Theory of coherent two-dimensional vibrational spectroscopy

Jansen, Thomas la Cour; Saito, Shinji; Jeon, Jonggu; Cho, Minhaeng

doi:10.1063/1.5083966

Two-dimensional (2D) vibrational spectroscopy has emerged as one of the most important experimental techniques useful to study the molecular structure and dynamics in condensed phases. Theory and computation have also played essential and integral roles in its development through the nonlinear optical response theory and computational methods such as molecular dynamics (MD) simulations and electronic structure calculations. In this article, we present the fundamental theory of coherent 2D vibrational spectroscopy and describe computational approaches to simulate the 2D vibrational spectra. The classical approximation to the quantum mechanical nonlinear response function is invoked from the outset. It is shown that the third-order response function can be evaluated in that classical limit by using equilibrium or non-equilibrium MD simulation trajectories. Another simulation method is based on the assumptions that the molecular vibrations can still be described quantum mechanically and that the relevant molecular response functions are evaluated by the numerical integration of the Schrödinger equation. A few application examples are presented to help the researchers in this and related areas to understand the fundamental principles and to use these methods for their studies with 2D vibrational spectroscopic techniques. In summary, this exposition provides an overview of current theoretical efforts to understand the 2D vibrational spectra and an outlook for future developments.

I. INTRODUCTION

Resonance is a phenomenon ubiquitous in nature. Molecular spectroscopy is an experimental tool that utilizes one of the resonance phenomena that involves an interaction of oscillating charged particles in a given molecule with an external electromagnetic field whose frequency (ω_wave) is close to that (ω_molecule) of a vibrational or electronic oscillation determined by the associated interaction potential between constituent charged particles in a given polyatomic molecule in condensed phases. Such resonant field-matter interaction is manifested in the complex quantum transition amplitude that is approximately proportional to (ω_wave − ω_molecule + iγ_molecule)⁻¹, where γ_molecule determines the linewidth of such a transition band that reflects complicated molecular interactions with fluctuating internal and bath degrees of freedom. As the wave frequency ω_wave approaches the molecular oscillation frequency ω_molecule, the corresponding quantum transition amplitude increases, which has been referred to as a quantum resonant field-matter interaction process.

Vibrational spectroscopy, one of the molecular spectroscopic techniques, involves quantum transitions of vibrational degrees of freedom. The most widely used infrared (IR) spectroscopy measures the vibrational transition amplitude of normal modes via their resonant interaction with infrared (e.g., near IR, mid-IR, far-IR, and THz) radiation. The vibrational Raman spectroscopy is a resonance-enhanced scattering of the incident electronically non-resonant field when the beat frequency between the incident radiation field and the inelastic Raman scattering field is close to molecular vibrational frequency. Although IR-vis sum-frequency-generation (SFG) spectroscopy for molecular systems on the surface or at the interface with broken centrosymmetry is a nonlinear optical (three-wave-mixing) spectroscopy, it still involves a single vibrational transition whose amplitude is essentially determined by the quantum correlation between the transition dipole and the transition polarizability of a given vibrational mode.

Coherent multidimensional vibrational spectroscopy^1–5 is an extension of these linear spectroscopic methods and involves multiple vibrational transitions that are temporally separated. To observe and analyze the correlation between the distinctive vibrational transitions, one usually applies a series of coherent laser pulses with specific phase relations to induce such transitions. The multiple transitions are detected and presented as a function of corresponding frequency variables, making the spectrum multidimensional. In particular, coherent two-dimensional (2D) vibrational spectroscopy employs three femtosecond laser pulses in the infrared (IR) or visible frequency range to induce the third-order polarization in molecular systems. The signal due to the polarization is measured and presented in two frequency dimensions conjugates to the time intervals between the first and second pulses (τ) and between the third pulse and the detection (t), respectively. A series of 2D spectra at different time intervals between the second and the third pulse (T) reveals the molecular dynamics (MD) taking place during that time interval T that is called the waiting or population time.

2D vibrational spectroscopy provides ultrafast time resolution along the waiting time (T) axis as well as high spectral resolution of individual peaks along the diagonal in the 2D frequency space and their couplings in the form of off-diagonal cross peaks. Therefore, the method conveys rich information on molecular systems such as homogeneous (anti-diagonal) and inhomogeneous (diagonal) spectral broadening effects, vibrational anharmonicity, spectral diffusion,⁶ and intermode coupling strength and its temporal variation. Over the past two decades, 2D vibrational spectroscopy has been extensively used to study the structure and dynamics of small peptides, proteins, DNA, and lipid bilayers, energy transfer dynamics, hydrogen-bonding (H-bonding) structure and dynamics of liquid water, and its isotopologues, and configurational and H-bonding dynamics of biomolecules.

The amplitudes of IR and Raman transitions critically depend on the transition dipole and polarizability, respectively. Since the molecular moments m generally depend on the vibrational coordinate q, they can be Taylor-expanded in the form m(t) = m₀ + (∂m/∂q)₀q + $\dots$ with respect to the equilibrium point denoted by subscript 0. Then, when the rotational motion of each vibrational chromophore is assumed to be very slow compared to vibrational dephasing rates, according to the time correlation function formalism of molecular spectroscopy,⁷ the corresponding line shapes can be approximately written as

\begin{aligned} I_{IR} (ω) & \propto \int_{- \infty}^{\infty} d t e^{i ω t} ⟨μ (t) \cdot μ (0)⟩ \\ \approx \bar{{|{(\partial μ / \partial q)}_{0}|}^{2}} \int_{- \infty}^{\infty} d t e^{i ω t} ⟨q (t) q (0)⟩, \\ I_{Raman} (ω) & \propto \int_{- \infty}^{\infty} d t e^{i ω t} ⟨α (t) : α (0)⟩ \\ \approx \bar{{|{(\partial α / \partial q)}_{0}|}^{2}} \int_{- \infty}^{\infty} d t e^{i ω t} ⟨q (t) q (0)⟩, \end{aligned}

(1)

where (∂μ/∂q)₀ and (∂α/∂q)₀ are the transition dipole and polarizability, respectively, and the bar over, for example, $\bar{{|{(\partial μ / \partial q)}_{0}|}^{2}}$ denotes the orientational average. Due to the first-order coordinate dependences of transition dipole and transition polarizability, linear spectra (IR, Raman, IR-vis SFG), which are in principle related to the dipole-dipole, polarizability-polarizability, dipole-polarizability correlation functions, are determined by the vibrational coordinate-coordinate correlation function $⟨q (t) q (0)⟩$ ⁠.⁸

For coupled multi-oscillator systems, vibrational couplings through space via intermolecular interactions or through bond via anharmonicities in the multi-dimensional potential energy surface are crucial in understanding vibrational dynamics.^9,10 However, since the linear vibrational spectrum is mainly determined by the harmonic properties, e.g., vibrational frequency and transition dipole moment, of the oscillators, it is difficult to quantitatively extract such weak features, e.g., vibrational coupling constants and potential anharmonic coefficients of coupled oscillators in condensed phases, from those linear spectra.⁸ On the other hand, the coherent multi-dimensional vibrational spectroscopic methods have been found to be of exceptional use because they enable one to explore these mode-mode couplings and provide crucial information on the structural dynamics through the time-dependent changes of the spatial proximity and relative orientation of chromophores. Note that the nonlinear vibrational response functions determining amplitudes and lineshapes of coherent multidimensional vibrational spectra vanish for perfect harmonic oscillators.^9,11–14 One of the most popular techniques is the 2D IR spectroscopy,¹ which is a four-wave-mixing method capable of providing information on the molecular nonlinear response function that is given by multi-time correlation functions of transition dipole moments. Again, using the linear expansion form of electric dipole moment with respect to vibrational coordinates and employing perturbation theory treating electric and mechanical anharmonicities as weak perturbations to harmonic oscillators, it was possible to rewrite the nonlinear response functions in terms of vibrational Green functions and such perturbation terms.^11,12,15,16 This method was known as a linked diagram theory for the multidimensional vibrational response function. Although such theoretical approaches were popular in the early time of theory on multidimensional vibrational spectroscopy, the intrinsic difficulties in describing the anharmonicity-induced frequency shift of the excited state absorption contribution to the 2D IR spectrum, for example, and in accurately calculating vibrational (both mechanical and electrical) anharmonicities for oscillators in solutions, prohibited its wide use later.

Although 2D IR spectroscopy employing three incident IR laser pulses is one of the most widely used forms of coherent multidimensional vibrational spectroscopy, different varieties of the method have also been developed and applied for specific purposes. Examples include surface specific 2D SFG spectroscopy,^17,18 2D Raman¹⁹ and terahertz spectroscopy,^20–24 2D IR-IR-visible spectroscopy,^{3,9,13,25–29} and so on. In this article, because the basic theories for these variants are similar to one another, we mainly focus on the theory and computation of 2D IR spectroscopy and reference is made to these variants when appropriate.

By necessity, the exposition in this paper cannot be exhaustive. However, it is hoped to serve as an up-to-date introduction and a basis for elaborate research especially with the aid of references provided. In Sec. II, we briefly introduce the third-order response function formalism, which has the central importance in the description of 2D vibrational spectroscopy. In Sec. III, classical mechanical approaches to calculate 2D vibrational spectra are presented based on the classical limit of the quantum vibrational third-order response function. In Sec. IV, we present a quantum mechanical (QM) method to simulate the spectra by numerical integration of the Schrödinger equation (NISE). Finally, we conclude with perspective on future development.

II. FUNDAMENTALS

In general, spectroscopic measurements are conducted in two steps: the preparation step where the molecular system is excited by incident radiations and the detection step where the generated signal is measured and presented. In 2D vibrational spectroscopy, the system is irradiated with three coherent laser pulses, usually in the infrared (IR) frequency region, in the preparation step and then the generated signal field is detected and presented in two frequency dimensions, representing two distinct coherence oscillations separated by a waiting (population) time. It is a kind of four-wave-mixing spectroscopy due to the fact that the signal field arises from three preceding field-matter interactions that are each linear in the field. In each of the four field-matter interaction events, a quantum transition takes place between vibrational states of the system, on either the ket or the bra side of the system density operator (see below). Depending on the configuration of the optical laser pulses such as the frequency, the direction of propagation (wave vector), and polarization, as well as the detection methods, different quantum transition pathways can be selectively generated and measured.⁸

The spectroscopic observables are determined by the third-order optical response function of the molecular system that includes all distinctive quantum transition sequences consistent with the incident radiations and the detection method. The response function naturally emerges from the quantum mechanical time-dependent perturbation theory treatment of the molecular system in the presence of the three perturbative light-matter interactions of the preparation step. In this section, we sketch the theoretical procedure and present key results relevant to 2D vibrational spectroscopy.

A. Third-order optical response function

In 2D IR spectroscopy, the molecular system interacts with the incident electric field and, in the electric dipole approximation, the interaction Hamiltonian can be written as

H_{int} (t) = - \hat{μ} \cdot E (r, t),

(2)

where $\hat{μ}$ is the electric dipole operator and E(r, t) is the superposition of the electric fields of the three incident (IR, THz, visible, or X-ray) pulses denoted as E₁, E₂, and E₃. The total Hamiltonian of the system is the sum of the system Hamiltonian H₀ in the absence of radiation and H_int(t). In the cases of the other 2D vibrational spectroscopy utilizing visible, UV, etc., one needs to consider the corresponding effective field-matter interaction Hamiltonian, e.g., $- \hat{α} : E (r, t) E (r, t)$ for Raman spectroscopy.

The system evolves in time according to the quantum Liouville equation for the density operator ρ(t) of the system as follows:

\frac{\partial ρ (t)}{\partial t} = - \frac{i}{ℏ} [H_{0} + H_{int} (t), ρ] .

(3)

The solution of this equation provides quantitative information about any physical observable of the system A(t) through the expectation value $Tr [Â ρ (t)]$ ⁠, where Tr denotes the trace of a matrix and $Â$ is the operator for observable A. A diagonal element ρ_aa of the density matrix in a basis set $\{|a⟩\}$ represents the probability that the system is in state a or the population of the system in state a. The off-diagonal element ρ_ab of the density matrix, which is related to coherence or super-position state evolution of two states a and b, gives rise to the temporal oscillation of the aforementioned probability with a frequency ω ≈ ω_ab ≡ (E_a − E_b)/ℏ determined by the energy difference of the two states.

Treating H_int(t) as the perturbation to the reference Hamiltonian H₀, Eq. (3) can be solved by applying the time-dependent perturbation theory. The solution is expressed as a power series expansion of ρ(t), the zeroth-order term of which is the equilibrium density operator for the unperturbed system ρ⁽⁰⁾(t) = ρ_eq. Each of the higher-order terms ρ⁽ⁿ⁾(t) contains n factors of H_int(t) and is given by

\begin{matrix} ρ^{(n)} (t) & = & {(- \frac{i}{ℏ})}^{n} \int_{t_{0}}^{t} d τ_{n} \int_{t_{0}}^{τ_{n}} d τ_{n - 1} \dots \int_{t_{0}}^{τ_{2}} d τ_{1} G_{0} (t - τ_{n}) \\ \times L_{int} (τ_{n}) G_{0} (τ_{n} - τ_{n - 1}) L_{int} (τ_{n - 1}) \dots G_{0} (τ_{2} - τ_{1}) \\ \times L_{int} (τ_{1}) G_{0} (τ_{1} - t_{0}) ρ (t_{0}), \end{matrix}

(4)

where G₀(t) = exp(−iL₀t/ℏ) is the time evolution operator in the absence of radiation and the Liouville operators are defined as L_aA = [H_a, A] for a = 0 or int. According to Eq. (4), the system initially defined by ρ(t₀) evolves freely without perturbation for τ₁ − t₀, as given by G₀(τ₁ − t₀), and then interacts with the radiation at time τ₁, as given by L_int(τ₁). This sequence is repeated n times until the final field-matter interaction at τ_n, as given by L_int(τ_n). Finally, the system evolves freely until the observation time t according to G₀(t − τ_n). The multiple integrals over τ₁, …, τ_n account for all possible interaction times under the time ordering condition t₀ ≤ τ₁ ≤ … ≤ τ_n ≤ t.

Each term of the power series expansion of ρ(t) in Eq. (4) gives rise to the corresponding polarization $P^{(n)} (r, t) = Tr [\hat{μ} ρ^{(n)} (t)]$ in the system as follows:

\begin{matrix} P^{(n)} (r, t) & = & \int_{0}^{\infty} d t_{n} \dots \int_{0}^{\infty} d t_{1} R^{(n)} (t_{n}, \dots, t_{1}) \\ ⋮ E (r, t - t_{n}) \dots E (r, t - t_{n} \dots - t_{1}) \end{matrix}

(5)

in terms of the nth order optical response function given by

\begin{matrix} R^{(n)} (t_{n}, \dots, t_{1}) & = & {(\frac{i}{ℏ})}^{n} θ (t_{n}) \dots θ (t_{1}) ⟨) μ (t_{n} + \dots + t_{1}) \\ \times [μ (t_{n - 1} + \dots + t_{1}), [\dots [μ (t_{1}), [μ (0), ρ_{eq}]] \dots]] ⟩), \end{matrix}

(6)

where $μ (t) = \exp (i H_{0} t / ℏ) \hat{μ} \exp (- i H_{0} t / ℏ)$ is the dipole operator in the interaction picture and the angular bracket denotes the trace of a matrix. We obtain the linear response by setting n = 1 in Eqs. (5) and (6). The signal of 2D IR spectroscopy is determined by the third order polarization P⁽³⁾(r, t) and the third-order response function R⁽³⁾(t₃, …, t₁), the latter being the fourth rank tensor. Note that the time variables t₁, …, t_n−1 in Eqs. (5) and (6) are time intervals between consecutive field-matter interactions related to τ₁, …, τ_n in Eq. (4) as t_m = τ_m+1 − τ_m (1 ≤ m ≤ n − 1), while t_n = t − τ_n is the time elapsed since the last field-matter interaction. Therefore, t₁, …, t_n are all positive, and the response function must vanish if any of its time arguments are negative in accordance with the causality principle, as imposed by the Heaviside step functions θ(t) in Eq. (6). In addition, R⁽ⁿ⁾ is a real function because P⁽ⁿ⁾(r, t) and E(r, t) in Eq. (5) are both real quantities, although individual terms comprising R⁽ⁿ⁾ are complex in general and represent different quantum transition pathways.

