An intrinsic measure on path space for Feynman integration: Feynman's 1948 requirements after Cameron's no-go theorem

In his 1948 formulation of non-relativistic quantum mechanics as a sum over histories, Feynman listed as central open problems the lack of an intrinsic measure on the space of paths $x(t)$ — the formal weight $e^{iS[x]/\hbar}$ is given meaning only through a time-slicing limit — together with the absence of manifest unitary and canonical invariance, of a direct prescription for evaluating observables on classes of paths, and of genuinely path-based treatments of spin, relativity, and actions nonlocal in time. Cameron's 1960 no-go theorem settled the most literal version of the measure question negatively: the natural complex set function on cylinder sets of path space is only finitely additive and admits no σ-additive extension. The subsequent literature — oscillatory-integral and analytic-continuation theories, white-noise (Hida) distributions, rigorous time-slicing, and Wiener-measure constructions — responds by replacing the measure with a linear functional, thereby abandoning σ-additivity, intrinsicness, or manifest invariance. The open core is precise: construct, or prove impossible, an integration theory on a space of paths that is defined intrinsically (not via time-slicing limits or analytic continuation from imaginary time), genuinely measure-like after Cameron's theorem, manifestly invariant under unitary/canonical (and, in the relativistic extension, Lorentz) transformations, and capable of direct path-class evaluation of general observables such as momentum. A strengthened impossibility theorem would equally resolve the question, establishing that the existing linear-functional replacements are the only possible route.

Background

Feynman's 1948 formulation of non-relativistic quantum mechanics expresses the transition amplitude between position $y$ at time $0$ and position $x$ at time $t$ as a sum over histories: formally, an integral over all paths connecting them, each weighted by $\exp(iS[x]/\hbar)$, where $S[x]=\int_0^t L(x,\dot{x},\tau)\,d\tau$ is the classical action and $L$ the Lagrangian. Feynman gave this expression meaning only operationally: subdivide $[0,t]$ into $N$ time slices, insert a complete set of position states at each slice, and pass to a limit. In Section 11 of the paper he lists this dependence on subdivision as a serious drawback and states the need for an appropriate measure for the space of the argument functions $x(t)$. He further notes that the formulation is not manifestly invariant under unitary transformations of the state space or canonical transformations of the coordinates, and that quantities such as momentum can be extracted only indirectly, through limiting sequences of position measurements, rather than by a direct prescription for paths of class $R$ — specified classes of paths to which amplitudes would be assigned directly. His closing remarks flag the path-based derivations for spin and relativity as purely formal, and Section 12 observes that for actions nonlocal in time (involving, for example, $\int x(t)x(t+T)\,dt$) no wave function is available to describe a state.

