Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

On the motivic class of an algebraic group

On the motivic class of an algebraic group FEDERICO SCAVIA Abstract. Let F be a field of characteristic zero admitting a biquadratic field extension. We give an example of a torus G over F whose classifying stack −1 BG is stably rational and such that {BG} 6= {G} in the Grothendieck ring of algebraic stacks over F . We also give an example of a finite ´etale group scheme A over F such that BA is stably rational and {BA} 6= 1. 1. Introduction Let F be a field. The Grothendieck ring of algebraic stacks K (Stacks ) was 0 F introduced by Ekedahl in [8], following up on earlier works [1], [14], [20]. It is a variant of the Grothendieck ring of varieties K (Var ). By definition, K (Stacks ) 0 F 0 F is generated as an abelian group by the equivalence classes {X} of all algebraic stacks X of finite type over F with affine stabilizers. These classes are subject to the scissor relations {X} = {Y } + {X \ Y } for every closed substack Y ⊆ X, and the relations {E} = {A × X} for every vector bundle E of rank n over X. The product is defined by {X} · {Y } := {X × Y }, and extended by linearity. Given a group scheme G over F , we may consider the class {BG} of its classifying stack in K (Stacks ). The problem of computing {BG} appears to be related to the 0 F problem of the stable rationality of BG, although no direct implications are known. Recall that BG is stably rational if for one (equivalently, every) generically free representation V of G, the rational quotient V/G is stably rational. An equivalent terminology is that the Noether problem for stable rationality has a positive solution for G; see [12, §3]. The case of a finite (constant) group G was considered in [7]: it frequently happens that {BG} = 1 (notably for the symmetric groups, see [7, Theorem 4.3]), although there are examples of finite groups G for which {BG} = 6 1; see [7, Corollary 5.2, Corollary 5.8]. Further work on the triviality of {BG} for finite groups G has been done in [16] and [17]. So far, all the known examples of finite group schemes G for which {BG} 6= 1 are such that BG is not stably rational. This suggests the following question. Question 1.1. (cf. [7, §6]) Is it true that, for a finite group scheme G, the following two conditions are equivalent? • BG is stably rational; • {BG} = 1 in K (Stacks ). 0 F We will answer Question 1.1 in the negative in Theorem 1.6. Now let G be a connected linear algebraic group. Recall that G is special if every G-torsor is Zariski-locally trivial. For example, GL , SL and Sp are special; see n n [5]. It was shown by Ekedahl that if P → S is a torsor under the special group G, then {P} = {G}{S}. This is immediate if S is a scheme, but less obvious arXiv:1808.00056v2 [math.AG] 30 Dec 2020 2 FEDERICO SCAVIA when S is a stack; see [3, Corollary 2.4]. Applying this to the universal G-torsor Spec F → BG, one obtains {BG}{G} = 1. −1 The equality {BG} = {G} appears to be the analogue for connected groups of the relation {BG} = 1 for finite group schemes. In [3], these equalities are referred to as expected class formulas, and there is a sense in which they are “almost” true. In [8, §2] Ekedahl defines a generalized Euler characteristic χ : K (Stacks ) → K (Coh ) c 0 F 0 F taking values in a Grothendieck ring K (Coh ) of Galois representations over F . If 0 F G is a finite group scheme, the equality χ ({BG}) = 1 always holds [7, Proposition 3.1]. On the other hand, if G is connected, then χ ({BG}{G}) = 1; see [3, §2.2]. Since {BG} 6= 1 for some finite groups G, the following question naturally arises. Question 1.2. Let F be a field. Is it true that −1 (1.3) {BG} = {G} in K (Stacks ) for every connected group G? 0 F In Theorem 1.5, we show that the answer to Question 1.2 is also negative. Com- putations for non-special G have been carried out for PGL and PGL in [3], for 2 3 SO and n odd in [6], for SO and n even and O for any n in [19], and for n n n Spin , Spin and G in [18]. In each of these cases, (1.3) was found to be true. The 7 8 expectation was that, for a connected linear algebraic group G over a field F of characteristic 0, Question 1.4 below should have an affirmative answer. If F is an algebraically closed field, then there are no examples of connected G where BG is known not to be stably rational. If F is not assumed to be algebraically closed, then such examples exist. The following variant of Question 1.1 seems natural in this context. Question 1.4. (cf. [19, §1] and [18, Remark 4.1]) Is it true that, for a connected linear algebraic group G, the following two conditions are equivalent? • BG is stably rational; −1 • {BG} = {G} in K (Stacks ). 0 F Our first result gives a negative answer to Question 1.2 and Question 1.4. Theorem 1.5. Let F be a field of characteristic zero which admits a biquadratic field extension K, let E and E be two distinct quadratic subextensions of K/F , 1 2 (1) and set G := R (G ). Then E ×E /F 1 2 (a) BG is stably rational, and −1 (b) {BG} 6= {G} in K (Stacks ). 0 F The torus G is an example of a norm-one torus; see Section 2 for the definition. It follows from Theorem 1.5 that counterexamples H to (1.3) exist in any dimension dim H ≥ 3: consider for example H := G × G for r ≥ 0. The key ingredient in the proof of Theorem 1.5 is the refined Euler characteristic of Ekedahl, introduced in [8, §6, 3]; see Section 4. Our second result gives a negative answer to Question 1.1. Theorem 1.6. Let F be a field of characteristic zero which admits a biquadratic field extension K, and let E and E be two distinct quadratic subextensions of 1 2 (1) K/F . Define G := R (G ), and let A := G[2] be the 2-torsion subgroup of E ×E /F 1 2 G. Then ON THE MOTIVIC CLASS OF AN ALGEBRAIC GROUP 3 (a) BA is stably rational, and (b) {BA} 6= 1 in K (Stacks ). 0 F Questions 1.1, 1.2 and 1.4 remain open in the case, where the base field F is assumed to be algebraically closed. Our arguments do not shed any new light in this setting. The remainder of this paper is structured as follows. In Section 2 we review well known computations of motivic classes for non-split tori. In Section 3 we obtain explicit formulas for the motivic classes of G and BG, and in Section 4 we give the required background on the refined Euler characteristic. In Section 5 we prove Theorem 1.5, and in Section 6 we prove Theorem 1.6. 2. Preliminaries Let F be a field. We will write L for the class {A } in K (Var ) or K (Stacks ). 