The signal electric field $E_{s}^{(n)} (r, t)$ detected in nonlinear spectroscopy is obtained by solving Maxwell’s equation taking the nonlinear polarization P⁽ⁿ⁾(r, t) as the source. After making simplifying assumptions that (i) the signal field is only weakly absorbed by the medium, (ii) the envelopes of polarization and signal fields vary slowly in time compared to the optical period, (iii) the signal field envelope spatially varies slowly compared to its wave length, (iv) the frequency dispersion of the medium refractive index is weak, the approximate solution can be obtained as^30,31

E_{s}^{(n)} (t) \propto \frac{i ω_{s}}{n (ω_{s})} P_{s}^{(n)} (t) .

(7)

Here, n(ω) is the refractive index of the medium and $P_{s}^{(n)} (t)$ is the polarization component propagating with wave vector k_s and frequency ω_s that are one of the combinations ±k₁ ± k₂ ± ⋯ ±k_n and ±ω₁ ± ω₂ ± ⋯ ±ω_n, respectively. Note that Eq. (7) gives the approximate signal field arising from a single Fourier component of the nth order polarization expanded as^8,31

P^{(n)} (r, t) = \sum_{l} P_{l}^{(n)} (t) \exp (i k_{l} \cdot r - i ω_{l} t) .

(8)

By changing the location of the detector appropriately, individual components of the polarization with different k_s can be selectively measured. Note that the assumption (ii) above could become invalid for the far-IR and THz spectroscopy. See Ref. 32 and references therein for more general approaches to the nonlinear signal field calculation with wider applicability.

B. Response function of a three-level system

2D vibrational spectroscopy usually induces transitions up to the second vibrational excited state. Therefore, a three-level system with eigenstates $|g⟩$ ⁠, $|e⟩$ ⁠, and $|f⟩$ is a useful model for the response function relevant to 2D vibrational spectroscopy. Because the third-order response function vanishes for a harmonic oscillator, the model system must represent an anharmonic oscillator where the fundamental transition frequency ω_eg is slightly larger than ω_fe.

The evaluation of a realistic response function critically depends on the accurate description of the system-bath interaction that is responsible for important spectroscopic phenomena such as dephasing, relaxation, reorientation, spectral diffusion, and population and coherence transfers. Methods to incorporate the effect of environment as well as the multimode vibrational coupling are discussed in Secs. III and IV. In this section, to highlight the structure of the response function, we consider a simple model where a single three-level chromophore interacts with the environment according to the following Hamiltonian:

H_{0} = \sum_{m = g, e, f} [ℏ ω_{m} + V_{m} (q) + H_{B} (q)] |m⟩ ⟨m| .

(9)

Here, ℏω_m is the energy of state m in the absence of bath, V_m(q) is the chromophore-bath interaction energy of the state m that depends on the bath degrees of freedom q, H_B(q) is the energy of the bath, and the basis states $|m⟩$ (m = g, e, f) are chosen as eigenstates of an isolated chromophore. Note that the off-diagonal elements of the chromophore-bath interaction are assumed negligible for the sake of simplicity.³³ Using this Hamiltonian, the three nested commutators in the response function in Eq. (6) can be expanded as the sum of eight terms^31,34

\begin{matrix} R^{(3)} (t_{3}, t_{2}, t_{1}) & = & {(\frac{i}{ℏ})}^{3} θ (t_{3}) θ (t_{2}) θ (t_{1}) \\ \times \sum_{i = 1}^{4} [R_{i} (t_{3}, t_{2}, t_{1}) - R_{i}^{*} (t_{3}, t_{2}, t_{1})], \end{matrix}

(10)

where the components R_i(t₃, t₂, t₁) are given by^8,35

\begin{matrix} R_{1} (t_{3}, t_{2}, t_{1}) & = & μ_{g e} μ_{e g} μ_{g e} μ_{e g} \exp [i (- {\bar{ω}}_{e g} t_{3} - {\bar{ω}}_{e g} t_{1})] F_{1}^{g e g e} (t_{3}, t_{2}, t_{1}) \\ + μ_{g e} μ_{e f} μ_{f e} μ_{e g} \exp [i ({\bar{ω}}_{f e} t_{3} - {\bar{ω}}_{e g} t_{1})] F_{1}^{g e f e} (t_{3}, t_{2}, t_{1}), \\ R_{2} (t_{3}, t_{2}, t_{1}) & = & μ_{g e} μ_{e g} μ_{g e} μ_{e g} \exp [i (- {\bar{ω}}_{e g} t_{3} + {\bar{ω}}_{e g} t_{1})] F_{2}^{g e g e} (t_{3}, t_{2}, t_{1}) \\ + μ_{g e} μ_{e f} μ_{f e} μ_{e g} \exp [i ({\bar{ω}}_{f e} t_{3} + {\bar{ω}}_{e g} t_{1})] F_{2}^{g e f e} (t_{3}, t_{2}, t_{1}), \\ R_{3} (t_{3}, t_{2}, t_{1}) & = & μ_{g e} μ_{e g} μ_{g e} μ_{e g} \exp [i (- {\bar{ω}}_{e g} t_{3} + {\bar{ω}}_{e g} t_{1})] F_{3}^{g e g e} (t_{3}, t_{2}, t_{1}) \\ + μ_{g e} μ_{e f} μ_{f e} μ_{e g} \exp [i ({\bar{ω}}_{f e} t_{3} + {\bar{ω}}_{f g} t_{2} + {\bar{ω}}_{e g} t_{1})] \\ \times F_{3}^{g e f e} (t_{3}, t_{2}, t_{1}), \\ R_{4} (t_{3}, t_{2}, t_{1}) & = & μ_{g e} μ_{e g} μ_{g e} μ_{e g} \exp [i (- {\bar{ω}}_{e g} t_{3} - {\bar{ω}}_{e g} t_{1})] F_{4}^{g e g e} (t_{3}, t_{2}, t_{1}) \\ + μ_{g e} μ_{e f} μ_{f e} μ_{e g} \exp [i (- {\bar{ω}}_{e g} t_{3} - {\bar{ω}}_{f g} t_{2} - {\bar{ω}}_{e g} t_{1})] \\ \times F_{4}^{g e f e} (t_{3}, t_{2}, t_{1}), \end{matrix}

(11)

assuming that the system is initially in the ground state g. Here, μ_ab is the transition dipole between states a and b obtained by the repeated insertion of the closure relation $\sum_{m} |m⟩ ⟨m| = 1$ in Eq. (6), which is assumed to be independent of the bath coordinate (Condon approximation), $ℏ {\bar{ω}}_{a b} = ℏ (ω_{a} - ω_{b}) + {⟨V_{a} (q) - V_{b} (q)⟩}_{B}$ is the energy gap averaged over bath degrees of freedom, and $F_{n}^{g a b c} (t_{3}, t_{2}, t_{1})$ is the line shape function expressed in terms of time-ordered exponentials of the fluctuations in the system-bath interactions, $U_{m} (q) = V_{m} (q) - {⟨V_{m} (q)⟩}_{B}$ ⁠. Therefore, the response function is composed of multiple quantum transition pathways represented by individual R_i, each of which is the product of three factors determining the transition strength (products of transition moments), the transition frequency (coherence oscillation), and the line shape (F₁₋₄). Throughout this paper, third order double-quantum or zero-quantum spectroscopies are not taken into consideration.

To facilitate computation of $F_{n}^{g a b c} (t_{3}, t_{2}, t_{1})$ ⁠, we can approximately replace the time-ordered exponential operators with normal exponential functions containing difference potential energies U_ab(q) = U_a(q) − U_b(q). Alternatively, we can invoke the second-order cumulant expansion approximation, which becomes exact when the fluctuation of the energy gap $ℏ {\bar{ω}}_{a b}$ obeys the Gaussian statistics, to obtain⁸

\begin{aligned} R_{1 A} (t_{3}, t_{2}, t_{1}) & = μ_{g e} μ_{e g} μ_{g e} μ_{e g} \exp (- i {\bar{ω}}_{e g} t_{3} - i {\bar{ω}}_{e g} t_{1}) \exp [- g^{*} (t_{3}) - g (t_{1}) - f_{+} (t_{3}, t_{2}, t_{1})], \\ R_{1 B} (t_{3}, t_{2}, t_{1}) & = μ_{g e} μ_{e f} μ_{f e} μ_{e g} \exp (i {\bar{ω}}_{f e} t_{3} - i {\bar{ω}}_{e g} t_{1}) \\ \times \exp [- g^{*} (t_{3}) - g (t_{1}) + g^{*} (t_{2}) - g (t_{1} + t_{2}) - g^{*} (t_{2} + t_{3}) + g (t_{1} + t_{2} + t_{3})], \\ R_{2 A} (t_{3}, t_{2}, t_{1}) & = μ_{g e} μ_{e g} μ_{g e} μ_{e g} \exp (- i {\bar{ω}}_{e g} t_{3} + i {\bar{ω}}_{e g} t_{1}) \exp [- g^{*} (t_{3}) - g^{*} (t_{1}) + f_{+}^{*} (t_{3}, t_{2}, t_{1})], \\ R_{2 B} (t_{3}, t_{2}, t_{1}) & = μ_{g e} μ_{e f} μ_{f e} μ_{e g} \exp (i {\bar{ω}}_{f e} t_{3} + i {\bar{ω}}_{e g} t_{1}) \\ \times \exp [- g^{*} (t_{3}) - g^{*} (t_{1}) - g (t_{2}) + g^{*} (t_{1} + t_{2}) + g (t_{2} + t_{3}) - g^{*} (t_{1} + t_{2} + t_{3})], \\ R_{3} (t_{3}, t_{2}, t_{1}) & = μ_{g e} μ_{e g} μ_{g e} μ_{e g} \exp (- i {\bar{ω}}_{e g} t_{3} + i {\bar{ω}}_{e g} t_{1}) \exp [- g (t_{3}) - g^{*} (t_{1}) + f_{-}^{*} (t_{3}, t_{2}, t_{1})], \\ R_{4} (t_{3}, t_{2}, t_{1}) & = μ_{g e} μ_{e g} μ_{g e} μ_{e g} \exp (- i {\bar{ω}}_{e g} t_{3} - i {\bar{ω}}_{e g} t_{1}) \exp [- g^{*} (t_{3}) - g (t_{1}) - f_{-} (t_{3}, t_{2}, t_{1})], \end{aligned}

(12)

where the auxiliary functions are given by

\begin{matrix} \begin{gathered} g (t) = \frac{1}{ℏ^{2}} \int_{0}^{t} d τ_{1} \int_{0}^{τ_{1}} d τ_{2} {⟨U_{e g} (τ_{2}) U_{e g} (0)⟩}_{B}, \\ U_{e g} (t) = \exp (\frac{i}{ℏ} H_{0}^{g} t) [U_{e} (q) - U_{g} (q)] \exp (- \frac{i}{ℏ} H_{0}^{g} t), H_{0}^{g} = ⟨g| H_{0} |g⟩, \\ f_{+} (t_{3}, t_{2}, t_{1}) = g^{*} (t_{2}) - g^{*} (t_{2} + t_{3}) - g (t_{1} + t_{2}) + g (t_{1} + t_{2} + t_{3}), \\ f_{-} (t_{3}, t_{2}, t_{1}) = g (t_{2}) - g (t_{2} + t_{3}) - g (t_{1} + t_{2}) + g (t_{1} + t_{2} + t_{3}) . \end{gathered} \end{matrix}

(13)

The second terms of R₃ and R₄ in Eq. (11), that represent coherence evolution during t₂, are not included in Eq. (12) and the energy fluctuation between states g and f is assumed to be twice that between g and e, i.e., U_fg(t) ≅ 2U_eg(t). A few approximate expressions for the line-broadening function g(t) can be found in other review articles and books. The six response functions in Eq. (12) can be illustrated using double-sided Feynman diagrams (Fig. 1), which highlight their physical interpretation in terms of ground state bleach, stimulated emission, and excited state absorption contributions. These are further classified as rephasing (top row in Fig. 1) and non-rephasing (bottom row in Fig. 1) versions depending on whether the coherence oscillations during t₁ and t₃ have the opposite or the same signs, respectively.

FIG. 1.

View large Download slide

Double-sided Feynman diagrams illustrating the six contributions of Eq. (12). Time is running from bottom to the top with the vertical lines illustrating the ket and the bra wave functions. Interactions with applied laser fields are illustrated with full arrows, while emitted light is illustrated with dashed arrows. The numbers (zero, one, and two) indicate the number of vibrations excited. The response functions R_2A and R₄ are identified as ground state bleach as the system is in the ground state during the waiting time t₂, while R₃ and R_1A are stimulated emission contributions, and R_1B and R_2B are excited state absorption contributions.

This general formulation for three-level systems is very powerful and can be used to understand the effect of dynamics on the two-dimensional lineshapes,³⁶ which, for example, allows the distinction between homogeneous and inhomogeneous line-broadenings and defines rules for extracting the correlation functions describing the bath dynamics in the limit of the Gaussian dynamics, for example, using the nodal or center line slope.^6,37 This is illustrated in Fig. 2.

FIG. 2.

View large Download slide

Illustration of the 2D IR spectra with slow (left) and fast (right) dynamics. The corresponding linear spectra, which have little sensitivity to dynamics, are shown on the top. The diagonal peaks shown with red contours are bleaching signals (R_1A, R_2A, R₃, and R₄), while the peaks with blue contours, which are shifted by the anharmonicity below the diagonal, arise from the absorption of the excited state (R_1B and R_2B). When the bath dynamics scramble the vibrational frequencies faster than the waiting time, all correlation between excitation and detection frequencies is lost and round peaks are obtained, while elongated peaks arise from residual memory.

III. CLASSICAL MECHANICAL APPROACHES TO NONLINEAR RESPONSE FUNCTIONS

In Sec. II, we derived the quantum mechanical linear and nonlinear response functions. Although these equations are formally exact, fully quantum mechanical simulations are still impractical for systems with many degrees of freedom even with state-of-the-art supercomputers. Thus, applicable and efficient methods are required for the calculation of nonlinear response functions for such systems. In this section, we first summarize the classical mechanical approaches to the calculation of the linear and nonlinear response functions, which can be expressed as (multi-)time correlation functions on equilibrium and nonequilibrium trajectories, and then present some results of the 2D IR and pump-probe spectra obtained from the approaches.

A. Equilibrium molecular dynamics approach

The classical mechanical response functions can be derived by using the relationship between the commutator and the Poisson bracket,³⁸

\frac{1}{i ℏ} [X, Y] = {X, Y}_{PB} .

(14)

Here, X and Y are physical variables and the Poisson bracket is defined as follows:

{X, Y}_{PB} = \frac{\partial X}{\partial q} \frac{\partial Y}{\partial p} - \frac{\partial X}{\partial p} \frac{\partial Y}{\partial q} .

