Submind YouTube summaries
Thumbnail for Basic Notions Seminar - Speakers: M. Jardin, L.Silva de Oliveira

Basic Notions Seminar - Speakers: M. Jardin, L.Silva de Oliveira

Watch on YouTube

Video summary

The seminar introduces fundamental concepts in algebraic geometry, focusing on moduli spaces that classify geometric objects up to equivalence and sheaves which generalize vector bundles to accommodate singularities. A central theme is the Donaldson-Uhlenbeck-Yau theorem, establishing a deep link between algebraic stability conditions for holomorphic vector bundles or locally free sheaves and the existence of Hermitian-Einstein metrics solutions to partial differential equations. The discussion transitions from well-behaved moduli spaces on curves to more complex structures on surfaces and threefolds like $\mathbb{P}^3$, where higher-dimensional spaces often exhibit multiple irreducible components, varying dimensions, and singularities described by Vakil's Murphy Law. Recent research highlights the structure of rank-two sheaves on $\mathbb{P}^3$ with maximal third Chern class ($C_3$), revealing that stable reflexive sheaves exist only when $C_3 \leq C_2^2$, while specific transformations involving global sections allow for the construction of new families of irreducible components in torsion-free sheaves, effectively lowering $C_3$ and generating novel geometric structures. Stability conditions are rigorously defined through Hilbert polynomials, distinguishing between $\mu$-stability, Gieseker-stability, and semi-stability, with a clear hierarchy where $\mu$-stable implies stable, which in turn implies semi-stable. In the specific context of rank two sheaves on $\mathbb{P}^3$ with first Chern class $-1$, all stability notions coincide for $c_2 \geq 3$, ensuring that every semi-stable sheaf is fully stable without proper subobjects. The analysis utilizes the spectrum of sheaves to determine cohomology groups via Euler characteristics, proving that torsion-free sheaves in this range admit non-zero global sections. This property renders all pairs delta-stable and very stable under a stability notion defined by inequalities involving reduced Hilbert polynomials perturbed by $\Delta$, eliminating intermediate walls between saturated and non-saturated states found in other contexts. The geometric implications of these findings are illustrated through the mapping of moduli spaces to specific families of curves, such as twisted cubics associated with $c_2=3$ or planar cubics plus a point for higher values where reflexive sheaves vanish. For instance, when $c_2 \geq 4$, only the component corresponding to planar curves and points remains, while at $c_2=3$, the Hilbert scheme comprises two distinct irreducible components linked to twisted cubics and planar configurations respectively. Although Schmidt has demonstrated smoothness and irreducibility for higher degrees using alternative methods, extending these results to products like $\mathbb{P}^1 \times \mathbb{P}^2$ or computing intersection theory on Hilbert schemes remains a significant challenge due to the inherent complexity of understanding the underlying curve Hilbert schemes. Ultimately, the seminar concludes by emphasizing how correspondence between bundles with sections and curves allows for a precise classification of moduli space components based on their geometric realizations as families of plane curves plus points.
Read the full video transcript
Okay, you're clear to start. So, thanks everyone for showing up. I was asked to do a basic notions talk. So, I'll try to explain a few things behind the the work that I do in the double research that I do in algebraic geometry. So, I'm assuming that some of you, perhaps most of you are not familiar with these words, moduli spaces, sheaves. So, I'll try to first explain what a moduli space What is the idea behind a moduli space and the idea behind this notion of sheaves. And end up with some of the results that we have proved on this topic in the past couple of years. Uh try to click on the screen once. Maybe it has to be activated. With the With the mouse, just points on the Exactly, give it one click and now try using the pointer. Yeah. Okay. Okay. So, in in geometry or in mathematics in general, right? So, mathematicians like to classify stuff. The The dream of any any mathematician is to say, "Okay, so any object of that type is in this list." Sometimes you can make a finite list or some Even if the list is not finite, you can do a neat list like the classification of of finite abelian groups. You know, for every Let's say this the the order of the the you can tell how many groups there are, and this is the final list. Um also in representation theory, you can classify representations of um of [clears throat] certain groups. Um And the in in geometry also have this type of classification problem, uh classification [clears throat] of geometric objects, except that uh more often than not, the objects we are trying to classify are not finite. They come in infinite families. >> [clears throat] >> And uh we want to study collection of objects like curves, uh manifolds, or vector bundles in the in uh I'll explain what these are in a moment, which is what I do. And the the answer that algebraic geometry particular gives to this classification problem is called a moduli space. Um so a moduli space is some sort of geometric space. This is in the broadest sense of the word uh or a space, uh in which each point corresponds to one class one isomorphism class of the objects you want to uh to to classify. Right? So we start with a collection of some objects, um a big bag full of stuff inside, um and then you need also to define when two objects are the same, when they are equivalent in some sense. And uh as I said, the moduli space is just this uh set of uh equivalence classes of these objects you're trying to to classify. In such a way that each point in this space corresponds to a unique a unique [clears throat] uh equivalence class, right? So you can find a representative for each point, you can find an object, and then uh that the equivalence class of that object will uniquely be assigned with that point. Um and in addition, we would like this space, this moduli space, to be equipped with some sort of geometric structure. So that we can do geometry on the space that classifies geometric objects. So in this sense, this is the geometry of geometry, right? So we want to [snorts] do geometry in a space that classifies geometric objects. Let me give you a very simple example. Uh say you want to classify all triangles in a plane, right? In the regular Euclidean plane, you want to classify all triangles up to congruence. Right? So up to congruence. So one way to define a congruence class of a triangle is give uh two sides, right? And an angle between them. So you can think of this space of all congruence classes of triangles to be uh well, I have to give three parameters, right? So one positive number for the side of one of one of the sides, another positive number, plus an angle, right? So plus something in the interval between 0 and 180°. So the product of these three uh things will be the set of all congruence classes of triangles. Okay? Because for any any triangle, we have this uh uh side-angle-side case of congruence of triangles and I then I can find the unique address of any given triangle in this address book. So, you can think of a the moduli space as an address book and the and the coordinates here as the address of the particular object you're trying to to classify. And now in this space, right, which is some sort of strip, sort of a I don't know. Sort