Existence and uniqueness of stationary distributions for finite irreducible CTMC generators
ProvedClassicalGaps.ctmc_stationary_uniquecontinuous-timemarkov-chainsprobabilitystationary-distribution
Let be a finite square real matrix whose off-diagonal entries are nonnegative and whose rows sum to zero (a continuous-time Markov chain generator). If is irreducible, then there is a unique probability vector with for all , and this vector is strictly positive.
[Retired 2026-09-22: vacuous hypotheses.] Mathlib's Matrix.IsIrreducible carries entrywise nonnegativity as a field; combined with zero row sums and nonnegative off-diagonal entries it forces Q = 0, which is never irreducible, so this statement says nothing about CTMC generators. Superseded by ClassicalGaps.ctmc_stationary_unique_v2, which puts irreducibility on the off-diagonal rate matrix. Credit: reviewer flag, Shuze Chen, 2026-09-20.
Preamble
import Mathlib.Algebra.BigOperators.Group.Finset.Defs import Mathlib.LinearAlgebra.Matrix.Irreducible.Defs import Mathlib.Data.Real.Basic open scoped BigOperators
Formal statement
theorem ClassicalGaps.ctmc_stationary_unique {n : Type*} [Fintype n] [Nonempty n] [DecidableEq n]
(Q : Matrix n n ℝ) (hoff : ∀ i j, i ≠ j → 0 ≤ Q i j)
(hrowsum : ∀ i, ∑ j, Q i j = 0) (hirr : Q.IsIrreducible) :
∃! π : n → ℝ, (∀ i, 0 < π i) ∧ ∑ i, π i = 1 ∧ ∀ j, ∑ i, π i * Q i j = 0 := by sorrySource
J. R. Norris, Markov Chains, Cambridge University Press, 1997, Chapter 3 (continuous-time chains); see https://en.wikipedia.org/wiki/Stationary_distribution
Human review
Matrix.IsIrreducible(Mathlib.LinearAlgebra.Matrix.Irreducible.Defs, at c5ea003) carries entrywise nonnegativity as a field:nonneg : ∀ i j, 0 ≤ A i j, the diagonal included. Combined withhoffandhrowsumthat forcesQ i i = 0for every i, henceQ = 0. The quiver ofQ = 0has no edges, andIsSStronglyConnectedasks every ordered pair — including i to itself — for a path of positive length, sohirris unsatisfiable for everyn. The hypotheses are contradictory and the theorem holds vacuously, saying nothing about CTMC generators.Putting irreducibility on the off-diagonal rates says what the description means:
Equivalently, state it for
Q + c • 1withcchosen so that∀ i, -Q i i ≤ c.