This paper was converted on www.awesomepapers.org from LaTeX by an anonymous user.
Want to know more? Visit the Converter page.

77-location, weak systolicity and isoperimetry

Nima Hoda Department of Mathematics, Cornell University
Ithaca, NY 14853, USA
[email protected]
 and  Ioana-Claudia Lazăr Dept. of Mathematics
Politehnica University of Timişoara
Victoriei Square 22, 300006300006-Timişoara, Romania
[email protected]
Abstract.

mm-location is a local combinatorial condition for flag simplicial complexes introduced by Osajda. Osajda showed that simply connected 88-located locally 55-large complexes are hyperbolic. We treat the nonpositive curvature case of 77-located locally 55-large complexes.

We show that any minimal area disc diagram in a 77-located locally 55-large complex is itself 77-located and locally 55-large. We define a natural CAT(0)\operatorname{CAT}(0) metric for 77-located disc diagrams and use this to prove that simply connected 77-located locally 55-large complexes have quadratic isoperimetric function. Along the way, we prove that locally weakly systolic complexes are 77-located locally 55-large.

Key words and phrases:
7-located complex, combinatorial nonpositive curvature, weakly systolic complex
2010 Mathematics Subject Classification:
20F65, 20F67

1. Introduction

Metric and combinatorial notions of negative and nonpositive curvature have seen significant study and application in recent decades, especially in the areas of geometric group theory and metric graph theory [9, 7, 10, 12, 2, 5, 1, 11]. These notions are often given as local conditions that have global implications on spaces. Of particular note on the combinatorial side are the developments of CAT(0)\operatorname{CAT}(0)-cubical complexes (also known as median graphs) [9, 7], systolic complexes (also known as bridged graphs) [7, 10, 12] and Helly graphs [5].

The combinatorial nonpositive curvature condition of mm-location for simplicial complexes was introduced by Osajda in [16] as a method for proving hyperbolicity for 33-manifolds. While local systolicity of a complex depends only on the combinatorics around a single vertex, mm-location depends on the combinatorics around pairs of adjacent vertices, allowing more flexibility. Osajda proved that simply connected 88-located locally 55-large complexes are Gromov hyperbolic and used this to obtain a new solution to a problem of Thurston. The second named author studied a version of 88-location, suggested by Osajda [16, Subsection 5.1] and showed that simply connected, 88-located simplicial complexes are Gromov hyperbolic [13]. In later work. the second named author introduced another combinatorial curvature condition, called the 5/95/9-condition, and showed that the complexes which satisfy it, are 88-located and therefore Gromov hyperbolic as well [14].

The current paper continues the study of mm-located complexes. As noted by Osajda [16], systolic complexes are 77-located. We strengthen this statement by showing that weakly systolic complexes are 77-located. This provides many more examples of 77-located locally 55-large complexes, including thickenings of CAT(0)\operatorname{CAT}(0) cube complexes whose links do not contain induced 44-cycles [15].

Theorem A (Theorem 3.1).

Locally weakly systolic complexes are 77-located locally 55-large.

We also prove the following two theorems which show that 77-location and local 55-largeness have significant global consequences: disc diagrams with particularly nice combinatorial and geometric properties.

Theorem B (Theorem 5.9).

Minimal area disc diagrams in 77-located locally 55-large complexes are 77-located locally 55-large.

Let PnP_{n} be the regular Euclidean nn-gon subdivided into nn congruent triangles meeting at the barycenter of PnP_{n}. Let TnT_{n} be a triangle of PnP_{n}.

Theorem C (Theorem 4.6).

Let DD be a 77-located simplicial disc. There is a natural way to metrize the triangles of DD using only isometry types of T4T_{4}, T5T_{5} and T6T_{6} such that the resulting piecewise Euclidean triangle complex is CAT(0)\operatorname{CAT}(0). The choice of metrization for each triangle depends only on the local combinatorics of DD around the triangle.

From Theorem B and Theorem C we obtain the following global coarse geometric and algebraic consequences.

Corollary D (Corollary 4.7).

A simply connected 77-located locally 55-large complex has quadratic isoperimetric function. Consequently, the fundamental group of a 77-located locally 55-large complex has quadratic isoperimetric function

1.1. Structure of the paper

In Section 2 we present basic definitions and notation, in Section 3 we prove Theorem A, in Section 4 we prove Theorem C and in Section 5 we prove Theorem B.

2. Preliminaries

Let XX be a simplicial complex. If v1,v2,,vkv_{1},v_{2},\ldots,v_{k} are vertices of XX then we use the notation v1,v2,,vk\langle v_{1},v_{2},\ldots,v_{k}\rangle to denote the simplex spanned on the vertices v1,v2,,vkv_{1},v_{2},\ldots,v_{k}, should it exist. A subcomplex LL of XX is full if any simplex of XX whose vertices are contained in LL is contained in LL. For distinct vertices vv and vv^{\prime} of XX we write vvv\sim v^{\prime} if vv and vv^{\prime} are adjacent in the 11-skeleton of XX and otherwise we write vvv\nsim v^{\prime}. The simplicial complex XX is flag if any finite set of pairwise adjacent vertices of XX spans a simplex of XX. Going forward we will always assume that XX is a flag complex.

Unless otherwise stated, when referring to distances between vertices of a simplicial complex we will mean the graph metric, i.e., the number of edges in the shortest 11-skeleton path joining the vertices.

A cycle or loop γ\gamma in XX is a subcomplex of XX isomorphic to a subdivision of S1S^{1}. We may denote a cycle γ\gamma by a sequence (v1,v2,,vk)(v_{1},v_{2},\ldots,v_{k}) where {v1,v2,,vk}\{v_{1},v_{2},\ldots,v_{k}\} is the set of vertices of γ\gamma and where vi,vi+1\langle v_{i},v_{i+1}\rangle is an edge of γ\gamma for each ii, indices taken modulo kk. The length of γ\gamma, denoted by |γ||\gamma|, is the number of edges of γ\gamma. A kk-wheel in XX is a subcomplex W=(v0;v1,v2,,vk)W=(v_{0};v_{1},v_{2},\ldots,v_{k}) of XX where (v1,,vk)(v_{1},\ldots,v_{k}) is a cycle and v0,vi,vi+1\langle v_{0},v_{i},v_{i+1}\rangle is a triangle of WW for each ii, indices taken modulo kk. That is, the kk-wheel WW is a simplicial cone on its boundary cycle W=(v1,v2,,vk)\partial W=(v_{1},v_{2},\ldots,v_{k}).

Let γ\gamma be a cycle in XX. A disc diagram (D,f)(D,f) for γ\gamma is a nondegenerate simplicial map f:DXf:D\rightarrow X where DD is a triangulated disc, and f|Df|_{\partial D} maps D\partial D isomorphically onto γ\gamma. The disc diagram (D,f)(D,f) has minimal area if DD has the least possible number of triangles among all disc diagrams for γ\gamma.

By van Kampen’s Lemma, every nullhomotopic cycle in XX has a disc diagram. It is not difficult to see that when XX is a flag simplicial complex, any minimal area disc diagram in XX is also a flag simplicial complex.

Refer to caption
Figure 1. A planar (6,5)(6,5)-dwheel and a nonplanar (6,5)(6,5)-dwheel.

A (k,l)(k,l)-dwheel W=W1W2W=W_{1}\cup W_{2} in XX is the union of two wheels W1=(wl;v1,v2,v3=wl1,,W_{1}=(w_{l};v_{1},v_{2},v_{3}=w_{l-1},\ldots, vk)v_{k}) and W2=(v2;w1,w2,,wl)W_{2}=(v_{2};w_{1},w_{2},\ldots,w_{l}) with v1=w1v_{1}=w_{1} or v1w1v_{1}\sim w_{1}. See Figure 1. If v1=w1v_{1}=w_{1}, we call WW a planar dwheel. If v1w1v_{1}\sim w_{1}, we call WW a nonplanar dwheel. The boundary W\partial W of WW is the cycle (w1,w2,,wl1=v3,v4,,vk)(w_{1},w_{2},\ldots,w_{l-1}=v_{3},v_{4},\ldots,v_{k}). The boundary W\partial W is the subcomplex of WW induced on its vertices excluding the wheel centers wlw_{l} and v2v_{2}. Note that W\partial W has length k+l4k+l-4 if WW is planar and k+l3k+l-3 if it is nonplanar.

The link of XX at a vertex vv, denoted XvX_{v}, is the subcomplex of XX consisting of all simplices of XX that are disjoint from vv and that, together with vv, span a simplex of XX.

Definition 2.1.

A simplicial complex XX is mm-located for m4m\geq 4 if it is flag and whenever a dwheel subcomplex W=W1W2W=W_{1}\cup W_{2} of XX satisfies the conditions

  1. (1)

    W\partial W has length at most mm, and

  2. (2)

    the wheels W1W_{1} and W2W_{2} are full subcomplexes of XX,

the dwheel subcomplex WW is contained in the link XvX_{v} of some vertex vv.

Remark 2.2.

In the definition of mm-locatedness, we may weaken the requirement that WW is contained in the link XvX_{v} of a vertex by requiring only that WW is contained in the 11-ball B1(v)B_{1}(v) centered at a vertex. This results in an equivalent definition since if WW is contained in B1(v)B_{1}(v) then either it is contained in XvX_{v} or vWv\in W, which implies that the wheels of WW are not both full.

Definition 2.3.

Let σ\sigma be a simplex of XX. A flag simplicial complex is kk-large if there are no full jj-cycles in XX, when 4jk14\leq j\leq k-1. We say XX is locally kk-large if all its links are kk-large. Note that if XX is kk-large then it is locally kk-large.

For the purposes of this paper, we need only refer to local weak systolicity, a condition which together with simple connectedness is equivalent to weak systolicity by the local-to-global theorem of Chepoi and Osajda [8, Theorem A]. In order to define local weak systolicity, we first need to introduce some terminology. An extended 55-wheel W^5=(c;v1,v2,v3,v4,v5;a)\widehat{W}_{5}=(c;v_{1},v_{2},v_{3},v_{4},v_{5};a) in a flag simplicial complex XX is a subcomplex consisting of a 55-wheel W5=(c;v1,v2,v3,v4,v5)W_{5}=(c;v_{1},v_{2},v_{3},v_{4},v_{5}) together with an additional vertex aa that spans a triangle a,v1,v2\langle a,v_{1},v_{2}\rangle with v1v_{1} and v2v_{2} but that is not adjacent to any other vertex of W5W_{5}.

Definition 2.4.

A flag simplicial complex XX is locally weakly systolic if it is 55-large and every extended 55-wheel is contained in the link of a vertex. This latter condition is called the W^5\widehat{W}_{5}-condition.

3. Weakly systolic complexes are 77-located

In this section we prove that locally weakly systolic complexes are 77-located. Because locally weakly systolic complexes are 55-large, they thus provide a large class of examples of 77-located locally 55-large complexes.

w5w_{5}v1=w1v_{1}=w_{1}v2v_{2}v3=w4v_{3}=w_{4}v4v_{4}v5v_{5}w2w_{2}w3w_{3}
Figure 2. A planar (5,5)(5,5)-dwheel.
Theorem 3.1.

Let XX be a locally weakly systolic complex. Then XX is 77-located.

Proof.

Let W=W1W2W=W_{1}\cup W_{2} be a (k,l)(k,l)-dwheel subcomplex of XX of boundary length at most 77 such that W1W_{1} and W2W_{2} are full subcomplexes. By 55-largeness of XX, we have k,l5k,l\geq 5. Then, since WW has boundary length at most 77, either WW is a planar or nonplanar (5,5)(5,5)-dwheel or WW is a planar (5,6)(5,6)-dwheel. By Remark 2.2 it suffices to show that WW is contained in the 11-ball centered at a vertex of XX.

Claim 3.2.

If W=W1W2W=W_{1}\cup W_{2} is a planar (5,5)(5,5)-dwheel then WW is contained in a 11-ball of XX.

Proof.

