WorksPhysics and Astronomy
Detection Defines Dephasing in Two-Dimensional Electronic Spectroscopy of Materials: Coherent Field Emission versus Incoherent Population Observables
Abstract
The homogeneous spectral linewidth associated with light-matter interactions is a fundamental descriptor of the optical properties of materials, governed by the quantum dynamics of the condensed-matter system. We discuss here that the homogeneous linewidth measured by means of two-dimensional electronic spectroscopy depends not only on the intrinsic microscopic dynamics of the material, but also on the observable through which those dynamics are projected onto the measurement. In this Perspective, we develop a unified framework showing that identical microscopic dynamics can yield different experimentally inferred dephasing times because different detection operators project different sectors of the nonequilibrium dynamics. For coherent emitted-field measurements, the observed linewidth largely retains its conventional connection to the optical coherence time . By contrast, in population-detected modalities such as photoluminescence, photocurrent, and other action-detected two-dimensional spectroscopies, the apparent linewidth can additionally encode excited-state population redistribution dynamics, leading naturally to an effective coherence time . Using a coupled-mode model propagated under a common Liouvillian, we show that identical microscopic dynamics yield distinct apparent dephasing times when projected onto coherent-emission and population-derived observables. The detection observable is therefore not merely part of the experimental implementation, but determines what dynamical information remains experimentally observable and how homogeneous linewidths should be interpreted as materials descriptors.
Detection defines dephasing
The optical response of molecular and condensed-matter systems is governed by the timescale over which electronically excited states retain the phase memory imposed by the driving electromagnetic wavepacket. The decay of this phase coherence defines the homogeneous spectral linewidth—the intrinsic width of an optical transition, inversely proportional to the coherence time and revealed by ultrafast dephasing dynamics. Homogeneous lineshapes therefore encode the fundamental quantum dynamics underlying the optical transition and provide a direct connection between microscopic fluctuations, scattering processes, and macroscopic optical function. Dephasing is often treated as an intrinsic material property determined by the system Hamiltonian and its coupling to the surrounding environment. Here we argue that, while the underlying microscopic dynamics are indeed intrinsic to the material, the experimentally inferred linewidth reflects the dynamics of the projected observable rather than the microscopic dynamics alone. In other words, the experiment does not reveal the complete density-matrix dynamics; it probes its projection onto a detection observable.
Two-dimensional electronic spectroscopy (2DES) has transformed the study of condensed-phase molecular and materials systems by separating homogeneous and inhomogeneous broadening, resolving spectral correlations in energy and time, and providing direct access to ultrafast coherence and population dynamics.[1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11] In its conventional implementation (Figure 1a), the measured signal arises from coherent emission of the time-varying third-order nonlinear polarization, and the resulting linewidths are naturally interpreted in terms of the optical decoherence time , with distinct contributions from pure dephasing and population relaxation. Over the past two decades, 2DES has expanded to include implementations that detect not the emitted field itself, but rather an incoherent materials response generated by a four-pulse sequence (Figure 1b).[5, 6, 12] Photoluminescence (PL)-detected, photocurrent-detected,[12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27] [28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38] photoinduced-absorption-detected,38 and other action-based variants have greatly broadened the range of accessible systems, extending 2DES into increasingly complex condensed-phase architectures, including in-operando device environments.38–47 In these approaches, the measured signal is not the radiated nonlinear field itself, but rather a population-derived action observable.
Figure 1. Detection defines what linewidth means in multidimensional spectroscopy. (a) Schematic of a phase-matched coherent emitted-field 2DES experiment, in which three pulses generate a third-order nonlinear polarization that radiates a signal field detected by heterodyne interference with a local oscillator (LO). The signal is parameterized by three time intervals: the first coherence time , the population waiting time , and the detection-time coherence , corresponding to the evolution of a radiating optical coherence. The emitted signal field is detected by heterodyne interference with a local oscillator, yielding a time-domain signal proportional to the field amplitude, which is Fourier transformed (FT) with respect to to obtain the response along emission frequency . (b) Schematic of a phase-tagged, population-detected (action) 2DES experiment. Four pulses generate a nonlinear population, which is subsequently projected onto an action observable such as photoluminescence (PL) or photocurrent (PC) and, if applicable, integrated over a detection gate time . The signal is likewise parameterized by three intervals , , and , where and retain their usual meanings, while the final interval corresponds to the delay preceding population preparation by the final interaction, after which the signal is projected onto the action observable . Phase modulation of the pulse pairs allows isolation of rephasing and nonrephasing contributions by lock-in detection. (c) Conceptual hierarchy illustrating that identical photoexcitation dynamics, governed by a common Hamiltonian and dissipative environment , are projected onto different observables . Coherent emitted-field detection probes radiating optical coherences, while action detection probes population-derived observables. Consequently, any experimentally inferred linewidth reflects the evolution of the projected observable rather than a unique microscopic coherence timescale.
This distinction raises a central conceptual question: what does dephasing mean when the experiment does not directly observe radiating coherence? In conventional phase-matched coherent spectroscopy, assuming excited-state population redistribution is negligible on the timescale of the ultrafast experiment, the homogeneous neous spectral linewidth is directly linked to the optical phase memory time [10] [48, 49, 50, 51] through
where is the pure-dephasing contribution with time constant , and the contribution from population relaxation with lifetime . In population-detected 2DES, however, the same spectral feature may additionally reflect excited-state redistribution into the measured action channel, such as integrated PL intensity.
The central thesis of this Perspective is that the operational definition of dephasing depends on the detection operator , and therefore on the experimental detection modality . This distinction follows directly from the general structure of quantum measurement in nonlinear spectroscopy, where the experimentally recorded signal is obtained as the expectation value of a chosen detection operator acting on the optically prepared density matrix,
where denotes the number of light-matter interactions. Different detection modalities correspond to different choices of the operator , each defining a distinct probe of the same nonequilibrium density matrix . We define the first interaction to occur at , and to denote the interval between the final interaction and the measurement event. For the third-order response, denotes the first coherence interval and the waiting time in both experiments, while the meaning of the final interval depends on the detection scheme: for coherent emitted-field detection, , and corresponds to the evolution of a radiating optical coherence, whereas for action-detected measurements, , and denotes the final pulse-delay interval preceding projection onto the measured population observable (Figure 1).
The central question addressed here is therefore not whether different detection observables produce different nonlinear spectra—a point that is well established—but rather what dynamical information remains observable under a given measurement projection. We refer to this principle as observability in coherent spectroscopy. Because different detection operators interrogate different sectors of the nonequilibrium dynamics, they assign different operational meanings to linewidths, cross peaks, and even the dephasing time itself. A central practical question therefore arises: Do action-detected spectroscopies measure the same homogeneous linewidth as coherent emitted-field 2DES? The prevailing view in the ultrafast spectroscopy community has been that, provided nonlinear recombination dynamics do not substantially distort the 2D coherent lineshape, the answer is tentatively yes.[15] Subsequent work has shown that additional population dynamics can alter lineshapes, and therefore homogeneous linewidths inferred from action detection must be modeled explicitly to enable quantitative comparison with coherent-emission 2DES.[20, 27] [28, 31, 32] More fundamentally, once the measured signal depends on a population-derived observable rather than directly on the radiated field, the homogeneous linewidth no longer admits a unique observable-independent interpretation.
Here, we develop a unified framework for understanding how coherent emission and incoherent population observables encode dephasing in multidimensional spectroscopy of materials. We therefore define an observable-dependent effective coherence time
Equation (3) should therefore be viewed as an operational definition of the coherence time associated with a specified measurement protocol rather than as a universal material constant (Figure 1). This framework provides a route to disentangle genuine coherence loss from population redistribution and action-channel filtering in complex molecular and condensed-matter systems. The central objective is not to compare detection modalities per se, but rather to establish how the choice of observable defines the physical meaning of dephasing and homogeneous linewidth in 2DES.
