Summary
This lesson reverses the question of Calculus I: given a function , find the functions whose derivative is , called antiderivatives of . The foundation is a single corollary, the one on functions with the same derivative: on an interval, two antiderivatives of the same function differ by a constant, so that finding one is enough to have them all. On a domain that is not an interval, such as , there is an independent constant for each interval. The family of antiderivatives is written , and the lesson makes precise what that equals sign between families asserts. Each formula of the table of basic antiderivatives is proved by differentiating, together with the interval on which it holds, and linearity allows them to be combined. Finally, Darboux’s theorem shows that not every function has an antiderivative: the sign function has none on .
Learning objectives
- Define the antiderivative of a function on an interval and prove that two antiderivatives on an interval differ by a constant, and that on a union of disjoint open intervals there is one constant for each interval.
- Write and read the indefinite integral as the family of antiderivatives on an interval, and interpret the equals sign between two families.
- Prove, by differentiating, linearity and the table of basic antiderivatives, with the interval on which each formula holds, and use them to find antiderivatives that satisfy given conditions.
- Recognize, with Darboux’s theorem, functions that have no antiderivative, and distinguish the existence of an antiderivative from the possibility of writing it with elementary functions.
The inverse problem of differentiation
Calculus I solved one problem: given a function, find its derivative. The differentiation rules solve it whenever the function is built from the elementary functions by operations and compositions. This unit studies the inverse problem: given a function , find a function whose derivative is . If the slope of the graph of is known at each point, what is the graph? If the rate of change of a quantity is known at each instant, what is the quantity?
For , the result comes from memory; but and also have derivative , because a constant has zero derivative. Two questions arise: whether there are other functions that work besides those of the form , and whether every function has one. The first is settled precisely, in a way that depends on the set on which one works; the second, not always in the affirmative.
Let be an interval with more than one point and a function defined on . A function is an antiderivative of on if it is differentiable at each point of and its derivative at each point is the value of :
If contains one of its endpoints, the derivative of at that endpoint is, by the proposition on the derivative and one-sided derivatives, from the lesson on the derivative as a limit, the corresponding one-sided derivative. A function defined on a set that contains is an antiderivative of on if its restriction to is one.
By the lemma on the local character of the derivative, from the lesson on differentiation rules, if a function is differentiable at a point of , its restriction to is too, with the same derivative. Thus is an antiderivative of on and on any interval. Every antiderivative is continuous on , by the theorem that differentiable implies continuous. And the definition requires the equality also at the endpoints that belong to : the function is an antiderivative of on , because at each its derivative is , by the derivative of a power with a real exponent, and at its difference quotient, , tends to from the right.
Two antiderivatives differ by a constant
The theorem below settles the first question: on an interval, adding constants is the only freedom. The whole unit rests on it, because it reduces the search for all antiderivatives to the search for a single one.
Let be an antiderivative of on an interval . A function is an antiderivative of on if and only if there is a number such that for every in :
Adding a constant does not change the derivative; and the difference of two antiderivatives has zero derivative on an interval, so it is constant. The figure shows, with , an antiderivative of , some of the functions , and an antiderivative whose vertical distance to is the same at all points.
Proof
If , the constant function has zero derivative at every point, by the corollary on the derivative of a polynomial, from the lesson on differentiation rules. By the theorem on the derivative of a linear combination, from the same lesson, which only requires the point to be an accumulation point of the domain, is differentiable at each point of , including the endpoints that belong to it.
Conversely, let be an antiderivative of on . By the theorem that differentiable implies continuous, from the lesson on the derivative as a limit, and are continuous on , and at each interior point of they have the same derivative, . By the corollary on functions with the same derivative, from the lesson on the theorems of Fermat, Rolle, Lagrange and Cauchy, there is a number such that:
Geometrically, the graphs of the antiderivatives of on are the vertical translations of one of them: at each abscissa they have the same slope, , and none comes closer to another or moves away from it. To choose one, it suffices to fix its height at a point.
Let be an antiderivative of on an interval , a point of and a real number. There is a unique antiderivative of on such that , and it is . The condition at determines the constant.
Proof
By the theorem, the antiderivatives of on are the functions , and the condition at holds for a single value of . That value gives an antiderivative, by the same theorem, and no other satisfies the condition.
For example, the only antiderivative of on that takes the value at is . If the rate of change of a quantity is known at each instant of a time interval, together with its value at one instant, the quantity is determined on the whole interval.
