Integration
Partial Fractions Integration
Decompose P(x)/Q(x) into simpler fractions, integrate each piece
To integrate a rational function P(x)/Q(x), factor the denominator, decompose into a sum of A/(x−a) terms plus (Bx+C)/(x²+bx+c) terms, and integrate term-wise. The Heaviside cover-up trick reads off coefficients in one line.
- Distinct linear (x − a)A/(x − a) → A · ln|x − a|
- Repeated (x − a)ⁿA₁/(x − a) + A₂/(x − a)² + … + Aₙ/(x − a)ⁿ
- Irreducible quadratic(Bx + C)/(x² + bx + c) → log + arctan
- Heaviside cover-upA = P(a)/Q'(a) for simple linear poles
- Prerequisitedeg P < deg Q (else polynomial long division first)
- OutputSum of logarithms and arctangents
Watch the 60-second explainer
A condensed visual walkthrough — narrated, captioned, under a minute.
The idea
Adding fractions with different denominators creates a single fraction with a complicated denominator: A/(x−1) + B/(x+2) becomes (A(x+2) + B(x−1))/((x−1)(x+2)). Partial fractions runs that arithmetic backwards. Given a rational integrand P(x)/Q(x), we split it back into the simple pieces whose denominators are the factors of Q(x). Each simple piece has a known antiderivative — log or arctan — and the original integral falls out as a sum.
The technique is mechanical once you factor Q(x). The skill is recognizing which decomposition template to use and how to extract the coefficients.
The three decomposition templates
| Factor of Q(x) | Contributes | Each piece integrates to |
|---|---|---|
| Distinct linear (x − a) | A/(x − a) | A · ln|x − a| |
| Repeated linear (x − a)ⁿ | A₁/(x − a) + A₂/(x − a)² + ⋯ + Aₙ/(x − a)ⁿ | log (n = 1) or power rule (n ≥ 2) |
| Irreducible quadratic x² + bx + c | (Bx + C)/(x² + bx + c) | log + arctan after completing the square |
| Repeated irreducible quadratic (x² + bx + c)ⁿ | Stacked (Bᵢx + Cᵢ)/(x² + bx + c)ⁱ terms | Reduction formulas; rarely seen in practice |
"Irreducible" means the quadratic has no real roots — i.e., its discriminant b² − 4c < 0. Reducible quadratics factor into two linear pieces and reduce to the linear cases.
How to apply it
- Check degree — if deg P ≥ deg Q, do polynomial long division first.
- Factor Q(x) into linear and irreducible quadratic factors.
- Write the decomposition template based on the factors.
- Solve for the unknown constants — by Heaviside cover-up (for distinct linear factors) or by matching coefficients.
- Integrate each piece — log for simple linear, power rule for repeated, log + arctan for quadratic.
- Combine. Don't forget + C.
Worked example — ∫ dx / ((x − 1)(x + 2))
The denominator has two distinct linear factors. Template:
1/((x − 1)(x + 2)) = A/(x − 1) + B/(x + 2).
Apply the Heaviside cover-up. To find A, cover up (x − 1) in the original and substitute x = 1:
A = 1/(x + 2) |_{x = 1} = 1/(1 + 2) = 1/3.
B = 1/(x − 1) |_{x = −2} = 1/(−2 − 1) = −1/3.
Two lines, no algebra. The decomposition is:
1/((x − 1)(x + 2)) = (1/3)/(x − 1) − (1/3)/(x + 2).
Integrate:
∫ dx/((x − 1)(x + 2))
= (1/3) ∫ dx/(x − 1) − (1/3) ∫ dx/(x + 2)
= (1/3) ln|x − 1| − (1/3) ln|x + 2| + C
= (1/3) ln| (x − 1)/(x + 2) | + C.
Worked example — ∫ (3x + 1) / (x − 1)² dx (repeated factor)
The denominator has a repeated linear factor (x − 1)². Template:
(3x + 1) / (x − 1)² = A/(x − 1) + B/(x − 1)².