Detection observables
The key distinction between conventional phase-matched 2DES that heterodyne-detects coherent field emission (Figure 1a) and phase-tagged population-based detection variants (Figure 1b) lies in the measurement operator acting on the nonequilibrium state generated by the pulse sequence.
In coherent detection, the measured third-order nonlinear signal following three phase-matched, time-ordered ultrafast light-matter interactions is proportional to the heterodyne interference between the emitted nonlinear field due to the time-varying third-order polarization and a local oscillator (LO), derived from an attenuated replica of the femtosecond pulse train,
where is the LO field that is arranged to copropagate with the phase-matched coherent emission, and
is the time-varying, third-order mesoscopic polarization, which is the source of coherent emission.52
In contrast, action-detected measurements project the nonequilibrium density matrix produced by a sequence of four phase-tagged femtosecond pulses, which produce a final population, onto a material action observable,
where is the detection operator and denotes the detection time gate.
The operator may correspond to a time-integrated action variable such as PL intensity, photocurrent, photoinduced absorption, or any other experimentally accessible population-space observable as outlined in Sec. I. Using phase-sensitive detection with a lock-in amplifier, we measure the action-detected signal integrated over the population lifetime, which is long compared to the evolution periods within the Liouville path:
Equation (7) emphasizes that action detection projects the optically prepared state through its subsequent population evolution before the signal is recorded. The integration over therefore contributes to the effective observable through which population redistribution is sampled;[20, 27] [28, 44, 53, 54] it is not introduced here as an independent additive dephasing channel.
More generally, the measured nonlinear signal may be written as Eq. (2), where specifies the observable. Because these operators interrogate different sectors of the nonequilibrium dynamics, changing changes the operational definition of dephasing. This rationale supports the view that dephasing is not only a dynamical property of the material Hamiltonian; it is also a property of how the system-bath quantum dynamics project onto the observable intrinsic to the experiment.
It is worth noting that the experimental scheme and strategy shown in Figure 1b can be implemented to realize the two measured observables discussed in this section. As depicted in this figure, it represents the projection of Eq. (6) if the four-pulse sequence generates the final population on which the action observation is performed, but the fourth pulse can also be used as a LO to extract the time-varying polarization to project as in Eq. (4), as exemplified in Ref. 48. The collinear geometry with a phase-tagging, phase-sensitive detection strategy can be implemented for both projection schemes discussed in this Perspective, offering a powerful option for microscopy and other applications in which the phase-matching beam geometry, as in Figure 1a, is inconvenient.
Observable-Dependent Dephasing Dynamics
The observable dependence of dephasing can be simply illustrated using a two-level system with Hamiltonian
The system undergoes identical microscopic dynamics but is subjected to different measurements, namely coherent field-emission and action-detected measurements. In the context of multidimensional spectroscopy, the expectation value of an observable operator to -th order at time can be expressed in the interaction representation as
The interaction representation is equivalent to the Schrödinger picture but is convenient for analyzing the final measurement . Because is time-independent in the absence of perturbations, we can focus on the dynamics of .
Taking the Lindblad master equation
we consider two Lindblad operators,
that characterize dissipation due to population relaxation and energy-gap fluctuations, respectively. The population-relaxation rate is , and is the pure-dephasing rate. The raising and lowering operators are , and are the Pauli matrices.
For coherent field-emission detection, the measurement operator is the dipole operator , which satisfies
The dephasing of the polarization is characterized by the second term on the right-hand side of Eq. (10), with decay rate
consistent with the linewidth specified in Eq. (1).
For action detection, the relevant operator is the excited-state projection operator
with evolution
The relaxation operator affects the excited-state population , whereas pure dephasing associated with energy-gap fluctuations makes no contribution. The population decay rate is therefore
Equations (11) and (13) demonstrate the central observable dependence in its minimal form. The dynamics contain the same population relaxation and the same pure dephasing, but the dipole operator and population projector sample different Liouville-space sectors, leading to different decay rates. The coherent polarization is sensitive to both processes, whereas the population observable is blind to pure dephasing. Thus, the operator-specific decay constants differ even though the Hamiltonian and dissipative dynamics are identical.
Furthermore, the minimal model provides a qualitative estimate of the homogeneous linewidths associated with the two spectroscopic approaches. When pure dephasing prevails, , the homogeneous linewidth of the coherent field-emission spectrum is broader than that of the action-detected spectrum. Conversely, when population relaxation dominates, the action-detected spectrum exhibits a broader homogeneous lineshape.
This example establishes the organizing principle used throughout the Perspective: experimentally inferred dynamical quantities are determined jointly by the underlying quantum dynamics and by the observable through which those dynamics are measured. In practice, multidimensional spectroscopy deals with systems containing more states that involve population redistribution, competing decay channels, and state-dependent action yields, as discussed in § 3. The analysis of the two-level model should therefore be viewed as a minimal operator-level illustration rather than as a complete description of dephasing in 2D spectroscopies.
Shared pathway selectivity across detection modalities
We return to the question: How are the two experimental modalities in Figure 1 fundamentally different? An imimportant point of commonality between coherent emitted-field and population-detected 2DES is that both detection schemes retain access to selected Liouville pathway classes.[10] In phase-matched three-pulse experiments, wavevector selection geometrically isolates rephasing and nonrephasing contributions through their distinct phase-matching conditions. In collinear phase-tagged population detection, the same pathway classes can be isolated through appropriate phase cycling or phase demodulation, enabling direct construction of rephasing-like and nonrephasing-like responses from the same underlying nonequilibrium quantum dynamics. For either modality, one may formally write
where denotes the selected rephasing or nonrephasing density-matrix component.
This shared pathway selectivity is conceptually important because, once a pathway class has been isolated, the detected signal still depends on the observable operator through which the trajectory is read out. Rephasing and nonrephasing spectra may therefore remain formally analogous across modalities while still encoding different operational meanings for linewidths, cross peaks, and effective coherence times.
Pathway-summed pump-probe implementations
A useful point of comparison is provided by pump-probe implementations of 2DES based on passively phase-locked pump pulse pairs generated, for example, with birefringent wedges and detected through white-light probe self-heterodyning.55 In this geometry, the broadband white-light probe pulsetrain serves as an intrinsic local oscillator for the third-order differential transmission signal, preserving high phase stability and experimental simplicity while directly encoding spectral correlations along the pump coherence and probe detection axes.
Unlike the explicitly pathway-resolved coherent phase-matched and phase-tagged population-detected modalities discussed above, however, this implementation typically reports a pathway-summed correlation function rather than separated Liouville components.55 Unless additional phase cycling, pulse-order discrimination, or interferometric separation is introduced,56–60 the measured spectrum is more naturally viewed as the total response,
This distinction is important for the present Perspective because the linewidth and lineshape nuances discussed here are most rigorously defined when specific Liouville pathway classes are explicitly resolved. The observable dependence emphasized throughout this article therefore applies to the total projected response in an averaged sense, rather than to individually resolved sectors.61 This does not diminish the utility of pump-probe 2D methods, but it does place the present discussion of effective observed linewidths in the narrower context of explicitly pathway-resolved 2DES measurements; it is our strong statement here that pathway-specific lineshape analysis is one of the most compelling advantages of 2DES,61 and pathway-summed 2DES is a useful way to enhance general pump-probe methods, but does not realize the full potential of 2DES. The hierarchy of pathway resolution provides the natural starting point for asking how linewidths and apparent dephasing times inherit distinct observable-dependent contributions.
