Skip to content
Topos Uranos

Numerical series: synthesis and challenges

Summary

This lesson closes the unit on numerical series. It gathers into a single deductive chain what the six content lessons proved: a series is the sequence of its partial sums, and its sum, the limit of that sequence; the Cauchy criterion decides convergence without knowing the sum, and for series with nonnegative terms the completeness axiom reduces it to the boundedness of the partial sums, whence follow comparison, the integral test, the Euler-Mascheroni constant and condensation; comparison with the geometric series gives the ratio and root tests, and comparison of ratios, the finer tests of Kummer, Raabe, Bertrand and Gauss; Abel’s summation by parts gives the Leibniz, Dirichlet and Abel tests, which govern the series that converge by cancellation; and absolute convergence is the condition that allows a series to be rearranged and multiplied without changing its sum. It explains how the unit rests on Calculus I and on the preceding units, which of their promises it fulfills and what it leaves to the unit on Taylor polynomials and power series. It ends with an exit assessment, parallel to the entry assessment, and with ten challenges, harder than the problems of the lessons, each of which combines results from several of them.

Learning objectives

  1. Set out in logical order the results of the unit, indicating on which results of Calculus I and of the preceding units each one depends and in which proofs the Cauchy criterion for series, the comparison test for series, the geometric series and Abel’s summation by parts are reused.
  2. Distinguish absolute from conditional convergence, and recognize which operations on a series (grouping, ungrouping, rearranging, multiplying) preserve its nature and its sum, and under which hypotheses.
  3. Relate the unit to Calculus I, to the units on sequences, the definite integral and improper integrals, and to the unit on Taylor polynomials and power series.
  4. Solve non-routine problems that combine several lessons: choosing the test that decides a series, comparing it with a well-chosen scale, grouping its terms, summing by parts and rearranging.

What the unit achieved

The unit on sequences studied the limit of a sequence, and the unit on improper integrals, the limit of the integrals of a function over ever larger intervals. This unit applied the same passage to the limit to sums: the series with general term ana_n is the sequence of its partial sums, and its sum, when it exists, is the following limit:

∑n=1∞an=lim⁡n→∞Sn,Sn=∑k=1nak\sum_{n=1}^{\infty} a_n = \lim_{n \to \infty} S_n, \qquad S_n = \sum_{k=1}^{n} a_k

If this is so, everything proved for sequences holds for series, and the central question of the unit was not how to compute a sum, but when it exists without knowing it. The responses repeated those of improper integrals (the Cauchy criterion, monotonicity with the completeness axiom, comparison and absolute convergence) and added two that have no continuous counterpart: comparison with the geometric series, which gives the ratio and root tests and, refined, the finer tests; and the question of order, because a series is not a set of numbers that are added, but a sequence of numbers that are added in an order. Therefore, the chain has six segments: the series and its sum, series with nonnegative terms, the ratio and root tests, the finer tests, conditional convergence and, finally, rearrangements and products.

The series and its sum

The lesson on numerical series and their convergence defined a series, its partial sums and its sum, and distinguished divergence to infinity from oscillation. It summed the geometric series, which converges if and only if its common ratio has absolute value less than 11, and the telescoping series; it proved that the general term of a convergent series tends to 00, and that this condition is not sufficient, because the harmonic series diverges: its partial sums H2mH_{2^m} do not fall below 1+m21 + \frac{m}{2}. The Cauchy criterion for series, which is the criterion for sequences applied to the partial sums, decides convergence by looking only at the stretches of terms; linearity, monotonicity and the tail of a series showed that changing finitely many terms does not change its nature. Finally, the lesson studied grouping: grouping without rearranging preserves the sum of a convergent series, the converse fails, as Grandi’s series shows, and it holds if the terms tend to 00 and the groups have bounded length, or if the terms of each group have the same sign.

Series with nonnegative terms

The lesson on series with positive terms observed that their partial sums increase, so that, by the completeness axiom, such a series converges if and only if they are bounded. From this follow the comparison test for series and the limit comparison test, with its cases of limit 00 and of infinite limit. The integral test enclosed the area under the graph of a positive decreasing function between two families of rectangles, as the units on the definite integral, its applications and improper integrals had promised: the series and the integral converge together, and the remainder lies between two integrals. With it, the series of reciprocals of the powers with exponent pp converges if and only if p>1p > 1, and the difference between a partial sum and the corresponding integral decreases toward a limit; for the harmonic series, that limit is the Euler-Mascheroni constant, γ\gamma, with 12(n+1)<Hn−ln⁡n−γ<12n\frac{1}{2(n + 1)} < H_n - \ln n - \gamma < \frac{1}{2n}, and from it follows that Hqn−Hn→ln⁡qH_{qn} - H_n \to \ln q. The Cauchy condensation test replaced a series with decreasing terms by another that uses only the terms of index 2n2^n, and decided the series with a logarithm in the denominator.

The ratio and root tests

The lesson on the ratio and root tests compared a series with the geometric one without knowing its ratio. With the limits superior and inferior of the unit on sequences, the root test decides if lim sup⁡n→∞ann\limsup_{n \to \infty} \sqrt[n]{a_n} is different from 11, and the ratio test, if the ratios an+1an\frac{a_{n + 1}}{a_n} stay below a number less than 11 or do not fall below 11 from some index on. The inequality between the ratios and the roots encloses the two limits of the roots between the two limits of the ratios, whence the root test decides whenever the ratio test decides, and not the other way around. When they decide convergence, the general term tends to 00 at a geometric rate and the remainder is bounded by that of a geometric series; and the series of reciprocals of powers give 11 in both tests, which in that case do not distinguish convergence from divergence.

