MATH2111 Higher Several Variable Calculus Integration: School of Mathematics and Statistics University of New South Wales
MATH2111 Higher Several Variable Calculus Integration: School of Mathematics and Statistics University of New South Wales
Integration
Semester 1, 2014
For a bounded function f : [a, b] → R, if there exists a unique number I such that
S P (f ) ≤ I ≤ S P (f )
for every partition P of [a, b], then f is Riemann integrable on [a, b] and
Z Z b
I = f = f (x)dx.
[a,b] a
Let
P1 = {a = x0 , x1 , x2 , . . . , xn = b}
y4
be a partition of [a, b] and y3
y2 R R42
P2 = {c = y0 , y1 , y2 , . . . , ym = d } y1
a partition of [c, d ].
y0
Then R is the union of the mn
subrectangles x0 x1 x2x3 x4 x5
Integration
The upper and lower Riemann sums of f with respect to these partitions are
X
S P1 ,P2 (f ) = f jk ∆xj ∆yk y4
j,k
and
y3
y2 R R42
X y1
S P1 ,P2 (f ) = f jk ∆xj ∆yk .
j,k
y0
The sums are over all pairs (j, k) with
x0 x1 x2x3 x4 x5
1≤j ≤n and 1 ≤ k ≤ m.
The numbers f jk and f jk are the infimum and supremum of f on Rjk and ∆xj and
∆yk its width and height.
S P1 ,P2 (f ) ≤ I ≤ S P1 ,P2 (f )
Riemann Integral
For a function of one variable, the Riemann integral is interpreted as the (signed)
area bounded by the graphRRy = f (x) and the x-axis over the interval [a, b]. For a
function of two variables, R f is the (signed) volume bounded by the graph
z = f (x, y ) and the xy -plane over the rectangle R.
Properties:
z = f (x, y )
y
R
x
Upper sum
z = f (x, y )
y
R
x
Note that Z b Z d
f (x, y )dydx
a c
means !
Z b Z d
f (x, y )dy dx.
a c
The integral inside the brackets is, for each x, a one variable integral.
Fubini’s theorem for rectangular regions says gives two ways of calculating this
integral.
Z Z 2 Z 3 Z 2 3
3 1 2 3 y
f = xy dx dy = x y dy
R 1 0 1 2 0
Z 2 2
9 3
= y dy
1 2 1
2
9 1 4
= y
2 4 1 x
3
135
= .
8
and it is tempting to consider the sum in the brackets as a one variable Riemann
lower sum of f (x, y ) for a fixed value of y . However, f j,k is not necessarily an
infimum for f (x, y ) as a function of x for any particular fixed value of y .
One
Pn way around this problem is to use the continuity of f to show that
j=1 f j,k ∆xj can be made to be within ε of the one variable lower sum of
f (x, yk ) by requiring the spacing in the partition P2 to be sufficiently small.
A more eligant approach can be found on pages 65–70 of the Internet Supplement
to Marsden and Tromba 5th Edition.
JM Kress (UNSW Maths & Stats) MATH2111 Integration Semester 1, 2014 12 / 58
Riemann integral over more general regions
The proof of this theorem is exercises 4, 5 and 6 from the Internet Supplement to
Marsden and Tromba. The essence of the proof is that the contribution to the
upper and lower Riemann sums from subrectangles containing the the lines of
discontinuity can be made arbitrarily small by taking the width of subintervals in
the partitions to be small enough.
Z b
Similarly, if the integral f (x, y )dx exists for each y ∈ [c, d ], then
a
!
ZZ Z d Z b
f = f (x, y )dx dy .
R c a
Rd
Since f is not continuous there is no guarantee that c
f (x, y )dy exists for each x
Rb
or that a f (x, y )dx exists for each y .
JM Kress (UNSW Maths & Stats) MATH2111 Integration Semester 1, 2014 14 / 58
Riemann integral over more general regions
An elementary region is a region of the type illustrated below.