0 F 0 F If E is an ´etale algebra over F , we will denote by {E} the class {Spec E} in K (Var ) or K (Stacks ). If X is a quasi-projective scheme over E, we will denote 0 F 0 F by R (X) the Weil restriction of X to F . By definition, for every F -scheme S E/F one has R (X)(S) = X(S ). We refer the reader to [21, §3.12] for an account E/F E of the main properties of the Weil restriction. Let G be a linear algebraic group over F , and α ∈ H (F, G) be represented by a G-torsor P → Spec F . For every quasi-projective F -scheme Z, we denote by Z the twist of Z by P , that is, Z := (Y × P )/G, where G acts diagonally. We refer the reader to [10, Section 2] for the definition and the basic properties of the twisting operation. We will write C for the cyclic group of two elements, and S for the symmetric 2 n group on n symbols. The following observations will be repeatedly used in the sequel. Lemma 2.1. Let X be a scheme over F , E an ´etale algebra of degree n over F , α ∈ H (F, S ) the class corresponding to E/F . (a) Let S act on the disjoint union ∐ X by permuting the n copies of X. i=1 Then (∐ X) = X . i=1 (b) Let S act on X by permuting the n factors. Then (X ) = R (X). E/F Proof. (a) Let Y := ∐ X, and let S act on Y by permuting the copies of X. By i=1 definition, Y = (Y × Spec E)/S (Y × X )/S , n X E n where S acts diagonally. This shows that Y is the twist of X by the trivial n E S -torsor Y → X in the category of X-schemes, which implies Y = X . n E (b) See the bottom of page 5 in [11]. Lemma 2.2. Let 1 → N → G → H → 1 be an exact sequence of group schemes over F , and assume that G is special. Then {BN} = {H}/{G}. 4 FEDERICO SCAVIA Proof. See [3, Proposition 2.9]. Let F be a separable closure of F . Recall that a group scheme T over F is called a torus if T = G for some n ≥ 0. The character lattice of T is the finitely m,F ,G ). The character lattice induces generated Z-free Gal(F )-module Hom (T F F m,F s s s an anti-equivalence between the category of F -tori and the category of Gal(F )- lattices, i.e., Z-free continuous Gal(F )-modules; see [9, §2]. Similarly, for every separable finite extension L/F , we have an anti-equivalence between Gal(L/F )- lattices and F -tori T split by L, i.e., such that T G for some n ≥ 0. The m,L dual torus of T is the torus T whose character lattice is dual to that of T . Let E be an ´etale algebra over F . If G is a group scheme over E, then R (G) is E/F a group scheme over F . The group R (G ) := R (G ) is an F -torus. Tori m m,E E/F E/F of this kind are called quasi-split. They are special groups, and they correspond to permutation Gal(F )-lattices, that is, lattices admitting a Z-basis that is permuted by Gal(F ); see [21, §3.12, Example 19]. Lemma 2.3. Let T be an algebraic torus over F , and let T be its dual. Assume that T is stably rational. Then (a) BT is stably rational; (b) {BT }{T} = 1 in K (Stacks ). 0 F Proof. Since T is stably rational, by [21, §4.7, Theorem 2] there is a short exact sequence (2.4) 1 → T → T → T → 1 1 2 where T and T are quasi-split. Since quasi-split tori are isomorphic to their dual, 1 2 the sequence dual to (2.4), (2.5) 1 → T → T → T → 1, 2 1 shows that T embeds in T . We may view T as a maximal torus inside GL , where 2 2 n n = rank T . This gives a faithful representation of T with quotient birational to T . Since quasi-split tori are rational, it follows that BT is stably rational. Quasi-split tori are special, so we may apply Lemma 2.2 to (2.4) and (2.5). We ′ ′ obtain {T} = {T }/{T } and {BT } = {T }/{T }, so {BT }{T} = 1. 2 1 1 2 Let E/F be an ´etale algebra, and let R (G ) be the associated quasi-split E/F m torus. The kernel of the norm homomorphism R (G ) → G is called a E/F m m (1) norm-one torus, and is denoted by R (G ). Its dual torus is isomorphic to E/F R (G )/ G . m m E/F Lemma 2.6. Assume that char F 6= 2. Let E := F ( m) be a separable quadratic field extension, and let α denote the class of E/F in H (F, C ). Then: (1) (a) R (G ) R (G )/ G . m E/F m m E/F 1 −1 1 1 (b) Let Gal(E/F ) act on P via z 7→ z . Then P P . (c) R (G )/ G is rational and m m E/F −1 {R (G )/ G } = {B(R (G )/ G )} = L − {E} + 1. E/F m m E/F m m −1 (d) {R (G )} = {BR (G )} = (L − 1)(L − {E} + 1). m m E/F E/F 1 2 (e) {R (P )} = L + {E}L + 1. E/F ON THE MOTIVIC CLASS OF AN ALGEBRAIC GROUP 5 Proof. (a) Both tori correspond to the unique non-trivial Gal(E/F )-lattice of rank 1. Here Gal(E/F ) = C . 1 1 (b) The C -action on P has a fixed point z = 1, hence P has an F -point. By Chaˆtelet’s Theorem [13, Theorem 5.1.3], a form of P which admits an F -point is trivial (the case n = 1 is particularly simple, see [13, Remark 1.3.5]). We conclude 1 1 that P = P . (1) (c) Let T := R (G ) = R (G )/ G . The open embedding G ֒→ P , m E/F m m m E/F as the complement of Z := {0, ∞}, is equivariant under the C -action on G and 2 m 1 −1 P given by z 7→ z . Twisting by α, we obtain by (b) an open embedding of T 1 α in P as the complement of Z. In particular, T is rational. By Lemma 2.1(a), Z = Spec E, so 1 α {T} = {P } − { Z} = L + 1 − {E}. Now (c) follows from Lemma 2.3(b). (d) The first equality holds because R (G ) is special. Consider the short E/F exact sequence 1 → G → R (G ) → T → 1. m E/F m Since R (G ) is special, Lemma 2.2 yields E/F −1 {R (G )} = (L − 1){BT} , E/F m thus (d) follows from (c). 1 1 (e) Write P = A ∪ {∞}, and consider the C -equivariant decomposition 1 2 1 2 1 1 (P ) = (A ) ∐ (A × {∞} ∪ {∞} × A ) ∐ {(∞, ∞)}. By Hilbert’s Theorem 90 and Lemma 2.1(a), twisting by α gives 1 2 1 R (P ) = A ∐ A ∐ Spec F, E/F 1 2 thus {R (P )} = L + {E}L + 1. E/F 3. The classes of G and BG Let F be a field of characteristic not 2, and assume that there exists a biquadratic extension √ √ K := F ( m , m ) 1 2 of F . Let √ √ √ E := F ( m ), E := F ( m ), E := F ( m m ), E := E × E , 1 1 2 2 12 1 2 1 2 and let Γ := Gal(K/F ) = C be the Galois group of K/F . We define the torus (1) G := R (G ) E/F and let G := R (G )/ G E/F m m be the dual torus of G. By definition, we have a short exact sequence (3.1) 1 → G → R (G ) −→ G → 1, m m E/F where N is the norm homomorphism. The purpose of this section is the proof of Proposition 3.7, which expresses {BG} and {G} as rational functions in L, with coefficients classes of ´etale algebras. 6 FEDERICO SCAVIA σ σ 1 2 Let σ and σ be generators for Γ such that E = K and E = K . Consider 1 2 1 2 2 −1 −1 the Γ-action on G , where σ (u, v) = (v , u ) and σ (u, v) = (v, u), and set 1 2 (3.2) T := (G ), where α ∈ H (F, Γ) corresponds to the extension K/F . Lemma 3.3. We have {T} = L + ({E } − {K})L + {K} − {E } − {E } + 1. 