VShankar Stability Intro

Introduction
Linear stability theory
A toy example
Kelvin-Helmholtz instability
Capillary Instability
Parallel shear flows
Introduction to Hydrodynamic Stability

V. Shankar
Department of Chemical Engineering
IIT Kanpur
vshankar@iitk.ac.in
home.iitk.ac.in/~vshankar
SADEAFFP-2014, Department of Mathematics, IIT Kanpur
V. Shankar (ChE, IITK)
Hydrodynamic Stability
SADEAFFP-2014
1 / 63
Introduction
A toy example
Outline
Introduction
A toy example
SADEAFFP-2014
2 / 63
Introduction
A toy example
References
P. G. Drazin, Introduction to Hydrodynamic Stability, Cambridge

(2002).
F. Charru, Hydrodynamic Instabilities, Cambridge (2011).
P. Schmid and D. Henningson, Stability and Transition in Shear

Flows, Springer (2001).
P. G. Drazin and W. H. Reid, Hydrodynamic Stability, Cambridge

(1981).
S. Chandrasekhar, Hydrodynamic and Hydromagnetic Stability,

Oxford (1961).
National Committee for Fluid Mechanics Films, Video on Instability

and Transition, http://web.mit.edu/hml/ncfmf.html
SADEAFFP-2014
3 / 63
Introduction
A toy example
Outline
Introduction
A toy example
SADEAFFP-2014
4 / 63
Introduction
A toy example
Instabilities are everywhere
Water flow from a tap.

Smoke from an incence stick.
Flow between two concentric
cylinders.
A layer of heated liquid.
SADEAFFP-2014
5 / 63
Introduction
A toy example

cylinders.
SADEAFFP-2014
5 / 63
Introduction
A toy example

cylinders.
SADEAFFP-2014
5 / 63
Introduction
A toy example

cylinders.
SADEAFFP-2014
5 / 63
Introduction
A toy example
Hydrodynamic stability: connection with Mathematics
Pioneers: Applied Mathematicians and Mathematical Physicists in

late 1800s and early 1900s.
E.g. von Helmholtz, Lord Kelvin, Lord Rayleigh, Neils Bohr,
A Sommerfeld, W Heisenberg (PhD thesis, 1923), S Chandrasekhar,
V I Arnold...
One of the 7 Millennium Prize problems of Clay Mathematical
Institute.
SADEAFFP-2014
6 / 63
Introduction
A toy example

V I Arnold...
Institute.
SADEAFFP-2014
6 / 63
Introduction
A toy example

V I Arnold...
Institute.
SADEAFFP-2014
6 / 63
Introduction
A toy example
Laminar-turbulent transition
Osborne Reynolds (1883), An experimental investigation of the
circumstances which determine whether the motion of water shall be
direct or sinuous..
Discontinuous transition from laminar to a turbulent flow when

Re VD/ > 2000.
For rectangular channels, transition at Re 1200.
SADEAFFP-2014
7 / 63
Introduction
A toy example
direct or sinuous..
Re VD/ > 2000.

SADEAFFP-2014
7 / 63
Introduction
A toy example

direct or sinuous..
Re VD/ > 2000.
SADEAFFP-2014
7 / 63
Introduction
A toy example
Practical perspective
Newtonian fluids: Navier-Stokes equations have all the information

about fluid flow.
Laminar flows: simple solutions to governing equations.
Examples: plane and pipe Poiseuille flows; plane Couette flow.
Flow in a pipe: laminar flow unstable at Re 2000.
Instability leads to turbulence.
Turbulent flows: high mixing and drag.
Laminar flows: low mixing and drag.
When does a given laminar flow become unstable ?
SADEAFFP-2014
8 / 63
Introduction
A toy example

about fluid flow.
SADEAFFP-2014
8 / 63
Introduction
A toy example

about fluid flow.
SADEAFFP-2014
8 / 63
Introduction
A toy example

about fluid flow.
SADEAFFP-2014
8 / 63
Introduction
A toy example

about fluid flow.
SADEAFFP-2014
8 / 63
Introduction
A toy example

about fluid flow.
SADEAFFP-2014
8 / 63
Introduction
A toy example

about fluid flow.
SADEAFFP-2014
8 / 63
Introduction
A toy example

about fluid flow.
SADEAFFP-2014
8 / 63
Introduction
A toy example
Laminar flows
Navier-Stokes equations
v =0
Re[t v + v v] = p + 2 v
Laminar flows: flows with relatively simple kinematics and are usually
time-independent.
Laminar flow solutions satisfy Navier-Stokes equations at any Re.
Landau & Lifshitz

Yet not every solution of the equations of motion, even if it is exact, can
actually occur in Nature. The flows that occur in Nature must not only
obey the equations of fluid dynamics, but also be stable
Need to probe the stability of laminar flows to external disturbances.

SADEAFFP-2014
9 / 63
Introduction
A toy example
Laminar flows
v =0
Re[t v + v v] = p + 2 v
time-independent.
Landau & Lifshitz


SADEAFFP-2014
9 / 63
Introduction
A toy example
Laminar flows
v =0
Re[t v + v v] = p + 2 v
time-independent.
Landau & Lifshitz


SADEAFFP-2014
9 / 63
Introduction
A toy example
Laminar flows
v =0
Re[t v + v v] = p + 2 v
time-independent.
Landau & Lifshitz


SADEAFFP-2014
9 / 63
Introduction
A toy example
Why do instabilities occur ?
Real flows subjected to disturbances of

various types.
Disturbances distort the existing force
equilibrium.
Thermal convection & Circular Couette
flows.
After instability
New complex laminar states.
Direct transition to turbulence.
SADEAFFP-2014
10 / 63
Introduction
A toy example

various types.
equilibrium.
flows.
After instability
SADEAFFP-2014
10 / 63
Introduction
A toy example

various types.
equilibrium.
flows.
After instability
SADEAFFP-2014
10 / 63
Introduction
A toy example

various types.
equilibrium.
flows.
After instability
SADEAFFP-2014
10 / 63
Introduction
A toy example

various types.
equilibrium.
flows.
After instability
SADEAFFP-2014
10 / 63
Introduction
A toy example
How to analyze and predict instabilities ?
Experiments, computation and theory.

