AFTD Auto-Formalizing Theoretical Domains

ose_spherical_shrink_scalar_bounds

theoremverified

Under the small-s parameter assumptions for the spherical two-layer construction, the two radial scaling factors have the ordering and cross-distance bounds required to attain distortion q.

Statement

theorem ose_spherical_shrink_scalar_bounds (q r s : ℝ) (hq : 0 < q) (hq2 : 2 ≤ q ^ 2) (hr : 0 < r) (hs : 0 < s) (hid : s ^ 2 + 2 * r * s = 1) (hsmall : s * q ≤ 1) : let α := s * q; let β := min q (1 / α); 1 ≤ q ∧ 0 < α ∧ α ≤ 1 ∧ 1 ≤ β ∧ β ≤ q ∧ α * β ≤ 1 ∧ (β - α)^2 ≤ 4 * r^2 ∧ r^2 + q^2 * (1 - α * β) ≤ q^2 * r^2
Source
arXiv:2611.08546, Section 6.1; scalar estimates for the proposed spherical construction
Verified
15 Nov 2026
Axioms
Classical.choiceQuot.soundpropext
Built on 1
  • ose_cube_scalar_certificateFor real p at least two and u between zero and one, four times p cubed times u squared times one minus u is at most (p minus one) times (p…
Used by 1

The proof, explained

The theorem verifies eight ordering and cross-distance bounds on the radial scaling factors $\\alpha = sq$ and $\\beta = \\min(q, 1/\\alpha)$ under the spherical parameter constraint $s^2 + 2rs = 1$. The first seven bounds follow by direct algebraic inequalities. The proof then splits on whether $sq^2 \\ge 1$: when $sq^2 \\ge 1$, $\\beta = 1/\\alpha$ and the final distortion bound trivially collapses to $r^2 \\le q^2 r^2$; when $sq^2 \\le 1$, $\\beta = q$ and the bound follows by substituting $p = q^2, u = sq^2$ into the cubic certificate inequality $4p^3 u^2(1-u) \\le (p-1)(p^2-u^2)^2$ via the identity $p^2 - u^2 = 2pur$.

  1. $1 \\le q$ and the radial scale $\\alpha = sq$ is strictly positive.

    We record the cubic certificate lemma `ose_cube_scalar_certificate` and unfold the definitions of $\\alpha = sq$ and $\\beta = \\min(q, 1/\\alpha)$. From $q > 0$ and $q^2 \\ge 2$, it follows that $1 \\le q$. Since $s > 0$ and $q > 0$, their product $\\alpha = sq$ is strictly positive by `mul_pos`.

    lines 2–6

    uses ose_cube_scalar_certificate, mul_pos

      have hcert (p u : ℝ) (hp : 2 ≤ p) (hu0 : 0 ≤ u) (hu1 : u ≤ 1) :=
        ose_cube_scalar_certificate p u hp hu0 hu1
      dsimp only
      have hq1 : 1 ≤ q := by nlinarith
      have ha : 0 < s*q := mul_pos hs hq
  2. The radial scale $\\beta = \\min(q, 1/\\alpha)$ satisfies $1 \\le \\beta \\le q$, $\\beta \\le 1/\\alpha$, and $\\alpha\\beta \\le 1$.

    Since $1 \\le q$ and the assumption $sq \\le 1$ implies $1 \\le 1/(sq)$ via `le_div_iff₀`, we obtain $1 \\le \\beta$ by `le_min`. The definition of the minimum immediately gives $\\beta \\le q$ and $\\beta \\le 1/(sq)$ via `min_le_left` and `min_le_right`. Multiplying $\\beta \\le 1/\\alpha$ by $\\alpha > 0$ yields $\\alpha\\beta \\le \\alpha \\cdot (1/\\alpha) = 1$ using `mul_le_mul_of_nonneg_left` and `field_simp` with `ne_of_gt`.

    lines 7–14

    uses le_min, le_div_iff₀, min_le_left, min_le_right, mul_le_mul_of_nonneg_left, le_of_lt, ne_of_gt

      have hbeta1 : 1 ≤ min q (1 / (s*q)) :=
        le_min hq1 ((le_div_iff₀ ha).2 (by simpa using hsmall))
      have hbetaq : min q (1 / (s*q)) ≤ q := min_le_left _ _
      have hbetainv : min q (1 / (s*q)) ≤ 1 / (s*q) := min_le_right _ _
      have hab : (s*q) * min q (1 / (s*q)) ≤ 1 := by
        calc (s*q) * min q (1 / (s*q)) ≤ (s*q) * (1 / (s*q)) :=
              mul_le_mul_of_nonneg_left hbetainv (le_of_lt ha)
          _ = 1 := by field_simp [ne_of_gt ha]
  3. The difference between the two scales satisfies $(\\beta - \\alpha)^2 \\le 4r^2$.

