Incompressible vs. Compressible Navier-Stokes

Short answer: incompressible flow preserves the volume of each fluid parcel and, in the standard homogeneous model, treats density as constant. Compressible flow allows density to change as parcels expand or contract. Pressure varies in both models; incompressible does not mean constant pressure.

That is the practical difference between compressible and incompressible Navier-Stokes equations: one removes density as a dynamical unknown, while the other must track mass, momentum, and energy together. The sections below unpack both systems and explain when each model applies.

Short answer: the standard homogeneous incompressible Navier-Stokes model imposes $\rho=\rho_0$ and $\nabla \cdot u=0$. The compressible Navier-Stokes-Fourier model solves for $\rho(x,t)$ together with momentum and energy. Pressure varies in both systems, but it plays a different mathematical role in each.

Strictly, incompressibility means local volume preservation, $\nabla\cdot u=0$; inhomogeneous incompressible models can advect a spatially varying density. This page compares the homogeneous constant-density system used in the Clay formulation with the full compressible system. These aren't two notations for the same PDE: they have different unknowns, constraints, and mixed elliptic, hyperbolic, and parabolic structures.

The incompressible Navier-Stokes equations describe fluids whose density is constant. They're the version that appears in the Clay Millennium Problem and the version this site focuses on. For a focused treatment, see Incompressible Navier-Stokes Equations.

The system has two parts. The momentum equation:

$$\partial_t u + (u \cdot \nabla)u = -\nabla p + \nu \Delta u + f$$

and the incompressibility constraint:

$$\nabla \cdot u = 0$$

The constraint $\nabla \cdot u = 0$ says the velocity field is divergence-free: fluid neither piles up nor thins out anywhere. Whatever flows into a tiny region must flow out at the same rate. This single condition replaces the entire density equation. Density doesn't change, so you don't need an equation to track it.

Pressure plays a special role here. It isn't determined by a thermodynamic law (like the ideal gas law). Instead, it adjusts instantaneously everywhere to keep the flow divergence-free. Mathematically, $p$ solves a Poisson equation derived from the constraint. Pressure changes propagate infinitely fast. There's no "speed of sound" in incompressible flow.

The incompressible Navier-Stokes system has two unknown fields: velocity $u$ and pressure $p$. That simplicity is deceptive. The nonlinear term $(u \cdot \nabla)u$ still makes the system extremely difficult in three dimensions.

The incompressible Navier-Stokes system on $\mathbb{R}^3$ with kinematic viscosity $\nu > 0$:

$$\partial_t u + (u \cdot \nabla)u = -\nabla p + \nu \Delta u + f, \qquad x \in \mathbb{R}^3,\; t > 0,$$

$$\nabla \cdot u = 0, \qquad u(x,0) = u_0(x).$$

The divergence-free condition $\nabla \cdot u = 0$ is a pointwise constraint, not an evolution equation. It encodes local volume preservation: the flow map $\Phi_t$ satisfies $\det(D\Phi_t) = 1$ for all $t$.

Applying the divergence operator to the momentum equation and using incompressibility yields the pressure Poisson equation:

$$-\Delta p = \partial_i \partial_j (u_i u_j) - \nabla \cdot f.$$

This is an elliptic equation for $p$ at each fixed time. The pressure isn't an independent thermodynamic variable; in the standard PDE interpretation, it acts as a Lagrange multiplier enforcing the divergence-free constraint, determined globally and instantaneously by the velocity field. Information propagates at infinite speed through the pressure, a structural difference from the compressible system that can't be papered over.

The unknowns are $u : \mathbb{R}^3 \times [0,T) \to \mathbb{R}^3$ and $p : \mathbb{R}^3 \times [0,T) \to \mathbb{R}$. For the Clay formulation (Fefferman, 2000), $u_0 \in C^\infty(\mathbb{R}^3)$ is divergence-free and the question is whether $u$ remains in $C^\infty(\mathbb{R}^3 \times [0,\infty))$ with bounded energy.

The compressible Navier-Stokes equations govern flows where density varies. Bigger system. More unknowns. More equations. For the full focused page, see Compressible Navier-Stokes Equations.

