Bivariate dimension quasi-polynomials of difference-differential field extensions with weighted basic operators (Q2319452)

The present review is about the four following papers by Alexander Levin, one coauthored with Alexander Egrafov: \begin{itemize} \item I: \textit{A. Levin}, Math. Comput. Sci. 13, No. 1--2, 157--168 (2019); \item II: \textit{A. Levin}, J. Symb. Comput. 102, 173--188 (2021; Zbl 1456.12004); \item III: \textit{A. Levin}, Math. Comput. Sci. 14, No. 2, 361--374 (2020; Zbl 1456.12003) \item IV: \textit{A. Evgrafov} and \textit{A. Levin}, Math. Comput. Sci. 14, No. 2, 347--360 (2020; Zbl 1456.12001). \end{itemize} We shall also have to refer to some more ancient works: \begin{itemize} \item \textit{A. Einstein}, The meaning of relativity. Fourth ed. Princeton: Princeton University Press (1953; Zbl 0050.21208) \item \textit{E. R. Kolchin}, Bull. Am. Math. Soc. 70, 570--573 (1964; Zbl 0144.03702) \item \textit{M. V. Kondratieva} et al., Differential and difference dimension polynomials. Dordrecht: Kluwer Academic Publishers (1999; Zbl 0930.12005) \item \textit{A. B. Levin}, Difference algebra. New York, NY: Springer (2008; Zbl 1209.12003). \end{itemize} The four papers under review are related to systems of difference, resp. difference-differential equations; and, more precisely, to the number of arbitrary parameters in a generic solution of such systems. Differential algebra evolved from commutative algebra, especially in its algebraic geometry component; and difference algebra evolved from differential algebra. We briefly recall the history of the relevant concepts. The number of arbitrary parameters in a generic solution of a system of algebraic equations in \(m\) variables is the \textit{dimension} of the corresponding algebraic subvariety of the affine space of dimension \(m\). This dimension, as well as other important geometric invariants (such as the multiplicity of a solution) are encoded in the \textit{Hilbert polynomial}, a powerful tool rooted in commutative algebra, but resting in a linear point of view as follows: for some graded module \(\bigoplus M_n\), the function \(n \mapsto \dim_K M_n\) (\(K\) the base field) is polynomial for \(n \gg 0\) (meaning \(n\) ``big enough''). See the book of \textit{D. Eisenbud}, ``Commutative algebra. With a view toward algebraic geometry.'' Berlin: Springer-Verlag (1995; Zbl 0819.13001) for details. The corresponding concepts for systems of differential equations were introduced and developed by Kolchin in 1964: the \textit{differential dimension} is ``the number of arbitrary functions [...] in the solution of the system [...]'' [Kolchin (loc. cit.)]. The \textit{differential dimension polynomial} also appears there, in order to replace the Hilbert polynomial. Note that a more down-to-earth, yet maybe more detailed notion had been previously introduced by Einstein in 1953 under the name of \textit{strength} of a system of partial differential equations [Einstein (loc. cit.)]. For a precise definition of the \textit{Einstein strength}, see [Kondratieva et al., (loc. cit.)] and [Levin I]. The former book develops Kolchin theory and builds a difference analogue; the latter lies entirely within the frame of difference algebra. The four papers under review here deal with the computation of dimension polynomials. The first three involve systems of difference-differential equations, while the fourth involves ``pure'' difference equations. As is the case for Hilbert polynomials in commutative algebra and algebraic geometry, the techniques required include: theory of polynomials taking only integer values; some combinatorics of lattices (used to model the properties of multi-index exponents, weighted or not, of multi-variate polynomials); and a theory of reduction of polynomials (compare e.g. with Gröbner bases) in which a key ingredient is the construction of \textit{characteristic sets} [\textit{M. V. Kondratieva} et al., (loc. cit.); Levin I]. We now describe more precisely the framework of the first article [ Levin I]. Then we shall explain the variations for the three other ones. A \textit{difference-differential field} is a field \(K\) endowed with a set \(\Delta := \{\delta_1,\ldots,\delta_m\}\) of derivations and a set \(\sigma := \{\sigma_1,\ldots,\sigma_n\}\) of automorphisms (Thus the setting described here is that of \textit{inversive} difference fields; but actually some parts of the theory are valid for rings, assuming only that the \(\sigma_j\) are injective endomorphisms (which is of course automatic in the case of fields).), all the \(\delta_i\) and \(\sigma_j\) commuting among themselves. More precise terminology is: \(\Delta-\sigma\)-field. Notions of \(\Delta-\sigma\)-subfield, extension field, homomorphism are easily defined; as well as those of \(\Delta-\sigma\)-polynomials (generalizing differential polynomials), independence and transcendence degree. Write \(T := \{\delta_1^{k_1} \cdots \delta_m^{k_m} \sigma_1^{l_1} \cdots \sigma_n^{l_n}\}\), all \(k_i, l_j \in \mathbf{N}\), the free commutative semigroup on \(\Delta \cup \sigma\). For integers \(r,s\), let \(T(r,s)\) the subset of those \(\tau := \delta_1^{k_1} \cdots \delta_m^{k_m} \sigma_1^{l_1} \cdots \sigma_n^{l_n}\) such that \(\text{ord}(\tau) \leq r\) and \(\text{ord}'(\tau) \leq s\), where \(\text{ord}(\tau) := \sum k_i\) and \(\text{ord}'(\tau) := \sum l_j\). For any \(\Delta-\sigma\)-extension \(L\) of \(K\) generated by \(\eta_1,\ldots,\eta_p\), and any integers \(r,s\), let \(L_{r,s}\) the subextension generated by all the \(\tau \eta_1,\ldots,\tau \eta_p\), \(\tau \in T(r,s)\). Then the main theorem of [Levin I] article is: \textit{For \(r,s >>0\), the transcendence degree of \(L_{r,s}\) is polynomial as a function of \(r,s\).} In [II], the global setting is the same but there is only one difference automorphism \(\sigma\). Then the above theorem is refined in the following way: The polynomial above has degree \(1\) in \(s\); moreover an explicit formula is given for its two coefficients (polynomials in \(r\)) from some combinatorial computations. In [Levin III], the global setting is that of [Levin I], but the sets \(\Delta\) and \(\sigma\) are respectively partitioned in families of subsets \((\Delta_i)_{i= 1,\ldots,p}\), \((\sigma_j)_{j = 1,\ldots,q}\). All weights are taken to be \(1\) and partial weights \(\text{ord}_i(\tau)\), resp. \(\text{ord}'_j(\tau)\) are defined relative to the \(\Delta_i\), resp. the \(\sigma_j\), yielding subsets (We change slightly the notations to emphasize homogeneity with [Levin I].) \(T(r_1,\ldots,r_p,s_1,\ldots,s_q)\) of \(T\) defined by conditions \(\text{ord}_i(\tau) \leq r_i\), resp. \(\text{ord}'_j(\tau) \leq s_j\). Subextensions \(L(r_1,\ldots,r_p,s_1,\ldots,s_q)\) of \(L\) are defined accordingly. Then: \textit{The transcendence degree of \(L(r_1,\ldots,r_p,s_1,\ldots,s_q)\) is polynomial as a function of the \(r_i,s_j\) (for values \(>> 0\)).} Finally, in [Levin IV], \(\Delta\) is empty (no derivation, ``pure difference'' case). But then the corresponding dimension polynomial is related to Einstein's strength and application is given to equation arising from natural sciences.