From Ideal Gas Law,
P
V
=
N
k
T
⟹
P
=
ρ
μ
m
p
k
T
PV=NkT \implies P = \frac{\rho}{\mu m_p}kT
PV=NkT⟹P=μmpρkT
Hydrostatic Atmosphere
d
P
d
z
=
−
ρ
g
=
−
μ
m
p
k
T
g
P
⟹
P
=
P
0
e
−
z
/
H
\newcommand{\D}[2]{\frac{d#1}{d#2}} \D{P}{z} = -\rho g = -\frac{\mu m_p}{kT}gP \implies P = P_0e^{-z/H}
dzdP=−ρg=−kTμmpgP⟹P=P0e−z/H
H
H
H is the scale height of atmosphere, approximately 8.4km
Hypsometric equation:
z
=
H
log
P
0
P
(
z
)
z = H\log\frac{P_0}{P(z)}
z=HlogP(z)P0
Velocity Field
The velocity of a fluid element: v ( x , y , z , t ) \boldsymbol{v}(x,y,z,t) v(x,y,z,t)
Notations:
a j b j = ∑ j a j b j Einstein Notation a_jb_j = \sum_ja_jb_j\qquad\text{Einstein Notation} ajbj=j∑ajbjEinstein Notation
∂ j = ∂ ∂ x j ∂ t = ∂ ∂ t (and same for other derivatives) \newcommand{\DP}[2]{\frac{\partial#1}{\partial#2}} \partial_j = \DP{}{x_j}\qquad \partial_t = \DP{}{t} \qquad\text{(and same for other derivatives)} ∂j=∂xj∂∂t=∂t∂(and same for other derivatives)
Convectiveon Differentiation
consider a scalar function
T
(
x
,
y
,
z
,
t
)
T(x,y,z,t)
T(x,y,z,t). For a small fluid element, the infinitesimal change is
δ
T
=
T
(
x
+
v
x
δ
t
,
y
+
v
y
δ
t
,
z
+
v
z
δ
t
,
t
+
δ
t
)
−
T
(
x
,
y
,
z
,
t
)
=
∂
T
∂
t
δ
t
+
∂
T
∂
x
v
x
δ
t
+
∂
T
∂
y
v
y
δ
t
+
∂
T
∂
z
v
z
δ
t
\newcommand{\D}[2]{\frac{d#1}{d#2}}\newcommand{\DP}[2]{\frac{\partial#1}{\partial#2}} \begin{aligned} \delta T &= T(x+v_x\delta t, y+v_y\delta t, z+v_z\delta t, t+\delta t) - T(x,y,z,t) \\ &= \DP{T}{t}\delta t + \DP{T}{x}v_x\delta t + \DP{T}{y}v_y\delta t + \DP{T}{z}v_z\delta t \end{aligned}
δT=T(x+vxδt,y+vyδt,z+vzδt,t+δt)−T(x,y,z,t)=∂t∂Tδt+∂x∂Tvxδt+∂y∂Tvyδt+∂z∂Tvzδt
Therefore, we define convective derivative (or fluid dervivative) as
D
T
D
t
=
δ
T
δ
t
=
∂
t
(
T
)
+
(
v
⋅
∇
)
T
D
D
t
=
∂
t
+
(
v
⋅
∇
)
\newcommand{\D}[2]{\frac{d#1}{d#2}}\newcommand{\FD}[2]{\frac{\mathcal{D}#1}{\mathcal{D}#2}} \FD{T}{t} = \frac{\delta T}{\delta t} = \partial_t(T) + (\boldsymbol{v}\cdot\nabla)T \qquad \FD{}{t} = \partial_t + (\boldsymbol{v}\cdot\nabla)
DtDT=δtδT=∂t(T)+(v⋅∇)TDtD=∂t+(v⋅∇)
The acceleration:
D v D t = ( ∂ t + ( v ⋅ ∇ ) ) v \newcommand{\D}[2]{\frac{d#1}{d#2}}\newcommand{\FD}[2]{\frac{\mathcal{D}#1}{\mathcal{D}#2}} \FD{\boldsymbol{v}}{t} = (\partial_t + (\boldsymbol{v}\cdot\nabla))\boldsymbol{v} DtDv=(∂t+(v⋅∇))v
which is not linear, resulting super complicated consequences.
Continuity
Define the current as
J
=
ρ
v
⟹
∂
t
(
ρ
)
+
∇
⋅
J
=
0
\boldsymbol{J} = \rho\boldsymbol{v} \implies \partial_t(\rho) + \nabla\cdot\boldsymbol{J} = 0
J=ρv⟹∂t(ρ)+∇⋅J=0
by the conservation of mass
For inconpressible fluid,
∇
⋅
v
=
0
⟹
D
t
(
v
)
=
0
\nabla\cdot\boldsymbol{v} = 0 \implies \mathcal{D}_t(\boldsymbol{v}) = 0
∇⋅v=0⟹Dt(v)=0
Euler’s Equation
For an infinitesimal volume
d
V
=
d
x
d
y
d
z
dV = dxdydz
dV=dxdydz
in z direction,
F
=
z
^
(
−
P
(
x
,
y
,
z
+
d
z
)
+
P
(
x
,
y
,
z
)
)
d
x
d
y
\boldsymbol{F} = \hat{z}(-P(x,y,z+dz)+P(x,y,z))dxdy
F=z^(−P(x,y,z+dz)+P(x,y,z))dxdy
F
=
−
(
∂
x
P
x
^
+
∂
y
P
y
^
+
∂
z
P
z
^
)
d
V
=
(
−
∇
P
)
d
V
\boldsymbol{F} = -(\partial_xP\hat{x} + \partial_yP\hat{y} + \partial_zP\hat{z})dV = (-\nabla P)dV
F=−(∂xPx^+∂yPy^+∂zPz^)dV=(−∇P)dV
Apply the Newtonian Mechanics to fluid
a
=
D
t
v
=
F
m
=
−
∇
P
ρ
⟹
∂
t
v
+
(
v
⋅
∇
)
v
=
−
∇
P
ρ
+
g
\boldsymbol{a} = \mathcal{D}_{t}\boldsymbol{v} = \frac{\boldsymbol{F}}{m} = \frac{-\nabla P}{\rho} \implies \partial_t\boldsymbol{v} + (\boldsymbol{v}\cdot\nabla)\boldsymbol{v} = \frac{-\nabla P}{\rho} + \boldsymbol{g}
a=Dtv=mF=ρ−∇P⟹∂tv+(v⋅∇)v=ρ−∇P+g
which is known as Euler’s Equation. This equation has nothing to do with viscosity.
