CHAPTER 8

Three-Dimensional Small-Disturbance
Solutions

In this chapter, three-dimensional small-disturbance solutions will be derived for

some simple cases such as the large aspect ratio wing, the slender pointed wing, and the
slender cylindrical body. Flow problems requiring more detailed geometries will be treated
in the forthcoming chapters.

8.1 Finite Wing: The Lifting Line Model

The three-dimensional lifting wing problem was formulated in Chapter 4, and it is
clear that obtaining an analytic solution of the integral equations is difficult. However, it is
possible to approximate the lifting properties of a wing by a single lifting line, an approx-
imation that will allow a closed-form solution. In spite of the considerable simplifications
in this model it captures the basic features of three-dimensional lifting flows and predicts
the reduction of lift slope and the increase in induced drag with decreasing aspect ratio.

8.1.1 Definition of the Problem

Consider a lifting, thin, finite wing (described in Section 4.5 and shown in Fig. 8.1)
that is moving at a constant speed in an otherwise undisturbed fluid. The free stream of speed
Q ∞ has a small angle of attack α, relative to the coordinate system attached to the wing.
The velocity field for this potential flow problem can be obtained by solving Laplace’s
equation for the perturbation potential :
∇ 2 = 0 (8.1)
Following Section 4.5, the boundary condition requiring no flow across the wing solid
surface will be approximated at z = 0, for the case of small angle of attack, by
 
∂ ∂η
(x, y, 0±) = Q ∞ −α (8.2)
∂z ∂x
where η = ηc (x, y) is the camber surface placed on the x–y plane and for simplicity the
subscript c is omitted in this chapter. For modeling the lifting surface, a vortex distribution
is selected (as formulated in Section 4.5). The unknown vortex distribution γx (x, y) and
γ y (x, y) (shown in Fig. 4.9) is placed on the wing’s projected area at the z = 0 plane. The
resulting integral equation is
  
−1 γ y (x − x0 ) − γx (y − y0 ) ∂η
d x 0 dy0 = Q ∞ − α (8.3)
4π wing+wake [(x − x0 )2 + (y − y0 )2 ]3/2 ∂x
which should hold for any point on the wing. A proper (and unique) solution for the vortex
distribution will have to fulfill the Kutta condition along the trailing edge, such that the
vorticity component parallel to the trailing edge (γT.E. ) is zero:
γT.E. = 0 (8.4)
167
168 8 / Three-Dimensional Small-Disturbance Solutions

Figure 8.1 Far field horseshoe model of a finite wing.

Also, since vortex lines do not begin or end in a fluid (Eq. (4.64)), the solution must comply
with
   
 ∂γx   ∂γ y 
 =  (8.5)
 ∂x   ∂y 

8.1.2 The Lifting-Line Model

The simplest model that can be suggested to solve this problem is where the
chordwise circulation, at any spanwise station, is replaced by a single concentrated vortex.
Also, these local vortices of circulation (y) will be placed along a single spanwise line.
Based on the results of Section 5.5 for the two-dimensional lumped-vortex element, this
vortex line will be placed at the wing’s quarter-chord line along the span, −b/2 < y < b/2
(this bound vortex line is assumed to be straight and parallel to the y axis, as shown in
Fig. 8.1). The above positioning of the vortex line at the wing’s quarter-chord line effectively
satisfies the Kutta condition of Eq. (8.4) as was shown in Section 5.5.
At this point, attention needs to be focused on the vortex theorems requiring that a
vortex line cannot start or end abruptly in a fluid (or Eq. (8.5)). Therefore, if any change
of the vortex line strength d(y)/dy is introduced, it must be followed by introducing a
similar vorticity component in the other direction γx . In other words, the vortex line does
not terminate at this point but it changes direction, and its strength remains constant.
The most physical application of these principles is to “shed” these trailing vortices into
the flow and create a “wake” such that there will be no force acting on these free vortices.
Following Section 4.7, this requirement reduces to the condition that the flow along this
segment must be parallel to Γ (where positive Γ is according to the right-hand rule):

q × Γwake = 0 (8.6)

The most basic element that will fit these requirements will have the shape of a “horseshoe”
vortex (Fig. 8.1), which will have constant “bound vorticity”  along its quarter-chord line,
will turn backward at the wing tips and will continue far behind the wing, and eventually
will be closed by the starting vortex. It is assumed here that the flow is steady and therefore
the starting vortex is far downstream and its influence can be neglected.
8.1 Finite Wing: The Lifting Line Model 169

Figure 8.2 Lifting-line model consisting of horseshoe vortices. The bound vortex segment of all
vortices is placed on the y axis.

