Worldline Formalism in Quantum Field Theory

• Worldline formalism is a first-quantized approach that reformulates quantum field theory by expressing propagators and effective actions as path integrals over particle trajectories.
• It unifies open and closed trajectories, enabling compact master formulas that absorb many Feynman diagram orderings and naturally extend to spins, colors, and curved spaces.
• The method provides practical tools for computing one-loop effective actions, heat kernels, and scattering amplitudes under various boundary and external field conditions.

The worldline formalism is a first-quantized reformulation of perturbative quantum field theory in which propagators, effective actions, heat kernels, and scattering amplitudes are written as path integrals over relativistic particle trajectories rather than as sums of second-quantized Feynman diagrams. Open worldlines represent propagation between fixed endpoints, closed worldlines represent one-loop traces and effective actions, and external particles are inserted by vertex operators along the line or loop. In this representation the Schwinger proper time $T$ is explicit, many diagram orderings are absorbed into a single parameter integral, and the same framework extends to spin, color, curved space, boundaries, external fields, and several semiclassical or numerical constructions (Edwards et al., 2019, Edwards et al., 2022, Schubert, 2023).

1. First-quantized basis and proper-time representation

At its most basic level, the formalism starts from Schwinger’s proper-time representation of inverse operators and functional determinants. For a scalar particle in a background gauge field $A_\mu$, the propagator can be written as

$$D^{xx'}[A] = \left\langle x' \left| \int_0^\infty dT\, e^{-T\left[-(\partial+ieA)^2+m^2\right]} \right| x \right\rangle = \int_0^\infty dT\, e^{-m^2T} \int_{x(0)=x}^{x(T)=x'} \mathcal D x(\tau)\, e^{-\int_0^T d\tau \left(\frac14 \dot x^2 + ie\,\dot x\cdot A(x(\tau))\right)} .$$

For a closed worldline, the corresponding one-loop scalar effective action is

$$\Gamma_{\rm scal}[A] = \int_0^\infty \frac{dT}{T}\, e^{-m^2T} \int_{x(0)=x(T)} \mathcal D x(\tau)\, e^{-\int_0^T d\tau \left(\frac14 \dot x^2 + ie\,\dot x\cdot A(x(\tau))\right)} .$$

The same proper-time logic underlies the heat-kernel representation of one-loop determinants in curved or bounded settings, where $\mathrm{Tr}\,e^{-T(-\triangle)}$ or $\mathrm{Tr}\,e^{-T(-\slashed D^2)}$ becomes the transition amplitude of an auxiliary particle evolving for proper time $T$ (Edwards et al., 2022, Corradini et al., 2019, Manzo, 2024).

This representation makes the distinction between open and closed worldlines structural rather than incidental. Open lines are the natural language for propagators and dressed external legs; closed loops are the natural language for trace logs and one-loop amplitudes. On closed loops one removes the translational zero mode of the kinetic operator, while on open lines Dirichlet-type endpoint conditions make the quadratic operator invertible without zero-mode subtleties. The proper-time measure also produces the standard free determinant factor $(4\pi T)^{-D/2}$, which is the particle analogue of a one-loop Gaussian determinant (Edwards et al., 2022, Edwards et al., 2019).

A central consequence is that the formalism reorganizes perturbation theory before any explicit integration is done. Instead of assigning an independent loop momentum to each diagram, one integrates over a single trajectory and over the insertion positions of external quanta. This is the point-particle counterpart of the string-inspired viewpoint developed in the Bern–Kosower and Strassler tradition, and it is the source of the compact master formulas that distinguish worldline calculations from ordinary graph-by-graph perturbation theory (Edwards et al., 2022, Edwards et al., 2019).

