arXiv:quant-ph/0212023v2 7 Jul 2003 Contents Theory Relativity and Information Quantum I.Terltvsi esrn process measuring relativistic The III. V unu nrp n pca relativity special and entropy Quantum IV. I eodseilrelativity special Beyond VI. I h custo finformation of acquisition The II. .Rltvsi ttstransformations states Relativistic A. .Terl fqatmfil theory field quantum of role The V. .Treisprbetheories inseparable Three I. cnweget n apologies and Acknowledgments .Casclanalogies Classical D. .Aclrtddtcos22 detectors Accelerated D. .Temaigo rbblt 3 probability of meaning The D. .Entanglement D. .Kasmtie n OM 8 POVMs and matrices Kraus D. .Rdcddniymtie 13 matrices density Reduced A. .Rltvt n nomto 1 information and Relativity A. .Etnlmn eiie 24 revisited Entanglement A. .Gnrlproperties General A. .Gnrltheorems General A. .Teabvln unu bevr5 observer quantum ambivalent The A. .Qatmnnoaiy 13 nonlocality? Quantum C. .Etnlmn nqatmfil hoy21 theory field quantum in Entanglement C. .Rltvt n unu hoy3 theory quantum and Relativity C. .Oe problems Open C. .Photons C. .Decoherence C. .Terl frltvt 11 relativity of role The B. .Prilsadlclzto 20 localization and Particles B. .Msieparticles Massive B. .Qatmmcaisadifrain2 information and mechanics Quantum B. .Tetemdnmc fbakhls25 holes black of thermodynamics The B. .Temauigpoes6 process measuring The B. .Terl ftopology of role The E. .Ten-omncto hoe 8 theorem no-communication The E. .Cmuiaincanl 18 channels Communication E. .Teesneo unu nomto 4 information quantum of essence The F. oios r novd oto h urn ocpsi quan in concepts current the reassessment. of a Most when require particular involved. in are cons effects, horizons) a reliabili counterintuitive for detector new, necessary between produces is trade-off which fundamental theory, a field entangled Lor implies Quantum for a matrices not maps. density is tive entropy reduced Quantum of transformations systems. sever distant imposes between ar relativity description mation Special its (POVMs). for quan measures tools valued a Essential from physics. information quantum of and acquisition the The relativity physics. and retical theory, information mechanics, Quantum Ont Waterloo, Physics, Theoretical for Institute Perimeter Terno R. Daniel Techno of and Institute Israel — Technion Physics, of Department Peres Asher 27 27 27 13 13 23 10 10 19 19 18 15 14 4 1 7 5 yadlclzblt.Gnrlrelativity General localizability. and ty ru arcsadpstv operator positive and matrices Kraus e u ytmi h nefc fclassical of interface the is system tum r r h ai onain ftheo- of foundations basic the are ory ytm a o ecmltl posi- completely be not may systems lc oe o oegnrly event generally, more (or holes black etitoso h rnfro infor- of transfer the on restrictions e ro aaaNJ2W9 N2J Canada ario, .TREISPRBETHEORIES INSEPARABLE THREE I. aedffrn ieodrnswe eerdt observers to referred might when events orderings distant time that different and have simultaneity meaning, that that absolute showed or no Einstein sound, had service). of imagine postal speed well the as the of could by in- (one limited were Einstein communications do that Albert not concepts by does new troduced effects fundamentally physical the of to justice propagation of speed the 1991). a (Landauer, physical needs is information Information by because physics. bounded — light carrier is of material signals velocity these The the of spacetime. be- — in transmission events human locate of and as speed label to of detectors. is and role These emitters thought Their physical ordinary be rather signals. but receive not ings, and should send “observers” who observers titious information and Relativity A. related. inseparably are they that this show In shall other? we these each review, do affect How disciplines analyzing disparate methods. seemingly for communication (1948) of theory Shannon efficiency information the Claude later, by electrodynamics years developed the Many was nuclei, bodies. and moving to spectrum, atoms of answers body of black give structure the to the physics: century in twentieth issues unexplained the of beginning u nomto hoymythen may theory information tum .Bakhl radiation hole Black B. setdsrpino interactions, of description istent oee,temr xsec fa pe on on bound upper an of existence mere the However, fic- employ theory relativity of presentations Common the at emerged theory relativity and theory Quantum oy 20 af,Israel Haifa, 32000 logy, nzcvratcnet Lorentz concept. covariant entz References n h atrms bytelw of laws the obey must latter the and , 29 28 c 2 in relative motion. Relativistic kinematics is all about in- The experimenter controls the emission process and formation transfer between observers in relative motion. observes detection events. The theorist’s problem is to Classical information theory involves concepts such as predict the probability of response of this or that de- the rates of emission and detection of signals, and the tector, for a given emission procedure. It often happens noise power spectrum. These variables have well defined that the preparation is unknown to the experimenter, and relativistic transformation properties, independent of the then the theory can be used for discriminating between actual physical implementation of the communication different preparation hypotheses, once the detection out- system. A detailed analysis by Jarett and Cover (1981) comes are known. showed that the transmission rates for observers with rel- ative velocity v were altered by a factor (c + v)/(c v), Quantum mechanics tells us that whatever comes from − the emitter is represented by a state ρ (a positive oper- namely the square of the familiar Doppler factor for fre- 3 quencies of periodic phenomena. We shall later derive ator, usually normalized to unit trace). Detectors are the same factor from classical electromagnetic theory, see represented by positive operators Eµ, where µ is an arbi- trary label which identifies the detector. The probability Eq. (36) below. Physics has a remarkably rigid theoret- ical structure: you cannot alter any part of it without that detector µ be excited is tr (ρEµ). A complete set of E , including the possibility of no detection, sums up having to change everything (Weinberg, 1992). µ to the unit matrix and is called a positive operator val- ued measure (POVM). The various Eµ do not in general commute, and therefore a detection event does not cor- B. Quantum mechanics and information respond to what is commonly called the “measurement of an observable.” Still, the activation of a particular de- Einstein’s theory elicited strong opposition when it was tector is a macroscopic, objective phenomenon. There is proposed, but is generally accepted by now. On the other no uncertainty as to which detector actually clicked. hand, the revolution caused by quantum theory still pro- duces uneasy feelings among some physicists.1 Standard Many physicists, perhaps a majority, have an intuitive texbooks on quantum mechanics tell you that observ- realistic worldview and consider a quantum state as a able quantities are represented by Hermitian operators, physical entity. Its value may not be known, but in prin- their possible values are the eigenvalues of these opera- ciple the quantum state of a physical system would be well defined. However, there is no experimental evidence tors, and that the probability of detecting eigenvalue λn, 2 whatsoever to support this naive belief. On the contrary, corresponding to eigenvector un, is un ψ , where ψ is the (pure) state of the quantum system|h | thati| is observed. if this view is taken seriously, it may lead to bizarre con- With a bit more sophistication to include mixed states, sequences, called “quantum paradoxes.” These so-called paradoxes originate solely from an incorrect interpreta- the probability can be written in a general way un ρ un . h | | i tion of quantum theory. The latter is thoroughly prag- This is nice and neat, but this does not describe what matic and, when correctly used, never yields two contra- happens in real life. Quantum phenomena do not occur dictory answers to a well posed question. It is only the in a Hilbert space; they occur in a laboratory. If you misuse of quantum concepts, guided by a pseudorealistic visit a real laboratory, you will never find there Hermi- philosophy, that leads to paradoxical results. tian operators. All you can see are emitters (lasers, ion guns, synchrotrons, and the like) and appropriate detec- In this review we shall adhere to the view that ρ is tors. In the latter, the time required for the irreversible only a mathematical expression which encodes informa- act of amplification (the formation of a microscopic bub- tion about the potential results of our experimental in- ble in a bubble chamber, or the initial stage of an electric terventions. The latter are commonly called “measure- discharge) is extremely brief, typically of the order of an ments” — an unfortunate terminology, which gives the atomic radius divided by the velocity of light. Once irre- impression that there exists in the real world some un- versibility has set in, the rest of the amplification process known property that we are measuring. Even the very is essentially classical. It is noteworthy that the time and existence of particles depends on the context of our ex- space needed for initiating the irreversible processes are periments. In a classic article, Mott (1929) wrote “Until incomparably smaller than the macroscopic resolution of the final interpretation is made, no mention should be 2 the detecting equipment. made of the α-ray being a particle at all.” Drell (1978) provocatively asked “When is a particle?” In particular, observers whose world lines are accelerated record differ- ent numbers of particles, as will be explained in Sec. V.D 1 The theory of relativity did not cause as much misunderstanding (Unruh, 1976; Wald, 1994). and controversy as quantum theory, because people were care- ful to avoid using the same nomenclature as in nonrelativistic physics. For example, elementary textbooks on relativity the- ory distinguish “rest mass” from “relativistic mass” (hard core relativists call them simply “mass” and “energy”). needed to facilitate the work of the experimenter. 2 The “irreversible act of amplification” is part of the quantum 3 Positive operators are those having the property that hψ|ρ|ψi ≥ 0 folklore, but it is not essential to physics. Amplification is solely for any state ψ. These operators are always Hermitian. 3

C. Relativity and quantum theory tions (the probabilities of specified sets of events) must be Lorentz invariant. The theory of relativity deals with the geometric struc- As a simple example, consider our two observers, con- ture of a four-dimensional spacetime. Quantum mechan- ventionally called Alice and Bob,4 holding a pair of spin- 1 ics describes properties of matter. Combining these two 2 particles in a singlet state. Alice measures σz and finds theoretical edifices is a difficult proposition. For exam- +1, say. This tells her what the state of Bob’s particle ple, there is no way of defining a relativistic proper time is, namely the probabilities that Bob would obtain 1 if ± for a quantum system which is spread all over space. A he measures (or has measured, or will measure) σ along proper time can in principle be defined for a massive any direction he chooses. This is purely counterfactual apparatus (“observer”) whose Compton wavelength is so information: nothing changes at Bob’s location until he small that its center of mass has classical coordinates and performs the experiment himself, or receives a message follows a continuous world-line. However, when there is from Alice telling him the result that she found. In par- more than one apparatus, there is no role for the private ticular, no experiment performed by Bob can tell him proper times that might be attached to the observers’ whether Alice has measured (or will measure) her half of world-lines. Therefore a physical situation involving sev- the singlet. eral observers in relative motion cannot be described by A seemingly paradoxical way of presenting these re- a wave function with a relativistic transformation law sults is to ask the following naive question: suppose that (Aharonov and Albert, 1981; Peres, 1995, and references Alice finds that σz = 1 while Bob does nothing. When therein). This should not be surprising because a wave does the state of Bob’s particle, far away, become the one function is not a physical object. It is only a tool for com- for which σz = 1 with certainty? Though this question − puting the probabilities of objective macroscopic events. is meaningless, it may be given a definite answer: Bob’s Einstein’s principle of relativity asserts that there are particle state changes instantaneously. In which Lorentz no privileged inertial frames. This does not imply the frame is this instantaneous? In any frame! Whatever necessity or even the possibility of using manifestly sym- frame is chosen for defining simultaneity, the experimen- metric four-dimensional notations. This is not a pecu- tally observable result is the same, as can be shown in a liarity of relativistic quantum mechanics. Likewise in formal way (Peres, 2000b). Einstein himself was puzzled classical canonical theories, time has a special role in the by what seemed to be the instantaneous transmission of equations of motion. quantum information. In his autobiography, he wrote the The relativity principle is extraordinarily restrictive. words “telepathically” and “spook” (Einstein, 1949). For example, in ordinary classical mechanics with a fi- Examples like the above one, taken from relativistic nite number of degrees of freedom, the requirement that quantum mechanics, manifestly have an informational the canonical coordinates q have the meaning of posi- nature. We cannot separate the three disciplines: rel- tions, so that particle trajectories q(t) transform like ativity, quantum mechanics, and information theory. four-dimensional world lines, implies that these lines con- sist of straight segments. Long range interactions are for- bidden; there can be only contact interactions between D. The meaning of probability point particles (Currie, Jordan, and Sudarshan, 1963; Leutwyler, 1965). Nontrivial relativistic dynamics re- In this review, we shall often invoke the notion of prob- quires an infinite number of degrees of freedom which ability. Quantum mechanics is fundamentally statistical are labelled by the spacetime coordinates (this is called (Ballentine, 1970). In the laboratory, any experiment a field theory). has to be repeated many times in order to infer a law; in a theoretical discussion, we may imagine an infinite Combining relativity and quantum theory is not only number of replicas of our gedankenexperiment, so as to a difficult technical question on how to formulate dynam- have a genuine statistical ensemble. Yet, the validity of ical laws. The ontologies of these theories are radically the statistical nature of quantum theory is not restricted different. Classical theory asserts that fields, velocities, to situations where there are a large number of similar etc., transform in a definite way and that the equations systems. Statistical predictions do apply to single events. of motion of particles and fields behave covariantly. For When we are told that the probability of precipitation example if the expression for the Lorentz force is written tomorrow is 35%, there is only one tomorrow. This tells f = F uν in one frame, the same expression is valid in µ µν us that it may be advisable to carry an umbrella. Prob- any other frame. These symbols (f , etc.) have objective µ ability theory is simply the quantitative formulation of values. They represent entities that really exist, accord- how to make rational decisions in the face of uncertainty ing to the theory. On the other hand, wave functions are not defined in spacetime, but in a multidimensional Hilbert space. They do not transform covariantly when there are interventions by external agents, as will be seen 4 Alice and Bob joined the quantum information community after in Sec. III. Only the classical parameters attached to a distinguished service in classical cryptography. For example, each intervention transform covariantly. Yet, in spite of they appeared in the historic RSA paper (Rivest, Shamir, and the non-covariance of ρ, the final results of the calcula- Adleman, 1978). 4

(Fuchs and Peres, 2000). A lucid analysis of how prob- “Information theory, as it is understood in abilistic concepts are incorporated into physical theories this paper and as it usually understood by is given by Emch and Liu (2002). mathematicians and engineers following the pioneer paper of Shannon, is not only a the- ory of the entropy concept itself (in this as- E. The role of topology pect information theory is most interesting for physicists), but also a theory of transmis- Physicists often tend to ignore the topological struc- sion and coding of information, i.e., a theory ture of the concepts that they use, or turn to it only as a of information sources and channels.” last resort. Actually, there is a “bewildering” multitude In other words: the goals of quantum information the- of topologies (Reed and Simon, 1980). Many of them ory are the intersection of those of quantum mechan- have a direct physical meaning (Emch 1972; Haag, 1996; ics and information theory, while its tools are the union Araki, 1999). In particular, since measurements can ac- of those of these two theories. Actually, the tools be- tually be performed only with a finite accuracy, a finite longing to quantum theory were developed under the in- number of outcomes, and a finite number of times, only fluence of nascent quantum information, “when it was bounded ranges of values are ever registered. Suppose necessary to consider communication problems for the that we measure N times the value q of an observable Q, needs of quantum of quantum electronic and optics” (In- and a value q is obtained n times. A relative frequency j j garden, 1976). Work of Sudarshan et al. (1961), and w = n /N is either used to extract a probability esti- j j later those of Davies, Kossakowski, Kraus, Lindblad, and mate, or it is taken at face value and interpreted as the Lewis established the formalism of quantum mechanics estimate. Thus the information about a state ρ can be of open systems, expressed by POVMs and completely formulated as (Araki, 1999; Peres and Terno, 1998) positive maps, while works of Helstrom, Holevo, Lebe- dev, and Levitin, produced important results in what pQ(q ) w <ǫ , (1) | ρ j − j | j became quantum information theory.7 We shall discuss these subjects in Sec. II of this review. for some positive ǫ . These inequalities induce a natural j Some trends in modern quantum information theory topology on the space of states, which is called a “physi- may be traced to security problems in quantum com- cal topology” (Emch, 1972; Araki, 1999). More precisely, munication. A very early contribution was Wiesner’s they define a weak-* topology on the observables and a seminal paper Conjugate Coding, which was submitted weak topology on the states. This is a trace-norm topol- circa 1970 to IEEE Transactions on Information Theory, ogy5 (Reed and Simon, 1980). These structures are nat- and promptly rejected because it was written in a jar- urally accommodated in the algebraic approach to quan- gon incomprehensible to computer scientists (this actu- tum theory. That approach consists in the characteriza- ally was a paper about physics, but it had been submitted tion of the theory by a net of algebras of local observ- to a computer science journal). Wiesner’s article was fi- ables, and is especially suited for the analysis of infinite nally published (Wiesner, 1983) in the newsletter of ACM systems in quantum statistical mechanics and quantum SIGACT (Association for Computing Machinery, Special field theory. We will use results based on algebraic field Interest Group in Algorithms and Computation Theory). theory in Sec. V and VI.6 That article tacitly assumed that exact duplication of an unknown quantum state was impossible, well before the no-cloning theorem (Wootters and Zurek, 1982; Dieks, F. The essence of quantum information 1982) became common knowledge. Another early arti- cle, Unforgeable Subway Tokens (Bennett et al., 1983), In an early review of quantum information theory, In- also tacitly assumed the same. garden (1976) distinguished two fundamental aspects: The standard method for quantum cryptography was invented by Bennett and Brassard (1984), using two mutually unbiased bases, namely two bases such that √ 5 um vµ = 1/ d, where d is the number of Hilbert Since probabilities in quantum mechanics are given by the ex- spaceh | dimensions.i Security may be improved by us- pression tr (ρEµ), and physically acceptable states are trace class positive operators, the trace norm topology is the concrete real- ing three bases (Bruß, 1998; Bechmann-Pasquinucci and ization of the physical topology. Gisin, 1999), and even more by going to higher dimen- 6 References whose primary interest is field theory include Bogol- sions (Bechmann-Pasquinucci and Peres, 2000; Bruß and ubov et al. (1990), Haag (1996) and Araki (1999). On the Macchiavello 2002). Gisin, Ribordy, Tittel and Zbinden other hand, Davies (1976), Bratteli and Robinson (1987), and Ingraden, Kossakowski and Ohaya (1997), consider mainly ap- plications to open quantum systems, statistical mechanics and thermodynamics. Emch (1972) is concerned with both. Emch (1972), Bratelli and Robinson (1987), and Baumgartel and Wol- 7 The books of Davies (1976), Holevo (1982), and Ingarden, Kos- lenberg (1992) give a rigorous, and yet readable exposition of the sakowski and Ohaya (1997), contain historical surveys and ex- subject. haustive lists of references. 5

