First decide what function is being integrated
Definition 5.7.1
(Rational function and its real domain)
A rational function is a quotient of real polynomials with nonzero denominator polynomial. Its original real domain excludes every denominator zero. An algebraic simplification can reveal that an excluded input is a removable hole rather than a pole, but it does not silently change the function's declared domain.
Polynomial division comes before partial fractions.
Theorem 5.7.2
(Division with remainder)
Let be real polynomials with . There are unique polynomials with , where or . Consequently wherever .
Proof
(Cancel the leading term, then subtract two answers)
Repeated cancellation of the leading term of the current remainder, by subtracting a monomial multiple of , produces and . At each nonterminal step the remainder degree falls, so the algorithm stops. For uniqueness, subtract two answers : then . A nonzero multiple of cannot equal a polynomial of smaller degree, so and then .
The polynomial part integrates by powers; the proper rational remainder is the part needing decomposition.
Do not start by placing unknown constants over whatever factors look convenient. The denominator multiplicities and the degrees of the irreducible factors determine exactly which numerators are allowed. This is a finite algebra problem before it is a calculus problem.
The precise decomposition being used
Assume a factorization is supplied:
where the factors are distinct monic real irreducible polynomials, each of degree one or two, and the multiplicities are positive integers. We use this given factorization; we are not proving the general real factorization theorem. Absorb the scalar into the numerator of the proper part. Its unique partial-fraction form has one term for every , with each numerator zero or of degree smaller than .
Thus a real linear factor requires a constant numerator at every power, whereas an irreducible quadratic requires a numerator of degree at most one at every power. A missing lower power is allowed only when its coefficient actually turns out to be zero. Guessing it away beforehand restricts the space of possible functions.
Theorem 5.7.3
(Partial fraction decomposition)
Let be as above and let be proper, so or . Then can be written in exactly one way as a sum of terms over and , where each is zero or has degree smaller than .
Proof
(Coprime blocks, then repeated division)
Here is how the algebraic dependency supplies existence and uniqueness. Distinct irreducible factors, and their powers, are coprime. The coprime-block argument uses polynomial division and Bézout identities to separate a proper fraction over two coprime blocks with the required numerator bounds. Applying it repeatedly separates the finitely many power blocks. Within one block, repeated division of its numerator by the underlying factor gives
with zero remainders permitted. Dividing by supplies the complete chain of powers. Uniqueness follows first for the coprime blocks and then for these successive division remainders. This explains the standard form without assuming that an arbitrary coefficient system must be solvable. The result is inherited algebra; the integration consequences are our next task.
Common mistake
(Leaving out lower powers or giving a quadratic a constant numerator)
For , the guess would need for every , which fails: forces and forces . The correct form is , and here , . Likewise cannot be written as with a constant ; an irreducible quadratic needs a numerator of degree at most one.
Linear factors and the logarithm's absolute value
For a real root , the first power integrates to on any interval avoiding . On the right this is the derivative rule for the positive logarithm; on the left write the absolute value as and apply the chain rule. The same derivative results on both sides.
Higher powers behave differently. For an integer ,
Differentiation verifies this formula and its sign. Only the first power produces a logarithm. A repeated pole does not mean that every term in its decomposition has the same type of primitive.
Common mistake
(Expecting a logarithm from every power of a linear factor)
The derivative of is , not . Hence has no logarithm. For example, on the derivative of is , so the formula is verified. Only the first power in the decomposition contributes a logarithm.
Concept lensAlgorithmic
A simple pole's coefficient from one evaluation
Let be a simple real root of , so with . Multiply by and let : . For this gives at and at , as found by solving.
Quadratics: why both logarithms and arctangents appear
A monic irreducible real quadratic can be written with , by completing the square. Its positivity excludes real zeros. Set . Split any allowed numerator as . The first part is proportional to the derivative of the denominator; the remaining constant part is not.
For the first power, this gives
The quadratic is positive, so its logarithm needs no absolute value. The arctangent is its usual real principal function, whose derivative is . Substitution and the chain rule verify every scale factor. A linear numerator does not mean that the whole fraction is a logarithmic derivative.
For higher powers, the derivative-of-denominator part still integrates directly. The remaining ingredient is
Lemma 5.7.4
(Reduction formula for powers of a quadratic)
Let and on the real line. Then and, for ,
Proof
(Differentiate a one-term guess)
Choose primitives on the real line. The base is . To reduce a higher power, differentiate an expression whose numerator has degree one:
Indeed the quotient/product rule gives a numerator ; replacing by gives the displayed identity. Rearranging and integrating proves the recurrence
These equalities identify primitives up to a final arbitrary constant. Each step lowers the positive integer index, so it ends at arctangent after finitely many steps. The numerator part proportional to integrates to a logarithm when the power is one, and to a rational expression for higher powers, by substituting .
Proof X-ray
How the reduction proof chooses its guess
The proof does not integrate directly; it differentiates a different expression and reads the answer off. The non-obvious move is the guess : a numerator of degree one is the right size, because differentiating it produces a numerator of degree two, which splits into the target power and the lower power . The hypothesis keeps positive, so no denominator vanishes and the identity holds on the whole line. The substitution is where the two powers separate; the coefficient in front of is nonzero for , which is why the recurrence can be solved for . It closes because each step lowers by one, ending at the base case , the arctangent. Without the restriction the division by would be undefined.