2. Green functions, vertex operators, and master formulas

The Gaussian evaluation of worldline path integrals is controlled by one-dimensional Green functions. For a closed bosonic loop with zero mode removed, the bosonic Green function is

$$G_B(\tau_i,\tau_j)=|\tau_i-\tau_j|-\frac{(\tau_i-\tau_j)^2}{T},$$

with derivatives

$$\dot G_B(\tau_i,\tau_j)=\mathrm{sign}(\tau_i-\tau_j)-2\frac{\tau_i-\tau_j}{T}, \qquad \ddot G_B(\tau_i,\tau_j)=2\delta(\tau_i-\tau_j)-\frac{2}{T}.$$

For open lines, the corresponding Green function is

$A_\mu$0

These kernels encode all Wick contractions of position and velocity insertions along the worldline (Edwards et al., 2022, Ahmadiniaz et al., 2022, Edwards et al., 2019).

External photons are represented by vertex operators. In scalar QED the basic insertion is

$A_\mu$1

After Gaussian contraction one obtains the standard scalar-QED one-loop $A_\mu$2-photon master formula

$A_\mu$3

$A_\mu$4

The notation $A_\mu$5 means projection onto the term multilinear in all polarizations. On open lines an analogous master formula generates the scalar propagator dressed by an arbitrary number of photons (Edwards et al., 2022, Edwards et al., 2019).

Spin is incorporated either by a spin factor,

$A_\mu$6

or, more efficiently, by Grassmann worldline fields $A_\mu$7 with anti-periodic boundary conditions,

$A_\mu$8

This replacement turns path ordering into ordinary Gaussian contraction and leads to the Bern–Kosower cycle replacement rule: after integration by parts removes all $A_\mu$9 terms, every bosonic $$D^{xx'}[A] = \left\langle x' \left| \int_0^\infty dT\, e^{-T\left[-(\partial+ieA)^2+m^2\right]} \right| x \right\rangle = \int_0^\infty dT\, e^{-m^2T} \int_{x(0)=x}^{x(T)=x'} \mathcal D x(\tau)\, e^{-\int_0^T d\tau \left(\frac14 \dot x^2 + ie\,\dot x\cdot A(x(\tau))\right)} .$$0-cycle

$$D^{xx'}[A] = \left\langle x' \left| \int_0^\infty dT\, e^{-T\left[-(\partial+ieA)^2+m^2\right]} \right| x \right\rangle = \int_0^\infty dT\, e^{-m^2T} \int_{x(0)=x}^{x(T)=x'} \mathcal D x(\tau)\, e^{-\int_0^T d\tau \left(\frac14 \dot x^2 + ie\,\dot x\cdot A(x(\tau))\right)} .$$1

is replaced by

$$D^{xx'}[A] = \left\langle x' \left| \int_0^\infty dT\, e^{-T\left[-(\partial+ieA)^2+m^2\right]} \right| x \right\rangle = \int_0^\infty dT\, e^{-m^2T} \int_{x(0)=x}^{x(T)=x'} \mathcal D x(\tau)\, e^{-\int_0^T d\tau \left(\frac14 \dot x^2 + ie\,\dot x\cdot A(x(\tau))\right)} .$$2

thereby converting scalar-loop formulas into spinor-loop formulas without fresh Dirac algebra (Schubert, 2023, Ahmadiniaz et al., 2022, Edwards et al., 2019).

3. Diagram summation and the analytic integration problem

The formalism’s most distinctive structural feature is that the $$D^{xx'}[A] = \left\langle x' \left| \int_0^\infty dT\, e^{-T\left[-(\partial+ieA)^2+m^2\right]} \right| x \right\rangle = \int_0^\infty dT\, e^{-m^2T} \int_{x(0)=x}^{x(T)=x'} \mathcal D x(\tau)\, e^{-\int_0^T d\tau \left(\frac14 \dot x^2 + ie\,\dot x\cdot A(x(\tau))\right)} .$$3-integrations already sum over large families of Feynman diagrams. At one loop, diagrams that differ only by the ordering of external legs around the loop are encoded by a single master integral. In the four-photon case, for example, six inequivalent cyclic orderings that would appear as separate one-loop diagrams in a standard approach are unified by the same unordered $$D^{xx'}[A] = \left\langle x' \left| \int_0^\infty dT\, e^{-T\left[-(\partial+ieA)^2+m^2\right]} \right| x \right\rangle = \int_0^\infty dT\, e^{-m^2T} \int_{x(0)=x}^{x(T)=x'} \mathcal D x(\tau)\, e^{-\int_0^T d\tau \left(\frac14 \dot x^2 + ie\,\dot x\cdot A(x(\tau))\right)} .$$4-integral (Edwards et al., 2022).

