Invariants for a family of discrete equations with the Laurent property
Andrei K. Svinin · 2026
Read Invariants for a family of discrete equations with the Laurent property by Andrei K. Svinin (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2606-24925.txt.
Introduction
Andrei Svinin constructs invariants for a recently found family of discrete equations. The equations are homogeneous and of odd order, and they hold the Laurent property. The family's first member is the famous Somos-5 recurrence. Its integer sequence, catalogued as A006721, is widely used. A companion paper introduced the family; here the focus is hidden conserved quantities. Svinin builds a finite set of independent invariants for the equations.
Examples show what they buy. The Laurent property alone guarantees a certain form, but whether the sequences stay integral can fail or hold for subtler reasons. With the invariants in hand, a sharper test for integrality becomes available. Svinin's construction thus sharpens the theory behind a class of recurrences known for producing integers against expectations. The Somos-5 sequence sits inside the family as the guiding example.
Catalog record
| CREATOR | Andrei K. Svinin |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 42 KB |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2606.24925
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Invariants for a family of discrete equations with the Laurent property Andrei K. Svinin Abstract . Recently, we have found an infinite family of homogeneous discrete equations of odd order possessing the Laurent property. The first representative of this family is the well-known Somos-5 equation, which under certain conditions generates the integer sequence A006721, which has numerous applications. In this work, we construct a finite set of independent invariants for our equations. We show, through examples, that the presence of these invariants allows us to find a more general criterion for the integrality of sequences compared to what the usual Laurent property provides. Introduction In 1989, Michael Somos, in his research on elliptic functions, discovered several nonlinear recurrence relations which, despite their nonlinearity, generate integer sequences for suitable initial data. One of them, which is directly relevant to the present work, is the bilinear recurrence relation t_{n}t_{n+5}=t_{n+1}t_{n+4}+t_{n+2}t_{n+3}, now known as Somos-5. Somos posed the problem of proving – and, even better, explaining – this integrality. Specifically for the Somos-5 equation, one needed to prove the integrality of the sequence defined by the initial conditions t_{j}=1 for j=0,1,2,3,4 . In the OEIS online encyclopedia, this sequence is identified as A006721. It later turned out that this sequence has numerous diverse applications, listing which would require a separate survey. We give key references here. In [ 2 ] , a connection was found between the sequence A006721 and an infinite family of Heronian triangles with two rational medians. The same paper mentions a relation of this sequence to an elliptic curve and related special functions. In [ 4 , 13 ] , one can find combinatorial interpretations of the terms of the sequence A006721. This sequence is also of interest from the viewpoint of number theory [ 3 , 12 ] . Somos’s discovery sparked a wave of research, since the nature of the integrality of these sequences remained unclear for some time. David Gale, in his article [ 7 ] , gave a survey of the state of affairs at that time. One of his observations was that there actually exist quite many such polynomial recurrence relations and several examples had been found. In particular, he mentioned an infinite family of recurrence relations of Raphael Robinson: t_{n}t_{n+N}=t_{n+1}t_{n+N-1}+t_{n+2}t_{n+N-2},\;\;N\geq 4. Gale, in turn, conjectured that the recurrence relation t_{n}t_{n+N}=\alpha t_{n+a}t_{n+N-a}+\beta t_{n+c}t_{n+N-c}, (0.1) for any N\geq 4 and 1\leq a<c\leq\left\lfloor N/2\right\rfloor , with arbitrary integer coefficients \alpha and \beta and unit initial conditions, also generates an integer sequence. Nowadays, this relation is called the Gale–Robinson equation. More precisely, it is an infinite family of equations, where N is the order of the corresponding equation. Subsequently, the Gale–Robinson equations gained enormous popularity due to applications in combinatorics and theoretical physics. There currently exists an extensive amount of literature devoted to the combinatorial meaning of the integer sequences generated by Gale–Robinson equations, including connections to theoretical physics [ 4 ] . Perhaps the most intuitive interpretation of the sequences generated by