calque

Density Functional Theory and Beyond with Numeric Atom-Centered Orbitals - FHI-aims Developers' and Users' Meeting 2012

August 28, 2012 to August 31, 2012
Location : Free University Berlin, Germany

Organisers

  • Volker Blum (Duke University, Durham, NC, USA, USA)
  • Mariana Rossi (Oxford University, United Kingdom)
  • Matthias Scheffler (Fritz Haber Institute of the Max Planck Society (FHI), Berlin, Germany)

Supports

   CECAM

   Psi-k

Description

 

Density functional theory and approaches that go beyond its standard capabilities are the most widely used electronic structure methods for the description of solids and molecules. The European scientific community is particularly strong in this area, not least to the fact that a number of strong, integrated computer code packages are being developed by many groups in healthy competition and cooperation. In particular, a number of strong frameworks have embraced algorithms based on numeric atom-centered basis functions. FHI-aims (the Fritz Haber Institute ab initio molecular simulations package) is one of them; others include Siesta, Dmol3, or PLATO. Many other frameworks are related and use related methods, however, for example Gaussian-type methods from QC (cp2k, Turbomole, ...) and even plane-wave methods, for example, many recent efforts focused around explicit localization of plane wave derived orbitals for many-body perturbation approaches.
Such developments can only be of use to their own and the wider community if the communication among its developers, with developers of other, related projects, and between developers and users is organized as effective and attractive as possible. The goal of the proposed workshop „FHI-aims Developers' and Users' meetings“ is to bring together developers from the immediate FHI-aims community with speakers and participants from other projects in the field and scientists from the wider community. We emphasize that this workshop is not just a code-centric, closed activity, but instead is expressly intended to foster information flow with the wider community, and in relation with other projects.
We here request support for a focused meeting with a target number of 40 to (at most) 50 participants. We have held a very successful earlier meeting in Berlin in 2010, which included scientific talks as well as focused „hands-on discussions“ with computers on site to facilitate any exchange on development and methods as much as possible. Especially the "hands-on" discussion is an extremely effective vehicle to facilitate the communication between expert developers, expert users, scientists from other communities, and newcomers alike. We here suggest a similar scope, mixing talks, tutorials, and a „hands-on discussion“, to be held at CECAM in Lausanne. 

Density functional theory and approaches that go beyond its standard capabilities are the most widely used electronic structure methods for the description of solids and molecules. The European scientific community is particularly strong in this area, not least to the fact that a number of strong, integrated computer code packages are being developed by many groups in healthy competition and cooperation. In particular, a number of strong frameworks have embraced algorithms based on numeric atom-centered basis functions. FHI-aims (the Fritz Haber Institute ab initio molecular simulations package) is one of them; others include Siesta, Dmol3, or PLATO. Many other frameworks are related and use related methods, however, for example Gaussian-type methods from QC (cp2k, Turbomole, ...) and even plane-wave methods, for example, many recent efforts focused around explicit localization of plane wave derived orbitals for many-body perturbation approaches.
Such developments can only be of use to their own and the wider community if the communication among its developers, with developers of other, related projects, and between developers and users is organized as effective and attractive as possible. The goal of the proposed workshop „FHI-aims Developers' and Users' meetings“ is to bring together developers from the immediate FHI-aims community with speakers and participants from other projects in the field and scientists from the wider community. We emphasize that this workshop is not just a code-centric, closed activity, but instead is expressly intended to foster information flow with the wider community, and in relation with other projects.
We here request support for a focused meeting with a target number of 40 to (at most) 50 participants. We have held a very successful earlier meeting in Berlin in 2010, which included scientific talks as well as focused „hands-on discussions“ with computers on site to facilitate any exchange on development and methods as much as possible. Especially the "hands-on" discussion is an extremely effective vehicle to facilitate the communication between expert developers, expert users, scientists from other communities, and newcomers alike. We here suggest a similar scope, mixing talks, tutorials, and a „hands-on discussion“, to be held at CECAM in Lausanne. 

 

Again, density functional theory and approaches that go beyond its standard capabilities are the most widely used electronic structure methods for the description of solids and molecules. Development activities in Europe have been especially strong and pushed the community to a leading role in this area worldwide. We presently see three key developments in that drive and shape the future development of electronic structure theory for simulations of real materials:

 

(1) Accuracy of numerical algorithms and methods. The accuracy of methods employed for the description of electronic structure theory is a sustained focus of the community. DFT-LDA/GGAs as the workhorse methods of the past approx. 20 years are becoming increasingly standardized, with a trend towards converging the numerical accuracy requirements between different codes and methods. On the other hand, the limitations of these workhorse method are by now well understood and, in some areas, severe, with a strong push to overcome them by use of „higher-level“ methods for realistically large system sizes, often many-body perturbation theory based or inspired.

(2) Push towards large system sizes. From system sizes 10-20 atom only some 15 years ago, the mainstream of electronic structure theory (DFT-based) is now moving towards more realistic simulations that routinely encompass 100's or 1000's or atoms. Similarly, the complexity of simulations is increasing, for instance the use of molecular dynamics methods now into the 100s of picoseconds / low nanoseconds range. While there is still much room to grow, appropriately scaling (algorithmic and memory) developments are an ongoing focus across the community.

(3) Growth and change of computational resources. A key driving factor behind the above developments is the ongoing growth in computational resources, of which all strong development projects in the fields are attempting to make use. Even the present paradigm (multi-core, multinode architectures) leaves much room for improvements, as do new paradigms such as GPU-based architecture. 