Adiabatic Flow (air)
P
V
γ
=
C
⟹
D
t
(
P
ρ
−
γ
)
=
0
T
γ
P
γ
−
1
=
C
PV^\gamma = C \implies \mathcal{D}_t(P\rho^{-\gamma}) = 0 \qquad \frac{T^\gamma}{P^{\gamma-1}}=C
PVγ=C⟹Dt(Pρ−γ)=0Pγ−1Tγ=C
Streamline
The curve which the velocity is tangent to at every point.
In stationary state, streamlines are fixed in space (which is also known as pathline)
Sink and Source are the point where streamline disappears / appears, respectively.
Consider a streamline tube, if there is no sink or source, by the conservation of mass, we can get the Bernoulli’s Principle by the conservation of total energy (kinetic, potential and internal)
γ
P
(
γ
−
1
)
ρ
+
1
2
v
2
+
g
h
=
constant
\frac{\gamma P}{(\gamma-1)\rho} + \frac{1}{2}v^2 + gh = \text{constant}
(γ−1)ργP+21v2+gh=constant
Vorticity
ω
=
∇
×
v
\boldsymbol{\omega} = \nabla\times\boldsymbol{v}
ω=∇×v
By the identity
(
a
×
b
)
×
c
=
(
c
⋅
a
)
b
−
(
c
⋅
b
)
a
(a\times b)\times c = (c\cdot a)b - (c\cdot b)a
(a×b)×c=(c⋅a)b−(c⋅b)a
ω × v = ( ∇ × v ) × v = ( v ⋅ ∇ ) v − 1 2 ∇ ( v 2 ) \newcommand{\V}[1]{\boldsymbol{#1}} \V{\omega}\times\V{v} = (\nabla\times\V{v})\times\V{v} = (\V{v}\cdot\nabla)\V{v} - \frac {1}{2}\nabla(\V{v}^2) ω×v=(∇×v)×v=(v⋅∇)v−21∇(v2)
Rewrite the Euler’s Equation
∂
t
v
+
ω
×
v
=
−
1
2
∇
v
2
−
1
ρ
∇
P
−
∇
Φ
\newcommand{\V}[1]{\boldsymbol{#1}}\partial_t\V{v} + \V{\omega}\times\V{v} = -\frac{1}{2}\nabla\V{v}^2-\frac{1}{\rho}\nabla P - \nabla\Phi
∂tv+ω×v=−21∇v2−ρ1∇P−∇Φ
In steady flow, dot multiply both sides by
v
\boldsymbol{v}
v
v
∂
t
v
+
v
(
ω
×
v
)
=
−
v
1
2
∇
v
2
−
v
1
ρ
∇
P
−
v
∇
Φ
0
=
v
⋅
∇
[
1
2
v
2
+
Φ
+
P
ρ
]
\newcommand{\V}[1]{\boldsymbol{#1}}\begin{aligned} \V{v}\partial_t\V{v} + \V{v}(\V{\omega}\times\V{v}) &= -\V{v}\frac{1}{2}\nabla\V{v}^2-\V{v}\frac{1}{\rho}\nabla P - \V{v}\nabla\Phi \\ 0 &= \V{v}\cdot\nabla[\frac{1}{2}\V{v}^2 + \Phi + \frac{P}{\rho}] \end{aligned}
v∂tv+v(ω×v)0=−v21∇v2−vρ1∇P−v∇Φ=v⋅∇[21v2+Φ+ρP]
which indicates that, along a streamline,
1
2
v
2
+
Φ
+
P
ρ
\frac{1}{2}v^2 + \Phi + \frac{P}{\rho}
21v2+Φ+ρP is constant, which is the Bernoulli’s Principle again.
For compressible fluid, we have
P = k ρ γ ⟹ d P = k γ ρ γ − 1 d ρ = ρ γ γ − 1 d P ρ P = k\rho^\gamma \implies dP = k\gamma\rho^{\gamma-1}d\rho = \rho\frac{\gamma}{\gamma-1}d\frac{P}{\rho} P=kργ⟹dP=kγργ−1dρ=ργ−1γdρP
Therefore, the Bernoulli’s Equation becomes
0 = v ⋅ ∇ [ 1 2 v 2 + Φ + γ P ( γ − 1 ) ρ ] 0 = \boldsymbol{v}\cdot\nabla[\frac{1}{2}v^2 + \Phi + \frac{\gamma P}{(\gamma - 1)\rho}] 0=v⋅∇[21v2+Φ+(γ−1)ργP]
If the vorticity is zero, then this conclusion applies everywhere instead of only along the streamline, because we don’t have to multiply both sides by v \boldsymbol{v} v.
