Download as pdf or txt
Download as pdf or txt
You are on page 1of 6

IOP PUBLISHING REPORTS ON PROGRESS IN PHYSICS

Rep. Prog. Phys. 71 (2008) 022001 (6pp) doi:10.1088/0034-4885/71/2/022001

Realism and the physical world


A J Leggett
Department of Physics, National University of Singapore, Block S 12, 2 Science Drive 3,
Singapore 117542
and
Department of Physics, University of Illinois, 1110 W Green Street, Urbana, IL 61801-3080, USA

Received 19 December 2007


Published 29 January 2008
Online at stacks.iop.org/RoPP/71/022001

Abstract
I consider the extent to which the applicability of the concept of classical realism is
constrained, irrespective of the validity or not of the quantum formalism, by existing
experiments both in the EPR–Bell setup, including recent experiments testing ‘nonlocal
realistic’ theories, and in the area of ‘macroscopic quantum coherence’. Unless we are willing
to sacrifice one or more other intuitively plausible notions such as that of the conventional
‘arrow of time’, it appears impossible, in either context, to maintain the classical notion of
realism.
This article was invited by Professor L H Greene

Ever since the earliest days of quantum mechanics (hereafter to the macroscopic objects made up of them, including any
QM) it has been appreciated that there are major difficulties devices which we have set up as measuring apparatus. Thus,
in reconciling the account it gives of the behaviour of the we are forced to conclude, with Schrödinger, that under
microscopic world of atoms and electrons with the classic appropriate circumstances the description of a macroscopic
notion of ‘realism’, which crudely speaking is the postulate that object (measuring apparatus, cat . . .) is inconsistent with the
physical objects have definite properties, and occupy definite hypothesis that it is in a definite macroscopic state (in the case
states, independently of whether or not they are observed. of the cat, is either ‘dead’ or ‘alive’).
This was, of course, explicitly realized by Nils Bohr, whose It is worth stating the ‘cat’ (‘measurement,’ or better
solution was to postulate that microscopic objects (such as ‘realization’) paradox a little more formally. At the
electrons, photons and atoms) are simply not the kind of thing microscopic level, the formalism of QM assigns probability
that can possess properties independently of the macroscopic amplitudes to various possible behaviors of the system (e.g. in
apparatus used to observe them; thus, for example, in the a Young’s slits experiment, to passage through each of the two
standard Young’s slits experimental setup, an electron which slits in the intermediate screen). In view of the experimentally
is inspected to see which of the two slits in the intermediate observed phenomenon of interference at the final screen (the
screen it passed through is simply not the same object as total diffraction pattern is not the sum of the two patterns which
an electron which passed through uninspected, and thus it would result from passage through one slit, the other being
is not at all surprising that the diffraction pattern (or lack of blocked), we can prima facie draw, in a situation in which
it) which is produced on the final screen differs in the two the assigned amplitude at each slit is nonzero, the (negative)
cases. If one accepts, as Bohr seems to, a division of the conclusion that it is incorrect to say that each electron of
world into the ‘microscopic’ regime of atoms and electrons the ensemble in question passed either through the upper or
and the ‘macroscopic’ level of measuring apparatus, which through the lower slit. The same conclusion can be drawn for
is postulated to behave classically, then this picture seems other microscopic-level examples when we have simultaneous
internally consistent, and it certainly seems to have satisfied nonzero amplitudes for different behaviors and see the effects
a generation of physicists and even of philosophers. of interference between them, e.g. oscillations in the Ko –K̄o
However, as pointed out by Schrödinger in 1935 in his system. Now, when we extrapolate the QM formalism to
famous ‘cat’ paper [1], there is no good reason to accept this the macroscopic level of cats and detectors, it is unarguable
division of the world into a microscopic regime where QM that in suitable circumstances (such as those postulated by
reigns and a macroscopic one governed by classical physics; Schrödinger in his original paper) it assigns simultaneous
QM is a very ‘totalitarian’ theory, and if it applies to individual nonzero amplitudes to two or more macroscopically distinct
atoms and electrons, then it should prima facie equally apply states. Of course, it is generally agreed that, as a result

