1 Introduction to Incompressible Potential Flow

This lecture is not meant to be an exhaustive dive into potential flow; it is only
meant as an introduction to incompressible potential flow and how we can use
potential flow theory to get simplified solutions to some otherwise complicated
problems.
When solving complex problems in aerodynamics, we can start with the most
general equations we know, for instance, the Navier-Stokes equations (which
themselves already have a couple of assumptions built in, but are not very
limiting). These equations require detailed numerical solutions, so we would
like to simplify them down by making some more valid assumptions. We will
be making two assumptions here:

• Irrotational flow (angular velocity/vorticity is zero)

• Incompressible flow (density is constant)

The first assumption results in what we call potential flow theory, while
the second assumption gives us the more specific incompressible potential flow,
which we can use to calculate, for example, the flow around an airfoil.

2 Assumption: Irrotational
Lets start with the assumption of irrotational flow. The vorticity is equal to
the curl of the velocity (∇ × V ~ ), which we can derive in another post at a later
date (or see Chapter 2.12 in Fundamentals of Aerodynamics by Anderson). For
irrotational flow, this is equal to zero at every point in the flow. Physically,
this means that the fluid elements have no angular velocity, and that they
only translate through space (though you can have equal and opposite angular
velocities of the fluid element’s connecting sides).
Since we now know for irrotational flow that ∇ × V ~ = 0, we need to find
a form of V ~ that always satisfies this equation. Note that V ~ = 0satisfies the
equation, but this is a trivial solution. Recall the vector identity that the curl of
the gradient of a scalar is always equal to zero, ∇×(∇ϕ) = 0. In this equation we
are assuming that ϕ is a scalar. Let’s run through the cross product calculation
to see if this is true.

î ĵ k̂
∇ × ∇ϕ = ∂/∂x ∂/∂y ∂/∂z (1)
∂ϕ/∂x ∂ϕ/∂y ∂ϕ/∂z
 
∂ϕ ∂ϕ
∇ × ∇ϕ = − î
∂y∂z ∂z∂y
 
∂ϕ ∂ϕ
− − ĵ (2)
∂x∂z ∂z∂x
 
∂ϕ ∂ϕ
+ − k̂ = 0
∂x∂y ∂y∂x
From the symmetry of the second derivatives (∂/∂x∂y = ∂/∂y∂x), which will
be true for most cases you ever encounter, we can see that the above equation is

1
indeed equal to zero when ϕ is a scalar field. Comparing the two main equations,
we can see that the velocity vector is equal to the gradient of the scalar field.

~ = ∇ϕ
V (3)
The significance of this equation is that instead of solving three equations for
the three unknown components of velocity (u, v, w), we can solve one equation
for the scalar ϕ, and then find the velocity components by taking the gradient
of ϕ.

3 Assumption: Incompressible
But what equation will we solve to find ϕ? If we don’t use our second assumption
yet, we can manipulate some conservation equations to get the compressible
potential flow equation. We would like to go even simpler though. Now it’s
time to make our second assumption, namely that the flow is incompressible.
This physically means that the density is constant throughout the flow (not
changing with time or space). Let’s write the mass conservation equation (also
called the continuity equation).
∂ρ  
~ =0
+ ∇ · ρV (4)
∂t
For incompressible flow, the time derivative term disappears, and the density
can be taken out of the dot product and divided out of the equation. The
incompressible mass conservation equation simplifies down to the following.

~ =0
∇·V (5)
Plugging in the velocity vector from Eq. 3, we get the following.

∇ · (∇ϕ) = 0 (6)
Simplifying this leads to the well-known Laplace equation. This is the equa-
tion we will be solving in order to find ϕ.

∇2 ϕ = 0 (7)
After finding ϕ, we can find the velocity components using Eq. 3.

~ = ∇ϕ
V so ~ = ∂ϕ î + ∂ϕ ĵ + ∂ϕ k̂
V (8)
∂x ∂y ∂z
We can also define the x, y, and z velocity components in separate equations.
∂ϕ ∂ϕ ∂ϕ
u= v= w= (9)
∂x ∂y ∂z
Because Laplace’s equation is linear (and even though it is second-order),
we can use the principle of superposition, which says that the sum of solutions
of Laplace’s equation is also a solution of Laplace’s equation.

