30.01.2013 Views

THÈSE DE DOCTORAT ELLOUMI Mounir Espaces duaux de ...

THÈSE DE DOCTORAT ELLOUMI Mounir Espaces duaux de ...

THÈSE DE DOCTORAT ELLOUMI Mounir Espaces duaux de ...

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

44 Dual topology of the motion groups SO(n) ⋉ R n<br />

for all l = 1, ..., s. It is easy to see that if λk = µpl then P (±λk) = Q(±λk) =<br />

0. On the other hand for all λk �= µpl<br />

� �s<br />

P (±λk) = Q(±λk) −<br />

−<br />

� � �s<br />

= Q(±λk) 1 −<br />

−<br />

� � p1 p2<br />

i=1 i=q1+1<br />

l=1<br />

... �d i=qs+1 (λ2i − µ 2 pl )<br />

� � p1 p2<br />

i=1 i=q1+1 ...<br />

i�=pl i�=pl � � ps<br />

d<br />

i=qs−1+1 i=qs+1<br />

i�=pl (µ2i − µ2pl )<br />

p1−1 p2−1 � � d� �<br />

...<br />

j=1 j=q1+1 j=qs+1<br />

� � p1 p2<br />

i=1 i=q1+1 ... �d i=qs+1 (λ2i − µ 2 j)<br />

� � p1 p2<br />

i=1 i=q1+1 ...<br />

i�=j i�=j<br />

�d i=qs+1(µ<br />

i�=j<br />

2 i − µ2j )Qj(±λk)<br />

�<br />

� � p1 p2<br />

i=1 i=q1+1 ...<br />

i�=k i�=k<br />

l=1<br />

�d i=qs+1(λ<br />

i�=k<br />

2 i − µ 2 pl )<br />

� � p1 p2<br />

i=1 i=q1+1 ...<br />

i�=pl i�=pl � � ps<br />

d<br />

i=qs−1+1 i=qs+1<br />

i�=pl (µ2i − µ2pl )<br />

�<br />

� � p1 p2<br />

p1−1 p2−1 � � d� � i=1 i=q1+1 ...<br />

i�=k i�=k<br />

...<br />

j=1 j=q1+1 j=qs+1<br />

�d i=qs+1(λ<br />

i�=k<br />

2 i − µ 2 j)<br />

� � p1 p2<br />

i=1 i=q1+1 ...<br />

i�=j i�=j<br />

�d i=qs+1(µ<br />

i�=j<br />

2 i − µ2j )<br />

��<br />

� � p1 p2<br />

p1 p2 � � d� � i=1 i=q1+1 ...<br />

i�=k i�=k<br />

...<br />

j=1 j=q1+1 j=qs+1<br />

�d i=qs+1<br />

i�=k<br />

� � p1 p2<br />

i=1 i=q1+1 ...<br />

i�=j i�=j<br />

�d i=qs+1(µ<br />

i�=j<br />

2 i − µ2j )<br />

�<br />

= Q(±λk) 1 −<br />

= 0<br />

�<br />

˜Ql(±λk)<br />

(λ2 i − µ 2 j) ��<br />

by using the preceding Lemma. Thus, <strong>de</strong>t(Jµ + B ± iλkI) = 0 for all k =<br />

1, ..., d.<br />

Démonstration. (of the theorem 2.4.3) Let n = 2d + 1 be odd. Suppose that<br />

λ1 ≥ µ k 1 ≥ λ2 ≥ µ k 2 ≥ ... ≥ λd ≥ |µ k d | for k large enough. So there is at least<br />

one subsequence (µ kj )j∈I such that µ kj = µ for all j in I where µ <strong>de</strong>pends on<br />

I and λ1 ≥ µ1 ≥ λ2 ≥ µ2 ≥ ... ≥ λd ≥ |µd|. We have proved in the preceding<br />

Lemma that there exists a skew-symmetric matrix B <strong>de</strong>fined in (2.19) such<br />

that the spectrum of the matrix Jµ + B is given by zero and the complex<br />

numbers ±iλ1, ±iλ2, ..., ±iλd. On the other hand, there is an orthogonal<br />

matrix A such that A(Jµ + B)At = Jλ (c.f. for instance [BJLR], Proposition<br />

7.3 for a similar statement in the complex case). If A ∈ SO(2d + 1), then we<br />

can take Akj = A (if not, we take Akj = −A) and Bkj = B for all j in I.<br />

Conversely, it is clear that lim rk = lim �Akℓrk� = 0, and for all j = 1, 2, ..., d<br />

k→∞ k→∞<br />

one has lim<br />

k→∞ <strong>de</strong>t(J µ k + Bk ± iλjI) = lim<br />

k→∞ <strong>de</strong>t(Ak(J µ k + Bk)A t k ± iλjI) = 0.<br />

Then, by the preceding Lemma, λ1 ≥ µ k 1 ≥ λ2 ≥ µ k 2 ≥ ... ≥ λd ≥ |µ k d | for k<br />

large enough large.

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!