Linear degree growth in lattice equations (Q2297884): Difference between revisions

The authors present a characterisation of the linear growth of quad equations. Quad equations are special partial difference equations where the dependent variable take values on the vertices of a quadrilateral figure, namely: \[ Q\left( u_{l,m},u_{l+1,m},u_{l,m+1},u_{l+1,m+1} \right)=0. \] The background of the considered problem relies on the concept of \textit{algebraic entropy} introduced in [\textit{M. P. Bellon} and \textit{C. M. Viallet}, Commun. Math. Phys. 204, No. 2, 425--437 (1999; Zbl 0987.37007)] for ordinary difference equations and later extended to other kind of discrete systems, including multi-dimensional lattice equations. In short, given a system with some discrete degrees of freedom which is \textit{birational}, fixed an initial condition it is possible to compute the degrees of the iterates of the system. Calling the degree of the \(n\)-th iterate \(d_{n}\) the algebraic entropy is then defined as the limit: \[ \varepsilon = \lim_{n\to\infty} \frac{1}{n} \log d_{n}. \] An equation is said to be \textit{integrable} when \(\varepsilon=0\), that is if the asymptotic behaviour of \(d_{n}\) is sub-exponential. In particular when \(d_n\sim n\) as \(n\to\infty\), that is the growth is linear, the equation is said to be linearisable. Quad equations fall into the class of discrete systems which can be tackled with this method because the function \(Q\) is taken to be \textit{multi-linear} in the \(u_{l+i,m+j}\) variables. In the paper, initial conditions along two semi-infinite lines and staircases are considered. Then, the drop of degrees in the sequence of the iterations is characterised by the presence of the same factors in the numerator and denominator of \(u_{l,m}\). Explicitly, this can be explained in terms of projective coordinates \(u_{l,m}=x_{l,m}/z_{l,m}\) and then dividing by the g.c.d. of \(x_{l,m}\) and \(z_{l,m}\). This allows one to define a sequence of degrees \(\bar{d}_{l,m}\) which takes into account the factorisation. In [the authors, Nonlinearity 32, No. 2, 622--653 (2019; Zbl 1410.39016)] it was conjectured a recursive formula for the g.c.d.. This recursive formula allows one to prove the following linear recurrence relation for the sequence of degrees: \[ \bar{d}_{l-1,m-1} + \bar{d}_{l,m+1} + \bar{d}_{l+1,m} = \bar{d}_{l+1,m+1} + \bar{d}_{l-1,m} + \bar{d}_{l,m-1}. \] In general this recurrence relation gives quadratic grows. However, the authors prove that a specialisation of this formula in several cases yields a linear behaviour. This kind of behaviour also reflects on the periodic reduction of these quad equations. Several explicit examples of linear growth are given, and some are proved through the explicit linearisation known from the literature, like for instance the Ramani-Joshi-Grammaticos-Tamizhmani (RJGT) equation, and the Adler-Startsev discretisation of the Liouville equation. Finally, the author search for quad equations with linear growth, starting from the most general quad equation. This search yields a possibly new quad equation with linear growth expressed as: \[ \frac{\left( a_{11}u_{l+1,m+1} +a_{6} \right)}{\left( a_{11}u_{l+1,m}+a_{6} \right)} \frac{\left( u_{l,m} +s_{2} \right)}{\left( u_{l,m+1}+s_{2} \right)} = -\frac{a_{11}}{a_{12}}\frac{\left( u_{l,m} +t_{2} \right)}{\left( u_{l+1,m}+t_{2} \right)}, \] where \(a_{11}\), \(a_{6}\), \(s_{2}\), and \(t_{2}\) are free parameters. This equation is linearisable through a Cole-Hopf transformation and it possesses a nonautonomous version which is a slight generalisation of the RJGT equation.