Skip to content
Topos Uranos

Antiderivatives and integration techniques: introduction

Summary

This lesson opens the first unit of the course, which travels the road of Calculus I in reverse: given a function, find the functions whose derivative it is. It presents the route of the unit, whose foundation is a single corollary of Calculus I, the one on functions with the same derivative, and whose techniques are so many theorems proved by differentiating: the change of variable inverts the chain rule, integration by parts inverts the product rule, and the families of integrands (trigonometric, with roots of degree two, rational, rationalizable) are reduced to them. It reconstructs the history of these techniques, from Newton’s fluents and Leibniz’s sign to the work of Ostrogradsky, Chebyshev and Hermite in the nineteenth century; it states the problem that runs through the unit, that of writing an antiderivative with known functions, and shows with an example that differentiation does not leave the rational functions while integration does. It ends by indicating the place of the unit in the course and where antiderivatives are used in other courses of the site and in science and technology.

Learning objectives

  1. Describe the route of the unit and explain what each lesson proves and which result of Calculus I or of Algebra and Geometry it relies on.
  2. Explain why an antiderivative is determined only up to a constant on each interval, why not every function has an antiderivative and why integrating forces one to leave the family of functions of the integrand.
  3. Place the techniques of the unit in their historical context, from Newton and Leibniz to Ostrogradsky, Chebyshev and Hermite, and distinguish the order in which they were discovered from the order in which they are proved.
  4. Indicate the place of the unit in the course and where antiderivatives come into play in other courses and in scientific and technical applications.

Entry assessment

Before working through the unit, it is advisable to measure the starting point. This self-assessment is not graded: it serves to know where to pause. The first seven questions cover, in order, the seven content lessons of the unit, one each, and measure a skill (choosing a technique, fixing the interval on which an antiderivative holds, recognizing a typical error), not a statement to be recalled; the same skills are measured again at the end, in the synthesis lesson, with other data. The last five check what this unit takes as known from Calculus I and from Algebra and Geometry: the zero derivative on an interval, the chain rule, the derivative of an inverse function, partial fractions and the tangent half-angle substitution.

Entry assessment

Before starting the unit, measure your starting point. It is a short self-assessment: it is not graded and its result is stored only in your browser.

  1. Question 1A function FF is differentiable on R∖{0}\mathbb{R} \setminus \{0\}, with F′(x)=2x3F'(x) = \frac{2}{x^3} for every x≠0x \neq 0, and F(1)=1F(1) = 1. What can be asserted about F(−1)F(-1)?

  2. Question 2Compute ∫x1+x4 dx\int \frac{x}{1 + x^4} \, \mathrm{d}x on R\mathbb{R}.

  3. Question 3Compute ∫x2ln⁡x dx\int x^2 \ln x \, \mathrm{d}x on (0,∞)\ioo{0}{\infty}.

  4. Question 4Compute ∫sin⁡2xcos⁡2x dx\int \sin^2 x \cos^2 x \, \mathrm{d}x on R\mathbb{R}.

  5. Question 5Which of these functions is an antiderivative of 1x2x2−4\frac{1}{x^2 \sqrt{x^2 - 4}} on (−∞,−2)\ioo{-\infty}{-2}?

  6. Question 6Compute ∫2x2+x+1x3+x dx\int \frac{2x^2 + x + 1}{x^3 + x} \, \mathrm{d}x, with the intervals on which it holds.

  7. Question 7Compute ∫cos⁡xsin⁡2x+4 dx\int \frac{\cos x}{\sin^2 x + 4} \, \mathrm{d}x.

  8. Question 8From earlier unitsWhat can be asserted about the function f(x)=arctan⁡x+arctan⁡1xf(x) = \arctan x + \arctan \frac{1}{x}, defined on R∖{0}\mathbb{R} \setminus \{0\}?

  9. Question 9From earlier unitsCompute the derivative of f(x)=ln⁡(x+x2+1)f(x) = \ln\left(x + \sqrt{x^2 + 1}\right) on R\mathbb{R}.

