🧮 Brain Teaser

The Harmonic Series Is Never an Integer

Why it's beautiful: Hn=1+12+13++1nH_n = 1 + \frac{1}{2} + \frac{1}{3} + \cdots + \frac{1}{n} grows without bound, so you might expect it to eventually hit an integer. It never does (for n2n \geq 2). The proof uses a clever "lonely prime" argument: there is always one prime that appears exactly once in the denominators, and that prime poisons the entire sum, preventing it from being an integer.


Problem

Let Hn=1+12+13++1nH_n = 1 + \frac{1}{2} + \frac{1}{3} + \cdots + \frac{1}{n}.

Prove that HnH_n is not an integer for any n2n \geq 2.


Hint: Let pp be the largest prime n\leq n. By Bertrand's postulate, p>n/2p > n/2, so 2p>n2p > n — meaning pp appears only once among {1,2,,n}\{1, 2, \ldots, n\}. Now multiply HnH_n by L=lcm(1,2,,n)L = \text{lcm}(1, 2, \ldots, n) and think about divisibility by pp.

Harmonic seriesPrimesBertrand's postulateLCMDivisibility

Harmonic Series Is Never an Integer — Answer


Setup: The "Lonely Prime" Idea

Bertrand's Postulate states: for every integer n1n \geq 1, there exists a prime pp with n<p2nn < p \leq 2n.

Equivalently: the largest prime pnp \leq n satisfies p>n/2p > n/2, so 2p>n2p > n.

This means pp is the only multiple of pp in {1,2,,n}\{1, 2, \ldots, n\} — hence "lonely."


Proof

Let n2n \geq 2, and let pp be the largest prime n\leq n.

Step 1: Multiply through by L=lcm(1,2,,n)L = \text{lcm}(1, 2, \ldots, n).

LHn=L1+L2++LnL \cdot H_n = \frac{L}{1} + \frac{L}{2} + \cdots + \frac{L}{n}

Every term L/kL/k is an integer (since kLk \mid L by definition of lcm). So LHnL \cdot H_n is an integer.

If HnH_n were also an integer, then pLHnp \mid L \cdot H_n (since pLp \mid L). We will show this is impossible.


Step 2: Determine the exact power of pp in LL.

Since p>n/2p > n/2, we have p2>np^2 > n, so p2p^2 does not divide any knk \leq n. Therefore:

p1L(i.e., pL but p2L)p^1 \,\Big\|\, L \quad \text{(i.e., } p \mid L \text{ but } p^2 \nmid L\text{)}


Step 3: Analyze each term L/kL/k modulo pp.

  • If kpk \neq p: since pkp \nmid k (because pp is the only multiple of pp in {1,,n}\{1,\ldots,n\}), and p1Lp^1 \| L, we get pL/kp \mid L/k.

  • If k=pk = p: since p1Lp^1 \| L, removing one factor of pp gives pL/pp \nmid L/p.

So modulo pp:

LHn=Lp≢0+kpLk0≢0(modp)L \cdot H_n = \underbrace{\frac{L}{p}}_{\not\equiv\, 0} + \underbrace{\sum_{k \neq p} \frac{L}{k}}_{\equiv\, 0} \not\equiv 0 \pmod{p}


Step 4: Conclude.

We have shown pLHnp \nmid L \cdot H_n.

But if Hn=mH_n = m were an integer, then LHn=LmL \cdot H_n = L \cdot m, and since pLp \mid L, we'd have pLmp \mid L \cdot m — contradiction.

Therefore HnH_n is not an integer. \square


Why each step is needed

| Step | Role | |------|------| | Bertrand's postulate | Guarantees pp appears exactly once in {1,,n}\{1,\ldots,n\} | | p1Lp^1 \| L | Ensures L/pL/p is not divisible by pp, but all other L/kL/k are | | Mod pp analysis | Shows LHn≢0(modp)L \cdot H_n \not\equiv 0 \pmod{p}, killing integrality |

The lonely prime pp is the obstruction. Every other denominator cancels cleanly — but pp has no partner to cancel with, so it leaves a remainder that can never be an integer.

Type: PutnamEdit on GitHub ↗