The finer tests

The lesson on the finer convergence tests studied the series in which the ratio tends to 11. Kummer’s test compares the ratios of the series with those of an auxiliary sequence (cn)(c_n); with cn=nc_n = n it gives Raabe’s test, and with cn=nln⁡nc_n = n \ln n, Bertrand’s. Gauss’s test, which looks at the following term of the expansion of the ratio, brings them together and decides in particular the series whose ratios are quotients of two polynomials of the same degree and tend to 11. The logarithmic test writes with logarithms the comparisons with the series of reciprocals of powers and with those with a logarithm in the denominator, and Ermakov’s test, with a change of variable in the integral, decides the scale of iterated logarithms. The lesson ended with a reasoned guide for choosing the test.

Conditional convergence

The lesson on alternating series and conditional convergence distinguished absolute convergence, which implies convergence by the Cauchy criterion, from conditional convergence. The Leibniz test gives the convergence of an alternating series whose terms decrease toward 00, with the remainder enclosed between 00 and the first omitted term, and the alternating harmonic series is the example of a series that converges conditionally. Abel’s summation by parts, a discrete version of integration by parts, gave Dirichlet’s and Abel’s tests for series, parallel to those for improper integrals, and with them the series of sin⁡(nx)n\frac{\sin(nx)}{n} and of cos⁡(nx)n\frac{\cos(nx)}{n} converge when xx is not a multiple of 2π2\pi.

Rearrangements and products

The lesson on absolute convergence, rearrangements and products of series defined the rearrangement of a series by means of a bijection σ\sigma from N\mathbb{N} to N\mathbb{N}. Dirichlet’s theorem on rearrangements ensures that an absolutely convergent series can be rearranged without changing its sum; on the other hand, in a conditionally convergent series the series of the positive terms and of the negative terms diverge, and the Riemann rearrangement theorem rearranges it so that it sums to any real number, or diverges to ∞\infty or to −∞-\infty. With the harmonic sums of the lesson on series with positive terms, the sum of the alternating harmonic series turned out to be ln⁡2\ln 2, without power series. Finally, the Cauchy product of two series converges to the product of their sums if one of them converges absolutely and the other converges (Mertens’s theorem), and it may diverge if both converge only conditionally.

What is admitted, what is proved and what remains

The balance of the unit can be summarized as follows:

  • Admitted are the completeness axiom of Algebra and Geometry I, with the Archimedean property that is deduced from it there; the results of Algebra and Geometry on summations, telescoping sums and the trigonometric functions; those of Calculus I on limits, equivalent infinitesimals, L’Hôpital’s rule, the exponential and the logarithm; and those of the preceding units on sequences, the definite integral and improper integrals.
  • Proved are the definition of the sum of a series and its properties, all the convergence tests of the unit, the theorems on rearrangements and products, the Euler-Mascheroni constant and the sum ln⁡2\ln 2. Several promises are thus fulfilled: that of the lesson on Cauchy sequences, which announced the Cauchy criterion applied to partial sums; those of the introduction and the synthesis of the unit on sequences, which announced the ratio and root tests with the limits superior and inferior; those of the introduction and the synthesis of the unit on the definite integral, of the introduction and the synthesis of the unit on its applications and of the introduction of the unit on improper integrals, which announced the integral test with rectangles below and above a decreasing graph; and those of the lesson on conditional convergence of improper integrals and the synthesis of that unit, which announced the harmonic series, the same Dirichlet argument with sums and the plan of this unit.
  • Four instruments are reused instead of repeating arguments: the Cauchy criterion for series, which gives absolute convergence, Dirichlet’s and Abel’s tests, Dirichlet’s theorem on rearrangements and Mertens’s theorem; the comparison test for series, which decides almost all concrete convergences and underlies the limit comparison test, Kummer’s test and the decomposition into positive and negative parts; the geometric series, which gives the majorant of the ratio and root tests; and Abel’s summation by parts, which gives Dirichlet’s and Abel’s tests.
  • Left for later is what depends on a variable: the unit on Taylor polynomials and power series applies these tests to the series whose terms are powers of xx. Also left, as facts from later courses on which nothing here depends, are the value ∑n=1∞1n2=π26\sum_{n=1}^{\infty} \frac{1}{n^2} = \frac{\pi^2}{6}, the open question of the irrationality of γ\gamma, Fourier series and infinite products.

The following figure gathers the six segments into a single chain. Each node of the column is a result, numbered according to the logical order and joined by an arc to the earlier links that its proof uses; around it float the supports, that is, the principles and tools to which the chain resorts again and again, each one joined by a solid line to the links that rest on it and by a dotted one to the link where it is established, except those that come from Calculus I, from Algebra and Geometry or from the preceding units, which are drawn as rings. Clicking a support highlights all the links that use it, and each link refers to the lesson in which it is proved.

