Corrected exercises: substitution, integration by parts and partial fractions (201-SN3-RE)
Here is a long set of 20 corrected exercises on the three techniques that make up most of an integral calculus course, as 201-SN3-RE (formerly 201-NYB) requires them: substitution (change of variable), integration by parts and partial fraction decomposition. The sets organized by chapter introduce them; this one exhausts them, with the cases textbooks dispatch in one line, such as the cover-up method, the repeated irreducible quadratic, the tabular method or the clever choice of the constant in v.
Work through it in four sessions of about 75 minutes, one per part. Part A, substitution: the five forms where the derivative is already present, the case where an x survives the substitution, rationalizing substitutions that turn an exponential or a root into a rational function, handling the limits of a definite integral with the monotonicity condition, and the three situations where substitution cannot work. Part B, integration by parts: the LIATE rule and why it works, repeated parts with the tabular method, the cyclic case and its general formula, the boundary term on a definite integral, and the invisible factor 1 beyond the logarithm. Part C, partial fractions: the four cases of decomposition, the cover-up method and its exact limits, long division of improper fractions, the irreducible quadratic that produces a logarithm and an arctangent, and the three settings where the decomposition really pays off. Part D, strategy.
The thread of the set fits in one question, asked before any other in front of an unfamiliar integral: is the factor next to the complicated expression its DERIVATIVE? If so, a substitution is enough. If not, check whether it is a product of different families, in which case it is integration by parts, or a quotient of polynomials, in which case it is partial fractions. The order in which you ask these questions matters as much as the techniques themselves.
Self-checking setType your answers and check them question by question: the page tells you right or wrong before you open the solution. With an account, every answer right on the first try earns points, and your right answers stay filled in when you come back.
•The five forms of direct substitution: ∫ff′=ln∣f∣; ∫f′fn=n+1fn+1; ∫f′ef=ef; ∫ff′=2f; ∫f′cosf=sinf. The test: differentiate the most complicated expression and look for the result elsewhere in the integrand.
•Substitution with a way back: if an x survives, express it in terms of u. Rationalizing substitutions: x=un in front of nx, x=um with m the LCM in front of several roots, u=ex in front of an exponential. In a definite integral, the substitution must be MONOTONIC on the interval.
•Parts: ∫udv=uv−∫vdu, and on a definite integral the boundary term [uv]ab is evaluated at BOTH limits. Priority order for u: LIATE, Logarithmic, Inverse trigonometric, Algebraic, Trigonometric, Exponential. Tabular method as soon as a polynomial factor eventually differentiates to zero.
•Cyclic case: ∫eaxsin(bx)dx=a2+b2eax(asinbx−bcosbx) and ∫eaxcos(bx)dx=a2+b2eax(acosbx+bsinbx). Keep the SAME choice of u in both rounds, otherwise the calculation goes back to where it started.
•Partial fractions: divide if the numerator reaches the degree of the denominator, factor, then one term per simple factor, as many terms as the multiplicity for a repeated factor, and a numerator Bx+C for each irreducible quadratic. The number of constants always equals the DEGREE of the denominator.
•Cover-up method: for the coefficient of x−a1, cover (x−a) and evaluate the rest at a. On a repeated factor it gives only the highest power, and nothing at all on an irreducible quadratic. Remember: ∫x2−a2dx=2a1lnx+ax−a, whereas ∫x2+a2dx=a1arctanax.
Part A: Substitution in all its forms (/50)
Exercise 1: The five forms where the derivative is already there
A substitution is not invented: it is SPOTTED. In almost every exam case, the integrand already contains, up to a constant factor, the derivative of the expression to substitute. Five forms keep coming back.
a) Compute ∫x3+2x+13x2+2dx and ∫2+sinxcosxdx. Which form do you recognize?
b) Compute ∫x(x2+1)5dx and ∫x(lnx)3dx.
c) Compute ∫esinxcosxdx and ∫sec2xetanxdx.
d) Compute ∫x2+4xdx and ∫x(1+x)2dx.
e) State the five forms in their general form, then say in one sentence the test that decides, in three seconds, whether a direct substitution will work.
Show the solution
Answers
a)ln∣x3+2x+1∣+C; ln(2+sinx)+C
b)12(x2+1)6+C; 4(lnx)4+C
c)esinx+C; etanx+C
d)x2+4+C; −1+x2+C
e)Is the derivative of the inside a factor?
a)
Form ∫ff′, whose antiderivative is ln∣f∣. The numerator 3x2+2 is exactly the derivative of x3+2x+1: the answer is lnx3+2x+1+C. Likewise cosx is the derivative of 2+sinx, hence ln(2+sinx)+C, the absolute value being unnecessary since 2+sinx stays positive.
b)
Form ∫f′fn, with antiderivative n+1fn+1. Here x is half the derivative of x2+1: 21∫u5du=12(x2+1)6+C. For the second, x1 is the derivative of lnx, hence 4(lnx)4+C.
c)
Form ∫f′ef, with antiderivative ef. Since cosx=(sinx)′: esinx+C. Since sec2x=(tanx)′: etanx+C. These two integrals look fearsome and are settled without writing a single line of substitution, provided you have the table of derivatives in your head.
d)
Form ∫ff′, with antiderivative 2f. Here x=21(x2+4)′, hence x2+4+C. For the second, 2x1 is the derivative of 1+x: the integral equals 2∫u2du with u=1+x, that is −1+x2+C.
e)
The five forms: ∫ff′=ln∣f∣; ∫f′fn=n+1fn+1 for n=−1; ∫f′ef=ef; ∫ff′=2f; and ∫f′cosf=sinf, with its trigonometric variants. The test fits in one question: mentally DIFFERENTIATE the most complicated expression in the integrand, the one inside a power, a logarithm, an exponential or a root, and look at whether the result appears elsewhere in the integrand, up to a constant factor. If it does, the substitution is immediate; if not, no direct substitution will work and you must change technique.
Tick the exercises you have done or want to review: a free account, no password, keeps your ticks and your right answers from one visit to the next, tells you which chapter to tackle next and lets you ask for the exercise you are missing. Create your space, an email is enough.
Exercise 2: When an x is left standing
Sometimes the substitution cleans up the hard part but leaves an isolated x that the du does not absorb. The move is then to express this x IN TERMS OF u, which the substitution relation always allows.
a) Compute ∫xx+1dx.
b) Compute ∫(x+2)3xdx.
c) Compute ∫x3x2+1dx. How does this case differ from the previous two?
d) Compute ∫1+xdx.
e) Compare ∫xx+1dx and ∫xx2+1dx: why is the second one much faster?
