Pengfei
Huo
and
Thomas F.
Miller III
*
Division of Chemistry and Chemical Engineering, California Institute of Technology, Pasadena, CA 91125, USA. E-mail: tfm@caltech.edu
First published on 1st June 2015
We apply real-time path-integral dynamics simulations to characterize the role of electronic coherence in inter-complex excitation energy transfer (EET) processes. The analysis is performed using a system-bath model that exhibits the essential features of light-harvesting networks, including strong intra-complex electronic coupling and weak inter-complex coupling. Strong intra-complex coupling is known to generate both static and dynamic electron coherences, which delocalize the exciton over multiple chromophores and potentially influence the inter-complex EET dynamics. With numerical results from partial linearized density matrix (PLDM) real-time path-integral calculations, it is found that both static and dynamic coherence are correlated with the rate of inter-complex EET. To distinguish the impact of these two types of intra-complex coherence on the rate of inter-complex EET, we use Multi-Chromophore Förster Resonance Energy Transfer (MC-FRET) theory to map the original parameterization of the system-bath model to an alternative parameterization for which the effects of static coherence are preserved while the effects of dynamic coherence are largely eliminated. It is then shown that both parameterizations of the model (i.e., the original that supports dynamic coherence and the alternative that eliminates it), exhibit nearly identical EET kinetics and population dynamics over a wide range of parameters. These observations are found to hold for cases in which either the EET donor or acceptor is a dimeric complex and for cases in which the dimeric complex is either symmetric or asymmetric. The results from this study suggest that dynamic coherence plays only a minor role in the actual kinetics of inter-complex EET, whereas static coherence largely governs the kinetics of incoherent inter-complex EET in light-harvesting networks.
Light-harvesting systems are typically comprised of multiple complexes, which are in turn comprised of multiple light-absorbing chromophores. For example, Photosystem II is comprised of the CP43, CP47, and reaction center complexes, each of which includes multiple chromophores.18–20 In most light-harvesting systems, chromophores within a complex exhibit strong electronic coupling due to their close-packed configurations.11,12,18,19 The electronic coupling between chromophores in different complexes, however, is typically much weaker due to the larger distances of separation, leading to incoherent dynamics for EET between complexes. The manifestation of electronic coherences in EET has been the subject of significant experimental2,3 and theoretical11,12,21–32 attention. In the current work, we specifically focus on the way in which intra-complex electronic coherence affects inter-complex EET kinetics, which is crucial to the efficiency of light-harvesting systems. We address this issue by distinguishing between static and dynamic types of electronic coherence and by examining the relative impact of these two types of intra-complex coherence on the kinetics of inter-complex EET.
Fig. 1 introduces the simple model for inter-complex energy transfer that is employed in this study.11,12,22 The model exhibits an EET donor complex that consists of a pair of chromophores and an acceptor complex that consists of a single chromophore (Fig. 1a). Inter-complex electronic coupling is described via the parameter δ and intra-complex electronic coupling is given by Δ, where the latter is a function of the tilt angle, θ, between the two chromophores of the dimer (Fig. 1b). The intra-complex coupling can be varied via the chromophore-dipole tilt angle (Fig. 1c); previous theoretical work has found that the sign of the intra-complex coupling has a significant influence on both the rate of inter-complex EET and non-linear spectroscopic features.22,33,34 Despite its simplicity, this model exhibits many of the essential features of EET in light-harvesting systems for regimes in which the intra-complex coupling is relatively large (βΔ > ∼1, where β = 1/kBT) and the inter-complex coupling is small (βδ ≪ 1).
