Chapter 2: Transport phenomena in fluid dynamics

Similarity and Transport Phenomena in Fluid Dynamics

Christophe Ancey
 Chapter 2: Transport phenomena in fluid dynamics

• Transport phenomena
• Advection
• Diffusion
• Wave
• Boundaries
• Classification of PDEs

Similarity and Transport Phenomena in Fluid Dynamics 2

o
 A short detour: conservation equations
n

Equations of conservation describe how the spatial and temporal variations of a

function f are related. Let us consider a control volume V (and its bounding
surface S), then
∂f
Z Z
dV + f u · ndS = Φ
V ∂t S
n normal oriented, n local velocity of the control surface, Φ rate of variation of f
(if ϕ = 0, then f is conserved).
Similarity and Transport Phenomena in Fluid Dynamics 3
o
 A short detour: conservation equations

When f is continuous overZ V , then the Green-Ostrogradski

Z theorem tell us that
f u · ndS = ∇ · (f u)dV
S V
If this holds true for any volume V , then we can derive the local form of
conservation
∂f
+ ∇ · (f u) = ϕ
R ∂t
where ϕ is such that Φ = V ϕdV .

Similarity and Transport Phenomena in Fluid Dynamics 4

o
 Characteristic form of conservation equations

A reminder of a simple The ODE

du
geometrical interpretation a(x, u) = b(x, u),
dx
of the derivative. has an implicit solution ψ(x, u(x)) = c. Taking the x derivative
of this equation gives us
dψ ∂ψ du ∂ψ ′ ψx
= + =0⇒u =−
dx ∂x dx ∂u ψu
In other words, if we introduce the tangent vector t = (a, b),
′
then the ODE is equivalent to ∇ψ · t = 0 Point O lies on the
curve ψ(x, u(x)) = c if (dx, dy) is collinear to t: λt = (dx, dy),
and thus
dx dy
λ= =
a b
Similarity and Transport Phenomena in Fluid Dynamics 5
o
 Characteristic form of conservation equations

Generalization to partial differential equations

Quasi-linear first-order partial differential equations are linear in
the differential terms. They can be put in the form:
P (x, y, u)∂xu + Q(x, y, u)∂y u = R(x, y, u).

Similarity and Transport Phenomena in Fluid Dynamics 6

o
 Characteristic form of conservation equations

The implicit solution can be written as ψ(x, y, u(x, y)) = c (with c a constant).
ψ is a first integral of the vector field (P, Q, R). We have:
∂xψ(x, y, u(x, y)) = 0 = ψx + ψuux,
∂y ψ(x, y, u(x, y)) = 0 = ψy + ψuuy .
We deduce that ux = −ψx/ψu et uy = −ψy /ψu, and thus:
P ψx + Qψy + Rψu = 0,
or in a vector form
(P, Q, R) · ∇ψ = 0.

Similarity and Transport Phenomena in Fluid Dynamics 7

o
 Characteristic form of conservation equations

Geometrical interpretation: At point O, the normal to the solution surface is

normal to the vector field (P, Q, R). If point O (x, y, u) and neighbour point O’
′
(x + dx, y + dy, u + du) belong to the solution surface, then the vector 00
(dx, dy, du) is normal to ∇ψ : ψxdx + ψy dy + ψudu = 0. This means that
(dx, dy, du) is collinear to the tangent t = (P, Q, R). There exists a scalar λ such
that
(dx, dy, du) = λt = λ(P, Q, R)
We then deduce the characteristic form.

Similarity and Transport Phenomena in Fluid Dynamics 8

o
 Characteristic form of conservation equations

The characteristic equations is

dx dy du
= =
P (x, y, u) Q(x, y, u) R(x, y, u)
Each pair of equations defines a curve in the space (x, y, u). These curves define a
two-parameter family (there are 3 equations, so 3 invariants but only 2 are
independent): for example, if p is a first integral of the first pair of equations, an
integral surface of the first pair is given by an equation of the form p(x, y, u) = a,
with a a constant. Similarly for the second pair: q(x, y, u) = b. The functional
relation F (a, b) = 0 defines the integral curve.

