PX264 Physics of Fluids

Nicholas d’Ambrumenil
2023
CONTENTS 1

Contents

Preface i

1 What is a fluid? 1

2 Laminar, steady, viscous flow 6

2.1 Poiseuille flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.2 Viscosity and Reynolds number . . . . . . . . . . . . . . . . . . . . . . . . 10

3 Navier-Stokes equation 13
3.1 Core equation summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
3.2 Boundary conditions and boundary layers . . . . . . . . . . . . . . . . . . . 17

A Essential Differential Calculus 20

B Viscous Stresses 23
Preface

The field of fluids is one of the richest and most easily appreciated in physics. Tidal waves,
cloud formation and the weather generally are some of the more spectacular phenomena
encountered in fluids. The module establishes the basic equation of motion for a fluid - the
Navier-Stokes equation - and shows that in many cases they can yield simple and intuitively
appealing explanations of fluid flows. The module concentrates on incompressible fluids.
The module should explain why PDEs (with associated boundary conditions) are an appropriate
model for fluids. You should learn how physical ideas and limiting cases can help analyse these
PDEs which, in general, cannot be solved. These include the role of the Reynolds number,
laminar viscous flow, the boundary layer concept and irrotational flow. The module also
prepares you for future applied mathematics modules.
Assumed Mathematical Knowledge
The module makes extensive use of vector calculus. What you need to know is summarised
in Appendix A.
Recommended texts

• LD Landau and EM Lifshitz, Fluid Mechanics, Pergamon;

• DJ Tritton, Physical Fluid Dynamics, OUP;
• TE Faber Fluid Dynamics for Physicists, CUP

Acknowledgements
The notes for this module have been developed over time by Tony Arber (who first typed
the notes and produced almost all the ”artwork”, 2019-2022), Julie Staunton (2011-2018),
Mario Nicodemi (2009-2010, now in Naples), Boris Mouzykantskii (1996-2008, later ad-tech
business tycoon) and myself (NdA, 1990-1995, 2023).

i
Chapter 1

What is a fluid?

What is a fluid? - Any substance which cannot resist a shear force without motion and does
not return to its initial state when the stress is removed, e.g. liquids and gases.

• For simple steady flows analytic progress possible;

• Complex (real) flows often require numerical solution and, even then, finding accurate
solutions can be intractable.

This first module on the physics of fluids will introduce the principles and equations of fluid
mechanics. The main objectives are to understand the core equations and how they help
explain, for example, aerodynamic drag and flight.
Stress (τ ) is defined as the force per unit area, and so has the dimensions of pressure. The
force for a shear is tangential to the area.

F
τ=
A

If a stress is applied to a solid, for example the bottom surface is held fixed and a tangential
force is applied to the top, then it will deform. This deformation is measured through the
strain e which is defined as
∂X
e=
∂y
where X(y) is the displacement of the upper surface due to an applied stress τ .

1
CHAPTER 1. WHAT IS A FLUID? 2

For small displacements, or alternatively differential distances in y,

∂X
= tan θ ' θ
∂y

In general the strain has components due to ∂X

∂z
, ∂Z
∂x
etc. which is why we have used partial
derivatives even though for this example X = X(y) only. For small displacements in a solid
stress is proportional to strain so that

τsolid = G e

where G is the shear modulus. If this stress is removed the solid object returns to its original
shape.

Solid: Internal forces

balance the applied stress

Fluid: No restoring force and

the strain grows with 9me as
it flows.

For a solid it is the binding of atoms into the solid structure which gives rise to the stress
when subjected to a strain and hence no stress. A fluid state does not have this bonding
and there is no restoring force to the initial position. The only force which can act now is
friction between layers slipping over each other. If the strain, i.e. θ, remains constant the
fluid experiences no force (stress). Hence for a fluid τf luid 6= Ge. Instead the strain must be
changing in time so that layers of fluid experience a frictional force.
For a fluid the relation between stress and strain must be
de
τf luid = µ (1.1)
dt
as there must be a time dependent strain e(t). Here the new constant µ is called the viscosity.
Since this module deals only with fluids we drop the subscript fluid on τ from now on. Note
CHAPTER 1. WHAT IS A FLUID? 3

that we still have e = ∂X

∂y
so that

de d ∂X ∂ux
τ =µ =µ =µ (1.2)
dt dt ∂y ∂y

where we have introduced the fluid velocity in the x-direction ux = dX

dt
. Hence for a fluid

∂ux
τ =µ
∂y
for the case where we only have a ux = ux (y). In general there will be 9 components of
stress and strain due to the possible shear flows ∂u
∂x
x
, ∂u
∂y
x
, ∂u
∂z
x
, ∂u
∂x
y
. Actually these are not all
relevant as discussed in Appendix B.
We could set up an experiment with parallel plates, fix the base and slowly move the top plate
then measure the applied force and fluid motion as below.

