Quasi-Classical Doorway-Window Approximation

• The paper introduces quasi-classical reduction by factorizing quantum response functions into distinct doorway and window components.
• It employs a mixed quantum-classical framework with Wigner sampling and trajectory propagation to model nonlinear spectroscopic signals.
• The method is validated through pump-probe and multidimensional spectroscopies, enabling efficient on-the-fly ab initio simulations.

Searching arXiv for the cited topic and papers to ground the article in current literature. The quasi-classical doorway-window approximation is a factorized representation of time-resolved nonlinear spectroscopic signals in which the preparation of a nonstationary molecular ensemble by one set of pulses is encoded in a “doorway” function and its later interrogation by another set of pulses is encoded in a “window” function. In the formulation summarized in "WaveMixings.jl: a Julia package for performing on-the-fly time-resolved nonlinear electronic spectra from quasi-classical trajectories" (Vasquez et al., 4 Sep 2025) and "Ab Initio Simulation of Femtosecond Time-Resolved Multi-Pulse Spectroscopies applied to the Heptazine$\cdots$H$_2$O Complex" (Pios et al., 24 Mar 2026), the exact quantum response is reduced to averages over quasi-classical trajectories by combining a Wigner representation of the initial nuclear density with classical or surface-hopping propagation of nuclei. The resulting framework is designed for pump-probe, dispersed transient absorption, two-dimensional, 2D-FLEX, strong-field, pump-push-probe, and pump-induced two-dimensional spectroscopies, and is explicitly intended for on-the-fly electronic-structure simulations (Vasquez et al., 4 Sep 2025).

1. Response-function formulation and the doorway-window factorization

In the third-order response-function formalism, the heterodyne-detected four-wave mixing signal can be written, under the rotating-wave and impulsive approximations and assuming well-separated non-overlapping pulses, as a sum over ground-state bleach (GSB), stimulated emission (SE), and excited-state absorption (ESA) pathways (Vasquez et al., 4 Sep 2025). For pump-probe spectroscopy this takes the form

$$S^{(3)}(T_1,T_2,T_3)\propto \mathrm{Re}\sum_{k=\mathrm{GSB,SE,ESA}} a_k\,\mathrm{Tr}\!\left[W_k(T_3)\,e^{iLT_2}\,D(T_1)\right],$$

where $T_1$ is the coherence time between the first two pump pulses, $T_2\equiv T$ is the population delay, $T_3\equiv T'$ is the detection-time coherence interval, $L\rho\equiv[H,\rho]/\hbar$ is the Liouvillian, $D(T_1)$ is the doorway operator, and $W_k(T_3)$ is the window operator for pathway $k$. The coefficients are $_2$0 for GSB and SE and $_2$1 for ESA (Vasquez et al., 4 Sep 2025).

A conceptually equivalent formulation appears in the multi-pulse generalization, where the detected signal is reorganized so that the preparation stage and the detection stage enter as separate phase-space factors (Pios et al., 24 Mar 2026). In that representation, the pump-probe signal is written schematically as

$_2$2

with nuclear phase-space point $_2$3, doorway $_2$4, and window $_2$5 (Pios et al., 24 Mar 2026). This factorized structure is the defining feature of the approximation.

The terminology is operational. The doorway encodes how the initial pulse sequence prepares an excited-state population or coherence, while the window encodes how the later pulse sequence detects the system through bleach, emission, or absorption channels (Pios et al., 24 Mar 2026). In this sense, the approximation separates state preparation from readout without discarding the dependence of either object on nuclear geometry, transition dipoles, or pulse spectra.

2. Quasi-classical reduction from quantum response to trajectory averages

The quasi-classical approximation is obtained by reorganizing the trace so that nuclei evolve classically on adiabatic surfaces while electronic coherence contributions are evaluated at fixed nuclear geometries (Vasquez et al., 4 Sep 2025). After a Wigner representation of the ground-state nuclear density and Monte Carlo sampling of initial conditions $_2$6, the pump-probe signal becomes

$_2$7

with explicit doorway and window functions built from adiabatic energy gaps, transition-dipole magnitudes, spectral envelopes, and the ground-state Wigner distribution (Vasquez et al., 4 Sep 2025).

