• Aucun résultat trouvé

An isomorphic characterization of the Schmidt class

N/A
N/A
Protected

Academic year: 2022

Partager "An isomorphic characterization of the Schmidt class"

Copied!
6
0
0

Texte intégral

(1)

C OMPOSITIO M ATHEMATICA

D. R. L EWIS

An isomorphic characterization of the Schmidt class

Compositio Mathematica, tome 30, n

o

3 (1975), p. 293-297

<http://www.numdam.org/item?id=CM_1975__30_3_293_0>

© Foundation Compositio Mathematica, 1975, tous droits réservés.

L’accès aux archives de la revue « Compositio Mathematica » (http:

//http://www.compositio.nl/) implique l’accord avec les conditions géné- rales d’utilisation (http://www.numdam.org/conditions). Toute utilisation commerciale ou impression systématique est constitutive d’une infrac- tion pénale. Toute copie ou impression de ce fichier doit contenir la présente mention de copyright.

Article numérisé dans le cadre du programme Numérisation de documents anciens mathématiques

http://www.numdam.org/

(2)

293

AN ISOMORPHIC CHARACTERIZATION OF THE SCHMIDT CLASS

D. R. Lewis 1

Noordhoff International Publishing

Printed in the Netherlands

This note

gives

an

isomorphic

characterization of the Hilbert-Schmidt

norm within the class of

unitarily

invariant crossnorms on the space R of finite rank operators on

l2 . Specifically

if the

completion

of R under a

has an unconditional basis

(or

more

generally

local unconditional struc-

ture)

than a is

equivalent

to the Hilbert-Schmidt norm.

For convenience

only

real Banach spaces are considered. Notation and

terminology

is standard, with the

following possible exceptions.

For E finite dimensional the unconditional basis constant of E is denoted

by x(E),

and the parameter q is defined

by ri(E)

=

inf ~u~~v~~(F),

where the

infimum is taken over all spaces F and

pairs of operators

u : E ~ F, v : F - E

satisfying

vu =

identity.

The definition

of îl

is extended to infinite dimen- sional spaces

by setting ~(G)

=

infDsuPE~D~(E),

with the infimum taken

over all confinal collections of finite dimensional

subspaces

of G. G has

local unconditional structure if

~(G)

~.

R

(respectively, R(n))

is the space of finite rank operators

on l2 (resp., l2).

A norm a on R

(or R(n))

is called a

unitarily

invariant crossnorm

if,

for

each

u ~ R, ~u~ ~ 03B1(u) ~ i1(u)

and

a(guh)

=

a(u)

for all isometries g and h.

This coincides with Schatten’s definition

[7].

The

completion

of R under

a is written

R(a),

and

R(a, n)

is

R(n)

with norm a.

Finally,

03C0p and

ip

denote

the

p-absolutely summing

and

p-integral

norms,

respectively ([6]).

THEOREM :

If a is

a

unitarily

invariant crossnorm on

12

and

if R(a)

has local

unconditional structure, then a is

equivalent

to the Hilbert-Schmidt norm.

PROOF : Fix n and let P

(respectively, Q)

be the natural inclusion

of R(a, n) (resp., R(a, n)’)

into

R(n2’ n).

G denotes the group of isometries

of l2 and dg

is the normalized Haar measure on G. Let Q) E

R(n)

with

03C9 ~

0, and write

1 The author was supported in part by NSF GP-42666.

(3)

294

(03C9*03C9)1 2 =

where

(ei)

is some orthonormal basis

for l2

and

each

Ai

~ 0. The dual of

R(a, n)

is

naturally

identified with

R(a’, n),

where

a’ is the associate crossnorm of

a([7])

and the action is the trace of the com-

position.

Define a

probability

measure y on the closed unit ball

of R(a, n) by setting

Then for u E

R(a, n)

the first since Q) may be translated to

(03C9*03C9)1 2 by

an

isometry.

Now in the last

integral replace g by

gg£, where 8 =

(03B51, ···, 8n)

is an

n-tuple

of

signs

and

g03B5(ei)

= 03B5iei.

Averaging

over all such

n-tuples 8

and

applying

Khinchin’s

inequality yields

for any vector

e ~ ln2

with norm one. Let dm be the normalized

(n -1 )-di-

mensional rotational invariant measure on the unit

sphere of 1’2

and en the constant

satisfying

This

equality

shows

that for ~e~

= 1

(4)

and the last is

nothing

but cn

21C2(U). Combining inequalities

now

gives

so that

To

estimate 7rl(P)

let y be the

probability

measure on the closed unit ball of

R(a, n)’

defined

by

where e is some fixed vector of norm one. The

previous inequality

shows

for

all

MëR(n),

and

consequently 03C01(P) ~ cn . By

the

proof of Theorem

3.5

of [2]

and by

, so

This first

inequality

is also valid if a is

replaced by

its associate a’, and hence

by duality (using 03C0’2

=

712)

The constants in the last two

inequalities

are both dominated

by 03C02~(R(03B1)),

so that a and 03C02 are

03C04~(R(03B1))2 - equivalent

on R. This

completes

the

proof.

