The Pilot-Wave Perspective on Spin

Travis Norsen
Smith College
Northampton, MA 01060
tnorsen@smith.edu
(Dated: May 6, 2013)
The alternative pilot-wave theory of quantum phenomena – associated especially with Louis de
Broglie, David Bohm, and John Bell – reproduces the statistical predictions of ordinary quantum
mechanics, but without recourse to special ad hoc axioms pertaining to measurement. That (and
how) it does so is relatively straightforward to understand in the case of position measurements
and, more generally, measurements whose outcome is ultimately registered by the position of a
arXiv:1305.1280v2 [quant-ph] 10 Sep 2013

pointer. Despite a widespread belief to the contrary among physicists, the theory can also account
successfully for phenomena involving spin. The main goal of the paper is to explain how the pilot-
wave theory’s account of spin works. Along the way, we provide illuminating comparisons between
the orthodox and pilot-wave accounts of spin and address some puzzles about how the pilot-wave
theory relates to the important theorems of Kochen and Specker and Bell.

I. INTRODUCTION Critics of the theory often argue that the addition of

these definite particle positions is pointless (or “meta-
In an earlier paper, which readers are encouraged to physical”) because at the end of the day the theory’s
examine first and which I refer to hereafter as “the ear- empirical predictions match those of ordinary quantum
lier paper,” I attempted to give a physicists’ introduction mechanics, which latter predictions are of course made
to the alternative, de Broglie - Bohm pilot-wave theory using the wave function alone. What the critics fail to
of quantum phenomena. [1] As a so-called “hidden vari- appreciate, however, is that adding the particle positions
able theory” the pilot-wave theory adds something to the allows something to be subtracted elsewhere in the sys-
state descriptions of ordinary quantum mechanics: in ad- tem. In particular, the dynamical laws sketched above –
dition to the usual wave function Ψ obeying the usual namely, Equations (1) and (2) – constitute the entirety
Schrödinger equation of the dynamical postulates of the theory. No additional
axioms or special exceptions to the usual rules – such as
∂Ψ the collapse postulate of ordinary QM – need to be in-
i~ = ĤΨ (1)
∂t troduced in order to understand measurement or, more
generally, the emergence of the familiar everyday classical
one also has definite positions for each particle in the world.
system. For example, for a system of N spinless non-
relativistic particles, the position X ~ n of the nth particle In the earlier paper, this point was developed in the
will evolve according to context of some simple scattering experiments involving
single particles. In such situations, the fact that the scat-
tered particle is found, whole, at some particular place

dX~ n (t) ~jn (~x1 , ~x2 , ..., ~xN , t) 
= (2) at the end of the experiment is explained in the simplest
dt ρ(~x1 , ~x2 , ..., ~xN , t) 

~
xk =Xk (t)∀k
~ imaginable way: there really is a particle, following a def-
  inite trajectory and hence possessing a perfectly definite
where ~jn = 2mi~ ~ n Ψ − Ψ∇
Ψ∗ ∇ ~ n Ψ∗ is (what in ordi- position at all times. The detector finds it at a particular
place (even when its wave function is spread out across
nary QM is termed) the probability current associated
several different places) because it is already at a par-
with particle n and ρ = Ψ∗ Ψ is (what in ordinary QM is
ticular place before interacting with the detector. And
termed) the probability density.
equivariance, in light of the QEH, guarantees that the
As explained in the earlier paper, the pilot-wave the-
probability for the particle to be at a certain place at the
ory involves the quantum equilibrium hypothesis (QEH)
end of the experiment perfectly matches what ordinary
according to which the particle positions are assumed to
QM would instead describe as the probability that the
be random, with initial (t = 0) probability distribution
measurement intervention triggers a collapse that makes
~ N = ~xN ] = |Ψ(~x1 , ..., ~xN , 0)|2 .
~ 1 = ~x1 , ..., X the particle suddenly materialize at that place. It is thus
P [X (3)
clear, in pattern, how the pilot-wave theory reproduces
It is then a purely mathematical consequence of Equa- the statistical predictions of ordinary QM for experi-
tions (1) and (2) that the probability distribution will ments which end with the measurement of the position
remain |Ψ|2 -distributed for all times. The family of pos- of a particle.
sible particle trajectories thus “flows along with” ρ, a That the pilot-wave theory makes the same predictions
property that has been dubbed the “equivariance” of the as ordinary QM for other kinds of measurements as well
|Ψ|2 probability distribution. [2] can be understood by including the particles constituting
 2

the measuring device in the system under study. Con- pilot-wave type of analysis just sketched. Indeed, it is
sider for example a simple schematic treatment of the sometimes reported that the pilot-wave theory simply
measurement of, say, the energy of a particle with initial cannot deal with spin phenomena in a plausible way. [5]
wave function ψ0 (x). [3] The measurement apparatus is This view was expressed very eloquently by one of the
imagined to have a pointer with spatial coordinate y and anonymous referees of the earlier paper:
initial wave function φ(y) = φ0 (y) where φ0 is a narrow
packet centered at the origin, corresponding to a “ready” “It has indeed been shown that Bohmian me-
state for the apparatus. chanics is equivalent to non-relativistic quan-
An interaction between the particle and the pointer is tum mechanics with respect to its predictions
regarded as a “measurement of the energy of the parti- concerning particle positions. However the
cle” just in case the Schrödinger equation (for the parti- scheme encounters major problems: in spite
cle+pointer system) produces the following sort of time- of a half-century of effort by its adherents, it
evolution in the case that the particle is initially in an has not been possible to incorporate spin into
“energy eigenstate” ψ0 (x) = ψi (x) with eigenvalue Ei : the theory in a convincing way...”
ψi (x)φ0 (y) → ψi (x)φo (y − λEi ). (4) This type of view is perhaps motivated by the various “no
That is, at the end of the experiment, the wave function hidden variables” theorems, such as that due to Kochen
for the pointer is now sharply peaked around some new and Specker, which show quite clearly that it is mathe-
point, y = λEi , proportional to the initial energy Ei of matically impossible to assign pre-existing values to all
the particle. The pointer, in short, indicates the energy relevant spin-components of an ensemble of particles such
of the particle. that the quantum mechanical predictions are reproduced.
But the linearity of the Schrödinger equation then im- [6, 7] That is, it is known to be impossible to do, for all of
mediately implies that, in the general case in which the the non-commuting components of spin, what the pilot-
particle is initially in an arbitrary superposition of energy wave theory, as sketched above, does for position.
eigenstates, the evolution goes as follows: And yet, in fact, it is entirely false that the pilot-wave
! theory cannot deal with spin. Indeed, the truth is that
X
ci ψi (x) φ0 (y) →
X
ci ψi (x)φ0 (y − λEi ). (5) the pilot-wave theory deals with spin in an almost shock-
i i
ingly natural – certainly a shockingly trivial – way. The
goal of the rest of this paper is to resolve this paradox
The resulting “Schrödinger cat” type state is of course and explain how.
problematic from the point of view of ordinary QM: in- The main ideas of the paper are not new. The pilot-
stead of registering some one definite outcome for the wave theory was applied to spin already in 1955 by Bohm
experiment, the pointer itself ends up in an entangled et al. [8], and numerical simulations of the theory’s ac-
superposition. Enter the collapse postulate to save us count of spin measurements were carried out by Dewd-
from the troubling implications of universal Schrödinger ney et al. in the 1980s. [9] Our approach here, following
evolution! But this is no problem at all in the pilot-wave Bell’s more elegant treatment in Ref. [10], instead aims at
theory, according to which the (directly perceivable) fi- simplicity and accessibility. In particular, the proposed
nal disposition of the pointer is not to be found in the delta-function model of a Stern-Gerlach experiment (the
wave function, but instead in the actual position Y of one genuine novelty in the paper) allows one to solve
the pointer particle at the end of the experiment. It the Schrödinger equation and determine the pilot-wave
follows from the equivariance property discussed earlier particle trajectories (without any recourse to numerical
that, assuming the initial particle positions are random in simulations) using the “plane-wave packet” methods de-
accordance with the QEH, there will be a probability |ci |2 veloped in the earlier paper.
that the actual value of Y at the end of the experiment
That model is developed in the following section; the
is (near) λEi . That is, the pointer will point to a defi-
associated pilot-wave particle trajectories are then dis-
nite place, with probabilities given by the usual quantum
played and analyzed in Section III. Section IV explains
rules, but, here, by rigorously following the basic dynam-
how the pilot-wave theory deals with cases of repeated
ical rules, rather than making up special exceptions when
spin measurements, while Sections V and VI develop
they produce embarrassing results.
examples to illustrate the “contextuality” and “non-
One can thus understand how the pilot-wave theory
locality” exhibited by the theory. Some concluding re-
manages to reproduce the statistical predictions of or-
marks about the relationship between the pilot-wave and
dinary QM, at least in cases where one measures posi-
orthodox points of view are then made in Section VII.
tion or measures some other quantity but by means of
the position of a pointer. But... what about spin? Of
all the phenomena surveyed in undergraduate quantum
mechanics courses, those involving intrinsic spin and its II. THE STERN-GERLACH EXPERIMENT
measurement – which Pauli famously described as indi-
cating a mysterious, non-classical Zweideutigkeit or “two- The Stern-Gerlach experiment, first and most fa-
valuedness” [4] – seem perhaps the least amenable to the mously performed in 1922, involves subjecting a beam of
 3

