Bohmian Mechanics (the de Broglie–Bohm Pilot-Wave Theory)

Bohmian mechanics is the most fully worked-out deterministic completion of quantum theory. It takes the interpretations survey's one-paragraph sketch of "hidden variables" and turns it into a precise, complete dynamical theory: point particles with definite positions at all times, moving along continuous trajectories under a velocity law dictated by the wavefunction, which itself evolves by the ordinary Schrödinger equation. Nothing is added to the predictions of standard quantum mechanics — every experimental result is reproduced exactly — yet the measurement problem simply does not arise, because measurements have outcomes in virtue of where the particles actually are.

The theory was proposed by Louis de Broglie in 1927, abandoned after the Solvay conference, rediscovered and completed by David Bohm in 1952, and given its modern mathematically careful form (equivariance, quantum equilibrium, typicality) by John Bell and by Dürr, Goldstein, and Zanghì from the 1980s onward. This page develops the technical formulation in full and then turns to the philosophical questions it raises.

Part I — Technical Formulation

1. The State and the Two Laws

Setup and notation. Fix $N$ particles with masses $m_1,\dots ,m_N>0$. A point of configuration space is written

$$\mathbf{q}=({\mathbf{q}}_1,\dots ,{\mathbf{q}}_N)\in {\mathbb{R}}^{3N},{\mathbf{q}}_k\in {\mathbb{R}}^3,$$

with ${\mathbf{q}}_k$ the position coordinate of the $k$-th particle. Let ${\nabla }_k:=\partial /\partial {\mathbf{q}}_k$ be the gradient in the three coordinates of particle $k$, and ${\nabla }_k^2:={\Delta }_k$ the corresponding Laplacian. The wavefunction is a map

$$\psi :{\mathbb{R}}^{3N}\times \mathbb{R}\to \mathbb{C},\psi (\cdot ,t)\in \mathcal{H}:=L^2({\mathbb{R}}^{3N};\mathbb{C}),\Vert \psi (\cdot ,t)\Vert =1,$$

i.e. a normalized element of the same separable Hilbert space $\mathcal{H}$ as in standard QM (inner product linear in the second argument). The Hamiltonian

$$\hat{H}:=-\sum _{k=1}^N\frac{{\hslash }^2}{2m_k}{\nabla }_k^2+V(\mathbf{q})$$

is the usual operator, taken self-adjoint on its natural domain, with $V:{\mathbb{R}}^{3N}\to \mathbb{R}$ a real potential for which $\hat{H}$ is (essentially) self-adjoint. Spin and magnetic vector potentials are suppressed here and reinstated in §8. The theory is fixed by three postulates.

Postulate B1 (State). The complete physical state at time $t$ is the ordered pair

$$(\psi (\cdot ,t),\mathbf{Q}(t)),\mathbf{Q}(t)=({\mathbf{Q}}_1(t),\dots ,{\mathbf{Q}}_N(t))\in {\mathbb{R}}^{3N},$$

consisting of the wavefunction $\psi$ and the actual configuration $\mathbf{Q}(t)$ — the real positions ${\mathbf{Q}}_k(t)\in {\mathbb{R}}^3$ of the particles. These positions are the theory's "hidden variables," though positions are the least hidden thing in it: they are precisely what we directly observe.

Postulate B2 (Wavefunction dynamics). The wavefunction evolves by the time-dependent Schrödinger equation, unchanged and without collapse:

$$i\hslash {\partial }_t\psi =\hat{H}\psi =\left(-\sum _{k=1}^N\frac{{\hslash }^2}{2m_k}{\nabla }_k^2+V\right)\psi ,\psi (\cdot ,t)=e^{-i\hat{H}t/\hslash }\psi (\cdot ,0).\text{\hspace{1em}(B2)}$$

This is identical to standard unitary evolution; nothing about the ordinary quantum dynamics of $\psi$ is altered.

Postulate B3 (Guidance). The actual configuration obeys the first-order guidance equation: the velocity of particle $k$ is $\hslash /m_k$ times the imaginary part of the logarithmic derivative of $\psi$, evaluated at the actual configuration,

$$\boxed{{\dot{\mathbf{Q}}}_k(t):={\mathbf{v}}_k^{\psi }(\mathbf{Q}(t),t)=\frac{\hslash }{m_k}\mathrm{Im}\frac{{\nabla }_k\psi }{\psi }{|}_{\mathbf{q}=\mathbf{Q}(t)}.}\text{\hspace{1em}(B3)}$$

Postulates B1–B3 are the entire dynamics; there is no separate collapse law and no primitive notion of "measurement."

Equivalent forms of the velocity. On the open set $\{\mathbf{q}:\psi (\mathbf{q},t)\ne 0\}$, multiplying numerator and denominator by ${\psi }^{\ast }$ gives

$${\mathbf{v}}_k^{\psi }=\frac{\hslash }{m_k}\frac{\mathrm{Im}({\psi }^{\ast }{\nabla }_k\psi )}{|\psi {|}^2}=\frac{{\mathbf{j}}_k}{\rho },{\mathbf{j}}_k:=\frac{\hslash }{m_k}\mathrm{Im}({\psi }^{\ast }{\nabla }_k\psi ),\rho :=|\psi {|}^2,$$

where ${\mathbf{j}}_k$ is exactly the $k$-th probability current and $\rho$ the position density of ordinary QM. Thus B3 promotes the standard kinematic identity $\mathbf{j}=\rho \mathbf{v}$ to a law of motion: the particles are carried along by the quantum probability current.

