09.07.04 · classical-mech / continuum

Viscous Flow: The Navier-Stokes Equations, Reynolds Number, and Laminar-Turbulent Transition

shipped3 tiersLean: none

Anchor (Master): Foias, Manley, Rosa & Temam, Navier-Stokes Equations and Turbulence (2001); Frisch, Turbulence: The Legacy of A. N. Kolmogorov (1995)

Intuition Beginner

Pour honey and water from the same height. The water flows freely; the honey oozes. Both are liquids, but honey resists internal motion far more than water. This resistance is viscosity -- the internal friction of a fluid. Every real fluid has viscosity. The ideal fluid studied in the previous unit 09.07.03 is a mathematical simplification that ignores it. Adding viscosity back transforms the Euler equation into the Navier-Stokes equations, which govern almost every real fluid flow -- from blood in arteries to air over aircraft wings to magma beneath volcanoes.

Viscosity arises because adjacent layers of fluid moving at different speeds exchange momentum. Imagine a stack of playing cards on a table. Put your hand on the top card and slide it sideways. The top card drags the one below it, which drags the one below that, and so on. Each card experiences a frictional force from its neighbours. A fluid works the same way: if one layer moves faster than the layer next to it, the faster layer pulls the slower one forward and the slower one holds the faster one back. This shearing interaction is viscosity.

In 1883, Osborne Reynolds performed a decisive experiment. He set up a long glass pipe carrying water and injected a thin stream of dye at the centre of the inlet. At low flow speeds, the dye formed a smooth, straight thread running the length of the pipe -- the flow was orderly and layered, each cylindrical layer of water sliding past its neighbours without mixing. Reynolds called this laminar flow (from the Latin lamina, meaning layer).

When he increased the flow speed past a critical value, the dye thread suddenly broke up, swirling and mixing with the surrounding water in chaotic, irregular motion. This is turbulent flow. The transition depends on a single dimensionless quantity: the ratio of inertial forces (which tend to keep the flow moving and amplify disturbances) to viscous forces (which tend to smooth them out). This ratio is the Reynolds number:

R e = \frac{ρ v L}{μ},

where $ρ$ is the fluid density, $v$ is a characteristic speed, $L$ is a characteristic length (the pipe diameter, say), and $μ$ is the dynamic viscosity (a material property of the fluid). At low $R e$ , viscous forces dominate and the flow is laminar. At high $R e$ , inertial forces dominate and the flow becomes turbulent. For pipe flow, the transition typically occurs around $R e \approx 2300$ . Honey, with its high viscosity, has a low Reynolds number in most everyday situations and flows laminarly. Air, with its low viscosity, almost always has a high Reynolds number and flows turbulently.

The Navier-Stokes equations are Newton's second law for a viscous fluid. They say: the mass of a fluid element times its acceleration equals the sum of the pressure force, the gravitational force, and the viscous force. The viscous force is new -- it was absent in the Euler equation for ideal fluids 09.07.03 -- and it is the term that makes honey different from water.

The viscous force is proportional to the viscosity $μ$ and to second derivatives of the velocity field: it depends on how fast the velocity changes from point to point. Where the velocity varies sharply (near a wall, or in a thin boundary layer), the viscous force is strong. Where the velocity is nearly uniform (far from boundaries in a fast flow), the viscous force is negligible and the flow behaves almost ideally.

Four take-aways before the details:

Viscosity is internal friction. It resists shearing motion between adjacent fluid layers.
The Navier-Stokes equations extend the Euler equation by adding viscous stresses to the momentum balance.
The Reynolds number $R e = ρ v L / μ$ determines whether viscous or inertial effects dominate.
Laminar flow (low $R e$ , smooth and orderly) transitions to turbulent flow (high $R e$ , chaotic and mixing) at a critical Reynolds number.

Visual Beginner

Concept	Physical meaning	Key formula
Dynamic viscosity $μ$	Internal friction of the fluid	Shear stress $= μ \times$ velocity gradient
Reynolds number $R e$	Ratio of inertial to viscous forces	$R e = ρ v L / μ$
Laminar flow	Smooth, layered, orderly	Dye thread stays coherent
Turbulent flow	Chaotic, mixing, irregular	Dye thread breaks up and disperses

Worked example Beginner

A horizontal pipe of diameter 2 cm carries water at a mean speed of 0.5 m/s. The dynamic viscosity of water is $μ = 1.0 \times 1 0^{- 3} Pa s$ and the density is $ρ = 1000 kg/m^{3}$ . Is the flow laminar or turbulent?

The Reynolds number is $R e = ρ v D / μ = 1000 \times 0.5 \times 0.02/0.001 = 10, 000$ . This is well above the critical value of approximately 2300 for pipe flow, so the flow is turbulent. If the pipe carried honey ( $μ \approx 2 Pa s$ , $ρ \approx 1400 kg/m^{3}$ ) at the same speed, the Reynolds number would be $R e = 1400 \times 0.5 \times 0.02/2 = 7$ -- deeply laminar. The same geometry, the same speed, but a different fluid changes the flow character entirely.

Check your understanding Beginner

Formal definition Intermediate+

The viscous stress tensor

In an ideal fluid, the stress is purely isotropic: $σ_{ij} = - p δ_{ij}$ , where $p$ is the pressure 09.07.03. A Newtonian viscous fluid adds a stress proportional to the rate of deformation. The full Cauchy stress tensor is:

σ_{ij} = - p δ_{ij} + η (\frac{\partial v _{i}}{\partial x _{j}} + \frac{\partial v _{j}}{\partial x _{i}} - \frac{2}{3} δ_{ij} \frac{\partial v _{k}}{\partial x _{k}}) + ζ δ_{ij} \frac{\partial v _{k}}{\partial x _{k}},

where $η$ (often written $μ$ ) is the dynamic viscosity (shear viscosity) and $ζ$ is the bulk viscosity (volume viscosity). The term proportional to $η$ is the deviatoric stress (shape-changing, shear) and the term proportional to $ζ$ is the volumetric stress (compression and expansion). For an incompressible fluid, $\partial v_{k} / \partial x_{k} = \nabla \cdot v = 0$ and the stress simplifies to:

σ_{ij} = - p δ_{ij} + η (\frac{\partial v _{i}}{\partial x _{j}} + \frac{\partial v _{j}}{\partial x _{i}}) .