This compression produces a non-standard analytic problem. After rescaling $$D^{xx'}[A] = \left\langle x' \left| \int_0^\infty dT\, e^{-T\left[-(\partial+ieA)^2+m^2\right]} \right| x \right\rangle = \int_0^\infty dT\, e^{-m^2T} \int_{x(0)=x}^{x(T)=x'} \mathcal D x(\tau)\, e^{-\int_0^T d\tau \left(\frac14 \dot x^2 + ie\,\dot x\cdot A(x(\tau))\right)} .$$5, one encounters integrals of the form

$$D^{xx'}[A] = \left\langle x' \left| \int_0^\infty dT\, e^{-T\left[-(\partial+ieA)^2+m^2\right]} \right| x \right\rangle = \int_0^\infty dT\, e^{-m^2T} \int_{x(0)=x}^{x(T)=x'} \mathcal D x(\tau)\, e^{-\int_0^T d\tau \left(\frac14 \dot x^2 + ie\,\dot x\cdot A(x(\tau))\right)} .$$6

with

$$D^{xx'}[A] = \left\langle x' \left| \int_0^\infty dT\, e^{-T\left[-(\partial+ieA)^2+m^2\right]} \right| x \right\rangle = \int_0^\infty dT\, e^{-m^2T} \int_{x(0)=x}^{x(T)=x'} \mathcal D x(\tau)\, e^{-\int_0^T d\tau \left(\frac14 \dot x^2 + ie\,\dot x\cdot A(x(\tau))\right)} .$$7

The absolute values in $$D^{xx'}[A] = \left\langle x' \left| \int_0^\infty dT\, e^{-T\left[-(\partial+ieA)^2+m^2\right]} \right| x \right\rangle = \int_0^\infty dT\, e^{-m^2T} \int_{x(0)=x}^{x(T)=x'} \mathcal D x(\tau)\, e^{-\int_0^T d\tau \left(\frac14 \dot x^2 + ie\,\dot x\cdot A(x(\tau))\right)} .$$8 and sign functions in $$D^{xx'}[A] = \left\langle x' \left| \int_0^\infty dT\, e^{-T\left[-(\partial+ieA)^2+m^2\right]} \right| x \right\rangle = \int_0^\infty dT\, e^{-m^2T} \int_{x(0)=x}^{x(T)=x'} \mathcal D x(\tau)\, e^{-\int_0^T d\tau \left(\frac14 \dot x^2 + ie\,\dot x\cdot A(x(\tau))\right)} .$$9 make ordinary symbolic integration ineffective unless one decomposes the domain into ordered sectors. Such sector decomposition reconstructs the individual Feynman graphs that the worldline representation was designed to avoid, so the search for worldline-native integration technology has become a central subproblem of the subject (Ahmadiniaz et al., 2022, Guzman et al., 19 Mar 2026).

One important class of exact results concerns bosonic cycle integrals,

$$\Gamma_{\rm scal}[A] = \int_0^\infty \frac{dT}{T}\, e^{-m^2T} \int_{x(0)=x(T)} \mathcal D x(\tau)\, e^{-\int_0^T d\tau \left(\frac14 \dot x^2 + ie\,\dot x\cdot A(x(\tau))\right)} .$$0

for which

$$\Gamma_{\rm scal}[A] = \int_0^\infty \frac{dT}{T}\, e^{-m^2T} \int_{x(0)=x(T)} \mathcal D x(\tau)\, e^{-\int_0^T d\tau \left(\frac14 \dot x^2 + ie\,\dot x\cdot A(x(\tau))\right)} .$$1