Supports of the chain

  • ACompleteness axiom. Axiom of Algebra and Geometry I: every nonempty set of real numbers that is bounded above has a supremum. Here it gives the sum of a series with nonnegative terms as the supremum of its partial sums, and the limit of the monotonic and bounded sequences of the unit. Admitted without proof in this unit.
  • BCalculus I and Algebra and Geometry. Proved in Calculus I and in Algebra and Geometry: summations and telescoping sums, the triangle inequality, limits of functions, equivalent infinitesimals, L’Hôpital’s rule, the exponential and the logarithm, powers with real exponents, monotonicity from the sign of the derivative and the trigonometric formulas. Admitted without proof in this unit.
  • CSequences. From the unit on sequences: the limit theorems, the convergence of monotonic and bounded sequences, subsequences, the limits superior and inferior, the Cauchy criterion and the Stolz-Cesàro lemma. Admitted without proof in this unit.
  • DIntegrals. From the units on the definite integral and improper integrals: the integrability of monotonic functions, the bounds and the additivity of the integral, Barrow’s rule, the change of variable and the improper integrals of nonnegative functions and of powers. Admitted without proof in this unit.
  • EGeometric series. If a≠0a \neq 0, the series ∑n=1∞arn−1\sum_{n=1}^{\infty} a r^{n - 1} converges if and only if ∣r∣<1\abs{r} < 1, with sum a1−r\frac{a}{1 - r}: the scale against which the ratio and root tests are compared. Established in: 1. The geometric series.
  • FCauchy criterion. The series ∑n=1∞an\sum_{n=1}^{\infty} a_n converges if and only if for every ε>0\varepsilon > 0 there is NN with ∣∑k=n+1mak∣<ε\abs{\sum_{k=n + 1}^{m} a_k} < \varepsilon whenever m>n≥Nm > n \geq N: the way to prove convergence without knowing the sum. Established in: 5. Cauchy criterion for series.
  • GComparison. If 0≤an≤bn0 \leq a_n \leq b_n from some index on and ∑n=1∞bn\sum_{n=1}^{\infty} b_n converges, then ∑n=1∞an\sum_{n=1}^{\infty} a_n converges: the usual way to decide a concrete convergence. Established in: 11. Comparison test for series.
  • HSummation by parts. If Ak−Ak−1=akA_k - A_{k - 1} = a_k, the sum ∑k=n+1makbk\sum_{k=n + 1}^{m} a_k b_k is equal to Ambm−Anbn+1−∑k=n+1m−1Ak (bk+1−bk)A_m b_m - A_n b_{n + 1} - \sum_{k=n + 1}^{m - 1} A_k \, (b_{k + 1} - b_k): the way to transfer the oscillation of some terms to their sums. Established in: 42. Abel summation by parts.
  1. Segment 1 The series and its sum

    • Define the sum of a series and sum the geometric series and the telescoping series.
    • Prove the necessary condition for convergence, the divergence of the harmonic series and the Cauchy criterion for series.
    • Decide when grouping or ungrouping the terms preserves the sum.
    1. The geometric series. If a≠0a \neq 0, the series ∑n=1∞arn−1\sum_{n=1}^{\infty} a r^{n - 1} converges if and only if ∣r∣<1\abs{r} < 1, and then its sum is a1−r\frac{a}{1 - r}; if r≥1r \geq 1 it diverges to ∞\infty or to −∞-\infty, according to the sign of aa, and if r≤−1r \leq -1 it oscillates.

      Rests on: C (Sequences), B (Calculus I and Algebra and Geometry). Proved in: Numerical series and their convergence.

    2. Telescoping series. The series ∑n=1∞(bn−bn+1)\sum_{n=1}^{\infty} \left(b_n - b_{n + 1}\right) converges if and only if (bn)(b_n) converges, and its sum is b1−lim⁡n→∞bnb_1 - \lim_{n \to \infty} b_n.

      Rests on: C (Sequences), B (Calculus I and Algebra and Geometry). Proved in: Numerical series and their convergence.

    3. Necessary condition for convergence. If the series ∑n=1∞an\sum_{n=1}^{\infty} a_n converges, then an→0a_n \to 0; if (an)(a_n) does not tend to 00, the series diverges.

      Rests on: C (Sequences). Proved in: Numerical series and their convergence.

    4. The harmonic series diverges. For every integer m≥0m \geq 0, H2m≥1+m2H_{2^m} \geq 1 + \frac{m}{2}; therefore, ∑n=1∞1n=∞\sum_{n=1}^{\infty} \frac{1}{n} = \infty, although its term tends to 00.

      Rests on: C (Sequences), B (Calculus I and Algebra and Geometry). Proved in: Numerical series and their convergence.

    5. Cauchy criterion for series. The series ∑n=1∞an\sum_{n=1}^{\infty} a_n converges if and only if for every ε>0\varepsilon > 0 there is a natural number NN such that ∣∑k=n+1mak∣<ε\abs{\sum_{k=n + 1}^{m} a_k} < \varepsilon for all natural numbers m>n≥Nm > n \geq N.

      Rests on: C (Sequences), A (Completeness axiom), B (Calculus I and Algebra and Geometry). Proved in: Numerical series and their convergence.

    6. Linearity and monotonicity of series. If ∑n=1∞an=S\sum_{n=1}^{\infty} a_n = S and ∑n=1∞bn=T\sum_{n=1}^{\infty} b_n = T, then ∑n=1∞(αan+βbn)=αS+βT\sum_{n=1}^{\infty} \left(\alpha a_n + \beta b_n\right) = \alpha S + \beta T, and if an≤bna_n \leq b_n for every nn, S≤TS \leq T; a convergent series plus a divergent one diverges.

      Rests on: C (Sequences), B (Calculus I and Algebra and Geometry). Proved in: Numerical series and their convergence.

    7. The tail of a series. A series converges if and only if its tail from any index converges, and then its remainder RnR_n tends to 00; two series whose terms coincide from some index on converge or diverge together.