0034-4885/08/022001+06$90.00 1 © 2008 IOP Publishing Ltd Printed in the UK


Rep. Prog. Phys. 71 (2008) 022001 A J Leggett

of the decoherence predicted, under any normally realistic Ma Mb


conditions, by the QM formalism itself, it is no longer possible
to see the effects of interference between the two or more
~ ~
‘branches’ of the superposition. However, the quantum
formalism is exactly the same at the microscopic and the switch switch
(a)
macroscopic level; if, therefore, a given interpretation is Ma′ M b′
excluded in the former case, it cannot become permitted in
the latter! The above formulation shows clearly that the
(+)
phenomenon of decoherence, while it may no doubt be an = Da
essential ingredient in any future resolution of the realization Ma
paradox, cannot by itself constitute such a resolution. (-)
Da Pa
In the rest of this paper, I shall not be concerned with
the numerous attempts to resolve the cat paradox within (b)
the assumption of the universal validity of QM which have Figure 1. (a) Schematic diagram of an EPR–Bell experiment.
appeared in the literature over the last 70 years. Rather, I shall (b) Details of the ‘measurement device’ Ma . The devices Ma , Mb
use the paradox as a motivation for the question: irrespective and Mb are similar mutatis mutandis.
of the validity or not of QM, what can we say from experiment
about the validity, or not, of the concept of realism as applied made on a pair of systems which have interacted in the past
to the physical world? And what can we expect to be able to but are now separated to a distance such that, according
say in the foreseeable future? to the postulates of special relativity, the outcome of a
Two cautions before we start: first, we will never, by any measurement as one system cannot be causally influenced by
finite sense of experiments, be able to establish unequivocally the choice of what to measure on the other. A schematic
the truth either of realism or of any specific alternative to it diagram of an idealized EPR–Bell experiment is shown in
(e.g. QM). This is a matter of simple logic: the fact that theory figure 1. The essential points are (1) the atomic source emits
T predicts experimental consequence E, and that experiment photons in pairs: while in real life most of the pairs are not
finds E, does not establish that T is correct. (The contrary emitted back-to-back, we concentrate on the subensemble
assertion is known to logicians as the fallacy of ‘affirmation of which is. (2) At each ‘station’ (S1 or S2) a randomly
the consequent’: that so many papers in experimental physics activated device makes a choice to switch the corresponding
appear to commit this fallacy probably signals that there is photon into one of two different measuring devices, which
an extra unspoken assumption in the authors’ minds, namely will measure different properties (e.g. the device Ma consists
something like ‘E is so surprising that it is highly unlikely that it of a polarizer Pa with transmission axis set in direction a
would be predicted by any theory other than T.’ Depending on (and the nontransmitted photons reflected), plus detectors
context, such an implicit argument may no doubt be somewhat Da(+) , Da(−) to register the transmitted and the reflected photons,
persuasive, but it is hardly logically conclusive.) On the respectively, (see figure 1); the device Ma is similar except
other hand, it may be possible to establish unequivocally, by that the transmission axis of the polarizer is set in direction
experiment, the falsity of a set of propositions (the fact that a ). (3) The spatial dimensions are such that the ‘event’ of
theory T predicts experimental consequence E, and that ‘not- detection at station S2 is spacelike separated from the ‘event’
E’ is found, does logically establish the falsity of T). of switching (choice of measurement) at S1 (and vice versa).
Secondly, I deliberately said in the last sentence ‘a set of (4) The detectors are 100% efficient. No existing experiment
propositions.’ I know of no set of experiments, and can imagine satisfies (at least to my knowledge) all of the above conditions
none, which could establish that the hypothesis of realism taken simultaneously, but it is useful to consider this idealized version
in isolation is false. Indeed, the Bohm–de Broglie ‘pilot wave’ as a basis for discussion.
(or ‘hidden variable’) interpretation of the QM formalism Let us first set up a notation to describe the data obtained in
reproduces all the standard experimental predictions of that this experiment. Suppose that a given photon 1 is switched into
formalism, yet claims to maintain the concept of realism; thus, the measuring device Ma , and the ‘transmission’ counter Da(+)
even if we should find that the experiments continue forever clicks: then we define the variable A to take the value +1 for
to verify the predictions of QM, realism per se would not be that photon. If the ‘reflection’ counter Da(−) clicks, we define
refuted. (Whether the additional assumptions which have to A to take the value −1. We take it as an experimental fact (of
be made in the Bohm–de Broglie interpretation effectively course easily verified) that for each photon switched into Ma
devalue the concept of ‘realism’ to the point where it is as either Da(+) or Da(−) clicks (but not both); hence for each photon
it were no longer recognizable is a matter of opinion.) Thus, switched into Ma , either A = +1 or A = −1. Similarly, for
we shall never attempt to test realism alone, but always in photon 1 switched into device Ma , we define the variable A to
conjunction with one or more other prima facie plausible take the value +1(−1) accordingly as the counter Da(+) (−)
 (Da )

