Birational automorphisms of Severi-Brauer surfaces
Constantin Shramov

TL;DR
This paper investigates the structure of finite groups acting birationally on non-trivial Severi-Brauer surfaces, establishing bounds on their composition and providing explicit order limits when roots of unity are present.
Contribution
It proves that such groups have a normal abelian subgroup of index at most 3 and derives explicit bounds for their orders over fields containing all roots of unity.
Findings
Finite groups have a normal abelian subgroup of index ≤ 3.
Explicit bounds for group orders are provided when the field contains all roots of unity.
The results apply to non-trivial Severi-Brauer surfaces over characteristic zero fields.
Abstract
We prove that a finite group acting by birational automorphisms of a non-trivial Severi-Brauer surface over a field of characteristic zero contains a normal abelian subgroup of index at most 3. Also, we find an explicit bound for orders of such finite groups in the case when the base field contains all roots of 1.
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Birational automorphisms of Severi–Brauer surfaces
Constantin Shramov
Steklov Mathematical Institute of Russian Academy of Sciences, 8 Gubkina st., Moscow, 119991, Russia
National Research University Higher School of Economics, Laboratory of Algebraic Geometry, 6 Usacheva str., Moscow, 119048, Russia
Abstract.
We prove that a finite group acting by birational automorphisms of a non-trivial Severi–Brauer surface over a field of characteristic zero contains a normal abelian subgroup of index at most . Also, we find an explicit bound for orders of such finite groups in the case when the base field contains all roots of .
1. Introduction
A Severi–Brauer variety of dimension over a field is a variety that becomes isomorphic to the projective space of dimension over the algebraic closure of . Such varieties are in one-to-one correspondence with central simple algebras of dimension over . They have many nice geometric properties. For instance, it is known that a Severi–Brauer variety over is isomorphic to the projective space if and only if it has a -point. We refer the reader to [Art82] and [Kol16] for other basic facts concerning Severi–Brauer varieties.
Automorphism groups of Severi–Brauer varieties can be described in terms of the corresponding central simple algebras, see Theorem E on page 266 of [Châ44], or [Art82, §1.6.1], or [SV18, Lemma 4.1]. As for the group of birational automorphisms, something is known in the case of surfaces. Namely, let be a field of characteristic zero (more generally, one can assume that either is perfect, or its characteristic is different from and ). Let be a non-trivial Severi–Brauer surface over , i.e. one that is not isomorphic to . In this case generators for the group of birational automorphisms of are known, see [Wei89], or [Wei19], or [Isk96, Theorem 2.6], or Theorem 2.10 below. Moreover, relations between the generators are known as well, see [IT91, §3]. This may be thought of as an analog of the classical theorem of Noether describing the generators of the group over an algebraically closed field, and the results concerning relations between them (see [Giz83], [Isk91], [IKT94]).
Regarding finite groups acting by automorphisms or birational automorphisms on Severi–Brauer surfaces, the following result is known.
Theorem 1.1** (see [SV18, Proposition 1.9(ii),(iii)], [SV18, Corollary 1.5]).**
Let be a non-trivial Severi–Brauer surface over a field of characteristic zero. Suppose that contains all roots of . The following assertions hold.
- (i)
If is a finite subgroup, then every non-trivial element of has order , and is a -group of order at most .
- (ii)
There exists a constant such that for any finite subgroup one has .
In this paper we prove the result making Theorem 1.1 more precise.
Theorem 1.2**.**
Let be a non-trivial Severi–Brauer surface over a field of characteristic zero, and let be a finite group. The following assertions hold.
- (i)
The order of is odd.
- (ii)
The group is either abelian, or contains a normal abelian subgroup of index .
- (iii)
If contains all roots of , then is an abelian -group of order at most .
