When a Rationalizing Substitution Has Not Worked
A substitution for an algebraic integral is useful when it turns the complete transformed integrand into a rational function of the new parameter. A substitution can fail by leaving an irrational factor in that integrand, by making the algebra more complicated than the original expression, or by using an inverse formula that does not match the domain being integrated over.
It is important to judge the complete transformed integrand, not each piece in isolation. The substituted expression for \(x\), the radical, and the differential may each contain a radical, yet those factors can cancel when multiplied. Conversely, a substitution that makes one radical simple can leave a different irrational expression in the differential.
What to Check After Substitution
Suppose the integral is \(\int f(x)\,dx\), and a proposed substitution expresses \(x\) as a differentiable function \(x=\phi(t)\) on an interval. The change of variable gives
The expression to inspect is \(f(\phi(t))\phi'(t)\), not just \(f(\phi(t))\), and not just \(\phi'(t)\). Simplify products and quotients before deciding whether radicals remain. Then check that the parameter interval and inverse formulas correspond to the original \(x\)-interval. A formula obtained by squaring an equation may have more than one possible branch.
Identify the interval of \(x\) under consideration and any sign restrictions that follow from it.
Write the substituted expressions for \(x\), every radical, and \(dx\), using the same branch throughout.
Multiply and simplify the full coefficient of \(dt\). Radical factors may cancel, or a residual radical may remain.
Choose a parameter that expresses the relevant radical relations rationally and is reversible on the interval.
Worked Example: A Substitution That Leaves an Irrational Differential
Worked Example: Repairing a Residual Radical
Consider \(\int \frac{dx}{\sqrt{x^2-1}}\) on \(x>1\). A tempting choice is \(t=\sqrt{x^2-1}\), which makes the radical itself equal to \(t\). Since \(x=\sqrt{1+t^2}\) on this domain, differentiation gives
Substitution into the whole integrand yields
The residual factor \(1/\sqrt{1+t^2}\) is not rational in \(t\), so this parameter has not rationalized the integral. It has simplified the displayed radical but left an irrational differential.
On \(x>1\), use the Euler parameter \(u=x+\sqrt{x^2-1}\). It satisfies \(u>1\). Multiplying \(u=x+\sqrt{x^2-1}\) by its conjugate gives \(1/u=x-\sqrt{x^2-1}\). Adding and subtracting these equations gives
Therefore
The transformed integrand is rational. An antiderivative is \(\ln u\). To check it, \(du/dx=1+(d/dx)\sqrt{x^2-1}=1+x/\sqrt{x^2-1}=u/\sqrt{x^2-1}\), so
Why the Complete Integrand Matters
A common diagnostic error is to reject a parameter as soon as one substituted piece still contains a radical. That conclusion does not follow: factors in the complete transformed integrand may cancel. The next example uses precisely such a cancellation.
Worked Example: Radicals That Cancel in the Full Expression
Evaluate \(\int \frac{\sqrt{x^2-1}}{x}\,dx\) on \(x>1\), using \(t=\sqrt{x^2-1}\). Here \(t>0\) and \(x=\sqrt{t^2+1}\), so
Although \(x\) and \(dx/dt\) involve \(\sqrt{t^2+1}\), the entire integrand becomes
This is rational. Since \(t^2/(t^2+1)=1-1/(t^2+1)\), integration gives \(t-\arctan t+C\). Check by differentiating with respect to \(x\). Because \(dt/dx=x/t\),
The final equality uses \(1+t^2=x^2\) and \(x>1\). The example shows why the complete transformed integrand, rather than its separate ingredients, decides whether the substitution rationalizes the integral.
Replacing an Unnecessarily Complicated Parameter
A parameter can be algebraically valid but still be a poor choice. For example, a first substitution may turn a simple radical into an expression containing another radical inside a radical. This is a sign to reconsider which combination of the original expressions should be the parameter. The aim is not to simplify one isolated piece at any cost; it is to make the relationships needed by the whole integrand rational.