(2002) recently reviewed theoretical and experimental re- have done and what we have learned and that, sults in quantum cryptography. therefore, the account of the experimental ar- A spectacular discovery was that of quantum telepor- rangement and of the results of the observa- tation (Bennett et al.., 1993), which effectively turned tions must be expressed in unambiguous lan- quantum entanglement into a communication resource. guage, with suitable application of the termi- Soon afterwards, it also became a computational resource nology of classical physics.” (Shor, 1994) and since then it continues to attract consid- erable attention. Various aspects of entanglement theory Note the words “we can tell.” Bohr was concerned with information, in the broadest sense of this term. He are reviewed in special issues of Quantum Information and Computation (2001) 1 (1) and Journal of Mathe- never said that there were classical systems or quantum systems. There were physical systems, for which it was matical Physics (2002) 43 (9). Experimental results were reviewed by Zeilinger (1999). appropriate to use the classical language or the quantum language. There is no guarantee that either language Quantum binary channels were introduced by Schu- macher (1995), who also generalized Shannon’s coding gives a perfect description, but in a well designed exper- iment it should be at least a good approximation. theorems to the quantum domain, and coined the word “qubit” (quantum bit) for elementary carriers of quan- Bohr’s approach divides the physical world into “en- dosystems” (Finkelstein, 1988) that are described by tum information. Quantum channels are discussed by Holevo (1999), Amosov, Holevo, and Werner (2000), quantum dynamics, and “exosystems” (such as measur- ing apparatuses) that are not described by the dynam- King and Ruskai (2001), and in the special issue of Jour- ical formalism of the endosystem under consideration. nal of Mathematical Physics (2002) 43 (9). An extensive review of the mathematical aspects of quantum informa- A physical system is called “open” when parts of the universe are excluded from its description. In different tion theory was given by Keyl (2002). Our review deals with many interrelated issues. Lorentz frames used by observers in relative motion, dif- ferent parts of the universe may be excluded. The sys- Causality constraints on POVMs are discussed in Sec. II.E. Relativistic extensions of the formalism appear tems considered by these observers are then essentially different, and no Lorentz transformation exists that can in Sec. III and VI.A. In Sec. IV we discuss how relativistic considerations modify basic notions of quantum informa- relate them (Peres and Terno, 2002). It is noteworthy that Bohr never described the measur- tion theory: qubits, entanglement, and quantum chan- nels. In Sec. V we investigate the implications of quan- ing process as a dynamical interaction between an exo- physical apparatus and the system under observation. He tum field theory on the construction of POVMs and the detection of entanglement. Section VI.A deals with rel- was of course fully aware that measuring apparatuses are made of the same kind of matter as everything else, and ativistic extensions of quantum information theory, and in Sec. VI.B we discuss its applications to the black hole they obey the same physical laws. It is therefore tempt- ing to use quantum theory in order to investigate their physics. behavior during a measurement. However, if this is done, the quantized apparatus loses its status of a measuring II. THE ACQUISITION OF INFORMATION instrument. It becomes a mere intermediate system in the measuring process, and there must still be a final in- A. The ambivalent quantum observer strument that has a purely classical description (Bohr, 1939). Quantum mechanics is used by theorists in two differ- Measurement was understood by Bohr as a primitive ent ways: it is a tool for computing accurate relation- notion. He could thereby elude questions which caused ships between physical constants, such as energy levels, considerable controversy among other authors. A quan- cross sections, transition rates, etc. These calculations tum dynamical description of the measuring process was are technically difficult, but they are not controversial. first attempted by John von Neumann, in his treatise on Besides this, quantum mechanics also provides statisti- the mathematical foundations of quantum theory (1932). cal predictions for results of measurements performed on In the last section of that book, as in an afterthought, physical systems that have been prepared in a specified von Neumann represented the apparatus by a single de- way. The quantum measuring process is the interface of gree of freedom, whose value was correlated to that of the classical and quantum phenomena. The preparation and dynamical variable being measured. Such an apparatus measurement are performed by macroscopic devices, and is not, in general, left in a definite pure state, and it does these are described in classical terms. The necessity of not admit a classical description. Therefore, von Neu- using a classical terminology was emphasized by Niels mann introduced a second apparatus which observes the Bohr (1927) since the very early days of quantum me- first one, and possibly a third apparatus, and so on, until chanics. Bohr’s insistence on a classical description was there is a final measurement, which is not described by very strict. He wrote (1949): quantum dynamics and has a definite result (for which quantum mechanics can only give statistical predictions). “. . . by the word ‘experiment’ we refer to a The essential point that was suggested, but not proved by situation where we can tell others what we von Neumann, is that the introduction of this sequence 6 of apparatuses is irrelevant: the final result is the same, by classical information received from past interventions, irrespective of the location of the “cut” between classical or they may be chosen arbitrarily by the observer who and quantum physics.8 prepares that intervention, or by a local random device These different approaches of Bohr and von Neumann acting in lieu of the observer. were reconciled by Hay and Peres (1998), who introduced An intervention has two consequences. One is the ac- a dual description for the measuring apparatus. It obeys quisition of information by means of an apparatus that quantum mechanics while it interacts with the system produces a record. This is the “measurement.” Its out- under observation, and then it is “dequantized” and is come, which is in general unpredictable, is the output described by a classical Liouville density which provides of the intervention. The other consequence is a change the probability distribution for the results of the mea- of the environment in which the quantum system will surement. Alternatively, the apparatus may always be evolve after completion of the intervention. For example treated by quantum mechanics, and be measured by a the intervening apparatus may generate a new Hamilto- second apparatus which has such a dual description. The nian which depends on the recorded result. In particular, question raised by Hay and Peres is whether these two classical signals may be emitted for controlling the execu- different methods of calculation give the same result, or tion of further interventions. These signals are of course at least asymptotically agree under suitable conditions. limited to the velocity of light. They showed that a sufficient condition for agreement The experimental protocols that we consider all start between the two methods is that the dynamical variable in the same way, with the same initial state ρ0, and the used as a “pointer” by the first apparatus be represented first intervention is the same. However, later stages of the by a “quasi-classical” operator of the Weyl-Wigner type experiment may involve different types of interventions, (Hillery et al., 1984). possibly with different spacetime locations, depending on To avoid any misunderstanding, we emphasize that the the outcomes of the preceding events. Yet, assuming that classical description of a pointer is not by means of a each intervention has only a finite number of outcomes, point in phase space, but by a Liouville density. Quan- there is for the entire experiment only a finite number tum theory makes only statistical predictions, and any of possible records. (Here, the word “record” means the semiclassical treatment that simulates it must also be complete list of outcomes that occurred during the exper- statistical. iment. We do not want to use the word “history” which has acquired a different meaning in the writings of some quantum theorists.) B. The measuring process Each one of these records has a definite probability in the statistical ensemble. In the laboratory, experimenters Dirac (1947) wrote “a measurement always causes the can observe its relative frequency among all the records system to jump into an eigenstate of the dynamical vari- that were obtained; when the number of records tends able being measured.” Here, we must be careful: a quan- to infinity, this relative frequency is expected to tend to tum jump (also called collapse) is something that hap- the true probability. The aim of theory is to predict the pens in our description of the system, not to the system probability of each record, given the inputs of the vari- itself. Likewise, the time dependence of the wave func- ous interventions (both the inputs that are actually con- tion does not represent the evolution of a physical sys- trolled by the local experimenter and those determined tem. It only gives the evolution of probabilities for the by the outputs of earlier interventions). Each record is outcomes of potential experiments on that system (Fuchs objective: everyone agrees on what happened (e.g., which and Peres, 2000). detectors clicked). Therefore, everyone agrees on what the various relative frequencies are, and the theoretical Let us examine more closely the measuring process. probabilities are also the same for everyone. First, we must refine the notion of measurement and extend it to a more general one: an intervention. An Interventions are localized in spacetime, but quantum intervention is described by a set of parameters which systems are pervasive. In each experiment, irrespective include the location of the intervention in spacetime, re- of its history, there is only one quantum system, which ferred to an arbitrary coordinate system. We also have to may consist of several particles or other subsystems, cre- specify the speed and orientation of the apparatus in the ated or annihilated at the various interventions. Note coordinate system that we are using, and various other that all these properties still hold if the measurement input parameters that control the apparatus, such as the outcome is the absence of a detector click. It does not strength of a magnetic field, or that of an rf pulse used matter whether this is due to an imperfection of the de- in the experiment. The input parameters are determined tector or to a probability < 1 that a perfect detector would be excited. The state of the quantum system does not remain unchanged. It has to change to respect uni- tarity. The mere presence of a detector that could have 8 At this point, von Neumann also speculated that the final step been excited implies that there has been an interaction involves the consciousness of the observer — a bizarre statement between that detector and the quantum system. Even if in a mathematically rigorous monograph (von Neumann, 1955). the detector has a finite probability of remaining in its 7 initial state, the quantum system correlated to the latter C. Decoherence acquires a different state (Dicke, 1981). The absence of a click, when there could have been one, is also an event. Up to now, the quantum evolution is well defined and Interventions, as defined above, start by an interaction it is in principle reversible. It would remain so if the envi- with a measuring apparatus, called “premeasurement” ronment could be perfectly isolated from the macroscopic (Peres, 1980). The quantum system and the apparatus degrees of freedom of the apparatus. This demand is of are initially in a state c s A , and become entan- s s | i⊗ | i course self-contradictory, since we have to read the re- gled into a single compositeP system : C sult of the measurement if we wish to make any use of it. A detailed analysis of the interaction with the environ- c s A c U λ , (2) ment, together with plausible hypotheses (Peres, 2000a), s | i ⊗ | i → s sλ | i Xs Xs,λ shows that states of the environment that are correlated to subspaces of with different labels µ can be treated C where λ is a complete basis for the states of . It as if they were orthogonal. This is an excellent approx- {| i} C imation (physics is not an exact science, it is a science is the choice of the unitary matrix Usλ that determines which property of the system under study is correlated to of approximations). The resulting theoretical predictions the apparatus, and therefore is measured. When writing will almost always be correct, and if any rare small de- the above equation, we tacitly assumed that the quantum viation from them is ever observed, it will be considered system and the measuring apparatus were initially in a as a statistical quirk, or an experimental error. pure state. Since a mixed state is a convex combination The density matrix of the quantum system thus is ef- of pure states, no new feature can result from taking fectively block-diagonal and all our statistical predictions mixed states (which would admittedly be more realistic). are identical to those obtained for an ordinary mixture Relativistic restrictions on the allowed forms of Usλ will of (unnormalized) pure states be discussed in Sec. III. The measuring process involves not only the physical system under study and a measuring apparatus (which ψ = c U µ, ξ , (3) | µi s sµξ | i together form the composite system ) but also their “en- Xs,ξ vironment” which includes unspecifiedC degrees of free- dom of the apparatus and the rest of the world. These unknown degrees of freedom interact with the relevant where the statistical weight of each state is the square ones, but they are not under the control of the experi- of its norm. This process is called decoherence. Each menter and cannot be explicitly described. Our partial subspace µ is stable under decoherence — it is their rel- ignorance is not a sign of weakness. It is fundamental. If ative phase that decoheres. From this moment on, the everything were known, acquisition of information would macroscopic degrees of freedom of have entered into the be a meaningless concept. classical domain. We can safely observeC them and “lay A complete description of involves both macroscopic on them our grubby hands” (Caves, 1982). In particu- and microscopic variables. TheC difference between them lar, they can be used to trigger amplification mechanisms is that the environment can be considered as adequately (the so-called detector clicks) for the convenience of the isolated from the microscopic degrees of freedom for the experimenter. duration of the experiment and is not influenced by them, Some authors claim that decoherence may provide a the environment is not isolated from the macro- while solution of the “measurement problem,” with the partic- scopic degrees of freedom . For example, if there is a ular meaning that they attribute to that problem (Zurek, macroscopic pointer, air molecules bounce from it in a 1991). Others dispute this point of view in their com- way that depends on the position of that pointer. Even if ments on the above article (Zurek, 1993). A reassessment we can neglect the Brownian motion of a massive pointer, of this issue and many important technical details were its influence on the environment leads to the phenomenon recently published by Zurek (2002, 2003). Yet, decoher- of decoherence, which is inherent to the measuring pro- ence has an essential role, as explained above. It is es- cess. sential to distinguish decoherence, which results from the An essential property of the composite system , which C disturbance of the environment by the apparatus (and is is necessary to produce a meaningful measurement, is a quantum effect), from noise, which would result from that its states form a finite number of orthogonal sub- the disturbance of the system or the apparatus by the en- spaces which are distinguishable by the observer. Each vironment and would cause errors. Noise is a mundane macroscopically distinguishable subspace corresponds to classical phenomenon, which we ignore in this review.9 one of the outcomes of the intervention and defines a POVM element Eµ, given explicitly by Eq. (8) below. Let us therefore introduce a complete basis for , namely µ, ξ , where µ labels a macroscopic subspace,C and ξ la- 9 {| i} The so-called “quantum noise” which is discussed in Sec. IV.C bels microscopic states in that subspace. has a different nature. 8

D. Kraus matrices and POVMs completely positive. If we have two systems, it is physi- cally meaningless to reverse the direction of time for only The final step of the intervention is to discard part of one of them. One can write a formal expression for this the composite system . The discarded part may de- impossible process, but the resulting “density matrix” pend on the outcome µC. We therefore introduce in the is unphysical because it may have negative eigenvalues subspace µ two sets of basis vectors µ, σ and µ, m for (Peres, 1996). The case for consideration of completely the new system and the part that is| discarded,i | respec-i positive maps was made by Kraus (1971), Davies (1976) tively. We thus obtain for the new system a reduced and Lindblad (1976), and since than they are part of the density matrix toolbox of quantum information. In Sec. IV.E we discuss apparent exceptions to this approach.

(ρµ′ )στ = (Aµm)σs ρst (Aµm∗ )τt, (4) It follows from Eq. (6) that the probability of occur- Xm Xs,t rence of outcome µ is where ρst csct∗ is the initial state, and the notation pµ = tr (Aµm ρ Aµm† ) = tr (ρEµ). (7) ≡ Xm (A ) U , (5) µm σs ≡ sµσm The positive operators was introduced for later convenience. Recall that the Eµ = Aµm† Aµm, (8) indices s and σ refer to the original system under study Xm and to the final one, respectively. Omitting these indices, Eq. (4) takes the familiar form whose dimensions are the same as those of the initial ρ, satisfy µ Eµ = 1l owing to the unitarity of Usµσm. ThereforeP they are the elements of a POVM. Conversely, ρ ρµ′ = Aµm ρ Aµm† , (6) → given Eµ (a positive matrix of order k) it is always pos- Xm sible to split it in infinitely many ways as in the above where µ is a label that indicates which detector was in- equation. volved and the label m refers to any subsystem that was In the special case where the POVM elements Eµ com- discarded at the conclusion of the interaction. Clearly, mute, they are orthogonal projection operators, and the POVM becomes a projection valued measure (PVM). the “quantum jump” ρ ρµ′ is not a dynamical pro- cess that occurs in the quantum→ system by itself. It re- The corresponding intervention is sometimes called a sults from the introduction of an apparatus, followed by von Neumann measurement. Rigorous treatment of the its deletion or that of another subsystem. A jump in POVM formalism can be found in the books of Davies the quantum state occurs even when there is no detector (1976), Holevo (1982), and Kraus (1983). click or other macroscopic amplification, because we im- pose abrupt changes in our way of delimiting the object that we consider as the quantum system under study. E. The no-communication theorem The initial ρ is usually assumed to be normalized to We now derive a sufficient condition that no instan- unit trace, and the trace of ρ′ is the probability of oc- µ taneous information transfer can result from a distant currence of outcome µ. Note that each symbol Aµm in the above equation represents a matrix (not a matrix intervention. We shall show that the condition is element). Explicitly, the Kraus operators Aµm (Kraus, [Aµm,Bνn]=0, (9) 1983) are given by Eq. (5), where Usµσm is the matrix element for the unitary interaction between the system where Aµm and Bνn are Kraus matrices for the observa- under study and the apparatus, including any auxiliary tion of outcomes µ by Alice and ν by Bob. Indeed, the systems that are subsequently discarded (Peres, 2000a). probability that Bob gets a result ν, irrespective of what Alice found, is Equation (6) is sometimes written ρµ′ = ρ, where is a linear superoperator which acts on densityS matricesS like ordinary operators act on pure states. Note however pν = tr Bνn Aµm ρ Aµm† Bνn† . (10)   that these superoperators have a very special structure, Xµ Xm,n explicitly given by Eq. (6). We now make use of Eq. (9) to exchange the positions It is noteworthy that Eq. (6) is the most general com- of Aµm and Bνn, and likewise those of Aµm† and Bνn† , pletely positive linear map (Stinespring, 1955; Davies, and then we move Aµm from the first position to the last 1976; Kraus, 1983). This is a crucial property: a linear one in the product of operators in the traced parenthesis. map T (ρ) is called positive if it transforms any positive We thereby obtain expressions as in Eq. (8). These are matrix ρ (namely, one without negative eigenvalues) into elements of a POVM that satisfy Eµ = 1l. Therefore completely positive µ another positive matrix. It is called if Eq. (10) reduces to P (T 1l) acting on a bipartite ρ produces a valid bipartite ρ. For⊗ instance, complex conjugation of ρ (whose mean- pν = tr Bνn ρBνn† , (11) ing is time reversal) is a positive map. However, it is not Xn  9