The viscous part $τ_{ij} = η (\partial v_{i} / \partial x_{j} + \partial v_{j} / \partial x_{i})$ is the viscous stress tensor. It is symmetric (six independent components) and traceless for incompressible flow.

The Navier-Stokes equations

Substituting the Newtonian stress tensor into Cauchy's momentum equation $ρ D v_{i} / D t = \partial σ_{ij} / \partial x_{j} + ρ g_{i}$ gives the Navier-Stokes equations for a compressible viscous fluid:

ρ \frac{D v _{i}}{D t} = - \frac{\partial p}{\partial x _{i}} + \frac{\partial}{\partial x _{j}} [η (\frac{\partial v _{i}}{\partial x _{j}} + \frac{\partial v _{j}}{\partial x _{i}} - \frac{2}{3} δ_{ij} \frac{\partial v _{k}}{\partial x _{k}})] + \frac{\partial}{\partial x _{i}} (ζ \frac{\partial v _{k}}{\partial x _{k}}) + ρ g_{i} .

For an incompressible fluid with constant viscosity $η$ , this reduces to:

ρ (\frac{\partial v}{\partial t} + (v \cdot \nabla) v) = - \nabla p + η \nabla^{2} v + ρ g .

The term $η \nabla^{2} v$ is the viscous force per unit volume. It is the Laplacian of the velocity field multiplied by the viscosity, and it represents the diffusion of momentum due to viscous shearing. This is the only term that distinguishes the Navier-Stokes equation from the Euler equation 09.07.03. The viscous term introduces dissipation: kinetic energy is irreversibly converted into heat. This is the thermodynamic cost of internal friction.

The incompressible limit

The incompressible Navier-Stokes system consists of two equations for two unknowns ( $v$ and $p$ ):

\nabla \cdot v = 0 (incompressibility),

ρ \frac{D v}{D t} = - \nabla p + η \nabla^{2} v + ρ g (momentum) .

The pressure $p$ no longer has an independent thermodynamic meaning; it is determined implicitly by the incompressibility constraint. Mathematically, $p$ is a Lagrange multiplier enforcing $\nabla \cdot v = 0$ . This is why the pressure in an incompressible flow adjusts instantaneously to satisfy the divergence-free condition: pressure waves propagate at infinite speed in the incompressible limit.

The Reynolds number and nondimensionalisation

Define dimensionless variables by scaling lengths by $L$ , velocities by $U$ , time by $L / U$ , and pressure by $ρ U^{2}$ . Substituting into the incompressible Navier-Stokes equation yields:

\frac{\partial v ^{*}}{\partial t ^{*}} + (v^{*} \cdot \nabla^{*}) v^{*} = - \nabla^{*} p^{*} + \frac{1}{R e} \nabla^{* 2} v^{*} + \frac{1}{F r ^{2}} \overset{z}{^},

where $R e = ρ U L / η$ is the Reynolds number and $F r = U / g L$ is the Froude number. The single parameter $R e$ controls the ratio of inertial to viscous effects:

$R e ≪ 1$ : viscous forces dominate (creeping flow, e.g., micro-organisms in water, sediment in oil).
$R e \sim 1$ : inertial and viscous forces are comparable.
$R e ≫ 1$ : inertial forces dominate; viscous effects are confined to thin boundary layers near walls.

Two geometrically similar flows with the same Reynolds number (and Froude number) are dynamically similar: the dimensionless velocity and pressure fields are identical. This is the basis of scale-model testing in wind tunnels and water channels. A 1:10 scale model of an aircraft tested at ten times the wind speed (or with a fluid of different viscosity chosen to match $R e$ ) reproduces the full-scale flow pattern.

Energy dissipation

Taking the dot product of the Navier-Stokes equation with $v$ and integrating over the flow domain gives the energy equation:

\frac{d}{d t} \int \frac{1}{2} ρ ∣ v ∣^{2} d V = - \int Φ d V + (boundary flux terms),

where $Φ = η (\frac{\partial v _{i}}{\partial x _{j}} + \frac{\partial v _{j}}{\partial x _{i}})^{2} /2$ is the viscous dissipation function (always positive). Kinetic energy is monotonically converted to heat by viscosity. In a closed system with no external forcing, the flow eventually comes to rest. This is why real fluids always stop moving if left alone -- unlike ideal fluids, which conserve kinetic energy indefinitely.

Boundary conditions

Viscosity enforces the no-slip condition: the fluid velocity matches the wall velocity at every point on a solid boundary. For a stationary wall: $v = 0$ on the wall. This contrasts with the ideal-fluid case 09.07.03, where only the no-penetration condition ( $v \cdot \overset{n}{^} = 0$ ) is imposed and the fluid can slip along the wall. The no-slip condition is the source of boundary layers: at high $R e$ , the flow adjusts from the free-stream velocity to zero across a thin layer of thickness $δ \sim L / R e$ .

Key derivation Intermediate+

Derivation of the Navier-Stokes equation from the stress tensor

Start from Cauchy's momentum equation for a continuum 09.07.01:

ρ \frac{D v _{i}}{D t} = \frac{\partial σ _{ij}}{\partial x _{j}} + ρ g_{i} .

Insert the Newtonian viscous stress tensor for an incompressible fluid: $σ_{ij} = - p δ_{ij} + η (\partial v_{i} / \partial x_{j} + \partial v_{j} / \partial x_{i})$ . Computing the divergence of the stress:

\frac{\partial σ _{ij}}{\partial x _{j}} = - \frac{\partial p}{\partial x _{i}} + η \frac{\partial}{\partial x _{j}} (\frac{\partial v _{i}}{\partial x _{j}} + \frac{\partial v _{j}}{\partial x _{i}}) .

