You are on page 1of 5

Lecture 1.

Fluid Kinematics

Description of Fluid Motion.

Before we look at the full 3D problem, let’s consider the 1D case – it contains all the basic ideas.
Consider a continuum that can move only in one direction, say x. At some initial time, say t = 0,
suppose we know the positions of all the particles of the continuum. The position of a particle
initially at ξ at t = 0 is then a function x = g(ξ, t). We assume that g can always be inverted to
give ξ = g −1 (x, t), i.e., to tell where the particle that is at x at time t was at time 0. The velocity
of this particle is
∂g
v(ξ, t) = .
∂t
It is a function of ξ, t. For a continuum, it is often convenient to describe a velocity field by giving the
velocity at each fixed point x in space as a function of time, i.e., by giving u(x, t) = v(g −1 (x, t), t).
The inverse of this relation is: v(ξ, t) = u(g(ξ, t), t) If we also want to express the acceleration of a
fluid particle in the same way, we proceed as follows:
∂v ∂u(g(ξ, t), t) ∂u ∂u ∂g ∂u ∂u
a(ξ, t) = = = +( ) = + v(ξ, t) ,
∂t ∂t ∂t ∂x ∂t ∂t ∂x
and if we now substitute ξ = g −1 (x, t) for ξ we find the acceleration at a fixed point in the continuum
is
∂u ∂u Du
a(x, t) = +u ≡ .
∂t ∂x Dt
The expression DQ/Dt is called the substantial derivative of a continuum property Q(x, t), and
represents the change in Q as the particle carrying Q moves. The two ways of describing the
motion of a fluid or continuum kinematically are known as the Lagrangian description (ξ, t the
basic independent variables) and Eulerian description (x, t the basic independent variables).

Before moving on to higher space dimensions, lets look a an important result which can be used to
derive the fundamental conservation laws for any continuum. Again, we work in 1D. Suppose Q is
any property, associated with the continuum (say its mass density). Then the amount of Q in a
moving layer is
Z b(t)
I(t) = Q(x, t) dx,
a(t)

where a(t) = g(α, t), b(t) = g(β, t). We want to find an expression for dI/dt. Since the limits of
integration depend on t we introduce the change of variables x = g(ξ, t). Then we have
Z β
∂g
I= Q(g(ξ, t), t) dξ.
α ∂ξ
Differentiating we obtain
Z β ² ³
dI ∂Q ∂Q ∂g ∂g ∂v
= ( + ) +Q dξ,
dt α ∂t ∂x ∂t ∂ξ ∂ξ
where we used the fact that
∂ ∂g ∂ ∂g ∂v
= = .
∂t ∂ξ ∂ξ ∂t ∂ξ

1
Now we use the fact that v(ξ, t) = u(g(ξ, t), t) to write

∂v ∂u ∂g
=
∂ξ ∂x ∂ξ
. Then we substitute into the above integral to obtain
Z β² ³
dI ∂Q ∂Q ∂g ∂u ∂g
= + +Q dξ.
dt α ∂t ∂x ∂t ∂x ∂ξ

Finally, we change back to the original variable of integration using ξ = g −1 (x, t) and introduce the
substantial derivative notation to obtain
Z b(t) ² ³
dI DQ ∂u
= +Q dx
dt a(t) Dt ∂x

The three dimensional version of this result is called the Reynold’s Transport theorem. It is proved
in much the same way allowing for the increased complications of 3D.

Now we turn to the full 3D case. The Lagrangian description gives the position ~x of each fluid particle
~ x0 , t).
at time t in terms of its position ~x0 at an arbitrarily selected initial time t = 0, i.e., ~x = Φ(~
~ ~
Given the function Φ, the velocity of the particle at ~x = Φ(~x0 , t) at time t is determined1 from
~ ~
x0 , t), and the acceleration from ~aL (~x0 , t) = ∂ 2 Φ/∂t
~u L (~x0 , t) ≡ ∂ Φ/∂t(~ 2
(~x0 , t).

The Eulerian description gives the velocity and acceleration of the fluid at a point in the flow as
functions of the coordinates of the point and time. Thus, in the Eulerian description

~u (~x, t) = (u(~x, t), v(~x, t), w(~x, t)), ~a(~x, t) = (ax (~x, t), ay (~x, t), az (~x, t)),

where ~x can be any point in the flow domain. In general, different fluid particles are at the point at
each time so we must actually develop expressions for the Eulerian functions u(~x, t), etc., in terms
of the Lagranginan quantities which have well defined meanings. The function ~u can be determined
from the Lagrangian velocity, ~u L (and vice-versa) using the relations

~u (~x, t) ~
= ∂ Φ/∂t( ~ −1 (~x, t), t) = ~u L (Φ
Φ ~ −1 (~x, t), t),
~
~u L (~x0 , t) = ~u (Φ(~x0 , t), t),

~ −1 is the inverse of Φ,
where Φ ~ i.e., ~x = Φ(~
~ x0 , t) ⇔ ~x0 = Φ
~ −1 (~x, t) or Φ
~ −1 (Φ(~
~ x0 , t), t) = ~x0 .

