3 tests that pin down the radius of convergence of any power series

Radius and interval of convergence sit in a small corner of the AP Calculus BC syllabus, but they ask a candidate to do something most other units do not: hold a single algebraic expression in two different ways at once. The series must be analysed globally, to find a radius, and then locally, to test each endpoint of the resulting interval. That double demand is the reason the topic appears on the BC exam as a free-response staple, and the reason students who race through Unit 10 of the College Board framework often lose points they expected to keep. This article walks through the working method I teach for these problems, the three tests that decide nearly every radius-of-convergence answer, and the endpoint logic that turns a half-correct solution into full marks. Throughout, the framing borrows deliberately from LNAT preparation habits, because the same disciplined reading of a passage — isolate the claim, then test the boundary — translates almost one-for-one into the way a careful student should attack a power series on test day.

What "radius" and "interval" of convergence actually mean

A power series in x has the form ∑ aₙ (x − c)ⁿ, where aₙ are the coefficients, x is the variable, and c is the centre of the series. For any fixed value of x the series becomes an ordinary numerical series, and the central question is: for which x does it converge, and for which x does it diverge? The answer is never a scattered set of points; it is always a single interval, which can degenerate to a point or expand to the entire real line. The radius of convergence, conventionally written R, is the distance from the centre c to the nearest point where the series stops converging. The interval of convergence is the actual set of x values, including the centre and possibly the endpoints, where the series converges.

This is the bit students often miss. The radius is a non-negative number, possibly zero, possibly infinite, and it is measured in the same units as x. The interval of convergence is a set, written in the form (c − R, c + R), [c − R, c + R), (c − R, c + R], or [c − R, c + R], depending on what happens at the two boundaries. The radius answers "how far can I go from the centre?"; the interval answers "given that distance, what specific x values are inside?" Both questions appear on the AP Calculus BC exam, and the College Board's scoring guidelines routinely award a point for the radius and a separate point for each correctly classified endpoint.

A useful mental picture is a torch beam pointed at a wall. The torch sits at c, the beam has half-width R, and the lit-up strip on the wall is the open interval. Endpoints behave like the very edges of the beam: they may or may not actually catch light, and you have to test them individually. The arithmetic that produces R is almost always mechanical; the judgement that decides the interval lives at the boundary, and that is where most lost points accumulate.

The three convergence tests that decide nearly every AP problem

For AP Calculus BC purposes, three tests cover roughly 95% of radius-of-convergence prompts. A student who is fluent with all three, and who knows when to switch between them, will rarely face a problem that does not bend to one of them. The tests are the ratio test, the root test, and direct comparison with a known geometric or p-series.

Ratio test as the default

The ratio test is the workhorse. Given ∑ aₙ (x − c)ⁿ, compute the limit L of |aₙ₊₁ / aₙ|, multiplied by |x − c|. Converges when L < 1, diverges when L > 1, and is inconclusive when L = 1. The condition L < 1 is what produces the radius. Solve |x − c| < 1 / (limit) to get the open interval, then probe the two endpoints separately. For an AP free-response, write out the limit calculation explicitly; the readers want to see the absolute value signs, the cancellation, and the final inequality in x.

Root test when factorials meet powers

When the general term contains a power of n alongside a power of (x − c), the root test is often faster than the ratio test. The form nth-root(|aₙ (x − c)ⁿ|) is the natural one, and factorials disappear cleanly because nth-root(n!) approaches 1 as n grows. The College Board framework labels this as an alternative method, and a student who recognises the trigger — terms like nⁿ or nⁿ inside the coefficient — can save a full algebraic minute per problem.

Comparison with a known series as a sanity check

Direct comparison with the geometric series ∑ rⁿ, which converges when |r| < 1, is the third route. It rarely produces the radius on its own in AP problems, but it is the cleanest way to justify what happens at an endpoint. If a candidate substitutes x = c + R and gets a series whose behaviour is obvious, comparison lets them argue convergence or divergence without invoking integral or alternating-series machinery.

Working a radius-of-convergence problem step by step

Take the series ∑ (x − 3)ⁿ / (n · 2ⁿ), a textbook AP-style example. The first move is mechanical: set up the ratio test by writing aₙ = 1 / (n · 2ⁿ). Then aₙ₊₁ / aₙ simplifies to n / ((n + 1) · 2), which tends to 1/2 as n grows. Multiplying by |x − 3| gives L = |x − 3| / 2. Convergence requires |x − 3| / 2 < 1, which simplifies to |x − 3| < 2. The radius is 2, and the open interval is (1, 5).

The second move is the one most candidates skip or rush. Substitute x = 1, giving ∑ (−2)ⁿ / (n · 2ⁿ) = ∑ (−1)ⁿ / n. This is the alternating harmonic series, which converges by the alternating series test. Substitute x = 5, giving ∑ 2ⁿ / (n · 2ⁿ) = ∑ 1/n, which is the harmonic series and diverges. The interval of convergence is therefore [1, 5). On the AP exam this is the difference between earning 1 point (radius only) and earning 4 points (radius plus three endpoint decisions).

