ABCD optical path generating function Created 2026-10-25 Updated 2026-11-05
Consider a lossless optical system in the paraxial approximation with the same refractive index at its input and output. In ray-height/slope coordinates, write
For , its optical path length generating function is
The factor is required for physical optical path length; omitting it defines the index-normalized length .
Fermat's principle gives endpoint derivatives and . Substituting and and integrating proves the expression. For uniform free space , the extra optical length is , fixing both sign and normalization. Its quadratic wave phase generates a linear canonical transform. The case is a separate imaging limit.
Fermat's principle 2026-11-05
Fermat's principle characterizes a light ray between fixed endpoints by stationary travel time, or stationary optical path length in a stationary isotropic medium. With refractive index and path element , the travel time is , so an allowed ray has vanishing first variation of this integral. Stationary does not always mean the globally shortest time.
For a planar interface between two uniform media, vary the point at which a broken straight path crosses the interface. Setting the travel-time derivative to zero gives , Snell's law, with angles measured from the normal. This makes the link between a variational statement and an observable direction of refraction explicit. The geometrical optics description requires scales large enough compared with the wavelength; diffraction supplies effects that a single ray cannot describe.
Parabolic light rays in a reciprocal-square-root speed profile Created 2026-10-18 Updated 2026-11-05
For speed of light , where and , the travel-time integrand in Fermat's principle is . For nonzero , the Beltrami identity gives
where is the turning height. The corresponding stationary rays are the parabolas
For equal-height endpoints with , symmetry gives and . Real turning heights therefore require , not merely a one-sided condition on signed . When the inequality is strict, the two algebraic values are ; only rays remaining inside the specified optical layer are admissible. Equality merges the two branches. At , the medium has constant speed of light and the ray joining these endpoints is the straight horizontal segment; the formulas containing must be replaced by that limit.
Use the forward-slope convention established in part a, with and , where . Assume . Solve for the endpoint slopes:
Fermat's principle makes the optical path stationary against variations at internal interfaces. Consequently only endpoint terms survive when its endpoints change. For optical path , these are
Integrate the first equation with respect to , then match the second and impose . The resulting ABCD optical path generating function is
Thus the geometrical kernel defined in the source would have the form
The printed minus sign is inconsistent with forward propagation under these conventions. A concrete counterexample is free space: , , . Its true geometrical path is
whereas the printed expression predicts a path shorter than for nonzero transverse separation. For example, , , gives , rather than the printed . Reversing the slope convention would also reverse the free-space entry and must be done consistently; it cannot justify the minus sign with from part a.
The printed form can be recovered by explicitly using the opposite slope coordinate . Then , free space has , and the same physical path is . Thus the sign can be a convention choice, but the printed minus sign must not be combined with the positive- free-space ray transfer matrix used here.
For physical field propagation, the paraxial diffraction kernel retains this quadratic optical-path phase while using the leading spreading amplitude. In one transverse dimension the normalized linear canonical transform is
with a consistent square-root branch and . The source's prefactor and kernel are a schematic one-dimensional version of the Fresnel-Kirchhoff diffraction integral, not this normalized general-system propagator. The quadratic denominator is normally replaced by its leading value when making the Fresnel approximation. When , propagation is an imaging limit rather than an ordinary integral with division by .
In each layer a minimum-time path is straight because the speed is constant there. The total time as a function of the two crossing positions is
Setting its two partial derivatives to zero gives
The angles in the held diagram are measured from the vertical interface normals, so these ratios are
This is the layered-speed form of Fermat's principle and Snell's law. Signed horizontal displacements can be retained if the endpoint lies on the other side of the origin.
Fermat's principle says that the travel time, or equivalently the optical path length, is stationary under small path variations at fixed endpoints. For a flat interface, let the ray cross at horizontal coordinate , with source and detector at heights and horizontal separation . Apart from , its travel time is
Setting gives
The angles are measured to the normal. This is Snell's law; “least time” is more generally a stationarity condition.
Fermat's principle states that the actual ray makes the travel time stationary under small variations of its path with endpoints fixed. Equivalently, for a stationary medium it makes the optical path length stationary. “Stationary” allows minima, maxima, or other stationary paths; it does not require a strict minimum in every optical system.
Fermat's principle states that the travel time, or equivalently the optical path length in a stationary medium, is stationary under small variations of the path with endpoints held fixed. It need not be a global minimum.
Let a light ray join a point at height above a plane interface to a point at depth below it, with horizontal separation . If the crossing point is horizontally from the first endpoint, its optical path length is
Setting its derivative to zero gives
The two geometric ratios are the sines of the incident and refracted angles, measured from the interface normal vector. Hence Snell's law is
The paraxial approximation restricts light rays to small slopes relative to the optical axis, so and geometric lengths can be expanded to second order in height. Treat the source's optical lens as a thin lens located at , with and small surface slopes.