Show the solution
Answers
a)52(x+1)5/2−32(x+1)3/2+C
b)−x+21+(x+2)21+C
c)5(x2+1)5/2−3(x2+1)3/2+C
d)2x−2ln(1+x)+C
e)The factor x is the hidden derivative
a)
Let u=x+1, so x=u−1 and dx=du: ∫(u−1)udu=∫(u3/2−u1/2)du=52u5/2−32u3/2, that is 52(x+1)5/2−32(x+1)3/2+C. The remaining x was replaced by u−1: that is the whole trick.
b)
u=x+2, x=u−2: ∫u3u−2du=∫(u−2−2u−3)du=−u1+u21, that is −x+21+(x+2)21+C.
c)
u=x2+1 gives du=2xdx: one factor x is ABSORBED by the du, and what remains is x2=u−1, which is expressed without a root. 21∫(u−1)udu=5u5/2−3u3/2, that is 5(x2+1)5/2−3(x2+1)3/2+C. The difference from the first two cases is that the power of x was ODD: one factor goes into the du and the rest, an even power, converts cleanly. This is exactly the mechanism behind odd powers of sine.
d)
u=x, so x=u2 and dx=2udu: ∫1+u2udu=2∫(1−1+u1)du=2u−2ln∣1+u∣, that is 2x−2ln(1+x)+C. Note the division done along the way: as soon as a quotient has comparable degrees, divide before integrating.
e)
In the second one, the derivative of the inside is 2x: the factor x in front of the root IS the hidden derivative, the substitution absorbs everything and you get 3(x2+1)3/2+C in two lines. In the first one, the derivative of x+1 is 1: the factor x has nothing to do with it, it survives the substitution and must be re-expressed, which doubles the length of the calculation. The general rule is that the cost of a substitution is measured by what it leaves standing.
Exercise 3: Rationalizing: nested roots and exponentials
A well-chosen substitution turns an irrational or exponential integral into a simple RATIONAL FUNCTION, which Part C will then be able to decompose. This is the most underrated job of substitution: it does not compute, it changes the family.
a) Compute ∫1+exdx by letting u=ex.
b) Compute ∫1+exe2xdx.
c) Compute ∫x+3xdx. Which exponent should you choose for the substitution, and why?
d) Compute ∫ex+e−xdx.
e) State the rule for choosing the rationalizing substitution in front of nx, in front of ex and in front of nax+b.
Show the solution
Answers
a)x−ln(1+ex)+C
b)ex−ln(1+ex)+C
c)x=u6: 2u3−3u2+6u−6ln∣u+1∣+C
d)arctan(ex)+C
e)x=un, LCM of the indices, u=ex, u=nax+b
a)
u=ex gives du=exdx, so dx=udu: ∫u(1+u)du. The decomposition u(1+u)1=u1−1+u1 gives ln∣u∣−ln∣1+u∣=ln1+uu, that is x−ln(1+ex)+C. An exponential integral has become a rational function in two lines.
b)
u=ex and dx=udu: ∫1+uu2⋅udu=∫1+uudu=∫(1−1+u1)du=u−ln∣1+u∣, that is ex−ln(1+ex)+C.
c)
You need a substitution that makes BOTH exponents whole numbers at once: let x=u6, because 6 is the least common multiple of 2 and 3. Then x=u3, 3x=u2 and dx=6u5du: ∫u3+u26u5du=∫u+16u3du. Long division gives u+1u3=u2−u+1−u+11, hence 2u3−3u2+6u−6ln∣u+1∣+C with u=x1/6. A smaller exponent would leave a root, and a larger one would make the polynomial needlessly heavy.
d)
Multiply top and bottom by ex: ∫e2x+1exdx. With u=ex, this is ∫u2+1du=arctanu, that is arctan(ex)+C. The preliminary multiplication was essential: without it, no substitution gives a clean result. Notice along the way that ex+e−x1=2coshx1, which explains why this result is related to the hyperbolic functions.
e)
In front of nx, let x=un. In front of several roots with different indices, let x=um where m is the LCM of the indices. In front of ex, let u=ex, which always gives dx=udu and turns any fraction in ex into a rational function of u. In front of nax+b, let u=nax+b, so x=aun−b. In all four cases the goal is the same: make the fractional exponents disappear to land on a rational function, the only family we know how to integrate systematically.
Exercise 4: Substituting in a definite integral
In a definite integral, two strategies are correct: change the limits of integration along with the variable, or go back to x before evaluating. The first is almost always shorter, but it requires a check that the second does not.
a) Compute ∫1exlnxdx by changing the limits.
b) Compute ∫01x2+1xdx and ∫0ln21+e2xexdx.
c) Compute ∫041+xdx.
d) A student computes ∫−11x3(x4+1)2dx by letting u=x4+1 and finds zero because both limits become u=2. Is the answer right? Is the reasoning?
e) What condition must a substitution satisfy for changing the limits to be enough? Give a counterexample.
Show the solution
Answers
a)21
b)2ln2≈0.3466; arctan2−4π≈0.3218
c)2(2−ln3)≈1.803
d)Right because the integrand is odd, wrong reasoning
e)Monotonic substitution; u=x2 on [−1,1] fails
a)
u=lnx, du=xdx; x=1 gives u=0 and x=e gives u=1: ∫01udu=21.
b)
For the first, u=x2+1 goes from 1 to 2: 21∫12udu=2ln2≈0.3466. For the second, u=ex goes from 1 to 2: ∫121+u2du=arctan2−arctan1≈1.1071−0.7854≈0.3218.
c)
u=x goes from 0 to 2, with dx=2udu: ∫021+u2udu=2∫02(1−1+u1)du=2[u−ln(1+u)]02=2(2−ln3)≈1.803.
d)
The answer is RIGHT, but the reasoning is wrong, and that is worse. The integrand x3(x4+1)2 is an ODD function integrated over a symmetric interval: that is why it equals zero. The argument of equal limits is worthless here: the substitution u=x4+1 is not one-to-one on [−1,1], since x and −x give the same u. Writing ∫22 amounts to folding the interval onto itself, merging two pieces that do cancel here, but that would ADD UP for an even function. The same reasoning applied to ∫−11x3(x4+1)2⋅xdx, an even integrand that is positive everywhere except at 0, would again give zero, which would be absurd.
e)
The substitution must be MONOTONIC, hence one-to-one, on the interval of integration: that is what guarantees a correspondence without folding between the values of x and those of u. Counterexample: on [−1,1], the substitution u=x2 fails, because it sends −1 and 1 to the same u=1. For ∫−11x2dx, it would give ∫11, so zero, whereas the true value is 32. The remedy is to split the interval at x=0, where the substitution stops being monotonic, and to handle the two pieces separately.
Exercise 5: When substitution cannot work
Knowing that a technique will fail saves as much time as knowing which one to apply. Three families of integrands resist substitution completely, and each one points to another technique.
a) Compare ∫xex2dx and ∫ex2dx. What can be said about the second one?
b) Compare ∫xlnxdx and ∫lnxdx. Which technique does the second one call for?
c) Compare ∫1−x2xdx and ∫1−x2dx, then ∫x2−1xdx and ∫x2−1dx.
d) Facing ∫x2−4x2+1dx, a student spends ten minutes looking for a substitution. What should they have seen immediately?
e) Write the list of signals that, on first reading, RULE OUT substitution and point to the technique to use.
