Chapter 6: Similarity solutions of partial

differential equations

Similarity and Transport Phenomena in Fluid Dynamics

Christophe Ancey
 Chapter 6: Similarity solutions of partial differential equations

• Similarity solutions
• Associated stretching groups
• Asymptotic behaviour
• Case studies

Similarity and Transport Phenomena in Fluid Dynamics 2

o
 Stretching groups

Let us consider a partial differential equation in the form

 
∂c ∂c
f , , . . . = 0.
∂t ∂x
We study the existence and properties of similarity solutions. Not all
solutions to PDEs are similarity solutions, PDEs do not always have similar
solutions, but when they exist, they shed light on the behaviour of more
general solutions. We consider the one-parameter stretching group
0 0 β 0 α
x = λx, t = λ t, and c = λ c
with λ the group parameters, and α and β the family parameters of the
group that label different groups of the family. The values of α and β are
usually fixed by the initial and boundary conditions.
Similarity and Transport Phenomena in Fluid Dynamics 3
o
 Stretching groups: similarity solutions

The solution to the PDE is a surface in the (x, t, c) space. Its general
equation is
F (x, t, c) = 0
The equation is invariant to the stretching group
β α
F (λx, λ t, λ c) = 0
Differentiating it with respect to λ, then setting λ = 1 leads to
xFx + βtFt + αcFc = 0
whose characteristic form is
dx dt dc dF
= = =
x βt αc 0
Similarity and Transport Phenomena in Fluid Dynamics 4
o
 Stretching groups: similarity solutions

Three integrals of this equation are

x c
ζ = 1/β , f = α/β and F
t t
The most general solution is F = F(ζ, f ) where F is an arbitrary
function. Since F = 0, we obtain an explicit solution in the form
α/β x
c = t f (ζ) with ζ = 1/β
t
When this form is substituted into the PDE, an ODE results. This ODE
(called the principal equation) is usually simpler to solve than the original
PDE.

Similarity and Transport Phenomena in Fluid Dynamics 5

o
 Asymptotic behaviour of similarity solutions

If the initial and boundary conditions impose that M α + N β = L (with

M , N , and L constants), then the ODE is invariant to the associated
stretching group
0 0 L/M
ζ = λζ and f = λ f
If L/M < 0, then the principal equation admits an asymptotic solution
L/M
f = Aζ . The value of A is found by substituting this form into the
PDE.
(Note: usually in fluid mechanics, but not always, we are looking for
solutions that represent spreading of a material/quantity, this is why we
are interested in L/M < 0.)

Similarity and Transport Phenomena in Fluid Dynamics 6

o
 Exercises 1 and 2

1. Consider the linear diffusion equation:

2
∂c ∂c
=D 2
∂t ∂x R
with D the diffusivity. Mass conservation implies that R c(x, t)dx = c0.
Show that this PDE is invariant to a stretching group. Solve the problem.
2. Consider the same linear diffusion equation subject to the boundary
conditions c(0, t) = sin(ωt). Does the problem admit a similarity solution?
Similarity and Transport Phenomena in Fluid Dynamics 7
o
 Exercise 3: the dam-break problem

3. Consider the shallow water equations:

∂th + ∂x(uh) = 0,
∂tu + u∂xu + g∂xh = 0,
subject to
for u, −∞ < x < ∞ u(x, 0) = 0
for h, x > 0 h(x, 0) = h0
x<0 h(x, 0) = 0
Show that these PDEs admit a similarity solution and determine this
solution.
Ritter, A., Die Fortpflanzung der Wasserwellen, Zeitschrift des Vereines Deutscher Ingenieure, 36 (33), 947-954, 1892.

Similarity and Transport Phenomena in Fluid Dynamics 8

o
 Exercise 4: unsteady impulsive flows

4. We address the Stokes’ first problem (also called Rayleigh problem).

Consider an incompressible Newtonian fluid bounded by an upper solid
boundary. At t = 0, the plate is suddenly and instantaneously set into
α
motion. Its velocity is U (t) = At with A a constant and α > 0. The
fluid velocity adjusts to this change in the boundary condition. Write the
Navier-Stokes equations and the boundary conditions. Reduce the
equations and determine their type (hyperbolic, parabolic, elliptic). Seek
similarity solutions.
Drazin, P.G., and N. Riley, The Navier-Stokes Equations: A classification of Flows and Exact Solutions, Cambridge University Press, Cambridge, 2006.

Similarity and Transport Phenomena in Fluid Dynamics 9

o
 Exercise 5: round jet

5. We consider a steady turbulent round jet. There is a dominant flow

direction z, and the flow spreads gradually in the r direction. The axial
gradients are assumed small compared with lateral gradients. These
features allow boundary-layer equations to be used in place of the
Reynolds equation (even in the absence of a boundary).
Pope, S.B., Turbulent Flows, 771 pp., Cambridge University Press, Cambridge, 2000.

