Bilinear Path Integral in Wave Optics

• Bilinear path integral is a generalization of classical path integrals that computes interference effects using a double integral over paired paths.
• It enforces symmetry, non-negativity for self-pairs, and bounded cross terms to ensure physical consistency in wave-optical and stochastic analyses.
• The framework offers actionable insights into covariant discretization, quadratic response-field couplings, and unifies methods across optical and quantum applications.

to=arxiv.search 天天中彩票怎么json {"^^^^^^^^4query4 path integral^^^^^^^^4\4 OR ^^^^^^^^4\4 a path-integral calculus^^^^^^^^4\4 OR ^^^^^^^^4\4 Tracing: Generalizing The Path Integral To Wave Optics^^^^^^^^4\4 A bilinear path integral is, in the most explicit recent usage, a generalization of the classical path-space formulation in which the measured quantity is not an integral over individual paths but an integral over pairs of paths. In this form, the domain is PRESERVED_PLACEHOLDER_^^^^4query4^^^^ rather than PRESERVED_PLACEHOLDER_^^^^4\4^^^^, and the integrand is a mutual contribution function PRESERVED_PLACEHOLDER_^^^^4 OR \4^^^^ that encodes pairwise interference and coherence. The term is not standard across the broader path-integral literature: several closely related papers do not introduce a named object called a “bilinear path integral,” but they study quadratic or bilinear structures that are directly relevant, including quadratic actions, bilinear response-field couplings, products of kernels, and bilinear oscillatory integrals (&&&^^^^4query4^^^^&&&).

^^^^4\4^^^^. Terminological scope and principal meaning

In the classical path-space formulation of light transport, the measurement is written as

PRESERVED_PLACEHOLDER_^^^^4 OR \4^^^^

where $\vbar{x}\in\Omega$ is a path, $\mu$ is the path-space measure, and $f(\vbar{x})\ge 0$ is the measurement contribution function. The bilinear generalization replaces this linear integral by

$I = \int_{\Omega\cross\Omega} F\qty(\vbar{x},\vbar{y}) \dd{\mu\qty(\vbar{x})} \dd{\mu\qty(\vbar{y})},$

with $\Omega\times\Omega$ the space of path pairs and $F$ the mutual contribution function. This is the paper’s explicit definition of a bilinear path integral in wave-optical transport (&&&^^^^4query4^^^^&&&).

Outside this setting, the phrase is better understood descriptively than as a universally adopted formal term. One paper states that it does not explicitly study an object called a “bilinear path integral,” but is highly relevant because its main concern is determining which quadratic or bilinear manipulations are legitimate after discretization, especially for actions containing terms quadratic in the velocity or bilinear couplings between a response field and a trajectory variable (&&&^^^^4 OR \4^^^^&&&). Another paper likewise does not introduce a named object called a “bilinear path integral,” but its path-integral construction is built from Riemannian quadratic forms and their induced bilinear pairings, including the metric on the total space, the orbit metric, the mechanical connection, and a reduction Jacobian built from a quadratic form PRESERVED_PLACEHOLDER_^^^^4\4query4^^^^ (&&&^^^^4 OR \4^^^^&&&).

A common misconception is therefore that “bilinear path integral” denotes a single established formalism across quantum mechanics, stochastic analysis, and harmonic analysis. The literature represented here suggests a narrower conclusion: the phrase is explicit and foundational in the wave-optics path-pair formulation, while elsewhere it functions as an interpretive label for quadratic and bilinear structures embedded in more standard path-integral frameworks.

^^^^4 OR \4^^^^. Path-pair formulation in wave optics

The wave-optical motivation is that classical transport integrates contributions of individual paths, whereas measured power in wave optics depends on interference between paths. The path-pair formulation is therefore

PRESERVED_PLACEHOLDER_^^^^4\4\4^^^^

where PRESERVED_PLACEHOLDER_^^^^4\4 OR \4^^^^ is the mutual contribution function. The formulation is “bilinear” because transport is expressed in terms of pairwise interactions between two path contributions rather than the contribution of a single path (&&&^^^^4query4^^^^&&&).

