arXiv:0903.

2601v1 [quant-ph] 15 Mar 2009

Bohmian Mechanics

Detlef Dürr1, Sheldon Goldstein2,

Roderich Tumulka3, and Nino Zanghı̀4

January 6, 2008

1 Mathematisches Institut, Ludwig-Maximilians-Universität, Theresienstraße 39, 80333

München, Germany. E-mail: [email protected]
2 Departments of Mathematics and Physics, Rutgers University, Hill Center, 110 Frelinghuy-

sen Road, Piscataway, NJ 08854-8019, USA. E-mail: [email protected]

3 Department of Mathematics, Rutgers University, Hill Center, 110 Frelinghuysen Road,

Piscataway, NJ 08854-8019, USA. E-mail: [email protected]

4 Dipartimento di Fisica and INFN sezione di Genova, Via Dodecaneso 33, 16146 Genova,

Italy. E-mail: [email protected]

Bohmian mechanics is a theory about point particles moving along trajectories. It
has the property that in a world governed by Bohmian mechanics, observers see the
same statistics for experimental results as predicted by quantum mechanics. Bohmian
mechanics thus provides an explanation of quantum mechanics. Moreover, the Bohmian
trajectories are defined in a non-conspiratorial way by a few simple laws.

Overview. Bohmian mechanics is a version of quantum mechanics for nonrelativistic

particles in which the word “particle” is to be understood literally: In Bohmian mechan-
ics quantum particles have positions, always, and follow trajectories. These trajectories
differ, however, from the classical Newtonian trajectories. Indeed, the law of motion, see
eq. (1) below, involves a wave function. As a consequence, the role of the wave function
in Bohmian mechanics is to tell the matter how to move.
Bohmian mechanics constitutes a quantum theory without observers, i.e., a theory
that is formulated not in terms of what observers see but in terms of objective events,
regardless of whether or not they are observed. Bohmian mechanics provides a con-
sistent resolution of all paradoxes of quantum mechanics, in particular of the so-called
measurement problem. In particular, the collapse of the wave function can be derived
from Bohmian mechanics.
Bohmian mechanics is sometimes called a hidden variables theory because it involves
variables besides the wave function. However, there is a danger of confusion here because
the term “hidden variables theory” is often used to convey the idea that every “quantum
measurement” of an “observable” reveals a pre-existing value of that observable, which
is not the case in Bohmian mechanics.
Bohmian mechanics is deterministic. But the motivation behind Bohmian mechanics
is not to obtain a deterministic theory, but rather to obtain a coherent account of
the nature of physical reality. In this regard, we note that some variants of Bohmian
mechanics, developed by its proponents, are stochastic rather than deterministic, for
example Bell’s proposal for lattice quantum field theory [4].
Historically, the “Bohmian” law of motion, see eq. (1) below, was first proposed by
de Broglie [6]. However, Bohm [5] was the first to recognize that this theory explains
all of the phenomena of (non-relativistic) quantum mechanics.

Defining Equations. Bohmian mechanics is a non-relativistic theory governing the

behavior of a system of N point particles moving in physical space R3 along trajectories.
Let Qi (t) ∈ R3 denote the
position of the i-th particle of the system at time t, and
Q(t) = Q1 (t), . . . , QN (t) ∈ R3N its configuration.
The trajectories are governed by Bohm’s law of motion [5, 2]

dQi ~ Ψ ∗ ∇i Ψ t
Im t ∗


= Q(t) , (1)
dt mi Ψt Ψt

where mi is the mass of particle i, Im denotes the imaginary part, Ψt : R3N → Ck (i.e., a
function of the configuration with k complex components) is the wave function at time
t, Φ∗ Ψ is the scalar product in Ck , and ∇i is the gradient relative to the 3 coordinates

1
of particle i. (In case k = 1, i.e., for complex-valued wave functions, a factor Ψ∗t cancels
on the right hand side of (1).)
The wave function evolves according to the Schrödinger equation
N
X ~2
∂Ψt
i~ =− ∇2i Ψt + V Ψt , (2)
∂t i=1
2mi

where V : R3N → R is the potential function. (The potential, while often assumed to be
real-valued, may take values in the space of self-adjoint complex k × k matrices instead
of R.) The wave function is postulated to belong to the Hilbert space H = L2 (R3N , Ck )
of square-integrable functions (and to be sufficiently smooth).

Deterministic Evolution. Since the Schrödinger equation does not involve the particle
positions Qi (t), it can be solved first and determines the wave function Ψt for every
time t once an initial wave function Ψt0 is specified for any time t0 that we choose to
regard as the initial time. Next note that the right hand side of (1) consists of the 3
components corresponding to particle i out of the 3N components of a vector field v Ψt
on configuration space R3N . As a consequence, the equations (1) for all i = 1, . . . , N
can be summarized by
dQ
= v Ψt Q(t) .


(3)
dt
Regarding Ψt as known, this is a (time-dependent) ordinary differential equation (ODE)
of first order, and as such determines the entire history t 7→ Q(t) once an initial con-
figuration Q(t0 ) is specified. That is why Bohmian mechanics is deterministic: once
Q(t0 ) and Ψt0 are specified, the entire history is fixed by the equations (1) and (2). This
fact also implies that the pair (Q(t0 ), Ψt0 ) can be regarded as the state of the Bohmian
particle system at time t0 . Since the choice of t0 is arbitrary, the state at any time t is
the pair (Q(t), Ψt ), and the phase space of Bohmian mechanics is R3N × H .

System or Universe. The equations of Bohmian mechanics could be applied to a

familiar system (e.g., an atom) or to the universe as a whole. Of course, one cannot
expect that the equations hold for every system, for example for systems that interact
with their environments. So let us begin with the system for which the equations
are primarily intended: the universe. In this setting, N is the number of particles in
the universe, and Ψt is the wave function of the universe. To consider such a wave
function is unusual; after all, the quantum formalism never refers to a wave function
of the universe; the quantum formalism, providing the probabilities for the results of
observations performed on a system by an external observer, involves the wave function
of that system and not of the entire universe. In the context of Bohmian mechanics,
however, the wave function of the universe is not at all a meaningless concept, as it
influences the motion of the particles according to (1).
When (1) and (2) hold for the universe, it follows that equations of the same type
(but with smaller N) hold for certain subsystems. (We shall assume here for simplicity
that k = 1, i.e., that we are dealing with spinless particles.) Consider a subsystem of

2
the universe with configuration X (the x-system), so that the configuration Q of the
universe is of the form Q = (X, Y ) with Y the configuration of the environment of the
x-system. Then a natural notion of the wave function of the x-system is provided by its
conditional wave function
ψ(x) = Ψ(x, Y ) , (4)
where Ψ(q) = Ψ(x, y) is the wave function of the universe. It is easy to see that the
x-system obeys (3) (with Q = X and Ψ = ψ).
Moreover, if the x-system is suitably decoupled from its environment, (2) will hold
as well. For example, this is the case when there is no interaction between the x-system
and its environment, and the wave function of the universe is of the form

Ψ(x, y) = ψ(x) ϕ(y) + Φ(x, y) (5)

with ϕ and Φ having macroscopically disjoint y-supports (so that they will never again
overlap appreciably), and with Y lying in the support of ϕ. Such a situation often arises
after a “quantum measurement.”

Equivariance. If the initial configuration Q(t0 ) is chosen at random with probability

density |Ψt0 |2 then the configuration Q(t) at any other time t is random with proba-
bility density |Ψt |2 . (Whenever speaking of probabilities, we assume Rthat Ψ has been
normalized, by multiplication by a suitable constant, so that hΨ|Ψi = |Ψ(q)|2dq = 1.)
This fact, known as equivariance, follows from the continuity equation
∂ρ
= −∇ · (ρ v) (6)
∂t
for ρ = |Ψ|2 and with the Bohmian velocity vector field v = v Ψ as in (3). The continuity
equation (6) is in turn a consequence of the Schrödinger equation; it is usually written
(in standard quantum mechanics) in terms of the quantum probability current J = ρ v.

