THE JOURNAL OF CHEMICAL PHYSICS 124, 044108 2006

Effects of anharmonicity on nonadiabatic electron transfer: A model

Sina Yeganeha and Mark A. Ratnerb
Department of Chemistry and Center for Nanofabrication and Molecular Self Assembly, Northwestern University, Evanston, Illinois 60208-3113

Received 28 September 2005; accepted 6 December 2005; published online 27 January 2006 The effect of anharmonicity in the intramolecular modes of a model system for exothermic intramolecular nonadiabatic electron transfer is probed by examining the dependence of the transition probability on the exoergicity. The Franck-Condon factor for the Morse potential is written in terms of the Gauss hypergeometric function both for a ground initial state and for the general case, and comparisons are made between the rst-order perturbation theory results for transition probability for harmonic and Morse oscillators. These results are veried with quantum dynamical simulations using wave-packet propagations on a numerical grid. The transition-probability expression incorporating a high-frequency quantum mode and low-frequency medium mode is compared for Morse and harmonic oscillators in different temperature ranges and with various coarse-graining treatments of the delta function from the Fermi golden rule expression. We nd that signicant deviations from the harmonic approximation are expected for even moderately anharmonic quantum modes at large values of exoergicity. The addition of a second quantum mode of opposite displacement negates the anharmonic effect at small energy change, but in the inverted regime a signicantly atter dependence on exoergicity is predicted for anharmonic modes. 2006 American Institute of Physics. DOI: 10.1063/1.2162172
I. INTRODUCTION

The classical and semiclassical theories of electron transfer ET developed by Marcus and others13 have been successful in describing and predicting many aspects of the ET rate constant including its dependence on driving force the negative free-energy change, G0 Ref. 4 and on temperature. However, the neglect of quantum-mechanical tunneling causes the classical theory to disagree with experimental results in signicant ways. The Marcus expression gives a Gaussian dependence on free energy, resulting in too steep a drop-off in the inverted region. In addition, the theory predicts that the rate constant approaches zero as temperature falls to zero while experimental results have shown a nonvanishing rate constant at low temperatures.5 These discrepancies can be dealt with by treating the problem quantum mechanically, proceeding from the golden rule result of rstorder perturbation theory and in the simplest approach assuming a single coupled vibrational mode represented by displaced harmonic potentials with same frequency.6 Several authors have extended this treatment to include multiple modes of differing frequency,7,8 and the physically important case of change in both displacement and frequency has been addressed.9 Starting from the Kubo-Lax generating function for multiple harmonic oscillators with displacement but no frequency change, Jortner and co-workers1012 formulated the ET rate constant in terms of a two-mode model with a high-frequency mode, representing an average of the relevant quantum modes, and a lower-frequency solvent mode. A very important early result came in applying the single-mode limit
a b

to the experimental data of DeVault and Chance5 in chromatium, providing an explanation for the temperature dependence of the ET rate. While the harmonic approximation is generally appropriate for describing the low-frequency solvent modes,13 it is signicantly unrealistic for the quantum modes at large values of as signicantly anharmonic behavior is expected. Several authors have accounted for anharmonicity using Morse potentials14 in treatments of nonradiative decay in crystals,15 proton transfer reactions,16,17 as well as nonadiabatic ET.11,1821 In Sec. II of this paper we derive a more compact form of the transition probability for nonadiabatic ET with anharmonic quantum modes, making use of a nite sum form of the Gauss hypergeometric functions.22 In Sec. III the results of computations with this closed form are compared with the corresponding harmonic results, and the golden rule results are compared with grid-based wavepacket propagations using the methods of Kosloff and coworkers. Finally, the two-mode case of anharmonic quantum mode and a harmonic low-frequency phonon mode is considered in Sec. IV, and comparisons are made with the harmonic results from the treatment by Jortner.12 Finally, the expected effects of signicantly anharmonic modes on the ET rate constant are discussed for parameters of interest in experimental systems such as the exoergicity in the optimal regime. Regions are identied in which the anharmonic effect on the rate constant is expected to be largest.
II. GOLDEN RULE FORMULATION: MORSE POTENTIALS

Electronic mail: s-yeganeh@northwestern.edu Electronic mail: ratner@chem.northwestern.edu

The starting point for the nonadiabatic quantummechanical treatments is the Fermi golden rule result,23
124, 044108-1 2006 American Institute of Physics

0021-9606/2006/124 4 /044108/9/$23.00

Downloaded 28 May 2009 to 129.105.215.213. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/jcp/copyright.jsp