If the vorticity of a flow is zero, then v = ∇ ϕ \boldsymbol{v} = \nabla \phi v=∇ϕ, which is known as potential flow. If imcompressible, we have a Laplace Equation ∇ ϕ = 0 \nabla \phi = 0 ∇ϕ=0
Circulation
For an arbitrary loop in the fluid, the circulation is defined as
Γ
=
∮
c
v
⋅
d
l
=
∬
S
ω
⋅
d
A
\newcommand{\V}[1]{\boldsymbol{#1}}\Gamma = \oint_{c} \V{v}\cdot d\V{l} = \iint_S \V{\omega}\cdot d\V{A}
Γ=∮cv⋅dl=∬Sω⋅dA
by Storker Theorem.
Baroclinity
Calculate the curl of both sides of the vorticity form of Euler’s Equation
∇
×
(
∂
t
v
+
ω
×
v
=
−
1
2
∇
v
2
−
1
ρ
∇
P
−
∇
Φ
)
\nabla\times \qquad (\newcommand{\V}[1]{\boldsymbol{#1}}\partial_t\V{v} + \V{\omega}\times\V{v} = -\frac{1}{2}\nabla\V{v}^2-\frac{1}{\rho}\nabla P - \nabla\Phi)
∇×(∂tv+ω×v=−21∇v2−ρ1∇P−∇Φ)
Since the curl of gradient is zero, and the curl of velocity is vorticity, we have
∂
t
ω
+
∇
×
(
ω
×
v
)
=
−
∇
×
1
ρ
∇
P
D
t
ω
=
∂
t
ω
+
(
v
⋅
∇
)
ω
=
1
ρ
2
∇
ρ
×
∇
P
\newcommand{\V}[1]{\boldsymbol{#1}}\begin{aligned} \partial_t\V{\omega} + \nabla\times (\V{\omega}\times\V{v}) &= -\nabla\times\frac{1}{\rho}\nabla P \\ \mathcal{D}_t\V{\omega} = \partial_t\V{\omega} + (\V{v}\cdot\nabla)\V{\omega} &= \frac{1}{\rho^2}\nabla\rho\times\nabla P \end{aligned}
∂tω+∇×(ω×v)Dtω=∂tω+(v⋅∇)ω=−∇×ρ1∇P=ρ21∇ρ×∇P
Define Baroclinity as ∇ P × ∇ ρ \nabla P \times \nabla \rho ∇P×∇ρ, which measures the misalignment between gradient of density and pressure.
If pressure is a function of density (which is known as Barotropic Flow) or the fluid is imcompressible, the RHS is zero. This indicates, the vorticity will stay zero without voscosity.
Stress Tensor
by continuity equation and the Euler’s Equation,
∂
t
(
ρ
v
)
=
v
∂
t
ρ
+
ρ
∂
t
v
=
−
v
∇
⋅
(
ρ
v
)
−
ρ
(
v
⋅
∇
)
v
−
∇
P
=
−
v
i
∂
j
(
ρ
v
j
)
−
ρ
v
j
∂
j
v
i
−
∂
i
P
=
−
∂
j
(
ρ
v
j
v
i
)
−
∂
i
P
=
−
∂
j
Π
j
i
−
∂
i
P
\newcommand{\V}[1]{\boldsymbol{#1}} \begin{aligned} \partial_t(\rho\V{v}) &= \V{v}\partial_t\rho + \rho\partial_t\V{v} = -\V{v}\nabla\cdot(\rho\V{v}) - \rho(\V{v}\cdot\nabla)\V{v} - \nabla P \\ &= -v_i\partial_j(\rho v_j) - \rho v_j\partial_j v_i - \partial_i P = -\partial_j(\rho v_j v_i) - \partial_i P \\ &= -\partial_j\Pi_{ji} - \partial_i P \end{aligned}
∂t(ρv)=v∂tρ+ρ∂tv=−v∇⋅(ρv)−ρ(v⋅∇)v−∇P=−vi∂j(ρvj)−ρvj∂jvi−∂iP=−∂j(ρvjvi)−∂iP=−∂jΠji−∂iP
which is known as the Stress Tensor
Π
j
i
=
ρ
v
j
v
i
−
δ
j
i
P
\Pi_{ji} = \rho v_jv_i - \delta_{ji}P
Πji=ρvjvi−δjiP
Water Waves
Assume that the water is imcompressible, no voscosity, no surface tension.
Plane Wave
Assume no vorticity, velocity has no y-dependent, with the Euler’s Equation
∂
t
v
+
(
v
⋅
∇
)
v
=
−
1
ρ
∇
P
−
g
z
^
\newcommand{\V}[1]{\boldsymbol{#1}}\partial_t\V{v}+(\V{v}\cdot\nabla)\V{v} = -\frac{1}{\rho}\nabla P - g\V{\hat{z}}
∂tv+(v⋅∇)v=−ρ1∇P−gz^
When amplitude is much smaller then wavelength, we can ignore the second term to make the equation linear. Therefore, we can use superposition principle and complex notation.
