the Lumer-Phillips theorem
/ LOO-mer FILL-ips /
Hille-Yosida tells you when an operator generates a contraction semigroup, but its resolvent-power condition can be awkward to check. The Lumer-Phillips theorem repackages the same conclusion into a condition that is often immediate to verify from the physics: just confirm that the operator never increases energy. If applying A can only push the energy of a state down (or hold it level), and one mild range condition holds, then A generates a semigroup that never grows — a contraction semigroup. It trades a quantitative resolvent estimate for a one-line energy inequality.
Precisely, work in a Hilbert space H with inner product. An operator A is dissipative if Re (A x, x) <= 0 for every x in D(A) — the rate of change of the squared norm, d/dt of (norm of u)^2 = 2 Re (A u, u), is never positive, so energy cannot grow. The Lumer-Phillips theorem states: a densely defined operator A generates a C_0 contraction semigroup if and only if A is dissipative AND the range of (lambda I - A) is all of H for some (hence every) lambda > 0. The second clause is just enough surjectivity to make the resolvent exist; in practice it follows from an elliptic existence theorem. So the recipe is: integrate by parts to show (A u, u) <= 0, then quote elliptic solvability for the range condition, and you are done.
This is the workhorse generation theorem for PDEs, because dissipativity is exactly what an integration-by-parts energy estimate delivers. For the Laplacian, (Laplacian u, u) = - integral of |grad u|^2 <= 0 falls right out of the divergence theorem with the boundary condition; the energy of heat can only decrease. Beyond contractions, the same idea handles unitary groups (skew-adjoint A, where energy is exactly conserved — the wave and Schrodinger cases), making Lumer-Phillips the unifying generation criterion across the equation types.
For A = Laplacian on L^2(Omega) with Dirichlet conditions and u in D(A): (A u, u) = integral of (Laplacian u) u = - integral of |grad u|^2 <= 0 by Green's identity (the boundary term vanishes since u = 0 there). So A is dissipative. Elliptic solvability of lambda u - Laplacian u = f gives the range condition. Lumer-Phillips then yields the heat contraction semigroup in two short lines.
Dissipativity (energy cannot grow) plus a range condition generates a contraction semigroup.
Dissipativity alone is not enough — the range/surjectivity condition is essential and is precisely where elliptic existence theory enters; skip it and you can have a dissipative operator that generates nothing. Also note dissipative gives contractions (no growth); to get genuine decay you additionally need a spectral gap or a Poincare-type strict inequality.