\(u\)-invariants for forms of higher degree (Q1004604): Difference between revisions

In this paper, the author considers finite dimensional forms of degree \(d\geq 2\) over fields \(k\) (generally assumed to be of characteristic zero), in other words homogeneous polynomials of degree \(d\) in \(n<\infty\) variables, that are assumed to be nondegenerate, i.e. the symmetric \(d\)-linear map that is naturally associated to such a form is nondegenerate. \(n\) is called the dimension of the form. A form of degree \(d\) in the variables \(x_i\), \(1\leq i\leq n\), is called diagonal if it is of the shape \(a_1x_1^d+\ldots +a_nx_n^d\), \(a_i\in k\). The focus is the study of the invariants \(u(d,k)\) (resp. \(u_{\text{diag}}(d,k)\)), the supremum of the dimensions of all forms (resp. diagonal forms) of degree \(d\) over \(k\). In the case \(d=2\), these invariants are equal as quadratic forms can always be diagonalized in characteristic not \(2\), and they have been studied extensively. However, forms can generally not be diagonalized if \(d\geq 3\), so the values of these two invariants might differ. Most of the many results in this paper are concerned with relations between these invariants for a fixed field, and giving estimates for these invariants in the case of particular types of fields, such as finite and local fields (or more generally, fields with a discrete or Henselian valuation), and for rational function fields over such fields. A nice feature of this article is that it also gives an overview of the numerous results on or related to this question that are scattered throughout the existing literature.