Similarity and Transport Phenomena in Fluid Dynamics 9

o
 Exercise 1

1. Solve equation
∂u ∂u 2
x −y =u .
∂x ∂y

Similarity and Transport Phenomena in Fluid Dynamics 10

o
 Advection
f
at time t1 at time t2

u(t2 − t1 )

The simplest convection equation is the following one

∂f ∂f
+ u = 0,
∂t ∂x
Linear advection with no source term.
Similarity and Transport Phenomena in Fluid Dynamics 11
o
 Advection

The characteristic equation associated to this PDE is

dx dt df
= = .
u 1 0
As u is assumed to be constant, this means that the solution of the
characteristic equation is x − ut = const and any function F (x − ut)
whose argument is x − ut is a solution. A feature of this solution is that
the original form F (x) (at t = 0) is conserved in the course of movement:
it is simply translated by ut.

Similarity and Transport Phenomena in Fluid Dynamics 12

o
 Advection: methods of characteristics

Let us consider the solution F (x, t) to the linear advection equation

∂f ∂f
+u =0
∂t ∂x
and a continuous path X(t). The time derivative of F along X is
dF ∂F ∂F dX
(X(t), t) = +
dt ∂t ∂x dt
′
If we select X such that X = u then
dF
(X(t), t) = 0
dt
This means that F is constant along the path X. This shows the
equivalence
∂f ∂f df dx
+u =0⇔ = 0 along x(t) such that =u
∂t ∂x dt dt
Similarity and Transport Phenomena in Fluid Dynamics 13
o
 Methods of characteristics: information propagation

Initial value problem Equivalence of a PDE and a system of ODEs

∂f ∂f df dx
+ u(f, x, t) = s(f, x, t) ⇔ = s on =u
∂t ∂x dt dt
Information is propagated along the characteristic
curve. When s = 0 (homogenous equation), then the
characteristic curve are straight lines of slope u. For
linear advection, u = cst and all characteristic curves
are parallel. On each curve, the value of f is fixed by
the initial condition
f (x, t) = f0(x0) = f0(x − ut)
See chap. 8...
Similarity and Transport Phenomena in Fluid Dynamics 14
o
 Diffusion

An example of one-dimensional diffusion equations is the heat equation

2
∂T ∂T
= α 2,
∂t ∂x
with α thermal diffusion, T (x, t) temperature, x abscissa in the bar
direction. This is a second-order linear equation, which describes heat
diffusion along the bar.

Example: spread of a heat source. Conservation of thermal energy E

imposes Z ∞
E
T (x, t)dx = V = .
−∞ cS
with c heat capacity

Similarity and Transport Phenomena in Fluid Dynamics 15

o
 Diffusion: Dimensional analysis and similarity solution

There are n = 5 variables: T , x, t, α, and V ; the other variables (E, c,

and S) are introduced through V .

The dimensional matrix is the following

T x t α V
2
homogeneous to K m s m /s m· K
power decomposition :
power of m 0 1 0 2 1
power of s 0 0 1 −1 0
power of K 1 0 0 0 1

Similarity and Transport Phenomena in Fluid Dynamics 16

o
 Diffusion: Dimensional analysis and similarity solution

This is a 3 × 5 matrix of rank 3 (the fourth column is obtained by linear

combination of columns 2 and 3, column 5 is the sum of columns 1 and 2). We can
therefore form k = n − r = 2 dimensionless numbers. Let us pose
a b c a′ b′ c′
Π1 = xα t V et Π2 = T α t V .
To get [Π1] = 0, we must have
2 a b c
[m (m /s) s (mK) ] = 0.
This leads to the following system of equations
for m : 0 = 2a + c + 1,
for s : 0 = −a + b,
for K : 0 = c,
Similarity and Transport Phenomena in Fluid Dynamics 17
o
 Diffusion: Dimensional analysis and similarity solution

1 1
The solution is a = −2, b= −2, and c = 0. The first dimensionless group is
x
Π1 = √ .
αt
To get [Π2] = 0, we must have
2 a′ b′ c′
[K (m /s) s (mK) ] = 0,
which leads to the following system of equations
′ ′
for m : 0 = 2a + c ,
′ ′
for s : 0 = −a + b ,
′
for K : 0 = c + 1.

