Integration by parts is the formula ∫u dv = uv − ∫v du: you split the integrand into a part u that gets simpler when differentiated and a part dv you can integrate, and trade the original integral for an easier one. The tabular method (also called the DI method) is a fast, low-error way to do the same thing when parts would otherwise be repeated several times. This guide shows both, with six worked examples that rise from one step to the "loop back" integrals that catch most students.
Where the formula comes from
The product rule says d/dx (uv) = u · dv/dx + v · du/dx. Integrate both sides and rearrange:
∫u dv = uv − ∫v du
For definite integrals the boundary term is evaluated at the limits:
∫ₐᵇ u dv = [uv]ₐᵇ − ∫ₐᵇ v du
So integration by parts is the product rule run backwards, in the same way that u-substitution is the chain rule run backwards. It is one of the core techniques of Calculus II; our college maths tutoring page for US students explains how online sessions with Ajay Vatsyayan cover the whole integration-techniques unit.
Choosing u, and when to ignore LIATE
The usual rule of thumb is LIATE: pick u from whichever type comes first in the list Logarithmic, Inverse trig, Algebraic (polynomials), Trig, Exponential. Logs and inverse trig functions have simple derivatives but awkward integrals, so they make good u's; exponentials and sin/cos integrate easily, so they make good dv's.
LIATE is a guide, not a law. The real test is: after differentiating u and integrating dv, is ∫v du easier than what you started with? Two cases break the rule:
- ∫x³ e^(x²) dx. LIATE suggests u = x³, dv = e^(x²) dx, but e^(x²) has no elementary antiderivative. Instead take u = x² and dv = x e^(x²) dx, so v = ½e^(x²). The answer is ½e^(x²)(x² − 1) + C.
- ∫x e^(x²) dx. This is not a parts question at all. The substitution w = x² gives ½e^(x²) + C in one line.
Which method to use
| integrand looks like | method | u | dv |
|---|---|---|---|
| polynomial × e^(kx), sin kx or cos kx | tabular (repeat until the polynomial hits 0) | polynomial | the exponential or trig part |
| polynomial × ln x | single step of parts | ln x | polynomial dx |
| ln x, arcsin x or arctan x alone | parts with dv = dx | the function | dx |
| e^(ax) × sin bx or cos bx | parts twice, or tabular with row stop, then solve | either (stay consistent) | the other |
| function of x times its own inner derivative | substitution, not parts | none | none |
| e^(√x), sin(√x) | substitute w = √x first, then parts | after substitution | after substitution |
The tabular method in three moves
- Make two columns. Under D (differentiate) put u; under I (integrate) put dv.
- Differentiate down the D column until you reach 0. Integrate down the I column the same number of times.
- Multiply diagonally (each D entry with the I entry one row below), attaching signs +, −, +, − in turn, and add.
It works because each diagonal product is the uv term of one round of parts, and the alternating signs come from the minus in the formula.
Worked example 1: a single round (polynomial × exponential)
Find ∫x e^(3x) dx.
u = x, dv = e^(3x) dx, so du = dx and v = (1/3)e^(3x).
∫x e^(3x) dx = (x/3)e^(3x) − ∫(1/3)e^(3x) dx = (x/3)e^(3x) − (1/9)e^(3x) + C
The 1/3 from integrating e^(3x) appears twice: once in v and once more in the second integral. Losing the second one is the most common slip here.
Worked example 2: tabular with a quadratic
Find ∫x² cos 2x dx.
| sign | D (differentiate x²) | I (integrate cos 2x) |
|---|---|---|
| + | x² | cos 2x |
| − | 2x | ½ sin 2x |
| + | 2 | −¼ cos 2x |
| 0 | −⅛ sin 2x |
Diagonal products with signs:
(+) x² · ½ sin 2x, (−) 2x · (−¼ cos 2x), (+) 2 · (−⅛ sin 2x)
∫x² cos 2x dx = ½x² sin 2x + ½x cos 2x − ¼ sin 2x + C
Without the table this needs parts twice and careful sign bookkeeping. With it, the work fits in four lines.
Worked example 3: a definite integral with ln x
Evaluate ∫₁ᵉ x² ln x dx.
LIATE: u = ln x, dv = x² dx, so du = (1/x) dx and v = x³/3. (Tabular doesn't help, because ln x never differentiates to 0.)
∫₁ᵉ x² ln x dx = [(x³/3) ln x]₁ᵉ − ∫₁ᵉ (x³/3)(1/x) dx
The boundary term is e³/3 − 0 = e³/3, since ln 1 = 0. The remaining integral is ∫₁ᵉ x²/3 dx = (e³ − 1)/9.
Result: e³/3 − (e³ − 1)/9 = (2e³ + 1)/9 ≈ 4.575
Worked example 4: an inverse trig function alone
Evaluate ∫₀^(1/2) arcsin x dx.
Take u = arcsin x and dv = dx, so du = dx/√(1 − x²) and v = x.
∫arcsin x dx = x arcsin x − ∫x/√(1 − x²) dx = x arcsin x + √(1 − x²) + C
(the last integral is a substitution with w = 1 − x²). Now apply the limits:
[x arcsin x + √(1 − x²)]₀^(1/2) = (½)(π/6) + √3/2 − 1 = π/12 + √3/2 − 1 ≈ 0.1278
Worked example 5: the loop-back integral (row-stop tabular)
Find I = ∫e^(2x) sin 3x dx.