Gale–Robinson equations with coefficients \left(\alpha,\beta\right) equal to unity was found in the works [ 1 , 13 ] . Definition 0.1 . We say that ( 0.1 ) is a Gale–Robinson equation of type \left(N,a,c\right) . A major breakthrough in the problem of explaining the integrality of sequences generated by certain polynomial recurrence relations was achieved by Andrei Zelevinsky and Sergey Fomin. In the early 2000s, they were developing their theory of cluster algebras. In [ 5 ] , using this theory, they proved, as a special case, the Gale–Robinson conjecture [ 5 ] . The essence is that these equations have the Laurent property, which Fomin and Zelevinsky discovered in their developing theory of cluster algebras, and the Gale–Robinson equations with \left(\alpha,\beta\right)=\left(1,1\right) represent recurrence relations naturally arising in the theory of cluster algebras. Consider the recurrence relation t_{n}t_{n+N}=L\left(t_{n+1},\ldots,t_{n+N-1}\right), (0.2) where it is assumed that L is some (Laurent) polynomial in its variables. Definition 0.2 . Equation ( 0.2 ) has the Laurent property if t_{n}\in\mathbb{Z}[t_{0}^{\pm 1},\ldots,t_{N-1}^{\pm 1}],\;\;\forall n\in\mathbb{Z} . Remark 0.3 . In the literature, for virtually all equations with the Laurent property of the form ( 0.2 ), the right-hand side L is a polynomial in its variables. As an elementary consequence of the Laurent property, we obtain that if t_{j}=\pm 1 for j=0,\ldots,N-1 , then the corresponding recurrence relation generates some integer sequence. This condition is sufficient for integrality of the sequence, though it is not necessary. Despite the fact that the condition in Definition 0.2 looks extremely restrictive, it turns out that there exist quite many recurrence relations with the Laurent property. Such equations either follow from the theory of cluster algebras, as for example in [ 6 ] , or within the framework of the approach of Lam and Pylyavskyy [ 11 ] , which generalizes the theory of cluster algebras. Both approaches are, in essence, combinatorial. In a recent paper [ 18 ] , we presented an infinite family of homogeneous recurrence relations of the form t_{n}t_{n+2g+3}=\frac{\sum_{j=0}^{g}\alpha_{j}B^{j}_{2g-j}(n+1)}{\prod_{j=3}^{2g}t_{n+j}},\;g\geq 1, (0.3) for which, as we proved, the Laurent property holds. Note that relation ( 0.3 ) represents an equation of the form ( 0.2 ) with parameters \alpha_{j} , which are considered arbitrary. The right-hand side of ( 0.3 ) is defined by certain homogeneous discrete polynomials B^{k}_{s}(n) in t_{n} . 1 1 1 We regularly use the term discrete polynomial. In this context it means that B^{k}_{s}(n) is some polynomial in a finite number of variables t_{n},t_{n+1},\ldots These polynomials are defined by means of a recurrence relation. We will write this relation in the next section; for now we only note that B^{j}_{2g-j}(n) is a homogeneous polynomial in the variables \left(t_{n},\ldots,t_{n+2g+1}\right) . The degree of this polynomial is 2g . Thus, relation ( 0.3 ), for any g\geq 1 , is a homogeneous equation. Note that the right-hand side of our equations of the form ( 0.3 ) is, generally speaking, not a polynomial, but a Laurent polynomial. For clarity, we write out the first two representatives of this family. For g=1 , the product in the denominator on the right-hand side of ( 0.3 ) should be taken as unity. In this case, ( 0.3 ) turns into the Somos-5 equation with parameters, namely 2 2 2 Hereafter we call this equation simply the Somos-5 equation. t_{n}t_{n+5}=\alpha_{0}t_{n+2}t_{n+3}+\alpha_{1}t_{n+1}t_{n+4}. (0.4) The next representative of this family is the equation t_{n}t_{n+7}=\alpha_{0}t_{n+2}t_{n+5}+\alpha_{1}\frac{t_{n+1}t_{n+4}^{2}t_{n+5}+t_{n+2}^{2}t_{n+5}^{2}+t_{n+2}t_{n+3}^{2}t_{n+6}}{t_{n+3}t_{n+4}}+\alpha_{2}t_{n+1}t_{n+6}. (0.5) As already noted, Somos-5, like all Gale–Robinson equations ( 0.1 ), fits well into the theory of cluster algebras [ 6 ] . However, for our equations of the form ( 0.3 ), such a combinatorial interpretation is not yet known. Let us highlight a key fact about our equations of the form ( 0.3 ), which can be viewed, on the one hand, as a property of these equations, and on the other hand, as their definition. Let \left(u_{n}\right)_{n\in\mathbb{Z}} be the unknown sequence. For each g\geq 1 , we define the associated recurrence relation \prod_{j=0}^{2g}u_{n+j}=\sum_{j=0}^{g}\alpha_{j}T^{j}_{2g-j}(n+1), (0.6) where T^{k}_{s}(n) are certain discrete homogeneous polynomials in u_{n} , which will be defined in the next section. The fact is that by means of the substitution u_{n}=\frac{t_{n}t_{n+3}}{t_{n+1}t_{n+2}} (0.7) the relation ( 0.6 ) reduces to the homogeneous form ( 0.3 ). In