Can Bell’s Prescription for Physical Reality Be Considered Complete?

Joy Christian∗
Department of Physics, University of Oxford, Parks Road, Oxford OX1 3PU, United Kingdom

An experiment is proposed to test Bell’s theorem in a purely macroscopic domain. If realized, it

would determine whether Bell inequalities are satisfied for a manifestly local, classical system. It is
stressed why the inequalities should not be presumed to hold for such a macroscopic system without
actual experimental evidence. In particular, by providing a purely classical, topological explanation
for the EPR-Bohm type spin correlations, it is demonstrated why Bell inequalities must be violated
in the manifestly local, macroscopic domain, just as strongly as they are in the microscopic domain.

PACS numbers: 03.65.Ud, 03.67.-a, 02.20.Sv

arXiv:0806.3078v2 [quant-ph] 29 Jun 2009

Despite the existence of an explicit counterexample [1], local realistic prescription provided by Bell, namely
Bell’s theorem is still widely believed to have proved that Z
no physical theory can be reconciled with the notion of E(a, b) = Aa (λ) Bb (λ) dρ(λ), (2)
local reality espoused by Einstein, Podolsky, and Rosen E3
(EPR) [2][3]. It therefore seems worthwhile to investigate
the expectation values of the individual variables Aa (λ)
the very foundations of Bell’s theorem experimentally, in
or Bb (λ) can be easily shown to vanish identically [5].
a purely macroscopic domain. If realized, the experiment
Their joint correlation function on the other hand would
described below would test whether or not a manifestly
not vanish in general, and is usually worked out to be [5]
local, macroscopic system can violate Bell inequalities, as
implied by the arguments of Ref. [1]. A physical scenario 2
well suited for this purpose is that of the local model first E(a, b) = −1 + cos−1 (a · b) . (3)
π
considered by Bell himself [4]. The details of this model
can be found also in some standard textbooks [5]. To our If we now substitute this linear correlation function into
knowledge, no real experiment has ever been performed the CHSH string of expectation values for four arbitrarily
to check whether Bell inequalities do indeed hold in the chosen detector directions a, a′ , b, and b′ , giving
manifestly local and realistic domain of Bell’s model.
E(a, b) + E(a, b′ ) + E(a′ , b) − E(a′ , b′ ), (4)
The central contention of Ref. [1] is that the sinusoidal
EPR-Bohm correlations observed in the laboratory have then it is easy to check that the absolute value of this
nothing to do with entanglement or nonlocality per se, string never exceeds the bound of 2, thus saturating but
but stem entirely from the topological properties of the not violating the celebrated Bell-CHSH inequalities [6].
physical space. This viewpoint can be explained clearly This often quoted result is usually considered to be well
by a closer examination of the model for spin considered established, but in fact it is simply incorrect. The trouble
by Bell. In this model the space of complete states of spin is that the local realistic prescription for spin correlations
consists of unit vectors λ in three-dimensional Euclidean provided by Bell—namely, Eq.(2) above—is incapable of
space E3 . The local beables Aa (λ) and Bb (λ), existing accounting for the elements of physical reality envisaged
at freely chosen unit directions a and b, are defined by by EPR in the topologically correct order. The situation
is analogous to having taken a photograph apart pixel
An (λ) = − Bn (λ) = sign (λ · n) , (1) by pixel, keeping count of each pixel correctly, and then
trying to put it back together. If the geometrical order of
the pixels has been neglected in the process, then there
provided λ · n 6= 0 for n = a or b, and otherwise equal to would be little chance of recovering the photograph back.
the sign of the first nonzero term from {nx , ny , nz }. This Similarly, what is missing from the prescription (2) is
simply means that An (λ) = + 1 if the two unit vectors not so much the operational accounting of the elements
n and λ happen to point through the same hemisphere of physical reality, but how these elements are coalesced
centered at the origin of n, and An (λ) = − 1 otherwise. together topologically. As we shall see, the correct result
As a visual aid to Bell’s model [5] one can think of a bomb for the spin correlations, derived using both operationally
at rest exploding into two freely moving fragments with and topologically complete prescription, works out to be
angular momenta λ = J1 = −J2 , with J1 + J2 = 0. The
two functions Aa (J1 ) and Bb (J2 ) can then be taken as E(a, b) = − a · b , (5)
sign (λ · a) and sign ( − λ · b), respectively. If the initial
directions of the two angular momenta are uncontrollable which extends
√ the bound on the Bell-CHSH inequality
but describable by an isotropic probability distribution from 2 to 2 2. To fully understand this classical result
ρ(λ) (normalized on the space E3 ), then, employing the let us take a closer look at the derivation of Eq.(3).
 2

