• Aucun résultat trouvé

Spectral Theorem for compact self-adjoint operators

N/A
N/A
Protected

Academic year: 2022

Partager "Spectral Theorem for compact self-adjoint operators"

Copied!
53
0
0

Texte intégral

(1)

Spectral Theorem for compact self-adjoint operators

Philippe Jaming

Universit ´e de Bordeaux

http://www.u-bordeaux.fr/˜ pjaming/enseignement/M1.html

Master Math ´ematiques et Applications

M1 Math ´ematiques fondamentales & M1 Analyse, EDP, probabilit ´es Lecture : Introduction to spectral analysis

(2)

Preliminary warning

This video is a complement to the lecture notes available at

http://www.u-bordeaux.fr/˜ pjaming/enseignement/M1.html

1. Statement of the Spectral Theorem 2. An example

3. Toolbox 4. The proof

Philippe Jaming (Universit ´e de Bordeaux) Spectral Theorem 2 Master Math & Applications 2 / 11

(3)

Preliminary warning

This video is a complement to the lecture notes available at

http://www.u-bordeaux.fr/˜ pjaming/enseignement/M1.html

1. Statement of the Spectral Theorem 2. An example

3. Toolbox 4. The proof

(4)

Preliminary warning

This video is a complement to the lecture notes available at

http://www.u-bordeaux.fr/˜ pjaming/enseignement/M1.html

1. Statement of the Spectral Theorem 2. An example

3. Toolbox 4. The proof

Philippe Jaming (Universit ´e de Bordeaux) Spectral Theorem 2 Master Math & Applications 2 / 11

(5)

Preliminary warning

This video is a complement to the lecture notes available at

http://www.u-bordeaux.fr/˜ pjaming/enseignement/M1.html

1. Statement of the Spectral Theorem 2. An example

3. Toolbox 4. The proof

(6)

Preliminary warning

This video is a complement to the lecture notes available at

http://www.u-bordeaux.fr/˜ pjaming/enseignement/M1.html

1. Statement of the Spectral Theorem 2. An example

3. Toolbox 4. The proof

Philippe Jaming (Universit ´e de Bordeaux) Spectral Theorem 2 Master Math & Applications 2 / 11

(7)

Spectral Theorem for compact self-adjoint operators

Aim: give the main tools to prove the spectral theorem Theorem (Spectral Theorem)

H , infinite dimensional, Hilbert space. T :H →Hcompact, self-adjoint, one-to-oneoperator.

∃(ek)k∈Northonormal basis of H;

∃(λk)k∈Nreal, ,λk decreasingλk →0;

Tx =X

k∈N

λkhx,ekiek.

(8)

Spectral Theorem for compact self-adjoint operators

Aim: give the main tools to prove the spectral theorem Theorem (Spectral Theorem)

H , infinite dimensional, Hilbert space. T :H →Hcompact, self-adjoint, one-to-oneoperator.

∃(ek)k∈Northonormal basis of H;

∃(λk)k∈Nreal, ,λk decreasingλk →0;

Tx =X

k∈N

λkhx,ekiek.

Philippe Jaming (Universit ´e de Bordeaux) Spectral Theorem 2 Master Math & Applications 3 / 11

(9)

Spectral Theorem for compact self-adjoint operators

Aim: give the main tools to prove the spectral theorem Theorem (Spectral Theorem)

H , infinite dimensional, Hilbert space. T :H →Hcompact, self-adjoint, one-to-oneoperator.

∃(ek)k∈Northonormal basis of H;

∃(λk)k∈Nreal, ,λk decreasingλk →0;

Tx =X

k∈N

λkhx,ekiek.

(10)

Spectral Theorem for compact self-adjoint operators

Aim: give the main tools to prove the spectral theorem Theorem (Spectral Theorem)

H , infinite dimensional, Hilbert space. T :H →Hcompact, self-adjoint, one-to-oneoperator.

∃(ek)k∈Northonormal basis of H;

∃(λk)k∈Nreal,,λk decreasingλk →0;

Tx =X

k∈N

λkhx,ekiek.