I do not know if the bounds in Theorem 1.2(ii),(iii) are optimal. In particular, I am not aware of an example of a finite non-abelian group acting by birational (or biregular, cf. Proposition 1.3 below) automorphisms on a non-trivial Severi–Brauer surface. Note that it is easy to construct an example of a non-trivial Severi–Brauer surface with an action of a group of order , see Example 4.7 below. In certain cases this is the largest finite subgroup of the automorphism group of a Severi–Brauer surface; see Lemma 4.6, which improves the result of Theorem 1.1(i). It would be interesting to obtain a complete description of finite groups acting by biregular and birational automorphisms on Severi–Brauer surfaces, cf. [DI09].
Theorem 1.2(ii) can be reformulated by saying that the Jordan constant (see e.g. [Yas17, Definition 1.1] for a definition) of the birational automorphism group of a non-trivial Severi–Brauer surface over a field of characteristic zero is at most . This shows one of the amazing differences between birational geometry of non-trivial Severi–Brauer surfaces and the projective plane, since in the latter case the corresponding Jordan constant may be much larger. For instance, if the base field is algebraically closed, the Jordan constant of the group of birational automorphisms of equals , see [Yas17, Theorem 1.9]. Moreover, by the remark made after Theorem 5.3 in [Ser09] the multiplicative analog of this constant for equals in the case of the algebraically closed field of characteristic zero, while by Theorem 1.2(ii) a similar constant for a non-trivial Severi–Brauer surface also equals either or .
To prove Theorem 1.2, we establish the following intermediate result that might be of independent interest (see also Proposition 3.7 below for a more precise statement).
Proposition 1.3**.**
Let be a non-trivial Severi–Brauer surface over a field of characteristic zero, and let be a finite non-abelian group. Then is conjugate to a subgroup of .
It would be interesting to find out if Proposition 1.3 holds for finite abelian subgroups of birational automorphism groups of non-trivial Severi–Brauer surfaces.
The plan of the paper is as follows. In §2 we study surfaces that may be birational to a non-trivial Severi–Brauer surface. In §3 we study finite groups acting on such surfaces. In §4 we prove Proposition 1.3 and Theorem 1.2.
Notation and conventions. Throughout the paper assume that all varieties are projective. We denote by the cyclic group of order , and by the symmetric group on letters.
Given a field , we denote by its algebraic closure. For a variety defined over , we denote by its scalar extension to . By a point of degree on a variety defined over some field we mean a closed point whose residue field is an extension of of degree ; a -point is a point of degree .
By a linear system on a variety over a field we mean a twisted linear subvariety of a linear system on defined over ; thus, a linear system is not a projective space in general, but a Severi–Brauer variety.
By a degree of a subvariety of a Severi–Brauer variety over a field we mean the degree of the subvariety of with respect to the hyperplane in .
For a Severi–Brauer variety corresponding to the central simple algebra , we denote by the Severi–Brauer variety corresponding to the algebra opposite to .
A del Pezzo surface is a smooth surface with an ample anticanonical class. For a del Pezzo surface , by its degree we mean its (anti)canonical degree .
Let be a variety over an algebraically closed field with an action of a group , and let be an element of . By and we denote the loci of fixed points of the group and the element on , respectively.
Acknowledgements. I am grateful to A. Trepalin and V. Vologodsky for many useful discussions. I was partially supported by the HSE University Basic Research Program, Russian Academic Excellence Project “5-100”, by the Young Russian Mathematics award, and by the Foundation for the Advancement of Theoretical Physics and Mathematics “BASIS”.
2. Birational models of Severi–Brauer surfaces
In this section we study surfaces that may be birational to a non-trivial Severi–Brauer surface.
The following general result is sometimes referred to as the theorem of Lang and Nishimura.
Theorem 2.1** (see e.g. [Kol96, Proposition IV.6.2]).**
Let and be smooth projective varieties over an arbitrary field . Suppose that is birational to . Then has a -point if and only if has a -point.
Corollary 2.2**.**
Let and be smooth projective varieties over an arbitrary field , and let be a positive integer. Suppose that is birational to , and has a point of degree not divisible by . Then has a point of degree not divisible by .
The following result concerning Severi–Brauer surfaces is well-known.