of a three-dimensional space. Strip like like this that comes off infinitely from the from the board, you can start to do paths, right? So, you can walk around also in this in this space and walking around in this space, you are actually doing a family of triangles. So, you are deforming the triangle when one of the legs of the triangle increases, perhaps one of the the angles of the triangle increases and then decreases. So, you you're building you're you're moving your triangle around. And this movement of the triangle can be translated as a path in this in this space. R plus times R plus times uh Let me repeat. The second arc length is the second length to be bigger or smaller than this one. No, because I'm I can give any sides and the angle between them. So, this always gives me a triangle. Yeah, you know, it's just switch the two. Yeah, okay, I'm lying a little bit because I have to attach labels to to these. Otherwise, you have these automorphisms and I didn't want to to to say that, but since you Yeah, but yes, you can make one smaller than the other, but when they are equal, uh you have this ambiguity ambiguity. Otherwise, then you have to say side number one and side number two and uh kill the automorphisms. >> [clears throat] >> Good. >> [snorts] >> Now, what is a a vector bundle? A vector bundle now is a family of vector spaces parametrized by some uh by some manifold, by some geometric object. For me, the manifolds will be projective algebraic varieties. So, what is that? That is the common zeros of a homo- of homogeneous polynomials. So, I take uh a bunch of uh homogeneous polynomials in [snorts] n plus one variables. These uh kappa here is some field that is the uh that gives me the coefficients of these polynomials. Okay, so I have a bunch of homogeneous polynomials with coefficients in kappa, right? And uh n plus one variables, and the common zeros of these polynomials gives me a projective variety. And this is a uh And And that's what we are going to call space for the purposes of this talk. And uh a vector bundle over an R is something that uh is well, a family of vector spaces. So, for each point, I attach a vector space uh to that point in such a way that it looks locally like a product, right? So, for every open set or there is uh a covering, right? Um I can find a a covering of X by open sets that I'm going to denote so that uh the uh the pre-image of this open set looks like a product. That's what they say when they say that this vibration here, this is a projection map, is locally trivial. Um but once you you know that locally it looks like just a product, you have to know how to glue them up in the varying the various intersections of the uh of the open sets. And these gluing are are done by these so-called transition functions, which are just polynomial functions from the intersections of the various sets in the covering to this group. That tells you how to change the base, right? So, this is a GL GL here is supposed to be how how you move, right? So, how do you glue or compare two different uh vector vector spaces of the same dimension n or R. They are used R. >> [snorts] >> So, informally, vector bundle is a family of vector spaces parameterized by X in such a way that locally it looks like a product. Um and these admit a topological classification in terms of some invariants called the Chern classes, right? I'm not going to explain what a Chern class is, but it's some kind of topological invariants that classifies these these invariants up to uh up to topology. So, why care, right? So, vector bundles are very important for various things, in particular in remaining geometry and the non-linear analysis on on manifolds. Right? So, they are the right mathematical language to describe general relativity and many things in physics, like gauge theory and string theory. So, it turns out that it's the right mathematical object in on which over which you can do physics on manifolds. And physics here is both high energy physics and also general relativity. Sheaves, right? So, a sheaf will be a A sheaf is something that you can think of a sheaf as a vector bundles with singularities. A vector bundles that is singular at some points. Now, how do we go from a vector bundle to a sheaf? So, for each open covering I can look at the set of all local sections. And by that I mean maps from the open set to the to the product. So, maps from each open set in the covering to the product of this open set with the the vector space kappa R copies. And that in such a way that the image of some point in X will be that point in X plus a vector. And this vector changes with with X. Right? So, you can think of if you have a family of vector spaces, this is a sections, these local sections is a family of vectors in those vector spaces. Right? So, so as you move around in your space, the vector is also moving around in the various vector spaces. Well, but this oops, yeah. But this vector is actually well, this F is a function from the the open set to the vector space capital R and and it actually can be regarded as R functions from the open set to the field. So this field can be the reals or the complex numbers for instance. And these are polynomial maps. So that's another way to say this fancy word regular maps, right? So you can think of a polynomial polynomials from the from your open set which is which you can regard as a some CN or an RN if you prefer to to R. So it's a bunch of polynomials. Okay. Um which means that each set of all local sections on an open set which is this curly E of UI, the set of all local local sections on an open set UI is a free module over the set over the ring of regular functions, the ring of polynomial functions. >> [clears throat] >> So together all these uh modules of local sections, they form what is technically called a sheaf, a locally free sheaf of OX modules. And more generally you can consider sheaves, right? So collections of of uh, of modules over each open set, so collection of a collection of, uh, all UI modules. Uh, but instead of these, each OUI modules to, uh, to be free as as it occurs in the case of vector bundles, they can be more general modules like reflexive or torsion-free modules. Right, a reflexive module is a module, uh, over a ring that is isomorphic to its double dual. And a torsion-free module is a module that injects into into its double dual. There's always a map from any module to its double dual. When this map is injective, we call that module a torsion-free module. And when this, uh, natural map into the double dual is an isomorphism, we call it reflexive. So, if you if you allow for these more general modules instead of just, uh, free modules over, uh, for each open set, you get what is called, uh, you know, a sheaf, a reflexive or torsion-free sheaf. Uh, and these are the objects that we'd like to classify. And coherent, I'm not going to explain the meaning of this word, coherent, uh, sheaves of OX modules over projective varieties. Right, so something that to each open set in your projective variety associates a module over the ring of regular functions, um, on that open set. Good. Uh, and, uh, moduli spaces of sheaves, um, which is, you know, now I want to classify all such sheaves up to some, uh, equivalence, some isomorphism or some other equivalence, actually. But we know that they, uh, that there are some, uh, as I mentioned, there are some discrete, uh, topological invariants, which are these Chern classes. So, the first thing we do is fix the Chern classes. This is fixing, I don't know, like the zip code of the Once you fix the zip code, there's still more bits to the