(15)

When Y is the equilibrium distribution function of the system ρ_eq = e^−βH/Z, where Z is the canonical partition function, Eq. (15) becomes ${X, ρ_{eq}}_{PB} = - β \dot{X} ρ_{eq}$ ⁠. Here, β is the reciprocal temperature of the system and $\dot{X}$ is the time derivative of X. As a result, we obtain the following well-known general expression for the classical linear response function of a physical quantity A to a perturbation B⁷ by using Eq. (14) in the quantum linear response function $R^{(1)} (t) = \frac{i}{ℏ} θ (t) Tr \{A (t) [B (0), ρ_{eq}]\}$ ⁠, which corresponds to n = 1 in Eq. (6),

R^{(1)} (t) = β ⟨ \dot{B} (0) A (t) ⟩ .

(16)

Here, the angular bracket indicates ensemble average. For example, the classical expression of the Raman response function is given as³⁹

R_{R a m a n} (t) = β ⟨ \dot{α} (0) α (t) ⟩,

(17)

where α is the total polarizability of the system. Equation (17) shows that the response function for the optical Kerr effect is expressed as the time derivative of time correlation of the polarizability along the equilibrium classical trajectory.

By using Eq. (14), we can also derive any classical nonlinear response function. For example, the fifth-order nonlinear Raman spectroscopy,^19,38 which is the second-order response function of the polarizability, is given as^39–45

\begin{matrix} R^{(2)} (t_{1}, t_{2}) & = & - β ⟨ {α (t_{1} + t_{2}), α (t_{1})}_{PB} \dot{α} (0) ⟩ \\ = & β ⟨ α (t_{1} + t_{2}) (β \dot{α} (t_{1}) \dot{α} (0) - {α (t_{1}), \dot{α} (0)}_{PB}) ⟩ . \end{matrix}

(18)

These equations have been used for the analyses of the fifth-order nonlinear Raman spectra of liquids.^39–41,46

Similarly, the classical third-order response function for 2D IR spectroscopy is expressed as^47–49

\begin{matrix} R^{(3)} (t_{1}, t_{2}, t_{3}) & = & β ⟨ {{μ (t_{1} + t_{2} + t_{3}), μ (t_{1} + t_{2})}_{PB}, μ (t_{1})}_{PB} \dot{μ} (0) ⟩ \\ = & - β ⟨) {μ (t_{1} + t_{2} + t_{3}), μ (t_{1} + t_{2})}_{PB} \\ \times (β \dot{μ} (t_{1}) \dot{μ} (0) - {μ (t_{1}), \dot{μ} (0)}_{PB}) ⟩), \end{matrix}

(19)

where μ denotes the dipole moment of the system.

It should be noted here that the Poisson brackets of physical variables at different times, e.g., ${μ (t), μ (t^{'})}_{PB}$ ⁠, are required for the calculation of nonlinear response functions based on the equilibrium molecular dynamics approach. The explicit expression of ${μ (t), μ (t^{'})}_{PB}$ is

{μ (t), μ (t^{'})}_{PB} = \sum_{α β} \frac{\partial μ (t)}{\partial q_{α} (t)} \frac{\partial μ (t^{'})}{\partial q_{β} (t^{'})} \frac{\partial q_{β} (t^{'})}{\partial p_{α} (t)},

(20)

where the dipole moment is assumed to be a function of particle positions only and we utilized the invariance of Poisson brackets under canonical transformation^43,50 (of which the Newtonian time evolution is an example) and the chain rule of the form $\partial μ (t^{'}) / \partial p_{α} (t) = \sum_{β} [\partial μ (t^{'}) / \partial q_{β} (t^{'})] [\partial q_{β} (t^{'}) / \partial p_{α} (t)]$ ⁠. The partial derivative matrix, ∂q(t′)/∂p(t), represents the variation of the position at t′ induced by a small change of momentum at t. This matrix is a sub-matrix of the so-called stability matrix, J(t) = ∂γ(t)/∂γ(0), where γ(t) = (q(t), p(t)). The time-evolution of the stability matrix is expressed as^38,39

\frac{d}{d t} J (t) = (\begin{matrix} \partial^{2} H_{0} / \partial q (t) \partial p (t) & \partial^{2} H_{0} / \partial p (t) \partial p (t) \\ - \partial^{2} H_{0} / \partial q (t) \partial q (t) & - \partial^{2} H_{0} / \partial q (t) \partial p (t) \end{matrix}) J (0)

(21)

with the initial condition of J(0) = 1. The stability matrix represents the transformation of the phase space along the trajectory and plays an important role in nonlinear mechanics and semiclassical theory.^51–53 The nonlinear response functions are highly sensitive to the trajectory, and thus, they can provide more detailed information on the dynamics than the linear response function.³⁹

Jeon and Cho have investigated the 2D IR spectra of intramolecular vibrations in water.^49,54 In their studies, the quantum mechanical/molecular mechanical (QM/MM) approach has been employed for the accurate description of the intramolecular vibrations of solute molecules: deuterated N-methylacetamide (d-NMA) and HOD molecules have been described with the semiempirical PM3 method and scc-DFTB potential, respectively. The flexible TIP3P and SPC/Fw models have been used for solvent water molecules, respectively. Figure 3 shows the 2D IR spectra of the OD stretch of the HOD molecule in a water cluster. In these calculations,⁵⁴ Jeon and Cho have improved the computational efficiency by exploiting the time reversibility of trajectory and have successfully calculated the 2D IR spectra. The positive and negative peaks are found at (∼2650 cm⁻¹, ∼2700 cm⁻¹) and (∼2600 cm⁻¹, ∼2550 cm⁻¹), respectively. The positive peak corresponds to the stimulated emission and the ground state bleaching of the amide I band, whereas the negative peak is related to the excited state absorption. Note that various general features of the experimental 2D IR spectra are reproduced in the calculated 2D spectra based on classical mechanics: the vertical splitting of positive and negative peaks due to the vibrational anharmonicity, the diagonal elongation of the signal reflecting the inhomogeneity of the solvation environment, the change in lineshape with waiting time due to the spectral diffusion, and the decrease in the tilt angle of the nodal line. The classical 2D spectra also show that the main decay component of the nodal line with a ∼1.6 ps time scale obtained from the calculated 2D spectra is comparable to experimental results.⁵⁵

FIG. 3.

View large Download slide

Absorptive 2D IR spectra of the OD stretch of the hydrated HOD molecule at different waiting times T. All figures are drawn on the same amplitude scale. Diminishing tilt angles of the nodal line and the signal amplitudes can be noticed with increasing waiting time T, indicating the loss of frequency memory and the vibrational population relaxation, respectively. Reprinted with permission from J. Jeon and M. Cho, J. Phys. Chem. B 118, 8148 (2014). Copyright 2014 American Chemical Society.

B. Nonequilibrium molecular dynamics approach

1. Nonequilibrium finite-field method

In Subsection III A, we described the prescription for the linear and nonlinear response functions based on the equilibrium molecular dynamics approach. Since the stability matrix is required in the approach, high computational costs and numerical instability would make it difficult to calculate nonlinear response functions, when many degrees of freedom have to be considered explicitly, for example, the low frequency intermolecular motions and the OH stretches in liquid H₂O. Thus, other approaches circumventing the use of the stability matrix would be required for such systems. In this subsection, we introduce the nonequilibrium molecular dynamics approach in which the response functions are evaluated by considering (sequential) external fields, as in real experiments.

By introducing a Liouville operator B₋ and a dimensionless parameter ε, we have the following identity:⁵⁶

\frac{i}{ℏ} [B,] \equiv \frac{i}{ℏ} B_{-} = lim_{ε \to 0} \frac{1}{ε} [\exp (i ε B_{-} / ℏ) - 1] .

(22)

By applying Eq. (22) to the linear response function R⁽¹⁾(t) (for t > 0) of a physical variable A to a perturbation B, we have^56–58

\begin{matrix} R^{(1)} (t) & = & \frac{i}{ℏ} Tr \{A (t) [B (0), ρ_{e q}]\} \\ = & lim_{ε \to 0} \frac{1}{ε} (Tr \{A (t) (e^{i ε B_{-} / ℏ} - 1) ρ_{e q}\}) \\ \equiv & lim_{ε \to 0} \frac{1}{ε} \{{⟨ A (t) ⟩}_{B (0)} - ⟨ A ⟩\} . \end{matrix}

(23)

Here, the first term is the expectation value of A(t) on the perturbed trajectory given by H₀ − εBδ(t), whereas the second term is the expectation value of A on the trajectory expressed by the unperturbed Hamiltonian H₀.

In the nonequilibrium finite-field method, the second-order response function of 2D Raman spectroscopy, corresponding to Eq. (18) in the equilibrium molecular dynamics approach, is⁴⁶

\begin{matrix} R^{(2)} (t_{1}, t_{2}) & = & lim_{ε \to 0} \frac{1}{ε^{2}} \{) {⟨ Π (t_{1} + t_{2}) ⟩}_{E (0) E (0), E (t_{1}) E (t_{1})} \\ - {⟨ Π (t_{1} + t_{2}) ⟩}_{E (0) E (0)} - {⟨ Π (t_{1} + t_{2}) ⟩}_{E (t_{1}) E (t_{1})} \\ + ⟨ Π (t_{1} + t_{2}) ⟩ \}) . \end{matrix}

(24)

Jansen et al. have applied the nonequilibrium finite-field method to the 1D and 2D Raman spectra of liquids.^57,58

Based on the nonequilibrium finite-field method, the third-order response function for 2D IR spectroscopy is expressed as⁴⁸

\begin{matrix} R^{(3)} (t_{1}, t_{2}, t_{3}) & = & lim_{ε \to 0} \frac{1}{ε^{3}} \{) {⟨ μ (t_{1} + t_{2} + t_{3}) ⟩}_{E (0), E (t_{1}), E (t_{1} + t_{2})} \\ - {⟨ μ (t_{1} + t_{2} + t_{3}) ⟩}_{E (0), E (t_{1})} - {⟨ μ (t_{1} + t_{2} + t_{3}) ⟩}_{E (0), E (t_{1} + t_{2})} \\ - {⟨ μ (t_{1} + t_{2} + t_{3}) ⟩}_{E (t_{1}), E (t_{1} + t_{2})} \\ + {⟨ μ (t_{1} + t_{2} + t_{3}) ⟩}_{E (0)} + {⟨ μ (t_{1} + t_{2} + t_{3}) ⟩}_{E (t_{1})} \\ + {⟨ μ (t_{1} + t_{2} + t_{3}) ⟩}_{E (t_{1} + t_{2})} - ⟨ μ (t_{1} + t_{2} + t_{3}) ⟩ \}) . \end{matrix}

(25)

Later, Jansen et al., have developed the efficient method to subtract higher-order responses from calculated signals by combining the nonequilibrium molecular dynamics simulations with the positive and negative electric fields.⁵⁹ By using this method, the third-order nonlinear response function can be recast in the following form:

\begin{matrix} R^{(3)} (t_{1}, t_{2}, t_{3}) & = & lim_{ε \to 0} \frac{1}{4 ε^{3}} \{{⟨ μ (t_{1} + t_{2} + t_{3}) ⟩}_{E (0), E (t_{1}), E (t_{1} + t_{2})} - {⟨ μ (t_{1} + t_{2} + t_{3}) ⟩}_{E (0), E (t_{1}), Ē (t_{1} + t_{2})} - {⟨ μ (t_{1} + t_{2} + t_{3}) ⟩}_{E (0), Ē (t_{1}), E (t_{1} + t_{2})}) \\ + {⟨ μ (t_{1} + t_{2} + t_{3}) ⟩}_{E (0), Ē (t_{1}), Ē (t_{1} + t_{2})} - {⟨ μ (t_{1} + t_{2} + t_{3}) ⟩}_{Ē (0), E (t_{1}), E (t_{1} + t_{2})} + {⟨ μ (t_{1} + t_{2} + t_{3}) ⟩}_{Ē (0), E (t_{1}), Ē (t_{1} + t_{2})} \\ (+ {⟨ μ (t_{1} + t_{2} + t_{3}) ⟩}_{Ē (0), Ē (t_{1}), E (t_{1} + t_{2})} - {⟨ μ (t_{1} + t_{2} + t_{3}) ⟩}_{Ē (0), Ē (t_{1}), Ē (t_{1} + t_{2})}\}, \end{matrix}

(26)

where the bar on the subscript denotes the application of the external field with the opposite direction.

2. Hybrid approach combining equilibrium and nonequilibrium finite-field methods

The nonequilibrium finite-field method does not need the stability matrix for the calculation of nonlinear responses. However, the approach still demands high computational costs; i.e., extra three and seven nonequilibrium trajectories as well as the equilibrium trajectory are required for the second- and third-order nonlinear response functions, respectively.

In order to reduce the number of trajectories, Hasegawa and Tanimura have developed an efficient computational method for higher-order response functions, i.e., the hybrid method combining the equilibrium molecular dynamics and nonequilibrium finite-field methods.⁶⁰ In the hybrid method, the second-order response function for the 2D Raman spectra given by Eqs. (18) and (24) can be written as⁶¹

\begin{matrix} R^{(2)} (t_{1}, t_{2}) & = & lim_{ε \to 0} \frac{1}{2 ε} \{) β () {⟨ \dot{Π} (0) Π (t_{1} + t_{2}) ⟩}_{E (t_{1}) E (t_{1})} \\ - {⟨ \dot{Π} (0) Π (t_{1} + t_{2}) ⟩}_{E (t_{1}) Ē (t_{1})})) \}) . \end{matrix}

(27)

In Eq. (27), the inverse force method is also used, as indicated by the overbar in the last subscript. Equation (27) shows that, in the hybrid method, the second-order response function can be calculated from the time correlation function of $\dot{Π} (0)$ on the equilibrium trajectory and Π(t₁ + t₂) on the nonequilibrium trajectory generated by the external fields applied at t₁.

The third-order response function of the 2D IR spectra can be calculated with the following four terms:⁶¹

\begin{matrix} R^{(3)} (t_{1}, t_{2}, t_{3}) & = & lim_{ε \to 0} \frac{1}{4 ε^{2}} \{) β () {⟨ \dot{μ} (0) μ (t_{1} + t_{2} + t_{3}) ⟩}_{E (t_{1}), E (t_{1} + t_{2})} \\ - {⟨ \dot{μ} (0) μ (t_{1} + t_{2} + t_{3}) ⟩}_{E (t_{1}), \bar{E} (t_{1} + t_{2})} \\ - {⟨ \dot{μ} (0) μ (t_{1} + t_{2} + t_{3}) ⟩}_{\bar{E} (t_{1}), E (t_{1} + t_{2})} \\ + {⟨ \dot{μ} (0) μ (t_{1} + t_{2} + t_{3}) ⟩}_{\bar{E} (t_{1}), \bar{E} (t_{1} + t_{2})})) \}), \end{matrix}

(28)

where the inverse forces are denoted with overbars.