044108-2

S. Yeganeh and M. A. Ratner

J. Chem. Phys. 124, 044108 2006

WRP =

2
P R

EP ,

cally when = . The eigenfunctions of this anharmonic potential14 can be written as27
n

where WRP, the transition probability per unit time from reactants to products, is given in terms of the product density of states DOS , E P , and the matrix element over the reactant R and product P wave functions. Making the Condon approximation24 that the electronic coupling is independent of the nuclear motion, the transition rate can be written as25 WRP = 2 HRP 2
r p 2

N, ;r r0 =

M n,N Nw + w exp

N/2n

N+1 w LN2n Nw + w , n 2

where Lk is the generalized Laguerre polynomial and n

n

w = e

rr0

M n,N =
j=0

N 2n + j j!

1/2

, 6

EP ,

2 N=2 2 D 1.

where the electronic coupling matrix element HRP has been factored out and is assumed to be independent of separation, leaving the Franck-Condon FC overlap integral over reactant and product vibrational states, r and p , respectively. In this single-mode expression, energy conservation requires that the product vibrational state be given by E n p = + E nr , where E n is the energy of the nth vibrational eigenstate. The energy conservation requirement can be relaxed in the single-mode case by introducing coarse graining, assuming a vibrational coupling to environmental modes.12,26 For the multimode case, a more general form is used where the delta function allows contributions only from energyconserving transitions primed and unprimed variables refer to the nal and initial states, respectively , WRP = 2 HRP
2 k k nk=0 n =0 k
k

The parameter N is physically important because N / 2 gives the highest bound level in the Morse potential. The FC overlap integral, In,n =
n

N , ;r r0

N, ;r r0 dr,

was rst given for Morse eigenfunctions by Fraser and Jarmain.28,29 The expression simplies as N = N for our case of identical strengths and ranges, and we can write in the notation of Iachello and Ibrahim,27 In,n = M n
,N M n,N n n N/2n

e k In ,n 2 k k .

k En / kBT
k

2 1+

Nn n

Qk 3

k En En
k

m=0 m =0 m

1 m+m N n m!m ! n m
m+m

Nn nm , 8

Here the product is over the k different modes, Qk is the k partition function for the kth mode, In ,n is the FC overlap
k k integral given by n n , and En and En are the vibrak k k k tional energies. The summations over nk and nk in Eq. 3 represent the possible transitions between initial and nal vibrational levels weighted by the fractional population in each reactant initial vibrational level, as given by k eEnk/ kBT / Qk. The delta function then picks out combinations of nk and nk which result in energy-conserving transitions. As noted by Jortner,12 this delta function should be replaced by a Lorentzian to account for energy-time uncertainty broadening resulting from nite lifetimes. Kubo7 and Lax8 were able to reduce Eq. 3 to give a generating function for the FC factors for multiple harmonic oscillators. For our treatment of anharmonicity in the quantum mode, we consider the eigenfunctions of two Morse potentials centered at r0 and r0, k
k k

2 1+

Nnn +m+m

with = e r0r0 . We now proceed to simplify Eq. 8 in the lowtemperature limit where the initial state is entirely in its ground vibrational level. In keeping with our notation we set n = 0, yielding I0,n = M 0,NM n 2 1+
N/2

,N

2 1+

Nn

m =0

1 m Nn m ! n m 9

Nn +m

The binomial coefcient can be expanded in its gamma function representation, and some grouping produces I0,n = M 0,NM n
n ,N N/2

V r =D 1e

rr0 2

V r =D 1e

rr0 2

, 4

2 1+

Nn

Nn +1
m

and proceeding in the usual fashion, we take the ranges = and strengths D = D of the potentials to be the same. Unlike the harmonic case, however, in which analytic forms can be written for the FC factor of oscillators of different frequencies, the Morse result can only be calculated analyti-

m =0

2/ 1 + m!

Nn +m n m + 1 N 2n + m + 1

10

Downloaded 28 May 2009 to 129.105.215.213. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/jcp/copyright.jsp

044108-3

Nonadiabatic electron transfer

J. Chem. Phys. 124, 044108 2006

The summation can be rewritten using a nite sum form of the Gauss 2F1 hypergeometric series that holds when n is a positive integer,30
n 2F 1

taking n continuous and differentiating with respect to the energy E,35,36 E = dn = dE 2

1

DE

16

n,b;c;z =
m=0

m mz , m!

