On a question of Milnor concerning singularities of maps (Q4487706)

Milnor has in his book discussed a map \(f:\mathbb{R}^m\to \mathbb{R}^k\) satisfying the following Milnor's hypothesis: The derivative \(df_x:\mathbb{R}^m\to \mathbb{R}^k\) of \(f\) at \(x\in\mathbb{R}^m\) has rank \(k\) for every \(x\) in some punctured neighborhood of \(0\in\mathbb{R}^m\), but \(df_0\) has rank less than \(k\). Let \(\rho(m)\) denote the Radon-Hurwitz number, which can be defined as (i) \(\rho(mn)=\rho(m)\) if \(n\) is odd;(ii) \(p(m)=m\) if \(m=1,2,4,8\); and (iii) \(\rho(16m)=\rho(m)+8\) for \(m\geq 1\). Rees' main results are the following. Theorem 1.1. If \(f:\mathbb{R}^m\to\mathbb{R}^k\) is an even map satisfying Milnor's hypothesis, then \(k\leq \rho(m)\). Theorem 1.2. Let \(f_1,f_2,\dots f_k\) be polynomials of degree two over \(\mathbb{R}^m\). Assume that the map \(f=(f_1,f_2,\dots f_k): \mathbb{R}^m\to\mathbb{R}^k\) satisfies Milnor's hypothesis. Then \(k\leq\rho(m/2)+1\). (If \(m\) is odd, then \(\rho(m/2)=0.\)) Besides, the restrictions are shown to be best possible in the quadratic case.