The current work focuses on characterizing the effect of electronic coherence on the kinetics of EET in light-harvesting systems. In doing so, we will follow earlier work by distinguishing between static coherence and dynamic coherence.10,12,35
Static coherence (also known as “exciton delocalization”) is a time-independent feature associated with shifts in the intra-complex energy eigenspectrum.10,18,36,37 It can be characterized in terms of the off-diagonal elements of the projector for the donor-complex eigenstates, such that
〈D1|![]() | (1) |
Dynamic coherence is a time-dependent feature associated with transient superpositions among the local excitations on individual chromophores.12,18,38 It is most commonly characterized in terms of the off-diagonal elements of the projector for the time-dependent wavefunction of the exciton, such that
〈D1|![]() | (2) |
To characterize the relative contributions of static and dynamic electronic coherence in facilitating inter-complex EET, we use MC-FRET theory15–17 to construct an alternative parametrization of the model in Fig. 1, such that the effects of static coherence are preserved, while dynamic coherence is eliminated via the removal of the intra-complex electronic coupling. Using the partial linearized density matrix (PLDM) path-integral method,13,14 we investigate the EET for the model system, considering both the original parameterization that supports dynamic coherence and the alternative parameterization that eliminates dynamic coherence. The numerical results show nearly identical inter-complex EET kinetics for both sets of parameters over a broad range of EET regimes. These results suggest that static coherence largely governs the kinetics of inter-complex EET in light-harvesting networks, whereas dynamic coherence plays only a minimal role.
Ĥ = Ĥs + Ĥint + Ĥb. | (3) |
Ĥs = ĤDs + ĤAs + ĤCs. | (4) |
![]() | (5) |
![]() | (6) |
Within the dipole approximation, the intra-complex electronic coupling can be related to the relative orientation of the transition dipole associated with individual chromophores using
![]() | (7) |
The model incorporates the effect of a dissipative environment via linear coupling of each chromophore site to a corresponding harmonic bath, using
![]() | (8) |
Ĥb = ĤDb + ĤAb | (9) |
![]() | (10) |
The system-bath coupling constants cDj,i and cAk,i are sampled from the Debye–Drude form of the spectral density,21,41
![]() | (11) |
For use in the following sections, we introduce notation for the Hamiltonian operators associated with the donor and acceptor complexes,
![]() | (12) |
The general expression for the time-dependent reduced density matrix is13,42
ρij(t) = Tr[![]() | (13) |
![]() | (14) |
The time correlation functions in eqn (13) and (14) are computed using the PLDM method.13,14 Expressing the total Hamiltonian in terms of its nuclear kinetic and total potential energy contributions, Ĥ = + Vnm(
)|n〉〈m|, where |n〉, |m〉 ∈ {{|Dj〉}, {|Ak〉}} are the site basis functions, we use the mapping representation of Meyer–Miller–Stock–Thoss47 to transform the discrete electronic states into continuous variables
, where
. Applying a linearization approximation42 to the nuclear degrees of freedom and keeping the explicit propagation of the electronic degrees of freedom, we arrive at the general PLDM expression13,14
![]() | (15) |
Classical trajectories are used to evaluate the approximate quantum time correlation function in eqn (15). These trajectories are propagated using the equations of motion13
![]() | (16) |
In recent work, we have demonstrated the accuracy of the PLDM method for non-adiabatic reaction dynamics associated with electron transfer.14 To further benchmark the accuracy of the method for EET processes, Fig. 2 presents a comparison of PLDM with numerically exact results for the EET rate associated with excitation transfer from a single donor chromophore to a single acceptor chromophore. These results provide a clear picture of how different quantities control the EET rate, as well as the accuracy of the PLDM method in various regimes. Unless otherwise specified, the excitation energy gap between the donor and acceptor is ε = εD − εA = 100 cm−1, the inter-complex electronic coupling is δ = 5 cm−1, the solvent reorganization energy is λ = λD = λA = 100 cm−1, and the solvent response time is τ = 0.1 ps. The solvent bath are discretized following the description in Section 4.
