Many "fractal" konkur problems (Koch snowflake, Cantor-style lengths, Sierpiński perimeters) hide a very simple engine: a self-similar substitution rule applied to every piece of the figure, over and over. If you spot the rule, the answer is just a geometric sequence — no calculus, no figure-counting.
The substitution rule
Fix attention on one straight segment of length . A single iteration is described by two numbers:
- = how many segments each segment is replaced by,
- = the length scale factor of each new segment (, usually or ).
So one iteration sends
meaning the length of that piece is multiplied by the constant ratio . Because every segment obeys the same rule, the total length of the whole figure obeys the same multiplier:
Solving the recurrence
This is a first-order linear recurrence with constant ratio , so it is purely geometric:
where is the length/perimeter of the starting figure (step ). Watch the indexing: "after steps" means multiply times. Equivalently, tracking the number of segments and the length of each separately gives the same product .
The behavior as is decided by :
| Condition | Total length |
|---|---|
| (infinite perimeter, e.g. Koch) | |
| constant | |
| (e.g. the Cantor set's length) |
How to apply it
- Look at a single segment after one step. Count (new segments) and read off (scale).
- Form the ratio .
- Identify = perimeter of the initial figure (don't forget it has several sides).
- Plug into , being careful whether the question counts from step or step .
Mini-example: Koch curve from one segment
Koch replaces the middle third of a segment by the two sides of a small equilateral triangle, so one segment becomes segments each scaled by . Thus . Starting from a single segment of length , after steps the length is
For the full Koch snowflake the seed is an equilateral triangle, so for side , and the perimeter after steps is — which diverges, encircling a finite area with an infinite boundary.
Exam trick: The whole problem reduces to the ratio of one segment — never try to count all the little edges in the final figure. And mind the step count: the figure "obtained at the -th step" has been transformed times, so the exponent is , not (and not ).