Научная статья на тему 'When are the nonstandard hulls of normed lattices discrete or continuous?'

When are the nonstandard hulls of normed lattices discrete or continuous? Текст научной статьи по специальности «Математика»

CC BY
50
7
i Надоели баннеры? Вы всегда можете отключить рекламу.
Область наук
Ключевые слова
НОРМИРОВАННАЯ РЕШЕТКА / ДИСКРЕТНЫЙ ЭЛЕМЕНТ / НЕСТАНДАРТНАЯ ОБОЛОЧКА. / NORMED LATTICE / DISCRETE ELEMENT / NONSTANDARD HULL

Аннотация научной статьи по математике, автор научной работы — Troitsky Vladimir Georgievich

This note is a nonstandard analysis version of the paper "When are ultrapowers of normed lattices discrete or continuous?" by W. Wnuk and B. Wiatrowski.

i Надоели баннеры? Вы всегда можете отключить рекламу.
iНе можете найти то, что вам нужно? Попробуйте сервис подбора литературы.
i Надоели баннеры? Вы всегда можете отключить рекламу.

Текст научной работы на тему «When are the nonstandard hulls of normed lattices discrete or continuous?»

Vladikavkaz Mathematical Journal 2009, Vol. 11, No 2, pp. 43-45

UDC 517.98

WHEN ARE THE NONSTANDARD HULLS OF NORMED LATTICES DISCRETE OR CONTINUOUS?

Dedicated to Safak Alpay on the occasion of his sixtieth birthday

V. G. Troitsky

This note is a nonstandard analysis version of the paper «When are ultrapowers of normed lattices discrete

or continuous?» by W. Wnuk and B. Wiatrowski.

Mathematics Subject Classification (2000): 46S20, 46B42.

Key words: normed lattice, discrete element, nonstandard hull.

In Functional Analysis, the ultrapower and the nonstandard analysis approaches are equivalent: results obtained by one of these two methods can usually be translated into the other. In this short note, we present nonstandard analysis versions of the main results of [5], where they were originally presented in the ultrapower language. We believe that in this new form the ideas of the proofs are more transparent.

Suppose that E is a Archimedean vector lattice. Recall that an element 0 < e G E is said to be discrete if 0 ^ x ^ e implies that x is a scalar multiple of e or, equivalently, the interval [0, e] doesn't contain two non-zero disjoint vectors (see [3, Theorem 26.4]). We say that E is continuous if it contains no discrete elements and discrete if every non-zero positive vector dominates a discrete element or, equivalently, E has a complete disjoint system consisting of discrete elements (see [1, p. 40]).

If E is a normed space. We will write *E for the nonstandard extension of E and E for the nonstandard hull of E. We refer the reader to [2, 6] for terminology and details on nonstandard hulls of normed spaces and normed lattices. We will use the following standard fact (see, e. g., [4, Remark 4]).

Lemma 1. Suppose that E is a normed lattice and a,x,b G *E such that a ^ b and a ^ X ^ b. Then there exists y G *E such that y ~ x and a ^ y ^ b.

The following is a variant of Theorem 2.2 of [5]:

Theorem 2. Let E be a normed lattice. Then the following are equivalent.

(i) E is continuous;

(ii) 3e > 0 Vx G E+ 3a, b G [0,x] a ± b and ||a|| A ||b|| ^ e||x||.

< (i) ^ (ii) Suppose that E fails (ii). Let e be a positive infinitesimal. Then there exists a vector x G *E+ such that for all a, b G *[0, x] with a ± b we have ||a|| A ||b|| < e||x||. Without loss of generality, ||x|| = 1. Let a, b G [0, x] and a ± b. By Lemma 1, we may assume that

© 2009 Troitsky V. G.

44

Troitsky V. G.

a, b £ *[0, x]. Furthermore, a ±6 implies that a A b w 0. Let u = a — a A b and v = b — a A b, then u, v £ *[0, x] and u ± v, so that ||u|| A ||v|| < e. It follows that either ||u|| or ||v|| is infinitesimal. Say, u ~ 0. Then a = u + a A b is infinitesimal as well, so that a = 0. Thus, x is discrete in E

(ii) ^ (i) Suppose that (ii) holds for some (standard) e > 0. Let x £ E+, show that x is not discrete. Without loss of generality, x £ *E+ and ||x|| = 1. By (ii), we can find a, b £ *[0, x] such that a ± b and ||a|| A ||b|| ^ e. It follows that neither a nor b is infinitesimal, so that a, b are two non-zero disjoint elements of [0, x]. Hence, x is not discrete. >

Recall that a normed lattice satisfies the Fatou property if 0 ^ xa | x implies ||xa|| ^ ||x||, and the a-Fatou property if 0 ^ xn | x implies ||xn|| ^ ||x||, see, e. g., [1]. We will use the following simple lemma.

Lemma 3. Suppose that E is a normed lattice with the Fatou property and S C E+ such that x = sup S exists. Then for every e > 0 there is a finite subset y of S such that || sup Y|| ^ (1 — e)||x||. The same is true for countable families if E satisfies the a-Fatou property.