The rate constant in Fig. 2a is calculated from an exponential fit of the population decay of an excitation on the donor. The initial condition for the excitation is 0 = |D〉〈D|(
b)w. The exciton transfer rate constants in Fig. 2b–d are calculated as the long-time plateau value of the flux-side correlation function, as described in eqn (14). To evaluate the partial-Wigner transformation of the thermal flux operator (e−βĤ
)w, we use the approximation14,42,45
![]() | (17) |
![]() | (18) |
![]() | (19) |
Fig. 2a presents the numerical results of EET rate over a range of solvent reorganization energies, λ, using an inter-complex coupling of δ = 20 cm−1. The EET rate is computed using the PLDM method, as well as with classical Marcus Theory (MT)45 and a Fermi Golden Rule (FGR) expression41,45 that accounts for nuclear quantization (see Appendix A for details). Also included are numerical exact results obtained from the hierarchy equation of motion (HEOM).41 Whereas the FGR and MT descriptions only agree with the exact results for large reorganization energies, PLDM works well over the entire range of λ. Note that a maximal transfer rate is obtained with changing λ; this behavior has been previously discussed in terms of “environment-assisted quantum transport”.12,25,48,49
Fig. 2b presents the numerical results for the EET rate upon varying the inter-complex electronic coupling, δ, and with εD = εA. For small couplings (βδ ≪ 1), the system exhibits non-adiabatic EET dynamics and PLDM agrees with the FGR result. For larger couplings (βδ > 1), PLDM correctly deviates from FGR,14,45 which assumes weak coupling.
Fig. 2c presents PLDM results for the EET rate over a range of values of the energy gap between donor and acceptor states, ε = εD − εA. Here, inter-complex coupling is small (βδ ≪ 1) such that FGR provides an accurate description. The FGR results predict a turnover with maximal rate at ε = λ. Classical MT gives the correct qualitative trend of the EET rate, but it underestimates the rate in the inverted regime (ε > λ). This effect is well understood from the fact that quantized vibrational levels provide additional transfer channels between donor and acceptor states, especially in the inverted regime.50
In Fig. 2d, we explore the effect of the memory time of the bath, τ. By varying τ for the asymmetric model system (ε = 500 cm−1), the EET rate exhibits classical Kramers behavior, as expected.51 For the more interesting symmetric case (ε = 0), the rate increases at the low friction regime (small τ) due to interference effects, as been discussed in previous semi-classical52 and numerically exact51 studies.
The Debye–Drude form of the spectral density described in eqn (11) is discretized into 500 oscillators,53 with frequencies and coupling constants
, where j = 1,…,N and ωm = 20/τ is the cutoff frequency for the oscillators. The solvent reorganization energy for the donor and the acceptor baths is λD = λA = 100 cm−1, and the solvent response time is τ = 0.1 ps. These parameters are chosen to be consistent with natural light harvesting systems.11,41
The time-dependent reduced density matrix ρij(t) is calculated as described in eqn (13), using (0)w = 1/2[|D1〉〈D1| + |D2〉〈D2|](
b)w for the Δ = 0 case and
(0)w = |D1〉〈D1|(
b)w for the Δ = ±200 cm−1 cases where (
b)w are evaluated in the manner of eqn (19).
The EET rate constant is calculated using eqn (14). Although the Wigner transformed thermal flux operator (e−βĤ)w could be sampled exactly from a path-integral Monte-Carlo based procedure,54,55 we use an approximate expression,14,42 as described in eqn (17)–(19).
For the case of dimeric donor and monomeric acceptor, the site and flux operators are ĥ = |A〉〈A| and
![]() | (20) |
![]() | (21) |
Fig. 3a and b presents the real and imaginary components of the intra-complex dynamic coherence 〈D1|(t)|D2〉, as defined in eqn (2). These results demonstrate that the amount of dynamic coherence depends upon the magnitude and sign of the intra-complex electronic coupling, Δ. As the sign of Δ changes from positive to negative values, the dynamic coherence changes correspondingly.
Fig. 3c presents the time-dependence of the exciton population on the donor complex, 〈D1|(t)|D1〉 + 〈D2|
(t)|D2〉. This result indicates that the EET kinetics are also strongly influenced by intra-complex coupling. By varying Δ from 200 to −200 cm−1, the EET rate accelerates by nearly a factor of five.
