 A⫺ 具 A 典 ⫹ 具 A 典 兲共 B⫺ 具 B 典 ⫹ 具 B 典 兲共 C⫺ 具 C 典 ⫹ 具 C 典 兲 典
A⬅ 兵 P̂, Q̂, P̂ 2 , Q̂ 2 , 共 P̂Q̂ 兲 s , . . . 其 . 共8兲
⫽ 具 共 A⫺ 具 A 典 兲共 B⫺ 具 B 典 兲共 C⫺ 具 C 典 兲 典 ⫹ 具 AB 典具 C 典
The cross terms are Weyl symmetrized,1,60 e.g.,
⫹ 具 AC 典具 B 典 ⫹ 具 BC 典具 A 典 ⫺2 具 A 典具 B 典具 C 典
P̂Q̂⫹Q̂ P̂
共 P̂Q̂ 兲 s ⬅ . ⬇ 具 AB 典具 C 典 ⫹ 具 AC 典具 B 典 ⫹ 具 BC 典具 A 典 ⫺2 具 A 典具 B 典具 C 典 ,
2
共12兲
The dynamical variables of the QHD method are defined by
the quantum mechanical expectation values of the observ- since all the pairings
ables in the set taken with respect to the time-independent 具 共 A⫺ 具 A 典 兲共 B⫺ 具 B 典 兲共 C⫺ 具 C 典 兲 典
wave function
⬇ 具 共 A⫺ 具 A 典 兲共 B⫺ 具 B 典 兲 典具 共 C⫺ 具 C 典 兲 典
具 A 典 ⬅ 兵 具 P̂ 典 , 具 Q̂ 典 , 具 P̂ 2 典 , 具 Q̂ 2 典 , 具 P̂Q̂ 典 s , . . . 其 ⫹ 具 共 A⫺ 具 A 典 兲共 C⫺ 具 C 典 兲 典具 共 B⫺ 具 B 典 兲 典
with 具 P̂ 典 ⫽ 具 ⌿ 兩 P̂ 兩 ⌿ 典 , etc. 共9兲 ⫹ 具 共 B⫺ 具 B 典 兲共 C⫺ 具 C 典 兲 典具 共 A⫺ 具 A 典 兲 典 ⫽0 共13兲
The set must contain the first order terms 具 P̂ 典 and 具 Q̂ 典 . vanish. Fourth order terms are decomposed into
6560 J. Chem. Phys., Vol. 113, No. 16, 22 October 2000 O. V. Prezhdo and Y. V. Pereverzev