Identical Particles. Bohmian mechanics can be formulated for identical particles,

despite a fact that could be felt to contradict their indistinguishability, namely that
the particle trajectories in R3 determine “who is who” at different times, i.e., select a
one-to-one association between the N points at any time t1 and the N points at another
time t2 . Taking the notion of a particle seriously, as one should in Bohmian mechanics,
one recognizes that the configuration space for N identical particles is best regarded
as the manifold of all sets of N points in physical space R3 . This manifold has non-
trivial topological properties, as its fundamental (homotopy) group is isomorphic to the
group of permutations of N objects. On such manifolds there arise several versions
of Bohmian mechanics corresponding to the different 1-dimensional representations of
the fundamental group; for the permutation group, there are two such representations,
corresponding to bosons (with symmetric wave functions on the covering space R3N )
and fermions (with anti-symmetric wave functions). Thus, Bohmian mechanics lends
support to the modern view that the symmetrization postulate emerges as a topological
effect, due to the non-trivial topology of configuration space.

3
Quantum Equilibrium Hypothesis. This is the assertion that whenever a sys-
tem has wave function ψ then its configuration is (or can be taken to be) random with
probability distribution |ψ|2 . Equivariance implies that this hypothesis is consistent
with the time evolution of isolated systems, and it is not hard to show that it is also
consistent with the time evolution if the system is not isolated, provided we take ψ to
mean the conditional wave function. An important consequence of the quantum equilib-
rium hypothesis is the empirical equivalence between Bohmian mechanics and quantum
mechanics: For every conceivable experiment, whenever quantum mechanics makes an
unambiguous prediction, Bohmian mechanics makes exactly the same prediction. Thus,
the two cannot be tested against each other.

Typicality. The quantum equilibrium hypothesis follows from typicality: As shown

in [7] using the law of large numbers, results of experiments are as predicted by the
quantum equilibrium hypothesis for typical initial configurations Q(t0 ) of the universe
relative to the |Ψt0 |2 distribution, i.e., for the overwhelming majority, counted using the
|Ψt0 |2 distribution, of the initial configurations.

Operators. Given that it makes the same predictions as quantum mechanics, what
is the status in Bohmian mechanics of the non-commuting operators of the quantum
formalism (the self-adjoint “observables”), with which the predictions of quantum me-
chanics seem exclusively concerned? The answer is that operators do in fact arise nat-
urally in Bohmian mechanics, but with a different meaning than the one attributed to
them in orthodox quantum mechanics (which regards them as more or less the same
thing as their classical counterparts: as “observables” that can be “measured”). In-
stead, operators in Bohmian mechanics are mathematical tools encoding statistics. Let
us explain.
The statistics of the random outcome Z of an experiment in a world governed by
Bohmian mechanics on a system with wave function ψ can be shown [8] always to be of
the form (in Dirac notation)

Prob(Z = α) = hψ|E(α)|ψi , (7)

where E(α) is a suitable positive operator. (Together, the E(α) form a positive-operator-
valued measure, or POVM.) In relevant cases, E(α) is a family of projection operators
which are mutually orthogonal (a projection-valued measure, or PVM), and thus corre-
spond to the one self-adjoint operator
X
A= α E(α) , (8)
α

which, by the spectral theorem, contains precisely the same information as the PVM
E(α). Thus, operators encode the functional dependence of the outcome statistics on
the system’s wave function ψ. With this understanding, which is opposite to thinking
of operators as representing quantities whose values can be “measured,” it is no longer
surprising that one cannot associate actual values with all “observables” in a consistent

4
way. With this understanding, contextuality is not surprising either, since it no longer
means that the same quantity can choose different values depending on what happens
to another system, but rather that, unspectacularly, different experiments can have the
same statistics.

Collapse of the Wave Function. Here is an analysis, for Bohmian mechanics, of an

“ideal measurement” of a quantum observable, given by a self-adjoint operator A on the
Hilbert space of the relevant system. For simplicity we assume that A has pure point
spectrum with non-degenerate eigenvalues α, corresponding to (8) for E(α) = |ψα ihψα |
with normalized eigenstates ψα (x) = |A = αi. The experiment is implemented by hav-
ing the system interact with an apparatus in a suitable way. To avoid unimportant
complications, we shall assume that the relevant “universe” for the problem at hand
consists entirely of the system, with configuration X, and the apparatus, with configu-
ration Y . The measurement begins, say, at time 0, with the initial (“ready”) state of
the apparatus given by a wave function ϕ0 (y), and ends at time t. The interaction is
such that when the state of the system is initially ψα it produces a normalized apparatus
state ϕα (y), that registers that the value found for A is α without having affected the
state of the system,
t
ψα (x)ϕ0 (y) → ψα (x)ϕα (y). (9)
t
Here → indicates the unitary evolution induced by the interaction. If the measurement
is to provide useful information, the apparatus states must be noticeably different, cor-
responding, say, to a pointer on the apparatus pointing in different directions. We thus
assume that the ϕα have disjoint supports in the configuration space for the apparatus,

supp(ϕα ) ∩ supp(ϕβ ) = ∅, α 6= β. (10)

