Master2019 Proposal

From Werner KRAUTH

(Difference between revisions)
Jump to: navigation, search
Revision as of 20:18, 2 February 2019
Werner (Talk | contribs)

← Previous diff
Revision as of 20:18, 2 February 2019
Werner (Talk | contribs)

Next diff →
Line 7: Line 7:
While the existence of a paramagnetic-ferromagnetic phase transition in the two-dimensional Ising model was proven some 80 years ago (Peierls, 1936), the rigorous proof of a phase transition in the hard-disk model (and more generally in hard-sphere models in dimension d>1) is still missing (the phase diagram of the hard-disk model was obtained in <1>). The same is true for general particle models. In fact, the most recent rigorous result for hard disks is due to Lebowitz and Penrose <2>, in 1964: It states that the liquid phase in this model exists. It extends at least from density =0 to density = 0.03619. (This is to be compared to the empirical value 069, obtained in <1>).) While the existence of a paramagnetic-ferromagnetic phase transition in the two-dimensional Ising model was proven some 80 years ago (Peierls, 1936), the rigorous proof of a phase transition in the hard-disk model (and more generally in hard-sphere models in dimension d>1) is still missing (the phase diagram of the hard-disk model was obtained in <1>). The same is true for general particle models. In fact, the most recent rigorous result for hard disks is due to Lebowitz and Penrose <2>, in 1964: It states that the liquid phase in this model exists. It extends at least from density =0 to density = 0.03619. (This is to be compared to the empirical value 069, obtained in <1>).)
-In 2003, however, [www.cs.yale.edu/homes/kannan/Papers/Hardcore.pdf| Kannan, Mahoney, and Montenegro] proved that simple Monte Carlo algorithms showed the property of `fast' mixing up to much larger densities (density = 1/8= 0.125), more than three times larger than those indicated by Lebowitz and Penrose. Fast mixing appears incompatible with any presence of order or slower than exponential correlation functions. The task of this practical is to rigorously prove this conjecture.+In 2003, however, [[www.cs.yale.edu/homes/kannan/Papers/Hardcore.pdf| Kannan, Mahoney, and Montenegro]] proved that simple Monte Carlo algorithms showed the property of `fast' mixing up to much larger densities (density = 1/8= 0.125), more than three times larger than those indicated by Lebowitz and Penrose. Fast mixing appears incompatible with any presence of order or slower than exponential correlation functions. The task of this practical is to rigorously prove this conjecture.
The "stage" will be located at LPENS, and supervised by Werner Krauth. A colloboration with the mathematicians C. Moore (Santa Fe Institute, USA) and T. Hayes (University of New Mexico, USA) appears likely. Moore and Hayes <4> recently proved an improved rapid-mixing bound, without obtaining a connection between the mixing times and the physical properties of the system. The "stage" will be located at LPENS, and supervised by Werner Krauth. A colloboration with the mathematicians C. Moore (Santa Fe Institute, USA) and T. Hayes (University of New Mexico, USA) appears likely. Moore and Hayes <4> recently proved an improved rapid-mixing bound, without obtaining a connection between the mixing times and the physical properties of the system.

Revision as of 20:18, 2 February 2019

Title: An improved bounding density for the liquid phase in hard-sphere systems.

Stage de M2 2019, ENS (Werner Krauth)

Summary

While the existence of a paramagnetic-ferromagnetic phase transition in the two-dimensional Ising model was proven some 80 years ago (Peierls, 1936), the rigorous proof of a phase transition in the hard-disk model (and more generally in hard-sphere models in dimension d>1) is still missing (the phase diagram of the hard-disk model was obtained in <1>). The same is true for general particle models. In fact, the most recent rigorous result for hard disks is due to Lebowitz and Penrose <2>, in 1964: It states that the liquid phase in this model exists. It extends at least from density =0 to density = 0.03619. (This is to be compared to the empirical value 069, obtained in <1>).)

In 2003, however, Kannan, Mahoney, and Montenegro proved that simple Monte Carlo algorithms showed the property of `fast' mixing up to much larger densities (density = 1/8= 0.125), more than three times larger than those indicated by Lebowitz and Penrose. Fast mixing appears incompatible with any presence of order or slower than exponential correlation functions. The task of this practical is to rigorously prove this conjecture.

The "stage" will be located at LPENS, and supervised by Werner Krauth. A colloboration with the mathematicians C. Moore (Santa Fe Institute, USA) and T. Hayes (University of New Mexico, USA) appears likely. Moore and Hayes <4> recently proved an improved rapid-mixing bound, without obtaining a connection between the mixing times and the physical properties of the system.

<1> Bernard and Krauth, PRL 2011

<2> Lebowitz and Penrose, JMP 1964

<4> Hayes and Moore, unpublished

Personal tools