Key Takeaways: Mathematical Induction

A one-page reference card. Reread it before an exam or before writing your next induction proof.

The technique at a glance

Mathematical induction proves $P(n)$ for all integers $n \ge \text{start}$ using two finite steps.

Part What you write Why it's there
State $P(n)$ "Let $P(n)$ be the statement that …" Pins down exactly what you're proving
Base case Prove $P(\text{start})$ directly Push the first domino
Inductive step Fix arbitrary $k \ge \text{start}$; assume $P(k)$; prove $P(k+1)$ Each domino topples the next
Conclude "By induction, $P(n)$ holds for all $n \ge \text{start}$. $\blacksquare$" Chains the dominoes to infinity

The assumption $P(k)$ is the inductive hypothesis. You prove an implication — assuming $P(k)$ to get $P(k+1)$ is not circular.

Recursion ↔ induction (the same shape)

Induction proof Recursive program
Base case $P(\text{start})$ Non-recursive (terminating) branch
Inductive hypothesis $P(k)$ "the recursive call returns the right answer"
Inductive step Combine the recursive result correctly
Conclusion: correct for all $n$ Function correct for all inputs

Iterative code uses a loop invariant: initialization = base case, maintenance = inductive step, plus a termination argument.

Proof engines (memorize these moves)

  • Summation induction: peel off the last term, then apply the hypothesis. $$\sum_{i=1}^{k+1} a_i = \left(\sum_{i=1}^{k} a_i\right) + a_{k+1} \xrightarrow{\text{hyp.}} (\text{closed form at }k) + a_{k+1}.$$
  • Divisibility induction: unpack "$d \mid x$" as "$x = d m$", do algebra, repackage as "$d \mid (\dots)$".
  • Inequality induction: bound $f(k+1)$ using $f(k)$ (the hypothesis), then close the remaining gap.

Quick-reference results proven this chapter

Identity Holds for
$\sum_{i=1}^{n} i = \dfrac{n(n+1)}{2}$ $n \ge 1$
$\sum_{i=1}^{n} (2i-1) = n^2$ $n \ge 1$
$2^n > n^2$ $n \ge 5$ (note the base case!)
$3 \mid (4^n - 1)$ $n \ge 0$
$\sum_{i=1}^{n} F_i = F_{n+2} - 1$ $n \ge 1$

When to reach for induction

Is the claim "for all integers n >= something"?
  ├─ No  → induction is probably the wrong tool.
  └─ Yes → Does P(k+1) naturally build on P(k)?
            ├─ Yes, on the single previous case      → ordinary induction (this chapter)
            ├─ Yes, but needs several previous cases  → STRONG induction (Chapter 7)
            └─ It's about a recursively-built object  → STRUCTURAL induction (Chapter 7)

Common pitfalls

  • Wrong base case. Test small $n$ first; start where the claim becomes true ($2^n > n^2$ starts at 5).
  • Step that secretly needs large $k$. The inductive step must hold for every $k \ge \text{start}$. The "all horses same color" fallacy breaks because its overlap argument fails at $k=1$.
  • Skipped/false base case. A valid step with no valid base proves nothing.
  • Needing two cases back. If you must assume $P(k-1)$ and $P(k)$, you need strong induction.

Toolkit addition

recurrences.py: check_identity(lhs, rhs, upto, start) (tests a conjectured identity, returns the first counterexample or None) and a naive fib(n). Remember: a test that finds no counterexample is evidence, not a proof.