PROCEEDINGS OF THE AMERICAN MATHEMATICAL SOCIETY Volume 129, Number 7, Pages 20052010 S 0002-9939(00)05781-6 Article electronically published on November

21, 2000

STAR RING HOMOMORPHISMS BETWEEN COMMUTATIVE BANACH ALGEBRAS

TAKESHI MIURA (Communicated by Dale Alspach) Abstract. We consider a -ring homomorphism from a commutative Banach algebra with an involution to a commutative Banach algebra with a symmetric involution. We give the Gelfand transform of the -ring homomorphism image.

1. Introduction Denition 1.1. Let A and B be commutative Banach algebras with a -involution and a -involution, respectively. We say that : A B is a -ring homomorphism if the following equalities hold for every f, g A: (f + g) = (f ) + (g), (f g) = (f ) (g), (f ) = (f ) . Semrl [4] proved the following theorem on a structure of a -ring homomorphism between two commutative C -algebras. Theorem ([4]). Let X and Y be compact Hausdor spaces, C(X) and C(Y ) the Banach algebras of all complex-valued continuous functions on X and Y , respectively. If : C(X) C(Y ) is a -ring homomorphism, then there exist clopen decomposition {Y1 , Y0 , Y1 } of Y and a continuous map : Y1 Y1 X such that the equality f ((y)), y Y1 , (f )(y) = 0, y Y0 , f ((y)), y Y1 , holds for every f C(X). Kaplansky [1] proved that every ring isomorphism between semisimple complex Banach algebras can be decomposed into a linear part, a conjugate-linear part, and a non-continuous part on a nite-dimensional ideal. We consider a -ring homomorphism from a commutative Banach algebra A with an involution to a non-radical commutative Banach algebra B with a symmetric involution. We prove that the (Jacobson) radical of A is mapped into the (Jacobson)
Received by the editors May 17, 1999 and, in revised form, October 22, 1999. 1991 Mathematics Subject Classication. Primary 46J10. Key words and phrases. Commutative Banach algebras, ring homomorphisms.
c 2000 American Mathematical Society

2005

2006

TAKESHI MIURA

radical of B by the -ring homomorphism. If A is non-radical, we show that there exists a decomposition {Y1 , Y0 , Y1 } of the maximal ideal space of B such that the Gelfand transform of the -ring homomorphism image of f is 0 on Y0 and a composition of a non-zero continuous ring homomorphism on C and the Gelfand transform of f on Y1 Y1 for every f A. 2. Main results Let A be a commutative Banach algebra. We say that A is a radical algebra if there is no non-zero complex homomorphism on A. Then we dene the radical of A, denoted by radA, to be A. Unless A is a radical algebra, then we dene radA to be the intersection of all the maximal regular ideals in A. For convenience, we say that A is non-radical if A is not a radical algebra. Let B be a commutative Banach algebra with a -involution. We say that a -involution is symmetric if the equality x =x holds for every x B, where denotes the Gelfand transform, and denotes the complex conjugate. Theorem 2.1. Let A be a commutative Banach algebra with a -involution, B a non-radical commutative Banach algebra with a symmetric -involution, and MA and MB the maximal ideal spaces of A and B, respectively. If : A B is a -ring homomorphism, then (radA) radB holds. Therefore (f ) = 0 (f A)

holds if A is a radical algebra. If A is non-radical, there exist a decomposition {Y1 , Y0 , Y1 } of MB and a continuous map : Y1 Y1 MA such that the equality f (()), Y1 , (f )() = 0, Y0 , f (()), Y1 , holds for every f A. Then Y1 and Y1 are open subsets in MB . In particular, if A has a unit element, then Y1 , Y0 and Y1 are clopen, and Y1 and Y1 are compact subsets in MB . Before we turn to the proof, we show the following lemma. Lemma 2.2. Let : A C be a -ring homomorphism on a commutative Banach algebra A with a -involution. Then = 0 or MA or MA holds. Therefore = 0 if A is a radical algebra. Proof. First we consider the case where A has a unit element e. If we dene e : C C to be e () = (e) ( C),