Nodes and well-posedness. The velocity field ${\mathbf{v}}_k^{\psi }$ is smooth wherever $\psi \ne 0$ but is singular on the nodal set $\{\psi =0\}$, which has Lebesgue measure zero. One can prove (Berndl–Dürr–Goldstein–Peruzzi–Zanghì 1995; Teufel–Tumulka 2005) that for $|\psi {|}^2$-almost-every initial configuration $\mathbf{Q}(0)$ the guidance equation (B3) has a unique global solution that neither reaches a node nor escapes to infinity in finite time. So the flow is well-defined for almost every initial condition — exactly the set to which quantum equilibrium (§3) assigns full probability.

Character of the theory. The dynamics B1–B3 is

• first-order: the state $(\psi ,\mathbf{Q})$ fixes the velocity $\dot{\mathbf{Q}}$ directly, with no independent momentum degree of freedom;
• deterministic: the initial pair $(\psi (\cdot ,0),\mathbf{Q}(0))$ determines $(\psi (\cdot ,t),\mathbf{Q}(t))$ for all $t$; and
• Galilean covariant, with $\psi$ carrying the usual projective representation of the Galilei group.

Everything beyond B1–B3 — the Born rule, operators-as-observables, the uncertainty relations, effective collapse — is derived, not postulated, once one adjoins the quantum equilibrium hypothesis of §3, which is itself justified (§4) by a typicality theorem rather than assumed as an independent axiom.

2. The Polar Decomposition and the Quantum Potential

Bohm's original 1952 route makes the dynamics vivid and exhibits its relation to classical mechanics. All manipulations below are carried out on the open set $\{\psi \ne 0\}$.

Polar (Madelung) variables. Where $\psi \ne 0$, write

$$\psi =R{e}^{iS/\hslash },R:=|\psi |>0,S:=\hslash \arg \psi .$$

Here $R$ is single-valued, while $S$ is real but defined only locally, modulo $2\pi \hslash$ (the phase is a section of a circle bundle over $\{\psi \ne 0\}$). Its gradient ${\nabla }_{k}S$ is nonetheless globally single-valued and smooth, and only ${\nabla }_{k}S$ enters the equations below.

Madelung equations. Substituting $\psi =R{e}^{iS/\hslash }$ into the Schrödinger equation (B2) and separating real and imaginary parts yields two exact real equations on $\{\psi \ne 0\}$. The imaginary part is the continuity equation for $\rho =R^2=|\psi {|}^2$,

$$\frac{\partial \rho }{\partial t}+\sum _k{\nabla }_k\cdot \left(\rho {\mathbf{v}}_k\right)=0,{\mathbf{v}}_k=\frac{{\nabla }_{k}S}{m_k},$$

which confirms that the guidance velocity (B3) is the phase gradient, ${\mathbf{v}}_k={\nabla }_{k}S/m_k$ (indeed $\mathrm{Im}({\nabla }_k\psi /\psi )={\nabla }_{k}S/\hslash$). The real part is the quantum Hamilton–Jacobi equation

$$\frac{\partial S}{\partial t}+\sum _k\frac{({\nabla }_{k}S{)}^2}{2m_k}+V+Q=0,$$

identical to the classical Hamilton–Jacobi equation for the action $S$ except for the single extra term, the quantum potential

$$\boxed{Q(\mathbf{q},t):=-\sum _{k=1}^N\frac{{\hslash }^2}{2m_k}\frac{{\nabla }_k^{2}R}{R}=-\sum _{k=1}^N\frac{{\hslash }^2}{2m_k}\frac{{\nabla }_k^2|\psi |}{|\psi |},}$$

defined wherever $R=|\psi |\ne 0$. Together, the continuity and quantum Hamilton–Jacobi equations are equivalent to the Schrödinger equation (B2) on $\{\psi \ne 0\}$: the change of variables $\psi \leftrightarrow (R,S)$ neither adds nor loses information.

Newtonian second-order form. Differentiating ${\mathbf{v}}_k$ along an actual trajectory with the material (convective) derivative $\frac{d}{dt}={\partial }_t+{\sum }_j{\mathbf{v}}_j\cdot {\nabla }_j$, and taking ${\nabla }_k$ of the quantum Hamilton–Jacobi equation, gives a Newtonian law along the trajectory $\mathbf{Q}(t)$,

$$m_k{\ddot{\mathbf{Q}}}_k=-{\nabla }_k(V+Q){|}_{\mathbf{q}=\mathbf{Q}(t)}.$$

A Bohmian particle thus feels the classical force $-{\nabla }_{k}V$ plus a quantum force $-{\nabla }_{k}Q$. This second-order equation is a consequence of the guidance postulate B3, not an independent axiom (see the note below). All non-classical behaviour — interference, tunneling, quantization, non-locality — is carried by $Q$, whose two decisive features are:

• It is not a fixed external field. $Q$ depends on the shape of $R=|\psi |$, not its magnitude, so it is unchanged by $\psi \to c\psi$. A wave of tiny amplitude can exert a large quantum force; intensity and influence are decoupled. This is why the guiding wave acts as information rather than as a push proportional to energy density.
• It couples the particles non-locally. For entangled $\psi$, $Q(\mathbf{q},t)$ does not decompose into a sum of one-particle terms, so the force on particle $k$ depends instantaneously on the positions of all the others, however far away (see §7).

Two formulations, one theory. The first-order guidance equation (§1) and the second-order Newtonian picture (§2) describe the same trajectories, but the first-order form is fundamental: it needs only $\psi$ and $\mathbf{Q}$, whereas the Newtonian form requires also specifying the initial velocity — which the guidance equation fixes. Modern treatments (Bell, Dürr–Goldstein–Zanghì) take the guidance equation as primary and regard $Q$ as a useful diagnostic, not a foundational entity.

3. Equivariance and Quantum Equilibrium

The theory must recover the Born rule: that the statistics of positions are $|\psi {|}^2$-distributed. In Bohmian mechanics this is not a postulate about a single system — a single system has a definite configuration — but a statement about a probability distribution over an ensemble of systems all guided by the same $\psi$. The key structural fact is equivariance.

