WorksMathematics
Minimal norm of regular vectors in the root lattice of a crystallographic root system
Abstract
For a crystallographic root system with Weyl group and root lattice , we determine the minimum norm of a vector of having trivial stabilizer in , and we prove that the vectors attaining it form a single orbit of . Although this is a minimization of a quadratic function over the lattice points of a cone, the minimizer in the dominant chamber is always unique: it is , where is the Weyl vector and is or the minuscule weight in the class of modulo . In particular, the optimal vector is the Weyl vector exactly when the Weyl vector lies in . We give explicit formulas for the families , , and , with elementary proofs, and the values for the exceptional types , , , and . Finally, we explain the coincidences between the minimal norms by the folding of Dynkin diagrams.
Introduction
Let be a crystallographic root system in a Euclidean space with Weyl group , and let be its root lattice. We refer to Humphreys [1 Chapters 1 and 2] for the definitions and basic properties of root systems, Weyl groups and root lattices; crystallographic root systems, root lattices and weight lattices are treated in [1 §2.9]. For the simply laced types, with roots of norm 2, the root lattices are , , , and . For the types and we use the standard realizations in :
whose root lattices are and respectively. In both cases the Weyl group is the group of all signed permutations of the coordinates, which for is larger than . For we take, in the plane of , the short roots and the long roots with ; the root lattice is and is the dihedral group of order 12. For we take, in , the short roots and the long roots and ; the root lattice is and has order 1152. The root lattice alone therefore does not determine the problem considered below: and share the lattice , and and share the lattice , but with different Weyl groups.
The reflection hyperplanes , for , partition into open Weyl chambers. Since the stabilizer of a vector is generated by the reflections it contains [1 Theorem 1.12], a lattice vector has trivial stabilizer in if and only if lies in the interior of some Weyl chamber, equivalently, for all roots . Such vectors are called regular.
We consider the following optimization problem.
Definition 1. For a root system , define
By symmetry, we may restrict the search to a single closed Weyl chamber. Fixing a system of simple roots , the dominant chamber is for all , and we seek a lattice point in its interior with minimum norm. In the realizations above the inner product takes integral values on , so that this is
where is the Gram matrix of the simple roots and is expressed in the basis of simple roots.
This is an integer quadratic programming problem that can be solved computationally, and a priori nothing guarantees that its optimal solution is unique. Our main result, Theorem 4, shows that it always is: the optimal vector in the dominant chamber is , where is the Weyl vector and is the minimal dominant weight in the class of modulo , which is 0 or a minuscule weight. For the families , , and , we also give elementary proofs of closed-form solutions (Theorems 1 and 2). For the exceptional types, the solutions were first obtained computationally and are reported in § 1.1. In § 3 we compare the optimal vectors with the Weyl vector and prove Theorem 4, and in § 4 we explain the equalities between the minimal norms of different root systems by folding. The results are summarized in Table 1.
Table 1. The minimal norm for each irreducible root system , the weight such that the optimal vector in the dominant chamber is (Theorem 4), and where the optimal vector is described. The fundamental weights are numbered as in Bourbaki; means that the optimal vector is the Weyl vector .
Remark 1. The problem reduces to irreducible root systems. Let be the decomposition of into irreducible components, which are pairwise orthogonal. Then with , and , each factor acting on the span of its own component and trivially on the others. For with , we have and . Hence
and the optimal vectors of are the sums of optimal vectors of the components. In particular, the optimal solution is again unique up to the action of , and the optimal vector in the dominant chamber is , where and are the sums of the corresponding vectors of the components. In the rest of the paper we assume that is irreducible.
Exceptional root systems
For , and , we solve the integer quadratic programming problem (1) computationally using the SCIP optimization suite [6]. For and , we enumerate all lattice vectors of small norm. The optimal vectors are given in Figure 1, together with the one for , which is the smallest case of Theorem 2.
Figure 1. Vectors realizing in the dominant chamber for the types , , , , , . Each node of the Dynkin diagram carries the coefficient of the vector on the corresponding simple root; for and the arrow points from the long to the short simple roots. The inner product of the vector with each simple root is , except for the node of with coefficient , where it is .
