Calculus

Divergence Theorem

Цели урока

  • State and apply the Divergence Theorem to compute flux through surfaces
  • Understand divergence as source density and its physical meaning in electrodynamics and fluid mechanics
  • Know the continuity equation as a consequence of the divergence theorem
  • See the theorem as the 3D case of the generalized Stokes theorem

Предварительные знания

  • Green's Theorem
  • Surface integrals
  • Divergence and curl
  • Green's Theorem

How can the total net outward flow through a closed surface be computed without integrating over every square centimeter of that surface?

  • NVIDIA PhysX in 4000 game engines: divergence theorem converts expensive surface integrals to volume integrals, cutting fluid simulation cost by up to 100x per frame
  • Gauss's law: oiint E.dS = Q_enclosed/eps_0 - electric flux through any surface determined solely by enclosed charge
  • OpenFOAM and Airbus CFD: Finite Volume Method discretizes divergence theorem to model aerodynamics on millions of cells
  • FNO neural operators: learn to solve Navier-Stokes at 1000x speedup over traditional solvers by implicitly encoding conservation laws from divergence theorem

Three independent discoveries

Mikhail Ostrogradsky (1801-1862) proved the theorem in 1826 and presented it to the Paris Academy of Sciences in 1828. Carl Friedrich Gauss published a version around 1813 for special cases. George Green proved the 2D case in 1828. Three people, three formulations, one idea. The theorem is known by different names in different traditions: Gauss's theorem in physics, Ostrogradsky's theorem in Russian mathematics, the Divergence Theorem in modern analysis. All three names refer to the same result.

The Divergence Theorem

1831. Ostrogradsky proves the theorem in Saint Petersburg. Gauss publishes a version in Germany the same year. Green had the 2D case three years earlier. Three countries, three people, one idea. NVIDIA PhysX, running inside over 4000 game engines in 2023, replaces expensive surface integrals with volume divergence integrals using this theorem - cutting fluid simulation cost by up to 100x per frame.

Gauss's law in electrostatics: div E = rho / epsilon_0 (Maxwell's equation). By the divergence theorem: oiint_S E dS = Q_enclosed / epsilon_0. For a spherically symmetric charge Q: E = Q/(4*pi*epsilon_0 r^2). The surface shape does not matter - only the enclosed charge.

Gauss's law for gravity

Gravitational flux depends only on enclosed mass

Gravitational field of mass M: g = -GM/r^2 * r_hat. div g = 0 everywhere except r=0. By the divergence theorem: for any surface enclosing M, flux = -4*pi*G*M. For a surface not enclosing M: flux = 0. The shape of the surface is irrelevant - only the enclosed mass matters. This is why orbital mechanics needs only the mass inside the orbit, not the exact mass distribution.

Field F = (x, 0, 0). Flux through the unit sphere?

div(x,0,0) = 1. Divergence theorem: flux = iiint_V 1 dV = volume of unit ball = 4*pi/3.

Divergence in Physics: Maxwell's Equations

The four Maxwell's equations contain two divergence conditions: div E = rho/eps_0 and div B = 0. The first says electric field lines originate from charges. The second says magnetic field lines never begin or end - there are no magnetic monopoles. Both conditions are stated in differential form; the divergence theorem converts them to integral form, which is how they are measured experimentally.

The divergence theorem requires F to have continuous partial derivatives in the closed volume V. If F has a singularity inside V (like the gravitational field of a point mass at the origin), exclude the singularity by a small sphere and take the limit.

Identity: div(curl F) = 0 for any smooth F. This follows from d^2 = 0 in the language of differential forms. Consequence: magnetic field lines never end - if B = curl A (vector potential), then div B = div(curl A) = 0 automatically.

div B = 0 for the magnetic field means:

Divergence theorem: oiint B dS = iiint div B dV = 0. No sources or sinks of magnetic flux - magnetic field lines form closed loops.

