On universal stably free modules in positive characteristic (Q2065606): Difference between revisions

Let \(K\) be a field. In [\textit{M. Raynaud}, Invent. Math. 6, 1--26 (1968; Zbl 0216.32601)], the rings \[ A_{n,q} = \frac{K[X_{ji} , Y_{ij} ; 1 \leq i \leq n , 1 \leq j \leq n-q]}{\left( \sum_{i=1}^n X_{ji} Y_{ik} = \delta_{jk} ; 1 \leq j,k \leq n-q \right)} \] are defined. The matrix \([[ \overline{X}_{ji} ]]\) defines a map of free modules \( A_{n,q}^n \rightarrow A_{n,q}^{n-q}\) and the matrix \([[ \overline{Y}_{ik} ]]\) defines a splitting. The kernel of the map defined by \([[ \bar{X}_{ji} ]]\) is thus a stably free projective module, denoted \(P_{n,q}\). It is clear that the pair \((A_{n,q} , P_{n,q} )\) is universal with this property over pairs of \(K\)-algebras \(A\) and stably free \(A\)-modules \(P\) such that \(P \oplus A^{n-q} \stackrel{\sim}{\rightarrow} A^n\). Also, \(\mathrm{Spec}(A_{n,q}) = \mathrm{GL}_n/ \mathrm{GL}_q\) and it follows from universality that \(P_{n,q}\) is free if and only if the fibration \(GL_n \rightarrow \mathrm{GL}_n/\mathrm{GL}_q\) has a section. Similarly, it follows that for \(r < q\) , \(P_{n,q}\) has a free direct summand of rank \(r\) if and only if the fibration \(GL_n/ GL_{q-r} \rightarrow \mathrm{GL}_n/ \mathrm{GL}_q\) has a section. It was shown in [\textit{M. Raynaud}, Invent. Math. 6, 1--26 (1968; Zbl 0216.32601)] that the following fibrations do not admit a section : \begin{itemize} \item[1.] \(\mathrm{GL}_n \rightarrow \mathrm{GL}_n/ \mathrm{GL}_{n-1}\) ; unknown exceptions : \(\mathrm{char}(K) = 2\) , \(n=3\) and \(\mathrm{char}(K) = 3\) , \(n=4\). \item[2.] \(\mathrm{GL}_n/ \mathrm{GL}_{n-q} \rightarrow \mathrm{GL}_n/ \mathrm{GL}_{n-1}\) ; unknown exceptions depending on \(\mathrm{char}(K)\) and \(n\). \item[3.] \(\mathrm{Sp}_{2n} \rightarrow \mathrm{Sp}_{2n}/ \mathrm{Sp}_{2n-2}\) , \(\mathrm{SO}_{2n+1} \rightarrow \mathrm{SO}_{2n+1}/ \mathrm{SO}_{2n-1}\) ; unknown exceptions : \(\mathrm{char}(K) = 3\) , \(n=2\) and \(\mathrm{char}(K) = 5\) , \(n=3\). \end{itemize} The proof is based on the computation of étale cohomology with finite coefficients and using Steenrod operations thereon. Note that if follows from work of Mohan Kumar and Nori that for \(n \geq 3\), the projective module \(P_{n,n-1}\) is not free (described in [\textit{R. G. Swan}, Ann. Math. Stud. 113, 432--522 (1987; Zbl 0708.55014)], or a more direct proof available at [\url{http://www.math.uchicago.edu/~swan/MKN.pdf}]) thus showing (1) above without exceptions. Relevant computations of motivic cohomology with finite coefficients and the study of Steenrod operations thereon for \(\mathrm{GL}_n\) were undertaken in [\textit{B. Williams}, J. \(K\)-Theory 10, No. 1, 141--163 (2012; Zbl 1264.19004)] when the characteristic of the coefficient field differs from \(\mathrm{char}(K)\) and some of the unknown exceptions above can be resolved using those computations. In a previous article [\textit{E. Primozic}, Forum Math. Sigma 8, Paper No. e52, 25 p. (2020; Zbl 1460.14054)] by the author, they define and study Steenrod operations over motivic cohomology with \(\mathbb{F}_p\) coefficients for smooth varieties over \(K\) when \(\mathrm{char}(K)=p\), thus laying the groundwork to apply similar techniques as in the original proof of the above results ((1), (2), (3)) without any exceptions. Indeed, in the article under review, the author carries out this program and gives a unified proof of the above results, removing all exceptions, by computing motivic cohomology with finite coefficients and Steenrod operations thereon without any assumptions on \(\mathrm{char}(K)\).