Finster and Tolksdorf (J. Math. Phys. 53, 052305 (2012)) showed that perturbative solutions of the classical ϕ4\phi^4 field equation with a Lorentz-invariant Gaussian stochastic background field generate bosonic loop diagrams whose weights can match individual quantum amplitudes — but only with topology-dependent values of the stochastic-strength parameter β\beta (β=1/(2π)\beta=1/(2\pi) for the tadpole and the simple loop, β=1/π\beta=1/\pi for the triangle loop), so no single Gaussian ensemble reproduces all quantum loop combinatorics in ϕ4\phi^4 theory. The paper leaves open a connected research program: (i) a full combinatorial classification of which theories and which (Gaussian or non-Gaussian) stochastic ensembles on classical solution space make all classical-loop amplitudes coincide with the quantum ones; (ii) whether the mismatches disappear in theories without parallel identical internal lines between the same vertex pair, as conjectured for QED; (iii) incorporation of fermions; (iv) non-Gaussian ensembles arising from pre-interaction nonlinear dynamics; (v) a conditional constructive program toward a rigorous non-perturbative interacting Lorentzian quantum field theory (UV regularization, uniform global existence for random initial data, regulator removal, non-perturbative renormalization); (vi) the physical meaning of β\beta and its relation to ℏ\hbar; and (vii) unitary completion of the classical scattering operator. The classification, the QED conjecture, the fermionic extension, and the constructive program remain open.

Background

In perturbative quantum field theory, scattering amplitudes are organized as Feynman diagrams: tree diagrams represent the classical nonlinear response of the field, while loop diagrams — diagrams containing closed internal lines, such as the tadpole (a single line leaving and re-entering the same vertex) — encode genuinely quantum effects and combinatorial factors. A long-standing question is how much of this structure can be reproduced by purely classical dynamics with randomness added by hand, a line of thought that goes back to stochastic electrodynamics (the program of deriving quantum phenomena from classical electrodynamics supplemented by a random zero-point radiation background) and related classical statistical models of quantum theory.

Finster and Tolksdorf (J. Math. Phys. 53, 052305 (2012); arXiv:1201.5497) made this question concrete for the interacting scalar theory with quartic self-coupling (ϕ4\phi^4 theory). They solve the classical field equation perturbatively around a background field that is a random variable: the background is drawn from a Gaussian probability distribution (a 'stochastic ensemble') on the space of classical solutions. Requiring Lorentz invariance forces the ensemble's covariance (two-point correlation function) to be C=(1/β)P0C=(1/\beta)P_0, where P0P_0 is the real-valued on-shell fundamental solution of the free wave equation — the symmetric combination of positive- and negative-frequency modes, given by the Fourier transform of δ(k2)\delta(k^2) — and β>0\beta>0 is a single real parameter, the stochastic strength (Prop. 4.2 of the paper). Averaging the perturbative classical solution over the ensemble generates loop diagrams: each closed line of the corresponding Feynman graph is traced back to one power of the Gaussian contraction. The inner lines of the classical n-point functions carry the real causal Green's function S0S_0 (the principal-value Green's function of the Klein–Gordon operator), while the quantum lines carry the complex Feynman propagator ΔF=S0−iπP0\Delta_F=S_0-i\pi P_0, which combines both distributions; the paper's Lemma 4.3, a contour-integral identity, is the technical bridge that allows an order-by-order comparison of classical and quantum loop diagrams.

Their central computation is a mixed result. For the value β=1/(2π)\beta=1/(2\pi) the tadpole contribution and the one-loop self-energy diagram exactly reproduce the corresponding quantum amplitudes (eq. 4.20), and with β=1/π\beta=1/\pi the mean energy ω/(2πβ)\omega/(2\pi\beta) of each field mode equals the quantum zero-point energy ℏω/2\hbar\omega/2 (Section 4.4). But the triangle loop agrees with the quantum amplitude only for β=1/π\beta=1/\pi (eq. 4.21), so within Lorentz-invariant Gaussian ensembles no single β\beta makes all loop topologies match; the authors note explicitly that longer loops carry twice the weight of a tadpole or simple loop, and that the higher-loop combinatorics are beyond the paper's scope. They observe that the obstruction is tied to diagrams with several parallel identical internal lines connecting the same pair of vertices, and conjecture that theories without such lines — QED is the proposed example — might admit full agreement. The paper's final section (Section 5, 'Outlook') frames this as a program. Besides the combinatorial classification and the QED test, it asks how fermionic degrees of freedom could enter a classical or hybrid classical–quantum stochastic framework (Pauli exclusion and the spin-statistics connection have no obvious classical analogue; the suggested routes are coupling the classical bosonic field to second-quantized fermions — fermions treated in the many-particle Fock-space formalism — or using the fermionic projector, a framework developed by Finster that reconstructs quantum theory from a system of occupied Dirac wave functions); whether non-Gaussian probability measures on classical solutions — heuristically generated by nonlinear dynamics before the scattering region — could cure the combinatorial mismatch or yield other quantum features; whether the classical scattering operator, which fails to be unitary because it is built only from connected diagrams (Section 3.6), admits a unitary completion; and, most ambitiously, a conditional constructive route to a rigorous non-perturbative interacting Lorentzian quantum field theory: introduce an ultraviolet (UV) regularization of the ensemble's covariance (a short-distance cutoff in momentum space), prove global existence of classical solutions uniformly in the regularization, integrate over the random initial data to define n-point functions, and control the removal of the regulator (a 'non-perturbative renormalization').

