The fundamental theorem of calculus: derivatives of integrals and definite integrals
Ten problems on both parts of the fundamental theorem: an accumulation function checked against its closed form, a chain-rule upper limit, a variable lower limit, two moving limits and the maximum they give, critical points of an integral, definite integrals with an absolute value, the derivative of a Gaussian and an exponential CDF, and a limit of an integral by L'Hôpital's rule, with worked solutions and the mistakes of a missing chain-rule factor, a lost minus sign and confusing F with f.
Before you start
The fundamental theorem is the link between the two halves of calculus: the derivative of an accumulated area is the height of the curve, and a definite integral is a difference of antiderivatives. In machine learning it is the reason a probability density is the derivative of its cumulative distribution function, so the density of any transformed variable comes from the chain rule on a CDF. The work goes wrong in a few places: a chain-rule factor left off a limit like ,x3, the minus sign of a variable lower limit dropped, the value of the integrand at a constant lower limit subtracted, an absolute value integrated as if it were not there, and the accumulation function F confused with its integrand .f. These ten problems cover each of them.
Part 1. If f is continuous and ,F(x)=∫axf(t)dt, then .F′(x)=f(x). The constant a does not appear in the answer.
Moving limits. By the chain rule, .dxd∫g(x)h(x)f(t)dt=f(h(x))h′(x)−f(g(x))g′(x). A constant limit contributes ;0; a variable lower limit contributes a minus sign.
Part 2. If f is continuous on [a,b] and ,F′=f, then .∫abf(x)dx=F(b)−F(a). Any antiderivative works, so no +C is needed.
F(x)=∫axf(t)dt is a function of ,x, not of :t:t is a dummy variable. F is increasing where ,f>0, so its critical points are the zeros of .f.
A cumulative distribution function is F(x)=∫−∞xp(t)dt for a density ,p, so .F′=p. For the standard normal, Φ(z)=∫−∞zφ(t)dt with .φ(t)=2π1e−t2/2.
∣x−c∣ is c−x for x<c and x−c for :x>c: split the integral at .c.
Let .F(x)=∫1x(t3−2t)dt. Find F′(x) by Part 1, then find F(x) in closed form and check.
··
Find dxd∫1x3lntdt for ,x>0, and its value at .x=2.
··
Find ,dxd∫x51+t4dt, and its value at .x=1.
···
Let G(x)=∫x2xe−tdt for .x≥0. Find ,G′(x), and the x at which G is largest.
··
Let .F(x)=∫0x(t2−4t+3)dt. Find the local maximum and minimum values of .F.
·
Evaluate .∫14(2x+x1)dx.
··
Evaluate .∫03∣x−1∣dx.
··
Φ is the standard normal CDF. Find ,dxdΦ(2x−1), and its value at .x=21.
···
The waiting time T has density p(t)=2e−2t for .t≥0. Find its CDF ,F(x)=P(T≤x), check that ,F′=p, and find the median.
···
Find .x→0limx31∫0xsin(t2)dt.
Answers
,F′(x)=x3−2x, and F(x)=4x4−x2+43 has that derivative
;9x2lnx; at x=2 it is 36ln2≈24.95
;−1+x4; at x=1 it is −2≈−1.414
;G′(x)=2e−2x−e−x;G is largest at ,x=ln2≈0.693, where G=41
Local maximum 34 at ;x=1; local minimum 0 at x=3
17
25
;2φ(2x−1); at x=21 it is 2/π≈0.798
;F(x)=1−e−2x; median 2ln2≈0.347
31
Worked solutions
Problem 1
Let .F(x)=∫1x(t3−2t)dt. Find F′(x) by Part 1, then find F(x) in closed form and check.
F′(x)=x3−2xPart 1: replace t with the upper limit.
F(x)=[4t4−t2]1x=4x4−x2−(41−1)Part 2 with the antiderivative .4t4−t2.
F(x)=4x4−x2+43The lower limit only adds the constant .43.
,F′(x)=x3−2x, and F(x)=4x4−x2+43 has that derivativeThe constant 43 differentiates to ,0, which is why the lower limit never appears in .F′. Also ,F(1)=0, as it must.