This explains why Bernoulli numbers and Bernoulli polynomials recur throughout worldline amplitudes. In the Hilbert space of periodic functions orthogonal to constants, the inverse derivative kernels satisfy

$$\Gamma_{\rm scal}[A] = \int_0^\infty \frac{dT}{T}\, e^{-m^2T} \int_{x(0)=x(T)} \mathcal D x(\tau)\, e^{-\int_0^T d\tau \left(\frac14 \dot x^2 + ie\,\dot x\cdot A(x(\tau))\right)} .$$2

and diagonal matrix elements reduce to Bernoulli numbers. This algebra of inverse derivatives underlies a program for evaluating full-momentum one-loop integrals without ordered-sector decomposition, particularly in scalar $$\Gamma_{\rm scal}[A] = \int_0^\infty \frac{dT}{T}\, e^{-m^2T} \int_{x(0)=x(T)} \mathcal D x(\tau)\, e^{-\int_0^T d\tau \left(\frac14 \dot x^2 + ie\,\dot x\cdot A(x(\tau))\right)} .$$3 theory (Edwards et al., 2022, Ahmadiniaz et al., 2022).

Further techniques extend these ideas to constant external fields. In a magnetic background one packages the field-dependent bosonic Green functions into

$$\Gamma_{\rm scal}[A] = \int_0^\infty \frac{dT}{T}\, e^{-m^2T} \int_{x(0)=x(T)} \mathcal D x(\tau)\, e^{-\int_0^T d\tau \left(\frac14 \dot x^2 + ie\,\dot x\cdot A(x(\tau))\right)} .$$4

whose iterated integrals satisfy closed folding identities such as

$$\Gamma_{\rm scal}[A] = \int_0^\infty \frac{dT}{T}\, e^{-m^2T} \int_{x(0)=x(T)} \mathcal D x(\tau)\, e^{-\int_0^T d\tau \left(\frac14 \dot x^2 + ie\,\dot x\cdot A(x(\tau))\right)} .$$5

This reduces repeated worldline integrations to algebraic manipulations in the field parameters $$\Gamma_{\rm scal}[A] = \int_0^\infty \frac{dT}{T}\, e^{-m^2T} \int_{x(0)=x(T)} \mathcal D x(\tau)\, e^{-\int_0^T d\tau \left(\frac14 \dot x^2 + ie\,\dot x\cdot A(x(\tau))\right)} .$$6. A plausible implication is that the analytic tractability of worldline formulas depends less on conventional special-function technology than on the specific Green-function algebra generated by the circle or interval worldline (Guzman et al., 19 Mar 2026).

4. Bounded manifolds, heat kernels, and singular interface geometry

When the underlying field theory is defined on a bounded region, the worldline path integral must be restricted to trajectories compatible with the boundary. For a scalar field confined to the $$\Gamma_{\rm scal}[A] = \int_0^\infty \frac{dT}{T}\, e^{-m^2T} \int_{x(0)=x(T)} \mathcal D x(\tau)\, e^{-\int_0^T d\tau \left(\frac14 \dot x^2 + ie\,\dot x\cdot A(x(\tau))\right)} .$$7-dimensional ball $$\Gamma_{\rm scal}[A] = \int_0^\infty \frac{dT}{T}\, e^{-m^2T} \int_{x(0)=x(T)} \mathcal D x(\tau)\, e^{-\int_0^T d\tau \left(\frac14 \dot x^2 + ie\,\dot x\cdot A(x(\tau))\right)} .$$8, one implementation proceeds by conformally mapping the ball to the half-space $$\Gamma_{\rm scal}[A] = \int_0^\infty \frac{dT}{T}\, e^{-m^2T} \int_{x(0)=x(T)} \mathcal D x(\tau)\, e^{-\int_0^T d\tau \left(\frac14 \dot x^2 + ie\,\dot x\cdot A(x(\tau))\right)} .$$9, reflecting across the interface $\mathrm{Tr}\,e^{-T(-\triangle)}$0, and replacing the original bounded geometry by a doubled geometry $\mathrm{Tr}\,e^{-T(-\triangle)}$1 with reflected metric

