Entropic Dynamics Approach

Updated 29 November 2025

Entropic Dynamics is an inference-based framework that derives quantum mechanics from maximum-entropy updates under physical constraints.
It employs short-step transitions and information geometry to reconstruct the Schrödinger equation, gauge symmetry, and charge quantization.
The approach bridges classical and quantum regimes, enabling extensions to quantum field theory and relational dynamics through epistemic variables.

Entropic Dynamics (ED) is a framework in which dynamical laws—especially those of quantum mechanics and quantum field theory—are derived from entropic methods of inference, constrained by physical information and formulated within an information-geometric phase space. In ED, the only ontic variables are the definite but unknown particle positions; all other physical variables (wave function, phase, energy) are epistemic constructs encoding knowledge and updating rules. The central methodology is to formulate motion as a succession of short steps, assigning transition probabilities by maximizing relative entropy subject to appropriate physical constraints. The resulting dynamical equations recover the standard structure of quantum mechanics, including the Schrödinger equation, by imposing that the evolution preserve not just a symplectic (Hamiltonian) flow but also the Riemannian (information-geometric) metric—a requirement that produces a Kähler geometry identical to projective Hilbert space. ED affords a natural, inference-based derivation of gauge symmetry, charge quantization, and the role of quantum phases, as well as providing a foundation for extension to quantum field theory, gauge interactions, and relativistic models.

1. Ontology, Epistemology, and the Entropic Inference Principle

The ED framework posits that the only fundamental ("ontic") variables are particle positions $x\in\mathbb{R}^{3N}$ ; all other quantities—wave function $\Psi(x)$ , density $\rho(x)$ , phase $\phi(x)$ , and operators—are epistemic, reflecting knowledge and its logical updates (Caticha, 2017). The key tool of inference is the maximization of relative entropy,

$S[P,Q] = -\int dx'\, P(x'|x) \log \frac{P(x'|x)}{Q(x'|x)}$

where $Q(x'|x)$ is a prior enforcing continuity and isotropy (short, uncorrelated steps). Physical information is implemented by constraints—such as expected drift and gauge interactions—that encode directionality, correlations, and coupling to external fields (Caticha, 2017, Carrara et al., 2017).

In this context, the wave function is epistemic in both magnitude and phase: $|\Psi(x)|^2 = \rho(x)$ gives the epistemic probability density, while the phase $\phi(x)$ parametrizes the flow of probability and is also epistemic, controlling the "drift" component in the transition probability (Caticha, 2017, Caticha, 2022).

2. Maximum-Entropy Update and Derivation of Evolution Equations

Motion is modeled as a sequence of infinitesimal transitions $x\to x'$ , each governed by the maximum-entropy distribution $P(x'|x)$ . The Gaussian form of $P(x'|x)$ is dictated by the short-step prior,

$Q(x'|x) \propto \exp\left[-\frac{1}{2}\sum_n \alpha_n\,\delta_{ab}\,\Delta x_n^a \Delta x_n^b\right],$

with $\alpha_n\to\infty$ and $\Delta x_n^a = x_n^{\prime\,a} - x_n^a$ . Physical constraints take the form of expected drifts:

Directionality: impose an average shift along the gradient of a "drift potential" $\phi(x)$ :

$\langle \Delta\phi \rangle = \sum_n \langle \Delta x_n^a \rangle\,\partial_{x_n^a}\phi(x) = \kappa'$

Gauge interaction: for electromagnetic coupling, introduce a local angle (gauge variable) $\chi(x_n)$ and connection $A_a(x_n)$ , imposing

$\langle \Delta x_n^a [\partial_a\chi(x_n) - A_a(x_n)] \rangle = \mathrm{const}$

Maximizing entropy under these constraints yields

$P(x'|x) \propto \exp\left\{ -\sum_n \frac{\alpha_n}{2} \delta_{ab} \left[\Delta x_n^a - \langle \Delta x_n^a \rangle\right] \left[\Delta x_n^b - \langle \Delta x_n^b \rangle\right]\right\}$

with

$\langle \Delta x_n^a \rangle = \frac{\hbar}{m_n} \delta^{ab}[\partial_b(\phi+\chi) - q_n A_b] \Delta t, \quad \langle \Delta w_n^a \Delta w_n^b \rangle = \frac{\hbar}{m_n}\, \delta^{ab}\, \Delta t$

(Caticha, 2017, Carrara et al., 2017).

Entropic time is constructed such that

$\rho_{t+\Delta t}(x') = \int dx\, P(x'|x)\, \rho_t(x),$

with $\Delta t$ chosen so that the diffusion coefficient is $\hbar / m_n$ . In the $\Delta t\to 0$ limit, this leads to the Fokker–Planck (continuity) equation,

$\partial_t \rho = - \sum_n \partial_a [\rho\, v_n^a]$

with current velocity

$v_n^a = \frac{\hbar}{m_n}\left(\partial^a\phi - q_n A^a\right) - \frac{\hbar}{m_n}\partial^a \log\rho^{1/2}$

(Caticha, 2017).

