Reflection, Refraction, Bounded Waves, Ray Theory and Seismic Tomography
2. Amplitudes at an interface: the Zoeppritz equations in outline
Amplitudes follow from boundary conditions. At a welded solid–solid interface the displacement and the traction (σ_zz, σ_zx, σ_zy) are continuous. For an incident P or SV wave, four waves are involved (reflected P and SV, transmitted P and SV), so four unknown amplitudes are fixed by four equations: the Zoeppritz equations. SH motion decouples, with only reflected and transmitted SH and two equations. At a liquid–solid boundary shear traction vanishes and no S wave exists in the liquid, and at a free surface all traction vanishes.
The results depend on the contrast and on the angle of incidence, which is the basis of AVO (amplitude versus offset) analysis. At normal incidence nothing is converted, and with the acoustic impedance Z = ρv the amplitude reflection coefficient is R = (Z₂ − Z₁)/(Z₂ + Z₁) (the exploration convention, positive for an increase of impedance). The reflected energy fraction is R² and the transmitted fraction is 1 − R². A negative R means a polarity reversal, as at the top of a gas sand. Worked example. Z₁ = 2000 × 2000 = 4.0×10⁶ and Z₂ = 2500 × 3000 = 7.5×10⁶ give R = 3.5/11.5 = 0.304, so R² = 0.0926: 9.26% of the energy is reflected.
3. Travel-time curves: direct, reflected and head waves, and the Earth’s curve
- Direct wave: t = x/v₁, a straight line through the origin.
- Reflected wave from a flat interface at depth h: t = √(x² + 4h²)/v₁, a hyperbola, t² = t₀² + x²/v₁² with t₀ = 2h/v₁.
- Head wave: t = x/v₂ + t_i, with intercept time t_i = 2h cosθ_c / v₁ = 2h √(v₂² − v₁²)/(v₁ v₂). It first arrives beyond the crossover distance x_c = 2h √((v₂ + v₁)/(v₂ − v₁)). The slope of this branch is 1/v₂ and the intercept then gives h.
Worked example. v₁ = 2000 m/s, v₂ = 4000 m/s, h = 500 m. θ_c = 30°, t_i = 2 × 500 × cos 30° / 2000 = 0.433 s and x_c = 1000 × √(6000/2000) = 1732 m, where the direct and head-wave times are both 0.866 s. A layer with a lower speed than the one above it gives no head wave and stays hidden in refraction work.
For the whole Earth the travel-time curve T(Δ) has slope dT/dΔ = p. In a sphere the ray parameter p = r sin i / v is constant along a ray, and at the deepest point (i = 90°) p = r_t / v(r_t); a larger distance Δ means a smaller p and a deeper turning point. The Herglotz–Wiechert inversion converts T(Δ) into v(r) if speed increases steadily with depth, whereas a low-velocity zone gives no first arrivals and stays hidden. The core’s low velocity produces the P shadow zone from about 103° to 142°.
4. Inhomogeneous, evanescent and bounded waves
A homogeneous plane wave has the same amplitude over each surface of constant phase. An inhomogeneous plane wave has an amplitude that varies across that surface, for example u ∝ exp(−ωηz) exp[iω(px − t)] with η = √(p² − 1/v²). When p > 1/v, η is real and the wave is evanescent: it runs along the interface with apparent speed 1/p < v and decays exponentially with distance from it, over about a wavelength or less except very near the critical angle. Beyond the critical angle the transmitted wave in the faster medium is evanescent. It carries no net energy away from the interface, so the reflection is total.
- Rayleigh waves exist at the free surface of a half-space as coupled evanescent P and SV waves. They occur at all frequencies, are non-dispersive in a homogeneous half-space, move in a retrograde ellipse at the surface and travel at about 0.92 β (0.919 β for a Poisson solid).
- Love waves are SH waves trapped in a surface layer of speed β₁ over a faster half-space β₂ by repeated post-critical reflection. They cannot exist in a homogeneous half-space, are dispersive, and have a phase speed c between β₁ and β₂ (near β₁ at short periods, near β₂ at long periods).
- Stoneley waves are guided along a solid–solid or solid–liquid interface, and channel waves in a low-speed seam. All of these are bounded (guided) waves: inhomogeneous waves whose energy is held near a boundary.
Dispersion. The phase speed is c = ω/k and the group speed, at which energy travels, is U = dω/dk = c + k dc/dk. Longer periods penetrate deeper into faster rock, so for periods of tens of seconds the longer ones arrive first, and measured dispersion curves of Rayleigh and Love waves give the shear-speed structure of crust and upper mantle. Surface waves fall off as 1/√r, more slowly than body waves.
