4 ms·

Next sentence: "With this theory, the properties of a many-electron system can be determined by using functionals, i.e. functions of another function, which in

Next sentence: "With this theory, the properties of a many-electron system can be determined by using functionals, i.e. functions of another function, which in this case is the spatially dependent electron density." The object of study in DFT is the distribution of particles in the system rather than the particles themselves. Which is all well and good, but the article is technically correct in that exactly simulating n-body systems is currently out of reach (in all problem domains) if n is huge.

semi-extrinsic 11y ago
Sure, there's a limit to how large you can go, but people have recently been doing DFT e.g. for the interaction between DNA bases and graphene sheets, which is quite respectable. Quantum computing may offer better scaling at large n, but since quantum computing is still largely unproven in practice, and in particular since quantum hardware gets much harder to design and implement as you increase the system size, it's quite probable that classical computing will remain faster for the next 50 or 100 years. You can compare it with matrix multiplication: we know of several algorithms that are better than the naïve O(n^3), but they're essentially never used in practice.
- kolinko 11y ago
  Correct me if I'm wrong, but interaction with graphene sheets might me relatively simple due to a simple structure of those. Once we begin to deal with anything more complicated (e.g. two enzymes, protein folding, etc) we very quickly reach a boundary of what traditional computers can do.
  - semi-extrinsic 11y ago
    Well, you have to keep your tounge straight here. We're a very very far way from doing DFT of protein folding, that's correct. But we have little reason to believe that DFT is useful for studying the interactions of macromolecules; these interactions are dominated by typical van der Waals forces, along with repulsion and electrostatics, and this doesn't need a quantum mechanical treatment. Then you're in the world of molecular dynamics etc., where we have no problems simulating two enzymes interacting. See this paper where they did fully atomistic simulations of an entire HIV capsule back in 2013: http://www.nature.com/nature/journal/v497/n7451/full/nature12162.html http://www.nature.com/nature/journal/v497/n7451/full/nature1...

semi-extrinsic 11y ago

Sure, there's a limit to how large you can go, but people have recently been doing DFT e.g. for the interaction between DNA bases and graphene sheets, which is quite respectable. Quantum computing may offer better scaling at large n, but since quantum computing is still largely unproven in practice, and in particular since quantum hardware gets much harder to design and implement as you increase the system size, it's quite probable that classical computing will remain faster for the next 50 or 100 years. You can compare it with matrix multiplication: we know of several algorithms that are better than the naïve O(n^3), but they're essentially never used in practice.

kolinko 11y ago
Correct me if I'm wrong, but interaction with graphene sheets might me relatively simple due to a simple structure of those. Once we begin to deal with anything more complicated (e.g. two enzymes, protein folding, etc) we very quickly reach a boundary of what traditional computers can do.
- semi-extrinsic 11y ago
  Well, you have to keep your tounge straight here. We're a very very far way from doing DFT of protein folding, that's correct. But we have little reason to believe that DFT is useful for studying the interactions of macromolecules; these interactions are dominated by typical van der Waals forces, along with repulsion and electrostatics, and this doesn't need a quantum mechanical treatment. Then you're in the world of molecular dynamics etc., where we have no problems simulating two enzymes interacting. See this paper where they did fully atomistic simulations of an entire HIV capsule back in 2013: http://www.nature.com/nature/journal/v497/n7451/full/nature12162.html http://www.nature.com/nature/journal/v497/n7451/full/nature1...