Appl. Magn. Reson.

20, 111135 (2001)

Applied Magnetic Resonance

Springer-Verlag 2001 Printed in Austria

Spin Dynamics in Strongly Coupled Spin-Correlated Radical Pairs: Stochastic Modulation of the Exchange Interaction and ST1 Mixing in Different Magnetic Fields
A. Osintsev1, A. Popov1, M. Fuhs2, and K. Mbius2
2

Kemerovo Technological Institute, Kemerovo, Russian Federation Institute of Experimental Physics, Free University Berlin, Berlin, Germany

Received November 13, 2000

Abstract. We investigate the effect of stochastic modulation of the exchange interaction Jex on singlet (S)-triplet (T) transitions in radical pairs. These transitions limit the lifetime of the photo-generated radical pairs in covalently linked porphyrin-quinone systems that have been developed for biomimetic modeling of photosynthetic electron transfer processes. In order to explain transient electron paramagnetic resonance (EPR) results in different magnetic fields, i.e., with X-band ((0.34 T)/ (9.5 GHz)) and W-band ((3.4 T)/(95 GHz)) time-resolved EPR, we have to assume that Jex is modulated over a range of 20000 G, which is wide enough that S is temporarily almost degenerate with T0 as well as with T1. This large modulation of Jex is caused by restricted rotational diffusion of the quinone subunit with respect to the porphyrin subunit. However, because of the small interradical distance of about 1.01.4 nm, the radical pair is continuously kept in the strong coupling limit and, therefore, we observe only EPR transition between the triplet sublevels. We find an approximation to solve the stochastic Liouville equation valid for rotational diffusion on an intermediate time scale, i.e., the diffusion rate DR is smaller than the singlet electron recombination rate KS ~ 109 s1, but larger than the ST transition rates < 106 s1.

1 Introduction Time-resolved transient electron paramagnetic resonance (TREPR) is a well established method for investigating dynamics and structure of photoinduced shortlived spin-correlated radical pairs (for recent reviews, see [1, 2]). We applied this method to biomimetic porphyrin-quinone model systems for photosynthetic electron transfer [36]. They were synthesized by H. Kurreck and collaborators [710]; the chemical structures of two typical systems considered in the present article is shown in Fig. 1. In these systems, after light excitation, an electron is transferred from a singlet excited porphyrin (P) donor to a covalently linked quinone (Q) acceptor yielding a spin-correlated radical pair. In polar solvents below the melting point (e.g., 150 K in ethanol), after intersystem crossing to

112

A. Osintsev et al.

porphyrin

bridge

quinone

R1 = phenyl R2 = tolyl

ZnTPP-transCH-BQ

ZnTPP-transCH-TMQ Fig. 1. Molecular structures of studied bulky (ZnTPP-transCH-TMQ) and nonbulky (ZnTPP-transCHBQ) porphyrin-cyclohexylene-benzoquinones [610]. The methyl substituent on the bulky benzoquinone marked with S restricts rotation about the bond to the cyclohexylene bridge.

the PT triplet state, the TREPR-detectable radical pair (P+Q)T is formed via triplet electron transfer. The electron recombination to the singlet ground state takes place from the singlet radical pair state (P+Q)S. Because the recombina(PQ)S is very fast (>108 s1 [11]), the triplet radical pair tion rate (P + Q ) S lifetime is mainly limited by singlet-triplet (ST) transitions, (P+Q)T ~ (P+Q)S. The triplet radical pair is observed by TREPR and, in the chosen environment and temperature conditions, the overall molecular motion is slowed down in such a way that anisotropic contributions of the zero-field and the g-tensors are resolved in the spectra. Nevertheless, inside the solvent cage the P and Q constituents of the molecular complex PQ can still undergo fast restricted rotations. In order to relate the experimental TREPR results with the molecular dynamics and, thereby, to learn about the lifetime-limiting ST transitions, we investigate in the present paper the role of rotational diffusion of the two molecular constituents of PQ relative to each other (see Fig. 2). This motion leads to a stochastic modulation of the overlap integral of the wave functions of the two unpaired electrons in the radical pair and, therefore, to a modulation of the exchange interaction. The effect of a modulated exchange interaction has been studied before for unlinked rapidly diffusing radical pairs (radical pair mechanism [RPM]), see, for example, refs. 12-15. In all these systems, by TREPR one observes the noninteracting diffusing radicals. For the covalently linked PQ systems, however, the overall diffusion is slowed down in the solvent, and the diffusion of the Q moiety

Spin Polarization in Radical Pairs with Modulated Exchange Interaction

113

Fig. 2. Definition of the molecular reference frame and the rotational conformation angle . P and Q are the porphyrin and quinone molecular planes, respectively. CH indicates the cyclohexylene bridge. Rotational diffusion about modulates the exchange interaction Jex.

with respect to the P moiety is restricted by the covalent bonds to rotational diffusion about the bond axis. In this case, the two radicals are permanently kept in the strong coupling limit, and with TREPR one observes magnetic resonance transitions in a triplet radical pair. The description of this situation requires a mathematical treatment different from that of standard RPM. The situation is related to the generation of electron spin polarization in flexible-chain biradicals produced by photolysis [16, 17]. In these biradicals the distance and, therefore, the exchange interaction of the two radicals is modulated by dynamic folding of the chain, and the chain length determines the upper distance limit. The TREPR results could be explained with Redfield theory to calculate the effect of modulation of the exchange interaction on various relaxation rates [16, 17]. This approach is correct only in the case of fast diffusion and when no level crossing occurs, i.e., when ST mixing terms, such as differences in hyperfine interactions and g-values of the pair partners, are negligible. The dynamics of the radical-pair spin system is determined by the intramolecular rotational diffusion rate DR, defined via the correlation time , by the transition rates between levels and by level depopulation rates (DR is the coefficient for diffusion about the bond between P and Q of the radical. More strictly, , with as the amplitude of the restricted diffusion. In our case , see below). Such a depopulation occurs via electron recombination, KS, that, in our case, defines the lifetime of the singlet level, and via depopulation of triplet levels, with rates , that define averaged lifetimes of triplet levels. The calculation of the latter one, which consists of the ST transition probability and the singlet recombination rate, is one goal of the present article. Thus, when analyzing quantum transitions and taking into account molecular motion, one has to distinguish the limiting cases. 1. Slow-diffusion limit: the rotational diffusion rates are small compared to transition rates between levels or level depopulation rates. Thus, slow rotational