A more refined model of the finite wing was first proposed by the German scientist
Ludwig Prandtl (see Ref. 5.2) during World War 1 and it uses a large number of such
spanwise horseshoe vortices, as illustrated by Fig. 8.2 (the following analysis is in the spirit
of this early model). The straight bound vortex (y) in this case is placed along the y axis
and at each spanwise station the leading edge is one-quarter chord ahead of this line and
the local trailing edge is three-quarter chord behind the vortex line. Now, let us examine the
integral equation (Eq. (8.3)) for the case of the flat lifting surface, where ∂η/∂ x = 0. The
equation now simply states the boundary condition of Eq. (8.2):
∂wing ∂wake
+ + Q∞α = 0 (8.2a)
∂z ∂z
That is, the sum of the normal velocity components induced by the wing wb = ∂wing /∂z
and wake vortices wi = ∂wake /∂z, plus the normal velocity component of the free-stream
flow Q ∞ α, must be zero on the wing’s solid boundary:

wb + wi + Q ∞ α = 0 (8.7)

where w is considered to be positive in the +z direction. The subscripts ( )b and ( )i stand

for bound (on wing) and induced (by wake) influences, respectively.
The velocity component wb induced by the lifting line on the section with a chord
c(y) can be estimated by using the lumped-vortex model with the downwash calculated at
the collocation point located at the 3/4 chord. Consider the spanwise component (−y0 ≤
y ≤ y0 ) of a typical horseshoe vortex in Fig. 8.3 with strength (y0 ) (where (y0 ) =
−(d(y0 )/dy)dy0 ). The downwash wb at the collocation point (c/2, y) due to this segment
is given by Eq. (2.69) (see Fig. 2.15, which defines the angles β in this formula) as
− 
wb = (cos β1 − cos β2 )
4πd
 
−  y + y0 y0 − y
=  +
4π [c(y)/2] (c/2)2 + (y + y0 )2 (c/2)2 + (y0 − y)2
170 8 / Three-Dimensional Small-Disturbance Solutions

Figure 8.3 Velocity induced by the segments of a typical horseshoe element.

For a wing of large aspect ratio, we can neglect (c/2)2 in the square root terms to get

− (y0 )
wb = [1 + 1]
4π[c(y)/2]

The result for the complete lifting line (evaluated at y) is obtained by summing the results
for all the horseshoe vortices and is

−(y)
wb = (8.8)
2π [c(y)/2]

Note that this is identical to the result obtained by applying a locally two-dimensional
lumped-vortex model at each spanwise station, where the downwash wb is measured at the
3/4 chord due to a vortex (y) placed at the 1/4 chord position (see inset in the left-hand
side of Fig. 8.2).
Next, the downwash due to the wing trailing vortices must be evaluated. Since a change
in the spanwise circulation (y) is allowed, and since no vortex can begin or end in the flow,
the local change in this circulation is shed into the wake. Thus, the wake is now constructed
from semi-infinite vortex lines with the strength of (d/dy)dy (Fig. 8.2). Before we proceed
with the solution, the velocity induced by a single, semi-infinite trailing vortex line with
a strength of  = −(d(y0 )/dy)dy0 is evaluated (note that for positive  on the +y
side of the wing, negative d/dy is needed). The right-hand-side wake vortex line is
located at a spanwise location y0 , as shown in Fig. 8.3, and the downwash induced by this
vortex at the collocation point (c/2, y) is given by the result for a semi-infinite vortex line
from Eq. (2.71). Since for a large aspect ratio wing β1 ≈ π/2, β2 ≈ π (in Fig. 2.15) and
therefore

(y0 ) 1
w(y) = (8.9)
4π (y − y0 )

which is exactly one half of the velocity induced by an infinite (two-dimensional) vortex of
strength (y0 ). With the aid of this equation the normal velocity component induced by
8.1 Finite Wing: The Lifting Line Model 171

Figure 8.4 Two-dimensional section (in the y = const. plane) of a three-dimensional wing. The angle
of attack α is reduced by the induced downwash of the trailing vortices by αi .

