209 (1978) 209-215 VO,. ON THE RANGE OF COMPLETELY BOUNDED MAPS RICHARD h LOEBL Department of Mathematics Wayne State University Detroit, Michigan 48202 (Received March 28, 1977) It is shown that if every bounded linear map from a ABSTRACT. to a yon Neumann algebra dimensional or and Mn ; @ Mn, is the algebra of Let 67, be is completely bounded, then either where n C*-algebras, For every positive integer n, the entry-wise map from 67 (R) M n of ng n complex matrices. C*-algebra 67 67 is finite- is a commutative von Neumann algebra complex matrlces. n and let : 67 be a bounded linear map. we define the map to (R) We say that IInll Mn n where is Mn to be n= (R) denotes the comple.tely bounde idn’ C*-algebra if < m [i]. It is not a priori evident that there are bounded maps which sup n fall to be completely bounded. It follows from the results of this paper that there are almost always such maps. R.I. LOBEL 210 BI[2 Let us denote by the set of all completely bounded maps from 7 ] B[7, and by the set of all bounded maps from d ] B[7 ] shall describe some of the structure of C*-algebra conmutatlve S c ] for all and integer d, We Further, in a B (R) Mn then for some n then either 2 B[2, ], ] BliP7, If (2) B[, 1 BI[, If . , 3], we made the following Conjectures: previous paper (I) below. to to ,.(R) Mn . is finite-dimenslonal or We shall give an affirmative answer to both these conjectures under the is a yon Neumann algebra hypothesis that We should remark that the converses to (1) and (2) hold; i.e., if 7 is flnite-dlmenslonal or @ n’ B[, ] Bl[, then ] (see below). Although our proof depends heavily on the hypothesis that the range is avon Neumann algebra, we feel that this is merely a shortcoming of our proof, and not a true reflection of the facts. We begin with what is, to the best of our knowledge, the only example in the literature of a bounded THEOREM I: Further, if C(X) : c C(X) d n M2n where is not completely bounded. such that d is a Then there C*-algebra, then has an extension The proof of the first assertion can be found in [4, Lemma 2.1 PROOF: . and Theorem 2.2], and to produce is not completely bounded. X be an infinite compact Hausdorff space. Let is a bounded map mapat the second assertion follows from the construction used We will sketch the highlights of the construction, both for the convenience of the reader and for later reference. Let such that: C(X). (I) Ai For every integer Ai*; (2) AiAj n, + AjAi there exist elements 25ijI 2n; and (3) AI,...,An E M 2 n Tr(Ai) 0. 211 COHPLETELY BOUNDED MAPS (n) (n) Pi Let be positive linear functionals on n (n) (f) (n) p Then (f)A %0 i__l n) supports and let n)ll that if n 3/4 for all , 67 extensions to IlPn)ll that the 2 n" 14 II(n)II pn) by Krein’s Theorem [5, p. 227], the 2 4 [[p I I i. i, with disjoint closed n) I 2 SO We now remark hat have norm-preserving positive and the extension assertion rests on our demonstrating are the keys to computing sp It is true that for a positive linear functional [I, Prop. 1.2.10]. " I < >II --n Let Thus, "n) n i=l llpi(n)ll-nI/4. sp llk(n)ll joint closed supports, which is possible since II,II-- ,Np 1/4 sup n SUPn n" +, and thus n COROLLARY 2 p-i(n) where all the functionals If / =r2. X But in fact, are chosen to have dis- is infinite. I > SnUP ][q2n II- n, But Then fails to be completely bounded. is an infinite dimensional Hilbert space, and is an infinite compact Hausdorff space, there is a bounded map the algebra of all bounded operators on /, such that : X , C(X) is not completely bounded. , , II*[I If is a linear map 0:7 by *(A) =(A*)*. Then ’’II11" We say that but Bm * Bm[7, ] is a we define the adjoint of is a linear map from self-adJoint if I + i2’ be written uniquely as PROPOSITION 3: is C*-algebras, of where -.*. I’ 2 self-adJoint are 7 to , Every map and can self-adJoint. linear subspace of BI[, need not be norm-closed. PROOF: It is elementary that for all k, I]kH I]1] and that the sum or scalar multiples of completely bounded maps are completely bounded. For the second assertion, let us re-examine the proof of Theorem i. N ,(N) (n) Let ], 212 . (N) so R.I. LOBEL sp II*(N)II k However (N) I/4 (n) so each is completely bounded. We remark that if one defines in III’III. We suspect that lli llJ Bm[., 8] k __l[ kll, B.[, he is always dense in BI[ is closed ], at least in the weak (i. e., pointwise) topology. We also should remark that if ll0o@ll 1[8ooll llll 8 is a and more generally, Bl[d ] unchanged under isomorphism; e.g., if Z Thus the classes k. integers B.