Clear denominators: 3x + 1 = A(x − 1) + B. Substitute x = 1: 4 = B, so B = 4. Match x-coefficients: 3 = A, so A = 3.
(3x + 1)/(x − 1)² = 3/(x − 1) + 4/(x − 1)².
∫ (3x + 1)/(x − 1)² dx = 3 ln|x − 1| − 4/(x − 1) + C.
The first piece integrates to a log; the second uses the power rule (∫1/(x − 1)² dx = −1/(x − 1) + C).
Worked example — ∫ dx / (x · (x² + 1))
Denominator has one linear factor x and one irreducible quadratic x² + 1 (discriminant −4 < 0). Template:
1/(x(x² + 1)) = A/x + (Bx + C)/(x² + 1).
Clear: 1 = A(x² + 1) + (Bx + C)x = Ax² + A + Bx² + Cx = (A + B)x² + Cx + A.
Match coefficients: x² gives A + B = 0; x gives C = 0; constant gives A = 1. So A = 1, B = −1, C = 0.
1/(x(x² + 1)) = 1/x − x/(x² + 1).
∫ dx/(x(x² + 1)) = ∫ dx/x − ∫ x/(x² + 1) dx
= ln|x| − (1/2) ln(x² + 1) + C
= ln| x / √(x² + 1) | + C.
The second integral uses u = x² + 1, du = 2x dx — the (Bx + C)/(x² + 1) piece has a "derivative of denominator" part that goes to log. When C ≠ 0, the remaining 1/(x² + bx + c) piece completes the square and integrates to arctan.
Partial fractions vs other techniques
| Partial fractions | U-substitution | Integration by parts | Trig substitution | |
|---|---|---|---|---|
| Best when | Rational function P(x)/Q(x) | Inner function + derivative | Product of dissimilar factors | Square roots of quadratics |
| Underlying principle | Algebraic decomposition | Chain rule reversal | Product rule reversal | Pythagorean identity |
| When to try | For every rational integrand | First — always | Second | When square roots appear |
| Algebra-heaviness | High (decomposition) | Low | Medium | Medium |
| Output forms | Logs and arctangents | Anything | uv-style mixed forms | Arctrig and logs |
| Shortcut tool | Heaviside cover-up | Pattern recognition | LIATE mnemonic | Right-triangle picture |
When partial fractions is the right tool
- Any rational integrand. When the integrand is P(x)/Q(x) and u-substitution doesn't apply directly, partial fractions is the systematic next step.
- Inverse Laplace transforms. Most rational Laplace expressions decompose into simple pieces whose inverse transforms are tabulated.
- Partial-fraction expansion in signal processing. Transfer functions H(s) = N(s)/D(s) decompose into modes, each contributing one exponential or sinusoidal response.
- Generating functions in combinatorics. Rational generating functions decompose; each term gives one sequence component.
- Probability — characteristic and moment-generating functions. Decomposition gives separable contributions to distribution computations.
- Differential equations. Method of undetermined coefficients and Laplace methods both lean on partial fractions for the final step.
Common mistakes
- Forgetting to long-divide first. If deg P ≥ deg Q, the decomposition template is wrong. Long-divide to reduce to a polynomial plus a proper rational, then decompose the proper part.
- Wrong template for repeated factors. Repeated (x − a)ⁿ requires n separate terms, one for each power. Using only A/(x − a) gives an underdetermined system.
- Missing the linear numerator for quadratics. Irreducible x² + bx + c needs (Bx + C), not just B. The numerator must span all polynomials of degree < 2.
- Heaviside on the wrong factor type. Cover-up only works for distinct simple linear poles. Repeated factors and quadratics need other methods.
- Sign errors in completing the square. Converting (x² + bx + c) to ((x + b/2)² + (c − b²/4)) requires careful sign tracking.
- Forgetting absolute values in logs. ∫dx/(x − a) = ln|x − a| + C — the absolute value is essential for negative arguments.
Connections to other techniques
U-substitution. After decomposition, some pieces (the derivative-of-denominator part of a quadratic term) reduce by u-sub to logs. Partial fractions sets up u-sub at the term level.