-RINGHOMOMORPHISM

2007

then e is a -ring homomorphism. In fact, it is easy to see that e is a ring homomorphism. Since e = e , the equality e () = (e) = ((e) ) = holds for every C. In particular, e (t) = e (t) (t R). That is, Im e = 0 on R, where Im e denotes the imaginary part of e . Now, Kestelman [2] proved that if a ring homomorphism : C C is unbounded, then (R) is dense in C. Therefore e must be bounded. If e = 0, then e () = ( C) or e () = ( C), since e (r) = r for every rational number r, and since e is bounded. (i) In case e = 0, we have (f ) = = = (e)(f ) e (1)(f ) 0 (e) = e ()

for every f A. Therefore, = 0 on A. (ii) In case e () = for every C, we have, for every f A and every C, (f ) = = = (e)(f ) e ()(f ) (f ).

Therefore, MA . (iii) In case e () = for every C, we see that (f ) = (f ) for every f A and every C, in a way similar to the above. Therefore, MA . We have proved the case that A is unital. In case that A does not have a unit element, put Ae = {(f, ) : f A, C} the commutative Banach algebra adjoining a unit element to A. Moreover we can extend the -involution to Ae : (f, ) = (f , ) (f, ) Ae . Suppose = 0. That is, there exists a g0 A such that (g0 ) = 0. Put : Ae C as follows: (g0 ) (f, ) Ae . (f, ) = (f ) + (g0 ) It is easy to see that is additive and an extension of . Also ((f, )(g, )) = (f, ) (g, ) holds for every (f, ), (g, ) Ae , since the equalities (g0 ) (g0 ) (f ) = = (g0 ) (g0 ) , (g0 ) (g0 ) (g0 ) (f ) (g0 )

2008

TAKESHI MIURA

hold for every f A, , C. Hence, is multiplicative. Moreover, it is easy to see that ((f, ) ) = (f, ) holds for every (f, ) Ae , since (g0 ) (g0 ) = (g0 ) (g0 ) ( C).

Therefore, : Ae C is a -ring homomorphism. Then satises only one of the following, by the result proved above: = 0 or MAe or MAe . Since is an extension of , and = 0, we see that MA or MA . In particular, = 0 if A is a radical algebra. This completes the proof. Proof of Theorem 2.1. For every MB , put : A C as follows: (f ) = (f )() (f A).

Then we see that is a -ring homomorphism. Therefore, for every MB , = 0 or MA or MA , by Lemma 2.2. Therefore, = 0 for every MB , if A is a radical algebra. By the denition of radA, (radA) = (A) radB holds, if A is a radical algebra. If A is non-radical, we dene Y1 , Y0 and Y1 as follows: Y1 = { MB : MA }, Y0 = { MB : = 0}, Y1 = { MB : MA }. Then it is easy to see that {Y1 , Y0 , Y1 } is a decomposition of MB , and that Y1 and Y1 are open subsets in MB . Finally, we dene : Y1 Y1 MA as follows: () = , ( Y1 ), , ( Y1 ). For every f A, the equality ( Y1 ), ( Y0 ), ( Y1 ),

Then we see that is continuous on Y1 Y1 . f (()), (f )() = 0, f (()), holds, by the denition of . Therefore,

(radA) radB holds, even if A is non-radical.

-RINGHOMOMORPHISM

2009

In particular, if A has the unit element e, for every MB , we have (ie)() = i or (ie)() = 0 or (ie)() = i, by the formula of (f )() above. Consequently, the equalities Y1 = { MB : (ie)() = i}, Y0 = { MB : (ie)() = 0}, Y1 = { MB : (ie)() = i} hold for the decomposition {Y1 , Y0 , Y1 } of MB . Since (ie)is continuous on MB , Y1 , Y0 and Y1 are all clopen subsets in MB . Hence, Y1 , Y0 and Y1 are compact if MB is compact. If MB is locally compact, Y1 Y1 = { MB : |(ie)()| = 1} is a compact subset in MB , since B C0 (MB ) the algebra of all complex-valued continuous functions on MB which vanish at innity. Therefore, Y1 and Y1 are compact. If is a -homomorphism between C -algebras A and B, then is norm decreasing [3, Theorem 1.5.7], where -homomorphism is a linear -ring homomorphism. We consider a -ring homomorphism from a commutative Banach algebra with an involution to a commutative Banach algebra with a symmetric involution. If the Gelfand transform on B is an isometry, the following result holds. b Corollary 2.3. In addition to the assumptions in Theorem 2.1, if b B = holds for every b B, then is norm decreasing. Proof. For every f A (f ) holds, where
A B

