Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Definitions: A partially ordered group or po-group is a po-set $(G,\leq)$, such that $G$ is a group and $\forall x,y,a,b\!\in\!G\!:x\!\leq\!y\Rightarrow axb\!\leq\!ayb$, i.e. a po-set that is a group in which left and right translations are isotone (= order preserving). A directed group or do-group is a po-group that is an (up and down) directed set. A lattice ordered group or lo-group is a po-group that is a lattice. A totally ordered group or to-group is a po-group that is a totally ordered set. The (positive) cone of $G$ is $G_{\geq1}\!=\!G_+\!:=\!\{g\!\in\!G;\, g\!\geq\!1\}$.

For a group $G$ and subsets $A,B\!\subseteq\!G$, let $\langle A\rangle$ denote the subgroup generated by $A$, and let $AB\!=\!\{ab; a\!\in\!A,b\!\in\!B\}$ and $A^{-1}\!=\!\{a^{-1}; a\!\in\!A\}$.

Theorem (order/cone correspondence for groups) [Steinberg: Lattice-ordered Rings and Modules p.34-35; Blyth: Lattices and Ordered Algebraic Structures p.144-145]: Let $G$ be a group and $P\!\subseteq\!G$ a subset. Consider the following conditions on $P$: $$\text{(i) } PP\!\subseteq\!P; \hspace{3mm} \text{(ii) } \forall g\!\in\!G\!: gPg^{-1}\!\subseteq\!P; \hspace{3mm} \text{(iii) } P\!\cap\!P^{-1}\!=\!\{1\}; \hspace{3mm} \text{(iv) } PP^{-1}\!=\!G; \hspace{3mm} \text{(v) } P\!\cup\!P^{-1}\!=\!G.$$

a) Define the relation $x\leq_P\,y\Leftrightarrow x^{-1}y\!\in\!P$ on $G$. Call $P$ a po-cone / do-cone / lo-cone / to-cone on $G$ when (i)-(iii) hold / (i)-(iii),(iv) hold / (i)-(iii),(iv) hold and $(P,\leq_P\!)$ is a lattice / (i)-(iii),(v) hold. Then the maps $\begin{smallmatrix} \scriptstyle\leq &\!\scriptstyle \mapsto \! &\scriptstyle G_{\geq1}\\ \scriptstyle\leq_P &\!\scriptstyle \leftarrow\! &\scriptstyle P \end{smallmatrix}$ are mutually inverse bijections on the following pairs of sets: $$\begin{array}{r @{\hspace{1mm}} c @{\hspace{1mm}} l} \{\leq\subseteq\!G^2; \text{ the pair }(G,\leq)\text{ is a po-group}\} & \rightleftarrows & \{P\!\subseteq\!G; \text{ the subset }P\text{ is a po-cone on }G\}\\ \bigcup\!\mathbf{|}\hspace{2.6cm} & & \hspace{0.0cm}\bigcup\!\mathbf{|}\\ \{\leq\subseteq\!G^2; \text{ the pair }(G,\leq)\text{ is a do-group}\} & \rightleftarrows & \{P\!\subseteq\!G; \text{ the subset }P\text{ is a do-cone on }G\}\\ \bigcup\!\mathbf{|}\hspace{2.6cm} & & \hspace{0.0cm}\bigcup\!\mathbf{|}\\ \{\leq\subseteq\!G^2; \text{ the pair }(G,\leq)\text{ is a lo-group}\} & \rightleftarrows & \{P\!\subseteq\!G; \text{ the subset }P\text{ is a lo-cone on }G\}\\ \bigcup\!\mathbf{|}\hspace{2.6cm} & & \hspace{0.0cm}\bigcup\!\mathbf{|}\\ \{\leq\subseteq\!G^2; \text{ the pair }(G,\leq)\text{ is a to-group}\} & \rightleftarrows & \{P\!\subseteq\!G; \text{ the subset }P\text{ is a to-cone on }G\}. \end{array}$$

b) Let $(G,\leq)$ be a po-group with $P\!=\!G_{\geq1}$. T.f.a.e.: $(G,\leq)$ is a do-group; $(G,\leq)$ is directed up; $(G,\leq)$ is directed down; $\langle P\rangle \!=\!G$. T.f.a.e.: $(G,\leq)$ is an lo-group; $\forall g\!\in\!G\,\exists g\!\wedge\!1 \in G$; $\forall g\!\in\!G\,\exists g\!\vee\!1 \in G$.