Neither factor ever differentiates to 0, so stop the table at the row where the product is a multiple of the original integrand, and integrate that row instead of multiplying diagonally.
| sign | D (differentiate sin 3x) | I (integrate e^(2x)) |
|---|---|---|
| + | sin 3x | e^(2x) |
| − | 3 cos 3x | ½e^(2x) |
| + | −9 sin 3x | ¼e^(2x) |
The two diagonals give ½e^(2x) sin 3x − ¾e^(2x) cos 3x. The last row, integrated straight across with its sign +, gives ∫(−9 sin 3x)(¼e^(2x)) dx = −(9/4) I. So
I = ½e^(2x) sin 3x − ¾e^(2x) cos 3x − (9/4) I
(13/4) I = e^(2x)(½ sin 3x − ¾ cos 3x)
I = e^(2x)(2 sin 3x − 3 cos 3x)/13 + C
Check by differentiating: the derivative of the answer simplifies back to e^(2x) sin 3x.
Worked example 6: a cubic in the tabular method
Find ∫(x³ − 2x)e^(−x) dx.
| sign | D | I |
|---|---|---|
| + | x³ − 2x | e^(−x) |
| − | 3x² − 2 | −e^(−x) |
| + | 6x | e^(−x) |
| − | 6 | −e^(−x) |
| 0 | e^(−x) |
Diagonal products:
−(x³ − 2x)e^(−x) − (3x² − 2)e^(−x) − 6x e^(−x) − 6e^(−x)
Collect inside one bracket:
∫(x³ − 2x)e^(−x) dx = −e^(−x)(x³ + 3x² + 4x + 4) + C
Doing this with the formula would mean three rounds of parts, each with a chance to drop a sign. The table makes every sign visible.
Errors we see most often
- Losing a constant in v. ∫e^(3x) dx = (1/3)e^(3x), and ∫cos 2x dx = ½ sin 2x. In the tabular column these fractions multiply up (½, ¼, ⅛), so write each row in full.
- Signs in the tabular method. The signs belong to the D column and alternate starting with +. Don't also flip signs you already have from integrating (example 6 has both).
- Stopping tabular too early or too late. With a polynomial, continue until D reaches 0. With e^(ax) × trig, stop as soon as the original integrand reappears; continuing just goes round in circles.
- Switching choices halfway through. In loop-back integrals, if the first round uses u = trig, the second must too. Swapping u and dv in round two undoes round one and gives 0 = 0.
- Boundary terms in definite integrals. Evaluate [uv] at both limits; the lower limit often matters (here ln 1 = 0 made it vanish, but it won't always).
- Using parts when substitution is quicker. If one factor is the derivative of the inside of the other, substitute.
- No check. Differentiate your answer. It takes a minute and catches most slips.
Practice set
- ∫x sin 4x dx
- ∫x² e^(x/2) dx
- ∫₁² ln x dx
- ∫eˣ cos 2x dx
- ∫e^(√x) dx (hint: substitute first)
- ∫₀^π x² sin x dx
Answers:
- −(x/4) cos 4x + (1/16) sin 4x + C
- e^(x/2)(2x² − 8x + 16) + C
- 2 ln 2 − 1 ≈ 0.386
- eˣ(cos 2x + 2 sin 2x)/5 + C
- With w = √x, x = w², dx = 2w dw: ∫2w eʷ dw = 2(√x − 1)e^(√x) + C
- π² − 4 ≈ 5.870
All worked answers and practice answers were checked by differentiation or exact integration in Python (sympy).
Where integration by parts comes back
Parts is not just a Calculus II exercise. Later courses rely on it constantly:
- Series. The integral test often needs parts, for example to show ∑ n e^(−n) converges; see the series convergence tests flowchart.
- Laplace transforms. The rule for the transform of a derivative is one application of parts; our Laplace transforms exam guide builds on it.
- Fourier series. Coefficients of functions like x or x² on an interval are tabular integrals with sin and cos; see Fourier series for engineering maths and the engineering mathematics tutoring page.
- Reduction formulas and the Gamma function. Repeating parts on ∫₀^∞ xⁿ e^(−x) dx gives n!, the starting point for the Gamma function.
If you came from the IB, our post on IB Maths AA HL integration techniques shows how much of this you have already seen. Earlier Calculus I topics, such as related rates, are where the chain and product rules behind parts and substitution first get used.
How Maths Bodhi helps
Choosing u, keeping signs straight and spotting when to switch methods come from practising mixed integrals, not lists of one type. In online sessions we set mixed sets, watch how you choose a method and review your working, with a quick derivative check each time. We teach and review your own attempts; we don't complete graded work. For other first-year courses, see the university maths tutoring hub.
Home tuition is available in Gurgaon (Gurugram). Students elsewhere in India and abroad learn with us online. You can book a free demo class.
Written and checked by the Maths Bodhi team (November 2026).
Frequently asked questions
- What is the integration by parts formula?
- ∫u dv = uv − ∫v du, and for definite integrals ∫ₐᵇ u dv = [uv]ₐᵇ − ∫ₐᵇ v du.
- How do you choose u in integration by parts?
- Use LIATE (logs, inverse trig, algebraic, trig, exponential) as a first guess, but choose whatever makes ∫v du simpler.
- When can you use the tabular method?
- When one factor differentiates to 0 after a few steps (a polynomial) and the other is easy to integrate repeatedly, or with the row-stop version for e^(ax) times sin or cos.
- How do you do integration by parts with e^x sin x?
- Apply parts twice (or stop the table at the third row) until the original integral reappears, then solve for it algebraically.
- Why does integration by parts give me the same integral back?
- In a loop-back integral that is expected; move it to the left side and solve, but if you get 0 = 0 you swapped u and dv in the second round.
- Is integration by parts in Calc 1 or Calc 2?
- It is usually taught in the Calculus II integration techniques unit, though some courses introduce it at the end of Calculus I.
