Definition

[(2.7.7) The Alternating Series Test] Let \(a_n\) be a sequence satisfying
  1. \(a_1 \geq a_2 \geq a_3 ... \geq a_n \geq a_{n+1} \geq ... \) and
  2. \((a_n) \rightarrow 0\).
Then, the alternating series \(\sum_{n=1}^{\infty} (-1)^{n+1}a_n\) converges.


Discussion

We will prove this using the nested interval property by defining the following intervals

$$ \begin{align*} I_1 &= [a_1-a_2, a_1] \\ I_2 &= [a_1-a_2, a_1-a_2+a_3] \\ I_3 &= [a_1-a_2+a_3-a_4, a_1-a_2+a_3] \\ I_4 &= [a_1-a_2+a_3-a_4, a_1-a_2+a_3-a_4+a_5] \\ I_5 &= [a_1-a_2+a_3-a_4+a_5-a_6, a_1-a_2+a_3-a_4+a_5] \\ I_6 &= [a_1-a_2+a_3-a_4+a_5-a_6, a_1-a_2+a_3-a_4+a_5-a_6+a_7]. \\ ... \end{align*} $$

A few facts about these intervals. First, In each interval above, the start point is less than or equal to the end point. This is because when \(n\) is even, \(a_n\) is negative and so \(a_1 - a_2 \leq a_1\). Similarly \(a_1 - a_2 + a_3 - a_4 \leq a_1 - a_2 + a_3\) because \(a_4\) is negative. Second, we defined them in a way such that the length of each of these intervals is exactly \(a_{n+1}\). For example (\(a1 - (a_1 - a_2) = a_2\)). And finally, we defined them in a way such that \(I_1 \supseteq I_2 \supseteq I_3 ... \supseteq I_n\). This will allow us to use the nested interval property in order to come up with a candidate that the series will converge to. These intervals can be written compactly using the terms from the sequence of partial sums.

Proof

Let \((a_n)\) be a sequence satisfying the conditions above. To prove that \(\sum_{n=1}^{\infty} (-1)^{n+1}a_n\) converges, we need to prove that the sequence of partial sums converges. Let \((s_n)\) be the partial sum sequence where \(s_n\) is defined as

$$ \begin{align*} s_n = a_1 - a_2 + a_3 - a_4 + ... \pm a_n. \end{align*} $$

To do this, Let the following intervals be constructed using the terms from the sequence of partial sums such that,

$$ \begin{align*} I_{2k-1} &= [s_{2k}, s_{2k-1}] \\ I_{2k} &= [s_{2k}, s_{2k+1}]. \\ \end{align*} $$

We can also see that the length of each interval \(I_n\) is exactly \(a_{n+1}\). Also note here that \(I_1 \supseteq I_2 \supseteq I_3 ... \supseteq I_n\). Moreover, the length of these intervals will approach 0 as \(n\) approaches infinity. This can also be proved by the fact that \((a_n) \rightarrow 0\). Now, we can use the Nested Interval Property to conclude that the intersection of these intervals is not empty. In other words,

$$ \begin{align*} \bigcap_{n=1}^{\infty} I_n \neq \emptyset. \end{align*} $$

Therefore, let \(s \in \bigcap_{n=1}^{\infty} I_n\). We claim that the sequence of partial sums \((s_m)\) converges to \(s\) or \(\lim s_m = s\). To prove this, let \(\epsilon > 0\). We will find \(N \in \mathbf{N}\) such that whenever \(n \geq N\) we have

$$ \begin{align*} |s_n - s| < \epsilon. \end{align*} $$

To see this, we know that \(s \in I_n\) and also \(s_n \in I_n\) by how we constructed the intervals. So both \(s\) and \(s_n\) are members of \(I_n\). Furthermore, we know that the length of interval \(I_n\) is \(a_{n+1}\). Therefore, the distance between \(s\) and \(s_{n+1}\) can not be more than \(a_{n+1}\). In other words,

$$ \begin{align*} |s_n - s| \leq a_{n+1}. \end{align*} $$

But we also know that the sequence \((a_n)\) converges to 0. This means that for \(\epsilon > 0\), there exists an \(N \in \mathbf{N}\) such that when \(n \geq N\), we have \(|a_n| < \epsilon\). This means that \(|a_{n+1}| < \epsilon\) too. But \(a_{n+1} \geq 0\) so we can just write \(a_{n+1} < \epsilon\). Finally, we can see that

$$ \begin{align*} |s_n - s| \leq a_{n+1} < \epsilon. \end{align*} $$

as we wanted to show. \(\blacksquare\).

Other Definitions and Properties

  • Fo the absolute value function definition and other properties, see here.
  • For the definitions of sequences and what it means to for a sequence to converge, see this, for subsequences see this.
  • For the definitions of series, and what it means to for a series to converge, see this.
  • For the "show the limit" template and an example, see this.


References: