perspective_affine_hasDerivAt_zero
For Q0 > 0, the perspective t ↦ (Q0 + t Q1) f((C0 + t C1)/(Q0 + t Q1)) has derivative Q1 f(C0/Q0) + (C1 Q0 - C0 Q1) f'(C0/Q0)/Q0 at 0.
Statement
theorem perspective_affine_hasDerivAt_zero (f f1 : ℝ → ℝ) (Q0 Q1 C0 C1 : ℝ) (hQ : 0 < Q0) (hf : HasDerivAt f (f1 (C0 / Q0)) (C0 / Q0)) : HasDerivAt (fun t : ℝ => (Q0 + t * Q1) * f ((C0 + t * C1) / (Q0 + t * Q1))) (Q1 * f (C0 / Q0) + (C1 * Q0 - C0 * Q1) * f1 (C0 / Q0) / Q0) 0
- Source
- arXiv:2610.21231 (On the subadditivity of mutual f-information), Appendix B, proof of Lemma 4
- Verified
- 24 Oct 2026
- Axioms
Classical.choiceQuot.soundpropext- Used by 1
- mutual_f_theta_concaveOn_of_weight_boundLemma 4 of arXiv:2610.21231: under the assumptions of Theorem 1 (g = x² f'' with g'' < 0 and g + 2(g' - η)²/(-g'') ≤ 2η + (x-1)ℓ), for…
The proof, explained
The proof computes the derivative of the perspective function t ↦ (Q0 + t * Q1) * f((C0 + t * C1) / (Q0 + t * Q1)) at t = 0 by systematically applying standard calculus rules. First, the derivatives of the linear numerator and denominator at 0 are established, and the quotient rule yields the derivative of their ratio, which is valid because Q0 > 0. Next, the chain rule differentiates the composition with f, and the product rule combines this with the denominator factor. Finally, the resulting expression is simplified algebraically at t = 0 to match the stated formula.
The affine function t ↦ Q0 + t * Q1 has derivative Q1 at t = 0.
The identity function has derivative 1 everywhere by hasDerivAt_id'. Scaling by the constant Q1 and adding the constant Q0 shows that t ↦ Q0 + t * Q1 has derivative Q1 at 0.
uses
hasDerivAt_id'have hQt : HasDerivAt (fun t : ℝ => Q0 + t * Q1) Q1 0 := by simpa using ((hasDerivAt_id' (0:ℝ)).mul_const Q1).const_add Q0
The affine function t ↦ C0 + t * C1 has derivative C1 at t = 0.
Similarly, scaling the identity by C1 and adding C0 shows that t ↦ C0 + t * C1 has derivative C1 at 0.
uses
hasDerivAt_id'have hCt : HasDerivAt (fun t : ℝ => C0 + t * C1) C1 0 := by simpa using ((hasDerivAt_id' (0:ℝ)).mul_const C1).const_add C0
The quotient function t ↦ (C0 + t * C1) / (Q0 + t * Q1) has derivative (C1 * (Q0 + 0 * Q1) - (C0 + 0 * C1) * Q1) / (Q0 + 0 * Q1)^2 at t = 0.
Applying the quotient rule to the affine numerator and denominator gives the derivative of their ratio at 0. The denominator is nonzero at 0 because Q0 + 0 * Q1 = Q0 > 0.
have hdiv := hCt.fun_div hQt (by simp; exact hQ.ne')
The composite function t ↦ f((C0 + t * C1) / (Q0 + t * Q1)) has derivative f1(C0 / Q0) * ((C1 * (Q0 + 0 * Q1) - (C0 + 0 * C1) * Q1) / (Q0 + 0 * Q1)^2) at t = 0.
