MATH 141 Calculus 2 • McGill University, Montreal

Revision sheet: power series (MATH 141)

This sheet is not a summary of sections 11.8 and 11.9 of Stewart: you already have the course notes. It answers one question only, what makes students lose marks on power series in MATH 141 at McGill University, and which precise gesture avoids each loss.

Every number below is done by hand, as on the exam, and every value quoted has been checked by an independent computation. The rule that runs through the whole sheet: the radius comes from one test, each endpoint from its own.

The thread of the chapter

The ratio test finds the RADIUS, never the interval: each endpoint is a separate numerical series with its own test, every operation on a series (substitute, recentre, multiply by a power of x, differentiate, integrate) keeps track of R but not of the endpoints, and outside the interval the series is worth nothing, even where the function exists.

This chapter is part of MATH 141, Calculus 2 (McGill)

The essentials

Radius from one test, each endpoint from its own

  • • A power series ∑cn(x−a)n\sum c_n(x - a)^n converges absolutely for ∣x−a∣<R|x - a| < R and diverges for ∣x−a∣>R|x - a| > R. It always converges at its centre aa; RR can be 00 or ∞\infty.
  • • RR: ratio test on ∣an∣|a_n| with xx fixed, ∣an+1an∣→∣x−a∣R\left|\frac{a_{n+1}}{a_n}\right| \to \frac{|x - a|}{R}. When every factor is a power nn, the root test is faster.
  • • At ∣x−a∣=R|x - a| = R the ratio limit is exactly 11: the test is silent BY CONSTRUCTION. Substitute x=a−Rx = a - R and x=a+Rx = a + R, write the two numerical series, test each one.
  • • Four possible answers: (a−R,a+R)(a - R, a + R), [a−R,a+R)[a - R, a + R), (a−R,a+R](a - R, a + R], [a−R,a+R][a - R, a + R]. Only the endpoint tests choose between them.
  • • Read the centre after factoring: (2x−3)n=2n(x−32)n(2x - 3)^n = 2^n\left(x - \frac{3}{2}\right)^n, centre 32\frac{3}{2}, and ∣2x−3∣<1|2x - 3| < 1 gives R=12R = \frac{1}{2}.
−13centre 1ratio test: |x − 1| < 2AST: convergesΣ 1/√n: diverges
For ∑(x−1)nn 2n\sum \frac{(x - 1)^n}{\sqrt{n}\,2^n} the ratio test gives only (−1,3)(-1, 3); the alternating series test closes the bracket at −1-1, the pp-series ∑1n\sum \frac{1}{\sqrt{n}} leaves it open at 33.

The midpoint of the interval must be the centre, and its half-length the radius: a ten-second check that catches every misread centre.

Everything comes from 1/(1 - u)

  • • 11−u=∑n=0∞un\frac{1}{1 - u} = \sum_{n=0}^\infty u^n for ∣u∣<1|u| < 1, and never at ∣u∣=1|u| = 1. Put the function in the form constant1−u\frac{\text{constant}}{1 - u} and read uu: the interval is ∣u∣<1|u| < 1.
  • • Create the 11 by factoring: 12+x=12⋅11+x/2=∑(−1)nxn2n+1\frac{1}{2 + x} = \frac{1}{2} \cdot \frac{1}{1 + x/2} = \sum \frac{(-1)^n x^n}{2^{n+1}}, ∣x∣<2|x| < 2.
  • • Recentre at aa by rewriting around aa: 1x+5=17+(x−2)=∑(−1)n(x−2)n7n+1\frac{1}{x + 5} = \frac{1}{7 + (x - 2)} = \sum \frac{(-1)^n (x - 2)^n}{7^{n+1}}, ∣x−2∣<7|x - 2| < 7.
  • • Differentiate or integrate term by term inside (a−R,a+R)(a - R, a + R): same RR, endpoints re-tested, and a constant CC found at the centre.
  • • ln⁡(1+x)=∑n≥1(−1)n+1xnn\ln(1 + x) = \sum_{n\ge1} (-1)^{n+1}\frac{x^n}{n} on (−1,1](-1, 1], arctan⁡x=∑n≥0(−1)nx2n+12n+1\arctan x = \sum_{n\ge0} (-1)^n\frac{x^{2n+1}}{2n + 1} on [−1,1][-1, 1], 1(1−x)2=∑n≥0(n+1)xn\frac{1}{(1 - x)^2} = \sum_{n\ge0} (n + 1)x^n on (−1,1)(-1, 1).