      Rests on: C (Sequences). Proved in: Numerical series and their convergence.

    8. Grouping the terms of a convergent series. Every grouped series of a convergent series converges, with the same sum; if the series diverges to ∞\infty or to −∞-\infty, so does the grouped one.

      Rests on: C (Sequences). Proved in: Numerical series and their convergence.

    9. Ungrouping the terms of a series. If a grouped series converges, the series converges with the same sum, provided that the terms tend to 00 and the groups have bounded length, or that the terms of each group have the same sign.

      Rests on: C (Sequences). Proved in: Numerical series and their convergence.

  2. Segment 2 Series with nonnegative terms

    • Reduce the convergence of a series with nonnegative terms to the boundedness of its partial sums, with the completeness axiom, and deduce the comparison tests.
    • Prove the integral test and decide the series of reciprocals of powers.
    • Define the Euler-Mascheroni constant and prove the Cauchy condensation test.
    1. Series with nonnegative terms. If an≥0a_n \geq 0 for every nn, the series converges if and only if its partial sums are bounded, and its sum is their supremum; otherwise, it diverges to ∞\infty. Every grouped series has the same nature and the same sum.

      Uses: 8. Grouping the terms of a convergent series, 9. Ungrouping the terms of a series. Rests on: A (Completeness axiom), C (Sequences). Proved in: Series with positive terms.

    2. Comparison test for series. If 0≤an≤bn0 \leq a_n \leq b_n from some index on and ∑n=1∞bn\sum_{n=1}^{\infty} b_n converges, then ∑n=1∞an\sum_{n=1}^{\infty} a_n converges; if ∑n=1∞an\sum_{n=1}^{\infty} a_n diverges, ∑n=1∞bn=∞\sum_{n=1}^{\infty} b_n = \infty.

      Uses: 10. Series with nonnegative terms, 7. The tail of a series. Rests on: C (Sequences). Proved in: Series with positive terms.

    3. Limit comparison test for series. If an≥0a_n \geq 0, bn>0b_n > 0 and anbn→L\frac{a_n}{b_n} \to L: with 0<L<∞0 < L < \infty, the two series converge or diverge together; with L=0L = 0, the convergence of the series of the bnb_n gives that of the series of the ana_n; with L=∞L = \infty, the divergence of the series of the bnb_n gives that of the series of the ana_n.

      Uses: 6. Linearity and monotonicity of series. Rests on: G (Comparison), C (Sequences). Proved in: Series with positive terms.

    4. Integral test. If ff is nonnegative and decreasing on [m,∞)\ico{m}{\infty} and an=f(n)a_n = f(n), the series ∑n=m∞an\sum_{n=m}^{\infty} a_n converges if and only if ∫m∞f\int_m^{\infty} f converges, and then ∫N+1∞f≤∑n=N+1∞an≤∫N∞f\int_{N + 1}^{\infty} f \leq \sum_{n=N + 1}^{\infty} a_n \leq \int_N^{\infty} f for every natural number N≥mN \geq m.

      Uses: 10. Series with nonnegative terms, 7. The tail of a series, 6. Linearity and monotonicity of series. Rests on: D (Integrals), C (Sequences). Proved in: Series with positive terms.

    5. The series of reciprocals of powers. The series ∑n=1∞1np\sum_{n=1}^{\infty} \frac{1}{n^p} converges if and only if p>1p > 1; if p>1p > 1, its remainder satisfies 1(p−1) (N+1)p−1≤RN≤1(p−1) Np−1\frac{1}{(p - 1) \, (N + 1)^{p - 1}} \leq R_N \leq \frac{1}{(p - 1) \, N^{p - 1}}.

      Uses: 3. Necessary condition for convergence, 13. Integral test. Rests on: D (Integrals), B (Calculus I and Algebra and Geometry). Proved in: Series with positive terms.

    6. The difference between the series and the integral. Under the hypotheses of the integral test, Dn=∑k=mnak−∫mnfD_n = \sum_{k=m}^{n} a_k - \int_m^n f decreases, satisfies an≤Dn≤ama_n \leq D_n \leq a_m and converges to some ℓ\ell with 0≤Dn−ℓ≤an0 \leq D_n - \ell \leq a_n.

      Uses: 13. Integral test. Rests on: A (Completeness axiom), C (Sequences), B (Calculus I and Algebra and Geometry). Proved in: Series with positive terms.

    7. The Euler-Mascheroni constant. The sequence Hn−ln⁡nH_n - \ln n is strictly decreasing and converges to γ\gamma, with 12(n+1)<Hn−ln⁡n−γ<12n\frac{1}{2(n + 1)} < H_n - \ln n - \gamma < \frac{1}{2n}; in particular, 12<γ<34\frac{1}{2} < \gamma < \frac{3}{4}.

      Uses: 15. The difference between the series and the integral. Rests on: D (Integrals), B (Calculus I and Algebra and Geometry). Proved in: Series with positive terms.

    8. The harmonic sums from n + 1 to qn. For every natural number qq, Hqn−Hn→ln⁡qH_{qn} - H_n \to \ln q.

      Uses: 16. The Euler-Mascheroni constant. Rests on: C (Sequences). Proved in: Series with positive terms.

    9. Cauchy condensation test. If (an)(a_n) is decreasing and nonnegative, ∑n=1∞an\sum_{n=1}^{\infty} a_n converges if and only if ∑n=1∞2na2n\sum_{n=1}^{\infty} 2^n a_{2^n} converges, and the first sum lies between half of the second and a1a_1 plus the second.

