Calculus3Sequences and Series6 of 6
Sequences and Series · 6

Ratio and Root Tests

1

Comparing each term with the one before it

Let's find out whether the series of 3nn!\tm{\frac{3^n}{n!}} converges or diverges. The tests we have don't help much here. We can't use the integral test because n!n! is defined only for whole numbers, so there's no function f(x)f(x) to integrate. And for direct comparison, it isn't obvious which convergent series we'd compare 3nn!\tm{\frac{3^n}{n!}} with.

n!=n(n−1)⋯2⋅1,\displaystyle {n! = n(n-1)\cdots 2\cdot 1,}
4!=24,\displaystyle {4! = 24,}
5!=5⋅4!\displaystyle {5! = 5\cdot 4!}

Instead, let's see how each term compares with the one before it. In a geometric series, every term is r\geo{r} times the previous term, and the series converges whenever −1<r<1-1 < \geo{r} < 1.

an+1=r an,\displaystyle {\nx{a_{n+1}} = \geo{r}\,\tm{a_n},}
a+ar+ar2+⋯=a1−r\displaystyle {\geo{a + ar + ar^2 + \cdots} = \geo{\frac{a}{1-r}}}
(−1<r<1)\displaystyle {(-1 < \geo{r} < 1)}
Determine whether ∑n=1∞3nn!\sum_{n=1}^{\infty}\tm{\frac{3^n}{n!}} converges or diverges
1\htmlClass{lc-mute}{1}
2\htmlClass{lc-mute}{2}
3\htmlClass{lc-mute}{3}
4\htmlClass{lc-mute}{4}
5\htmlClass{lc-mute}{5}
6\htmlClass{lc-mute}{6}
7\htmlClass{lc-mute}{7}
1\htmlClass{lc-mute}{1}
2\htmlClass{lc-mute}{2}
3\htmlClass{lc-mute}{3}
4\htmlClass{lc-mute}{4}
nn
an\htmlClass{lc-pink}{a_n}
Text solution
Find the multiplierThe numerator gains a factor of 3\mul{3}, and (n+1)!=(n+1)⋅n!\nx{(n+1)!} = \mul{(n+1)}\cdot\tm{n!}, so each term is the one before times 3n+1\mul{\frac{3}{n+1}}. We call an+1an\frac{\nx{a_{n+1}}}{\tm{a_n}} the multiplier.
an+1\displaystyle \nx{a_{n+1}} =an⋅3n+1\displaystyle = \tm{a_n}\cdot\mul{\frac{3}{n+1}}
3⋅32\displaystyle \tm{3}\cdot\mul{\frac32} =4.5\displaystyle = \tm{4.5}
4.5⋅33\displaystyle \tm{4.5}\cdot\mul{\frac33} =4.5\displaystyle = \tm{4.5}
4.5⋅34\displaystyle \tm{4.5}\cdot\mul{\frac34} =3.375\displaystyle = \tm{3.375}
Cap the multiplier at 34\geo{\frac34}Unlike a geometric series, the multiplier changes at every step, and it approaches 0\mul{0}. Once n≥3n \ge 3, it's at most 34\geo{\frac34}, so from a3=4.5\tm{a_3 = 4.5} on, each term is at most 4.5(34)n−3\geo{4.5\left(\frac34\right)^{n-3}}, the matching term of a geometric series with r=34\geo{r = \frac34}. The picture below adds those up.

Here is the geometric series laid over the terms from the third term on. Its sum is 18\geo{18}, so the whole series is at most 25.525.5. All the terms are positive, so by direct comparison, the series converges.

1\htmlClass{lc-mute}{1}
2\htmlClass{lc-mute}{2}
3\htmlClass{lc-mute}{3}
4\htmlClass{lc-mute}{4}
5\htmlClass{lc-mute}{5}
6\htmlClass{lc-mute}{6}
7\htmlClass{lc-mute}{7}
8\htmlClass{lc-mute}{8}
1\htmlClass{lc-mute}{1}
2\htmlClass{lc-mute}{2}
3\htmlClass{lc-mute}{3}
4\htmlClass{lc-mute}{4}
nn
an\htmlClass{lc-pink}{a_n}
×32\htmlClass{lc-green}{\times\dfrac32}
×321\htmlClass{lc-green}{\times\vphantom{\dfrac32}1}
×34\htmlClass{lc-green}{\times\dfrac34}
×35\htmlClass{lc-green}{\times\dfrac35}
×12\htmlClass{lc-green}{\times\dfrac12}
×37\htmlClass{lc-green}{\times\dfrac37}
×38\htmlClass{lc-green}{\times\dfrac38}
4.5(34)n−3\htmlClass{lc-violet}{4.5\left(\dfrac34\right)^{n-3}}
n\displaystyle n ≥3:an+1\displaystyle \ge 3:\quad \nx{a_{n+1}} =an⋅3n+1\displaystyle = \tm{a_n}\cdot\mul{\frac{3}{n+1}} ≤34 an\displaystyle \le \geo{\frac34}\,\tm{a_n}
4.5+4.5⋅34+4.5(34)2+⋯\displaystyle \geo{4.5 + 4.5\cdot\frac34 + 4.5\left(\frac34\right)^2 + \cdots} =4.51−34\displaystyle = \geo{\frac{4.5}{1-\frac34}} =18\displaystyle = \geo{18}
∑n=1∞3nn!\displaystyle \sum_{n=1}^{\infty}\tm{\frac{3^n}{n!}} ≤3+4.5+18\displaystyle \le \tm{3} + \tm{4.5} + \geo{18} =25.5\displaystyle = 25.5

The terms rose at first, but the first two only add a finite amount. What decides convergence is that the multiplier approaches 0\mul{0}, so eventually a convergent geometric series covers the rest.

