A new computational system featuring GPU-accelerated molecular dynamics simulation and 3D/VR visualization, analysis and manipulation of nanostructures has been implemented, representing a novel approach to advance materials research and promote innovative investigation and alternative methods to learn about material structures with dimensions invisible to the human eye.
The increasing development of computing (hardware and software) in the last decades has impacted scientific research in many fields including materials science, biology, chemistry and physics among many others. A new computational system for the accurate and fast simulation and 3D/VR visualization of nanostructures is presented here, using the open-source molecular dynamics (MD) computer program LAMMPS. This alternative computational method uses modern graphics processors, NVIDIA CUDA technology and specialized scientific codes to overcome processing speed barriers common to traditional computing methods. In conjunction with a virtual reality system used to model materials, this enhancement allows the addition of accelerated MD simulation capability. The motivation is to provide a novel research environment which simultaneously allows visualization, simulation, modeling and analysis. The research goal is to investigate the structure and properties of inorganic nanostructures (e.g., silica glass nanosprings) under different conditions using this innovative computational system. The work presented outlines a description of the 3D/VR Visualization System and basic components, an overview of important considerations such as the physical environment, details on the setup and use of the novel system, a general procedure for the accelerated MD enhancement, technical information, and relevant remarks. The impact of this work is the creation of a unique computational system combining nanoscale materials simulation, visualization and interactivity in a virtual environment, which is both a research and teaching instrument at UC Merced.
Materials science is an interdisciplinary field that examines the structure-property relationships in matter for their application to many areas of science and engineering. As structure-property relationships are investigated through computer simulations in addition to experimentation, computational tools offer complementary features that can enhance research efforts. While nanomaterials are of interest to scientists and have redeeming value for their potential social impact, this size regime is fraught with many challenges found particularly in experimentation.
Computer simulations allow scientists and engineers to perform specialized tests in a large variety of environments limited only by time and computational resources. Molecular dynamics (MD) simulations allow the appropriate time and length scales to study the phenomena of interest in many nanomaterials. Simulations expand the study of materials by removing the constraints of the physical laboratory, however many computational tools lack accessible, intuitive interfaces for research. Enhancement with the graphical display of models, efficient computational algorithms, and graphical processing unit (GPU) based computing complement current simulation efforts. These new graphics devices combine with central processing units efficiently to allow mathematically intensive calculations to be accomplished by the GPU. The result is an effective acceleration of computation on the order of 10x accompanied by a reduction in power consumption of up to 20x.
The goal of this research project was to develop and implement a novel tool for nanoscience investigation that directly connects an interactive interface to MD simulations, materials science analysis and 3D visualization. This innovative system with unique and powerful analysis capabilities has been used for nanoscale research and education at UC Merced, with direct implications to other related STEM fields such as nanotechnology, physics, biology, and geology, and ultimate benefit to education and society.
The 3D/VR Visualization System was implemented as both a research and teaching instrument which allows creation and manipulation of atomic structures in an interactive 3D virtual reality (VR) environment. The system was created from a set of relatively low-cost and accessible components following the model originally developed by Dr. Oliver Kreylos at UC Davis1.
Below is a photo of the final 3D/VR Visualization System layout, with important components labeled (Figure 1). This system was originally established for education purposes at UC Merced in 2009. The implementation of the original 3D/VR system resulted in peer-reviewed publications2-3. Table 1 below summarizes key characteristics for each element of the 3D/VR Visualization System.
Figure 1. 3D/VR Visualization System and main components (left) in the Davila Research Laboratory at UCM and visualization devices (right). Please click here to view a larger version of this figure.
Item | Component | Functionality in System |
A | 3D TV | 3D display of modeled molecular structures and on-screen menus. |
B | Infrared (IR) tracking cameras4 | IR cameras track positions of the Wiimote and 3D viewing goggles in the user workspace in front of 3D TV, allowing virtual 3D manipulation of displayed structures. |
C | Tracking PC | Runs IR camera tracking software and transmits Wiimote and 3D goggle positions to modeling computer. |
D | Wiimote | Used for on-screen management of modeling software and to manipulate structures in 3D virtual environment. |
E | 3D goggles5 | Synchronized with 3D TV IR signal, allow 3D view of structure. Position tracked by IR cameras for accurate 3D view. |
F | Modeling PC | Runs NCK/VRUI 3D modeling and display software6, accepts goggle / Wiimote position and control signals to create accurate 3D molecular structure view. |
Table 1. Functionality of main elements of the 3D/VR Visualization System at UCM.
