The radius of convergence and the interval of convergence of a power series are two of the most tested deliverables on the AP Calculus BC free-response section. Every power-series prompt ultimately asks the candidate to convert an expression of the form ∑ cₙ (x − a)ⁿ into a single inequality on |x − a| and then decide, endpoint by endpoint, whether the resulting numerical series converges. A clean answer does two things in this order: it computes R from the ratio test (or root test, when the index is set up for it), and it then tests each endpoint of the open interval (a − R, a + R) by direct substitution. The AP Calculus BC rubric scores each row separately, and the order in which the work is presented is the order in which the reader awards points. The article below walks through that scoring order, with the algebra a reader is expected to recognise, the two endpoint cases that decide whether the closed interval is reported, and the tactical choices that separate a 5 from a 3.
What a radius of convergence actually is, and why the exam insists on both R and the interval
A power series ∑ cₙ (x − a)ⁿ is a function of x whose behaviour depends on how far x sits from the centre a. The radius of convergence R is the non-negative real number, or +∞, such that the series converges absolutely whenever |x − a| < R and diverges whenever |x − a| > R. On the open disc where |x − a| < R the series is a perfectly tame function — it can be differentiated and integrated term by term — but at the two boundary points x = a − R and x = a + R the test gives no information. The interval of convergence is therefore reported as a subset of the real line: it always contains (a − R, a + R), and it may or may not contain one or both endpoints, depending on what happens when each endpoint is substituted in.
AP Calculus BC treats this as a two-part deliverable, and a candidate who reports only the radius misses the harder half of the question. In practice on the FRQ, the prompt usually asks for the interval of convergence and uses the word "interval" deliberately. A score-5 answer will state R, write the inequality |x − a| < R, solve it to expose a − R < x < a + R, then substitute x = a − R and x = a + R into the original series. The two substitutions become two independent numerical series, each of which is tested by an appropriate convergence test from the syllabus. The reader's pencil moves down the page in this exact order, and the rubric mirrors it: ratio test, interval inequality, left endpoint, right endpoint, final interval.
It is worth naming the three structural cases a candidate can encounter. First, the closed case: both endpoints converge, so the interval is [a − R, a + R]. Second, the half-open case: one endpoint converges and the other diverges, producing an interval of the form [a − R, a + R) or (a − R, a + R]. Third, the open case: both endpoints diverge, so the interval stays as (a − R, a + R). A fourth case, R = 0, gives a single-point interval {a}, and R = +∞ gives (−∞, +∞); both are tested occasionally on the multiple-choice section. Knowing which of these four cases a given series will land in is half of the scoring battle, because the reader can predict the answer shape before any algebra is done.
The ratio test row: setting up L = lim |aₙ₊₁ / aₙ| correctly
The ratio test is the workhorse on the AP Calculus BC exam because almost every FRQ power series has a polynomial, exponential, or factorial factor in cₙ that collapses cleanly under aₙ₊₁ / aₙ. The limit the rubric actually scores is L = limₙ→∞ |cₙ₊₁ (x − a)ⁿ⁺¹ / cₙ (x − a)ⁿ|, and the row the reader marks is the simplified expression inside the limit, not the chain of cancellations that produced it. Candidates who rewrite the ratio three different ways before simplifying often lose the row to a sign error, and a sign error on the ratio test is a sign error on every downstream row.
The mechanical pattern, for a series ∑ cₙ (x − a)ⁿ, is to write aₙ = cₙ (x − a)ⁿ, aₙ₊₁ = cₙ₊₁ (x − a)ⁿ⁺¹, and form the absolute-value ratio. The factor (x − a)ⁿ⁺¹ / (x − a)ⁿ simplifies to |x − a|, and the factor cₙ₊₁ / cₙ depends on the closed form of cₙ. For a polynomial in n, the ratio cₙ₊₁ / cₙ tends to 1, and the entire limit reduces to |x − a|. For a factorial cₙ = 1/n!, the ratio cₙ₊₁ / cₙ = 1/(n+1), which tends to 0 and forces L = 0 regardless of x. For an exponential cₙ = 1/2ⁿ, the ratio is 1/2, and L = |x − a| / 2. The third case is the one that most often appears on the FRQ, and the limit is scored as the constant times |x − a| rather than |x − a| alone — candidates who forget the constant 1/2 and report L = |x − a| lose the radius row to a factor of two.