The paper imposes three properties on PRESERVED_PLACEHOLDER_^^^^4\4 OR \4^^^^: PRESERVED_PLACEHOLDER_^^^^4\44^^^^

PRESERVED_PLACEHOLDER_^^^^4\4^^^^5

and

PRESERVED_PLACEHOLDER_^^^^4\4^^^^6

These conditions allow cross terms with PRESERVED_PLACEHOLDER_^^^^4\4^^^^7 to be negative while keeping the total measurement PRESERVED_PLACEHOLDER_^^^^4\4^^^^8 nonnegative, as required for power (&&&^^^^4query4^^^^&&&).

The physical interpretation is given in terms of statistical optics. If PRESERVED_PLACEHOLDER_^^^^4\4^^^^9 denotes a realization of a statistical wave ensemble and PRESERVED_PLACEHOLDER_^^^^4 OR \4query4^^^^ the field strength transported over path PRESERVED_PLACEHOLDER_^^^^4 OR \4\4^^^^, then

PRESERVED_PLACEHOLDER_^^^^4 OR \4 OR \4^^^^

Accordingly,

PRESERVED_PLACEHOLDER_^^^^4 OR \4 OR \4^^^^

This makes the bilinear structure a path-space expression of the fact that intensity is quadratic in the field and that coherence theory is inherently second-order (&&&^^^^4query4^^^^&&&).

The paper emphasizes two limiting cases. In the perfectly incoherent limit,

PRESERVED_PLACEHOLDER_^^^^4 OR \44^^^^

so only self-pairs contribute and the bilinear form collapses to the classical path integral. In the fully coherent case, the paper gives an example of the form

PRESERVED_PLACEHOLDER_^^^^4 OR \4^^^^5

with the intended structure a relative phase based on path-length difference. The interference content is made explicit by

PRESERVED_PLACEHOLDER_^^^^4 OR \4^^^^6

so the self-terms are supplemented by an interference correction PRESERVED_PLACEHOLDER_^^^^4 OR \4^^^^7 (&&&^^^^4query4^^^^&&&).

This framework also serves as a unifying interpretation for methods with phase-carrying rays, shooting-bouncing rays, UTD-based diffraction, and partially coherent models based on Wigner distributions or mutual coherence. The formulation is therefore not merely an alternative notation for classical transport, but a statement that in wave optics the fundamental object is the mutual contribution of path pairs rather than an independently meaningful contribution of each path.

^^^^4 OR \4^^^^. Quadratic actions, bilinear cross terms, and covariant calculus

A different but closely related sense of “bilinear path integral” arises in the calculus of stochastic path integrals. For multiplicative-noise Langevin dynamics with one degree of freedom,

PRESERVED_PLACEHOLDER_^^^^4 OR \4^^^^8

the central issue is not only the Itō-versus-Stratonovich ambiguity at the Langevin-equation level, but the failure of naive nonlinear changes of variables inside the path-integral action (&&&^^^^4 OR \4^^^^&&&).

In Stratonovich discretization, one has

PRESERVED_PLACEHOLDER_^^^^4 OR \4^^^^9

which is sufficient at the stochastic differential equation level. However, the Onsager–Machlup action contains the quadratic term

PRESERVED_PLACEHOLDER_^^^^4 OR \4query4^^^^

If one substitutes

PRESERVED_PLACEHOLDER_^^^^4 OR \4\4^^^^

then the cross term between the leading piece PRESERVED_PLACEHOLDER_^^^^4 OR \4 OR \4^^^^ and the neglected correction is

PRESERVED_PLACEHOLDER_^^^^4 OR \4 OR \4^^^^

which contributes at the same order as the action itself. The consequence is that subleading terms that can be discarded in the Langevin equation cannot be discarded inside a quadratic action, because bilinear cross terms promote them to leading relevance (&&&^^^^4 OR \4^^^^&&&).

