There is a well-known classification result which gives a complete list of all root lattices which can embed into the lattice $E_8$. Moreover, each of these root lattices admits a unique embedding modulo the action of the Weyl group $W(E_8)$ except for five exceptional types: $$A_7, \; A_3^2, \; A_5 \oplus A_1,\; A_3 \oplus A_1^2,\; A_1^4$$

I would be very interested to know whether a similar result is known for root lattice embeddings into the even unimodular lattice $E_{10}=E_8 \oplus U$ (where $U$ is the hyperbolic plane), and in particular how to count the number of inequivalent embeddings for a given root lattice.

EDIT: Is there any hope to solving this by playing around with the Dynkin diagram for $E_{10}$? My main concern is that $E_{10}$ does not arise from a semisimple Lie algebra, however Friedman - On the geometry of anticanonical pairs (p.83, Example 9.21) seems to imply that a classification should be possible, but unfortunately doesn't give any details.

  • $\begingroup$ Are the "exceptional types" for $E_8$ made unique by requiring that they embed primitively? For example, $A_7$ can embed either as an index-$2$ sublattice of the unique $E_7$, or as the orthogonal complement of a primitive vector of norm $8$; likewise $A_3^2$, $A_5 \oplus A_1$, $A_3 \oplus A_1^2$, and $A_1^4$ can be either primitive or contained with index $2$ in $D_6, E_6, D_5, D_4$ respectively. (For the $D_6$ and $D_5$ sublattices this can be seen more easily by writing $A_3^2 = D_3^2$ and $A_3 \oplus A_1^2 = D_3 \oplus D_2$.) $\endgroup$ May 3, 2021 at 14:18
  • 1
    $\begingroup$ Yes, for each of the exceptional types there exist exactly one primitive and one imprimitive embedding. $\endgroup$ May 3, 2021 at 14:34

1 Answer 1


The results I am familiar with in this direction are in the related area of regular subalgebras, or $\pi$-systems, as in Carbone et. al. here. There, they give a good amount of information about the possible subalgebras in Kac--Moody algebras (and the overextended cases like $E_{10}$) coming from an embedding of compatible root spaces; they also look at possible Weyl group orbits of these subsystems and the corresponding embeddings. The references within include a lot of what I have seen in this area.

A paper not in the references of Carbone et. al. by Felikson and Tumarkin, here, looks at the specific case of hyperbolic regular subalgebras of hyperbolic Kac--Moody algebras, classifying them by subgroups of the Weyl group. An interesting question posed in their paper (and might remain open, I am not sure) is whether, for an arbitrary symmetrizable Kac--Moody algebra, any reflection subgroup of W should correspond to a regular subalgebra embedding.

In terms of determining embeddings from the Dynkin diagram, I would be surprised if one could make such an algorithm. Viswanath showed here that any hyperbolic Kac--Moody algebra with a symmetric Cartan matrix can be embedded as a regular subalgebra into $E_{10}$ by explicitly constructing the $\pi$-system of roots; since these would include Dynkin diagrams with multiple lacings, I wouldn't expect a nice way to read this information off from the diagram for $E_{10}$.


Your Answer

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

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