Problem Set 8 - hu-berlin.de

12
Humboldt-Universit¨ at zu Berlin Institut f¨ ur Mathematik C. Wendl, S. Dwivedi, L. Upmeier zu Belzen Funktionalanalysis WiSe 2020–21 Problem Set 8 Due: Thursday, 14.01.2021 (21pts total) Problems marked with p˚q will be graded. Solutions may be written up in German or English and should be submitted electronically via the moodle before the ¨ Ubung on the due date. For problems without p˚q, you do not need to write up your solutions, but it is highly recommended that you think through them before the next Tuesday lecture. You may also use the results of those problems in your written solutions to the graded problems. Problem 1 Fix s 0 and a multi-index of order m :“||P N. (a) Use the Fourier transform and Fourier inverse operators F , F ˚ : L 2 pR n L 2 pR n q to write down an explicit formula for the unique extension of B : S pR n S pR n q to a bounded linear operator B : H s`m pR n H s pR n q. (b) p˚q Show that a sequence f j P H m pR n q converges in the H m -norm to f P H m pR n q if and only if B β f j ÑB β f in the L 2 -norm for all multi-indices β of order |β | § m. [3pts] (c) p˚q Show that for any scalar-valued function f P L 1 pR n q, the convolution with f defines a bounded linear operator H s pR n H s pR n q : g Ñ f ˚ g, and if g P H s`m pR n q, then B pf ˚ gq“ f ˚B g. [4pts] Hint: Problem Set 7 #6 proves the formula z f ˚ g p f p g for f,g P L 1 pR n q. For the present problem, you may assume this formula also holds when f P L 1 pR n q and g P L 2 pR n q; this case was omitted from the lecture due to lack of time, but is proved via an easy density argument as Theorem 8.18 in the lecture notes. (d) Show that for any f P H m pR n q and any approximate identity j : R n Ñr0, 8q with shrinking support, the functions f j :j ˚ f are in H m pR n qX C 8 pR n q and converge in the H m -norm to f as j Ñ8. Problem 2 Suppose P S pR n q satisfies 0 and R n pxq dx 1, and define j pxq :j n pjxq for j P N. (a) p˚q Show that for any s 0 and f P H s pR n q, the sequence j ˚ f P C 8 pR n q satisfies }j ˚ f } H s § }f } H s and j ˚ f H s Ñ f as j Ñ8. Note that the convergence does not follow from Problem 1(d) since we are not assuming s P N. [4pts] (b) Show that the same result holds if j P S pR n q is instead defined as F ˚ j for a sequence of smooth functions j : R n Ñr0, 1s with compact support in the increa- singly large ball B j `1 p0q Ä R n and j | B j p0q 1. Problem 3 Prove that for every m P N, functions f P C m pT n q are also in H m pT n q and the inclusion C m pT n q ã Ñ H m pT n q is continuous. 1 density of mi E spaces Pj f Hep E hfHµ Gm If g Hep E H ftp.HGKP 2

Transcript of Problem Set 8 - hu-berlin.de

Humboldt-Universitat zu Berlin

Institut fur Mathematik

C. Wendl, S. Dwivedi,

L. Upmeier zu Belzen

FunktionalanalysisWiSe 2020–21

Problem Set 8

Due: Thursday, 14.01.2021 (21pts total)

Problems marked with p˚q will be graded. Solutions may be written up in German orEnglish and should be submitted electronically via the moodle before the Ubung on thedue date. For problems without p˚q, you do not need to write up your solutions, but it ishighly recommended that you think through them before the next Tuesday lecture. Youmay also use the results of those problems in your written solutions to the graded problems.

Problem 1Fix s • 0 and a multi-index ↵ of order m :“ |↵| P N.

