Advanced Calculus (II)
WEN-CHINGLIEN
Department of Mathematics National Cheng Kung University
2009
Ch12: Integration on R
12.3: Iterate Integral
We call Z d
c
Z b a
f (x1, . . . ,xn)dxjdxk :=
Z d c
Z b a
f (x1, . . . ,xn)dxj
dxk
an iterated integral when the integrals on the right side exist.
Let R = [a, b] × [c, d ] be a two-dimensional rectangle and suppose that f : R →R is bounded. If f (x , ·) is integrable on [c, d ], then
(22) (L) Z Z
R
f dA ≤ (L) Z b
a
Z d c
f (x , y )dy
dx
≤ (U) Z b
a
Z d c
f (x , y )dy
dx ≤ (U) Z Z
R
f dA.
Proof.
Let Rij = [xi−1,xi] × [yj−1,yj], where {x0, . . . ,xk} is a partition of [a, b] and {y0, . . . ,y`} is a partition of [c, d].
Then G = {Rij :i = 1, 2, . . . , k , j = 1, 2, . . . , `} is a grid on R.
Let R = [a, b] × [c, d ] be a two-dimensional rectangle and suppose that f : R →R is bounded. If f (x , ·) is integrable on [c, d ], then
(22) (L) Z Z
R
f dA ≤ (L) Z b
a
Z d c
f (x , y )dy
dx
≤ (U) Z b
a
Z d c
f (x , y )dy
dx ≤ (U) Z Z
R
f dA.
Proof.
Let Rij = [xi−1,xi] × [yj−1,yj], where {x0, . . . ,xk} is a partition of [a, b] and {y0, . . . ,y`} is a partition of [c, d].
Then G = {Rij :i = 1, 2, . . . , k , j = 1, 2, . . . , `} is a grid on R.
Let R = [a, b] × [c, d ] be a two-dimensional rectangle and suppose that f : R →R is bounded. If f (x , ·) is integrable on [c, d ], then
(22) (L) Z Z
R
f dA ≤ (L) Z b
a
Z d c
f (x , y )dy
dx
≤ (U) Z b
a
Z d c
f (x , y )dy
dx ≤ (U) Z Z
R
f dA.
Proof.
Let Rij = [xi−1,xi] × [yj−1,yj], where {x0, . . . ,xk} is a partition of [a, b] and {y0, . . . ,y`} is a partition of [c, d].
Then G = {Rij :i = 1, 2, . . . , k , j = 1, 2, . . . , `} is a grid on R.
Let ε > 0,choose G so that
(23) U(f , G)−ε < (U) Z Z
R
f dA,
and set
(24) Mij = sup
(x ,y )∈Rij
f (x , y ).
Since (U)Rb
a φ(x )dx =Pk
i=1(U)Rxi
xi−1φ(x )dx and
(U) Z b
a
(φ(x ) + ψ(x ))dx ≤ (U) Z b
a
φ(x )dx + (U) Z b
a
ψ(x )dx
Let ε > 0, choose G so that
(23) U(f , G)−ε < (U) Z Z
R
f dA,
and set
(24) Mij = sup
(x ,y )∈Rij
f (x , y ).
Since (U)Rb
a φ(x )dx =Pk
i=1(U)Rxi
xi−1φ(x )dx and (U)
Z b a
(φ(x ) + ψ(x ))dx ≤ (U) Z b
a
φ(x )dx + (U) Z b
a
ψ(x )dx
Let ε > 0, choose G so that
(23) U(f , G)−ε < (U) Z Z
R
f dA,
and set
(24) Mij = sup
(x ,y )∈Rij
f (x , y ).
Since (U)Rb
a φ(x )dx =Pk
i=1(U)Rxi
xi−1φ(x )dx and (U)
Z b a
(φ(x ) + ψ(x ))dx ≤ (U) Z b
a
φ(x )dx + (U) Z b
a
ψ(x )dx
Let ε > 0, choose G so that
(23) U(f , G)−ε < (U) Z Z
R
f dA,
and set
(24) Mij = sup
(x ,y )∈Rij
f (x , y ).