The proposed cure is a covariant discretization. Time is sliced as

PRESERVED_PLACEHOLDER_^^^^4 OR \44^^^^

with PRESERVED_PLACEHOLDER_^^^^4 OR \4^^^^5, and the evaluation point is changed from the midpoint prescription to

PRESERVED_PLACEHOLDER_^^^^4 OR \4^^^^6

where

PRESERVED_PLACEHOLDER_^^^^4 OR \4^^^^7

Because PRESERVED_PLACEHOLDER_^^^^4 OR \4^^^^8, the correction is PRESERVED_PLACEHOLDER_^^^^4 OR \4^^^^9, exactly the order required to repair the calculus. The paper derives an exact all-orders covariant discretization in operator form and emphasizes that for path-integral construction the truncated correction

$\vbar{x}\in\Omega$^^^^4query4^^^^

is sufficient (&&&^^^^4 OR \4^^^^&&&).

This analysis is directly relevant to bilinear manipulations in path integrals because it identifies the precise reason that quadratic velocity terms and bilinear response-field couplings become noncovariant under naive variable changes. A plausible implication is that any formalism built from Gaussian rewritings, Hubbard–Stratonovich-type representations, or response-field bilinearizations must be controlled at the discretized level rather than by continuous-time chain-rule manipulations alone.

4. Quadratic versus non-quadratic time-derivative dependence

A second major context is the distinction between path integrals derived from actions quadratic in time derivatives and those derived from non-quadratic actions. For scalar fields, the canonical momentum is defined by

$\vbar{x}\in\Omega$^^^^4\4^^^^

and when this relation can be inverted to obtain $\vbar{x}\in\Omega$^^^^4 OR \4^^^^, the Hamiltonian density is

$\vbar{x}\in\Omega$^^^^4 OR \4^^^^

When the Lagrangian is quadratic in the time derivatives, for example $\vbar{x}\in\Omega$4, the momentum-velocity relation is linear and the Legendre transformation is straightforward. This is the standard case behind the usual path integral (&&&^^^^4\4\4^^^^&&&).

The conventional phase-space expression is

$\vbar{x}\in\Omega$5

which reduces, after integrating over momenta, to

$\vbar{x}\in\Omega$6

For the free scalar field,

$\vbar{x}\in\Omega$7

and the formalism reduces exactly to the familiar Gaussian case (&&&^^^^4\4\4^^^^&&&).

When $\vbar{x}\in\Omega$8 depends on $\vbar{x}\in\Omega$9 non-quadratically, the Legendre transform may be intractable. The generalized construction introduces an auxiliary velocity-like field $\mu$^^^^4query4^^^^ and yields

$\mu$^^^^4\4^^^^

The determinant may, when positive, be represented by Grassmann variables: $\mu$^^^^4 OR \4^^^^ The paper emphasizes that the price of avoiding the explicit Legendre transform is doubling of variables via the auxiliary $\mu$^^^^4 OR \4^^^^, a Jacobian determinant, and potentially “hidden fermionic variables” representing that determinant (&&&^^^^4\4\4^^^^&&&).

In relation to the bilinear theme, the key point is that the familiar path integral occupies the special case in which the kinetic structure is quadratic and the auxiliary-field integral becomes Gaussian. The generalized formalism is therefore a strict extension of the standard quadratic or bilinear setting, not a replacement for it.

5. Geometric, kernel, and oscillatory-integral analogues

Several additional constructions are adjacent to the bilinear path-integral idea because they are organized by quadratic forms or products of amplitudes rather than by a single-path scalar density.

For Wiener path integrals on a manifold with symmetry, the configuration space is

$\mu$4

with product-type Riemannian metric

$\mu$5

A compact semisimple unimodular Lie group acts freely, properly, and isometrically by

$\mu$6

The resulting reduction procedure leads to an integral relation between path integrals on the initial and reduced manifolds, and for reduction onto the zero-momentum level the reduction Jacobian is obtained as an additional potential term to the Hamiltonian. The paper is relevant here because the entire construction is built from Riemannian quadratic forms and induced bilinear pairings (&&&^^^^4 OR \4^^^^&&&).

A related kernel-level structure appears in the “hit function” transform of the Euclidean quantum-mechanical propagator

$\mu$7

The unnormalized hit function

$\mu$8

can be rewritten as

$\mu$9

This is a time average of a product of two kernels, obtained by forcing the path to pass through an intermediate point $f(\vbar{x})\ge 0$^^^^4query4^^^^ and then using path-integral composition. The paper does not call this bilinear, but it is the clearest product-of-kernels structure in the material (&&&^^^^4\4^^^^5&&&).