What is the effect of an initial disturbance on
laminar flow ?
Do perturbations grow or decay ?
At what value of Reynolds number ?
Infinitesimal vs. finite disturbances.
Infinitesimal disturbances: unavoidable.
Linear stability theory.
SADEAFFP-2014
11 / 63
Introduction
A toy example

laminar flow ?
SADEAFFP-2014
11 / 63
Introduction
A toy example

laminar flow ?
STABLE

UNSTABLE
SADEAFFP-2014
11 / 63
Introduction
A toy example

laminar flow ?
STABLE

UNSTABLE
SADEAFFP-2014
11 / 63
Introduction
A toy example

laminar flow ?
STABLE

UNSTABLE
SADEAFFP-2014
11 / 63
Introduction
A toy example

laminar flow ?
STABLE

UNSTABLE
SADEAFFP-2014
11 / 63
Introduction
A toy example

laminar flow ?
STABLE

UNSTABLE
SADEAFFP-2014
11 / 63
Introduction
A toy example
Outline
Introduction
A toy example
SADEAFFP-2014
12 / 63
Introduction
A toy example
Specify governing equations
Navier-Stokes equations (Newtonian fluid)

v = 0,
[t v + (v )v] = p + 2 v + g .
Boundary conditions at rigid surface: no-slip and no-penetration.
Find the base state: v x (z), and p(x) (steady, unidirectional).
For complex fluids

Specify appropriate constitutive relation.
SADEAFFP-2014
13 / 63
Introduction
A toy example
Specify governing equations
Navier-Stokes equations (Newtonian fluid)

v = 0,
[t v + (v )v] = p + 2 v + g .
Boundary conditions at rigid surface: no-slip and no-penetration.
Find the base state: v x (z), and p(x) (steady, unidirectional).
For complex fluids

Specify appropriate constitutive relation.
SADEAFFP-2014
13 / 63
Introduction
A toy example
Add a small perturbation
v(x, t) = vx (x) + v (x, t) ,
p(x, t) = p(x) + p (x, t) ,
|| 1.
Key questions
For a given control parameter , does v grow or decay with
time ?
What is the critical for instability ?
What is the spatial structure at the critical value ?
SADEAFFP-2014
14 / 63
Introduction
A toy example
Add a small perturbation
v(x, t) = vx (x) + v (x, t) ,
p(x, t) = p(x) + p (x, t) ,
|| 1.
Key questions
For a given control parameter , does v grow or decay with
time ?
What is the critical for instability ?
What is the spatial structure at the critical value ?
SADEAFFP-2014
14 / 63
Introduction
A toy example
Linearize about laminar state
Nonlinear term in Navier-Stokes:

v v = v v + (v v + v v ) + 2 v v
v v is the trivial laminar-flow contribution.
At O(), terms linear in the perturbations Include.
At O( 2 ), terms non-linear in the perturbations Neglected for
small perturbations.
Hence, linear stability.
Can predict only the onset of instability.
SADEAFFP-2014
15 / 63
Introduction
A toy example

v v = v v + (v v + v v ) + 2 v v
SADEAFFP-2014
15 / 63
Introduction
A toy example

v v = v v + (v v + v v ) + 2 v v
SADEAFFP-2014
15 / 63
Introduction
A toy example

v v = v v + (v v + v v ) + 2 v v
SADEAFFP-2014
15 / 63
Introduction
A toy example

v v = v v + (v v + v v ) + 2 v v
SADEAFFP-2014
15 / 63
Introduction
A toy example
Linearized PDEs: Fourier expansion

Collect terms of O():
v = 0
[t v + v v + v v] = p + 2 v
BCs: v = 0 at rigid boundaries; Initial condition v (x, t = 0).
y
x MEAN FLOW :
(Vx (y), 0, 0 )
Fourier expand the disturbances

A(x, t) =
dkx
dkz
An (k, t)Fn (y ) exp[i(kx x + kz z)]
n=1
SADEAFFP-2014
16 / 63
Introduction
A toy example
Linearized PDEs: Fourier expansion

Collect terms of O():
v = 0
[t v + v v + v v] = p + 2 v
BCs: v = 0 at rigid boundaries; Initial condition v (x, t = 0).
y
x MEAN FLOW :
(Vx (y), 0, 0 )
Fourier expand the disturbances

=
A(x, t) =
+
Z
dkx
+ ...
dkz
An (k, t)Fn (y ) exp[i(kx x + kz z)]
n=1
SADEAFFP-2014
16 / 63
Introduction
A toy example
Normal modes: Eigenvalue problem
Linearity Study the dynamics of all Fourier modes individually.