Let W1W2=(w5;v1,v2,v3,v4,v5)(v2;w1=v1,w2,w3,w4=v3,w5)W_{1}\cup W_{2}=(w_{5};v_{1},v_{2},v_{3},v_{4},v_{5})\cup(v_{2};w_{1}=v_{1},w_{2},w_{3},w_{4}=v_{3},w_{5}). Then W1W_{1} and W2W_{2} are full. See Figure 2.

Consider W1{w2}W_{1}\cup\{w_{2}\}. Note that w2v1w_{2}\sim v_{1} and w2v2w_{2}\sim v_{2} and that, since W2W_{2} is a full kk-wheel, the vertices w2w_{2} and w5w_{5} are distinct and w2w5w_{2}\nsim w_{5} and w2v3w_{2}\nsim v_{3}. If we had w2v4w_{2}\sim v_{4} then (w2,v2,w5,v4)(w_{2},v_{2},w_{5},v_{4}) would be a 44-cycle and then 55-largeness of XX would contradict fullness of either W1W_{1} or W2W_{2}. So w2v4w_{2}\nsim v_{4}. We also have w2v5w_{2}\nsim v_{5} arguing otherwise similarly using the 44-cycle (w2,v2,w5,v5)(w_{2},v_{2},w_{5},v_{5}). Thus W1{w2}=(w5;v1,v2,v3,v4,v5;w2)W_{1}\cup\{w_{2}\}=(w_{5};v_{1},v_{2},v_{3},v_{4},v_{5};w_{2}) is an extended 55-wheel. Thus, by the W5^\widehat{W_{5}}-condition, there is a vertex w¯2\bar{w}_{2} that is adjacent to but distinct from every vertex of W1{w2}W_{1}\cup\{w_{2}\}. If w¯2=w3\bar{w}_{2}=w_{3} then W1W2=W1{w2,w3}W_{1}\cup W_{2}=W_{1}\cup\{w_{2},w_{3}\} is contained in the 11-ball centered at w¯2\bar{w}_{2} and we are done. So we may assume that w¯2W1W2\bar{w}_{2}\notin W_{1}\cup W_{2}. By a similar argument, there is a vertex w¯3W1W2\bar{w}_{3}\notin W_{1}\cup W_{2} that is adjacent to every vertex of W1{w3}W_{1}\cup\{w_{3}\}. If w¯2=w¯3\bar{w}_{2}=\bar{w}_{3} then W1W2=W1{w2,w3}W_{1}\cup W_{2}=W_{1}\cup\{w_{2},w_{3}\} is contained in the 11-ball centered at w¯2\bar{w}_{2} and we are done. So we may assume that w¯2w¯3\bar{w}_{2}\neq\bar{w}_{3}. Then (v1,w¯2,v3,w¯3)(v_{1},\bar{w}_{2},v_{3},\bar{w}_{3}) is a 44-cycle and 55-largeness of XX and fullness of W1W_{1} imply that w¯2w¯3\bar{w}_{2}\sim\bar{w}_{3}. Finally, the 55-largeness of XX in consideration of the 44-cycle (w¯2,w2,w3,w¯3)(\bar{w}_{2},w_{2},w_{3},\bar{w}_{3}) implies that W1W2W_{1}\cup W_{2} is contained in the 11-ball centered at either w¯2\bar{w}_{2} or w¯3\bar{w}_{3}. ∎

Claim 3.3.

If W=W1W2W=W_{1}\cup W_{2} is a nonplanar (5,5)(5,5)-dwheel then WW is contained in a 11-ball of XX.

Proof.

Let W1W2=(w5;v1,v2,v3,v4,v5)W_{1}\cup W_{2}=(w_{5};v_{1},v_{2},v_{3},v_{4},v_{5}) (v2;w1,w2,w3,w4=v3,w5)\cup(v_{2};w_{1},w_{2},w_{3},w_{4}=v_{3},w_{5}). See Figure 3.

w5w_{5}v1v_{1}v2v_{2}v3=w4v_{3}=w_{4}v4v_{4}v5v_{5}w1w_{1}w2w_{2}w3w_{3}
Figure 3. A nonplanar (5,5)(5,5)-dwheel from the proof of Claim 3.3. The encircled vertices form an extended 55-wheel W^1\widehat{W}_{1} contained in the link XpX_{p} of the vertex pp. The ensquared vertices form an extended 55-wheel W^2\widehat{W}_{2} contained in XqX_{q}.

Let W^1=W1{w3}=(w5;v1,v2,v3,v4,v5;w3)\widehat{W}_{1}=W_{1}\cup\{w_{3}\}=(w_{5};v_{1},v_{2},v_{3},v_{4},v_{5};w_{3}). We will show that W^1\widehat{W}_{1} is an extended 55-wheel. Note that W1W_{1} and W2W_{2} are full because they form a dwheel. This implies that w3w_{3} is distinct from the vertices of W2W_{2}. It remains to show that w3w_{3} does not form an edge with any of w5w_{5}, v4v_{4}, v5v_{5} or v1v_{1}. That w3w5w_{3}\nsim w_{5} follows by fullness of W2W_{2}. The remaining cases would each introduce a 44-cycle:

  • w3v4w_{3}\sim v_{4} introduces the 44-cycle (w3,v2,w5,v4)(w_{3},v_{2},w_{5},v_{4}).

  • w3v5w_{3}\sim v_{5} introduces the 44-cycle (w3,v3,w5,v5)(w_{3},v_{3},w_{5},v_{5}).

  • w3v1w_{3}\sim v_{1} introduces the 44-cycle (w3,v3,w5,v1)(w_{3},v_{3},w_{5},v_{1}).