Problem Statement

The open problem is the exact-equivalence question of the Finster–Tolksdorf stochastic-classical framework, together with the dependent program the paper attaches to it. Core question: determine, by full combinatorial classification or proof, for which interacting field theories and which probability measures on the space of classical solutions (Gaussian measures with parameter β\beta, or non-Gaussian measures arising from pre-interaction nonlinear dynamics) the loop expansion of the perturbatively solved classical field equation coincides exactly with the corresponding quantum loop amplitudes — with the known obstruction that in ϕ4\phi^4 theory different loop topologies require different β\beta, and the concrete test case of the paper's conjecture that theories without parallel identical internal lines between the same pair of vertices, such as QED, admit full agreement. The same program further asks: how fermionic degrees of freedom (Pauli exclusion, spin-statistics) can be incorporated, via a classical bosonic field coupled to second-quantized fermions or the fermionic projector; whether the classical scattering operator, non-unitary because it is built only from connected diagrams, admits a unitary completion; what the physical meaning and scaling of β\beta is and whether a single natural choice can be motivated beyond perturbative matching of selected amplitudes; and, conditionally on exact diagram agreement, whether the construction can be made rigorous and non-perturbative — UV-regularize the ensemble covariance, prove global existence of classical solutions uniformly in the regularization for random initial data, define n-point functions by integrating over the initial data, and control removal of the regulator. These open questions form one connected program — establishing (and exploiting) exact classical–quantum equivalence at loop level — in which the core classification question is the load-bearing part on which the constructive payoff depends. Completion standard: the core classification is the mandatory deliverable — a proof or counterexample deciding, for which interacting field theories and which probability measures on classical solution space the classical loop expansion coincides exactly with the corresponding quantum loop amplitudes; the conjecture for theories without parallel identical internal lines between the same pair of vertices (QED, its proposed example) is the concrete test case of that classification, not a substitute for it. The remaining items (fermionic incorporation, unitary completion of the classical scattering operator, the interpretation of β\beta, and the conditional constructive program) are dependent and optional: resolving any of them is a valid contribution, but none is required to resolve the task.

References

  1. Felix Finster and Jürgen Tolksdorf, 'Bosonic Loop Diagrams as Perturbative Solutions of the Classical Field Equations in ϕ4\phi^4-Theory', Journal of Mathematical Physics 53, 052305 (2012), DOI 10.1063/1.4718333, arXiv:1201.5497, https://doi.org/10.1063/1.4718333, https://arxiv.org/abs/1201.5497
  2. Felix Finster, 'Perturbative Quantum Field Theory in the Framework of the Fermionic Projector', Journal of Mathematical Physics 55, 042301 (2014), DOI 10.1063/1.4871549, arXiv:1310.4121, https://doi.org/10.1063/1.4871549, https://arxiv.org/abs/1310.4121
  3. Claudio Dappiaggi, Felix Finster, Niky Kamran, and Moritz Reintjes, 'Holographic Mixing and Fock Space Dynamics of Causal Fermion Systems', Annales Henri Poincaré 27, 1885–1969 (2025), DOI 10.1007/s00023-025-01593-9, arXiv:2410.18045, https://doi.org/10.1007/s00023-025-01593-9, https://arxiv.org/abs/2410.18045
  4. Felix Finster, 'The Continuum Limit of Causal Fermion Systems: From Planck Scale Structures to Macroscopic Physics', Fundamental Theories of Physics vol. 186, Springer (2016), DOI 10.1007/978-3-319-42067-7, arXiv:1605.04742, https://doi.org/10.1007/978-3-319-42067-7, https://arxiv.org/abs/1605.04742
  5. Frédéric Hélein, 'First integrals for nonlinear dispersive equations', Transactions of the American Mathematical Society 368, 6939–6978 (2015), DOI 10.1090/tran/6573, arXiv:1311.0722, https://doi.org/10.1090/tran/6573, https://arxiv.org/abs/1311.0722