Once L is in the form k · |x − a| for some k > 0, convergence requires L < 1, which solves to |x − a| < 1/k. The radius is R = 1/k. Candidates should write this final inequality on its own line and circle R; on a multi-part FRQ the next row reads the radius from that circle, and a missing R cascades into a missing interval. If k = 0, the inequality is 0 < 1, which is always true, and R = +∞. If the algebra inside L produces something that does not depend on x, the radius is the constant 1 over that constant — a small but common stumbling block. The same scaffolding applies if the prompt specifies the root test; the limit is the n-th root of |cₙ| · |x − a|, and the same separation into "depends on x" and "does not" drives the radius.
Solving the inequality and writing the open interval
After R has been named, the next scored row is the solution of |x − a| < R as an explicit inequality on x. For R > 0 this expands to −R < x − a < R, and adding a throughout gives a − R < x < a + R. The rubric reads the centre a from the original series and the radius R from the previous row, so the line a − R < x < a + R is the deliverable. Candidates who write |x − a| < R and stop have not yet produced the interval, and the row that scores "interval as a − R < x < a + R" remains open.
Several notational traps cost marks here. The first is forgetting the absolute value: writing x − a < R on its own, or worse, x < a + R without the lower bound. The second is sign error on a − R when a is negative; a candidate who computes, say, R = 3 and a = −2 must write −5 < x < 1, not −1 < x < 5. The third is dropping the strict inequality. The endpoints have not been tested yet, so the open form is the only correct form on this row; a closed bracket on a − R at this stage is treated by the reader as a forward reference to a later row, and if the endpoint diverges the closed bracket becomes a deduction. A clean convention: state (a − R, a + R) on the interval row, then test the endpoints, then update the brackets in the final interval row.
For R = 0 the inequality is |x − a| < 0, which has no real solutions, and the interval is the singleton {a}. For R = +∞ the inequality holds for all real x, and the interval is (−∞, +∞). These two boundary cases are rare on the FRQ but appear on multiple choice, and writing them down explicitly is the kind of low-effort, high-payoff move that recovers points on questions the rest of the cohort skips. The interval row is the row that converts abstract radius into the line the test will draw at the end of the problem; getting the brackets right at this stage sets up the endpoint rows cleanly.
Testing the left endpoint: a separate series, a separate row
Once the open interval is in hand, the next scored deliverable is the left endpoint x = a − R. Substituting this value into the original series produces a numerical series ∑ cₙ (−R)ⁿ (or, more generally, ∑ cₙ (a − R − a)ⁿ = ∑ cₙ (−R)ⁿ). The candidate then chooses a convergence test from the syllabus — alternating series test, p-series, comparison, integral — and writes a one-line verdict: converges or diverges. Each endpoint is its own row on the rubric, and the verdict is what scores. The algebra of the substitution rarely scores; the named test and the conclusion do.
The most common left-endpoint substitution produces an alternating numerical series of the form ∑ (−1)ⁿ bₙ, where bₙ is a positive, decreasing sequence that tends to 0. This is exactly the shape the alternating series test accepts, and the rubric will give full credit for naming the test, showing that bₙ is decreasing (often by inspection for a 1/n or 1/n² type term), and stating lim bₙ = 0. Candidates who try to apply the ratio test to a numerical series at this stage are usually using a tool the rubric does not score for, and the row is read as "test not applicable" or, worse, "no test named". The series is no longer a power series; the ratio test has done its job and stepped out of the picture.
