On-the-Fly Lifting of Coarse Reaction-Coordinate Paths to Full-Dimensional Transition Path Ensembles

Christof Schuette, Alexander Sikorski, Jakob Kresse, Marcus Weber · 2026

Read On-the-Fly Lifting of Coarse Reaction-Coordinate Paths to Full-Dimensional Transition Path Ensembles by Christof Schuette, Alexander Sikorski, Jakob Kresse, Marcus Weber (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2603-23266.txt.

Introduction

Schütte and three co-authors solve a gap between cheap coarse models and full dynamics. Motion is often simulated on a few collective variables, which is fast but coarse. Turning a coarse path back into full atom-level paths is the hard part. Their method guides full-system runs to track the coarse path. A Girsanov reweighting then removes the bias from that guidance. Tests show coarse paths become realistic full paths with barrier crossings and detours. The weighted ensembles also speed up path statistics.

Catalog record

CREATORChristof Schuette, Alexander Sikorski, Jakob Kresse, Marcus Weber
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE69872
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2603.23266

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

On-the-Fly Lifting of Coarse Reaction-Coordinate Paths to Full-Dimensional Transition Path Ensembles Christof Schuette, Alexander Sikorski, Jakob Kresse, Marcus Weber Abstract Effective dynamics on a low-dimensional collective-variable (CV) or latent space can be simulated far more cheaply than the underlying high-dimensional stochastic system, but exploiting such coarse predictions requires lifting: turning a coarse CV trajectory into dynamically consistent full-dimensional states and path ensembles, without relying on global sampling of invariant or conditional fiber measures. We present a local, on-the-fly lifting strategy based on guided full-system trajectories. First an effective model in CV space is used to obtain a coarse reference trajectory. Then, an ensemble of full-dimensional trajectories is generated from a guided version of the original dynamics, where the guidance steers the trajectory to track the CV reference path. Because guidance biases the path distribution, we correct it via pathwise Girsanov reweighting, yielding a correct-by-construction importance-sampling approximation of the conditional law of the uncontrolled dynamics. We further connect the approach to stochastic optimal control, clarifying how coarse models can inform variance-reducing guidance for rare-event quantities. Numerical experiments demonstrate that inexpensive coarse transition paths can be converted into realistic full-system transition pathways (including barrier crossings and detours) and can accelerate estimation of transition pathways and statistics while providing minimal bias through weighted ensembles. 1 Introduction High-dimensional stochastic dynamics are ubiquitous in molecular simulation and related areas; a prototypical example for such dynamics is given by diffusion systems of the form dX_{t}\;=\;b(X_{t})\,dt+\sigma\,dW_{t},\qquad X_{t}\in\mathcal{X}\subset\mathbb{R}^{d}, (1) or similar forms like Langevin dynamics. In many applications, the dynamics is metastable : trajectories spend long times near a few dominant regions and only rarely transition between them. Direct numerical time-stepping in full dimension is then expensive, since small timesteps are required for stability and accuracy, while the rare transitions determine the relevant long-time statistics and mechanisms. Transfer-operator and variational perspectives provide a principled route to analyze such slow processes and to build reduced kinetic models such as Markov state models (MSMs); see, e.g., [ 45 , 44 , 5 , 27 ] . One of the main challenges is to reliably find the transition regions and construct representative transition paths between the few dominant regions. A standard route to reduce complexity is to introduce a low-dimensional collective variable (CV) map \xi:\mathcal{X}\to\mathbb{R}^{m} , m\ll d , introducing a coarse representation of the full system, Z_{t}:=\xi(X_{t}) , in a m -dimensional latent space. Among all the approaches to finding such CVs or reaction coordinates , some specific dynamically informative CVs are such that Z_{t} captures the essential slow degrees of freedom [ 53 ] . Once an informative CV is available, one may attempt to describe the evolution of Z_{t} by an effective (coarse) dynamics in \mathbb{R}^{m} , which can often be simulated at substantially reduced cost. The construction, analysis, and data-driven identification of such CVs and effective models has seen substantial recent progress. Representative examples include time-lagged