This action might not be possible to undo. Are you sure you want to continue?
C -Algebras and Quantum Mechanics
Draft: 8 April 1998
Korteweg-de Vries Institute for Mathematics, University of Amsterdam, Plantage Muidergracht 24, 1018 TV AMSTERDAM, THE NETHERLANDS
email: email@example.com homepage: http://turing.wins.uva.nl/ npl/ telephone: 020-5256282 o ce: Euclides 218a
1.1 1.2 1.3 1.4 1.5
1 Historical notes
Origins in functional analysis and quantum mechanics Rings of operators (von Neumann algebras) . . . . . . Reduction of unitary group representations . . . . . . The classi cation of factors . . . . . . . . . . . . . . . C -algebras . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5 5 6 8 11
2 Elementary theory of C -algebras
2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 2.9 2.10 2.11 2.12 2.13 2.14
Basic de nitions . . . . . . . . . . . . . . . Banach algebra basics . . . . . . . . . . . . Commutative Banach algebras . . . . . . . Commutative C -algebras . . . . . . . . . . Spectrum and functional calculus . . . . . . Positivity in C -algebras . . . . . . . . . . . Ideals in C -algebras . . . . . . . . . . . . . States . . . . . . . . . . . . . . . . . . . . . Representations and the GNS-construction The Gel'fand-Neumark theorem . . . . . . . Complete positivity . . . . . . . . . . . . . . Pure states and irreducible representations . The C -algebra of compact operators . . . . The double commutant theorem . . . . . .
12 15 20 25 28 30 32 35 37 39 40 43 47 54 56 58 61 63 65 67 68 73 75 78 79 81 83 84 85 86 88
3 Hilbert C -modules and induced representations
3.1 3.2 3.3 3.4 3.5 3.6 3.7 3.8 3.9 3.10 3.11 3.12 Vector bundles . . . . . . . . . . . . . . . . . Hilbert C -modules . . . . . . . . . . . . . . . The C -algebra of a Hilbert C -module . . . Morita equivalence . . . . . . . . . . . . . . . Rie el induction . . . . . . . . . . . . . . . . The imprimitivity theorem . . . . . . . . . . Group C -algebras . . . . . . . . . . . . . . . C -dynamical systems and crossed products . Transformation group C -algebras . . . . . . The abstract transitive imprimitivity theorem Induced group representations . . . . . . . . . Mackey's transitive imprimitivity theorem . .
4 Applications to quantum mechanics
4.1 4.2 4.3 4.4 4.5
The mathematical structure of classical and quantum mechanics Quantization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Stinespring's theorem and coherent states . . . . . . . . . . . . . Covariant localization in con guration space . . . . . . . . . . . . Covariant quantization on phase space . . . . . . . . . . . . . . .
spectral theory. F. the abstract concept of a Hilbert space was still missing. a subalgebra which is closed under the involution A ! A ) that is closed (i. in which ( A ) is the expectation value of the observable A in the state .2 Rings of operators (von Neumann algebras) . this book remains the de nitive account of the mathematical structure of elementary quantum mechanics. Aided by his assistants Nordheim and von Neumann. and various examples of L2-spaces emerged around the same time. developed the spectral theory of bounded as well as unbounded normal operators on a Hilbert space. at the age of 23. saw his way through all structures and mathematical di culties. However. and in nite-dimensional quadratic forms (1904-) The discovery of quantum mechanics by Heisenberg (1925) in Gottingen and (independently) by Schrodinger in Zurich (1926) The arrival of John von Neumann in Gottingen (1926) to become Hilbert's assistant. It was von Neumann alone who. and the three wrote a joint paper on the subject (now obsolete). It was clear from either approach that the body of work mentioned in the previous paragraph was relevant to quantum mechanics. which at rst sight looked completely di erent. Schmidt de ned the space `2 in the modern sense. Heisenberg discovered a form of quantum mechanics. culminating in his book Mathematische Grundlagen der Quantenmechanik (1932).5 1 Historical notes 1.e. In 1908 his student E. Hilbert's memoirs on integral equations appeared between 1904 and 1906. he formulated the abstract concept of a Hilbert space. 1. The relationship and possible equivalence between these alternative formulations of quantum mechanics.) In one of his papers on Hilbert space theory (1929).e. was much discussed at the time. In a series of papers written between 1927-1932.1 Origins in functional analysis and quantum mechanics The emergence of the theory of operator algebras may be traced back to (at least) three developments. Initiating and largely completing the theory of self-adjoint operators on a Hilbert space.. (von Neumann's book was preceded by Dirac's The Principles of Quantum Mechanics (1930). Born was one of the few physicists of his time to be familiar with the concept of a matrix in previous research he had even used in nite matrices (Heisenberg's fundamental equations could only be satis ed by in nite-dimensional matrices). which he called `wave mechanics'. The latter may be de ned by its notion of convergence: a sequence fAn g of bounded operators weakly converges to A when ( An ) ! ( A ) for all 2 H. which contains a heuristic and mathematically unsatisfactory account of quantum mechanics in terms of linear spaces and operators. and proved the mathematical equivalence between matrix mechanics and wave mechanics. Heisenberg. Riesz studied the space of all continuous linear maps on `2 (1912). von Neumann de nes a ring of operators M (nowadays called a von Neumann algebra) as a -subalgebra of the algebra B(H) of all bounded operators on a Hilbert space H (i. which at the time was called `matrix mechanics'. Hilbert had been interested in the mathematical structure of physical theories for a long time his Sixth Problem (1900) called for the mathematical axiomatization of physics. This type of convergence is partly motivated by quantum mechanics. provided that A is self-adjoint and has unit norm. and introducing notions such as density matrices and quantum entropy. Hilbert thus ran a seminar on the mathematical structure of quantum mechanics. Born turned to his former teacher Hilbert for mathematical advice. Heisenberg's paper initiating matrix mechanics was followed by the `Dreimannerarbeit' of Born. Schrodinger was led to a di erent formulation of the theory. sequentially complete) in the weak operator topology. The work of Hilbert and his pupils in Gottingen on integral equations. and Jordan (1926) all three were in Gottingen at that time.
In one direction. In the present context. not of the form B(H). von Neumann proves what is still the basic theorem of the subject: a subalgebra M of B(H). one trivially veri es that the commutant of any set of bounded operators is weakly closed. a theorem of Wedderburn says that a von Neumann algebra on a nite-dimensional Hilbert space is (isomorphic to) a direct sum of matrix algebras. This alternative characterization indicates why von Neumann algebras are important in physics: the set of bounded operators on H which are invariant under a given group representation U (G) on H is automatically a von Neumann algebra.3 Reduction of unitary group representations The (possible) reduction of U (G) is determined by the von Neumann algebras U (G)00 and U (G)0 . he knew that all Hilbert spaces of a given dimension are isomorphic. that is. so that one cannot characterize a physical system by saying that `its Hilbert space of (pure) states is L2 (R3 )'. but for in nite-dimensional Hilbert spaces this state may no longer be de ned. This programme has. This theorem is remarkable. conversely.6 1 HISTORICAL NOTES For example. When G is compact. Among von Neumann's interest in quantum mechanics was the notion of entropy he wished to de ne states of minimal information. von Neumann's motivation came from group representations. This hope has not materialized. it does not. but are all isomorphic to the unique hyper nite factor of type III1 (see below). Von Neumann wondered if this. and. is weakly closed i M00 = M. B(H) is itself a von Neumann algebra. or a similar result in which direct sums are replaced by direct integrals (see below). so that U is discretely reducible. von Neumann hoped that the divergences in quantum eld theory might be removed by considering algebras of observables di erent from B(H). the natural topology on a von Neumann algebra is neither the weak nor the uniform one. so that. The commutant of a group U of unitary operators on a Hilbert space is a von Neumann algebra. For example. to some extent been realized in algebraic quantum eld theory (Haag and followers). although in algebraic quantum eld theory the basic algebras of local observables are. Density matrices may be regarded as states on the von Neumann algebraB(H) (in the sense of positive linear functionals which map I to 1). When H = C n for n < 1. The representation U is called primary when U (G)00 has a trivial center. Furthermore. a von Neumann algebra becomes a C algebra when one changes the topology from the weak to the uniform one. this implies that U is a multiple of a xed irreducible . in relating a topological condition to an algebraic one one is reminded of the much simpler fact that a linear subspace K of H is closed i K?? . containing the unit operator I. In the opposite direction.) Finally. where K? is the orthogonal complement of K in H. when U (G)00 \ U (G)0 = C I. in terminology to be introduced later on. the factors of type II1 (see below). given M. still holds when the dimension of H is in nite. viz. where U is the identity map. and the bicommutant M00 is simply (M0 )0 . Motivation from a di erent direction came from the structure theory of algebras. Instead. indeed. U is irreducible i U (G)0 = C I (Schur's lemma). As we shall see. However. (Note that a given group U of unitaries on H may be regarded as a representation U of U itself. Von Neumann's motivation in studying rings of operators was plurifold. a von Neumann algebra is automatically norm-closed as well. one takes U to be the set of all unitaries in M0 . (As we shall see. (Since the weak topology is weaker than the uniform (or norm) topology on B(H). von Neumann hoped to characterize quantum-mechanical systems by algebraic conditions on the observables. Von Neumann's bicommutant theorem implies a useful alternative characterization of von Neumann algebras from now on we add to the de nition of a von Neumann algebra the condition that M contains I. such a state is given by the density matrix = I=n.) In the same paper. whereas the commutant of a set of bounded operators which is closed under the involution is a -algebra. His primary motivation probably came from quantum mechanics unlike many physicists then and even now. there are von Neumann algebras on in nite-dimensional Hilbert spaces which do admit states of minimal information that generalize I=n.) 1. Here the commutant M0 of a collection M of bounded operators consists of all bounded operators which commute with all elements of M. every von Neumann algebra arises in this way.
1. When the direct integral reduces to a direct sum. of course. When H is separable. and take the sequence f ng to consist of a single strictly positive measurable function. one should decompose it as a direct integral. in which case we. but these functions do not lie in L2 (R). (To do so. ^ of all To completely reduce U . one may take = R with Lebesgue measure . We then write A= Z d ( )A : Thus a unitary group representation U (G) on H is diagonal when U (x) ( ) = U (x) for all x 2 G. of all sections which contains all n . such that each H carries a representation U . so that H ' H K.) This means that rstly one has a measure space ( ) and a family of Hilbert spaces fH g 2 . and V U (x)V is diagonal in the above sense. and secondly that for each xed the n span H . Instead of decomposing H as a direct sum. and this central decomposition of U is essentially unique. and U ' U IK. For 2 . but still assumed to be locally compact. one needs the U to be irreducible. An operator A on this direct integral Hilbert space is said to be diagonal when A ( )=A ( ) for some (suitably measurable) family of operators A on H .0 of the space . so that is the space G equivalence classes of irreducible unitary representations of G. such as the regular representation of G = R on H = L2 (R) that is. The irreducible would-be subspaces of H would be spanned by the vectors p (y) := exp(ipy). To de ne the direct integral of the H with respect to the measure . and for which all sections ! ( ) are measurable. For example. one puts U (x) (y) = (y . x). with A = U (x).0 it then makes sense to de ne ( The direct integral ) := Z d ( ) ( ( ) ( )) : d ( )H is discrete.3 Reduction of unitary group representations 7 representation U on a Hilbert space H . but written in the thirties (the ideas in it must have guided von Neumann from at least 1936 on). Z is then by de nition the subset of .0 of functions for which ( ) < 1. This leads to the direct integral decomposition Z L2 (R) = dp Hp R . When G is not compact. A section of this family is a function : ! fH g 2 for which ( ) 2 H . This occurs already in the simplest possible cases. one needs a sequence of sections f n g satisfying the two conditions that rstly the function ! ( n ( ) m ( )) be measurable for all n m. Complete reduction therefore calls for a further direct integral decomposition of primary representations this will be discussed below. write U= Z d ( )U : Reducing a given representation U on some Hilbert space then amounts to nding a unitary map V between H and some direct integral Hilbert space. The solution to this problem was given by von Neumann in a paper published in 1949. one needs to assume that H is separable. unitary representations may be reducible without containing any irreducible subrepresentation. There then exists a unique maximal linear subspace . one may always reduce a unitary representation in such a way that the U occurring in the decomposition are primary.
di cult. A locally compact group is said to be type I or tame when every primary representation is a multiple of a xed irreducible representation in other words. V (p) = Z R dy e. respectively). i. it is said to be a representation of type II. t 2 R. so that U has been diagonalized: the U (x) above are now the one-dimensional operators Up (x) = exp(ipx) on Hp = C . and annihilates pH? . a wild group has primary unitary representations which contain no irreducible subrepresentations. the group is called non-type I or wild. the operator V is unitary from pH to qH. Moreover. representations of the latter type may be decomposed in two alternative ways U= Z ^ G d 1( ) U = Z ^ G d 2( ) U ^ ). that is. all nite-dimensional Lie groups and all discrete groups). As far as the reduction of unitary representations is concerned. In case that U is not equivalent to Uh . More bizarrely. and profound papers (3 of which were in collaboration with Murray) in which the study of his `rings of operators' was initiated. in which p q i there exists a partial isometry V in M such that V V = p and V V = q. for in that case one may take any V with the above properties. An example of a wild group. Another example.ipy (y): This leads to V U (x)V (p) = exp(ipx) (p). Between 1936 and 1953 von Neumann wrote 5 lengthy. where the measures 1 and 2 are disjoint (that is. Hence when M = B(H) one has p q i pH and qH have the same dimension.4 The classi cation of factors . for example. and z w 2 C .. By de nition. We have therefore completely reduced U .8 1 HISTORICAL NOTES in which each Hp is C .E. these papers form `perhaps the most original major work in mathematics in this century'. it is of type III. discovered at a later stage. one simply performs a Fourier transform V : L2 (R) ! L2 (R). Murray and von Neumann de ne a partial orderering on the collection of all projections in M by declaring that p q when pq = p. M is generated by its projections. are unitarily equivalent. is the group of matrices of the form 0 eit 0 z 1 @ 0 ei t w A 0 0 1 where is an irrational real number. To reduce the regular representation of R on L2 (R). and U is not of type I. When U is neither of type I nor of type II.e. If not. (According to I. Segal. when pH qH. there exist two radically di erent classes of locally compact groups (the class of all locally compact groups includes. all (proper) subrepresentations of a primary type III representation are equivalent.) The analysis of Murray and von Neumann is based on the study of the projections in a von Neumann algebra M (a projection is an operator p for which p2 = p = p) indeed. curious phenomena may occur. When G is wild. An equivalent characterization of arises when we write M = U (G)0 for some unitary representation U of a group G (as we have seen. When M B(H). supported by disjoint subsets of G A reducible primary representation U may always be decomposed as U = Uh Uh . well known to von Neumann. This induces a partial orderering on the set of equivalence classes of projections by putting p] q] when the 1. is the free group on two generators. In that case U is equivalent to Uh indeed. a group is type I when all its primary representations are of type I. They noticed that one may de ne an equivalence relation on the set of all projections in M. A primary representation is said to be of type I when it may be decomposed as the direct sum of irreducible subrepresentations these subrepresentations are necessarily equivalent. this always applies) then p q i the subrepresentations pU and qU (on pH and qH.
Murray and von Neumann now proved that on any factor M (acting on a separable Hilbert space) there exists a function d from the set of all projections in M to R+ 1. and d takes the values f0 1 : : : ng. Murray and von Neumann de ne a projection to be nite when it is not equivalent to any of its (proper) sub-projections an in nite projection is then a projection which has proper sub-projections to which it is equivalent.. The study of von Neumann algebras acting on separable Hilbert spaces H reduces to the study of factors. all projections are nite. type II1: M has no minimal projections. Here a projection in B(H) is called nite when pH is nite-dimensional. The range of d is 0 1]. Such a factor is isomorphic to B(H) for separable in nite-dimensional H.e. distinguishing the equivalence classes p]. More generally.4 The classi cation of factors 9 equivalence classes p] and q] contain representatives p ~ and q ~ such that p ~ q ~. d(p) = d(q) i p] q] 3. d is unique up to nite rescaling. For this to be the possible. for von Neumann proved that every von Neumann algebra M B(H) may be uniquely decomposed. this is independent of the choice of p 2 p] and q 2 q]. satisfying the above properties. A projection p is called minimal or atomic when there exists no q < p (i. as in H = M = Z Z d ( )H d ( )M where (almost) each M is a factor. yet some are nite in the sense of Murray and von Neumann. A factor of type In is isomorphic to the algebra of n n matrices. type I1 : M has minimal projections. As we have seen. Moreover. A von Neumann algebra M is a factor when M \ M0 = C I when M = U (G)0 this means that M is a factor i the representation U is primary. .e. A rst classi cation of factors (on a separable Hilbert space) is now performed by considering the possible niteness of its projections and the range of d. Normalizing d such that d(I) = 1. One may say that. the dimension function d is a `renormalized' version of the usual one. One then has the following possibilities for a factor M. q p and q 6= p). for the factor M = B(H) the dimension d of a projection is a complete invariant. all nonzero projections are in nite-dimensional in the usual sense. but M has nite-dimensional projections in the sense of Murray and von Neumann. where n < 1: M has minimal projections. and I is in nite. The dimension is a function from the set of all projections in B(H) to R+ 1. type II1 : M has no minimal projections. type In. but in other factors all projections may be in nite in the usual sense. in order to distinguish in nite-dimensional but inequivalent projections. and I is nite. all projections are in nite-dimensional in the usual sense.. Murray and von Neumann showed that the set of equivalence classes of projections in M is totally ordered by whenever M is a factor. For M = U (G)0 the decomposition of H amounts to the central decomposition of U (G). For M = B(H) this actually de nes a total ordering on the equivalence classes. the range of d is the interval 0 1]. in which p] q] when pH has the same dimension as qH as we just saw. For M = B(H) this generalized notion of niteness coincides with the usual one. when pH and qH are orthogonal 4. d(p) < 1 i p is nite.1. and d takes the values f0 1 : : : 1g. d(p) > 0 when p 6= 0. satisfying 1. and d(0) = 0 2. d(p + q) = d(p) + d(q) when pq = 0 (i.
which nowadays is called modular theory or Tomita-Takesaki theory (apart from clarifying Tomita's work. The operation O ! O? on causally closed regions in space-time is somewhat analogous to the operation M ! M0 on von Neumann algebras. they were unable to provide a further classi cation of all factors. he believed these de ned some form of continuous geometry. Haag. In view of the range of d. a subject that really got o the ground with papers by Haag's pupil Araki in 1963 and by Haag and Kastler in 1964.g. One of the result has been that in realistic quantum eld theories the local algebras M(O) must all be isomorphic to the unique hyper nite factor of type III1 discussed below. This idea initiated algebraic quantum eld theory. led by Haag. Takesaki made essential contributions to this theory). as we have seen. realized that operator algebras could be a useful tool in quantum eld theory and in the quantum statistical mechanics of in nite systems. Here a region O in space-time is said to be causally closed when O?? = O. and Winnink characterized states of thermal equilibrium of in nite quantum systems by an algebraic condition that had previously been introduced in a heuristic setting by Kubo. For he showed that on a II1 factor M the dimension function d. In the sixties. Martin. and gave von Neumann the state of minimal information he had sought. where. Murray and von Neumann managed to do the former. Kaplansky.) A few years later (1967). Among other things. and until the sixties only a handful of mathematicians worked on operator algebras (e. Moreover. but only 5 years after he and Murray had recognized that the existence of type III factors was a logical possibility. The precise connection between von Neumann algebras and the decomposition of unitary group representations envisaged by von Neumann was worked out by Mackey.10 1 HISTORICAL NOTES With M = U (G)0 . Kadison. de ned on the projections in M. and Schwinger. From then till the present day. and others). which has helped both subjects. One sometimes says that a factor is nite when I is nite (so that d(I) < 1) hence type In and type II1 factors are nite. Factors of type I1 and II1 are then called semi nite. Dixmier. Sakai. the representation U is primary i M is a factor. an extraordinary coincidence took place. the existence of a II1 factor solved one of the problems that worried him in quantum mechanics. where O? consists of all points that are spacelike separated from O. It is hard to construct an example of a II1 factor. and type III factors are purely in nite. and they admitted having no tools to study type III factors. algebraic quantum eld theory has attracted a small but dedicated group of mathematical physicists. all nonzero projections are in nite-dimensional and equivalent in the usual sense as well as in the sense of Murray and von Neumann. Tomita developed a technique in the study of von Neumann algebras. On the physics side. and even harder to write down a type III factor. and von Neumann did the latter by himself. in 1957 Haag observed a formal similarity between the collection of all von Neumann algebras on a Hilbert space and the set of all causally closed subsets of Minkowksi space-time. U is of a given type i M is of the same type. This `trace' satis es tr(I) = d(I) = 1. In particular. On the mathematics side. However. Von Neumann was fascinated by II1 factors. which was to play an essential role in the classi cation of factors of type III. Hugenholtz. a group of physicists. (Hence von Neumann's belief that physics should use II1 factors has not been vindicated. such that M(O1 ) M(O2 ) when O1 O2 . This condition leads to type III factors equipped with a time-evolution which coincided with the one of the Tomita-Takesaki theory. may be extended to a positive linear functional tr on M. Mautner.. . Thus Haag proposed that a quantum eld theory should be de ned by a net of local observables this is a map O ! M(O) from the set of all causally closed regions in space-time to the set of all von Neumann algebras on some Hilbert space. This has led to an extremely fruitful intercation between physics and mathematics. the Tomita-Takesaki theory and the examples of type III factors type III: M has no minimal projections. At the time not many people were familiar with the di cult papers of Murray and von Neumann. and is therefore called the KMS condition. and M(O)0 = M(O? ). In the hands of Connes. Segal. and d assumes the values f0 1g. with the property that tr(UAU ) = tr(A) for all A 2 M and all unitaries U in M. this theory leads to a natural time-evolution on certain factors. and Adel'son-Vel'skii. Godement. Partly for this reason he believed that physics should be described by II1 factors.
try and nd a C -algebra which encodes this structure in some way. Other examples of mathematical structures that may be analyzed through an appropriate C algebra are group actions. Here the strategy is to take a given mathematical structure. There is a family of type III0 factors. This was proved by constructing a C -algebra C (G) of the group G. He proceeded to de ne an intrinsic spectral theory. groupoids. In 1941 he de ned the concept of a Banach algebra. for each pair x 6= y there exists a unitary representation U for which U (x) 6= U (y). the theory of C -algebras turned out to be interesting both for intrinsic reasons (structure and representation theory of C -algebras). and proved the basic theorem that each C -algebra is isomorphic to the norm-closed -algebra of operators on a Hilbert space. combining operator algebras with the theory of Banach spaces. and then obtain information about the structure through proving theorems about the C -algebra of the structure. who generalized von Neumann's idea of a state as a positive normalized linear functional from B(H) to arbitrary C -algebras. and others) eventually led to the classi cation of all hyper nite factors of type II and III (the complete classi cation of all factors of type I is already given by the list presented earlier). and proved most basic results in the theory of commutative Banach algebras. for an in nite-dimensional separable Hilbert space K. in which the theory of inclusions of II1 factors was related to knot theory. and even led to a new knot invariant. the sixties brought a fruitful interaction between C -algebras and quantum physics. as well as because of its connections with a number of other elds of mathematics. The rst instance where this led to a deep result which has not been proved in any other way is the theorem of Gel'fand and Raikov (1943). As we have mentioned already. For 6= 0 the factor of type III is unique. The unique hyper nite II1 factor was crucial in a spectacular development.5 C -algebras .) There is a unique hyper nite factor of type II1 . and many natural examples of factors constructed by purely mathematical techniques are hyper nite as well. labeled by 2 0 1]. the second one to be given to the once obscure eld of operator algebras. Their paper also contained the rudiments of what is now called the GNS construction. and complex domains. Powers. and others. which in turn is has been classi ed in terms of concepts from ergodic theory. Moreover. for which he was awarded the Fields Medal in 1982. such that M is the weak closure of n Mn . namely the tensor product of the hyper nite II1 -factor with B(K). In 1943 Gel'fand and Neumark de ned what is now called a C -algebra (some of their axioms were later shown to be super uous). The same idea lies at the 1. a great admirer of von Neumann. Woods. stating that the unitary representations of a locally compact group separate the points of the group (that is. (Experience shows that all factors playing a role in physics are hyper nite. Moreover. led to the following classi cation of hyper nite factors of type II and III (up to isomorphism): There is a unique hyper nite factor of type II1 . (In physics this factor occurs when one considers KMS-states at in nite temperature. showing that representations of C (G) bijectively correspond to unitary representations of G.) The work of Connes. In its present form. in which multiplication is (separately) continuous in the norm-topology. These are factors containing a sequence of nite-dimensional subalgebras M1 M2 : : : M. In the midst of the Murray-von Neumann series of papers.1.5 C -algebras 11 provided by physicists (Araki. foliations. Gel'fand initiated a separate development. the unique hyper nite III1 factor plays a central role in algebraic quantum eld theory. Segal returned to von Neumann's motivation of relating operator algebras to quantum mechanics. and nally showing that the states of an arbitrary C -algebra A separate the elements of A. this construction is due to Segal (1947). In 1990 Jones was awarded a Fields medal for this work. connecting states to representations. As with von Neumann algebras. There is a family of type III factors.
2 A pre-inner product on a vector space V is a map ( ) : V V ! C such that 1. This strategy has been successful in K -theory. k v k= 0 i v = 0 3. which we will study in detail in these lectures. An equation of the type a := b means that a is by de nition equal to b. The equivalence between the two de nitions of a pre-inner product is elementary in fact. k v + w k k v k + k w k (triangle inequality). should be translated into tools pertinent to the C -algebra C (X ).12 2 ELEMENTARY THEORY OF C -ALGEBRAS basis of non-commutative geometry and non-commutative topology. The two main examples of Banach spaces we will encounter are Hilbert spaces and certain collections of operators on Hilbert spaces. (v v) 0 for all v 2 V . ( 1 v1 + 2 v2 1 w1 + 2 w2 ) = 1 1 (v1 w1 ) + 1 2 (v1 w2 ) + 2 1 (v2 w1 ) + 2 2 (v2 w2 ) for all 1 2 1 2 2 C and v1 v2 w1 w2 2 V 2. Finally. K -theory. De nition 2.1) . 2 Elementary theory of C -algebras All vector spaces will be de ned over C . and in (de Rham) cohomology theory. The abbreviation `i ' means `if and only if'.1.1 Basic de nitions Cauchy-Schwarz inequality j(v w)j2 (v v)(w w) (2. to derive the rst axiom of the second characterization from the rst set of conditions. one derives the 2. A norm on V de nes a metric d on V by d(v w) :=k v . De nition 2. Here the starting point is another theorem of Gel'fand. cohomology. (v 1 w1 + 2 w2 ) = 1 (v w1 ) + 2 (v w2 ) for all 1 2 2 C and v w1 w2 2 V 3. A pre-inner product for which (v v) = 0 i v = 0 is called an inner product. k v k= j j k v k for all 2 C and v 2 V 4. An equivalent set of conditions is 1. w k. when appropriate. The strategy is now that the basic tools in the topology of X . in its di erential geometry. stating that any commutative C -algebra (with unit) is isomorphic to C (X ).. whose non-commutative version is called cyclic cohomology. where X is a compact Hausdor space. (v v) 0 for all v 2 V . which is the same as the symbol . (v w) = (w v) for all v w 2 V 2. and all functions will be C -valued. it is enough to assume that (v v) 2 R for all v (use this reality with v ! v + iw). homology. unless we explicitly state otherwise. and.1. We will denote a generic Banach space by the symbol B. Either way. and index theory haven been uni ed and made non-commutative in the KK -theory of Kasparov. A vector space with a norm which is complete in the associated metric (in the sense that every Cauchy sequence converges) is called a Banach space. and that subsequently these tools should be generalized to non-commutative C -algebras.1 A norm on a vector space V is a map k k : V ! R such that 1. The basic tool in KK theory is the concept of a Hilbert C -module. whose non-commutative version is even simpler than its usual incarnation. k v k 0 for all v 2 V 2.
the sequence fAn vg is a Cauchy sequence in B. In other words.5) for n m > N ( ). Note that this inequality is valid even when ( ) is not an inner product. Since B is complete by assumption.5). The smallest such C is the norm k k:= sup fj (v)j v 2 B k v k= 1g: (2. Since A = (A .) When B is a Hilbert space H the expression (2. we obtain k Av . This map is obviously linear. and B(B) is a linear space. Let fAn g be a Cauchy sequence in B(B). as it follows almost immediately from its de nition (and from the properties of the the norm on B) that the operator norm is indeed a norm.3). Since we have just seen that A 2 B(B). one therefore studies operators on a Hilbert space. when for A 6= 0 there is a C > 0 such that k Av k C k v k for all v. Am v k k An . (It is easily shown that a linear map on a Banach space is continuous i it is a bounded operator. the sequence fAn vg converges to some w 2 B. To obtain a satisfactory mathematical theory. Am ) + Am . it is wise to restrict oneself to bounded operators.e. Am is bounded.4 A bounded operator on a Banach space B is a linear map A : B ! B for which k A k:= sup fk Av k j v 2 B k v k= 1g < 1: (2. It now follows from (2. Am k < when n m > N ( ). we infer that A is bounded. Moreover. For arbitrary v 2 B.2) becomes 1 k A k:= sup f(A A ) 2 j 2H ( ) = 1g: (2. A Hilbert space is completely characterized by its dimension (i. or simply the norm. De nition 2. but we will never use this result.1.2) that A ..3) When A is bounded. De nition 2.1. because k An v . Am k k v k k v k (2. This terminology is justi ed. and all v 2 B.4) for all v 2 B. rather than the Hilbert space itself. of A.2) The number k A k is the operator norm. We will usually denote Hilbert spaces by the symbol H.1. Am k for all m > N ( ).3 A Hilbert space is a vector space with inner product which is complete in the associated norm. It follows from these properties that an inner product on V de nes a norm on V by k v k:= p(v v) the triangle inequality is automatic.2) imply that k A . We de ne a functional on a Banach space B as a linear map : B ! C which is continuous in that (v)j C k v k for some C .1 Basic de nitions 13 for all v w 2 V . Proposition 2. To obtain an interesting theory. for any > 0 there is a natural number N ( ) such that k An . Now de ne a map A on B by Av := w = limn An v.2.6) for all m > N ( ) and all v 2 B.7) k Av k k A k k v k . it follows that (2. In view of the comments following (2. Am v k k v k (2. it only remains to be shown that B(B) is complete in the operator norm. then A is bounded. but merely a pre-inner product. this proves that B(B) is complete. with operator norm k A k equal to the smallest possible C for which the above inequality holds. We recall this concept in the more general context of arbitrary Banach spaces. Taking n ! 1 in (2. (2. in arguments involving continuous operators on a Banach space one almost always uses boundedness rather than continuity. by the cardinality of an arbitrary orthogonal basis).5 The space B(B) of all bounded operators on a Banach space B is itself a Banach space in the operator norm. so that fAn g converges to A.6) and (2. Conversely. Indeed.
The operator adjoint A ! A on a Hilbert space. For v 6= 0 we may de ne a functional 0 on C v by 0 ( v) = .13) 2 + i 2i every element of A is a linear combination of two self-adjoint elements. Substituting this in (2.2) we obtain k AB k k A k k B k. As in this case.4) twice.A A = A0 + iA00 := A + (2.1.1. we derive the crucial property k A A k=k A k2.14 2 ELEMENTARY THEORY OF C -ALGEBRAS k ABv k k A k k Bv k k A k k B k k v k : Hence from (2. In other words. without proof. and use the CauchySchwarz inequality and (2.8) The dual B of B is the space of all functionals on B.4) to estimate k A k2 = (A A ) = ( A A ) k k k A A k k A A k k k2 : Using (2. For later use.8). Moreover. Theorem 2. we pick 2 H. This algebra has additional structure.8 A Banach algebra is a Banach space A which is at the same time an algebra. we infer that k A k2 k A A k k A k k A k : (2.3) and (2. Hence B(H) is a -algebra.5. To see how the norm in B(H) is related to the involution.1.10) (AB ) = B A ( A) = A : (2. .7 When (v) = 0 for all 2 B then v = 0.11) A -algebra is an algebra with an involution. Similarly to the proof of 2. This motivates the following de nition.1. de nes an involution on B(H).14) This leads to k A k k A k.14). Replacing A by A and using (2. for any Banach space B the space B(B) of all bounded operators on B is a Banach algebra. one shows that B is a Banach space. In what follows. de ned by the property ( A ) := (A ). and extend it to a functional on B with norm 1. we will restrict ourselves to the case that B is a Hilbert space H this leads to the Banach algebra B(H).9) yields k A k k A k.9) (2. It is clear that B(B) is an algebra under operator multiplication.6 For a functional 0 on a linear subspace B0 of a Banach space B there exists a functional on B such that = 0 on B0 and k k=k 0 k.12) Since one may write A A. so that k A k=k A k. As we have just seen. using (2. De nition 2.9 An involution on an algebra A is a real-linear map A ! A such that for all A B 2 A and 2 C one has A = A (2. each functional de ned on a linear subspace of B has an extension to B with the same norm. in which for all A B 2 A one has k AB k k A k k B k : (2.1. the fundamental Hahn-Banach theorem. an element A of a C algebra A is called self-adjoint when A = A we sometimes denote the collection of all self-adjoint elements by AR := fA 2 Aj A = Ag: (2. Recall that an algebra is a vector space with an associative bilinear operation (`multiplication') : A A ! A we usually write AB for A B . we quote. Corollary 2. De nition 2. for each v 2 B one has It follows that multiplication in a Banach algebra is separately continuous in each variable.