assumptions about the world. clicks. A similar definition is made for the variables B and
The best known set of experiments which examine the B  measured on photon 2 at station S2 . Note carefully that
question of realism are those which stem from the theoretical if, for example, photon 1 is switched into device Ma , then
work of Einstein et al [2] and of Bell [3], and are usually known while the variable A is defined for this photon the variable A
as ‘EPR–Bell’ experiments. They refer to measurements is not (and vice versa). In this way, one can of course obtain

2
Rep. Prog. Phys. 71 (2008) 022001 A J Leggett

the mean value A of A, for the subensemble of photon 1 (or ‘objectivity’). A crude definition of this concept would
which is switched into Ma (let us denote averages taken on this be that ‘each photon of a given pair possesses (a complete
subensemble by Aa ); similarly we can measure A a , Bb , set of) properties in its own right.’ In other words, each
etc. More interestingly, we can measure the correlation photon 1 (say) carries with it information which is sufficient
to decide, either deterministically or statistically, how it will
N++ + N−− − N+− − N−+ behave both if switched into Ma and if switched into Ma . It is
ABab ≡ , (1)
N++ + N−− + N+− + N−+ natural to think of the information in question as embodied in
a set of ‘hidden’ variables, thus OLTs are often loosely known
where N++ is the number of pairs for which both A and B were as ‘hidden-variable’ theories, but this is not essential to the
measured to be +1, etc. Note that this correlation (average) is argument to be developed below. If one phrases the concept
defined on a particular subensemble, namely that set of pairs of realism in this way, so that the properties of interest are
for which photon 1 was switched into Ma and photon 2 into possessed by microscopic objects (photons, as in other variants
Mb . Although this paper is not concerned with QM as such, we of the experiment atoms, etc), it is natural to add to the word
note as an aside that that theory gives unambiguous predictions ‘realism’ the adjective ‘microscopic.’ Thus, it is microscopic
for the quantity ABab for any given type of atomic source: realism, in conjunction with locality and induction, which is
for instance, if the pair of atomic transitions involved in the tested by the Bell–EPR experiments.
emission of the photon pair is of the ‘0+ ’ type (no change of There is, however, an interesting alternative formulation
atomic angular momentum or parity) then a simple argument of postulates (3), which is usually called the hypothesis of
shows that the prediction is ‘macroscopic counterfactual definiteness’ (MCFD). To explain
it, let us imagine that a given photon 1 was actually switched
ABab = cos(2θab ), (2)
into Ma , thereby realizing a value of A . Given the ‘random’
where θab is the angle between the polarizer settings a and b at nature of the switching process, it could equally well have
S1 and S2 , respectively. been switched into Ma , and would then have realized a value
of A. The postulate of MCFD then states that the value of A
The class of theories about the physical world with
which ‘would have’ been realized is definite. That is, it can be
whose predictions the experimental data are most commonly
treated as a ‘fact’ about the physical world. It is fairly clear
compared is that of ‘objective local theories’ (OLTs) [4]. This
that microscopic realism implies MCFD (if the photon carried
