Tricki
a repository of mathematical know-how

Bounding the sum by an integral

Note iconIncomplete This article is incomplete. More examples are needed and more complicated situations should be analyzed.

Quick description

Suppose we want to estimate a sum

 \sum_{a<n<b} a_n

where the sequence \{a_n\} is a monotone sequence of non negative real numbers and a<b are real numbers. Later we can let for example b\rightarrow +\infty and likewise a\rightarrow -\infty but let us keep a,b finite for the moment. In this case is a standard tactic to look for a real function f with the same kind of monotonicity as the sequence \{a_n\}, such that f(n)=a_n for all the n we are interested in. Then we have

 \sum _{a<n<b} a_n \simeq \int_a ^b f(x) dx .

Usually the choice of the function should be obvious by looking at the sequence \{a_n\}. If for example the sequence \{a_n\} is explicitly given then the first obvious choice would be to replace the discrete parameter n with a continuous variable x and look at the resulting function f(x).

Prerequisites

calculus

Example 1

Let us look at the partial sums of the harmonic series

 \sum_{k=1} ^n \frac{1}{k}.

The sequence a_k=\frac{1}{k}, k=1,2,\ldots is a strictly decreasing sequence of positive numbers. Taking f(x)=\frac{1}{x} we recover the well known estimate

 \sum_{k=1} ^n \frac{1}{k}\simeq \int_1 ^n \frac{dx}{x}= \log n

Example 2

One can use the same technique to prove that for a positive real number p, the over-harmonic series

\sum_{k=1} ^\infty \frac{1}{n^p},

converge exactly when p>1 and we have the estimate

\sum_{k=1} ^\infty \frac{1}{n^p}\simeq  \frac{1}{p-1},

whenever p>1.

General discussion