Suppose that at some time the configuration is distributed according to some density $\rho (\mathbf{q},t)$. Because the particles move by the velocity field ${\mathbf{v}}^{\psi }$, $\rho$ evolves by the same continuity equation as $|\psi {|}^2$:

$$\frac{\partial \rho }{\partial t}=-\sum _k{\nabla }_k\cdot (\rho {\mathbf{v}}_k^{\psi }),\frac{\partial |\psi {|}^2}{\partial t}=-\sum _k{\nabla }_k\cdot (|\psi {|}^2{\mathbf{v}}_k^{\psi }).$$

Both densities are transported by the identical flow. Therefore, if $\rho (\mathbf{q},t_0)=|\psi (\mathbf{q},t_0){|}^2$ at one time, then $\rho (\mathbf{q},t)=|\psi (\mathbf{q},t){|}^2$ at all times. This is equivariance: the property "distributed as $|\psi {|}^2$" is dynamically preserved, a fixed point of the Bohmian flow.

The distribution $\rho =|\psi {|}^2$ is called quantum equilibrium. The quantum equilibrium hypothesis is that the actual configuration of a (sub)system is typically in quantum equilibrium; granting it, Bohmian mechanics reproduces all the statistical predictions of quantum mechanics, because equivariance guarantees position statistics remain $|\psi {|}^2$ and — since every measurement ultimately registers as a position (a pointer, a mark on a screen, an ink trace) — this suffices for all observable predictions.

4. Why Equilibrium? Typicality and the Origin of the Born Rule

Equivariance shows quantum equilibrium is stable, but not why our world is in it. Two lines of argument address this.

Typicality (Dürr–Goldstein–Zanghì). On the universal configuration space, the measure $|\Psi {|}^2$ (with $\Psi$ the wavefunction of the universe) is the natural, equivariant measure — the unique one, up to the flow, that is stationary under the dynamics. One shows that for the overwhelming majority of initial universal configurations — "overwhelming" as judged by this very $|\Psi {|}^2$ measure — the empirical distribution of positions in any subsystem ensemble is $|\psi {|}^2$, where $\psi$ is the subsystem's conditional wavefunction (§6). The Born rule thus holds for typical initial conditions, in exactly the sense that the second law of thermodynamics holds for typical microstates: it is a law about what happens in all but a measure-zero set of possible worlds. Quantum equilibrium is to Bohmian mechanics what thermal equilibrium is to statistical mechanics.

Dynamical relaxation (Valentini). Alternatively, define a coarse-grained "sub-quantum $H$-function," $H=\int \rho \ln (\rho /|\psi {|}^2)$, an analogue of Boltzmann's $H$. One can argue that for generic non-equilibrium initial $\rho \ne |\psi {|}^2$, mixing by the Bohmian flow drives $H\to 0$, i.e. $\rho \to |\psi {|}^2$, so equilibrium is an attractor. This view is bolder: it entertains that quantum non-equilibrium ($\rho \ne |\psi {|}^2$) might have existed in the early universe and could in principle survive in relic particles, allowing sub-quantum signalling and violations of the uncertainty bound — a speculative but empirically distinguishing possibility, unlike the typicality account, which is strictly equivalent to standard QM.

Either way, the Born rule is explained rather than posited — a genuine achievement, and one of the theory's principal selling points.

5. Recovering the Standard Formalism

Given quantum equilibrium, the rest of the quantum apparatus is recovered as effective description.

• Position measurements are direct: the particle is somewhere, and the pointer that records it is itself made of particles that end up correlated with it. Statistics are $|\psi {|}^2$ by equivariance.
• All other "observables" — momentum, energy, spin — are not properties the particle possesses prior to measurement. They are contextual: what a "momentum measurement" yields is determined by how the actual configuration flows through the particular experimental apparatus, and the outcome statistics coincide with $⟨\psi |\hat{P}|\psi ⟩$ etc. only because the apparatus is engineered so that the final pointer position correlates with the spectral decomposition of the self-adjoint operator. Operators are thus not observables in the naive sense; they are compact bookkeeping for the statistics of position outcomes of whole experiments. (Bell: "the word ['measurement'] should now be banned … in favour of the word experiment.")
• The uncertainty relations hold as statistical facts about $|\psi {|}^2$-ensembles, not as limits on what the particle has: a Bohmian particle has a perfectly definite position and a perfectly definite velocity ${\mathbf{v}}^{\psi }$ at every instant. The uncertainty principle constrains our ability to prepare and predict, because a sharp position measurement disturbs $\psi$ (§6), not because the underlying quantities are indefinite.

6. Measurement, the Conditional Wavefunction, and Effective Collapse

The measurement problem is dissolved by two constructions.

The conditional wavefunction. Split the universe into a subsystem (configuration $\mathbf{x}$) and its environment (configuration $\mathbf{y}$, actual value $\mathbf{Y}(t)$). The subsystem's conditional wavefunction is defined by plugging the actual environmental configuration into the universal wavefunction:

$${\psi }_t^{\text{sub}}(\mathbf{x})\equiv {\Psi }_t(\mathbf{x},\mathbf{Y}(t)).$$

This is the object that plays the role of "the wavefunction of the subsystem" and that appears in the guidance equation for the subsystem's particles. Crucially, it is well-defined even when the subsystem is entangled with its environment, where standard QM would assign only a reduced density matrix.