Problem 2
Find dxd∫1x3lntdt for ,x>0, and its value at .x=2.
ln(x3)⋅dxdx3Put the upper limit into the integrand, then multiply by its derivative.
=3lnx⋅3x2ln(x3)=3lnx and .(x3)′=3x2.
Check: ,∫1x3lntdt=[tlnt−t]1x3=3x3lnx−x3+1, whose derivative is 9x2lnx+3x2−3x2The long way agrees.
;9x2lnx; at x=2 it is 36ln2≈24.95Without the chain-rule factor the answer would be ,3lnx, off by a factor of .3x2.
Problem 3
Find ,dxd∫x51+t4dt, and its value at .x=1.
∫x51+t4dt=−∫5x1+t4dtSwapping the limits changes the sign, and puts x in the upper limit.
dxd(−∫5x1+t4dt)=−1+x4Part 1.
;−1+x4; at x=1 it is −2≈−1.414Moving the lower limit right removes area from a positive integrand, so the derivative must be negative.
Problem 4
Let G(x)=∫x2xe−tdt for .x≥0. Find ,G′(x), and the x at which G is largest.
G′(x)=e−2x⋅2−e−x⋅1Upper limit 2x with derivative ,2, minus the lower limit x with derivative .1.
G′(x)=0⟺2e−2x=e−x⟺e−x=21Divide by .e−x>0.
x=ln2G′>0 for x<ln2 and G′<0 after, so this is the maximum.
G(ln2)=e−ln2−e−2ln2=21−41G(x)=e−x−e−2x by Part 2.
;G′(x)=2e−2x−e−x;G is largest at ,x=ln2≈0.693, where G=41G(0)=0 and G→0 as :x→∞: the interval [x,2x] starts empty and ends where e−t is tiny.
Problem 5
Let .F(x)=∫0x(t2−4t+3)dt. Find the local maximum and minimum values of .F.
F′(x)=x2−4x+3=(x−1)(x−3)Part 1. The critical points of F are the zeros of its integrand.
,F′′(x)=2x−4, so F′′(1)=−2 and F′′(3)=2Local maximum at ,1, local minimum at .3.
F(x)=3x3−2x2+3xPart 2; the lower limit 0 adds nothing.
F(1)=31−2+3=34 and F(3)=9−18+9=0The values of ,F, not of the integrand.
Local maximum 34 at ;x=1; local minimum 0 at x=3F climbs while the integrand is positive, on ,(0,1), and loses exactly that area again on ,(1,3), where the integrand is negative.
Problem 6
Evaluate .∫14(2x+x1)dx.
An antiderivative is x2+2x.∫x−1/2dx=1/2x1/2=2x.
(16+2⋅2)−(1+2⋅1)=20−3.F(4)−F(1). The integrand is continuous on .[1,4].
17Term by term: ∫142xdx=15 and .∫14x−1/2dx=2.
Problem 7
Evaluate .∫03∣x−1∣dx.
∣x−1∣=1−x on [0,1] and x−1 on [1,3]Split where the inside changes sign.
∫01(1−x)dx=[x−2x2]01=21The first piece.
∫13(x−1)dx=[2x2−x]13=23−(−21)=2The second piece.
25Two triangles: one with legs 1 and ,1, one with legs 2 and .2.
Problem 8
Φ is the standard normal CDF. Find ,dxdΦ(2x−1), and its value at .x=21.
Φ(2x−1)=∫−∞2x−1φ(t)dtA CDF is an integral with a variable upper limit.
dxdΦ(2x−1)=φ(2x−1)⋅2Part 1 with the chain rule; the lower limit −∞ does not move.
At :x=21:2φ(0)=2π2.φ(0)=2π1.
;2φ(2x−1); at x=21 it is 2/π≈0.798This is the density of :X=2Z+1:,P(X≤x)=Φ(2x−1), and differentiating a CDF gives the density. It is the normal density with mean 21 and standard deviation .21.