具 ABCD 典 ⫽ 具 共 A⫺ 具 A 典 兲共 B⫺ 具 B 典 兲共 C⫺ 具 C 典 兲共 D⫺ 具 D 典 兲 典
⫹ 具 ABC 典具 D 典 ⫹ 具 ABD 典具 C 典 ⫹ 具 ACD 典具 B 典
⫹ 具 BCD 典具 A 典 ⫺ 具 AB 典具 C 典具 D 典 ⫺ 具 AC 典具 B 典具 D 典
⫺ 具 BC 典具 A 典具 D 典 ⫺ 具 AD 典具 B 典具 C 典 ⫺ 具 BD 典具 A 典具 C 典
⫺ 具 CD 典具 A 典具 B 典 ⫹3 具 A 典具 B 典具 C 典具 D 典 . 共14兲
The pairing procedure gives
具 共 A⫺ 具 A 典 兲共 B⫺ 具 B 典 兲共 C⫺ 具 C 典 兲共 D⫺ 具 D 典 兲 典
⬇ 具 共 A⫺ 具 A 典 兲共 B⫺ 具 B 典 兲 典具 共 C⫺ 具 C 典 兲共 D⫺ 具 D 典 兲 典
⫹ 具 共 A⫺ 具 A 典 兲共 C⫺ 具 C 典 兲 典具 共 B⫺ 具 B 典 兲共 D⫺ 具 D 典 兲 典 FIG. 1. Potential energy as a function of particle position 具 Q̂ 典 in the tun-
neling example, Eq. 共18兲. The barrier maximum is located at q⫽⫺3.33 and
⫹ 具 共 A⫺ 具 A 典 兲共 D⫺ 具 D 典 兲 典具 共 B⫺ 具 B 典 兲共 C⫺ 具 C 典 兲 典 . 共15兲
has the height of 1.85.
The 3⫹1 decompositions such as 具 (A⫺ 具 A 典 )(B⫺ 具 B 典 )
⫻(C⫺ 具 C 典 ) 典具 (D⫺ 具 D 典 ) 典 vanish, and the final closure for-
mula for the fourth order term reads
QHD simulations correspond to the Gaussian wave packet
具 ABCD 典 ⬇ 具 ABC 典具 D 典 ⫹ 具 ABD 典具 C 典 ⫹ 具 ACD 典具 B 典 that has the following form in the position representation:
⫹ 具 BCD 典具 A 典 ⫹ 具 AB 典具 CD 典 ⫹ 具 AC 典具 BD 典
⫹ 具 AD 典具 BC 典 ⫺2 具 AB 典具 C 典具 D 典 ⫺2 具 AC 典具 B 典具 D 典
⌿共 q 兲⫽ 冉 冊 冋
m␻
␲ប
1/4
exp ⫺
m ␻ 共 q⫺q 0 兲 2
2ប
⫹i
p 0q
ប 册 for t⫽0,
共17兲
⫺2 具 BC 典具 A 典具 D 典 ⫺2 具 AD 典具 B 典具 C 典 where q 0 is the center of the wave packet and p 0 is its initial
⫺2 具 BD 典具 A 典具 C 典 ⫺2 具 CD 典具 A 典具 B 典 momentum. The units in both examples are chosen to have
the Planck constant and particle mass equal one: ប⫽m⫽1.
⫹6 具 A 典具 B 典具 C 典具 D 典 . 共16兲
Higher order terms can be decomposed in a similar
manner.62 A. Example 1: Tunneling escape from a metastable
It is straightforward to show based on the Ehrenfest state
theorem31 that when the set of variables 共8兲 is restricted to
The first applications of QHD models decay of a meta-
contain only the momentum and position operators A
stable state.44,46 At low energies the decay is due to tunneling
⫽ 兵 P̂,Q̂ 其 , the QHD algorithm gives the Hamilton EOMs of through the potential barrier. Examples of such processes are
classical mechanics for any quantum-classical Hamiltonian numerous in chemistry and range from dissociation of elec-
共4兲. The Poisson bracket in the quantum-classical EOM 共6兲, tronically excited states of small molecules in gas phase9 to
共11兲 is essential for the classical limit in the presence of the biochemical processes such as enzyme mediated proton or
high order terms V in the Hamiltonian 共4兲, 共10兲. proton-coupled electron transfer10,40 and isomerization
reactions.67 Dynamical tunneling68 can provide important
III. APPLICATIONS contributions to the decay process even at high energies in
The QHD method is applied in present to two model competition with classically accessible pathways. The one
systems that exhibit tunneling and ZPE phenomena. The dimensional Hamiltonian of the metastable state is defined
Hamiltonians in both models are anharmonic only to low by
order, so that the quantum Hamiltonian H 0 is the full Hamil- p2 q2
tonian H, Eqs. 共4兲 and 共10兲. The QHD calculations are com- H 共 p,q 兲 ⫽ ⫹ ⫹aq 3 , a⫽0.1. 共18兲
2 2
pared with the classical mechanics results as well as with the
fully quantum mechanical calculations. The classical and It describes an oscillator with a cubic anharmonicity. The
QHD simulations are performed numerically using the fourth potential energy function is plotted in Fig. 1.
order Runge–Kutta integrator. The classical and QHD calcu- The QHD approach is applied to second order. The QHD
lations take about the same amount of time and are several set of variables includes momentum and position as in clas-
orders of magnitude faster than the quantum mechanical cal- sical mechanics, together with the squares of momentum,
culation. The quantum mechanical wave packets are propa- position and the symmetrized cross term. The set contains
gated by the split-operator method66 using fast Fourier trans- five variables:
form to switch between position and momentum
representations. The calculation results are tested for conver-
具 A 典 ⬅ 兵 具 P̂ 典 , 具 Q̂ 典 , 具 P̂ 2 典 , 具 Q̂ 2 典 , 具 P̂Q̂ 典 s 其 , 共19兲
gence with respect to the integration time step and, in the compared to two variables in classical mechanics. Closure of
quantum mechanical case, with respect to the grid size and the third and higher order terms results in the following
spacing. The initial conditions of quantum, classical and EOM:
J. Chem. Phys., Vol. 113, No. 16, 22 October 2000 Quantized Hamilton dynamics 6561

FIG. 2. Trajectory of the metastable particle in the tunneling example. FIG. 3. Phase space evolution of the metastable particle in the tunneling
共a兲 具 Q̂ 典 (t⫽0)⫽1, 具 P̂ 典 (t⫽0)⫽⫺1, 具 E 典 ⫽1.75; 共b兲 具 Q̂ 典 (t⫽0)⫽⫺1, example. The notation is the same as in Fig. 2.
具 P̂ 典 (t⫽0)⫽1, 具 E 典 ⫽1.25. The quantum, classical and QHD trajectories are
given by thick solid, thin solid, and dashed lines, respectively.

