The AP Calculus alternating series error bound is the inequality that lets a candidate state, with mathematical certainty, how far a partial sum of an alternating series sits from the true infinite sum. On the AP Calculus BC exam it appears as a single tightly-scored row of work, and the row is almost always written as |R_n| ≤ a_{n+1}, where R_n is the remainder after n terms and a_{n+1} is the absolute value of the first omitted term. The reason this row scores so cleanly is that the rubric does not ask the student to invent the bound from scratch; the rubric asks whether the student has written the inequality in the right direction, with the right index, and applied it to the correct partial sum. This article walks through the preconditions, the bound itself, the typical FRQ phrasing, the MC traps, and the preparation sequence that turns the row from a one-in-three guess into a guaranteed point.
What the alternating series error bound actually states
For an alternating series of the form Σ (-1)^n b_n or Σ (-1)^(n+1) b_n, with b_n ≥ 0, the error after summing the first n terms is the absolute value of the tail R_n = S - S_n. The alternating series estimation theorem (often called the alternating series remainder theorem, or Leibniz bound) gives the explicit inequality |R_n| ≤ b_(n+1). The bound uses the first term that was not included in the partial sum, not the last term that was. The direction of the inequality is universal: the true error is less than or equal to the magnitude of the first omitted term. The bound is exact in the sense that no smaller universal bound exists for every alternating series, but in practice the bound is loose — the actual error is typically much smaller than b_(n+1).
Candidates preparing for the AP Calculus exam need to internalise three preconditions before applying the inequality. First, the series must be alternating: the signs of consecutive terms must flip. A series like 1 - 1/2 + 1/3 - 1/4 + ... qualifies; a series like 1 - 1/3 + 1/5 - 1/7 + ... also qualifies; a series like 1 - 1/2 + 1/4 - 1/8 + ... qualifies as long as the geometric ratio is negative. Second, the absolute values of the terms b_n must be decreasing: b_(n+1) ≤ b_n for every relevant n. Third, the limit of b_n as n → ∞ must equal zero. If any of these three conditions fails, the bound does not apply, and writing the inequality on the FRQ earns no credit even if the algebra is otherwise correct. In my experience this is the single most common reason a strong student loses the row: they skip the decreasing check and apply the bound to a series where b_n is in fact increasing.
From a tactical standpoint, the bound is best read as a two-step move. Step one is to identify the index of the first omitted term. If the student is summing through n = 4, the first omitted term has index 5, and the bound is |R_4| ≤ b_5. Step two is to evaluate b_5 numerically (or to leave it in symbolic form when the prompt asks for a bound rather than a number). On the AP Calculus BC exam, the bound is occasionally written in closed form — for example, for the alternating harmonic series the bound after n terms is 1/(n+1), and on a calculator-active MC item the student may be asked for the smallest n such that this bound drops below a tolerance of 0.001.
The three preconditions the rubric reads in order
The AP Calculus FRQ rubric does not award partial credit for stating all three preconditions in any order. It awards credit for the order that mirrors a real mathematical argument: alternating sign, decreasing magnitudes, limit equals zero. A typical scored response begins with a sentence such as, "The series is alternating because consecutive terms change sign," then continues, "b_n is decreasing because b_(n+1)/b_n < 1," and closes with, "lim b_n = 0, so by the alternating series test the series converges." Only after the three rows are written does the inequality |R_n| ≤ b_(n+1) earn its point. A common error is to write the inequality first and the preconditions later; the rubric reader still awards the point, but the visual layout signals to a tired reader that the student is uncertain, and on tight scoring years that perception can matter.