Theorem 2.3** (see e.g. [Kol16, Theorem 53(2)]).**
Let be a non-trivial Severi–Brauer surface over an arbitrary field. Then does not contain points of degree not divisible by .
Corollary 2.4**.**
Let be a non-trivial Severi–Brauer surface over an arbitrary field. Then is not birational to any conic bundle, and not birational to any del Pezzo surface of degree different from , , and .
Proof.
Suppose that is birational to a surface with a conic bundle structure . Then is a conic itself, so that has a point of degree . This implies that the surface has a point of degree or , and by Corollary 2.2 the surface has a point of degree not divisible by . This gives a contradiction with Theorem 2.3.
Now suppose that is birational to a del Pezzo surface of degree not divisible by . Then the intersection of two general elements of the anticanonical linear system is an effective zero-cycle of degree defined over the base field. Thus has a point of degree not divisible by . By Corollary 2.2 this again gives a contradiction with Theorem 2.3. ∎
Corollary 2.5**.**
Let be a non-trivial Severi–Brauer surface over an arbitrary field. Then is not birational to any del Pezzo surface of degree of Picard rank greater than , and not birational to any del Pezzo surface of degree of Picard rank greater than .
Proof.
Suppose that is birational to a del Pezzo surface as above. Then there exists a birational contraction from to a del Pezzo surface of degree and Picard rank .
If and , then , and thus . This gives a contradiction with Corollary 2.4.
If and , then is either a del Pezzo surface of degree with , which is impossible by the above argument, or a del Pezzo surface of degree not divisible by , which is impossible by Corollary 2.4. ∎
To proceed we will need the following general fact about non-trivial Severi–Brauer surfaces.
Lemma 2.6**.**
Let be a non-trivial Severi–Brauer surface over an arbitrary field , and let be a curve on . Then the degree of is divisible by .
Proof.
It is well-known (see for instance [GS18, Exercise 3.3.5(iii)]) that there exists an exact sequence of groups
[TABLE]
where is the -torsion part of the Brauer group of . Furthermore, the image of is non-trivial since the Severi–Brauer surface is non-trivial, see for instance [GS18, Exercise 3.3.4]. This means that any line bundle on has degree divisible by , and the assertion follows. ∎
Remark 2.7*.*
For an alternative proof of Lemma 2.6, one can consider the image of under a general automorphism of (see [SV18, Lemma 4.1] for a description of the automorphism group of ). Then the zero-cycle is defined over and has degree coprime to , so that the assertion follows from Theorem 2.3.
Given distinct points on over an algebraically closed field , we will say that they are in general position if no three of them are contained in a line, and all six (in the case ) are not contained in a conic (cf. [Man86, Remark IV.4.4]). If is a point of degree on a Severi–Brauer surface over a perfect field , we will say that is in general position if the points of the set are in general position. Note that if is a point of degree in general position, then the blow up of at is a del Pezzo surface of degree , see for instance [Man86, Theorem IV.2.6].
Lemma 2.8**.**
Let be a non-trivial Severi–Brauer surface over a perfect field , and let be a point of degree . Suppose that or . Then is in general position.
Proof.
Suppose that and is not in general position. Then the three points of are contained in some line on . The line is -invariant and thus defined over , which is impossible by Lemma 2.6.
Now suppose that and is not in general position. Let . If at least four of the points are contained in a line, then a line with such a property is unique, and we again get a contradiction with Lemma 2.6.
Assume that no four of the points are contained in a line, but some three of them (say, the points , , and ) are contained in a line which we denote by . The line is not -invariant by Lemma 2.6, so that there exists a line that is -conjugate to . If does not pass through any of the points , , and , then and are the only lines in that contain three of the points . This gives a contradiction with Lemma 2.6. Hence, up to relabelling the points, we may assume that , and the points and are contained in . Let be the line passing through the points and , where and . Note that are pairwise different, none of them coincides with or , and no three of them intersect at one point. If the point is not contained in any of the lines , then and are the only lines that contain three of the points , which is impossible by Lemma 2.6. If is an intersection point of two of the lines , then there are exactly four lines that contain three of the points , which is also impossible by Lemma 2.6. Thus, we see that must be contained in a unique line among , say, in . Now there are exactly three lines that contain three of the points , namely, , , and . We see that each of the points , , and is contained in two of these lines, while each of the points , , and is contained in a unique such line. However, the Galois group acts transitively on the points , which gives a contradiction.