bracket of Step 3 of the QHD algorithm and the closure


d 具 Q̂ 典 procedure of Step 4 are unique to QHD and make it directly
⫽ 具 P̂ 典 , 共20兲
dt extendible to higher orders. The dynamics of the QHD
EOMs obtained to any order has another unique property due
d 具 P̂ 典 to the closure procedure of Step 4 of the QHD algorithm. In
⫽⫺ 具 Q̂ 典 ⫺3a 具 Q̂ 2 典 , 共21兲
dt the classical limit ប⫽0 with classical initial conditions for
the QHD variables, i.e., 具 Q̂ 2 典 ⫽ 具 Q̂ 典具 Q̂ 典 , etc., the QHD dy-
d 具 Q̂ 2 典 namics for the extended set of variables is identical to the
⫽2 具 P̂Q̂ 典 s , 共22兲
dt corresponding Hamilton dynamics for these variables. To-
gether with the energy conservation property discussed be-
d 具 P̂ 2 典 low, this property provides a useful test of QHD codes.
⫽⫺2 具 P̂Q̂ 典 s ⫺6a 共 2 具 P̂Q̂ 典 s 具 Q̂ 典 ⫹ 具 P̂ 典具 Q̂ 2 典
dt As follows from the Hamiltonian 共18兲, the total energy
of the metastable system is given in the second order QHD
⫺2 具 P̂ 典具 Q̂ 典 2 兲 , 共23兲 approach by
d 具 P̂Q̂ 典 s
⫽ 具 P̂ 2 典 ⫺ 具 Q̂ 2 典 ⫺3a 共 3 具 Q̂ 2 典具 Q̂ 典 ⫺2 具 Q̂ 典 3 兲 . 共24兲
dt 具 P̂ 2 典 具 Q̂ 2 典
具E典⫽ ⫹ ⫹a 共 3 具 Q̂ 2 典具 Q̂ 典 ⫺2 具 Q̂ 典 3 兲 , 共25兲
2 2
The derivation is described in the Appendix.
As pointed out in the review process, the second order where the anharmonic term is expressed using the available
QHD EOMs are reminiscent of the equations obtained by variables 共19兲 according to Eq. 共12兲. The total energy is con-
Heller19 and Mukamel.18 Both Heller’s and Mukamel’s ap- served during the evolution by Eqs. 共20兲–共24兲. The initial
proaches are based on the Gaussian ansatz for the quantum- quantum mechanical wave function is a Gaussian wave
mechanical wave function and Wigner distribution function packet, Eq. 共17兲. The center p 0 , q 0 of the wave packet de-
and produce dynamics that is different from QHD. For in- fines the initial conditions of the classical simulation. The
stance, the 具 P̂ 典 and 具 Q̂ 典 variables in Ref. 19 evolve purely initial conditions in QHD are the quantum mechanical aver-
classically and, therefore, do not tunnel. QHD does not re- ages from the set 共19兲 taken with respect to the the Gaussian
quire the Gaussian ansatz. The mixed quantum-classical wave packet 共17兲,
6562 J. Chem. Phys., Vol. 113, No. 16, 22 October 2000 O. V. Prezhdo and Y. V. Pereverzev