The rules in table form

Each row reads from left to right: the assumptions, then the result. A red cell is not an answer, it is the finding that the form settles nothing and the instruction to rewrite it. Every case is followed by a worked example.

What each operation does to the radius and to the endpoints

Start from a series of radius RR. The left column is the operation, the middle one what happens to RR, the right one what the endpoints need afterwards. The red line is the only one that produces no value at all.

OperationRadiusEndpoints after it
replace xx by cxcx R∣c∣\frac{R}{|c|} re-test

Example: 11+3x=∑(−3)nxn\frac{1}{1 + 3x} = \sum (-3)^n x^n: R=13R = \frac{1}{3}, interval (−13,13)\left(-\frac{1}{3}, \frac{1}{3}\right).

replace xx by x2x^2 R\sqrt{R} re-test

Example: ∑un2n\sum \frac{u^n}{2^n} has R=2R = 2, so ∑x2n2n\sum \frac{x^{2n}}{2^n} has R=2R = \sqrt{2}.

multiply by xkx^k unchanged unchanged

Example: x24−x=∑xn+24n+1\frac{x^2}{4 - x} = \sum \frac{x^{n+2}}{4^{n+1}} on (−4,4)(-4, 4), like 14−x\frac{1}{4 - x}.

differentiate unchanged may LOSE an endpoint

Example: ∑xnn2\sum \frac{x^n}{n^2} on [−1,1][-1, 1] gives ∑xn−1n\sum \frac{x^{n-1}}{n} on [−1,1)[-1, 1).

integrate unchanged may GAIN an endpoint

Example: 11+x2\frac{1}{1 + x^2} on (−1,1)(-1, 1) gives arctan⁡x\arctan x on [−1,1][-1, 1].

evaluate at ∣x−a∣>R|x - a| > R not applicable the series diverges no value

Example: ln⁡(1+x)\ln(1 + x) at x=2x = 2: the terms 2nn\frac{2^n}{n} tend to ∞\infty, so the series is not ln⁡3\ln 3.

Same form, other result: 11−x\frac{1}{1 - x} at x=−12x = -\frac{1}{2}: ∑(−12)n=23\sum \left(-\frac{1}{2}\right)^n = \frac{2}{3}, correct; at x=−2x = -2: ∑(−2)n\sum (-2)^n diverges while 11−(−2)=13\frac{1}{1 - (-2)} = \frac{1}{3}.

What to do: Find another series, or another point, inside its interval: ln⁡3=−ln⁡(1−23)=∑(2/3)nn\ln 3 = -\ln\left(1 - \frac{2}{3}\right) = \sum \frac{(2/3)^n}{n}.

The radius column is a free result from the theorem or from ∣u∣<1|u| < 1; the endpoint column is always work to do on the copy.

The mistakes that cost marks

These are the errors I correct most often in session. Each one costs marks on a paper, even when the reasoning behind it is right.

1. Giving the open interval as the answer

2 marks, the two endpoint marks of the question

What not to write

“∣an+1an∣→∣x∣3<1\left|\frac{a_{n+1}}{a_n}\right| \to \frac{|x|}{3} < 1, so the interval of convergence of ∑xnn 3n\sum \frac{x^n}{n\,3^n} is (−3,3)(-3, 3).”

What to write

“R=3R = 3. At x=3x = 3: ∑1n\sum \frac{1}{n} diverges. At x=−3x = -3: ∑(−1)nn\sum \frac{(-1)^n}{n} converges by the alternating series test. Interval [−3,3)[-3, 3).”

Why: The ratio test decides only strictly inside and strictly outside. The endpoints are two more series, and the answer is not complete until each has a verdict with its test named.

2. Testing an endpoint with the ratio test

1 mark per endpoint

What not to write

“At x=3x = 3, lim⁡∣an+1an∣=1\lim\left|\frac{a_{n+1}}{a_n}\right| = 1, so the series diverges there.”

