Skip to content
Topos Uranos

Summary

The ratio and root tests compare a series with a geometric one, and for that reason they cannot prove the convergence of any series whose terms are at least a positive multiple of a power of 1n\frac{1}{n}; in the series of reciprocals of powers, the ratio an+1an\frac{a_{n+1}}{a_n} tends to 11. This lesson compares the ratio with that of series that converge or diverge more slowly. Kummer’s test gives the common scheme: once a positive sequence (cn)(c_n) has been chosen, an inequality between cnanan+1c_n \frac{a_n}{a_{n+1}} and cn+1c_{n+1} makes the sequence (cnan)(c_n a_n) decreasing and bounds the partial sums, or makes it increasing and compares the series with that of the 1cn\frac{1}{c_n}. With cn=nc_n = n one obtains Raabe’s test, which measures the distance from the ratio to 11 in multiples of 1n\frac{1}{n}; with cn=nln⁡nc_n = n \ln n, Bertrand’s; and from the two, Gauss’s, which decides every series whose ratio is a rational function of the index tending to 11, such as the hypergeometric series. The logarithmic test compares the term with the powers of 1n\frac{1}{n}, and Ermakov’s test compares the integral of a decreasing function with that of its composition with the exponential; with it one proves the scale of the series with iterated logarithms in the denominator. The lesson ends with a reasoned guide for choosing the test.

Learning objectives

  1. Prove Kummer’s test and deduce from it Raabe’s, with Duhamel’s form, and Bertrand’s, understanding each one as the choice of a comparison sequence.
  2. Prove Gauss’s test, with the case μ=1\mu = 1, and use it to decide the series whose ratio is a rational function of the index tending to 11.
  3. Prove the logarithmic test and Ermakov’s test, and with the latter the scale of the series with iterated logarithms in the denominator.
  4. Choose with reasons the test that decides a series with positive terms, and recognize when a test does not decide.

Comparing the ratios

The lesson on ratio and root tests proved that, if an≥cnpa_n \geq \frac{c}{n^p} from some index on, with c>0c > 0, neither the root nor the ratio proves the convergence of the series, and it announced this lesson: the ratio an+1an\frac{a_{n+1}}{a_n} is compared here with that of the series of reciprocals of powers and with that of others that converge or diverge even more slowly, and not with that of the geometric series. The notation of the earlier lessons is kept: Sn=∑k=1nakS_n = \sum_{k=1}^{n} a_k, the sum SS, the remainder Rn=S−SnR_n = S - S_n, the series from an index, and the notation ∑n=1∞an<∞\sum_{n=1}^{\infty} a_n < \infty, which the lesson on series with positive terms reserves for saying that a series with nonnegative terms converges. Throughout the lesson the terms are positive, so that one can divide by them; if they are positive only from some index on, the proposition on the tail of a series, from the lesson on numerical series and their convergence, allows one to discard the first ones.

LemmaComparison of ratios

Let (an)(a_n) and (bn)(b_n) be sequences of positive terms and NN a natural number such that an+1an≤bn+1bn\frac{a_{n+1}}{a_n} \leq \frac{b_{n+1}}{b_n} for every n≥Nn \geq N. Then an≤aNbN bna_n \leq \frac{a_N}{b_N} \, b_n for every n≥Nn \geq N. Consequently, if ∑n=1∞bn\sum_{n=1}^{\infty} b_n converges, ∑n=1∞an\sum_{n=1}^{\infty} a_n also converges; and if ∑n=1∞an\sum_{n=1}^{\infty} a_n diverges, ∑n=1∞bn\sum_{n=1}^{\infty} b_n also diverges. If the terms of a series decrease, from each one to the following, in a proportion at least as strong as those of another, the first series stays below a multiple of the second.

Proof

  1. an+1bn+1≤anbn→an≤aNbN bn(n≥N)\frac{a_{n+1}}{b_{n+1}} \leq \frac{a_n}{b_n} \rightarrow \resaltar{a_n \leq \frac{a_N}{b_N} \, b_n} \quad (n \geq N)

    Let n≥Nn \geq N. On multiplying the hypothesis by the positive number anbn+1\frac{a_n}{b_{n+1}}, the quotient anbn\frac{a_n}{b_n} does not grow from one index to the following; by induction on nn, starting from n=Nn = N, it does not exceed its value at NN.

  2. 0<an≤aNbN bn(n≥N)→(∑n=1∞bn<∞→∑n=1∞an<∞)0 < a_n \leq \frac{a_N}{b_N} \, b_n \quad (n \geq N) \rightarrow \left(\sum_{n=1}^{\infty} b_n < \infty \rightarrow \resaltar{\sum_{n=1}^{\infty} a_n < \infty}\right)

    If ∑n=1∞bn\sum_{n=1}^{\infty} b_n converges, by the proposition on linearity and monotonicity of series, from the lesson on numerical series and their convergence, the series of the aNbN bn\frac{a_N}{b_N} \, b_n also converges, and the theorem of the comparison test for series, from the lesson on series with positive terms, with the inequality from NN on, gives the convergence of ∑n=1∞an\sum_{n=1}^{\infty} a_n. The second statement is the contrapositive of the first.

With bn=1npb_n = \frac{1}{n^p}, the laws of real exponents, from the lesson on the exponential function and powers with real exponents, give bn+1bn=(nn+1)p\frac{b_{n+1}}{b_n} = \left(\frac{n}{n + 1}\right)^p: if an+1an≤(nn+1)p\frac{a_{n+1}}{a_n} \leq \left(\frac{n}{n + 1}\right)^p from some index on, with p>1p > 1, the series converges, because the series of the 1np\frac{1}{n^p} converges, by the proposition on the series of reciprocals of powers, from the lesson on series with positive terms. Kummer’s test generalizes this idea: instead of fixing the series used for comparison, it weights the terms with a positive sequence (cn)(c_n) that is chosen according to the series.