v
(
x
,
z
,
t
)
=
Re
u
(
z
)
e
i
(
k
x
−
ω
t
)
P
(
x
,
z
,
t
)
=
P
0
−
ρ
g
z
+
δ
P
(
x
,
z
,
t
)
δ
P
(
x
,
z
,
t
)
=
Re
δ
P
~
(
z
)
e
i
(
k
x
−
ω
t
)
\newcommand{\V}[1]{\boldsymbol{#1}}\begin{aligned} \V{v}(x,z,t) &= \text{Re}\;\V{u}(z)e^{i(kx-\omega t)} \\ P(x,z,t) &= P_0 - \rho gz + \delta P(x,z,t) \\ \delta P(x,z,t) &= \text{Re}\;\delta \tilde P(z) e^{i(kx-\omega t)} \end{aligned}
v(x,z,t)P(x,z,t)δP(x,z,t)=Reu(z)ei(kx−ωt)=P0−ρgz+δP(x,z,t)=ReδP~(z)ei(kx−ωt)
Calculate divergence of both sides of the Euler’s Equation,
∇
2
δ
P
=
0
⟹
δ
P
~
(
z
)
=
A
e
k
z
+
B
e
−
k
z
\nabla^2 \delta P = 0 \implies \delta\tilde P(z) = Ae^{kz} + Be^{-kz}
∇2δP=0⟹δP~(z)=Aekz+Be−kz
Take this into the Euler’s Equation
u
x
=
k
ρ
ω
δ
P
~
=
k
ρ
ω
(
A
e
k
z
+
B
e
−
k
z
)
u
z
=
−
i
ρ
ω
∂
z
δ
P
~
=
−
i
k
ρ
ω
(
A
e
k
z
−
B
e
−
k
z
)
\begin{aligned} u_x &= \frac{k}{\rho\omega}\delta\tilde P = \frac{k}{\rho\omega}(Ae^{kz} + Be^{-kz}) \\ u_z &= \frac{-i}{\rho\omega}\partial_z\delta\tilde P = \frac{-ik}{\rho\omega}(Ae^{kz} - Be^{-kz}) \end{aligned}
uxuz=ρωkδP~=ρωk(Aekz+Be−kz)=ρω−i∂zδP~=ρω−ik(Aekz−Be−kz)
Boundary condition:
- Ocean bottom: u z ( − h ) = 0 ⟹ B = A e − 2 h k u_z (-h) = 0 \implies B = Ae^{-2hk} uz(−h)=0⟹B=Ae−2hk
- Water surface: D t P = 0 \mathcal{D}_tP = 0 DtP=0
For the surface, drop the second order term,
∂
t
P
+
v
⋅
∇
(
P
0
−
ρ
g
z
+
δ
P
)
=
0
⟹
−
i
ω
δ
P
~
=
ρ
g
u
z
\partial_tP + \boldsymbol{v}\cdot\nabla(P_0 - \rho gz + \delta P) = 0 \implies -i\omega\delta\tilde P = \rho g u_z
∂tP+v⋅∇(P0−ρgz+δP)=0⟹−iωδP~=ρguz
simplify to first order, we can set
z
=
0
z = 0
z=0 and substitute
B
B
B, we get the dispersion relation
ω
(
k
)
\omega(k)
ω(k) for water wave (linear first order).
ω
2
=
g
k
A
−
B
A
+
B
=
g
k
tanh
(
h
k
)
\omega^2 = gk\frac{A-B}{A+B} = gk\tanh(hk)
ω2=gkA+BA−B=gktanh(hk)
Different situations
- shallow wave h < < λ ⟹ ω = g h k h << \lambda \implies \omega = \sqrt{gh}k h<<λ⟹ω=ghk
- deep wave h > > λ ⟹ ω = g k h >> \lambda \implies \omega = \sqrt{gk} h>>λ⟹ω=gk
Therefore, we can get the velocities of the water waves:
v
p
h
a
s
e
=
ω
k
=
g
k
tanh
(
h
k
)
v
g
r
o
u
p
=
d
ω
d
k
=
1
2
v
p
h
a
s
e
(
1
+
2
k
h
sinh
(
2
k
h
)
)
\begin{aligned}v_{phase} &= \frac{\omega}{k} = \sqrt{\frac{g}{k}\tanh(hk)} \\ v_{group} &= \frac{d\omega}{dk} = \frac{1}{2}v_{phase}\Big(1+\frac{2kh}{\sinh(2kh)}\Big)\end{aligned}
vphasevgroup=kω=kgtanh(hk)=dkdω=21vphase(1+sinh(2kh)2kh)
Solution
Let
C
=
2
A
e
−
k
h
C = 2Ae^{-kh}
C=2Ae−kh, we have the solution
δ
P
=
C
cosh
[
k
(
z
+
h
)
]
cos
(
k
x
−
ω
t
)
v
x
=
C
k
ρ
ω
cosh
[
k
(
z
+
h
)
]
cos
(
k
x
−
ω
t
)
v
z
=
C
k
ρ
ω
sinh
[
k
(
z
+
h
)
]
sin
(
k
x
−
ω
t
)
\begin{aligned} \delta P &= C \cosh[k(z+h)]\cos(kx-\omega t) \\ v_x &= \frac{Ck}{\rho\omega}\cosh[k(z+h)]\cos(kx-\omega t) \\ v_z &= \frac{Ck}{\rho\omega}\sinh[k(z+h)]\sin(kx-\omega t) \end{aligned}
δPvxvz=Ccosh[k(z+h)]cos(kx−ωt)=ρωCkcosh[k(z+h)]cos(kx−ωt)=ρωCksinh[k(z+h)]sin(kx−ωt)
Deep Water waves: (
k
h
kh
kh is large)
x
>
>
1
⟹
cosh
(
x
)
≈
sinh
(
x
)
≈
1
2
e
x
x >> 1 \implies \cosh(x) \approx \sinh(x) \approx \frac{1}{2}e^{x}
x>>1⟹cosh(x)≈sinh(x)≈21ex
Therefore,
(
v
x
v
y
)
=
A
k
e
k
z
ρ
ω
(
sin
(
k
x
−
ω
t
)
cos
(
k
x
−
ω
t
)
)
\left(\begin{matrix}v_x \\ v_y\end{matrix}\right) = \frac{Ake^{kz}}{\rho\omega}\left(\begin{matrix}\sin(kx-\omega t) \\ \cos(kx-\omega t)\end{matrix}\right)
(vxvy)=ρωAkekz(sin(kx−ωt)cos(kx−ωt))
Energy
η ( x , t ) \eta (x,t) η(x,t) is the extra height above z = 0 z=0 z=0 due to the wave. ξ ( x , t ) \boldsymbol{\xi}(x,t) ξ(x,t) is the displacement of a water particle.