Extension to Higher-Order Response: Accessing Many-Body Correlations
While the discussion above has focused on third-order response, the role of the detection operator in selecting experimentally accessible dynamics extends naturally to higher-order nonlinear processes. Throughout this Perspective we distinguish the quantum order of the detected coherence, one-quantum or two-quantum, from the perturbative order of the nonlinear optical response. Although the leading contribution to the two-quantum signal considered here is fifth order, higher-order nonlinear responses may also contribute, depending on excitation conditions and detection geometry. Two-quantum coherences may also be accessed at third order in appropriate phase-matching directions; consequently, the designation “two-quantum” specifies the coherence sector being interrogated rather than the perturbative order of the measured response.
At the fifth-order level, the measured signal remains of the general form
but now reflects the evolution of higher-rank multi-time correlations in the system. As in the third-order case, the key distinction lies in the detection operator .
Experimentally, the two-quantum signal dominated by its leading fifth-order response in action detection can be isolated via higher-harmonic demodulation, which imposes selection rules in Liouville space. In our work on Frenkel biexcitons,[25] demodulation at harmonics corresponding to two interactions with an initial pulse pair selects pathways that generate two-quantum coherences, which are subsequently projected onto populations and converted back to single-quantum coherences by later interactions. This sequence defines a specific population–coherence–population conversion pathway that is directly encoded in the measured observable.
At this level, the signal encodes not only the presence of multi-exciton states, but also the dynamics that interconvert coherences and populations within these manifolds. In analogy to the decomposition of at third order, one can interpret the fifth-order multi-quantum signal in terms of distinct physical processes, including correlated dephasing, population transfer within the two-exciton manifold, and relaxation to single-exciton states. These processes contribute with different weights depending on .
Complementary information is obtained in coherent field-emission 2DES, where phase-matched detection isolates the leading nonlinear polarization pathways contributing to the selected phase-matching direction. In the present implementation, the dominant contribution to the two-quantum coherent signal is fifth order, although higher-order nonlinear responses may also contribute under sufficiently strong excitation conditions. In semiconductor quantum wells, this approach has been used to directly resolve a two-quantum signal dominated by its leading fifth-order response (e.g., Turner and Nelson62), providing access to coherent multi-exciton correlations that are not projected onto populations.
Formally, the underlying response involves higher-order dipole correlation functions,
but the experimentally accessible information is determined by how these correlations are projected by the detection operator. As a result, coherent emission and action detection select different components of the same many-body dynamics.
Thus, extending to the dominant contribution preserves the central conclusion established at third order: the observed signal does not directly reflect the intrinsic dynamics alone, but rather the dynamics as filtered by the detection observable. Higher-order spectroscopy therefore provides a more intricate route to selectively accessing many-body correlations, with detection modality acting as a control parameter for which processes are emphasized or suppressed. This opens new opportunities to significantly advance our understanding of multiquasiparticle physics.63
Detection defines 2DES lineshapes
The remainder of this Perspective develops the basic idea in three steps. We first derive an observable-dependent decomposition of the measured linewidth, then illustrate the concept using a minimal Liouville-space model with fixed microscopic dynamics, and finally compare the predictions with representative experimental observations in conjugated polymers.
The observed homogeneous linewidth
Building on the observable-specific definition of introduced in Eq. (3), we decompose the measured linewidth into contributions arising from coherence loss,
population relaxation, and excited-state redistribution. The normalized time-domain signal envelope is written as
where describes pure phase randomization, population relaxation, and excited-state population redistribution. The factorization is not intended as an exact microscopic decomposition of the Liouvillian dynamics. Rather, it provides an effective phenomenological framework for organizing the principal physical processes that contribute to the experimentally observed linewidth.
If these processes are approximately separable and exponential over the experimental window,
This decomposition is introduced as the minimal phenomenological framework required to distinguish contributions arising from coherence decay, population relaxation, population redistribution, and the detection channel. It is not intended as a microscopic derivation applicable to arbitrary open quantum systems. The effective linewidth is then
Here again, the factorization is not intended as an exact microscopic decomposition of the Liouvillian. Rather, it provides a phenomenological framework for organizing the principal physical processes contributing to the experimentally observed linewidth. In general, non-Markovian fluctuations, correlated vibronic dynamics, and nonlinear population mixing may lead to nonexponential dynamics that do not admit a unique separation. These regimes motivate the discussion in § 5, where incoherent population mixing and pseudo-dephasing are examined as experimentally important examples of the limitations of this phenomenological picture.
A detection-dependent generalization is
where denotes the detection modality. The first two terms describe intrinsic optical decoherence. The latter term describe how the chosen observable projects the common nonequilibrium dynamics onto the experimentally recorded signal: accounts for observable-weighted population redistribution. The weighting coefficients encode how a given experimental observable samples the underlying nonequilibrium dynamics rather than representing intrinsic properties of the material. Coherent emission samples transfer through its effect on evolving optical coherences, whereas action detection samples populations after redistribution and observable filtering, leading to distinct weights .
The observable dependence enters through , which describes how population redistribution under the common Liouvillian is projected onto the experimentally recorded signal. For action detection, this weighting incorporates state-dependent action yields and any postexcitation population evolution sampled by the detection protocol. These effects therefore modify the observable projection of redistribution rather than constituting an independent additive linewidth channel.
The inferred dephasing time is therefore an operational property of the measurement protocol acting on common microscopic dynamics, rather than a property of the system Hamiltonian alone.
This detection dependence is summarized schematically in Figure 1c, where a common set of microscopic dynamics is projected onto distinct observables, yielding different operational definitions of the homogeneous linewidth.
Beyond linewidths, the measurement modifies other aspects of the 2DES lineshape. Cross-peaks in coherent emission reflect phase-coherent pathways, whereas in population detection they are additionally shaped by redistribution and observable weighting. As a result, amplitudes and temporal evolution can differ even when the underlying dynamics are identical. To make this decomposition operational, we now construct a model that represents a fixed physical system probed by both experimental protocols.
Model: Common dynamics, distinct observables
To isolate the role of the detection observable independently of material-specific disorder, excitonic microstructure, and other system-dependent complexities, we consider a minimal model in which the Hamiltonian and dissipative dynamics are held fixed while only the detection operator is changed, as illustrated schematically in Figure 1c. The purpose of the simulations is not to validate the phenomenological decomposition itself, but to demonstrate that identical microscopic dynamics produce distinct experimentally inferred linewidths when projected onto different observables. The objective is therefore not to reproduce the full complexity of any particular condensed-matter system, but to isolate the observable dependence of the inferred linewidth under otherwise identical microscopic dynamics. Material-specific effects such as static disorder, energetic heterogeneity, and correlated transport can subsequently be incorporated within the same observable-projection framework without altering its central conceptual premise.
The model is naturally formulated in Liouville space, where the density matrix evolves under a common Liouvillian and different detection operators project distinct signals from the same underlying trajectory. We consider two weakly anharmonic coupled modes,
where and create and annihilate excitations in mode ; are the mode frequencies, are weak anharmonicities, and is the intermode coupling. We consider a weakly anharmonic excitonic manifold that ensures a nontrivial nonlinear response while preserving a transparent normal-mode structure, for which the excited-state absorption transition is only slightly displaced from the corresponding stimulated-emission pathway. This regime minimizes artificial linewidth differences arising solely from pathway separation and therefore provides a more stringent test of whether observable-dependent detection alone modifies the experimentally inferred homogeneous linewidth.
The light-matter interaction is treated perturbatively in the Coulomb gauge,
with dipole operator
where , with being the detection vector that encodes the measurement operator defining the observable signal.
The reduced density matrix evolves according to
where the dissipators describe pure dephasing, population relaxation, population redistribution, and radiative decay, respectively. The simulations therefore establish the existence of observable-dependent linewidths under identical microscopic dynamics, independently of the quantitative details of any particular material system.