114

A. Osintsev et al.

diffusion is characterized by DR , KS. This limiting case can be treated as the static problem. 2. Fast-diffusion limit: KS ; DR. In this case, the system is described with diffusionally averaged expressions, and we can calculate the relaxation transitions according to Redfield theory. Nevertheless, there might be additional ST transitions due to ST mixing in the level crossing region, which must be considered in a particular way (see below). 3. The intermediate case: KS. Here, the various elements of the density matrix have to be treated differently: the singlet depopulation is treated as a static problem, while the triplet-sublevel populations are calculated for the diffusionally averaged system. In our case, for KS a lower limit can be given from transient absorption measurements [11]. D^ can be estimated from the experimental results. It will be shown that the photoinduced radical pairs of the nonbulky (see below) PQs must be described by the intermediate case 3. However, it should be noted that in all three cases considered above, an explicit form of an average tripletlevel decay rate can principally be obtained only with one of the approximations. Therefore, the validity of the approximation can be verified post factum only. In earlier work on covalently linked porphyrin-quinones with relative orientational diffusion [18], it was possible to formulate a seminumerical solution for calculating effective lifetimes of radical pairs from the stochastic Liouville equation. All three cases, 13, were covered and basically the experimental results could be explained, although some inconsistencies remained, for example, in the choice of the parameter set [18]. The analytical solution for the intermediate case 3 presented here provides a deeper insight into the dependence of the triplet radical-pair lifetime on the various parameters. It will be shown that, with the new solution and a plausible set of parameters, the TREPR spectra of the photogenerated PQ radical pairs can be satisfactorily interpreted.

1.1 Experimental Results The detailed results of the TREPR experiments at (0.34 T)/(9.5 GHz) (X-band) and at (3.4 T)/(95 GHz) (W-band) have been reported elsewhere [36]. It was shown that the biomimetic systems, in which the porphyrin donor is covalently linked to the quinone acceptor via a cyclohexylene bridge, can be classified according to their TREPR spectra in two categories [18]. (1) When molecular structure and elevated temperature allow the quinone to have some freedom to rotate about the bond to the cyclohexylene bridge, the TREPR spectra are totally in emission, independent of the magnitude of the applied magnetic field [4]. An example is given by the case that the acceptor is a benzoquinone, linked in trans conformation by cyclohexylene to zinc triphenylporphyrin (ZnTPP-transCH-BQ in Fig. 1). (2) When the temperature is decreased and/or the quinone carries a bulky substituent (marked S in Fig. 1) close to the site where it is linked to the

Spin Polarization in Radical Pairs with Modulated Exchange Interaction

115

cyclohexylene, the rotational degree of freedom is further suppressed, and the TREPR spectrum shows an emissive-emissive-absorptive-absorptive (eeaa) lineshape with almost no net polarization [4]. An example is given by the case when the acceptor is a trimethylequinone (ZnTPP-transCH-TMQ in Fig. 1). Thus, the difference of the two categories can be described in terms of the slope of the potential well in which the rotational diffusion about the bond axis takes place. To the first category we refer as "nonbulky" quinones (flat potential), to the second category as "bulky" quinones (deep potential). TREPR lineshapes are determined by the transition energies and moments and by the respective populations of the four radical-pair states. The PQ systems form strongly coupled radical pairs in which the singlet state, S, and the three triplet states, T+1, T0, T1, are only approximately eigenstates. Therefore, only the two allowed transitions (T0 - T1) between the triplet sublevels are observed to first order. The pulsed-laser-generated system is spin polarized, i.e., the population differs from thermal equilibrium at early times after the laser pulse. Both transition energy and polarization depend on the orientation of the molecule relative to the external magnetic field. Consequently, the lineshape is given by the average over all possible orientations. In the radical pairs of both types, bulky and nonbulky systems, the polarization at early times after photo-excitation is due to triplet mechanism. The mechanism which generates the polarization in the bulky Q systems after excitation was briefly discussed in ref. 5 and in more detail described recently [6]. For this situation, the polarization at later times after the laser pulse arises from anisotropic spin-lattice relaxation effects which cause ST transitions followed by fast singlet recombination, thereby depopulating either T0 more efficiently than T+1 and T1, or vice versa. The experimental results in the case of the nonbulky Q systems (TREPR spectra with emissive contributions only) show that the lower the energy of the respective triplet state, the more it becomes depopulated, i.e., T1 is more depopulated than T0, which, for its part, is more depopulated than T+1. In the present paper, we discuss a model that describes the case of bulky PQ as well as the case of nonbulky PQ systems.

1.2 Outline of the Theoretical Model We use, as starting point, the model described in ref. 18 to explain the depopulation rates for the nonbulky quinones. ST1 mixing followed by singlet electron recombination provides the mechanism for depopulation of T1. This mechanism is effective only when the singlet state is energetically degenerate with T1, or at least close to it. The energy gap between S and T1 depends on the exchange interaction and the external magnetic field. For ST1 mixing to be effective in X- as well as in W-band EPR, the exchange interaction must be modulated in the course of rotational diffusion over a wide range of values, so that S may become temporarily degenerate with T1 at magnetic fields corresponding to both X- and W-band EPR. The ST mixing situation is indicated in Fig. 3. As matrix elements providing ST mixing we take into account contributions of hyperfine

116

A. Osintsev et al.

Conformational coordinate

Fig. 3. Variation of singlet and triplet energy levels of porphyrin-cyclohexylene-benzoquinone radical pairs with rotation angle between the P and Q planes. This scheme defines the S T0 and s T1 transition zones. The energy gap between the triplet sublevels is mainly determined by the electronic Zeeman energy with the mean g-factor of the porphyrin and quinone radicals, gB0 = . Jex is the exchange interaction with the dependence on given by Eq. (6).

interactions in both donor and acceptor radicals and, in contrast to ref. 18, also contributions of the anisotropic g-tensor of the quinone anion radical. The rotational diffusion, which modulates the exchange interaction, at the same time modulates the g-values by changing the orientation of the quinone g-tensor with respect to the magnetic field. In contrast to the derivation in ref. 18, we now use a potential for the rotational diffusion, which in the limit of freezing out the diffusion, results in a planar molecule conformation. This is in accordance with the X-ray structure [19], which was determined for PQ systems with bulky Q. In this conformation the exchange interaction | Jex | has its maximum value (| Jex | is estimated to reach at least 20000 G) and, consequently, the singlet state is well separated from all triplet states in X- and W-band EPR.