Therefore, we may assume that the points are contained in some irreducible conic . Obviously, such a conic is unique, and thus -invariant. This again gives a contradiction with Lemma 2.6. ∎
Let be a non-trivial Severi–Brauer surface over a perfect field , and let be the corresponding central simple algebra. Recall that the Severi–Brauer surface corresponding to the algebra opposite to . There are two classes of interesting birational maps from to which we describe below (cf. [IT91, Lemma 3.1], or cases (c) and (e) of [Isk96, Theorem 2.6(ii)]).
Let be a point of degree on . Then is in general position by Lemma 2.8. Blowing up and blowing down the proper transforms of the three lines on passing through the pairs of points of , we obtain a birational map to another Severi–Brauer surface . This map is given by the linear system of conics passing through .
Similarly, let be a point of degree on . Then is in general position by Lemma 2.8. Blowing up and blowing down the proper transforms of the six conics on passing through the quintuples of points of , we obtain a birational map to another Severi–Brauer surface . This map is given by the linear system of quintics singular at (or, in other words, at each point of ).
In both of the above cases one has . This follows from a well-known general fact about the degrees of birational maps between Severi–Brauer varieties with given classes, see for instance [GS18, Exercise 3.3.7(iii)]. For more details on the maps , see [Cor05, §2].
Remark 2.9*.*
It is known that and are the only Severi–Brauer surfaces birational to , see for instance [GS18, Exercise 3.3.6(v)].
The following theorem was first published in [Wei89] (see also [Wei19]). It can be also obtained as a particular case of a much more general result, see [Isk96, Theorem 2.6]. We provide a sketch of its proof for the reader’s convenience.
Theorem 2.10**.**
Let be a non-trivial Severi–Brauer surface over a perfect field , and let be a del Pezzo surface over with . Suppose that is birational to . Then either , or . Moreover, any birational map can be written as a composition
[TABLE]
where each of the maps is either an automorphism, or a map or for some point of or .
Sketch of the proof.
Let be a birational map, and suppose that is not an isomorphism. Choose a very ample linear system on , and let be its proper transform on . Then is a mobile non-empty (and in general incomplete) linear system on . Write
[TABLE]
for some positive rational number ; note that is an integer. By the Noether–Fano inequality (see [Isk96, Lemma 1.3(i)]), one has for some point on . Let be the degree of , and let and be two general members of the linear system (defined over ). We see that
[TABLE]
and thus . Hence one has or by Theorem 2.3, and the point is in general position by Lemma 2.8.
Consider a birational map defined as follows: if , we let , and if , we let . Let . Let be the proper transform of (or ) on the surface , and write
[TABLE]
for some positive rational number such that . Using the information about provided in [IT91, Lemma 3.1], we see that . Therefore, applying the same procedure to the surface , the birational map , and the linear system and arguing by induction, we prove the theorem. ∎
A particular case of Theorem 2.10 is the following result that we will need below.
Corollary 2.11**.**
Let be a non-trivial Severi–Brauer surface over a perfect field. Then is not birational to any del Pezzo surface of degree or with .
The part of Corollary 2.11 concerning del Pezzo surfaces of degree also follows from [Man86, Chapter V]. The part concerning del Pezzo surfaces of degree can be obtained from [IT91, §2].
Corollaries 2.5 and 2.11 show that a del Pezzo surface of degree birational to a non-trivial Severi–Brauer surface must have Picard rank equal to , and a del Pezzo surface of degree birational to a non-trivial Severi–Brauer surface must have Picard rank equal to or . In the next section we will obtain further restrictions on such surfaces provided that they are -minimal with respect to some finite group .
