4
$\begingroup$

Let $r>1$ be real and \begin{align*} f_1(x) &= 1,\\ f_2(x) &= (x+4)^r,\\ f_3(x) &=(x+4)^r(x+3)^r,\\ f_4(x) &= (x+4)^r(x+3)^r(x+2)^r,\\ f_5(x) &=(x+4)^r(x+3)^r(x+2)^r(x+1)^r. \end{align*} How can I show that these function form a T-system on $(0,\infty)$, i.e., $$\det[f_j(x_i)]\ne 0$$ for every mutually distinct $0<x_1,\ldots,x_5$?

I am looking for a good hint for a solution (not a full solution) to this, and its generalization.

$\endgroup$
4
  • $\begingroup$ Just a remark that $f_{6-j}(x)=(\Gamma(x+5)/\Gamma(x+j))^r$. So up to a non-zero factor, your determinant is $\det[\Gamma(x_i+j)^{-r}]$. Perhaps there are general positivity results for $\det[\Gamma(x_i+y_j)^{-r}]$ that could be applied? $\endgroup$ Commented Nov 11 at 14:12
  • $\begingroup$ positivity results, like what? $\endgroup$ Commented Nov 11 at 14:34
  • $\begingroup$ I mean like in your previous question on the case $r=-1$, mathoverflow.net/q/306366/10846 . Perhaps there are general results implying that $\det[\Gamma(x_i+y_j)^{-r}]$ is positive (or negative) for some appropriate range of the variables. $\endgroup$ Commented Nov 11 at 14:49
  • $\begingroup$ I have tried many, but it didn't help. I don't think I should go for those complex solutions. $\endgroup$ Commented Nov 11 at 14:52

1 Answer 1

1
$\begingroup$

A simple sufficient criteria is to verify that the system of functions is $M_{+}$ system, i.e., for each $k=1, \ldots, 4$ the determinants of the matrices $\{f_{i}^{(j-1)}(t)\}_{i,j=1}^{k+1}$ are positive on $(0, \infty)$ (see the discussion right before Theorem 1.1 here and the references therein https://arxiv.org/pdf/2201.12932 )

One could write a script to verify signs of these simple determinants. I used LLMs to do this computation. I checked it and it looks correct to me (but please also double check for yourself). Here are the Wronskians:

Let $$ f_1(x)=1,\quad f_2(x)=(x+4)^r,\quad f_3(x)=(x+4)^r(x+3)^r,\quad f_4(x)=(x+4)^r(x+3)^r(x+2)^r,\quad f_5(x)=(x+4)^r(x+3)^r(x+2)^r(x+1)^r . $$ We denote by $ \Delta_k(x)=\det\big(f_i^{(j-1)}(x)\big)_{i,j=1}^{k+1} $ the $(k+1)\times (k+1)$ Wronskian determinant.
Since $f_1\equiv1$, each $\Delta_k$ reduces to the usual Wronskian of $(f_2,\dots,f_{k+1})$.

Order $k=1$ $$ \Delta_1(x)=W(f_2)=f_2'(x)=r(x+4)^{r-1}>0. $$

Order $k=2$. Using the identity $W(u,ug)=u^2 g'$ with $u=(x+4)^r$, $g=(x+3)^r$, we get $$ \Delta_2(x)=W(f_2,f_3) =r(x+3)^{\,r-1}(x+4)^{\,2r}>0. $$

Order $k=3$. A direct computation yields $$ \Delta_3(x) =\frac{r^2\,(x+2)^{\,r}(x+3)^{\,2r}(x+4)^{\,3r}}{(x+2)^2(x+3)^2}\, \bigl(2r x+5r-1\bigr), $$ and since $r>1$ one has $2r x+5r-1>0$ for all $x>0$. Hence $\Delta_3(x)>0$.

Order $k=4$. A longer but straightforward calculation gives $$ \Delta_4(x) =\frac{r^3\,(x+1)^{\,r}(x+2)^{\,2r}(x+3)^{\,3r}(x+4)^{\,4r}} {(x+1)^3(x+2)^4(x+3)^3}\, Q_r(x), $$ where $$ \begin{aligned} Q_r(x)=&\;12r^3x^4+(96r^3-24r^2)x^3+(281r^3-144r^2+19r)x^2\\ &\quad+(356r^3-284r^2+76r-4)x+(165r^3-184r^2+75r-8). \end{aligned} $$ For every $r>1$ all coefficients of $Q_r(x)$ are positive, hence $Q_r(x)>0$ on $(0,\infty)$.

Therefore, for $r>1$ we have $\Delta_k(x)>0$ for $k=1,2,3,4$ and all $x>0$, so $(f_1,\dots,f_5)$ satisfies the sufficient Wronskian positivity condition for an $M_{+}$ (or $T_{+}$) system on $(0,\infty)$.

$\endgroup$
4
  • $\begingroup$ This proof has a lot of calculations, and I am sure the generalisation of this problem (for 6, 7, ..... functions) will have more complicated steps using this method. I am looking for some induction method kind of proof. $\endgroup$ Commented Nov 12 at 14:26
  • 1
    $\begingroup$ Yes, it has some calculation in it. But from my experience working with various Tchebyshev and Markov systems I think checking signs of Wornskians is really the least amount of computation for this kind of problem one has to do. $\endgroup$ Commented Nov 12 at 14:40
  • $\begingroup$ Do you have an alternate solution, or hint? $\endgroup$ Commented Nov 16 at 12:02
  • $\begingroup$ I asked for a hint, which could be useful for this and its generalisation as well, but this proof is not helpful. $\endgroup$ Commented Nov 16 at 12:04

You must log in to answer this question.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.