Green's Theorem, Gauss' Law, Continuation and the Equivalent Stratum
CSIR NET Earth Sciences Part C: Green's identities, Gauss' law and total excess mass, upward and downward continuation integral, equivalent stratum.
- 18 min
On this page · 13 sections
Prerequisites
- Newtonian Potential, Laplace and Poisson Equations: harmonic functions, the sphere and cylinder anomalies, uniqueness. Everything here builds on that page.
- The divergence theorem and Fourier transforms (see Mathematical Sciences).
- Bouguer anomalies and regional fields from Gravity, Geomagnetism, Heat Flow and Isostasy.
The Hook: A Survey at the Wrong Height
Green’s Theorem and Green’s Identities
Green’s theorem in the plane
For functions \( L(x, y) \) and \( M(x, y) \) with continuous first derivatives on a region \( R \) bounded by a closed curve \( C \) traversed anticlockwise:
$$\boxed{\oint_C (L\, dx + M\, dy) = \iint_R \left(\frac{\partial M}{\partial x} - \frac{\partial L}{\partial y}\right) dx\, dy}$$Taking \( L = -y/2 \), \( M = x/2 \) gives the area of \( R \) as \( \frac{1}{2}\oint (x\,dy - y\,dx) \). This is the 2D form of the divergence (Gauss) theorem, and it is how the gravity of a 2D polygon (the Talwani-type line-integral method) is reduced to a sum over its edges.
The divergence theorem
For a vector field \( \mathbf{F} \) in a volume \( V \) with closed surface \( S \) and outward normal \( \hat{\mathbf{n}} \):
$$\int_V \nabla\cdot\mathbf{F}\, dV = \oint_S \mathbf{F}\cdot\hat{\mathbf{n}}\, dS$$Green’s first and second identities
Put \( \mathbf{F} = \phi\nabla\psi \):
$$\text{First:}\quad \int_V \left(\phi\,\nabla^2\psi + \nabla\phi\cdot\nabla\psi\right) dV = \oint_S \phi\,\frac{\partial\psi}{\partial n}\, dS$$Write the same with \( \phi \) and \( \psi \) swapped and subtract:
$$\text{Second:}\quad \int_V \left(\phi\,\nabla^2\psi - \psi\,\nabla^2\phi\right) dV = \oint_S \left(\phi\,\frac{\partial\psi}{\partial n} - \psi\,\frac{\partial\phi}{\partial n}\right) dS$$Uses you should recognise:
| Choice | Result |
|---|---|
| First identity with \( \phi = \psi = U \), \( U \) harmonic | \( \int_V \lvert\nabla U\rvert^2 dV = \oint U\,\partial U/\partial n\, dS \): proves uniqueness of the Dirichlet and Neumann problems |
| First identity with \( \phi = 1 \) | \( \int_V \nabla^2\psi\, dV = \oint \partial\psi/\partial n\, dS \): Gauss’ law when \( \psi = U \) |
| Second identity with \( \psi = 1/r \) | Green’s third identity (below) |
Uniqueness in one line. If \( U_1 \) and \( U_2 \) are harmonic with the same boundary values, their difference \( W \) is harmonic and zero on \( S \). The first identity gives \( \int_V \lvert\nabla W\rvert^2 dV = 0 \), so \( W \) is constant, and since it is zero on \( S \), \( W = 0 \).
Green’s third identity
Take \( \psi = 1/r \), where \( r \) is the distance from a point \( P \) inside \( V \), cut out a small sphere around \( P \), and let it shrink. For \( U \) harmonic in \( V \):
$$\boxed{U(P) = \frac{1}{4\pi}\oint_S \left[\frac{1}{r}\frac{\partial U}{\partial n} - U\,\frac{\partial}{\partial n}\!\left(\frac{1}{r}\right)\right] dS}$$(the integral is zero for \( P \) outside \( V \), and gives \( \tfrac{1}{2}U(P) \) for \( P \) on a smooth part of \( S \)).
Read it physically. The first term is the potential of a single layer (a surface mass sheet) with surface density proportional to \( \partial U/\partial n \). The second is the potential of a double layer (a sheet of dipoles) with moment proportional to \( U \). So:
That one sentence leads to both continuation and the equivalent stratum.