Similarity and Transport Phenomena in Fluid Dynamics 10

o
 Exercise 5: round jet

We use a simple closure equation for the velocity cross-correlation

0 0 ∂hui
hu v i = −νt
∂z
with νt ∼ 0.028. The momentum balance equation reduces to
 
∂hui ∂hui νt ∂ ∂hui
hui + hvi = r
∂z ∂r r ∂r ∂r
subject to the boundary conditions u = v = 0 along the z-axis and
u = v = 0 in the limit r → ∞. Introduce the stream function ψ
−1 −1
(hui = r ψr and hvi = −r ψz ). Find a similarity solution to the
momentum balance equation.

Similarity and Transport Phenomena in Fluid Dynamics 11

o
 Exercise 6: front propagation in viscoplastic fluids

6. Let us consider a viscoplastic material. In time-dependent simple-shear

flow, the momentum balance equation reduces to
∂u ∂ ∂u
% = (µ(γ̇)γ̇) with γ̇ =
∂t ∂z ∂z
where µ is the bulk viscosity given by Carreau’s model
µ 0 − µ∞
µ(γ̇) = µ∞ + 2 (1−n)/2
(1 + (γ/γc) )
with n the shear-thinning/thickening index, µ∞ viscosity in the high
shear-rate limit, µ0 viscosity in low high shear-rate limit, and γ̇c a constant.
Duffy, B.R., D. Pritchard, and S.K. Wilson, The shear-driven Rayleigh problem for generalised Newtonian fluids, J. Non-Newt. Fluid Mech., 206, 11-17, 2014.

Similarity and Transport Phenomena in Fluid Dynamics 12

o
 Exercise 6: front propagation in viscoplastic fluids

The nonlinear diffusion equation is subject to the boundary condition

τ = µγ̇ = τ0 at z = 0 and u → 0 when z → −∞
Show that the equation is invariant to a similarity group. Under which
conditions, is there a front propagating downward? Hint: Show that
a
u = t f (ξ) with ξ the similarity variable to be determined. Then consider
two possible behaviours for f : (i) a boundary-layer approximation
k +
f ∼ A(ξ − ξf ) when ξ → ξf (ξf front position), (ii) algebraic decay in
p
the far field f ∼ B(−ξ) when ξ → −∞.

Similarity and Transport Phenomena in Fluid Dynamics 13

o
 Exercise 7: propagation of a gravity current

7. Consider the shallow equations for a gravity current

propagating over a horizontal boundary
∂h ∂hu
+ = 0,
∂t ∂x
∂u ∂u 0 ∂h
+u +g = 0.
∂t ∂x ∂x
0
with g = g(% − %a)/% the reduced gravity constant, %a
the ambient fluid’s density.

Rottman, J.W., and J.E. Simpson, Gravity currents produced by instantaneous releases of a heavy fluid in a rectangular channel, J. Fluid Mech., 135, 95-110, 1

Gratton, J., and C. Vigo, Self-similar gravity currents with variable inflow revisited: plane currents, J. Fluid Mech., 258, 77-104, 1994.

Similarity and Transport Phenomena in Fluid Dynamics 14

o
 Exercise 7: propagation of a gravity current

The gravity is supplied from a source located at x = 0, with an inflow rate

α−1
Q(t) = uh|x=0 = αqt ,
with α ≥ 0. The Froude number at the source is fixed
u0
Fr|x=0 = √ 0 F0,
g h0
with F0 a constant. The front takes the form of a blunt nose (shock), with
a Benjamin-like condition that related the flow depth to the front velocity
2 2
ẋf = β ghf ,
Show that the governing equations are invariant to a stretching group.
How to study the phase portrait?

Similarity and Transport Phenomena in Fluid Dynamics 15

o
 Exercise 8: nonlinear diffusion equation

8. Consider the nonlinear diffusion equation

 
∂h(x, t) ∂ n ∂h(x, t)
= h
∂t ∂x ∂x
with the auxiliary equation (mass conservation)
Z xf (t)
M= h(x, t)dx = A,
0
and
h(xf , t) = 0
at the front xf . Study the phase portrait and deduce
the late-time behaviour.
Grundy, R.E., Similarity solutions of the

nonlinear diffusion equation, Quarterly of Applied Mathematics, 79, 259-280, 1979.

