SLIDE 1
MISG 2018: Instability in Fluids
Yilun Wang, Nolwazi Nkomo, Shina D Oloniiju, Jessica Ihesie, Williams Chukwu, Keegan Anderson, Mojalefa Nchupang, Saul Hurwitz Supervisor: Prof David Mason
January 13, 2018
SLIDE 2 Rayleigh-Taylor Instability
y j x i y = η(x, t)
∂φ1 ∂y ∂φ2 ∂y
HEAVIER FLUID ρ2, p2, φ2 LIGHTER FLUID ρ1, p1, φ1 g
SLIDE 3
Rayleigh-Taylor Instability
The problem proposed is a situation with 2 fluids, one atop the other with different densities. Between them is the interface η(x, t) which is a perturbation across y = 0. Some assumptions are made: The vorticity is 0 (it is irrotational) so ∇ × v = 0 It is incompressible, meaning the volume is constant. This results in ∇ · v = 0
SLIDE 4
Equations of State and Boundary Conditions
∂2φ1 ∂x2 + ∂2φ1 ∂y2 = 0 ∂2φ2 ∂x2 + ∂2φ2 ∂y2 = 0 y = 0 : ∂φ1 ∂y (x, 0, t) = ∂η ∂t (x, t) y = 0 : ∂φ2 ∂y (x, 0, t) = ∂η ∂t (x, t) y = 0 : ρ1[∂φ1 ∂t (x, 0, t)+gη(x, t)] = ρ2[∂φ2 ∂t (x, 0, t)+gη(x, t)]
SLIDE 5 Form of Solution and Dispersion Relation
φ1(x, y, t) = F1(y)exp[i(kx − ωt)] φ2(x, y, t) = F2(y)exp[i(kx − ωt)] ω = ±i
ρ1 + ρ2
SLIDE 6 Solution and Analysis
Re(η) = A1 cos(kx)
g(ρ2 − ρ1)(eβt − e−βt) This is an unstable exponentially growing standing wave if ρ2 > ρ1, and is a stable standing wave if ρ1 > ρ2
SLIDE 7 Rayleigh-Taylor Instability with Surface Tension
y j x i y = η(x, t)
∂φ1 ∂y ∂φ2 ∂y
T ∂2η
∂x2
HEAVIER FLUID ρ2, p2, φ2 LIGHTER FLUID ρ1, p1, φ1 g
SLIDE 8
Instability of Fluids with Interfacial Tension
Net upward force per unit area due to interfacial tension T ∂2y ∂x2 ∂2φ1 ∂x2 + ∂2φ1 ∂y2 = 0 ∂2φ2 ∂x2 + ∂2φ2 ∂y2 = 0 y = 0 : ∂φ1 ∂y (x, 0, t) = ∂η ∂t (x, t) y = 0 : ∂φ2 ∂y (x, 0, t) = ∂η ∂t (x, t) y = 0 : p1(x, 0, t) + T ∂2y ∂x2 = p2(x, 0, t)
SLIDE 9 Form of Solution and Dispersion Relation
The form of solution: η(x, t) = η0exp[i(kx − ωt)] φ1(x, y, t) = F1(y)exp[i(kx − ωt)] φ2(x, y, t) = F2(y)exp[i(kx − ωt)] Dispersion Relation: ω = ±
(ρ2 + ρ1)(−g(ρ2 − ρ1) + Tk2)
SLIDE 10 Solution and Analysis
Stable if k2 > (ρ2 − ρ1)g T λ <
(ρ2 − ρ1)g = 2π
(ρ2 − ρ1)g .
SLIDE 11
Kelvin-Helmholtz Instability
y j x i y = η(x, t) ρ, p2, φ2 ρ, p1, φ1 g v2 v1
SLIDE 12 Kelvin-Helmholtz Instability
V (1)
x
= V1 + ∂φ1 ∂x , V (1)
y
= ∂φ1 ∂y V (2)
x
= V2 + ∂φ2 ∂x , V (2)
y
= ∂φ2 ∂y ∂2φ1 ∂x2 + ∂2φ1 ∂y2 = 0; ∂2φ2 ∂x2 + ∂2φ2 ∂y2 = 0, ∂φ1 ∂y (x, 0, t) = ∂η ∂t + V1 ∂η ∂y . ∂φ2 ∂y (x, 0, t) = ∂η ∂t + V2 ∂η ∂y . V1 ∂φ1 ∂x + ∂φ1 ∂t = V2 ∂φ2 ∂x + ∂φ2 ∂t .
SLIDE 13
Form of Solution and Dispersion Relation
The form of solution: η(x, t) = η0exp[i(kx − ωt)] φ1(x, y, t) = F1(y)exp[i(kx − ωt)] φ2(x, y, t) = F2(y)exp[i(kx − ωt)] Dispersion Relation: ω = k(V2 + V1) ± ik(V1 − V2) 2
SLIDE 14 Solution and Analysis
The perturbation solution is η(x, t) = η0 exp
2(V2 + V1)t
2(V1 − V2)t
η0 exp
2(V2 + V1)t
2(V1 − V2)t
- Re[η(x, t)] = η0 cos
- kx − k
2(V2 + V1)t exp
2(V2 + V 1)t
2(V1 − V 2)t
- This is unstable for V1 < V2 and V2 < V1.
