NEWS UPDATE and PROJECT OVERVIEW 12/2009
Dear QMC@HOME supporters!
- thank you very much for your continued support of our research in Quantum Chemistry! Our Concept has undergone some changes in the last year and our time was especially scarce this summer, so it is definitely time for us to let you know about the current state of the project and what our future plans are:
* PROJECT STAFF *
In the last two years the project was essentially operated by two persons, Robert for ORCA (general Quantum Chemistry calculations - more information in the ORCA forum:
http://qah.uni-muenster.de/forum_forum.php?id=22 ) and me for Amolqc (Quantum Monte Carlo calculations - see everywhere else ;-) and general project stuff. Robert will finish his PhD at the beginning of next year, but we already have a young and motivated person who will take over his work, which is likely to further follow the path of 'ab initio' (i.e. first principle, that is purely relying on the underlying quantum theory principles) modeling of organic and then biomolecular systems (still very basic science - but someone has to produce good data for the development of more approximate methods that can then be used in larger simulations like protein folding etc. ;-) The QMC calculations will again be my field of research. I finished my PhD at the beginning of this year and went to Prague for 7 months during this summer to get more insight into (and write a nice publication about :-) biomolecular modeling - but now I am fully back for QMC research, this time at the University of Cambridge (see
http://www.tcm.phy.cam.ac.uk/~mk642/ for more information soon), but of course working in cooperation with Stefan Grimme at the University of Muenster, where QMC@HOME is hosted. That directly leads us to science ...
* QUANTUM CHEMISTRY *
a) The QMC calculations of last year were again devoted to benchmark sets of non-covalently complexes. Non-covalently means that these systems are not bound by the overlap of orbitals like in the chemical bonds one draws as sticks between atoms, but by interactions between charges, by hydrogen bonds or by dispersion interactions. While being hard to model purely from (standard) theory approaches, these interactions are very important for virtually everything that happens in biochemistry. As a result non-covalent interaction are quite a hot topic in Theoretical Chemistry and related fields. We have published our QMC results in this research paper:
http://pubs.acs.org/cgi-bin/abstract.cgi/j.../jp077592t.htmlb) But now we have entered the next stage of our project: The next year will be devoted to covalently bound systems with the 'Mindless Benchmarking' set (http://pubs.acs.org/doi/abs/10.1021/ct800511q). While being ideally suited for non-covalently bound systems, QMC (or to be more precise Fixed Node Diffusion Monte Carlo) has a problem with breaking bonds in chemical reactions (quite opposite to the 'standard' methods) and we hope to find a cure for it, to make QMC the (future ;-) method of choice also in Quantum Chemistry (it is already in Physics, but the problems treated are quite different from the ones in Chemistry). The Mindless Benchmark set is composed of 165 'random' chemical reactions to exclude 'human intuition' from the results as far as possible. Benchmarking then works in the following way: If you develop a method you want to make sure, it is really an improvement. Therefore you take a benchmark, with a defined set of problems and very accurate reference values to which you can compare your values. You can then assess the quality of your method with the benchmark set by trying to reproduce the benchmark set reference values. BENCHMARKING HERE DOES NOT MEAN TO DO SIMPLE TIMING TESTS ETC. It means showing that you were able to achieve a substantial improvement. Some improvement can already be expected from a new QMC application which we would like to run on QMC@HOME: CASINO, the QMC code of the TCM group at the University of Cambridge:
http://www.tcm.phy.cam.ac.uk/~mdt26/casino...troduction.html* PROJECT STATE *
Albeit being Beta we seem to run really stable now for quite some time (so I'm just looking for the right occasion to officially leave beta behind us) and we just changed to very nice new server hardware, so we do not expect major technical problems in the next future.
Things I would like to work on in the next year for QMC@HOME are:
- add CASINO QMC application with several new (scientific) features and 64bit and GPU support
- give better support via email and forum posts ;-)
- update the webpage, especially the scientific part
- succeed with the 'Mindless Benchmark' ...
* SUMMARY *
Let me finish with a short summary:
- we will have more time again for our users in the next year (yippee!)
- our research into non-covalent systems with QMC is finished for first:
http://pubs.acs.org/cgi-bin/abstract.cgi/j.../jp077592t.html- we will continue with QMC calculations on the 'Mindless Benchmark':
http://pubs.acs.org/doi/abs/10.1021/ct800511q(to look into the performance of QMC for more complicated systems)
- hopefully soon we will have a second QMC application with a lot of new features and 64bit and GPU support
- at the same time, the ORCA side-project will be further developed, making the QMC@HOME grid available for general Quantum Chemistry
* LAST BUT NOT LEAST *
If you have any questions about the above - please post them!
Thank you again for supporting our research!
Best wishes,
Martin
http://qah.uni-muenster.de/forum_thread.ph...281&nowrap=true--------------------------------------------
I've finished the first Mindless units. The graphics shows two molecules, one seems to be H2O and the other CH4. But why two molecules.and do they interact in any way?
Tullio
-------------------------------------------
@tullio: Yes, that's two molecules and they indeed interact!
Two make this a little bit clearer it might be good to say a few words about our 'Mindless Benchmarking' approach:
- if people want to find out about the performance and/or problems of computation methods they do benchmarking (see below for more)
- normally they pick up some reactions they think to be important or for which good experimental reference values are available
- this introduces a strong bias into the benchmark sets: most of them contain only very well-behaved systems
- our idea was to 'randomize' the whole thing, so that we can benchmark our methods against really complicated cases (and exclude any 'human bias' as far as possible) from the results
- all 'mindless' systems are randomly generated within a few boundary conditions (otherwise we would end up with a lot of atoms distributed in a space as large as we let it become ...)
- when we came up with this concept we tested it on the established methods of Quantum Chemistry and it indeed seems to be a very valuable tool (a link to the publication can be found on the main page and below in the news update)
So back to Tullio's question: The system consists of two molecules just by chance, you will also encounter single molecules etc. in the mindless units with sometimes very 'normal' structures (like H2O and CH4 - Tullio learned his chemistry, well done!) and sometimes very weird structures (e.g. simply a 'string' of atom symbols). All atoms interact with each other in some way: some through 'covalent' interactions, that are shown as bonds and some through non-covalent interactions that are not illustrated in the graphics. The later interactions were of high importance for the systems we looked into so far: you might have recognized that almost all units up to now consisted of two molecules or of one half of a 'two molecule sandwich'. [yeah, know that my sentences are too long - bad German habit ...] And now we turn our attention to the 'real' bonds by calculation all these 'weird' molecules ...
Best wishes,
Martin