particles to an inhomogenous magnetic field so the parti- z

cles experience a force proportional to a certain compo-
nent of their intrinsic spin angular momenta. [11] Since,
empirically, the incident beam gets split into two or more
S
discrete sub-beams (as opposed to a continuous distribu-
tion), the experiment is usually understood as demon- +z
c +χ
strating the quantization of spin angular momentum.
c+ χ+z + c− χ−z
In a standard textbook treatment of the experiment, y
we assume a magnetic field c− χ
−z

~ ≈ βz ẑ
B (6) N

as might be produced, for example, by the magnets in-

dicated in Figure 1. [12] For a neutral spin-1/2 particle
(such as the silver atoms used in the original S-G exper- FIG. 1: Neutral spin-1/2 particles with initial spin state
iment) the interaction Hamiltonian is c+ χ+z + c− χ−z move in the y-direction toward a Stern-
Gerlach apparatus with magnetic field gradient along the z-
~ = −µβzσz
Ĥ = −µ~σ · B (7) direction. The field delivers an impulse in the ±z-direction to
the χ±z components of the beam, so downstream there arise
where µ is the magnitude of the particle’s magnetic mo- two spatially non-overlapping sub-beams with amplitudes c+
ment and the components of ~σ are, in the usual repre- and c− . In ordinary quantum mechanics, a measurement of
sentation, the Pauli matrices. The eigenstates of this the position of the particle downstream from the S-G device
interaction Hamiltonian will obviously

be (proportional will find the particle in (or more accurately, cause it to mate-
to) spinors χ+z = 10 and χ−z = 01 . satisfying rialize in) the upper/lower sub-beam with probability |c± |2 .

σz χ±z = ±χ±z . (8)

Now let us consider the spatial degrees of freedom of the

particle’s wave function, focusing in particular on the di- But then, since Schrödinger’s equation is linear, it is im-
mension (here, z) parallel to the magnetic field gradi- mediately obvious that the S-G device will split the in-
ent. The beam is initially propagating, say, in the +y- coming wave function into two sub-beams – one with rel-
direction, so let us assume that, in the z-direction, the ative amplitude c+ that is deflected up, and one with
wave function is initially constant (over the range where relative amplitude c− that is deflected down. The net
it is nonzero) and that the particles are in eigenstates of effect is pictured in Figure 1.
σz : For purposes of discussing the pilot-wave account of
these phenomena, it will be helpful to work with a sim-
Ψ(z, 0) = Aχ±z . (9) plified model that captures all the physics we’ve just re-
viewed, but includes also an explicit treatment of the par-
Now suppose the particles experience the magnetic field ticle’s other spatial degrees of freedom. This can be done,
for a finite period of time T during which the interaction in the spirit of Figure 1, by imagining that the field B~ is
Hamiltonian above dominates all other terms. Then the very strong, but is non-zero only in a vanishingly small
Schrödinger equation reads region around y = 0. That is, the interaction Hamilto-
nian of Equation (7) is replaced with Ĥ = −µbzδ(y)σz ,
∂Ψ so that the total Hamiltonian governing the particle’s
i~ = −µβzσz Ψ (10)
∂t motion in the y − z-plane is
which, for the assumed initial condition, has solution
~2 ∂ 2 ~2 ∂ 2
Ĥ = − − − µbzδ(y)σz . (13)
Ψ(z, T ) = Ae±iκz χ±z (11) 2m ∂y 2 2m ∂z 2

where ~κ = µβT is the magnitude of the (upward or Now the situation can be treated like an elementary 2D
downward) momentum that the particle acquires as a scattering problem in wave mechanics.
result of traversing the field. For a beam of particles ini- To begin with, we imagine a “plane-wave packet” like
tially propagating in the y-direction, this impulse in the those described in the earlier paper: the initial wave func-
±z-direction deflects the beam either upward or down- tion should be a packet of length L along the y-direction
ward depending on whether the particle was initially in and width w along the z-direction, with constant ampli-
the spin state χ+z or χ−z . Of course, in general, the ini- tude in the region where its amplitude doesn’t vanish,
tial spin might be an arbitrary linear combination of the propagating initially in the +y-direction with (reason-
two z-spin eigenstates: ably sharply-defined) energy E = ~2 k 2 /2m. Let us again
start by assuming that the incident particle’s spin degrees
χ0 = c+ χ+z + c− χ−z . (12) of freedom are described by one of the eigenspinors, χ±z .
 4

Then, in the time period during which the incident packet incident packet should be small compared to 1/κ. Then
is interacting with the field at y = 0, the wave function Equation (14) provides an acceptable description of the
can be taken to be (up to an overall time-dependent phase wave function in the vicinity of the Stern-Gerlach appa-
which we omit for simplicity) ratus while the length-L packet interacts with it. Note
that, in these limits, the (z-dependent) amplitude of the
A eiky χ±z + Bz e−iky χ±z for y < 0

Ψ(y, z) = ′ (14) reflected wave is (everywhere) small. Since it also plays
C eik y e±iκz χ±z for y > 0 no significant role in the physics to be discussed, we there-
fore drop it. (Those interested in the details should see
in the appropriate regions of the y − z plane. (Note that note [13].)
the factor of z – in the term, involving B, representing a So far we have assumed that the spin degrees of free-
reflected wave – is put in for future convenience.) This dom are given by one of the (z-direction) eigenspinors,
expression is a valid solution to the time-independent χ±z . But it is now straightforward to appeal to the lin-
Schrödinger equation for y < 0 and y > 0 as long as earity of the Schrödinger equation in order to write a
solution appropriate for an arbitrary initial spinor, like
~2 k 2 ~2 (k ′2 + κ2 )
E= = . (15) that of Equation (12):
2m 2m
A eiky (c+ χ+z + c− χ−z )

In addition, the above expression should solve the Ψ(y, z) = ik′ y

,y<0
Schrödinger equation at y = 0. This will be the case Ae e+iκz c+ χ+z + e−iκz c− χ−z , y > 0
if the wave function is continuous at y = 0 (21)
where, of course, the expression for y < 0 applies only
A + Bz = C e±iκz (16) within the range −w/2 < z < w/2 where the plane-wave
packet has support, and the expression for y > 0 applies
and if the following condition on the y-derivatives only within the triangular “overlap region” – see Figure
of ψ (arrived at by integrating the time-independent 1 – where the two (separating) components of the wave
Schrödinger equation from y = 0− to y = 0+ ) is sat- coincide.
isfied:
  !
∂ψ  ∂ψ  2mµb
− = ∓ 2 zψ(0, z). (17) III. PARTICLE TRAJECTORIES IN THE
∂y y=0+ ∂y y=0− ~
PILOT-WAVE THEORY