2

The ratio test

The same idea works for any series with positive terms. We take the limit of each term divided by the term right before it, and call that limit L\mul{L}.

L\displaystyle \mul{L} =lim⁡n→∞an+1an(an>0)\displaystyle = \lim_{n\to\infty}\frac{\nx{a_{n+1}}}{\tm{a_n}} \qquad (\tm{a_n} > 0)
L<1:  ∑an convergesL>1:  ∑an divergesL=1:  no conclusion\displaystyle \begin{aligned} \mul{L} < 1 &:\ \ {\textstyle\sum} \tm{a_n} \text{ converges} \\ \mul{L} > 1 &:\ \ {\textstyle\sum} \tm{a_n} \text{ diverges} \\ \mul{L} = 1 &:\ \ \text{no conclusion} \end{aligned}

If L<1\mul{L} < 1, we can pick a fixed number r\geo{r} between L\mul{L} and 11. The ratios approach L\mul{L}, so from some nn onward they stay less than r\geo{r}, and the remaining terms are bounded by a geometric series with ratio r<1\geo{r} < 1. For ∑3nn!\sum \tm{\frac{3^n}{n!}}, L\mul{L} was 0\mul{0}, and we used r=34\geo{r = \frac34}. For a geometric series, every ratio is the same, so the test agrees with what we already know.

∑(23)n:an+1an=23,\displaystyle {\sum\tm{\left(\frac23\right)^n}:\quad \frac{\nx{a_{n+1}}}{\tm{a_n}} = \mul{\frac23},}
L=23<1\displaystyle {\mul{L = \frac23} < 1}

Now let's test a series where the ratio changes at every step.

Determine whether ∑n=1∞2nn2\sum_{n=1}^{\infty}\tm{\frac{2^n}{n^2}} converges or diverges
∑n=1∞2nn2\displaystyle \sum_{n=1}^{\infty}\tm{\frac{2^n}{n^2}}
lim⁡n→∞an+1an\displaystyle \lim_{n\to\infty}\frac{\nx{a_{n+1}}}{\tm{a_n}} =  ?\displaystyle = \;?
Text solution
Write an+1\nx{a_{n+1}}Replace every nn in an\tm{a_n} with n+1n + 1, including the exponent.
an=2nn2\displaystyle \tm{a_n = \frac{2^n}{n^2}}
an+1=2n+1(n+1)2\displaystyle \nx{a_{n+1} = \frac{2^{n+1}}{(n+1)^2}}
Divide by an\tm{a_n}Multiply by the reciprocal n22n\tm{\frac{n^2}{2^n}}, then group the powers of 22 and the squares.
an+1an\displaystyle \frac{\nx{a_{n+1}}}{\tm{a_n}} =2n+1(n+1)2⋅n22n\displaystyle = \nx{\frac{2^{n+1}}{(n+1)^2}}\cdot\tm{\frac{n^2}{2^n}}
=2⋅(nn+1)2→2\displaystyle = \mul{2}\cdot\pw{\left(\frac{n}{n+1}\right)^2} \to \mul{2}
The terms growL=2>1\mul{L = 2} > 1, so the series diverges. The terms dip from 2\tm{2} to 1\tm{1} to 89\tm{\frac89}, but from n=3n = 3 on the ratio is greater than 11, so each term is larger than the one before and the terms can't approach 00.
n\displaystyle n =3:2⋅(34)2\displaystyle = 3:\quad \mul{2}\cdot\pw{\left(\frac34\right)^2} =98>1\displaystyle = \mul{\frac98} > 1
3

When the limit of the ratio is 11

The ratio test decides a series when L<1\mul{L} < 1 or L>1\mul{L} > 1. But what happens when L=1\mul{L} = 1? Let's try it on two p-series we already know.

Why the ratio test gives no conclusion when L=1\mul{L = 1}, and which series give L=1\mul{L = 1}
L\displaystyle \mul{L} =lim⁡n→∞an+1an\displaystyle = \lim_{n\to\infty}\frac{\nx{a_{n+1}}}{\tm{a_n}}
L<1:convergesL>1:divergesL=1:?\displaystyle \begin{aligned} \mul{L} &< 1{:} &&\text{converges} \\ \mul{L} &> 1{:} &&\text{diverges} \\ \mul{L} &= 1{:} &&? \end{aligned}
Text solution
The harmonic series: L=1\mul{L = 1}Every ratio 12,23,34,…\mul{\frac12, \frac23, \frac34, \dots} is less than 11, but they approach 11, and the series diverges.
∑1n:an+1an\displaystyle \sum\tm{\frac1n}:\quad \frac{\nx{a_{n+1}}}{\tm{a_n}} =nn+1→1\displaystyle = \mul{\frac{n}{n+1} \to 1}
diverges (p=1)\displaystyle \text{diverges} \ (p = 1)
No fixed r<1\geo{r} < 1 worksWhatever fixed r<1\geo{r} < 1 we pick, the ratios eventually pass it, so no convergent geometric series covers the terms. Each term is smaller than the last, but not at most a fixed fraction of it.
nn+1>0.9  (n≥10)\displaystyle \mul{\frac{n}{n+1}} > \geo{0.9} \ \ (n \ge 10)
nn+1>0.99  (n≥100)\displaystyle \mul{\frac{n}{n+1}} > \geo{0.99} \ \ (n \ge 100)
∑1n2\sum\tm{\frac{1}{n^2}}: L=1\mul{L = 1} tooThis one converges, so L=1\mul{L = 1} can go either way, and the test gives no conclusion.
an+1an\displaystyle \frac{\nx{a_{n+1}}}{\tm{a_n}} =(nn+1)2→1\displaystyle = \mul{\left(\frac{n}{n+1}\right)^2 \to 1}
converges (p=2)\displaystyle \text{converges} \ (p = 2)