Philippe Jaming (Universit ´e de Bordeaux) Spectral Theorem 2 Master Math & Applications 3 / 11

(11)

Spectral Theorem for compact self-adjoint operators

Aim: give the main tools to prove the spectral theorem Theorem (Spectral Theorem)

H , infinite dimensional, Hilbert space. T :H →Hcompact, self-adjoint, one-to-oneoperator.

∃(ek)k∈Northonormal basis of H;

∃(λk)k∈Nreal, ,λk decreasingλk →0;

Tx =X

k∈N

λkhx,ekiek.

(12)

Spectral Theorem for compact self-adjoint operators

Aim: give the main tools to prove the spectral theorem Theorem (Spectral Theorem)

H , infinite dimensional, Hilbert space. T :H →Hcompact, self-adjoint, one-to-oneoperator.

∃(ek)k∈Northonormal basis of H;

∃(λk)k∈Nreal, ,λk decreasingλk →0;

Tx =X

k∈N

λkhx,ekiek.

Philippe Jaming (Universit ´e de Bordeaux) Spectral Theorem 2 Master Math & Applications 3 / 11

(13)

Spectral theorem - comments

T has a matrix of the form

kerT E1 E2 · · ·

e1,1, . . . ,en1,1 e1,2, . . . ,en2,2

kerT 0 0 0 · · ·

e1,1 ... en1,1

0

λ1 0

. ..

0 λ1

0 e1,2

... en2,2

0

λ2 0

. ..

0 λ2

0 0

. .. 0

. ..

0 . ..

(14)

First step

The proof is an induction

i F1=EkTk⊕E−kTk

ii π1±⊥-projection onE±kTkΠ11+1projection onF1;

iii m1= dimEkTk,m2= dimE−kTk etd1=m1+m2≥1;

iv e1, . . . ,em1 ONB ofEkTk &em1+1, . . . ,ed1 ONB ofE−kTk e1, . . . ,ed1 ONB ofF1;

v λ1=· · ·=λm1 =kTk,λm1+1=· · ·=λd1 =−kTk.

T = Π11

| {z }

Pd1

k=1λkhx,ekiek

+ (I−Π1)T(I−Π1)

| {z }

=T1

Philippe Jaming (Universit ´e de Bordeaux) Spectral Theorem 2 Master Math & Applications 5 / 11

(15)

First step

The proof is an induction

i F1=EkTk⊕E−kTk

ii π1±⊥-projection onE±kTkΠ11+1projection onF1;

iii m1= dimEkTk,m2= dimE−kTk etd1=m1+m2≥1;

iv e1, . . . ,em1 ONB ofEkTk &em1+1, . . . ,ed1 ONB ofE−kTk e1, . . . ,ed1 ONB ofF1;

v λ1=· · ·=λm1 =kTk,λm1+1=· · ·=λd1 =−kTk.

T = Π11

| {z }

Pd1

k=1λkhx,ekiek

+ (I−Π1)T(I−Π1)

| {z }

=T1

(16)

First step

The proof is an induction

i F1=EkTk⊕E−kTk

ii π1±⊥-projection onE±kTkΠ11+1projection onF1;

iii m1= dimEkTk,m2= dimE−kTk etd1=m1+m2≥1;

iv e1, . . . ,em1 ONB ofEkTk &em1+1, . . . ,ed1 ONB ofE−kTk e1, . . . ,ed1 ONB ofF1;

v λ1=· · ·=λm1 =kTk,λm1+1=· · ·=λd1 =−kTk.

T = Π11

| {z }

Pd1

k=1λkhx,ekiek

+ (I−Π1)T(I−Π1)

| {z }

=T1

Philippe Jaming (Universit ´e de Bordeaux) Spectral Theorem 2 Master Math & Applications 5 / 11

(17)

First step

The proof is an induction

i F1=EkTk⊕E−kTk

ii π1±⊥-projection onE±kTkΠ11+1projection onF1;

iii m1= dimEkTk,m2= dimE−kTk etd1=m1+m2≥1;

iv e1, . . . ,em1 ONB ofEkTk &em1+1, . . . ,ed1 ONB ofE−kTk e1, . . . ,ed1 ONB ofF1;

v λ1=· · ·=λm1 =kTk,λm1+1=· · ·=λd1 =−kTk.