3. -birational models of Severi–Brauer surfaces
In this section we study finite groups acting on surfaces birational to a non-trivial Severi–Brauer surface.
We start with del Pezzo surfaces of degree . Recall that over an algebraically closed field of characteristic zero a del Pezzo surface of degree is unique up to isomorphism, and its automorphism group is isomorphic to . More details on this can be found in [Dol12, Theorem 8.4.2].
Given a del Pezzo surface of degree over an arbitrary field of characteristic zero, we will call its Weyl group. For every element we will refer to the image of under the composition of the embedding with the natural homomorphism
[TABLE]
as the image of in the Weyl group. Similarly, we will consider the image of the Galois group in the Weyl group.
Let be the standard Cremona involution, that is, a birational involution acting as
[TABLE]
in some homogeneous coordinates , , and . This involution becomes regular on the del Pezzo surface of degree obtained as a blow up of at the points , , and . Let be its image in the Weyl group . Then is the generator of the center of the Weyl group. If one thinks about as the group of symmetries of a regular hexagon, then is an involution that interchanges the opposite sides of the hexagon or, in other words, a rotation by . If is a del Pezzo surface of degree over some field of characteristic zero and is its automorphism, we will say that is of Cremona type if its image in the Weyl group coincides with .
Lemma 3.1**.**
Let be a del Pezzo surface of degree over a field of characteristic zero, and let be its automorphism of Cremona type. Then is an involution that has exactly four fixed points on .
Proof.
Using a birational contraction , consider the automorphism as a birational automorphism of . Then can be represented as a composition of the standard Cremona involution with some element of the standard torus acting on . Thus, one can choose homogeneous coordinates , , and on so that acts as
[TABLE]
for some non-zero . Therefore, is conjugate to the involution via the automorphism
[TABLE]
This shows that has the same number of fixed points on as , while it is easy to see that the fixed points of the latter are the four points
[TABLE]
It remains to notice that a fixed point of on cannot be contained in a -curve, and thus all of them are mapped to fixed points of on . ∎
Lemma 3.2**.**
Let be a non-trivial Severi–Brauer surface over a field of characteristic zero, and let be a del Pezzo surface of degree over . Suppose that there exists a finite group such that . Then is not birational to .
Proof.
Suppose that is birational to . We know from Corollary 2.5 that . Hence by Corollary 2.11 one has . Since does not have a structure of a conic bundle by Corollary 2.4, we see that has a contraction on a del Pezzo surface of larger degree. Again by Corollary 2.4, this means that is a blow up of a Severi–Brauer surface at a point of degree . Hence the image of the Galois group in the Weyl group contains an element of order .
Since the cone of effective curves on has two extremal rays and , we see that must contain an element whose image in the Weyl group has order . On the other hand, the image of in the Weyl group commutes with . Since is the only element of order in that commutes with an element of order , we conclude that must contain an element of Cremona type. The element has exactly fixed points on by Lemma 3.1. This implies that there is a point of degree not divisible by on . Thus the assertion follows from Corollary 2.2 and Theorem 2.3. ∎
Now we deal with del Pezzo surfaces of degree , i.e. smooth cubic surfaces in . Recall from that for a del Pezzo surface of degree over a field the action of the groups and on -curves on defines homomorphisms of these groups to the Weyl group . Furthermore, the homomorphism is an embedding. The order of the Weyl group equals . We refer the reader to [Dol12, Theorem 8.2.40] and [Man86, Chapter IV] for details.
Lemma 3.3**.**
Let be a non-trivial Severi–Brauer surface over a field of characteristic zero, and let be a del Pezzo surface of degree over . Suppose that there exists a non-trivial automorphism such that the order of is not a power of . Then is not birational to .
Proof.
