On a conjecture of Kato and Kuzumaki concerning Fano hypersurfaces (Q498673)

The paper under review proves that \(p\)-adic fields and totally imaginary number fields have the \(C^1_1\) property in the sense of \textit{K. Kato} and \textit{T. Kuzumaki} [J. Number Theory 24, 229--244 (1986; Zbl 0608.12029)]. More generally, given a finite field extension \(L/k\) and an integer \(q\geq 0\), one has the norm map \(N_{L/k}: K_q(L)\to K_q(k)\) on the Milnor \(K\)-groups. (For \(q=1\), this is the usual norm map \(L^*\to k^*\).) A field \(k\) is said to satisfy the \(C^q_i\) property if the following holds: For every finite extension \(K/k\) and every homogeneous polynomial \(f\in K[x_0,\dotsc, x_n]\) of degree \(d\) such that \(d^i\leq n\), the group \(K_q(K)\) is generated by the norm subgroups \(N_{L/K}(K_q(L))\), where \(L\) runs through finite extension of \(K\) in which \(f\) admits a nontrivial zero. Kato and Kuzumaki noticed that this notion has close relations with several dimension theories of fields (cohomogical dimension, Diophantine dimension, etc.) and as an analogue of a remarkable transition property in those theories, they proved the following fact: Let \(K\) be a complete discrete valuation field with perfect residue field \(F\) and let \(\ell\) be a prime number. If \(F\) satisfies the \(C_{i-1}^q\) and the \(C^{q-1}_i\) properties for forms of degree \(\ell\), then \(K\) satisfies the \(C_i^q\) property for forms of degree \(\ell\). In particular, it was shown that \(p\)-adic fields satisfy \(C^1_1\) for forms of \textit{prime} degrees. But whether this result extends to forms of arbitrary degrees was left open. To the reviewer's knowledge, no more progress has been made until the present paper. A key innovation of this paper is to relate the \(C^q_1\) property to a geometric invariant of proper varieties. For an algebraic scheme \(X\) over a field \(K\), denote by \(N_q(X/K)\) the subgroup of \(K_q(K)\) generated by norms from the \(K_q\)-groups of residue fields of closed points of \(X\). To say that a field \(k\) satisfies \(C_1^q\) is to say that for every finite extension \(K/k\) and every Fano hypersurface \(X/K\) one has \(N_q(X/K)=K_q(K)\), or equivalently, the exponent of the (torsion) group \(K_q(K)/N_q(X/K)\) is 1, which equals the Euler-Poincaré characteristic \(\chi(X,\,\mathcal{O}_X)\) of the structural sheaf \(\mathcal{O}_X\). From this point of view, the author introduces a generalized notation as follows: A field \(k\) satisfies the strong \(C_1^q\) property if for every finite extension \(K/k\), every proper scheme \(X/K\) and every coherent sheaf \(E\) on \(X\), the exponent of \(K_q(K)/N_q(X/K)\) divides the Euler-Poincaré characteristic \(\chi(X,\,E)\). If \(\ell\) is a prime number, one can define the restricted property at \(\ell\) (resp. outside \(\ell\)) by considering only the \(\ell\)-primary (resp. prime-to-\(\ell\)) torsion part. One of the main results of this paper is the following theorem: Theorem. Let \(K\) be a Henselian excellent discrete valuation field with residue field \(F\) and let \(\ell\) be a prime number invertible in \(F\). If \(F\) satisfies the strong \(C_1^{q-1}\) property at \(\ell\) and the \(C_0^q\) property at \(\ell\), then \(K\) has the property \(C^q_1\) property at \(\ell\). As a corollary, the field \(\mathbb{C}((x_1))\cdots((x_n))\) satisfies the strong \(C_1^{n-1}\) property, and the fields \(\mathbb{F}_p((x_1))\cdots((x_n))\) and \(\mathbb{Q}_p((x_1))\cdots((x_{n-1}))\) satisfy the strong \(C_1^n\) property outside \(p\). To obtain the above theorem, the author first proves that for every proper scheme \(X/K\) the exponent \(n_X\) of the \(\ell\)-primary component of \(K_q(K)/N_q(X)\) satisfies a general dévissage principle. By that dévissage method, it is sufficient to prove, for every coherent sheaf \(E\) on \(X\), the divisibility relation \(n_X\,|\,\chi(X,\,E)\) under the additional hypothesis that \(X\) admits a connected regular proper model over the ring of integers of \(K\) (so in particular \(X\) itself is regular and connected). In this last case the problem is solved by \(K\)-theoretic analysis. For \(p\)-adic fields, there are some subtleties in dealing with the \(p\)-primary part. This is done by studying the so-called residue index, another invariant which is shown to satisfy the dévissage principle and for which models with quotient singularities can be used to refine the previous method. The case of totally imaginary number fields is established by using a local-global principle obtained in [\textit{K. Kato} and \textit{S. Saito}, Ann. Math. (2) 118, 241--275 (1983; Zbl 0562.14011)]. Another remarkable application of the author's dévissage principle is the following result related to a conjecture of Ax: Let \(p\) be a prime number. If \(k\) is a perfect PAC field whose absolute Galois group is a pro-\(p\)-group, then \(k\) is a \(C_1\) field. Last but not least, this nicely written paper contains many interesting examples and remarks. For instance, the author notices that the field \(\mathbb{C}(x)\) does not satisfies the strong \(C_1^0\) property although it is \(C_1\) (and hence \(C_1^0\)). On the other hand, he exhibits an example of a field satisfying the strong \(C_1^0\) property but not the \(C_1\) property. The following is a relevant question that is left open in this paper: Do the fields \(\mathbb{C}(x)((y))\), \(\mathbb{C}((x))(y)\), \(\mathbb{C}((x,\,y))\) and \(\mathbb{C}(x,\,y)\) satisfy the \(C_1^1\) property?