1 whence all expressions involving Alice’s operators Aµm Φ± := ( 0 0 1 1 ). (14) have totally disappeared. The statistics of Bob’s result | i √2 | i| i ± | i| i are not affected at all by what Alice may simultaneously Other classes of bipartite operations are defined as fol- do somewhere else. This proves that Eq. (9) indeed is lows: Bob performs a local operation T just before the a sufficient condition for no instantaneous information B global operation T . If no local operation of Alice can transfer.10 reveal any information about T , i.e., Bob cannot signal Note that any classical communication between distant B to Alice, then the operation T is semicausal. If the oper- observers can be considered as a kind of long range inter- ation is semicausal in both directions, it is called causal. action. Indeed, it is always possible to treat their appa- In many cases it is easier to prove causality than local- ratuses as quantum systems (von Neumann, 1932; Bohr, izability. To check the causality of an operation T whose 1939) and then any signals that propagate between these outcomes are the states ρ = T (ρ)/p with probabilities apparatuses are a manifestation of their mutual interac- µ µ µ p = tr T (ρ), it is enough to consider the corresponding tion. The propagation of signals is of course bounded by µ µ superoperator the velocity of light. As a result, there exists a partial time ordering of the various interventions in an exper- T ′(ρ) := Tµ(ρ). (15) iment, which defines the notions earlier and later (we Xµ assume that there are no closed causal loops). The input parameters of an intervention are deterministic (or pos- Indeed, assume that Bob’s action prior to the global op- sibly stochastic) functions of the parameters of earlier in- eration leads to one of the two different states ρ1 and ρ2. terventions, but not of the stochastic outcomes resulting Then the states T ′(ρ1) and T ′(ρ2) are distinguishable from later or mutually spacelike interventions (Blanchard if and only if some of the pairs of states Tµ(ρ1)/pµ1 and and Jadczik, 1996 and 1998; Percival, 1998). Tµ(ρ2)/pµ2 are distinguishable. Such probabilistic distin- Even these apparently simple notions lead to non- guishability shows that the operation T is not semicausal. trivial results. Consider a separable bipartite superop- These definitions of causal and localizable operators ap- erator T , pear equivalent. It is easily proved that localizable oper- ators are causal. It was shown that semicausal operators T (ρ)= M ρM †,M = A B , (12) are always semilocalizable (Eggeling, Schlingemann, and k k k k ⊗ k Xk Werner, 2002). However, there are causal operations that are not localizable (Beckman et al., 2001). where the operators Ak represent operations of Alice, It is curious that while a complete Bell measurement et al. and Bk those of Bob. It was shown by Bennett is causal, the two-outcome incomplete Bell measurement (1999) that not all such superoperators can be imple- is not (Sorkin, 1993). Indeed, consider a two-outcome mented by local transformations and classical communi- PVM cation (LOCC). For more on this subject, see Walgate + + and Hardy (2002). E1 = Φ Φ , E2 = 1l E1, (16) A classification of bipartite state transformations was | ih | − introduced by Beckman et al. (2001). It consists of where Φ+ = ( 00 + 11 )/√2 (and the Kraus matri- | i | i | i the following categories. There are localizable opera- ces are the projectors Eµ themselves). If the initial state tions that can be implemented locally by Alice and Bob, is 01 AB, then the outcome that is associated with E2 | i possibly with the help of prearranged entangled auxil- always occurs and Alice’s reduced density matrix after iary systems (ancillas), but without classical comunica- the measurement is ρA = 0 0 . On the other hand, | ih | tion. Ideally, local operations are instantaneous, and the if before the joint measurement Bob performs a unitary whole process can be viewed as performed at a definite operation that transforms the state into 00 AB, then the | i time. For semilocalizable operations, the requirement of two outcomes are equiprobable, the resulting states after no communication is relaxed and one-way classical com- the measurement are maximally entangled, and Alice’s 1 munication is possible. It is obvious that any tensor- reduced density matrix is ρA = 2 1l. It can be shown that product operation T T is localizable. The converse two input states 00 AB and 01 AB after this incomplete A B | i | i is not always true, for⊗ example in Bell measurements Bell measurement are distinguished by Alice with a prob- (Braunstein, Mann, and Revzen, 1992) which distinguish ability of 0.75. between the four standard bipartite entangled states, Here is another example of a semicausal and semilocal- izable measurement which can be executed with one-way 1 Ψ± := ( 0 1 1 0 ), (13) classical communication from Alice to Bob. Consider a | i √2 | i| i ± | i| i PVM measurement, whose complete orthogonal projec- tors are 0 0 , 0 1 , 1 + , 1 , (17) 10 An algebraic approach to statistical independence and to related | i ⊗ | i | i ⊗ | i | i ⊗ | i | i ⊗ |−i topics is discussed by Florig and Summers (1997), while Neu- where = ( 0 1 )/√2. The Kraus matrices are mann and Werner (1983) specifically address the issue of causal- |±i | i ± | i ity between preparation and registration processes. Aµj = Eµδj0. (18) 10

From the properties of complete orthogonal measure- state Ψ− and, contrary to the scheme of Beckman et al. ments (Beckman et al., 2001), it follows that this opera- (2001),| theyi do not have to coordinate their moves. Alice tion cannot be performed without Alice talking to Bob. and Bob perform their tasks independently and convey A protocol to realize this measurement is the following. their results to a common center, where the final analy- Alice measures her qubit in the basis 0 , 1 , and tells sis is made. In the first step of this measurement, Alice her result to Bob. If Alice’s outcome{| wasi |0i}, Bob mea- performs a Bell measurement as in the teleportation of a sures his qubit in the basis 0 , 1 , and if| iti was 1 , in state Ψ from her site to Bob (see below). However, Al- the basis + , . {| i | i} | i ice and| Bobi do not perform the full teleportation which Beckman{| eti al.|−i}(2001) derived necessary and sufficient requires a classical communication between them. The conditions to check the semicausality (and therefore, the second step of the verification is executed by Bob. He causality) of PVM measurements. Groisman and Reznik measures the spin of his particle in the z direction. Ac- (2002) allowed for more complicated conditional state cording to whether that spin is up or down, he measures evolutions. In particular, they were interested in verifica- the spin of his ancilla in the z or x direction, respectively. tion measurements, i.e., those yielding µ with certainty This completes the measurement and it only remains to if the state prior to the classical intervention is ρ Eµ, combine the local outcomes to get the result of the non- but without making any specific demand on the resulting∝ local measurement (Vaidman, 2003). This method can state ρµ′ . They showed that all PVM verification mea- be extended to arbitrary Hilbert space dimensions. surements on 2 2 dimensional systems are localizable. × Vaidman (2003) proposed a realization of verification In the teleportation of an unknown state Ψ 0 of a 1 | i measurements by means of a shared entangled ancilla, spin- 2 particle located at Alice’s site, Alice and Bob use and Bell-type measurements by one of the parties. A a prearranged pair in a singlet state, namely Ψ− 12 = | i verification measurement of the states in Eq. (17) will ( 0 1 1 2 1 1 0 2)/√2. The procedure is based on the illustrate his construction. Alice and Bob share a Bell identity| i | i (Bennett− | i | i et al., 1993)

1 + ˜ (z) ˜ (x) + ˜ (y) Ψ 0 Ψ− 12 = 2 Ψ− 01 Ψ 2 + Ψ 01 Ψ 2 + Φ− 01 Ψ 2 + Φ 01 Ψ 2 , (19) | i | i | i | i | i | i | i | i | i | i  where the four Bell states are given by Eqs. (13) and III. THE RELATIVISTIC MEASURING PROCESS (14), and the symbol Ψ˜ (z) means the state Ψ rotated by π around the z-axis,| etc.i Thus, the Bell measurement| i A. General properties performed on the two particles at Alice’s site leads to one of the branches of the superposition on the rhs of Quantum measurements are usually considered as Eq. (19). To complete the teleportation, Bob performs quasi-instantaneous processes. In particular, they affect a rotation by π around one of the axes according to the the wave function instantaneously throughout the en- classical information he gets from Alice. tire configuration space. Measurements of finite duration (Peres and Wootters, 1985) make no essential difference Gauge theories also lead to interesting questions about in this respect. Is this quasi-instantaneous change of the measurability. Wilson loops, which are nonlocal objects quantum state, caused by a local intervention of an ex- by definition, are often invoked in their presentation (Pe- ophysical agent, consistent with relativity theory? The skin and Schroeder, 1995) and are the backbone of lattice answer is not obvious. The wave function itself is not gauge theories (Makeenko, 2002). Beckman et al. (2002) a material object forbidden to travel faster than light, investigated the measurability of the Wilson loop opera- but we may still ask how the dynamical evolution of an tors. extended quantum system that undergoes several mea- surements in distant spacetime regions is described in The impossibility of instantaneous communication al- different Lorentz frames. lows to circumvent the theoretical impossibility of quan- Difficulties were pointed out long ago by Bloch (1967), tum bit commitment (Mayers, 1997; Lo and Chau, 1997). Aharonov and Albert (1981, 1984), and many others Kent (1999, 2003) developed protocols based on the finite (Peres, 1995 and references therein). Still before them, speed of communication and evaluated their communica- in the very early years of quantum mechanics, Bohr and tion costs and security. In particular Kent’s RBC2 proto- Rosenfeld (1933) had given a complete relativistic theory col allows a bit commitment to be indefinitely maintained of the measurement of quantum fields, but these authors with unconditional security against all classical attacks, were not concerned about the properties of the new quan- and at least for some finite amount of time against quan- tum states that resulted from these measurements and tum attacks (Kent, 2003). their work does not answer the question that was raised 11 above. Other authors (Scarani et al., 2000; Zbinden et uinely different prescriptions for choosing the sequence al., 2001) considered detectors in relative motion, and of mathematical operations in our calculation. Therefore therefore at rest in different Lorentz frames. These works these different orderings ought to give the same set of also do not give an explicit answer to the above ques- probabilities, and this demand is not trivial. tion: a detector in uniform motion is just as good as one that has undergone an ordinary spatial rotation. (Ac- celerated detectors involve new physical phenomena, see t1 t2 Sec. V.D.) The point is not how individual detectors happen to move, but how the effects due to these detec- tors are described in different ways in one Lorentz frame or another. Testof Testof S S To become fully relativistic, the notion of intervention 1x 2y requires some refinement. The precise location of an in- tervention, which is important in a relativistic discussion, is the point from which classical information is sent that may affect the input of other interventions. More pre- cisely, it is the earliest small region of spacetime from which classical information could have been sent. More- over, in the conventional presentation of non-relativistic quantum mechanics, each intervention has a (finite) num- FIG. 1 In this spacetime diagram, the origins of the coordi- ber of outcomes, for example, this or that detector clicks. nate systems are the locations of the two tests. The t1 and In a relativistic treatment, the spatial separation of the t2 axes are the world lines of the observers, who are receding detectors is essential and each detector corresponds to a from each other. In each Lorentz frame, the z1 and z2 axes different intervention. The reason is that if several de- are isochronous: t1 =0 and t2 = 0, respectively. tectors are set up so that they act at a given time in one Lorentz frame, they would act at different times in an- other Lorentz frame. However, a knowledge of the time ordering of events is essential in our dynamical calcula- B. The role of relativity tions, so that we want the parameters of an intervention to refer unambiguously to only one time (indeed to only A typical example of relativistic measurement is the one spacetime “point”). Therefore, an intervention can detection system in the experimental facility of a mod- involve only one detector and it can have only two possi- ern high energy accelerator. Following a high energy col- ble outcomes: either there was a “click” or there wasn’t. lision, thousands of detection events occur in locations What is the role of relativity theory here? We may that may be mutually space-like. Yet, some of the detec- likewise ask what is the role of translation and/or rota- tion events are mutually time-like, for example when the tion invariance in a nonrelativistic theory. The point is world line of a charged particle is recorded in an array that the rules for computing quantum probabilities in- of wire chambers. In a relativistic context, the term “de- volve explicitly the spacetime coordinates of the inter- tector” strictly means an elementary detecting element, ventions. Lorentz invariance (or rotational invariance, as such as a bubble in a bubble chamber, or a small segment a special case) says that if the classical spacetime coordi- of wire in a wire chamber.11 nates are subjected to a particular linear transformation, A much simpler example of space-like separated in- then the probabilities remain the same. This invariance terventions, which is amenable to a complete analysis, is not trivial because the rule for computing the proba- is Bohm’s version of the Einstein-Podolsky-Rosen “para- bility of occurrence of a given record involves a sequence dox” (hereafter EPRB; Einstein, Podolsky, and Rosen, of mathematical operations corresponding to the time or- 1935; Bohm 1951) which is sketched in Fig. 1, with two dered set of all the relevant interventions. coordinate systems in relative motion (Peres, 1993). In 1 If we only consider the Euclidean group, all we have that experiment, a pair of spin- 2 particles, prepared in to know is how to transform the classical parameters, a singlet state, move apart and are detected by two ob- and the wave function, and the various operators, under servers. Each observer measures a spin component along translations and rotations of the coordinates. However, an arbitrarily chosen direction. The two interventions are when we consider genuine Lorentz transformations, we have not only to Lorentz-transform the above symbols, but we are faced with a new problem: the natural way of calculating the result of a sequence of interventions, 11 High energy physicists use a different language. For them, an namely by considering them in chronological order, is dif- “event” is one high energy collision together with all the subse- quent detections that are recorded. This “event” is what we call ferent for different inertial frames. The issue is not only a here an experiment (while they call “experiment” the complete matter of covariance of the symbols at each intervention experimental setup that may be run for many months). And and between consecutive interventions. There are gen- their “detector” is a huge machine weighing thousands of tons. 12

mutually space-like as shown in the figure. The test of measurement-induced state transformation that can be S1x occurs first when recorded in t1-time, and the test of written in the Kraus form is permitted or makes sense. S2y is the first one in t2-time. The evolution of the quan- Relativity theory prohibits superluminal velocity for ma- tum state of this bipartite system appears to be genuinely terial objects. Consistency with the requirements of co- different when recorded in two Lorentz frames in relative variance and causality is an intrinsic feature of quantum motion. The quantum states are not Lorentz-transforms field theories. Nevertheless, to make problems solvable, of each other. Yet, all the observable results are the same. a patchwork of relativistic and non-relativistic theories Consistency of the theoretical formalism imposes definite is employed. For example, a measurement on relativis- relationships between the various operators used in the tic systems is usually treated by introducing detectors calculations (Peres, 2000b). In particular, it is sufficient that are described by non-relativistic quantum mechan- for consistency that the Kraus operators satisfy equal- ics. Often these detectors are stripped to only a few dis- time commutation relation as in Eq. (9). The analogy crete degrees of freedom (Unruh and Wald, 1984; Levin, with relativistic quantum field theory is manifest. Peleg and Peres, 1992; Wald, 1994). In general, consider the quantum evolution from an An external probe which is not described by field the- initial state ρ0 to a final state ρf . It is a completely ory and whose coupling to the fields of interest is arbitrar- positive map, ily adjustable is obviously an idealization. Beckman et al. (2001) assert that if the probe variables are “heavy,” ρf = An ρ0 An† . (20) with rapidly decaying correlations and the field variables Xn are “light,” then this idealization is credible. Still, causal- The Lorentz transformation of the Kraus matrices An ity requirements like the absence of signalling should be can be obtained as follows. We have ρ0′ = Uρ0U † and checked for any proposed measurement scheme (Sec. II.E ρf′ = Vρf V †, where U and V are unitary representations also discusses causality requirements). of Lorentz transformations for the systems represented Consider again the descriptions of the EPRB by ρ0 and ρf (which may be of different nature and even gedankenexperiment in two coordinate systems in rela- of different dimensions). tive motion. There exists a Lorentz transformation con- Lorentz invariance means that, in another frame, the necting the initial states ρ0 and ρ0′ before the two inter- Kraus matrices An′ satisfy ventions, and likewise there is a Lorentz transformation connecting the final states ρ and ρ after completion ρ = A ρ A . (21) f f′ f′ n′ 0′ n′† of the two interventions. On the other hand, there is Xn no Lorentz transformation relating the states at inter- A simple solution is mediate times that are not in the past or future of both interventions (Peres, 2000b). The various Kraus opera- A = V A U , (22) n′ n † tors, acting at different times, appear in different orders. but this is not the most general one. The latter is Nevertheless the overall transition from initial to final state is Lorentz invariant (Peres, 2001). m An′ = Wn V Am U †, (23) In the time interval between the two interventions, Xm nothing actually happens in the real world. It is only m where Wn is a unitary matrix that acts on the labels m,n in our mathematical calculations that there is a deter- (not on the Hilbert spaces of ρ0 and ρf ). This arbitrari- ministic evolution of the state of the quantum system. ness is a kind of gauge freedom, and can be resolved only This evolution is not a physical process.13 What distin- by a complete dynamical description of the intervention guishes the intermediate evolution between interventions process. This, however, is an arduous problem. Rela- from the one occurring at an intervention is the unpre- tivistic interactions necessarily involve field theory, and dictability of the outcome of the latter: either there is the question is how to generalize the quantum informa- a click or there is no click of the detector. This un- tion tools (POVMs, completely positive maps) into ob- predictable macroscopic event starts a new chapter in jects that are described by quantum field theories (Terno 2002). At this stage we consider only field theories in Minkowski spacetime where a unique vacuum state Ω quantities”) are finite real numbers is sufficient to construct prob- is defined. The discrete indices that appear in the above| i ability measures. For the exact formulation see Davies (1976) and Holevo (1982). Similar arguments justify the inclusion of equations can still be used, owing to the fact that the un- only bounded operators into algebras of local observables (Haag, derlying Hilbert space is separable (Streater and Wight- 1996; Araki, 1999). man, 1964). Therefore the formalism is valid without 13 Likewise, the quantum state of Schr¨odinger’s legendary cat, change in the relativistic domain.12 However, not every doomed to be killed by an automatic device triggered by the de- cay of a radioactive atom, evolves into a superposition of “live” and “dead” states. This is a manifestly absurd situation for a real cat. The only meaning that such a quantum state can have is that of a mathematical tool for statistical predictions on the 12 The fact that the values of classical parameters (“measurable fates of numerous cats subjected to the same cruel experiment. 13 the history of the quantum system which acquires a new sen components of J1 and J2. (They can measure all state, according to Eq. (6). the components, since these have objective values.) Yet, Bob’s measurement tells him nothing of what Alice did, nor even whether she did anything at all. He can only C. Quantum nonlocality? know with certainty what would be the result found by Alice if she measures her J along the same direction as Phenomena like those illustrated in Fig. 1 are often at- him, and make statistical inferences for other possible tributed to “quantum nonlocality” and have led some directions of Alice’s measurement. authors to speculate on the possibility of superlumi- The classical-quantum analogy becomes complete if we nal communication (actually, instantaneous communica- use classical statistical mechanics. The distribution of tion). One of these proposals (Herbert, 1981) looked rea- bomb fragments is given by a Liouville function in phase sonably serious and arose enough interest to lead to inves- space. When Alice measures J1, the Liouville function tigations disproving this possibility (Glauber, 1986) and for J2 is instantly altered, however far Bob is from Al- in particular to the discovery of the no-cloning theorem ice. No one finds this surprising, since it is universally (Wootters and Zurek, 1982; Dieks, 1982). Let us examine agreed that a Liouville function is only a mathematical more closely the origin of these claims of nonlocality. tool representing our statistical knowledge. Likewise, the Bell’s theorem (1964) asserts that it is impossible to wave function ψ, or the corresponding Wigner function mimic quantum theory by introducing a set of objec- (Wigner, 1932) which is the quantum analogue of a Li- tive local “hidden” variables. It follows that any classical ouville function, are no more than mathematical tools imitation of quantum mechanics is necessarily nonlocal. for computing probabilities. It is only when they are re- However Bell’s theorem does not imply the existence of garded as physical objects that superluminal paradoxes any nonlocality in quantum theory itself. In particular arise. relativistic quantum field theory is manifestly local. The The essential difference between the classical and quan- simple and obvious fact is that information has to be tum functions which change instantaneously as the result carried by material objects, quantized or not. Therefore of measurements is that the classical Liouville function quantum measurements do not allow any information to is attached to objective properties that are only imper- be transmitted faster than the characteristic velocity that fectly known. On the other hand, in the quantum case, appears in the Green’s functions of the particles emitted the probabilities are attached to potential outcomes of in the experiment. In a Lorentz invariant theory, this mutually incompatible experiments, and these outcomes limit is the velocity of light. do not exist “out there” without the actual interventions. In summary, relativistic causality cannot be violated Unperformed experiments have no results. by quantum measurements. The only physical assump- tion that is needed to prove this assertion is that Lorentz transformations of the spacetime coordinates are imple- IV. QUANTUM ENTROPY AND SPECIAL RELATIVITY mented in quantum theory by unitary transformations of the various operators. This is the same as saying that the A. Reduced density matrices Lorentz group is a valid symmetry of the physical system (Weinberg, 1995). In our discussion of the measuring process, decoherence was attributed to the unability of accounting explicitly for the degrees of freedom of the environment. The envi- ronment thus behaves an exosystem (Finkelstein, 1988) D. Classical analogies and the system of interest is “open” because parts of the universe are excluded from its description. Are relativity and quantum theory really involved in This leads to the introduction of reduced density ma- these issues? The matter of information transfer by trices: let us use Latin indices for the description of the means of distant measurements is essentially nonrela- exosystem (that is, if we were able to give it a description) tivistic. Replace “superluminal” by “supersonic” and and Greek indices for the subsystem that we can actually the argument is exactly the same. The maximal speed describe. The components of a state vector would thus of communication is determined by the dynamical laws be written Vmµ and those of a density matrix ρmµ,nν . that govern the physical infrastructure. In quantum field The reduced density matrix of the system of interest is theory, the field excitations are called “particles” and given by their speed over macroscopic distances cannot exceed the speed of light. In condensed matter physics, linear exci- τ = ρ . (24) tations are called phonons and the maximal speed is that µν mµ,mν Xm of sound. As to the EPRB setup, consider an analogous classi- Even if ρ is a pure state (a matrix of rank one), τ is in cal situation: a bomb, initially at rest, explodes into two general a mixed state. Its entropy is defined as fragments carrying opposite angular momenta. Alice and Bob, far away from each other, measure arbitrarily cho- S = tr (τ log τ). (25) − 14

