Detection of Electronic Coherence via Two-Dimensional Electronic Spectroscopy in Condensed Phase

Yu-xiang Weng

Laboratory of Soft Matter Physics,Institute of Physics,Chinese Academy of Sciences,Beijing 100190,China

I.INTRODUCTION

Electronic coherence arising from the superposition of the excited-state wave functions of the atoms or molecules has been gradually shaped from a theoretical concept of quantum to the experimentally detectable reality,thanks to the advent of the ultrafast pulsed lasers.Among the various time-resolved spectroscopic methods[1?4],two-dimensional electronic nonlinear spectroscopy is one of the powerful tools in detection of electronic coherence,which has been widely used to study the coherent electronic energy transfer in a variety of photosynthetic light-harvesting systems[5?10],organic coherent intrachain energy migration in a conjugated polymer[11],exciton valley coherence in monolayer WSe2[12].Interestingly,many quantum coherent effects were observed at room temperature in condensed phase,for the time scales involved in the relaxation and dephasing processes of excited electronic states are in the femtosecond regime in the condensed phase systems[13].In solution,a spectroscopically labeled molecule constantly interacts with its surroundings,leading to fluctuations of its eigenstate energies on a time scale that is comparable to or longer than the“homogeneous” dephasing time,and the spectroscopically “l(fā)abeled” subpopulation would be randomized after excitation due to the system-bath coupling.The time scale of frequency randomization varies,but for typical condensed phase systems it is generally below a few picoseconds,and often it is subpicosecond[14].Thus,in principle the electronic coherence should be detected with a high resolution in both frequency and time domains.However,due to the problem arising from the Fourier transformation limit imposed by the laser pulse,i.e.,a short pulse has a broad spectrum[15],both pump and probe frequency resolved ultrafast relaxation process can hardly be realized by means of conventional spectrometer,hence multi(two)-dimensional electronic spectroscopy has been developed.The basic idea of multi-dimensional spectroscopy is to perform multidimensional optical Fourier transform spectroscopy to reduce the spectral congestion that results from strong system-bath interactions by spreading frequency information on two or more axes.Furthermore,owing to the coherent coupling of the excited and ground state of the molecules,the spectral information regarding excitation and the detection is always entangled.The obvious advantage of 2DES over traditional one dimensional spectroscopy is that it can disentangle congested spectrum by spreading the spectrum out over two dimensions,i.e.,excitation and probe frequencies,and the coupling between resonances can be identi fied by the presence of cross-peaks in the 2D spectrum,which provides an ideal tool for unraveling coupling among resonances and dynamics in complex systems[16].2D spectroscopy has been developed experimentally by Jonas’s group in the near IR region[17,18].In the visible region,2D spec-troscopy was implemented by the groups of Fleming[19],Miller[20],Scholes[6],Ogilvie[21],et al.,and 2DES has become a powerful tool suitable for study of electronic couplings in multichromophore molecular aggregates[7],coupled quantum dots[22],and other systems.2D spectroscopy can trace the spectral evolution caused by energy fluctuation,energy transfer and delocalization with a high temporal resolution by detecting the correlation between the excitation and probe light,it becomes a powerful tool for detection of the electronic,vibronic and vibrational coherences within the molecules generated by coherent excitation.By recording 2D spectra as functions of the“waiting” or“population”time,the pathways by which a complex dynamical system evolves may be observed directly.Finally,because the 2D method is a form of photon echo spectroscopy,the inhomogeneous broadening is removed in the anti-diagonal direction[23].In this review,we confine our scope to the 2D electronic spectroscopy based on the three-photon echo technique.

II.PHYSICAL LIMITATIONS IMPOSED ON COHERENCE DETECTION

In measuring quantum coherence with ultrafast laser pulses,the information expected from the measurement at least should include(i)coherence time which can be as short as tens of femtosecond,and(ii)the multiple energy levels comprising the coherence states.However,it seems that,in principles,there are two physical barriers that have to be overcome.One is the uncertainty principle;the other is the collapse of the wave-packet when subjected to any external perturbation like measurement.For the uncertainty principle,roughly speaking,it states that one cannot simultaneously determine the exact values of a pair of conjugate observables such as momentum and position of a physical system.Conjugate variables are pairs of variables mathematically de fined in such a way that they become Fourier transform duals.The duality relations lead naturally to an uncertainty relation between them,in physics called the Heisenberge uncertainty principle[24].While in matrix mechanics,this corresponds to that pair of observables whose operators do not commute[25].The fact that short pulses probe dynamics and long pulses probe energy levels is indicative of the uncertainty principle,that time and energy resolution are related to each other through the Fourier transform[26].

A.Uncertainty Principle for ΔE and Δt[27]

The Heisenberg uncertainty principle is a relationship between certain types of physical variables like position and momentum.The most well-known expression takes the position and momentum to be the conjugate variables.Another uncertainty relation which is often referenced in discussion of quantum mechanics is the energy-time uncertainty principle,which is directly related to the coherence detection.It is tempting to interpret this energy-time relation as the statement that a system may fluctuate in energy by an arbitrarily large amount over a sufficiently short time scale.This explanation is often given as a description for particleantiparticle production and annihilation,where a particle and its antiparticle appear spontaneously from the vacuum briefly via “borrowed” energy before colliding and returning to vacuum.However,this explanation is not very precise and the given inequality is not so welldefined in quantum mechanics despite the nice physical interpretation.The reason that is not well-defined is because there is no operator in quantum mechanics corresponding to the measurement of time,although the Hamiltonian is the operator corresponding to energy.Nevertheless,there are some ways to make sense of an energy-time uncertainty principle by considering how the measurement of an arbitrary operator changes in time.

Since time is not an operator,it is unclear how time enters quantum mechanics at all.The answer is that time is incorporated into the Schr?dinger equation,where it describes the time rate of change of a wave function.Physically,the passage of time is recorded by noting that certain physical observables are changing over time:for instance,perhaps the position of a particle is changing,which one interprets as motion over time,or the momentum of a particle is changing,which one interprets as accelerating or decelerating over time.

To quantify this statement,consider the Ehrenfest theorem governing the dynamics of the expectation value of an operator in terms of the commutator with the Hamiltonian[27]:

Notably,since the Hamiltonian is the energy operator,σHcorresponds to the uncertainty in energy.