具 Q̂ 典 t⫽0 ⫽q 0 , 具 P̂ 典 t⫽0 ⫽p 0 ,

具 Q̂ 2 典 t⫽0 ⫽q 20 ⫹ ,
2m ␻
共26兲
mប ␻
具 P̂ 2 典 t⫽0 ⫽p 20 ⫹ ,
2

具 P̂Q̂ 典 s,t⫽0 ⫽p 0 q 0
with ប⫽m⫽ ␻ ⫽1.
The results of the quantum-mechanical, classical and
QHD simulations are presented in Figs. 2 and 3. The evolu-
tion of classical position and momentum is compared with
the evolution of the corresponding QHD variables and the
average values from the quantum simulations. Figure 2
shows time dependence of coordinate. Figure 3 shows mo-
mentum vs position in the phase space. Two sets of initial
conditions are considered. The trajectories in Figs. 2共a兲 and
3共a兲 originate on the right slope of the pseudoharmonic re-
gion, Fig. 1, with momentum to the left. The trajectories in
Figs. 2共b兲 and 3共b兲 originate on the left slope with momen-
tum to the right. Quantum mechanically, any state localized
inside the pseudoharmonic region of the potential is meta-
stable and will decay in time. Classically, only particles with
the total energy higher than the energy of the barrier can
escape. In both 共a兲 and 共b兲 cases the classical trajectories are FIG. 4. Deep tunneling QHD evolution of the metastable particle. The ini-
periodic and are not able to reproduce the quantum mechani- tial conditions are 具 Q̂ 典 (t⫽0)⫽0, 具 P̂ 典 (t⫽0)⫽1.05 and the total QHD en-
cal decay of the metastable state. Second order QHD cor- ergy is 具 E 典 ⫽1.051 25.
rectly describes the tunneling escape in both cases. The QHD
trajectory starting at q 0 ⫽1 is higher in energy and more
strong intermode coupling usually leads to faster reactions.
closely agrees with the quantum mechanical data than the
Efficient energy exchange between modes is the necessary
trajectory starting at q 0 ⫽⫺1.
condition for applicability of transition state theories. In pho-
Similar to classical mechanics, QHD exhibits a threshold
tochemistry, direct excitation of the reactive mode decays by
energy below which the nature of the dynamics changes.
coupling to other modes and the reaction is slowed down.
Formation of the cusp in Fig. 3共b兲 indicates that the q 0
Energy exchange is significantly influenced by ZPE which
⫽⫺1 trajectory is getting close to the threshold. The thresh-
comprises a large fraction of energy stored in high frequency
old energy is notably less in second order QHD than in clas-
modes. At room temperature, high frequency modes involve
sical mechanics and can be decreased by extension of the
not only the low mass C–H and O–H vibrations, but also
closure procedure to higher orders. In contrast to the classical
C–C, C–O, C–N and other bonds.
motion which for the present model is periodic below the
The Hamiltonian of the ZPE model is a simplified ver-
threshold energy, the QHD dynamics becomes chaotic below
sion of the well-known Henon–Heiles Hamiltonian44,69,70
the corresponding threshold.69 An example of a chaotic tra-
jectory is given in Fig. 4. Here, tunneling is deep and takes p 21 ␻ 21 q 21 p 22 ␻ 22 q 22
place after many oscillations inside the well. The trajectory H 共 p 1 ,q 1 ,p 2 ,q 2 兲 ⫽ ⫹ ⫹ ⫹ ⫹␭q 1 q 22
2 2 2 2
suggests that deep quantum mechanical tunneling is repre- 共27兲
sented in QHD by a chaotic motion in the extended phase
space. The QHD trajectory explores the extended phase ␻ 1 ⫽1.0, ␻ 2 ⫽1.1, ␭⫽⫺0.11.
space, finds the tunneling bottleneck and escapes. The parameters of the model are the same as in Ref. 44.
Similar to the previous example, QHD is developed to sec-
ond order in each degree of freedom. In addition to the mo-
B. Example 2: ZPE and energy exchange between
two oscillators menta and coordinates of the two modes 具 P̂ 1 典 , 具 Q̂ 1 典 , and
具 P̂ 2 典 , 具 Q̂ 2 典 that are present in classical mechanics, six new
The second model system illustrates the effect of zero variables are added. These are the squares of position and
point motion in energy exchange between two harmonic os-
momentum 具 P̂ 21 典 , 具 Q̂ 21 典 , 具 P̂ 22 典 , 具 Q̂ 22 典 and the symmetrized
cillators coupled by an anharmonic term. Energy exchange
has wide implications in chemistry. Chemical reaction rates position-momentum correlations 具 P̂ 1 Q̂ 1 典 s ⫽ 具 ( P̂ 1 Q̂ 1
are determined by the amount of energy stored in the reac- ⫹Q̂ 1 P̂ 1 )/2典 , 具 P̂ 2 Q̂ 2 典 s ⫽ 具 ( P̂ 2 Q̂ 2 ⫹Q̂ 2 P̂ 2 )/2典 . The cross-
tive mode. In thermally activated reactions where the energy terms between the modes such as P̂ 1 P̂ 2 , P̂ 1 Q̂ 2 , etc., are
reaches the reactive mode through coupling to a heat bath, neglected. If included, the cross terms would dominate QHD
J. Chem. Phys., Vol. 113, No. 16, 22 October 2000 Quantized Hamilton dynamics 6563