Similarity and Transport Phenomena in Fluid Dynamics 18

o
 Diffusion: Dimensional analysis and similarity solution

′ 1 ′ 1 ′
The solution is a = 2 , and c = −1. The second dimensionless group is
b = 2 ,
√
T αt
Π2 = .
V
Dimensional analysis leads us to pose the solution in the form Π2 = F (Π1). We
substitute T into the PDE, with T defined by
V
T = √ F (ξ),
√ αt
with ξ = x/ αt. We get
∂T 1 V 1 V ′
= − 3/2√ F (ξ) − ξ 3/2√ F (ξ)
∂t 2t α 2 t α
2
∂T V ′ ∂T V ′′
= F (ξ) and = F (ξ),
∂x tα ∂x2 (tα)3/2
Similarity and Transport Phenomena in Fluid Dynamics 19
o
 Diffusion: Dimensional analysis and similarity solution

This leads us to write the heat equation in the form of a second-order ordinary
differential equation
1 1 ′ ′′
− F − ξF = F ,
2 2
which is easy to integrate
1 ′
ξF + F = a0,
2
with a0 a constant of integration. If propagation occurs in both directions, then the
′
solution is even (heat spreads equally in both directions), and for ξ = 0, F = 0
(horizontal tangent), so a0 = 0.

Similarity and Transport Phenomena in Fluid Dynamics 20

o
 Diffusion: Dimensional analysis and similarity solution

We get
′
 
F 1 1 2
= − ξ ⇒ F = a1 exp − ξ
F 2 4
with a1 a constant of integration. Using the conservation of heat and since
R √
F dξ = 1, we deduce a1 = 1/(2 π).

The solution reads  2


V 1x
T = √ exp − .
2 παt 4 αt

Similarity and Transport Phenomena in Fluid Dynamics 21

o
 Diffusion: Dimensional analysis and similarity solution

0.8

f ( x,t ) 0.6

0.4

0.2

0
-10 -5 0 5 10
x

Similarity and Transport Phenomena in Fluid Dynamics 22

o
 Exercise 2

2. Consider the diffusion equation:

2
∂f ∂f
= D 2,
∂t ∂x
with D the diffusion coefficient (which is constant).

Similarity and Transport Phenomena in Fluid Dynamics 23

o
 Exercise 2 (continued)

It is subject to the following initial and boundary conditions

f (x, 0) = 0,
f (0, t) = a for t > 0,
f (x, t) = 0 for x → ∞ and t > 0,
with a a constant.
Solve it using the Laplace transform
Z ∞in t
ˆ
f (x, s) = −st
e f (x, t)dt.
0

Similarity and Transport Phenomena in Fluid Dynamics 24

o
 Advection-diffusion equation

Convection-diffusion is a combination of two phenomena. The linear case is

2
df ∂f ∂f ∂f
= + u = D 2,
dt ∂t ∂x ∂x
where D and u are assumed constant. This equation can be reduced to a linear
diffusion problem by making the following change of variable
ζ = x − ut,
τ = t.

Similarity and Transport Phenomena in Fluid Dynamics 25

o
 Advection-diffusion equation

We get
∂· ∂· ∂ζ ∂· ∂τ
= + ,
∂x ∂ζ ∂x ∂τ ∂x
∂·
= ,
∂ζ
∂· ∂· ∂ζ ∂· ∂τ
= + ,
∂t ∂ζ ∂t ∂τ ∂t
∂· ∂·
= −u + .
∂ζ ∂τ
and so 2
∂f ∂f
= D 2,
∂τ ∂ζ

Similarity and Transport Phenomena in Fluid Dynamics 26

o
 Advection-diffusion equation: Burger’s equation

A special case of convection-diffusion is Burger’s equation

2
∂u ∂u ∂u
+ u = D 2,
∂t ∂x ∂x
which can also be transformed into a diffusion equation using the Cole-Hopf
transformation:
2D ∂ϕ
u=− ,
ϕ ∂x
with ϕ(x, t) an auxiliary function.

Similarity and Transport Phenomena in Fluid Dynamics 27

o
 Advection-diffusion equation: Burger’s equation

We have 2
2

∂u 2D ∂ ϕ 2D ∂ϕ
=− + ,
∂x ϕ ∂x2 ϕ2 ∂x
2
∂u 2D ∂ ϕ 2D ∂ϕ ∂ϕ
=− + 2 ,
∂t ϕ ∂x∂t ϕ ∂x ∂t
2 3
  3 2
∂u 2D ∂ ϕ 4D ∂ϕ 6D ∂ ϕ ∂ϕ
= − − + .
∂x2 ϕ ∂x3 ϕ3 ∂x ϕ2 ∂x2 ∂x