Since the initial distribution of the angular momenta the latter case on purely operational basis, is simply an
is supposed to have been isotropic, in Bell’s model the illusion. This peculiar property of the ordinary objects is
space of all possible directions of J1 —that is, both the not respected by the structure of SO(3). That is to say,
configuration space as well as the phase space of J1 —is SO(3) is capable of providing only tensor representations
traditionally [5] taken to be a unit 2-sphere, defined by of the rotation group, and not its spinor representations.
This is fine as long as one is concerned with rotations of
n2x + n2y + n2z = 1. (6)
only isolated objects, but it is anything but fine in our
Next, since each point λ on this surface represents an case, since we are concerned with correlations between
EPR element of physical reality, the integration in Eq.(2) two macroscopic bomb fragments rotating in tandem.
is carried out over this surface, yielding the result (3). Fortunately [7][9], all of the above difficulties can be
Now the group of linear transformations that leave such resolved by representing rotations in physical space by
a quadratic form invariant is the orthogonal group O(3), elements of the universal covering group of SO(3), namely
which includes both rotational and reflective symmetries the group SU(2) of unit quaternions (or spinors [8], or
of the 2-sphere. On the other hand, we know that angular rotors [10]). This group can be constructed by taking
momentum is not an ordinary polar vector, but a pseudo two copies of SO(3), and gluing their boundaries together
vector that changes sign upon reflection. One only needs point by point, so that each −π rotation-point on the
to compare a spinning object with its image in a mirror to boundary of one copy is identified with the respective
confirm this fact. This familiar fact is sufficient, however, +π rotation-point on the boundary of the second copy.
to divulge the first sign of trouble with Bell’s chosen set The resulting space is a topological 3-sphere defined by
of observables—namely, sign (λ · n). Clearly, since λ is
n2o + n2x + n2y + n2z = 1, (7)
supposed to be the spin angular momentum whereas n
is simply an ordinary polar vector, the dot product in where the quadruple ( no , nx , ny , nz ) defines a non-pure
the observable sign (λ · n) cannot be a true scalar, but a unit quaternion [7]. The 3-sphere is well known of course
pseudo scalar—one that changes sign in the mirror. to have exceptionally special properties [11]. It is the only
There is of course an easy way out of this problem. three-dimensional manifold without boundary that is not
All one has to do is to restrict the symmetry group of only compact and connected, but also simply-connected.
the 2-sphere to the subgroup SO(3)—i.e., to the group And it is the only simply-connected, parallelizable sphere
of non-reflective symmetries. This seems straightforward that is homeomorphic to a Lie group, namely SU(2) (it is
enough, but one must bear in mind that, although their also worth noting the obvious that the usefulness of this
Lie algebras are identical, globally the groups O(3) and group is not exclusive to quantum mechanics [9]).
SO(3) are profoundly different from each other. Globally Despite its being contained in R4 , it is in fact possible
the group SO(3) is a highly non-trivial subgroup of O(3). to “see” inside this sphere by means of a Hopf fibration
Indeed, topologically the space SO(3) is homeomorphic [12]. This provides us an opportunity to appreciate the
to the real projective space RP3 , which is a connected, true topological structure of the elements of reality for
but not simply-connected manifold [7][8]. That is to say, our bomb fragments. As illustrated in Fig. 1, the 2-sphere
there are loops in SO(3) that cannot be contracted to a we started out with, namely the one defined by Eq.(6),
point. In physical applications this fact is well known to turns out to be only the base manifold of this profound
give rise to unavoidable singularities, discontinuities, and structure. The points of this base manifold, namely S 2 ,
wildly spinning trajectories [9]. In addition to this fatal now correspond to elements of the Lie algebra su(2), and
defect, the group SO(3) also harbors a related conceptual are in fact pure quaternionic numbers [7][8]. The product
defect, which is of profound significance for our concerns. of two such numbers on S 2 are then general quaternionic
The trouble is that SO(3) does not always respect the numbers, defined by (7), and belong to the group SU(2)
true rotational symmetries of the physical space. itself. That is to say, they are points on the bundle space
To appreciate this well known fact, consider a rock in S 3 , which is completely made up of the preimages of the
an otherwise empty universe. If such a rock is allowed to points on the base S 2 [12]. These preimages are 1-spheres,
rotate by 2π radians about some axis, then it will return S 1 , called Hopf circles, or Clifford parallels ([8], p 335).
back to its original state. This, however, will not happen Since these 1-spheres are the fibers of the bundle, they do
if there is at least one other object present in the universe. not share a single point in common. And yet each circle
The rock will then have to rotate by another 2π radians threads through every other circle in the bundle, making
(i.e., a total of 4π radians) to return back to its original them all linked together in a highly intricate fashion. In
state, relative to that other object. This well known fact particular, although locally the bundle S 3 is a product
is often demonstrated by a “belt trick” (cf. [8], p 205), space S 2 × S 1 , globally it has no cross-section at all.
which shows that what is an identity transformation for It should be fairly clear by now that topologically the
an isolated object is not an identity transformation for an EPR elements of reality have far deeper structure than
object that is rotating in the presence of other objects. has been hitherto appreciated. Clearly, no prescription
Thus, what appears to be an identity transformation in that ignores this structure can be expected to provide the
 3