< Let A be the collection of all finite subsets of S, ordered by inclusion. Clearly, supaeA sup a = x. Let xa = sup a, then (xa)aeA is an increasing net and 0 ^ xa | x. It follows from the Fatou property that | xa 11 ^ 11 x 11, so that there exists y £ A with ||x71| ^ (1 — e)||x||.

Now suppose that E satisfies a-Fatou property and x = V¿=1 x%. Let Zk = V(=1 xi, then xk ^ Z( ^ x, so that x = \/¿=1 Z(. Now a-Fatou property guarantees that ||z(|| ^ ||x||, so that (1 — e)||x|| ^ ||zm|| = ||x1 V ■ ■ ■ V xm|| for some m. >

The following is a variant of Theorem 3.1 of [5].

Theorem 4. Let E be a discrete normed lattice, and D the set of all discrete elements of norm one in E. If E satisfies the Fatou property (or the a-Fatou property if D is countable) then the discrete elements of E are exactly the positive scalar multiples of the elements of {e | e £ *D}.

< It suffices to show that given x £ *E with ||x|| = 1, then x is discrete in E if and only if x = e for some e £ *D. Suppose that x = e for some e £ *D. Take any a £ *E such that 0 ^ a ^ x. By Lemma 1, we may assume that 0 ^ a ^ e. It follows that a is a scalar multiple of e, hence a is a scalar multiple of x.

Conversely, suppose that x is discrete in E. Note that the set D is a complete disjoint system in E. By [1, Theorem 1.75], we have x = sup{Pex | e £ *D}. For every e £ *D, the vector Pex is a scalar multiple of e, and 0 ^ Pex ^ x, hence 0 ^ Pex ^ x. Therefore, if Pex is not infinitesimal for some e £ *D then x is a scalar multiple of Pex, hence of e.

Suppose now that Pex is infinitesimal for every e £ *D. It follows from x = sup{Pex | e £ *D} and Lemma 3 that there exist n £ *N and e1,...,en £ *D such that ||z|| ^ |, where z = ||Peix V ■ ■ ■ V Penx||. Choose k ^ n in *N so that ||Peix V ■ ■ ■ V Pek-i x|| < 4, while ||Peix V ■ ■ ■ V Pekx|| ^ 4. put u = Peix V ■ ■ ■ V Pekx = Peix +-----+ Pekx. Then

4 < ||u|| < 11Peix V ■ ■ ■ V Pek-ix|| + ||Pefcx|| < 4,

hence ||u|| ~ 4. Put v = z — u, then u ± v, 0 ^ u, v ^ z, and ||u||, ||v|| ^ 4. Therefore, u and e are non-zero and disjoint elements of [0, x]; a contradiction. >

Corollary 5. Suppose that E is an AM-space with a strong unit, and H is a discrete regular sublattice of E. Then H is discrete.

< Let D be a complete disjoint system of discrete elements of norm one in H. Suppose that x £ H+. We will show that x majorizes a discrete vector. Without loss of generality,

When are hulls discrete or continuous?

45

x £ *H+ with ||x|| = 1. Then x = sup{Pex | e £ } by [1, Theorem 1.75]. Since E is an AM-space, we can apply Lemma 3 with e ~ 0 and find n £ *N and ei,...,en £ such that ||Peix V ■ ■ ■ V Penx|| ^ (1 — e)||x|| ~ 1. Again, since E is an AM-space, we have ||Peix V ■ ■ ■ V Penx|| = ||Peix|| V ■ ■ ■ V ||Penx||, so that ||Pekx|| w 1 for some k ^ n. Then Pekx is non-zero. It is discrete by Theorem 4 because Pek x is a multiple of . Finally, notice that pekx ^ x. >

References

1. Aliprantis C. D., Burkinshaw O. Locally solid Riesz spaces with applications to economics (Math. Surveys and Monographs, Vol. 105). 2 ed.—Providence (R.I.): Amer. Math. Soc., 2003.—344 p.

2. Emel'yanov E. Yu. Infinitesimals in vector lattices // Nonstandard analysis and vector lattices (Math. Appl., Vol. 525).—Dordrecht: Kluwer, 2000.—P. 161-230.

3. Luxemburg W. A. J., Zaanen A. C. Riesz spaces.—Amsterdam-London: North-Holland Publishing Co, 1971.—Vol. 1.

4. Troitsky V. G. Measures on non-compactness of operators on Banach lattices // Positivity.—2004.— Vol. 8, № 2.—P. 165-178.

5. Wnuk W., Wiatrowski B. When are ultrapowers of normed lattices discrete or continuous? // Positivity IV—theory and applications—Dresden: Tech. Univ. Dresden, 2006.—P. 173-182.

6. Wolff M. P. H. An introduction to nonstandard functional analysis // Nonstandard analysis (Edinburgh, 1996).—Dordrecht: Kluwer Acad. Publ., 1997.—P. 121-151.

Received May 19, 2009. Troitsky Vladimir G.

Department of Mathematical and Statistical Sciences, University of Alberta, Associate Professor Edmonton, AB, T6G2G1. Canada E-mail: [email protected]

i Надоели баннеры? Вы всегда можете отключить рекламу.