Selected Problems from Evans

Leonardo Abbrescia
November 20, 2013

Chapter 2
Problem 3
Modify the proof of the mean value formulas to show for n 3 that
 
1 1 1
u(0) = g dS + f dy
B(0,r) n(n 2)(n) B(0,r) |y|n2 rn2
provided
in B 0 (0, r)

u = f
.
u=g on B(0, r)
Proof. First we notice that the formula makes intuitive sense. The first term is just the MVP and the second
term is contributed from the inhomogeneity. As per the hint to look at the proof for the MVP, I define

1
(r) = u dS = g dS.
n(n)rn1 B(0,r) B(0,r)

Differentiating, the proof of Theorem 2 yields

r 1
0 (r) = u dy = f dy.
n B(0,r) n(n)rn1 B(0,r)

Fix  > 0. Then notice that r

() = (r) 0 (t) dt. (1)

In order to evaluate the second term of (1), I do integration by parts:
r r
1 1
0 (t) dt = n1
f dy dt (2)
 n(n)  B(0,t) t
r
r
!
1 1 1  1 1
= f dy  f dydt (3)

n(n) 2 n tn2 B(0,t)   B(0,t) n 2 t
n2

r !
1 1 1 1
= f n2 dy dt n2 f dy + n2 f dy . (4)
n(n 2)(n)  B(0,t) t r B(0,r)  B(0,)

It is obvious that the last term is bounded by C 2 0 as  0. As  0, the first term becomes
r
1 1
lim n2
f dy dt = n2
f dy, (5)
0  B(0,t) t B(0,r) |y|

as desired. So in the end, letting 0, (1) yields the desired

 
1 1 1
u(0) = (0) = g dS + f dy
B(0,r) n(n 2)(n) B(0,r) |y|n2 rn2

1
Problem 4
We say v C 2 (U ) is subharmonic if v 0 v 0 in U .

(a) Prove for subharmonic v that

v(x) v dS
B(x,r)

for all B(x, r) U .

(b) Prove that maxU v = maxU v .
(c) Let : R R be smooth and convex. Assume u is harmonic and v := (u). Prove v is subharmonic.
(d) Prove v := |Du|2 is subharmonic, whenever u is harmonic.
Proof. The proof starts out being essentially the same as the MVP for harmonic functions. I define

(r) = v dy.
B(x,r)

From the assumption that v 0, we have

v
0 v dy = dS.
B(x,r) B(x,r)

From the differentiation of (seen in Theorem 2), we have

v
0 (r) = dS 0.
B(x,r)
ffl
Hence, is increasing = v(x) = (0) (r) = B(x,r) v dS. Transforming to polar coordinates in the
same way as the proof on page 26 yields the desired results for (a).

The proof for (b) is identical to that of Theorem 4, except for one little inequality that doesnt matter
in the proof.

(c) Recall Jensens inequality: if f : R R is convex, U Rn is open and bounded, and u : U R

is summable, then  
f u dx f (u) dx.
U U
Then from the MVP of harmonic functions, we have
 
v = (u) = u dx (u) dx = v dx.
U U U

2
Pn(c), (uxi ) is subharmonic. Obviously
(d) We know that if u is harmonic, uxi is harmonic. Then from part
the sum of subharmonic functions is subharmonic, so v = |Du|2 = i=1 (uxi )2 is subharmonic.

Problem 5
Prove that there exists a constant C, depending only on n, such that
 
max |u| C max |g| + max |f | ,
B(0,1) B(0,1) B(0,1)

whenever u is a solution of
in B 0 (0, 1)

u = f
.
u=g on B(0, 1)

2
 M
Proof. Let M = maxB(0,1) f and define v(x) = u(x)+ 2n |x|2 . Then we see that v = uM = f M
0 = v is subharmonic. Also notice that u v x B(0, 1). Then
1 1
max u max v = max v max |g| + M = max |g| + max |f |,
B(0,1) B(0,1) B(0,1) B(0,1) 2n B(0,1) 2n B(0,1)

where we have used part (b) from Problem 4.

Problem 6
Use Poissons formula for the ball to prove

r |x| r + |x|
rn2 u(0) u(x) rn2 u(0)
(r + |x|)n1 (r |x|)n1

whenever u is positive and harmonic in B 0 (0, r). This is an explicit form of Harnacks inequality.
Proof. We recall Poissons formula:

r2 |x|2 g
u(x) = dS (6)
n(n)r B(0,r) |x y|n

solves u = 0 in B 0 (0, r), u = g on B(0, r). Since we are integrating around the boundary, y B(0, r) =
|x y| |x| + r. Then we easily see

r2 |x|2 g (r + |x|)(r |x|) g
u(x) = n
dS = n
dS
n(n)r B(0,r) |x y| n(n)r B(0,r) |x y|

(r + |x|)(r x) g r |x| rn2
n
dS = g dS
n(n)r B(0,r) (|x| + r) (|x| + r)n1 n(n)rn1 B(0,r)
r |x| r |x|
= rn2 g dS = rn2 u(0).
(|x| + r)n1 B(0,r) (|x| + r)n1

The other inequality is easily seen the same way by noticing that |x y| |x| r.