In this way we could measure the viscosity µ. Note that for a fluid if we stop applying the
force to the top plate, i.e. remove the stress, the fluid does not return to its original position
- unlike a solid.
In many text books the fluid velocity is represented as u. In this course we will use both u
and v for the fluid velocity interchangeably. There are no occasions where this should lead
to any confusion. Many fluid textbooks use (u, v, w) as the components of the fluid flow so
that u = ux etc. This convention will not be used in this module.
Newtonian and Non-newtonian behaviour
If µ 6= µ(u), i.e. the viscosity is not a function of fluid velocity, then this is a Newtonian
fluid. We will only deal with Newtonian fluids in this module. To a good approximation air
and water are Newtonian fluids.
If µ = µ(u) then this is a non-Newtonian fluid. Non-newtonian behaviour is found for example
in polymer melts, solutions of polymers, emulsions (solids dispersed through a fluid). It can
be important in biophysics (blood flow in capillaries) and paints amongst others.
Thinking through a sketch derivation of 1.2 can help see why it is an approximation. For very
fast changes fluids will resist shear and behave like a solid. Fast here means faster than the
system can respond. For example, skimming stones thrown onto the sea will bounce off the
surface as it were solid. Alternatively don’t try belly-flopping into a diving pool from the 10m
platform.
OneNote 23/11/2022, 15)17

Relaxation
CHAPTER 1. WHAT IS A FLUID?
Wednesday, 23 November 2022 09:40 4

The figure shows schematically how a fluid would respond to an instantaneous shear. Initially
the stress will be as in a solid. As the fluid starts to flow the stress relaxes into the new
sheared state. The stress response at t to the shear applied at t0 would be

τ (t) = Gf (t − t0 )e

where the function f describes the relaxation process. For a general applied strain e = e(t0 ),
the stress would be
Z t
de t
Z
0 de 0 0 de
τ (t) = Gf (t − t ) 0 (t ) dt ≈ Gf (t − t0 )dt0 ≡ µ .
−∞ dt dt −∞ dt

This expression is the result we assumed, see 1.2.

Lagrangian & Eulerian fluid elements
There are two common descriptions of fluid motion. These are usually defined in terms of
fluid elements. A fluid element is a section of the fluid, which we can visualise as a cube,
which may move with the fluid or be fixed in space. The element may be deformable if it
moves with the fluid. We consider the limit of small elements to derive differential equations.
A Lagrangian fluid element moves with the fluid in such a way that there is no flow of fluid into
or out off the element. Hence the fluid mass is constant within a Lagrangian fluid element. If
we imagine a 2D flow and set up an initial grid of elements then this grid becomes deformed
https://ukc-onenote.officeapps.live.com/o/onenoteframe.aspx?ui=en…rectionreason=Force_SingleStepBoot&rct=Medium&ctp=LeastProtected Page 1 of 2

as the fluid moves (see below). Moving with the fluid the velocity is only a function of time
so that the velocity of a fluid element is
d dx dy dz
u= (r − r0 ) = î + ĵ + k̂
dt dt dt dt
CHAPTER 1. WHAT IS A FLUID? 5

This is the direct analogy with normal Newtonian mechanics of point particles where we
need only consider u = u(t). The Lagrangian coordinate system is fixed with the fluid so
the Lagrangian coordinates themselves move. In other words the fluid is stationary on the
Lagrangian grid. However, fluid the motion may rapidly become complex as all the fluid
elements begin to mix. Hence an alternative approach is often used - Eulerian fluid elements.
Eulerian fluid elements are fixed in space and do not move with the fluid. Hence the mass
in a fluid element may vary with time. The initial Eulerian grid of fluid elements is now fixed
so that, in the Eulerian coordinate system, the fluid velocity u = u(x, y, z, t) where (x, y, z)
defines the initial Eulerian grid.

Figure 1.1: Sketch of motion through an Eulerian grid (top) and with a Lagrangian grid
(bottom)
 Chapter 2

Laminar, steady, viscous flow

Now for a first calculation of a fluid flow. This is for laminar, viscous and steady flow between
two stationary plates. The flow is driven by a pressure gradient and we will assume well-
behaved planar flows. Specifically:

Laminar: The flow has a smooth profile, usually matching the shape of material surfaces. For
this example of flow between two parallel plates, the flow is along surfaces parallel to
those plates.

Viscous: Viscosity is important. Many of the flow problems we will look at later, will have not
treat viscosity. This is an important distinction.

Steady: The fluid velocity vector flow field is not a function of time.
OneNote
u(r, t) = u(r).

The configuration with labels is shown in the diagram below. Note that, in this diagram, we
have assumed Poiseuille1
that the fluid flow is zero on the boundaries (more on this later) and that the
flow is entirely in the x-direction. The flow is driven by a pressure difference of P1 − P2 over
a length L. With these assumptions we have that u = ux (y)î
Friday, 23 December 2022 17:54

6
CHAPTER 2. LAMINAR, STEADY, VISCOUS FLOW 7

The fluid will flow as there is a pressure difference. Since the flow is zero on the boundaries,
there must be a flow gradient (the fluid travels faster at y = 0 as shown). The flow gradient
(v depends on y) means that fluid elements will experience a viscous force. Consider a small
fluid element as shown below such that the top and bottom surfaces have area A = ∆x∆z
The viscous force on the fluid element will be
∂ux
OneNote Fv = Aµ . 23/12/2022, 18)07
∂y
This will be in the positive x-direction on the top and negative x-direction on the bottom of
Stress
the element of fluid for the flow profile below.
Wednesday, 21 December 2022 17:44

