Q & A in ME2135

AY 2023/2024
Here I compile and log some selected questions asked by you and my answers. I will regularly update this
document when new questions are asked. The questions asked by previous students are attached also.

Q: When deriving the boundary layer equations, you seemed to talk

about the flow inside the boundary layer. But the boundary conditions
for these equations include u = U∞ at y = ∞. Why you can specify the
b.c. at y = ∞ for the equations you derived “inside the boundary layer”?
A: First, it may be not rigorous to emphasise the boundary layer equations derived are strictly only applicable to
the flow inside the boundary layer. When deriving these equations, the most important assumption is x  y. In the
literature, you will see people call this kind of equations (derived under x  y) the boundary layer equation, even if
the studied problem is not boundary layer.
For the boundary condition at y = ∞, here, this infinity should not be understood as the conventional infinity.
Because the flow outside the boundary layer is actually uniform and has the same profile U∞ along the y direction,
which means that the velocity just outside the boundary layer will have the same velocity in the far outer region.
So, the infinity here can be understood as the flow outside the boundary layer where the velocity profile is U∞ .

Q: When you were discussing the continuity equation, you expanded

∇ · (ρV) as V · ∇ρ + ρ∇ · V. Why it’s like this?
A: ∇ is the spatial derivative. Here, I used the product role of derivative. If this is difficult for you to understand,
you can use the index form, that is,
∂
∇ · (ρV) = (ρuj ) according to the definition of inner product, using uj for V
∂xj
∂ρ ∂uj
= uj +ρ product role of derivative
∂xj ∂xj
= V · ∇ρ + ρ∇ · V again, the definition of inner product (1)

Du
Q: Why in Lagrangian framework it’s Dt and in the Eulerian framework
it’s ∂u
∂t ?
A: I realised I didn’t make this point very clear in the class. In the schematic on the slide, I just drew one stickman
(red) for the Eulerian description. It doesn’t mean that we only observe the flow from that single point. In fact, if
you learn in the future computational fluid mechanics, you will know that a grid (or a mesh of grid points) is used
to numerically discritise the NS equations (see the red circle points in the illustration below). Figuratively each grid
point is an observer. At each point, you can observe the flow; if the velocity at a certain point is changing in time,
then you can calculate ∂u∂t . How do you calculate this in a numerical simulation or in an experiment? Let’s say, you
are sitting at the grid point 11. At time t = t1 , you record the velocity to be u1 and at time t2 , you record the
u2 −u1
velocity to be u2 (2 is not the power). Then, you can approximate ∂u ∂t = t2 −t1 at the point 11.
If you listened to me carefully, I also mentioned that in the Eulerian framework, you can also calculate the spatial
information, for example, ∂u ∂x . Now I will illustrate how to calculate this quantity at the point 12. We assume a
uniform grid with the x-distance between two adjacent points being δx. Say, at a certain instant, the value of u at
the grid point 13 is u13 and at the point 11 is u11 . Then, we can use these two values to approximate the spatial
u13 −u11
derivative ∂u ∂u
∂x at the point 12 using the equation ∂x 12 = 2δx . How did I get this equation? Taylor’s series. I

1
have
∂u
u13 = u12 + 12
δx + H.O.T. (2)
∂x
∂u
u11 = u12 − 12
δx + H.O.T. (3)
∂x
Neglecting the high-order terms (H.O.T.) and subtracting the second equation from the first one, we have
∂u ∂u u13 − u11
u13 − u11 = 2 12
δx, after rearranging, we have 12
= (4)
∂x ∂x 2δx
When δx is smaller, this approximation is getting better. Similarly, you can get ∂u ∂u
∂y , ∂z information. So, this is the
Eulerian description. The NS equations are casted in this form.
In the Lagrangian description, in the schematic, I just drew one stickman (blue) but in general there may be
millions of them. In principle, you can track the trajectories of each of them, but it’s computationally expensive.
Now, I drew the trajectory of the fluid particle on the illustration above passing the grid of points, and you can
understand that the Lagrangian observer should be able to not only obtain the time derivative (similar to the method
I just described above, you record the two velocity values at two time instants and calculate the time derivative) and
also obtain the spatial information (because the fluid particle is moving in space, so it can collect the information
regarding the variation in space). This means that for the Lagrangian observer, s/he can obtain at the same time
the temporal and spatial time derivatives; this is a vague way of saying the following equation
Du ∂u ∂u ∂u ∂u
LHS is Lagrangian = +u +v +w RHS is Eulerian. (5)
Dt ∂t ∂x ∂y ∂z
In the above explanation, I used u as the example. The above equations apply to any flow variable.