T = Π11

| {z }

Pd1

k=1λkhx,ekiek

+ (I−Π1)T(I−Π1)

| {z }

=T1

(18)

First step

The proof is an induction

i F1=EkTk⊕E−kTk

ii π1±⊥-projection onE±kTkΠ11+1projection onF1;

iii m1= dimEkTk,m2= dimE−kTk etd1=m1+m2≥1;

iv e1, . . . ,em1 ONB ofEkTk &em1+1, . . . ,ed1 ONB ofE−kTk e1, . . . ,ed1 ONB ofF1;

v λ1=· · ·=λm1 =kTk,λm1+1=· · ·=λd1 =−kTk.

T = Π11

| {z }

Pd1

k=1λkhx,ekiek

+ (I−Π1)T(I−Π1)

| {z }

=T1

Philippe Jaming (Universit ´e de Bordeaux) Spectral Theorem 2 Master Math & Applications 5 / 11

(19)

First step

The proof is an induction

i F1=EkTk⊕E−kTk

ii π1±⊥-projection onE±kTkΠ11+1projection onF1;

iii m1= dimEkTk,m2= dimE−kTk etd1=m1+m2≥1;

iv e1, . . . ,em1 ONB ofEkTk &em1+1, . . . ,ed1 ONB ofE−kTk e1, . . . ,ed1 ONB ofF1;

v λ1=· · ·=λm1 =kTk,λm1+1=· · ·=λd1 =−kTk.

T = Π11

| {z }

Pd1

k=1λkhx,ekiek

+ (I−Π1)T(I−Π1)

| {z }

=T1

(20)

First step

The proof is an induction

i F1=EkTk⊕E−kTk

ii π1±⊥-projection onE±kTkΠ11+1projection onF1;

iii m1= dimEkTk,m2= dimE−kTk etd1=m1+m2≥1;

iv e1, . . . ,em1 ONB ofEkTk &em1+1, . . . ,ed1 ONB ofE−kTk e1, . . . ,ed1 ONB ofF1;

v λ1=· · ·=λm1 =kTk,λm1+1=· · ·=λd1 =−kTk.

T = Π11

| {z }

Pd1

k=1λkhx,ekiek

+ (I−Π1)T(I−Π1)

| {z }

=T1

Philippe Jaming (Universit ´e de Bordeaux) Spectral Theorem 2 Master Math & Applications 5 / 11

(21)

First step

The proof is an induction

i F1=EkTk⊕E−kTk

ii π1±⊥-projection onE±kTkΠ11+1projection onF1;

iii m1= dimEkTk,m2= dimE−kTk etd1=m1+m2≥1;

iv e1, . . . ,em1 ONB ofEkTk &em1+1, . . . ,ed1 ONB ofE−kTk e1, . . . ,ed1 ONB ofF1;

v λ1=· · ·=λm1 =kTk,λm1+1=· · ·=λd1 =−kTk.

T = Π11

| {z }

Pd1

k=1λkhx,ekiek

+ (I−Π1)T(I−Π1)

| {z }

=T1

(22)

First step

The proof is an induction

i F1=EkTk⊕E−kTk

ii π1±⊥-projection onE±kTkΠ11+1projection onF1;

iii m1= dimEkTk,m2= dimE−kTk etd1=m1+m2≥1;

iv e1, . . . ,em1 ONB ofEkTk &em1+1, . . . ,ed1 ONB ofE−kTk e1, . . . ,ed1 ONB ofF1;

v λ1=· · ·=λm1 =kTk,λm1+1=· · ·=λd1 =−kTk.