Domains that are not intervals
The hypothesis that be an interval cannot be omitted. The lesson on the theorems of Fermat, Rolle, Lagrange and Cauchy showed this with , which has zero derivative on and equals on the positive numbers and on the negative ones. In terms of antiderivatives: and have the same derivative, , on , and they do not differ by a constant, but by a different constant on each ray. The domains of the elementary functions are usually unions of disjoint open intervals, and on them the rule is as follows.
Let be the union of a family of open intervals , pairwise disjoint, let be a function defined on and let be a function differentiable at each point of , with . A function defined on is differentiable at each point of , with , if and only if for each there is a number such that for every in . The constants are arbitrary and independent of one another. On each interval the theorem applies; since the intervals are open and disjoint, the derivative at a point only looks at the interval that contains it. The figure shows this with , which has derivative on .
Proof
Suppose that on , and let be one of the intervals. Every point of is an accumulation point of ; by the lemma on the local character of the derivative, the restrictions of and of to are differentiable at each point of , with derivative . They are antiderivatives of on , and the theorem on antiderivatives on an interval gives a constant for that interval.
Conversely, let be arbitrary numbers, one per interval, and let on each . Let be a point of , which lies in a single . Since is open, there is such that is contained in ; on that interval, coincides with , which is differentiable at with derivative . By the second statement of the lemma on the local character, is differentiable at with the same derivative, whatever the other constants may be.
By the theorem on the derivative of the natural logarithm, from the lesson on the elementary functions, is differentiable on with derivative . By the proposition, the functions differentiable on with derivative are exactly those of the form:
with and arbitrary. The expression , with a single constant, does not describe them all: it leaves out, for example, the function that equals on the negative numbers and on the positive ones.
The indefinite integral
The theorem on antiderivatives on an interval makes it possible to name the set of all antiderivatives by means of a single one of them. Leibniz’s notation for that set is as follows.
Let be a function that has an antiderivative on an interval . The indefinite integral of on is the set of all antiderivatives of on , which by the theorem on antiderivatives on an interval is . It is written:
The sign is the integral sign; the function , the integrand; , the variable of integration; and , the constant of integration, which ranges over the real numbers.
The equality in the definition is not an equality between numbers or between functions, but between sets of functions: it asserts that the antiderivatives of on are exactly the functions . By the theorem, writing is correct if and only if is also an antiderivative of on , that is, if and only if is constant on . That is why two results of different appearance can both be correct. For example, on , as seen in the lesson on the theorems of Fermat, Rolle, Lagrange and Cauchy; the functions and differ by the constant , and on :
The letter does not denote the same number on the two sides: each side is a set, and the two sets are equal. From , as an equality of families, it does not follow that , but only that is constant. To check a result it suffices to differentiate it: if its derivative is the integrand on all of , the result is correct.
The variable of integration is a dummy variable: is the same set of functions as . The symbol indicates which variable is meant; the lesson on integration by change of variable will give it an operational use. The name integral and the sign , an elongated , allude to a sum: the definite integral of unit 3, which is defined with sums, will be connected with the indefinite integral in the fundamental theorem of calculus. This unit does not use the definite integral: each antiderivative is obtained and checked by differentiating.
When the integrand is defined on a union of intervals, this unit adopts the following convention: the formula is written once, the intervals on which it holds are indicated, and an independent constant is understood for each of them. Thus the equality
holds on and on , with constants and that may be different. An antiderivative given without its interval is an incomplete result.
Linearity
The derivative of a linear combination is the linear combination of the derivatives. Read backward, that rule gives the first tool for computing antiderivatives.
Let and be antiderivatives of and of on an interval , and let and be two real numbers. Then is an antiderivative of on . If and are not both zero, as an equality of families:
where the right-hand side is the set of the functions , with an antiderivative of and an antiderivative of on . The linear combination is differentiated; then one counts which constants the right-hand side produces.
Proof
By the theorem on the derivative of a linear combination, applied first to (with the zero function as the second summand) and then to , at each point of :
By the theorem on antiderivatives on an interval, and . As and range over the reals, the number ranges over all the reals if or ; then the right-hand side is the family of the functions , which is the left-hand side.
If , the left-hand side is the set of all constants and the right-hand side contains only the zero function: the equality of families fails, although the statement about antiderivatives still holds. There is, on the other hand, no product rule for antiderivatives: is an antiderivative of , but the product of the antiderivatives and has derivative . Products will be handled with integration by parts, which reverses the product rule, and compositions with the change of variable, which reverses the chain rule.
