Linear and bilinear operators and their zero-sets (Q2228339)

The authors generalise certain results on the kernels of linear functionals to zero sets of linear operators and bilinear mappings. Given subspaces \(M\) and \(N\) of a vector space \(X\) and \(\varepsilon>0\), they say that \(d(M,N)\le \varepsilon\) if \( B_X\cap M\subset N+\varepsilon B_X\) and \( B_X\cap N\subset M+\varepsilon B_X\) and show that if \(f\) and \(g\) are surjective norm-one linear operators from \(X\) onto \(\mathbb{R}^n\) with \(d(\ker f,\ker g)\le\varepsilon\), then there is an \(n\times n\) matrix \(W\) such that \(\|f-Wg\|\le\sqrt{n}\varepsilon\). Setting \(M(g)=\max_{i\le n}d(0,g^{-1}(e_i))\), they show that if \(f\) and \(g\) are surjective norm-one linear operators from \(X\) onto \(\mathbb{R}^n\) with \(d(\ker f,\ker g)\le\frac{\varepsilon}{\sqrt{n}M(g)}\), then there is an invertible linear operator \(T\colon X\to X\) such that \(\|T-I\|\le\varepsilon\) and \(T(\ker f)=\ker g\). Given a bilinear mapping \(A\colon X\times X\to \mathbb{R}^n\), the set \(Z(A)\) is defined as \(Z(A)=\{(x,y):A(x,y)=0\}\) while the set \(\Omega\) is defined as \(\Omega=\{x\in X:A_x$ is surjective$\}\), where \(A_x(y)=A(x,y)\). The authors prove that if \(A\) and \(B\) are bilinear operators from \(X\times X\) into \(\mathbb{R}^n\) with \(Z(A)=Z(B)\) such that \(\Omega =X-\{0\}\) and \(\dim X>n^2\), then there is an invertible \(n\times n\) matrix \(W\) such that \(B=WA\). An example shows that the result need not hold when \(\dim X<n^2\). Finally, the authors show that if \(A\) and \(B\) are norm-one bilinear operators from \(\mathbb{R}^k\times \mathbb{R}^k\) to \(\mathbb{R}^n\) such that (i) \(k>n^2\), (ii) \(\Omega=\mathbb{R}^k-\{0\}\), (iii) \(d(\ker A_x,\ker B_x)\le\varepsilon\) for all \(x\) in the unit sphere of \(\mathbb{R}^k\), then there are a constant \(C\), depending on \(n\) and \(A\), and an \(n\times n\) matrix \(W\) such that \(\|B-WA\|\le C\varepsilon\).