Reynolds Number

A ratio of two forces

Every flow contains two competing effects. Inertia carries fluid forward and resists changes in its motion; viscosity drags neighbouring layers along with each other and damps disturbances. The ratio of these two effects is the Reynolds number:

Re = ρ v L / μ = v L / ν

ρ  fluid density        (kg/m³)
v  characteristic speed (m/s)
L  characteristic length (m)
μ  dynamic viscosity    (Pa·s)
ν  kinematic viscosity  (m²/s)

The formula appears in the textbook with deceptive simplicity, but reading it through dimensional analysis is illuminating. The quantity ρv² is a dynamic pressure — how hard a piece of fluid hits something at speed v. The quantity μv/L is a viscous stress — how strongly the fluid resists shearing across a length L. Their ratio ρvL/μ is therefore "how much momentum the flow has versus how strongly viscosity can dissipate it." When inertia wins by a wide margin, small perturbations get amplified rather than damped, and the flow becomes turbulent.

The number is named for Osborne Reynolds, the Manchester engineer who in 1883 ran a now-classic experiment: water flowing through a glass tube with a thin filament of dye injected at the inlet. At low flow rates the dye traced a perfectly straight line down the centre of the pipe. At a critical speed the filament suddenly broke up into eddies that filled the pipe. Reynolds plotted his results in terms of the ratio that now bears his name and identified the dimensionless threshold near 2000 for circular pipes — a number that appears in every fluids textbook a century and a half later.

Laminar, transitional, turbulent

The split is not always sharp, but the canonical pipe-flow boundaries are clean enough to memorise:

Regime	Re range (pipe)	Behaviour	Example
Stokes / creeping	Re < 1	Viscous forces dominate; reversible streamlines	Bacterium swimming, Re ≈ 10⁻⁴
Laminar	1 < Re < 2300	Smooth parallel layers, parabolic velocity profile	Capillary in tissue, Re ≈ 0.001–1
Transitional	2300 < Re < 4000	Intermittent puffs and slugs of turbulence	Slow water in 25 mm pipe at 0.1 m/s
Turbulent (smooth)	4000 < Re < 10⁵	Fully chaotic with self-similar eddies	Household tap, Re ≈ 5000
Turbulent (rough)	Re > 10⁵	Friction factor independent of Re	Water main, Re ≈ 10⁶
High-Re geophysical	Re > 10⁸	Multi-scale cascade, no laminar base state	Atmosphere, Re ≈ 10⁹

The transition values are not universal constants. They depend on inlet conditions, surface roughness and any vibrations or acoustic noise that might trigger an instability. With a long settling chamber, polished walls and very quiet supply, laminar pipe flow has been sustained to Re ≈ 10⁵ in the lab. The textbook 2300 is a robust working value — the Re above which any realistic disturbance is likely to grow.

Worked example: Reynolds number in your aorta

Blood flowing through the human aorta is the canonical biomedical example. Take typical values for an adult at rest:

Density of blood     ρ = 1060 kg/m³
Aortic diameter      L = 0.025 m  (25 mm)
Peak systolic speed  v = 0.6 m/s
Dynamic viscosity    μ = 0.0035 Pa·s  (≈3.5 cP)

Re = ρ v L / μ
   = (1060)(0.6)(0.025) / 0.0035
   = 15.9 / 0.0035
   = 4540

This is in the transitional regime. In healthy aorta the flow is mostly laminar during diastole and develops only short turbulent bursts during the peak of systole. In aortic stenosis (a narrowed valve) the local velocity can triple and the local diameter halves, pushing Re above 10⁴ — fully turbulent flow. That turbulence is what produces the audible heart murmur a stethoscope picks up.

Now repeat the calculation for a capillary, where blood reaches its single-cell-file end stations:

Capillary diameter   L = 8×10⁻⁶ m
Capillary speed      v = 5×10⁻⁴ m/s
Same fluid           ρ = 1060, μ = 0.0035

Re = (1060)(5×10⁻⁴)(8×10⁻⁶) / 0.0035
   = 4.24×10⁻⁶ / 0.0035
   = 1.2×10⁻³

That is six orders of magnitude lower than the aorta. Capillaries live in the deep Stokes regime where viscosity dominates utterly — flow is slow, reversible, and nearly free of inertia. The same fluid that goes turbulent at the heart behaves like syrup at the capillary wall. This is why the geometry of the circulation is so different at different scales: the heart needs valves to prevent inertia-driven backflow, but capillaries do not.