2 Theoretical Model 2.1 The Basic Assumptions and Main Formulae The spin Hamiltonian consists of the magnetic interactions of the two electron spins, SA and SB [20]: the electron Zeeman interaction containing the g-tensors gA and gB, the hyperfine interaction containing the splittings aAi and aBj by the magnetic nuclei, i and j, of the radicals A and B, respectively, the spin-spin in-

Spin Polarization in Radical Pairs with Modulated Exchange Interaction

117

teraction containing the zero-field tensor D, and the exchange interaction containing the exchange integral Jex:

where is the Bohr magneton, B0 is the external field vector, S and I are the electron and nuclear spin vector operators, respectively. Because in the systems of interest the interradical distance is larger than 1.0 nm, we assume (what is consistent with the experimental results) that the zero-field tensor is axial, i.e., the zero-field parameter E = 0. Then the term describing the zero-field splitting reduces to d = D/3(3cos2 1) with the second zero-field parameter D and the angle between the dipolar axis and B0. The time development of the spin system is given by the stochastic Liouville equation for the spin-density matrix (q, t) of the radical pairs, in which electron recombination and rotational diffusion are included [21, 22]. Using superoperator formalism, we obtain [23]:

(2)

where q is a set of generalized coordinates in the molecular conformation space, and (q, t) is the dynamic Liouvillian for the spin Hamiltonian. is the recombination superoperator and L(U(q), q) describes the rotational diffusion in the potential U(q). Since recombination to the singlet ground state is possible from the radical-pair singlet state only, has the form

(3)

where PS is the projection operator to the singlet state and KS(q) is the recombination rate from the singlet state. As was mentioned in Sect. 1, the overall diffusional motion is so slow that the anisotropic contributions of the zero-field tensor are not averaged out. Therefore, we assume that the molecule is staying in a solvent cage with its orientation to B0 given by the polar angles and On the other hand, inside the solvent cage rotational diffusion of the quinone around its bond to the cyclohexylene may take place. We refer to the angle between the molecular planes of the porphyrin and the quinone as (see Fig. 2 for the various molecular axes and angles). In this model, the set q of generalized coordinates contains , and . We assume, as before [18], that the electron recombination is independent of these angles and that the diffusion operator acts only on . Neverthe-

118

A. Osintsev et al.

less, the Hamiltonian dW depends on all three coordinates. Then the stochastic operator L(U(q),q) has the form:

(4)

where DR is the diffusional rotation coefficient and = represents the corresponding rotational correlation time. This operator leads only to a redistribution of the particles over the coordinate , without changing the integral value of radical-pair triplet and singlet population. The potential we choose is essentially different from that of ref. 18, because MO energy minimization calculations and X-ray structure data show that in equilibrium the PQ molecule is planar, i.e., = 0 [19]. The potential

(5)
fulfills the requirement of such an equilibrium conformation. Furthermore, the results obtained in ref. 19 also allow us to restrict the angle to values ranging from /2 to /2, since at these border values the conformational potential becomes so high that it may be considered as repulsive wall. The angular dependence of the exchange interaction is caused by the change in overlap of the two orbitals of the radical-pair partners, and we approximate it by [18]

(6)
With these assumptions and a J1 value of about 20000 G we end up with a strongly coupled radical pair in the case of the bulky Q being characterized by a large positive U0. In the case of the nonbulky Q, which we treat here, U0 is between 0 and 3 k B T and, by means of rotational diffusion, the singlet state gets temporarily close to T0 and T1 in X- and W-band EPR fields (Fig. 3). Details of the parameter set are discussed in Sect. 4. The allowed EPR transitions take place between the triplet substates, and there are only minor effects of the rotational diffusion on the transition energies. We therefore write the equations in the singlet-triplet basis and have to calculate s T1 and S - T0 transitions. Because of the negative values for Jex(), we do not have to consider S - T+1 transitions. Generally, it is possible to transfer population between T0 and T1 via transition to and from S. But for high recombination rates KS, as is shown below, the singlet state is always virtually empty. In this case, to calculate S -> T transition probabilities, we can use a twolevel system, one singlet and one triplet state. With the notation Tm for either T0 or T1, we can rewrite the Hamiltonian in the STm basis (m = 1, 0) as
(7)

Spin Polarization in Radical Pairs with Modulated Exchange Interaction

119

) are the energies (in angular frequency units) Here ES(, ) and ) is the maof the singlet and triplet states depicted in Fig. 3, and trix element which causes ST mixing. A detailed discussion of () is given in Sect. 4. The coordinates , may be considered to be "frozen" on the time scale of the TREPR experiment. They define the orientation of the external magnetic field B0 with respect to the dipolar axis of the molecule. Hence, in the following derivation of analytical expressions for ST transitions, we omit the dependence on and <j>, and keep only the dependence on . With the definition for the recombination term (Eq. (3)) and the relation between the matrix elements of and matrix elements of the corresponding Hamiltothe dynamic Liouvillian nian (Eq. (7))
(8)

we obtain the following four coupled differential equations which have to be solved (omitting index m and coordinate dependence of matrix elements of Eq. (7)):

(9)

where () = ET() ES(). It should be mentioned that also V depends on . This will be taken into account below. In order to solve this system of equations we use the Laplace transform for the density matrix with s as Laplace variable [24]:
(10)

We do not take into account explicitly the electron transfer at the very beginning of the formation of the radical pair (i.e., for times tobs < 106 s), but we assume instead that the initial state is the triplet radical-pair state. Then the system of equations becomes:

120

A. Osintsev et al.

The quantity p0() is the initial population of the triplet state and, therefore, we set
(12)