We may assume that the order is prime. Thus one has or , since the order of the Weyl group is not divisible by primes greater than . The action of on can be of one of the three types listed in [Tre16, Table 2]; in the notation of [Tre16, Table 2] these are types 1, 2, and 6. It is straightforward to check that if is of type , then the fixed point locus consists of a smooth elliptic curve and one isolated point; if is of type , then consists of a -curve and three isolated points; if is of type , then consists of a four isolated points. Note that a -invariant -curve on always contains a -point, since such a curve is a line in the anticanonical embedding of . Therefore, in each of the above three cases we find a -invariant set of points on of cardinality coprime to . Thus the assertion follows from Corollary 2.2 and Theorem 2.3. ∎
Corollary 3.4**.**
Let be a non-trivial Severi–Brauer surface over a field of characteristic zero, and let be a del Pezzo surface of degree over birational to . Suppose that there exists a subgroup such that . Then .
Proof.
We know from Corollaries 2.5 and 2.11 that either , or . Suppose that , so that the cone of effective curves on has two extremal rays. Since , the group must contain an element of even order. Therefore, the assertion follows from Lemma 3.3. ∎
Corollary 3.5**.**
Let be a non-trivial Severi–Brauer surface over a field of characteristic zero, and let be a del Pezzo surface of degree over birational to . Let be a subgroup such that . Then is isomorphic to a subgroup of .
Proof.
We know from Corollary 3.4 that . Hence is a blow up of a Severi–Brauer surface at two points of degree . This means that the image of the Galois group in the Weyl group contains an element conjugate to
[TABLE]
in the notation of [Tre16, §4]. We may assume that contains itself, so that the image of in is contained in the centralizer of . By [Tre16, Proposition 4.5] one has
[TABLE]
On the other hand, we know from Lemma 3.3 that the order of is a power of . Since the Sylow -subgroup of is isomorphic to , the required assertion follows. ∎
Remark 3.6*.*
For an alternative proof of Corollary 3.5, suppose that is a non-abelian -group. One can notice that in this case the image of in is either trivial, or is generated by an element of type in the notation of [Car72]. In the former case , which is impossible by Corollary 2.5. In the latter case , which is impossible by Corollary 2.11. Thus, is an abelian -group. The rest is more or less straightforward, cf. [DI09, Theorem 6.14].
Let us summarize the results of this section.
Proposition 3.7**.**
Let be a non-trivial Severi–Brauer surface over a field of characteristic zero, and let be a finite subgroup. Then is conjugate either to a subgroup of , or to a subgroup of , or to a subgroup of , where is a del Pezzo surface of degree over birational to such that and . In the latter case is isomorphic to a subgroup of .
Proof.
Regularizing the action of and running a -Minimal Model Program (see [Isk80, Theorem 1G]), we obtain a -surface birational to , such that is either a del Pezzo surface with , or a conic bundle. The case of a conic bundle is impossible by Corollary 2.4. Thus by Corollary 2.4 and Lemma 3.2 we conclude that is a del Pezzo surface of degree or . In the former case is a Severi–Brauer surface itself, so that is isomorphic either to or to by Remark 2.9. In the latter case we have by Corollary 3.4. Furthermore, in this case is isomorphic to a subgroup of by Corollary 3.5. ∎
I do not know the answer to the following question.
Question 3.8**.**
Does there exist an example of a finite abelian group acting on a smooth cubic surface over a field of characteristic zero, such that is birational to a non-trivial Severi–Brauer surface and ? In other words, does there exist a non-trivial Severi–Brauer surface over a field of characteristic zero and a finite abelian group , such that the action of can be regularized on some smooth cubic surface, but is not conjugate to a subgroup of ?
4. Automorphisms of Severi–Brauer surfaces
In this section we prove Proposition 1.3 and Theorem 1.2. We start with a couple of simple auxiliary results.
Let denote the Heisenberg group of order ; this is the only non-abelian group of order and exponent . Its center is isomorphic to , and there is a non-split exact sequence
[TABLE]
On the other hand, one can also represent as a semi-direct product .
