IS2202 – Engineering Mathematics IV

Department of Interdisciplinary Studies

Faculty of Engineering

Practical 3

Solving the 2-D Wave Equation in Polar Coordinates

Recall the 2-D Wave equation in Polar coordinates that was derived in the lectures;

𝜕 2 𝑢 1 𝜕𝑢 1 𝜕 2 𝑢
𝑢𝑡𝑡 = 𝑐 2 ( 2 + + )
𝜕𝑥 𝑟 𝜕𝑟 𝑟 2 𝜕𝜃 2

Now, consider a circle with radius R. We will only model radially symmetric solutions,
meaning solutions only depending on r and t, such that u(r,t). So, in this instance,  will be
constant causing u = 0. So, the above wave equation reduces to;

2
𝜕 2 𝑢 1 𝜕𝑢
𝑢𝑡𝑡 = 𝑐 ( 2 + )
𝜕𝑥 𝑟 𝜕𝑟

Assuming the membrane is clamped at the edges, then the boundary condition is;

𝑢(𝑅, 𝑡) = 0

Then, assume the initial conditions;

𝑢(𝑟, 0) = 𝑓(𝑟)
𝑢𝑡 (𝑟, 0) = 𝑔(𝑟)

Department of Interdisciplinary Studies IS2202: Engineering Mathematics IV 1

Faculty of Engineering, University of Sri Jayawardenapura
Question 1

Use separation of variables to derive a solution for the function u(r,t), with given BCs and ICs.

i. Let u(r,t) = W(r)G(t). Use -k2 as your separation constant. Derive the 2 ODEs required
for this solution.

ii. For the ODE that is a function of (W(r)), show that it may be re-written as the Bessel
equation through this substitution.

𝑠 = 𝑘𝑟
𝑑2𝑊 1 𝑑𝑊
+ + 𝑊 = 0 − −→ 𝐵𝑒𝑠𝑠𝑒𝑙 ′ 𝑠 𝐸𝑞𝑢𝑎𝑡𝑖𝑜𝑛 𝑓𝑜𝑟 𝜈 = 0
𝑑𝑠 2 𝑠 𝑑𝑠

𝑑𝑊 𝑑𝑊 𝑑𝑠
Hint: Use the chain rule to replace derivatives in w.r.t r to s. i.e = .
𝑑𝑟 𝑑𝑠 𝑑𝑟

The Bessel equation at nu = 0, has solutions Jo and Yo. (You do not have to worry about the
derivation of these functions). However, Yo →  when x → 0. Since the deflection of the
membrane at x=0 cannot be infinite, we do not consider the Yo solution. Therefore, the above
ODE has the solution;

𝑊(𝑟) = 𝐽𝑜 (𝑠) = 𝐽𝑜 (𝑘𝑟)

Where the Bessel function J is defined as;


𝜈
(−1)𝑚 𝑥 2𝑚
𝐽𝜈 (𝑥) = 𝑥 ∑ 2𝑚+𝜈
2 𝑚! Γ(𝜈 + 𝑚 + 1)
𝑚=0

Now, if we apply the given boundary condition;


(−1)𝑚 (𝑘𝑅)2𝑚
𝑊(𝑅) = 𝐽𝑜 (𝑘𝑅) = ∑ 2𝑚 =0
2 𝑚! Γ(𝑚 + 1)
𝑚=0

Department of Interdisciplinary Studies IS2202: Engineering Mathematics IV 2

Faculty of Engineering, University of Sri Jayawardenapura
This series converges for all kR but fortunately, we don’t have to solve this by hand.
Thankfully, we can plot 𝐽𝑜 (𝑘𝑅) vs some arbitrary values of kR through Matlab and find the
points where 𝐽𝑜 (𝑘𝑅) = 0.

%define kR values. Can increase spacing 0.01 to improve approximation;

kR = 0:0.01:12;

%Define an arraj J to store the Bessel function output

Jo=besselj(0,kR);

figure(1)
plot(kR,Jo)
grid on
legend('J_0')
title('Bessel Functions of the First Kind for \nu = 0 ')
xlabel('kR')
ylabel('J_0(kR)')

%Find the values of J that is approximately = 0. "Find" function allows us

to
%find the location in the array where J ~ 0. Refer to the graph
%to establish a suitable range

I=find(Jo>-xx & Jo<0.0025);

%Our solution is, at what values of kR is J ~ 0. Since kR and J are the

%same size, we can use the location positions in I, to find the kR values
%at which J~0. Use “alpha=kR(I)” in the command window to find alpha_m
vals.

%Create an array of the 4 alpha_m values that best approximates J~0.

alpha_m = [xx, xx, xx, xx];

You will find an infinite number of such points that we will refer to as m. Note, how here, m
actually corresponds to the vibrational modes, but in 2-D. Similar to how F(x) defined the mode
shape for the 1-D case, W(r) ends up defining the mode shape here as well.

Figure 1: 2-D Vibrational Modes in Polar Coordinates

Department of Interdisciplinary Studies IS2202: Engineering Mathematics IV 3

Faculty of Engineering, University of Sri Jayawardenapura
 iv. Find the 𝛼𝑚 values for m = 1, 2, 3 and 4. Increase spacing to increase size of array kR
to improve your approximation. Report to 4 decimal places. Attach your modified
Matlab code.

