In AP Calculus BC, representing functions as power series is the unit where a candidate who has handled Taylor polynomials all year is suddenly asked to do the opposite operation: instead of differentiating or integrating a known series to find coefficients, the student is handed a closed-form function such as 1/(1+x), arctan(x), or ln(1+x) and is expected to produce the series itself. The exam measures this skill through one or two free-response items, typically embedded in a 6-question FRQ section that also covers Taylor polynomial approximation, Lagrange error bounds, and convergence tests. The marks ride on three rubric rows in order: the coefficients row, the radius-of-convergence row, and the interval-of-convergence row, with integration-by-series scoring on top when the prompt asks for an antiderivative.
What "representing a function as a power series" actually means on the FRQ
A power series is an expression of the form a_0 + a_1(x-c) + a_2(x-c)^2 + a_3(x-c)^3 + … that converges to a function f(x) on some interval centred at c. On the AP Calculus BC exam, the phrase "represent f(x) as a power series" almost always points to one of three families of functions: geometric forms 1/(1-u), logarithmic forms such as ln(1+u) obtained by integrating the geometric form, and arctangent forms obtained by integrating the alternating geometric form. The exam rarely asks a candidate to invent a series from scratch; it gives a closed-form function and expects the student to recognise which known series it derives from.
The operative word is recognition. Most candidates who lose points on this FRQ do not fail because they cannot differentiate or integrate a series — they fail because they cannot see that 1/(2-x) is a geometric series in disguise. The standard move is to rewrite the function so the denominator matches the template 1 - u, where u is a small expression in x. For 1/(2-x), the move is to factor 2 out of the denominator, giving 1/2 · 1/(1 - x/2). The series is then the geometric template evaluated at u = x/2, namely 1/2 · (1 + x/2 + (x/2)^2 + (x/2)^3 + …) = 1/2 + x/4 + x^2/8 + x^3/16 + ….
The AP exam expects this manipulation in a single line. The coefficients row on the rubric typically reads something like: "Writes the series in the form Σ a_n (x - c)^n and identifies the general term." Candidates who leave the answer in unsimplified nested form, such as 1/2 · 1/(1 - x/2) expanded only partway, are docked the row because the rubric reads simplified coefficients. A useful habit: after you write the first three non-zero terms, jump straight to the general term a_n = (1/2)(x/2)^n for n = 0, 1, 2, …. That way the reader sees both the pattern and the formula, which is exactly what the row is looking for.
Three function families the exam cycles through
- Geometric: 1/(1-u) = Σ u^n. Variants include 1/(1+u), 1/(1-au), and the k/(b-x) form above.
- Logarithmic: ln(1+u) = Σ (-1)^{n+1} u^n / n. The series alternates, which is why convergence at the endpoints is a routine test question.
- Arctangent: arctan(u) = Σ (-1)^n u^{2n+1} / (2n+1). This is the one with only odd powers, and it is the classic integrate-the-geometric form.
The three rubric rows, in the order readers actually score them
Most AP Calculus BC power-series FRQs are scored in the same three-row order, and once a candidate knows the order, the time-budgeting on the 15-minute slot becomes automatic. The first row is the coefficients row, the second is the radius-of-convergence row, and the third is the interval row. A fourth row sometimes appears as a follow-up, asking for the sum of the series evaluated at a particular x-value or for an integral of the function represented as a series. That fourth row is where the exam tests whether a student can plug in correctly without re-deriving the whole series.
The coefficients row is the easiest to earn and the easiest to lose. The reader is checking two things: that the candidate has written a Σ with a clean general term, and that the radius of convergence is not hidden inside the algebra. A series written as Σ (x/2)^n is acceptable as a step, but a series written as Σ (1/2^n) x^n is the form the rubric actually credits, because the (1/2^n) is the coefficient a_n and the x^n makes the powers of x visible. The general term should be readable at a glance. Most readers will spend 8 to 12 seconds on this row; a clean answer is the difference between a 1 and a 0 in about that window.
The radius-of-convergence row follows immediately. For a geometric form Σ a · u^n where u is a linear expression in x, the radius is determined by |u| < 1. For 1/(1 - x/2), the condition is |x/2| < 1, giving R = 2. The rubric typically reads: "States the radius of convergence and shows the inequality that defines it." A common error is to write R = 2 without showing |x/2| < 1, which costs the row on stricter readers. The exam wants to see the candidate recognise that the radius comes from the convergence of the geometric template, not from a separate application of the ratio test. Writing the ratio test on a series that is obviously geometric is a time sink that costs 2 to 3 minutes for no extra credit.
