Quantum-chemistry methods for materials science
CECAM-HQ-EPFL, Lausanne, Switzerland
Density-functional theory (DFT) [1,2] has been the method of choice for electronic-structure calculations in materials science over two decades. However, certain well-documented failures such as unsatisfactory prediction of atomization energies and underestimation of weak interactions and reaction barriers, limit the predictive power of current density-functional approximations, including (semi)-local and hybrid functionals in materials science [3,4].
The desire for general-purpose electronic-structure methods with high accuracy is pressing, especially in materials science. Together with the rapid growth of computational capacity, this has drawn attention to the sophisticated quantum-chemistry methodologies rooted in wave-function theory (WFT) [5,6]. WFT offers a systematic hierarchy to approach the exact solution of the many-electron Schrödinger equation. The Møller-Plesset perturbation theory and the coupled-cluster approach are two popular choices in quantum chemistry. In contrast to DFT, these WFT-based quantum-chemistry methods go beyond the single-electron mean-field model and take the correlation effects into account in an explicit many-body picture. The improvable accuracy together with potentially richer electronic-structure information make them very promising in materials science.
The implementation of popular quantum-chemistry methods to condensed matter systems, including the second-order Møller-Plesset perturbation method (MP2) and the coupled-cluster approach with singles, doubles, and perturbative triples (CCSD(T)), has been done in several mainstream computational platforms. Their applications in solids and surfaces have been presented by the world's leading researchers and their groups [5,6]. As a great progress towards an exact description of solids, Booth et al. have demonstrated the possibility to perform a full configuration interaction quality calculation in periodic boundary conditions with the aid of the Quantum Monte Carlo stochastic strategy . Recently, The Journal of Chemical Physics issued a Special Topic Section on “Advanced electronic structure methods for solids and surfaces” , which shows up the enormous potential of quantum-chemistry methods for condensed matter systems, and provides a timely snapshot of this rapidly developing area in computational materials science.
However, there is still a long way to go before making quantum-chemistry methods practical for solids. Compared to popular density functionals, the quantum-chemistry methods are often much more expensive, and encounter more difficulties ensuring the numerical accuracy, efficiency and reproducibility. In quantum chemistry, these challenges are well documented, stimulating a bunch of novel methodologies and algorithms with higher accuracy and efficiency. However, in periodic boundary conditions, these challenges become severe and have not been fully studied . In this context, the central goal of the workshop is to discuss the state of the art and challenges of using quantum chemistry methods in materials science, to share the recent progresses in quantum chemistry, and to deepen the coalescence of two communities and also of two theories: density-functional theory and wave-function theory.
The proposed workshop brings together 12 internationally renowned experts and young scientists with an outstanding expertise in computational materials science or quantum chemistry, to highlight, discuss, and advance the state of the art of quantum-chemistry methods in materials science. This workshop will feature 4 sessions, each starting with a 10-minute introduction of a chairperson. She or he will also be responsible to consciously guide the discussion after each talk and each session.
The workshop will start with two sessions on the recent developments and new insights of applying quantum-chemistry methods to solids, covering:
(1) Numerical algorithms and implementation of the periodic MP2, random-phase approximation, and CCSD(T) for solids, in mainstream computational platforms, including VASP, CP2K, CRYSCOR, and FHI-aims [8-10].
(2) Numerical accuracy and reproducibility of these advanced electronic-structure methods with respect to the choice of basis sets, the treatment of core electrons, and the k-grid choice to sample the first Brillouin zone in reciprocal space [7,11,12].
(3) Low-scaling algorithms and cutting-edge applications towards large systems and computational resources [5, 13-16].
Six selected talks in the first two sessions, given by Georg Kresse (University of Vienna), Jürg Hutter (University of Zürich), Dennis Usvyat (University Regensburg), Andreas Grüneis (MPI for solid State Research), Beate Paulus (Free University, Berlin), and Xinguo Ren (USTC, China), will focus on open problems more than on results. We hope that the discussion in these sessions will facilitate an in-depth and comprehensive understanding of the timely challenges to achieve a practical applicability of the quantum-chemistry methods for solids.
The next two sessions will focus on some selective topics in quantum chemistry, aiming at building on and enhancing the benign interdisciplinary collaboration between quantum-chemistry and computational materials science. The issues that will be covered are:
(4) Recent efforts in quantum chemistry to solve the slow basis-set convergence problem, including the efficient correlation-consistent basis sets , and the explicitly correlated F12/R12 methods .
(5) Lower-scaling algorithms, to explore the locality in real space (or the sparsity in configuration space) of the electronic correlations in different levels of approximations [19-22].
(6) New embedding concepts and algorithms with higher accuracy and efficiency, including the force-based (or mixed-basis set) extension of the popular ONIOM methods [23, 24] and embedding approaches towards a correct description of molecular crystals , or even strongly correlated systems .
As one of the biggest and most active communities in computational science, quantum chemistry is constantly giving rise to progress and new concepts in the field. In these two sessions, we carefully select 6 topics which either have been exhibiting a promoting role in solids, or are considered to be feasible for the solution of the timely challenges in computational materials science as outlined in the first two sessions. The corresponding 6 talks will be offered by Edward F. Valeev (Virginaia Tech, USA), Ali Alavi (MPI for Solid State Research), Frank Neese (MPI for Chemical Energy Coversion), Xin Xu (Fudan University, Shanghai), Frederick R. Manby (University of Bristol, UK), and Garnet Kin-Lic Chan (Princeton/Caltech, USA).
We have contacted 12 key speakers for this proposed workshop, and received positive responses from all of them. (see section 2, participant list).
Matthias Scheffler (Fritz-Haber-Institut der Max-Planck-Gesellschaft) - Organiser
Igor Ying Zhang (Fritz Haber Institute of the Max Planck Society (FHI), Berlin) - Organiser