Skip to main content

Strong induction

Another way to use a stronger induction hypothesis is called strong induction: To prove P(k)P(k), instead of assuming that P(k1)P(k-1) is true, we assume that P(1),,P(k1)P(1),\ldots, P(k-1) are all true and try to derive the validity of P(k)P(k). This again goes into the principle of making the hypothesis stronger so that we can prove the statement we want.

Example 17

Every natural number n>1n >1 can be written as a product of primes. (Equivalent to Lemma 1)

Proof:

We use the hypothesis:

P(n):nP(n): n can be written as a product of primes

Notice that if we want to prove P(100)P(100), we may want to write 100=2×50100 = 2 \times 50. But then we need the fact that 5050 can be written as a prime product, so the validity of P(99)P(99) is not enough for us. Let us try to prove this via a strong induction. The base case is when n=2n=2, and P(2)P(2) holds trivially. Now assume that P(2),,P(n0)P(2), \ldots, P(n_0) hold. To see P(n0+1)P(n_0+1), there are two cases. If n0+1n_0+1 is prime, we are done. Otherwise, we can factor it into n0+1=xyn_0+1 = x\cdot y where 1<x,y<n0+11 < x, y <n_0+1. By using the validity of P(x)P(x) and P(y)P(y), we can write xx and yy as a product of primes, and therefore the same holds for n0+1n_0+1.

Exercise 6

We are interested in writing a rational number p/q<1p/q <1 as the sum of fractions with numerators 11. For instance, 3/5=1/2+1/103/5 = 1/2+1/10 and 4/5=1/2+1/4+1/204/5= 1/2+1/4+1/20. Consider the following algorithm. To write p/qp/q in this form, calculate pq1q/p\frac{p}{q} - \frac{1}{\lceil q/p \rceil}. If the result is zero, we are done; otherwise, repeat the same step on the result. Prove that this algorithm always terminates in finitely many steps.