$\mathrm{Tr}\,e^{-T(-\triangle)}$2

The price is that the reflected metric is only $\mathrm{Tr}\,e^{-T(-\triangle)}$3, so curvature becomes concentrated at the interface: $\mathrm{Tr}\,e^{-T(-\triangle)}$4 Boundary effects are therefore encoded as singular $\mathrm{Tr}\,e^{-T(-\triangle)}$5- and $\mathrm{Tr}\,e^{-T(-\triangle)}$6-type interactions in the worldline Hamiltonian rather than by an explicit restriction of the Gaussian measure on the original ball (Corradini et al., 2019).

Dirichlet and Neumann conditions are then imposed by an image decomposition. If $\mathrm{Tr}\,e^{-T(-\triangle)}$7 is the reflection of $\mathrm{Tr}\,e^{-T(-\triangle)}$8 across the interface, the heat trace on the physical region is

$\mathrm{Tr}\,e^{-T(-\triangle)}$9

with the upper or lower sign corresponding to Dirichlet or Neumann boundary conditions. The direct contribution sums over loops returning to the same point in the doubled space, while the indirect contribution sums over paths ending at the reflected point; these indirect trajectories correspond to physical paths that touch the boundary. In $\mathrm{Tr}\,e^{-T(-\slashed D^2)}$0 this construction reproduces

$\mathrm{Tr}\,e^{-T(-\slashed D^2)}$1

in agreement with standard heat-kernel coefficients for the disk (Corradini et al., 2019).

The same logic has been extended to spinors. For a Dirac field on a curved two-dimensional half-plane with MIT bag boundary conditions, the heat kernel is written on a doubled manifold as

$\mathrm{Tr}\,e^{-T(-\slashed D^2)}$2

where $\mathrm{Tr}\,e^{-T(-\slashed D^2)}$3 is a spinorial reflection operator and the doubled Hamiltonian contains a projector-valued boundary $\mathrm{Tr}\,e^{-T(-\slashed D^2)}$4-interaction acting on the $\mathrm{Tr}\,e^{-T(-\slashed D^2)}$5 sector. This realizes the mixed MIT bag boundary problem as a direct-plus-image decomposition in which $\mathrm{Tr}\,e^{-T(-\slashed D^2)}$6 behaves as a Dirichlet sector and $\mathrm{Tr}\,e^{-T(-\slashed D^2)}$7 as a Robin sector. The resulting Seeley–DeWitt coefficients are

$\mathrm{Tr}\,e^{-T(-\slashed D^2)}$8

again matching known heat-kernel results (Manzo, 2024).

These constructions show that boundaries in the worldline formalism are not merely exclusions of paths. After doubling, the boundary is converted into a singular interface whose geometry enters as explicit interaction vertices in the first-quantized Hamiltonian. This is the mechanism by which heat-kernel asymptotics, boundary anomalies, and mixed boundary conditions are recovered in a first-quantized language (Corradini et al., 2019, Manzo, 2024).

5. Color, representations, and higher-spin particle models

Non-Abelian gauge couplings introduce path ordering, which is awkward to impose directly in a first-quantized path integral. A standard resolution is to add auxiliary worldline color fields. For a Dirac particle in a non-Abelian background, one introduces color variables $\mathrm{Tr}\,e^{-T(-\slashed D^2)}$9 with first-order kinetic term $T$0, so that their contractions generate step functions and thereby reproduce ordered Wilson-line couplings dynamically. A single Grassmann family produces the direct sum of all totally antisymmetric tensor powers of the fundamental; a single bosonic family produces all totally symmetric tensor powers. Gauging a worldline $T$1 symmetry with a Chern–Simons coupling projects onto fixed occupation number, and partially gauging a family $T$2 symmetry with the appropriate Faddeev–Popov measure projects onto an arbitrary irreducible $T$3 representation (Edwards et al., 2016).

