### November 27, 2006

1991 Mathematics Subject Classification. Primary 53C23; Secondary 55M30, 57N65 . ${}^{*}$ Supported by the Israel Science Foundation (grants no. 620/00-10.0 and 84/03) . ${}^{**}$ Supported by NSF, grant 0406311 .
<ph f="cmbx">Bounding volume by systoles of 3-manifolds</ph>

### Yuli B. Rudyak * *

Department of Mathematics, Bar Ilan University, Ramat Gan 52900 Israel E-mail address : katzmik@math.biu.ac.il Department of Mathematics, University of Florida, PO Box 118105, Gainesville, FL 32611-8105 USA E-mail address : rudyak@math.ufl.edu
• Abstract. We prove a new systolic volume lower bound for non-orientable $n$  -manifolds, involving the stable 1-systole and the codimension 1 systole with coefficients in ${\mathbb{Z}}_{2}$  . As an application, we prove that Lusternik-Schnirelmann category and systolic category agree for non-orientable closed manifolds of dimension 3, extending our earlier result in the orientable case. Finally, we prove the homotopy invariance of systolic category.

Contents

1 Introduction

The systolic project in its modern form was initiated by M. Gromov [Gr83, when he proved a volume lower bound for a closed essential Riemannian manifold $M$  , which is curvature-free and depends only on the least length of a non contractible loop in $M$  , i.e. the 1-systole ${sys\pi }_{1}\left(M\right)$  : $\left({sys\pi }_{1}\left(M\right){\right)}^{n}\le {C}_{n}{vol}_{n}\left(M\right).$  See [BeCG03for a recent application. More generally, one can consider higher dimensional systoles and try to find similar volume lower bounds.
The defining text for this material is [Gr99(which is an extended English version of [Gr81), with more details in [Gr83, Gr96. Recently there has been a considerable amount of activity related to systolic inequalities. The Loewner inequality and its generalisations are studied in [Am04, IK04, KL04, BCIK05, BCIK04, KS04. Near-optimal asymptotic bounds are studied in [Bab04, Bal04, BB05, KS05.
The notion of systolic category ${cat}_{sys}$  was introduced by the authors in [KR04, cf.  2.4 below. It can be regarded as a differential-geometric analogue of the Lusternik-Schnirelmann category ${cat}_{LS}$    , cf. [LS34. We will use the modern definition of the Lusternik-Schnirelmann category, which differs by a unit from the original definition, cf. [CLOT03. Thus, ${cat}_{LS}$    of the point is equal to zero. The two invariants share a number of characteristics, including lower bound by cup-length, sensitivity to Massey products, etc., cf. [KR04,Theorem 11.1, [Ka05.
We prove that systolic category and Lusternik-Schnirelmann category agree for non-orientable closed manifolds of dimension 3, extending our earlier result in the orientable case [KR04,Corollary 6.2. The required lower bound for the systolic category of a non-orientable manifold follows from a new inequality  3.2 involving systoles of dimension and codimension one. A different but similar inequality in the orientable case was studied in [BaK03, BaK04.
The paper is organized as follows. In Section  2 , we recall the definition of the systolic invariants. In Section  3 , we recall the definition of the Bergé-Martinet constant, and present an optimal systolic inequality  3.2 combining systoles of dimension and codimension 1, valid for non-orientable manifolds. Its proof appears in Section  4 . The positivity of systoles is proved in Section  5 . In Section  6 , we prove the homotopy invariance of systolic category, which parallels that of the Lusternik-Schnirelmann category. Some open questions are posed in Section  7 .
All manifolds are assumed to be closed, connected, and smooth. All polyhedra are assumed to be compact and connected, unless explicitly mentioned otherwise. To the extent that our paper aims to address both a topological and a geometric audience, we attempt to give some indication of proof of pertinent results that may be more familiar to one audience than the other.