20) take A = I . De nition 2. We often write z for z I.15) k A A k = k A k2 : (2. it is enlightening to see concepts such as the spectrum in their general and appropriate setting. although the axioms make no reference to Hilbert spaces at all. Lemma 2.11 A Banach -algebra in which k A k2 k A A k is a C -algebra. which we will formulate precisely and prove in due course. however. but because of its role in the theory of C -algebras. We have just shown that B(H) is a C -algebra.1.16).1. as follows.2. Here a Banach -algebra is. Moreover. a Banach algebra with involution.16) In other words. Hence the axioms in 2.8. and k I k= 1: (2. where z 2 C . Form the vector space AI := A C (2. Note that in a C -algebra the property IA = AI = A already implies. Two C -algebras are isomorphic when there exists an isomorphism between them. It is remarkable.2 Banach algebra basics 15 De nition 2. that an injective morphism (and hence an isomorphism) between C -algebras is automatically isometric. a C -algebra is a Banach -algebra in which (2. so that (2. of course.10 A C -algebra is a complex Banach space A which is at the same time a algebra.2 Banach algebra basics (2.12 A morphism between C -algebras A B is a (complex-) linear map ' : A ! B such that '(AB ) = '(A)'(B ) '(A ) = '(A) (2. so that I I = I taking the adjoint.16) and (2. (2.15). for some Hilbert space H. such that for all A B 2 A one has k AB k kAk kBk (2. Combining (2.1.22) .20) follows from (2. An isomorphism is a bijective morphism.1. For later use we state some self-evident de nitions. each (operator) norm-closed algebra in B(H) is a C -algebra by the same argument.1.16) holds. When a Banach algebra A does not contain a unit. For this reason the condition that an isomorphism be isometric is not included in the de nition.21) and make this into an algebra by means of (A + I)(B + I) := AB + B + A + I 2. we can always add one.20) A Banach algebra with unit is called unital. we infer that for all elements A of a C -algebra one has the equality k A k=k A k : (2. this implies I = I.19) for all A B 2 A. Even in that special context.17) The same argument proves the following. A unit in a Banach algebra A is an element I satisfying IA = AI = A for all A 2 A. One immediately checks that the inverse of a bijective morphism is a morphism. A much deeper result. one derives k A k k A k as in the preceding paragraph. Recall De nition 2. states the converse of this: each C -algebra is isomorphic to a norm-closed -algebra in B(H). The material in this section is not included for its own interest.10 characterize norm-closed -algebras on Hilbert spaces.18) (2.
A) k=k An+1 k k A kn+1 . A) . m X k=0 Ak k=k n X k=m+1 Ak k n X k=m+1 k Ak k n X k=m+1 k A kk : For n m ! goes to 0 by the theory of the geometric series.16 2 ELEMENTARY THEORY OF C -ALGEBRAS where we have written A + I for (A ). z I) always exists when jz j > k A k. since it follows from (2. for n > m one has k (A + I)(B + I) k k A k k B k +j j k B k +j j k A k +j j j j =k A + I k k B + I k so that AI is a Banach algebra with unit. 1 Hence (A . The resolvent (A) of A 2 A is the set of all z 2 C for which A .1. De nition 2.23) that the norm of A + 0I in AI. The spectrum (A) of A 2 A is the complement of (A) in C in other words. Ak (I . De nition 2. etc.4. contained in the set fz 2 C j jz j k A kg 2. Furthermore. k . z I has a (two-sided) inverse in A.2 Let A be a unital Banach algebra. z I has no (two-sided) inverse in A. As we shall see at the end of section 2. Now compute n X k=0 Ak (I . the number 1 in C is identi ed with I.2 reproduces the usual notion of the spectrum of an operator on a Hilbert space.22) that A never has an inverse in AI. We rst show that the sum is a Cauchy sequence. Theorem 2. (A) is the set of all z 2 C for which A . the Cauchy P1 this k sequence n k=0 A converges for n ! 1. Indeed. A) = n X k=0 n X k=0 (Ak .1. compact 3. When A has no unit.4 When k A k < 1 the sum Pn k=0 A converges to (I . Proposition 2. The proof uses two lemmas.23) In particular.3 The spectrum (A) of any element A of a Banach algebra is 1. When A is the algebra of n n matrices. Since by (2. such that AI=A ' C . k I k= 1. the spectrum of A is just the set of eigenvalues. Using (2. not empty.1 Lemma 2. For A = B(H).15) in A. Ak+1 ) = I . de ne a norm on AI by k A + I k:=k A k +j j: (2. the resolvent and the spectrum are de ned through the embedding of A in AI = A C .2. the unitization AI with the given properties is not unique.2. An+1 : Hence k I. When A has no unit. we have shown the following.2.23) the norm of A 2 A in A coincides with k n X k=0 Ak . In other words. We assume that A is unital. as well as 2. the spectrum (A) of A 2 A always contains zero. .1 For every Banach algebra without unit there exists a unital Banach algebra AI and an isometric (hence injective) morphism A ! AI.2. one sees from (2.3.22) and (2.2. Since A is complete.
we see that z 2 (A). To resume the proof of Theorem 2. A k=0 z0 . From the proof of Lemma 2.4. A 1 n z .4.1 k. z ).4. f (z ) k= .2 Banach algebra basics which ! 0 for n ! 1. z0 .z = z.2. But f .1(G(A)) is open in A.1. Since k f (z + ) .27) .z X 0 k (z0 . A). Finally. its complement (A) is closed. which exists by (2. A). z )k (z0 .1 ! 1 z 1 = g(z ): 1 X 0 0.1 . take a B 2 A for which k B k < k A. for k A. For xed z0 2 (A). choose z 2 C such that jz . A) n X Ak = I: so that. z0 .1 k.1B ) has an inverse.2. one nally has nlim !1 k=0 Ak = (I . A)k. it remains to be shown that it is closed.26) and Lemma 2. the limit n ! 1 of this power series is 1 z . z0 and C ! A . Lemma 2. (A . z ) k= jz . so that f .2. A has an inverse. z .1 k k B k < 1: (2.1 k < 1: By Lemma 2. by continuity of multiplication in a Banach algebra.25) k A. de ne g : (A) ! A by g(z ) := (z .1 (G(A)) is the set of all z 2 C where z . A.1B ). A = z0 . Thus nlim !1 n X k=0 17 Ak (I .1 k.1 (2. nlim !1(I .1 . given A 2 A we now de ne a function f : C ! A by f (z ) := z . Given A 2 G(A). A=z ). This set being open.1 .24) The second claim of the lemma follows because (A . which exists because k A=z k < 1 when jz j > k A k. it follows from the topological de nition of a continuous function that f .1 A.8) this implies (2. By (2.2. namely (I + A. z0 j. because z0 .5 The set G(A) := fA 2 Aj A. A k=0 z0 .1 = .1 k= jz0 .1 (G(A)) = (A). To prove that (A) is compact.2.z . Moreover.1 B k k A. z0 j < k (A .2.5. z )(z0 . A 1 .1. A) = I: n X k=0 By a similar argument. as k A .1(I .26) Hence A + B = A(I + A. z0 .2. A).3.2. C k < lie in G(A). the power series converges for n ! 1 by Lemma 2. with A ! A . A).1 existsg of invertible elements in A is open in A. A A Hence g(z ) = 1 X k=0 (z0 . as k A k < 1 by assumption. we see that f is continuous (take = in the de nition of continuity).z .1 . Because G(A) is open in A by Lemma 2.1 : (2.5. z0 ). It follows that all C 2 A for which k A . z j k (z0 .
is then de ned on where it is analytic and vanishes at in nity.2. The fact that the spectrum is never empty leads to the following Gel'fand-Mazur theorem.A=z ). By (2.8 For a polynomial p on C .A=z = I.27) implies that the function g : z ! (g(z )) is given by a convergent power series. which will be essential in the characterization of commutative C -algebras. Since (A) 6= .2.28) z!1 lim g (z ) = 0: (2. Since k A k=k z I k= jz j.2.1. A=z ). this sharpens what we know from Lemma 2.34) To derive a second inequality we use the following polynomial spectral mapping property. g is bounded.32) r(A) k A k : Proposition 2.7 For each A in a unital Banach algebra one has n 1=n r(A) = nlim !1 k A k : 1 g(z ) = z 1 A k X z : k=0 By Lemma 2. In particular.2.1. zA I is not invertible. Hence limz!1 (I.35) . This is absurd. De ne the spectral radius r(A) of A 2 A by r(A) := supfjz j z 2 (A)g: From Theorem 2. accordingly. for each A 6= 0 there is a zA 2 C for which A . so that (A) 6= C hence (A) 6= . de ne p( (A)) as fp(z)j z 2 (A)g. and (2. The same theory says that (2.7. and the map A ! zA is the desired algebra isomorphism.28) implies that z!1 C. For z 6= 0 we write k g(z ) k= jz j.6 If every element (except 0) of a unital Banach algebra A is invertible.18 2 ELEMENTARY THEORY OF C -ALGEBRAS Let 2 A be a functional on A since is bounded. so that (A) = C . Comparing with (2. At this point the proof relies on the theory of analytic functions with values in a Banach space. and hence g .33) is norm-convergent for jz j > r(A). The function g.2.2. Lemma 2. Hence A . is a norm-convergent power series in z . If jz j > r(A) then z 2 (A).4.3. and lim k g(z ) k= 0: (2. (2.2. Corollary 2. for jz j >k A k the function g in the proof of Lemma 2.1 one immediately infers (2.5 has the normconvergent power series expansion (2. this isomorphism is isometric. so that g = 0 by Corollary 2.30) (2. so that lim sup k A k1=n r(A): (2.31) (2.27) converges for jz j > r(A).4.31). Then n!1 p( (A)) = (p(A)): (2.29) this constant is zero. zAI = 0 by assumption. which says that. This is true for all z for which jz j > r(A).3. so that by Liouville's theorem it must be constant. we have seen that for any z 2 (A) one may nd a z0 2 (A) such that the power series (2.1 k (I .33) cannot norm-converge uniformly in z unless k An k =jz jn < 1 for large enough n. uniformly in z . then A ' C as Banach algebras. since limz!1 k A=z k= 0 by 2.33) On the other hand.29) Now suppose that (A) = .27) converges. (2.1.1 = I. so (2.1 k and observe that limz!1 I.
so that A . with unit (I).8.35) follows. Conversely. If A is unital then A=I is unital.37) and the multiplication (A) (B ) := (AB ): (2.1 A = I must lie in I. p(A) . Hence i ( ) 2 (A). when 2 p( (A)) then = p(z ) for some z 2 (A). When 2 (p(A)) then p(A) . which implies that all A . i ( )I is not invertible. for given A 2 A. so that 2 (p(A)).2. = 0. exists which contains I. I is not invertible.38) 2I nlim !1 k A k 1=n n 1=n = inf n k A k = r(A): (2. since A. A right-ideal of A is a closed linear subspace I for which A 2 I implies AB 2 I for all B 2 A. implying that p(A) . i. observe that.36) Here the coe cients c and i ( ) are determined by p and . To conclude the proof of Proposition 2. Proposition 2.2 Banach algebra basics To prove this equality. an ideal is itself a Banach algebra. This shows the need for considering Banach algebras with and without unit it is usually harmless to add a unit to a Banach algebra A.34) yields lim sup k A k1=n r(A) k An k1=n : n!1 Hence the limit must exist.38).7. i ( )I must be invertible. Hence k An k1=n j j = r(A).39) is well de ned: when J1 J2 2 I one has (A + J1 ) (B + J2 ) = (AB + AJ2 + J1 B + J1 J2 ) = (AB ) = (A) (B ) since AJ2 + J1 B + J1 J2 2 I by de nition of an ideal.9 An ideal in a Banach algebra A is a closed linear subspace I A such that A 2 I implies AB 2 I and BA 2 I for all B 2 A. i ( )I): (2. A left-ideal of A is a closed linear subspace I for which A 2 I implies BA 2 I for all B 2 A. This proves the inclusion (p(A)) p( (A)). choose z 19 2 C and compare the factorizations = c I p(z ) .31). ~ 6= A. and one cannot add I to I without ruining the property that it is a proper ideal. An ideal I that contains an invertible element A must coincide with A.40) . i ( ) is not invertible. for each > 0 there exists a J 2 I such that k (A) k + k A + J k : (2. i ( )) (A . To prove (2.2. and (J ) = 0 for all J 2 I.. I is invertible. 2 p( (A)). by de nition of the in mum. Hence p( i ( )) . Hence 2 (p(A)) implies that at least one of the A . and (2. and De nition 2. rst note that (2. so that all B = B I must lie in I. so that for some i one must have i ( ) = z for this particular z .e. n Y = c i=1 n Y i=1 (z .8). so that i ( ) 2 (A) for at least one i. one has j n j k An k by (2. As far as the Banach algebra structure is concerned.2. we note that since (A) is closed there is an 2 (A) for which j j = r(A).39) Here : A ! A=I is the canonical projection. Since n 2 (An ) by Lemma 2. This shows that p( (A)) (p(A)). We omit the standard proof that A=I is a Banach space in the norm (2.2.2. but a given proper ideal I 6= A does not contain I. A maximal ideal is an ideal I 6= A for which no ideal I In particular. Combining this with (2.10 If I is an ideal in a Banach algebra A then the quotient A=I is a Banach algebra in the norm k (A) k:= Jinf kA+J k (2. I ~ 6= I.
41) (2. 1. the norm in A=I could not be given by (2. Theorem 2.2.8) with B = I one derives k (I) k 1.2. For the second. we know from Lemma 2.45) (2. so that ker(!) is a maximal ideal.38) that When A has a unit. since !(IA) = !(I)!(A) = !(A). in that the kernel ker(!) of each ! 2 (A) is a maximal ideal I! . z is invertible when jz j > k A k. De nition 2. Again on any vector space. By (2. Hence k (I) k= 1.44). (2.1 The structure space (A) of a commutative Banach algebra A is the set of all nonzero linear maps ! : A ! C for which !(AB ) = !(A)!(B ) for all A B 2 A. since ! is a homomorphism. from (2. On the other hand.41) with A = I one has k (I) k k I k= 1.44) (2.46) j!(A)j k A k The rst claim is obvious. AB = BA for all A B 2 A). and estimate k (A) (B ) k = k (A + J1 ) (B + J2 ) k=k ((A + J1 )(B + J2 )) k k (A + J1 )(B + J2 ) k k A + J1 k k B + J2 k (k (A) k + )(k (B ) k + ): Letting ! 0 yields k (A) (B ) k k (A) k k (B ) k.3.3. each ! 2 (A) is continuous.42) 2.3 Let A be a unital commutative Banach algebra. dim(V = ker(!)) = 1).3. Each ! 2 (A) satis es !(I) = 1 2. The kernel of each ! 2 (A) is closed. We say that such an ! is multiplicative. In other words.20 2 ELEMENTARY THEORY OF C -ALGEBRAS For if such a J would not exist. Furthermore. The kernel of every linear map ! : V ! C on a vector space V has codimension one (that is. z 6= 0.46) follows. and (2.43). with norm k ! k= 1 hence for all A 2 A. (A) consists of all nonzero homomorphisms from A to C . and !1 = !2 i I!1 = I!2 . each maximal ideal is the kernel of some ! 2 (A). for any J 2 I it is clear from (2. when ker(!1 ) = ker(!2 ) then !1 is a multiple of !2. (2. On the other hand. . Proposition 2.40) holds for A B .3. it is obvious from (2. For !i 2 (A) this implies !1 = !2 because of (2.2. since ! is continuous by 2. ker(!) is an ideal since ! satis es (2. There is a bijective correspondence between (A) and the set of all maximal ideals in A.4 that A . k (A) k=k (A + J ) k k A + J k : For A B 2 A choose > 0 and J1 J2 2 I such that (2. z ) = !(A) .39) that (I) is a unit in A=I.43) (2.38). and there is an A for which !(A) 6= 0 because ! is not identically zero.3 Commutative Banach algebras We now assume that the Banach algebra A is commutative (that is.2 Let A have a unit I. so that !(A . Hence j!(A)j 6= jz j for jz j > A k.
Hence ! 2 (A).3.46) and the triangle inequality. the claim follows. A 2 A. !(A)!(B ) = ( (A)) ( (B )) = ( (A) (B )) = ( (AB )) = !(AB ) because of (2.1 in A=I. Finally.2 we have (A) A . hence I = BA + J for suitable A 2 A J 2 I. Also. or because !(I) = 1. the w When ! 2 (A). Taking A = 0 we see I IB .39) and the fact that is a homomorphism. !(A)j+ k A k j!n (B ) . !(B )j: All in all. By the Banach-Alaoglu theorem. Therefore. on the dual B of a Banach space B is de ned by the convergence !n ! ! i !n (v) ! !(v) for all v 2 B. the absolute value of the right-hand side is bounded by 1). Since B was arbitrary (though nonzero). Hence IB = A. giving (I) = I = (BA) = (B ) (A) because of (2. there is a nonzero B 2 A which is not in I. IB is even an ideal.2. when !n ! ! in the w -topology we obtain j!(AB ) . This implies that topology on A is the weakest topology for which all A . (A) is w -compact. as I is maximal.6 this yields A=I ' C . the unit ball in A is w -compact. Now de ne a map ! : A ! C by !(A) := ( (A)). one has In the second term we write j!(AB ) . so that actually I = ker(!). Recall that the weak -topology.39) and (J ) = 0. !(A)!(B )j: !n (A)!n (B ) . Taking A = I and J = 0 we see that B 2 IB . !n(AB )j + j!n (A)!n (B ) . !(A)!(B )j j!(AB ) . In particular. This map is clearly linear. !(A)!(B )j = 0. this de nes A ^. ^. The convergence !n ! ! in the w -topology by de nition means that !n (A) ! !(A) for all A 2 A.4 The structure space (A) of a unital commutative Banach algebra A is compact and Hausdor in the Gel'fand topology.3. Hence (A) is w -closed. !(A))!n (B ) + !(A)(!n (B ) . From (2. so that the limit ! 2 (A).2. The Gel'fand topology on (A) is the relative w -topology. where We embed A in A by A ! A k B k j!n (A) . Being a closed subset of this unit ball.47) ^ as a function on (A). hence (A) = (B ). this shows that every nonzero element of A=I is invertible. ! is multiplicative it is nonzero because !(B ) 6= 0. !(B )): By (2. Since the w -topology is Hausdor (as is immediate from its de nition). By Corollary 2. !(A)!(B ) = (!n (A) .2. consisting of all functionals with norm ^(!) := !(A): A (2. Apply the canonical projection : A ! A=I to this equation.45) we have (A) 2 A1 (the unit ball in A . Form IB := fBA + J j A 2 A J 2 Ig: This is clearly a left-ideal since A is commutative. When !n 2 (A) for all n. so that there is a homomorphism : A=I ! C . so that IB 6= I.3 Commutative Banach algebras 21 We now show that every maximal ideal I of A is the kernel of some ! 2 (A). Proposition 2. !n (AB ) + !n (A)!n (B ) . since and are. I 2 IB . Since I 6= A. are continuous. By elementary functional analysis. !(A)!(B )j = j!(AB ) . also called w -topology. By 2. I ker(!) since I = ker( ) but if B 2 = I we saw that !(B ) 6= 0.
3. Theorem 2. we have showed that A 2 G(A) is equivalent to !(A) 6= 0 for all ! 2 (A). In this connection. secondly it concerns a commutative Banach algebra which is not a C -algebra. The image of A under the Gel'fand transform separates points in (A). When A 2 = G(A) the ideal IA := fAB j B 2 Ag does not contain I. the map A ! A transform. The Gel'fand transform is a homomorphism from A to C ( (A)). and thirdly the Banach algebra in question has no unit. so that it is contained in a maximal ideal I (this conclusion is actually nontrivial. so the example illustrates what happens to the structure theory in the absence of a unit.47) and (2. 2. for which !1 (A) 6= !2 (A).1 (O) = f! 2 (A)j !(A) 2 Og A (2.31). Convergence in the sup-norm is the same as uniform convergence.e.50) Hence the function 1X which is 1 for every x is the unit I. (2. The spectrum of A 2 A is the set of values of A ^) = fA ^(!)j ! 2 (A)g: (A) = (A (2.52) The rst property immediately follows from (2.47). 3.2. a basis for this topology is formed by all open sets of the form ^.51) 4. This proves 2.43).49) x2 X A basic fact of topology and analysis is that C (X ) is complete in this norm. (2. that is. ^ on (A) in other words. we regard the space C (X ) of all continuous functions on X as a Banach space in the sup-norm de ned by k f k1:= sup jf (x)j: (2. we obtain (2. When !1 6= !2 there is an A 2 A ^(!1 ) 6= A ^(!2 ). and using (2.54) .3. the Gel'fand topology on (A) is the weakest topology for which all functions A In particular.51). One checks that the spectrum of f 2 C (X ) is simply the set of values of f . We regard C ( (A)) as a commutative Banach algebra in the manner explained. ^ de ned by (2. 1. We now look at an example..48) where A 2 A and O is an open set in C . so that !(A) 6= 0 for all ! 2 (A). and (2. The Gel'fand transform is a contraction. Thus the resolvent is !(A) = fz 2 C j z 6= !(A) 8! 2 (A)g: (2. ( f + g)(x) := f (x) + g(x) (fg)(x) := f (x)g(x): (2. Hence by Theorem 2. What's more.53) Taking the complement. that is. (2. which is included for three reasons: rstly it provides a concrete illustration of the Gel'fand transform. All in all. A is invertibe).52) then follows from (2. Eq.3.1 ) = 1. then !(A)!(A.30). let us note in general that each ! 2 (A) has a suitable extension ! ~ to AI.49). z 2 G(A) i !(A) 6= z for all ! 2 (A). so that A If A 2 G(A) (i.47). relying on the axiom of choice in the guise of Hausdor 's maximality priciple). it is easily veri ed that C (X ) is even a commutative Banach algebra under pointwise addition and multiplication. namely !(A + I) := !(A) + : (2.5 Let A be a unital commutative Banach algebra.3 there is a ! 2 (A) for which !(A) = 0. ^ k1 k A k : kA (2. Hence A .22 2 ELEMENTARY THEORY OF C -ALGEBRAS ^ are continuous.47) is called the Gel'fand Seen as a map from A to C ( (A)).5. For any compact Hausdor space X .
43) then implies that ! ^ (x + y ) = ! ^ (x)^ ! (y) for almost all R dx f (x)^ ! (x): (2.56) f g(x) := Z Strictly speaking. there is no equality in the bound ^ 2 C0 (R). Indeed. Even if one does not actually extend A to AI.58) 1 for some p 2 C . Finally.54) that every multiplicative functional ! 2 (L1 (R)) is continuous.58) that f^(p) = Z R dx f (x)eipx : (2. The space C0 (X ) consists of all continuous functions on X which vanish at in nity in the sense that for each > 0 there is a compact subset K X such that jf (x)j < for all x outside K .52). as can be seen from (2. This is an important function space. i.. the existence of ! ~ shows that ! satis es (2.22) and the de nition (2. and norm k f k1 := dx jf (x)j: The associative product de ning the Banach algebra structure is convolution.5.8). which does not lie in L1 (R). we estimate R dy f (x .6 Let X be a Hausdor space X which is locally compact (in that each point has a compact neighbourhood). which (2.3. the Riemann-Lebesgue lemma states that f 2 L1 (R) implies f is the space of continuous functions on R that go to zero when jxj ! 1. Hence for each ! 2 (L1 (R)) there is a function ! ^ 2 L1(R) such that R dy jg(y)j dx jf (x)j =k f k1 k g k1 R Z R dy jg(y)j dx jf (x . Consider A = L1 (R). using Fubini's theorem on product integrals.3.56) into the pointwise product is then a restatement of Theorem 2. This extension is clearly unique. and subsequently be extended by continuity to L1(R). with the usual linear structure. many of the integral transforms of classical analysis may be seen as special cases of the Gel'fand transform).5. There is no unit in L1 (R). De nition 2. since from (2.3.58) is simply called p. y)g(y): k f g k1 = dx j dy f (x . and therefore certainly for the restriction ! of ! ~ to A. this should rst be de ned on the dense subspace Cc (R). that is. since this property (which was proved for the unital case) holds for ! ~ . Note that the Gel'fand transform is strictly a contraction.46).. so that (L1 (R)) may be identi ed with R.59) Hence the Gel'fand transform is nothing but the Fourier transform (more generally. The well-known fact that the Fourier transform maps the convolution product (2. The functional ! corresponding to (2.3 that the spectrum (f ) of f in L1 (R) is just the set of values of its Fourier transform.57) and (2.e. This implies Z The multiplicativity condition (2.3 Commutative Banach algebras 23 The point is that ! ~ remains multiplicative on AI. and since ! ^ is bounded (being in L (R)) it must be that p 2 R.43). we see from (2. Standard Banach space theory says that the dual of L1 (R) is L1 (R).57) the Gel'fand transform (2. R Z (2. We know from the discussion following (2. using the inequality below. It is clear that di erent p's yield di erent functionals.56) one sees that the unit should be Dirac's deltafunction (i.55) (2.e. . y)g(y)j R Z Z Z = Z R which is (2. With this notation. Moreover.47) reads ! ^ (x) = exp(ipx) (2. y)j R Z !(f ) = x y 2 R. the measure on R which assigns 1 to x = 0 and 0 to all other x).2.1. whose de nition may be generalized as follows. we see from 2.
2 ELEMENTARY THEORY OF C -ALGEBRAS
So when X is compact one trivially has C0 (X ) = C (X ). When X is not compact, the sup-norm (2.49) can still be de ned, and just as for C (X ) one easily checks that C0 (X ) is a Banach algebra in this norm. We see that in the example A = L1 (R) the Gel'fand transform takes values in C0 ( (A)). This may be generalized to arbitrary commutative non-unital Banach algebras. The non-unital version of Theorem 2.3.5 is Theorem 2.3.7 Let A be a non-unital commutative Banach algebra. 1. The structure space (A) is locally compact and Hausdor in the Gel'fand topology. 2. The space (AI) is the one-point compacti cation of (A). 3. The Gel'fand transform is a homomorphism from A to C0 ( (A)). ^ on (A), with zero added (if 0 is not already 4. The spectrum of A 2 A is the set of values of A contained in this set). 5. The claims 2 and 4 in Theorem 2.3.5 hold. ~ of a non-compact topological space X is the Recall that the one-point compacti cation X set X 1, whose open sets are the open sets in X plus those subsets of X 1 whose complement ~ is a compact Hausdor space, the removal of some point is compact in X . If, on the other hand, X ~ nf1g in the relative topology (i.e., the open `1' yields a locally compact Hausdor space X = X ~ sets in X are the open sets in X minus the point 1), whose one-point compacti cation is, in turn, ~. X To prove 2.3.7 we add a unit to A, and note that (AI) = (A) 1 (2.60) (2.61) where each ! 2 (A) is seen as a functional ! ~ on AI by (2.54), and the functional 1 is de ned by
1(A + I) := :
There can be no other elements ' of (AI), because the restriction of ' has a unique multiplicative extension (2.54) to AI, unless it identically vanishes on (A). In the latter case (2.61) is clearly the only multiplicative possibility. By Proposition 2.3.4 the space (AI) is compact and Hausdor by (2.61) one has (A) = (AI)nf1g (2.62) as a set. In view of the paragraph following 2.3.7, in order to prove 18.104.22.168 and 2, we need to show that the Gel'fand topology of (AI) restricted to (A) coincides with the Gel'fand topology of (A) itself. Firstly, it is clear from (2.48) that any open set in (A) (in its own Gel'fand topology) is the restriction of some open set in (AI), because A AI. Secondly, for any A 2 A, 2 C , and open set O C , from (2.54) we evidently have
f' 2 (AI)j '(A + I) 2 Ognf1g = f! 2 (A)j !(A) 2 O ; g: (When 1 does not lie in the set f: : :g on the left-hand side, one should here omit the \nf1g".)
With (2.48), this shows that the restriction of any open set in (AI) to (A) is always open in the Gel'fand topology of (A). This establishes 22.214.171.124 and 2. It follows from (2.3.5) and (2.61) that ^(1) = 0 (2.63) A ^ leads to 126.96.36.199. for all A 2 A, which by continuity of A The comment preceding Theorem 2.2.3 implies 188.8.131.52. The nal claim follows from the fact that it holds for AI.
2.4 Commutative C -algebras
The Banach algebra C (X ) considered in the previous section is more than a Banach algebra. Recall De nition 2.1.9. The map f ! f , where f (x) := f (x) (2.64) evidently de nes an involution on C (X ), in which C (X ) is a commutative C -algebra with unit. The main goal of this section is to prove the converse statement cf. De nition 2.1.12 Theorem 2.4.1 Let A be a commutative C -algebra with unit. Then there is a compact Hausdor space X such that A is (isometrically) isomorphic to C (X ). This space is unique up to homeomorphism. The isomorphism in question is the Gel'fand transform, so that X = (A), equipped with the Gel'fand topology, and the isomorphism ' : A ! C (X ) is given by ^ '(A) := A: (2.65) We have already seen in 184.108.40.206 that this transform is a homomorphism, so that (2.18) is satis ed. To show that (2.19) holds as well, it su ces to show that a self-adjoint element of A is mapped into a real-valued function, because of (2.13), (2.64), and the fact that the Gel'fand transform is complex-linear. We pick A 2 AR and ! 2 (A), and suppose that !(A) = + i , where 2 R. By (2.44) one has !(B ) = i , where B := A ; I is self-adjoint. Hence for t 2 R one computes j!(B + itI)j2 = 2 + 2t + t2 : (2.66) On the other hand, using (2.46) and (2.16) we estimate j!(B + itI)j2 k B + itI k2 =k (B + itI) (B + itI) k=k B 2 + t2 k k B k2 +t2 : Using (2.66) then yields 2 + t k B k2 for all t 2 R. For > 0 this is impossible. For < 0 we repeat the argument with B ! ;B , nding the same absurdity. Hence = 0, so that !(A) ^ is real-valued, and (2.19) follows as is real when A = A . Consequently, by (2.47) the function A announced. We now prove that the Gel'fand transform, and therefore the morphism ' in (2.65), is isometric. When A = A , the axiom (2.16) reads k A2 k=k A k2 . This implies that k A2m k=k A k2m for all m 2 N . Taking the limit in (2.32) along the subsequence n = 2m then yields r(A) =k A k : (2.67) In view of (2.30) and (2.51), this implies ^ k1 =k A k : kA (2.68) For general A 2 A we note that A A is self-adjoint, so that we may use the previous result and (2.16) to compute ^ k1 =k A ^ k2 d ^A k A k2=k A A k=k A A k1 =k A 1: c =A ^ , which we just proved, and in the fourth we exploited the In the third equality we used A fact that C (X ) is a C -algebra, so that (2.16) is satis ed in it. Hence (2.68) holds for all A 2 A. It follows that ' in (2.65) is injective, because if '(A) = 0 for some A 6= 0, then ' would fail to be an isometry. (A commutative Banach algebra for which the Gel`fand transform is injective is called semi-simple. Thus commutative C -algebraa are semi-simple.) We nally prove that the morphism ' is surjective. We know from (2.68) that the image ^ is closed in C ( (A)), because A is closed (being a C -algebra, hence a Banach space). '(A) = A In addition, we know from 220.127.116.11 that '(A) separates points on (A). Thirdly, since the Gel`fand transform was just shown to preserve the adjoint, '(A) is closed under complex conjugation by ^ = 1X by (2.44) and (2.47), the image '(A) contains 1X . The surjectivity (2.64). Finally, since I of ' now follows from the following Stone-Weierstrass theorem, which we state without proof.