Energy density (comes from the time average of Potential energy and Kinetic energy)
E
=
1
2
ρ
g
η
2
E = \frac{1}{2}\rho g\eta^2
E=21ρgη2
Energy flux is carried by the group velocity
F
=
E
v
g
r
F = Ev_{gr}
F=Evgr
When the wave goes onto beach, where
h
h
h is gradually decreasing, by the conservation of energy(flux), we have the Green’s Law
η
=
C
h
1
/
4
\eta = \frac{C}{h^{1/4}}
η=h1/4C
Break down of Approximation
In shallow waves,
η
=
ξ
z
(
z
=
0
)
=
C
k
2
h
ρ
ω
2
cos
(
k
x
−
ω
t
)
=
h
k
ω
v
x
=
h
v
x
v
p
h
\eta = \xi_z(z=0) = \frac{Ck^2h}{\rho\omega^2}\cos(kx-\omega t) = h\frac{k}{\omega}v_x = h\frac{v_x}{v_{ph}}
η=ξz(z=0)=ρω2Ck2hcos(kx−ωt)=hωkvx=hvphvx
When η ≳ h \eta \gtrsim h η≳h, v x ≳ v p h v_x \gtrsim v_{ph} vx≳vph, which is breaking down.
In deep waves,
v
x
∼
v
z
∼
∂
t
η
≈
ω
η
v_x \sim v_z \sim \partial_t\eta \approx \omega\eta
vx∼vz∼∂tη≈ωη
When η ≳ 1 / k \eta \gtrsim 1/k η≳1/k, v x ∼ v z ≳ v p h v_x \sim v_z \gtrsim v_{ph} vx∼vz≳vph, which is breaking down.
type | wave shape |
---|---|
spilling | skewed waves |
turnel | for surfing |
Profile steepen: velocity on the crest is larger, making the front side steeper. Therefore, in shallow water
v
p
h
=
3
g
h
e
f
f
e
c
t
i
v
e
−
2
g
h
v_{ph} = 3\sqrt{gh_{effective}} - 2\sqrt{gh}
vph=3gheffective−2gh
which is result of super complex derivation.
v
c
r
e
s
t
−
v
t
r
o
u
g
h
=
2
g
(
h
+
η
)
−
3
g
(
h
−
η
)
=
3
g
h
η
h
v_{crest} - v_{trough} = 2\sqrt{g(h+\eta)}-3\sqrt{g(h-\eta)} = 3\sqrt{gh}\frac{\eta}{h}
vcrest−vtrough=2g(h+η)−3g(h−η)=3ghhη
The time for the crest to catch the trough
t
≈
λ
/
2
3
g
h
η
/
h
=
λ
h
6
η
g
h
=
h
6
η
T
t \approx \frac{\lambda/2}{3\sqrt{gh}\eta/h} = \frac{\lambda h}{6\eta\sqrt{gh}} = \frac{h}{6\eta}T
t≈3ghη/hλ/2=6ηghλh=6ηhT
This forms the effect called Hydrolic Jump, which causes the instability even from a perfect sin/cos wave.
In deep water, Stalling effect acts against steepening, because from Fourier Transform, steeper waves have smaller k, which is slower.
When stalling and steepening balance each other, it is known as Solitary Wave or Solition.
Korteweg-DeVries Equation
2 Dutch in 1895
Correction on the dispersion
tanh
x
≈
x
−
x
3
3
⟹
ω
=
g
h
k
(
1
−
h
2
k
2
6
)
\tanh x \approx x-\frac{x^3}{3} \implies \omega = \sqrt{gh}k(1-\frac{h^2k^2}{6})
tanhx≈x−3x3⟹ω=ghk(1−6h2k2)
Substitute
ω
\omega
ω and
k
k
k with the partial derivatives of
η
\eta
η
∂
t
η
+
g
h
∂
x
η
+
g
h
h
2
6
∂
x
3
η
=
0
\partial_t\eta + \sqrt{gh}\partial_x\eta + \sqrt{gh}\frac{h^2}{6}\partial^3_x\eta = 0
∂tη+gh∂xη+gh6h2∂x3η=0
Adding a non-linear correction (should be something about
η
/
h
\eta/h
η/h), we get to the Korteweg-DeVries Equation
∂
t
η
+
g
h
(
1
+
3
η
2
h
)
∂
x
η
+
g
h
h
2
6
∂
x
3
η
=
0
\partial_t\eta + \sqrt{gh}(1+\frac{3\eta}{2h})\partial_x\eta + \sqrt{gh}\frac{h^2}{6}\partial^3_x\eta = 0
∂tη+gh(1+2h3η)∂xη+gh6h2∂x3η=0
obviously, the factor
3
/
2
3/2
3/2 comes from the effective
h
h
h equation in the previous section.
The solution could be
η
=
η
0
sech
2
[
3
η
0
4
h
3
(
x
−
u
t
)
]
with
u
=
g
h
(
1
+
η
0
2
h
)
\eta = \eta_0 \text{sech}^2\Big[\sqrt{\frac{3\eta_0}{4h^3}}(x-ut)\Big] \quad\text{with}\; u = \sqrt{gh}(1+\frac{\eta_0}{2h})
η=η0sech2[4h33η0(x−ut)]withu=gh(1+2hη0)
Surface Tension
leads to ripples and capillary waves (large k k k, short wavelength)
The effect which trys to minimize the surface, because surface causes energy.