12 1 2 Proof. The embedding of G in P as the complement of Z := {0, ∞} gives an 2 1 2 2 1 2 open embedding G ֒→ (P ) such that the Γ-action on G extends to (P ) . By m m definition 1 2 1 2 (P ) = ((P ) × Spec K)/Γ, where Γ = hσ , σ i acts diagonally. We first take the quotient by the subgroup 1 2 −1 −1 σ σ 1 2 hσ σ i. Since σ σ (u, v) = (u , v ) and E = K , by Lemma 2.6(b) 1 2 1 2 12 1 2 1 2 (P ) = ((P ) × Spec E )/C , 12 2 1 2 where C acts on (P ) by switching the two factors. Here we are using the fact 1 2 that every automorphism of (P ) must respect the ruling (because it respects 1 2 1 2 the intersection form), and so Aut((P ) ) = (Aut(P )) ⋊ C , where C switches 2 2 1 2 1 the two factors. By Lemma 2.1(b) we deduce that (P ) = R (P ), so by E /F Lemma 2.6(e) 1 2 2 (3.4) { (P ) } = L + {E }L + 1. 1 2 2 We may partition (P ) \ G in two strata Z := Z × Z, Z := (Z × G ) ∐ (G ×Z). 1 2 m m The Γ-action on Z has two orbits, and Γ acts on Z by transitively permuting 1 2 the components as the Klein subgroup of S . By Lemma 2.1(a), Z = Spec E ∐ 4 1 1 Spec E and Z = G × Spec K. By (3.4) 2 2 m 1 2 α α {T} = { (P ) } − { Z } − { Z } 1 2 = L + {E }L + 1 − {E } − {E } − {K}(L − 1) 12 1 2 = L + ({E } − {K})L + {K} − {E } − {E } + 1. 12 1 2 Proposition 3.5. There is a short exact sequence of tori 1 → G → G → T → 1, where T is the torus of (3.2). Proof. Let P , M and Z be the character lattices of R (G ), G and G , respec- m m E/F tively. We may view P as the Γ-lattice with a basis e , e , e , e , such that σ acts 1 2 3 4 1 by switching e with e and fixing e and e , and σ switches e with e and fixes 1 2 3 4 2 3 4 e and e . The sequence of Γ-lattices dual to (3.1) identifies M with the cokernel 1 2 of the Γ-homomorphism Z → P given by 1 7→ e + e + e + e ; denote by e ∈ M 1 2 3 4 i the projection of e . Following Kunyavski˘ı [15, §3, Proposition 1(b)], we consider an exact sequence of Γ-lattices (3.6) 0 → N → M −→ Z → 0. The map π is defined by π( a e ) = a + a − a − a , and N := Ker π. A basis i i 1 2 3 4 for N is given by v := e +e and v := e +e . With respect to the basis (v , v ), 1 1 3 2 1 4 1 2 ON THE MOTIVIC CLASS OF AN ALGEBRAIC GROUP 7 the Γ-action on N is given by σ (a, b) = (−b, −a) and σ (a, b) = (b, a). It is now 1 2 clear that N is the character lattice of the torus T of (3.2), hence the proof is complete. Proposition 3.7. (a) BG is stably rational. (b) {BG}{G } = 1 in K (Stacks ). 0 F Proof. Consider the sequence (3.8) 1 → G → G → (R (G )/ G ) × (R (G )/ G ) → 1, m E /F m m E /F m m 1 2 which exhibits G as a G -torsor over a rational variety, by Lemma 2.6(c). We deduce that G is rational, and now (a) and (b) follow from Lemma 2.3. Proposition 3.9. We have (3.10) {G} = (L − 1)(L + ({E } − {K})L + {K} − {E } − {E } + 1) 12 1 2 and −1 (3.11) {BG} = (L − 1)(L − {E } + 1)(L − {E } + 1) 1 2 in K (Stacks ). 0 F Proof. By Proposition 3.5, G is a G -torsor over T . Since G is special, {G} = m m (L − 1){T}. The class of T was determined in Lemma 3.3. −1 ′ ′ By Proposition 3.7(b), {BG} = {G }. Since G is special, by (3.8), {G } = (1) (1) (L − 1){R (G )}{R (G )}. Now (3.11) follows from Lemma 2.6(c). m m E /F E /F 1 2 4. The refined Euler characteristic Let F be a field of characteristic zero. Using the computations of the previous section, we will reduce Theorem 1.5(b) to the assertion that a certain polynomial in L with coefficients motivic classes of ´etale algebras is a non-zero element of K (Var ). To prove the assertion, we will use a simplified version of the refined 0 F Euler characteristic, introduced by Ekedahl in [8]. Fix a prime number p, and let G be a profinite group. The representation ring a (G) of G is the Grothendieck ring of continuous G-representations [M] of finite dimension over F , subject to the relations [M ⊕ N] = [M] + [N]. Note that no relations for non-split short exact sequences are imposed. The product structure on a (G) is given by tensor product of representations. The next observation is well known when G is assumed to be finite; see [2, §5.1]. Lemma 4.1. As an abelian group, a (G) is freely generated by the set of isomor- phism classes of indecomposable representations. Proof. It is clear that a (G) is generated by isomorphism classes of indecomposable P P representations. Assume that a [M ] − b [N ] = 0 in a (G), for some positive i i j j p integers a , b and some pairwise non-isomorphic indecomposable G-representations i j M and N . i j As a group, a (G) is the quotient group F/I, where F is the free abelian group with one generator hPi for every isomorphism class of G-representations P , and I is the subgroup generated by all elements of the form hP ⊕ Qi−hPi−hQi. It follows that we may find a G-representation X such that ⊕b ⊕a j (⊕ M ) ⊕ X (⊕ N ) ⊕ X. i j i j 8 FEDERICO SCAVIA Let G be a finite quotient of G such that G acts on M , N and X through G . 0 i j 0 Then M ⊕ X = N ⊕ X as G -representations. By the Krull-Schmidt Theorem applied to the group algebra F [G ], this implies M N as G -modules, hence as p 0 0 G-modules. This is impossible, because the indecomposable representations M and N are pairwise non-isomorphic. Proposition 4.2. Let F be a field of characteristic zero, let Gal(F ) be the absolute Galois group of F , and let R := a (Gal(F )). There is a ring homomorphism p p µ : K (Var ) → R [t] 0 F p i i such that for every smooth complete variety X we have µ (X) = [H (X ,F )]t . ´ et p Proof. See the proof of [8, Proposition 3.2(i)]. To show that µ is well-defined, one needs to assume that charF = 0 in order to invoke Bittner’s presentation of K (Var ); see [4, Theorem 3.1]. 0 F 5. Proof of Theorem 1.5 Theorem 1.5(a) was proved in Proposition 3.7(b), so we will focus on Theo- rem 1.5(b). We maintain the notation given at the beginning of Section 3. (1) Proof of Theorem 1.5(b). Assume by contradiction that G = R (G ) satisfies E/F (1.3). Then by Proposition 3.9 we have (L − 1)(L − {E } + 1)(L − {E } + 1) = 1 2 = (L − 1)(L + ({E } − {K})L + {K} − {E } − {E } + 1) 12 1 2 in K (Stacks ). Since L − 1 is invertible in K (Stacks ), we may divide by L − 1 0 F 0 F on both sides. Subtracting L on the left and on the right, we arrive to (2−{E }−{E })L+(1−{E })(1−{E }) = ({E }−{K})L+{K}−{E }−{E }+1, 1 2 1 2 12 1 2 that is ({K} − {E } − {E } − {E } + 2)L = 0 1 2 12 in K (Stacks ). 