address. So, the Chern classes are the first invariants that you fix. Uh, once you know the the fixed code the the the the zip code, there you still have to go around in the different streets to look for, uh, the the the correct house, the correct address, uh, you want to go. Uh, and it's been proved in the, started with, uh, Mumford, David Mumford, in the late, uh, '60s and then continued in the '70s and early '80s, that these moduli spaces actually have the structure of a projective scheme. Right? So, it's, uh, they are also, they can also be, uh, regarded as a zeros of a bunch of, uh, homogeneous polynomials. So, this is the small, uh, miracle. It It didn't have to be that way. It's a small miracle of nature that you're classifying these objects and they are again the type of objects that you started with in some sense. They are again, uh, projective varieties or at least projective schemes. Um, one very, uh, very well-known result in this area is the so-called Donaldson-Uhlenbeck-Yau, uh, theorem that provides a bridge uh, between differential geometry and, uh, mathematical physics with, uh, algebraic geometry in the middle. So, a holomorphic vector bundle, so, that's the statement of the Donaldson-Uhlenbeck-Yau theorem, that any holomorphic vector bundle or locally free, uh, sheaf on a smooth compact killer manifold is uh is slope stable. This uh it's some some property some algebraic property of a locally free chief if and only if it admits a compatible hermitian Einstein metric. So, it's uh telling you that uh some bit of despite the technical word here is that this statement telling you that some bit of uh algebraic geometry is some uh is the uh necessary condition for the existence of solutions of a partial differential equation on a manifold, right? Something that is very scary and uh horrible, right? So, solving partial differential equations on curved spaces is very scary, but this algebraic uh condition stability tells you or uh uh uh gives you a necessary condition for the existence of solutions for these uh scary PDEs. So, this is one example why uh sheaves and and moduli spaces of sheaves uh became uh fashionable and important um from the more or less the the '80s uh so on so forth the past 40 to 50 years. Uh and this is also the bridge with uh string theory and a bunch of mathematical physics that is done nowadays, right? Because uh it turns out that the moduli space of uh uh stable sheaves is uh the same as the as the moduli space of solutions of a given PDE, right? So, we can classify solutions of a PDE using algebraic geometry. However, uh uh these moduli spaces can be very nasty um very nasty spaces, right? They are even in uh It's not because they are there they are particularly nice. And uh even proving that they are non-empty, proving the existence of uh of a stable uh bundles uh with given uh turn classes and uh given you know that there is a a blue house in a given uh for a given zip code, right? Is not a completely trivial uh matter, right? So, you have to walk in all the streets to check if one of these houses are is blue. So. >> [clears throat] >> Good. So, let me uh now uh start talking about some of the the results that are known, you know, the in the in the case of of curves, which are uh projective varieties of dimension one, this is the simplest case, and and that's uh the situation where you can tell the most about these spaces. So, I'm going to use this notation here uh to mean the moduli space, right? So, the set of all uh semi-stable vector bundles uh of rank R and degree D, right? So, this D here is playing the role of the turn class. Now, there's only one turn class that is important, right? It uh this turn class is called the degree, and this is the invariant we fix for curves. Never [clears throat] mind what semi-stable is. It's just some property of a vector bundles. Uh we actually need need this property in order to classify because classifying all bundles um doesn't give you a very good geometric structure, but classic but if you throw out some uh bad bundles, some uh uh bundles that are called unstable, and you keep just the semi-stable ones, then uh this uh moduli space this classifying problem, this classification problem is solved by a uh a decent uh variety. Uh so, this is always a a uh a projective variety, right? And this is irreducible projective variety of a given dimension that that depends on the um rank and the genus of the curve. It should depend on the degree also. No? Oh, yeah, this is right. Yeah. Sorry. Yeah. So, that depends only on the rank and the genus of the curve, which is this G. Uh and this is for this situation where when the when uh rank and degree are coprime. So, when rank and degree are coprime, these uh projective varieties in addition uh smooth. So, this is the uh best possible situation in which your objects are classified by a by some neat object as well. So, some smooth uh space. Some smooth projective variety. Um And then uh there are some some uh there is a very influential paper by Atiyah and Bott from the um mid-80s that tells you how to compute the cohomology of these spaces. So, you can tell something about the topology of these spaces. Um The Narasimhan-Seshadri theorem provides you a bridge with uh representation theory telling that uh stable bundles of degree zero uh correspond to irreducible representations of the uh fundamental group of the curve. So, this is an another uh unexpected link between uh a little bit of representation theory representing these uh these these particular groups um fundamental groups of uh Riemann surfaces into uh into unitary uh groups, right? Into UN. They are uh they are related to these bundles, and they are classified by the same uh class of objects. So, again, it's not only the classification of geometric objects, but also the classification of algebraic uh objects. You're you're also classifying representations of a certain group uh in uh the space of unitary matrices, the group of unitary matrices, for instance. And uh And Donaldson also gives a this is uh you know, the the first version of the Donaldson-Uhlenbeck-Yau theorem, which it gives uh uh the existence of a unitary connection uh with constant curvature on the the stable bundles. So, these objects are actually in the middle of uh a bunch of other things, representation theory, differential geometry, algebraic geometry. They are uh really uh quite important uh in in in in maths in general, because there are many things that you can relate to these moduli spaces of sheaves on a curve. Even if you're not interested in algebraic geometry, this might be a a nice uh bit of mathematics to learn. Okay. So, the moral of the story is that on curves, moduli spaces are well-behaved, and it's possible to study their topology and geometry through algebraic means, right? So, we can really understand the geometry of the space that classifies these bundles uh on curves. The situation is less rosy for surfaces, but still um for some in class of surfaces, there is a lot you can say about these are moduli spaces. Now, we need not only the rank and and but also two turn classes that we have I'm going to call C1 and C2. Um these are again the the zip code the the the topological invariants of the sheaves. Um And you can say something about the existence. So, when this is always non-empty when C2 is sufficiently large and how large depends on the surface. We have some some some estimates given by these are so-called Bogomolov inequality. Um When S is a special kind of surface called a funnel surface. So, that's the So, if you tell me if you give me more information about the surface, then I can