Now suppose that the system is initially, not in an eigenstate of A, but in a general
state, given by a superposition
X
ψ(x) = cα ψα (x). (11)
α

We then have, by (9) and the linearity of the unitary evolution, that
t
X
Ψ0 (x, y) = ψ(x)ϕ0 (y) → Ψt (x, y) = cα ψα (x)ϕα (y), (12)
α

so that the final wave function Ψt of system and apparatus is itself a superposition. The
fact that the pointer ends up pointing in a definite direction, even a random one, is not
discernible in this final wave function. Insofar as orthodox quantum theory is concerned,
we have arrived at the measurement problem.
However, insofar as Bohmian mechanics is concerned, we have no such problem,
because in Bohmian mechanics particles always have positions and pointers, which are
made of particles, always point—in a direction determined by the final configuration Yt
of the apparatus. Moreover, in Bohmian mechanics we find that the state of the system

5
is transformed in exactly the manner prescribed by textbook quantum theory, as the
final wave function of the system, i.e., its conditional wave function at time t, see (4), is
X
ψt (x) = Ψt (x, Yt ) = cα ψα (x)ϕα (Yt ) = cβ ψβ (x)ϕβ (Yt ) = N ψβ (x) (13)
α

when Yt ∈ supp(ϕβ ), i.e., when the value β is registered. (Here N is a constant that
depends upon Y but not on x. According to (13) the wave function of the system at
time t, when normalized, is ψβ .) The probability for this event is, by the quantum
equilibrium hypothesis,
Z Z
dx dy |Ψt (x, y)|2 = |cβ |2 . (14)
supp(ϕβ )

The upshot of the analysis is this: It is a consequence of Bohmian mechanics that in

the course of an ideal measurement of A the (normalized) wave  function
2 of the system
2
is transformed from ψ (11) to ψβ with probability |cβ | = hψβ |ψi . That is how the
 
projection postulate arises from Bohmian mechanics. (The fact that the contributions
with α 6= β will never again overlap with what evolves from ψβ (x)ϕβ (y), and thus will
not influence the future motion of the particles, is the reason why they can be ignored
from time t onwards, or “collapsed away,” without consequences for the trajectories of
the particles.)

The Double Slit Experiment. In Bohmian mechanics, “wave–particle duality” can be

taken literally: there is a wave (ψ) and there are particles. Accordingly, in a double slit
experiment the wave passes through both slits, whereas the particle passes only through
one slit. Since the motion of the particle depends on the wave, it matters whether or not
the other slit is open. The possible trajectories, when both slits are open, are depicted
in Fig. 1; by virtue of the quantum equilibrium hypothesis, the actual trajectory will be
random with the appropriate |ψ|2 distribution. Thus, the place of the particle’s arrival
at a screen on the right will have a probability distribution featuring interference fringes.
As John Bell commented [10, p. 191]: “This idea seems to me so natural and simple [...]
that it is a great mystery to me that it was so generally ignored.”

Spin. One may easily get the impression that spin cannot be explained in a realist
way, given its “non-classical two-valuedness.” But actually it can be incorporated into
Bohmian mechanics very easily, and Bell discovered how [2]: Do not assume that there
is an “actual value” associated with the spin observable σ̂z in the z (or any other)
direction! Instead, take the equation of motion (1) seriously, with Ck the spin space,
i.e., k = (2s + 1)N for N spin-s particles. (In particular, it is useful here to regard the
wave function ψt for, say, a single spin- 12 particle not as a function ψt : R3 ×{−1, 1} → C
of a continuous (position) variable and a discrete (spin) variable, but rather as a spinor-
valued function of position, ψt : R3 → C2 .)
As a consequence of (1), the motion of a particle with spin is influenced by both the
“spin-up” and the “spin-down” component of the wave function. While the particle has

6
Figure 1: Possible Bohmian trajectories in the double-slit experiment (from C. Philip-
pidis, C. Dewdney and B.J. Hiley, Il Nuovo Cimento 52, 15 (1979))

an actual position (and a wave function) but no additional actual spin degrees of freedom,
these are sufficient to completely account for all quantum phenomena associated with
spin.