What to write

“At x=3x = 3 the ratio limit is 11, which decides nothing. The series is ∑1n\sum \frac{1}{n}, a pp-series with p=1p = 1: it diverges.”

Why: L=1L = 1 is the silent case of the ratio test, and at an endpoint it is 11 by construction. The conclusion must come from a test that CAN decide: divergence test, pp-series, comparison, alternating series test.

3. Misreading the centre or the radius

the whole question, since every endpoint is then wrong

What not to write

“∑(2x−3)nn\sum \frac{(2x - 3)^n}{n} is centred at 33 with radius 11.”

What to write

“(2x−3)n=2n(x−32)n(2x - 3)^n = 2^n\left(x - \frac{3}{2}\right)^n: centre 32\frac{3}{2}, and ∣2x−3∣<1  ⟺  ∣x−32∣<12|2x - 3| < 1 \iff \left|x - \frac{3}{2}\right| < \frac{1}{2}, so R=12R = \frac{1}{2} and the interval is [1,2)[1, 2).”

Why: The centre is the point where the base of the power vanishes, and the radius is measured in xx, not in 2x2x. The check: the midpoint of [1,2)[1, 2) is 32\frac{3}{2} and its half-length 12\frac{1}{2}.

4. Evaluating a series outside its interval

the whole question

What not to write

“11−x=∑xn\frac{1}{1 - x} = \sum x^n, so at x=−1.5x = -1.5 the series gives 12.5=0.4\frac{1}{2.5} = 0.4.”

What to write

“The identity holds only for ∣x∣<1|x| < 1. At x=−1.5x = -1.5 the terms (−1.5)n(-1.5)^n grow without bound: the series diverges, although 11−x=0.4\frac{1}{1 - x} = 0.4 exists.”

-1.8-1.6-1.4-1.2-1-0.8-0.6-0.4-0.20.20.40.60.8-3-2-112341/(1 − x)S₄S₅
Left of the dashed line x=−1x = -1, S4S_4 and S5S_5 leave 11−x\frac{1}{1 - x} in opposite directions, although the function is perfectly smooth there.

Why: A power series knows its function only inside its interval. Beyond it the partial sums swing further and further from the function, and no value can be read off them.

5. Forgetting the constant after integrating

1 to 2 marks

What not to write

“ddxln⁡(3+x)=13+x=∑(−1)nxn3n+1\frac{d}{dx}\ln(3 + x) = \frac{1}{3 + x} = \sum \frac{(-1)^n x^n}{3^{n+1}}, so ln⁡(3+x)=∑(−1)nxn+1(n+1)3n+1\ln(3 + x) = \sum \frac{(-1)^n x^{n+1}}{(n + 1)3^{n+1}}.”

What to write

“ln⁡(3+x)=C+∑(−1)nxn+1(n+1)3n+1\ln(3 + x) = C + \sum \frac{(-1)^n x^{n+1}}{(n + 1)3^{n+1}}, and at x=0x = 0: ln⁡3=C\ln 3 = C. So ln⁡(3+x)=ln⁡3+∑n≥0(−1)nxn+1(n+1)3n+1\ln(3 + x) = \ln 3 + \sum_{n\ge0} \frac{(-1)^n x^{n+1}}{(n + 1)3^{n+1}} on (−3,3](-3, 3].”

Why: Term-by-term integration gives an antiderivative, so a constant. At the centre the series is worth 00 and the function ln⁡3\ln 3: the check at x=0x = 0 finds CC in one line.

6. Differentiating a number instead of a function

the whole question

What not to write

“∑n≥0(13)n=32\sum_{n\ge0} \left(\frac{1}{3}\right)^n = \frac{3}{2}, and the derivative of 32\frac{3}{2} is 00, so ∑n≥1n(13)n−1=0\sum_{n\ge1} n\left(\frac{1}{3}\right)^{n-1} = 0.”

What to write

“On (−1,1)(-1, 1), ∑n≥1nxn−1=1(1−x)2\sum_{n\ge1} n x^{n-1} = \frac{1}{(1 - x)^2}; at x=13x = \frac{1}{3} this gives 94\frac{9}{4}, so ∑n3n=13⋅94=34\sum \frac{n}{3^n} = \frac{1}{3} \cdot \frac{9}{4} = \frac{3}{4}.”