variable sets in large systems. Decomposition of such terms


into products of first order classical terms significantly im-
proves dimensionality scaling of the method. The total of 10
QHD variables in the present calculation should be com-
pared with the total of four classical variables.
The QHD EOMs are derived using the closure procedure
described earlier. The cubic terms are decomposed using Eq.
共12兲. The cross-terms of the type 具 P̂ 1 P̂ 2 典 are approximated
by the products 具 P̂ 1 典具 P̂ 2 典 . The final result is

d 具 Q̂ 1 典
⫽ 具 P̂ 1 典 , 共28兲
dt
FIG. 5. Harmonic mode energies 共40兲 as functions of time in the ZPE
d 具 Q̂ 2 典 example. Thick, thin solid, and dashed lines depict the energies calculated
⫽ 具 P̂ 2 典 , 共29兲 quantum mechanically, classically, and by QHD, respectively.
dt

d 具 P̂ 1 典
⫽⫺ ␻ 21 具 Q̂ 1 典 ⫺␭ 具 Q̂ 22 典 , 共30兲 where i⫽1,2, and n i ⫽0 refers to the initial quantum state of
dt
the ith oscillator. The values of the angle variables ␾ i are
d 具 P̂ 2 典 selected randomly in the interval 共0,2␲兲.
⫽⫺ ␻ 22 具 Q̂ 2 典 ⫺2␭ 具 Q̂ 1 典具 Q̂ 2 典 , 共31兲 The dynamical quantities of interest are the the harmonic
dt mode energies defined by
d 具 Q̂ 21 典 具 P̂ 2i 典
␻ 2i 具 Q̂ 2i 典
⫽2 具 P̂ 1 Q̂ 1 典 s , 共32兲 具 E i典 ⫽ ⫹ , i⫽1,2. 共40兲
dt 2 2

d 具 Q̂ 22 典 The quantum mechanical expectation values are computed


⫽2 具 P̂ 2 Q̂ 2 典 s , 共33兲 with respect to the evolving wave packet. The harmonic
dt
mode energies in the QHD approach are given by the 具 P̂ 2i 典
d 具 P̂ 21 典 and 具 Q̂ 2i 典 variables. Classically, the mode energies are cal-
⫽⫺2 ␻ 21 具 P̂ 1 Q̂ 1 典 s ⫺2␭ 具 Q̂ 1 典具 Q̂ 22 典 , 共34兲 culated by averaging over an ensemble of 1500 trajectories
dt
initiated using different random numbers ␾ i , Eq. 共39兲. The
d 具 P̂ 22 典 results are presented in Fig. 5. A substantial disagreement is
⫽⫺2 ␻ 22 具 P̂ 2 Q̂ 2 典 s ⫺4␭ 具 Q̂ 1 典具 P̂ 2 Q̂ 2 典 s , 共35兲 observed between the quantum and classical results due to
dt
the ZPE effect. Classically, the energy is free to flow from
d 具 P̂ 1 Q̂ 1 典 s one mode to the other. At long times the energy can concen-
⫽ 具 P̂ 21 典 ⫺ ␻ 21 具 Q̂ 21 典 ⫺␭ 具 Q̂ 1 典具 Q̂ 22 典 , 共36兲 trate in a single mode. The classical energy flow is more
dt chaotic and with a larger amplitude in a single trajectory and
is smoothed in Fig. 5 by the ensemble averaging. Quantum
d 具 P̂ 2 Q̂ 2 典 s
⫽ 具 P̂ 22 典 ⫺ ␻ 22 具 Q̂ 22 典 ⫺2␭ 具 Q̂ 1 典具 Q̂ 22 典 . 共37兲 mechanically, each mode has to contain ZPE. Consequently,
dt the quantum mechanical energy exchange is significantly
The second order QHD expression for the total energy is weaker compared to the flow of the classical energy. QHD
obtained by decomposing the cubic term in the Hamiltonian reproduces the quantum mechanical behavior. The harmonic
共27兲 mode energies do not decrease below ZPE. The QHD com-
putational effort is roughly 2.5 of the computational effort
具 P̂ 21 典 ␻ 21 具 Q̂ 21 典 具 P̂ 22 典 ␻ 22 具 Q̂ 22 典 for a single classical trajectory. This is at least two orders of
具E典⫽ ⫹ ⫹ ⫹ ⫹␭ 具 Q̂ 1 典具 Q̂ 22 典 . magnitude better than for the ensemble of classical trajecto-
2 2 2 2
共38兲 ries and about four orders better than for the full quantum
simulation.
The initial conditions represent the product of the ground
state wave functions of each harmonic mode. The ground
state quantum wave packet of the harmonic oscillator is IV. CONCLUSIONS
given by Eq. 共17兲 with q 0 ⫽p 0 ⫽0. The QHD initial condi- The QHD formalism presented in the paper establishes a
tions are given by Eqs. 共26兲 for each mode. The initial con- natural extension of classical Hamilton dynamics. It provides
ditions for the classical trajectories are selected in the usual efficient means to incorporate quantum effects such as tun-
way44 to correspond to the harmonic potential expressed in neling and ZPE into classical calculations. The variables rep-
the action-angle variables resenting the quantum effects are treated on equal footing
q i,t⫽0 ⫽ 冑共 2n i ⫹1 兲 ប/ ␻ i sin ␾ i , with the classical position and momentum variables. In the
共39兲 lowest order, QHD is equivalent to classical mechanics.
p i,t⫽0 ⫽ 冑共 2n i ⫹1 兲 ប ␻ i cos ␾ i , QHD approaches the exact quantum limit when the QHD
6564 J. Chem. Phys., Vol. 113, No. 16, 22 October 2000 O. V. Prezhdo and Y. V. Pereverzev