11

Thus, for a single anharmonic mode in the low-temperature regime, the transition-probability rate is given by WRP = 2 HRP 2 I0,n
2

where the Pochhammer symbol is used, a k = a + k / a .31 After some manipulations of Eq. 10 to obtain an expression in the form of Eq. 11 , the summation can be written as
n

17

m =0

2/ 1 + m! =

Nn +m n m + 1 N 2n + m + 1

III. NUMERICAL SOLUTION OF THE SCHRDINGER EQUATION

Nn n + 1 N 2n + 1
2F 1

n ,N n ;N 2n + 1;

2 1+

12

In order to investigate the results of the previous section, one-dimensional wavefunctions were propagated on a numerical Cartesian grid with equally spaced points. The methods are only outlined here as numerous reviews have been written on the subject.3739 The Schrdinger equation given by i r,t =H t

The normalization constants M 0,N and M n ,N can be simplied as well, with the former given by N 1/2, and the latter by32 Mn = Nn +1 N 2n n +1
1/2

r,t

18

has the formal solution r,t = eiHt/ r,0 , 19

,N

13

Combining terms, we write the FC overlap integral for eigenfunctions of two displaced Morses with identical strengths and ranges as I0,n =
N/2

2 1+

Nn

Nn N 2n + 1
1/2

N 2n Nn +1 n +1 N
2F 1

n ,N n ;N 2n + 1;

2 1+

14

Equation 14 was compared for a number of parameters with the numerical method of Lopez et al.,33 and the values obtained agreed. For the multimode case at sufciently low temperatures such that the quantum mode is not thermally populated, all that remains is to substitute I0,n in Eq. 3 for the appropriate mode. For the single-mode case, the simpler form of Eq. 2 can be used after nding an expression for n and a suitable form for the product DOS. From the energy eigenvalues of the Morse potential, it can be shown that for a transition from the n = 0 initial state, n is given by energy conservation as n = 1 2
2

D D

1 + , 4

15

, with the reduced mass for the oswhere = 2 D / cillator. We use the classical form of the DOS Ref. 34 by

where the Hamiltonian is written in the usual way, H = T + V. The potential operator V is local in coordinate space, and its action is given simply by V x x . The operation in the kinetic term T = 2 / 2m 2 / x2 can be calculated with the Fourier method40,41 by a discrete fast Fourier transform FFT of the wave function into momentum space followed by multiplication by K2, where Ki is the wave number in i the frequency domain, and nally an inverse FFT back into coordinate space. With the use of the FFT algorithm, the calculation of T scales efciently as O N log N , where N is the number of grid points.39 The exponential in Eq. 19 is then expanded using a global propagator through the Newtonian interpolation method,38,42 where the sampling points are taken to be the zeros of the Chebychev polynomials. The initial wave function was prepared as the ground eigenstate of the initial potential as obtained through propagation in imaginary time.43 As shown in Fig. 1, two states, harmonic or Morse, were used and the dependence of the transition-probability rate on the exoergicity was probed by lowering the nal potential in small increments of . The population in the initial state was monitored over the course of the propagation and then t to an exponential, i 2 t = Aect, where c was taken to be the transition-probability rate. Since the rate expressions arise from the rst-order perturbation theory result, the population at short times was used. By applying the single-mode Morse treatment from Sec. II and the analogous result for harmonic oscillators,44,45 the perturbation theory results are compared with the numerical propagations Fig. 2 . The numerical propagation predicts a smaller transition probability than the golden rule result at small values of . This numerical calculation is somewhat articial since a dephasing process is implicit in

Downloaded 28 May 2009 to 129.105.215.213. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/jcp/copyright.jsp

044108-4

S. Yeganeh and M. A. Ratner

J. Chem. Phys. 124, 044108 2006

FIG. 1. Initial V and nal V potentials are shown for a harmonic and b Morse cases with parameters shown. All values are in a.u. unless otherwise stated. The electronic coupling matrix element HRP was taken as 0.02 eV. The reduced mass c was 2000 a . u. The potentials were constructed such that the frequencies, Morse 2D / c and Harmonic = kH / c, would be the same and equal to c 1870 cm1. The Morse potential has 26 bound states. The reorganization energies are H = 0.11 eV and iM = 0.18 eV. The ground-state eigenfunction for each case is also shown - - - . The exoergicity is i represented as the bottom-bottom vertical separation of the potentials.