When L=1\mul{L = 1}, we go back to p-series or a comparison test. We can often tell L\mul{L} will be 11 before doing any algebra. When an\tm{a_n} is a product of pieces, an+1an\frac{\nx{a_{n+1}}}{\tm{a_n}} is the product of the ratios of those pieces, and any fixed power np\pw{n^p} gives a ratio that approaches 11.

3n:  3,12n:  12n!:  n+1,1n!:  1n+1→0n2:  (n+1n)2→1,1n2:  (nn+1)2→1\displaystyle \begin{alignedat}{2} \mul{3^n} &:\ \ \mul{3}, &\qquad\qquad \mul{\frac{1}{2^n}} &:\ \ \mul{\frac12} \\[6pt] \mul{n!} &:\ \ \mul{n+1}, &\qquad\qquad \mul{\frac{1}{n!}} &:\ \ \mul{\frac{1}{n+1}} \to 0 \\[6pt] \pw{n^2} &:\ \ \pw{\left(\frac{n+1}{n}\right)^2} \to 1, &\qquad\qquad \pw{\frac{1}{n^2}} &:\ \ \pw{\left(\frac{n}{n+1}\right)^2} \to 1 \end{alignedat}
3n:  312n:  12n!:  n+11n!:  1n+1→0n2:  (n+1n)2→11n2:  (nn+1)2→1\displaystyle \begin{aligned} \mul{3^n} &:\ \ \mul{3} \\[6pt] \mul{\frac{1}{2^n}} &:\ \ \mul{\frac12} \\[16pt] \mul{n!} &:\ \ \mul{n+1} \\[6pt] \mul{\frac{1}{n!}} &:\ \ \mul{\frac{1}{n+1}} \to 0 \\[16pt] \pw{n^2} &:\ \ \pw{\left(\frac{n+1}{n}\right)^2} \to 1 \\[6pt] \pw{\frac{1}{n^2}} &:\ \ \pw{\left(\frac{n}{n+1}\right)^2} \to 1 \end{aligned}
(1110)2=1.21,\displaystyle {\pw{\left(\frac{11}{10}\right)^2} = 1.21,}
(101100)2=1.0201\displaystyle {\pw{\left(\frac{101}{100}\right)^2} = 1.0201}

That's why the n2\pw{n^2} in ∑2nn2\sum \tm{\frac{2^n}{n^2}} didn't change the limit 2\mul{2}. Even in n1001.01n\tm{\frac{n^{100}}{1.01^n}}, where the first 10,00010{,}000 terms increase, the n100\pw{n^{100}} only gives a factor that approaches 11, so the series converges.

n1001.01n:an+1an\displaystyle {\tm{\frac{n^{100}}{1.01^n}}:\quad \frac{\nx{a_{n+1}}}{\tm{a_n}}} =(n+1n)100⋅11.01→11.01<1\displaystyle = {\pw{\left(\frac{n+1}{n}\right)^{100}}\cdot\mul{\frac{1}{1.01}} \to \mul{\frac{1}{1.01}} < 1}

If an\tm{a_n} is built only out of powers of nn and roots, like nn2+1\tm{\frac{\sqrt n}{n^2 + 1}}, every piece has a ratio that approaches 11. So L=1\mul{L = 1}, and we skip the ratio test and use comparison or p-series right away.

Why isn't it enough for every multiplier to be below 1?

4

Dealing with (2n)!(2n)!

Next, a series with factorials on both the top and the bottom, so we use the ratio test. The new piece is (2n)!(2n)!. When nn goes up by 11, n!\tm{n!} gains one new factor, but 2n2n goes up by 22, so (2n+2)!\nx{(2n + 2)!} gains two.

