arXiv subjects

David M. Rogers

Publications and source records attributed to David M. Rogers.

At least 19 recordsLinked to original sources

arXiv2025

Variational Formulation of Local Molecular Field Theory

In this note, we show that the Local Molecular Field theory of Weeks et. al. can be re-derived as an extremum problem for an approximate Helmholtz free energy. Using the resulting free energy as a classical, fluid density functional yields an implicit solvent method identical in form to the Molecular Density Functional theory of Borgis et. al., but with an explicit formula for the 'ideal' free energy term. This new expression for the ideal free energy term can be computed from all-atom molecular dynamics of a solvent with only short-range interactions. The key hypothesis required to make the theory valid is that all smooth (and hence long-range) energy functions obey Gaussian statistics. This is essentially a random phase approximation for perturbations from a short-range only, 'reference,' fluid. This single hypothesis is enough to prove that the self-consistent LMF procedure minimizes a novel density functional whose 'ideal' free energy is the molecular system under a specific, reference Hamiltonian, as opposed to the non-interacting gas of conventional density functionals. Implementation of this new functional into existing software should be straightforward and robust.

cond-mat.soft↗

arXiv2025

Enabling Seamless Transitions from Experimental to Production HPC for Interactive Workflows

The evolving landscape of scientific computing requires seamless transitions from experimental to production HPC environments for interactive workflows. This paper presents a structured transition pathway developed at OLCF that bridges the gap between development testbeds and production systems. We address both technological and policy challenges, introducing frameworks for data streaming architectures, secure service interfaces, and adaptive resource scheduling for time-sensitive workloads and improved HPC interactivity. Our approach transforms traditional batch-oriented HPC into a more dynamic ecosystem capable of supporting modern scientific workflows that require near real-time data analysis, experimental steering, and cross-facility integration.

cs.DC↗

arXiv2024

A cast of thousands: How the IDEAS Productivity project has advanced software productivity and sustainability

Computational and data-enabled science and engineering are revolutionizing advances throughout science and society, at all scales of computing. For example, teams in the U.S. DOE Exascale Computing Project have been tackling new frontiers in modeling, simulation, and analysis by exploiting unprecedented exascale computing capabilities-building an advanced software ecosystem that supports next-generation applications and addresses disruptive changes in computer architectures. However, concerns are growing about the productivity of the developers of scientific software, its sustainability, and the trustworthiness of the results that it produces. Members of the IDEAS project serve as catalysts to address these challenges through fostering software communities, incubating and curating methodologies and resources, and disseminating knowledge to advance developer productivity and software sustainability. This paper discusses how these synergistic activities are advancing scientific discovery-mitigating technical risks by building a firmer foundation for reproducible, sustainable science at all scales of computing, from laptops to clusters to exascale and beyond.

cs.CY↗

arXiv2022

DFT-FE 1.0: A massively parallel hybrid CPU-GPU density functional theory code using finite-element discretization

We present DFT-FE 1.0, building on DFT-FE 0.6 [Comput. Phys. Commun. 246, 106853 (2020)], to conduct fast and accurate large-scale density functional theory (DFT) calculations (reaching ~ $100,000$ electrons) on both many-core CPU and hybrid CPU-GPU computing architectures. This work involves improvements in the real-space formulation -- via an improved treatment of the electrostatic interactions that substantially enhances the computational efficiency -- as well high-performance computing aspects, including the GPU acceleration of all the key compute kernels in DFT-FE. We demonstrate the accuracy by comparing the ground-state energies, ionic forces and cell stresses on a wide-range of benchmark systems against those obtained from widely used DFT codes. Further, we demonstrate the numerical efficiency of our implementation, which yields $\sim 20 \times$ CPU-GPU speed-up by using GPU acceleration on hybrid CPU-GPU nodes. Notably, owing to the parallel-scaling of the GPU implementation, we obtain wall-times of $80-140$ seconds for full ground-state calculations, with stringent accuracy, on benchmark systems containing ~ $6,000-15,000$ electrons.

physics.comp-ph↗

arXiv2021

Three Practical Workflow Schedulers for Easy Maximum Parallelism

Runtime scheduling and workflow systems are an increasingly popular algorithmic component in HPC because they allow full system utilization with relaxed synchronization requirements. There are so many special-purpose tools for task scheduling, one might wonder why more are needed. Use cases seen on the Summit supercomputer needed better integration with MPI and greater flexibility in job launch configurations. Preparation, execution, and analysis of computational chemistry simulations at the scale of tens of thousands of processors revealed three distinct workflow patterns. A separate job scheduler was implemented for each one using extremely simple and robust designs: file-based, task-list based, and bulk-synchronous. Comparing to existing methods shows unique benefits of this work, including simplicity of design, suitability for HPC centers, short startup time, and well-understood per-task overhead. All three new tools have been shown to scale to full utilization of Summit, and have been made publicly available with tests and documentation. This work presents a complete characterization of the minimum effective task granularity for efficient scheduler usage scenarios. These schedulers have the same bottlenecks, and hence similar task granularities as those reported for existing tools following comparable paradigms.

cs.DC↗

arXiv2020

Protein Conformational States: A First Principles Bayesian Method

Automated identification of protein conformational states from simulation of an ensemble of structures is a hard problem because it requires teaching a computer to recognize shapes. We adapt the naive Bayes classifier from the machine learning community for use on atom-to-atom pairwise contacts. The result is an unsupervised learning algorithm that samples a `distribution' over potential classification schemes. We apply the classifier to a series of test structures and one real protein, showing that it identifies the conformational transition with > 95% accuracy in most cases. A nontrivial feature of our adaptation is a new connection to information entropy that allows us to vary the level of structural detail without spoiling the categorization. This is confirmed by comparing results as the number of atoms and time-samples are varied over 1.5 orders of magnitude. Further, the method's derivation from Bayesian analysis on the set of inter-atomic contacts makes it easy to understand and extend to more complex cases.

physics.comp-ph↗

arXiv2019

Range separation: The divide between local structures and field theories

This work presents parallel histories of the development of two modern theories of condensed matter: the theory of electron structure in quantum mechanics, and the theory of liquid structure in statistical mechanics. Comparison shows that key revelations in both are not only remarkably similar, but even follow along a common thread of controversy that marks progress from antiquity through to the present. This theme appears as a creative tension between two competing philosophies, that of short range structure (atomistic models) on the one hand, and long range structure (continuum or density functional models) on the other. The timeline and technical content are designed to build up a set of key relations as guideposts for using density functional theories together with atomistic simulation.

cond-mat.stat-mech↗

arXiv2019

Unifying Theories for Nonequilibrium Statistical Mechanics

The question of deriving general force/flux relationships that apply out of the linear response regime is a central topic of theories for nonequilibrium statistical mechanics. This work applies an information theory perspective to compute approximate force/flux relations and compares the result with traditional alternatives. If it can be said that there is a consensus on the form of response theories in driven, nonequilibrium transient dynamics, then that consensus is consistent with maximizing the entropy of a distribution over transition space. This agreement requires the problem of force/flux relationships to be described entirely in terms of such transition distributions, rather than steady-state properties (such as near-equilibrium works) or distributions over trajectory space (such as maximum caliber). Within the transition space paradigm, it is actually simpler to work in the fully nonlinear regime without relying on any assumptions about the steady-state or long-time properties. Our results are compared to extensive numerical simulations of two very different systems. The first is a the periodic Lorentz gas under constant external force, extended with angular velocity and physically realistic inelastic scattering. The second is an $α$-Fermi-Pasta-Ulam chain, extended with a Langevin thermostat that couples only to individual harmonic modes. Although we simulate both starting from transient initial conditions, the maximum entropy structure of the transition distribution is clearly evident on both atomistic and intermediate size scales. The result encourages further development of empirical laws for nonequilibrium statistical mechanics by employing analogies with standard maximum entropy techniques -- even in cases where large deviation principles cannot be rigorously proven.