The second term: $η \frac{\partial ^{2} v _{i}}{\partial x _{j} \partial x _{j}} + η \frac{\partial}{\partial x _{i}} (\frac{\partial v _{j}}{\partial x _{j}})$ . For incompressible flow, $\partial v_{j} / \partial x_{j} = 0$ , so the second piece vanishes, leaving:

\frac{\partial σ _{ij}}{\partial x _{j}} = - \frac{\partial p}{\partial x _{i}} + η \nabla^{2} v_{i} .

In vector notation: $- \nabla p + η \nabla^{2} v$ . Substituting into Cauchy's equation gives the incompressible Navier-Stokes equation. The viscous term $η \nabla^{2} v$ arises from the assumption that the deviatoric stress is linearly proportional to the strain rate (the Newtonian hypothesis). Fluids obeying this linear relation are Newtonian fluids; water, air, and most simple liquids and gases are Newtonian to an excellent approximation. Non-Newtonian fluids (polymer solutions, blood, ketchup, mud) have more complicated constitutive relations and lie beyond the scope of this unit.

Poiseuille flow: exact solution for laminar pipe flow

Consider steady, fully developed, pressure-driven flow through a straight circular pipe of radius $R$ . By symmetry, the velocity is purely axial and depends only on the radial coordinate $r$ : $v = v (r) \overset{z}{^}$ . The Navier-Stokes equation reduces to a balance between the pressure gradient and the viscous force:

- \frac{d p}{d z} + η \frac{1}{r} \frac{d}{d r} (r \frac{d v}{d r}) = 0.

The pressure gradient $- d p / d z = Δ p / L$ is a constant (the pressure drops uniformly along the pipe). Integrating twice with boundary conditions $v (R) = 0$ (no-slip) and $d v / d r (0) = 0$ (symmetry on the axis) gives the Poiseuille velocity profile:

v (r) = \frac{Δ p}{4 η L} (R^{2} - r^{2}) .

The profile is a paraboloid: fastest on the axis, zero at the wall. The maximum velocity (at $r = 0$ ) is $v_{m a x} = Δ p R^{2} / (4 η L)$ . The volume flow rate is:

Q = \int_{0}^{R} v (r) 2 π r d r = \frac{π R ^{4} Δ p}{8 η L} .

This is Poiseuille's law (Hagen-Poiseuille law): the flow rate is proportional to the fourth power of the pipe radius and inversely proportional to the viscosity. Halving the pipe radius reduces the flow rate by a factor of 16. This result holds for laminar flow only; turbulent flow has a different (and higher) resistance law.

Stokes drag on a sphere

For a rigid sphere of radius $a$ moving at slow speed $U$ through a viscous fluid at $R e ≪ 1$ , the Stokes approximation drops the inertial terms from the Navier-Stokes equation, leaving the Stokes equations:

\nabla p = η \nabla^{2} v, \nabla \cdot v = 0.

The drag force on the sphere is:

F = 6 π η a U .

This is Stokes' law (1851). The drag is proportional to the velocity (not its square), a hallmark of the viscous-dominated regime. The drag coefficient $C_{D} = F / (\frac{1}{2} ρ U^{2} π a^{2}) = 24/ R e$ diverges as $R e \to 0$ , confirming that inertia is irrelevant in this limit. Stokes drag applies to sediment grains settling in water, fog droplets falling through air, and micro-organisms swimming -- all situations where $R e ≪ 1$ .

The Blasius boundary layer

At high Reynolds number, viscous effects are confined to a thin boundary layer of thickness $δ$ near a solid wall. For steady, incompressible flow at speed $U$ past a flat plate, the boundary layer thickness grows with distance $x$ from the leading edge as:

δ (x) \approx 5 \frac{η x}{ρ U} = \frac{5 x}{R e _{x}},

where $R e_{x} = ρ U x / η$ is the local Reynolds number based on $x$ . The Blasius similarity solution (1908) reduces the Navier-Stokes equation to a single ordinary differential equation for the similarity variable $η = y ρ U / (2 η x)$ :

f^{'''} + f f^{''} = 0, f (0) = f^{'} (0) = 0, f^{'} (\infty) = 1,

where $u = U f^{'} (η)$ . This ODE has no closed-form solution but is readily integrated numerically. The wall shear stress is $τ_{w} = 0.332 ρ U^{2} / R e_{x}$ , and the total drag on a plate of length $L$ and width $b$ is:

D = 1.328 b ρ η U^{3} L .

The Blasius solution is the foundational example of boundary-layer theory, introduced by Prandtl in 1904, which resolved d'Alembert's paradox (zero drag in potential flow 09.07.03) by showing that viscous effects, however small, always produce a thin layer near walls where the flow adjusts from the free-stream velocity to zero.

Worked example: Poiseuille flow in a capillary tube

Blood flows through a capillary of diameter 8 $μ$ m at a mean speed of 0.5 mm/s. The viscosity of blood is approximately $η = 3.5 \times 1 0^{- 3}$ Pa s and the density is $ρ = 1060$ kg/m $^{3}$ . Find the Reynolds number and the pressure drop over a length of 1 mm.

The Reynolds number: $R e = ρ v D / η = 1060 \times 5 \times 1 0^{- 4} \times 8 \times 1 0^{- 6} / (3.5 \times 1 0^{- 3}) = 1.2 \times 1 0^{- 3}$ . The flow is deeply laminar -- inertial effects are negligible. From Poiseuille's law: $Δ p = 8 η L Q / (π R^{4})$ . With $Q = v_{mean} \times π R^{2}$ and $v_{mean} = v_{m a x} /2$ for Poiseuille flow: $Δ p = 8 η L v_{mean} / R^{2} = 8 \times 3.5 \times 1 0^{- 3} \times 1 0^{- 3} \times 5 \times 1 0^{- 4} / (4 \times 1 0^{- 6})^{2} = 1.4 \times 1 0^{- 14} /1.6 \times 1 0^{- 11} = 875$ Pa $\approx$ 6.6 mmHg. This is consistent with measured pressure drops across capillary beds in the human body.