The 55-largeness of XX applied to any of these 44-cycles would contradict fullness of W1W_{1} or W2W_{2}. Thus W^1\widehat{W}_{1} is an extended 55-wheel and so, by the W^5\widehat{W}_{5}-condition, there exists a vertex pp of XX such that W^1Xp\widehat{W}_{1}\subset X_{p}. Fullness of W2W_{2} implies that pp is distinct w1w_{1} and w2w_{2} so pp is distinct from the vertices of W=W1W2W=W_{1}\cup W_{2}. By a similar argument, we have that W^2=W2{v4}=(v2;w1,w2,w3,w4,w5;v4)\widehat{W}_{2}=W_{2}\cup\{v_{4}\}=(v_{2};w_{1},w_{2},w_{3},w_{4},w_{5};v_{4}) is an extended 55-wheel contained in XqX_{q} for some vertex qq of XWX\setminus W. If p=qp=q then WXpW\subset X_{p} and we are done so we may assume that pp and qq are distinct. Thus we have a 44-cycle (p,q,w1,v1)(p,q,w_{1},v_{1}) to which we can apply 55-largeness. Without loss of generality, we have pw1p\sim w_{1} so that we have a 44-cycle (p,w1,w2,w3)(p,w_{1},w_{2},w_{3}). But w1w3w_{1}\nsim w_{3} by fullness of W2W_{2} so, applying 55-largeness, we have pw2p\sim w_{2} and so W=W1W2B1(p)W=W_{1}\cup W_{2}\subset B_{1}(p). ∎

Claim 3.4.

If W=W1W2W=W_{1}\cup W_{2} is a planar (5,6)(5,6)-dwheel then WW is contained in a 11-ball of XX.

Proof.

Let W1W2=(w6;v1,v2,v3,v4,v5)(v2;w1=v1,w2,w3,W_{1}\cup W_{2}=(w_{6};v_{1},v_{2},v_{3},v_{4},v_{5})\cup(v_{2};w_{1}=v_{1},w_{2},w_{3}, w4,w5=v3,w6)w_{4},w_{5}=v_{3},w_{6}). Then W1W_{1} and W2W_{2} are full. Let W^1=W1{w2}=(w6;v1,v2,v3,v4,v5;w2)\widehat{W}_{1}=W_{1}\cup\{w_{2}\}=(w_{6};v_{1},v_{2},v_{3},v_{4},v_{5};w_{2}). Let W^1=W1{w4}=(w6;v1,v2,v3,v4,v5;w4)\widehat{W}_{1}^{\prime}=W_{1}\cup\{w_{4}\}=(w_{6};v_{1},v_{2},v_{3},v_{4},v_{5};w_{4}). See Figure 4.

w6w_{6}v1=w1v_{1}=w_{1}v2v_{2}v3=w5v_{3}=w_{5}v4v_{4}v5v_{5}w2w_{2}w3w_{3}w4w_{4}
Figure 4. A planar (5,6)(5,6)-dwheel from the proof of Claim 3.4. The encircled vertices form an extended 55-wheel W^1\widehat{W}_{1} contained in the link XpX_{p} of the vertex pp. The ensquared vertices form an extended 55-wheel W^1\widehat{W}_{1}^{\prime} contained in XqX_{q}.

By fullness of W2W_{2}, we have w2w6w_{2}\neq w_{6}, w2w6w_{2}\nsim w_{6} and w2v3w_{2}\nsim v_{3}. Then, to show that W^1\widehat{W}_{1} is an extended 55-wheel, it suffices to show w2v4w_{2}\nsim v_{4} and w2v5w_{2}\nsim v_{5}. But w2v4w_{2}\sim v_{4} would introduce the 44-cycle (w2,v2,w6,v4)(w_{2},v_{2},w_{6},v_{4}) and w2v5w_{2}\sim v_{5} would introduce the 44-cycle (w2,v2,w6,v5)(w_{2},v_{2},w_{6},v_{5}). Applying 55-largeness to either of these 44-cycles would contradict fullness of either W1W_{1} or W2W_{2}. Thus W^1\widehat{W}_{1} is an extended 55-wheel and so is contained in the link XpX_{p} of a vertex pp of XX. Fullness of W2W_{2} ensures that pp is distinct from w3w_{3} and w4w_{4} so that pW=W1W2p\notin W=W_{1}\cup W_{2}. By symmetry, the same argument can be applied to W^1\widehat{W}_{1}^{\prime} to show that W^1Xq\widehat{W}_{1}^{\prime}\subset X_{q} for some vertex qq of XWX\setminus W.

If p=qp=q then we could apply 55-largeness to the 44-cycle (w2,w3,w4,p)(w_{2},w_{3},w_{4},p) and then the fullness of W2W_{2} would imply that pw3p\sim w_{3}. Then we would have WXpW\subset X_{p}, which would prove the claim. So we may assume that pqp\neq q. Then fullness of W1W_{1} and 55-largeness applied to the 44-cycle (v1,p,v3,q)(v_{1},p,v_{3},q) implies that pqp\sim q.

w6w_{6}v1=w1v_{1}=w_{1}v2v_{2}v3=w5v_{3}=w_{5}v4v_{4}v5v_{5}w2w_{2}w3w_{3}w4w_{4}ppqq
Figure 5. The extended 55-wheel W^\widehat{W} from the proof of Claim 3.4.

Next we need to show that W^=(v2;p,w2,w3,w4,q;v5)\widehat{W}=(v_{2};p,w_{2},w_{3},w_{4},q;v_{5}) is an extended 55-wheel. See Figure 5. Fullness of W1W_{1} implies that v5v2v_{5}\neq v_{2} and v5v2v_{5}\nsim v_{2}. We need to show that v5v_{5} is not adjacent to w2w_{2}, w3w_{3} or w4w_{4}. But adjacency of v5v_{5} to any of these vertices would introduce a 44-cycle in WW two which applying 55-largeness would either contradict fullness of W1W_{1} or fullness of W2W_{2}. Thus W^\widehat{W} is an extended 55-wheel and so is contained in the link XrX_{r} of a vertex rr of XX. Fullness of W1W_{1} and W2W_{2} ensure that rWr\notin W.

Since v5v2v_{5}\nsim v_{2}, the 55-largeness of XX applied to 44-cycles (v5,r,v2,v1)(v_{5},r,v_{2},v_{1}) and (v5,r,v2,w6)(v_{5},r,v_{2},w_{6}) imply that rr is adjacent to v1v_{1} and w6w_{6}. Then, similarly, the 44-cycle (w6,r,w4,v3)(w_{6},r,w_{4},v_{3}) implies that rv3r\sim v_{3}. Finally, the 44-cycle (v5,r,v3,v4)(v_{5},r,v_{3},v_{4}) implies that rv4r\sim v_{4}. Thus WB1(r)W\subset B_{1}(r). ∎

The above three claims together imply that XX is 77-located. ∎

4. A CAT(0)\operatorname{CAT}(0) metric for 77-located disc diagrams

In this section we show that the triangles of any 77-located disc DD can be metrized in a natural way as Euclidean triangles such that DD is CAT(0)\operatorname{CAT}(0). Only three isometry types of triangles are required and the choice of isometry type for each 22-simplex σ\sigma will depend only on the local combinatorics around σ\sigma.