2 Systoles and systolic category

Let $X$  be a (finite) polyhedron equipped with a piecewise Riemannian metric $\mathcal{G}$  . We will now define the systolic invariants of $\left(X,\mathcal{G}\right)$  .
Definition 2.1. The homotopy 1-systole, denoted ${sys\pi }_{1}\left(X,\mathcal{G}\right)$  , is the least length of a non-contractible loop in $X$  . The homology 1-systole, denoted ${sysh}_{1}\left(X,\mathcal{G}\right)$  , is defined in a similar way, in terms of loops which are not zero-homologous.
Clearly, ${sys\pi }_{1}\le {sysh}_{1}$  . Now let $k\in \mathbb{N}$  . Higher homology $k$  -systoles ${sysh}_{k}$  , with coefficients over a ring $A=\mathbb{Z}$  or ${\mathbb{Z}}_{2}$  , are defined similarly to ${sysh}_{1}$  , as the infimum of $k$  -areas of $k$  -cycles, with coefficients in $A$  , which are not zero-homologous. More generally, let $B$  be the group of a regular covering space of $X$  . Then the homology groups of the covering space of $X$  can be identified with the homology groups of $X$  with coefficients in the group ring $A\left[B\right]$  . Allowing more general coefficients in such a group ring, we can define the corresponding systole
 $\begin{array}{c}{sysh}_{k}\left(X,\mathcal{G};A\left[B\right]\right).\end{array}$ (2.1)
Note that we adopt the usual convention, convenient for our purposes, that the infimum over an empty set is infinity.
More detailed definitions appear in the survey [CK03by C. Croke and the first author. We do not consider higher “homotopy” systoles.
Definition 2.2 (cf. [Fe69, BaK03). Given a class $\alpha \in {H}_{k}\left(X;\mathbb{Z}\right)$  of infinite order, we define the stable norm $\parallel {\alpha }_{\mathbb{R}}\parallel$  by setting $\parallel {\alpha }_{\mathbb{R}}\parallel ={lim}_{m\to \infty }{m}^{-1}{inf}_{\alpha \left(m\right)}{vol}_{k}\left(\alpha \left(m\right)\right),$  where ${\alpha }_{\mathbb{R}}$  denotes the image of $\alpha$  in real homology, while $\alpha \left(m\right)$  runs over all Lipschitz cycles with integral coefficients representing $m\alpha$  . The stable homology $k$  -systole, denoted ${stsys}_{k}\left(\mathcal{G}\right)$  , is defined by minimizing the stable norm $\parallel {\alpha }_{\mathbb{R}}\parallel$  over all integral $k$  -homology classes $\alpha$  of infinite order.
Recall that, in our convention, the systolic invariants are infinite when defined over an empty set of loops or cycles.
Remark 2.3. M. Berger [Be72defined invariants which eventually came to be known as the $k$  -systoles, in the case of a Riemannian manifold $X$  . All systolic notions can be defined similarly for polyhedra, cf. [Gr96and [Bab02. Note that every smooth manifold is triangulable and therefore can be viewed as a polyhedron. When $k=n$  is the dimension, ${sysh}_{n}\left(M,\mathcal{G}\right)$  is equal to the volume ${vol}_{n}\left(M,\mathcal{G}\right)$  of a compact Riemannian $n$  -manifold $\left(M,\mathcal{G}\right)$  . For an $n$  -polyhedron $X$  , however, the volume may not agree with the $n$  -systole ${sys}_{n}\left(X\right)$  , as the former is always finite, while the latter may be infinite, when $X$  does not possess a fundamental class. Moreover, it can happen that ${sys}_{n}X\ne {vol}_{n}X$  even if ${sys}_{n}X$  is finite: for example, if $X$  is a wedge of two $n$  -spheres.
Definition 2.4. Given $k\in \mathbb{N},k>1$  we set ${sys}_{k}\left(X,\mathcal{G}\right)=inf\left\{{sysh}_{k}\left(X,\mathcal{G};\mathbb{Z}\left[B\right]\right),{sysh}_{k}\left(X,\mathcal{G};{\mathbb{Z}}_{2}\left[B\right]\right),{stsys}_{k}\left(X,\mathcal{G}\right)\right\},$  where the infimum is over all groups $B$  of regular covering spaces of $X$  . Furthermore, we define ${sys}_{1}\left(X,\mathcal{G}\right)=min\left\{{sys\pi }_{1}\left(X,\mathcal{G}\right),{stsys}_{1}\left(X,\mathcal{G}\right)\right\}.$
Note that the systolic invariants thus defined are positive (or infinite), see Section  5 .
Let $X$  be an $n$  -dimensional polyhedron, and let $d\ge 2$  be an integer.
Consider a partition
 $\begin{array}{c}n={k}_{1}+\dots +{k}_{d},\end{array}$ (2.2)
where ${k}_{i}\ge 1$  for all $i=1,\dots ,d$  . We will consider scale-invariant inequalities “of length $d$  ” of the following type:
 $\begin{array}{c}{sys}_{{k}_{1}}\left(\mathcal{G}\right){sys}_{{k}_{2}}\left(\mathcal{G}\right)\dots {sys}_{{k}_{d}}\left(\mathcal{G}\right)\le C\left(X\right){vol}_{n}\left(\mathcal{G}\right),\end{array}$ (2.3)
satisfied by all metrics $\mathcal{G}$  on $X$  , where the constant $C\left(X\right)$  is expected to depend only on the topological type of $X$  , but not on the metric $\mathcal{G}$  .
Here the quantity ${sys}_{k}$  denotes the infimum of all non-vanishing systolic invariants in dimension $k$  , as defined above.
Definition 2.5. Systolic category of $X$  , denoted ${cat}_{sys}\left(X\right)$  , is the largest integer $d$  such that there exists a partition  2.2 with ${\prod }_{i=1}^{d}{sys}_{{k}_{i}}\left(X,\mathcal{G}\right)\le C\left(X\right){vol}_{n}\left(X,\mathcal{G}\right)$  for all metrics $\mathcal{G}$  on $X$  . If no such partition and inequality exist, we define systolic category to be zero.
In particular, ${cat}_{sys}X\le dimX$  .
Remark 2.6. Clearly, systolic category equals one if and only if the polyhedron possesses an $n$  -dimensional homology class, but the volume cannot be bounded from below by products of systoles of positive codimension. Systolic category vanishes if $X$  is contractible. Another example of a 2-polyhedron $X$  with ${cat}_{sys}X=0$  is a wedge of the disk and the circle, cf. Corollary  6.4 .

