The ranks of the homotopy groups of a space of finite LS category (Q2371315)

For \(X\) a simply connected CW complex of finite type, set \(\rho_i = \dim( \pi_i(X)\otimes{\mathbb{Q}})\), the rank of the \(i\)th homotopy group of \(X\). The celebrated dichotomy theorem of the authors establishes that, if \(X\) has finite rational category (which is the case, for instance, if \(X\) is a finite complex), then either almost all of the \(\rho_i\) are zero, or the sequence \(n_k = \sum_{i=2}^k \rho_i\) grows exponentially: there are \(\lambda > 0\) and \(C > 1\) such that \(n_k \geq \lambda C^k\) for all \(k\). In the former case, \(X\) is referred to as an elliptic space; in the latter as a hyperbolic space. The elliptic case turns out to be rather exceptional: here, a wealth of precise information is known that greatly constrains the homotopy-theoretic structure of \(X\). In the hyperbolic case, our understanding is more fragmentary -- not surprisingly, since this is the ``generic'' case. A spectacular result of Halperin establishes that, for simply connected \(X\) of finite dimension \(N\), each \(\sum_{i=k+2}^{k+N} \rho_i\) is non-zero. Beyond this, much of the work on the homotopy-theoretic structure in the hyperbolic case has focused around sharpening the statement about exponential growth of the \(\rho_i\) that features in the dichotomy theorem. The present paper does two things in this area: it provides a welcome (although brief) survey of the current state of knowledge; and it makes a further contribution to advancing the authors' program. The survey (Section \(2\)) reviews \(25\) years' worth of work, principally that of the authors with important contributions by several others. It gives a ``snapshot from the field'' of work in this area that is especially useful when placed alongside the authors' carefully composed ``studio portrait'' given in their book (see especially chapters 31, 33, 35 and 36 of [\textit{Y. Félix, S. Halperin} and \textit{J.-C. Thomas}, Rational homotopy theory, Graduate Texts in Mathematics 205, Springer-Verlag, New York (2001; Zbl 0961.55002)]). The results reviewed are grouped around three conjectures, each of which expresses the theme of exponential growth. For instance, we have {Conjecture A:} Suppose \(X\) is a simply connected CW complex of finite type with finite rational category. If \(X\) is hyperbolic, then there is some \(d\) for which the sequence \(\nu_k = \sum_{i=k+2}^{k+1+d} \rho_i\) grows exponentially. As the authors point out, it makes sense to consider the growth of the \(\rho_i\) when sampled together in a ``window'' as in this conjecture, rather than individually (e.g.~\(S^{n+1} \vee S^{n+1}\) has \(\rho_i = 0\) unless \(i \equiv 1 (n)\)). Also, in the case in which \(X\) is a hyperbolic finite complex of dimension \(N\), the mysterious \(d\) in this conjecture may be replaced by \(N-1\), as in Halperin's result mentioned above (that this may be done is a result of Lambrechts, one of several other than the authors who have made significant contributions here). The authors have approached Conjecture A quite closely in [\textit{Y. Félix, S. Halperin} and \textit{J.-C. Thomas}, Homology Homotopy Appl. 4, 219--225 (2002; Zbl 1006.55008)], where they establish Conjecture A with the clause ``grows exponentially'' replaced with ``grows faster than any polynomial.'' Conjectures B and C of the survey provide more precise measures of the conjectured exponential growth, at the cost of somewhat tighter hypotheses. Since the so-called depth theorem of [\textit{Y. Félix, S. Halperin, C. Jacobsson, C. Löfwall} and \textit{J.-C. Thomas}, Am. J. Math. 110, 301--322 (1988; Zbl 0654.55011)], cohomological invariants of Lie algebras, including global dimension and depth, have been very important in the study of the \(\rho_i\). The new contribution to this program given in this paper (Theorem 1) uses some Lie theoretic results in a crucial way. Define the \textit{logarithmic index} (of the rational homotopy Lie algebra) of \(X\) to be \[ \lambda_X = \lim\;\sup_k\;\frac{\log\;\rho_{k+1}}{k}. \] Then we have Theorem 1: Suppose \(X\) is a simply connected CW complex of finite type with finite rational category, is hyperbolic, and has finite logarithmic index, \(\lambda_X < \infty\). If the depth of the rational homotopy Lie algebra of \(X\) either equals \(1\), or is equal to the rational category of \(X\), then for some \(d\) we may write \(\nu_k = \sum_{i=k+2}^{k+1+d} \rho_i = \text{e}^{(\lambda_k + \varepsilon_k)k}\), where \(\varepsilon_k \to 0\) as \(k \to \infty\). This result implies, in particular, that for \(X\) a simply connected finite complex of dimension \(N\) whose homotopy Lie algebra satisfies either of the two conditions on depth, we have that \(\nu_k\) grows exponentially. Moreover, the authors are able to apply their theorem to several interesting examples: \(X\) is a ``non-trivial'' connected sum or wedge of spaces; \(X\) is a punctured manifold whose rational cohomology algebra is not monogenic; \(X\) has rational category at most \(2\).