Reynolds similarity and wind tunnels

The most powerful consequence of the Reynolds number is dynamical similarity. If you take two flows over geometrically similar bodies and arrange them to have the same Re, their dimensionless velocity, pressure and force coefficients will match. A 1/20-scale model in a wind tunnel reproduces the full-scale flow exactly — provided the test conditions hit the same Re.

This is rarely free. Suppose you want to test a 60-metre commercial airliner cruising at 240 m/s in air with ν = 1.5×10⁻⁵ m²/s. The cruise Re is:

Re_cruise = (240 × 60) / 1.5×10⁻⁵ = 9.6×10⁸

For a 1/20-scale model with chord 3 m, matching that Re at sea-level air would require v = 4800 m/s — impossible without supersonic effects that change the physics. Engineers cheat in three ways: (1) increase ρ by pressurising the tunnel, (2) decrease μ and increase ρ together by cooling the gas to cryogenic temperatures, or (3) live with a Re mismatch and apply known corrections. NASA Langley's National Transonic Facility uses route (2): nitrogen at 110 K hits Re of 10⁹ in a 2.5 m test section.

Where Re lives in the Navier–Stokes equation

Non-dimensionalising the incompressible Navier–Stokes equation with characteristic scales L and v drops everything into a single coefficient — Re. The dimensionless equation reads:

∂u*/∂t* + (u*·∇*)u* = −∇*p* + (1/Re) ∇*²u*

The viscous term is divided by Re. When Re is small, the viscous diffusion term dominates and the equation reduces to the linear Stokes equation, which can be solved analytically for many geometries. When Re is large, the viscous term is a small singular perturbation that survives only inside thin boundary layers near walls — the entire framework of boundary-layer theory exists because of this scaling. The non-trivial mathematical content is that the limit Re → ∞ is singular, not regular: ignoring viscosity entirely (the Euler equation) gives a different solution from the Re → ∞ limit of Navier–Stokes.

What actually goes turbulent at the threshold

The bare statement "above 2300 the flow is turbulent" hides a lot of physics. In a pipe, the laminar Hagen–Poiseuille profile is linearly stable to infinitesimal disturbances at all Reynolds numbers — every textbook eigenvalue calculation confirms this. So how does turbulence get started?

The answer is that finite-amplitude disturbances can grow even when infinitesimal ones cannot. Above Re ≈ 1800, disturbances of a few percent in inlet velocity excite three-dimensional, non-linear structures called turbulent puffs that travel down the pipe at the mean speed. Below a second threshold near Re ≈ 2040 these puffs decay individually but can also split — and below another threshold near Re ≈ 2300 they split faster than they decay, producing fully turbulent flow downstream. Björn Hof's group at IST Austria measured these crossover Re values in glass pipes 30 m long.

For boundary layers and shear flows, the path to turbulence runs through a different mechanism — Tollmien–Schlichting waves, secondary instabilities, lambda vortices, and finally turbulent spots. The orderly progression is well-documented but extremely sensitive to free-stream turbulence and wall roughness. A modern aircraft wing might be designed for "natural laminar flow" up to Re_x ≈ 5×10⁶, but a single insect strike on the leading edge will trip the boundary layer to turbulence and kill the laminar advantage.

Where Reynolds number shows up

• Aircraft wing drag. A Boeing 737 wing has chord-Re ≈ 4×10⁷ at cruise. The boundary layer is turbulent over >90 % of the wing, contributing roughly half the total drag. A 1 % friction reduction by laminar-flow-control technology saves billions of dollars in fuel per year for the global fleet.
• Pipeline flow. A trans-Alaska oil pipeline carrying crude at 1.5 m/s through a 1.2 m diameter pipe has Re ≈ 4×10⁵. Designers pick pumps and pipe roughness to land in the fully turbulent rough regime where friction factor is independent of Re — the most predictable operating point.
• Cardiovascular pathology. Aortic stenosis raises local Re past 10⁴, producing audible turbulence and shear stresses high enough to lyse red blood cells. A Doppler-ultrasound peak velocity over 4 m/s is a clinical indication for valve replacement.
• Sports ball trajectories. A baseball at 40 m/s with diameter 7.4 cm has Re ≈ 2×10⁵, just below the drag crisis where the boundary layer transitions from laminar to turbulent on the ball. Seam orientation tips the local Re past this threshold on one side, generating asymmetric drag and the famous knuckleball or sinker behaviour.
• Atmospheric boundary layer. The lowest 1 km of the atmosphere has Re ≈ 10⁹. Weather and climate models cannot resolve the smallest turbulent eddies and instead parameterise their effect via subgrid-scale closure. The overall energy balance of Earth depends on getting the turbulent transport right.

