21

4. VORTICITY Integral: the magnitude of the circulation is the vor-

ticity integrated around a closed path:
For God’s Sake Let us sit upon the ground I Z
and tell sad stories K = v · dl = (∇ × v) · dA (4.2)
Of vortex filaments. A

How some have been ill-posed, some sin- where Stokes’ theorem is used in the second equality.
gular, As we noted above, K can be taken as a vector, with
Some poisoned by their self induction, direction given by the sense of the flow, using the right
some core size killed, hand rule. We can think of K as a “flux” of vorticity;
Some haunted by the mathematics they it measures the amount of ω passing through some sur-
have involved. face bounded by dl.
Let’s look at some examples. Solid body rotation is
All murderous.
described by vθ = Ωr;

For within the swirling motion that 1 ∂

ω= (rvθ ) = 2Ω ;
rounds the mortal circulation r ∂r
Of a vortex
Z 2π (4.3)
Keeps futility his court, K= vθ rdθ = 2πr 2 Ω
0
And there the non-linearity sits
Scoffing at his state and grinning at his so solid body rotation has a net vorticity, and a net cir-
theories culation.
Allowing him a breath, a little scene to An Irrotational vortex comes from potential flow.
linearize, compute and fill with ap- Pick vθ = C/r; then rvθ = constant, so that
proximations
And then at last he comes and with a ω = 0, r 6= 0 ;
little inconsistency bores through the Z 2π (4.4)
costly hopes and
K= vθ rdθ = 2πC
0
Farewell . . .
Thus, circular streamlines do not imply that the flow
has a net vorticity everywhere. In this example ω 6= 0
Shakespeare, Richard II, Act 3, Scene 2 only at the origin; elsewhere the flow is curl-free. The
“translated” by H. C. Yuen, quoted in divergeance at the origin, however, is enough to give
Vortex Dynamics, P. G. Saffman this vortex a net circulation.
A. Vortex Kinematics The Rankine vortex is a more physical extension of
the irrotational one. A real vortex must have a rotational
Vorticity is a large topic, and we can only treat it lightly core (we can’t have vθ → ∞). We can approximate this
here. We start with kinematics. We saw an example in with vθ ∝ r, r ≤ R (the radius of the core of the vor-
Chapter 3 of a flow field with a net circulation: in which tex); and vθ ∝ 1/r for r > R. Thus, the Rankine vortex
the line integral of the velocity, around a closed loop, is has uniform vorticity in its core, and zero elsewhere.
finite. Such a flow has a net circulation. We generalize
this here, with differential and integral definitions. B. Vortex Dynamics
We will only need this result in the nearly-
incompressible limit. That is, we take ∇ · v ≃ 0, but
Differential: vorticity is the curl of the velocity field: formally allow ∇ρ 6= 0. We start with the incompress-
ω = ∇×v (4.1) ible form of the Navier-Stokes equation, from (2.38),1
∂v 1
We can think of vortex lines – just like magnetic field + (v · ∇)v + ∇p = −∇Φ − ν∇ × (∇ × v) (4.5)
lines or streamlines, they are lines which are every- ∂t ρ
where tangent to the local vorticity. Also we will run
into vortex tubes, regions of high vorticity (picture a 1
The viscous term is written slightly differently, by the magic of
tornado; that is an extreme example). vector analysis.
 22

To proceed, take the curl of (4.5). Using the vector re- Proof #1 Kundu proves this as follows. Start with Eu-
lation v · ∇v = (∇ × v) × v + 12 ∇v 2 , we get the ler’s equation, (1.10) (note we aren’t worried about vis-
dynamical equation for vorticity: cosity right now):
Dv 1
ω
∂ω 1 = ∇p + ∇Φ (4.11)
ω × v) = ν∇2ω + 2 ∇ρ × ∇p (4.6)
+ ∇ × (ω Dt ρ
∂t ρ
Now, take the formal derivative of K:
or2
DK Dv D
I I
= · dl + v · dl (4.12)
ω
Dω 1 Dt Dt Dt
= ω · ∇v + ν∇2ω + 2 ∇ρ × ∇p (4.7)
Dt ρ But using Euler’s equation, and noting that ∇p · dl =
dp, the difference in pressure between two adjacent
A possible further step is to consider the barotropic points:
limit, in which the pressure is only a function of the
DK dp D
I I I
density [p = p(ρ)], for instance in an adiabatic gas with = ∇Φ · dl − + v· dl (4.13)
p/ργ = constant. In that case, ∇ρ k ∇p, so their cross Dt ρ Dt
product is zero and the equations simplify, to Now: the force, being from a potential, is conservative;
so the first term goes to zero. If the density is only a
ω
∂ω function of pressure, ρ = ρ(p) – that is if the fluid is
ω × v) = ν∇2ω
+ ∇ × (ω (4.8)
∂t barotropic – then dp/ρ is a perfect differential, so that
the second integral (around a closed curve) also goes to
and zero. Finally, the third term involves D(dl)/Dt = dv,
and thus v · D(dl)/Dt = D(v 2 /2)/Dt, also a perfect
ω
Dω differential, so the third term also vanishes. Thus, K is
= ω · ∇v + ν∇2ω (4.9)
Dt a constant of the motion in this flow.