In a relativistic system, whatever is outside the past spinor, light cone of the observer is unknown to him, but also a (p) cannot affect his system, therefore does not lead to de- ψ(p)= 1 , (26) coherence (here, we assume that no particle emitted by a2(p) an exosystem located outside the past cone penetrates where the amplitudes a satisfy a (p) 2dp = 1. into the future cone.) Since observers located at different r r | r | points have different past light cones, they exclude from The normalization of these amplitudesP R is a matter of con- their descriptions different parts of spacetime. Therefore venience, depending on whether we prefer to include a 2 p2 1/2 any transformation law between them must tacitly as- factor p0 = (m + ) in it, or to have such factors in sume that the part excluded by one observer is irrelevant the transformation law (29) below. Following Halpern to the system of the other observer. (1968), we shall use the second alternative, because it is closer to the nonrelativistic notation which appears in Another consequence of relativity is that there is a hi- the usual definition of entropy. In this section, we use erarchy of dynamical variables: primary variables have natural units: c = 1. relativistic transformation laws that depend only on the Here we emphasize that we consider normalizable Lorentz transformation matrix Λ that acts on the space- states, in the momentum representation, not momen- time coordinates. For example, momentum components tum eigenstates as usual in textbooks on particle physics. are primary variables. On the other hand, secondary The latter are chiefly concerned with the computation of variables such as spin and polarization have transforma- in out matrix elements needed to obtain cross sections tion laws that depend not only on Λ, but also on the mo- andh | otheri asymptotic properties. However, in general mentum of the particle. As a consequence, the reduced a particle has no definite momentum. For example, if density matrix for secondary variables, which may be well an electron is elastically scattered by some target, the defined in any coordinate system, has no transformation electron state after the scattering is a superposition that law relating its values in different Lorentz frames. A sim- involves momenta in all directions. ple example is given in Sec. IV.B. Appendix A gives a In that case, it still is formally possible to ask, in any summary of the relativistic state transformations for free Lorentz frame, what is the value of a spin component in a particles. given direction (this is a legitimate Hermitian operator). Moreover, an unambiguous definition of the reduced In quantum information theory, the important issue does density matrix by means of Eq. (24) is possible only if the not reside in asymptotic properties, but how entangle- secondary variables are unconstrained. For gauge field ment (a communication resource) is defined by different theories, that equation may be meaningless if it conflicts observers. Early papers on this subject used momentum with constraints imposed on the physical states (Beck- eigenstates, just as in particle physics (Czachor, 1997). man et al., 2002; Peres and Terno, 2003). In the absence However, radically new properties arise when localized of a general prescription, a case-by-case treatment is re- quantum states are considered. quired. A particular construction, valid with respect to a Let us define a reduced density matrix, τ = certain class of tests, is given in Sec. IV.C. A general way dp ψ(p)ψ (p), giving statistical predictions for the re- of defining reduced density matrices for physical states in † sults of measurements of spin components by an ideal gauge theories is an open problem. R apparatus which is not affected by the momentum of the particle. The spin entropy is

B. Massive particles S = tr (τ log τ)= λj log λj , (27) − − X We first consider the relativistic properties of the spin where λj are the eigenvalues of τ. 1 As usual, ignoring some degrees of freedom leaves the entropy for a single, free particle of spin 2 and mass m> 0. We shall show that the usual definition of quan- others in a mixed state. What is not obvious is that in tum entropy has no invariant meaning. The reason is the present case the amount of mixing depends on the that under a Lorentz boost, the spin undergoes a Wigner Lorentz frame used by the observer. Indeed consider an- rotation (Wigner, 1939; Halpern, 1968) whose direction other observer (Bob) who moves with a constant velocity and magnitude depend on the momentum of the particle. with respect to Alice who prepared state (26). In the 1 Even if the initial state is a direct product of a function Lorentz frame where Bob is at rest, the same spin- 2 par- of momentum and a function of spin, the transformed ticle has a state state is not a direct product. Spin and momentum ap- a (p) pear to be entangled. (This is not the familiar type of ψ (p)= 1′ . (28) ′ p entanglement which can be used for quantum communi- a2′ ( ) cation, because both degrees of freedom belong to the The transformation law is (Weinberg, 1995) same particle, not to distinct subsystems that could be widely separated.) 1 1/2 1 1 a′(p) = [(Λ− p)0/p0] Drs[Λ, (Λ− p)] as(Λ− p), 1 The quantum state of a spin- 2 particle can be written, Xs in the momentum representation, as a two-component (29) 15 where Drs is the Wigner rotation matrix for a Lorentz have no relativistic transformation law. Only the com- transformation Λ. Further details of this transformation plete Liouville functional, including the fields, has one. and its representation by a quantum circuit are given in Consider now a pair of orthogonal states that were Appendix A. prepared by Alice. How well can moving Bob distinguish As an example, take a particle prepared by Alice with them, if he is restricted to measuring discrete degrees of spin in the z direction, so that a2(p) = 0. Spin and freedom? We shall use the simplest criterion, namely the momentum are not entangled, and the spin entropy is probability of error PE, defined as follows: an observer zero. When that particle is described in Bob’s Lorentz receives a single copy of one of the two known states and frame, moving with velocity β in a direction at an angle θ performs any operation permitted by quantum theory in with Alice’s z-axis, a detailed calculation shows that both order to decide which state was supplied. The probability a1′ and a2′ are nonzero, so that the spin entropy is positive of a wrong answer for an optimal measurement is (Fuchs (Peres, Scudo, and Terno, 2002). This phenomenon is and van de Graaf, 1999) illustrated in Fig. 2. A relevant parameter, apart from 1 1 2 the angle θ, is, in the leading order in momentum spread, PE(ρ1,ρ2)= 2 + 4 tr (ρ1 ρ2) . (31) p − ∆ 1 1 β2 In Alice’s frame PE = 0. It can be shown that in Bob’s Γ= − − , (30) 2 m pβ frame, PE′ Γ , where the proportionality factor de- pends on the∝ angle θ defined above. Of course, the op- where ∆ is the momentum spread in Alice’s frame. The posite Lorentz transformation induces a change from a entropy has no invariant meaning, because the reduced positive PE in Bob’s frame to PE=0 in Alice’s frame. density matrix τ has no covariant transformation law, We discuss the resulting effective quantum channel in except in the limiting case of sharp momenta. Only the Sec. IV.E. complete density matrix transforms covariantly. How is the linearity of the transformation laws lost in 3 this purely quantum mechanical problem? The momenta p do transform linearly, but the law of transformation 2 of spin depends explicitly on p. When we evaluate τ θ by summing over momenta in ρ, all knowledge of these 1 momenta is lost and it is then impossible to obtain τ ′ by transforming τ. Not only is linearity lost, but the result is not nonlinearity in the usual sense of this term. It 0 S is the absence of any definite transformation law which 0.015 depends only on the Lorentz matrix. 0.01 It is noteworthy that a similar situation arises for a 0.005 classical system whose state is given in any Lorentz frame 0 by a Liouville function (Balescu and Kotera, 1967). Re- 0.1 0.08 call that a Liouville function expresses our probabilistic 0.06 description of a classical system — what we can pre- 0.04 dict before we perform an actual observation — just as Γ 0.02 a quantum state is a mathematical expression used for computing probabilities of events. FIG. 2 Dependence of the spin entropy S, in Bob’s frame, To avoid any misunderstanding, we emphasize that on the values of the angle θ and a parameter Γ ≈ [1 − (1 − there is no consistent relativistic statistical mechanics β2)1/2]∆/mβ, where ∆ is the momentum spread in Alice’s for N interacting particles, with a 6N-dimensional phase frame. space defined by the canonical coordinates pn and qn (n =1,...,N). Any relativistic interaction must be me- diated by fields, having an infinity of degrees of freedom. A complete Liouville function, or rather Liouville func- C. Photons tional, must therefore contain not only all the canonical variables pn and qn, but also all the fields. However, once The long range propagation of polarized photons is this Liouville functional is known (in principle), we can an essential tool of quantum cryptography (Gisin et al., define from it a reduced Liouville function, by integrating 2002). Usually, optical fibers are used, and the photons the functional over all the degrees of freedom of the fields. may be absorbed or depolarized due to imperfections. In The result is a function of pn and qn only (just as we some cases, such as communication with space stations, compute reduced density matrices in quantum theory). the photons propagate in vacuo (Buttler et al., 2000). The time evolution of such reduced Liouville functions The beam then has a finite diffraction angle of order cannot be obtained directly from canonical Hamiltonian λ/a, where a is the aperture size, and new deleterious dynamics without explicitly mentioning the fields. These effects appear. In particular a polarization detector can- functions are well defined in any Lorentz frame, but they not be rigorously perpendicular to the wave vector and 16 the transmission is never faithful, even with perfect de- The square root is the familiar relativistic Doppler factor. tectors. Moreover, this “vacuum noise” depends on the For large negative v, the diffraction angle becomes arbi- relative motion of the observer with respect to the source. trarily small, and sideway losses (which are proportional 2 These relativistic effects are essentially different from to θ′ ) can be reduced to zero. those for massive particles that were discussed above, It is noteworthy that the same Doppler factor was because photons have only two linearly independent po- obtained by Jarett and Cover (1981) who considered larization states. The properties that we discuss are only the relativistic transformations of bit rate and noise kinematical, not dynamical. At the statistical level, it intensity, without any specific physical model. This is not even necessary to involve quantum electrodynam- remarkable agreement shows that information theory ics. Most formulas can be derived by elementary clas- should properly be considered as a branch of physics. sical methods (Peres and Terno, 2003). It is only when In applications to secure communication, the ideal sce- we consider individual photons, for cryptographic appli- nario is that isolated photons (single particle Fock states) cations, that quantum theory becomes essential. The are emitted. In a more realistic setup, the transmission diffraction effects mentioned above lead to superselection is by means of weak coherent pulses containing on the rules which make it impossible to define a reduced den- average less than one photon each. A basis of the one- sity matrix for polarization. As shown below, it is still photon space is spanned by states of definite momentum possible to have “effective” density matrices; however, and helicity, the latter depend not only on the preparation process, k, ǫ± k ǫ± , (37) but also on the method of detection that is used by the | k i ≡ | i ⊗ | k i observer. where the momentum basis is normalized by q k = Assume for simplicity that the electromagnetic sig- 3 0 (3) h | i (2π) (2k )δ (q k), and helicity states ǫ± are ex- nal is monochromatic. In a Fourier decomposition, the k plicitly defined by− Eq. (40) below. | i Cartesian components of the wave vector k (with µ = µ As we know, polarization is a secondary variable: 0, 1, 2, 3) can be written in term of polar angles: states that correspond to different momenta belong to distinct Hilbert spaces and cannot be superposed (an ex- kµ = (1, sin θ cos φ, sin θ sin φ, cos θ), (32) ǫ ǫ pression such as k± + q± is meaningless if k = q). The complete basis (37)| i does| noti violate this superselection6 where we use units such that c = 1 and k0 = 1. Let us choose the z axis so that a well collimated beam has a rule, owing to the othogonality of the momentum basis. large amplitude only for small θ. Therefore, a generic one-photon state is given by a wave In a real experiment, the angles θ and φ are distributed packet in a continuous way around the z axis (exactly how de- pends on the properties of the laser) and one has to take Ψ = dµ(k)f(k) k, α(k) . (38) | i Z | i a suitable average over them. As the definition of polar- ization explicitly depends on the direction of k, taking The Lorentz-invariant measure is dµ(k)= d3k/(2π)32k0, the average over many values of k leads to an impure and normalized states satisfy dµ(k) f(k) 2 = 1. The | | polarization and may cause transmission errors. generic polarization state α(k)R corresponds to the geo- Let us consider the effect of a motion of the detec- metrical 3-vector | i tor relative to the emitter, with a constant velocity v = α ǫ+ ǫ (0, 0, v). The Lorentz transformation of kµ in Eq. (32) (k)= α+(k) k + α (k) k−, (39) − yields new components 2 2 ǫ where α+ + α = 1, and the explicit form of k± is given below.| | | −| k0′ = γ(1 v cos θ) and kz′ = γ(cos θ v), (33) − − Lorentz transformations of quantum states are most 2 1/2 where γ = (1 v )− . Considering again a single easily computed by referring to some standard momen- Fourier component,− we have, instead of the unit vector tum, which for photons is pν = (1, 0, 0, 1). Accordingly, k, a new unit vector standard right and left circular polarization vectors are ǫ± = (1, i, 0)/√2. For linear polarization, we take p ± sin θ cos θ v Eq. (39) with α+ = (α )∗, so that the 3-vectors α(k) k′ = , 0, − . (34) − γ(1 v cos θ) 1 v cos θ  are real. In general, complex α(k) correspond to elliptic − − polarization. In other words, there is a new tilt angle θ′ given by Under a Lorentz transformation Λ, these states be- come k , α(k ) , where k is the spatial part of a four- | Λ Λ i Λ sin θ′ = sin θ/γ(1 v cos θ). (35) vector k =Λk, and the new polarization vector can be − Λ obtained by an appropriate rotation given by Eq. (42) For small θ, such that θ2 v , we have below. For each k a polarization basis consists of the ≪ | | helicity vectors, 1+ v θ′ = θ . (36) ǫ ˆ ǫ r1 v k± = R(k) p±, (40) − 17 and the corresponding quantum states are k, ǫ± . To the direction xˆ corresponds a projection operator | k i As usual, kˆ denotes the unit 3-vector in the direc- tion of k. The standard matrix (Weinberg, 1995) P = xˆ xˆ 1l = xˆ xˆ dµ(k) k k , (46) x | ih |⊗ p | ih |⊗ | ih | that rotates the standard direction (0, 0, 1) to kˆ = Z

(sin θ cos φ, sin θ sin φ, cos θ) is where 1lp is the unit operator in momentum space. The ǫ ǫℓ action of Px on Ψ follows from Eq. (43) and k± k = 0. cos θ cos φ sin φ cos φ sin θ Only the transversal| i part of xˆ appears in theh | expecta-i ˆ − R(k)=  cos θ sin φ cos φ sin φ sin θ  , (41) tion value: | i sin θ 0 cos θ −   2 2 Ψ Px Ψ = dµ(k) f(k) x+(k)α+∗ (k)+x (k)α∗ (k) . and likewise for kˆΛ. h | | i Z | | | − − | Under a general Lorentz transformation, be it a rota- (47) It is convenient to write the transversal part of xˆ as tion or a boost, helicity is preserved, but quantum states | i and the corresponding geometric vectors acquire helicity- + + b (k) ( ǫ ǫ + ǫ− ǫ− ) xˆ , (48) dependent phases (see Appendix A for more details): | x i ≡ | k ih k | | k ih k | | i ǫ+ ǫ = x+(k) k + x (k) k− . (49) ǫ+ ǫ iξ(Λ,kˆ)ǫ+ iξ(Λ,kˆ)ǫ | i − | i α+ k + α k− α+e kΛ + α e− k−Λ , (42) − → − Likewise define b (k) and b (k) . These three state | y i | z i where the explicit expressions for ξ(Λ, kˆ) are given by vectors are neither of unit length nor mutually orthogo- Lindner, Peres, and Terno (2003) and Bergou, Gingrich nal. For k = (sin θ cos φ, sin θ sin φ, cos θ) we have and Adami (2003). b (k) = 1 (cos θ cos φ + i sin φ) ǫ+ + (50) The superselection rule that was mentioned above | x i √2 | k i 1 makes it impossible to define a reduced density matrix (cos θ cos φ i sin φ) ǫ− c(θ, φ) k, e (k) , (51) √2 k x in the usual way (Peres and Terno, 2003; Lindner, Peres − | i≡ | i and Terno, 2003). We can however define an “effective” where ex(k) is given by Eq. (44), and c(θ, φ) = reduced density matrix for polarization, as follows. The 2 2 n x+ + x . labelling of polarization states by Euclidean vectors ek, q − and the fact that photons are spin-1 particles, suggest Finally, a POVM element Ex which is the physical the use of a 3 3 matrix with entries labelled x, y and z. part of Px, namely is equivalent to Px for physical states Classically, they× correspond to different directions of the (without longitudinal photons) is electric field. For example, when kˆ = zˆ, only ρxx, ρxy, ρ are non-zero. For a generic photon state Ψ , let us Ex = dµ(k) k, bx(k) k, bx(k) , (52) yy | i Z | ih | try to construct a reduced density matrix ρxx that gives the expectation value of an operator representing the po- and likewise for the other directions. The operators Ex, larization in the x direction, irrespective of the particle’s Ey and Ez indeed form a POVM in the space of physical momentum. states, owing to Eq. (45). The above derivation was, ad- To have a momentum-independent polarization is to mittedly, a rather circuitous route for obtaining a POVM tacitly admit longitudinal photons. Unphysical concepts for polarization. This is due to the fact that the latter are often used in intermediate steps in theoretical physics. is a secondary variable, subject to superselection rules. Momentum-independent polarization states thus consist Unfortunately, this is the generic situation. of physical (transversal) and unphysical (longitudinal) The entire effective density matrix is reconstructed us- parts, the latter corresponding to a polarization vector ing techniques of Chuang and Nielsen (1997), and we get ǫℓ = kˆ. For example, a generalized polarization state a simple expression for the reduced density matrix cor- along the x-axis is responding to the polarization state α(k) : | i ǫ+ ǫ ǫℓ xˆ = x+(k) k + x (k) k− + xℓ(k) k , (43) 2 α α | i | i − | i | i ρmn = dµ(k) f(k) (k) bm(k) bn(k) (k) (53) Z | | h | ih | i ǫ ˆ where x (k)= k± xˆ, and xℓ(k)= xˆ k = sin θ cos φ. It ± 2 · 2 2 · It is interesting to note that this derivation gives a direct follows that x+ + x + xℓ = 1, and we thus define | | | −| | | physical meaning to the naive definition of a reduced den- ǫ+ ǫ x+(k) k + x (k) k− sity matrix, ex(k)= − , (44) 2 2 x+ + x naive 2 − ρ = dµ(k) f(k) αm(k)α∗ (k)= ρmn. (54) q mn Z | | n as the polarization vector associated with the x direction. It follows from (43) that xˆ xˆ = 1 and xˆ yˆ = xˆ yˆ = 0, Since polarization 3-vectors transform under rotations re- and likewise for other directions,h | i so thath | i · gardless of momentum, the effective 3 3 polarization density matrix has a standard transformation× law under xˆ xˆ + yˆ yˆ + ˆz ˆz = 1l. (45) rotation R as well, ρ RρRT . | ih | | ih | | ih | → 18