  10. Question 10From earlier unitsCompute the derivative of f(x)=arcsin⁡x2f(x) = \arcsin \frac{x}{2} on (−2,2)\ioo{-2}{2}.

  11. Question 11From earlier unitsWhat is the partial fraction decomposition of 5x−1x2−1\frac{5x - 1}{x^2 - 1}?

  12. Question 12From earlier unitsIf θ\theta is not an odd multiple of π\pi and t=tan⁡θ2t = \tan \frac{\theta}{2}, what is the value of cos⁡θ\cos \theta?

The contents of the unit

Calculus I proved all the differentiation rules: the sum rule and the product rule, the chain rule, the rule for the inverse function, and the derivatives of the elementary functions. Its last unit also left the corollary that sustains this one: two functions with the same derivative on an interval differ by a constant. This unit travels those rules in reverse. Each integration technique is a theorem whose demonstration consists in differentiating the result, and the demonstrations cite the rules of Calculus I instead of redoing them. What Algebra and Geometry left admitted about partial fractions with quadratic factors is proved here. Throughout the unit, an antiderivative is given with the interval on which it holds, and the definite integral, with its areas and its sums, does not yet exist. The lessons follow this order:

  1. Antiderivatives and the indefinite integral define the antiderivative of a function on an interval and prove, with the corollary on functions with the same derivative, that two antiderivatives differ by a constant; on a domain that is not an interval, there is one constant for each interval. The lesson fixes the meaning of ∫f(x) dx=F(x)+C\int f(x) \, \mathrm{d}x = F(x) + C as an equality between families, proves by differentiating the table of basic antiderivatives and linearity, and shows with Darboux’s theorem that the sign function has no antiderivative on the real line.
  2. Integration by change of variable inverts the chain rule: if FF is an antiderivative of ff, then F(g(x))F\left(g(x)\right) is one of f(g(x))g′(x)f\left(g(x)\right) g'(x). The theorem justifies the notation du=g′(x) dx\mathrm{d}u = g'(x) \, \mathrm{d}x, gives the linear change of variable and the forms g′g\frac{g'}{g}, (g(x))rg′(x)\left(g(x)\right)^r g'(x) and eg(x)g′(x)e^{g(x)} g'(x), and is completed by the change of variable by the inverse function, x=h(t)x = h(t), which requires a derivative that has no zeros in order to return to the original variable.
  3. Integration by parts inverts the product rule and exchanges one problem for another; repeated, it gives the tabular method, which is proved by induction, cyclic integrals and reduction formulas, among them the one for the powers of 1a2+x2\frac{1}{a^2 + x^2}; with the mean value theorem, it also gives the antiderivative of an inverse function.
  4. Trigonometric integrals reduce to the change of variable and to the identities: an odd exponent is split off as the derivative of the other factor, and even exponents are lowered by power reduction. The lesson proves the antiderivative of the secant, with its constant on each interval, and treats the powers of the tangent and of the secant and the products of sines and cosines.
  5. Trigonometric and hyperbolic substitutions are cases of the change of variable by the inverse function that turn the roots a2−x2\sqrt{a^2 - x^2}, a2+x2\sqrt{a^2 + x^2} and x2−a2\sqrt{x^2 - a^2} into expressions without a root; the lesson chooses the interval of the new variable so that the substitution is invertible and the sign of the root known, compares the trigonometric version with the hyperbolic one and, by completing the square, reduces any root of a polynomial of degree two with nonzero discriminant to one of the three.
  6. Integration of rational functions proves Bézout’s identity for polynomials and, with it, the existence and uniqueness of the partial fraction decomposition, also with quadratic factors; from this it follows that every rational function has, on each interval without roots of the denominator, an antiderivative made up of a rational function, logarithms and arctangents. The fundamental theorem of algebra is used as admitted, as in Algebra and Geometry, and will be proved in the unit on sequences. The lesson closes with the Hermite-Ostrogradsky method, which obtains the rational part of the antiderivative without knowing the roots of the denominator.
  7. Integration by rationalization turns other families into rational functions of a new variable: the rational functions of sine and cosine, with the tangent half-angle substitution or, if parity allows it, with shorter substitutions, and with the joining of the antiderivatives of adjacent intervals; those that contain the root of a linear fractional function; those that contain the root of a polynomial of degree two, with the three Euler substitutions; and differential binomials, in Chebyshev’s three cases.