Similarity and Transport Phenomena in Fluid Dynamics 28

o
 Advection-diffusion equation: Burger’s equation

After simplification, we obtain

2 3 2
 
∂ϕ ∂ϕ ∂ϕ ∂ ϕ ∂ϕ ∂ ϕ
−ϕ + 2D ϕ 3 − = 0,
∂t ∂x ∂x∂t ∂x ∂x ∂x2
2
which can be transformed—by dividing it by ϕ , then integrating with respect to x,
and ultimately by multiplying it again by ϕ—into a linear diffusion equation
2
∂ϕ ∂ϕ
= D 2.
∂t ∂x

Similarity and Transport Phenomena in Fluid Dynamics 29

o
 Wave equation

Dynamic waves are solutions to a differential equation such as:

2 2
∂ϕ 2∂ ϕ
= c ,
∂t2 ∂x2
with c the (phase) velocity. This form is not exhaustive. For example, the equation
of gravity waves reads
2
∂ϕ ∂ϕ
2
= −g ,
∂t ∂y
with here ϕ the velocity potential (u(x, y, t) = ∇ϕ) and g gravity acceleration.

Similarity and Transport Phenomena in Fluid Dynamics 30

o
 Wave equation
crest

amplitude A

wavelength λ

trough

Shape: A is the amplitude, k wave number (λ = 2π/k wavelength), ω

angular frequency; we also introduce a frequency f defined as
f = ω/(2π): the number of complete oscillations during a second at a
given position. The period is defined as T = λ/c.

Similarity and Transport Phenomena in Fluid Dynamics 31

o
 Wave equation

Solutions are sought in the form of harmonics (periodic wave)

ϕ(t) = A exp[ı(kx − ωt)] = Re(A) cos(kx − ωt) − Im(A) sin(kx − ωt),
The wave velocity is
c = ω/k.
The dispersion relation ω(k) is here linear (ω(k) = ck), i.e. the wave
crests move at a constant speed regardless of the wavelength. The phase
velocity cp
ω(k)
cp = = c.
k

Similarity and Transport Phenomena in Fluid Dynamics 32

o
 Wave equation

Note: In a physical process where waves result from the superposition of

many harmonic waves of different wavelength, each harmonic component
moves at its own speed, which ultimately leads to a separation or
dispersion of the wave, hence the name dispersion relation for ω(k).

There is a third velocity, called group velocity, which represents the speed
at which the energy associated with the wave propagates:
dω
cg = =c
dk
for linear waves. In general for most physical processes, we have cg ≤ cp.

Similarity and Transport Phenomena in Fluid Dynamics 33

o
 Wave equation

There are two directions of propagation:

• forward wave f = f (x − ct): the wave goes in the x > 0 direction;
• backward wave f = f (x + ct): the wave goes in the x < 0 direction.

The general solution to the wave equation reads

f = a(x − ct) + b(x + ct),
with a and b two functions. This is the d’Alembert solution.

Similarity and Transport Phenomena in Fluid Dynamics 34

o
 Exercise 3

3. Calculate the phase velocity and group velocity of the following

equation:
ut + ux + uxxx = 0.

Similarity and Transport Phenomena in Fluid Dynamics 35

o
 Classification of second-order partial differential equations

The only general classification of partial differential equations concerns linear

equations of second order. These equations are of the following form
auxx + 2buxy + cuyy + dux + euy + f u = g,
where a, b, c, d, e, f , and g are real-valued functions x and y. When g = 0, the
equation is said to be homogeneous. Linear equations are classified depending on
2
the sign of ∆ = b − ac.

Similarity and Transport Phenomena in Fluid Dynamics 36

o
 Classification of second-order partial differential equations

2
If ∆ = b − ac > 0, the PDE is hyperbolic. The wave equation is an example. In
fluid mechanics, transport equations are often hyperbolic. The canonical form is
uxx − uyy + · · · = 0 or, equivalently, uxy + · · · = 0,
where dots represent terms related to u or its first-order derivatives;
2
If ∆ = b − ac < 0, the PDE is elliptic. The Laplace equation is an example.
Equations describing equilibrium of a process are often elliptic. The canonical form
is
uxx + uyy + · · · = 0

Similarity and Transport Phenomena in Fluid Dynamics 37

o
 Classification of second-order partial differential equations

2
If ∆ = b − ac = 0, the PDE is parabolic. The heat equation is an example.
Diffusion equations are often parabolic. The canonical form is
uyy + · · · = 0.
2
Link between the classification and conics: If we substitute uxx with x , ux
2
with x, uyy with y , uy with y, and uxy with xy into the PDE, we obtain the
2
general equation of a conic, which depending on the sign ∆ = b − ac gives a
parabola (∆ = 0), an ellipse (∆ < 0), or a hyperbola (∆ > 0).