Basic antiderivatives
Each known derivative, read backward, gives an antiderivative. The three propositions below gather those that will be used throughout the unit; each is proved by differentiating the right-hand side, and the interval on which it holds is included in the statement.
For every real number other than and on the intervals indicated afterward:
The first holds on for every ; on if is a nonnegative integer, with ; and on each of the intervals and if is an integer less than . The second holds on each of the intervals and . The exponent is not zero, and the derivative of brings that exponent down, where it cancels with the denominator; the exponent , which that formula excludes, is covered by the logarithm.
Proof
Let . By the theorem on the derivative of a power with a real exponent, from the lesson on the variable in the exponent, with the exponent :
If is a nonnegative integer, is a positive integer, and if is an integer less than , is a negative integer. By the proposition on powers with an integer exponent, from the lesson on differentiation rules, is differentiable on its whole domain, in the first case and in the second, with the same formula; in the second case, the restriction to each ray is an antiderivative.
By the theorem on the derivative of the natural logarithm, is differentiable on with derivative , and its restriction to each ray is an antiderivative of .
Powers with a rational exponent whose denominator is odd, such as , are defined on all of , and the proposition only gives their antiderivative on ; the proposed problem “The cube root on the whole real line” extends it to the whole real line. On , the formula for the logarithm is the limiting case of the one for powers in a precise sense: the antiderivative , which vanishes at , tends to as tends to , by the corollary on the standard limits of the logarithm and of powers, from the lesson on the continuity of the elementary functions, with and the exponent in place of the variable.
Let , with . On :
On each interval , with an integer, the first two equalities below, and on each interval , the last two:
And on :
Each formula is a derivative from Calculus I read backward.
Proof
By the theorem on the derivative of the natural exponential, from the lesson on the elementary functions, and the one on the derivatives of the exponential and of the logarithm to base , from the lesson on the variable in the exponent, since :
By the theorem on the derivatives of sine and cosine, from the lesson on the elementary functions, and the one on the derivative of a linear combination:
By the proposition on the derivatives of the other trigonometric functions, from the same lesson, the tangent and the secant are differentiable where , and the cotangent and the cosecant where . Those domains are the unions of the intervals in the statement, and the restriction to each interval is an antiderivative.
From the same proposition, the derivatives of the secant and of the cosecant; and by the one on the derivatives of the inverse trigonometric functions, from the lesson on the derivative of the inverse function, the arctangent is differentiable on and the arcsine on .
The arcsine is continuous on , but it is not differentiable at or at , and the integrand is not defined there: the interval of the last formula is open. The antiderivatives of and of are not basic, because no derivative in the list is the tangent or the secant; they will be obtained in the lesson on the change of variable and in the one on trigonometric integrals.
On :
On each of the intervals and the first equality below, and on each of the intervals , and the second:
On , the first antiderivative is , and on , the second is . The logarithmic forms of the inverse hyperbolic functions also hold outside their domains, with the absolute value; they are checked by differentiating with the corollary on compositions with the logarithm.
Proof
By the proposition on the derivatives of the hyperbolic functions, from the lesson on the elementary functions, and the one on the derivatives of the inverse hyperbolic functions, from the lesson on the derivative of the inverse function; the logarithmic form is the proposition on the inverse hyperbolic sine, from the lesson on the hyperbolic functions, in Algebra and Geometry.
Let and . Since , the function does not vanish. By the corollary on standard compositions, from the lesson on the chain rule, and the one on compositions with the exponential and the logarithm, from the lesson on the variable in the exponent:
Let . By the law of the logarithm of a quotient, , and by the same corollary:
On and on , the antiderivatives coincide with the logarithmic forms of the propositions on the inverse hyperbolic cosine and on the inverse hyperbolic tangent, from the lesson on the hyperbolic functions; on , is positive, and on , so is .
On , the antiderivative is , because is the reciprocal of ; the logarithmic form gathers into a single formula what the inverse hyperbolic functions would express with two. The formulas for and for differ only in a sign, as do their antiderivatives and , and the same holds for those of and inside the root.
Functions without an antiderivative
Every derivative on an open interval has the intermediate value property, by Darboux’s theorem, from the lesson on monotonicity and the first derivative test, even if it is not continuous. A function without that property cannot be the derivative of anything.
Let be the sign function: if , and if . There is no function differentiable on whose derivative is . On the other hand, has antiderivatives on and on , the restrictions of plus a constant. An antiderivative on would have a derivative that passes from to without taking the value .