Gauss’ Law and Total Excess Mass
The closed-surface statement
From Poisson’s equation and the divergence theorem (or directly from the inverse-square law):
$$\oint_S \mathbf{g}\cdot\hat{\mathbf{n}}\, dS = -4\pi G\, M_{\text{enc}}$$The flux depends only on the enclosed mass. Shape, depth and distribution do not enter.
From a closed surface to a survey plane
Enclose all the anomalous mass below the ground in a surface made of the survey plane plus a hemisphere of radius \( R \to \infty \) underneath. Far away, the anomalous field looks like that of a point mass \( M \), so its flux through the hemisphere tends to \( G M/R^2 \times 2\pi R^2 = 2\pi G M \): half of the total \( 4\pi G M \). The other half must pass through the plane. With \( \Delta g \) the vertical anomaly (positive downward attraction):
$$\boxed{\iint_{\text{plane}} \Delta g\; dx\, dy = 2\pi G\, M_{\text{excess}}}$$Check with the sphere: \( \int_0^\infty \frac{GMz}{(s^2+z^2)^{3/2}}\,2\pi s\, ds = 2\pi G M z \cdot \frac{1}{z} = 2\pi G M \). The depth \( z \) cancels.
For a 2D (elongated) body measured along a profile, the same argument with a half-cylinder gives
$$\iint \to \int:\qquad \boxed{\int_{-\infty}^{\infty} \Delta g\; dx = 2\pi G\, \lambda}$$where \( \lambda \) is the excess mass per unit length.
The Continuation Integral
Statement
Let \( U \) be harmonic above a horizontal plane \( z = 0 \) (all sources below), regular at infinity, with \( z \) measured upward. Then its value at height \( h > 0 \) follows from its values on the plane:
$$\boxed{U(x, y, h) = \frac{h}{2\pi}\iint_{-\infty}^{\infty} \frac{U(x', y', 0)}{\left[(x - x')^2 + (y - y')^2 + h^2\right]^{3/2}}\, dx'\, dy'}$$This is the upward continuation integral (the Poisson integral for a half-space). It comes from Green’s third identity with a Green’s function that vanishes on the plane (the image method), so only the boundary values, not the normal derivatives, are needed: a Dirichlet problem.
The 2D (profile) form is
$$U(x, h) = \frac{h}{\pi}\int_{-\infty}^{\infty} \frac{U(x', 0)}{(x - x')^2 + h^2}\, dx'$$Properties of the kernel:
| Property | Consequence |
|---|---|
| Positive, integrates to 1 over the plane | upward continuation is a weighted average: it smooths |
| Width grows with \( h \) | the higher you go, the wider the average |
| Tends to a delta function as \( h \to 0 \) | consistent at \( h = 0 \) |
| Same form for any harmonic quantity | applies to \( U \), \( g_z \), the total-field magnetic anomaly, and their derivatives |
The wavenumber picture
From the Cartesian harmonics \( e^{i(k_x x + k_y y)}e^{-\lvert k\rvert z} \), each Fourier component simply decays with height. The Fourier transform of the kernel is \( e^{-\lvert k\rvert h} \), so the convolution becomes a multiplication:
$$\boxed{\mathcal{F}[U](k, h) = \mathcal{F}[U](k, 0)\; e^{-\lvert k\rvert h}}, \qquad \lvert k\rvert = \frac{2\pi}{\lambda_w}$$| Operation | Filter in the wavenumber domain | Behaviour |
|---|---|---|
| Upward continuation by \( h \) | \( e^{-\lvert k\rvert h} \) | low-pass, stable |
| Downward continuation by \( h \) | \( e^{+\lvert k\rvert h} \) | high-pass, amplifies noise, unstable |
| Vertical derivative \( \partial/\partial z \) | \( \lvert k\rvert \) (up to sign) | enhances short wavelengths |
| Horizontal derivative \( \partial/\partial x \) | \( i k_x \) | enhances edges |
Downward continuation
Continuing towards the sources multiplies each component by \( e^{+\lvert k\rvert h} \). Short-wavelength noise is magnified enormously:
$$\text{noise at } \lambda_w = 1\ \text{km}, \text{ continued down } 0.5\ \text{km}: \quad e^{2\pi(0.5)/1} = e^{\pi} \approx 23$$Downward continuation is therefore an ill-posed problem: small errors in the data produce large errors in the result. It is valid only in the source-free region, so you cannot continue through or below the top of a source (the field there obeys Poisson, not Laplace). Practical schemes damp the high wavenumbers (regularisation) and stop well above the shallowest source. Continuation to an uneven surface (drape to level, level to drape) uses the equivalent-layer idea below.