The acceleration of a fluid particle in the Lagrangian description is simply the time derivative of
the velocity holding ~x0 constant. Thus, ~aL (~x0 , t) = ∂~u L (~x0 , t)/∂t. It is important to translate the
acceleration into Eulerian variables so it can be used in our presentation of Newton’s second law.
~ x0 , t), t), we find via the chain rule that
Using the equation ~u L (~x0 , t) = ~u (Φ(~

∂~u ∂~u ∂~u ∂~u


~aL (~x0 , t) = +u +v +w ,
∂t ∂x ∂y ∂z
1 In order that the velocity and acceleration make sense, we need to assume that Φ ~ has at least two deriva-
tives with respect to t. We will also require that for every t, the system of equations ~ ~ x0 , t) =
x = (x, y, z) = Φ(~
x0 , t), φ2 (~
(φ1 (~ x0 , t), φ3 (~
x0 , t)) be solvable for ~
x0 . We assure this by the requiring that the first partial derivatives of
~ = (φ1 , φ2 , φ3 ) with respect to the components of ~
Φ x0 = (x01 , x02 , x03 ) exist and be continuous, and that the jacobian
matrix J = [∂φi /∂~ x0i ] be positive for all t > 0 (At t = 0, J = 1 since φi (~ x0 , 0) = x0i .)

2
where all the quantities on the right hand side are evaluated at the point ~x = Φ(~ ~ x0 , t). Now
~ −1 ~ −1
substituting ~x0 = Φ (~x, t) for ~x0 and using ~a(~x, t) = ~aL (Φ (~x, t), t) we have

D~u ∂~u ∂~u ∂~u ∂~u


~a(~x, t) = ≡ +u +v +w .
Dt ∂t ∂x ∂y ∂z
The operator
D() ∂() ∂() ∂() ∂()
= +u +v +w
Dt ∂t ∂x ∂y ∂z
introduced above occurs in a number of places in our work and is called the substantial or material
derivative or the derivative following a fluid particle.

The Eulerian description is the preferred one in the overwhelming number of applications since we
are not usually interested in individual fluid particles.

Kinematic Concepts

We first consider three types of paths or curves associated with fluid motions: particle paths, stream
lines, streak lines. We work in 2D (1D is too simple to be illustrative in this context). Let the
motion be described by ~x = (x, y) = G(ξ,~ t) with ξ~ = (ξ, η). Now

∂~x ~ t),
= ~v (ξ,
∂t

and substituting ξ~ = G−1 (~x, t) we have ∂~


x
∂t = ~
u(~x, t). If we consider ~u as known, then these two
equations become a system of first order ODEs for the particle paths
dx dy
= u(x, y, t) = v(x, y, t)
dt dt
~
with initial conditions x(0) = ξ, y(0) = η for a particle starting at ξ.

As an example, consider a flow described by u = x/(1 + t), v = y, t ≥ 0. The particle paths are
determined from
dx x dy
= = y.
dt 1 + t dt
The solutions of these equations are: x = ξ(1 + t), y = ηet or eliminating the parameter t,

x−ξ
y = η exp( ).
ξ

Next, we define the stream lines as curves everywhere tangent to the velocity vector. Again, assuming
~u(~x, t) to be known, we obtain a system of two first order ODEs for the stream lines by setting
d~x/ds = ~u(~x, t) with t fixed and s a parameter along the stream line, say distance from some fixed
point. That is in a non steady flow there are different streamlines at each instant in general. In
component form
dx dy dx dy
= u(x, y, t), = v(x, y, t) or =
ds ds u v

3
with t a fixed parameter. For the example considered above,
dx x dy
= =y
ds 1+t ds
s
with solutions x = a exp( 1+t ), y = bes . Eliminating the parameter s we have y = b(x/a)1+t . Since
this is a time dependent flow we have a new set of stream lines at every t. For time independent
flows path lines and stream lines obviously coincide.

