Upper Bound for Harmonic Number
Jump to navigation
Jump to search
Theorem
- $H_{2^m} \le 1 m$
where $H_{2^m}$ denotes the $2^m$th harmonic number.
Proof
- $\ds \sum_{n \mathop = 1}^\infty \frac 1 n = \underbrace 1_{s_0} \underbrace {\frac 1 2 \frac 1 3}_{s_1} \underbrace {\frac 1 4 \frac 1 5 \frac 1 6 \frac 1 7}_{s_2} \cdots$
where $\ds s_k = \sum_{i \mathop = 2^k}^{2^{k 1} \mathop - 1} \frac 1 i$
From Ordering of Reciprocals:
- $\forall m, n \in \N_{>0}: m > n: \dfrac 1 m < \dfrac 1 n$
so each of the summands in a given $s_k$ is less than $\dfrac 1 {2^k}$.
The number of summands in a given $s_k$ is $2^{k 1} - 2^k = 2 \times 2^k - 2^k = 2^k$, and so:
- $s_k < \dfrac {2^k} {2^k} = 1$
Hence the harmonic sum $H_{2^m}$ satisfies the following inequality:
\(\ds \sum_{n \mathop = 1}^{2^m} \frac 1 n\) | \(=\) | \(\ds \sum_{k \mathop = 0}^m \paren {s_k}\) | ||||||||||||
\(\ds \) | \(<\) | \(\ds \sum_{a \mathop = 0}^m 1\) | ||||||||||||
\(\ds \) | \(=\) | \(\ds 1 m\) |
Hence the result.
$\blacksquare$
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 $2$