The limitations of the Poincar{é} inequality
We examine the validity of the Poincaré inequality for degenerate, second-order, elliptic operators H in divergence form on $L_2(\Ri^{n}\times\Ri^{m})$. We assume the coefficients are real symmetric and a1Hδ≥H≥a2Hδ for some a1,a2>0 where Hδ is a generalized Grušin operator, Hδ=−∇x1∣x1∣(2δ1,2δ1′)∇x1−∣x1∣(2δ2,2δ2′)∇x22. Here $x_1\in\Ri^n$, $x_2\in\Ri^m$, δ1,δ1′∈[0,1⟩, δ2,δ2′≥0 and ∣x1∣(2δ,2δ′)=∣x1∣2δ if ∣x1∣≤1 and ∣x1∣(2δ,2δ′)=∣x1∣2δ′ if ∣x1∣≥1. \smallskip We prove that the Poincaré inequality, formulated in terms of the Riemannian geometry corresponding to H, is valid if n≥2, or if n=1 and δ1∨δ1′∈[0,1/2⟩ but it fails if n=1 and δ1∨δ1′∈[1/2,1⟩. The failure is caused by the leading term. If δ1∈[1/2,1⟩ it is an effect of the local degeneracy ∣x1∣2δ1 but if δ1∈[0,1/2⟩ and δ1′∈[1/2,1⟩ it is an effect of the growth at infinity of ∣x1∣2δ1′. If n=1 and δ1∈[1/2,1⟩ then the semigroup S generated by the Friedrichs' extension of H is not ergodic. The subspaces x1≥0 and x1≤0 are S-invariant and the Poincaré inequality is valid on each of these subspaces. If, however, n=1, δ1∈[0,1/2⟩ and δ1′∈[1/2,1⟩ then the semigroup S is ergodic but the Poincaré inequality is only valid locally. \smallskip Finally we discuss the implication of these results for the kernel of the semigroup S.