Since (U)Rb
a φ(x )dx =Pk
i=1(U)Rxi
xi−1φ(x )dx and (U)
Z b a
(φ(x ) + ψ(x ))dx ≤ (U) Z b
a
φ(x )dx + (U) Z b
a
ψ(x )dx
for any bounded function φ and ψ defined on [a, b] (see Exercise 7, p.116),we can write
(U) Z b
a
Z d c
f (x , y )dy
! dx
=
k
X
i=1
(U) Z xi
xi−1
`
X
j=1
Z yj yj−1
f (x , y )dy
! dx
≤
k
X
i=1
`
X
j=1
(U) Z xi
xi−1
Z yj
yj−1
f (x , y )dy
! dx
≤
k
X
i=1
`
X
j=1
Mij(xi− xi−1)(yj− yj−1) =U(f , G).
for any bounded function φ and ψ defined on [a, b] (see Exercise 7, p.116), we can write
(U) Z b
a
Z d c
f (x , y )dy
! dx
=
k
X
i=1
(U) Z xi
xi−1
`
X
j=1
Z yj yj−1
f (x , y )dy
! dx
≤
k
X
i=1
`
X
j=1
(U) Z xi
xi−1
Z yj
yj−1
f (x , y )dy
! dx
≤
k
X
i=1
`
X
j=1
Mij(xi− xi−1)(yj− yj−1) =U(f , G).
for any bounded function φ and ψ defined on [a, b] (see Exercise 7, p.116), we can write
(U) Z b
a
Z d c
f (x , y )dy
! dx
=
k
X
i=1
(U) Z xi
xi−1
`
X
j=1
Z yj yj−1
f (x , y )dy
! dx
≤
k
X
i=1
`
X
j=1
(U) Z xi
xi−1
Z yj
yj−1
f (x , y )dy
! dx
≤
k
X
i=1
`
X
j=1
Mij(xi− xi−1)(yj− yj−1) =U(f , G).
for any bounded function φ and ψ defined on [a, b] (see Exercise 7, p.116), we can write
(U) Z b
a
Z d c
f (x , y )dy
! dx
=
k
X
i=1
(U) Z xi
xi−1
`
X
j=1
Z yj yj−1
f (x , y )dy
! dx
≤
k
X
i=1
`
X
j=1
(U) Z xi
xi−1
Z yj
yj−1
f (x , y )dy
! dx
≤
k
X
i=1
`
X
j=1
Mij(xi− xi−1)(yj− yj−1) =U(f , G).
It follows from (23) that
(U) Z b
a
Z d c
f (x , y )dy
!
dx < (U) Z Z
R
f dA + ε.
Taking the limit of this inequality as ε → 0, we obtain
(U) Z b
a
Z d c
f (x , y )dy
!
dx ≤ (U) Z Z
R
f dA.
Similarly,
(L) Z b
a
Z d c
f (x , y )dy
!
dx ≥ (L) Z Z
R
f dA.
It follows from (23) that
(U) Z b
a
Z d c
f (x , y )dy
!
dx < (U) Z Z
R
f dA + ε.
Taking the limit of this inequality as ε → 0, we obtain
(U) Z b
a
Z d c
f (x , y )dy
!
dx ≤ (U) Z Z
R
f dA.
Similarly,
(L) Z b
a
Z d c
f (x , y )dy
!
dx ≥ (L) Z Z
R
f dA.
It follows from (23) that
(U) Z b
a
Z d c
f (x , y )dy
!
dx < (U) Z Z
R
f dA + ε.
Taking the limit of this inequality as ε → 0, we obtain
(U) Z b
a
Z d c
f (x , y )dy
!
dx ≤ (U) Z Z
R
f dA.
Similarly,
(L) Z b
a
Z d c
f (x , y )dy
!
dx ≥ (L) Z Z
R
f dA.
It follows from (23) that
(U) Z b
a
Z d c
f (x , y )dy
!
dx < (U) Z Z
R
f dA + ε.
Taking the limit of this inequality as ε → 0, we obtain
(U) Z b
a
Z d c
f (x , y )dy
!
dx ≤ (U) Z Z
R
f dA.
Similarly,
(L) Z b
a
Z d c
f (x , y )dy
!
dx ≥ (L) Z Z
R
f dA.
