Integration by Parts
The product rule for derivatives reverses into integration by parts, trading the integral of for the integral of whenever the second is easier. The LIATE ordering fixes which factor to differentiate.
╌╌╌╌
The substitution rule is the chain rule read backward. Integration by parts is the product rule read backward. Both convert a hard integral into an easier one, and between them they handle a large fraction of the integrals that have elementary antiderivatives.
Start from the product rule for differentiable and :
Integrate both sides. The left side integrates to , and the right side splits into two integrals:
Solving for the first integral on the right gives the formula.1
The formula is easier to remember with the substitutions and , so that and :
Reading it as a strategy: an integrand is split into a part that will be differentiated and a part that will be integrated. The exchange is only worthwhile when is simpler than . Choosing and well is the entire skill.
The formula as areas
For a positive increasing function the formula has a direct reading as areas. Let run from to . The corner rectangle of area splits along the curve into the region under it and the region to its left, and subtracting the inner rectangle leaves the standard statement .
Choosing and
Two requirements constrain the choice: must be something integrable to get , and should get simpler (or at least no worse) when differentiated. A useful ordering for picking is the acronym LIATE: whichever factor comes first on this list is the one to differentiate, so it becomes ; the rest is .
| Priority for | Class | Example factor | Why differentiate it |
|---|---|---|---|
| 1 | L — logarithmic | has no simple antiderivative; its derivative is algebraic | |
| 2 | I — inverse trig | , | same reason; derivative is algebraic |
| 3 | A — algebraic | differentiating lowers the degree toward | |
| 4 | T — trigonometric | , | cycles under both operations; usually goes in |
| 5 | E — exponential | unchanged by both; usually goes in |
LIATE is a guideline, not a theorem; it fails on constructed examples, but it picks the right split on nearly every integral encountered in practice.
A polynomial against a trigonometric factor
Consider . The integrand is algebraic times trigonometric, so LIATE selects (the A factor) and . Then
Applying the formula,
The exchange replaced with the elementary . The opposite choice , produces , which is worse: differentiating did not simplify it, and integrating raised the degree.
A single logarithm
Some integrands have no obvious second factor. For , take and let be the whole integrand:
Then
Parts works here because has a simpler derivative than itself. The same choice evaluates and , the other functions whose derivatives are algebraic.
When a power of multiplies the logarithm, still differentiate the logarithm; one pass turns it algebraic.
Repeated parts
When one application leaves a still-difficult integral, applying parts again may finish the job. For , the factor becomes simpler on differentiation while is unchanged, so , :
The remaining is simpler but not immediate, so apply parts once more with , :
Substituting back,
Each pass lowers the polynomial degree by one; a degree- polynomial against needs applications.
The tabular method
When is a polynomial (so its derivatives reach ) and integrates easily any number of times, repeated parts can be organized into a table. Differentiate the left column down to zero, integrate the right column at each step, and multiply along the diagonals with alternating signs.
Written as a procedure:
- 1build column by differentiating repeatedly until reaching
- 2build column by integrating once per row
- 3attach signs down the rows
- 4for each entry except the last do
- 5form the product with the sign of row
- 6result is the sum of these signed products, plus
A trigonometric works the same way, since integrates cyclically without growing.
Cyclic integrals
Sometimes neither factor simplifies, yet repeated parts still succeeds because the original integral reappears. For , choose , :
The new integral is no simpler, but apply parts again with , :
Substituting the second line into the first returns the starting integral on the right:
This is an equation for the unknown integral. Add to both sides and divide by :
Definite integrals by parts
Combining the formula with Part 2 of the Fundamental Theorem gives a version for definite integrals: evaluate the boundary term and integrate the correction term over the same interval.2
For , take , , so and :
The remaining integral yields to the substitution , , giving . Therefore
Since on , the value is the area under the arctangent curve there.
Reduction formulas
Parts can also express an integral with exponent in terms of the same integral with a smaller exponent, producing a reduction formula that peels the exponent down by recursion. Take with and . Then and , so
Replace and the integral splits, reproducing on the right. Solving for it gives the reduction formula.3
Applied repeatedly, this drives the exponent down to when is odd, or when is even.
The same pattern gives reduction formulas for , , , and , all of which appear among the trigonometric integrals.
When parts applies
| Integrand shape | Choice of | Outcome |
|---|---|---|
| , , | polynomial degree drops each pass; use the table | |
| , | or | one pass turns the transcendental factor algebraic |
| , , alone | the function, | one pass gives an elementary integral |
| , | either factor, twice | cyclic; solve the resulting equation |
| , , | the power | reduction formula lowering |
Integration by parts and substitution are the two general methods. The other techniques are these two specialized to a particular integrand shape.
Footnotes
- Stewart, Calculus, §7.1 — Integration by Parts: the formula derived from the product rule, in both the and the notation. ↩
- Stewart, Calculus, §7.1 — the definite-integral form obtained by evaluating both sides between and with the Fundamental Theorem. ↩
- Stewart, Calculus, §7.1, Example 6 — derivation of the reduction formula for by parts. ↩
╌╌ END ╌╌