ϕ = ϕ1 + ϕ2 + ϕ3 + · · · + ϕn (10)

2
 NAME VELOCITY POTENTIAL

Uniform Flow ϕ = V∞ cos (α) x + V∞ sin (α) y

Λ
Source Flow ϕ= 2π ln (r)

−Γ
Vortex Flow ϕ= 2π θ

There are many elementary flow solutions that provide simple analytical
solutions to Laplace’s equation. By combining the solutions of these elementary
flows (using superposition), we can construct more complicated flows, such as
the flow around an airfoil. The solutions of three elementary flows (using the
velocity potential, not the stream function) can be seen in the table below.
Future lectures on potential flow elementary flows will discuss these in more
detail.

3
Elementary Flow: Uniform Flow
For this first (and simplest) flow, our discussion will appear a little circular.
That is, first we will specify the x and y velocity components of a uniform flow
(at an angle of attack, α), then we will determine the velocity potential (ϕu ),
and then we will find the velocity components from that velocity potential. The
reason is just to show that we can go forward or backwards. In the remaining
elementary potential flow cases, we will start with the velocity potential and
find the x and y velocity components.
Our uniform flow needs to be, well, uniform. That is, the velocity (magnitude
and orientation) needs to be the same at every point in the flow. This means
that we can specify the magnitude and orientation once, and it will be the same
for every (x, y) point in the flow. Let’s define the magnitude as V∞ , and the
orientation as α, which can take any value from 0 to 2π radians (0o to 360o ).
This means we can get the x and y components of the velocity vectors (see Fig.
1) using the following equations.

Vx = V∞ cos (α)
(1)
Vy = V∞ sin (α)

Figure 1: Uniform flow vector definition.

From the definition of the velocity potential, we know the following.

∂ϕu
Vx =
∂x
(2)
∂ϕu
Vy =
∂y
This means we can integrate the velocity we defined with respect to the
appropriate variables (dx for Vx and dy for Vy ).
Z
V∞ cos (α) dx → ϕu = V∞ cos (α) x + f (y)
Z (3)
V∞ sin (α) dy → ϕu = V∞ sin (α) y + g (x)

1
 We can find the velocity potential by noting that the first term in the first
equation is a function of x(g (x)), and that the first term in the second equation
is a function of y(f (y)).

ϕu = V∞ cos (α) x + V∞ sin (α) y (4)

Now we will come full circle and find the velocity components from the
velocity potential.
∂ϕu
Vx = = V∞ cos (α)
∂x
(5)
∂ϕu
Vy = = V∞ sin (α)
∂y
Note that these are the same as the velocity components we defined at the
beginning of this section, and that makes sense. If they were not the same,
then we would have known that we made a mistake in our integration when
solving for the velocity potential. We can see that the velocity components are
independent of x and y, which means they will always be the same over the
entire grid. Let’s put this to practice in the code below. The two knowns that
we are specifying are the velocity magnitude and velocity angle.
% Velocity magnitude and angle
Vinf = 1; % Velocity magnitude
alpha = 0; % Velocity vector angle [deg]
% Create the grid
numX = 10; % Number of X points
numY = 10; % Number of Y points
X = linspace(-10,10,numX)’; % Create X points array
Y = linspace(-10,10,numY)’; % Create Y points array
% Solve for velocities
Vx = zeros(numX,numY); % Initialize X velocity
Vy = zeros(numX,numY); % Initialize Y velocity
for i = 1:1:numX % Loop over X-points
for j = 1:1:numY % Loop over Y-points
Vx(i,j) = Vinf*cosd(alpha); % X velocity (Eq. 5)
Vy(i,j) = Vinf*sind(alpha); % Y velocity (Eq. 5)
end
end
% Plot the velocity on the grid
figure(1); % Create figure
cla; hold on; grid off; % Get ready for plotting
set(gcf,’Color’,’White’); % Set background to white
set(gca,’FontSize’,12); % Labels font size
quiver(X,Y,Vx,Vy,’r’); % Velocity vector arrows
xlim([min(X) max(X)]); % Set X-axis limits
ylim([min(Y) max(Y)]); % Set Y-axis limits
xlabel(’X Axis’); % Set X-axis label
ylabel(’Y Axis’); % Set Y-axis label
title(’Uniform Flow’); % Set title

2
 The results of the code can be seen in the plots in Figs. 2 and 3. In Fig. 2,
the angle of the velocity vector is zero (aligned with the x-axis), while in Fig.
3, the angle of the velocity vector is 30o . We can see in both plots that the
velocity vectors at every grid point are the same (uniform), and do not depend
on their x and y locations, which is what we expected.

