2023-2024

Fritz Haber Research Center of Molecular Dynamics,
Institute of Chemistry, Edmond J. Safra Campus,
The Hebrew University of Jerusalem, 9190401 Israel

Statistical mechanics tells us that atomistic systems, such as molecules, nanocrystals, and periodic solids, simplify as they grow in size and when their temperature rises. So, our calculations should also be eased by these limits. However, traditional methods to study the structure, equations of state, and physical properties of materials within these limits are becoming increasingly complex and expensive. In this talk, we will review stochastic vector computational methods developed in collaboration between my group and that of Daniel Neuhauser (UCLA) and Eran Rabani (UC Berkeley) to study the electronic structure of extended condensed matter systems. These techniques help reduce algorithmic complexity, facilitate efficient parallelization, simplify computational tasks, accelerate calculations, and diminish memory requirements [1]. We will focus on finite-temperature stochatic-density functional theory (DFT) [2-4] as well as stochastic many-body perturbation theory [5-7].

Bibliography:
[1] R. Baer, D. Neuhauser, and E. Rabani, Stochastic Vector Techniques in Ground-State Electronic Structure, Annu. Rev. Phys. Chem. 73, 12.1 (2022).
[2] R. Baer, D. Neuhauser, and E. Rabani, Self-Averaging Stochastic Kohn-Sham Density-Functional Theory, Phys. Rev. Lett. 111, 106402 (2013).
[3] E. Arnon, E. Rabani, D. Neuhauser, and R. Baer, Equilibrium Configurations of Large Nanostructures Using the Embedded Saturated-Fragments Stochastic Density Functional Theory, J. Chem. Phys. 146, 224111 (2017).
[4] Y. Cytter, E. Rabani, D. Neuhauser, and R. Baer, Stochastic Density Functional Theory at Finite Temperatures, Phys. Rev. B 97, 115207 (2018).
[5] D. Neuhauser, Y. Gao, C. Arntsen, C. Karshenas, E. Rabani, and R. Baer, Breaking the Theoretical Scaling Limit for Predicting Quasiparticle Energies: The Stochastic G W Approach, Phys. Rev. Lett. 113, 076402 (2014).
[6] D. Neuhauser, R. Baer, and D. Zgid, Stochastic Self-Consistent Second-Order Green`s Function Method for Correlation Energies of Large Electronic Systems, J. Chem. Theory Comput. 13, 5396 (2017).
[7] W. Dou, J. Lee, J. Zhu, L. Mejía, D. R. Reichman, R. Baer, and E. Rabani, Time-Dependent Second-Order Green’s Function Theory for Neutral Excitations, J. Chem. Theory Comput. 18, 5221 (2022).

Chicago Center for Theoretical Chemistry, Department of Chemistry, James Franck Institute, and Institute for Biophysical Dynamics, The University of Chicago, Chicago, IL, USA

Advances in theoretical and computational methodology will be presented that are designed to simulate complex (biomolecular and other soft matter) systems across multiple length and time scales. This bottom-up approach provides a systematic connection between all-atom (AA) molecular dynamics, coarse-grained (CG) modeling, and mesoscopic phenomena. At the heart of these concepts are methods for deriving CG models from molecular structures and their underlying atomic-scale interactions. An important component of our work in the past few years has been the concept of the “ultra-coarse-grained” (UCG) model and its associated computational implementation. In the UCG approach, the CG sites or “beads” can have internal states, much like quantum mechanical states, so the UCG model involves a conceptual abstraction beyond simply Newtonian or Langevin dynamics for the CG beads. These internal states help to self-consistently quantify a more complicated set of possible interactions within and between the CG sites, while still maintaining a high degree of coarse-graining in the modeling. The presence of the CG site internal states also greatly expands the possible range of systems amenable to accurate CG modeling, including quite heterogeneous systems such as aggregation of hydrophobes in solution, liquid-vapor and liquid-solid interfaces, and complex self-assembly processes such as occur for large multi-protein complexes. The development of bottom-up CG models from the underlying atomistic interactions also addresses special challenges in terms of the treatment of solvation, multi-body correlations, representability, transferability, and the missing entropy in CG models. Recent breakthroughs in addressing these issues – in particular by employing developments in machine learning – will be a focus of my talk. As time allows, one “pay-off” application from our multi-year effort will focus on processes in HIV-1 virus replication, and especially on the assembly of the HIV-1 virus capsid from over one thousand proteins – a phenomenon involving a billion atoms or more over long timescales that cannot be approached through AA MD simulation.

