AP Calculus BC: Alternating Series Error Estimation

Approximating an infinite series with a finite sum is a practical necessity in engineering and science, but it introduces error. For alternating series—those whose terms alternate in sign—we are gifted with an exceptionally simple and powerful tool for quantifying this error. Understanding the Alternating Series Error Bound transforms you from someone who just calculates a sum to someone who can confidently state, "My approximation is accurate to within 0.0001."

The Nature of Alternating Series and Convergence

An alternating series is one whose successive terms have opposite signs. The most common form is ${\sum }_{n=1}^{\infty }(-1{)}^{n-1}{b}_n$ or ${\sum }_{n=1}^{\infty }(-1{)}^n{b}_n$, where $b_n>0$. Not all alternating series converge; their sum only approaches a finite limit if they satisfy specific conditions. The Alternating Series Test (also known as the Leibniz Test) provides these criteria: a series $\sum (-1{)}^{n-1}{b}_n$ converges if both 1) $b_{n+1}\le b_n$ for all $n$ (the terms are non-increasing in magnitude), and 2) ${\lim }_{n\to \infty }{b}_n=0$.

Consider the classic alternating harmonic series: $1-\frac{1}{2}+\frac{1}{3}-\frac{1}{4}+\frac{1}{5}-...={\sum }_{n=1}^{\infty }(-1{)}^{n-1}\frac{1}{n}$. Here, $b_n=\frac{1}{n}$. The terms clearly decrease ($\frac{1}{n+1}<\frac{1}{n}$) and approach zero, so the series converges (to $\ln 2$, though that fact isn't needed for error estimation). This test is your first checkpoint: the error bound we will study only applies to alternating series that pass this test.

The Alternating Series Error Bound Theorem

The power of a convergent alternating series lies in the behavior of its partial sums. Let $S$ be the true sum of the convergent alternating series $\sum (-1{)}^{n-1}{b}_n$, and let $S_n$ be its $n$th partial sum (the sum of the first $n$ terms). The theorem states:

The absolute error $|S-S_n|$—the difference between the true sum and your partial sum approximation—is less than or equal to the absolute value of the first omitted term. In mathematical terms:

$$|S-S_n|\le b_{n+1}$$

This is both intuitive and remarkable. It means the error you make by stopping at $S_n$ is no larger than the very next term in the series, $b_{n+1}$. Geometrically, as you calculate partial sums, they "bounce" back and forth, closing in on the true sum $S$. Each new partial sum overshoots or undershoots the limit, but by less than the previous jump. The error bound is a guaranteed ceiling for your approximation's inaccuracy.

For the alternating harmonic series $S=1-\frac{1}{2}+\frac{1}{3}-\frac{1}{4}+...$, approximating with $S_4=1-\frac{1}{2}+\frac{1}{3}-\frac{1}{4}\approx 0.5833$ gives an error bound of $b_5=\frac{1}{5}=0.2$. We can state that the true sum $S$ lies within the interval $(S_4-0.2,S_4+0.2)$, or approximately $(0.3833,0.7833)$. This is a wide range, reflecting that we've only used four terms.

Application 1: Determining Terms for a Desired Accuracy

The most common application is working backwards: given a required precision (e.g., error < 0.001), find the smallest number of terms $n$ you need to sum. You simply find the first term $b_{n+1}$ that is less than your allowed error.

Example: How many terms of the alternating series ${\sum }_{n=1}^{\infty }\frac{(-1{)}^{n-1}}{n^2}$ are required to approximate its sum with an error less than $0.0005$?

Here, $b_n=\frac{1}{n^2}$. We need $b_{n+1}<0.0005$. $$\frac{1}{(n+1{)}^2}<0.0005$$ Solving: $(n+1{)}^2>\frac{1}{0.0005}=2000$, so $n+1>\sqrt{2000}\approx 44.72$. Thus, $n+1\ge 45$, meaning $n\ge 44$. You must sum at least the first 44 terms. You would check $b_{45}=\frac{1}{45^2}\approx 0.000494$, which is indeed less than $0.0005$. Notice you solve for $n+1$, not $n$, because the error bound is in terms of the first omitted term.

