Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Theorem 2, p. 16 — for ρ(P) ≤ 1, equilibria are unique on classes with ρ < 1, and on classes with ρ = 1 non-unique iff (35)–(36), forming the segment (37)

Proved
SaturatedNet.Uniqueness.theorem_2

by mikedeng1 · Oct 22, 2026 · Mathlib 0df444a (Lean v4.33.1)

financial-networksnetwork-equilibriump2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1perron-frobeniussaturationuniqueness

Let (P,w)(P, w)(P,w) be a network (P≥0P \ge 0P≥0, w≥0w \ge 0w≥0) with ρ(P)≤1\rho(P) \le 1ρ(P)≤1, and ccc any exogenous flow. Let Vl\mathcal V_lVl​ be a class of PPP (a strongly connected component of GP\mathcal G_PGP​) with principal block P(ll)P^{(ll)}P(ll), and fix a network equilibrium xxx. Consider the set of projections x′(l)x'^{(l)}x′(l) on Vl\mathcal V_lVl​ of all network equilibria x′x'x′ that coincide with xxx on the nodes upstream of Vl\mathcal V_lVl​ (the paper's "given (x(1),…,x(l−1))(x^{(1)}, \dots, x^{(l-1)})(x(1),…,x(l−1))"), and write

u(l)=c(l)+∑1≤i<lP(il)⊤x(i),uj(l)=cj+∑k∉VlPkjxk.u^{(l)} = c^{(l)} + \sum_{1 \le i < l} P^{(il)\top}x^{(i)}, \qquad u^{(l)}_j = c_j + \sum_{k \notin \mathcal V_l} P_{kj}x_k .u(l)=c(l)+1≤i<l∑​P(il)⊤x(i),uj(l)​=cj​+k∈/Vl​∑​Pkj​xk​.
  1. If ρ(P(ll))<1\rho(P^{(ll)}) < 1ρ(P(ll))<1, the projection is unique: the set is {x(l)}\{x^{(l)}\}{x(l)}.
  2. If ρ(P(ll))=1\rho(P^{(ll)}) = 1ρ(P(ll))=1, let p(l)p^{(l)}p(l) and π(l)\pi^{(l)}π(l) be right and left dominant eigenvectors of P(ll)P^{(ll)}P(ll) (nonnegative, nonzero, P(ll)p(l)=p(l)P^{(ll)}p^{(l)} = p^{(l)}P(ll)p(l)=p(l), π(l)⊤P(ll)=π(l)⊤\pi^{(l)\top}P^{(ll)} = \pi^{(l)\top}π(l)⊤P(ll)=π(l)⊤). The projection is non-unique if and only if
p(l)⊤u(l)=0andmin⁡i∈Vl{νi(l)πi(l)}+min⁡i∈Vl{wi−νi(l)πi(l)}>0p^{(l)\top}u^{(l)} = 0 \quad\text{and}\quad \min_{i \in \mathcal V_l}\Big\{\frac{\nu^{(l)}_i}{\pi^{(l)}_i}\Big\} + \min_{i \in \mathcal V_l}\Big\{\frac{w_i - \nu^{(l)}_i}{\pi^{(l)}_i}\Big\} > 0p(l)⊤u(l)=0andi∈Vl​min​{πi(l)​νi(l)​​}+i∈Vl​min​{πi(l)​wi​−νi(l)​​}>0

for a solution ν(l)\nu^{(l)}ν(l) of ν(l)=P(ll)⊤ν(l)+u(l)\nu^{(l)} = P^{(ll)\top}\nu^{(l)} + u^{(l)}ν(l)=P(ll)⊤ν(l)+u(l) ((35)–(36)). In that case, for every such solution ν(l)\nu^{(l)}ν(l), the set of projections is exactly

{ν(l)+απ(l): −min⁡i∈Vl{νi(l)πi(l)}≤α≤min⁡i∈Vl{wi−νi(l)πi(l)}}(37).\Big\{\nu^{(l)} + \alpha\pi^{(l)} :\ -\min_{i \in \mathcal V_l}\Big\{\frac{\nu^{(l)}_i}{\pi^{(l)}_i}\Big\} \le \alpha \le \min_{i \in \mathcal V_l}\Big\{\frac{w_i - \nu^{(l)}_i}{\pi^{(l)}_i}\Big\}\Big\} \qquad (37).{ν(l)+απ(l): −i∈Vl​min​{πi(l)​νi(l)​​}≤α≤i∈Vl​min​{πi(l)​wi​−νi(l)​​}}(37).