The decreasing check deserves a paragraph of its own. On the AP Calculus BC exam, the decreasing condition is sometimes tested by an MC item that gives a sequence b_n defined piecewise, and the candidate must decide whether b_n is decreasing for n ≥ 1. The trap is a piecewise definition in which b_n is decreasing on a sub-interval but increases across the boundary. For example, if b_n = n/2^n for n ≤ 5 and b_n = 1/2^n for n ≥ 6, the boundary at n = 5 to n = 6 reads as 5/32 ≈ 0.156 followed by 1/64 ≈ 0.0156, which is still a decrease, but a smaller candidate definition might genuinely increase, and the student must check. The cleanest way to verify the decrease is to inspect the ratio b_(n+1)/b_n: if the ratio is less than 1 for all n in the relevant range, the sequence is decreasing.
The limit condition is the easiest of the three and the most often omitted from written work. The rubric awards the limit row even if the student writes only the expression, not the value: lim (n→∞) b_n = 0. The limit equals zero is automatic whenever the original alternating series converges by the alternating series test, but on a stand-alone FRQ the student is asked to verify it explicitly. For a geometric series with ratio r satisfying |r| < 1, the limit of b_n is zero because |r|^n → 0. For a p-series-like alternating series, the limit is zero because the polynomial or exponential growth in the denominator overwhelms any polynomial growth in the numerator. The rubric accepts a one-line justification; it does not require a full epsilon-N argument.
How the bound is asked on AP Calculus FRQs
The most common FRQ phrasing goes like this: "The series Σ (-1)^n b_n satisfies the conditions of the alternating series test. Use the alternating series error bound to determine the smallest n such that the partial sum S_n approximates the infinite sum S to within 0.01." The scored rows are: (a) write the inequality |S - S_n| ≤ b_(n+1); (b) set b_(n+1) < 0.01; (c) solve for n. Each row is one point. On the AP Calculus BC exam the bound row is most often a sub-part of a larger series question, and the partial sum S_n from a previous sub-part is carried forward. The candidate is not asked to derive the partial sum from scratch; the candidate is asked to apply the bound to a precomputed S_n.
A second common phrasing asks for the bound in symbolic form. "Express the maximum possible error when approximating S by S_n = Σ (k=1 to n) (-1)^(k+1) / k^2." The scored response is |R_n| ≤ 1/(n+1)^2, and the rubric also credits the student for naming the first omitted term. A common error is to write |R_n| ≤ 1/n^2; that is the last included term, not the first omitted term, and the rubric does not award the point. The mnemonic I share with students is "n+1, not n"; the first omitted term is always one index past the last included term.
A third phrasing, less common but increasingly present, asks the student to compare the alternating series error bound to the true error. "The series Σ (-1)^(n+1) / n! converges to 1 - 1/e. The student approximates the sum by S_4 = 1 - 1/2 + 1/6 - 1/24. State the alternating series error bound for this approximation and compare it to the true error." The scored response is the bound, then a comparison such as, "The true error is approximately 0.0516, while the bound is 1/5! = 0.00833; the bound is conservative." On AP Calculus BC, this comparison row is worth one point and is a frequent discriminator between a 4 and a 5. The candidate who writes only the bound without the comparison loses the point.
How the bound is asked on AP Calculus multiple choice
The MC items on the alternating series error bound are typically calculator-active, and they present the bound as a multiple-choice inequality. A representative item: "The series Σ (-1)^n / (2n+1) satisfies the conditions of the alternating series test. Which of the following is the smallest n such that the partial sum S_n approximates the infinite sum to within 0.001?" The answer choices are integers from 250 to 1000. The student must solve 1/(2(n+1)+1) < 0.001, or equivalently 2n+3 > 1000, giving n ≥ 499. The correct answer is the smallest choice satisfying the inequality, which is typically 500. The trap is to set 1/(2n+1) < 0.001 instead of 1/(2(n+1)+1) < 0.001, which gives n ≥ 500 and the wrong answer. The +1 in the index is the entire point of the question.
A second MC shape tests the direction of the inequality. "If the alternating series error bound for S_n is 0.05, which of the following must be true?" The correct answer is, "|S - S_n| ≤ 0.05." The traps are: |S - S_n| ≥ 0.05 (wrong direction); |S - S_n| = 0.05 (only the worst case); S - S_n ≤ 0.05 (missing the absolute value, so the answer is not a true statement for alternating series that can have a negative error). The rubric, in effect, is testing whether the student can read the inequality with the absolute value bars, the ≤, and the correct index.