Why: Differentiation acts on functions of xx. Keep xx a variable until the identity between functions is established, and substitute the number on the last line only.

7. Assuming a derived series keeps its endpoints

1 mark per endpoint

What not to write

“∑xnn\sum \frac{x^n}{n} converges on [−1,1)[-1, 1), so its derivative ∑xn−1\sum x^{n-1} converges on [−1,1)[-1, 1) too, by the theorem.”

What to write

“The theorem gives the same radius 11. At x=−1x = -1, ∑(−1)n−1\sum (-1)^{n-1} has terms that do not tend to 00: it diverges. The derivative converges on (−1,1)(-1, 1).”

−11Σ xⁿ/n[−1, 1)−11derivative(−1, 1)−11antiderivative[−1, 1]
Same radius 11 on all three lines, three different brackets: the derivative loses x=−1x = -1, the antiderivative gains x=1x = 1.

Why: Differentiating multiplies the nnth term by nn, which can only make things worse at an endpoint; integrating divides by n+1n + 1, which can only help. The theorem speaks about RR, never about the brackets.

8. A coefficient formula on a series with missing powers

the whole question

What not to write

“∑x2n4n\sum \frac{x^{2n}}{4^n} has cn=14nc_n = \frac{1}{4^n}, so R=lim⁡∣cncn+1∣=4R = \lim\left|\frac{c_n}{c_{n+1}}\right| = 4.”

What to write

“Ratio test on the terms: ∣an+1an∣=x24<1  ⟺  ∣x∣<2\left|\frac{a_{n+1}}{a_n}\right| = \frac{x^2}{4} < 1 \iff |x| < 2. So R=2R = 2, and at x=±2x = \pm 2 every term is 11: interval (−2,2)(-2, 2).”

Why: 14n\frac{1}{4^n} is the coefficient of x2nx^{2n}, not of xnx^n: the odd coefficients are 00 and the shortcut compares the wrong numbers. Always apply the test to the terms as written.

Which method to choose

Which test at an endpoint, by the FORM of the series

The ratio test has given ∣x−a∣<R|x - a| < R. Substitute x=a−Rx = a - R and x=a+Rx = a + R, write each numerical series in full, and read its form

  • If the terms do not tend to 0 (they are all plus or minus 1, or they grow like n) → divergence test: the endpoint is excluded in one line

    Example: ∑nxn4n\sum \frac{n x^n}{4^n} at x=4x = 4: the terms are nn, diverges

  • If the absolute values behave like 1np\frac{1}{n^p} with p>1p > 1 → absolute convergence by p-series or comparison, which settles BOTH endpoints at once

    Example: ∑xnn2 5n\sum \frac{x^n}{n^2\,5^n} at x=±5x = \pm 5: 1n2\frac{1}{n^2}, both included

  • If constant sign, absolute values like 1np\frac{1}{n^p} with p≤1p \le 1 → p-series or limit comparison: diverges

    Example: ∑(−1)nxnn+1\sum \frac{(-1)^n x^n}{\sqrt{n + 1}} at x=−1x = -1: ∑1n+1\sum \frac{1}{\sqrt{n + 1}} diverges

  • If alternating sign, absolute values decreasing to 00 but like 1np\frac{1}{n^p} with p≤1p \le 1 → alternating series test: converges, conditionally

    Example: ∑xnn 3n\sum \frac{x^n}{n\,3^n} at x=−3x = -3: ∑(−1)nn\sum \frac{(-1)^n}{n} converges

  • If the series came from 11−u\frac{1}{1 - u} with nothing else done to it → ∣u∣=1|u| = 1 at both endpoints: both excluded, no further test

    Example: 11+3x\frac{1}{1 + 3x} at x=±13x = \pm\frac{1}{3}: terms ±1\pm 1

  • If the term is a power nn of a base that tends to 11 → form 1∞1^\infty: compute the limit of the terms with a logarithm before concluding

    Example: (1−52n+6)n→e−5/2≠0\left(1 - \frac{5}{2n + 6}\right)^n \to e^{-5/2} \ne 0: diverges

Never the ratio or root test at an endpoint: there its limit is 11 by construction, the only value on which it is silent.

How the answer is expected to be written

A marker ticks steps. Here they are in order, with the concluding sentence expected word for word.