Our basis states k, ǫk are direct products of momen- tangled states C = 1, while for non-entangled states tum and polarization.| Owingi to the transversality re- C = 0. quirement ǫk k = 0, they remain direct products un- Alsing and Milburn (2002) considered bipartite states der Lorentz transformations.· All the other states have with well-defined momenta. They showed that while their polarization and momentum degrees of freedom en- Lorentz transformations change the appearance of the tangled. As a result, if one is restricted to polarization state in different inertial frames and the spin directions measurements as described by the POVM elements (52), are Wigner rotated, the amount of entanglement remains there do not exist two orthogonal polarization states. It intact. The reason is that Lorentz boosts do not cre- follows that photon polarization states cannot be cloned ate spin-momentum entanglement when acting on eigen- perfectly, because the no-cloning theorem (Wootters and states of momentum, and the effect of a boost on a pair Zurek, 1982; Dieks, 1982) forbids an exact copying of un- is implemented on both particles by local unitary trans- known non-orthogonal states. In general, any measure- formations, which are known to preserve entanglement. ment procedure with finite momentum sensitivity will The same conclusion is valid for photon pairs. lead to the errors in identification. In particular, Hacyan (2001) showed that since the Our present problem is the distinguishability by our polarization angle remains constant in the polarization observer, Bob, of a pair of different quantum states that plane, the directions of perfect correlation for two pho- were prepared by Alice. The probability of an error by tons still exist in any reference frame, even if they are Bob is given by Eq. (31). The distinguishability of polar- different from the laboratory directions. Terashima and ization density matrices depends on the observer’s mo- Ueda (2003) showed that in a quite general setting for tion. We again assume that Bob moves along the z-axis both massive and massless particles, allowing for relative with a velocity v. Let us calculate his reduced density motion, it is always possible to find directions of perfect matrix. Recall that reduced density matrices have no (anti)correlations. transformation law (only the complete density matrix has However, realistic situations involve wave packets. For 1 one) except in the limiting case of sharp momenta. To example, a state of two spin- 2 particles is calculate Bob’s reduced density matrix, we must trans- form the complete state, and only then take a partial Υ12 = dµ(p1)dµ(p2)g(σ1σ2, p1, p2) p1, σ1; p2, σ2 , trace. A detailed calculation (Peres and Terno, 2003) | i Z | i σX1,σ2 leads to (57) 1+ v where dµ(p)= d3p/16π3p0 as usual. P ′ = P , (55) E 1 v E For typical particle beams, g is sharply peaked at some − values p10, p20. Again, a boost to any Lorentz frame will which may be either larger or smaller than PE . As ex- result in a unitary U(Λ) U(Λ) acting on each particle pected, we obtain for one-photon states the same Doppler separately, thus preserving⊗ the entanglement. Neverthe- effect as in the classical equation (36). less, since boosts can change entanglement between dif- ferent degrees of freedom of each particle, the spin-spin entanglement is frame-dependent as well. D. Entanglement Gingrich and Adami (2002) investigated the reduced density matrix for Υ and made explicit calculations An important problem is the relativistic nature of 12 for the case where|g isi a Gaussian, as in the work of quantum entanglement when there are several particles. Peres, Scudo, and Terno (2002). They showed that if For two particles, an invariant definition of the entangle- two particles are maximally entangled in a common, ap- ment of their spins would be to compute it in the Lorentz proximate rest frame (Alice’s frame), then C(ρ), as seen “rest frame” where p = 0. However, this simple def- by a Lorentz-boosted Bob, decreases when the boost ve- inition is not adequateh wheni there are more than two P locity tends to c. Of course, the inverse transformation particles, because there appears a problem of cluster de- from Bob to Alice will increase the concurrence. Thus, we composition: each subset of particles may have a different see that that spin-spin entanglement is not a Lorentz in- rest frame. This is a difficult problem, still awaiting for variant quantity, exactly as spin entropy is not a Lorentz a solution. We shall mention only a few partial results. scalar. Relativistic properties of the polarization entan- First, we have to define a convenient measure of entan- glement we investigated by Bergou, Gingrich and Adami glement. For two spin- 1 particles, the concurrence, C(ρ), 2 (2003). is defined as follows (Wootters, 1998). Introduce a spin- flipped stateρ ˜ = (σy σy)ρ∗(σy σy). The concurrence is ⊗ ⊗ E. Communication channels

C(ρ) = max(0, λ1 λ2 λ3 λ4), (56) − − − Although reduced polarization density matrices have where λi are the eigenvalues, in decreasing order, of the no general transformation rule, the above results show Hermitian matrix [√ρρ˜√ρ]1/2. The larger the concur- that such rules can be established for particular classes rence, the stronger the entanglement: for maximally en- of experimental procedures. We can then ask how these 19 effective transformation rules, τ ′ = T (τ), fit into the Theorem. The set of states ( ) Ω , generated from framework of general state transformations. Are they the vacuum Ω by the (polynomial)A O | algebrai of operators completely positive (CP) as in Eq. (6)? It can be proved in any bounded| i region, is dense in the Hilbert space of that distinguishability, as expressed by natural measures all field states.  like P , cannot be improved by any CP transforma- E This is the Reeh-Schlieder theorem (Reeh and tion (Fuchs and van de Graaf, 1999). However, the CP Schlieder, 1961; Streater and Wightman, 1964; Haag requirement may fail if there is a prior entanglement 1996; Araki, 1999). It asserts that there are local op- with another system and the dynamics is not factorizable erators Q ( ) which, applied to the vacuum, pro- (Pechukas, 1994; Stelmachoviˇcandˇ Buˇzek, 2001; Salgado duce a state∈ AwhichO is arbitrarily close to any arbitrary and S´anchez-G´omez, 2002). Υ (the vacuum state can be replaced by any state of Since in Eq. (55) and in the discussion following finite| i energy). Thus in principle any entangled state can Eq. (31) we have seen that distinguishability can be im- be arbitrarily closely approximated by suitable local op- proved, we conclude that these transformations are not erations on any other state. completely positive. The reason is that the Lorentz trans- The theorem reveals a surprising amount of entangle- formation acts not only on the “interesting” discrete vari- ment that is present in the vacuum state Ω . The corol- ables, but also on the primary momentum variables that lary below shows that if a local operator is| usedi to model we elected to ignore and to trace out, and its action on a detector, that detector must have “dark counts”: it has the interesting degrees of freedom depends on the “hid- a finite probability to “click” in a vacuum. den” primary ones. Of course, the complete state, with all the variables, transforms unitarily and distinguisha- Corollary. No operator that is localized in a bounded bility is preserved. spacetime region annihilates the vacuum (nor any other This technicality has one important consequence. In physical state).  quantum information theory quantum channels are de- Another important theorem is due to Epstein, Glaser scribed by completely positive maps that act on qubit and Jaffe (1965): states (Holevo, 1999; Keyl, 2002). Qubits themselves are realized as discrete degrees of freedom of various parti- Theorem. If a field Q(x) satisfies Ψ Q(x) Ψ 0 for cles. If relativistic motion is important, then not only all states, and if Ω Q(x) Ω = 0 forh the| vacuum| i ≥ state, does the vacuum behave as a noisy quantum channel, then Q(x)=0. h | | i  but the very representation of a channel by a CP map This implies that no POVM constructed from local fails. or quasi-local operators can have zero vacuum response. The theorem predicts for any local field Q(x) that has a zero vacuum expectation value, namely Ω Q(x) Ω = 0, V. THE ROLE OF QUANTUM FIELD THEORY there exists a state for which the expectationh | | valuei of Q(x) is negative. Further details can be found in the The POVM formalism is an essential tool of quantum original article and in Tippler (1978). information theory. Entanglement is a major resource Another implication is a violation of the classical en- for quantum communication and computation. In this ergy conditions (Hawking and Ellis, 1973; Wald, 1984). section we present results of quantum field theory that Classically, energy density is always positive and the are important for the relativistic generalization of these stress-energy tensor for all classical fields satisfies the concepts. Mathematical results are stated in an informal µ ν µ weak energy condition (WEC) Tµν u u 0, where u is way. Rigorous formulations and fine mathematical points any timelike or null vector. The Epstein-Glaser-Jaffe≥ the- can be found in the references that are supplied for each orem shows that this is impossible for the renormalized concept or theorem we introduce. stress-energy tensor of quantum field theories. Since it has by definition a null vacuum expectation value, there µ ν are states Υ such that Υ Tµν u u Υ < 0. For exam- A. General theorems ple, squeezed| i states of theh | electromagnetic| i field (Man- del and Wolf, 1995), or the scalar field (Borde, Ford, First, we define the notions of local and quasi-local and Roman, 2002), have locally negative energy densi- operators (Emch, 1972; Bogolubov et al., 1990; Haag, ties. The violation of WEC raises doubts on the use of 1996; Araki, 1999). Local operators are associated with energy density for the description of particle localization, bounded regions of spacetime. For example, they may as discussed in Sec. V.B. be field operators that are smeared with functions of While any entangled state can be approximated by the bounded support (that is, functions that vanish if their action of local operators on Ω , the clustering property argument is outside of a prescribed bounded region of the vacuum14 asserts that| statesi created by local oper- of spacetime). Smeared renormalized stress-energy ten-O sors also belong to this category. Quasi-local operators are obtained when the smearing functions have exponen- tially decaying tails. 14 Its relation to the cluster property of the S-matrix is discussed 20 ations, namely Q Ω ,Q ( ), tend to look practically an operational meaning only owing to their localization: like a vacuum with| i respect∈ A toO measurements in distant, particles are what is registered by detectors. causally unconnected regions. The behavior of detectors When quantum mechanics was a new science, most that are far away from each other is ruled by the fol- physicists wanted to preserve the notions with which lowing theorems, where, for a local operator B ( ), they were familiar, and considered particles as real ob- ∈ A O we denote by Bx its translate by a spatial vector x, i.e., jects having positions and momenta that were possibly Bx = U(x)BU †(x). unknown, and/or subject to an “uncertainty principle.” Still, a few writers expressed critical opinions, for exam- Theorem. If A, B ( ) are local operators and Ω ple “...no scheme of operations can determine experi- is the vacuum state,∈ then A O | i mentally whether physical quantities such as position and x momentum exist. . . we get into a maze of contradictions Ω ABx Ω | |→∞ Ω A Ω Ω Bx Ω . (58) h | | i −→ h | | ih | | i as soon as we inject into quantum mechanics such con- cepts carried over from the language of our ancestors...” There are estimates on the rate of convergence of the (Kemble, 1937). above expression as a function of the spacelike separa- More recently, Haag (1996) wrote tion for the cases of massive and massless particles. The asymptotic behavior depends on that of the Wightman “...it is not possible to assume that an elec- function W (x , x ) for x x 2 (Streater and 1 2 | 1 − 2| → ∞ tron has, at a particular instant of time, any Wightman, 1964; Bogolubov et al. 1990; Haag, 1996). position in space; in other words, the concept Theorem. If A ( ) and B ( ), where of position at a given time is not a meaningful ∈ A O1 ∈ A O2 O1 and 2 are mutually spcelike regions with a spacelike attribute of the electron. Rather, ‘position’ separationO r, then is an attribute of the interaction between the electron and a suitable detection device.” Ω AB Ω Ω A Ω Ω B Ω (59) |h | | i − h | | ih | | i| We shall first briefly examine some aspects of the old for a massless theory is bounded by fashioned approach to localization. First we note that even when we construct a local probability density (and, f( , ,A,B)/r2, (60) O1 O2 possibly, a corresponding current) it is impossible to in- where f is a certain function that depends on the regions terpret ρ(x,t)d3x as the probability to find a particle in and the operators, but not on the distance between the the volume d3x at the space point x. It was argued by regions; for a massive theory it is bounded by Landau and Peierls (1931) that a particle may be local- ized only with uncertainty ∆x> ~c/ E , where E is the mr h i h i e− g(A, B), (61) particle’s expected energy. Intuitively, confinement of a particle to a narrower domain by “high walls” requires where m is the relevant mass and g depends on the op- a very strong interaction which leads to pair production. erators only. In this case , may be unbounded.  O1 O2 Haag and Swieca (1965) have shown that restriction to a The explicit derivation of the coefficients requires a compact region of spacetime makes it impossible to de- more detailed treatment. Particular cases and values of tect with certainty any state. Hegerfeldt (1985) proved numerical constants are given by Emch (1972), Freden- that if a one-particle POVM leads to probability distribu- hagen (1985), Haag (1996), and Araki (1999). tions such that the total probability of finding a particle While it seems that vacuum correlations for massless outside a sphere of radius R at time t is bounded by fields decay much slower, the difference disappears if the 2 finite sensitivity of detectors for soft photons is taken into Prob R < C exp( 2γR), (62) 6∈ − account. It was shown by Summers and Werner (1987) that if a detector has an energy threshold ǫ, the latter where C is some constant and γ>m, then at later times serves as an effective mass in correlation estimates, and the probability distribution will spread faster than light. ǫr an additional e− factor appears in Eq. (60). Furthermore, Giannitrapani (1998) and Toller (1999) proved that a spacetime localized POVM cannot be con- structed even from quasi-local operators. General discus- B. Particles and localization sions of localization from the point of view of algebraic quantum field theory can be found in the works of Buch- Classical interventions in quantum systems are local- holz and Fredenhagen (1982), Roberts (1982), Neumann ized in space and time. However, the principles of quan- and Werner (1983), Werner (1986) and Haag (1996). tum mechanics and relativity dictate that this localiza- Much earlier, Newton and Wigner (1949) had at- tion is only approximate. The notion of particles has tempted to define a position operator, whose spectral de- composition (Wightman, 1962) gives a rough indication of the particle localization. However, it was shown by Rosenstein and Usher (1987) that Gaussian-like Newton- by Weinberg (1995). Wigner wave functions lead to superluminal propagation 21 of probability distributions. Busch (1999) reviewed the We now consider correlation experiments with devices problems involved in the construction of POVMs for par- a and b placed in spacelike-separated regions L and R, ticle localization. so all local operators pertaining to these regionsO com-O Energy density is directly related to photon local- mute: [ ( L), ( R)] = 0. In each region, there are A O A O ization in quantum optics (Mandel and Wolf, 1995; two such devices, labelled a1,a2,b1,b2, which yield out- Bialynicki-Birula, 1996). If the electrons in a detector comes “yes” or “no” in each individual experiment. We interact with the electric field of light, then in a simple denote the probabilities for positive outcomes as p(aj ) model the detection probability is proportional to the and p(bk), and by p(aj bk) the probability of their joint ∧ expectation value of the normal-ordered electric field in- occurrence. tensity operator I(x,t) (Mandel, 1966), and the latter is The measuring apparatus aj is described by a POVM proportional to the energy density. This probability dis- element Fj ( L) and the probability of the “yes” ∈ A O tribution decays asymptotically as the seventh power of outcome for a state ρ is tr (ρFj ). If Gk is the POVM distance, or even slower (Amrein, 1969). Despite it suc- for apparatus bk then the probability of the “yes-yes” cess in these examples, the notion of localization based outcome is tr (ρ Fj Gk). Let us to introduce operators on the energy density cannot have a universal validity, Aj =2Fj 1l and Bk =2Gk 1l, and define − − because the violation of WEC makes it unsuitable for ζ(a,b,ρ)= 1 tr ρ[A (B + B )+ A (B B )] . (64) the construction of POVMs. 2 { 1 1 2 2 1 − 2 } The real physical problem is how localized detectors This quantity, which is experimentally measurable, has a can be. The idealization of “one detector per spacetime classical analogue whose value is bounded: ζ 1. This point” is obviously impossible. How can we manage to is the CHSH inequality (Clauser et al., 1969),≤ which is ensure that two detectors have zero probability to over- one of the variants of the Bell inequality (Bell, 1964).15 lap? There appears to be a fundamental trade-off be- The above definition of ζ can be extended to tween detector reliability and localizability. The bottom line is how to formulate a relativistic interaction between ζ( , ,ρ) = sup ζ(a,b,ρ), (65) a detector and the detected system. A true detector A B should be amenable to a dual quantum-classical descrip- where = ( ), = ( ), and the supremum A A OL B A OR tion, as in the Hay-Peres model (1998). This problem is taken over all operators Aj ,Bk. It was shown by seems to be very far from a solution. Completely new Cirel’son (1980) that there is also a quantum bound on notions may have to be invented. correlations: for commuting algebras and and any Although states with a definite number of particles are state ρ, A B a useful theoretical concept, a look at quantum optics techniques or at the Table of Particle Properties shows ζ( , ,ρ) √2. (66) that experimentally accesible quantum states are usually A B ≤ not eigenstates of particle number operators. In general Further results of Summers and Werner (1985, any process that is not explicitly forbidden by some con- 1987a,b) and Landau (1987) establish that a violation servation law has a non-zero amplitude (Weinberg, 1995; Bell’s inequalities is generic in quantum field theory. For Peskin and Schroeder, 1995; Haag, 1996). There are mul- any two spacelike separated regions and any pairs of op- tiple decay channels, extra soft photons may always ap- erators, a, b, there is a state ρ such that the CHSH in- pear, so that the so-called ‘one-photon’ states are often equality is violated, namely, ζ(a,b,ρ) > 1. With addi- accompanied by soft multiphoton components, tional technical assumptions the existence of a maximally violating state ρm can be proved: α Ω + β 1ω + γ 2ω′ω′′ + ..., β 1. (63) | i | i | i | |∼ ζ(a,b,ρm)= √2, (67) Thus the physical realization of a single qubit is itself necessarily an idealization. for any spacelike separated regions L and R. It follows from convexity arguments that statesO thatO maximally vi- olate Bell inequalities are pure. What are then the op- erators that lead to maximal violation? Summers and C. Entanglement in quantum field theory Werner (1987a) have shown that operators Aj and Bk that give ζ = √2 satisfy A2 = 1l and A A + A A = 0, Recall that while the Reeh-Schlieder theorem ensures j 1 2 2 1 and likewise for Bk. If we define A3 := i[A1, A2]/2, then that any state can be approximated by local operations, these three operators have the same algebra− as Pauli spin the clustering property of the vacuum implies that lo- matrices (Summers,1990). Even if we ignore the problem cally created states look almost like a vacuum for distant measurements. The Reeh-Schlieder and Epstein-Glaser- Jaffe theorems entail dark counts for local detectors. The responses of spatially separated detectors are correlated, 15 Recall that the Bell inequalities are essentially classical (Peres, but these correlations decay fast due to cluster proper- 1993). Their violation by a quatum system is a sufficient condi- ties. tion for entanglement, but not a necessary one. 22