Exercise: Let $G$ be a group with $H\!\unlhd\!G$, and suppose that $H$ and $G/H$ are po-groups. Prove that $$P:=H_+\cup\{g\!\in\!G;\, H\!\neq\!gH\!\in\!(G/H)_+\}$$ is a po-cone on $G$ iff $\forall g\!\in\!G\!: gH_+g^{-1}\!\subseteq\!H_+$. From now on, assume that $P$ is a po-cone on $G$, and $H,G/H$ have po-cones $H_+,(G/H)_+$. Prove: a) $H$ is convex in $G$ [and $P$ induces $H_+,(G/H)_+$]. b) $H\!<\!P\!\setminus\!H$. c) $G$ is a to-group iff $H$ and $G/H$ are to-groups. d) If $H\!\neq\!\{1\}$, then $G$ is an lo-group iff ($H$ is an lo-[sub]group and $G/H$ is a to-group). e) If $G$ is an lo-group, then the inclusion $H\!\rightarrow\!G$ is a complete lattice morphism.

Comment: I added the [...] parts (they were not originally part of the exercise).

Partial Solution: $(\Rightarrow)$: If $P$ is a po-cone, then $\forall g\!\in\!G\!: gPg^{-1}\!\subseteq\!P$, so for $h\!\in\!H_+$, we have $ghg^{-1}\!\in\!P$. From $H\!\unlhd\!G$ we get $ghg^{-1}\!\in\!H$, so we cannot have $H\!\neq\!ghg^{-1}H\!\in\!(G/H)_+$, hence $ghg^{-1}\!\in\!H_+$. $(\Leftarrow)$: (i): If $h,h'\!\in\!H_+$, then $hh'\!\in\!H_+$. If $h\!\in\!H_+$ and $H\!\neq\!gH\!\in\!(G/H)_+$, then $H\!\neq\!hgH\!\in\!(G/H)_+$ (if $hg\!\in\!H$, then $h^{-1}hg\!=\!g\!\in\!H$, $\rightarrow\leftarrow$; since $gH\!\in\!(G/H)_+$ and $hH\!=\!1H\!\in\!(G/H)_+$, we have $hHgH\!=\!hgH\!\in\!(G/H)_+$), and similarly, $H\!\neq\!ghH\!\in\!(G/H)_+$. If $H\!\neq\!gH\!\in\!(G/H)_+$ and $H\!\neq\!g'H\!\in\!(G/H)_+$, then $gg'H\!=\!gHg'H\!\in\!(G/H)_+$, and if $gg'\!\in\!H$, then ???. (ii): By assumption, $gH_+g^{-1}\!\subseteq\!H_+$. Furthermore, if $H\!\neq\!g'H\!\in\!(G/H)_+$, then $H\!\neq\!gg'g^{-1}H$ (since $H\!\unlhd\!G$) and $gg'g^{-1}H\!=\!gH(g'H)g^{-1}H\!\in\!(G/H)_+$. Thus $gPg^{-1}\!\subseteq\!P$. (iii): Assume that $g,g^{-1}\!\in\!P$. If $g,g^{-1}\!\in\!H_+$, then $g\!=\!1$. If $g\!\in\!H_+$ and $H\!\neq\!g^{-1}H\!\in\!(G/H)_+$, then $g^{-1}\!\in\!H$, $\rightarrow\leftarrow$. If $H\!\neq\!gH\!\in\!(G/H)_+$ and $H\!\neq\!g^{-1}H\!\in\!(G/H)_+$, then $gH\!\in\!(G/H)_+\!\cap\!(G/H)_+^{-1}$, so $gH\!=\!H$, $\rightarrow\leftarrow$.

a) If $h\!\leq\!g\!\leq\!h'$ and $h,h'\!\in\!H$, then $h^{-1}g, g^{-1}h'\!\in\!P$. If $h^{-1}g\!\in\!H_+$ or $g^{-1}h'\!\in\!H_+$, then $g\!\in\!H$. But if $h^{-1}gH\!\in\!(G/H)_+$ and $g^{-1}h'H\!\in\!(G/H)_+$, then $gH\!=\!hHh^{-1}gH\!\in\!(G/H)_+$ and $g^{-1}H\!=\!g^{-1}h'Hh'^{-1}H\!\in\!(G/H)_+$, i.e. $gH\!\in\!(G/H)_+\!\cap\!(G/H)_+^{-1}$, which implies $gH\!=\!H$, i.e. $g\!\in\!H$.