T = Π11

| {z }

Pd1

k=1λkhx,ekiek

+ (I−Π1)T(I−Π1)

| {z }

=T1

Philippe Jaming (Universit ´e de Bordeaux) Spectral Theorem 2 Master Math & Applications 5 / 11

(23)

Spectral theorem - comments

T has a matrix of the form

E1 E2 · · ·

e1, . . . ,en1 en1+1, . . . ,en2

e1 ... en1

kTk 0

. ..

0 −kTk

0

en1+1 ... en2

λ2 0

. ..

0 λ2

0

. .. 0

. ..

0 . ..

(24)

Second step

T1= (I−Π1)T(I−Π1)self-adjoint,compact,kT1k ≤ kTk.

Observation kT1k<kTk

Proof: If not,∃x,kxk=1,T1x =kTkx (orT1x =−kTkx)

—(I−Π1)x =x otherwisek(I−π1)xk<kxkand then

kTk=kT1xk=k(I−Π1)T(I−Π1)xk ≤ kTkk(I−Π1)xk<kTk

— sokTkx =T1x = (I−Π1)T(I−Π1)x = (I−Π1)Tx =Tx−Π1Tx Tx = kTkx

| {z }

∈Im(I−Π1)

1Tx

Philippe Jaming (Universit ´e de Bordeaux) Spectral Theorem 2 Master Math & Applications 7 / 11

(25)

Second step

T1= (I−Π1)T(I−Π1)self-adjoint,compact,kT1k ≤ kTk.

Observation kT1k<kTk

Proof: If not,∃x,kxk=1,T1x =kTkx (orT1x =−kTkx)

—(I−Π1)x =x otherwisek(I−π1)xk<kxkand then

kTk=kT1xk=k(I−Π1)T(I−Π1)xk ≤ kTkk(I−Π1)xk<kTk

— sokTkx =T1x = (I−Π1)T(I−Π1)x = (I−Π1)Tx =Tx−Π1Tx Tx = kTkx

| {z }

∈Im(I−Π1)

1Tx

(26)

Second step

T1= (I−Π1)T(I−Π1)self-adjoint,compact,kT1k ≤ kTk.

Observation kT1k<kTk

Proof: If not,∃x,kxk=1,T1x =kTkx (orT1x =−kTkx)

—(I−Π1)x =x otherwisek(I−π1)xk<kxkand then

kTk=kT1xk=k(I−Π1)T(I−Π1)xk ≤ kTkk(I−Π1)xk<kTk

— sokTkx =T1x = (I−Π1)T(I−Π1)x = (I−Π1)Tx =Tx−Π1Tx Tx = kTkx

| {z }

∈Im(I−Π1)

1Tx

Philippe Jaming (Universit ´e de Bordeaux) Spectral Theorem 2 Master Math & Applications 7 / 11

(27)

Second step

T1= (I−Π1)T(I−Π1)self-adjoint,compact,kT1k ≤ kTk.

Observation kT1k<kTk

Proof: If not,∃x,kxk=1,T1x =kTkx (orT1x =−kTkx)

—(I−Π1)x =x otherwisek(I−π1)xk<kxkand then

kTk=kT1xk=k(I−Π1)T(I−Π1)xk ≤ kTkk(I−Π1)xk<kTk

— sokTkx =T1x = (I−Π1)T(I−Π1)x = (I−Π1)Tx =Tx−Π1Tx Tx = kTkx

| {z }

∈Im(I−Π1)

1Tx

(28)

Second step

T1= (I−Π1)T(I−Π1)self-adjoint,compact,kT1k ≤ kTk.

Observation kT1k<kTk

Proof: If not,∃x,kxk=1,T1x =kTkx (orT1x =−kTkx)

—(I−Π1)x =x otherwisek(I−π1)xk<kxkand then

kTk=kT1xk=k(I−Π1)T(I−Π1)xk≤ kTkk(I−Π1)xk<kTk

— sokTkx =T1x = (I−Π1)T(I−Π1)x = (I−Π1)Tx =Tx−Π1Tx Tx = kTkx

| {z }

∈Im(I−Π1)

1Tx

Philippe Jaming (Universit ´e de Bordeaux) Spectral Theorem 2 Master Math & Applications 7 / 11

(29)

Second step

T1= (I−Π1)T(I−Π1)self-adjoint,compact,kT1k ≤ kTk.