3 Inequality combining dimension and codimension 1

Given a maximal rank lattice $L$  in a normed space $\left({\mathbb{R}}^{b},||\cdot ||\right)$  , let ${\lambda }_{1}\left(L\right)$  denote the least length of a non-zero vector of $L$  .
Definition 3.1. The Bergé-Martinet constant [BeM89, denoted ${\gamma }_{b}^{\prime }$  , is defined as follows:
 $\begin{array}{c}{\gamma }_{b}^{\prime }=sup\left\{{\lambda }_{1}\left(L\right){\lambda }_{1}\left({L}^{*}\right)|L\subseteq {\mathbb{R}}^{b}\right\},\end{array}$ (3.1)
where the supremum is extended over all lattices $L$  in ${\mathbb{R}}^{b}$  with its Euclidean norm. A lattice attaining the supremum is called dual-critical.
Like the Hermite constant ${\gamma }_{b}$  , the Bergé-Martinet constant ${\gamma }_{b}^{\prime }$  is asymptotically linear in $b$  . Its value is known in dimensions up to 4.
Example 3.2. In dimension 3, the value of the Bergé-Martinet constant, ${\gamma }_{3}^{\prime }=\sqrt{\frac{3}{2}}=1.2247\dots$  , is slightly below the Hermite constant ${\gamma }_{3}={2}^{\frac{1}{3}}=1.2599\dots$  . It is attained by the face-centered cubic lattice, which is not isodual [MH73,p. 31, [BeM89,Proposition2.13(iii), [CoS94.
We generalize an inequality proved in the orientable case in [BaK03, BaK04.
Theorem 3.3. Let $M$  be an $n$  -dimensional manifold with first Betti number $b\ge 1$  . Then every metric $\mathcal{G}$  on $M$  satisfies the systolic inequality
 $\begin{array}{c}{stsys}_{1}\left(\mathcal{G}\right){sys}_{n-1}\left(\mathcal{G};{\mathbb{Z}}_{2}\right)\le {\gamma }_{b}^{\prime }{vol}_{n}\left(\mathcal{G}\right),\end{array}$ (3.2)
where ${\gamma }_{b}^{\prime }$  is the Bergé-Martinet constant of  3.1 . Furthermore, inequality  3.2 is optimal.
This inequality is proved in Section  4 .
Corollary 3.4. We have ${cat}_{sys}\left(M\right)\ge 2$  for all manifolds $M$  with positive first Betti number.
Corollary 3.5. If $M$  is a closed $3$  -dimensional manifold with free fundamental group, then ${cat}_{sys}M=2$  .
• Proof. In the orientable case, we proved the assertion in [KR04,Corollary7.3. Also, for $M$  non-orientable we proved there that ${cat}_{sys}M\le 2$  .
Thus, in view of Corollary  3.4 , ${cat}_{sys}M=2$  .
Corollary 3.6. Systolic category and Lusternik-Schnirelmann category coincide for all closed connected $3$  -manifolds, orientable or not.
Remark 3.7. The class of 3-dimensional Poincaré complexes is essentially larger than the class of 3-manifolds. For example, the Sphere Theorem does not hold for 3-dimensional Poincaré complexes by J. Hillman's work [Hi04. Hillman's example $Y$  is irreducible, essential, and virtually free. Thus $Y$  does not easily fit into the algebraic dichotomy in the context of 3-manifolds discussed in [KR04,Proposition 7.2, cf. the “Tits alternative” of [Hil03. It remains to be seen how the existence of such an example affects the calculation of the two categories.
Question 3.8. Does the conclusion of Corollary  3.6 hold more generally for 3-dimensional Poincaré complexes?

