This report contains the scientific program, abstracts, references and views from the US and European scientists participating in a workshop on Methods in Computational Materials Science jointly organized by the US-National Science Foundation and the European Community in San Francisco on April 15 and 16, 2004. The joint workshop was the first on computational methods. It is hoped that it will lay the foundations for several active and exciting research areas for US-EU collaborations dealing with modeling the complex behavior of materials, and spanning length scales from the atomic level to the continuum.
The objectives of the workshop were two-fold: (i) to explore, identify and assess current and future opportunities in computational materials science, and (ii) to discuss the form and mechanisms appropriate for collaborative efforts between US researchers and their European counterparts.
There were nineteen invited contributors from Europe and twenty-three from the US. New computational methods for understanding and predicting materials properties and phenomena were presented. Contributions spanned a broad range of approaches, length and time-scales, and applications. These electronic and structural properties, solidification, surface adsorption, structural defects, processing-microstructure and structure-properties links, functional and nano-structured materials, crystal growth, microstructural evolution in heterogeneous materials, mechanistic methods to predict crack initiation and growth, diffusion, spintronics, molecular and nano-electronics, and quantum dots. A broad range of computational materials modeling approaches were covered, including ab initio methods, molecular dynamics simulations, Monte Carlo techniques, statistical mechanics, crystal and polycrystal plasticity, and other continuum mechanics approaches.
The format of the meeting included keynote presentations in a common session, and invited contributions in three parallel sessions. On both days, the scientific sessions were followed by discussions, which were held in small groups in areas of common interest. The discussions groups gathered in terms of material / application types on the first day, namely:
and according to length scales on the second day, namely
Not surprisingly, the grouping of the breakout discussions by length scale and by materials types/applications resulted in a similar grouping of the participants. For example, the electronic properties of materials are dependent on the electronic structure of the material, e.g., the nature of the bond between neighboring atoms, and as such they are determined at the atomistic scale. In contrast, even though structural properties of a material are generally macroscopic, they depend on microstructural features (e.g. grain size, grain boundary thickness, precipitates), which need to be considered at different length scales.
There was overall agreement amongst the participants that methods at all length scales are important and that to no one method will be sufficient to understand wide ranging properties. In fact, it is difficult to envision a single method that would be able to predict the properties of materials at the atomistic and macroscopic scales simultaneously.
Currently, there are several approaches that are able to handle disparate length scales, under some circumstances. For instance, methods that are focused on different length scales can be connected through parameter transfer from one method to the other. As an example, atomistic codes can be used to determine accurate interatomic forces, which can in turn be relied upon to construct molecular dynamics interatomic potentials. Molecular dynamics codes can then be interfaced with continuum mechanics methods. Such linkages have been implemented with some success in the study of fracture, but interfacing between codes remains a problematic issue, as the transcription of classical to quantum mechanical forces is not straightforward. Another example is the use of atomistic codes to determine constitutive relations or equations of state. These relations can then be input into macroscopic continuum codes based on, e.g., finite element techniques. In this procedure, electronic properties are not considered. Moreover, if the atomistic configurations change in time, this may not be an appropriate approach. A related procedure involves "atomistically informed models of surfaces, interfaces and grain boundaries. Ideally, one would like to scale up current ab initio methods for very large systems, e.g., millions of atoms, using the so-called order-N methods. These approaches would not invoke any ad hoc assumptions. However, even with ever increasing computational power, ab initio approaches are not likely to exceed the nanoscale (e.g., 100 nm) in the near future, although even at this scale such methods will clearly have a profound impact on nano science and technology.
Given the difficulties involved with multiscale phenomena, all the participants agreed that this problem takes on a "grand challenge" character.
A key issue in any discussion of complex systems is the nature and use of approximate methods. It will be crucial to test and assess such methods by resorting to either more rigorous theories, which might span a limited length and time scale, or to experimental validation of the approximate models.