(5)

296

REMARK 1: An

absolutely summing

operator on a

subspace

of an

L1(03BC)-

space or a

quotient

of a

C(K)-space

must factor

through

an

L1-space [2].

Thus the same

proof

as above shows that a is

equivalent

to the Hilbert- Schmidt norm if

R(a)

is

isomorphic

to a

subspace

of an

L1-space

or to a

quotient

of a

C(K)-space.

In

addition, by applying

a basis selection theorem of

Kadec-Pelczynski [3]

and the result of

McCarthy [5]

that

R(cp) Í Lp,

it can be shown that ods

equivalent

to n2

if R(a)

c=

L p for some p

&#x3E; 2. It seems

a reasonable

conjecture

that a is

equivalent

to 1r2 whenever

R(a)

embeds

in a space with an unconditional basis.

REMARK 2: Given 03BB ~ 1 there is a

unitarily

invariant crossnorm a on R which is

equivalent

to 1r2 and such that the unconditional basis constant

of R(a)

is at least 03BB. To see this fix n and let

~n

be the

symmetric

gauge function

([7], Chapter V)

on the space

of finitely

non-zero scalar sequences defined

by

The crossnorm a on R induced

by 0,,

is

clearly n2-equivalent

to 03C02 and

by

the

inequalities

of the

proof ~(R(03B1)) ~

03C0-2n1 2.

Similarly

it is

possible

to

produce unitary crossnorms 03B2

for which

~(R(03B2, n))

increases very

slowly.

For

instance,

the crossnorm

03B2

induced

by

the

symmetric

gauge function

is, asymptotically, (log n) 2-equivalent

to 03C02 on

ln2.

REMÂRK3:

Suppose R(a, n)

has a basis with the property that each per- mutation of the basis vectors

naturally

defines an

isomorphism

of norm at

most 03BB.

By [4] q(R(a, n)) ~ 303BB,

so a is

9n4 À2-equivalent

to 2 on

ln2.

(6)

REFERENCES

[1] Y. GORDON: On p-absolutely summing constants of Banach spaces. Israeal J. Math. 7

(1969) 151-162.

[2] Y. GORDON and D. R. LEWIS: Absolutely summing operators and local unconditional structures. Acta Math. (to appear).

[3] M. I. KADEC and A. OELCZYNSKI: Bases, lacunary sequences and complemented subspaces in the space Lp. Studia Math. 21 (1962) 161-176.

[4] D. R. LEWIS: A relation between diagonal and unconditional basis constants (to appear

[5] Ch. A. MCCARTHY: Cp. Israel J. Math. 5 (1967) 249-271.

[6] A. PLETSCH: Theorie der operatorenideale. Jena, 1972.

[7] R. SCHATTEN: Norm ideals of completely continuous operators. Berlin-Göttingen- Heidelberg, 1960.

(Oblatum 11-XI-1974) Department of Mathematics

University of Florida

Gainesville, Florida 32601 U.S.A.

Références

Documents relatifs

Finally, based on the empirical results, we think that the norm of the confusion matrix is quite an useful tool for measuring the performance of a model in the multi-class

Trotter, Critically indecomposable partially ordered sets, graphs, tournaments and other binary relational structures, Discrete Math. 113

We prove that the Schwartz class is stable under the magnetic Schrödinger flow when the magnetic 2-form is non-degenerate and does not oscillate too much at infinity.. Motivation

We show that every positive Hilbert-Schmidt operator on ~, and more generally every positive self-adjoint unbounded operator on determines a standard generalized

It has been pointed out [1] that the Universal Connection Theorem could possibly provide connections between Yang-Mills theories and so-called a-models which concern themselves with

Then ^p is a Banach algebra such that the strong operator sequential convergence and the norm convergence coincide.. Note that under the strong operator topology, J ^ p is

The main purpose of this paper is to construct a toy class of Feller semigroups, hence, Feller processes, with state space R n × Z m using pseudo- dierential operators not

Universit´ e de Tunis El Manar, Facult´ e des Sciences de Tunis, D´ epartement de Math´ ematiquess, 2092, Tunis, Tunisie. E-mail address