Lemma 4.1**.**
Let be an algebraically closed field of characteristic zero, and let
[TABLE]
be a finite subgroup. The following assertions hold.
- (i)
If the order of is odd, then is either abelian, or contains a normal abelian subgroup of index .
- (ii)
If the order of is odd and is non-trivial, then contains a subgroup such that either , or has a unique isolated point. Moreover, one can choose with such a property so that either , or is a normal subgroup of index in .
- (iii)
If , then contains an element such that has a unique isolated point.
- (iv)
The group is not isomorphic to .
Proof.
Let be the preimage of under the natural projection
[TABLE]
and let be the corresponding three-dimensional representation of . We will use the classification of finite subgroups of , see for instance [Bli17, Chapter V].
Suppose that the order of is odd. Then it follows from the classification that either is abelian, so that splits as a sum of three one-dimensional -representations such that not all of them are isomorphic to each other; or is non-abelian and there exists a surjective homomorphism whose kernel is an abelian group, so that splits as a sum of three one-dimensional -representations, and transitively permutes these -representations. In other words, the group cannot be primitive (see [Bli17, §60] for the terminology), and cannot split as a sum of a one-dimensional and an irreducible two-dimensional -representation. This proves assertion (i).
If is abelian, let . Thus, if is odd and is non-trivial, in all possible cases we see that the group is a group with the properties required in assertion (ii).
Suppose that . Then it follows from the classification (cf. [Bor61, 6.4]) that
[TABLE]
and . The elements of can be simultaneously diagonalized. Hence the image of in contains an element such that consists of a line and an isolated point. This proves assertion (iii).
Assertion (iv) directly follows from the classification (cf. [Bor61, 6.4]). ∎
Most of our remaining arguments are based on the following observation.
Lemma 4.2**.**
Let be a non-trivial Severi–Brauer surface over a field of characteristic zero, and let be a finite subgroup. Then the order of is odd.
Proof.
Suppose that the order of is even. Then contains an element of order . Consider the action of on . The fixed point locus is a union of a line and a unique isolated point . Since the Galois group commutes with , the point is -invariant, which is impossible by assumption. ∎
Corollary 4.3**.**
Let be a non-trivial Severi–Brauer surface over a field of characteristic zero, and let be a finite subgroup. Then
- (i)
the group is either abelian, or contains a normal abelian subgroup of index ;
- (ii)
there exists a -invariant point of degree on .
Proof.
By Lemma 4.2, the order of is odd. The action of on gives an embedding . By Lemma 4.1(i) the group is either abelian, or contains a normal abelian subgroup of index . This proves assertion (i).
To prove assertion (ii), we may assume that is non-trivial. By Lemma 4.1(ii), the group contains a subgroup such that either , or has a unique isolated point; moreover, one can choose with such a property so that either coincides with , or is a normal subgroup of index in . In any case, cannot have a unique isolated point, because this point would be -invariant, which is impossible by assumption. Hence . Since is a normal subgroup in , the set is -invariant. This gives a -invariant point of degree on and proves assertion (ii). ∎
Remark 4.4*.*
It is interesting to note that the analogs of Lemma 4.2 and Corollary 4.3 do not hold for Severi–Brauer curves, i.e. for conics. Thus, a conic over the field of real numbers defined by the equation
[TABLE]
in with homogeneous coordinates , , and has no -points. However, it is acted on by all finite groups embeddable into , that is, by cyclic groups, dihedral groups, the tetrahedral group , the octahedral group , and the icosahedral group . From this point of view finite groups acting on non-trivial Severi–Brauer curves are more complicated than those acting on non-trivial Severi–Brauer surfaces. It would be interesting to obtain a complete classification of finite groups acting on Severi–Brauer surfaces similarly to what is done for conics in [GA13].
Recall from §2 that a Severi–Brauer surface is birational to the surface . In particular, the groups and are (non-canonically) isomorphic, and the group is (non-canonically) realized as a subgroup of . Note also that , although these groups are not conjugate in .
Corollary 4.3 has the following geometric consequence.