The result is confirmed by Moorhouse [4], who reports: “The 310th shell is the smallest containing a regular -orbit; i.e. the shortest lattice vectors having trivial stabilizer in have norm 620.”
In the realizations of the introduction, the optimal vector in the dominant chamber is for and for , for the simple roots , of and , , , of , in the order of Figure 1.
For , , , and these vectors are the Weyl vectors, which are optimal and unique by Remark 4, while for the optimal vector is not the Weyl vector. In all cases, the optimal vector is unique and is given by Theorem 4.
Remark 2. We have and , and in Figure 1 the coefficients for and are those of the nodes of and that are identified by the folding of their Dynkin diagrams. This is explained in § 4.
The case
The root lattice can be represented as the set of integer vectors with . The Weyl group acts by permuting coordinates. A vector has trivial stabilizer if and only if for all . The norm is .
Thus reduces to minimizing over distinct integers summing to zero. Setting , we prove:
Theorem 1. The problem has a unique optimal solution (up to the action of ). In the ambient representation, the optimal coordinates are:
For even ( odd, ): the set .
For odd ( even, ): the set .
In both cases, .
Expressed in the basis of simple roots , the optimal vector in the dominant Weyl chamber has coefficients:
For even: for .
For odd: where , for .
The proof relies on two exchange lemmas that constrain the structure of optimal solutions.
Lemma 1 (No large gaps). Let be a set of distinct integers with that minimizes . Then for all .
Proof. Suppose for some . Replace by and by . Both new values lie strictly between and , so they are distinct from each other and from all other elements of . The sum is preserved, and the change in is
contradicting optimality. ▢
Lemma 2 (At most one gap). Under the same hypotheses as Lemma 1, there is at most one index with .
Proof. Suppose and for some . The integers and are absent from . Replace by and by . The two new values are distinct from each other (since implies ) and from all remaining elements of . The sum is preserved since . The change in is
contradicting optimality. □
Proof of Theorem 1. Let be an optimal set. By Lemmas 1 and 2, the set is either consecutive integers or consecutive integers with one element removed.
If has no gap, then and , giving . This is an integer if and only if is odd, i.e., is even. In that case and .
If has exactly one gap, then for some integer . The sum condition gives
The constraint forces .
If (i.e., odd): The only integer in that range is , giving . Thus .
If (i.e., even): Then . For , we get , so the gap is at the top endpoint. For , we get , so the gap is at the bottom endpoint. In both cases, removing an endpoint reduces to the case without gap: .
In all cases, the optimal norm is .
The simple root coordinates (where the ambient coordinates are arranged in decreasing order for the dominant chamber) yield the stated formulas. □
The , and cases
The three families are realized in with standard basis :
(): roots , ; root lattice ; Weyl group the signed permutations.
(): roots , ; root lattice ; Weyl group the signed permutations.
(): roots ; root lattice ; Weyl group the signed permutations with an even number of sign changes.
We take the simple roots for together with for , for , and for . The interior of the dominant chamber is for and , and for . Writing and , the coefficients on the simple roots are
and, for , for , and .
Lemma 3. A vector in the root lattice has trivial stabilizer (1) for and , if and only if are nonzero and pairwise distinct; (2) for , if and only if are pairwise distinct.
Proof. The stabilizer is trivial if and only if for every root . We have , which is nonzero for both signs if and only if , and are nonzero if and only if . □
Lemma 4. Let and let be integers. Then
with equality if and only if for all . If , then
with equality if and only if for and .
Proof. Since the are strictly increasing integers, , which gives the first inequality and its equality case. If , then for all , so for all . Hence if we have , and the second inequality follows from for . □
Write .
Theorem 2. For each of , , , the problem Triv has a unique optimal solution up to the action of the Weyl group. The optimal vector in the dominant chamber, its coefficients on the simple roots, and the minimum norm are the following.
(1) , all : , for , and .
(2) , : , for , , and .
(3) , : , for , , and .
(4) , : , for , , and .
(5) , : , for , , and .
Proof. Let have trivial stabilizer and let be the absolute values in increasing order; they are pairwise distinct by Lemma 3. The norm of is , and since changing the sign of changes by the even number , we have .