of localization (Sec. V.B), a violation of Bell inequalities number of degrees of freedom. A usuful device in the de- is not at all trivial, as the analysis of various relativis- scription of infinite sytems is the notion of singular states, tic spin operators shows (Terno, 2003). For example, for which cannot be represented by density operators: states moving observers, if the observables are constructed by are considered to be just positive linear functionals on means of the Pauli-Lubanski operator, the amount of vio- the space of POVMs, and only non-singular states are lation of Bell’s inequality decreases with increasing veloc- represented by density operators (Emch, 1972; Bratteli ity, and the inequality is satisfied in the ultra-relativistic and Robinson, 1987). One of their results is a rigorous limit (Czachor, 1997; Ahn et al., 2003). description to the original EPR (1935) state, which can The violation of Bell’s inequalities by the vacuum state be modeled as a sequence of more and more squeezed does not mean that it is enough to have two detectors and two-mode states, and actually is a singular state. check their dark count coincidences. The cluster theorem Pachos and Solano (2003) discussed the generation of predicts a strong damping of the violations with distance. entangled states and performed ab initio QED calcula- 1 When the lowest relevant mass is m> 0, clustering leads tions for the case of two interacting spin- 2 charged par- to the estimate ticles. They obtained particular results for low energy scattering, and more general situations are under inves- ζ( ( L), ( R), Ω) 1+ 4exp[ mr( L, R)], (68) tigation. A O A O ≤ − O O where r( , ) is the separation between the regions OL OR (Summers and Werner, 1985, 1987a,b). For massless par- D. Accelerated detectors ticles, the energy threshold for photodetection serves as an effective mass. Therefore, a direct observation of vac- In quantum field theory, the vacuum is defined as the uum entanglement should be extremely difficult. Reznik lowest energy state of a field. A free field with linear (2000) proposed a method to convert vacuum entangle- equations of motion can be resolved into normal modes, ment into conventional bipartite entanglement. It re- such as standing waves. Each mode has a fixed frequency quires to switch on and off in a controllable way the and behaves as a harmonic oscillator. The zero point mo- interaction between two-level systems and a field. Ap- tion of all these harmonic oscillators is called “vacuum propriately tailored local interaction Hamiltonians can fluctuations”and the latter, under suitable conditions, then transfer vacuum entanglement to atoms. may excite a localized detector that follows a trajectory The classification of entangled states and their manip- xν (τ) parametrized by its proper time τ. The internal ulation are current research topics in quantum informa- structure of the detector is described by non-relativistic tion theory. Up to now we have dealt with entangle- quantum mechanics, so that we can indeed assume that it ment of a finite number of degrees of freedom, or spin- is approximately localized, and it has discrete energy lev- momentum entanglement. After introducing Lorentz els En. Furthermore, we assume the existence of a linear transformations, we were still able to use the standard coupling of an internal degree of freedom, µ, of the de- techniques of the non-relativistic theory. However, in the tector, with the scalar field φ(x(τ)) at the position of the general case, infinite-dimensional Hilbert spaces are in- detector. First-order perturbation theory gives the fol- volved. Recently Parker, Bose, and Plenio (2000), Eisert, lowing expression for the transition probability per unit Simon, and Plenio (2002), and Keyl, Schlingemann, and proper time: Werner (2003) investigated the entanglements of forma-

tion and of distillation in infinite-dimensional systems. 2 2 i(E E0)τ g En µ E0 dτe− − W (τ), (69) When the Hilbert space of a bipartite system is infinite |h | | i| Z Xn dimensional, some peculiarities arise. For pure states, a natural measure of entanglement is the von Neumann where g is a coupling constant and entropy S = tr ρ ln ρ of either one of the reduced density − matrices. It can be shown (Eisert, Simon and Plenio, W (τ) W (x(τ1), x(τ2)), τ = τ1 τ2, (70) ≡ − 2002) that in an arbitrarily small neighborhood of any state there is an infinity of entangled states. The reason is is the Wightman function, defined by W (x1, x2) = Ω ψ(x )ψ(x ) Ω for two arbitrary points on the detec- that in the neighborhood of any state with finite energy, h | 1 2 | i there are states of infinite entropy (Wehrl, 1978).16 This tor’s trajectory (Streater and Wightman, 1964). The in- seems paradoxical, but if we consider states with bounded tegral in Eq. (69) is the Fourier transform of the auto- energy only, the continuity of the degree of entanglement correlation. In other words, it gives the power spectrum is restored. of the Wightman function. ν ν Keyl, Schlingemann and Werner (2003) applied tech- For inertial detectors (that is, x = v τ with a con- ν niques of operator algebra to systems with an infinite stant four-velocity v ) the transition probability is zero, as one should expect. However, the response rate does not vanish for more complicated trajectories. Consider in particular one with constant proper acceleration a. 16 The set of states with infinite entropy is trace-norm dense in the With an appropriate choice of initial conditions, it corre- state space. sponds to the hyperbola t2 + x2 =1/a2, shown in Fig. 3. 23

Then the transition rate between levels appears to be ated cavity will be excited. The relevant property in all the same as for an inertial detector in equilibrium with these cases is the relative acceleration of the detector and ~ thermal radiation at temperature T = a/2πckB . This the field normal modes. phenomenon is called the Unruh (1976) effect. It was We now consider the evolution of an arbitrary quan- also discussed by Davies (1975) and it is related to the tum system. An observer at rest (Alice) can describe the fluctuation-dissipation theorem (Candelas and Sciama, quantum evolution on consecutive parallel slices of space- 1977) and to the Hawking effect that will be dicussed in 17 time, t = const. What can Bob, the accelerated observer, the next section. A rigorous proof of the Unruh ef- do? From Fig. 3, one sees that there is no communica- fect in Minkowski spacetime was given by Bisognano and tion whatsoever between him and the region of spacetime Wichmann (1976) in the context of axiomatic quantum that lies beyond both horizons. Where Alice sees a pure field theory, thus establishing that the Unruh effect is not state, Bob has only a mixed state. Some information is limited to free field theory. lost. We shall return to this subject in the next, final section. t

pasthorizon VI. BEYOND SPECIAL RELATIVITY

It took Einstein more than ten years of intensive work to progress from special relativity to general relativity. x Despite its name, the latter is not a generalization of the special theory, but a radically different construct: spacetime is not only a passive arena where dynamical processes take place, but has itself a dynamical nature. At this time, there is no satisfactory quantum theory of futurehorizon gravitation (after seventy years of efforts by leading the- oretical physicists). In the present review on quantum information theory, we shall not attempt to use the full machinery of general FIG. 3 Dependence of the spin entropy S, in Bob’s frame, relativity, with Einstein’s equations.18 We still consider on the values of the angle θ and a parameter Γ ≈ [1 − (1 − spacetime as a passive arena, endowed with a Rieman- β2)1/2]∆/mβ, where ∆ is the momentum spread in Alice’s nian metric, instead of the Minkowski metric of special frame. relativity. The difference between them is essential: it is necessary to introduce notions of topology, because it For any reasonable acceleration, the Unruh tempera- may be impossible to find a single coordinate system that ture is incomparably smaller that the black-body temper- covers all of spacetime. To achieve that result, it may be ature of the cosmic background, or any temperature ever necessary to use several coordinate patches, sewed to each attained in a laboratory, and is not observable. Levin, other at their boundaries. Then in each patch, the metric Peleg, and Peres (1992) considered the effect of shield- is not geodesically complete: a geodesic line stops after a ing a hypothetical experiment from any parasitic sources. finite length, although there is no singularity there. The This, however, creates a radically new situation, because presence of singularities (points of infinite curvature) is the presence of a boundary affects the dynamical prop- another consequence of Einstein’s equations. It is likely erties of the quantum field by altering the frequencies of that these equations, which were derived and tested for its normal modes. Finite-size effects on fields have been the case of moderate curvature, are no longer valid under known for a long time, both theoretically (Casimir, 1948) such extreme conditions. We shall not speculate on this and experimentally (Spaarnay, 1958). Levin, Peleg, and issue, and we shall restrict our attention to the behav- Peres showed that if the detector is accelerated together ior of quantum systems in the presence of horizons, in with the cavity that shields it, it will not be excited by particular of black holes. Before we examine the latter, the vacuum fluctuations of the field. On the other hand, let us first return to entanglement, now in curved space- an inertial detector freely falling within such an acceler- time, and to the Unruh effect, still in flat spacetime, but described now in an accelerated coordinate system.

17 Properties of detectors undergoing circular acceleration, as in high energy accelerators, were investigated by Bell and Leinaas 18 Concepts of quantum information were recently invoked in sev- (1983), Levin, Peleg, and Peres (1993), and by Davies, Dray, and eral problems of quantum gravity and quantum cosmology, but Manogue (1996). we restrict ourselves to conventional black hole physics. 24

A. Entanglement revisited a Cauchy surface for specifying initial values, whose do- main of development is the entire spacetime (Hawking Calculations on EPRB correlations require a common and Ellis, 1973; Wald, 1984 and 1994). The vacuum state reference frame. Only then can statements such as “if 0R obtained in this construction is called a Rindler vac- 1 1 uum.| i It is a natural vacuum for observers who move on m1z= 2 , then m2z=– 2 ” have an operational meaning. In a curved space we can choose an arbitrary frame at one orbits like in Fig. 3, with different positive values of the spacetime point and then translate it parallel to itself acceleration a. 1 As a consequence of the Reeh-Schlieder theorem, it along a geodesic path. For example, spin- 2 particles may be sent to Alice and Bob, far away. After a reference follows that a Minkowski vacuum Ω corresponds to a | i frame is chosen at the emission point, local frames are mixed state in the Rindler spacetime. To relate the established for them by parallel transport along the parti- Minkowski and Rindler Hilbert spaces, fields in both cles’ trajectories. However, particles only approximately wedges are required. The relation between the standard follow classical geodesic trajectories, and this inevitably Minkowski Fock space and a tensor product of Rindler introduces uncertainties in the definition of directions. Fock spaces is given by a formally unitary operator U, Using path integral methods, von Borzeszkowski and whose action on the Minkowski vacuum is Mensky (2000) have shown that if certain conditions are ∞ met, approximate EPR correlations still exist, but “the U Ω = exp( nπω /a) n n , (71) | i − i | iLi ⊗ | iRi longer the propagation and the stronger the gravitational Yi nX=0 field, the poorer is the correlation”. where ωi denotes the frequencies of the modes of the One of the difficulties of quantum field theory in curved Rindler fields, and ni are the corresponding occupa- spacetimes is the absence of a unique (or preferred) tion numbers. The above expression suggests that the Hilbert space, the reason being that different represen- Minkowski vacuum has the structure of a maximally tations of canonical commutation or anticommutation entangled state when viewed by accelerated observers. relations lead to unitarily inequivalent representations When restricted to only one wedge, the state becomes (Emch, 1972; Bogolubov et al., 1990; Haag 1996). For the Minkowski spacetime, the existence of a preferred ∞ 1 ρ = exp( nπω /a)Z− n n , (72) vacuum state enables us to define a unique Hilbert space − i i | iRih iR| representation. A similar construction is also possible in Yi nX=0 stationary curved spacetimes (Fulling, 1989; Wald, 1994). where the ith mode was normalized by Zi = However, in a general globally hyperbolic spacetime this exp( nπω /a). That state indeed produces a ther- i − i is impossible, and one is faced with multiple inequivalent malP density matrix ρ exp( HR/T ), where HR is the ∝ − ~ representations. field Hamiltonian for region R, and T = a /2πckB . We Genuinely different Hilbert spaces with different den- can now calculate the entanglement of the Minkowski sity operators and POVMs apparently lead to predictions vacuum as seen by an accelerated observer. A natural that depend on the specific choice of the method of cal- reduced density matrix is ρ itself, which is a singular culation. The algebraic approach to field theory can re- state (in the sense of Sec. V.C) of an infinite thermal solve this difficulty for PVMs. The essential ingredient bath. Its entropy is infinite, which is in agrement with is the notion of physical equivalence (Emch, 1972; Araki, the previous discussion, since the energy of such a system 1999; Wald, 1994), which allows to extend the formalism is also infinite. of POVMs and CP maps to general globally hyperbolic The relationship between Minkowski and Rindler wave spacetimes (Terno, 2002). packets was analyzed by Audretsch and M¨uller (1994a). The simplest example of inequivalent representations These authors also discussed local detection by Rindler occurs in the discussion of the Unruh effect, when we observers and EPR-like correlations (Audretsch and wish to use quantum field theory in the Rindler wedge M¨uller, 1994a, b). x > t where the detector moves, or in the opposite Alsing and Milburn (2002, 2003) examined the fidelity wedge| x| < t , which is causally separated from it, or of teleportation from Alice in an inertial frame to Bob in both wedges−| | together. Each one of the two wedges, or who is uniformly accelerated. Assume that qubits are both together, can be considered as spacetimes on their realized by some mode ω of the electromagnetic field, own right (Rindler spaces), where a global timelike field and that Alice’s state is Ψ = α Ω + β 1 , where Ω | i | i | i | i is obtained from the set of all hyperbolas with different is the Minkowski vacuum. Then the best state that Bob values of the acceleration (Wald, 1984). can hope to get is The transformation between Minkowski and Rindler Ψ′ = α 0R + β 1R , (73) wedge descriptions are unitary only formally (Unruh and | i | i | i Wald 1984; Wald 1994) and algebraic field theory should where 0R is the Rindler vacuum, and some mode ω′ (as be used to give a rigorous interpretation to these for- seen by| Bob)i was chosen for his realization of qubits. The mal expressions (Emch, 1972; Haag, 1996). A quantum fidelity of teleportation Ψ Ψ′ then decreases with field theory can be defined in a standard way because Bob’s acceleration. It also| i depends → | i on time: the fidelity the Rindler spaces are globally hyperbolic. They admit of course vanishes when Alice is behind Bob’s horizon. 25

B. The thermodynamics of black holes a black hole is present: if we drop ordinary matter into a black hole, it will disappear into a spacetime singularity, Black holes result from concentrations of matter so together with its entropy S. No compensating gain of large that their gravity pull prevents the escape of light entropy occurs, so that the total entropy in the universe (Michell, 1784; Laplace, 1795). In other words, a fu- decreases. One could attempt to salvage the second law ture horizon is formed. While Unruh’s horizons were for by invoking the bookkeeping rule that one must continue observers whose asymptotic speed approaches c, a black to count the entropy of matter dropped into a black hole hole horizon affects every observer. We now present ba- as still contributing to the total entropy of the universe. sic facts of black hole physics, limiting ourselves almost However, the second law would then be observationally exclusively to spherically-symmetric spacetimes. The lit- unverifiable. erature on black holes is voluminous, and our sketch gives It was noted by Bekenstein (1972, 1974) that proper- just a glimpse of this fascinating subject. Our main ties of the horizon area of a stationary black hole resemble sources for classical black hole physics were Landau and those of entropy. In the most general case, a stationary Lifshitz (1975), Hawking and Ellis (1973), Wald (1984) black hole is characterized by three parameters: its mass and Frolov and Novikov (1998). For quantum aspects, we M, angular momentum J and charge Q. The first law consulted Birrell and Davies (1982), Wald (1994), Brout of black hole dynamics (Bardeen, Carter, and Hawking, et al. (1995), and Frolov and Novikov (1998). An ex- 1973; Iyer and Wald, 1994) states that tensive survey of black hole thermodynamics was given κ by Wald (1999, 2001). In this section, unless otherwise dM = dA +ΩdJ +ΦdQ, (76) stated, c = G = ~ = 1. 8π Spacetime outside a spherically symmetric distribu- where Ω is the angular velocity and Φ the electric poten- tion of matter (and hence outside an incipient black tial. This relation is formally identical to the first law of hole during all stages of its collapse) is described by the thermodynamics, if we identify temperature with surface Schwarzschild metric, gravity and entropy with horizon area. We would then

