Turbulence is a complex state of fluid motion characterised by irregular, three dimensional and strongly interacting fluctuations in velocity and other flow quantities across a broad range of spatial and temporal scales. It commonly develops when inertial effects are sufficiently strong relative to viscous effects, although transition depends on geometry, disturbances, boundary conditions and flow history. Turbulent flow is not simply synonymous with randomness because it possesses organised statistical structure, coherent motions and energy transfer across scales. Turbulence occurs in atmospheric flows, rivers, oceans, industrial pipes, combustion systems, aircraft wakes and many other fluid environments.
The Reynolds number is a key dimensionless parameter associated with turbulence because it compares inertial and viscous effects. At low Reynolds numbers, viscosity can damp disturbances effectively and flow may remain laminar. At sufficiently high Reynolds numbers, disturbances can amplify and produce increasingly complex motion. There is no single universal Reynolds number at which every flow becomes turbulent because the transition depends on configuration, surface roughness, inlet disturbances and geometry. Pipe flow, flow over a flat plate and free shear layers have different transition characteristics. Consequently, Reynolds number should be interpreted in relation to the specific physical configuration rather than as an absolute turbulence switch.
Turbulence contains a hierarchy of interacting scales. Large eddies extract kinetic energy from the mean flow, while nonlinear interactions transfer energy towards progressively smaller structures in what is often called the energy cascade. At sufficiently small scales, viscous effects become dominant and convert kinetic energy into thermal energy. The smallest dynamically important scales are associated with the Kolmogorov length, time and velocity scales under classical assumptions of homogeneous isotropic turbulence. The range between large energy containing scales and small dissipative scales is called the inertial range under appropriate conditions. The number and separation of scales increase strongly with Reynolds number, making direct numerical resolution computationally expensive.
Turbulent flow is often described statistically because instantaneous velocity fields contain enormous detail. A velocity component can be decomposed into a mean component and a fluctuating component, allowing averaged equations to be derived. Reynolds averaging produces additional terms involving correlations of velocity fluctuations, commonly called Reynolds stresses. These terms represent the effects of turbulence on mean momentum transport but introduce unknown quantities, producing the closure problem. Turbulence models approximate these unknown terms using additional assumptions. The quality of a model depends on the flow regime and the quantity being predicted, so no single turbulence model is universally accurate.
Computational treatment of Turbulence can use direct numerical simulation, large eddy simulation or Reynolds averaged methods. Direct numerical simulation attempts to resolve all dynamically relevant scales but can require enormous computational resources at high Reynolds numbers. Large eddy simulation resolves larger turbulent structures while modelling the effects of smaller scales. Reynolds averaged approaches solve equations for mean quantities and model turbulent fluctuations, making them much less computationally demanding for many engineering applications. Hybrid approaches combine features of different methods. The choice involves trade-offs among spatial and temporal resolution, computational cost, geometry, Reynolds number and required prediction accuracy.
Turbulence strongly influences transport processes. It enhances mixing of momentum, heat and chemical species by moving fluid parcels across regions that would exchange material much more slowly through molecular diffusion alone. This can increase effective transport coefficients in engineering models. Turbulent boundary layers usually produce greater wall shear stress and drag than corresponding laminar flows. In combustion systems, turbulence can strongly affect the mixing of fuel and oxidiser and therefore influence reaction rates and flame structure. Atmospheric turbulence contributes to the dispersion of pollutants and heat, while ocean turbulence affects mixing and transport of nutrients and dissolved substances.
Turbulence is difficult to predict because the governing Navier Stokes equations are nonlinear and couple structures across a wide range of scales. Small changes in initial or boundary conditions can produce substantially different instantaneous flow states while statistical properties remain comparatively predictable. This distinction between instantaneous predictability and statistical predictability is important. Turbulent flow does not mean that the governing equations cease to apply; rather, the equations generate highly complex solutions whose full resolution can be computationally prohibitive. Experimental measurements therefore commonly use statistical quantities such as mean velocity, root mean square fluctuations, spectra and correlation functions rather than attempting to record every microscopic feature.
Turbulence has major practical significance in aerospace, environmental science, energy systems and industrial fluid transport. It can increase drag and therefore energy consumption, but it can also enhance mixing and heat transfer in applications where those effects are desirable. Engineers may deliberately promote turbulence in heat exchangers to improve transfer while attempting to limit unnecessary pressure loss. Atmospheric and oceanic turbulence affects weather prediction, pollutant dispersion and climate modelling. Understanding turbulence requires combining fluid dynamics, statistical analysis, dimensional reasoning, experimentation and computation because no simple single variable completely characterises its behaviour.
