Master techniques to determine convergence of improper integrals without explicit evaluation: comparison tests, limit comparison, Cauchy criterion, and specialized tests for logarithmic and exponential integrands.
Many improper integrals cannot be evaluated explicitly using elementary functions. For example:
No Elementary Antiderivative
Cannot find F(x) in closed form
Complex Expression
Explicit evaluation is very difficult
Convergence tests allow us to determine whether an integral converges or diverges without computing its exact value. This is analogous to convergence tests for infinite series.
The main approaches for testing convergence:
Let for all . Then:
Given: for
To Show: Convergence of implies convergence of
Proof:
Define and .
Since , is increasing in .
Since , we have for all .
If converges to , then .
An increasing function bounded above has a limit, so exists.
Therefore converges. ∎
The comparison test requires (non-negative integrands). For oscillating functions, use the absolute convergence test or Dirichlet/Abel tests (covered in the next section).
Problem: Show converges.
Solution:
For :
Since converges (p = 2 > 1), by comparison test:
converges.
Problem: Show diverges.
Solution:
For :
Since diverges (p = 1), by comparison test:
diverges.
Let for all where both have a singularity at . Then:
Problem: Show converges.
Solution:
Near , there's a singularity. For :
Since converges (p = 1/2 < 1), by comparison:
converges.
Let and for . If
where (finite and positive), then:
and either both converge or both diverge.
Given: with
To Show: Both integrals have the same convergence behavior
Proof:
Choose . There exists such that for :
This gives:
By direct comparison test applied in both directions, and have the same behavior. ∎
Problem: Determine convergence of
Solution:
As , the integrand behaves like:
Check the limit:
Since converges (p = 2 > 1), the original integral converges.
Let and near (singularity). If
where , then and have the same convergence behavior.
Problem: Determine convergence of
Solution:
Near : , so:
Verify the limit:
Since converges (p = 1/2 < 1), the original integral converges.
Problem: Analyze
Solution:
Factor:
Singularities at and . Split at .
Near x = 0: → p = 1, diverges
The integral diverges (no need to check x = 1).
The improper integral converges if and only if:
In words: the integral over any sufficiently far interval can be made arbitrarily small.
(⇒) If converges to , then .
For : by Cauchy criterion for limits.
(⇐) If the Cauchy condition holds, then satisfies Cauchy criterion, so the limit exists.
∎
The Cauchy criterion is useful when:
If converges, then:
Warning: This is necessary but NOT sufficient. The integrand need not go to 0.
Problem: Show converges using Cauchy criterion.
Solution:
For :
Given , choose . Then for :
By Cauchy criterion, the integral converges.
If for and converges, then converges absolutely.
The integral with logarithmic factors:
Substitute , so :
This is the standard p-integral, which converges iff . ∎
For the integral near 1 (where ):
Key fact: as for any and .
This means logarithmic factors don't change the convergence threshold of the p-test; they only affect borderline cases.
Problem: Determine convergence of
Solution:
This matches the logarithmic p-test with p = 2 > 1.
Alternatively, substitute :
The integral converges with value .
Problem: Show diverges.
Solution:
Logarithmic p-test with p = 1 (boundary case).
Substitute :
The integral diverges.
Problem: Determine convergence of
Solution:
Substitute , then :
This is logarithmic p-test with p = 2 > 1.
The integral converges.
The p-integrals serve as the primary comparison standards because:
Most rational and algebraic functions can be compared to appropriate p-integrals.
If as for some constant , then:
Problem: Determine convergence of
Solution:
Dominant terms as :
This is equivalent to p = 2 > 1.
The integral converges.
Problem: Determine convergence of
Solution:
As :
This is p = 3/2 > 1.
The integral converges.
Problem: Determine convergence of
Solution:
Two singularities: x = 0 and x = 1. Split at x = 1/2.
Near x = 0:
This is p = 2/3 < 1 ✓
Near x = 1:
This is p = 1/3 < 1 ✓
Both parts converge, so the integral converges.
To analyze at a problem point:
Problem: Evaluate
Solution:
Problem: Evaluate
Solution:
Substitute , :
Problem: Evaluate
Solution:
Note as , so this is Type II.
Since (by L'Hôpital):
Problem: Evaluate
Solution:
Singularities at both and .
Near :
Near :
Both are p = 1/2 < 1, so both parts converge.
Using substitution : the integral equals .
Wrong: Treating as a regular integral and computing .
Right: Recognize the singularity at and evaluate as .
Wrong: Thinking converges for .
Right: For Type II (singular), it's . The conditions are opposite for Type I and Type II!
Wrong: Claiming by symmetry.
Right: Both halves diverge independently. The integral is undefined, not zero. (The symmetric limit gives the Cauchy Principal Value, which is different.)
Wrong: Evaluating as one limit.
Right: Split at . Check (singular) and (infinite) separately.
Wrong: Computing directly.
Right: There's a singularity at inside the interval. Split into .
Problem: Evaluate
Solution:
Use integration by parts twice, or recognize this as the real part of:
Taking the real part:
Problem: Evaluate
Solution:
This is related to the Gamma function. Using integration by parts twice:
The boundary term vanishes. For the remaining integral:
So
Problem: Analyze
Solution:
Singularities at both and .
