Today we consider PDE problem in cylindrical coordinate and spherical coordinate.
Cylindrical Coordinate
Recall that Laplacian in cylindrical coordinate r,θ,z is written as
Δu=r1∂r(r∂r(u))+r21∂θ2u+∂z2u.
We shall look for eigenfunctions of Δ of the following form
u(r,θ,z)=R(r)Θ(θ)Z(z),
Then we have
u1Δu=R1r1∂r(r∂r(R))+Θ1r21∂θ2Θ+Z1∂z2Z.
such that
Θ−1∂θ2Θ=λθR1r1∂r(r∂r(R))+r−2λθ=λrZ1∂z2Z=λz.
eigenvalue problem for Θ. λθ=−n2,Θ(θ)=cos(nθ),sin(nθ)
eigenvalue problem for R(r).
Assume λθ=−n2, then we get
r∂r(r∂r(R))+(−λrr2−n2)R=0.
Compare with Bessel equation (12.2) from Boas
x(xy′)′+(x2−p2)y=0⇒y(x)=aJp(x)+bYp(x).
And notice that we want R(0) finite, and ∣Yp(0)∣=∞, we see
R(r)={Jn(−λrr)rnλr=0λr=0
Note that, the −λr has an ambiguity of ±1 sign, but Jn(x)=(−1)nJn(−x), hence you can take either sign for the square root.
If λr<0, say λr=−k2 for k>0, then
R(r)=Jn(−λrr)=Jn(kr)R(r)willoscillateasr$ increase.
If λr>0, say λr=k2 for k>0, then we will take
R(r)=In(kr)=inJn(ikr)
where In(kr) is the 'hyperbolic' Bessel function. R(r) looks like expoential function, with no oscillation.
Eigenvalue problem for Z(z) Z(z)={eλZz1,zλz=0λz=0
Hence, for any λθ=−n2,λr∈R,λz∈R, we have an eigen-function of Laplacian
u=Rλr(r)Θn(θ)Zλz(z),Δu=(λr+λz)u.
where the R,Θ,Z are given as above.
Solve the Steady State Temperature problem
Suppose we are given a cylinder of radius 1 and height 1 (z∈[0,1]), and suppose we specify the temperature on the boundary of the cylinder. We need to solve the equation
Δu=0
satisfying the boundary condition.
Note that λz+λr=0, hence we will use λr and n to label the solutions.
We consider two special cases of boundary value specification.
Case 1: Bottom and Side boundary value = 0. u(r,θ,z=1) is given
Since R(r) will have a zero at r=1, we should have λr=−k2<0, and
uk,n(r,θ,z)=Jn(kr)cos(nθ)sinh(kz),Jn(kr)sin(nθ)sinh(kz)
For each n, we need to choose those k such that Jn(k1)=0, hence the choice of k is discrete as well.
More precisely, let kn,m>0 be the m-th root of Jn(x). Then the general solution is
u(r,θ,z)=n=1∑∞m=1∑∞Jn(kn,mr)sinh(kn,mz)(an,mcos(nθ)+bn,msin(nθ))
Set z=1, we can use boundary condition to determine the coefficients an,m,bn,m. Recall that
∫01Jn(kn,m1r)Jn(kn,m2r)rdr=0,m1=m2
Hence
an,m=sinh(kn,m)∫01∫02πJn(kn,mr)2cos2(nθ)rdrdθ∫01∫02πu(r,θ,1)Jn(kn,mr)cos(nθ)rdrdθ
and bn,m is obtained similarly, replacing cos(nθ) by sin(nθ).
Case 2: Bottom and Top face boundary value = 0. u(r=1,θ,z) is given
This forces λz<0. In particular, Z(z)=sin(mπz), hence λz=−(mπ)2 and λr=(mπ)2. The general solution is written as
u(r,θ,z)=n=1∑∞m=1∑∞In(mπr)sin(mπz)(an,mcos(nθ)+bn,msin(nθ)).
We figure out the coefficient an,m,bn,m using the boundary value at r=1.
u(1,θ,z)=n=1∑∞m=1∑∞In(mπ)sin(mπz)(an,mcos(nθ)+bn,msin(nθ)).
Hence by integrating θ,z we get
an,m=In(mπ)∫01∫02πsin2(mπz)cos2(nθ)dθdz∫01∫02πu(1,θ,z)sin(mπz)cos(nθ)dθdz
and similarly for bn,m.
General case
Decompose the boundary condition into three parts, top face, bottom face and side face. Break the original problem into 3 smaller problems in which only one part is non-zero. Then add up the solution from the three smaller problems. You end up with a solution that satisfies Δu=0 and has the desired boundary value.
math121b/04-03.txt · Last modified: 2020/04/03 10:39 by pzhou