Problem 7
Prove Theorem 15 in section 2.2.4. (Hint: Since u 1 solves (44) for g 1, the theory automatically implies

K(x, y) dS = 1
B(0,1)

for each x B 0 (0, 1).)

Proof. First recall that we must prove that u C (B 0 (0, r)), u = 0 in B 0 (0, r), and limxx0 u(x) = g(x0 )
for each x0 B(0, r). The book defines u by my (6) and

G r2 |x|2 1
(x, y) = K(x, y) =
n(n)r |x y|n

as Poissons kernel. Let G(x, y) be Greens function for the ball. Fixing x, we see that y 7 G(x, y) is
harmonic for x 6= y. Since G(x, y) = G(y, x), we see that x 7 G(x, y) is also harmonic for x 6= y. This
implies that x 7 K(x, y) is harmonic. Then since K(x, y) is both smooth and harmonic, we have that
u C (B 0 (0, r)) and u = 0 by (6) since x K(x, y) = 0.

3
Now we try to show that limxx0 u(x) = g(x0 ) x0 B(0, r). Fix x0 B(0, r). Fix  > 0. Choose
0
> 0 small such that |g(y) g(x )| <  if |y x0 | < . This just comes from continuity of g. Recall that
0
B(0,1)
K = 1. Then for |x x | < , we see
   
   
0 0  0


|u(x) g(x )| =  K(x, y)g(y) dS(y) g(x ) =  K(x, y) g(y) g(x ) dS(y)
  
 B(0,r)   B(0,r) 
   
   
K(x, y) g(y) g(x0 ) dS(y) +  K(x, y) g(y) g(x0 ) dS(y)




   
 B(0,r)|xx0 |   B(0,r)|xx0 |> 

K(x, y) g(y) g(x0 ) dS(y) + K(x, y) g(y) g(x0 ) dS(y)
   

B(0,r)|xx0 | B(0,r)|xx0 |>

= I + J.

Then
I K  dy = .
B(0,r)

For J, recall that g is continuous and bounded:

J kgkL K dS(y)
B(0,r)|xx0 |>
2 2
r |x| 1
= kgkL dS(y)
n(n)r B(0,r)|xx0 |> |x y|n
0 as |x| r.

Then we finally get that |u(x) g(x0 )| 2 for small |x x0 |, and we are done proving the last part.

Problem 9
Let U + denote the open half-ball {x Rn ||x| < 1, xn > 0}. Assume u C 2 (U + ) is harmonic in U + with
u = 0 on U + {xn = 0}. Set

u(x) if xn 0
v(x) :=
u(x1 , . . . , xn1 , xn ) if xn < 0

for x U = B 0 (0, 1). Prove v is harmonic in U .

Proof. If xn 0, uxi xi = vxi xi = v = 0. Let xn < 0. Then notice that vxi = uxi (x1 , . . . , xn1 , xn ), i =
1, . . . , n1. Similarly, vxi xi = uxi xi (x1 , . . . , xn1 , xn ), i = 1, . . . , n1. For i = n, vxn = uxn (x1 , . . . , xn1 , xn )
and vxn xn = uxn xn (x1 , . . . , xn1 , xn ). Since u C 2 (U + ) = v is too = all vxi xi extend continuously
to U + {xn = 0}. So we get that v = u = 0 = v is harmonic for all x U .

Chapter 5
Problem 2
Let U, V be open sets, with V U . Show there exists a smooth function such that 1 on V, = 0
near U .
Proof. Take V W U , and consider

1 xW
W (x) =
0 x
/ W.

4
Let  (x) be the standard mollifier, and define the following convolution:

w := W  (x) = W  (x) dx
R
n

=  (x) dx.
W

It is obvious that w is smooth and has support in W B(0, ). Since W is compact, we can cover it by open
balls Wi such that
[n
W Wi .
i=1
Letting
P i (x) be the standard mollifier for each respective ball Wi , define wi := W i (x). Note that
w
i i is nowhere zero. Then
N
w
P i (x)
X
(x) :=
i=1 i wi

is the desired smooth function such that 1 on V (because V W ), and = 0 near U (because
W U ).

Problem 4
SN
Assume U is bounded and U i=1 Vi . Show there exists C functions i such that

0 1, spt i Vi (i = 1, . . . , N )
PN i
i=1 i = 1 on U .