For the integral pump-probe signal, the doorway is

$_2$8

and the three windows are

$_2$9

$$S^{(3)}(T_1,T_2,T_3)\propto \mathrm{Re}\sum_{k=\mathrm{GSB,SE,ESA}} a_k\,\mathrm{Tr}\!\left[W_k(T_3)\,e^{iLT_2}\,D(T_1)\right],$$0

$$S^{(3)}(T_1,T_2,T_3)\propto \mathrm{Re}\sum_{k=\mathrm{GSB,SE,ESA}} a_k\,\mathrm{Tr}\!\left[W_k(T_3)\,e^{iLT_2}\,D(T_1)\right],$$1

Here $$S^{(3)}(T_1,T_2,T_3)\propto \mathrm{Re}\sum_{k=\mathrm{GSB,SE,ESA}} a_k\,\mathrm{Tr}\!\left[W_k(T_3)\,e^{iLT_2}\,D(T_1)\right],$$2 are transition-dipole magnitudes, $$S^{(3)}(T_1,T_2,T_3)\propto \mathrm{Re}\sum_{k=\mathrm{GSB,SE,ESA}} a_k\,\mathrm{Tr}\!\left[W_k(T_3)\,e^{iLT_2}\,D(T_1)\right],$$3 are adiabatic energy gaps, $$S^{(3)}(T_1,T_2,T_3)\propto \mathrm{Re}\sum_{k=\mathrm{GSB,SE,ESA}} a_k\,\mathrm{Tr}\!\left[W_k(T_3)\,e^{iLT_2}\,D(T_1)\right],$$4 and $$S^{(3)}(T_1,T_2,T_3)\propto \mathrm{Re}\sum_{k=\mathrm{GSB,SE,ESA}} a_k\,\mathrm{Tr}\!\left[W_k(T_3)\,e^{iLT_2}\,D(T_1)\right],$$5 are pump and probe spectral envelopes, and $$S^{(3)}(T_1,T_2,T_3)\propto \mathrm{Re}\sum_{k=\mathrm{GSB,SE,ESA}} a_k\,\mathrm{Tr}\!\left[W_k(T_3)\,e^{iLT_2}\,D(T_1)\right],$$6 is the ground-state Wigner distribution (Vasquez et al., 4 Sep 2025).

The same reduction is described in the multi-pulse treatment as the replacement of the exact quantum trace by a Wigner phase-space integral,

$$S^{(3)}(T_1,T_2,T_3)\propto \mathrm{Re}\sum_{k=\mathrm{GSB,SE,ESA}} a_k\,\mathrm{Tr}\!\left[W_k(T_3)\,e^{iLT_2}\,D(T_1)\right],$$7

together with classical phase-space evolution $$S^{(3)}(T_1,T_2,T_3)\propto \mathrm{Re}\sum_{k=\mathrm{GSB,SE,ESA}} a_k\,\mathrm{Tr}\!\left[W_k(T_3)\,e^{iLT_2}\,D(T_1)\right],$$8 (Pios et al., 24 Mar 2026). This places the doorway-window approximation at the interface of nonlinear response theory, semiclassical initial-state sampling, and mixed quantum-classical dynamics.

3. Core assumptions and their domain of validity

The derivation in the quasi-classical doorway-window framework invokes four explicit approximations (Vasquez et al., 4 Sep 2025). First, pump and probe are assumed non-overlapping and separated by $$S^{(3)}(T_1,T_2,T_3)\propto \mathrm{Re}\sum_{k=\mathrm{GSB,SE,ESA}} a_k\,\mathrm{Tr}\!\left[W_k(T_3)\,e^{iLT_2}\,D(T_1)\right],$$9 pulse durations, which allows mixed-time integrals to factorize into separate doorway and window operators. Second, nuclear motion during each ultrashort pulse is neglected, so nuclear coordinates are frozen at the time of each light-matter interaction. Third, electronic coherences during the short coherence intervals $T_1$0 and $T_1$1 evolve at fixed $T_1$2, contributing phase factors $T_1$3 but no nuclear dynamics. Fourth, the population evolution during $T_1$4 is replaced by classical or surface-hopping trajectories, so the quantum trace becomes a Monte Carlo average over sampled initial conditions and stochastic hops (Vasquez et al., 4 Sep 2025).

The range of validity is stated directly. Assumptions (i)–(iii) hold when pulses are $T_1$5 fs and $T_1$6 are controlled by pulse durations in pump-probe conditions (Vasquez et al., 4 Sep 2025). Assumption (iv) inherits the accuracy of the chosen mixed quantum-classical propagation scheme, with Tully’s surface hopping given as an example (Vasquez et al., 4 Sep 2025).