Time dependence An (k, t) = An (k) exp[sn t]
Complex growth rate s = sr + isi
sr > 0 Instability; sr < 0 Stability; sr = 0 Neutral stability.
If any Fourier mode grows with time unstable, exponential growth
as t .
If all Fourier modes decay stable as t .
Need to solve coupled ODEs (for F (y )) with an eigenvalue s for
various values of control parameter .
SADEAFFP-2014
17 / 63
Introduction
A toy example

as t .
SADEAFFP-2014
17 / 63
Introduction
A toy example

as t .
SADEAFFP-2014
17 / 63
Introduction
A toy example

as t .
SADEAFFP-2014
17 / 63
Introduction
A toy example

as t .
SADEAFFP-2014
17 / 63
Introduction
A toy example

as t .
SADEAFFP-2014
17 / 63
Introduction
A toy example

as t .
SADEAFFP-2014
17 / 63
Introduction
A toy example
Outline
Introduction
A toy example
SADEAFFP-2014
18 / 63
Introduction
A toy example
A simple toy example

Governing equation: t f = f f 2 + 1 y2 f
Boundary conditions: f (y = 0) = f (y = 1) = 0.
Base state: f = 0
Add perturbation: f (y , t) = f + f (y , t)
Linearize:

t (f + f ) = f f2 + 1 y2 f + f 2ff + 1 y2 f + O( 2 )
At O(): t f = f 2ff + 1 y2 f
+ 1 y2 f , f (y
modes: f (y , t) =
Normal
sF = F +
1 2
dy F
= 0) = f (y = 1) = 0
F (y ) exp[st]
F (y = 0) = F (y = 1) = 0
SADEAFFP-2014
19 / 63
Introduction
A toy example

Base state: f = 0
Linearize:

t (f + f ) = f f2 + 1 y2 f + f 2ff + 1 y2 f + O( 2 )
At O(): t f = f 2ff + 1 y2 f
+ 1 y2 f , f (y
modes: f (y , t) =
Normal
sF = F +
1 2
dy F
= 0) = f (y = 1) = 0
F (y ) exp[st]
F (y = 0) = F (y = 1) = 0
SADEAFFP-2014
19 / 63
Introduction
A toy example

Base state: f = 0
Linearize:

t (f + f ) = f f2 + 1 y2 f + f 2ff + 1 y2 f + O( 2 )
At O(): t f = f 2ff + 1 y2 f
+ 1 y2 f , f (y
modes: f (y , t) =
Normal
sF = F +
1 2
dy F
= 0) = f (y = 1) = 0
F (y ) exp[st]
F (y = 0) = F (y = 1) = 0
SADEAFFP-2014
19 / 63
Introduction
A toy example

Base state: f = 0
Linearize:

t (f + f ) = f f2 + 1 y2 f + f 2ff + 1 y2 f + O( 2 )
At O(): t f = f 2ff + 1 y2 f
+ 1 y2 f , f (y
modes: f (y , t) =
Normal
sF = F +
1 2
dy F
= 0) = f (y = 1) = 0
F (y ) exp[st]
F (y = 0) = F (y = 1) = 0
SADEAFFP-2014
19 / 63
Introduction
A toy example

Base state: f = 0
Linearize:

t (f + f ) = f f2 + 1 y2 f + f 2ff + 1 y2 f + O( 2 )
At O(): t f = f 2ff + 1 y2 f
+ 1 y2 f , f (y
modes: f (y , t) =
Normal
sF = F +
1 2
dy F
= 0) = f (y = 1) = 0
F (y ) exp[st]
F (y = 0) = F (y = 1) = 0
SADEAFFP-2014
19 / 63
Introduction
A toy example

Base state: f = 0
Linearize:

t (f + f ) = f f2 + 1 y2 f + f 2ff + 1 y2 f + O( 2 )
At O(): t f = f 2ff + 1 y2 f
+ 1 y2 f , f (y
modes: f (y , t) =
Normal
sF = F +
1 2
dy F
= 0) = f (y = 1) = 0
F (y ) exp[st]
F (y = 0) = F (y = 1) = 0
SADEAFFP-2014
19 / 63
Introduction
A toy example

Base state: f = 0
Linearize:

t (f + f ) = f f2 + 1 y2 f + f 2ff + 1 y2 f + O( 2 )
At O(): t f = f 2ff + 1 y2 f
+ 1 y2 f , f (y
modes: f (y , t) =
Normal
sF = F +
1 2
dy F
= 0) = f (y = 1) = 0
F (y ) exp[st]
F (y = 0) = F (y = 1) = 0
SADEAFFP-2014
19 / 63
Introduction
A toy example

Base state: f = 0
Linearize:

t (f + f ) = f f2 + 1 y2 f + f 2ff + 1 y2 f + O( 2 )
At O(): t f = f 2ff + 1 y2 f
+ 1 y2 f , f (y
modes: f (y , t) =
Normal
sF = F +
1 2
dy F
= 0) = f (y = 1) = 0
F (y ) exp[st]
F (y = 0) = F (y = 1) = 0
SADEAFFP-2014
19 / 63
Introduction
A toy example

Base state: f = 0
Linearize:

t (f + f ) = f f2 + 1 y2 f + f 2ff + 1 y2 f + O( 2 )
At O(): t f = f 2ff + 1 y2 f
+ 1 y2 f , f (y
modes: f (y , t) =
Normal
sF = F +
1 2
dy F
= 0) = f (y = 1) = 0
F (y ) exp[st]
F (y = 0) = F (y = 1) = 0
SADEAFFP-2014
19 / 63
Introduction
A toy example
Solution of toy problem
F (y ) = c sin[ny ] n integer.
sn = 1
n2 2
Stable s < 0 for < n2 2