The first contains the factor x, half the derivative of x2: immediate substitution, 2ex2+C. The second does not, and no substitution can handle it. Better still: its antiderivative is not an ELEMENTARY function, it cannot be expressed with any finite combination of the usual functions. This is not a limit of our skill but a theorem, and the only acceptable answer is numerical, or by a power series.
b)
The first has the form ∫f′f with f=lnx: it equals 2(lnx)2+C. The second has no factor x1: it calls for integration by PARTS, with the invisible factor 1, and gives xlnx−x+C. The presence or absence of the x1 tips you from one technique to the other.
c)
With the factor x: −1−x2+C by substitution, and 21lnx2−1+C likewise. Without it: arcsinx+C, which requires a TRIGONOMETRIC substitution, and 21lnx+1x−1+C, which requires PARTIAL FRACTIONS. A single factor x therefore separates four different techniques, and it is the first detail to look at in the whole chapter.
d)
That the degree of the numerator EQUALS that of the denominator: the fraction is improper, and no substitution can apply before the division. x2−4x2+1=1+x2−45, and the remainder is decomposed into partial fractions. The rule is absolute: as soon as a rational function has a numerator of degree greater than or equal to that of the denominator, LONG DIVISION comes before everything else.
e)
First signal: the integrand is a PRODUCT of two functions from different families, a polynomial and an exponential, a polynomial and a logarithm, an inverse trigonometric function and something else, without one being the derivative of the inside of the other. That is integration by parts. Second signal: the integrand is a RATIONAL FUNCTION, a quotient of two polynomials, without the numerator being proportional to the derivative of the denominator. That is partial fractions, preceded by a division if the degree of the top reaches that of the bottom. Third signal: a ROOT of the form a2±x2 or x2−a2 remains, with no factor x to absorb it. That is trigonometric substitution. And if none of the three applies and substitution fails too, consider that the antiderivative may simply not exist in elementary form.
Part B: Integration by parts (/50)
Exercise 6: Choosing u: the LIATE rule and why it works
The formula ∫udv=uv−∫vdu does not say which factor to take as u, and yet that is the whole problem: a good choice simplifies the remaining integral, a bad one makes it worse. The acronym LIATE, for Logarithmic, Inverse trigonometric, Algebraic, Trigonometric, Exponential, gives the priority order for u.
a) Compute ∫xlnxdx, justifying the choice of u with LIATE.
b) Compute ∫xe−xdx and ∫xcosxdx.
c) Compute ∫lnxdx. Why is the choice dv=dx forced here rather than clever?
d) Redo ∫xexdx with the WRONG choice, u=ex and dv=xdx. Write the new integral and say precisely how the situation got worse.
e) Justify the LIATE order: what property do the functions at the front share, and which one do those at the back share?
Show the solution
Answers
a)2x2lnx−4x2+C
b)−(x+1)e−x+C; xsinx+cosx+C
c)xlnx−x+C, the only possible choice
d)2x2ex−21∫x2exdx: worse
e)At the front: simple derivative; at the back: antiderivative that reproduces itself
a)
The logarithm is at the front of LIATE: take u=lnx and dv=xdx, so du=xdx and v=2x2. Hence 2x2lnx−∫2x2⋅xdx=2x2lnx−21∫xdx=2x2lnx−4x2+C. The remaining integral has become a polynomial one: the logarithm has disappeared, which is exactly what we wanted.
b)
Here A comes before E: u=x and dv=e−xdx, so v=−e−x. ∫xe−xdx=−xe−x+∫e−xdx=−xe−x−e−x=−(x+1)e−x+C. Likewise, u=x and dv=cosxdx give xsinx−∫sinxdx=xsinx+cosx+C. In both cases, differentiating x makes it disappear.
c)
There is only one factor: to apply the formula you MUST make a second one, and the only one available is the factor 1, invisible in the notation. So let u=lnx and dv=1⋅dx, hence v=x: xlnx−∫x⋅xdx=xlnx−x+C. This is not a trick but the only possible route, because no antiderivative of the logarithm is known when you start, which rules out taking it as dv.
d)
With u=ex and dv=xdx: v=2x2 and du=exdx, so ∫xexdx=2x2ex−21∫x2exdx. The equality is perfectly TRUE, but useless: the degree of the polynomial went from 1 to 2, the remaining integral is harder than the original, and repeating the operation would only make it worse indefinitely. Integration by parts is never wrong; it is only fruitless or productive.
e)
The functions at the front, logarithms and inverse trigonometric functions, have two properties: their DERIVATIVE is algebraic, hence simpler than they are, and their antiderivative is not known at the start. They are therefore the obvious candidates for u, the role of the one you differentiate. The functions at the back, exponentials and trigonometric functions, have the opposite property: they reproduce themselves when integrated, without getting more complicated, which makes them convenient choices for dv. Algebraic functions sit in the middle because they lose one degree at each differentiation: they eventually disappear, provided you differentiate them. LIATE is therefore not an arbitrary recipe, it is the ranking of functions according to what differentiation does to them.
Exercise 7: Repeated parts and the tabular method
When the algebraic factor has degree n, it takes n successive integrations by parts to use it up. Laying them out in a TABLE lets you carry them all out at once, in a single line of calculation.
a) Compute ∫x2exdx by two successive integrations by parts.
b) Describe the tabular method: two columns, which operations, which signs, and when do you stop?
c) Compute ∫x3e2xdx by the tabular method.
d) Compute ∫x2sin3xdx.
e) Does the tabular method apply to ∫xlnxdx? To ∫excosxdx? Justify in both cases.
Show the solution
Answers
a)ex(x2−2x+2)+C
b)Differentiate on the left, integrate on the right, alternating signs, stop at 0
c)e2x(2x3−43x2+43x−83)+C
d)−3x2cos3x+92xsin3x+272cos3x+C
e)No in both cases
a)
First round: u=x2, dv=exdx, hence x2ex−2∫xexdx. Second round on the remaining integral: ∫xexdx=xex−ex. Altogether, x2ex−2xex+2ex+C=ex(x2−2x+2)+C. The polynomial lost one degree at each round, until it disappeared.
b)
Draw two columns. On the left, the ALGEBRAIC factor, which you DIFFERENTIATE from row to row until you get zero. On the right, the other factor, which you INTEGRATE from row to row as many times. Then multiply along the diagonals, each product taking an alternating sign +, −, +, −, starting with +. Stop as soon as the left column reaches zero, which happens after n+1 rows for a polynomial of degree n.
c)
Left column: x3, 3x2, 6x, 6, 0. Right column: e2x, 2e2x, 4e2x, 8e2x, 16e2x. Diagonal products, with alternating signs: +x32e2x−3x24e2x+6x8e2x−616e2x, that is e2x(2x3−43x2+43x−83)+C. Three integrations by parts carried out in one go.
d)
Left column: x2, 2x, 2, 0. Right column: sin3x, −3cos3x, −9sin3x, 27cos3x. Hence −3x2cos3x+92xsin3x+272cos3x+C, remembering that the alternating sign of the third term, +, combines with the sign of the antiderivative.
e)
For ∫xlnxdx, NO: the column to differentiate would have to contain the logarithm, which never reaches zero by differentiation, and the column to integrate would contain x, whose degree would grow indefinitely. The tabular method requires a polynomial factor that eventually becomes zero. For ∫excosxdx, NO as well, but for another reason: neither factor becomes zero by differentiation, so the two columns would go on forever. This is the CYCLIC case, which is solved differently, as the next exercise shows.