4 Proof of optimal inequality

With the proof of Theorem  3.3 in mind, let ${H}_{1}\left(M;\mathbb{Z}{\right)}_{\mathbb{R}}$  be the integer lattice in ${H}_{1}\left(M;\mathbb{R}\right)$  , and similarly for cohomology. Given a metric $\mathcal{G}$  , one defines the stable norm $\parallel \cdot \parallel$  in homology and the comass norm $\parallel \cdot {\parallel }^{*}$  in cohomology. The normed lattices $\left({H}_{1}\left(M;\mathbb{Z}{\right)}_{\mathbb{R}},\parallel \cdot \parallel \right)\text{and}\left({H}^{1}\left(M;\mathbb{Z}{\right)}_{\mathbb{R}},\parallel \cdot {\parallel }^{*}\right)$  are dual, whether or not $M$  is orientable [Fe74,item 5.8. Let $\parallel \cdot {\parallel }_{2}^{*}$  be the ${L}^{2}$  -norm in ${H}^{1}\left(M;\mathbb{R}\right)$  , i.e. $\parallel \omega {\parallel }_{2}^{*}={inf}_{\xi \in \omega }|\xi {|}_{2}$  where the infimum is over all closed forms $\xi \in \omega$  , and $|\cdot {|}_{2}$  is the ${L}^{2}$  -norm for forms. We have $\parallel \omega {\parallel }_{2}^{*}=|\eta {|}_{2}$  for the harmonic form $\eta \in \omega$  [LM89, and in particular the norm $\parallel \cdot {\parallel }_{2}^{*}$  is Euclidean. We will consider the invariant ${\lambda }_{1}\left(L\right){\lambda }_{1}\left({L}^{*}\right)$  for the lattice $L={H}_{1}\left(M;\mathbb{Z}{\right)}_{\mathbb{R}}\subset {H}_{1}\left(M;\mathbb{R}\right)$  .
Lemma 4.1. Let $\omega \in {H}^{1}\left(M;\mathbb{Z}{\right)}_{\mathbb{R}}$  be a cohomology class whose modulo $2$  reduction $\overline{\omega }\in {H}^{1}\left(M;{\mathbb{Z}}_{2}\right)$  is nonzero. Then ${sys}_{n-1}\left(\mathcal{G};{\mathbb{Z}}_{2}\right)\le \parallel \omega {\parallel }_{2}^{*}\left({vol}_{n}\left(\mathcal{G}\right){\right)}^{1/2}.$
• Proof. Let $\eta \in \omega$  be the harmonic 1-form for the metric $\mathcal{G}$  on $M$  .
Then $\eta$  can be represented as $df$  for some map $f:M\to {S}^{1}=\mathbb{R}/\mathbb{Z}.$  Using the Cauchy-Schwartz inequality, we obtain  $\begin{array}{c}\begin{array}{c}\parallel \omega {\parallel }_{2}^{*}\left({vol}_{n}\left(\mathcal{G}\right){\right)}^{1/2}=|\eta {|}_{2}\left({vol}_{n}\left(\mathcal{G}\right){\right)}^{1/2}\ge {\int }_{M}|df|d{vol}_{n},\end{array}\end{array}$ (4.1)
where $|\cdot |$  is the pointwise norm defined by the Riemannian metric.
We now use the coarea formula, cf. [Fe69,3.2.11, [Ch93,p. 267:  $\begin{array}{c}{\int }_{M}|df|d{vol}_{n}={\int }_{{S}^{1}}{vol}_{n-1}\left({f}^{-1}\left(t\right)\right)dt.\end{array}$ (4.2)
Note that, for every regular value $t$  of $f$  , the ${\mathbb{Z}}_{2}$  -homology class of the hypersurface ${f}^{-1}\left(t\right)\subset M$  is Poincaré ${\mathbb{Z}}_{2}$  -dual to $\overline{\omega }$  . Hence, ${vol}_{n-1}\left({f}^{-1}\left(t\right)\right)\ge {sys}_{n-1}\left(\mathcal{G};{\mathbb{Z}}_{2}\right)$  for all regular values $t$  of $f$  . By Sard's Theorem, the set of regular values of $f$  has measure 1 in ${S}^{1}$  . Thus,  $\begin{array}{c}\begin{array}{c}{\int }_{M}|df|d{vol}_{n}\ge {sys}_{n-1}\left(\mathcal{G};{\mathbb{Z}}_{2}\right).\end{array}\end{array}$ (4.3)
The lemma results by combining inequalities  4.1 and  4.3 .
• Proof of Theorem  3.3    . Let $\parallel \cdot {\parallel }_{2}$  be the norm in homology dual to the ${L}^{2}$  -norm $\parallel \cdot {\parallel }_{2}^{*}$  in cohomology. Let $\alpha \in {H}_{1}\left(M;\mathbb{Z}{\right)}_{\mathbb{R}}$  be an element of least norm, so that $\parallel \alpha {\parallel }_{2}={\lambda }_{1}\left({H}_{1}\left(M;\mathbb{Z}{\right)}_{\mathbb{R}},\parallel \cdot {\parallel }_{2}\right).$  Clearly, $\parallel \cdot {\parallel }_{2}^{*}\le \parallel \cdot {\parallel }^{*}{vol}_{n}\left(\mathcal{G}{\right)}^{1/2}$  , and so, dually, $\parallel \alpha \parallel \le \parallel \alpha {\parallel }_{2}{vol}_{n}\left(\mathcal{G}{\right)}^{1/2}$  .
Choose $\omega$  so that $\parallel \omega {\parallel }_{2}^{*}={\lambda }_{1}\left({H}^{1}\left(M;\mathbb{Z}{\right)}_{\mathbb{R}},\parallel \cdot {\parallel }_{2}^{*}\right).$  By Lemma  4.1 , we obtain  $\begin{array}{cc}{stsys}_{1}\left(\mathcal{G}\right){sys}_{n-1}\left(\mathcal{G};{\mathbb{Z}}_{2}\right)& =\parallel \alpha \parallel {sys}_{n-1}\left(\mathcal{G};{\mathbb{Z}}_{2}\right)\end{array}$
 $\begin{array}{cc}& \le \parallel \alpha \parallel \parallel \omega {\parallel }_{2}^{*}{vol}_{n}\left(\mathcal{G}{\right)}^{1/2}\end{array}$
 $\begin{array}{cc}& \le \parallel \alpha {\parallel }_{2}\parallel \omega {\parallel }_{2}^{*}{vol}_{n}\left(\mathcal{G}\right).\end{array}$
The theorem now follows from the inequality $\parallel \alpha {\parallel }_{2}\parallel \omega {\parallel }_{2}^{*}={\lambda }_{1}\left({H}_{1}\left(M;\mathbb{Z}{\right)}_{\mathbb{R}},\parallel \cdot {\parallel }_{2}\right){\lambda }_{1}\left({H}^{1}\left(M;\mathbb{Z}{\right)}_{\mathbb{R}},\parallel \cdot {\parallel }_{2}^{*}\right)\le {\gamma }_{b}^{\prime }$  by Definition  3.1 of the Bergé-Martinet constant. The optimality of the inequality results by considering a suitable product metric on a dual-critical torus ${\mathbb{T}}^{b}$  times a copy of $\mathbb{R}{P}^{2}$  .