Neutral s = 0 for = n2 2
Unstable s > 0 for > n2 2
Most unstable mode:
f (y , t) = h
c sin[y ] exp 1
i
t
SADEAFFP-2014
20 / 63
Introduction
A toy example
sn = 1
n2 2

Most unstable mode:
f (y , t) = h
c sin[y ] exp 1
i
t
SADEAFFP-2014
20 / 63
Introduction
A toy example
sn = 1
n2 2

Most unstable mode:
f (y , t) = h
c sin[y ] exp 1
i
t
SADEAFFP-2014
20 / 63
Introduction
A toy example
sn = 1
n2 2

Most unstable mode:
f (y , t) = h
c sin[y ] exp 1
i
t
SADEAFFP-2014
20 / 63
Introduction
A toy example
Outline
Introduction
A toy example
SADEAFFP-2014
21 / 63
Introduction
A toy example
Kelvin-Helmholtz instability - I
Two incompressible, inviscid fluids in horizontal parallel infinte
streams of different densities and velocities, one stream above the
other.
U(z) = U2 i, (z) = 2 , P(z) = p0 g 2 z for z > 0.
U(z) = U1 i, (z) = 1 , P(z) = p0 g 1 z for z < 0.
Irrotational perturbations: restrictive, but enough for a proof of
instability.
Does not prove stability as the analysis gives no information about
rotational disturbances.
Perturbed interface z = (x, y , t).
u = where = 1 for z > , and = 2 for z < .
SADEAFFP-2014
22 / 63
Introduction
A toy example
Kelvin-Helmholtz instability - II
u = 0, u = 0 u = .
2 1 = 0 for z > and 2 2 = 0 for z < .
BC: = U as z = .
Kinematic condition at moving interface z = , k = 1, 2:
k
k
D
k
Dt = uz = t + x x + y y
Normal
stress continuity (Bernoulli
theorem)
at z = :
h
h
i
i
1
2
1
1
2
gz = 2 C2 2 (2 )2
gz
1 C1 2 (1 )
t
t
The base flowalso satisfies thiscondition (at z = 0):
1 C1 21 U12 = 2 C2 21 U22
SADEAFFP-2014
23 / 63
Introduction
A toy example
Kelvin-Helmholtz instability - Linearization

2 = U2 x + 2 for z > , 1 = U1 x + 1 for z < . Neglect products
of small perturbations 1 , 2 and .

,
1 and g U12 , U22 .
How small ?
x y
Taylor-expand z = about z = 0:
k |z= = k |z=0 + k |z=0 +

z
2 2 = 0, 2 1 = 0
BC: k 0 for z for k = 1, 2
k
=
+ Uk
at z = 0 for k = 1, 2.
z
t
x

2 2
1 1
1 U1
+
+ g = 2 U2
+
+ g at z = 0.
x
t
x
t
SADEAFFP-2014
24 / 63
Introduction
A toy example
Kelvin-Helmholtz instability - Normal Modes
1 , 2 ) exp[i(kx + ly ) + st]
(, 1 , 2 ) = (,
h 2
i
2 k ddz2k (k 2 + l 2 )k = 0
+ B2 exp[kz],
k 2 = (k 2 + l 2 ).
2 = A2 exp[kz]
BC at y , uz2 = 0 B2 = 0.
and 1 = A1 exp[kz].
So: 2 = A2 exp[k]z
Using the kinematic condition at interface:
k,
k.
A2 = (s + ikU2 )/
A1 = (s + ikU1 )/
Using the normal stress condition at interface:
+ (s + ikU1 )2 ] = 2 [kg
(s + ikU2 )2 ]
1 [kg
SADEAFFP-2014
25 / 63
Introduction
A toy example
Kelvin-Helmholtz instability - Growth Rate
2 U2
s = ik 1 U11 +
+2
k 2 1 2 (U1 U2 )2
(1 +2 )2
i
(1 2 ) 1/2
kg
1 +2
Both roots are neutrally stable if

(2 2 ) k 2 1 2 (U1 U2 )2 . When the equality holds, marginal
kg
2
1
stability.
One root is unstable (with Re[s] > 0) if
(2 2 ) < k 2 1 2 (U1 U2 )2 .
kg
1
2
SADEAFFP-2014
26 / 63
Introduction
A toy example
Surface Gravity Waves
2 = 0 and U1 = 0, U2 = 0. Surface gravity waves on deep water.

Stable with phase velocity:
1/2
c = is/k = (g /k)
Oscillatory, stable normal modes.
Waves a special case of hydrodynamic stability.
SADEAFFP-2014
27 / 63
Introduction
A toy example
Internal Gravity Waves

No basic flow: U1 = 0, U2 = 0.
(2 1 )/(1 + 2 )2 ]1/2
s = [kg
Instability if 1 < 2 (heavey fluid rests above light fluid).
If 1 > 2 , there is stability, and there are waves with phase velocity:
1 + 2 )]1/2 .
c = [g (1 2 )/k(
The eigenfunctions decay exponentially away from the interface.
Motion confined to the vicinity of the interface.
Observed between layers of fresh and salt water that occur in
estuaries.
Rayleigh-Taylor instability when the whole system has an upward
acceleration f . Same result with g = f + g . If 2 > 1 , then
instability occurs only of g < 0.
SADEAFFP-2014
28 / 63
Introduction
A toy example
Instability Due to Shear