Exercise 8: The cyclic case: when the integral comes back
If neither factor simplifies when differentiated, two integrations by parts bring back the original integral. Far from being a failure, this is an EQUATION whose unknown is the integral.
a) Compute ∫excosxdx by carrying out two integrations by parts, then solving the resulting equation.
b) Why must you keep the SAME kind of choice in the second round? What happens if you swap the roles?
c) Compute ∫e2xsin3xdx.
d) Derive the general formula for ∫eaxsin(bx)dx and check that it gives back the result of c).
e) Compute ∫sin(lnx)dx. In what way is this case cyclic too?
Show the solution
Answers
a)2ex(sinx+cosx)+C
b)Swapping gives I=I
c)13e2x(2sin3x−3cos3x)+C
d)a2+b2eax(asinbx−bcosbx)+C
e)2x(sin(lnx)−cos(lnx))+C
a)
Let I=∫excosxdx. With u=ex and dv=cosxdx: I=exsinx−∫exsinxdx. On the latter, the same choice: ∫exsinxdx=−excosx+∫excosxdx=−excosx+I. Substituting back: I=exsinx+excosx−I, so 2I=ex(sinx+cosx) and I=2ex(sinx+cosx)+C.
b)
Because swapping the roles in the second round UNDOES exactly what the first one did: you land back on the identity I=I, true but empty, and the calculation literally goes in circles. So decide once and for all, for example always differentiate the exponential, and stick to it until the integral reappears. This is the most common error on this type of exercise, and it is frustrating because nothing is wrong: the calculation simply stands still.
c)
Let I=∫e2xsin3xdx. With u=e2x and dv=sin3xdx: I=−3e2xcos3x+32∫e2xcos3xdx. Then ∫e2xcos3xdx=3e2xsin3x−32I. Substituting back: I=−3e2xcos3x+92e2xsin3x−94I, so 913I=e2x(92sin3x−3cos3x) and I=13e2x(2sin3x−3cos3x)+C.
d)
The same calculation carried out with letters gives ∫eaxsin(bx)dx=a2+b2eax(asinbx−bcosbx)+C, and symmetrically ∫eaxcos(bx)dx=a2+b2eax(acosbx+bsinbx)+C. With a=2 and b=3: denominator 4+9=13 and numerator e2x(2sin3x−3cos3x), exactly c). The denominator a2+b2 also betrays the complex origin of the result: it is the squared modulus of a+ib, and the integral is computed in one line through ∫e(a+ib)xdx, of which you take the imaginary part.
e)
With u=sin(lnx) and dv=dx, so v=x: I=xsin(lnx)−∫cos(lnx)dx. On the latter, the same choice: ∫cos(lnx)dx=xcos(lnx)+∫sin(lnx)dx=xcos(lnx)+I. Hence I=xsin(lnx)−xcos(lnx)−I, that is I=2x(sin(lnx)−cos(lnx))+C. The case is cyclic because differentiating a sine of a logarithm gives back a cosine of a logarithm divided by x, which the v=x exactly compensates: the structure reproduces itself identically, just as with the exponential.
Exercise 9: Parts on a definite integral: the boundary term
On a definite integral, the formula becomes ∫abudv=[uv]ab−∫abvdu. The bracket, called the boundary term, is evaluated AT BOTH LIMITS, and forgetting it is the costliest error of the chapter.
a) Compute ∫01xexdx.
b) Compute ∫1elnxdx and interpret the result geometrically.
c) Compute ∫0πxsinxdx.
d) Compute ∫0+∞xe−xdx. What happens to the boundary term, and why is a limit needed?
e) A student writes ∫01xexdx=[xex]01−∫01exdx=e−ex. Point out the two faults.
Show the solution
Answers
a)1
b)1: area under ln between 1 and e
c)π
d)1, the boundary term tends to 0
e)ex instead of e−1, an answer that depends on x
a)
u=x, dv=exdx: [xex]01−∫01exdx=(e−0)−[ex]01=e−(e−1)=1. A perfectly round result, which is a good sign.
b)
u=lnx, dv=dx: [xlnx]1e−∫1edx=(e⋅1−1⋅0)−(e−1)=1. Geometrically, the area under the graph of the logarithm between 1 and e is exactly 1. You can see it another way: the rectangle [0,e]×[0,1] has area e, and the area to the left of the curve, between the vertical axis and the logarithm, is ∫01eydy=e−1; the difference is indeed 1.
The integral is improper, so you must write limb→+∞∫0bxe−xdx=limb→+∞([−xe−x]0b+∫0be−xdx). The boundary term equals −be−b, which tends to ZERO by comparison of growth rates, the decaying exponential winning over the linear factor. What remains is limb→+∞[−e−x]0b=1. The limit is essential: without it, you would write an expression containing ∞×0, an indeterminate form whose value cannot be guessed.
e)
The boundary term is right: [xex]01 equals 1⋅e−0⋅e0=e. First fault: the remaining integral ∫01exdx was replaced by the antiderivative ex, with no brackets and no limits. It had to be evaluated, [ex]01=e−1: the lower limit, which subtracts e0=1, vanished along with the brackets. Second fault, more serious: the final answer, e−ex, still contains the variable x, which is impossible for a definite integral, since it must produce a number. That is the alarm signal to know: an integral between limits can never depend on x. The correct value is e−(e−1)=1.
Exercise 10: The invisible factor 1, beyond the logarithm
Whenever an integrand is made of a SINGLE block whose derivative is known but whose antiderivative is not, integration by parts with dv=dx settles the question. The family is larger than you might think.
a) Compute ∫ln(x2+1)dx.
b) Compute ∫(lnx)2dx.
c) Compute ∫xln(x+1)dx, taking care with the division that follows.
d) Compute ∫ln(x+x2+1)dx.
e) Derive the reduction formula for ∫(lnx)ndx and deduce ∫(lnx)3dx.
u=ln(x2+1), dv=dx: xln(x2+1)−∫x2+12x2dx. The degree of the numerator equals that of the denominator, so divide: x2+12x2=2−x2+12. Hence xln(x2+1)−2x+2arctanx+C.
b)
u=(lnx)2, dv=dx: x(lnx)2−∫x⋅x2lnxdx=x(lnx)2−2∫lnxdx=x(lnx)2−2(xlnx−x), that is x(lnx)2−2xlnx+2x+C. Two successive applications, the second being the integral of the logarithm, already known.
c)
u=ln(x+1), dv=xdx, v=2x2: 2x2ln(x+1)−21∫x+1x2dx. Long division gives x+1x2=x−1+x+11, whose integral is 2x2−x+ln∣x+1∣. Result: 2x2ln(x+1)−4x2+2x−21ln∣x+1∣+C. You could also choose v=2x2−1, an equally valid antiderivative of x, which would bring out x+1x2−1=x−1 and remove the final logarithm: choosing the CONSTANT in v wisely is a little-known trick.
d)
u=ln(x+x2+1), whose derivative is x2+11, and dv=dx: xln(x+x2+1)−∫x2+1xdx=xln(x+x2+1)−x2+1+C, the last integral being a direct substitution. Note that this function is none other than the inverse hyperbolic sine, so you find the same structure as for the arcsine.
e)
Letting In=∫(lnx)ndx with u=(lnx)n and dv=dx: In=x(lnx)n−n∫(lnx)n−1dx=x(lnx)n−nIn−1. With I0=x and I1=xlnx−x, you get I2=x(lnx)2−2I1=x(lnx)2−2xlnx+2x, in agreement with b), then I3=x(lnx)3−3I2=x(lnx)3−3x(lnx)2+6xlnx−6x+C. The coefficients 1, 3, 6, 6 are those of the falling factorials, which is no accident: the recurrence multiplies by n at each step.