(n+1)!=(n+1)⋅n!(2n+2)!=(2n+2)(2n+1) (2n)(2n−1)⋯2⋅1⏟(2n)!=(2n+2)(2n+1)⋅(2n)!\displaystyle \begin{aligned} \nx{(n+1)!} &= \mul{(n+1)}\cdot\tm{n!} \\[10pt] \nx{(2n+2)!} &= \mul{(2n+2)(2n+1)}\,\tm{\underbrace{(2n)(2n-1)\cdots 2\cdot 1}_{(2n)!}} \\[2pt] &= \mul{(2n+2)(2n+1)}\cdot\tm{(2n)!} \end{aligned}
(n+1)!=(n+1)⋅n!(2n+2)!=(2n+2)(2n+1)⋅(2n)(2n−1)⋯2⋅1⏟(2n)!=(2n+2)(2n+1)⋅(2n)!\displaystyle \begin{aligned} \nx{(n+1)!} &= \mul{(n+1)}\cdot\tm{n!} \\[10pt] \nx{(2n+2)!} &= \mul{(2n+2)(2n+1)} \\ &\qquad\cdot\tm{\underbrace{(2n)(2n-1)\cdots 2\cdot 1}_{(2n)!}} \\[2pt] &= \mul{(2n+2)(2n+1)}\cdot\tm{(2n)!} \end{aligned}
Determine whether ∑n=1∞(n!)2(2n)!\sum_{n=1}^{\infty}\tm{\frac{(n!)^2}{(2n)!}} converges or diverges
∑n=1∞(n!)2(2n)!\displaystyle \sum_{n=1}^{\infty}\tm{\frac{(n!)^2}{(2n)!}}
L\displaystyle \mul{L} =lim⁡n→∞an+1an\displaystyle = \lim_{n\to\infty}\frac{\nx{a_{n+1}}}{\tm{a_n}} =  ?\displaystyle = \;?
Text solution
Write an+1\nx{a_{n+1}}On the bottom, 2n2n becomes 2(n+1)=2n+22(n + 1) = 2n + 2, so the denominator is (2n+2)!\nx{(2n + 2)!}, not (2n+1)!(2n + 1)!.
an+1\displaystyle \nx{a_{n+1}} =((n+1)!)2(2n+2)!\displaystyle = \nx{\frac{\big((n+1)!\big)^2}{(2n+2)!}}
CancelWith the new factors written out, (n!)2\tm{(n!)^2} and (2n)!\tm{(2n)!} cancel. Then 2n+2=2(n+1)2n + 2 = 2(n + 1), so one factor of n+1n + 1 cancels too.
an+1an\displaystyle \frac{\nx{a_{n+1}}}{\tm{a_n}} =(n+1)2(2n+2)(2n+1)\displaystyle = \mul{\frac{(n+1)^2}{(2n+2)(2n+1)}} =n+12(2n+1)\displaystyle = \mul{\frac{n+1}{2(2n+1)}}
Take the limitThe top is roughly nn and the bottom roughly 4n4n. Since 14<1\mul{\frac14} < 1, the series converges.
n+12(2n+1)→14\displaystyle \mul{\frac{n+1}{2(2n+1)}} \to \mul{\frac14}

Be careful not to confuse (2n)!(2n)! with 2⋅n!2\cdot n!. The parentheses mean we multiply nn by 22 first, then take the factorial.

(2n)!≠2⋅n!:\displaystyle {(2n)! \ne 2\cdot n!:}
4!=24,\displaystyle {4! = 24,}
2⋅2!=4\displaystyle {2\cdot 2! = 4}
5

The root test

Now a series where the whole term is raised to the nth power. Here the ratio test gets messy, so let's find another way to compare with a geometric series.

Determine whether ∑n=1∞(n+12n)n\sum_{n=1}^{\infty}\tm{\left(\frac{n+1}{2n}\right)^n} converges or diverges
∑n=1∞(n+12n)n\displaystyle \sum_{n=1}^{\infty}\tm{\left(\frac{n+1}{2n}\right)^{n}}
converges or diverges?\displaystyle \text{converges or diverges?}
Text solution
The ratio doesn't simplifyThe bases are different, the powers are different, and nothing cancels.
an+1an\displaystyle \frac{\nx{a_{n+1}}}{\tm{a_n}} =(n+22n+2)n+1⋅(2nn+1)n\displaystyle = \nx{\left(\frac{n+2}{2n+2}\right)^{n+1}}\cdot\tm{\left(\frac{2n}{n+1}\right)^n}
Take the nth rootThe nth root undoes the nth power, the same way it gives 34\geo{\frac34} back from (34)n\geo{\left(\frac34\right)^n}. What's left is the base.
(34)nn=34\displaystyle \geo{\sqrt[n]{\left(\frac34\right)^n} = \frac34}
ann=n+12n\displaystyle \mul{\sqrt[n]{a_n} = \frac{n+1}{2n}}
1, 34, 23, 58, ⋯→12\displaystyle \mul{1,\ \frac34,\ \frac23,\ \frac58,\ \dots \to \frac12}
Compare with (34)n\geo{\left(\frac34\right)^n}From n=2n = 2 on, the base is at most 34\geo{\frac34}, so by direct comparison with a convergent geometric series, our series converges.
n\displaystyle n ≥2:n+12n\displaystyle \ge 2:\quad \mul{\frac{n+1}{2n}} ≤34\displaystyle \le \geo{\frac34}
(n+12n)n\displaystyle \tm{\left(\mul{\frac{n+1}{2n}}\right)^n} ≤(34)n\displaystyle \le \geo{\left(\frac34\right)^n}
The root isn't the termThe nth root approaches 12\mul{\frac12}, not 00, but the nth-term test applies to an\tm{a_n} itself, and an\tm{a_n} approaches 00 very quickly.

This works for any series with positive terms, and it's called the root test.

L=lim⁡n→∞ann(an>0)\displaystyle \mul{L = \lim_{n\to\infty}\sqrt[n]{a_n}} \qquad (\tm{a_n} > 0)
L<1:  ∑an convergesL>1:  ∑an divergesL=1:  no conclusion\displaystyle \begin{aligned} \mul{L} < 1 &:\ \ {\textstyle\sum} \tm{a_n} \text{ converges} \\ \mul{L} > 1 &:\ \ {\textstyle\sum} \tm{a_n} \text{ diverges} \\ \mul{L} = 1 &:\ \ \text{no conclusion} \end{aligned}

If L<1\mul{L} < 1, the terms eventually sit under rn\geo{r^n} for a fixed r<1\geo{r} < 1, just as ours sat under (34)n\geo{\left(\frac34\right)^n}. If L>1\mul{L} > 1, then for large nn the terms are greater than 11, so they can't approach 00.