Observation kT1k<kTk

Proof: If not,∃x,kxk=1,T1x =kTkx (orT1x =−kTkx)

—(I−Π1)x =x otherwisek(I−π1)xk<kxkand then

kTk=kT1xk=k(I−Π1)T(I−Π1)xk ≤ kTkk(I−Π1)xk<kTk

— sokTkx =T1x = (I−Π1)T(I−Π1)x = (I−Π1)Tx =Tx−Π1Tx Tx = kTkx

| {z }

∈Im(I−Π1)

1Tx

(30)

Second step

T1= (I−Π1)T(I−Π1)self-adjoint,compact,kT1k ≤ kTk.

Observation kT1k<kTk

Proof: If not,∃x,kxk=1,T1x =kTkx (orT1x =−kTkx)

—(I−Π1)x =x otherwisek(I−π1)xk<kxkand then

kTk=kT1xk=k(I−Π1)T(I−Π1)xk ≤ kTkk(I−Π1)xk<kTk

— sokTkx =T1x = (I−Π1)T(I−Π1)x = (I−Π1)Tx =Tx−Π1Tx Tx = kTkx

| {z }

∈Im(I−Π1)

1Tx

Philippe Jaming (Universit ´e de Bordeaux) Spectral Theorem 2 Master Math & Applications 7 / 11

(31)

Second step

T1= (I−Π1)T(I−Π1)self-adjoint,compact,kT1k ≤ kTk.

Observation kT1k<kTk

Proof: If not,∃x,kxk=1,T1x =kTkx (orT1x =−kTkx)

—(I−Π1)x =x otherwisek(I−π1)xk<kxkand then

kTk=kT1xk=k(I−Π1)T(I−Π1)xk ≤ kTkk(I−Π1)xk<kTk

— sokTkx =T1x = (I−Π1)T(I−Π1)x = (I−Π1)Tx =Tx−Π1Tx Tx = kTkx

| {z }

∈Im(I−Π1)

1Tx

(32)

Second step

(I−Π1)x =x &Tx = kTkx

| {z }

∈Im(I−Π1)

1Tx – thus

kTxk2=kTk2kxk2+kΠ1Txk2 Π1Tx =0Tx =kTkx i.e. x∈EkTk a contradiction

Philippe Jaming (Universit ´e de Bordeaux) Spectral Theorem 2 Master Math & Applications 8 / 11

(33)

Second step

(I−Π1)x =x &Tx = kTkx

| {z }

∈Im(I−Π1)

1Tx – thus

kTxk2=kTk2kxk2+kΠ1Txk2 Π1Tx =0Tx =kTkx i.e. x∈EkTk a contradiction

(34)

Second step

(I−Π1)x =x &Tx = kTkx

| {z }

∈Im(I−Π1)

1Tx – thus

kTxk2=kTk2kxk2+kΠ1Txk2 Π1Tx =0Tx =kTkx i.e. x∈EkTk a contradiction

Philippe Jaming (Universit ´e de Bordeaux) Spectral Theorem 2 Master Math & Applications 8 / 11

(35)

Induction

T =

n−1

X

k=1

Πkk+ I−

n−1

X

k=1

Πk

!

T I−

n−1

X

k=1

Πk

!

| {z }

=Tn−1

i Fn=EkTn−1k⊕E−kTn−1k

ii πn±⊥-projection onE±kTn−1kΠn+nnprojection onFn;

iii m2n−1= dimEkTnk,m2n= dimE−kTnk etdn=m2n−1+m2n≥1;

iv edn−1+1, . . . ,edn−1+m2n−1 ONB ofEkTnk &edn−1+m2n−1+1, . . . ,edn ONB ofE−kTnkedn−1+1, . . . ,edn ONB ofF1;

v λdn−1+1=· · ·=λdn−1+m2n−1 =kTn−1k, λdn−1+m2n−1+1=· · ·=λdn =−kTn−1k.