tell you more about the moduli space of sheaves on the surface. Uh so, on a funnel funnel surface, these guys are always smooth. So, these are again an instance in which um there's a decent classification for these um on a surface. Uh and if you specify the surface a bit more like a K3 surface, which is specially kind of nice kind of surface, you can tell more about geometric structures on this moduli space because this is another trend trend also in in algebraic and differential geometry. Okay, so you give me you tell me what the space are. So, now I can you tell me what the moduli space is. Uh now, let me do geometry on this moduli space. And doing geometry is finding different different types of geometric structures um uh on uh, this moduli space. So, in particular, uh, Mukai in this, uh, very important, uh, example tells you that these moduli spaces are hyperkähler manifolds. Which is maybe, uh, the type of, uh, geometric structure that is the most rigid and and then has the most, uh, uh, amount of information needed in order to define what is a hyperkähler, um, manifold. So, uh, these the moral here that I that that I'm trying to, uh, to to say is that, uh, these moduli spaces they answer a classification problem, but they are also very interesting, uh, varieties on their own, right? Because they also have, uh, very rich, uh, geometric structures and, uh, and and and properties, topological properties and geometric properties that are somehow inherited [clears throat] and can be, uh, and can be read from the, uh, original manifold that you started with. >> [snorts] >> So, they're also good ways to produce new, uh, interesting examples of varieties and, um, and geometric structures on manifolds. Um, one very interesting bit of theory that was done in the in the the early '90s is the the the so-called, uh, topological, uh, Donaldson invariants, right? So, they, um, >> [clears throat] >> So, the the the the the the story's more or less like this. The Donaldson invariants are sort of, uh, second order topological invariants of your original, uh, surface. Start with a surface and then you make this other space, a moduli space, uh, you make this other, uh, projective variety, this, uh, moduli space of uh, of bundles on a surface. Uh now you look at the topological invariants of this new space. And it turns out that this the topological invariants of this new space are also topological invariants of the original space. So going to to the moduli space is also a way to produce new topological invariants of your original manifold. And Donaldson was able to to prove very ground many groundbreaking um results in differential topology using this this technique. All right, so starting on a manifold, looking at a moduli space of stuff on the original manifold, do topology on this new space, and discover something about the topology of the original manifold. Discover some highly non-trivial topological invariants of the original manifold that you started with. So in in this sense, the moduli spaces became a tool to understand the differential topology, to understand the algebraic geometry of the original manifold that you that you were interested in. Uh now, for varieties of higher dimension, dimension greater than two, all of the questions above are uh open and often challenging. All right, so and this is and this is the threefold. So I explained what a moduli space is. I said a little bit about sheaves. Now let let me get to the threefolds. So a threefold is a projective variety of dimension three. Okay. Um And here I'm I'm going to restrict myself to my personal favorite to threefold which is the P3. Uh which is the the simplest possible the projective spaces are the simplest possible projective varieties, right? Um >> [clears throat] >> and it's the it's the situation where uh as far as I know, you can say the most about uh these these moduli spaces. The fact is that these moduli spaces are neither irreducible nor smooth in general. Irreducible here means that whether or not the moduli space uh has a has one or various pieces. Something that looks like uh like this, the union of two lines, this is not irreducible, right? Because it has two two bits. And these These are called irreducible components. And these irreducible components they also have different dimensions. Like a plane with a straight line stuck on it. Uh >> [clears throat] >> and uh And in general, the the moduli spaces of sheaves on the projective space uh may have several irreducible components. They look like stuff like this. >> [clears throat] >> Uh However, there are some uh well-behaved cases. There are some cases for for the Chern classes that we can say something uh about these moduli spaces. In principle, they are very uh very wild, but there are some cases in which we can uh they are not so wild and we can say something about them. Here it should be C1 uh equals zero or minus one. It should be a minus one, sorry. >> [snorts] >> So, first is the question of of non-emptiness. Right? So, when I'm always on range two here. >> [clears throat] >> So, if you fix these two numbers, the first and the second turn class, the third turn class cannot be too big. Right? So, if it's larger than something, we know that this moduli space is always going to be empty. Depending on the the the C1 and this and the C2, if C3 is very large, this moduli space is empty. This is the existence. Otherwise, otherwise, if if the C3 is a smaller than the this number here, then this is non-empty provided this parity condition is satisfied. This parity condition is satisfied. I'm implicitly assuming that C2 is positive here. Even though C3 can be whatever, it can be negative. In the case of projective spaces, these turn classes are just numbers. Right? They can I can present them just as numbers. Uh When the third turn class, when this number C3 is the largest possible, then this moduli space is smooth, irreducible. It has a known formula for the dimension that I'm not going to write here. Right? But, it's This is the nicest possible situation. It's just one piece that This and you can classify these uh uh different decent enough variety. And also when for these when C2 is zero, C1 and C2 are zero. Uh you can also say something about these modular spaces. Uh they are irreducible of a certain dimension, and they are they are going to be irreducible provided N is not large enough. It's It's not too large, right? Between two and N. It has to be any number between two and N. But in general modular spaces of sheaves on threefolds obey what it's what became known as Vakil's Murphy Law. Every conceivable singularity can be found on some Hilbert scheme or modular space of sheaves. Right? So, modular space of sheaves on threefolds can be as bad as you want. You can find any type of singularity in these modular spaces. Uh Sorry? Uh Yes. No, yeah. Even if you start with a good variety, >> [clears throat] >> you can on P3, you can find any conceivable singularity on as as a module in some modular space of stuff on that variety. Now, Uh, some cases in which the the moduli space is not, you know, ultra nice, but we can still tell something about it. Um, so this situation, this this moduli space is connected connected because for for this case the rank two sheaves it's not even know whether the moduli space is connected. Can have a various disconnected pieces. And it has two irreducible components. So let me Let's see. So this is the case of uh M 2 0 2 uh So four is the max. So C3 max here is equal to four. And in this case we know that it's irreducible. So if we fix C2 equals two and C1 equals zero max is four. So there are two