A power of nn outside the nth power, like the n3\pw{n^3} in ∑n3(ln⁡n)n\sum \tm{\frac{n^3}{(\ln n)^n}}, doesn't change the limit, because n1/n\pw{n^{1/n}} approaches 11.

n3n=(n1/n)3\displaystyle \pw{\sqrt[n]{n^3} = \left(n^{1/n}\right)^3}
n=10:  n1/n≈1.26,\displaystyle {n = 10:\ \ \pw{n^{1/n}} \approx 1.26,}
n=100:  n1/n≈1.047,\displaystyle {n = 100:\ \ \pw{n^{1/n}} \approx 1.047,}
n=1000:  n1/n≈1.007\displaystyle {n = 1000:\ \ \pw{n^{1/n}} \approx 1.007}
ln⁡(n1/n)=ln⁡nn→0,\displaystyle {\ln\left(\pw{n^{1/n}}\right) = \frac{\ln n}{n} \to 0,}
n1/n→1\displaystyle {\pw{n^{1/n} \to 1}}

So when you see factorials, use the ratio test, because (n+1)!\nx{(n+1)!} cancels with n!\tm{n!}. When the whole term is raised to the nth power, use the root test, because the nth root gets rid of the power. When you only have powers or roots of nn, both tests give L=1\mul{L = 1}, so use a p-series or a comparison test instead.

(n!)2(2n)!\displaystyle \tm{\frac{(n!)^2}{(2n)!}}
lim⁡an+1an\displaystyle \lim\frac{\nx{a_{n+1}}}{\tm{a_n}} =14\displaystyle = \mul{\frac14}
(n+12n)n\displaystyle \tm{\left(\frac{n+1}{2n}\right)^n}
lim⁡ann=12\displaystyle \mul{\lim\sqrt[n]{a_n} = \frac12}
nn2+1\displaystyle \tm{\frac{\sqrt n}{n^2+1}}
L=1\displaystyle \mul{L = 1}
6

Series with negative terms

So far, every series had positive terms. What if some terms are negative? In the series below, (−4)n(-4)^n is negative when nn is odd, so the terms alternate in sign.

Determine whether ∑n=1∞(−4)n3n n\sum_{n=1}^{\infty}\tm{\frac{(-4)^n}{3^n\,n}} converges or diverges
1\htmlClass{lc-mute}{1}
−1\htmlClass{lc-mute}{-1}
1\htmlClass{lc-mute}{1}
2\htmlClass{lc-mute}{2}
3\htmlClass{lc-mute}{3}
4\htmlClass{lc-mute}{4}
5\htmlClass{lc-mute}{5}
6\htmlClass{lc-mute}{6}
nn
an\htmlClass{lc-pink}{a_n}
Text solution
Use the sizes of the termsTake the ratio of the absolute values, so the minus sign is gone.
∣an+1an∣\displaystyle \left|\frac{\nx{a_{n+1}}}{\tm{a_n}}\right| =4n+13n+1(n+1)⋅3n n4n\displaystyle = \nx{\frac{4^{n+1}}{3^{n+1}(n+1)}}\cdot\tm{\frac{3^n\,n}{4^n}}
=43⋅nn+1→43\displaystyle = \mul{\frac43\cdot\frac{n}{n+1}} \to \mul{\frac43}
Compare with 11L=43>1\mul{L = \frac43} > 1. For n>3n > 3 the ratio is greater than 11, so the terms grow in size and can't approach 00, and the series diverges.
Don't drop the absolute valuesWithout them the ratio approaches −43-\frac43, which is less than 11 only because of the sign. The sign means the terms alternate, not that they get smaller.
an+1an\displaystyle \frac{\nx{a_{n+1}}}{\tm{a_n}} =−43⋅nn+1→−43\displaystyle = {-\frac43\cdot\frac{n}{n+1} \to -\frac43}

So for a series with negative terms, both tests use the sizes ∣an∣|\tm{a_n}|. If L<1\mul{L} < 1, the sum of the absolute values converges, so the series converges absolutely.

L=lim⁡n→∞∣an+1an∣\displaystyle {\mul{L} = \lim_{n\to\infty}\left|\frac{\nx{a_{n+1}}}{\tm{a_n}}\right|}
or\displaystyle {\text{or}}
L=lim⁡n→∞∣an∣n\displaystyle {\mul{L = \lim_{n\to\infty}\sqrt[n]{|a_n|}}}
L<1:  ∑an converges absolutelyL>1:  ∑an divergesL=1:  no conclusion\displaystyle \begin{aligned} \mul{L} < 1 &:\ \ {\textstyle\sum} \tm{a_n} \text{ converges absolutely} \\ \mul{L} > 1 &:\ \ {\textstyle\sum} \tm{a_n} \text{ diverges} \\ \mul{L} = 1 &:\ \ \text{no conclusion} \end{aligned}

For ∑(−3)nn!\sum \tm{\frac{(-3)^n}{n!}}, taking the absolute values gives 3nn!\tm{\frac{3^n}{n!}}. We already found that its ratio approaches 0\mul{0}, so L=0<1\mul{L = 0} < 1, and the series converges absolutely.