2.4 Commutative C -algebras
2 ELEMENTARY THEORY OF C -ALGEBRAS
algebra as explained above. A C -subalgebra of C (X ) which separates points on X and contains 1X coincides with C (X ). Being injective and surjective, the morphism ' is bijective, and is therefore an isomorphism. The uniqueness of X is the a consequence of the following result. Proposition 2.4.3 Let X be a compact Hausdor space, and regard C (X ) as a commutative C algebra as explained above. Then (C (X )) (equipped with the Gel`fand topology) is homeomorphic to X . Each x 2 X de nes a linear map !x : C (X ) ! C by !x (f ) := f (x), which is clearly multiplicative and nonzero. Hence x ! !x de nes a map E (for Evaluation) from X to (C (X )), given by E (x) : f ! f (x): (2.69) Since a compact Hausdor space is normal, Urysohn's lemma says that C (X ) separates points on X (i.e., for all x 6= y there is an f 2 C (X ) for which f (x) 6= f (y)). This shows that E is injective. We now use the compactness of X and Theorem 2.3.3 to prove that E is surjective. The maximal ideal Ix := I!x in C (X ) which corresponds to !x 2 (C (X )) is obviously
Lemma 2.4.2 Let X be a compact Hausdor space, and regard C (X ) as a commutative C -
Ix = ff 2 C (X )j f (x) = 0g: (2.70) Therefore, when E is not surjective there exists a maximal ideal I C (X ) which for each x 2 X contains at a function fx for which fx (x) 6= 0 (if not, I would contain an ideal Ix which thereby would not be maximal). For each x, the set Ox where fx is nonzero is open, because f is continuous. This gives a covering fOx gx2X of X . By compactness, there exists a nite P N subcovering fOxi gi=1 ::: N . Then form the function g := i=1 jfxi j2 . This function is strictly positive by construction, so that it is invertible (note that f 2 C (X ) is invertible i f (x) 6= 0 for all x 2 X , in which case f ;1 (x) = 1=f (x)). But I is an ideal, so that, with all fxi 2 I (since all fx 2 I) also g 2 I. But an ideal containing an invertible element must coincide with A (see the comment after 2.2.9), contradicting the assumption that I is a maximal ideal. Hence E is surjective since we already found it is injective, E must be a bijection. It remains to be shown that E is a homeomorphism. Let Xo denote X with its originally given topology, and ^ E = f by (2.69) and (2.47), and the write XG for X with the topology induced by E ;1 . Since f ^ are continuous, Gel'fand topology on (C (X )) is the weakest topology for which all functions f we infer that XG is weaker than Xo (since f , lying in C (Xo ), is continuous). Here a topology T1 is called weaker than a topology T2 on the same set if any open set of T1 contains an open set of T2 . This includes the possibility T1 = T2 . Without proof we now state a result from topology. Lemma 2.4.4 Let a set X be Hausdor in some topology T1 and compact in a topology T2 . If T1 is weaker than T2 then T1 = T2 . Since Xo and XG are both compact and Hausdor (the former by assumption, and the latter by Proposition 2.3.4), we conclude from this lemma that X0 = XG in other words, E is a homeomorphism. This concludes the proof of 2.4.3. Proposition 2.4.3 shows that X as a topological space may be extracted from the Banachalgebraic structure of C (X ), up to homeomorphism. Hence if C (X ) ' C (Y ) as a C -algebra, where Y is a second compact Hausdor space, then X ' Y as topological spaces. Given the isomorphism A ' C (X ) constructed above, a second isomorphism A ' C (Y ) is therefore only possible if X ' Y . This proves the nal claim of Theorem 2.4.1. The condition that a compact topological space be Hausdor is su cient, but not necessary for the completeness of C (X ) in the sup-norm. However, when X is not Hausdor yet C (X ) is complete, the map E may fail to be injective since in that case C (X ) may fail to separate points on X .
2.4 Commutative C -algebras
On the other hand, suppose X is locally compact but not compact, and consider A = Cb (X ) this is the space of all continuous bounded functions on X . Equipped with the operations (2.49), (2.50), and (2.64) this is a commutative C -algebra. The map E : X ! (Cb (X )) is now injective, but fails to be surjective (this is suggested by the invalidity of the proof we gave for C (X )). Indeed, it can be shown that (Cb (X )) is homeomorphic to the Ceh-Stone compacti cation of X . Let us now consider what happens to Theorem 2.4.1 when A has no unit. Following the strategy we used in proving Theorem 2.3.7, we would like to add a unit to A. As in the case of a general Banach algebra (cf. section 2.2), we form AI by (2.21), de ne multiplication by (2.22), and use the natural involution (2.71) (A + I) := A + I: However, the straightforward norm (2.23) cannot be used, since it is not a C -norm in that axiom (2.16) is not satis ed. Recall De nition 2.1.4. Lemma 2.4.5 Let A be a C -algebra. 1. The map : A ! B(A) given by (A)B := AB (2.72) establishes an isomorphism between A and (A) B(A). 2. When A has no unit, de ne a norm on AI by k A + I k:=k (A) + I k (2.73) where the norm on the right-hand side is the operator norm (2.2) in B(A), and I on the right-hand side is the unit operator in B(A). With the operations (2.22) and (2.71), the norm (2.73) turns AI into a C -algebra with unit. By (2.15) we have k (A)B k=k AB k k A kk B k for all B , so that k (A) k k A k by (2.2). On the other hand, using (2.16) and (2.17) we can write
(2.74) Being isometric, the map must be injective it is clearly a homomorphism, so that we have proved 18.104.22.168. It is clear from (2.22) and (2.71) that the map A + I ! (A) + I (where the symbol I on the left-hand side is de ned below (2.22), and the I on the right-hand side is the unit in B(A)) is a morphism. Hence the norm (2.73) satis es (2.15), because (2.8) is satis ed in B(A). Moreover, in order to prove that the norm (2.73) satis es (2.16), by Lemma 2.1.11 it su ces to prove that k (A) + I k2 k ( (A) + I) ( (A) + I) k (2.75) for all A 2 A and 2 C . To do so, we use a trick similar to the one involving (2.40), but with inf replaced by sup. Namely, in view of (2.2), for given A 2 B(B) and > 0 there exists a v 2 V , with k v k= 1, such that k A k2 ; k Av k2 . Applying this with B ! A and A ! (A) + I, we infer that for every > 0 there exists a B 2 A with norm 1 such that k (A) + I k2 ; k ( (A) + I)B k2 =k AB + B k2 =k (AB + B ) (AB + B ) k : Here we used (2.16) in A. Using (2.72), the right-hand side may be rearranged as k (B ) (A + I) (A + I)B k k (B ) k k ( (A) + I) ( (A) + I) k k B k : Since k (B ) k=k B k=k B k= 1 by (2.74) and (2.17), and k B k= 1 also in the last term, the inequality (2.75) follows by letting ! 0. Hence the C -algebraic version of Theorem 2.2.1 is
k A k=k AA k = k A k=k (A) kA Ak k in the last step we used (2.4) and k (A = k A k) k= 1. Hence k (A) k=k A k :
The spectrum (A) is a subset of R. It is su cient for our purposes to restrict ourselves to this case. in terms of I = 1X we may therefore write 1 ff k : 1= f X (2.23) and (2. for any f 2 C (X ) that is nonzero throughout X we have 0 <k f k.2.3.64). (A ). A .7 may then be repeated in the C -algebraic setting the only nontrivial point compared to the situation for Banach algebras is the generalization of Lemma 2.5 to Theorem 2. an element A 2 A is called normal when A A ] = 0. A.1.3 below. 1. (A ).5 Spectrum and functional calculus f k f k2 1 k=0 k f k2 1 .6. Hence C (A A. and I. so that C (A I) is isomorphic ^ : (A) ! R is the identity to C ( (A)). having established the existence of the unitization of an arbitrary non-unital C algebra. we conclude that the unital Banach algebra AI called for in that proposition is not. The passage from Theorem 2. The crucial property of a normal operator is that C (A I) is commutative.4. The uniqueness of AI follows from Corollary 2. 12 ff 12 ff < 1 pointwise. This space is unique up to homeomorphism. and regard C0 (X ) as a commutative C -algebra as explained below De nition 2. we see that. k f k. A ^ is invertible in C (X ) (i. 2. in particular. and is such that for each x 2 X there is an f 2 A such that f (x) 6= 0.1 we have C (A A. C (A I) is simply the closure of the space of all polynomials in A.7 Let X be a locally compact Hausdor space.1 . 2. A.73) de ne a norm on AI satisfying the claims of Proposition 2.28 2 ELEMENTARY THEORY OF C -ALGEBRAS Proposition 2.8 Let A be a commutative C -algebra without unit. in general.1 . Following the terminology for operators on a Hilbert space. A(x) 6= 0x for all x 2 X ). and I.25). One has (A. but analyze properties of A by studying certain commutative subalgebras. This now reads Lemma 2. In particular. 1 pointwise. and consider the C -algebra C (A A. when A is self-adjoint.e.1 .2.4. The structure space (C (A I)) is homeomorphic with (A). and A. in view of the fact that both (2.3. hence k 1X .76) I. For each element A 2 A there is a smallest C -subalgebra C (A I) of A which contains A and I. Using Lemma 2. Since A is invertible and the Gel'fand transform (2. However.4. Under this isomorphism the Gel'fand transform A function id (A) : t ! t.3. coincides with C0 (X ).1 ) = (A ). ff = k f k2 1 k1 < 1: Here f is given by (2. a commutative non-unital C -algebra has a unitization.5. Recall (2. At the end of the day we then nd Theorem 2.6 For every C -algebra without unit there exists a unique unital C -algebra AI and an isometric (hence injective) morphism A ! AI. so that 0 1X . Let A 2 G(A) be normal in A. In any case. such that AI=A ' C . A. 3. namely the closure of the linear span of I and all operators of the type A1 : : : An .1 Let A = A be a self-adjoint element of a unital C -algebra. unique. There is a locally compact Hausdor space X such that A is (isometrically) isomorphic to C0 (X ). where Ai is A or A . A . A C -subalgebra A of C0 (X ) which separates points on X . By Theorem 2. This will lead to important results. On the other hand. The spectrum A (A) of A in A coincides with the spectrum C (A I)(A) of A in C (A I) (so that we may unambiguously speak of the spectrum (A)).2.47) is an isomorphism..1 I) ' C (X ) for some compact ^ Hausdor space X .4.1 I) generated by A.1 and I all commute with each other.4.1 . Theorem 2.5.1 . We return to the general case in which a C -algebra A is not necessarily commutative (but assumed unital).1 I) is commutative it is the closure of the space of all polynomials in A.4.
Theorem 2. the sup-norm of id (A) is r(A).3 yields an isomorphism C ( (A)) ! C (A I).80) Since the spectrum is determined by the algebraic structure alone. We now prove injectivity.5 Spectrum and functional calculus 29 ^.3 the According to Theorem 2. we conclude by linearity that !1 = !2 on all polynomials in A.1.77) follows from (2.5. So if we assume that A . Since also !1 (I) = !2(I) = 1 by (2. k A k= r(A A): p (22.214.171.124 The norm in a C -algebra is unique (that is. Gel'fand transHence A forming this result back to C (A A. In 2.31)).2) this implies that !1 = !2 on C (A I). and apply (2. Using (2.5.44). (126.96.36.199. Corollary 2. so that by the previous result (A) is real.77) (2.78).3.3 the function A When !1 !2 2 X and !1 (A) = !2 (A).1. To prove continuity of the inverse. Hence A. Hence by 2.5.43) with B = A.1 (z ) 2 (C (A I)) maps A to z (and hence An to z n .1 is continuous.1) yields (2. leading to the conclusion that the resolvent A (A) in A coincides with the resolvent C (A I)(A) in C (A I).1 is a norm-convergent limit of a sequence of polynomials in A and A .3.2.1 is a norm-convergent limit of a sequence of polynomials in A ^ and A ^ . The last claim follows by combining 2.4. When A is normal A . that for z 2 (A) the functional A ^.3. one checks Since A 2 C (X ) by 2. according to 2.1 I) = C (A I).79) Since A A is self-adjoint for any A. that A ^ ^ is continuous. for all n 2 N .77) and (2.1 (where X = (C (A I))). A ^.). ^ is real-valued when A = A . we infer that A. etc.1 lies in C (A I). 2. in fact.48). which is the obvious expression when f is a polynomial (and in general is given via the uniform approximation of f by polynomials). z is normal. Looking at (2.78). By de nition (cf. ^ is a surjective map from X to (A). A ^ is a homeomorphism. By De nition 2.80) shows that the norm is determined by the algebraic structure as well. In conclusion. we have !1 (An ) = !1 (A)n = !2 (A)n = !2 (An ) by iterating (2.16). (2.78) with f = id (A) . z 2 G(A) the argument above applies. By continuity (cf. Corollary 2.1 with (2. the norm of f (A) in C (A I) coincides with its norm in A. Since f ( (A)) is the set of values of f on (A). Note that Corollary 2.2 we then conclude that A (A) = C (A I)(A).5.51).3 is then obvious. given the isomorphism C (A I) ' C (X ) of Theorem 2. and C (A A. An immediate consequence of this theorem is the continuous functional calculus.5. (2. then. Now replace A by A . First assume A = A . for general A 2 A we have. such that (f (A)) = f ( (A)) k f (A) k = k f k1 : (2. which is precisely the map f ! f (A) of the continuous functional calculus.2.1 I). given a C -algebra A there is no other norm in which A is a C -algebra). Finally. with A ! f (A).5. z .2 For each self-adjoint element A 2 A and each f 2 C ( (A)) there is an operator f (A) 2 A.1.47). using (2.5. . one then sees that A The nal claim in 2.4. we have proved ^(!1 ) = A ^(!2 ) implies !1 = !2 .3 the completeness of A is assumed from the outset.78) In particular. where z 2 C .5. so that k A k= r(A) (A = A ): (2. the function A spectrum C (A I)(A) is real.3 does not imply that a given -algebra can be normed only in one way so as to be completed into a C -algebra (we will. encounter an example of the opposite situation).5. since the linear span of all polynomials is dense in C (A I).3. The fact that this isomorphism is isometric (see 2.
81) When z 2 (A). z ). Hence (A) ('(A)). transform A Proposition 2.1 An element A of a C -algebra A is called positive when A = A and its spectrum is positive i.30 2 ELEMENTARY THEORY OF C -ALGEBRAS Corollary 2.6. then '(A . 1. which we would like to generalize to arbitrary abstract C -algebras. so that (A . play a central role in the proof of the Gel'fand Neumark theorem.18) implies that ('(A .3 and 2. (A) R+ .82)). z ).18). z ) is certainly invertible in B. In quantum mechanics this means that the expectation value of the observable A is always positive. A bounded operator A 2 B(H) on a Hilbert space H is called positive when ( A ) 0 for all 2 H this property is equivalent to A = A and (A) R+ .6 Positivity in C -algebras De nition 2.6.1 ).e.3.5. (2. when A 2 A+ and t 2 R+ then tA 2 A+ 2. a function f on some space X is positive simply when f (x) 0 for all x 2 X . We write A 0 or A 2 A+ . to elements of the commutative C -algebra C0 (X ) (where X is a locally compact Hausdor space). Classically. for example. for (2. This applies.2 The set A+ of all positive elements of a C -algebra A is a convex cone that is.5. and so is f ('(A)) in view of (2. For general f the result then follows by continuity.83) for all f 2 C ( (A)) (here f (A) is de ned by the continuous functional calculus.5 When ' : A ! B is a morphism and A = A then f ('(A)) = '(f (A)) (2. when A B 2 A+ then A + B 2 A+ 3. since for those f has its naive meaning.5. in particular.1.84) It is immediate from Theorems 2.A+ = 0.1 exists in A.82) Hence r('(A)) r(A). Lemma 2.4 A morphism ' : A ! B between two C -algebras satis es k '(A) k k A k and is therefore automatically continuous. where A+ := fA 2 ARj (A) R+ g: (2.. The property is true for polynomials by (2. A+ \ . z )).1 = '((A .3 that A 2 AR is positive i its Gel'fand ^ is pointwise positive in C ( (A)). In particular.5.81) follows from (2. . so that ('(A)) For later use we note (A): (2.80). and clearly also applies to closed subalgebras of B(H). 2. so that (2. Positivity is one of the most important features in a C -algebra it will. Hence we have a notion of positivity for certain concrete C -algebras. one is interested in nding a number of equivalent characterizations of positivity.
. because a convex cone in a real vector space is equivalent to a linear partial ordering. where id (A) (t).88) p When (A) R+ andp A = A then A 2 AR is de ned by the continuous functional calculus p for f = . The real vector space in question is the space AR of all self-adjoint elements of A. A ^ k1 c. B BA. this implies k cI.5. k B k k B k]. f. which is a special case of (2.A. ) BA. which also implies that kI . Then k cI .86) then follows.2. = . This is important..77) and 2. we have jc . For later use we also record Lemma 2.kI 0 (since k 0) we obtain . when A = A one checks the validity of .77) with f (t) = t3 that (A. Then (A.77). Apply the continuous functional calculus with f = id (A) = f+ .3 and 2.2. so that (A) . when A 2 A+ and A 2 .85) by taking the Gel'fand transform of C (A I). Hence A+ fA2 j A 2 ARg.5 Every self-adjoint element A has a decomposition A = A+ .kI A kI. Moreover. tj c for all t 2 (A) and all c r(A). We now come to the central result in the theory of positivity in C -algebras. a partial ordering in which A B implies A + C B + C for all C and A B for all 2 R+ . hence k A k k B k by (2. The implication . The opposite inclusion follows from (2. where A+ A.6. so that k c1 (A) .87) = fB B j B 2 Ag: (2.B ) k c where in the last step we used the previous paragraph for A and for B separately.1. hence 0 A kI . B by the linearity of the partial ordering. ) R+ because A. k A k k A k.2.3. which generalizes the cases A = B(H) and A = C0 (X ).78) with A ! A . and f. .87). hence A = 0 by (2. = . A k c for some c k A k. A 2 AR .30).A.kAk I A kAk I (2.t 0g. By (2.6 Positivity in C -algebras 31 The rst property follows from (tA) = t (A).AA. as 0 B . k B k I A k B k I.(BA.B A B and (2. (A+ . Hence . this inequality implies A + B 2 A+ .3 When A B 2 A+ and k A + B k k then k A k k.87). A k c for all c k A k by 2.(BA.79) and (2. i.A+ it must be that (A) = 0.6. Inverting this argument. and given one puts AR = fA 2 AR j 0 Ag. Gel'fand transformsupt2 (A) jc1 (A) .(t) = maxf. f+ (t) = maxft 0g. 2 A+ and A+ A. The inclusion A+ fB B j B 2 Ag is is trivial from (2. and satis es A2 = A.6.5. = 0. )3 = . the bound follows from (2. We use this lemma to prove that fB B j B 2 Ag A+. Since (A) 0 r(A)].3. from which the lemma follows by (2. : Since (A. Hence. because . B kI.86).5.4 One has A+ = fA2 j A 2 ARg (2. A. then (A) R+ . A. A ing back to C (A I). one sees that if k cI .79) and (2. (A + B ) k k (k A k . we see from (2. = . )A. As we have seen. . Theorem 2.4. 0. )3 0.B A B =) k A k k B k (2. Apply the lemma to A = B B (noting that A = A ). Finally.5.1.A. Lemma 2.85) we have A + B kI. For example.85) for A ! B yield .e. ) BA.30). using . This proves (2.1.1.A) k + k (k B k . Hence ^j c by 2.79)). + The equivalence between these two structures is as follows: given A+ B if R := A one de nes A + + B . Use this with A ! A + B and c =k A k + k B k clearly c k A + B k by 2. is positive. Since k f k1 r(A) =k A k (where we used (2.
but an approximate unit may be constructed as follows: take = N . so that A1 A2 implies B A1 B B A2 B . and use (2. I =I (2. i. For example. When A is separable (in containing a countable dense subset) then may be taken to be countable. z . and this is nothing but B AB B A2 B . An important consequence of (2. and (I ) 2. (2. for all A 2 A.6 If .3. .1 An approximate unit in a non-unital C -algebra A is a family fI g 2 . yielding A A k A k2 I. As we have seen. where is some directed set (i.e.93) (2.CC ) R+ . B A AB k A k2 B B (2. Hence C = 0 by 2. )3 = . Namely.e. ) BA. By (2.91) An ideal I in a C -algebra A is de ned by 2.32 2 ELEMENTARY THEORY OF C -ALGEBRAS Lemma 2. and nally A. But clearly (A3 B ) A3 B 0.16). z implies the invertibility of BA .1 A .n n] and vanishes for jxj > n + 1.85) by A A.87). = 0. = 0 we see that (A. A k= 0 lim !1 !1 For example. so that kI k 1 k I A .13) we can write C = D + iE ..2.A1 0 by (2.92) (2. A1 = A3 A3 . This is because A1 A2 is the same as A2 . hence (. Applying this with A ! C and B ! C we see that the assumption (C C ) R. (C C ) R+ . the C -algebra C0 (R) has no unit (the unit would be 1R.90) This is because for z = 6 0 the invertibility of AB . . z ). = 0 by the continuous functional calculus with f (t) = t1=3 .6. By (2. and take In to be a continuous function which is 1 on . Proposition 2. D E 2 AR. One checks the axioms. Applying the above principle gives for all A B 2 A.7.2.6. so that the assumption . implies (CC ) R. z ).6. A k= lim k AI . with the following properties: 1. )3 = 0. one computes that (BA . a set with a partial order and a sense in which ! 1). a proper ideal cannot contain I in order to prove properties of ideals we need a suitable replacement of a unit.88) there is an A3 2 A such that A2 .1I.2 we see that C C 0. and notes that one certainly does not have In ! 1R in the sup-norm.2 Every non-unital C -algebra A has an approximate unit. and 2.9.88) is the fact that inequalities of the type A1 A2 for A1 A2 2 AR are stable under conjugation by arbitrary elements B 2 A.7.89). z .C C 2 A+ for some C 2 A then C = 0.89) Now for any A B 2 A one has (AB ) f0g = (BA) f0g: (2. The last claim before the lemma therefore implies BA.1 = B (AB . 2.2.(BA. Hence B B = A+ . replace A in (2..C C 2 A+ now yields (C C ) = 0. which lies in A+ . CC : (2.7 Ideals in C -algebras De nition 2.94) 0 1]. which does not vanish at in nity because it is constant). so that C C = 2D2 + 2E 2 . As (A.
3 Let I be an ideal in a C -algebra A.92).1 : (2. A k2 = lim k (I A . from which we build the element B := i Ai Ai . JJ I ) . JJ I ) k k (JJ . one has I = B C + B .1 ).2 it has an approximate unit fI g.78) with A ! B and f (t) = n.4 Let fI g be an approximate unit in I. I (JJ . it follows from (2.10. we estimate k J . Ai .2 t(n. I \ I is a C -subalgebra of A. 1 . one has supt2R+ jf (t)j = 1=4n.78).96) We now apply (2. Since f 0 and f assumes its maximum at t = 1=n.39).1 + t).93). JJ I ) k + k I (JJ . AI k : !1 (2. one has I = I . Clearly B is self-adjoint. Using (2. In view of 2.6.77) and the positivity of B that (I ) 0 1]. so that k i Ci Ci k 1=4n by (2. so that it is of the form C + I for some C 2 A and 2 C . hence a left-ideal.4 and 2. Hence lim !1 k J .3.7 Ideals in C -algebras 33 One takes to be the set of all nite subsets of A. J I k= 0.38). 2 from a Theorem 2. 1. also using (2. it follows that J 2 I. JJ I ) k k (J J . which lies in A. A) I A .2 . Hence J is a norm-limit of elements in I since I is closed.97) Note that (2.1 I + B ).7.2 one has (B ) R+ .3.7. JJ I ) k : As we have seen. Although (n. A k= lim k C C k= 0: !1 !1 !1 i i The other equality in (2.7.1 I+ B ). so that n. P .3 then shows that k Ci Ci k 1=4n for each i = 1 : : : n. I J )(J .2.2. the multiplication (2. Finally.6. J I k2 =k (J . so certainly I 2 I. Note that J 2 I implies J J 2 I \ I : it lies in I because I is an ideal. This proves 2. and since I is an ideal it must be that J I 2 I for all .98) . so that. Hence by 2. with f (t) = t=(n + t). J I ) k=k (JJ . Hence we may form I := B (n. Take J 2 I. 2 . Lemma 2.2.1 I + B ). Then k (A) k= lim k A . hence a right-ideal. But I lies in I \ I . Since any A 2 A sits in some directed subset of with n ! 1. Putting Ci := I Ai . so that is countable. 2. Put I := fA j A 2 Ig.1 I + B ). The quotient A=I is a C -algebra in the norm (2.7.94) follows analogously. every ideal in a C -algebra is self-adjoint. and it lies in I because I is an ideal.188.8.131.52 B (n. The main properties of ideals in C -algebras are as follows. a simple computation shows that X i Ci Ci = n. Since I is an ideal.2 is the property (2. partially ordered byP inclusion.1 I+ B is invertible in AI. it follows that k f k1 1=4n. If A 2 I then A 2 I in other words.95) Since B is self-adjoint and B commutes with functions of itself (such as (n. As (B ) R+ . J J I ) k + k I k k (JJ . Using the continuous functional calculus on B . one sees from (2. 2 hence k n B (n I + B ) k 1=4n by (2. Hence 2 is of the form = fA1 : : : An g. and let A 2 A. J J 2 I \ I .1. and the involution (A) := (A ): (2.1 is computed in AI.96). This uses Lemma 2.16). when A is separable one may draw all Ai occurring as elements of countable dense subset.97) is well de ned because of 2.2 : (2.16) that lim k I A . both terms vanish for ! 1.7. all we need to prove to establish 2.16) and (2.6. and according to 2.
I ) k !1 !1 !1 =k (A A) k=k (A) (A ) k=k (A) (A) k : Lemma 2. I ) k lim k I .3. Hence every morphism has norm 1. AI k2 = lim !1 !1 k A . In particular. is a C -algebra. Then (2. is a C -algebra morphism. since I is the kernel of the canonical projection : A ! A=I. We now prove (2.5. pick any J 2 I. since k k= 1. and write k A . we nd k (A . and (2.5. and ' = . .99). since J 2 I. inheriting all operations in B.7. The nal claim follows from the preceding one. so that f ('(A)) = 0. By Lemma 2. AI k : !1 Letting ! 0 proves (2. The rst claim is almost trivial. where A=I is a C -algebra. J k : Note that k I. and (2. since '(A) = 0 implies '(AB ) = '(BA) = 0 for all B by (2.7.98).40) holds.7.101) !1 For each > 0 we can choose J 2 I so that (2.2.18).11 then implies (2. contradicting the injectivity of '.7 The image of a morphism ' : A ! B between two C -algebras is closed. is a vector space isomorphism between A= ker(') and '(A). This seemingly technical result is very important. I ) k lim k A A(I .80) and (2.2. and (2.7.38) that 2 ELEMENTARY THEORY OF C -ALGEBRAS To derive the opposite inequality. According to 2. Assume there is an B 2 A for which k '(B ) k6=k B k. and (2. By Urysohn's lemma there is a nonzero f 2 C ( (A)) which vanishes on ('(A)).16) in AI.6 An injective morphism between C -algebras is isometric. For the next consequence of 2. Since ' is a morphism. Put A := B B . (2.5 and 2.97).16) in A=I. Since ' and are morphisms. AI k k (A) k : (2. its image is a -algebra in B.6.7. Corollary 2.101). (2.39). Also.1 and the proof of 2.7. I k + k J I . AI k k A + J k : (2. Conversely. (2. so that ' has closed range in B. I k k A A(I . I k k A + J k k I .7.3 we need a Lemma 2.3. By the theory of vector spaces. is injective.100). (2.30) we must have (A) 6= ('(A)). and is a morphism by (2. k (A) k k A . By (2.100) by 2.7. AI k=k (A + J )(I . add a unit I to A if necessary. Corollary 2.16). In particular. AI k . since ' is continuous (see 2. But (A= ker(')) = '(A). (2. Hence (A= ker(')) has closed range in B by 2. (2.16). its range is closed.82) implies ('(A)) (A).97). the space A= ker(') is a C -algebra. For this speci c J we combine (2. I ) + J (I .98).5 The kernel of a morphism between two C -algebras is an ideal. The second term on the right-hand side goes to zero for ! 1.99) = lim k (I .40) to nd lim k A . every ideal in a C -algebra is the kernel of some morphism. AI ) k k (A) k2 = lim k A . '(A) is a C -subalgebra of B. The converse follows from Theorem 2.18).39). I )A A(I . I k 1 (2.6.1. noting that A = A.7. which by the preceding sentence is closed in the norm of B. Successively using (2. where A] is the equivalence class in A= ker(') of A 2 A. In particular. AI ) (A . By (2.4) its kernel is closed.1. De ne : A= ker(') ! B by ( A]) = '(A).5 we have '(f (A)) = 0. Hence lim k A .19) this implies k '(B B ) k6=k B B k. Hence '(A).98).34 It is obvious from (2.