class is defined by the conjunction of three postulates, which
with it sufficient ‘instructions’ to determine the value of A,
may be labeled crudely as
then we can say that this value ‘would have’ occurred), but the
(1) locality converse is not true: it is entirely conceivable that properties are
(2) induction indeterminate at the microscopic level but become determinate
(3) realism. when as it were amplified, as in a photodetector, to some
level of ‘macroscopicness’; indeed, many people hold (in the
Locality (often denoted ‘Einstein locality’ in the literature)
opinion of the present author incorrectly) that the formalism of
is the postulate that an event (in the sense of special relativity)
QM itself somehow achieves this result. In any case, it is clear
cannot be causally affected by any past events which lie outside
that the hypothesis of MCFD is somewhat weaker than that of
its past light cone (i.e. which cannot transmit information to the
microscopic realism, so that any experiment which excludes
spacetime point of the event in question by any signal whose
the former must automatically exclude the latter. It is also
velocity is less than or equal to that of light). This is, of course,
worth noticing that the postulate of MCFD is insensitive to
a fundamental postulate of the theory of special relativity as
whether the way in which the instructions carried by the photon
the latter is usually formulated.
decide the value of A is deterministic or only statistical.
Induction is, crudely speaking, the postulate that ‘causality
It is well known that any theory satisfying the conjunction
propagates only forward in time,’ that is, that an event cannot be
of postulates (1)–(3) (with either form of (3)) must predict
offered by any future events, whether or not the latter lie inside
values of the experimentally measured correlations ABab ,
its light cone. Since within the framework of special relativity
which satisfy the celebrated ‘CHSH’ inequality, derived by
this postulate can be derived as a consequence of locality
Clauser et al [5] by extending the result of Bell:
and ‘transitivity of causality,’ it is often not listed separately;
however, since we wish to examine the implications of the ABab + AB  ab + A Ba b − A B  a b  2. (3)
experimental data with a minimum of a priori assumptions,
it is convenient to list it explicitly. In the present context, A simple proof (one of many) goes as follows:
its significance is that the statistical properties of the complete 1. By postulate (3) (in either form) the values of A and
ensemble of emitted pairs are determined only by conditions at A simultaneously exist for each photon 1, and similarly
the source. In particular, the photons of the ‘ab’ subensemble values of B and B  exist for each photon 2.
do not know in advance that they will be switched into 2. By postulate (1), the value of A cannot be affected
Ma and Mb , respectively, so the statistical properties of this by whether it was B or B  which was chosen for the
subensemble cannot be affected by this fact and so must be measurement in S2 and vice versa.
identical to those of the ensemble of all pairs. 3. Therefore, the quantities AB, AB  , A B, A B  exist for
Let us move then to the trickiest, and in the present context each photon pair, with A, A , B, B  each taking the same
most interesting, ingredient in an OLT, namely ‘realism’ value ±1 in each of the combinations in which it occurs.