You still have a momentum equation, but now density $\rho$ appears explicitly:

$$\partial_t (\rho u) + \nabla \cdot (\rho u \otimes u) = -\nabla p + \nabla \cdot \tau + \rho f$$

The constraint $\nabla \cdot u = 0$ is gone. In its place, you get a continuity equation that tracks how density evolves:

$$\partial_t \rho + \nabla \cdot (\rho u) = 0$$

This says mass is conserved: density changes because the flow compresses or expands fluid parcels.

The system also needs an energy equation and an equation of state, a thermodynamic relation like $p = \rho R T$ (the ideal gas law) that ties pressure to density and temperature. Pressure is no longer a passive enforcer of a constraint. Its acoustic part travels at the speed of sound, while the full viscous system also contains diffusive effects.

The compressible system is essential for aerodynamics at high speeds, astrophysical gas dynamics, combustion, and any flow where density changes matter. But it's a genuinely different mathematical object from the incompressible equations. More unknowns, more equations, different PDE structure entirely.

The compressible Navier-Stokes system couples the velocity $u(x,t)$, density $\rho(x,t)$, pressure $p(x,t)$, and specific internal energy $e(x,t)$ (or temperature $\theta$). In conservation form:

Continuity: $$\partial_t \rho + \nabla \cdot (\rho u) = 0$$

Momentum: $$\partial_t (\rho u) + \nabla \cdot (\rho u \otimes u) + \nabla p = \nabla \cdot \tau + \rho f$$

Energy: $$\partial_t (\rho E) + \nabla \cdot ((\rho E + p)u) = \nabla \cdot (\tau \cdot u) + \nabla \cdot (\kappa \nabla \theta) + \rho f \cdot u$$

where $E = e + \tfrac{1}{2}|u|^2$ is the total specific energy, $\tau$ is the viscous stress tensor (for a Newtonian fluid, $\tau = \mu(\nabla u + \nabla u^T) + \lambda (\nabla \cdot u)I$ with second viscosity coefficient $\lambda$; see Stokes' hypothesis), and $\kappa$ is thermal conductivity.

Closure requires an equation of state, e.g., $p = (\gamma - 1)\rho e$ for an ideal gas with adiabatic index $\gamma$.

In the inviscid or hyperbolic part of the compressible system, acoustic disturbances propagate at finite speed, with characteristic sound speed $c = \sqrt{\partial p / \partial \rho |_s}$. The full viscous Navier-Stokes-Fourier system is mixed hyperbolic-parabolic, so viscous and thermal diffusion introduce infinite propagation mathematically. The inviscid Euler limit admits shock, rarefaction, and contact discontinuities; viscosity and heat conduction replace ideal discontinuities with thin internal layers under standard nondegenerate assumptions.

When does compressibility matter? A useful first screening parameter is the Mach number.

$$\text{Ma} = \frac{|u|}{c}$$

$|u|$ is flow speed and $c$ is the speed of sound. Their ratio measures how fast the flow moves relative to pressure disturbances. For many gas flows without strong heating, combustion, or stratification, $\text{Ma}<0.3$ is a useful engineering rule for treating pressure-driven density changes as small.

It is not a guarantee. Density can vary substantially even at low Mach number because of heating, composition changes, phase change, or buoyancy. Conversely, water in a pipe is often modeled as incompressible because its density changes little under the pressures involved, not because Mach number alone decides the model.

As $\text{Ma}$ approaches one in gas dynamics, acoustic and density effects become prominent and local supersonic regions or shocks may appear. The right choice is therefore based on both Mach number and the physical sources of density variation.

The Mach number $\text{Ma} = |u|/c$ parametrizes pressure-driven compressibility, where $c = \sqrt{\partial p / \partial \rho |_s}$ is the isentropic sound speed. Formally, the incompressible equations arise as the low-Mach limit of the compressible system under additional assumptions.

The asymptotic expansion in powers of $\text{Ma}^2$ (see Klainerman & Majda, 1981, 1982; Schochet, 1986) shows that as $\text{Ma} \to 0$ with suitable initial data, the compressible solutions converge to an incompressible solution. In standard low-Mach nondimensionalizations with well-prepared data, one often writes pressure as a nearly spatially uniform thermodynamic background plus a smaller dynamic correction that enforces the incompressibility constraint in the limit.

This is a singular limit: the sound speed becomes asymptotically large on the flow time scale and the compressible system's hyperbolic pressure dynamics degenerates to the elliptic pressure constraint of incompressible flow. Acoustic modes decouple from the leading vortical dynamics.