Proof
Suppose that is differentiable on , with for every . Then and . Since is an open interval, Darboux’s theorem, with , gives a in with ; but only takes the values , and .
On , , with derivative ; on , , with derivative . By the theorem on antiderivatives on an interval, these are all the antiderivatives, up to a constant on each ray.
The value assigned at has no influence: if were any other number , the same argument, with a in different from , would show that there is no antiderivative. The lesson on monotonicity reached the same conclusion for the Heaviside function, in its solved problem “A jump that is not a derivative”, and its corollary on the discontinuities of a derivative generalizes it.
Let be a function defined on an open interval that has, at a point of , a removable or jump discontinuity. Then has no antiderivative on . A derivative can only have essential discontinuities.
Proof
If were an antiderivative of on , would be the derivative of a function differentiable at each point of . By the corollary on the discontinuities of a derivative, from the lesson on monotonicity and the first derivative test, every one-sided limit of at would be equal to , and if both exist, is continuous at : would not have at a removable or jump discontinuity, contrary to the hypothesis.
Continuity is not necessary for having an antiderivative: the solved problem “Changing a single value” shows a discontinuous function that has one. It is, however, sufficient: every function continuous on an interval has an antiderivative on it. That result will not be used in this unit, because its proof requires the definite integral; it is the fundamental theorem of calculus, from unit 3. Until then, each antiderivative is exhibited and checked by differentiating it, and each technique of this unit is a theorem that is proved by differentiating its result.
That an antiderivative exists does not mean that it can be written with the elementary functions. The function is continuous on , and it will have an antiderivative by the fundamental theorem; but none of its antiderivatives is obtained from polynomials, exponentials, logarithms, the trigonometric functions and their inverses by finitely many operations and compositions. The same happens with and with . It is a fact that is proved in later courses, with the theory that Joseph Liouville initiated in the nineteenth century, and that this unit does not use. The techniques of the lessons that follow find elementary antiderivatives, when they exist, for broad families of functions, not for all.
Solved problems
The problems increase in difficulty: the first rewrites two integrands as combinations from the table; the second compares three antiderivatives of the same function and derives an identity from them; the third determines an antiderivative on a split domain with one condition on each interval; the fourth constructs a piecewise-defined antiderivative; and the last shows a discontinuous function with an antiderivative, which loses it when a single value is changed.
Compute and , indicating on which interval each result holds.
Solution
The first integrand is defined on . Expanding the square and dividing each term by leaves a linear combination of powers.
By linearity and the proposition on antiderivatives of powers, with and , on , where :
The second integrand is defined on , and subtracting and adding in the numerator splits it into a constant and a multiple of , whose antiderivative is the arctangent.
Neither of the two integrands was in the table, nor was a new technique needed: algebra was enough. It is always worth looking at the integrand before searching for a technique.
Prove that , and are antiderivatives of on , find the constants by which they differ and deduce a trigonometric identity.
Solution
By the corollary on standard compositions, from the lesson on the chain rule, and the double-angle formula for sine:
By the solved problem “A linear change of variable” from the lesson on the elementary functions, the derivative of is :
By the theorem on antiderivatives on an interval, each difference is constant on , and its value is computed at .
The first constant is the Pythagorean identity. The second gives, on solving, the power-reducing formula that the lesson on the sum and difference formulas obtained from the cosine of the double angle:
The three results for are correct and of different appearance: as an equality of families, . Comparing two antiderivatives on an interval is also a method for proving identities: it suffices that they have the same derivative and that they coincide at one point.
Find the function , differentiable on , such that , and . Can it be written in the form , with a single formula on the whole domain and a single constant?
Solution
Expanding the numerator, the integrand is a combination of powers. By linearity and the proposition on antiderivatives of powers, an antiderivative on each of the two rays is:
The domain is the union of the open rays, and by the proposition on antiderivatives on a union of disjoint open intervals there is one constant on each of them. The condition at fixes the one on the right, and the condition at , the one on the left.
The constants are different, so the function found is not of the form with a single constant. With only one of the two conditions, the other constant would remain free, and there would be infinitely many such functions.
On a split domain, the number of conditions needed is the number of intervals: each condition gives information about the interval in which its point lies, and not about the others.
Let if and if . Find all the antiderivatives of on and, among them, the one that vanishes at .
Solution
Let be an antiderivative of on with . Its restriction to is an antiderivative of on that interval, where equals , even at the endpoints. By the corollary on the antiderivative taking a given value:
On , , and the restriction of is an antiderivative of ; by the theorem on antiderivatives on an interval, there. The point belongs to both intervals, and fixes the constant; on , in the same way, with .