Let R = [a, b] × [c, d ] be a two-dimensional rectangle and let f : R →R. Suppose that f (x , ·) is integrable on [c, d ] for each x ∈ [a, b], that f (·, y ) is integrable on [a, b] for each y ∈ [c, d ], and that f is integrable on R (as a function of two variables). Then
Z Z
R
f dA = Z b
a
Z d c
f (x , y )dydx = Z d
c
Z b a
f (x , y )dxdy .
Note: These hypotheses hold if f is continuous on the rectangle [a, b] × [c, d ].
For each x ∈ [a, b],set g(x ) =Rd
c f (x , y )dy . Since f is integrable on R, Lemma 12.30 implies
Z Z
R
f dA= (U) Z b
a
g(x )dx = (L) Z b
a
g(x )dx .
Hence, g is integrable on [a, b] and the first identity in (25) holds. Reversing the roles of x and y , we obtain
Z Z
R
f dA = Z d
c
Z b a
f (x , y )dx dy .
Hence, the second identity in (25) holds.
For each x ∈ [a, b], set g(x ) =Rd
c f (x , y )dy .Since f is integrable on R, Lemma 12.30 implies
Z Z
R
f dA = (U) Z b
a
g(x )dx =(L) Z b
a
g(x )dx .
Hence, g is integrable on [a, b] and the first identity in (25) holds. Reversing the roles of x and y , we obtain
Z Z
R
f dA = Z d
c
Z b a
f (x , y )dx dy .
Hence, the second identity in (25) holds.
For each x ∈ [a, b], set g(x ) =Rd
c f (x , y )dy . Since f is integrable on R, Lemma 12.30 implies
Z Z
R
f dA= (U) Z b
a
g(x )dx = (L) Z b
a
g(x )dx .
Hence, g is integrable on [a, b] and the first identity in (25) holds. Reversing the roles of x and y , we obtain
Z Z
R
f dA = Z d
c
Z b a
f (x , y )dx dy .
Hence, the second identity in (25) holds.
For each x ∈ [a, b], set g(x ) =Rd
c f (x , y )dy . Since f is integrable on R, Lemma 12.30 implies
Z Z
R
f dA = (U) Z b
a
g(x )dx =(L) Z b
a
g(x )dx .
Hence, g is integrable on [a, b] and the first identity in (25) holds.Reversing the roles of x and y , we obtain
Z Z
R
f dA = Z d
c
Z b a
f (x , y )dx dy .
Hence, the second identity in (25) holds.
For each x ∈ [a, b], set g(x ) =Rd
c f (x , y )dy . Since f is integrable on R, Lemma 12.30 implies
Z Z
R
f dA = (U) Z b
a
g(x )dx = (L) Z b
a
g(x )dx .
Hence, g is integrable on [a, b] and the first identity in (25) holds. Reversing the roles of x and y ,we obtain
Z Z
R
f dA = Z d
c
Z b a
f (x , y )dx dy .
Hence, the second identity in (25) holds.
For each x ∈ [a, b], set g(x ) =Rd
c f (x , y )dy . Since f is integrable on R, Lemma 12.30 implies
Z Z
R
f dA = (U) Z b
a
g(x )dx = (L) Z b
a
g(x )dx .
Hence, g is integrable on [a, b] and the first identity in (25) holds.Reversing the roles of x and y , we obtain
Z Z
R
f dA = Z d
c
Z b a
f (x , y )dx dy .
Hence, the second identity in (25) holds.
For each x ∈ [a, b], set g(x ) =Rd
c f (x , y )dy . Since f is integrable on R, Lemma 12.30 implies
Z Z
R
f dA = (U) Z b
a
g(x )dx = (L) Z b
a
g(x )dx .
Hence, g is integrable on [a, b] and the first identity in (25) holds. Reversing the roles of x and y ,we obtain
Z Z
R
f dA = Z d
c
Z b a
f (x , y )dx dy .
Hence, the second identity in (25) holds.
For each x ∈ [a, b], set g(x ) =Rd
c f (x , y )dy . Since f is integrable on R, Lemma 12.30 implies
Z Z
R
f dA = (U) Z b
a
g(x )dx = (L) Z b
a
g(x )dx .
Hence, g is integrable on [a, b] and the first identity in (25) holds. Reversing the roles of x and y , we obtain
Z Z
R
f dA = Z d
c
Z b a
f (x , y )dx dy .
Hence, the second identity in (25) holds.