The interval-of-convergence row is where the alternating-sign and boundary cases get tested. For 1/(1 - x/2), the interval is (-2, 2) open on both ends, because the geometric series diverges at x = 2 and x = -2. For arctan(x), the interval is [-1, 1] closed at both ends, because the alternating series converges at both endpoints. The exam loves asking candidates to test the endpoints, and the rubric almost always has a sub-row: "Tests x = R and x = -R for convergence, states the result of each test." Candidates who skip the endpoint tests or write "converges at both endpoints" without showing the substitution lose this row even when the rest is correct.
Manipulation tactics the exam rewards
Every power-series FRQ on AP Calculus BC is built on one of three manipulations: algebraic rewrite into geometric form, integration of a known series, and differentiation of a known series. The exam rarely asks a candidate to derive a series by long division or by taking derivatives and solving for a_n; that work is what Taylor polynomials are for, and the two topics are kept distinct in the scoring guidelines. Recognising which manipulation the question wants is the first 30 seconds of the item.
Algebraic rewrite is the most common. Given 1/(1+3x), the move is to write it as 1/(1 - (-3x)) and read off the geometric series with u = -3x. The series is 1 - 3x + 9x^2 - 27x^3 + …, and the radius of convergence comes from |-3x| < 1, which gives R = 1/3. Candidates who forget the minus sign in u and write 1 + 3x + 9x^2 + … are punished twice: the sign pattern is wrong, and the radius is over-stated. A useful self-check: when you have finished writing the first three terms, plug x = 1/3 into the original function and into your series; both should give the same numerical value. If they do not, the sign is wrong.
Integration is the second most common manipulation. The exam will hand a candidate a series such as Σ x^n and ask for the power series representation of -ln(1-x). The work is a single line: integrate term-by-term, and the constant of integration is C, which the rubric typically scores as the next row. For -ln(1-x), the answer is x + x^2/2 + x^3/3 + x^4/4 + …. The radius of convergence is preserved by term-by-term integration, so R = 1 from the original geometric form. The interval of convergence becomes (-1, 1) but the rubric usually wants the endpoint tests: at x = 1 the series is the harmonic series and diverges, at x = -1 the series is the alternating harmonic series and converges. Candidates who write the interval as (-1, 1] without showing the endpoint test lose the row.
Differentiation is rarer but appears at least once every few years. The standard problem gives a function written as a power series and asks for the derivative, again as a power series. The procedure is mechanical: differentiate each term, simplify the coefficients, and keep the Σ notation. The radius of convergence is preserved by differentiation on the interior of the interval. The exam rarely asks for the interval of convergence after a differentiation, so candidates should not waste time on it unless the prompt specifies.
Common pitfalls and how to avoid them
- Forgetting the radius of convergence comes from the geometric template. If the series is geometric in u = ax + b, then the radius is the set of x with |ax + b| < 1. A quick way: solve |ax + b| = 1 for x; the absolute value of the answer is the radius.
- Skipping endpoint tests. The alternating series converges at one or both endpoints more often than candidates expect. Test both endpoints and write the result next to each one.
- Dropping the constant of integration. When integrating a series to recover a function, write "+ C" exactly once. The rubric row for the constant is often worth a full point on its own.
- Writing the answer as a closed-form fraction rather than a series. The question asks for a power series representation. A candidate who writes 1/(1-x) without expanding is scoring zero on the coefficients row.
Worked example: 1/(2+3x) as a power series
Take the function f(x) = 1/(2+3x) and walk through the rubric. The first move is to factor the denominator to expose the geometric form. 2 + 3x = 2(1 + 3x/2), so f(x) = (1/2) · 1/(1 - (-3x/2)). The geometric template 1/(1-u) = Σ u^n then gives the series (1/2) Σ (-3x/2)^n, valid for |−3x/2| < 1. The coefficients row is satisfied with the general term a_n = (1/2)(-3/2)^n. The radius-of-convergence row is satisfied by writing |−3x/2| < 1, which solves to |x| < 2/3, so R = 2/3. The interval-of-convergence row requires the endpoint tests: at x = 2/3, the series becomes (1/2) Σ (-1)^n, which diverges because the partial sums oscillate. At x = -2/3, the series becomes (1/2) Σ 1, which diverges because the terms do not go to zero. The interval is therefore (-2/3, 2/3), open at both ends. Total scoring: coefficients 1, radius 1, interval 1, endpoints 1. The item is worth 4 points and the candidate has earned all of them in roughly 4 minutes of writing.
