New directions in ab initio modeling of materials properties

In many areas, technological advance depends increasingly on the capability to control the structure and composition of materials, devices and biological systems on an atomic level. This is particularly true in technological key areas, such as micro- and the rapidly advancing nano-electronics, nano-structured magnetic and magneto-optic storage devices, catalysis or biotechnology. Due to the reduced dimensions, which are often approaching few atoms, many local properties and processes are no longer accessible in the laboratory. Furthermore, materials properties depend to a large extent on quantum effects. The small, atomic dimensions entail the importance of atomic-scale computer simulations, in which the properties of single atoms are calculated and predicted. The importance of quantum effects requires a full quantum mechanical description of the electrons. Although the basis for such a quantum mechanical treatment has been put in place by Erwin Schrödinger 70 years ago (Nobel price 1933), exact calculations for systems containing many interacting electrons remained intractable until recently. Only the development of ab initio density functional theory, for which the scientist Walter Kohn received a Nobel price in 1998, combined with algorithmic advances, and the implementation of these algorithms on large scale parallel computers has led to a revolution in atomic-scale computational modelling in the last 20 years. Today these methods are used transdisciplinarily for the investigation of metallic, semiconducting and insulating materials, as well as nano-structured and biological systems. With the development of the software package VASP (Vienna Ab initio Simulation Package), important contributions to this breakthrough have been made in Vienna. This success is documented by the large user basis of VASP: presently 400 research institutions are using and applying VASP, among them are such prestigious centers as the MIT (Massachusetts Institute of Technology), or the research and development departments of General Motors, Ford, Sony, Intel, Motorola and TOTAL. The success also demonstrates that basic research, carried out in Austria, can have an extremely high impact, even against stiff international competition. The Start price will make it possible for Austria to remain a center for the development of ab initio methods, and possibly, the international leadership can be even extended. The projected developments cover relatively many aspects, and can be summarized under the two headings "higher precession", and "larger and more complex" systems. "More precise" calculations: The applied density functional methods describe the interaction between electrons only approximately. In certain areas, such as magnetic devices, which are used in hard discs, these approximations can yield significant deviations between the real and computer experiment. Also the properties of semiconductors and insulators and defects in semiconductors, which are important for the production of highly integrated chips (Intel, Motorola), can not be predicted with sufficient accuracy. The proposal aims to develop methods that solve these problems, and therefore increase the precession of density functional based calculations. "Larger and more complex" systems: A fundamental problem of the existing methods is that their computational effort scales as the cube of the system size. Whenever the number of atoms is doubled, the computational time therefore increases by a factor eight (2x2x2=8). This implies that many important and complicated processes can not be simulated directly in the computer, even though the performance of computers has increased dramatically in the last 20 years. It is clearly necessary to develop algorithms that improve upon the existing scaling properties, and theoretically it has been shown indeed that so called linear scaling methods must exist, i.e. methods in which the computational demand scales only linearly with the number of atoms. To date, computer codes with sufficient stability and accuracy, that apply such methods do not exist. The proposal aims to implement a linear scaling density functional code, by means of a strong interdisciplinary cooperation, including computer scientists and applied mathematicians. The new methods will allow to address a variety of important open materials science problems. Realistic predictions for the properties of defects in technologically relevant insulators and semiconductors will become possible. Furthermore, predictions on chemical and catalytic processes can be expected to become more precise. Finally linear scaling algorithms will make density functional theory applicable to important new fields, such as nano-structures (quantum dots, nano-wires, nano-electronic devices) and biological systems.

In many areas, technological advance depends increasingly on the capability to control the structure and composition of materials, devices and biological systems on an atomic level. This is particularly true in technological key areas, such as micro- and the rapidly advancing nano-electronics, nano-structured magnetic and magneto-optic storage devices, catalysis or biotechnology. Due to the reduced dimensions, which are often approaching few atoms, many local properties and processes are no longer accessible in the laboratory. Furthermore, materials properties depend to a large extent on quantum effects. The small, atomic dimensions entail the importance of atomic-scale computer simulations, in which the properties of single atoms are calculated and predicted. The importance of quantum effects requires a full quantum mechanical description of the electrons. Although the basis for such a quantum mechanical treatment has been put in place by Erwin Schrödinger 70 years ago (Nobel price 1933), exact calculations for systems containing many interacting electrons remained intractable until recently. Only the development of ab initio density functional theory, for which the scientist Walter Kohn received a Nobel price in 1998, combined with algorithmic advances, and the implementation of these algorithms on large scale parallel computers has led to a revolution in atomic-scale computational modelling in the last 20 years. Today these methods are used transdisciplinarily for the investigation of metallic, semiconducting and insulating materials, as well as nano-structured and biological systems. With the development of the software package VASP (Vienna Ab initio Simulation Package), important contributions to this breakthrough have been made in Vienna. This success is documented by the large user basis of VASP: presently 400 research institutions are using and applying VASP, among them are such prestigious centers as the MIT (Massachusetts Institute of Technology), or the research and development departments of General Motors, Ford, Sony, Intel, Motorola and TOTAL. The success also demonstrates that basic research, carried out in Austria, can have an extremely high impact, even against stiff international competition. The Start price will make it possible for Austria to remain a center for the development of ab initio methods, and possibly, the international leadership can be even extended. The projected developments cover relatively many aspects, and can be summarized under the two headings "higher precession", and "larger and more complex" systems. "More precise" calculations: The applied density functional methods describe the interaction between electrons only approximately. In certain areas, such as magnetic devices, which are used in hard discs, these approximations can yield significant deviations between the real and computer experiment. Also the properties of semiconductors and insulators and defects in semiconductors, which are important for the production of highly integrated chips (Intel, Motorola), can not be predicted with sufficient accuracy. The proposal aims to develop methods that solve these problems, and therefore increase the precession of density functional based calculations. "Larger and more complex" systems: A fundamental problem of the existing methods is that their computational effort scales as the cube of the system size. Whenever the number of atoms is doubled, the computational time therefore increases by a factor eight (2x2x2=8). This implies that many important and complicated processes can not be simulated directly in the computer, even though the performance of computers has increased dramatically in the last 20 years. It is clearly necessary to develop algorithms that improve upon the existing scaling properties, and theoretically it has been shown indeed that so called linear scaling methods must exist, i.e. methods in which the computational demand scales only linearly with the number of atoms. To date, computer codes with sufficient stability and accuracy, that apply such methods do not exist. The proposal aims to implement a linear scaling density functional code, by means of a strong interdisciplinary cooperation, including computer scientists and applied mathematicians. The new methods will allow to address a variety of important open materials science problems. Realistic predictions for the properties of defects in technologically relevant insulators and semiconductors will become possible. Furthermore, predictions on chemical and catalytic processes can be expected to become more precise. Finally linear scaling algorithms will make density functional theory applicable to important new fields, such as nano-structures (quantum dots, nano-wires, nano-electronic devices) and biological systems.