Clear Filters
Clear Filters

Numerical integration of Airy functions

4 views (last 30 days)
carlos g
carlos g on 22 Apr 2018
Commented: carlos g on 24 Apr 2018
I have a problem when I try to numerically integrate the Airy function at "extreme values". Here's a MWE:
clc
clear
%%Parameters
M=4.5
lambda_0=0.332;
tic
for pp=4:4
for m=1:1
beta1=(pp-1)/10;
alpha1=m;
omega1=7
gamma1=sqrt((M^2-1)*alpha1^2-beta1^2);
N=25;
YMAX=80;
eta01=-i*omega1/((i*alpha1*lambda_0)^(2/3))
eta_inf1=((i*alpha1*lambda_0)^(1/3))*YMAX+eta01;
%%Operations with Airy function
syms lu
aiprime(lu) = airy(1,lu);
aisecond(lu)= diff(airy(1,lu));
airy_D1=airy(1,eta01)
airy_DD1=vpa(aisecond(eta01));
airy_INT1=integral(@(n) airy(n),eta01,eta_inf1)
%%Reflection coefficient
inte1=alpha1*(alpha1^2+beta1^2)*airy_INT1;
deri1=gamma1*lambda_0*((i*alpha1*lambda_0)^(2/3))*airy_D1;
Ref_coeff1=-(inte1+deri1)/(inte1-deri1);
q1=(alpha1^2+beta1^2)*(1+Ref_coeff1)/((i*alpha1*lambda_0)^(2/3)*airy_D1);
p1=1+Ref_coeff1;
%%Coefficients
aa1=2*pi*complex(0,1)*gamma1*beta1*lambda_0*airy_D1/(alpha1*(alpha1^2+beta1^2)*airy_INT1-gamma1*lambda_0*airy_D1*(i*alpha1*lambda_0)^(2/3));
bb1=-aa1*airy(2,eta01)*airy_INT1/airy(eta01);
eta1=@(y) ((i*alpha1*lambda_0)^(1/3))*y+eta01;
Gi1=@(x) -(airy(2,x)*integral(@(n) airy(n),eta_inf1,x)-airy(x)*integral(@(n) airy(2,n),eta01,x));
V1=@(eta) ((i*alpha1*lambda_0)^(-1/3))*q1*integral2(@(x, y) airy(y),...
eta01, eta, eta01, @(x) x);
DV1=@(eta) q1*integral(@(n) airy(n),eta01,eta);
W1=@(eta) aa1*Gi1(eta)+bb1*airy(eta);
U1=@(eta) i/alpha1*DV1(eta)-beta1/alpha1*W1(eta);
toc
end
end
ve=0:0.01:N;
U1VEC=arrayfun(U1,arrayfun(eta1,0:0.01:N));
V1VEC=arrayfun(V1,arrayfun(eta1,0:0.01:N));
W1VEC=arrayfun(W1,arrayfun(eta1,0:0.01:N));
figure(1)
subplot(1,3,1)
plot(abs(U1VEC),ve,'-k','LineWidth',2)
xlabel('$\hat{U}_1$','Interpreter','latex','Fontsize',15)
ylab=ylabel('$Y$','Interpreter','latex','Fontsize',15,'rot', 0);
set(ylab, 'Units', 'Normalized', 'Position', [-0.4, 0.45, 0]);
set(gca,'linewidth',1)
subplot(1,3,2)
plot(abs(V1VEC),ve,'-k','LineWidth',2)
xlabel('$\hat{V}_1$','Interpreter','latex','Fontsize',15)
set(gca,'linewidth',1)
subplot(1,3,3)
plot(abs(W1VEC),ve,'-k','LineWidth',2)
xlabel('$\hat{W}_1$','Interpreter','latex','Fontsize',15)
hold on
plot(abs(-i*beta1*p1/((i*alpha1*lambda_0)^(2/3))./eta1(5:0.01:N)),5:0.01:N,'--k')
set(gca,'linewidth',1)
The problem is concretely in
Gi1=@(x) -(airy(2,x)*integral(@(n) airy(n),eta_inf1,x)-airy(x)*integral(@(n) airy(2,n),eta01,x));
In this case, eta01 and eta_inf1 take very extreme values, making the U1 and W1 solutions get wrong values.
Do you have any idea on how to solve this problem? Without increasing the computational time too much, so avoiding increasing the precision artificially with vpa.
  3 Comments
carlos g
carlos g on 22 Apr 2018
Hi John BG,
Yes, I did it on purpose. Sometimes I am using the Ai function and, in the first case you mentioned, it's the Bi function. Is this your question?

Sign in to comment.

Answers (0)

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!