cases to consider. The case C3 equals two and also the case 2 0 2 0. And then all the negative cases because this C3 can be as negative as you want. But the these two cases um So this now has two components. One has dimension 13 um Yes. One has the has dimension 13 and the other has dimension 17. And this guy here has three components. One has dimension 13, the other um 17 and finally 21. They go by by four. >> [clears throat] >> So they have two and three irreducible components and we can tell what is the genetic point in each of the components. I can uh we can tell uh what they look like. Um now, if instead of uh C1 = 0, you put C1 = -1 here, then again, C3 max is again four. And there are two cases that we studied. These are two cases. Uh This again is going to have two components of dimensions 11 and 15. And this will have four components. Uh 11 15 19 and the other I don't remember. I should have checked. I want to say that it's uh 17, but I'm not sure. Um And I drew it like that because I don't know whether this component intersects this other component. I only know that it is intersecting one of these two. Possibly the other one, but uh we only proved for one of these two, but not for uh the intersection with that one. Uh So, yeah, so these are the four irreducible components and two irreducible components that we found. Uh Already in the this other space with second chain class equals three and third chain class zero has at least 11 irreducible components and we don't know the total number. Still haven't Still don't know the total number of components. We found 11. We don't know if if we found them all. We also know that the number of irreducible components grows to infinity with C2. Right, so the larger the second trend class the higher the number of components. But we don't know how this growth occurs. If it's a polynomial, exponential, logarithmic. If it only grows, if it grows diminishes and grows again, if it's more whether or not this growth is mono is monotone. Right, so we don't know how how it grows. And as I said, it's not even know in gen known in general whether these moduli spaces are are connected. And and the one of the new things we've been trying to explore is what happens when the third trend class is negative. Right, this is also allowed. This is also you know there are such shapes with negative C3, but we haven't we just started to study these moduli spaces now. Maybe the last slide I'll mention one of the the key techniques or the key idea behind trying to to understand these results. This is done by the so-called sacred springs. Sacred correspondence is a very nice link between vector bundles and curves on P3. It tells you that whenever you consider a pair of a bundle with a section of a global section, this is the same as considering a curve or more generally a one-dimensional scheme because it can be a curve with a bunch of points. They don't have to be pure dimensional. So, it's it's a one-dimensional scheme together with a section of this order dualizing sheaf. So, there is this correspondence. And then you can actually study the moduli space of pairs of bundle with section. And this has been studied. There's literature about that. This is a projective scheme. And what we found recently is that for each irreducible component of this moduli space of pairs, you have rational maps. One map goes to goes to the moduli space of sheaves, which is this M here. And another rational map takes you to the Hilbert scheme of curves. Hilbert scheme is something that classifies curves inside P3. And these are And And you have these rational maps and this is This is one way to understand this. So, for instance, in this particular situation, these these irreducible component here, the the sheaves in this irreducible component, they will correspond to twisted cubics inside P3. And I think it's twisted cubics. Yeah, they should have degree three. So, one is a twisted cubic and the and this guy is a twisted cubic together with a with a point. >> [clears throat] >> Is it a conic and a line or just a conic? Conic and a conic and a point, yeah. I'm I'm confusing with this is the other case, yeah. So, this is a conic and this is a conic with a with a floating point, yes. This is the the the the the planner in the So, the the the sheaves, the generic sheaf in this component here via this same correspondence corresponds to plane plane curves of degree two, conics. And this component has to do with conics with a with an added point somewhere. This one is again conics, conics and a point. This is conic and two points. And this third component is is something like a line and a or I guess it's a double line. I I don't remember it. Maybe it's something like a double line with some point as well. Or the double line with one point or But it it comes from some non-reduced schemes. So, it's understanding the the So, And Leonardo's presentation will give you more details on that, right? So, that's what he will present next. So, the way to to study to understand these irreducible components it's tying them down to the irreducible components of the Hilbert scheme. And that's the one of the the beautiful things about this topic that it gives this relation between the uh classification of sheaves and classification of curves and how to go from uh one to the to the other. Um and that's all that I have for today. Thank you very much. >> [applause] >> So, these conics are they're intersecting at their common points. No, they don't. Depending on the So, yeah. So, so this is just one conic. There's no other curve. Just uh just a And two of them Yeah, at this point may or may not. No, these these are two different conics. So, one uh family of sheaves correspond to just one conic and the other family of sheaves correspond to one conic plus a point. Right. So, these are uh two different families. Are there any general results about the moduli space in dimension greater than two? As far as I know, only that uh they are projective schemes. Otherwise, they can be pretty much anything. And connectedness only for the Hilbert scheme, which is the rank one case. Yeah. Are there non-commutative analogues of the rank of the LPS directions and Um I think there should be non-commutative analogs of moduli spaces, so for instance, people do in super geometry it's the kind of non-commutative geometry where the non-commutative non-commutativity is very mild, just has these Z2 grading on the algebra. And I know that in the case of super geometry, there is a theory of moduli spaces for bundles on super manifolds. So it's possible that this is viable also for non-commutative non-commutative as well. It would be classifying all modules and and this has also been done for associative algebras. Classification of of a modules over associative algebras, they also admit moduli spaces. So it's possible that there is something in the more generally in non-commutative geometry. Thanks everyone. I think we take a little break before Take a 5-minute Yeah, 5-minute break. >> [cough] >> Before the last [clears throat] bit. Okay. Yeah. Okay. >> [snorts] [snorts] >> Uh Professor, please share the screen on Zoom. Okay, sorry. Thank you. Yeah, I'm Whereabouts? Um there should be maybe in the top left a little box with Zoom. Uh the Z in there? Uh hold on. Okay, and then you have gone view on the top. Uh top left, you've got view. And [snorts] uh full screen. Here. Um No, it's not there. You have to okay something here. I don't know why it's not giving you full screen. Okay, I'll be there in a second. Hold on. Okay. Oh, here it is. Here it goes. We managed to do it here. You don't have to come