Figure 2: Uniform flow with α = 0o

Figure 3: Uniform flow with α = 30o , and blue line for line integral to compute
circulation.

In Fig. 3, the blue line represents the curve that we are computing the
circulation along (see my circulation lecture). The line integral (and code)
from my circulation post is used to calculate the circulation, which is equal to
Γ = 7.702 × 10−16 . This is expected that the circulation is nearly zero, since
there is no vortex in the flow.

3
Elementary Flow: Source/Sink Flow
In case of uniform flow, we saw how we can use the velocity potential to find
the velocity component of the flow at discrete grid points. We will use this
methodology again here to find the velocity induced at discrete grid points from
a source/sink in the flow. The velocity of source/sink flow is given below (with
no derivation here).
Λ
ϕs = ln (r) (1)
2π
We can see that the velocity potential is a function of the radial distance
from the origin of the source (r), and so most textbooks will solve for the radial
and angular components of the velocity (Vr and Vθ ). We will stick with the
Cartesian coordinates here though, because it will show another method and it
will also be helpful in later derivations of the source panel method (SPM) and
vortex panel method (VPM). The source strength (uppercase lambda), Λ, is
constant here, so we can take it out of the derivative term.

∂ϕs Λ ∂ln (r) Λ ∂r

Vx = = =
∂x 2π ∂x 2πr ∂x
(2)
∂ϕs Λ ∂ln (r) Λ ∂r
Vy = = =
∂y 2π ∂y 2πr ∂y
The velocity induced at point P (XP , YP ) depends on the distance from the
sourse/sink origin (X0 , Y0 ), rP , which is defined below.
q
2 2
rP = (XP − X0 ) + (YP − Y0 ) (3)
Performing the derivative using the chain rule, we get the following expres-
sion for the partial derivatives.

∂rP 1h i−1/2
2 2
= (XP − X0 ) + (YP − Y0 ) (2) (XP − X0 )
∂x 2
(XP − X0 )
=
rP
i−1/2 (4)
∂rP 1h 2 2
= (XP − X0 ) + (YP − Y0 ) (2) (YP − Y0 )
∂y 2
(YP − Y0 )
=
rP
Plugging these expressions into the Vx and Vy equations from Eq. 2, we
obtain the following.

Λ (XP − X0 )
Vx = 2
2πrP
(5)
Λ (YP − Y0 )
Vy = 2
2πrP
We can code the source/sink flow in the same way we coded the uniform
flow. To create a source flow, the value of Λ must be positive, while for sink
flow Λ is negative.

1
% Source/sink knowns
lambda = 1; % Source/sink magnitude
X0 = 0; % Source/sink X origin
Y0 = 0; % Source/sink Y origin
% Create the grid
numX = 20; % Number of X points
numY = 20; % Number of Y points
X = linspace(-10,10,numX)’; % Create X points array
Y = linspace(-10,10,numY)’; % Create Y points array
[XX,YY] = meshgrid(X,Y); % Create the meshgrid
% Solve for velocities
Vx = zeros(numX,numY); % Initialize X velocity
Vy = zeros(numX,numY); % Initialize Y velocity
for i = 1:1:numX % Loop over X-points
for j = 1:1:numY % Loop over Y-points
x = XX(i,j); % X-value of current point
y = YY(i,j); % Y-value of current point
dx = x - X0; % X distance
dy = y - Y0; % Y distance
r = sqrt(dx^2 + dy^2); % Distance (Eq. 3)
Vx(i,j) = (lambda*dx)/(2*pi*r^2); % X velocity (Eq. 4)
Vy(i,j) = (lambda*dy)/(2*pi*r^2); % Y velocity (Eq. 4)