Worked Example: Avoiding a Nested Radical
Consider \(\int \frac{dx}{\sqrt{x+\sqrt{x^2-1}}}\) on \(x>1\). If \(v=\sqrt{x^2-1}\), then \(x=\sqrt{1+v^2}\), and the denominator becomes \(\sqrt{\sqrt{1+v^2}+v}\). This introduces a nested radical and does not give a rational transformed integrand.
Instead set \(q=\sqrt{x+\sqrt{x^2-1}}\). The domain \(x>1\) implies \(q>1\). Since \(q^2=x+\sqrt{x^2-1}\), the conjugate relation gives \(q^{-2}=x-\sqrt{x^2-1}\). Thus
The denominator in the original integrand is \(q\), so the full transformed integrand is
It is rational, and an antiderivative is \(q+\frac{1}{3q^3}\). Differentiating with respect to \(q\) gives \(1-q^{-4}\), exactly the transformed integrand. The parameter was selected from the full expression under the radical, avoiding the nested expression produced by the first choice.
Domain and Branch Errors
Substitution formulas must respect the interval of integration. For example, solving \(t=\sqrt{x^2-1}\) for \(x\) gives \(x=\sqrt{1+t^2}\) only on the positive branch. On \(x<-1\), the correct inverse is \(x=-\sqrt{1+t^2}\). Ignoring that distinction can make a formally plausible calculation represent the wrong part of the domain.
Worked Example: Repairing a Branch Mismatch
Consider again \(\int \frac{dx}{\sqrt{x^2-1}}\), now on \(x<-1\). Taking \(t=\sqrt{x^2-1}>0\), the correct inverse is \(x=-\sqrt{1+t^2}\), so \(dx=-t/\sqrt{1+t^2}\,dt\). The transformed integrand is \(-dt/\sqrt{1+t^2}\), which remains irrational.
Use instead \(u=x+\sqrt{x^2-1}\). For \(x<-1\), \(u\) lies in \((-1,0)\). The algebraic relations are still
In this interval \(u-u^{-1}>0\), as required for the positive square root. Combining the factors gives
Thus an antiderivative is \(\ln|u|\). To check it, \(du/dx=1+x/\sqrt{x^2-1}=u/\sqrt{x^2-1}\), and \(u\ne0\) on the interval. Therefore
The negative interval requires its own branch-aware setup; the positive inverse \(x=\sqrt{1+t^2}\) cannot be used there.
A Practical Pitfall and Its Correction
Wrong: “The differential contains a square root, so this substitution cannot rationalize the integral.” That judgment examines \(dx\) by itself. For the integral in the cancellation example, it would reject \(t=\sqrt{x^2-1}\), even though the full expression simplifies to \(t^2/(t^2+1)\,dt\).
Correction: Substitute into every factor, include the differential, and simplify their product. Then decide whether the resulting coefficient of \(dt\) is rational. Separately verify that the parameter and inverse formulas are valid on the interval in question. These checks distinguish a genuinely failed setup from one whose apparent radicals cancel, and from one that works only on a different branch.
Rationalization is valuable because it converts an algebraic integral into an integral of rational functions, where algebraic integration methods can be applied. Diagnosing the setup before proceeding can save work: a residual radical or a branch mismatch is usually easier to correct at the substitution stage than after a long calculation.
Check Your Understanding
For each question, focus on the complete transformed integrand and the interval on which the substitution is intended to apply.
- What expression should be checked to determine whether \(x=\phi(t)\) rationalizes \(\int f(x)\,dx\)?
- Why does a radical in \(dx/dt\) alone not prove that a substitution has failed?
- For \(t=\sqrt{x^2-1}\), which inverse branch applies when \(x<-1\)?
- In the nested-radical example, what choice of parameter avoids the nested expression?
- What should be checked after algebraically solving a substitution equation for \(x\)?