Because of the symmetry of Eqs. (5) and (6), the lower limit in all integrals over may be set equal to zero. 2.2 The Approach To Solve the Stochastic Liouville Equation We want to solve the system of coupled differential equations for the approxiKS, because otherwise we do not expect large enough depopulamation DR tion rates [18]. Therefore, we omit the diffusion operators in the first and two last equations of (11). This is similar to the approximations in refs. 25 and 26. It means that SS, ST and TS decay, due to recombination, faster than they diffuse about . On the other hand, we will see later that TT diffuses before it decays. Therefore, the decay rate for TT is averaged over . We are interested in calculating the polarization for observation times tobs of about 1 s after the photoexcitation, when the emissive radical-pair spectra are . For the Laplace transdetected. Therefore, we can use the condition tobs KS. Therefore, we neglect formed functions this condition means that s s in Eqs. (11) whenever it is added to KS. After these simplifications we can rewrite Eqs. (11) as:

(13)

Spin Polarization in Radical Pairs with Modulated Exchange Interaction

121

Actually, neglecting the term s compared with the decay rate KS will set the in the inverse Laplace derivatives and transform (Eqs. (9)) to zero. Formally, this looks like the quasistationary approximation, which is widely used in chemical kinetics [27]. Nevertheless, the classic quasistationary approximation deals with concentrations (or populations) and has nothing to do with coherences. Hence, in the following we give a condition for the validity of this approximation in the quantum mechanical case. in Eqs. (9), and In order to neglect one has to show that these derivatives are small compared to the other terms in the respective equation. This has to be done separately for real and imaginary parts. Neglecting L(), as discussed above, one has to prove the following conditions to be fulfilled:

(14)

With the solution of the quasistationary approximation in Eqs. (20) and (27) (see below) for TT(,t) and with Eqs. (13), one can calculate the other elements of the density matrix and its derivatives. The necessary condition of validity of Eqs. (14) and thereby of the quasistationary approximation is

(15)
and it follows that KS and KS (see Eqs. (27)). This ensures that all three quantities, , can be treated in the same way as the intermediate quantity B in a reaction A C with k1 : k2. In this case, the quasistationary condition can be easily proved [27]. For example, with Eqs. (13) the quasistationary solution for SS(,t) is

(16)
and we see that the singlet population is very small. This confirms also the assumption, made with Eq. (7), that S ** T0 and S <-> T1 transitions do not interfere and can be treated separately. After having shown that the approximations in Eqs. (13) are justified, it is now possible to solve this system of coupled differential equations. Excluding off-diagonal elements and the variable ( , s ), we obtain

122

A. Osintsev et al.

(17)
where

(18)
Since V : KS we can simplify this expression:

(19)
An estimate for Eq. (19) gives () < 108 s1 and, consequently, KS. This means, consistent with Eq. (16), that the rate *: for the triplet-singlet transitions cannot sufficiently compensate the decay of the singlet population due to recombination. In other words: in our case, the singlet-state level performs simply as a reservoir of infinite capacity for the triplet-level populations. In general, Eq. (17) cannot be solved analytically. However, when the decay of TT(,t) is much slower than the rotational diffusion, i.e., () DR, we may consider the coordinate dependence of this matrix element to be defined by a probability function p(, t) (probability of finding the molecule with conformation at time t [18]). Moreover, for the time tobs , stochastic molecular motion leads to dynamic equilibrium concerning the coordinate , i.e., p0(). Thus, fast diffusion compared to the decay time, leads p(,t the spin system to be always distributed concerning according to Boltzmann equilibrium, and we can put
(20)

with
(21)

and

(t) defined by
(22)

giving the overall population of the triplet state. Its Laplace transform is

(23)

We now integrate Eq. (17) over and make use of the boundary condition Eq. (12). Because the kinematic flow integrated over full space is zero:

Spin Polarization in Radical Pairs with Modulated Exchange Interaction

123

(24)

we obtain
(25)

With approximation Eq. (20) it follows that

(26)

Thus, with Eqs. (19) and (26), we obtain from inverse Laplace transformation

(27)

As a final result, the complex network of population and depopulation rates, which is described by the coupled differential equations, may be represented symbolically by analogy to chemical reactions according to Fig. 4a, and its simplification may now be rewritten as depicted in Fig. 4b.
a

Fig. 4. a Graphical representations of the complex network described by the coupled differential Eqs. (9). b The approximations described in the text lead to this simplified network with given by Eqs. (27). As discussed in the text, to calculate the singlet-triplet transitions it is sufficient to consider only two states, the relevant triplet sublevel and the singlet state. The transition rate for T0 is labeled , the transition rate for T1 is

124

A. Osintsev et al.

2.3 Further Approximations In Eqs. (27), both the TS and VTS depend on . We now want to discuss a special case which will help to get an idea of the different dependences of the rates and on various parameters. Let us assume that TS is constant. Then the singlet-triplet transitions are due to relaxation caused by stochastic fluctuations of . In this case, for Eq. (27) can be rewritten as

(28)

with

This looks like the expression for Redfield relaxation, with (KS/2)1 as correlation time. The derivation is valid for the case KS DR and, therefore, the stochastic fluctuations of FT()S are static on the time scale of the evolution of the spin system. Then the transition rate is calculated like for the case of static ST mixing and is proportional to When KS is further increased, the mixing process cannot evolve because of the fast decay of the coherences, and the depopulation decreases with K$l. Now we want to examine the case when level crossing occurs for = c. We expect that for in Eqs. (27) mainly the molecules with the conformation = c contribute to . This leads to another simplification. Usually increases very fast when the system passes the S-T1 crossing point, i.e., has a great value. Then, expanding () into a power series and retaining only the linear term, we obtain

(29)

With the definition

(30)

it can be rewritten as
(31)

R defines the effective size of the crossing zone and, when it is very narrow, we . Then can use the representation for the Dirac -function: (x) =

Spin Polarization in Radical Pairs with Modulated Exchange Interaction

125

(32)

and integration yields the depopulation rate

(33)
The Eqs. (32) and (33) are repeatedly used for describing the S-T1 transitions in the term crossing zone [2830]. The condition for the validity of Eq. (33) is R : /2. Our estimations show that this condition works fine with a cosine-dependent exchange integral defined 3 2 by Eq. (6), provided the values of R are within the range of 10 10 rad in the region that covers X- and W-band EPR crossing points. Thus, can be expressed as a function of the values of different quantities at the crossing point = c, such as the slope of the energy separation, the matrix element and the probability of finding the system with = c. We used the approximation K(O) <SC DR, because the differential Eq. (17) could not be solved for the general case. However, the necessary condition is, in fact, weaker because not only the absolute value of the rate for the conformation with angle , (), is important, but also the time c R 2 / D R which the system stays in a conformation with high rate, i.e., in the crossing zone. To retain S-T1 transitions weaker than diffusion mixing, we reformulate the condition above to
(34)