Part C: Partial fractions (/50)
Exercise 11: The four cases of decomposition
Every PROPER rational function decomposes into a sum of simple terms, and there are only four types of terms, dictated by the factorization of the denominator: distinct linear factor, repeated linear factor, distinct irreducible quadratic factor, repeated irreducible quadratic factor.
a) Decompose and integrate ∫(x−1)(x+3)5x−3dx: DISTINCT linear factors.
b) Decompose and integrate ∫x(x−1)23x2−x+1dx: REPEATED linear factor. How many terms does the repeated factor generate, and why?
c) Decompose and integrate ∫(x+1)(x2+1)3x2+xdx: IRREDUCIBLE quadratic factor. What form must its numerator take?
d) Compute ∫(x2+1)2dx, the case of a REPEATED quadratic.
e) Draw up the table of the four cases: which factor of the denominator generates which terms, and which function each one produces after integration.
Show the solution
Answers
a)21ln∣x−1∣+29ln∣x+3∣+C
b)ln∣x∣+2ln∣x−1∣−x−13+C
c)ln∣x+1∣+ln(x2+1)−arctanx+C
d)21(arctanx+x2+1x)+C
e)Log; log and powers; log and arctangent; reduction
a)
Write (x−1)(x+3)5x−3=x−1A+x+3B. Multiplying by (x−1) and evaluating at x=1: A=1+35−3=21. Likewise at x=−3: B=−3−1−15−3=29. Hence ∫=21ln∣x−1∣+29ln∣x+3∣+C.
b)
A factor (x−1)2 generates TWO terms, x−1B and (x−1)2C, because you need as many constants as the total degree of the factor to have enough freedom. Write x(x−1)23x2−x+1=xA+x−1B+(x−1)2C. At x=0: A=(−1)21=1. At x=1: C=13−1+1=3. Comparing the coefficients of x2: 3=A+B, so B=2. Hence ∫=ln∣x∣+2ln∣x−1∣−x−13+C. Note that the squared term gives a power, not a logarithm.
c)
An irreducible quadratic factor requires a numerator of degree 1, so of the form Bx+C and not a mere constant: otherwise there would not be enough parameters to represent every possible fraction. Write (x+1)(x2+1)3x2+x=x+1A+x2+1Bx+C. At x=−1: A=1+13−1=1. Substituting back, 3x2+x−(x2+1)=2x2+x−1=(Bx+C)(x+1), and since 2x2+x−1=(2x−1)(x+1), you get B=2 and C=−1. Hence ∫=ln∣x+1∣+∫x2+12x−1dx=ln∣x+1∣+ln(x2+1)−arctanx+C.
d)
This case decomposes no further: (x2+1)2 is already a simple term. Handle it by trigonometric substitution, x=tanθ: ∫sec4θsec2θdθ=∫cos2θdθ=2θ+4sin2θ=2θ+2sinθcosθ. Going back through the triangle: 21(arctanx+x2+1x)+C. This is why repeated quadratics are rare on exams: integrating them goes beyond partial fractions.
e)
Factor (x−a): one term x−aA, which gives a LOGARITHM. Factor (x−a)k: k terms x−aA1+⋯+(x−a)kAk, the first giving a logarithm and the others negative POWERS. Irreducible factor (x2+bx+c): one term x2+bx+cBx+C, whose part proportional to the derivative gives a logarithm and the rest an ARCTANGENT after completing the square. Factor (x2+bx+c)k: k similar terms, the higher powers requiring a trigonometric substitution or a reduction formula. The total number of constants to determine always equals the DEGREE of the denominator, which gives an immediate check before you start.
Exercise 12: The cover-up method, and when it is no longer enough
For distinct linear factors, there is a shortcut that gives each coefficient in one second, without any system: the cover-up method, or Heaviside's method. It has precise limits, which you must know as well as the method itself.
a) State the cover-up method and justify why it works.
b) Decompose (x−1)(x+2)(x−3)2x+7 by this method, then integrate.
c) Check your decomposition by comparing the coefficients of x2. What must the sum of the three constants be, and why?
d) Apply the method to x(x−1)2x+2. Which coefficients do you get directly, which ones resist, and how do you get them?
e) Why does the method fail completely on an irreducible quadratic factor? What are the two strategies available then?
Show the solution
Answers
a)Cover (x−a), evaluate at a
b)−23ln∣x−1∣+51ln∣x+2∣+1013ln∣x−3∣+C
c)A+B+C=0
d)2ln∣x∣−2ln∣x−1∣−x−13+C
e)No real root: convenient values or equating coefficients
a)
To find the coefficient associated with the factor (x−a), COVER this factor in the denominator of the original fraction and evaluate everything else at x=a. It works because multiplying the identity by (x−a) removes this factor from the term you want while leaving a factor (x−a) in front of all the others: when you then evaluate at a, all the other terms vanish and only the coefficient you want remains. The method is therefore nothing but the full calculation, done mentally.
b)
Coefficient of x−11: cover (x−1) and evaluate at 1, giving (1+2)(1−3)2+7=−69=−23. Coefficient of x+21: at −2, (−2−1)(−2−3)−4+7=153=51. Coefficient of x−31: at 3, (3−1)(3+2)6+7=1013. Hence ∫=−23ln∣x−1∣+51ln∣x+2∣+1013ln∣x−3∣+C. Three coefficients in three evaluations, without writing a single system.
c)
Putting everything back over the common denominator, the coefficient of x2 in the numerator is A+B+C. But the original numerator, 2x+7, has NO x2 term: the sum must therefore be zero. Check: −23+51+1013=−1.5+0.2+1.3=0. This check is free and catches almost all sign errors; it applies whenever the degree of the numerator is at least two less than the degree of the denominator.
d)
Cover-up immediately gives the coefficient of x1, by evaluating (x−1)2x+2 at x=0, which gives 2; and that of (x−1)21, the HIGHEST power of the repeated factor, by evaluating xx+2 at x=1, which gives 3. On the other hand, the coefficient of x−11 RESISTS: cover-up cannot reach it, because multiplying by (x−1) is not enough to isolate this term. You get it by comparing the coefficients of x2: 0=A+B, so B=−2. The decomposition is x2−x−12+(x−1)23, and the integral is 2ln∣x∣−2ln∣x−1∣−x−13+C.
e)
Because an irreducible quadratic factor has NO real root: there is no value of x at which to evaluate so as to cancel the other terms, and the method loses its mechanism. Two strategies remain. The first is to substitute CONVENIENT values of x, often 0, 1 and −1, as many times as there are remaining unknowns, which gives a small, easy system. The second is to expand and EQUATE the coefficients of each power of x, which is systematic but longer. In practice, you combine the three tools: cover-up for the real roots, convenient values to go fast, equating coefficients to finish.