T =Pn

k=1Πkk + (I−

n

X

k=1

Πk)T(I−

n

X

k=1

Πk)

(36)

Induction

T =

n−1

X

k=1

Πkk+ I−

n−1

X

k=1

Πk

!

T I−

n−1

X

k=1

Πk

!

| {z }

=Tn−1

i Fn=EkTn−1k⊕E−kTn−1k

ii πn±⊥-projection onE±kTn−1kΠn+nnprojection onFn;

iii m2n−1= dimEkTnk,m2n= dimE−kTnk etdn=m2n−1+m2n≥1;

iv edn−1+1, . . . ,edn−1+m2n−1 ONB ofEkTnk &edn−1+m2n−1+1, . . . ,edn ONB ofE−kTnkedn−1+1, . . . ,edn ONB ofF1;

v λdn−1+1=· · ·=λdn−1+m2n−1 =kTn−1k, λdn−1+m2n−1+1=· · ·=λdn =−kTn−1k.

T =Pn

k=1Πkk + (I−

n

X

k=1

Πk)T(I−

n

X

k=1

Πk)

| {z }

=Tn

Philippe Jaming (Universit ´e de Bordeaux) Spectral Theorem 2 Master Math & Applications 9 / 11

(37)

Induction

T =

n−1

X

k=1

Πkk+ I−

n−1

X

k=1

Πk

!

T I−

n−1

X

k=1

Πk

!

| {z }

=Tn−1

i Fn=EkTn−1k⊕E−kTn−1k

ii πn±⊥-projection onE±kTn−1kΠn+nnprojection onFn;

iii m2n−1= dimEkTnk,m2n= dimE−kTnk etdn=m2n−1+m2n≥1;

iv edn−1+1, . . . ,edn−1+m2n−1 ONB ofEkTnk &edn−1+m2n−1+1, . . . ,edn ONB ofE−kTnkedn−1+1, . . . ,edn ONB ofF1;

v λdn−1+1=· · ·=λdn−1+m2n−1 =kTn−1k, λdn−1+m2n−1+1=· · ·=λdn =−kTn−1k.

T =Pn

k=1Πkk + (I−

n

X

k=1

Πk)T(I−

n

X

k=1

Πk)

(38)

Induction

T =

n−1

X

k=1

Πkk+ I−

n−1

X

k=1

Πk

!

T I−

n−1

X

k=1

Πk

!

| {z }

=Tn−1

i Fn=EkTn−1k⊕E−kTn−1k

ii πn±⊥-projection onE±kTn−1kΠn+nnprojection onFn;

iii m2n−1= dimEkTnk,m2n= dimE−kTnk etdn=m2n−1+m2n≥1;

iv edn−1+1, . . . ,edn−1+m2n−1 ONB ofEkTnk &edn−1+m2n−1+1, . . . ,edn ONB ofE−kTnkedn−1+1, . . . ,edn ONB ofF1;

v λdn−1+1=· · ·=λdn−1+m2n−1 =kTn−1k, λdn−1+m2n−1+1=· · ·=λdn =−kTn−1k.

T =Pn

k=1Πkk + (I−

n

X

k=1

Πk)T(I−

n

X

k=1

Πk)

| {z }

=Tn

Philippe Jaming (Universit ´e de Bordeaux) Spectral Theorem 2 Master Math & Applications 9 / 11

(39)

Induction

T =

n−1

X

k=1

Πkk+ I−

n−1

X

k=1

Πk

!