0 F Recall that K (Stacks ) is the localization of K (Var ) at L and the cyclotomic 0 F 0 F polynomials in L; see [8, Theorem 1.2]. It follows that (5.1) ({K} − {E } − {E } − {E } + 2)f(L) = 0 1 2 12 in K (Var ), where f(x) ∈ Z[x] is a monic polynomial of some degree n. 0 F In order to obtain a contradiction, we now want to apply the homomorphism µ of (4.2), with respect to the prime p = 2. If L/F is an ´etale algebra of degree n, µ ({L}) consists of the permutation representation of Gal(F ) associated to L, concentrated in degree 0. Since we have chosen p = 2, µ ({P }) consists of one copy of the trivial representation in degree 0 and 2 (in the case p > 2 one would need a 1 2 Tate twist in degree 2). Since L = {P } − 1, we deduce that µ (L) = t , and hence µ (f(L)) = f(t ). If X is a finite Gal(F )-set, we denote by F [X] the permutation representation over F associated to X. Recall from Section 3 that we denote Gal(K/F ) by Γ = hσ , σ i. Applying µ to (5.1) and looking at degree 2n, we obtain 1 2 [F [Γ]] − [F [Γ/ hσ i]] − [F [Γ/ hσ i]] − [F [Γ/ hσ i]] + 2[F ] = 0 2 2 1 2 2 2 12 2 ON THE MOTIVIC CLASS OF AN ALGEBRAIC GROUP 9 in R . This is a non-trivial relation of linear dependence in R among classes of 2 2 indecomposable representations. This is in contradiction with Lemma 4.1, hence −1 {BG} 6= {G} , as desired. Remark 5.2. By [21, §4.9, Example 7] every torus of rank 2 is rational, so by Proposition 3.5 the torus G is rational. By Lemma 2.3, BG is stably rational and ′ −1 ′ −1 ′ ′ {BG } = {G} . By Proposition 3.7(b) we have {BG} = {G } , so {BG }{G } = −1 −1 {BG} {G} . Since {BG}{G} 6= 1, the conclusions of Theorem 1.5(a) and (b) hold for G as well. 6. Proof of Theorem 1.6 We maintain the notation of Section 3. Proof of Theorem 1.6. Let Γ := Gal(K/F ), let M be the character lattice of G, so that M/2M is the character module of A, and let P be the character lattice of R (G ). As in the proof of Proposition 3.5, we view P as the lattice freely E/F m generated by e , e , e , e , such that σ acts by switching e with e , and σ by 1 2 3 4 1 1 2 2 switching e with e . Using (3.1), we may construct a commutative diagram of 3 4 Γ-modules 0 Z P M 0 (6.1) ι 0 N P M/2M 0. with exact rows. Here Z denotes the trivial one-dimensional Γ-lattice, ι(1) := e + e + e + e , and N is the kernel of ϕ, that is, 1 2 3 4 N = { a e : a ≡ a ≡ a ≡ a (mod 2)}. i i 1 2 3 4 i=1 Applying the snake lemma to (6.1), we obtain a short exact sequence 0 → Z − → N → M → 0. Define π : N → Z by sending a e to (a + a )/2. Then π is a Γ-homomorphism i i 1 2 and ι is a section of π. Therefore, we have an isomorphism N Z ⊕ M. Let S be an F -torus with character lattice N. Since N Z ⊕ M, we have S G × G. The bottom row of (6.1) corresponds to the short exact sequence of group schemes 1 → A → R (G ) → G × G → 1. E/F m m By Lemma 2.2, we have {BA} = {G }{G}/{R (G )}. Applying Lemma 2.2 m E/F m to (3.1), we see that {BG} = {G }/{R (G )}. Therefore, {BA} = {BG}{G}. m E/F m −1 By Theorem 1.6 we have {BG} 6= {G} , hence {BA} 6= 1, as desired. Acknowledgments I would like to thank my advisor Zinovy Reichstein for his guidance and for greatly improving the exposition, Mattia Talpo and Angelo Vistoli for helpful com- ments, and Boris Kunyavski˘ı for sending me a copy of his paper [15]. I am very grateful to the anonymous referee for finding a mistake in a previous version of the proof of Theorem 1.5, and for suggesting a fix. 10 FEDERICO SCAVIA References [1] Kai Behrend and Ajneet Dhillon. On the motivic class of the stack of bundles. Advances in Mathematics, 212(2):617–644, 2007. [2] D. J. Benson. Representations and cohomology. II, volume 31 of Cambridge Studies in Ad- vanced Mathematics. Cambridge University Press, Cambridge, second edition, 1998. Coho- mology of groups and modules. [3] Daniel Bergh. Motivic classes of some classifying stacks. Journal of the London Mathematical Society, 93(1):219–243, 2015. [4] Franziska Bittner. The universal Euler characteristic for varieties of characteristic zero. Com- positio Mathematica, 140(4):1011–1032, 2004. [5] S´eminaire Claude Chevalley and JP Serre. Espaces fibr´es alg´ebriques. S´eminaire Claude Chevalley, 3:1–37, 1958. [6] Ajneet Dhillon, Matthew B Young, et al. The motive of the classifying stack of the orthogonal group. The Michigan Mathematical Journal, 65(1):189–197, 2016. [7] Torsten Ekedahl. A geometric invariant of a finite group. arXiv preprint arXiv:0903.3148, [8] Torsten Ekedahl. The Grothendieck group of algebraic stacks. arXiv preprint arXiv:0903.3143, 2009. [9] Giordano Favi and Mathieu Florence. Tori and essential dimension. J. Algebra, 319(9):3885– 3900, 2008. [10] Mathieu Florence. On the essential dimension of cyclic p-groups. Invent. Math., 171(1):175– 189, 2008. [11] Mathieu Florence and Zinovy Reichstein. On the rationality problem for forms of moduli spaces of stable marked curves of positive genus. arXiv preprint arXiv:1709.05696, 2017. [12] Mathieu Florence and Zinovy Reichstein. The rationality problem for forms of M . Bulletin 0,n of the London Mathematical Society, 50(1):148–158, 2018. [13] Philippe Gille and Tam´as Szamuely. Central Simple Algebras and Galois Cohomology. Cam- bridge Studies in Advanced Mathematics. Cambridge University Press, 2006. [14] Dominic Joyce. Motivic invariants of Artin stacks and stack functions. The Quarterly Journal of Mathematics, 58(3):345–392, 2007. [15] B. E. Kunyavski˘ı. Three-dimensional algebraic tori. In Investigations in number theory (Rus- sian), pages 90–111. Saratov. Gos. Univ., Saratov, 1987. Translated in Selecta Math. Soviet. 9 (1990), no. 1, 1–21. [16] Ivan Martino. The Ekedahl invariants for finite groups. Journal of Pure and Applied Algebra, 220(4):1294–1309, 2016. [17] Ivan Martino. Introduction to the Ekedahl Invariants. Mathematica scandinavica, 120(2):211– 224, 2017. [18] Roberto Pirisi and Mattia Talpo. On the motivic class of the classifying stack of G and the spin groups. To appear in International Mathematics Research Notices. arXiv preprint arXiv:0903.3143. [19] Mattia Talpo and Angelo Vistoli. The motivic class of the classifying stack of the special orthogonal group. Bulletin of the London Mathematical Society, 49(5):818–823, 2017. [20] Bertrand To¨en. Grothendieck rings of Artin n-stacks. arXiv preprint math/0509098, 2005. [21] V. E. Voskresenski˘ı. Algebraic groups and their birational invariants, volume 179 of Trans- lations of Mathematical Monographs. American Mathematical Society, Providence, RI, 1998. Translated from the Russian manuscript by Boris Kunyavski [Boris E. Kunyavski˘ı]. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Mathematics arXiv (Cornell University)

On the motivic class of an algebraic group

Scavia, Federico
Mathematics , Volume 2021 (1808) – Jul 31, 2018
Download PDF
10 pages

Loading next page...