Trig substitution. A Weierstrass substitution t = tan(θ/2) converts rational trig integrals to rational t integrals — partial fractions then finishes them. Trig sub and partial fractions compose.
Laplace transforms. The inverse Laplace transform of a rational H(s) is computed by partial fraction decomposition. The denominator factors give the poles; the residues become the partial-fraction coefficients.
Complex analysis — residues. The Heaviside cover-up is the real-variable face of the residue theorem. A = P(a)/Q'(a) is exactly the residue of P/Q at the simple pole x = a.
Polynomial algebra. Decomposition is finite-dimensional linear algebra in disguise. The space of proper rationals with denominator Q is finite-dimensional, and the partial-fraction basis is a natural decomposition.
Frequently asked questions
What is partial fraction decomposition?
A way to rewrite a rational function P(x)/Q(x) as a sum of simpler fractions. Each factor of Q(x) gives one term in the decomposition. Distinct linear factor (x − a) contributes A/(x − a); a repeated linear factor (x − a)ⁿ contributes A₁/(x − a) + A₂/(x − a)² + … + Aₙ/(x − a)ⁿ; an irreducible quadratic x² + bx + c contributes (Bx + C)/(x² + bx + c). Solving for the constants A, B, C by matching coefficients or substituting strategic values reduces the original integral to a sum of trivial integrals.
How do I use the Heaviside cover-up trick?
For distinct linear factors, the cover-up is the fastest route. To find A in A/(x − a), "cover up" the factor (x − a) in the original P(x)/Q(x) and evaluate the remaining expression at x = a. Equivalent formula — A = P(a)/Q'(a). For 1/((x − 1)(x + 2)) split as A/(x − 1) + B/(x + 2) — A = 1/(1 + 2) = 1/3, B = 1/(−2 − 1) = −1/3. Two lines, no algebra. Only works for distinct (simple) linear poles.
What happens with repeated linear factors?
Each power of the repeated factor gets its own term. (x − a)² in Q gives A₁/(x − a) + A₂/(x − a)². A₂ comes from the cover-up at x = a using (x − a)²·P(x)/Q(x) evaluated at a. A₁ requires differentiating once before substituting — or, more commonly, solving the system by matching coefficients on the cleared expression.
What about irreducible quadratic factors?
An irreducible quadratic x² + bx + c (no real roots) contributes (Bx + C)/(x² + bx + c). The numerator is linear, not constant — because the quadratic has two-dimensional space of polynomial factors. After decomposition, the term integrates by completing the square in the denominator — (Bx + C)/(x² + bx + c) splits into a (derivative of the quadratic)/(quadratic) part (integrates to log) and a constant/(quadratic) part (integrates to arctan after completing the square).
What if the numerator's degree is greater than or equal to the denominator's?
Do polynomial long division first. Write P(x) = D(x)·Q(x) + R(x) where deg R < deg Q. Then P/Q = D + R/Q. The polynomial D integrates trivially; partial fractions handles R/Q. Skipping this step gives wrong decompositions. Always check the degree before decomposing.
How does each piece integrate?
Simple linear A/(x − a) integrates to A·ln|x − a| + C. A repeated power A/(x − a)ⁿ for n ≥ 2 integrates by the power rule to −A/((n − 1)(x − a)^(n−1)) + C. The linear-over-quadratic (Bx + C)/(x² + bx + c) splits into a log piece and an arctan piece via completing the square. So every partial-fraction integral ends up being a sum of logs and arctangents — no transcendental functions needed.
Why doesn't u-substitution work directly on rational functions?
Sometimes it does — ∫1/(x² + 1) dx is u = x, giving arctan directly; ∫(2x)/(x² + 1) dx is u = x² + 1, giving a log. But for ∫1/((x − 1)(x + 2)) dx there's no inner function with its derivative present. The denominator's factorization is essential structure that u-substitution can't access. Partial fractions exposes that structure first; integration is mechanical afterwards.