The Equivalent Stratum
Green’s equivalent layer
Apply Green’s third identity to the region outside a closed surface \( S \) that encloses all the sources, and choose \( S \) to be an equipotential of the body’s own field. On such a surface \( U \) is constant, so the double-layer term contributes nothing outside (its integral over a closed surface with constant moment gives zero external potential), and only a single layer remains. Its surface density is
$$\boxed{\sigma = \frac{1}{4\pi G}\left\lvert\frac{\partial U}{\partial n}\right\rvert = \frac{\lvert\mathbf{g}\rvert_S}{4\pi G}}$$and its total mass equals the body’s mass (Gauss’ law). Outside \( S \), this sheet produces exactly the same potential and field as the original body. More generally, for any closed surface enclosing the sources there is a single layer on it that reproduces the external field (found by solving an integral equation), though for a non-equipotential surface the density has no simple closed form.
This is the equivalent stratum (equivalent layer).
The planar equivalent layer
Most surveys are on a plane. Compare the upward continuation integral for \( g_z \) with the field of a thin sheet of surface density \( \sigma(x', y') \) lying on the plane:
$$g_z^{\text{sheet}}(x, y, h) = G\iint \frac{\sigma(x', y')\, h}{\left[(x-x')^2 + (y-y')^2 + h^2\right]^{3/2}}\, dx'\, dy'$$They are identical if
$$\boxed{\sigma(x, y) = \frac{\Delta g(x, y)}{2\pi G}}$$So a sheet on the survey plane with density \( \Delta g/(2\pi G) \) reproduces the whole anomalous field everywhere above the plane. For 1 mGal, \( \sigma = 10^{-5}/(4.193\times 10^{-10}) = 2.39\times 10^{4} \) kg/m². (The infinite slab in the previous topic is the special case of a uniform sheet: \( 2\pi G\sigma \) on each side.)
Why it matters
| Use | How the equivalent layer helps |
|---|---|
| Ambiguity | A thin layer at the surface and a deep body can give identical data: gravity alone cannot fix depth or thickness. |
| Gridding and interpolation | Fit a layer of point sources to scattered data; evaluate it on a regular grid. |
| Continuation to uneven surfaces | Fit sources to data on a draped survey, recompute on a level surface (or the reverse). |
| Field transformations | From the fitted sources, compute any component, a vertical derivative, or a reduced-to-pole magnetic map. |
| Merging surveys | Bring surveys flown at different heights to one common surface. |
Memory Tricks
Don’t Confuse
| This | vs | That |
|---|---|---|
| Closed-surface Gauss law: flux \( = 4\pi G M \) | survey plane: \( \iint\Delta g\,dS = 2\pi G M \) | |
| Excess mass (fixed by data) | actual mass (needs body and host densities) | |
| Upward continuation: smoothing, stable, \( e^{-\lvert k\rvert h} \) | downward continuation: sharpening, ill-posed, \( e^{+\lvert k\rvert h} \) | |
| Dirichlet data (values) suffice for continuation over a plane | Green’s third identity in general uses values and normal derivatives | |
| Equipotential equivalent layer: \( \sigma = \lvert g\rvert/(4\pi G) \) | planar equivalent layer on the survey plane: \( \sigma = \Delta g/(2\pi G) \) | |
| Single layer: mass sheet, potential continuous, normal field jumps | double layer: dipole sheet, potential itself jumps |
Common Mistakes
Practice
All questions below are original practice questions written for this page. They are not from any CSIR NET paper. Use \( G = 6.674\times 10^{-11} \) SI, \( 2\pi G = 4.193\times 10^{-10} \) SI, 1 mGal \( = 10^{-5} \) m/s².
