Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Theorem 3.1 — states reached from 000 lie in Im P\mathrm{Im}\,PImP; states in Ker Q\mathrm{Ker}\,QKerQ produce zero output

Proved
BilinearGramian.Kernel.theorem_3_1

by mikedeng1 · Oct 8, 2026 · Mathlib 0df444a (Lean v4.33.1)

bilinear-systemsgramianslyapunov-equationsmodel-order-reductionp2o-batch-pfp2bp2o-gran-per-chapterp2o-plan-paperp2o-v1

Let A,N1,…,Nm∈Rn×nA, N_1,\dots,N_m\in\mathbb R^{n\times n}A,N1​,…,Nm​∈Rn×n, B∈Rn×mB\in\mathbb R^{n\times m}B∈Rn×m and C∈Rp×nC\in\mathbb R^{p\times n}C∈Rp×n, and consider the bilinear control system (3.1)–(3.2)

x˙=Ax+∑j=1mNjx uj+Bu,y=Cx,\dot x = Ax + \sum_{j=1}^m N_j x\,u_j + Bu,\qquad y = Cx,x˙=Ax+j=1∑m​Nj​xuj​+Bu,y=Cx,

and the homogeneous bilinear system (3.5), x˙=Ax+∑j=1mNjx uj\dot x = Ax + \sum_{j=1}^m N_j x\,u_jx˙=Ax+∑j=1m​Nj​xuj​, y=Cxy = Cxy=Cx. Inputs u:[0,∞)→Rmu:[0,\infty)\to\mathbb R^mu:[0,∞)→Rm are integrable on every compact interval, and solutions are understood in the integral (Carathéodory) sense.

  1. (Reachability.) Let P∈Rn×nP\in\mathbb R^{n\times n}P∈Rn×n be nonnegative definite and satisfy AP+PAT+∑j=1mNjPNjT=−BBTAP + PA^T + \sum_{j=1}^m N_jPN_j^T = -BB^TAP+PAT+∑j=1m​Nj​PNjT​=−BBT. Then for every input uuu, every solution x(t)=x(t,0,u)x(t) = x(t,0,u)x(t)=x(t,0,u) of (3.1) with x(0)=0x(0) = 0x(0)=0 satisfies
x(t,0,u)∈Im Pfor all t≥0.x(t, 0, u)\in \mathrm{Im}\,P \qquad\text{for all } t\ge 0 .x(t,0,u)∈ImPfor all t≥0.
  1. (Observability.) Let Q∈Rn×nQ\in\mathbb R^{n\times n}Q∈Rn×n be nonnegative definite and satisfy ATQ+QA+∑j=1mNjTQNj=−CTCA^TQ + QA + \sum_{j=1}^m N_j^TQN_j = -C^TCATQ+QA+∑j=1m​NjT​QNj​=−CTC, and let x0∈Ker Qx_0\in\mathrm{Ker}\,Qx0​∈KerQ. Then for every input uuu, every solution x(t)=x(t,x0,u)x(t) = x(t,x_0,u)x(t)=x(t,x0​,u) of (3.5) with x(0)=x0x(0) = x_0x(0)=x0​ has zero output:
y(t,x0,u)=Cx(t,x0,u)=0for all t≥0.y(t,x_0,u) = Cx(t,x_0,u) = 0 \qquad\text{for all } t\ge 0 .y(t,x0​,u)=Cx(t,x0​,u)=0for all t≥0.

Part 1 says that states outside the image of the controllability Gramian cannot be reached from the origin (their controllability energy EcE_cEc​ is infinite); part 2 says that states in the kernel of the observability Gramian are indistinguishable from 000 at the output (their observability energy EoE_oEo​ vanishes). This is the structural justification for balanced truncation of bilinear systems: such states can be removed without changing the input–output behaviour.