Writing an interval of convergence

When to use it: Any question that says find the radius and the interval of convergence

  1. 1 Name the term ana_n, with xx fixed, and apply the ratio test (or the root test) to ∣an∣|a_n|, with the limit written.
  2. 2 Solve limit <1< 1 in the form ∣x−a∣<R|x - a| < R: state the centre and the radius separately.
  3. 3 Substitute each endpoint, write the numerical series in full, name its test and verify the hypotheses of that test.
  4. 4 Conclude with the interval, each bracket justified by the verdict of its endpoint.

Concluding sentence

“By the ratio test the series converges absolutely for ∣x∣<3|x| < 3 and diverges for ∣x∣>3|x| > 3, so R=3R = 3. At x=3x = 3 it is ∑1n\sum \frac{1}{n}, which diverges; at x=−3x = -3 it is ∑(−1)nn\sum \frac{(-1)^n}{n}, which converges by the alternating series test. The interval of convergence is [−3,3)[-3, 3).”

The trap: Writing the two endpoint verdicts without the series they come from: the marker cannot tell which test applies to which series, and gives nothing for a verdict alone.

Marking: Typically 2 marks for the radius with the ratio test written, 1 to 2 marks per endpoint (series written, test named, hypotheses checked), and the interval with the right brackets.

Check before you hand in

Five minutes of checking recover more marks than one more problem started in a hurry.

The typical problem, taken apart

Interval, closed form and a numerical sum

Let g(x)=∑n=1∞(x−1)nn 2ng(x) = \sum_{n=1}^\infty \frac{(x - 1)^n}{n\,2^n}. Find the interval of convergence, a closed form for gg, and deduce the sum of ∑n=1∞(−1)nn 2n\sum_{n=1}^\infty \frac{(-1)^n}{n\,2^n}.

No calculator. Every step must be justified as on a MATH 141 final.

Step 1

Ratio test, xx fixed: ∣an+1an∣=∣x−1∣2⋅nn+1→∣x−1∣2\left|\frac{a_{n+1}}{a_n}\right| = \frac{|x - 1|}{2} \cdot \frac{n}{n + 1} \to \frac{|x - 1|}{2}, which is <1< 1 iff ∣x−1∣<2|x - 1| < 2. Centre 11, R=2R = 2, open interval (−1,3)(-1, 3).

Why

The limit written in full is the first mark; the centre and the radius are stated separately because the next two steps use them.

Step 2

At x=3x = 3: ∑2nn 2n=∑1n\sum \frac{2^n}{n\,2^n} = \sum \frac{1}{n}, diverges (p=1p = 1). At x=−1x = -1: ∑(−2)nn 2n=∑(−1)nn\sum \frac{(-2)^n}{n\,2^n} = \sum \frac{(-1)^n}{n}, alternating with 1n\frac{1}{n} decreasing to 00: converges. Interval [−1,3)[-1, 3).

Why

Each endpoint is its own numerical series with its own test; the ratio test gives 11 at both and could decide neither.

Step 3

For ∣x−1∣<2|x - 1| < 2, differentiate term by term: g′(x)=∑n≥1(x−1)n−12n=12⋅11−x−12=13−xg'(x) = \sum_{n\ge1} \frac{(x - 1)^{n-1}}{2^n} = \frac{1}{2} \cdot \frac{1}{1 - \frac{x - 1}{2}} = \frac{1}{3 - x}.

Why

Differentiating removes the nn of the denominator and leaves a geometric series with u=x−12u = \frac{x - 1}{2}, whose sum is known.

Step 4

Integrate: g(x)=−ln⁡(3−x)+Cg(x) = -\ln(3 - x) + C. At the centre, g(1)=0g(1) = 0 (every term contains (x−1)n(x - 1)^n with n≥1n \ge 1), so 0=−ln⁡2+C0 = -\ln 2 + C, C=ln⁡2C = \ln 2 and g(x)=ln⁡23−xg(x) = \ln\frac{2}{3 - x}.

Why

The constant is fixed at the centre, where the series is worth its constant term; without it, g(1)g(1) would be −ln⁡2-\ln 2, not 00.