In the multi-family Grassmann construction, the color partition function on the circle factorizes over angular moduli $T$4, and the Faddeev–Popov determinant

$T$5

acts as a representation-theoretic projector. The resulting worldline path integral produces the Wilson-loop character in the chosen representation rather than merely the correct Hilbert-space dimension. This removes the restriction to fundamental or purely symmetric/antisymmetric matter and makes arbitrary mixed-symmetry multiplets accessible to worldline calculations (Edwards et al., 2016).

Open non-Abelian worldlines provide the corresponding technology for dressed propagators. For a colored scalar propagating in a non-Abelian background, an open-worldline path integral with auxiliary color fields and endpoint coherent states yields a master formula for the scalar propagator with an arbitrary number of gluons attached directly to the scalar line. The same formalism simultaneously describes a particle in the fundamental or in arbitrarily chosen symmetric or antisymmetric tensor products of the fundamental. Because the worldline is an interval rather than a circle, the result is a propagator-like building block rather than a trace, but the color-ordering mechanism remains the same (Ahmadiniaz et al., 2015).

Higher spin is described by extended worldline supersymmetry. Free massless spin $T$6 particles are represented by $T$7 supersymmetric spinning-particle models with $T$8 symmetry; in particular, the graviton corresponds to an $T$9 model. At tree level, graviton scattering can be formulated by inserting background-graviton vertex operators on the worldline. A specific controversy appears for spin $(4\pi T)^{-D/2}$0: reproducing the Einstein three-graviton vertex with one graviton off shell requires breaking the $(4\pi T)^{-D/2}$1 symmetry of the free graviton worldline to $(4\pi T)^{-D/2}$2, and the coefficient $(4\pi T)^{-D/2}$3 of the counterterm $(4\pi T)^{-D/2}$4 then differs from previous results in the literature. The same analysis reveals a squaring relation between linearized photon and graviton emission operators, leading for MHV amplitudes to double-copy-like relations between worldline numerators (Du et al., 2023).

6. External fields, strong coupling, noncommutative geometry, and classical scattering

External electromagnetic backgrounds are one of the most developed application areas. In a constant field $(4\pi T)^{-D/2}$5, the free worldline Green functions are replaced by matrix-valued field-dependent kernels $(4\pi T)^{-D/2}$6, and the Gaussian normalization acquires determinant factors such as

$(4\pi T)^{-D/2}$7

These determinants generate the Weisskopf and Euler–Heisenberg effective Lagrangians. The same machinery yields compact one-loop representations for photon splitting in constant magnetic fields, plane-wave background amplitudes, and open-line quantities relevant to Compton scattering, while the semiclassical worldline-instanton method gives the imaginary part of the effective action governing Schwinger pair creation in general electric backgrounds (Schubert, 2023).

The formalism has also been used outside weak-coupling amplitude theory. In QCD-like theories, the fermion determinant can be rewritten as

$(4\pi T)^{-D/2}$8

with $(4\pi T)^{-D/2}$9 represented as a sum over closed super-Wilson loops. Differentiating with respect to mesonic sources selects loop ensembles constrained to pass through insertion points, so mesonic correlators become gauge averages of Wilson loops passing through those points. In large-$$G_B(\tau_i,\tau_j)=|\tau_i-\tau_j|-\frac{(\tau_i-\tau_j)^2}{T},$$0 two-dimensional QCD, this leads to a mapping of the asymptotic meson spectrum to a harmonic oscillator or Landau problem; for the Peskin $$G_B(\tau_i,\tau_j)=|\tau_i-\tau_j|-\frac{(\tau_i-\tau_j)^2}{T},$$1-parameter it yields a strong-coupling scaling estimate

