MEASURE AND INTEGRATION: LECTURE 24

Inequalities Generalized Minkowski inequality. Let Rn = R Rm and z = (x, y) Rn . If Rn C is measurable, then |f (x, y)|p dx : Rm R = fy Lp (R ) : Rm R
R

is R measurable for 1 p < . Assume that fy Lp (R ) dy < .

Rm

Then for a.e. x R , fx (y) : Rm C is in L1 (Rm ). Let F (x) = fx (y) dy.

Rm

Then F (x) : R C is R measurable and we have

F Lp (R ) fy Lp (R ) dy.
Rm

Note this is p 1/p

1/p p |f (x, y)| dx dy.
m f (x, y) dy
dx m
R R R R

We could replace by X, Y nite measure space and Y = {p1 , . . . , pn }, dy the counting measure and get old Minkowskis. Proof. We have

|F (x)|
Rm

|fx (y)| dy, 0. If p = 1, then Fubinis

so without loss of generality, assume f theorem applies; so now let p > 1. Dene g : R Rm R0 by 1/p f (x, y) fy Lp (R ) g(x, y) = 0
Then, for each y,
Date: December 9, 2003.
1

if 0 < fy p < ; if fy p = 0; if fy p = .

MEASURE AND INTEGRATION: LECTURE 24

1/p (1) f (x, y) g(x, y) fy p (2)

for a.e. x, and

gy Lp (R ) = fy 1/p . p Then
F (x) =
Rm

f (x, y) dy
1/p g(x, y) fy p dy

Rm

gx Lp (Rm ) fy p dy

1/p

= gx Lp (Rm ) C 1/p , where C = Rm fy p dy. We now use Fubinis theorem: F (x)p p (R ) L dx

p1 p =c g(x, y) dy dx R Rm p1 p =c g(x, y) dx dy Rm R p1 =c gy p p (R ) dy L m R = cp1 fy Lp (R ) dy Rm
p1 p fy Lp (R ) dy. =c c=c =
R Rm

p/p

gx p p (Rm ) L

Thus,
F (x)Lp (R ) and so
f (x, y) dy

m
p
R Rm

fy Lp (R ) dy

Rm

fy Lp (R ) dy.

L (R )

MEASURE AND INTEGRATION: LECTURE 24

Application. Let f L1 (Rn and g Lp (Rn ). Then f g Lp (Rn ) since p

1/p dx
n f (y)g(x y) dy

Rn R 1/p p |f (y)g(x y)| dx dy
Rn Rn

=
Rn

|f (y)|
Rn

|g(x y)| dx

1/p dy

=
Rn

|
(y)| gp dy
f

= f 1 gp .
Distribution functions. Suppose f : X [0, ] and let {f > t} = (x | f (x) > t}. Theorem 0.1.
and f d =
X 0

{f > t} dt

f d = p
0

{f > t}tp1 dt

More generally, if is dierentiable, then f d = {f > t} (t) dt.

X 0

Proof. We have
|f | dx =

Rn Rn 0

|f (x)|

dt dx

= R =
R Rn
n

[0,f (x)] (t) dt dx

R

[0,f (x)] (t) dx dt

Then

Rn

|f | dx =

{|f |p > t} dt {|t| > t1/p } dt {|f | > s}sp1 ds,

=
0

=p
0

MEASURE AND INTEGRATION: LECTURE 24

letting s = t1/p , so sp = t and dt = psp1 ds. Marcinkiewicz interpolation. Recall the maximal function 1 |f (y)| dy. M f (x) = sup 0<r< (B(x, r)) B(x,r)

Note if f L , then M f f . Thus, M maps L into itself: M : L L . On the other hand, by HardyLittlewood, if f L1 , then
(0.1) {M f > t} 3n
f 1 , t

and M maps L1 to weak L1 . Using a method called Marcinkiewicz interpolation, we prove the following. Theorem 0.2. Let 1 < p < and f Lp . Then M f Lp , and (0.2) M f p C(n, p) f p ,

where C(n, p) is bounded as p and C(n, p) as p 1. Proof. Observe that M f = M |f |, so assume f 0. Choose a constant 0 < c < 1 (we will choose the best c later). For t (0, ), write f = gt + ht , where f (x) f (x) > ct; gt (x) = 0 f (x) ct. So, 0 ht (x) ct for every x, and thus ht L . We have M f M gt + M ht M gt + ct from (0.2). Thus, M f ct M gt , o if M f (x) > t, then (1 c)t M gt (x). Let Et = {f > ct}. Then {M f > t} {M gt > (1 c)t} 3n gt 1 from (0.1) (1 c)t 3n = f dx. (1 c)t Et

MEASURE AND INTEGRATION: LECTURE 24

Thus,
Rn

(M f ) dx = p

{M f > t}tp1 dt 0 3n p p2 t f dx dt 1c 0 Et f (x)/c 3n p p2 = f (x) t dt dx 1 c Rn 0 p1 1 f (x) 3n p dx = f (x) 1 c Rn p1 c 3n p c1p = f (x)p dx 1 c p 1 Rn = C(n, p) f p . p

1/p

Thus, 3n pc1p M f p f p . (1 c)(p 1)

1 as p

Choose c = 1/p = (p 1)/p; this gives the best constant.