1 A state on a unital C -algebra A is a linear map ! : A ! C which is positive in that !(A) 0 for all A 2 A+. the bound j!(A)j !(I) k A k. Choose s > 0 small enough.106) In particular: 1. Partly in order to extend the de nition of a state to non-unital C -algebras. we show that A 0 implies !(A) 0.2 The state space of A = C (X ) consists of all probability measures on X . we have Proposition 2.4. By the Riesz theorem of measure theory. . and normalized in that !(I) = 1: (2. Next.16).104) with A = I.8.105) !(A ) = !(A) as !(A ) = !(A I) = (A I)! = (I A)! = !(A). Hence from (2.106). we rst note that the argument around (2. sA)j : 1 k I . each positive linear map ! : C (X ) ! C is given by a regular positive measure ! on X .8. sA k= k !(I) k I . we have (2. The positivity of ! with 2. A state on a unital C -algebra is bounded. we look at a state in a more general context.66) may be copied. and normalization is obvious from (I) = ( ) = 1. The normalization of ! implies that !(1X ) = ! (X ) = 1.6.8 States 35 2. When ! is positive and A = A we have.8 States The notion of a state in the sense de ned below comes from quantum mechanics. but states play a central role in the general theory of abstract C -algebras also. Since the upper bound is reached by B = I.3 A linear map ! : A ! C on a unital C -algebra is positive i ! is bounded and k ! k= !(I): (2. sA k 1. since (B B ) =k B k2 0. For general A we use (2. and the bound just derived to nd j!(B )j2 !(B B )!(I) !(I)2 k B B k= !(I)2 k B k2 : Hence k ! k !(I). We now pass to states on C -algebras without unit.85). with norm 1. for all A 2 A one has (2. (2. when A B(H) then every 2 H with norm 1 de nes a state by (A) := ( A ): (2.8. Firstly. s!(A)j !(I). 2.2. Theorem 2. so that k I . Then (assuming ! 6= 0) !k j!(I . To prove the converse claim.1) we obtain j!(A B )j2 !(A A)!(B B ) (2. De nition 2.102) The state space S (A) of A consists of all states on A.103) Positivity follows from Theorem 2.104) which will often be used. sA k !(I) Hence j!(I) .4 implies that (A B )! := !(A B ) de nes a pre-inner product on A.6. For example. Moreover. so that ! is a probability measure. using (2. which is only possible when !(A) 0. An element ! 2 A for which k ! k= !(I) = 1 is a state on A. showing that ! is real on AR.
iA.8. Hence in the non-unital case we may replace the normalization condition in 2. one still has j!(A)j2 !(A A).8. Since a state ! on A is bounded by 2.8. Since k A0 k k A k and k A00 k k A k by (2. Since Q is positive. we have j!(A . There are lots of states: 0: Lemma 2. then 00 k Q(A) k k Q(A0+ ) k + k Q(A0.8.1 as follows.86) A+. and therefore on A by the previous paragraph.36 2 ELEMENTARY THEORY OF C -ALGEBRAS De nition 2. where A0+ etc.106). The extension in question is de ned by !I(A + I) := !(A) + : (2.104) and (2. De nition 2.108) that !I((A + I) (A + I)) j!(A) + j2 Hence ! is positive by (2.105).6 A state on a C -algebra A is a linear map ! : A ! C which is positive and has norm 184.108.40.206 because of (2.5 A positive map between two C -algebras is bounded (continuous).8. When A = A there exists a state ! such that j!(A)j =k A k. which is impossible since k Q(A) k is some nite number. ) k + k Q(A00 4C k A k : + ) k + k Q(A. This de nition is possible by 2. we obtain from (2.2An obviously converges to some A 2 A+.8. and is consistent with 2.13) and 2.8.8. Thus Q is bounded on Choosing B = C .5 then provides an alternative proof of 2. A0. .7 A state ! on a C -algebra without unit has a unique extension to a state !I on the unitization AI. The following result is very useful cf.107) + . are positive. A00 C k B k for all B 2 A+ and some + .108) This obviously satis es (2. Let us rst show that boundedness on A+ implies boundedness on A.5.5.102) it remains to prove positivity.8 For every A 2 A and a 2 (A) there is a state !a on A for which !(A) = a. AI )j ! 0 for any approximate unit in A. Combining this with (2. we can write 00 A = A0+ .8.105) may then be copied from the unital case in particular. Hence if k Q(B ) k C > 0. Proposition 2. The derivation of (2.4 A positive map Q : A ! B between two C -algebras is a linear map with the property that A 0 implies Q(A) 0 in B.220.127.116.11 k Q(An ) k n for all n 2 N . The series n=0 n. 2. by 2. Proposition 2. so that + for each n 2 N there isP an An 2 A+ n3 (here A+ 1 so that k Q(An ) k 1 consists of all A 2 A with 1 k A k 1). we see that a state on a unital C -algebra is a special case of a positive map between C -algebras Proposition 2. + iA00 (2.8. Using (2. ) k Now assume that Q is not bounded by the previous argument it is not bounded on A+ . we have Q(A) n. k Q(A) k n. we have k B k k A k 00 for B = A0+ A0. Hence by (2.5.6. or A.4.2 Q(An ) 0 for each n.1.13).88).
there exists an extension !a of ! ~ to A of norm 1.110) for all A B 2 A. Since a 2 (A) one has a + 2 ( A + I) this easily follows from the de nition of . Repeating this process.41) with A ! A + I implies j! ~a ( A + I)j k ( A + I) k.8.81). An important feature of a state space S (A) is that it is a convex set. Hence (2.8. k (A) k=k A k when is faithful by Lemma 2.2.3 that each element ! of S (A) is continuous. We conclude that S (C C ) may be identi ed with the interval 0 1]. which clearly satis es !a (A) = ! ~ a (A) = a.9 Representations and the GNS-construction 37 where x y z 2 R. in which case S (A) is a point.2. The positivity of this matrix then corresponds to the constraint x2 + y2 + z 2 1. We saw in 2.8. Positive elements of C C are fo the form + _ with of the form !( + 0 _ and 0. We return to the unital case. Let S (A) be the state space of a unital C -algebra A. satisfying the bound k (A) k k A k : (2.2. Moreover.111) This is because is a morphism cf.109) y . normalization yields c1 + c2 = 1. In the unital case it is clear that S (A) is convex.79).7). respectively. iz 1 . Since (A) is closed by 2. In particular. The next case is A = C C = C 2 . De nition 2. We identify M2 (C ) with its dual through the pairing !(A) = Tr !A.8. and all vi 2 C .6. It follows that the state space of a unital C -algebra is a compact convex set. we parametrize 1 + x y + iz 1 =2 (2.9 Representations and the GNS-construction The material of this section explains how the usual Hilbert space framework of quantum mechanics emerges from the C -algebraic setting. If necessary we add a unit to A (this is justi ed by 2.1 A representation of A on a Hilbert space H is a (complex) linear map : A ! B(H) satisfying (A B ) = (A) (B ) (A ) = (A) (2. The very simplest example is A = C . so that S (A) A . In the non-unital case one arrives at this conclusion most simply via 2. so that a positive functional must have c1 0 and c2 0. x 2. there is an a 2 (A) for which r(A) = jaj. By the Hahn-Banach Theorem 2.3.7. since both positivity and normalization are clearly preserved under convex sums.9. (2. For this a one has j!(A)j = jaj = r(A) =k A k by (2. so that S (A) is compact in the (relative) w -topology by the Banach-Alaoglu theorem.3. it follows that k ! ~ k= 1.2 !a is a state. we see that S (A) is closed in A if the latter is equipped with the w -topology.1.8. . De ne a linear map ! ~a : C A + C I ! C by ! ~a ( A + I) := a + . Finally.6.3. It follows that S (A) consists of all positive 2 2 matrices with Tr = 1 these are the density matrices of quantum mechanics. since I = 1+1.7. To identify S (A) with a familiar compact convex set. A convex set C in a . Since ! ~ a(I) = 1. There is a natural equivalence relation in the set of all representations of A: two representations 1 2 on Hilbert spaces H1 H2 .1. Now consider A = M2 (C ). Hence S (M2 (C )) is the unit ball in R3 . it follows that i pi vi belongs to C when all P pi 0 and i pi = 1. so that each element of (C 2 ) is _ ) = c1 1 + c2 2 . The dual is C 2 as well. )w belongs to C whenever vector space C is a subset of V such that the convex sum v + (1P v w 2 C and 2 0 1]. Since w -limits obviously preserve positivity and normalization. By 2. A representation is automatically continuous. are called equivalent if there exists a unitary isomorphism U : H1 ! H2 such that U 1 (A)U = 2 (A) for all A 2 A. S (A) is a closed subset of the unit ball of A by 2.
Conversely. The cyclic vector is de ned by ! = V I.9. and let p be the projection onto the closure of M . When A = A then (Ap) = pA = (pAp) = pAp = Ap so that A p] = 0.7. etc. by means of (2.115).116) . p A] = 0 for all A 2 M). take a nonzero vector 2 H. where it satis es (2. 4. If V : A ! A=N! is the canonical projection. this form is positive semi-de nite (this means that (A A)0 0 for all A). Given ! 2 S (A). If A 2 M then ApH pH by de nition of p.113) is a closed left-ideal in A. The representation ! (A) is rstly de ned on A=N! H! by ! (A)V B := V AB (2. from any state ! 2 S (A) one can construct a cyclic representation ! on a Hilbert space H! with cyclic vector ! in the following way. hence a positive functional.110). so that ( ! ! (A) ! ) = !(A) 8A 2 A: (2.112) N! = fA 2 A j !(A A) = 0g (2. Its null space (2.2 Any non-degenerate representation is a direct sum of cyclic representations. We restrict ourselves to the unital case the general case follows by adding a unit to A and using 2. Hence ! (A) may be de ned on all of H! by continuous extension of (2.13) this is true for all A 2 M. The proof uses a lemma which appears in many other proofs as well. 2. then any unit vector 2 H de nes a state 2 S (A). Apply this lemma with M = (A) the assumption of non-degeneracy guarantees that p is nonzero. By (2. If is a non-degenerate representation of A on H.3 Let M be a -algebra in B(H). with cyclic vector . and the conclusion implies that A ! p (A) de nes a subrepresentation of A on pH.115) it follows that ! is continuous. one says that a representation is non-degenerate if 0 is the only vector annihilated by all representatives of A. This process may be repeated on p? H. Proposition 2.9. then by de nition (V A V B )! := (A B )0 : The Hilbert space H! is the closure of A=N! in this inner product. Lemma 2. This subrepresentation is clearly cyclic. p this reads Ap = pAp. To exclude this possibility.114) (2. 3.8. such trivial may occur as a summand. Construction 2.38 2 ELEMENTARY THEORY OF C -ALGEBRAS The map (A) = 0 for all A 2 A is a representation more generally. referred to as a vector state relative to . A representation is called cyclic if its carrier space H contains a cyclic vector for this means that the closure of (A) (which in any case is a closed subspace of H) coincides with H. de ne the sesquilinear form ( )0 on A by (A B )0 := !(A B ): Since ! is a state. Hence p? Ap = 0 with p? = I . The form ( )0 projects to an inner product ( )! on the quotient A=N! .9.103). Then p 2 M0 (that is.4 1.
110) on the dense subspace A=N! of H! . then 1 1 1 1 2 each contain a (A) 1 ) = ( 2 2 (A) 2 ) =: !2 (A) (A) and 2 (A) are equivalent.10 The Gel'fand-Neumark theorem 39 We now prove the various claims made here. 2. The GNS-construction leads to a simple proof this theorem.2 The universal representation u of a C -algebra A is the direct sum of all its GNS-representations ! . The equality (2. for ! (A) ! = 0 implies (A) = 0 by the GNS-construction. we compute k ! (A) k2 for = V B . For later use we mention that the GNS-construction yields ( ! (A) ! ! (B ) ! ) = !(A B ): Putting B = A yields (2. !1 and !2 are induced by the same state.118) (2. and 2 is equivalent to !2 . upon which k ! (A) k k A k follows from (2. By 2. for some Hilbert space H. because it implies that the map (A) : A ! A de ned in (2.116) that U is unitary as a map from H! to U H! .119) (2. By (2. one has k ! (A) k2 = !(B AA B ).5 the representation 1 is equivalent to the GNS-representation !1 . it is easily checked that ! is a morphism as well.10 The Gel'fand-Neumark theorem One of the main results in the theory of C -algebras is Theorem 2.2. so that k ! (A) k k A k k k. which is closed because of the continuity of !.9.72) quotients well to a map from A=N! to A=N! the latter map is ! de ned in (2. where A B 2 A. so they must coincide. satisfying (2. 2 2 H2 .1 A C -algebra is isomorphic to a subalgebra of B(H). It is trivial to verify that U intertwines ! and . But !(B B ) =k k2.104).3). H2 of two cyclic representations cyclic vector 1 2 H1 .114) and step 2 above.9. De nition 2. Proposition 2. This is important. Since is a morphism. which uses the following notion. To prove that ! is continuous on A=N! . but since is cyclic for the image of U is H. Hence U is unitary. and !1 (A) := ( for all A 2 A. .117) implies that N! is a left-ideal. On the other hand. Corollary 2.116) and the fact that ! is a representation.5 If ( (A) H) is cyclic then the GNS-representation ( ! (A) H! ) de ned by any vector state (corresponding to a cyclic unit vector 2 H) is unitarily equivalent to ( (A) H).10.91) and the positivity of ! one has !(B AA B ) k A k2 !(B B ). By (2.115). It follows from (2.9.120) k ! (A) ! k2 = !(A A) which may alternatively be derived from (2. ! 2 S (A) hence it is de ned on the Hilbert space Hu = !2S (A) H! .6 If the Hilbert spaces H1.10. This is very simple to prove: the operator U : H! ! H implementing the equivalence is initially de ned on the dense subspace ! (A) ! by U ! (A) ! = (A) this operator is well-de ned. First note that the null space N! of ( )0 can be de ned in two equivalent ways N! := fA 2 A j (A A)0 = 0g = fA 2 A j (A B )0 = 0 8 B 2 Ag: (2.117) The equivalence follows from the Cauchy-Schwarz inequality (2.
We have seen that a positive map Q (cf. Being injective. where each vi 2 H. When is a faithful representation of A (which exists by Theorem 2. The norm k M k of M 2 Mn (A) is then simply de ned to be the norm of n (M ). De nition 2. in which the involution in A replaces the usual complex conjugation in C . which is positive in Mn (B).7. Another consequence of the GNS-construction.11. any representation of A on H is a completely positive map from A to B(H). which is of central importance for quantization theory.120) this means ! !(A A) = 0 for all states !.e. and that Q is normalized. is Corollary 2. then '(A ) = '(B ) '(B ). The norm is unique by Corollary 2.6.10. The elements of Mn (A) are P n n matrices with entries in A multiplication is done in the usual way. By Theorem 2. seen as map not from H to H but as a map from H to H is unitary. of course. is positive. In particular. with p := WW (the target projection of W on H ).16). Since n (Mn (A)) is a closed algebra in B(H C n ) (because n < 1). Hence ! (A) ! = 0. In particular.8.11. de ned by linear extension of n (M )vi := (Mij )vj we here look at elements of H C n as n-tuples (v1 : : : vn ). To prove that is injective. we get an interesting generalization of the GNS-construction. For example. We rst introduce the C -algebra Mn (A) for a given C -algebra A and n 2 N . and nally A = 0 by the de nition of a norm. so that this procedure does not depend on the choice of . suppose that u (A) = 0 for some A 2 A.1).10.1 then follows by taking H = Hu the desired isomorphism is u . While the universal representation leads to a nice proof of 2.10. Theorem 2. De nition 2. There exists a Hilbert space H .40 2 ELEMENTARY THEORY OF C -ALGEBRAS Theorem 2. we may assume that B is faithfully represented as a subalgebra B ' (B) B(H ). By de nition of a direct sum.3 An operator A 2 A is positive (that is. one obtains a faithful realization n of Mn (A) on H C n .10. A 2 A+ (A) 0 for all cyclic R) i representations . the morphism u is isometric by Lemma 2.5. so that U H ! H de ned as W .1. de ned by (Qn (M ))ij := Q(Mij ). since when A = B B in Mn (A). One may identify Mn (A) with A Mn (C ) in the obvious way. hence k ! (A) ! k2= 0 by (2. for some Hilbert space H . it is obvious that Mn (A) is a C -algebra in this norm. It turns out that for this purpose one needs to impose a further condition on Q. or rather of 2. which implies k A A k= 0 by Lemma 2. We would like to see if one can generalize the GNS-construction. in that the C in ! : A ! C is replaced by a general C -algebra B in Q : A ! B.2. If we also assume that A and B are unital. so that k A k= 0 by (2. For example.2 Let Q : A ! B be a completely positive map between C -algebras with unit.8. this implies u ( A ) = 0 for all states !. i. a morphism ' is a completely positive map. and U : ~ ~ .1 A linear map Q : A ! B between C -algebras is called completely positive if for all n 2 N the map Qn : Mn (A) ! Mn (B).122) 2.4 generalizes the notion of a state.1 = p (A)p: (2. the best faithful representation of B(H) is simply its de ning one.11 Complete positivity .8. given by (M )ij = Mji . This generalization will appear as the proof of the following Stinespring theorem. such that Q(I) = I. H H . a representation of A on H . one has U (Q(A))U . with the di erence that one now multiplies elements of A rather than complex numbers. the order has to be taken into account. such that (Q(A)) = W (A)W 8A 2 A: (2.121) ~ := pH Equivalently.3.1. (MN )ij := k Mik Nkj .10. the Hilbert space Hu is absurdly large in practical examples a better way of obtaining a faithful representation always exists. The involution in Mn (A) is. and a partial isometry W : H ! H (with W W = I).10.
91) in Mn (A). Its adjoint W : H ! H is given by (continuous extension of) W V A v = (Q(A))v from which the properties W W = I and (2. Repeating the above argument with A and replaced by B and . 3. This is a linear map W : H1 ! H2 between two Hilbert spaces.126) so that (A) may be de ned on all of H by continuous extension of (2. then by de nition N.121) follow. with inner product (v w) . One has the bound k (A) k k A k (2.126) one uses (2. We denote elements of H by v w.11 Complete positivity 41 The proof consists of a modi cation of the GNS-construction.125). Namely. Looking in a faithful representation n as explained above. with the property that H1 contains a closed subspace K1 such that (W W )2 = ( )1 for all 2 K1 .125) satis es (A ) = (A) .127) (2. de ned by i i i its matrix elements B ij := Q(Ai Aj ). 4. this form is positive semi-de nite denote its null space by (V (A v) V (B w)) := (A v B w)0 : The Hilbert space H is the closure of A H =N in this inner product.128) is a partial isometry. To show that the form de ned by (2. This extension Wv := V I (A)N v (2.124) (A)V (B w) := V (AB w) this is well-de ned. respectively. de ned by N . The map W : H ! H .123) is positive. we write X ij (Ai vi Aj vj )0 = X ij (vi (Q(Ai Aj ))vj ) : (2. Hence W is unitary from K1 to W K1 . De ne the sesquilinear form ( )0 on A H (algebraic tensor product) by (sesqui-)linear extension of (A v B w)0 := (v (Q(A B ))w) : (2. is positive in Mn (B).123) 2. whose rst row is (B1 : : : Bn ).129) is positive. one sees that (z A z ) = X ij (zi (Ai Aj )zj ) = X ij ( (Ai )zi (Aj )zj ) =k Az k2 0 0. for arbitrary A B1 : : : Bn 2 A we conjugate the inequality 0 A AIn k A k2 In with the matrix B .3 1. Construction 2. If V : A H ! A H =N is the canonical projection. Since Q is completely positive. Since Q is completely positive.129) Now consider the element A of Mn (A) with matrix elements A ij = Ai Aj . To prove (2. The representation (A) is initially de ned on A H =N by linear extension of (2. and W = 0 ? . Hence A .2. It uses the notion of a partial isometry. one concludes that the right-hand side of (2.11. It follows that WW = K2 ] and W W = K1 ] are on K1 projections onto the image and the kernel of W . it must be that B . P where Az = A z . because (2. The form ( )0 projects to an inner product ( ) on A H =N . respectively.
121).1 we may assume that A = C (X ) for some locally compact Hausdor space X . i. where F (x) = i fi (x)Mi for fi 2 C (X ) and Mi 2 Mn (C ).. Since Q is completely positive. When the extension exists but does not preserve the unit. The relevance of Stinespring's theorem for quantum mechanics stems from the following result. (w W all w 2 H and 2 H . 3. 2. W W = I for the projection onto the subspace of H on which W is isometric this subspace is H itself Hence (2. Hence in any representation (B) and any vector (v1 : : : vn ) 2 H C n one has X ij (vi (Q(Bi A ABj ))vj ) k A k2 X ij (vi (Q(Bi Bj ))vj ): (2. .11. are dense in C (X Mn (C )).e.1 = W (Q(A))W = WW (A)WW = p (A)p: When Q fails to preserve the unit. To check (2. The proof then proceeds in the following steps: P 1. when G 0 there is a sequence Fk 0 such that limk Fk = G. and all other entries zero. Thus it is no longer possible to regard H as a subspace of H . but W is no longer a partial isometry one rather has k W k2=k Q(I) k. Such F is positive i all fi and Mi are positive. one is in the situation of the previous paragraph. We may then identify Mn (C (X )) with C (X Mn (C )). use the de nition to compute (Wv Ww) = (V vV I w) = (I v w ) 0 = (v w ) ) = (Ww ) for where we used (2. This leads to 0 B A AB k A k2 B B .4. one has p = WW for the projection p onto the image of W . Positive elements G of C (X Mn (C )) can be norm-approximated by positive F 's.130) one then has P k (A) k2 = k A k2 X ij (ABi vi ABj vj )0 = X ij (vi (Q(Bi A ABj ))vj ) (Bi vi Bj vj )0 X ij (vi =k A k2 (V X i (Q(Bi Bj ))vj ) =k A k2 X ij Bi vi V I X j Bj vj ) =k A k2 k k2 : I To show that W is a partial isometry. from (2.121). Proposition 2.128). we use (2. since U (Q(A))U . By Theorem 2.122) follows from (2.4 Let A be a commutative unital C -algebra. and. one merely uses the de nition of the adjoint. To verify (2.123) and Q(I) = I.42 2 ELEMENTARY THEORY OF C -ALGEBRAS and which has zeros everywhere else the adjoint B is then the matrix whose rst column is (B1 : : : Bn )T . one has Qn (B A AB ) k A k2 Qn (B B ). and (2. Then any positive map Q : A ! is completely positive. in this case.127) and (2.128) to compute W (A)Wv = W (A)V (I v) = W V (A v) = (Q(A))v: Being a partial isometry. the above construction still applies. and the sum is nite. If A and perhaps B are non-unital the theorem holds if Q can be extended (as a positive map) to the unitization of A. (2. This trivially veri ed. such that the extension preserves the unit I (perhaps relative to the unitization of B). viz.125). Elements of the form F .130) With = i V Bi vi .123).
When G in item 1 is positive then each G(xi ) is positive. In particular.8. A norm-limit A = limn An of positive elements in a C -algebra is positive.2. because by (2. and lim Bn = B exist because of (2. G(x) k < : x2X 2. G k= sup k Fl (x) . 43 6. Hence k Fl . A partition of unity subordinate to the given cover is a collection of continuous positive P functions i 2 C (X ). P Since Q is positive. Finally. which by the compactness of X has a nite subcover fOx1 : : : Oxl g. If Fk ! G 0 in C (X Mn (C )) then Q(G) = limk Q(Fk ) is a norm-limit of positive elements.88) we have An = Bn Bn . P 2. The intrinsic de nition of this boundary is as follows.12 Pure states and irreducible representations 4. This gives an open cover of X .131) Since k G(xi ) . i. We know from 2.e. A = B B by continuity of multiplication. .15).12 Pure states and irreducible representations We return to the discussion at the end of 2.8. G(x) k=k 'i (x)(G(xi ) . Now de ne Fl 2 C (X Mn (C )) by Fl (x) := l X i=1 l X i=1 'i (x)G(xi ): l X i=1 l X i=1 (2. G(y) k < g is open for each x 2 X . We now prove each of these claims. On F as speci ed in 2. Since G is continuous.16). 4. where i = 1 : : : l. Qn is continuous. An element F 2 C (X Mn (C )) is positive i F (x) is positive in Mn (C ) for each x 2 X . it follows that Qn maps each positive element of the form F = i fi Mi into a positive member of Mn (B). Take G 2 C (X Mn (C )) and pick > 0. By 2. 6. when F (x) = f (x)M for some f 2 C (X ) and M 2 Mn (C ) then F is positive i f is positive in C (X ) and M is positive in Mn (C ). G(x) k < for all x 2 Oxi .5 that Q is continuous the continuity of Qn follows because n < 1.. Such a partition of unity exists. such that the support of 'i lies in Oxi and li=1 'i (x) = 1 for all x 2 X . 5. Qn (F ) is positive when F is positive.2 we infer that F de ned by P F (x) = i fi (x)Mi is positive when all fi and Mi are positive. 3. 5. the set Ox := fy 2 X k G(x) . hence is positive.6. G(x) k < 'i (x) = : Here the norm is the matrix norm in Mn (C ). by (2.2. One sees that the compact convex sets in the examples have a natural boundary. as we have just seen.1 one has Qn (F ) = i Q(fi ) Mi . one has k Fl (x) .4. G(x)) k 'i (x) k G(xi ) . 1.11. Now each operator Bi M is positive in Mn (B) when Bi and M are positive (as can be checked in a faithful representation).
When there exists a vector 2 H for which (A) is not dense in H. we will prove in 2. Hence when (A)0 is nontrivial it contains a nontrivial projection p. Proposition 2. If p ! = 0 then Ap ! = pA ! = 0 for all A 2 A. Hence \ (A)0 = C I ) irreducible".1 An extreme point in a convex set K (in some vector space) is a member ! ! = !1 + (1 . there is a nontrivial projection p 2 ! (A)0 by Schur's lemma. so that cannot be irreducible by Schur's lemma.3 Each of the following conditions is equivalent to the irreducibility of (A) on H: 1. We may now ask what happens to the GNS-construction when the state ! one constructs the representation ! from is pure.12. but we will not prove this.4 The GNS-representation ! (A) of a state ! 2 S (A) is irreducible i ! is pure. we can form the projection onto the closure of (A) . Theorem 2.3 below and the spectral theorem. By Lemma 2. But then pH is stable under (A). where 0 is identi ed with _ to .12. Hence the pure states on A = C C are the points 0 and 1 in 0 1]. It is a deep fact of C -algebras that the quali er `closed' may be omitted from this de nition. The rst part of De nition 2. for @e K .18.104.22.168. Using Theorem 2. or simply P . whereas 1 maps it to .44 of K which can only be decomposed as 2 ELEMENTARY THEORY OF C -ALGEBRAS De nition 2. with M = (A). The commutant (A)0 is a -algebra in B(H). An extreme point in the state space K = S (A) of a C -algebra A is called a pure state. (A)00 = B(H) (Schur's lemma) 2. this projection lies in (A)0 . Every vector in H is cyclic for (A) (recall that this means that (A) is dense in H).13.132) 2 (0 1). the de ning representation d of the matrix algebra MN on C N is irreducible.14.2 A representation of a C -algebra A on a Hilbert space H is called irreducible if a closed subspace of H which is stable under (A) is either H or 0. We are now in a position to answer the question posed before 2. Let ! be the cyclic vector for ! . so when it is nontrivial it must contain a selfadjoint element A which is not a multiple of I. it can be shown that the projections in the spectral resolution of A lie in (A)0 if A does.10 that the state space of Mn (C ) consists of all positive matrices with unit trace the pure state space of Mn (C ) then consists of all one-dimensional projections.1 is due to Minkowski it was von Neumann who recognized that this de nition is applicable to quantum mechanics. This de nition should be familiar from the theory of group representations. In preparation: De nition 2.9. contradicting irreducibility. Hence \ irreducible ) (A)0 = C I". A state that is not pure is called a mixed state. When ! is pure yet ! (A) reducible. When K = S (A) is a state space of a C -algebra we write P (A).109) for which x2 + y2 + z 2 = 1 these are the projections onto one-dimensional subspaces of C 2 . if !1 = !2 = !. )!2 (2. The commutant of (A) in B(H) is f I j 2 C g in other words.12. The converse is trivial. when (A)0 = C I and is reducible one nds a contradiction because the projection onto the alleged nontrivial stable subspace of H commutes with (A). The pure states on A = M2 (C ) are the the functional mapping + matrices in (2.2. . More generally. Conversely. In the in nite-dimensional case. This precisely reproduces the notion of a pure state in quantum mechanics. Hence \ irreducible ) every vector cyclic". referred to as the pure state space of A. The collection @e K of extreme points in K is called the extreme boundary of K . the de ning representations d of B(H) on H is irreducible as well. Clearly.
132) for !1 !2 2 S (A) and 2 0 1].119) show that !1 is proportional to !. suppose ! is irreducible. one has ^( ) = Q ^( ) Q (2.12.105) with A ! A B and ! ! !1 .133) in other words. Q ^ Furthermore.137) ( ! (A) ! Q ! (B ) ! ) = !1 (A B ): 0 It is the immediate from (2.133) and (2. in that (2. De ne Q by Q = . A2 )) = 0 by (2. From 2.120). with (2. whence k Q k C .12.110) that Q ! (C )] = 0 for all C 2 A. A2 ) B )j2 0 jQ ^ ( ! (A1 ) ! ! (B ) ! ) = Q ^ ( ! (A2 ) ! ! (B ) ! ). ^ on a Hilbert space H be bounded.137). Continuing with the proof of 2. so that ^ ( ! (A1 ) ! ! (B ) ! ) . so that by the Riesz-Fischer Hold xed. for all 2 ! (A) ! .9. Q ^ ( ! (A2 ) ! ! (B ) ! )j2 = j !1 ((A1 . ^ ( ) is then bounded by (2. By (2. by (2. Hence ! cannot be pure. ! = (1 .12.4 yields Corollary 2.6 If ( (A) H) is irreducible then the GNS-representation ( ! (A) H! ) de ned by any vector state (corresponding to a unit vector 2 H) is unitarily equivalent to ( (A) H). Combining this with 2. ) ? .e.12 Pure states and irreducible representations 45 so that p = 0 as ! is cyclic.134) This is well de ned: when ! (A1 ) ! = ! (A2 ) ! then !((A1 . This proves \pure ) irreducible".104) this yields j !1 (A B )j2 2 !1 (A A)!1 (B B ) !(A A)!(B B ) (2. (2. )!2. and k Q k C . Then !1 . a sesquilinear map) Q ^ ( ! (A) ! ! (B ) ! ) := !1 (A B ): Q (2. In the opposite direction. with C = 1.4. Similarly.. When (2.2. where and ? are states de ned as in (2. The map ! Q ^ ( ) = ( ).103).135) holds Lemma 2.5 Let a quadratic form Q for all 2 H. There is a bounded operator Q on H such that ^ ( ) = ( Q ) for all Q 2 H. It follows that Q Moreover.136) by (2. The theorem there exists a unique vector such that Q self-adjointness of Q in case that (2. ? := p? ! = k p? ! k. Now use (2. and some constant C 0.7 Every irreducible representation of a C -algebra comes from a pure state via the GNS-construction.136) is satis ed Q is self-adjoint.5 we have the Corollary 2. which is positive hence !1 (A A) !(A A) for all A 2 A. and (2. p? ! = 0 is impossible. and therefore equal to ! by normalization. We may then decompose ! = + (1 .135).135) ^ can be extended to all of H! by continuity. we see that there is a self-adjoint operator Q on H! such that (2.120) imply that Q is bounded in that ^ ( )j C k k k k jQ (2. This allows us to de ne a quadratic form (i.133) ^ on ! (A) ! by for all A B .12. and =k p? ! k2 . (2.12. so that ! is pure. with := p ! = k p ! k. Similarly for B .134).136) holds is obvious. A2 ) (A1 . . Hence Q 2 ! (A) since ! is irreducible one must have Q = tI for some t 2 R hence (2.135) to estimate ^ (Q ) C k Q k k k 2 k Q k2= (Q Q ) = Q taking the supremum over all in the unit ball yields k Q k2 C k Q k2 .
For = 0 or = ! the claim is obvious.8. hence k ! . the proof of 2. then 0 < (I) < 1. K = co(@e K ). since ! .3).12. Since ! is pure. applied to C (X ) it shows that arbitrary probability measures on X may be approximated by nite convex sums of point (Dirac) measures.12.8 as follows Theorem 2. (I ) = 1 . Hence !g = t! for some t 2 R+ by 2.8. and 0 1 .4. then fg 2 ker(!) for all g 2 C (X ). The convex hull co(V ) of a subset V of a vector space is de ned by co(V ) := f v + (1 . and . Fortunately. It could be that no pure states exist in S (A) think of an open convex cone.54) coincides with the extension !I of a state de ned in (2. )w j v w 2 V 2 0 1]g: (2.12. + (1 .3 and Theorems 2. the extension of state on C0 (X ) to a state on C (X a multiplicative functional de ned in (2.46 2 ELEMENTARY THEORY OF C -ALGEBRAS A useful reformulation of the notion of a pure state is as follows. it follows from 2. use 2.3. Hence (I) would imply ! = . whereas (I) = 0 implies = 0. But ker(!x) is a maximal ideal. which is maximal because the kernel of a functional on any vector space has codimension 1.8. .3. In other words. and ker( ) is an ideal. Let !x 2 (C (X )) (cf.7 remains pure. (I)) and = (I) are states.8 that !x is pure. this possibility is excluded by the Krein-Milman theorem in functional analysis. and the functional 1 in (2. ker(!g ) = ker(!).132) holds then 0 !1 ! (cf. 2. (I) (I) with = 1 .11 For every A 2 AR and a 2 (A) there is a pure state !a on A for which !a(A) = a.10 A compact convex set K embedded in a locally convex vector space is the closure of the convex hull of its extreme points. (I). )=(1 . k= !(I) . Since two functionals on any vector space are proportional when they have the same kernel.138) Theorem 2. Conversely. g 1X . In view of Proposition 2. is positive.6 and 2. so that !1 = t! by assumption normalization gives t = . Proposition 2. and ! is pure.7. it guarantees that a C -algebra has lots of pure states. the proof of 2. It follows that arbitrary states on a C -algebra may be approximated by nite convex sums of pure states. It would follow that such a C -algebra has no irreducible representations.12.108). g)). Hence (! . This is a spectacular result: for example. with 0 6= 6= !. one has 0 !g !.12. In general. The case that X is not compact may be reduced to the compact case by passing from A = C0 (X ) ~ ) cf. There exists a pure state ! such that j!(A)j =k A k.4. so when 6= 0 it must be that ker(!x ) = ker( ). if (2. We assume that A is unital if not. When ! is pure and 0 !. since any function is a linear combination of functions for which 0 g 1X . Hence ker(!) is an ideal. hence !1 = ! = !2 .8. Hence ! is multiplicative by Theorem 2. and pick a g 2 C (X ) with 0 g 1X .4). (I). and suppose a functional satis es 0 !x .6 and 2. De ne a functional !g on C (X ) by !g (f ) := !(fg). This is possible because the unique extension of a pure to AI = C (X ~ ) guaranteed by 22.214.171.124. Thus we put A = C (X ). contrary to assumption.9 The pure state space of the commutative C -algebra C0(X ) (equipped with the relative w -topology) is homeomorphic to X . Conversely. Then ker(!x) ker( ). we merely need to establish a bijective correspondence between the pure states and the multiplicative functionals on C0 (X ). The simplest application of this proposition is Theorem 2.1 and 2.4.8 A state is pure i 0 ! for a positive functional implies = t! for some t 2 R+ . It follows that when f 2 ker(!).4.12.7 etc. which we state without proof.3. ) ! = 1! . let ! be a pure state.8. Since !(f ) . For example. Moreover. In particular. by 2. we may now re ne Lemma 2.1 we have = (I)!.61) clearly de nes a pure state. !g (f ) = !(f (1 .4.
