What “Non-Elementary” Means
An antiderivative problem asks for a function whose derivative is a given integrand. Many familiar integrands have antiderivatives expressible with a finite combination of standard functions. But this does not happen for every integrand. The important distinction is between having no antiderivative in the elementary-function class and having no antiderivative at all.
Elementary functions include algebraic functions, exponentials, logarithms, trigonometric and inverse trigonometric functions, and functions formed from these using finitely many arithmetic operations and compositions, wherever the expressions are defined. An elementary antiderivative of \(f\) on an interval \(I\) is an elementary function \(F\) defined on \(I\) such that \(F'(x)=f(x)\) for every \(x\in I\).
The phrase is about the available form of an antiderivative, not whether the integrand is defined, whether an integral exists, or whether a numerical approximation can be found. Keeping those questions separate prevents a common misunderstanding.
An Elementary Antiderivative for Comparison
Worked Example: An Exponential Composite with an Elementary Antiderivative
Consider \(f(x)=2x e^{x^2}\). Since the derivative of \(x^2\) is \(2x\), use \(u=x^2\), so \(du=2x\,dx\). Then
Check the proposed antiderivative by differentiating:
The derivative agrees with the integrand, so \(e^{x^2}+C\) is an elementary antiderivative. The substitution works here because the factor \(2x\), the derivative of the inner function, is present.
By contrast, \(e^{x^2}\) by itself does not have an elementary antiderivative. This is a statement about the entire elementary-function class, not merely a claim that a familiar substitution did not work. A standard theorem about elementary integration can establish the distinction.
A Criterion for One Non-Elementary Antiderivative
A special case of Liouville’s theorem gives a useful test. For a polynomial \(p(x)\), an elementary antiderivative of \(e^{p(x)}\) can exist only if there is a rational function \(R(x)\) satisfying \(R'(x)+p'(x)R(x)=1\). In this setting, this condition is also sufficient: if such an \(R\) exists, then differentiating \(R(x)e^{p(x)}\) gives \(e^{p(x)}\). The criterion reduces an elementary-antiderivative question to an equation for a rational function.
For \(p(x)=x^2\), the required rational-function equation is \(R'+2xR=1\). We can show directly that no rational function satisfies it. If \(R\) had a finite pole of order \(m\), then \(R'\) would have a pole of order \(m+1\), while \(2xR\) would have a pole of order at most \(m\). The highest-order pole could not cancel, contradicting the fact that the right-hand side is the constant \(1\). Thus \(R\) has no finite poles. A rational function with no finite poles is a polynomial.
If \(R\) is a nonzero polynomial of degree \(n\), then \(2xR\) has degree \(n+1\), while \(R'\) has degree \(n-1\). Their sum therefore has degree \(n+1\), so it cannot equal the constant \(1\). The zero polynomial also fails, since \(0'+2x\cdot0=0\). Hence there is no rational solution, and the criterion shows that \(e^{x^2}\) has no elementary antiderivative.
Worked Example: Testing Whether \(e^{x^2}\) Has an Elementary Antiderivative
Apply the criterion with \(p(x)=x^2\). First compute \(p'(x)=2x\), so a rational candidate \(R\) would have to satisfy
Assume \(R\) has a finite pole of order \(m\). Then \(R'\) has a pole of order \(m+1\), and \(2xR\) has a pole of order at most \(m\). Their sum still has a pole, so it cannot equal \(1\). Therefore \(R\) has no finite poles and must be a polynomial.
For a nonzero polynomial of degree \(n\), the term \(2xR\) has degree \(n+1\), while \(R'\) has degree \(n-1\). The degree-\((n+1)\) term cannot cancel, so the sum cannot be \(1\). If \(R=0\), the left side is \(0\), not \(1\). There is no rational solution; by the stated criterion, \(\int e^{x^2}\,dx\) has no elementary antiderivative. This conclusion concerns elementary forms, not the existence of an antiderivative by other means.