Department of Chemistry, Duke University, Durham, NC

Electron transfer reactions produce the fuel that drives living systems, and biological electron transport occurs on nanometer to macroscopic length scales. In the first part of my talk, I will describe the development and status of biological electron-transfer theories. Early misperceptions about electron transfer mechanisms motivated the development of modern electron-transfer theories, and theories of vibronically-coupled electron tunneling resolved misconceptions surrounding photosynthetic mechanisms. These theories continue to frame thinking about tunneling in living matter, and our theories of superexchange mediated tunneling in proteins, flickering resonance transport in DNA, and direct cofactor-to-cofactor hopping in bacterial nanowires have added molecular-scale detail and richness. Atomistic theories allow us to understand the function of complex biological systems, and the theories are also predictive: they empower experimental design. In the second part of my talk, I will describe recent progress toward understanding the mechanism of electron bifurcation reactions – correlated double-electron transfer reactions that lie at the heart of photosynthesis, respiration, and biocatalysis. Electron bifurcation separates a pair of electrons that is initially co-localized on one cofactor, typically a quinone or flavin. Energy conserving electron bifurcation delivers one electron to a low energy pathway and one electron to a high energy pathway, while dissipating little free energy. Remarkably, the high energy electrons do not short circuit into the low energy pathway; the physical source of protection against short circuiting was a puzzle for over 50 years. We recently developed a theory for electron bifurcation reactions that explains their high efficiency. We find that a privileged free-energy landscape shuts down electron short-circuiting, making electron flow on this landscape highly efficient. Our studies seem to resolve the short-circuiting mystery and point to strategies for the design of bioinspired structures to deliver strongly reducing electrons on demand.

Department of Applied Physics, Stanford University, Stanford, California.

Mutations can alter the short-term fitness of an organism, as well as the rates and benefits of future mutations. While numerous examples of these evolvability modifiers have been observed in rapidly adapting microbial populations, existing theory struggles to predict when they will be favored by natural selection. In this talk, I will discuss how “traveling wave” approaches from statistical physics can help shed light on this problem. I will first review existing theory for predicting how the fates of mutations emerge from the stochastic competition between large numbers of linked genetic variants. I will then show how these approaches can be generalized to incorporate mutations that alter the evolutionary process itself, by changing the rates and benefits of future mutations. We derive analytical expressions for the fixation probabilities of these variants, and how they vary as a function of the population size and the diversity of competing mutations. We find that competition between linked mutations can dramatically enhance selection for modifiers that increase the benefits of future mutations, even when they impose a strong direct cost on fitness. However, this evolutionary foresight is surprisingly asymmetric: large populations can still greedily select for mutations that lower their overall rate of adaptation, even while they are better able to endure short-term fitness costs to realize long-term evolutionary gains. These results suggest that subtle differences in evolvability could play an important role in shaping the success of genetic variants in rapidly evolving microbial populations like cancer or SARS-CoV-2, which harbor large amounts of genetic diversity.

Department of Chemistry, Stony Brook University, Stony Brook, New York.

Many important problems in chemistry and materials science involve nonadiabatic dynamics on large numbers of electronic states. Phenomena important for strong-field physics, energy conversion, hot carrier cooling, and relaxation of plasmonic excitations fall into this category. Some of these phenomena involve long lived coherences, which are challenging to accurately model with many mixed quantum-classical methods. We will present recent theoretical developments towards an accurate and broadly applicable simulation method for modeling dynamics in this regime. Specifically, we will present the development of the Ehrenfest with collapse to a block (TAB) method and a derivative designed for dense manifolds of states (DMS). The primary achievement of TAB-DMS is that it is able to accurately describe decoherence effects without requiring explicit computation of individual electronic eigenstates. Coupling to graphics processing unit accelerated time-dependent configuration interaction software to TAB and TAB-DMS enables ab initio nonadiabatic molecular dynamics simulations on many electronic states, in full nuclear dimensionality, and without prior knowledge of reaction mechanism. The utility of this approach will be demonstrated by application to long-lived electronic coherences observed in recent ultrafast experiments.

Department of Chemistry, University of Minnesota, Twin Cities.