No effect of buoyancy: 1 = 2 , but U1 6= U2 .
s = 12 ik(U1 + U2 ) 12 k(U1 U2 )
Flow always unstable if U1 6= U2 . Waves of all wavelengths are
unstable.
SADEAFFP-2014
29 / 63
Introduction
A toy example
Outline
Introduction
A toy example
SADEAFFP-2014
30 / 63
Introduction
A toy example
Capillary Instability of a Jet

A cylindrical jet of liquid moving with uniform velocity in air (e.g.
water jet from a slightly-open tap).
Surface tension at the liquid-air interface.
Assume density of outside fluid is zero, and inviscid dynamics for the
liquid.

u=0
u + uu = p
t
Pressure inside the jet: P = P + n at r = (x, , t)
Perturbed unit normal to the jet:
,1, )
(
r #
n = " x

1/2
2
+1+ 2
x
r
SADEAFFP-2014
31 / 63
Introduction
A toy example

Kinematic condition: ur =
D
Dt
at r = .
Base flow: U = 0, P = p + /a for 0 r a, as n = 1/r when

n = er .
Disturbances: u = U + u , p = P + p , and = a +
u = p and
u = 0.
t
2
2
1
n = 1r
2
r 2
x 2

2
Normal stress BC at r = a: p = a2 +
+ a2 2
x 2
2
2
2
2 p = 0, where 2 = 2 + 2 + 1r + r12 2 .
r
x
r
exp[st + i(kx + n)]

Normal modes: (u , p , ) = (
u(r ), p(r ), )
SADEAFFP-2014
32 / 63
Introduction
A toy example

d2 p
dr 2
p
1 d
r dr

k2 +
n2
r2
p = 0
Linearly independent solutions: modified Bessel functions In (kr ),

Kn (kr ); take n 0 without loss of generality.
Physically allowed solution: p(r ) = AIn (kr ).
= A(s)1 (ikIn (kr ), kIn (kr ), inr 1 In (kr ))
u
2,
Linearized BCs: AIn () = (1 2 n2 )/a
1
A(as) In () = s where = ak.

Eigenvalue relation:
s2 =
In ()
(1
a3 In ()
2 n 2 )
In ()/In () > 0 for all 6= 0. So s 2 < 0 if n 6= 0.

s 2 > 0 for 1 < < 1 if n = 0.
SADEAFFP-2014
33 / 63
Introduction
A toy example
Dispersion relation: theory and experiments
Jet stable to all nonaxisymmetric disturbances (n 6= 0).

Jet unstable to axisymmetric modes with wavelengths
= 2/k > 2a.
If km is the wavenumber at which s is maximum, km = 0.7/a.
Jets of all radii are unstable. No critical parameter that marks the
domain of stability.
In experiments the liquid jet will break up with wavelength about
2/km 9a
SADEAFFP-2014
34 / 63
Introduction
A toy example
Physical interpretation
Displacement of the jet radius: Rnew = R + cos(kz), k = 2/.
Surface area of the perturbed jet:
A=
V =
2Rnew ds
0
2
Rnew
dz
#1/2

dRnew 2
+1
ds =
dz
dz
h
2 i
dz
For small , ds = 1 + 21 dRdznew
"
1
A = 2R + R2 k 2
2
SADEAFFP-2014
35 / 63
Introduction
A toy example
V =
dz [R + cos(kz)]2
1
V
= R 2 + 2
2
i
h
2 1/2
Require V / = R02 , so R = R0 1 21 R 2
To O(2 ), R = R0
1 2
4 R02
Then, change in surface area of the jet (per unit wavelength) due to the
displacement Rnew = R + cos(kz) is then
1 2
[(2R0 )2 2 )]
2 R02
SADEAFFP-2014
36 / 63
Introduction
A toy example
In terms of , the change in surface area is

1 2
[(2R0 )2 2 ]
2 R02 2
For > 2R0 , the surface area decreases.
SADEAFFP-2014
37 / 63
Introduction
A toy example
Linear stability analysis: success stories

Rayleigh-Benard thermal convection
Critical Rayleigh number Rac = 1708. Experiments: 1705 10.

SADEAFFP-2014
38 / 63
Introduction
A toy example
Linear stability analysis: success stories
Taylor-Couette Centrifugal Instability
SADEAFFP-2014
39 / 63
Introduction
A toy example
Outline
Introduction
A toy example
SADEAFFP-2014
40 / 63
Introduction
A toy example
Stability of Parallel Shear Flows
Parallel Base Flow: U = (U(y ), 0, 0)

Perturbations: u, v , w
u
u
+U
+ vU
t
x
v
v
+U
t
x
w
w
+U
t
x
u v
w
+
+
x
y
z
p
1 2
+
u
x
Re
p
1 2
=
+
v
y
Re
p
1 2
=
+
w
z
Re
=
= 0
SADEAFFP-2014
41 / 63
Introduction
A toy example
Stability of Parallel Shear Flows
v
2 p = 2U
x
h

i
+ U 2 U 1 4 v = 0
Re
t
x
x
h
i
+ U 1 2 = U v
Re
t
x
z
u
w
Normal vorticity: =
.
z
x
Boundary conditions: v = v = = 0 at solid walls and in the far
field.
SADEAFFP-2014
42 / 63
Introduction
A toy example
Orr-Sommerfeld and Squire equations
v (x, y , z, t) = v (y ) exp[i(x + z t)]

(x, y , z, t) = (y ) exp[i(x + z t)]
k 2 = (2 + 2 )

1
(D2 k 2 )2 v = 0
(i + iU)(D k 2 ) iU
Re

1
2
2
(D k ) = iU v
(i + iU)
Re
BCs: v = D
v = = 0 at solid walls and in free stream.
SADEAFFP-2014
43 / 63
Introduction
A toy example
Orr-Sommerfeld and Squire equations

