Feynman’s path integrals as strictly hyperfinite sums
The path-integral formulation of quantum mechanics by Richard Feynman is physically very intuitive: a particle is not described by one single distinguished path, but by a superposition of possible paths through space and time. Mathematically, however, the continuous path integral is subtle. The formal expression contains a measure over an infinite-dimensional function space; an ordinary translation-invariant measure of the Lebesgue type does not exist there in naive form.
Established physics therefore works with limiting constructions, regularisations, discrete time slices, or with Wick rotation to the Euclidean theory. In the Euclidean formulation, many models can be treated rigorously by means of the Wiener measure and the Feynman-Kac formula.
In the discretised nonstandard space, this measure problem can be organised differently. The path integral is not regarded as a direct integral over an already given infinite-dimensional measure space, but as a hyperfinite sum, or as a hyperfinite product, over an extremely fine lattice. The formal expression thereby first becomes an exactly defined algebraic lattice quantity.
The macroscopically measurable part does not then arise by subsequently discarding hyperreal quantities. Rather, the hyperfinite quantity is normalised, expanded in the infinitesimal scales \(\tilde{\omega}\) and \(\tilde{\nu}\), and structurally deflated. The physical measured value is the regular scale-free part of this expansion.
The measure problem in the continuum
In the standard continuum, summation over all paths requires an integral over infinitely many degrees of freedom. The formal expression for the propagator is
\[ K(b, a) \propto \uparrow \mathcal{D}x \exp(\tilde{\hbar}\underline{S}). \]Here \(K(b,a)\) denotes the propagator, that is, the complex transition amplitude from the initial point \(a=(x_a,t_a)\) to the final point \(b=(x_b,t_b)\). It is the kernel of quantum-mechanical time evolution. Here \(S\) is the classical action, \(\hbar\) is the Planck constant, and \(\mathcal{D}x\) formally denotes the measure over all paths.
In this form, the expression is not an ordinary Lebesgue integration. In particular, there is no naive translation-invariant Lebesgue measure on the infinite-dimensional path space that possesses all the desired properties at once. For this reason, actual calculations must be made precise by means of discrete approximations, limiting procedures, oscillatory integrals, distributions, operator methods, or Euclidean methods.
The hyperfinite discretisation of space and time
In nonstandard space, the time interval between the initial and final points is subdivided into an exactly specified midfinite number \(\omega\) of time steps of infinitesimal duration \(\Delta t\). Space is likewise modelled by a hyperfine, but strictly discrete, lattice.
A single path \(x(t)\) is then no longer an arbitrary continuous functional object, but an exactly specified vector consisting of \(\omega\) spatial coordinates. The space of all lattice paths is extremely large, but hyperfinite. In this lattice formulation, the formal path measure \(\mathcal{D}x\) is replaced by a hyperfinite product of the individual lattice contributions:
\[ \mathcal{D}x \to {\LARGE{\textbf{$\times$}}}_{k=1}^{\omega} \Delta x_k. \]It is important here that, in physical applications, this product usually also has to be supplemented by normalisation factors depending on mass, time step, dimension, and the chosen discretisation. The product therefore describes the structural measure kernel, not yet the fully normalised propagator formula.
The exact algebraic sum
Because of the lattice structure, there is no immediate need first to construct an infinite-dimensional Lebesgue measure. The path integral becomes a hyperfinite sum over the discrete set of all lattice paths \(P\). The amplitude then has the schematic form
\[ K(b, a) \propto {\LARGE{\textbf{+}}}_{P} \exp(\tilde{\hbar}\underline{S}[P]). \]Since each summand is a complex phase factor of modulus \(1\), each individual contribution is exactly defined. The hyperfinite sum is therefore algebraically well-defined. Its physical evaluation nevertheless still requires a suitable normalisation and the determination of the regular part.
It is therefore more accurate to say: the hyperfinite path integral avoids the naive infinite-dimensional measure problem, but it does not automatically replace all analytical questions. Rather, these are translated into questions of normalisation, expansion in the scales \(\tilde{\omega}\) and \(\tilde{\nu}\), structural deflation, and stability of the macroscopic limiting transition.
A typical normalised hyperfinite quantity has the form
\[ X_{\omega,\nu} = P_{\omega,\nu} + R + I_{\omega,\nu}. \]Here \(P_{\omega,\nu}\) denotes the hyperreal pole or background part, \(R\) the regular scale-free part, and \(I_{\omega,\nu}\) the infinitesimal remainder, consisting of positive powers of \(\tilde{\omega}\) or \(\tilde{\nu}\). The macroscopic measured value is then not an externally formed standard part, but the regular part \(R\) isolated by deflation.
Stationary phase and structural cancellation
How does classical physics arise from this enormous hyperfinite sum? The decisive mechanism is stationary phase. For macroscopic systems one typically has \(S \gg \hbar\). The phase factor \(\exp(\tilde{\hbar}\underline{S})\) then oscillates very rapidly for most neighbouring paths. In the hyperfinite sum, these contributions largely cancel by destructive interference.
In the neighbourhood of those paths for which the action is stationary, by contrast, the phase changes only slowly between neighbouring paths. There the contributions add constructively. These stationary paths satisfy the principle of least action, or more precisely the principle of stationary action.
In this view, the sum deflates structurally: the strongly oscillating non-stationary paths form a background whose contributions cancel one another in the macroscopic approximation. The leading contribution comes from the neighbourhood of the stationary paths. It does not necessarily mark exactly one single classical path, but rather the classical trajectory together with its local fluctuations.
Quantum fluctuations as the regular remainder
The neighbouring paths around the stationary trajectory do not cancel completely. They form the regular remainder of the summation. This remainder contains the quantum corrections to classical motion, the spread around the classical path, and contributions that become essential, for example, in the tunnel effect.
On a hyperreal lattice, such remainder contributions can be formulated as hyperfinite sums over discrete paths. In suitable Euclidean formulations they are closely related to random-walk models, Brownian motion, and the Feynman-Kac formula.
The Schrödinger equation is not necessarily the starting point of this representation. Rather, it can be understood as a macroscopic equation equivalent to the correctly normalised path integral. In this sense, the path-integral formulation connects the local differential equation with a global summation over paths.
Conclusion of the derivation
Feynman’s path integrals need not be understood as naive integrals over an infinite-dimensional Lebesgue measure. Precisely this naive reading is mathematically problematic.
In nonstandard space, the path integral can be formulated as a hyperfinite algebraic summation over discrete lattice paths. This gives precision to Feynman’s original intuition: the quantum amplitude arises from the superposition of many path contributions, while the classical trajectory is singled out by stationary phase and destructive interference of the non-stationary paths.
Nonstandard analysis does not solve the measure problem by producing an ordinary infinite-dimensional Lebesgue measure. It shifts the basis of the calculation to a hyperfinite lattice, on which sums and products are first algebraically exactly defined.
The physical theory then arises through normalisation, structural deflation, and the proof that the macroscopic limiting transition yields the known equations of quantum mechanics. The macroscopically measurable quantity is not obtained by subsequently discarding hyperreal parts, but by expansion in the infinitesimal scales \(\tilde{\omega}\) and \(\tilde{\nu}\). The regular part is the scale-free coefficient of this expansion, after the hyperreal poles have been structurally separated off.
© 2026 by Boris Haase