smoothing the delta function from Eq. 3 . When becomes substantially greater than c, the Morse results from these two treatments agree as the density of states becomes large. The chosen potentials for this quantum mode correspond to a stretch with the nal state at larger displacement than the initial state. While the symmetry of the harmonic potential makes the distinction between a stretching or compression mode irrelevant in calculating the FC factor and transition probability, this distinction has an important effect for a Morse potential.11 The plot for the Morse oscillator in this case exhibits a atter dependence on exoergicity and is lower valued than in the harmonic case, as is expected from the FC factor with r0 r0. For a compression mode with r0 r0 the

opposite would be true as the plot would be more sharply peaked and higher valued than the corresponding harmonic, as seen in Fig. 3. In addition, the computed transition probability stops abruptly at D for the Morse potential, as energy conservation requires a transition from a bound to unbound state when approaches D. The expression in Eq. 14 holds only for bound states, when n N / 2 . Beyond this point, the FC factor for a bound/unbound state of the anharmonic oscillator is required.
IV. TWO-MODE TREATMENT

Although the single-mode treatment was found to be sufcient in treating some situations such as the chromatium data,12 it is often the case that the relevant modes in a system can be separated into high-frequency quantum modes and a collection of low-frequency phonon modes. The lowfrequency modes can be treated accurately as an average

FIG. 2. Golden rule results and wave-packet propagations for the transition probability as a function of for a single-quantum mode from Fig. 1. Harmonic potentials: golden rule result and wave packet ; Morse potentials: golden rule result - - - and wave packet . Propagation parameters are given: nr is the number of spatial grid points, dr is the distance between grid points, nt is the number of time steps taken for each propagation, dt is the length of a time step, and nint is the number of interpolation points taken in the Newton scheme.

FIG. 3. Transition probability for single harmonic mode , Morse stretching mode - - - with r0 r0 = 1.5, and Morse compression mode with r0 r0 = 1.5 potentials from Fig. 1 are plotted.

Downloaded 28 May 2009 to 129.105.215.213. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/jcp/copyright.jsp

044108-5

Nonadiabatic electron transfer

J. Chem. Phys. 124, 044108 2006

over many modes of similar frequencies, but incorporating the effect of many quantum modes explicitly changes the transition probability signicantly.46 We focus our attention on a system with only one relevant quantum mode with frequency c and one collective solvent phonon mode with frequency s. For harmonic oscillators in both modes, Jortner12 has identied several limiting cases in which simpler forms for the ET transition probability can be written, and we proceed analogously. Ulstrup and Jortner11 have shown that anharmonicity and frequency change can be treated by simply substituting the appropriate FC overlap integrals and energy eigenvalues into the harmonic form. a For the zero-temperature limit, when kT s 12 c,
Harmonic WRP =

HRP 2
s m=0

SmeSc c m!

p Ss m eSs , pm!

20
FIG. 4. The zero-temperature transition probability for two modes with values from Fig. 1 for harmonic and Morse oscillators in the quantum mode and with the solvent mode parameters shown. The solvent frequency s is approximately 30 cm1 and o = 0.006 eV. The results of Eq. 20 and Eq. 21 - - - as well as with a Lorentzian, Eq. 22 , for a harmonic and Morse mode are plotted. For the Lorentzian, = 1 105 hartree.

where p m is taken as m c / s and Sc and Ss are the Huang-Rhys factors for the quantum and solvent modes, respectively, given in terms of the inner- and outer-sphere reorganization energies, i and o, by Sc = i / c and Ss 2 2 2 = o/ with and s i = c c r0 r0 / 2 o = s s z0 2 z0 / 2. In the summation, the two groups of terms in parentheses are the squared FC overlap integrals for the quantum and solvent modes, respectively. We can incorporate the effect of anharmonicity in the quantum mode by substituting11 I0,m from Eq. 14 to get
Morse WRP =

HRP 2

N/2

I0,m 2

s m=0

p Ss m eSs , p m!

21

Em m + Em 0 / where p m is s and Em n is the nth Morse energy eigenvalue. This result can be compared with that obtained from the zero-temperature limit of Eq. 3 with two modes,
Morse WRP =

N/2

HRP

2 m=0

I0,m 2
n=0

n Ss eSs n! s

L Em m Em 0 + n

22

where the delta function has been replaced by the Lorentzian, x L x = / / x2 + 2 .12,47 Note that for the Morse the summation over m is only over the bound levels formally, the full spectrum of possible transitions should be represented by the sum over bound levels and an integral over the dissociative states, but the latter will typically be small in comparison and is ignored throughout. The results of both expressions are shown in Fig. 4. The substitution of Morse FC factors and eigenvalues into Eq. 20 Jortners12 Eq. 3 gives the same result as the more general form of Eq. 22 ; the similarity of the two results can be understood by rewriting the delta function, Em m Em 0 + n = 1
s s