Later, Hasegawa and Tanimura⁴⁷ have proposed another efficient method to calculate the third-order response function by exploiting backward nonequilibrium trajectories. In this method, the third-order response function is re-expressed as

\begin{matrix} R^{(3)} (t_{1}, t_{2}, t_{3}) & = & β ⟨) lim_{ε \to 0} \frac{1}{2 ε} (μ {(t_{2} + t_{3})}_{E (t_{2})} - μ {(t_{2} + t_{3})}_{\bar{E} (t_{2})}) \\ \times {β \dot{μ} (- t_{1}) \dot{μ} (0) + lim_{ε \to 0} \frac{1}{2 ε} (\dot{μ} {(- t_{1})}_{E (0)} - \dot{μ} {(- t_{1})}_{\bar{E} (0)})} ⟩) \\ - β ⟨ lim_{ε \to 0} \frac{1}{2 ε} (μ {(t_{2} + t_{3})}_{E (t_{2})} - μ {(t_{2} + t_{3})}_{\bar{E} (t_{2})}) ⟩ \\ \times ⟨ β \dot{μ} (- t_{1}) \dot{μ} (0) + lim_{ε \to 0} \frac{1}{2 ε} (\dot{μ} {(- t_{1})}_{E (0)} - \dot{μ} {(- t_{1})}_{\bar{E} (0)}) ⟩, \end{matrix}

(29)

where μ(t₂ + t₃)_E(t2) and $\dot{μ} {(- t_{1})}_{E (0)}$ denote the dipole moment at t₂ + t₃ on the perturbed trajectory with the Hamiltonian H₀ − εμEδ(t − t₂) and the time derivative of dipole moment at −t₁ on the backward perturbed trajectory H₀ − εμEδ(t), respectively, while $\dot{μ} (- t_{1}) \dot{μ} (0)$ is evaluated from the equilibrium trajectory. In practical computations, the second term in Eq. (29) is required due to the breakdown of equipartition.⁶¹ The backward-forward sampling method given as Eq. (29) is particularly efficient for the sampling of the response function at non-zero t₂ values because the third-order response functions at different t₂ values can be efficiently calculated by choosing the pair of initial configurations along the forward and backward trajectories.

It should be noted that the experimental third-order nonlinear spectra are obtained under different phase matching conditions [see Eq. (8)]. For example, the 2D IR spectra consist of the responses emitted along the wave vector directions k_I = −k₁ + k₂ + k₃ and k_II = +k₁ − k₂ + k₃. However, the calculated third-order response function consists of all the wave vector components generated by three incident electric fields, i.e., k_III = +k₁ + k₂ − k₃ and k_IV = +k₁ + k₂ + k₃ as well as k_I and k_II. Therefore, the components with k_III and k_IV have to be eliminated from the calculated third-order responses for the 2D IR spectrum. The responses arising from the individual phase matching conditions can easily be obtained by using the nonequilibrium molecular dynamics approach in which external electric fields are explicitly considered. Thus, various third-order nonlinear spectroscopies, such as the three-pulse photon echo peak shift and the pump-probe spectrum, can also be calculated.⁶¹ In the earlier studies, k_III and k_IV components have been eliminated by exploiting the fact that the three-dimensional Fourier transformed spectra of k_III and k_IV components show higher frequency oscillation in ω₂ than those of k_I and k_II.^47–49,54

Hasegawa and Tanimura have applied the backward-forward sampling method to the 2D IR spectra of liquid HF.⁴⁷ As shown in Fig. 4, the positive and negative peaks of intermolecular libration motion are located at (550 cm⁻¹, 550 cm⁻¹) and (550 cm⁻¹, 400 cm⁻¹), respectively. The frequency correlation of the libration motion of liquid HF was found to be lost with a time scale of ∼200 fs. It was also found that the width of the anti-diagonal line of the peak is narrow compared with that of liquid water, indicating the suppressed homogeneous broadening in liquid HF.

FIG. 4.

View large Download slide

Correlation spectra of liquid HF at t₂ = (a) 0 fs, (b) 20 fs, (c) 100 fs, (d) 200 fs, (e) 300 fs, and (f) 400 fs. The diagonal and off-diagonal peaks in each spectrum arise from fundamental and anharmonic oscillations, respectively, of the hydrogen bond network in liquid HF. Reprinted with permission from T. Hasegawa and Y. Tanimura, J. Chem. Phys. 128, 064511 (2008). Copyright 2008 AIP Publishing LLC.

Yagasaki and Saito investigated the fluctuation and relaxation of the intermolecular motions in liquid water by using the hybrid method with the SPC/E model for water molecules.⁴⁸ Figure 5 shows the 2D IR spectra of the intermolecular motions of liquid water at several waiting times. The intermolecular translational and librational motions are found below and above 300 cm⁻¹, respectively. As shown in Fig. 5, the tilt angle of the nodal line between the positive and negative peaks of the librational motion decays with a time constant of ∼110 fs. Yagasaki and Saito found that the frequency fluctuation of librational motion becomes three times slower when the translational motions of individual molecules are frozen, indicating that the ultrafast frequency fluctuation of the librational motion arises from the coupling between the translational and librational motions. It was found that the three-pulse stimulated photon-echo signal of the librational motion in water decays with time scales of ∼20 and ∼100 fs, in which the slower time scale is similar to that obtained from the change of the nodal line tilt angle.⁴⁸

FIG. 5.

View large Download slide

2D IR correlation spectra of intermolecular motion of liquid water at several waiting times. The tilt angle is defined as the angle between the ν₁ axis and the nodal line of the libration peak as shown in (a) and is found to decay with a time scale of 115 fs. Reprinted with permission from T. Yagasaki and S. Saito, J. Chem. Phys. 128, 154521 (2008). Copyright 2008 AIP Publishing LLC.

The energy relaxation process of the librational modes has been examined by the pump-probe signals calculated from the third-order response functions.⁶¹ It was found that the absorption changes at 700 and 800 cm⁻¹ are induced by the initial decrease of absorption due to the ground state bleach and stimulated emission followed by the fast energy relaxation of the librational motion to low frequency modes within ∼60 fs and the subsequent slow thermalization that corresponds to its energy relaxation to hot ground states occurs in 500 fs. These time scales of the two relaxation processes are consistent with the experimental results.^62–64

As mentioned above, once third-order response functions with individual phase-matching conditions are obtained, any third-order nonlinear spectrum can be calculated in principle. However, despite these developments of efficient computational approaches, the calculation of the third-order response function is still expensive and computationally demanding. In this regard, it should be noted that other efficient computational methods to estimate mode-to-mode vibrational energy relaxation rates have been developed by Yagasaki and Saito^65,66 and Jeon and Cho.^67,68

Recently, the nonequilibrium molecular dynamics method has also been applied to the intramolecular vibrational modes of H₂O in water.⁶⁹ In this study, the 2D IR spectra of the OH stretch and the HOH bending modes were calculated with the TTM3-F interaction potential, which is the ab initio-based charge transferable, flexible, and polarizable Thole-type model to describe the inter- and intra-molecular interactions in liquid water. The waiting time-dependent changes of the OH stretch spectra indicate the heterogeneous dynamics of the local H-bonding network which has also been found in an experimental result.⁷⁰ The spectrum loses diagonal correlation at waiting times larger than 200 fs, which is consistent with the other quantum mechanical calculation results.^71,72 The 2D IR spectra of the HOH bend are found to have a smaller nodal line tilt angle at T = 0 fs compared to the OH stretch spectra and become parallel to the ω₁ axis within ∼400 fs.

The pump-probe spectra of the OH stretch and the HOH bend in water have also been calculated with the TTM3-F interaction potential.⁶⁹ The positive and negative peaks of the pump-probe spectra of the OH stretch at (ω_pump, ω_probe) = (3600 cm⁻¹, 3225 cm⁻¹) and (3600 cm⁻¹, 3600 cm⁻¹), corresponding to the 1 → 2 and 0 → 1 transitions, show the initial decay with a time constant of 240 fs followed by the slow relaxation to the hot-ground state. The calculated time scale of the initial decay is in agreement with experimental results of 200-270 fs.^73–75 The pump-probe spectra of the HOH bend at (1650 cm⁻¹, 1475 cm⁻¹) and (1650 cm⁻¹, 1650 cm⁻¹) decay with a time constant of 250 fs, which is in good agreement with the experimental results of 170–260 fs.^62,64 Furthermore, the frequency-resolved kinetic energy analysis revealed the detailed relaxation processes after the excitations of the OH stretch and HOH bend of H₂O in water.

C. Limitations

In this section, the classical simulation methods for the nonlinear response functions were presented with their applications to the calculations of the 2D IR and pump-probe spectra of liquids. Before closing this section, we summarize the limitations of the classical treatment of nonlinear spectral simulations.

The validity of the classical 2D IR spectra has been investigated by several groups.^76–81 It should be noted that the classical nonlinear response functions are not stable, for example, for integrable systems and systems without dissipation.^77,78 Sakurai and Tanimura examined the quantum effects on the simulated IR and 2D IR spectra of a Morse oscillator in a harmonic bath by solving the quantum and classical hierarchical equations of motion (HEOM).⁸⁰ They found that the classical 2D IR spectra are a good approximation of the quantum 2D IR spectra when the system (vibration) is largely modulated by the bath, i.e., via a strong system-bath coupling or a fast bath modulation even in a weak system-bath coupling. On the other hand, there are significant differences between the quantum and classical mechanically calculated spectra when the modulation due to the system-bath interactions is weak or slow. Very recently, Reppert and Brumer have also shown that the classical 2D IR spectra of a Morse oscillator, which mimics the amide I vibration, can reproduce the qualitative features of the quantum 2D IR spectra very well.⁸¹ As shown in these studies, the validity of the classical mechanically calculated spectra depends on the system-bath coupling. Notably, in their study, it was shown that the anharmonicity of the classical signal is determined almost entirely by the frequency resolution afforded by the simulated scan time t₃.

Classical nonlinear spectral simulations can reproduce spectra calculated with quantum methods qualitatively well when the system vibration is largely modulated by the bath, such as those examples discussed in this section. It should be noted, however, that any vibrational spectra are described as transitions from one vibrational level to another in quantum mechanics, whereas those are obtained from the fluctuation of dipole moment in classical mechanics. In quantum mechanics, wave functions or density matrices are determined by non-local information; i.e., the kinetic energy operator is expressed as the second derivative of coordinates in the coordinate representation. The anharmonic shift in quantum 2D IR spectra is thus determined by information about the whole potential energy profile. On the other hand, no discrete vibrational levels are considered in any classical approaches. In classical mechanics, a trajectory is determined by local information on coordinates and momenta. Consequently, the anharmonic shift corresponding to the frequency difference between the positive and negative peaks in the classical 2D IR spectra arises from the difference between the curvatures of trajectories perturbed by one and two electric fields. As a result, in many examples including those shown in this section, the anharmonic shifts in the classical 2D IR spectra tend to be smaller than those in the experimental or quantum mechanically calculated 2D IR spectra. In addition to the potential profile, the transition dipole moment is also important quantity for quantitative simulations of the IR and 2D IR spectra. In quantum mechanics, a spectral intensity is related to the transition dipole moments between discrete vibrational levels. On the other hand, the classical spectra are obtained by the dipole moment induced by vibrational and conformational changes. Accurate model potential and dipole moment are therefore extremely important to accurately simulate the 2D IR spectra of complicated molecular systems.

A fully quantum mechanical simulation of nuclear degrees of freedom including all the effects of surrounding thermal bath remains unpromising, despite the technological advancement of computers. Therefore, semiclassical approaches, such as (linearized) semiclassical-initial value representation,⁸² centroid molecular dynamics,⁸³ and ring-polymer molecular dynamics,⁸⁴ have been developed to accurately calculate the static and dynamic properties of molecules in condensed phases; for example, the IR spectrum of liquid water.⁸⁵ A formalism for two-time correlation functions has also been developed.⁸⁶ It is hoped that these semiclassical approaches will be able to provide a novel framework to calculate the 2D IR spectra incorporating the nuclear quantum effects by developing the formalism of nonlinear three-time correlation functions.⁸⁷

IV. NUMERICAL INTEGRATION OF THE SCHRÖDINGER EQUATION

The response functions described in Sec. II can be calculated with quantum-classical simulations. This essentially involves solving the time-dependent Schrödinger equation. In the first application of the method to systems involving more than one vibration, it was referred to as the Numerical Integration of the Schrödinger Equation (NISE) approach.^88–90 To evaluate the response functions, one must identify the essential coordinates directly coupled with light and treat those as weakly anharmonic oscillators represented by coupled three-level systems. All other coordinates are treated classically and only included through their time-dependent modulation of the parameters in the quantum system. In reality, the full vibrational quantum Hamiltonian will contain anharmonicities of all orders. For efficient calculations, one typically employs an effective model Hamiltonian, where all anharmonicities are collected in one effective quartic anharmonicity term sufficient to describe the energy fluctuations of the three energy levels crucial for the 2D IR spectroscopy. This approximation leads to an effective expression without anharmonicity terms that can cause relaxation between the three vibrational levels. The quantum system is, thus, typically described by the time-dependent effective model Hamiltonian

\begin{matrix} H (t) & = & \sum_{n}^{N} ℏ ω_{n} (t) B_{n}^{†} B_{n} + \sum_{n, m}^{N} J_{n m} (t) B_{n}^{†} B_{m} - \frac{1}{2} \sum_{n}^{N} Δ_{n} (t) B_{n}^{†} B_{n}^{†} B_{n} B_{n} \\ + \sum_{n}^{N} \vec{E} (t) \cdot {\vec{μ}}_{n} (t) (B_{n}^{†} + B_{n}) + \sum_{n}^{N} \vec{E} (t) \cdot {\bar{\bar{α}}}_{n} (t) \cdot \vec{E} (t) (B_{n}^{†} + B_{n}) . \end{matrix}

(30)

Here, $B_{n}^{†}$ and B_m are the Bosonic creation and annihilation operators for the N vibrations considered quantum mechanically, which are numbered with indices n and m, respectively. The individual local vibrations are characterized by their frequencies, ω_n(t), transition dipoles, ${\vec{μ}}_{n} (t)$ ⁠, transition polarizabilities, ${\bar{\bar{α}}}_{n} (t)$ ⁠, and anharmonicities, Δ_n(t). The different local vibrations are mixed by their mutual coupling, J_nm(t). This is a generalization of Eq. (9) applicable to coupled multi-chromophore systems. The time-dependence arises from the couplings of the quantum oscillators with the classical bath coordinates. The last two terms account for the interaction with the applied electric field(s), $\vec{E} (t)$ ⁠, and are included through the time-dependent perturbation theory that results in the response function formulation. Determining the fluctuating quantities in the Hamiltonian depends on the system under consideration and will be discussed in Sec. IV A. If we assume that these are known, the response functions can be calculated through the calculation of the time-evolution operators. To determine these, the solution of the time-dependent Schrödinger equation is needed. However, if time is divided into short enough intervals (Δt) that the Hamiltonian can be considered constant the trivial solution of the time-evolution during one time interval is given by the solution of the time-dependent Schrödinger equation with a time-independent Hamiltonian and the time-evolution operators for longer time intervals are obtained by successive application of time-evolution operators for neighboring time intervals.

In practical calculations, the Hamiltonian in Eq. (30) only includes a change in the number of vibrations in the terms involving the coupling with the external electric field(s) and thermalization is not included. Therefore, the remaining part of the Hamiltonian is block diagonal. This allows treating the ground state, single excited states, and double excited states, separately. As there is only one ground state and the energy of this is set to zero, the time-evolution operator is the unit operator. When N vibrational degrees of freedom are treated, there are N single excited states and N(N + 1)/2 double excited states. The time-evolution in each excitation manifold can be evaluated independently in the corresponding harmonic basis, which, in principle, requires the evaluation of matrix exponentials of the corresponding dimensionalities, which is typically done by diagonalizing the Hamiltonian, multiplying the eigenvalues with −iΔt/ℏ, taking the exponent and transforming back to the original basis. However, in practice, the time-evolution matrix for the double excited states can be made more efficient as the time required for the direct evaluation of the time-evolution operator scales as N,⁶ which is a costly procedure to repeat even for moderately sized systems. Among the solutions to this problem are the nonlinear exciton propagation scheme,⁹¹ which utilizes an anharmonic perturbation of the harmonic solution, and an approach based on the Trotter algorithm making use of the sparse nature of the Hamiltonian of the doubly excited states.^92,93 The latter allowed the calculation of the 2D IR spectra of systems up to 864 coupled vibrations for water ices^94,95 as well as applications to the amide I band of full proteins.^96–98 Numerous applications of the NISE method for 2D vibrational spectroscopy have been implemented when only one vibration is involved.^99–102 A few different implementations applicable to coupled systems have been presented.^88–93,103

A. Hamiltonian parameterization, frequency maps

The terms in the Hamiltonian in Eq. (30) have different origins, and their bath dependence can be parameterized in different ways. In the following, the different terms will be considered one by one. First, the vibrational frequency of a given mode depends on the local environment. The well-known Stark shift is a simple way to describe such frequency shift and fluctuation. Local electric fields modulate the vibrational frequency, and this type of effect can be parameterized by ab initio calculations of the vibrational frequency in different point charge environments of inhomogeneous electric fields. In this way, the solvent dependence of the vibrational frequencies of a number of common vibrations has been parameterized expressing the vibrational frequency in an expansion of the electric potential^104–108 or electric field^{102,109–125} and sometimes the electric field gradient^126–130 on specific points inside a molecule. The parameterization of the solvatochromic vibrational frequency shift parameters can either be obtained from ab initio calculations or empirical fitting to experimental data. These mappings then allow obtaining a frequency trajectory from a molecular dynamics trajectory. As typical frequencies fluctuate on a sub-picosecond time scale, the trajectories need to be sampled and saved every ∼20 fs and typical lengths are on the order of nanoseconds.