The multi-pulse treatment emphasizes analogous constraints. It states that the framework is based on the usual “short, non-overlapping pulses” assumption and identifies limitations that follow from the approximation: it neglects nuclear and electronic coherences beyond classical transport, cannot describe interference between pathways, requires non-overlapping short pulses, and produces signal artifacts at overlap times $T_1$7 that are not quantitative (Pios et al., 24 Mar 2026). Accuracy also inherits that of the classical-trajectory and electronic-structure level; the example given is that ADC(2) may mis-locate high states (Pios et al., 24 Mar 2026).

These statements delimit a common misconception. The method is not a full quantum treatment of nonlinear response, even though it starts from quantum response functions. It is instead a quasi-classical reduction that preserves explicit pulse, energy-gap, and transition-dipole dependence while replacing exact nuclear quantum propagation by classical or mixed quantum-classical dynamics (Vasquez et al., 4 Sep 2025).

4. Signal classes and generalizations beyond pump-probe

Within the doorway-window construction, transient absorption pump-probe is the basic third-order case. The multi-pulse summary gives the pump-probe signal as

$T_1$8

with

$T_1$9

and windows

$T_2\equiv T$0

$T_2\equiv T$1

The minus sign of ESA is explicitly identified as reflecting its opposite phase contribution in heterodyne detection (Pios et al., 24 Mar 2026).

The framework extends to dispersed-probe, 2D, 2D-FLEX, and strong-field expressions by introducing electronic dephasing rates $T_2\equiv T$2, Fourier transforms in detection times, or non-linear Rabi phase factors (Vasquez et al., 4 Sep 2025). In the WaveMixings.jl presentation, the same doorway-window logic is stated to apply to dispersed pump-probe, 2D spectroscopy, 2D-FLEX, and strong-field formulas by changing the doorway and window definitions in a controlled way (Vasquez et al., 4 Sep 2025).

The 2026 generalization broadens the method to multi-pulse spectroscopies (Pios et al., 24 Mar 2026). For pump-push-probe (PPP), the preparation consists of a pump step followed by a push step, so the doorway becomes

$T_2\equiv T$3

with $T_2\equiv T$4, and the probe windows are evaluated at $T_2\equiv T$5. The signal is then

$T_2\equiv T$6

(Pios et al., 24 Mar 2026).

For pump-induced two-dimensional spectroscopy (P-2D), the quasi-classical doorway-window form is

$T_2\equiv T$7

where $T_2\equiv T$8 and $T_2\equiv T$9 are analytic expressions almost identical to those in standard 2D spectroscopy, but with the reference ensemble $T_3\equiv T'$0 taken from the pump-prepared distribution at $T_3\equiv T'$1 (Pios et al., 24 Mar 2026). This suggests that the doorway-window architecture is not tied to a specific pulse count; rather, it provides a modular factorization pattern whose complexity grows through nested preparation stages and modified detection functionals.

5. Computational realization in on-the-fly simulations

The approximation is explicitly intended for on-the-fly simulations driven by ab initio electronic-structure data (Vasquez et al., 4 Sep 2025). In WaveMixings.jl, the required input for each trajectory $T_3\equiv T'$2 is the time series of adiabatic energies $T_3\equiv T'$3 and transition dipoles $T_3\equiv T'$4, together with the sampling weights (Vasquez et al., 4 Sep 2025). The high-level workflow for an integral pump-probe signal consists of seven steps: generation of $T_3\equiv T'$5 initial conditions from the Wigner distribution, sampling of the initial excited state $T_3\equiv T'$6 from the pump-weighted transition probability, doorway evaluation, ground-state propagation, GSB-window assembly, excited-state propagation on manifold $T_3\equiv T'$7 using Tully’s surface hopping, and SE/ESA window evaluation, followed by Monte Carlo averaging,

$T_3\equiv T'$8

(Vasquez et al., 4 Sep 2025).

The Julia implementation is organized around a main composite type, $_2$06 with modules for data import, constructors and unit conversion, post-processing, and core signal evaluation (Vasquez et al., 4 Sep 2025). The package contains parsing of surface-hopping outputs, filtering of completed trajectories, direct import of preprocessed arrays of energies and dipoles, unit converters, pulse-shape generators for Gaussian, $T_3\equiv T'$9, and Lorentzian profiles, normalization, interpolation and smoothing, difference maps, orientational averages, and plotting routines (Vasquez et al., 4 Sep 2025). Core functions include pathway-resolved and modality-resolved methods such as gsb!, se!, esa!, gsb_dispersion!, gsb_strong!, gsb_2D!, se_2D!, esa_2D!, flex_2D!, and trf! (Vasquez et al., 4 Sep 2025).