Application 2: Estimating the Series Sum Within a Bound

Once you know how many terms to use, you can compute the partial sum $S_n$ and then state your final estimate as an interval: $S_n\pm b_{n+1}$. Often, the question will ask you to find $S_n$ and the associated error bound.

Example: Estimate the sum of ${\sum }_{n=1}^{\infty }\frac{(-1{)}^n}{2^n\cdot n}$ using the first five terms. Find the error bound and an interval containing the true sum.

First, identify the series in standard form: It is ${\sum }_{n=1}^{\infty }(-1{)}^n{b}_n$ where $b_n=\frac{1}{2^n\cdot n}$. We are using $S_5$. $$S_5=-\frac{1}{2\cdot 1}+\frac{1}{4\cdot 2}-\frac{1}{8\cdot 3}+\frac{1}{16\cdot 4}-\frac{1}{32\cdot 5}$$ Calculate stepwise: $-\frac{1}{2}=-0.5$ $+\frac{1}{8}=+0.125\to S_2=-0.375$ $-\frac{1}{24}\approx -0.041667\to S_3\approx -0.416667$ $+\frac{1}{64}=+0.015625\to S_4\approx -0.401042$ $-\frac{1}{160}=-0.00625\to S_5\approx -0.407292$

The error bound is $b_6=\frac{1}{2^6\cdot 6}=\frac{1}{64\cdot 6}=\frac{1}{384}\approx 0.002604$.

Therefore, the true sum $S$ is in the interval: $S_5-b_6<S<S_5+b_6$ $-0.407292-0.002604<S<-0.407292+0.002604$ $-0.409896<S<-0.404688$

You can report your estimate as $S\approx -0.407$ with an error less than $0.00261$.

Common Pitfalls

1. Applying the bound to a non-alternating series. This is the cardinal sin. The theorem only works for series that satisfy the conditions of the Alternating Series Test. If the series is not alternating or if $b_n$ is not decreasing, this error bound does not apply. Always verify the AST conditions first.
2. Confusing $n$ with $n+1$. The error bound is $b_{n+1}$, the term after the last one you summed. If you use $S_{10}$, the error bound is $b_{11}$, not $b_{10}$. When solving for the number of terms needed, you are solving the inequality $b_{n+1}<\text{(tolerance)}$.
3. Forgetting the absolute value. The theorem states $|S-S_n|\le b_{n+1}$. The error bound is a positive number representing a distance. When writing the interval for $S$, you correctly apply it as $S_n\pm b_{n+1}$.
4. Assuming the bound equals the error. The theorem gives an upper bound ($\le$), not necessarily the exact error. The actual error is often less than $b_{n+1}$, but $b_{n+1}$ is a safe, guaranteed ceiling. In many problems, especially on the AP exam, you will use the inequality $|S-S_n|<b_{n+1}$ (strictly less than), as the typical phrasing is "error less than" a given value.

Summary

• The Alternating Series Error Bound is a simple, powerful tool: for a convergent alternating series, the error in using the $n$th partial sum $S_n$ satisfies $|S-S_n|\le b_{n+1}$.
• Its primary applications are 1) determining the minimum number of terms needed to achieve a specified accuracy by solving $b_{n+1}<\text{tolerance}$, and 2) providing an interval estimate ($S_n\pm b_{n+1}$) for the true sum of the series.
• This bound applies only to series that pass the Alternating Series Test (terms are positive, decreasing, and approach zero). It is not valid for other types of series.
• Always remember that $b_{n+1}$ refers to the magnitude of the first omitted term, providing a guaranteed worst-case error scenario that is invaluable for ensuring precision in scientific and engineering calculations.