For positive real conjugate distances , the collapsed-plane geometric path has length
To thin-element accuracy, a segment of this path lies inside the optical lens, replacing the same length of air. The geometric air path is , the material path is , and the optical path length and travel time are therefore
This distinguishes geometric length from optical path length; adding to without subtracting the replaced air would count that segment twice.
For a nonaxial light ray, Fermat's principle gives and hence
The focal length is obtained with the incident object at infinity:
The constant thickness supplies a common delay but does not change the leading focal length. Since has units of inverse length, this formula has the correct dimensions. These are thin lens results; paraxial approximation alone does not remove finite-thickness corrections in a general thick lens.
Fermat's principle makes the optical path length, or transit time in a stationary medium, stationary under small variations of the light ray with fixed endpoints. It need not be a global minimum. In the result of part (a), the coefficient of is
For a family of paraxial light rays from the object to its image, stationarity with respect to requires this coefficient to vanish. Therefore the thin lens equation is
For a distant object, , and the image position defines the focal length. Hence
The axial member is stationary by symmetry for arbitrary ; requiring the neighbouring non-axial light rays to meet the same image is what determines .
In surrounding refractive index , the geometrical segments contribute to the optical path length, while replacement by the lens adds . Fermat's principle therefore gives
The physical focal length measured in the surrounding medium is
For and , increasing from towards makes the positive focal length grow monotonically from its vacuum value to infinity: . At , there is no refraction at the boundary, so the element has zero optical power and no finite optical focus. If , its optical power changes sign and the same shape becomes diverging.
Seismic surface-wave dispersion. Rayleigh waves combine vertical and longitudinal motion in the vertical plane of propagation; Love waves have horizontal transverse motion and can be guided by a low-shear-speed surface layer. A homogeneous elastic half-space supports a Rayleigh wave with a frequency-independent speed in the ideal elastic limit, but depth-dependent Earth structure makes the surface-wave eigenfunctions and their propagation speeds depend on frequency. This structural dispersion can occur even when each constituent material is itself nondispersive.
Distinguish phase velocity from group velocity , which controls a narrow-band envelope in the appropriate weakly attenuating limit. Longer periods generally have deeper sensitivity, while shorter periods emphasize shallower structure. In a profile whose shear speed increases with depth this often makes long-period waves faster, but low-velocity zones, anisotropy and higher modes can change that trend. Frequency dependence need not be monotonic in every Earth model.
Measure phase delays or wave-packet arrival times across known source–receiver paths for many periods, accounting for source phase, cycle ambiguity, multipathing and mode identification. Compute predicted dispersion from trial elastic profiles and compare with observations. Linearization expresses the sensitivity schematically as
where are compressional and shear speeds. A regularized inverse problem then estimates depth structure; many crossing paths additionally provide surface-wave tomography. Joint Love wave and Rayleigh wave measurements help distinguish shear-wave polarizations and constrain radial seismic anisotropy. Finite bandwidth, depth averaging and correlated sensitivities mean that one dispersion curve does not uniquely determine a detailed Earth profile.
Seismic ray theory, the eikonal equation and Fermat's principle. In a smoothly varying isotropic medium, insert a high-frequency ansatz into the scalar wave equation. The leading powers of frequency give
The travel-time surfaces are wavefronts and rays are their normal trajectories, with slowness vector . The next-order equation transports amplitude, including geometrical spreading; attenuation is an additional material effect.
Fermat's principle says that the ray makes stationary under path variations with fixed endpoints. Stationarity, rather than always the global minimum, permits multiple arrivals. Its Euler–Lagrange equation, with arclength and unit tangent , is
For horizontal layering, the horizontal slowness is conserved, giving Snell's law at interfaces and smooth refraction between them. Here is measured from vertical. A turning point satisfies ; increasing speed with depth bends a downgoing ray back upwards. Travel-time tomography uses these ray integrals to infer spatial slowness variations.
The approximation requires wavelength to be small compared with structural variation scales. Caustics, turning-point amplitudes, diffraction, scattering and finite-frequency sensitivity require wave-based corrections. Thus seismic ray theory gives an economical geometric description and useful travel times, while surface-wave dispersion retains essential interference and boundary-mode physics.
Figure 1.
Original illustrative Love-wave dispersion and depth sensitivity, alongside a turning ray
. Details.
Seismic ray theory 2026-10-20
Seismic ray theory approximates high-frequency propagation through smoothly varying media. The travel-time field satisfies the eikonal equation, and its characteristics are paths of stationary travel time according to Fermat's principle. Diffraction and caustics require refinements beyond a simple geometrical-ray approximation.