 
/lp/arxiv-cornell-university/on-the-motivic-class-of-an-algebraic-group-bpgQRUKR6D

References (23)

ISSN
1944-7833
eISSN
ARCH-3343
DOI
10.2140/ant.2020.14.855
Publisher site
See Article on Publisher Site

Abstract

FEDERICO SCAVIA Abstract. Let F be a field of characteristic zero admitting a biquadratic field extension. We give an example of a torus G over F whose classifying stack −1 BG is stably rational and such that {BG} 6= {G} in the Grothendieck ring of algebraic stacks over F . We also give an example of a finite ´etale group scheme A over F such that BA is stably rational and {BA} 6= 1. 1. Introduction Let F be a field. The Grothendieck ring of algebraic stacks K (Stacks ) was 0 F introduced by Ekedahl in [8], following up on earlier works [1], [14], [20]. It is a variant of the Grothendieck ring of varieties K (Var ). By definition, K (Stacks ) 0 F 0 F is generated as an abelian group by the equivalence classes {X} of all algebraic stacks X of finite type over F with affine stabilizers. These classes are subject to the scissor relations {X} = {Y } + {X \ Y } for every closed substack Y ⊆ X, and the relations {E} = {A × X} for every vector bundle E of rank n over X. The product is defined by {X} · {Y } := {X × Y }, and extended by linearity. Given a group scheme G over F , we may consider the class {BG} of its classifying stack in K (Stacks ). The problem of computing {BG} appears to be related to the 0 F problem of the stable rationality of BG, although no direct implications are known. Recall that BG is stably rational if for one (equivalently, every) generically free representation V of G, the rational quotient V/G is stably rational. An equivalent terminology is that the Noether problem for stable rationality has a positive solution for G; see [12, §3]. The case of a finite (constant) group G was considered in [7]: it frequently happens that {BG} = 1 (notably for the symmetric groups, see [7, Theorem 4.3]), although there are examples of finite groups G for which {BG} = 6 1; see [7, Corollary 5.2, Corollary 5.8]. Further work on the triviality of {BG} for finite groups G has been done in [16] and [17]. So far, all the known examples of finite group schemes G for which {BG} 6= 1 are such that BG is not stably rational. This suggests the following question. Question 1.1. (cf. [7, §6]) Is it true that, for a finite group scheme G, the following two conditions are equivalent? • BG is stably rational; • {BG} = 1 in K (Stacks ). 0 F We will answer Question 1.1 in the negative in Theorem 1.6. Now let G be a connected linear algebraic group. Recall that G is special if every G-torsor is Zariski-locally trivial. For example, GL , SL and Sp are special; see n n [5]. It was shown by Ekedahl that if P → S is a torsor under the special group G, then {P} = {G}{S}. This is immediate if S is a scheme, but less obvious arXiv:1808.00056v2 [math.AG] 30 Dec 2020 2 FEDERICO SCAVIA when S is a stack; see [3, Corollary 2.4]. Applying this to the universal G-torsor Spec F → BG, one obtains {BG}{G} = 1. −1 The equality {BG} = {G} appears to be the analogue for connected groups of the relation {BG} = 1 for finite group schemes. In [3], these equalities are referred to as expected class formulas, and there is a sense in which they are “almost” true. In [8, §2] Ekedahl defines a generalized Euler characteristic χ : K (Stacks ) → K (Coh ) c 0 F 0 F taking values in a Grothendieck ring K (Coh ) of Galois representations over F . If 0 F G is a finite group scheme, the equality χ ({BG}) = 1 always holds [7, Proposition 3.1]. On the other hand, if G is connected, then χ ({BG}{G}) = 1; see [3, §2.2]. Since {BG} 6= 1 for some finite groups G, the following question naturally arises. Question 1.2. Let F be a field. Is it true that −1 (1.3) {BG} = {G} in K (Stacks ) for every connected group G? 0 F In Theorem 1.5, we show that the answer to Question 1.2 is also negative. Com- putations for non-special G have been carried out for PGL and PGL in [3], for 2 3 SO and n odd in [6], for SO and n even and O for any n in [19], and for n n n Spin , Spin and G in [18]. In each of these cases, (1.3) was found to be true. The 7 8 expectation was that, for a connected linear algebraic group G over a field F of characteristic 0, Question 1.4 below should have an affirmative answer. If F is an algebraically closed field, then there are no examples of connected G where BG is known not to be stably rational. If F is not assumed to be algebraically closed, then such examples exist. The following variant of Question 1.1 seems natural in this context. Question 1.4. (cf. [19, §1] and [18, Remark 4.1]) Is it true that, for a connected linear algebraic group G, the following two conditions are equivalent? • BG is stably rational; −1 • {BG} = {G} in K (Stacks ). 0 F Our first result gives a negative answer to Question 1.2 and Question 1.4. Theorem 1.5. Let F be a field of characteristic zero which admits a biquadratic field extension K, let E and E be two distinct quadratic subextensions of K/F , 1 2 (1) and set G := R (G ). Then E ×E /F 1 2 (a) BG is stably rational, and −1 (b) {BG} 6= {G} in K (Stacks ). 0 F The torus G is an example of a norm-one torus; see Section 2 for the definition. It follows from Theorem 1.5 that counterexamples H to (1.3) exist in any dimension dim H ≥ 3: consider for example H := G × G for r ≥ 0. The key ingredient in the proof of Theorem 1.5 is the refined Euler characteristic of Ekedahl, introduced in [8, §6, 3]; see Section 4. Our second result gives a negative answer to Question 1.1. Theorem 1.6. Let F be a field of characteristic zero which admits a biquadratic field extension K, and let E and E be two distinct quadratic subextensions of 1 2 (1) K/F . Define G := R (G ), and let A := G[2] be the 2-torsion subgroup of E ×E /F 1 2 G. Then ON THE MOTIVIC CLASS OF AN ALGEBRAIC GROUP 3 (a) BA is stably rational, and (b) {BA} 6= 1 in K (Stacks ). 0 F Questions 1.1, 1.2 and 1.4 remain open in the case, where the base field F is assumed to be algebraically closed. Our arguments do not shed any new light in this setting. The remainder of this paper is structured as follows. In Section 2 we review well known computations of motivic classes for non-split tori. In Section 3 we obtain explicit formulas for the motivic classes of G and BG, and in Section 4 we give the required background on the refined Euler characteristic. In Section 5 we prove Theorem 1.5, and in Section 6 we prove Theorem 1.6. 2. Preliminaries Let F be a field. We will write L for the class {A } in K (Var ) or K (Stacks ). 