I would like to find the relation between the solutions of a differential equation obtained by two different authors. The first solution is given in terms of the hypergeometric function $_2F_1$:

$$\left(1+\beta p^2\right)^{\frac{1}{4} (-1-2 n-2 \xi)} \, _2F_1\left(-n,-n-2 \xi ;-n-\xi +\frac{1}{2};\frac{1}{2}\left(1 + i p \sqrt{\beta}\right)\right),$$

where $p$ is a real variable, $\beta>0$, $n$ is a positive integer and $\xi= \sqrt{4 + \beta^2}/(2\beta)$.

The second one involves the Gegenbauer polynomials $C_n^\lambda(z)$

$$ \left(1 + \beta p^2\right)^{-\frac{1}{2} \left(\xi +\frac 12\right)} C_n^{\left(\xi +\frac{1}{2}\right)}\left(\frac{\sqrt{\beta } p}{\sqrt{\beta p^2+1}}\right).$$

I have omitted some normalization constants. Perhaps it is easy, but I do not see it. I know the relations between the hypergeometric function and the Gegenbauer polynomials but I can't apply them to this problem. In particular, I'm not sure how to transform the argument of $_2F_1$ to match the argument of $C_n^\lambda(z)$.

  • $\begingroup$ @MartinNicholson $n$ is an arbitrary positive integer (and also zero): $n = 0,1,2,\ldots$. But if you have a relation that involves even or odd $n$ it's a good start. $\endgroup$ – Pincopallino Nov 24 '15 at 18:48
  • $\begingroup$ formulas 20-23 mathworld.wolfram.com/GegenbauerPolynomial.html $\endgroup$ – Nemo Nov 24 '15 at 19:46
  • $\begingroup$ @MartinNicholson I have already seen these formulas but I can't use them, or at least I can't find the proper variable change from $(1+i\sqrt{\beta}p)/2$ to $\sqrt{\beta} p / \sqrt{\beta p^2 +1}$ $\endgroup$ – Pincopallino Nov 25 '15 at 15:53

I will consider the case when $n$ is even. The fact that these two solutions are equivalent when $n$ is odd can be proved in an analogous manner.

Let $n=2\nu$ be even nonnegative integer, $\lambda=\xi+\frac{1}{2}$ and let's absorb $\sqrt{\beta}$ into the definition of $p$. Then the ratio of the two expressions above can be written as $$ \displaystyle\frac{ \displaystyle\phantom{}_2F_1\left(-2\nu,-2\nu-2\lambda+1,-2\nu-\lambda+1;\frac{1+pi}{2}\right)}{ \displaystyle(1+p^2)^\nu C_{2\nu}^\lambda\left(\frac{p}{\sqrt{p^2+1}}\right)}. $$ By eq. (20): $$ C_{2\nu}^\lambda\left(\frac{p}{\sqrt{p^2+1}}\right)=\left({\textstyle 2\nu+2\lambda-1\atop \textstyle 2\nu}\right)\phantom{}_2F_1\left(-\nu,\nu+\lambda,\lambda+\frac{1}{2};\frac{1}{p^2+1}\right). $$ By eq. (15.8.2) this is equal, up to a constant coefficient: $$ \frac{1}{(p^2+1)^\nu}\cdot \phantom{}_2F_1\left(-\nu,-\nu-\lambda+\frac{1}{2},-2\nu-\lambda+1;p^2+1\right). $$ By Kummer's quadratic transformation, eq. 15.8.18: $$ \phantom{}_2F_1\left(-\nu,-\nu-\lambda+\frac{1}{2},-2\nu-\lambda+1;p^2+1\right)=\phantom{}_2F_1\left(-2\nu,-2\nu-2\lambda+1,-2\nu-\lambda+1;\frac{1\pm pi}{2}\right). $$ So we proved that the expression $$ \displaystyle\frac{ \displaystyle\phantom{}_2F_1\left(-2\nu,-2\nu-2\lambda+1,-2\nu-\lambda+1;\frac{1+pi}{2}\right)}{ \displaystyle(1+p^2)^\nu C_{2\nu}^\lambda\left(\frac{p}{\sqrt{p^2+1}}\right)}. $$ is constant, i.e. doesn't depend on $p$, as required.

  • $\begingroup$ Thank you! But I think that you used 15.8.2 instead of 15.8.1 and 15.8.18 instead of 15.8.17, am I right? Otherwise I can't explain the passages. $\endgroup$ – Pincopallino Nov 30 '15 at 16:16
  • $\begingroup$ @Pincopallino You are right, thanks for mentioning that. $\endgroup$ – Nemo Nov 30 '15 at 16:37
  • $\begingroup$ While I was working out the normalization constant, I came up with an interesting thing. The normalization constant of the two functions should be a number close to unity. After applying Eq. (15.8.2) of the NIST DLMF I obtained a very large $\Gamma(2 \lambda)$ in the normalization. But by using the corresponding formula on the Abramowitz & Stegun, (15.3.7), which is slightly different, then everything is correct. I wonder why the two formulas are different and why the older one works for me and the other doesn't $\endgroup$ – Pincopallino Dec 1 '15 at 20:59
  • $\begingroup$ @Pincopallino try to use (1.8.6) instead of (1.8.2). $\endgroup$ – Nemo Dec 1 '15 at 21:03
  • $\begingroup$ Yes (1.8.6) is the right one to use, thank you. Still, I don't understand the problem with (15.8.2)... $\endgroup$ – Pincopallino Dec 2 '15 at 11:13

Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.