where u, v, and w are vectors of arbitrary length, and

3 I is the fundamental volume form on the physical space.
S h−1 (q)
h : S3 → S2 b q The specification of the complete state µ predetermines
b p
the entire geometry of the three-dimensional Euclidean
Hopf fibration space E3 (encapsulated in the Clifford algebra Cl3,0 ). It
h−1(p) determines all scalars by their duality relations with µ,
h−1 : S 2 → S 3 S2 all vectors x by definition µ ∧ x = 0, all bivectors by the
base space duality relation µ · x = µ x, and all quaternions by the
Clifford product x y = x · y + µ · (x × y). The locally
specified beables of the model are then taken to be the
FIG. 1: The tangled web of linked Hopf circles depicting the unit bivectors µ · n, which are elements of the Lie algebra
topological intricacies of the EPR elements of physical reality. su(2) of the group SU(2), with the following properties:

µ · n = ± 1 about the dual vector n, (13)

correct correlation function for our bomb fragments. In and ( µ · a) ( µ · b) = − a · b − µ · (a × b). (14)
particular, no Bell-type scalar functions of the form
Note that the Clifford product (µ · a)(µ · b) within Cl3,0
An (λ) : V × Λ −→ S 0 ≡ {−1, +1}, (8)
is also a group product within SU(2), yielding a non-pure
with Aa (λ)Bb (λ) ∈ S 0 × S 0 = S 0 ≡ {−1, +1} (9) unit quaternion [7][12]. This can be verified by comparing
(where V is a vector space and Λ is a space of “complete” the decomposition of the above product with Eq.(7). A
states), can account for the topological intricacies of the generalized expectation functional analogous to (2) then
elements of physical reality, even for our purely classical gives the correct correlation function for our rotors:
rotors. Surely, no elements of a 0-sphere can imitate the Z
topological profundities of the elements of a 3-sphere. E(a, b) = ( µ · a )( µ · b ) dρ(µ) = − a · b, (15)
V3
On the other hand, from the above picture, and from
the fact that the groups O(3), SO(3), and SU(2) all share where the integral is a directed integral, V3 is a manifold
the same Lie algebra structure, it is clear that simply whose “points” are vectors in E3 , and the distribution
promoting the variables An (λ) to be the elements of Lie ρ(µ) is assumed to be normalized on this vector manifold.
algebra su(2) cannot be sufficient to capture the global, Clearly, unlike Eq.(2), the above prescription is not only
topological features of the group SU(2). Capturing these operationally complete, but also topologically complete.
features is mandatory, however, if we are to represent As shown in Ref. [2], if we now substitute this correlation
the EPR elements of physical reality faithfully within the function into the CHSH string of expectation values, √ then
choice of our dynamical variables. Hence, recalling the the bound on its absolute value is extended to 2 2.
definition of a group, what we must ensure is not only It is worth noting here that the model described above
that the local variables—which we write as An (µ)—are is not only manifestly realistic, but also intrinsically local.
functions of the elements of the Lie algebra su(2), but There are several independent ways to verify the latter
that their group products, Aa (µ) Bb (µ), appearing in the fact [2]. To begin with, it is evident from their bivectorial
integrand of Eq.(2), are themselves genuine elements of constitution that the two remote beables µ · a and µ · b
the group SU(2); for what is captured by the Lie algebra have nothing to do with each other. In fact, since they are
su(2) is only the tangent structure of the group SU(2). two genuine elements of the Lie algebra su(2), they are
At the same time, we must also ensure of course that the necessarily two independent points on the corresponding
revised variables An (µ) remain operationally identical to 2-sphere. Moreover, as rigorously proved in Ref. [2], the
the original variables An (λ). We may then have a chance above model satisfies not only the condition of parameter
of evaluating their correlations correctly, by employing an independence, but also that of outcome independence.
appropriately generalized expectation functional [13]. In The central message of Refs. [1] and [2] and the above
sum, the necessary and sufficient conditions on the local discussion is that EPR-Bohm correlations have nothing
beables for obtaining the correct correlation function are: to do with entanglement or non-locality per se, but are a
vestige of geometry and topology of the physical space.
su(2) ∋ An (µ) = ±1 about n, (10)
This recognition almost immediately leads to prediction
and SU(2) ∋ Aa (µ) Bb (µ) = a unit quaternion. (11) (15), which differs from the prediction (3) derived on the
A local realistic model for the EPR-Bohm correlations basis of Bell’s prescription (2). These two predictions are
satisfying precisely these conditions has been proposed in clearly distinguishable. The experiment described below
Ref.[1]. The complete state specifying all of the elements to distinguish them is essentially a realization of Bell’s
of reality in this model is taken to be the unit trivector own local model discussed above [5]. It can be performed
either in the outer space or in a terrestrial laboratory. In
µ = u ∧ v ∧ w = ± I ≡ ± ex ∧ ey ∧ ez , (12) the latter case, however, the effects of gravity and air
 4

