a.__ 6 February l!!M 1995 -33 ELSEVIER PHYSICS LETTERS A PhysicsLettersA 197 (1995) 367-371 Physical motivation of the modal interpretation of quantum mechanics Dennis Dieks zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJ FounaWions of Science Unit, Vniversiteit Received 18 July 1994; revised manuscript Vtrecht, PO. Box 80.000, 3.508 TA Vtrecht, The Netherlands received 25 November 1994; accepted for publication 12 December 1994 Communicated by P.R. Holland zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPON Abstract We show that the modal interpretation of quantum mechanics follows from essentially two demands: ( 1) Definite properties are to be ascribed to physical systems in such a way that the magnitudes which are definite are definable solely from the quantum state and the structure of Hilbert space; (2) There is a one-to-one relation between properties possessed by a system and properties of its environment. 1. Introduction vector of a composite system. Consider such a system &p represented by It,@) E W @ Yip; I-P and 7@ are the Hilbert spaces associated with o and p, respectively. Consider the bi-orthonormal decomposition of I@@), The modal interpretation of quantum mechanics [ 11 is a realist interpretation of the quantum mechanical formalism. It is realist in the sense that it assigns definite values to a set of magnitudes pertaining to a physical system. In other words, the interpretation is (1.1) about zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA properties of physical systems, not about meaj surement results. Measurement outcomes are just specific examples of properties; as in the case of pointer where{h>Ij and{IPj>Ij are subsets of orthonormal positions, exhibited by a measuring device. A second bases of 7-P and 7@, respectively. This decomposiimportant feature of the interpretation is that it is eni tion generates two sets of projectors operating on 7-P tirely based on the usual formalism of quantum meandonWs,respectively: {~~~)(~~J}j~d{~~j)(~j~}~~ chanics - nothing is added to the mathematical strucIf there is no degeneracy among the numbers { Icj 12} ture. The new aspect of the interpretation is exactly these sets of one-dimensional projectors are uniquely that this well-known structure is not regarded merely determined by the decomposition. If degeneracy ocas a codification of possible measurement results and curs this is no longer so; but the projectors belongtheir probabilities, but is also considered to contain ing to one value of { ]cj12} can be added to form a information about physical properties and the probamulti-dimensional projector. In this way a new set of bilities of their presence. projectors, including multi-dimensional ones, can be The property ascription is based on the bidefined. This set is again uniquely determined by the orthonormal (Schmidt) decomposition of the state bi-orthonormal decomposition. IS”‘>=~cjlaj)lP.i)~ 0375-9601/95/$09.50 @ 1995 Elsevier Science B.V. All rights reserved SSDIO375-9601(94)01002-l 368 D. Dieks /Physics Letters A 197 (I 995) 367-371 The modal interpretation assigns definite values to the same set of properties cannot be definite in all cirthe physical magnitudes represented by the projectors cumstances (as will turn out to be the case): it will in the just-described sets. Also all functions of them depend on the entangled state of system plus environare taken to represent magnitudes with definite valment which magnitudes become definite. ues (“well-defined” or “applicable” physical magniThe charge has been leveled that the modal intertudes). zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA Which value of all possible values of a given pretation’s way of determining the set of definite magapplicable magnitude is actually realized is not fixed nitudes is ad hoc. Why for example, so the objection by the interpretation: the interpretation is probabilistic. goes, concentrate on the bi-orthonormal decomposiFor each possible value of an applicable magnitude a tion and not on one of the countless other possible probability is specified. The probability that the magrepresentations of the total state vector? Our result can nitude represented by 1cq) (q 1 (and also Ipi) (pi I) has be seen as a justification of the modal prescription. Inthe value 1 is given by Ici12. In the case of degenerdeed, the first of the above requirements leading to the acy it is stipulated that the magnitude represented by modal imerpretation seems natural in giving any interCiEh IczJ(czil has value 1 with probability CiEll Ici12 pretation to the mathematical formalism of quantum (Zl is an index-set containing indices j, k such that zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHG mechanics without adding anything to the mathematical structure. The second requirement can be seen as lC j12 = IC k12). The above prescription for finding the definite quana way to generalize (and make rigorous) a significant tities pertaining to a system can also be given in a form part of Bohr’s interpretation of quantum mechanics. in which the Schmidt decomposition and the state vecAccording to Bohr the applicability of concepts detor of the composite system do not occur explicitly pends on the type of