Plugging in our explicit expression for ψ(y, z) converts

Equation (17) into Following the methods introduced in the earlier pa-
per, we can use the structure of the wave function in
2mµb (especially) the “overlap region” to analyze the motion
ik ′ Ce±iκz − ikA + ikBz = ∓ zC e±iκz . (18) of the particles in the pilot-wave theory and to demon-
~2
strate explicitly that the theory reproduces the usual
It is obvious that these two conditions, Equations (16) quantum mechanical predictions for the outcome of a
and (18), cannot be satisfied exactly for all z. How- Stern-Gerlach spin measurement. The first question that
ever, they can be approximately satisfied over the narrow must be addressed, though, is this: what is the analog
range of z where the (width-w) wave packet has support. Ψ+


of Equation (2) when the wave function Ψ = Ψ is a
Thus, for example, Equation (16) is valid to first order −
multi-component spinor? The answer is simply that the
in κw if A = C and equation is exactly the same, except we must sum over
B = ±iκC. (19) the spin index in both the numerator and denominator.
Or equivalently, writing Ψ† = (Ψ∗+ Ψ∗− ), we say
Then Equation (18) is satisfied to the same degree of ap-     
proximation if k ≈ k ′ (which implies, in light of Equation ~ Ψ † ~
∇Ψ − ~ † Ψ 
∇Ψ
dX(t) ~
(15) that κ ≪ k) and =  . (22)
dt 2mi Ψ† Ψ 
 ~
x=X(t)
~
mµb
κ= . (20)
~2 k Consider, for example, the wave function Ψ(y, z) in the
Notice that there are two distinct physical assumptions y < 0 region from Equation (21). We have that
here. First, the impulse (of magnitude ~κ) imparted to  
the particles by the Stern-Gerlach fields must be small iky c+
Ψ(y, z) = Ae (23)
compared to the initial momentum ~k of the particles. c−
Or equivalently, κ/k = µb/2E should be small. (Notice
that b has the same units as a magnetic field even though so that
it is not one!) That is, the magnetic fields should be ap-
propriately “gentle”. And second, the width w of the Ψ† (y, z) = A∗ e−iky (c∗+ c∗− ). (24)
 5

S S

wP+ ∆y
w
∆z y

Critical
Trajectory
N
N

FIG. 2: Representative sample of possible particle trajec- FIG. 3: Particles which begin above the critical trajectory
tories through a z-oriented Stern-Gerlach apparatus p for an will be shunted into the upper (“spin-up”) wave packet down-
initial p
wave function with spinor components c+ = 2/3 and stream of the Stern-Gerlach magnets, while those which be-
c− = 1/3. The particle will move along with the incident gin below the critical trajectory will be shunted into the lower
wave until crossing over into the overlap region where the + (“spin-down”) packet. For initial particle positions that are
and − components of the wave are beginning to separate. In (in accordance with the QEH) uniformly distributed across
this region, the particle velocity is the weighted average, given the lateral extent of the incident packet, the probability for
in Equation (26), of the group velocities associated with the a particle to emerge with the “spin-up” packet will be the
two separating components of the wave. Depending on its fraction of the lateral width (w) of the packet that is above
initial lateral position within the incoming wave packet, the the critical trajectory – i.e., the distance from the critical tra-
particle will either be shunted into the upper (+) or lower (−) jectory to the top of the incident packet is wP+ . The vertical
fork. Subsequent detection of the particle in the upper/lower displacement ∆z of the critical trajectory as it traverses the
fork will result in its being identified as “spin up/down along triangular overlap region is then wP+ − w/2. By considering
z”. the right triangle that is the lower half of the overlap region
(and whose hypotenuse is one edge of the “spin-up” packet),
one can see that the horizontal extent of the overlap region,
∆y, is given by w2 kκ′ : the slope, (w/2)/∆y, of the hypotenuse
Plugging into Equation (22) then gives should match the ratio κ/k′ of the z- and y-components of
the group velocity associated with the “spin-up” part of the
dX~ ~k wave.
= ŷ. (25)
dt m
No matter its exact location within the wave, the particle
will simply drift in the positive y-direction with the group velocity of the − component and all incoming particles
velocity of the incident packet. [13] will be shunted downward.
Eventually of course the particle will encounter the The equivariance property mentioned in the introduc-
fields at y = 0 and pass over into the overlap region on tion implies that if the position of the particle within the
the y > 0 side. Here, because the + and − components wave is random and |Ψ|2 -distributed at t = 0, it will re-
of Ψ have different dependencies on z, we end up with main |Ψ|2 -distributed for all subsequent times. This has
the immediate consequence that the probability for the
dX~ ~k ′ ~κ particle to be located, downstream of the Stern-Gerlach
|c+ |2 − |c− |2 ẑ


= ŷ + (26) device, in the upper (respectively, lower) sub-beam – and
dt m m
hence counted as “spin-up” (respectively, “spin-down”)
which can be understood as a weighted average of the if its position there is detected – is |c+ |2 (respectively,
group velocities associated with the two (now differently- |c− |2 ). It is thus clear that the pilot-wave theory repro-
directed) components of the wave. duces the usual quantum mechanical predictions for the
The family of possible particle trajectories thus looks statistics of such spin measurements.
something like that illustrated in Figure 2 for the partic- One of the nice things about the way we have set things
ular case that |c+ | is a little bigger than |c− |. It should up, though, is that this claim can be demonstrated ex-
be clear, for example, that if c+ = 1 and c− = 0, then, plicitly by considering the “critical trajectory” [14] which
while in the overlap region, the particles will just move divides the trajectories emerging into the upper and lower
with the group velocity of the + component of the wave beams. By definition, the criticial trajectory arrives just
function, and all incoming particles (no matter their lat- at the vertex on the right of the triangular overlap region
eral position within the wave) will be shunted into the behind the magnets, as shown in Figure 3. The slope of
upward sub-beam. Whereas if c+ = 0 and c− = 1, the the criticial trajectory is given by the ratio of the z−
particles in the overlap region will move with the group and y−components of the velocity in the overlap region,
 6

Equation (26): 50%

SGz
|c+ |2 − |c− |2 κ SGx 50%
slope = . (27) SGz
|c+ |2 + |c− |2 k ′
On the other hand, as explained in the Figure caption,
′
the critical trajectory traverses a distance ∆y = wk FIG. 4: Schematic representation of a series of Stern-Gerlach
2κ in
the y−direction and a distance ∆z = wP+ − w/2 in the spin measurements, in which the devices are replaced by
“black boxes” with two output ports, one for “spin-up” along
z−direction, where P+ is the fraction of the lateral width
the direction n indicated by the box’s label “SGn ”, and one
w of the beam with the property that, should the particle for “spin-down”. Here, particles which emerge as “spin-up”
begin there, it wil end up going into the upper sub-beam. along the z-direction and then “spin-up” along the x-direction
In short, P+ represents the probability that the S-G spin are subjected to a further measurement of z-spin. Half of the
measurement will yield the outcome “spin-up”. Setting particles exiting this final SGz device are “spin-up”, with the
other half being “spin-down”. As discussed in the text, both
∆z ordinary QM and the pilot-wave theory are able to account
slope = (28)
∆y for the observed statistics, because in both theories the state
of a particle is in general affected by subjecting it to a spin
and solving for P+ one indeed recovers that measurement. In particular, the state of a particle exiting
the first SGz device is not the same as the state of that same
P+ = |c+ |2 (29) particle when it exits the intermediate SGx device.
confirming explicitly that, indeed, the pilot-wave dynam-
ics for the wave and particle are compatible in the way
expressed by the formal equivariance property: the par- Perhaps the theory will fail, however, when we consider
ticle trajectories “bend just the right amount” to yield the possibility of subsequent and/or repeated measure-
a final probability distribution for the particle position ments? The pilot-wave theory is, after all, a deterministic
that is consistent with the usual QM predictions and, hidden variable theory: for a given experimental setup,
more importantly, what is seen in experiments. the outcome of a spin measurement is determined by the
initial state of the “particle” (meaning here the parti-
cle+wave combination). Standard textbook discussions
IV. ADDITIONAL MEASUREMENTS
might perhaps suggest that there would be a problem.
For example, suppose that particles emerging from
So far we have concentrated on the measurement of the “spin-up” port of a z-oriented Stern-Gerlach de-
the z-component of the spin using a Stern-Gerlach device vice (SGz ) are subsequently sent through an SGx de-
whose magnetic field gradient is along the z-direction. By vice; those particles emerging as “spin-up” along x are
simply rotating the device, however, we can also measure then subjected to a further measurement of their spin’s z-
the component of the spin along (say) an arbitrary direc- component. (See Figure 4.) The result, of course, is that
tion n̂ in the x − z-plane: fully half of the particles entering it will emerge from
n̂ = cos θẑ + sin θx̂. (30) the final SGz device as “spin-down”. The usual inter-
pretation is that spin-along-z and spin-along-x are in-
The whole measurement procedure can be analyzed in compatible properties, corresponding to non-commuting
exactly the same way we’ve done above, mutatis mutan- quantum mechanical operators, so the determination of
dis: the eigenspinors χ±n will now be the eigenvectors a definite value for spin-along-x completely erases the
of previously-definite value for spin-along-z. Surely a non-
  quantum, deterministic hidden variable theory could not
cos θ sin θ reproduce this paradigmatically quantum result?
σn = , (31)
sin θ − cos θ But a little thought reveals that, yes, the pilot-wave
theory reproduces this prediction quite trivially. The
an arbitrary initial wave-packet (proportional to initial crucial point is that, although the particles entering the
spinor χ0 ) can be written as a linear combination of these SG
x device were being guided by waves proportional to
eigen-spinors 1
0 , the wave guiding those particles which successfully
χ0 = c+n χ+n + c−n χ−n (32) emerge from the “spin-up”
port of the SGx device is pro-
portional to χ+x = √12 11 . The χ+x and χ−x components
and the whole analysis of Sections II and III goes of the wave get separated by the SGx device, whereas
through, with z being simply everywhere replaced by n. the particle goes one way or the other. And, by defi-
[15] So it should be immediately clear that the pilot wave nition, if the particle ends up in the “spin-up-along-x”
theory reproduces the usual quantum statistical predic- beam, downstream of the SGx device, the component of
tions for spin measurements, whether it is the z- or some the wave now surrounding and guiding it, is the “spin-
other component of spin that is to be measured. up-along-x” part. So it is as if the particle’s quantum
 7