V(i,j) = sqrt(Vx(i,j)^2+Vy(i,j)^2); % Total velocity

Vr(i,j) = lambda/(2*pi*r); % Radial velocity
end
end
% Plot the velocity on the grid
figure(1); % Create figure
cla; hold on; grid off; % Get ready for plotting
set(gcf,’Color’,’White’); % Set background to white
set(gca,’FontSize’,12); % Change font size
quiver(X,Y,Vx,Vy,’r’); % Velocity vector arrows
xlim([min(X) max(X)]); % Set X-axis limits
ylim([min(Y) max(Y)]); % Set Y-axis limits
xlabel(’X Axis’); % Set X-axis label
ylabel(’Y Axis’); % Set Y-axis label
title(’Source/Sink Flow’); % Set title

Since the source and sink flow is usually described by the radial and angular
components, let’s compute their velocities below (Vr computed on line 27 in the
code).
∂ϕ Λ
Vr = =
∂r 2πr (6)
1 ∂ϕ
Vθ = =0
r ∂θ
Since the angular component of the velocity is zero, we can compare the
magnitude of the radial component to the magnitude of the velocity from the
Cartesian Vx and Vy components. In the code (line 26), the Cartesian velocity

2
magnitude is V . Comparing these two matrices, it is clear that they are the
same, which is a good check that our Cartesian velocity derivation is correct.
The results for source and sink flow can be seen in Fig. 1 and Fig. 2,
respectively. The velocity vector arrows either point directly out from the origin
(source) or directly into the origin (sink).

Figure 1: Source flow with a strength of Λ = 1. Blue line is curve along which
the circulation is calculated

Figure 2: Sink flow with a strength of Λ = −1

We can also now compute the circulation around the solid blue curve as
shown in Fig. 1. Again, the expected circulation is zero since there is no vortex
in the flow, and this agrees with the computed circulation of Λ = −2.11419 ×
10−18 . The same circulation can be computed for the sink in Fig. 2, and the
same results are obtained.

3
Combined Flow: Uniform and Source/Sink Flow
Now we can try to combine two elementary flows using the principle of super-
position. We will combine a uniform flow at zero angle of attack (α = 0) with a
source flow (Λ > 0). For brevity, only the loop where the velocity components
are computed for this combined flow are shown in the code. The Cartesian
velocity components are shown below.

Λ (XP − X0 )
Vx = V∞ cos (α) + 2
2πrP
(1)
Λ (YP − Y0 )
Vy = V∞ sin (α) + 2
2πrP

for i = 1:1:numX
for j = 1:1:numY
x = XX(i,j);
y = YY(i,j);
dx = x - X0;
dy = y - Y0;
r = sqrt(dx^2 + dy^2);
Vx(i,j) = Vinf*cosd(alpha) + ((lambda*dx)/(2*pi*r^2));
Vy(i,j) = Vinf*sind(alpha) + ((lambda*dy)/(2*pi*r^2));
end
end

The only difference between this code and the others we have looked at
is that Vx and Vy are equal to the sum of the uniform flow and source flow
solutions. The resulting flow field can be seen in Fig. 1. The circulation is
computed the same way as before, and results again in a value nearly zero
(Λ = 5.69315 × 10−16 ), which makes sense because we still have no vortex in
this flow (no source of circulation).

Figure 1: Uniform and source flow, with the blue line showing the curve used
for the circulation integral.

1
Elementary Flow: Vortex Flow
Finally, we arrive at the elementary vortex flow. The velocity potential of the
vortex is given below (without derivation).
Γ
ϕv = − θ (1)
2π
Instead of being a function of the radius, it is now a function of the angle
θ. The term (capital gamma) Γ is called the vortex strength, and is the same
thing as the circulation defined earlier. As with the source/sink flow, we will
stick with Cartesian coordinates because that’s what we’ve been using so far,
and also because it will be useful for the panel method derivations later in
this document. Similar to the source flow, we can find the x and y velocity
components by taking the appropriate derivatives of the velocity potential.
∂ϕv Γ ∂θ
Vx = =−
∂x 2π ∂x
(2)
∂ϕv Γ ∂θ
Vy = =−
∂y 2π ∂y
In order to take the derivative, we need to know what θ is. If we look at a
point on a circle at θ degrees from the positive x-axis, we can draw a line from
the origin to this point (see Fig. 1).