The implementation details are paired with explicit performance choices: all inner loops are written in type-stable Julia code; hot functions are annotated with @inbounds and @simd; heavy linear algebra uses BLAS from LinearAlgebra.jl; multiple dispatch on WaveMixing and signal types eliminates if/else at runtime; and external dependencies are minimized to LinearAlgebra and DelimitedFiles from the standard library (Vasquez et al., 4 Sep 2025). The package is open-source on Codeberg.org, and its docstrings are accessible via Julia help mode (Vasquez et al., 4 Sep 2025).

The on-the-fly protocol described for the heptazine$L\rho\equiv[H,\rho]/\hbar$0H$L\rho\equiv[H,\rho]/\hbar$1O complex provides a complementary ab initio realization (Pios et al., 24 Mar 2026). It starts by sampling the $L\rho\equiv[H,\rho]/\hbar$2 harmonic Wigner distribution of the electronic ground state using an MP2 Hessian, computing vertical excitation energies and transition dipoles to manifold I at the ADC(2) level with basis cc-pVDZ(+aug), selecting pump-resonant trajectories weighted by $L\rho\equiv[H,\rho]/\hbar$3, and propagating them with velocity-Verlet plus Landau-Zener surface hopping using the Belyaev-Lebedev formula (Pios et al., 24 Mar 2026). For PPP, higher-state energies and transition dipoles to manifold II are evaluated at the push time, push-resonant trajectories are selected, and propagation continues on $L\rho\equiv[H,\rho]/\hbar$4 before probe windows are accumulated into the signal average (Pios et al., 24 Mar 2026). Practical details are explicitly specified: MP2/cc-pVDZ for the ground state, ADC(2)/aug-cc-pVDZ for excitations, TURBOMOLE for electronic structure, and ZagHop for surface hopping (Pios et al., 24 Mar 2026).

6. Benchmarks, applications, and interpretive scope

An illustrative benchmark in WaveMixings.jl reproduces the transient absorption pump-probe spectrum of pyrazine from Gelin et al., J. Chem. Theory Comput. 17, 2394 (2021) (Vasquez et al., 4 Sep 2025). The setup uses 591 completed surface-hopping trajectories, obtained from 600 launched trajectories filtered to 591, with manifold $L\rho\equiv[H,\rho]/\hbar$5 containing 4 states up to 5.2 eV and manifold $L\rho\equiv[H,\rho]/\hbar$6 containing 30 states up to 10 eV (Vasquez et al., 4 Sep 2025). Gaussian pump and probe pulses with $L\rho\equiv[H,\rho]/\hbar$7 fs are employed; the pump is centered at $L\rho\equiv[H,\rho]/\hbar$8 eV with 200 steps over $L\rho\equiv[H,\rho]/\hbar$9 eV, and the probe spans 2.0–6.0 eV in 0.02 eV steps (Vasquez et al., 4 Sep 2025).

The reported performance for the full GSB+SE+ESA integral pump-probe signal on 591 trajectories and a $D(T_1)$0 $D(T_1)$1 grid is a total wall-clock time of approximately 128 s on a 6-core i5-12500H laptop with Julia 1.13.0-DEV, with a typical memory footprint below 2 GB (Vasquez et al., 4 Sep 2025). Convergence is quantified: signals converge within 5% RMS error for $D(T_1)$2 trajectories, and doubling to 1,000 trajectories reduces noise to $D(T_1)$3 (Vasquez et al., 4 Sep 2025). Accuracy is described in pathway-resolved and modality-resolved terms: the package reproduces peak positions and oscillatory GSB and ESA trace features with period $D(T_1)$4 fs in perfect agreement with published results; dispersed pump-probe and TRF spectra with dephasing $D(T_1)$5 eV fully match reference line shapes; 2D nonrephasing and rephasing signals at $D(T_1)$6 fs show correct cross-peak patterns; and 2D-FLEX reveals vibronic doublets at $D(T_1)$7 in line with theory (Vasquez et al., 4 Sep 2025). Parameter studies further state that narrowing the pulse duration to $D(T_1)$8 fs broadens spectral features by $D(T_1)$9 meV, that dephasing $W_k(T_3)$0 eV yields $W_k(T_3)$1 fs $W_k(T_3)$2 coherence lifetimes in dispersed signals, and that alternative state-sampling distributions affect early-time GSB amplitude by $W_k(T_3)$3 (Vasquez et al., 4 Sep 2025).