resistance would complicate matters. For simplicity we bomb fragments can then be calculated using the formula
shall assume that experimental parameters can be chosen
1 X
N
sufficiently carefully to compensate such effects.
E(a, b) = {sign (λj · a)} {sign (−λj · b)}, (16)
With this assumption, consider a “bomb” made out of N j =1
a hollow toy ball of diameter, say, three centimeters. The
thin hemispherical shells of uniform density that make where N is the number of trials. This result, which would
up the ball are snapped together at their rims in such a give purely local correlations, should then be compared
manner that a slight increase in temperature would pop (in N → ∞ limit) with the predictions (3) and (15).
the ball open into its two constituents with considerable It is worth recalling here that the variables sign (λ · n)
force [5]. A small lump of density much grater than the and µ · n used in the respective derivations of equations
density of the ball is attached on the inner surface of each (3) and (15) are operationally identical to each other:
shell at a random location, so that, when the ball pops
open, not only would the two shells propagate with equal sign (λ · n) ∼
= ± 1 about n ∼
= µ · n. (17)
and opposite linear momenta orthogonal to their common This can be easily verified by noting that the variables
plane, but would also rotate with equal and opposite spin sign (λ · n) are simply the normalized components of the
momenta about a random axis in space. The volume of angular momenta J along the directions n, and so are
the attached lumps can be as small as a cubic millimeter, the variables µ · n (albeit in the bivector basis [10]). In
whereas their mass can be comparable to the mass of the other words, although mathematically sign (λ · n) and
ball. This will facilitate some 106 possible spin directions µ · n are elements of two different grades in the algebra
for the two shells, whose outer surfaces can be decorated Cl3,0 (one is a scalar and the other a bivector), physically
with colors to make their rotations easily detectable. they represent one and the same rotor quantity [2].
Now consider a large ensemble of such balls, identical Undoubtedly, there would be many different sources of
in every respect except for the relative locations of the systematic errors in an experiment such as this. If it is
two lumps (affixed randomly on the inner surface of each performed carefully enough, however, then—in the light
shell). The balls are then placed over a heater—one at a of the discussion above—we believe the experiment will
time—at the center of an EPR-Bohm type setup [6], with vindicate prediction (15) and refute prediction (3).
the common plane of their shells held perpendicular to
I am grateful to Simon Saunders for raising the issue of
the horizontal direction of the setup. Although initially
macroscopic violation of Bell inequalities, and to Abner
at rest, a slight increase in temperature of each ball will