Exercise 13: Divide first: improper fractions
Partial fraction decomposition exists only for a PROPER fraction, whose numerator has a degree strictly less than that of the denominator. Otherwise, polynomial long division is a required first step, and forgetting it makes the decomposition impossible.
a) Compute ∫x2−1x3dx.
b) Compute ∫x2−1x2+1dx.
c) Derive the formula ∫x2−a2dx=2a1lnx+ax−a+C and compare it with ∫x2+a2dx.
d) Compute ∫x2+1x3+2xdx.
e) What happens if you attempt a partial fraction decomposition WITHOUT dividing first? Illustrate with b).
Show the solution
Answers
a)2x2+21ln∣x2−1∣+C
b)x+lnx+1x−1+C
c)2a1lnx+ax−a versus a1arctanax
d)2x2+21ln(x2+1)+C
e)The identity is false: the 1 from the division is missing
a)
The numerator has degree 3, the denominator degree 2: divide. x2−1x3=x+x2−1x. The first piece integrates to 2x2, the second has the form ff′ up to a factor 21: 21lnx2−1. Total: 2x2+21lnx2−1+C. No decomposition was needed.
b)
Equal degrees, so divide again: x2−1x2+1=1+x2−12. The remainder decomposes: (x−1)(x+1)2=x−11−x+11, by cover-up. Hence ∫=x+ln∣x−1∣−ln∣x+1∣+C=x+lnx+1x−1+C.
c)
By cover-up on (x−a)(x+a)1: the coefficient of x−a1 is 2a1 and that of x+a1 is −2a1. Hence ∫x2−a2dx=2a1(ln∣x−a∣−ln∣x+a∣)=2a1lnx+ax−a+C. The comparison is instructive: ∫x2+a2dx=a1arctanax gives an ARCTANGENT, whereas the minus sign gives a LOGARITHM. A single sign separates two families of functions that have nothing in common, and it is the most profitable detail of the whole chapter.
d)
Division: x2+1x3+2x=x+x2+1x, since x(x2+1)=x3+x and x is left over. Hence ∫=2x2+21ln(x2+1)+C. Note that here the denominator is irreducible but the division is enough: there was no need to decompose at all.
e)
You would write (x−1)(x+1)x2+1=x−1A+x+1B and find, by cover-up, A=21+1=1 and B=−21+1=−1. But recombining, x−11−x+11=x2−12, which is NOT the original fraction: exactly the 1 from the division is missing. The proposed identity is therefore false, and integrating it would give an answer missing the term x. The symptom can be spotted before any calculation: if the degree of the numerator reaches that of the denominator, no sum of proper terms can ever reproduce it, since any sum of proper fractions is itself proper.
Exercise 14: The irreducible quadratic: logarithm plus arctangent
A term x2+bx+cBx+C with an irreducible denominator always splits into TWO pieces: the one proportional to the derivative of the denominator, which gives a logarithm, and the constant remainder, which gives an arctangent after completing the square.
a) Compute ∫x2+2x+52x+3dx by separating the two pieces.
b) Compute ∫x3+xdx.
c) Compute ∫x3+x2x2+1dx.
d) Compute ∫(x+1)(x2+4)xdx.
e) State the complete five-step procedure for integrating any rational function.
The derivative of the denominator is 2x+2. So write 2x+3=(2x+2)+1: ∫x2+2x+52x+2dx+∫x2+2x+5dx. The first piece has the form ff′ and equals ln(x2+2x+5), with no absolute value since the quadratic is always positive. The second requires completing the square, x2+2x+5=(x+1)2+4, and equals 21arctan2x+1. Total: ln(x2+2x+5)+21arctan2x+1+C.
b)
Factor first: x3+x=x(x2+1), with one linear factor and one irreducible quadratic. x(x2+1)1=xA+x2+1Bx+C; cover-up at x=0 gives A=1, then 1−(x2+1)=−x2=(Bx+C)x gives B=−1 and C=0. Hence ∫=ln∣x∣−21ln(x2+1)+C, which can be written lnx2+1∣x∣+C.
c)
x3+x2=x2(x+1): a REPEATED linear factor and a distinct linear factor. x2(x+1)x2+1=xA+x2B+x+1C. Cover-up: B=0+10+1=1 and C=11+1=2. Coefficients of x2: 1=A+C, so A=−1. Hence ∫=−ln∣x∣−x1+2ln∣x+1∣+C.
d)
(x+1)(x2+4)x=x+1A+x2+4Bx+C. Cover-up at x=−1: A=1+4−1=−51. Substituting back: x+51(x2+4)=(Bx+C)(x+1), that is 5x2+x+54=(Bx+C)(x+1). Comparing the highest-degree terms, B=51, and the constant terms, C=54. Hence ∫=−51ln∣x+1∣+101ln(x2+4)+52arctan2x+C, splitting 5(x2+4)x+4 into its logarithmic part and its arctangent part.
e)
Step one: check the DEGREES, and divide if the numerator reaches the denominator. Step two: FACTOR the denominator completely into linear and irreducible quadratic factors, which is often the real difficulty. Step three: write the form of the decomposition, with as many constants as the degree of the denominator, a numerator Bx+C for each quadratic and as many terms as the multiplicity for each repeated factor. Step four: determine the constants, by cover-up for the real roots, then by convenient values or equating coefficients for the rest. Step five: integrate term by term, each logarithm being read off directly, each negative power too, and each quadratic splitting into a logarithmic part and an arctangent after completing the square.
Exercise 15: Where partial fractions really pay off
The decomposition is not an algebra exercise for its own sake: it is the tool that unlocks the logistic model, telescoping series and the evaluation of definite integrals. Three uses, three different chapters.
a) The logistic model leads to ∫P(K−P)dP. Decompose and integrate.
b) Decompose n(n+1)1 and compute ∑n=1∞n(n+1)1.
c) Decompose n(n+2)1 and compute ∑n=1∞n(n+2)1. Why is the result no longer 1?
d) Compute ∫35x2−4dx and give a decimal value.
e) In what way is partial fraction decomposition the SAME operation in these three settings, despite very different appearances?