the trailing vortices of the wing becomes

 b/2
1 [−d(y0 )/dy] dy0
wi = (8.10)
4π −b/2 y − y0
(Note that since the trailing vortices are assumed to lie in the z = 0 plane, their induced
spanwise velocity component is zero from the Biot–Savart law, Eq. (2.68b).) Assuming that
the wing aspect ratio is large (b/c(y)  1) has allowed us to treat a spanwise station as a
two-dimensional section and to transfer the boundary condition to the local three-quarter
chord. Substituting Eqs. (8.8) and (8.10) into Eq. (8.7) yields
 b/2
−(y) 1 [d(y0 )/dy] dy0
− + Q∞α = 0 (8.11)
2π [c(y)/2] 4π −b/2 y − y0
Dividing Eq. (8.11) by the free-stream speed Q ∞ results in
 b/2
−(y) 1 [d(y0 )/dy] dy0
− +α =0 (8.11a)
π c(y)Q ∞ 4π Q ∞ −b/2 y − y0
This is the Prandtl lifting-line integrodifferential equation for the spanwise load distri-
bution (y). This equation can be viewed as a combination of the angles (as shown in
Fig. 8.4):
−αe − αi + α = 0 (8.12)
where the induced downwash angle is (note that w is positive in the positive z direction)
−wi
αi ≈ (8.13)
Q∞
Equation (8.12) can be rearranged as
αe = α − αi (8.12a)
This means that in the case of the finite wing the effective angle of attack of a wing section
αe (the angle between the modified free-stream velocity q in Fig. 8.4 and the chord) is
smaller than the actual geometric angle of attack α by αi , which is a result of the downwash
induced by the wake.
172 8 / Three-Dimensional Small-Disturbance Solutions

It is possible to generalize the result of this equation by assuming that the two-dimensional
section has a local lift slope of m 0 and its local effective angle of attack is αe . Now, if camber
effects are to be accounted for too, then this angle is measured from the zero-lift angle of
the section, such that
ρ Q ∞ (y)
Cl (y) = = m 0 (y)αe (y) (8.14)
(1/2)ρ Q 2∞ c(y)
Consequently, Eq. (8.12a) becomes
αe = α − αi − α L0 (8.15)
where α L0 is the angle of zero lift due to the section camber (for cambered airfoils, usually
α L0 is a negative number). A more general form of Eq. (8.11a) that allows for section
camber and wing twist α(y) is now
 b/2
−2(y) 1 [d(y0 )/dy] dy0
− + α(y) − α L0 (y) = 0 (8.16)
m 0 (y)c(y)Q ∞ 4π Q ∞ −b/2 y − y0
In this equation α(y) is the local angle of attack relative to Q∞ and α L0 (y) is the airfoil
section zero-lift angle. If it is assumed that these geometrical quantities are known, then
(y) becomes the unknown in this equation. Also, at the wing tips the pressure difference
[or the lift ρ Q ∞ (y = ±b/2)] must reduce to zero, that is,
 
b
 y=± =0 (8.17)
2

8.1.3 The Aerodynamic Loads

The solution of Eq. (8.16) will provide the spanwise bound circulation distribution
(y). To obtain the aerodynamic forces, the two-dimensional Kutta–Joukowski theorem
will be applied (in the y = const. plane). However, because of the wake-induced velocity,
the free-stream vector will be rotated by αi (y), as shown in Fig. 8.5. This angle can be
calculated for a known (y) by using Eqs. (8.10) and (8.13):
 b/2
1 [d(y0 )/dy] dy0
αi = (8.18)
4π Q ∞ −b/2 y − y0
If we assume that αi is small, then cos αi ≈ 1 and sin αi ≈ αi , and the lift of the wing is
given by an integration of the local two-dimensional lift (see Eq. (3.113)) as

Figure 8.5 Tilting of the local (of section y = const.) lift vector by the angle αi induced by the trailing
vortices.
8.1 Finite Wing: The Lifting Line Model 173

 b/2
L = ρ Q∞ (y) dy (8.19)
−b/2

while the drag force, which is created by turning the two-dimensional lift vector by the
wake-induced flow, becomes
 b/2
Di = ρ Q ∞ αi (y)(y) dy (8.20)
−b/2

This drag is called induced drag because it is induced by the trailing vortices. This finite
wing’s drag is directly related to the lift and will diminish as the wingspan increases (b → ∞).
Equation (8.20) can also be rewritten in terms of the wake-induced downwash wi :
 b/2
Di = −ρ wi (y)(y) dy (8.20a)
−b/2

From the engineering point of view, the total drag D of a wing includes the induced drag
Di and the viscous drag D0 :

D = Di + D0

For example, the two-dimensional viscous drag of a NACA 0009 section is presented in
Fig. 5.20.