Δ
E
=
γ
Δ
A
\Delta E = \gamma \Delta A
ΔE=γΔA
Capillary Waves
waves with large k
ω
2
=
γ
ρ
k
3
\omega^2 = \frac{\gamma}{\rho}k^3
ω2=ργk3
can be infered from dimensional analysis
Overrall, for large k
ω
2
=
g
k
+
γ
ρ
k
3
⟹
v
p
h
2
=
h
k
+
γ
ρ
k
\omega^2 = gk + \frac{\gamma}{\rho}k^3 \implies v_{ph}^2 = \frac{h}{k} + \frac{\gamma}{\rho}k
ω2=gk+ργk3⟹vph2=kh+ργk
which has a minimum value:
k
=
ρ
g
γ
v
p
h
2
=
2
g
γ
ρ
⟹
v
p
h
≈
0.23
m/s
k = \sqrt{\frac{\rho g}{\gamma}} \quad v_{ph}^2 = 2\sqrt{\frac{g\gamma}{\rho}} \implies v_{ph} \approx 0.23 \text{m/s}
k=γρgvph2=2ρgγ⟹vph≈0.23m/s
The following figure is
v
p
h
v_{ph}
vph versus
k
k
k (from shallow to deep to capillary)
[外链图片转存失败,源站可能有防盗链机制,建议将图片保存下来直接上传(img-svt4y8rs-1609270197968)(https://i.loli.net/2020/11/18/FijW5swYdUayGNE.png)]
Interface
in the interface between two liquid (
ρ
1
,
ρ
2
\rho_1, \rho_2
ρ1,ρ2 from up to down)
ω
2
=
g
k
ρ
2
−
ρ
1
ρ
2
+
ρ
1
+
γ
k
3
ρ
2
+
ρ
1
\omega^2 = gk\frac{\rho_2-\rho_1}{\rho_2+\rho_1} + \frac{\gamma k^3}{\rho_2+\rho_1}
ω2=gkρ2+ρ1ρ2−ρ1+ρ2+ρ1γk3
If the ω \omega ω is imarginary, the waves become exponential.
Rayleigh-Taylor Instability
ω
2
<
0
⟹
k
2
<
g
∣
ρ
2
−
ρ
1
∣
γ
⟹
λ
>
2
π
γ
g
∣
ρ
2
−
ρ
1
∣
\omega^2 < 0 \implies k^2 < \frac{g|\rho_2-\rho_1|}{\gamma} \implies \lambda > 2\pi\sqrt{\frac{\gamma}{g|\rho_2-\rho_1|}}
ω2<0⟹k2<γg∣ρ2−ρ1∣⟹λ>2πg∣ρ2−ρ1∣γ
which implies longer wave length causes more instability.
Viscosity
Parity: x → − x , y → − y , z → − z ⟹ v → − v x\to -x,\; y \to -y,\; z\to -z \implies v\to -v x→−x,y→−y,z→−z⟹v→−v
Time reverse: t → − t ⟹ v → − v t \to -t \implies v\to -v t→−t⟹v→−v
Newtonian holds parity invariance, but friction does not hold time reverse invariance.
Ignore gravity, Euler’s Equation respects parity and time reverse invariance. However, we can add an extra term that respects parity but violates time reverse invariance. The possible and reasonable choice is Laplacian.
Navier–Stokes equations
Rewrite the Euler’s Equation with the extra term known as the Kinematic Viscosity, we have an equation which deserves the title (also known as N-S Equation):
∂
v
∂
t
+
(
v
⋅
∇
)
v
=
−
∇
P
ρ
−
∇
Φ
+
ν
∇
2
v
\newcommand{\V}[1]{\boldsymbol{#1}}\frac{\partial\V{v}}{\partial t} + (\V{v}\cdot\nabla)\V{v} = -\frac{\nabla P}{\rho} - \nabla\Phi + \nu\nabla^2\V{v}
∂t∂v+(v⋅∇)v=−ρ∇P−∇Φ+ν∇2v
- ν \nu ν is called kinematic viscosity coefficient.
- η = ρ ν \eta = \rho\nu η=ρν is called dynamic viscosity coefficient.
The curl of the equation gives, even if the liquid is barotropic, viscosity generates vorticity at low velocity (boundary)
D
t
ω
=
ν
∇
2
ω
\mathcal{D}_t\boldsymbol{\omega} = \nu\nabla^2\boldsymbol{\omega}
Dtω=ν∇2ω
Reynolds Number
the relative importance of inertial and viscosity
R
e
=
inertial
viscosity
=
(
v
⋅
∇
)
v
ν
∇
2
v
=
v
l
ν
Re = \frac{\text{inertial}}{\text{viscosity}} = \frac{(v\cdot\nabla)v}{\nu\nabla^2v} = \frac{vl}{\nu}
Re=viscosityinertial=ν∇2v(v⋅∇)v=νvl
This is not a precisely defined quantity, since the length is any kinds of length scale in the system.
- Low Re: slow, high viscosity (Laminar Flow)
- High Re: Fast, low viscosity (Turbulence Flow)
Hagen-Poisieulle Flow
The steady incompressible flow through a pipe with radius R R R and length l l l.