This means that the transition probability during one passage of the crossing zone is very small. 3 Calculation of the EPR Spectra For the simulations of in a rate equation fig. 9). This allows us simple way. Thus, we TREPR spectra we include the depopulation rates and for the triplet-level populations (see, for example, ref. 6: to take into account additional relaxation processes in a obtain
(35)

where are the populations of the T0 and T1 levels, respectively, and r0 and r1 are the rates for the irreversible decay of the triplet level populations. This decay may be caused by spin-lattice-interaction-induced incoherent transi-

126

A. Osintsev et al.

tions to the singlet state, or by weak triplet recombination via intersystem crossing to the ground state [31, 32]. The rates are defined by Eqs. (27). We did not take into account dipolar relaxation between the triplet sublevels that might be caused either by slow-angle fluctuations of the whole molecule or by stochastic modulation of the distance between porphyrin and quinone. Estimations show that these motions provide only minor contributions as long as the mean amplitude of the dipolar local magnetic-field fluctuations, D, remains smaller than 1 G, which is probably the case. For low enough microwave field amplitude the spectrum is simply given by the differences of the populations of the triplet levels and, of course, the transition energies. Because in our case the lines are inhomogeneously broadened by unresolved hyperfine structure [3], we use the Voigt line-shape representation [33]. It describes the mixture of a homogeneously broadened EPR line with Lorentzian shape and an inhomogeneously broadened line with Gaussian shape. The TREPR spectra are recorded in a direct absorption/emission (a/e) mode of detection, i.e., resonance amplitude I as a function of the external magnetic field, and not as first derivatives with respect to field. Accordingly, for spectra simulations, the calculated absorption amplitude, exhibiting Voigt line shape, is given by [33]:
(36)

is the normalized distance from the line center where (B) = 2 . Here, are the half-widths of Lo(B = 0), and b = rentzian and Gaussian lines, respectively. B is the offset from the center of the resonance line. According to ref. 34, Eq. (36) can be approximated by a sum of Gaussian and Lorentzian line shapes with different weights, but equal half-width B1/2:

(37)

and the half-width with the weight factor 2 1/2 ) ] . As mentioned before, the transition energies depend on the orientation of the molecule with respect to the magnetic field, which is given by the polar angles and (/). Accordingly, we have to integrate I(B; , </>) over all orientations. The TREPR line shape is then given by:

(38)

Spin Polarization in Radical Pairs with Modulated Exchange Interaction

127

4 Results and Discussion After the derivation of equations describing the effect of rotational diffusion, at an intermediate time scale, on the spin polarization in strongly coupled radical pairs has been presented, we now discuss numerical spectra simulations. In particular, we will discuss the choice of parameter values for the simulations of the TREPR results from the nonbulky porphyrin-cyclohexylene-benzoquinone (ZnTPPtransCH-BQ) system. The main characteristics of the TREPR results are that 1 s and later after the laser excitation pulse the spectra are totally emissive, both at X-band and at W-band microwave frequencies (see Fig. 5, solid lines). This is very uncommon for radical pairs which are covalently linked and cannot diffuse apart. Now, the question is whether the presented model can explain this observation and in which range parameters have to be chosen that are not known a priori by independent measurements. The parameters for the spectra simulation calculations shown in Fig. 5 are given in Table 1. The spectroscopic parameters, which determine the transition energies, can be estimated by independent measurements. The g-tensor of Q was measured by high-field EPR [35]. The averaged g-value of P+ has also been measured by high-field EPR [35] to be close to the value for the free electron. The dipolar coupling can be estimated from the spectral width of the TREPR spectra of strongly coupled radical pairs [5, 6]. This value is about 25% higher as compared to the 22 G which have been calculated for trans-PQ with the pointdipole approximation and the X-ray structure [19], but it remains within the limits

Fig. 5. Calculated and experimental TREPR spectra of the laser-excited nonbulky cyclohexylenebridged ZnTPP-transCH-BQ at X- and W-band in frozen solution (X-band: ethanol, 150 K; W-band: ethanol plus toluene mixture, 166 K; 5 s after laser pulse). The solid lines show the experimental data, the dots give the results of our calculations. For the parameters used in the calculations, see text and Table 1. Note that the TREPR spectra are recorded in a direct a/e mode of detection, and not as first field derivatives. The experimental spectra are adopted from refs. 3 and 6.

128

A. Osintsev et al. Table 1. Parameters used for spectra simulations. Parameters Hyperfine contribution to the local magnetic field (G) g-tensor(Q ) [35] g-tensor(P+) Dipolar coupling (G) Exchange interaction (G)

Values

gxx = 2.00645, gyy = 2.00526, gzz = 2.00229 D = 27 (ZnTPP-transCH-BQ) D = 38 (ZnTPP-cisCH-BQ) J1 = 20000 J0 = 400 (ZnTPP-transCH-BQ) J0 = 1200 (ZnTPP-cisCH-BQ) U0 = kBT (nonbulky Q) U0 = 10 kBT (bulky Q) KS = 5109 r(0) = 2105 a = 0 (nonbulky Q) a = 105 (bulky Q) X-band: B1/2 = 7 W-band: B1/2 = 11

Potential Singlet electron recombination rate (s1) Electron relaxation (s1) (see Eqs. (41), (42)) Linewidth (G) Gaussian vs. Lorentzian line-shape contribution (see Eq. (37))

of the approximation. For cis-PQ we chose the dipolar coupling equal to 38 G. Also the line width was adjusted for a best fit to the experimental spectrum. The X4 G, which is close to band line shape contains a Gaussian line width of the value estimated in ref. 3. The larger W-band line width is reasonable because of the enhanced T2 relaxation in higher fields caused by a fluctuating g tensor of Q . The matrix element V causes ST mixing which leads to the depopulation of T0 and T1. Only two interactions are considered to contribute to V. These are isotropic hyperfine (hf) and anisotropic g interactions. The matrix elements, in angular frequency units, for S -> T0 and S -* T1 transitions have the form:

(39)

where subscripts _L and || indicate normal and tangential components of the difference in local field, and superscripts g or hf indicate the origin of the local

