Ordinary Differential Equations 9 | Lipschitz Continuity [dark version]

TL;DR

Local Lipschitz continuity requires a neighborhood-dependent bound: for each x there exist ε>0 and L>0 so that ‖V(y)−V(z)‖ ≤ L‖y−z‖ for all y,z within the ε-ball.

Briefing Cornell Notes

Briefing

Uniqueness of solutions to an initial value problem for ordinary differential equations hinges on a “middle-ground” regularity condition: local Lipschitz continuity. Continuous functions alone do not guarantee uniqueness, while continuous differentiability is often stronger than necessary. Local Lipschitz continuity sits between those extremes—strong enough to control how fast the right-hand side can change, but weaker than requiring derivatives to be continuous.

The core definition is built around bounding how much a function’s outputs can differ in terms of how much its inputs differ. For a map V: R^N → R^N (or on an open subset), local Lipschitz continuity means: for every point x, there exists a radius ε > 0 and a constant L > 0 such that for all inputs y and z inside the ε-ball around x, the norm of the output difference satisfies ‖V(y) − V(z)‖ ≤ L ‖y − z‖. The constant L is the key: it works uniformly for all y and z in that neighborhood, not just for a single pair of points. This “local” requirement matters because the bound only needs to hold near each point, not globally across the entire domain.

From that definition, two immediate consequences follow. First, local Lipschitz continuity implies ordinary continuity. If y_n → y, then the Lipschitz estimate forces V(y_n) → V(y), since the right-hand side L‖y_n − y‖ goes to zero. Second, local Lipschitz continuity prevents slopes from blowing up. By rearranging the inequality for distinct y and z, the difference quotient is bounded by L, meaning the function cannot develop arbitrarily steep behavior inside the neighborhood.

The transcript then connects Lipschitz continuity to differentiability. Starting with a C^1 function in one dimension, the mean value theorem turns the secant slope into a tangent slope: for points y and z near a fixed x, the difference quotient equals f′(c) for some intermediate point c between y and z. Because f′ is continuous, it is bounded on the ε-neighborhood, so one can choose a finite Lipschitz constant L as the supremum of |f′| over that neighborhood. That yields a local Lipschitz bound, proving that every C^1 function is locally Lipschitz. The same reasoning extends to higher dimensions.

This chain of facts sets up the next step in the ODE course: once the right-hand side V in an initial value problem is locally Lipschitz, the controlled change rate implied by the Lipschitz estimate becomes the ingredient needed to establish uniqueness of solutions. In short, local Lipschitz continuity is the practical condition that turns “well-behaved” dynamics into “uniquely determined” trajectories.

Cornell Notes

Local Lipschitz continuity is the regularity condition that guarantees uniqueness for ODE initial value problems. A function V is locally Lipschitz if, around every point x, there exists an ε-ball and a constant L such that ‖V(y)−V(z)‖ ≤ L‖y−z‖ for all y,z in that ball. This condition automatically implies ordinary continuity and also bounds difference quotients, preventing slopes from becoming unbounded. For C^1 functions, the mean value theorem plus boundedness of the derivative on a neighborhood produces exactly the Lipschitz estimate, showing that C^1 ⇒ locally Lipschitz (and the argument extends to higher dimensions).

What does “locally Lipschitz” mean in terms of an inequality?

For each point x in the domain, there must exist an ε > 0 and a constant L > 0 such that for all y and z inside the ε-ball around x, the output variation is controlled by the input variation: ‖V(y) − V(z)‖ ≤ L‖y − z‖. The constant L must work uniformly for every pair y,z in that neighborhood.

Why does local Lipschitz continuity imply ordinary continuity?

If y_n → y, then the Lipschitz estimate gives ‖V(y_n) − V(y)‖ ≤ L‖y_n − y‖. Since ‖y_n − y‖ → 0, the right-hand side goes to 0, forcing ‖V(y_n) − V(y)‖ → 0. That is exactly the definition of continuity at y.

How does local Lipschitz continuity bound “slopes” or difference quotients?

Rewriting the inequality for distinct y and z yields ‖V(y) − V(z)‖/‖y − z‖ ≤ L within the ε-neighborhood. In one dimension this is the bound on the absolute value of the secant slope. The key point is that L is finite and uniform locally, so the function cannot develop arbitrarily large slopes near x.

How does the mean value theorem show that a C^1 function is locally Lipschitz (1D case)?

For a C^1 function f, the mean value theorem states that for y and z near a fixed x, the secant slope (f(y)−f(z))/(y−z) equals f′(c) for some intermediate point c between y and z. Taking absolute values gives |(f(y)−f(z))/(y−z)| = |f′(c)|. Since f′ is continuous, it is bounded on the ε-neighborhood, so one can choose L as the supremum of |f′| there, producing the Lipschitz inequality.

Why does boundedness of f′ matter for the Lipschitz constant?

The Lipschitz constant L must be finite. Continuity of f′ on a neighborhood implies it attains a finite supremum (or at least has a finite bound) on that neighborhood. That finite bound becomes the L in ‖f(y)−f(z)‖ ≤ L|y−z|, ensuring the Lipschitz estimate holds for all y,z in the neighborhood.

Review Questions

State the formal definition of local Lipschitz continuity for V: R^N → R^N and explain what “local” means in the quantifiers.
Prove (using the Lipschitz inequality) that local Lipschitz continuity implies continuity.
Explain how the mean value theorem and boundedness of f′ on a neighborhood produce a Lipschitz constant for a C^1 function.

Key Points

1
Local Lipschitz continuity requires a neighborhood-dependent bound: for each x there exist ε>0 and L>0 so that ‖V(y)−V(z)‖ ≤ L‖y−z‖ for all y,z within the ε-ball.
2
The Lipschitz constant L must work uniformly for all pairs of points in that neighborhood, not just for one pair.
3
Local Lipschitz continuity implies ordinary continuity because the Lipschitz inequality forces ‖V(y_n)−V(y)‖ → 0 whenever y_n → y.
4
Local Lipschitz continuity bounds difference quotients, preventing slopes from becoming arbitrarily large near any point.
5
Every C^1 function is locally Lipschitz: the mean value theorem converts secant slopes into values of f′, and continuity of f′ yields a finite bound on neighborhoods.
6
The bounded-slope control provided by local Lipschitz continuity is the ingredient needed later to establish uniqueness of solutions for ODE initial value problems.

Highlights

Local Lipschitz continuity is defined by a neighborhood-wise inequality: ‖V(y)−V(z)‖ ≤ L‖y−z‖ for all y,z in an ε-ball around each point.

The condition automatically implies continuity and also enforces a local bound on difference quotients, ruling out infinite steepness.

For C^1 functions, the mean value theorem plus boundedness of f′ on a neighborhood produces the exact Lipschitz estimate.

Local Lipschitz continuity is positioned as the “right” regularity level for uniqueness in ODE initial value problems—stronger than continuity, weaker than requiring continuous derivatives everywhere.