© Del documento, de los autores. Digitalización realizada por ULPGC. Biblioteca Universitaria, 2017
Rev.Acad.Canar.Cienc . . , VIII (Núm. 1), 107-112 (1996)
FACTORIZATION OF UNBOUNDED W~AKLY COMPACT OPERATORS
T. Alvarez.
Opto. Matematicas. Universidad de Oviedo. Calvo Sotelo sin 33007 Oviedo. Spain.
Abstract. In this paper the factorísation result of Davis, Figiel, Johnson and
Pelczynski [DFJP] from bounded operators is generalise for unbounded operators .
Resumen. En esta nota generalizamos el resultado de factorizacion de Davis,
Figiel, Johnson y Pelczynski [DFJP] para operadores acotados debilmente compactos
actuando entre espacios de Banach al caso general de operadores lineales debilmente
compactos no necesariamente continuos y actuando entre espacios normados arbitrarios.
Keywords: weakly compact operator, reflexiva space, factorization of unbounded
operator.
1. INTRODUCTION
Let X and Y be normed spaces and T is a linear operator with domain a linear
subspace D(T) of X and range R(T) contained in Y. The class of all such operators T will
be denotad by L(X,Y).
We shall denote by XT the vector space D(T) with the norm 11 xll T = llxll + llTxll and
the operator GT ( or simply G ) denotes the canonical injection from XT into D(T).
Continuous operators T with D(T) = X will be referred to as bounded; TG is thus a
bounded operator. The adjoint of T is the operator defined by T' : D(T') e Y' -+ D(T)'
where D(T') = {y' e Y' : y' T is continuous on D(T) } and T'y'(x) = y'(Tx) for xe D(T) and
y'e D(T') (11; 11.2.2].
We shall write T 0 = JyT where Jy is the natural injection of Y into its
completion v- • and J denotes the canonical isometry of a given normed space into its
seconddual.
lf T is continuous, then T- is the closure of T regarded as an element of L(X- ,Y-).
In particular (TG)-is a bounded operator in L((XT)-,v-).
lf M is a linear subspace of X we write T/M for the restriction of T to M; by the
usual convention T/M = T/Míl D(T).
The operator T is called closable if T has an extension T whose graph is G(T). lf T =
T then T is said to be closed. A closed operator T is called a 0+·operator (0.-operator,
107
© Del documento, de los autores. Digitalización realizada por ULPGC. Biblioteca Universitaria, 2017
Fredholm operator) if R(T) is closed and its null space is finite dimensional (if
R(T) is closed and of finite codimension,if Te 0+ íl 0-). We call Ta F +- operator
(partially continuous) if there exists a linear subspace M of X of finite codimension such
that (T/M)-1 exists and is continuous (T/M is continuous) [4],and T is an F- -operator
if T'e 0+) [7].
1 will be recalled (see,e.g [3), [8), [9]) sorne facts about bounded operators. Let T
be bounded and X and Y be Banach spaces. Then T is callad weakly compact·if TBx is
relatively cr(Y,Y')-compact ( Bx denotes the closed unit ball of X) and the following
properties are equivalent:
(1) T is weakly compact.
(2) T' is cr(Y' ,Y)-cr(X'.x'') continuous.
(3) T"X"CJY.
(4) T' is weakly compact.
(5) T factors through a reflexiva space.
In the general case of Te L(X,Y) we observe that the second adjoint of T presents as
an operator T" : D(T") e D(T)" ~ D(T')' and thus R(T")C IQY" where a denotes the
quotient map of Y" onto Y"/D(T')J.. and 1 is the canonical isometry mapping of QY" onto
D(T')'.
The operator Te L(X,Y) is said to be weakly compact if TBo(T) is relatively
cr(Y, D(T'))-compact (1 ). Sorne characterizations analogous to (1 )-(2) and (3) in the
general case of an arbitrary operator T are given in (1) where also it is investigated the
connection between weak compactness of T and that of T'.
In the present papar we will analyze the equivalence of properties (1) and (5) in
the general situation.
Throughout the remainder of the papar Tel(X,Y) where X and Y are normed spaces.