5 Positivity of systoles

Proposition 5.1. The homotopy $1$  -systole and the stable systoles defined in Section  2    are nonzero for all polyhedra $X$  .
• Proof. We cover $X$  by a finite number of open, contractible sets ${U}_{i}$  . By the Lebesgue Lemma, there exists $\delta >0$  such that every subset of $X$  of diameter at most $\delta$  is contained in some ${U}_{i}$  . Therefore the diameter of any non-contractible loop $L$  must be more than $\delta$  , and thus the length of $L$  must exceed 2 $\delta$  , proving the positivity of the 1-systole.
For the stable $k$  -systoles, the positivity follows by a “calibration” argument. Namely, suppose classes $\alpha \in {H}_{k}\left(X;\mathbb{Z}{\right)}_{\mathbb{R}}$  and $\omega \in {H}^{k}\left(X;\mathbb{Z}{\right)}_{\mathbb{R}}$  pair non-trivially and positively. (For the theory of differential forms on polyhedra see e.g. [Bab02.) Then $1\le {\int }_{\alpha }\omega \le C\parallel \omega {\parallel }^{*}{vol}_{k}\left(\alpha \right),$  and, moreover, $1\le {\int }_{\alpha }\omega \le {m}^{-1}C\parallel \omega {\parallel }^{*}{vol}_{k}\left(\alpha \left(m\right)\right)$  for all $m\in \mathbb{N}$  . Minimizing over all singular Lipschitz cycles $\alpha \left(m\right)\in m\left[{\alpha }_{\mathbb{R}}\right]$  , we obtain the necessary bound $1\le C\parallel \omega {\parallel }^{*}\parallel {\alpha }_{\mathbb{R}}\parallel$  . Now the result follows because the abelian group ${H}^{k}\left(X;\mathbb{Z}{\right)}_{\mathbb{R}}$  is of finite rank.
For the ordinary $k$  -systoles, one cannot use differential forms, as in the proof of Proposition  5.1 , due to possible torsion classes in homology.
Nevertheless, the the positivity of ${sysh}_{k}$  holds as well. In fact, H. Federer [Fe69,item 4.2.2(1)proved that cycles with small mass are homologically trivial. However, we need a slightly stronger conclusion, to obtain a uniform lower bound for systoles of covering spaces. The following lemma is a consequence of the construction used in the proof of the deformation theorem of Federer and Fleming [FF60, Fe69, cf. [Wh99. The proof was summarized in [Gr83,Prop.3.1.A.
Lemma 5.2. Let $V$  be a $k$  -dimensional polyhedron in ${\mathbb{R}}^{N}$  . Then there exists a continuous map $f$  of $V$  into a $\left(k-1\right)$  -dimensional polyhedron ${K}^{k-1}$  in ${\mathbb{R}}^{n}$  such that $dist\left(v,f\left(v\right)\right)\le {C}_{N}\left(volV{\right)}^{1/k}$  for all $v\in V$  and for some constant ${C}_{N}$  depending only on the ambient dimension, where $dist$  denotes the Euclidean distance in ${\mathbb{R}}^{N}$  .
Proposition 5.3. The $k$  -systoles ${sysh}_{k}\left(X,\mathcal{G};A\right)$  defined in Section  2    are nonzero for all polyhedra $X$  . In fact, a uniform lower bound for ${sysh}_{k}$  is valid for all covering spaces of $X$  .
• Proof. We may view $X$  as a polyhedron in ${\mathbb{R}}^{N}$  , since the metric $\mathcal{G}$  is bilipschitz equivalent to the restricted metric. Note that $X$  has a regular neighborhood $U\subset {\mathbb{R}}^{N}$  , i.e. $X$  is a deformation retract of $U$  .
Since $X$  is compact, there exists $\varepsilon >0$  such that $dist\left(X,{\mathbb{R}}^{N}\U\right)>\varepsilon$  .
Let $j:X\to U$  denote the inclusion.
Note that we can regard any singular chain in $X$  , with coefficients in $A$  with $A=\mathbb{Z}$  or ${\mathbb{Z}}_{p}$  , as a singular, not necessarily connected polyhedron. Consider a such a polyhedron $\phi :{F}^{k}\to X$  representing a non-zero $k$  -dimensional homology with coefficients in $A$  . We now view $\phi$  as a singular polyhedron in ${\mathbb{R}}^{N}$  .
Choose $\sigma >0$  such that ${C}_{N}{\sigma }^{1/k}<\varepsilon$  where ${C}_{N}$  is the constant from Lemma  5.2 . Assume that $vol\left(\phi \left(F\right)\right)<\sigma$  . Then by Lemma  5.2 , there is a map $f:\phi \left(F\right)\to {K}^{k-1}\subset {\mathbb{R}}^{N}$  such that $dist\left(a,f\left(a\right)\right)<\varepsilon$  for all $a\in \phi \left(F\right)$  . Thus $f\left(\phi \left(F\right)\right)\subset U$  . Moreover, for all $a\in \phi \left(F\right)$  , the segment joining $a$  and $f\left(a\right)$  is contained in $U$  . Therefore the maps $F---\to \phi X---\to jU\text{and}F---\to \phi \phi \left(F\right)---\to fU$  are homotopic. But $f\left(\phi \left(F\right)\right)\subset {K}^{k-1}$  , and so $f\circ \phi :F\to U$  represents the zero element in ${H}_{k}\left(U;A\right)$  . Hence $j\circ \phi$  represents the zero element in ${H}_{k}\left(U;A\right)$  . But then $\phi$  represents the zero element in ${H}_{k}\left(X;A\right)$  . This is a contradiction.
The deformation of $F$  into a polyhedron of positive codimension is contained in $U$  . Since $X$  is a retract of $U$  , and because of the Cellular Approximation Theorem, we can map the deformation into $X$  in a way that the last-moment map sends $F$  into the $\left(k-1\right)$  -skeleton ${X}^{\left(k-1\right)}$  of $X$  .
It follows that the same $\delta$  works for the systoles of arbitrary covering spaces of $X$  , by the covering homotopy property, cf. [BCIK04,Section 2.