If $h,h'\!\in\!H$, then $h\!\leq_{H_+}\!h' \Leftrightarrow h^{-1}h'\!\in\!H_+ \Leftrightarrow h^{-1}h'\!\in\!P \Leftrightarrow h\!\leq_{P}\!h'$. If $gH,g'H\!\in\!G/H$, then $gH\!\leq_{\!(G/H)_+}\!\!g'H \Leftrightarrow g^{-1}g'H\!\in\!(G/H)_+ \Leftrightarrow \big((\exists h\!\in\!H\!:g^{-1}g'h\!\in\!H_+)\text{ or }(H\!\neq\!g^{-1}g'H\!\in\!(G/H)_+)\big) \Leftrightarrow \exists h\!\in\!H\!:\big(g^{-1}g'h\!\in\!H_+\text{ or }H\!\neq\!g^{-1}g'hH\!\in\!(G/H)_+)\big) \Leftrightarrow \exists h\!\in\!H\!: g\!\leq_P\!g'h$. In the second equivalence, $\Leftarrow$ is clear, but for $\Rightarrow$, either $g^{-1}g'H\!\neq\!H$ or $g^{-1}g'\!=\!h^{-1}$ for some $h\!\in\!H$, and then $g^{-1}g'h\!=\!1\!\in\!H_+$.

b) Let $h\!\in\!H$ and $H\!\neq\!gH\!\in\!(G/H)_+$. To prove $h\!<\!g$, we must show that $h^{-1}g\!\in\!P$. We have $h^{-1}gH\!=\!gH\!\neq\!H$ and $h^{-1}gH\!=\!gH\!\in\!(G/H)_+$.

c) $(\Rightarrow)$: Assume $P\!\cup\!P^{-1}\!=\!G$. For any $h\!\in\!H$, either $h\!\in\!P$ or $h^{-1}\!\in\!P$, but since $H\!=\!hH\!=\!h^{-1}H$, either $h\!\in\!H_+$ or $h^{-1}\!\in\!H_+$, i.e. $h\!\in\!H_+\!\cup\!H_+^{-1}$. For any $gH\!\in\!G/H$, either $g\!\in\!P$ or $g^{-1}\!\in\!P$. If $g\!\in\!H_+$ or $g^{-1}\!\in\!H_+$, then $gH\!=\!H\!\in\!(G/H)_+$ or $g^{-1}H\!=\!H\!\in\!(G/H)_+$. Otherwise $gH\!\in\!(G/H)_+$ or $g^{-1}H\!\in\!(G/H)_+$. $(\Leftarrow)$: Assume $H_+\!\cup\!H_+^{-1}\!=\!H$ and $(G/H)_+\!\cup\!(G/H)_+^{-1}\!=\!G/H$. For any $g\!\in\!G$, either $g\!\in\!H$ (then $g\!\in\!H_+\!\cup\!H_+^{-1}\!\subseteq\!P\!\cup\!P^{-1}$) or $gH\!\neq\!H$ (then $g^{-1}H\!\neq\!H$, and either $gH\!\in\!(G/H)_+$ or $g^{-1}H\!\in\!(G/H)_+$), so either $g\!\in\!P$ or $g^{-1}\!\in\!P$.

d) $(\Rightarrow)$: Let $G$ be an lo-group. If $h\!\in\!H$, then $\exists h\!\vee_P\!1\!=:\!g\!\in\!P$, and ???, so $g\!\in\!H$, hence $g\!=\!h\!\vee_{H_+}\!1$. $(\Leftarrow)$: Let $H$ be an lo-group and $G/H$ a to-group. It suffices to prove that $\forall g\!\in\!G\!\setminus\!H\!: \exists g\!\vee_P\!1$. We have either $gH\!\in\!(G/H)_+$ or $g^{-1}H\!\in\!(G/H)_+$, i.e. $hg\!\in\!P$ or $g^{-1}h\!\in\!P$ for some $h\!\in\!H$. ???

e) We must prove that $\exists\inf_H\{h_i; i\!\in\!I\}\!=:\!h_0\Rightarrow \inf_G\{h_i; i\!\in\!I\}\!=\!h_0$ and $\exists\sup_H\{h_i; i\!\in\!I\}\!=:\!h_1\Rightarrow \sup_G\{h_i; i\!\in\!I\}\!=\!h_1$. If $\exists g\!\in\!G\!\setminus\!H\!: g\!\leq\!\{h_i; i\!\in\!I\}$, then ???, so $g\!\leq\!h_0$.