For , the are positive, and Lemma 4 with shows that , with equality if and only if . All such vectors are signed permutations of , hence form a single orbit of . For , the are positive and must be even. If , then is even, and we conclude as for . If , then is odd, so and the second part of Lemma 4 gives , with equality if and only if . This set has even sum , so all the corresponding vectors lie in , and they form a single orbit of .
For , at most one is zero, and we apply Lemma 4 with . The sum is even if and only if . In that case the optimum is , realized exactly when . Otherwise the second part of Lemma 4 gives the bound , realized exactly when , whose sum is even. In both cases one of the is zero, so every signed permutation of can be obtained with an even number of sign changes: the optimal vectors form a single orbit of .
The vectors listed are the representatives of these orbits in the dominant chamber, and the coefficients follow from (2) and the partial sums and .
Remark 3. The root systems and are the same up to a rotation and a scaling by , in agreement with . Also, for all , the optimal vector of being the optimal vector of with a coordinate appended, and for all . These equalities are explained in § 4.
Comparison with the Weyl vector
The Weyl vector of is
It lies in the interior of the dominant chamber and is characterized by for every simple root , since the simple reflection permutes the positive roots other than [1 Proposition 1.4]. We compare it with the optimal vector in the dominant chamber given by Theorems 1 and 2 and Figure 1.
Theorem 3. Let be one of (), (), (), (), , , , , . Then if and only if , and otherwise is not a multiple of . We have exactly for with even, with , with , , , and .
Proof. We compare the two vectors case by case.
: . For even, . For odd, , and is not a multiple of since its consecutive differences are not all equal.
: , and is not a multiple of .
: lies in if and only if is even, that is , and then . Otherwise is not a multiple of .
: lies in if and only if is even, that is , and then . Otherwise is not a multiple of $\rho.
E, E, F, G: by Figure 1, for every simple root, so , and .
E: by Figure 1, is 1 for all simple roots but one, where it is 2, so is not a multiple of . The coefficients of on the simple roots are 17, 33, 48, , 26, along the long branch and on the short one, so E. □
Remark 4. If , then is optimal and unique, without computation. For , with , so if is in the open dominant chamber, then is dominant. Since is a sum of positive roots, and , with equality if and only if . This proves the results for E, E, F, G.
The argument of Remark 4 extends to the case , and gives a uniform description of the optimal vector, together with its uniqueness. Let be the weight lattice [1 §2.9], with fundamental weights , and let be the set of dominant weights. We write if is a nonnegative integral combination of simple roots. A dominant weight is minuscule if it is nonzero and for all . By Stembridge [5, Corollary 1.4, Proposition 1.12 and Corollary 1.13], each coset of in contains a unique dominant weight with for every dominant weight , and this weight is 0 for the coset and the unique minuscule weight of the coset otherwise.
Theorem 4. Let be the minimal dominant weight of the coset . Then is the unique vector of of minimum norm in the open dominant chamber, and
In particular, the optimal solution is unique up to the action of .
Proof. Let be in the open dominant chamber. Since , the integers are at least 1, and since , the vector is a dominant weight. As is the sum of the positive roots, . Hence with a nonnegative integral combination of simple roots,
Since is in the open dominant chamber, for . So , with equality if and only if , and is in and in the open dominant chamber. □
Remark 4 is the case , that is . In the cases where , the weight and the optimal vector of Theorems 1 and 2 are recovered as follows:
A: and ;
B: and ;
C, : and ;
D, : and ;
E: is the minuscule weight , dual to the simple root at the end of the long branch, which is the node with coefficient 15 in Figure 1. With , and , we get .
In particular, the results for the exceptional types do not depend on the computations of Section 2, which serve as a confirmation.
Folding
The equalities between minimal norms observed in § 1.1 and § 2, namely
come from the automorphisms of the Dynkin diagrams of the simply laced root systems. These diagram automorphisms and the corresponding folding of Dynkin diagrams are at the basis of the construction of the twisted affine Lie algebras [2 §8.3].