Each discussion group identified fundamental scientific issues at each length scale, which was considered to be highly appropriate for EU-US collaborations. At the nanoscale, areas identified were
Many of these areas of study involve new forms of materials, such as spintronic materials, which may result in new electronic devices that will utilize both electronic charge and electronic spin. Other forms of matter such as nanotubes or nano wires may result in unprecedented control of electronic devices at the molecular scale. This issue can have significant technological impact. For example, silicon based technology dominates the ongoing miniaturization of electronic components. For the past thirty years, Moore's law, which states that every 18 months, the number of transistors in a processor chip doubles, has characterized progress. However, this law cannot hold indefinitely. There are serious scientific and technological issues that must be resolved as device features shrink to nanoscale dimensions, where quantum mechanical effects become important. For example, design rules for transport based on simple Ohmic behavior and field-effect transistor digital function will become suspect as a consequence of quantum effects.
"Computational materials science" networks have been set up in both the EU and US within this area. In the EU, these networks provide repositories for computer codes for a variety of electronic structure problems, e.g., codes such as VASP, ABINIT and SIESTA. In the US, comparable repositories or coding projects do not exist. In the US, sites such as those associated with the NSF-ITR program, e.g., the Materials Computation Center at Illinois, and the DOE, such as NERSC, provide links to codes. However, no concerted effort is made to update these codes on a continual or coordinated basis. A synergistic effort could be built on these differing strengths by encouraging collaborative efforts between the networks. These repositories fill a need for workers, whose efforts may focus on only one length scale regime, to have access to codes that are applicable to other length scale regimes. A related issue centers on "cyber-infrastructure." Some joint collaborations involving computational platforms maintained by an EC-NSF would enhance the exchange and implementation of codes in the US and EU.
Soft materials — including complex fluids, liquids, biological materials, polymers, colloids, granular materials, and hierarchically-structured materials - pose unique opportunities and challenges for materials theory, modeling and simulation arising from the complexity of the material building blocks, and/or the complexity of the processes involved. In many cases, relevant physics occurs over many length and time scales. The theories and methodologies for bridging those scales is often lacking. Slow dynamics, complex structures, frustration and/or competition, and dissimilar materials are examples of just a few of the issues that plague simulations of soft materials. Advances in atomistic, molecular, and mesoscale simulation methods and capabilities over the past decade have enabled breakthroughs in our understanding of soft matter systems. However, when compared to the enormous range of problems in this area, these studies represent only a partial list.
Several representative areas on the forefront of soft matter research were identified. These included:
Each of these areas has strong and complementary research efforts on both sides of the Atlantic, which would benefit from close collaboration. In many of these topics, the need for length and time scale-bridging theories and algorithm development was identified as a high priority item.
At the intermediate scale between the atomistic and the continuum, or mesoscale and at the macroscale, the following were identified as important scientific areas for any future US-EU collaborations:
For instance, most current alloy design and development procedures rely heavily on "over-simplified" phenomenological approaches, providing little insight into the dominant physical processes, hence leading to incremental, rather than inspirational new materials. Computational modeling of materials is becoming a reliable tool to underpin .materials by design. approaches and complement such traditional analytical and experimental methods. A great deal of opportunities exists to develop multi-scale mechanistic models to predict microstructural evolution. For instance, the evolution of the microstructure during processing strongly influences the final mechanical properties of materials and its mechanical behavior during service. Recent interest in nanotechnology is also driving the development of novel and sophisticated approaches for design and performance prediction of nano to micron size devices for many potential applications, e.g. electronics, photonics,drug delivery systems. The need to understand and predict the role of microstructural changes (e.g. in shape memory alloys, active materials, etc.) and degradation (e.g. defect nucleation and subsequent stress driven diffusion, grain coarsening) at the relevant scales in such types of devices constitute one of the most formidable challenges for discrete dislocation and continuum-based approaches.
In summary, the forthcoming NSF-EC collaborative research program in the topic of the workshop received heartfelt support from all the workshop participants. The general view was that such program would considerably enhance the pre-eminent international position of the EU and the US in computational materials science, and promote genuine interdisciplinary collaborations between scientists from the EU and from the US. It is hoped that that future joint US-EU scientific collaborations would drive scientific discoveries through the application of materials modeling to new and emerging areas of chemistry, physics, material science and materials engineering, and will enable the development of new capabilities to integrate appropriate modeling approaches to describe material phenomena involving different length and time scales. Collaborations between US and EU scientists would also enhance educational opportunities to young scientists through international research collaborations.