2.FACTORIZATION OF AN UNBOUNDED WEAKLY COMPACT OPERATOR
Glven an operator Te L(X, Y) we shall obtain a factorisation T 0 = jA through a
Banach space Zp (1sps oo ) where j is a tauberian injection and AGT is a bounded
operator. We factorise T 0 instead of T in order to retain the character of the [DFJP)
factorisation. The space Zp is constructed by setting W:= {T 0x : llxllT ~ 1, xe D(T) } and
108
© Del documento, de los autores. Digitalización realizada por ULPGC. Biblioteca Universitaria, 2017
then proceeding in the usual way from the gauges of the set Un= 2nw + 2-nsy-.
We have the bounded operators
,...,
and (TG) = jB is the usual factorisation. Hence T 0 = jA where A= BJX.y.G-1 so
that AG is bounded . Since T 0 GBXT is contained in (TGBxi" the gauges of the sets
obtained from W = T 0 GBXT coincide with those used in the [DFJP] construction, based on
"" (TG) B(XTl· Hence the following theorem:
2.1 Tbeorem.([DFJP] factorization of T 0 ). Let TE L(X, Y) be given . Then
corresponding to each 1~~00 there is a Banach space Zp anda factorisation
A: D(T)-+ Zp,
in which jp is a bounded Tauberian injection, AG is bounded, and (TG)-= jp(AG)-.
Moreover if p>1 then the second adjoint of jp is injective. The map jp coincides with the
Tauberian injection in the [DFJP] factorisation of (TG)- corresponding to p.
Proof. ( see [2))
2.2. Propos!t!on. Let T be dense/y defined and partially cbntinuous. Then T' is
weakly compact if ánd only if T0 lactors through a reflexiva space.
Proof. Suppose that T' is weakly compact . By 2.1 Theorem it is enough to show that
(TG)-factors through a reflexiva space and for this it suffices, by the bounded case, to
preve that ( (TG)"' )' = (TG)' is weakly compact. But T' = (G-1 )'(TG)' [11; lemma 2.4)
and since T is partialiy continuous, so G-1 and hence (G-1 )' is continuous [5; corol.5].
Moreover (G-1 )' = (G')-1 ( as D(T) is dense ) and consequently R(G') is closed [10;
11.3.1). Therefore G'T' = IR(G')(TG)' and the hypothesis on T' now gives G'T' is weakly
compact [1 ;prop. 2.3) and by [12; 4.6.9) we conclude that (TG)' is weakly compact as
required.
Conversely, assume that T 0 factors through a reflexiva space. Then since the maps
ip in the [DFJP) factorisation of T 0 and (TG)-coincide by 2.1 Theorem it follows that
109
© Del documento, de los autores. Digitalización realizada por ULPGC. Biblioteca Universitaria, 2017
(TG)' is weakly compact and an argument similar to the preceding implication
shows that T' = (G-1 )'(TG)' is weakly compact.
2.3. Theorem.Let T be dense/y defined and partial/y continuous. Consider the
following properties:
(i) T is weak/y compact.
(ii) T0 factors through a reflexive space.
Then (i) => (ii), and if Y is complete then the two properties (i) and (ii) are
equivalent.
Proof. (i) => (ii). lt follows readily from the above proposition and that if T is
weakly compact then so is T' (1; corol. 4.4].
(ii)=> (i). Suppose that Y is Banach. Then by (1; Th. 2.7] it is sufficient to show
that T' is a(D(T'),Y)-a(D(T)', D(T)") continuous ( as D(T)" = D(T") by the continuity
of T' ). But if T satisfies (ii) then by 2.1 Theorem and the bounded result (4) ~ (5) we
deduce that (TG)' is a(Y',Y)-a((Xr)'. (Xrt) continuous (*) and since (G-1 )' is
continuous we have (G-1 )' is a(D(G-1 )', (D(G-1 )')')- a(D(T)', D(T)") continuous
(**) [13; p. 234. ex.9].
Let (y'cx) be a net in D(T') which is a(D(T'),Y)-convergent to sorne point y'eY'.
Since Y is complete and T' is continuous is D(T') a(Y',Y)-closed [6; Th. 5.8] and so,
y' e D(T'). Since T' = (G-1 )'(TG)' (11; lemma 2.4] it follows from (*) that
a((Xr)'. (Xr)")-lim(TG)'y'cx = a((Xr)'. (Xr)")-lim i(TG)'y'a = (TG)'y' = i(TG)'y'
where i is the inclusion of D(G-1 )' in (Xr)' and since i' is surjective (10; IV.1.2] we
deduce that a(D(G-1 )', (D(G-1 )')')-lim(TG)'y'a = (TG)'y' and by (**) we conclude
that a (D(T)',D(T)")-lim (G-1 )(TG) 'y'cx = a (D(T)', D(T)")-limT'y'a = T'y' as
required.