For each x ∈ [a, b], set g(x ) =Rd
c f (x , y )dy . Since f is integrable on R, Lemma 12.30 implies
Z Z
R
f dA = (U) Z b
a
g(x )dx = (L) Z b
a
g(x )dx .
Hence, g is integrable on [a, b] and the first identity in (25) holds. Reversing the roles of x and y , we obtain
Z Z
R
f dA = Z d
c
Z b a
f (x , y )dx dy .
Hence, the second identity in (25) holds.
Find
Z 1 0
Z 1 0
y3exy2dydx .
There exists a function f :R2→ R such that f (x, ·) and f (·, y ) are both integrable on [0, 1], but iterated integrals are not equal.
Proof.
Set
f (x , y ) =
22n (x , y ) ∈ [2−n,2−n+1) × [2−n,2−n+1), n ∈N,
−22n+1 (x , y ) ∈ [2−n−1,2−n) × [2−n,2−n+1), n ∈N,
0 otherwise.
There exists a function f :R2→ R such that f (x, ·) and f (·, y ) are both integrable on [0, 1], but iterated integrals are not equal.
Proof.
Set
f (x , y ) =
22n (x , y ) ∈ [2−n,2−n+1) × [2−n,2−n+1), n ∈N,
−22n+1 (x , y ) ∈ [2−n−1,2−n) × [2−n,2−n+1), n ∈N,
0 otherwise.
Notice that for each fixed y0∈ [0, 1), f (x, y0)takes on only two nonzero values and is integrable on [0, 1) in x .For example, if y0 ∈ [2−n,2−n+1), then f (x , y0) =22n for x ∈ [2−n,2−n+1), and f (x , y0) = −22n+1for x ∈ [2−n−1, 2−n);hence, f (x , y0)is bounded on [0, 1), and
(26) Z 1
0
f (x , y0)dx =
Z 2−n+1 2−n
22ndx − Z 2−n
2−n−1
22n+1dx
=2n− 2n =0.
the same is true for f (x0,y ) when x0∈ [0,12), but when x0 ∈ [12,1), f (x0,y ) takes on only nonzero value, namely, f (x0,y ) = 4 when y ∈ [12,1), and equals zero otherwise. It follows that
Notice that for each fixed y0∈ [0, 1), f (x, y0)takes on only two nonzero values and is integrable on [0, 1) in x . For example, if y0 ∈ [2−n,2−n+1),then f (x , y0) =22n for x ∈ [2−n,2−n+1), and f (x , y0) = −22n+1for x ∈ [2−n−1, 2−n); hence, f (x , y0)is bounded on [0, 1), and
(26) Z 1
0
f (x , y0)dx =
Z 2−n+1 2−n
22ndx − Z 2−n
2−n−1
22n+1dx
=2n− 2n =0.
the same is true for f (x0,y ) when x0∈ [0,12), but when x0 ∈ [12,1), f (x0,y ) takes on only nonzero value, namely, f (x0,y ) = 4 when y ∈ [12,1), and equals zero otherwise. It follows that
Notice that for each fixed y0∈ [0, 1), f (x, y0)takes on only two nonzero values and is integrable on [0, 1) in x . For example, if y0 ∈ [2−n,2−n+1), then f (x , y0) =22n for x ∈ [2−n,2−n+1), and f (x , y0) = −22n+1for x ∈ [2−n−1, 2−n);hence, f (x , y0)is bounded on [0, 1), and
(26) Z 1
0
f (x , y0)dx =
Z 2−n+1 2−n
22ndx − Z 2−n
2−n−1
22n+1dx
=2n− 2n =0.
the same is true for f (x0,y ) when x0∈ [0,12), but when x0 ∈ [12,1), f (x0,y ) takes on only nonzero value, namely, f (x0,y ) = 4 when y ∈ [12,1), and equals zero otherwise. It follows that
Notice that for each fixed y0∈ [0, 1), f (x, y0)takes on only two nonzero values and is integrable on [0, 1) in x . For example, if y0 ∈ [2−n,2−n+1), then f (x , y0) =22n for x ∈ [2−n,2−n+1), and f (x , y0) = −22n+1for x ∈ [2−n−1, 2−n); hence, f (x , y0)is bounded on [0, 1), and
(26) Z 1
0
f (x , y0)dx =
Z 2−n+1 2−n
22ndx − Z 2−n
2−n−1
22n+1dx
=2n− 2n =0.