state has “collapsed”, as per the usual quantum mechan- (2), defining the theory’s dynamics) whereby the physical
ical rules, even though, in fact, no collapse – no violation measurement intervention affects the state of the mea-
of the usual Schrödinger evolution of the wave – has oc- sured system. There is thus a sense in which, for the
curred. The other, “spin-down-along-x”, part of the wave pilot-wave theory just as for ordinary QM, the measure-
is still present – it’s just “over there”, far away from the ment cannot be thought of as passively revealing some
actual location of the particle, and hence dynamically ir- pre-existing quantity, but should instead be thought of
relevant to the present or future motion of the particle. as an active intervention which brings about the new,
[16] This separation of the different spin components of final state of the particle corresponding to the measure-
the wave packet of course happens in ordinary quantum ment outcome.
mechanics as well. But in ordinary QM, there is nothing But there is an even deeper sense in which, for the
like the actual position of the particle in addition to the pilot-wave theory, the measurement cannot be thought
wave function, to warrant talk of the particle actually of as passively revealing a pre-existing value. Let us dis-
being up there (i.e., talk of the spin measurement having cuss this in terms of a concrete example. Recall the SGz
actually had the outcome “spin-up”), and so additional device sketched in Figure 1: the magnets produce a mag-
dynamical postulates are required. This is an illustration netic field near the origin which can be approximated by
of the point made in the introduction, that although the the expression in Equation (6). The field is such that a
pilot-wave theory adds something to the physical state classical particle, with magnetic dipole moment in the
descriptions, it is not really more (let alone pointlessly +z direction, will feel a force in the +z direction and
more) complicated than ordinary QM, since this addi- hence be deflected up upon passing through the SG de-
tion allows for the subtraction of the otherwise rather vice. Likewise, a classical particle with magnetic dipole
dubious measurement axioms. moment in the −z direction will feel a force in the −z
What is actually showed by this kind of example is direction and will hence be deflected down. This is, in
not that one cannot explain the observed statistics of essence, the basis for saying that, when a particle emerges
spin measurements with a non-quantum, deterministic, from the SG device having been deflected up, it is “spin-
hidden variable theory, but rather only that the physi- up along z”, etc.
cal state of a particle (known, say, to be “spin-up-along- But other ways of measuring the z-component of a
z”) must be affected by a measurement of (for example) particle’s spin can also be contemplated. For example,
its spin-along-x. Ordinary quantum theory of course ex- imagine a device – let us call it here an SG′z device –
hibits this behavior: in quantum mechanics, the state of like that indicated in Figure 5. The structure is identical
the particle, after a measurement, is postulated to “col- to the original SGz device, except that the polarity of
lapse” to the eigenstate of the appropriate operator corre- the magnets has been reversed and hence Equation (6)
sponding to the actually-observed outcome. (It is worth is replaced with
noting here that the “actually-observed outcome” is yet
~ ′ ≈ −βz ẑ.
B (33)
another additional postulate of the theory – for Bohr,
for example, this was realized in some classical macro-
This reversal of the magnetic field gradient reverses the
scopic pointer somewhere, whose physical relationship to
effect on particles passing through the device: now, a
the particle in question is, at best, obscure.) The point
classical particle with a magnetic dipole moment in the
here is that the pilot-wave theory is also exhibits this
+z direction will feel a force in the −z direction and hence
behavior: the physical state of (in particular) the wave
be deflected down upon passing through the device, while
guiding the particle is different before, and after, the in-
a classical particle with magnetic dipole moment in the
termediate SGx device. But in the pilot-wave theory, this
−z direction will feel a force in the +z direction and will
difference – the physical influence of the measurement on
hence be deflected up. The SG′z device is a perfectly valid
the properties of the system in question – is a natural and
one for measuring the z-component of the spin of a par-
straightforward consequence of the usual – the universal,
ticle; it merely has a different “calibration” (one might
the exceptionless – dynamical laws.
say) than the original device. Whereas, with the original
device, particles deflected up are declared to be “spin-up
along z”, particles deflected up by the modifed device
V. CONTEXTUALITY are instead declared to be “spin-down along z”, and vice
versa. And indeed, a full quantum mechanical analysis,
In the previous section, I stressed that the pilot-wave parallel to that undertaken for the original SGz device in
theory is not the “naive” sort of hidden-variable theory Section III, leads (in a rather obvious and trivial way) to
that is sometimes discussed (and, when discussed, always the conclusion that an incident wave packet proportional
refuted with great pomp!) in textbooks. In these “naive” to c+ χ+z + c− χ−z will be split, upon passage through
theories, measurements simply passively reveal some pre- the SG′z device, into two sub-beams, one proportional to
existing value of the property being measured, without c+ χ+z that has been deflected down and the other pro-
affecting the state of the particle. In the pilot-wave the- portional to c− χ−z that has been deflected up.
ory, by contrast, there is a natural mechanism (which, The significance of these two distinct pieces of exper-
quite literally, is nothing but the Equations, (1) and imental apparatus – both equally qualified to “measure
 8

z theory, not just on the initial state of the particle but

also on the overall experimental context – i.e., on the
particular “way” that the measurement is implemented.
This “contextuality” should be contrasted to the “non-
N
contextual” type of hidden variable theory in which, for
c −χ
−z each (Hermitian) quantum mechanical operator (corre-
sponding to some “observable” property), each particle
c+ χ+z + c− χ−z
y in an ensemble possesses a definite value which is simply
c+ χ
+z
revealed by the appropriate kind of experiment, inde-
pendent of which specific experimental implementation
S is used to measure the observable in question. In partic-
ular, a non-contextual hidden variable theory would (by
definition) assign a definite value to the z-component of
each particle’s spin, independent of whether the spin is
FIG. 5: An alternative way to “measure the z-component of measured using an SGz or an SG′z device. More generally,
the spin of a particle” is to pass the particle through the SG′z non-contextuality can be understood as the requirement
device shown here. This is the same as an SGz device, but that each observable should possess a particular definite
with the polarity of the magnets – and hence the calibration value, that will be revealed by a measurement of that ob-
of the device – reversed: particles that deflect up are now servable, no matter what compatible observables are per-
counted as “spin-down along z”, while particles that deflect haps being measured simultaneously. (See the following
down are now counted as “spin-up along z”.
section for a concrete example of this more general sort
of contextuality.)
That the “no hidden variables” theorems of von Neu-
the z-component of the spin of a particle” – becomes clear mann, Jauch-Piron, and Gleason tacitly assume non-
when we consider what happens when, for example, a
contextuality (or an even stronger and less innocent re-
particle with initial wave function proportional to √12 11 quirement) was first clearly pointed out in Bell’s land-
mark 1966 paper, “On the Problem of Hidden Variables
is incident on the two devices. For this case, the parti-
in Quantum Mechanics.” [18] (The paper, incidentally,
cle velocity in the overlap region is proportional to ŷ, i.e.,
was written in 1964, prior to Bell’s more famous 1964
the particle trajectories simply continue in a straight line
paper proving what is now called “Bell’s Theorem”, but
all the way through the overlap region until they finally
remained unpublished until 1966 due to an editorial acci-
emerge into one of the now spatially-separated sub-beams
dent.) The somewhat more famous Kochen-Specker “no
corresponding to their initial lateral positions within the
hidden variables” theorem, which appeared the year after
incident packet. That part is identical, whether the orig-
Bell’s paper, also assumes non-contextuality, as discussed
inal SGz , or instead the alternative SG′z , device is in-
in the beautiful review paper by Mermin. [6, 7]
volved. But there is a crucial difference between the two
As Bell explains, these proofs all rely on relating “in
devices: if (say) the particle happens to start in the up-
a nontrivial way the results of experiments which cannot
per half of the incident packet, it is destined to eventually
be performed simultaneously.” (For example, they as-
find itself in the upward-deflected sub-beam and hence be
sume that the results of an SGz -based and an SG′z -based
counted as “spin-up along z” – if the measurement is car-
measurement should be the same.) Bell elaborates:
ried out using the original SGz device. But if instead the
measurement is carried out using the alternative SG′z de- “It was tacitly assumed that measurement
vice, the exact same “particle” – that is, the exact same of an observable must yield the same value
wave function and the exact same particle location in independently of what other measurements
the upper half of the wave packet – is instead destined to may be made simultaneously. Thus as well
eventually find itself in the upward-deflected sub-beam as [some observable with corresponding QM
and hence be counted as “spin-down along z”. [17] operator Â] say, one might measure either [B̂]
In short, for the (deterministic, hidden variable) pilot- or [Ĉ], where [B̂ and Ĉ commute with Â, but
wave theory, the outcome of “measuring the z-component not necessarily with each other]. These differ-
of the spin of the particle” is not simply a function of ent possibilities require different experimen-
the initial state of the “particle” (i.e., the particle+wave tal arrangements; there is no a priori reason
complex). The exact same initial state can yield either to believe that the results for [Â] should be
of the two possible measurement outcomes, depending on the same. The result of an observation may
which of two possible experimental devices is chosen for reasonably depend not only on the state of
performing the experiment. the system (including hidden variables) but
This is a concrete illustration of the fact that is usually also on the complete disposition of the appa-
put as follows: for the pilot-wave theory, spin is “contex- ratus...” [18]
tual”. That is, the outcome of a measurement of a certain
component of a particle’s spin depends, in the pilot-wave Bell then references Bohr’s insistence on remembering
 9