Effective collapse. Consider a measurement that couples system states $|a_n⟩$ to macroscopically distinct, non-overlapping apparatus configurations, producing the entangled

$$\Psi =\sum _n{c}_n|a_n⟩{\varphi }_n(\mathbf{y}),{\varphi }_n\text{ supported in disjoint regions }{\Omega }_n\subset \text{config space}.$$

The actual apparatus configuration $\mathbf{Y}$ lands in exactly one region, say ${\Omega }_m$, because the particles have definite positions. Since the ${\varphi }_n$ have disjoint supports, all terms $n\ne m$ vanish at $\mathbf{y}=\mathbf{Y}$, so the conditional wavefunction collapses to the single branch

$${\psi }_t^{\text{sub}}(\mathbf{x})={\Psi }_t(\mathbf{x},\mathbf{Y})\propto c_m|a_m⟩.$$

Collapse is thus a theorem, not an axiom: the "other branches" of $\Psi$ are still present in configuration space, but they are empty (contain no particles) and, once decoherence has driven their supports apart so they never again overlap, they can never again influence the actual trajectory. They become dynamically irrelevant "ghost" or "empty" waves. This is precisely the role decoherence plays for Bohm: it does not select an outcome (the particle position does that) but it guarantees the branches stay separated, making the effective collapse permanent. The outcome that occurs is the one the pre-existing configuration was already headed for; the Born-rule probability of getting outcome $m$ is $|c_m{|}^2$ by quantum equilibrium.

7. Non-Locality and Bell's Theorem

Bohmian mechanics is explicitly, manifestly non-local, and this is a feature, not a bug: Bell's theorem guarantees that any hidden-variable theory reproducing quantum predictions must be non-local. Bohm's theory wears its non-locality openly.

For an entangled two-particle state $\psi ({\mathbf{x}}_1,{\mathbf{x}}_2)$ that does not factorize, the guidance velocity of particle 1,

$${\mathbf{v}}_1=\frac{\hslash }{m_1}\mathrm{Im}\frac{{\nabla }_1\psi }{\psi }{|}_{({\mathbf{x}}_1,{\mathbf{x}}_2)=({\mathbf{Q}}_1,{\mathbf{Q}}_2)},$$

depends on the instantaneous actual position ${\mathbf{Q}}_2$ of particle 2, no matter how far away. Adjusting a distant apparatus that changes $\psi$ changes particle 1's motion at once. This is exactly the EPR–Bohm correlation, now mechanistically explained: the singlet's two particles are choreographed by a single wavefunction on the joint configuration space, and that wave links them rigidly.

Two points reconcile this with relativity's prohibition on signalling:

• No superluminal signalling. Although the trajectories respond non-locally, the statistics do not: because the configuration is in quantum equilibrium and one cannot control the initial position within the $|\psi {|}^2$ distribution, the marginal statistics at particle 1 are independent of the distant setting. Quantum equilibrium is exactly the condition that makes the manifest non-locality unobservable at the statistical level — the same no-signalling theorem as in standard QM. (Valentini's non-equilibrium, by contrast, would permit signalling — which is why it is empirically distinguishable.)
• A preferred foliation. The instantaneous dependence on distant configurations requires a notion of "simultaneous," i.e. a preferred slicing of spacetime. Non-relativistic Bohmian mechanics simply uses absolute Newtonian time. The relativistic extension (§9) must posit a preferred foliation — a real tension with the spirit of relativity, even though the observable predictions remain Lorentz-invariant.

Bohm as a hidden-variable theory: mapping Bell's $\lambda$ and $\rho (\lambda )$. It is worth making explicit how Bohmian mechanics instantiates the abstract local-hidden-variable template whose factorization Bell's theorem constrains — because Bohm reproduces every ingredient of that template except the one it must, factorizability. The dictionary is:

• the hidden-variable space $\Lambda$ is the configuration space ${\mathbb{R}}^{3N}$;
• the complete state $\lambda$ is the actual initial configuration $\mathbf{Q}(0)$ — the particle positions of Postulate B1, with the pilot wave $\psi$ carried along as a shared, non-hidden field;
• the probability measure $\rho (\lambda )d\lambda$ is the quantum-equilibrium density $|\psi (\mathbf{q},0){|}^2{d}^{3N}q$ (§3);
• the normalization ${\int }_{\Lambda }\rho (\lambda )d\lambda =1$ is therefore nothing but the normalization of the wavefunction,

$${\int }_{\Lambda }\rho (\lambda )d\lambda ⟼{\int }_{{\mathbb{R}}^{3N}}|\psi (\mathbf{q}){|}^2{d}^{3N}q=\Vert \psi {\Vert }^2=1,$$

a condition equivariance preserves for all time. Bell's ensemble average over the hidden state then becomes an average over the initial configuration weighted by $|\psi {|}^2$,

$$E(a,b)={\int }_{\Lambda }d\lambda \rho (\lambda )A(a,\lambda )B(b,\lambda )⟼{\int }_{{\mathbb{R}}^{3N}}{d}^{3N}q|\psi (\mathbf{q}){|}^{2}A(a,b,\mathbf{q})B(a,b,\mathbf{q}),$$

the outcomes $A,B$ being definite functions of $\mathbf{Q}(0)$ because the dynamics is deterministic (§3).

The decisive point is where Bohm departs from the template. It honours the measure exactly, and it also satisfies measurement independence, $\rho (\lambda \mid a,b)=\rho (\lambda )$: the initial $|\psi {|}^2$ distribution over $\mathbf{Q}(0)$ does not depend on the future settings, so Bohm is emphatically not superdeterministic. What it violates is factorizability alone — the outcomes are written $A(a,b,\mathbf{q})$ and $B(a,b,\mathbf{q})$ deliberately: since $\psi$ lives on the joint configuration space, each wing's trajectory depends on the distant setting, so

$$P(A,B\mid a,b,\mathbf{Q})\ne P(A\mid a,\mathbf{Q})P(B\mid b,\mathbf{Q}).$$

The CHSH derivation needs factorizability in addition to the normalized measure, so the bound $|S|\le 2$ simply does not apply, and the same $|\psi {|}^2$-average reaches $2\sqrt{2}$. The distant-setting dependence washes out only after averaging over quantum equilibrium, restoring no-signalling at the level of marginals (first bullet above).