(a) Use the Fourier transform and Fourier inverse operators F ,F ˚ : L2pRnq Ñ L2pRnq

to write down an explicit formula for the unique extension of B↵ : S pRnq Ñ S pRnqto a bounded linear operator B↵ : Hs`mpRnq Ñ H

spRnq.(b) p˚q Show that a sequence fj P H

mpRnq converges in the Hm-norm to f P H

mpRnqif and only if B�

fj Ñ B�f in the L

2-norm for all multi-indices � of order |�| § m.[3pts]

(c) p˚q Show that for any scalar-valued function f P L1pRnq, the convolution with f

defines a bounded linear operator HspRnq Ñ H

spRnq : g fiÑ f ˚ g, and if g PH

s`mpRnq, then B↵pf ˚ gq “ f ˚ B↵g. [4pts]

Hint: Problem Set 7 #6 proves the formula zf ˚ g “ pfpg for f, g P L1pRnq. For the

present problem, you may assume this formula also holds when f P L1pRnq and

g P L2pRnq; this case was omitted from the lecture due to lack of time, but is proved

via an easy density argument as Theorem 8.18 in the lecture notes.

(d) Show that for any f P HmpRnq and any approximate identity ⇢j : Rn Ñ r0,8q with

shrinking support, the functions fj :“ ⇢j ˚ f are in HmpRnq XC

8pRnq and convergein the H

m-norm to f as j Ñ 8.

Problem 2Suppose ⇢ P S pRnq satisfies ⇢ • 0 and

≥Rn ⇢pxq dx “ 1, and define ⇢jpxq :“ j

n⇢pjxq for

j P N.

(a) p˚q Show that for any s • 0 and f P HspRnq, the sequence ⇢j ˚ f P C

8pRnq satisfies

}⇢j ˚ f}Hs § }f}Hs and ⇢j ˚ fH

s›Ñ f as j Ñ 8.

Note that the convergence does not follow from Problem 1(d) since we are notassuming s P N. [4pts]

(b) Show that the same result holds if ⇢j P S pRnq is instead defined as F ˚ j for a

sequence of smooth functions j : Rn Ñ r0, 1s with compact support in the increa-singly large ball Bj`1p0q Ä Rn and j |Bjp0q ” 1.

Problem 3Prove that for every m P N, functions f P C

mpTnq are also in HmpTnq and the inclusion

CmpTnq ãÑ H

mpTnq is continuous.

1

density

of mi

Espaces

Pj fHepE hfHµ Gm If gHepE H

ftp.HGKP

2

Problem Set 8

Problem 4Fix constants a, b ° 1 with b P N and consider the periodic function fpxq :“ ∞8

k“01ake2⇡ibkx,

which is continuous since the series converges absolutely and uniformly. Prove:

(a) p˚q f P HspS1q if and only if s † logb a. [4pts]

(b) f P C0,↵pS1q for every ↵ P p0, 1q with ↵ § logb a.

Hint: Use Lemma 9.27 from the lecture notes. Note that the partial sums are conti-

nuously di↵erentiable: estimate their C0,1

-norms.

Remark: f is a variant of the function famously introduced by Weierstrass in 1872, which

is of class C1if b † a, but nowhere di↵erentiable if b • a. Part (a) establishes a weak

version of the latter statement by proving f R H1pS1q, which implies f R C

1pS1q via

Problem 3. Notice that while the Sobolev embedding theorem provides a continuous in-

clusion HspS1q ãÑ C

0,↵pS1q whenever ↵ § s ´ 1{2, f turns out to be in a wider range of

Holder spaces than is guaranteed by that theorem. (That is just a coincidence—there is

no interesting phenomenon behind it that I am aware of.)

Problem 5Assume ⌦ is a compact subset of either Rn or Tn, and 0 † ↵ † � § 1.

(a) Prove via the Arzela-Ascoli theorem that the inclusion C0,�p⌦q ãÑ C

0p⌦q is compact.

(b) Show that if fk P C0,�p⌦q is a uniformly C

0,�-bounded sequence that is C0-convergentto f P C

0p⌦q, then f is also in C0,�p⌦q.

Caution: Do not try to prove that fk is also C0,�

-convergent to f—that is not gene-

rally true.

(c) Show that the inclusion C0,�p⌦q ãÑ C

0,↵p⌦q is also compact.Hint: Given fk Ñ f as in part (b), use the relation

|gpxq ´ gpyq||x ´ y|↵ “

ˆ |gpxq ´ gpyq||x ´ y|�

˙↵{�

¨ |gpxq ´ gpyq|1´↵� .

for the functions g :“ f ´ fk.

