# Orthogonal polynomials

**URL:** <https://boards.straightdope.com/t/orthogonal-polynomials/649396>\
**Category:** Factual Questions\
**Created:** [February 5, 2013, 4:20pm UTC](https://boards.straightdope.com/t/orthogonal-polynomials/649396 "2013-02-05T16:20:29Z")\
**Posts on this page:** 3\
**Page:** 1

<div class="post-metadata">

**Author:** ![MikeS](https://avatars.discourse-cdn.com/v4/letter/m/919ad9/32.png) [@MikeS](https://boards.straightdope.com/u/MikeS)\
**Post date:** [February 5, 2013, 4:20pm UTC](https://boards.straightdope.com/t/orthogonal-polynomials/649396/1 "2013-02-05T16:20:29Z")

</div>

I’ve been playing around with orthogonal polynomials as part of my research recently, and a question occurred to me. It seems like a simple enough question to pose, but I’m not sure how to go about addressing it.

Background: the Legendre polynomials {1, x, 3x[sup]2[/sup] - 1, …} are orthogonal if you integrate them over the domain [-1,1] with no weighting. The Hermite polynomials {1, 2x, 4x[sup]2[/sup] - 2, …} are orthogonal if you integrate them over the domain (-∞,+∞) with a weighting of e[sup]-x[sup]2[/sup]/2[/sup]. The Chebyshev polynomials of the first kind {1, x, 2x[sup]2[/sup]-1, …} are orthogonal over the domain [-1,1] with a weighting function of 1/√(1-x[sup]2[/sup]). My questions are:  
[ul][_]Does there exist a weighting/measure and domain of integration such that the monomials {1, x, x[sup]2[/sup], x[sup]3[/sup], …} are orthogonal? If so, can an explicit expression for this weighting be given?[_]More generally, if you hand me a countably infinite set of polynomials of degrees 0, 1, 2, 3, …, can I always find an inner product (i.e., domain of integration and weighting) such that these polynomials are orthogonal? Does the answer change if I let this inner product be indefinite?[/ul]I tried playing around with Sturm-Liouville theory for a bit, but it’s been a while since I dealt with such things. I have a niggling suspicion that I asked this question of one of my professors once, but I can’t recall what the answer was (or if he had one.)

---

<div class="post-metadata">

**Author:** ![ZenBeam](https://avatars.discourse-cdn.com/v4/letter/z/3ab097/32.png) [@ZenBeam](https://boards.straightdope.com/u/ZenBeam)\
**Post date:** [February 5, 2013, 5:32pm UTC](https://boards.straightdope.com/t/orthogonal-polynomials/649396/2 "2013-02-05T17:32:45Z")

</div>

If the weighting is to be non-negative, you wouldn’t be able to, since 1, x^2, x^4 are all non-negative as well, so they couldn’t be made orthogonal.

Even with arbitrary weighting, they couldn’t be made orthonormal. The self term of x would give the same result as the cross-term between 1 and x^2, for example.

That leaves you with Integral{w(x) \* x^n dx} = 0 for all n \> 0. I guess W(x) = delta(x) works, but that’s boring. Maybe something oscillating infinitely many times as x -\> 1, like W(x) = cos(1/(1-x))\*, could work over the range +/- 1.

- I didn’t work out those integrals, so no complaining if they aren’t zero.

ETA: For your more general question, I think any set polynomials where more than one is non-negative would have the same sort of problems.

---

<div class="post-metadata">

**Author:** ![Hari\_Seldon](https://sea3.discourse-cdn.com/straightdope/user_avatar/boards.straightdope.com/hari_seldon/32/5173_2.png) [@Hari\_Seldon](https://boards.straightdope.com/u/Hari_Seldon)\
**Post date:** [February 5, 2013, 9:08pm UTC](https://boards.straightdope.com/t/orthogonal-polynomials/649396/3 "2013-02-05T21:08:22Z")

</div>

While Zen Beam is obviously correct, a better place to ask such technical questions is a site called Math Overflow where I have found useful answers to questions.