The regime $\text{Ma}<0.3$ is an engineering heuristic for flows in which pressure-driven density variation dominates. Isentropic ideal-gas estimates give relative density changes of order $\text{Ma}^2$, motivating the familiar few-percent rule. Low Mach number alone does not justify constant density when heating, combustion, composition, phase change, or buoyancy produces density variations. Rigorous low-Mach convergence also requires suitable thermodynamic scaling, initial data, and boundary conditions.

The Clay Millennium Problem asks a precise question: given a smooth, divergence-free initial velocity on $\mathbb{R}^3$, does the incompressible Navier-Stokes system always produce a smooth solution that exists for all time?

Why incompressible specifically? Three reasons.

First, it's already hard enough. The incompressible 3D equations have resisted proof of global regularity since Leray's foundational work in 1934. Adding variable density, thermodynamics, and shock waves would make the problem vastly harder, not more tractable.

Second, the difficulty is pure fluid mechanics. The incompressible system isolates the core mathematical challenge, the competition between nonlinear advection $(u \cdot \nabla)u$ and viscous dissipation $\nu \Delta u$, without thermodynamic or acoustic complications. It's the cleanest arena to ask the regularity question.

Third, the physics is clean. The incompressible equations model the most common everyday flows. Whether they can produce singularities from smooth data is a fundamental question about the mathematical consistency of classical fluid mechanics.

The compressible system has its own deep open problems (existence of global solutions with large data, formation and interaction of shocks), but those are different problems with different structures. The Clay prize targets the incompressible case because that's the specific regularity question Fefferman formulated for 3D Navier-Stokes.

The official Clay formulation (Fefferman, 2000) specifies the incompressible system on $\mathbb{R}^3$:

$$\partial_t u + (u \cdot \nabla)u = -\nabla p + \nu \Delta u, \qquad \nabla \cdot u = 0, \qquad u|_{t=0} = u_0,$$

with $u_0 \in C^\infty(\mathbb{R}^3)$ divergence-free, and the question is whether $u \in C^\infty(\mathbb{R}^3 \times [0,\infty))$ with $\int_{\mathbb{R}^3} |u(x,t)|^2\,dx$ bounded for all $t \geq 0$.

The choice of the incompressible system is mathematically motivated. The key open problem, the gap between Leray-Hopf weak solutions (which exist globally but may not be unique or smooth) and classical smooth solutions (which exist locally but may blow up), is specific to the incompressible 3D equations. In 2D, global regularity for smooth incompressible Navier-Stokes solutions is known; the 3D case remains open.

The compressible system introduces qualitatively different difficulties: shock formation (which occurs even for Euler equations with smooth data), vacuum states ($\rho \to 0$), and the coupling between vorticity and acoustic modes. These are important open problems, but they're structurally distinct from the incompressible regularity question.

The incompressible problem isolates the competition between the energy-supercritical nonlinearity and viscous dissipation. The natural energy estimate gives $u \in L^\infty_t L^2_x \cap L^2_t \dot{H}^1_x$, which falls short of the scaling-critical space $L^\infty_t \dot{H}^{1/2}_x$ by half a derivative in 3D. Closing this gap, or proving it can't be closed, is the heart of the Millennium Problem.

Start with the focused pages: Incompressible Navier-Stokes Equations and Compressible Navier-Stokes Equations.

Want every term in the main system pulled apart? What Are the Navier-Stokes Equations?

Where does this system come from? Derivation of the Navier-Stokes Equations.

Drop viscosity and you get the Euler equations. Euler vs. Navier-Stokes.

The prize. The Navier-Stokes Existence and Smoothness Problem.

Suggested paths from here:

• Incompressible Navier-Stokes Equations, the constant-density Clay-problem system
• Compressible Navier-Stokes Equations, the density-momentum-energy system
• What Are the Navier-Stokes Equations?, the main incompressible system with term-by-term analysis
• Derivation of the Navier-Stokes Equations, from continuum mechanics axioms to the PDE system
• Euler vs. Navier-Stokes, the inviscid limit $\nu \to 0$ and the role of viscosity in regularity
• The Navier-Stokes Existence and Smoothness Problem, the precise Clay formulation, Leray-Hopf theory, and the gap between weak and strong solutions