energy_filter_gaussian_upper_step
For n>0, t≥0, and t+1≤n, (n−t)/(n+t+1) is at most exp(−((t+1)²/(n+t+1)−t²/(n+t))).
Statement
theorem energy_filter_gaussian_upper_step (n t : ℝ)
(hn : 0 < n) (ht : 0 ≤ t) (_htn : t + 1 ≤ n) :
(n-t)/(n+t+1) ≤
Real.exp (-(((t+1)^2/(n+t+1))-(t^2/(n+t))))- Source
- OP-892: centered binomial fourth-mass asymptotic via rescaled step functions and dominated convergence.
- Verified
- 14 Nov 2026
- Axioms
Classical.choiceQuot.soundpropext- Used by 1
- energy_filter_centered_binomial_gaussian_sandwichFor natural k≥1 and 0≤j≤k, choose(2k,k+j)/choose(2k,k) lies between exp(−j²/(k−j+1)) and exp(−j²/(k+j)).
The proof, explained
The argument rewrites the ratio (n - t) / (n + t + 1) as 1 - (2t + 1) / (n + t + 1) and upper bounds it by exp(-(2t + 1) / (n + t + 1)) via the standard inequality 1 + x ≤ exp(x). Expanding (t + 1)² = t² + 2t + 1 and using the fact that t² / (n + t + 1) ≤ t² / (n + t) shows that the exponent difference (t + 1)² / (n + t + 1) - t² / (n + t) is at most (2t + 1) / (n + t + 1). Negating this inequality and applying the monotonicity of the exponential function completes the bound.
The denominators n + t and n + t + 1 are strictly positive.
Because n > 0 and t ≥ 0, their sum n + t is strictly positive, and adding 1 ensures n + t + 1 > 0 as well.
have hp : 0 < n+t := by positivity have hp' : 0 < n+t+1 := by positivity
t² / (n + t + 1) ≤ t² / (n + t)
Since t² ≥ 0 and 0 < n + t ≤ n + t + 1, increasing the positive denominator produces a smaller or equal fraction.
uses
div_le_div_of_nonneg_left,sq_nonneghave hdiv : t^2/(n+t+1) ≤ t^2/(n+t) := div_le_div_of_nonneg_left (sq_nonneg t) hp (by linarith)
(t + 1)² / (n + t + 1) - t² / (n + t) ≤ (2 * t + 1) / (n + t + 1)
Expanding (t + 1)² as t² + (2 * t + 1) and distributing over the denominator splits the fraction into t² / (n + t + 1) + (2 * t + 1) / (n + t + 1). Since t² / (n + t + 1) ≤ t² / (n + t), subtracting t² / (n + t) leaves at most (2 * t + 1) / (n + t + 1).
uses
add_divhave hdelta : (t+1)^2/(n+t+1) - t^2/(n+t) ≤ (2*t+1)/(n+t+1) := by have hpoly : (t+1)^2 = t^2 + (2*t+1) := by ring rw [hpoly, add_div] linarith
(n - t) / (n + t + 1) = 1 - (2 * t + 1) / (n + t + 1)
Clearing the non-zero denominator reduces this to n - t = (n + t + 1) - (2 * t + 1), which holds by basic algebra.
have hid : (n-t)/(n+t+1) = 1 - (2*t+1)/(n+t+1) := by field_simp; ring
-(2 * t + 1) / (n + t + 1) + 1 ≤ Real.exp (-(2 * t + 1) / (n + t + 1))
This is the standard inequality x + 1 ≤ exp(x) instantiated at x = -(2 * t + 1) / (n + t + 1).
uses
Real.add_one_le_exphave hexp := Real.add_one_le_exp (-((2*t+1)/(n+t+1)))(n - t) / (n + t + 1) ≤ Real.exp (-((t + 1)² / (n + t + 1) - t² / (n + t)))
Chaining the identity and the exponential inequality gives (n - t) / (n + t + 1) ≤ exp(-(2 * t + 1) / (n + t + 1)). Negating the inequality from step 3 reverses it to -(2 * t + 1) / (n + t + 1) ≤ -((t + 1)² / (n + t + 1) - t² / (n + t)), and taking the exponential on both sides preserves this order by monotonicity of exp.
uses
Real.exp_le_exp.mprcalc _ = 1 - (2*t+1)/(n+t+1) := hid _ ≤ Real.exp (-((2*t+1)/(n+t+1))) := by linarith _ ≤ Real.exp (-(((t+1)^2/(n+t+1))-(t^2/(n+t)))) := by apply Real.exp_le_exp.mpr linarith
add_div- Division distributes over addition from the right: (a + b) / c = a / c + b / c.
div_le_div_of_nonneg_left- For a non-negative numerator a ≥ 0 and positive denominators with c ≤ b, dividing by the larger denominator gives a smaller or equal quotient: a / b ≤ a / c.
sq_nonneg- The square of any real number (or element of an ordered semiring) is non-negative: 0 ≤ a².
Real.add_one_le_exp- For any real number x, x + 1 ≤ exp(x).
Real.exp_le_exp.mpr- The real exponential function is monotonically non-decreasing: if a ≤ b, then exp(a) ≤ exp(b).
- The hypothesis `_htn : t + 1 ≤ n` is unused in the proof; the inequality holds for all real numbers n > 0 and t ≥ 0.
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 energy_filter_gaussian_upper_step (n t : ℝ) (hn : 0 < n) (ht : 0 ≤ t) (_htn : t + 1 ≤ n) : (n-t)/(n+t+1) ≤ Real.exp (-(((t+1)^2/(n+t+1))-(t^2/(n+t)))) := by have hp : 0 < n+t := by positivity have hp' : 0 < n+t+1 := by positivity have hdiv : t^2/(n+t+1) ≤ t^2/(n+t) := div_le_div_of_nonneg_left (sq_nonneg t) hp (by linarith) have hdelta : (t+1)^2/(n+t+1) - t^2/(n+t) ≤ (2*t+1)/(n+t+1) := by have hpoly : (t+1)^2 = t^2 + (2*t+1) := by ring rw [hpoly, add_div] linarith have hid : (n-t)/(n+t+1) = 1 - (2*t+1)/(n+t+1) := by field_simp; ring have hexp := Real.add_one_le_exp (-((2*t+1)/(n+t+1))) calc _ = 1 - (2*t+1)/(n+t+1) := hid _ ≤ Real.exp (-((2*t+1)/(n+t+1))) := by linarith _ ≤ Real.exp (-(((t+1)^2/(n+t+1))-(t^2/(n+t)))) := by apply Real.exp_le_exp.mpr linarith