Tutorials › Real Analysis › Proof of the Alternating Series Test

Infinite Series · Tutorial 526 of 1000

Proof of the Alternating Series Test

See why alternating partial sums converge by tracking their even and odd terms separately, and learn how those two subsequences locate the sum.

Advanced 9 min read

What You'll Learn

  • Prove convergence under the Alternating Series Test hypotheses using monotone even and odd partial sums
  • Show that the two subsequences approach the same limit
  • Establish how even and odd partial sums bracket the sum
  • Apply the proof to alternating series with decreasing magnitudes
  • Recognize why alternating signs alone do not guarantee convergence

The Proof Strategy: Separate the Partial Sums by Parity

The Alternating Series Test states that an alternating series converges when its nonnegative magnitudes are nonincreasing and tend to zero. The key to proving this is to separate the partial sums into those with an even number of terms and those with an odd number. Each subsequence moves in just one direction: the even partial sums increase, while the odd partial sums decrease.

These two subsequences do not merely converge independently. The difference between the odd and even partial sums with neighboring indices is exactly one term, whose magnitude tends to zero. Thus the two subsequences approach the same limit. Since every partial sum belongs to one of them, the full sequence of partial sums converges as well.

We use the positive-first form of the alternating series, \(\sum_{n=1}^{\infty}(-1)^{n-1}b_n\), with \(b_n\geq0\), \(b_{n+1}\leq b_n\), and \(b_n\to0\). Write \(S_0=0\) and \(S_N=\sum_{n=1}^{N}(-1)^{n-1}b_n\) for \(N\geq1\). For each \(k\geq1\), the even partial sum \(S_{2k}\) is followed by the odd partial sum \(S_{2k+1}\); the even and odd subsequences are \((S_{2k})\) and \((S_{2k-1})\), respectively.

Monotonicity of the Even and Odd Partial Sums

The change from one even partial sum to the next is a pair of terms. The change from one odd partial sum to the next is also a pair, but starts with a negative term. The hypothesis \(b_{n+1}\leq b_n\) gives the signs of both changes.

Lemma (Monotonicity and Separation of Partial Sums): Under the assumptions above, for every \(k\geq1\), $$ S_{2k-2}\leq S_{2k}\leq S_{2k-1} \quad\text{and}\quad S_{2k+1}\leq S_{2k-1}. $$ Moreover, $$ S_{2k-1}-S_{2k}=b_{2k}\longrightarrow0. $$

Proof. The change between consecutive even partial sums is $$ S_{2k}-S_{2k-2}=b_{2k-1}-b_{2k}\geq0, $$ because \(b_{2k-1}\geq b_{2k}\). Thus the even partial sums are nondecreasing. The change between consecutive odd partial sums is $$ S_{2k+1}-S_{2k-1}=-b_{2k}+b_{2k+1}\leq0, $$ because \(b_{2k+1}\leq b_{2k}\). Thus the odd partial sums are nonincreasing. Also, $$ S_{2k-1}-S_{2k}=b_{2k}\geq0. $$ Finally, \(b_{2k}\to0\) since \(b_n\to0\). These identities give all the stated inequalities and the separation limit. \(\square\)

The inequalities show why neither the even nor the odd partial sums need move toward the limit from the same direction: the even sums move upward, and the odd sums move downward. The distance between them shrinks to zero, so they close in on one another.

Proof of the Alternating Series Test

Theorem (Alternating Series Test): Let \((b_n)_{n=1}^{\infty}\) be a sequence of nonnegative real numbers such that \(b_{n+1}\leq b_n\) for every \(n\geq1\), and \(b_n\to0\). Then $$ \sum_{n=1}^{\infty}(-1)^{n-1}b_n $$ converges. The same conclusion holds if the first sign is negative.

Proof. First consider the positive-first series, and use the partial sums \(S_N\) defined above. The even partial sums are nondecreasing by the lemma. They are bounded above: for every \(k\geq1\), $$ S_{2k}\leq S_{2k-1}\leq S_1=b_1, $$ because the odd partial sums are nonincreasing. The even partial sums are therefore a nondecreasing sequence bounded above, so they converge to some real number \(A\).