“the impossibility of any sharp distinction between the These CWFs are convenient because the dynamical law
behavior of atomic objects and the interaction with the for the particle positions, Equation (2), can be reformu-
measuring instruments which serve to define the condi- lated to give, for each particle, an expression for the par-
tions under which the phenomena appear.” [19] Abner ticle velocity in terms of its own CWF:
Shimony has aptly described Bell’s invocation of Bohr 
(the arch-opponent of hidden variables) in Bell’s defense dX~n ~ n ) − (∇Ψ
~ Ψ†n (∇Ψ ~ †n )Ψn 
= . (37)
of hidden variables (against the theorems supposedly, but dt 2mi Ψ†n Ψn

 ~
~x =X
not actually, showing them to be impossible) as a “judo- n

like manoeuvre”. [20] The CWFs can thus be thought of as “one-particle wave
For our purposes, the upshot of all this is the following. functions”, each of which guides its corresponding parti-
In the literature on hidden variables, one often finds the cle, in exactly the way we have discussed previously for
implication that the requirement of non-contextuality is the pilot-wave theory of single particles.
quite reasonable, in the sense that contextuality would The CWFs have however several perhaps-unexpected
supposedly involve putting in “by hand” an obviously properties that should be acknowledged. First, it should
implausible ad hoc kind of apparatus-dependence. But be appreciated that, in general, the CWFs do not obey
as the simple example of the contextuality of the pilot- some simple one-particle Schrödinger equation. [22] Eval-
wave theory (involving the distinct outcomes for SGz - uating Equation (35) at ~x2 = X ~ 2 gives an overall (time-
and SG′z -based measurements of z-spin) makes clear, this dependent) multiplicative factor which means that (even
is not the case at all. The “context-dependence” arises in the case of non-entangled particles) the overall phase
in a perfectly straightforward and natural way, again as and normalization of the CWF can change erratically.
a direct consequence of the fundamental dynamical pos- This, however, is irrelevant to the motion of the associ-
tulates of the theory, Equations (1) and (2). ated particle, since any multiplicative factor simply can-
cels out in Equation (37). In addition, the CWF of each
particle carries spin indices for both particles. In the case
VI. NON-LOCALITY
where the two-particle wave function is factorizable, as
in Equation (35), the particle 2 spinor (χ20 ) is irrelevant
Bell’s 1966 paper closes by noting that Bohm’s 1952 to the motion of particle 1, and vice versa. The “extra”
pilot-wave theory eludes the various “no hidden vari- spinor thus acts just like the overall multiplicative con-
ables” theorems by means of its contextuality – but that stant. It is thus clear that when the two-particle wave
the context-dependence implies a non-local action at a function factorizes – when the two particles are not en-
distance in situations involving entangled but spatially- tangled – the motion of the two particles will be fully in-
separated particles: “in this theory an explicit causal dependent: if the particles subsequently encounter Stern-
mechanism exists whereby the disposition of one piece Gerlach devices, each will (independently and separately)
of apparatus affects the results obtained with a distant act exactly as described in the previous sections.
piece.” [18] Let us explore this type of situation, in which If, however, the spins of the two particles are initially
the pilot-wave theory’s contextuality implies a non-local entangled, as for example in
dependence on a remote context.
Consider then a pair of spin-1/2 particles, propagating 1
Ψ(~x1 , ~x2 ) = ψ(~x1 , ~x2 ) √ χ1+z χ2−z − χ1−z χ2+z


(38)
in opposite directions along (and already widely sepa- 2
rated along) the y-axis. The initial spatial part of the
then the non-locality appears. Suppose for example that
2-particle wave function can be taken to be
both particles are to be subjected to SGz -based mea-
(0,d,0) (0,−d,0) surements of their z-spins. And suppose that particle
ψ(~x1 , ~x2 ) = Φk (~x1 )Φ−k (~x2 ) (34)
1 encounters its SGz device first. A simple calculation
(0,d,0)
where Φk (~x) represents a plane-wave packet [1] cen- shows that, in the overlap region behind the magnets,
tered at ~x = (0, d, 0) with central wave vector ~k = kŷ. the velocity of particle 1 will be proportional to ŷ (i.e., it
Consider first the case in which the spin part of the ini- will simply continue in a straight horizontal line through
tial state also factorizes, so that the total state can be the overlap region). It will then exit the overlap region
written into one or the other of the two downstream sub-beams,

(0,d,0)

(0,−d,0)
 depending on its initial z-coordinate: if the particle hap-
Ψ(~x1 , ~x2 ) = Φk (~x1 )χ10 Φk (~x2 )χ20 . (35) pens to have started in the upper half of the packet it
will end up deflecting up and being counted as “spin-
We may introduce the so-called “conditional wave func- up”, whereas if it happens to have started in the lower
tion” (CWF) for particle 1 as follows: half of the packet it will end up deflecting down and being
Ψ1 (~x) = Ψ(~x, ~x2 )|~x2 =X~ 2 (36) counted as “spin-down”. Supposing, for notational sim-
plicity, that the two now spatially-separated sub-beams
~ 2 is, of course, the actual position of particle 2.
where X are bent (say, by additional appropriate SG type mag-
The CWF for particle 2 is, obviously, defined in a parallel nets) so that they again propagate in the +y-direction
way. but with displacements ±∆ in the z-direction [21], the
 10

two-particle wave function after particle 1 has passed

through its SGz device will take the form: S S

1 h (0,d′ ,∆) (0,−d′ ,0)

Ψ = √ Φk (~x1 ) Φ−k (~x2 ) χ1+z χ2−z (39)
2
i
(0,d′ ,−∆) (0,−d′ ,0)
− Φk (~x1 ) Φ−k (~x2 ) χ1−z χ2+z .

