Activities: Research, Education, Training, and Industrial Application
Goals: De Novo Design of Industrial Catalysts, Drugs, Nanoscale Materials, and Processes
Using a Hierarchy of Quantum, Atomistic, Mesoscale, and Continuum Simulations
There has long been the dream that theory (quantum mechanics, molecular dynamics, and statistical mechanics) properly incorporated into computer software could be used to design new drugs, new catalysts, new chemicals, and new materials. The pursuit of this dream has led to sufficient advances in quantum chemistry and molecular dynamics (collectively referred to as atomistic simulations) that there are enormous potential opportunities for application to numerous important industrial problems involving the materials sciences, biotechnology, and chemical technology.
This opportunity is to design, characterize, and optimize materials before beginning the expensive experimental processes of synthesis, characterization, processing, assembly, and testing. We refer to this as de novo or first principles design. With reliable de novo simulations on real materials, industry could save enormously by cutting years off development cycles, while achieving designs that are more efficient. Moreover, such de novo design would allow efficient consideration of completely new materials and designs, which is particularly important for the challenges of environmentally benign industrial chemistry.
Over the last decade the potentially enormous payoffs from such activities have led to major investments by pharmaceutical, chemical, oil, and materials oriented industrial organizations to develop in-house capabilities for theory and computer simulations of molecules and materials. There have been significant advances and these industries can point to successes justifying the investments. However, these advances leave us far short of the potential impact theory could have on practical issues in the product and process development activities critical to the chemical, oil, and materials industries. The difficulty here is the enormous gap between the capabilities of current software/hardware for applying the theory to molecules and materials and the practical problems facing the designers of new materials and new processes. Thus the designer may be faced with eliminating an additive or solvent involved with a polymer blend because of regulatory issues (environment, toxicity, flame retardation, recycling) while needing to retain or enhance the materials properties (glass temperature, modulus, heat resistance, oxidation resistance, etc.). The molecular weight of the polymer may involve about 100,000 atoms per chain so that atomistic calculations would require 500,000 to 1,000,000 atoms per unit cell to properly predict the properties of these materials. To optimize the design, the engineer needs to test quickly the properties for various strategies requiring results back from any computer experiment quickly (ideally in seconds, but certainly in a few minutes). This contrasts with the situation using current theoretical technology where 1 microsecond (of simulations) on 1,000,000 atoms per unit cell (of the pseudo-amorphous material) might require ~150 years. Thus, direct de novo simulations on the systems of greatest industrial importance seems impossible. Consequently such de novo theory plays little role in the practical product and process development activities so important to industry.
The Materials and Process Simulation Center (MSC) was established in 1990 as part of the Beckman Institute at Caltech to bridge this gap between de novo theory and industrial application. The MSC strategy is to focus on the two equally important aspects of bringing theory to industrial practice:
To provide the most rigorous and reliable predictions on new materials requires first-principles quantum mechanics (QM) -- the solution of the Schrodinger equation to obtain electronic wavefunctions and from these wavefunctions all other properties. Unfortunately, the practical time and length scales for first-principles QM are often a million times too small for industrial design. To overcome this difficulty we use a hierarchy of methodologies (see Figure 1) with QM at the foundation. We ascend this hierarchy by successive coarsening in which each step involves averaging over the elements of the finer steps to obtain effective parameters for the next. This leads to telescoping whereby one can transcend from QM to engineering with just a few levels of simulation.
Averaging over the electrons to obtain spring constants, discrete charges, and van der Waals parameters, allows QM to be replaced with Molecular Dynamics (MD) where one solves coupled Newtonian equations to predict the motions of systems 100 to 10000 times larger for periods about 1000 to 1,000,000 times longer. For some industrial applications, such MD simulations are adequate.
However most real industrial materials applications require consideration of length scales of at least a micron to a millimeter for time scales of a microsecond to milliseconds or longer. For such scales, atomistic MD simulations remain inadequate by many orders of magnitude. Instead, we use mesoscale techniques in which collections of atoms (segments or grains) are considered as the elementary particles. The parameters for this mesoscale description are obtained from large scale atomistic MD. Thus, the mesoscale is still based on de novo simulations.
Finally, to describe the properties at the macroscopic relevant to many industrial processes, one averages over the mesoscale to obtain constitutive equations and parameters for use in continuum modeling (finite element calculations, continuum mechanics, Navier-Stokes equations, Maxwell's equations). Such continuum studies are the meeting ground between de novo atomistic theory and methods of engineering design.
To simulate and design the chemical engineering processes and manufacturing requires parameters describing how the properties of materials change with temperature, pressure, flow, etc. The weakest aspect of current chemical engineering software is the inadequate data available (essentially all based on experiment) and the inadequate models for relating chemical structure and composition to the important properties. For optimum design, it is essential to base such calculations on the results of de novo simulations. The required macroscopic properties are obtained by averaging over the atomic and mesoscale levels to obtain constitutive equations, parameters, phase diagrams, etc. for finite element calculations, continuum mechanics, Navier-Stokes equations, Maxwell's equations). This would provide correlations between structure, composition, and properties not available from experiment.