0 F 0 F If E is an ´etale algebra over F , we will denote by {E} the class {Spec E} in K (Var ) or K (Stacks ). If X is a quasi-projective scheme over E, we will denote 0 F 0 F by R (X) the Weil restriction of X to F . By definition, for every F -scheme S E/F one has R (X)(S) = X(S ). We refer the reader to [21, §3.12] for an account E/F E of the main properties of the Weil restriction. Let G be a linear algebraic group over F , and α ∈ H (F, G) be represented by a G-torsor P → Spec F . For every quasi-projective F -scheme Z, we denote by Z the twist of Z by P , that is, Z := (Y × P )/G, where G acts diagonally. We refer the reader to [10, Section 2] for the definition and the basic properties of the twisting operation. We will write C for the cyclic group of two elements, and S for the symmetric 2 n group on n symbols. The following observations will be repeatedly used in the sequel. Lemma 2.1. Let X be a scheme over F , E an ´etale algebra of degree n over F , α ∈ H (F, S ) the class corresponding to E/F . (a) Let S act on the disjoint union ∐ X by permuting the n copies of X. i=1 Then (∐ X) = X . i=1 (b) Let S act on X by permuting the n factors. Then (X ) = R (X). E/F Proof. (a) Let Y := ∐ X, and let S act on Y by permuting the copies of X. By i=1 definition, Y = (Y × Spec E)/S (Y × X )/S , n X E n where S acts diagonally. This shows that Y is the twist of X by the trivial n E S -torsor Y → X in the category of X-schemes, which implies Y = X . n E (b) See the bottom of page 5 in [11]. Lemma 2.2. Let 1 → N → G → H → 1 be an exact sequence of group schemes over F , and assume that G is special. Then {BN} = {H}/{G}. 4 FEDERICO SCAVIA Proof. See [3, Proposition 2.9]. Let F be a separable closure of F . Recall that a group scheme T over F is called a torus if T = G for some n ≥ 0. The character lattice of T is the finitely m,F ,G ). The character lattice induces generated Z-free Gal(F )-module Hom (T F F m,F s s s an anti-equivalence between the category of F -tori and the category of Gal(F )- lattices, i.e., Z-free continuous Gal(F )-modules; see [9, §2]. Similarly, for every separable finite extension L/F , we have an anti-equivalence between Gal(L/F )- lattices and F -tori T split by L, i.e., such that T G for some n ≥ 0. The m,L dual torus of T is the torus T whose character lattice is dual to that of T . Let E be an ´etale algebra over F . If G is a group scheme over E, then R (G) is E/F a group scheme over F . The group R (G ) := R (G ) is an F -torus. Tori m m,E E/F E/F of this kind are called quasi-split. They are special groups, and they correspond to permutation Gal(F )-lattices, that is, lattices admitting a Z-basis that is permuted by Gal(F ); see [21, §3.12, Example 19]. Lemma 2.3. Let T be an algebraic torus over F , and let T be its dual. Assume that T is stably rational. Then (a) BT is stably rational; (b) {BT }{T} = 1 in K (Stacks ). 0 F Proof. Since T is stably rational, by [21, §4.7, Theorem 2] there is a short exact sequence (2.4) 1 → T → T → T → 1 1 2 where T and T are quasi-split. Since quasi-split tori are isomorphic to their dual, 1 2 the sequence dual to (2.4), (2.5) 1 → T → T → T → 1, 2 1 shows that T embeds in T . We may view T as a maximal torus inside GL , where 2 2 n n = rank T . This gives a faithful representation of T with quotient birational to T . Since quasi-split tori are rational, it follows that BT is stably rational. Quasi-split tori are special, so we may apply Lemma 2.2 to (2.4) and (2.5). We ′ ′ obtain {T} = {T }/{T } and {BT } = {T }/{T }, so {BT }{T} = 1. 2 1 1 2 Let E/F be an ´etale algebra, and let R (G ) be the associated quasi-split E/F m torus. The kernel of the norm homomorphism R (G ) → G is called a E/F m m (1) norm-one torus, and is denoted by R (G ). Its dual torus is isomorphic to E/F R (G )/ G . m m E/F Lemma 2.6. Assume that char F 6= 2. Let E := F ( m) be a separable quadratic field extension, and let α denote the class of E/F in H (F, C ). Then: (1) (a) R (G ) R (G )/ G . m E/F m m E/F 1 −1 1 1 (b) Let Gal(E/F ) act on P via z 7→ z . Then P P . (c) R (G )/ G is rational and m m E/F −1 {R (G )/ G } = {B(R (G )/ G )} = L − {E} + 1. E/F m m E/F m m −1 (d) {R (G )} = {BR (G )} = (L − 1)(L − {E} + 1). m m E/F E/F 1 2 (e) {R (P )} = L + {E}L + 1. E/F ON THE MOTIVIC CLASS OF AN ALGEBRAIC GROUP 5 Proof. (a) Both tori correspond to the unique non-trivial Gal(E/F )-lattice of rank 1. Here Gal(E/F ) = C . 1 1 (b) The C -action on P has a fixed point z = 1, hence P has an F -point. By Chaˆtelet’s Theorem [13, Theorem 5.1.3], a form of P which admits an F -point is trivial (the case n = 1 is particularly simple, see [13, Remark 1.3.5]). We conclude 1 1 that P = P . (1) (c) Let T := R (G ) = R (G )/ G . The open embedding G ֒→ P , m E/F m m m E/F as the complement of Z := {0, ∞}, is equivariant under the C -action on G and 2 m 1 −1 P given by z 7→ z . Twisting by α, we obtain by (b) an open embedding of T 1 α in P as the complement of Z. In particular, T is rational. By Lemma 2.1(a), Z = Spec E, so 1 α {T} = {P } − { Z} = L + 1 − {E}. Now (c) follows from Lemma 2.3(b). (d) The first equality holds because R (G ) is special. Consider the short E/F exact sequence 1 → G → R (G ) → T → 1. m E/F m Since R (G ) is special, Lemma 2.2 yields E/F −1 {R (G )} = (L − 1){BT} , E/F m thus (d) follows from (c). 1 1 (e) Write P = A ∪ {∞}, and consider the C -equivariant decomposition 1 2 1 2 1 1 (P ) = (A ) ∐ (A × {∞} ∪ {∞} × A ) ∐ {(∞, ∞)}. By Hilbert’s Theorem 90 and Lemma 2.1(a), twisting by α gives 1 2 1 R (P ) = A ∐ A ∐ Spec F, E/F 1 2 thus {R (P )} = L + {E}L + 1. E/F 3. The classes of G and BG Let F be a field of characteristic not 2, and assume that there exists a biquadratic extension √ √ K := F ( m , m ) 1 2 of F . Let √ √ √ E := F ( m ), E := F ( m ), E := F ( m m ), E := E × E , 1 1 2 2 12 1 2 1 2 and let Γ := Gal(K/F ) = C be the Galois group of K/F . We define the torus (1) G := R (G ) E/F and let G := R (G )/ G E/F m m be the dual torus of G. By definition, we have a short exact sequence (3.1) 1 → G → R (G ) −→ G → 1, m m E/F where N is the norm homomorphism. The purpose of this section is the proof of Proposition 3.7, which expresses {BG} and {G} as rational functions in L, with coefficients classes of ´etale algebras. 6 FEDERICO SCAVIA σ σ 1 2 Let σ and σ be generators for Γ such that E = K and E = K . Consider 1 2 1 2 2 −1 −1 the Γ-action on G , where σ (u, v) = (v , u ) and σ (u, v) = (v, u), and set 1 2 (3.2) T := (G ), where α ∈ H (F, Γ) corresponds to the extension K/F . Lemma 3.3. We have {T} = L + ({E } − {K})L + {K} − {E } − {E } + 1. 