We have therefore proved, conditional on the stated real factorization and the established algebraic decomposition, that a rational function has a primitive assembled from a polynomial, rational terms, logarithms and arctangents on each permitted real interval. The calculation is finite; it makes no claim that every nonrational elementary integrand has an elementary primitive.
Organize the calculation before solving coefficients
After division, record each distinct factor once together with its multiplicity. A quadratic factor of multiplicity three contributes three linear numerators, hence six scalar coefficients; it does not contribute one constant. The total number of scalar coefficients equals the degree of the denominator of the proper part. Clearing denominators produces a polynomial identity of smaller degree, so its coefficient equations account for exactly those degrees of freedom. Existence and uniqueness above justify solving that system; counting alone would not prove that its equations are independent.
Substitution at a real root can quickly reveal the numerator attached to the highest corresponding linear power. It does not determine every coefficient of a repeated factor, and an irreducible quadratic has no real root to substitute. Coefficient comparison completes the calculation. Finally, recombine the fractions or differentiate the primitive: a decomposition with the correct-looking denominators can still have a wrong coefficient or sign.
There is also a reason that the final answer contains precisely arbitrary constants rather than some additional unknown functions. If two differentiable primitives agree in derivative on an interval, apply the mean value theorem to their difference between any two points. That difference has derivative zero, so its values are equal. On separate domain intervals the argument gives separate constants, because there is no intervening interval on which to apply the theorem.
Worked example: Division, a repeated pole and a quadratic
Worked example 5.7.5
(Division, a repeated pole and a quadratic)
Question. Find all real primitives, on their connected domains, of
Solution. The original domain excludes . Division gives polynomial part and remainder . Write the proper part as
Clearing denominators and setting gives . Comparing the cubic, quadratic and linear coefficients gives , and . Hence , and ; the constant coefficient also agrees. The decomposition is
A primitive is
Differentiate each term: the rational term contributes the negative repeated-pole term, and the arctangent contributes . Recombination recovers the original numerator. All primitives are on and on , with independent constants. The logarithm's absolute value permits each side separately; it does not connect them across the pole.
Worked example: A squared irreducible factor needs a reduction
Worked example 5.7.6
(A squared irreducible factor needs a reduction)
Question. Evaluate the proper integral
Solution. Put , carrying the endpoints to and . The numerator becomes . The odd numerator part has primitive . The recurrence with and gives
Consequently a primitive of the full transformed integrand is
Its rational part has derivative ; the arctangent part adds . Their sum is exactly . Evaluating at the endpoints yields
The denominator is positive and the integrand continuous on the entire interval. This is a proper application of the fundamental theorem, with no pole or convergence limit hidden inside the substitution.
Worked example: Real poles and the choice of interval
Worked example 5.7.7
(Distinct real poles and a definite integral)
Question. Evaluate , and explain why the antiderivative found cannot be applied on .
Solution. The numerator and denominator have equal degree, so division gives . The denominator factors as with distinct real roots, so and . Setting gives , and gives . Thus
The interval lies in , where the integrand and satisfy the fundamental theorem. So
As a check, the integrand is about at and the interval has length . The interval contains the pole at . There the integrand is unbounded, is undefined at , and subtracting proves nothing; the integral is improper and, because is not integrable on either side of , it diverges.
A domain check is part of the answer
Cancellation gives only when . If the original domain is retained, constants on its two components are independent. If one deliberately fills the removable hole and asks for a primitive of the extended function on the whole line, a single constant is required. Those are different domain specifications.
A genuine pole is different again: an endpoint subtraction across it does not justify a proper integral, and any improper integral must use separate limits. Those conditions belong to the theory of improper integrals. Algebra supplies an identity on its domain; the fundamental theorem still requires an interval on which the integrand and primitive meet its hypotheses.
Here the need for a connected interval becomes a concrete failure.
Counterexample mode
One constant is not enough across a pole
Claim: every primitive of on its domain has the form for a single constant . Define for and for . On each of the two intervals , so is a primitive of on the domain. But while , so no single works. The mean value theorem only applies on an interval, so it gives one constant on and an independent one on . The repair is to state the primitive as on the left and on the right.
Exercises
Quick check
(Write the form of the proper partial fraction decomposition of a fraction with denominator , and count the unknown coefficients.)
Use one term for each power of each irreducible factor.
Solution · Answer
The form is . It has five unknowns, equal to the degree of the denominator.
Quick check
(Find on an interval avoiding .)
Use the power formula with .
Solution · Answer
. Differentiating gives .
Exercise 5.7.8
(Find all real primitives of on .)
Core
Solution(3)
- 1
Write the allowed form
The denominator has a double root at and a simple root at , so .
- 2
Solve for the coefficients
Key step
Clearing denominators, . At , . At , . The coefficient of gives , so . The coefficient checks: .
Because: Substituting the roots finds and ; the coefficient then gives .
- 3
Integrate each term
On the primitives are . Differentiating confirms .