The net viscous force on the fluid element, assuming µ is constant, is given by
(   )
∂ux  ∂ux 
Fv = µ −µ ∆x∆z.
∂y y+∆y ∂y y

From the definition of a derivative in the limit of vanishing fluid element volume this gives
∂ 2 ux
https://ukc-onenote.officeapps.live.com/o/onenoteframe.aspx?ui=en%2DUS&rs=en%2DGB&wopisrc=http…tc=1&cac=1&mtf=1&sfp=1&wdredirectionreason=Force_SingleStepBoot&rct=Normal&ctp=LeastProtected Page 1 of 2
Fv = µ ∆x∆y∆z.
∂y 2
(The final result will be correct for any flow profile as all sign changes will work through to
give the same result for a negative flow gradient.)
The flow is assumed to be steady (there is no acceleration anywhere) so this force must balance
the force on the fluid element due to the pressure gradient. The force in the x-direction due
to the pressure is given by
∂P
Fp = {P (x) − P (x + ∆x)}∆y∆z = −
∆x∆y∆z
∂x
Balancing the viscous and pressure forces, Fp + Fv = 0, gives
∂P ∂ 2 ux
=µ 2 .
∂x ∂y

The LHS is only a function of x and the RHS only a function of y so both sides must equal
the same constant. For convenience later, we choose this to be −Q giving
∂ 2 ux Q 2
µ = −Q giving ux (y) = − y + Ay + B.
∂y 2 2µ
CHAPTER 2. LAMINAR, STEADY, VISCOUS FLOW 8

The constants of integration A, B are found from the boundary conditions that
Q 2
ux (a) = ux (−a) = 0 resulting in ux (y) = a − y2 .


2µ
The constant Q is determined from the pressure difference along the plates
∂P P1 − P2
Q=− = .
dx L
The final result is that the steady, viscous, laminar flow between two parallel plates has a
parabolic profile. This example is illustrative but not very practical as this is a 1D problem and
we have assumed that the z-direction is infinite and ignored any boundaries in the z−direction.

2.1 Poiseuille flow

Poiseuille flow is cylindrically symmetric, viscous, steady, laminar flow through a cylindrical
pipe. This is essentially a repeat of the calculation for steady, viscous, laminar flow between
parallel plates but in cylindrical geometry. We will use the usual (r, φ, z) coordinates. The
pipe radius is a and a pressure difference is maintained along the pipe by having the left hand
end attached to a deep water reservoir (whose depth we assume constant) and the right hand
end open to the atmosphere.

As before we use the laminar assumption to set u(r, t) = uz (r)k̂ (sorry but will use ẑ and k̂
for unit vector in z-direction for no obvious reason) and consider the force balance on a fluid
element required for steady flow. Now the fluid element needs to be in cylindrical geometry.
The inner radius of the element is at r − 0.5∆r and its outer radius at r + 0.5∆r as shown
below.
CHAPTER 2. LAMINAR, STEADY, VISCOUS FLOW 9

The viscous force on this element is given by

 
∂uz  ∂uz 
Fv = ∆z(r + ∆r/2)∆φµ − ∆z(r − ∆r/2)∆φµ .
δr r+∆r ∂r r

Note that the area used to find the force from the stress is different on the inner and outer
surfaces. We can now Taylor expand

∂ 2 uz 
  
∂uz  ∂uz 
' + ∆r + ··· .
∂r r+∆r ∂r r ∂r2 r

Substituting this into the expression for the viscous force and ignoring all terms in ∆r2 , we
obtain
∂ 2 uz
 
∂uz
Fv = µ∆z∆φ∆r +r 2
∂r ∂r
which can be rewritten as  
∂ ∂uz
Fv = µ∆z∆φ∆r r .
δr ∂r
The total pressure force in the z-direction on the fluid element is
∂P
Fp = r∆φ∆r(P (z) − P (z + ∆z)) ' −r∆φ∆z∆r . .
∂z
As before these forces must balance for steady flow so Fp + Fv = 0 which gives
 
∂ ∂uz ∂P
µ r =r
∂r ∂r ∂z

which can be re-arranged to give

 
∂P 1 ∂ ∂uz
=µ r .
∂z r ∂r ∂r

Note the factors of r not present in the planar problem. Once again both sides of this equation
can be set equal to the constant −Q and integrated. After doing this and finding the constants
of integration using the boundary conditions

uz (a) = 0 and uz (0) is finite

gives
Q 2
(P1 − P2 ) 2
a − r2 = a − r2 .


uz (r) =
4µ 4µL
As for the planar flow the Poiseuille flow profile is a parabola but now in radius r.
Poiseuille flow rate
What is the volume or mass of water flowing through the pipe per second for Poiseuille flow?
Assume the mass per unit volume (the mass density) ρ is constant. In unit time, the mass of
fluid flowing through the element cross-sectional area r∆φ∆r is ρuz r∆φ∆r.
 Cylinder
Wednesday, 21 December 2022 17:49

CHAPTER 2. LAMINAR, STEADY, VISCOUS FLOW 10

The total mass flowing down the pipe per second is

Z 2π Z a
M= ρuz (r) r drdφ
0 0 .
πρ
= (P1 − P2 ) a4
8µL

Remember this formula only applies to steady, viscous and laminar flows. It clearly fails for a
fluid where the viscosity can be ignored as then µ = 0 and the flow rate diverges.
Steady, viscous flow
The general equation for fluid steady, viscous flow where the only forces are viscosity and
pressure in 3D is
∇P = µ∇2 u.

