You are currently browsing the monthly archive for July 2022.

Let denote the space of matrices with integer entries, and let be the group of invertible matrices with integer entries. The Smith normal form takes an arbitrary matrix and factorises it as , where , , and is a rectangular diagonal matrix, by which we mean that the principal minor is diagonal, with all other entries zero. Furthermore the diagonal entries of are for some (which is also the rank of ) with the numbers (known as the *invariant factors*) principal divisors with . The invariant factors are uniquely determined; but there can be some freedom to modify the invertible matrices . The Smith normal form can be computed easily; for instance, in SAGE, it can be computed calling the function from the matrix class. The Smith normal form is also available for other principal ideal domains than the integers, but we will only be focused on the integer case here. For the purposes of this post, we will view the Smith normal form as a primitive operation on matrices that can be invoked as a “black box”.

In this post I would like to record how to use the Smith normal form to computationally manipulate two closely related classes of objects:

- Subgroups of a standard lattice (or
*lattice subgroups*for short); - Closed subgroups of a standard torus (or
*closed torus subgroups*for short).

The above two classes of objects are isomorphic to each other by Pontryagin duality: if is a lattice subgroup, then the orthogonal complement

is a closed torus subgroup (with the usual Fourier pairing); conversely, if is a closed torus subgroup, then is a lattice subgroup. These two operations invert each other: and .

Example 1The orthogonal complement of the lattice subgroup is the closed torus subgroup and conversely.

Let us focus first on lattice subgroups . As all such subgroups are finitely generated abelian groups, one way to describe a lattice subgroup is to specify a set of generators of . Equivalently, we have

where is the matrix whose columns are . Applying the Smith normal form , we conclude that so in particular is isomorphic (with respect to the automorphism group of ) to . In particular, we see that is a free abelian group of rank , where is the rank of (or ). This representation also allows one to trim the representation down to , where is the matrix formed from the left columns of ; the columns of then give a basis for . Let us call this a*trimmed representation*of .

Example 2Let be the lattice subgroup generated by , , , thus with . A Smith normal form for is given by so is a rank two lattice with a basis of and (and the invariant factors are and ). The trimmed representation is There are other Smith normal forms for , giving slightly different representations here, but the rank and invariant factors will always be the same.

By the above discussion we can represent a lattice subgroup by a matrix for some ; this representation is not unique, but we will address this issue shortly. For now, we focus on the question of how to use such data representations of subgroups to perform basic operations on lattice subgroups. There are some operations that are very easy to perform using this data representation:

- (Applying a linear transformation) if , so that is also a linear transformation from to , then maps lattice subgroups to lattice subgroups, and clearly maps the lattice subgroup to for any .
- (Sum) Given two lattice subgroups for some , , the sum is equal to the lattice subgroup , where is the matrix formed by concatenating the columns of with the columns of .
- (Direct sum) Given two lattice subgroups , , the direct sum is equal to the lattice subgroup , where is the block matrix formed by taking the direct sum of and .

One can also use Smith normal form to detect when one lattice subgroup is a subgroup of another lattice subgroup . Using Smith normal form factorization , with invariant factors , the relation is equivalent after some manipulation to

The group is generated by the columns of , so this gives a test to determine whether : the row of must be divisible by for , and all other rows must vanish.

Example 3To test whether the lattice subgroup generated by and is contained in the lattice subgroup from Example 2, we write as with , and observe that The first row is of course divisible by , and the last row vanishes as required, but the second row is not divisible by , so is not contained in (but is); also a similar computation verifies that is conversely contained in .

One can now test whether by testing whether and simultaneously hold (there may be more efficient ways to do this, but this is already computationally manageable in many applications). This in principle addresses the issue of non-uniqueness of representation of a subgroup in the form .

Next, we consider the question of representing the intersection of two subgroups in the form for some and . We can write

where is the matrix formed by concatenating and , and similarly for (here we use the change of variable ). We apply the Smith normal form to to write where , , with of rank . We can then write (making the change of variables ). Thus we can write where consists of the right columns of .

Example 4With the lattice from Example 2, we shall compute the intersection of with the subgroup , which one can also write as with . We obtain a Smith normal form so . We have and so we can write where One can trim this representation if desired, for instance by deleting the first column of (and replacing with ). Thus the intersection of with is the rank one subgroup generated by .