T I−

n−1

X

k=1

Πk

!

| {z }

=Tn−1

i Fn=EkTn−1k⊕E−kTn−1k

ii πn±⊥-projection onE±kTn−1kΠn+nnprojection onFn;

iii m2n−1= dimEkTnk,m2n= dimE−kTnk etdn=m2n−1+m2n≥1;

iv edn−1+1, . . . ,edn−1+m2n−1 ONB ofEkTnk &edn−1+m2n−1+1, . . . ,edn ONB ofE−kTnkedn−1+1, . . . ,edn ONB ofF1;

v λdn−1+1=· · ·=λdn−1+m2n−1 =kTn−1k, λdn−1+m2n−1+1=· · ·=λdn =−kTn−1k.

T =Pn

k=1Πkk + (I−

n

X

k=1

Πk)T(I−

n

X

k=1

Πk)

(40)

Induction

T =

n−1

X

k=1

Πkk+ I−

n−1

X

k=1

Πk

!

T I−

n−1

X

k=1

Πk

!

| {z }

=Tn−1

i Fn=EkTn−1k⊕E−kTn−1k

ii πn±⊥-projection onE±kTn−1kΠn+nnprojection onFn;

iii m2n−1= dimEkTnk,m2n= dimE−kTnk etdn=m2n−1+m2n≥1;

iv edn−1+1, . . . ,edn−1+m2n−1 ONB ofEkTnk &edn−1+m2n−1+1, . . . ,edn ONB ofE−kTnkedn−1+1, . . . ,edn ONB ofF1;

v λdn−1+1=· · ·=λdn−1+m2n−1 =kTn−1k, λdn−1+m2n−1+1=· · ·=λdn =−kTn−1k.

T =Pn

k=1Πkk + (I−

n

X

k=1

Πk)T(I−

n

X

k=1

Πk)

| {z }

=Tn

Philippe Jaming (Universit ´e de Bordeaux) Spectral Theorem 2 Master Math & Applications 9 / 11

(41)

Induction

T =

n−1

X

k=1

Πkk+ I−

n−1

X

k=1

Πk

!

T I−

n−1

X

k=1

Πk

!

| {z }

=Tn−1

i Fn=EkTn−1k⊕E−kTn−1k

ii πn±⊥-projection onE±kTn−1kΠn+nnprojection onFn;

iii m2n−1= dimEkTnk,m2n= dimE−kTnk etdn=m2n−1+m2n≥1;

iv edn−1+1, . . . ,edn−1+m2n−1 ONB ofEkTnk &edn−1+m2n−1+1, . . . ,edn ONB ofE−kTnkedn−1+1, . . . ,edn ONB ofF1;

v λdn−1+1=· · ·=λdn−1+m2n−1 =kTn−1k, λdn−1+m2n−1+1=· · ·=λdn =−kTn−1k.

T =Pn

k=1Πkk + (I−

n

X

k=1

Πk)T(I−

n

X

k=1

Πk)

(42)

Induction

T =

n−1

X

k=1

Πkk+ I−

n−1

X

k=1

Πk

!

T I−

n−1

X

k=1

Πk

!

| {z }

=Tn−1

i Fn=EkTn−1k⊕E−kTn−1k

ii πn±⊥-projection onE±kTn−1kΠn+nnprojection onFn;

iii m2n−1= dimEkTnk,m2n= dimE−kTnk etdn=m2n−1+m2n≥1;

iv edn−1+1, . . . ,edn−1+m2n−1 ONB ofEkTnk &edn−1+m2n−1+1, . . . ,edn ONB ofE−kTnkedn−1+1, . . . ,edn ONB ofF1;

v λdn−1+1=· · ·=λdn−1+m2n−1 =kTn−1k, λdn−1+m2n−1+1=· · ·=λdn =−kTn−1k.

T =Pn

k=1Πkk + (I−

n

X

k=1

Πk)T(I−

n

X

k=1

Πk)

| {z }

=Tn

Philippe Jaming (Universit ´e de Bordeaux) Spectral Theorem 2 Master Math & Applications 9 / 11

(43)

Induction

T =

n−1

X

k=1

Πkk+ I−

n−1

X

k=1

Πk

!

