10
$\begingroup$

Consider a set of $n$ elements $S=\lbrace 1,\dots,n\rbrace$ and $\mathcal{P}(S)$ to be the power set of $S$, which is a well-defined poset with respect to the inclusions. Now consider $\emptyset\neq T\varsubsetneq S$ and define $A=\lbrace \emptyset\neq U\in\mathcal{P}(S)\mid T\nsubseteq U\rbrace$. I want to compute the homotopy type of the geometric realization $\left|A\right|$.

My intuition tells me that $\left|A\right|\simeq\mathbb{S}^{\left|T\right|-2}$, however I don't know if this is right and how to prove it.

I know some facts such that if the poset has a maximal element or a minimal element then the geometric realization is contractible. Also, for some small cases I am able to do it by hand, however I don't know how to do it in general and whether there is some ''general'' procedure to study this problem.

Any help or references about this problem will be appreciated.

$\endgroup$
7
  • 1
    $\begingroup$ @SamHopkins Thanks!! You can add this as an answer if you want me to acept it. $\endgroup$
    – Marcos
    Jun 6 at 21:07
  • $\begingroup$ Thanks! That's an interesting example. $\endgroup$
    – Dry Bones
    Jun 6 at 23:16
  • $\begingroup$ How is the geometric realization of a poset defined? Is it the realization of its order complex (p6 of "Poset Topology: Tools and Applications" arxiv.org/pdf/math/0602226.pdf)? $\endgroup$
    – nasosev
    Jun 14 at 0:10
  • $\begingroup$ @nasosev: yes, that's right. $\endgroup$ Jun 14 at 13:04
  • $\begingroup$ @SamHopkins thanks. I was wondering, a topological space is itself a poset. So is there a relationship between the homotopy types of a space, and the realisation of the order complex of its poset of opens ? $\endgroup$
    – nasosev
    Jun 15 at 16:19

2 Answers 2

6
$\begingroup$

First recall that geometric realisation of posets preserves products: the projections $P\xleftarrow{p}P\times Q\xrightarrow{q}Q$ give a map $(|p|,|q|)\colon |P\times Q|\to|P|\times|Q|$, and it is a standard fact that this is a homeomorphism.

Next, if $f_0,f_1\colon P\to Q$ are morphisms of posets and $f_0(x)\leq f_1(x)$ for all $x$ then we can define a poset morphism $h\colon \{0,1\}\times P\to Q$ by $h(i,x)=f_i(x)$, and this gives a map $|h|\colon [0,1]\times|P|\to|Q|$ which is a homotopy between $|f_0|$ and $|f_1|$.

Now put $A=\{U\subseteq S\;:\;U\neq\emptyset, U\not\supseteq T\}$ as in the question. We can define $f_0,f_1,f_2\colon A\to A$ by $f_0(U)=U$ and $f_1(U)=U\cup T^c$ and $f_2(U)=T^c$. Then $f_0\leq f_1\geq f_2$, so the identity is homotopic to the constant map $|f_2|$, so $|A|$ is contractible.

Now suppose instead we consider the poset $B=\{U\subseteq S\;:\;T\cap U\neq\emptyset,U\not\supseteq T\}\subset A$, and the poset $C=\{U\;:\;\emptyset\subset U\subset T\}$. We have an inclusion $i\colon C\to B$ and a retraction $r\colon B\to C$ given by $r(U)=T\cap U$. These have $ri=1$ and $ir\leq 1$ so $|i|$ and $|r|$ are mutually inverse homotopy equivalences. Alternatively, we can say that $B\simeq C\times D$, where $D$ is the poset of all subsets of $T^c$. This gives $|B|\simeq|C|\times|D|$, where $|D|$ is homeomorphic to $[0,1]^{|T^c|}$ and in particular is contractible.

Now put $M=\{m:T\to\mathbb{R}\;:\;\sum_{t\in T}m(t)=0\}$ (which is a vector space of dimension $|T|-1$ with an obvious inner product). For $U\in C$ define $f(U)=\chi_U-|U||T|^{-1}\in M$. Extend this linearly over simplices to get a map $f\colon |C|\to M$. One can check that this is nowhere zero, so we can define $f_1(x)=f(x)/\|f(x)\|$, and this gives a homeomorphism $|C|\to S(M)\simeq S^{|T|-2}$.