The differential operators ∇·, ∇, ∇× and ∇2 take on different forms depending on whether
the problem is solved in Cartesian, cylindrical or spherical geometry. For example in Cartesian
geometry
∂ 2 ux
(∇2 u)x =
∂y 2
if u = ux (y)x̂. But if instead we use cylindrical geometry with u = uz (r)ẑ then
 
2 1 ∂ ∂uz
(∇ u)z = r .
r ∂r ∂r
These will be given in the rubric of the exam paper if not in Cartesian coordinates.

2.2 Viscosity and Reynolds number

What is the microscopic origin of viscosity? First define some basic quantities.

• Average molecular speed umol due to random thermal motion. This is a function of
temperature and the same magnitude as the sound speed in the fluid.

https://ukc-onenote.officeapps.live.com/o/onenoteframe.aspx?ui=en…irectionreason=Force_SingleStepBoot&rct=Normal&ctp=LeastProtected
CHAPTER 2. LAMINAR, STEADY, VISCOUS FLOW 11

• Fluid centre of mass speed u (or u). In this course we will mostly deal with subsonic
motion so |u| < umol .

At constant temperature umol is constant but u can be a function of space and time as in the
examples above. As different layers of fluid are moving at different speeds molecular diffusion
can transfer momentum between fluid layers. This is the microscopic source of viscosity. In
reality these random processes also convert some centre of mass kinetic energy into random
motion, i.e. viscosity leads to heating, but we won’t consider this effect in this course.
Viscosity is a property of the fluid. For example For gases µ increases with temperature, for
liquids it decreases. Often this course and textbooks will use the kinematic viscosity ν defined
through
µ
ν= .
ρ
where as usual ρ is the fluid mass density - the mass per unit volume. Confusingly ν is also
often called the viscosity and we drop the ‘kinematic’. This is sloppy but so common it is best
to be aware of this usage. It is rarely a problem and usually clear from the context.
Reynolds number
For the viscous flow solutions dealt with so far the viscosity needs to be ‘large’ but large
compared to what? How do we quantify this? The formula for one component of the stress
is
∂ux
τ =µ
∂y
and τ has the units of pressure. Hence the dimensions of viscosity µ are [M ][L]−1 [T ]−1 . For
a fluid with a typical mass density ρ0 , typical flow speed u0 and for flows on a characteristic
length-scale L0 the product ρ0 L0 u0 has the same dimensions as µ. We use this to define a
dimensionless parameter Re, called the Reynolds number, through
ρ 0 L0 u0
Re = .
µ
The flow is viscous-dominated when Re is small. How small is determined from experiments
- usually u0 is varied until the flow is no longer laminar (and steady). Note that the definition
of Re is only a rough estimate as there is often a range of sensible values we could choose for
L0 and u0 . For example for the Poiseuille flow the typical speed could be defined at v max /2
or as the average speed
Z 2π Z a
1 Qa2
u0 = 2 u(r)rdrdφ = .
πa 0 0 8µ
If we then choose L0 = 2a = d0 this gives
Qd30 ρ0
Re = .
32µ2
Experiments show that water flow in a pipe is laminar up to around Re < 3200 so for a pipe
with a 5 cm diameter
ρwater = 103 kgm−3
µw = 10−3 Ns m−2
µw Re
u0 = = 5 cm s−1 .
ρ0 L0
OneNote 18/01/2023, 09*13

CHAPTER 2. LAMINAR, STEADY, VISCOUS FLOW 12

PipeFlowInstability
For flow with u17:52
0 < 5 cm/s the flow is laminar but above this speed the flow may become
Wednesday, 21 December 2022
complex.

We need an equation set which can describe motion for all values of Re, i.e. non-steady,
non-laminar flows with or without viscosity. This is the Navier-Stokes equation.

https://ukc-onenote.officeapps.live.com/o/onenoteframe.aspx?ui=en…irectionreason=Force_SingleStepBoot&rct=Normal&ctp=LeastProtected Page 1 of 1
Chapter 3

Navier-Stokes equation

We will treat the fluid as a continuum - this is called the continuum hypothesis. It ignores
anything at the molecular level and is a good approximation for fluid length scales much larger
than the typical mean-free-path (the distance between molecular/atomic collisions). In this
case we need the equation for the conservation of mass ignoring individual particles.
Consider a small volume of the fluid with volume V bounded by a surface S. The total surface
area is A and the outward normal to a small element of the surface is dS

The mass flow per second through the vector element dS is ρu.dS. This follows from noting
that for uniform flow of speed v through an area A normal to the flow the mass flow rate is
vρA and applying this to a small vector area. Integrating over the whole surface area then
gives the mass outflow rate I
ρ u · dS.
S
Since mass is conserved this
R must equal the rate of mass loss in the volume. At any time t
the mass in the volume is V ρdV so combining this give
Z I
∂
ρdV = − ρu · dS.
∂t V s

For a surface which is stationary, i.e. an Eulerian fluid element, the divergence theorem for a
smooth, differentiable vector field A states that
I Z
A · dS = (∇ · A)dV
s V

13
CHAPTER 3. NAVIER-STOKES EQUATION 14

so the mass conservation equation can be re-written as

Z  
∂ρ
+ ∇ · (ρu) dV = 0.
v ∂t

As this must be true for any volume in the fluid

∂ρ
+ ∇ · (ρu) = 0.
∂t
This is the continuity equation and expresses conservation of mass for a fluid. Note this is an
equation for the mass density ρ but in order to solve this equation we need the fluid velocity
vector field u. This we derive in the next section. This equation applies in contexts other
than fluids. Normally we think of ρu as a mass current density. In quantum mechanics, ρ
becomes the probability density and j the probability current density. In EM theory ρ is the
charge density and j the current density.
Note that if the fluid is incompressible, i.e. the density cannot change, then the continuity
equation simplifies to
∇ · u = 0.
It is surprising how often this is a good approximation and we will use this approximation
many times.
Navier-Stokes equation
Up to this point we have found solutions for steady flow but what if the flow isn’t steady? In
this case we need to apply F = ma to a Lagrangian fluid element. In such an element the
mass is constant so that ρdV is constant where dV is the fluid element volume. By applying
Newton’s law to a Lagrangian element the rate of change of momentum can be written as
d d
(ρ dV u) = ρ dV u.
dt dt

We need to be careful about taking the total derivative of the velocity vector field. Following
the usual convention in fluid mechanics we use dtd = DtD
and introduce capital ‘D’ for total
derivative. There as absolutely no good reason for doing this other than convention. Since
the fluid element is moving we have
 
Du u(r(t + ∆t), t + ∆t) − u(r(t), t)}
= lim
Dt ∆t→0 ∆t
which can be Taylor expanded in Cartesian coordinates to give
  
Du 1 ∂u ∂u ∂u ∂u
= lim u(r, t) + ∆x + ∆y + ∆z + ∆t − u(r, t) .
Dt ∆t→0 ∆t ∂x ∂y ∂z ∂t
In this expression, noting that  
∆x(t)
lim = ux
∆→0 ∆t
gives
Du ∂u ∂u ∂u ∂u
= + ux + uy + uz
Dt ∂t ∂x ∂y ∂z
∂u
= + (u.∇)u .
∂t
CHAPTER 3. NAVIER-STOKES EQUATION 15

For Cartesian coordinates

∂ ∂ ∂
∇ = î + ĵ + k̂ .
∂x ∂y ∂z

Throughout this module note that the following are all equivalent.
d D ∂
= = + u.∇
dt Dt ∂t
and is called the advective derivative (although it is also just the total derivative as usually
used in calculus). The justification for introducing a new name and notation is to emphasise
the meaning of the two terms in the advective derivative. The first term ∂t is the rate of
change of the fluid property at a fixed point in space. The second term u.∇ is the additional
rate of change due to the fluid element moving - hence the total advective derivative name.
We now have that the ma side of Newton’s second law for the Lagrangian fluid element is
Du
ρ dV .
Dt
This needs to be set equal to the forces. We already have some of these from the steady,
viscous flow calculations earlier. Here we will drop the factors of element volume dV . We are
actually calculating force per unit volume, or force density, but we will call it the force as a
shorthand. This is common in fluids.

• The pressure force is −∇P

• The force due to gravity is ρg = −ρg k̂ where we have assumed that the gravity points
downward in the -z direction. This will be true for all examples in this module.

• We have calculated one component of the viscous force due to ∂y ux but there are also
components due to ∂x uy etc. and if we included all possible terms then we would find
that the full viscous force is (see Appendix B for more on this)

Fvisc = µ∇2 u + λ∇(∇.u).

All of this can be combined to give “F = ma” for a Lagrangian fluid element.
∂u
ρ + ρ(u · ∇)u = −∇p + µ∇2 u + λ∇(∇.u) − ρg k̂.
∂t
This is usually called the Navier-Stokes equation although we will always use a simplified
version in this module. We will assume that the viscous term, which includes ∇ · u, can be
ignored. This is true if the fluid is incompressible but for simplicity we will always assume it is
a good approximation. (The quantity λ is usually called the volume or bulk visosity.) Hence
for this module the Navier-Stokes (N-S) equation is

∂u
ρ + ρ(u · ∇)u = −∇p + µ∇2 u − ρg k̂.
∂t

Note that the full continuity equation for ρ involves u. Similarly the N-S for u includes the
pressure P so we now need an equation for P .
CHAPTER 3. NAVIER-STOKES EQUATION 16

Pressure equation
In this module we will assume that if we need a pressure equation (which will not always be
the case), then changes to pressure will be assumed to be adiabatic. This means that changes
to the fluid are sufficiently quick that heat flow into/out of a fluid element can be ignored.
For an element of volume V moving with the fluid the adiabatic equation we will use is that
of an ideal gas, P V γ = constant. (Non-ideal gases need a knowledge of the equation of state
for the fluid. Often only approximate equations are known.) Here γ is the ratio of the specific
heat at constant pressure and constant volume. Moving with a Lagrangian fluid element the
mass is conserved so that V = m/ρ where m is the element mass. Putting this together gives
that P/ργ is constant moving with the fluid. Hence
 
d P
=0
dt ργ

This completes our set of equations for ρ, u and P but note they are coupled and non-linear.

3.1 Core equation summary

Let’s summarise the core equations we have just derived so they are in one place for reference.
The continuity equation is
∂ρ
+ ∇ · (ρu) = 0
∂t
If the fluid is incompressible (ρ is constant) this reduces to

∇ · u = 0.

The momentum equation (F = ma for a fluid) is

∂u µ
ρ + ρ(u · ∇)u = −∇p + µ∇2 u + ∇(∇.u) − ρg k̂.
∂t 3
Here we have assumed that gravity acts downwards in the negative z-direction. In this course
we will assume incompressibility for the viscous term so that the Navier-Stokes equation
(formally the incompressible N-S) we use is

∂u
ρ + ρ(u · ∇)u = −∇p + µ∇2 u − ρg k̂
∂t
If we need to determine the pressure then we assume adiabatic processes and
 
d P
=0
dt ργ

For the flow over objects, we will need to satisfy the boundary conditions. We will see in
Section 3.2 that u = 0 on a stationary solid boundary or u = ub for a solid boundary moving
with velocity ub .
There are sub-sets of the equation set above in boxes.
CHAPTER 3. NAVIER-STOKES EQUATION 17

• Euler’s equations - These are the set of boxed equations above but with no viscosity.
Also called the inviscid N-S equations.

• Incompressible - The mass density remains constant. We have had this already and it
replaces the continuity equation with ∇ · u = 0

• Laminar, steady, incompressible, viscous flow - for example Poiseuille flow. Here steady
means ∂t = 0. Incompressible means we don’t need the continuity equation and can
just set ρ to a constant. Laminar for Poiseuille meant that u = uz (r)k̂ so (u·∇)u = 0.

Curious fact: this isn’t examinable but you may be curious why I keep saying ‘if’ we need
the pressure. If the fluid flow is incompressible ∇ · u = 0. We can take the curl of the N-S
equation. This gives an equation for ∇ × u so we know both the divergence and curl of the
velocity. By Helmholtz’s theorem (it is possible that you haven’t seen this before, which is
why this paragraph isn’t examinable) this is enough to uniquely specify any vector field. But
note that taking the curl of the N-S equation removes the pressure as ∇ × ∇P = 0 so we
don’t need to know the pressure either to find an incompressible flow!

3.2 Boundary conditions and boundary layers

Differential equations need boundary conditions to find unique solutions. In time this means
we have to specify the initial conditions. In space we usually assume we know the fluid density,
velocity and pressure far from an object, the far-field boundary condition. This just leaves
the boundary conditions on any object in the fluid flow. We will always assume the ‘no-slip’
boundary condition for objects in the flow as well as the ‘no-flux’ condition (nothing can flow
into a solid). This means that on the surface of a material in the flow the fluid vector velocity
field matches the velocity of the boundary. Hence if the object is stationary the fluid flow
must be zero on the boundary.
Boundary layers
We have solved for steady, laminar, viscous flow and stated that viscosity is ‘large’ and im-
portant. This was quantified by saying that the Reynolds number
ρ0 Lo u0
Re =
µ
is ‘small’. How small can only be found by experiment. When in this regime we solved the
equation
∇P = µ∇2 u.
However, we now know the full equation when it is not steady is given by N-S
∂u
ρ + ρ(u · ∇)u = −∇p + µ∇2 u − ρg k̂.
∂t
OneNote 23/12/2022, 18)16

CHAPTER 3. NAVIER-STOKES EQUATION

Object L0 18
Wednesday, 21 December 2022 17:54

Let’s look at the Reynolds number again for flow over a simple obstacle. Where u0 is the
typical flow speed, ρ0 the typical density (or just the constant density for incompressible flow)
and L0 is some scale-length of interest.
The term ρ(u · ∇)u in the N-S equation is usually called the inertial term—it’s to do with
the well-known inertial effect called acceleration(!). Note though that this doesn’t include the
time derivative. Consider the typical size of the inertial and the viscous terms.

ρu20 µu0
|ρ(u · ∇)u| ∼ , |µ∇2 u| ∼ .
L0 L20

Here we have taken |∇| ∼ L−1 0 and |∇ | ∼ L0 so we are considering the obstacle length to
2 −2

be the important scale-length. If we take the ratio of the estimated magnitude of the intertial
and viscous terms we get

Inertial u2 L2 ρu0 L0
∼ρ 0 · 0 = = Re
Viscous L0 µu0 µ
So the Reynolds number is an estimate of the ratio of the magnitude of the intertial to viscous
terms in the N-S equation. There are two important limits.