The functions {i }N
i=1 form a partition of unity.
SN SN
Proof. Take U P i=1 Wi i=1 Vi . For each i, we apply Problem 2 and find a wi 1 on Wi and 0
near Vi . Note that wi is nowhere zero. Hence
wi
i := P
i wi

is smooth and the sum is obviously 1. Also supp i Vi because Wi Vi and {x|i 6= 0} = {x|wi 6= 0
Wi .

Problem 6
Prove directly that if u W 1,p (0, 1) for some 1 < p < , then
 1 1/p
1
1 p 0 p
|u(x) u(y)| |x y| |u | dt
0

for a.e. x, y [0, 1].

Proof. Let v(x) be the weak derivative of u. I define
x
u(x) := v(x) dx.
0

Let Cc (0, 1) be arbitrary. Note that since u, v Lp (0, 1), they are also in L1 (0, 1) = u is absolutely
continuous = u0 = v a.e. Then,
1 1 1 1
u0 dx = v dx = u0 dx = u0 dx,
0 0 0 0

5
where we used the definition of a weak derivative, fundamental theorem of calculus, and integration by parts,
respectively. This implies that
1
(u u)0 dx = 0 = u = u0 a.e.
0

Without loss of generality, let y < x and x, y [0, 1].

|u(x) u(y)| = |u(x) u(y)|

 x 
 
= v dx
y
 x 1/p  x 1 p1
1
1 p
|v|p dx 1 dx
y y
 1 1/p
1
|x y|1 p |v|p dx ,
0

as desired.

Problem 8
Integrate by parts to prove the interpolation inequality:
 1/2  1/2
|Du|2 dx C u2 dx |D2 u|2 dx
U U U

for all u Cc (U ). Assume U is smooth, and prove this inequality if u H 2 (U ) H01 (U ).

Proof. Let u Cc (U ). Then integrating by parts and applying Holders inequality for p = 2 yields

|Du|2 dx = uu dx C |u||D2 u| dx
U U U
 1/2  1/2
2 2 2
C |u | dx |D u| .
U U

There is no boundary term because u has compact support and so the integral along the boundary is zero.
Now let u H01 (U ) H 2 (U ). We will approximate by smooth functions: let {vk }
k=1 Cc (U ) converge to
u in H01 (U ) and {wk }
k=1 C
(U ) converge to v in H 2
(U ). Then Greens formula implies

Dwk Dvk dx = vk wk dx
U
U
C |vk ||D2 wk | dx
U
 1/2  1/2
2 2
C |vk | dx |D wk | dx .
U U

There is no boundary term because vk has compact support. As k , the RHS of the inequality above
converges to CkukL2 (U ) kD2 ukL2 (U ) by the definitions of our sequences vk and wk . All that is left is to show

6
that the LHS of the inequality above converges to kDuk2L2 (U ) as k :

(Dwk Dvk |Du|2 ) dx = (Dwk (Dvk Du) + Du (Dwk Du)) dx
U
U
(|Dwk ||Dvk Du| + |Du||Dwk Du|) dx
U
kDwk kL2 (U ) kDvk DukLp (U ) + kDukL2 (U ) kDwk DukL2 (U )
0 as k .

So for u H01 (U ) H 2 (U ) we are left with

kDuk2L2 (U ) kukL2 (U ) kD2 ukL2 (U ) ,

as desired.

Problem 10
Suppose U is connected and u W 1,p (U ) satisfies

Du = 0 a.e. in U .

Prove u is constant a.e. in U .

Proof. Fix  > 0 such that dist(B(x, r), U ) <  and consider the convolution u :=  u. Recall that
 Cc (U ) and we know that u is smooth. Then I claim that the regular derivative of u is equal to the
convolution of  and the weak derivative of u:

Du =  Du.

We see this explicitly:


Du (x) = D  (x y)u(y) dy = Dx  (x y)u(y) dy
U U

= Dy  (x y)u(y) dy = (1) (1) Du(y) (x y) dy
U U
= [ Du](x).

But from the given that Du = 0 a.e. in U , we conclude that u is constant almost everywhere in our ball
1,p
= u is too because u u in Wloc (U ). The fact that U is connected implies that any two points can be
connected through a single path; a path that can be finitely covered by open balls. Applying the fact that
u is constant a.e. in open balls inside U , and since any one of these balls has a non-empty intersection with
its neighbor, u must be constant in all of U a.e.

Problem 14
Let U be bounded, with a C 1 boundary. Show that a typical function u Lp (U ) (1 p < ) does not
have a trace on U . More precisely, prove that there does not exist a bounded linear operator

T : Lp (U ) Lp (U )

such that T u = u|U whenever u C(U ) Lp (U ).

