Maths WorkshopMaths problems with full solutions
Exercise 1557Calculus › Single-variable calculus › Differentiation › Taylor expansionsM1Added 3 Nov 2026

Bounding f′ by f and f″ (Landau's inequality)

#Taylor-Lagrange#inequalities#optimisation of a parameter

Statement

Let f:R→Rf:\mathbb{R}\to\mathbb{R} be twice differentiable, with ∣f∣≤M0|f|\le M_0 and ∣f′′∣≤M2|f''|\le M_2 on R\mathbb{R}. Prove that
∣f′(x)∣≤2M0M2for all x.|f'(x)|\le\sqrt{2M_0M_2}\quad\text{for all }x.

Hint

Write Taylor–Lagrange at order 22 for f(x+h)f(x+h) and f(x−h)f(x-h), subtract, and then choose hh optimally.

Show solutionHide solution

Solution

Taylor–Lagrange. For h>0h>0 there exist c1∈(x,x+h)c_1\in(x,x+h) and c2∈(x−h,x)c_2\in(x-h,x) with
f(x+h)=f(x)+hf′(x)+h22f′′(c1),f(x−h)=f(x)−hf′(x)+h22f′′(c2).f(x+h)=f(x)+hf'(x)+\tfrac{h^2}2f''(c_1),\qquad f(x-h)=f(x)-hf'(x)+\tfrac{h^2}2f''(c_2).

Subtracting.
2hf′(x)=f(x+h)−f(x−h)−h22(f′′(c1)−f′′(c2)).2hf'(x)=f(x+h)-f(x-h)-\tfrac{h^2}2\bigl(f''(c_1)-f''(c_2)\bigr).
Using the bounds, 2h∣f′(x)∣≤2M0+h2M22h|f'(x)|\le2M_0+h^2M_2, that is
∣f′(x)∣≤M0h+hM22for every h>0.|f'(x)|\le\frac{M_0}h+\frac{hM_2}2\qquad\text{for every }h>0.

Optimising. The right side is minimal at h=2M0/M2h=\sqrt{2M_0/M_2}, where it equals M0M22+M0M22=2M0M2\sqrt{\frac{M_0M_2}2}+\sqrt{\frac{M_0M_2}2}=\sqrt{2M_0M_2}. (If M2=0M_2=0, let h→∞h\to\infty: f′=0f'=0.)

Check. For f=sin⁡f=\sin: M0=M2=1M_0=M_2=1 and the bound is 2≥1=sup⁡∣cos⁡∣\sqrt2\ge1=\sup|\cos| ✓.

Remark. A function cannot be both small and have a small second derivative while having a large first derivative: this is a quantitative version of "a smooth bounded function cannot oscillate quickly with large amplitude". The sharp constant is 2M0M2\sqrt{2M_0M_2}.