Numerical simulations are performed using our quantum dynamics package QuDPy, modeling the system described by Eq. (20), truncated to three levels per mode. This nine-dimensional Hilbert space provides a minimal description of the ground , one-quantum , and two-quantum manifolds required for both coherent response and population redistribution dynamics. Both the phase-matched emitted-field response and the phase-tagged PL-detected signal are propagated under the same common Liouvillian as in Figure 1c, and the corresponding signals are determined according to Eqs. (4) and (6).
Pure dephasing is modeled as
where projects onto excited state , and is the corresponding dephasing rate. Population relaxation is described as a cascade from the two-quantum manifold into the one-quantum manifold , followed by decay to the ground state ,
with and . Although these jump operators are not Hermitian, the Lindblad form guarantees preservation of both Hermiticity and trace of .
Population redistribution within the one-quantum manifold is described by
with and . This term captures processes such as energy transfer or exciton migration between excited states. For thermally activated reversible exchange, the rates satisfy detailed balance,
so that downhill transfer is favored according to the Boltzmann population ratio.
Radiative decay from the one-quantum manifold is given by
, where is the spontaneous emission rate of state . These radiative channels define the intrinsic emission processes, while the detection operator determines how strongly each channel contributes to the measured signal.
Within this common dynamical framework, we compare two observables. In coherent emitted-field detection, the signal is obtained from the third-order polarization , Eq. (5), which directly probes the evolution of optical coherences. In contrast, in PL-detected 2DES the signal is obtained from a population-weighted action operator acting on the emissive manifold,
From the standpoint of quantum measurement theory, changing from coherent detection to action detection corresponds only to replacing the final observable while leaving the Liouvillian propagation unchanged. For the present two-mode model, Eq. (29) reduces to
The coefficients encode the emissive yield and detection weight of each state, so that the observable reflects both intrinsic radiative decay and any population redistribution preceding emission.
The key point is that the Hamiltonian and all dissipative processes are identical in the two cases; only the observable differs. This distinction is reflected experimentally in 2DES of silicon-vacancy centers in diamond, where PL- and heterodyne-detected measurements yield markedly different spectra.65 In that system, PL detection effectively post-selects a subset of radiatively efficient emitters, whereas heterodyne detection probes a broader ensemble, leading to different linewidths and coherence times. Within the present framework, this corresponds to observable-dependent projection of a common dynamical ensemble. The work of Smallwood et al.65 therefore illustrates an important general principle: changing the observable changes the experimentally accessible projection of the underlying dynamics without changing the underlying Hamiltonian itself, as expressed in Figure 1. Here we extend this principle beyond ensemble selection, showing that even for a fixed set of microscopic dynamics, different observables can weigh coherence, population redistribution, and relaxation processes differently. As a result, the measured homogeneous linewidth becomes an observable-dependent quantity, rather than a unique property of the underlying system.
We show analytically in the Supplemental Material that a two-level realization of the model of Eq. (20), propagated according to Eq. (23), results in a deviation of the observed homogeneous linewidth for resonance upon action detection, in the low-temperature limit ,
We assume , such that in the low-temperature limit population transfer is unidirectional and downhill, . In the high-temperature limit , detailed balance gives , so that the initial-slope result becomes
with . Population transfer contributes to the initial observed linewidth through the downhill component of the thermally parametrized exchange rate, weighted by the detection ratio , and vanishes identically for . At finite evolution times, the full bidirectional exchange dynamics are retained by the Liouvillian propagation.
We now model the 2DES lineshapes numerically64 to explore this parameter landscape. The numerical simulations presented below verify that this transport-induced observable divergence persists under experimentally realistic pure-dephasing conditions and increases systematically with population transport, demonstrating that it is not an artifact of lifetime-limited dynamics.
2DES simulated lineshapes
Figure 2 compares simulated rephasing 2DES amplitude spectra generated from the same underlying quantum dynamics of Eq. (23), but projected onto two different detection observables. The top row shows coherent emitted-field detection, while the bottom row shows action detection. In the reference case Figure 2(a), both detection schemes recover the same basic spectral structure, with diagonal features associated with the two excitonic resonances and cross peaks reflecting correlations between them. Despite the common Hamiltonian and Liouvillian dynamics, however, the two observables produce markedly different lineshapes. The action-detected resonances are systematically broader than their coherent-emission counterparts, demonstrating that the experimentally inferred homogeneous linewidth depends on the observable through which the common nonequilibrium dynamics are projected. The relative cross-peak amplitudes also differ substantially between the two detection modalities, with the action-detected spectrum exhibiting considerably stronger cross-peak intensity relative to the diagonal features. Thus, the observable dependence extends beyond linewidth broadening to the experimentally accessible manifestation of spectral correlations.
Figure 2. Simulated 2DES amplitude rephasing spectra in a rephasing Liouville pathway, illustrating the effect of detection modality on the observed nonlinear lineshape in a low-temperature regime of Eq. (26) (no uphill excited-state population transfer), and with , , and . Transition dipole moments were normalized such that . The top row shows field-emission-detected spectra (), while the bottom row shows action-detected spectra () calculated from the same underlying excitonic Hamiltonian and dissipative dynamics, Eq. (23). The columns correspond to three parameter regimes: (a) the leftmost column is the baseline reference case with microscopic parameters for both modes, and for both modes, , for both modes, and in the action operator; (b) the middle column represents a case with modified population-transfer rate , with all other parameters as in the baseline simulation; (c) the rightmost column uses the same parameters as the baseline simulation but with unbalanced action-detection weights . These parameter values were chosen to avoid the strong-coupling regime; the rates and coupling strength are varied as described in Table I.
Increasing the downhill population-transfer rate Figure 2(b) redistributes spectral weight within both detection schemes, modifying the relative intensities of the diagonal and cross-peak features. In particular, the higher-energy resonance is suppressed relative to the lower-energy resonance, consistent with downhill population redistribution. Importantly, the two detection observables respond differently to this common change in the underlying dynamics: the action-detected features remain substantially broader and display a different redistribution of cross-peak intensity than the corresponding coherent-emission features. This provides a direct spectral manifestation of the observable-dependent projection of population transport.
Further observable-dependent effects emerge when the action-detection weights are made unequal Figure 2(c). As expected, changing these weights has little effect on the coherent-emission spectrum because its detection operator is unchanged. By contrast, the action-detected spectrum exhibits pronounced changes in both linewidth and relative spectral weight, including enhanced broadening of the population-derived features. The comparison therefore isolates the role of the final measurement projection: identical microscopic dynamics can produce different experimentally inferred lineshapes when the evolving state is projected onto observables with different state-dependent action weights.
Throughout this Perspective we assume the impulsive excitation limit, in which the antidiagonal width of the rephasing spectrum provides the standard measure of the homogeneous linewidth. Within this limit, Figure 3 quantifies the observable dependence of the apparent homogeneous linewidth through normalized antidiagonal cuts of the simulated spectra. The observed linewidth is defined operationally by Eq. (19). In the reference case Figure 3(a), both excitonic resonances are systematically broader under action detection than under coherent emitted-field detection, despite the identical Hamiltonian and Liouvillian dynamics used to generate the two signals. The effect is particularly clear because the revised simulations are performed in the pure-dephasing-limited regime, where the direct contribution of excited-state population relaxation to the homogeneous linewidth is negligible. The difference therefore cannot be attributed to the artificially short excited-state lifetime employed in the original illustrative parameter set, but instead reflects the distinct projection of the common nonequilibrium dynamics onto the two detection observables.

Figure 3. Antidiagonal cuts through the simulated 2DES quantify the detection dependence of the apparent homogeneous linewidth. Shown are cuts for (a) the reference simulations Figure 2(a), (b) the enhanced excited-state transfer simulations Figure 2(b), and (c) the unbalanced detection weights Figure 2(c).