 

Attaining the goals (1) and (2) depends heavily on leveraging the more technical aspects (3) as effectively as possible.

 

The goals of our own developments in FHI-aims [1] are to pursue all three goals (1)-(3) aggressively, and this has been our focus over the past three years. FHI-aims is designed as an efficient, accurate all-electron computer code for electronic structure theory, especially, the development of accurate, hierarchical numeric atom-centered basis sets up to very tightly converged accuracy has been an invaluable foundation for our work [1]. The workhorse methods are DFT-LDA/GGA, with a strong push towards methods „beyond“ (Hartree-Fock, hybrid functionals, MP2, RPA and many-body perturbation theory based on it, GW self-energies etc.). The code treats cluster-type (non-periodic) and periodic systems on equal footing. To give a few examples that are specifically relevant for (1)-(3):

We have demonstrated the use of FHI-aims for accurate DFT-LDA/GGA based benchmark calculations against pertinent experiments both for solids (large-scale surface reconstruction energies of Au and Pt [2]) as well as for  biomolecular systems (computational vs. experimental spectroscopy [3] and need for van der Waals based methods [4] in a 180-atom benchmark polypeptide).

Many-body developments within FHI-aims have gone well beyond the RPA by now [5], and ongoing developments include self-consistent GW and efficient periodic hybrid functionals for large-scale systems, among many others. 

Finally, our efforts towards scalable computing are perhaps best documented in a significant, collaborative development effort of the scalable, massively parallel eigensolver library ELPA (http://elpa.rzg.mpg.de), which significantly enhances the reach of algebraic eigensolvers through algorithmic and numerical enhancements [6]. In our view, this library fills a significant need in the wider community, and this development (that has seen sustained production use within FHI-aims over the past two years) is now available as a standalone, open-source library (http://elpa-lib.fhi-berlin.mpg.de/) that others are adopting. 

 

Many other projects and methods developers work towards similar goals, and our goal here is an information exchange both within and outside our immediate community. Examples of strong related projects that we hope to represent at the meeting are the EStest verification & validation database (F. Gygi) [7], transport developments from the Turbomole [8] community (A. Bagrets), strong correlation many-body approaches (S. Biermann) [9], or the interaction with the phonopy project (by way of a tutorial, J. Meyer, TU Munich) [10].

 

However, a continued push towards the goals (1)-(3) can only work and be successful if we include developers and users (and their experience and needs) from the wider community in Europe and elsewhere in our developments, interchange of information and ideas. We are well aware that we are not the only ones doing such developments, and we must continue to engage and sustain relations between our and related projects. Our first Users' and Developers' meeting was held in Berlin with this goal in mind. Not only were we able to include the existing community around FHI-aims, but we were also able to develop some entirely new contacts through this meeting. A particularly effective aspect of the meeting were afternoon „hands-on discussions“ by focused subgroups of the participants, with computers on hand. We would here like to repeat this successful strategy and enhance it by two tutorials on FHI-aims: First, the basic use and handling of the code, and second, on a specific aspect that demonstrates how to couple FHI-aims and an existing, outside open-source framework for a specific application (phonopy [10] for phonons). 

References

[1] Volker Blum, Ralf Gehrke, Felix Hanke, Paula Havu, Ville Havu, Xinguo Ren, Karsten Reuter, and Matthias Scheffler, "Ab initio molecular simulations with numeric atom-centered orbitals", Computer Physics Communications 180, 2175-2196 (2009)

[2] P. Havu, V. Blum, V. Havu, P. Rinke, and M. Scheffler, "Large-scale surface reconstruction energetics of Pt(100) and Au(100) by all-electron density functional theory", Phys. Rev. B 82, 161418(R) (2010).

[3] M. Rossi, V. Blum, P. Kupser, G. von Helden, F. Bierau, K. Pagel, G. Meijer, and M. Scheffler, "Secondary structure of Ac-Alan-LysH+ polyalanine peptides (n=5,10,15) in vacuo: Helical or not?", J. Phys. Chem. Lett. 1, 3465-3470 (2010).

[4] A. Tkatchenko, M. Rossi, V. Blum, J. Ireta, and M. Scheffler, "Unraveling the stability of polypeptide helices: Critical role of van der waals interactions", Phys. Rev. Lett. 106, 118102 (2011).

[5] X. Ren, A. Tkatchenko, P. Rinke, and M. Scheffler, "Beyond the Random Phase Approximation for the Electron Correlation Energy: Importance of Single Excitations",
Phys. Rev. Lett. 106, 153003 (2011).

[6] T. Auckenthaler, V. Blum, H.-J. Bungartz, T. Huckle, R. Johanni, L. Krämer, B. Lang, H. Lederer, and P. R. Willems, "Parallel solution of partial symmetric eigenvalue problems from electronic structure calculations", Parallel Computing (2011), doi:10.1016/j.parco.2011.05.002.

[7] Gary Yuan and François Gygi, "ESTEST: a framework for the validation and verification of electronic structure codes ", Comput. Sci. Disc. 3, 015004 (2010)

[8] http://www.turbomole.com/ .

[9] Ferdi Aryasetiawan and Silke Biermann "Generalized Hedin equations and sGsW approximation for quantum many-body systems with spin-dependent interactions", Journal of Physics: Condensed Matter 21, 064232 (2010)

[10] Atsushi Togo, Fumiyasu Oba, and Isao Tanaka, “First-principles calculations of the ferroelastic transition between rutile-type and CaCl2-type SiO2 at high pressures”, Phys. Rev. B 78, 134106 (2008); http://phonopy.sourceforge.net