In reality, the solvatochromic vibrational frequency shift is not just an electrostatic effect. Charge transfer, dispersion forces, Pauli repulsion, polarization, and multipole effects may further play a role.^{121,125,131–137} Empirical models may compensate for this as long as other effects are correlated with electrostatic potential/field fluctuations. Ab initio calculations based on clusters, where the electrons of the solvent molecules are treated explicitly, may also be able to account for such effects in an efficient way. Detailed analysis of the amide I vibrations has shown that the various other effects tend to cancel out.^132,136 For other systems, this may not be the case.

The anharmonicity also depends on the solvent, and this dependence has been included in mappings.^102,126,128 Still, for many systems such as the amide I vibrations in proteins, the variation in the anharmonicity is small relative to the variation in the frequencies and the solvent dependence can be neglected using a constant anharmonicity value for each type of vibrational mode.

The couplings between different local modes, for long-range interactions, are well approximated by the transition-dipole coupling model.^138–143 Transition charge coupling models have been developed for intermediate range couplings.^130,144,145 These models account for some multipole contributions to the electrostatic interaction. For vibrational modes directly connected through covalent bonds, ab initio based coupling models have been developed. This has both been done for the OH stretch vibrations in the water molecule,¹⁴⁶ for neighboring amide I modes in a peptide chain,^{117,118,130,145,147,148} and amide I and II modes in the same and neighboring peptide units of proteins.^129,149 Ab initio modeling of the vibrational couplings in small peptides and DNA has also been reported based on the polarizable continuum model (PCM) description of solvent.^150,151 For the peptide units, this coupling depends on the local configuration, as characterized by the Ramachandran angles. The developed mappings were thus made by varying these angles systematically for small di- or tri-peptides and then calculate the vibrational coupling through ab initio calculations. Just as for the vibrational frequencies, the couplings can be extracted from molecular dynamics trajectories using these mappings.

Figure 6 illustrates a comparison between the spectra simulated for the OH stretch of water molecules diluted in acetonitrile and experimental data. The evolution of cross-peaks between the high-frequency asymmetric OH-stretch and the low-frequency symmetric OH-stretch is well described by the theory. Furthermore, the difference in the dynamics of the two peaks, which was the focus of the study,¹⁵² is revealed by the center line slope analysis. From the simulations, this could be interpreted in terms of non-Gaussian dynamics as the bath dynamics for strong hydrogen bond environments is faster than for weak hydrogen bond environments. As strong hydrogen bonded OH-stretches absorb at lower frequencies, the fast dynamics contributes more to the low frequency symmetric vibration peak. While the anharmonicity of OH-stretch modes is around 200 cm⁻¹, a sequential transition is observed with a much smaller apparent anharmonicity. This phenomenon arises as the large anharmonicity largely localizes the double excited states.

FIG. 6.

View large Download slide

2D IR spectra for parallel laser polarization of the OH-stretch of water dissolved in acetonitrile at different waiting times t₂₃ reproduced with permission from Jansen et al. J. Phys. Chem. A 113, 6260 (2009), Copyright 2009 American Chemical Society. (a) and (b) present experimental and simulated data, respectively. Dotted lines show the frequencies of symmetric and asymmetric stretching modes of the H₂O molecule. The 2D spectra are normalized to the maximal amplitude. Red contours illustrate bleach signals, while the cyan contours illustrate excited state absorption. The contour lines are drawn at 10% steps of the maximal amplitude in each individual plot. Solid black illustrated the max lines.

B. Mappings for interactions and the related techniques

A linear absorption spectrum can be calculated from the one-time (two-point) transition-dipole response function, while the 2D IR spectra are governed by a combination of six three-time (four-point) transition-dipole response functions related to rephasing and non-rephasing versions of the ground state bleach, stimulated emission, and excited state absorption pathways (Sec. II). The transition-dipoles of different states are reflected in the relative intensity of the peaks. Ultrafast reorientation of the transition-dipoles resulting from molecular reorientation^153–156 and vibrational energy transfer^72,153,157 can be monitored by measuring 2D IR spectra using different polarizations^158,159 of the applied laser fields. Cross peak intensities in the 2D IR spectra depend on the relative orientation of the transition dipoles of the two involved resonances. This allows for the determination of molecular geometries^160–162 and is powerful in the peak assignment. For many vibrations, the transition-dipole magnitude is fairly independent of the solvent environment. A clear exception of this is the OH-stretch vibration, which in water shows a five-fold change across the spectrum,¹⁶³ with the very weak dipole strength of the free OH-stretch vibration and very strong absorption of a strongly H-bonded OH species. Therefore, it is crucial to include the variation of the transition-dipole in the modeling. This has been done very similar to how the solvent effects are accounted for through mapping for the vibrational frequencies.^{102,128,130,163}

Raman spectra,^{102,146,164,165} sum-frequency generation,^166–169 and 2D sum-frequency generation^{17,18,168,170–173} spectroscopy signals can be calculated using similar response functions as for linear absorption and 2D IR spectroscopies. The only major change is that one needs to replace transition-dipoles in the expressions with transition-polarizabilities. The possible polarization setups available increase as the transition-polarizabilities represent two interactions with external visible laser fields, while the transition-dipoles represent one interaction with an infrared laser field. This opens a possibility of directly probing molecular orientational distribution. Most importantly in the sum-frequency generation type experiments, which are only sensitive to non-centrosymmetric signal contributions,^{167,168,174,175} the orientation of the vibrations relative to the sample surface can be determined. Just as the transition-dipole, the transition-polarizability depends on the environment. This has been studied for the OH-stretch of water using an electric field mapping. Still, the solvent effect both for the OH-stretch and other systems is relatively small.^102,146 The difference between the transition-dipole and the transition-polarizability solvent effects results in a non-coincidence effect^147,176,177 making the infrared and Raman type spectra exhibit significantly different lineshapes. In particular, Raman and SFG spectra in the OH-stretch region are much more sensitive to free OH-species than infrared spectra.¹⁴⁶

C. Limitations—Bandwidth, high-frequency modes

The quantum-classical methods discussed here have a number of important advantages and limitations. The methods are powerful in providing a quantum description of a given system and allow for treating essentially arbitrary bath dynamics, as provided by the combination of mappings and molecular dynamics simulations. For example, the methods allow describing non-Gaussian dynamics^{152,178–182} and chemical exchange^{101,155,179,183–192} arising when the bath coordinates are not harmonic. Commonly used methods invoking the second-order cumulant approximation^34,193 or methods assuming a harmonic bath spectral density^194–197 cannot account for such effects. The methods can use input from molecular dynamics simulations and are thus able to predict spectra starting from first principles. Sometimes it is not desirable to perform a full molecular dynamics simulation, and then stochastic models can be employed instead.^183,198

The Hamiltonian in Eq. (30) does not allow for relaxations between the different excitation manifolds. This means that vibrational energy relaxation is intrinsically neglected. Furthermore, the methods account for the effect that the bath exerts on the system, while the feedback of the system to the bath when the system is on an excited state is not included.^{140,199–201} This results in a lack of correct thermalization in the quantum system, which leads to an equal population of all quantum states after equilibration.²⁰¹ In practice, this means that artifacts can be expected at low temperatures, and when broad spectral ranges are considered, where the thermal energy is comparable to the width of the considered spectral region. In systems like water and ice, the OH-stretch band is broader than the thermal energy at room temperature, but still fairly good approximations of the spectra are found for such systems as long as the waiting time is not too long.⁷² For longer waiting times, persistent ground state bleach signals are observed in experiments resulting from the thermalization effect. A number of attempts to overcome this issue have been reported;^140–143 however, including feedback to the bath and thermalization effects that, generally, require simultaneous propagation of the bath modes and the system Hamiltonian resulting in the need for extensive computational time.

Another limitation of the method is that the modes treated quantum mechanically should be well defined. OH-stretch vibrations are an example of such well-defined modes as the OH-bond under normal circumstances does not break at ambient temperatures. On the contrary, hydrogen bonds form and break all the time and such low-frequency volatile bond vibrations are not suited for this quantum-classical treatment as their nature is dynamically changed on the short time-scale of the spectroscopy. If a vibration of interest is strongly coupled with many other modes, this also poses a problem as all the strongly coupled modes would need to be included in the system Hamiltonian for an accurate description requiring extensive parameterization and long simulation times. In such case, the methods discussed in Sec. III of this paper may be worth considering.

V. PERSPECTIVES AND A FEW CONCLUDING REMARKS

While 2D IR spectra contain valuable information on couplings, anharmonicity, and frequency correlation, more information can be obtained through higher-order methods generating multidimensional spectra. Three-dimensional infrared spectroscopy has been explored experimentally²⁰² and simulated^203,204 with a method based on that described in Sec. IV. Still the interpretation of such spectra remains very challenging. Infrared methods can also be combined with electronic spectroscopy^205,206 to reveal the vibrations on the electronic state or to utilize the longer lifetime of the electronic state to probe reaction dynamics. With fluorescence detection,²⁰⁷ 2D IR spectroscopy and microscopy would also be possible.²⁰⁸

In the NISE method described in Sec. IV, an essential step is the modeling and parameterization of the multi-chromophore Hamiltonian from electronic structure calculations on their static structures. By contrast, the classical mechanical approach in Sec. III relies on the accurate description of the molecular vibrational property in the context of classical dynamics. This often requires the use of sophisticated molecular mechanics potential models^21,209 or quantum mechanical potentials. In this regard, recently proposed ab initio theories of vibrational solvatochromism^135–137 and direct QM/MM or full ab initio MD simulations of vibrational spectra^210–213 can be fruitfully adapted for either approach to improve their efficiency or accuracy.

2D vibrational spectra can be calculated using the exact hierarchical equations of motion (HEOM) approach.^80,214–215 It should be, however, noted that this HEOM approach is based on the assumption of a spectral density description of the bath, which means that the bath is modelled as a collection of independent harmonic oscillators, i.e., the Caldeira-Leggett quantum dissipative bath model.²¹⁶ Under this assumption, the quantum correlation between the system and the bath can be fully accounted for. However, this method scales rather unfavorably, limiting its application to relatively small systems even though a number of approximations have been developed to improve efficiency.²¹⁷ Another limitation is that the method requires the parameterization of the spectral density.

In summary, we have presented a brief introduction to the fundamental theory and computational methods used to numerically simulate linear and nonlinear vibrational response functions and the corresponding spectra. One of the widely used methods is to approximate the associated quantum mechanical nonlinear response function with invoking classical approximations so that the linear and nonlinear vibrational response functions can be evaluated by using classical MD simulation trajectories of either equilibrium or non-equilibrium systems. The second method of choice is to solve the time-dependent Schrödinger equation of coupled oscillators where their frequencies, coupling constants, and anharmonicities that are fluctuating in time due to the system-bath interactions can be obtained from independent computational methods. Here, a few representative cases were presented and discussed. Over the past decade, a variety of 2D vibrational spectroscopic techniques that utilize femtosecond IR, THz, and visible pulses have been developed and used to study large-scale delocalized modes in condensed phases and ultrafast reaction dynamics during chemical and biological reactions. Thus, we anticipate that the computational and theoretical methods for accurate calculations of coherent multidimensional vibrational spectroscopic signals of coupled multi-chromophore systems could be incorporated into various MD simulation and quantum chemistry calculation program packages to help the experimentalists in this and related areas to interpret their experimental results and to further understand underlying principles, mechanisms, and functions of materials and molecules in condensed phases.

ACKNOWLEDGMENTS

This work was supported by IBS-R023-D1 (M.C.) and JSPS KAKENHI under Grant No. JP16H02254 (S.S.).

The authors declare no competing financial interests.

REFERENCES

1.

P.

Hamm

,

M.

Lim

, and

R. M.

Hochstrasser

,

J. Phys. Chem. B

102

,

6123

(

1998

).

https://doi.org/10.1021/jp9813286

Crossref

2.

M.

Khalil

and

A.

Tokmakoff

,

Chem. Phys.

266

,

213

(

2001

).

https://doi.org/10.1016/s0301-0104(01)00230-0

Crossref

3.

J. C.

Wright

,

Int. Rev. Phys. Chem.

21

,

185

(

2002

).

https://doi.org/10.1080/01442350210124506

Crossref

4.

M.

Khalil

,

N.

Demirdöven

, and

A.

Tokmakoff

,

J. Phys. Chem. A

107

,

5258

(

2003

).

https://doi.org/10.1021/jp0219247

Crossref

5.

M.

Cho

,

Chem. Rev.

108

,

1331

(

2008

).

https://doi.org/10.1021/cr078377b

Crossref

PubMed

6.

S. T.

Roberts

,

J. J.

Loparo

, and

A.

Tokmakoff

,

J. Chem. Phys.

125

,

084502

(

2006

).

https://doi.org/10.1063/1.2232271

Crossref

PubMed

7.

D. A.

McQuarrie

,

Statistical Mechanics

(

Harper & Row

,

New York

,

1976

),

Google Scholar

8.

M.

Cho

,

Two-Dimensional Optical Spectroscopy

(

CRC Press

,

Boca Raton

,

2009

).

Google Scholar

Crossref

9.

M.

Cho

, in

Advances in Multi-Photon Processes and Spectroscopy

, edited by

S. H.

Lin

,

A. A.

Villaeys

, and

Y.

Fujimura

(

World Scientific Publishing Co.

,

Singapore

,

1999

), p.

229

.

Google Scholar

Crossref

10.

M.

Cho

,

PhysChemComm

5

,

40

(

2002

).

https://doi.org/10.1039/b110898k

Crossref

11.

K.

Okumura

and

Y.

Tanimura

,

J. Chem. Phys.

106

,

1687

(

1997

).

https://doi.org/10.1063/1.473284

Crossref

12.

K.

Okumura

and

Y.

Tanimura

,

J. Chem. Phys.

107

,

2267

(

1997

).

https://doi.org/10.1063/1.474604

Crossref

13.

K.

Park

and

M.

Cho

,

J. Chem. Phys.

109

,

10559

(

1998

).

https://doi.org/10.1063/1.477756

Crossref

14.

J.

Sung

and

M.

Cho

,

J. Chem. Phys.

113

,

7072

(

2000

).

https://doi.org/10.1063/1.1312276

Crossref

15.

K.

Park

and

M.

Cho

,

J. Chem. Phys.

112

,

5021

(

2000

).

https://doi.org/10.1063/1.481091

Crossref

16.

K.

Park

and

M.

Cho

,

J. Chem. Phys.

112

,

10496

(

2000

).

https://doi.org/10.1063/1.481684

Crossref

17.

