Let $x$ and $y$ be variables. Consider the following recurrence: \begin{equation} u_{n}:= \begin{cases} \displaystyle{\frac{1+u_{n-1}^b}{u_{n-2}}} & if\ n\ \text{is even},\\ &\\ \displaystyle{\frac{1+u_{n-1}^c}{u_{n-2}}} & if\ n\ \text{is odd}. \end{cases} \end{equation} with $b,c\in\mathbb{N}^{+}$ (non zero naturals numbers), $u_{0}=x$ and $u_{1}=y$. Suppose $bc\geq 3$. Consider the homogeneous linear recurrence relations of order 4 given by the formulas: $$\alpha_{n}=(bc-2)\cdot\alpha_{n-2}-\alpha_{n-4},\ \ \forall n\geq 5$$ with initial conditions: $$(\alpha_{1},\alpha_{2},\alpha_{3},\alpha_{4})=(0,1,c,bc-1);$$
and $$ \beta_{n}=(bc-2)\cdot\beta_{n-2}-\beta_{n-4},\ \ \forall n\geq 5 $$ with initial conditions: $$ (\beta_{1},\beta_{2},\beta_{3},\beta_{4})=(0,1,b,bc-1).$$ I found $\alpha_n$ and $\beta_n$ experimentally. Notice that both recurrences $\alpha_n$ and $\beta_n$ reduce to the formula of Definition 1 from https://arxiv.org/pdf/1106.0952.pdf when $b=c$ (they call $r$ the common value when $b=c$). Using Maple I verified several cases where for some polynomials $P_{n,b,c}, Q_{n,b,c}\in\mathbb{R}[x,y]$ and for $n\geq 4$, we have the reduced expressions: $$ u_n=\frac{\displaystyle{P_{n,b,c}(x,y)}}{x^{\beta_{n-1}}y^{\alpha_{n-2}}},\quad n\geq 4$$ and also $$ u_{3-n}=\frac{\displaystyle{Q_{n,b,c}(x,y)}}{y^{\beta_{n-1}}x^{\alpha_{n-2}}},\quad n\geq 4$$

I was trying to tackle the following questions:

  1. Can we show that if the recurrence $u_n$ is periodic (that is, $u_{n+p}=u_{n}$ for all $n\in \mathbb{Z}$ and some fixed positve integer $p$) then $bc\leq 3$ without using Cluster Algebras and the Fomin-Zelevinsky results ? It is clear that $bc\leq 3$ implies that the recurrence for $u_n$ is periodic since it is well know that: $$bc=1 \Rightarrow p=5,\ \textrm{that is}\ \ u_{n+5}=u_{n},\ \textrm{for all}\ n\in\mathbb{Z}$$ $$bc=2 \Rightarrow p=6,\ \textrm{that is}\ \ u_{n+6}=u_{n},\ \textrm{for all}\ n\in\mathbb{Z}$$ $$bc=3 \Rightarrow p=8,\ \textrm{that is}\ \ u_{n+8}=u_{n},\ \textrm{for all}\ n\in\mathbb{Z}$$ for any non zero initial conditions $u_0=x$ and $u_1=y$ I proved this directly from definition. The converse result is a well known result of Fomin-Zelevinsky but I was wondering if we can find a method to prove it without Cluster Algebras.

  2. Can we find formulas for $P_{n,b,c}$ and $Q_{n,b,c}$ like: $$P_{n,b,c}(x,y)=\sum_{\gamma\in\mathcal{F}_n} x^{b\vert\gamma\vert_{1}}y^{c(\alpha_{n-1}-\vert\gamma\vert_{2})}$$ $$Q_{n,b,c}(x,y)=\sum_{\gamma\in\mathcal{F}_n} y^{b\vert\gamma\vert_{1}}x^{c(\alpha_{n-1}-\vert\gamma\vert_{2})}$$ where the sum is taken over a family $\mathcal{F}_n$ of subpaths of Dyck pahts ?

  3. How to show that $x^{\beta_{n-1}}y^{\alpha_{n-2}}$ is the denominator of the reduced fraction $u_n$. The same for $u_{3-n}$.

  4. Can we find a solution for $P_{n,b,c}$ and $Q_{n,b,c}$ using the notion of Greedy elements (see https://arxiv.org/pdf/1208.2391.pdf) ? Can we express these polynomials as a sum over a set of compatible pairs in $\mathcal{D}^{\beta_{n-1}\times\alpha_{n-2}}$ ?

  5. From that fact that both recurrences $\alpha_n$ and $\beta_n$ have the same characteristic polynomial: $$C_{b,c}(x)=x^4+(2-bc)x^2+1$$ whose factorization ca be written as: $$ x^4+(2-bc)x^2+1=(x^2-\sqrt{bc}\,x+1)(x^2+\sqrt{bc}\,x+1), $$ can we show somehow that $u_{n}$ is periodic for some $p$ if $bc\leq 3$, that is equivalent to the fact that $(x^2\pm\sqrt{bc}\,x+1)$ have complex roots, because the discriminant of $(x^2\pm\sqrt{bc}\,x+1)$ is $bc-4$. As $b$ and $c$ are naturals $bc<4$ iff $bc\leq 3$.

This is motivated by the work of Lee and Schiffler see https://arxiv.org/pdf/1106.0952.pdf and https://arxiv.org/pdf/1208.2391.pdf


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.