Bridge. The Navier-Stokes equations complete the momentum balance by adding viscous stress to the Euler equation. The structural relationship is: the Euler equation is the $R e \to \infty$ limit of the Navier-Stokes equation; the Stokes equation is the $R e \to 0$ limit. Between these two extremes, the full nonlinear equation governs. The Reynolds number is the single parameter that interpolates between the viscous-dominated and inertia-dominated regimes. At high $R e$ , the flow is nearly ideal in the bulk but viscous in thin boundary layers, and the transition from the boundary layer to the free stream is governed by the Prandtl equations (a singular perturbation of the Navier-Stokes equation). The laminar-to-turbulent transition in the boundary layer is governed by linear stability analysis (Orr-Sommerfeld equation) and by nonlinear secondary instabilities, connecting to the broader theory of dynamical systems and chaos.

Exercises Intermediate+

Exercise 5 (medium, symbolic).

Derive the energy dissipation identity for the incompressible Navier-Stokes equations (without body forces): $d E / d t = - \int Φ d V$ where $E = \frac{1}{2} \int ρ ∣ v ∣^{2} d V$ is the total kinetic energy and $Φ = η \sum_{i, j} (\partial v_{i} / \partial x_{j} + \partial v_{j} / \partial x_{i})^{2} /2$ is the viscous dissipation function. Assume $v = 0$ on the boundary.

Hint

Take the dot product of the Navier-Stokes equation with $v$ , integrate over the domain, and use the divergence theorem with the boundary condition to eliminate boundary terms.

Answer

$\frac{d E}{d t} = \int ρ v \cdot \frac{D v}{D t} d V = \int v \cdot (- \nabla p + η \nabla^{2} v) d V$ . The pressure term: $\int - v \cdot \nabla p d V = \int - \nabla \cdot (p v) d V + \int p \nabla \cdot v d V$ . The first integral vanishes by the divergence theorem and $v = 0$ on the boundary. The second vanishes by incompressibility. The viscous term: $\int η v \cdot \nabla^{2} v d V = \int η [\nabla \cdot (v \cdot \nabla v) - ∣\nabla v ∣^{2}] d V$ . The divergence term vanishes by the boundary condition. After careful expansion using incompressibility, the remaining term becomes $- \int Φ d V$ . Hence $d E / d t = - \int Φ d V \leq 0$ : kinetic energy is always dissipated.

Exercise 6 (hard, symbolic).

Couette flow: a viscous fluid fills the gap between two concentric cylinders of radii $R_{1}$ (inner, rotating at angular velocity $Ω$ ) and $R_{2}$ (outer, stationary). The flow is steady, axisymmetric, and purely azimuthal: $v = v_{θ} (r) \hat{θ}$ . Solve the Navier-Stokes equation for $v_{θ} (r)$ and find the torque per unit length that must be applied to the inner cylinder to maintain the rotation. This torque is the basis of the Couette viscometer, a device for measuring viscosity.

Hint

Write the $θ$ -component of the Navier-Stokes equation in cylindrical coordinates for purely azimuthal flow. The equation reduces to $d (r^{2} d v_{θ} / d r) / d r = 0$ after using the $r$ -component to determine the pressure. Integrate with $v_{θ} (R_{1}) = Ω R_{1}$ and $v_{θ} (R_{2}) = 0$ .

Answer

The $r$ -momentum equation gives $\partial p / \partial r = ρ v_{θ}^{2} / r$ (centripetal acceleration balanced by the radial pressure gradient). The $θ$ -momentum equation (steady, no axial dependence) reduces to:

\frac{d}{d r} (r^{2} \frac{d}{d r} (\frac{v _{θ}}{r})) = 0.

Wait -- more directly, the azimuthal component of the viscous term in cylindrical coordinates for $v_{θ} (r)$ only is $η (\frac{d}{d r} (\frac{1}{r} \frac{d ( r v _{θ} )}{d r}))$ , and the momentum equation is $\frac{d}{d r} (\frac{1}{r} \frac{d ( r v _{θ} )}{d r}) = 0$ . Integrating twice: $v_{θ} (r) = A r + B / r$ . Boundary conditions: $v_{θ} (R_{1}) = Ω R_{1}$ gives $A R_{1} + B / R_{1} = Ω R_{1}$ ; $v_{θ} (R_{2}) = 0$ gives $A R_{2} + B / R_{2} = 0$ . Solving: $A = - Ω R_{1}^{2} / (R_{2}^{2} - R_{1}^{2})$ and $B = Ω R_{1}^{2} R_{2}^{2} / (R_{2}^{2} - R_{1}^{2})$ . The torque per unit length on the inner cylinder: $T = 2 π R_{1} \times η r d (v_{θ} / r) / d r ∣_{R_{1}} \times R_{1} = 4 π η Ω R_{1}^{2} R_{2}^{2} / (R_{2}^{2} - R_{1}^{2})$ . Measuring the torque gives $η$ , which is the operating principle of the Couette viscometer.

Exercise 7 (hard, symbolic).

Consider the linear stability of plane Poiseuille flow $U (y) = U_{m a x} (1 - y^{2} / h^{2})$ between parallel plates at $y = \pm h$ . The Orr-Sommerfeld equation for a two-dimensional perturbation of the form $ψ^{'} = ϕ (y) e^{i α (x - c t)}$ is:

(U - c) (ϕ^{''} - α^{2} ϕ) - U^{''} ϕ = \frac{1}{i α R e} (ϕ^{''''} - 2 α^{2} ϕ^{''} + α^{4} ϕ),

where $R e$ is based on the half-channel height and the centreline velocity. Show that for $R e \to \infty$ , the right side vanishes and the equation reduces to the Rayleigh stability equation. Then explain why the presence of viscosity (finite $R e$ ) can be destabilising, despite the intuition that viscosity should damp perturbations.

Hint

The Rayleigh equation is inviscid and requires an inflection point in $U (y)$ for instability (Rayleigh's inflection-point theorem). Plane Poiseuille flow has no inflection point, so it is inviscidly stable. But viscous effects can redistribute the perturbation energy in a way that feeds the instability. Look up the critical Reynolds number for plane Poiseuille flow.