$$G_B(\tau_i,\tau_j)=|\tau_i-\tau_j|-\frac{(\tau_i-\tau_j)^2}{T},$$2

with $$G_B(\tau_i,\tau_j)=|\tau_i-\tau_j|-\frac{(\tau_i-\tau_j)^2}{T},$$3 depending on the $$G_B(\tau_i,\tau_j)=|\tau_i-\tau_j|-\frac{(\tau_i-\tau_j)^2}{T},$$4-ality of the representation. A related but distinct application uses the convergence of the worldline/Wilson-loop expansion of the fermion determinant to motivate the heuristic conformal-window criterion

$$G_B(\tau_i,\tau_j)=|\tau_i-\tau_j|-\frac{(\tau_i-\tau_j)^2}{T},$$5

which was proposed as a universal estimate for the lower boundary of the conformal window in non-supersymmetric QCD-like theories (Armoni et al., 2011, 0907.4091).

Noncommutative geometry produces another extension. In a linearized Snyder space, the one-loop effective action of the scalar $$G_B(\tau_i,\tau_j)=|\tau_i-\tau_j|-\frac{(\tau_i-\tau_j)^2}{T},$$6 theory can be recast in worldline form after expanding the nonlocal fluctuation operator to first order in the Snyder parameter $$G_B(\tau_i,\tau_j)=|\tau_i-\tau_j|-\frac{(\tau_i-\tau_j)^2}{T},$$7. The resulting master formula for one-loop $$G_B(\tau_i,\tau_j)=|\tau_i-\tau_j|-\frac{(\tau_i-\tau_j)^2}{T},$$8-point functions shows that the two-point function renormalizes only the mass, the four-point function renormalizes both $$G_B(\tau_i,\tau_j)=|\tau_i-\tau_j|-\frac{(\tau_i-\tau_j)^2}{T},$$9 and $$\dot G_B(\tau_i,\tau_j)=\mathrm{sign}(\tau_i-\tau_j)-2\frac{\tau_i-\tau_j}{T}, \qquad \ddot G_B(\tau_i,\tau_j)=2\delta(\tau_i-\tau_j)-\frac{2}{T}.$$0, and the six-point function generates a divergent $$\dot G_B(\tau_i,\tau_j)=\mathrm{sign}(\tau_i-\tau_j)-2\frac{\tau_i-\tau_j}{T}, \qquad \ddot G_B(\tau_i,\tau_j)=2\delta(\tau_i-\tau_j)-\frac{2}{T}.$$1 term,

$$\dot G_B(\tau_i,\tau_j)=\mathrm{sign}(\tau_i-\tau_j)-2\frac{\tau_i-\tau_j}{T}, \qquad \ddot G_B(\tau_i,\tau_j)=2\delta(\tau_i-\tau_j)-\frac{2}{T}.$$2

raising the question of whether the linearized theory is renormalizable. The momentum-quadratic term in the corresponding worldline Hamiltonian also admits an interpretation in terms of an effective metric proportional to $$\dot G_B(\tau_i,\tau_j)=\mathrm{sign}(\tau_i-\tau_j)-2\frac{\tau_i-\tau_j}{T}, \qquad \ddot G_B(\tau_i,\tau_j)=2\delta(\tau_i-\tau_j)-\frac{2}{T}.$$3 (Franchino-Viñas et al., 2018).

In the classical limit of long-range scattering, the formalism reorganizes amplitudes into worldline quantum field theory. For $$\dot G_B(\tau_i,\tau_j)=\mathrm{sign}(\tau_i-\tau_j)-2\frac{\tau_i-\tau_j}{T}, \qquad \ddot G_B(\tau_i,\tau_j)=2\delta(\tau_i-\tau_j)-\frac{2}{T}.$$4 scattering with massless mediators, the exact worldline amplitude reduces in the classical limit to the WQFT rules of Mogull, Plefka, and Steinhoff. The asymptotic vertex operators shift the worldline to a straight classical trajectory,

$$\dot G_B(\tau_i,\tau_j)=\mathrm{sign}(\tau_i-\tau_j)-2\frac{\tau_i-\tau_j}{T}, \qquad \ddot G_B(\tau_i,\tau_j)=2\delta(\tau_i-\tau_j)-\frac{2}{T}.$$5