= (f )

and

denote the norms on A and B, respectively.

Corollary 2.4. In addition to the assumptions in Theorem 2.1, if is surjective, then is dened on MB into MA and injective. Proof. First, we show that A must be non-radical. Suppose not. Then B = radB holds, since is surjective. This is a contradiction. Hence, A is non-radical. Moreover, MB = Y1 Y1 holds. In fact, assume to the contrary that there exists a MB such that = 0. Since is surjective, = 0 on B. We arrived at a contradiction. Hence, is dened on MB into MA . Suppose 1 = 2 (1 , 2 MB ). We show that there exists a x B with x = x such that 1 (x) = 2 (x). In fact, by the hypothesis, there exists a y B such that 1 (y) = 2 (y). We can write y = u + iv for some u, v B with u = u and v = v . Then j (y) = j (u) + ij (v) (j = 1, 2).

Since -involution is symmetric, j (u) = j (u), j (v) = j (v) holds for j = 1, 2. Hence, 1 (u) = 2 (u) or 1 (v) = 2 (v) holds. Therefore, we proved that there exists a x B with x = x such that 1 (x) = 2 (x). Since is surjective, there exists an f A such that (f ) = x. Then (f )is real valued on MB . Therefore, x(j ) = f ((j )) (j = 1, 2) holds, by the Gelfand transform formula of (f ) in Theorem 2.1. Hence, (1 ) = (2 ). This completes the proof.

2010

TAKESHI MIURA

Example 2.1. In Theorem 2.1, Y1 and Y1 need not be closed subsets in MA , unless A is unital. In fact, put A = {f C([ 0, 1 ]) : f ( 1 ) = 0 = f ( 2 )}, 3 3 B = C([ 0, 1 ]), f (x) (x [0, 1 )) 3 (f )(x) = 0 (x [ 1 , 2 ] ) 3 3 f (x) (x ( 2 , 1] ) 3

(f A).

Then, A and B are commutative Banach algebras with respect to the pointwise operations and the supremum norm, and : A B is a -ring homomorphism, where involutions on A and B are both complex conjugates. Then the decomposition {Y1 , Y0 , Y1 } of MB = [0, 1] is as follows: 1 1 2 2 Y1 = [0, ), Y0 = [ , ], Y1 = ( , 1]. 3 3 3 3 Hence Y1 and Y1 are not closed subsets in [0, 1]. References
[1] I. Kaplansky, Ring isomorphisms of Banach algebras, Canad. J. Math. 6 (1954), 374-381. MR 16:49e [2] H. Kestelman, Automorphisms of the eld of complex numbers, Proc. London Math. Soc. (2) 53 (1951), 1-12. MR 12:812b [3] G. K. Pedersen, C -Algebras and their Automorphism Groups, Academic Press, London 1979. MR 81e:46037 [4] P. Semrl, Non linear perturbations of homomorphisms on C(X), Quart. J. Math. Oxford Ser. (2) 50 (1999), no. 197, 87-109. MR 2000a:46074 [5] S.-E. Takahasi and O. Hatori, A structure of ring homomorphisms on commutative Banach algebras, Proc. Amer. Math. Soc. 127 (1999) no. 8, 2283-2288. MR 99j:46058 Department of Mathematical Sciences, Graduate School of Science and Technology, Niigata University, 8050 Ikarashi 2-no-chou, Niigata 950-2181, Japan E-mail address: miura@scux.sc.niigata-u.ac.jp