Increasing the downhill population-transfer rate Figure 3(b) modifies the antidiagonal lineshapes while preserving the systematic distinction between the two detection modalities. Population redistribution contributes to the measured linewidth in both experiments, as represented phenomenologically by Eq. (19), but its spectroscopic manifestation depends on how the resulting nonequilibrium dynamics are projected onto the measured observable. Consequently, the action-detected and coherent-emission linewidths respond differently to the same underlying population-transfer dynamics. The systematic dependence of this observable-specific broadening on the transfer rate is quantified below in Figure 4.
Figure 4. Observable-dependent homogeneous linewidth as a function of population transfer. Extracted homogeneous linewidth as a function of the population-transfer contribution for coherent emitted-field (open squares) and action-detected (filled circles) spectra at the (top) and (bottom) resonances. All microscopic parameters are held fixed as described in Figure 2(a) while is varied, so that the two detection modalities probe the same underlying Hamiltonian and Liouvillian dynamics.
A complementary observable-selective effect appears when the action-detection weights are made unequal Figure 3(c). Because the coherent-emission detection operator is unchanged, its antidiagonal linewidths remain nearly unchanged relative to the reference case. By contrast, the action-detected lineshapes respond directly to the modified state-dependent detection weights and remain substantially broader than their coherent-emission counterparts. Taken together, the three parameter regimes demonstrate that identical microscopic dynamics need not yield identical experimentally inferred homogeneous linewidths: the measured linewidth reflects both the underlying dynamics and the observable through which those dynamics are projected.
We report the values of for both experimental modalities in Table I. Taken together, these values, with Figure 2 and 3, establish that the experimentally extracted homogeneous linewidth is not an intrinsic property of the excitonic system alone, but depends explicitly on the detection operator through which the nonlinear response is projected. The simulations therefore support the central thesis of this Perspective: the experimentally observed dephasing time is an observable-dependent quantity that reflects both the intrinsic quantum dynamics and the measurement channel used to interrogate them.
| Simulation label (Fig. 2) | Mode frequency | (meV) | (meV) |
| (a) Reference | 5.16 | 8.63 | |
| 7.10 | 10.99 | ||
| (b) Enhanced transfer rate | 5.18 | 8.60 | |
| 8.91 | 12.68 | ||
| (c) Unbalanced action detection | 5.16 | 8.44 | |
| 7.10 | 10.75 |
Table 1. Observed homogeneous linewidths extracted from the antidiagonal cuts of Figure 3.
We quantify the dependence of on the downhill population-transfer rate in Figure 4. At [Figure 4(a)], the linewidths remain nearly independent of population transfer but exhibit a pronounced observable-dependent
offset, with action detection yielding the broader feature. At [Figure 4(b)], both linewidths increase with , while their separation grows systematically with increasing population transfer. Thus, the observable dependence itself is resonance-specific: at it appears primarily as a nearly transport-independent linewidth offset, whereas at population redistribution progressively enhances the divergence between the two detection modalities. This behavior demonstrates that population redistribution is projected differently by the two detection observables and provides direct numerical evidence that the homogeneous linewidth inferred from a 2DES spectrum depends jointly on the underlying microscopic dynamics and on the observable through which those dynamics are measured.
The simulations presented here demonstrate that observable-dependent linewidth broadening survives under parameter regimes representative of molecular and polymeric systems. The polymer examples discussed below therefore serve not as validation of the model itself, but as experimental contexts in which these operational distinctions are expected to become relevant.
HJ aggregates in conjugated polymers
This Perspective is motivated by measurements of homogeneous linewidths in conjugated polymers, carried out in the context of understanding photophysical aggregates in this class of materials, for which we have implemented both experimental schemes discussed here — see Refs. 14,25,38,41,66,67. Before discussing the experimental observations, it is useful to distinguish two conceptually separate aspects of the problem. The first concerns the intrinsic microscopic dynamics governing optical coherence, which are determined by the material Hamiltonian, system–bath interactions, and excitonic structure. In the polymer systems discussed below, these intrinsic dynamics are responsible for the remarkably weak temperature dependence of the homogeneous linewidth, as previously attributed to HJ-aggregate excitonic structure. The second concerns how those same nonequilibrium dynamics are projected into an experimentally recorded signal through a specified detection observable. The present Perspective addresses this latter question. Accordingly, the observable-dependent framework proposed here should be viewed as complementary to, rather than replacing, the microscopic mechanisms responsible for the underlying coherence dynamics.
The examples discussed below are intended to illustrate how the observable-dependent framework provides a unified interpretation of representative experimental observations reported across a variety of material systems. Because these studies were not originally designed as controlled comparisons between coherent-emission and action-detected measurements performed on identical samples under identical conditions, they should not be interpreted as definitive experimental validation of the framework. Rather, they illustrate that the proposed interpretation is consistent with a broad range of reported observations. Across these studies, we observe an apparently anomalous weak temperature dependence of over 4–300 K68,69 in a progression of solid-state microstructures with increasing chain flexibility, structural order, donor–acceptor character, and aggregate coherence. A systematic comparison of coherent-emission and action-detected measurements on the same family of conjugated polymer systems is presented in Ref. 70, providing a direct experimental realization of the observable-dependent framework proposed here. The observed homogeneous linewidth varies from to 90 meV across the series, yet remains nearly temperature independent, within , for each material. This behavior is striking because strongly vibronically active systems are generally expected to exhibit pronounced thermal pure dephasing. Representative values are summarized in Table II, highlighting the systematic difference between coherent emitted-field and action-detected measurements.
We rationalized this behavior within an exciton–vibrational framework as a consequence of the hybrid HJ aggregate electronic structure characteristic of these polymers.71 Here, “HJ” denotes the coexistence of J-like intrachain excitonic delocalization and H-like interchain Coulomb coupling, producing an excitonic manifold that supports both bright band-edge states and nearby dark relaxation channels. The key point is that exciton–vibrational coupling alone does not determine the quantum dynamics; rather, the aggregate band structure determines which sectors of the environmental fluctuation spectrum are sampled by the optically prepared excitation. To test this interpretation, we implemented a triangulation strategy mirroring the central thesis of this Perspective: selected materials were interrogated using both detection modalities together with resonance Raman intensity analysis as an independent probe of exciton–vibrational coupling.68,69 Systematically, we find that , supporting the conclusion that the inferred linewidth depends on the observable through which the nonequilibrium dynamics are projected. The nearly temperature-independent intrinsic coherence dynamics have previously been attributed to the HJ-aggregate excitonic structure. Within the present framework, these intrinsic dynamics constitute the common underlying evolution that is subsequently projected onto different detection observables, leading to observable-dependent experimentally inferred linewidths.
The semiconductor polymer PBTTT provides an especially clear example. This polymer forms liquid-crystalline-like mesophases associated with its ribbon-like backbone, promoted by the rigid thienothiophene unit and side-chain interdigitation. We measure and (Table II), with both values remaining nearly unchanged over the full 4–300 K range.68,69 The films used in both measurements were prepared from the same polymer batch under closely matched processing conditions, supporting that the observed difference is intrinsic to the detection modality.
A similarly instructive contrast is provided by P3HT (Table II), the canonical semicrystalline conjugated polymer, whose high-molecular-weight films () adopt a two-phase morphology consisting of crystalline domains embedded in amorphous chain-entangled regions.72,73 Contrary to a common simplification in the literature, the high crystallinity of P3HT does not imply a narrow electronic density of states. In PL-detected multidimensional measurements, we reported at 8 K,[14] whereas coherent emitted-field measurements yield over the full 4–300 K range.69 The persistence of this factor-of-two discrepancy across two chemically and morphologically distinct polymer systems strongly supports the detection-dependent linewidth inequality.
