11
$\begingroup$

Is there a way to get a clean presentation of the orthogonal polynomials w.r.t. the Maxwell distribution https://en.wikipedia.org/wiki/Maxwell%E2%80%93Boltzmann_distribution ? If you separate the problems to an even and odd problems, the even part of the orthogonal basis behaves nice enough: one simply takes the quotient of the odd Hermite polynomials by x (since the Maxwell distribution is given by $Cx^2\exp(-x^2/2)$, and Hermite polynomials are orthogonal w.r.t. $\exp(-x^2/2)$) ; as for the even part, I described a semi-solution in the comment below. However, I have no ide how the two parts interact.

The analog question for Laguerre polynomials (i.e. when one moves from the measure $\exp(-x)$ to the measure $C_{\alpha}x^{\alpha}\exp(-x)$ yields a very satisfying result in the form of generalized Laguerre polynomials. Due to the relation between Laguerre polynomials and Hermite polynomials, I'd expect something similar here.

$\endgroup$
  • $\begingroup$ For what it's worth, setting $F_{2n+1}=(H_{2n+2} + (2n+1)H_{2n})/x$ for the odd degree polynomials, (which - as mentioned in the question - are the only problem), the inner products form a symmetric tri-diagonal matrix. $\endgroup$ – David Lehavi Jun 12 '16 at 15:17
  • $\begingroup$ According to this reference link.springer.com/chapter/10.1007/978-94-017-9454-1_2 page 30, these are Sonine polynomials with $\alpha=1/2$. But I am afraid that these are only polynomials in $x^2$, that is not what you are asking about. $\endgroup$ – Fedor Petrov Jun 14 '16 at 9:46
  • $\begingroup$ @FP, thanks, but indeed I want the full orthogonal basis - odd powers included - and I can't believe I'm the first one asking this question, it's very natural if you are analyzing random points in 3D. $\endgroup$ – David Lehavi Jun 14 '16 at 9:53
10
+50
$\begingroup$

These are the socalled Maxwell polynomials $M_n^{(p)}(x)$, see page 75 and following of Spectral Methods in Chemistry and Physics. The coefficients for $p=2$, which is the case you need, are given in Table 2.14 of that reference.

| cite | improve this answer | |
$\endgroup$
  • $\begingroup$ Specifically (following your references) it seems that what I looked for is referred to as "half range generalized Hermite polynomials". Thanks $\endgroup$ – David Lehavi Jun 14 '16 at 10:23

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.