Q: When you were explaining the following equation

i j k
U × V = U1 U2 U3 = U2 V3 i + U3 V1 j + U1 V2 k − U3 V2 i − U1 V3 j − U2 V1 k = ijk Uj Vk ,
V1 V2 V 3
where does i come from?
A: The index
 i on the
 RHS
 is infact
 implicitly represented by the i, j, k vectors. In three dimensional problems, we
1 0 0
have i = 0 , j = 1 , k = 0.
0 0 1

2
 Then, for the original equation, I can rewrite as

i j k
U × V = U1 U2 U3 = U2 V3 i + U3 V1 j + U1 V2 k − U3 V2 i − U1 V3 j − U2 V1 k
V1 V2 V3
             
1 0 0 1 0 0 U2 V3 − U3 V2
= U2 V3 0 + U3 V1 1 + U1 V2 0 − U3 V2 0 − U1 V3 1 − U2 V1 0 = U3 V1 − U1 V3  (6)
0 0 1 0 0 1 U1 V2 − U2 V1
= ijk Uj Vk .

Note the vector in equation (6). Now we look at ijk Uj Vk , and will prove that we can get the same vector in
equation (6).
On the RHS, for ijk Uj Vk , the index i is the free index, meaning that the RHS’s a vector. Now I will write down
all the 9 terms represented by this term based on Einstein’s summation convention (recall my demonstration in the
lecture that this term ijk Uj Vk represents 9 terms)

ijk Uj Vk = i11 U1 V1 + i21 U2 V1 + i31 U3 V1

+ i12 U1 V2 + i22 U2 V2 + i32 U3 V2
+ i13 U1 V3 + i23 U2 V3 + i33 U3 V3 . (7)

The free index i is free to take a value from 1 to 2 to 3. Now, we take i = 1 to interrogate the first entry in this
vector ijk Uj Vk , that is,

1jk Uj Vk = 
111
U
1 V1 + 
121
U2 V1 + 
131
U3 V1
  

+112
U1 V2 + 
122
U 2 V2 + 132 U3 V2
 

+113
U1 V3 + 123 U2 V3 + 
133
U 3 V3 (8)
 

where the cancelled terms are due to the value of  = 0 (recall the definition of Levi-Civita symbol on the slide).
Also because 132 = −1 and 123 = 1, we are left with 1jk Uj Vk = −U3 V2 + U2 V3 .
Similarly, for i = 2, we have

2jk Uj Vk = 
211
U1 V1 + 
221
U
2 V1 + 231 U3 V1
 

+212
U 1 V2 + 
222
U2 V2 + 
232
U 3 V2
  

+ 213 U1 V3 + 
223
U2 V3 + 
233
U 3 V3
 

= 231 U3 V1 + 213 U1 V3 = U3 V1 − U1 V3 (9)

which means that the second entry in the matrix is U3 V1 − U1 V3 .

Finally for i = 3, we have

311
3jk Uj Vk =  U1 V1 + 321 U2 V1 + 
331
U
3 V1
 

322
+ 312 U1 V2 +  U 2 V2 + 
332
U3 V2
 

+313
U 1 V3 + 
323
U 2 V3 + 
333
U3 V3
  

= 321 U2 V1 + 312 U1 V2 = −U2 V1 + U1 V2 (10)

which is the third entry in the matrix.