cond-mat.stat-mech↗

arXiv2019

Dual Characterization of the Ornstein-Zernike Equation in Moment Space

The molecular density functional theory of fluids provides an exact theory for computing solvation free energies in implicit solvents. One of the reasons it has not received nearly as much attention as quantum density functional theory for implicit electron densities is the paucity of basis set expansions for this theory. This work constructs a minimal Hilbert space version of the Ornstein-Zernike theory over the complete spatial, rotational, and internal conformational space that leaves the choice of basis open. The basis is minimal in the sense that it is isomorphic to a choice of molecular property space (i.e. moments of the distribution), and does not require auxiliary grids. This can be exploited, since there are usually only a few `important' properties for determining the structure and energetics of a molecular fluid. Two novel basis expansions are provided which emphasize either the multipolar expansion (most useful for flexible molecules) or the rotational distribution (most useful for rigid bodies described using quaternions). The perspective of this work shows that radial truncation of the Bessel series over translation space determines an analytical extrapolation of these functions to the origin in reciprocal space. We provide a new density functional theory that naturally fits the moment-based, matrix approach. Three diverse applications are presented: relating the present approach to traditional rotational invariants, demonstrating the stability of convex optimization on the density functional, and finding analytical expression for dispersion contributions to the solvation free energies of point polarizable dipoles.

cond-mat.stat-mech↗

arXiv2017

Fluctuation Theory of Ionic Solvation Potentials

This work presents a rigorous statistical mechanical theory of solvation free energies, specifically useful for describing the long-range nature of ions in an electrolyte solution. The theory avoids common issues with field theories by writing the excess chemical potential directly as a maximum-entropy variational problem in the space of solvent 1-particle density functions. The theory was developed to provide a simple physical picture of the relationship between the solution's spatial dielectric function, ion screening, and the chemical potential. The key idea is to view the direct correlation function of molecular Ornstein-Zernike theory as a Green's function for both longitudinal and transverse electrostatic dipole relaxation of the solvent. Molecular simulation data is used to calculate these direct correlation functions, and suggests that the most important solvation effects can be captured with only a screened random phase approximation. Using that approximation predicts both the Born solvation free energy and a Debye-Huckel law in close agreement with the mean spherical approximation result. These limiting cases establish the simplicity and generality of the theory, and serve as a guide to replacing local dielectric and Poisson-Boltzmann approximations.

cond-mat.stat-mech↗

arXiv2017

Extension of Kirkwood-Buff Theory to the Canonical Ensemble

Kirkwood-Buff (KB) integrals are notoriously difficult to converge from a canonical simulation because they require estimating the grand-canonical radial distribution. The same essential difficulty is encountered when attempting to estimate the direct correlation function of Ornstein-Zernike theory by inverting the pair correlation functions. We present a new theory that applies to the entire, finite, simulation volume, so that no cutoff issues arise at all. The theory gives the direct correlation function for closed systems, while smoothness of the direct correlation function in reciprocal space allows calculating canonical KB integrals via a well-posed extrapolation to the origin. The present analysis method represents an improvement over previous work because it makes use of the entire simulation volume and its convergence can be accelerated using known properties of the direct correlation function. Using known interaction energy functions can make this extrapolation near perfect accuracy in the low-density case. Because finite size effects are stronger in the canonical than the grand-canonical ensemble, we state ensemble correction formulas for the chemical potential and the KB coefficients. The new theory is illustrated with both analytical and simulation results on the 1D Ising model and a supercritical Lennard-Jones fluid. For the latter, the finite-size corrections are shown to be small.