sum over coupled values of m and p m . The rounding of p m or p m to the nearest integer gives width to the delta function,48 as does the substitution of the Lorentzian, and so the results are nearly identical. The oscillatory structure in Fig. 4 is an interesting feature that will always be present at zero temperature when there are two modes of signicantly different frequencies and 11 i o. For the harmonic oscillator, the peaks occur at integer multiples of c shifted by o. The Morse oscillator plot shows similar spacing at small values of , but for larger values of exoergicity the peaks become more closely spaced together, corresponding to the DOS approaching a continuum at the dissociative limit. In addition, as expected from the stretching nature of the quantum mode, the anharmonic system has a smaller transition rate near the optimal regime and broader dependence on the exoergicity. At higher temperatures, broadening of the peaks is expected as excited states in the solvent contribute more allowed combinations of m and n which are energy conserving. b For the low-temperature regime with s kT c, the solvent mode is treated classically, resulting in the following expression for harmonic oscillators in both modes:12

Harmonic WRP =

HRP 2
m=0

exp

m 4 okT
o

n , 23

eScSm c m!

okT

1/2

24

Em m + Em 0
s

where the delta function is nonzero only when n = p m , and so the summation over m and n is subsumed into a single

The corresponding expression for a Morse oscillator can be written once again by a simple substitution of the appropriate FC overlap integrals and energies,

Downloaded 28 May 2009 to 129.105.215.213. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/jcp/copyright.jsp

044108-6

S. Yeganeh and M. A. Ratner

J. Chem. Phys. 124, 044108 2006

Morse WRP =

HRP 2
N/2

exp
m=0

Em m + Em 0 4 okT .

I0,m 2 4

okT

1/2

25

For the Morse result with a Lorentzian and the exact form for the solvent modes, we make use of Eq. 3 :
Morse WRP =

HRP 2
N/2

I0,m 2
m=0 n=0 n =0

In,n

21

e s/ kT e sn/ kT
s

L Em m Em 0 + n n
H In,n

26

is the FC overlap integral for the harmonic oscilwhere lators in the solvent mode given by45,49 In,n
H 2 n = n ! n ! Ss n j=0 n Ss

FIG. 5. The transition probability in the low-temperature regime. The same potentials as in Fig. 4, at a temperature of 150 K. The classical approximation for the solvent modes with Eq. 24 for the harmonic quantum mode or Eq. 25 - - - for a Morse quantum mode as well as the full sum over both modes using Eq. 3 for the harmonic and Morse mode are plotted. = 2.5 105 hartree.

Ss j j! nj ! n n+j !

27

Figure 5 shows the result for the transition probability at 150 K. The quantum and solvent modes have characteristic / k, of 2700 and 43 K, respectively, so temperatures, Tc = in this regime the low-temperature approximation for quantum modes and classical approximation for the solvent modes are appropriate. Although broader than in Fig. 4, oscillations are still visible; Ulstrup and Jortner placed the con-

dition on the presence of these peaks as 2 okT c, which is satised for the parameters in Fig. 5 with the small value of o chosen. The low-temperature regime described here is valid for many temperatures of interest in biological systems. c For higher temperatures such that s kT c, the full form of the Morse FC overlap integral In,n is required as the molecular modes become thermally excited. This general equation can be obtained by similar manipulations of Eq. 8 to give

N/2n

In,n =

N 2n + 1
n m=0

2 1+
m

Nnn

N 2n N 2n

n +1 Nn+1 n+1

Nn +1

1/2

2 1+

m+1

Nnn +m 2 2F1 n,N n n + m;N 2n + 1; n m + 1 N 2n + m + 1 1+

28

Equipped with this result, transition rates at any temperature can be calculated within the classical approximation for the solvent modes. First, for comparison, the harmonic oscillator expression can be reduced to12,49

1 where = exp and Im is the modied Bessel v c / kT 1 function. We write the Morse ET transition probability,

Morse WRP = Harmonic WRP =

HRP 2 4
N/2 N/2

okT

1/2

HRP 2

exp Sc 2 + 1 v 4
cm/ 2kT

okT

exp
m=0 m =0 2e Em m / kT

Em m + Em m 4 okT

e
m=

Im 2Sc + 1 vv
c 2

Im,m 29

Qm

30

exp

m 4 okT
o

where Qm is the Morse partition function, which we take as

Downloaded 28 May 2009 to 129.105.215.213. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/jcp/copyright.jsp