Which half of factorizability fails: PI, not OI. Splitting factorizability into Parameter and Outcome Independence (Jarrett–Shimony) locates Bohm's nonlocality precisely. Its failure is one of Parameter Independence: writing the outcomes as $A(a,b,\mathbf{q})$ lets the distant setting enter each wing's response. Outcome Independence, by contrast, holds trivially — determinism fixes both outcomes as definite functions of $\lambda =\mathbf{Q}(0)$, so once $\lambda$ and the two settings are given the results are just numbers, leaving no residual outcome–outcome correlation to break. This is the exact mirror of orthodox QM, which keeps PI (no-signalling) and violates OI: Bohm keeps OI and violates PI, and both respect signal-locality because Bohm's PI-violation stays hidden in the quantum-equilibrium marginals.

For a measurement proper, $\lambda$ is enlarged to include the apparatus configuration and $\rho =|\Psi {|}^2$ on the joint space — the (harmless) enlargement Fine's theorem licenses to make the response functions deterministic.

How and why Bohmian mechanics obeys the no-cloning theorem. Because the theory posits definite positions, one might expect it to permit copying an unknown state by reading the hidden variable — in apparent conflict with the no-cloning theorem. Bohmian mechanics obeys no-cloning nonetheless, and seeing why is instructive about where the theorem's content actually lives. No-cloning is a statement about the linear dynamics of $\psi$: a copying device is a unitary $\hat{U}$ generated by some Hamiltonian, and no linear $\hat{U}$ can satisfy $\hat{U}(|\psi ⟩|b⟩)=|\psi ⟩|\psi ⟩$ for all $|\psi ⟩$. Bohmian mechanics keeps the ordinary Schrödinger equation for $\psi$ untouched (Postulate B2), so it inherits no-cloning verbatim — the guidance equation (B3) plays no role in the proof and requires no special adjustment to comply. The particles ride on $\psi$; they cannot produce a copy the wavefunction dynamics itself forbids.

Nor does the hidden variable open a back door. "Cloning" means reproducing the quantum state — the conditional wavefunction ${\psi }^{\text{sub}}$ — not the position $\mathbf{Q}$, which is merely one point guided by it. Reading $\mathbf{Q}$ finely enough to reconstruct ${\psi }^{\text{sub}}$ is precisely what quantum equilibrium forbids: in equilibrium the configuration is $|\psi {|}^2$-distributed and no measurement extracts more than the Born statistics of a single ordinary measurement. So the theory obeys the theorem for two complementary reasons, and the division of labour is sharp:

$$\text{no-cloning barrier}=\underset{\text{forbids the copying unitary}}{\underbrace{\text{linear }\psi \text{-dynamics}}}+\underset{\text{hides the position }\mathbf{Q}}{\underbrace{\rho =|\psi {|}^2}},$$

and neither term is a tuning of the guidance law. The guidance velocity is fixed once and for all as the Schrödinger current over the density, ${\mathbf{v}}^{\psi }=\mathbf{j}/|\psi {|}^2$ (§1), and equivariance (§3) makes $|\psi {|}^2$ a dynamical fixed point automatically — the "tuning" is an initial condition on the distribution of $\mathbf{Q}$, justified by typicality (§4), not a parameter in the law of motion. That this is where the obedience comes from is confirmed by Valentini's quantum non-equilibrium ($\rho \ne |\psi {|}^2$): with the same guidance equation but an out-of-equilibrium distribution, sub-quantum measurements would resolve $\mathbf{Q}$ below the $|\psi {|}^2$ scale, distinguish non-orthogonal states, and permit both signalling (§7, first bullet) and effective cloning — showing that Bohmian mechanics respects no-cloning by virtue of the linear $\psi$-dynamics plus equilibrium, and by nothing in the guidance law that would need to be tuned by hand.

8. Spin, Contextuality, and Kochen–Specker

Bohmian mechanics is a theory of positions only; it introduces no "spin variable." Spin is carried entirely by the wavefunction, which is spinor-valued: for a spin-$\frac{1}{2}$ particle $\psi =({\psi }_{\uparrow },{\psi }_{\downarrow }{)}^{\top }$, and the guidance equation uses the spinor inner product,

$$\mathbf{v}=\frac{\hslash }{m}\mathrm{Im}\frac{{\psi }^{†}\nabla \psi }{{\psi }^{†}\psi },$$

with ${\psi }^{†}\chi$ the spinor product. In a Stern–Gerlach experiment, the inhomogeneous field splits the wave packet into an up-deflected and a down-deflected part; the particle, riding one of them according to where it started in the packet, is deflected up or down — and that deflection (a position) is what "measuring $S_z$" records. The particle never "had" a spin; the outcome is jointly created by the wavefunction and the initial position, given the apparatus orientation.

This makes vivid the contextuality that the Kochen–Specker theorem shows is forced on any hidden-variable theory: the result of "measuring $\hat{A}$" can depend on which other compatible observables are measured alongside it (the experimental context), because the result is a joint product of state, position, and apparatus, not the revelation of a pre-existing value attached to $\hat{A}$ alone. Bohmian mechanics evades the no-hidden-variables theorems precisely because it does not assign context-independent values to all observables — only positions are beables; everything else is contextual.