      Uses: 10. Series with nonnegative terms. Rests on: B (Calculus I and Algebra and Geometry). Proved in: Series with positive terms.

    10. Series with a logarithm in the denominator. The series ∑n=2∞1n (ln⁡n)p\sum_{n=2}^{\infty} \frac{1}{n \, (\ln n)^p} converges if and only if p>1p > 1.

      Uses: 18. Cauchy condensation test, 14. The series of reciprocals of powers, 4. The harmonic series diverges, 7. The tail of a series, 6. Linearity and monotonicity of series. Rests on: G (Comparison), B (Calculus I and Algebra and Geometry). Proved in: Series with positive terms.

  3. Segment 3 The ratio and root tests

    • Translate the limits superior and inferior into inequalities from some index on, and bound a series by a geometric one.
    • Prove the root and ratio tests and the inequality between the ratios and the roots.
    • Extend the tests to terms of arbitrary sign.
    1. Bounds from some index on. If lim sup⁡n→∞xn<c\limsup_{n \to \infty} x_n < c, then xn<cx_n < c from some index on; if lim sup⁡n→∞xn>c\limsup_{n \to \infty} x_n > c, infinitely many terms exceed cc; and analogously for the limit inferior.

      Rests on: C (Sequences), A (Completeness axiom). Proved in: Ratio and root tests.

    2. Geometric majorant. If 0≤ak≤Crk0 \leq a_k \leq C r^k for every k>Nk > N, with 0≤r<10 \leq r < 1, the series converges and 0≤Rn≤Crn+11−r0 \leq R_n \leq \frac{C r^{n + 1}}{1 - r} for n≥Nn \geq N.

      Rests on: E (Geometric series), C (Sequences). Proved in: Ratio and root tests.

    3. Root test. If an≥0a_n \geq 0 and ρ=lim sup⁡n→∞ann\rho = \limsup_{n \to \infty} \sqrt[n]{a_n}: with ρ<1\rho < 1 the series converges; with ρ>1\rho > 1 it diverges to infinity; with ρ=1\rho = 1 the test is inconclusive.

      Uses: 20. Bounds from some index on, 21. Geometric majorant, 3. Necessary condition for convergence, 4. The harmonic series diverges, 2. Telescoping series. Rests on: B (Calculus I and Algebra and Geometry), C (Sequences). Proved in: Ratio and root tests.

    4. Ratio test. If an>0a_n > 0: with lim sup⁡n→∞an+1an<1\limsup_{n \to \infty} \frac{a_{n + 1}}{a_n} < 1 the series converges; if an+1an≥1\frac{a_{n + 1}}{a_n} \geq 1 from some index on, it diverges to infinity; if the ratios tend to 11 and are less than 11 for every nn, the test is inconclusive.

      Uses: 20. Bounds from some index on, 21. Geometric majorant, 22. Root test, 3. Necessary condition for convergence. Rests on: B (Calculus I and Algebra and Geometry), C (Sequences). Proved in: Ratio and root tests.

    5. Inequality between the ratios and the roots. If an>0a_n > 0, then lim inf⁡n→∞an+1an≤lim inf⁡n→∞ann≤lim sup⁡n→∞ann≤lim sup⁡n→∞an+1an\liminf_{n \to \infty} \frac{a_{n + 1}}{a_n} \leq \liminf_{n \to \infty} \sqrt[n]{a_n} \leq \limsup_{n \to \infty} \sqrt[n]{a_n} \leq \limsup_{n \to \infty} \frac{a_{n + 1}}{a_n}.

      Uses: 20. Bounds from some index on. Rests on: B (Calculus I and Algebra and Geometry), C (Sequences). Proved in: Ratio and root tests.

    6. The root test decides whenever the ratio test decides. If lim sup⁡n→∞an+1an<1\limsup_{n \to \infty} \frac{a_{n + 1}}{a_n} < 1, then also lim sup⁡n→∞ann<1\limsup_{n \to \infty} \sqrt[n]{a_n} < 1; if lim inf⁡n→∞an+1an>1\liminf_{n \to \infty} \frac{a_{n + 1}}{a_n} > 1, then also lim sup⁡n→∞ann>1\limsup_{n \to \infty} \sqrt[n]{a_n} > 1.

      Uses: 24. Inequality between the ratios and the roots. Proved in: Ratio and root tests.

    7. Geometric rate of the general term. If lim sup⁡n→∞ann=ρ<1\limsup_{n \to \infty} \sqrt[n]{a_n} = \rho < 1, for every rr with ρ<r<1\rho < r < 1 one has an<rna_n < r^n from some index on, and npan→0n^p a_n \to 0 for every real pp.

      Uses: 22. Root test, 20. Bounds from some index on, 25. The root test decides whenever the ratio test decides. Rests on: C (Sequences), B (Calculus I and Algebra and Geometry). Proved in: Ratio and root tests.

    8. The tests for terms of arbitrary sign. If lim sup⁡n→∞∣an∣n<1\limsup_{n \to \infty} \sqrt[n]{\abs{a_n}} < 1, or the terms do not vanish and the limit superior of the ratios of the absolute values is less than 11, the series of the ∣an∣\abs{a_n} and of the ana_n converge; if lim sup⁡n→∞∣an∣n>1\limsup_{n \to \infty} \sqrt[n]{\abs{a_n}} > 1, or the terms do not vanish and ∣an+1∣≥∣an∣\abs{a_{n + 1}} \geq \abs{a_n} from some index on, the series diverges.