The heptazine$W_k(T_3)$4H$W_k(T_3)$5O application demonstrates the extension to multi-pulse spectroscopy (Pios et al., 24 Mar 2026). The protocol yields approximately 300 trajectories on manifold I after pump selection and approximately 287 trajectories on $W_k(T_3)$6 after push selection, with $W_k(T_3)$7 fs, $W_k(T_3)$8 up to 100 fs, and $W_k(T_3)$9 fs (Pios et al., 24 Mar 2026). The reported spectral interpretation is that pump-stimulated experiments provide much richer information on ultrafast radiationless relaxation dynamics than conventional pump-probe and 2D experiments (Pios et al., 24 Mar 2026). Specifically, pump-probe SE and ESA reproduce ultrafast $k$0 $k$1 internal conversion and the appearance of a vibronically hot $k$2 state at $k$3 eV; PPP shows a push-induced re-excitation of the hot $k$4 ensemble followed by rapid cascade back to $k$5, with an intense oscillating SE band near 2.8 eV that red-shifts and decays over $k$6 fs and a broad ESA from 0–4 eV that builds up over $k$7 fs; and P-2D maps show a $k$8 eV-wide emission band below the diagonal even at $k$9 fs, together with broad elongated ESA attributed to population cycling among bright $_2$00–$_2$01 and dark charge-transfer states (Pios et al., 24 Mar 2026).

These examples define the practical scope of the approximation. It is presented as numerically efficient because only a single ensemble of classical trajectories is needed and no multi-dimensional integrals over Liouville pathways are required; it is described as conceptually transparent because $_2$02 specifies preparation and $_2$03 specifies emission or absorption at the probe; and it is stated to be fully compatible with on-the-fly electronic structure for systems with many modes and many electronic states (Pios et al., 24 Mar 2026). A plausible implication is that the method is particularly well suited to studies where electronic-structure cost, rather than formal response-theory algebra, is the limiting factor.

7. Relation to neighboring methodologies and common points of confusion

The quasi-classical doorway-window approximation belongs to the broader class of mixed quantum-classical methods for nonlinear spectroscopy, but its distinctive feature is the explicit factorization of the signal into preparation and detection objects evaluated on trajectory ensembles (Vasquez et al., 4 Sep 2025). In the WaveMixings.jl formulation, this structure is preserved across integral and dispersed pump-probe, 2D, 2D-FLEX, TRF, and strong-field signals by changing the definitions of the relevant doorway and window functions rather than replacing the overall framework (Vasquez et al., 4 Sep 2025). In the multi-pulse extension, the same logic is carried over to PPP and P-2D by introducing additional preparation stages before the final window evaluation (Pios et al., 24 Mar 2026).

A frequent source of confusion is the role of coherence. The approximation does include electronic coherence phase factors during short coherence intervals $_2$04 and $_2$05, evaluated at fixed nuclear geometry (Vasquez et al., 4 Sep 2025). However, it does not retain nuclear and electronic coherences beyond classical transport and therefore cannot describe interference between pathways in the fully quantum sense (Pios et al., 24 Mar 2026). Another common confusion concerns pulse overlap. The factorization relies on short, non-overlapping pulses; signal artifacts at overlap times below 20 fs are explicitly stated to be non-quantitative (Pios et al., 24 Mar 2026). A further point is that the method is not independent of the chosen dynamics engine: its accuracy inherits that of the mixed quantum-classical propagation and the electronic-structure level employed, whether exemplified by Tully’s surface hopping (Vasquez et al., 4 Sep 2025) or by Landau-Zener surface hopping at the ADC(2) level (Pios et al., 24 Mar 2026).

Taken together, these features place the quasi-classical doorway-window approximation in a specific methodological niche. It is a reduction of nonlinear optical response theory to phase-space averages that preserves pathway resolution, pulse dependence, and direct compatibility with on-the-fly simulations, while accepting the standard semiclassical compromises associated with frozen-coordinate interactions, classical population transport, and non-overlapping ultrashort pulses (Vasquez et al., 4 Sep 2025).