hierarchy of equations is extended to higher orders. Due to which is the same as Eq. 共24兲. The EOM of the new dynami-
flexibility of the QHD formalism, quantum effects can be cal variable 具 P̂Q̂ 典 s is
incorporated into a classical calculation in a systematic man-
ner and to any desired degree of accuracy. As exemplified by d 具 P̂Q̂ 典 s
i ⫽ 具 关共 P̂Q̂⫹Q̂ P̂ 兲 /2,Ĥ 兴 典
the model applications, the lowest order QHD approximation dt
that goes beyond classical mechanics already gives a dra-
matic improvement over the purely classical theory. The ⫽⫺i 具 Q̂ 2 典 ⫺i3a 具 Q̂ 3 典 ⫹i 具 P̂ 2 典 . 共A3兲
computational cost of the lowest order QHD is extremely It spawns two more variables 具 P̂ 典 and 具 Q̂ 典 . P̂ belongs to
2 3 2
modest and is only 3 times greater than that of a classical the set 共19兲 and is also treated as an independent variable
calculation. with the EOM
Due to its formal consistency and computational effi-
ciency, the QHD approach deserves further exploration. De- d 具 P̂ 2 典
composition of the Hamiltonian 共4兲 into the quantum and i ⫽ 具 关 P̂ 2 ,Ĥ 兴 典 ⫽⫺i2 具 P̂Q̂ 典 s ⫺i3a 具 P̂Q̂ 2 ⫹Q̂ 2 P̂ 典 .
dt
classical parts is not unique and can be optimized depending 共A4兲
on application. The closure procedure that terminates the hi-
erarchy of QHD EOM should be tested with more general The variables 具 Q̂ 3 典 and 具 P̂Q̂ 2 ⫹Q̂ 2 P̂ 典 , however, do not be-
potentials. Various extensions of the basic QHD algorithm long to the set 共19兲. The hierarchy of equations is terminated
are possible, for example, to systems with separate quantum at these variables, which are approximated according to Eq.
and classical modes as well as to non-adiabatic systems that 共12兲 by
evolve on several potential energy surfaces. Similar to clas- 具 Q̂ 3 典 ⫽3 具 Q̂ 2 典具 Q̂ 典 ⫺2 具 Q̂ 典 3 共A5兲
sical molecular dynamics, QHD can be employed in statisti-
60
cal mechanical studies of equilibrium and nonequilibrium and, with proper symmetrization,
properties of matter. Statistical mechanical applications of
具 P̂Q̂ 2 ⫹Q̂ 2 P̂ 典 ⫽4 具 Q̂ 典具 P̂Q̂ 典 s ⫺4 具 P̂ 典具 Q̂ 典 2 ⫹2 具 P̂ 典具 Q̂ 2 典 .
QHD will require coupling of the deterministic quantized 共A6兲
dynamics with a heat or pressure bath. Last, but not the least,
the formal group theoretical foundations of QHD should be Equations 共A5兲 and 共A6兲 combined with Eqs. 共A3兲 and 共A4兲
investigated. As it presently stands, QHD is a consistent dy- give Eqs. 共23兲 and 共24兲.
namical theory of quantum effects and is close to classical
Hamilton mechanics both in formal spirit and computational 1
O. V. Prezhdo and V. V. Kisil, Phys. Rev. A 56, 162 共1997兲.
efficiency. 2
O. V. Prezhdo and P. J. Rossky, J. Chem. Phys. 107, 825 共1997兲.
3
O. V. Prezhdo, J. Chem. Phys. 111, 8366 共1999兲.
4
A. D. McLachlan, Mol. Phys. 8, 39 共1964兲.
ACKNOWLEDGMENTS 5
P. Pechukas, Phys. Rev. 181, 174 共1969兲.
6
F. A. Webster et al., Phys. Rev. Lett. 66, 3172 共1991兲.