That is, the overall wave function is a superposition of N N

two terms – one, proportional to χ1+z χ2−z , in which the
particle 1 wave packet has been displaced a distance ∆ Particle 2 Particle 1
in the positive z-direction, and the other, proportional
to χ1−z χ2+z , in which the particle 1 wave packet has been S
displaced a distance ∆ in the negative z-direction.
The actual location X ~ 1 of particle 1 will be either near
z = ∆ or near z = −∆ (again, depending on its random
initial position). It is thus meaningful already at this
stage to speak of the actual outcome of the SGz -based
measurement of the z-spin of particle 1. The crucial point N
is now that the CWF for particle 2 – the thing that will
determine how particle 2 behaves when it subsequently
encounters its SGz device – depends on where particle 1
FIG. 6: Illustration of the two scenarios discussed in the
ended up. If particle 1 went up (that is, if X ~ 1 is now in
text. In the top frame, particle 1 (on the right) encounters
(0,d′ ,∆)
the support of Φk ) then the CWF of particle 2 is its SGz device first; the particle is found to be spin-up (solid
(proportional to) trajectories) or spin-down (dashed trajectories) depending on
the initial z-coordinate of the particle. If particle 1 goes up,
(0,−d′ ,0) the collapse suffered by the CWF of particle 2 (on the left)
Ψ2 (~x) = Φ−k (~x) χ−z (40)
causes it to go down (solid trajectories) regardless of its initial
and its subsequent interaction with an SGz device will, z-coordinate. On the other hand, if particle 1 goes down, the
independent of the exact position X ~ 2 , result in its be- collapse causes particle 2 instead to go up (dashed trajecto-
ries) regardless of its initial z-coordinate. In the lower frame,
ing deflected down. Whereas if instead particle 1 ended
particle 1 is not subjected to any measurement. The result of
up going down (that is, if X~ 2 is now in the support of
measuring the z-spin of particle 2 is then determined by the
(0,d′ ,−∆)
Φk ) then the CWF of particle 2 is instead (pro- initial z-coordinate of particle 2. The difference between the
portional to) two scenarios exemplifies the contextuality of the pilot-wave
theory (since the result of measuring the spin of particle 2
(0,−d′ ,0) depends not only on the initial state but on whether or not
Ψ2 (~x) = Φ−k (~x) χ+z (41)
the spin of particle 1 is measured jointly) and also its non-
and its subsequent interaction with an SGz device will, locality (since the choice of whether or not to measure particle
independent of the exact position X ~ 2 , instead result in 1 could be made at space-like relativistic separation from the
measurement of particle 2).
its being deflected up. That is, the CWF for particle 2
collapses, as a result of the measurement on particle 1, to
a state of definite z-spin. This happens, however, despite
the fact that the two-particle wave function obeys the
(unitary) Schrödinger equation without exception! And of the way. Then, the two-particle wave function remains
the usual quantum mechanical statistics are reproduced: in a state described by Equation (38) until particle 2 ar-
the two possible joint outcomes (particle 1 is spin-up and rives at its SGz device. But then particle 2 will behave in
particle 2 is spin-down, or particle 1 is spin-down and par- just exactly the way we previously described for particle
ticle 2 is spin-up) each occur with 50% probability. Of 1 (when it was the first to encounter an SGz device): its
course, the pilot-wave theory being after all determinis- velocity in the overlap region behind the magnets will be
tic, these probabilities are quite reducible. In particular, proportional to ŷ and the outcome of the spin measure-
which of the two joint outcomes occurs depends on the ment will depend on the initial position (in particular,
random initial position of particle 1 (specifically, its ini- the z-coordinate) of particle 2.
tial z-coordinate). The non-locality is thus clear: even for the same ini-
In order to make the non-local character of the theory tial state – two-particle wave function given by Equation
absolutely clear, it is helpful to now consider an alterna- (38) and, say, both particle positions, X ~ 1 and X~ 2 , in the
tive scenario in which particle 1 is not subjected to any upper halves of their respective packets – the outcome
measurement. For example, perhaps Alice, who is sta- of the measurement on particle 2 depends on what Alice
tioned there next to it, decides (at the last possible sec- chooses to do in the vicinity of particle 1. If Alice subjects
ond before particle 1 arrives) to yank the SGz device out particle 1 to an SGz -based measurement, that measure-
 11

ment will have outcome “spin-up”, the CWF of particle simply recall, again from ordinary QM, that for particles
2 will collapse to a state proportional to χ−z , and the with spin the definitions of both ~jn and ρ involve sum-
subsequent measurement of particle 2’s z-spin will have ming over the spin indices. Thus understood, the pilot-
outcome “spin-down”. On the other hand, if Alice re- wave theory reproduces the usual quantum statistical
moves the SGz device, the measurement of particle 2’s predictions for spin measurements (including sequences
z-spin will instead have the outcome “spin-up”. [23] It of measurements performed on a single particle, and joint
is clear that the non-locality is a form of contextuality measurements performed on pairs of perhaps-entangled
in which the part of the experimental “context” affecting particles) but without any additional postulates or ad hoc
the realized outcome of the experiment is remote. exceptions to the usual dynamical laws. And it does this,
Although we have focused here on the concrete ex- incidentally, in a way that could have been anticipated:
ample in which the z-components of the spins of both spin measurements too (like position measurements and
particles are (perhaps) measured, it should now be clear measurements whose results are registered by the posi-
that, and how, the pilot-wave theory accounts for the em- tion of a pointer) ultimately come down to the position of
pirically observed correlations in the more general sort of a particle (downstream from an appropriate SG device).
case in which arbitrary components of spin are measured. The theory exhibits “contextuality” as illustrated in
One of the particles will encounter its measuring device the two examples discussed: the result of a measurement
first ; the outcome of this first measurement will be de- depends not only on the state of the particle prior to mea-
termined by the random initial position of this measured surement, but also on the measurement’s specific exper-
particle within its wave packet; the completion of this imental implementation. In particular, the outcome of a
first measurement induces a collapse in the distant par- measurement of the z-component of the spin of a parti-
ticle’s CWF; this in turn determines the statistics for a cle can depend on whether the spin is measured using an
subsequent measurement on the distant particle. (Show- SGz or an SG′z device, and/or can depend on which (com-
ing that, in addition, the expected statistical results are muting) observable is also being jointly measured. Con-
reproduced even if one or both of the measuring devices sidered in the abstract, the idea of a contextual hidden
are replaced with alternative SG′n devices is left as an variable theory perhaps sounds contrived and implausi-
exercise for the reader.) ble. Such, at least, has apparently been the thinking
That so much hangs on which measurement happens behind the suggestions that the various “no hidden vari-
first (or more mathematically, that the CWF of each par- ables” theorems (due to von Neumann, Kochen-Specker,
ticle is the two-particle wave function evaluated at the and so on) rule out the possibility of (or even an impor-
actual current position of the other particle, no matter tant class of) hidden variables theories. But the example
how distant) makes it clear that the non-locality of the of the pilot-wave theory shows, quite simply and conclu-
pilot-wave theory will be difficult to reconcile with rel- sively, that they don’t. And furthermore, like it or lump
ativity. This is, however, in principle no different from it, the theory (backed up by Bell’s theorem) shows that
the non-locality implied by the collapse postulate of ordi- there is nothing the least bit contrived or intolerable in
nary QM. (Indeed, the collapse of the pilot-wave theory’s the sort of contextuality that is required to eliminate the
CWFs can and should be understood as a mathematically “unprofessionally vague and ambiguous” aspects of ordi-
precise derivation of the usual textbook approach to mea- nary QM in favor of something clear and mathematically
surement, in a theory where imprecise notions like “mea- precise. [28]
surement” play no fundamental role.) And of course, The idea that there should be something contrived or
the “grossly non-local structure” of the pilot-wave the- intolerable about contextuality undoubtedly arises from
ory and ordinary QM turns out to be “characteristic ... the idea that, if a property really exists, measurement
of any such theory which reproduces exactly the quan- of it should – by definition – simply reveal its value. It
tum mechanical predictions” – as shown in Bell’s other would be hard, actually, to disagree with this sentiment.
landmark paper, of 1964. [24] The key question, though, is precisely whether any such
property exists. As has been discussed in illuminating
detail in Ref. [25], the real lesson to be taken away from
VII. CONCLUSIONS examining the pilot-wave perspective on spin is that so-
called “contextual properties” (like the individual spin
components in the pilot-wave theory) are not properties
Contrary to an apparently widespread belief, the pilot- at all. [26] They simply do not exist and there is noth-
wave theory has no trouble accounting for phenomena in-
ing mysterious about this at all, just as there is nothing
volving spin. It does so, actually, in the most straightfor- mysterious in the fact that the eventual flavor of a loaf of
ward imaginable way: the wave function (a scalar-valued
bread (which depends not just on the ingredients but also
field on the configuration space in the case of spinless on how it is later baked!) is not a pre-existing property
particles) is replaced with the appropriate spinor-valued
of the raw dough:
function, evolving according to the appropriate general-
ization of the Schrödinger equation, just exactly as in “Note that one can completely understand
ordinary QM. The dynamical law for the motion of the what’s going on in [a] Stern-Gerlach experi-
particles, Equation (2), doesn’t change at all – one need ment without invoking any putative property
 12