    Multiplying the sphere condition $s^2 + 2rs = 1$ by $q$ and combining with $(q - 1)(1 + qs^2) \\ge 0$ yields $1 \\le sq(2r + sq)$. Dividing by $sq > 0$ gives $1/(sq) \\le 2r + sq$, so $\\beta - \\alpha \\le 1/\\alpha - \\alpha \\le 2r$. Because $\\alpha \\le 1 \\le \\beta$, the difference $\\beta - \\alpha$ is nonnegative. Squaring $0 \\le \\beta - \\alpha \\le 2r$ via `pow_le_pow_left₀` proves $(\\beta - \\alpha)^2 \\le 4r^2$.

    lines 15–24

    uses mul_nonneg, congrArg, div_le_iff₀, mul_comm, pow_le_pow_left₀

      have hsec : 1 ≤ (s*q)*(2*r+s*q) := by
        have hp := mul_nonneg (show 0 ≤ q-1 by linarith) (show 0 ≤ 1+q*s^2 by positivity)
        have hmul := congrArg (fun z : ℝ => q * z) hid
        nlinarith [hp, hmul]
      have hinv : 1 / (s*q) ≤ 2*r+s*q := (div_le_iff₀ ha).2 (by simpa [mul_comm] using hsec)
      have hdiff0 : 0 ≤ min q (1 / (s*q)) - s*q := by linarith
      have hdiff : min q (1 / (s*q)) - s*q ≤ 2*r := by linarith
      have hdiffsq : (min q (1 / (s*q)) - s*q)^2 ≤ 4*r^2 := by
        have h := pow_le_pow_left₀ hdiff0 hdiff 2
        nlinarith
  4. The first seven conditions hold, reducing the goal to the final distortion inequality $r^2 + q^2(1 - \\alpha\\beta) \\le q^2 r^2$.

    We supply the proofs of $1 \\le q$, $0 < \\alpha$, $\\alpha \\le 1$, $1 \\le \\beta$, $\\beta \\le q$, $\\alpha\\beta \\le 1$, and $(\\beta - \\alpha)^2 \\le 4r^2$, leaving only $r^2 + q^2(1 - \\alpha\\beta) \\le q^2 r^2$ to be verified.

    line 25

      refine ⟨hq1, ha, hsmall, hbeta1, hbetaq, hab, hdiffsq, ?_⟩
  5. If $1 \\le sq^2$, the final distortion inequality $r^2 + q^2(1 - \\alpha\\beta) \\le q^2 r^2$ holds.

    When $1 \\le sq^2$, we have $1/(sq) \\le q$, so $\\beta = 1/(sq)$ by `min_eq_right` and `div_le_iff₀`. Then $\\alpha\\beta = 1$, reducing the inequality to $r^2 \\le q^2 r^2$, which holds because $q^2 \\ge 2 \\ge 1$ and $r^2 \\ge 0$ by `sq_nonneg`.

    lines 26–34

    uses min_eq_right, div_le_iff₀, ne_of_gt, sq_nonneg

      by_cases hu : 1 ≤ s*q^2
      · have hb : min q (1 / (s*q)) = 1 / (s*q) := by
          apply min_eq_right
          apply (div_le_iff₀ ha).2
          nlinarith [hu]
        rw [hb]
        have hprod : (s*q) * (1 / (s*q)) = 1 := by field_simp [ne_of_gt ha]
        rw [hprod]
        nlinarith [sq_nonneg r]
  6. If $sq^2 \\le 1$, the final distortion inequality $r^2 + q^2(1 - \\alpha\\beta) \\le q^2 r^2$ holds.