Problem 6The linear inhomogeneous Cauchy-Riemann equation for functions f : C Ñ C of a complexvariable z “ x ` iy is a first-order PDE taking the form

Bf :“ Bxf ` iByf “ g.

Use the coordinates px, yq to identify C with R2 and consider functions that are fullyperiodic on R2; these are equivalent to complex-valued functions on the torus T2. Prove:

(a) p˚q If f P H1pT2q and g “ Bf P H

mpT2q for some m P N, then f P Hm`1pT2q. [6pts]

(b) If f P C1pT2q and g “ Bf is smooth, then f is smooth.

2

Problem Session 9

f c Cm Tn f is m times continuouslydifferentiable on Tn compact D If12km is bounded D Zdf c L2 Fn

for Klem Fn is of finite measure

f E CmcTn f e HMCFn

want HfHµm E Chthon

tHtm E Kiki

m IIP I LI I al2Tie54FPKEIN

Z It 1k123mN 2

k 322 10ft

2 I 11ktm

gsa12412

C 11811cm

emcFn HMCTT is continuous

a HHuses 241 1542 tax

E 22 b atc 2 ba cc Ea

awant forf c Hscs

lose aTaking logy we get

f c Hscs 0 0 Ss logiathe

Lemma9o2F_ Suppose f k is a sequence of continuous

functions on Sc IR converging uniformly to f

and there are constants a L b Id C O

and BE Oil s t

Hf fall E Ea and lfklco.ptCbk

Then f e 42 for a BItlogab

4lb f x Z Ia e2tibx

Cx 2 ate2Tibx

If tiako Ine

b

atEatnf a.ttN ktl

aAAktica

ECaK

B L

that I tan e Neo

find bounds on the derivative of fincxwhich are finitely many terms

K n

ffx 2mi I t.aze2itibnx

n O

Hhalloo E C k

Using the lemma

f c Cold At Oil d BItlogab

2 1

Itloga ZL

Xt logab logaaf

1

togaslog a

i f E co'd d E logba

Proof of theLemma s want to estimate

H If EC for a figs

ga

we'll prove for oclx yl EC

If fly I fix fkex fish tidy 14dgfi's I

Hex fi I Ife tie IHey fk

I Cbklx yp

Choose KEIN big enoughso that

ie 1 210 1ab

s

Ifan that E 2 Cbk

3aCaKt1

Ifan fusil tix test

Ix y K Ix Tioga

3aeakIx y Hogba

3aC1xy11tl0gbaektl1x

y1Pz3aC1xy1Hl09ba.ab Kal

Zac Ix y11109babktl

a logab ab a Cab osab

H

A alogab

b

ft z C1 212

where X BHloga D

b

a cable eyed compact

uniform boundednessfk t B bounded

dquicontinuous

Fk is uniformly bounded

If Cx Lacy E E whenever Ix y 1 8

o fk e QB

It t.ly EClx y1B

choose c GYPC

D Lk are equicontinuous

7 Arzela Ascoli fee has a convergent

Gubsequence lie CO

COP CO is compact

b f e Bcs

Want Han tealis

E C

Ifan fas I Hix fax 1 few teas fans

ft2 Hex fiacre I ItisCx Tia I

THI HysteresisEc Ixyp mrC Ix YIP

Han Hel E C Ix YIP

af _gIfgeHmf cHM12

2 f 2xf tidyf is a 1storder operator

a If g2 fti2yf g

T2xfti2yf ga

a

2itik I t n IITs Kzfk k

Lignik I 217192Ik k

11h12 Ki'tk

We'll look at Ho norm off sent 1

for vs Smt l to 2 1 Egm

Emi It tE e l j

1412

c z ma 21g I

CE 1k12gD 1g 12

E CE 1142M 19 12

HgHµm

f e Hn F provided gettman

b THEME Tn if g c co getsFS

by parta f e H't felt Fs

0 fe RCF's

If the operator is elliptic the solution will be

as regular as the problem allows

Lf L his order k

tf ge Wn't f e wm kiP

If 0Wm Pcs CK'd