11.4. MAXIMAL FUNCTIONS 325

Now let g ∈ Cc (Rp) and x /∈ Z. Then from the above observations about continuousfunctions,

µ

([x /∈ Z : limsup

r→0

1µ (B(x,r))

∫B(x,r)

| f (y)− f (x)|dµ (y)> ε

])(11.6)

≤ µ

([x /∈ Z : limsup

r→0

1µ (B(x,r))

∫B(x,r)

| f (y)−g(y)|dµ (y)>ε

2

])+µ

([x /∈ Z : |g(x)− f (x)|> ε

2

]).

≤ µ

([M ( f −g)>

ε

2

])+µ

([| f −g|> ε

2

])(11.7)

Now∫[| f−g|> ε

2 ]| f −g|dµ ≥ ε

2 µ([| f −g|> ε

2

])and so using Claim 1 in 11.7, it follows

that 11.6 is dominated by(

2ε+

Npε

)∫| f −g|dµ. But by Theorem 10.8.7, g can be chosen

to make this as small as desired. Hence 11.6 is 0. Now observe that

µ

([x /∈ Z : limsup

r→0

1µ (B(x,r))

∫B(x,r)

| f (y)− f (x)|dµ (y)> 0])

≤∞

∑k=1

µ

([x /∈ Z : limsup

r→0

1µ (B(x,r))

∫B(x,r)

| f (y)− f (x)|dµ (y)>1k

])= 0

By completeness of µ this implies[x /∈ Z : limsup

r→0

1µ (B(x,r))

∫B(x,r)

| f (y)− f (x)|dµ (y)> 0]

is a set of µ measure zero. ■The following corollary is the main result referred to as the Lebesgue Besicovitch Dif-

ferentiation theorem.

Definition 11.4.3 f ∈ L1loc (Rp,µ) means f XB is in L1 (Rp,µ) whenever B is a

ball.

Theorem 11.4.4 If f ∈ L1loc (Rp,µ), then for µ a.e.x /∈ Z,

limr→0

1µ (B(x,r))

∫B(x,r)

| f (y)− f (x)|dµ (y) = 0 . (11.8)

Proof: If f is replaced by f XB(0,k) then the conclusion 11.8 holds for all x /∈ Fk whereFk is a set of µ measure 0. Letting k = 1,2, · · · , and F ≡ ∪∞

k=1Fk, it follows that F is aset of measure zero and for any x /∈ F , and k ∈ {1,2, · · ·}, 11.8 holds if f is replaced byf XB(0,k). Picking any such x, and letting k > |x|+1, this shows

limr→0

1µ (B(x,r))

∫B(x,r)

| f (y)− f (x)|dµ (y)

= limr→0

1µ (B(x,r))

∫B(x,r)

∣∣ f XB(0,k) (y)− f XB(0,k) (x)∣∣dµ (y) = 0

because for all r small enough, B(x,r)⊆ B(0,k). ■

11.4. MAXIMAL FUNCTIONS 325Now let g € C, (R”) and x ¢ Z. Then from the above observations about continuousfunctions,u(|eez timsup cay dyn fo) —F@IlaH >e]) 16< (|e EZ: ‘lim sup 7B (ar) a) [..ifw)-etwlawo) > 5])ee sei)[wr—2) > 5]) +#([Ir-1> 5]) ui<(Now Siur-ei> lf gldu> a g| > §]) and so using Claim 1 in 11.7, it followsthat 11.6 is dominated by +") J \f—g|du. But by Theorem 10.8.7, g can be chosento make this as small as desired, Hence 11.6 is 0. Now observe thatu(|eez slimsup ae) Ine If (y) — f (w)| du (y) ~0])Le ([ee2 tmp rey Ian lt w—Feiano)> 5] ) =oBy completeness of u this implies1 .x Z :limsup ———— | fy)—fla au(y) >0|fe ez stimsup or J, Mew) Falla)is a set of LW measure zero.The following corollary is the main result referred to as the Lebesgue Besicovitch Dif-ferentiation theorem.Definition 11.4.3 f < L},. (R?,L) means f 2 is in L' (R?,W) whenever B is aball.Theorem 11.4.4 7 f € 1). (R’,u), then for ua.e.x ¢Z,1lim | — f(«)|du(y)=0. 118lim Taceay Ina yf ~F @)IdH0) (11.8)Proof: If f is replaced by f.25(9,,) then the conclusion 11.8 holds for all x ¢ F;, whereFy is a set of uw measure 0. Letting k = 1,2,---, and F = Uy, Fy, it follows that F is aset of measure zero and for any 2 ¢ F, and k € {1,2,---}, 11.8 holds if f is replaced byf %(o,x)- Picking any such z, and letting k > |a| + 1, this shows1lim Ee Ihe pf —F@)|du)1= lim aan) Thee If Zao.) (Y) — f avo, (w) | du (y) =because for all r small enough, B(a,r) C B(0,k).