Level 1: Foundation
- What is the flux of \( \mathbf{g} \) through a closed surface that contains no mass?
- Write the wavenumber-domain filter for upward continuation by a height \( h \).
- What surface density on the survey plane reproduces an anomaly \( \Delta g(x, y) \) above the plane?
Answers
- Zero. Masses outside a closed surface give no net flux.
- \( e^{-\lvert k\rvert h} \), with \( \lvert k\rvert = 2\pi/\lambda_w \).
- \( \sigma = \Delta g/(2\pi G) \).
Level 2: Part B Style
Practice Q1. A residual gravity anomaly over a buried body integrates to \( 2\pi G M \) over the whole survey plane. The quantity \( M \) is:
(a) the actual mass of the body (b) the excess mass relative to the host (c) the mass per unit volume (d) half the excess mass
Answer: (b)
Gauss’ law sees only the anomalous (contrast) mass, independent of shape and depth. (a) needs the densities of body and host. (d) is the \( 4\pi G \) trap.
Practice Q2. A magnetic anomaly component of wavelength 4 km is upward continued by 1 km. Its amplitude becomes about:
(a) 78% of the original (b) 61% (c) 21% (d) 4%
Answer: (c) 21%
\( e^{-2\pi(1)/4} = e^{-\pi/2} = e^{-1.571} = 0.208 \approx 21\% \).
(a) is \( e^{-1/4} \): it forgets the \( 2\pi \).
Practice Q3. Downward continuation of potential-field data is unstable mainly because:
(a) the continuation kernel is negative (b) it multiplies each wavenumber by \( e^{+\lvert k\rvert h} \), amplifying short-wavelength noise (c) Laplace’s equation has no solution below the survey plane (d) the Fourier transform does not exist for gridded data
Answer: (b)
High wavenumbers (short wavelengths, mostly noise) grow exponentially. (c) is wrong: Laplace holds down to the top of the shallowest source.
Practice Q4. Green’s equivalent layer on an equipotential surface enclosing a body has surface density:
(a) \( \lvert g\rvert/(2\pi G) \) (b) \( \lvert g\rvert/(4\pi G) \) (c) \( U/(4\pi G) \) (d) \( 4\pi G \lvert g\rvert \)
Answer: (b)
From Green’s third identity, the single-layer term has density \( (1/4\pi G)\lvert\partial U/\partial n\rvert = \lvert g\rvert/(4\pi G) \) on a closed surface. (a) is the planar layer on the survey plane, where the field above sees a sheet that sends half its flux each way.
Level 3: Part C Style
Practice Q5. A Bouguer residual over a sulphide body integrates to 21 mGal km² over a sufficiently large area. Sulphide density is 3700 kg/m³ and host density 2700 kg/m³. Which statements are correct?
- The excess mass is about \( 5.0 \times 10^{11} \) kg.
- The actual mass of sulphide is about \( 1.85 \times 10^{12} \) kg.
- The volume of the body is about \( 5.0 \times 10^{8} \) m³.
- If the body were twice as deep, the integrated anomaly would be halved.
(a) 1 and 2 only (b) 1, 2 and 3 only (c) 1 and 4 only (d) 1, 2, 3 and 4
Answer: (b) 1, 2 and 3 only
- \( 210\ \text{m}^3/\text{s}^2 \div 4.193\times10^{-10} = 5.01\times 10^{11} \) kg. Correct.
- \( \times\, 3700/1000 = 1.85\times 10^{12} \) kg. Correct.
- Volume \( = M_{\text{excess}}/\Delta\rho = 5.01\times10^{11}/1000 = 5.0\times 10^8 \) m³ (check: \( 1.85\times10^{12}/3700 \) gives the same). Correct.
- The integral is independent of depth: a deeper body gives a lower, broader anomaly with the same area under it. Wrong.