Let us refer to the ensuing set of pure states as P (A)].9 that this extension is pure. as will follow from this section.14. that there are many other pure states whose GNS-representation is not equivalent to the de ning representation . For one thing.8. when A 2 B(H) and a 2 (A).6 that any vector state on B(H) de nes an irreducible representation of B(H) which is equivalent to the de ning representation. Namely.12. and taking only one pure state in each equivalence class.132). it must be that (A) is isomorphic to an algebra acting on a nite-dimensional vector space. On the other hand. we know from 2. If !a were not an extreme point in S (A).12. where X is connected). but a is not in the discrete spectrum of A as an operator on H (i.12. has only one irreducible representation up to equivalence).3.8 to C (A I) by multiplicativity and continuity. For it is easy to show from (2. B(H) has a huge number of inequivalent representations even when H is separable. it follows from 2.116) that !a 2 H!a is an eigenvector of !a (A) with eigenvalue a. one has the isomorphism '(A) ' A= ker(').8. in that case. We then have A ' r (A) := !2 P (A)] ! (A): (2.1 implies that (A) must be a matrix algebra (as B(H) is the algebra of n n matrices for H = C n ). In other words. Since A= ker( ) is nite-dimensional. One easily checks that the set of all extensions of ! ~ a to A (which extensions we know to be states see the proof of 2.e.1 Let H be a Hilbert space.139.10 it has at least one extreme point !a. But it is clear that.12. Furthermore.10.12 Every nite-dimensional C -algebra is a direct sum of matrix algebras.12. concluding that the universal representation u may be replaced by r := !2P (A) ! . upon which 2. The -algebra Bf (H) of nite-rank operators on H is the ( nite) linear span of all nite-dimensional projections on H. De nition 2. For any morphism '. By the Krein-Milman theorem 2.13 The C -algebra of compact operators .11 and the existence of bounded self-adjoint operators with continuous spectrum (such as any multiplication operator on L2 (X ).10. In other words. It follows from 2.12.8 by 2.13 The C -algebra of compact operators 47 We extend the state in the proof of 2.13. For example. In non-commutative geometry elements of this C -algebra play the role of in nitesimals in general. it would be decomposable as in (2.139) It is obvious that the proof of 2. Another argument against B(H) is that it is non-separable in the nom-topology even when H is separable. It would appear that the appropriate generalization of the C -algebra Mn (C ) of n n matrices to in nite-dimensional Hilbert spaces H is the C -algebra B(H) of all bounded operators on H. an operator A 2 B(H) lies in Bf (H) when AH := fA j 2 Hg is nite-dimensional.8. so that !a cannot be an extreme point of Ka . which excludes the possibility that (A) and !a (A) are equivalent (as the spectrum is invariant under unitary transformations).11 in the proof of the Gel'fand-Neumark Theorem 2. The simplest application of this re nement is Proposition 2.12.. hence certainly for any representation ' = . a is in the continuous spectrum of A = (A) but in the discrete spectrum of !a (A). that is. The appropriate generalization of Mn (C ) to an in nite-dimensional Hilbert space H turns out to be the C -algebra B0 (H) of compact operator on H. 2. Then apply the isomorphism 2. We may further restrict this direct sum by de ning two states to be equivalent if the corresponding GNS-representations are equivalent. B0 (H) is a basic building block in the theory of C -algebras. most of these are realized on non-separable Hilbert spaces. unlike Mn (C ) (which. then !a cannot be equivalent to . !1 and !2 would coincide on C (A I).2.8) is a closed convex subset Ka of S (A) hence it is a compact convex set. there is no eigenvector a 2 H for which A a = a a ).1.3 below that (A)00 = (A) in every nite-dimensional representation of A. We may now replace the use of 2. This section is devoted to an exhaustive study of this C algebra. we put ! ~a (An ) = an etc.1 still goes through. This is not the case. it follows from Theorem 2.
Now if An ! A then An B ! AB and BAn ! BA by continuity of multiplication in B(H). I) k 1. hence k An . Thirdly. Then k A n .e. The C -algebra B0 (H) is an ideal in B(H). . Hence supk k=1 k (An . with equality i A A is proportional to . ) k 1 3 +1 3 + k Af k k p( n .13.3). In particular. Since k k2 = ( ) = lim ( n n) k kk n k= one has k k 1. )) . and the second goes to zero by de nition of weak convergence. the norm in B0 (H) is the operator norm (2. it follows that AB1 is compact.3) in B(H) (hence in B0 (H)). save for the fact that it is not norm-closed (unless H has nite dimension). ( . . the nite-dimensional projection onto the image of Af . they coincide on pH. A k=k (A . n A A )j: The rst term goes to zero by the proof of 2. n A A )j k A A( n . ) k : Since the weak and the norm topology on a nite-dimensional Hilbert space coincide. choose Af 2 Bf (H) such that k A . fA ( )j = j( n A A( n . where a2 =k A k2 .48 2 ELEMENTARY THEORY OF C -ALGEBRAS The C -algebra B0 (H) of compact operators on H is the norm-closure of Bf (H) in B(H) in other words. Hence when A = A the property k A k2 =k A a k2 with k a k= 1 implies 2 A2 a = a a . Now the CauchySchwarz inequality with = 1 gives k A k2 = ( A A ) k A A k. 2. and Bf (H) is a -algebra. Hence B0 (H) is an ideal by virtue of its de nition. Hence fA is continuous. But since BA = (A B ) . fA assumes its maximum at some a. Firstly. then jfA( n ) . I) k= 1. when A 2 Bf (H) and B 2 B(H) then AB 2 Bf (H). so that k (An . then A is continuous from H with the weak topology to H with the norm-topology since compactness is preserved under continuous maps. Indeed. Af ) n + (A . When n ! weakly with k n k= 1. De ne fA : B1 ! R by fA ( ) :=k A k2 . ) k +j( . 3. This maximum is k A k2 by (2. I) = . Proposition 2. Hence Bf (H) is an ideal in B(H).3). let n ! in the weak topology.13. for any sequence (or net) An 2 Bf (H) we may choose a unit vector n 2 (An H)? .2 1. it is the smallest C -algebra of B(H) containing Bf (H). Secondly.3 (noting that A A 2 B0 (H)). It is clear that Bf (H) is a -algebra. i.. and put p := Af H]. If A 2 B0 (H) then AB1 is compact in H (with the norm-topology). Given > 0. Hence the unit ball B1 is compact in the weak topology by the Banach-Alaoglu theorem. ) k < =3 for all n > N . note that the weak topology on H (in which n ! i ( n ) ! ( ) for all 2 H) is actually the w -topology under the duality of H with itself given by the Riesz-Fischer theorem.3 A self-adjoint operator A 2 B0 (H) has an eigenvector a with eigenvalue a such that jaj =k A k. with k n k= 1 for all n. So if we can show that A 2 B0 (H) maps weakly convergent sequences to norm-convergent sequences. since p = p for any projection p. An operator A 2 B(H) lies in B0 (H) when it can be approximated in norm by nite-rank operators. The third item in the next proposition explains the use of the word `compact' in the present context. A k < . Then (An . the set of all 2 H with k k 1.13. The unit operator I lies in B0 (H) i H is nite-dimensional. Af ) + Af ( n . Corollary 2. Since B1 is weakly compact.2. I k! 0 is impossible by de nition of the norm (2. Hence k A n . The spectral theorem or the continuous functional calculus with p f (A2 ) = A2 = A implies A a = a a. Clearly jaj =k A k. Af k < =3. since AB H = AH. so that we can nd N such that k p( n . Here B1 is the unit ball in H. one has A B 2 Bf (H) and hence BA 2 Bf (H).
9. which implies that all vectors in p? H are eigenvectors with eigenvalue zero.2.13.13. By 2. In other words. This proves \A compact and self-adjoint ) A diagonalizable". Let A 2 B0 (H) be self-adjoint. Then limi!1 ( i ) = 0 for all 2 H.3. This contradicts the de nition of p? H unless p? H = 0. which must lie in p? H. so that ( )= X i j( i )j2 (2. and can only have 0 as a possible accumulation point.142) where (A A) 4 is de ned by the continuous functional calculus.3 there is a largest eigenvalue. Let now A be self-adjoint and diagonalizable. and compact by 2. so that (pA) = pA.2. Let A be compact and self-adjoint.13. N X i=1 ai i ]) k2 =k 1 X i=N +1 ai ( i ) i k2 1 X Using (2.13. To show that (2.2.3 the compact self-adjoint operator p? A has an eigenvector.141) is independent Pof the basis. hence diagonalizable. This ordering is possible because by 2. Normalize the eigenvectors i := ai to unit length. Hence i=1 i i \A self-adjoint and diagonalizable with limi!1 jai j = 0 ) A compact". Hence p?A = (I. so that limN !1 k A. Hence i ! 0 weakly.13 The C -algebra of compact operators 49 Theorem 2. The trace class B1 (H) consists of all A 2 B(H) for which k A k2 < 1.140). which by 2. since the i form a basis.4 A self-adjoint operator A 2 B(H) is compact i A = Pi ai i ] (norm-convergent sum). Ordering the eigenvalues so that ai aj when i > j .2. p)A is self-adjoint. except for the fact that they are not closed. We now wish to compute the state space of B0 (H). because limN !1 jaN j = 0. The Hilbert-Schmidt class B2 (H) consists of all A 2 B(H) for which k A k2 < 1.1 proves the claim about the multiplicity of the eigenvalues. N i=1 i i=N +1 jai j2 j( i )j2 jaN j2 1 X i=N +1 j( i )j2 : k A k2 2 := X i k Aei k2 (2. this is P jaN j2 ( i ] k= 0. the set of eigenvalues is discrete. so k A i k= jai j ! 0 by (the proof of) 2.13. This proves \A compact and self-adjoint ) A diagonalizable with limi!1 jai j = 0". Hence k p? A k= 0 2.3 one sees that A p] = 0.13. Aince I = i ei ] we then have k A k2 2 := X ij (ej ui )(ui A Aej ) = X ij (A Aui ej )(ej ui ) = X i k Aui k2 : . we take a second P basis fuigi.13. By assumption this eigenvector can only be zero. For N < 1 and 2 H one then has k (A .5 The Hilbert-Schmidt norm k A k2 of A 2 B(H) is de ned by P a ). this proves that A is compact.140) which clearly converges. De nition 2. which implies that A restricted to p? H is zero. As in Lemma 2.2. Since the operator N a ] is clearly of nite rank. This involves the study of a number of subspaces of B(H) which are not C -algebras.3. Hence limi!1 jai j = 0. but which are ideals of B(H). one has limi!1 jai j = 0. when A is compact its restriction to any closed subspace of H is compact. and let p be the projection onto the closure of the linear span of all eigenvectors of A. with limi!1 jai j = 0.13. with corresponding resolution of the identity I = i ui ] (weakly). and must therefore be an eigenvector of A in p? H. where each eigenvalue ai has nite multiplicity.141) where fei gi is an arbitrary basis of H the right-hand side is independent of the choice of the basis. The trace norm k A k1 of A 2 B(H) is de ned by 4 k2 k A k1 :=k (A A) 1 2 1 (2. Finally.
143).153) and that U U is the projection onto the closure of ran(jAj). which does not depend on the basis.143) by manipulations similar to those establishing the basis-independence of (2. In particular.145) where p jAj := A A: (2. whether or not A 2 B1 (H).16). In any case.141).151) To prove this. where Tr + is de ned in terms of the decomposition (2.13. and note that 4 k2 =k A k by (2. it may happen that Tr A depends on the basis it may even be nite in one basis and in nite in another). Conversely. and will not be needed.150). for every > 0 1 there is a 2 H of norm 1 such that k B k2 k B k2 + . + iTr A00 + . the linearity property Tr (A + B ) = Tr A + Tr B (2.148) for all positive A 2 B(H) and all unitaries U . One always has the equalities k A k1 = Tr jAj (2. One easily veri es that jAj = U A (2. The same trick with k A k k A k + establishes (2.149) for all A B 2 B1 (H) is immediate from (2. Completing k (A A ) 1 to a basis fei gi . when A 0 one simply has k A k1 = Tr A. it can be shown that A 2 B1 (H) when Tr + A < 1. U is well de ned. . Put B = (A A) 4 . follow from (2. The properties Tr A A = Tr AA (2. and Tr UAU = Tr A (2. and put U = 0 on ran(jAj)? . For A 2 B1 (H) one has Tr + A = Tr A. Then compute (U jAj U jAj ) = (A A ) = ( A A ) = ( jAj2 ) = (jAj jAj ): Hence U is an isometry on ran(jAj). we have 4 k2 4 k A k=k (A A) 1 k (A A) 1 k2 + X i 4 ei k2 + =k A k1 + : k (A A) 1 Letting ! 0 then proves (2.151).6 Every operator A 2 B(H) has a polar decomposition A = U jAj (2. we use our old trick: although k B k k B k for all unit vectors . Tr A0. The following decomposition will often be used. . Then extend U to the closure of ran(jAj) by continuity. and that B2 (H) is complete in this norm.146)) and U is a partial isometry with the same kernel as A.144) 2 k A k2 = Tr jAj = Tr A A (2. It is easy to see that the Hilbert-Schmidt norm is indeed a norm. whereas UU is the projection onto the closure of ran(A).146) In particular.147) for all A 2 B(H). Also.107) by 00 Tr + A := Tr A0+ . The corresponding properties for the trace norm are nontrivial (but true).152) p where jAj = A A (cf.143) is nite and independent of the basis (when A 2 = B1 (H). First de ne U on the range of jAj by U jAj := A .50 If A 2 B1 (H) then 2 ELEMENTARY THEORY OF C -ALGEBRAS Tr A := X i (ei Aei ) (2. Lemma 2.150) kAk k A k2 : (2. for all A 2 B(H) one has kAk k A k1 (2. for this property implies that if jAj 1 = jAj 2 then U jAj 1 = U jAj 2 . (2. iTr A.
this implies p? N ( ) jAj 2 ! jAj 2 . use (2.2. Since i (ei jAjei ) < 1.154) with equalities i H is nite-dimensional.13. we use Theorem 2. with the same conclusion.4 and (2.144) and (2. and A The basic ingredient in the proof is the following lemma.154) is sometimes seen as the non-commutative analogue of `c (X ) `1 (X ) `2(X ) `0 (X ) `1 (X ) where X is an in nite discrete set.17). Using (2. Since the star is norm-continuous by 1 1 1 ? (2.155). this analogy is strengthened by the following result. Proposition 2.157) ^ 2 B1 (H) is identi ed with A 2 B(H).156) where the ai are the eigenvalues of jAj. so 1 1 1 that jAj 2 is compact by De nition 2.13. Let pN ( ) be the projection onto the linear span of all ei .13.150).16) and (2.152). Here ^ 2 B0 (H) is identi ed with 2 B1 (H). Let A 2 B1 (H).1. Since A = U jAj 2 jAj 2 by (126.96.36.199) To show that this is well de ned. Theorem 2. the claim about proper inclusions is trivially established by producing examples on the basis of 2.4 to rewrite (2. whose proof is based on the fact that B2 (H) is a Hilbert space in the inner product (A B ) := Tr A B: (2. The proof that B2 (H) B0 (H) is similar: this time we have k jAjpN ( ) k2=k pN ( )jAj2 pN ( ) k k pN ( ) jAj2 pN ( ) k2 < 1 1 so that jAjp? N ( ) ! jAj.13. P We rst show that B1 (H)P B0 (H). Finally.9 For 2 B1 (H) and A 2 B(H) one has jTr A j k A k k k1 : (2.141) as k A k1 = k A k2 = X i i X ai a2 i (2.1) and (2. Lemma 2. Since `1 (X ) = `0(X ) and `1 (X ) = `1 (X ) = `0 (X ) .13.145). for every > 0 we can nd N ( ) such that i>N ( ) (ei jAjei ) < .155) (2.13 The C -algebra of compact operators 51 Proposition 2.8 One has B0 (H) = B1 (H) and B1 (H) = B0 (H) = B(H) under the pairing ^( ): ^(A) = Tr A = A (2. Now pN ( ) jAj 2 obviously has nite rank for every > 0. i > N ( ).7 One has the inclusions Bf (H) B1 (H) B2 (H) B0 (H) B(H) (2.2 implies that A 2 B0 (H). we have 1 pN ( ) k2 =k pN ( )jAjpN ( ) k k pN ( )jAjpN ( ) k1 < k jAj 2 so that jAj 2 p? N ( ) ! jAj 2 in the operator-norm topology.159) . This immediately gives k A k2 k A k1 implying B1 (H) B2 (H). The chain of inclusions (2.13. Finally.
13. Now choose a basis of H. Then use (2.143) is nite.143) to be nite in a suitable basis. pick A 2 H.147) and (2. Hence we have (2. For ^ 2 B0 (H) and A 2 B2 (H) k ^ k k k1 : Hence ^ 2 B2 (H) since B2 (H) is a Hilbert space. we have j ^(A)j k ^ k k A k k ^ k k A k2 : jTr pj j j k ^ k k pU k k ^ k k p k=k ^ k (2.155).159).162) we then have jTr pj j j = j N X i=1 (ei j jei )j k ^k: It follows that the sequence sN := j N i=1 (ei j jei )j is bounded. 1 1 From (2. It is clear from 2.160) (2.13.163) ^ 2 B1 (H) and To establish the converse. In view of (2. since on account of 2.4 one has A1 A2 i all eigenvalues of A1 are all eigenvalues of A2 . we need to sharpen 2 B2 (H) to 2 B1 (H). and take p to be the projection onto the subspace spanned by the rst N elements from (2. Using the original assumption ^ 2 B0 (H) .160) and (2.164) P . and de ne a quadratic form QA on H by ^(j >< j): QA( ) := A (2.161) B0 (H) we B1 (H). We turn to the proof of B1 (H) = B(H).153) to write Tr pj j = Tr pU = Tr pU = ^(pU ) changing the order inside the trace is justi ed by naive arguments. since the sum in (2.162) and (2.162) since U is a partial isometry. and note that pj j 2 Bf (H) B1 (H) the presence of p even causes the sum in (2. we estimate 1 1 1 2 j j 2 j = j((AU j j 2 ) j j 2 )j jTr A j2 = jTr AU j j 1 1 2 1 1 2 ) k2 =k k1 Tr (j j 2 U A AU j j 2 ): k j j2 k2 k (AU j j 1 2 Now observe that if 0 A1 A2 then Tr A2 Tr A2 for all A1 A2 2 B0 (H). We now prove B0 (H) = B1 (H).144) this means that k k1 k ^ k.151).152) for and (2. we conclude that B0 (H) = B1 (H) and k k1=k ^ k. by Riesz-Fischer there is an operator 2 B2 (H) such that ^(A) = Tr A for all A 2 B2 (H). so from the above insight we arrive at Tr (j j 2 U A AU j j 2 ) k k1 k AU k2 k A k2 1 1 since U is a partial isometry.161). with ^k kAk: kA (2.9 that B1 (H) B0 (H) with To prove that B0 (H) therefore have (2.13.9 that B(H) B1 (H) . It is clear from 2.154).142).143) and (2. whereas k p k= 1 in view of (2.91) we have j j 2 U A AU j j 2 k AU k2 . and since it is positive it must have a limit. choose a nite-dimensional projection p.16) and p = p2 = p . we use (2. By (2.52 2 ELEMENTARY THEORY OF C -ALGEBRAS Using (2. hence B0 (H) B1 (H).158). Now use the polar decomposition = U j j with (2. To do so. so that 2 B1 (H). as well as (2. Combining this with (2.1) for the inner product (2.
144) over a basis containing k k. with ^k kAk kA (2. Conversely.10.1 j >. for any projection p. The C -algebra B0 (H) possesses only one irreducible representation.13.6.168) ^(j >< j) = ( A ). Using A = i ].163) and (2.10.13 The C -algebra of compact operators 53 Here the operator j >< j is de ned by j >< j := ( ) .160). The state space of the C -algebra B0 (H) of all compact operators on some Hilbert space H consists of all density matrices. for which by de nition the left-ideal N de ned in (2. and in general j >< j =k k2 ].13.e.2.4 and 2.13. 2. we obtain A B(H). and subsequently by ^ = Tr A.154.13. where a density matrix is an element 2 B1 (H) which is positive ( 0) and has unit trace (Tr = 1).1. it shows that the C -algebra Mn (C ) of n n matrices has only one irreducible representation. ^ (A)B := AB: (2. Combining (2.165). For example. the number Tr p is the dimension of pH (take a basis whose elements lie either in pH or in p?H).164). and (2. when has unit length.166) (2.155)) yields 2. Note that (j >< j) = j >< j. we have ^kk kk k: jQA ( )j k A Hence by Lemma 2.12. i. Applied to the nite-dimensional case. Corollary 2.112) is zero. Hence from (2. one has Since A ^(j >< j)j k A ^ k k j >< j k1 : jA Combining (2.7 and 2.13. j >< j is the projection ]. we have Tr ] = 1.10. Hence B1 (H) continuity to B1 (H).. (2. ^ k. Hence A Extending this equation by linearity to the span Bf (H) of all j >< j.12. 3.167) (2. so that. 2. The normalization P condition k ^ k=k k1 = pi = 1 (see 2.10. In other words. which is positive. The opposite extreme to a pure state on B0 (H) is a faithful state ^.2 is then obvious from 2. P Proposition 2. so that j j >< j j = (j >< j) j >< j = ( p p )j >< j =k k k k ]: Since. one has Tr A A > 0 for all A 6= 0. up to unitary equivalence.1. Finally.168). Corollary 2. Diagonalize = i pi i ] cf.13.12. the condition ^(A) 0 yields pi 0. The next item 2.11 The GNS-representation corresponding to a faithful state ^ on B0 (H) is unitarily equivalent to the representation ^ (B0 (H)) on the Hilbert space B2 (H) of Hilbert-Schmidt operators given by left-multiplication.165) (2.13. Now note that ( A ) = Tr j >< jA this follows by such that A ^(j >< j) = Tr j >< jA. so that the we obtain B1 (H) = B(H).12.8.6) and (2.5 and (2.164) there is an operator A.10 1.3 follows from 2. The pure state space of B0 (H) consists of all one-dimensional projections. when all pi 0 the operator is positive. we nd k A k=k A identi cation of B1 (H) with B(H) is isometric. namely the de ning one. with (2. 2.2 and Corollaries 2.10.169) . evaluating (2.166).144) we obtain k j >< j k1=k k k k : ^ 2 B1 (H) by assumption.3 is one of the most important results in the theory of C -algebras.13.
Recall that the commutant M0 of a collection M of bounded operators consists of all bounded operators which commute with all elements of M the bicommutant M00 is (M0 )0 .145).173) n 0 n 0 For arbitrary B 2 M (B(H)).9. and form the linear subspace M of H. given A 2 M00 . We rst give the nite-dimensional version of the theorem this is already nontrivial. The map A ! A 1=2 .6 or 2. and.141) that for A 2 B(H) and B 2 B2 (H) one has (2. with 2 B1 (H).174) n=0 in (B + inA ) (B + in A ): (2.54 2 ELEMENTARY THEORY OF C -ALGEBRAS It is obvious from (2. The idea of the proof is to take n arbitrary vectors 1 : : : n in C n . and regard 1 + (the direct sum of n copies of H).9. so that (M)00 = (M00 ): (2.171) This follows from (2. Moreover.9.14. Hence given 1 + _ :::+ _ n and (A) 2 (M) there exists A 0 2 (M)00 such and ! 1 + that _ :::+ _ n) = A 0 ( 1+ _ :::+ _ n ): (A)( 1 + (2. By Lemma 2. although it is a statement about von Neumann algebras. compute ( B (A)])ij = Bij A]. The so-called double commutant theorem was proved by von Neumann in 1929. It is easy to see that Mn (M0 )0 = Mn (M00 ). It is easily seen that 1=2 is cyclic for ^ (B0 (H)) when ^ is faithful.158) and (2. Hence A 2 M00 commutes with p. maps B0 (H) into B2 (H).169) is well-de ned (even for A 2 B(H) rather than merely A 2 B0 (H)). A ! A := (A). use the GNS construction itself. _ :::+ _ n . Then M00 = M. we therefore have = I 2 M . and if ^ is faithful the closure (in norm derived from the inner product (2. We will write H for C n .170) so that the representation (2.14 The double commutant theorem . so = p .144) and (2. it controls the (ir)reducibility of representations. and remains a central result in operator algebra theory. Proposition 2. Hence A = A0 2 M.3 one has p 2 M0 . and its proof contains the main idea of the proof of the in nite-dimensional case as well. Using (2.5. with the substitutions H ! Hn . Now use the rst part of the proof. when A B 2 B2 (H) one has Tr AB = Tr BA: (2.171) we compute ( 1=2 ^ (A) 1=2 ) = Tr 1=2 ^ (A) 1=2 = Tr A = ^(A): The equivalence between and ^ now follows from 2. Hence (M) = M (M ). _ :::+ _ n as an element of Hn := n H ' H C n Now choose 1 : : : n 2 H. For example. and we may consider the projection p = M ] onto this subspace. Choose some = 1 2 H. Since H is nitedimensional.1 Let M be a -algebra (and hence a C -algebra) in Mn (C ) containing I (here n < 1). where i lies in the i'th copy.158)) of the image of this map is B2 (H). embed M in B(Hn ) ' Mn (B(H)) by A ! (A) := AIn (where In is the unit in Mn (B(H))) this is the diagonal matrix in Mn (B(H)) in which all diagonal entries are A. this subspace is closed. M ! (M). Furthermore. For an alternative proof.147) and the identity k AB k2 k A k k B k2 AB = 1 4 3 X When 2 B1 (H) and 0 then 1=2 2 B2 (H) see (2. and A = Ap = pA 2 M . construct a matrix A0 2 M such that A i = A0 i for all i = 1 : : : n.172) 2. Since I 2 M. Hence A = A0 for some A0 2 M.
which may not be closed.14.2 and the topologies de ned by the open sets in question is given in theory of locally convex topological vector spaces.1 is not valid when Mn (C ) is replaced by B(H). so that p = M ].1. and > 0. In other words. respectively.1. Hence \1 ) 2".1.177) where A 2 B(H). and > 0. \2 ) 3" is trivial. A) )j ! 0 for all 2 H. The estimate (2. and B 2 N. Its open sets are generated by -balls in the operator norm. Proposition 2. A basis for the strong topology is given by all sets of the form (2.14.1 to the in nite-dimensional situation. The strong topology on B(H) is de ned by the convergence A ! A i k (A . where dim(H) = 1. De nition 2. A) i )j < 8i = 1 : : : ng A B ] ) = ( AB ) . we de ne two topologies on B(H) which are weaker than the norm-topology we have used so far (and whose de nition we repeat for convenience). The following are equivalent: 1. It is easily veri ed from the de nition of weak convergence that the commutant N0 of a -algebra N is always weakly closed: for if A ! A weakly with all A 2 N. Using the CauchSchwarz inequality (2. M00 = M 2. namely the operator norm. 1 : : : n 1 : : : n 2 H. The equivalence between the de nitions of convergence stated in 2. A basis for the norm-topology is given by all sets of the form On (A) := fB 2 B(H)j k B . 1 : : : n 2 H.176) where A 2 B(H). A) i k < 8i = 1 : : : ng (2.14. which is even a norm. (B A ) = lim( A B ] ) = 0: If M00 = M then M = N0 for N = M0 . Hence A 2 M00 . which in turn is stronger than the weak topology. Since the weak topology is weaker than the strong topology. Hence (2. (B A ) = lim( A B ) . A k! 0. v) ! 0 for all . To prove \3 ) 1". we consider its closure M . whereas the strong and the weak topologies are generated by nite intersections of -balls de ned by semi-norms of the form ps (A) :=k A k and pw (A) := j( A )j. A) k! 0 for all 2 H. Instead of M . A k < g (2. A net fv g in V converges to v in the topology generated by a given i p (v . The weak topology on B(H) is de ned by the convergence A ! A i j( (A . we adapt the proof of 2. A basis for the weak topology is given by all sets of the form Ow (A 1 : : : n 1 : : : n ) := fB 2 B(H)j j( i (A . M is closed in the strong operator topology. To describe the appropriate re nement.4) shows that norm-convergence implies strong convergence. The norm-topology is de ned by a single semi-norm.1) one sees that strong convergence implies weak convergence. so that M is weakly closed.175) where A 2 B(H) and > 0.14.2 The norm-topology on B(H) is de ned by the criterion for convergence A ! A i k A . the norm topology is stronger than the strong topology. These are vector spaces whose topology is de ned by a family fp g of semi-norms recall that a semi-norm on a vector space V is a function p : V ! R satisfying 2. These topologies should all be seen in the light of the general theory of locally convex topological vector spaces. M is closed in the weak operator topology 3.2. A 0 = (A)0 for some A0 2 M.3 Let M be a -algebra in B(H). 3.173) reads A i = A0 i for all i = 1 : : : n. containing I.14. As it stands. and 4.14 The double commutant theorem 55 Therefore. Theorem 2. then ( Os (A 1 : : : n ) := fB 2 B(H)j k (B .
A ) i k2 < 2 : 1 f n X i=1 k (A . so that F restricted to . For Hn this means that _ :::+ _ n ) k2 = k (A . The space Vp inherits a topology and a projection (onto the rst co-ordinate) from X C n .56 3 HILBERT C -MODULES AND INDUCED REPRESENTATIONS implies A 2 M in other words. One should verify that a matrix-valued function p 2 C (X Mn (C )) is an idempotent (that is. it follows that A ! A for ! 0. any bundle is locally trivial in the speci ed sense. This bundle leads to possibly nontrivial sub-bundles. (2. If X is a compact Hausdor space. this implies that A 2 M.2 A vector bundle is a bundle in which 1. X (the base).1 (x) provides a homeomorphism between the latter and the typical ber F .1. whose ber above x is . One may think of B as X with a copy of F attached at each point.1 (N ) ! N F X F for which = X (where X : X F ! X is the projection onto the rst factor). The simplest vector bundle over X with ber V = C n is the trivial bundle V = X C n . Such an idempotent p de nes a vector bundle Vp .1 A bundle B(X F ) consists of topological spaces B (the total space). The maps are called local trivializations. with . for every > 0 there is an A 2 M such that k (A . so that M00 M. De nition 3.4.11. Note that the dimension of p(x) is independent of x. each ber is a nite-dimensional vector space. specialized to A = C (X ) in the proof of 2. 3 Hilbert C -modules and induced representations This chapter is concerned with the `non-commutative analogue' of a vector bundle.1 (x) ! F (where x 2 N ) is linear.1. then Mn (C (X )) ' C (X Mn (C )) is a C -algebra.1 (x) := p(x)C n . with typical ber F = V . We factorize = ( F ). A complex vector bundle is a vector bundle with typical ber C m .1 (x) is called a ber of B. Let us rst recall the notion of an ordinary vector bundle this is a special case of the following De nition 3. B ) i k2 < 2 g Os (A : : : n) (cf. F (the typical ber). as follows. each local trivialization F : . with (x f ) := x. such that the relative topology of each ber coincides with its topology as a vector space 2. for some m 2 N . A )( 1 + Noting the inclusion n X i=1 k (A . because p is continuous and X is connected. There is an integer n m and an idempotent p 2 C (X Mn (C )) such that V X C n . and a continuous surjection : P ! X with the following property: each x 2 X has a neighbourhood N such that there is a homeomorphism : .1. With the trivial inclusion M M00 . Proposition 3.11. The converse is also true. p2 = p) i each p(x) is an idempotent in Mn (C ). this proves that M00 = M.1 Vector bundles . with typical ber C m .176)). Each subset . Recall the de nition of Mn (A) in 2. 3.3 Let V be a complex vector bundle over a connected compact Hausdor space X . Since all A 2 M and M is strongly closed. relative to which all axioms for a vector bundle are satis ed. Let X in addition be connected. A ) k < . According to the de nition. We will generically denote vector bundles by the letter V.1 (x) = p(x)C n . The simplest example of a bundle over a base X with typical ber F is the trivial bundle B = X F .