If the substitution at the left endpoint produces a p-series ∑ 1/nᵖ, the verdict is read directly from p > 1 (converges) or p ≤ 1 (diverges). A geometric series at an endpoint is scored as convergent if |r| < 1 and divergent otherwise, and the row asks for both r and the comparison. If the substitution produces a series with terms that do not tend to 0 — for instance, terms that tend to a non-zero constant because the n-th power of (−1) or 1 does not shrink the magnitude — the n-th term test is the right test, and the conclusion "diverges because lim aₙ ≠ 0" is the full row. Whichever test is chosen, the verdict at the left endpoint is independent of the right endpoint, and the rubric awards the two endpoint rows independently. A common error is to assume the series is symmetric: it almost never is, because the centre a is rarely 0 and the parity of n in the (x − a) factor is rarely symmetric across the two endpoints.
Testing the right endpoint: a different test, possibly a different verdict
The right endpoint x = a + R substitutes into the series as ∑ cₙ Rⁿ, with no sign flip. The verdict at this endpoint is decided by a different test in most cases, because the resulting numerical series has all-positive (or all-negative) terms, and the alternating series test is no longer available. The typical decision is between a p-series comparison, a geometric-series comparison, and a comparison or limit-comparison test against a known series.
For the typical exam series, the right endpoint becomes ∑ cₙ Rⁿ with cₙ decaying like 1/n or 1/n² and R a small positive number, so the series looks like a scaled p-series. The rubric's preferred test here is usually a direct comparison or limit comparison against ∑ 1/nᵖ, because the candidate is expected to recognise the asymptotic shape of cₙ Rⁿ. Naming the comparison, writing the inequality, and stating the conclusion is the full row. A common scoring miss is to write "the series converges by the alternating series test" at the right endpoint, where all terms have the same sign; the reader marks the test row down and the conclusion row down, and the candidate loses both.
The right endpoint is also the endpoint where the n-th term test is most often the correct test. If cₙ Rⁿ does not tend to 0 — for instance, if R is large enough that cₙ Rⁿ → 1, or if cₙ does not decay fast enough — the verdict is divergence, and the test is the n-th term test. The rubric scores the named test and the conclusion "diverges because lim aₙ ≠ 0". A useful diagnostic: at each endpoint, compute limₙ→∞ aₙ first. If the limit is not 0, divergence is automatic and the candidate saves a minute by not searching for a deeper test. If the limit is 0, the test is one of comparison, alternating series, or integral.
Writing the final interval: closing the brackets in the order the rubric reads
The final interval is reported as a single line of answer, in interval notation, with brackets or parentheses set by the endpoint verdicts. The convention on the FRQ is to write the interval last, in a clearly labelled line, and to use a closed bracket on each endpoint that converged and a parenthesis on each endpoint that diverged. The reader scans this line and confirms it is consistent with the two endpoint rows above; any mismatch is a deduction on the final-interval row.
Three worked shapes cover the bulk of AP Calculus BC prompts. The first is the closed case: both endpoints converge, and the answer is [a − R, a + R]. This shape is the most common on prompts built around ∑ (x − a)ⁿ / n or similar, where the right endpoint becomes a divergent harmonic series and the left endpoint becomes a convergent alternating harmonic series — a half-open case in fact, scored as (a − R, a + R] or [a − R, a + R) depending on which endpoint converges. The second is the half-open case: one endpoint converges and the other diverges, giving one closed and one open bracket. The third is the open case: both endpoints diverge, and the interval is (a − R, a + R).
For prompts where R = +∞, the final interval is (−∞, +∞) and no endpoint testing is required. For R = 0, the final interval is {a}, again without endpoint rows. The rubric's design is that endpoint rows are scored only when they exist; a candidate who invents an endpoint test for an infinite-radius series loses time but does not lose points, because the rows are read as "not applicable". The same tolerance applies to the singleton case. Candidates who try to apply the alternating series test to an all-positive endpoint series, by contrast, do lose the endpoint row, because the test is named and the rubric marks it as not satisfied. The discipline is to name the test that fits the sign of the series and the size of the terms.
Common pitfalls and how to avoid them
The first pitfall is dropping the absolute value in the ratio test. The limit the rubric scores is L = lim |aₙ₊₁ / aₙ|, and writing L = lim aₙ₊₁ / aₙ without the absolute value bars loses the row the moment aₙ₊₁ / aₙ is negative. The fix is mechanical: wrap the ratio in absolute value bars before taking the limit, and never strip them on the way to a simplification.