7
O.V.P. is grateful to Bill Reinhardt, Bill Miller, and Eric B. J. Schwartz, E. R. Bittner, O. V. Prezhdo, and P. J. Rossky, J. Chem.
Bittner for comments on the paper and acknowledges finan- Phys. 104, 5942 共1996兲.
8
O. V. Prezhdo and P. J. Rossky, J. Chem. Phys. 107, 5863 共1997兲.
cial support of the New Faculty Award from the Camille and 9
J. C. Tully and R. K. Preston, J. Chem. Phys. 55, 562 共1971兲.
Henry Dreyfus Foundation and Award No. R10246 from the 10
S. Hammes-Schiffer and J. C. Tully, J. Chem. Phys. 101, 4657 共1994兲.
Research Corporation. 11
D. S. Sholl and J. C. Tully, J. Chem. Phys. 109, 7702 共1998兲.
12
J. C. Tully, in Classical and Quantum Dynamics in Condensed Phase
Simulations, edited by B. J. Berne, G. Ciccotti, and D. F. Coker 共World
APPENDIX: DERIVATION OF THE QHD EOMS FOR Scientific, Singapore, 1998兲, pp. 489–514.
13
THE TUNNELING EXAMPLE D. F. Coker, in Computer Simulations in Chemical Physics, edited by M.
P. Allen and D. J. Tildesley 共Kluwer Academic, The Netherlands, 1993兲,
We illustrate the derivation of equations of motion 共20兲– pp. 315–377.
14
W. H. Miller and T. F. George, J. Chem. Phys. 56, 5637 共1972兲.
共24兲. Given the Hamiltonian 共18兲, we first compute the time 15
H. D. Meyer and W. H. Miller, J. Chem. Phys. 72, 2272 共1980兲.
derivatives of the position and momentum observables 16
N. Makri and W. H. Miller, J. Chem. Phys. 87, 5781 共1987兲.
17
X. Sun and W. H. Miller, J. Chem. Phys. 106, 916 共1997兲.
18
d P̂ J. Grad, Y. Jing, A. Haque, and S. Mukamel, J. Chem. Phys. 86, 3441
i ⫽ 关 P̂,Ĥ 兴 ⫽⫺iQ̂⫺i3aQ̂ 2 , 共1987兲.
dt 19
E. J. Heller, J. Chem. Phys. 62, 1544 共1975兲.
共A1兲 20
E. J. Heller, J. Chem. Phys. 64, 63 共1976兲.
dQ̂ 21
E. J. Heller, J. Chem. Phys. 75, 2923 共1981兲.
i ⫽ 关 Q̂,H 兴 ⫽i P̂. 22
M. Braun, H. Metiu, and V. Engel, J. Chem. Phys. 108, 8983 共1998兲.
dt 23
A. Warshel and M. Karplus, Chem. Phys. Lett. 17, 7 共1972兲.
Quantum mechanical averages of Eqs. 共A1兲 produce EOM of 24
25
R. B. Gerber, V. Buch, and M. A. Ratner, J. Chem. Phys. 77, 3022 共1982兲.
the first two 共classical兲 dynamical variables, Eqs. 共20兲 and N. C. Blais, D. G. Truhlar, and C. A. Mead, J. Chem. Phys. 89, 6204
共1988兲.
共21兲. The time derivative of the momentum variable contains 26
G. D. Billing, Int. Rev. Phys. Chem. 13, 309 共1994兲.
具 Q̂ 2 典 that is treated in QHD as an independent dynamical 27
28
M. F. Herman and J. C. Arce, Chem. Phys. 183, 335 共1994兲.
variable. Its equation of motion is R. Kosloff and A. D. Hammerich, Faraday Discuss. Chem. Soc. 91, 239
共1991兲.
29
I. V. Aleksandrov, Z. Naturforsch. A 36A, 902 共1981兲.
d 具 Q̂ 2 典
i ⫽ 具 关 Q̂ 2 ,Ĥ 兴 典 ⫽i 具 P̂Q̂⫹Q̂ P̂ 典 ⫽i2 具 P̂Q̂ 典 s , 共A2兲
30
W. Boucher and J. Traschen, Phys. Rev. D 37, 3522 共1988兲.
dt 31
P. Ehrenfest, Z. Phys. 45, 455 共1927兲.
J. Chem. Phys., Vol. 113, No. 16, 22 October 2000 Quantized Hamilton dynamics 6565