SLIDE 15 Kelvin-Helmholtz and Rayleigh-Taylor Instability with Interfacial Tension
y j x i p1 p2 T ∂2η
∂x2
ρ2, p2 ρ1, p1 v2 v1
SLIDE 16 Kelvin-Helmholtz and Rayleigh-Taylor Instability with Interfacial Tension
V (1)
x
= V1 + ∂φ1 ∂x , V (1)
y
= ∂φ1 ∂y V (2)
x
= V2 + ∂φ2 ∂x , V (2)
y
= ∂φ2 ∂y ∂2φ1 ∂x2 + ∂2φ1 ∂y2 = 0; ∂2φ2 ∂x2 + ∂2φ2 ∂y2 = 0, ∂φ1 ∂y (x, 0, t) = ∂η ∂t + V1 ∂η ∂y . ∂φ2 ∂y (x, 0, t) = ∂η ∂t + V2 ∂η ∂y . y = 0 : p1(x, 0, t) + T ∂2y ∂x2 = p2(x, 0, t)
SLIDE 17 Form of Solution and Dispersion Relation
The form of solution: η(x, t) = η0exp[i(kx − ωt)] φ1(x, y, t) = F1(y)exp[i(kx − ωt)] φ2(x, y, t) = F2(y)exp[i(kx − ωt)] ω k = ρ1V1 + ρ2V2 ± √Q ρ1 + ρ2 where Q = −ρ1ρ2(V1 − V2)2 + (ρ1 + ρ2)(Tk − g/
k(ρ2 − ρ1))
SLIDE 18 Solution and Analysis
It is unstable when (V1 − V2)2 >
(ρ1+ρ2)(Tk− g
k (ρ2−ρ1))
ρ1ρ2
k >
T (ρ2 − ρ1) is the first necessary condition for stability
SLIDE 19 Graphical Analysis for ρ2 > ρ1
(v1 − v2)2 k f (k) f (k) = (ρ1 + ρ2)(Tk − g
k (ρ2 − ρ1))
ρ1ρ2
SLIDE 20 Graphical Analysis for ρ1 > ρ2
(v1 − v2)2 k f (k) f (k) = (ρ1 + ρ2)(Tk + g
k (ρ1 − ρ2))
ρ1ρ2
SLIDE 21
Benard Problem
z = d z = 0 cold hot k j i TL − ∆T TL VISCOUS FLUID AT REST
SLIDE 22 Benard Problem
We wish to study the instability in the fluid owing to heat transfer - that is, when does the heat transfer mechanism switch from conduction to convection Temperature Gradient: dT
dz = − ∆T d
Navier-Stokes Momentum Eqn: ρDv
Dt = −∇p + µ∇2v + ρg
Energy Eqn: DT
Dt = κ∇2T+viscous 2nd order terms
Equation of State: ρ = ˜ ρ(1 − α(T − ˜ T))
SLIDE 23
Benard Problem - Unperturbed State
v0 = 0 T = T0(z) ρ = ρ0(z) p = p0(z) Simplified Constitutive Eqns: dp0 dz = −ρ0g κd2T0 dz2 = 0 ρ0 = ˜ ρ(1 − α(T0 − ˜ T))
SLIDE 24
Benard Problem - Unperturbed State
Solving yields: T0(z) = TL − ∆Tz d ρ0(z) = ˜ ρ(1 − α(TL − ∆Tz d − ˜ T)) p0(z) = C + g ˜ ρz(α(TL − ∆Tz 2d − ˜ T) − 1)
SLIDE 25
Perturbation and Boussinesq Approximation
We make the Boussinesq Approximation: we take ρ to be constant and approximately ˜ ρ unless it gives rise to buoyancy forces in the Navier-Stokes eqn We perturb the state by making it no longer at rest: v(x, y, z, t) = 0 + v1(x, y, z, t) T(x, y, z, t) = T0(z) + T1(x, y, z, t) ρ(x, y, z, t) = ρ0(z) + ρ1(x, y, z, t) p(x, y, z, t) = p0(z) + p1(x, y, z, t)
SLIDE 26 New Constitutive Equations
We still have incompressibility: ∇ · v = ∇ · v1 = 0 ˜ ρ∂v1 ∂t = −∇p1 + µ∇2v1 + ρ1g ∂T1 ∂t − v(z)
1
∆T d = κ∇2T1 ρ1 = −˜ ραT1
SLIDE 27 Form of Solution and Eigenvalue Problem
We want to find v(z)
1
to analyse the stability of the system. We assume the form v(z)
1
= w(z)f (x, y)est We obtain an eigenvalue problem with eigenfunction w(z)
z − a2) − s
ν(D2
z − a2) − s
z − a2) − αa2g dT0
dz
= 0
SLIDE 28 Boundary Conditions and Form of Solution
w(0) = w(d) = 0 as v(z)
1
would be 0 at the boundary, as the boundaries are not moving. d2w dz2 (0) = d2w dz2 (d) = 0 d4w dz4 (0) = d4w dz4 (d) = 0 To satisfy the boundary conditions, we let w(z) = sin(nπz d ); n = 1, 2, 3..
SLIDE 29 Stability for Negative Temperature Gradient
To analyse stability we need to focus on the est factor. Solving the eigenvalue problem for w(z) yields an equation for s. Solving (ν − κ)2a6
∗ + 4αa2g ∆T d
> 0 for a yields that in this case, s is always real However, for stability, we need s < 0. This is true when αgd3∆T νκ < 27π4 4 Where Rayleigh’s number is defined as R = αgd3∆T
νκ
SLIDE 30 Stability for Positive Temperature Gradient
Solving (ν − κ)2a6
∗ − 4αa2g ∆T d
> 0 for a yields that in this case, s is not always real. We require αg∆Td3 (ν − κ)2 > 27π4 16 in order for s to be real, otherwise we get an oscillating velocity, where Mason’s number is defined as M = αg∆Td3
(ν−κ)2 > 27π4 16
If s is real, for stability, we need s < 0. This is true when αgd3∆T νκ > −27π4 4 Which is always true, as the Rayleigh number is always positive
SLIDE 31
Convective Unstable Fluid
z = d z = 0 cold hot TL − ∆T TL