---
title: 'Classical Eikonal: Wavefronts & Scattering'
url: https://www.emergentmind.com/topics/classical-eikonal
type: topic
---

# Classical Eikonal: Wavefronts & Scattering

Classical eikonal denotes a family of closely related structures that arise when wave propagation or scattering is reduced to phase data. In geometrical optics and Hamilton–Jacobi theory, it is the first-order nonlinear equation satisfied by a rapidly varying phase, typically \( |\nabla S(\mathbf r)|^2=n^2(\mathbf r) \) or, in relativistic notation, \(g^{\mu\nu}\partial_\mu S\,\partial_\nu S=0\) [1906.09467][1011.5074]. In kinetic theory, a hyperbolic WKB limit of transport with BGK relaxation produces a non-quadratic Hamilton–Jacobi equation that is only asymptotically equivalent to the classical quadratic eikonal for small gradients [1202.2342]. In high-energy scattering, the same term refers to an impact-parameter phase \(\delta(s,b)\) entering \(\tilde{\mathcal S}(s,b)\approx e^{2i\delta(s,b)}\), or, in a stricter canonical formulation, to a generator \(\chi\) of the classical scattering map \(A_{\rm out}=e^{\{\chi,\}}A_{\rm in}\) [2604.18004][2511.05649]. The modern literature therefore treats “classical eikonal” not as a single equation, but as a unifying Hamilton–Jacobi and phase-space paradigm linking wavefronts, rays, transport limits, and classical observables in scattering.

## 1. Wavefront equations and geometric-optics foundations

In optical and PDE settings, the classical eikonal equation is the leading-order WKB reduction of a wave equation. For the scalar Helmholtz equation,
\[
\nabla^2 \psi(\mathbf r)+k_0^2 n^2(\mathbf r)\psi(\mathbf r)=0,
\]
the short-wavelength ansatz
\[
\psi(\mathbf r)=A(\mathbf r)e^{ik_0 S(\mathbf r)}
\]
yields, at order \(O(k_0^2)\),
\[
|\nabla S(\mathbf r)|^2=n^2(\mathbf r),
\]
while the next order gives the transport equation
\[
2(\nabla A)\!\cdot\!(\nabla S)+A\nabla^2 S=0
\]
[1906.09467]. The level sets \(S(\mathbf r)=\mathrm{const}\) are wavefronts, and rays are the characteristic curves orthogonal to those fronts.

The relativistic version appears in the classical theory of wavefronts in Maxwell electrodynamics. If a wavefront is described by
\[
\omega(t,\mathbf x)=f(\mathbf x)-t=0,
\]
then the Lorentz-invariant eikonal equation is
\[
\left(\frac{\partial \omega}{\partial t}\right)^2-(\nabla \omega)^2=0,
\]
equivalently
\[
g^{\mu\nu}\partial_\mu \omega\,\partial_\nu \omega=0,
\qquad g^{\mu\nu}=\mathrm{diag}(1,-1,-1,-1),
\]
which is the Hamilton–Jacobi equation for massless propagation [1011.5074].

A parallel mathematical tradition studies the relativistic and unit eikonal equations
\[
u_\mu u_\mu=0,
\qquad
u_\mu u_\mu=1,
\]
with
\[
u_\mu u_\mu=u_0^2-u_1^2-\cdots-u_n^2
\]
in Minkowski signature [2212.09914]. In this language, the null equation describes lightlike gradients, while the unit equation fixes the Minkowski norm of the gradient to \(+1\). The same source notes that these are classical eikonal equations in the sense of geometrical optics and Hamilton–Jacobi theory.

| Context | Representative object | Role |
|---|---|---|
| Geometrical optics | \( |\nabla S|^2=n^2(\mathbf r) \) | Wavefronts and rays |
| Relativistic PDE | \(u_\mu u_\mu=0\), \(u_\mu u_\mu=1\) | Null and unit eikonal equations |
| Kinetic transport | \(\partial_t\varphi^0+H(\nabla_x\varphi^0)=0\) | Velocity-averaged eikonal analogue |
| Impact-parameter scattering | \(\tilde{\mathcal S}(s,b)\approx e^{2i\delta(s,b)}\) | Classical phase shift |
| Canonical scattering map | \(A_{\rm out}=e^{\{\chi,\}}A_{\rm in}\) | Classical scattering generator |

These forms are not interchangeable, but they are structurally linked by the fact that each encodes propagation through a phase function whose derivatives determine observable trajectories or fronts.