The pattern holds for any centre c. Converges for |x − c| < R, then test x = c − R and x = c + R separately, using the simplest test that applies at each endpoint. The LNAT parallel is deliberate: the LNAT multiple-choice section rewards the candidate who first locates the structural feature of a passage and then probes its boundary cases. Power-series problems reward the candidate who first locates the radius and then probes its boundary points. Same cognitive move, different subject.

Endpoint behaviour: the place where most points are lost

Three endpoint archetypes cover almost every AP prompt. A candidate who can recognise each by sight, and who knows which test to reach for, will not be ambushed by a College Board free-response. The archetypes are: the alternating geometric or alternating p-series, the non-alternating geometric or p-series, and the series that reduces to a constant non-zero term.

Alternating archetypes

When the boundary value of x produces a factor of (−1)ⁿ, the alternating series test is usually decisive. Convergence requires the absolute value of the terms to be decreasing and to approach zero. The classic trap is to forget the "decreasing" condition; a candidate who writes "terms go to zero, so it converges" is implicitly assuming monotonic decrease, which is a separate step and must be shown or at least noted.

Non-alternating archetypes

When the boundary value produces only positive terms — for example x = 5 in the example above — comparison with a known series is the natural test. The integral test, where applicable, is also acceptable on the AP, but comparison is faster and earns the same credit. If the candidate can rewrite the boundary series as a multiple of a p-series or a geometric series, that single observation is enough to justify divergence or convergence in one line.

Constant-term archetypes

When the boundary value makes |x − c| = 1 and the coefficient does not contain n in the denominator, the series becomes a sum of constants, which diverges by the nth-term test. The candidate who writes "the terms do not approach zero" has just earned the point. This is the easiest endpoint to handle, and the one most often missed because students assume endpoint questions must involve clever tests.

The scoring pattern in AP Calculus BC free-response makes the asymmetry plain. The radius earns one point, and each of two endpoints earns one point. A student who gets the radius right but blows both endpoints loses three of the four available points on the problem. Drilling endpoint archetypes for a single focused week, working twenty problems with the endpoint as the only variable, closes that gap faster than any other intervention I have seen.

Common pitfalls and how to avoid them

Five errors account for the majority of lost points on radius and interval of convergence problems. The first is forgetting absolute values around the ratio. The ratio test requires |aₙ₊₁ / aₙ|; without the absolute value, the limit can be negative and the inequality L < 1 loses meaning. The second is cancelling x − c before taking the limit; the limit as n grows should leave |x − c| untouched, because x is fixed during the limit, only n varies. The third is treating R as a half-interval width when the centre is non-zero; the centre shifts the interval horizontally, and students who assume c = 0 will write the wrong interval. The fourth is forgetting to test the endpoints at all, having decided that the ratio test "covers everything". It does not. The fifth is mixing up the alternating series test conditions and writing "terms go to zero" as if that alone implies convergence. The terms of an alternating series must decrease in absolute value; zero alone is not enough.

Why this topic is also a strong LNAT training ground

LNAT preparation and AP Calculus preparation look unrelated on the surface, but the working method that produces a clean radius-of-convergence answer is the same method that produces a clean LNAT multiple-choice answer. Both demand that the candidate read a structured stimulus — a passage in one case, a series in the other — and identify the moving part that controls the result. In an LNAT argument question, the moving part is usually the conclusion of an implicit argument; in a radius problem, the moving part is the absolute value of x − c. Once the moving part is isolated, the candidate applies a small fixed test: for LNAT, it is a six-step argument audit; for power series, it is one of the three convergence tests. The boundary check is the part that distinguishes a good answer from a complete one. In an LNAT essay, the candidate must also test the boundary of their argument — what happens at the strongest objection? — before the response counts as more than adequate. AP students who have been drilled to probe endpoints will find that LNAT essay revision is the same cognitive habit applied to a different surface.

For LNAT scoring, the multiple-choice section is scaled to a 0–42 range, and a candidate who reliably reaches the boundary cases of an argument gains the small but consistent edge that separates a 28 from a 32. For AP Calculus, the BC exam scores the free-response section on a 0–9 per-question scale, and the student who handles endpoints correctly on a convergence problem gains the same kind of edge. The shared habit is the habit of finishing the test, not skipping the last 10% of the question. The LNAT preparation strategy that drills the final objection is, mechanically, the same strategy that drills the endpoint.

Drills that move a candidate from "finds the radius" to "writes the full interval"

A four-week drill plan produces measurable improvement on this topic. In week one, the candidate works fifteen ratio-test problems with the centre c = 0, writing out the limit in full each time. No endpoints are tested. The aim is mechanical fluency with the algebraic pattern. In week two, the candidate works another fifteen problems, this time with the centre shifted to non-zero values such as c = 2 or c = −3. The aim is to internalise the horizontal shift. In week three, the candidate introduces endpoints: ten problems with full interval-of-convergence solutions, with the ratio test leading to the radius and a separate endpoint test at each boundary. In week four, the candidate works ten mixed problems under timed conditions, with a soft cap of ten minutes per problem, mirroring AP free-response pacing.

