European Commission logo
English English
CORDIS - EU research results
CORDIS

Unrestricted Divide-Expand-Consolidate Coupled Cluster Methods for Large Open-Shell Molecules

Periodic Reporting for period 1 - DECOS (Unrestricted Divide-Expand-Consolidate Coupled Cluster Methods for Large Open-Shell Molecules)

Reporting period: 2015-04-01 to 2017-03-31

The MSCA “DECOS” project addresses one of the most challenging problems of contemporary computational chemistry – the ability of theoretical modeling to describe accurately large open-shell molecular systems (molecules with one or more unpaired electrons) at affordable computational costs. To this end the Divide-Expand-Consolidate (DEC) scheme was extended to unrestricted formalism. The new method constitutes crucial breakthrough because it combines linear scaling and rigorous error control of the established DEC scheme with the ability of treating large open-shell systems for biological and industrial use. Indeed, magnetic and catalytic properties, reactivity and spectroscopy often must be entirely attributed to special electronic structure of open-shell systems. Additionally, through unique collaboration with Oak Ridge National Laboratory in Tennessee, US, the new methods were implemented, tested and run on “TITAN” – one of the largest super computers in the world.
The MSCA fellow acquired new expertise in the wide family of the Coupled Cluster (CC) correlated methods. He took the advanced courses at Aarhus University. He got knowledge in cutting edge theoretical chemistry developments. For example Lagrangian-based triples perturbative corrections, general time dependent response functions theory. Thus, new competences in the mathematical formulation of new quantum chemical methods helped quickly understand the details of the DEC method and reach the scientific objective of the MSCA “DECOS” action. During the “DECOS” project the fellow was able to publish 6 scientific papers, 3 scientific talks were delivered. The LSDalton program was extended in the functionality with newly developed methods.
Additionally, through unique collaboration with Oak Ridge National Laboratory in Tennessee, US, the new methods were implemented in the massively parallel manner. The fellow acquired new technical knowledge in parallel implementations (through research visits to Oak Ridge National Laboratory in USA) and graphical processing units (GPU) programming (at the summer school in Switzerland). Eventually, the program was tested and run on “TITAN” – one of the largest super computers in the world. As a result the largest correlational calculation to that date was performed using DEC methodology. The fellow within the team from Aarhus University took part in the Gordon-Bell supercomputer competition.
Besides scientific component, the project also became crucial step towards reaching professional maturity and independence in the fellow’s career. During the project new long-term collaborations were established with USA (Oak Ridge National Laboratory) and Germany (University of Duisburg-Essen, Max Planck Institute for Coal Research). The developments in the “DECOS” project will be used and maintained for many years beyond the lifetime of this grant as LSDalton program getting more popularity and becoming open-source.
The MSCA “DECOS” project dissemination activities were realized to make high impact. The new methods were implemented in quantum chemistry program LSDalton, and are now at the program forum and also distributed among the users of the programs as well as available for the general public. Published peer-reviewed articles were popularized on FACEBOOK and LINKEDIN; A website was made for the “DECOS” project. The project announcement was made at the website of the Aarhus University and at the info screens at the faculty. The MSCA fellow presented “DECOS” project results during research visits to US, Germany and Netherlands.
The greatest impact for the MSCA fellow was gaining experience in mathematical formulation of new quantum chemistry methods. Most importantly was to acquire new expertise in the wide family of the correlated CC methods. These methods are unavoidable in solving the electron correlation problems in theoretical chemistry, but limitedly used to date. The fellow’s competencies were pushed greatly forward by acquiring a number of technical skills at Aarhus University and Oak Ridge National Laboratory, such as programming of massively parallel algorithms, GPU programming, code generators.
As a results of MSCA action fellow’s soft skills were also trained. To this end, complementary training program was designed and executed. The fellow gained experience in management, teaching, and organizational skills.
It is also very important that all the skills are “perfectly transferable”. They should not depend on special equipment or professional staff in a special location. Thus the skills make direct impact to the fellow’s career profile and stay with him for many years after the proposal timeframe.
Reduction of the scaling makes possible calculation of large molecular systems