Quantum Field Theory and Relativity. Bohmian mechanics does not account for
phenomena such as particle creation and annihilation characteristic of quantum field
theory. This is not an objection to Bohmian mechanics but merely a recognition that
quantum field theory explains a great deal more than does nonrelativistic quantum
mechanics, whether in orthodox or Bohmian form. There are extensions of Bohmian
mechanics to general quantum field theories based on a particle ontology, as well as
other approaches. Moreover, like nonrelativistic quantum theory, Bohmian mechanics
is incompatible with special relativity, a central principle of physics: it is not Lorentz
invariant. Nor can Bohmian mechanics easily be modified to become Lorentz invariant.
For an overview of recent proposals aimed at finding a Lorentz invariant extension of
Bohmian mechanics, see [13].

Nonlocality. In Bohmian mechanics the motion of a particle may depend on the posi-
tions of distant particles, at spacelike separation. This is an instance of nonlocality. It is
worth noting that this dependence is of a kind that does not allow superluminal commu-
nication. Orthodox quantum mechanics features nonlocality as well, associated with the
instantaneous collapse of the wave function for all particles, even distant ones. In 1964,
John Bell asked whether nonlocality could be avoided by any version of quantum me-
chanics, and his celebrated (but often misunderstood) argument [3, 10], involving Bell’s
inequality, proves that the answer is no. His argument shows that certain correlations
predicted by quantum mechanics (and Bohmian mechanics) and confirmed in experi-
ment [1] cannot be explained in a local way, i.e., without allowing influences travelling
faster than light. Thus, nonlocality is a feature of our world.

7
Literature

Primary

[1] A. Aspect, J. Dalibard, G. Roger: Experimental Test of Bell’s Inequalities using

Time-Varying Analyzers. Physical Review Letters 49, 1804–1807 (1982)

[2] J. S. Bell: On the problem of hidden variables in quantum mechanics. Reviews of

Modern Physics 38, 447–452 (1966); reprinted as chapter 1 of [10].

[3] J. S. Bell: On the Einstein-Podolsky-Rosen Paradox. Physics, 1, 195–200 (1964);

reprinted as chapter 2 of [10].

[4] J. S. Bell: Quantum field theory without observers. Physics Reports 137, 49–54
(1986); reprinted under the title “Beables for quantum field theory” as chapter 19
of [10].

[5] D. Bohm: A Suggested Interpretation of the Quantum Theory in Terms of “Hid-

den” Variables, I and II. Physical Review 85, 166–193 (1952)

[6] L. de Broglie: In Electrons et Photons: Rapports et Discussions du Cinquième

Conseil de Physique tenu à Bruxelles du 24 au 29 Octobre 1927 sous les Auspices
de l’Institut International de Physique Solvay, (Gauthier-Villars, Paris 1928); En-
glish translation in [9].

[7] D. Dürr, S. Goldstein, N. Zanghı̀: Quantum Equilibrium and the Origin of Abso-
lute Uncertainty. Journal of Statistical Physics 67, 843–907 (1992)

[8] D. Dürr, S. Goldstein, N. Zanghı̀: Quantum Equilibrium and the Role of Operators
as Observables in Quantum Theory. Journal of Statistical Physics 116, 959–1055
(2004)

Secondary

[9] G. Bacciagaluppi, A. Valentini: Quantum Theory at the Crossroads (Cambridge

University Press 2009)

[10] J. S. Bell: Speakable and unspeakable in quantum mechanics (Cambridge Univer-

sity Press 1987)

8
[11] D. Bohm, B.J. Hiley: The Undivided Universe (Routledge, London 1993)

[12] S. Goldstein: Bohmian Mechanics. In E. N. Zalta (ed.), Stanford Encyclopedia of

Philosophy, published online by Stanford University (2001) at
http://plato.stanford.edu/entries/qm-bohm/

[13] R. Tumulka: The ‘Unromantic Pictures’ of Quantum Theory. Journal of Physics

A: Mathematical and Theoretical 40, 3245–3273 (2007)