(Vp ) as a nitely generated projective C (X )-module by (3. for all 2 . For example. . R (B )A1 : : : An := A1 B : : : An B: (3.(V) as a certain modi cation of a nitely generated free module for C (X ). Following the philosophy of non-commutative geometry. and idempotent p 2 Mn (A). it is easily shown that any complex vector bundle over a paracompact space can be equipped with such a Hermitian structure. a Hermitian structure is simply given by restricting the natural inner product on each ber C n of X C n to Vp .1. the action of p on n A commutes with the action by A given by right-multiplication on each component. Theorem 3.e.3) If this number is nite one says that the free module is nitely generated. given such a module p n C (X ).3.1 (x).1 Vector bundles 57 The essence of the proof is the construction of a complex vector bundle V0 such that V V0 is trivial (where the direct sum is de ned berwise) this is the bundle X C n ..1. Conversely.(V) by R (f ) (x) := f (x) (x): (3. Describing the bundle as Vp . In other words.A-module.3: any vector bundle is of the form Vp . Thus we have achieved our goal of describing vector bundles over X purely in terms of concepts pertinent to the C -algebra C (X ). There is a bijective correspondence between complex vector bundles V over X and nitely generated projective modules E (V) = . One may then choose the idempotent p 2 C (X C n ) so as to be a projection with respect to the usual involution on C (X C n ) (i. of course.1.(X C m ) is a nitely generated free module for C (X ).. one has p 2 C (X Mn (C )).1) Since C (X ) is commutative.2) A fancy way of saying this is that . De nition 3. a left-action as well. called projective. a section maps a point in the base space into the ber above the point. For any algebra A.(X C m) A itself. when V = X C n and Vp is the vector bundle described prior to 3. one obtains .3. This is a vector space under pointwise addition and scalar multiplication (recall that each ber of V is a vector space). When V is non-trivial.(V) for C (X ). Moreover. i.(V) is a right-module for the commutative C -algebra C (X ): one obtains a linear action R of C (X ) on . on which A acts by right-multiplication. we now try to describe vector bundles in terms of C -algebras. we see that . This is an immediate consequence of Proposition 3.4 Let X be a connected compact Hausdor space.4) under the obvious (right-) action of C (X ). this is. This lead to the Serre-Swan theorem. one calls p m A a nitely generated projective module for A. which continuously depends on x. The rst step is the notion of a section of V this is a map : X ! V for which ( (x)) = x for all x 2 X .e.1. When m < 1. in the trivial case one has the obvious isomorphisms ' C (X C m ) ' C (X ) C m ' mC (X ): (3. Here a free (right-) module E for an algebra A is a direct sum E = nA of a number of copies of .3. Thus one de nes the space .(V) of all continuous sections of V. and thereby a vector bundle Vp as described prior to 3.(V) the function x ! ( (x) (x))x lies in C (X ). one . Using the local triviality of V and the existence of a partition of unity.5 A Hermitian vector bundle is a complex vector bundle V with an inner product ( )x de ned on each ber . leading to . Hence the vector space p m A is a right.(Vp ) = p n C (X ) (3. More precisely.1.4). when X is a connected compact Hausdor space. In particular. Let us now add further structure.
and x]H = xH is the equivalence class in G=H of x 2 G. Any other Hermitian structure on Vp may be shown to be equivalent to this canonical one. R maps B linearly into the space of all linear operators on E .1 by `Hilbert space'. implies that every q 2 G=H has a neighbourhood N .58 3 HILBERT C -MODULES AND INDUCED REPRESENTATIONS has p = p in addition to p2 = p).e.6) and typical ber H . This is an example of the strategy of non-commutative geometry. This map is well de ned because of (3. which satis es H (xh) = H (x)h: (3.9) h B iB = h i B B (3. The proof relies on the fact that G is a bundle over G=H with projection (x) = x]H (3.1.12) for all 2 E and B 2 B. and take a unitary representation U of H on a Hilbert space H .10) h iB 0 (3.1.1 U (h)v). where 2 E and B 2 B. All required properties are easily checked. so that = ( H ) : .2. and satis es R (AB ) = R (B ) R (A)). A Hilbert bundle is de ned by replacing ` nite-dimensional vector space' in 3. whose proof we omit. The following class of Hilbert bundles will play a central role in the theory of induced group representations. for which we shall write B := R (B ) .8). What follows generalizes the notion of a Hilbert bundle in such a way that the commutative C algebra C (X ) is replaced by an arbitrary C -algebra B.5) by this action is a Hilbert bundle over X = G=H . 3.2 Hilbert C -modules . still requiring that all bers have the same dimension (which may be in nite).1 (N ) ! N H . The simplest example of a Hilbert bundle is a Hilbert space. There is no reason to restrict the dimension of the bers so as to be nite-dimensional..8) This leads to a map : . A right-action R of B on E (i. A Hilbert bundle with nite-dimensional bers is evidently the same as a Hermitian vector bundle. Here x v]H is the equivalence class in G H H of (x v) 2 G H .1 A Hilbert C -module over a C -algebra B consists of A complex linear space E . seen as a bundle over the base space consisting of a single point. and the quotient H := G H H = (G H )=H (3.2. linear in the second and anti-linear in the rst entry. This fact. Then H acts on G H by h : (x v) ! (xh.11) h iB = 0 . Proposition 3. A sesquilinear map h iB : E E ! B.6 Let H be a closed subgroup of a locally compact group G.7) and typical ber H . and is a local trivialization of G H H .1 (N ) ! N H is a di eomorphism. with projection ( x v]H ) := x]H (3. Note that the projection is well de ned. De nition 3. given by ( x v]H ) := ( x]H U ( H (x))v). satisfying h iB = h i B (3. = 0 (3.
It is easily checked that the required properties hold by continuity.4 A pre-Hilbert B ~ in the norm (3.2 Hilbert C -modules The space E is complete in the norm 59 except for positive de niteness) by (3. For (3. The space of continuous 2 sections E = . It is useful to note that (3.16) (3.13) is only a semi-norm. 2. Using (3.86). so that = 0 by (3. The inequality (3. the B ~ -action on E ~ extends One rst completes E ~ -valued to a B-action on E . (2.12). Any C -algebra A is a A-module A A over itself.18) is immediate from (3. Any Hilbert space H is a Hilbert C -module H C in its inner product. obtaining a Hilbert C -module in that way. except ~ B Given a pre-C -algebra B that the nal completeness condition is omitted. When all conditions in 3.13) coincides with the C -norm by (2.2. ~ -module E ~ can be completed to a Hilbert B-module.12).15) h k Bk iB h iB k h iB k k kk B k k k 2 h iB k kk k : (3. which is actually positive de nite because of Lemma 2. Recall that a pre-C -algebra is a -algebra satisfying all properties of a C -algebra except perhaps completeness.3.16). the rst term equals h iB h iB h iB .. The right-action of C (X ) on . The completeness of B and (3.9) and (3. . de ne a pre-Hilbert B ~ -module E ~ as in De nition 3. 2.18) To prove (3.16). Since each map ! !(h iB ) is a semi-norm (i. obtaining E . with hA B iA := A B .3 In a pre-Hilbert B inequalities 1 iC (X ) : x ! ( (x) (x))x (3.2. The B-action on E is automatically non-degenerate: the property B = 0 for all B 2 B implies that h iB B = 0 for all B .11).2.13) is indeed a norm: k k2 equals p supf!(h iB)g.8 and (3.13).10) imply that 1 k k:=k h iB k 2 : (3. and write E B.16) one uses (3.2. Expanding.17).12).2. where k (x) k= ( (x) (x))x so that it is easily seen that E is complete.1 (x). hence h iB = 0 (when B is unital this is follows by taking B = I otherwise one uses an approximate unit in B).17) we substitute h iB . and replace by = k k.14).(H) is k k = supx k (x) k. 3.1). Then use (2.1. a norm h B iB = B h iB : (3.0 (H) the function h iC (X ) is de ned by h where the inner product is the one in the ber .14) Example 3. ~ .2 1. Many Hilbert C -modules of interest will be constructed in the following way. and (2.17) (3. ~ B Corollary 3.(H) is de ned by (3.16).91). (2.18) then allow one to extend the B ~ sesquilinear form on E to a B-valued one on E . one simply takes the quotient of E by its subspace of all null vectors and completes. for in the inequality h iB 0. Note that the norm (3. ~ -module (and hence in a Hilbert B-module) one has the Proposition 3.8.(H) C (X ) over B = C (X ) for .13) We say that E is a Hilbert B-module. the supremum is a semi-norm.(H) of H is a Hilbert C -module . where the supremum is taken over all states ! on B. Let H be a Hilbert bundle H over a compact Hausdor space X . One checks that (3.1 are met except (3. In the third example the norm in .e.91). so that k k de ned by (3.
19) for all 2 E is called adjointable. When the right-hand side vanishes for all it must be that h Z iB = 0 for all . the space C (E B) is a C -algebra.2.20) for all 2 E . de ned with respect to the norm (3. To prove boundedness of a given adjointable map A. Equipped with this involution. Each element A 2 C (E B) satis es the bound hA A iB k A k2 h iB (3. It then follows from (3. The space C (E B) is norm-closed. using (3. one may wonder what C (A A) is. It is clear from (3.13) that if An ! A then An converges to some element. By self-duality this must equal h Z iB for some . it follows from (3.19) that for xed 2 E the map A ! h A iB from C (E B) to B is positive. and (3.11). As a norm-closed space of linear maps on a Banach space. Theorem 3.13) and the de nition (3.20). In the context of Example 3. we note that. To prove the nal claim.2). which by de nition is A . and bounded. A Hilbert C -module over B is called self-dual when every bounded B-linear map ' : E ! B is of the form '( ) = h iB for some 2 E .15).88).13) on E . the latter in addition requires (3. Finally.6 In a self-dual Hilbert C -module E B the C -algebra C (E B) coincides with the space L(E )B of all bounded C -linear and B-linear maps on E . Hence supfk T k j k k= 1g < 1 by the principle of uniform boundedness (here it is essential that E is complete). Indeed.2. so that its satis es (2. To establish B-linearity one uses (3.18) that k T k k A A k. one has T = h A A iB . (A )B = A( B) for all 2 E and B 2 B). We write C (E B) ! E B.16) and (3.16) one infers from (3.19) of the adjoint that k A k2 k A A k then use Lemma 2. Here we used the fact that B = 0 for all and B in the linear span of hE EiB implies B = 0. which is precisely A . C (E B) is a Banach algebra. respectively.18) once again. x 2 E and de ne T : E ! B by T := hA A iB . On the other hand. This map is isometric (hence injective). The (de ning) action of C (E B) on E is non-degenerate.10) it implies that h iB B = 0. it is almost trivial to see that A and H are the closures of A ~ ) and of a dense subspace D. Under a further assumption (which is by no means always met in our examples) one can completely characterize C (E B).1. for xed 2 E de ne 'A : E ! B by 'A (Z ) := h AZ iB .72) is easily seen to map A into C (A A). B-linear (that is. since A is adjointable. so that.2. A map A : E ! E for which there exists a map A : E ! E such that h A iB = hA iB (3.11. . which plays an important role in the induction theory in 3. hence for = Z .2.5 we only need to show that a given map A 2 L(E )B is adjointable. The map : A ! B(A) given by (2. The property of C -linearity is immediate. A A Hilbert C -module E B de nes a certain C -algebra C (E B).12).5.13) that k A k< 1.60 3 HILBERT C -MODULES AND INDUCED REPRESENTATIONS ~ (de ned over In Example 3.14) this also shows that A 2 C (E B) when A 2 C (E B).2.2.19) and (3. To check (2. and with the norm (2. for xed 2 E .85) and using (2.1.19) then leads to (3.2. so that T is bounded. In view of Theorem 3. and the map A ! A de nes an involution on the space C (E B) of all adjointable maps on E . Replacing A by A A in (2.9). so that Z = 0. Uniqueness and involutivity of the adjoint are proved as for Hilbert spaces the former follows from (3. the map Z ! h Z iB is in C (E B). as one easily veri es from (3. one has k T k k A A k k k. for by (3. (2. Proposition 3.5 An adjointable map is automatically C -linear. The adjoint of an adjointable map is unique.
28) C0 (H C ) = B0 (H) whence the dual pair B0 (H) (3. From (3.2. When A has no unit.2.24) B B AT = TA (188.8.131.52.21) C (B0 (H) B0 (H)) = B(H): (3.27) One sees from the nal part of the proof of Theorem 3. and also as a matter of independent interest.3 below.13. Proposition 3. The signi cance of the notation introduced at the end of the de nition will emerge from Theorem 3. In Example 3. The word \compact" appears between quotation marks because in general elements of C0 (E B) need not be compact operators.21) follows by taking (C0 (X )) to be the representation on L2 (X ) by multiplication operators (where L2 is de ned by a measure with support X ).22) Eq.3. (3. Its subalgebra B0 (H) of compact operators has an analogue in the general setting of Hilbert C -modules as well. so that it is a C -algebra by Theorem 3. and (3.5. . One may compute this object by taking a faithful non-degenerate representation : A ! B(H) it can be shown that C (A A) is isomorphic to the idealizer of (A) in B(H) (this is the set of all B 2 B(H) for which B (A) 2 (A) for all A 2 A). For E = H and B = C (see Example 3. Using the (trivially proved) properties = TB (3.5.22) is obtained by taking (B0 (H)) to be the de ning representation see the paragraph following 2. When C0 (E B) has a unit it must coincide with C (E B).3.2.2 1.29) H C. the proof of 2. C (A A) is the so-called multiplier algebra of A.3 The C -algebra of a Hilbert C -module 61 Using (3.1) one has This leads to the dual pair A A 2.19).3.1 The C -algebra C0 (E B) of \compact"operators on a Hilbert C -module E B is the C -subalgebra of C (E B) generated by the adjointable maps of the type T B . In preparation for the imprimitivity theorem. When A has a unit.3 The C -algebra of a Hilbert C -module k TB k k kk k : (3. one veri es without di culty that C0 (E B) is a (closed 2-sided) ideal in C (E B).2. because every bounded operator has an adjoint.2. De nition 3. one infers that A (B ) = (AB ) for all A B 2 A.18) one nds the bound (T B ) 3. (3. one therefore has C (A A) = (A) ' A cf. we introduce the analogue for Hilbert C -modules of the C -algebra B0 (H) of compact operators on a Hilbert space.3. where 2 E . and call this a dual pair.2.5 that C0 (E B) acts non-degenerately on E.26) where A 2 C (E B).2) one obtains C0 (A A) ' A: A.2 the C -algebra C (H C ) coincides with B(H).23) We write C0 (E B) E B. When E = B = A (see Example 3. One thus obtains C (C0 (X ) C0 (X )) = Cb (X ) (3.16) and (3.4. Hence (A) is an ideal in C (A A). This is the C -algebra most canonically associated to a Hilbert C -module.25) B B T A = T A (3. and T B Z := h Z iB : (3.
It is.34) For 2 E we now use (3. but has the conjugate action of complex scalars. whose norm is majorized by the norm of B in B. 1 k kB=k T A k 2 : C0 (E C0 (E B)) ' B: . The left-action L (B ) := B of B on E implements the isomorphism (3. It follows from the previous paragraph that L (B ) extends to an operator on E c (denoted by the same symbol). and (3.7 that '(C0 (A A)) = A. The space E is equipped with two norms by applying (3. It follows from (3. respectively. Eq. Hence E c = E . de nes E as a full Hilbert C -module over C0 (E B). where A 2 C0 (E B). is dense in B. Since h h iB = 0.2. and (3. and (3.11) follow from (3. we use (3.32) to nd that (3.13. (3. so that k kA =k kB . From (3. The properties (3. the isometric nature of L . in particular.33) The fullness of E B and the de nition of C0 (E c A) imply that L : B ! C0 (E c A) is an isomorphism. where run over E . (3.5.16) one nds that L (B ) is a bounded operator on E with respect to k kA .12).10).3 Let E be a full Hilbert B-module.6. (3.26). The map L is injective because E is non-degenerate as a right-B-module.31) We call A := C0 (E B) in the references to (3. applying (3. When A has a unit it is obvious that ' is surjective in the non-unital case the existence of an approximate unit implies that the linear span of all is dense in A. (3.14). It is trivial from its de nition that L is a morphism.27).13).9). To prove (3.10).34). and Lemma 3.1 and the fact that the linear span of all T C is Bf (H).12). injective. this implies B k= 0. (3. A similar de nition applies to pre-Hilbert C -modules. as E is complete in k kB by assumption.14) and (3. The completeness of E as a Hilbert Bmodule is equivalent to the completeness of E as a Hilbert A-module.29) follows from De nition 2.35) From (3. is an isometric morphism as well. and (3.3.32) L (h for the de nitions in question imply that B = Zh T A Z = h Z iA = TZ iB : (3. (3.26) that each L (B ) is adjointable with respect to h iA .30) in combination with the right-action R (A) := A . the map ' from the linear span of all T A to A.24). (3. hence = 0 by (3. Let E c be the completion of E in k kA we will shortly prove that E c = E .62 3 HILBERT C -MODULES AND INDUCED REPRESENTATIONS One has T A = ( ) see (2. below one should substitute A for B when appropriate.13) with B or with A we write k kB and k kA .13) to show that h iA = 0 implies k h i3 iB is positive by (3.23) with Z = . Given a complex linear space E . Since : A ! B(A) is an isometric morphism. The expression h iC0 (E B) := T B (3. A Hilbert C -module E over B is called full when the collection fh iB g. and that L maps B into C (E c A). Extending ' to C0 (A A) by continuity.30) and (3. Moreover.30) with = . from E B one obtains E C0 (E B). the conjugate space E is equal to E as a real vector space.13).9) etc. Now observe that iB ) = T A (3. In particular.27) with B ! A one then derives the converse inequality to (3.30). de ned by linear extension of '(T A ) = . Theorem 3.7. one sees from Corollary 2.7. (3.72).11). In other words.27) one derives k kA k kB : (3. it is norm-preserving by Lemma 2.
41) C0 (E C) = L (A): Using (3.37) this satis es (3.42) This leads to the inclusion C0 (E CP ) L (A). (Remarkably. De nition 3.38). denoted by E C. Using B = L (B ) (see 3.2 Morita equivalence is an equivalence relation in the class of all C -algebras.23) with B ! C.40) (3.4 Morita equivalence 63 We have now proved (3. and (3.1 Two C -algebras A and B are Morita-equivalent when there exists a full Hilbert C -module E B under which A ' C0 (E B).37) Moreover.3 is ready. When A M B and B M C we have the chain of dual pairs A E1 B E2 C: We then form the linear space E1 B E2 (which is the quotient of E1 E2 by the ideal IB generated by all vectors of the form 1 B 2 .36) Proposition 3.38) as de ned on E1 E2 . Finally noticing that as a Hilbert C -module over A the space E is full by de nition of C0 (E B).33). (3. this yields h iB = T C .23) with B ! C.41).28). Symmetry is implied by (3. 1 B 2 ).39) (3.31). To prove the opposite inclusion. proving that A E B implies B E A.4. (3. the space E carries a left-action L (A): the operator is bounded on E1 ) := (A 1 ) B B E2 and extends to E . with B ! C.13) is easily seen to contain IB . For later reference we record the remarkable identity h which is a restatement of (3. we can de ne a sesquilinear map h iC on E1 B E2 by h 1 B 2 1 B 2 iC := h 2 h 1 1 iB 2 iC : (3. We now claim that L (A)( 1 B 2 2 (3. The proof of transitivity is more involved. the proof of Theorem 3.10) as explained prior to (3.30). one picks a Ci i is an approximate unit in B = C (E2 C). we eventually obtain T C1 B 2 1 B 2 = L (T B1 h 2 2 iB 1 ): (3.) Apart from the right-action R (C). which establishes the dual pair B B B. One T double sequence f i2 i2 g such that N 0 i 2 2 . and (3. the null space of (3. one may therefore construct a Hilbert C -module. it is easily shown that B B 2 = 1 B h 2 2 h 1 1 iB iB 2 : L (T 1 h 2 2 iB 1 ) 1 Now use the assumption C0 (E2 C) = B as in (3.3.14). so that in constructing E one only needs to complete E1 B E2 . Substituting this in the right-hand side of (3. if one looks at (3. We write A M B and A E B.9) and (3. and using (3. and E ! E2 . iC0 (E B)Z = h Z iB (3. 3.38) With (3.4. but in fact coincides with it.3.3. the right-hand side of (3. which carries a right-action R (C) given by R (C )( 1 B 2 ) := 1 B( 2 C ): (3. The imprimitivity theorem establishes an isomorphism between the respective representation theories of two C -algebras that stand in a certain equivalence relation to each other.4 Morita equivalence The re exivity property B M B follows from (3.23). the de nition of B .10).31).3).19) with B ! C.41) becomes 1 B 2h 2 h 1 1 iB 2 iC .
such that E ~ can be made into a full pre-Hilbert A ~ -module with respect a left-action of A to the right-action R (A) := A the identity h iA (3. Proposition 3.2.44) ~ Z = h Z iB ~ ~) relating the two Hilbert C -module structures (for all Z2E the bounds h B B iA ~ hA A iB ~ ~ and B 2 B ~. Therefore. and therefore Morita equivalences. the action of A ~ extends to an action of A on k A k k A k k k for all A 2 A ~ E . where Mn (C ) and C act on C n in the usual way. In the nite-dimensional case one has Mn (C ) C n C . and a short computation using (3. This equivalence is implemented by the dual pair Mn (C ) Mn m (C ) Mm (C ). the matrix algebra Mn (C ) is Morita-equivalent to C .3. implying that A M C. one derives (3. for all A 2 A k B k2 h k A k2 h iA ~ iB ~ (3.23) with (3. Similarly.46). By (3.4. one has the dual pair A E C. with dual pair B0 (H) H C .45) (3.29). so that the B-completion E of E ~ coincides with the A-completion E c of E ~ that converse for 2 E is.23). where Mn m (C ) is the space of complex matrices with n rows and m columns.64 3 HILBERT C -MODULES AND INDUCED REPRESENTATIONS P i h i Z i = Z from (3. This proves transitivity. where E is the completion of E ~ to a Hilbert B-module E . which implies Using Corollary 3. we complete E to a Hilbert A-module E c by (3.43) Since Mn (C ) M C and C M Mm (C ).45) the left-action L (B ) := B extends to an action of B on E c . Here is a simple example of this concept. In practice the following way to construct dual pairs. and combining both inclusions one nds (3. We leave the details as an exercise.36).4. As in the proof of Theorem 3. Then A M B. In particular. is useful. one has Mn (C ) M Mm (C ).38) then has limN N i 2 2 C yields lim N N X i T C1 B i 2 1 = B i 2 L (T B1 1 ): Hence L (A) C0 (E C).4 we rst complete E ~ and 2 E ~ on E ~.3 The C -algebra B0 (H) of compact operators is Morita-equivalent to C . E c = E . This is immediate from (3. Proposition 3.3.46) ~ as a Hilbert B-module.4 Suppose one has ~ and B ~ two pre-C -algebras A ~ -module E ~ a full pre-Hilbert B ~ on E ~.34) and its ~. with dual pair A E B. hz wiC = zi wi (3. The double Hilbert C -module structure is completed by specifying (hz wiMn (C ) )ij = z i wj from which one easily veri es (3.42). .
1). The form ( is the canonical projection.3 and 3.50) and ~ k2= (A^ (3. one infers a useful property.20). Clearly.9.2. Theorem 3.4 is a special case of 3. The Hilbert space H ~ H ~ by 3.44). Since E =N ~ de ned on all of H by continuous extension of (3.3. from the proof of Theorem 2.g. each A 2 A automatically satis es (3. which imply that k ~ (A)V A )0 . The GNS-construction 2. The continuity of ~ follows from (3. Given a Hilbert B-module E .10. (3. the A-action on E is injective. then by de nition ~^ ~ ) := (^)0 : (V V ~ is the closure of E =N ~ in this inner product. the operator ~ (A) may be it follows that ~ is continuous.5 Rie el induction To formulate and prove the imprimitivity theorem we need a basic technique.3. 3. obtained by choosing E = B = A. implies that k A k= supfk ~ (A) k ! 2 S (B)g: (3. used with the de nition of k A k for A 2 C (E B). which will be used. . one nds that k ~ (A) k k A k : (3.3 shows that ~ is faithful (hence norm-preserving) when ! is. Construction 3. that A ' C0 (E B). into which it is turned after use of this proposition. In order to explicate that the induction procedure is a generalization of the GNS-construction 2.1 one sees that k A k2 = supfj!(A A)j j ! 2 S (A)g: Applying (3. In particular.5.49).9. in the proof of Theorem 3. hence faithful.3.2.4. Its null space is ~ = f 2 E j (^)0 = 0g: N (3. which is of interest also in a more general context. (3. It follows Since E is a full pre-Hilbert A from (3.. where it satis es (2.5 Rie el induction 65 ~ -module. e. We will repeatedly use Proposition 3. and (once again) the fullness of E .4.18) in succession.10.47) g)0 projects to an inner product (g) on the quotient E =N ~ .1.11)). rather than from a representation .50).47).4 in what follows see 3. g)0 on E by 1.49) (3. this form is positive semi-de nite. Using (3. (184.108.40.206.1. the goal of the Rie el induction procedure described in this section is to construct a representation of C (E B) from a representation of B. and (3.19). The analogue of (2.51) On the other hand.117) of course applies here.50) ~ is dense in H ~ . If V ~ : E ! E =N ~ 2.1. As a corollary. de ne the sesquilinear form ( (^)0 := ! (h i B ): (3.5. we rst induce from a state ! on B.36).48) (3. The representation ~ (C (E B)) is rstly de ned on E =N ~ := V ~A (A)V Since ! and h iB are positive (cf.52) (3.3.52) to B.3.44) is inspired by (3.53) A similar argument combined with Corollary 2. Given a state ! on B. as explained in Example 3.1 Suppose one has a Hilbert C -module E B.
To interrelate the above two formulations.47). this map has the property ~ ~ )0 : ~ U ~ )0 = (^ (3.116). we assume that (B) P is cyclic (if not. The situation is depicted in Figure 1.9. 1. with cyclic vector . one then uses (3. with inner product ( ) .57) and (3. (3.57) where V : E H ! E H =N is the canonical projection. The representation (C (E B)) is then de ned on H by continuous extension of (A)V ~ := V (A I ~ ) (3.54). Take a faithful state ! on B the condition implies that ~ (A) 0. Construction 3.14). Paraphrasing the comment after the rst version of the construction.9.59).117). is faithful when is.54) is positive semi-de nite.58) where I is the unit operator on H the extension in question is possible.56) N = f ~ 2 E H j ( ~ ~ )0 = 0 8 ~ 2 E H g: 2.49) and (3.54). the argument below is repeated for each cyclic summand see 2. Since the second factor equals k V ~ k2 .20) and the property k (A) k k A k (cf.58) according to (3.50) and (3. With slight abuse of notation we will write (A) for ( (A)). 3. Given a representation (B) on a Hilbert space H . Of course. With ~ = i i vi and vi = (Bi ) (where is a cyclic vector for P (B)). ~ : E ! E H by Then de ne a linear map U ~ := U : (3.10) to nd ( ~ ~ )0 = (v (h iB )v) with := i i Bi .10) and the positivity of : B ! B(H ).60) According to (3.5. This particularly applies when one has a given (pre-) C -algebra A and a -homomorphism : A ! C (E B). since k (A) k k A k : (3. This form is positive semi-de nite.59) To prove that the form de ned in (3. Also. because ( ) and h iB are. de ned by (V ~ V ~ ) := ( ~ ~ )0 (3. one veri es that is non-degenerate when and are. one computes k (A)V ~ k2 = (v (hA A iB )v) from (3.2). .1). it is not di cult to verify that is non-degenerate when is. (3.3 Start from a Hilbert C -module E B. Further to an earlier comment. When one starts from a representation (B) rather than from a state.57) the map U by (3. this proves (3.2 Let A 2 C (E B) satisfy h A iB 0 for all 2 E . one assumes that is cyclic. Then A 0. The Hilbert space H is the closure of E H =N in this inner product. The null space is N = f ~ 2 E H j ( ~ ~ )0 = 0g: (3. this is bounded by k A k2 (v (h iB )v) . the text after 2. the sesquilinear form ( )0 is de ned on E H (algebraic tensor product) by sesquilinear extension of ( v w)0 := (v (h iB )w) (3.5. Similarly. and (2. any subspace of C (E B) may be subjected to the induced representation . and (3.58) duly intertwines ~ and . we may equally well write (3.61) (U ~ therefore quotients to a unitary isomorphism U : H ~ ! H . leading to the representation (A) on H .66 3 HILBERT C -MODULES AND INDUCED REPRESENTATIONS Lemma 3. the general construction looks as follows.55) As in (2.54) where v w 2 H . The form ( )0 projects to an inner product ( ) on the quotient E H =N . Hence ( ~ ~ )0 0 by (3. which By (3.
one has (B) ' (B). one has (A) ' (A). de ned with respect to some representation (A) and the dual pair B E A. Taking (A) = (A) as just de ned. Theorem 3.1 There is a bijective correspondence between the non-degenerate representations of Morita-equivalent C -algebras A and B. Starting with (B). Conversely.6.58) and the above dual pair. Let the pertinent dual pair be A E B. relabel this representation as (A). ~ : E E H ! H by linear extension of We rst de ne U ~ U v := (h iB )v: (3.62) . This correspondence is as follows. We then construct a unitary map U : H ! H which intertwines and . When (A) is a representation on a Hilbert space H there exists a representation (B) on a Hilbert space H such that is equivalent to the Rie el-induced representation de ned by (3. See Figure 2. and move on to construct (B) from Rie el induction with respect to the dual pair B E A.6 The imprimitivity theorem 67 A HH HH HHH H j H - E B - H H ? induction Figure 1: Rie el induction A E B @@ @ @ R @ @@ R @ - H B E A @ B H := H @ H ' Figure 2: Quantum imprimitivity theorem: H ' H and 3. taking (B) = (B). In the opposite direction.3. preserving direct sums and irreducibility.6 The imprimitivity theorem After this preparation. we construct (A) with Rie el induction from the dual pair A E B. a given representation (B) is equivalent to the Rie el-induced representation . we pass to the imprimitivity theorem.
The right-hand side of (3. (3. A representation of C (G) on a Hilbert space H is de ned as a morphism : C (G) ! B(H). with ! .3. Now put = and H = H .65) from right to left.1 at the group level. Combined with Proposition 3. Using the proof that the Morita equivalence relation is symmetric (see 3.19) and (2.7 Group C -algebras f g(x) := and the involution X y z2Gjyz=x f ( y )g (z ) (3.62).3. Hence we see from the construction rstly of H = H from E H .3. and (3. This is made into a -algebra by the convolution 3.4. the operation de ned by (3. and subsequently (3. using (3.4.2).58).18) with ' ! Note that U and (3.1 ): . C (G) consist of all complexvalued functions on G. so that U is (B) is non-degenerate.57) from right to left. and thus arrives at a contradiction.63) ~ is indeed C -linear. and also in the theory of induced representation as originally formulated for groups by Frobenius and Mackey. since ' . with ! . one immediately sees that the construction works in the opposite direction as well. Thus writing L (B ) = B .13.64) ~ annihilates ~ .65) and (3. all read from right to left.67) (3.10. one obtains ~ 1 1 v1 U ~ 2 2 v2 ) = ( 1 v1 T B1 2 2 v2 )0 : (U Now use the assumption A = C0 (E B) to use (3.54). the right-hand side of (3.68 3 HILBERT C -MODULES AND INDUCED REPRESENTATIONS . we use (3. As a vector space. (3. de ned by linear extension of (cf. one puts = as above. that U UV ( V ~ v) := U v = (h iB )v: (3.5.66) is seen to be (B )UV ( V v). and use (3. Using (3.65) Using the assumptions that the Hilbert C -module E B is full and that the representation ~ and hence of U is dense in H . and secondly of H from E H ~ descends to an isometry U : H ! H . In many interesting applications.63) is then seen to be equal to (V 1 v1 (h 1 2 iA )V 2 v2 ) . we see that the range of U unitary. where ~ 2 E H . This shows that ~ 1 1 v1 U ~ 2 2 v2 ) = (V ( 1 V 1 v1 ) V ( 2 V 2 v2 )) : (U (3. 2 2 leads to It is easy to verify that = 1 = 1 .54) and (3.1 = x and (xy).23). Hence U (B ) = (B )U for all B 2 B. f (x) := f (x. decomposes = 1 2 representation (B) as = 1 . We start with the de nition of the group algebra C (G) of a nite group with n(G) elements one then usually writes C (G) instead of C (G).3). whenever ~ 2 N or In particular.14). This also proves that the bijective correspondence (B) $ (A) preserves irreducibility: when is irreducible and 2 .66) where the left-action of B 2 B on 2 E is as de ned in 3. to compute U (B ) V ( V v) = (h L (B ) iB )v (3.68) is an involution because of the properties (x.1 x. then decomposes the induced isn't.1 ).68) It is easy to check that the multiplication is associative as a consequence of the associativity of the product in G. U V ~ 2 N .57) and (3.58).36). In similar vein. this theorem leads to a new proof of Corollary 2.18) with ' ! .1 = y. the C -algebra B featuring in the de nition of a Hilbert C -module and in Rie el induction is a so-called group C -algebra. the properties (2. 3. so that C (G) = C n(G) . (2. To verify that U intertwines and .