[, ]. Mm We will denote by the algebra *-isomorphism of C*-algebras, then llkll II(O@)kll ll(@O)kll - % - Bm[, ] and and ), for all are essentially i’ where Bm[6Z then ] - is a separable ) M n. n 2 We will now do an analysis of von Neumann algebras, based on their type, infinite-dimenslonal Hilbert space. By Corollary 2, that will identify the characteristics we need. cation of Let be a von Neumann algebra of type (an isomorphic copy of) LEMMA 5: Let be a [6, p. 45]. J, and {Pn]n=l Let E Foreach n) for projections. is spatially isomorphic to yon Neumann algebra of type II I. -lj (R) M but Then be family of non-zero, orthogonal projections in 2n n) n) for where n, write Pn i--1 [V)] E En)E )E is equivalent to Let or III. of))n M2n" (an isomorphic copy PROOF: !, II M. By [6, Cor. 14], PROOF: i We follow the type classlfi- [6, pp. 24-25]. LEMMA 4: Then Mm E n) in the usual sense of equivalence be partial isometries (in -lJ -iJ E Pn,Pn [p. 46, Remark]) so that we can take V )* COMPLETELY BOUNDED MAPS 213 Theft n P-Pn-- n M2n" LEMMA 6: In of degree n of) Let e = %, n < w. I % where If iS a homogeneous yon Neumann algebra n sup w, D then (an isomorphic copy M n. 2 PROOF: Each is (isomorphically> of the form is a commutative von Neumann algebra sp n If Thus 2 " [6, p. 98], and thus there is a subsequence M ai e M Then % (R) Mn ni such that 1 @ 2 ni % where i, Mn-- Mn. and thus i. We are now ready to prove the main result of this paper. THEOREM 7. Let be a yon Neumann algebra, and suppose that for some infinite compact Hausdorff space X, C*-algebra a commutative )5 where type III, and integer We can write PROOF: i and n ] Bin[ C(X), ]. such that of type !m, 2 In [6, p. 25]. of type lira’ 3 5 degree , n we see that C (R) is of type [6, p. 42]. sup n N I so Then there is (R) M n I 2 )3 4 of type 4 III, of By using Theorem i, Corollary 2, , , . Mn Lemma 4, and Len.na 5, we see that the hypothesis forces Thus u as a (unique) direct sum is of type 5 BI[ C(X), where i 2 3 4 0. is homogeneous of By applying Theorem I, Lenxna 6, and the hypothesis, < ". But then for an appropriate commutative (C(X) @ ) ) (C(X) @ Mn) C*-algebra We can now give our answer to the first conjecture. COROLLARY 8: 7, BI[7, ] If Bm[7, ], is a yon Neumann.algebra, and for all then C*-algebras @ Mn. For the sake of completeness, we state a converse to Theorem 7. may be found in [3, Lenna 7]. The proof R.I. LOBEL 214 PROPOSITION 9: then for all Mn, C*-algebras 6, B[, ]. ] BI[, (R) If In order to present our answer to the second conjecture, we need the following result. PROPOSITION i0: for some infinite compact Hausdorff space C(X) a variation upon [2]. THEOREM ii: If BI[ C*-algebra. ] E], and is a von Neumann algebra, and for some Bm[, ], then either = (R) is C*-algebra C (R) M n. is finite-dimensional or The proof of Theorem 7 shows that either PROOF: Then X. This fact is established in the proof of [3, Thin. PROOF: 6, be an infinlte-dlmenslonal Let Mn or M n. is infinite-dimensional, we see Suppose the latter; then if 2 from Proposition i0 that C(X) for some infinite compact Hausdorff space n X. But then by Theorem i, there is a map bounded but not completely bounded. , : , which is M n 2 This contradicts the hypothesis, and n completes the proof. For the sake of completeness, we state a result, which along with Proposition 9, yields a converse to Theorem ii. The proof may be found in 3, Lemma 5]. , PROPOSITION 12: B i[, 1 is finite-dimensional, then for all If C*- algebras BtdT, ]. We remark that, by Theorem ii, there are almost always bounded maps between C*-algebras that fail to be completely bounded. We should mention that there is another interesting consequence of the methods of this paper. - THEOREM 13. Let be an infinite-dimensional is a bounded self-adjoint map + where / M n C*- algebra Then there such that cannot be written 2 are bounded positive linear maps from ). q0. n COMPLETELY BOUNDED MAPS PROOF: the map : 215 The actual assertion of [4, Lemna 2.1 and Theorem 2.2] is that C(X) M n n described in the proof of Theorem 1 does not admit 2 such a decomposition. . The extension statement in Theorem i and Proposition i0 allow us to pass to an arbitrary infinite-dimensional Arguments simila to those of Lemnas 4, 5, 6 and Theorem 7 in this paper were used in the dissertation of Sze-Kai Jack Tsui, University of Pennsylvania, 1975. He also obtained results relating to Theorem 13 of this paper. BIBLIOGRAPHY Subalgebras of C*-algebras, Acta Math. 123 (1969), 141-224. i. Arveson, W. B. 2. Hirschfeld, R. A. and Johnson, B. E. Spectral characterization of finitedimensional algebras, Inda. Math. 34 (1972), 19-23. 3. Loebl, R. I. Contractive linear maps on 22 (1975), 361-366. 4. C*-algebras, Michigan Math. J. A Hahn decomposition for linear maps, Pacific J. Math., 65 (1976), 119-133. 5. Rickart, C. E. 1960. General Theory of Banach Algebras, Van Nostrand, New York, 6. Topping, D. M. Lectures on yon Neumann Algebras, Van Nostrand Reinhold, London, 1971. KEY WORDS AND PHRASES. Comply bounded maps, C*-algebras, von Neumann lgbas, extension of maps. AMS(MOS) SUBJECT CLASSIFICATION (1970) CODES. 46L05, 46L10, 46J10.