Similarity and Transport Phenomena in Fluid Dynamics 38

o
 Classification of second-order partial differential equations
4
The differential terms are linked and vary
according to the constraints intrinsic to each
2 type of curve. For example, for hyperbolic
equations, there are two branches and part of
the x − y plane is not crossed by the curve,
0
which allows for discontinuous jumps from one
branch to another: a hyperbolic equation is
2 able to generate solutions that become
discontinuous, i.e. undergo a shock even if
initially they were continuous.
4
4 2 0 2 4
Similarity and Transport Phenomena in Fluid Dynamics 39
o
 Boundaries

In mechanics, we solve equations with space and time variables. In general, in order
to work out a particular solution u to a partial differential equation, we need
• boundary conditions that specify how u varies along the domain boundaries at any
time;
• the initial conditions that specify how varies u at the initial instant for any point in
the domain.
We must solve what is called a boundary-value problem with initial conditions or,
said differently, initial boundary-value problem. In some cases, we do not need as
much information. For example, for certain hyperbolic equations, one needs only
the initial conditions, whereas elliptic problems require only boundary conditions
(they generally reflect stationary processes).
Similarity and Transport Phenomena in Fluid Dynamics 40
o
 Boundaries

We distinguish:
• Dirichlet boundary conditions: the boundary condition specifies the value u0 that
the function takes at a point or a curve
u(x ; t) = u0(t)
along a curve Γ.
• Neuman boundary conditions: the boundary conditions specify the derivative that
the function takes at a point or a series of points. Physically, this reflects a flux
condition across the domain boundary:
∂u
(n ; t) = ϕ(t)
∂n
along a curve Γ, with n the normal to Γ and ϕ(t) a flux function.
Similarity and Transport Phenomena in Fluid Dynamics 41
o
 Exercise 4

4. Consider the following initial-value problem

∂u ∂u
t − x = 0,
∂x ∂t
with u(x, 0) = f (x) for x > 0. What type is this equation? Solve it after
determining the associate characteristic equation.

Similarity and Transport Phenomena in Fluid Dynamics 42

o
 Exercises 5 and 6

5. The Euler-Darboux equation reads

aux − buy
uxy + = 0.
x−y
Characterise this equation.

6. The Helmholtz equation reads

2 2
∇ u + k u = 0.
Characterise this equation.

Similarity and Transport Phenomena in Fluid Dynamics 43

o
 Exercises 7

7. Let us consider the kinematic wave approximation for a water flow induced by
rainfall of intensity I over a sloping bed (starting from x0)
∂h ∂hu
+ = I.
∂t ∂x
n
where h is the water flow depth and u = ah the mean flow velocity (a and n two
parameters). Initially the soil is dry, and there is no inflow: h(x0, t) = 0. Use the
method of characteristics and solve the equation. Similarity and Transport Phenomena in Fluid Dynamics 44
o
 Exercises 8 and 9

8. The Klein-Gordon equation is a variant of the Schrödinger equation, which

describes how an elementary particle behaves. It reads
2 2
∂u 2∂ u 2
2
− γ 2
+ c u = 0.
∂t ∂x
Characterise this equation. Seek periodic solutions in the form
u(x, t) = a(k) exp(ıkx + λ(k)t) with a the amplitude of the wave and where λ
and k are the modes. Determine the mode λ? Is the solution stable?

9. Find and sketch the regions in the (x, y)-plane where the equation
2
(1 + x)ϕxx + 2xyϕxy + y ϕyy = 0
is elliptic, parabolic, and hyperbolic.

Similarity and Transport Phenomena in Fluid Dynamics 45

o
 Temporary page!
LATEX was unable to guess the total number of pages correctly. As there was some unprocessed data that should have been added to the final page this extra page has been added to receive it.
If you rerun the document (without altering it) this surplus page will go away, because LATEX now knows how many pages to expect for this document.