Question: How can I finish the '???' parts? I'm stuck at $PP\!\subseteq\!P$, d), e).

share|cite|improve this question
up vote 3 down vote accepted

For the first gap: if $gg'\in H$, then $g'\in g^{-1}H$, so $g'H=(gH)^{-1}$, $g'H\in(G/H)_+\cap\big((G/H)_+\big)^{-1}$, and therefore $g'H=H$, $\to\leftarrow$.

(d) Let $\pi:G\to G/H:g\mapsto gH$. For $g,g'\in G$, $g\le_G g'$ iff $g^{-1}g'\in P$; that that this is the case iff either

  • $\pi(g)<_{G/H}\pi(g')$, or
  • $\pi(g)=\pi(g')$ and $1\le_H g^{-1}g'$.

Note in particular that if $g,g'\in H$, then $g\le_G g'$ iff $g\le_H g'$.

Suppose that $G/H$ is a to-group, and let $g\in G$. Then exactly one of the following holds: $\pi(g)=\pi(1)$, $\pi(g)<_{G/H}\pi(1)$, or $\pi(g)>_{G/H}\pi(1)$. If $\pi(g)>_{G/H}\pi(1)$, then $g>_G 1$, and $g\lor_G 1=g$. If $\pi(g)<_{G/H}\pi(1)$, then $g<_G 1$, and $g\lor_G 1=1$. And if $\pi(g)=\pi(1)$, then $g\in H$.

We’ll show now that $G$ is a lo-group iff $H$ is a lo-group. Assume first that $H$ is a lo-group, and let $g\in G$. We just saw that if $g\notin H$, then $g\lor_G 1$ exists and is either $g$ or $1$, so assume now that $g\in H$, and let $u=g\lor_H 1$. Then if $g'\in G$ is any upper bound for $g$ and $1$ in $G$, either $\pi(g')>_{G/H}\pi(1)$, in which case $u<_G g'$, or $g'\in H$, in which case $u\le_G g'$, so $u=g\lor_G 1$.

Now suppose that $H$ is not a lo-group, and let $h\in H$ be such that $h$ and $1$ have no supremum in $H$; I claim that $h$ and $1$ can have no supremum in $G$, either. This is clear if there is a $u\in H$ such that $h,1\le_H u$: if $g$ is any upper bound of $h$ and $1$ in $G\setminus H$, $u<_G g$, so $g\ne h\lor_G 1$. Thus, we may assume that $\{h,1\}$ has no upper bound in $H$. Suppose that $g$ is an upper bound for $\{h,1\}$ in $G$; clearly $\pi(g)>_{G/H}\pi(1)$. If $H_+\ne\{1\}$, choose $p\in H_+\setminus\{1\}$; then $h,1<_G gp^{-1}<_G g$, so $g\ne h\lor_G 1$. The only remaining possibility is that $H_+=\{1\}$. In that case $\pi(gh)=\pi(g)$, so $gh$ is an upper bound for $\{h,1\}$. However, it’s clear that $h\ne 1$, so $g^{-1}(gh)=h\notin P$, $g\not\le_G gh$, and $g\ne h\lor_G 1$. Thus, $h$ and $1$ have no supremum in $G$, and $G$ is not a lo-group.

To complete the proof of (d) we must show that if $G/H$ is not a to-group, then $G$ is not a lo-group. Suppose, then, that $G$ is not a to-group; then there is a $g\in G$ such that $\pi(g)\not\le_{G/H}\pi(1)\not\le_{G/H}\pi(g)$. Suppose that $u$ an upper bound for $\{g,1\}$ in $G$; then $\pi(u)>_{G/H}\pi(g),\pi(1)$. As before, if $H$ has a strictly positive element $p$, then $up^{-1}$ is an upper bound for $\{g,1\}$ strictly smaller than $u$. If $H_+=\{1\}$, let $h\in H\setminus\{1\}$, and observe that $uh$ is an upper bound for $\{g,1\}$ such that $u\not\le_G uh$. In either case $u\ne g\lor_G 1$, and $G$ is not a lo-group.