A very special opportunity now exists for de novo design and simulation to play a role in the pharmaceutical, genome therapy, cancer therapy, and bioengineering industries. Here the critical gap has been in the ability to predict structure (and function) from first principles. Amazing progress is being made here (using hierarchical techniques) for both globular (aqueous) and membrane bound proteins. We believe that successes will be achieved here by around the end of this millennium (Dec. 31, 2000), providing the basis for de novo design.
Figure 1. The hierarchical strategy for coupling atomistic simulations to engineering design.
Summarizing, despite the tremendous recent progress in the methods for QM and MD simulations, they remain inadequate for the most important industrial problems in materials science, biotechnology, chemical technology, and nanotechnology. Even so, recent progress in theoretical methodologies bridging the various length and time scales, combined with the continuing progress in scalable massively parallel hardware convinces us that the bottlenecks to developing theory and simulation tools suitable for de novo design and characterization of real materials will be overcome in the next few years.
Developing the theoretical methods and software is only half the job. Most advances using theory to help solve real industrial problems have had finesse as a critical component, and this is likely to remain the case for the most interesting developments. The scientist with a thorough understanding of both the basic theory and the application simplifies the theory to make the calculations practical while ensuring that the simulation still adequately describes the application of interest. As a result, progress is most effectively made by coupling the theorists at the cutting edge of theory and simulation with the chemists, biotechnologists, or material scientists driven to design new or improved materials, catalysts, processes, etc.
There is a difficulty in achieving this coupling. The basic theoretical research is concentrated in universities, and these groups often have little direct knowledge of the most important industrial applications. In addition, the theoretical efforts tend to be fragmented with no central location where a variety of theoretical efforts are coordinated to focus on providing the breakthroughs needed for real applications. On the other hand, even the largest industrial organization usually has at most only a handful of theorists, far short of the critical mass needed to keep up with the research advances in the various fundamental areas of theory or with the applications to various fields. It was to resolve this dilemma that the MSC was established.
The Materials and Molecular Simulation Center (MSC) was established in July 1990 as part of the Beckman Institute (BI) at Caltech with two major goals:
The overall objectives of the MSC are to:
This philosophy for the MSC evolved over the 1970's and 1980's as a result of consulting and research activities between Professor Goddard and various companies (including Shell Development Corp., Allied-Signal, ICI, General Electric, Molecular Simulations Inc., General Motors, SOHIO, Nippon Steel, Dow Chemical, Xerox Corp.).
Particularly important in establishing the MSC was the Biocatalysis and Chemical Technology Research (BCTR) program of the Advanced Industrial Concepts Design (AICD) Division of the Office of Industrial Processes of the Office of Industrial Technologies of the United States Department of Energy (DOE); hereafter referred to as DOE/AICD. (Initially this was called ECUT for Energy Conservation and Utilization Technologies.) This part of DOE was particularly concerned with bridging the gap between basic research and industrial practice and encouraged technology transfer. Program directors for DOE during this period included Jim Eberhardt and David Boron from DOE and Minoo Dastoor and Gene Peterson of JPL.
In addition, funding from the AFOSR/DARPA project for tribology/ceramics and from various DARPA/ONR/URI programs played an essential role, as did funding of the basic research activities from National Science Foundation (Chemistry and Materials Research Divisions).
The MSC started officially in July 1990 with seed funding from DOE/AICD. The MSC and the Goddard research group moved into the current quarters of the Beckman Institute in December 1990. The dedication was held on January 28 and 29, 1991, with a one and one-half day workshop aimed at industrial research managers. Mary Good gave the keynote address and many from industry, government, and academia attended.
Various industrial laboratories started official collaborations with the MSC, beginning with BP America and Asahi Chemical in summer 1990. Dr. Siddharth Dasgupta, a Beckman Institute Fellow, assisted Prof. Goddard in establishing and managing these industrial collaborations in atomistic simulations. Trained as an experimental Bioinorganic Spectroscopist, Dr. Dasgupta helped provide an experimentalists perspective in our collaborations.
A focus on highly parallel computing for large scale atomistic simulations (million atom molecular dynamics) was initiated October 1992 by the NSF with the formation of a Grand Challenge Application Group centered at the MSC (Program Managers at NSF: Al Thaylor and Dick Hilderbrandt).