At t = 0, the quotient evaluates to C0 / Q0. Since f has derivative f1(C0 / Q0) at that point by assumption, the chain rule yields the derivative of the composition at 0.
have hf' : HasDerivAt f (f1 (C0 / Q0)) ((fun t : ℝ => (C0 + t * C1) / (Q0 + t * Q1)) 0) := by simpa using hf have hcomp := hf'.comp (h := fun t : ℝ => (C0 + t * C1) / (Q0 + t * Q1)) 0 hdiv
The product t ↦ (Q0 + t * Q1) * f((C0 + t * C1) / (Q0 + t * Q1)) has derivative Q1 * f(C0 / Q0) + (Q0 + 0 * Q1) * [f1(C0 / Q0) * ((C1 * (Q0 + 0 * Q1) - (C0 + 0 * C1) * Q1) / (Q0 + 0 * Q1)^2)] at t = 0.
Applying the product rule to the factor Q0 + t * Q1 and the composite factor f((C0 + t * C1) / (Q0 + t * Q1)) gives the derivative of their product at 0.
have := hQt.fun_mul hcompThe derivative of the perspective function at 0 is Q1 * f(C0 / Q0) + (C1 * Q0 - C0 * Q1) * f1(C0 / Q0) / Q0.
The proof finishes by replacing the derivative value obtained from the calculus rules with the algebraically simplified target form. Evaluating the zero terms reduces the second term's coefficient to Q0 * (C1 * Q0 - C0 * Q1) / Q0^2; since Q0 is nonzero, clearing the common factor of Q0 in the denominator leaves (C1 * Q0 - C0 * Q1) / Q0, which field_simp verifies.
uses
zero_mul,add_zero,Function.comp_apply,Pi.sub_apply,sub_selfexact this.congr_deriv (by simp only [zero_mul, add_zero, Function.comp_apply, Pi.sub_apply, sub_self] field_simp try ring)
hasDerivAt_id'- The identity function has derivative 1 at any point.
zero_mul- Multiplying any element by zero on the left yields zero: 0 * a = 0.
add_zero- Adding zero to any element on the right leaves it unchanged: a + 0 = a.
Function.comp_apply- The definition of function composition: (f ∘ g)(x) = f(g(x)).
Pi.sub_apply- Pointwise subtraction of functions evaluates as (f - g)(x) = f(x) - g(x).
sub_self- Subtracting any element from itself yields zero: a - a = 0.
- On line 17, 'try ring' is redundant because 'field_simp' on line 16 already completely simplifies the rational expressions and closes the goal.
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
Choose a tactic — a marked line — to see the goals Lean had before and after it, as in Lean’s infoview. Step with ← → or j k.
theorem perspective_affine_hasDerivAt_zero (f f1 : ℝ → ℝ) (Q0 Q1 C0 C1 : ℝ) (hQ : 0 < Q0) (hf : HasDerivAt f (f1 (C0 / Q0)) (C0 / Q0)) : HasDerivAt (fun t : ℝ => (Q0 + t * Q1) * f ((C0 + t * C1) / (Q0 + t * Q1))) (Q1 * f (C0 / Q0) + (C1 * Q0 - C0 * Q1) * f1 (C0 / Q0) / Q0) 0 := by have hQt : HasDerivAt (fun t : ℝ => Q0 + t * Q1) Q1 0 := by simpa using ((hasDerivAt_id' (0:ℝ)).mul_const Q1).const_add Q0 have hCt : HasDerivAt (fun t : ℝ => C0 + t * C1) C1 0 := by simpa using ((hasDerivAt_id' (0:ℝ)).mul_const C1).const_add C0 have hdiv := hCt.fun_div hQt (by simp; exact hQ.ne') have hf' : HasDerivAt f (f1 (C0 / Q0)) ((fun t : ℝ => (C0 + t * C1) / (Q0 + t * Q1)) 0) := by simpa using hf have hcomp := hf'.comp (h := fun t : ℝ => (C0 + t * C1) / (Q0 + t * Q1)) 0 hdiv have := hQt.fun_mul hcomp exact this.congr_deriv (by simp only [zero_mul, add_zero, Function.comp_apply, Pi.sub_apply, sub_self] field_simp try ring)