# Harmonic Number is not Integer

## Theorem

Let $H_n$ be the $n$th harmonic number.

Then $H_n$ is not an integer for $n \ge 2$.

That is, the only harmonic numbers that are integers are $H_0$ and $H_1$.

## Proof

As $H_0 = 0$ and $H_1 = 1$, they are integers.

The claim is that $H_n$ is not an integer for all $n \ge 2$.

Aiming for a contradiction, suppose otherwise:

- $(P): \quad \exists m \in \N: H_m \in \Z$

By the definition of the harmonic numbers:

- $\displaystyle H_m = 1 + \frac 1 2 + \frac 1 3 + \ldots + \frac 1 m$

Let $2^t$ denote the highest power of two in the denominators of the summands.

Then:

\(\displaystyle H_m\) | \(=\) | \(\displaystyle 1 + \frac 1 2 + \frac 1 3 + \ldots + \frac 1 m\) | $\quad$ | $\quad$ | |||||||||

\((1):\quad\) | \(\displaystyle \implies \ \ \) | \(\displaystyle H_m - \frac 1 {2^t}\) | \(=\) | \(\displaystyle 1 + \frac 1 2 + \frac 1 3 + \ldots + \frac 1 {2^t - 1} + \frac 1 {2^t + 1} + \ldots + \frac 1 m\) | $\quad$ | $\quad$ | |||||||

\(\displaystyle \implies \ \ \) | \(\displaystyle 2^{t-1} H_m - \frac 1 2\) | \(=\) | \(\displaystyle 2^{t-1} + \frac {2^{t-1} } 2 + \frac {2^{t-1} } 3 + \frac {2^{t-1} } 4 + \frac {2^{t-1} } 5 + \frac {2^{t-1} } 6 + \ldots + \frac {2^{t-1} } m\) | $\quad$ multiplying by $2^{t-1}$ | $\quad$ | ||||||||

\((2):\quad\) | \(\displaystyle \) | \(=\) | \(\displaystyle 2^{t-1} + 2^{t-2} + \frac {2^{t-1} } 3 + 2^{t-3} + \frac {2^{t-1} } 5 + \frac {2^{t-2} } 3 + \ldots + \frac {2^{t-1} } m\) | $\quad$ cancelling powers of $2$ | $\quad$ |

Let $S$ be the set of denominators on the right hand side of $(2)$.

Then no element of $S$ can have $2$ as a factor, as follows.

Consider an arbitrary summand:

- $\dfrac {2^{t-1}}{2^j \times k}$

for some $k \in \Z$, where $j \ge 0$ is the highest power of $2$ that divides the denominator.

For any $2$ to remain after simplification, we would need $j > t - 1$.

Were this to be so, then $2^j\times k$ would have $2^t$ as a factor, and some denominator would be a multiple of $2^t$.

By Greatest Power of Two not Divisor, the set of denominators of the right hand side of $(1)$:

- $\left\{ {1, 2, 3, \ldots, 2^t - 1, 2^t + 1, \ldots, m}\right\}$

contains no multiple of $2^t$.

Therefore there can be no multiple of $2$ in the denominators of the right hand side of $(2)$.

Let:

- $\ell = \lcm \left({S}\right)$

be the least common multiple of the elements of $S$.

Because all the elements of $S$ are odd, $\ell$ is likewise odd.

We have:

\(\displaystyle 2^{t-1} H_m - \frac 1 2\) | \(=\) | \(\displaystyle 2^{t-1} + 2^{t-2} + \frac {2^{t-1} } 3 + 2^{t-3} + \frac {2^{t-1} } 5 + \frac {2^{t-2} } 3 + \ldots + \frac {2^{t-1} } m\) | $\quad$ from $(2)$ | $\quad$ | |||||||||

\(\displaystyle \implies \ \ \) | \(\displaystyle \frac {2^t H_m - 1} 2\) | \(=\) | \(\displaystyle \frac{2^{t-1} \ell + 2^{t-2} \ell + 2^{t-1}\left({\ell/3}\right) + 2^{t-3} \ell + 2^{t-1}\left({\ell/5}\right) + \ldots + 2^{t-1} \left({\ell/m}\right)} \ell\) | $\quad$ multiplying top and bottom by $\ell$ | $\quad$ | ||||||||

\(\displaystyle \implies \ \ \) | \(\displaystyle \ell \left({2^t H_m - 1}\right)\) | \(=\) | \(\displaystyle 2 \left({2^{t-1} \ell + 2^{t-2} \ell + 2^{t-1}\left({\ell/3}\right) + 2^{t-3} \ell + 2^{t-1}\left({\ell/5}\right) + \ldots + 2^{t-1} \left({\ell/m}\right)}\right)\) | $\quad$ multiplying both sides by $2 \ell$ | $\quad$ |

But the left hand side of that last equation is odd, while its right hand side is even.

As this is a contradiction, it follows that our assumption $(P)$ that such an $m$ exists is false.

That is, there are no harmonic numbers apart from $0$ and $1$ which are integers.

$\blacksquare$

## Sources

- 1997: Donald E. Knuth:
*The Art of Computer Programming: Volume 1: Fundamental Algorithms*(3rd ed.) ... (previous) ... (next): $\S 1.2.7$: Harmonic Numbers: Exercise $19$