Corollary 4.5**.**
Let be a non-trivial Severi–Brauer surface over a field of characteristic zero, and let be a finite subgroup. Then is conjugate to a subgroup of .
Proof.
By Corollary 4.3(ii) the group has an invariant point of degree on , and by Lemma 2.8 this point is in general position. Blowing up and blowing down the proper transforms of the three lines on passing through the pairs of the three points of , we obtain a (regular) action of on the surface together with a -equivariant birational map , cf. §2. This means that is conjugate to a subgroup of . ∎
Now we prove Proposition 1.3.
Proof of Proposition 1.3.
We know from Proposition 3.7 that is conjugate to a subgroup of , where or . In the former case we are done. In the latter case is conjugate to a subgroup of by Corollary 4.5. ∎
Similarly to Lemma 4.2, we prove the following.
Lemma 4.6**.**
Let be a field of characteristic zero that contains all roots of . Let be a non-trivial Severi–Brauer surface over , and let be a finite subgroup. Then is isomorphic to a subgroup of .
Proof.
We know from Theorem 1.1(i) that every non-trivial element of has order , and . Assume that is not isomorphic to a subgroup of . Then either or . The former case is impossible by Lemma 4.1(iv). Thus, we have . By Lemma 4.1(iii) the group contains an element such that has a unique isolated point. This latter point must be -invariant, which is impossible by assumption. ∎
Finally, we prove our main result.
Proof of Theorem 1.2.
Assertion (i) follows from Proposition 3.7 and Lemma 4.2. Assertion (ii) follows from Proposition 1.3 and Corollary 4.3(i). Assertion (iii) follows from Proposition 3.7 and Lemma 4.6. ∎
I do not know if the bound provided by Theorem 1.2(iii) (or Lemma 4.6) is optimal. However, in certain cases it is easy to construct non-trivial Severi–Brauer surfaces with an action of the group .
Example 4.7**.**
Let be a field of characteristic different from that contains a primitive cubic root of unity . Let be the elements such that is not a cube in , and is not contained in the image of the Galois norm for the field extension . Consider the algebra over generated by variables and subject to relations
[TABLE]
Then is a central division algebra, see for instance [GS18, Exercise 3.1.6(ii),(iv)]. One has , so that corresponds to a non-trivial Severi–Brauer surface . Conjugation by defines an automorphism of order of (sending to and to ). Together with conjugation by it generates a group acting by automorphisms of and .
The reference list from the paper itself. Each links out to its DOI / PubMed record.
- 1[Art 82] M. Artin. Brauer–Severi varieties. In Brauer groups in ring theory and algebraic geometry (Wilrijk, 1981) , volume 917 of Lecture Notes in Math. , pages 194–210. Springer, Berlin-New York, 1982.
- 2[Bli 17] H. F. Blichfeldt. Finite collineation groups . University of Chicago, Chicago, 1917.
- 3[Bor 61] A. Borel. Sous-groupes commutatifs et torsion des groupes de Lie compacts connexes. Tôhoku Math. J. (2) , 13:216–240, 1961.
- 4[Car 72] R. W. Carter. Conjugacy classes in the Weyl group. Compositio Math. , 25:1–59, 1972.
- 5[Châ44] F. Châtelet. Variations sur un thème de H. Poincaré. Ann. Sci. École Norm. Sup. (3) , 61:249–300, 1944.
- 6[Cor 05] P. Corn. Del Pezzo surfaces of degree 6 6 6 . Math. Res. Lett. , 12(1):75–84, 2005.
- 7[DI 09] I. V. Dolgachev and V. A. Iskovskikh. Finite subgroups of the plane Cremona group. In Algebra, arithmetic, and geometry: in honor of Yu. I. Manin. Vol. I , volume 269 of Progr. Math. , pages 443–548. Birkhäuser Boston, Inc., Boston, MA, 2009.
- 8[Dol 12] I. V. Dolgachev. Classical algebraic geometry . Cambridge University Press, Cambridge, 2012.