Step 5

x=0x = 0 lies in [−1,3)[-1, 3) and gives (x−1)n=(−1)n(x - 1)^n = (-1)^n: ∑(−1)nn 2n=g(0)=ln⁡23\sum \frac{(-1)^n}{n\,2^n} = g(0) = \ln\frac{2}{3}. Check: −12+18−124+164≈−0.40-\frac{1}{2} + \frac{1}{8} - \frac{1}{24} + \frac{1}{64} \approx -0.40, and ln⁡2−ln⁡3≈0.69−1.10=−0.41\ln 2 - \ln 3 \approx 0.69 - 1.10 = -0.41.

Why

A point is substituted only once it is checked to be inside the interval; the first-terms test then confirms the value.

The conclusion, written out

“The series converges exactly on [−1,3)[-1, 3), where g(x)=ln⁡23−xg(x) = \ln\frac{2}{3 - x}; hence ∑n=1∞(−1)nn 2n=ln⁡23\sum_{n=1}^\infty \frac{(-1)^n}{n\,2^n} = \ln\frac{2}{3}.”

The classic mistake on this problem: Stopping at (−1,3)(-1, 3), or writing g(x)=−ln⁡(3−x)g(x) = -\ln(3 - x) without the constant, which gives ∑(−1)nn 2n=−ln⁡3\sum \frac{(-1)^n}{n\,2^n} = -\ln 3: negative, plausible, and wrong, since the first-terms test gives about −0.40-0.40, not −1.10-1.10.

Learn by heart

  • • The ratio test gives RR, never the interval: at ∣x−a∣=R|x - a| = R its limit is 11 and it is silent.
  • • Each endpoint: write the numerical series, name its test, check its hypotheses.
  • • 11−u=∑un\frac{1}{1 - u} = \sum u^n for ∣u∣<1|u| < 1 only; never at ∣u∣=1|u| = 1.
  • • Factor to create the 11: 12+x=12⋅11+x/2\frac{1}{2 + x} = \frac{1}{2} \cdot \frac{1}{1 + x/2}, and ∣x∣<2|x| < 2.
  • • Substituting x2x^2 takes the square root of RR; substituting cxcx divides it by ∣c∣|c|.
  • • Differentiating and integrating keep RR; differentiating may lose an endpoint, integrating may gain one.
  • • After integrating, find CC at the centre. After differentiating, evaluate on the last line only.
  • • Outside its interval a series is worth nothing, even where the function exists.

Frequently asked questions

How do I find the interval of convergence of a power series?

Apply the ratio test to the absolute value of the general term, with x fixed, and solve limit less than one: this gives the centre and the radius. Then substitute each endpoint into the series and test the two numerical series separately, with the divergence test, a p-series, a comparison or the alternating series test. The interval is written with each bracket decided by its endpoint.

Why can't the ratio test decide the endpoints of a power series?

Because the radius is defined as the place where the ratio limit equals one, and a limit equal to one is exactly the case where the ratio test gives no conclusion. The same holds for the root test. At an endpoint the series becomes an ordinary numerical series, and only the tests that can decide it, such as p-series or the alternating series test, may be used.

Does differentiating a power series change its interval of convergence?

It keeps the radius but not necessarily the endpoints. Differentiating multiplies each term by n, so a series that converged at an endpoint can diverge there after differentiation. Integrating divides by n plus one and can only gain an endpoint. The series of x to the n over n converges at minus one, its derivative does not.

How do I write 1 over (a + x) as a power series?

Factor a out of the denominator to create the form one over one minus u: one over (a plus x) equals one over a, times one over (1 plus x over a). Then u is minus x over a, the series is the geometric series in u multiplied by one over a, and it converges when the absolute value of x is less than the absolute value of a. Check at x equal to zero: the series must give one over a.

Practise it

Corrected exercises: Power series, MATH 141 at McGill

A method is proved on a paper, not on a sheet. The set for the same chapter takes each of these traps into a problem, with the solution written out step by step.

  • 10 corrected exercises
  • 100 points
  • 150 minutes
Do the exercises
Previous sheet Ratio and root tests Next sheet Taylor and Maclaurin series

See also

Looking for a MATH 141 tutor in Montreal?

Get in touch for a first session. Power series are where every test of the series chapters is used at once, and where the endpoint marks are won or lost.

Site by Studio Squalli