## 2. Solution theory, characteristics, and symmetry structure

For the unit relativistic eikonal equation \(u_\mu u_\mu=1\), a general local solution can be organized by the rank of the Hessian \(U=(u_{\mu\nu})\) [2212.09914]. The rank-zero sector consists of linear solutions
\[
u(x)=c_\mu x^\mu+c,
\qquad
c_\mu c_\mu=1.
\]
For maximal spatial rank, the general solution is given parametrically by
\[
u=-x_aT_a+x_0\sqrt{1+T_aT_a}+\mathcal Y(T_a),
\]
together with the implicit constraints
\[
x_a-x_0T_a(1+T_bT_b)^{-1/2}-\mathcal Y_{T_a}(T_a)=0,
\]
where \(T_a\) are parameters and \(\mathcal Y\) is an arbitrary sufficiently smooth function. Intermediate ranks \(1\le k\le n-1\) admit analogous parametric forms involving auxiliary functions \(w_m(T_b)\). This rank stratification makes the general solution resemble characteristic data encoded by a generating function.

The same paper develops the solution theory by hodograph and contact transformations. A general contact transformation is written as
\[
x'^i=\Phi^i(x^\mu,u,u_\mu),
\qquad
u'=\Psi(x^\mu,u,u_\mu),
\qquad
u'_i=Q_i(x^\mu,u,u_\mu),
\]
and is used to convert first-order nonlinear eikonal equations into simpler equations in new variables. In two dimensions, the Euclidean eikonal
\[
u_1^2+u_2^2=1
\]
is reduced to a parametric solution through a Legendre-type transformation; the higher-dimensional relativistic case follows the same pattern.

The symmetry theory is equally rich. For the null equation \(u_\mu u_\mu=0\), the maximal Lie symmetry algebra is infinite dimensional and contains transformations
\[
X=\xi^\mu(x)\partial_{x^\mu}+\beta(u)x_\mu\partial_{x^\mu}+\alpha(u)\partial_u,
\]
with \(\xi^\mu(x)\) a conformal Killing field and \(\alpha(u),\beta(u)\) arbitrary smooth functions. A distinctive property is that any function of a solution is again a solution. For \(u_\mu u_\mu=1\), the maximal Lie symmetry algebra is finite dimensional but contains translations, Lorentz transformations, dilations, and special conformal transformations in an extended \((n+2)\)-dimensional space where the dependent variable \(u\) is treated as an additional coordinate. Discrete symmetries include time reflection, space reflections, sign reversal \(u\mapsto -u\), permutations of spatial coordinates, and local hodograph transformations.

In optical applications, the same characteristic structure becomes computational. The numerical study of Maxwell and Luneburg lenses rewrites the eikonal equation as a characteristic ODE system. In Cartesian coordinates, with \(p_i=\partial_i u\),
\[
\frac{dx}{dt}=\frac{p_1}{n^2(x,y)},
\qquad
\frac{dy}{dt}=\frac{p_2}{n^2(x,y)},
\qquad
\frac{dp_1}{dt}=\frac{1}{n(x,y)}\frac{\partial n}{\partial x},
\qquad
\frac{dp_2}{dt}=\frac{1}{n(x,y)}\frac{\partial n}{\partial y},
\]
and one further has
\[
\frac{du}{dt}=1.
\]
This identifies the ray parameter \(t\) with the eikonal \(u\) up to an additive constant [1906.09467].

## 3. Kinetic and transport generalizations

A nontrivial generalization of the classical eikonal arises from kinetic transport with BGK relaxation [1202.2342]. The starting model is
\[
\partial_t f+v\cdot\nabla_x f=M(v)\rho-f,
\qquad
\rho(t,x)=\int_V f(t,x,v)\,dv,
\]
with \(V\subset\mathbb R^n\) bounded and symmetric, and \(M(v)\) a symmetric normalized Maxwellian. Under the hyperbolic scaling
\[
(t,x)\mapsto \left(\frac{t}{\varepsilon},\frac{x}{\varepsilon}\right),
\]
the rescaled equation becomes
\[
\partial_t f^\varepsilon+v\cdot\nabla_x f^\varepsilon=\frac1\varepsilon\left(M(v)\rho^\varepsilon-f^\varepsilon\right).
\]