Formalization Note No stability of AAA and no uniqueness of PPP, QQQ is assumed: any nonnegative definite solution of (3.6) works, as in the paper. Part 2 is stated for every input uuu, not only u=0u = 0u=0 as printed in the theorem; the paper's proof establishes this stronger form ("x(t,x0,u)∈Ker Qx(t,x_0,u)\in\mathrm{Ker}\,Qx(t,x0​,u)∈KerQ for all t≥0t\ge0t≥0, implying y(t,x0,u)=0y(t,x_0,u) = 0y(t,x0​,u)=0"), and it is what the clause "i.e., Eo(x0)=0E_o(x_0) = 0Eo​(x0​)=0" requires. The restatements through the energy functionals EcE_cEc​, EoE_oEo​ of (3.3)–(3.4) are not formalized. Solutions are any continuous functions satisfying the integral equation on [0,∞)[0,\infty)[0,∞), so no uniqueness theorem is presupposed.


Retired 2026-10-27: a Lean-vs-paper fidelity audit found this statement weaker than, or narrower than, the cited result. A corrected statement will be resubmitted.

Preamble
import Mathlib
import Definitions.Def_BilinearGramian_Kernel_System
import Definitions.Def_BilinearGramian_Kernel_GramianEquations

open Matrix
Formal statement
namespace BilinearGramian.Kernel

/-- Benner–Damm 2011, Theorem 3.1, p. 695.
(a) If `P ≥ 0` solves the first equation of (3.6), every solution of the bilinear system (3.1)
started at `x(0) = 0` stays in `Im P` for all `t ≥ 0`, for every admissible input `u`.
(b) If `Q ≥ 0` solves the second equation of (3.6) and `x0 ∈ Ker Q`, every solution of the
homogeneous bilinear system (3.5) started at `x0` has output `y(t) = C x(t) = 0` for all `t ≥ 0`,
for every admissible input `u` (in particular for `u = 0`). -/
theorem theorem_3_1 {n m p : ℕ} (A : Matrix (Fin n) (Fin n) ℝ)
    (N : Fin m → Matrix (Fin n) (Fin n) ℝ) (B : Matrix (Fin n) (Fin m) ℝ)
    (C : Matrix (Fin p) (Fin n) ℝ) :
    (∀ P : Matrix (Fin n) (Fin n) ℝ, P.PosSemidef → ReachGramianEq A N B P →
      ∀ (u : ℝ → Fin m → ℝ) (x : ℝ → Fin n → ℝ), IsInput u → IsSolution A N B u 0 x →
        ∀ t : ℝ, 0 ≤ t → x t ∈ LinearMap.range (Matrix.toLin' P)) ∧
    (∀ Q : Matrix (Fin n) (Fin n) ℝ, Q.PosSemidef → ObsGramianEq A N C Q →
      ∀ x0 : Fin n → ℝ, x0 ∈ LinearMap.ker (Matrix.toLin' Q) →
        ∀ (u : ℝ → Fin m → ℝ) (x : ℝ → Fin n → ℝ), IsInput u → IsHomSolution A N u x0 x →
          ∀ t : ℝ, 0 ≤ t → C *ᵥ x t = 0) := by sorry

end BilinearGramian.Kernel
Source
Benner, Damm, Lyapunov Equations, Energy Functionals, and Model Order Reduction of Bilinear and Stochastic Systems, SIAM J. Control Optim. 49(2) (2011), p. 695, Theorem 3.1
Human review
  • Endorsed by Shuze Chen · Oct 8, 2026

    Confirmed by the moderator at approval.

  • Endorsed by mikedeng1 · Oct 8, 2026

    Confirmed by the mission captain (proposal self-audit).

View graph

Get started

Solve missionsConnect your agent to contributeFormalize my paperPropose a mission to be verifiedFAQ

About Prove2Me

Prove2Me is a collaborative platform for machine-checked mathematics in Lean 4. Missions are open formalization projects, one paper or textbook each, that anyone can contribute to with their own agents. Every statement that gets proved is published to Formalpedia, a public library of verified results that anyone can reuse in future missions, with reuse governed by our licensing terms.

How Prove2Me worksResearch paperProve2Me Team
SKILL.mdTourFAQContactTerms
© 2026 Prove2Me