This five-year project involved five Principle Investigators:
By Mid 1993 the advances in atomistic simulations (quantum mechanics and molecular dynamics), the lower two boxes in Figure 1 were sufficient to begin considering strategies for developing methods toward process simulation. Thus in November 1993, Dr. Mario Blanco was recruited to spearhead and manage these industrially oriented activities. Dr. Blanco had ten years of experience in industry (Rohm and Haas and MSI) and brought considerable expertise in these areas. This led to changing the name of the center from the Materials and Molecular Simulation Center to the current name, the Materials and Process Simulation Center (MSC). It continues to be directed by Professor William A. Goddard III.
In May 1995, MSC recruited Dr. Tahir Cagin from Molecular Simulations Inc. where he managed and developed simulation and modeling software for the PolyGraf, BioGraf, and Cerius2 commercial software packages. Trained in physics and materials science, with a focus on advanced molecular dynamics, Dr. Tahir helped coordinate the development of massively parallel simulation and modeling software, while extending the methods toward mesoscale simulations, metallic alloys, and ceramics.
As it became important to coordinate an increasing emphasis on advanced quantum methods for catalysis and advanced materials, Dr. Richard Muller was recruited in June 1997 to join the Management team of the MSC. Dr Muller had obtained his PhD in quantum chemical methods with Prof. Goddard and had postdocked at USC with advanced simulations of biological systems.
As it became important to coordinate an increasing emphasis on biomacromolecular systems and drug design, Dr. Nagarajan Vaidehi (also known as Dr. Vaidehi Nagarajan) was promoted from postdoctoral fellow to a senior MSC manager in September 1997.
Since 1997 two of the managers, Dr. Dasgupta and Dr. Muller, have left the group for opportunities in the private sector and the Sandia National Labs, respectively. Dr. Dasgupta was suceeded by Dr. Alejandro Strachan, who subsequently left for a position at the Los Alamos National Labs.
Dr. Strachan was replaced by Dr. Adri van Duin, a long-time collaborator in the field of reactive force fields. Dr. van Duin was previously a Fellow at University of Newcastle upon Tyne, United Kingdom, where he in addition to force fields chiefly worked with organic geochemical research. With reactive force fields taking a more and more integral part as the connector between first principle and molecular dynamics, Dr. van Duin is involved in projects from all parts of the MSC.
Dr. Muller was replaced by Dr. Jonas Oxgaard, a physical organic chemist with a background in synthesis and applied computational chemistry, Dr. Oxgaard now coordinates work in the fields of catalysis, organic chemistry and quantum chemistry.
In addition to the above five directors, two more position have been added. Dr. Mamadou Diallo, a faculty member at Howard University, Washington DC, coordinates our environmental chemistry.
In recent years the group has had an increasing focus on fuel cells. Since the design of more efficient fuel cells involve research in atom/ion transport, catalysis, surface science and electrochemistry, a director with experience in fuel cell research was recruited, Dr. Boris Merinov.
This staff of seven senior managers work with Prof. Goddard to guide the development of new technology and to apply it to projects funded by industry or government.
Current industrial collaborations with the MSC include:
Previous projects have been completed for
Currently there are ~35 full-time scientists in the MSC, with primary training in Chemistry, Materials Science, Applied Physics, Physics, Chemical Engineering, Computer Science, and Biology. Further, there are collaborations with 13 other research groups at Caltech (Chemistry, Chemical Engineering, Materials Science, Biology, Geophysics, Mechanical Engineering, JPL, and the Beckman Institute). In addition, there are numerous outside academic collaborators and ~10 collaborative industrial projects. Generally there are ~ 3 industrial visitors on site in any one week. The net result is that the MSC has a critical mass for research and applications with the deepest and most comprehensive program anywhere in the world for microscopic simulations of chemical, materials, biological, and nanotechnology problems.
Generally, the government funding supports graduate student activities and development of new methods, while the industrial support is mainly toward postdoctoral fellows and specific applications.
The MSC has two levels of industrial participation:
Once a year the MSC presents a two-day workshop to highlight the advances at the MSC over the previous year. This was established primarily to benefit the corporate collaborators, each of which can send several employees with no additional registration fee (only charged for meals). These workshops are also open to the Caltech Industrial Associates, but otherwise are by invitation only (for such invitees the registration fee is $175 plus meals). The program and other details for MSC2000 are now available.
In addition, once a year the MSC is willing to organize a special one-day workshop for personnel of a specific CP. This would highlight the activities relevant to the interests of the CP and can involve one-on-one, hand-on experience with the software. It may be oriented toward management, scientists, or engineers depending upon the interest of the CP.
The research activities of the MSC are summarized in the programs of the various annual workshops and in other documents on the MSC website.
The breadth and depth of these activities and of the collaborations allows a critical mass of personnel in each activity involving scientists and engineers with training in a number of different fields (chemistry, physics, applied physics, chemical engineering, material science, biology, computer science) and involving experimentalists in addition to theorists. Indeed there has been enormous progress in many of these areas since establishing the MSC progress that stems directly from the variety of activities and perspectives brought together in the MSC.