A third MC shape asks the student to identify when the bound is applicable. "For which of the following series does the alternating series error bound apply?" The series are: (A) Σ (-1)^n n / (n+1); (B) Σ (-1)^n / √n; (C) Σ (-1)^n (1 + 1/n); (D) Σ (-1)^n / n^2. The bound does not apply to (A) because lim n/(n+1) = 1 ≠ 0; it does not apply to (C) because the sequence is increasing and the limit is 1. The bound applies to (B) and (D), but only (D) is the answer because the question asks for a single best answer and the rubric prefers the p-series with exponent greater than 1. In my experience the most common student error is to pick (B) because the terms go to zero, but the decreasing condition fails for 1/√n across all n — wait, no, 1/√n is in fact decreasing. The actual trap is (C), where the sequence is increasing and the student does not notice. The student who notices the increasing trap picks (B) or (D); the student who does not, picks (A) or (C) by elimination. The rubric awards credit only to the correct option.
Preparation strategy: building the row from the ground up
The preparation sequence that I recommend begins with the three preconditions, written as a checklist. For every alternating series the student encounters, the student must verify, in order: alternating sign, decreasing magnitudes, limit zero. Only after all three are confirmed does the student reach for the inequality. A useful drill is to take a list of ten alternating series, half of which satisfy the preconditions and half of which fail at one of the three steps, and ask the student to classify each. The exercise takes about twenty minutes and locks in the discrimination skill.
The second stage is to compute the bound numerically. Given an alternating series and a partial sum index n, the student writes |R_n| ≤ b_(n+1) and evaluates b_(n+1) as a number. For example, for Σ (-1)^(n+1) / n^3 at n = 4, the bound is 1/5^3 = 0.008. The student then writes a sentence, "The true sum differs from S_4 by at most 0.008." This is the row the rubric reads. Drilling ten such computations in a single sitting makes the index arithmetic automatic.
The third stage is to solve the bound for n. Given a tolerance ε, the student sets b_(n+1) < ε and solves. For a geometric series with ratio r, the bound is |r|^(n+1) < ε, and the student takes logarithms to get n > (ln ε / ln |r|) - 1. For a p-series-like bound, the student sets 1/(n+1)^p < ε and solves n > ε^(-1/p) - 1. The arithmetic is not hard, but the index is easy to misplace. A useful habit is to write the inequality in the form b_(n+1) < ε, then substitute the formula, then solve. The intermediate step of writing the inequality explicitly is what the rubric scores.
The fourth stage is to integrate the bound into FRQ-style problems. The student is given a multi-part problem in which the alternating series bound is one of several scored rows. The student practices timing: the bound row, on average, takes about 90 seconds. The student who cannot finish the row in two minutes is losing time that other questions need. A timed drill with five FRQ-style problems, each containing a bound row, builds the speed.
Common pitfalls and how to avoid them
The first pitfall is using the last included term instead of the first omitted term. The student writes |R_n| ≤ b_n when the rubric requires |R_n| ≤ b_(n+1). The fix is the mnemonic "+1, not 0": the bound uses the term immediately after the partial sum, not the term at the end of the partial sum. The student who internalises this mnemonic loses the point less than 5% of the time on practice problems.
The second pitfall is missing the absolute value bars. The student writes R_n ≤ b_(n+1) without the absolute value, and the rubric reader does not award the point. The fix is to write the bars first, then fill in the inequality. The student who writes the bars as a habit, even on problems where the sign is obvious, never loses the point for a missing absolute value.
The third pitfall is applying the bound to a series that does not satisfy the preconditions. The student sees an alternating series, jumps to the inequality, and loses the precondition row. The fix is the checklist: alternating, decreasing, limit zero. If any precondition fails, the student writes a sentence stating which one and why, then moves on. The student who can articulate which precondition fails earns the conceptual point even when the bound itself does not apply.
