Triple linkage of quadratic Pfister forms (Q1785560)

A binary bilinear form over a field \(F\) with diagonalization \(\langle 1,\beta\rangle\) is called a bilinear \(1\)-fold Pfister form, denoted by \(\langle\!\langle \beta\rangle\!\rangle\) for short, and an \(n\)-fold tensor product \(\bigotimes_{i=1}^n\langle\!\langle \beta_i\rangle\!\rangle=\langle\!\langle \beta_1,\ldots,\beta_n\rangle\!\rangle\) is called a bilinear \(n\)-fold Pfister form. If the characteristic of \(F\) is different from \(2\), quadratic and bilinear forms can be identified in a natural way. In characteristic \(2\), there is also the notion of quadratic \(n\)-fold Pfister forms that are of shape \(\langle\!\langle \beta_1,\ldots,\beta_{n-1},\alpha ]]=\langle\!\langle \beta_1,\ldots,\beta_{n-1}\rangle\!\rangle\otimes [1,\alpha]\), where \([1,\alpha]\) denotes the nondegenerate binary quadratic form \(x^2+xy+\alpha y^2\). The Witt classes of \(n\)-fold bilinear Pfister forms generate \(I^nF\), the \(n\)-th power of the fundamental ideal \(IF\) of even-dimensional forms in the Witt ring \(WF\) of \(F\). In characteristic \(2\), one considers in addition the subgroups \(I^n_qF=I^{n-1}F\otimes W_qF\) of the ``quadratic'' Witt group \(W_qF\) generated by scalar multiples of quadratic \(n\)-fold Pfister forms. Let us call \(F\) to be \((n,m)\)-linked if any \(m\) bilinear \(n\)-fold Pfister forms share some common bilinear \((n-1)\)-fold Pfister form as a factor. If \(F\) is of characteristic \(2\), one considers also inseparable \((n,m)_i\)-linkage (resp. separable \((n,m)_s\)-linkage) defined by the property that any \(m\) quadratic \(n\)-fold Pfister forms share some common bilinear \((n-1)\)-fold Pfister form (resp. some common quadratic \((n-1)\)-fold Pfister form) as a factor, and \((n,m)_{si}\)-linkage means that any such \(m\) quadratic \(n\)-fold Pfister forms are either separably or inseparably linked in the above sense. Recall that the \(u\)-invariant of \(F\) is the supremum of the dimensions of anisotropic quadratic forms that are torsion in \(W_qF\) (we may identify \(W_qF\) and \(WF\) in characteristic not \(2\)), and that in characteristic not \(2\), the Hasse number \(\tilde{u}\) of \(F\) is the supremum of the dimensions of anisotropic quadratic forms that are indefinite at each ordering of \(F\) (in the absence of orderings, we have \(u=\tilde{u}\)). A famous result by \textit{R. Elman} and \textit{T. Y. Lam} [Invent. Math. 21, 125--137 (1973; Zbl 0267.10029)] states that in characteristic \(2\), if \(F\) is \((n,2)\)-linked then \(I^{n+2}F=0\), and \((2,2)\)-linkage implies that \(u=\tilde{u}\in\{ 0,1,2,4,8\}\). In characteristic \(2\), it is known that \((2,2)_i\)-linkage implies \((2,2)_s\)-linkage but not vice versa. \textit{R. Baeza} [Bol. Soc. Bras. Mat. 13, No. 1, 105--114 (1982; Zbl 0573.10014)] showed that \((2,2)_i\)-linkage is equivalent to \(u\leq 4\), and the first two authors showed in [Arch. Math. 109, No. 2, 133--142 (2017; Zbl 1427.11037)] that \((2,2)_s\)-linkage implies \(u\in\{ 0,2,4,8\}\). For nonreal \(F\), \textit{K. J. Becher} [Ann. \(K\)-Theory 3, No. 3, 369--378 (2018; Zbl 1422.11073)] proved that \((n,3)\)-linkage implies \(I^{n+1}F=0\), and he also showed that for any field \(F\) of characteristic not \(2\), \((2,3)\)-linkage implies \(\tilde{u}\leq 4\), and that the converse holds if \(F\) is nonreal. The present paper includes also this latter result with a similar proof that was found independently by the authors. They also prove that in characteristic \(2\), \((n,3)_{si}\)-linkage implies that \(I^{n+1}_qF=0\). As a corollary, \((2,3)_{si}\)-linkage implies \(u\leq 4\). They ask if conversely, \(u\leq 4\) implies \((2,3)_{si}\)-linkage.