The Continuity Equation

Squeeze the end of a garden hose and the water shoots out faster. Watch a river narrow between two rock walls and the current quickens. Blow smoke through a pinched straw and it races. In every case nothing is created and nothing is destroyed — the same mass of fluid per second has to get through a smaller opening, so it has no choice but to speed up. That single, stubborn bookkeeping fact — mass is conserved — is the most reliable statement in all of fluid dynamics, and when we write it down as a differential equation for a flowing continuum it is called the continuity equation.

This page builds that equation from scratch. We fix a little box of space, tally the mass flowing in and out through its faces, and demand that any imbalance show up as the mass inside changing. The divergence theorem turns that tally into a beautifully compact local law, \partial\rho/\partial t + \nabla\cdot(\rho\mathbf{u}) = 0. We will read off its meaning, recast it in convective form using the material derivative, and pin down the single most useful special case — an incompressible flow, for which the velocity field must be divergence-free, \nabla\cdot\mathbf{u} = 0.

Fix a box and count what crosses it

The cleanest way to enforce conservation of mass is to stop chasing the fluid and instead nail down a region of space. Pick any fixed volume V — a control volume — bounded by a closed surface S. Fluid pours through this imaginary box freely; we are going to keep the books on it. Two things, and only two things, can change the mass currently inside:

Reveal the figure step by step: the fixed box, the mass streaming in one face, the mass streaming out another, and the balance that ties them to the mass stored inside.

Put that in symbols. The mass inside the box is \int_V \rho\,dV. The mass leaving per second through a patch of surface with outward normal \mathbf{n} is the density times the velocity component crossing it, \rho\,\mathbf{u}\cdot\mathbf{n}, integrated over the whole surface. Conservation of mass says the rate at which stored mass drops equals the net rate it flows out:

\frac{d}{dt}\int_V \rho\,dV = -\oint_S \rho\,\mathbf{u}\cdot\mathbf{n}\,dS.

This is the integral (control-volume) form of the continuity equation — already complete and exact. The quantity \rho\mathbf{u} is the mass flux (mass per unit area per unit time); its surface integral is the net mass flow rate out of the box.

From the box to a point: the divergence theorem

The integral form is true for every control volume, of every size and shape. That is a huge amount of information, and we can squeeze it down to a statement at a single point. Two moves do it. First, because V is fixed in space, the time derivative slides inside the integral as a partial derivative:

\frac{d}{dt}\int_V \rho\,dV = \int_V \frac{\partial \rho}{\partial t}\,dV.

Second — the key step — the divergence theorem converts the surface flux into a volume integral of the divergence:

\oint_S \rho\,\mathbf{u}\cdot\mathbf{n}\,dS = \int_V \nabla\cdot(\rho\mathbf{u})\,dV.

Substitute both into the balance and gather everything under one integral:

\int_V \left[\frac{\partial \rho}{\partial t} + \nabla\cdot(\rho\mathbf{u})\right]dV = 0.

Here is the punchline. This holds for any region V whatsoever. If a continuous integrand integrated to zero over every possible box, it must be zero everywhere — otherwise draw a tiny box around a point where it isn't zero and the integral there wouldn't vanish. So the bracket itself must be identically zero, and we arrive at the differential form:

The convective form, and what divergence means

There is a second, equally useful way to write the same law. Expand the divergence of the product \rho\mathbf{u} with the product rule, \nabla\cdot(\rho\mathbf{u}) = \mathbf{u}\cdot\nabla\rho + \rho\,\nabla\cdot\mathbf{u}, and drop it into the continuity equation:

\frac{\partial \rho}{\partial t} + \mathbf{u}\cdot\nabla\rho + \rho\,\nabla\cdot\mathbf{u} = 0.

The first two terms are exactly the material derivative of the density — D\rho/Dt = \partial\rho/\partial t + \mathbf{u}\cdot\nabla\rho, the rate of change of density following a parcel. So the law collapses to the compact convective form:

\frac{D\rho}{Dt} + \rho\,\nabla\cdot\mathbf{u} = 0.

This version wears its physical meaning on its sleeve. Rearrange it to \nabla\cdot\mathbf{u} = -\dfrac{1}{\rho}\dfrac{D\rho}{Dt}. Now recall that a parcel of fixed mass m = \rho\,\delta V has D\rho/Dt = -(m/\delta V^2)\,D(\delta V)/Dt, so a short calculation gives

\nabla\cdot\mathbf{u} = \frac{1}{\delta V}\frac{D(\delta V)}{Dt}.

In words: the divergence of the velocity is the fractional rate at which a fluid parcel's volume is expanding. Positive divergence means the parcel is swelling (its density falling); negative divergence means it is being squeezed. That is the concrete meaning of \nabla\cdot\mathbf{u} you should carry everywhere — it is a local volumetric expansion rate, not an abstract piece of notation.

