Griffiths monotonicity lemma
Statement (monotonicity in parameters)
Consider the ferromagnetic Ising model on a finite graph/finite region with spins and Hamiltonian
with ferromagnetic couplings and nonnegative external field (and any fixed boundary condition, typically plus).
Let denote expectation with respect to the associated finite-volume Gibbs measure .
For any finite set , define the spin product . Then:
For every edge ,
For every site ,
Equivalently, all correlation functions are nondecreasing functions of each coupling and each field in the ferromagnetic, regime.
Key hypotheses
- The model is ferromagnetic: for all interacting pairs.
- External field is nonnegative: (a common hypothesis ensuring the relevant correlation inequalities hold in the strongest form).
- Differentiability of finite-volume expectations in the parameters (automatic for finite systems).
Key conclusions
- Monotonicity of magnetization and higher correlations with respect to strengthening couplings/fields.
- Useful monotone limits as volume grows (e.g., with plus boundary conditions), supporting construction of infinite-volume Gibbs measures by monotone convergence.
- Parameter comparison and bounds for observables such as two-point correlations .
Cross-links to relevant definitions and tools
- The model: Ising model .
- Expectation as an integral w.r.t. a probability measure: expectation .
- Positivity inequalities typically used in the proof:
Proof idea (sketch)
Differentiate the expectation in the parameter. For example,
Similarly, $$ \frac{\partial}{\partial J_{xy}}\langle \sigma_A\rangle_{J,h}
={} \operatorname{Cov}_{J,h}(\sigma_A,\sigma_x\sigma_y). $$
In the ferromagnetic regime with , Griffiths/GKS/FKG-type inequalities imply the relevant covariances are nonnegative, giving the stated monotonicity.