macroscopic measuring device [4] : the definite properties of a system can be dethat is present; given a “phenomenon” there is a onetermined from the spectral resolution of the reduced to-one correspondence between the properties of the density operator of the system alone. measuring device and those of the object system. Our Physical magnitudes are represented in the usual second requirement implements this idea also in sitway by operators in Hilbert space. There is conseuations in which there is no macroscopic measuring quently no contextualism in the sense that the repdevice, but only a correlation with the (possibly miresentation of physical magnitudes by operators decroscopic) environment. pends on the context. However, there is contextualThe idea that there is a correspondence between ism in another sense: the set of definite properties properties of a system and those of its environment is also physically motivated by other approaches to picked out depends on how the system under considthe interpretation of quantum mechanics, especially eration is “entangled” with its surroundings. In fact, we shall demonstrate that the modal interpretation is the decoherence approach (the “monitoring” of a systhe only possible realist interpretation satisfying estem by its environment, see Ref. [5] and references therein). sentially two demands, one of which embodies this “entanglement”. However, it should be noted that the modal interThe first demand is that the Hilbert space formalpretation can also be justified in another way, indeism, with the usual representation of physical magnipendent of the second of the above requirements. As shown by Clifton [3] (see also Ref. [2] ) the modal tudes by observables, should be completely respected; no additional structure should be used in defining the interpretation is uniquely determined by a number of applicable physical magnitudes (it seems appropriate natural assumptions about the structure of the set of definite properties which is definable from the statisto call this a “no-hidden-variables” demand). We will tical operator W pertaining to a system. Our alternatake this requirement to mean that each individual deftive approach is meant to highlight the specific way inite magnitude should be definable from the quantum in which the modal interpretation is contextual, and to state and the structure of Hilbert space alone. show that it is the natural realist interpretation to be The second requirement says that there should be combined with decoherence or “Copenhagen” ideas. a one-to-one correlation between the definite properties of a system and the definite properties of its environment. This will lead to a type of contextuality if D. Dieks / Physics Letters A 197 (1995) 367-371 2. The problem 369 zyxwvut that I$) represents a composite system, we can take it that the Hilbert space has a dimension greater than 2. In that case Gleason’s theorem says that every probability measure on the closed subspaces of Hilbert space has the form p(P) = Tr WP, with P a projection operator (p(P) is the probability assigned to the subspace on which P projects) and W a semi-definitely positive Hermitian operator with trace 1. Now, the requirement of definability in terms of I+) has the consequence that all probabilities should remain the same under any unitary transformation of zyxwvutsrqponmlkjihgf W that leaves I$) invariant. That means that W itself must be invariant under any such transformation. The subspaces of the Hilbert space on which the projection operators in the spectral resolution of W project should consequently also be invariant. However, the group of unitary transformations which leave I+) invariant has only two invariant subspaces: the one-dimensional subspace ‘FI+ spanned by ]I+%)and the subspace 7-$ orthogonal to I@). It follows that W must have the following form: W = qP+ + c2Pi/rn, with P$ and Pk the projection Every realist interpretation of quantum mechanics has to specify which physical magnitudes possess definite values. Our aim is to formulate a rule for such a specification which is based on the standard formalism of quantum mechanics alone. In line with this we assume that all physical magnitudes are represented in the standard way by Hermitian operators on Hilbert space. Not all physical magnitudes thus represented can simultaneously possess definite values: that would lead to inconsistencies, for example via Gleason’s theorem. Therefore, a selection has to be made from the set of all magnitudes. Our objective is to define the definite physical magnitudes from the state vector and the structure of Hilbert space. In our argument we will make use of probabilities defined on Hilbert space. In accordance with an unproblematical and generally accepted part of the standard interpretation of quantum mechanics, we assume that a ket I$) induces the usual probability measure on the (closed) subspaces of Hilbert space. This is operators on 7-L*and 7-L;, respectively, m the number equivalent to assuming that all values of all physical of dimensions of Xi, and with ct + c:! = 1, cl, c2 > magnitudes are assigned a probability in the situation 0. This general form of W is sufficient for our further described by I$). As