32
V. S. Batista and D. F. Coker, J. Chem. Phys. 106, 6923 共1997兲. 54
S. John and J. W. Wilson, Phys. Rev. E 49, 145 共1994兲.
33
J. Faeder and R. Parson, Science 276, 1660 共1997兲. 55
V. S. Filinov, Y. V. Medvedev, and V. L. Kamskyi, Mol. Phys. 85, 711
34
J. Faeder and R. Parson, J. Chem. Phys. 108, 3909 共1998兲. 共1995兲.
35
M. Head-Gordon and J. C. Tully, J. Chem. Phys. 103, 10137 共1995兲. 56
G. A. Voth, D. Chandler, and W. H. Miller, J. Chem. Phys. 91, 7749
36
O. V. Prezhdo and P. J. Rossky, J. Phys. Chem. 100, 17094 共1996兲. 共1989兲.
37
O. V. Prezhdo and P. J. Rossky, Phys. Rev. Lett. 81, 5294 共1998兲. 57
V. A. Benderskii, D. E. Makarov, and C. A. Wight, Chemical Dynamics at
38
A. Warshel and J. K. Hwang, J. Chem. Phys. 84, 4938 共1986兲. Low Temperatures 共Wiley, New York, 1994兲, Vol. 88.
39
A. Warshel and W. W. Parson, Annu. Rev. Phys. Chem. 42, 279 共1991兲. 58
V. V. Kisil 共private communication兲.
40
J.-K. Hwang and A. Warshel, J. Am. Chem. Soc. 118, 11745 共1996兲. 59
J. Cnops and V. V. Kisil, LANL preprint archive, 1995,
41
B. A. Waite and W. H. Miller, J. Chem. Phys. 73, 3713 共1980兲. quant-ph/9502022.
42
N. Makri and W. H. Miller, J. Chem. Phys. 91, 4026 共1989兲. 60
V. I. Tatarskii, Sov. Phys. Usp. 26, 311 共1983兲.
43
W. H. Miller, W. L. Hase, and C. L. Darling, J. Chem. Phys. 91, 2859 61
N. N. Bogoliubov, The Dynamic Theory in Statistical Physics 共Hindustan,
共1989兲.
44 Delhi, India, 1965兲.
Y. Guo, D. L. Thompson, and T. D. Sewell, J. Chem. Phys. 104, 576 62
K. Kawasaki, Ann. Phys. 共N.Y.兲 61, 1 共1970兲.
共1996兲. 63
45
Y. Guo and D. L. Thompson, J. Chem. Phys. 105, 7480 共1996兲. R. L. Liboff, Kinetic Theory: Classical, Quantum and Relativistic De-
46
Y. Guo and D. L. Thompson, J. Chem. Phys. 105, 1070 共1996兲. scriptions 共Prentice-Hall, Englewood Cliffs, NJ, 1990兲.
47
M. Ben-Nun and T. J. Martinez, J. Chem. Phys. 112, 6113 共2000兲.
64
G. C. Wick, Phys. Rev. 80, 268 共1950兲.
48
M. Ben-Nun and R. D. Levine, J. Chem. Phys. 101, 8768 共1994兲.
65
S. Wilson, Electronic Correlation in Molecules 共Oxford University Press,
49
D. Huber and E. J. Heller, J. Chem. Phys. 89, 4752 共1988兲. Oxford, England, 1984兲.
50
D. Huber, S. Ling, D. Imre, and E. J. Heller, J. Chem. Phys. 90, 7317
66
M. D. Feit, J. J. A. Fleck, and A. Steiger, J. Comput. Phys. 47, 412 共1982兲.
共1989兲.
67
H. Stecher et al., Biochemistry 38, 13542 共1999兲.
51
E. J. Heller, J. Chem. Phys. 94, 2723 共1990兲.
68
E. J. Heller and M. J. Davis, J. Phys. Chem. 85, 307 共1981兲.
52
R. C. Brown and E. J. Heller, J. Chem. Phys. 75, 186 共1981兲.
69
C. Jaffe and W. P. Reinhardt, J. Chem. Phys. 77, 5191 共1982兲.
53 70
K. B. Moller, J. P. Dahl, and N. E. Henriksen, J. Phys. Chem. 98, 3272 E. J. Heller, E. B. Stechel, and M. J. Davis, J. Chem. Phys. 73, 4720
共1994兲. 共1980兲.

You might also like