3. Information Geometry of E-Phase Space: Metric, Symplectic, and Complex Structures

ED assigns a natural geometry to the phase space of epistemic variables $(\rho,\phi)$ . The Fisher information metric on the probability space,

$ds^2 = \sum_i \frac{dp_i^2}{p_i} \quad \to \quad G_{\rho\rho}(x,x') = \frac{\hbar}{2\rho(x)} \delta(x-x'),$

is extended to e-phase space as

$ds^2 = \int dx \left[ \frac{\hbar}{2\rho(x)} (d\rho)^2 + \frac{2\rho(x)}{\hbar} (d\phi)^2 \right]$

(Caticha, 2017). Lowering an index with the metric gives the symplectic form,

$\Omega = \int dx\, d\rho(x) \wedge d\phi(x), \quad d\Omega=0.$

The compatible complex structure $J$ acts as

$J: (d\rho, d\phi) \mapsto \left( \frac{2\rho}{\hbar}d\phi, -\frac{\hbar}{2\rho}d\rho \right),$

satisfying $J^2 = -1$ and $G(J\cdot, J\cdot) = G(\cdot,\cdot)$ . Thus, $(G, \Omega, J)$ form a Kähler triple, establishing a geometric structure identical (up to normalization) to the projective Hilbert-space geometry of conventional quantum mechanics (Caticha, 2017, Carrara et al., 2017).

4. Reconstruction of the Schrödinger Equation and Quantum Features

ED reconstructs the quantum dynamics by combining the derived continuity equation for the probability density $\rho$ with a Hamilton–Jacobi-type equation for the phase $\phi$ : $\partial_t \phi + \sum_n \left[ \frac{(\partial_a\phi - q_n A_a)^2}{2m_n} + V\right] - \frac{\hbar^2}{2m_n} \frac{\nabla^2 \sqrt{\rho}}{\sqrt{\rho}} = 0$ (Caticha, 2017). By defining the complex wave function as

$\Psi = \sqrt{\rho}\, e^{i\phi/\hbar},$

these coupled real equations are algebraically equivalent to the linear Schrödinger equation,

$i\hbar \partial_t \Psi = \sum_n \left[ \frac{1}{2m_n}(-i\hbar\partial_a - q_n A_a)^2 + V \right]\Psi$

with minimal coupling to the vector potential (Caticha, 2017, Carrara et al., 2017, Caticha et al., 2014).

Linearity, the use of complex numbers, and the superposition principle are all consequences of the requirement that the quantum flow is simultaneously symplectic (Hamiltonian) and isometric (Killing) in the information-geometric sense (Caticha, 2019).

5. Gauge Symmetry, Quantum Phases, and Charge Quantization

ED gives a systematic, inferential origin to gauge invariance and charge quantization. The phase $\phi$ enters the dynamics only via its gradient $\partial_a\phi - qA_a$ , so global shifts $\phi\to\phi+q\chi$ leave physical predictions invariant if accompanied by $A_a\to A_a+\partial_a\chi$ . Consistency with single-valuedness of $\Psi=\sqrt{\rho}e^{i\phi/\hbar}$ around any closed loop $\Gamma$ in configuration space enforces

$\oint_\Gamma d\ell\cdot\nabla\phi = 2\pi\hbar n, \quad n \in \mathbb{Z}$

(Caticha, 2017, Carrara et al., 2017). For a loop encircling the $n$ th particle, the Lagrange multiplier associated to the gauge constraint must be an integer, $B_n\in\mathbb{Z}$ , and the corresponding charge $q_n = \hbar B_n$ is quantized (Caticha, 2017). This mechanism ties together gauge symmetry, the necessity of quantum phases, and the quantization of electric charge. The requirement of single-valuedness is thus a necessary and sufficient condition for both linearity and probabilistic consistency of the quantum theory within the ED approach (Carrara et al., 2017).

6. Classical and Relational Extensions; Limitations and Generalizations

ED naturally encompasses classical mechanics as a limiting case: when the quantum (Fisher information) term in the Hamiltonian functional is dropped, the equations reduce to classical diffusion or Hamilton-Jacobi theory (Caticha, 2017, Caticha, 2017, Caticha, 2008). The framework can be extended to incorporate additional symmetries (translation, rotation, gauge), as in relational ED, where best-matching procedures and constraints on expectation values govern relational dynamics and offer a quantum implementation of Mach's principles (Ipek et al., 2016, Caticha et al., 9 Jun 2025).

Applications of ED extend to quantum field theory (scalar and gauge fields), economic modeling (e.g., option pricing), and statistical manifolds (e.g., Gibbs distributions), with the information-geometric metric governing the dynamical flows (Ipek et al., 2014, Ipek et al., 2018, Pessoa et al., 2020, Abedi et al., 2019). In quantum field contexts, the preservation of Kähler geometry, entropic time, and path independence (commutation properties of local deformations) yield manifestly covariant field dynamics and highlight the epistemic status of quantum states (Ipek et al., 2018, Ipek et al., 2019, Ipek et al., 2020).

ED’s principal limitation lies in the identification and physical justification of appropriate constraints: the structure and updating of these constraints dictate the resulting physical theory. The formalism recovers standard quantum mechanics only for choices that ensure conservation of a Hamiltonian functional with the correct Fisher-information (quantum) contribution (Caticha, 2017, Caticha, 2019). In the context of quantum gravity and dynamical geometry, ED predicts—and conceptually explains—conditional breakdowns of the superposition principle in fully coupled evolution (Ipek et al., 2020).

References: