Introduction

Processes of turbulence generation

This chapter is about turbulence within the stratified body of the ocean beyond the direct effects, described in Chapter 3, of its boundaries. The ultimate sources of energy leading to mixing in the ocean are external. The processes causing mixing in the stratified regions of the ocean derive their energy internally, as illustrated in Fig. 3.2, from sources (e.g., radiating internal waves) that may themselves be directly or indirectly driven by external forcing at the boundaries.

Two very different processes usually dominate in the generation of turbulence and diapycnal mixing in the stably stratified ocean. The first is instability resulting from the shear or differential motion of water, i.e., the vertical gradient of the horizontal current, dU/dz, which is often caused by internal waves. This is described in Section 4.2 and some aspects and evidence of the related turbulent motion are presented in Sections 4.3–4.7. The second process is more subtle, a form of convection that results from the different molecular diffusion coefficients of heat and salinity. How these lead to instability is explained in Section 4.8.

The first observations of turbulence in the thermocline

The first published measurements of turbulence within the stratified waters of the thermocline were reported in 1968 by Grant, Moilliet and Vogel. They were made off the west coast of Vancouver Island using hot-film anemometers mounted on a submarine. Grant and his colleagues compared their measurements of turbulence with those made in the mixed layer near the sea surface.

These are possible answers to the problems set at the end of each chapter in An Introduction to Ocean Turbulence. Alternative or simpler answers may be possible in some cases. Please advise the author if any errors are found.

Chapter 1

P1.1. Integrating over the area of the tube, the net flow is ∫0au ⋅ 2π r. dr = 4π U ∫ (r − r3/a2)dr = 4π U[r2/2 − r4/(4a2)]0a=π Ua2. By definition, this must be equal to the mean flow times the cross-sectional area, π a2, so U is equal to the mean flow. By conservation of the volume flux, the mean flow downstream of the transition from laminar to turbulent flow must also be equal to U.

The flux of kinetic energy upstream of the transition from laminar to turbulent flow is u(ρ u2/2) integrated over the cross-section of the tube, i.e., ∫0a (ρ u3/2)⋅ 2π r dr = 8π ρ U3∫0ar(1 − r2/a2)3 dr = − π ρ U3a2[(1 − r2/a2)4]0a = π ρ U3a2.

In the same way, the flux of the kinetic energy of the mean flow is (ρ U3/2)π a2 within the turbulent flow downstream of the transition, so the reduction in the flux is the difference between the flux upstream of the transition and that downstream, or (ρ U3/2) π a2. This represents a flux of kinetic energy to the turbulent motion, ignoring any work done by pressure forces.

Introduction

Turbulence is the dominant physical process in the transfer of momentum and heat, and in dispersing solutes and small organic or inorganic particles, in the lakes, reservoirs, seas, oceans and fluid mantles of this and other planets. Oceanic turbulence has properties that are shared by turbulence in other naturally occurring fluids and in flows generated in civil, hydraulic and chemical engineering installations and in buildings. The study of turbulence consequently has applications well beyond the particular examples in the ocean that are selected for description below.

Figure 1.1 shows the sea surface in a wind of about 26 m s− 1. It is covered by waves, many of them breaking and injecting their momentum and bubbles of air from the overlying atmosphere into the underlying seawater. Immediately below the surface, and even at great depths, the water is generally in the state of irregular and variable motion that is referred to as ‘turbulence’, although there is no simple and unambiguous definition of the term. Turbulence has, however, characteristics that, as will be explained, can be quantified and which make it of vital importance. Many of the figures in this book illustrate the nature of turbulent motion, the processes that drive turbulence, or the measurements that can be made to determine its effects.

• Turbulence is generally accepted to be an energetic, rotational and eddying state of motion that results in the dispersion of material and the transfer of momentum, heat and solutes at rates far higher than those of molecular processes alone.

Boundary layers are an elemental concept of high Reynolds number flow. They are a framework for discussing viscous fluid dynamics by separating the flow into distinct regions. That is the essential nature of boundary layer theory. The seminal ideas were described by Prandtl in 1904. He recognized that viscous flow along surfaces could be divided into two regions: a vortical layer next to the wall and a potential flow farther from the surface. Modern theories have expanded that to multilayered structure; but the basic notion always is of a thin, vortical layer next to the surface and an inviscid outer flow.

The boundary layer concept brought clarity to the puzzle of the high Reynolds number limit. High Reynolds number can be interpreted as low viscosity. Is inviscid flow the correct limit? Without viscosity, fluid flows freely over a surface, slipping relative to the wall. Hence, the tangential velocity is discontinuous between the stationary wall and the flowing fluid. The shear is infinite. Adding the smallest amount of viscosity would cause an infinite stress. Inviscid flow cannot be the correct high Re limit of viscous flow.

Any amount of viscosity will diffuse the velocity discontinuity: the fluid velocity will be brought smoothly to zero at the stationary wall. Even at the highest Reynolds numbers, viscous stresses cannot be ignored. How, then, is high Reynolds number flow to be constructed?

This puzzle is solved by recognizing that viscous influence is confined to a very thin layer next to the wall. The layer of viscous influence becomes increasingly thin as Re becomes increasingly large.

Separation, Vorticity, and Vortex Shedding

Most classical theory of viscous flow is based on approximations valid at low or high Reynolds number. Although there are a few exact solutions that illustrate aspects of the intermediate range, they are rather limited. The range of phenomena that fall under this heading is commonly illustrated by photographs taken in laboratories. Computational fluid dynamics makes the intermediate Reynolds number range quite accessible.

Inertia is now comparable to viscous stress, and all terms in the Navier–Stokes equations must be retained. Convection destroys the upstream–downstream symmetry of creeping flow (Chapter 3). A distinct wake can be identified leaving the downstream side of a body in an incident flow. Forces on blunt bodies become increasingly due to pressure rather than to viscous stress.

Vorticity is increasingly confined to regions near to walls on the upstream portions of a body and to wakes on the downstream side. The upstream vortical regions become boundary layers in the high Reynolds number limit. As vorticity diffuses away from the surface it is convected downstream, ultimately to form the wake. The upstream–downstream asymmetry leads to another important idea, that of separation. For example, flows into and out of a nozzle are quite different. The flow into a trumpet shaped orifice, say, will follow the walls. As the opening narrows, the flow accelerates to conserve mass [loosely, ρUA = constant, per Eq. (1.33), implies U increases as A decreases]. The accelerating flow convects vorticity toward the wall, keeping it confined near the surface.

Why Study Fluid Dynamics?

Fluid dynamics is a branch of classical physics. It is an instance of continuum mechanics. A fluid is a continuous, deformable material. It is a material that flows in response to imposed forces. This is embodied in the everyday experience of draining water from a sink. The water flows under the action of gravity. It does not have a fixed shape; it fills the sink, conforming to its shape. The water flows with variable velocity, depending on its distance from the drain. All these distinguish fluid motion from solid dynamics. As another example, a pump propels water through a pipe or through the cooling system of a car. How does the reciprocating movement of the pump produce directed flow, extending to distant parts of the cooling circuit? One way or the other, the pump must be exerting forces on the fluid; one way or the other, these forces are communicated to distant portions of the fluid and sets them in motion. It is far from obvious what the nature of that flow will be, especially in a complex geometry. It may be laminar, it may be turbulent; it may be unidirectional, it may be recirculating.

Recirculation is the occurrence of backflow, opposite to the direction of the primary stream. This can be seen behind the pedestals supporting a bridge in a swift river. Despite the strong current, the flow direction reverses, and a circulating eddy forms in a region behind the pedestal. How is such behavior understood and predicted? An understanding requires knowledge of viscous action, of vorticity, of turbulence, and of the governing equations.

Two principles distinguish compressible flow: gases heat when compressed and cool when expanded; disturbances propagate at the speed of sound. The first alludes to thermodynamics. The second alludes to gas dynamics.

Thermodynamics

Heating by compression converts work into thermal energy. This is a reversible conversion in the sense that the thermal energy can be converted back into work. Heating also occurs by frictional dissipation of fluid kinetic energy into thermal energy. That is an irreversible process; viscosity cannot convert the thermal energy back into ordered flow. Friction increases entropy.

Compression and expansion occur in the course of the motion of a gas. For instance, on approaching a blunt body, the flow will slow, and fluid elements will be compressed. That is the ultimate motive for reviewing basic thermodynamics: the governing equations of compressible flow must be consistent with thermodynamics, extended to a spatially distributed system. However, we start with the thermodynamic description of compression and expansion of a homogeneous gas and then proceed to discuss compressible fluid dynamics. Comprehensive texts (Saad, 1997) can be consulted if the reader desires a thorough treatment of thermodynamics. The following is an informal treatment that provides background to compressible flow analysis.

Define a fluid element as a fixed mass, M, of gas. This occupies a volume element, V, which contains that mass. The volume defined in this way is termed specific volume – specific properties are those associated with a given quantity of mass. The mass of the fluid element is invariant, because that is how the element is defined: its volume can change. Indeed, compressibility is the property of volume change in consequence of pressure variations.

When two fluids occupy the domain, with a sharp boundary between them, we speak of fluid–fluid interfaces or just interfaces. To the extent that the fluids are immiscible, their interface is a type of boundary. The governing laws are unchanged; the new features are boundary conditions. They are of a different nature from those at fixed, solid walls. They depend on the flow on either side of the interface; indeed, the position of the interface is itself a variable. Interface conditions are alternatively described as matching conditions: velocities and stresses on either side must properly match at the interface. Despite this complicating aspect, the view that only boundary conditions are at issue provides some clarity.

The interface may be between liquid and gas – say, water and air. Often the matching conditions are simplified in this case. The density of air is three orders of magnitude smaller than that of water. For many purposes, the forces exerted by the air on the water can be neglected; then the interface is a force-free surface, insofar as the hydrodynamics are concerned. It nevertheless is a moveable surface, whose position must be solved as part of the analysis.

Or the interface could be between two viscous fluids – say, oil and water. The viscosity jumps across their common boundary. Conditions of stress continuity then determine the interaction between the fluid motions.

Oil and water might be placed in a vertical tube. The interface then curves in consequence of surface tension and the angle of contact with the tube. The line of contact is a three-phase boundary, among water, oil, and solid wall.

The basic laws of fluid dynamics are the Navier–Stokes momentum equations described in Chapter 1. Computational fluid dynamics (CFD) is the practice of solving those equations,* along with the mass conservation equation, by numerical algorithms. The ability of such seemingly simple governing equations to describe a wealth of complex fluid motions is quite remarkable. That remarkable capability is revealed most notably by computer simulation.

Numerical solution of Navier–Stokes equations nowadays has become almost routine. A variety of algorithms and solution methods for both incompressible and compressible flow have been developed over time and successfully implemented in a large number of computational codes (Ferziger and Peric, 2002; Fletcher, 1991; Tannehill et al., 1997). Initially, this software was primarily for research, mainly in academic institutions, government labs, and corporate research centers. But the appearance (and disappearance) of a number of general purpose, commercial CFD codes has been seen since the early 1990s. These were developed for use by nonexperts, as well as by those experienced in the practice of computation. Some of these codes have matured over time, becoming increasingly powerful as the latest techniques, methods, and analytical models were adapted to their requirements, and as high-speed computing power became increasingly available. Computational capabilities, previously mastered only in the research environment (higher-order numerical schemes, multigrid methods, advanced modeling capabilities, parallel processing) are now being used widely, through the medium of software packages. The engineer, student, or scientist no longer needs to have an intimate familiarity with computational methods to make productive use of CFD. Other technologies that have facilitated CFD include the graphic user interface, software for geometry and mesh creation, and techniques for plotting and visualization.

This is a book on fluid dynamics. It is not a book on computation. Many excellent books on fluid dynamics are available: why is another needed?

In recent decades, numerical algorithms and computer power have advanced to the point that computer simulations of the Navier–Stokes equations have become routine. This vastly expands our ability to solve these equations, further extending our understanding of fluid flow and providing a tool for engineering analysis. Computer simulations are solutions of a different nature from classical exact and approximate solutions. They are numerical data rather than formulas. One of our objectives in this text is to relate computer solutions to theoretical fluid dynamics. Indeed, it is this goal, rather than computation as a tool for complex engineering analysis, that provides the guideline for this text. Computer solutions can reproduce closed-form and approximate solutions; they can illuminate the merits and limits of simple analyses; and they can provide entirely new solutions of varying degrees of complexity. The time is ripe to integrate computer solutions into fluid dynamics education.

From a pedagogical perspective, readily available, commercial computational fluid dynamics (CFD) software provides a new resource for teaching fluid dynamics. This software converts CFD from a technique used by researchers and engineers in industry into a readily accessible facility. It is a challenge to integrate such software packages into the educational structure. Most of the examples in this book have been computed with commercial software, and exercises to be solved with such software have been suggested. How far to go in this direction was a true quandary.

The terminologies creeping flow, Stokes flow, or low Reynolds number hydrodynamics are used synonymously to refer to flows in which inertia is negligible compared to viscous and pressure forces. The formal requirement is that the Reynolds number be small: Re ≪ 1. However, in practice the low Reynolds approximation often remains satisfactory for Reynolds of order unity: Re ∼ 1.

With inertia neglected, momentum is transported by viscous diffusion but not by convection. Some ideas about fluid dynamics must be rethought in this limit. Without convection, there is no wake on the downstream side of an object; pressure scales on viscosity not on kinetic energy; when they occur, eddies are as likely upstream as downstream of a blunt body.

Low Reynolds number can mean highly viscous; hence, one can imagine objects moving through syrupy fluid or syrupy fluid being pumped through a conduit. The dominant forces are frictional in origin. Of course, low Reynolds number can also mean very low velocity or very small scale. One application of creeping flow is to locomotion of microorganisms through a fluid. These animals are a few microns in size. They do not move by propulsion; they drag themselves through the fluid, pushing or pulling by frictional forces. In some cases, they use spiral flagella to corkscrew themselves along. On their relative scale, the fluid appears to be very viscous. One can imagine pushing against a very thick fluid to move forward. To do so, the frictional force pushing forward must be greater than the frictional force resisting motion. Swimming is possible if the organism can produce motions that create more pushing friction than impeding friction.

It is likely that most questions the reader might pose about turbulent flow have no satisfactory answer - questions like. What is its cause? How can equations as innocuous as the Navier-Stokes momentum equations produce such complex solutions? How can we describe it? How do we predict its properties? and so on. The phenomenon is common experience: turbulent eddying is seen in smoke billowing above a large fire, in dust clouds rising from an explosion, in the wake of a last-moving boat; it is heard in the roar of a jet engine, in the wind rushing over an automobile; it is fell when an airplane bobs up and down in it or when a stiff breeze blows in one's face. Turbulence is an essential element of many processes. A text on fluid mechanics is not complete without a chapter on turbulence. That said, we provide, in this chapter, an introduction to computation of turbulent flow. The reader interested in a more thorough treatment of the subject can consult books entirely devoted to turbulence, such as Pope (2000).

The word turbulence conjures up the notion of randomness. It has entered everyday vocabulary, divorced from the field of fluid mechanics. It evokes images of roiling, churning, and disorder. These are valid definitions, but in fluid flow it is often less severe than vernacular usage suggests. A 10% level of velocity fluctuation may be considered to be substantial. Turbulence is best defined as the irregular component of motion that occurs in fluids when the Reynolds number is sufficiently high. The irregularity may be mild or it may be severe.

Introduction

In Chapter 4e, a computational method based on vorticity confinement (VC) is described that has been designed to capture thin vortical regions in high-Reynolds-number incompressible flows. The principal objective of the method is to capture the essential features of these small-scale vortical structures and model them with a very efficient difference method directly on an Eulerian computational grid. Essentially, the small scales are modeled as nonlinear solitary waves that “live” on the lattice indefinitely. The method allows convecting structures to be modeled over as few as two grid cells with no numerical spreading as they convect indefinitely over long distances, with no special logic required for merging or reconnection. It can be used to provide very efficient models of attached and separating boundary layers, vortex sheets, and filaments. Further, the method easily allows boundaries with no-slip conditions to be treated as “immersed” surfaces in uniform, nonconforming grids, with no requirements for complex logic involving “cut” cells.

There are close analogies between VC and well-known shock- and contact-discontinuity-capturing methodologies. These were discussed in Chapter 4e to explain the basic ideas behind VC, since it is somewhat different than conventional computational fluid dynamics (CFD) methods. Some of the possibilities that VC offers toward the very efficient computation of turbulent flows, which can be considered to be in the implicit large eddy simulation (ILES) spirit, were explored. These stem from the ability of VC to act as a negative dissipation at scales just above a grid cell, but that saturates and does not lead to divergence.