Temporal problem: , real, complex.
Spatial problem: real, , complex.
We will consider the temporal problem: c = / is the complex
wavespeed (eigenvalue) of the OS equation, and the associated v are
the eigenfunctions.
OS equation is homogeneous, while the Squire equation is forced by
the solutions of the OS equation.
Two classes of eigenmodes: OS modes and Squire modes.
OS modes: Find vn and n by solving OS equation and then find np
by solving the inhomogeneous Squire equation.
Squire modes: v = 0, m m
SADEAFFP-2014
44 / 63
Introduction
A toy example
Squires transformation and Squires theorem
(U c)(D k 2 )
v U v
1
2
iRe (D
k 2 )2 v = 0
OS equation with = 0 (no variation in the z direction):

2 )2 v
2 )
=0
(U c)(D 2D
v U v i2D1Re2D (D2 2D
Comparing: the two equations will have
p identical solutions if the
following relations hold: 2D = k = 2 + 2
2D Re2D = Re
Re2D = Re k < Re
To each 3D OS mode, there is a corresponding 2D OS mode at a
lower Reynolds number.
SADEAFFP-2014
45 / 63
Introduction
A toy example
Squires transformation and Squires theorem

Damped Squire modes: The solutions to the Squire equation are
always damped with ci < 0 for all , , and Re.
To prove, multiply Squire equation by and integrate from y = 1
to y = 1 (the fluid domain), and take imaginary part.
ci
1
dy |
| =
Re
1
2
dy |D|2 + k 2 |
|2 < 0
Squires theorem
Given ReL as the critical Reynolds number for the onset of linear instability
for a given , , the Reynolds number Rec below which no exponential
instabilities exisit for any wavenumber satisfies
Rec min, ReL (, ) = min ReL (, 0)
SADEAFFP-2014
46 / 63
Introduction
A toy example
Inviscid Analysis
Rayleigh Equation: Neglect viscous terms in OS equation
(U c)(D2 k 2 )
v U v = 0
with k 2 = 2 + 2 , and BCs: v = 0 at y = 1 at solid boundaries.
We have to forgo the no-slip BC due to the reduced order of the ODE.
Since the coefficients of the Rayleigh equation are real, any complex
eigenvalue will appear in conjugate pairs. If c is an eigenvalue, so is
c .
Regular singular point in the complex y plane when U(y ) = c. The
corresponding real part of this location yc is the critical layer where
U(y ) = cr .
SADEAFFP-2014
47 / 63
Introduction
A toy example
Inviscid Analysis
Singularity is logarithmic, and Frobenius series can be used to find the
solution about yc .
v1 (y ) = (y yc )P1 (y )
U
v2 (y ) = P2 (y ) + c ln(y yc )
Uc
Here, P1 and P2 are analytic.
The second solution is multivalued due to the logarithmic term.
When ci = 0, the critical layer is on the real axis,
ln(y yc ) = ln |y yc | i for y < yc .
The correct sign of the imaginary part cannot be determined within
the inviscid analysis.
Must do a matched asymptotic expansion of the Rayleigh invisicd
solution with the full OS solution about y = yc .
SADEAFFP-2014
48 / 63
Introduction
A toy example
Classical theorems in inviscid stability

Multiply Rayleigh equation by v and integrate from y = 1 to
y = 1. Then integrate by parts:
Z
dy |D
v |2 + k 2 |
v |2 +
dy
1
U
|
v |2 = 0
U c
Take imaginary part:

ci
|
v |2
dy U
1
|
v |2
=0
|U c|2
c|2
Both
and |U
are nonnegative. If ci is positive, then U has
to change sign in order for the integral to be zero.
Raylieghs inflexion point theorem only a necessary condition.
SADEAFFP-2014
49 / 63
Introduction
A toy example

Rayleighs inflexion point theorem
If there exist perturbations with ci > 0, then U (y ) must vanish for some
y [1, 1] for instability.
Z
dy |D
v | + k |
v| +
1
Take real part:

Z
dy
1
U
|
v |2 = 0
U c
Z 1
U (U cr )
dy
dy |D
v |2 + k 2 |
v |2
=
2
|U
c|
1
1
Then add the following expression to the left side of above equation:
1
|
v |2
=0
|U c|2
1
The above expression is zero
due to inflexion point theorem.
SADEAFFP-2014
(cr Us )
dy U
50 / 63
Introduction
A toy example

Z
U (U Us) 2
|
v| =
dy
|U c|2
1
dy |D
v |2 + k 2 |
v |2
For the integral to be negative, U (U Us ) < 0 in the flow field.
Fjortofts Criterion
Given a monotonic mean velocity profile U(y ), a necessary condition for
instability is that U (U Us ) < 0, with Us = U(ys ) as the mean velocity
at the inflexion point, i.e., U (ys ) = 0.
SADEAFFP-2014
51 / 63
Introduction
A toy example

Howards semicircle theorem
The unstable eigenvalues of the Rayleigh equation satisfy

1
cr (Umax + Umin )
2
2
ci2
(Umax Umin )
2
2
For ci 0+ , Umin < cr < Umax for marginally stable modes.