Answer

As $R e \to \infty$ , the viscous term (right side) vanishes, leaving the Rayleigh equation: $(U - c) (ϕ^{''} - α^{2} ϕ) - U^{''} ϕ = 0$ . Rayleigh's inflection-point theorem states that a necessary condition for inviscid instability ( $Im (c) > 0$ ) is that $U^{''}$ changes sign somewhere in the flow. For plane Poiseuille flow, $U = U_{m a x} (1 - y^{2} / h^{2})$ , so $U^{''} = - 2 U_{m a x} / h^{2} < 0$ everywhere -- no inflection point. The flow is inviscidly stable.

Yet the full Orr-Sommerfeld equation predicts instability at finite $R e$ . The critical Reynolds number for plane Poiseuille flow is $R e_{c} \approx 5772$ (Orszag 1971). Viscosity is destabilising because it redistributes the perturbation Reynolds stress. In the inviscid limit, the phase relationship between $u^{'}$ and $v^{'}$ perturbations is locked by the Rayleigh equation in a way that produces zero net energy transfer from the mean flow to the perturbation. Viscous diffusion shifts this phase relationship, allowing a net energy transfer that feeds the perturbation. This is a counterintuitive result: adding dissipation can promote instability by changing the phase between velocity components in the perturbation. The mechanism is related to the critical layer, where $U (y) = c_{r}$ (the real part of the wave speed), and the perturbation dynamics are singular in the inviscid limit but regularised by viscosity.

Exercise 8 (hard, symbolic).

A sphere of radius $a$ and density $ρ_{s}$ is released from rest in an infinite fluid of density $ρ_{f}$ and viscosity $η$ . Using Stokes' drag law, derive the equation of motion for the sphere and show that the terminal velocity is $U_{t} = 2 a^{2} (ρ_{s} - ρ_{f}) g / (9 η)$ . Then estimate the time constant $τ$ for the sphere to reach terminal velocity (the time for the velocity to reach $1 - 1/ e$ of its terminal value).

Hint

Newton's second law for the sphere: $(4/3) π a^{3} ρ_{s} d v / d t = (4/3) π a^{3} (ρ_{s} - ρ_{f}) g - 6 π η a v$ (weight minus buoyancy minus drag). This is a first-order linear ODE.

Answer

The equation of motion is $m_{s} d v / d t = W - B - F_{D}$ where $m_{s} = (4/3) π a^{3} ρ_{s}$ is the sphere mass, $W - B = (4/3) π a^{3} (ρ_{s} - ρ_{f}) g$ is the net weight (weight minus buoyancy), and $F_{D} = 6 π η a v$ is the Stokes drag. This gives:

\frac{4}{3} π a^{3} ρ_{s} \frac{d v}{d t} = \frac{4}{3} π a^{3} (ρ_{s} - ρ_{f}) g - 6 π η a v .

Rearranging: $d v / d t + v / τ = g_{eff}$ where $τ = (2 ρ_{s} a^{2}) / (9 η)$ is the time constant and $g_{eff} = (ρ_{s} - ρ_{f}) g / ρ_{s}$ . The solution with $v (0) = 0$ is $v (t) = U_{t} (1 - e^{- t / τ})$ where $U_{t} = g_{eff} τ = 2 a^{2} (ρ_{s} - ρ_{f}) g / (9 η)$ . At $t = τ$ , $v = U_{t} (1 - 1/ e) \approx 0.63 U_{t}$ .

For a 10 $μ$ m dust particle in air ( $ρ_{s} = 2500$ , $η = 1.8 \times 1 0^{- 5}$ ): $τ = 2 \times 2500 \times 1 0^{- 10} / (9 \times 1.8 \times 1 0^{- 5}) = 5 \times 1 0^{- 7} /1.62 \times 1 0^{- 4} = 3.1 \times 1 0^{- 3}$ s. The particle reaches terminal velocity in a few milliseconds -- nearly instantaneously on any practical time scale.

Lean formalization Intermediate+

Mathlib has no formalisation of the Navier-Stokes equations, the Newtonian viscous stress tensor, the Reynolds number, Poiseuille flow, Stokes drag, or boundary-layer theory. The closest infrastructure is the vector calculus library (Mathlib.Analysis.Calculus.Curl, Mathlib.Analysis.Calculus.Divergence) which defines divergence, curl, and gradient on EuclideanSpace, and the PDE framework for Laplace's equation and the heat equation. What is missing: the Newtonian stress tensor as a linear map from the symmetric part of the velocity gradient to the stress; the Navier-Stokes equation as a parabolic PDE with the incompressibility constraint enforced by a pressure Lagrange multiplier; the energy dissipation identity and its proof using integration by parts; existence of weak solutions via the Leray projection and Galerkin approximation (Leray 1934); the uniqueness of weak solutions in two dimensions and the open uniqueness question in three dimensions (the Clay Millennium Problem); exact solutions including Poiseuille, Couette, and Stokes flows; the Blasius similarity solution; the Reynolds number as a consequence of nondimensionalisation; and the Kolmogorov turbulence theory including the energy cascade and the $- 5/3$ law for the energy spectrum. Formalising even the existence of Leray's weak solutions would require substantial development in Sobolev-space theory, the Helmholtz decomposition of $L^{2}$ vector fields into divergence-free and gradient parts, and compactness arguments for sequences of approximate solutions. This unit ships without a Lean module.

Advanced results Master

Navier-Stokes existence, uniqueness, and regularity: the Millennium Prize

The Navier-Stokes existence and smoothness problem is one of the seven Clay Mathematics Institute Millennium Prize Problems ( $1 M p r i z e) . T h e q u es t i o ni s : g i v e n s m oo t h, d i v er g e n ce - f r ee ini t ia l d a t a$ \mathbf{v}_0 \in C^\infty(\mathbb{R}^3) $w i t h$ \nabla \cdot \mathbf{v}_0 = 0 $, an d s m oo t h, co m p a c tl y - s u pp or t e df or c in g$ \mathbf{f}(x,t) $, d oes t h e t h r ee - d im e n s i o na l in co m p r ess ib l e N a v i er - S t o k ese q u a t i o nha v e a u ni q u e, s m oo t h (c l a ss$ C^\infty $) so l u t i o n f or a l l t im e$ t > 0$?

