20.07.2013 Views

Estimation optimale du gradient du semi-groupe de la chaleur sur le ...

Estimation optimale du gradient du semi-groupe de la chaleur sur le ...

Estimation optimale du gradient du semi-groupe de la chaleur sur le ...

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

540 A. Olofsson / Journal of Functional Analysis 236 (2006) 517–545<br />

L(z, ζ ) =<br />

1<br />

(1 − ¯ζz) n IDn,T − Dn,T (I − zT ) −n (I − ¯ζT ∗ ) −n Dn,T<br />

− 1<br />

1 − ¯ζz Wn,T (z)Wn,T (ζ ) ∗ , (z,ζ)∈ D × D,<br />

is positive <strong>de</strong>finite on D × D. In particu<strong>la</strong>r, we have the inequality<br />

1<br />

1 −|z| 2 Wn,T (z)Wn,T (z) ∗ + Dn,T (I − zT ) −n (I −¯zT ∗ ) −n Dn,T<br />

<br />

1<br />

(1 −|z| 2 ) n IDn,T in L(Dn,T ), z ∈ D.<br />

Proof. Let the space W be as in Lemma 4.1. By Theorem 4.1 and Remark 4.1 the space W<br />

is contractively embed<strong>de</strong>d into In,T . By this we mean that W ⊂ In,T and f 2 An f 2 W for<br />

f ∈ W. Recall that the space An(Dn,T ) is the orthogonal sum of the subspaces Vn(H) and In,T .<br />

The repro<strong>du</strong>cing kernel function for the space In,T is given by<br />

KIn,T (z, ζ ) = Kn(z, ζ ) − KVn(H)(z, ζ )<br />

=<br />

1<br />

(1 − ¯ζz) n IDn,T − Dn,T (I − zT ) −n (I − ¯ζT ∗ ) −n Dn,T , (z,ζ)∈ D × D,<br />

where the <strong>la</strong>st equality follows by Proposition 4.2 and (4.4). The repro<strong>du</strong>cing kernel function<br />

KW for the space W was computed in Lemma 4.1. It is known that a contractive embedding<br />

W ⊂ In,T is equiva<strong>le</strong>nt to the domination re<strong>la</strong>tion KW ≪ KIn,T of repro<strong>du</strong>cing kernel functions<br />

(see [7, Section I.7]). We conclu<strong>de</strong> that the function<br />

L(z, ζ ) = KIn,T (z, ζ ) − KW(z, ζ ), (z, ζ ) ∈ D × D,<br />

is positive <strong>de</strong>finite on D × D. This gives the positive <strong>de</strong>finiteness assertion in the theorem. The<br />

<strong>la</strong>st inequality in the theorem follows by setting z = ζ noticing that L(z, z) 0inL(Dn,T ) by<br />

positive <strong>de</strong>finiteness of the function L. This comp<strong>le</strong>tes the proof of the theorem. ✷<br />

In the case n = 2 of sca<strong>la</strong>r-valued Bergman inner functions the inequality in Theorem 4.2<br />

seems first to have appeared in Zhu [27, Theorem 4.2].<br />

Let us examine the case n = 1 somewhat closer.<br />

Corol<strong>la</strong>ry 4.2. Let T ∈ L(H) be a contraction in the c<strong>la</strong>ss C0·. Then<br />

1<br />

1<br />

IDT =<br />

1 − ¯ζz 1 − ¯ζz WT (z)WT (ζ ) ∗ + DT (I − zT ) −1 (I − ¯ζT ∗ ) −1 DT ,<br />

(z, ζ ) ∈ D × D.<br />

Proof. The space A1(DT ) is the orthogonal sum of the subspaces V1(H) and I1,T . By Corol<strong>la</strong>ry<br />

4.1 we know that the characteristic operator function WT is an isometric multiplier from<br />

A1(DT ∗) onto I1,T . The repro<strong>du</strong>cing kernel functions <strong>de</strong>compose accordingly. ✷

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

Saved successfully!

Ooh no, something went wrong!