AFTD Auto-Formalizing Theoretical Domains

op_1560_entropy_upper_bound_of_half_integral

theoremverified

For any finite simple graph G with orientation o and any half-integral stationary current j on G.edgeFinset with sum of |j(e)| equal to the number of vertices n, and any 0 < J < n, the infimal stationary entropy production S_G(J) is at most the number of edges with |j(e)| = 1 times phi(J/n) plus the number of edges with |j(e)| = 1/2 times phi(J/(2n)).

Statement

theorem op_1560_entropy_upper_bound_of_half_integral
    {V : Type*} [Fintype V] [DecidableEq V]
    (G : SimpleGraph V) [DecidableRel G.Adj]
    (o : SimpleGraphOrientation G)
    (j : G.edgeFinset → ℝ)
    (hj_stat : is_stationary_current G o j)
    (hj_half : is_half_integral_current G j)
    (hj_sum : ∑ e : G.edgeFinset, |j e| = Fintype.card V)
    (J : ℝ) (hJ_pos : 0 < J) (hJ_lt : J < Fintype.card V) :
    cubic_stationary_entropy_inf G o J ≤
      (∑ e : G.edgeFinset, if |j e| = 1 then (1 : ℝ) else 0) *
        cubic_edge_entropy (J / Fintype.card V) +
      (∑ e : G.edgeFinset, if |j e| = 1/2 then (1 : ℝ) else 0) *
        cubic_edge_entropy (J / (2 * Fintype.card V))
Source
arXiv:2204.00875
Verified
13 Nov 2026
Axioms
Classical.choiceQuot.soundpropext
Built on 5
  • cubic_stationary_entropy_infFor 0 < J < n, S_G(J) is the infimum of sum_e phi(j_e) among stationary currents j with |j_e| < 1 and sum_e |j_e| = J.
  • cubic_edge_entropyThe entropy production contribution phi(x) of an edge carrying stationary current x under unit total forward-plus-reverse flux is |x| *…
  • is_half_integral_currentFor a finite simple graph G with decidable equality on vertices and decidable adjacency, a function j on the edge finset of G into the…
  • is_stationary_currentA current j on the edges of an oriented finite simple graph G is stationary if at every vertex, the sum of currents leaving the vertex…
  • SimpleGraphOrientationAn orientation of a finite simple graph G assigns to each edge an ordered pair of its endpoints.

Read back from the Lean

Let $V$ be a finite type with decidable equality, and let $G$ be a simple graph on $V$ with decidable adjacency. Let $o$ be an orientation of $G$ (assigning to each edge $e$ an ordered pair $(u, v)$ with $\{u, v\} = e$). Let $j : G\text{.edgeFinset} \to \mathbb{R}$ be a real-valued edge current such that: 1. $j$ is stationary with respect to $o$ (the net flow at every vertex $v \in V$ is zero: $\sum_{e \in G\text{.edgeFinset}} (\mathbf{1}_{(o(e))_1 = v} - \mathbf{1}_{(o(e))_2 = v}) j(e) = 0$); 2. $j$ is half-integral (for all $e \in G\text{.edgeFinset}$, $j(e) \in \{0, 1/2, -1/2, 1, -1\}$); 3. the total absolute current is $|V|$: $\sum_{e \in G\text{.edgeFinset}} |j(e)| = |V|$. Let $J \in \mathbb{R}$ satisfy $0 < J < |V|$. Then the infimum cubic stationary entropy of all stationary currents on $(G, o)$ with edge values strictly in $(-1, 1)$ and total absolute sum $J$, denoted $\text{cubic\_stationary\_entropy\_inf}(G, o, J)$, is bounded above by: $$|\{e \in G\text{.edgeFinset} : |j(e)| = 1\}| \cdot \text{cubic\_edge\_entropy}\left(\frac{J}{|V|}\right) + |\{e \in G\text{.edgeFinset} : |j(e)| = 1/2\}| \cdot \text{cubic\_edge\_entropy}\left(\frac{J}{2|V|}\right)$$ where $\text{cubic\_edge\_entropy}(x) = |x| \ln\left(\frac{1 + |x|}{1 - |x|}\right)$.

Written by a model that saw only the Lean, never the English above. If the two disagree, that is worth a challenge.