3
Rep. Prog. Phys. 71 (2008) 022001 A J Leggett

4. By trivial algebra it then follows that for each pair AB + figure 1, and relax postulate (1) by allowing arbitrary nonlocal
AB  + A B − A B   2. effects in the detection process. On the other hand, we will in
5. If ABall indicates the average over the complete some sense strengthen postulate (3), in the following sense: we
ensemble of pairs emitted by the photon source, then (4) require that the ensemble of pairs of photons emitted by the
leads directly to the result source is a disjoint union of subensembles in each of which
each photon of the pair has a definite polarization, a statement
ABall + AB  all + A Ball − A B  all  2. which is operationally defined by the requirement that the
photon obeys Malus’s law, i.e. for any given photon there
6. Finally, by postulate (2), ABab is identical to ABall , exists a (complex) unit polarization vector e such that when it
etc, and thus we obtain the CHSH inequality, equation (3). is presented with a polarizer set with (complex) transmission
Thus, if an ideal EPR–Bell experiment were to produce ‘axis’ c, the probability of transmission is |e ∗ ·c|2 . (In such
results which violate the inequality (3) (in particular, were it a theory the transmission/rejection of any individual photon
to reproduce the QM prediction (2), which violates (3) for is determined inter alia by nonlocal effects, but the statistics
(example) settings such that a · b = a · b = a · b = π/8, a · for the ensemble of photons, obtained by averaging over these
b = 3π/8), then we should know for sure that one of the effects and others, is required to satisfy Malus’s law.) Using the
postulates (1)–(3) cannot hold in the real world. Unfortunately, elementary inequalities, applicable for variables A, B taking
there is at present no single experiment which satisfies all values ±1
of the defining criteria for an ‘ideal’ experiment; there are
always various ‘loopholes’ (imperfect detector efficiencies, − 1 + |A + B|  AB  1 − |A − B|,
lack of adequate spatial separation, questionable ‘randomness’
in the switching process, etc). However, while no existing it is straightforward to show [7] that such a theory, called
experiment has blocked all existing loopholes simultaneously, a ‘crypto-nonlocal hidden-variable (CNLHV) theory’ must
with one exception each of them individually has been blocked predict inequalities for the experimental correlations which
in at least one experiment, so that it would seem to require are violated by the QM predictions. These inequalities are
a very peculiar conspiracy of nature to allow an OLT to be however of a different nature from the Bell–CHSH inequalities,
maintained. The exception is what is sometimes called the and a test of them requires a set of measurements with
‘collapse locality’ loophole: if one postpones the ‘event’ of elliptically polarized analyzers which had not been previously
realization of a particular outcome at each of the stations to done. Recently, the experimental group in Vienna carried
a sufficiently late stage (perhaps long after the dectector has, out just such an experiment [8], finding agreement with the
in our usual way of thinking, produced a macroscopic output) predictions of QM and violation of the CNLHV predictions by