5. The eikonal equation and ray theory
Ray theory is the high-frequency solution of the wave equation, valid when the wavelength is much smaller than the scale over which velocity changes. Try u = A(x) exp[iω(T(x) − t)]. To leading order in ω the wave equation gives the eikonal equation |∇T|² = 1/v² = s², where T(x) is the travel time of the wavefront and s the slowness. The next order gives the transport equation, 2∇A·∇T + A∇²T = 0 for constant density, which keeps the energy flux constant along a ray tube. Rays are the curves perpendicular to the wavefronts T = constant and obey the ray equation d/ds (s dx/ds) = ∇s, which expresses Fermat’s principle: the travel time is stationary along the true path.
Ray theory fails at caustics, in shadow zones, at sharp boundaries and where heterogeneity is as small as the wavelength (diffraction). In a medium with a constant velocity gradient v = v₀ + g z, rays are circular arcs of radius R = 1/(p g) that turn where v = 1/p, at depth z_t = (1/p − v₀)/g. Worked example. v₀ = 2.0 km/s, g = 0.5 s⁻¹ and take-off angle i₀ = 30° from the vertical give p = sin 30°/2.0 = 0.25 s/km, R = 1/(0.25 × 0.5) = 8 km and turning depth z_t = (4.0 − 2.0)/0.5 = 4 km. The ray returns to the surface at range X = 2R cos i₀ = 13.9 km.
6. Seismic tomography
Tomography images the interior from many crossing rays. After linearising about a reference model, the travel-time residual of ray i is δt_i = Σ_j L_ij δs_j, where L_ij is the length of ray i in cell j and δs_j is the slowness perturbation (δs = −δv/v²). In matrix form this is d = G m, solved by least squares, m = (GᵀG)⁻¹Gᵀd, in practice damped, m = (GᵀG + ε²I)⁻¹Gᵀd, or by iterative sparse methods (ART, SIRT, LSQR), with the rays re-traced and the process repeated because paths depend on the model. The data can be body-wave delays, surface-wave dispersion or ambient-noise correlations, and full-waveform inversion uses the whole seismogram.
Resolution depends on ray coverage. Where no rays pass, or all pass in one direction, parts of the model lie in the null space and smear along the rays; a checkerboard test shows what can be recovered. Worked example. Take a 2 × 2 grid of cells 10 km wide, numbered 1 2 over 3 4, with four rays along row 1, row 2, column 1 and column 2. G has rank 3: the pattern (+, −, −, +) changes no ray sum, so these rays cannot recover it. Let the residuals be 0.10, 0.30, 0.20 and 0.20 s, and let cell 1 be known to be unperturbed. Column 1 gives 10 δs₃ = 0.20, so δs₃ = 0.020 s/km, row 1 gives δs₂ = 0.010 s/km, and row 2 gives 10(δs₃ + δs₄) = 0.30, so δs₄ = 0.010 s/km; column 2 then checks. With a reference speed of 5.0 km/s (0.200 s/km) cell 4 has 0.210 s/km, or 4.76 km/s.
Key takeaways
- Snell’s law: sin i/v = p is constant along a ray and equals 1/(apparent speed). The critical angle is arcsin(v₁/v₂) for v₂ > v₁, beyond which the reflection is total and a head wave runs along the interface.
- The Zoeppritz equations come from continuity of displacement and traction: four equations for P–SV, two for SH. At normal incidence R = (Z₂ − Z₁)/(Z₂ + Z₁) with Z = ρv, and the reflected energy is R²; state the convention.
- Travel-time curves: direct t = x/v₁, reflected hyperbola t² = t₀² + x²/v₁², head wave t = x/v₂ + t_i with t_i = 2h cosθ_c/v₁. In the Earth dT/dΔ = p, and a low-velocity zone or the core produces hidden layers and shadow zones.
- Beyond the critical angle the transmitted wave is evanescent (exponential decay, no net energy flux). Bounded waves are trapped near boundaries: Rayleigh waves at about 0.92β on a free surface, and dispersive Love waves in a slow surface layer with β₁ < c < β₂.
- Ray theory is the high-frequency limit: the eikonal equation |∇T|² = 1/v² gives wavefronts, the transport equation gives amplitude, and Fermat’s principle gives the paths. In v = v₀ + g z rays are circles that turn at z_t = (1/p − v₀)/g.
- Tomography solves d = G m for slowness perturbations by damped least squares or iteration. Poor coverage leaves a null space (a row-and-column grid cannot see a checkerboard), and slow or fast anomalies need not mean hot or cold.
Practice questions (17)
Attempt each one before opening the answer. Every explanation names the tempting wrong option as well as the right one, because that is where marks are lost.
A seismic array records a P wave that crosses it with an apparent horizontal speed of 10 km/s. The P speed in the crust beneath the array is 6 km/s. What angle does the ray make with the vertical?