Notes
The properties of the integral over rectangular regions also apply to more general
regions.
(a) Find the integral of f (x, y ) = x 2 y over the triangular region Ω with vertices
(0, 0), (0, 1) and (1, 0).
(b) For each of the following two integrals, find the region over which the
integral is calculated and reverse the order of integration.
(i)
Z 1 Z 2−2x
I1 = f (x, y )dydx.
0 0
(ii)
Z 1 Z √1−x 2
I2 = f (x, y )dydx.
−1 x−1
Fubini’s theorem
(a) Find the integral of f (x, y ) = x 2 y over the triangular region Ω with vertices
(0, 0), (0, 1) and (1, 0).
(b) For each of the following two integrals, find the region over which the
integral is calculated and reverse the order of integration.
(i)
Z 1 Z 2−2x
I1 = f (x, y )dydx.
0 0
(ii)
Z 1 Z √1−x 2
I2 = f (x, y )dydx.
−1 x−1
So,
Z 1 Z 2−2x Z 2 Z 1− 12 y
I1 = f (x, y ) dydx = f (x, y ) dxdy .
0 0 0 0
y
To reverse the order of integration and integrate p
with respect to x first, it is convenient to split x =− 1 − y2 p
the region into two pieces. x= 1 − y2
1
Z Z √ 2 1 1−y
I2A = √ f (x, y ) dx dy .
0 − 1−y 2 Ω2 Ω2A
Z 0 Z y +1 −1 Ω2B 1 x
I2B = f (x, y ) dx dy .
−2 −1 x = −1 x =y +1
−1
So, √
Z 1 Z 1−x 2
I2 = f (x, y ) dydx −2
−1 x−1
Z 1 Z √1−y 2 Z 0 Z y +1
= √ f (x, y ) dxdy + f (x, y ) dxdy .
0 − 1−y 2 −2 −1
JM Kress (UNSW Maths & Stats) MATH2111 Integration Semester 1, 2014 26 / 58
Uniform continuity
Definition
A function f : Ω ⊂ Rn → Rm is uniformly continuous on Ω if for all x, y ∈ Ω and
for all positive ∈ R there exists δ such that
d (x, y) < δ ⇒ d f(x), f(y) < .
In the definition of continuity, δ may depend on x, but here, given , the same δ
must work for all x.
Theorem
A continuous function on a compact set Ω is uniformly continuous on Ω.
Leibniz’ rule
then
Z b
d
F 0 (x) = f (x, y )dy
dx a
Z b
∂f
= (x, y )dy .
a ∂x
[Given any of Leibniz’ rule, Fubini’s theorem or Clairot’s theorem, the others can
be proved. See Q9 on tutorial sheet 4.]
JM Kress (UNSW Maths & Stats) MATH2111 Integration Semester 1, 2014 28 / 58
Proof of Leibniz’ rule
Given Z b
F (x) = f (x, y )dy ,
a
we want to show that
b
F (x + h) − F (x)
Z
0 ∂f
F (x) = lim = (x, y )dy .
h→0 h a ∂x
So consider
F (x + h) − F (x) Z b ∂f
− (x, y )dy
h a ∂x
R b
f (x + h, y )dy − R b f (x, y )dy Z b ∂f
a a
= − (x, y )dy
h a ∂x
Z
b f (x + h, y ) − f (x, y ) ∂f
= − (x, y ) dy
a h ∂x
Z
b ∂f ∂f [for some c between x and x + h
= (c, y ) − (x, y ) dy
a ∂x ∂x by the Mean Value Theorem]
JM Kress (UNSW Maths & Stats) MATH2111 Integration Semester 1, 2014 29 / 58
Z
F (x + h) − F (x) Z b ∂f b ∂f ∂f
− (x, y )dy = (c, y ) − (x, y ) dy
h a ∂x a ∂x ∂x
Z b
∂f ∂f
≤ (c, y ) −
∂x (x, y ) dy
a ∂x
Z Z
g ≤ g
Ω Ω
∂f
Since is uniformly continuous, for any 0 > 0 there is a δ (independent of c, x
∂x
and y ) such that
∂f ∂f
(x, y ) < 0
(c, y ) − (x, y )
< δ ⇒ (c, y ) −
∂x ∂x
Z b
∂f ∂f dy < 0 (b − a)
⇒ (c, y ) −
∂x (x, y )
a ∂x
So,
F (x + h) − F (x) Z b ∂f
lim − (x, y )dy = 0,
h→0 h a ∂x
that is,
b
F (x + h) − F (x)
Z
∂f
F 0 (x) = lim = (x, y )dy .
