1
$\begingroup$

This question is somewhat related to this.

I am looking for an operator $f:\mathbb{R}[x]\to\mathbb{R}[x]$, that is, $f$ is an operator that maps polynomials in one variable to polynomials in one variable.

The requirement of $f$ is that it must map the monomial $x^q$ to the monomial $\frac{1}{q+1}x^q$, that is $f(x^q)=\frac{1}{q+1}x^q$. This operator should be independent of $q$.

I first thought on using integration, however I do not want this (see below).

My proposal is the following:

Let $P\in\mathbb R[x]$. We define $\phi:\mathbb R[x]\to\mathbb R[x]$ by $\phi(P)=\frac{d}{dx}(xP)$ (I think I can freely use $x$).

Next we define $f:\mathbb R[x]\to\mathbb R[x]$ by $f(P)=\frac{P^2}{\phi(P)}$. Note that in the particular case $P=x^q$ we do satisfy our requirement $$f(x^q)=\frac{x^{2q}}{(q+1)x^{q}}=\frac{1}{q+1}x^q.$$

--

Now, let us extend the previous idea to operators $f:\mathbb R[x_1,\ldots,x_n]\to\mathbb R[x_1,\ldots,x_n]$ .

Let $P\in \mathbb R[x_1,\ldots,x_n]$. Following the above stuff we define a family of maps $\phi_k:\mathbb R[x_1,\ldots,x_n]\to\mathbb R[x_1,\ldots,x_n]$ by $\phi_k(P)=\frac{d}{dx_k}(x_kP)$ (again I think I can freely use $x_k$, also $\phi_k(\cdot)=\phi(\cdot,k) $ with $k\in\mathbb N\backslash \left\{ 0 \right\}$).

Next, we define similarly $f_k(P)=\frac{P^2}{\phi_k(P)}$.

In the particular case that $P=x^Q=x_1^{q_1}\cdots x_n^{q_n}$, we have $$ f_k(x^Q)=\frac{(x_1^{2q_1}\cdots x_k^{2q_k}\cdots x_n^{2q_n}) }{ (q_k+1)(x_1^{q_1}\cdots x_k^{q_k}\cdots x_n^{q_n}) }=\frac{1}{q_k+1}x^Q. $$

What do you think?, any criticism is appreciated. Specially: is it true that if I define an operator on $P$, I can use $x_k$?


If $x\in\mathbb{R}$ it is OK to define $f(x^q)=x\int x^q dx=\frac{1}{q+1}x^q+c$.

However if $x\in\mathbb{R}^n$ then we would have $f_k(x^q)=x_k\int x^q dx_k=\frac{1}{q+1}x^q+\phi(x_1,\cdots,x_{k-1},x_{k+1},\cdots,x_n)$, where $\phi$ is a function of all terms $x_j$ except $x_k$.

$\endgroup$
1
$\begingroup$

It seems to me that the operator you seek would be: $$f(x^p)=\frac{1}{x}\int_0^x t^p dt$$ this has the desired properties.

Note that this operator is actually independant of $p$, and can be written more generally, for some $g\in\Bbb R[x]$ $$f(g)=\frac{1}{x}\int_0^x g(t)dt$$

$\endgroup$
  • $\begingroup$ This is indeed good for $g\in\mathbb R[x]$ as you say. What about $g\in\mathbb R[x_1,\ldots,x_n]$? ... is it possible to define $f_k(g)=\frac{1}{x_k}\int_0^{x_k}g(x_1,\ldots,x_{k-1},t,x_{k+1}\ldots,x_n)dt$ ? $\endgroup$ – PepeToro Jul 27 '14 at 17:35
  • $\begingroup$ If we consider that a monomial would now be $g(x_1,...,x_n)=x_1^{p_1}...x_n^{p_n}$ the let $$f(g)=\frac{1}{x^n}\int_0^{x_1}...\int_0^{x_n}g(t_1,...,t_n)dt_1...dt_n$$ but this depends on what you consider a monomial to now be, in honesty ive never considered what a multi variable monomial would be $\endgroup$ – Ellya Jul 27 '14 at 17:48
  • $\begingroup$ Just did a quick google, and what i have put for g is a multivariable monomial. $\endgroup$ – Ellya Jul 27 '14 at 17:51
  • $\begingroup$ Thanks, I think this helps a lot. I'll accept the answer once I've checked everything carries on well in my project. $\endgroup$ – PepeToro Jul 27 '14 at 18:05
1
$\begingroup$

EDIT: I'm quite sure that I misunderstood the OP's intention. However, the comments to this answer may still be of interest and so have made this a wiki answer.

No such function exists. We want $f(x^q)=x^q/(q+1)$ for all $q$; consider in particular $q=6$. Then we require \begin{align} f(x^{6}) &=f((x^2)^3))=f((x^3)^2))\\ &=\frac{x^{6}}{7}=\frac{(x^{2})^3}{3}=\frac{(x^{3})^2}{4}\end{align} which cannot be true since $7\neq 4\neq 3$.

$\endgroup$
  • $\begingroup$ Thanks. Do you think there is some way to require that the function "acts on $x^q$ and not in factorisations"? Or something like that? Does it makes a difference that I am not defining $f(P^q)=P/(q+1)$ but requiring that in the particular case that the argument is something "exactly like" $x^q$ I then obtain $x^q/(q+1)$ ? I set $f(P)=P^2/\phi(P)$, this is well defined right? $\endgroup$ – PepeToro Jul 27 '14 at 16:40
  • $\begingroup$ @user58533: I suspect the problem is that you're wanting to work with $f(x)$ as part of a polynomial ring, rather than as a function unto itself. That is: you want to find an operator $\hat{f}$ which acts on the monomials $x^q$, not a function which receives numerical inputs. Somehow that needs to be clear at the level of your definition, and I'm not sure what the right way is. $\endgroup$ – Semiclassical Jul 27 '14 at 16:43
  • $\begingroup$ that is correct, I am assuming/thinking of $f$ as an operator that receives polynomials, not numbers. $\endgroup$ – PepeToro Jul 27 '14 at 16:45
  • $\begingroup$ @user58533: Probably it's better to think of $\mathbb{R}[x]$ explicitly as a vector space. In those terms, you want an operator that maps the basis vector $e_q$ to $(q+1)^{-1} e_q.$ But what we're wrestling with is a notational issue. $\endgroup$ – Semiclassical Jul 27 '14 at 16:46
  • $\begingroup$ I need to think on this as I do not understand very well your argument. I thought I was sure I was working on "maps from the space of polynomials onto itself". But it seems I am doing something wrong. Thanks. $\endgroup$ – PepeToro Jul 27 '14 at 16:51

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.