Reynolds number and flow similarity

Choose a representative speed U, length L, and constant kinematic viscosity nu. Convective acceleration scales as U²/L, while viscous acceleration scales as nu U/L² when velocity changes over length L. Their ratio is UL/nu. Equivalently, viscous diffusion across L takes time L²/nu, while advection across L takes L/U. Reynolds number is the diffusion time divided by the advection time. Large values indicate that diffusion across the full chosen length is relatively slow, not that viscosity has disappeared everywhere.

Scale position by L, velocity by U, time by L/U, and pressure differences by rho U². The incompressible momentum equation then displays a coefficient 1/Re in front of its viscous term. This makes precise which parameter a dynamically similar experiment must match. The pressure scale is a convenient inertial scale, not a claim that pressure is constant. At very small Reynolds number a viscous pressure scale can be more informative, provided all terms are consistently rescaled.

Dynamic similarity requires geometrically similar boundaries and corresponding dimensionless forcing, initial conditions, and boundary conditions. For a rigid submerged body in approximately incompressible single-phase flow, matching Reynolds number is often the main target once geometry and relative roughness are fixed. A free surface introduces gravitational effects and a Froude number; surface tension introduces a Weber number; compressibility introduces a Mach number. Matching one group while changing another can produce a visually convincing but physically misleading experiment.

When the similarity conditions hold, dimensionless observables such as drag coefficient can be compared. State the convention: here drag coefficient is D divided by one-half rho U² A, with a specified reference area A. For similar geometry that area scales with L². Equal drag coefficients therefore imply a force ratio determined by density, speed squared, and length squared. Equal Reynolds numbers do not imply equal dimensional forces, nor do they guarantee that an unsteady flow will be reproduced point by point rather than statistically.

For slow steady motion with negligible inertial terms, the Stokes approximation balances pressure, viscosity, and any retained body forces. A small Reynolds number based on U and L supports neglecting convective inertia, but rapid imposed oscillation can still make the unsteady term important. At large Reynolds number, thin wall regions and wakes can retain strong viscous gradients. The inviscid limit near no-slip boundaries is therefore not obtained by simply deleting viscosity from every part of the domain.

Transition to turbulence is not governed by one universal critical Reynolds number. The relevant definition of U and L, flow geometry, surface roughness, and disturbance environment matter. Pipe-flow thresholds cannot be copied into a boundary layer or rotating apparatus without justification. Treat Reynolds number as an organizing parameter and report the length and velocity used to calculate it. A similarity claim becomes testable when it names the observables being compared and the dimensionless conditions that were actually controlled.