Since;
𝑘𝑅 = 𝛼𝑚

𝛼𝑚
𝑘=
𝑅

So, for our ODE in terms of W(r) we end up with an infinite number of solutions in the form;

𝛼𝑚
𝑊𝑚 (𝑟) = 𝐽𝑜 ( 𝑟)
𝑅

v. Solve the time dependent PDE as usual and show that the m th solution for our function
is;

𝛼𝑚
𝑢𝑚 (𝑟, 𝑡) = [𝐴𝑚 cos(𝑐 (𝛼𝑚 /𝑅) ) 𝑡 + 𝐵𝑚 sin(𝑐 (𝛼𝑚 /𝑅) ) 𝑡 ] 𝐽𝑜 ( 𝑟)
𝑅

So,


𝛼𝑚
𝑢(𝑟, 𝑡) = ∑ [𝐴𝑚 cos(𝑐 (𝛼𝑚 /𝑅) ) 𝑡 + 𝐵𝑚 sin(𝑐 (𝛼𝑚 /𝑅) ) 𝑡 ] 𝐽𝑜 ( 𝑟)
𝑅
𝑚=1

Now, we satisfy the initial conditions;

 1 0 𝛼𝑚
𝑢(𝑟, 0) = ∑ [𝐴𝑚 cos(𝑐 (𝛼𝑚 /𝑅) ) 𝑡 + 𝐵𝑚 sin(𝑐 (𝛼𝑚 /𝑅) ) 𝑡 ] 𝐽𝑜 ( 𝑟)
𝑅
𝑚=1

Department of Interdisciplinary Studies IS2202: Engineering Mathematics IV 4

Faculty of Engineering, University of Sri Jayawardenapura
 
𝛼𝑚
= ∑ 𝐴𝑚 𝐽𝑜 ( 𝑟) = 𝑓(𝑟)
𝑅
𝑚=1

Now, if you recall the lecture, you will see the similarity this function has with the Fourier
manipulation we did last time. What this actually resembles is the Fourier-Bessel expansion of
f(r). We are not going to derive it from scratch, but just know that the co-efficient Am can be
found as;

𝑅
2 𝛼𝑚
𝐴𝑚 = 2 2 (𝛼 ) ∫ 𝑟 𝑓(𝑟) 𝐽𝑜 ( 𝑟) 𝑑𝑟
𝑅 𝐽1 𝑚 0 𝑅

Similarly, Bm can be found by substituting the other IC; 𝑢𝑡 (𝑟, 0) = 𝑔(𝑟).

Now, let’s assume that;

𝑓(𝑟) = (1 − 𝑟 2 ) 𝑎𝑛𝑑 𝑔(𝑟) = 0 and R=1

The Bessel function derivative theory states that;

𝑑(𝑥 𝜈 𝐽𝜈 (𝜆𝑥))
= 𝜆 𝑥 𝜈 𝐽𝜈−1 (𝜆𝑥) (𝑇ℎ𝑒𝑜𝑟𝑦 1)
𝑑𝑥

Also,
2𝜈
𝐽𝜈−1 (𝑥) + 𝐽𝜈+1 (𝑥) = 𝐽 (𝑥) (𝑇ℎ𝑒𝑜𝑟𝑦 2)
𝑥 𝜈
Eg:
2𝜈
𝐽𝑜 (𝑥) + 𝐽2 (𝑥) = 𝐽 (𝑥)
𝑥 1

Department of Interdisciplinary Studies IS2202: Engineering Mathematics IV 5

Faculty of Engineering, University of Sri Jayawardenapura
vi. Use integration by parts to show that;

8
𝐴𝑚 = 3
𝛼𝑚 𝐽1 (𝛼𝑚 )

𝛼𝑚
(Hint: Use dV = 𝑟 𝐽𝑜 ( 𝑟) , 𝑢 = (1 − 𝑟 2 ) 𝑎𝑛𝑑 𝑇ℎ𝑒𝑜𝑟𝑦 1 𝑎𝑛𝑑 𝑇ℎ𝑒𝑜𝑟𝑦 2)
𝑅

vii. Find Am for m = 1,2,3 and 4 and plot the full solution u (r,,t) at c = 1 m/s. For  = 0
→ 2 and t = 1000s. Attach your modified Matlab code. What mode shape is
dominating? How does this relate to the shape of the Initial condition?

%Find an array for Am. Define J1(alpha_m)

Am= zeros([1 length(alpha_m)]);
J1=besselj(1,alpha_m);

for m=1:length(alpha_m)
Am(m)= 8/(((alpha_m(m))^3)*J1(m));
end

%Full solution plotting.

c=1;
R=1;
t=0;
r=linspace(0,R,150);
theta=linspace(0,2*pi,150);
Jo_r=zeros([length(r) length(alpha_m)]);
u=[];

for n=1:length(r)
for m=1:length(alpha_m)
Jo_r(n,m)=besselj(0,(alpha_m(m)*r(n)));
u(n)= sum(Am.*(Jo_r(n,:)).*(cos(c*(alpha_m)*t)));

end
end

figure(2)
plot(r,u)

figure(3)
xx = bsxfun(@times,r',cos(theta)); %convert all the radial coords to x y
yy = bsxfun(@times,r',sin(theta));
zz = repmat(u',1,length(r)); %turn vector into 150x150 matrix
surf(xx,yy,zz)

Department of Interdisciplinary Studies IS2202: Engineering Mathematics IV 6

Faculty of Engineering, University of Sri Jayawardenapura