In my research I obtained as result for a real function the following expression:

$$u(r) = H_0\left((-1)^{1/3} \, r \right) - Y_0\left((-1)^{1/3} \, r \right) + H_0\left(- (-1)^{2/3}\, r \right)- Y_0\left(- (-1)^{2/3}\, r \right)- H_0\left(r \right)- Y_0\left(r \right) $$

Where $H_0$ is the zeroth order StruveH function and $ Y_0$ is the zeroth order Bessel function of the second kind, and $r$ a real number greater than zero. I was quite happy about my result so i decided to plot it with mathematica and the result is the following:

enter image description here

The solution starts to behave strangely for large $r$, but i am not sure if this is just Mathematica not properly plotting, or if actually the solution is diverging at large $r$. Is there a way of evaluate such asymptotic behaviour? do you have any reference for that?

  • $\begingroup$ Recall that $K=H-Y$ and use this asymptotic expansions: dlmf.nist.gov/11.6 $\endgroup$ – tired Apr 20 '17 at 17:55
  • 2
    $\begingroup$ It definitely looks like numerical error. You can probably show it rigorously by using the known asymptotics for the Bessel functions and Struve functions. $\endgroup$ – Antonio Vargas Apr 20 '17 at 17:56
  • $\begingroup$ how are the third-roots of negatives defined? $\endgroup$ – mathreadler Apr 20 '17 at 18:10
  • $\begingroup$ @tired in the summation you also have a -H -Y, how do you treat that? $\endgroup$ – SSC Napoli Apr 20 '17 at 18:39
  • $\begingroup$ If it is numerical error, do Bessel and Struve have any "move" relations like sin and cos families do? Say for example being able to calculate $\cos(x)/x$ from $\cos(x+2\pi)/(x+2\pi)$. If they do maybe that could reduce the numerical error for expansions. $\endgroup$ – mathreadler Apr 20 '17 at 18:46

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.