The LNAT-equivalent drill plan is structurally similar. In week one, the candidate works argument-analysis questions in isolation, identifying the conclusion and the main premise. In week two, the candidate adds the step of testing the strongest objection. In week three, the candidate works essay prompts with a five-minute planning window that explicitly allocates time to the boundary case. In week four, the candidate works full-length mixed sections under timed conditions. Both plans exploit the same principle: build the easy half of the skill to fluency, then layer the harder half on top.

For most candidates the breakthrough moment in week three is the realisation that the endpoint test is not an extra task; it is a different task that uses the same setup. The candidate has already computed R, and the two values x = c − R and x = c + R are sitting in front of them. The endpoint substitution is a single line, and the test is a single further line. What looked like a doubling of the work is actually a 30% addition. That realisation usually arrives in week three, which is why the plan reserves week three for full interval problems rather than scattering them across all four weeks.

Reading the AP problem stem: a checklist that prevents the last-minute error

AP Calculus BC free-response prompts for this topic typically include three phrases that change the answer. "Find the interval of convergence" means endpoints must be classified. "Find the radius of convergence" means only the radius is required, and the candidate can stop there. "Find the values of x for which the series converges" is functionally identical to "find the interval of convergence". Reading the verb carefully costs two seconds and saves a point. A second phrase to watch is "in interval notation"; this is a College Board instruction that requires a specific answer format, and a candidate who writes R = 3 and stops has not answered the question. A third is "include reasoning for the endpoints" or its equivalent; this is the explicit signal that endpoint work will be scored, and skipping it forfeits the points.

The LNAT analogy is the question-stem audit. LNAT multiple-choice stems contain words like "weaken", "strengthen", "assumption", "flaw", "must be false", and "most strongly supported". Each of these words changes what counts as a correct answer. A candidate who reads "assumption" but answers as if the question asked "conclusion" loses the point for a misread, not for a missing skill. The same diagnostic — re-read the verb, identify the format, allocate the work — applies to both exams, and the student who develops the habit in one subject will carry it into the other.

Pulling it together: a single worked example end to end

Work ∑ (n · (x + 2)ⁿ) / 4ⁿ. Start with the ratio test. The ratio of successive terms is |(n+1) · (x+2)ⁿ⁺¹ / 4ⁿ⁺¹| divided by |n · (x+2)ⁿ / 4ⁿ|. Simplify to (n+1)/n · |x+2|/4. The first factor tends to 1 as n grows, leaving L = |x+2|/4. Convergence when L < 1, so |x+2| < 4. The radius is 4, and the open interval is (−6, −2). Now test the endpoints. At x = −6, the series becomes ∑ n · (−4)ⁿ / 4ⁿ = ∑ n · (−1)ⁿ. The terms do not approach zero, so the series diverges by the nth-term test. At x = −2, the series becomes ∑ n · 0ⁿ. The terms are zero for n ≥ 1, so the series converges (it is the single term n=0, which is 0 · 1 = 0). The interval of convergence is (−6, −2].

That single problem exercises every component: ratio test with a non-zero centre, two endpoints with opposite behaviour, and a boundary case where one endpoint collapses the entire series to a single term. Worked problems of this density, four or five per week for a month, will lock in the pattern more durably than twenty problems of the easier c = 0 variety. The LNAT parallel is the mixed-difficulty full-section drill, where the candidate sees argument questions, assumption questions, and flaw questions interleaved rather than in topic blocks. Both training plans trade single-topic volume for cross-topic transfer, and both produce larger score gains per hour than a single-topic grind.

Where this topic sits in the broader BC exam strategy

Radius and interval of convergence appear on Section II of the AP Calculus BC exam, typically as a six-point free-response problem. The scoring weight is large enough that a single careless endpoint can cost a full letter grade on the five-point AP scale. The topic also feeds into the Taylor and Maclaurin series unit that precedes it, because the radius of a Taylor series determines the domain on which the function and its polynomial approximation are interchangeable. A student who treats the topic as a standalone drill will earn the points on this question and lose the points on the Taylor follow-up; a student who treats it as the gateway to a larger idea about the size of the convergence domain will earn both.

LNAT preparation strategy mirrors this layering. The argument-analysis question type is the standalone drill that produces the immediate score gain; the essay is the larger structure that depends on the same skill but rewards a deeper use of it. Candidates who treat argument analysis as a means to a better essay, rather than as a self-contained topic, tend to score higher on both sections. The same principle — small drill first, then bigger structure that reuses the same skill — is the right way to schedule radius-of-convergence work in an AP Calculus BC plan.

Radius and interval of convergence is the AP Calculus BC topic where mechanical fluency meets boundary judgement, and the small habit of probing the endpoints is what separates a 6 from a 9 on the free-response. TestPrep Europe's targeted drill on power-series endpoints is the natural next step for BC candidates who can find the radius on demand but lose marks at the boundaries of the interval.

Frequently asked questions