Computer usage Notur 2007
Chemistry: a many-body problem At the deepest level, molecules are simple: charged particles in motion governed by the laws of quantum mechanics HΨ=EΨ …but it is a many-body problem… “The underlying physical laws necessary for the mathematical treatment of a large part of physics and the whole of chemistry are thus completely known and the difficult is only that the exact application of these laws leads to equations that are too complicated to be soluble” Dirac (1927)
Computers came to our rescue…
Quantum chemistry HΨ=EΨ Simulations of chemical systems and processes approximate solutions of the Schrödinger equation Journal of Americal Chemical Society 40% of all articles supported by computation most of these are quantum chemical This is an amazing development for an experimental science “Every attempt to to employ mathematical methods in the study of chemical questions must be considered profoundly irrational” August Comte (1798–1857)
Computation: the third way Theory, experiment and computation interpretation and prediction of experiment alternative to experimental measurements
Example: Reaction pathways
Example: NMR spectra 200 MHz NMR spectrum of vinyllithium
Methods development: DALTON Computational models are being constantly improved increase accuracy and predictive power broaden the range of applicability and lower the cost Dalton program system Scandinavian collaboration 25 years of development broad functionality 1300 research groups 250 computer centers
Towards higher accuracy…
Towards larger systems… Real-world problems are typically large computational cost typically scales cubically or high with increasing system size however, in large systems, nearly all contributions are insignificant these should be recognized and avoided in the computations ideally, cost should increase in proportion to system size
Towards larger systems… Energies and structure of large molecules Quantum chemistry is typically done on 50 or less atoms Energy and forces for this 392-atom molecule can now be done in about one hour This is more than an order of magnitude improvement over a few years On a parallel computer, we should be able to do this in about one minute We should reach 10.000 atoms in 2010 (currently at a few thousand) Our bottleneck is memory
Computer developments Our requirements CPU power and memory—little or no data transfer or storage Encouraging developments: Powerful multicore chips Graphical processing units (GPUs) Improvements in bandwidths of interconnects Discouraging developments: Chips are not getting faster (3GHz) Multicore chips hard to program effectively GPU/CPU communication slow Software and algorithm development necessary This is our job!
Running on Blue Gene Dalton scales well to over 20.000 processing cores Argonne’s Blue Gene/P 1-PFLOPS computing with 294 192 PowerPC 450 850 MHz processors If you provide the hardware, we shall put it to good use…