These two case studies provide direct experimental support for the central claim of this Perspective. The persistence of the linewidth inequality across heterogeneous HJ-aggregate microstructures identifies a regime in which pulse ordering and delayed population conversion can mimic additional dephasing and obscure the interpretation of , motivating the discussion of pseudo-dephasing and its experimental diagnostics in the next section. The dedicated same-material comparison reported in Ref. 70 provides an experimental bridge between the conceptual framework developed in this Perspective and quantitative observable-dependent spectroscopy of complex molecular systems. Future experiments combining coherent-emission and action-detected multidimensional spectroscopy on the same sample under identical experimental conditions will provide the most direct quantitative tests of the observable-dependent framework developed here.
Incoherent population mixing and pseudo-dephasing
The polymer case studies above point to a broader interpretive challenge: in population-detected multidimensional spectroscopy, linewidth broadening does not necessarily reflect loss of optical phase coherence. Delayed population dynamics can instead generate spectral structure that mimics dephasing.
The clearest manifestation of this effect arises from incoherent population mixing. In this regime, populations generated by different pulse interactions evolve independently and subsequently interact through nonlinear kinetics during or after the pulse sequence. The measured signal therefore contains contributions that do not originate from a single coherent Liouville pathway, but from population-level mixing. Consequently, cross peaks, linewidths, and waiting-time dynamics can acquire contributions that are not directly related to optical coherence decay.
| Material | Detection modality | (meV) | Temperature range | Ref. |
| PBTTT | Coherent emission | 4–300 K | 68–70 | |
| PBTTT | Action (PL-detected) | 4–300 K | 25,68–70 | |
| P3HT | Coherent emission | 4–300 K | 69 | |
| P3HT | Action (PL-detected) | 8 K | 14,70 |
Table II. Representative homogeneous linewidths extracted from coherent emitted-field and action-detected multidimensional spectroscopy in conjugated polymer systems. Values are reported for comparable temperature ranges; see cited references for full experimental details.
This behavior has been documented explicitly in our earlier work on phase-modulated 2DES, where incoherent population mixing was shown to generate cross peaks and apparent spectral broadening without corresponding additional coherent pathway content.[15] [46] In semiconducting polymer films, processes such as exciton migration, exciton–exciton annihilation, and downhill relaxation into lower-energy emissive states can continue on timescales comparable to, or longer than, the interpulse delays.[15] The detected population can therefore continue to evolve after the coherent light–matter interactions have occurred, so that the time-integrated action signal reflects both the coherent response encoded by the pulse sequence and its subsequent population processing.
When this additional spectral structure can be represented phenomenologically as linewidth broadening, one may write
where denotes the observable-dependent optical de-coherence described by equation (18), while represents an effective broadening associated with nonlinear population kinetics entering the time-integrated observable of equation (7). This decomposition is phenomenological: incoherent population mixing can also generate new spectral features rather than simple broadening. In either case, should not be interpreted as an additional intrinsic contribution to the material coherence decay rate.
A key practical question is therefore how to determine when such incoherent contributions are present. A particularly direct diagnostic arises from cross-talk between modulation channels in the four-pulse sequence, as demonstrated in ref. 46. In photocurrent-detected two-dimensional spectroscopy, incoherent population mixing produces measurable correlations between signals associated with nominally independent coherence-delay variables, providing a fingerprint of nonlinear population mixing contributions to the measured lineshape.
Such effects motivate the useful distinction between observable-dependent dephasing and pseudo-dephasing. The former describes genuine loss of optical coherence as projected through a specified detection observable; the latter denotes apparent spectral broadening or related lineshape changes generated by population dynamics after the relevant optical coherence has been created. Recognizing[25] [46] and mitigating34 this regime is therefore essential for interpreting action-detected 2DES of heterogeneous materials, where population redistribution can occur throughout the detection window represented by equation (7) and is intrinsic to the photoexcited-state landscape.
Perspective: designing observables to probe materials function
The central message of this Perspective is that the detection observable must be treated as a fundamental part of the 2DES experiment. The apparent homogeneous linewidth and inferred dephasing time do not arise solely from the material Hamiltonian and its bath-induced fluctuations, but from how the nonequilibrium density matrix—the photoexcited state—is projected onto an experimentally accessible observable defined by the detection scheme. The choice of detection modality therefore directly shapes which aspects of the underlying dynamics and materials function are experimentally accessible.
This recognition suggests a broader conceptual shift for multidimensional spectroscopy. Historically, the field has focused primarily on pulse-sequence design, phase matching, and pathway selection as the principal means of isolating coherent dynamics. The framework developed here shows that the detection operator is equally fundamental. The opportunity is therefore not only to engineer excitation pathways, but also to engineer observables that are maximally sensitive to the functional degrees of freedom most relevant to the material under study.
In semiconducting polymers, such observable engineering may emphasize exciton migration, multiexciton interactions, or relaxation into dark-state manifolds associated with hybrid HJ aggregate structure;68,69 in polymer-based solar cells, it may distinguish carrier-generation and charge-recombination pathways;41 and in organic and hybrid organic–inorganic semiconductors, it may enhance sensitivity to biexciton correlations and binding.[25] [63, 66, 74, 75] In this broader view, observable engineering transforms 2DES from a general probe of coherence into a targeted probe of structure–dynamics–function relationships.
A particularly promising extension is toward the identification and quantification of many-body correlations in quantum materials. In strongly correlated systems, including candidate quantum spin liquids, excitonic insulators, and low-dimensional magnets, the relevant degrees of freedom cannot generally be reduced to independent single-particle excitations. Multidimensional spectroscopies can encode higher-order correlations in cross peaks, spectral correlations, and nonlinear response pathways, but the extent to which these features reveal specific many-body dynamics depends critically on the observable through which the response is detected. By comparing or designing coherent-emission, action-detected, and hybrid measurement schemes, it may become possible to enhance sensitivity to selected correlated sectors of the dynamics while suppressing competing incoherent contributions. Developing operational mappings between multidimensional spectra and quantitative measures of many-body correlations—and, ultimately, experimentally accessible witnesses of quantum correlations—represents an important direction for connecting coherent spectroscopy with quantum-materials discovery.
More generally, the concept of observability developed here is not restricted to coherent spectroscopy. It raises the broader question of which dynamical properties of an open quantum system remain experimentally accessible under a specified measurement protocol. A nonlinear spectroscopic experiment does not measure microscopic dynamics directly; it measures a projection of those dynamics onto a chosen observable. Consequently, experimentally inferred quantities such as dephasing rates, population-transfer contributions, and effective linewidths should be understood operationally rather than assumed to be intrinsic material constants independent of measurement.
The analytical framework and numerical simulations considered here show that the divergence between detection observables persists across experimentally relevant parameter regimes. Observable-dependent linewidths are therefore not a peculiarity of a specific model, but a general consequence of projecting nonequilibrium dynamics onto different measured quantities. This distinction extends beyond the specific comparison between coherent-field-emission and action-detected multidimensional spectroscopies considered here.
We therefore propose observable engineering as a general principle for multidimensional spectroscopy: experimentally accessible information is determined jointly by the system dynamics and by the observable through which those dynamics are projected. The detection observable is not simply the final stage of the experiment; it helps define the physical meaning of the measured spectrum and determines which aspects of the underlying dynamics become experimentally accessible. Treating it as a first-class design variable provides a framework for interpreting existing measurement schemes, designing new ones, and establishing more rigorous structure–dynamics–function relationships in complex molecular and condensed-matter systems.
This perspective also leads naturally from observability to identifiability. Once the measurement projection has been specified, the next question is not merely whether a dynamical process contributes to the signal, but whether its underlying physical parameters can be uniquely inferred from the available observables. Determining which parameters remain identifiable under different measurement projections provides a natural next step toward a more general theory of information extraction from nonlinear spectroscopy.
Supplementary Material
The Supplementary Material file contains section SI: ”OBSERVABLE-WEIGHTED POPULATION EXCHANGE AND FINITE-TEMPERATURE LIMITS”, containing derivation of equations (31) and (32), and section SII: ”2DES LINESHAPE AS A FUNCTION OF EXCITED-STATE TRANSPORT”, preprinted spectra used to construct Figure 4.