044108-7
N/2 j=0 exp

Nonadiabatic electron transfer

J. Chem. Phys. 124, 044108 2006

Em j / kT . For the Lorentzian result, we use the full two-mode form of Eq. 3 ,
Morse WRP =

N/2 N/2

HRP

2 m=0 m =0 H 21

Im,m

2e

Em m / kT

Qm

In,n
n=0 n =0

e s/ kT e sn/ kT
s

L Em m Em m + n n

31

The results of these expressions at 300 K are given in Fig. 6. There is signicantly more broadening at this temperature as is expected from the 1 / T temperature dependence in the Gaussian. Examining Figs. 5 and 6, we note the following things. In Fig. 5, there is good agreement between the classical approximation for the solvent modes and the fully quantummechanical treatment of the solvent modes. In Fig. 6, there is also good agreement between the two forms; at this higher temperature, the numerical calculation of Eq. 27 for the solvent modes is avoided for large values of n and n by using the recursion relations for the harmonic oscillator FC factors.45,50
V. CONCLUSIONS

Applying the theory of nonradiative multiphonon transitions to nonadiabatic ET has yielded forms for the transition probability that incorporate anharmonicity in the molecular mode. The anharmonicity is treated by substituting the Morse FC overlap integral and energies into the analogous relations for harmonic oscillators. At low temperatures and small values of o, interesting oscillatory behavior in the transition probability as a function of exoergicity is exhibited. Regardless of whether the mode corresponds to a stretch or compression, signicant deviations are seen from the harmonic result for a relatively anharmonic mode. For increas-

ing values of , as the density of states for the Morse oscillator rapidly increases, the peaks will no longer be equally separated by approximately c but will move closer to each other. As other authors have pointed out,11,18 in a system with two anharmonic quantum modes with opposite displacements, a stretching and compression mode, the changes in the FC factors from the harmonic values will roughly cancel and anharmonic effects will disappear. However, at low temperatures and small o such that the oscillatory structure discussed in Sec. IV is present, anharmonic effects will be visible through the nonuniform spacing of peaks. We are not aware, however, of a condensed-phase experimental system which possesses these characteristics. Turning our attention to a more realistic system with = 0.16 eV at 150 K, the transition probability for a haro monic quantum mode is contrasted with stretching and compression anharmonic modes in Fig. 7. One signicant feature is in the inverted regime: the transition probability for the Morse compression mode falls off steeply as a Gaussian as would be predicted from a classical harmonic oscillator treatment , while the harmonic oscillator and Morse stretching mode are described by exponential decay. The sharp drop leads to an ET rate that is several orders of magnitude smaller than for a harmonic or stretching mode, even at = 0.5 eV where the inverted regime is just beginning. In addition, the stretching mode has a signicantly broader optimal regime than the harmonic case, and in the inverted regime this results in a rate increase of an order of magnitude relative to the harmonic case. At this larger value of o, all oscillatory structure is smoothed outfor distinct peaks a temperature of 10 K is required. For this single anharmonic mode, we see that the optimal exoergicity opt can be orcompression stretch 51 dered as: opt opt . We cannot in general predict harmonic the position of opt relative to the Morse optimal exoergicities because this relation depends on the magnitude of the horizontal displacement r0 r0 . For harmonic oscillators at high temperatures for which Jortners two-mode form converges to the Marcus expression, the optimal regime is given

FIG. 6. The transition probability with excited initial states in the quantum modes at 300 K same values as Fig. 4 . The full form of Eq. 3 using the appropriate partition functions is plotted for harmonic and Morse potentials = 1 104 hartree as well as Eq. 29 for the harmonic and Eq. 30 for the Morse - - - mode.

FIG. 7. Logarithmic plot of transition probability in the low-temperature regime at 150 K. Same as Fig. 5 with o = 0.16 eV. Using Eqs. 24 and 25 , a harmonic mode , Morse stretching mode - - - , and Morse compression mode are plotted.

Downloaded 28 May 2009 to 129.105.215.213. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/jcp/copyright.jsp

044108-8

S. Yeganeh and M. A. Ratner

J. Chem. Phys. 124, 044108 2006

FIG. 8. Transition probability with two quantum modes and one solvent mode at 150 K. Using Eqs. 24 and 25 with an additional quantum mode same D and with r0 r0 = 1.5 , harmonic and Morse - - - with o = 0.006 eV, and harmonic and Morse with o = 0.16 eV are plotted.