then the ‘events’ at the two stations would not have fulfilled the several standard deviations. Actually, to obtain this violation
condition of spacelike separation in any existing experiment, it was necessary for the authors of [8] to assume that the
and indeed to fulfill it would require conditions which border experimental correlations, while a function of the relative angle
on science fiction: cf for example [6]. For the purposes of this a · b of the polarizer settings, are independent of the ‘center-
discussion, let me from now on assume that such an ‘ideal’ of-mass’ variable, i.e. invariant under simultaneous rotation of
(loophole-free) Bell–EPR experiment will someday be done, a and b through the same angle, something which is certainly
will turn out consistently with the QM prediction (2) and will plausible but is not directly tested. However, in very recent
thereby definitively invalidate the whole class of OLTs. What work both these authors [9] and the Singapore group [10] have
are the implications? generalized the CNLHV inequalities and conducted further
Of the three defining postulates (1)–(3) of the class of experiments so as to eliminate this loophole; once again the
OLTs, the most impervious to challenge would seem at first data are consistent with the QM predictions and violate those of
sight to be (2), in the sense that once we give up our ‘common- the CNLHV class of theories, in the Vienna case by more than
sense’ notions concerning the ‘arrow of time’ it seems very 80 standard deviations. While those experiments are subject to
difficult to continue to do physics at all in the mold to some of the same ‘loopholes’ as the standard EPR–Bell ones,
which we have been accustomed. Actually, in this writer’s the outcome makes it extremely plausible that CNLHV theories
opinion it is entirely possible, indeed probable, that the next cannot describe the physical world. What is the significance
major revolution in physics will force us to do just that, of this result? Can we regard it as establishing (subject, of
but unfortunately it is in the nature of scientific revolutions course, to the above reservations concerning induction) that it
that their content is difficult or impossible to forecast, so is indeed realism rather than locality which has to be sacrificed?
speculation along these lines seems pointless at this time. We It is certainly suggestive in this respect; on the other hand, a
are then left with postulates (1) and (3), and it is an amusing critic might argue that by formulating our ‘realistic’ postulate
sociological observation that while popular writers on the in a way which requires the realism to refer to the properties
subject of the EPR–Bell experiments almost without exception of individual photons of the pair rather than to the pair as a
choose to give up (1) (locality), professional physicists usually whole, we have in effect smuggled the concept of locality
opt to sacrifice realism. Is there any experimental way of back in again. Perhaps the lesson is that while the concept
deciding the question? of ‘local realism’ is clear-cut, to try to analyze it in terms
A partial answer comes from a very recent experiment and of its two prima facie components may not in the end be a
the associated theory. Consider the standard ‘ideal’ setup of particularly meaningful exercise. Certainly, in QM itself the