Lemma 4.1.

Let DD be a 77-located disc. Then any pair of adjacent internal vertices u,vu,v of DD satisfy deg(v)+deg(w)12\deg(v)+\deg(w)\geq 12.

Proof.

Since DD is a simplicial complex and vv is an interior vertex, it is the center of a full wheel WvW_{v}. Similarly, we have a full wheel WwW_{w} in DD with center ww and WvWwW_{v}\cup W_{w} is a planar dwheel of boundary length deg(v)+deg(w)4\deg(v)+\deg(w)-4. Since DD is planar, the dwheel WvWwW_{v}\cup W_{w} cannot be contained in a 11-ball. Thus deg(v)+deg(w)48\deg(v)+\deg(w)-4\geq 8. ∎

Definition 4.2.

Let WW be a kk-wheel with central vertex vv. We metrize each triangle of WW as a Euclidean triangle whose angle at vv is 2πk\frac{2\pi}{k}, whose remaining two angles are both equal to (k2)π2k\frac{(k-2)\pi}{2k} and whose boundary edge eWe\subset\partial W has length 11. We then metrize WW as a Euclidean polygonal complex. We call the resulting metric the flattened wheel metric on WW.

Remark 4.3.

A kk-wheel WW with the flattened wheel metric is isometric to a regular Euclidean kk-gon of side length 11. Under this isometry, the central vertex is sent to the center of the kk-gon.

Definition 4.4.

Let DD be a simplicial disc such that for any k1k_{1}-wheel W1DW_{1}\subset D and any k2k_{2}-wheel W2DW_{2}\subset D, if k1<6k_{1}<6 and k2<6k_{2}<6 then either W1=W2W_{1}=W_{2} or W1W2W1W2W_{1}\cap W_{2}\subset\partial W_{1}\cap\partial W_{2}. The flattened wheel metric on DD is the Euclidean polygonal complex metric obtained by metrizing each kk-wheel of DD for which k<6k<6 with the flattened wheel metric and metrizing any remaining triangles as regular Euclidean triangles of side length 11. With this metric, the flattened wheels of DD are its kk-wheels with k<6k<6.

Lemma 4.5.

Let DD be a 77-located disc. Let W1W_{1} and W2W_{2} be distinct wheels of DD of boundary length less than 66. Then W1W2W1W2W_{1}\cap W_{2}\subset\partial W_{1}\cap\partial W_{2}. In particular, the flattened wheel metric is well-defined for DD.

Proof.

If W1W_{1} and W2W_{2} intersect in their interiors then they must share a triangle σ\sigma. Thus the central vertices v1v_{1} and v2v_{2} of W1W_{1} and W2W_{2} are adjacent. But then, by Lemma 4.1, we have |W1|+|W2|=deg(v1)+deg(v2)12|\partial W_{1}|+|\partial W_{2}|=\deg(v_{1})+\deg(v_{2})\geq 12 so that one of the wheels must have boundary length at least 66. ∎

In the next theorem we establish a connection between some 77-located discs and CAT(0)\operatorname{CAT}(0) spaces.

Theorem 4.6.

Let DD be a 77-located simplicial disc endowed with the flattened wheel metric. Then DD is CAT(0)\operatorname{CAT}(0).

Proof.

Since DD is simply connected, it suffices to show that the sum of the angles of the corners of triangles incident to any interior vertex vv of DD is at least 2π2\pi [4, Theorem II.5.4 and Lemma II.5.6]. Because DD is flag, it has no interior vertices of degree 33. Any interior vertex of degree 44 or 55 is the central vertex of a flattened wheel and so has angle sum exactly 2π2\pi. Away from flattened wheel centers, corner angles of triangles are either π4\frac{\pi}{4}, 3π10\frac{3\pi}{10} or π3\frac{\pi}{3}. Thus any interior vertex of degree at least 88 has angle sum at least 2π2\pi.

Thus we may assume that vv is an interior vertex of degree 66 or 77. If vv is not incident to any flattened wheel then its incident triangles are all regular Euclidean and so have corner angle π3\frac{\pi}{3} so that vv has angle sum 6π3=2π\frac{6\pi}{3}=2\pi or 7π3>2π\frac{7\pi}{3}>2\pi. So we may assume that vv is incident to a flattened wheel. Since deg(v)7\deg(v)\leq 7, by Lemma 4.1, any central vertex of a flattened wheel WW incident to vv has degree 55. Then vv is incident to a vertex of degree 55 and any triangle corner incident to vv has angle at least 3π10\frac{3\pi}{10}. So, by Lemma 4.1, we have deg(v)=7\deg(v)=7 and so vv has angle sum at least 21π10>2π\frac{21\pi}{10}>2\pi. ∎

Corollary 4.7.

There is a uniform quadratic function f:f\colon\mathbb{N}\to\mathbb{N} such that any 77-located disc DD of boundary length |D|=n|\partial D|=n has at most f(n)f(n) triangles.

Proof.

The area of a minimal disc diagram for a loop of length nn in a CAT(0)\operatorname{CAT}(0) space is bounded by the area of a circle of circumference nn in the Euclidean plane [4, Theorem III.2.17]. It follows that if DD is a 77-located disc of boundary length nn, it has area at most n24π\frac{n^{2}}{4\pi} when endowed with the flattened wheel metric. Each triangle of DD has area at least 14\frac{1}{4} so the number of triangles of DD is at most f(n)=n2πf(n)=\frac{n^{2}}{\pi}. ∎

5. A Minimal Disc Diagram Lemma for 77-located locally 55-large complexes

In this section we prove that a minimal area disc diagram in a 77-located locally 55-large complex is 77-located.

Remark 5.1.