8.1.4 The Elliptic Lift Distribution

The spanwise circulation distribution (y) for a given planform shape can be
obtained by solving Eq. (8.16). In the particular case of an elliptic distribution of the
circulation, the solution becomes rather simple since the downwash wi becomes constant
along the wing span. Also, as will be shown later, wings having such a spanwise distribution
will have minimum induced drag. The proposed distribution of (y) is shown in Fig. 8.6
and is
 
y 2 1/2
(y) = max 1 − (8.21)
b/2

This must be substituted into Eq. (8.16) so that the constant max can be evaluated. For
simplicity, let us first calculate the downwash integral (second term in Eq. (8.16)). The term

Figure 8.6 Elliptic spanwise distribution of the circulation (y).

174 8 / Three-Dimensional Small-Disturbance Solutions

d(y)/dy is evaluated by differentiating Eq. (8.21):

   
d(y) max y 2 −1/2 4
= 1− −2 2 y
dy 2 b/2 b
and the downwash wi is obtained by substituting this result into Eq. (8.10):
  
max b/2 y0 2 −1/2 y0
wi (y) = 1− dy0 (8.22)
πb2 −b/2 b/2 y − y0

Note that when y = y0 , this integral is singular and therefore must be evaluated based on
Cauchy’s principal value. It is possible to arrive at Glauert’s integral (Eq. (5.22)) by the
transformation
b
y= cos θ (8.23)
2
b
dy = − sin θ dθ (8.23a)
2
and at the wingtips y = −b/2, θ = π and at y = b/2, θ = 0. This reduces Eq. (8.21) to

(θ ) = max [1 − cos2 θ ]1/2 = max sin θ (8.21a)

Substituting Eq. (8.23) into Eq. (8.22) yields


max 0 (b/2) cos θ0 [(−b/2) sin θ0 ]dθ0
wi = [1 − cos2 θ0 ]−1/2
π b2 π (b/2)(cos θ − cos θ0 )
The principal value of this integral can be obtained by using the Glauert integral, Eq. (5.22):

−max π cos θ0 dθ0 −max π sin θ
wi = =
2πb 0 (cos θ0 − cos θ ) 2π b sin θ
Consequently, wi and αi become
max
wi = − (8.24)
2b
max
αi = (8.24a)
2bQ ∞
and are constant along the wing span.
Another feature of the elliptic distribution is that the spanwise integral is simply half the
area of an ellipse (with semi-axes max and b/2):
 b/2
π b πb
(y) dy = max = max (8.25)
−b/2 2 2 4
Consequently, the lift and the drag of the wing can be evaluated as
 b/2
πb
L = ρ Q∞ (y) dy = ρ Q ∞ max (8.26)
−b/2 4
 b/2  
max π b π
Di = ρ Q ∞ αi (y) dy = αi L = ρ Q ∞ max = ρmax
2
(8.27)
−b/2 2bQ ∞ 4 8
8.1 Finite Wing: The Lifting Line Model 175

The lift and drag coefficients become

L π b max
CL ≡ 2
= (8.28)
(1/2)ρ Q ∞ S 2 S Q∞
Di π max
2
1 S 2
C Di ≡ = = C (8.29)
2
(1/2)ρ Q ∞ S 4S Q 2∞ π b2 L
Substituting the spanwise downwash (Eq. (8.24)) and the elliptic circulation distribution
(Eq. (8.21)) into Eq. (8.16) yields
 
−2max y 2 1/2 max
1− − + α(y) − α L0 (y) = 0 (8.30)
m 0 (y)c(y)Q ∞ b/2 2bQ ∞
This equation provides the relation between the local chord c(y) and the local angle of attack
α(y) for the wing with the elliptic circulation distribution. If the chord c(y) has an elliptic
form, too, the constant max can be easily evaluated. Thus, assume
 
y 2 1/2
c(y) = c0 1 − (8.31)
b/2
where c0 is the root chord. Substituting Eq. (8.31) into Eq. (8.30) cancels the elliptic variation
and we have
−2max max
− + α(y) − α L0 (y) = 0 (8.32)
m 0 (y)c0 Q ∞ 2bQ ∞
For an elliptic planform with constant airfoil shape, all terms but α(y) in this equation are
constant, and therefore this wing with an elliptic planform and load distribution is untwisted
(α(y) = α = const.). The value of max is then
2bQ ∞ (α − α L0 )
max = (8.33)
1 + 4b/m 0 c0
The area S of the elliptic wing is
c0 b
S=π (8.34)
2 2
Also, it is common to define the wing aspect ratio as
b2
≡ (8.35)
S
Using the and the area S for the elliptic wing and substituting into Eq. (8.33), we obtain
2bQ ∞ (α − α L0 )
max = (8.33a)
1 + π /m 0
With this expression for max and using m 0 = 2π , the lift coefficient (Eq. (8.28)) becomes
2π
CL = (α − α L0 ) ≡ C L α (α − α L0 ) (8.36)
1 + 2/
Here C L α is the three-dimensional wing lift slope and the most important conclusion of this
analysis is that this slope becomes less as the wingspan becomes smaller due to the induced
downwash. This is illustrated by Fig. 8.7, where for a wing with given α L0 the effective
176 8 / Three-Dimensional Small-Disturbance Solutions