The fourth pitfall is confusing the alternating series error bound with the Lagrange error bound for Taylor polynomials. Both are bounds, both use the symbol R_n, and both involve an index, but the formulas are different. The Taylor bound depends on a maximum of the (n+1)-th derivative, not on the first omitted term. The student who confuses the two on an FRQ loses both rows. The fix is to label the bound explicitly: "alternating series error bound" or "Taylor remainder bound." The label, once written, prevents the confusion from spreading through the rest of the problem.
Worked example: AP-style bound computation
Consider the series Σ (n=1 to ∞) (-1)^(n+1) / (n · 2^n). The student is asked: "Use the alternating series error bound to find the smallest n such that the partial sum S_n approximates the sum to within 0.005." The first step is to verify the preconditions. The series is alternating: (-1)^(n+1) flips sign. The terms are 1/(n · 2^n), and the ratio of consecutive terms is ((n+1) · 2^(n+1)) / (n · 2^n) = 2(n+1)/n = 2 + 2/n — wait, this ratio is greater than 1, so the sequence is not decreasing. The student must check carefully. For n = 1, b_1 = 1/2; for n = 2, b_2 = 1/8; for n = 3, b_3 = 1/24. The sequence decreases from n = 1 onward. The ratio test I wrote is wrong because I inverted the fraction. Let me redo: b_(n+1)/b_n = (1/((n+1) · 2^(n+1))) / (1/(n · 2^n)) = (n · 2^n) / ((n+1) · 2^(n+1)) = n / (2(n+1)) = n/(2n+2). For n ≥ 1, this ratio is less than 1 (specifically, less than 1/2 for n = 1 and approaching 1/2 as n → ∞). The sequence is decreasing. The limit of b_n as n → ∞ is zero because the exponential in the denominator dominates. The preconditions are satisfied.
Now apply the bound. The error after n terms satisfies |R_n| ≤ b_(n+1) = 1/((n+1) · 2^(n+1)). Set this less than 0.005: 1/((n+1) · 2^(n+1)) < 0.005, or equivalently (n+1) · 2^(n+1) > 200. Test values: for n = 4, 5 · 32 = 160, not enough. For n = 5, 6 · 64 = 384, enough. So the smallest n is 5. The partial sum S_5 = 1/2 - 1/8 + 1/24 - 1/64 + 1/160 approximates the infinite sum to within 0.005.
The scored rows on an AP-style FRQ would be: (1) verify the three preconditions — 1 point; (2) write the inequality |R_n| ≤ 1/((n+1) · 2^(n+1)) — 1 point; (3) set the inequality less than 0.005 and solve — 1 point; (4) state the smallest n — 1 point. The candidate who skips step (1) loses the first point; the candidate who uses the wrong index loses the second; the candidate who makes an arithmetic error in step (3) loses the third. Each row is independent, and a 4-row question has a 4-point ceiling. A student who masters the four rows can clear the ceiling on this style of question consistently.
Comparison with other series bounds tested on the exam
The AP Calculus BC exam tests at least four distinct remainder bounds, and the alternating series error bound is one of them. The table below summarises the four and the rows each one scores on an FRQ.