Now we assemble all of them, we have the vector represented by the term ijk Uj Vk , that is
 
−U3 V2 + U2 V3
ijk Uj Vk =  U3 V1 − U1 V3  (11)
−U2 V1 + U1 V2

which is the same as the vector in equation (6).

3
Q: What’s the physical meaning of the variable s in the Eulerian de-
scription ds = |dx| dx
|u| or ds = u
.
A: s here is some dummy variable. It’s not important as s is usually cancelled out in the final result as you see
in the example on slide 34. It seems to be related to time, because ds = |dx|
|u| where |dx| should have the unit of a
length or m and |u| should have the unit of a velocity or m/second. So the unit of ds or s is second. But this s is
not the physical time t. Not to be confused.

4
 Questions from previous years.

Q: Slide 12 in Week 7. Why is the large vortex split into smaller ones?
A: That is because of nonlinearity. The viscosity will only act linearly as dissipation, making the vortex smaller. As
you will see in Week 8, the nonlinearity in the Navier-Stokes equations comes from the convective term and it is one
of the reasons why the NS equations are so difficult to solve. The interaction of two vortices can be considered as a
nonlinear process. Depending on their interaction, they may merge into one big vortex or split into several smaller
ones. This is a nonlinear process. If a process is linear, the number of the vortices in the flow field remains the same.

Q: Week 11. How do I understand the similarity function of the Blasius

boundary layer?
A: The similarity function Uu∞ = f 0 (η) means that the profile of u does not depend explicitly on either x or y. It is
√
instead a function of η, which is η = √ yνx = xy Rex . This means that for a pair of (x, y), if we change the value of
U∞
√
y from y to y1 while keeping the value of η constant, we can always find x1 , which satisfies xy Rex = xy11 Rex1 . And
p

both of them are equal to η, meaning that the velocity u in the Blasius profile at (x, y) is equal to (x1 , y1 ) because
they have the same η.
A visual explanation can be found as follows. As shown in the figure below, two Blasius profiles for Uν∞ = 100
are plotted at x = 2 and x = 12. The horizontal axis is f 0 = u/U∞ . The vertical axis is the y axis. You can see
that the two velocity profiles look different. Especially, the boundary layer thickness at x = 2 is smaller than that
at x = 12. (You can roughly pinpoint the edge of the boundary layer where the velocity becomes constant). This is
consistent with our conclusion that the boundary layer thickness increases along the x direction.

3.5

2.5

2
y

1.5

0.5

0
2 4 6 8 10 12 14
df/d

But the two boundary layer profiles would appear the same if one plots them together as η being the vertical axis.
To do that, for the blue profile, we have η = √ yνx = √y 2 , which means that we scale the value of y by dividing
U∞ 100
q q
2 √ y √ y 12
100 ; for the red profile, we have η = νx = 12
, which means that we scale the value of y by dividing 100
U∞ 100
in this case. When their y coordinate are rescaled respectively, we have the following graph. The dashed curves are
the rescaled profiles of the original ones. And they are the same, even though the solid curves are different.
The above example demonstrates the velocity profiles at two x-stations. See the figure below for the velocity
profiles ranging from x =1 to 20. The red dots are the 99% boundary layer thickness obtained numerically at each x
station. Note that I put the red dot at x =1,2, ..., 20 because the velocity profiles are pertaining to these x stations.
The black line is the result of y = √5x Rex
, which is the boundary layer thickness according to our derivation in the
lecture. The black line (theoretical derivation) is very close to the red dots (numerical implementation).