For plane-Poiseuille flow and many internal channel flows, U does
not vanish in the domain of the flow, and so these flows are stable in
the inviscid limit to 2D infinitesimal perturbations.
Unbounded jets and free shear layers are unstable in the inviscid limit.
SADEAFFP-2014
52 / 63
Introduction
A toy example
Viscous instability
SADEAFFP-2014
53 / 63
Introduction
A toy example
Numerical Eigenspectrum
SADEAFFP-2014
54 / 63
Introduction
A toy example
Numerical Eigenspectrum
SADEAFFP-2014
54 / 63
Introduction
A toy example
Comparison with experiments

Plane-Poiseuille flow (numerical solution): Rec = 5722.2
Experiments: Rec could be as low as 1000.
Pipe-Poiseuille flow (asymptotic/numerical solution): Rec =
Experiments: Rec 2000.
Plane Couette flow (numerical solution): Rec = .
Can a unavoidable disturbance in an experiment be treated as
infinitesimal ?
Very careful experiments in pipe flow: Re for transition could be 105 .
Still need to explain the usual value of Re 2000.
SADEAFFP-2014
55 / 63
Introduction
A toy example

infinitesimal ?
SADEAFFP-2014
55 / 63
Introduction
A toy example

infinitesimal ?
SADEAFFP-2014
55 / 63
Introduction
A toy example

infinitesimal ?
SADEAFFP-2014
55 / 63
Introduction
A toy example

infinitesimal ?
SADEAFFP-2014
55 / 63
Introduction
A toy example

infinitesimal ?
SADEAFFP-2014
55 / 63
Introduction
A toy example
Nonlinear stability
A sufficiently large disturbance could destabilize or stabilize.

A toy nonlinear ODE:
dx
dt
= ax bx 3
Base-state: x = xB1 = 0 , x = xB2 = +

a/b > 0.
p
p
a/b , xB3 = a/b for
Stability: x = xB1 + x , dx
dt = ax , x (t) = A exp[at] , Unstable for
a > 0 and Stable for a < 0.
Stability of xB2 and xB3 :

s = 2a.
dx
dt
= ax 3bxB2 x , x (t) = A exp[st] ,
xB2 and xB3 stable if a > 0 and are unstable if a < 0.

SADEAFFP-2014
56 / 63
Introduction
A toy example
Nonlinear stability

dx
dt
= ax bx 3

a/b > 0.
p
p
a/b , xB3 = a/b for

s = 2a.
dx
dt

SADEAFFP-2014
56 / 63
Introduction
A toy example
Nonlinear stability

dx
dt
= ax bx 3

a/b > 0.
p
p
a/b , xB3 = a/b for

s = 2a.
dx
dt

SADEAFFP-2014
56 / 63
Introduction
A toy example
Nonlinear stability

dx
dt
= ax bx 3

a/b > 0.
p
p
a/b , xB3 = a/b for

s = 2a.
dx
dt

SADEAFFP-2014
56 / 63
Introduction
A toy example
Nonlinear stability

dx
dt
= ax bx 3

a/b > 0.
p
p
a/b , xB3 = a/b for

s = 2a.
dx
dt

SADEAFFP-2014
56 / 63
Introduction
A toy example
Supercritical and Subcritical Bifurcation
Supercritical bifurcation for b > 0: Solutions xB2 and xB3 exist only
for a > 0; for a < 0, x = 0 is the only solution.
For a > 0 x = 0 becomes unstable and gives rise to two new stable
solutions xB2 or xB3 . (RayleighBenard convection).
Subcritical bifurcation for b < 0: For b < 0, xB2 and xB3 exist only
for a < 0.
For a < 0, xB2 and xB3 are unstable.
The x = 0 solution could become unstable even for a < 0 to a
sufficiently large disturbance (Shear flow instabilities).
SADEAFFP-2014
57 / 63
Introduction
A toy example
Supercritical and Subcritical Bifurcation
Supercritical bifurcation for b > 0: Solutions xB2 and xB3 exist only
for a > 0; for a < 0, x = 0 is the only solution.
For a > 0 x = 0 becomes unstable and gives rise to two new stable
solutions xB2 or xB3 . (RayleighBenard convection).
Subcritical bifurcation for b < 0: For b < 0, xB2 and xB3 exist only
for a < 0.
For a < 0, xB2 and xB3 are unstable.
The x = 0 solution could become unstable even for a < 0 to a
sufficiently large disturbance (Shear flow instabilities).
SADEAFFP-2014
57 / 63
Introduction
A toy example
Weakly nonlinear effects: Landau equation
Linear theory:
1 dA
A dt
= s0 A A(t) = A(0) exp[s0 t]
When s0 > 0 Flow unstable: Only onset predicted.

Exponential growth: Cannot neglect nonlinearities anymore.
Nonlinearities can saturate exponential growth, or further accelerate
exponential growth.
Weakly nonlinear theory:
Landau Equation:
1 dA
A dt
= s0 + s1 A2 ;
s1 : Landau constant effect of nonlinearities.
SADEAFFP-2014
58 / 63
Introduction
A toy example
Supercritical Equilibrium
Landau Constant negative
A
1 dA
A dt
New non-laminar
steady states
> c
< c
= s0 |s1 |A2 A2s = s0 /|s1 |.

|{z}
+ve
A linearly unstable mode saturates to a new steady state.