The simplicial complex XX in Figure 6 is 77-located but it is not locally 55-large. However, it has a minimal area disc diagram (D,f)(D,f) that is not 77-located. This example justifies the additional hypothesis of local 55-largeness for the minimal disc diagram lemma.

Refer to caption
Figure 6.

We shall refer to the following lemmas frequently in the rest of this section.

Lemma 5.2.

Let XX be a flag simplicial complex and let γ\gamma be a homotopically trivial loop in XX. Let (D,f)(D,f) be a disc diagram for γ\gamma. Let (u,a,v,b)(u,a,v,b) be a 44-cycle in XX. If f(u)=f(v)f(u)=f(v) then there exists a disc diagram (D,f)(D^{\prime},f^{\prime}) for γ\gamma of lesser area than DD.

Proof.

The subdisc bounded by (u,a,v,b)(u,a,v,b) has at least two triangles. We delete this subdisc. We glue uu to vv, the edge a,u\langle a,u\rangle to the edge a,v\langle a,v\rangle, and the edge b,u\langle b,u\rangle to the edge b,v\langle b,v\rangle. Thus we get a new disc DD^{\prime}. See Figure 7. Let ff^{\prime} be the map induced by the gluing. This is well defined since f(u)=f(v)f(u)=f(v), f(a,u)=f(a,v)f(\langle a,u\rangle)=f(\langle a,v\rangle) and f(b,u)=f(b,v)f(\langle b,u\rangle)=f(\langle b,v\rangle). Thus (D,f)(D^{\prime},f^{\prime}) is a disc diagram for γ\gamma of lesser area than DD.

Refer to caption
Figure 7.

Lemma 5.3.

Let XX be a flag simplicial complex and let γ\gamma be a homotopically trivial loop in XX. Let (D,f)(D,f) be a disc diagram for γ\gamma. Let u,vu,v be vertices of a 55-cycle α=(a,v,b,c,u)\alpha=(a,v,b,c,u) of DD with f(u)=f(v)f(u)=f(v). Then there exists a disc diagram (D,f)(D^{\prime},f^{\prime}) for γ\gamma of lesser area than DD.

Proof.

The subdisc bounded by α\alpha has at least three 22-simplices. We delete this subdisc. We obtain a new disc diagram DD^{\prime} by glueing uu to vv, gluing the edge a,u\langle a,u\rangle to the edge a,v\langle a,v\rangle and gluing a 22-simplex u,b,c\langle u,b,c\rangle. See Figure 8. Let ff^{\prime} be the map induced by the gluing. This is well defined since f(u)=f(v)f(u)=f(v), f(a,u)=f(a,v)f(\langle a,u\rangle)=f(\langle a,v\rangle) and, by flagness, f(u),f(b),f(c)\langle f(u),f(b),f(c)\rangle is a triangle in XX. Thus (D,f)(D^{\prime},f^{\prime}) is a disc diagram for γ\gamma of lesser area than DD.

Refer to caption
Figure 8.

Lemma 5.4.

Let XX be a flag simplicial complex and let γ\gamma be a homotopically trivial loop in XX. Let (D,f)(D,f) be a disc diagram for γ\gamma. Let u,vu,v be vertices of a 66-cycle α=(u,a,b,v,c,d)\alpha=(u,a,b,v,c,d) of DD with d(u,v)=3d(u,v)=3 such that f(u)=f(v)f(u)=f(v). Then there exists a disc diagram (D,f)(D^{\prime},f^{\prime}) for γ\gamma of lesser area than DD.

Proof.

The subdisc bounded by α\alpha has at least four 22-simplices. We obtain a new disc diagram DD^{\prime} by deleting the subdisc bounded by α\alpha, gluing uu to vv and then gluing in two triangles a,b,u\langle a,b,u\rangle and u,c,d\langle u,c,d\rangle. See Figure 9. Let ff^{\prime} be the map induced by gluing. This is well defined since f(u)=f(v)f(u)=f(v) and, by flagness, f(a),f(b),f(u)\langle f(a),f(b),f(u)\rangle, f(u),f(c),f(d)\langle f(u),f(c),f(d)\rangle are triangles in XX. Hence (D,f)(D^{\prime},f^{\prime}) is a disc diagram for γ\gamma of lesser area than DD.

Refer to caption
Figure 9.

Lemma 5.5.

Let XX be a flag and locally 55-large simplicial complex. Let (D,f)(D,f) be a minimal area disc diagram for a cycle γ\gamma in XX. Then DD does not contain a 44-wheel.

Proof.

Suppose WW is a 44-wheel in DD. If the restriction f|Wf|_{W} is not injective then some pair of antipodal vertices u,vWu,v\in\partial W have a common image f(u)=f(v)f(u)=f(v). This contradicts minimality by Lemma 5.2.

If the restriction f|Wf|_{W} is injective then, since XX is locally 55-large, a pair of antipodal vertices u,vWu,v\in\partial W have images f(u)f(u) and f(v)f(v) that are joined by an edge ee in XX. Since XX is flag, the image f(W)f(\partial W) along with ee span two triangles so that f(W)f(\partial W) has a disc diagram of area 22. Thus we may remove the interior of WW in DD and replace it with an edge joining uu and vv and a pair of triangles to obtain a disc diagram for γ\gamma of lesser area, contradicting minimality of DD. ∎

Lemma 5.6.

Let XX be a flag and locally 55-large simplicial complex. Let (D,f)(D,f) be a minimal area disc diagram for a cycle γ\gamma in XX and let WW be a 55-wheel in DD. Then the image f(W)f(W) is a full 55-wheel of XX.

Proof.

It follows from Lemma 5.3 that f|Wf|_{W} is injective. Then f(W)f(W) is a 55-wheel in XX. We need only verify that it is full. If not then some pair of vertices u,vWu,v\in\partial W that are not adjacent in W\partial W have images f(u)f(u) and f(v)f(v) that are joined by an edge ee. Since XX is flag, the edge ee spans a triangle with each of f(u)f(u) and f(v)f(v). Let aa be the unique vertex of W\partial W adjacent to both uu and vv. Let cc be the central vertex of WW. Then (u,a,v,c)(u,a,v,c) is a 44-cycle bounding a subdisc of DD with two triangles. We delete this subdisc, join uu and vv by an edge ee^{\prime} and add triangles a,u,v\langle a,u,v\rangle and c,u,v\langle c,u,v\rangle to obtain a new disc diagram DD^{\prime} for γ\gamma of the same (minimal) area. In particular, DD^{\prime} is a minimal area disc diagram. But the triangle c,u,v\langle c,u,v\rangle along with the three remaining original triangles of WW form a 44-wheel, which contradicts minimality by Lemma 5.5. ∎

Lemma 5.7.