A similar calculation allows one to represent the pullback of a subgroup via a linear transformation , since

where is the concatenation of the identity matrix and the zero matrix. Applying the Smith normal form to write with of rank , the same argument as before allows us to write where consists of the right columns of .Among other things, this allows one to describe lattices given by systems of linear equations and congruences in the format. Indeed, the set of lattice vectors that solve the system of congruences

for , some natural numbers , and some lattice vectors , together with an additional system of equations for and some lattice vectors , can be written as where is the matrix with rows , and is the diagonal matrix with diagonal entries . Conversely, any subgroup can be described in this form by first using the trimmed representation , at which point membership of a lattice vector in is seen to be equivalent to the congruences for (where is the rank, are the invariant factors, and is the standard basis of ) together with the equations for . Thus one can obtain a representation in the form (1), (2) with , and to be the rows of in order.

Example 5With the lattice subgroup from Example 2, we have , and so consists of those triples which obey the (redundant) congruence the congruence and the identity Conversely, one can use the above procedure to convert the above system of congruences and identities back into a form (though depending on which Smith normal form one chooses, the end result may be a different representation of the same lattice group ).

Now we apply Pontryagin duality. We claim the identity

for any (where induces a homomorphism from to in the obvious fashion). This can be verified by direct computation when is a (rectangular) diagonal matrix, and the general case then easily follows from a Smith normal form computation (one can presumably also derive it from the category-theoretic properties of Pontryagin duality, although I will not do so here). So closed torus subgroups that are defined by a system of linear equations (over , with integer coefficients) are represented in the form of an orthogonal complement of a lattice subgroup. Using the trimmed form , we see that giving an explicit representation “in coordinates” of such a closed torus subgroup. In particular we can read off the isomorphism class of a closed torus subgroup as the product of a finite number of cyclic groups and a torus:

Example 6The orthogonal complement of the lattice subgroup from Example 2 is the closed torus subgroup using the trimmed representation of , one can simplify this a little to and one can also write this as the image of the group under the torus isomorphism In other words, one can write so that is isomorphic to .

We can now dualize all of the previous computable operations on subgroups of to produce computable operations on closed subgroups of . For instance:

- To form the intersection or sum of two closed torus subgroups , use the identities and and then calculate the sum or intersection of the lattice subgroups by the previous methods. Similarly, the operation of direct sum of two closed torus subgroups dualises to the operation of direct sum of two lattice subgroups.
- To determine whether one closed torus subgroup is contained in (or equal to) another closed torus subgroup , simply use the preceding methods to check whether the lattice subgroup is contained in (or equal to) the lattice subgroup .
- To compute the pull back of a closed torus subgroup via a linear transformation , use the identity Similarly, to compute the image of a closed torus subgroup , use the identity

Example 7Suppose one wants to compute the sum of the closed torus subgroup from Example 6 with the closed torus subgroup . This latter group is the orthogonal complement of the lattice subgroup considered in Example 4. Thus we have where is the matrix from Example 6; discarding the zero column, we thus have

[*This post is collectively authored by the ICM structure committee, whom I am currently chairing – T.*]

The ICM structure committee is responsible for the preparation of the Scientific Program of the International Congress of Mathematicians (ICM). It decides the structure of the Scientific Program, in particular,

- the number of plenary lectures,
- the sections and their precise definition,
- the target number of talks in each section,
- other kind of lectures, and
- the arrangement of sections.

(The actual selection of speakers and the local organization of the ICM are handled separately by the Program Committee and Organizing Comittee respectively.)

Our committee can also propose more radical changes to the format of the congress, although certain components of the congress, such as the prize lectures and satellite events, are outside the jurisdiction of this committee. For instance, in 2019 we proposed the addition of two new categories of lectures, “special sectional lectures” and “special plenary lectures”, which are broad and experimental categories of lectures that do not fall under the traditional format of a mathematician presenting their recent advances in a given section, but can instead highlight (for instance) emerging connections between two areas of mathematics, or present a “big picture” talk on a “hot topic” from an expert with the appropriate perspective. These new categories made their debut at the recently concluded virtual ICM, held on July 6-14, 2022.

