AFTD Auto-Formalizing Theoretical Domains

ose_sphere_two_scale_distance_identity

theoremverified

For two vectors on the same sphere of squared radius q²/4, the squared distance after independent radial scalings beta and alpha equals (beta-alpha)²q²/4 plus alpha*beta times their original squared distance.

Statement

theorem ose_sphere_two_scale_distance_identity (H : Type) [NormedAddCommGroup H] [InnerProductSpace ℝ H] (x y : H) (α β q : ℝ) (hx : ‖x‖ ^ 2 = q ^ 2 / 4) (hy : ‖y‖ ^ 2 = q ^ 2 / 4) : ‖β • x - α • y‖ ^ 2 = (β - α) ^ 2 * q ^ 2 / 4 + α * β * ‖x - y‖ ^ 2
Source
arXiv:2611.08546, Section 6.1; spherical two-layer construction
Verified
15 Nov 2026
Axioms
Classical.choiceQuot.soundpropext
Used by 1

The proof, explained

The proof expands both squared distances, ‖x - y‖² and ‖β • x - α • y‖², in terms of vector norms and the inner product ⟨x, y⟩. Multiplying the identity for ‖x - y‖² by α * β aligns its inner product cross-term with that of ‖β • x - α • y‖². Substituting the sphere hypothesis ‖x‖² = ‖y‖² = q² / 4, the cross-terms cancel and the remaining terms simplify algebraically to (β - α)² * q² / 4.

  1. ‖x - y‖² = ‖x‖² - 2 ⟨x, y⟩ + ‖y‖²

    In a real inner product space, the square of the norm of the difference x - y expands into the sum of the squared norms minus twice their inner product.

    line 2

    uses norm_sub_sq_real

      have hxy := norm_sub_sq_real x y
  2. ‖β • x - α • y‖² = β² ‖x‖² - 2 * (α * (β * ⟨x, y⟩)) + α² ‖y‖²

    Expand ‖β • x - α • y‖² using the same difference formula, then simplify the scalar multiples: factor the scalars β and α out of the inner product by bilinearity, and rewrite ‖β • x‖² as |β|² ‖x‖² = β² ‖x‖² (and similarly for α • y).

    lines 3–4

    uses norm_sub_sq_real, real_inner_smul_left, real_inner_smul_right, norm_smul, Real.norm_eq_abs, mul_pow, sq_abs

      have hscaled := norm_sub_sq_real (β • x) (α • y)
      simp only [real_inner_smul_left, real_inner_smul_right, norm_smul, Real.norm_eq_abs, mul_pow, sq_abs] at hscaled
  3. (α * β) * ‖x - y‖² = (α * β) * (‖x‖² - 2 ⟨x, y⟩ + ‖y‖²)

    Multiply both sides of the expanded identity for ‖x - y‖² by α * β.

    line 5

    uses congrArg

      have hm := congrArg (fun z : ℝ => (α*β)*z) hxy
  4. ‖β • x - α • y‖² = (β - α)² * q² / 4 + α * β * ‖x - y‖² automation

    Substitute the sphere radius assumptions ‖x‖² = q² / 4 and ‖y‖² = q² / 4. The inner product cross-term -2 * α * β * ⟨x, y⟩ appears with the same coefficient in both expansions, so subtracting α * β * ‖x - y‖² cancels it, leaving (β² - 2 * α * β + α²) * (q² / 4) = (β - α)² * q² / 4, which is verified by nonlinear arithmetic.

    line 6

      nlinarith [hm]
norm_sub_sq_real
In a real inner product space, the squared norm of a difference expands as ‖x - y‖² = ‖x‖² - 2 ⟨x, y⟩ + ‖y‖².
real_inner_smul_left
The real inner product is homogeneous in its left argument: ⟨r • x, y⟩ = r * ⟨x, y⟩.
real_inner_smul_right
The real inner product is homogeneous in its right argument: ⟨x, r • y⟩ = r * ⟨x, y⟩.
norm_smul
The norm of a scalar multiple is the product of norms: ‖r • x‖ = ‖r‖ * ‖x‖.
Real.norm_eq_abs
For any real number r, its norm equals its absolute value |r|.
mul_pow
In a commutative monoid, taking powers distributes over multiplication: (a * b)ⁿ = aⁿ * bⁿ.
sq_abs
For any element a of an ordered ring, |a|² = a².
congrArg
If two inputs are equal, applying any function to both preserves equality: a₁ = a₂ implies f(a₁) = f(a₂).

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_sphere_two_scale_distance_identity (H : Type) [NormedAddCommGroup H] [InnerProductSpace ℝ H] (x y : H) (α β q : ℝ) (hx : ‖x‖ ^ 2 = q ^ 2 / 4) (hy : ‖y‖ ^ 2 = q ^ 2 / 4) : ‖β • x - α • y‖ ^ 2 = (β - α) ^ 2 * q ^ 2 / 4 + α * β * ‖x - y‖ ^ 2 := by  have hxy := norm_sub_sq_real x y  have hscaled := norm_sub_sq_real (β • x) (α • y)  simp only [real_inner_smul_left, real_inner_smul_right, norm_smul, Real.norm_eq_abs, mul_pow, sq_abs] at hscaled  have hm := congrArg (fun z : ℝ => (α*β)*z) hxy  nlinarith [hm]
See it in the graphDiscuss this resultChallenge it