12 1 2 Proof. The embedding of G in P as the complement of Z := {0, ∞} gives an 2 1 2 2 1 2 open embedding G ֒→ (P ) such that the Γ-action on G extends to (P ) . By m m definition 1 2 1 2 (P ) = ((P ) × Spec K)/Γ, where Γ = hσ , σ i acts diagonally. We first take the quotient by the subgroup 1 2 −1 −1 σ σ 1 2 hσ σ i. Since σ σ (u, v) = (u , v ) and E = K , by Lemma 2.6(b) 1 2 1 2 12 1 2 1 2 (P ) = ((P ) × Spec E )/C , 12 2 1 2 where C acts on (P ) by switching the two factors. Here we are using the fact 1 2 that every automorphism of (P ) must respect the ruling (because it respects 1 2 1 2 the intersection form), and so Aut((P ) ) = (Aut(P )) ⋊ C , where C switches 2 2 1 2 1 the two factors. By Lemma 2.1(b) we deduce that (P ) = R (P ), so by E /F Lemma 2.6(e) 1 2 2 (3.4) { (P ) } = L + {E }L + 1. 1 2 2 We may partition (P ) \ G in two strata Z := Z × Z, Z := (Z × G ) ∐ (G ×Z). 1 2 m m The Γ-action on Z has two orbits, and Γ acts on Z by transitively permuting 1 2 the components as the Klein subgroup of S . By Lemma 2.1(a), Z = Spec E ∐ 4 1 1 Spec E and Z = G × Spec K. By (3.4) 2 2 m 1 2 α α {T} = { (P ) } − { Z } − { Z } 1 2 = L + {E }L + 1 − {E } − {E } − {K}(L − 1) 12 1 2 = L + ({E } − {K})L + {K} − {E } − {E } + 1. 12 1 2 Proposition 3.5. There is a short exact sequence of tori 1 → G → G → T → 1, where T is the torus of (3.2). Proof. Let P , M and Z be the character lattices of R (G ), G and G , respec- m m E/F tively. We may view P as the Γ-lattice with a basis e , e , e , e , such that σ acts 1 2 3 4 1 by switching e with e and fixing e and e , and σ switches e with e and fixes 1 2 3 4 2 3 4 e and e . The sequence of Γ-lattices dual to (3.1) identifies M with the cokernel 1 2 of the Γ-homomorphism Z → P given by 1 7→ e + e + e + e ; denote by e ∈ M 1 2 3 4 i the projection of e . Following Kunyavski˘ı [15, §3, Proposition 1(b)], we consider an exact sequence of Γ-lattices (3.6) 0 → N → M −→ Z → 0. The map π is defined by π( a e ) = a + a − a − a , and N := Ker π. A basis i i 1 2 3 4 for N is given by v := e +e and v := e +e . With respect to the basis (v , v ), 1 1 3 2 1 4 1 2 ON THE MOTIVIC CLASS OF AN ALGEBRAIC GROUP 7 the Γ-action on N is given by σ (a, b) = (−b, −a) and σ (a, b) = (b, a). It is now 1 2 clear that N is the character lattice of the torus T of (3.2), hence the proof is complete. Proposition 3.7. (a) BG is stably rational. (b) {BG}{G } = 1 in K (Stacks ). 0 F Proof. Consider the sequence (3.8) 1 → G → G → (R (G )/ G ) × (R (G )/ G ) → 1, m E /F m m E /F m m 1 2 which exhibits G as a G -torsor over a rational variety, by Lemma 2.6(c). We deduce that G is rational, and now (a) and (b) follow from Lemma 2.3. Proposition 3.9. We have (3.10) {G} = (L − 1)(L + ({E } − {K})L + {K} − {E } − {E } + 1) 12 1 2 and −1 (3.11) {BG} = (L − 1)(L − {E } + 1)(L − {E } + 1) 1 2 in K (Stacks ). 0 F Proof. By Proposition 3.5, G is a G -torsor over T . Since G is special, {G} = m m (L − 1){T}. The class of T was determined in Lemma 3.3. −1 ′ ′ By Proposition 3.7(b), {BG} = {G }. Since G is special, by (3.8), {G } = (1) (1) (L − 1){R (G )}{R (G )}. Now (3.11) follows from Lemma 2.6(c). m m E /F E /F 1 2 4. The refined Euler characteristic Let F be a field of characteristic zero. Using the computations of the previous section, we will reduce Theorem 1.5(b) to the assertion that a certain polynomial in L with coefficients motivic classes of ´etale algebras is a non-zero element of K (Var ). To prove the assertion, we will use a simplified version of the refined 0 F Euler characteristic, introduced by Ekedahl in [8]. Fix a prime number p, and let G be a profinite group. The representation ring a (G) of G is the Grothendieck ring of continuous G-representations [M] of finite dimension over F , subject to the relations [M ⊕ N] = [M] + [N]. Note that no relations for non-split short exact sequences are imposed. The product structure on a (G) is given by tensor product of representations. The next observation is well known when G is assumed to be finite; see [2, §5.1]. Lemma 4.1. As an abelian group, a (G) is freely generated by the set of isomor- phism classes of indecomposable representations. Proof. It is clear that a (G) is generated by isomorphism classes of indecomposable P P representations. Assume that a [M ] − b [N ] = 0 in a (G), for some positive i i j j p integers a , b and some pairwise non-isomorphic indecomposable G-representations i j M and N . i j As a group, a (G) is the quotient group F/I, where F is the free abelian group with one generator hPi for every isomorphism class of G-representations P , and I is the subgroup generated by all elements of the form hP ⊕ Qi−hPi−hQi. It follows that we may find a G-representation X such that ⊕b ⊕a j (⊕ M ) ⊕ X (⊕ N ) ⊕ X. i j i j 8 FEDERICO SCAVIA Let G be a finite quotient of G such that G acts on M , N and X through G . 0 i j 0 Then M ⊕ X = N ⊕ X as G -representations. By the Krull-Schmidt Theorem applied to the group algebra F [G ], this implies M N as G -modules, hence as p 0 0 G-modules. This is impossible, because the indecomposable representations M and N are pairwise non-isomorphic. Proposition 4.2. Let F be a field of characteristic zero, let Gal(F ) be the absolute Galois group of F , and let R := a (Gal(F )). There is a ring homomorphism p p µ : K (Var ) → R [t] 0 F p i i such that for every smooth complete variety X we have µ (X) = [H (X ,F )]t . ´ et p Proof. See the proof of [8, Proposition 3.2(i)]. To show that µ is well-defined, one needs to assume that charF = 0 in order to invoke Bittner’s presentation of K (Var ); see [4, Theorem 3.1]. 0 F 5. Proof of Theorem 1.5 Theorem 1.5(a) was proved in Proposition 3.7(b), so we will focus on Theo- rem 1.5(b). We maintain the notation given at the beginning of Section 3. (1) Proof of Theorem 1.5(b). Assume by contradiction that G = R (G ) satisfies E/F (1.3). Then by Proposition 3.9 we have (L − 1)(L − {E } + 1)(L − {E } + 1) = 1 2 = (L − 1)(L + ({E } − {K})L + {K} − {E } − {E } + 1) 12 1 2 in K (Stacks ). Since L − 1 is invertible in K (Stacks ), we may divide by L − 1 0 F 0 F on both sides. Subtracting L on the left and on the right, we arrive to (2−{E }−{E })L+(1−{E })(1−{E }) = ({E }−{K})L+{K}−{E }−{E }+1, 1 2 1 2 12 1 2 that is ({K} − {E } − {E } − {E } + 2)L = 0 1 2 12 in K (Stacks ). 