What is known:

Two dimensions. Global existence and uniqueness of smooth solutions in 2D was proved by Ladyzhenskaya (1959) and by Leray and others. The vorticity equation in 2D is $D ω / D t = η \nabla^{2} ω$ , which gives an a priori bound on $∣ ω ∣$ that prevents blow-up.
Three dimensions -- weak solutions. Leray (1934) proved the existence of weak solutions (solutions in the distributional sense, belonging to $L^{\infty} (0, T; L^{2}) \cap L^{2} (0, T; H^{1})$ ) for all time. The proof uses the energy dissipation identity $d E / d t = - \int Φ d V$ to obtain a priori bounds, Galerkin approximation to construct approximate solutions, and compactness to extract a convergent subsequence. These solutions satisfy the energy inequality $E (t) \leq E (0) - \int_{0}^{t} \int Φ d V d s$ (not necessarily equality, because the passage to the limit may lose energy).
Three dimensions -- uniqueness. Weak solutions are known to be unique if they are sufficiently regular (e.g., if $v \in L^{4} (0, T; L^{4})$ , the Serrin condition). The Leray weak solutions are known to be unique in 2D and are not known to be unique in 3D. If uniqueness fails in 3D, the energy inequality becomes strict for some solutions, meaning energy is "lost" in singularities.
Three dimensions -- regularity. If a smooth solution develops a singularity at time $T^{*}$ , then the maximum velocity and the enstrophy (integral of $∣\nabla v ∣^{2}$ ) must blow up at $T^{*}$ . The Beale-Kato-Majda criterion (1984) states that a smooth solution remains smooth as long as the vorticity remains bounded in $L^{1} (0, T; L^{\infty})$ . No blow-up has been found numerically or analytically, but no proof rules it out.

The mathematical heart of the problem is that the a priori estimates available from the energy equation control $v$ in $L^{\infty} (0, T; L^{2})$ and $\nabla v$ in $L^{2} (0, T; L^{2})$ , but this is not enough to prevent blow-up in 3D. The nonlinearity $(v \cdot \nabla) v$ couples all scales of motion, and the a priori bounds do not control the $L^{\infty}$ norm of $v$ . The gap between what is known and what is needed is the gap between $L^{2}$ and $L^{\infty}$ in three dimensions -- the Sobolev embedding $H^{1} ↪ L^{6}$ in 3D is not enough.

The connection to the ideal-fluid unit 09.07.03 is that the Euler equation is the $R e \to \infty$ limit of the Navier-Stokes equation. For the Euler equation, the regularity question is even harder: the energy dissipation identity is unavailable (no viscosity), and the available a priori bounds are weaker. The theory of weak solutions for the Euler equation 09.02.16 pending is intimately connected to the Navier-Stokes regularity problem through the vanishing-viscosity limit.

Kolmogorov's theory of turbulence (1941)

At high Reynolds numbers, the flow becomes turbulent: the velocity field is chaotic, containing a wide range of length and time scales. Kolmogorov's 1941 theory (often called K41) provides a universal description of the statistical properties of turbulence in the inertial range -- the range of scales larger than the dissipation scale but smaller than the scale at which energy is injected.

The key ideas:

The energy cascade. Energy is injected at large scales (by stirring, boundary layers, or instabilities) and transferred to progressively smaller scales by the nonlinear $(v \cdot \nabla) v$ term. This is the Richardson cascade: "big whirls have little whirls that feed on their velocity, and little whirls have lesser whirls, and so on to viscosity." At the smallest scales, viscous dissipation converts kinetic energy to heat.
Kolmogorov's hypotheses. In the inertial range, the statistics of the flow depend only on the energy dissipation rate $ε$ (energy per unit mass per unit time) and the viscosity $η$ . The Kolmogorov length scale at which viscosity becomes important is $ℓ_{η} = (η^{3} / ρ^{3} ε)^{1/4}$ and the Kolmogorov velocity scale is $v_{η} = (η ε / ρ)^{1/4}$ , giving a Kolmogorov Reynolds number $R e_{η} = ρ v_{η} ℓ_{η} / η = 1$ .
The $- 5/3$ law. The energy spectrum $E (k)$ , defined so that $\int_{0}^{\infty} E (k) d k = \frac{1}{2} ⟨ ∣ v ∣^{2} ⟩$ , obeys in the inertial range:

E (k) = C_{K} ε^{2/3} k^{- 5/3},

where $k$ is the wavenumber (inverse length scale) and $C_{K} \approx 1.5$ is the Kolmogorov constant. This is one of the most accurately verified predictions in all of physics: the $- 5/3$ power law has been confirmed in atmospheric turbulence, wind tunnels, ocean currents, and numerical simulations over many decades of wavenumber.

The derivation follows from dimensional analysis: in the inertial range, $E (k)$ can depend only on $ε$ (the only parameter characterising the energy flow) and $k$ . The dimensions $[ε] = m^{2} s^{- 3}$ and $[k] = m^{- 1}$ uniquely determine $E (k) = C ε^{2/3} k^{- 5/3}$ .

Kolmogorov's theory predicts that the second-order structure function (mean square velocity difference between two points separated by distance $r$ ) scales as $⟨ ∣ v (x + r) - v (x) ∣^{2} ⟩ \propto ε^{2/3} r^{2/3}$ in the inertial range. Higher-order structure functions exhibit intermittency corrections (Kolmogorov's refined theory, 1962): the dissipation rate $ε$ is not spatially uniform but fluctuates, concentrating into intense, sparse structures. The intermittency corrections grow with the order of the structure function, deviating from the K41 prediction. This intermittency is one of the central open problems in turbulence research.

DNS and LES: computing turbulent flows

Direct Numerical Simulation (DNS) solves the full Navier-Stokes equations on a grid fine enough to resolve all scales down to the Kolmogorov scale $ℓ_{η}$ . The number of grid points required scales as $N \sim (L / ℓ_{η})^{3} \sim R e^{9/4}$ (in three dimensions). For $R e = 1 0^{6}$ (a typical engineering flow), $N \sim 1 0^{13.5}$ -- far beyond current computational capacity. DNS is currently feasible up to $R e \sim 1 0^{5}$ on the largest supercomputers.