Figure 8.7 The reduction of lift slope for three-dimensional wings.

angle of attack is reduced by αi according to Eq. (8.15). Consequently, for finite span wings,
more incidence is needed to achieve the same lift coefficient as the wingspan decreases.
The induced drag coefficient is obtained by substituting Eq. (8.35) into Eq. (8.29):
1
C Di = C L2 (8.37)
π
which indicates that as the wing aspect ratio increases the induced drag becomes smaller.
Also, the induced drag for the finite elliptic wing will increase with a rate of C L2 as shown
in Fig. 8.8.
The lift slope C L α versus for the elliptic wing (Eq. (8.36)) is shown in Fig. 8.9. The lift
slope of a two-dimensional wing is the largest (2π) and as the wingspan becomes smaller
C L α decreases too.
The spanwise loading L (y) (lift per unit span) of the elliptic wing is obtained by using
the Kutta–Joukowski theorem:
 
y 2 1/2
L (y) = ρ Q ∞ (y) = ρ Q ∞ max 1 − (8.38)
b/2
Figure 8.10 shows an elliptic planform and the spanwise lift distribution L (y) that is elliptic
too. As Eq. (8.24) indicated, the downwash of such a wing is constant, and, combined with

Figure 8.8 Lift polar for an elliptic wing.

8.1 Finite Wing: The Lifting Line Model 177

Figure 8.9 Variation of lift coefficient slope versus aspect ratio for thin elliptic wings.

the velocity induced by the bound vortex wb , must be equal to the upwash of the free-stream
Q ∞ α so that the combined normal velocity component is zero, according to Eq. (8.7). (Note
that it is possible to have elliptic loading with other than an elliptic planform, but in that
case, local twist or camber needs to be adjusted so that wi will remain constant.)
The section lift coefficient Cl is defined by using the local chord from Eq. (8.31) and is

L (y) 2max
Cl ≡ = = CL
(1/2)ρ Q 2∞ c(y) c0 Q ∞

Figure 8.10 Chord and load distribution for a thin elliptic wing. Note that the induced downwash is
constant and combined with the downwash of the bound vortex is equal to the free-stream upwash,
resulting in zero velocity normal to the wing surface (Eq. (8.7)).
178 8 / Three-Dimensional Small-Disturbance Solutions

Figure 8.11 Schematic description of the pressure difference and wake vortex distribution of a thin
elliptic wing.

Thus, for the elliptic wing, both section lift coefficient and wing lift coefficient are the same,
that is,
2π
Cl = (α − α L0 ) ≡ Clα (α − α L0 ) (8.39)
1 + 2/
Similarly, the section induced drag coefficient is
L (y)αi 1 S 2
Cdi ≡ = C = C Di (8.40)
(1/2)ρ Q 2∞ c(y) π b2 l
The strength of the circulation in the wake is simply the spanwise derivative of (y) (see
Eq. (8.21)),
d(y) 4max y
=− 2  (8.41)
dy b [1 − (y/(b/2))2 ]
This spanwise wake vortex strength is shown schematically in Fig. 8.11. It is clear from this
figure that near the wingtips, where |(y)/dy| is the largest, the wake vortex will be the
strongest. Therefore, owing to the induced velocity at the wake it will roll up, mostly near
the wingtips, to form two concentrated trailing vortices as shown by the flow visualization
in Fig. 8.12. The induced effect of this wake rollup on (y) is assumed to be negligible in
this model; but this effect can be investigated by the numerical methods of later chapters.

8.1.5 General Spanwise Circulation Distribution

A more general solution for the spanwise circulation (y) in Eq. (8.16) can be
obtained by describing the unknown distribution in terms of a trigonometric expansion.
Using the spanwise coordinate θ , as defined in Eq. (8.23), we select the following Fourier
expansion:
∞
(θ) = 2bQ ∞ An sin nθ (8.42)
n=1

The shapes of the first three symmetric terms in this expansion are shown schematically in