      Uses: 22. Root test, 23. Ratio test, 3. Necessary condition for convergence. Rests on: F (Cauchy criterion), B (Calculus I and Algebra and Geometry). Proved in: Ratio and root tests.

  4. Segment 4 The finer tests

    • Prove Kummer’s test and deduce from it Raabe’s and Bertrand’s tests.
    • Prove Gauss’s test and apply it to ratios of polynomials.
    • Prove the logarithmic test, Ermakov’s test and the scale of iterated logarithms.
    1. Comparison of ratios. If an,bn>0a_n, b_n > 0 and an+1an≤bn+1bn\frac{a_{n + 1}}{a_n} \leq \frac{b_{n + 1}}{b_n} from NN on, then an≤aNbN bna_n \leq \frac{a_N}{b_N} \, b_n from NN on, and the convergence of the series of the bnb_n gives that of the series of the ana_n.

      Uses: 6. Linearity and monotonicity of series. Rests on: G (Comparison). Proved in: Finer convergence tests.

    2. Kummer’s test. If an,cn>0a_n, c_n > 0 and Kn=cnanan+1−cn+1K_n = c_n \frac{a_n}{a_{n + 1}} - c_{n + 1}: if Kn≥h>0K_n \geq h > 0 from some index on, the series converges; if Kn≤0K_n \leq 0 from some index on and ∑n=1∞1cn\sum_{n=1}^{\infty} \frac{1}{c_n} diverges, the series diverges.

      Uses: 10. Series with nonnegative terms, 7. The tail of a series, 6. Linearity and monotonicity of series. Rests on: G (Comparison). Proved in: Finer convergence tests.

    3. Raabe’s test. If n(anan+1−1)≥q>1n \left(\frac{a_n}{a_{n + 1}} - 1\right) \geq q > 1 from some index on, the series converges; if it does not exceed 11 from some index on, it diverges; if it tends to LL, it converges with L>1L > 1 or L=∞L = \infty and diverges with L<1L < 1.

      Uses: 29. Kummer’s test, 4. The harmonic series diverges. Rests on: C (Sequences). Proved in: Finer convergence tests.

    4. Duhamel’s form of Raabe’s test. If an+1an=1−βn+o⁡(1n)\frac{a_{n + 1}}{a_n} = 1 - \frac{\beta}{n} + \operatorname{o}\left(\frac{1}{n}\right), with the notation of the definition of comparison of sequences, the series converges if β>1\beta > 1 and diverges if β<1\beta < 1.

      Uses: 30. Raabe’s test. Rests on: C (Sequences), B (Calculus I and Algebra and Geometry). Proved in: Finer convergence tests.

    5. Bertrand’s test. With Bn=ln⁡n(n(anan+1−1)−1)B_n = \ln n \left(n \left(\frac{a_n}{a_{n + 1}} - 1\right) - 1\right): if Bn≥q>1B_n \geq q > 1 from some index on, the series converges; if BnB_n does not exceed 11 from some index on, it diverges.

      Uses: 29. Kummer’s test, 30. Raabe’s test, 19. Series with a logarithm in the denominator, 7. The tail of a series. Rests on: B (Calculus I and Algebra and Geometry). Proved in: Finer convergence tests.

    6. The Raabe and Bertrand quantities with a logarithm. For an=1n (ln⁡n)pa_n = \frac{1}{n \, (\ln n)^p}, the Raabe quantity tends to 11 and Bn→pB_n \to p.

      Uses: 32. Bertrand’s test. Rests on: C (Sequences), B (Calculus I and Algebra and Geometry). Proved in: Finer convergence tests.

    7. Gauss’s test. If anan+1=1+μn+O⁡(1n1+ε)\frac{a_n}{a_{n + 1}} = 1 + \frac{\mu}{n} + \operatorname{O}\left(\frac{1}{n^{1 + \varepsilon}}\right) with ε>0\varepsilon > 0, the series converges if and only if μ>1\mu > 1.

      Uses: 30. Raabe’s test, 32. Bertrand’s test. Rests on: C (Sequences), B (Calculus I and Algebra and Geometry). Proved in: Finer convergence tests.

    8. Ratios of polynomials. If anan+1=P(n)Q(n)\frac{a_n}{a_{n + 1}} = \frac{P(n)}{Q(n)} from some index on, with P(x)=xk+p1xk−1+⋯P(x) = x^k + p_1 x^{k - 1} + \cdots and Q(x)=xk+q1xk−1+⋯Q(x) = x^k + q_1 x^{k - 1} + \cdots, the series converges if and only if p1−q1>1p_1 - q_1 > 1.

      Uses: 34. Gauss’s test. Rests on: C (Sequences), B (Calculus I and Algebra and Geometry). Proved in: Finer convergence tests.

    9. Logarithmic test. If ln⁡1anln⁡n≥q>1\frac{\ln \frac{1}{a_n}}{\ln n} \geq q > 1 from some index on, the series converges, and if it does not exceed 11, it diverges; the same holds with ln⁡1nanln⁡ln⁡n\frac{\ln \frac{1}{n a_n}}{\ln \ln n}.

      Uses: 14. The series of reciprocals of powers, 19. Series with a logarithm in the denominator, 7. The tail of a series. Rests on: G (Comparison), B (Calculus I and Algebra and Geometry). Proved in: Finer convergence tests.

