On self-maps of complex flag manifolds (Q2106532)

It is well known that for any \(n\in\mathbb N\) and any continuous map \[ f:\mathbb C\mathbb P^{2n}\to\mathbb C\mathbb P^{2n}, \] there exists \(x \in \mathbb C\mathbb P^{2n}\) such that \(f(x) = x\), i.e. \(\mathbb C\mathbb P^{2n}\) has the fixed-point property (FPP). Moreover, \(\mathbb C\mathbb P^{2n+1},\) does not have the FPP. A natural generalization of complex projective spaces are (complex) Grassmann manifolds \(G_{k,n}\) (here \(G_{k,n}\) denotes the Grassmann manifold of \(k\)-dimensional spaces in \(\mathbb C^{n+k}\); one has \(G_{1,n} = \mathbb C\mathbb P^n\)). Somewhat surprisingly, classification of Grassmann manifolds that have the FPP is still open. It was conjectured that \(G_{k,n}\) has the FPP if and only if \(n\neq k\) and \(nk\) is even. This conjecture is proven only for \(k = 2\) (in [\textit{L. O'Neill}, On the fixed point property for Grassmann manifolds. Ohio State University (PhD Thesis) (1974)]), \(k = 3\) [\textit{H. Glover} and \textit{B. Homer}, Lect. Notes Math. 657, 170--193 (1978; Zbl 0385.57011)], and if \(n\geq 2k2-k-1\) [Zbl 0385.57011]. They note that there are similar conjectures and results for real and quaternionic Grassmannians in [Zbl 0385.57011]. The (complex) flag manifold \(F(n_1,\dots, n_r)\), where \(r \geq 2, \, n_1, \dots , n_r \in \mathbb N\) and \(n = n_1 + \dots + n_r\), consists of complex flags in \(\mathbb C^n\) of type \((n_1, \dots , n_r)\), that is, \(r\)-tuples \((V_1, \dots , V_r)\) of mutually orthogonal complex vector subspaces of \(\mathbb C^n\) with \(\dim_{\mathbb C} (V_i) = n_i\) for \(1 \leq i \leq r\). Note that for \(r = 2\) the flag manifold \(F(n_1, n_2)\) is actually the Grassmann manifold \(G_{n_1,n_2}\) , while the flag manifold with \(n_1 = \dots = n_r = 1\) is the complete complex flag manifold. \textit{H. H. Glover} and \textit{W. D. Homer} [Pac. J. Math. 101, 303--306 (1982; Zbl 0458.55002)] conjectured that \(F(n_1,\dots, n_r)\) has the FPP if and only if the numbers \(n_i\) are distinct and at most one of them is odd (in the same paper they prove the ``only if'' part of this conjecture). Other than for the above-mentioned cases of Grassmann manifolds, the only flag manifolds for which the ``if'' part of this conjecture is proven are \(F(1, p, q)\) for \(p \geq 2\) and \(q \geq 2p^2-1\) (see [\textit{H. H. Glover} and \textit{W. D. Homer}, Trans. Am. Math. Soc. 267, 423--434 (1981; Zbl 0479.55014)]). In this paper, the authors consider a related question. Let \(M\) be a connected topological manifold. We say that self-maps \(f\) and \(g\) of \(M\) are coincident if there exists \(x \in M\) such that \(f(x) = g(x)\). Let \(m\) be the maximum number of self-maps of \(M\) such that none of them has the FPP and that no two of them are coincident. Then the noncoincidence index of \(M\) is \(NI(M) =m+1\) if \(m\) is finite, otherwise \(\infty\). Obviously, \(M\) has the FPP if and only if \(NI(M) = 1\). In this paper, they calculate \(NI(F)\) for certain flag manifolds F. To do so, they consider endomorphisms \[ \varphi : H^*(F;\mathbb Z)\rightarrow H^*(F;\mathbb Z). \] It was conjectured in [Zbl 0479.55014] that any such endomorphism is either a grading endomorphism or a projective endomorphism. The main result of this paper as follows: For \(l \in \mathbb N\) and \(a \in \mathbb Z\) we denote \[ a^{...l} = a, a,\dots, a (l \text{ times}). \] Theorem. Let \(F := F(1^{...j}, k,m)\), where \(k,m \in \mathbb N, j \geq 0, k \geq 2\) and \(m \geq 2k^2-1\). Then every endomorphism \(\varphi: H^*(F;\mathbb Z)\rightarrow H^*(F;\mathbb Z))\) is either a grading endomorphism or a projective endomorphism.