6
$\begingroup$

I'm very new to umbral caluclus and I have come across a paper that makes use of some results in this area, which I do not quite understand.

The problem I have is the following.

Consider the following operator

\begin{equation} \mathcal{S} = a^{-1}(bI\Delta_{-1} + c\Delta_{1}), \quad I \in \mathbb{Z^+} \end{equation}

where $\Delta_h[f(I)] =f(I+h)-f(I), h \in \mathbb{Z}$

It is claimed by the paper that given that $\Delta_{1}(I)_m=m(I)_{m-1}$ and $I\Delta_{-1}(I)_m = -m(I)_m$ [where with $(I)_m$ we denote the falling factorial] the eigenfunctions $\psi(I)$ of the operator are computable as:

\begin{equation} \psi_n(I) = \sum_{m=0}^{n} {n \choose m} \left(-\frac{c}{b}\right)^m(I)_{n-m} \end{equation} and the eigenvalues

\begin{equation} \lambda_n=-n\frac{b}{a} \end{equation}

Note that $\mathcal{S}(I)_m = -m\frac{b}{a}[(I)_m - \frac{c}{b}(I)_{m-1}]$

How is this caluculation performed? I have looked at Rota's book but I see no reference to the computation of the eigenfunctions. Can anybody point me out in the right direction?

Thank you all in advance

$\endgroup$
2
  • $\begingroup$ Pardon once again I have been wirting in a haste. I have edited the post appropriately, i.e. $I\Delta_{-1}(I)_m=-m(I)_m$ $\endgroup$ – Jpk Feb 18 '20 at 18:41
  • 1
    $\begingroup$ Can you prove a link to this paper? $\endgroup$ – PackSciences Feb 18 '20 at 20:01
2
$\begingroup$

Begin with the study of polynomials in $X$ and some linear operators. Define the shift linear operator $$ E_h[f(X)] \!:=\! f(X\!+\!h), \tag{1} $$ and the difference linear operator $$ \Delta_h[f(X)] := f(X\!+\!h)\!-\!f(X), \tag{2} $$ and the derivative linear operator $$ D[f(X)] := \frac{d}{dX} f(X), \tag{3} $$ where $\,f()\,$ is any polynomial.

Define the falling factorial linear operator for monomials $$ L[X^n] := X(X-1)\dots(X-n+1). \tag{4} $$ Without loss of generality, define the $\,S\,$ linear operator by $$ S[f(X)] := (X\Delta_{-1}+c\,\Delta_{1})[f(X)]. \tag{5}$$ Applying this to falling factorials gives $$ S[(X)_n] = -n ( (X)_n - c\, (X)_{n-1}). \tag{6} $$ Rewriting this using the $\,L\,$ operator gives $$ S[L[X^n]] = -n L[ X^{n-1} (X-c)]. \tag{7} $$

Rewrite this using the derivative operator gives $$ S[L[f(X)]] = -L[ D[f(X)](X-c)]. \tag{8} $$ Apply this to the case $\,f(X)=E_h[X^n]\,$ to get $$ S[L[E_h[X^n]]] = -L[D[E_h[X^n]](X-c)]. \tag{9} $$ Apply this to the case $\,h=-c\,$ to get $$ S[L[(X-c)^n]] = -L[D[(X-c)^n](X-c)]. \tag{10} $$ But we know the derivative of $\,(X-c)^n\,$ and so get $$ S[L[(X-c)^n]] = -n\,L[(X-c)^n]. \tag{11} $$ Define the eigenvector function $$ \psi_n(X) \!:=\! L[(X\!-\!c)^n] \!=\! \sum_{m=0}^n {n \choose m} (-c)^m(X)_{n-m}. \tag{12} $$ with eigenvalue $\,-n\,$ and is expanded into a finite sum using the binomial theorem.

$\endgroup$
4
  • $\begingroup$ Thank you very much for this. This clearly required some practice and knowledge of linear operators with which I am not familiar. Do you happen to have a good reference on the matter? Thank a lot in advance. $\endgroup$ – Jpk Feb 19 '20 at 10:33
  • 1
    $\begingroup$ @Jpk I suggest the Wikipedia article Umbral calculus as a starting point and its references. $\endgroup$ – Somos Feb 19 '20 at 16:48
  • $\begingroup$ If I set $c$ to be dependent on x the same relationship does not hold as the derivative of the composed function breaks the relation you have written. Is the any other way to compute the eigen functions when $c =c(x)$? $\endgroup$ – Jpk Jun 14 '20 at 23:34
  • 1
    $\begingroup$ @Jpk That is a different question which you may want to submit. I don't know the answer. $\endgroup$ – Somos Jun 15 '20 at 1:28
1
$\begingroup$

In the article mentionned in comment and not in the question, the author give details about the eigenvalues of $S$.

$$S_{x,y} = -y \frac{b}{a} \delta_{x,y} + y I \frac{b}{a} \delta_{x,y-1}$$

Now the author is implicitely using the fact that $S$ is triangular, therefore its eigenvalues are the diagonal components. I can provide a proof if needed but it is part of the properties you can find easily on any article about triangular matrices.

Now I did not perform the calculation of the eigenfunction, but I believe it is the general method.

$\endgroup$
1
  • $\begingroup$ The computation of the eigenvalues once the change of basis is done is clear. I'm more perplexed by the calculation of the eigenfunctions; what do you mean by general method? [I'm not familar with this part of maths] $\endgroup$ – Jpk Feb 18 '20 at 21:25

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.