Let be a simply laced root system with simple roots and let be a nontrivial automorphism of its Dynkin diagram. Since the Gram matrix of is determined by the Dynkin diagram, extends to an isometry of , still denoted , which permutes . It therefore preserves , the root lattice and the dominant chamber . Let be the subspace of vectors fixed by and . For a -orbit , set . Since permutes the basis of , the vectors form a basis of . Two roots in the same orbit are orthogonal, except for the middle orbit of , where they are adjacent; so , except for that orbit, where it is 2. A direct computation of the Gram matrix of the shows that they form a system of simple roots of a root system in , whose type and root lattice are given in the following table.
| order of | ||||
| , | 2 | 2, 4 | ||
| , | 2 | 2, 4 | ||
| , | 2 | 2, 4 | ||
| 3 | 2, 6 | |||
| 2 | 2, 4 |
Table 2.
Since a root system is determined up to isometry by the Gram matrix of its simple roots, is isometric to the realization of the introduction, except for , which appears scaled by , so that . For and , the subalgebra of the simple Lie algebra of type fixed by the automorphism induced by has type and respectively [2 Proposition 8.3 e)], the dual of . A root system and its dual have the same reflection hyperplanes, hence the same regular vectors; it is the lattice that selects among the two.
Proposition 1. Let and be as above. Then , with equality if and only if some vector of in the open dominant chamber of norm is fixed by . If the optimal solution for is unique up to the action of , then equality holds, the optimal solution for is unique up to the action of , and the optimal vector in the dominant chamber is the same for and . Its coefficient on is the common coefficient of on the roots of .
Proof. For and , we have , so is constant on and . Hence the open dominant chamber of , defined by for all orbits , is , where is the open dominant chamber of . Since can be computed in the open dominant chamber,
with equality if and only if a minimizer of the right-hand side lies in . Every -orbit meets in at most one point, so if the optimal solution for is unique up to , there is a unique minimizer . Since preserves , and the norm, is also a minimizer, so and equality holds. Any minimizer for in is then a minimizer for , hence equal to . Finally, writing , the relation shows that is constant on each orbit , with value say, and .
By Theorem 4, or by Theorems 1 and 2 and Remark 4, the optimal solutions for , and are unique, so Proposition 1 gives the equalities above. It also gives a second proof of the values of for , , and , and of the uniqueness of their optimal solutions. The last assertion of Proposition 1 explains the coefficients in Figure 1, and the fact that the coefficients of the optimal vector of are those of and for in Theorem 1. For , the space is the hyperplane , which explains why the optimal vector of is that of with a coordinate 0 appended. Note that admits diagram automorphisms of order 2 and 3, giving , while and have no nontrivial diagram automorphism.
Tool and computational resource disclosure
Following the Leiden Declaration on Artificial Intelligence and Mathematics [3], we disclose the automated tools used in this work. The optimal vectors for the exceptional types were computed with programs written in GAP, using the SCIP optimization suite [6] for , and and the enumeration of short lattice vectors for and . The AI coding assistant Claude Code (Anthropic), with the large language model Claude Opus 5.5, was used to draft and edit parts of the text and of the proofs, including the treatment of the types , , and , the comparison with the Weyl vector, Theorem 4 and § 4, to search the literature, and to check the results numerically by independent enumerations. All statements and proofs were checked by the author, who takes full responsibility for the content of the paper.
MATHIEU DUTOUR SIKIRIĆ, MSM PROGRAMMING, KARLOVAČKA CESTA 28B, 10452 KLINCA SELO
Email address: [email protected]
References
- [1]J. E. Humphreys, Reflection groups and Coxeter groups, Cambridge Studies in Advanced Mathematics 29, Cambridge University Press, Cambridge, 1990.DOI
- [2]V. G. Kac, Infinite-dimensional Lie algebras, 3rd edition, Cambridge University Press, Cambridge, 1994.
- [3]Leiden Declaration on Artificial Intelligence and Mathematics, 2 June 2026, https://leidendeclaration.ai/, doi:10.5281/zenodo.20302944.DOI
- [4]E. Moorhouse, E₈, http://ericmoorhouse.org/e8/.
- [6]S. Bolusani, M. Besançon, K. Bestuzheva, et al., The SCIP Optimization Suite 9.0, Technical Report, Optimization Online, 2024. https://www.scipopt.org.arxiv.org/abs/2402.17702