The unit ends with the lesson of synthesis and challenges, which gathers what was achieved into a deductive chain, repeats in its exit assessment the skills of the entry assessment and proposes ten problems of greater difficulty.

The historical background

The order of the unit is not that of history. Quadratures, the computation of areas enclosed by curves, had been practiced since Antiquity, with the method of exhaustion of Eudoxus and Archimedes; and when Newton and Leibniz recognized that the problem of quadratures is the inverse of the problem of tangents, integration became, for more than a century, the art of finding antiderivatives. This unit follows that second idea, which turns each differentiation rule into an integration technique; the integral as area and as limit of sums will come in unit 3, and the fundamental theorem of calculus will show there that the two ideas coincide.

Fluents and fluxions

Isaac Newton wrote in October 1666 a treatise on fluxions. He called fluents the quantities that flow with time, such as the coordinates xx and yy of a point moving along a curve, and fluxions the velocities with which they change, what today are their derivatives with respect to time. In that treatise he already poses the inverse problem: given the relation between xx and the ratio of the fluxions, which is the slope of the curve, find yy; and he solves it by running the rules of differentiation in reverse. His Method of Fluxions and Infinite Series, written in 1671, was only published in 1736, in an English translation. For Newton, then, finding a fluent from its fluxion was exactly what this unit calls finding an antiderivative.

The Leibniz sign

Gottfried Wilhelm Leibniz reached the same calculus by another road, that of sums of differences. In his manuscripts of October and November 1675 he replaced the abbreviation omn., from omnia (“all”), which he placed before the ordinate of a curve to indicate the sum of all of them, by an elongated s, the initial of summa: this is the sign ∫\int, which together with the d\mathrm{d} of differences gave the notation ∫f(x) dx\int f(x) \, \mathrm{d}x. Leibniz published his differential calculus in the Acta Eruditorum in 1684, and the sign ∫\int appeared in print for the first time in June 1686, in an article in the same journal on hidden geometry and the analysis of indivisibles and infinites. Leibniz called this branch of the calculus calculus summatorius; the name integral calculus was proposed by Jacob Bernoulli in 1690. The sign thus keeps the trace of the sum, although in this unit it denotes a family of antiderivatives: the fundamental theorem of unit 3 will justify using the same sign for both things.

Rational fractions

Johann Bernoulli, Jacob’s younger brother, understood integration simply as the inverse operation of differentiation, the same point of view as this unit. In 1702, he, in the memoirs of the Paris Academy of Sciences, and Leibniz, in the Acta Eruditorum, showed independently that a rational function is integrated by decomposing it into partial fractions, whose antiderivatives are logarithms and arctangents; in the language of the time, they reduce to the quadrature of the hyperbola and to that of the circle. The difficulty lay in factoring the denominator. Leibniz believed that x4+a4x^4 + a^4 could not be decomposed into two real factors of degree two, and concluded that its reciprocal required a quadrature of another kind; but it suffices to complete the square:

x4+a4=(x2+a2)2−2a2x2=(x2+ax2+a2)(x2−ax2+a2)x^4 + a^4 = \left(x^2 + a^2\right)^2 - 2a^2 x^2 = \left(x^2 + a x \sqrt{2} + a^2\right)\left(x^2 - a x \sqrt{2} + a^2\right)

The lesson on the integration of rational functions uses this factorization, and the guarantee that such a factorization always exists is given by the fundamental theorem of algebra, which is admitted in this unit.

