DLR existence theorem
Statement
Let be a translation-invariant interaction (potential) on with finite single-spin space (e.g. ) and assume is uniformly absolutely summable (in particular, any finite-range interaction qualifies). Let be the associated Gibbs specification constructed from the lattice Hamiltonian .
Then the set of infinite-volume Gibbs measures consistent with is nonempty. Equivalently, there exists at least one probability measure on such that satisfies the DLR equation for every finite .
Moreover, for any fixed boundary condition and any increasing sequence of finite volumes , every weak limit point of the corresponding finite-volume Gibbs measures is an infinite-volume Gibbs measure for .
Key hypotheses
- Configuration space: with finite (or more generally compact Polish).
- Interaction: is translation-invariant and uniformly absolutely summable (e.g. finite range).
- Specification: is the Gibbs specification induced by (quasilocal, consistent, proper).
(Probability-theoretic background: is a probability measure on .)
Key conclusions
- Existence: , where denotes the set of satisfying the DLR equation .
- Compactness mechanism: the family has weakly convergent subnet/subsequence along any exhaustion, and each limit point is in .
- Equilibrium exists at all parameters: for every inverse temperature (absorbed into ) and external parameters appearing in , at least one infinite-volume equilibrium state exists.
Proof idea / significance (sketch)
For finite spin spaces, is compact in the product topology (Tychonoff), hence the space of probability measures on is weak-* compact. Fix a boundary condition and consider . Any sequence along admits weak limit points. Quasilocality/consistency of the specification implies that the DLR consistency relations pass to the limit for local test functions, yielding a measure that satisfies the DLR equations.
Conceptually: this theorem guarantees that the DLR formalism is not empty—there is always at least one infinite-volume Gibbs state to study for standard lattice interactions (e.g. the Ising model ).