Acknowledgments
CSA and ERB dedicate this Perspective to the memory of Mark Ratner, a role-model, a mentor, and a friend.
HL and CSA acknowledge funding from the Government of Canada (Canada Excellence Research Chair CERC-2022-0055). CSA acknowledges support from the Institut Courtois, Faculté des arts et des sciences, Université de Montréal (Chaire de recherche de direction de l’Institut Courtois) and from the Natural Science and Engineering Research Council of Canada (NSERC Discovery Grant RGPIN-2024-05893). CSA and SPO gratefully acknowledge support from the training program CREATE for Accelerated Discovery (AccelID) hosted by the Acceleration Consortium and co-led by the Institut Courtois (NSERC Grant #596133-2025). ERB acknowledges funding from the National Science Foundation (CHE-2404788), Robert A. Welch Foundation (E-1337), the Department of Energy supported this research through Award No. 11937-PO147716. ERB gratefully acknowledges funding from IVADO for a Visiting Professorship at the Institut Courtois, Université de Montréal.
Author declarations
Conflict of Interest
The authors have no conflicts to disclose.
Author Contributions
SPO: Investigation (numerical simulations), Software; HL: Supervision, Conceptualization; ERB: Conceptualization; CSA: Conceptualization (lead), Supervision, Project administration, Writing—original draft; All authors: Writing—review & editing.
Use of Generative Artificial Intelligence
In compliance with institutional guidelines of the Université de Montréal, generative artificial intelligence tools were used to assist with the editing of language and stylistic refinement of parts of the manuscript and to assist in the synthesis of the literature. These tools were not used to generate scientific content, perform analysis, or influence the interpretation of results. All content has been reviewed and validated by the authors, who assume full responsibility for the manuscript.
Data availability
In accordance to the Data Management Plan of the Canada Excellence Research Chair in Light-Matter interactions (http://hdl.handle.net/1866/33427), the numerical data and code that support the findings of this study are openly available in the Borealis Dataverse Repository at https://doi.org/10.5683/SP4/UKSLI2.
Observable-weighted population exchange and finite-temperature limits
In the main text, the experimentally inferred linewidth is interpreted as an observable-dependent quantity obtained by projecting a common nonequilibrium dynamics onto a specified detection operator. Here we derive this result for population exchange between two excited states. The derivation is formulated first for an arbitrary initial population distribution and then specialized to the optically prepared pure-state initial condition used in the main text.
Throughout this section, the rate constants have units of inverse time. The corresponding contribution to an energy linewidth is obtained by multiplication by . Equivalently, may be absorbed into the definition of the rates when they are reported in energy units.
Population dynamics and detection observable
Consider two excited states and , with populations collected into the vector
The populations evolve according to the kinetic equation
with rate matrix
Here, and denote irreversible population-loss rates from states and , respectively, while and describe bidirectional population exchange within the excited-state manifold.
For an action-detected measurement, the signal is a weighted projection of the population vector,
where
contains the state-dependent action yields and detection efficiencies. The formal solution of Eq. (35) is
so that
General observable-dependent logarithmic decay rate
We define the instantaneous effective decay rate of the measured action signal by
Using Eqs. (35) and (37), this may be written in the general form
At the initial time,
Equation (S10) is the general observable-projection result for the initial logarithmic slope. It shows explicitly that the measured decay rate depends jointly on the population dynamics, the initial state, and the detection observable. For an arbitrary initial condition
Eq. (S10) becomes
The first term is the detection-weighted contribution from irreversible population loss. The second is the contribution from population exchange. It depends on both the net population flux and the difference between the detection weights.
Several consequences follow directly from Eq. (S12). First, exchange does not contribute when
because a balanced observable is insensitive to redistribution between the two states. Second, exchange does not contribute to the initial slope when the initial populations satisfy
because the exchange dynamics initially carry no net population flux.
Optical preparation of state
The simulations and main-text analysis consider selective optical preparation of the higher-energy state ,
For this initial condition,
The uphill rate does not enter the initial derivative because . The initial logarithmic decay rate is therefore
The observable-dependent exchange contribution is
or, expressed as an energy linewidth,
This result is the microscopic realization of the observable-weighted population-redistribution contribution introduced phenomenologically in the main text.
If optical dephasing and population relaxation are included, the total observed linewidth associated with resonance may be written as
within the approximately exponential and separable regime discussed in the main text.
The final term in Eq. (53) is the microscopic realization of the observable-weighted population-redistribution contribution introduced in the main text. The state-dependent action yields and subsequent population evolution enter through the measurement projection and therefore do not define an additional independent linewidth contribution.
Finite-temperature parametrization
At finite temperature, both downhill and uphill exchange rates may be nonzero. Detailed balance gives
We define the total exchange rate
and the thermodynamic asymmetry
Using Eq. (54),
The individual rates are therefore
For the pure- initial condition of Eq. (48), only the downhill rate enters the initial logarithmic slope. Substitution of Eq. (58) into Eq. (52) gives
Equation (S27) is a finite-temperature parametrization of the initial downhill-transfer contribution. It should not be interpreted as an independent sum of uphill and downhill contributions to the initial slope. The uphill process enters the subsequent finite-time dynamics only after population has accumulated in state .
Limiting cases
Equation (S27) satisfies the relevant limiting cases.
a. Low-temperature limit. When
one has
and therefore
b. High-temperature limit. When
one has
so that
Although the uphill and downhill rates are equal in this limit, the initial slope following pure- preparation contains only the instantaneous downhill flux out of state .
c. Balanced detection. For
the exchange contribution vanishes at every temperature,
Population redistribution remains present dynamically, but it is invisible to an observable that assigns equal weight to the two participating states.
Thermal and mixed initial populations
A genuinely bidirectional contribution to the initial logarithmic slope requires an initial population distribution with both and . For such a mixed nonequilibrium initial state, Eq. (45) gives
Both uphill and downhill rates now enter explicitly.
For a thermal distribution within the two-state excited manifold,
detailed balance implies
Consequently,
At thermal equilibrium there is no net exchange flux, so population exchange does not contribute to the initial logarithmic decay even when the detection weights are unequal. Observable-dependent exchange broadening therefore requires a nonequilibrium population distribution, as is naturally produced by optical excitation.
Finite-time bidirectional dynamics
Even when the system is initially prepared entirely in state , both exchange rates influence the measured signal at finite times. Once downhill transfer produces a nonzero population , the uphill channel enters the evolution through the full propagator . The corresponding instantaneous effective decay rate is
Unlike the initial slope, Eq. (73) generally depends on both and . A linewidth inferred from a finite temporal window may therefore encode the full bidirectional exchange dynamics and need not coincide with the initial-slope result.
Physical interpretation
The derivation establishes three distinct points. First, the contribution of population exchange to an action-detected linewidth is not determined by the exchange kinetics alone. It is weighted by the detection observable through the ratio . Accordingly, within the minimal model considered here, the observed linewidth contains three distinct contributions: pure dephasing, population relaxation, and observable-weighted population redistribution. The action-detection dependence resides in the projection of the redistribution dynamics onto the chosen observable rather than in a separate additive decay channel. Second, following selective optical preparation of state , the initial logarithmic slope contains only the instantaneous downhill rate . Finite temperature changes this rate through detailed balance, but the uphill rate does not enter independently at . Third, both uphill and downhill rates enter either for a mixed nonequilibrium initial population or through the finite-time evolution of the action signal. These results distinguish the initial-slope linewidth used in the minimal analytic treatment from the more general bidirectional dynamics contained in the full Liouvillian propagation.