5
 10

y
4

0
2 4 6 8 10 12 14
df/d

4.5

3.5

2.5
y

1.5

0.5

0
0 5 10 15 20
df/d

Q: On slide 22 in week 7, you mentioned τw (and τw = µ ∂u ∂y which was

shown in the video clip on slide 23). Which comes first, τw or u?
A: τw here is a defined quantity and is called wall shear stress. The velocity profile should come first, in the following
sense. Imagining an experimentalist doing a flow experiment, s/he will be able to measure the velocity field. There
is no such way to accurately measure the wall shear stress in experiments. When s/he gets the velocity field, s/he
can calculate its y-derivative and this result, ∂u
∂y , evaluated at the wall and multiplied by µ will give the value of τw ,
according to the above definition. We will discuss this in detail in week 10.

dV
Q: Exercise 5 in week 7. Why is the acceleration a function like a = dt =
∂V ∂V ∂V ∂V ∂V ∂V ∂V
∂t + u ∂x + v ∂y + w ∂z ? How did the u ∂x + v ∂y + w ∂z come about?
A: This is the definition of the substantial derivative that we talked about in week 7. For the substantial derivative
or total derivative, you have to consider the derivative of all the variables. We can understand this derivative using
Taylor’s series, as follows.
Consider V = V(x, y, z, t), then we can expand the function V around a certain point in space and time
x0 , y0 , z0 , t0 , which is the Taylor’s series of V,
∂V ∂V ∂V ∂V
V(x, y, z, t) = V(x0 , y0 , z0 , t0 ) + ∆x + ∆y + ∆z + ∆t + H.O.T. (12)
∂x 0 ∂y 0 ∂z 0 ∂t 0
where H.O.T. means high order terms and ∆x = x − x0 , similarly for the other small variables ∆y, ∆z, ∆t. We only
show the first order terms in the above expansion.
Then according to the definition of the derivative, the acceleration is
dV V(x, y, z, t) − V(x − ∆x, y − ∆y, z − ∆z, t − ∆t)
a= = lim . (13)
dt ∆t→0 ∆t

6
We have this definition for the time derivative because when t changes, x, y, z will also change since they are function
of t. Then
dV V(x, y, z, t) − V(x0 , y0 , z0 , t0 )
a= = lim (14)
dt ∆t→0 ∆t
∂V ∂V ∂V ∂V
∂x ∆x + ∂y ∆y + ∂z ∆z + ∂t dt
0 0 0 0
= lim (15)
∆t→0 ∆t
∂V ∆x ∂V ∆y ∂V ∆z ∂V ∆t
= lim + lim + lim + lim (16)
∆t→0 ∂x ∆t ∆t→0 ∂y ∆t ∆t→0 ∂z ∆t ∆t→0 ∂t
0 0 0 0 ∆t
∂V dx ∂V dy ∂V dz ∂V
take the limit = + + + (17)
∂x 0 dt ∂y 0 dt ∂z 0 dt ∂t 0

Because x0 , y0 , z0 , t0 are actually arbitrary, we can simply neglect the subscript 0 , so

dV ∂V dx ∂V dy ∂V dz ∂V
a= = + + + (18)
dt ∂x dt ∂y dt ∂z dt ∂t
dx
It should be straightforward to realise that the time derivative of the coordinate is the velocity, that is, dt = u;
similarly for v, w. Thus, we have
dV ∂V ∂V ∂V ∂V
a= =u +v +w + (19)
dt ∂x ∂y ∂z ∂t
which is the equation given on the slides.

Q: Exercise 5 in week 8. In the video, you mentioned that the last

two terms in the conservation of horizontal momentum equation are the
buoyancy force. However, I thought buoyancy force is vertical?
A: A buoyancy force is a pressure force. For a certain object in the air, it should also experience the buoyancy force,
only that now in this case because the density of the air is small, the object does not feel a strong buoyancy force.
Indeed, the RESULTANT buoyancy force is acting upwards because of the dependency of the buoyancy force on the
height. For example, look at the object in the figure below. In this case, I agree with you that the RESULTANT
buoyancy force is vertical. However, the lateral surfaces of the object also experience buoyancy force, but because
the two lateral surfaces are identical there is no net buoyancy force in the horizontal direction. You can imagine the
following: when the water in the red dashed box suddenly vanishes, the buoyancy force on the right lateral surface
will push the object to the left. Thus, your understanding that the buoyancy force is vertical is probably because in
most examples, we have a zero net horizontal buoyancy force.
When you understand the above explanation, go back to the control volume in Ex. 5 of week 8 and you understand
that for the control volume, it does experience buoyancy forces on the two lateral surfaces and they are not equal.
This results in a net horizontal force pointing to the RHS. So, to sum up, a buoyancy force is a pressure force and
its direction is perpendicular to the control surface (and pointing towards the control volume).

