ARBFN: Arbitrary Externally-Computed Closed-Loop Force Fixes in LAMMPS

Spring 2025, paper in preparation for journal submission

Jordan Dehmel (CMU), Jarrod Schiffbauer (CMU)

Abstract: The molecular dynamics simulation software LAMMPS (Large-scale Atomic/Molecular Massively Parallel Simulator) provides a scripting language for the easy implementation of experiments: However, it is not all-encompassing. There are many situations in which LAMMPS alone is not sufficient and some external computation must be used, for example, including quantum effects [Kohlmeyer, Ippolito, Cavazzoni, 2014] and machine learning control [Rohskopf et. al., 2023] of a simulation. Such situations provide some custom interface for a specific program to apply forces on the simulated particles. This project outlines the development of a generic protocol for this process. Specifically, we introduce an externally-controlled arbitrary atomic forcing fix within the existing MPI (Message Passing Interface) LAMMPS framework [OpenMPI, 2003] [Thompson et. al., 2024]. This involves an arbitrary-language controller program being instantiated alongside LAMMPS in an MPI runtime, then communicating with all LAMMPS instances whenever the desired fix must be computed. The protocol communicates in JSON (JavaScript Object Notation) strings and assumes no controller linguistic properties beyond a valid MPI implementation.

Study and Investigation of behavior of Janus particles in different concentration KCl electrolytes mixed with glycerol

November 2024, Poster for the 2024 APS 77th Annual Meeting of the Division of Fluid Dynamics

Sandeep Ramteke (FIU), Jordan Dehmel (presenter, CMU), Alicia Boymelgreen (FIU), Jarrod Schiffbauer (CMU)

Vulnerability and Data Flow Analysis on Arbitrary Languages with Code Property Graphs and Taint Traversals

Summer 2024, paper in-progress

Jordan Dehmel (CMU) and Warren MacEvoy (CMU)

Abstract: When assessing the security of code, it is paramount to ensure that security-privileged routines do not disclose secret information to would-be attackers. Similarly, when interacting with important and abstract APIs like databases, it must be ensured that no unverified user input be allowed direct access (EG SQL injection attacks). These problems both involve tracking the contributors and dependents of variables at debugging time, a common task in analysis and linting. This paper examines the use of Code Property Graphs (CPGs) [Yamaguchi, Golde, Arp and Rieck, 2014] and taint-style vulnerability scanning [Sabelfeld and Myers, 2003] in such a context. CPGs are dense and expressive representations of program syntax, dependency, and control flow often used in linting. Taint-style vulnerability scanning, on the other hand, tracks possible vulnerabilities along variable modifications and control flow alterations such that sensitive data does not flow beyond its desired scope. This paper describes a language-agnostic algorithm for the construction of CPGs with embedded information-flow security measures and details the development of a VSCode extension utilizing them.

Expanding Active Matter to the Third Dimension: Exploring Short and Long-Range Particle-Wall Interactions.

Fall 2024, paper, in review at PRL.

Sandeep Ramteke, Jordan Dehmel, Touvia Miloh, Jarrod Schiffbauer, Alicia Boymelgreen

Resulted in the launching of extraterrestrial ISS experiments on the effects of microgravity on colloidal active matter in August 2024, as well as novel experiments involving density matching on Earth for the avoidance of gravity-related error.

Colloidal Active Matter Under Applied Electric Fields

Fall 2024, Poster for the 2024 RMACC HPC Symposium Student Poster Competition.

Jordan Dehmel (CMU), Sandeep Ramteke (FIU), Jarrod Schiffbauer (CMU), Alicia Boymelgreen (FIU)

Individual and Collective Motion in Colloidal Active Matter Under Applied Electric Fields: Experimental and Computational Perspectives

Fall 2024, Presentation for the 2024 Colorado Mesa University Student Showcase