Taking square roots now gives the relation:

De fineσtby the relation:

B.Wave-packet collapse

Ultrashort pulse excitation creates a linear superposition of the eigenstates within the spectrum of the laser rather than a single eigenstate of the system excited by continuous working laser.The linear superposition is such as to create wave packet.The superposition principle states that a system is in all possible states at the same time,until it is measured[27].In fact,such states are very fragile in the presence of dissipation,and would rapidly collapse to eigenstates after measurement,thus destroying the original con figuration leaving no unusual interference features.This is also known as the collapse of the wave packet:when the system is excited with a laser pulse expressed as a pulse electric field,in the optical frequency,α=2π2/ln2,τis the pulse width.The resulting wave function of the system after excitation can be written as a superposition of the eigenstates of the system:

whereanare the coefficients(or amplitudes)of the eigenstatesφn.These coefficients,whose squared moduli represent the populations or occupation probabilities of the individual eigenstates after the pulse is applied,can be determined from the following first-order perturbation-theory formula:

is the transition dipole moment,is the Franck-Condon factor,hωnis the eigen energy value of the individual excited stateφn,andhω0is that of the ground state.Under the case of very short pulse excitation where,one has

where

When considering the time-evolution,one has

This wave packet is a coherent superposition of excited eigenstates.Over time,these states dephase with respect to one another.

III.HOMOGENEOUS AND INHOMOGENEOUS SPECTRAL BROADENING CAUSED BY ELECTRONIC DEPHASING

The following describes the concepts of line broadening and optical dephasing pertaining to an ensemble comprising of two-level systems[28].Considering an ensemble of identical oscillators that do not interact with the surrounding environment or each other,the system is free to evolve over time and will be correlated for long times,limited only by the natural lifetimeτewhich is approximately 10?8s(typically).In the time domain(for time≤τe)the system will evolve as an undamped sinusoid which would approach an infinitesimally narrow line shape in the frequency domain after Fourier transformation(FIG.1,the left panel).If the same ensemble is coupled to a weakly interacting local environment,the local environment will cause the system to depart from a perfect sinusoidal behavior as each two-level system interacts differently with the surrounding environment.This causes each oscillator in the ensemble to become out of phase,a process called optical dephasing.The system now begins to lose the long time correlation and decays to zero asThus,the process of optical dephasing of the oscillators in the ensemble causes damping in the time domain and broadens the absorption and fluorescence spectra.The two major types of dephasing or broadening mechanisms for molecules in solution are homogeneous and inhomogeneous broadening(FIG.1).Two typical cases of oscillators interacting with a liquid solvent will now be considered.In the first scenario the surroundings of each oscillator change rapidly to produce a similar environment for each oscillator.The time evolution of each oscillator appears more stochastic in this case than in the natural lifetime limit due to the interaction between the oscillators and their surroundings changing rapidly with time.The optical dephasing in this case is said to be homogeneous and typically produces an exponential decay in the time-dependent correlation functionC(t)which results in a Lorentzian form of the broadened absorption and fluorescence line shape(FIG.1,the middle panel).Dynamic interactions between a two-level system and its environment can be inelastic,in which an exchange of energy occurs,or elastic,where no energy is exchanged but the frequency is modulated in some way.The optical dephasing from inelastic collisions which contributes to the decay of the time-dependent correlation function is referred to as energy relaxation and is typically characterized by an exponential decay rate ofwhereT1is the population relaxation time.The term pure dephasing is used to describe the effect of elastic interactions on the decay of the correlation function,or equivalently the broadening of the spectrum.The decay of the time-dependent correlation function due to pure dephasing is often characterized by an exponential decay rate ofwhereis the pure dephasing time.The decay rates of energy relaxation and pure dephasing are combined to produce a total homogeneous dephasing rate,

FIG.1 Schematic diagram illustrating the homogeneous and inhomogeneous spectral broadening in time and frequency domain for three hypothetical two-level systems.Left panel:a system of non-interacting harmonic oscillators;middle panel:a homogenously broadened system;right panel:an inhomogenously broadened system.νi(t):sample coordinate trajectories;M(t):time-dependent correlation function;Ia(ω):absorption spectra[29].

In the second hypothetical scenario each oscillator has a slightly different local environment,which does not change significantly over the course of time.In this case,the interaction between the inhomogeneous environments and the oscillators results in a shift in the energy separation away from the unperturbed energyeach frequency shift is different given that the ensemble of oscillators experiences different static environments.As the oscillators evolve over time they still exhibit sinusoidal behavior but oscillate at different frequencies due to their inhomogeneous surroundings.Assuming the distribution of frequencies is Gaussian,the time-dependent correlation function will have a Gaussian decay which results in a Gaussian broadened absorption and fluorescence lineshape when transformed(FIG.1,the right panel).This hypothetical scenario is a limiting case of an inhomogeneously broadened system.

An absorption or fluorescence spectrum of a sample in a real solution may have both homogeneous and inhomogeneous characteristics.One could imagine a hypothetical two-level system between the two limiting cases in which a distribution of inhomogeneous environments arises from extremely slow solvent motions while dynamical interactions from faster solvent motions occur within each environment.Unfortunately,the linear response of system does not distinguish between dephasing processes of different time scales.The advantage of some nonlinear techniques is their ability to selectively eliminate inhomogeneous contributions and extract only the part for which dynamical(homogeneous)interactions dominate.

IV.PRINCIPLE OF 3PPE

The molecular system for nonlinear optical measurement is not a molecule of eigenstates,but a statistical ensemble of the inhomogeneous system,the temporal evolution of the ensemble can be described by the density function theory.Due to the ensemble inhomogeneity,any optical information retrieval occurs in the form of a photon echo[30],where photon echo is one of the first examples of an optical analogue of NMR.

For an ensemble of two-level systems characterized by a ground state|g〉and an excited state|e〉,time independent eigen functions of the Hamiltonian with eigen energies beingEgandEesuperposition states,the wave function superposition state is

It should be noted that the observables corresponding to operators that do not commute with the Hamiltonian,such as the dipole operator,will oscillate over time with frequencyω=ωe?ωg.This phase factor is a unique feature of quantum mechanics.We anticipate here that the quantum beating in such observables is a direct consequence of the superposition character[31].To generalize this treatment for ensembles containing a very large number of particles,the common approach is to introduce the density operator de fined asso that:

where

A.Evolution of density matrix interpreted as photo echoes[32?34]

Considering an optical transition in a two-level system,e.g.,the ground state and an excited state of a dye molecule having two different energiesEgandEe,are involved in the interaction. The external force isF=e·E(r,t),whereeis the charge of the electron,ris the position vector in the three dimensional space,andEis the electric field.The electric field is:E(r,t)=E0cos(ωt).

UsingF=??Vand the electric dipole moment isμ=?e·r,the external potential can be expressed as:

When molecules interact with electrical fieldE(r,t),the ensemble average of〈μ〉would be

Then consider the radiation field whose frequency fulfillsAs a result,the density matrix is reduced to a 2×2 matrix with the elementsρgg,ρge,ρegandρee.