cond-mat.stat-mech↗

arXiv2016

The EPR Paradox Implies A Minimum Achievable Temperature

We carefully examine the thermodynamic consequences of the repeated partial projection model for coupling a quantum system to an arbitrary series of environments under feedback control. This paper provides observational definitions of heat and work that can be realized in current laboratory setups. In contrast to other definitions, it uses only properties of the environment and the measurement outcomes, avoiding references to the `measurement' of the central system's state in any basis. These definitions are consistent with the usual laws of thermodynamics at all temperatures, while never requiring complete projective measurement of the entire system. It is shown that the back-action of measurement must be counted as work rather than heat to satisfy the second law. Comparisons are made to stochastic Schrödinger unravelling and transition-probability based methods, many of which appear as particular limits of the present model. These limits show that our total entropy production is a lower bound on traditional definitions of heat that trace out the measurement device. Examining the master equation approximation to the process at finite measurement rates, we show that most interactions with the environment make the system unable to reach absolute zero. We give an explicit formula for the minimum temperature achievable in repeatedly measured quantum systems. The phenomenon of minimum temperature offers a novel explanation of recent experiments aimed at testing fluctuation theorems in the quantum realm and places a fundamental purity limit on quantum computers.

quant-ph↗

arXiv2015

Towards a Direct, By-Need Evaluator for Dependently Typed Languages

We present a C-language implementation of the lambda-pi calculus by extending the (call-by-need) stack machine of Ariola, Chang and Felleisen to hold types, using a typeless- tagless- final interpreter strategy. It has the advantage of expressing all operations as folds over terms, including by-need evaluation, recovery of the initial syntax-tree encoding for any term, and eliminating most garbage-collection tasks. These are made possible by a disciplined approach to handling the spine of each term, along with a robust stack-based API. Type inference is not covered in this work, but also derives several advantages from the present stack transformation. Timing and maximum stack space usage results for executing benchmark problems are presented. We discuss how the design choices for this interpreter allow the language to be used as a high-level scripting language for automatic distributed parallel execution of common scientific computing workflows.

cs.PL↗

arXiv2015

Thermodynamics of Maximum Transition Entropy for Quantum Assemblies

This work presents a general unifying theoretical framework for quantum non-equilibrium systems. It is based on a re-statement of the dynamical problem as one of inferring the distribution of collision events that move a system toward thermal equilibrium from an arbitrary starting distribution. Using a form based on maximum entropy for this transition distribution leads to a statistical description of open quantum systems with strong parallels to the conventional, maximum-entropy, equilibrium thermostatics. A precise form of the second law of thermodynamics can be stated for this dynamics at every time-point in a trajectory. Numerical results are presented for low-dimensional systems interacting with cavity fields. The dynamics and stationary state are compared to a reference model of a weakly coupled oscillator plus cavity supersystem thermostatted by periodic partial measurements. Despite the absence of an explicit cavity in the present model of open quantum dynamics, both the relaxation rates and stationary state properties closely match the reference. Additionally, the time-course of energy exchange and entropy increase is given throughout an entire measurement process for a single spin system. The results show the process to be capable of initially absorbing heat when starting from a superposition state, but not from an isotropic distribution. Based on these results, it is argued that logical inference in the presence of environmental noise is sufficient to resolve the paradox of wavefunction collapse.

quant-ph↗

arXiv2014

Real-space quadrature: a convenient, efficient representation for multipole expansions

Multipolar expansions are a foundational tool for describing basis functions in quantum mechanics, many-body polarization, and other distributions on the unit sphere. Progress on these topics is often held back by complicated and competing formulas for calculating and using spherical harmonics. We present a complete representation for supersymmetric 3D tensors that replaces spherical harmonic basis functions by a dramatically simpler set of weights associated to discrete points in 3D space. This representation is shown to be space optimal. It reduces tensor contraction and the spherical harmonic decomposition of Poisson's operator to pairwise summations over the point set. Moreover, multiplication of spherical harmonic basis functions translates to a direct product in this representation.