∑(−3)nn!:∣an∣=3nn!,\displaystyle {\sum\tm{\frac{(-3)^n}{n!}}:\quad |\tm{a_n}| = \tm{\frac{3^n}{n!}},}
L=0<1\displaystyle {\mul{L = 0} < 1}

A series that converges conditionally, like ∑(−1)nn\sum \tm{\frac{(-1)^n}{\sqrt n}}, always gives L=1\mul{L = 1} when the limit exists. If L\mul{L} were less than 11, the series would converge absolutely, and if it were greater than 11, the series would diverge. So for these we need another test, like the alternating series test.

Summary

The ratio test
L\displaystyle \mul{L} =lim⁡n→∞∣an+1an∣\displaystyle = \lim_{n\to\infty}\left|\frac{\nx{a_{n+1}}}{\tm{a_n}}\right|
L<1:  converges absolutelyL>1:  divergesL=1:  no conclusion\displaystyle \begin{aligned} \mul{L} < 1 &:\ \ \text{converges absolutely} \\ \mul{L} > 1 &:\ \ \text{diverges} \\ \mul{L} = 1 &:\ \ \text{no conclusion} \end{aligned}

If L<1\mul{L} < 1, the ratio ∣an+1an∣\left|\frac{\nx{a_{n+1}}}{\tm{a_n}}\right| stays below a fixed number r<1\geo{r} < 1 for large nn. The sizes of the terms are bounded by a convergent geometric series, so the series converges absolutely. If L>1\mul{L} > 1, the terms ∣an∣|\tm{a_n}| increase for large nn and can't approach 00, so the series diverges. If L=1\mul{L} = 1, the test gives no conclusion.

When the ratio test gives L=1L = 1
∑1n, ∑1n2: L=1\displaystyle {\sum\tm{\frac1n},\ \sum\tm{\frac{1}{n^2}}:\ \mul{L = 1}}
nn2+1: L=1\displaystyle {\tm{\frac{\sqrt n}{n^2+1}}:\ \mul{L = 1}}

Both ∑1n\sum \tm{\frac1n} and ∑1n2\sum \tm{\frac{1}{n^2}} give L=1\mul{L = 1}, but ∑1n\sum \tm{\frac1n} diverges and ∑1n2\sum \tm{\frac{1}{n^2}} converges. For any power np\pw{n^p}, the ratio (n+1n)p\pw{\left(\frac{n+1}{n}\right)^p} approaches 11. If a term is built only from powers of nn, roots, and polynomials, every piece gives a ratio that approaches 11, so the overall limit is always L=1\mul{L = 1}. Skip the ratio test for these series and use a p-series or a comparison test instead. Use the ratio test when terms contain factorials like n!\mul{n!} or exponentials like 3n\mul{3^n}.

Cancelling factorials
(n+1)!\displaystyle \nx{(n+1)!} =(n+1)⋅n!\displaystyle = \mul{(n+1)}\cdot\tm{n!}
(2n+2)!\displaystyle \nx{(2n+2)!} =(2n+2)(2n+1)⋅(2n)!\displaystyle = \mul{(2n+2)(2n+1)}\cdot\tm{(2n)!}

To simplify a ratio of factorials, write out factors of the larger factorial until you reach the smaller one, then cancel. Going from nn to n+1n + 1 adds one new factor, n+1\mul{n + 1}. Going from 2n2n to 2(n+1)=2n+22(n + 1) = 2n + 2 adds two new factors, 2n+1\mul{2n + 1} and 2n+2\mul{2n + 2}.

The root test
L=lim⁡n→∞∣an∣n\displaystyle \mul{L = \lim_{n\to\infty}\sqrt[n]{|a_n|}}
n1/n→1\displaystyle \pw{n^{1/n} \to 1}

When an entire term is raised to the nth power, take the nth root to remove the power. The root test has the same three outcomes as the ratio test. If a fixed power like n3\pw{n^3} is outside the nth power, its nth root (n1/n)3\pw{\left(n^{1/n}\right)^3} approaches 11, so powers of nn don't change L\mul{L}.

Negative terms and conditional convergence

Take the absolute value of each term before applying the ratio test or root test. If L<1\mul{L} < 1, the series converges absolutely, and if L>1\mul{L} > 1, it diverges. For a conditionally convergent series, ∑∣an∣\sum |\tm{a_n}| diverges, so L\mul{L} cannot be less than 11. But the terms approach 00, so L\mul{L} cannot be greater than 11 either. So if the limit exists, L\mul{L} must be 11, and the test tells us nothing. When L=1\mul{L = 1}, use another test, such as the alternating series test.

7

Practice

1
Determine whether the series ∑n=1∞n2 3nn!\displaystyle\sum_{n=1}^{\infty}\frac{n^2\,3^n}{n!} converges or diverges.

The term has three pieces, n2n^2, 3n3^n, and n!n!. When we set up an+1an\frac{a_{n+1}}{a_n}, what factor does each piece bring?

If you got diverges Check whether you found an+1an\frac{a_{n+1}}{a_n} or anan+1\frac{a_n}{a_{n+1}}. The early terms increase at first, but convergence depends only on what happens as nn gets large.