Lean source view module on GitHub

private lemma op_1560_entropy_upper_bound_of_half_integral_neg (x : ℝ) :    cubic_edge_entropy (-x) = cubic_edge_entropy x := by  simp [cubic_edge_entropy, abs_neg]private lemma op_1560_entropy_upper_bound_of_half_integral_nonneg {x : ℝ} (hx : |x| < 1) :    0 ≤ cubic_edge_entropy x := by  unfold cubic_edge_entropy  apply mul_nonneg (abs_nonneg x)  apply Real.log_nonneg  rw [le_div_iff₀ (by linarith : (0:ℝ) < 1 - |x|)]  have := abs_nonneg x  linarithprivate lemma op_1560_entropy_upper_bound_of_half_integral_edge_scale {t : ℝ} (_ht0 : 0 < t) (x : ℝ)    (hx : x = 0 ∨ x = 1/2 ∨ x = -1/2 ∨ x = 1 ∨ x = -1) :    cubic_edge_entropy (t * x) =      (if |x| = 1 then (1:ℝ) else 0) * cubic_edge_entropy t +      (if |x| = 1/2 then (1:ℝ) else 0) * cubic_edge_entropy (t/2) := by  rcases hx with h|h|h|h|h  · subst h; simp [cubic_edge_entropy]  · subst h    have h1 : cubic_edge_entropy (t * (1/2:ℝ)) = cubic_edge_entropy (t/2) := by congr 1; ring    rw [h1]; norm_num  · subst h    have h1 : cubic_edge_entropy (t * (-1/2:ℝ)) = cubic_edge_entropy (t/2) := by      have hh : t * (-1/2:ℝ) = -(t/2) := by ring      rw [hh, op_1560_entropy_upper_bound_of_half_integral_neg]    rw [h1]; norm_num  · subst h; simp [cubic_edge_entropy]  · subst h    have h1 : cubic_edge_entropy (t * (-1:ℝ)) = cubic_edge_entropy t := by      have hh : t * (-1:ℝ) = -t := by ring      rw [hh, op_1560_entropy_upper_bound_of_half_integral_neg]    rw [h1]; norm_num/-- Scaling a half-integral stationary current provides an explicit upper bound on the continuous entropy infimum. -/theorem op_1560_entropy_upper_bound_of_half_integral    {V : Type*} [Fintype V] [DecidableEq V]    (G : SimpleGraph V) [DecidableRel G.Adj]    (o : SimpleGraphOrientation G)    (j : G.edgeFinset → ℝ)    (hj_stat : is_stationary_current G o j)    (hj_half : is_half_integral_current G j)    (hj_sum : ∑ e : G.edgeFinset, |j e| = Fintype.card V)    (J : ℝ) (hJ_pos : 0 < J) (hJ_lt : J < Fintype.card V) :    cubic_stationary_entropy_inf G o J ≤      (∑ e : G.edgeFinset, if |j e| = 1 then (1 : ℝ) else 0) *        cubic_edge_entropy (J / Fintype.card V) +      (∑ e : G.edgeFinset, if |j e| = 1/2 then (1 : ℝ) else 0) *        cubic_edge_entropy (J / (2 * Fintype.card V)) := by  classical  set n : ℝ := (Fintype.card V : ℝ) with hn_def  have hn_pos : 0 < n := by    rw [hn_def]    by_contra h    have : (Fintype.card V : ℝ) ≤ 0 := le_of_not_gt h    linarith  have hn_ne : n ≠ 0 := ne_of_gt hn_pos  set t : ℝ := J / n with ht_def  have ht_pos : 0 < t := by rw [ht_def]; exact div_pos hJ_pos hn_pos  have ht_lt : t < 1 := by    rw [ht_def, div_lt_one hn_pos]    simpa using hJ_lt  let j' : G.edgeFinset → ℝ := fun e => t * j e  have hstat : is_stationary_current G o j' := by    intro v    have hv := hj_stat v    have hterm : ∀ e : ↥G.edgeFinset,        ((if (o.dir e).1 = v then j' e else 0) - if (o.dir e).2 = v then j' e else 0)        = t * ((if (o.dir e).1 = v then j e else 0) - if (o.dir e).2 = v then j e else 0) := by      intro e      by_cases h1 : (o.dir e).1 = v <;> by_cases h2 : (o.dir e).2 = v <;>        simp [j', h1, h2]    calc ∑ e, ((if (o.dir e).1 = v then j' e else 0) - if (o.dir e).2 = v then j' e else 0)        = ∑ e, t * ((if (o.dir e).1 = v then j e else 0) - if (o.dir e).2 = v