The incompressible special case: a divergence-free field

Liquids, and gases moving well below the speed of sound, barely change their density at all — a parcel of water keeps essentially the same volume no matter how you push it. Such a flow is called incompressible, meaning each parcel's density is constant along its path: D\rho/Dt = 0. Feed that straight into the convective form D\rho/Dt + \rho\,\nabla\cdot\mathbf{u} = 0, and since \rho \neq 0 the second term must vanish on its own:

This is one of the most-used equations in engineering. It is a single scalar constraint that the three velocity components must satisfy together, and it is what makes the flow of water, blood and slow air tractable: whatever else the velocity field does, its divergence has to cancel to zero at every point.

Worked examples

Example 1 — is this field incompressible? A two-dimensional flow is \mathbf{u} = (3x + 2y,\ -3y + 5x). Compute its divergence:

\nabla\cdot\mathbf{u} = \frac{\partial u}{\partial x} + \frac{\partial v}{\partial y} = 3 + (-3) = 0.

The divergence is zero, so this field is incompressible — even though it stretches fluid in one direction, it squeezes it by exactly as much in the other, conserving each parcel's volume. Notice the cross terms (2y and 5x) never entered: only \partial u/\partial x and \partial v/\partial y matter for the divergence.

Example 2 — completing an incompressible field. A flow has u = 4x in the x-direction. What y-component v makes it incompressible (in 2-D)? We need \partial u/\partial x + \partial v/\partial y = 0, and \partial u/\partial x = 4, so we need \partial v/\partial y = -4. The simplest choice is

v = -4y \quad\Longrightarrow\quad \nabla\cdot\mathbf{u} = 4 + (-4) = 0. \checkmark

Example 3 — a pipe that narrows (the hose). Water (incompressible) flows steadily through a pipe whose area shrinks from A_1 = 0.02\ \text{m}^2 to A_2 = 0.005\ \text{m}^2. The integral form over the pipe reduces to A_1 v_1 = A_2 v_2 (equal volume flow rate in and out). If v_1 = 1.5\ \text{m/s}, then

v_2 = \frac{A_1 v_1}{A_2} = \frac{(0.02)(1.5)}{0.005} = 6\ \text{m/s}.

Quarter the area, quadruple the speed. This is the hose-nozzle rule, and it is nothing but continuity applied to a control volume wrapping the pipe.

Example 4 — a compressible nozzle. Now let the fluid be a gas whose density changes. Steady continuity through a duct reads \rho_1 A_1 v_1 = \rho_2 A_2 v_2 (equal mass flow rate). If air enters at \rho_1 = 1.2\ \text{kg/m}^3,\ A_1 = 0.10\ \text{m}^2,\ v_1 = 40\ \text{m/s} and leaves a section where \rho_2 = 0.80\ \text{kg/m}^3,\ A_2 = 0.10\ \text{m}^2, then

v_2 = \frac{\rho_1 A_1 v_1}{\rho_2 A_2} = \frac{(1.2)(0.10)(40)}{(0.80)(0.10)} = 60\ \text{m/s}.

Because the gas thinned out, it must speed up even though the area stayed the same — the mass flux \rho v is what is conserved, not the speed.

No — and this is the classic trap. Incompressible means each fluid parcel keeps its own density as it moves, D\rho/Dt = 0; it does not mean the density is uniform in space (\nabla\rho = 0). The ocean is very nearly incompressible, yet warm surface water is genuinely less dense than the cold deep — the density varies from place to place. What is forbidden is a parcel changing its density as it drifts along.

The velocity-side statement is the clean one to lean on: incompressible is exactly \nabla\cdot\mathbf{u} = 0, a divergence-free velocity field. That follows from D\rho/Dt = 0 in the convective continuity equation, and it holds whether or not the density happens to be uniform. Only in the further, stronger case of a constant-density fluid do the two ideas coincide.

The continuity equation is the template for every conservation law in physics: a stored quantity plus the divergence of its flux equals a source, \partial_t(\text{stuff}) + \nabla\cdot(\text{flux}) = \text{source}. For heat you can add a source (a flame, a resistor); for electric charge in a reacting plasma you might too. But mass has no sources in ordinary fluid mechanics — you cannot create or annihilate matter by stirring water — so the right-hand side is exactly zero. That single "0" is a deep statement: it is the fluid-dynamical face of the conservation of mass, the same principle Lavoisier weighed out in the 1770s. (Push to relativistic or nuclear regimes and mass-energy can convert, and even this equation grows a term — but for a garden hose, zero it is.)