Introducing Rabi frequency

Apparently the solutions ofρge(t)andρeg(t)would contain a time varying factor of the type:What we want is a time-independent,or at least slowly varying solution when the frequency of the perturbing field is close or equal to the transition frequency,i.e.,To realize this,the rotating wave approximation(RWA)is employed by multiplyingρge(t)andρeg(t)with the complex conjugate of their time depending factor:

Then we have

Rewrite the cosine form of the electrical field in the Rabi frequency into a complex form:

The resulting Liouville-von Neumann equation can now be recast in the form

In this model,no relaxation is introduced so far.Now we introduce relaxation by two phenomenological constants.The population of the excited state decays to the ground state at a decay rate written asand the decay rate of the coherence as,this relaxation happens due to phonons in the molecule.is called the irreversible loss off coherence time.T2is a combination of the two relaxation times above according to Eq.(12),which is called as the coherence decay rate.Then we have

B.Bloch vector

The Bloch vector formalism consists in replacing the density matrix elements by three real components(u,v,w)with a population conservation equationρgg+ρee=1.The components are de fined as

or

Notice thatu,v,andware real.wrepresents the population difference anduandvare the real and imaginary parts of the off-diagonal elements in the density matrix,and thus represents the real and imaginary parts of the optical coherence.Alsouandvare sometimes called the in-phase and put-of-phase components of Bloch vector.

The Bloch vector equations of motion thus can be written as

If the relaxation process is disregarded,the optical Bloch equation has the same form as that on NMR by setting

And the coherent evolution part can be represented by

C.Bloch diagram

FIG.2 Bloch diagrams in a rotation frame with w as the rotating frequency.(a)The Bloch vector,S,pointing out with a length of unit in the space(u,v,and w).(b)The Bloch vector starts to smear out in the(u,v)-plane,because of the slightly different excitation energies of the molecules in the ensemble.(c)and(d)How the Bloch vector is turned when a π/2-pulse,respectively,a π-pulse is interacting with the ensemble of molecules,all in the ground level.

FIG.3 How the Bloch vectors change according to the excitation pulses separated in time by τ in 2PPE.The blue arrows indicate the direction of precessing for each individual vector.

The whole Bloch diagram is rotating around thewaxis at the field frequencyω,where the RWA is introduced.When the whole population is in the ground level,the arrow points tow=?1.How do the relaxation processes alter the direction and length of the Bloch vector?The factor 1/T1relaxes the component in thew-direction towards the valuew=?1.The factor 1/T2strives to letu→0 andv→0.The latter leads to a decrease of the length of the Bloch vector.

The detuning factor Δ will causesSto start to smear out like a fan in the(u,v)-plane.This happens since all the different molecules in the ensemble have slightly different excitation energies,due to individual surroundings interaction with the matrix,etc.Because the Bloch diagram is rotating around thew-axis on the field frequencyω,a molecule excited with a slightly different frequency,ωm,will have its own Bloch vector rotating on the detuning frequency Δ=ωm?ωwith respect to the coordinate system.

An electric field turns the Bloch vector around theu-axis by a certain number of degrees depending on its integrated intensity.Aπ/2-pulse causes the Bloch vector to make aπ/2 degree turn around theu-axis,while aπ-pulse turns the Bloch vectorπdegree around theu-axis as shown in FIG.2(c,d)[35].

In photo echo experiments,theπ-orπ/2-pulse excitation is realized by the sequenced laser pulses separated by a time delayτ.FIG.3 shows how Bloch vectors change in a two-photon echo process.The initial state is atw=?1 withSvector pointing at the south pole,and the firstπ/2 pulse excitation coherently turn theSvector arounduaxis byπ/2 to overlap withvaxis,then the decoherence occurs by smearing outSfor each member of the inhomogeneous ensemble from their initial overlapped sate at a rate proportional to the frequency detuning Δ=ωm?ωwithin theuvplane owing to the frequency detuning during the intervalτ.Then the secondπ-pulse comes,which flips the smeared vectorsSarounduaxis byπto overlap with?vaxis while keeping the precessing direction of each individualSsame as that before flipping.It should be noted that after theπ-pulse excitation,the smearing motion becomes reversed as a converging process,which is exactly the reverse process of the decoherence.It is expected,after an evolution time same as the intervalτ,the coherence is resumed and a stimulated emission occurs as the photo echo.Photo echo is a physical vision of the time evolution of the density matrix elements,however in reality,the direction in which the echo is emittedis determined by the wave vectors of the two excitation pulses,and,according to the following equation based on the momentum conservation in the four-wave mixing of nonlinear optics as shown in FIG.4[36]:

FIG.4 Phase-matched optical con figuration in 3PPE.(a)The echo is emitted in the above shown direction.The direction is given by the two excitation pulses wave vectors.(b)Various phase-matched stimulated echo beams on the projection plane[34].

If the temporal order of the two writing beams is changed,the echo is emitted in the direction implied as a dashed arrow above,which provides a convenient way to acquire the rephasing and non-rephaisng signals in the 2DES.

Although in the time sequence,there appear two excitation pulses called two-pulse photo echo.However,in terms of nonlinear wave mixing,it consists of three beams,i.e.,one beam 1 and two degenerated beam 2 having the same propagating direction and a zero respective delay time.Therefore,by nature,2PPE is a simpli fied version of three-photo echo emission.

The three-pulse photon echo(3PPE)experiment involves three pulses with time sequence asτ,T,andtas shown in FIG.3,whereT=0.Like 2PPE,the firstπ/2 pulse again creates a macroscopic polarization,then at delay timeτis aπ/2-pulse rather than aπ-pulse.This secondπ/2-pulse creates population in either the ground or excited states in which the coherence is stored[37].A thirdπ/2-pulse then interacts with the sample at delay timeTand causes the system to rephase the inhomogeneous component of the polarization decay to produce a photon echo at timet.The echo signal is measured in the phase matching directionand is maximum for threeπ/2-pulses[28].

FIG.5 Illustration of the evolution for the density matrix elements of two-level system in 3PPE.During τ,the system evolves in a coherence.The system is stored in a population on the ground or excited state during T,and the final interaction induces a second coherence,which radiates a stimulated photo echo signal in the phase match direction[14].