Putting x = e in (3. each left Haar measure is also right-invariant. but simpli es most of the formulae. When U1 (x) = V U2 (x)V for all x 2 G then evidently 1 (f ) = V 2 (f )V for all f 2 C (G). The involution is given by (3. It is elementary to verify that is indeed a representation of C (G) when U is a unitary representation of G. G dy jg(y)j Z G dy jg(y)j Z G dx jf (xy.3.1 )j G dx jf (x)j =k f k1 k g k1 . We denote Haar measure by dx it is unique up to normalization. It is obvious from invariance of the Haar measure under x ! x.3. and putting k f k:=k (f ) k.72).71) and (3. we construct a faithful representation on a Hilbert space. initially on Cc (G).73) Hence L1 (G) is a Banach -algebra under the continuous extensions of (3. (f ) = 1 (f ) 2 (f ) for all f i U (x) = U1 (x) U2 (x) for all x.7. We also assume that G is unimodular that is. Unimodularity implies that the Haar measure isR invariant under inversion x ! x. Proposition 3. by f g(x) := Z it is evident that for a nite group this expression specializes to (3. The converse follows by choosing f = x . Similarly. The convolution product is de ned.71) and (3. Let now G be an arbitrary locally compact group (such as a nite-dimensional Lie group).56). In order to equip L1 (G) with a C -norm. The proof of (3.68).69) and in the other by x2G where x (y) := (xy).7. The Banach space L1 (G) and the Hilbert space L2(G) are de ned with respect to the Haar measure.67).1.72) is a straightforward generalization of the case G = R cf.2 The operations (3.1 There is a bijective correspondence between non-degenerate representations of the -algebra C (G) and unitary representations U of G.1 )g(y) (3.1 )g(y)j Z which is (3.5.68) from Cc (G) to L1 (G). As in the nite case.1 . one checks that these operations make C c (G) a -algebra this time one needs the invariance of the Haar measure at various steps of the proof. This correspondence is given in one direction by X (f ) := f (x)U (x) (3.72) k f k1 = k f k1 : (3. and vice versa.70) We can de ne a C -norm on C (G) by taking any faithful representation . U (x) := ( x ) (3. so that cannot be degenerate. This assumption is not necessary.1 that k f k1 =k f k1 . When G is compact we choose the normalization so that G dx = 1. Recall the de nition of a Banach -algebra below 2. This time we have G dy f (xy. so that the involution is certainly continuous.70) yields ( e ) = I.71) k f g k1= Z G dx j Z = Z G dy f (xy. which preserves unitary equivalence and direct sums (and therefore preserves irreducibility). Since C (G) is a nite-dimensional vector space it is complete in this norm.10. which therefore is independent of the choice of by Corollary 2.68) are continuous in the L1-norm one has k f g k1 k f k1 k g k1 (3.7 Group C -algebras 69 Proposition 3. (2.
(2.1 x) and f I (x) = N N dy f (xy. we now note from (3. seen as a Banach -algebra as in 3.76) Z G dx f (x)UL (x): The boundedness of L (f ) then follows from Lemma 3. but this distribution is not in L1 (G).3. When G is nite.2. In other words.7.93) holds as an equality.70 3 HILBERT C -MODULES AND INDUCED REPRESENTATIONS Proposition 3. For f 2 Cc (G) one therefore has lim I f = f and lim f I = f pointwise (i.4. In order to apply Lemma 2. which is the characteristic function of N times a normalization factor ensuring that k I k1 = 1.7.79) Since the Fourier transform turns convolution into pointwise multiplication.68) and the invariance of N under R inversion. . and therefore for all A 2 L1 (G). since Cc (G) is dense in L1 (G).92) then holds by virtue of (3.59) maps L1 (G) into a subspace of C0 (R) which separates points on R. for xed x). One has I f (x) = N N dy f (y. so that each N is invariant under x ! x. To prove that L is faithful.94) hold for all A 2 L1(G). and k k=k k1 . (220.127.116.11.78) follows from Lemma 2. satisfying k L (f ) k k f k1 .75) (3. Lemma 3. It is clear that for every p 2 R there is an f 2 L1(R) for which f^(p) 6= 0.4. so that f = 0 and To nish the proof of 3. 2 L2 (G). For general locally compact groups one would like to take the Dirac -`function' as a unit.7.94) for all A 2 Cc (G). Pick a basis of neighbourhoods N of e. we need to verify that k f kr =k f^ k1 : (3. Eq.77) Cr (R) ' C0 (R): (3.3 For f 2 L1(G) the operator L(f ) on L2(G). Take I = N N .7.4. the left-regular representation L on L2(R) is Fourier-transformed into the action on L2 (R) by multiplication operators. One easily veri es that L (f g) = L (f ) L (g ) and L (f ) = L (f ) . By construction. and hence certainly for = I .92) . Proposition 3.1 this basis is partially ordered by inclusion.6 that the Fourier transform (2.7.e.7.6 One has the isomorphism (3. L is injective.7 below. we rst show that L1 (G) possesses the analogue of an approximate unit (see 2.1x) it follows that L (f ) = (3.74) that L (f ) = 0 implies f = 0 for all containing L (Cc (G)).1 for C -algebras).7. Hence k f k1 = 0 by Lemma 3. The Lebesgue dominated convergence theorem then leads to (2. the delta-function e is a unit in C (G). Cr (G) is the closure of the latter in the norm De nition 3. theR inequality (2.74) is bounded. de ned by L (f ) := f : (3..5 The reduced group C -algebra Cr (G) is the smallest C -algebra in B(L2 (G)) k f kr :=k L (f ) k : Perhaps the simplest example of a reduced group algebra is obtained by taking G = R. The linear map L : L1(G) ! B(L2 (G)) is a faithful representation of L1 (G).7. Hence (3. Introducing the left-regular representation UL of G on L2 (G) by UL(y) (x) := (y.78) It follows from the discussion preceding 2.4 The Banach -algebra L1(G) has an approximate unit I in the sense that (2.3.1 .
For example. that is. one obtains ^ ): Cr (G) ' C0 (G (3. Lemma 2. It will always be directly clear from this whether a given operator. The Bochner integral X d (x)f (x) can be manipulated as if it were an ordinary (Lebesgue) integral.75) and L in (3.84) Thus reading (3.80) k (f ) k k f k1 : ( (f ) ) := (3. De nition 3. . with G G dx f (x)U (x): (3.82) is bounded. A function f : X ! B is Bochner-integrable with respect to a measure on X i f is weakly measurable (that is.or vector-valued integral may be read as a Bochner integral if not. one may de ne (3. where F (x) := k k jf (x)j.83) is immediate from (3.81) We return to the general case.7 Let U be an arbitrary unitary representation of G on a Hilbert space H.84).3.7. where G is not necessarily abelian. Lemma 3. we have j( (f ) )j (F F )L2 (G) for all 2 H.7. We have now found a C algebra which may play the role of C (G) for locally compact groups. by its matrix elements Z G dx f (x)( U (x) ): Since U is unitary. a unitary representation of a topological group is always meant to be continuous.5 then leads to (3. one has p k Z X d (x) f (x) k Z X d (x) k f (x) k : (3. Then (f ). for each functional ! 2 B the function x ! !(f (x)) is measurable) there is a null set X0 X such that ff (x)jx 2 X nX0g is separable the function de ned by x !k f (x) k is integrable.8 Let X be a measure space and let B be a Banach space.82) as a Bochner integral. de ned as f^( ) := Z By the same arguments as for G = R. We explain this notion in a more general context. Let us agree that here and in what follows. the analogue of Proposition 3.82) is most simply de ned weakly. Unfortunately. in a sense always R obvious from the context. Hence we need a di erent construction.83). it is understood as a weak integral.7 Group C -algebras 71 ^ be the set of all This example generalizes to arbitrary locally compact abelian groups.7. de ned by Z (f ) := dx f (x)U (x) (3.82) as a Bochner integral.76) to arbitrary representations. ^ of ^ is nothing but the set of characters on G. Alternatively. Let G irreducible unitary representations U of G such representations are necessarily one-dimensional.83) The integral (3. The following result generalizes the correspondence between UL in (3. The generalized Fourier transform f so that G 1 ^ f 2 L (G) is a function on G. The Cauchy-Schwarz inequality then leads to j( (f ) )j k f k1 k k2 . (3.12.1 only holds for a limited class of groups.
k (f ) k k f k k f k1 : . proving that must be non-degenerate.83) by 3.9. This correspondence preserves irreducibility. and only one representative of each equivalence class of such representations is included. given U (G) one nds (L1 (G)) satisfying (3. To go in the opposite direction we use the approximate unit once more it follows from (3.82). The equivalence between the two de nitions follows from (2. if U is reducible there is a projection E such that E U (x)] = 0 for all x 2 G. Since is continuous. Since k (Ix) k k Ix k1 = 1. the Banach -algebra L1 (G) which satisfy (3.88) that one may extend to a representation of C (G) by continuity. and unitary representations U of G. given by (continuous extension of) (3.85).12. and therefore irreducibility. we infer that k U (x) k 1 for all x.82) and (3. It follows from (18.104.22.168 3 HILBERT C -MODULES AND INDUCED REPRESENTATIONS Theorem 3.82) indeed de nes a representation.4.1 y).10 The group C -algebra C (G) is the closure of the Banach -algebra algebra L1(G) in the norm (3.82) de nes U on a dense subspace of this summand it will be shown that U is unitary. C (G) is the closure of L1 (G) in the norm This theorem suggests looking at a di erent object from Cr (G). Hence also k U (x.1 ) k 22.214.171.124 that any non-degenerate representation of a C -algebra is a direct sum of cyclic representations the proof also applies to L1 (G).1. It is bounded by Lemma 3. and therefore corresponds to U (G) by Theorem 3.10.7. which is the same as k U (x). Thus in (3. The unitarity of each U (x) follows by direct calculation.7.85) (from which the continuity of U is obvious) that U (x) (f ) = lim (Ix) (f ) .111) and (3. Conversely.9. in which U is an irreducible unitary representation of G.7.83). if is reducible then E (Ix)] = 0 for all x 2 G by the previous paragraph this implies E U (x)] = 0 for all x. hence is reducible.82).139) and Theorem 3. and (3. Conversely.71).87) where the sum is over all representations (L1 (G)) of the form (3. Finally. Theorem 3.3. it follows from easy calculations that (f ) in (3.9.85) Recall from 2.11 There is a bijective correspondence between non-degenerate representations has of the C -algebra C (G) and unitary representations U of G.82) that (f ) E ] = 0 for all f . Hence U (x) = lim (Ix) strongly on a dense domain. one has lim (I ) = I strongly.83).85) and (3.86) that for any representation (C (G)) and f 2 L1 (G) one (3. This correspondence is given in one direction by (3. or from the following argument.86) where u is the direct sum of all non-degenerate representations of L1(G) which are bounded as in (3.1 k 1 We see that U (x) and U (x).85) stands for a cyclic vector of a certain cyclic summand of H. The nal claim then follows from Schur's lemma 2. Inspired by 2.2 one puts k f k:=k u (f ) k k f k:= sup k (f ) k (3. so that it can be extended to all of H by continuity. Given U . 2 De nition 3.9 There is a bijective correspondence between non-degenerate representations of U (x) (f ) := (f x ) where f x(y) := f (x. and in the other by (3.9 it then follows from (3.1 are both contractions this is only possible when U (x) is unitary.7. It is obvious from (2. Equivalently. This bijection preserves direct sums.88) Hence the restriction (L1 (G)) satis es (3.7. The proof of non-degeneracy makes use of Lemma 3.83). The property U (x)U (y) = U (xy) then follows from (3.
8.79).96) for all x y 2 G and A B 2 A.1 x)) G x (f (x. When A has no unit.1 that ( (A)) = (A) for any A 2 A any automorphism hence k (A) k=k A k (3.93) De nition 3. Proposition 3. one has C (Rn ) ' C0 (Rn ): (3.2 Let (G A ) be a C -dynamical system.94) x y (A) = xy (A) ( AB ) = ( A ) ( A ) (3. L (C (G)) is faithful i G is amenable.7.90) Cr (G) = L (C (G)) ' C (G)= ker( L ): A Lie group group is said to be amenable when the equality Cr (G) = C (G) holds in other words.2 and 3.87) and (3.80). looking at 3.89) For general locally compact groups. The operations Z G dx k f (x) k (3.18) and the uniqueness of the unit. However. Z dy f (y) y (g(y.95) x x x (A ) = (A) (3.1 An automorphic action of a group G on a C -algebra A is a group homomorphism x ! x such that each x is an automorphism of A.5 we see that (3.98) (3. Hence a general C -dynamical system is a non-commutative analogue of a dynamical system. the second de nition of C (G) stated in 3.8.97) (3. The term `dynamical system' comes from the example G = R and A = C0 (S ). The reason is that for 2 G ^ f ( ) 2 C . and de ne L1(G A ) as the space of all measurable functions f : G ! A for which 3.92) by (2.79) coincide.1 ) ) . a C -algebra A. one has (3. In particular. one may extend to an automorphism I of the unitization AI by I(A + I) := (A) + I: (3. It follows from De nitions 2.8 C -dynamical systems and crossed products k f k1 := is nite. One A has a unit. de ned by x !k x (A) k. In other words. It follows from the Peter-Weyl theorem that all compact groups are amenable as well.11 implies that for ^ one has (f ) = abelian groups C (G) always coincides with Cr (G). is continuous. and t (f ) : ! (t).8. non-compact semi-simple Lie groups are not amenable.7. We have just seen that all locally compact abelian groups are amenable. and an automorphic action of G on A such that for each A 2 A the function from G to A. An automorphism of a C -algebra A is an isomorphism between A and A.8 C -dynamical systems and crossed products 73 In conjunction with (3. A C -dynamical system (G A ) consists of a locally compact group G. so that the norms (3.2.99) f g(x) := f (x) := turn L1 (G A ) into a Banach -algebra.3.91) by (2. where R acts on S by t : ! (t). one has (I) = I (3.
8 are satis ed as a consequence of the continuity assumption in the de nition of a C -dynamical system. Proposition 3. (3.4 describes the way unitary representations of the Poincare group are constructed in algebraic quantum eld theory.72) and (3. that not all covariant representations of a C -dynamical system arise in this way a given unitary representation U (G) may may not contain the trivial representation as a subrepresentation cf. where U is a unitary representation of G. in which ! is then taken to be the vacuum state on the algebra of local observables of the system in question.8. B )) = 0 by (3. Theorem 3. Furthermore.9 is as follows. and ~ is a non-degenerate representation of A which for all x 2 G and A 2 A satis es U (x)~ (A)U (x) = ~( x (A)): (3. Hence U (x) is unitary. Note that (3. so that U (x) ! (A) ! = U (x) ! (B ) ! .101).120).8.103).73) one follows the same derivation as for L1 (G).84) and (3.101) for all x 2 G and A 2 A. The completeness of L1 (G A ) is proved as in the case A = C . In order to generalize Theorem 3. for which L1(G A ) = L1 (G). the generalization of Theorem 3.9. B ) (A . and covariant representations (U (G) ~ (A)).4 Let (G A ) be a C -dynamical system. B )) ! k2 = 0 by (2. Recall (3. and suppose one has a state ! on A which is G-invariant in the sense that !( x (A)) = !(A) (3.101) imply that (U (x) ! (A) ! U (x) ! (B ) ! ) = ( ! (A) ! ! (B ) ! ): This shows rstly that U (x) is bounded on ! (A) ! . whereas (3.93) implies that U (x) ! = ! : (3. de ne an operator U (x) on the dense subspace ! (A) ! of H! by U (x) ! (A) ! := ! ( x (A)) ! : (3. however.98) is de ned as a Bochner integral the assumptions in De nition 3.102) This operator is well de ned. For x 2 G. (3.103) Proposition 3. which is unitary from H! to the closure of U (x)H! .1 (B ) in (3.102) with (3. B )) = 0 by (2. Secondly.7. and de nes a unitary representation of G on H! .120).7. If ! (A) ! = ! (B ) ! then !((A . B ) x (A .74 3 HILBERT C -MODULES AND INDUCED REPRESENTATIONS As usual.5 Let (G A ) be a C -dynamical system.92) or (3.8. There is a bijective correspondence between non-degenerate representations of the Banach -algebra L1 (G A ) which satisfy (3.8. U (x) is a partial isometry. using (3. so that it may be extended to H! by continuity. Hence ! ( x (A)) ! = ! ( x (B )). The integral (3. whose closure is H! because ! is cyclic.94) implies that U (x)U (y) = U (xy). so that k ! ( x (A . Note.102).102) and (3.104) .7. This correspondence is given in one direction by (f ) = Z G dx ~ (f (x))U (x) (3. one sees that U (x)H! = ! (A) ! . Taking A = x.97). Consider the GNS-representation ! (A) on a Hilbert space H! with cyclic vector ! .100) Here is an elegant and useful method to construct covariant representations.3 A covariant representation of a C -dynamical system (G A ) consists of a pair (U ~ ). To verify the properties (3. we have assumed that G is unimodular with a slight modi cation one may extend these formulae to the non-unimodular case. In any case.91). Hence !( x (A . we need De nition 3.83).
7. We write Lx(q) = xq := L(x q). and therefore irreducibility. The proof of this theorem is analogous to that of 3.104).104) and (3. C (G A ) is the closure of L1 (G A ) in the norm k f k:= sup k (f ) k (3.107) where u is the direct sum of all non-degenerate representations of L1 (G A ) which are bounded as in (126.96.36.199) where the sum is over all representations (L1 (G A )) of the form (3. If G is a Lie group and Q is a manifold we assume that L is smooth.109) is then simply the usual action on R3 . where Q is a locally compact Hausdor space.9. Here we simply say that a covariant representation (U ~ ) is irreducible when the only bounded operator commuting with all U (x) and ~ (A) is a multiple of the unit. The crossed product C (G A ) of G and A is the closure of the Banach -algebra algebra L1 (G A ) in the norm k f k:=k u (f ) k (3.10. 3.9 Transformation group C -algebras We now come to an important class of crossed products. For example. and therefore irreducibility. De nition 3. when G = SO(3) and H = SO(2) is the subgroup of rotations around the z -axis. restricted to S 2 . The approximate unit in L1(G A ) is constructed by taking the tensor product of an approximate unit in L1 (G) and an approximate unit in A.7 Let (G A ) be a C -dynamical system. This action is given by x y]H := xy]H (3.1.7. we put De nition 3. . where is a cyclic vector for a cyclic summand of (C (G A This bijection preserves direct sums. If G and Q are locally compact we assume that L is continuous.83). The main example we shall consider is the canonical action of G on the coset space G=H (where H is a closed subgroup of G). Theorem 3. This correspondence preserves direct sums.1 A (left-) action L of a group G on a space Q is a map L : G Q ! Q.139) and Theorem 3. Equivalently. and puts U (x) (f ) = (~x (f )) (3. (3.9.105). in which (U ~ ) is an irreducible covariant representation of (G A ).106) ~ )).6 Let (G A ) be a C -dynamical system.8.6 etc. Generalizing 3.5. The proof is identical to that of 3.8. and x is de ned as follows. at least at a heuristic level. There is a bijective correspondence between non-degenerate representations of the crossed product C (G A ) and covariant representations (U (G) ~(A)). and only one representative of each equivalence class of such representations is included.7.1 y)).105) ~ (A) (f ) = (Af ) (3. We assume the reader is familiar with this concept.9 Transformation group C -algebras 75 in the other direction one de nes Af : x ! Af (x) and ~x (f ) : y ! x (f (x. The SO(3)-action (3. satisfying L(e q) = q and L(x L(y q)) = L(xy q) for all q 2 Q and x y 2 G. The equivalence between the two de nitions follows from (2.109) where x]H := xH cf.8.3. This correspondence is given by (continuous extension of) (3.9. The rest of the proof may then essentially be read o from 3. in which A = C0 (Q).106). one may identify G=H with the unit two-sphere S 2 in R3 . 3.
(3.114) Xf dt where Exp : g ! G is the usual exponential map.117) .121) (3. and may be seen as a generalization of the canonical These equations hold on the domain HU commutation relations of quantum mechanics. Since Cc (Q) is dense in C0 (Q) in the sup-norm. 2. and (3. the l'th co-ordinate function on Rn . and ~ 2 Rnf0g we put re-expressing it in in nitesimal form. respectively. and that the operator idU (X ) is essentially ! . Eq.91). Using (3.115) Q~ (f~) := ~ (f~): (3.112) 1 .112) makes sense when is a smooth vector for a U this is an element 2 H 1 of for which the map x ! U (x) from G to H is smooth.1 q): x (f for all f~ 2 Cc (Q). (3. one therefore obtains l .100) on the pair (U ~ ) may now be clari ed by 1 (Q). (3.119) i Q (qk ) Q (ql )] = 0 ~ ~ ~ i Q (T ~ ) Q (T ~ )] = 0 ~ ~ (3.76 3 HILBERT C -MODULES AND INDUCED REPRESENTATIONS Assume that Q is a locally compact Hausdor space. it is easy to prove that the continuity of the G-action on Q implies that ~) .113).120) (3. we then obtain i Q (f~) Q (~ (3. It can be shown that the set HU smooth vectors for U is a dense linear subspace of H. given a unitary representation U of a Lie group G on a Hilbert space H one can construct a representation of the Lie algebra g by When H is in nite-dimensional this de nes an unbounded operator. Taking f = q l .116) From the commutativity of C0 (Q). The meaning of the covariance condition (3. f ~ k= 0 lim k (f (3.118) (3. so that one may form the commutative C -algebra C0 (Q) cf. For X 2 g. ~). which is not de ned on all of H. The relations (3. It is quite instructive to look at covariant representations (U ~ ) of (G C0 (Q) ) in the special case that G is a Lie group and Q is a manifold. the same is true for f~ 2 C0 (Q). one sees that this function is continuous on all of G. Firstly. ^ X Y ]) ~ ~ i Q (X ~ ) Q~ (f~)] = Q~ ( X f~): ~ ~ (3.113) Secondly. Hence (G C0 (Q) ) is a C -dynamical system. To see this. given a Lie group action one de nes a linear map X ! X from g to the space of all vector elds on Q by ~(q) := d f~(Exp(tX )q)jt=0 (3.111) x!e x ~) : q ! f~(x.117) ~ ~ g)] = 0 ~ d U (Exp(tX )) dU (X ) := dt jt=0 : (3.119) then become k ql = k i Q (X ~ ) Q~ (Y ~ )] = Q~ (. consider the case G = Q = Rn .122) i Q (T~ ) Q (ql )] = l : k ~ k ~ ~ k ~ l .100). A G-action on Q leads to an automorphic action of G on C0 (Q). so that e has a basis of compact neighbourhoods. f~ 2 Cc ~ ) := i~dU (X ) Q ~ (X (3. where the G-action is given by L(x q) := q + x. If X = Tk is the k'th generator of Rn one has k := Tk = @=@q k .110) Using the fact that G is locally compact. on H1 one has self-adjoint on HU U dU (X ) dU (Y )] = dU ( X Y ]): (3. Moreover.94) and Hence the function x ! x (f~) from G to C0 (Q) is continuous at e (as e (f~) = f (3.4. given by (3.
L1(G C0 (Q)) is then identi ed with the space of all such functions f for which k f k1 = f g(x q) = Z is nite cf. Hence one may identify Q~ (qk ) and Q~ (T respectively.124) (3. ~ = Cc (G G). Since (Cc (G G)) is dense in B2 (L2 (G)) in the Hilbert-Schmidt-norm (which is a standard fact of Hilbert space theory).98) and (3. To prove this. the operations (3.9. Proposition 3.126) . so that is a representation of the -algebra Cc (G G). It is easily veri ed that the Hilbert-Schmidt-norm (2. in which we take A ~ = Cc (G). but one may show ~k ) always possess a from the representation theory of the Heisenberg group that Q~ (qk ) and Q~ (T common dense domain on which (3.1 x. One may identify L1 (G C0 (Q)) with a subspace of the space of all (measurable) functions from G Q to C an element f of the latter de nes F 2 L1 (G C0 (Q)) by F (x) = f (x ). Since is evidently faithful.141) of (f ) is Z Z k (f ) k2 = dx dy jf (xy.3. G is here assumed to be unimodular.13. we conclude from (2. B ~ = C.1 q) Z One veri es from (3.129) h iCc (G G) := (y) (x.) De nition 3.1 x y. whereas the Cc (G G)-valued inner product The left-action of A ~ on Cc (G) is given by (3.3 Let a locally compact group G act on Q = G by L(x y) := xy. Here is a simple example.128) ~ on E ~ is as de ned in (3.4.1 x) (y): (3. Clearly. We start from Cc (G G).45) and (3. The pre-Hilbert C -module Cc (G) C is de ned by the obvious C -action on and E Cc (G).1 x)j2 : (3.154) that (Cc (G G)) B0 (L2 (G)).122) are valid.120) .123) (3. although the construction clearly depends on it.1 y): It is not necessary to consider the bounds (3.1 even Bf (L2 (G)) is dense in B0 (L2 (G))).124) and (3.1 q): G dy f (y q)g(y. The transformation group C -algebra C (G Q) is the crossed product C (G C0 (Q) ) de ned by the automorphic action (3.9 Transformation group C -algebras 77 ~k ) with the quantum position and momentum observables. This.46).126). is the case if all irreducible representations of the -algebra Cc (G G) are unitarily equivalent to .110). Following the proof of Theorem 3.(3.127) 2 Since this is clearly nite for f 2 Cc (G G). and B2 (L2 (G)) is dense in B0 (L2 (G)) in the usual operator norm (since by De nition 2. one shows directly that there is a bijective correspondence between the representations of Cc (G G) and of C . we conclude that the closure of (Cc (G G)) in the operator norm coincides with B0 (L2 (G)). Conventionally.108) coincides with the operator norm of ( ).97). G G G dy f (xy. (It should be remarked that Q~ (qk ) is an unbounded operator. (3. we proceed as in Proposition 3. regarded as a dense subalgebra of C (G G).6.125) that (f ) (g) = (f g) and (f ) = (f ) . We de ne a linear map : Cc (G G) ! B(L2 (G)) by (f ) (x) := f (x q) = f (x.9. in turn.4. the G-action L on Q is not indicated in the notation C (G Q). and therefore the isomorphism C (G G) ' B0 (L2 (G)).99) read G dx sup jf (x q)j q 2Q (3. and the inner product h iC := ( )L2 (G) : (3.125) Z As always. the equality (C (G G)) = B0 (L2 (G)).2 Let L be a continuous action of a locally compact group on a locally compact space Q. Then C (G G) ' B0 (L2 (G)) as C -algebras. follows from the previous paragraph if we can show that the norm de ned by (3. In this realization.1.
11). given by UR (h) (x) := (xh): (3.4.71).130) h iCc (H ) : h ! Z Interestingly.7.134) The properties (3. For simplicity we assume that both G and H are unimodular.9) and (3. Combining (3.4. (3. respectively. Theorem 3. we nd G G=H H ! (h iCc(H ) ) = Z G=H d (q) k Z H dh (s(q)h)U (h) k2 : (3.11 (with G ! H ).123) ~ = Cc (H ).109).136).131) h iCc (H ) : h ! ( U (h) )L2 (G): (3.1 The transformation group C -algebra C (G G=H ) is Morita-equivalent to C (H ). we take a vector state ! on C (H ). This will be done on the basis of Proposition 3. In 3. We note that the Haar measure on G and the one on H de ne a unique measure on G=H .10) are easily veri ed from (3. seen as a dense subalgebra of A = C (G G=H ) as explained prior to A (3. Condition .78 3 HILBERT C -MODULES AND INDUCED REPRESENTATIONS 3.1 ) (3. This proves (3.10.136) for any f 2 Cc (G). one has Z dh f (h)U (h. so that h iCc(H ) is positive in C (H ) by Corollary 2.82).131).10. E We make a pre-Hilbert Cc (H )-module Cc (G) Cc (H ) by means of the right-action = f :x! R (f ) Z Here f 2 Cc (H ) and 2 Cc (G). where H is a closed subgroup of G. satisfying Z Z Z dx f (x) = d (q) dh f (s(q)h) (3.135). with corresponding unit vector 2 H .135) where U is the unitary representation of H corresponding to (C (H )) see Theorem 3. both formulae may be written in terms of the right-regular representation UR of H on L2 (G). this proves that (h iCc (H ) ) is positive for all representations of C (H ).4 we take ~ = Cc (G G=H ).10 The abstract transitive imprimitivity theorem We specialize to the case where Q = G=H .4.3.133) R (f ) = which should be compared with (3. and H G (3. The Cc (H )-valued inner product on Cc (G) is de ned by H dh (xh. We need to construct a full Hilbert C -module E C (H ) for which C0 (E C (H )) is isomorphic to C (G G=H ). seen as a dense subalgebra of B = C (H ) B ~ = Cc (G). and any measurable map s : G=H ! G for which s = id (where : G ! G=H is the canonical projection (x) := x]H = xH ).132) Namely.11). and the G-action on G=H is given by (3.1 )f (h): dx (x) (xh): (3.137) Since this is positive. To prove (3.68) and (3. and (3. Hence for f 2 Cc (H ) L1 (H ) one has ! (f ) = ( (f ) ) = Z H dh f (h)( U (h ) ) (3. This leads to the transformation group C -algebra C (G G=H ).
Math. and that Cc (G) Cc (G G=H ) is a pre-Hilbert C -module with respect to the right-action of Cc (G G=H ) given by := L (f ) cf. This section is evidently continuous.6). in the sense that it gives a necessary and su cient criterion for a unitary representation to be induced. when G and H are unimodular).46) are satis ed.141) One should check that U ( ) (x) ( ) is again a section.3. These are given by L (f ) (x) = Z h iCc (G G=H ) : (x y]H ) ! G dy f (xy.141) satis es (U ( ) (x) ( ) U ( ) (x) ( ) )=( ( ) ( ) ): (3. ~ = Cc (G G=H ) on Cc (G) and the Cc (G G=H )-valued We now come to the left-action L of A inner product h iCc (G G=H ) on Cc (G). Enthusiastic readers may nd the proof in M.1 x]H ) (y) Z (3. one may check that L is indeed a left-action. since the G-action on G=H is continuous.138). and 2 Cc (G). given by ( ( ) ( ) ) := Z G=H d (q) ( ( ) (q ) ( ) (q)) (3. In order to explain the idea of an induced group representation from a geometric point of view.1 q)): (3.A.1 then turns out to be equivalent to a complete characterization of induced group representations.137) as well. and that the bounds (3. since h iCc(H ) = 0 implies that the right-hand side of (3. Since one may choose s so as to be piecewise continuous. and ( ) is the inner product in the ber .(H ) by 3.45) and (3. using (3. The G-action U on the vector bundle H induces a natural G-action U ( ) on the space of continuous sections . de ned on ( ) 2 . it is easy to verify R (f ) The theory of induced group representations provides a mechanism for constructing a unitary representation of a locally compact group G from a unitary representation of some closed subgroup H .44).139) H dh (yh) (x. There is a natural inner product on the space of sections . in that (U ( ) (x) ( ) (q)) = q see (3. the operator U ( ) (x) de ned by (3.124) and (3.(H ). one needs to show that the Hilbert C -modules Cc (G) Cc (H ) and Cc (G) Cc (G G=H ) are full.137) vanishes for all .142) where is the measure on G=H de ned by (3. (3.6.10. we return to Proposition 3. The group G acts on the Hilbert bundle H de ned by (3.125). To complete the proof. but an argument that is su ciently elementary for inclusion in these notes does not seem to exist.12) easily follows from (3.11 Induced group representations 79 (3. Rie el. and (3. This is indeed the case.c(H ) of continuous sections of H with compact support (in the norm derived from this inner product). this implies that = 0. (3.140) Since the left-action x : y ! xy of G on itself commutes with the right-action h : y ! yh of H on G.1 yh): the crucial condition (3.143) . Induced representations of C -algebras.139).131). 13 (1974) 176-257. Note that di erent identi cations of the ber with H lead to the same inner product.4. the action (3.1 (q) ' H . This implies that the function (q h) ! (s(q)h) vanishes almost everywhere for arbitrary sections s.11 Induced group representations U ( ) (x) ( ) (q) := U (x) ( ) (x.5) by means of U (x) : y v]H ! xy v]H : (3. Adv. Theorem 3. When the measure is G-invariant (which is the case.140) is clearly well de ned. for example.136). Using (3.(H ) of H .138) (3. Also.1. 3. The Hilbert space L2 (H ) is the completion of the space .4.130).
