Abstract
We establish the satisfiability threshold for random k-sat for all k≥k0, with k0 an absolute constant. That is, there exists a limiting density αsat(k) such that a random k-sat formula of clause density α is with high probability satisfiable for α < αsat, and unsatisfiable for α > αsat. We show that the threshold αsat(k) is given explicitly by the one-step replica symmetry breaking prediction from statistical physics. The proof develops a new analytic method for moment calculations on random graphs, mapping a high-dimensional optimization problem to a more tractable problem of analyzing tree recursions.
Original language | English (US) |
---|---|
Pages (from-to) | 1-388 |
Number of pages | 388 |
Journal | Annals of Mathematics |
Volume | 196 |
Issue number | 1 |
DOIs | |
State | Published - Jul 2022 |
All Science Journal Classification (ASJC) codes
- Statistics and Probability
- Statistics, Probability and Uncertainty
Keywords
- Boolean satisfiability
- Random constraint satisfaction problem
- Replica symmetry breaking
- Spin glasses
- Warning propagation