What Is U-Substitution?
U-substitution (also called integration by substitution or the change-of-variables method) is the integration technique that reverses the chain rule. The chain rule manufactures derivatives shaped like g'(x)·F'(g(x)); u-substitution recognizes that shape inside an integrand and unwinds it:
∫ f(g(x))·g'(x) dx = ∫ f(u) du, where u = g(x)
The move rewrites a tangled integral in x as a clean one in the new variable u — clean precisely because the g'(x) factor gets absorbed into du = g'(x) dx and disappears. Integrate in u with basic rules, then substitute u = g(x) back, and the original integral is solved. Spotting the pattern — an inner function whose derivative is also present as a factor — is the entire skill, and it's the single most-used integration technique in Calculus II.
How to Use This Calculator
Enter an integrand. The calculator hunts for the substitution automatically: it collects every candidate inner function (arguments of trig/exp/log functions, bases of powers, denominators), and for each candidate u = g(x) it divides the integrand by g'(x) and checks whether what remains can be written purely in terms of u. When a candidate works, you get the complete textbook write-up: the choice of u, the differential du, the rewritten integral in u, the antiderivative in u, and the back-substituted final answer — plus an independent verification step, where the answer is differentiated and matched against the original integrand numerically. If no substitution fits, the calculator says so honestly and points you to the technique more likely to succeed.
Worked Example
Integrate ∫2x·cos(x²) dx. The inner function candidate is x² (the argument of cosine), so set u = x². Then du = 2x dx — and the integrand conveniently contains exactly that 2x factor:
∫2x·cos(x²) dx = ∫cos(u) du = sin(u) + C = sin(x²) + C
Verification by the chain rule: d/dx[sin(x²)] = cos(x²)·2x ✓ — the derivative regenerates the original integrand, which is what "substitution reverses the chain rule" means concretely.
Choosing u: the Patterns That Work
| Integrand shape | Choose u = | Example |
|---|---|---|
| x·(stuff with x²) | x² | x·e^(x²), x·cos(x²) |
| x²·(stuff with x³) | x³ | x²·e^(x³) |
| cos(x)·(stuff with sin x) | sin(x) | sin³x·cos x |
| (stuff)/(inner function) | the denominator | x/(x²+1) → ½ln(x²+1) |
| g'(x)·√(g(x)) | g(x) | 2x·√(x²+4) |
The unifying signature: some factor of the integrand is (up to a constant) the derivative of another part. Constant mismatches are fine — if du is off by a factor of 3, compensate with 1/3 outside the integral.
Common Mistakes to Avoid
- Forgetting to convert dx. Every x-item must leave the integral, including dx itself — du = g'(x)dx is the exchange rate, not an optional garnish.
- Leaving a stray x behind. If, after substituting, the integral still contains x, the substitution hasn't worked — either compensate algebraically or pick a different u. An integral mixing u and x is meaningless.
- Skipping back-substitution. The answer to a question posed in x must be delivered in x: sin(u) + C is unfinished until u becomes x².
- Mishandling definite integrals. Either convert the bounds to u-values (u(a) and u(b)) and never return to x, or back-substitute first and use the original bounds — mixing the two conventions double-counts the transformation.
Real-World Applications
Substitution is less a topic than a load-bearing wall: most integrals that arise in science need at least one change of variables before any rule applies. Standardizing a normal distribution in statistics — converting any Gaussian to the standard z-form — is exactly the substitution u = (x−μ)/σ inside a probability integral, performed millions of times daily inside statistical software. Physics leans on it for every change of coordinates: switching an integral from Cartesian to polar or spherical coordinates is multivariable u-substitution, and the "Jacobian factor" that appears is the multidimensional version of du = g'(x)dx.
In engineering practice, substitution tames integrals of damped oscillations (u = the exponent), reactor kinetics (u = concentration expressions), and RC-circuit charging laws (u = the voltage difference), while in finance the log-substitution u = ln(S) converts the Black–Scholes equation into a form solvable with standard tools. Learning to see the g'·F(g) pattern is learning to read the integral's structure — which is why it's the first technique taught and the last one abandoned.
Frequently Asked Questions
How do I know u-substitution applies to my integral?
Scan for an inner function whose derivative (up to a constant) appears as a multiplying factor. If differentiating your candidate u produces something visible in the integrand, you're in business; if not — say ∫x·sin(x)dx, where u = sin(x) leaves an unabsorbed x — you likely need integration by parts instead.
What if the constant doesn't match exactly?
Adjust with constants freely: for ∫x·e^(x²)dx, du = 2x dx but only x dx is present, so write x dx = du/2 and carry the ½ outside. Only constant mismatches can be fixed this way — a leftover variable factor cannot be moved outside an integral.
How does the calculator verify its answers?
It differentiates the final antiderivative symbolically and compares it against the original integrand at several sample points. Only answers that reproduce the integrand are reported — the same differentiate-to-check habit worth adopting for hand computations.
Does substitution work on definite integrals?
Yes, elegantly: transform the bounds along with the variable — ∫₀¹ 2x·cos(x²)dx becomes ∫₀¹ cos(u)du with u-bounds u(0)=0 and u(1)=1 — and evaluate entirely in u, never returning to x. For numeric evaluation of a specific definite integral, the Definite Integral Calculator is the quickest route.
Why is it called the "reverse chain rule"?
Because differentiating F(g(x)) by the chain rule yields F'(g(x))·g'(x) — exactly the integrand shape substitution targets. The technique recognizes chain-rule output and reconstructs the F(g(x)) that produced it, the same way integration by parts reverses the product rule.
Are there substitutions beyond u = g(x)?
Yes — trigonometric substitution runs the machinery in reverse, setting x = sin(θ) or x = tan(θ) to dissolve square roots like √(1−x²) into trig identities, and Weierstrass's t = tan(x/2) rationalizes any trig integrand. Both are changes of variable at heart, sharing the same du bookkeeping this page demonstrates, just with the roles of old and new variable swapped.