The second pitfall is forgetting the constant that lives next to |x − a| in L. When cₙ is geometric, factorial, or polynomial, the ratio cₙ₊₁ / cₙ contributes a constant multiplier that the rubric expects to see inside L. Reporting L = |x − a| when the correct L is |x − a| / 2 changes R from 2 to 1 and propagates the error through the interval. The fix is to write L in the form (constant) · |x − a| before solving the inequality, and to read R from the reciprocal of the constant.
The third pitfall is applying the alternating series test to an all-positive series. The test requires alternating signs, and a numerical series whose terms are all positive or all negative cannot satisfy that precondition. The rubric marks the test row as not satisfied and the conclusion row as unsupported. The fix is to inspect the sign of the first term of the substituted series and choose the test accordingly: alternating signs go to the alternating series test, all-positive terms go to p-series or comparison.
The fourth pitfall is writing a closed bracket before the endpoint has been tested. Brackets are decisions about the endpoints, and they belong on the final-interval row, not on the open-interval row. The fix is to keep the open-interval row in parentheses and to update the brackets in a separate, clearly labelled final line. Candidates who mix the two rows lose the final-interval row even when their endpoint verdicts are correct.
The fifth pitfall is a sign error in the centre. When a is negative, a − R is more negative than −R, and a + R is closer to zero. A candidate who reads a as 0 when the series is centred at, say, x = −2 produces an interval that is symmetric about the wrong point. The fix is to copy the centre directly from the original series into the open-interval line, without re-reading it from a calculator or from memory.
Worked example: a typical AP Calculus BC power-series prompt
Consider the series ∑ₙ₌₁^∞ (x − 2)ⁿ / (n · 3ⁿ). The first step is the ratio test: aₙ = (x − 2)ⁿ / (n · 3ⁿ), aₙ₊₁ = (x − 2)ⁿ⁺¹ / ((n+1) · 3ⁿ⁺¹), and the absolute-value ratio is |x − 2| / 3 · n / (n+1). The factor n / (n+1) tends to 1, so L = |x − 2| / 3. The convergence condition L < 1 gives |x − 2| < 3, hence R = 3 and the open interval is (2 − 3, 2 + 3) = (−1, 5).
The left endpoint is x = −1, substituting (x − 2) = −3, so the numerical series is ∑ (−3)ⁿ / (n · 3ⁿ) = ∑ (−1)ⁿ / n. This is the alternating harmonic series, the terms 1/n are positive and decreasing, and lim 1/n = 0. The alternating series test applies, and the verdict is converges.
The right endpoint is x = 5, substituting (x − 2) = 3, so the numerical series is ∑ 3ⁿ / (n · 3ⁿ) = ∑ 1/n. This is the harmonic series, all-positive terms, no alternating structure. The p-series test gives p = 1, hence diverges. Equivalently, the n-th term test gives lim 1/n = 0, so the test is inconclusive, and the comparison is needed to reach divergence — the p-series test is the more efficient row to write.
The final interval is [−1, 5), reflecting a closed left bracket and an open right bracket. The rubric reads: ratio row, R row, open interval row, left endpoint test row, left endpoint verdict row, right endpoint test row, right endpoint verdict row, final interval row. That is eight rows on a single prompt, and each one is a place to score or to lose a point.
| Step | What is written | What the rubric scores |
|---|---|---|
| Ratio test limit L | L = |x − 2| / 3 | Simplified limit with absolute value |
| Solve L < 1 | |x − 2| < 3, so R = 3 | Radius R |
| Open interval | −1 < x < 5 | Inequality in x |
| Left endpoint x = −1 | ∑ (−1)ⁿ / n, alternating series test, converges | Substitution, named test, verdict |
| Right endpoint x = 5 | ∑ 1/n, p-series with p = 1, diverges | Substitution, named test, verdict |
| Final interval | [−1, 5) | Brackets consistent with endpoint verdicts |
Exam-day strategy: time budget, layout, and the order of the rows
A power-series FRQ on the AP Calculus BC exam is typically a 15-minute question, and the budget that holds up under pressure is roughly two minutes on the ratio test, one minute on the open interval, four minutes on the two endpoints, and one minute on the final interval. The remaining time is a buffer for algebra slips. Candidates who spend six minutes on the ratio test, which is the most algebra-heavy row, often run out of time on the endpoints, and the endpoint rows together are worth as much as the ratio row on most prompts. The trade is to do the ratio test efficiently, write the inequality in one line, and move on.