Over the next year or so, our committee will conduct our deliberations on proposed changes to the structure of the congress for the next ICM (to be held in-person in Philadelphia in 2026) and beyond. As part of the preparation for these deliberations, we are soliciting feedback from the general mathematics community (on this blog and elsewhere) on the current state of the ICM, and any proposals to improve that state for the subsequent congresses; we had issued a similar call on this blog back in 2019. This time around, of course, the situation is complicated by the extraordinary and exceptional circumstances that led to the 2022 ICM being moved to a virtual platform on short notice, and so it is difficult for many reasons to hold the 2022 virtual ICM as a model for subsequent congresses. On the other hand, the scientific program had already been selected by the 2022 ICM Program Committee prior to the invasion of Ukraine, and feedback on the content of that program will be of great value to our committee.

Among the specific questions (in no particular order) for which we seek comments are the following:

- Are there suggestions to change the format of the ICM that would increase its value to the mathematical community?
- Are there suggestions to change the format of the ICM that would encourage greater participation and interest in attending, particularly with regards to junior researchers and mathematicians from developing countries?
- The special sectional and special plenary lectures were introduced in part to increase the emphasis on the quality of exposition at ICM lectures. Has this in fact resulted in a notable improvement in exposition, and should any alternations be made to the special lecture component of the ICM?
- Is the balance between plenary talks, sectional talks, special plenary and sectional talks, and public talks at an optimal level? There is only a finite amount of space in the calendar, so any increase in the number or length of one of these types of talks will come at the expense of another.
- The ICM is generally perceived to be more important to pure mathematics than to applied mathematics. In what ways can the ICM be made more relevant and attractive to applied mathematicians, or should one not try to do so?
- Are there structural barriers that cause certain areas or styles of mathematics (such as applied or interdisciplinary mathematics) or certain groups of mathematicians to be under-represented at the ICM? What, if anything, can be done to mitigate these barriers?
- The recently concluded virtual ICM had a
*sui generis*format, in which the core virtual program was supplemented by a number of physical “overlay” satellite events. Are there any positive features of that format which could potentially be usefully adapted to such congresses? For instance, should there be any virtual or hybrid components at the next ICM?

Of course, we do not expect these complex and difficult questions to be resolved within this blog post, and debating these and other issues would likely be a major component of our internal committee discussions. Nevertheless, we would value constructive comments towards the above questions (or on other topics within the scope of our committee) to help inform these subsequent discussions. We therefore welcome and invite such commentary, either as responses to this blog post, or sent privately to one of the members of our committee. We would also be interested in having readers share their personal experiences at past congresses, and how it compares with other major conferences of this type. (But in order to keep the discussion focused and constructive, we request that comments here refrain from discussing topics that are out of the scope of this committee, such as suggesting specific potential speakers for the next congress, which is a task instead for the 2022 ICM Program Committee. Comments that are specific to the recently concluded virtual ICM can be made instead at this blog post.)

I’m currently in Helsinki, Finland for the General Assembly meeting of the International Mathematical Union (IMU), which runs the International Congress of Mathematicians (ICM) as well as several other events and initiatives. In particular the assembly voted on the location of the 2026 ICM; it will be held in Philadelphia, USA (with the general assembly being held in New York, USA).

Tomorrow the IMU award ceremony will take place, where the recipients of the various IMU awards (such as the Fields medal) will be revealed and honored. Event information can be found at this Facebook Event page, and will also be streamed at this Youtube page; participants who have registered at the virtual ICM can also view it from the web page links they would have received in email in the last few days. (Due to high demand, registration for the virtual ICM has unfortunately reached the capacity of the live platform; but lectures will be made available on the IMU Youtube channel a few hours after they are given. The virtual ICM program will begin the day after the award ceremony, beginning with the lectures of the prize laureates.

We have an unofficial ICM Discord server set up to follow the virtual ICM as it happens, with events set up for the prize ceremony and individual days of the congress, as well as for individual sections, as well as more recreational channels, such as a speculation page for the IMU prize winners. There are also a number of other virtual ICM satellite events that are being held either simultaneously with, or close to, the virtual ICM; I would like to draw particular attention to the satellite public lectures by Williamson (July 8), Giorgi (July 11), and Tokieda (July 13), which was also highlighted in my previous blog post. (EDIT: I would also like to mention the now-live poster room for the short communic

After the virtual ICM concludes, I will solicit feedback on this blog (in my capacity as chair of the IMU Structure Committee) on all aspects of that congress, as well as suggestions for future congresses; but I am not formally requesting such feedback at this present time.

## Recent Comments