Practice Q6. A gravity profile across a long, buried horizontal body integrates to \( \int \Delta g\, dx = 1.0 \) mGal km. The density contrast is 400 kg/m³. Treating the body as a horizontal cylinder, its radius is closest to:
(a) 70 m (b) 100 m (c) 140 m (d) 240 m
Answer: (c) 140 m
\( \int \Delta g\, dx = 10^{-5} \times 10^{3} = 10^{-2}\ \text{m}^2/\text{s}^2 \).
\( \lambda = \dfrac{10^{-2}}{2\pi G} = \dfrac{10^{-2}}{4.193\times 10^{-10}} = 2.385\times 10^{7} \) kg/m.
Cross-section \( A = \lambda/\Delta\rho = 5.96\times 10^{4} \) m², so \( R = \sqrt{A/\pi} = \sqrt{1.898\times 10^{4}} = 138 \) m \( \approx 140 \) m.
Note what you did not need: the depth. Gauss’ law gives mass per unit length directly.
Practice Q7. A gridded aeromagnetic dataset is to be (i) upward continued by 2 km and (ii) separately downward continued by 0.5 km. The data contain random noise concentrated at wavelengths near 1 km. Which statements are correct?
- In (i), the 1 km noise is reduced by a factor of more than \( 10^{5} \).
- In (ii), the 1 km noise is amplified about 23 times.
- Operation (i) can be carried out by convolving with a kernel that integrates to 1.
- Operation (ii) remains valid even if some magnetic sources lie 0.3 km below the survey surface.
(a) 1, 2 and 3 only (b) 2 and 3 only (c) 1, 2, 3 and 4 (d) 1 and 4 only
Answer: (a) 1, 2 and 3 only
- \( e^{-2\pi(2)/1} = e^{-4\pi} = e^{-12.57} \approx 3.5\times 10^{-6} \): reduction by about \( 2.9\times 10^{5} \). Correct.
- \( e^{+2\pi(0.5)/1} = e^{\pi} = 23.1 \). Correct.
- The upward continuation kernel \( \frac{h}{2\pi}(r^2 + h^2)^{-3/2} \) is positive and integrates to 1. Correct.
- Continuing 0.5 km down passes sources at 0.3 km; below them the field is not harmonic and the operation is invalid. Wrong.
Quick Revision Table
| Idea | Formula or fact |
|---|---|
| Green’s theorem (plane) | \( \oint(L\,dx + M\,dy) = \iint(M_x - L_y)\,dx\,dy \) |
| Green’s second identity | \( \int(\phi\nabla^2\psi - \psi\nabla^2\phi)dV = \oint(\phi\,\partial_n\psi - \psi\,\partial_n\phi)dS \) |
| Green’s third identity | \( U(P) = \frac{1}{4\pi}\oint[\frac{1}{r}\partial_n U - U\partial_n\frac{1}{r}]dS \): single + double layer |
| Gauss, closed surface | flux \( = -4\pi G M_{\text{enc}} \) |
| Gauss, survey plane | \( \iint\Delta g\,dS = 2\pi G M_{\text{excess}} \); profile: \( \int\Delta g\,dx = 2\pi G\lambda \) |
| Actual mass | \( M_{\text{excess}}\,\rho_b/(\rho_b - \rho_h) \) |
| Upward continuation | \( \frac{h}{2\pi}\iint U(x',y',0)\,[r^2+h^2]^{-3/2}dx'dy' \); filter \( e^{-\lvert k\rvert h} \) |
| Downward continuation | \( e^{+\lvert k\rvert h} \): ill-posed, only above sources |
| Equivalent layer, equipotential | \( \sigma = \lvert g\rvert/(4\pi G) \) |
| Equivalent layer, survey plane | \( \sigma = \Delta g/(2\pi G) \approx 2.4\times10^4 \) kg/m² per mGal |
Key Takeaways
What’s Next
- Maxwell’s Equations in Geophysics: from static potentials to time-varying electromagnetic fields, diffusion and skin depth.
- Helmholtz Theorem and Seismic Wave Propagation.
- Field Theory: Quick Revision.
- Back to the Field Theory overview.