The process through which metastable liquid transforms into crystalline solids has significance in almost every field from formation of rocks on the earth to the formation of bones in the human body. This process often occurs via nucleation – which entails the formation of the solid embryo of a critical size (referred to as critical nucleus) often associated with a high free energy barrier. Understand the process of nucleation and its role in affecting the liquid-to-solid transition is important to modulate crystallization – both in terms of rates and the desired polymorphs. Capturing the nucleation process is rather difficult in wet-lab experiments. In molecular simulations, while the molecular scales are aptly suited for nucleation, the associated free energy barrier makes sampling nucleation a challenge. Therefore, advanced sampling techniques are often used to study nucleation. In our work, we focus on developing and applying path sampling techniques that enable sampling of processes with large free energy barriers in simulations without applying an external energy bias. In my talk, I will provide a pedagogical introduction to the path sampling techniques and discuss the challenges in applying these methods to systems with complex energy landscapes. In the second half of the talk, I will discuss the application of path sampling techniques to study nucleation of gas hydrates and Lennard Jones particles. I will also briefly talk about our work on heterogeneous ice nucleation and methods developed to enable high throughput studies of heterogenous nucleation in simulations.

Department of Chemistry, University of Colorado, Boulder.

Obtaining a solution to the many-electron Schrödinger equation stands as one of the grand challenges in chemistry and physics. Highly accurate solutions are achievable for materials containing elements from the first two rows of the periodic table, enabling us to predict the energies and properties, sometimes even achieving accuracy on par with experimental results. However, the predictive power diminishes significantly when attempting to study the properties of transition metal-containing clusters or transition states. This difficulty makes it difficult to understand and predict a wide array of fascinating phenomena, including magnetism, catalysis, and spin-forbidden reactions, among others. In this presentation, I will describe a promising new method known as Auxiliary Field Quantum Monte Carlo (AFMQC), which has the potential to address many of the challenges that conventional quantum chemistry methods face. Nevertheless, this method encounters three significant challenges: (a) it is relatively expensive, (b) obtaining properties beyond energy values can be challenging, and (c) systematically improving its accuracy can also be costly. It’s worth noting that some of these challenges are not unique to AFQMC but afflict many other quantum Monte Carlo methods. During this talk, I will introduce novel techniques that have allowed us to overcome these three challenges. I will demonstrate that with the help of these developments, we have been able to tackle chemical problems that were previously inaccessible or exceptionally demanding. I will argue that AFQMC can be used in a plug-and-play manner, even by experimentalists.

Department of Chemistry, University of Michigan, Ann Arbor.

A variety of important dynamical processes that take place in biologically and technologically important complex molecular systems involve an intricate network of interrelated photo-induced electronic energy and charge transfer pathways. The simulation of the inherently quantum-mechanical electronic dynamics underlying these pathways poses multiple formidable challenges associated with simulating the intrinsically quantum dynamics of a system that consists of a large number of coupled electronic and nuclear degrees of freedom (DOF). One strategy for tackling this challenge is to treat the electronic DOF of interest as an open quantum system and restrict the input regarding the remaining electronic and nuclear DOF (the so-called “bath”) to the minimum necessary in order to account for their effect on the electronic DOF of interest. Quantum master equations (QMEs) provide a flexible general-purpose framework for formulating the effect of the bath DOF on the dynamics of the electronic DOF of interest in terms of temporally and dimensionally compact reduced quantities known as memory kernels, whose matrix elements are associated with electronic energy/charge transfer and decoherence rates. A wide variety of different types of QMEs is available, ranging from the formally exact Nakajima–Zwanzig generalized quantum master equation (GQME) to Redfield/Lindblad QMEs, which are based on assuming weak-coupling between the system of interest and bath DOF (within second-order perturbation theory), and often also Markovianity and the secular approximation. In this talk, I will overview QME-based computational approaches that were explored by my group and their application to simulating photo-induced electronic energy and charge transfer dynamics in complex molecular systems.

Department of Chemistry, University of California, Berkeley

Most of our intuition for how chemical reactions proceed comes from classical mechanics in equilibrium settings. The Arrhenius rate law and the transition state theory that underpins it, conceive of a reaction as the motion across a barrier mediated by thermal environmental fluctuations. Increasingly, systems of current interest violate the equilibrium assumptions built into these theories, either because molecules are manipulated directly or because they evolve in environments that are constantly dissipating energy. In this seminar, I will discuss some recent ideas to extend reaction rate theories away from equilibrium using principles from stochastic thermodynamics so that they can be applied to living and driven systems. Trajectory reweighting principles and a thermodynamic speed limit provide formal results that can be understood with some simple, exactly solvable models. Variational path sampling and nonequilibrium instanton methodologies allow for these formal results to be brought to bear to complex systems. Throughout this talk, I will highlight results related to force spectroscopy of proteins and ion pair dissociation in electrolyte solutions, as well as active matter and driven self-assembly.