4
Rep. Prog. Phys. 71 (2008) 022001 A J Leggett

two concepts, or rather their absence, in some sense appear does indeed always produce the value ±1. We then note that
blended, in that once one has the (nonrealistic) concept of for this simple ‘2-state’ system, if it is ideally isolated from
a quantum superposition then the idea of applying it to the its environment (a term which in this context includes any
coupled state of two spatially remote objects is an entirely internal dissipative mechanism), QM predicts, independently
natural development. of the density matrix of the system (or more accurately of the
I now turn to a different class of experiments, which ‘time ensemble’ formed by a sequence of experiments starting
attempts to examine the question of realism at a level much from the same initial conditions) 2-time correlations of the
closer to our own everyday consciousness than that of electrons form
and atoms. Since I wrote a review [11] of this area of research Q(t)Q(t  ) = cos (t − t  ), (4)
in 2002 and not much has changed qualitatively since then, I
where  is a tunneling frequency between the two states
will be rather brief. The experiments of this type done to date
Q = ±1. The formal similarity of equation (4) to equation (2)
look, explicitly or implicitly, for evidence that the predictions
above suggests that it might be possible to devise a CHSH-
of the QM formalism continue to work when the states involved
type inequality for this system, with the different times of
are quantum superpositions of two or more states which measurement playing the role of the different polarizer settings,
are in some intuitively reasonable sense ‘macroscopically’ and we shall now see that this is so.
distinct; in most cases, the evidence for this conclusion comes The class of theories whose predictions we propose to
from the time-dependent behavior, in particular from the compare with those of QM and with experiment go under
‘Ramsey-fringe’ effects which are the characteristic signature the name of ‘macrorealism.’ Similarly to the case of OLTs,
of a quantum superposition. The systems in question range this class of theories is defined by the conjunction of three
from fullerenes and other complex molecules to magnetic postulates:
biomolecules, quantum-optical systems and superconducting
devices such as SQUIDs; for details see [11]. It should be (1) macroscopic realism per se
emphasized that from the point of view of the ‘logic’ of realism (2) noninvasive measurability
this class of experiments is less advanced than the EPR–Bell (3) induction.
ones, in the following sense: in the EPR–Bell case the existing The third postulate needs no special comment; it plays
experiments not only establish that the QM predictions are essentially the same role as in the EPR–Bell case, assuring
correct, but also (modulo the ‘loopholes’, see above) refute us that the outcome of a measurement on the system cannot
those of the class of OLTs (and, now, of CNLHV theories). In be affected by what will or will not be measured on it later.
the case of the present class of experiments (sometimes called The assumption of macroscopic realism simply says that: a
‘macroscopic quantum coherence’ (MQC) experiments), what macroscopic object which has available to it two or more
has been established so far is that if the raw data are interpreted macroscopically distinct states is at ‘almost all’ times in one
according to the QM formalism, then the states generated of these states (the ‘almost all’ is necessary because one has to
must be coherent quantum superpositions of macroscopically allow, in such a theory, for transits between the states; however,
distinct states rather than classical mixtures of such states (the this complication can be taken into account in the analysis,
latter description would be entirely consistent with a picture see [12]). In the present context, the postulate of macroscopic
in which each system of the ensemble has realized a definite realism is equivalent to the statement that Q(t) possesses either
state). One might then go on to argue as follows: ‘The notion the value +1 or the value −1 at (almost) all times t, irrespective
of QM superpositions at any given level is inconsistent with of whether or not it is actually measured (we saw above that
the assumption of realism at that level: we see evidence for QM when measured, it certainly realizes one or other of these two
superposition at the level of SQUIDs (etc): therefore, realism values).
is false at the level of SQUIDs.’ As pointed out above, this In this case, in contrast to the EPR–Bell experiments, all
argument, while perhaps plausible, is logically unsound; in measurements are made on a single system, so there is no
order to exclude realism at the level of SQUIDs, we need to question of invoking locality. Instead, one uses in a rather
do an experiment which goes beyond the existing ones (just similar role the postulate of ‘noninvasive measurability.’ This
as in the EPR–Bell case, it was necessary to do more than postulate (which is emphatically not a QM one!) states that it
simply verify that the QM predictions were correct for a few is possible, at least in principle, to perform a measurement
randomly chosen polarizer settings). Such an experiment was of Q(t) in such a way that neither the state of the system
proposed by Garg and the present author in [12], and I believe at time t nor its subsequent behavior is influenced by the
that a number of experimental groups are currently working measurement. Given the generally very invasive nature of most
towards it. measurement procedures on macroscopic solid-state systems
The general structure of the experiment is conceptually such as SQUIIDs, this postulate might at first sight seem rather
very similar to that of the EPR–Bell experiments. One deals unlikely to be fulfilled in a real-life situation. However, we can
with a single system, e.g. a SQUID ring in an appropriate make it a lot more plausible by invoking an ‘ideal negative
external magnetic flux, which possesses a dichotomic variable result’ procedure: we arrange our measuring apparatus so
Q(t) (in practice the value of the circulating current in the that if Q(t) is (say) +1 it is triggered, while if Q(t) = −1
ring), i.e. one which can take only one of two discrete values nothing happens. We then do a series of runs in which Q
which we label by convention +1 and −1, respectively. We is measured at some time to ; these runs on which Q(to ) is
verify in a preliminary experiment that measurement of Q(t) measured to be +1 we throw away, the rest we keep. We