of the electron such as its actual z-component such thing as a spin angular momentum vector, pointing
of spin that is supposed to be revealed in the in some particular direction, at all.
experiment. For a general initial wave func- And yet – even for Townsend, who is far more care-
tion there is no such property. What is more, ful about this issue than most authors [32] – there is a
the transparency of the [pilot-wave] analysis tendency to occasionally slip into suggesting that there
of this experiment makes it clear that there is is, or at least that (even though there isn’t) it is some-
nothing the least bit remarkable (or for that times helpful to pretend that there is. Thus for example,
matter ‘nonclassical’) about the nonexistence Townsend later explains that “placing [a spin-1/2] parti-
of this property.” [27] cle in a magnetic field in the z direction rotates the spin
of the particle about the z axis as time progresses...” We
As explained by Bell, the appearance to the contrary –
also find statements about how, for example, in the weak
that is, the tacit assumption that there must be some
decay of the muon, “the positron is preferentially emitted
real “z-component of spin” property that the measure-
in a direction opposite to the spin direction of the muon”
ments unveil – seems to arise from the unfortunate and
and even a description of h+x|ψ(t)i and h+y|ψ(t)i – the
inappropriate connotations of the word “measurement”:
amplitudes for a particle in spin state |ψ(t)i to be found,
“the word comes loaded with meaning from respectively, spin-up along the x- and y-directions – as
everyday life, meaning which is entirely inap- “the components of the intrinsic spin in the x − y plane”.
propriate in the quantum context. When it is So perhaps after all particles do have definite spin vec-
said that something is ‘measured’ it is difficult tors? No, Townsend reminds us that this language and
not to think of the result as referring to some the associated physical picture are not to be taken too
pre-existing property of the object in ques- seriously:
tion. This is to disregard Bohr’s insistence
that in quantum phenomena the apparatus “However, we should be careful not to carry
as well as the system is essentially involved. over too completely the classical picture of a
If it were not so, how could we understand, magnetic moment precessing in a magnetic
for example, that ‘measurement’ of a compo- field since in the quantum system the angu-
nent of ‘angular momentum’ – in an arbitrar- lar momentum ... of the particle cannot ac-
ily chosen direction – yields one of a discrete tually be pointing in a specific direction be-
set of values? When one forgets the role of the cause of the uncertainty relations...” [empha-
apparatus, as the word ‘measurement’ makes sis added]
all too likely, one despairs of ordinary logic
It is not my intention here to criticize Townsend’s text,
– hence ‘quantum logic’. When one remem-
which I really do think is wonderful. Surely it would
bers the role of the apparatus, ordinary logic
be made much worse if all the linguistic shortcuts crit-
is just fine.” [29]
icized above were “fixed”. (Just imagine the dreariness
Unfortunately, even some of the people in the best possi- and impenetrability of a textbook which explained, for
ble position to understand this important point – namely, example, that “the positron is preferentially emitted in a
proponents of the pilot-wave theory – have missed it and direction opposite to the direction along which measure-
have thus pointlessly laden the theory with additional ment of the muon’s spin, should such a measurement have
variables corresponding to such actual spin components. been performed prior to its decay, would have given the
[30] One goal of the present paper is thus to present, in highest probability of yielding the outcome ‘spin-up’.”)
a clear and accessible way, the simplest possible pilot- Nevertheless, there really is a sense – highlighted espe-
wave account of spin, in order to rectify misconceptions cially by Townsend’s use of the word “too” in the passage
like that held by the anonymous referee quoted in the just quoted – in which the orthodox view insists on re-
introduction. taining the classical picture (of a little spinning ball of
Finally, note it is by no means only champions of hid- charge with definite spin angular momentum vector) but
den variable theories that are sometimes seduced into simultaneously apologizing for this, by demanding that
thinking of spin components as real properties. Although the picture not be carried over too completely, not be
the official party line of the orthodox quantum theory, as taken too seriously.
expressed for example in standard textbooks, is of course The reason for this schizophrenia, I suspect, is that
that there are no such things, one often finds that the orthodox quantum physicsts are, after all, physicists.
authors, perhaps swept away by all the talk of “measur- They cannot just “shut up and calculate” – not com-
ing” “observables”, are somewhat conflicted about this. pletely. They need some sort of visualizable picture of
Townsend, for example, whose beautiful textbook begins what, physically, the mathematical formalism describes,
with five full chapters on spin, stresses early on that, be- or they simply cannot keep track of what in the world
cause the operators corresponding to distinct spin com- they are talking about. [33] So they retain the classical
ponents fail to commute, “the angular momentum never picture while simultaneously, out of the other sides of
really ‘points’ in any definite direction.” [31] This already their mouths, rejecting it.
implies the usual, orthodox view that there is simply no Perhaps then, at the end of the day, the most impor-
 13

tant thing about the pilot-wave perspective on spin is Acknowledgements: Thanks to George Greenstein for or-
simply that it provides a picture of what might actually ganizing the wonderful series of meetings at which I had
be going on physically in phenomena involving spin – a the opportunity to first try out some of this material on
picture that does not involve any spinning balls of charge, a live audience. Philip Pearle, John Townsend, Roderich
but which is nevertheless completely and absolutely clear Tumulka, and two anonymous referees provided helpful
and precise and which can be taken seriously, without comments on an earlier draft.
apologies or double-speak.

[1] T. Norsen, “The pilot-wave perspective on scattering and idealized: a more precise expression would involve field
tunneling” American Journal of Physics 81(4), pp. 258 components perpindicular to ẑ – which would allow the
(2013) divergence of B ~ to vanish, as it of course should – as
[2] S. Goldstein, D. Dürr, and N. Zanghı́’, “Quantum Equi- well as a large constant field in the z-direction. As it is
librium and the Origin of Absolute Uncertainty” Journal usually explained, this large constant field causes rapid
of Statistical Physics, 67, 843-907 (1992) precession about the z-axis, and thus makes any trans-
[3] Note that here, in discussing the measurement from the verse forces (that would otherwise arise from the trans-
point of view of orthodox QM, the word “particle” is used verse magnetic field gradients) average to zero. It is much
in an extended sense – meaning, for example, “an elec- simpler, though, to simply look the other way in regard
tron.” Of course, according to orthodox QM, an electron to the violation of Maxwell’s equations, and thus avoid
is not literally a particle in the sense of being a point-like needing to talk about any of these subtleties, by sim-
object following a definite trajectory. Whereas, accord- ply pretending that Equation (6) accurately describes the
ing to the pilot-wave theory, a single electron comprises field to which the particles are subjected.
both a quantum wave and a (literal) particle. Readers are [13] If the reflected wave term from Equation (14) is retained,
hereafter assumed to be able to judge, from the context, the detailed particle trajectories are a little more com-
in which sense the word “particle” is used. plicated. During the time period when the incident and
[4] W. Pauli, “Über den Zusammenhang des Abschlusses der reflected packets overlap, the particle velocity in that
Elektronengruppen im Atom mit der Komplexstruktur (y < 0) overlap region is slightly oscillatory, with an
der Spektren” Z. Physik 31, 765-783 (1925). average velocity – in precisely the sense explained and
[5] See, for example, Lubos Motl’s comments used in the earlier paper – that is slightly less than ~k/m
here: http://motls.blogspot.com/2007/12/ by an amount proportional to (κz)2 . Thus, particles very
david-bohm-born-90-years-ago.html near the trailing edge of the incident packet will in fact
[6] S. Kochen and E. Specker, “The problem of hidden vari- be overtaken by the trailing edge of the packet before
ables in quantum mechanics” Journal of Mathematics reaching y = 0 and will hence be swept away, back out
and Mechanics 17, pp. 5987 (1967) to the left, by the reflected packet. In short, particles
[7] N.D. Mermin, “Hidden variables and the two theorems which happen to begin in a small (parabolic) slice near
of John Bell”, Reviews of Modern Physics, Vol. 65, No. the trailing edge of the incident packet will not behave as
3, July 1993, pp. 803-815. described in the main text, but will instead reflect. The
[8] D. Bohm, R. Schiler, and J. Tiomno, Suppl. Nuovo Ci- behavior of the particles which do transmit through to
mento 1 (1955) 48; see also T. Takabayasi, Prog. Theor. the y > 0 region, however, is (in the applicable κw → 0
Phys 14 (1955) 283. limit) unaffected. So we will continue to suppress discus-
[9] See, e.g., C. Dewdney, P.R. Holland, and A. Kyprianidis, sion at this level of detail in the main text.
“What happens in a spin measurement”, Physics Letters [14] D. Bohm and B. Hiley, The Undivided Universe (Rout-
A 119 (1986) 259-267; C. Dewdney, P.R. Holland, and ledge, London and New York, 1993), pp. 73-78.
A. Kyprianidis, “A causal account of non-local Einstein- [15] Actually there is one subtle point that deserves to be
Podolsky-Rosen spin correlations”, J. Phys. A 20 (1987) addressed. In analyzing the measurement of z-spin, we
4717-4732. have assumed that the incident wave packet has an es-
[10] J.S. Bell, “Introduction to the hidden variable question,” sentially rectangular cross-sectional profile, with (as men-
Proceedings of the International School of Physics ‘En- tioned explicitly) some width w along the z-direction and
rico Fermi’, course IL: Foundations of Quantum Me- (as only really assumed tacitly) some width w′ along the
chanics, New York, Academic (1971) 171-81; a some- x-direction. This makes the problem “effectively 2D” in
what more detailed elaboration can be found in J.S. the sense that (for the width-w′ range of x-values where
Bell, “Quantum mechanics for cosmologists” in Quan- the wave function has support and where the particle
tum Gravity 2, C. Isham, R. Penrose, and D. Sciama, might therefore be) nothing depends on x. If, however,
eds., Oxford, Clarendon Press (1981) pp. 611-637; both the incident wave function is the same but the SG ap-
essays are reprinted in J.S. Bell, Speakable and Unspeak- paratus is rotated through some arbitrary angle (not an
able in Quantum Mechanics, 2nd ed., Cambridge Univer- integer multiple of 90◦ ), this nice symmetry is spoiled.
sity Press (2004) Everything still works as claimed; it’s just that one has to
[11] W. Gerlach and O. Stern, “Der experimentelle Nachweis think a little bit (about, e.g., slicing the 3D wave packet
der Richtungsquantelung im Magnetfeld” Zeitschrift für up into a stack of planes perpindicular to n̂) to convince
Physik 9 pp. 349-352 (1922) oneself of this. Alternatively, one can recognize that the
[12] Note that this expression for the field is simplified and statistics of the measurement outcomes will not depend
 14