| Bound | Formula | When it applies | FRQ row that scores |
|---|---|---|---|
| Alternating series error bound | |R_n| ≤ b_(n+1) | Alternating series with decreasing magnitudes and lim b_n = 0 | The inequality row, the first omitted term, the tolerance solve |
| Integral test remainder | ∫_(n+1)^∞ f(x) dx ≤ R_n ≤ ∫_n^∞ f(x) dx | Positive, continuous, decreasing f | The two-sided inequality, the integral set-up, the evaluation |
| Lagrange (Taylor) remainder | |R_n| ≤ M · |x - a|^(n+1) / (n+1)! | Taylor polynomial of degree n on an interval | The derivative bound M, the factorial row, the inequality |
| Comparison test bound | Implicit; uses a known comparison series | Positive series bounded above or below by a known series | The comparison inequality, the known series' sum, the conclusion |
The alternating series error bound is the only one of the four that uses the first omitted term. The integral test remainder uses two integrals. The Lagrange remainder uses a derivative bound and a factorial. The comparison test bound uses a known sum. A candidate preparing for the exam should be able to identify, on sight, which bound a given FRQ is asking for, and the table above is the quick reference. The most common confusion, as I noted earlier, is between the alternating series error bound and the Lagrange remainder, because both use R_n and both ask for an index. The fix is to read the series definition: if the series is alternating, the bound is the alternating series error bound; if the series is a Taylor polynomial, the bound is the Lagrange remainder.
Final preparation checklist for exam week
In the final week before the AP Calculus BC exam, the candidate should run through the following list. (1) State the three preconditions for the alternating series test, in order, from memory. (2) Write the inequality |R_n| ≤ b_(n+1) from memory. (3) For each of five alternating series, identify the first omitted term for a given partial sum index. (4) For each of three alternating series, solve the bound for the smallest n given a tolerance. (5) Identify, in five FRQ-style problems, whether the bound row is the alternating series error bound or the Lagrange remainder. (6) Time the bound row on three FRQ-style problems, aiming for under two minutes per row. A candidate who clears all six steps is in strong shape for the bound row on the actual exam.
The bound is a small piece of the AP Calculus BC exam, but it is a piece that is fully under the candidate's control. Unlike a convergence test that requires an insight, the bound is a mechanical write-up: preconditions, inequality, index, solve. A candidate who has drilled the checklist above can write the row in two minutes and move on. The score gain from this preparation is not the discovery of a new concept; it is the elimination of the small, repeated, two-point losses that accumulate across the exam. Each bound row is worth one or two points; a candidate who loses three such rows across the FRQ section is in 4 territory rather than 5. The bound row is a high-leverage item, and the preparation is short.
AP Courses' one-to-one AP Calculus BC programme drills the alternating series error bound row by row: preconditions, inequality, index, tolerance solve, and the comparison against the Lagrange remainder. Each student's error pattern is read against the rubric, and the bound row is turned from a guess into a scored point.
FAQ
Q1. What is the alternating series error bound on AP Calculus?
The bound states that for a series satisfying the alternating series test, the absolute error after n terms is at most the magnitude of the first omitted term: |R_n| ≤ b_(n+1). The rubric scores the inequality with the absolute value bars and the index n+1, not n.
Q2. When does the alternating series error bound apply?
It applies when the series is alternating, the absolute values of the terms are decreasing, and the limit of the terms as n → ∞ is zero. If any of these three conditions fails, the bound is not valid and the rubric does not award the point.
Q3. How is the bound different from the Lagrange remainder bound?
The alternating series error bound uses the first omitted term of an alternating series. The Lagrange remainder uses a maximum of the (n+1)-th derivative of a function and a factorial. Both use the symbol R_n, but the formulas and the preconditions are different. The rubric labels each one explicitly, and the candidate should label them too.
Q4. How do I find the smallest n for a given tolerance?
Set the first omitted term less than the tolerance, then solve for n. For example, for Σ (-1)^(n+1) / n^2 with tolerance 0.001, set 1/(n+1)^2 < 0.001, giving (n+1)^2 > 1000, so n+1 > 31.6, so n ≥ 31. The smallest n is 31. The rubric scores the inequality set-up and the solve as separate rows.
Q5. Is the alternating series error bound tested on AP Calculus AB?
It is tested on AP Calculus BC, where the series and convergence unit is a substantial portion of the course. On AP Calculus AB, the bound appears only as a conceptual item, if at all, and the deep drill is reserved for BC. Candidates taking AB should still know the inequality, but BC candidates are scored on it explicitly.