Rayleigh Bernard convection cells, Taylor vortices . . .
SADEAFFP-2014
59 / 63
Introduction
A toy example
Subcritical Instability
Landau constant positive
A
c
No nearby equilibrium states
1 dA
A dt
Stable to
Small disturbances
Unstable to
finite disturbances
= |s0 | + s1 A2 Instability for < c if A2intial > |s0 |/s1 .

|{z}
ve
A linearly stable mode becomes unstable to finite amplitude

disturbances.
Plane Poiseuille flow in a rigid channel: 65% reduction in critical Re

for vx /V = 0.025; Strongly subcritical.
SADEAFFP-2014
60 / 63
Introduction
A toy example
Subcritical Instability
weakly
subcritical
strongly
A subcritical
1 dA
A dt
Unstable
= |s0 | + s1 A2 Instability for < c if A2intial > |s0 |/s1 .

|{z}
ve
A linearly stable mode becomes unstable to finite amplitude

disturbances.
Plane Poiseuille flow in a rigid channel: 65% reduction in critical Re
for vx /V = 0.025; Strongly subcritical.
SADEAFFP-2014
60 / 63
Introduction
A toy example
Transient or algebraic growth: A simple example

d
1
2
Coupled ODEs: dv
dt = Re v ,
dt = v Re
Solution: Exponential decay of v (t) and (t)
v (t) = v0 exp[t/Re]
(t) = v0 Re(exp[t/Re] exp[2t/Re]) + 0 exp[2t/Re]
For small times, series expand:
Rev0 (exp[t/Re] exp[2t/Re]) = v0 t
3v0 2
t +
Re
Growth t for small times t < O(Re), and exponential decay for
long times.
During this transient growth nonlinearities could become important
and lead to instabilities.
Bypass transition.
SADEAFFP-2014
61 / 63
Introduction
A toy example

d
1
2
Coupled ODEs: dv
dt = Re v ,
dt = v Re
3v0 2
t +
Re
long times.
Bypass transition.
SADEAFFP-2014
61 / 63
Introduction
A toy example

d
1
2
Coupled ODEs: dv
dt = Re v ,
dt = v Re
3v0 2
t +
Re
long times.
Bypass transition.
SADEAFFP-2014
61 / 63
Introduction
A toy example
Non-modal (transient) growth
In rigid tubes, pipe flow always (asymptotically) stable at any Re as

per normal mode analysis.
Plane-Poiseuille flow unstable at Rec = 5772, but experiments show
instability at Re 1200.
Underlying differential operators are non-normal, and eigenfunctions
are non-orthogonal.
Possibility of algebraic or transient growth at early times, which
eventually decay as t .
Schmid and Henningson, 2001
SADEAFFP-2014
62 / 63
Introduction
A toy example
Non-modal (transient) growth

In rigid tubes, pipe flow always (asymptotically) stable at any Re as
per normal mode analysis.
Plane-Poiseuille flow unstable at Rec = 5772, but experiments show
instability at Re 1200.
Underlying differential operators are non-normal, and eigenfunctions
are non-orthogonal.
Possibility of algebraic or transient growth at early times, which
eventually decay as t .
transient
growth
disturbance
energy
asymptotic
decay
time
Schmid and Henningson, 2001

SADEAFFP-2014
62 / 63
Introduction
A toy example
Spatio-temporal analysis
Response to impulse forcing at x = 0, t = 0.
t
x
Convective
G (x, t) =
I (x, ) =
d
L 2
d
2
Absolute
I (x, ) exp(it)
D(k, )1 exp(ix)
Absolute instability: Contour deformation in the L-domain until there

is a pinching of branches in the F -domain.
SADEAFFP-2014
63 / 63

VShankar Stability Intro

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

VShankar Stability Intro

Uploaded by

Copyright:

Available Formats

Introduction

Linear stability theory

Parallel shear flows

Introduction to Hydrodynamic Stability

SADEAFFP-2014, Department of Mathematics, IIT Kanpur

V. Shankar (ChE, IITK)

Linear stability theory

Parallel shear flows

Linear stability theory

Parallel shear flows

V. Shankar (ChE, IITK)

Linear stability theory

Parallel shear flows

P. G. Drazin, Introduction to Hydrodynamic Stability, Cambridge

F. Charru, Hydrodynamic Instabilities, Cambridge (2011).

P. Schmid and D. Henningson, Stability and Transition in Shear

P. G. Drazin and W. H. Reid, Hydrodynamic Stability, Cambridge

S. Chandrasekhar, Hydrodynamic and Hydromagnetic Stability,

National Committee for Fluid Mechanics Films, Video on Instability

V. Shankar (ChE, IITK)

Linear stability theory

Parallel shear flows

Linear stability theory

Parallel shear flows

V. Shankar (ChE, IITK)

Linear stability theory

Parallel shear flows

Instabilities are everywhere

Water flow from a tap.

V. Shankar (ChE, IITK)

Linear stability theory

Parallel shear flows

Instabilities are everywhere

Water flow from a tap.

V. Shankar (ChE, IITK)

Linear stability theory

Parallel shear flows

Instabilities are everywhere

Water flow from a tap.

V. Shankar (ChE, IITK)

Linear stability theory

Parallel shear flows

Instabilities are everywhere

Water flow from a tap.

V. Shankar (ChE, IITK)

Linear stability theory

Parallel shear flows

Hydrodynamic stability: connection with Mathematics

Pioneers: Applied Mathematicians and Mathematical Physicists in

V. Shankar (ChE, IITK)

Linear stability theory

Parallel shear flows

Hydrodynamic stability: connection with Mathematics

Pioneers: Applied Mathematicians and Mathematical Physicists in

V. Shankar (ChE, IITK)

Linear stability theory

Parallel shear flows

Hydrodynamic stability: connection with Mathematics

V. Shankar (ChE, IITK)

Linear stability theory

Parallel shear flows

Discontinuous transition from laminar to a turbulent flow when

Linear stability theory

Parallel shear flows