What the Reynolds number actually compares
Two things fight for control of any flow. Inertia carries fluid onward and lets small disturbances grow into eddies; viscosity resists shear and damps those disturbances out. The Reynolds number is the ratio of the two: inertial force per unit volume scales as ρv²/L, viscous force as μv/L², and dividing one by the other leaves ρvL/μ. Everything cancels, which is why Re is a pure number.
A small Reynolds number means viscosity wins. Fluid moves in orderly layers that slide over one another without mixing; a dye filament injected upstream stays as a thin line all the way down the pipe. This is laminar flow, and Osborne Reynolds demonstrated exactly that experiment in Manchester in 1883.
A large Reynolds number means inertia wins. Any disturbance amplifies, the flow fills with eddies of many sizes, and the dye filament breaks up and stains the whole pipe within a few diameters. This is turbulent flow. The switch is not gradual: below about Re = 2,300 in a pipe you get laminar flow, above about 4,000 you get turbulence, and the band between is intermittent, with turbulent slugs travelling through otherwise laminar fluid.
Because Re is dimensionless, it also licenses model testing. Two flows with the same Reynolds number and the same geometry behave identically no matter their absolute size — a principle called dynamic similarity. A 1:20 scale model tested at twenty times the velocity reproduces the full-scale flow field, which is what makes wind tunnels and towing tanks useful at all.
Choosing the characteristic length, which is where most errors live
The Reynolds number is not defined until you say which length you used, and different conventions give answers that differ by large factors. Three cases cover almost everything.
Circular pipe. Use the internal diameter, not the radius. Using the radius halves Re and can move a genuinely turbulent flow into the laminar band on paper. And use the actual bore rather than the nominal size: a 50 mm nominal steel pipe has an internal diameter near 52.5 mm, and the difference is 5% on Re.
Non-circular duct. Use the hydraulic diameter D_h = 4A/P, four times the cross-sectional area divided by the wetted perimeter. For a circle that reduces to D exactly, which is why the factor of four is there. For a rectangle it becomes 2wh/(w + h), so a 100 × 50 mm duct has Dh = 66.7 mm — not 100, not 50, and not the 75 mm average. For an annulus it is the outer diameter minus the inner one. Hydraulic diameter reproduces pipe correlations well for aspect ratios up to about 4:1 and less well for very flat passages.
External flow over a surface. Use the distance from the leading edge. This means Re grows along the plate, and so the boundary layer starts laminar at the nose and becomes turbulent somewhere downstream. For a smooth plate in low-turbulence flow the transition is conventionally taken at Rex ≈ 5 × 105, but that figure is much softer than the pipe values: careful experiments delay it well past 106, and roughness, vibration or free-stream turbulence bring it forward. For a cylinder or sphere, the diameter is used instead, and the important transitions sit near Re = 1 (creeping flow ends) and Re ≈ 2 × 105 (the drag crisis, which is why golf balls are dimpled).
Worked example: sizing a chilled-water branch
A chilled-water branch carries 4.0 L/s of water at 20 °C through a 50 mm internal-bore pipe. Water at 20 °C has ρ = 998.2 kg/m³ and μ = 1.002 × 10−3 Pa·s.
- Kinematic viscosity. ν = μ ÷ ρ = 1.002×10−3 ÷ 998.2 = 1.00381 × 10−6 m²/s.
- Cross-sectional area. A = πD²/4 = π × 0.05² ÷ 4 = 1.96350 × 10−3 m².
- Mean velocity. v = Q ÷ A = 0.004 ÷ 1.96350×10−3 = 2.0372 m/s.
- Reynolds number. Re = vD/ν = 2.0372 × 0.05 ÷ 1.00381×10−6 = 0.101860 ÷ 1.00381×10−6 = 101,473.
- Classify. Well above 4,000, so the flow is fully turbulent. Friction loss will follow the Colebrook or Moody route, not the laminar
64/Relaw. - Find the laminar limit. vcrit = 2300 × 1.00381×10−6 ÷ 0.05 = 0.0462 m/s, which is 0.091 L/s. To keep this pipe laminar you would have to throttle the flow to 2.3% of its design rate — which is the general lesson: water in pipes of engineering size is almost always turbulent.
Now change the fluid to an ISO VG 46 lubricant at 40 °C, ν = 46 cSt = 4.6 × 10−5 m²/s. At the same 2.0372 m/s: Re = 0.101860 ÷ 4.6×10−5 = 2,214 — just below the laminar limit, in the same pipe at the same speed. That factor of 46 in viscosity moved the flow across the entire transition, which is why oil and water pipework are designed with different rules of thumb.
What the number means for pressure drop, heat transfer and mixing
Pressure drop. In laminar pipe flow the Darcy friction factor is exactly f = 64/Re, so pressure loss is proportional to velocity and independent of wall roughness — a smooth pipe and a rough one of the same bore lose the same head. In turbulent flow the friction factor comes from the Colebrook equation or a Moody chart, depends strongly on relative roughness, and pressure loss scales roughly with velocity squared. Doubling the flow rate in a turbulent line roughly quadruples the pumping head and multiplies the power by about eight.
Velocity profile. Laminar pipe flow is exactly parabolic, and the centreline velocity is twice the mean. Turbulent flow is far blunter, with the centreline typically about 1.2 times the mean. If you are measuring flow with a single-point pitot on the centreline, you need to know which regime you are in before converting to a flow rate.