higher or lower exoergicities depending on the sign of R P , the shape of the relation is generally the same. Thus, displaced-distorted harmonic oscillators cannot account for the anharmonic deviations of broadening or narrowing for Morse stretching or compression modes discussed earlier. In this paper, we have written the FC factors for Morse oscillators in condensed form using the hypergeometric 2F1 function, and the expected deviations from harmonic behavior for anharmonic stretching and compression modes have been discussed. We predict that signicant differences in the electron transfer rate constant are expected, especially in the inverted regime. As seen in the logarithmic plot in Fig. 7, the Morse stretching and compression modes will differ by many orders of magnitude from the harmonic result. In addition, at large values of i the Morse compression mode is expected to have a much larger transition rate than the harmonic and Morse stretching modes. It is expected that these anharmonic effects will be signicant and experimentally observable in a properly constructed system.
ACKNOWLEDGMENT

harmonic by opt = o + H with H = kH r0 r0 2 / 2. This relationi i ship can be used to approximate the optimal regime at lower temperatures. The Morse reorganization energy is given by M Morse M r0 r0 2 , but opt o + i is a poor approxii =D 1e mation to the actual optimal region. This is due to the fact that the governing factor for the optimal region is the exoergicity at which energy conservation requires the maximum FC factor within o , and this will not necessarily peak near the activationless exoergicity given by i. While the harmonic oscillator FC factor does peak close to H, the maxii mum in the Morse FC overlap integral is not well approximated by iM . An alternative prediction of the location of the optimal region is not possible without specic parameters. We can predict that the ET transition probability at the optimal regime for the Morse compression mode will decrease much more gradually than the harmonic and the stretching mode as a function of increasing displacement r0 r0 , suggesting a possible role for anharmonic compression modes in nonadiabatic ET reactions with large inner-sphere reorganization. In Fig. 8 the transition probability for a system with two quantum modes and one solvent mode is plotted; the two Morse quantum modes are chosen to be stretching and compression modes, while the two harmonic modes are identical. The combination of compression and stretch causes the Morse and harmonic results to be nearly identical at low values of , but at higher exoergicities the Morse result is dominated by the stretching mode, and a more gradual decrease in transition rate is seen. Even at the higher o, anharmonic effects are noticeable. This effect is even more pronounced at larger values of i. It is interesting to consider the case of displaceddistorted harmonic oscillators in the quantum mode where both displacement and frequency change are considered. Using the appropriate FC factors,52 the effect of frequency change can be compared to the anharmonic effect, but while frequency change does slightly shift the W vs plot to

The authors would like to thank Abraham Nitzan for helpful discussions. This work was supported by the Chemistry Division of the NSF and the MURI/DURINT program of the DoD.
R. A. Marcus, J. Chem. Phys. 24, 966 1956 . R. A. Marcus, Discuss. Faraday Soc. 29, 21 1960 . 3 R. A. Marcus, Annu. Rev. Phys. Chem. 15, 155 1964 . 4 For the harmonic case with same frequencies the entropy change vanishes, and G = T S = . This, however, is not true for the case of anharmonic oscillators with a nonuniform density of states. We therefore consider only the transition probability dependence on exoergicity and not the free-energy change. 5 D. DeVault and B. Chance, Biophys. J. 6, 825 1966 . 6 K. Huang and R. Rhys, Proc. R. Soc. London, Ser. A 204, 406 1950 . 7 R. Kubo, Phys. Rev. 86, 929 1952 . 8 M. Lax, J. Chem. Phys. 20, 1752 1952 . 9 T. Kakitani and H. Kakitani, Biochim. Biophys. Acta 635, 498 1981 . 10 N. R. Kestner, J. Logan, and J. Jortner, J. Phys. Chem. 78, 2148 1974 . 11 J. Ulstrup and J. Jortner, J. Chem. Phys. 63, 4358 1975 . 12 J. Jortner, J. Chem. Phys. 64, 4860 1976 . 13 M. Marchi, J. N. Gehlen, D. Chandler, and M. Newton, J. Am. Chem. Soc. 115, 4178 1993 . 14 P. M. Morse, Phys. Rev. 34, 57 1929 . 15 M. D. Sturge, Phys. Rev. B 8, 6 1973 . 16 N. Brniche-Olsen and J. Ulstrup, J. Chem. Soc., Faraday Trans. 2 74, 1690 1978 . 17 N. Brniche-Olsen and J. Ulstrup, J. Chem. Soc., Faraday Trans. 1 75, 205 1979 . 18 N. C. Sndergaard, J. Ulstrup, and J. Jortner, Chem. Phys. 17, 417 1976 . 19 J. Ulstrup, Charge Transfer Processes in Condensed Media, Lecture Notes in Chemistry, Vol. 10 Springer-Verlag, Berlin, 1979 . 20 R. Islampour and S. H. Lin, Chem. Phys. Lett. 179, 147 1991 . 21 R. Islampour and S. H. Lin, J. Phys. Chem. 95, 10261 1991 . 22 The hypergeometric function appears often when dealing with matrix elements of the Morse eigenstates. See, for example, N. M. Avram and G. E. Draganescu, Int. J. Quantum Chem. 64, 655 1997 ; D. A. Morales, J. Math. Chem. 22, 255 1997 ; E. F. de Lima and J. E. M. Hornos, J. Phys. B 38, 815 2005 . 23 See, for example, C. Cohen-Tannoudji, B. Diu, and F. Lalo, Quantum Mechanics Wiley, New York, 1977 , Vol. 2, p. 1299. 24 E. Condon, Phys. Rev. 32, 858 1928 . 25 P. F. Barbara, T. J. Meyer, and M. A. Ratner, J. Phys. Chem. 100, 13148 1996 . 26 A. M. Kuznetsov and J. Ulstrup, Electron Transfer in Chemistry and
2 1