9. The Classical Limit

Classical behaviour emerges when the quantum potential and quantum force are negligible compared with the classical terms, $|Q|\ll |V|$ and $|\nabla Q|\ll |\nabla V|$. Then the quantum Hamilton–Jacobi equation of §2 reduces to the classical one, and Bohmian trajectories coincide with Newtonian ones. This happens for wave packets that stay narrow and quasi-Gaussian (so ${\nabla }^2|\psi |/|\psi |$ is small and slowly varying), i.e. for heavy, decohered, well-localized bodies. Decoherence does double duty: it keeps the effective (conditional) wavefunction a localized packet, and for such packets $Q$ is nearly trivial, so the center of the packet — carrying the actual particle — follows Newton's laws. Unlike in standard QM, there is no interpretive puzzle about "how definite trajectories emerge from a spread-out wavefunction": the trajectory was always there; the classical limit is just the regime where the quantum force switches off.

10. Relativistic and Field-Theoretic Extensions

Extending pilot-wave theory beyond the non-relativistic $N$-particle case is where its difficulties concentrate.

• Single Dirac particle. A one-particle Dirac theory admits a natural Bohmian velocity $\mathbf{v}=c{\psi }^{†}\mathbf{\alpha }\psi /{\psi }^{†}\psi$ (the Dirac current over the density), which is even bounded by $c$. This works cleanly.
• Many particles / entanglement. The trouble is non-locality plus relativity: the guidance law needs the simultaneous positions of all particles, forcing a preferred foliation of spacetime (§7). One can make the observable predictions fully Lorentz-invariant while the underlying trajectories secretly single out a rest frame; some regard the foliation as an additional dynamical field, others as an unavoidable cost.
• Quantum field theory. Two main strategies exist. (i) Bohm–Dirac field beables: take the field configuration (e.g. the value of a bosonic field on space) as the beable, guided by a wave functional $\Psi [\varphi ]$ — natural for bosons, awkward for fermions. (ii) Bell-type / Dürr–Goldstein–Tumulka–Zanghì "Bell beables": keep particle positions as beables and add a stochastic law for particle creation/annihilation, so the particle number jumps as the field-theoretic $\Psi$ dictates. Both reproduce the QFT predictions in their domains but neither is as clean or canonical as the non-relativistic theory, and none has been extended convincingly to a Bohmian version of the Standard Model or gravity. This is the honest weak point of the programme.

Part II — Philosophical Analysis

Bohmian mechanics is philosophically important out of proportion to the number of physicists who adopt it, because it is a proof of concept: it demonstrates, by construction, that a clear, deterministic, observer-free, "classical-realist" picture of the quantum world is possible. Whatever one thinks of its cost, it refutes the once-standard claim — associated with von Neumann's flawed no-hidden-variables theorem and with Copenhagen orthodoxy — that no such picture can exist. Every serious argument against it is therefore an argument about price, not possibility.

11. What the Theory Buys: A Solution to the Measurement Problem

The measurement problem is the incoherence of having two evolution laws — unitary and collapse — with no principled account of when each applies. Bohmian mechanics has one dynamical law for the wavefunction (Schrödinger, always) and one for the particles (guidance, always). Measurement is not a primitive category; it is ordinary interaction, and "collapse" is the derived, effective phenomenon of §6. There is:

• no observer playing a physical role — the theory never mentions measurement in its axioms;
• no vague "classical/quantum cut" of the Heisenberg kind;
• no need to interpret superpositions of pointers as anything mysterious — the pointer is where its particles are, and the other branches are empty waves.

This is a real and rare virtue. Bohmian mechanics, GRW, and many-worlds are the three programmes that state a precise microscopic theory with no primitive "measurement," what Bell called theories "without observables, only beables." Copenhagen and its epistemic descendants, by contrast, arguably do not solve the problem so much as decline to state a theory of the world at all.

12. Determinism, and Why the Randomness Is Epistemic

In Bohmian mechanics the universe is fully deterministic: fix $({\Psi }_0,{\mathbf{Q}}_0)$ and all of history follows. Quantum randomness is entirely epistemic — it reflects our irreducible ignorance of the exact initial configuration within the $|\psi {|}^2$ distribution, in precise analogy to how classical statistical-mechanical probabilities reflect ignorance of the exact microstate. This is philosophically attractive to those who find fundamental, lawless chance (Copenhagen, GRW) unpalatable: it restores the Laplacean picture at the fundamental level while explaining, via §4, why the world nonetheless looks irreducibly random to agents embedded in it who cannot know or control ${\mathbf{Q}}_0$ better than $|\psi {|}^2$ permits.

Note the subtlety: the determinism is not in tension with the impossibility of prediction. Absolute uncertainty is itself a theorem — the very equilibrium that makes the theory empirically adequate also forbids any Bohmian observer from exploiting the underlying determinism to out-predict standard QM. The hidden variables are hidden for a reason the theory itself supplies.

13. The Ontology: Primitive Ontology and the Status of the Wavefunction

Bohmian mechanics is the clearest case study for the notion of primitive ontology — the part of the theory's furniture that lives in ordinary three-dimensional space and directly constitutes the physical world we see. For Bohm the primitive ontology is the particle configuration: matter is, literally, point particles tracing world-lines in space. Tables and pointers are patterns of these particles. Because the beables live in 3-space, the theory has an immediate, unproblematic account of why our experience is of things arranged in space — a question that is notoriously delicate for many-worlds, whose fundamental object is a wavefunction on high-dimensional configuration space.

But then: what is the wavefunction? Here Bohmians divide, and the debate is one of the richest in philosophy of physics.