The odd partial sums are nonincreasing. They are bounded below by zero, since $$ S_{2k-1}\geq S_{2k}\geq S_0=0. $$ Here the first inequality follows from the lemma, and the second follows because the even partial sums are nondecreasing from \(S_0=0\). Thus the odd partial sums converge to some real number \(B\). By the separation identity in the lemma, $$ S_{2k-1}-S_{2k}=b_{2k}\longrightarrow0. $$ Since \(S_{2k}\to A\) and \(S_{2k-1}\to B\), taking limits in this difference gives \(B-A=0\). Hence \(A=B\).

It remains to show that convergence of the even and odd subsequences to the same limit implies convergence of the full sequence. Let \(\varepsilon>0\). There are indices beyond which both \(|S_{2k}-A|<\varepsilon\) and \(|S_{2k-1}-A|<\varepsilon\). Every sufficiently late partial sum is either an even partial sum or an odd partial sum, so every sufficiently late \(S_N\) satisfies \(|S_N-A|<\varepsilon\). Therefore \(S_N\to A\), which proves convergence of the series.

If the first sign is negative, its \(N\)th partial sum is the negative of the \(N\)th partial sum for the positive-first series with the same magnitudes. The negative of a convergent sequence converges, so the negative-first series converges as well. \(\square\)

The Two Subsequences Bracket the Sum

The proof also tells us where the limit lies relative to the partial sums. The even partial sums increase to the limit, while the odd partial sums decrease to it. This gives a useful structural conclusion independent of any particular formula for the sum.

Theorem (Even–Odd Partial-Sum Bracketing): Under the hypotheses of the Alternating Series Test, let \(S\) be the sum of the positive-first series. Then for every \(k\geq1\), $$ S_{2k}\leq S\leq S_{2k-1}. $$

Proof. The even partial sums are nondecreasing and converge to \(S\). Therefore each even partial sum is at most \(S\): \(S_{2k}\leq S\). The odd partial sums are nonincreasing and converge to \(S\), so each odd partial sum is at least \(S\): \(S\leq S_{2k-1}\). Combining these inequalities proves the claim. \(\square\)

This bracketing says, in particular, that consecutive partial sums on either side of an even index can give lower and upper bounds: \(S_{2k}\leq S\leq S_{2k-1}\). The width of this bracket is $$ S_{2k-1}-S_{2k}=b_{2k}. $$ The Alternating Series Test’s hypotheses ensure that this width tends to zero. The Remainder Bound for a Convergent Alternating Series, established in the previous tutorial, gives a separate bound on the error after a specified number of terms.

Worked Examples

Worked Example: Bracketing a Series with Rational Magnitudes

Consider $$ \sum_{n=1}^{\infty}(-1)^{n-1}\frac{1}{2n+1}. $$ Set \(b_n=1/(2n+1)\). Since \(2n+3>2n+1>0\), $$ 0<b_{n+1}=\frac{1}{2n+3}<\frac{1}{2n+1}=b_n. $$ Also \(b_n\to0\). The Alternating Series Test therefore proves convergence.

The first two partial sums are $$ S_1=\frac13 \quad\text{and}\quad S_2=\frac13-\frac15=\frac{2}{15}. $$ The bracketing theorem gives $$ \frac{2}{15}\leq S\leq\frac13. $$ The gap between these bounds is \(\frac13-\frac{2}{15}=\frac{1}{5}\), which is exactly \(b_2\), the magnitude of the second term. Later even and odd partial sums give further brackets.

Worked Example: Even and Odd Partial Sums for a Square-Denominator Series

Consider $$ \sum_{n=1}^{\infty}(-1)^{n-1}\frac{1}{n^2+1}. $$ For every \(n\geq1\), $$ b_n-b_{n+1} =\frac{1}{n^2+1}-\frac{1}{(n+1)^2+1} =\frac{(n+1)^2+1-(n^2+1)}{(n^2+1)((n+1)^2+1)} =\frac{2n+1}{(n^2+1)((n+1)^2+1)}>0. $$ Thus \(b_n=1/(n^2+1)\) decreases, and \(b_n\to0\). The test applies.

Here \(S_1=1/2\) and \(S_2=1/2-1/5=3/10\), so \(3/10\leq S\leq1/2\). The next pair changes the even partial sum by $$ S_4-S_2=\left(\frac{1}{3^2+1}-\frac{1}{4^2+1}\right) =\frac{1}{10}-\frac{1}{17} =\frac{7}{170}>0. $$ The even partial sums move upward, as the proof predicts. The odd partial sums move downward; specifically, $$ S_3-S_1=-\frac15+\frac{1}{10}=-\frac{1}{10}<0. $$ Both subsequences converge to the same sum.

Worked Example: A Negative-First Alternating Series

Consider $$ \sum_{n=1}^{\infty}(-1)^n\frac{1}{\sqrt{n+5}}. $$ The magnitudes \(b_n=1/\sqrt{n+5}\) are positive, decrease because \(n+6>n+5\), and tend to zero. The Alternating Series Test applies to the positive-first series with these magnitudes; the series here is its negative and therefore converges.

The positive-first partial sums have even terms below the sum and odd terms above it. Negating all these quantities reverses the inequalities. In particular, if \(T\) is the sum of this negative-first series and \(T_N=-S_N\), then $$ -T_{2k-1}\ \text{is not the appropriate comparison; rather,}\quad T=-S,\qquad T_{2k}=-S_{2k},\qquad T_{2k-1}=-S_{2k-1}. $$ From \(S_{2k}\leq S\leq S_{2k-1}\), multiplication by \(-1\) yields $$ T_{2k-1}\leq T\leq T_{2k}. $$ Thus the order of the bounds reverses when the first sign is negative.

What the Hypotheses Do—and Do Not—Say

The proof depends on two distinct properties. Nonincreasing magnitudes make the even partial sums increase and the odd partial sums decrease. The limit-zero condition makes the distance between those subsequences vanish. Without that last condition, monotonicity alone need not make the two subsequences converge to the same value.

Worked Example: Decreasing Magnitudes That Do Not Tend to Zero

Consider $$ \sum_{n=1}^{\infty}(-1)^{n-1}\left(2+\frac{1}{n}\right). $$ The magnitudes decrease, since $$ \left(2+\frac1n\right)-\left(2+\frac{1}{n+1}\right) =\frac{1}{n(n+1)}>0. $$ But they tend to \(2\), not zero. The terms of the series therefore do not tend to zero: their absolute values tend to \(2\). By the Necessary Condition for Series Convergence, the series diverges. In terms of the proof, the gap between the odd and even partial sums is \(b_{2k}=2+1/(2k)\), which tends to \(2\), not to zero.

The test is a sufficient condition, not a claim that every convergent alternating series must have nonincreasing magnitudes. If the magnitudes fail to be nonincreasing, this proof no longer establishes the required opposite monotonicity of the even and odd partial sums; the test then gives no conclusion. Also, convergence by this test does not by itself establish absolute convergence. That requires a separate examination of the series of magnitudes.

Takeaway: Pairing terms makes the even partial sums nondecreasing and the odd partial sums nonincreasing. Their gap is a single magnitude \(b_{2k}\), so the limit-zero hypothesis forces them to share a limit and proves convergence.

Check Your Understanding

Use the parity argument and the bracketing theorem to answer the following questions.

  1. What is the difference \(S_{2k}-S_{2k-2}\), and why is it nonnegative?
  2. What is the difference \(S_{2k+1}-S_{2k-1}\), and why is it nonpositive?
  3. Why does \(b_n\to0\) imply that the even and odd partial-sum subsequences have the same limit?
  4. For a positive-first series, which partial sum gives the lower bound in the even–odd bracket?
  5. What does the Necessary Condition for Series Convergence tell you if the alternating term magnitudes decrease to a positive number?