    When $sq^2 \\le 1$, we have $q \\le 1/(sq)$, so $\\beta = q$ by `min_eq_left` and $\\alpha\\beta = sq^2$. Setting $p = q^2$ and $u = sq^2 = sp$, we have $p \\ge 2$ and $0 < u \\le 1$. Using the spherical relation $1 - s^2 = 2rs$, we compute $p^2 - u^2 = p^2(1 - s^2) = 2pur$. Rewriting this into the cubic certificate $4p^3 u^2(1 - u) \\le (p - 1)(p^2 - u^2)^2$ and cancelling the positive factor $4p^2 u^2$ via `le_of_mul_le_mul_left` gives $p(1 - u) \\le (p - 1)r^2$, which rearranges to $r^2 + q^2(1 - sq^2) \\le q^2 r^2$.

    lines 35–59

    uses min_eq_left, le_div_iff₀, mul_pos, le_of_lt, le_of_mul_le_mul_left

      · have hu_le : s*q^2 ≤ 1 := by linarith
        have hb : min q (1 / (s*q)) = q := by
          apply min_eq_left
          apply (le_div_iff₀ ha).2
          nlinarith [hu_le]
        rw [hb]
        let p : ℝ := q^2
        let u : ℝ := s*p
        have hp : 2 ≤ p := hq2
        have hp0 : 0 < p := by linarith
        have hu0 : 0 < u := mul_pos hs hp0
        have hu1 : u ≤ 1 := by dsimp [u,p]; exact hu_le
        have hcert' := hcert p u hp (le_of_lt hu0) hu1
        have hpuid : p^2-u^2 = 2*p*u*r := by
          have hrel : 1-s^2 = 2*r*s := by nlinarith [hid]
          calc p^2-u^2 = (q^2)^2 * (1-s^2) := by dsimp [p,u]; ring
            _ = (q^2)^2 * (2*r*s) := by rw [hrel]
            _ = 2*p*u*r := by dsimp [p,u]; ring
        rw [hpuid] at hcert'
        have hf : 0 < 4*p^2*u^2 := by positivity
        have hfactor : (4*p^2*u^2) * (p*(1-u)) ≤
            (4*p^2*u^2) * ((p-1)*r^2) := by nlinarith [hcert']
        have hresult := le_of_mul_le_mul_left hfactor hf
        dsimp [p,u] at hresult
        nlinarith [hresult]
ose_cube_scalar_certificate
For real numbers $p \\ge 2$ and $0 \\le u \\le 1$, $4p^3 u^2(1 - u) \\le (p - 1)(p^2 - u^2)^2$.
mul_pos
If $a > 0$ and $b > 0$, then $a \\cdot b > 0$.
le_min
If $c \\le a$ and $c \\le b$, then $c \\le \\min(a, b)$.
le_div_iff₀
For $c > 0$, $a \\le b / c \\iff a \\cdot c \\le b$.
min_le_left
For any $a$ and $b$, $\\min(a, b) \\le a$.
min_le_right
For any $a$ and $b$, $\\min(a, b) \\le b$.
mul_le_mul_of_nonneg_left
If $b \\le c$ and $0 \\le a$, then $a \\cdot b \\le a \\cdot c$.
le_of_lt
If $a < b$, then $a \\le b$.
ne_of_gt
If $b < a$, then $a \\ne b$.
mul_nonneg
If $0 \\le a$ and $0 \\le b$, then $0 \\le a \\cdot b$.
congrArg
If $a_1 = a_2$, then $f(a_1) = f(a_2)$ for any function $f$.
div_le_iff₀
For $c > 0$, $b / c \\le a \\iff b \\le a \\cdot c$.
mul_comm
Multiplication is commutative: $a \\cdot b = b \\cdot a$.
pow_le_pow_left₀
If $0 \\le a \\le b$, then $a^n \\le b^n$ for every natural number $n$.
min_eq_right
If $b \\le a$, then $\\min(a, b) = b$.
min_eq_left
If $a \\le b$, then $\\min(a, b) = a$.
sq_nonneg
For any real number $a$, $0 \\le a^2$.
le_of_mul_le_mul_left
If $a \\cdot b \\le a \\cdot c$ and $a > 0$, then $b \\le c$.

Written by a model (gemini-3.8-flash) from the Lean proof and the goal states Lean reported at each step. Lean verified the proof; this explanation is a reading aid and is not itself machine-checked.

Lean source view module on GitHub

theorem ose_spherical_shrink_scalar_bounds (q r s : ℝ) (hq : 0 < q) (hq2 : 2 ≤ q ^ 2) (hr : 0 < r) (hs : 0 < s) (hid : s ^ 2 + 2 * r * s = 1) (hsmall : s * q ≤ 1) : let α := s * q; let β := min q (1 / α); 1 ≤ q ∧ 0 < α ∧ α ≤ 1 ∧ 1 ≤ β ∧ β ≤ q ∧ α * β ≤ 1 ∧ (β - α)^2 ≤ 4 * r^2 ∧ r^2 + q^2 * (1 - α * β) ≤ q^2 * r^2 := by  have hcert (p u : ℝ) (hp : 2 ≤ p) (hu0 : 0 ≤ u) (hu1 : u ≤ 1) :=    ose_cube_scalar_certificate p u hp hu0 hu1  dsimp only  have hq1 : 1 ≤ q := by nlinarith  have ha : 0 < s*q := mul_pos hs hq  have hbeta1 : 1 ≤ min q (1 / (s*q)) :=    le_min hq1 ((le_div_iff₀ ha).2 (by simpa using hsmall))  have hbetaq : min q (1 / (s*q)) ≤ q := min_le_left _ _  have hbetainv : min q (1 / (s*q)) ≤ 1 / (s*q) := min_le_right _ _  have hab : (s*q) * min q (1 / (s*q)) ≤ 1 := by    calc (s*q) * min q (1 / (s*q)) ≤ (s*q) * (1 / (s*q)) :=          mul_le_mul_of_nonneg_left hbetainv (le_of_lt ha)      _ = 1 := by field_simp [ne_of_gt ha]  have hsec : 1 ≤ (s*q)*(2*r+s*q) := by    have hp := mul_nonneg (show 0 ≤ q-1 by linarith) (show 0 ≤ 1+q*s^2 by positivity)    have hmul := congrArg (fun z : ℝ => q * z) hid    nlinarith [hp, hmul]  have hinv : 1 / (s*q) ≤ 2*r+s*q := (div_le_iff₀ ha).2 (by simpa [mul_comm] using hsec)  have hdiff0 : 0 ≤ min q (1 / (s*q)) - s*q := by linarith  have hdiff : min q (1 / (s*q)) - s*q ≤ 2*r := by linarith  have hdiffsq : (min q (1 / (s*q)) - s*q)^2 ≤ 4*r^2 := by    have h := pow_le_pow_left₀ hdiff0 hdiff 2    nlinarith  refine ⟨hq1, ha, hsmall, hbeta1, hbetaq, hab, hdiffsq, ?_⟩  by_cases hu : 1 ≤ s*q^2  · have hb : min q (1 / (s*q)) = 1 / (s*q) := by      apply min_eq_right      apply (div_le_iff₀ ha).2      nlinarith [hu]    rw [hb]    have hprod : (s*q) * (1 / (s*q)) = 1 := by field_simp [ne_of_gt ha]    rw [hprod]    nlinarith [sq_nonneg r]  · have hu_le : s*q^2 ≤ 1 := by linarith    have hb : min q (1 / (s*q)) = q := by      apply min_eq_left      apply (le_div_iff₀ ha).2      nlinarith [hu_le]    rw [hb]    let p : ℝ := q^2    let u : ℝ := s*p    have hp : 2 ≤ p := hq2    have hp0 : 0 < p := by linarith    have hu0 : 0 < u := mul_pos hs hp0    have hu1 : u ≤ 1 := by dsimp [u,p]; exact hu_le    have hcert' := hcert p u hp (le_of_lt hu0) hu1    have hpuid : p^2-u^2 = 2*p*u*r := by      have hrel : 1-s^2 = 2*r*s := by nlinarith [hid]      calc p^2-u^2 = (q^2)^2 * (1-s^2) := by dsimp [p,u]; ring        _ = (q^2)^2 * (2*r*s) := by rw [hrel]        _ = 2*p*u*r := by dsimp [p,u]; ring    rw [hpuid] at hcert'    have hf : 0 < 4*p^2*u^2 := by positivity    have hfactor : (4*p^2*u^2) * (p*(1-u)) ≤        (4*p^2*u^2) * ((p-1)*r^2) := by nlinarith [hcert']    have hresult := le_of_mul_le_mul_left hfactor hf    dsimp [p,u] at hresult    nlinarith [hresult]
See it in the graphDiscuss this resultChallenge it