Set up the termsWe have a factorial and an exponential, so let's use the ratio test. Replace every nn with n+1n + 1 to write down an+1\nx{a_{n+1}}.
an=n2 3nn!,\displaystyle {\tm{a_n = \frac{n^2\,3^n}{n!}},}
an+1=(n+1)2 3n+1(n+1)!\displaystyle {\nx{a_{n+1} = \frac{(n+1)^2\,3^{n+1}}{(n+1)!}}}
Find the ratioGroup the powers of nn, the powers of 33, and the factorials together.
an+1an\displaystyle \frac{\nx{a_{n+1}}}{\tm{a_n}} =(n+1)2 3n+1(n+1)!⋅n!n2 3n\displaystyle = \nx{\frac{(n+1)^2\,3^{n+1}}{(n+1)!}}\cdot\tm{\frac{n!}{n^2\,3^n}} =(n+1n)2⋅3⋅1n+1\displaystyle = \pw{\left(\frac{n+1}{n}\right)^2}\cdot\mul{3}\cdot\mul{\frac{1}{n+1}} =3(n+1)n2→0\displaystyle = \mul{\frac{3(n+1)}{n^2}} \to \mul{0}
The limit is less than 11, so the series converges.
L=0<1 → converges\displaystyle \mul{L = 0} < 1 \ \to\ \text{converges}
2
Determine whether the series ∑n=1∞(2n)!3n (n!)2\displaystyle\sum_{n=1}^{\infty}\frac{(2n)!}{3^n\,(n!)^2} converges or diverges.

When nn becomes n+1n + 1, (2n)!(2n)! becomes (2n+2)!(2n + 2)!. How many new factors does that add?

If you got converges Check how (2n+2)!(2n + 2)! expands. It adds two new factors, (2n+2)(2n+1)(2n + 2)(2n + 1), not just 2n+22n + 2.

Expand the factorials(2n+2)!\nx{(2n + 2)!} adds two new factors before reaching (2n)!\tm{(2n)!}, and (n+1)!\nx{(n + 1)!} adds one.
(2n+2)!=(2n+2)(2n+1)⋅(2n)!,\displaystyle {\nx{(2n+2)!} = \mul{(2n+2)(2n+1)}\cdot\tm{(2n)!},}
(n+1)!=(n+1)⋅n!\displaystyle {\nx{(n+1)!} = \mul{(n+1)}\cdot\tm{n!}}
Simplify the ratioCancel (2n)!\tm{(2n)!} and (n!)2\tm{(n!)^2}, then factor 2n+2=2(n+1)2n + 2 = 2(n + 1) to cancel one factor of n+1n + 1.
an+1an\displaystyle \frac{\nx{a_{n+1}}}{\tm{a_n}} =(2n+2)!3n+1((n+1)!)2⋅3n (n!)2(2n)!\displaystyle = \nx{\frac{(2n+2)!}{3^{n+1}\big((n+1)!\big)^2}}\cdot\tm{\frac{3^n\,(n!)^2}{(2n)!}} =(2n+2)(2n+1)3(n+1)2\displaystyle = \mul{\frac{(2n+2)(2n+1)}{3(n+1)^2}} =2(n+1)(2n+1)3(n+1)2\displaystyle = \mul{\frac{2(n+1)(2n+1)}{3(n+1)^2}} =2(2n+1)3(n+1)→43\displaystyle = \mul{\frac{2(2n+1)}{3(n+1)}} \to \mul{\frac43}
The limit is 43\mul{\frac43}, which is greater than 11, so the series diverges.
L=43>1 → diverges\displaystyle \mul{L = \frac43} > 1 \ \to\ \text{diverges}
3
Determine whether the series ∑n=1∞nn2+1\displaystyle\sum_{n=1}^{\infty}\frac{\sqrt n}{n^2+1} converges or diverges.

The terms are built only from powers of nn. What will the ratio test give, and which test should you use instead?

If you got diverges Check what L=1L = 1 means for the ratio test. It gives no conclusion, not divergence. Try comparing with a pp-series.

Try the ratio testThe terms are made only of powers of nn, so the ratio test gives L=1\mul{L = 1}, which tells us nothing.
an+1an\displaystyle \frac{\nx{a_{n+1}}}{\tm{a_n}} =n+1(n+1)2+1⋅n2+1n\displaystyle = \nx{\frac{\sqrt{n+1}}{(n+1)^2+1}}\cdot\tm{\frac{n^2+1}{\sqrt n}} =n+1n⋅n2+1(n+1)2+1→1\displaystyle = \pw{\sqrt{\frac{n+1}{n}}}\cdot\pw{\frac{n^2+1}{(n+1)^2+1}} \to \mul{1}
Use limit comparisonFor large nn, nn2+1\tm{\frac{\sqrt n}{n^2+1}} behaves like nn2=1n32\frac{\sqrt n}{n^2} = \frac{1}{n^{\frac{3}{2}}}, so we use the limit comparison test.
nn2+1⋅n32\displaystyle \tm{\frac{\sqrt n}{n^2+1}}\cdot n^{\frac32} =n2n2+1→1\displaystyle = \frac{n^2}{n^2+1} \to 1
The limit is 11, and ∑1n32\sum \frac{1}{n^{\frac{3}{2}}} is a convergent pp-series, so our series converges too.
∑1n32: p\displaystyle \sum\frac{1}{n^{\frac32}}:\ p =32>1 → ∑nn2+1 converges\displaystyle = \frac32 > 1 \ \to\ \sum\tm{\frac{\sqrt n}{n^2+1}} \text{ converges}
4
Determine whether the series ∑n=2∞n3(ln⁡n)n\displaystyle\sum_{n=2}^{\infty}\frac{n^3}{(\ln n)^n} converges or diverges.

The whole denominator is raised to the nthn\text{th} power. What do you get if you take the nthn\text{th} root of the whole term?