Spin Polarization in Radical Pairs with Modulated Exchange Interaction

129

fields. Because the hf interaction is not resolved in the spectra, the averaged effective contribution to the local magnetic field was estimated to be approximately Bhf = 8 G, following the way it was done in ref. 3. Bg is caused by the difference of the g-tensors of P and Q, g = g(Q) g(P+). The parallel component given by gzzB0, the perpendicular component by (gxx + gyy)B0 in the + laboratory frame. The g-tensor values of P are approximated by that of the free electron ge = 2.0023, and the main values of the g-tensor of Q are given in Table 1. We should mention that the anisotropy of the g-tensor of Q is the main contribution to singlet-triplet mixing in W-band EPR. The local magnetic field, and its stochastic fluctuations due to modulation of the g-tensor of Q, becomes one order of magnitude higher than in X-band. Hence, due to the squared matrix elements V 2 in Eqs. (27) and (33), the transition rates become two orders of magnitude higher. In the present PQ systems, one of the main features of the exchange interaction is that it must be modulated over a broad range of values in order to cause S-T1 crossing both in X- and W-band EPR. Therefore, we chose for the values in Eq. (6) | J1 | = 20000 G and | J0 | = 400 G for trans-PQ and | J0 | = 1200 G for cis-PQ. There are two limiting conditions which determine J0. On the one hand, it must be large enough to separate S and T0 levels in W-band EPR sufficiently, i.e., it must be large compared to the respective matrix element V. On the other hand, it must not be too large in order to still allow sufficient S-T0 mixing in X-band EPR. For J1 the condition is that it must be large enough to allow ST1 crossing in W-band. On the other hand, it should not be too large, because the magnitude for S-T1 mixing in Eq. (33) is determined by the slope of the exchange energy along the conformational coordinate, . Semi-empirical theoretical estimations of the distance dependence of the exchange interaction are not helpful in the case of our PQ systems because, dependent on the systems used for an estimation of the parameters, we yield values between 0 and 5 10 G. For this reason, we prefer an estimation of the exchange interaction on the basis of the experimentally determined value for the radical pairs of 2-N,N,N',N'tetraalkyl-p-phenylenediamine-Zn-porphyrin-2-naphthoquinone (TAPD-Zn-NQ) [36]. TREPR investigations of the light-induced charge-separated state provided for this system an exchange integral of about 1 G, which is unexpectedly large. It is at least two orders of magnitude larger than in donor-acceptor radical pairs in photosynthetic reaction centers. The most likely reason for this difference in Jex is that in TAPD-Zn-NQ there is a strong through-bond contribution to the exchange interaction. We assume that there are high through-bond contributions also for our Zn-triphenylporphyrin-benzoquinone (ZnTPP-CH-BQ) radical pairs. In this situation we can make an estimation of Jex on the basis of the reference system TAPD-Zn-NQ. Assuming an exponential dependence of the exchange integral on the distance R between the unpaired electrons [37], one can write
(40)

130

A. Osintsev et al.

With a commonly used attenuation parameter of = 10 nm1 [18, 38] and 1 nm, J ex TAPD-Zn-NQ = 1 G [36] and R TAPD-Zn-NQ = 2 nm, we obtain 2.2 104 G, a value which is close to the chosen one for J1 in Table 1, thus justifying our choice. We assume that the rotational diffusion of the quinone about its bond to the cyclohexylene spacer takes place in a potential of the form given in Eq. (5). The value of U0 mainly controls the probability of the molecule to be in the conformation where the S and T1 levels are degenerate. When U0 is decreased too much, the depopulation rate for the T1 level is not able to create noticeable emissive polarization in W-band EPR. On the other hand, with a potential barrier higher than kBT the depopulation of T1 in X-band EPR is too small. We found that a value of U0 = kBT (at 160 K) works well to simulate the experimental TREPR spectra. It compares favorably with results obtained by molecular modeling calculations [19]. For the electron recombination rate, KS, a lower limit of 108 s1 was measured with transient absorption [11]. To simulate the results for both EPR frequency bands, X- and W-band, it was necessary for the recombination rate to be at least 510 9 s1. Lower values for KS result in too small depopulation rates of T0 in X-band. On the other hand, when decreasing J0 the depopulation of T0 is enhanced, but then in W-band Zeeman fields the S and T0 levels are not sufficiently separated. In Eqs. (35) the rates r0,1 were introduced in order to describe the effect of additional depopulation of the triplet sublevels. We divide the rates r0,1 into two parts, an isotropic and an anisotropic part:

(41)
A value r = 210 s was chosen to fit qualitatively the polarization decay of the experimental TREPR spectra (Fig. 6). This additional depopulation may be caused either by relaxational transitions followed by electron recombination or by slow triplet recombination (i.e., via reversed triplet mechanism [31, 32]). , will be discussed Contributions from additional anisotropic relaxation rates, below. is very critical because only in the case when T1 is The ratio depopulated slightly faster than T0, both EPR transitions between the three tripin X- and Wlet sublevels become emissive. Because all four rates, band EPR, depend differently on the g-tensor anisotropy via the matrix elements, the simulated lineshapes become very sensitive to variations of the parameters. The approximate expressions Eqs. (27) and (33) provide a deeper insight into the dependence of on the various parameters. For S and T0 levels separated by | J0 | > 100 G, there is no effective ST0 mixing and the depopulation of T0 is caused by the contribution approximated by Eq. (28). With the chosen parameters, the time-averaged value of 2Jex. Therefore, smaller absolute values of J0 and J1, i.e., 1, where The only , and larger values of KS lead to higher smaller values for values between X- and W-band EPR is the matrix element difference for the
(0) 5 1

Spin Polarization in Radical Pairs with Modulated Exchange Interaction

131

experimental

calculated

experimental

calculated

Fig. 6. Comparison of experimental (ethanol 150 K) and calculated X-band TREPR spectra for the nonbulky trans- (a) and cis-cyclohexylene-bridged ZnTPP-CH-BQ (b) for various delay times after laser excitation. For the parameters used in the calculations, see text and Table 1. Note that the TREPR spectra are recorded in a direct a/e mode of detection and not as first field derivatives. The experimental spectra are adopted from reference 3.