Exact Integral Notation Still Works
Suppose \(f\) is continuous on an interval containing \(a\) and \(x\). The Fundamental Theorem of Calculus gives an antiderivative defined by an integral:
The integral defines \(F\) exactly, even when there is no elementary formula for it. The variable \(t\) is a dummy variable of integration; the resulting function depends on \(x\). Thus “no elementary antiderivative” does not mean “no antiderivative.”
Worked Example: Defining an Antiderivative Exactly
Let \(f(x)=e^{x^2}\) and define
The integrand is continuous for every real \(t\), so the Fundamental Theorem of Calculus applies. Differentiating the function defined by the integral gives
Thus \(F\) is an antiderivative of \(e^{x^2}\), checked directly by differentiation. The preceding criterion shows that \(F\) is not elementary. The integral definition is nevertheless exact: it specifies one particular antiderivative, with \(F(0)=0\). Any other antiderivative on the same interval differs from \(F\) by a constant.
A definite integral can likewise be an exact value without being expressible using elementary functions. For a continuous function on a finite closed interval, the definite integral exists as a finite real number. Its integral notation is an exact specification of that number; it is not a claim that the number can be rewritten in a simpler elementary form.
Worked Example: An Exact Definite Integral and Bounds
Consider \(I=\int_0^1 e^{x^2}\,dx\). The integrand is continuous on \([0,1]\), so this definite integral exists. For \(0\leq x\leq1\), we have \(0\leq x^2\leq1\). Since the exponential function is increasing,
Integrating these inequalities over an interval of length \(1\) gives
In fact, the inequalities are strict for the integral: \(e^{x^2}>1\) for \(0<x\leq1\), and \(e^{x^2}<e\) for \(0\leq x<1\). Therefore \(1<I<e\). The notation \(\int_0^1 e^{x^2}\,dx\) specifies the exact value, and the inequalities check that it is finite and lies in the stated range. Neither the exact notation nor the bounds assert an elementary antiderivative.
Do Not Confuse Non-Elementary with Undefined
For a definite integral, ask first whether the integrand is defined on the interval and whether the integral exists. Only then ask whether an elementary antiderivative is available. A continuous integrand on a finite closed interval gives an ordinary definite integral, even if its antiderivative is non-elementary. A singularity at an endpoint or inside the interval instead requires an improper-integral analysis; an elementary antiderivative on either side of a singularity does not guarantee convergence.
Worked Example: A Singular Integral Is Not a Non-Elementary Antiderivative Problem
The function \(1/x\) is not defined at \(x=0\), so \(\int_{-1}^1 1/x\,dx\) is not an ordinary definite integral. Its improper interpretation requires both one-sided integrals to converge separately:
Both one-sided integrals fail to converge to finite values, so the improper integral does not exist. Although \(\ln|x|\) is an elementary antiderivative on each side of zero, it does not make the integral across zero converge. The symmetric Cauchy principal value is \(0\), but that is a different limiting procedure and does not change the divergence of the improper integral.
Why the Distinction Matters
When an elementary antiderivative cannot be found, the integral has not become meaningless or useless. An exact integral can define a function or number; continuity can establish existence; and bounds or numerical methods can provide useful information. Conversely, writing an integral symbol does not by itself establish that the integral exists, especially when the integrand has a singularity or the interval is unbounded.
Before deciding what an integral means, separate these questions: Is the integrand defined on the relevant domain? Does the ordinary or improper integral exist? Is an elementary antiderivative available? An answer to one does not automatically answer the others.
Check Your Understanding
Use the distinctions and results in this tutorial to answer the following questions.
- What does it mean for an antiderivative to be non-elementary?
- Why does \(F(x)=\int_0^x e^{t^2}\,dt\) still define an antiderivative of \(e^{x^2}\)?
- What rational-function equation does the stated Liouville criterion require for \(e^{x^2}\)?
- What must be checked before treating an integral with a singularity as an improper integral?
- Why does a Cauchy principal value of \(0\) not establish convergence of \(\int_{-1}^1 1/x\,dx\)?
