Koch Snowflake: Infinite Perimeter, Finite Area

One rule, applied forever

The Swedish mathematician Helge von Koch introduced this curve in a 1904 paper, looking for a continuous curve that had no tangent line anywhere — a "monster" curve, in the language mathematicians of the era used for the handful of pathological examples that were breaking their intuitions about smoothness. The construction rule is almost embarrassingly simple: take any straight line segment, divide it into three equal thirds, remove the middle third, and replace it with the other two sides of an equilateral triangle built outward on that gap. Apply that same replacement to every one of the resulting four segments, then to every segment of the new shape, forever. Starting from an equilateral triangle instead of a single segment and applying the rule to all three sides at once gives the six-pointed Koch snowflake.

The perimeter runs away to infinity

Every application of the rule replaces one segment of length L with four segments of length L/3 each, so the total length after that step is 4 × (L/3) = 4L/3 — a multiplication by exactly 4/3 every single iteration, regardless of how many segments exist by that point. Starting from a triangle with perimeter P, after n iterations the perimeter is P × (4/3)ⁿ, and since 4/3 is greater than 1, that quantity grows without bound as n increases. There is no largest snowflake perimeter and no limiting finite value — in the idealized mathematical limit, after infinitely many iterations, the boundary has become infinitely long, even though it still fits inside a bounded region of the plane you could draw a finite circle around.

// recursive construction: replace a segment (p1 → p2) with 4 shorter segments
function koch(p1, p2, depth):
  if depth == 0: return [p1, p2]
  a = p1 + (p2 - p1) / 3                       // first third-point
  b = p1 + (p2 - p1) * 2 / 3                   // second third-point
  peak = midpoint(a, b) + outwardNormal(p2 - p1) * (side / (2 * sqrt(3)))  // equilateral bump
  return koch(p1, a, depth-1) + koch(a, peak, depth-1)
       + koch(peak, b, depth-1) + koch(b, p2, depth-1)

// perimeter(n) = P0 * (4/3)^n         → diverges as n → ∞
// area(n)      = A0 * (1 + (3/5) * (1 - (4/9)^n))  → converges to A0 * 8/5

The area converges anyway

The area behaves completely differently, and that contrast is the whole point of the construction. Each new bump adds a small triangle whose area is proportional to the square of the segment it replaces, and because each generation's segments are 1/3 the length of the previous generation's, each new triangle's area is (1/3)² = 1/9 the area of the triangle added at the corresponding position one generation earlier — but there are 4× as many segments getting a new bump each time, so the total area added per iteration shrinks by a net factor of 4/9 compared with the iteration before. A geometric series with ratio 4/9 (which is less than 1) converges to a finite sum: starting from an initial equilateral triangle of area A0, the snowflake's total area converges to exactly 8/5 of A0 — a clean, closed-form limit, in stark contrast to the perimeter's runaway growth.

A dimension that isn't a whole number

Benoit Mandelbrot later used the Koch curve as a founding example when developing fractal geometry, in part because its self-similarity dimension is not an integer. Each iteration replaces one segment with four self-similar copies, each scaled down by a factor of 1/3; the general formula for self-similarity dimension is D = log(N)/log(1/r), where N is the number of self-similar pieces and r is the scaling ratio, giving D = log(4)/log(3) ≈ 1.2619. That value sitting strictly between 1 (the dimension of an ordinary smooth curve) and 2 (the dimension of a filled two-dimensional region) is a formal way of capturing the informal observation that the Koch curve is "rougher" than a line but never actually fills any area — exactly the kind of intermediate, non-integer roughness that gave the field of fractal geometry its name and its founding intuition.

Nowhere smooth, at any zoom level

Von Koch's original motivation — a curve with no tangent anywhere — follows directly from the same recursive rule. Zoom into any point on the finished curve, at any magnification, and you will always find a fresh triangular kink introduced at some later iteration; the curve never locally straightens out the way a smooth curve does as you zoom in closer and closer. Formally, this means the Koch curve is continuous everywhere but differentiable nowhere — there is no point on it where a well-defined tangent line, and therefore a well-defined slope or velocity, exists. It was one of the first explicit, easy-to-visualize constructions of such a curve, predating the more abstract pathological examples analysts had been using, and it remains one of the clearest ways to see why "continuous" and "smooth" are not the same property.

Frequently asked questions

Try it live

The full recursive construction runs live in Koch Snowflake. Step through iterations one at a time from a triangle, square, or hexagon base and watch the perimeter counter climb while the area counter visibly settles down.