Heat and mass transfer. Turbulence mixes, so heat exchangers are almost always designed to run turbulent even at the cost of extra pumping power: the Nusselt number climbs steeply with Re once past transition. A laminar exchanger relies on conduction across the fluid and performs poorly by comparison.
Entrance length. The profile takes distance to develop. Laminar flow needs about 0.05·Re·D of straight pipe, which at Re = 2,000 is 100 diameters. Turbulent flow needs only 10–60 diameters. Flow meters specify an upstream straight-run requirement for exactly this reason.
Finally, treat the transition values as conventions rather than physical constants. In extremely smooth, vibration-free pipes with a carefully shaped inlet, laminar flow has been maintained past Re = 10,000; a disturbed inlet or a rough wall trips it much sooner. Design so you are not sitting inside 2,300–4,000, where neither correlation is reliable and the flow can flip between states.
Reynolds numbers for everyday flows
| Situation | ν (m²/s) | v (m/s) | L (m) | Re | Regime |
|---|---|---|---|---|---|
| Glycerine in a 50 mm pipe | 1.121 × 10⁻³ | 1.0 | 0.050 | 45 | Laminar |
| Water at the laminar limit, 50 mm pipe | 1.004 × 10⁻⁶ | 0.0462 | 0.050 | 2,300 | Transition onset |
| Water in a 15 mm domestic pipe | 1.004 × 10⁻⁶ | 1.0 | 0.015 | 14,940 | Turbulent |
| Air in a 300 mm HVAC duct | 1.516 × 10⁻⁵ | 5.0 | 0.300 | 98,945 | Turbulent |
| Water in a 50 mm pipe | 1.004 × 10⁻⁶ | 2.0 | 0.050 | 99,602 | Turbulent |
| Water in a 100 mm main | 1.004 × 10⁻⁶ | 1.5 | 0.100 | 149,402 | Turbulent |
| Air over a 0.5 m plate at 10 m/s | 1.516 × 10⁻⁵ | 10.0 | 0.500 | 329,815 | Laminar boundary layer |
Water and glycerine values are at 20 °C; air is at 20 °C and one atmosphere. The plate row uses the flat-plate transition value of 5 × 10⁵ rather than the pipe values.
Mistakes that give a wrong Reynolds number
- Using radius instead of diameter. This halves Re and can turn a turbulent design into an apparently laminar one on paper.
- Mixing dynamic and kinematic viscosity.
Re = ρvL/μorvL/ν, neverρvL/ν. Getting this wrong is off by a factor of the density — three orders of magnitude for water. - Leaving viscosity in cP or cSt. Convert to Pa·s and m²/s. 1 cP = 10−3 Pa·s and 1 cSt = 10−6 m²/s.
- Ignoring temperature. Water's viscosity nearly halves between 20 °C and 50 °C, so the same flow can move a long way across the transition band on a hot day. Oil viscosity is even more temperature-sensitive.
- Using the arithmetic mean of duct sides. The hydraulic diameter of a 100 × 50 mm duct is 66.7 mm, not 75 mm.
- Applying the pipe transition value to external flow. A flat-plate boundary layer transitions near 5 × 105, more than two hundred times the pipe figure.
- Using the nominal pipe size. Nominal diameters are labels, not bores. Look up the actual internal diameter for the schedule you are using.
- Quoting an Re without saying which length it used. A cylinder Re on diameter and a plate Re on chord length are not comparable numbers.
Where the Reynolds number sits among the other dimensionless groups
Re is the first of a family. The Mach number compares speed to the speed of sound and governs compressibility; below about Mach 0.3 you can treat a gas as incompressible and ignore it. The Froude number compares inertia to gravity and governs free-surface flows, ship waves and open channels. The Prandtl number compares momentum diffusion to thermal diffusion and, combined with Re, gives the Nusselt number for convective heat transfer. The Weber number compares inertia to surface tension and governs droplet break-up.
In pipework, Re is the input to the friction factor and therefore to the pressure drop, which is the head a pump must supply on top of any static head from elevation. For gases the density needed for Re comes from the ideal gas law, so a duct calculation at altitude or at temperature has to start there. And when the flow is carrying heat rather than just fluid, the energy balance is the Q = mcΔT calculation, with Re deciding how effectively that heat crosses to the wall.
Turbulence also generates broadband noise, which is why the same Reynolds number that tells you a duct will be lossy also tells you it will be audible — the level of that noise is the decibel calculation. At the other end of the scale, flows at Re far below 1 are dominated entirely by viscosity: bacteria swimming, lubrication films and sedimenting particles all live there, where reversing the driving force exactly retraces the motion and inertia plays no part at all.
Key terms
- Dynamic viscosity (µ)
- Resistance to shear, in pascal-seconds. Water at 20 °C is 1.002 mPa·s; SAE 30 oil is a few hundred times higher.
- Kinematic viscosity (ν)
- Dynamic viscosity divided by density, in m²/s. Lubricant grades are specified this way, in centistokes at 40 °C.
- Hydraulic diameter
- Four times the flow area divided by the wetted perimeter — the equivalent circular bore that reproduces pipe correlations for a non-circular passage.
- Critical Reynolds number
- The value at which the flow changes regime: about 2,300 for the onset of transition in a pipe and about 5 × 10⁵ for a smooth flat-plate boundary layer.
- Dynamic similarity
- Two geometrically similar flows with the same Reynolds number have the same flow pattern, which is what makes scale-model testing valid.