down. Thanks. Uh so, first I want [clears throat] to thank you to invite this presentation and Marcus to invite me to be at here at I speak. So, I going to talk about the moduli space of torsion free sheaves with quasi maximal C3, where is C3 is the third Chern class that Marcus talked before. This is [clears throat] a joint work with Charles Almeida, my co-advisor, and Marcus, my advisor. And first we going to talk about the historical background of this how developed this study of moduli spaces, some basic definitions, the moduli space it corresponds to the question that Marcus talked before, and the moduli space we are going to show you how it is, and some next steps. So, first we going to denote by M C1, C2, and C3 the Gieseker-Maruyama moduli scheme that parametrize semistable rank two sheaves. So, Marcus has a R before the C1 for the rank. Here we going to talk about just rank two sheaves >> [clears throat] >> on projective space P3. C CI denotes the Chern classes and we can consider this subscheme that's denoted by R and C1, C2, and C3. This one are going to be just the stable reflexive sheaves. So, the double dual the sheaves is of mark is double dual. So, this is a projective scheme. And for Maruyama that there are the Maruyama bounds that he can be very bad. And these two papers say there is some examples how this happen. The moduli space have reducible components of different dimensions. So, we can fix the Chern class as minus one. We have two options for sheaves on P3 of rank two is zero or minus one. So, we're going to fix minus one. And we have that C2 and C3 have some possibilities. For example, C2 need to be positive. And then we have this condition on parity of C3. And this bound for C3 it's always um less or equal than C2 squared to be a stable reflexive to be a stable sheaf. We can have sheaves with more larger but it's not going to be stable or semi-stable. So, first Hartshorne proved that for when we have C1 equals minus one, all these stable reflexive sheaves are going to have a third Chern class less or equal than the second Chern class squared. And 5 years after O'Grady and Spindler proved that this bound remains true for free sheaves. So, Hartshorne proves just for reflexive and 5 years after we have this bound for torsion-free sheaves. And Hartshorne computes the moduli spaces when we have the maximal C3 for reflexive sheaves. And O'Grady and Spindler computes the moduli space with maximal C3 for torsion-free sheaves with C2 greater or equal than 6. And recently we have this paper by Schmidt that with different techniques he computed the moduli space for all C2 with the third Chern class being the maximal. But when we talk about reflexive sheaves, they are that things that I that the means that I say before that uh uh >> [clears throat] >> C3 is less or equal than C2 squared, but there are some C3 we don't have stable reflexive sheaves. For example, Schenk studied this for fixing C1 equals minus one and C2 greater than zero, we have that for this range we don't have a stable reflexive sheaves. Some years after uh Miro-Roig extended this result. So, if you fix this speed as this number, so for C2 greater or equal than four, we're going to have a range that is no stable reflexive sheaves. So, >> [clears throat] >> until now we have those moduli spaces. For uh C2 C3 greater than C2 squared we don't have stable sheaves and for reflexive sheaves there is some gap in C3 that is no stable reflexive sheaves. But what about uh the torsion-free sheaves? So, there is a paper uh where Almeida, Jardim, and Kumar construct a possibly new family of irreducible components for certain moduli spaces. This is this big theorem and we can say like that. We start with a sheaf, reflexive sheaf with C1, C2, and C3, and you take a subjective map to O. S >> [applause] >> where S is just S points on the tree. So, you can take the kernel of this map. I going to put a prime here. So, we can take this kernel of this map and we going to have a sheet here that we going to have Chern class C1, C2, C3 - 2S. What is S is the amount of weights we have here. So, they construct this new family. So, if you have a a sheet in this reflexive the moduli space of this reflexive sheets, we can construct some sheet in another moduli space and this one going to be a torsion-free sheet. So, but [clears throat] the it's not just a stable reflexive sheet, we need to be a smooth point in the moduli space. So, if you have a smooth point in our moduli space of stable reflexive sheets, we can do a transformation by a by points and have a sheet that is a torsion-free with this Chern classes. So, this this theorem is like that. So, for any reducible component of our >> [clears throat] >> the reflexive sheets, we can construct a new new sheets on this new Chern class. And we have the dimension of this component. It's going to be 8 * n - 3 + 2 e + 4S on n and n, e, and S are a relation as with the Chern classes of the first weight. >> [clears throat] >> In particular, if you the moduli space of the reflexive sheets is reducible, so the this component is going to be reducible, too. >> [clears throat] >> So, now I can go to some basic definitions. I'm going to start with what is a stable sheaf? To do that, first uh we remember what is the Euler characteristic of a sheaf. So, for this we're going to take the homology, the alternate sum of the dimension of the homology of the sheaves. So, this [clears throat] homology stops in the dimension of the our variety. In our case, it's three. And we can define the Hilbert polynomial of a sheaf. So, we're going to take uh this polynomial as just take the Euler characteristic of the sheaf twisted by the integer. And this is going to be a polynomial this function going to be a polynomial, and we can define the reduced Hilbert polynomial. We're going to take the Hilbert polynomial and divide by the rank of the sheaf. >> [clears throat] >> So, what's going to be a stable sheaf? Uh a torsion-free sheaf is going to be Gieseker-semistable if for any proper subsheaf you have the polynomial the reduced Hilbert polynomial of the uh subsheaf is lesser or equal than the reduced polynomial of the original sheaf. And if this inequality is strict, we're going to say that E is just stable. There is another uh definition of uh stability that is the slope stability. And we start with a torsion-free sheaf, we take the degree of the sheaf is invariant, and the rank of the sheaf, and when you take this fraction, we have uh what we call the slope of the sheaf. So, they're going to be the same as the Gieseker-stable. We're going to say that a sheaf is stable if for semistable if for all subsheaves, we have a rank less or equal than our sheaf or zero, [clears throat] we have that the mu-stability the mu of F is less less or equal than mu of E. And it going to be stable if the inequality is strict. We have these implications. So, mu-stable implies stable implies semi-stable, that implies mu-semi-stable. But we are working with rank two sheaves with the first Chern class minus one. In this case, all the notions of stabilities are equal. So, mu-stable it's equal to um Gieseker stable and all semi-stable sheaves is in fact a stable sheaf. Now, I going to talk about a little about the moduli stack of polystable >> [clears throat] >> This can be found in this paper. So, here we put the X with a smooth projective fold. But we can all the things I do is just for P3. So, first thing we can define is a coherent pair. Going to be a sheaf and a global section of the sheaf. Possible zero. It's possible or you can take this global section as zero. So, we going to say that this pair is