Show answer
Answer: B — 37°
The apparent speed is v/sin i, so sin i = 6/10 = 0.6 and i = 37° from the vertical. The tempting 53° is arccos 0.6, the angle from the horizontal, and the remaining values do not follow from the ratio 0.6.A P wave in a slow medium meets a faster medium beyond the critical angle. What is the transmitted wave in the faster medium?
Show answer
Answer: D — An evanescent wave that decays exponentially away from the interface and carries no net energy into the medium
The boundary conditions are still satisfied, by an inhomogeneous (evanescent) wave with exp(−ωηz) decay that travels along the interface. Its time-averaged normal energy flux is zero, which is why the reflection is total. A refracted wave at exactly 90° applies only at the critical angle, and a converted S wave cannot account for the missing transmitted P energy.A P wave is obliquely incident on a welded solid–solid interface. How many wave amplitudes are unknown, and what conditions fix them?
Show answer
Answer: A — Four: reflected P, reflected SV, transmitted P and transmitted SV, fixed by continuity of displacement and traction
In the P–SV system the incident P wave generates reflected P and SV and transmitted P and SV; continuity of the two displacement components and the two traction components in the plane gives four equations, the Zoeppritz equations. SH is decoupled from P and SV, and the two-wave option with pressure continuity describes a fluid–fluid interface.A P pulse travels in rock (ρ = 2500 kg/m³, v = 3500 m/s) and meets, at normal incidence, a gas-bearing sand (ρ = 2000 kg/m³, v = 2500 m/s). Take R = (Z₂ − Z₁)/(Z₂ + Z₁) with Z = ρv. Which statement is correct?
Show answer
Answer: C — R ≈ −0.27: the reflected pulse has reversed polarity and carries about 7.4% of the incident energy
Z₁ = 2500 × 3500 = 8.75×10⁶ and Z₂ = 2000 × 2500 = 5.0×10⁶, so R = (5.0 − 8.75)/(5.0 + 8.75) = −0.273. The sign is negative because impedance decreases, a polarity reversal. The energy fraction is R² = 0.0744, about 7.4%, not 27% (which confuses amplitude with energy).Which description of Love and Rayleigh waves is correct?
Show answer
Answer: A — Love waves are horizontally polarised shear waves that need a slow layer; Rayleigh waves are retrograde ellipses in the vertical plane
Love waves are SH-type and exist only when a slow layer overlies a faster medium; Rayleigh waves are coupled P–SV motion at a free surface, travelling at about 0.92 β. The option that makes Rayleigh waves horizontally polarised swaps the two sets of properties; surface waves are slower than S waves, and Love waves need layering and are dispersive.On a long continental path, surface waves of period 100 s arrive before those of period 20 s. What is the reason?
Show answer
Answer: D — The long periods sample deeper, faster layers, so their speed is higher, which is normal dispersion
Penetration depth grows with wavelength, and speed generally increases with depth, so longer periods have higher speed and arrive first (normal dispersion). Frequency is the inverse of period, so the 100 s waves have the lower frequency, and their arrival order is set by propagation, not by the time of generation.The eikonal equation |∇T|² = 1/v² follows from the elastic wave equation in the limit in which
Show answer
Answer: B — the wave frequency is high, so that the wavelength is much smaller than the scale over which the velocity varies
Ray theory is the high-frequency limit: substituting u = A exp[iω(T − t)] and keeping the leading order in ω gives the eikonal equation, valid when the wavelength is small compared with the scale of velocity variation. The equation holds in heterogeneous media, so homogeneity is not required, and the low-frequency limit is the opposite regime.A 2 × 2 grid of equal cells is crossed only by rays along its two rows and its two columns. Which perturbation pattern changes none of the four travel times and so cannot be resolved?
Show answer
Answer: B — The checkerboard pattern (+, −, −, +), because every row sum and every column sum is zero
With rays along rows and columns, G has rank 3, and the vector (+, −, −, +) lies in its null space because each row and column ray gets +1 and −1 in equal measure. A uniform change adds to every travel time, a single cell changes two rays, and a change in one row alters that row’s ray and both column rays.On a refraction travel-time plot, the branch of the head wave is a straight line that does not pass through the origin. What do its slope and intercept give?
Show answer
Answer: B — The slope is 1/v₂ of the lower layer, and with v₁ and v₂ the intercept time gives the layer thickness h
The head-wave time is t = x/v₂ + t_i, so the slope is 1/v₂, and t_i = 2h cosθ_c/v₁ with sin θ_c = v₁/v₂ lets h be solved once v₁ (from the direct wave) and v₂ are known. The direct wave, not the head wave, has slope 1/v₁, and the intercept is a time, not a distance or a speed.Which of the following statements about bounded and evanescent waves are correct?