6 Homotopy invariance of systolic category

The homotopy invariance of systolic category follows from the techniques developed by I. Babenko in [Bab93, cf. the compression theorem of [KR04,Theorem 5.1. More precisely, we have the following theorem, proved in [Bab93for the 1-systole by essentially the same argument.
Theorem 6.1. The optimal systolic ratio associated with a partition of $n=dim\left(M\right)$  is a homotopy invariant of a closed manifold $M$  .
• Proof. Let $f:{M}^{n}\to {N}^{n}$  be a homotopy equivalence of closed PL manifolds. By A. Wright [Wr74,Theorem7.3, $f$  is homotopic to a PL monotone map. Recall that a continuous map is called monotone if the inverse image of every point is connected and compact. Thus, we can assume the every $n$  -dimensional simplex of $N$  has exactly one inverse image simplex. Then we can pull back systolic inequalities from one manifold to the other, in the following sense. The pullback metric has the same volume as the target metric. Meanwhile, the projection map is distance decreasing, and therefore the target manifold has smaller systoles than the source manifold.
Now suppose that $M$  satisfies a systolic inequality  $\begin{array}{c}{\prod }_{i}{sys}_{{k}_{i}}\left(X\right)\le C{vol}_{n}\left(X\right),\end{array}$ (6.1)
relative to a suitable partition $n={k}_{1}+\dots +{k}_{d}$  , for all metrics, with a constant $C$  . In particular, it satisfies it for all pullback metrics. Then the target manifold will satisfy the same inequality (with the same partition of $n$  ) with exactly the same constant $C$  . Thus the associated optimal systolic ratio, which is the least such constant $C$  , is a homotopy invariant.
Since Wright's result is not available for polyhedra, we have to make do with a fixed simplicial map. The disadvantage here is that the pullback metric may have greater volume than the target, but anyway it is controlled by the number of simplices in the inverse image of a top dimensional simplex. Thus the optimal systolic ratio is no longer a homotopy invariant, but nevertheless the homotopy invariance of systolic category persists, in the following sense.
Theorem 6.2. Given two $n$  -dimensional polyhedra $X$  and $Y$  , assume that there exists a simplicial map $f:X\to Y$  that induces an isomorphism in ${\pi }_{1}$  and a monomorphism in homology with coefficients in $\mathbb{Z}\left[B\right]$  and ${\mathbb{Z}}_{2}\left[B\right]$  where $B$  runs over all groups of regular covering maps of $X$  $\left(\text{and therefore}Y\right)$  .
Then ${cat}_{sys}X\le {cat}_{sys}Y$  .
• Proof. The case ${cat}_{sys}X=0$  is trivial, so assume that ${cat}_{sys}X>0$  .
Consider a partition $n={k}_{1}+\dots +{k}_{d}$  . Suppose that $X$  satisfies a systolic inequality ${\prod }_{i}{sys}_{{k}_{i}}\left(X,\mathcal{G}\right)\le C\left(X\right){vol}_{n}\left(X,\mathcal{G}\right)$  for all piecewise Riemannian metrics. Choose a piecewise Riemannian metric on $Y$  and consider the pull back (degenerate) metric $\mathcal{ℋ}$  induced by $f$  on $X$  .
By the monomorphism hypothesis, we have ${sys}_{k}\left(X,\mathcal{ℋ}\right)\ge {sys}_{k}\left(Y\right)$  .
Meanwhile, ${vol}_{n}\left(X,\mathcal{ℋ}\right)\le k{vol}_{n}\left(Y\right)$  where $k$  is number of $n$  -simplices in $X$  . A small perturbation of $\mathcal{ℋ}$  will yield a non-degenerate metric $\mathcal{G}$  , satisfying the inequality ${vol}_{n}\left(X,\mathcal{G}\right)\le 2k{vol}_{n}\left(Y\right)$  . Moreover, we can also assume that ${sys}_{k}\left(X,\mathcal{G}\right)\ge {sys}_{k}\left(Y\right)-\delta$  where $\delta \le {sys}_{k}\left(X\right)$  for all $k$  .
Hence, $\prod {sys}_{{k}_{i}}\left(Y\right)\le \prod \left({sys}_{{k}_{i}}\left(X,\mathcal{G}\right)+\delta \right)\le 2C{vol}_{n}\left(X,\mathcal{G}\right)\le 4kC{vol}_{n}\left(Y\right),$  and thus ${cat}_{sys}X\le {cat}_{sys}Y$  .
Corollary 6.3. The systolic category of $n$  -dimensional polyhedra is a homotopy invariant.
Corollary 6.4. Let $X$  be an $n$  -dimensional polyhedron which is homotopy equivalent to a polyhedron $Y$  of dimension at most $n-1$  . Then ${cat}_{sys}X=0$  .
• Proof. Consider the wedge $Y\vee {\Delta }^{n}$  of $Y$  with the $n$  -simplex ${\Delta }^{n}$  . Then, given a metric on $Y$  , we can extend it to a metric on $Y\vee {\Delta }^{n}$  of arbitrary small volume. Hence, ${cat}_{sys}\left(Y\vee {\Delta }^{n}\right)=0$  , and thus ${cat}_{sys}X=0$  .

