On the mod \(p\) Lannes-Zarati homomorphism (Q2324647)

Suppose \(X\) is a pointed \(CW\)-complex of finite type. One of the interesting and challenging problems in stable homotopy theory/homology of infinite loop spaces, is the computation of the image of the unstable mod \(p\) Hurewicz homomorphism \[ {_p\pi_*^s}X\simeq{_p\pi_*}QX\longrightarrow H_*(QX;\mathbb{Z}/p) \] where \({_p\pi_*^s}X\) denotes the \(p\)-component of \({\pi_*^s}X\), the stable homotopy of \(X\), and \(Q\) is the destablization defined by \(QX=\operatorname{colim} \Omega^i\Sigma^iX\). \textbf{Curtis conjecture.} In the case of \(p=2\) and \(X=S^0\), this conjecture due to \textit{E. B. Curtis} [Ill. J. Math. 19, 231--246 (1975; Zbl 0311.55007), Theorem 7.1], predicts that the image of \[ {_2\pi_*^s}\simeq{_2\pi_*}QS^0\longrightarrow H_*(QS^0;\mathbb{Z}/2) \] is generated by \(h(\eta)\), \(h(\nu)\), \(h(\sigma)\), and \(h(\theta_i)\) where \(\eta\), \(\nu\) and \(\sigma\) are Hopf invariant one elements and \(\theta_i\in{_2\pi_{2^{i+1}-2}^s}\) is a Kervaire invariant one element; indeed after the work of Hill, Hopkins and Ravenel [\textit{M. A. Hill} et al., Ann. Math. (2) 184, No. 1, 1--262 (2016; Zbl 1366.55007)] we know that \(\theta_i\) elements do not exist for \(i>6\) which implies that the image of \(h\) is finite. In general, for a given \(X\) there are two conjectures which describe the image of \(h\) which we state below. Suppose \(X\) is a pointed space, and consider the unstable Hurewicz homomorphism \[ {_2\pi_*^s}X\simeq{_2\pi_*}QX\longrightarrow H_*(QX;\mathbb{Z}/2). \] \textbf{Eccles conjecture.} If \(X\) is path connected, \(n>0\), and \(f\in{_2\pi_n}QX\) maps nontrivially under \(h\) then the stable adjoint of \(f\), say \(\widetilde{f}:S^n\to X\) is either detected by homology or by a primary operation in its mapping cone [\textit{H. Zare}, ``Filtered finiteness of the image of the unstable Hurewicz homomorphism with applications to bordism of immersions'', Q. J. Math. 70, No. 3, 859--878 (2019), Conjecture 1.2]. \textbf{Hung conjecture.} For \(n>0\) if \(h(f)\neq 0\) then \(f\) is of Adams filtration at most \(2\) [\textit{N. H. V. Hung} and \textit{G. Powell}, J. Algebra 517, 186--206 (2019; Zbl 1405.55018), Conjecture 1]. The existing literature on the \(2\)-primary case is very rich. However, as far as the reviewer is aware, we even do not have any conjecture which predicts the image at an odd prime. Perhaps, one of the first (maybe the first one) attempts to look at the case of odd primes is due to \textit{R. J. Wellington} [The unstable Adams spectral sequence for free iterated loop spaces. Providence, RI: American Mathematical Society (AMS) (1982; Zbl 0488.55006)] who tried to generalise Curtis's work to odd primes. In the case of \(p>2\), the general attitude, at least to the reviewer, has been to show that at odd primes the Hurewicz homomorphism enjoys properties analogous to its \(2\)-primary counterpart. For instance, as noted above, at the prime \(2\) and \(X=S^0\), the Curtis conjecture implies that the image of \(h\) is finite. So, one could expect that at an odd prime at least the image should be finite. A very recent work of \textit{N. J. Kuhn} [Invent. Math. 214, No. 2, 957--998 (2018; Zbl 1403.55007)] verifies this latter conjecture for a range of spectra [loc. cit., Theorem 1.7 (Finiteness Theorem)]. The paper under review adds to our knowledge on the case of \(p>2\) and identifies some differences that we might expect, at least from an algebraic point of view. We note that one of the main tools to study the unstable Hurewicz homomorphism \[ {_p\pi_*^s}X\simeq{_p\pi_*}QX\longrightarrow H_*(QX;\mathbb{Z}/p) \] is to consider the Lannes-Zarati-homomorphism \[ \varphi^M_s:\mathrm{Ext}^{s,s+t}_A(M,\mathbb{F}_p)\to\mathrm{Ann}(\mathcal{R}_sM)^{\#})_t \] where \(M\) is any unstable module over the mod \(p\) Steenrod algebra, \(N^{\#}\) denotes the linear dual of \(N\), \(\mathrm{Ann}(N^{\#})\) is the submodule of \(N^{\#}\) annihilated by \(Sq^i\) with \(i>0\), and \(\mathcal{R}_s\) is Singer's functor (see Section 2.3 of the paper for the exact definition of this functor). For \(M=\widetilde{H}^*X\), \(\mathrm{Ext}^{s,s+t}_A(M,\mathbb{F}_p)\) is the \(E_2\)-term of the Adams spectral sequence converging to \({_p\pi_*^s}X\). In this case \(\varphi^M_s\) is considered as the graded associated homomorphism to \(h\) and is used to study the behaviour of \(h\). However, it is possible to have \(h(f)\neq 0\) where as \(\varphi^M_s(c)=0\) on a permanent cycle mapping to \(f\). A weak conjecture on spherical classes is that for \(M=\widetilde{H}^*S^0\), \(\varphi^M_s=0\) for \(s>0\). The authors' strategy to study \(\varphi^M_s\) is to work with a linear dual of a chain level representation of \(\varphi^M_s\), denoted by \[ ({\widetilde{\varphi}}^M_s)^{\#}:\mathcal{R}_sM\longrightarrow(\Gamma^+M)_s \] where \(\Gamma^+M\{(\Gamma^+M)_s\}\) is a complex which computes the homology of the mod \(p\) Steenrod algebra. The authors take first steps in this direction and study \(\varphi^{\widetilde{H}^*S^0}_s\) for \(s=1,2\). Their main results are recorded as Theorem 4.1 and Theorem 4.2. We note that the case of \(X=S^0\) is one which does not live under the umbrella of Kuhn's finiteness theorem. It is recorded that \(\varphi_s^{\widetilde{H}^*S^0}\) is an isomorphism for \(s=1\) (Theorem 4.1) and in the case of \(s=2\) the image is nontrivial if \(t=0\) or \(t=2(p-1)p^{i+1}-2\) with \(i\geqslant 0\) (Theorem 4.2). This observation does support an analogue of Curtis's conjecture as the odd primary analogues of Kervaire invariant one elements live in dimensions \(2(p-1)p^{i+1}-2\). However, the authors note that this identifies a difference of the behaviour of the Lannes-Zarati homomorphism as in the case of \(s=2\) and \(p>2\) this homomorphism is not an epimorphism unlike its \(2\)-primary counterpart. The paper is written in very algebraic terms as many other papers on the Lannes-Zarati homomorphism.