D.3PPE in view of density matrix

FIG.5 shows the stimulated photon echo pulse sequence along with the 2×2 density matrix for a twolevel system[14].The sequence has three time intervals,τ,T,andt,where due to interactions with the three fieldsE1,E2,andE3,the system evolves alternately in coherences(ρjk)and populations(ρjj).An“echo” is the phase conjugation between the two coherence evolution periods.Duringτthe density isρgeand duringtit isρeg,thus effectively reversing the sign of time in the field-free Green function responsible for the system’s evolution during the second coherence interval.The lower set of density matrices shown duringτandtindicate an alternative pathway where the system returns to the ground state following the second excitation pulse.Both the excited state and ground state paths are required to evaluate the system response to the applied fields.The fact that echo signalρgeis a phase-conjugated term underlies the principle for 3PPE as a quantum data storage technique in time sequence.

In summary,the conventional interpretation of a photon echo experiment represents the state of the system using a pseudo spin vector on the Bloch sphere.By transforming to the rotating frame,each member of the inhomogeneous ensemble precesses away from the origin at a rate proportional to the difference between the speci fic frequencyωmand the frame rotation frequencyω,i.e., Δ.WhenT=0,the second and third pulses flip the vectors leading to the phase conjugation.The echo is emitted when the vectors precess freely back to the origin,hence the gradual increase and decrease of the echo signal can be detected.ChoosingT＞0 permits incoherent relaxation to occur before the refocusing pulse,providing access to lifetime measurements and observation of spectral diffusion processes due to microscopic site randomization.Indeed,it is precisely theT-dependence that provides much of the chemical information in condensed phase photon echo experiments.The homogenous and inhomogeneous decay process of the system in response to respective onepulse,two-pulses and three-pulses excitation is depicted schematically in FIG.6[28].

可以看出，對目標(biāo)可靠度較大、非線性程度較高、預(yù)測誤差較大的功能函數(shù)，構(gòu)建較大的序列采樣區(qū)域以提高局部Kriging擬合精度；反之，構(gòu)建較小的序列采樣區(qū)域以提高采樣效率。

FIG.6 Schematic diagram of coherent techniques involving one,two and three pulses.The electric field pulses(blue solid peaks)excite the sample.The resulting coherent signal is shown for homogeneously broadened transitions(red dashed line)and for inhomogenously broadened transitions(green dotted and dashed line).(a)A single excitation pulse E1with a wave vector.Homogeneously broadened transitions lead to an exponential decay with dephasing time ofwhile interference of an inhomogenously distribution of the resonant frequencies causes a faster decay.(b)Two excitation pulses E1and E2with wave vectors ofandgives an echo in the direction 2.For inhomogeneously broadened transitions,the echo is emitted at t=τ.The echo amplitude measured as a function of τ allows determination of the homogeneously dephasing time constant(c)When three excitation pulses E1,E2,and E3are applied,the coherent signal emitted in the direction allows determination of the dephasing timeand the energy relaxation time T1[38].

E.3PPE as time-domain data storage

3PPE has been shown as potential technique for timedomain data storage[39]and word by word logic processing[40]possibly used in quantum computation.The idea of photon-echo memory is to perform spectral hole-burning memory[37]in the time domain by utilizing the temporal interference effect of two optical pulses.From the viewpoint of memory,the three excitation pulses are renamed as WRITE,DATA and READ pulses and,rather than being a delta-function pulse,the DATA pulse carries some kind of temporal information as shown in FIG.7.Calculation shows that this echo signal simply replicates the shape of the DATA pulse.In terms of memory,the WRITE pulse as a trigger and the DATA pulse with temporal information complete the storage of the data into the medium.The readout process starts by applying a trigger called a READ pulse,and the echo then appears as the OUTPUT DATA.The DATA can be understood in spectral region,since pulse 1 and pulse 2 would create spectral interference which provides spectral selectivity by tuning the relative delayτ.

FIG.7 Pulse sequence and expected signal in the stimulated photon echo(upper panel)and the simulated photon-echo memory(lower panel)[39].

V.THIRD ORDER NONLINEAR POLARIZABILITY THEROY FOR 3PPE

3PPE is one of implementations of 2DES[41].In experiment,three ultrashort pulses are arranged in a unique temporal sequence to excite the sample,creating a 3rd order polarizationP(3)(τ,T,t)which can be expressed as the convolution of the nonlinear response functionR(3)(τ,T,t)with the respective three field envelope

where the third order nonlinear response function is expressed as a function of the dipole moment and the density matrix of the unperturbed system

FIG.8 Pulse sequence in 3PPE.LO:local oscillator for heterodyne detection.

FIG.9 Schematic illustration of the information contained in absorptive two-dimensional Fourier transform spectra.The observed spectral range shown in the unshaded region is determined by the excitation and detection laser bandwidths.(a)The T=0 correlation spectrum reveals homogeneous and inhomogeneous line widths,excitonic coupling,and excited state absorption(ESA)features.(b)At T＞0,the broadening of peaks in the anti-diagonal direction reflects spectral diffusion.The growth of cross peaks indicates energy transfer.The emergence of an entirely new peak at later waiting times,such as the B′D cross peak in panel b,represents the formation of a new product species,populated upon excitation of B′,that absorbs at D(cited from Ref.[41]).

Apparently,P(3)(τ,T,t)contains three interacting electric fields,when the laser pulsesE1(t),E2(t),andE3(t)are shorter than the time separation between them(semi-impulsive limit),they do not overlap in time which gives rise to a strict time ordering.In that case,the first interactionμ(t0)originates from the pulseE1(t),μ1(t)fromE2(t),and so on.This reduces the number of terms inP(3)(τ,T,t)to 32 corresponding to all possible contributions from three incoming electric fields to polarization.Each of these terms can be associated with a distinct time evolution of the density matrix,known as Liouville pathways[34].In spectroscopy these pathways are usually graphically visualized by means of diagrams capable of highlighting the temporal sequence of the field interactions and the transitions promoted in the systems by such interactions,while the most famous are the double-sided Feynman diagrams.

The fourth homologous laser pulse serves as a local oscillator to detect the emitted signal by means of optical heterodyne detection as shown in FIG.8[42].The detected signal by the spectrometerS(τ,T,ωt)is related to the 3rd polarizationP(3)(τ,T,t)via one-dimensional Fourier transformation on the detection periodtat the phase-matched condition

where the third-order polarization is related to the measured signal field byHere the spectrometer acts as a hardware of performing Fourier transformation along the detection timetresulting in a probed spectrum spanned alongωt,leaving the pump frequency to be resolved from the time spaceτ.Further performing the 2D-FT of the experimentally measured nonlinear optical response function along the evolution timeτ,one has the 2D-FT spectrum at a fixedT[43?45]:

The 2D spectrum at a particularTinvolves frequency dimensionsωτandωt,corresponding to the evolution timeτ(coherence time)and the detection timetrespectively.In 2D-FT spectroscopy,two different phasematched signals are typically recorded:the rephasing and non-rephasing signals in thek2?k1+k3andk1?k2+k3directions,respectively. When summed,these signals yield the absorptive 2D-FT spectrum,free from broadening refractive contributions.Another signal,with the phase-matching directionk1+k2?k3,yields information about double-quantum coherences[43].Typical 2DES is shown in FIG.9.

