Summary:
The past three years have seen exciting new developments in the calculation of the energetics and spectroscopy of condensed matter using wave-function-based methods, rather than density-functional theory (DFT). These developments are coming from several directions: First, there are vigorous efforts to apply methods that evolved in the condensed-matter community (e.g. plane-wave basis sets) to quantum chemistry techniques (e.g. the MP2 and coupled-cluster approximations); second, major improvements in the scaling of well-established quantum chemistry methods with number of atoms are making it possible to apply these methods to condensed-matter systems; and third, there is major progress in embedding techniques, which allow high-level quantum chemistry or DFT techniques to be embedded in techniques of lower accuracy. The time is now ripe for a discussion workshop that will bring together researchers from both the condensed-matter physics and quantum chemistry communities, to exchange ideas and to explore directions for future developments. We propose a four-day workshop to be held at CECAM in 2008, and we seek joint CECAM/Psi-k funding for this workshop.
Background to proposal:
Since its primitive beginnings 50 years ago, the atomic-scale computer simulation of materials has undergone an extraordinary evolution, in which CECAM has played an important part. In the early days, the simulations were based on simple empirical
models of the interactions between atoms, but these already sufficed to give a reasonable description of some materials, and to calculate a wide range of properties, including structure and diffusion in liquids and solids, static and dynamics surface and defect properties in solids, etc. A major revolution came in 1985, when the Car-Parrinello paper showed how the formulation of quantum mechanics known as density-functional theory (DFT) could be used to model materials as collections of nuclei and electrons, so that the making and breaking of chemical bonds could be modelled in dynamic simulations for the first time. At that time, DFT was associated mainly with the condensed-matter physics community, but it is now widely used by quantum chemists. Over the past 20 years, DFT has gained a dominant position in materials modelling, for several reasons: it
achieves good enough accuracy to identify chemical trends in a very wide range of materials; its computer requirements scale fairly mildly with number of atoms, so that large complex systems can be treated; it is straightforward to achieve basis-set convergence; and atomic forces can be calculated at almost no extra cost, so that high-temperature dynamical and thermodynamic properties can be calculated by molecular dynamics. In particular, DFT statistical mechanics for both bulk materials and for surface
processes is becoming well established.
At the same time, there is a large and important community engaged in tackling materials problems using techniques traditionally associated with the quantum chemistry community. It has been possible for many years to make calculations on large complex
systems using the Hartree-Fock approximation, and with "hybrid" functionals, which combine the DFT and Hartree-Fock theories. These approaches have, until recently, employed Gaussian basis sets, though that situation is now rapidly changing. Furthermore, there is a long-standing and successful effort to calculate the
properties of bulk materials using the so-called "incremental" approach, with high-level wave-function-based techniques, including Moller-Plesset-2 (MP2) and the coupled-cluster hierarchy. A further important approach is the application of quantum-chemistry
techniques to embedded clusters, which in some cases allows the accurate calculation of excited states in the bulk or at surfaces.
All the foregoing developments are well known in their respective communities. But in the past two or three years a new and important current of thought has begun to emerge, and there is a renewed impetus to build bridges between the communities. On the one hand, increasing numbers of DFT practitioners are acknowledging the inadequacies of DFT. (A famous example is the wrong prediction of adsorption sites of molecules on surfaces, and the poor prediction of adsorption energies, but there are many, many other
examples.) This is stimulating some DFT groups to use quantum chemistry to benchmark or correct DFT calculations, and there are examples of DFT and quantum-chemistry research groups teaming up to do this. Other research groups are working to bring key
advantages of the DFT approach, for example automatic basis-set convergence using plane-wave basis sets, into high-level quantum chemistry techniques. On the other hand, there is a drive from the quantum-chemistry side, to exploit recent dramatic improvements in the scaling of wave-function-based methods (so called "local-MP2",
or "local-coupled-cluster") to large systems, including condensed matter. Furthermore, recent reports of wave-function based methods used for molecular dynamics simulations on condensed matter make it realistic to envisage that ways may soon be found of doing most of the things that DFT can do (e.g. DFT statistical mechanics on large complex
systems), but using wave-function based quantum-chemistry techniques that will deliver much better accuracy for many systems.
In our workshop, we want to seize this exciting opportunity to bring together researchers from the DFT and quantum-chemistry communities (or researchers belonging to both), who are actively engaged now in developing and promoting these newly emerging ambitions. We note that the three organisers cover together the three-fold theme of DFT, quantum chemistry and statistical mechanics, and the research groups of all three are working in the directions we have described. The general aims of the workshop are to examine and compare the different strategies that are being developed for using quantum chemistry techniques to study condensed matter, to consider future possibilities that are now emerging, and hopefully to stimulate the formation of new collaborations.
We note that a workshop having something in common with what we propose was held in September 2007 at the Max-Planck-Institut fuer Physik komplexer Systeme in Dresden. This was entitled "Local correlation methods: From molecules to crystals", and was organised by Birkenheuer, Schuetz, Pisani and Paulus. However, the workshop proposed here will, we believe, be broader, since it will place a strong emphasis on drawing together researchers from different backgrounds. For this reason, there will be a major difference in the topics presented, and we expect there to be only a small overlap of the lists of participants of the two workshops.