Near : → p = 1/2 < 1 ✓
Near : → p = 1/2 < 1 ✓
Both singularities are integrable. Using substitution : the integral equals .
Problem: Evaluate
Solution:
Factor:
Partial fractions:
Problem: Evaluate
Solution:
Note
Substitute , :
Before evaluating any integral, scan the integrand for points where it might blow up. Check denominators, square roots, and logarithms.
: needs p > 1. : needs p < 1. Opposite!
Many improper integrals require L'Hôpital's rule or known limits like .
For mixed integrals, choose splitting points that separate singularities and infinite limits. Often x = 1 works well.
Substitutions can convert Type I to Type II or vice versa. For example, swaps 0 and ∞.
Many convergence tests for improper integrals parallel those for series. This helps with intuition.
| Integral | Type | Converges? | Value |
|---|---|---|---|
| Type I | Yes (p=2>1) | 1 | |
| Type I | No (p=1) | — | |
| Type II | Yes (p=½<1) | 2 | |
| Type II | No (p=1) | — | |
| Type I | Yes | 1 | |
| Type I (both) | Yes | π | |
| Type II | Yes | −1 | |
| Mixed | No | — |
Challenge 1
Determine convergence and evaluate if convergent:
Substitute u = 1 + x².
Challenge 2
For what values of α does the following converge?
Near x = 1, use L'Hôpital. Near x = 0, compare with x^α.
Challenge 3
Evaluate the following integral:
Split at x = 1 and use the substitution u = 1/x on one part.
Challenge 4
Prove that the following integral converges:
Find the singularities and check the exponent at each.
| Test | When to Use | Key Requirement |
|---|---|---|
| Direct Comparison | Can establish or | Non-negative functions |
| Limit Comparison | Functions with similar asymptotic behavior | |
| p-Test (Type I) | Power functions at infinity | Converges iff p > 1 |
| p-Test (Type II) | Power functions at singularity | Converges iff p < 1 |
| Logarithmic Test | Products with | Same as p-test after substitution |
| Cauchy Criterion | No explicit antiderivative | Tail integrals → 0 |
| Absolute Convergence | Oscillating integrands | converges |
Identify Problem Points
Check for infinite limits (Type I) and singularities (Type II)
Split if Mixed
Separate each problem so each piece has only one issue
Find Dominant Behavior
Determine or
Apply Appropriate Test
Use p-test, comparison, or limit comparison based on the form
Conclude
All pieces must converge for the whole integral to converge
Problem: Determine convergence of
Solution:
Exponentials dominate any polynomial. For any :
Since converges for any :
The integral converges.
Problem: Determine convergence of
Solution:
Singularity at . Near 0:
This is p = 1 (borderline). The integral diverges.
Problem: Determine convergence of
Solution:
Singularity at . Near 0:
This is p = 3/2 > 1. For Type II, we need p < 1.
The integral diverges.
Problem: Determine convergence of
Solution:
Compare with a p-integral. For any :
Choose . Then .
The integral converges.
where deg(Q) - deg(P) = k
Behaves like . Converges at ∞ iff k > 1.
for a > 0
Always converges at ∞. Dominates any polynomial.
Log factors don't affect convergence threshold for p-test.
,
Use Taylor expansions to find the equivalent power.
Use direct comparison when you can easily show f(x) ≤ g(x) or f(x) ≥ g(x). Use limit comparison when the functions have similar asymptotic behavior but the inequality is hard to establish directly.
The p-integrals: ∫₁^∞ 1/xᵖ (converges for p>1), ∫₀¹ 1/xᵖ (converges for p<1), and exponentials like e^{-x} which decay faster than any polynomial.
Logarithms grow slower than any positive power: (ln x)^k / x^ε → 0 for any ε > 0. So ∫₁^∞ 1/(x(ln x)^p) converges iff p > 1, similar to the p-test.
∫ₐ^∞ f converges iff for every ε > 0, there exists M such that |∫ᵦ^c f| < ε for all c > b > M. It's useful when you can't find the antiderivative explicitly.
Not directly. For oscillating functions like sin(x)/x, use the absolute value: if ∫|f| converges, then ∫f converges absolutely. For conditional convergence, use Dirichlet or Abel tests.
Find the dominant terms as x→∞ (or x→a for singularities). For rational functions, compare the degrees of numerator and denominator. The difference gives you p.
If L = 0 and ∫g converges, then ∫f converges (f is 'smaller'). If L = ∞ and ∫g diverges, then ∫f diverges (f is 'larger'). Other combinations are inconclusive.
Split the integral at a convenient point (like x=1) so each piece has only one problem. Then analyze each piece separately. The integral converges iff ALL pieces converge.
It's about how fast the function must decay. At ∞, you need f→0 fast enough (p>1). At a singularity, the blow-up can't be too severe (p<1). The threshold p=1 is borderline for both.
Yes! For example, ∫₁^∞ sin(x²) dx converges even though sin(x²) oscillates. However, for non-negative integrands, f(x)→0 is necessary but not sufficient for convergence.
: converges ⟺ p > 1
: converges ⟺ p < 1