Problem 9
The waiting time T has density p(t)=2e−2t for .t≥0. Find its CDF ,F(x)=P(T≤x), check that ,F′=p, and find the median.
F(x)=∫0x2e−2tdt=[−e−2t]0xThe density is 0 for ,t<0, so the integral starts at .0.
F(x)=1−e−2x for x≥0.−e−2x−(−1).F(0)=0 and .F→1.
F′(x)=2e−2x=p(x)Part 1, confirmed by differentiating the closed form.
1−e−2m=21⟺e−2m=21⟺m=2ln2The median is where the CDF reaches .21.
;F(x)=1−e−2x; median 2ln2≈0.347The mean is ,21, larger than the median: the long right tail pulls it up.
Problem 10
Find .x→0limx31∫0xsin(t2)dt.
As x→0 both ∫0xsin(t2)dt and x3 go to 0The form is ,00, so L'Hôpital's rule applies.
x→0lim3x2sin(x2)Differentiate the top by Part 1 and the bottom by the power rule.
x2sin(x2)→1With ,u=x2→0,.usinu→1.
31Near ,0,,sin(t2)≈t2, and :∫0xt2dt=3x3: the same answer without the rule.
Where this goes wrong
1. Leaving out the chain-rule factor
Part 1 is learned as "put the upper limit in", which is the whole rule only when the limit is .x.
,dxd∫1x3lntdt,x>0Right so far: Problem 2.
“Replace t with the upper limit.”The habit that causes the mistake: the limit is ,x3, not ,x, so it moves 3x2 times as fast.
,ln(x3)=3lnx, so at x=2 the derivative is 3ln2≈2.079The edge of the area moves at the rate ,(x3)′=3x2, and the derivative is ,9x2lnx, which is 36ln2≈24.95 at ,x=2, twelve times as much. Differentiating the closed form 3x3lnx−x3+1 shows the factor.
2. Dropping the sign of a variable lower limit
In most problems x is the upper limit, so the minus sign is easy to forget.
dxd∫x51+t4dtRight so far: Problem 3.
“Put x into the integrand.”The habit that causes the mistake: Part 1 applied without first moving x to the upper limit.
The derivative is ,1+x4, so 2≈1.414 at x=1,∫x5=−∫5x, so the derivative is ,−1+x4, which is −1.414 at .x=1. As x increases, the interval [x,5] gets shorter and the integrand is positive, so the integral is decreasing.
3. Subtracting the integrand at the lower limit
Part 2 is "top minus bottom", and the pattern carries over to Part 1, where it does not belong.
F(x)=∫1x(t3−2t)dtRight so far: Problem 1.
“The derivative is f at the top minus f at the bottom.”The habit that causes the mistake: Part 2's subtraction copied into Part 1.
F′(x)=(x3−2x)−(1−2)=x3−2x+1The lower limit is a constant, and it contributes a constant, ,43, to ,F, which differentiates to .0.;F′(x)=x3−2x; at x=2 that is ,4, not .5. Only a lower limit that moves with x contributes to the derivative.
4. Integrating an absolute value without splitting
The bars look like decoration once the integral is set up.
∫03∣x−1∣dxRight so far: Problem 7.
“For most of the interval ,x−1≥0, so use .x−1.”The habit that causes the mistake: one formula used on both sides of .x=1.
∫03(x−1)dx=29−3=23On ,[0,1],x−1 is negative, so the triangle there, of area ,21, is subtracted instead of added. Split at :x=1:.21+2=25. The integral of a function that is never negative is a total area, and here 23=2−21 is a net one.
5. Reporting the integrand's value instead of F's
F and its integrand share the same critical points, which makes it easy to evaluate the wrong one.
,F(x)=∫0x(t2−4t+3)dt, local maximum at x=1Right so far: Problem 5, step 2.
“Plug x=1 into the function.”The habit that causes the mistake: the function on the page is the integrand.
The local maximum value is 1−4+3=00 is the integrand at :x=1: it is ,F′(1), which is 0 at every critical point. The value of F is the area accumulated up to ,1,.F(1)=31−2+3=34. The local minimum value is ,F(3)=0, which happens to be 0 for a different reason.