Yet another possible step is to work in a rotating Proof #2 I have not seen this in a text, but it follows
frame. Referring back to chapter 1, the same analysis directly from magnetic field flux freezing arguments
gives here, (which will be presented in a later chapter). Here, we
keep the viscosity term around, but retain the incom-
ω
Dω 1 pressible, barotropic limit. That means we are again
= (ω Ω) · ∇v + ν∇2ω + 2 ∇ρ × ∇p (4.10)
ω + 2Ω using (4.5) as the governing equation, and (4.6) derived
Dt ρ
from it, but now assuming ∇ρ × ∇p = 0. The formal
derivative of K can now be written
C. Conservation of Circulation: Kelvin’s
ω
Z  
∂K ∂ω
Z
Theorem
= − ∇ × (v × ω ) ·dA = ν∇2ω ·dA
∂t A ∂t
A basic result here is Kelvin’s theorem: in an invis- (4.14)
cid, barotropic flow with conservative body forces, the In expanding out the K integral, we have allowed for
circulation K around a closed curve, moving with the intrinsic changes of ω with time, and also for motion
fluid, is a constant of the motion. This means that vor- of the boundary of the surface, l. But now the result
tex lines can be thought of as “frozen into” the fluid: is clear: if ν → 0, K → constant. Furthermore, a
if you stretch or compress a piece of fluid, any vortex finite viscosity will lead to eventual dissipation of the
tube running through it must also be stretched or com- circulation.
presses.
This is worth two slightly different proofs, as fol- The Helmholtz theorems are related; they can be de-
lows. rived from the same assumptions that we used for
Kelvin’ theorem. The theorems are:
2
If we want to eliminate the nearly incompressible assumption,
• Vortex lines move with the fluid.
we simply need to start with (2.37) and carry out the full vector
curl, to go from v to ω ; students with time on their hands are • The strength of a vortex tube – the circulation – is
welcome to do this. constant along its length.
 23

• A vortex tube cannot end within the fluid. It must

either end at a solid boundary, or form a closed loop (a
vortex ring).
• The strength of a vortex tube is constant in time.

D. The Magnus Effect

Let’s return to potential flow. Consider, again, flow past
a cylinder, but this time let the cylinder rotate. This
situation leads to a net force on the cylinder.
To be specific, consider a potential, and stream func-
tion, given by

K K
φ= θ; ψ=− ln r (4.15)
2π 2π
Figure 4.1. Potential flow past a circular cylinder for
which recovers the velocities vr = 0 ; vθ = different values of the circulation Γ (K in our notation).
K/2πr and streamlines r = constant, as we expect. Note the location of the stagnation points S depends on the
(Note, the r term in the log in the expression for ψ must ratio of circulation to flow speed U . From Kundu figure 6.14
contain a scaling parameter to make it dimensionless;
this will not matter as we only take derivatives of ψ).
Thus, this potential-stream pair describes counterclock- symmetric above and below the cylinder; the flow sim-
wise flow (if K > 0) about the origin. Or, put in a ply moves faster “over the top”. Figure 4.1 illustrates
minus sign to make clockwise flow, and add this to our various possibilities. At low K values, there are two
previous solution for irrotational flow around a cylin- stagnation points, at the surface of the cylinder. At high
der: enough K, there is one stagnation point, well below
the cylinder. The streamline passing through this stag-
a2 nation point contains a region of fluid which remains
 
K
φ=U r+ cos θ + θ; separated from the outer flows; it simply circulates and
r 2π
(4.16) never reaches large distances away.
a2
 
K
ψ=U r− sin θ − ln r And now: there is a net surface pressure. We can
r 2π
find it as usual from Bernoulli, and show that it con-
We recall a is a length scale, corresponding to the ra- tributes a net lift:
dius of the cylinder. This potential/stream pair has the Z 2π
velocity field, L=− p(a, θ) sin θadθ = ρU K (4.18)
0

a2
 
We can remember the direction of the force by noting
vr = U 1 − 2 cos θ
r that the lift L in this example is upwards. If we define
(4.17) the direction of the circulation K as along the axis of
a2
 