FIG.10 Feynman diagrams.(a)Diagrammatic representation of the interaction of an electromagnetic field with a material system,the interactions at t1and t2represent absorption and the interactions at t3and t4emissions at the ket and bra side respectively of the density matrix for a two-level system with|0〉and|1〉,time increases from bottom to top.The very last interaction(usually emission)corresponds to the coherently induced polarization in the ensemble.(b),(c)and(d)Feynman diagrams for excitonic system with|0〉,|α〉and|β〉as the ground, first,and second electronic excited state respectively showing non-oscillation and oscillation properties.

The time resolution of dynamics over the waiting time,in both pump-probe and 2D experiments,is dependent on the duration of the pump pulse,with shorter pulses giving higher time resolution.The commensurately larger bandwidth of a shorter pulse leads to a lower spectral resolution as restricted by the uncertainty principle.2D-FT spectroscopy use the Fourier transform methodology by converting time to the frequency,thus the restriction imposed by the uncertainty principle in 2D spectroscopy is circumvented[43],and the result has both high temporal and spectral resolution,limited only by the signal-to-noise ratio.However,the wave-packet collapse to the eigenstates as reflected in both excitation and detection frequency axis at any fixed waiting time,i.e.,coherence cannot be informed from the corresponding 2D spectra.

VI.ELECTRONIC COHERENCE(QUANTUM BEAT)ALONG WAITING TIME AND DOUBLE-SIDED FEYNMAN DIAGRAM

A.Double-sided Feynman diagram[31]

Double-sided Feynman diagrams describe the evolution of the density matrix in the presence of several coherent fields in time sequence[46,47]by depicting interaction sequences between external perturbations such as electromagnetic fields and ensemble of molecules.One advantage of using Feynman diagrams is to physically capture the most important interaction paths without referring to the actual form of either the density matrix or the Hamiltonians describing the ensemble.

The drawing of double-sided Feynman diagrams derives from taking separately each term of the commutator expansion(Eq.(47)and Eq.(48))to create the corresponding diagram,using specific rules:(i)two vertical lines represent the time evolution of the ket(left line)and bra(right line)of the density matrix,with time running from the bottom to the top.(ii)The interactions between the density matrix and the electric field are indicated by arrows acting either on the bra or ket side pointing towards or away from the system corresponding to absorption or emission.(iii)The ket and bra sides of the diagram are the complex conjugate of each other and the last emission is always from the ket side,by convention.(iv)The last interaction must end in a population state.(v)An arrow pointing to the right represents an electric field with e?iωt+ikr(+k),an arrow pointing to the left represents an electric field with e+iωt?ikr(?k).The emitted light,i.e.the last interaction,has a frequency and wavevector which is the sum of the input frequencies and wavevectors,in accord to the energy conservation and phase matching condition respectively.

Considering a molecular system with|0〉,|α〉,and|β〉as the electronic ground, first and second electronic states respectively,with further assumption that the energy difference between|α〉and|β〉is smaller than the spectral bandwidth of the excitation laser pulses,i.e.,|α|and|β|can be simultaneously excited.The first and the second electric field interactions create an absorption resulting in a population state.Therefore,the possible population states can be|α〉〈α|,|β〉〈β|,and|α〉〈β|together with their conjugated complexes.Obviously,|α〉〈α|and|β〉〈β|are the excited eigenstates while|α〉〈β|is the electronic coherent state,their corresponding double-sided Feynman diagrams are shown in FIG.10.In measuring of the echo signal by scanning the waiting timeT,static population decay kinetics are observed for either|α〉〈α|or|β〉〈β|only.However,for the coherent excited state,except for the static population relaxation,there is an additional coherent oscillation term as exp[±(ωα?ωβ)T]de fining the quantum beating signal overriding on the population relaxation envelope.An example of observation of the electronic quantum coherence is illustrated in FIG.11 and FIG.12 respectively.Therefore,whether a speci fic Li-ouville pathway is represented by the Feynman diagram leads to coherent oscillation can be easily determined by the generated population state after interaction with the first two pulses.Consequently,in 2DES the coherence is observed at the energy difference between the two eigenstates in the excited wave packet along the waiting timeTaxis,and the restriction arising from the collapse of the wave packet is thus circumvented.

FIG.11 Double-sided Feynman diagrams describing the dominant contributions to electronic beats in 2DES spectra at diagonal and off-diagonal positions.Not all Feynman diagrams are included,the focus being on those important for the electronic beats.(a)and(b)for rephasing pathways while(c)and(d)for non-rephasing pathways(with modi fication to match the convention of Feynman diagram)[5].

FIG.12 Typical example of 2DES showing the quantum beat signals.(a)The 2DES spectrum depicts the signal amplitude on an arcsinh scale(color scale,arbitrary units)plotted as the excitation frequency ωτ and detection frequency ωtat a population delay time of T=100 fs for the antenna complex PC45 recorded at room temperature.(b)Amplitude of the cross-peaks at positions indicated in the 2D map with open squares[(ωτ,ωt)=(2.185,2.06)eV in black and(ωτ,ωt)=(2.06,2.185)eV in red]as a function of time T.The dashed lines interpolate the data points(solid circles).The solid line is a fit to a sum of damped sine functions revealing the quantum beats along T.For further information,see Refs.[6,48].

B.Electronic versus vibronic coherence

Two-dimensional spectroscopy has the capability to reveal the quantum coherence as the oscillatory behavior of the excitation dynamics of molecular systems.However,the situation is complicated regarding the exact origin of the observed coherent phenomena,i.e.,what is actually being observed:excitonic or vibrational wave-packet motion or evidence of quantum transport.Especially,in some molecules and their aggregates,electronic transitions are coupled to various intra-and intermolecular vibrational modes,with the magnitudes of the resonant couplingsJin excitonic aggregates being in the same range as those of vibrational energies.Thus,vibronic and excitonic systems show considerable spectroscopic similarities,and the presence of electronic and/or vibrational beats in the 2DES spectrum is ex-pected.Indeed,similar spectral beats originating entirely from a high-energy vibrational wave-packet motion have been studied theoretically for weak electronphonon coupling[49]and have been observed experimentally[50].Number of groups have contributed to addressing this important issue in both theory and experiments[49?52].