It remains to check that the function so constructed is differentiable at . Let and . By the proposition on the derivative at a junction point, from the lesson on differentiation rules, it suffices that the two formulas give at the same value and the same derivative; at the reasoning is the same with , which equals and has derivative .
By the theorem on antiderivatives on an interval, the antiderivatives of on are the functions : a single constant, because is an interval, even though the formula has three pieces. The constants of the pieces are not free: they are tied by the continuity of the antiderivative at the junction points. The antiderivative that vanishes at is odd, and is even; the proposed problem “Antiderivatives of even and odd functions” explains why.
Let if , and . Prove that has an antiderivative on , although it is discontinuous at . Let be the function that coincides with outside and satisfies . Prove that has no antiderivative on .
Solution
Let if , and . The solved problem “A derivative that has no limit” from the lesson on the elementary functions computed its derivative outside and showed that it has no limit at ; at , the derivative is zero, by the solved problem “Continuous and without one-sided derivatives” from the lesson on the derivative as a limit.
Then on all of , and is an antiderivative of . The discontinuity of at is essential, as the corollary on the discontinuities of a derivative requires.
Suppose that is differentiable on , with . The difference is differentiable on , and its derivative is : zero outside , and at .
By Darboux’s theorem, in there would be a point with , because lies between and . But is zero on . There is no .
Having an antiderivative does not depend on continuity, but on something more fragile: a single changed value is enough to lose it. The function differs from a derivative at only one point, and the difference, which equals at and at the other points, has a removable discontinuity, which no derivative can have.
Proposed problems
The problems increase in difficulty: the first extends the antiderivative of the cube root to the whole real line; the second shows that the sign function has no antiderivative on a closed ray either, where Darboux’s theorem does not apply; the third constructs a periodic antiderivative with infinitely many intervals; the fourth relates the parity of a function to that of its antiderivatives; and the last finds the function that comes closest to an antiderivative of the integer part. Each one has a hint, which is best opened only after having attempted it.
The proposition on antiderivatives of powers gives only on . Prove that is an antiderivative of on all of , and write the indefinite integral of on .
Hint
For , differentiate the product with the proposition on the derivative of the nth root, from the lesson on the derivative of the inverse function, and simplify with . At , the root is not differentiable and the product rule does not apply: compute the limit of the difference quotient .
Answer
Let be the sign function. Prove that has no antiderivative on , although it has one on and although Darboux’s theorem does not apply on an interval that contains one of its endpoints. What if the value of at is changed to ?
Hint
If were an antiderivative of on , it would be continuous on , with on . Apply the corollary on the limit of the derivative, from the lesson on the theorems of Fermat, Rolle, Lagrange and Cauchy, and compare with .
Answer
It has none: an antiderivative would satisfy . With the value at , the function is constant and equal to on , and its antiderivatives there are .
Find the function , differentiable at each with , such that and for every integer . Prove that on its whole domain, although is not periodic.
Hint
Use and linearity. The domain is the union of the intervals , and lies in the one with index : there is one constant per interval, and each condition fixes one. For the periodicity, observe that lies in the interval with index if lies in the one with index .
Answer
Let be an interval symmetric about (if is in , so is ), with more than one point, and let be a function with an antiderivative on . Prove that if is odd, every antiderivative of on is even, and that if is even, the antiderivative that vanishes at is odd. Is every antiderivative of an even function odd?
Hint
Differentiate with the solved problem “A linear change of variable” from the lesson on the elementary functions, with and , and apply the theorem on antiderivatives on an interval to , or to , and to . The constant is computed at , which is in .
Answer
If is odd, for every antiderivative ; if is even and , . Not every antiderivative of an even function is odd: is an antiderivative of and is not odd.
Worked solution
Solution
Let be odd and let , defined on . By the linear change of variable, is differentiable at each point of , and since is odd, is an antiderivative of on .
By the theorem on antiderivatives on an interval, ; at , which is in , , so and is even.
Let be even, with , and let . Now , so , and at one obtains .
Prove that the integer part, , has no antiderivative on any open interval that contains an integer. Find the function , continuous on and differentiable at each non-integer , such that for every non-integer and .
Hint
At each integer, the integer part has a jump: apply the corollary on functions with a jump. On each interval , is continuous and has derivative at the interior points; the corollary on functions with the same derivative gives there. Continuity at the integers ties each constant to the one before it, starting from .
Answer
Completed
Sign in to save your progress