Skip to content
Tutorials Center
My learning

Topics

Lesson 18 of 37 · Series and Convergence

Rearrangements and the Cauchy Condensation Test

See why rearranging a conditionally convergent series can change its sum, and use the Cauchy condensation test to decide series like 1/(k ln k).

Addition of finitely many numbers is commutative: order does not matter. For infinite sums that is not automatically true. This lesson shows exactly when order matters, with a spectacular result: a conditionally convergent series can be rearranged to add up to any number. It then finishes the chapter with a sharp test, Cauchy condensation, for delicate series like Σ 1/(k ln k).

Rearrangements

A rearrangement of Σ a_k is a series Σ a_{σ(k)} where σ: ℕ → ℕ is a bijection: the same terms in a different order.

Theorem 1. If Σ a_k converges absolutely, every rearrangement converges to the same sum.

Sketch. Let S be the sum. Given ε, choose N with Σ_{k>N} |a_k| < ε. Pick M so the rearranged series' first M terms include a_1, ..., a_N. Beyond that, the difference between the rearranged partial sum and S involves only terms with index > N, so it is at most ε in absolute value (twice, to include both tails). ∎

Theorem 2 (Riemann rearrangement). If Σ a_k is conditionally convergent, then for every real number T (and also ±∞) there is a rearrangement converging to T.

Idea. A conditionally convergent series has positive terms summing to +∞ and negative terms summing to −∞ (otherwise it would converge absolutely), while a_k → 0. To hit the target T: add positive terms in order until the running total exceeds T, then add negative terms until it drops below T, and repeat. Both kinds of terms are used forever, and since terms shrink to 0 the overshoot shrinks to 0, so the partial sums converge to T.

Python Rearranging the alternating harmonic series

Press Run (or Ctrl + Enter) to see the result here.

Program input · what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

Each target is approached, although the original order gives ln 2 ≈ 0.6931. Notice the ratio of positive to negative terms used: it differs by target, which is how the sum shifts.

A concrete classical rearrangement: take two positive terms for each negative one, 1 + 1/3 − 1/2 + 1/5 + 1/7 − 1/4 + .... This sums to (3/2) ln 2, which is not ln 2.

Python A 2-positive-to-1-negative rearrangement

Press Run (or Ctrl + Enter) to see the result here.

Program input · what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

Never reorder the terms of a series unless you have proved absolute convergence. Computer programs summing floating-point series in different orders get different answers for the same reason, and many numerical libraries sum the smallest terms first.

The Cauchy condensation test

The ratio and root tests fail on series like Σ 1/k and Σ 1/(k ln k). A sharper tool applies to decreasing nonnegative terms.

Theorem (Cauchy condensation). Let a_k ≥ 0 be decreasing. Then Σ_{k=1}^∞ a_k converges if and only if the condensed series Σ_{j=0}^∞ 2^j a_{2^j} converges.

Proof idea. Group the terms into blocks [2^j, 2^{j+1} − 1]. Because a_k is decreasing, the block has 2^j terms, each between a_{2^{j+1}} and a_{2^j}. So

(1/2) · 2^{j+1} a_{2^{j+1}} ≤ (block j) ≤ 2^j a_{2^j}.

Summing over j shows the two series are bounded by constant multiples of each other, so they converge or diverge together. ∎

Applications.

  • Σ 1/k^p: condensed terms 2^j / 2^{jp} = (2^{1−p})^j, a geometric series that converges iff p > 1. This re-proves the p-series test.
  • Σ 1/(k ln k) (from k = 2): condensed terms 2^j / (2^j · j ln 2) = 1/(j ln 2), a harmonic series, so it diverges.
  • Σ 1/(k (ln k)²): condensed terms 1/(j² (ln 2)²), convergent, so it converges.
Python Condensed series for 1/(k ln k)

Press Run (or Ctrl + Enter) to see the result here.

Program input · what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

The sums grow roughly like ln(ln n), so astronomically many terms are needed to pass even modest values.

flowchart LR
  A["Decreasing a_k ≥ 0"] --> B["Σ a_k"]
  A --> C["Σ 2^j a_(2^j)"]
  B <--> C

Cauchy condensation: the two series converge or diverge together.

Check your understanding

Quick check · 4 questions

of 4 answered You scored / 4 · perfect!

  1. 1. If Σ a_k converges absolutely, what happens under rearrangement?

    Absolutely convergent series are insensitive to the order of terms.
  2. 2. What does the Riemann rearrangement theorem say about conditionally convergent series?

    Positive terms and negative terms each sum to infinity, so you can steer the partial sums to any target.
  3. 3. Cauchy condensation applies to

    Monotonicity is what allows comparing each block with its first or last term.
  4. 4. Does Σ 1/(k ln k) converge?

    Condensing gives Σ 1/(j ln 2), a harmonic series, which diverges.

Summary

  • Rearranging an absolutely convergent series never changes its sum.
  • Rearranging a conditionally convergent series can give any prescribed sum (Riemann).
  • Cauchy condensation: for decreasing a_k ≥ 0, Σ a_k and Σ 2^j a_{2^j} converge together.
  • It settles borderline cases: Σ 1/(k ln k) diverges and Σ 1/(k (ln k)²) converges.
  • You have now seen limits of sequences and series; the next topic is continuity of functions.

Next: Epsilon-Delta Continuity