on the exact cross-sectional profile of the incident packet ing from a hidden-variable account of quantum mechan-
and hence imagine that the cross-sectional profile of the ics need not very much resemble the traditional classi-
incident packet is rotated in tandem with the SG device cal picture that the researcher may, secretly, have been
so that the analysis remains transparent. keeping in mind. The electron need not turn out to be a
[16] Of course, the mere fact that the two sub-beams are small spinning yellow sphere.” (J.S. Bell, “Introduction
spatially-separated is not sufficient to prove that the to the hidden-variable question” ...) Gestures in the di-
empty packet is forevermore dynamically irrelevant to the rection of this same point were made even earlier by Ren-
motion of the particle. The “empty” spin-down sub-beam ninger (“Zum Wellen-Korpuskel-Dualismus” Zeitschrift
could be made to become, again, dynamically relevant – für Physik 136, 251-261, 1953; “On Wave-Particle Du-
for example, by arranging, by use (say) of appropriate ality”, W. De Baere, trans., quant-ph/0504043) in 1953,
magnetic fields, to deflect the spin-down beam back up and even earlier by Lorentz, who discussed in 1922 a
toward the spin-up beam, at which point interference ef- pilot-wave theory of photons, previously suggested by
fects (including perhaps
the re-constitution of a guiding Einstein (Problems of Modern Physics, 1922).
wave proportional to 10 !) could occur. But if, as sug- [27] D. Dürr, S. Goldstein, and N. Zanghı́, “Quantum Equi-
gested in Figure 4, the spin-down sub-beam is blocked by librium and the Role of Operators as Observables in
a brick wall, then the usual process of decoherence will Quantum Theory” Journal of Statistical Physics 116,
render it impossible, at least for all practical purposes, 959-1055 (2004). Reprinted as Chapter 3 of Quantum
to arrange the appropriate overlapping of the two com- Physics Without Quantum Philosophy, Springer-Verlag,
ponents of the wave function, and treating the wave as Berlin/Heidelberg, 2013.
having “effectively collapsed” is entirely warranted. [28] J.S. Bell, “Beables for quantum field theory”, 1984;
[17] This particular illustration of the “contextuality” of spin reprinted in Bell (2004) op cit.
(for the pilot-wave theory) is due, in essence, to David [29] J.S. Bell, “Against ‘Measurement’,” 1989; reprinted in
Z. Albert: see Quantum Mechanics and Experience, Har- Bell (2004) op cit.
vard University Press, 1994, pp. 153-5. [30] See, for example: P. Holland, The Quantum Theory of
[18] J.S. Bell, “On the Problem of Hidden Variables in Quan- Motion, Cambridge University Press, 1993, Chapter 9;
tum Mechanics” Reviews of Modern Physics 38 pp. 447- C. Dewdney, P. Holland, and A. Kyprianidis, “A causal
452 (1966), reprinted in Bell (2004), op cit. account of non-local Einstein-Podolsky-Rosen spin corre-
[19] N. Bohr, “Discussions with Einstein on Epistemolog- lations” J. Phys. A: Math. Gen. 20 4717-4732 (1987); D.
ical Problems in Atomic Physics” in Albert Einstein: Bohm and B. Hiley, The Undivided Universe, Routledge,
Philosopher Scientist, P. A. Schillp, ed., Library of Living London, 1993, Chapter 10; C. dewdney, “Constraints
Philosophers, Evanston, Illinois, 1949. on quantum hidden-variables and the Bohm theory”, J.
[20] A. Shimony, “Contextual hidden variables theories and Phys. A 25 (1992) 3615-3626.
Bell’s inequalities” Brit. J. Phil. Sci., 35(1), 25-45 (1984) [31] J.S. Townsend, A Modern Approach to Quantum Me-
[21] We assume here that ∆ > w so that the two sub-beams chanics, 2nd ed., University Science Books, Mill Valley,
downstream of the SG device do not overlap spatially. California, 2012. All quotes are from chapters 3 and 4.
[22] T. Norsen, “The Theory of (Exclusively) Local Beables” [32] In his E&M text, for example, David Griffiths writes: “all
Foundations of Physics 40 (2010) pp. 1858. magnetic fields are due to electric charges in motion, and
[23] It is perhaps worth noting explicitly that the non-locality in fact, if you could examine a piece of magnetic material
cannot be used for signaling: if a large ensemble of pairs on an atomic scale you would find tiny currents: elec-
are prepared in the state (38), and Alice chooses at the trons orbiting around nuclei and electrons spinning on
last minute whether to subject all of the particles on her their axes.” And later: “Since every electron constitutes
side to an SGz measurement, or some other spin measure- a magnetic dipole (picture it, if you wish, as a tiny spin-
ment, or no measurement, Bob will see the same 50/50 ning sphere of charge), you might expect paramagnetism
statistics on his side regardless of her choice. to be a universal phenomenon.” (Introduction to Electro-
[24] J.S. Bell, “On the Einstein Podolsky Rosen Paradox” dynamics, 2nd ed., Prentice Hall, Englewood Cliffs, New
Physics Vol. 1, No. 3, pp. 195-200 (1964), reprinted Jersey, 1989) Interestingly, Griffiths is more carefully or-
in Bell (2004) op cit. For a detailed analysis see the thodox about spin in his Quantum text.
article by S. Goldstein, T. Norsen, D. Tausk, and [33] In this respect it is somewhat telling that the worst of-
N. Zanghı́ at http://www.scholarpedia.org/article/ fenses against the official orthodox view tend to occur in
Bell’s_theorem. figures and illustrations. Almost all books for example in-
[25] M. Daumer, D. Dürr, S. Goldstein, and N. Zanghı́, clude diagrams depicting the precession, induced by the
“Naive Realism about Operators” Erkenntnis 45, 379- presence of a magnetic field, of a particle’s spin vector
397 (1996), about some axis. Townsend’s text avoids that particular
[26] The authors of Ref. [25] describe themselves as following misleading suggestion of a classical picture. But his front
Bell in concluding that, for the pilot-wave theory, “spin cover art – depicting a Stern-Gerlach spin measurement
is not real”. Bell, for example, writes: “We have here much like our Figure 1 – includes little circles with arrows
a picture in which although the wave has two compo- (pointing respectively up and down) in the downstream
nents, the particle has only position.... The particle does sub-beams. The circles are even yellow, inviting us to
not ‘spin’, although the experimental phenomena associ- recall Bell’s warning, quoted earlier in note [26].
ated with spin are reproduced. Thus the picture result-