Show the solution
Answers
a)K1lnK−PP+C
b)n1−n+11; sum 1
c)21(n1−n+21); sum 43
d)41ln715≈0.1905
e)Break up into elementary pieces, then read the right table
a)
P(K−P)1=PA+K−PB; cover-up at P=0 gives A=K1, and at P=K gives B=K1. Hence ∫P(K−P)dP=K1(ln∣P∣−ln∣K−P∣)=K1lnK−PP+C. It is this antiderivative which, once inverted, produces the S-shaped logistic curve: without the decomposition, the differential equation would stay stuck at the separation of variables step.
b)
n(n+1)1=n1−n+11, by cover-up. The partial sum TELESCOPES: SN=(1−21)+(21−31)+⋯+(N1−N+11)=1−N+11, whose limit is 1. The series is therefore worth exactly 1.
c)
n(n+2)1=21(n1−n+21). This time the telescoping is SHIFTED by two places: SN=21(1+21−N+11−N+21), whose limit is 21⋅23=43. The result is no longer 1 because TWO terms survive at the start instead of one: the larger the shift, the more terms remain uncancelled, and the sum of a telescoping series is always read from what survives at the two ends.
d)
∫x2−4dx=41lnx+2x−2, from the formula derived earlier with a=2. Between 3 and 5: 41(ln73−ln51)=41ln715≈40.7621≈0.1905.
e)
In all three cases, the idea is to BREAK UP a complicated object into a sum of elementary objects about which everything is known. For the integral, the elementary objects are x−a1, whose antiderivative is a logarithm. For the series, they are n1 and n+k1, whose partial sums cancel each other. For the logistic model, they are P1 and K−P1, whose antiderivatives, logarithms again, unlock the separation of variables. The algebraic operation is strictly the same; only the table in which you read the result once the decomposition is done changes. That is why it is worth mastering the decomposition for its own sake: it is reused in three chapters that, on the surface, have nothing to do with each other.
Part D: Strategy, combinations and synthesis (/50)
Exercise 16: Look-alike pairs
Nothing wastes more time than applying the right technique to the wrong integral. Here are four pairs whose members differ only by a detail, and whose techniques have nothing in common.
a) Compute ∫xex2dx and ∫xexdx.
b) Compute ∫x2+1xdx and ∫x+1xdx.
c) Compute ∫xlnxdx and ∫lnxdx.
d) Compute ∫xx2+1dx and ∫xx+1dx.
e) For each of the four pairs, say in one sentence which DETAIL tips you from one technique to the other.
Show the solution
Answers
a)2ex2+C; (x−1)ex+C
b)21ln(x2+1)+C; x−ln∣x+1∣+C
c)2(lnx)2+C; xlnx−x+C
d)3(x2+1)3/2+C; 52(x+1)5/2−32(x+1)3/2+C
e)Is the outside factor the derivative of the inside?
a)
In the first, x is half the derivative of x2: SUBSTITUTION, and you get 2ex2+C. In the second, x is the derivative of nothing useful, the exponent already being x: integration by PARTS, with u=x, and you get (x−1)ex+C.
b)
The first has the form ff′ up to the factor 21: 21ln(x2+1)+C. The second is an IMPROPER rational function, equal degrees: divide first, x+1x=1−x+11, hence x−ln∣x+1∣+C.
c)
The first contains x1, the derivative of lnx: substitution, 2(lnx)2+C. The second has only one block: parts with the invisible factor 1, xlnx−x+C.
d)
The first has the factor x that the derivative of x2+1 calls for: substitution, 3(x2+1)3/2+C. The second does have an x, but the derivative of x+1 is 1: the x survives the substitution and must be re-expressed as u−1, hence 52(x+1)5/2−32(x+1)3/2+C.
e)
Pair a: is the exponent of e a SQUARE, of which x would be the derivative, or x itself? Pair b: is the denominator of degree 2, which makes the numerator its derivative, or of degree 1, which makes the fraction improper and forces a division? Pair c: is the factor x1, the derivative of the logarithm, present or absent? Pair d: does the derivative of what is under the root contain an x, in which case the outside factor is absorbed, or is it 1, in which case the factor survives? The four questions are really one: IS the outside factor the derivative of the inside?
Exercise 17: Combining: substitute first, integrate by parts next
Many integrals give way to no single technique but fall in two steps: a substitution cleans up the structure, then an integration by parts finishes the job. Recognizing that TWO steps are needed is the difficulty.
a) Compute ∫exdx.
b) Compute ∫cosxdx.
c) Compute ∫x3ex2dx.
d) Compute ∫xln(lnx)dx.
e) What clue, in the way an integrand is written, announces that a substitution alone will not be enough?
Show the solution
Answers
a)2(x−1)ex+C
b)2(xsinx+cosx)+C
c)2(x2−1)ex2+C
d)lnx⋅ln(lnx)−lnx+C
e)A u survives, multiplied by another function
a)
Let u=x, so x=u2 and dx=2udu: ∫eu⋅2udu=2∫ueudu. This is now a polynomial times exponential product, so integration by parts: 2(ueu−eu)=2(u−1)eu, that is 2(x−1)ex+C. The substitution did not compute the integral, it made it workable.
b)
Same substitution: 2∫ucosudu=2(usinu+cosu), that is 2(xsinx+cosx)+C.
c)
Let u=x2, so du=2xdx and x2=u: ∫x2⋅ex2⋅xdx=21∫ueudu=21(u−1)eu, that is 2(x2−1)ex2+C. One factor x went into the du, and the remaining x2 became the u of the product: the odd power did both jobs at once.
d)
Let u=lnx, so du=xdx: ∫lnudu=ulnu−u, that is lnx⋅ln(lnx)−lnx+C. The substitution removed the x1 and turned a double composition into the best-known integral of the chapter.
e)
The clue is a COMPOSITE function for which the count of factors does not come out right: its inner derivative is missing, or it is there but comes with an extra factor that will not go away. In ex, the derivative of x is 2x1, which appears nowhere: the substitution will therefore leave a u standing, and this u multiplied by an exponential calls for integration by parts. In other words, as soon as a substitution is structurally indicated but the count of factors does not come out right, plan for a second step. The useful reflex is to carry the substitution through to the end WITHOUT being discouraged by a surviving factor, then to look at what remains with fresh eyes.
Exercise 18: Rational functions in disguise
Many integrals that look like nothing special become rational functions after a well-chosen substitution. This is where the three previous parts meet, and one of the most frequent situations on exams.
a) Compute ∫sin2x+3sinx+2cosxdx.
b) Compute ∫x(x4+1)dx.
c) Compute ∫e2x−1exdx.
d) Compute ∫x(1+x)dx.
e) What do the four substitutions used have in common? State the rule in one sentence.
Show the solution
Answers
a)lnsinx+2sinx+1+C
b)41lnx4+1x4+C
c)21lnex+1ex−1+C
d)2arctanx+C
e)u=g(x) makes the fraction rational
a)
u=sinx gives du=cosxdx: ∫u2+3u+2du=∫(u+1)(u+2)du. By cover-up, u+11−u+21, hence lnu+2u+1, that is lnsinx+2sinx+1+C. A trigonometric integral handled entirely by partial fractions.
b)
Multiply top and bottom by x3 to bring out the derivative of x4: ∫x4(x4+1)x3dx. With u=x4 and du=4x3dx: 41∫u(u+1)du=41(ln∣u∣−ln∣u+1∣), that is 41lnx4+1x4+C. The preliminary multiplication is the decisive move.
c)
u=ex gives du=exdx: ∫u2−1du=21lnu+1u−1, from the formula of Part C with a=1. Result: 21lnex+1ex−1+C.
d)
u=x gives x=u2 and dx=2udu: ∫u(1+u2)2udu=2∫1+u2du=2arctanu, that is 2arctanx+C. The u in the numerator and the one in the denominator cancel, which is precisely the effect you want.
e)
In all four cases, the substitution replaces a building block of the integrand, the sine, the fourth power, the exponential or the root, by a single variable, turning the integrand into a quotient of POLYNOMIALS. The rule: as soon as an integrand is a fraction whose numerator and denominator depend on x only through one and the same function g(x), and g′(x) is present in the numerator or can be made to appear there, let u=g(x) and you will get a rational function, which Part C can always integrate. That is the deep reason why partial fractions are worth mastering: they are the end point of a great many calculations that did not look like it.