• Re  1 and safe to ignore viscosity. Inviscid flow. µ = 0.

• Re  1 in which case
∂u
= −∇p + µ∇2 u − ρg k̂.
ρ
∂t
If in addition the flow is steady and we can ignore gravity then ∇P = µ∇2 u.

For many flows Re  1 is satisfied and viscosity can be ignored. However, this cannot be true
near the surface of a solid boundary as the boundary condition is u = 0. The NS equation
https://ukc-onenote.officeapps.live.com/o/onenoteframe.aspx?ui=en…irectionreason=Force_SingleStepBoot&rct=Normal&ctp=LeastProtected Page 1 of 1

is second order and its solutions satisfy both the no flux condition u⊥ = 0 and the no-slip
condition uk = 0. The Euler equation, which is the NS equation without the viscous term, is
only first order and its solutions cannot normally satisfy both boundary conditions.
For example consider flow over a plate.
CHAPTER 3. NAVIER-STOKES EQUATION 19

If the Reynolds number is large then viscosity is unimportant for the main flow but the velocity
must be zero on the boundary. We would therefore expect a sharp velocity gradient near the
boundary and since the viscous stress is proportional to the velocity gradient viscous terms
may be important near a boundary. As a result of this the scale-length of interest for viscosity
near a boundary is not d from the figure above but δ - some small scale transverse to the
boundary. Hence
ρu2
|ρ(u · ∇)u| ∼ 0 ,
d
µu0
|µ∇2 u| ∼ 2 .
δ
We would expect viscosity to be important when these two estimates are roughly equal, i.e.
when
ρu20 µu0
∼ 2
d δ
which can be re-arranged to give
δ 1
∼ .
d Re1/2
For large Reynolds number flows we expect a thin layer near the boundary where viscosity
cannot be ignored. This is called the boundary layer . Outside the boundary layer we can take
µ = 0 as an accurate model for the flow but this must be matched to the viscous solution
near the boundary.
Appendix A

Essential Differential Calculus

You should look through the material given here—we will not go through this in lectures. The
background to these results should have been covered either in PX275 Mathematical Methods
for Physicists or MA259 Multivariable Calculus/PX276 Methods of Mathematical Physics.
Here’s the differential calculus, which will be assumed in this module and the examination.
All symbols have their usual meaning. We consider an arbitrary vector field A.
You need to remember the formulae for the divergence and curl of A in Cartesian coordinates.
This module uses (î, ĵ, k̂) for the (x, y, z) unit vectors.
Gauss’s theorem I Z
A · dS = (∇ · A) dV
S V

Stokes’s Theorem I Z
A · dl = (∇ × A) · dS
c S

In cylindrical coordinates (r, θ, z) the vector is A = Ar r̂ + Aθ θ̂ + Az ẑ

20
APPENDIX A. ESSENTIAL DIFFERENTIAL CALCULUS 21

and the unit vectors are related through

r̂ = cos θî + sin θĵ
θ̂ = − sin θî + cos θĵ.

Note that r̂ and θ̂ are both functions of θ. In this coordinate systems you should remember
that the grad operator is given by
∂ 1 ∂ ∂
∇ = r̂ + θ̂ + ẑ .
∂r r ∂θ ∂z

Cylindrical geometry results

The module often uses cylindrical geometry. The expressions for differential operators like the
curl and the divergence will be given in an examination.
Since r̂ and θ̂ are functions of θ we can use the expressions above to show
∂ r̂
= θ̂
∂θ
and
∂ θ̂
= −r̂.
∂θ
The fact that the unit vectors vary with position makes divergence and curl more complex in
cylindrical geometry. For example
     
∂ 1 ∂ ∂
∇ · A = r̂ · A + θ̂ · A + ẑ ·A
∂r r ∂θ ∂z
The derivatives of the unit vectors must be computed. For example
   
1 ∂ 1 ∂ Ar 1 ∂Aθ
θ̂ · A = θ̂ · (Ar r̂ + Aθ θ̂) = +
r ∂θ r ∂θ r r ∂θ
Here we have applied the chain rule to the derivatives and used the equations above for the
derivatives of the unit vectors.
Repeating this procedure and re-arranging gives the divergence as
1 ∂ 1 ∂Aθ ∂Az
∇·A= (rAr ) + +
r ∂r r ∂θ ∂z
and the curl as
1 ∂Az ∂Aθ
(∇ × A)r = −
r ∂θ ∂z
∂Ar ∂Az
(∇ × A)θ = −
∂z ∂r
1 ∂ 1 ∂Ar
(∇ × A)z = (rAθ ) −
r ∂r r ∂θ
or in terms of a determinant as
 
 r̂ rθ̂ ẑ 
1 ∂

∂ ∂ 
∇ × A =  ∂r ∂θ ∂z
.
r 
 Ar rAθ Az 
APPENDIX A. ESSENTIAL DIFFERENTIAL CALCULUS 22

The Laplacian for scalar f is

1 ∂ 2f ∂ 2f
 
1 ∂
2 ∂f
∇f= r + 2 2 + 2
r ∂r ∂r r ∂θ ∂z

from which we can find the Laplacian of a vector

2 ∂Aθ Ar
∇2 A = ∇2 Ar −


r
− 2
r2 ∂θ r
2 ∂A r A θ
∇2 A θ = ∇2 Aθ + 2


− 2
r ∂θ r
∇2 A z = ∇2 Az

The Advective Term

The advective term which appears in fluid mechanics (u · ∇)u can be evaluated in cylindrical
coordinates from the general formula