Let XX be a 77-located locally 55-large simplicial complex. Let (D,f)(D,f) be a minimal area disc diagram for a cycle γ\gamma in X. Let WW be a (5,5)(5,5)- or (5,6)(5,6)-dwheel of DD. Then the restriction f|Wf|_{W} is not injective.

Proof.

For the sake of deriving a contradiction, we assume that f|Wf|_{W} is injective. Since DD is planar, the dwheel WW is also planar. We consider first the case where WW is a planar (5,5)(5,5)-dwheel. By Lemma 5.6, the (5,5)(5,5)-dwheel f(W)f(W) has full wheels. Then, since XX is 77-located, the image f(W)f(W) is contained in the 11-ball centered at a vertex vv of XX. Since XX is flag, it follows that f(W)f(\partial W) has a disc diagram with at most 66 triangles, contradicting minimality of DD.

w6w_{6}v1=w1v_{1}=w_{1}v2v_{2}v3=w5v_{3}=w_{5}v4v_{4}v5v_{5}w2w_{2}w3w_{3}w4w_{4}
Figure 10. A planar (5,6)(5,6)-dwheel.

We now consider the case where WW is a planar (5,6)(5,6)-dwheel. Let W=W1W2=(w6;v1,v2,v3,v4,v5)(v2;w1=v1,w2,w3,W=W_{1}\cup W_{2}=(w_{6};v_{1},v_{2},v_{3},v_{4},v_{5})\cup(v_{2};w_{1}=v_{1},w_{2},w_{3}, w4,w5=v3,w6)w_{4},w_{5}=v_{3},w_{6}); see Figure 10. By Lemma 5.6, the image f(W1)f(W_{1}) is a full 55-wheel. If the 66-wheel f(W2)f(W_{2}) is also a full then we can argue as in the (5,5)(5,5)-dwheel case that f(W)f(\partial W) has a disc diagram with at most 77 triangles, which leads to a contradiction with the minimality of DD. Thus we need only consider the case where f(W2)f(W_{2}) is not full. Then some pair of vertices u,vW2u,v\in\partial W_{2} that are not adjacent in W2\partial W_{2} have images f(u)f(u) and f(v)f(v) that are adjacent.

Claim 5.8.

Either {u,v}={w6,w2}\{u,v\}=\{w_{6},w_{2}\} or {u,v}={w6,w4}\{u,v\}=\{w_{6},w_{4}\}.

Proof.

We first rule out the possibility that uu and vv are antipodal in W2\partial W_{2}. If this were so, we could cut DD open along the path (u,v2,w)(u,v_{2},w) and fill the resulting boundary path with an edge joining uu and ww (mapping to the edge f(u),f(v)\langle f(u),f(v)\rangle) and a pair of triangles (which map to triangles in XX by flagness). In the resulting disc diagram for γ\gamma, the cycle W2\partial W_{2} bounds a disc with two 44-wheels. But, by arguments similar to those in the proof of Lemma 5.5, each 44-wheel boundary (as a cycle in XX) has a disc diagram with at most two triangles. Thus W2\partial W_{2} has a disc diagram with at most four triangles, contradicting minimality of (D,f)(D,f).

Thus uu and vv are at distance 22 in W2\partial W_{2}. Since f(W1)f(W_{1}) is full in XX, we have {u,v}{w1,w5}\{u,v\}\neq\{w_{1},w_{5}\}. It remains to rule out the remaining possibilities: {u,v}={w1,w3}\{u,v\}=\{w_{1},w_{3}\}, {u,v}={w2,w4}\{u,v\}=\{w_{2},w_{4}\} and {u,v}={w3,w5}\{u,v\}=\{w_{3},w_{5}\}. In these cases, we perform a disc diagram surgery which turns v2v_{2} into the center of a 55-wheel, similar to the surgery performed in the proof of Lemma 5.6 that produced a 44-wheel. The resulting disc diagram has the same (minimal) area but has a 55-wheel and thus the argument reduces to the (5,5)(5,5)-dwheel case. ∎

Thus the additional edges in the full subcomplex induced on the 66-wheel f(W2)f(W_{2}) are either f(w6),f(w2)\langle f(w_{6}),f(w_{2})\rangle or f(w6),f(w4)\langle f(w_{6}),f(w_{4})\rangle or both. If just one of these is present then f(W)f(W) spans a non-planar (5,5)(5,5)-dwheel with full wheels so that W\partial W has a disc diagram with at most 77 triangles, contradicting minimality of (D,f)(D,f). If, on the other hand, both edges are present then we perform two disc diagram surgeries as in the proof of Lemma 5.5 transforming WW from a (5,6)(5,6)-dwheel to a (7,4)(7,4)-dwheel. The resulting disc diagram has the same (minimal) area and yet contains a 44-wheel, contradicting Lemma 5.5. ∎

We prove next the minimal disc diagram lemma for 77-located locally 55-large simplicial complexes.

Theorem 5.9.

Let XX be a 77-located locally 55-large simplicial complex. Let γ\gamma be a homotopically trivial loop in XX. Any minimal area disc diagram (D,f)(D,f) for γ\gamma is 77-located and locally 55-large.

Proof.

By Lemma 5.5, there are no 44-wheels in DD. It follows that DD is locally 55-large so we need only prove that DD is 77-located. Suppose for the sake of finding a contradiction that DD is not 77-located. Then DD has a dwheel W=W1W2W=W_{1}\cup W_{2} of boundary length at most 7 that is not contained in a 11-ball. Since DD is planar, so is WW. Since DD has no 44-wheels, the dwheel WW must be a planar (5,5)(5,5)- or (5,6)(5,6)-dwheel. By Lemma 5.7, the restriction f|Wf|_{W} is not injective. Then there exist distinct vertices v,wWv,w\in W such that f(v)=f(w)f(v)=f(w).

