🔬 physics

A partially linearized mapping approach to surface hopping (MASH-PLDM) for nonadiabatic dynamics and nonlinear spectroscopy

The paper introduces MASH-PLDM, a partially linearized mapping approach that utilizes two spin vectors to improve the simulation of nonadiabatic dynamics and coherence effects in molecular systems while enabling the calculation of multi-time correlation functions for nonlinear spectroscopy.

Original authors: Jan Obermeier, Kasra Asnaashari, Jeremy O. Richardson

Published 2026-09-30

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Jan Obermeier, Kasra Asnaashari, Jeremy O. Richardson

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

In the microscopic world of molecules, atoms are never truly still. They vibrate, stretch, and twist, constantly reshaping the molecule's structure. When a molecule absorbs energy, such as from a flash of light, its electrons can jump to a higher energy state. This creates a delicate situation where the heavy, slow-moving atomic nuclei and the light, fast-moving electrons must interact. Sometimes, the molecule finds itself in a position where two different electronic states are equally possible, a moment known as a "nonadiabatic" crossing. At this precise instant, the molecule can split its path, with some parts of the wave-like electron cloud staying on one energy path while others jump to another. Predicting exactly how this happens is one of the most difficult challenges in chemistry. If scientists cannot model this splitting accurately, they cannot understand how molecules break apart, how they release energy, or how they interact with light in complex ways like those seen in modern spectroscopy.

For decades, researchers have tried to simulate these events by treating the heavy atoms as classical particles, like tiny billiard balls, while keeping the electrons as quantum waves. A common approach, known as surface hopping, imagines the molecule traveling on one specific energy path until it suddenly "hops" to another. Another method, called mapping, tries to smooth out the quantum behavior by translating it into continuous variables that look more like classical motion. Each method has its strengths and weaknesses. The surface hopping approach is good at describing the splitting of the wave but struggles with the subtle quantum interference patterns that occur when the paths cross. The mapping approach handles the interference well but often fails to capture the sharp, distinct splitting of the wave, instead forcing the molecule to travel on a blurry, average path that doesn't physically exist.

In a new study, researchers at ETH Zurich have developed a hybrid technique that combines the best features of both worlds. They call this new method MASH-PLDM. It is designed to simulate how molecules move and change when their electrons and nuclei interact in complex ways. The core idea is to let the molecule travel on a single, well-defined energy path, just like in the surface hopping method, but to use two separate "spin" vectors to guide the motion. Think of these spin vectors as two internal compasses. In most situations, both compasses point in the same direction, telling the molecule to travel on either the high-energy or low-energy path. However, when the molecule is in a state of quantum coherence—a state where it is effectively in two places at once—the compasses point in opposite directions. In this specific scenario, the molecule travels on an average path between the two energy levels. This allows the simulation to capture the delicate interference effects that usually get lost in simpler models, while still maintaining the sharp, realistic splitting of the wave that other methods miss.

The researchers tested this new approach on a variety of mathematical models that represent different types of molecular behavior. They started with simple scattering problems, where a particle moves toward a barrier and either bounces back or passes through. In these tests, the new method proved to be more accurate than previous techniques, correctly predicting how the particle would split its path without needing complex, expensive corrections. They also tested the method on models that simulate how molecules interact with a surrounding environment, such as a liquid solvent. In these cases, the new method was able to reproduce the results of exact quantum calculations with high precision, particularly in situations where the temperature was low and the quantum effects were strongest. The simulations showed that the method could handle the rapid oscillations of energy and the transfer of heat between the molecule and its surroundings far better than older methods that relied on averaging the paths.

One of the most significant achievements of this work is the ability to simulate multi-time correlation functions. In the real world, scientists often use ultrafast lasers to probe molecules, firing a first pulse to excite them and a second pulse to measure the result. The time between these pulses reveals how the molecule changes. To simulate this, a computer model must track the molecule's history over several distinct time intervals, not just a single moment. Previous methods struggled with this, often breaking down when asked to look at more than one time step. The new MASH-PLDM approach handles these multi-step calculations naturally. The researchers demonstrated this by simulating a pump-probe experiment on a model of the sodium iodide molecule. They were able to track how the molecule's spectrum changed over time as it crossed energy barriers and split into different states. The results matched the exact quantum mechanical predictions almost perfectly, capturing the splitting and reforming of spectral peaks as the molecule moved through its cycle.

The study also addressed a specific problem found in earlier methods: the issue of "unphysical" trajectories. In some simulations, molecules would end up on energy paths that they could not physically reach, or they would contribute to the final result in ways that made no sense. The researchers introduced a refinement to their method that acts like a filter, ensuring that only the physically relevant paths contribute to the final answer. By using specific mathematical rules to select the correct starting and ending points for the simulation, they eliminated these errors. This allowed the method to produce clean, accurate results even in difficult scenarios where the molecule had very little energy to begin with. The team found that while the method requires more computational power than some older techniques because it tracks two internal compasses instead of one, the trade-off is worth it for the accuracy it provides.

The implications of this work extend beyond just getting better numbers for simple models. The ability to accurately simulate how molecules split their paths and maintain quantum coherence over multiple time steps opens the door to understanding complex chemical reactions in full detail. This is particularly important for studying processes like bond breaking, where the smooth, averaged paths of older methods fail completely. By providing a rigorous way to track these events, the new method offers a powerful tool for exploring the photochemistry of molecules in their full complexity. While the researchers note that their current results are based on simulations of specific models, they believe the method will prove essential for studying real-world molecular systems where the interplay between electrons and nuclei determines the outcome of chemical reactions. The work stands as a significant step forward in the quest to bridge the gap between the quantum world of electrons and the classical world of moving atoms.

Problem Statement
Mixed quantum–classical methods are essential for simulating nonadiabatic molecular dynamics, where nuclear degrees of freedom are treated classically and electronic degrees of freedom quantum mechanically. However, existing approaches face significant limitations. The Ehrenfest approximation, while simple and capable of handling multi-time correlation functions required for nonlinear spectroscopy, violates detailed balance and fails to thermalize correctly. Conversely, surface-hopping methods like Tully's fewest-switches surface hopping (FSSH) and the more rigorous Mapping Approach to Surface Hopping (MASH) successfully describe wavepacket branching and detailed balance but are inherently mean-field in character regarding coherences and are traditionally limited to single-time correlation functions. Partially linearized methods, such as the Partially Linearized Density Matrix (PLDM) and spin-PLDM, offer improved accuracy for coherences and can naturally extend to multi-time correlation functions. However, they inherit the mean-field character of their underlying mapping approaches, leading to unphysical potentials (linear combinations of states) and failures in describing wavepacket branching, particularly in anharmonic systems or bond dissociation scenarios. There is a need for a method that combines the rigorous treatment of surface hopping (deterministic hops, correct branching) with the capability of partially linearized methods to handle coherences and multi-time correlation functions.

Methodology: MASH-PLDM
The authors derive a new method, MASH-PLDM (Partially Linearized Mapping Approach to Surface Hopping), which synthesizes the MASH formalism with the PLDM framework.

Theoretical Foundation: The derivation begins with the quantum-classical Liouville equation (QCLE) expressed via path integrals. Following the PLDM approach, the nuclear degrees of freedom are linearized (replaced by an average path), while the electronic mapping variables are treated fully.
Mapping Kernels: A critical innovation is the introduction of specific mapping kernels. Unlike previous spin-mapping approaches that use Stratonovich–Weyl kernels, MASH-PLDM employs kernels designed to reproduce the MASH force function. The mapping kernel $\hat{w}_q(S)$ maps operators to functions of spin vectors, while the inverse kernel $\hat{w}^{-1}_q(S)$ maps functions back to operators.
Dual Spin Vectors: The method utilizes two independent spin vectors, $S$ $S$ and $S'$ $S^{'}$ , corresponding to the forward and backward propagators of the density matrix.
- Nuclear Dynamics: The nuclear force is determined by the average of the forces generated by both spin vectors: $\dot{\bar{p}} = \frac{1}{2}[F(\bar{q}, S) + F(\bar{q}, S')]$ .
- Surface Hopping Behavior: If both spins reside in the same hemisphere (e.g., both "up"), the nuclei move on a single adiabatic surface (excited or ground). If the spins are in opposite hemispheres, the nuclei move on the average potential energy surface. This allows the method to describe electronic coherences by propagating on the average surface, a feature absent in standard MASH.
Momentum Rescaling: To ensure energy conservation, the method incorporates a momentum-rescaling procedure when a spin vector crosses the equator (the boundary between hemispheres). This is analogous to the hop in FSSH and MASH but adapted for the dual-spin system.
Multi-Time Correlation Functions: The formalism naturally extends to multi-time correlation functions (e.g., $R(t_1, t_2, t_3)$ ) by propagating the two spin vectors through multiple interaction points, making it suitable for simulating nonlinear spectroscopy.
Kernel Variants: The authors explore three variants of the inverse kernel:
1. Linear ( $\hat{w}^{-1}_{lin}$ ): The basic form used in the derivation.
2. Initial Projection ( $\hat{w}^{-1}_{init}$ ): Projects trajectories based on their initial spin configuration.
3. Final Projection ( $\hat{w}^{-1}_{fin}$ ): Projects trajectories based on their final spin configuration.
  The projection variants are designed to filter out "unphysical" trajectories where spins are in the "wrong" hemispheres relative to the electronic state being measured, thereby improving accuracy in scattering and coherence problems.

Key Results
The method was benchmarked against exact quantum results (QUAPI, split-operator) across several model systems:

Tully Scattering Models:
- Model I (Single Avoided Crossing): MASH-PLDM with the linear kernel initially failed to capture wavepacket branching due to contributions from trajectories on the "wrong" surfaces. However, using the initial or final projection kernels recovered the exact quantum result almost perfectly.
- Model II (Dual Avoided Crossing): MASH-PLDM with energy-based initial conditions and the final projection kernel corrected errors found in spin-PLDM and the original MASH method, providing superior accuracy in transmission probabilities.
- Model III (Extended Coupling): The method successfully captured the step-like nature of scattering probabilities and electronic coherences without the need for ad hoc decoherence corrections, outperforming standard MASH and FSSH.
Initialization in a Coherence:
- In a model initialized with an electronic coherence, MASH-PLDM correctly described the short-time decoherence and the peak at 20 fs, whereas standard MASH failed to capture the correct short-time decay.
- The method demonstrated that propagating on the average surface is a valid and beneficial approximation for describing coherences.
Spin–Boson Models:
- Across symmetric/asymmetric, high/low temperature, and weak/strong coupling regimes, MASH-PLDM showed results comparable to or more accurate than spin-PLDM and MASH.
- Notably, at low temperatures where MASH exhibited overdamped oscillations, MASH-PLDM captured the amplitude of oscillations with higher accuracy.
Nonlinear Spectroscopy (NaI Pump–Probe):
- The authors performed a proof-of-principle simulation of time-resolved pump–probe spectroscopy for a 3-state NaI model.
- MASH-PLDM (specifically with the final projection) reproduced the exact quantum mechanical spectra, including the splitting and reformation of spectral peaks as the wavepacket crossed the nonadiabatic region.
- In contrast, spin-PLDM and the linear-kernel MASH-PLDM showed artificial broadening due to the failure to describe wavepacket branching.

Significance and Claims
The paper claims that MASH-PLDM represents a significant advancement by unifying the strengths of surface hopping and partially linearized methods:

Wavepacket Branching: Unlike mean-field PLDM variants, MASH-PLDM describes wavepacket branching deterministically without requiring expensive iterative quantum-jump procedures.
Coherence Treatment: Unlike the original MASH, MASH-PLDM can propagate trajectories on an average surface, providing a more accurate description of electronic coherences.
Multi-Time Correlation: The method retains the ability of PLDM to treat multi-time correlation functions, enabling the simulation of nonlinear spectroscopy (e.g., pump–probe) where nonadiabatic dynamics occur during all time delays ( $t_1, t_2, t_3$ ), not just the waiting time.
Rigorous Connection: The method maintains a rigorous connection to the QCLE in the short-time limit and conserves energy through its momentum-rescaling procedure.

The authors note that while MASH-PLDM requires more trajectories to converge than MASH (due to sampling two spin spheres), it is particularly well-suited for short-time properties relevant to ultrafast spectroscopy. They suggest that the method's true potential will be realized in complex molecular systems involving bond breaking, where mean-field dynamics fail, and where the ability to handle both coherences and multi-time correlations is critical.

More like this