The WKB/Hopf–Cole ansatz
\[
f^\varepsilon(t,x,v)=M(v)e^{-\varphi^\varepsilon(t,x,v)/\varepsilon}
\]
leads to a kinetic phase equation. As \(\varepsilon\to0\), \(\varphi^\varepsilon\) converges locally uniformly to a limit \(\varphi^0(t,x)\) independent of \(v\), and \(\varphi^0\) is the viscosity solution of the implicit Hamilton–Jacobi equation
\[
\int_V \frac{M(v)}{1-\partial_t\varphi^0(t,x)-v\cdot\nabla_x\varphi^0(t,x)}\,dv=1.
\]
By monotonicity, this is equivalent to
\[
\partial_t\varphi^0+H(\nabla_x\varphi^0)=0,
\]
where the effective Hamiltonian is defined implicitly by
\[
\int_V \frac{M(v)}{1+H(p)-v\cdot p}\,dv=1.
\]

The paper explicitly presents this as a kinetic analogue of the classical eikonal equation obtained from the heat equation with small diffusivity,
\[
\partial_t\phi^0+\theta^2|\nabla_x\phi^0|^2=0.
\]
The relation is asymptotic rather than exact:
\[
H(p)\sim \theta^2|p|^2
\qquad \text{as }|p|\to0.
\]
Outside the small-gradient regime, the kinetic Hamiltonian is generally non-quadratic. In one dimension with \(V=(-1,1)\) and \(M(v)\equiv \tfrac12\),
\[
H(p)=\frac{p-\tanh(p)}{\tanh(p)},
\]
while in the two-velocity model
\[
M(v)=\tfrac12(\delta_1+\delta_{-1}),
\qquad
H(p)=\frac{\sqrt{1+4p^2}-1}{2}.
\]

Several structural properties distinguish the kinetic Hamiltonian from the classical quadratic one. It is convex, globally Lipschitz, and satisfies
\[
\|\nabla H\|_\infty \le V_{\rm max}.
\]
This bounded gradient expresses finite propagation speeds inherited from the bounded velocity set \(V\). A plausible implication is that the kinetic eikonal retains the Hamilton–Jacobi form while encoding microscopic transport and scattering information through the effective Hamiltonian rather than through a universal quadratic norm.

## 4. Impact-parameter phases and the scattering eikonal

In scattering theory, the classical eikonal appears in high-energy, small-angle, large-impact-parameter regimes where multiple soft exchanges exponentiate into an impact-parameter phase [2604.18004][2211.00085]. For gravitational \(2\to2\) scattering, one writes
\[
\tilde{\mathcal S}(s,b)\approx e^{2i\delta(s,b)},
\]
with \(b\) the transverse impact parameter. In the nonlocal-gravity analysis of scalar scattering, the leading eikonal phase is extracted from the Born amplitude by
\[
2i\delta_0(s,b)=i\tilde A_0(s,b)
\equiv
i\int \frac{d^{D-2}q}{(2\pi)^{D-2}}\,e^{ib\cdot q}\,\frac{A_0(s,-q^2)}{2s}.
\]
Classical momentum transfer and deflection then follow from
\[
Q^\mu=-\frac{\partial}{\partial b_\mu}(2\delta_0),
\qquad
\Theta\simeq -\frac1p\,\partial_b(2\delta_0).
\]

The gravitational case admits a direct geometric interpretation. In massless scattering, the eikonal phase can be viewed as the phase shift of a particle crossing an Aichelburg–Sexl shock wave, with
\[
2\delta_0(s,b)=E^{(2)}f(b).
\]
In the nonlocal theory with Gaussian form factor \(H(\alpha q^2)=\alpha q^2\), the resulting eikonal regularizes the short-distance behavior and is reinterpreted in terms of generalized Aichelburg–Sexl geometries and smeared linearized Schwarzschild metrics. This same eikonal data is then used to motivate regular black-hole spacetimes with de Sitter cores [2604.18004].