Euler’s treatise

Leonhard Euler gathered the methods of the century in his Institutiones calculi integralis, published in St. Petersburg in three volumes, between 1768 and 1770. The work studies systematically which integrals can be expressed with the elementary functions, and also treats differential equations, which this course does not address. The substitutions that rationalize the functions of xx and of ax2+bx+c\sqrt{ax^2 + bx + c} bear his name today, and the lesson on rationalization proves them with the conditions that each one requires.

Which antiderivatives can be written down

The nineteenth century turned the techniques into theorems about what can and what cannot be integrated. Joseph Liouville began in his memoirs of 1833 the theory that makes it possible to prove that certain antiderivatives, such as those of e−x2e^{-x^2}, cannot be written with the elementary functions. Mikhail Ostrogradsky published in 1845, in the bulletin of the St. Petersburg Academy of Sciences, a method for obtaining the rational part of the antiderivative of a rational function without finding the roots of the denominator, and Charles Hermite returned to it in 1872, in an article on the integration of rational fractions published in the Nouvelles Annales de mathématiques; that is why the method bears both names. Pafnuty Chebyshev published in 1853, in Liouville’s Journal de mathématiques pures et appliquées, a study on the integration of irrational differentials, in which he proved that the antiderivative of a differential binomial xm(a+bxn)px^m \left(a + b x^n\right)^p, with rational exponents, is elementary only in three cases, those studied in the lesson on rationalization. The unit proves the constructive half of these results, that in each case the antiderivative exists and can be found; the negative half, that outside them it is not elementary, it mentions as a fact.

The problem that gave rise to them

Integrating is running differentiation in reverse, and the way back raises two questions that the way there did not raise. The first is whether an antiderivative exists and how many there are: the lesson on antiderivatives replies that on an interval there is a family differing by constants, and that there are functions, such as the sign function, that have none; that every continuous function has one is the fundamental theorem of calculus, in unit 3. The second is whether the antiderivative can be written with known functions. When differentiating there is no need to leave the family of the function: the derivative of a polynomial is a polynomial, and by the corollary on the derivative of a polynomial and of a rational function, from the lesson on differentiation rules, that of a rational function is rational. When integrating, on the other hand, there is such a need, and the example below proves it with the simplest rational integrand that has no rational antiderivative.

ExampleA rational function without a rational antiderivative

Consider f(x)=11+x2f(x) = \frac{1}{1 + x^2}, defined on R\mathbb{R}. The figure shows the function, its antiderivatives, their asymptotes and, at the end, a rational function with a finite limit at infinity.

Proof

  1. (arctan⁡x)′=11+x2,x∈R\left(\arctan x\right)' = \frac{1}{1 + x^2}, \qquad x \in \mathbb{R}

    The function ff is rational, and its denominator does not vanish. By the proposition on antiderivatives of the exponentials and of the trigonometric functions, from the lesson on antiderivatives, the arctangent is an antiderivative of ff on R\mathbb{R}.

  2. R(x)=arctan⁡x+C,x∈RR(x) = \arctan x + \dato{c}{C}, \qquad x \in \mathbb{R}

    Suppose that a rational function R=PQR = \frac{P}{Q}, with QQ without real roots, were an antiderivative of ff on R\mathbb{R}. By the theorem on the antiderivatives of a function on an interval, it differs from the arctangent by a constant.

  3. lim⁡x→∞R(x)=π2+C,lim⁡x→−∞R(x)=−π2+C\lim_{x \to \infty} R(x) = \frac{\pi}{2} + C, \qquad \lim_{x \to -\infty} R(x) = -\frac{\pi}{2} + C

    By the corollary on monotonic functions at infinity, from the lesson on limits as the variable tends to infinity, the arctangent tends to π2\frac{\pi}{2} at ∞\infty and to −π2-\frac{\pi}{2} at −∞-\infty. Therefore, RR has finite and distinct limits at both ends.