Rewrite n3n\sqrt[n]{n^3} as (n1/n)3(n^{1/n})^3. What does n1/nn^{1/n} approach as nn goes to infinity?

Apply the root testThe denominator is an nthn\text{th} power, so we apply the root test and rewrite the numerator using n1/n\pw{n^{1/n}}.
ann\displaystyle \mul{\sqrt[n]{a_n}} =n3nln⁡n\displaystyle = \frac{\pw{\sqrt[n]{n^3}}}{\mul{\ln n}} =(n1/n)3ln⁡n\displaystyle = \frac{\pw{\left(n^{1/n}\right)^3}}{\mul{\ln n}}
Evaluate the limitAs nn goes to infinity, n1/n\pw{n^{1/n}} approaches 11 while ln⁡n\mul{\ln n} goes to infinity.
(n1/n)3→1,\displaystyle {\pw{\left(n^{1/n}\right)^3} \to 1,}
ln⁡n→∞,\displaystyle {\mul{\ln n} \to \infty,}
(n1/n)3ln⁡n→0\displaystyle {\frac{\pw{\left(n^{1/n}\right)^3}}{\mul{\ln n}} \to \mul{0}}
The limit is 0\mul{0}, which is less than 11, so the series converges by the root test.
L=0<1 → converges\displaystyle \mul{L = 0} < 1 \ \to\ \text{converges}
5
A sequence is defined by a1=1a_1 = 1 and an+1=5n+14n+3 ana_{n+1} = \dfrac{5n+1}{4n+3}\,a_n for n≥1n \ge 1. Does ∑n=1∞an\displaystyle\sum_{n=1}^{\infty} a_n converge or diverge?

The problem defines an+1a_{n+1} in terms of ana_n. What is an+1an\frac{a_{n+1}}{a_n}?

If you got converges Check whether the terms keep decreasing. The ratio an+1an\frac{a_{n+1}}{a_n} starts at 67\frac67, but for n≥2n \ge 2 it's at least 11.

Find the ratioDivide both sides of the recursive formula by an\tm{a_n} to get the ratio directly.
an+1an\displaystyle \frac{\nx{a_{n+1}}}{\tm{a_n}} =5n+14n+3→54\displaystyle = \mul{\frac{5n+1}{4n+3}} \to \mul{\frac54}
Check the early ratiosThe ratio starts below 11 at n=1n = 1, but it reaches 11 at n=2n = 2 and stays above 11 from n=3n = 3 on.
n\displaystyle n =1,2,3:an+1an\displaystyle = 1, 2, 3:\quad \frac{\nx{a_{n+1}}}{\tm{a_n}} =67, 1, 1615\displaystyle = \mul{\frac67,\ 1,\ \frac{16}{15}}
The limit is 54>1\mul{\frac54} > 1, so the series diverges by the ratio test.
L=54>1 → diverges\displaystyle \mul{L = \frac54} > 1 \ \to\ \text{diverges}
6
Determine whether the series ∑n=1∞2n n!nn\displaystyle\sum_{n=1}^{\infty}\frac{2^n\,n!}{n^n} converges or diverges. (Challenge)

Write out an+1a_{n+1} and split (n+1)n+1(n+1)^{n+1} into (n+1)(n+1)n(n+1)(n+1)^n. What cancels?

Rewrite (nn+1)n\left(\frac{n}{n+1}\right)^n as 1(1+1n)n\frac{1}{\left(1 + \frac1n\right)^n}. What does (1+1n)n\left(1 + \frac1n\right)^n approach as nn goes to infinity?

If you got diverges Look closely at (nn+1)n\left(\frac{n}{n+1}\right)^n. The base approaches 11 while the exponent goes to infinity, so this is an indeterminate form. The expression approaches 1e\frac1e, not 11.

Write out an+1\nx{a_{n+1}}Replace every nn with n+1n + 1.
an+1=2n+1 (n+1)!(n+1)n+1\displaystyle \nx{a_{n+1} = \frac{2^{n+1}\,(n+1)!}{(n+1)^{n+1}}}
Simplify the ratioSplit (n+1)n+1(n+1)^{n+1} into (n+1)(n+1)n(n+1)(n+1)^n so the n+1n + 1 from the factorial cancels.
an+1an\displaystyle \frac{\nx{a_{n+1}}}{\tm{a_n}} =2n+1 (n+1)!(n+1)n+1⋅nn2n n!\displaystyle = \nx{\frac{2^{n+1}\,(n+1)!}{(n+1)^{n+1}}}\cdot\tm{\frac{n^n}{2^n\,n!}} =2(n+1) nn(n+1)n+1\displaystyle = \mul{\frac{2(n+1)\,n^n}{(n+1)^{n+1}}} =2(nn+1)n\displaystyle = \mul{2\left(\frac{n}{n+1}\right)^n}
Evaluate the limitDivide the top and bottom of nn+1\frac{n}{n+1} by nn to get 11+1/n\frac{1}{1 + 1/n}, and recall that (1+1n)n→e\left(1 + \frac1n\right)^n \to e.
(nn+1)n\displaystyle \left(\frac{n}{n+1}\right)^n =1(1+1n)n→1e\displaystyle = \frac{1}{\left(1+\frac1n\right)^n} \to \mul{\frac1e}
Since e≈2.718e \approx 2.718, the limit is 2e<1\mul{\frac2e} < 1, so the series converges by the ratio test.
L=2e≈0.74<1 → converges\displaystyle \mul{L = \frac2e} \approx 0.74 < 1 \ \to\ \text{converges}