0 F Recall that K (Stacks ) is the localization of K (Var ) at L and the cyclotomic 0 F 0 F polynomials in L; see [8, Theorem 1.2]. It follows that (5.1) ({K} − {E } − {E } − {E } + 2)f(L) = 0 1 2 12 in K (Var ), where f(x) ∈ Z[x] is a monic polynomial of some degree n. 0 F In order to obtain a contradiction, we now want to apply the homomorphism µ of (4.2), with respect to the prime p = 2. If L/F is an ´etale algebra of degree n, µ ({L}) consists of the permutation representation of Gal(F ) associated to L, concentrated in degree 0. Since we have chosen p = 2, µ ({P }) consists of one copy of the trivial representation in degree 0 and 2 (in the case p > 2 one would need a 1 2 Tate twist in degree 2). Since L = {P } − 1, we deduce that µ (L) = t , and hence µ (f(L)) = f(t ). If X is a finite Gal(F )-set, we denote by F [X] the permutation representation over F associated to X. Recall from Section 3 that we denote Gal(K/F ) by Γ = hσ , σ i. Applying µ to (5.1) and looking at degree 2n, we obtain 1 2 [F [Γ]] − [F [Γ/ hσ i]] − [F [Γ/ hσ i]] − [F [Γ/ hσ i]] + 2[F ] = 0 2 2 1 2 2 2 12 2 ON THE MOTIVIC CLASS OF AN ALGEBRAIC GROUP 9 in R . This is a non-trivial relation of linear dependence in R among classes of 2 2 indecomposable representations. This is in contradiction with Lemma 4.1, hence −1 {BG} 6= {G} , as desired. Remark 5.2. By [21, §4.9, Example 7] every torus of rank 2 is rational, so by Proposition 3.5 the torus G is rational. By Lemma 2.3, BG is stably rational and ′ −1 ′ −1 ′ ′ {BG } = {G} . By Proposition 3.7(b) we have {BG} = {G } , so {BG }{G } = −1 −1 {BG} {G} . Since {BG}{G} 6= 1, the conclusions of Theorem 1.5(a) and (b) hold for G as well. 6. Proof of Theorem 1.6 We maintain the notation of Section 3. Proof of Theorem 1.6. Let Γ := Gal(K/F ), let M be the character lattice of G, so that M/2M is the character module of A, and let P be the character lattice of R (G ). As in the proof of Proposition 3.5, we view P as the lattice freely E/F m generated by e , e , e , e , such that σ acts by switching e with e , and σ by 1 2 3 4 1 1 2 2 switching e with e . Using (3.1), we may construct a commutative diagram of 3 4 Γ-modules 0 Z P M 0 (6.1) ι 0 N P M/2M 0. with exact rows. Here Z denotes the trivial one-dimensional Γ-lattice, ι(1) := e + e + e + e , and N is the kernel of ϕ, that is, 1 2 3 4 N = { a e : a ≡ a ≡ a ≡ a (mod 2)}. i i 1 2 3 4 i=1 Applying the snake lemma to (6.1), we obtain a short exact sequence 0 → Z − → N → M → 0. Define π : N → Z by sending a e to (a + a )/2. Then π is a Γ-homomorphism i i 1 2 and ι is a section of π. Therefore, we have an isomorphism N Z ⊕ M. Let S be an F -torus with character lattice N. Since N Z ⊕ M, we have S G × G. The bottom row of (6.1) corresponds to the short exact sequence of group schemes 1 → A → R (G ) → G × G → 1. E/F m m By Lemma 2.2, we have {BA} = {G }{G}/{R (G )}. Applying Lemma 2.2 m E/F m to (3.1), we see that {BG} = {G }/{R (G )}. Therefore, {BA} = {BG}{G}. m E/F m −1 By Theorem 1.6 we have {BG} 6= {G} , hence {BA} 6= 1, as desired. Acknowledgments I would like to thank my advisor Zinovy Reichstein for his guidance and for greatly improving the exposition, Mattia Talpo and Angelo Vistoli for helpful com- ments, and Boris Kunyavski˘ı for sending me a copy of his paper [15]. I am very grateful to the anonymous referee for finding a mistake in a previous version of the proof of Theorem 1.5, and for suggesting a fix. 10 FEDERICO SCAVIA References [1] Kai Behrend and Ajneet Dhillon. On the motivic class of the stack of bundles. Advances in Mathematics, 212(2):617–644, 2007. [2] D. J. Benson. Representations and cohomology. II, volume 31 of Cambridge Studies in Ad- vanced Mathematics. Cambridge University Press, Cambridge, second edition, 1998. Coho- mology of groups and modules. [3] Daniel Bergh. Motivic classes of some classifying stacks. Journal of the London Mathematical Society, 93(1):219–243, 2015. [4] Franziska Bittner. The universal Euler characteristic for varieties of characteristic zero. Com- positio Mathematica, 140(4):1011–1032, 2004. [5] S´eminaire Claude Chevalley and JP Serre. Espaces fibr´es alg´ebriques. S´eminaire Claude Chevalley, 3:1–37, 1958. [6] Ajneet Dhillon, Matthew B Young, et al. The motive of the classifying stack of the orthogonal group. The Michigan Mathematical Journal, 65(1):189–197, 2016. [7] Torsten Ekedahl. A geometric invariant of a finite group. arXiv preprint arXiv:0903.3148, [8] Torsten Ekedahl. The Grothendieck group of algebraic stacks. arXiv preprint arXiv:0903.3143, 2009. [9] Giordano Favi and Mathieu Florence. Tori and essential dimension. J. Algebra, 319(9):3885– 3900, 2008. [10] Mathieu Florence. On the essential dimension of cyclic p-groups. Invent. Math., 171(1):175– 189, 2008. [11] Mathieu Florence and Zinovy Reichstein. On the rationality problem for forms of moduli spaces of stable marked curves of positive genus. arXiv preprint arXiv:1709.05696, 2017. [12] Mathieu Florence and Zinovy Reichstein. The rationality problem for forms of M . Bulletin 0,n of the London Mathematical Society, 50(1):148–158, 2018. [13] Philippe Gille and Tam´as Szamuely. Central Simple Algebras and Galois Cohomology. Cam- bridge Studies in Advanced Mathematics. Cambridge University Press, 2006. [14] Dominic Joyce. Motivic invariants of Artin stacks and stack functions. The Quarterly Journal of Mathematics, 58(3):345–392, 2007. [15] B. E. Kunyavski˘ı. Three-dimensional algebraic tori. In Investigations in number theory (Rus- sian), pages 90–111. Saratov. Gos. Univ., Saratov, 1987. Translated in Selecta Math. Soviet. 9 (1990), no. 1, 1–21. [16] Ivan Martino. The Ekedahl invariants for finite groups. Journal of Pure and Applied Algebra, 220(4):1294–1309, 2016. [17] Ivan Martino. Introduction to the Ekedahl Invariants. Mathematica scandinavica, 120(2):211– 224, 2017. [18] Roberto Pirisi and Mattia Talpo. On the motivic class of the classifying stack of G and the spin groups. To appear in International Mathematics Research Notices. arXiv preprint arXiv:0903.3143. [19] Mattia Talpo and Angelo Vistoli. The motivic class of the classifying stack of the special orthogonal group. Bulletin of the London Mathematical Society, 49(5):818–823, 2017. [20] Bertrand To¨en. Grothendieck rings of Artin n-stacks. arXiv preprint math/0509098, 2005. [21] V. E. Voskresenski˘ı. Algebraic groups and their birational invariants, volume 179 of Trans- lations of Mathematical Monographs. American Mathematical Society, Providence, RI, 1998. Translated from the Russian manuscript by Boris Kunyavski [Boris E. Kunyavski˘ı].

Journal

Mathematics – arXiv (Cornell University)

Published: Jul 31, 2018

There are no references for this article.