Finally, we consider streak lines. Let ~x0 be a fixed point and t a fixed time. The curve connecting
all particles which passed through ~x0 at some time s, 0 ≤ s ≤ t is the streak line through ~x0 . If we
release a neutrally buoyant non diffusing dye into the flow at point ~x0 , then the path traced out by
the colored fluid particles is the streak line through ~x0 . If 0 ≤ s ≤ t, ξ~ = G−1 (~x0 , s) is position at
time 0 of the particle which passed ~x0 at time s. At time t this particle is at ~x = G(G−1 (~x0 , s), t),
and as s varies between 0 and t we get the streak line. In our example, ξ = x0 /(1 + s), η = y0 es ,
and substituting into G gives
1+t
x = x0 , y = y0 et−s
1+s

Next we turn to a consideration of rotation and distortion of fluid elements and how to measure
it. For simplicity we restrict our discussion to 2D motion parallel to the x, y-plane. Consider a
small rectangular element of fluid at time t with lower left corner at (x, y) and upper right corner
at (x + dx, y + dy). Because the velocities at all corners are slightly different the element will distort
as it moves during a small time increment dt. At t + dt the position of the corners will change as
follows (dependence of the velocity components on time is not indicated explicitly)

(x, y) → (x + u(x, y)dt, y + v(x, y)dt),


(x, y + dy) → (x + u(x, y + dy)dt, y + dy + v(x, y + dy)dt),
(x + dx, y) → (x + dx + u(x + dx, y)dt, y + v(x + dx, y)dt),
(x + dx, y + dy) → (x + dx + u(x + dx, y + dy)dt, y + dy + v(x + dx, y + dy)dt).

For suitably small dx, dy we can assume the velocity variation along the sides is linear and given
by Taylor’s theorem, i.e.,
u(x + dx, y) = u(x, y) + ∂u/∂xdx,
u(x, y + dy) = v(x, y) + ∂v/∂ydy,
etc.. To this degree of approximation, the sides will remain straight and the original rectangle will be
transformed to a quadrilateral. The angle dθx between the x-axis and the side originally occupying
[x, x + dx] becomes
∂v/∂xdxdt
tan(dθx ) ≈ dθx = ,
dx + ∂u/∂xdxdt
and we see that this angle is increasing (in the counterclockwise sense) at approximately the rate

dθx ∂v
=
dt ∂x
Similarly, the angle dθy between the y-axis and the side originally occupying [y, y + dy] becomes

∂u/∂ydydt
tan(dθy ) ≈ dθy = − ,
dy + ∂v/∂ydydt

4
and we see that this angle is increasing (again in the counterclockwise sense) at approximately the
rate
dθy ∂u
=−
dt ∂y
. The angular velocity of the element is defined as the mean of these two rates of change, i.e.,
1 dθx dθy 1 ∂v ∂u
ωz ≡ ( + )= ( − )
2 dt dt 2 ∂x ∂y
Of course, this is for a two dimensional situation. In general we find the following: The angular
~ of a fluid element is one half of curl(~u ) = ∇ × ~u . In fluid mechanics, the curl of the
velocity ω
velocity is also referred to as the vorticity vector, and is defined by
~k ¬
¬ ¬
¬ ~ı ~
ω = ζ~ = ¬¬ ∂x
¬ ¬
2~ ∂ ∂ ∂ ¬
∂y ∂z ¬
¬ u v w¬

where the determinant is expanded using cofactors of the first row. Note that the z-component of
the angular velocity as given by this formula agrees with the result computed above.

The rates of angular deformation of a fluid element can also be determined from the velocity field.
For the 2D element just described, the rate of decrease in angle between the two segments originally
parallel to the x and y axes is called 2γxy = 2γyx and the above discussion shows that

1 dθx dθy 1 ∂v ∂u
2γxy = ( − )= ( + )
2 dt dt 2 ∂x ∂y
. It may be shown that the corresponding rates in 3D motion are given by the formulas:
1 ∂v ∂u 1 ∂w ∂v 1 ∂u ∂w
γxy = γyx = ( + ), γyz = γzy = ( + ), γzx = γxz = ( + ).
2 ∂x ∂y 2 ∂y ∂z 2 ∂z ∂x
In these expressions, 2γab , a, b = x, y, z, a 6= b is the rate of decrease of angle between elements of
fluid which are parallel to the coordinate axes a, b. The rates of change of length per unit length of
fluid elements parallel to the coordinate axes x, y and z are
∂u ∂v ∂w
γxx = , γyy = γzz = .
∂x ∂y ∂z

p for the 2D motion considered, the length of the segment originally parallel to the x axis becomes
E.g.,
(dx + ∂u/∂xdxdt)2 + (∂v/∂xdxdt)2 ≈ dx(1 + ∂u/∂xdt). The rate of change of volume of a fluid
element per unit volume is given by the divergence of the velocity field

div(~u ) = ∇ · ~u

You might also like