These results are consistent with recent theoretical results12,21 which illustrate that intra-complex dynamic coherence correlates with the inter-complex EET kinetics. The results might be interpreted to suggest that the EET rate is sensitive to an important role of dynamic coherence; however, in the next sections, we show that if we derive a parameterization of the model that preserves the effects of static coherence in Fig. 3 while largely eliminating the intra-complex dynamic coherence, then the EET kinetics are essentially unchanged.
In order to develop an alternative parameterization of the model that eliminates the intra-complex electronic coupling, we follow previous theoretical work40,56 that compares MC-FRET theory,15–17 a rate theory for energy transfer among weakly-coupled multimeric complexes, in both the site and exciton bases. We begin by expressing MC-FRET rate for the system with the original parameterization
![]() | (22) |
![]() | (23) |
We now switch from the MC-FRET expression in the chromophore site basis (eqn (22) and (23)) to the corresponding expression in terms of the subsystem exciton representation, for which the basis functions {|α〉} and {|γ〉} satisfy
ĤDs|α〉 = εα|α〉 and ĤAs|γ〉 = εγ|γ〉. | (24) |
![]() | (25) |
![]() | (26) |
![]() | (27) |
Comparison of the MC-FRET rate expressions in the site and exciton bases reveals two ways in which static coherence impacts the EET rate. Firstly, static coherence simply shifts the eigenspectrum εα and εγ associate with donor and acceptor complexes, due to the electronic coupling between the sites. Secondly, comparison of eqn (22) and (26) reveals that static coherence leads to the rescaling of the inter-complex coupling δjk by a factor of cjα*ckγ. Both of these effects have been emphasized in previous work.11,12,40
This analysis shows that for any original parameterization of the model system in the site basis, we can prescribe an alternative parameterization also in the site basis such that the effects of static coherence are preserved and intra-complex electronic coupling is strictly eliminated. Table 1 summarizes the way in which this Static Coherence Preserving (SCP) parameterization of the model can be obtained from the original parameterization; more general expressions for the SCP parameterization are provided in Appendix B.
Note that in the SCP parameterization of the model, the bath modes remain diagonally coupled to the single-chromophore states; the Hamiltonian in the SCP parameterization is thus not an exact similarity transform of the Hamiltonian in the original parameterization. The SCP parameterization should be viewed as a procedure to construct a new system with site energies and inter-complex couplings that preserve the effects of static coherence in an original system, while (as is now shown) largely eliminating dynamic coherence.
Finally, we numerically demonstrate that the SCP parametrization largely eliminates intra-complex dynamic coherence during the EET process. In Fig. 4 we present the modulus of dynamic coherence ∥〈D1|(t)|D2〉∥ for the SCP-parametrized model. We consider the SCP parametrization that corresponds to the original parameterization with Δ = −200 cm−1, donor–acceptor energetic gap ε = εD − εA = 500 cm−1, inter-complex electronic coupling δ = 5 cm−1, and a range of values for the intra-complex energy gap ZD = εD1 − εD2. The initial condition for these calculations is 〈D1|
|D1〉 = 〈D2|
|D2〉 = 1/2.
It is clear from the figure that in all cases the intra-complex is dramatically reduced from the levels seen in Fig. 3a and b, and for the special case of degenerate donor chromophores (ZD = 0) and symmetric donor–acceptor coupling (δD1A = δD2A), the intra-complex dynamic coherence vanishes completely. Similar results are obtained for Δ = 200 cm−1 with a range of values for ZD (not shown).
For any original parameterization of the model, we have derived an alternative parametrization that preserves the effects of static coherence while largely (and some cases exactly) eliminating dynamic coherence. In the next section, we will directly compare the EET kinetics for the two different parameterizations to examine the isolated effects of dynamic coherence in the EET process.
Fig. 5 compares the EET dynamics obtained for the two parameterizations with a range of values for the electronic coupling. In each case, the results for the SCP parameterization are obtained from the mapping of the original parameterization described in Table 1; furthermore, the initial condition 〈D1||D1〉 = 〈D2|
|D2〉 = 1/2 employed for the SCP parameterization was chosen such that the initial exciton population matches that which was considered for the original parameterization (Fig. 3).