  4. π2+C=−π2+C→π=0\frac{\pi}{2} + C = -\frac{\pi}{2} + C \rightarrow \resaltar{\pi = 0}

    Now, by the theorem on rational functions at infinity, from the same lesson, if RR has a finite limit at ∞\infty, the degree of PP does not exceed that of QQ, and then the limit at −∞-\infty is the same, 00 or the quotient of the leading coefficients. The two equalities contradict each other: no antiderivative of ff on R\mathbb{R} is rational.

If this is so, integrating a rational function forces one to leave the rational functions; it follows that integration techniques cannot be mere computational rules, but must say in which family the antiderivative lies. Now, for rational functions the way out is short: the corollary on the antiderivatives of rational functions, from the lesson on their integration, proves that it suffices to add logarithms and arctangents, which is what Leibniz and Bernoulli had found. The other lessons of the unit reduce other families to that one: the change of variable and integration by parts transform one integrand into another; trigonometric integrals, the substitutions for roots and rationalization lead, in their cases, to a rational function of a new variable. Therefore, the unit settles the second question for broad families, not for all: outside Chebyshev’s cases, or for e−x2e^{-x^2}, the antiderivative exists, because the integrand is continuous, but it cannot be written with the elementary functions. Hence the fundamental theorem of calculus and the numerical methods of unit 4 are not a supplement to this unit, but the way to reach what it cannot.

The relevance of antiderivatives

In the course

This unit relies on Calculus I and on Algebra and Geometry, and not on the rest of the course. The following unit, on sequences, prepares the definite integral: the Riemann integral is defined with sums and their limits, and its existence for continuous functions is proved with uniform continuity, which that unit establishes with the Bolzano-Weierstrass theorem; the same unit proves the fundamental theorem of algebra that is admitted here. In unit 3, the fundamental theorem of calculus joins the two integrals: it proves that every continuous function on an interval has an antiderivative and that the definite integral is computed with any antiderivative, and it carries over to the definite integral the change of variable and integration by parts; the lesson on techniques for definite integrals adds parity, periodicity, the orthogonality of sines and cosines and the Wallis integrals. The areas, the volumes of solids and the arc length of unit 4 call for antiderivatives from the families of this unit; the length of an arc of a parabola, for example, leads to the antiderivative of a2+x2\sqrt{a^2 + x^2}. The improper integrals of unit 5 and the gamma function are computed with antiderivatives and integration by parts, and in the last unit power series are integrated term by term.

In other courses

In Calculus III, double and triple integrals are computed as iterated integrals, each with an antiderivative in a single variable, and integrals over curves are reduced to integrals in one variable by a change of variable. In Foundations of Mechanics, the position of a moving body is an antiderivative of its velocity, and the velocity an antiderivative of its acceleration; the constant left free is fixed by the initial state, as in the corollary on the antiderivative taking a given value. In Financial Mathematics I, the accumulation factor of a capital with a variable instantaneous interest rate is the exponential of an antiderivative of that rate. In Data Analysis with R, the distribution function of a continuous variable is the antiderivative of its density that vanishes at the left end; the normal density is a multiple of e−x22e^{-\frac{x^2}{2}}, whose antiderivative is not elementary, and that is why its values are computed with numerical methods or looked up in tables. In Programming in C++ I bisection is programmed, and with the same tools those numerical methods can be programmed; comparison with an exact antiderivative, when there is one, measures the error.

In science and technology

Many quantities are known through their rate of change, and the quantity itself is an antiderivative of that rate: the amount of water in a tank from the flow rate coming in and going out, the population of a culture from its growth rate, the accumulated charge from the current. In all those cases, uniqueness up to a constant, from the lesson on antiderivatives, says that one datum, the value at one instant, suffices to fix the antiderivative. Computer algebra systems integrate rational functions with the Hermite-Ostrogradsky method, which avoids finding the roots of the denominator, and rely on the theory begun by Liouville to recognize, in many cases, that an antiderivative is not elementary; the computation then passes to numerical methods.