math-ph↗

arXiv2011

Information Theory and Statistical Mechanics Revisited

The statistical mechanics of Gibbs is a juxtaposition of subjective, probabilistic ideas on the one hand and objective, mechanical ideas on the other. In this paper, we follow the path set out by Jaynes, including elements added subsequently to that original work, to explore the consequences of the purely statistical point of view. We show how standard methods in the equilibrium theory could have been derived simply from a description of the available problem information. In addition, our presentation leads to novel insights into questions associated with symmetry and non-equilibrium statistical mechanics. Two surprising consequences to be explored in further work are that (in)distinguishability factors are automatically predicted from the problem formulation and that a quantity related to the thermodynamic entropy production is found by considering information loss in non-equilibrium processes. Using the problem of ion channel thermodynamics as an example, we illustrate the idea of building up complexity by successively adding information to create progressively more complex descriptions of a physical system. Our result is that such statistical mechanical descriptions can be used to create transparent, computable, experimentally-relevant models that may be informed by more detailed atomistic simulations. We also derive a theory for the kinetic behavior of this system, identifying the nonequilibrium `process' free energy functional. The Gibbs relation for this functional is a fluctuation-dissipation theorem applicable arbitrarily far from equilibrium, that captures the effect of non-local and time-dependent behavior from transient driving forces. Based on this work, it is clear that statistical mechanics is a general tool for constructing the relationships between constraints on system information.

cond-mat.stat-mech↗

arXiv2011

A First and Second Law for Nonequilibrium Thermodynamics: Maximum Entropy Derivation of the Fluctuation-Dissipation Theorem and Entropy Production Functionals

A theory for non-equilibrium systems is derived from a maximum entropy approach similar in spirit to the equilibrium theory given by Gibbs. Requiring Hamilton's principle of stationary action to be satisfied on average during a trajectory, we add constraints on the transition probability distribution which lead to a path probability of the Onsager-Machlup form. Additional constraints derived from energy and momentum conservation laws then introduce heat exchange and external driving forces into the system, with Lagrange multipliers related to the temperature and pressure of an external thermostatic system. The result is a fully time-dependent, non-local description of a nonequilibrium ensemble. Detailed accounting of the energy exchange and the change in information entropy of the central system then provides a description of the entropy production which is not dependent on the specification or existence of a steady-state or on any definition of thermostatic variables for the central system. These results are connected to the literature by showing a method for path re-weighting, creation of arbitrary fluctuation theorems, and by providing a simple derivation of Jarzynski relations referencing a steady-state. In addition, we identify path free energy and entropy (caliber) functionals which generate a first law of nonequilibrium thermodynamics by relating changes in the driving forces to changes in path averages. Analogous to the Gibbs relations, the variations in the path averages yield fluctuation-dissipation theorems. The thermodynamic entropy production can also be stated in terms of the caliber functional, resulting in a simple proof of our microscopic form for the Clausius statement. We find that the maximum entropy route provides a clear derivation of the path free energy functional, path-integral, Langevin, Brownian, and Fokker-Planck statements of nonequilibrium processes.

cond-mat.stat-mech↗

arXiv2010

Resolution and Scale Independent Function Matching Using a String Energy Penalized Spline Prior

The extension of the classical Bayesian penalized spline method to inference on vector-valued functions is considered, with an emphasis on characterizing the suitability of the method for general application.We show that the standard quadratic penalty is exactly analogous to the energy of a stretched string, with the penalty parameter corresponding to its tension. This physical analogy motivates a discussion of resolution independence, which we define as the convergence of a computational function estimate to arbitrary accuracy with increasing resolution.The multidimensional context makes direct application of standard procedures for choosing the penalty parameter difficult, and a new method is proposed and compared to the established generalized cross-validation (GCV) and Akaike information criterion (AIC) functions.Our Bayesian method for choosing this parameter is derived by introducing a scal e independence criterion to ensure that simultaneously scaling the function samples and their variances does not significantly change the posterior parameter distribution. Due to the possibility of an exact polynomial fit, numerical issues prevent the use of this prior, and a solution is presented based on adding a st ring zero-point energy. This makes more complicated approaches recently propose d in the literature unnecessary, and eliminates the requirement for sensitivity analysis when the function deviates from the above mentioned polynomial. An important class of problems which can be analyzed by this method are stochastic numerical integrators, which are considered as an example problem. This work represents the first extension of penalized spline methods to inference on multidimensional numerical integrators reported in the literature. Several numerical calculations illustrate the above points and address practical application issues.

stat.ML↗