pure if the sheaf is pure and S is a non-trivial section. And we can define now stability notion on this pair. So, we are going to take the If you forget about the epsilon and the delta, it's just a reduced Hilbert polynomial. But you are perturbing by this delta. Because if the section is non-zero, so we have that epsilon is one, so you are perturbing our reduced Hilbert polynomial and the section is non-zero, you don't do nothing, Just take the same Hubert pair, reduce it to Hubert pair. No, no. And this is the paint of the delta. Delta is a just a polynomial that is positive. Here are ordinary polynomials if the voting is one or other. So, >> [clears throat] >> this is what we call the delta stability. So, when I say that it's stable, semi-stable, if for any proper super pair, we have this inequality here. And if it's strict, we call that it's delta stable. >> [clears throat and cough] >> Uh Okay, we have that for example in our case in a Gieseker with rank two and first Chern class minus one, we don't have any proper semi-stable sheaf. But here we can have. And when we have, it's depends on the delta. And if the delta if for this delta, there is semi-stable sheaf, proper semi-stable sheaf, we call this delta a wall. >> [clears throat] >> And we can define the collapsing wall. What What we call collapsing? Because if you take delta as this polynomial, every pair is going to be unstable. So, we have this >> [clears throat] >> delta stability and for every delta greater than this one, everyone is unstable. And we have another definition that's now depends of the our section. So, if I have a pure coherent coherent pair, it's called saturated if the co-kernel of the monomorphism of the section is torsion free. So, you have a pair, look to the section which is this induced a monomorphism from O to the my sheaf. And if the co-kernel of this monography's is torsion free, we call the pair saturated. >> [clears throat] >> And we can define now a second wall. This wall is the maximum possible wall that is not greater than the collapsing wall. So, what happens is between these two walls, this new wall and the the collapsing wall, all the sheaves are torsion free are saturated. So, >> [clears throat] >> we have This line is our delta. And this is the collapsing wall. So, after this, every pair is unstable and we have this [clears throat] one. So, everyone here is saturated. And you know how big is the this chamber. Chamber's thing between two walls. And we can define So, we define the last wall, the second last wall. Now, we're going to to the first wall. Because we start with with no positive polynomials. So, we we have zero here. And this first chamber and we can have chambers here. And this first chamber we're going to be the >> [clears throat] >> a chamber that every stable Oh, sorry. Define the very stable first. We're going to say that a pair is very stable if the is delta stable as a pair and my sheaf is is curve stable. What happens [clears throat] is between this chamber, everyone that um Oh, sorry. Every sheaf can that is stable a section, we're going to build up the stable. So, until now we have this structure, this line of polynomials, and this wall in chambers here. And that is this theorem that when we consider the space of pairs here, we can have these two maps. Uh when we choose delta between these two walls in this chamber, we're going to have a map from the from the space of pairs to the Hilbert scheme. And when I choose delta between zero and this first wall, we're going to have our map from the space of pairs to the moduli space of sheaves. And [clears throat] we know uh how is this map. We know the fibers. So, now what we're going to do? Uh If you don't have any wall, because just the collapsing wall uh it's always exist, but the other ones maybe or not it exist. If you don't have any walls, so these chambers coincide. So, we're going to have these two maps at the same time. And how you know this uh the Hilbert scheme, if you know the Hilbert scheme, we can use this map to have information about the geometry and topology of this this curve. I don't know what happened. I don't know. Oh, okay. Uh so, you can use the geometric and topological information of the Hilbert scheme. Sorry, I'm going to back. Of the Hilbert scheme to understand better what is the disk curve moduli space. So, we want to see how this work in our case. The first one we're going to see that every pair with a C2 greater or equal than three the torsion the torsion-free sheaf with quasi-maximal Chern class have a non-zero section. I put [clears throat] greater or equal than three, but the demonstration I have here is just for the C2 equals three. Sorry. Uh to do this demonstration, first we need to recall what is the spectrum of the sheaf. It's going to be a collection of numbers, integers in fact, that the cohomology of our sheaf twisted by some number is the base of these numbers. So, the case we're testing here is the the second one. Uh Okay. So, the second cohomology of our sheaf E depends of the first cohomology of the structural sheaf of P1 twisted by the spectrum plus the first twist here plus one. So, you can use this this cohomology to compute the cohomology of E if you know what is the spectrum. And there is more about the spectrum. For if you look it's just uh some numbers. We don't have how many numbers, for example, the first case with this theorem and any condition about the number. >> [snorts] >> Yeah, yeah, you to construct this number you restrict to a plane first and look to the plane the sheaf in the plane. I do know. Okay, I don't know. I It doesn't have to do with the splitting on two ones. No, it's not because if you one for a two, you would get just two ones. No, yeah, that's the general genetic splitting type. >> Yeah. For a rank r. This this is the spectrum. The K I AM is just a the generic splitting type. So, the spectrum is this number K and [snorts] this number K that will happen is in the homology. What does M mean? M is The the amount of numbers we have in the spectrum. They are going to be the C2. Okay. So, the number of in the spectrum is C2. Uh okay. So, you have this uh proposition that say if you have the spectrum of the sheaf and you take uh negative number so, every number between this negative number and minus one has to be in the spectrum too. So, we have a condition to say how the sheaf how uh how many how many numbers do you are right now is always the C2, but now we can say more about how negative you can have a number in the spectrum. So, going back to our lemma first we have that uh Euler characteristic of our sheaf is going to be that and you can use Heisenberg to my heart to calculate the Euler characteristic of the sheaf and it's going to be one. As the sheaf is stable, we have the third homology is zero. So, we have that the first the global sections minus the first homology plus the second one equals one. If the second homology of our sheaf is going zero, so we have that the our sheaf admits at least one global section. And we are doing that by analyzing the spectrum. So, when you look to the spectrum of the sheaf, is that firm? Uh so, for the H2 not be zero, we have that minus four is going to be in the spectrum because of the homology of P1. But, by our uh proposition we see before, if minus four in the spectrum, so I have minus three, minus two, minus one. But, we have only M uh C2 numbers in the spectrum. So, we can have minus four in the spectrum. So, this homology need to be zero. And as you as you see, if it's zero, so our sheaf admits a global section. >> [clears throat] >> Uh we have this proposition that if I if