    10. Ermakov’s test. If ff is continuous, positive and decreasing and an=f(n)a_n = f(n): if exf(ex)≤qf(x)e^x f(e^x) \leq q f(x) from some point on, with q<1q < 1, the series converges; if exf(ex)≥f(x)e^x f(e^x) \geq f(x) from some point on, it diverges.

      Uses: 13. Integral test, 6. Linearity and monotonicity of series. Rests on: D (Integrals), B (Calculus I and Algebra and Geometry). Proved in: Finer convergence tests.

    11. The scale of iterated logarithms. For each natural number kk, the series of the 1n L1(n)⋯Lk−1(n) (Lk(n))p\frac{1}{n \, L_1(n) \cdots L_{k - 1}(n) \, \left(L_k(n)\right)^p}, where LkL_k is the logarithm iterated kk times, converges if and only if p>1p > 1.

      Uses: 37. Ermakov’s test. Rests on: G (Comparison), B (Calculus I and Algebra and Geometry). Proved in: Finer convergence tests.

  5. Segment 5 Conditional convergence

    • Prove that absolute convergence implies convergence, and the Leibniz test with the remainder bound.
    • Prove Abel’s summation by parts and, with it, Dirichlet’s and Abel’s tests for series.
    • Prove the convergence of the series of the sines and cosines of the multiples divided by their index.
    1. An absolutely convergent series converges. If ∑n=1∞∣an∣\sum_{n=1}^{\infty} \abs{a_n} converges, then ∑n=1∞an\sum_{n=1}^{\infty} a_n also converges, and ∣∑n=1∞an∣≤∑n=1∞∣an∣\abs{\sum_{n=1}^{\infty} a_n} \leq \sum_{n=1}^{\infty} \abs{a_n}.

      Rests on: F (Cauchy criterion), C (Sequences), B (Calculus I and Algebra and Geometry). Proved in: Alternating series and conditional convergence.

    2. Leibniz test. If (an)(a_n) is decreasing and an→0a_n \to 0, the series ∑n=1∞(−1)n+1an\sum_{n=1}^{\infty} (-1)^{n + 1} a_n converges, and 0≤(−1)nRn≤an+10 \leq (-1)^n R_n \leq a_{n + 1} for every integer n≥0n \geq 0.

      Rests on: C (Sequences), B (Calculus I and Algebra and Geometry). Proved in: Alternating series and conditional convergence.

    3. The alternating harmonic series converges conditionally. The series ∑n=1∞(−1)n+1n\sum_{n=1}^{\infty} \frac{(-1)^{n + 1}}{n} converges conditionally, its sum lies between 12\frac{1}{2} and 56\frac{5}{6}, and ∣S−Sn∣<1n+1\abs{S - S_n} < \frac{1}{n + 1}.

      Uses: 40. Leibniz test, 4. The harmonic series diverges. Rests on: B (Calculus I and Algebra and Geometry). Proved in: Alternating series and conditional convergence.

    4. Abel summation by parts. If Ak−Ak−1=akA_k - A_{k - 1} = a_k for every natural number kk, then, for m>n≥0m > n \geq 0, ∑k=n+1makbk=Ambm−Anbn+1−∑k=n+1m−1Ak (bk+1−bk)\sum_{k=n + 1}^{m} a_k b_k = A_m b_m - A_n b_{n + 1} - \sum_{k=n + 1}^{m - 1} A_k \, (b_{k + 1} - b_k).

      Rests on: B (Calculus I and Algebra and Geometry). Proved in: Alternating series and conditional convergence.

    5. Dirichlet’s test for series. If (bn)(b_n) is monotonic with bn→0b_n \to 0 and the stretches of the aka_k have sum bounded by KK, the series ∑n=1∞anbn\sum_{n=1}^{\infty} a_n b_n converges, and its tail from n+1n + 1 does not exceed K∣bn+1∣K \abs{b_{n + 1}} in absolute value.

      Uses: 7. The tail of a series. Rests on: H (Summation by parts), F (Cauchy criterion), B (Calculus I and Algebra and Geometry). Proved in: Alternating series and conditional convergence.

    6. Abel’s test for series. If ∑n=1∞an\sum_{n=1}^{\infty} a_n converges and (bn)(b_n) is monotonic and bounded, the series ∑n=1∞anbn\sum_{n=1}^{\infty} a_n b_n converges.

      Rests on: H (Summation by parts), F (Cauchy criterion). Proved in: Alternating series and conditional convergence.

    7. Sums of sines and cosines. If xx is not a multiple of 2π2\pi, the sums of sin⁡(kx)\sin(kx) and of cos⁡(kx)\cos(kx) over any stretch of indices do not exceed 1∣sin⁡x2∣\frac{1}{\abs{\sin \frac{x}{2}}} in absolute value.

      Rests on: B (Calculus I and Algebra and Geometry). Proved in: Alternating series and conditional convergence.

    8. The series of sines and cosines of multiples. If xx is not a multiple of 2π2\pi, the series of sin⁡(nx)n\frac{\sin(nx)}{n} and of cos⁡(nx)n\frac{\cos(nx)}{n} converge, and their tails from n+1n + 1 do not exceed 1(n+1)∣sin⁡x2∣\frac{1}{(n + 1) \abs{\sin \frac{x}{2}}}.

      Uses: 45. Sums of sines and cosines, 43. Dirichlet’s test for series. Proved in: Alternating series and conditional convergence.