Finite Volume Method: Discrete Divergence Theorem

The Finite Volume Method (FVM) is the divergence theorem applied to each mesh cell. The domain is split into small control volumes. For each cell: iiint_V div F dV = oiint_{faces} F.n dS. Replace the volume integral with the source term and the surface integral with flux between cells. This is the exact discrete divergence theorem. OpenFOAM uses this for 3D CFD simulations of aerodynamics. Airbus A350 aerodynamic design relied on it.

Heat equation via divergence theorem

Fourier's law and conservation of energy

Heat flux q = -k grad T (Fourier's law). By the divergence theorem: d/dt iiint rho*c*T dV = -oiint q.n dS = -iiint div q dV = iiint k Delta T dV. Locally: rho*c dT/dt = k Delta T. This is the heat equation. Every CFD solver and FEM code for thermal analysis derives the equation this way - from the divergence theorem applied to energy conservation.

Neural operators (FNO - Fourier Neural Operator, from Anandkumar et al. 2021) learn to solve PDEs by operating in Fourier space. Under the hood, they implicitly encode the divergence theorem: the conservation laws that define the PDE. FNO achieves 1000x speedup over numerical solvers for Navier-Stokes on 64x64 grids.

In the Finite Volume Method, the divergence theorem is used to:

FVM: for each cell V_i, apply divergence theorem: iiint div F dV = oiint F.n dS. Replace volume source with known forcing; replace surface fluxes with interpolated face values. This is the discrete divergence theorem.

div(curl F) = 0 and the Hierarchy

The identity div(curl F) = 0 is not a coincidence - it is d^2 = 0 in disguise. In the language of differential forms: gradient = d on 0-forms, curl = *d on 1-forms, divergence = *d* on 1-forms. The composition div(curl) = *d**d* = *d^2* = 0. Three vector calculus identities collapse into one algebraic fact.

The four Maxwell equations in compact form using differential forms: dF = 0 and d*F = J, where F is the electromagnetic 2-form on 4D spacetime. The divergence theorem and Stokes theorem are both hidden inside the operator d.

Why is div(curl F) = 0 for any smooth F?

div(curl F) = 0 follows from both d^2 = 0 (differential forms language) and Schwarz's theorem (mixed partials are equal). Two ways of seeing the same algebraic fact.

Connections to other topics

The Divergence Theorem appears throughout mathematical physics, from Maxwell's equations to fluid mechanics to numerical methods.

  • Electrodynamics — Related topic
  • Fluid mechanics — Related topic
  • Numerical methods — Related topic
  • Gravity — Related topic

Итоги

  • Divergence theorem: oiint_{boundary V} F.dS = iiint_V div F dV - surface flux = volume integral of divergence
  • Divergence at a point = limit of flux through a small sphere / volume - coordinate-independent definition
  • div B = 0: flux of B through any closed surface = 0 - no magnetic monopoles
  • div(curl F) = 0: follows from d^2 = 0 - Schwarz theorem on mixed partial derivatives
  • FVM: discrete divergence theorem on each mesh cell - foundation of CFD in aviation and engineering

Вопросы для размышления

  • Gauss's law says the flux through any surface surrounding a charge depends only on the charge, not the surface shape. What does this reveal about the structure of 1/r^2 fields?
  • The continuity equation drho/dt + div(rho v) = 0 expresses conservation of mass. How does the divergence theorem connect the differential form to the integral conservation statement?
  • Why does the identity div(curl F) = 0 mean that any curl field can serve as the magnetic field B in Maxwell's equations - satisfying div B = 0 automatically?

Связанные уроки

  • calc-25-green-theorem — Green's theorem is the 2D predecessor
  • calc-27-diff-forms — Differential forms express this as int_M d omega = int_{boundary} omega in 3D
  • calculus-27 — Classical Stokes theorem is the complementary result
  • calc-28-manifold-int
  • la-04-matrix-ops
Divergence Theorem

0

1

Sign In