VTS and, therefore, the rate is always larger in W-band than in X-band. The amplitude of the potential, U0, is included via the probability p0(). Larger U0 leads to smaller ic0, because the probability of S to be energetically close to T0 decreases, and the time-averaged value of Jex gets larger. S and T1 are degenerate for = c. Therefore, the main contribution to ic_l is given according to Eq. (33). In contrast to the calculation of ic0, now ic_l depends differently on J1 and U0 in X- and W-band EPR. J1 must be large enough to make S and T1 degenerate in W-band. However, the relation between J1 and is rather complex, because it depends on the slope U0 determines the probability of finding the system degenerate. For U0 < 5kBT, increasing U0 leads to larger rates in W-band, but smaller rates in X-band EPR. U0 therefore controls the ratio of in X- and W-band. With this parameter choice the values for K(d) in Eqs. (27) become smaller than 108 s1. The values for the relevant

132

A. Osintsev et al.

triplet sublevel depopulation rates, , are between 103 and 106 s1, depending on the orientation of the molecule with respect to B0. When increasing the potential U0 to a value of 20kBT, the molecule gets frozen in the conformation with = 0, and = 0. This is the case for the bulky systems, for which the rotational motion is locked, as was mentioned in Sect. 1. Then the polarization is caused by electron relaxation, followed by electron recombination. It was observed that in this situation spin-rotational interaction causes anisotropic relaxation [5, 6]. With the additional phenomenological relaxation rates

(42)
and a potential U0 = 20kBT the experimental spectra of the bulky porphyrincyclohexylene-trimethylbenzoquinone can be simulated without other parameter changes (Fig. 7). In Eqs. (42) a is a free parameter which determines the relaxation rates. is the angle between B0 and the axis perpendicular to the z-axis of the zero-field splitting in the porphyrin-quinone plane. This is discussed in more detail in ref. 6. In the intermediate-diffusion-limit approximation used, the diffusional constant DR for the rotational diffusion about does not enter explicitly. For the approximation to be valid, DR must be smaller than KS. In the opposite case DR to become much smaller. This is because, during we expect the rates the nonadiabatic passage, the system has no time to oscillate coherently, and only a relaxation-type contribution remains. The other approximation was DR (), X-band W-band

Fig. 7. Comparison of experimental (solid lines) and calculated (dots) TREPR spectra for the bulky cyclohexylene-bridged ZnTPP-transCH-BQ (where the quinone rotation is chemically locked) at Xand W-band (ethanol plus toluene, 166 K, 5 s after laser excitation). Parameters in the calculation a = 105 s1, U0 = 10kBT. For other calculation parameters, see text and Table 1. Note that the TREPR spectra are recorded in a direct a/e mode of detection and not as first field derivatives. The experimental spectra are adopted from refs. 4 and 6.

Spin Polarization in Radical Pairs with Modulated Exchange Interaction

133

1 in Eq. (34). When this is not valid the depopu(), weakened to lation rates are different at different and, therefore, the overall decay of TT is no longer monoexponential. The approximation we used is reasonable. Nevertheless, it is preferable to use a general solution for all values of DR, like it was presented previously [18]. The authors calculated generalized lifetimes, g, by integrating the stochastic Liouville Eqs. (9) over time. In the reKS the rate can be identified with because of gion monoexponential decay. As was mentioned in Sect. 1, in ref. 18 the spin-dynamics problem was solved for the potential U = U0 cos2 instead of that of Eq. (5), i.e., for large U0 the molecule gets frozen in a conformation in which the quinone plane is perpendicular to the porphyrin plane. This means that for large U0 the exchange interaction is small and, consequently, no S-T1 level crossing occurs, and the depopulation rate of T1 is expected to drop. Nevertheless, comparing fig. 3b and d (top line) in ref. 18 the depopulation rates do not drop significantly, and even increasing the potential up to 2000kBT does not make any changes. The origin of this discrepancy still remains an open question.

5 Conclusions The totally emissive radical-pair spectra of photoexcited covalently linked porphyrin-quinone donor-acceptor systems in time-resolved X- and W-band EPR can be explained by a theoretical model in which rotational diffusion of the two radical sites with respect to each other is allowed on an intermediate time scale. This intermediate-diffusion-limit approximation is valid as long as the diffusional constant DR is small compared to the singlet-recombination rate KS of 510 9 s1 and large compared to the triplet-sublevel depopulation rates () that are smaller than 108 s1. It could be shown that the latter condition is weakened according to Eq. (34). This result supports the notion that, in the case of a nonbulky quinone, it can undergo restricted rotational diffusion about its bond to the cyclohexylene bridge in a rather flat potential. In the present model, the stochastic Liouville equation is solved neglecting the rotational diffusion operator in its combination with the recombination operator. The solution is valid only for observation times tobs after photoexcitations which are larger than Z)^1. It was necessary to introduce this restriction on tobs in order to use the quasistationary approximation (Eqs. (13)), i.e., the populations of the intermediate states, SS, ST and TS, must have assumed the quasistationary value. Furthermore, the restriction on tobs was used for solving Eq. (17): tobs ensures that the system is in dynamic equilibrium concerning the diffusion coordinate . This is verified by the experimental results because otherwise not all spectral transitions could be emissive due to ST1 mixing. It should be mentioned that either one of the assumptions, , guarantees monoexponential decay of the TREPR signal. Moreover, it is possible to find analytic expressions for the singlet-triplet transition rates. They are the limiting rates which determine the lifetime of strong-

