Atom-centered electric multipole moments dynamically generated from QM/MM MD simulations
Andrea Levy, Andrej Antalík, Jógvan Magnus Haugaard Olsen, Ursula Rothlisberger · 2025
Read Atom-centered electric multipole moments dynamically generated from QM/MM MD simulations by Andrea Levy, Andrej Antalík, Jógvan Magnus Haugaard Olsen, Ursula Rothlisberger (2025) 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-2511-02630.txt.
Introduction
Chemists often sum up the electric charge pattern of a molecule with markers placed on each atom. Andrea Levy and colleagues present xDRESP, a method that computes such markers from mixed quantum and classical simulations, in a 2025 paper with Andrej Antalik, Jogvan Olsen, and Ursula Rothlisberger. Unlike fixed charge models, their markers update as the simulated molecule moves.
The team checks how well the new charges reproduce the true electric field around a molecule, plus its overall multipole pattern. They also show cases where the moving markers reveal useful chemistry. The markers can flag when polarization matters in a system. During a reaction, they can track shifts in electron density as the change unfolds, live on screen.
Catalog record
| CREATOR | Andrea Levy, Andrej Antalík, Jógvan Magnus Haugaard Olsen, Ursula Rothlisberger |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 82 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2511-02630
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Atom-centered electric multipole moments dynamically generated from QM/MM MD simulations Andrea Levy, Andrej Antalík, Jógvan Magnus Haugaard Olsen, Ursula Rothlisberger Abstract Atom-centered electric multipole moments can be extremely useful in chemistry as they enable the systematic mapping of a complex electrostatic problem to a simpler model. However, since they do not correspond to physical observables, there is no unique way to define them. In this work, we present an extension of the dynamically generated RESP charges (D-RESP) method, referred to as xDRESP, where atom-centered multipoles are computed from mixed quantum mechanics/molecular mechanics (QM/MM) molecular dynamics simulations. We compare the ability of xDRESP charges to reproduce the electrostatic potential, as well as molecular multipoles, against the performance of fixed point-charge models commonly used in force fields. Moreover, we highlight cases where xDRESP atomic multipoles can provide valuable information about chemical systems, such as indicating when polarization plays a significant role, and chemical reactions, in which xDRESP atomic multipoles can be used as an on-the-fly analysis tool to track changes in electron density. † † affiliation: Laboratory of Computational Chemistry and Biochemistry, Ecole Polytechnique Fédérale de Lausanne, Lausanne, Switzerland † † affiliation: Laboratory of Computational Chemistry and Biochemistry, Ecole Polytechnique Fédérale de Lausanne, Lausanne, Switzerland † † affiliation: DTU Chemistry, Technical University of Denmark, Lyngby, Denmark † † affiliation: Laboratory of Computational Chemistry and Biochemistry, Ecole Polytechnique Fédérale de Lausanne, Lausanne, Switzerland † † email: [email protected] † † suppinfo: S 1 Introduction Atomic charges or, more generally, atom-centered electric multipole moments, are an extremely useful concept in chemistry that can help to describe the electrostatic properties of complex molecular systems by using a simple model. Yet, as they do not correspond to a physical observable, i.e., there is no quantum mechanical operator whose eigenvalues correspond to such quantities Oliferenko et al. 2006 ; Cho et al. 2020 , they cannot be defined in a unique way in a quantum mechanical framework. Over the years, researchers have proposed different methods to define effective atomic charges, which can be classified into three main categories. The first are methods based on the partitioning of the wavefunction, such as Mulliken Mulliken 1955 or Löwdin Löwdin 1950 charges, natural population analysis Reed et al. 1985 , or CM x charge models Storer et al. 1995 ; Li et al. 1998 ; Winget et al. 2002 ; Kelly et al. 2005 ; Olson et al. 2007 ; Marenich et al. 2012 ( x =1–5 or 4M). Another class is represented by methods that partition the electron density distribution, such as charges from Bader’s Atoms-In-Molecules (AIM) analysis Bader 1985 , Voronoi deformation density (VDD) charges Bickelhaupt et al. 1996 ; Fonseca Guerra et al. 2004 , as well as Hirshfeld (or stockholder) partitioning Hirshfeld 1977 and its subsequent variants Bultinck et al. 2007 ; Verstraelen et al. 2013 ; Lillestolen and Wheatley 2009 . Finally, the last category includes methods based on the