Lipschitz condition
/ LIP-shits /
A Lipschitz condition is a controlled cap on how fast a function is allowed to change. A function f of y satisfies it if there is a single constant L (the Lipschitz constant) so that for any two inputs y1 and y2, the outputs differ by at most L times the inputs: |f(y1) - f(y2)| <= L |y1 - y2|. In words, the graph can never be steeper than a line of slope L anywhere — no vertical cliffs, no infinitely sharp corners. It is a notch stronger than plain continuity but weaker than having a derivative; in fact a function with a bounded derivative is automatically Lipschitz, with L the bound on |f'|.
In differential equations the condition is asked of the right-hand side f(x, y) as a function of its second slot, y, usually uniformly for x in some interval: |f(x, y1) - f(x, y2)| <= L |y1 - y2|. This is precisely the extra ingredient that turns existence into existence-and-uniqueness. The reason is intuitive: if two solutions started close and tried to peel apart, the difference in their slopes is bounded by L times the difference in their heights, so they cannot separate faster than exponentially — and if they start at the same point, the bound forces them to stay glued together forever. That bounding-the-spread argument is made rigorous by Gronwall's inequality.
The honest caveat is where the condition fails. At points where f has an infinite slope in y, no finite L works, and uniqueness can collapse. The canonical case is f(y) = y^(2/3): near y = 0 its derivative (2/3) y^(-1/3) blows up, so there is no Lipschitz constant there, and indeed infinitely many solutions sprout from the origin. So a Lipschitz condition is best read as a precise statement of exactly what must be true to forbid that branching — it rules out the steepness that lets solutions split apart.
f(y) = 3y is Lipschitz with L = 3: any two y-values feed slopes that differ by 3 times their gap. But f(y) = y^(2/3) is not Lipschitz near 0, because its slope (2/3) y^(-1/3) grows without bound as y -> 0 — and that is exactly where uniqueness breaks.
A bounded slope (a finite L) is the precise guard against solutions branching; an infinite slope removes it.
Lipschitz sits strictly between continuous and differentiable: every differentiable-with-bounded-derivative function is Lipschitz, but a Lipschitz function need not be differentiable (think of |y|, with a corner at 0).