MathOverflow is a question and answer site for professional mathematicians. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I would like to ask if there are any set of functions $u_n(x)$ which is orthogonal to $x^n$? i.e.:

$\int_0^1 x^n u_m(x) dx = \delta_{n,m}$

Edit: For clarification, this question asked for all non-negative integer m and n.

share|cite|improve this question
You should be able to prove easily that a continuous function which is orthogonal to all xn is zero. Using that, you can answer your question. (This applies to continuous functions; if you want other kind of regularity, well, you'll have to adapt the argument) – Mariano Suárez-Alvarez Jul 18 '10 at 15:17
the answers seem to reflect some confusion over whether the orthogonality is also over all n or is fixed for each n. – Suresh Venkat Jul 19 '10 at 3:01
up vote 8 down vote accepted

If $f \in L^2([0, 1])$ and $\int_0^1 x^n f(x)\, dx=0$ for all $n\ge N$ where $N$ is a nonnegative integer then $f$ is zero almost everywhere. To see this note that $x\mapsto x^N f(x)$ is an $L^2$ function orthogonal to all polynomials, and the polynomials are dense in $L^2([0,1])$. So the answer to your question is "no" for $L^2$ functions.

share|cite|improve this answer
Thank you for you answer. A follow up question: what if the interval changes? maybe to [-1,1] , or other interval? – Ross Tang Jul 21 '10 at 4:24
The argument remains exactly the same for any bounded interval. – Robin Chapman Jul 21 '10 at 5:20
Thank you very much. – Ross Tang Jul 22 '10 at 15:53

The answer is no, and the main reason is that $\{x^n\}_{n = 0}^{\infty}$ form a total set in $L^2([0,1])$ so the set of their finite linear combinations is dense. But $\int x^n dx > 0$ for $n \geq 0$.

I believe, the best one can do is apply Gram--Schmidt to $x^n$ and obtain a sequence of polynomials $p_n$ (the orthogonal polynomials) of degree $n$ such that $$ p_n \perp x^m,\quad m > n. $$ Here $f \perp g$ means $\int f(x) g(x) dx= 0$.

However, the notion of "best" here is not well-defined. It's just the usual choice.

share|cite|improve this answer

I'll answer this question for the interval [-1,1], but you should be able to extrapolate/specialize it to your case.

Let me define the inner product $\langle f,g\rangle=\int_{-1}^1 f(x)g(x) dx$ and normalized Legendre polynomials $P_m(x) = \sum_{n=0}^m p_{mn}x^n$ such that $\langle P_m,P_{m'}\rangle=\delta_{mm'}$. Note that $p_{mn}=0$ for $m>n$, so that the matrix $P$ with coefficients $p_{mn}$ is upper triangular. Let $Q$ be the inverse matrix, with components $q_{nm}$, that is, $x^n = \sum_{m=0}^n q_{nm}P_m(x)$. Clearly, $Q$ is also upper triangular. The triangularity matters because $Q$ can be found by explicit calculation, even though both $P$ and $Q$ are infinite dimensional.

It easily follows that $\langle x^n,P_m\rangle = q_{nm}$. It immediately follows that $\langle x^n,\sum_m p_{mn'} P_m\rangle = \sum_m q_{nm}p_{mn'} = \delta_{nn'}$. In other words, the functions you want are $u_{n}(x) = \sum_{m=0}^n p_{mn} P_m(x)$.

share|cite|improve this answer
These $u_m$ will satisfy $\int x^n u_m(x) dx=\delta_{m,n}$ for $n\le m$ but not necessarily for $n>m$. – Robin Chapman Jul 19 '10 at 6:21

Derivatives of the Dirac distribution (appropriately normalized).

share|cite|improve this answer
I believe the question was about functions, not distributions. – S. Carnahan Dec 7 '10 at 2:54

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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