T I−

n−1

X

k=1

Πk

!

| {z }

=Tn−1

i Fn=EkTn−1k⊕E−kTn−1k

ii πn±⊥-projection onE±kTn−1kΠn+nnprojection onFn;

iii m2n−1= dimEkTnk,m2n= dimE−kTnk etdn=m2n−1+m2n≥1;

iv edn−1+1, . . . ,edn−1+m2n−1 ONB ofEkTnk &edn−1+m2n−1+1, . . . ,edn ONB ofE−kTnkedn−1+1, . . . ,edn ONB ofF1;

v λdn−1+1=· · ·=λdn−1+m2n−1 =kTn−1k, λdn−1+m2n−1+1=· · ·=λdn =−kTn−1k.

T =Pn

k=1Πkk + (I−

n

X

k=1

Πk)T(I−

n

X

k=1

Πk)

(44)

Conclusion

We want to show thatT =

+∞

X

k=1

Πkk that isTn→0.

We havekTnk<kTn−1k(λk decreases).

WriteDn=d1+· · ·+dnand note that|λDn|=· · ·=|λDn+1−1|=kTnk If not,(eDn)infinite orthogonal system soeDn *0 (weakly),thus TeDn →0 strongly (T compact).

Dn|=|λDn|keDnk=kTeDnk →0.

Philippe Jaming (Universit ´e de Bordeaux) Spectral Theorem 2 Master Math & Applications 10 / 11

(45)

Conclusion

We want to show thatT =

+∞

X

k=1

Πkk that isTn→0.

We havekTnk<kTn−1k(λk decreases).

WriteDn=d1+· · ·+dnand note that|λDn|=· · ·=|λDn+1−1|=kTnk If not,(eDn)infinite orthogonal system soeDn *0 (weakly),thus TeDn →0 strongly (T compact).

Dn|=|λDn|keDnk=kTeDnk →0.

(46)

Conclusion

We want to show thatT =

+∞

X

k=1

Πkk that isTn→0.

We havekTnk<kTn−1k(λk decreases).

WriteDn=d1+· · ·+dnand note that|λDn|=· · ·=|λDn+1−1|=kTnk If not,(eDn)infinite orthogonal system soeDn *0 (weakly),thus TeDn →0 strongly (T compact).

Dn|=|λDn|keDnk=kTeDnk →0.

Philippe Jaming (Universit ´e de Bordeaux) Spectral Theorem 2 Master Math & Applications 10 / 11

(47)

Conclusion

We want to show thatT =

+∞

X

k=1

Πkk that isTn→0.

We havekTnk<kTn−1k(λk decreases).

WriteDn=d1+· · ·+dnand note that|λDn|=· · ·=|λDn+1−1|=kTnk If not,(eDn)infinite orthogonal system soeDn *0 (weakly),thus TeDn →0 strongly (T compact).

Dn|=|λDn|keDnk=kTeDnk →0.

(48)

Conclusion

We want to show thatT =

+∞

X

k=1

Πkk that isTn→0.

We havekTnk<kTn−1k(λk decreases).

WriteDn=d1+· · ·+dnand note that|λDn|=· · ·=|λDn+1−1|=kTnk If not,(eDn)infinite orthogonal system soeDn *0 (weakly),thus TeDn →0 strongly (T compact).

Dn|=|λDn|keDnk=kTeDnk →0.

Philippe Jaming (Universit ´e de Bordeaux) Spectral Theorem 2 Master Math & Applications 10 / 11

Références

Documents relatifs

(iii) ´ Ecrire pour chacune de ces familles un champ de vecteurs lisse sur H dont les courbes int´ egrales sont les droites de la famille. (iv) Calculer le crochet des deux champs

Alors si l’on cherche la distance sur ce trajet entre le village et le randonneur, elle est de 0 quand le randonneur est au village, et de a quand il est au refuge... Les

M1 info UFR mathématiques et informatique Université Paris Descartes Examen « apprentissage automatique ».. 17 mai 2011 ­ Durée 2h

[r]

[r]

[r]

[r]

[r]