  6. Segment 6 Rearrangements and products

    • Prove that an absolutely convergent series can be rearranged without changing its sum, and sum the alternating harmonic series.
    • Decompose a series into its positive and negative parts and prove the Riemann rearrangement theorem.
    • Define the Cauchy product and prove Mertens’s theorem.
    1. Rearranging a sequence does not change its limit. If xn→Lx_n \to L and σ\sigma is a bijection from N\mathbb{N} to N\mathbb{N}, then xσ(n)→Lx_{\sigma(n)} \to L.

      Rests on: C (Sequences). Proved in: Absolute convergence, rearrangements and products of series.

    2. Dirichlet’s theorem on rearrangements. If ∑n=1∞an\sum_{n=1}^{\infty} a_n converges absolutely, every rearrangement of it converges absolutely, with the same sum.

      Uses: 47. Rearranging a sequence does not change its limit. Rests on: F (Cauchy criterion), C (Sequences), B (Calculus I and Algebra and Geometry). Proved in: Absolute convergence, rearrangements and products of series.

    3. The sum of the alternating harmonic series. One has ∑n=1∞(−1)n+1n=ln⁡2\sum_{n=1}^{\infty} \frac{(-1)^{n + 1}}{n} = \ln 2: its partial sums of even index are H2n−HnH_{2n} - H_n.

      Uses: 17. The harmonic sums from n + 1 to qn, 41. The alternating harmonic series converges conditionally. Rests on: C (Sequences). Proved in: Absolute convergence, rearrangements and products of series.

    4. The positive and negative parts of a series. With pn=max⁡(an,0)p_n = \max(a_n, 0) and qn=max⁡(−an,0)q_n = \max(-a_n, 0), the series converges absolutely if and only if the series of the pnp_n and of the qnq_n converge; if it converges conditionally, both diverge to ∞\infty.

      Uses: 6. Linearity and monotonicity of series, 10. Series with nonnegative terms. Rests on: G (Comparison). Proved in: Absolute convergence, rearrangements and products of series.

    5. The terms of each sign of a conditionally convergent series. In a conditionally convergent series there are infinitely many terms of each sign; taken in their order, they tend to 00, the nonnegative ones sum to ∞\infty and the negative ones to −∞-\infty.

      Uses: 50. The positive and negative parts of a series, 7. The tail of a series, 3. Necessary condition for convergence. Rests on: C (Sequences), B (Calculus I and Algebra and Geometry). Proved in: Absolute convergence, rearrangements and products of series.

    6. Riemann rearrangement theorem. If ∑n=1∞an\sum_{n=1}^{\infty} a_n converges conditionally and LL is a real number, ∞\infty or −∞-\infty, there is a bijection σ\sigma from N\mathbb{N} to N\mathbb{N} with ∑n=1∞aσ(n)=L\sum_{n=1}^{\infty} a_{\sigma(n)} = L.

      Uses: 51. The terms of each sign of a conditionally convergent series, 7. The tail of a series, 6. Linearity and monotonicity of series. Rests on: B (Calculus I and Algebra and Geometry). Proved in: Absolute convergence, rearrangements and products of series.

    7. The partial sums of the Cauchy product. With cn=∑k=0nakbn−kc_n = \sum_{k=0}^{n} a_k b_{n - k}, the partial sums of the product are Cn=∑k=0nakBn−kC_n = \sum_{k=0}^{n} a_k B_{n - k}.

      Rests on: B (Calculus I and Algebra and Geometry). Proved in: Absolute convergence, rearrangements and products of series.

    8. Mertens’s theorem. If ∑n=0∞an\sum_{n=0}^{\infty} a_n converges absolutely, with sum AA, and ∑n=0∞bn\sum_{n=0}^{\infty} b_n converges, with sum BB, their Cauchy product converges, with sum ABAB.

      Uses: 53. The partial sums of the Cauchy product. Rests on: F (Cauchy criterion), C (Sequences), B (Calculus I and Algebra and Geometry). Proved in: Absolute convergence, rearrangements and products of series.

    9. The product of two absolutely convergent series. If both series converge absolutely, their Cauchy product converges absolutely, with sum ABAB.

      Uses: 54. Mertens’s theorem. Rests on: G (Comparison). Proved in: Absolute convergence, rearrangements and products of series.

    10. A divergent Cauchy product. With an=bn=(−1)nn+1a_n = b_n = \frac{(-1)^n}{\sqrt{n + 1}}, both series converge conditionally and their Cauchy product diverges: ∣cn∣≥1\abs{c_n} \geq 1 for every nn.

      Uses: 40. Leibniz test, 14. The series of reciprocals of powers, 3. Necessary condition for convergence. Rests on: B (Calculus I and Algebra and Geometry). Proved in: Absolute convergence, rearrangements and products of series.

Toward Taylor polynomials and power series. The chain stops where the terms begin to depend on a variable. The unit that follows studies the series whose terms are powers of xx multiplied by coefficients: for each value of xx, a power series is a numerical series, and the lesson on power series will decide, with the ratio and root tests of this unit applied to the absolute values of its terms, for which values it converges; where it converges absolutely, Dirichlet’s theorem on rearrangements and Mertens’s theorem allow it to be rearranged and multiplied. A new question appears with the variable: whether the sum of a series of continuous functions is continuous, and whether one can integrate or differentiate term by term; the convergence of each numerical series is not enough to answer it, and the lesson on sequences and series of functions will settle it with uniform convergence.