JOVANA
Library Glossary Getting Started Three Levels Fields How it works Mission
Join the mission
All guides

Comparison, p-Series, and the Integral Test

For series of non-negative terms, convergence is bounded partial sums. We exploit that with the comparison and limit-comparison tests, settle the p-series once and for all with the integral test, and learn which standard yardstick to compare against.

Positive terms: bounded equals convergent

When every a_n >= 0, the partial sums s_N can only increase: s_{N+1} = s_N + a_{N+1} >= s_N. A monotone increasing sequence converges if and only if it is bounded above, by the monotone convergence theorem. So for non-negative series, convergence is exactly boundedness of the partial sums. This is what makes every comparison test work.

The comparison test then reads: if 0 <= a_n <= b_n for all large n and sum b_n converges, so does sum a_n (its partial sums are trapped below a bound). And if a_n >= b_n >= 0 with sum b_n divergent, then sum a_n diverges too. Squeeze from above to prove convergence; push from below to prove divergence.

When the inequality is awkward to set up by hand, the limit comparison test is friendlier: if a_n, b_n > 0 and a_n/b_n -> L with 0 < L < infinity, then sum a_n and sum b_n converge or diverge together. You only need to know how a_n behaves to leading order.

The integral test settles the p-series

The standard yardstick is the p-series sum 1/n^p. To decide it cleanly we use the integral test: if f is positive, decreasing, and continuous with f(n) = a_n, then sum a_n and the improper integral of f from 1 to infinity converge or diverge together. The picture is rectangles trapped between the curve, comparing a sum to an area.

p-series sum_{n=1}^∞ 1/n^p, decided by the integral test.

Let f(x) = 1/x^p on [1, infinity): positive, continuous, decreasing
for p > 0. Compare the series to integral_1^infinity x^(-p) dx.

Case p != 1:
   integral_1^R x^(-p) dx = [ x^(1-p)/(1-p) ]_1^R
                          = (R^(1-p) - 1)/(1 - p).
   As R -> infinity:
      if p > 1, then 1-p < 0, so R^(1-p) -> 0  => integral = 1/(p-1)  (finite) => CONVERGES
      if p < 1, then 1-p > 0, so R^(1-p) -> infinity                 => DIVERGES

Case p = 1:
   integral_1^R (1/x) dx = ln R -> infinity                          => DIVERGES
   (this recovers the harmonic series)

Conclusion:  sum 1/n^p converges  <=>  p > 1.
One improper integral decides every p-series at once: convergence exactly when p > 1.

Putting comparison to work

  1. Look at the leading order of a_n for large n — strip away constants and lower-order terms.
  2. Guess a p-series 1/n^p (or geometric r^n) that matches that leading order.
  3. Confirm with the limit comparison test: compute a_n / (1/n^p) -> L and check 0 < L < infinity.
  4. Read off the answer from the known behavior of the yardstick series.
Test sum (2n + 1)/(n^3 + 5).

Leading order: numerator ~ 2n, denominator ~ n^3,
so a_n ~ 2n / n^3 = 2/n^2. Compare with b_n = 1/n^2 (p = 2 > 1).

Limit comparison:
   a_n / b_n = [ (2n+1)/(n^3+5) ] * n^2
             = (2n^3 + n^2) / (n^3 + 5)
             -> 2   as n -> infinity   (0 < 2 < infinity).

Since sum 1/n^2 converges (p = 2 > 1),
by limit comparison sum (2n+1)/(n^3+5) CONVERGES.
Strip to leading order 2/n^2, then limit-compare against the convergent p = 2 series.