h→0 h a ∂x
to find
Z 1
xdx
(a)
0 (ax + b)3/2
Z 1
dx
(b)
0 (ax + b)3/2
1 ∂f x
f (a, x) = √ and (a, x) = − for b > 0
ax + b ∂a 2(ax + b)3/2
are bounded and continuous and hence are both uniformly continuous on Ω.
∂
Z 1
1
∂ 2 √
√
√ dx = a+b− b
∂a 0 ax + b ∂a a
Z 1
∂
1
2 √ √ 1
⇒ √ dx = − 2 a+b− b + √
0 ∂a ax + b a a a+b
Z 1
1
x
2 √
√ 1
⇒ − dx = − a + b − b + √
0 2 (ax + b)3/2 a2 a a+b
Z 1
x 4 √ √ 2
⇒ dx = a + b − b − √ .
0 (ax + b)
3/2 a2 a a+b
1 ∂f 1
f (b, x) = √ and (b, x) = − for a > 0
ax + b ∂b 2(ax + b)3/2
are bounded and continuous and hence are both uniformly continuous on Ω.
∂
Z 1
1
∂ 2 √
√
√ dx = a+b− b
∂b 0 ax + b ∂b a
Z 1
∂ 1 1 1 1
⇒ √ dx = √ −√
0 ∂b ax + b a a+b b
Z 1
1 1 1 1 1
⇒ − 3/2
dx = √ −√
0 2 (ax + b) a a + b b
Z 1
1 2 1 1
⇒ 3/2
dx = − √ −√ .
0 (ax + b) a a+b b
(a) Assuming suitable continuity and uniform continuity of the integrand and its
partial derivative with respect to a,
Z ∞
∂ −ax ∂ k
e sin(kx) dx =
∂a 0 ∂a a2 + k 2
Z ∞
∂ 2ak
e −ax sin(kx) dx
⇒ =− 2
0 ∂a (a + k 2 )2
Z ∞
2ak
⇒ −xe −ax sin(kx)dx =− 2
0 (a + k 2 )2
Z ∞
2ak
⇒ xe −ax sin(kx)dx = 2 .
0 (a + k 2 )2
JM Kress (UNSW Maths & Stats) MATH2111 Integration Semester 1, 2014 35 / 58
Z ∞
k
e −ax sin(kx) dx =
0 a2 + k 2
(b) Assuming suitable continuity and uniform continuity of the integrand and its
partial derivative with respect to k,
Z ∞
∂ −ax ∂ k
e sin(kx) dx =
∂k 0 ∂k a2 + k 2
∞
2k 2
Z
∂ −ax
1
⇒ e sin(kx) dx = 2 − 2
0 ∂k a + k2 (a + k 2 )2
Z ∞
a2 − k 2
⇒ xe −ax cos(kx)dx = 2 2 )2
.
