Every point is the center, nested balls, and the p-adic Cantor set fractal
The p-adic absolute value satisfies a condition far stronger than the usual triangle inequality. Instead of |x + y| ≤ |x| + |y|, we have the ultrametric inequality:
|x + y|p ≤ max(|x|p, |y|p)
This single inequality reshapes all of geometry. In an ultrametric space, our Euclidean intuition fails spectacularly: triangles are always isosceles (with the unequal side being the shortest), every point inside a ball is its center, and two balls can never partially overlap. The resulting topology is totally disconnected, giving the p-adic integers a fractal, Cantor-set-like structure.
These are not mere curiosities. The ultrametric structure is what makes p-adic analysis work so differently from real analysis, and understanding it is essential for everything that follows.
In ordinary geometry, a ball has a unique center. In the p-adic world, this is no longer true. The ball B(a, r) = { x : |x − a|p ≤ r } consists of all integers whose p-adic expansion agrees with a in the last k digits. Pick any other point b in that ball, and the ball centered at b with the same radius contains exactly the same set of points.
Click any point inside the ball to re-center on it. The ball contains the same set of members regardless of which point you choose as center.
Key insight: In an ultrametric space, every point of a ball is its center. If b ∈ B(a, r), then B(b, r) = B(a, r). This follows directly from the ultrametric inequality: if |a − b|p ≤ r and |b − x|p ≤ r, then |a − x|p ≤ max(|a − b|p, |b − x|p) ≤ r.
In Euclidean space, two circles can intersect in a crescent-shaped region -- they partially overlap. In an ultrametric space, this is impossible. Two p-adic balls are either completely disjoint (sharing no points) or one is entirely contained inside the other. There is no in-between.
In the Euclidean panel, circles can partially overlap freely. In the p-adic panel, when you drag a ball into a partial overlap, it snaps to either fully nested or fully disjoint.
Key insight: If two p-adic balls share even a single point, then one ball is a subset of the other. This means the set of all balls forms a tree (partially ordered by inclusion), with no "crossing branches." Compare this with the Euclidean case on the left, where partial overlaps create complicated intersection patterns.
The p-adic integers Zp can be visualized as a fractal. Start with the interval [0, 1] and subdivide it into p equal parts, one for each possible first digit. Then subdivide each part into p more pieces for the second digit, and so on. The resulting limit set is a Cantor-like fractal that is homeomorphic to Zp. For p = 2, this looks like a Cantor set (but keeping both halves); for p = 3, you keep all three thirds (unlike the classical Cantor set, which removes the middle third).
Click an interval at the deepest level to zoom into it, revealing self-similar structure. Use the Zoom In / Zoom Out buttons to add or remove levels.
Key insight: The fractal structure of Zp reflects its total disconnectedness. Every open set is also closed, and the space "falls apart" into clopen (simultaneously closed and open) subsets at every scale. Each zoom level corresponds to fixing one more p-adic digit, and the self-similarity captures the recursive nature of the p-adic expansion.