already stated, we do not wish argument. to be committed to the position that all physical magnitudes actually possess a value in that situation - it 3. Determination of the physical magnitudes that should be stressed that our assumption about the expossess a definite value istence of a probability measure does not imply that position. We can just follow the standard interpretaOur second requirement directs our attention to systion, namely that the probabilities pertain to the outtems plus environment. We therefore consider a quancomes of possible ideal measurements on the system tum state I$@) of a composite system, an “object” described by the ket I+). When, at a later stage, we will have determined the restricted set of magnitudes that system (Yand its environment p. (This will also cover the generalized formulation of the modal interpretado possess definite values in the. situation described tion discussed in Ref. 141, as long as a pure state is by I+), the probabilities of the values of this special assumed to apply at some level of “compositeness”.) set zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA will be interpreted as probabilities of a value actul To determine the definite properties of our system a, ally being there (for those properties that were already we look for projection operators Pa that are candidefinite before the measurement, ideal measurements dates for representing such properties. Because our will just reveal the pre-existing values). aim is to determine a description of our physical sysIn fact, a weaker assumption would suffice for our tem which is as “fine-grained” as possible, we are insubsequent determination of the set of definite magterested in projection operators with a minimum numnitudes. We could just demand that a probability disber of dimensions - preferably one-dimensional protribution on Hilbert space is induced by I$), without jectors. Because a and p occur in the formalism in specifying its form. The first of our above main requirements then says that the probabilities should be a completely symmetrical way, there should also be projection operators Pp representing properties of the definable in terms of I$) and the structure of Hilbert environment /3. space only. Because we will be considering the case 370 D. Dieks /Phy sics Letters A 197 (1995) 367- 371 Making the assumption of the previous section, we pend on whether all coefficients in the bi-orthonormal find that the probability of joint occurrence (Pa and zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA decomposition of Iqafi) are unique or whether degenPp both taking the value 1) can be calculated as eracies occur. To show this, we make use of the following. If an applicable physical magnitude can be de(V?P”P%@V ( or, with the more general form of W: Tr(WPaPp) = zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA c~(s,b” filPaP~lqV@) + cz( 1 fined solely from I@@) and the structure of the Hilbert space 3-1”@ 7-@, all transformations which leave these (V?PcuP%@))). We now make use of the idea that there is a oneelements invariant also leave the magnitude in questo-one correlation between the properties of the systion invariant. All applicable magnitudes should theretem and its environment. We express this demand as fore be invariant under all unitary transformations of follows: P(P;) = P(Pf) = P(PfPf) for two corthe form Un @ Up that leave I+@) invariant (Ua is a unitary operator in 7-i*, Up a unitary operator in related properties represented by Pf and Pf, respec7#). To determine the class of such operators, apply tively. It follows, if we use the above expression for u= @ up to IQ. (l.l), joint probabilities (either the usual one or the one in terms of W), that (+I”plPfPkpI$@) = (+I”fljPt @ lly@) = (+@I1123 P,Pl$,nP). Write P; = ~ao)(~l U” @ UpI+ap) = CCj j U” lfXj) @ U’ lpj). (3.1) and P,f = Ibo)(bol, and I@) = zijcijlai)Ibj), with {IQ)} and { lbj)} orthogonal bases in 7-f” and HP, reBy virtue of the unitarity of U” and U* the right-hand spectively. The correlation condition yields [coal2 = side of Eq. (3.1) is again a bi-orthonormal decompoCjJCjO12, SO CjfJ= COj = 0 ifj # 0. We zjlCO j12 = zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA sition. If U” @3Up]@@) = I@@) it follows from the find therefore that either P (PF) = P( Pkp) = 0, or uniqueness properties of the bi-orthonormal decom11/I@) = clac)Ibc) + c’l&@) with lc12 > 0 and with position that Vlaj) = Iaj) for each j belonging to a Ic/@) a ket whose decomposition in the product basis value of Icj12 which occurs only once in the decomonly contains components in which {l”i>l~j)} O f zap position. However, if lck12 = (cl12 for k,Z E Zm, with j a~) or Ibo) do not occur. The ket IqF@) of course has Z, an index-set, the restriction of U” to the subspace a bi-orthogonal decomposition; from the above it folspanned by IaJ, i E Zmr can be an arbitrary unitary lows that all kets from 3-1” occurring in this decomoperator. position are orthogonal to I%), and all kets from tip It follows that the one-dimensional projectors assoorthogonal to Ibo). The bi-orthogonal decomposition ciated with non-degenerate values of Icj12 are invariof I@p), completed with the term c IUO)Ibo), therefore