the same is true for f (x0,y ) when x0∈ [0,12), but when x0 ∈ [12,1), f (x0,y ) takes on only nonzero value, namely, f (x0,y ) = 4 when y ∈ [12,1), and equals zero otherwise. It follows that
Notice that for each fixed y0∈ [0, 1), f (x, y0)takes on only two nonzero values and is integrable on [0, 1) in x . For example, if y0 ∈ [2−n,2−n+1), then f (x , y0) =22n for x ∈ [2−n,2−n+1), and f (x , y0) = −22n+1for x ∈ [2−n−1, 2−n); hence, f (x , y0)is bounded on [0, 1), and
(26) Z 1
0
f (x , y0)dx =
Z 2−n+1 2−n
22ndx − Z 2−n
2−n−1
22n+1dx
=2n− 2n =0.
the same is true for f (x0,y ) when x0∈ [0,12),but when x0 ∈ [12,1), f (x0,y ) takes on only nonzero value, namely, f (x0,y ) = 4 when y ∈ [12,1), and equals zero otherwise. It follows that
Notice that for each fixed y0∈ [0, 1), f (x, y0)takes on only two nonzero values and is integrable on [0, 1) in x . For example, if y0 ∈ [2−n,2−n+1), then f (x , y0) =22n for x ∈ [2−n,2−n+1), and f (x , y0) = −22n+1for x ∈ [2−n−1, 2−n); hence, f (x , y0)is bounded on [0, 1), and
(26) Z 1
0
f (x , y0)dx =
Z 2−n+1 2−n
22ndx − Z 2−n
2−n−1
22n+1dx
=2n− 2n =0.
the same is true for f (x0,y ) when x0∈ [0,12), but when x0 ∈ [12,1), f (x0,y ) takes on only nonzero value, namely, f (x0,y ) = 4 when y ∈ [12,1), and equals zero otherwise. It follows that
Notice that for each fixed y0∈ [0, 1), f (x, y0)takes on only two nonzero values and is integrable on [0, 1) in x . For example, if y0 ∈ [2−n,2−n+1), then f (x , y0) =22n for x ∈ [2−n,2−n+1), and f (x , y0) = −22n+1for x ∈ [2−n−1, 2−n); hence, f (x , y0)is bounded on [0, 1), and
(26) Z 1
0
f (x , y0)dx =
Z 2−n+1 2−n
22ndx − Z 2−n
2−n−1
22n+1dx
=2n− 2n =0.
the same is true for f (x0,y ) when x0∈ [0,12),but when x0 ∈ [12,1), f (x0,y ) takes on only nonzero value, namely, f (x0,y ) = 4 when y ∈ [12,1), and equals zero otherwise. It follows that
Notice that for each fixed y0∈ [0, 1), f (x, y0)takes on only two nonzero values and is integrable on [0, 1) in x . For example, if y0 ∈ [2−n,2−n+1), then f (x , y0) =22n for x ∈ [2−n,2−n+1), and f (x , y0) = −22n+1for x ∈ [2−n−1, 2−n); hence, f (x , y0)is bounded on [0, 1), and
(26) Z 1
0
f (x , y0)dx =
Z 2−n+1 2−n
22ndx − Z 2−n
2−n−1
22n+1dx
=2n− 2n =0.
the same is true for f (x0,y ) when x0∈ [0,12), but when x0 ∈ [12,1), f (x0,y ) takes on only nonzero value, namely, f (x0,y ) = 4 when y ∈ [12,1), and equals zero otherwise. It follows that
Z 1 0
Z 1 0
f (x , y )dy dx = Z 1
1 2
Z 1
1 2
4dy dx = 1.
On the other hand, by (26) we have Z 1
0
Z 1 0
f (x , y )dx dy = 0.
Thus the iterated integrals of f are not equal.(Of course, by Fubini’s Theorem, f itself cannot be Riemann
integrable on [0, 1) × [0, 1). In fact, f is not even bounded.)
Z 1 0
Z 1 0
f (x , y )dy dx = Z 1
1 2
Z 1
1 2
4dy dx = 1.
On the other hand, by (26) we have Z 1
0
Z 1 0
f (x , y )dx dy = 0.