∂Br Aφ ∂Br ∂Br Aφ Bφ

((A · ∇)B)r = Ar + + Az −
∂r r ∂φ ∂z r
∂Bφ Aφ ∂Bφ ∂Bφ Aφ Br
((A · ∇)B)φ = Ar + + Az +
∂r r ∂φ ∂z r
∂Bz Aφ ∂Bz ∂Bz
((A · ∇)B)z = Ar + + Az .
∂r r ∂φ ∂z

While these look complex they all follow from the basic definition of the grad operator ∇ and
the fact that the unit vectors are functions of φ.
One result that we will use is that
1
(u · ∇)u = ∇v 2 − u × (∇ × u). (A.1)
2
While the derivation is not examinable, it can be established from the usual rules for the
triple vector product in cartesian coordinates but without shifting the position of the partial
derivatives:
 
X ∂ X ∂ 1 2
[u × (∇ × u)]i = uj uj − uj ui = ∇u − (u · ∇)u
j
∂xi j
∂xj 2 i

In irrotational flows (∇ × u = 0), this enables us to simplify the advective derivative.

Appendix B

Viscous Stresses

We know that, in the special case u = (v(y), 0, 0) in an isotropic Newtonian fluid like water,
there is a stress acting in the x−direction across surfaces perpendicular to ŷ:
∂u
τ =µ .
∂y
What happens in the general case when u = (u1 (x1 , x2 , x3 ), u2 (x1 , x2 , x3 ), u3 (x1 , x2 , x3 ))?
We need a systematic notation. We write x = (x1 , x2 , x3 ) and
∂u1
τ21 = µ ≡ µ ζ12 . (B.1)
∂x2
The stress here, τ , acts across surfaces perpendicular to the ŷ or x2 direction, hence the
subscript 2, and acts in the x̂ or x1 direction, hence the second subscript 1.
We should consider all possible τij and ζij . However, not all combinations of the ζij give rise
to viscous effects. In particular, rigid rotations keep all lengths the same and hence give rise
to no viscous effects. Consider (assume the summation convention that repeated indices are
summed over)
D(δl)2 D(δxi δxi ) D(δxi )
= = 2δxi = 2δxi δui
Dt Dt Dt 
∂ui ∂ui ∂uj
= 2δxi δxj = δxi δxj + .
∂xj ∂xj ∂xi
If
∂ui ∂uj D(δl)2
=− then =0
∂xj ∂xi Dt
and all lengths in the fluid are constant. We therefore need only consider the symmetric part
of ζij , which we call the rate of strain tensor, Eij :
ζij + ζji
Eij = .
2
(The antisymmetric part
ζij − ζji
Wij = .
2
is related to the vorticity and is an important quantity in many flows.)
One may also show that τij = τji . Consider the element of fluid shown.

23
 OneNote 23/11/2022, 15)20

Stresses
APPENDIX B.Wednesday,
VISCOUS 23 NovemberSTRESSES
2022 10:55 24

Setting the net torque on the element equal to the rate of change of angular momentum gives
I ∆V →0
(τxz − τzx ) ∆V = I ω̇ ⇒ τxz − τzx = ω̇ → 0.
∆V

We assume a linear relation between stress and rate of strain (again assuming the summation
convention)
τij = Λijkl Ekl .
In an isotropic fluid, the tensor Λijkl must be isotropic (it cannot depend on the choice of a
coordinate system). It can be shown that this means that
Λijkl = λδij δkl + ξδik δjl + χδil δjk
⇒ τij = λδij Ekk + (ξ + χ)Eij .
Comparing this with B.1 gives
∂u1 ∂u2
τ21 = µ = 2µE12 (remember =0 for that case),
∂x2 ∂x1
and we can identify
(ξ + χ) = 2µ.

The quantity Ekk = ∂uk /∂xk = ∇ · u, so that

 
∂ui ∂uj
τij = µδij Ekk + µ + + λ∇ · u.
∂xj ∂xi
The quantity λ is called the bulk viscosity. We will concentrate on incompressible fluids
(∇ · u = 0) and will not normally retain this term.
The net force, f , on an infinitesimal element of fluid with volume, ∆V , has components
https://ukc-onenote.officeapps.live.com/o/onenoteframe.aspx?ui=en…rectionreason=Force_SingleStepBoot&rct=Medium&ctp=LeastProtected Page 1 of 2
 2
∂ 2 ui

∂τik ∂ ui
fk = ∆V = µ ∆V +
∂xi ∂xi ∂xk ∂x2i
∂ 2 ui
   
∂ ∂ui
= µ ∆V +
∂xk ∂xi ∂x2i
Recognising that ∂ui /∂xi = ∇ · u = 0 for the cases we are interested in leaves us with

fk = µ ∆V ∇2 uk (B.2)