J.

Bredenbeck

,

A.

Ghosh

,

M.

Smits

, and

M.

Bonn

,

J. Am. Chem. Soc.

130

,

2152

(

2008

).

https://doi.org/10.1021/ja710099c

Crossref

PubMed

18.

W.

Xiong

,

J. E.

Laaser

,

R. D.

Mehlenbacher

, and

M. T.

Zanni

,

Proc. Natl. Acad. Sci. U. S. A.

108

,

20902

(

2011

).

https://doi.org/10.1073/pnas.1115055108

Crossref

PubMed

19.

Y.

Tanimura

and

S.

Mukamel

,

J. Chem. Phys.

99

,

9496

(

1993

).

https://doi.org/10.1063/1.465484

Crossref

20.

W.

Kuehn

,

K.

Reimann

,

M.

Woerner

,

T.

Elsaesser

,

R.

Hey

, and

U.

Schade

,

Phys. Rev. Lett.

107

,

067401

(

2011

).

https://doi.org/10.1103/physrevlett.107.067401

Crossref

PubMed

21.

J.

Savolainen

,

S.

Ahmed

, and

P.

Hamm

,

Proc. Natl. Acad. Sci. U. S. A.

110

,

20402

(

2013

).

https://doi.org/10.1073/pnas.1317459110

Crossref

PubMed

22.

A.

Shalit

,

S.

Ahmed

,

J.

Savolainen

, and

P.

Hamm

,

Nat. Chem.

9

,

273

(

2016

).

https://doi.org/10.1038/nchem.2642

Crossref

PubMed

23.

H.

Ito

,

T.

Hasegawa

, and

Y.

Tanimura

,

J. Chem. Phys.

141

,

124503

(

2014

).

https://doi.org/10.1063/1.4895908

Crossref

PubMed

24.

H.

Ito

,

J.-Y.

Jo

, and

Y.

Tanimura

,

Struct. Dyn.

2

,

054102

(

2015

).

https://doi.org/10.1063/1.4932597

Crossref

PubMed

25.

K.

Park

,

M.

Cho

,

S.

Hahn

, and

D.

Kim

,

J. Chem. Phys.

111

,

4131

(

1999

).

https://doi.org/10.1063/1.479710

Crossref

26.

M.

Cho

,

Phys. Rev. A

61

,

023406

(

2000

).

https://doi.org/10.1103/physreva.61.023406

Crossref

27.

W.

Zhao

and

J. C.

Wright

,

Phys. Rev. Lett.

84

,

1411

(

2000

).

https://doi.org/10.1103/physrevlett.84.1411

Crossref

PubMed

28.

P. M.

Donaldson

,

R.

Guo

,

F.

Fournier

,

E. M.

Gardner

,

L. M. C.

Barter

,

C. J.

Barnett

,

I. R.

Gould

,

D. R.

Klug

,

D. J.

Palmer

, and

K. R.

Willison

,

J. Chem. Phys.

127

,

114513

(

2007

).

https://doi.org/10.1063/1.2771176

Crossref

PubMed

29.

F.

Fournier

,

R.

Guo

,

E. M.

Gardner

,

P. M.

Donaldson

,

C.

Loeffeld

,

I. R.

Gould

,

K. R.

Willison

, and

D. R.

Klug

,

Acc. Chem. Res.

42

,

1322

(

2009

).

https://doi.org/10.1021/ar900074p

Crossref

PubMed

30.

Y. R.

Shen

,

The Principle of Nonlinear Optics

(

Wiley

,

New York

,

1984

).

Google Scholar

31.

S.

Mukamel

,

Principles of Nonlinear Optical Spectroscopy

(

Oxford University Press

,

Oxford

,

1995

).

Google Scholar

32.

N.

Belabas

and

D. M.

Jonas

,

J. Opt. Soc. Am. B

22

,

655

(

2005

).

https://doi.org/10.1364/josab.22.000655

Crossref

33.

J.

Sung

,

R. J.

Silbey

, and

M.

Cho

,

J. Chem. Phys.

115

,

1422

(

2001

).

https://doi.org/10.1063/1.1379751

Crossref

34.

S.

Mukamel

,

Phys. Rev. A

28

,

3480

(

1983

).

https://doi.org/10.1103/physreva.28.3480

Crossref

35.

M.

Cho

,

J. Chem. Phys.

115

,

4424

(

2001

).

https://doi.org/10.1063/1.1389844

Crossref

36.

K.

Kwac

and

M.

Cho

,

J. Phys. Chem. A

107

,

5903

(

2003

).

https://doi.org/10.1021/jp034727w

Crossref

37.

K.

Lazonder

,

M. S.

Pshenichnikov

, and

D. A.

Wiersma

,

Opt. Lett.

31

,

3354

(

2006

).

https://doi.org/10.1364/ol.31.003354

Crossref

PubMed

38.

S.

Mukamel

,

V.

Khidekel

, and

V.

Chernyak

,

Phys. Rev. E

53

,

R1

(

1996

).

https://doi.org/10.1103/physreve.53.r1

Crossref

39.

S.

Saito

and

I.

Ohmine

,

J. Chem. Phys.

108

,

240

(

1998

).

https://doi.org/10.1063/1.475375

Crossref

40.

A.

Ma

and

R. M.

Stratt

,

Phys. Rev. Lett.

85

,

1004

(

2000

).

https://doi.org/10.1103/physrevlett.85.1004

Crossref

PubMed

41.

S.

Saito

and

I.

Ohmine

,

Phys. Rev. Lett.

88

,

207401

(

2002

).

https://doi.org/10.1103/physrevlett.88.207401

Crossref

PubMed

42.

J.

Cao

,

S.

Yang

, and

J.

Wu

,

J. Chem. Phys.

116

,

3760

(

2002

).

https://doi.org/10.1063/1.1445746

Crossref

43.

A.

Ma

and

R. M.

Stratt

,

J. Chem. Phys.

116

,

4962

(

2002

).

https://doi.org/10.1063/1.1453401

Crossref

44.

R. A.

Denny

and

D. R.

Reichman

,

J. Chem. Phys.

116

,

1987

(

2002

).

https://doi.org/10.1063/1.1431279

Crossref

45.

R.

DeVane

,

C.

Ridley

,

B.

Space

, and

T.

Keyes

,

J. Chem. Phys.

119

,

6073

(

2003

).

https://doi.org/10.1063/1.1601607

Crossref

46.

S.

Saito

and

I.

Ohmine

,

J. Chem. Phys.

125

,

084506

(

2006

).

https://doi.org/10.1063/1.2232254

Crossref

PubMed

47.

T.

Hasegawa

and

Y.

Tanimura

,

J. Chem. Phys.

128

,

064511

(

2008

).

https://doi.org/10.1063/1.2828189

Crossref

PubMed

48.

T.

Yagasaki

and

S.

Saito

,

J. Chem. Phys.

128

,

154521

(

2008

).

https://doi.org/10.1063/1.2903470

Crossref

PubMed

49.

J.

Jeon

and

M.

Cho

,

New J. Phys.

12

,

065001

(

2010

).

https://doi.org/10.1088/1367-2630/12/6/065001

Crossref

50.

H.

Goldstein

,

Classical Mechanics

, 2nd ed. (

Addison-Wesley

,

Reading

,

1980

).

Google Scholar

51.

A.

Lichtenberg

and

M.

Lieberman

,

Regular and Chaotic Dynamics

(

Springer-Verlag

,

New York

,

1992

).

Google Scholar

Crossref

52.

W. H.

Miller

,

J. Chem. Phys.

53

,

1949

(

1970

).

https://doi.org/10.1063/1.1674275

Crossref

53.

H.

Nakamura

,

J. Phys. Chem. A

110

,

10929

(

2006

).

https://doi.org/10.1021/jp063643v

Crossref

PubMed

54.

J.

Jeon

and

M.

Cho

,

J. Phys. Chem. B

118

,

8148

(

2014

).

https://doi.org/10.1021/jp501182d

Crossref

PubMed

55.

J. B.

Asbury

,

T.

Steinel

,

K.

Kwak

,

S. A.

Corcelli

,

C. P.

Lawrence

,

J. L.

Skinner

, and

M. D.

Fayer

,

J. Chem. Phys.

121

,

12431

(

2004

).

https://doi.org/10.1063/1.1818107

Crossref

PubMed

56.

S.

Mukamel

and

J. B.

Maddox

,

J. Chem. Phys.

121

,

36

(

2004

).

https://doi.org/10.1063/1.1756582

Crossref

PubMed

57.

T. L. C.

Jansen

,

J. G.

Snijders

, and

K.

Duppen

,

J. Chem. Phys.

113

,

307

(

2000

).

https://doi.org/10.1063/1.481795

Crossref

58.

T. L. C.

Jansen

,

J. G.

Snijders

, and

K.

Duppen

,

J. Chem. Phys.

114

,

10910

(

2001

).

https://doi.org/10.1063/1.1374959

Crossref

59.

T. l. C.

Jansen

,

K.

Duppen

, and

J. G.

Snijders

,

Phys. Rev. B

67

,

134206

(

2003

).

https://doi.org/10.1103/physrevb.67.134206

Crossref

60.

T.

Hasegawa

and

Y.

Tanimura

,

J. Chem. Phys.

125

,

074512

(

2006

).

https://doi.org/10.1063/1.2217947

Crossref

PubMed

61.

T.

Yagasaki

and

S.

Saito

,

Acc. Chem. Res.

42

,

1250

(

2009

).

https://doi.org/10.1021/ar900007s

Crossref

PubMed

62.

S.

Ashihara

,

N.

Huse

,

A.

Espagna

,

E. T. J.

Nibbering

, and

T.

Elsaesser

,

J. Phys. Chem. A

111

,

743

(

2007

).

https://doi.org/10.1021/jp0676538

Crossref

PubMed

63.

L.

Chieffo

,

J.

Shattuck

,

J. J.

Amsden

,

S.

Erramilli

, and

L. D.

Ziegler

,

Chem. Phys.

341

,

71

(

2007

).

https://doi.org/10.1016/j.chemphys.2007.06.024

Crossref

64.

J.

Lindner

,

D.

Cringus

,

M. S.

Pshenichnikov

, and

P.

Vöhringer

,

Chem. Phys.

341

,

326

(

2007

).

https://doi.org/10.1016/j.chemphys.2007.07.051

Crossref

65.

T.

Yagasaki

and

S.

Saito

,

J. Chem. Phys.

134

,

184503

(

2011

).

https://doi.org/10.1063/1.3587105

Crossref

PubMed

66.

T.

Yagasaki

and

S.

Saito

,

J. Chem. Phys.

135

,

244511

(

2011

).

https://doi.org/10.1063/1.3671993

Crossref

PubMed

67.

J.

Jeon

and

M.

Cho

,

J. Chem. Phys.

135

,

214504

(

2011

).

https://doi.org/10.1063/1.3658876

Crossref

PubMed

68.

J.

Jeon

,

J. H.

Lim

,

S.

Kim

,

H.

Kim

, and

M.

Cho

,

J. Phys. Chem. A

119

,

5356

(

2015

).

https://doi.org/10.1021/jp510157y

Crossref

PubMed

69.

S.

Imoto

,

S. S.

Xantheas

, and

S.

Saito

,

J. Chem. Phys.

139

,

044503

(

2013

).

https://doi.org/10.1063/1.4813071

Crossref

PubMed

70.

J. D.

Eaves

,

J. J.

Loparo

,

C. J.

Fecko

,

S. T.

Roberts

,

A.

Tokmakoff

, and

P. L.

Geissler

,

Proc. Natl. Acad. Sci. U. S. A.

102

,

13019

(

2005

).

https://doi.org/10.1073/pnas.0505125102

Crossref

PubMed

71.

A.

Paarmann

,

T.

Hayashi

,

S.

Mukamel

, and

R. J. D.

Miller

,

J. Chem. Phys.

128

,

191103

(

2008

).

https://doi.org/10.1063/1.2919050

Crossref

PubMed

72.

T. L. C.

Jansen

,

B. M.

Auer

,

M.

Yang

, and

J. L.

Skinner

,

J. Chem. Phys.

132

,

224503

(

2010

).

https://doi.org/10.1063/1.3454733

Crossref

PubMed

73.

A. J.

Lock

and

H. J.

Bakker

,

J. Chem. Phys.

117

,

1708

(

2002

).

https://doi.org/10.1063/1.1485966

Crossref

74.

D.

Kraemer

,

M. L.

Cowan

,

A.

Paarmann

,

N.

Huse

,

E. T.

Nibbering

,

T.

Elsaesser

, and

R. J.

Miller

,

Proc. Natl. Acad. Sci. U. S. A.

105

,

437

(

2008

).

https://doi.org/10.1073/pnas.0705792105

Crossref

PubMed

75.

K.

Ramasesha

,

L. D.

Marco

,

A.

Mandal

, and

A.

Tokmakoff

,

Nat. Chem.

5

,

935

(

2013

).

https://doi.org/10.1038/nchem.1757

Crossref

PubMed

76.

C.

Dellago

and

S.

Mukamel

,

Phys. Rev. E

67

,

035205

(

2003

).

https://doi.org/10.1103/physreve.67.035205

Crossref

77.

W. G.

Noid

,

G. S.

Ezra

, and

R. F.

Loring

,

J. Phys. Chem. B

108

,

6536

(

2004

).

https://doi.org/10.1021/jp036749o

Crossref

78.

M.

Kryvohuz

and

J.

Cao

,

Phys. Rev. Lett.

96

,

030403

(

2006

).

https://doi.org/10.1103/physrevlett.96.030403

Crossref

PubMed

79.

S. V.

Malinin

and

V. Y.

Chernyak

,

Phys. Rev. E

77

,

056201

(

2008

).

https://doi.org/10.1103/physreve.77.056201

Crossref

80.

A.

Sakurai

and

Y.

Tanimura

,

J. Phys. Chem. A

115

,

4009

(

2011

).

https://doi.org/10.1021/jp1095618

Crossref

PubMed

81.

M.

Reppert

and

P.

Brumer

,

J. Chem. Phys.

148

,

064101

(

2018

).

https://doi.org/10.1063/1.5017985

Crossref

PubMed

82.

W. H.

Miller

,

J. Phys. Chem. A

105

,

2942

(

2001

).

https://doi.org/10.1021/jp003712k

Crossref

83.

G. A.

Voth

, “

Path-integral centroid methods in quantum statistical mechanics and dynamics

,” in

New Methods in Computational Quantum Mechanics

,

Advances in Chemical Physics

Vol. 93, edited by

I.

Prigogine

and

S. A.

Rice

(

Wiley, 1996

), pp.

135-218

.

https://doi.org/10.1002/9780470141526.ch4

84.

S.

Habershon

,

D. E.

Manolopoulos

,

T. E.

Markland

, and

T. F.

Miller

,

Annu. Rev. Phys. Chem.

64

,

387

(

2013

).

https://doi.org/10.1146/annurev-physchem-040412-110122

Crossref

PubMed

85.

J.

Liu

,

W. H.

Miller

,

G. S.

Fanourgakis

,

S. S.

Xantheas

,

S.

Imoto

, and

S.

Saito

,

J. Chem. Phys.

135

,

244503

(

2011

).

https://doi.org/10.1063/1.3670960

Crossref

PubMed

86.

D. R.

Reichman

,

P.-N.

Roy

,

S.

Jang

, and

G. A.

Voth

,

J. Chem. Phys.

113

,

919

(

2000

).

https://doi.org/10.1063/1.481872

Crossref

87.

D. R.

Moberg

,

M.

Alemi

, and

R. F.

Loring

,

J. Chem. Phys.

143

,

084101

(

2015

).

https://doi.org/10.1063/1.4929377

Crossref

PubMed

88.