• The wavefunction as a physical field. Read $\Psi$ as a real physical entity — a "pilot wave" — that pushes the particles. Objection: it lives on $3N$-dimensional configuration space, not on 3-space; taking it as physically real seems to commit one to configuration-space realism (the fundamental arena is ${\mathbb{R}}^{3N}$, and three-dimensional space is derivative), which many find extravagant. It is also odd as a field: it acts on the particles but they exert no back-reaction on it — a violation of the action–reaction reciprocity that all other physical fields respect.
• The wavefunction as nomological (Dürr–Goldstein–Zanghì). Read $\Psi$ not as a thing but as a law — a component of the dynamical law for the particles, more like the Hamiltonian or the classical potential than like matter. On this view the strange features dissolve: laws are supposed to live in configuration space (the classical Hamiltonian is a function on phase space and no one reifies phase space), laws are supposed to act without being acted on, and the universal $\Psi$'s not being "in space" is no more troubling than Newton's $1/r^2$ not being "in space." The tension is that $\Psi$ is time-dependent and contingent (it obeys its own dynamical Schrödinger equation), which laws are usually not; the nomological view is cleanest for a timeless universal wavefunction of the sort a Wheeler–DeWitt cosmology might supply.
• Intermediate views: the wavefunction as a disposition/property of the particles, or as a "nomological but non-fundamental" quasi-law. The literature here is active and unsettled.

The upshot: Bohmian mechanics forces the general question "is the wavefunction a thing or a law?" into the open, and different answers carry different metaphysical bills. This is often counted a virtue — the theory is precise enough to have a sharp ontology to argue about.

Why the trajectory does not replace the wavefunction. A natural objection presses in the opposite direction: if the particle configuration is the whole material ontology, and the actual history $\mathbf{Q}(t)$ records where all matter ever is, why retain $\psi$ at all — does $\mathbf{Q}(t)$ not already contain everything? The reply sharpens exactly what $\psi$ is. Ontologically the objection has a point: the particles are the "stuff," $\psi$ is not a second substance, and $\mathbf{Q}(t)$ is indeed the complete inventory of matter. But a theory is a law that generates and explains the history, not the history itself, and here $\mathbf{Q}(t)$ is strictly insufficient for three reasons.

• There is no autonomous law for $\mathbf{Q}$ alone. The guidance equation sets ${\dot{\mathbf{Q}}}_k=(\hslash /m_k)\mathrm{Im}({\nabla }_k\psi /\psi )$ evaluated at the actual point; there is no closed $\dot{\mathbf{Q}}=\mathbf{F}(\mathbf{Q})$. Knowing the velocity even an instant later requires $\psi$ in a neighborhood, and propagating $\psi$ by Schrödinger couples the point to all of configuration space through ${\nabla }^2\psi$. The trajectory cannot be its own dynamics.
• $\psi$ encodes the guiding field everywhere; $\mathbf{Q}(t)$ samples it along one curve. The wavefunction defines a velocity field over the entire space ${\mathbb{R}}^{3N}$, while the actual trajectory reads that field only along the single streamline $\mathbf{q}=\mathbf{Q}(t)$. The unvisited regions are not idle: a branch of $\psi$ where the particle isn't can later steer where it goes. In the two-slit experiment the particle threads one slit but $\psi$ passes through both, and the empty branch is exactly what produces the interference that bends the trajectory (equivalently, the quantum potential depends on the shape of $|\psi |$ around the particle, not on its location). That influence is recorded only in $\psi$, never in a single $\mathbf{Q}(t)$.
• The reconstruction asymmetry. From one trajectory $\psi$ cannot be recovered — it fixes the field along a single curve. From the whole quantum-equilibrium ensemble it essentially can: the density gives the amplitude $|\psi |=\sqrt{\rho }$ and the current gives the phase via ${\mathbf{j}}_k=\rho {\nabla }_{k}S/m_k$, reconstructing $\psi =|\psi |e^{iS/\hslash }$ up to a global phase. So $\psi$ is a field / ensemble-level object — what an entire family of possible trajectories has in common — and a single realization $\mathbf{Q}(t)$ cannot contain it.

The contrast with classical mechanics is instructive: there one can often read the force off a single orbit via $\mathbf{F}=m\ddot{\mathbf{x}}$, because the law is second-order and local. In Bohm even that fails — the quantum force at the particle depends on the curvature of the surrounding, possibly empty, wave — so the trajectory is more dependent on the field, not less. $\psi$ is thus indispensable as the law-like structure that generates $\mathbf{Q}(t)$, whatever one's verdict (physical field vs. nomological) on its ultimate status.

14. Non-Locality and the Confrontation with Relativity

The deepest philosophical cost is non-locality and its uneasy relation to relativity. Bell's theorem makes clear that non-locality is not peculiar to Bohm — nature is non-local, and any empirically adequate theory must be — so one cannot fault Bohm for having it. What is peculiar is that Bohm's non-locality is explicit at the level of the beables' dynamics and appears to require a preferred foliation of spacetime (§7, §9). This raises pointed questions:

• Is a preferred frame a defect or a discovery? Bohmians can argue that since Bell has shown non-locality is real, and since non-locality most naturally lives on a foliation, the preferred frame is simply what nature is telling us — the Lorentz invariance of the observable statistics being an emergent, "conspiratorial" symmetry (compare the way Lorentzian relativity recovers all of special relativity's predictions atop a hidden ether frame). Critics reply that positing an in-principle-undetectable frame — one the theory's own no-signalling theorem forbids us ever to find — is exactly the kind of empirically idle structure that Einstein's relativity taught us to excise (an "ether by another name").
• Fundamental vs. phenomenal Lorentz invariance. The situation dramatizes a general question: is it acceptable for a theory's fundamental ontology to violate a symmetry that its empirical predictions perfectly respect? Everett and GRW-flash proponents claim advantages here; the debate over whether a fully Lorentz-invariant pilot-wave theory (with no preferred foliation, perhaps with a foliation determined by $\Psi$ itself) is possible remains open and technical.