FIG.13 Fourier maps and the Feynman diagrams for the excitonic dimer.(a)Energy levels.(b)and(c)Fourier maps for rephasing and non-rephasing con figuration respectively.(d)Feynman diagrams leading to oscillation.ω:positive frequency;?ω:negative frequency;GB:ground-state bleaching;SE:stimulated emission.ESA:excited-state absorption.Black symbols for non-oscillating pathways,colored symbols for oscillating pathways.

Two generic model systems which exhibit distinct internal coherent dynamics have been compared[53].One model system of pure electronic transition which shows similar spectroscopic properties but has completely different coherent internal dynamics without vibrations,is an excitonic dimer(ED)[54,55].It consists of two two-level chromophores(sites)with identical transition energiesε.The two sites are coupled by the inter-site resonance couplingJ.As a result,the ED has one ground state|g〉,two single-exciton states|e1〉and|e2〉with energiesrespectively,and a single double-exciton state|f〉with energy,where Δbindingis the bi-exciton binding energy.The lowest transition frequency is denoted asthe transition frequency difference between|e1〉and|e2〉isthe corresponding energy diagram,and 2DES pattern constructed from all the possible Liouville paths revealing oscillation(colored)and non-oscillation(black)in rephasing and nonrephasing 2DES are shown in FIG.13(a),(b),and(c)respectively.

The other model is the vibronic system of an isolated molecular electronic excitation represented by two electronic states,|g〉and|e〉,which are coupled to a onedimensional nuclear coordinateqwith a vibrational frequency energy ofhν,assuming thatνkeeps the same for all the vibrational excited states either in the electronic ground or excited state with its magnitude equal to that of Δω.This model is denoted as a displaced oscillator(DO)system(FIG.14).Its energy diagram and 2DES pattern constructed from all the possible Liouville paths revealing oscillation and non-oscillation in 2DES are shown in FIG.14(a),(b)and(c)respectively.

FIG.14 Fourier map and the Feynman diagrams for the displaced oscillator model with coupling of a single vibrational mode.(a)Energy levels for displaced oscillator;(b)Fourier map for rephasing con figuration.(c)Feynman diagrams leading to oscillation in rephasing con figuration. ν:positive frequency,?ν:negative frequency,GB:ground-state bleaching,SE:stimulated emission.Black symbols for non-oscillating pathways,colored symbols for oscillating pathways.

To reveal oscillatory contributions in the ED and DO systems,all contributions into either oscillatory or static can be compared,shown in FIG.13(b)and 14(b).As a function ofT,the DO system has 8 oscillatory and 8 static con figurations,which organize into six peaks,while the ED system has only 4 oscillatory and 8 static contributions which give four peaks.The net result is that the diagonal peaks in the rephasing and crosspeaks in the non-rephasing signals are non-oscillatory in the ED,while all peaks except the upper diagonal peak in rephasing signals are oscillatory in the DO.Thus signi ficant differences in oscillatory peaks between ED and DO systems are found.The typical oscillatory pattern of 2DES spectrum appearing in the rephasing con figuration of DO model with a single vibrational mode exhibits as a “chair”,and is commonly referred as the“chair” pattern[56,48].

Generally,the 2DES spectra are presented as the slices of pump and probe frequency projection at a fixed waiting time,and the coherence is given as oscillating amplitude at a certain point on the projection. To study the coherent oscillations,it is more straightforward to view the 2DES spectra along the beating frequency ΔωTinstead ofTby performing further Fourier transformation alongT,leading to the so-called oscillation Fourier maps.They are constructed by applying the Fourier transformation on the real part of the rephasing spectrumSR(ωτ,T,ωt)with respect to the delay timeT[48,51,57]:

In this way,at a given oscillation,all the contributions from the non-oscillation pathways and those of oscillation at a frequency other than the given one would be filtered,providing a powerful method to look into the coherent system having multiple beating frequencies. It should be noted that the Fourier maps for oscillation/non-oscillation contributions in 2DES spectra are similar whether it is projected at a fixed waitingTor at a fixed beating frequency ΔωT,while only the latter can be used to trace the contribution from a given single beating frequency in the 2DES.

C.Single-vibrational mode versus multi-vibrational mode coupling in vibronic transition

FIG.15(a)Multi-vibrational mode-coupled displaced oscillator model with electronic transition energy ωegand two vibrational modes having a speci fic transition frequency of ν and ? respectively;(b)Rephasing Fourier map for the vibrational mode-coupled displaced model at the oscillation frequency ΔωT=ν displaying the uncoupled pattern(blue ellipses)and the two-vibrational mode coupled pattern(red ellipses).The squares represent the beating appearing in the ground-state bleaching(GB)signals while the circles in the stimulated emission(SE)signals.The filled symbols for the beating with+ν frequency while the open ones for the?ν frequency.(c)Feynman diagrams of seven typical coupling cases having a beating frequency of±ν.The subscript indicates ladder of the mode,and the superscript indicates the vibrational mode.

Single vibrational-mode coupling to the vibronic transition is the simplest model.When multi-mode coupling is considered,the Hamiltonian as well as the corresponding sets of wave functions for a single-vibrational mode displaced oscillator coupled to the electronic states[58?60]needs to be extended to the multi-mode case[61].Considering a multi-mode coupled displaced oscillator in a chromophore molecule with only two electronic states,i.e.,the ground state|g〉and the excited state|e〉,coupled to a collection of harmonic oscillators within the adiabatic approximation,the total Hamiltonian can be expressed as

where

Hereirepresents theith vibrational mode,piandqiare the corresponding momentum and the position operator respectively.μgi(μei)is the reduced mass of the ground(excited)state,ωgi(ωei)is the frequency of the oscillator on the ground(excited)state,diis the displacement of the origin on the excited-state potential relative to the ground state along the coordinateqi,andωegis the energy difference between the minima of the two electronic states.An assumption is made that the nuclear potential surfaces are well described as harmonic and that the excited energy surface is allowed to differ in curvature,giving rise to that the values ofωeiandωgican be different[62,63].The oscillator strength for the vibronic transition is determined by the Frank-Condon factor while each vibrational mode contributes with a weight given by its Huang-Rhys factor in terms of difference of the mode displacement of the initial and final nuclear con figurations,i.e.,the extent for the wavefuction overlap of nuclear part between the initial and the final states[64?66].