In AdS–Schwarzschild scattering of a highly energetic light probe off a heavy object, the eikonal phase is the classical action of a null geodesic,
\[
\delta=p^t\,\Delta t-p^\varphi\,\Delta\varphi,
\]
equivalently
\[
\delta=2|p^t|\int_{r_0}^\infty \frac{dr}{f(r)}
\sqrt{1-\frac{f(r)\alpha^2}{r^2}},
\]
and its derivatives yield classical observables,
\[
\Delta t=\frac{\partial\delta}{\partial p^t},
\qquad
\Delta\varphi=-\frac{\partial\delta}{\partial p^\varphi}.
\]
In the flat-space limit, the phase develops a large imaginary part for sufficiently small impact parameter, and the inelastic cross-section equals the classical absorption cross-section of the black hole [2011.06920].

The large-angular-momentum origin of eikonal exponentiation has also been reformulated group-theoretically. In the large-\(\ell\) limit, the contraction
\[
SU(2)\to ISO(2)
\]
maps the partial-wave description to the isometries of the transverse plane; the Wigner \(d\)-matrices become Bessel functions, and the eikonal transform becomes a two-dimensional Fourier transform in impact-parameter space [2211.00085]. That paper further distinguishes resolvable from non-resolvable corrections: pure quantum-gravity corrections \((\ell_{\rm Pl}/b)^{2n}\) are stated to be never resolvable in the transplanckian eikonal regime, whereas gauge or tidal corrections can dominate higher post-Minkowskian terms.

Infrared structure is part of the same classical story. Combining impact-parameter eikonal exponentiation with momentum-space IR exponentiation, one can determine the divergent part of the two-loop eikonal from tree-level and one-loop elastic amplitudes in \(\mathcal N=8\) supergravity and in general relativity. In that framework, \(\Re\delta\) controls conservative dynamics, while the IR-divergent part of \(\Im\delta_2\) is directly linked to soft radiation, the zero-frequency limit of the energy spectrum, and radiation-reaction effects [2105.04594].

## 5. Phase, generator, Magnusian, and generalized channel structure

A central terminological issue is that amplitudes literature uses “eikonal” for two inequivalent objects [2511.05649]. The **eikonal phase** is defined by
\[
i\delta=\lim_{\hbar\to0^+}\hbar\,
\ln\bigl(\langle \psi_{\rm in}|\hat S|\psi_{\rm in}\rangle\bigr),
\]
and in the classical limit it coincides with the on-shell action,
\[
\delta=S(Q_f,Q_i).
\]
By contrast, the **eikonal generator** or **Magnusian** is the classical limit of the logarithm of the \(S\)-matrix operator,
\[
\hat S=e^{\frac{i}{\hbar}\hat\chi},
\qquad
i\chi=\lim_{\hbar\to0^+}\hbar\,\langle\psi_{\rm in}|\ln \hat S|\psi_{\rm in}\rangle,
\]
and it generates the scattering map in phase space,
\[
A_{\rm out}=e^{\{\chi,\}}A_{\rm in}.
\]
These two objects are inequivalent in general. The exact relation is
\[
S(t_f,Q_f,t_i,Q_i)=
\int_0^1\Bigl(p(\varepsilon)\frac{dq(\varepsilon)}{d\varepsilon}+\chi(t_f,t_i,q(\varepsilon),p(\varepsilon))\Bigr)d\varepsilon,
\]
or equivalently
\[
S=
\frac{e^{\{\chi,\}}-1}{\{\chi,\}}
\bigl(P_i\{\chi,Q_i\}\bigr)+\chi.
\]
Only in integrable scattering do \(\chi\) and \(S\) coincide up to a Legendre transform; with spin or radiation, that simplification fails.

The same distinction is operationalized in the Magnus-expansion approach to the classical eikonal [2410.22988]. There the classical eikonal is defined as the generator of the canonical transformation from in-states to out-states,
\[
O_{\rm out}=e^{\{\chi,\bullet\}}[O_{\rm in}],
\]
and computed from a Magnus expansion in oriented tree graphs whose edges encode retarded and advanced worldline propagators. The paper develops a Hopf-algebra-based algorithm that computes the tree coefficients up to the \(12\)th order—“over half a million trees”—in less than an hour, applies the method to the 3PM gravitational eikonal, and emphasizes that the Magnus prescription yields a finite eikonal whereas the naïve time-symmetric one is IR-divergent from 3PM on.

The worldline formulation provides a bridge between amplitude and canonical perspectives. In the worldline/QFT comparison, the eikonal method and WQFT are shown to be equivalent descriptions of the classical limit of \(2\to2\) scattering with massless mediators; reducible and irreducible WQFT diagrams map one-to-one to the exponentiation structure of the impact-parameter amplitude [2409.17866]. This suggests that the classical eikonal can be treated either as a phase extracted recursively from amplitudes or as a directly computable generator in worldline variables.

