exp_ge_linear
Bernoulli's inequality states that for real a at least 1 and natural n, a^n is at least 1 plus n times (a minus 1). Growth bound arguments in the Recognition framework cite it when exponentials must dominate linear or polynomial terms on the phi-ladder. The proof runs by induction on n, multiplying the inductive hypothesis by a and closing the step with ring rewriting plus nlinarith on nonnegativity.
claimFor all real numbers $a$ with $a ≥ 1$ and all natural numbers $n$, $a^n ≥ 1 + n(a - 1)$.
background
The GrowthBounds module supplies pure real analysis results showing exponential growth dominates any polynomial, with instances for the phi-ladder versus cubic volume growth that close the Fermi chain. Bernoulli's inequality supplies the base comparison between exponential and linear terms needed for those dominance statements. The module imports ArithmeticFromLogic for successor structure and Recognition plus Cycle3 for the monolith and 3-cycle recognition structure that embed the inequality into the larger framework.
proof idea
Proof by induction on n. The zero case simplifies directly to equality. In the successor step, the inductive hypothesis is multiplied on the right by a (using nonnegativity of a), the product is rewritten by ring to a + k a (a-1), then nlinarith lowers the middle term using a ≥ 1 and nonnegativity of (a-1), after which ring and cast normalization recover the linear form for k+1.
why it matters in Recognition Science
This theorem is invoked directly by exponential_exceeds_bound (which produces N such that a^N exceeds any M) and by phi_four_power_lower (which bounds phi to the 4M against (M/2)^4). It advances the Recognition framework by enabling the demonstration that phi-exponential growth defeats polynomial growth, consistent with the eight-tick octave and D=3 in the forcing chain. No open scaffolding remains at this step.
scope and limits
- Does not apply when a is less than 1.
- Does not address equality cases or sharpness.
- Does not extend to non-integer exponents.
- Does not treat complex bases or vectors.
Lean usage
have hge := exp_ge_linear a (le_of_lt ha) N
formal statement (Lean)
21theorem exp_ge_linear (a : ℝ) (ha : 1 ≤ a) (n : ℕ) :
22 a ^ n ≥ 1 + (n : ℝ) * (a - 1) := by
proof body
Tactic-mode proof.
23 induction n with
24 | zero => simp
25 | succ k ih =>
26 have ha_nonneg : 0 ≤ a := by linarith
27 have hk_nn : (0 : ℝ) ≤ k := Nat.cast_nonneg k
28 calc a ^ (k + 1) = a ^ k * a := pow_succ a k
29 _ ≥ (1 + (k : ℝ) * (a - 1)) * a := by
30 exact mul_le_mul_of_nonneg_right ih ha_nonneg
31 _ = a + (k : ℝ) * a * (a - 1) := by ring
32 _ ≥ a + (k : ℝ) * 1 * (a - 1) := by
33 nlinarith [mul_nonneg hk_nn (sub_nonneg.mpr ha), sq_nonneg (a - 1)]
34 _ = 1 + ((k : ℝ) + 1) * (a - 1) := by ring
35 _ = 1 + (↑(k + 1) : ℝ) * (a - 1) := by push_cast; ring
36
37/-- For a > 1 and any M, there exists N such that a^N > M. -/