after Fourier transform to impact-parameter space, and the $$\dot G_B(\tau_i,\tau_j)=\mathrm{sign}(\tau_i-\tau_j)-2\frac{\tau_i-\tau_j}{T}, \qquad \ddot G_B(\tau_i,\tau_j)=2\delta(\tau_i-\tau_j)-\frac{2}{T}.$$6-expansion becomes a contraction expansion on the worldline. In this language reducible and irreducible contributions map directly onto the structure of the eikonal expansion, and the eikonal phase is identified with minimally connected WQFT diagrams (Ajith et al., 2024).

7. Phase-space reformulations and current directions

A recent development reformulates the worldline formalism directly in phase space, viewing the particle worldline as a sigma model into a symplectic manifold $$\dot G_B(\tau_i,\tau_j)=\mathrm{sign}(\tau_i-\tau_j)-2\frac{\tau_i-\tau_j}{T}, \qquad \ddot G_B(\tau_i,\tau_j)=2\delta(\tau_i-\tau_j)-\frac{2}{T}.$$7 with action

$$\dot G_B(\tau_i,\tau_j)=\mathrm{sign}(\tau_i-\tau_j)-2\frac{\tau_i-\tau_j}{T}, \qquad \ddot G_B(\tau_i,\tau_j)=2\delta(\tau_i-\tau_j)-\frac{2}{T}.$$8

For the relativistic scalar on $$\dot G_B(\tau_i,\tau_j)=\mathrm{sign}(\tau_i-\tau_j)-2\frac{\tau_i-\tau_j}{T}, \qquad \ddot G_B(\tau_i,\tau_j)=2\delta(\tau_i-\tau_j)-\frac{2}{T}.$$9, this gives

$A_\mu$00

In this framework the inverse symplectic form is the propagator kernel, noncanonical coordinates move interactions from the Hamiltonian into the symplectic potential, and using kinetic rather than canonical momentum makes gauge invariance more transparent. The same approach identifies interval, half-line, and full-line worldline topologies with bulk-to-bulk, bulk-to-boundary, and boundary-to-boundary objects, respectively, thereby automating LSZ reduction through the topology and boundary conditions of the worldline itself (Kim, 7 Sep 2025).

This phase-space viewpoint suggests a shift in emphasis from worldline formulas as dressed propagators in configuration space to worldline formulas as amplitude-oriented objects native to momentum space. The same paper uses this setup to compute multi-photon Compton amplitudes up to six points in the classical limit and to organize Yang–Mills and gravity amplitudes in a uniform way by supposing backgrounds of nonlinearly superposed plane waves (Kim, 7 Sep 2025).

Several current limitations are explicit in the literature. For arbitrary-momentum one-loop $A_\mu$01-photon amplitudes, the inverse-derivative/Bernoulli-polynomial algorithm has been described as solving the scalar circular-integration problem only “in principle,” while the full QED generalization remains in progress (Edwards et al., 2022). In strong-field QED, constant fields admit a simple Bern–Kosower-type replacement structure, but for plane waves no comparably simple replacement rule is known in the spinor case (Schubert, 2023). In boundary problems, explicit worldline constructions are available for the scalar ball and for two-dimensional spinors with MIT bag conditions, while extensions to higher dimensions, more general local boundary conditions, and broader numerical implementations remain open directions (Corradini et al., 2019, Manzo, 2024).

Across these developments, the formalism retains a stable core: proper time, one-dimensional Green functions, and first-quantized particle dynamics. What varies is the geometry of the target space, the internal worldline degrees of freedom, and the interpretation of the resulting parameter integrals. This suggests that “worldline formalism” is less a single technique than a family of first-quantized representations whose common purpose is to reorganize quantum-field-theoretic information into compact, trajectory-based structures (Edwards et al., 2019, Guzman et al., 19 Mar 2026).