Joshua A. Anderson, Ph.D.

View Joshua A. Anderson's profile on LinkedIn Research papers by Joshua A. Anderson

GPU Computing

BVH trees for neighbor lists

We implement a bounding volume hierarchy (BVH) acceleration structure for efficient neighbor searching in molecular dynamics simulations. The BVH structure allows fast lookup of neighbors even when particle size disparity is very large, enabling research that was previously inconceivable due to slow cell list performance.

AABB tree

Strong scaling MD on GPUs

Strong scaling of a monolayer

We impelment multi-GPU scaling in HOOMD-blue, available open source starting in v1.0. Strong scaling is enabled with GPU optimized communication routines, including optional use of GPUDirect RDMA. We demonstrate scaling of a 108 million particle system to more than 3000 nodes on Titan.

Massively Parallel Monte Carlo

We develop a massively parallel method to perform Monte Carlo simulations of off-lattice particles. Our GPU implementation is 18 times faster on an NVIDIA K20 GPU than on an 8-core Intel CPU. In our initial work, we apply our method to large scale simulations of hard disks in 2D and confirm the existence of the hexatic phase.

Massively parallel Monte Carlo

GPU Accelerated Rigid Body MD

GPU Accelerated Rigid Body MD

We expand HOOMD-blue with the capability to group particle together into rigid bodies. All steps of the computation are implemented on the GPU, offering significant speed ups over serial CPU implementations. Performance on a single GPU is significantly faster than possible even with a parallel CPU domain decomposition scheme.

GPU Accelerated DPD

We expand HOOMD-blue with Dissipative Particle Dynamics (DPD) thermostats. Hash-based random number generators provide excellent performance for this algorithm on the GPU.

GPU Accelerated Dissipative Particle Dynamics

GPU Accelerated MD

GPU Accelerated MD

We develop HOOMD-blue, a general purpose particle simulation tool written from the ground up for maximum performance on massively parallel GPUs.

Hard particle self assembly

Shape allophiles

Shape allophiles

We investigate a class of "shape allophiles" that fit together like puzzle pieces as a method to access and stabilize desired structures by controlling directional entropic forces. We examine the assembly characteristics of this system via the potential of mean force and torque, and the fraction of particles that entropically bind.

Melting in Two Dimensions

Hard Disk Hexatic Phase

We apply Massively Parallel Monte Carlo to systems of hard disks. Combined with results from Event Chain Monte Carlo and Event Driven Molecular Dynamics, we confirm the existence of the hexatic phase and the first order fluid to hexatic phase transition for hard disks.

Optimal filling of shapes

Optimization Problems

Shape filling

Optimal filling of shapes

We propose filling as a type of placement problem similar to covering and packing. Filling is the optimal placement of N overlapping objects entirely inside an arbitrary shape so as to cover the most interior volume. We attack the problem with a Genetic Algorithm and develop heuristics to efficiently find solutions in polygons.

Polymer Self-assembly

Nanoparticle Telechelics


We use a polymer tether to geometrically constrain a pair of nanoparticles into a nanoparticle telechelic. Our simulation results show how architectural features control the self-assembled morphologies. HOOMD-blue powers these simulations on NVIDIA GPUs.

Polymer Nanocomposites

We perform a systematic investigation of polymer functionalization for design of composites combining nanosize crystallites with multiblock polymers in solution. Functionalization is an example of active self-assembly, where the resulting polymer nanocomposite exhibits a different type of order than the original pure polymer system without inorganic components. HOOMD-blue powers these simulations on NVIDIA GPUs.


Pluronic Micelles


We study cubic phases of Pluronic F127 in solution using coarse-grained Molecular Dynamics.