134

A. Osintsev et al.

ly coupled radical pairs. When singlet and triplet levels cross due to the rotational modulation of the exchange interaction, the transitions mainly take place in this level-crossing region. When no level crossing occurs, the depopulation process becomes similar to relaxation, with 2/KS as the correlation time. When DR : KS, the matrix element VTS, which is modulated by the rotational diffusion, is stable on the time scale of the recombination. This is why in this approximation the rotational diffusion can be neglected. It is interesting to note that (corresponding to Eqs. (27)) a larger singlet-recombination rate may lead to smaller depopulation of the triplet sublevels because the singlet-triplet mixing has no time to evolve. Hence, an electron-transfer system may be trapped in its electron-separated (triplet) state by a large singlet-recombination rate. Although the investigated PQ systems fulfill the approximations made in the present theoretical model, in particular those concerning the time scale of the rotational diffusion, it is desirable to find a general solution for all diffusional constants, as was outlined before [18]. This approach, however, still needs refinement, as is discussed above. Also the effect of spin-rotational interaction is still an open point. This is because its stochastic modulation is caused by a diffusion process whose rate is much larger than the electron-recombination rate and, therefore, cannot be treated in the given approximation. Work on these points is in progress. Acknowledgements We acknowledge numerous contributions of all the other collaborators in our group at the FU Berlin who were or still are working on the porphyrin-quinone systems, in particular G. Elger, for measurements and helpful discussions in the course of the present work. We are indebted to H. Kurreck (FU Berlin) and his collaborators A. Wiehe, J. von Gersdorff, H. Diecks and M. Speck for synthetizing the porphyrin-quinone compounds, and to K. Salikhov (Kazan) for helpful suggestions and discussions. This work was supported by the High School Department of the Russian Federation (grant no. 97-0-14.0-90), the Deutsche Akademische Austauschdienst (DAAD), the Deutsche Forschungsgemeinschaft (SFB 337, SFB 498, SPP 1051), and the Volkswagen-Stiftung (I/73 145).
References 1. 2. 3. 4. 5. 6. 7. 8. Stehlik D., Mbius K.: Annu. Rev. Phys. Chem. 48, 745 (1997) Hoff A.J., Deisenhofer J.: Phys. Rep. 287, 1 (1997) Schlpmann J., Lendzian F., Plato M., Mbius K.: J. Chem. Soc. Faraday Trans. 89, 2853 (1993) Elger G., Kurreck H., Wiehe A., Johnen E., Fuhs M., Prisner T., Vrieze J.: Acta Chem. Scand. 51, 593 (1997) Elger G., Fuhs M., Mller P., von Gersdorff J., Wiehe A., Kurreck H., Mbius K.: Mol. Phys. 95, 1309 (1998) Fuhs M., Elger G., Osintsev A., Popov A., Kurreck H., Mbius K.: Mol. Phys. 98, 1025 (2000) Kurreck H., Huber M.: Angew. Chem. Int. Ed. Engl. 34, 949 (1995) von Gersdorff J., Huber M., Schubert H., Niethammer D., Kirste B., Plato M., Mbius K., Kurreck H., Eichberger R., Kietzmann R., Willig F.: Angew. Chem. Int. Ed. Engl. 29, 670 (1990)

Spin Polarization in Radical Pairs with Modulated Exchange Interaction

135

9. Dieks H., Sobek J., Tian P., Kurreck H.: Tetrahedron Lett. 33, 5951 (1992) 10. Wiehe A., Senge M.O., Kurreck H.: Lieb. Ann. Recueil. 9, 1951 (1997) 11. Fuhs M., von Gersdorff J., Dieks H., Kurreck H., Mbius K., Prisner T.: J. Chem. Soc. Faraday Trans. 92, 949 (1996) 12. Freed J.H., Pedersen J.B.: Adv. Magn. Reson. 8, 1 (1976) 13. Salikhov K.M., Molin Y.N., Sagdeev R.Z., Buchachenko A.L.: Spin Polarization and Magnetic Effects in Radical Reactions. Amsterdam: Elsevier 1984. 14. Adrian F.J., Monchick L.: J. Chem. Phys. 71, 2600 (1979) 15. Shushin A.I.: Chem. Phys. 144, 201 (1990) 16. Avdievich N.I., Forbes M.D.E.: J. Phys. Chem. 99, 9660 (1995) 17. Avdievich N.I., Dukes K.E., Forbes M.D.E., DeSimone J.M.: J. Phys. Chem. A 101, 617 (1997) 18. Salikhov K., Schlpmann J., Plato M., Mbius K.: Chem. Phys. 215, 23 (1997) 19. Fajer J., Barkigia K.M., Melamed D., Sweet R.M., Kurreck H., von Gersdorff J., Plato M., Rohland H.-C., Elger G., Mbius K.: J. Phys. Chem. 100, 14236 (1996) 20. Hore P.J. in: Advanced EPR, Applications in Biology and Biochemistry (Hoff A.J., ed.), chap. 12. Amsterdam: Elsevier 1989. 21. Evans G.T., Fleming P.D., Lawler R.G.: J. Chem. Phys. 58, 2071 (1973) 22. Pedersen J.B., Freed J.H.: J. Chem. Phys. 61, 1517 (1974) 23. Purtov P.A., Doktorov A.B.: Chem. Phys. 178, 47 (1993) 24. Mathews J., Walker R.L.: Mathematical Methods of Physics. Menlo Park: Benjamin W.A. 1964. 25. Doktorov A.B., Burshtein A.I.: Zh. Teor. Eksp. Fiz. 68, 1349 (1975) 26. Doktorov A.B.: Physica A 90, 109 (1978) 27. Eyring H., Lin S.H., Lin S.M.: Basic Chemical Kinetics, p. 18. New York: Wiley 1980. 28. Zharikov A.A., Shokhirev N.V.: Z. Physik. Chem. 37, 177 (1992) 29. Adrian F.J.: J. Chem. Phys. 102, 4409 (1995) 30. Popov A.V., Purtov P.A.: Chem. Phys. Rep. 16, 27 (1997) 31. Steiner U.E., Ulrich T.: Chem. Rev. 89, 51 (1989) 32. Sasaki S., Katsuki A., Akiyama K., Ikegami Y., Tero-Kubota S.: J. Am. Chem. Soc. 119, 1323 (1997) 33. Poole C.P.: Electron Spin Resonance, p. 811. New York: Wiley 1967. 34. Whiting E.E.: J. Quant. Spectrosc. Radiat. Transf. 18, 1379 (1968) 35. Burghaus O., Plato M., Rohrer M., Mbius K., MacMillan F., Lubitz W.: J. Phys. Chem. 97, 7639 (1993) 36. van der Est A.J., Fuechsle G., Stehlik D., Wasielewski M.R.: Ber. Bunsenges. Phys. Chem. 100, 1082 (1996) 37. Herring C., Flicker M.: Phys. Rev. 134 (1964) 38. Closs G.L., Miller J.R.: Science 240, 440 (1988)

Authors address: Michael Fuhs, Institute of Experimental Physics, Free University Berlin, Arnimallee 14, 14195 Berlin, Germany E-mail: michael.fuhs@physik.fu-berlin.de