Now consider a slightly harder prompt: write the power series for arctan(2x) and determine its interval of convergence. The function arctan(2x) is the integrate-the-series form: differentiate it, get 2/(1+4x^2) = 2 · 1/(1 - (-4x^2)) = 2 Σ (-4x^2)^n = 2 Σ (-1)^n 4^n x^{2n}. Integrate term-by-term to recover arctan(2x) = C + 2 Σ (-1)^n 4^n x^{2n+1} / (2n+1). The constant of integration is 0 because arctan(0) = 0, which the rubric usually scores as a separate row. The radius of convergence is the same as the geometric form, |−4x^2| < 1, giving x^2 < 1/4 and R = 1/2. The interval is (-1/2, 1/2) before testing endpoints; at x = 1/2 the series is 2 Σ (-1)^n 4^n / (2n+1) · (1/2)^{2n+1} = Σ (-1)^n / (2n+1), which is the alternating arctangent series and converges. At x = -1/2 the series is the negative of the same alternating series, which also converges. The interval is therefore [-1/2, 1/2].
Comparison: power series versus Taylor series on the FRQ
The exam draws a clear line between the two, and the rubric reflects the distinction. A Taylor series question asks the candidate to compute coefficients a_n = f^{(n)}(c)/n! and write a Σ. A power-series-representation question hands the candidate the closed-form function and asks for a series. The two skills are different: Taylor requires derivative calculation, power series requires pattern recognition. Mixing them up is one of the more common ways to lose time on the FRQ section.
| Feature | Power series representation | Taylor series |
|---|---|---|
| Starting point | Closed-form function f(x) | Function and its derivatives at a centre c |
| Main work | Algebraic rewrite, integrate, or differentiate a known series | Compute derivatives, divide by n! |
| General term | Often a simple exponential pattern like (1/2)^n | f^{(n)}(c)/n! times (x-c)^n |
| Rubric emphasis | Coefficients, radius, interval, endpoint tests | Coefficients, Lagrange error bound, interval of convergence |
| FRQ slot | Usually one full FRQ or one part of a multi-part item | Often combined with polynomial approximation and error bound |
The takeaway: when the prompt contains the phrase "as a power series centered at 0," the candidate should immediately think of the three templates — geometric, logarithmic, arctangent — and not reach for derivative calculations. The exam rarely combines the two operations in the same part of an FRQ; it gives one method per prompt and asks the candidate to recognise which.
How to study this unit with a scoring-conscious plan
Most candidates reading this have already worked through a unit on Taylor polynomials and are now expected to invert the skill. The preparation plan that works best, in my experience, is to drill the three template families until the recognition is automatic, and only then move to the radius and interval work. Spending a single 50-minute session on rewriting 1/(1-ax), 1/(1+ax), and k/(b-x) into series form cements the algebraic move. A second 50-minute session on integrating Σ x^n to recover -ln(1-x) and on integrating Σ (-1)^n x^{2n} to recover arctan(x) cements the calculus move. Only after those two sessions should the candidate move to radius and interval problems, because the mechanics of |u| < 1 and the endpoint tests make sense once the series itself is easy to write.
Past free-response items are the best practice material, because the rubric language is exactly what the reader will use in May. The 2016 BC exam FRQ 6, for instance, is a clean power-series problem: it asks for the series of x/(1+x^2) and the series of arctan(x) as an integrated form, then asks for the radius of convergence of each. Working this problem under timed conditions — 15 minutes, no notes — trains the candidate to allocate 4 minutes on the rewrite, 4 minutes on the integration, 4 minutes on the radii, and 3 minutes on the interval and endpoint tests. A second pass with the rubric in hand is what teaches the candidate which phrases the reader is scanning for: "the general term is," "the series converges when," "at x = R, the series becomes."