In harmonic analysis, the phrase “path integral” becomes metaphorical, but the bilinear structure is explicit. One paper studies the bilinear iterated Fourier inversion integral

$f(\vbar{x})\ge 0$^^^^4\4^^^^

and its variational truncation operator

$f(\vbar{x})\ge 0$^^^^4 OR \4^^^^

The ordered frequency simplex $f(\vbar{x})\ge 0$^^^^4 OR \4^^^^ makes the operator an iterated integral over a path of truncation parameters in frequency space (&&&^^^^4\4^^^^6&&&).

Another paper proves a bilinear $f(\vbar{x})\ge 0$4 estimate for products of two oscillatory integral operators

$f(\vbar{x})\ge 0$5

with the product expanding to

$f(\vbar{x})\ge 0$6

This is not a Feynman path integral, but it is a bilinear theory of oscillatory phase integrals closely tied to semiclassical propagation (&&&^^^^4\4^^^^7&&&).

These examples support a broad usage in which “bilinear path integral” designates either a literal double integral over path space or a path-integral construction whose fundamental algebraic objects are quadratic forms, bilinear couplings, or products of propagators.

6. Rigorous status and computational consequences

The mathematical and computational consequences of bilinearization are substantial. In the wave-optics path-pair formulation, the domain becomes $f(\vbar{x})\ge 0$7, the integrand is no longer nonnegative, and useful contribution depends on relative phase and coherence between complete paths. The paper states that this makes local path sampling fundamentally difficult: even if one path is sampled well according to local throughput, a second path may later annihilate it, so sampling becomes a global problem (&&&^^^^4query4^^^^&&&).

This difficulty motivates a weakly-local reformulation based on region-to-region transport: $f(\vbar{x})\ge 0$8 where $f(\vbar{x})\ge 0$9 is a sequence of bounded spatial regions and

$I = \int_{\Omega\cross\Omega} F\qty(\vbar{x},\vbar{y}) \dd{\mu\qty(\vbar{x})} \dd{\mu\qty(\vbar{y})},$^^^^4query4^^^^

The aim is to internalize interference inside local region operators so that global Monte Carlo sampling can again proceed over single region paths rather than pairs of point paths (&&&^^^^4query4^^^^&&&).

The local interference structure is written as

$I = \int_{\Omega\cross\Omega} F\qty(\vbar{x},\vbar{y}) \dd{\mu\qty(\vbar{x})} \dd{\mu\qty(\vbar{y})},$^^^^4\4^^^^

and the candidate scattered-beam power as

$I = \int_{\Omega\cross\Omega} F\qty(\vbar{x},\vbar{y}) \dd{\mu\qty(\vbar{x})} \dd{\mu\qty(\vbar{y})},$^^^^4 OR \4^^^^

This does not eliminate bilinear interference; it relocates it from global path-pair recombination to local interfering sub-interactions (&&&^^^^4query4^^^^&&&).

At a more foundational level, rigorous path integration remains delicate even before bilinearization. The Feynman path integral has been surveyed as a Henstock integral over

$I = \int_{\Omega\cross\Omega} F\qty(\vbar{x},\vbar{y}) \dd{\mu\qty(\vbar{x})} \dd{\mu\qty(\vbar{y})},$^^^^4 OR \4^^^^

with the key point that the absolute value of Feynman’s integrand is not integrable, so Lebesgue integration theory cannot be used in the usual way. The oscillatory integral may exist as a Henstock integral, but the survey also discusses the impossibility of justifying the exchange of the path integral with an infinite series in the way required for diagrammatic expansions (&&&^^^^4 OR \4\4^^^^&&&). This suggests that bilinear and path-pair generalizations inherit, and may amplify, already difficult issues of oscillation, cancellation, and limiting procedures.

Taken together, these results show that the bilinear path integral is best understood as a family of closely connected ideas rather than a single universally fixed object. In its most explicit modern form, it is a double integral over path-pair space designed to represent wave interference. In stochastic and quantum settings, it identifies the regime in which quadratic actions, bilinear couplings, Jacobians, and product kernels become decisive. In all of these settings, the central theme is the same: once the observable is quadratic in amplitudes or fields, the natural path-space description is no longer purely linear in individual paths.