Large-Eddy Simulation (LES) resolves only the large energy-carrying eddies and models the effect of the small scales through a subgrid-scale model. The Navier-Stokes equations are filtered (averaged over a grid cell), and the unclosed subgrid stress is modelled, most commonly by the Smagorinsky model $τ_{ij}^{sgs} = - 2 (C_{s} Δ)^{2} ∣ \overset{ˉ}{S} ∣ \overset{ˉ}{S}_{ij}$ , where $Δ$ is the filter width, $\overset{ˉ}{S}_{ij}$ is the resolved strain rate, and $C_{s}$ is a model constant. LES reduces the computational cost to $N \sim R e^{3/2}$ (still expensive for high- $R e$ flows but feasible for many engineering applications).

Synthesis

The Navier-Stokes equations extend the Euler equation by adding viscous stress, introducing the physical mechanism of momentum diffusion and energy dissipation. The Reynolds number $R e = ρ U L / η$ is the single dimensionless parameter that determines the flow character: at low $R e$ , viscous forces dominate and the flow is smooth and laminar (Stokes flow, Poiseuille flow); at high $R e$ , inertial forces dominate and the flow becomes turbulent, with the viscous effects confined to thin boundary layers (Blasius). The mathematical structure reflects the physics: the viscous term $η \nabla^{2} v$ is a linear, dissipative perturbation of the Euler equation, but its presence is essential for the well-posedness of the boundary-value problem (it raises the order of the PDE from first to second order in space, permitting the no-slip condition). The global regularity of the three-dimensional Navier-Stokes equation remains one of the outstanding open problems in mathematics (the Clay Millennium Prize), while Kolmogorov's 1941 theory provides a remarkably accurate statistical description of turbulent flow in the inertial range. The $- 5/3$ energy spectrum, confirmed across decades of experiment and simulation, is one of the most robust predictions in all of physics.

Connections Master

Ideal fluid mechanics 09.07.03 provides the inviscid limit ( $η \to 0$ , $R e \to \infty$ ) of the Navier-Stokes equations. The Euler equation is recovered when viscous stresses are dropped. The concepts of the material derivative, pressure gradient, and vorticity all carry over directly. D'Alembert's paradox (zero drag in potential flow) is resolved by viscosity: the no-slip condition creates boundary layers that produce drag even at high $R e$ .
Continuum mechanics field theory 09.07.01 provides the Cauchy stress principle from which the Navier-Stokes equation is derived by specifying a Newtonian constitutive relation. The general balance laws for mass and momentum specialise to the continuity equation and the Navier-Stokes equation.
Elastic waves 09.07.02 use the same Cauchy stress framework but with an elastic (spring-like) constitutive relation instead of a viscous (dashpot-like) one. The Navier equation for elasticity and the Navier-Stokes equation for viscous flow share a common origin in the stress tensor but differ in whether the stress responds to strain (elastic) or strain rate (viscous).
Weak solutions 09.02.16 pending provide the mathematical framework for Leray's weak solutions of the Navier-Stokes equation. The energy dissipation identity gives the a priori estimates needed for existence via Galerkin approximation and compactness. The question of uniqueness of these weak solutions in 3D is equivalent to the regularity question and is the subject of the Clay Millennium Prize.
Geophysical fluid dynamics 28.02.01 applies the Navier-Stokes equations on a rotating sphere. The Ekman boundary layer (where viscous and Coriolis forces balance) is the geophysical analogue of the Blasius boundary layer. Turbulence in the atmosphere and ocean follows Kolmogorov's theory in the inertial range, modified by stratification and rotation at large scales.
Renormalisation group methods from statistical physics have been applied to turbulence, treating the energy cascade as a scale-invariant process and using RG techniques to compute corrections to the Kolmogorov $- 5/3$ law from intermittency. The connection between turbulence and critical phenomena is an active area of research.

Historical and philosophical context Master

Claude-Louis Navier derived the equations bearing his name in 1822 (Memoire sur les lois du mouvement des fluides), motivated by the design of bridges and the need to understand the resistance of fluids to moving bodies. Navier's derivation was based on a molecular model of inter-particle forces rather than the continuum stress tensor; his viscous term was correct but his physical reasoning was rooted in the Laplacian programme of reducing all phenomena to central forces between particles.

George Gabriel Stokes gave the modern derivation in 1845 (On the theories of the internal friction of fluids in motion), based on the stress tensor and the constitutive hypothesis that the deviatoric stress is proportional to the rate of strain. Stokes also introduced the Stokes hypothesis ( $ζ = 0$ : the bulk viscosity is zero, so the mean normal stress equals the thermodynamic pressure) and derived the drag law for a sphere in slow flow (Stokes law, 1851). The Navier-Stokes equation is named for both, though neither Navier nor Stokes wrote the equation in its modern vector form.

Osborne Reynolds performed his pipe-flow experiment in 1883 (An experimental investigation of the circumstances which determine whether the motion of water shall be direct or sinuous, and of the law of resistance in parallel channels), identifying the Reynolds number as the criterion for the laminar-turbulent transition. Reynolds' insight was that the transition depends not on the absolute values of the parameters but on their dimensionless combination $ρ U L / μ$ . This was one of the first systematic uses of dimensional analysis in physics and laid the foundation for the theory of dynamical similarity.

Ludwig Prandtl introduced boundary-layer theory in 1904 (Uber Flussigkeitsbewegung bei sehr kleiner Reibung), resolving d'Alembert's paradox by showing that viscosity, however small, always produces a thin layer near walls where the flow adjusts from the free-stream velocity to zero. Prandtl's boundary-layer equations are a singular perturbation of the Navier-Stokes equation: the highest-order term $η \nabla^{2} v$ is small but cannot be dropped near the wall, because it is the only term that can satisfy the no-slip condition. The Blasius solution (1908) for the flat-plate boundary layer was the first exact solution of Prandtl's equations.