The simplest model of multi-mode coupling in displaced oscillator is the two vibtrational modes system with their vibrational frequencies denoted asνand ? respectively,which is schematically shown in FIG.15(a),while the expected Fourier map for a given oscillation frequencyνwith a coupled mode ? is in FIG.15(b)and the corresponding four Feynman diagrams leading to the mode coupling is in FIG.15(c).The “chair” pattern indicated by the blue ellipses in FIG.15(b)re flects the single vibrational mode coupling in DO model at the observing frequencyν,i.e.,the Fourier map is the projection at fixed value ofν.Except for the expected chair pattern for the single mode,there are also twomode coupled patterns appearing at the location of the rephasing frequency atωeg?ν+? excited atωeg,ωeg+νandωeg+?;rephasing frequency atωeg+ν?? excited atω0;rephasing frequency atωeg?νexcited atωeg+?;rephasing frequency atωeg?? excited atωeg+ν.These indicate that the observed beating frequencyνcan be excited at a vibrational frequency corresponding either toωeg+νor toωeg+?.The fact suggests that the vibrational frequenciesνand ? are coupled to each other as shown in the red ellipses.

Obviously,since the detected oscillation signals originate from all the possible Liouville pathways including the excitonic and single-mode or multi-mode vibrationally coupled coherences,new methods either of apparatus or data analysis are called for clear distinguishing the electronic coherence from those of vibronic.It has been shown that the polarization modulated[67,68]2DES can be an effective way toward this purpose as successfully demonstrated in a recent experiment,where the oscillation maps extracted from the doublecross polarized schemes(relative polarizations ofπ/4,?π/4,π/2,and 0 for laser pulses 1 to 4)strongly suppresses signals of intramolecular vibrational coherences produced by the Franck-Condon excitation of vibrational wave packets[69].On the other hand,2D singletime-domain exciton perturbation spectroscopy(2DSTEPS)and 2D time-resolved interexciton perturbation spectroscopy(2D-TRIPS)are the new approaches to extract information from 2DES spectra.These timedomain optical spectroscopies probe vibrational motion by examining correlated dynamics of each exciton’s spectral motion. 2D-STEPS measures how excitedstate nuclear motions systematically perturb individually excitonic transitions after excitation,whereas 2DTRIPS probes how nuclear motions spawned by exciting one exciton affect other unoccupied excitonic states[70].

VII.CONCLUSION

Quantum coherence manifests itself not only as a concept but also as a reality in many aspects.An emerging field of quantum computation involves coherence control and the quantum information storage,while 3PPE provides possible quantum information storage in time sequence.While in natural and arti ficial photosynthetic systems,it also has been suggested that coherent energy transport could be more efficient than classical transport.Development of advanced theories would provide more applicable feasibility while the new advances in the experimental methods and designs of novel quantum coherent systems would speed up the application of quantum coherence.Though quantum coherence was considered fragile,recent evidence of coherence in chemical and biological systems suggests that the phenomena are robust and can survive in the face of disorder and noise,suggesting that coherence can be used in complex chemical systems[71].

VIII.ACKNOWLEDGMENTS

This work was supported by the National Natural Science Foundation of China(No.21227003,No.21433014,No.11721404). Mr. Xuan Leng and Rui-dan Zhu are kindly acknowledged for proof-reading and Figure preparation.

[1]W.R.Lambert,P.M.Felker,and A.H.Zewail,J.Chem.Phys.75,5958(1981).

[2]A.H.Zewail,J.Phys.Chem.97,12427(1993).

[3]H.Katsuki,H.Chiba,B.Girard,C.Meier,and K.Ohmori,Science311,1589(2006).

[4]G.D.Reid and K.Wynne,Ultrafast Laser Technology and Spectroscopy,in Encyclopedia of Analytical Chemistry,Chichester:John Wiley&Sons Ltd.,(2000).

[5]G.S.Engel,T.R.Calhoun,E.L.Read,T.K.Ahn,T.Man?al,Y.C.Cheng,R.E.Blankenship,and G.R.Fleming,Nature446,782(2007).

[6]E.Collini,C.Y.Wong,K.E.Wilk,P.M.G.Curmi,P.Brumer,and G.D.Scholes,Nature463,644(2010).

[7]G.Panitchayangkoon,D.Hayes,K.A.Fransted,J.R.Caram,E.Harel,J.Wen,R.E.Blankenship,and G.S.Engel,Proc.Natl.Acad.Sci.USA107,12766(2010).

[8]E.Romero,V.I.Novoderezhkin,and R.van Grondelle,Nature543,355(2017).

[9]Y.X.Weng,Physics(in Chinese)39,331(2010).

[10]Y.X.Weng,Physics(in Chinese)36,817(2007).

[11]E.Collini and G.D.Scholes,Science323,369(2009).

[12]K.Hao,G.Moody,F.Wu,C.K.Dass,L.Xu,C.H.Chen,L.Sun,M.Y.Li,L.J.Li,A.H.MacDonald,and X.Li,Nature Phys.12,677(2016).

[13]M.L.Cowan,J.P.Ogilvie,and R.J.D.Miller,Chem.Phys.Lett.386,184(2004).

[14]J.P.Ogilvie and K.J.Kubarych,in:E.Arimondo,P.R.Berman,C.C.Lin(Eds.),Adv.At.Mol.Opt.Phys.57,249(2009).

[15]S.M.G.Faeder and D.M.Jonas,J.Phys.Chem.A103,10489(1999).

[16]T.H.Zhang,C.N.Borca,X.Q.Li,S.T.Cundiff,Optics Express13,7432(2005).

[17]J.D.Hybl,A.W.Albrecht,S.M.Gallagher Faeder,and D.M.Jonas,Chem.Phys.Lett.297,307(1998).

[18]J.D.Hybl,A.A.Ferro,and D.M.Jonas,J.Chem.Phys.115,6606(2001).

[19]T.Brixner,J.Stenger,H.M.Vaswani,M.Cho,R.E.Blankenship,and G.R.Fleming,Nature434,625(2005).

[20]V.I.Prokhorenko,A.Halpin,and R.J.D.Miller,Opt.Express17,9764(2009).

[21]P.F.Tekavec,J.A.Myers,K.L.M.Lewis,and J.P.Ogilvie,Opt.Lett.34,1390(2009).

[22]S.T.Cundiff,A.D.Bristow,M.Siemens,H.Li,G.Moody,D.Karaiskaj,X.Dai,and T.Zhang,IEEE J.Sel.Top.Quant.18,318(2012).

[23]M.Cho,Chem.Rev.108,1331(2008).

[24]D.Sen,Current Science107,203(2014).

[25]U.Fano,Rev.Modern Phys.29,74(1957).

[26]M.G.Dantus,P.Gross,Encyclopedia of Applied Physics,22,431(1998).

[27]D.J.Griffiths,Introduction to Quantum Mechanics,Cambridge University Press,(2016).