(e) Recall from (d) that if $g,g'\in H$, then $g\le_G g'$ iff $g\le_H g'$.

share|cite|improve this answer
Thank you for your reply. I do not understand your 5th (longest) paragraph. Why do we even need that "$h,1$ has no upper bound in $H$"? However, using your ideas, I found an analogous solution to (d) and (e) which I post below, since it's too long for a comment. Thank you again. – Leon Jul 27 '12 at 0:23

This is just a solution to (d) and (e) which I understand easier, but due to Brian M. Scott. It's too long for a comment.

d) Let $\pi:G\!\rightarrow\!G/H$, $g\!\mapsto\!gH$. For $g,g'\!\in\!G$, we have $g\!\leq_G\!g'$ iff $g^{-1}g'\!\in\!P$, iff either $\pi(g)\!<_{G/H}\!\pi(g')$ or ($\pi(g)\!=\!\pi(g')$ and $1\!\leq_H\!g^{-1}g'$).

$(\Leftarrow)$: Suppose that $G/H$ is a to-group. For $g\!\in\!G$, exactly one of the following holds: $\pi(g)\!=\!\pi(1)$, $\pi(g)\!<_{G/H}\!\pi(1)$, $\pi(g)\!>_{G/H}\!\pi(1)$. In the first case, $g\!\in\!H$. In the second, $g\!<_G\!1$, hence $g\!\vee_G\!1\!=\!1$. In the third, $g\!>_G\!1$, hence $g\!\vee_G\!1\!=\!g$. Thus if $g\!\notin\!H$, then $\exists\,g\!\vee_G\!1$. But if $g\!\in\!H$, let $s\!:=\!g\!\vee_H\!1$, so $\pi(s)\!=\!\pi(1)$. For any $g'\!\in\!G$ with $g,1\!\leq\!g'$, either $\pi(1)\!<_{G/H}\!\pi(g')$ (then $s\!<_G\!g'$), or $g'\!\in\!H$ (then $s\!\leq\!g'$). Thus $s\!=\!g\!\vee_G\!1$. We have proved that $g\!\vee_G\!1$ always exists, so by the correspondence theorem, $G$ is an lo-group.

$(\Rightarrow)$: Suppose that $G$ is an lo-group, but $H$ is not, i.e. $\exists h\!\in\!H\!:\nexists\,h\!\vee_H\!1$. Take $s\!:=\!h\!\vee_G\!1$. We see that $s\!\in\!G\!\setminus\!H$. By a), we have $H\!<\!s$, so $H\!<\!Hs$. If $H_+\!\neq\!\{1\}$, choose $p\!\in\!H_+\!\setminus\!\{1\}$, and then $h,1\!<\!p^{-1}\!s\!<\!s$, a contradiction with the minimality of $s$. But if $H_+\!=\!\{1\}$, then $h,1\!\leq\!hs$ gives (by minimality of $s$) $s\!\leq\!hs$, i.e. $h\!\in\!H_+$, so $h\!=\!1$, a contradiction.

Now suppose that $G$ is an lo-group, but $G/H$ is not a to-group, i.e. $\exists g\!: \pi(g)\!\nleq_{G/H}\!\pi(1)\!\nleq_{G/H}\!\pi(g)$. Take $s\!:=\!g\!\vee_G\!1$. Then $\forall h\!\in\!H\!:\pi(g),\pi(1)<_{G/H}\pi(s)=\pi(hs)$. If $H_+\!\neq\!\{1\}$, choose $p\!\in\!H_+\!\setminus\!\{1\}$, and then $g,1 < p^{-1}u < u$, a contradiction. If $H_+\!=\!\{1\}$, let $h\!\in\!H\!\setminus\!\{1\}$, and then $g,1\!\leq\!hs$, so $s\!\leq\!hs$, i.e. $h\!\in\!H_+$, a contradiction.

e) We must prove that $\exists\sup_H\{h_i;i\!\in\!I\}$ implies $\sup_H\{h_i;i\!\in\!I\}\!=\!\sup_G\{h_i;i\!\in\!I\}$, and that $\exists\inf_H\{h_i;i\!\in\!I\}$ implies $\inf_H\{h_i;i\!\in\!I\}\!=\!\inf_G\{h_i;i\!\in\!I\}$. By d), $H$ is an lo-group and $G/H$ is a to-group. We use the criterium from d). Suppose that $s\!:=\!\sup_H\{h_i\}\!\in\!H$ exists. For any $g\!\in\!G$ with $\{h_i\}\!\leq\!g$, either $\pi(s)\!=\!\pi(h_i)\!=\!\pi(g)$ (then $g\!\in\!H$, so $s\!\leq\!g$), or $\pi(s)\!=\!\pi(h_i)\!<\!\pi(g)$ (then $s\!<\!g$). Thus $s\!=\!\sup_G\{h_i\}$. The equality for infimums is proved analogously. $\blacksquare$

share|cite|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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