15. Empirical Equivalence and the Underdetermination Problem

Because Bohmian mechanics reproduces exactly the predictions of standard QM (in quantum equilibrium), it furnishes a textbook case of the underdetermination of theory by evidence: here are two (indeed several) empirically indistinguishable theories — Bohm, Everett, GRW-in-its-equivalent-regimes, Copenhagen — and no experiment can decide among them. Philosophically this is a live laboratory for questions about theory choice beyond evidence:

• If theories are empirically equivalent, are the grounds for choosing (simplicity, explanatory power, ontological clarity, unity) epistemic or merely pragmatic/aesthetic?
• Does the very existence of an empirically adequate deterministic-realist alternative undercut claims that quantum mechanics teaches us indeterminism, or the collapse of realism, or the essential role of the observer? Many philosophers think it does: one cannot read a metaphysical moral off the formalism when an equally good formalism carries the opposite moral.
• Valentini's escape hatch. The equivalence is contingent on quantum equilibrium. If quantum non-equilibrium ($\rho \ne |\psi {|}^2$) exists anywhere — relic cosmological particles, say — Bohm predicts deviations from the Born rule and even superluminal signalling. This would break the underdetermination empirically, turning an interpretive dispute into a physical one. That such a thing is even conceivable distinguishes Bohm from interpretations that are equivalent to QM by construction.

16. Objections and Replies

• "The empty branches are an extravagant ontology." After a measurement, all the terms of $\Psi$ persist as empty waves guiding no particles; the theory keeps the entire Everettian wavefunction and adds particles on top. Deutsch's jibe: Bohm is "many-worlds in a state of chronic denial" — it retains all the branches but declares only one "occupied." Reply: the empty branches are not worlds (they contain no matter, no beables, no observers — there is nothing in them for anyone to inhabit); they are structure in the law, and on the nomological reading (§13) not an "ontology" at all. Whether this reply succeeds is exactly the question of the wavefunction's status.
• "Surreal trajectories." Englert–Scully–Süssmann–Walther (1992) constructed which-path setups in which the Bohmian trajectory seems to go through one slit while a "which-path" detector registers the other — trajectories that appear to contradict the naive particle-detector correlation, hence "surreal." Reply: later analysis (and the very meaning of "detection" in Bohm) shows the detectors are triggered non-locally, so the apparent paradox is just non-locality (§7) plus the mistake of assuming detector records track local particle passage; recent weak-measurement experiments (Kocsis et al. 2011; Mahler et al. 2016) have even reconstructed the Bohmian trajectory ensemble and found it consistent. The objection reduces to discomfort with non-locality, which is not optional post-Bell.
• "Operators-as-observables is naive; Bohm has to reinterpret everything." Bohm must demote momentum, energy, and spin from properties to contextual outcomes (§5, §8). Reply: Bohmians turn this into a virtue — the Kochen–Specker and Bell theorems prove that the naive view (all operators have simultaneous context-independent values) is impossible, so any viable realism must be contextual; Bohm just makes explicit what the no-go theorems already force on everyone.
• "Why privilege position?" The choice of positions as the beables looks arbitrary — why not momenta? Reply: position is special because all measurements ultimately terminate in positions (pointer locations, marks, clicks localized in space), so a theory whose beable is position automatically has definite records; a "momentum-beable" theory would have to explain records some other way. Bell and Dürr–Goldstein–Zanghì argue this asymmetry is principled, not arbitrary.
• "Occam's razor / simplicity." Bohm adds particles and keeps the whole wavefunction, so it is (the objection goes) strictly more than Everett, which keeps only $\Psi$. Reply: Everett must derive the appearance of definite outcomes, a 3-dimensional world, and the Born probabilities from $\Psi$ alone — projects of contested success — whereas Bohm gets all three immediately from the particles. "Fewer entities" is not "fewer problems"; parsimony of ontology trades against parsimony of interpretive burden.

17. Where Bohm Sits Among the Interpretations

Relative to the survey of interpretations, Bohmian mechanics occupies a distinctive corner: it is the option that keeps determinism, keeps a definite single world, keeps realism about a spatial ontology, and pays for all of this in a single currency — explicit non-locality (with its foliation problem) and the metaphysical puzzle of the wavefunction. Its natural rivals trade differently:

• vs. Many-worlds: Bohm keeps one world and adds particles; Everett keeps only $\Psi$ and multiplies worlds. Both are deterministic and collapse-free; they differ on ontology and on whether the Born rule is derived by typicality (Bohm) or decision theory (Everett).
• vs. GRW/CSL: both banish the observer, but GRW modifies the dynamics (real, stochastic collapse, hence empirically distinguishable) while Bohm keeps the Schrödinger dynamics exact and adds variables. GRW is indeterministic; Bohm is deterministic.
• vs. Copenhagen/QBism: Bohm supplies exactly the observer-independent story of the world that the epistemic interpretations deny is available or necessary; the contrast is a clean test case of scientific realism vs. instrumentalism.

The lasting significance is less that Bohmian mechanics is true — few working physicists commit to it — than that it is possible: a fully precise, empirically adequate, deterministic account of the quantum world, which recasts every quantum "mystery" as a claim about price, and thereby sharpens what is really at stake in the measurement problem.

See also

• The Measurement Problem and Interpretations — the survey this page expands.
• Entanglement, EPR, and Bell's Theorem — why non-locality is unavoidable.
• Decoherence and the Classical Limit — how empty branches stay separated.
• Wave Mechanics — the probability current that becomes the guidance velocity.
• Postulates of Quantum Mechanics — the Born rule and collapse that Bohm derives.
• Uncertainty Relations — statistical, not ontic, in this theory.
• Spin and the SU(2) Representation — spin as a wavefunction property.
• general/quantum-mechanics.md — conceptual/philosophical overview.