2DES lineshape as a function of excited-state transport
Figure S1. Simulated 2DES normalized amplitude rephasing spectra used to extract the homogeneous linewidth for coherent emission in Figure 4. Alt text: Five simulated normalized rephasing 2DES amplitude spectra for coherent emitted-field detection, shown for increasing downhill population-transfer rates from 0.1 to 20 meV. Each spectrum contains two diagonal resonances and associated cross-peak features. As the transfer rate increases, the higher-energy resonance and cross-peak structure change progressively in intensity and lineshape. These spectra are used to extract the coherent-emission homogeneous linewidths reported in Figure 4.
Figure S2. Simulated 2DES normalized amplitude rephasing spectra used to extract the homogeneous linewidth for action detection in Figure 4.
Alt text: Five simulated normalized rephasing 2DES amplitude spectra for action detection, shown for increasing downhill population-transfer rates from 0.1 to 20 meV. The spectra contain two diagonal resonances and associated cross-peak features, with broader lineshapes than the corresponding coherent-emission spectra. Increasing population transfer progressively modifies the relative intensities and lineshapes of the resonances and cross peaks. These spectra are used to extract the action-detected homogeneous linewidths reported in Figure 4.
References
- [1]D. M. Jonas, “Two-dimensional femtosecond spectroscopy,” Annu. Rev. Phys. Chem. 54, 425–463 (2003).pubmed.ncbi.nlm.nih.gov/12626736
- [2]M. Cho, “Coherent two-dimensional optical spectroscopy,” Chem. Rev. 108, 1331–1418 (2008).DOI
- [3]N. S. Ginsberg, Y.-C. Cheng, and G. R. Fleming, “Two-dimensional electronic spectroscopy of molecular aggregates,” Acc. Chem. Res. 42, 1352–1363 (2009).DOI
- [4]P. Nuernberger, S. Ruetzel, and T. Brixner, “Multidimensional electronic spectroscopy of photochemical reactions,” Angew. Chem. Int. Ed. 54, 11368–11386 (2015).DOI
- [5]F. D. Fuller and J. P. Ogilvie, “Experimental implementations of two-dimensional fourier transform electronic spectroscopy,” Annu. Rev. Phys. Chem. 66, 667–690 (2015).DOI
- [6]G. Nardin, T. M. Autry, G. Moody, R. Singh, H. Li, and S. T. Cundiff, “Multi-dimensional coherent optical spectroscopy of semiconductor nanostructures: Collinear and non-collinear approaches,” J. Appl. Phys. 117 (2015), 10.1063/1.4913830.DOI
- [7]T. A. Oliver, “Recent advances in multidimensional ultrafast spectroscopy,” R. Soc. Open Sci. 5 (2018), 10.1098/rsos.171425.
- [8]Y.-x. Weng, “Detection of electronic coherence via two-dimensional electronic spectroscopy in condensed phase,” Chin. J. Chem. Phys. 31, 135–151 (2018).DOI
- [9]E. Collini, “2D electronic spectroscopic techniques for quantum technology applications,” J. Phys. Chem. C 125, 13096–13108 (2021).DOI
- [10]S. Biswas, J. Kim, X. Zhang, and G. D. Scholes, “Coherent two-dimensional and broadband electronic spectroscopies,” Chem. Rev. 122, 4257–4321 (2022).DOI
- [11]E. Fresch, F. V. Camargo, Q. Shen, C. C. Bellora, T. Pullerits, G. S. Engel, G. Cerullo, and E. Collini, “Two-dimensional electronic spectroscopy,” Nat. Rev. Methods Primers 3, 84 (2023).
- [12]P. F. Tekavec, G. A. Lott, and A. H. Marcus, “Fluorescence-detected two-dimensional electronic coherence spectroscopy by acousto-optic phase modulation,” J. Chem. Phys. 127, 214307 (2007).pubmed.ncbi.nlm.nih.gov/18067357
- [13]A. K. De, D. Monahan, J. M. Dawlaty, and G. R. Fleming, “Two-dimensional fluorescence-detected coherent spectroscopy with absolute phasing by confocal imaging of a dynamic grating and 27-step phase cycling,” J. Chem. Phys. 140, 194201 (2014).DOI
- [14]P. Grégoire, E. Vella, M. Dyson, C. M. Bazán, R. Leonelli, N. Stingelin, P. N. Stavrinou, E. R. Bittner, and C. Silva, “Excitonic coupling dominates the homogeneous photoluminescence excitation linewidth in semicrystalline polymeric semiconductors,” Phys. Rev. B 95, 180201 (2017).DOI
- [15]P. Grégoire, A. R. Srimath Kandada, E. Vella, C. Tao, R. Leonelli, and C. Silva, “Incoherent population mixing contributions to phase-modulation two-dimensional coherent excitation spectra,” J. Chem. Phys. 147, 114201 (2017).DOI
- [16]S. Draeger, S. Roeding, and T. Brixner, “Rapid-scan coherent 2D fluorescence spectroscopy,” Opt. Express 25, 3259–3267 (2017).DOI
- [17]S. Goetz, D. Li, V. Kolb, J. Pflaum, and T. Brixner, “Coherent two-dimensional fluorescence micro-spectroscopy,” Opt. Express 26, 3915–3925 (2018).DOI
- [18]L. Bruder, U. Bangert, M. Binz, D. Uhl, and F. Stienkemeier, “Coherent multidimensional spectroscopy in the gas phase,” J. Phys. B: At. Mol. Opt. Phys. 52, 183501 (2019).DOI
- [19]P. Malý and T. Mančal, “Signatures of exciton delocalization and exciton–exciton annihilation in fluorescence-detected two-dimensional coherent spectroscopy,” J. Phys. Chem. Lett. 9, 5654–5659 (2018).arxiv.org/abs/1807.04846
- [20]P. Malý, J. Lüttig, S. Mueller, M. H. Schreck, C. Lambert, and T. Brixner, “Coherently and fluorescence-detected two-dimensional electronic spectroscopy: Direct comparison on squaraine dimers,” Phys. Chem. Chem. Phys. (2020), 10.1039/D0CP03218B.DOI
- [21]D. Paleček, P. Edlund, E. Gustavsson, S. Westenhoff, and D. Zigmantas, “Potential pitfalls of early-time dynamics in two-dimensional electronic spectroscopy,” J. Chem. Phys. 151, 024201 (2019).DOI
- [22]A. A. S. Halsey, F. Damtie, and K. J. Karki, “Differentiation of true nonlinear and incoherent mixing of linear signals in action-detected 2D spectroscopy,” J. Phys. Chem. A 123, 4119–4124 (2019).DOI
- [23]S. Mueller and T. Brixner, “Molecular coherent three-quantum two-dimensional fluorescence spectroscopy,” J. Phys. Chem. Lett. 11, 5139–5147 (2020).DOI
- [24]D. Sharkangelou, A. Javed, F. Sessa, X. Solinas, M. Joffre, and J. P. Ogilvie, “Phase-modulated rapid-scanning fluorescence-detected two-dimensional electronic spectroscopy,” J. Chem. Phys. 155 (2021), 10.1063/5.0057649.
- [25]E. Gutiérrez-Meza, R. Malatesta, H. Li, I. Bargigia, A. R. Srimath Kandada, D. A. Valverde-Chávez, S.-M. Kim, H. Li, N. Stingelin, S. Tretiak, et al., “Frenkel biexcitons in hybrid HJ photophysical aggregates,” Sci. Adv. 7, eabi5197 (2021).DOI
- [26]A. Liu, D. B. Almeida, L. A. Padilha, and S. T. Cundiff, “Perspective: Multidimensional coherent spectroscopy of perovskite nanocrystals,” J. Phys. Mater. 5, 021002 (2022).
- [27]M. Bruschi, F. Gallina, and B. Fresch, “Simulating action-2D electronic spectroscopy of quantum dots: insights on the exciton and biexciton interplay from detection-mode and time-gating,” Phys. Chem. Chem. Phys. 24, 27645–27659 (2022).DOI