Exercise 19: Checking your result: three free checks
An antiderivative can always be checked, and in a few seconds. No other part of the course offers such an immediate check, and yet it is the one where the most wrong answers are handed in.
a) A student proposes ∫x2+4dx=arctan2x+C. Check by differentiating and correct it.
b) Another proposes ∫xx+1dx=2x2⋅32(x+1)3/2+C. Refute it in one line, without differentiating.
c) A third proposes ∫01x−2dx=[ln∣x−2∣]01=ln1−ln2=−ln2. Is the result right? Was the sign predictable?
d) A fourth proposes ∫−11x2dx=[−x1]−11=−1−1=−2. Where is the fault, and which check would have revealed it?
e) State the three checks to make systematically before handing in an answer.
Show the solution
Answers
a)21arctan2x+C
b)Product integrated factor by factor: wrong
c)Right, and negative like the integrand
d)Singularity at 0: the integral diverges
e)Differentiate, sign and order of magnitude, domain
a)
(arctan2x)′=1+4x21⋅21=4+x24⋅21=x2+42, that is TWICE the integrand. The correct answer is therefore 21arctan2x+C. This is the most common error of the whole chapter: the factor a1 in the arctangent formula gets forgotten because it cannot be guessed from the way the integrand is written.
b)
Without differentiating, it is enough to notice that the student integrated the two factors SEPARATELY, as if the integral of a product were the product of the integrals. That rule does not exist, and an immediate counterexample confirms it: ∫x⋅xdx=3x3, whereas the product of the integrals would give 2x2⋅2x2=4x4. The right answer is the one from Part A, obtained by substitution with u=x+1.
c)
The result is RIGHT, and its sign was entirely predictable: on [0,1], the quantity x−2 stays negative, so the integrand x−21 is negative everywhere, and the integral of a negative function can only be negative. The calculation is legitimate because the interval [0,1] does not contain the singularity x=2: the function is continuous there, and the absolute value in the logarithm handles the sign correctly.
d)
The integrand x21 is not defined at x=0, which is INSIDE the interval: the integral is improper and the fundamental theorem does not apply. You must split it in two, and each piece DIVERGES, so the integral does not exist. The check that would have revealed it is the simplest of all: x21 is strictly POSITIVE wherever it is defined, so no integral of it can be negative. A result of −2 is absurd before any analysis, and that glance alone was enough.
e)
First check, the safest: DIFFERENTIATE the proposed antiderivative and check that you get back the integrand. It is infallible and takes ten seconds; there is no excuse for skipping it on an indefinite integral. Second check, for a definite integral: the SIGN and the order of magnitude. A positive function gives a positive integral; a function between 1 and 2 on an interval of length 3 gives an integral between 3 and 6. Third check: the DOMAIN. Is the integrand defined and continuous on the whole interval? Is there a singularity inside? Is the proposed antiderivative valid on the whole interval, or only on one branch? These three checks catch almost all errors and cost almost nothing.
Exercise 20: Synthesis problem: twelve integrals, one decision tree
Twelve mixed integrals, with no hint about the method. As on the exam, the difficulty is not the calculation but the DIAGNOSIS. Work through them in order and write down, for each one, the question that decided the technique.
a) Compute ∫x2+x+72x+1dx, ∫x2+x−6dx and ∫x−2x2dx.
b) Compute ∫x2lnxdx, ∫x3lnxdx and ∫e3xsin2xdx.
c) Compute ∫xlnxdx, ∫ex+1e2xdx and ∫1+3xdx.
d) Compute ∫4−x2dx, ∫4−x2xdx and ∫x2−6x+13dx.
e) Write the complete decision tree of the set: which questions to ask, in what order, in front of an unfamiliar integral?
The first has a numerator exactly equal to the derivative of the denominator: form ff′, so ln(x2+x+7)+C. The second has a denominator that FACTORS, (x+3)(x−2): partial fractions, with by cover-up x+3−1/5+x−21/5, hence 51lnx+3x−2+C. The third is improper: divide first, x−2x2=x+2+x−24, hence 2x2+2x+4ln∣x−2∣+C.
b)
Polynomial times logarithm product: parts with u=lnx. ∫x2lnxdx=3x3lnx−9x3+C. The same with dv=x−3dx: ∫x3lnxdx=−2x2lnx−4x21+C. The third is the cyclic case: ∫e3xsin2xdx=13e3x(3sin2x−2cos2x)+C.
c)
The first contains x1, the derivative of lnx: substitution, ln∣lnx∣+C. The second becomes rational with u=ex, with dx=udu: ∫u+1udu=u−ln∣u+1∣, that is ex−ln(ex+1)+C. The third is rationalized by x=u3: ∫1+u3u2du=3∫(u−1+1+u1)du=23u2−3u+3ln∣1+u∣ with u=3x.
d)
Root a2−x2 with no factor x: basic form, arcsin2x+C. With the factor x: algebraic substitution, −4−x2+C. The third has an IRREDUCIBLE quadratic denominator, its discriminant being 36−52<0: completing the square, (x−3)2+4, hence 21arctan2x−3+C.
e)
Question 1: is the integrand a form from the TABLE, directly or after an elementary rewriting? Question 2: is there, in the numerator or as a factor, the DERIVATIVE of something that appears elsewhere? If so, substitution, and the matter is closed. Question 3: is it a PRODUCT of two different families, polynomial and exponential, polynomial and logarithm, exponential and trigonometric, or a single function of which only the derivative is known? If so, integration by parts, differentiating according to LIATE. Question 4: is it a RATIONAL FUNCTION? If so, compare the degrees, divide if needed, factor the denominator and decompose. Question 5: can it be MADE rational by a substitution, by letting u equal the exponential, the root or the trigonometric function that keeps coming back? Question 6, and only then: is there a root of the type a2±x2 or x2−a2 left, with no factor to absorb it? Trigonometric substitution. And if all six questions fail, consider that the antiderivative may not be elementary, or that a symmetry trick is expected. The order matters as much as the list: the costly techniques come last.
Something missing from this set?A type of exercise your teacher gives, a problem you are stuck on, a mistake you spotted: tell me, I read every request.
Sending goes through your free space: that is how I can answer you and tell you when the exercise is added. An email is enough, no password. Create your space.
Rather not create an account? Write to me directly.
Looking for an integral calculus tutor in Montreal?
Get in touch for a first session. We work on the decision tree rather than on recipes: recognizing in thirty seconds whether an integral calls for a substitution, integration by parts or a decomposition, and knowing why.