Velocity is only function of radius (
v
=
v
z
(
r
)
z
^
\boldsymbol{v} = v_z(r)\hat{\boldsymbol{z}}
v=vz(r)z^), which has no dependency on
z
z
z. Thus, the LHS of N-S Equation is zero. Hence
Δ
P
ρ
l
+
ν
d
d
r
[
r
d
v
z
d
r
]
=
0
⟹
d
d
r
[
r
d
v
d
r
]
=
−
K
<
0
\frac{\Delta P}{\rho l} + \nu \frac{d}{dr}[r\frac{dv_z}{dr}] = 0 \implies \frac{d}{dr}[r\frac{d\boldsymbol{v}}{dr}] = -K < 0
ρlΔP+νdrd[rdrdvz]=0⟹drd[rdrdv]=−K<0
Solve to get
v
z
(
r
)
=
−
K
4
(
R
2
−
r
2
)
,
K
=
Δ
P
ν
ρ
l
v_z(r) = -\frac{K}{4}(R^2-r^2), \qquad K = \frac{\Delta P}{\nu \rho l}
vz(r)=−4K(R2−r2),K=νρlΔP
The mass current (mass passed per unit time) is
J
=
∫
ρ
v
z
(
r
)
d
A
=
π
Δ
P
8
ν
l
R
4
J = \int \rho v_z(r)dA = \frac{\pi\Delta P}{8\nu l}R^4
J=∫ρvz(r)dA=8νlπΔPR4
Murray’s Law
1926
- Power needed per unit length ∼ J D l P = C J 2 / R 4 \sim J\mathcal{D}_lP= CJ^2/R^4 ∼JDlP=CJ2/R4
- Power needed to maintenance ∼ C ′ R 2 \sim C'R^2 ∼C′R2
To optimize, we have
R
o
p
t
i
o
n
a
l
=
k
J
1
/
3
R_{optional} = kJ^{1/3}
Roptional=kJ1/3
Stokes Drag
A sphere in the fluid in Low Re
F
=
6
π
η
R
v
\boldsymbol{F} = 6\pi \eta R\boldsymbol{v}
F=6πηRv
Terminal Velocity of a sphere (
ρ
s
\rho_s
ρs) falling through a fluid
6
π
η
R
v
+
4
3
π
R
3
ρ
g
=
4
3
π
R
3
ρ
s
g
⟹
v
=
2
(
ρ
s
−
ρ
)
R
2
g
9
ρ
ν
6\pi\eta Rv + \frac{4}{3}\pi R^3 \rho g = \frac{4}{3}\pi R^3\rho_s g \implies v = \frac{2(\rho_s-\rho)R^2g}{9\rho\nu}
6πηRv+34πR3ρg=34πR3ρsg⟹v=9ρν2(ρs−ρ)R2g
Rain Drop R = 2 m m , ν = 15 m m 2 / s ⟹ v ≈ 600 m / s R = 2mm, \; \nu = 15mm^2/s \implies v \approx 600m/s R=2mm,ν=15mm2/s⟹v≈600m/s
which is completely wrong! since the Reynolds Number is very large. Stokes Formula cannot be applied.
Boundary Layer
Turbulent flow is super complicated with no analytic solution.
For laminar flow, the thickness of boundary layer is
δ
∼
ν
t
=
5
ν
x
v
⟹
δ
x
∼
5
R
e
x
\delta \sim \sqrt{\nu t} = 5\sqrt{\frac{\nu x}{v}} \implies \frac{\delta}{x}\sim\frac{5}{\sqrt{Re_x}}
δ∼νt=5vνx⟹xδ∼Rex5
Turbulence
Turbulence is critical in everyday life, e.g. heat exchanger. Cost: increase drag and require more power.
Kelvin-Helmholtz Instability
Interface of two fluid ρ 1 \rho_1 ρ1 and ρ 2 \rho_2 ρ2 with velocity v 1 v_1 v1 and v 2 v_2 v2 respectively. If v 1 ≠ v 2 v_1\neq v_2 v1=v2, then the fluid is instable, known as Kelvin-Helmholtz Instability.
Without gravity/viscosity/surface tension, like water waves:
{
∂
x
v
x
+
∂
z
v
z
=
0
∂
t
v
x
+
v
x
∂
x
v
x
+
v
z
∂
z
v
x
=
−
∂
x
P
/
ρ
∂
t
v
z
+
v
x
∂
x
v
z
+
v
z
∂
z
v
z
=
−
∂
z
P
/
ρ
\left\{\begin{aligned}&\partial_xv_x+\partial_zv_z = 0 \\ &\partial_tv_x + v_x\partial_xv_x + v_z\partial_zv_x = -\partial_xP/\rho \\ &\partial_tv_z + v_x\partial_xv_z + v_z\partial_zv_z = -\partial_zP/\rho \end{aligned}\right.
⎩⎪⎨⎪⎧∂xvx+∂zvz=0∂tvx+vx∂xvx+vz∂zvx=−∂xP/ρ∂tvz+vx∂xvz+vz∂zvz=−∂zP/ρ
Consider derivation
v
x
=
u
+
δ
u
,
v
z
=
δ
v
,
P
=
P
0
+
δ
P
v_x = u + \delta u,\; v_z = \delta v,\; P = P_0 + \delta P
vx=u+δu,vz=δv,P=P0+δP, drop higher order terms:
{
∂
x
δ
u
+
∂
z
δ
v
=
0
∂
t
δ
u
+
u
∂
x
δ
u
=
−
∂
x
δ
P
/
ρ
∂
t
δ
v
+
u
∂
x
δ
v
=
−
∂
z
δ
P
/
ρ
\left\{\begin{aligned}&\partial_x\delta u+\partial_z\delta v = 0 \\ &\partial_t\delta u + u\partial_x\delta u = -\partial_x\delta P/\rho \\ &\partial_t\delta v + u\partial_x\delta v = -\partial_z\delta P/\rho \end{aligned}\right.
⎩⎪⎨⎪⎧∂xδu+∂zδv=0∂tδu+u∂xδu=−∂xδP/ρ∂tδv+u∂xδv=−∂zδP/ρ
Assume wave behavior,
δ
X
=
δ
X
~
(
z
)
e
i
(
k
x
−
ω
t
)
\delta X = \delta \tilde X(z)e^{i(kx-\omega t)}
δX=δX~(z)ei(kx−ωt)
{
i
k
δ
u
~
+
∂
z
δ
v
~
=
0
ρ
(
−
i
ω
+
u
i
k
)
δ
u
~
=
−
i
k
δ
P
~
ρ
(
−
i
ω
+
u
i
k
)
δ
v
~
=
−
∂
z
δ
P
~
\left\{\begin{aligned}&ik\delta\tilde u+\partial_z\delta\tilde v = 0 \\ &\rho(-i\omega+uik)\delta\tilde u = -ik\delta\tilde P \\ &\rho(-i\omega+uik)\delta\tilde v = -\partial_z\delta\tilde P \end{aligned}\right.