K
vθ = −U 1 + 2 sin θ + the cylinder, in the sense give by the right hand rule,
r 2πr
then the direction of the lift is L ∝ U × K. This allows
H a reasonable analogy with basic EM: a current I in an
R this velocity field has a net circulation: v ·
Note that
external magnetic field B feels a force F ∝ I × B.
dl = vφ rdφ = K, if the integral is taken around any
closed curve that encloses the r = 0 axis. It turns out that this result – L = ρU K – holds for
Now, at the surface of the cylinder, vr = 0 and any two-dimensional shape, not just a true cylinder. A
rotating body feels a lateral force: this is the Magnus
K effect. Why is this different from the potential flow
vθ = −2U sin θ − solutions of chapter 3? One answer would simply be,
2πa
“Bernoulli”; the asymmetric velocity above and below
so that the sign of K – the intrinsic circulation associ- the rotating object leads to an asymmetric pressure, thus
ated with the r ≤ a region – determines whether the a net lateral force. But one could also ask how a rotat-
vθ flow outside agrees with K or with the ambient flow ing body develops the circulation which is needed for
U . We also see from (4.17) that the flow is no longer the Magnus effect. The answer here is, again, viscous
 24

effects at the boundary of the cylinder. Any rotating

body must have a no-slip surface; this will generate a
small layer of fluid rotating with the body.

E. Vortex lines and their behavior

Vortex lines demonstrate very interesting behavior. In
these notes we only summarize; the literature is exten-
sive here.
• Free vortex lines are linear (but not necessarily
straight) structures, with a net circulation K, which is
constant along the line and also a constant of the motion
(from Kelvin’s theorem). By Kelvin’s theorem, a loop
in the fluid which at one point encloses the core (r → 0)
will have a finite circulation K; it must retain that K
value no matter how the fluid is distorted in later flow. Figure 4.3. A line vortex (A) near a wall, and its image
Thus, a vortex line must move with the fluid; the fluid (B). From Kundu figure 5.13
is effectively attached to the vortex, particularly to its
core (about which the line integral defining K is finite). direction perpendicular to their separation. If the two
• A vortex sheet is made up of an infinite number of lines have parallel K values, they will circle around
vortex filaments placed side by side, with all filaments each other – or twist into a helix if one end of each is
rotating in the same sense. This sheet generates a tan- somehow tied down.
gential velocity discontinuity, as in Figure 4.2. Such
structures are important in understanding the flow over
aircraft wings.

Figure 4.2. An idealized vortex sheet. From Kundu

figure 5.15

• The behavior of vortex lines can be very in- Figure 4.4. A vortex ring approching a wall. From
triguing. They move under their own power. A vor- Kundu figure 5.14
tex line will move in a transverse flow, in a direction
U × K. (Recall the Magnus effect). A bent vortex line • Vortex rings are formed when a vortex line con-
will move in a direction K × n̂, if n̂ is the direction nects back on itself. A vortex ring will move itself along
of its radius of curvature R; Batchelor shows its speed through the background fluid, at a speed ∼ K/4πb if
∼ K/4π. b is the radius of the ring. (Any two opposite points
• Images are useful in calculating the behavior of around the ring act like paired, anti-parallel K lines.
vortex lines near boundaries (remember your images Alternatively, this is a case of well-defined curvature of
in electrostatics?). A wall, such as in Figure 4.3, is a a single vortex line.) One vortex ring will spread out as
streamline, and must have zero normal velocity. This it approaches a wall. Two rings will move around and
means we can add an imaginary image vortex, behind through each other – Faber shows this in his figures.
the wall and paired with the real one, to understand the • Vortex lines can reconnect. Viscosity breaks the
motion of the vortex line. conservation of circulation, and allows adjacent vor-
• Paired vortices interact. Two nearby vortex tex lines or rings to reconnect, thereby changing their
lines, with opposite K values, will drift together in a topology.
 25

F. Generation of Vorticity Saffman (quite mathematical), and several useful ar-

ticles in Annual Review of Fluid Mechanics over the
We argued above that circulation is conserved; this was years.
Kelvin’s theorem. How, then, can a fluid which starts ir-
rotational, ever gain vorticity? The answer, once again,
is the combination of viscosity and boundary effects.
To illustrate, I combine a figure from Kundu with the
discussion from Batchelor.

Figure 4.5. Top view of a vortex pair generated by

moving the handle of a spoon through a cup of coffee. From
Kundu figure 5.12

Start with a quiescent cup of coffee, turn your

spoon upside down, and move its handle perpen-
dicular to itself through the cup. Vortices will be
generated, as shown. Why? Even though there
is a formal no-slip condition at the surface of the
handle, there must be a finite tangential velocity
somewhere close to the handle. In this problem,
it’s at the edges. As Batchelor says, “the motion
that would be generated from rest in the absence
of diffusion of vorticity across the boundary . .
.is accompanied by a nonzero tangential relative
velocity at the boundary. Since the no-slip con-
dition requries the tangential component of rela-
tive velocity to be zero at each point of the solid
boundary, however small the viscosity may be, the
vorticity in the flow is infinite at the boundary.”
Thus: the need to go from a no-slip boundary to
a tangential slip in the coffee, some tiny distance
away, generates vorticity local to the handle; dif-
fusion (as per 4.6) takes care of the rest.

References
Vorticity is a large, interesting and complex topic.
I’ve mostly followed Faber and Tritton (for the words)
and Kundu (for the math); but there’s also good ma-
terial in Shivamoggi, not to mention a whole book by