We consider first the case where vv and ww belong to a common wheel WiW_{i} of WW. Since ff is a nondegenerate map and XX is simplicial, the vertices vv and ww cannot be adjacent. Thus they must be contained in the boundary Wi\partial W_{i} and be at distance 22 or 33 in Wi\partial W_{i}. Note that WiW_{i} is necessarily a 66-wheel in the case that dWi(v,w)=3d_{\partial W_{i}}(v,w)=3. But then we contradict minimality by Lemma 5.2 if dWi(v,w)=2d_{\partial W_{i}}(v,w)=2 or by Lemma 5.4 if dWi(v,w)=3d_{\partial W_{i}}(v,w)=3.

We consider now the case where vv and ww do not belong to a common wheel of WW. It will be helpful here to name the vertices of WW so let W=W1W2=(wk;v1,v2,,v5)(v2;w1=v1,w2,,wk1=v3,wk)W=W_{1}\cup W_{2}=(w_{k};v_{1},v_{2},\dots,v_{5})\cup(v_{2};w_{1}=v_{1},w_{2},\dots,w_{k-1}=v_{3},w_{k}) with k{5,6}k\in\{5,6\}. We first consider the (5,5)(5,5)-dwheel case, i.e., the case where k=5k=5. See Figure 2. Up to symmetry, there are only two cases to consider: (v,w)=(v4,w3)(v,w)=(v_{4},w_{3}) and (v,w)=(v4,w2)(v,w)=(v_{4},w_{2}). In the case (v,w)=(v4,w3)(v,w)=(v_{4},w_{3}), the 55-cycle (v4,v3,w3,v2,w5)(v_{4},v_{3},w_{3},v_{2},w_{5}) contradicts minimality by Lemma 5.3. In the case (v,w)=(v4,w2)(v,w)=(v_{4},w_{2}), the 66-cycle W=(v4,v3,w3,w2,w1,v5)\partial W=(v_{4},v_{3},w_{3},w_{2},w_{1},v_{5}) contradicts minimality by Lemma 5.4.

We now consider the (5,6)(5,6)-dwheel case, i.e., the case where k=6k=6. See Figure 10. Up to symmetry, there are three cases to consider: (v,w)=(v4,w4)(v,w)=(v_{4},w_{4}), (v,w)=(v4,w3)(v,w)=(v_{4},w_{3}) and (v,w)=(v4,w2)(v,w)=(v_{4},w_{2}). As with the (5,5)(5,5)-dwheel, in each case we contradict minimality by applying either Lemma 5.3 to a 55-cycle or Lemma 5.4 to a 66-cycle. Specifically, for the case (v,w)=(v4,w4)(v,w)=(v_{4},w_{4}) we consider the 55-cycle (v4,v3,w4,v2,w6)(v_{4},v_{3},w_{4},v_{2},w_{6}), for the case (v,w)=(v4,w3)(v,w)=(v_{4},w_{3}) we consider the 66-cycle (v4,v3,w4,w3,v2,w6)(v_{4},v_{3},w_{4},w_{3},v_{2},w_{6}) and for the (v,w)=(v4,w2)(v,w)=(v_{4},w_{2}) case we consider the 66-cycle (v4,w6,v2,w2,w1,v5)(v_{4},w_{6},v_{2},w_{2},w_{1},v_{5}).

References

  • [1] H.-J. Bandelt. Hereditary modular graphs. Combinatorica, 8(2):149–157, 1988.
  • [2] H.-J. Bandelt and E. Pesch. Dismantling absolute retracts of reflexive graphs. European J. Combin., 10(3):211–220, 1989.
  • [3] B. Brešar, J. Chalopin, V. Chepoi, T. Gologranc, and D. Osajda. Bucolic complexes. Adv. Math., 243:127–167, 2013.
  • [4] M. R. Bridson and A. Haefliger. Metric spaces of non-positive curvature, volume 319 of Grundlehren der Mathematischen Wissenschaften [Fundamental Principles of Mathematical Sciences]. Springer-Verlag, Berlin, 1999.
  • [5] J. Chalopin, V. Chepoi, A. Genevois, H. Hirai, and D. Osajda. Helly groups. arXiv preprint arXiv:2002.06895, 2020.
  • [6] J. Chalopin, V. Chepoi, H. Hirai, and D. Osajda. Weakly modular graphs and nonpositive curvature. arXiv preprint arXiv:1409.3892, 2014.
  • [7] V. Chepoi. Graphs of some CAT(0){\rm CAT}(0) complexes. Adv. in Appl. Math., 24(2):125–179, 2000.
  • [8] V. Chepoi and D. Osajda. Dismantlability of weakly systolic complexes and applications. Trans. Amer. Math. Soc., 367(2):1247–1272, 2015.
  • [9] M. Gromov. Hyperbolic groups. In Essays in group theory, volume 8 of Math. Sci. Res. Inst. Publ., pages 75–263. Springer, New York, 1987.
  • [10] F. Haglund. Complexes simpliciaux hyperboliques de grande dimension. Preprint, 2003.
  • [11] N. Hoda. Quadric complexes. Michigan Math. J., 69(2):241–271, 2020.
  • [12] T. Januszkiewicz and J. Świa̧tkowski. Simplicial nonpositive curvature. Publ. Math. Inst. Hautes Études Sci., 104(1):1–85, 2006.
  • [13] I.-C. Lazăr. A combinatorial negative curvature condition implying gromov hyperbolicity. Preprint, arXiv:1501.05487, 2015.
  • [14] I.-C. Lazăr. Minimal disc diagrams of 5/95/9-simplicial complexes. Michigan Math. J., 69:793–829, 2020.
  • [15] D. Osajda. A combinatorial non-positive curvature I: weak systolicity. Preprint, arXiv:1305.4661, 2013.
  • [16] D. Osajda. Combinatorial negative curvature and triangulations of three-manifolds. Indiana Univ. Math. J., 64(3):943–956, 2015.