Embed one Coxeter System into another

What is a good reference that explains all the braid relations and diagrams for Coxeter systems concisely?

In particular, how do I embed $H_3$ inside $D_6$, or $H_4$ inside $E_8$? Any hints?

• If I remember correctly these embeddings are discussed (for the first time?) in Lusztig, "Some examples of square integrable representations of a p-adic group". See discussion around p. 636 and lovely diagram... Aug 15, 2014 at 7:36
• @Geordie: You should "promote" this comment to an answer. I suspect this is exactly what Qiao wants. Aug 15, 2014 at 12:52
• @Geordie: Lusztig certainly gives an insightful embedding of $H_4$ into $E_8$ (though not the first such). Here is an online link to the article: ams.org/journals/tran/1983-277-02/S0002-9947-1983-0694380-4 Aug 15, 2014 at 14:03
• @DavidSpeyer: I think that manuscript of Stembridge only addresses embeddings of crystollagraphic root systems in simply-laced ones, in which case there is an easy way to see the embedding via "folding" which arises from the fixed-point set of an automorphism of the Coxeter graph. But the embeddings discussed by the OP (of $H_3$ into $D_6$ or $H_4$ into $E_8$) are different and more complicated than this. Aug 15, 2014 at 21:20
• @Geordie: The history looks murky to me, but I have been revisiting the past literature. For example, the 1988 article by Shcherbak (translated in Russian Math Surveys) has a more complete series of diagrams including Lusztig's. But too little history is given there. I never looked deeply into these things 25 years ago, lacking motivation. Bourbaki apparently goes only as far as Witt's construction of $H_3, H_4$ using quaternions. In any case, Lusztig's example is just one of the many fascinating things in that (misleadingly named) paper. Aug 15, 2014 at 22:59

It's not clear exactly what your first sentence is asking for (there are a variety of surveys and books). For instance, are you only interested in finite Coxeter groups?

The embeddings of the noncrystallographic Coxeter groups $H_3, H_4$ indicated in the follow-up question probably originate in older work of Coxeter and du Val (though I don't recall the exact source). These embeddings have been studied over the years in different styles, but a single comprehensive reference may not exist. One direction is indicated in a series of papers by Matthew Dyer, e.g., Embeddings of root systems. I. Root systems over commutative rings, J. Algebra 321 (2009), no. 11, 3226–3248.

Another direction of research connects the embeddings with mathematical physics, sometimes in rather a computational style, e.g., Mehmet Koca, Ramazan Koç, Muataz Al-Barwani, Noncrystallographic Coxeter group $H_4$ in $E_8$,
J. Phys. A 34 (2001), no. 50, 11201–11213.

ADDED: Older references for the embeddings were given in the notes to Section 2.13 in my 1990 book Reflection Groups and Coxeter Groups, though in the published version I wrote $D_3$ where $D_6$ is intended (this and other corrections have been maintained on my homepage).

NOTE ON HISTORY: Having looked further into the tangled literature (which reinforces my decision not to become a historian of contemporary mathematics), I would replace the "probably" in my second paragraph by "possibly" given the lack of documentation.

(1) The earliest relevant paper seems to be the short note (in an obscure Japanese journal): Jiro Sekiguchi, Tamaki Yano, A note on the Coxeter group of type $H_3$., Sci. Rep. Saitama Univ. Ser. A 9 (1979), no. 2, 33-44. This follows their note on $F_4$ and both are motivated especially by invariant theory aspects. At this remove at least one author recalls they also discussed $H_4$ at that time but didn't write down the details. (2) In the short remark 3.9(b) of his 1983 paper linked above, Lusztig applied some of the previous Hecke algebra and $W$-graph development to the embedding of $H_4$ into $E_8$. His proof comebines elementary and more sophisticated arguments. The starting point is an assignment of vertices in the $H_4$ diagram to nonconnected pairs of vertices in the $E_8$ diagram, pictured via a "folding" of the latter graph; this induces a homomorphism of one Coxeter group into the other. (3) A unified treatment of embeddings of non-crystallographic finite Coxeter groups is given more expansively in $\S2$ of a Russian paper, with English translation: O.P. Shcherbak, Wavefronts and reflection groups, Russian Math. Surveys 43:3 (1988), 149-194. See Theorem 2 and the folding'' diagrams there for $H_2$ (dihedral of order 10), $H_3, H_4$ as well as comments on embedding other non-crystallographic dihedral groups. His method is mostly different from Lusztig's, whose work he may or may not have been aware of (but doesn't cite); he does acknowledge the note by Sekiguchi-Yano. Shcherbak's work is in the mode of the Arnol'd-Brieskorn work on singularities and extends it to non-crystallographic cases.

While I referred to (1) and (3) in my 1990 book, I overlooked Lusztig's remark at the time even though his paper is in the reference list. (Some of Scherbak's earlier work on $H_4$ appeared only in a non-translated short note in the same Russian journal in 1984. It is this work which figures in later developments by B. Dubrovin and others on classification of 2D topological field theories, etc.)

• Thanks, I totally forgot to think of this Dyer reference! Aug 14, 2014 at 19:11

(This is more of a longer comment without a proper answer to the question.)

I assume you mean "An embedding $(W,S) \hookrightarrow (W',S')$ is an injective map from $S$ to the set $T' = \{ w s w^{-1} : w \in W', s \in S' \}$ of reflections in $W'$ generating a Coxeter system isomorphic to $(W,S)$".

• Is that what you mean?

If yes, this is the same as asking which reflection subgroups are there in $(W',S')$ that are isomorphic to $(W,S)$. By "Dyer - Reflection subgroups of Coxeter systems" (and in parallel by a paper of Dheodar), every reflection subgroup it itself a Coxeter group with canonical Coxeter generators depending on $S'$. This gives a (computationally feasible) way to systematically search for reflection subgroups in $(W',S')$ isomorphic to $(W,S)$. On the other hand, I am not aware of a more clever way to determine how $(W,S)$ embeds into $(W',S')$.

• I don't think that this is what he means. The embedding of e.g. H_3 into D_6 doubles lengths, and so doesn't send simple reflections to reflections. Aug 15, 2014 at 11:45
• Oh yes, you are right. Thanks for the clarification! Aug 15, 2014 at 12:03