0 (a + k
Example:
t
√ Z t
cos(t t) d √ cos(t.t) dt
Z
d cos(tx) ∂ cos(tx)
√
dx = − √ t + + √ dx
dt t x t dt t dt t ∂t x
√ Z t
cos(t t) 1 cos(t 2 ) −x sin(tx)
=− √ √ + + √ dx
t 2 t t t x
√ t
cos(t t) cos(t 2 )
cos(tx)
=− + + √
2t t t t
√ √
cos(t t) cos(t 2 ) cos(t 2 ) cos(t t)
=− + + −
2t √ t t t
2
3 cos(t t) 2 cos t
=− +
2t t
Change of variables
x + y = 1, x + y = 3, v
x − y = 1, x − y = −1. 1
Ω′
Perhaps it might be easier to use
1 3 u
u =x +y and v = x − y
−1
as coordinates. In the uv -plane, the
corresponding region Ω0 has boundaries
u = 1, u = 3, v = −1 and v = 1.
JM Kress (UNSW Maths & Stats) MATH2111 Integration Semester 1, 2014 41 / 58
Change of variables
v
If
1
u =x +y and v = x − y
Ω′
then
1 1 3 u
(u + v ).
x=
2
Can we simply write the integral as −1
Z 1 Z 3
1
I = (u + v ) dudv ?
−1 1 2
1
No! The map dudv = 2dxdy ⇒ dxdy =
2
dudv .
u 1 1 x So,
=
v 1 −1 y Z 1 Z 3
1 1
I = (u + v ) dudv .
1 1
−1 2 2
scales areas by a factor of det = 2. 1
1 −1
JM Kress (UNSW Maths & Stats) MATH2111 Integration Semester 1, 2014 42 / 58
Change of variables
For a more general change of variables by a differentiable function f : R → R
u x x x0 x0 x − x0
=f 'T =f + Jf
v y y y0 y0 y − y0
∆x
y D C
A00 ' f x0 x0
0
+ Jf
∆y f
y0 y0 0
B 00 ' f x0 x0 ∆x
y0 + Jf y0 0
A(x0 , y0 ) B v
C′ C 00 ' f x0 x0 ∆x
D′ y0 + Jf y0 ∆y
x D 00 ' f x0
+ Jf x0
0
T ′
y0 y0 ∆y
B
u C ′′ A′
D ′′
u
Area A0 B 0 C 0 D 0
′′ ' Area A00 B 00 C 00 D 00
B
A′′ = Area ABCD × |detJF |.
v
JM Kress (UNSW Maths & Stats) MATH2111 Integration Semester 1, 2014 43 / 58
Change of variables
Alternative notation:
Z Z
∂(x, y )
f (x, y ) dxdy = f (x(u, v ), y (u, v )) dudv
Ω0 Ω ∂(u, v )
where
∂x ∂x
∂(x, y )
= det(JF ) = det ∂u ∂v
∂(u, v ) ∂y ∂y
∂u ∂v
x 2 − y 2 = 1, x 2 − y 2= 4
x x
y= , y= .
2 4
y
Let u = x 2 − y 2 and v = .
x
v
1
2
Ω′ ∂x ∂x
1
4 ∂(x, y ) ∂u ∂v
=
∂(u, v ) ∂y ∂y
1 4 u
∂u ∂v
Z ZZ ZZ
∂(x, y )
Area = 1= 1 dxdy = 1 dudv .
Ω Ω Ω0 ∂(u, v )
JM Kress (UNSW Maths & Stats) MATH2111 Integration Semester 1, 2014 45 / 58
Change of variables
x2 − y2
!
x u x
=F ⇒ F −1 = y
y v y
x
Note that F −1 is differentiable for x 6= 0. So
∂u ∂u
−2y
!
∂x ∂y 2x
J(F −1 ) = ∂v ∂v = y 1 .
− 2
∂x ∂y x x
−1
1 y y2
det J(F ) = 2x − (−2y ) − 2 = 2 − 2 2 = 2 − 2v 2 .
x x x
But we want
1 1 1
det(JF ) = = = 2
,
det (JF ) −1 −1
det J(F ) 2 − 2v
Change of variables
Integrate x over the part of the unit disc that lies in the first quadrant.