De ne the Hilbert space L2 (H ) of L2 -sections of the Hilbert bundle H as the completion of . For ( ) to lie in . In this realization the inner product on .1 ) = U (h) (x): Such a map de nes a section ( ) by Note that the representation (3. Since U ( ) (x) is invertible.1.c(H ) is given by ( ) := Z G=H d ( (x)) ( (x) (x)) (3. The section ( ) thus de ned is independent of the choice of x 2 .1 ) have the same null sets for all x 2 G.145) The pair (U ( ) (G) ~( ) (C0 (G=H ))) is a covariant representation of the C -dynamical system (G C0 (G=H ) ). de ne a representation ~( )(C0 (G=H )) on L2(H ) ~(q) ( ) (q): ~ ( ) (f~) ( ) (q) := f (3.11.147) by ( ) ( (x)) = x (x)]H where : G ! G=H is given by (3.80 3 HILBERT C -MODULES AND INDUCED REPRESENTATIONS When fails to be G-invariant.3 The Hilbert space H is the completion in the inner product (3.c (H ). it is therefore a unitary operator.144) simpli es to (3. Proposition 3.144) Proposition 3.11.143).146).1 (q))=d (q) exists as a measurable function on G=H . Hence U ( ) (x) is isometric on .146). where is given by (3. For this purpose we note that a section ( ) : Q ! H of the bundle H may alternatively be represented as a map : G ! H which is H -equivariant in that (xh. One easily veri es that the square-root precisely compensates for the lack of G-invariance of .110).146).1 Let G be a locally compact group with closed subgroup H .145) is nothing but the right-action (3. The map x ! U ( ) (x) given by (3. and the .11.136). the Radon-Nikodym derivative q ! d (x.140) de nes a unitary representation of G on 2 L (H ).2 In the context of 3.142).148) of the set of continuous functions : G ! H which satisfy the equivariance condition (3.1 q)): (3. it can be shown that it is still quasi-invariant in the sense that ( ) and (x.146) (3.7).1 q)) (q) := d (dx (q( ) U (x ) s ( ) (x.1. the expression (3.1 ).c (H ).11. the projection of the support of from G to G=H must be compact. The representation U ( ) (G) is said to be induced by U (H ).141) to U ( ) (x) ( ) .148) the integrand indeed only depends on x through (x) because of (3.1 ( (x)) because of (3.c (H ) in the inner product (3. guaranteeing the property (3.144) with (3.11. When is G-invariant. Consequently. this follows from a simple computation. The property U ( ) (x)U ( ) (y) = U ( ) (xy) is easily checked. and can be extended to L2 (H ) by continuity.1) of (C0 (G=H )) on L2(H ) this right-action is at the same time a left-action. and let U be a unitary representation of H on a Hilbert space H . so that it is bounded. One then modi es (3. where the measure on G=H is de ned by (3. Given 3. (3. with inverse U ( ) (x. because (C0 (G=H )) is commutative.141). projection of whose support to G=H is compact. We now give a more convenient unitarily equivalent realization of this covariant representation. De nition 3.
and consider the C -dynamical system (G C0 (G=H ) ). given by 3.11.151) that (3. and (3.1 y)]H : Hence we infer from (3. Theorem 3.11.149) Using (3. 3. Recall (cf.11.1 and 3.3) that a covariant representation of this system consists of a unitary representation U (G) and a representation ~ (C0 (G=H )).110). we de ne the induced G-action U on by y U (x) (y)]H := U ( ) (x) ( ) ( (y)): (3. Comparing (3. The pair (U (G) ~ (C0 (G=H ))). as well as the de nition x (y) = x y]H = xy]H = (xy) of the G-action on G=H (cf. states that all covariant pairs (U (G) ~ (C0 (G=H )) arise in this way. and (3.156) ~ 2 C0 (G=H here f ~x(q) := f~(x.110). The original imprimitivity theorem of Mackey.12 Mackey's transitive imprimitivity theorem 81 Given (3. we obtain U ( ) (x) ( ) ( (y)) = U (x) ( ) (x.1 Let G be a locally compact group with closed subgroup H .154) holds.153) in the sense that V U (y)V . where U is given by (3.155) V ~ (f~)V .154) for all y 2 G. 3.153) with (3. satisfying the covariance condition U (x)~ (f~)U (x).1. for all x 2 G and f Any unitary representation U (H ) leads to a covariant representation (U (G) ~ (C0 (G=H )) of (G C0 (G=H ) ).1 q). which preserves direct sums and therefore irreducibility (here the equivalence relation is unitary equivalence).149) that U (y) (x) = (y. where is given by (3. This leads to a bijective correspondence between the space of equivalence classes of unitary representations of H and the space of equivalence classes of covariant representations (U ~ ) of the C -dynamical system (G C0 (G=H ) ). Conversely.10.2. any covariant representation (U ~ ) of (G C0 (G=H ) ) is unitarily equivalent to a pair of this form.141) by (3.147).149) to (3.6. (3.151) Similarly. de ne a representation ~ (C0 (G=H )) on H (cf. An analogous but simpler calculation shows (3. In the preceding section we have seen that the unitary representation U (G) induced by a unitary representation U of a closed subgroup H G can be extended to a covariant representation (U (G) ~ (C0 (G=H )). we then have Proposition 3. it should be obvious from the argument leading from (3.141).147).145) reads ~ (f~) (x) := f~( x]H ) (x): (3.155).1 = ~ ( ) (f~) ~ for all f 2 C0 (G=H ). which historically preceded Theorems 188.8.131.52) by (3.152) Analogous to 3. 3.151).147).144) in the above derivation yields .3. is a covariant representation of the C -dynamical system (G C0 (G=H ) ).1 = ~ (f~x ) (3.1 = U ( ) (y) (3.140). where is given by (3.152).1 x)) U (y) (x) = d d( ((y(x)) s (y.150) Replacing (3.152).1 x): (3.7)).12 Mackey's transitive imprimitivity theorem .1 y (x. in the realization H the representation (3.1 y)]H = y (x.11. (3.8.151) and (3.4 In the context of 3.1 (y))) = U (x) x.1 x): (3.12. This pair is unitarily equivalent to the pair (U ( ) (G) ~( ) (C0 (G=H ))) by the unitary map V : H ! H( ) given by V ( (x)) := x (x)]H (3.3. (3.
1.136).162) G .8.82 3 HILBERT C -MODULES AND INDUCED REPRESENTATIONS The existence of the bijective correspondence with the stated properties follows by combining Theorems 3. so that. the representation (C (E B)) is equivalent to the representation (C (E B)).13). applies to H . starting from the pre-Hilbert C -module Cc (G) Cc (H ) de ned in the proof of 3. unitary between H and the closure of the image of U In any case.158) ~ ). Let us start with a technical point concerning Rie el induction in general.148). comparing with (3. The explicit form of the correspondence remains to be established.131). (3.1. where the norm on the left-hand side is in H ~ . with Theorems 3.3.58). which relate the representations of C (H ) and C (G G=H ). however.148) one has (3.2 Suppose one has a Hilbert space H (with inner product denoted by ( ) ) and a ~ : E H ! H satisfying linear map U ~~ U ~ ~ ) = ( ~ ~ )0 (U (3. of course. ~ : Cc (G) H ! H de ned by linear extension of Consider the map U ~ U v(x) := Z H dh (xh)U (h)v: (3. will play the role H in 3. Using (3. and (3.160) (3. one nds that the induced representation of C (G G=H ) on H is given by Z (f ) (x) = dy f (xy.1 x]H ) (y) (3.11 and 3.12.47).82).57). Using (3.6.10. Otherwise. Lemma 3.1 and 3. (3. one obtains ( v ~ (U w )0 = ~ vU Z H dh ( UR (h) )L2 (G) (v U (h)w) = Z H dh Z G dx (x) (xh)(v U (h)w) cf. which allow one to pass from (C (H )) to U (H ) and from (C (G G=H )) to (U (G) ~ (C0 (G=H )).158) and (3.157).5. It is clear that U desired unitary map U : H ! H . with G ! H . by H . using the invariance of the Haar measure on H . When Then U the image is dense this map extends to a unitary isomorphism U : H ! H .138). as follows from the invariance of the Haar measure.10.159) Note that the equivariance condition (3. one veri es ~ (Cc (G) H ) is dense in H .12.161) w)H = Z H dh Z G=H d ( (x)) Z H dk (xk) (xh)(U (k)v U (h)w) : Shifting h ! kh. de ned by continuous extension of ~ ~ := U ~ (A I ~ ): (A)U (3.159) and (3. The same comment. and the norm on the right-hand kV ~ obtained by Rie el-inducing side is the one de ned in (3. and using (3.12.13). respectively.54). It follows that the induced space H from a pre-Hilbert C -module is the same as the induced space constructed from its completion. one indeed has U It is obvious that N = ker(U U.7.146) is indeed satis ed by the left-hand side. U is ~.2. The Hilbert space H de ned below (3. ~ quotients to an isometric map between E H =N and the image of U ~ in H . from (3. so by Proposition 3. We will use a gerenal technique that is often useful in problems involving Rie el induction. we continue to denote the Hilbert space H de ned in Construction 3.132). and will therefore be denoted by this symbol. Using (3. (3.7. one shows that ~ k k k. and (3.157) for all ~ ~ 2 E H . To avoid notational confusion. On the other hand.2 one obtains the (3. = We use this lemma in the following way.
or some other other class of real-vaued functions. whereas in (3.1 see Theorem 2.146).83 this looks just like (3. The observables are taken to be either all unbounded self-adjoint operators A on H.4) A key di erence between classical and quantum mechanics is that even in pure states expectation values are generally unsharp. The observables of the theory are functions on S one could consider smooth. C0 (S R). continuous.8. The situation is quite neat if S and AR stand in some duality relation. with the di erence that in (3.2. and is an eigenvector of A.162) corresponds to the covariant representation (U (G) ~ (C0 (G=H )) by the correspondence (3.8. say that a physical theory consists of a convex set S . However. etc. but this is not our concern here.138) lies in Cc (G).1 The mathematical structure of classical and quantum mechanics 4 Applications to quantum mechanics h f i := (f ) = d ( ) f ( ): S Z (4. More generally. interpreted as the state space a real vector space AR. whose points are interpreted as the pure states of the system.8. For example.104) . then the space of all probability measures on S is precisely the state space of A = C0 (S ) in the sense of De nition 2. which assigns the expectation value h! f i to a state ! and an observable f . etc. one should specify the dynamics of the theory. Hence the space AR of observables may be taken to be C 1 (S R). This pairing is given by 4. There is a pairing h i : S AR ! R 1 between the state space S of probability measures on S and the space AR of observables f . In addition. This time the pairing between states and observables is given by h Ai = Tr A: (4. whereas the set of observables is a real vector space (barring problems with the addition of unbounded operators on a Hilbert space). in that h f i2 6= h f 2 i.3) In a pure state one has h Ai = ( A ): (4.162) satis es the equivariance condition (3.162) lies in H . therefore. In these examples. Finally. We may.2) In elementary quantum mechanics the state space consists of all density matrices on some Hilbert space H the pure states are identi ed with unit vectors . in the classical case. In the same sense. mixed states are identi ed with probability measures on S . or all bounded self-adjoint operators. one should check that the function (f ) de ned by (3. In classical mechanics one starts from a phase space S .5.138). consisting of the observables a pairing h i : S AR ! R 1. Indeed. bounded.1) The physical interpretation of this pairing is that in a state the observable f has expectation value h f i.106) of Theorem 3. the state space has a convex structure. if S is a locally compact Hausdor space. Cb (S R). or all compact self-adjoint operators. it is a simple exercise the verify that the representation (C (G G=H )) de ned by (3. In general. or L1 (S R). measurable. The only exception is when an observable A has discrete spectrum.(3. in a pure state (seen as the Dirac measure on S ) the observable f has sharp expectation value (f ) = f ( ): (4. this expectation value will be unsharp. in quantum mechanics the space of all density matrices on H is the state space of the C -algebra B0 (H) of all compact operators on H . and we take AR = C0 (S R).
then the space of observables is the dual of the (linear space spanned by the) state space. The physical interpretation of quantum mechanics is a delicate matter.5 that a positive map automatically preserves self-adjointness. with Borel structure . etc.1. It is desirable (though not strictly necessary) that Q preserves positivity. A PVM is usually written as ! E ( ) it follows that each E ( ) is a projection (take 1 = 2 in the de nition). Note that the above conditions force 0 A( ) I. such as B0 (H)R or B(H)R. A projection-valued measure or PVM is a POVM which in addition satis es A( 1 \ 2 ) = A( 1 )A( 2 ) for all 1 2 2 . and L(H) denotes some space of self-adjoint operators on H. When X is compact it is required that Q(1X ) = I. a general quantum system is speci ed by some C -algebra A. Ideally. On the other hand. This means that one has a classical phase space S and a linear 0 1 map Q : A0 R ! L(H).6) The map Q is a representation of C0 (X ) i S ! A( ) is a PVM. a given quantum system arises from a classical system by `quantization'. and POVM's ! A( ) on S in H. We are thus led to the following De nition 4. In practice. if we take AR to be the space B(H)R of all bounded self-adjoint operators on H. one then ascribes the same physical interpretation to the corresponding quantum observable Q(f ).2. A quantization of X consists of a Hilbert space H and a positive map Q : C0 (X ) ! B(H).6. as well as the (approximate) unit. It is quite convenient to assume that A0 R = C0 (S R).2 Let X be a set with a -algebra of subsets of X . There is an interesting reformulation of the notion of a quantization in the above sense. regarded as C -algebras.13. We will not go into this generalization of elementary quantum mechanics here. of course.1 Let X be a locally compact Hausdor space.5) for all f 2 C0 (X ) this implies that f 2 C0 (X R) is mapped into a self-adjoint operator.13.1. In the C -algebraic approach to quantum mechanics. whose self-adjoint elements in AR correspond to the observables of the theory. satisfying A( ) = 0. .6.2 Quantization Q(f ) = Q(f ) (4. The state space of AR is then given by De nition 2. De nition 4.10. Proposition 4.8.84 4 APPLICATIONS TO QUANTUM MECHANICS see Corollary 2. This general setting allows for the existence of superselection rules.8.2. in that 4. There is a bijective correspondence between quantizations Q : C0 (X ) ! B(H). where AR stands for C (S R).8. with the intrinsic notion of positivity given by 2.3 Let X be a locally compact Hausdor space. which choice discards what happens at in nity on S . This provides the physical meaning of al least all operators in the image of Q.2. rather then vice versa see Theorem 2. and concentrate on the choice A = B(H). Also recall De nition 2. A(X ) = I. It follows from 2.4 of a positive map.1. Given the physical meaning of a classical observable f . and A( i i ) = i A( i ) for any countable collection of disjoint i 2 (where the in nite sum is taken in the weak operator topology). and when X is non-compact one demands that Q can be extended to the unitization C0 (X )I by a unit-preserving positive map. one needs to specify the physical meaning of any observable A 2 AR. A positive-operatorvalued measure or POVM on X in a Hilbert space H is a map ! A( ) fromP to B(H)+ (the set of positive operators on H). or C0 (S R). This notion is familiar from the spectral theorem. given by Z Q(f ) = dA(x) f (x): (4. Here C0 (X ) and B(H) are. with the same choice of the state space.
Let S be a locally compact Hausdor space (interpreted as a classical phase space). For 2 one then puts ( A( ) ) := ( ). a unitary map U : H ! pH . a projection p on H . Given the assumptions.11. When X is a con guration space Q.8) By Proposition 2. The probability that.4.1 = pE ( )p for all 2 .2. S Z ]: (4. in a state . and consider an embedding ! of S into some Hilbert space H.6) is expressed by this equation.3 is given by Q(f ) = d ( ) f ( ) In particular. The ensuing map ! A( ) is easily checked to have the properties required of a POVM. Hence the Riesz representation theorem yields a probability measure on X .2.4.11. one veri es that the property E ( )2 = E ( ) is equivalent to Q(fg) = Q(f )Q(g). Moreover. Combine Theorem 2. de ning an operator A( ) by polarization.11.2. the situation is therefore described by a PVM ! E ( ) on Q in H. the physical interpretation of the map p ( ) := Tr A( ) (4. a quantization Q : C0 (X ) ! B(H) is a completely positive map. We will now construct a class of examples of quantization in which one can construct an illuminating explicit realization of the Hilbert space H and the partial isometry W . The positive map Q corresponding to the POVM ! A( ) by Proposition 4.3 Stinespring's theorem and coherent states A( ) = Z d ( ) ] (4.3.3. Corollary 4.7 and 2. it is usually su cient to take the positive map Q to be a representation of C0 (Q) on H.7) is the probability that. for each pair 2 H a POVM ! A( ) in H de nes a signed measure on X by means of ( ) := ( A( ) ). By Proposition 4. one has Q(1S ) = I.9) guarantees that we may de ne a POVM on S in H by S 4. Conversely.11) .2. such that each has unit norm (so that a pure classical state is mapped into a pure quantum state).4 Let ! A( ) be a POVM on a locally compact Hausdor space X in a Hilbert space H .3. Condition (4. in a state .2 are satis ed.3 Stinespring's theorem and coherent states 85 The precise meaning of (4. for arbitrary 2 H one constructs a functional ^ on C (X ) by ^ (f ) := ( Q(f ) ) Since Q is linear and positive.10) where ] is the projection onto the one-dimensional subspace spanned by (in Dirac's notation one would have ] = j >< j).9) 2) = ( 1 2 ): for all 1 2 2 H.4.6 we may as well assume that X is compact. in view of 2. this functional has the same properties. The are called coherent states for S . Given Q.1 implies that the conditions for Stinespring's Theorem 2. ! A( ) is contained in the statement that the number When X is the phase space S of a physical system. and De nition 4.6) will emerge shortly.2. and a PVM ! E ( ) on H such that UA( )U . This yields a positive map Q : C (X ) ! B(H) by R ( Q(f ) ) := X d (x) f (x) the meaning of (4. the system in question is localized in S. the system is localized in Q is p ( ) := Tr E ( ): (4. there should be a measure on S such that Z d ( )( 1 )( (4.2 with Proposition 4. Approximating f g 2 C (X ) by step functions. There exist a Hilbert space H .
86 For example. Eq. Extending the map Q from C0 (S ) to C 1 (S ) in a heuristic way.13) ): (4. where k = 1 2 3.4 Covariant localization in con guration space In elementary quantum mechanics a particle moving on R3 with spin j 2 N is described by the Hilbert space j = L2 (R3 ) Hj HQM (4.12) that V pV . one nds that Q(qi ) and Q(pi ) are just the usual position.12) in H = L2 (R3 ). and de nes the quantization of a classical observable f 2 C0 (S ) as multiplication by f .11. We then map L2 (T R3 d3 pd3 q=(2 )3 ) into K := L2(C 3 d3 zd3 z exp(.20) (4. where Q is the con guration space of the system.i~ kl S S Jk Ql ] = i~ klm QS m PkS PlS ] = 0 S J S ] = i~ klm J S Jk l m S P S ] = i~ klm P S Jk m l (4. For example. In speci c cases the projection p = WW can be explicitly given as well. for (2. These operators satisfy the commutation relations (say.24) . and S Jk (angular momentum). We may then verify the statement of the theorem by taking H = L2(S d ): The map W : H ! H is then given by (4.18) where Hj = C 2j +1 carries the irreducible representation Uj (SO(3)) (usually called Dj ). sandwiched between the projection onto the subspace in question. One realizes the state space as a closed subspace of L2 (S ) (de ned with respect to a suitable measure).and momentum operators in the Schrodinger representation. 4.16) W ( ) := ( It follows from (4. In Theorem 2.q) =2 (p q ) (4. so that = (p q). The basic S physical observables are represented by unbounded operators QS k (position). given by (4.(x.15) (4. in p the case S = T R3 considered above one may pass to complex variables by putting z = (q . 2 ipq+ipx e. B = B(H). on the domain S (R3 ) Hj ) S QS k Ql ] = 0 PkS QS l ] = . (4.n=4 e. ip)= 2. Pk (momentum).19) (4.14) and (4. (A) = A for all A.22) (4.1 is the projection onto the space of entire 1 p p functions in K. This should be contrasted with the usual way of doing quantum mechanics on L2 (Q).zz)=(2 i)3 ) by the unitary operator V .14) (4.9) that W is a partial isometry. The representation (C0 (S )) is given by ( f ) ( ) = f ( ) ( ): Finally.9) then holds with d (p q) = d3 pd3 q=(2 )3 .1 = pfp: Q Eqs. z )= 2 q = (z + z )= 2): One may then verify from (4. one may take 1 2 (x) = ( ).16) form the core of the realization of quantum mechanics on phase space.2 we now put A = C0 (S ).23) (4.17) V (z z) := e 2 zz (p = (z .13) and (4.122) one has the simple expression ~(f ) = U Q(f )U . when S = T R3 4 APPLICATIONS TO QUANTUM MECHANICS = R6 . (4.21) (4.
The commutation relations (4.152).112).12. a)): UQM (4. with given by (3. The commutation relations (4.110). (4.4.152).18). and (3.21) are a consequence of the covariance relation (3. Generalizing this approach to an arbitrary homogeneous con guration space Q = G=H .3 one may instead assume one has a PVM ! E ( ) on G=H in H and a unitary representation U (G). one then has PkS = j (P ) and J S = i~dU j (T ) see (3. respectively. a non-degenerate representation QM C -algebra C (E (3) R3 ) (cf.113) and the commutation relations in the Lie algebra of E (3). we de ne a representation ~QM 0 QM j (f ~) = f~ Ij ~QM (4.7 this is equivalent to a covariant representation (U (G) ~ (C0 (G=H ))) on H. a nondegenerate representation of C (G G=H ) on a Hilbert space H describes a quantum system which is localizable in G=H and covariant under the canonical action of G on G=H . j of C (R3 ) on Hj by Moreover. we have Theorem 4.2. as well as the fact that ~QM 3 Identifying Q = R with G=H = E (3)=SO(3) in the obvious way.1 An irreducible quantum system which is localizable in R3 and covariant under E (3) is completely characterized by its spin j 2 N .151).4.156).26). (4.26).25) In terms of the standard generators Pk and Tk of R3 and SO(3).152) with ~QM This is a neat explanation of spin in quantum mechanics. The associated PVM ! E ( ) on R3 where f j in HQM 4.3. which satisfy U (x)E ( )U (x). To see this. given by unitary representation UQM QM j (R a) (q ) = Uj (R) (R. 3 commutativity of C0 (R ). The physical interpretation of the PVM is given by (4. The momentum and angular momentum operators are most conveniently de ned in terms of a j of the Euclidean group E (3) = SO(3) n R3 on Hj . Pk . Such a tions. Generalizing Proposition 4. the spectral theorem for unbounded operators.11.29). given by 3.4 Covariant localization in con guration space 87 justifying their physical interpretation. Eq.1. and (4.(4.115) play the role of quantized momentum observables. (4.20). (3. .7.25) is unitarily This follows from Theorem 3.8.4. Proposition 4. then by de nition describes a quantum system which is jlocalizable in R3 . is equivalent to the one described by (4.25) that the pair (UQM 0 QM 3 dynamical system (E (3) C0 (R ) ). The system of imprimitivity (U (G) ~j (C0 (G=H ))) is equivalent to the one described by (3. representation (3.4. and J k . The corresponding covariant representation (U j (E (3)) ~j (C0 (R3 ))). one checks that the canonical left-action of E (3) on E (3)=SO(3) is identi ed with its de ning action on R3 .1 (q . j (E (3)) de ned in (4. The representation UQM j equivalent to the induced representation U . and covariant under the de ning action of E (3). in which case the quantum system itself is said to be irreducible. It is then not hard to j (E (3)) ~ j (C (R3 ))) is a covariant representation of the C verify from (4.8. in terms of which the position operators are given by R (see QS = dE ( x ) x cf. (4. 3.25). and their j j (C (R3 ))).22) . By 3.19) then re ects the k k R3 j is a representation. In addition.1 = E (x ) (4. or. It is natural to require that QM be irreducible.2 An irreducible quantum system which is localizable in Q = G=H and covariant under the canonical action of G is characterized by an irreducible unitary representation of H .151) and (3.3) is E ( ) = Ij.8) the operators de ned in (3.26) ~ is seen as a multiplication operator on L2 (R3 ).2. check that the unitary map V : Hj ! j de ned by V j (q ) := j (e q ) intertwines U j and U j . equivalently. and by Proposition 4. it intertwines the HQM QM j as de ned in (4.24) follow i~dUQM k k QM k from (3.1. S S Rather than using the unbounded operators QS commutation relak . we therefore state the situation in terms of the pair (UQM (E (3)) ~QM 0 j of the transformation group pair.27) for all x 2 G and 2 cf.
which for all x 2 G and f satis es the covariance condition U (x)Q(f~)U (x) = Q( x (f~)): (4. one checks that (U ~ ) is a covariant ~ .1 = A(x ): (4. . and ask what happens in the presence of a symmetry group.2 commutes with all U (2.88 4 APPLICATIONS TO QUANTUM MECHANICS This is immediate from Theorem 3.1. one infers that SO(3)-covariant quantum particles on S 2 are characterized by an integer n 2 Z.3. and a projection p on H in the commutant of U (G). and (4. where arises from a continuous G-action on X by means of (3.2 we therefore omit the representation occurring in 2.11.29) Every (ordinary) covariant representation is evidently a generalized one as well.2 and 2. A class of examples of truly generalized covariant representations arises as follows. ~ (x) on C0 (G=H ) H by linear extension of For x 2 G we de ne a linear map U ~ (x)f (4.11. (3. Indeed.8. Hence U ~ puting on C0 (G=H ) H and then passing to the quotient. and suppose that U (G) is reducible. 4.3 with B(H).3 by H induced representation (3.31) ~ (G) quotients to a representation U ~ (G) on H ~ .3 and the sentation in Construction 2.151) will still be called H and .12.152). Let (U (G) ~ (C0 (G=H )) be a covariant representation on a Hilbert space K. which generalizes De nition 3.5 Covariant quantization on phase space Let us return to quantization theory. There exists a unitary representation U (H ).2. U condition (4. The following notion.11.2.12. Theorem 184.108.40.206. is natural in this context. with corresponding covariant representation (U ~ ) of (G C0 (G=H ) ) on the Hilbert space H .1 A generalized covariant representation of a C -dynamical system (G C0 (X ) ). where U is a unitary representation of G on a Hilbert space H.11. Of course. Comwhere ( )0 is de ned in (2. We identify B above. as described by 3. where H is speci ed in 4. as in the formulation of the theorem ~ ) may be identi ed with ( H ). 4.2 etc.1. we denote the Hilbert space H and the repre~ and ~ . Pick a projection p in the commutant of U (G) then (pU (G) p ~ p) is a generalized covariant representation on H = pK. (2.1 to the case where the representation ~ is replaced by a quantization Q. We apply Theorem 2. For each unitary irreducible representation U of SO(2) is labeled by such an n. By Theorem 3.2.110).30) U := x (f ) U (x) : ~ is clearly a G-action. To avoid confusion.3. This is veri ed from Finally. putting H = H. such that (pU (G) p ~ p) and (U (G) Q(C0 (G=H ))) are equivalent.28) This condition may be equivalently stated in terms of the POVM ! A( ) associated to Q (cf. the projection p de ned in 2. this system must be of the form (U ~ ) (up to unitary representation on H equivalence).127). one veri es that ~ (x)f ~ (x)g )0 = (f (U U g )0 (4. This class actually turns out to exhaust all possibilities. What follows generalizes Theorem 3.128).28) and the unitarity of U (x). since a representation is a particular example of a quantization. (U ~ ) is described by Theorem 3. ~ (x). The claim follows. writing the two-sphere S 2 as SO(3)=SO(2).151).28). and U is a representation.5.11. de ned with respect to the canonical G-action on G=H .11.1).1. and given by Un ( ) = exp(in ). and Q : C0 (X ) ! B(H) ~ 2 C0 (X ) is a quantization of C0 (X ) (in the sense of De nition 4.12.123). For example. our goal is to show that (~ H in 2.2 Let (U (G) Q(C0 (G=H ))) be a generalized covariant representation of the C dynamical system (G C0 (G=H ) ).3) by U (x)A( )U (x). and (3. Using the covariance Since x y = xy . De nition 4. respectively the space de ned in 3. consists of a pair (U Q).12.. and must be of the form (U ~ ).11.
R. Amer. 1978] Unitary Group Representations in Physics Probability and Number Theory. Locally Compact Groups and Banach -Algebraic Bundles. In: Doran. (ed. Math. pp. Robinson 1981] Operator Algebras and Quantum Statistical Mechanics. Symp. Vol. Oxford University Press. (ed. M. Models in Statistical Mechanics. Symmetry Groups. Benjamin.W. Lecture Notes in Mathematics 582.and W -Algebras. In: Kadison. R. Takesaki. 1996] C -Algebras by Example.. Academic Press. New York.) C -algebras: 1943-1993. and J.R. N.G. . E. Cont. G. 1994] Notes on the Gelfand-Neumark theorem.W.W. Kadison. LMS Lecture Notes 210. 1979] Theory of Operator Algebras I. K. New York. J. Mackey. Davidson. Pedersen. R. R. Amer. Springer. New York.G. Berlin. 2nd ed.W. Kadison. Cambridge University Press. Benjamin. Kadison. Berlin. 1979] C -Algebras and their Automorphism Groups. Vol. Math. II: Equilibrium States.K. 1968] Induced Representations. Bratteli. and J. G.V. Fields Institute Monographs 6. R. Robinson 1987] Operator Algebras and Quantum Statistical Mechanics. Springer. North-Holland. Oxford. Decomposition of States. 1993] K-theory and C -algebras. J. R. 1-18. Academic Press.M.V. 167. New York. Mackey. 1977] C -Algebras. Mackey. Math. Proc. Amsterdam. 1995] Hilbert C -Modules. pp. Academic Press. New York.M. Springer. Connes. J. Benjamin. Providence (RI). Academic Press. Ringrose 1983] Fundamentals of the Theory of Operator Algebras I. 1978] Induced Representations and Banach -algebra Bundles. R. O. Berlin. Providence (RI). San Diego. Vol. Fell.V. Academic Press. Dixmier. American Mathematical Society.S. 1963] The Mathematical Foundations of Quantum Mechanics. Wegge-Olsen. 2. Doran 1988] Representations of -Algebras.C. Springer. G.K. 1982] Operator algebras . New York. 38(1). Fell. 1994] Noncommutative Geometry. A.V. and R.) Operator Algebras and Applications.R. A Toolkit for Operator Algebraists.W. I: C . Springer. Heidelberg. O. Providence. Boston. Kadison. London.. G.V.LITERATURE 89 Literature Bratteli.the rst forty years.S.E. Pure Math. Ringrose 1986] Fundamentals of the Theory of Operator Algebras II. Soc. 1989] Analysis Now. and D. G. and D. Cambridge. Pedersen. Lance. 21-53. Soc.
This action might not be possible to undo. Are you sure you want to continue?
We've moved you to where you read on your other device.
Get the full title to continue reading from where you left off, or restart the preview.