5
Rep. Prog. Phys. 71 (2008) 022001 A J Leggett

then ‘invert’ the measuring setup, so that a value of Q(t) To answer this question, it is essential to appreciate
equal to −1 triggers it while Q(t) = +1 does not; this time that the hypothesis of MCFD, which is refuted by the
we keep the runs on which Q(to ) is measured to be +1 and EPR–Bell experiments, while it is as noted weaker than
throw the rest away. In this way we can construct the 2-time that of microscopic realism, is stronger than the postulate
correlations which we are giving to require in the argument of macroscopic realism. MCFD is a statement about a
below. (Note that it is only the first measurement of any pair macroscopic event which ‘would have’ happened (or not) in a
which needs to be noninvasive.) Conceptually, the postulate situation which did not in fact arise; by contrast, macroscopic
of noninvasive measurability seems a very natural corollary realism is a statement about what actually happens in a
to that of macroscopic realism: would there be any sense in situation which is realized. Of course, the hypothesis of
affirming that the system ‘really has’ the value Q = +1 at some macroscopic realism does presumably imply that a second
time t, if a measuring device which is activated only when macroscopic object, set up explicitly as a measuring device,
Q = −1 could nevertheless affect its behavior? Of course, ‘would have,’ had it been activated at time ti , have given the
in real life it may well be difficult to guarantee that ‘nothing output corresponding to Q(ti ), and thus, one could, if desired,
happens’ in the literal sense, and one may have to be content replace postulate (1) by an MCFD formulation similar to the
with replacing it by the statement that whatever happens when one given in the microscopic case. However, it is clear that
Q(t) has the ‘negative’ value is reliably calculable. (On MCFD in a situation where the measured object is macroscopic
a possible procedure for setting up an ideal-negative-result is a much weaker assertion than MCFD in the microscopic
measurement, and for further discussion, see [13].) (e.g. EPR–Bell) case, and thus its denial is a much stronger
Given the above three postulates, the argument for a CHSH statement. Hence the experiment proposed in [12], should it
inequality of the form be done and come out according to the predictions of QM, will
be a considerably stronger denial of realism than follows from
Q(t1 )Q(t2 )t1 t2 + Q(t2 )Q(t3 )t2 t3 + Q(t3 )Q(t4 )t3 t4
the existing EPR–Bell work.
−Q(t1 )Q(t4 )t1 t4  2, (5)
Acknowledgment
(where the subscript ti tj indicates that the averages are taken,
as they must be, on the subensemble on which measurements This work was supported in part by the National Science
are made at times ti and tj (only)) goes through analogously to Foundation through grant no. NSF-EIA-03-50842.
the EPR–Bell case: postulates (1) and (2) justify the statement
that provided the first measurement of any pair is conducted in References
an ideally noninvasive way, Q(ti ) has for each run a definite
value which is unaffected, according to postulate (3), by what [1] Schrödinger E 1935 Naturwissenschaften 23 807
will or will not be measured subsequently. Thus, for each [2] Einstein A, Podolsky B and Rosen N 1935 Phys. Rev. 47 777
run the quantities Q(ti )Q(tj ) exist, with Q(ti ) having the [3] Bell J S 1965 Physics 1 195
[4] Clauser J F and Shimony A 1978 Rep. Prog. Phys. 41 1881
same value for any tj . The trivial algebra of step (4) of [5] Clauser J F, Horne M A, Shimony A and Holt R A 1969 Phys.
the proof of the original CHSH theorem goes through, so Rev. Lett. 23 880
equation (5) is satisfied provided the averages are replaced by [6] Leggett A J 2007 J. Phys. A: Math. Theor. 40 3141
Q(ti )Q(tj )all , where ‘all’ indicates that they are taken over [7] Leggett A J 2003 Found. Phys. 33 1469
all runs. Finally, postulates (2) and (3) together justify the Leggett A J 2008 Found. Phys. (erratum) in press (The erratum
inter alia improves the RHS of the final inequality (4.9) to
replacement of Q(ti )Q(tj )all by the experimentally measured 2 − π4 | sin χ|).
correlation Q(ti )Q(tj )ti tj , giving the inequality (5). [8] Gröblacher S, Paterek T, Kaltenbaek R, Brukner C,
I will not discuss here the rather delicate question of Zukowski M, Aspelmeyer M and Zeilinger A 2007 Nature
exactly how ‘macroscopically distinct’ are the two circulating- 446 871
current states realized in a typical SQUID ring; this is at least Gröblacher S, Paterek T, Kaltenbaek R, Brukner C,
Zukowski M, Aspelmeyer M and Zeilinger A 2007 Nature
partly a matter of subjective definition, and in this context it is 449 252 (corrigendum)
sufficient to note that by any reasonable definition these two [9] Paterek T, Fedrizzi A, Gröblacher S, Jennewein T,
states differ in the behavior of, at a minimum, several million Zukowski M, Aspelmeyer M and Zeilinger A 2007 Phys.
electrons. Rather, I would like to close by asking the question: Rev. Lett. 99 210406
if the proposed experiment is done, and comes out according [10] Branciard C, Ling A, Gisin N, Kurtsiefer C, Lamas-Linares A
and Scarani V 2007 Phys. Rev. Lett. 99 210407
to QM (hence in violation of the inequality (5)), what will we [11] Leggett A J 2002 J. Phys.: Condens. Matter 14 R415
have learned over and above what is already known from the [12] Leggett A J and Garg A 1985 Phys. Rev. Lett. 54 857
EPR–Bell experiments? [13] Leggett A J 1988 Found. Phys. 18 939

You might also like