![]() | ||
Fig. 5 EET dynamics for the model with a dimeric donor and monomeric acceptor, comparing the original parameterization (solid curves) and the SCP parameterization (dashed curves). (a) The exciton population on the donor complex. (b) The normalized flux–flux correlation function, Cff(t)/Cff(0). The solid curves in part (a) are identical to those in Fig. 3c. |
In Fig. 5a, we present the time-dependence of the exciton population on the donor complex, 〈D1|(t)|D1〉 + 〈D2|
(t)|D2〉, for the two different parameterizations; the results from the original parameterization are identical to those presented in Fig. 3c. The similarity of the exponential decay in each case makes clear that the kinetics of EET for the two parameterizations is nearly identical. Fig. 5b presents the normalized flux–flux correlation function Cff(t)/Cff(0) for the EET process, which shows that the transient dynamics associated with inter-complex EET is also nearly identical for the two parameterizations. Taken together, these results indicate the static coherence effects play a major role in the kinetics of EET, whereas dynamic coherence makes a much smaller contribution.
Fig. 6 further illustrates this point by presenting the EET rates for the system discussed in Fig. 5. Indeed, across a wide range of electronic couplings for the original parameterization, the corresponding SCP parameterization exhibits essentially identical rates for inter-complex EET across all regimes.
The turnover behavior seen in Fig. 6 as a function of electronic coupling21 has previously been attributed to a balance between intra-complex dynamic coherence and dissipation to the bath,25,28 which is seemingly inconsistent with the fact that the SCP parameterization (for which dynamic coherence is largely eliminated) still exhibits the turnover behavior. We now investigate the mechanistic origin of this turnover behavior.
The energy-level diagrams at the right of Fig. 6 illustrate how the electronic coupling in the original parameterization leads to a shift of the site energies in the SCP parameterization. For the original parametrization, the donor complex consists of a symmetric dimer (shown at center in red) with a donor–acceptor energy gap of ε = εD − εA, and the two donor chromophores are equally coupled to the acceptor chromophore by δ. In the corresponding SCP parameterization, only one of the two donor chromophores remains coupled to the acceptor chromophore (,
D2A = 0) and the energy gap associated with the donor state with non-zero coupling is shifted to a value of
= εD − εA + Δ. As in the turnover behavior illustrated in Fig. 2c (and also as in the familiar Marcus theory for electron transfer), the relative values of the energy gap
and the total solvent reorganization energy determine whether the system in Fig. 6 is in the normal, activationless, or inverted regimes. Indeed, the observed turnover in Fig. 6 occurs precisely at the Marcus theory prediction of
= λ, for which Δ = −300 cm−1, since εD − εA = 500 cm−1 and λ = 200 cm−1. We thus see that the turnover behavior in Fig. 6 is simply due to a static coherence effect, namely the alignment of the donor and acceptor energy levels that exhibit non-zero coupling, rather than a dynamic coherence effect.
Finally, Fig. 7 explores the competing roles of static and dynamic coherence in the EET kinetics for the model system comprised of a monomeric donor complex and an asymmetric dimeric acceptor complex. In the original parameterization, both acceptor chromophores are coupled to the donor chromophore by δ = 5 cm−1, the electronic coupling among the acceptor chromophores assumes values of Δ = ±200 cm−1, and a range of asymmetries between the acceptor site energies ZA = εA1 − εA2 is explored while keeping the donor–acceptor energetic gap εD − (εA1 + εA2)/2 fixed at a value of 500 cm−1. In addition to the EET rates for the system in the original parameterization (red, blue), the figure also presents the corresponding results for the SCP parameterization that largely eliminates dynamic coherence. It is clear that that the rates for the two parameterizations are nearly identical; as for the previously considered case of a dimeric donor complex and a monomeric acceptor complex (Fig. 6), this version of the model indicates that static coherence plays a dominant role in determining the EET kinetics, whereas dynamic coherence plays only a minimal role.
![]() | (28) |
![]() | (29) |
![]() | (30) |
![]() | (31) |
This journal is © the Owner Societies 2015 |