Jean Leray proved the existence of weak solutions of the Navier-Stokes equation in his 1934 thesis (Sur le mouvement d'un liquide visqueux emplissant l'espace), using what are now called Galerkin methods and energy estimates. Leray's weak solutions satisfy the energy inequality and exist for all time in three dimensions. The question of their uniqueness and regularity remains open.

Andrey Kolmogorov published his turbulence theory in 1941 (The local structure of turbulence in incompressible viscous fluid for very large Reynolds numbers, Doklady Akademii Nauk SSSR 30, 9-13), deriving the $- 5/3$ energy spectrum and the two-thirds law for structure functions from dimensional analysis and the hypotheses of local isotropy and a constant energy flux through the inertial range. Kolmogorov's theory is remarkable for its simplicity (the predictions follow from dimensional analysis alone) and its accuracy (the $- 5/3$ law is one of the best-verified predictions in all of physics).

A philosophical note: the Navier-Stokes equation is a deterministic PDE, but turbulent flows are effectively unpredictable at the level of individual realisations. This is not quantum indeterminacy but deterministic chaos: extreme sensitivity to initial conditions at high Reynolds number. The statistical description (Kolmogorov's theory) is predictive not for individual flow realisations but for averaged quantities. The gap between the deterministic equation and the statistical prediction mirrors the gap in statistical mechanics between microscopic determinism and macroscopic thermodynamics. In both cases, the coarse-grained description is more useful than the fine-grained one, and the mathematical challenge is to derive the coarse-grained laws from the fine-grained equation.

Bibliography Master

Navier, C.-L., "Memoire sur les lois du mouvement des fluides," Memoires de l'Academie Royale des Sciences 6 (1822), 389-440.
Stokes, G. G., "On the theories of the internal friction of fluids in motion, and of the equilibrium and motion of elastic solids," Trans. Cambridge Phil. Soc. 8 (1845), 287-319.
Reynolds, O., "An experimental investigation of the circumstances which determine whether the motion of water shall be direct or sinuous, and of the law of resistance in parallel channels," Phil. Trans. Roy. Soc. London 174 (1883), 935-982.
Prandtl, L., "Uber Flussigkeitsbewegung bei sehr kleiner Reibung," Verhandlungen des III. Internationalen Mathematiker-Kongresses, Heidelberg (1904), 484-491.
Blasius, H., "Grenzschichten in Flussigkeiten mit kleiner Reibung," Z. Math. Phys. 56 (1908), 1-37.
Leray, J., "Sur le mouvement d'un liquide visqueux emplissant l'espace," Acta Mathematica 63 (1934), 193-248.
Kolmogorov, A. N., "The local structure of turbulence in incompressible viscous fluid for very large Reynolds numbers," Doklady Akademii Nauk SSSR 30 (1941), 9-13.
Kolmogorov, A. N., "Dissipation of energy in the locally isotropic turbulence," Doklady Akademii Nauk SSSR 32 (1941), 16-18.
Ladyzhenskaya, O. A., The Mathematical Theory of Viscous Incompressible Flow, 2nd ed. (Gordon and Breach, 1969).
Serrin, J., "On the interior regularity of weak solutions of the Navier-Stokes equations," Arch. Rational Mech. Anal. 9 (1962), 187-195.
Beale, J. T., Kato, T. & Majda, A., "Remarks on the breakdown of smooth solutions for the 3-D Euler equations," Comm. Math. Phys. 94 (1984), 61-66.
Frisch, U., Turbulence: The Legacy of A. N. Kolmogorov (Cambridge University Press, 1995).
Foias, C., Manley, O., Rosa, R. & Temam, R., Navier-Stokes Equations and Turbulence (Cambridge University Press, 2001).
Schlichting, H. & Gersten, K., Boundary-Layer Theory, 8th ed. (Springer, 2000).
Batchelor, G. K., An Introduction to Fluid Dynamics (Cambridge University Press, 2000).
Landau, L. D. & Lifshitz, E. M., Fluid Mechanics, 2nd ed. (Course of Theoretical Physics Vol. 6, Pergamon, 1987).
Pope, S. B., Turbulent Flows (Cambridge University Press, 2000).

Prerequisites

09.07.03

Tier anchors

beginner: Taylor, Classical Mechanics (2005), Ch. 14 (fluid dynamics)
intermediate: Landau & Lifshitz, Fluid Mechanics, 2nd ed. (1987), §15-23
master: Foias, Manley, Rosa & Temam, Navier-Stokes Equations and Turbulence (2001); Frisch, Turbulence: The Legacy of A. N. Kolmogorov (1995)

References

Landau & Lifshitz — Fluid Mechanics, 2nd ed. (Course of Theoretical Physics Vol. 6, Pergamon, 1987) · §15 The general equations of motion of a viscous fluid; §16 Energy dissipation in an incompressible fluid; §17 Flow in a pipe (Poiseuille); §20 The drag coefficient; §23 Turbulence and the Reynolds number; §31-33 Laminar boundary layers
Taylor — Classical Mechanics (University Science Books, 2005) · Ch. 14 Fluid dynamics. Viscosity. Navier-Stokes equations. Reynolds number.
Batchelor — An Introduction to Fluid Dynamics (Cambridge University Press, 2000) · Ch. 3 The equations of motion for a Newtonian fluid; Ch. 4 Steady unidirectional flows (Poiseuille, Couette); Ch. 5 The boundary layer
Frisch — Turbulence: The Legacy of A. N. Kolmogorov (Cambridge University Press, 1995) · Ch. 2 The Richardson cascade; Ch. 5 Kolmogorov's 1941 theory; Ch. 7 Intermittency
Foias, Manley, Rosa & Temam — Navier-Stokes Equations and Turbulence (Cambridge University Press, 2001) · Ch. 2 Functional setting of the Navier-Stokes equations; Ch. 4 Existence and uniqueness; Ch. 8 Turbulence and attractors
Schlichting & Gersten — Boundary-Layer Theory, 8th ed. (Springer, 2000) · Ch. 1 The Navier-Stokes equations; Ch. 7 The laminar boundary layer (Blasius); Ch. 15 Turbulent boundary layers; Ch. 16 The stability of laminar boundary layers

Estimated time

beginner: 15m
intermediate: 55m
master: 80m