Proof.

7
Chapter 6
Problem 3
Assume U is connected. A function u H 1 (U ) is a weak solution of Neumanns problem

u = f in U
u (7)
= 0 on U
if
Du Dv dx = f v dx (8)
U U
for all v H 1 (U ). Suppose f L2 (U ). Prove that (7) has a weak solution if and only if

f dx = 0.
U

Proof. Suppose u is a weak solution. i.e., (8) holds. Then in particular, we can make v 1 = Dv = 0 =

0= f dx.
U

Now assume that the integral of f is zero. Note that if u is a weak solution to (7), then u + c is another one
because the c becomes zero when we plug it into (8). Note that by modifying this constant, we can find a
weak solution u satisfying
u dx = 0.
U
Define
A := {f L2 (U )| U f dx = 0}
A0 := {f H 1 (U )| U f dx = 0}.
Then A is a closed subspace because it can be defined as the kernel of a linear functional defined by
f 7 U f dx. Now I define the bilinear form on A0 as

B[u, v] = Du Dv dx,
U
0
v, u A . By Holders inequality, we have
|B[u, v]| kDukL2 (U ) kDvkL2 (U ) kukA0 kvkA0 . (9)
Also note that, by definition,
B[u, u] = kDuk2L2 (U ) .
By our construction of A, we have that u A,

(u)U = u dx = 0.
U

Then we can apply Poincares inequality:

kuk2L2 (U ) = u2 dx = (u (u)U )2 dx C |Du|2 dx = CB[u, u].
U U U

Putting these inequalities together gives us

2 2 2
kukA0 = u + |Du| dx = u2 dx + B[u, u] (C + 1)B[u, u]. (10)
U U

Then by the Lax-Milgram theorem, we have that there is a unique u A0 such that B[u, v] = hf, vi for any
v A0 . But note that we havent used the fact that the integral of f over U is zero! Applying this gives the
required results.

8
Problem 6
Assume u is a smooth solution of
Pn
aij uxi xj = f

Lu = i,j=1 in U
u=0 on U .

where f is bounded. Fix x0 U . A barrier at x0 is a C 2 function w such that

Lw 1 in U , w(x0 ) = 0, w 0 on U .

We want to show that if w is a barrier at x0 , there exists a constant C such that

 
0
 w 0 
|Du(x )| C   (x ) .

Proof. Note that since w(x0 ) = 0, w 0 on U and it is a supersolution, then the weakmaximum principle
imples that
min w = min w,
U U

so w(x0 ) is a minimum. Since f is bounded, we know that kf kL (U ) < . Define v1 := u + kf kL (U ) w and

v2 := u kf kL (U ) w. We now calculate Lv1 and Lv2 :

Lv1 = f + kf kL (U ) Lw f + kf kL (U ) 0
Lv2 = f kf kL (U ) Lw f kf kL (U ) 0.

This shows that v1 is a supersolution and v2 is a subsolution. By the weak maximum principle, v1 attains
its minimum on U , while v2 attains its maximum there. Finally note that
  
v1 U = 0 + kf kL (U ) wU = kf kL (U ) wU
  
v2  = 0 kf kL (U ) w = kf kL (U ) w
U U U

This means that v1 attains a minimum at x0 and v2 attains a maximum at x0 . Then we have

v1 0 u 0 w(x0 )
(x ) = (x ) + kf kL (U ) 0

v2 0 u 0 w(x0 )
(x ) = (x ) 2kf kL (U ) 0.

Putting these two together yield    
 u 0 
 (x ) kf kL (U )  w (x0 ) .
 
   

Problem 7
Assume U is connected. Use (1) energy methods and (2) the maximum principle to show that the only
smooth solutions of the Neumann boundary-value problem

u = 0 in U
u (11)
= 0 on U

are u C, for some constant C.

9
Proof. 1. Let u, v be solutions to the general Neumann boundary-value problem:

u = f in U
u , (12)
= g on U

and let w = u v. Then w solves (11). If w attains its maximum on the interior of U , then it is
a constant and we are done by the strong maximum principle. Assume that it does not attain its
maximum in its interior. By the weak maximum principle, w has a maximum at U , call it x0 . This
implies that w(x0 ) > w(x) x U because we assume that its maximum is not attained on the interior.
If w is not a constant, then we can now use Hopfs Lemma:
w u v
= > 0.

However, since w/ = 0, we have that w is a constant. Note that this proves uniqueness of (12) up
to a constant.
2. Let w solve (11). Then we can use greens formula:

2 w
|Dw| dx = ww dx + w dS.
U U U

The RHS is equal to zero because w solves (11). Then since U
|Dw|2 dx = 0, |Dw|2 is positive, and U
is connected, we have that w is a constant.

10