$\endgroup$
1
  • 1
    $\begingroup$ Thanks, I didn't realize that the above proof was wrong. This means that I have to work harder to get confortable with those tools. $\endgroup$
    – Marcos
    Jun 8 at 10:17
13
$\begingroup$

Once more, with feeling. Thanks to comments from Tyler Lawson and Neil Strickland.

Let me use $B_n$ for the finite Boolean lattice of subsets of $[n] := \{1,2,\ldots,n\}$ (your $\mathcal{P}(S)$), and let $k=|T|$. Then your $A=\{U\subseteq[n]\colon U\neq \varnothing, T \not \subseteq U\}=((B_k\setminus \{\hat{1}\}) \times B_{n-k}) \setminus \{\hat{0}\}$, where $\hat{0}$ means the minimum of a poset, and $\hat{1}$ the maximum. To understand the homotopy type of this product, we need a result of Quillen (stated as Theorem 5.1(b) in the paper "Canonical homeomorphisms of posets" by Walker cited below; see also the discussion in Section 5.1 of the notes of Wachs), which says that $$ |P\times Q \setminus \{\hat{0}\}| \simeq |P\setminus \{\hat{0}\}| \ast |Q\setminus\{\hat{0}\}|,$$ for posets $P,Q$ which have minimums, where $\ast$ is the join of topological spaces. So in our case $$ |A| \simeq |B_k \setminus \{\hat{0},\hat{1}\}| \ast |B_{n-k}\setminus \{\hat{0}\}|.$$ But $B_{n-k}\setminus \{\hat{0}\}$ is contractible (it has a maximum), so $|A|$ is also contractible.

Note that if we considered the poset $B=\{U\subseteq [n]\colon U \cap T \neq \varnothing, T \not\subseteq U\}$ as in the answer of Neil Strickland, then we have that $B = (B_k \setminus \{\hat{0},\hat{1}\}) \times B_{n-k}$. [This is the poset that the first version of my answer was implicitly about.] Here we need a slightly different fact about homotopy types of products, namely that $$|P\times Q| \simeq |P| \times |Q|,$$ where $\times$ means product of topological spaces (see Theorem 5.1(a) in the paper of Walker, or again Section 5.1 of the notes of Wachs). This means that $$|B| \simeq |B_k \setminus \{\hat{0},\hat{1}\}| \times |B_{n-k}|,$$ and since $B_{n-k}$ is contractible, this time we actually can say $|B| \simeq |B_k \setminus \{\hat{0},\hat{1}\}| \simeq \mathbb{S}^{k-2}$.

Quillen, Daniel, Homotopy properties of the poset of nontrivial p-subgroups of a group, Adv. Math. 28, 101-128 (1978). ZBL0388.55007.

Wachs, Michelle L., Poset topology: tools and applications, Miller, Ezra (ed.) et al., Geometric combinatorics. Providence, RI: American Mathematical Society (AMS); Princeton, NJ: Institute for Advanced Studies (ISBN 978-0-8218-3736-8/hbk). IAS/Park City Mathematics Series 13, 497-615 (2007). ZBL1135.06001.

Walker, James W., Canonical homeomorphisms of posets, Eur. J. Comb. 9, No. 2, 97-107 (1988). ZBL0661.06006.

$\endgroup$
3
  • 1
    $\begingroup$ I'm not able to see the equality of posets - it seems to suggest that the subsets of interest have non-empty intersection with T, rather than just being non-empty? $\endgroup$ Jun 7 at 16:36
  • 2
    $\begingroup$ @TylerLawson: Thanks for your comment. I've corrected the answer (or at least attempted to!). $\endgroup$ Jun 7 at 17:08
  • $\begingroup$ @NeilStrickland: argh, of course you are right. I've corrected again. $\endgroup$ Jun 8 at 13:20

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.