Department of Chemistry, Virginia Tech

Quantum mechanical simulations of molecules provide a sub-atomic understanding of chemical reactions and properties. Due to the exponential scaling of the simulation of many-body systems, these simulations are obtained only as approximations to the electronic Schrodinger equation. Although, in principle one can increase the sophistication of an approximation arbitrarily until a desired accuracy is reached, more sophisticated calculations rapidly become intractable for even the largest supercomputers. Quantum computers provide a promising route to bypass these limitations in molecular simulation. In this talk, I will discuss some of the basic ideas underlying the use of quantum computation for molecular simulation and describe some of the recent work from our group.

Department of Molecular Biology, Princeton University

An emerging mechanism for intracellular organization is biomolecular phase separation into condensates of proteins and nucleic acids. While condensate size can be key to proper biological function – in particular, compartmentalized biochemistry – the processes that govern condensate sizes inside cells remain unclear. I will discuss two examples of condensate size control: (1) an experimental model system, the algal pyrenoid, and (2) a theoretical proposal for self-organizing enzyme clusters. The talk will emphasize general biophysical principles underpinning condensate size control.

Department of Chemistry, University of Pittsburgh

The weighted ensemble path sampling strategy has helped transform what is feasible for molecular dynamics and related simulations. Highlights include atomistic simulations of rare events on the milliseconds to seconds timescale such as protein-ligand (un)binding, drug permeation of membranes, and the large-scale opening of the coronavirus spike protein. As demonstrated by benchmark studies, weighted ensemble simulations can be orders of magnitude more efficient than conventional simulations in generating rates and pathways for rare events. Importantly, the pathways are generated without the use of external biasing forces or modifications of the energy landscape. I will begin my talk with a pedagogical introduction to the weighted ensemble strategy and challenges in applying this strategy to complex biological processes. In the second part of my talk, I will present our recent advances in weighted ensemble methods and ambitious applications enabled by these advances.

Flatiron Institute, Simons Foundation

For many types of biological experiments, we can describe the underlying biophysical process using a forward model (i.e. a simulator) that recapitulates the essential physics and errors in the observations. However, comparing observations to the simulations, and inferring parameter distributions given an observation, is challenging and time consuming.

In the first seminar, I will give a background and historical perspective on Bayesian inference, its advantages, and limitations. One challenge is estimating the posterior probability for phenomena that have intractable likelihoods or likelihoods that are computationally expensive. Then, I will describe recent methods that use simulations and neural-posterior estimates to bypass the explicit likelihood calculation, referred to as simulation-based inference (SBI) techniques. I will demonstrate the potential of SBI for biophysics, first focusing on a simple example for studying force spectroscopy experiments (smFS). In smFS, the coupling of the molecule with the ever-present experimental device introduces artifacts that complicates the interpretation of these experiments. Performing statistical inference to learn hidden molecular properties is challenging because these measurements produce non-Markovian time-series, and even minimal models lead to intractable likelihoods. SBI enabled us to overcome these challenges by directly estimating the Bayesian posterior and extracting reduced quantitative models encoding a mechanistic model into a simulator in combination with probabilistic deep learning. For synthetic data, we could systematically disentangle the measurement of hidden molecular properties from experimental artifacts.

In the second seminar, I will apply SBI for inferring molecular conformations and their uncertainties from single-particle cryo-electron microscopy (cryo-EM) images. Given an observed image, SBI enables us to directly estimate the Bayesian posterior using forward model simulations, an embedding network, and a neural posterior estimation framework. The cryo-EM SBI training happens only once with the simulation of synthetic data, after which inference for each experimental particle takes only milliseconds to evaluate. This brings the great advantage that the posterior is amortized: the particle poses and imaging parameters do not have to be estimated, resulting in a high computational speed up in comparison to explicit likelihood methods. For both synthetic and real data, we could systematically disentangle the molecular conformation from the noisy observation with a confidence interval for the inference and learn about the most relevant features of the observed particles. We foresee that SBI will be widely used for studying biophysical experiments with intractable, or time consuming, likelihood calculations.