fitting of a set of point charges to reproduce an observable quantity, such as the molecular electrostatic potential (ESP). However, in these methods, charges often converge to different values upon small changes in the molecular geometry. This conformational dependency can be reduced by using multiple conformations during the fitting procedure, with a noticeable improvement in the charge set transferability Reynolds et al. 1992 . Moreover, the charges can converge to significantly different values, especially for buried atoms in the molecule, and nevertheless reproduce the ESP equally well. For this reason, restraints to regularize the problem are often introduced. One common strategy to solve this problem is to restrain the charges to zero via a hyperbolic penalty function, as done in the restrained ESP (RESP) approach Bayly et al. 1993 . Still, the restraint of atomic charges to zero lacks strong physical motivation and comes more from practical considerations, to regularize the problem. The RESP approach provides an accurate description of the electrostatic properties of molecular systems, which have a direct impact on, e.g., conformational energies Wang et al. 2000 ; Cornell et al. 2002 and intermolecular interactions. Therefore, RESP charges are commonly used in the parametrization of classical force fields (FFs) Cornell et al. 1995 , such as in the AMBER FFs Duan et al. 2003 ; Ponder and Case 2003 ; Wang et al. 2004 . In practice, RESP fitting is performed on small molecules or molecular fragments in the gas phase, using the ESP on grid points around the molecule as a reference. For example, in the Merz–Singh–Kollman scheme Singh and Kollman 1984 ; Besler et al. 1990 , the grid points are located on several layers around the molecule, constructed around each atom as a union of spheres of increasing radius, i.e., from 1.4 up to 2.0 times the van der Waals radius. Empirical observations have shown that combining the Hartree–Fock or B3LYP method with the 6-31G* basis set overestimates the polarization of the system in the gas phase, which makes it often suitable for use in the condensed phase Bayly et al. 1993 ; Cornell et al. 1995 . However, there are no rigorous guarantees that ensure the transferability of gas-phase derived charge values to a large variety of complex electrostatic environments. In the context of quantum mechanical/molecular mechanical (QM/MM) molecular dynamics (MD) simulations, Laio et al. 2002 Laio et al. 2002 introduced a scheme where RESP charges are dynamically generated (D-RESP) on the fly during the simulation, using the MM atom position as probe sites for the ESP. This method can provide a set of atomic point charges able to accurately reproduce the total molecular moments, as well as the ESP generated by the QM subsystem on the MM atoms. In this work, we extended the original D-RESP approach Laio et al. 2002 to fit atom-centered electric multipole moments of arbitrary order, resulting in the extended D-RESP (xDRESP) approach. An additional benefit of the presented xDRESP method is its implementation within the MiMiC framework for multiscale modeling Olsen et al. 2019 ; Antalík et al. 2024 , which combines the capabilities of different external programs responsible for computing energy and force contributions of different subsystems. MiMiC itself only calculates the subsystem interactions and enables communication between the different external programs. The modular design of MiMiC provides access to the xDRESP approach using any combination of available QM and MM external programs in QM/MM MD simulations. In the following, we discuss the theoretical details of the xDRESP approach, as well as practical implications of its implementation within the MiMiC framework (Sec. 2 ). We then focus on different examples to validate the method and show potential applications for different chemical problems (Sec. 3 ), and we conclude by highlighting possible new methodological advantages enabled by the implementation of the xDRESP approach (Sec. 4 ). 2 Theory and Methods 2.1 Extended D-RESP approach The xDRESP atom-centered electric multipole moments can be obtained by solving a least-squares problem on the ESP generated by the QM subsystem on the MM atoms ( V^{\mathrm{QM\to MM}} ) with additional restraints on the values of the point charges ( q^{\mathrm{ref}} ), with the option of imposing additional constraints ( \sigma ). The loss function for such optimization takes the form L=\sum_{p=1}^{N^{\mathrm{SR}}}{\left(V_{p}^{\mathrm{xDRESP}}-V_{p}^{\mathrm{QM\to MM}}\right)}^{2}+w_{R}\sum_{i=1}^{N^{\mathrm{QM}}}{\left(M_{i}^{\mathrm{[0]}}-q_{i}^{\mathrm{ref}}\right)}^{2}+\sigma, (1) where N^{\mathrm{SR}} is the number