Description of 3D/VR Visualization System and Basic Components:
3D/VR Visualization System Overview — The 3D/VR Visualization System consists of a set of IR cameras and tracking software operating in conjunction with 3D modeling software to allow a user to interactively create 3D molecular structures. The IR cameras and software track the 3D location of a Wiimote and 3D viewing goggles using IR markers, and pass this to the modeling software. The modeling software uses the Wiimote control signals and movement to generate 3D molecular structures viewable using the combination of a 3D-capable large format television with synchronized and tracked 3D goggles. This results in a 3D virtual reality workspace within which the user can dynamically create and manipulate virtual molecular structures which reflect real-world physical behavior based on inter-atomic forces used in the modeling software (Figure 2). Special considerations for setting up this system can be found in supplemental materials.
Figure 2. Investigating silica nanomaterials using the 3D/VR Visualization System. (a) A researcher creates an initial cristobalite model (crystalline) before GPU-based simulations. (b) Upon performing a simulated MD melt-quench procedure on model shown in (a), another researcher obtains a silica glass model (non-crystalline). Please click here to view a larger version of this figure.
3D/VR Visualization System Enhancement — MD Simulation Capability:
Molecular dynamics simulation systems are commonly implemented in a multi-nodal fashion, that is, a large workload is distributed or parallelized among tens to thousands of processors. Recently, additional opportunities for accelerated scientific computing have arisen out of developments in computer graphics processing. These advances include a software interface allowing scientists to take advantage of the highly parallel nature of the processing power intrinsic to graphics chips. With the advent of the Compute Unified Device Architecture or CUDA7, scientists can use GPUs8 to enhance the speed at which problems are solved while reducing the cost of infrastructure. A typical GPU may have the equivalent of hundreds to thousands of cores or “nodes” for processing information, and as these can each be used in parallel, a well-coded solution may provide up to 1,000x throughput acceleration against its multi-core counterpart. Though not every problem is well-suited to this approach, current MD simulations have seen up to 15x throughput performance gains9. Details on the 3D/VR visualization system MD-GPU enhancement can be found in supplemental materials.
1. Install 3D/VR Modeling Software on Modeling PC
2. Set Up Tracking System1
3. Prepare 3D Modeling System for Use
4. Test 3D/VR Visualization System Using NCK Software
The following set of instructions outlines how to use the NCK software on-screen menus to establish controller tool functions, and then how to build and manipulate a carbon nanotube in the 3D/VR workspace from constituent carbon atoms (Figure 4). Instructions on how to measure the resulting bond angles and distances (Step 4.4.10) are available online10.
Figure 4. Undergraduate student using the 3D/VR Visualization System to study carbon nanotubes (CNTs). Photos (A)-(F) show the building process of a single-walled CNT. Please click here to view a larger version of this figure.
5. Visualization of Molecular Dynamics Simulation Models
Animated Figure 1. Animation of helical nanostructure tensile simulation.
This 3D/VR Visualization System presents new opportunities for conducting materials science studies. As this immersive environment operates in real time, in the form of 3D input and display, the researcher is presented with a fully interactive nanoscaled instrument2. By following the protocol presented here, a silica helical nanoribbon was created in this step-by-step fashion. A snapshot of this structure produced from LAMMPS MD is shown in Figure 7. This structure was subjected to simulated tensile testing, and the results of this simulation are shown in Animated Figure 1 which illustrates the reorganization and failure of the structure under tensile forces.
By combining the real-time interactivity and visual nature of an immersive environment with powerful MD simulations15, researchers can benefit from intuitive control and full-featured analysis.
The enhanced 3D/VR Visualization System with MD capability was thoroughly tested and implemented in nanoscience research in the Davila Lab at UC Merced, focusing on tensile simulations of amorphous silica nanowires, nanoribbons and nanosprings15.
Critical elements in the successful installation and usage of the 3D/VR Visualization System are detailed in the Physical Environment and Design Considerations and Special Considerations in supplemental materials. Important installation considerations include 3D display height for comfortable long-term standing or seated usage, maximized tracking camera mounted height to create a large 3D working area, stable tracking camera and 3D display support to maintain configuration over time, and removal of IR-reflective elements from the 3D working area. As mentioned in the installation instructions, if the available tracking camera mounting height is constrained, alternate camera orientation may be necessary to create the largest 3D working area.
During configuration of the tracking software, the wand capture step is important for final tracking accuracy. Care should be taken to move the reflective wand thoroughly and smoothly throughout the tracking camera overlap area without blocking any camera or introducing any secondary reflective object, repeating this step as necessary until the required error values are achieved. As noted in the above-mentioned sections, during system usage it can be important to create a small shield on the 3D goggles to prevent interference from the tracking IR signal with the 3D synchronizing IR signal, and to use fresh 3D goggle batteries to maximize the goggle 3D synchronization. Additionally, consistent care should be taken to not touch or alter the 3D goggle and Wiimote IR-reflective spheres, and to not physically shift the tracking camera or 3D display positions in order to maintain accurate 3D tracking and imaging.