The following examples show that the properties " T weakly compact " and " T
factors through a reflexiva space " are not equivalen! for an arbitrary operator T.
There is a weakly compact operator T acting between Banach spaces such that T has
not a factorization through a reflexive space.
Let X= Y = L1 (0,1] and define T by D(T) = {fe L1 [0,1] : f' exists almos!
everywhere and f' eL1[0,1]}. Then T is densely defined and D(T') ={O} [7; ex. 2.1].
110
© Del documento, de los autores. Digitalización realizada por ULPGC. Biblioteca Universitaria, 2017
Hence T is not partially continuos [5; Th.3]. Now, let M be dense subspace of D(T)
consisting of the absolutely continuous functions. Then T/M is a surjective closed
Fredholm operator with one-dimensional null space (1 O; Vl.3.1] and so by [7; Th. 4.3]
(T/M)GT/M is a F_-operator. Hence ((T/M)GT/M)' is a F+-operator with non
reflexiva domain and by (1; prop. 3.5] is not weakly compact.
Suppose that T factors through a reflexiva space, then by 2.1 Theorem and the
bounded case it follows that ((T/M)GT/Ml' is weakly compact; a contradiction.
There exists an operator T which factors through a ref/exive space but is not
weak/y compact.
Let T be an unbounded closable operator in L(X,Y) with Y reflexiva. Then D(T') is a
dense proper subspace of Y' [6; Remark 5.12) and hence T is not partially continuous
since adjoints of partially continuos are continuous [5; Cor. 5). Moreover it is clear
that T factors through a reflexiva space .However T is not weakly compact since if Y is
complete and T is closable then T is weakly compact if and only if T' is weakly compact
( and then T is continuous ) (1 ; cor. 4.3). Let X= L 1(O,1 ), Y= L2[0, 1) and define T and
M as the previous example . Then T/M is surjective closed Fredholm operator with
one-dimensional null space (11; Vl.3.1 ). In particular T/M is an unbounded closable
operator.
REFERENCES
(1) T. ALVAREZ, R.W. CROSS, A. l. GOUVEIA., Adjoints characterisations of unbounded
weakly compact, weakly completely contin,uous and unconditionally converging operators.
Studia Mathematica, 113 (4 995), 283-298.
(2) T. ALVAREZ, R. W. CROSS, M. GONZALEZ., Factorization of unbounded !hin and cothin
operators. Preprint.
(3) J. B. CONWAY., A course in Functional Analysis. Springer, Heidelberg, 1985.
(4) R. W. CROSS., Properties of sorne norm relatad functions of unbounded linear
operators. Math. Z., 199 (1988), 285-302.
[5]R. W. CROSS, L. E. LABUSCHAGNE., Partially continuous and semicontinuous linear
operators in normed spaces. Expositiones Math. (1989).
(6) R. W. CROSS., Adjoints of non densely defined linear operators. Universidad de
Cantabria. Homenaje al Prof. V.M. Onieva. (1991).
(7) R. W. CROSS, L.E. LABUSCHAGNE. Characterisations of operators of lower
semiFredholm type. Q. M. (1992), 151-173.
111
© Del documento, de los autores. Digitalización realizada por ULPGC. Biblioteca Universitaria, 2017
[8] W. J. DAVIS, T. FIGIEL, W. B. JOHNSON, A. PELCZVNSKI. Factoring weakly compact
operators. J. Functional analysis 17 (1974), 311-327.
[9] N. DUNFORD, J. T. SCHWARTZ. Linear Operators l. lnterscience Publishers.1958.
[10] S. GOLDBERG. Unbounded linear operators. McGraw-Hill ,1966.
[11] L. E. LABUSCHAGNE. On the minimun modulus of an arbitrary linear operator. Q .
M. 14,(1991), 77-91 .
[12] A. PIETSCH. Operators ideals. North-Holland, 1979.
[13] A. TAYLOR, D. LAY. lntroduction to Functional Analysis. John Wiley, 1980.
Recibido: 26 Diciembre 1996
112