Once spin or inelasticity are admitted, the eikonal becomes matrix-valued or channel-valued. For spinning particles, one assumes a matrix eikonal exponentiation
\[
\exp\!\left(\frac{i}{\hbar}\chi\right)^{a_1'a_2'}{}_{a_1a_2},
\]
and classical observables are obtained from derivatives and Poisson brackets of the spinning eikonal, including the shifted variables \(\tilde p_i=p_i\pm Q/2\) and \(\tilde s_i=s_i+\Delta s_i/2\) [2312.09960]. For inelastic coupled-channel eikonal scattering, the diagonal elastic block becomes
\[
S_{aa}(b)=e^{i\chi_a(b)}\eta_a(b),
\]
while off-diagonal blocks encode channel transitions. Unitarity then implies
\[
|\eta_a(x_{*\perp},s)|^2+\int_0^\infty dw\,\rho(w)\,
\eta_b(x_{*\perp},s,w)\eta_b^*(x_{*\perp},s,w)=1,
\]
so the purely elastic unit-modulus phase is recovered only when inelastic channels are absent [2411.02294].

## 6. Causality, polarization, and effective geometry

Classical eikonal data often doubles as a causality diagnostic. In higher-derivative effective theories of gravity, helicity-flip processes can survive the eikonal limit and promote the phase to a \(2\times2\) matrix in helicity space [2006.02375]. For scattering of gravitons or photons of frequency \(\omega\) off a heavy scalar of mass \(m\), the impact-parameter phase matrix is diagonalized, and its eigenvalues are used to compute the classical deflection angle and Shapiro time delay/advance to 2PM order via
\[
\theta^{(i)}(b)=\frac1\omega\,\frac{\partial}{\partial b}\delta^{(i)}(b,\omega),
\qquad
t^{(i)}(b)=\frac{\partial}{\partial \omega}\delta^{(i)}(b,\omega).
\]
Whenever the classical expectation of helicity conservation is violated, the eikonal eigenvalues become non-degenerate, and time advance becomes a generic possibility at small impact parameter. In that analysis, graviton scattering in \(R^4\) and \(FFR\) avoids time advance if the couplings satisfy
\[
\beta_1>0,\qquad \beta_3>0,
\qquad
\alpha_\gamma>0,
\]
whereas graviton scattering in \(R^3\) and photon scattering in \(FFR\) exhibit unavoidable time advance in the EFT regime.

A related causality analysis appears in Lorentz-violating Maxwell theory [1011.5074]. There the wavefront method yields a modified eikonal condition \(\det M_E=0\), equivalent to \(\det M_B=0\), and in a simple anisotropic example the wavefront function \(\omega(t,\mathbf x)\) obeys
\[
\left(\frac{\partial \omega}{\partial t}\right)^2-(\nabla\omega)^2
+\alpha\left(\omega_1\omega_2
\pm \sqrt{(\omega_1^2+\omega_3^2)(\omega_2^2+\omega_3^2)}\right)=0.
\]
The same analysis shows that the wavefront velocity equals the group velocity and the energy-flow velocity, and that there always exists a mode with signal velocity \(v_s>1\); the paper therefore concludes that causality is violated classically in that model.

The causal interpretation of eikonal scattering in gravity can also be phrased positively. In the large-angular-momentum regime, analyticity and unitarity imply non-linear positivity bounds on the eikonal amplitude, with positivity of time delay as the simplest case [2211.00085]. In AdS–Schwarzschild, the onset of a large imaginary part in the eikonal phase at impact parameters below the photon sphere is not presented as superluminality, but as the transition from elastic scattering to absorption, and the resulting inelastic cross-section equals the classical geometric absorption cross-section [2011.06920].

These examples make clear that the classical eikonal is not merely a calculational shortcut. It encodes effective geometry, polarization dependence, propagation speed, absorption, and the distinction between acceptable and pathological EFT corrections. A plausible implication is that eikonal methods continue to serve as a common language for comparing wavefront causality, amplitude analyticity, and classical observables across PDE theory, kinetic limits, and gravitational scattering.

Source: https://www.emergentmind.com/topics/classical-eikonal