2 2 1 2 2 2 have ds = (1 2M/r)dt (1 2M/r)− dr r dΩ . (74) − − − − κ ~c3 A T = , S = , (77) The proper time of a stationary observer is dτ = 2 2π GkB 4lP √gttdt = 1 2M/rdt, and the radial distance is dl = − 1/2 √ grrdr =p (1 2M/r)− dr. This metric has a coordi- where l = ~G/c3 is the Planck length, and ordinary − − P nate singularity at r =2M, which can be removed by a units were restored.p transition to various alternative coordinate systems. As Bekenstein (1972, 1974) proposed to assign to a black we shall see, it is a kind of “boundary” of the black hole. hole of area A an entropy On the other hand, the singularity at r = 0 is physical: 3 ~ the spacetime curvature diverges there. SBH = Ac /4 G, (78) Spacetimes may have symmetries. If translation along a family of curves leaves the metric invariant, the field of thus elevating a formal analogy to the status of a physical tangent vectors to these curves is called a Killing field law. Hawking (1974) found that a black hole radiates like (Killing, 1892). Killing vectors χµ have many useful a black body at temperature T , and thereby put the anal- properties. ogy between black hole mechanics and thermodynamics For example, the Schwarzschild metric is invariant un- on firm ground. der time translations, t t + τ. The corresponding There are many ways to explain Hawking radiation Killing vector χµ = (1, 0, 0→, 0) is timelike for r> 2M and (Hawking, 1975; Wald, 1975; Birrell and Davies, 1982; spacelike for r < 2M. It becomes null on the horizon. Fredenhagen and Haag, 1990; Wald, 1994; Brout et al., The surface gravity κ, which characterizes the strength 1995). Here, we follow the informal presentation of Frolov of gravitational field near the horizon, is defined as and Novikov (1998), which is based on the analogy with pair creation by an external static field. Actually, space- κ = lim(aα), (75) time is not static when a star collapses into a black hole and later evaporates. However, usually it is an excellent where a is the norm of the proper four-acceleration of approximation to treat it as static. A rigorous analysis a stationary object, and α is a red-shift factor. For along these lines was made by Brout et al. (1995). Schwarzschild black holes, α = √gtt and κ = 1/4M. Similarities between pair production and Hawking ra- Hawking and Ellis (1973), and Wald (1984, 1999) de- diation were discussed by M¨uller, Greiner and Rafeski scribe many properties of the surface gravity. Bardeen, (1977). Let Γ be the field strength and g the charge. Carter, and Hawking (1973), have shown that κ is con- By analogy with the tunnel effect, the probability that a stant over the horizon of any stationary black hole. This virtual pair of particles be found at a distance l from l/λ is known as a the zeroth law of black hole mechanics. one another is approximately e− , where λ is the Even in classical general relativity, there is a serious Compton wavelength. A pair may turn to be real if difficulty with the second law of thermodynamics when gΓl 2mc2. Thus, the probability of particle creation is ≥ 26 w exp( ζm2c3/~gΓ), where the numerical constant ζ Hawking’s radiation resolved the thermodynamic diffi- can∝ be obtained− by a more detailed calculation. culty only to introduce another puzzle. An inevitable re- A naive application of this formula to particle creation sult of that radiation is the evaporation of the black hole in a static gravitational field turns out to give not only after a finite time (see Appendix B). Since the emitted the right result, but also some valuable insights. In par- particles are overwhelmingly massless, black hole evapo- ticular, conservation of energy implies that a static gravi- ration leads to baryon number non-conservation. tational field can create particles only if there are regions Hawking (1976, 1982) also introduced a superopera- with timelike Killing fields and others with spacelike ones; tor to describe the quantum state evolution during the a horizon is needed. A static gravitational field with- black hole formation and evaporation (see Appendix B). out horizons cannot create particles (Birrell and Davies, A detailed analysis of this superoperator was made by 1982; Wald, 1984). A black hole emits particles as if it Strominger (1996). It is (at least formally) completely were a black body with temperature positive and as such it is a perfectly normal operation of quantum information theory (Terno, 2002). T = κ/2πkB , (79) Yet, it has often been asserted that the evolution of an as in Eq. (77). initial pure state into a final mixed state conflicts with The generalized second law of thermodynamics quantum mechanics, and this issue is usually referred to (Bekenstein, 1974; Frolov and Page, 1993; Wald, 1994; as the “black hole information loss paradox.” These pes- Frolov and Novikov, 1998) states that simistic views are groundless. When black hole thermo- dynamics appeared in the 70’s, notions such as POVMs ∆S + ∆SBH 0. (80) and completely positive maps were unknown to the rel- ≥ ativistic community. Today, we know that the evolution An informational analysis of this law by Hosoya, Carlini, of pure states into mixtures is the general rule when a and Shimomura (2001) clarified its relation to classical classical intervention is imposed on a quantum system, bounds on accessible information (Levitin, 1969, 1987; as we have seen in Sec. II. In the present case, the clas- Holevo, 1973). Bekenstein and Mayo (2001) and Beken- sical agent is the spacetime metric itself, which is bor- stein (2002) gave a description of the information absorp- rowed from classical general relativity in the absence of tion and emission by black holes in terms of quantum a consistent quantum gravity theory. Attempts to in- channels. troduce a hybrid quantum-classical dynamics by using A natural question is what (and where) are the de- the Koopman (1931) formalism are not mathematically grees of freedom responsible for the black hole entropy. inconsistent, but they violate the correspondence prin- On this issue, there are conflicting views. It is not clear ciple and are physically unacceptable (Peres and Terno, whether we should think of these degrees of freedom as 2001). Anyway, the evolution of an initial pure state into residing outside the black hole in its thermal atmosphere, a final mixed state is naturally accomodated within the or on the horizon in Chern-Simons states, or inside the framework of the algebraic approach to quantum theory black hole, associated with what classically corresponds (Wald, 1994), and that of a generalized quantum theory to the singularity deep within it. Or perhaps the micro- (Hartle, 1998). scopic origin of SBH is the entanglement between Hawk- The final fate of black holes and its relation to the in- ing particles inside and outside the horizon (Bombelli formation paradox were discussed by Preskill (1993), ’t et al. et al. , 1986; Ashtekar , 1994; Iorio, Lambiase, and Hooft (1996, 1999) and Frolov and Novikov (1998). How- Vitiello, 2001). It is likely that in order to gain a bet- ever, this issue may be conclusively resolved only after ter understanding of the degrees of freedom responsible there is a consistent theory of quantum gravity, allow- for black hole entropy, it will be necessary to achieve ing meanwhile for a number of tantalizing speculations. a deeper understanding of the notion of entropy itself Here we present five of the most popular alternatives of (Zurek, 1990). what happens with the “information” when a black hole Suppose now that the matter that has fallen inside the evaporates. horizon had quantum correlations with matter that re- mained outside. How is such a state described by quan- Information is lost: Hawking’s superscattering that tum theory? Are these correlations observable? This • was described above is a fundamental feature of problem is not yet fully understood, although such cor- quantum theory and not just an effective descrip- relations play an essential role in giving to Hawking ra- tion. diation a nearly exact thermal character (Wald, 1975). It is hard to imagine a mechanism for restoring the cor- There is no information loss: if the spectrum is an- relations during the process of black hole evaporation. • alyzed carefully, there may be enough non-thermal On the other hand, if the correlations between the inside features to encode all the information. Bekenstein and the outside of a black hole are not restored during (1993) showed that deviations of the Hawking ra- the evaporation process, then by the time that the black diation from the black body spectrum may help hole has evaporated completely, an initial pure state will reconstruct part of the information. Hod (2002) have evolved to a mixed state, and some “information” estimated that, under suitable assumptions about will have been lost. black hole quantization, the maximal information 27

emission rate may be sufficient to recover all the in- between the loss of undetected signals, false alarms formation from the resulting discrete spectrum of (dark counts), and our knowledge of the location of the radiation. recorded events. A quantitative discussion of this problem would be most welcome. Information comes out at the end, at the Plank • scale physics. Frolov and Vilkovisky (1981) con- It is possible to indicate the approximate orienta- structed a model that provides for this possibility. • tion of a Cartesian frame by means of a few suitably prepared spins (Bagan, Baig, and Mu˜noz-Tapia, There is a stable black hole remnant with about the • 2001), or even a single hydrogen atom (Peres and Planck mass (0.02 µg) and information is somehow Scudo, 2001). Likewise, the quantum transmission encoded in it (Aharonov, Casher, and Nussinov, of the orientation of a Lorentz frame should be 1987). possible. This problem is much more difficult, be- Information escapes to baby universes, that are cre- cause the Lorentz group is not compact and has no • ated instead of true singularities (Zel’dovich, 1977; finite-dimensional unitary representations (Wigner, Hawking, 1988). The overall evolution of the entire 1939). multiverse is unitary, but since baby universes are Progressing from special to general relativity, what causally unconnected to our universe and the total • is the meaning of parallel transport of a spin? In a state is entangled, we perceive a loss of information. curved spacetime, the result is obviously path de- pendent. Then what does it mean to say that a Still a different scenario is implied by the works of Ger- pair of distant particles is in a singlet state? As lach (1976) and Boulware (1976): a particle that falls the rotation group O(3) is not a valid symmetry, into an eternal black hole crosses the horizon after an in- the classification of particles, even the usefulness of finite amount of the coordinate time t, but only a finite the concept of a particle, become doubtful. Meth- amount of its own proper time. On the other hand, the ods are known for quantization of higher spin fields evaporation of a black hole takes a finite amount of the in a curved background (Birrell and Davies, 1982; coordinate time, which is the physical time of a distant Wald, 1994), but what is the operational meaning observer (see Appendix B). From the point of view of of the resulting states and POVMs? the infalling observer, the horizon always appears to re- We still need a method for detection of relativistic cede before her, until it finally disappears (or shrinks to • the Planck scale) and the region “beyond the horizon” is entanglement that involves the spacetime proper- unattainable. The distant observer sees the infalling one ties of the quantum system, such as a combination quickly arrive arbitrarily close to the effective horizon, of localization and spin POVMs (in flat or curved then she is nearly “frozen” there for an exceedingly long metric backgrounds). time, and finally either the black hole evaporates or the After all these problems have been solved, we’ll still universe collapses. Therefore it makes no sense to assert • have to find a theory of the quantum dynamics for that states having (essential) support on the part of the the spacetime structure. Cauchy surface that lies beyond the horizon would be correlated with an outgoing Hawking radiation and then mysteriously disappear. There is no issue of information Acknowledgments and apologies loss at all (Sonego, Almergren, and Abramowitz, 2000; Alberghi et al., 2001).19 We are grateful to numerous friends for helping us lo- cate references. We apologize if we missed some relevant ones. Only in a few cases, the omission was intentional. C. Open problems Work by AP was supported by the Gerard Swope Fund The good news are that there is still plenty of work and the Fund for Promotion of Research. Part of the to be done. Here we shall mention a few problems that work by DRT was carried at the Technion—Israel Insti- appear interesting and from which more physics can be tute of Technology, and was supported by a grant from learnt. the Technion Graduate School. As mentioned in Sec. V.B, quantum field theory im- • plies a trade-off between the reliability of detectors APPENDIX A: Relativistic states transformations and their localization. This is an important prac- tical problem. A proper balance must be found In this Appendix we list the conventions we used and outline the transformation rules for free particle states. Details can be found in the treatises of Bogolubov, Lo- gunov and Todorov (1975), and Weinberg (1995). Ex- 19 We have listed this opinion last, because it is the one we tend to plicit forms of the transformation laws for massive parti- support. cles are given by Halpern (1968), Bogolubov et al.(1975), 28

and Ahn et al. (2003); for massless particles, by Lind- kS = (1, 0, 0, 1). Explicit formulas for L(p) in the mas- ner, Peres, and Terno (2003) and Bergou, Gingrich, and sive and massless cases are given in the books of Halpern Adami (2003). (1968), Bogolubov et al. (1990), and Weinberg (1995). Unless stated otherwise, we chose the following con- Wave functions having a distribution of momenta ventions for states and related operators: transform as

σ, p =ˆa† Ω , (A1) σp ψ′ (q) = ξ, q U(Λ) dµ(p)ψσ(p) σ, p , | i | i ξ h | Z | i Xσ and = dµ(p)ψσ(p)Dχσ[W (Λ,p)] ξ, q χ, Λp , σ, p ξ, q = (2π)3(2p0)δ δ(3)(p q), (A2) Z h | i h | i σξ − Xσ,χ 1 1 where p0 E(p)= m2 + p2. One-particle states are = Dξσ[W (Λ, Λ− p)]ψσ(Λ− p), (A9) ≡ Xσ p ∞ Ψ = ψσ(p) σ, p dµ(p), (A3) so the same state in the boosted frame is | i Z | i Xσ −∞ ∞ 1 1 Ψ′ = Dσξ[W (Λ, Λ− p)]ψξ(Λ− p) σ, p dµ(p). with the Lorentz-invariant measure | i Z | i Xσ,ξ −∞ (A10) dµ(p)= d3p/(2π)3(2p0). (A4) Explicit expressions for D[W ] are given in Section IV and The wave functions Ψ satisfy in the references cited above. | i σ, p Ψ = ψ (p), (A5) spin h | i σ D and Λ momentum Ψ Φ = ψ∗ (p)φσ(p)dµ(p). (A6) h | i Z σ classical info Xσ If we want to be more explicit about the spin degrees FIG. 4 Relativistic state transformation as a quantum cir- of freedom, we use 2-spinor notations: a pure state of cuit: the gate D which represents the matrix D [W (Λ,p)] definite momentum and arbitrary spin is α p . The one- ξσ β | i is controlled by both the classical information and the mo- to-one correspondence with Dirac’s notation
is explained mentum p, which is itself subject to the classical information by Bogolubov, Logunov and Todorov (1975). Λ. Under a classical, geometric Lorentz transformation µ µ ν y =Λ ν x , the unitary transformation of the basis vec- tors (A.1) is APPENDIX B: Black hole radiation U(Λ) σ, p = D [W (Λ,p)] ξ, Λp , (A7) | i ξσ | i Xξ The energy radiated by a black hole satisfies approxi- mately the Stefan-Boltzmann law, (Frolov and Novikov, where Dξσ are matrix elements of the unitary operator D 1998; Brout et al., 1995) so the rate of mass loss due to that corresponds to the Wigner rotation W (Λ,p), given energy conservation is by Eq. (A8) below. 4 2 Note that the spin rotation depends on the value of the M˙ T A M − , (B1) momentum (spin is a secondary variable, as defined in ∝− ∝− Sec. IV). The quantum circuit in Fig. 4 gives a graphical where A is the horizon area and time is that of a dis- 20 1 representation of primary vs. secondary variables. tant observer. It can be shown that a relation T M − The Wigner rotation matrix is given by holds in quasi-static changes of mass at all stages of∝ evap- oration. Numerical coefficients were calculated by Page 1 W (Λ,p) := L− (Λp)ΛL(p), (A8) (1976). A back hole of initial mass M0 (not too small) evaporates after a time where L(p) is a “standard boost” which transforms a “standard four-momentum” kS into p. For massive par- 3 tE = aM0 , (B2) ticles kS = (m, 0, 0, 0), while for massless ones it is 9 3 where a = 4.9 10− sec/kg . Together with (B1), this gives the following× expression for the mass

20 This representation was suggested to us by Barbara Terhal. M(t)= M (1 t/t )1/3. (B3) 0 − E 29

The duration of the steady-state radiation build-up is in- Baumgartel, H., and M. Wollenberg, 1992, Causal nets of op- comparably shorter than tE (Wald, 1994; Brout et al., erator algebras: mathematical aspects of algebraic quan- 1995), so that the above expression is a good approxi- tum field theory (Akademie, Berlin). mation. Hence it takes a time comparable to the age of Bechmann-Pasquinucci, H., and N. Gisin, 1999, Phys. Rev. the universe for a black hole of mass 5 1014g (and ra- A 59, 4238. dius of atomic size) to evaporate completely× (Frolov and Bechmann-Pasquinucci, H., and A. Peres, 2000, Phys. Rev. Novikov, 1998). Lett. 85, 3313. Hawking (1976, 1982) introduced a superoperator Beckman, D., D. Gottesman, M. A. Nielsen, and J. Preskill, (originally called “superscattering operator”), that was 2001, Phys. Rev. A 64, 052309. mentioned in Sec. VI.B to describe the quantum state Beckman, D., D. Gottesman, A. Kitaev, and J. Preskill, 2002, evolution during the black hole formation and evapora- Phys. Rev. D 65, 065022. tion. In standard scattering theory, a unitary S-matrix Bekenstein, J. D., 1972, Lett. Nuovo. Cim. 4, 737. relates the density matrix of final states with to of the Bekenstein, J. D., 1974, Phys. Rev. D 9, 3292. out in incoming states: ρ = Sρ S†. For a spacetime with an Bekenstein, J. D., 1993, Phys. Rev. Lett 70, 3680. evaporating black hole, S would map states from (the Hin Bekenstein, J. D., 2002, in Advances in the Interplay be- states in the distant past, when the black hole did not tween Quantum and Gravity Physics, edited by P. G. exist yet) to the tensor product of out (the states that Bergmann and V. de Sabbata (Kluwer, Dordrecht, The reach infinity and are accessible toH a distant observer) Netherlands) p. 1 [e-print gr-qc/0107049]. and the Hilbert space of states that fell into the black Bekenstein, J. D., and A. E. Mayo, 2001, Gen. Rel. Grav. hole. This splitting is a standard step in many deriva- 33, 2095. tions of the Hawking radiation (Wald, 1994). Since only Bell J. S., 1964, Physics 1, 195. the states that reach infinity are accessible to a distant Bell, J. S., and J. M. Leinaas, 1983, Nucl. Phys. B 212, 131. observer, the final density matrix is calculated by tracing Bennett, C. H. and G. Brassard, 1984, in Proceedings of out the black hole, IEEE International Conference on Computers, Systems and Signal Processing, Bangalore, India (IEEE, New out in ρ = tr BH Sρ S†. (B4) York, 1984) p. 175. Bennett, C. H., G. Brassard, S. Breidbart, and S. Wies- ner, 1983, in Advances in Cryptology (Proceedings of Crypto-82, Plenum, New York) p. 267. Bennett, C. H., G. Brassard, C. Cr´epeau, R. Jozsa, A. Peres, References and W. Wootters, 1993, Phys. Rev. Lett. 70, 1895. Aharonov Y., and D. Z. Albert, 1981, Phys. Rev. D 24, 359. Bennett, C. H., D. P. DiVincenzo, C. A. Fuchs, T. Mor, E. 29 Rains, P. W. Shor, J. A. Smolin, and W. K. Wootters, Aharonov Y., and D. Z. Albert, 1984, Phys. Rev. D , 228. 1999, Phys. Rev. A 59, 1070. Aharonov, Y., A. Casher, and S. Nussinov, 1987, Phys. Lett. Bergou, A. J., R. M. Gingrich, and C. Adami, 2003, e-print B 194, 38. quant-ph/0302095. Ahn, D., H.-J. Lee, Y. H. Moon, S. W. Hwang, 2003, Phys. Bialynicki-Birula, I., 1996, Progress in Optics XXXVI , Rev. A 67, 012103. edited by E. Wolf (Elsevier, Amsterdam), p. 245. Alberghi, G. L., R. Casadio, G. P. Vacca, and G. Venturi, Birrell, N. D., and P. C. W. Davies, 1982, Quantum Fields 2001, Phys. Rev. D 64, 104012. in Cureved Space (Cambridge University, Cambridge, Alsing, P. M., and G. J. Milburn, 2002, Quant. Inf. Comp. UK). 2, 487. Bisognano, J. J., and E. H. Wichmann, 1976, J. Math. Phys. Alsing, P. M., and G. J. Milburn, 2003, e-print 17, 303. quant-ph/0302179. Blanchard, Ph., and A. Jadczzik, 1996, Found. Phys. 26, Amosov, G. G., A. S. Holevo, and R. F. Werner, 2000, Probl. 1669. Info. Transmission 36, 305. 42 Blanchard, Ph., and A. Jadczzik, 1998, Int. J. Theor. Phys. Amrein, W. O., 1969, Helv. Phys. Acta , 149. 37, 227. Mathematical Theory of Quantum Fields Araki, H., 1999, Bloch, I., 1967, Phys. Rev. 156, 1377. (Oxford University, Oxford,UK). Bogolubov, N. N., A. A. Logunov, A. I. Oksak, and I. T. Ashtekar, A., J. Baez, A. Corichi, and K. Krasnov, 1998, Todorov, 1990, General Principles of Quantum Field Phys. Rev. Lett. 80, 904. Theory (Kluwer, Dordrecht, The Netherlands). Audretsch, J., and R. M¨uller, 1994a, Phys. Rev. D 49, 4056. Bogolubov N. N., A. A. Logunov, and I. T. Todorov, 1975, 49 Audretsch, J., and R. M¨uller, 1994b, Phys. Rev. D , 6566. Introduction to Axiomatic Quantum Field Theory (Ben- Bagan, E., M. Baig, and R. Mu˜noz-Tapia, 2001, Phys. Rev. jamin, New York, NY). 87 Lett. , 257903. Bohm, D., 1951, Quantum Theory (Prentice-Hall, New York, Balescu, R., and T. Kotera, 1967, Physica (Utrecht) 33, 558. NY) Ballentine, L. E., 1970, Rev. Mod. Phys. 42, 358. Bohr, N., 1927, in Atti del Congresso Internazionale dei Bardeen, J. M., B. Carter, and S. W. Hawking, 1973, Comm. Fisici, Como; reprinted in Nature (London) 121, 78, Math. Phys. 31, 161. 580 (1928). 30