Q: In week 10’s lecture, when you compare the velocity profiles of a lam-
inar boundary layer and a turbulent boundary layer, why the turbulent
b.l. has a “fuller” profile?
A: This has been mentioned on slide 13 of that week. It is because ”turbulent flow is characterized by greater
interchange of mass, momentum and energy within the fluid particles than in the case of laminar flow”. Thus, the
fluid particles in the outer region can penetrate into the interior region of the turbulent boundary layer, so that its
velocity profile is consequently “nearly uniform” or “fuller” than that in a laminar layer.

7
Q: Week 10, slide 27. In the middle paragraph, you said here that the
H value is larger when the adverse pressure gradient is stronger. This,
however, seems to contradict the paragraph below where it says the
separation occurs at H≈ 3.5 for the laminar flow and H≈ 2.5 for the
turbulent boundary layer flow.
Rδ
δ∗ 0
1− Uu∞ dy
A: The shape factor H is calculated according to the local velocity profile as in H = θ = Rδ . The fact
(1− Uu∞ )dy
u
0 U∞
that the turbulent profile is ”fuller” compared to the laminar flow gives rise to a larger value of H in the turbulent
boundary layer flows compared to the laminar flows (this deduction can be better understood by examples. For
instance, you can come up with two artificial analytical velocity profiles, one is ”fuller”, representing the turbulent
case, and one is less so for the laminar one. Then you can calculate the H values for them, you can see the above
result). These velocity profiles are results of the flow conditions, including the pressure gradient, being favourable
or adverse. We can understand that if you have a stronger adverse pressure gradient, which pushes against the flow
field, the velocity profile will be more like the laminar one. From this point of view, you can link the following items:
a stronger adverse pressure gradient, a laminar profile, and then a larger H value. Similarly for the turbulent flows.

Q: Week 8. I don’t understand the derivation on slide 22. Why is

pressure multiplied by δij ? What does δij mean here? How did the
equation of viscous stress tensor Tij come about, and why is it like this?
A: I believe I mentioned this in the
 lecture.Here, I used the tensor form δij to represent the identity
 matrix:
 δij
1 0 0 p 0 0
is essentially the identity matrix 0 1 0. Then, pressure multiplied by δij simply means 0 p 0, which
0 0 1 0 0 p
indicates that the pressure only appears in the normal stress component. I also discussed this in the lecture as I said
that the direction of the pressure is perpendicular to the surface (so it’s part of the normal stress).
The equation for Tij is the viscosity model Newton came up with. I again used the tensor notation here. If you
want, you can write down all the components of Tij , which are shown below (this is a snapshot from the text book
by Prof. Schlichting & Gersten, page 64, ninth edition)
In the snapshot, their τ is our T and their div ~v means the divergence of the velocity field, which is our ∂u
∂xk . It can
k

∂uj
∂ui
be clearly seen that our Tij = µ( ∂x j
+ ∂xi ) + λ ∂u
∂xk δij is essentially the same as their equations.
k

8
 ∂ui ∂uj
In the Newton’s law of viscosity, we have two terms, one being the stress generated by the flow shear ∂xj + ∂xi
and the other the stress engendered by the divergence of the velocity field ∂u
∂xk (whose physical meaning is the finite
k

expansion and compression of the fluid parcel. This effect, however, is not considered in the incompressible flows as
we study in this module. So I didn’t discuss too much about it. In the derivation on the next slides, we have used
the fact that ∂u
∂xk = 0 for incompressible flows). The second term is kept in the derivation is because I want the
k

equations to be more general.