Other previous efforts have focused on MD and real-time interactivity (e.g., via VMD, a popular molecular visualization and modeling software for biomolecular systems21) while newer approaches have implemented other user interfaces and 3D gesture and voice controls22. Another group23 has created software which integrates adaptive, incremental algorithms to update the potential energy and interatomic forces within nanosystems. The system described in this work includes a particular target as it consists of 3D visualization of nanomaterials via the open-source NCK software6, with interactivity in a virtual reality environment and MD simulations capability via the LAMMPS open-source code12. This code allows flexibility since various robust interatomic potentials are available to study nanomaterials, for materials science research. Thus, the system in this work includes similar elements of MD simulation and interactivity as some other approaches, but with a focus on nanoscale materials research.
The significance of the 3D/VR Visualization System described here is that it is simpler and lower-cost to set up, and more flexible to use for the average researcher or educator, than more expensive specialized immersive environments. The addition of GPU-accelerated MD simulation capability takes advantage of this rapidly evolving computing technology to create an energy and space conserving, high-performance computing environment within the laboratory. This novel immersive tool coupled with advanced analysis capabilities is powerful and efficient for use in fields such as materials science, and is uniquely suited for nanoscale research and education. This system was selected to be showcased in the June 2012 series “Our Digital Life”24 on UCTV (a public-serving media outlet and the first university-run YouTube original channel).
As both a research and educational tool, the 3D/VR Visualization System with accelerated MD capability promotes interdisciplinary collaboration and the integration of research and new learning approaches, including coach-style teaching, active learning, and multiple learning styles, including the use of interactive manuals developed for the system3. The implementation of the 3D/VR Visualization System has resulted in peer-reviewed publications, several conference presentations, a Master’s thesis, a NSF award, and interdisciplinary collaborations.
Potential future development and expansion of the described 3D/VR Visualization System could include the addition of menu-driven tools within the NCK 3D interface to facilitate direct interaction with the MD program (LAMMPS), while remaining in the virtual reality environment.
The authors have nothing to disclose.
We wish to gratefully acknowledge the original inspiration and extensive support provided to us toward the creation of this system from Dr. Oliver Kreylos of the UC Davis Institute for Data Analysis and Visualization. His advice and assistance were instrumental to our success.
We also wish to thank the NSF BRIGE program for providing funding for this project. This material is based upon work supported by the National Science Foundation under Grant No. 1032653.
Samsung 61" 3D-capable high definition DLP TV | Samsung | http://www.samsung.com/us/video/tvs | See Protocol Section 3 (Step 3.2) (Large format 3D-capable TV) |
Alienware Area51 750i modeling computer | Alienware | http://www.alienware.com | See Protocol Section 1 (Step 1.1) (Modeling computer) |
HP EliteBook 8530w tracking computer | HP | http://www.hp.com | See Protocol Section 2 (Step 2.3) (Tracking computer) |
V100:R2 IR tracking cameras (3) | Naturalpoint | http://www.naturalpoint.com/optitrack/products/v100-r2/ | See Protocol Section 2 (Step 2.1) and Reference [4] (Tracking cameras) |
OptiTrack Tracking Tools IR tracking software | Naturalpoint | http://www.naturalpoint.com/optitrack/software/ | See Protocol Section 2 (Step 2.3) and Reference [4] (Tracking software) |
3D Goggles and 3D TV IR sync emitter | Ilixco | http://www.i-glassesstore.com/dlp3d-wireless-2set.html | See Protocol Section 3 (Step 3.2) and Reference [5] (3D goggles) |
Wiimote 3D controller | Nintendo | http://www.nintendo.com/wii | See Protocol Section 3 (Step 3.2) (Wiimote) |
VRUI, NCK and associated 3D/VR modeling software | Open source software | http://idav.ucdavis.edu/~okreylos/ResDev/NanoTech/index.html | See Protocol Section 1 (Step 1.3) and References [1,6] (VRUI, NCK) |
LAMMPS molecular dynamics software | Open source software | http://lammps.sandia.gov/ | See Protocol Section 5 (Step 5.2) and Reference [12] (LAMMPS) |
NanospringCarver program code and files | UC Merced – open source | http://tinyurl.com/qame8dj | See Protocol Section 5 (Step 5.4) and References [16-17] (NanospringCarver) |
MATLAB GUI files | UC Merced – open source | http://tinyurl.com/qame8dj | See Protocol Section 5 (Step 5.4) and References [16-17] (NanospringCarver) |
Atomistic bulk glass input file | UC Merced – open source | http://tinyurl.com/qame8dj | See Protocol Section 5 (Step 5.4) and References [16-17] (NanospringCarver) |