using ApproxFun, Plots, ComplexPhasePortrait, ApproxFun, SingularIntegralEquations,
SpecialFunctions, DualNumbers, SO
using SingularIntegralEquations.HypergeometricFunctions
gr();
Dr Sheehan Olver
Office Hours: 3-4pm Mondays, 11-12am Thursdays, Huxley 6M40
We will consider the first order ODE u′+a(z)u=0
The simplest point is an ordinary point, where we know that u is analytic:
Definition (ordinary point) z0 is a ordinary point of the ODE if a(z) and b(z) are analytic at z0.
Suppose a has a simple pole, and w.l.o.g. take z0=0, that is: a(z)=a−1z+a0+a1z+a2z2+⋯
Note that u′(z)=αzα−1+zα((α+1)v+zv′)
Here is an example: as we can see, v(z)
is analytic in a disk of radius r
where r
is dictated by the function a
.
r = 0.9 # radius we are solving in, can be up to but not including 1
S = Taylor(Circle(r))
α = 0.3
a = z -> α/z + 1/(1-z)
ã = Fun( z -> a(z) - α/z, S)
za = Fun( z -> z*a(z), S)
z = Fun(S)
D = Derivative() : S → S
L = z*D + (α+1-za)
v = L \ ã
# v is analytic in disk of radius r
phaseplot(v, (-1,1), (-1,1))
Here is a plot of u
, which shows that it has a branch cut:
# u has a branch
u = z -> z^α + z^(α+1)*v(z)
phaseplot(u, (-1,1), (-1,1))
We can confirm it solves the ODE:
up = dualpart(u(dual(0.1,1.0))) # uses DualNumbers.jl to calculate derivative
up - a(0.1)u(0.1)
4.884981308350689e-15 - 7.521914234669674e-17im
Definition (first order regular singular point) z0 is a regular singular point of the first order ODE if (z−z0)a(z) is analytic at z0.
For second order equations, we will arrive at the following analogue:
Definition (second order regular singular point) z0 is a regular singular point of the second order ODE if (z−z0)a(z) and (z−z0)2b(z) are analytic at z0.
Assume a(z)=a−1z+a0+a1z+⋯
Definition (indicial equation) α(α−1)+a−1α+b−2=0
Now take as an ansatz: u(z)=zα(1+zv(z))=zα+zα+1v(z)
The question is when can forward substitution proceed. Let's write out the operators, first we have: z2v″≡(002612⋱)
We need to only worry about the diagonal to determine if forward substitution proceeds, in this case we have (2α+a−1×4α+2+2a−1××6α+3a−1+6×××8α+4a−1+12××××10α+5a−1+20⋮⋱⋱⋱⋱⋱)
Remark If this condition is satisfied, then v is analytic near the singular point, which follows by bounding the forward elimination as before, but we'll skip this for now.
Remark If α+k+1 is an integer, then using the ansatz zα+k+1(1+zv(z)) proceeds without problem. To get the second solution, we need to introduce logarithms, but we'll skip this for now.
We now check the results of solving z2v″+z(2(α+1)+za(z))v′+(α(α+1)+za(z)(α+1)+z2b(z))v=−α(α−1)z−a(z)α−zb(z)
We first choose a solutino to the an indicial equation:
S = Taylor()
z = Fun(S)
D = Derivative() : S
a₋₁ = -0.6
b₋₂ = 0.3
a = z -> a₋₁/z + 1/(2-z)
ã = Fun( z -> a(z) - a₋₁/z, S)
za = Fun( z -> z*a(z), S)
b = z -> b₋₂/z^2 + 0.4/z + 1/(3-z)
zb̃ = Fun(z -> z*b(z) - b₋₂/z, S)
z²b = Fun( z -> z^2*b(z), S)
α = ((1-a₋₁) + sqrt((a₋₁-1)^2-4b₋₂))/2
@show α
α*(α-1) + α*a₋₁ + b₋₂
α = 1.3830951894845303
2.7755575615628914e-16
We can solve for v
, which is analytic in a neighbourhood of 0:
L = z^2*D^2 + z*(2(α+1) + za)*D + (α*(α+1) + za*(α+1) + z²b)
v = L \ (-α*ã - zb̃)
phaseplot(v, (-3,3), (-3,3))
Thus one solution is:
u = z -> z^α + z^(α+1)*v(z)
phaseplot(u, (-3,3), (-3,3))
Let's double check it satisfies the original ODE:
# dual numbers don't work to second degree so let's check numerically:
uf = Fun(u, 0.1 .. 0.2)
norm(uf'' + Fun(a, space(uf))*uf' + Fun(b, space(uf))*uf)
3.3654295491028256e-12
The second, linearly independent solution can be found taking the other α:
α̃ = ((1-a₋₁) - sqrt((a₋₁-1)^2-4b₋₂))/2
@show α̃
α̃*(α̃-1) + α̃*a₋₁ + b₋₂
α̃ = 0.21690481051546984
1.6653345369377348e-16
L = z^2*D^2 + z*(2(α̃+1) + za)*D + (α̃*(α̃+1) + za*(α̃+1) + z²b)
ṽ = L \ (-α̃*ã - zb̃)
phaseplot(ṽ, (-3,3), (-3,3))
ũ = z -> z^α + z^(α+1)*ṽ(z)
phaseplot(ũ, (-3,3), (-3,3))
Any general solution to the ODE is a linear combination of these two.
Every isolated singularity that is not regular is an irregular singular point: even something as simple as du/dz=uz2 has an irregular singular point. Irregular singular points correspond to essential singularities: in this case the solution is u(z)=e−1/z
To determine the type of singularity at ∞ we do our usual change of variables u(z)=v(1/z). For example, if dudz=a(z)u
Definition (first order regular singular point at ∞) ∞ is a regular singular point of a first order ODE if za(z) is analytic at ∞.
For second order, we see that v satisfies v′(z)=−z−2u′(z−1)v″(z)=2z−3u′(z−1)+z−4u″(z−1)=−2z−1v′(z)+z−4u″(z−1)
Definition (second order regular singular point at ∞) ∞ is a regular singular point of a first order ODE if za(z) and z2b(z) are analytic at ∞.