I have a torsion-free sheaf with rank R and have a section, no new section, so for every non-trivial section, if our sheaf is stable, so the pair is going to be uh delta stable. And we see that very stable is delta stable plus the sheaf is going to be stable, so our pair is going to be very stable. Uh the [clears throat] other lemma we need is prove that okay, we have a global section, but uh this pair is saturated. And we can prove that in fact the Z the Z is saturated um and you have how we know what is the the co-kernel of this map. And as I I say before, we have to see if there is more another ones or just the classical one. So, a possible one going to be a sub object that have the same stability because of all this when I have a strictly a proper semi-stable pair. So, I need to see if it's possible in that case I have a proper semi-stable pair for any doubt. So, we going to be defined by a sheaf and a section. [clears throat] And the first thing we can look is that if the this sheaf don't have any global section, so the delta Hilbert polynomial coincides with the Hilbert polynomial. So, we going to have the same Hilbert polynomial for this case if they don't have a global section. As our sheaf we state start with a pair. Our sheaf L is going to be a sub sheaf of our initial sheaf E twisted by one. But E is stable. So, I know that the Hilbert polynomial of L is going to be strictly less than the Hilbert polynomial of delta of of E. So, we have this inequality here. But you can consider the following diagram. We start with E. Oh, okay. And take the double dual. And we going to have just a structural sheaf of a point here. And here we have L as a sub sheaf. As a double dual going to be the structural sheaf O. So, you have this map O to the double dual to the structural of a point. We take the composition. We have this composition going to be or new or no new. If this composition is zero, so we have that L is going to be just uh, this is sheaf. If it's no move, so it's subjective and we're going to see going to have a rank one sheaf into the structural sheaf to the structural of a point. So, L going to be the ideal sheaf of this point. So, the only three sweepers we have for this uh, our our pair is the structural and the zero and the loop section. The structural and the section that is the collapsing wall and the ideal sheaf the ideal sheaf of this point and don't have any global section, so zero. As E is a stable we have that uh, O and IP and the ideal of the the point not going to stabilize our sheaf because our sheaf is stable this two are same our sweeper our sub sheaves. As he is a stable not stabilizer E. Okay? So, [clears throat] uh we have don't have walls in our case. So, at the same time we have this morphism to the Gieseker the Gieseker moduli space and this one to the Hilbert uh, Hilbert scheme. >> [clears throat] >> That's Uh, but what Hilbert scheme is that? That one. We can calculate it. Uh, you know that going to be this section this uh, exact sequence the section and so the kernel is ideals of something. We can calculate with the Chern class. And looking to the structural the of the sub scheme sub scheme Y we can calculate it the Hilbert polynomial and we have this three three times T plus one. And this system scheme you know what what is. So is the union of two no singular rational varieties 80 and 8 prime. Where 80 is a twisted cubics. And 8 prime is planar cubics plus a point. And have a intersection. So it's connected. >> [clears throat] >> So if if this map going to be subjective we can use the properties of the Hilbert scheme to conclude things about the his command of spaces. So this is the position. We have if we take X1 of the ideal of the chief by O twisted by minus one, this is no loop. So we have excitations. So we're going to have our stable chief in the space of first. And [clears throat] we can complete that and prove that this in fact no loop. And now we have this map is subjective. We know this one have two components. So this one have two components. And what is this components? One is going to be chief associated with twisted cubics. And another one going to be a planar cubic plus a point. This one is about this T component I said before. And this one going to be the reflexive chiefs. Where you know what is. Is a reducible rational of dimension 19 and generically reduced. So [clears throat] you know what is the moduli spaces with different classes has two irreducible components. The component of the reflexive is associated to the twisted cubicity. and the thick component that is associated to the final cubic plus one point. And do you know the dimensions of uh each component? This is the references. And thank you. >> [applause] >> I think that you have to, you know, read your slides very carefully because they were they went a bit fast. Oh, sorry. I kind of anxious. Just one comment. If you go back a couple of slides. Okay. Which one? The the one that has the two maps. Yeah, this one. Uh except that here they are not right since you showed that there are no walls. Yeah. This is not uh these are not rational maps. These are really morphisms. Oh, okay. >> actual maps between Okay. Thank you. >> [clears throat] >> So, what's this result of Schmidt? Uh Okay. I go back to where I say. Yeah. Uh because Hartshorne uh do all these things for reflexive sheaves. So, 5 years after, Oprea and Klee proved that uh the bound is the same and calculated the >> [clears throat] >> the moduli space for the maximal ones with C2 greater or equal than 6. Schmidt uh made a different uh approach and he proved for every C2 in a different very different way of the connected spinner. Yeah, it shows that the moduli space is irreducible and smooth. It's smooth, irreducible, smooth, and say what is exactly. And it looks like a Grassmannian vibration over something. >> Yeah. It gives a very complete model. And this is the nicest possible space. >> Yeah. This is the nicest possible. Yeah, and then this case for C2 equal three, we have two components, but for C2 greater or equal than four, so there are no reflexive sheaves. So, the moduli space going to be just this C component that is a planar curve plus a point. It's a nice one, too. This is what Leonardo is planning to say tomorrow. >> Yeah. Explain the case C2 and larger than six. And any of this can it be extended to say P1 cross P2 like I don't know. Has there any study been done for moduli space on P1 cross P2? Say P1 cross P2 or Not that I know. Or blow up formula? The the limitation is understanding the Hilbert scheme of curves on whatever. Right? Because this result is like any any threefold you have these rational maps and so on. Yeah. But then you have to know the Hilbert scheme of curves on the what type of on X. So, we have two of these points. >> to apply this this idea. Okay, I think it'd be nice to have examples to apply this thing to the water two points. I actually think the intersection theory just on the surface P2 is hard. I don't know if you I mean, intersection numbers, for example, on the Hilbert scheme of points on P2 could be crazy. It's not So, P2 is not easy as a surface to apply, you know, you know? Yeah, and there is a kind of luck because there is a paper that say about Hilbert scheme with a fixed degree and the genus of the subscheme and say for a range is irreducible. And we know what is. It's just a planar curve with a strict So, this help a lot. And to compute topology intersection on these on these Yeah, even if you give me a let's just just one component. It's out of my hands. Out of my reach. It's not computable. I understand [applause] that It's a Marks.