Applied class by class along the block-triangular order (14), this describes the whole lattice of equilibria X\mathcal XX of a network with ρ(P)≤1\rho(P) \le 1ρ(P)≤1: unique on the classes of spectral radius below one, and a segment, or a point, on each class of spectral radius one.

Formalization Note. The statement is permutation-free: instead of assuming PPP in block triangular form, it quantifies over a class VVV and fixes the equilibrium on the nodes upstream of VVV (outside VVV, with a path into VVV). Fixing every earlier class of a compatible ordering gives the same set of projections, since earlier classes that are not upstream neither feed nor are fed by VVV. The equilibria x′x'x′ range over equilibria of the whole network. The third bullet of the paper prints ν(l)=P(ll)⊤ν(l)+∑i=1lP(jl)⊤x(i)+c(l)\nu^{(l)} = P^{(ll)\top}\nu^{(l)} + \sum_{i=1}^{l} P^{(jl)\top}x^{(i)} + c^{(l)}ν(l)=P(ll)⊤ν(l)+∑i=1l​P(jl)⊤x(i)+c(l), a slip for the sum over i<li < li<l used here (compare (35) and (38)). As in Proposition 5, "any solution ν(l)\nu^{(l)}ν(l)" is an existential in the equivalence and a universal in (37), and (37) is stated as an equality of sets (the proof reduces to Proposition 5, (30)). The hypothesis ρ(P)≤1\rho(P) \le 1ρ(P)≤1 ensures every class falls under case 1 or 2.

Preamble
import Mathlib
import Definitions.Def_SaturatedNet_Uniqueness_Model
open Matrix
Formal statement
namespace SaturatedNet.Uniqueness

theorem theorem_2 {n : Type*} [Fintype n] [DecidableEq n]
    (P : Matrix n n ℝ) (hP : ∀ i j, 0 ≤ P i j) (w : n → ℝ) (hw : 0 ≤ w) (c : n → ℝ) (hρ : specRad P ≤ 1)
    (V : Finset n) (hV : IsClass P V) (x : n → ℝ) (hx : x ∈ equilibria P w c) :
    (specRad (block P V) < 1 → classProj P w c V x = {fun i : V => x i}) ∧
    (specRad (block P V) = 1 →
      ∀ p π : V → ℝ, 0 ≤ p → p ≠ 0 → block P V *ᵥ p = p →
        0 ≤ π → π ≠ 0 → π ᵥ* block P V = π →
        ((classProj P w c V x).Nontrivial ↔
          (p ⬝ᵥ classInflow P c V x = 0 ∧
            ∃ ν : V → ℝ, ν = ν ᵥ* block P V + classInflow P c V x ∧
              minRatio ν π hV.univ_nonempty +
                minRatio (fun i : V => w i - ν i) π hV.univ_nonempty > 0)) ∧
        ((classProj P w c V x).Nontrivial →
          ∀ ν : V → ℝ, ν = ν ᵥ* block P V + classInflow P c V x →
            classProj P w c V x =
              {z | ∃ α : ℝ, -minRatio ν π hV.univ_nonempty ≤ α ∧
                α ≤ minRatio (fun i : V => w i - ν i) π hV.univ_nonempty ∧
                z = ν + α • π})) := by sorry

end SaturatedNet.Uniqueness
Source
Massai, Como & Fagnani, Equilibria and Systemic Risk in Saturated Networks, arXiv:1912.04815v2, p. 16, Theorem 2, (34)–(37)
Human review
  • Endorsed by Shuze Chen · Oct 22, 2026

    Confirmed by the moderator at approval.

  • Endorsed by mikedeng1 · Oct 22, 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