⎩⎪⎨⎪⎧ikδu~+∂zδv~=0ρ(−iω+uik)δu~=−ikδP~ρ(−iω+uik)δv~=−∂zδP~
Combine the three equation by differentiating the first w.r.t. z again:
d
2
δ
v
~
d
z
2
=
−
i
k
∂
z
δ
u
~
=
−
k
2
ρ
(
−
i
ω
+
u
i
k
)
∂
z
δ
P
~
=
k
2
δ
v
~
⟹
δ
v
~
=
C
e
±
k
z
\begin{aligned}\frac{d^2\delta\tilde v}{dz^2} &= -ik\partial_z\delta\tilde u = \frac{-k^2}{\rho(-i\omega + uik)}\partial_z\delta\tilde P \\ &= k^2\delta\tilde v \\ \implies&\delta\tilde v = Ce^{\pm kz}\end{aligned}
dz2d2δv~⟹=−ik∂zδu~=ρ(−iω+uik)−k2∂zδP~=k2δv~δv~=Ce±kz
For convergence, the upper layer has negative sign, the lower layer has positive sign. Apply boundary condition for a wave form displacement:
δ
v
(
z
=
0
)
=
D
t
η
⟹
δ
v
~
=
η
0
i
(
u
k
−
ω
)
e
±
k
z
\delta v(z=0) = \mathcal{D}_t \eta \implies \delta\tilde v = \eta_0i(uk-\omega)e^{\pm kz}
δv(z=0)=Dtη⟹δv~=η0i(uk−ω)e±kz
Apply the boundary condition for pressure, we get:
P
1
(
z
=
0
)
=
P
2
(
z
=
0
)
⟹
ρ
1
(
u
1
k
−
ω
)
2
+
ρ
2
(
u
2
k
−
ω
)
2
=
0
P_1(z=0) = P_2(z=0) \implies \rho_1(u_1k-\omega)^2 + \rho_2(u_2k-\omega)^2 = 0
P1(z=0)=P2(z=0)⟹ρ1(u1k−ω)2+ρ2(u2k−ω)2=0
which gives a complex solution:
ω
=
ρ
1
u
1
+
ρ
2
u
2
ρ
1
+
ρ
2
k
±
i
ρ
1
ρ
2
(
u
1
−
u
2
)
ρ
1
+
ρ
2
k
\omega = \frac{\rho_1u_1+\rho_2u_2}{\rho_1+\rho_2}k \pm i\frac{\sqrt{\rho_1\rho_2}(u_1-u_2)}{\rho_1+\rho_2}k
ω=ρ1+ρ2ρ1u1+ρ2u2k±iρ1+ρ2ρ1ρ2(u1−u2)k
Von Karman Vortex Street
with high Reynold number, the time reversed invariance can be held. The sphere in the flow will result in a turbulence, because the high speed fluid element passing by the sphere will not merge into a higher pressure region.
The increasing pressure along the sphere surface causes a Boundary Layer Separation (reversed flow)
Drag Coefficient
C
D
C_D
CD is defined by the equation below
F
D
=
1
2
ρ
v
2
A
C
D
F_D = \frac{1}{2}\rho v^2AC_D
FD=21ρv2ACD
where
A
A
A is the incident area.
- Stokes: C D = 24 / R e C_D = 24/Re CD=24/Re
- Constant: Pressure difference goes from ∼ v \sim v ∼v to ∼ v 2 \sim v^2 ∼v2
- Dip: Boundary layer becomes turbulent.
Kolmogoror Theory
ϵ
\epsilon
ϵ is the energy per unit mass per unit of time supplied into the fluid. For eddy size
d
d
d, the velocity of the eddy should be
v
(
d
)
=
(
ϵ
d
)
1
/
3
v(d) = (\epsilon d)^{1/3}
v(d)=(ϵd)1/3
the minimum eddy is determined by the viscosity and the energy supplied:
d
m
i
n
=
ν
3
/
4
ϵ
1
/
4
d_{min} = \frac{\nu^{3/4}}{\epsilon^{1/4}}
dmin=ϵ1/4ν3/4
Combine the two equation together, we have a surprising result
d
m
a
x
d
m
i
n
=
(
v
m
a
x
d
m
a
x
ν
)
3
/
4
=
R
e
3
/
4
\frac{d_{max}}{d_{min}} = (\frac{v_{max}d_{max}}{\nu})^{3/4} = Re^{3/4}
dmindmax=(νvmaxdmax)3/4=Re3/4
Fourier Transform gives
d
∼
λ
/
2
∼
π
/
k
d \sim \lambda/2 \sim \pi/k
d∼λ/2∼π/k, the energy spectrum is
E
=
[
ϵ
2
/
3
k
−
5
/
3
]
d
k
E = [\epsilon^{2/3}k^{-5/3}]dk
E=[ϵ2/3k−5/3]dk
Temperature
Diffusion Equation
∂
t
T
=
κ
∇
2
T
\partial_tT = \kappa\nabla^2T
∂tT=κ∇2T
with fluid derivatives
D
t
T
=
κ
∇
2
T
\mathcal{D}_tT = \kappa\nabla^2T
DtT=κ∇2T
Boussinesq Approximation: ρ = ( 1 − α T ) \rho = (1-\alpha T) ρ=(1−αT), only with g g g