7 New directions

Question 7.1. In the context of higher Massey products, is there a generalization of the result [KR04,Theorem 11.1, to a case where triple products vanish but there is a higher nontrivial product?
Question 7.2. Considering finite covers which are intermediate between the free abelian cover and the manifold. Is it true that if the fiber class is non-zero in the free abelian cover, then it is already nontrivial in a finite cover? In such case we would immediately get a lower bound of ${b}_{1}\left(X\right)+1$  for ${cat}_{LS}$    , which parallels the bound for systolic category resulting from [IK04.
Question 7.3. Consider an (absolute) degree 1 map $f:X\to {T}^{2}$  from a non-orientable surface to the torus. Consider also the product map $f:X×{S}^{2}\to {T}^{2}×{S}^{2}.$  The range has systolic category 3, by real cup length argument.
However, the domain has real cup length only 2. Therefore we do not have an immediate lower bound of 3 for systolic category, unlike LS category. Is there such a bound? What is the systolic category of $\mathbb{R}{\mathbb{P}}^{2}×{S}^{2}$  ?
Question 7.4. Given a function $f$  on a manifold, how does ${cat}_{sys}\left({f}^{c}\right)$  of the sublevel set ${f}^{c}$  change as a function of $c$  ?

8 Acknowledgment

We are grateful to Brian White for a discussion of the deformation theorem, exploited in Section  5 .
References

