Computational algebraic geometry of projective configurations (Q1176393)

The author presents an application of the Gröbner basis theory to the study of projective plane configurations. A configuration is a pair \({\mathcal C}=({\mathcal P},{\mathcal L})\) where \({\mathcal P}=\{1,2,\dots,n\}\) and \({\mathcal L}\) is a subset of \(2^{\mathcal P}\) with the condition ``\(l,l'\in{\mathcal L},l\neq l'\Rightarrow\)card \((l\cap l')\leq 1\)''. Let \(K\) be a field. A realization of \({\mathcal C}=({\mathcal P},{\mathcal L})\) over \(K\) is a mapping \(X:{\mathcal P}\to\mathbb{P}_ 2(K)\) (the projective plane over \(K)\) such that: \(\forall i,j,k\in{\mathcal P},\;i<j<k\Rightarrow(\text{det}(X(i),X(j),X(k))=0\Leftrightarrow\exists l\in{\mathcal L}|\{i,j,k\}\subset\ell)\). The main question in the paper may be formulated as follows: Let \({\mathcal C}=({\mathcal P},{\mathcal L})\) be a configuration. Is there a realization of \({\mathcal C}\) over \(\mathbb{C},\mathbb{R}\) or \(\mathbb{Q}\)? It can be supposed that \({\mathcal P}\) contains four points in ``general position'' (if not, the answer to the question is easy). The problem is to find a realization \(X\) with the conditions \(X(1)=(1,0,0)^ T,X(2)=(0,1,0)^ T\), \(X(3)=(0,0,1)^ T\), \(X(4)=(1,1,1)^ T\). Put \(X(i)=(x_{i1},x_{i2},x_{i3})^ T\) for \(5\leq i\leq n\) where \(x_{ij}\) is a variable and consider the polynomial ring \(K[(x_{ij})\), \(5\leq i\leq n,j=1,2,3]\). For each \(l\in{\mathcal L}\) and for each \(\{i,j,k\}\subset l\) put \(f_{ijk}=\text{det}(X(i),X(j),X(k))\). In general, given a finite subset \({\mathcal F}=\{f_ 1,\dots,f_ r\}\) of polynomials in \(K[\underline y]=K[y_ 1,\dots,y_ s]\), a final polynomial for \({\mathcal F}\) is a polynomial \(p(z_ 1,\dots,z_ r,y_ 1,\dots,y_ s)\in K[\underline z,\underline y]\) such that \(p(f_ 1,\dots,f_ r,\underline y)=0\) and \(p(\underline 0,\underline y)=1\) in \(K[\underline y]\). By Hilbert's Nullstellensatz, the set \(\mathcal F\) has either a final polynomial (there exists \(g_ 1(\underline y),\dots,g_ r(\underline y)\) such that \(\sum_ if_ ig_ i-1=0)\) or a common zero on \(\overline K^ s\) (where \(\overline K\) is an algebraic closure of \(K)\). --- The author uses the Buchberger algorithm to compute a Gröbner basis of a polynomial ideal and a final polynomial for \({\mathcal F}\) (when \({\mathcal F}\) has no common root in \(\overline K^ s)\). The proof of the following theorem is sketched: Let \({\mathcal F}=\{f_ 1,\dots,f_ r\}\) be a subset of \(K[\underline y]\) which has no zeros in \(\overline K^ s\) and let \({\mathcal G}'\) be a Gröbner basis for the set \({\mathcal F}':=\{f_ 1(\underline y)-z_ 1,\dots,f_ r(\underline y)-z_ r\}\) (with respect to the lexicographical order induced from \(z_ 1<\dots<z_ r<y_ 1<\cdots<y_ s)\). Then \({\mathcal G}'\) contains a final polynomial for \({\mathcal F}\). Given a configuration \({\mathcal C}=({\mathcal P},{\mathcal L})\), if the Gröbner basis \({\mathcal G}_{{\mathcal C}}'\), for the set \({\mathcal F}_{{\mathcal C}}':=\{f_{ijk}-z_{ijk}|\) \(\{i,j,k\}\subset l,l\in{\mathcal L}\}\) contains a final polynomial for \({\mathcal F}_{\mathcal C}:=\{f_{ijk}|\{i,j,k\}\subset l,l\in{\mathcal L}\}\) then \({\mathcal C}\) is not realizable over \(\mathbb{C}\). --- But even if \({\mathcal F}_{\mathcal C}\) has a common root in \(\overline K^{3(n-4)}\), but has not roots in \(K^{3(n- 4)}\), the computation of \({\mathcal G}'_{\mathcal C}\) can provide necessary conditions for field extension of \(K\) to allow roots of \({\mathcal F}_{\mathcal C}\). The author gives two examples: the configurations \({\mathcal C}_{\sqrt{-1}}\) and \({\mathcal C}_{\sqrt 2}\), which are not realizable over \(\mathbb{Q}\) but such that there exist polynomials \(x^ 2_{ij}+1\) in the ideal generated by \({\mathcal F}_{{\mathcal C}_{\sqrt{-1}}}\) and polynomials \(x^ 2_{ij}-2\) in the ideal generated by \({\mathcal F}_{{\mathcal C}_{\sqrt 2}}\). The article contains a discussion of some results concerning the complexity of realization algorithm and the computation of minimum degree for final polynomials. The last section contains an automated proof (based on a final polynomial) for the Desargues theorem.