Show answer
Answer: A — A Rayleigh wave is non-dispersive at the surface of a homogeneous half-space; C — Love waves cannot exist at the surface of a homogeneous half-space; D — The amplitude of an evanescent wave decays exponentially with distance from the interface
The three correct statements are standard. Surface waves spread over a surface and fall off as 1/√r, which is slower than the 1/r of body waves, so the claim of 1/r² is wrong.Which of the following statements about ray theory are correct?
Show answer
Answer: B — The eikonal equation gives the travel-time field T(x) and hence the wavefronts; C — The transport equation governs how the amplitude changes along a ray, keeping the energy flux in a ray tube constant; D — By Fermat’s principle the travel time along the true ray path is stationary with respect to small changes of path
The eikonal and transport equations and Fermat’s principle are the core of ray theory. At a caustic the ray-theory amplitude becomes infinite, so the method breaks down there along with shadow zones and diffraction.Which of the following statements about seismic tomography are correct?
Show answer
Answer: A — After linearisation the travel-time residuals obey d = G m, with G holding the path lengths of the rays in the cells; B — Poor ray coverage leaves part of the model in the null space, so damping or other regularisation is needed; C — A positive slowness perturbation corresponds to a lower velocity than the reference model
The linear system, the null space and the sign of slowness are all correct, since s = 1/v and a larger s means a smaller v. Composition, melt, water and anisotropy also alter speeds, so a fast anomaly is not always cold.A P wave travelling at 3.0 km/s meets a boundary with a medium of 5.0 km/s at an angle of incidence of 30° from the normal. Find the angle of the refracted P ray from the normal, in degrees, to the nearest degree.
Numerical answer — type the value.
Show answer
Answer: 56
Snell’s law: sin j = (v₂/v₁) sin i = (5.0/3.0) × 0.5 = 0.8333, so j = 56.4°, which rounds to 56°. The critical angle is arcsin(0.6) = 36.9°, and 30° is below it, so a transmitted wave exists.At normal incidence a P wave goes from a layer of ρ = 2000 kg/m³ and v = 2000 m/s into a layer of ρ = 2500 kg/m³ and v = 3000 m/s. Find the percentage of the incident energy that is reflected, to two decimal places. Keep at least four significant figures in R.
Numerical answer — type the value.
Show answer
Answer: 9.26
Z₁ = 2000 × 2000 = 4.0×10⁶ and Z₂ = 2500 × 3000 = 7.5×10⁶. R = (7.5 − 4.0)/(7.5 + 4.0) = 0.3043, and the energy fraction is R² = 0.0926, which is 9.26%. Quoting R × 100 = 30.4% would confuse amplitude with energy.In a refraction survey the upper layer has v₁ = 2000 m/s and the lower layer v₂ = 4000 m/s. The head-wave branch has an intercept time of 0.433 s. Use t_i = 2h cosθ_c / v₁ with sin θ_c = v₁/v₂. Find the thickness h of the upper layer in metres, to the nearest metre.
Numerical answer — type the value.
Show answer
Answer: 500
sin θ_c = 2000/4000 = 0.5, so θ_c = 30° and cos θ_c = 0.866. Then h = t_i v₁ / (2 cosθ_c) = 0.433 × 2000 / (2 × 0.866) = 500 m. Using sin θ_c in place of cos θ_c would give 866 m.The velocity increases linearly with depth as v = 2.0 + 0.5 z (v in km/s, z in km). A ray leaves the surface at 30° from the vertical. Find the depth in km at which the ray reaches its deepest point.
Numerical answer — type the value.
Show answer
Answer: 4
The ray parameter is p = sin 30° / 2.0 = 0.25 s/km. The ray turns where sin i = 1, that is where v = 1/p = 4.0 km/s. Then 2.0 + 0.5 z = 4.0 gives z = 4 km. Using the take-off speed 2.0 km/s at the turning point, instead of 1/p, is the usual slip.A 2 × 2 grid has cells of side 10 km, numbered 1 2 over 3 4. Four rays run along row 1 (cells 1, 2), row 2 (cells 3, 4), column 1 (cells 1, 3) and column 2 (cells 2, 4). The travel-time residuals are 0.10, 0.30, 0.20 and 0.20 s, respectively. Cell 1 is known to be unperturbed. Find the slowness perturbation of cell 4 in units of 10⁻³ s/km.
Numerical answer — type the value.
Show answer
Answer: 10
Each ray gives 10 km × (sum of the δs of its cells) = residual. Column 1: 10 (δs₁ + δs₃) = 0.20 with δs₁ = 0 gives δs₃ = 0.020 s/km. Row 2: 10 (δs₃ + δs₄) = 0.30 gives δs₄ = 0.030 − 0.020 = 0.010 s/km = 10 × 10⁻³ s/km. Column 2 checks: 10 (0.010 + 0.010) = 0.20 s.