Similarity and Transport Phenomena in Fluid Dynamics 16
o
 Exercise 9: heat diffusion in a cylindrical geometry

9. In industry, coaxial cylinder viscometers are used to

determine the rheological properties of fluid samples.
Let us imagine that you need to study the thermal
coupling by applying a constant heat flux Q at the inner
cylinder r = R. What is the resulting temperature at
this surface? To simplify the problem we assume that
the gap is very large (the outer cylinder is placed at
infinity). The governing equation is the heat equation
 
∂T D ∂ ∂T
= r
∂t r ∂r ∂r
with D the diffusivity.
Similarity and Transport Phenomena in Fluid Dynamics 17
o
 Exercise 9: heat diffusion in a cylindrical geometry

The boundary conditions are

∂T
(R, t) = −Q and T (∞, r) = 0
∂r
and the initial condition is
T (r, 0) = 0
• Show that the PDE is invariant to a stretching group.
• What is the problem with the boundary conditions?
• Seek an approximate solution at short times by considering that the heat has not
diffused very far.
• Seek an approximate solution at long times by assuming that the heat has spread
very far from r = R and so we can take R = 0.

Similarity and Transport Phenomena in Fluid Dynamics 18

o
 Homework: problem statement
x

Solve Huppert’s equation, which describes fluid motion

over a horizontal plane in the low Reynolds-number
h0
limit:  
x ∂h ρg ∂ 3 ∂h
− h = 0.
x= ∂t 3µ ∂x ∂x
The solution must also satisfy the mass conservation
Initially, the fluid is contained in a
equation
reservoir.
Z
h(x, t)dx = V0
Huppert, H.E., The propagation

two-dimensional and axisymmetric viscous gravity currents

where V0 is the initial volume V0 = `h0. Calculate the
1/5
Journal of Fluid Mechanics, 121, 43-58, 1982.
front position with time and show that it varies as t .
Solve the equation numerically and compare with the
analytical solution. Similarity and Transport Phenomena in Fluid Dynamics 19
o
 Homework: using pdepe in Matlab

Hint: Show that the PDE is invariant to a stretching group. Solve the
principal equation. Use the Matlab built-in function pdepe to solve the
equation numerically. You can also use an implicit scheme, e.g.
Crank-Nicolson. (Scripts on the website).
volume = 0.1;
temps = [0:0.1:1 2:0.5:20 21:5:40 40:10:100 200:100:1000];
[x,t,h] = newton(0.1,10,2500,temps);
figure
for i = 1:length(t)
hold on
plot (x,h(i,:))
end
hold off

Similarity and Transport Phenomena in Fluid Dynamics 20

o
 Homework: using pdepe in Matlab
function [x,t,h] = newton(hg,xmax,nPoints,t)
m = 0; xmin=0;
x = linspace(xmin,xmax,nPoints);
options = odeset(’InitialStep’,1e-12);
sol =
pdepe(m,@pdex1pde,@pdex1ic,@pdex1bc,x,t,options,hg);
h = sol(:,:,1);
function [c,f,s] = pdex1pde(x,t,h,DhDx,hg)
c = 1; f = hˆ 3 DhDx ; s = 0;
function h0 = pdex1ic(x,hg)
d=0.05;
for i = 1:length(x)
y = x;
if y(i) <= -d/2
h0(i) = 1;
elseif y(i) > -d/2 & y(i) < d/2
h0(i) = ((cos((y(i)-(-d/2))*2*pi/(2*d)))/2+0.5);
else
h0(i) = 0;
end
end
function [pl,ql,pr,qr] = pdex1bc(xl,ul,xr,ur,t,hg)
pl = ul-1; ql = 0;
pr = ur; qr = 0; Similarity and Transport Phenomena in Fluid Dynamics 21
o
 Homework: using a Crank-Nicolson scheme

In implicit schemes, the spatial derivative is discretized at time t = kδ and

t = (k + 1)δ
2
∂h r k+1 k+1 k+1 1−r k k k 2
= (hi+1 + hi−1 − 2h i ) + (h i+1 + h i−1 − 2h i ) + o(δx )
∂x2 δx2 δx2
Taking r = 0.5 gives the Crank-Nicolson (or Adams-Moulton) scheme.
The scheme is unconditionally stable. For nonlinear diffusion terms, we
write  
∂ ∂h r 
k+1 k+1 k+1 k+1 k+1 k+1

f (h) = 2
f i+1/2 (h i+1 − h i ) − f i−1/2 (hi − hi−1 ) +
∂x ∂x δx
1−r k 
k k k k k

f i+1/2 (hi+1 − h i ) − f i−1/2 (h i − h i−1 )
δx2
k+1 k+1 k+1
with fi+1/2 = (fi+1 + fi )/2
Similarity and Transport Phenomena in Fluid Dynamics 22
o