Change of variables
z z
z ρ
φ
y y
θ θ
x x
x = r cos θ x = ρ sin φ cos θ
∂(x, y , z) ∂(x, y , z)
y = r sin θ y = ρ sin φ sin θ = ρ2 sin φ
∂(r , θ, z) = r
∂(ρ, θ, φ)
z =z z = ρ cos φ
JM Kress (UNSW Maths & Stats) MATH2111 Integration Semester 1, 2014 52 / 58
Spherical and polar coordinates
Example (Exam 2008):
above by z = 4 − x 2 − y 2 .
m1
m2
x̄ m3
x1 x2 x3 x
We can find the approximate centre of mass using an upper or lower sum of ρ
with respect to a partition P.
ρ(x)
x̄
n
X Z b Z b
ρ̄i ∆xi xi ρ(x) x dx
i=1 a
M= ρ(x) dx is the total mass
x̄ ' n , x̄ = Z b
. a
X ρ(x)
ρ̄i ∆xi ρ(x) dx and is a probability density.
a M
i=1
JM Kress (UNSW Maths & Stats) MATH2111 Integration Semester 1, 2014 55 / 58
For a lamina occupying the region For a solid body occupying the region
Ω ⊂ R2 with density ρ : Ω → R, the Ω ⊂ R3 with density ρ : Ω → R, the
total mass is total mass is
ZZ ZZZ
M= ρ(x, y ) dxdy M= ρ(x, y , z) dxdydz
Ω Ω
The coordinates of the centre of mass are The coordinates of the centre of mass are
ZZ ZZZ
1 1
x̄ = xρ(x, y ) dxdy x̄ = xρ(x, y , z) dxdydz
M Ω M Ω
ZZ ZZZ
1 1
ȳ = y ρ(x, y ) dxdy ȳ = y ρ(x, y , z) dxdydz
M Ω M Ω
ZZZ
1
z̄ = zρ(x, y , z) dxdydz
M Ω
For a lamina or solid body of constant density, the centre of mass is called the
centroid and denoted (xc , yc , zc ).
JM Kress (UNSW Maths & Stats) MATH2111 Integration Semester 1, 2014 56 / 58
Mass, centre of mass, centroid
Example: Find the centre of mass of the trianglular lamina with vertices at
(−1, 0), (0, 1) and (1, 0) with density ρ(x, y ) = y .
y
The symmetry of Ω and ρ gives x̄ = 0.
1
Z Z 1 Z 1−y
x =y −1 Ω x =1−y yρ = y y dx dy
Ω 0 y −1
Z 1 h i1−y
−1 1 x = 2
y x dy
0 y −1
Z Z 1 Z 1−y Z 1
M= y= y dx dy = y 2 (1 − y ) − y 2 (y − 1) dy
Ω 0 y −1 0
Z 1 h i1−y 1
= yx dy = .
0 y −1 6
Z 1 Z
1 1 1 1
= y (1 − y ) − y (y − 1) dy ȳ = yρ = = .
0 M Ω
1 6 2
1 3
= .
3
JM Kress (UNSW Maths & Stats) MATH2111 Integration Semester 1, 2014 57 / 58
y
Z π
1 3
= θ cos θ d θ
1 0 3
= 4 − π2 (int by parts 3 times)
Ω
6(4 − π 2 )
−1 x xc = ' −1.1358.
π3
To calculate the centroid, take ρ = 1. A similar calculation gives
We have already calculated the total
π θ
mass (area) to be A = π 3 /6. π 3 − 6π
Z Z Z
y= r sin θ rdr d θ = .
Ω 0 0 3
Z Z πZ θ
x= r cos θ rdr d θ
Ω 0 0
2(π 2 − 6)
yc = ' 0.78415.
Z π
1 3
θ π2
= r cos θ dθ
0 3 0
JM Kress (UNSW Maths & Stats) MATH2111 Integration Semester 1, 2014 58 / 58