[28]C.N.Lincoln,Ph.D Thesis,Swinburne University of Technology,(2007).

[29]R.B.Williams,Ph.D thesis,Cornell University(2001).

[30]S.V.Poltavtsev,M.Salewski,Y.V.Kapitonov,I.A.Yugova,I.A.Akimov,C.Schneider,M.Kamp,S.Hoefling,D.R.Yakovlev,A.V.Kavokin,and M.Bayer,Phys.Rev.B93,121304(R)(2016).

[31]E.Collini,Chem.Soc.Rev.42,4932(2013).

[32]S.Mukamel,Principles of Nonlinear Optical Spectroscopy,Oxford University Press,(1999).

[33]C.Nilsson,Lund Reports in Atomic Physics,(1997).

[34]P.Hamm and M.Zanni,Concepts and Methods of 2D Infrared Spectroscopy,Cambridge University Press,(2011).

[35]J.H.Eberly,Optical Resonance and Two-level Atoms,Dover:(1987).

[36]D.Polli,I.Rivalta,A.Nenov,O.Weingart,M.Garavelli,and G.Cerullo,Photochem.Photobiol.Sci.14,213(2015).

[37]J.T.Fourkas,Annu.Rev.Phys.Chem.53,17(2002).

[38]W.Zinth and W.Kaiser,Ultrashort Laser Pulses,Springer,(1988).

[39]M.Mitsunaga,Optical and Quantum Electronics24,1137(1992).

[40]S.Kroll and U.Elman,Opt.Lett.18,1834(1993).

[41]F.D.Fuller,J.P.Ogilvie,in:M.A.Johnson,T.J.Martinez(Eds.),Annu.Rev.Phys.Chem.66,667(2015).

[42]M.H.Cho,T.Brixner,I.Stiopkin,H.Vaswani,and G.R.Fleming,J.Chin.Chem.Soc.53,15(2006).

[43]D.M.Jonas,Annu.Rev.Phys.Chem.54,425(2003).

[44]T.Brixner,T.Man?al,I.V.Stiopkin,and G.R.Fleming,J.Chem.Phys.121,4221(2004).

[45]Y.X.Weng,H.Chen,Ultrafast Spectroscopy:Principles and Techniques(in Chinese),Beijing:Chemical Industry Press,(2013).

[46]K.Wynne,M.Muller,D.Brandt,and J.D.W.Vanvoorst,Chem.Phys.125,211(1988).

[47]B.D.Fainberg and V.Narbaev,J.Chem.Phys.113,8113(2000).

[48]D.B.Turner,R.Dinshaw,K.K.Lee,M.S.Belsley,K.E.Wilk,P.M.G.Curmi,and G.D.Scholes,Phys.Chem.Chem.Phys.14,4857(2012).

[49]A.Nemeth,F.Milota,T.Mancal,V.Lukes,J.Hauer,H.F.Kauffmann,and J.Sperling,J.Chem.Phys.132,184514(2010).

[50]N.Christensson,H.F.Kauffmann,T.Pullerits,and T.Mancal,J.Phys.Chem.B116,7449(2012).

[51]V.Butkus,D.Zigmantas,D.Abramavicius,and L.Valkunas,Chem.Phys.Lett.587,93(2013).

[52]T.Mancal,A.Nemeth,F.Milota,V.Lukes,H.F.Kauffmann,and J.Sperling,J.Chem.Phys.132,184515(2010).

[53]V.Butkus,D.Zigmantas,L.Valkunas,and D.Abramavicius,Chem.Phys.Lett.545,40(2012).

[54]A.Sakurai and Y.Tanimura,J.Phys.Chem.A115,4009(2011).

[55]J.Xu,R.Xu,D.Abramavicius,H.Zhang,and Y.Yan,Chin.J.Chem.Phys.24,497(2011).

[56]S.S.Senlik,V.R.Policht,and J.P.Ogivie,J.Phys.Chem.Lett.6,2413(2015).

[57]T.R.Calhoun,N.S.Ginsberg,G.S.Schlau-Cohen,Y.C.Cheng,M.Ballottari,R.Bassi,and G.R.Fleming,J.Phys.Chem.B113,16291(2009).

[58]V.Butkus,L.Valkunas,and D.Abramavicius,J.Chem.Phys.140,034306(2014).

[59]E.Basinskaite,V.Butkus,D.Abramavicius,and L.Valkunas,Photosyn.Res.121,95(2014).

[60]Y.P.Wang,Z.Wang,and Y.X.Weng,Chin.Sci.Bull.(Chinese Version)57,2895(2012).

[61]A.F.Fidler and G.S.Engel,J.Phys.Chem.A117,9444(2013).

[62]S.Yue,Z.Wang,X.Leng,R.D.Zhu,H.L.Chen,and Y.X.Weng,Chem.Phys.Lett.683,591(2017).

[63]S.Yue,Z.Wang,X.C.He,G.B.Zhu,and Y.X.Weng,Chin.J.Chem.Phys.28,509(2015).

[64]K.Huang and A.Rhys,Proc.R.Soc.Lond.A204,406(1950).

[65]J.Pieper,J.Voigt,and G.J.Small,J.Phys.Chem.B103,2319(1999).

[66]M.Ratsep,M.Pajusalu,and A.Freiberg,Chem.Phys.Lett.479,140(2009).

[67]G.S.Schlau-Cohen,A.Ishizaki,T.R.Calhoun,N.S.Ginsberg,M.Ballottari,R.Bassi,and G.R.Fleming,Nature Chem.4,389(2012).

[68]M.T.Zanni,N.H.Ge,Y.S.Kim,and R.M.Hochstrasser,Proc.Natl.Acad.Sci.USA98,11265(2001).

[69]D.Palecek,P.Edlund,S.Westenhoff,and D.Zigmantas,Sci.Advances3,e1603141(2017).

[70]B.S.Rolczynski,H.Zheng,V.P.Singh,P.Navotnaya,A.R.Ginzburg,J.R.Caram,K.Ashraf,A.T.Gardiner,S.H.Yeh,S.Kais,R.J.Cogdell,and G.S.Engel,Chem.4,138(2018).

[71]G.D.Scholes,G.R.Fleming,L.X.Chen,A.Aspuru-Guzik,A.Buchleitner,D.F.Coker,G.S.Engel,R.van Grondelle,A.Ishizaki,D.M.Jonas,J.S.Lundeen,J.K.McCusker,S.Mukamel,J.P.Ogilvie,A.Olaya-Castro,M.A.Ratner,F.C.Spano,K.B.Whaley,and X.Zhu,Nature543,647(2017).