Downloaded 28 May 2009 to 129.105.215.213. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/jcp/copyright.jsp

044108-9

Nonadiabatic electron transfer

39

J. Chem. Phys. 124, 044108 2006 R. Kosloff, in Dynamics of Molecules and Chemical Reactions, edited by R. Wyatt and J. Zhang Marcel Dekker, New York, 1996 , pp. 185230. 40 M. D. Feit, J. A. Fleck, Jr., and A. Steiger, J. Comput. Phys. 47, 412 1982 . 41 D. Kosloff and R. Kosloff, J. Comput. Phys. 52, 35 1983 . 42 G. Ashkenazi, R. Kosloff, S. Ruhman, and H. Tal-Ezer, J. Chem. Phys. 103, 10005 1995 . 43 R. Kosloff and H. Tal-Ezer, Chem. Phys. Lett. 127, 223 1986 . 44 E. Hutchisson, Phys. Rev. 36, 410 1930 . 45 C. Manneback, Physica Amsterdam 17, 1001 1951 . 46 A. Sarai, Chem. Phys. Lett. 63, 360 1979 . 47 The parameter for the Lorentzian was found by taking the harmonicoscillator result at different values of and comparing with the equivalent expression from Jortner 1976 until the results were nearly identical. 48 J. J. Markham, Rev. Mod. Phys. 31, 956 1959 . 49 D. DeVault, Quantum-Mechanical Tunnelling in Biological Systems Cambridge University Press, Cambridge, 1984 . 50 J. Lerm, Chem. Phys. 145, 67 1990 . 51 This relation is true as long as r0 r0 is not too large, but for large stretch separations the relation will switch and compression opt opt at this large displacement, however, the transition rate is nearly zero for the stretch mode, and the discussion of optimal region is meaningless. 52 W. Siebrand, J. Chem. Phys. 46, 440 1967 .

Biology: An Introduction to the Theory Wiley, Chicester, UK, 1999 , Chap. 8. 27 F. Iachello and M. Ibrahim, J. Phys. Chem. A 102, 9427 1998 . 28 P. A. Fraser and W. R. Jarmain, Proc. Phys. Soc., London, Sect. A 66, 1145 1953 . 29 W. R. Jarmain and P. A. Fraser, Proc. Phys. Soc., London, Sect. A 66, 1153 1953 . 30 Handbook of Mathematical Functions with Formulas, Graphs, and Mathematical Tables, 9th ed., edited by M. Abramowitz and I. Stegun Dover, New York, 1972 , Chap. 15. 31 For negative values as we have for n m, the Pochhammer symbol is represented by a k = a + 1 / a k + 1 . 32 We assume that N is not precisely an even integer. 33 J. C. Lopez, A. L. Rivera, Y. F. Smirnov, and A. Frank, Int. J. Quantum Chem. 88, 280 2002 . 34 Formally, the density of states is given by En = En +1 En 1. To obtain a simpler form, we take the classical density of states since the results are nearly identical except very near the dissociative limit when En D. 35 P. C. Haarhoff, Mol. Phys. 7, 101 1963 . 36 J. P. K. Doye and D. J. Wales, J. Chem. Phys. 102, 9659 1995 . 37 R. Kosloff, J. Phys. Chem. 92, 2087 1988 . 38 R. Kosloff, Annu. Rev. Phys. Chem. 45, 145 1994 .

Downloaded 28 May 2009 to 129.105.215.213. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/jcp/copyright.jsp