The layout matters because the reader is reading downward. The candidate should present the work in the order the rubric reads it: ratio limit, radius, open interval, left endpoint substitution, left endpoint test, left endpoint verdict, right endpoint substitution, right endpoint test, right endpoint verdict, final interval. Putting the final interval at the very bottom, in a clearly labelled line, lets the reader match the brackets to the verdicts above. Labelling each step ("Left endpoint:", "Right endpoint:", "Final interval:") saves the reader's time and the candidate's points, because an unlabelled test sitting in the middle of a paragraph is read as an unplaced attempt and the row is scored as missing.
Two tactical choices separate a 5 from a 4 on a power-series prompt. The first is to name the test at each endpoint, explicitly, by name. "Converges by the alternating series test" scores the test row; "converges" alone does not. The second is to keep the closed and open brackets on the final interval consistent with the endpoint verdicts. A candidate who writes [−1, 5] when the right endpoint diverges, or (−1, 5] when the left endpoint converges, loses the final-interval row even if every other row is correct. The discipline is to update the brackets only after both endpoint verdicts are on the page.
What to practise in the weeks before the exam
The single most productive drill is to take ten power series, compute the radius and interval of convergence for each, and present the work in the rubric's order. The series should be chosen to cover the four cases — closed, half-open, open, infinite — and the centre should vary across positive, negative, and zero. A candidate who can run a clean ratio test on a polynomial-in-n cₙ, write the open interval, test both endpoints, and report a final interval in under 12 minutes is operating at a 5 level. The next-best drill is to time the ratio test alone, on a fresh series, with the goal of producing L and R in under three minutes; the ratio test is the bottleneck on most prompts, and shaving time there frees up the endpoint rows.
The MCQ section tests the same content with smaller algebra payloads, and a useful practice is to compute only R for a series and check the sign of L = 1 boundary behaviour from the structure of cₙ, without doing the endpoint algebra. For most MCQ prompts, the answer is among three intervals, and recognising the shape of cₙ — factorial, geometric, polynomial-in-n — narrows the bracket choices quickly. The full interval is rarely required on MCQ, but the radius is, and the radius can be read off the ratio test in one line.
Finally, candidates should review the syllabus list of convergence tests and be ready to name, for any numerical series, which one applies. The list on the BC exam is finite: nth-term test, integral test, p-series, geometric series, comparison, limit comparison, alternating series. Each endpoint is a numerical series, and the test chosen must match the sign and the limit behaviour of the terms. A candidate who has practised matching tests to series shapes will name the right test at the right endpoint on the first try, and the endpoint rows are scored independently of the ratio row — so endpoint accuracy is a recoverable source of points even on a prompt where the ratio row went sideways.
Conclusion and next steps
Radius of convergence and interval of convergence form a single, multi-row deliverable on the AP Calculus BC free-response section, and the score-5 answer is the one that presents its work in the order the rubric reads: ratio test limit, radius, open interval, left endpoint substitution with named test and verdict, right endpoint substitution with named test and verdict, and a final interval whose brackets match the two verdicts. The two endpoint rows are where most of the recoverable points live, and they are also where most candidates lose them, usually by naming the wrong test or by closing the brackets too early. Drilling the rubric's row order, on a fresh series every other day for the last three weeks before the exam, is the preparation that converts a 3 into a 5 on this question type. AP Courses' AP Calculus BC tutoring programme scores each student's radius-and-interval work against the published rubric, names the row on which points are being lost, and turns the eight-row deliverable into a single, repeatable procedure.