The stochastic counterpart makes the contrast sharp. For the heat equation, Kac (1949) showed that the analogous sum over histories is a rigorous integral against the Wiener measure — a genuine probability measure, hence σ-additive (countably additive on a σ-algebra of subsets of path space) — on the space of continuous paths. For the Schrödinger equation the amplitude oscillates rather than decays, and the naive analogue fails: define, on the algebra of cylinder sets (sets of paths constrained at only finitely many times), the complex set function obtained by multiplying free Schrödinger kernels; Cameron (1960) proved that this set function is only finitely additive and admits no σ-additive extension to the σ-algebra generated by the cylinder sets. Consequently the modern literature defines the Feynman integral not as a measure integral but as a linear functional of one of several kinds: an oscillatory integral — an integral with a rapidly oscillating complex phase $e^{i\Phi(x)}$, defined as a distributional pairing or a damped limiting value rather than an ordinary Lebesgue integral — obtained by analytic continuation from the Wiener integral (Albeverio and Høegh-Krohn 1976); a Hida distribution, meaning a distribution on the white-noise (Gaussian) space underlying the Wiener process (de Faria, Potthoff and Streit 1991); a limit of rigorously controlled time-slicing approximations (Fujiwara 2017); or a genuine but positive, imaginary-time Wiener-type measure reached by stochastic quantization — representing the quantum dynamics by an auxiliary diffusion process in an imaginary-time parameter — and change of variables (Daubechies and Klauder 1982; Klauder and Daubechies 1984; Klauder 2011). Cartier and DeWitt-Morette (1995) achieve canonical invariance within a distributional half-density framework, a half-density being an object that transforms with the square root of the coordinate Jacobian, so that integrals against it are automatically coordinate-invariant. In a different direction, Sorkin (1994) recast the sum over histories as a generalized measure theory in which weights $\mu$ on sets of histories satisfy the interference-aware grade-2 sum rule — for mutually disjoint sets $A$, $B$, $C$ of histories, $\mu(A\cup B\cup C)=\mu(A\cup B)+\mu(A\cup C)+\mu(B\cup C)-\mu(A)-\mu(B)-\mu(C)$ — rather than ordinary countable additivity. On the relativistic side, Feynman (1950) described Klein-Gordon particles (particles obeying the relativistic spin-zero wave equation) by a path integral over an extra parameter analogous to proper time (the Lorentz-invariant time measured along a worldline); Struckmeier (2024) and Varshovi (2022) are recent unrefereed proof-of-principle attempts at relativistic path-integral measures, neither of which supplies the intrinsic, measure-like object of requirements (i)-(ii): Struckmeier obtains relativistic covariance within a parametrized extended-Lagrangian formalism, by adding an evolution parameter rather than by a manifestly invariant measure, and Varshovi builds his “Wiener fractal measure” out of the Wiener (imaginary-time) stochastic process, i.e. by the analytic-continuation route that (i) excludes. For actions nonlocal in time, Feynman and Vernon (1963) developed influence functionals — generally nonlocal-in-time effective actions for a system coupled to an environment — which handle the open-system case.

Problem Statement

Construct, or prove impossible, an integration theory for the non-relativistic sum over histories that simultaneously satisfies Feynman's 1948 requirements:

(i) Intrinsicness: the integration object is defined directly on a space of paths (with its σ-algebra or an explicitly stated additivity structure), not as the limit of time-sliced approximations, and not by analytic continuation or distributional extension from a Wiener-process (imaginary-time) measure.

(ii) Measure character: on its domain the object is genuinely measure-like — σ-additive, or carrying an explicitly axiomatized additivity that either is itself σ-additive on its domain or is accompanied by a proof that it yields the same transition amplitudes as σ-additivity would — so that Cameron's theorem, which rules out σ-additive extension of the cylinder-set complex set function on $(\mathbb{R}^d)^{[0,t]}$ (the set of all paths mapping $[0,t]$ into $\mathbb{R}^d$), is circumvented by an explicitly identified change of path space, σ-algebra, or additivity axiom, not by silent weakening. An explicitly weakened additivity axiom — for example Sorkin’s grade-2 interference rule — is admissible under (ii) only in this second, proof-carrying case.

(iii) Manifest invariance: the construction is manifestly invariant under unitary transformations of the state space and canonical transformations of phase space; in the relativistic extension, under Lorentz transformations.

(iv) Direct observable prescriptions: the theory assigns values (amplitudes or their generalizations) to classes of paths directly, yielding in particular a prescription for momentum-type observables that does not proceed through auxiliary limiting sequences of position measurements.

The question is resolved by an explicit construction meeting (i)-(iv) that reproduces the Schrödinger propagator at least for the free particle and for one interacting system whose propagator is independently known, by a strengthened impossibility theorem showing that requirements (i)-(iv), or a precisely stated subset thereof, are jointly unsatisfiable, which would prove the existing linear-functional replacements are forced, or by a demonstration, checked against the primary texts, that an already-published framework satisfies (i)-(iv) as stated.