T. l. C.

Jansen

and

J.

Knoester

,

J. Phys. Chem. B

110

,

22910

(

2006

).

https://doi.org/10.1021/jp064795t

Crossref

PubMed

89.

H.

Torii

,

Vib. Spectrosc.

42

,

140

(

2006

).

https://doi.org/10.1016/j.vibspec.2006.04.002

Crossref

90.

M.

Kobus

,

R. D.

Gorbunov

,

P. H.

Nguyen

, and

G.

Stock

,

Chem. Phys.

347

,

208

(

2008

).

https://doi.org/10.1016/j.chemphys.2007.10.034

Crossref

91.

C.

Falvo

,

B.

Palmieri

, and

S.

Mukamel

,

J. Chem. Phys.

130

,

184501

(

2009

).

https://doi.org/10.1063/1.3120771

Crossref

PubMed

92.

A.

Paarmann

,

T.

Hayashi

,

S.

Mukamel

, and

R. J. D.

Miller

,

J. Chem. Phys.

130

,

204110

(

2009

).

https://doi.org/10.1063/1.3139003

Crossref

PubMed

93.

C.

Liang

and

T. L. C.

Jansen

,

J. Chem. Theory Comput.

8

,

1706

(

2012

).

https://doi.org/10.1021/ct300045c

Crossref

PubMed

94.

L.

Shi

,

J. L.

Skinner

, and

T. L. C.

Jansen

,

Phys. Chem. Chem. Phys.

18

,

3772

(

2016

).

https://doi.org/10.1039/c5cp07264f

Crossref

PubMed

95.

H.

Tran

,

A. V.

Cunha

,

J. J.

Shephard

,

A.

Shalit

,

P.

Hamm

,

T. L. C.

Jansen

, and

C. G.

Salzmann

,

J. Chem. Phys.

147

,

144501

(

2017

).

https://doi.org/10.1063/1.4993952

Crossref

PubMed

96.

C.

Liang

,

M.

Louhivuori

,

S. J.

Marrink

,

T. L. C.

Jansen

, and

J.

Knoester

,

J. Phys. Chem. Lett.

4

,

448

(

2013

).

https://doi.org/10.1021/jz3019258

Crossref

PubMed

97.

A. S.

Bondarenko

and

T. L. C.

Jansen

,

J. Chem. Phys.

142

,

212437

(

2015

).

https://doi.org/10.1063/1.4919716

Crossref

PubMed

98.

A. V.

Cunha

,

A. S.

Bondarenko

, and

T. L. C.

Jansen

,

J. Chem. Theory Comput.

12

,

3982

(

2016

).

https://doi.org/10.1021/acs.jctc.6b00420

Crossref

PubMed

99.

K.

Kwac

and

M.

Cho

,

J. Chem. Phys.

119

,

2256

(

2003

).

https://doi.org/10.1063/1.1580808

Crossref

100.

T. l. C.

Jansen

,

W.

Zhuang

, and

S.

Mukamel

,

J. Chem. Phys.

121

,

10577

(

2004

).

https://doi.org/10.1063/1.1807824

Crossref

PubMed

101.

M. F.

DeCamp

,

L.

DeFlores

,

J. M.

McCracken

,

A.

Tokmakoff

,

K.

Kwac

, and

M.

Cho

,

J. Phys. Chem. B

109

,

11016

(

2005

).

https://doi.org/10.1021/jp050257p

Crossref

PubMed

102.

B.

Auer

,

R.

Kumar

,

J. R.

Schmidt

, and

J. L.

Skinner

,

Proc. Natl. Acad. Sci. U. S. A.

104

,

14215

(

2007

).

https://doi.org/10.1073/pnas.0701482104

Crossref

PubMed

103.

C. R.

Baiz

,

K. J.

Kubarych

,

E.

Geva

, and

E. L.

Sibert

,

J. Phys. Chem. A

115

,

5354

(

2011

).

https://doi.org/10.1021/jp201641h

Crossref

PubMed

104.

S.

Ham

and

M.

Cho

,

J. Chem. Phys.

118

,

6915

(

2003

).

https://doi.org/10.1063/1.1559681

Crossref

105.

K.

Kwac

and

M.

Cho

,

J. Chem. Phys.

119

,

2247

(

2003

).

https://doi.org/10.1063/1.1580807

Crossref

106.

J.-H.

Choi

,

S.

Ham

, and

M.

Cho

,

J. Phys. Chem. B

107

,

9132

(

2003

).

https://doi.org/10.1021/jp034835i

Crossref

107.

J.-H.

Choi

,

K.-I.

Oh

,

H.

Lee

,

C.

Lee

, and

M.

Cho

,

J. Chem. Phys.

128

,

134506

(

2008

).

https://doi.org/10.1063/1.2844787

Crossref

PubMed

108.

J.-H.

Choi

and

M.

Cho

,

J. Chem. Phys.

134

,

154513

(

2011

).

https://doi.org/10.1063/1.3580776

Crossref

PubMed

109.

S. A.

Corcelli

,

C. P.

Lawrence

, and

J. L.

Skinner

,

J. Chem. Phys.

120

,

8107

(

2004

).

https://doi.org/10.1063/1.1683072

Crossref

PubMed

110.

Y. S.

Lin

,

J. M.

Shorb

,

P.

Mukherjee

,

M. T.

Zanni

, and

J. L.

Skinner

,

J. Phys. Chem. B

113

,

592

(

2009

).

https://doi.org/10.1021/jp807528q

Crossref

PubMed

111.

L.

Wang

,

C. T.

Middleton

,

M. T.

Zanni

, and

J. L.

Skinner

,

J. Phys. Chem. B

115

,

3713

(

2011

).

https://doi.org/10.1021/jp200745r

Crossref

PubMed

112.

M.

Reppert

and

A.

Tokmakoff

,

J. Chem. Phys.

138

,

134116

(

2013

).

https://doi.org/10.1063/1.4798938

Crossref

PubMed

113.

M.

Reppert

and

A.

Tokmakoff

,

J. Chem. Phys.

143

,

061102

(

2015

).

https://doi.org/10.1063/1.4928637

Crossref

PubMed

114.

T. M.

Watson

and

J. D.

Hirst

,

J. Phys. Chem. A

107

,

6843

(

2003

).

https://doi.org/10.1021/jp0344500

Crossref

115.

T. M.

Watson

and

J. D.

Hirst

,

Mol. Phys.

103

,

1531

(

2005

).

https://doi.org/10.1080/00268970500052387

Crossref

116.

R. D.

Gorbunov

,

D. S.

Kosov

, and

G.

Stock

,

J. Chem. Phys.

122

,

224904

(

2005

).

https://doi.org/10.1063/1.1898215

Crossref

PubMed

117.

R. D.

Gorbunov

and

G.

Stock

,

Chem. Phys. Lett.

437

,

272

(

2007

).

https://doi.org/10.1016/j.cplett.2007.02.033

Crossref

118.

H.

Maekawa

and

N.-H.

Ge

,

J. Phys. Chem. B

114

,

1434

(

2010

).

https://doi.org/10.1021/jp908695g

Crossref

PubMed

119.

K. E.

Furse

and

S. A.

Corcelli

,

J. Chem. Theory Comput.

5

,

1959

(

2009

).

https://doi.org/10.1021/ct9001416

Crossref

PubMed

120.

C. S.

Miller

and

S. A.

Corcelli

,

J. Phys. Chem. B

114

,

8565

(

2010

).

https://doi.org/10.1021/jp1028596

Crossref

PubMed

121.

S.

Li

,

J. R.

Schmidt

,

S. A.

Corcelli

,

C. P.

Lawrence

, and

J. L.

Skinner

,

J. Chem. Phys.

124

,

204110

(

2006

).

https://doi.org/10.1063/1.2200690

Crossref

PubMed

122.

N. M.

Levinson

,

E. E.

Bolte

,

C. S.

Miller

,

S. A.

Corcelli

, and

S. G.

Boxer

,

J. Am. Chem. Soc.

133

,

13236

(

2011

).

https://doi.org/10.1021/ja2042589

Crossref

PubMed

123.

C.

Liang

,

K.

Kwak

, and

M.

Cho

,

J. Phys. Chem. Lett.

8

,

5779

(

2017

).

https://doi.org/10.1021/acs.jpclett.7b02623

Crossref

PubMed

124.

K.

Cai

,

C.

Han

, and

J.

Wang

,

Phys. Chem. Chem. Phys.

11

,

9149

(

2009

).

https://doi.org/10.1039/b910269h

Crossref

PubMed

125.

E.

Małolepsza

and

J. E.

Straub

,

J. Phys. Chem. B

118

,

7848

(

2014

).

https://doi.org/10.1021/jp412827s

Crossref

PubMed

126.

T.

Hayashi

,

T. l. C.

Jansen

,

W.

Zhuang

, and

S.

Mukamel

,

J. Phys. Chem. A

109

,

64

(

2005

).

https://doi.org/10.1021/jp046685x

Crossref

PubMed

127.

T.

Hayashi

,

W.

Zhuang

, and

S.

Mukamel

,

J. Phys. Chem. A

109

,

9747

(

2005

).

https://doi.org/10.1021/jp052324l

Crossref

PubMed

128.

T. I. C.

Jansen

and

J.

Knoester

,

J. Chem. Phys.

124

,

044502

(

2006

).

https://doi.org/10.1063/1.2148409

Crossref

PubMed

129.

R.

Bloem

,

A. G.

Dijkstra

,

T. l. C.

Jansen

, and

J.

Knoester

,

J. Chem. Phys.

129

,

055101

(

2008

).

https://doi.org/10.1063/1.2961020

Crossref

PubMed

130.

S.

Roy

,

J.

Lessing

,

G.

Meisl

,

Z.

Ganim

,

A.

Tokmakoff

,

J.

Knoester

, and

T. l. C.

Jansen

,

J. Chem. Phys.

135

,

234507

(

2011

).

https://doi.org/10.1063/1.3665417

Crossref

PubMed

131.

T. L. C.

Jansen

,

J. Phys. Chem. B

118

,

8162

(

2014

).

https://doi.org/10.1021/jp5012445

Crossref

PubMed

132.

H.

Torii

,

J. Phys. Chem. Lett.

6

,

727

(

2015

).

https://doi.org/10.1021/acs.jpclett.5b00004

Crossref

PubMed

133.

H.

Torii

,

J. Phys. Chem. B

114

,

13403

(

2010

).

https://doi.org/10.1021/jp106952q

Crossref

PubMed

134.

H.

Torii

,

J. Phys. Chem. B

112

,

8737

(

2008

).

https://doi.org/10.1021/jp801364s

Crossref

PubMed

135.

B.

Błasiak

,

H.

Lee

, and

M.

Cho

,

J. Chem. Phys.

139

,

044111

(

2013

).

https://doi.org/10.1063/1.4816041

Crossref

PubMed

136.

B.

Błasiak

and

M.

Cho

,

J. Chem. Phys.

140

,

164107

(

2014

).

https://doi.org/10.1063/1.4872040

Crossref

PubMed

137.

B.

Błasiak

and

M.

Cho

,

J. Chem. Phys.

143

,

164111

(

2015

).

https://doi.org/10.1063/1.4934667

Crossref

PubMed

138.

S.

Krimm

and

J.

Bandekar

,

Adv. Protein Chem.

38

,

181

(

1986

).

https://doi.org/10.1016/s0065-3233(08)60528-8

Crossref

PubMed

139.

H.

Torii

and

M.

Tasumi

,

J. Chem. Phys.

96

,

3379

(

1992

).

https://doi.org/10.1063/1.461939

Crossref

140.

Q.

Shi

and

E.

Geva

,

J. Chem. Phys.

129

,

124505

(

2008

).

https://doi.org/10.1063/1.2981566

Crossref

PubMed

141.

G.

Hanna

and

E.

Geva

,

J. Phys. Chem. B

113

,

9278

(

2009

).

https://doi.org/10.1021/jp902797z

Crossref

PubMed

142.

K.

Kwac

and

E.

Geva

,

J. Phys. Chem. B

116

,

2856

(

2012

).

https://doi.org/10.1021/jp211792j

Crossref

PubMed

143.

C. P.

van der Vegte

,

S.

Knop

,

P.

Vöhringer

,

J.

Knoester

, and

T. L. C.

Jansen

,

J. Phys. Chem. B

118

,

6256

(

2014

).

https://doi.org/10.1021/jp504267h

Crossref

PubMed

144.

P.

Hamm

and

S.

Woutersen

,

Bull. Chem. Soc. Jpn.

75

,

985

(

2002

).

https://doi.org/10.1246/bcsj.75.985

Crossref

145.

T. l. C.

Jansen

,

A. G.

Dijkstra

,

T. M.

Watson

,

J. D.

Hirst

, and

J.

Knoester

,

J. Chem. Phys.

125

,

044312

(

2006

).

https://doi.org/10.1063/1.2218516

Crossref

146.

B. M.

Auer

and

J. L.

Skinner

,

J. Chem. Phys.

128

,

224511

(

2008

).

https://doi.org/10.1063/1.2925258

Crossref

PubMed

147.

H.

Torii

and

M.

Tasumi

,

J. Raman Spectrosc.

29

,

81

(

1998

).

https://doi.org/10.1002/(sici)1097-4555(199801)29:1<81::aid-jrs214>3.3.co;2-8

Crossref

148.

S.

Ham

,

S.

Cha

,

J.-H.

Choi

, and

M.

Cho

,

J. Chem. Phys.

119

,

1451

(

2003

).

https://doi.org/10.1063/1.1581855

Crossref

149.

A. G.

Dijkstra

,

T. L. C.

Jansen

, and

J.

Knoester

,

J. Phys. Chem. A

114

,

7315

(

2010

).

https://doi.org/10.1021/jp9111124

Crossref

PubMed

150.

A.

Biancardi

,

C.

Cappelli

,

B.

Mennucci

, and

R.

Cammi

,

J. Phys. Chem. B

114

,

4924

(

2010

).

https://doi.org/10.1021/jp100634p

Crossref

151.

A.

Biancardi

,

R.

Cammi

,

C.

Cappelli

,

B.

Mennucci

, and

J.

Tomasi

,

Theor. Chem. Acc.

131

,

1157

(

2012

).

https://doi.org/10.1007/s00214-012-1157-3

Crossref

152.

T. l. C.

Jansen

,

D.

Cringus

, and

M. S.

Pshenichnikov

,

J. Phys. Chem. A

113

,

6260

(

2009

).

https://doi.org/10.1021/jp900480r

Crossref

PubMed

153.

K. J.

Gaffney

,

I. R.

Piletic

, and

M. D.

Fayer

,

J. Chem. Phys.

118

,

2270

(

2003

).

https://doi.org/10.1063/1.1534580

Crossref

154.

A. A.

Bakulin

,

O.

Selig

,

H. J.

Bakker

,

Y. L. A.

Rezus

,

C.

Müller

,

T.

Glaser

,

R.

Lovrincic

,

Z.

Sun

,

Z.

Chen

,

A.

Walsh

,

J. M.

Frost

, and

T. L. C.

Jansen

,

J. Phys. Chem. Lett.

6

,

3663

(

2015

).

https://doi.org/10.1021/acs.jpclett.5b01555

Crossref

PubMed

155.

M.

Ji

and

K. J.

Gaffney

,

J. Chem. Phys.

134

,

044516

(

2011

).

https://doi.org/10.1063/1.3530783

Crossref

PubMed

156.

Y. S.

Lin

,

P. A.

Pieniazek

,

M.

Yang

, and

J. L.

Skinner

,

J. Chem. Phys.

132

,

174505

(

2010

).

https://doi.org/10.1063/1.3409561

Crossref

PubMed

157.

T. l. C.

Jansen

and

J.

Knoester

,

Biophys. J.

94

,

1818

(

2008

).

https://doi.org/10.1529/biophysj.107.118851

Crossref

PubMed

158.

R. M.

Hochstrasser

,

Chem. Phys.

266

,

273

(

2001

).

https://doi.org/10.1016/s0301-0104(01)00232-4

Crossref

159.

A.

Tokmakoff

,

J. Chem. Phys.

105

,

1

(

1996

).

https://doi.org/10.1063/1.471856

Crossref

160.

O.

Golonzka

,

M.

Khalil

,

N.

Demirdöven

, and

A.

Tokmakoff

,

J. Chem. Phys.

115

,

10814

(

2001

).

https://doi.org/10.1063/1.1417504

Crossref

161.

M. T.

Zanni

,

N.-H.

Ge

,

Y. S.

Kim

, and

R. M.

Hochstrasser

,

Proc. Natl. Acad. Sci. U. S. A.

98

,

11265

(

2001

).

https://doi.org/10.1073/pnas.201412998

Crossref

PubMed

162.

K.-K.

Lee

,

K.-H.

Park

,

S.

Park

,

S.-J.

Jeon

, and

M.

Cho

,

J. Phys. Chem. B

115

,

5456

(

2011

).

https://doi.org/10.1021/jp1102274

Crossref

PubMed

163.

J. R.

Schmidt

,

S. A.

Corcelli

, and

J. L.

Skinner

,

J. Chem. Phys.

123

,

044513

(

2005

).

https://doi.org/10.1063/1.1961472

Crossref

PubMed

164.

H.

Torii

,

Chem. Phys. Lett.

414

,

417

(

2005

).

https://doi.org/10.1016/j.cplett.2005.07.114

Crossref

165.

M.

Tsuboi

and

G. J.

Thomas

,

Appl. Spectrosc. Rev.

32

,

263

(

1997

).

https://doi.org/10.1080/05704929708003316

Crossref

166.

B. M.

Auer

and

J. L.

Skinner

,

J. Phys. Chem. B

113

,

4125

(

2009

).

https://doi.org/10.1021/jp806644x

Crossref

PubMed

167.

S. J.

Roeters

,

C. N.

van Dijk

,

A.

Torres-Knoop

,

E. H. G.

Backus

,

R. K.

Campen

,

M.

Bonn

, and

S.

Woutersen

,

J. Phys. Chem. A

117

,

6311

(

2013

).

https://doi.org/10.1021/jp401159r

Crossref

PubMed

168.

C.

Liang

and

T. l. C.

Jansen

,

J. Phys. Chem. B

117

,

6937

(

2013

).

https://doi.org/10.1021/jp403111j

Crossref

PubMed

169.

A.

Morita

,

J. Phys. Chem. B

110

,

3158

(

2006

).

https://doi.org/10.1021/jp058155m

Crossref

PubMed

170.

P. C.

Singh

,

S.

Nihonyanagi

,

S.

Yamaguchi

, and

T.

Tahara

,

J. Chem. Phys.

137

,

094706

(

2012

).

https://doi.org/10.1063/1.4747828

Crossref

PubMed

171.

J. E.

Laaser

and

M. T.

Zanni

,

J. Phys. Chem. A

117

,

5875

(

2013

).

https://doi.org/10.1021/jp307721y

Crossref

PubMed

172.

Y.

Ni

,

S. M.

Gruenbaum

, and

J. L.

Skinner

,

Proc. Natl. Acad. Sci. U. S. A.

110

,

1992

(

2013

).

https://doi.org/10.1073/pnas.1222017110

Crossref

PubMed

173.

S.

Roy

,

S. M.

Gruenbaum

, and

J. L.

Skinner

,

J. Chem. Phys.

141

,

22D505

(

2014

).

https://doi.org/10.1063/1.4895968

Crossref

PubMed

174.

J. E.

Laaser

,

D. R.

Skoff

,

J.-J.

Ho

,

Y.

Joo

,

A. L.

Serrano

,

J. D.

Steinkruger

,

P.

Gopalan

,

S. H.

Gellman

, and

M. T.

Zanni

,

J. Am. Chem. Soc.

136

,

956

(

2014

).

https://doi.org/10.1021/ja408682s

Crossref

PubMed

175.

E. H. G.

Backus

,

N.

Garcia-Araez

,

M.

Bonn

, and

H. J.

Bakker

,

J. Phys. Chem. C

116

,

23351

(

2012

).

https://doi.org/10.1021/jp306273d

Crossref

176.

H.

Torii

,

Chem. Phys. Lett.

323

,

382

(

2000

).

https://doi.org/10.1016/s0009-2614(00)00536-4

Crossref

177.

M.

Musso

,

H.

Torii

,

P.

Ottaviani

,

A.

Asenbaum

, and

M. G.

Giorgini

,

J. Phys. Chem. A

106

,

10152

(

2002

).

https://doi.org/10.1021/jp021440a

Crossref

178.

S.

Roy

,

M. S.

Pshenichnikov

, and

T. L. C.

Jansen

,

J. Phys. Chem. B

115

,

5431

(

2011

).

https://doi.org/10.1021/jp109742p

Crossref

PubMed

179.

K.

Kwac

,

H.

Lee

, and

M.

Cho

,

J. Chem. Phys.

120

,

1477

(

2004

).

https://doi.org/10.1063/1.1633549

Crossref

PubMed

180.

M. H.

Farag

,

B. J.

Hoenders

,

J.

Knoester

, and

T. L. C.

Jansen

,

J. Chem. Phys.

146

,

234201

(

2017

).

https://doi.org/10.1063/1.4985665

Crossref

PubMed

181.

M.

Dinpajooh

and

D. V.

Matyushov

,

J. Phys. Chem. B

118

,

7925

(

2014

).

https://doi.org/10.1021/jp500733s

Crossref

PubMed

182.

A.

Anda

,

L.

De Vico

,

T.

Hansen

, and

D.

Abramavičius

,

J. Chem. Theory Comput.

12

,

5979

(

2016

).

https://doi.org/10.1021/acs.jctc.6b00997

Crossref

PubMed

183.

T. L. C.

Jansen

and

J.

Knoester

,

J. Chem. Phys.

127

,

234502

(

2007

).

https://doi.org/10.1063/1.2806179

Crossref

PubMed

184.

J. F.

Cahoon

,

K. R.

Sawyer

,

J. P.

Schlegel

, and

C. B.

Harris

,

Science

319

,

1820

(

2008

).

https://doi.org/10.1126/science.1154041

Crossref

PubMed

185.

F.

Šanda

and

S.

Mukamel

,

J. Chem. Phys.

125

,

014507

(

2006

).

https://doi.org/10.1063/1.2205367

Crossref

PubMed

186.

Y. S.

Kim

and

R. M.

Hochstrasser

,

J. Phys. Chem. B

110

,

8531

(

2006

).

https://doi.org/10.1021/jp060935n

Crossref

PubMed

187.

J.

Zheng

,

K.

Kwak

,

J. B.

Asbury

,

X.

Chen

,

I. R.

Piletic

, and

M. D.

Fayer

,

Science

309

,

1338

(

2005

).

https://doi.org/10.1126/science.1116213

Crossref

PubMed

188.

Y. S.

Kim

and

R. M.

Hochstrasser

,

Proc. Natl. Acad. Sci. U. S. A.

102

,

11185

(

2005

).

https://doi.org/10.1073/pnas.0504865102

Crossref

PubMed

189.

S.

Woutersen

,

Y.

Mu

,

G.

Stock

, and

P.

Hamm

,

Chem. Phys.

266

,

137

(

2001

).

https://doi.org/10.1016/s0301-0104(01)00224-5

Crossref

190.

C.

Scheurer

and

T.

Steinel

,

ChemPhysChem

8

,

503

(

2007

).

https://doi.org/10.1002/cphc.200600765

Crossref

PubMed

191.

K.

Kwak

,

J.

Zheng

,

H.

Cang

, and

M. D.

Fayer

,

J. Phys. Chem. B

110

,

19998

(

2006

).

https://doi.org/10.1021/jp0624808

Crossref

PubMed

192.

O.

Kühn

and

Y.

Tanimura

,

J. Chem. Phys.

119

,

2155

(

2003

).

https://doi.org/10.1063/1.1582841

Crossref

193.

R.

Kubo

,

J. Phys. Soc. Jpn.

17

,

1100

(

1962

).

https://doi.org/10.1143/jpsj.17.1100

Crossref

194.

T.

Steffen

and

Y.

Tanimura

,

J. Phys. Soc. Jpn.

69

,

3115

(

2000

).

https://doi.org/10.1143/jpsj.69.3115

Crossref

195.

Y.

Tanimura

and

T.

Steffen

,

J. Phys. Soc. Jpn.

69

,

4095

(

2000

).

https://doi.org/10.1143/jpsj.69.4095

Crossref

196.

Y.

Suzuki

and

Y.

Tanimura

,

J. Chem. Phys.

119

,

1650

(

2003

).

https://doi.org/10.1063/1.1578630

Crossref

197.

T.

Kato

and

Y.

Tanimura

,

J. Chem. Phys.

120

,

260

(

2003

).

https://doi.org/10.1063/1.1629272

Crossref

198.

D.

Cringus

,

T. l. C.

Jansen

,

M. S.

Pshenichnikov

, and

D. A.

Wiersma

,

J. Chem. Phys.

127

,

084507

(

2007

).

https://doi.org/10.1063/1.2771178

Crossref

PubMed

199.

C. P.

van der Vegte

,

A. G.

Dijkstra

,

J.

Knoester

, and

T. L. C.

Jansen

,

J. Phys. Chem. A

117

,

5970

(

2013

).

https://doi.org/10.1021/jp311668r

Crossref

PubMed

200.

P. L.

McRobbie

,

G.

Hanna

,

Q.

Shi

, and

E.

Geva

,

Acc. Chem. Res.

42

,

1299

(

2009

).

https://doi.org/10.1021/ar800280s

Crossref

PubMed

201.

T. L. C.

Jansen

,

J. Phys. Chem. A

122

,

172

(

2018

).

https://doi.org/10.1021/acs.jpca.7b10380

Crossref

PubMed

202.

F.

Ding

and

M. T.

Zanni

,

Chem. Phys.

341

,

95

(

2007

).

https://doi.org/10.1016/j.chemphys.2007.06.010

Crossref

203.

S.

Garrett-Roe

and

P.

Hamm

,

J. Chem. Phys.

130

,

164510

(

2009

).

https://doi.org/10.1063/1.3122982

Crossref

PubMed

204.

S.

Garrett-Roe

and

P.

Hamm

,

Acc. Chem. Res.

42

,

1412

(

2009

).

https://doi.org/10.1021/ar900028k

Crossref

PubMed

205.

L. J. G. W.

van Wilderen

,

A. T.

Messmer

, and

J.

Bredenbeck

,

Angew. Chem., Int. Ed.

53

,

2667

(

2014

).

https://doi.org/10.1002/anie.201305950

Crossref

206.

L. J. G. W.

van Wilderen

and

J.

Bredenbeck

,

Angew. Chem., Int. Ed.

54

,

11624

(

2015

).

https://doi.org/10.1002/anie.201503155

Crossref

207.

J. N.

Mastron

and

A.

Tokmakoff

,

J. Phys. Chem. A

120

,

9178

(

2016

).

https://doi.org/10.1021/acs.jpca.6b09158

Crossref

PubMed

208.

C. R.

Baiz

,

D.

Schach

, and

A.

Tokmakoff

,

Opt. Express

22

,

18724

(

2014

).

https://doi.org/10.1364/oe.22.018724

Crossref

PubMed

209.

T.

Hasegawa

and

Y.

Tanimura

,

J. Phys. Chem. B

115

,

5545

(

2011

).

https://doi.org/10.1021/jp111308f

Crossref

PubMed

210.

P. L.

Silvestrelli

,

M.

Bernasconi

, and

M.

Parrinello

,

Chem. Phys. Lett.

277

,

478

(

1997

).

https://doi.org/10.1016/s0009-2614(97)00930-5

Crossref

211.

M. K.

Ghosh

,

J.

Lee

,

C. H.

Choi

, and

M.

Cho

,

J. Phys. Chem. A

116

,

8965

(

2012

).

https://doi.org/10.1021/jp306807v

Crossref

PubMed

212.

M.

Sulpizi

,

M.

Salanne

,

M.

Sprik

, and

M.-P.

Gaigeot

,

J. Phys. Chem. Lett.

4

,

83

(

2013

).

https://doi.org/10.1021/jz301858g

Crossref

PubMed

213.

M.

Thomas

,

M.

Brehm

,

R.

Fligg

,

P.

Vöhringer

, and

B.

Kirchner

,

Phys. Chem. Chem. Phys.

15

,

6608

(

2013

).

https://doi.org/10.1039/c3cp44302g

Crossref

PubMed

214.

A.

Ishizaki

and

Y.

Tanimura

,

J. Chem. Phys.

125

,

084501

(

2006

).

https://doi.org/10.1063/1.2244558

Crossref

PubMed

215.

A.

Ishizaki

and

Y.

Tanimura

,

J. Phys. Chem. A

111

,

9269

(

2007

).

https://doi.org/10.1021/jp072880a

Crossref

PubMed

216.

A. O.

Caldeira

and

A. J.

Leggett

,

Physica A

121

,

587

(

1983

).

https://doi.org/10.1016/0378-4371(83)90013-4

Crossref

217.

J.-J.

Ding

,

J.

Xu

,

J.

Hu

,

R.-X.

Xu

, and

Y.

Yan

,

J. Chem. Phys.

135

,

164107

(

2011

).

https://doi.org/10.1063/1.3653479

Crossref

PubMed

2019

Author(s)

Theory of coherent two-dimensional vibrational spectroscopy

I. INTRODUCTION

II. FUNDAMENTALS

A. Third-order optical response function

B. Response function of a three-level system

III. CLASSICAL MECHANICAL APPROACHES TO NONLINEAR RESPONSE FUNCTIONS

A. Equilibrium molecular dynamics approach

B. Nonequilibrium molecular dynamics approach

1. Nonequilibrium finite-field method

2. Hybrid approach combining equilibrium and nonequilibrium finite-field methods

C. Limitations

IV. NUMERICAL INTEGRATION OF THE SCHRÖDINGER EQUATION

A. Hamiltonian parameterization, frequency maps

B. Mappings for interactions and the related techniques

C. Limitations—Bandwidth, high-frequency modes

V. PERSPECTIVES AND A FEW CONCLUDING REMARKS

ACKNOWLEDGMENTS

REFERENCES

Citing articles via

Resources

Explore

pubs.aip.org

Connect with AIP Publishing

Theory of coherent two-dimensional vibrational spectroscopy

I. INTRODUCTION

II. FUNDAMENTALS

A. Third-order optical response function

B. Response function of a three-level system

III. CLASSICAL MECHANICAL APPROACHES TO NONLINEAR RESPONSE FUNCTIONS

A. Equilibrium molecular dynamics approach

B. Nonequilibrium molecular dynamics approach

1. Nonequilibrium finite-field method

2. Hybrid approach combining equilibrium and nonequilibrium finite-field methods

C. Limitations

IV. NUMERICAL INTEGRATION OF THE SCHRÖDINGER EQUATION

A. Hamiltonian parameterization, frequency maps

B. Mappings for interactions and the related techniques

C. Limitations—Bandwidth, high-frequency modes

V. PERSPECTIVES AND A FEW CONCLUDING REMARKS

ACKNOWLEDGMENTS

REFERENCES

Citing articles via

Related Content

Resources

Explore

pubs.aip.org

Connect with AIP Publishing

This Feature Is Available To Subscribers Only