Figure 1: Variable definitions for an arbitrary point (x,y) away from the vortex
center (x0 ,y0 ).

Dropping a perpendicular line from the point to the x-axis makes a triangle,
with the hypotenuse being the radius of the circle. Using trigonometry, we can
find the value of θ.
dy
tan (θ) =
dx   (3)
dy
θ = tan−1
dx
The distances (dx and dy) to the point on the circle can be found knowing
the vortex center point (x0 , y0 ).

1
 dx = x − x0
(4)
dy = y − y0
The
p radius of the circle can also be found using Pythagorean’s theorem:
r = dx2 + dy 2 . Now we can perform the partial derivative of θ with respect
to both x and y.
  " # 
∂θ ∂ −1 dy 1 −dy
= tan = 2
∂x ∂x dx 1 + (dy/dx) dx2
  " #  (5)
∂θ ∂ −1 dy 1 1
= tan = 2
∂y ∂y dx 1 + (dy/dx) dx
The term in the square brackets can be simplified to the following, by mul-
tiplying by (dx2 /dx2 ), and noting that r2 = dx2 + dy 2 .
 
 dx2 dx2 dx2
 
1
= = (6)

 2 
dx2 dx2 + dy 2 r2

dy
1 + dx

Now we can write the velocity component equations by plugging in the above
expressions. Note that the dy term appears in the x velocity equation, and vice
versa.

−Γ dx2
    
−Γ ∂θ −dy −Γ −dy
Vx = = =
2π ∂x 2π r2 dx2 2π r2
Γdy
Vx =
2πr2
(7)
dx2
    
−Γ ∂θ −Γ 1 −Γ dx
Vy = = =
2π ∂y 2π r2 dx 2π r2
−Γdx
Vy =
2πr2
Similar to the source/sink flow, we can also easily compute the radial and
angular velocity components to compare these values to.
∂ϕ
Vr = =0
∂r   (8)
1 ∂ϕ 1 −Γ ∂θ Γ
Vθ = = =−
r ∂θ r 2π ∂θ 2πr
We can code the vortex flow in the same way we coded the previous ele-
mentary flows. To create a vortex flow, the value of Γ must be specified. A
positive value results in a clockwise vortex while a negative value results in a
counter-clockwise vortex.
% Vortex knowns
gamma = 1; % Vortex strength
X0 = 0; % Vortex X origin
Y0 = 0; % Vortex Y origin

2
% Create the grid
numX = 100; % Number of X points
numY = 100; % Number of Y points
X = linspace(-10,10,numX)’; % Create X points array
Y = linspace(-10,10,numY)’; % Create Y points array
[XX,YY] = meshgrid(X,Y); % Create the meshgrid
% Solve for velocities
Vx = zeros(numX,numY); % Initialize X velocity
Vy = zeros(numX,numY); % Initialize Y velocity
r = zeros(numX,numY); % Radius
for i = 1:1:numX % Loop over X-points
for j = 1:1:numY % Loop over Y-points
x = XX(i,j); % X-value of current point
y = YY(i,j); % Y-value of current point
dx = x - X0; % X distance
dy = y - Y0; % Y distance
r = sqrt(dx^2 + dy^2); % Distance
Vx(i,j) = (gamma*dy)/(2*pi*r^2); % X velocity (Eq. 7)
Vy(i,j) = (-gamma*dx)/(2*pi*r^2); % Y velocity (Eq. 7)

% Total, tangential, radial velocity (Eq. 8)