For candidates targeting a 5, the extra work is to practice problems that combine power series with Taylor polynomial error bounds, because the two topics are tested in the same free-response section and the time-budgeting across them is a real constraint. A 15-minute slot for a power-series item is generous if the candidate is fluent; the same 15 minutes is impossible if the candidate is hesitating on the algebraic rewrite. The preparation strategy that produces a 5 is fluency on the rewrite, not deeper theory.
What the AP reader actually marks, line by line
A close reading of released scoring guidelines shows that AP readers credit specific phrases and dock others. The coefficients row typically requires the candidate to write a Σ with an explicit general term, not just the first three non-zero terms. "I see the pattern" is not a credited phrase; "a_n = (1/2)(-3/2)^n for n ≥ 0" is. The radius row typically requires an inequality of the form |u| < 1 written out, not just the value of R. "R = 2/3" by itself is not always credited; "|x/2| < 1, so R = 2" is. The interval row typically requires the endpoint tests written next to the endpoints, not appended as a footnote. "Converges at x = 1" is weaker than "at x = 1, the series is Σ 1/n, which diverges by the integral test."
The constant of integration row, when it appears, is a single line: "+ C." Readers credit the row whether the candidate wrote "+ C" or wrote the constant explicitly, but they do not credit a missing constant when the function requires one. The exam will sometimes ask for the series of ln(1-x) starting from the geometric form for 1/(1-x); the constant is 0 because ln(1-0) = 0, and the candidate who writes "+ C" without evaluating C loses the row. A safe habit: after writing "+ C," substitute x = 0 on both sides and solve for C, even if the answer is obviously 0.
Time is the second-order constraint. The FRQ section is 6 items in 90 minutes, which works out to 15 minutes per item on average. A power-series item that runs longer than 15 minutes is taking time from another item, and the candidate is usually better off writing a clean partial answer than grinding for a 4-point item while leaving a 4-point item untouched. The fastest-scoring approach is to write the coefficients in the first 4 minutes, the radius in the next 2, the interval and endpoints in the next 4, and the optional follow-up (sum evaluation, integral, second function) in the last 5. Candidates who reverse the order — interval before radius — frequently find that they have to redo the radius after discovering the endpoints matter, and the time cost is severe.
Putting it all together: a 15-minute exam plan
The plan for a power-series FRQ slot is short enough to memorise. Read the prompt and identify the function family in the first 60 seconds. If the function is 1/(1±u) or a constant multiple, plan an algebraic rewrite. If the function is logarithmic or arctangent in form, plan to integrate a known geometric series. If the function is already a series and the prompt asks for the derivative, plan to differentiate term by term. Once the family is identified, the rest of the work is mechanical.
Write the coefficients row in the next 3 to 4 minutes. The general term should be on the page within the first 5 minutes of the slot. Compute the radius in the next 1 to 2 minutes by writing |u| < 1 and solving for the absolute value of x. Test the endpoints and write the interval in the next 3 to 4 minutes. Reserve the last 3 to 4 minutes for any follow-up part: evaluating the series at a specific x, integrating the series, or finding a second function. The plan is rigid on purpose; the rubric is also rigid, and the candidate who matches the rigidity of the rubric wins the time.
For students reading this who are still a few months out from the exam, the single most productive habit is to do one power-series rewrite every other day for the next 8 weeks. The repetition makes the recognition automatic, and the rest of the rubric rows fall into place once the series itself is easy to write. For students who are within 4 weeks of the exam, the most productive habit is timed past-FRQ drills: 15 minutes per item, no notes, then grade against the rubric. The 15-minute clock is the real constraint on exam day, and the candidate who has trained under that clock performs noticeably more cleanly than the candidate who has only done untimed practice.
AP Calculus BC's power-series unit is short in the curriculum but heavy in the scoring, and the FRQ items built on it are predictable enough that a focused preparation plan pays off. The three rubric rows, the three template families, the four endpoint cases, and the 15-minute slot are the entire skill set; everything else is mechanics. The candidate who walks into the exam with the plan above and 30 to 40 power-series problems under timed conditions has converted a 3-or-4 score into a 5 in roughly the time it takes to drill the algebra of the geometric template.
AP Courses' AP Calculus BC preparation programme pairs each student with a reader-trained tutor who grades every power-series FRQ against the released scoring guidelines and rebuilds the radius-of-convergence and endpoint-test rows until the 15-minute slot runs cleanly to a 4-out-of-4.