1. Ammann, B.: Dirac eigenvalue estimates on two-tori. J. Geom. Phys. 51 (2004), no. 3, 372–386.
2. Babenko, I.: Asymptotic invariants of smooth manifolds. Russian Acad. Sci. Izv. Math. 41 (1993), 1–38.
3. Babenko, I.: Forte souplesse intersystolique de variétés fermées et de polyèdres. Annales de l'Institut Fourier 52 (2002), no. 4, 1259-1284.
4. Babenko, I.: Géométrie systolique des variétés de groupe fondamental ${\mathbb{Z}}_{2}$  , Sémin. Théor. Spectr. Géom. Grenoble, 22 (2004), 25-52.
5. Babenko, I.; Balacheff, F.: Géométrie systolique des sommes connexes et des revêtements cycliques, Mathematische Annalen (to appear).
6. Balacheff, F.: Sur des problèmes de la géométrie systolique. Sémin. Théor. Spectr. Géom. Grenoble 22 (2004), 71–82.
7. Bangert, V; Croke, C.; Ivanov, S.; Katz, M.: Boundary case of equality in optimal Loewner-type inequalities, Trans. Amer. Math. Soc., to appear. See arXiv:math.DG/0406008
8. Bangert, V; Croke, C.; Ivanov, S.; Katz, M.: Filling area conjecture and ovalless real hyperelliptic surfaces, Geometric and Functional Analysis (GAFA), to appear. See arXiv:math.DG/0405583
9. Bangert, V.; Katz, M.: Stable systolic inequalities and cohomology products, Comm. Pure Appl. Math. See arXiv:math.DG/0204181
10. Bangert, V; Katz, M.: An optimal Loewner-type systolic inequality and harmonic one-forms of constant norm. Comm. Anal. Geom. 12 (2004), number 3, 701–730. See arXiv:math.DG/0304494
11. Bergé, A.-M.; Martinet, J.: Sur un problème de dualité lié aux sphères en géométrie des nombres. J. Number Theory 32 (1989), 14–42.
12. Berger, M.: Du côté de chez Pu. Ann. Sci. École Norm. Sup. (1972), 4, 1–44; A l'ombre de Loewner, ibid, 241–265.
13. Besson, G.; Courtois, G.;, Gallot, S.: Hyperbolic manifolds, amalgamated products and critical exponents. C. R. Math. Acad. Sci. Paris 336 (2003), no. 3, 257–261.
14. Chavel, I.: Riemannian geometry – a modern introduction. Cambridge Tracts in Mathematics, 108. Cambridge University Press, Cambridge, 1993.
15. Conway, J. H.; Sloane, N. J. A.: On lattices equivalent to their duals. J. Number Theory 48 (1994), no. 3, 373–382.
16. Cornea, O.; Lupton, G.; Oprea, J.; Tanré, D.: Lusternik-Schnirelmann category. Mathematical Surveys and Monographs, 103. American Mathematical Society, Providence, RI, 2003.
17. Croke, C.; Katz, M.: Universal volume bounds in Riemannian manifolds, Surveys in Differential Geometry VIII (2003), 109–137. Available at arXiv:math.DG/0302248
18. Federer, H.: Geometric measure theory. Grundlehren der mathematischen Wissenschaften, 153. Springer–Verlag, Berlin, 1969.
19. Federer, H.: Real flat chains, cochains, and variational problems. Indiana Univ. Math. J. 24 (1974), 351–407.
20. Federer, H.; Fleming, W. Normal and integral currents. Ann. of Math. (2) 72 (1960) 458–520.
21. Gómez-Larran͂aga, J.; González-Acun͂a, F.: Lusternik-Schnirel'mann category of $3$  -manifolds. Topology 31 (1992), no. 4, 791–800.
22. Gromov, M.: Structures métriques pour les variétés riemanniennes. Edited by J. Lafontaine and P. Pansu. Textes Mathématiques, 1. CEDIC, Paris, 1981.
23. Gromov, M.: Filling Riemannian manifolds, J. Diff. Geom. 18 (1983), 1–147.
24. Gromov, M.: Systoles and intersystolic inequalities, Actes de la Table Ronde de Géométrie Différentielle (Luminy, 1992), 291–362, Sémin. Congr., 1, Soc. Math. France, Paris, 1996.
25. Gromov, M.: Metric structures for Riemannian and non-Riemannian spaces, Progr. in Mathematics, 152, Birkhäuser, Boston, 1999.
26. Gromov, M.; Lawson, H. B., Jr.: Positive scalar curvature and the Dirac operator on complete Riemannian manifolds. Inst. Hautes Etudes Sci. Publ. Math., 58 (1983), 83–196 (1984).
27. Hillman, J. A.: Tits alternatives and low dimensional topology. J. Math. Soc. Japan 55 (2003), no. 2, 365–383.
28. Hillman, J. A.: An indecomposable PD ${}_{3}$  -complex: II, Algebraic and Geometric Topology 4 (2004), 1103-1109.
29. Ivanov, S.; Katz, M.: Generalized degree and optimal Loewner-type inequalities, Israel J. Math. 141 (2004), 221–233. Available at the site arXiv:math.DG/0405019
30. Katz, M.: The filling radius of two-point homogeneous spaces, J. Diff. Geom. 18 (1983), 505–511.
31. Katz, M.: The first diameter of 3-manifolds of positive scalar curvature. Proc. Amer. Math. Soc. 104 (1988), no. 2, 591–595.
32. Katz, M.: Systolic inequalities and Massey products in simply-connected manifolds, Geometriae Dedicata, to appear.
33. Katz, M.; Lescop, C.: Filling area conjecture, optimal systolic inequalities, and the fiber class in abelian covers. Proceedings of conference and workshop in memory of R. Brooks, held at the Technion, Israel Mathematical Conference Proceedings (IMCP), Contemporary Math., Amer. Math. Soc., Providence, R.I. (to appear). See arXiv:math.DG/0412011
34. Katz, M.; Rudyak, Y.: Lusternik-Schnirelmann category and systolic category of low dimensional manifolds. Communications on Pure and Applied Mathematics, to appear. See arXiv:math.DG/0410456
35. Katz, M.; Sabourau, S.: Hyperelliptic surfaces are Loewner, Proc. Amer. Math. Soc., to appear. See arXiv:math.DG/0407009
36. Katz, M.; Sabourau, S.: Entropy of systolically extremal surfaces and asymptotic bounds, Ergodic Theory and Dynamical Systems, to appear. See arXiv:math.DG/0410312
37. Katz, M.; Sabourau, S.: An optimal systolic inequality for CAT(0) metrics in genus two, preprint. See arXiv:math.DG/0501017
38. Lawson, H. B.; Michelsohn, M.-L.: Spin geometry. Princeton Mathematical Series 38. Princeton University Press, Princeton, NJ, 1989.
39. Lusternik, L. A., Schnirelmann, L. G.: Methodes topologiques dans le problèmes variationels. Hermann, Paris 1934.
40. Milnor, J.; Husemoller, D.: Symmetric bilinear forms. Springer, 1973.
41. White, B.: The deformation theorem for flat chains. Acta Math. 183 (1999), no. 2, 255–271.
42. Wright, A. H.: Monotone mappings and degree one mappings between $PL$  manifolds. Geometric topology (Proc. Conf., Park City, Utah, 1974), pp. 441–459. Lecture Notes in Math. 438, Springer, Berlin, 1975.

Department of Mathematics, Bar Ilan University, Ramat Gan 52900 Israel E-mail address : katzmik@math.biu.ac.il Department of Mathematics, University of Florida, PO Box 118105, Gainesville, FL 32611-8105 USA E-mail address : rudyak@math.ufl.edu