References

1. Feynman, R. P. (1948). Space-Time Approach to Non-Relativistic Quantum Mechanics. Reviews of Modern Physics 20, 367-387. doi:10.1103/RevModPhys.20.367. https://doi.org/10.1103/RevModPhys.20.367
2. Kac, M. (1949). On distributions of certain Wiener functionals. Transactions of the American Mathematical Society 65, 1-13. doi:10.1090/s0002-9947-1949-0027960-x. https://doi.org/10.1090/s0002-9947-1949-0027960-x
3. Cameron, R. H. (1960). A Family of Integrals Serving to Connect the Wiener and Feynman Integrals. Journal of Mathematics and Physics 39, 126-140. doi:10.1002/sapm1960391126. https://doi.org/10.1002/sapm1960391126
4. Albeverio, Sergio A.; Høegh-Krohn, Raphael J. (1976). Mathematical Theory of Feynman Path Integrals. Lecture Notes in Mathematics 523, Springer. doi:10.1007/bfb0079827. https://doi.org/10.1007/bfb0079827
5. de Faria, M.; Potthoff, J.; Streit, L. (1991). The Feynman integrand as a Hida distribution. Journal of Mathematical Physics 32, 2123-2127. doi:10.1063/1.529184. https://doi.org/10.1063/1.529184
6. Fujiwara, Daisuke (2017). Rigorous Time Slicing Approach to Feynman Path Integrals. Mathematical Physics Studies, Springer. doi:10.1007/978-4-431-56553-6. https://doi.org/10.1007/978-4-431-56553-6
7. Klauder, John R. (2011). A Modern Approach to Functional Integration. Applied and Numerical Harmonic Analysis, Birkhäuser/Springer. doi:10.1007/978-0-8176-4791-9. https://doi.org/10.1007/978-0-8176-4791-9
8. Daubechies, Ingrid; Klauder, John R. (1982). Constructing measures for path integrals. Journal of Mathematical Physics 23, 1806-1822. doi:10.1063/1.525234. https://doi.org/10.1063/1.525234
9. Klauder, John R.; Daubechies, Ingrid (1984). Quantum Mechanical Path Integrals with Wiener Measures for all Polynomial Hamiltonians. Physical Review Letters 52, 1161-1164. doi:10.1103/PhysRevLett.52.1161. https://doi.org/10.1103/PhysRevLett.52.1161
10. Cartier, Pierre; DeWitt-Morette, Cécile (1995). A new perspective on functional integration. Journal of Mathematical Physics 36, 2237-2312. doi:10.1063/1.531039. https://doi.org/10.1063/1.531039
11. Sorkin, Rafael D. (1994). Quantum Mechanics as Quantum Measure Theory. Modern Physics Letters A 9, 3119-3127. doi:10.1142/S021773239400294X. arXiv:gr-qc/9401003. https://arxiv.org/abs/gr-qc/9401003
12. Feynman, R. P. (1950). Mathematical Formulation of the Quantum Theory of Electromagnetic Interaction. Physical Review 80, 440-457. doi:10.1103/PhysRev.80.440. https://doi.org/10.1103/PhysRev.80.440
13. Feynman, R. P.; Vernon, F. L. (1963). The theory of a general quantum system interacting with a linear dissipative system. Annals of Physics 24, 118-173. doi:10.1016/0003-4916(63)90068-x. https://doi.org/10.1016/0003-4916(63)90068-x
14. Drago, N.; Mazzucchi, S.; Moretti, V. (2025). Feynman path integrals on compact Lie groups with bi-invariant Riemannian metrics. Annales Henri Poincaré. doi:10.1007/s00023-025-01625-4. arXiv:2307.03282. https://doi.org/10.1007/s00023-025-01625-4
15. Struckmeier, Jürgen (2024). Relativistic generalization of Feynman's path integral on the basis of extended Lagrangians. arXiv:2406.06530. https://arxiv.org/abs/2406.06530
16. Varshovi, A. A. (2022). Brownian Motion in the Hilbert Space of Quantum States and the Stochastically Emergent Lorentz Symmetry: A Fractal Geometric Approach from Wiener Process to Formulating Feynman's Path-Integral Measure for Relativistic Quantum Fields. arXiv:2201.09855. https://arxiv.org/abs/2201.09855