independent component analysis (TICA) [ 41 ] , variational approaches for Markov processes (VAMP) and their neural implementations (VAMPnets) [ 35 ] , reaction-coordinate flow ideas [ 51 ] , as well as Koopman/dominant-subspace learning approaches such as ISOKANN [ 42 ] . These algorithmic developments connect to theoretical work on optimal reaction coordinates and exact dynamical coarse-graining without time-scale separation [ 33 ] , and to quantitative analyses of effective dynamics and pathwise error bounds [ 52 , 30 , 31 ] . For related perspectives on choosing CVs by minimizing the deviation between effective and full dynamics and on spectral properties of effective dynamics from conditional expectations, see [ 53 , 38 ] . Our central goal is to use knowledge about CVs and the related effective dynamics for computing transition rates and pathways of the full system under consideration in diffusive and molecular processes. There is extensive literature on computing transition kinetics and paths, including transition path sampling (TPS) [ 4 ] , interface-based approaches such as milestoning [ 20 ] and transition interface sampling (TIS) [ 37 ] , and transition path theory (TPT) [ 15 , 16 , 36 ] , for a conceptual review see [ 3 ] . Some of these techniques like milestoning or TIS really make use of knowledge about CVs. However, all of these techniques suffer from the fact that they always explore the full-dimensional system and are not guided by a dynamically reliable low-dimensional dynamics as given, e.g., by the effective dynamics. From coarse prediction to full-dimensional realization: the lifting bottleneck. Even when an effective model for Z_{t} is available, a fundamental difficulty remains: lifting . Concretely, suppose the full dynamics ( 1 ) has been simulated up to time T , yielding a state X_{T} , and that the effective dynamics has been advanced further to a later time T_{\mathrm{new}} , producing a coarse path \{Z_{t}\} over [T,T_{\mathrm{new}}] . The lifting problem asks: How can we generate a dynamically consistent full-dimensional path \{X_{t}\} over [T,T_{\mathrm{new}}] conditioned on the coarse path \{Z_{t}\} ? In particular, if \{Z_{t}\} is a transition path of the effective dynamics, how can we guarantee that \{X_{t}\} is a representative transition path of the full-dimensional system? Main idea: guided trajectories driven by a cheap coarse reference path. This work develops a local, on-the-fly lifting strategy that uses the effective dynamics to provide a coarse reference path in CV space and then reconstructs full-dimensional transition paths by simulating guided trajectories . The key mechanism is to introduce a controlled process dX_{t}^{u}\;=\;\bigl(b(X_{t}^{u})+\sigma u(t,X_{t}^{u})\bigr)\,dt+\sigma\,dW_{t}, (2) where the feedback control u drives the full-system dynamics in a way that is driven by the CV \xi and its effective dynamics on the latent space. Herein, we consider control forces of the general form u(t,x)\;=\;J_{\xi}(x)^{\top}\nabla\Psi(\xi(x),Z_{t}), (3) with J_{\xi} denoting the Jacobian of \xi and where Z_{t} may be additional information on the latent space. As we will see, the choice of the function \Psi depends on what problem is to be solved: (1) Fast exploration , where one aims at finding previously unknown part of state space like relevant transition channels of the system, or (2) accurate computation , where we want an efficient way to compute certain quantities like transition probabilities, or transition rates. For fast exploration , we herein propose the guided diffusion brigde algorithm with the effective tracking choice u(t,x)\;=\;J_{\xi}(x)^{\top}G_{t}\cdot\bigl(Z_{t}-\xi(x)\bigr), (4) so that \xi(X_{t}^{u}) tracks a prescribed path Z_{t} , e.g., given by the effective dynamics on [T,T_{\mathrm{new}}] , and G_{t}\succeq 0 is a (possibly time-dependent) gain (matrix or scalar). Intuitively, ( 4 ) nudges the full system toward the CV path produced by the cheap effective dynamics, while the uncontrolled drift b and the physical noise \sigma still shape the microscopic fluctuations. When \xi is informative, the resulting guided trajectories can realize physically plausible transition mechanisms in the full state space, even when the coarse path was generated at much lower cost. Similar approaches utilizing guided diffusion systems have been studied in a general context, cf. [ 40 , 12 , 43 ] , and for molecular dynamics [ 9 , 39 , 10 ] . In these approaches, however, the diffusion bridge is not guided by a dynamically-informative effective dynamics but by endpoints or respective distributions. For accurate computation , we will see that good choices of

Read the complete text