Bohr, N., 1939, in New Theories in Physics, edited by Inter- Einstein A., B. Podolsky and N. Rosen, 1935, Phys. Rev. 47, national Institute of Intellectual Cooperation (Paris). 777. Bohr, N. 1949, in Albert Einstein: Philosopher-Scientist, Eisert J., C. Simon, and M. B. Plenio, 2002, J. Phys. A: edited by P. A. Schilpp (Library of Living Philosophers, Math. Gen. 35, 3911. Evanston, IL). Emch, G. G., 1972, Algebraic Methods in Statistical Mechan- Bohr, N., and L. Rosenfeld, 1933, Mat. Fys. Medd. Dan. ics and Quantum Field Theory (Wiley-Interscience, Vidensk. Selsk. 12 (8). New York, NY). Bombelli, L. R., R. Koul, J. Lee, and R. Sorkin, 1986, Phys. Emch, G. G., and C. Liu, 2002, The Logic of Thermostatis- Rev. D 34, 373. tical Physics (Springer, Berlin). Borde, A., L. H. Ford, and T. A. Roman, 2002, Phys. Rev. Epstein, H., V. Glaser, and A. Jaffe, 1965, Nuovo Cim. 36, D. 65, 084002. 1016. Boulware, D. G., 1976, Phys. Rev. D 13, 2169. Finkelstein, D., 1988, in The Universal Turing Machine, A Bratteli, O., and D. W. Robinson, 1987, Operator Algebras Half-Century Survey, edited by R. Herken (Oxford Uni- and Quantum Statistical Mechanics (Springer, New versity, Oxford, UK) p. 349. York, NY) 2 volumes, 2nd edition. Florig, M., and Summers, S. J., 1997, J. Math. Phys. 38, Braunstein, S. L., A. Mann, and M. Revzen, 1992, Phys. Rev. 1318. 68 Lett. , 3259. Fredenhagen, K., 1985, Commun. Math. Phys. 97, 461. Brout, R., S. Massar, R. Parentani, Ph. Spindel, 1995, Phys. Frolov, V. P., and I. D. Novikov, 1998, Black Hole Physics 260 Rep. , 329. (Kluwer, Dordrecht, The Netherlands). 81 Bruß, D., 1998, Phys. Rev. Lett. , 3018. Frolov, V. P., and D. N. Page, 1993, Phys. Rev. Lett. 71, Bruß, D., and Macchiavello, C., 2002, Phys. Rev. Lett. 88, 3902. 127901. Frolov, V. P., and G. A. Vilkovisky, 1981, Phys. Lett. B 106, Buchholz, D., and K. Fredenhagen, 1982, Commun. Math. 307. 84 Phys. , 1. Fuchs, C. A, and A. Peres, 2000, Physics Today 53 (3), 70. Busch, P., 1999, J. Phys. A: Math. Gen. 32, 6535. Fuchs, C. A., and J. van de Graaf, 1999, IEEE Trans. Info. Buttler, W. T., R. J. Hughes, S. K. Lamoreaux, G. L. Mor- Theory gan, J. E. Nordholt, and C. G. Peterson, 2000, Phys. Fulling, S. A., 1989, Aspects of Quantum Field Theory Rev. Lett. 84, 5652. in Curved Space-Time (Cambridge University, Cam- 38 Candelas, P., and D. W. Sciama, 1977, Phys. Rev. Lett. , bridge, UK). 1372. 14 51 Gerlach, U. H., 1976, Phys. Rev. D , 1479. Casimir, H. G. B., 1948, Proc. Kon. Akad. Wetenschap. , 39 793. Giannitrapani, R., 1998, J. Math. Phys. , 5180. 89 Caves, C. M., 1982, Phys. Rev. D 26, 1817. Gingrich, R. M., and C. Adami, 2002, Phys. Rev. Lett. , 270402. Chuang, I. L, and M. A. Nielsen, 1997, J. Mod. Opt. 44, 2455. Gisin, N., G. Ribordy, W. Tittel, and H. Zbinden, 2002, Rev. Mod. Phys. 74, 145. Cirel’son, B. S., 1980, Lett. Math. Phys. 4, 93. Glauber, R. J., 1986, in New Techniques and Ideas in Quan- Clauser, J. F., M. A. Horne, A. Shimony, and R. A. Holt, tum Measurement Theory, edited by D. M. Green- 1969, Phys. Rev. Lett. 23, 880. berger, Ann. New York Acad. Sci. 480, 336. Currie, D. G., T. F. Jordan, and E. C. G. Sudarshan, 1963, Groisman, B., and B. Reznik, 2002, Phys. Rev. A 66, 022110. Rev. Mod. Phys. 35, 350. Local Quantum Physics: Fields, Particles, Czachor M., 1997, Phys. Rev. A 55, 72. Haag, R., 1996, Algebras (Springer, Berlin). Davies, E. B., 1976, Quantum Dynamics of Open Systems 1 (Academic Press, New York, NY). Haag, R., and J. A. Swieca, 1965, Commun. Math. Phys. , 308. Davies, P. C. W., 1975, J. Phys. A: Math. Gen. 8, 609. Hacyan, S., 2001, Phys. Lett. A 288, 59. Davies, P. C. W., T. Dray, and C. A. Manogue, 1996, Phys. Rev. D 53, 4382. Halpern, F. R., 1968 Special Relativity and Quantum Me- chanics (Prentice-Hall, Englewood Cliffs, NJ), pp. 80 Detweiler, S., 1982, Black Holes. Selected Reprints, (Am. As- and 134. soc. of Phys. Teachers, Stony Brook, NY) Black Holes and Relativistic Stars Dicke, R. H., 1981, Am. J. Phys. 49, 925. Hartle, J. B., 1998, in , edited by R. M. Wald (University of Chicago Press, Dieks, D., 1982, Phys. Lett. A 92, 271. Chicago, IL). Dirac, P. A. M., 1947, The Principles of Quantum Mechanics Hawking, S. W., 1974, Nature 248, 30. (Oxford University, Oxford, England) p. 36. Hawking, S. W., 1975, Commun. Math. Phys. 43, 199. Drell, S., 1978, Am. J. Phys. 46, 597; Physics Today 31 (6), 14 23. Hawking, S. W., 1976, Phys. Rev. D , 2460. 87 Eggeling, T., D. Schlingemann, and R. F. Werner, 2002, Eu- Hawking, S. W., 1982, Commun. Math. Phys. , 395. rophys. Lett. 57, 782. Hawking, S. W., 1988, Phys. Rev. D 37, 904. Einstein, A., 1949, in Albert Einstein: Philosopher-Scientist, Hawking, S. W., and G. F. R. Ellis, 1973, The Large Scale edited by P. A. Schilpp (Library of Living Philosophers, Structure of Space-Time (Cambridge University, Cam- Evanston, IL) pp. 85, 683. bridge, UK). 31

Hay, O., and A. Peres, 1998, Phys. Rev. A 58, 116. Makeenko, Y., 2002, Methods of Contemporary Gauge The- Hegerfeldt, G. C., 1985, Phys. Rev. Lett. 54, 2395. ory (Cambridge University, Cambridge, UK). Herbert, N., 1981, Found. Phys. 12, 1171. Mandel, L., 1966, Phys. Rev. 144, 1071. Hillery, M., R. F. O’Connell, M. O. Scully, and E. P. Wigner, Mandel, L., and E. Wolf, 1995, Optical Coherence and Quan- 1984, Phys. Rep. 106, 121. tum Optics (Cambridge University, Cambridge, UK). Hod, S., 2002, Phys. Lett. A 299, 144. Mayers, D., 1997, Phys. Rev. Lett. 78, 3414. Holevo, A. S., 1973, Probl. Inform. Transmission 9, 110, 177 Michell, J., 1784, Phil. Trans. R. Soc. (London) 74, 35 [transl. from the Russian]. [reprinted in Detweiler (1982)]. Holevo, A. S., 1982, Probabilistic and Statistical Aspects of Mott, N. F., 1929, Proc. Roy. Soc. London A 126, 79. Quantum Theory (North-Holland, Amsterdam) Neumann, H., and R. Werner, 1983, Int. J. Theor. Phys. 22, Holevo, A. S., 1999, Russ. Math. Surveys 53, 1295. 781. Hosoya A., A. Carlini, and T. Shimomura, 2001, Phys. Rev. Newton, T. D., and E. P. Wigner, 1949, Rev. Mod. Phys 21, D 63, 104008. 400. Ingarden, R. S., 1976, Rep. Math. Phys. 10, 43. Pachos J., and E. Solano, 2003, Quant. Inf. Comp. 3, 115. Ingarden, R. S., A. Kossakowski, and M. Ohaya, 1997, In- Parker, S., S. Bose, and M. B. Plenio, 2000, Phys. Rev. A 61 formation Dynamics and Open Systems (Kluwer, Dor- , 032305. drecht, The Netherlands). Pechukas, P., 1994, Phys. Rev. Lett. 73, 1060. Iorio, A., G. Lambiase, and G. Vitiello, 2001, Ann. Phys. Percival, I. C., 1998, Phys. Lett. A 244, 495. (NY) 294. Peres, A., 1980, Phys. Rev. D 22, 879. Jarett, K., and T. Cover, 1981, IEEE Trans. Info. Theory, Peres A., 1993 Quantum Theory: Concepts and Methods IT-27, 152. (Kluwer, Dordrecht, The Netherlands). Kemble, E. C., 1937, The Fundamental Principles of Peres, A., 1995, in Fundamental Problems in Quantum The- Quantum Mechanics (McGraw-Hill, New York, NY; ory, edited by D. M. Greenberger and A. Zeilinger, reprinted by Dover) p. 244. Ann. New York Acad. Sci. 755, 445. Kent, A., 1999, Phys. Rev. Lett. 83, 1447. Peres A., 1996, Phys. Rev. Lett. 77, 1413. Kent, A., 2003, Phys. Rev. Lett. 90, 237901. Peres A., 2000a, Phys. Rev. A 61, 022116. Keyl, M., 2002, Phys. Rep. 369, 431. Peres A., 2000b, Phys. Rev. A 61, 022117. Keyl, M., D. Schlingemann, and R. F. Werner, 2003, Quant. Peres A., 2001, Phys. Rev. A 64, 066102. 3 Info. Comp. , in press [e-print quant-ph/0212014]. Peres, A., and P. F. Scudo, 2001, Phys. Rev. Lett. 87, King, C., and M. B. Ruskai, 2001, IEEE Trans. Info. Theory 167901. IT-47 , 192. Peres, A., P. F. Scudo, and D. R. Terno, 2002, Phys. Rev. Killing, W. K. J., 1892, J. Reine Angew. Math. 109, 121. Lett. 88, 230402. Koopman, B. O., 1931, Proc. Natl. Acad. Sci. U.S.A. 17, Peres, A., and D. R. Terno, 1998, J. Phys. A: Math. Gen. 315. 31, L671. Kraus, K., 1971, Ann. Phys. 64, 311. Peres, A., and D. R. Terno, 2001, Phys. Rev. A 63, 022101. Kraus, K., 1983, States, Effects, and Operations: Fundamen- Peres, A., and D. R. Terno, 2002, J. Mod. Optics 49, 1255. tal Notions of Quantum Theory (Springer, Berlin). Peres, A., and D. R. Terno, 2003, J. Mod. Optics 50, 1165. 69 Landau, L., and R. Peierls, 1931, Z. Phys. , 56. Peres, A., and W. K. Wootters, 1985, Phys. Rev. D 32, 1968. Landau, L. J., 1987, Phys. Lett. A, 120, 54. Peskin, M. E., and D. V. Schroeder, 1995, An Introduction Landauer, R., 1991, Physics Today 44 (5), 23. to Quantum Field Theory (Addison-Wesley, Reading, Laplace, P.-S., 1795, Exposition du Syst`eme du Monde, (Im- MA). primerie du Cercle-Social, Paris), Vol. 2, p. 305 [English Preskill, J., 1993, in Black Holes, Membranes, Wormholes translation in hawking and Ellis (1973)]. and Superstrings, Superstrings, edited by S. Kalara and Leutwyler, H., 1965, Nuovo Cim. 37, 556. D. V. Nanopoulos, World Scientific, Singapore) p. 22 [e- Levin, O., Y. Peleg, and A. Peres, 1992, J. Phys. A: Math. print gr-qc/9209058]. Gen. 25, 6471. Reed, M., and B. Simon, 1980, Functional Analyis, vol. 1 of Levin, O., Y. Peleg, and A. Peres, 1993, J. Phys. A: Math. Methods of Modern Mathematical Physics (Academic, Gen. 26, 3001. New York, NY). 22 Levitin, L. B., 1969, in Proc. Fourth All-Union Conf. on In- Reeh, H., and S. Schlieder, 1961, Nuovo Cim. , 1051. formation and Coding Theory, (Tashkent, 1969) p. 111 Reznik, B., 2000, e-print quant-ph/0008006. [in Russian]. Rivest, R., A. Shamir, and L. Adleman, 1978, Commun. Levitin, L. B., 1987, in Information Complexity and Control ACM 21 (2), 120. in Quantum Physics, ed. by A. Blaqui`ere, S. Diner, and Roberts, J. E., 1982, Commun. Math. Phys. 85, 87. G. Lochak (Springer, Vienna, 1987) p. 15. Rosenstein, B., and M. Usher, 1987, Phys. Rev. D 36, 2381. Lindblad, G., 1976, Commun. Math. Phys. 48, 119. Salgado D., and J. L. S´anchez-G´omez, 2002, e-print Lindner, N. H., A. Peres, and D.R. Terno, 2003, J. Phys. A quant-ph/0211164. 36, Lxxx [e-print hep-th/0304017]. Scarani, V., W. Tittel, H. Zbinden, and N. Gisin, 2000, Phys. Lo, H.-K., and H. Chau, 1997, Phys. Rev. Lett. 78, 3410. Lett. A 276, 1. 32

Schumacher, B., 1995, Phys. Rev. A 51, 2738. von Borzeszkowski, H., and M. B. Mensky, 2000, Phys. Lett. Shannon, C. E., 1948, Bell Syst. Tech. J. 27, 379, 623. A 269, 197. Shor, P., 1994, in Proceedings, 35th Annual Symposium on von Neumann, J., 1932, Mathematische Grundlagen der Foundations of Computer Science (IEEE, Los Alami- Quantenmechanik (Springer, Berlin), p. 236. tos, CA). von Neumann, J., 1955, Mathematical Foundations of Quan- Sonego, S., A. Almergren, M. A. Abramowicz, 2000, Phys. tum Mechanics, translated by R. T. Beyer (Princeton Rev. D 62, 064010. University, Princeton, NJ), p. 418. Sorkin, R. D., 1993, in Directions in General Relativity, Wald, R. M., 1975, Comm. Math. Phys. 45, 9. edited by L. Hu and T. A. Jacobson (Cambridge Uni- Wald, R. M., 1984, General Relativity (University of versity, Cambridge, UK). Chicago, Chicago, IL). Spaarnay, M. J., 1958, Physica 24, 751. Wald, R. M., 1994, Quantum Field Theory in Curved Space- Stelmachoviˇc,ˇ P., and V. Buˇzek, 2001, Phys. Rev. A 64, time and Black Hole Thermodynamics (University of 062106. Chicago, Chicago, IL). Stinespring, W. F., 1955, Proc. Amer. Math. Soc. 6, 211. Wald, R. M., 1999, Class. Quant. Grav. 16, A177. Streater, R. F, and A. S. Wightman, 1964, PCT, Spin and Wald, R. M., 2001, The Thermodynamics of Black Statistics, and all that (Benjamin, New York). Holes, in Living Rev. Relativity 4, 6. [Online article: Strominger, A., 1996, in Fluctuating Geometries in Statis- http://www.livingreviews.org/Articles/Volume4/2001-6wald]. tical Mechanics and Field Theory, 1994 Les Houches Walgate, J., and L. Hardy, 2002, Phys. Rev. Lett. 89, 147901. Summer School, edited by F. David, P. Ginsparg, and Wehrl, A., 1978, Rev. Mod. Phys. 50, 221. J. Zinn-Justin (North-Holland, Amsterdam) [e-print Weinberg, S., 1992, Dreams of a Final Theory (Pantheon, hep-th/9501071]. New York, NY). Sudarshan, E. C. G., P. M. Mathews and J. Rau, 1961, Phys. Weinberg, S., 1995, The Quantum Theory of Fields (Cam- Rev. 121, 920. bridge University, Cambridge, UK) Vol. I. Summers, S. J., 1990, in Quantum Probability and Applica- Werner, R., 1986, J. Math. Phys. 27, 793. tions V, Lecture Notes in Mathematics 1442 , edited by 34 L. Accardi and W. von Waldenfels (Springer, Berlin), Wightman, A. S., 1962, Rev. Mod. Phys. , 845. p. 393. Wiesner, S., 1983, SIGACT News, 15, 78. Summers, S. J., and R. Werner, 1985, Phys. Lett. A 110, Wigner, E., 1932, Phys. Rev. 40, 749. 257. Wigner, E., 1939, Ann. Math. 40, 149. Summers, S. J., and R. Werner, 1987a, J. Math. Phys. 28, Wootters, W. K., 1998, Phys. Rev. Lett. 80, 2245. 2440. Wootters, W. K., and W. H. Zurek, 1982, Nature 299, 802. Summers, S. J., and R. Werner, 1987b, J. Math. Phys. 28, Zbinden, H., J. Brendel, N. Gisin, and W. Tittel, 2000, Phys. 2447. Rev. A 63, 022111. Terashima, H., and M. Ueda, 2003, Int. J. Quant. Info. 1,93. Zeilinger, A., 1999, Rev. Mod. Phys. 71, S288. Terno, D. R, 2002, in Quantum Theory: Reconsideration of Zel’dovich, Ya. B., 1977, Sov. Phys. JETP 45, 9. Foundations, edited by A. Khrennikov (V¨axj¨oUniver- sity, V¨axj¨o, Sweden) p. 397. [e-print quant-ph/011144]. Zurek, W. H., 1990, editor, Complexity, Entropy, and the Physics of Information Terno, D. R., 2003, Phys. Rev. A 67, 014102. (Santa Fe Institute Studies in 11 the Sciences of Complexity, vol. VIII, Addison-Wesley, ’t Hooft, G., 1996, Int. J. Mod. Phys. A , 4623. Reading, MA). ’t Hooft, G., 1999, Class. Quant. Grav. 16, 3263. Zurek, W. H., 1991, Physics Today 44 (10), 36. Tippler, F. G, 1978, Phys. Rev. D 17, 2521. Zurek, W. H., 1993, Physics Today 46 (4), 13. Toller, M., 1999, Phys. Rev. A 59, 960. Zurek, W. H., 2002, Los Alamos Science 27, 2 [e-print 14 Unruh, W. G., 1976, Phys. Rev. D , 870. quant-ph/0306072]. 29 Unruh, W. G., and R. M. Wald, 1984, Phys. Rev. D , 1047. Zurek, W. H., 2003, Rev. Mod. Phys. 75, 715. Vaidman, L., 2003, Phys. Rev. Lett. 90, 010402.