Thus the iterated integrals of f are not equal. (Of course, by Fubini’s Theorem, f itself cannot be Riemann
integrable on [0, 1) × [0, 1). In fact, f is not even bounded.)
Z 1 0
Z 1 0
f (x , y )dy dx = Z 1
1 2
Z 1
1 2
4dy dx = 1.
On the other hand, by (26) we have Z 1
0
Z 1 0
f (x , y )dx dy = 0.
Thus the iterated integrals of f are not equal.(Of course, by Fubini’s Theorem, f itself cannot be Riemann
integrable on [0, 1) × [0, 1). In fact, f is not even bounded.)
Z 1 0
Z 1 0
f (x , y )dy dx = Z 1
1 2
Z 1
1 2
4dy dx = 1.
On the other hand, by (26) we have Z 1
0
Z 1 0
f (x , y )dx dy = 0.
Thus the iterated integrals of f are not equal. (Of course, by Fubini’s Theorem, f itself cannot be Riemann
integrable on [0, 1) × [0, 1). In fact, f is not even bounded.)
Let R = [a1,b1] × . . . × [an,bn]be an n-dimensional
rectangle and let f : R →R be integrable on R. If, for each x := (x1, . . . ,xn−1) ∈Rn := [a1,b1] × . . . × [an−1,bn−1], the function f (x.·) is integrable on [an,bn], then
Z bn
an
f (x, t)dt
is integrable on Rn, and
(27)
Z
R
f (x, t)d (x, t) = Z
Rn
Z bn
an
f (x, t)dt d x.
By repeating the argument of Lemma 12.30,we have
(L) Z
R
f (x, t)d (x, t) ≤ (L) Z
Rn
Z bn
an
f (x, t)dt d x
≤ (U) Z
Rn
Z bn
an
f (x, t)dt d x
≤ (U) Z
R
f (x, t)d (x, t)
for any bounded f . Since f is integrable on R, it follows that (27) holds.
By repeating the argument of Lemma 12.30, we have
(L) Z
R
f (x, t)d (x, t)≤ (L) Z
Rn
Z bn
an
f (x, t)dt d x
≤ (U) Z
Rn
Z bn
an
f (x, t)dt d x
≤ (U) Z
R
f (x, t)d (x, t)
for any bounded f . Since f is integrable on R, it follows that (27) holds.
By repeating the argument of Lemma 12.30, we have
(L) Z
R
f (x, t)d (x, t) ≤ (L) Z
Rn
Z bn
an
f (x, t)dt d x
≤ (U) Z
Rn
Z bn
an
f (x, t)dt d x
≤ (U) Z
R
f (x, t)d (x, t)
for any bounded f .Since f is integrable on R, it follows that (27) holds.
By repeating the argument of Lemma 12.30, we have
(L) Z
R
f (x, t)d (x, t) ≤ (L) Z
Rn
Z bn
an
f (x, t)dt d x
≤ (U) Z
Rn
Z bn
an
f (x, t)dt d x
≤ (U) Z
R
f (x, t)d (x, t)
for any bounded f . Since f is integrable on R, it follows that (27) holds.
By repeating the argument of Lemma 12.30, we have
(L) Z
R
f (x, t)d (x, t) ≤ (L) Z
Rn
Z bn
an
f (x, t)dt d x
≤ (U) Z
Rn
Z bn
an
f (x, t)dt d x
≤ (U) Z
R
f (x, t)d (x, t)
for any bounded f .Since f is integrable on R, it follows that (27) holds.
By repeating the argument of Lemma 12.30, we have
(L) Z
R
f (x, t)d (x, t) ≤ (L) Z
Rn
Z bn
an
f (x, t)dt d x
≤ (U) Z
Rn
Z bn
an
f (x, t)dt d x
≤ (U) Z
R
f (x, t)d (x, t)
for any bounded f . Since f is integrable on R, it follows that (27) holds.
Let E be a projectable region inRn generated by j, H, φ, and ψ. Then E is a Jordan region inRn. Moreover, if f : E →R is continuous on E , then
(28) Z
E
f (x)d x
= Z
H
Z ψ(x1,...,bxj,...,xn) φ(x1,...,bxj,...,xn)
f (x1, . . . ,xn)dxj
d (x1, . . . ,xbj, . . . ,xn).