V(i,j) = sqrt(Vx(i,j)^2 + Vy(i,j)^2); % Total velocity
Vt(i,j) = -gamma/(2*pi*r); % Tangential velocity
Vr(i,j) = 0; % Radial velocity
end
end
% Plot the velocity on the grid
figure(1); % Create figure
cla; hold on; grid off; % Get ready for plotting
set(gcf,’Color’,’White’); % Set background to white
set(gca,’FontSize’,12); % Change font size
quiver(X,Y,Vx,Vy,’r’); % Velocity vector arrows
axis(’equal’); % Set axis to equal sizes
xlim([min(X) max(X)]); % Set X-axis limits
ylim([min(Y) max(Y)]); % Set Y-axis limits
xlabel(’X Axis’); % Set X-axis label
ylabel(’Y Axis’); % Set Y-axis label
title(’Vortex Flow Flow’); % Set title
The results for both a positive and negative vortex can be seen in Fig. 2 and
Fig. 3, respectively. Note that in Fig. 2, we specified a positive value for the
vortex strength, which results in the clockwise flow, while in Fig. 3, we specified
a negative value for the vortex strength which results in counter-clockwise flow.
The circulation is again computed in the normal way along the blue line in
Fig. 2. The result of the integration is a value that very closely matches the
vortex strength that we specified in the equations above. The correct sign is
also included. For the positive vortex, we specified a strength of 1, and the
computed circulation is Γ = 0.999357. For the negative vortex, we specified a
strength of -1, and the compute circulation is Γ = −0.999357. Note that for
these calculations, we have broken the curve into 100 segments. If less segments

3
Figure 2: Positive strength vortex of Γ = 1. Blue line is the curve used for
circulation calculation.

Figure 3: Negative strength vortex of Γ = −1.

are used, the results are less accurate (Γ = 0.997281 using 50 segments).

4
Combined Flow: Uniform and Vortex Flow
Just as we had used a combined uniform and source flow in a previous example,
we will now combine the uniform flow at zero angle of attack with vortex flow.
The main point of this post is to compute the circulation over a couple of closed
loops and see how they change depending on where they are in the flow-field.
The Cartesian velocity components can be seen in the equations below.
Γdy
Vx = V∞ cos (α) +
2πr2
(1)
−Γdx
Vy = V∞ sin (α) +
2πr2
The code can be seen below.
% Vortex knowns
Vinf = 1; % Velocity
alpha = 0; % Angle of attack [deg]
gamma = 30; % Vortex strength
X0 = 0; % Vortex X origin
Y0 = 0; % Vortex Y origin
% Create the grid
numX = 50; % Number of X points
numY = 50; % Number of Y points
X = linspace(-10,10,numX)’; % Create X points array
Y = linspace(-10,10,numY)’; % Create Y points array
[XX,YY] = meshgrid(X,Y); % Create the meshgrid
% Solve for velocities
Vx = zeros(numX,numY); % Initialize X velocity
Vy = zeros(numX,numY); % Initialize Y velocity
r = zeros(numX,numY); % Radius
for i = 1:1:numX % Loop over X-points
for j = 1:1:numY % Loop over Y-points
x = XX(i,j); % X-value of current point
y = YY(i,j); % Y-value of current point
dx = x - X0; % X distance
dy = y - Y0; % Y distance
r = sqrt(dx^2 + dy^2); % Distance
Vx(i,j) = Vinf*cosd(alpha) + (gamma*dy)/(2*pi*r^2);
Vy(i,j) = Vinf*sind(alpha) + (-gamma*dx)/(2*pi*r^2);
end
end
Two plots are shown in Fig. 1 and Fig. 2. Figure 1 is for the combined
uniform and vortex flow when the vortex strength is Γ = 10, whereas Fig.
2 is for the combined uniform and vortex flow when the vortex strength is
Γ = 30. In both plots, the red arrows are the velocity vectors, the black lines
are streamlines, and the blue and magenta curves are used to compute the line
integral in the circulation equation.
We can see from this resulting circulation calculations for both plots that
when the vortex is enclosed in the loop, the circulation/vortex strength is accu-
rately predicted, but when the loop does not enclose the vortex, the resulting

1
Figure 1: Combined uniform and vortex flow, where the vortex strength is
Γ = 10. Blue and magenta contours are used for circulation line integral, with
respective values in the legend.

Figure 2: Combined uniform and vortex flow, where the vortex strength is
Γ = 30. Blue and magenta contours are used for circulation line integral, with
respective values in the legend.

circulation is zero. This will be important for our lift calculation for the airfoils
we will be doing in the vortex panel method (VPM). It also helps shed some
light on why the source panel method (SPM), which does not contain any vortex
flows in its calculation, does not result in any circulation, and thus no lift.