ant under the unitary transformations leaving I@@) is a bi-orthogonal decomposiEion of I$@). In other invariant. However, in the case of degeneracy only the words, the term lua)lbu) occurs in the bi-orthogonal multi-dimensional projector on the subspace spanned decomposition of I@@). Repeating this argument for by IQJ, corresponding to one single value of IcJ2, is other possible definite projectors, we find that all oneinvariant. All smaller projectors will change under the dimensional projection operators that may represent transformations Ua . definite properties, with a non-zero probability of takSo we have now found the one-dimensional projecing the value 1, project on mutually orthogonal kets tion operators corresponding to the terms in the biIaj), and lpi), that occur in the decomposition of orthonormal decomposition with unique coefficients, I@“P) in the way shown in Eq. ( 1.1) . and the multi-dimensional projection operators proAs a result, of all one-dimensional projectors only jecting on the subspaces associated with the terms with the ones which project on kets which occur in the equal coefficients, as the “finest-grained” candidates bi-orthogonal decomposition of IJ,@), with non-zero for applicable physical magnitudes. Conversely, it is coefficients, are candidates for representing definite clear that these operators indeed satisfy our correlamagnitudes with a non-zero probability of being 1; tion requirement and can be individually defined in conversely, it is clear that all such projectors satisfy terms of I$“fi) and the structure of the Hilbert space our correlation requirement. 7-Z”@ tip alone - they are the projectors occurring in But are all these one-dimensional projection operathe spectral resolution of the reduced state W” (obtors in fact definable from \lfi@)? This turns out to detained from Ifi@) by partial tracing). These projec- D. Dieks /Phy sics Letters A 197 (1995) 367- 371 371 tom can be ranked (and thus individually identified) this property ascription the bi-orthonormal decompoby the associated values of Icj12. sition of the state vector of a composite system plays Finally, straightforward repetition of the correlation a crucial role. It follows from the above that this role argument for higher-dimensional projectors shows that of the bi-orthonormal decomposition (or, equivalently, the only projectors pertaining to a and p, respectively, the spectral resolution of the reduced density operaand possessing a l-l correlation are sums of the protor) need not be postulated in an ad hoc way, but can jectors we already have found; this set is therefore be derived from physically motivated requirements. complete. We still have to discuss the case in which the probAcknowledgement ability of a projection operator taking the value 1 is zero. Here the situation is the same as in the case of I am much indebted to Guido Bacciagaluppi, Tim degeneracy we have just discussed. The set of all uniBudden, Pieter Vermaas and especially Rob Clifton tary transformations leaving leap) invariant, and refor their suggestions and criticism of earlier versions specting the factorisation of 1-l”s into ‘FI” and ‘7-f@, of this paper. includes arbitrary transformations U” and Up in the null-spaces of zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA W” and Wp, respectively. Only the operators projecting on the whole of these null-spaces References can therefore represent definite properties ’ . [ 11 D. Dieks, Phys. Rev. A 49 (1994) 2290, and references therein, [2] J. Bub, On the structure of quanta1 proposition systems, to be published in Found. Phys. (1994). [3] R. Clifton, Independently motivating the Kochen-Dieks modal interpretation of quantum mechanics, to be published 4. zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA Conclusion in Brit. J. Philos. Sci. ( 1994). [4] P.R. Vermaas, and D. Dieks, The modal interpretation of quantum mechanics and its generalization to density We have shown that the property ascription furoperators, to be published in Found. Phys. ( 1994). nished by the modal interpretation is the only one that [5] W.H. Zumk, and J.P. Paz, Decoherence, chaos, the quantum solely relies on the Hilbert space structure and the and the classical, in: Symposium on the foundations of state vector in assigning properties that are in one-tomodem physics, eds. l? Busch, P. Lahti and I? Mittelstaedt one relation to the properties of the environment. In (World Scientific, Singapore, 1993). 1This is in agreement with Ref. [ 11, and deviates slightly from the proposal by Clifton [ 31. Clifton takes all one-dimensional projectors in the null-spaces of Wn and Wp as definite, on the grounds that their definiteness best explains the fact that the value 0 will be found with certainty in a measurement of these observables. It seems, however, that the fact that the “bigger” projector on the total null-space has the value 0 with certainty can serve the same explanatory purpose. For example, if a system with certainty has the property that its position is outside a given region, this by itself seems sufficient to explain that a measurement which is sensitive to individual positions inside that region will with certainty give a null-result.