then j e else 0) :=          Finset.sum_congr rfl (fun e _ => hterm e)      _ = t * ∑ e, ((if (o.dir e).1 = v then j e else 0) - if (o.dir e).2 = v then j e else 0) := by          rw [Finset.mul_sum]      _ = t * 0 := by rw [hv]      _ = 0 := by ring  have hbound : ∀ e : G.edgeFinset, |j' e| < 1 := by    intro e    have he := hj_half e    have habs : |j e| ≤ 1 := by      rcases he with h|h|h|h|h <;> rw [h] <;> norm_num    have h1 : |j' e| = t * |j e| := by      simp [j', abs_mul, abs_of_pos ht_pos]    rw [h1]    calc t * |j e| ≤ t * 1 := mul_le_mul_of_nonneg_left habs (le_of_lt ht_pos)      _ = t := by ring      _ < 1 := ht_lt  have hsum : ∑ e : G.edgeFinset, |j' e| = J := by    have h1 : ∑ e : G.edgeFinset, |j' e| = ∑ e : G.edgeFinset, t * |j e| := by      apply Finset.sum_congr rfl; intro e _; simp [j', abs_mul, abs_of_pos ht_pos]    rw [h1, ← Finset.mul_sum, hj_sum]    rw [ht_def, div_mul_cancel₀ J hn_ne]  have hentropy_t : ∑ e : G.edgeFinset, cubic_edge_entropy (j' e) =      (∑ e : G.edgeFinset, if |j e| = 1 then (1:ℝ) else 0) * cubic_edge_entropy t +      (∑ e : G.edgeFinset, if |j e| = 1/2 then (1:ℝ) else 0) * cubic_edge_entropy (t/2) := by    have hper : ∀ e ∈ (Finset.univ : Finset ↥G.edgeFinset), cubic_edge_entropy (j' e) =        (if |j e| = 1 then (1:ℝ) else 0) * cubic_edge_entropy t +        (if |j e| = 1/2 then (1:ℝ) else 0) * cubic_edge_entropy (t/2) := by      intro e _      exact op_1560_entropy_upper_bound_of_half_integral_edge_scale ht_pos (j e) (hj_half e)    calc ∑ e : G.edgeFinset, cubic_edge_entropy (j' e)        = ∑ e : G.edgeFinset, ((if |j e| = 1 then (1:ℝ) else 0) * cubic_edge_entropy t +            (if |j e| = 1/2 then (1:ℝ) else 0) * cubic_edge_entropy (t/2)) :=          Finset.sum_congr rfl hper      _ = (∑ e : G.edgeFinset, (if |j e| = 1 then (1:ℝ) else 0) * cubic_edge_entropy t) +            (∑ e : G.edgeFinset, (if |j e| = 1/2 then (1:ℝ) else 0) * cubic_edge_entropy (t/2)) :=          Finset.sum_add_distrib      _ = (∑ e : G.edgeFinset, if |j e| = 1 then (1:ℝ) else 0) * cubic_edge_entropy t +            (∑ e : G.edgeFinset, if |j e| = 1/2 then (1:ℝ) else 0) * cubic_edge_entropy (t/2) := by          rw [Finset.sum_mul, Finset.sum_mul]  have hentropy : ∑ e : G.edgeFinset, cubic_edge_entropy (j' e) =      (∑ e : G.edgeFinset, if |j e| = 1 then (1:ℝ) else 0) * cubic_edge_entropy (J / n) +      (∑ e : G.edgeFinset, if |j e| = 1/2 then (1:ℝ) else 0) * cubic_edge_entropy (J / (2 * n)) := by    rw [hentropy_t]    have ht2 : t / 2 = J / (2 * n) := by rw [ht_def]; field_simp    rw [ht2, ht_def]  have hbdd : BddBelow {s : ℝ | ∃ jj : ↥G.edgeFinset → ℝ,      is_stationary_current G o jj ∧ (∀ e, |jj e| < 1) ∧ ∑ e, |jj e| = J ∧      ∑ e, cubic_edge_entropy (jj e) = s} := by    refine ⟨0, ?_⟩    rintro s ⟨jj, _, h1, _, rfl⟩    exact Finset.sum_nonneg (fun e _ => op_1560_entropy_upper_bound_of_half_integral_nonneg (h1 e))  have hmem : (∑ e : G.edgeFinset, cubic_edge_entropy (j' e)) ∈      {s : ℝ | ∃ jj : ↥G.edgeFinset → ℝ,        is_stationary_current G o jj ∧ (∀ e, |jj e| < 1) ∧ ∑ e, |jj e| = J ∧        ∑ e, cubic_edge_entropy (jj e) = s} :=    ⟨j', hstat, hbound, hsum, rfl⟩  rw [cubic_stationary_entropy_inf]  exact le_trans (csInf_le hbdd hmem) (le_of_eq hentropy)
See it in the graphDiscuss this resultChallenge it