Shimony for correspondence concerning Refs. [1] and [2].
eventually eject its two shells towards the observation
stations, situated at a chosen distance in the mutually
opposite directions. Instead of selecting the directions a
and b for observing spin components, however, one or
more contact-less rotational motion sensors—capable of ∗
Electronic address: [email protected]
determining the precise direction of rotation—are placed [1] J. Christian, Disproof of Bell’s Theorem by Clifford Al-
near each of the two stations, interfaced with a computer. gebra Valued Local Variables: arXiv:quant-ph/0703179.
These sensors will determine the exact direction of the [2] J. Christian, Disproof of Bell’s Theorem: Further Con-
solidations: arXiv:0707.1333; See also arXiv:0904.4259.
angular momentum λj (or −λj ) for each shell, without
[3] A. Einstein, B. Podolsky, and N. Rosen, Phys. Rev. 47,
disturbing them otherwise, at a designated distance from 777 (1935); A. Einstein, Dialectica 2, 320 (1948).
the center. The interfaced computers can then record this [4] J. S. Bell, Physics 1, 195 (1964).
data, in the form of a 3D map of all such directions. [5] A. Peres, Am. J. Phys. 46, 745 (1978); A. Peres, Quan-
Once the actual directions of the angular momenta for tum Theory: Concepts and Methods (Kluwer, Dordrecht,
a large ensemble of shells on both sides are fully recorded, 1993), p 160; M. Redhead, Incompleteness, Nonlocality,
the two computers are instructed to randomly choose the and Realism (Clarendon Press, Oxford, 1989), p 86.
[6] J. F. Clauser, M. A. Horne, A. Shimony, and R. A. Holt,
reference directions, a for one station and b for the other
Phys. Rev. Lett. 23, 880 (1969).
station—from within their already existing 3D maps of [7] Y. Choquet-Bruhat, C. DeWitt-Morette, and M. Dillard-
data—and then calculate the corresponding dynamical Bleick, Analysis, Manifolds and Physics: Parts I and II,
variables sign (λj · a) and sign ( − λj · b). This “delayed Revised Editions (North Holland, Amsterdam, 2000).
choice” of a and b will guarantee that the conditions [8] R. Penrose, The Road to Reality: A Complete Guide to
of parameter independence and outcome independence the Laws of the Universe (Jonathan Cape, London, 2004).
are strictly respected within the experiment [2]. It will [9] W. B. Heard, Rigid Body Mechanics (Wiley, Weinheim,
2006); F. S. Grassia, J. Graphic Tools 3, 29 (1998).
ensure, for example, that the local outcome sign (λj · a)
[10] D. Hestenes, Am. J. Phys. 71, 104 (2003).
remains independent not only of the remote parameter b, [11] J. W. Morgan, Bull. Am. Math. Soc. 42, 57 (2004).
but also of the remote outcome sign (−λj · b). If in any [12] L. H. Ryder, J. Phys. A 13, 437 (1980); D. W. Lyons,
doubt, the two computers can be located at a sufficiently Mathematical Magazine 76, 87 (2003).
large distance from each other to ensure local causality [13] I. E. Segal, Am. J. Math. 76, 721 (1954).
while selecting a and b. The correlation function for the