The drag crisis on a sphere

One of the most striking Re-driven transitions is the drag crisis on a sphere. As Re increases from 10² to 10⁵ the drag coefficient C_d slowly drifts down from about 1.0 to about 0.5. Then, between Re ≈ 2×10⁵ and Re ≈ 4×10⁵, C_d collapses to about 0.1 — a five-fold drop. The reason is that the laminar boundary layer on the windward face of the sphere transitions to turbulent before the flow separates, and the more energetic turbulent boundary layer stays attached further around the back of the sphere. The wake is narrower, the pressure recovery is better, and form drag plummets.

This is exactly what golf-ball dimples exploit. A smooth golf ball at typical drive Re ≈ 1.5×10⁵ would be in the high-drag regime. The dimples trip the boundary layer to turbulent at lower Re, putting the ball in the post-crisis low-drag regime and adding 30–40 % to the carry distance. The same trick — turbulators, vortex generators, leading-edge roughness — appears in cricket-ball seams, sailplane wings and submarine hulls.

Variants and extensions

• Local Reynolds number Re_x. For boundary-layer flow over a flat plate, the relevant length is the streamwise distance x from the leading edge: Re_x = ρvx/μ. Transition typically occurs near Re_x ≈ 5×10⁵, which is why airfoil pressure distributions are often plotted as a function of x/c rather than fixed Re.
• Reynolds number for non-circular ducts. The relevant length is the hydraulic diameter D_h = 4A/P, where A is cross-sectional area and P is wetted perimeter. The transition value remains Re ≈ 2300 for most cross-section shapes.
• Particle Reynolds number Re_p. For sedimentation and aerosol dynamics, the length scale is the particle diameter and the velocity is the relative velocity between particle and fluid. A 10 μm raindrop falling at terminal velocity has Re_p ≈ 0.01 (Stokes regime), while a 5 mm raindrop has Re_p ≈ 1500 (well into wake-shedding territory).
• Magnetic Reynolds number Rm. In magnetohydrodynamics the analogue is Rm = vL/η, where η is magnetic diffusivity. Rm controls whether magnetic field lines move with the fluid (high Rm — solar plasma) or diffuse independently (low Rm — liquid metals in lab experiments).
• Reynolds-averaged Navier–Stokes (RANS). For engineering CFD, the Reynolds decomposition splits velocity into mean and fluctuating parts and time-averages the equations, leaving an unknown Reynolds stress tensor that must be modelled. k-ε, k-ω and Spalart–Allmaras are the work-horse closures used in aerospace and automotive design.

Common pitfalls

• Picking the wrong length scale. Reynolds number is meaningless without specifying L. Pipe flow uses diameter; airfoils use chord; spheres use diameter; flat plates use streamwise distance. Comparing Re values that use different length conventions is a common student error and a not-uncommon paper-review error.
• Treating 2300 as a universal constant. The pipe-flow transition value depends on inlet quality and disturbance amplitude. Boundary-layer transition can occur anywhere from Re_x ≈ 6×10⁴ (with surface roughness) to Re_x ≈ 5×10⁶ (with very low free-stream turbulence). Always quote the geometry alongside the number.
• Confusing dynamic and kinematic viscosity. ν = μ/ρ has units of m²/s. μ has units of Pa·s. The two formulations of Re look different but give the same number when you keep the bookkeeping consistent. Mixing them will produce a Re that is wrong by a factor of ρ ≈ 10³ for water.
• Assuming compressibility doesn't matter. For Mach numbers above 0.3, density is no longer constant along streamlines and the simple Re analysis must be augmented by Mach-similarity arguments. A sphere in transonic flow does not show a clean drag crisis.
• Forgetting that turbulence has memory. A pipe whose entrance is downstream of a 90° bend may stay turbulent at Re = 1500 even though the textbook says it should re-laminarise. Geometry upstream of the test section can dominate the local Re criterion for tens of pipe diameters downstream.