62 3. TUBULAR SHIFTS AND PRIME ELEMENTSProof. Let M ∈H + be of rank r>0. Let S = S x be the simple objectconcentrated in x. There is the universal extension0 −→ M α M−→ σ x (M) −→ Ext 1 (S, M) ⊗ End(S) S −→ 0.The morphism M ·π x−→ M(d) is a monomorphism and its cokernel is concentrated inthe point x and of length r · e(x): using a line bundle filtration of M and induction,it suffices to show this for line bundles L ′ . If L ′ is a shift of L then the cokernelis Sxe(x) . In orbit case III we have to consider also the case L ′ = L. There is anirreducible map L −→ uL, and the cokernel S 0 is a simple object. One can assumethat S 0 ≄ S x . It then follows that the cokernel of the map L ·π x−→ L(d) isalsoisomorphic to Sx e(x) .We have to show that the cokernel C of M ·π x−→ M(d) is semisimple: The mapC ·π x−→ C(d) is zero, which follows from the commutative exact diagram0 M ·π xM(d)·π x·π x (d)0 M(d) ·π x(d)M(2d)p(d)pC·π x0C(d) 0.Applying the exact functor φ ◦ q x from 2.2.6 we get a short exact sequence ofright R P -modules:·π0 −→ MxP −→ M(d)P −→ C P −→ 0.Since the map C ·π x−→ C(d) is zero, we see that C P is a graded R P /P P -module,hence semisimple by 2.2.8. It follows that C is a direct sum of copies of S.Hence we get a commutative, exact diagram0 M α MM(x)β MM x 0i M0 M ·π xM(d)j M0,Cwith isomorphisms i M and j M (compare 0.4.2 (2)). With the uniqueness property0.4.2 (1) of σ x (f) it follows easily that for each f ∈ Hom(M,N)(withN ∈H + )we have σ x (f)◦i M = i N ◦f(d). Therefore, the functors σ d and σ x are naturally isomorphicon H + . (Actually, the argument shows, that this holds everywhere outsideU x .) With Lemma 1.2.2, presenting each object in H 0 as cokernel of a monomorphismbetween objects from H + , the result follows by diagram chasing. □Corollary 3.1.3. Let x, y ∈ X such that the corresponding prime elementsπ x and π y are central of the same degree in R. Then the tubular shifts σ x and σ yare isomorphic.□3.2. Non-central prime elements and ghosts3.2.1 (Normal multiplication). We generalize Theorem 3.1.2 to arbitrary primeelements. Therefore, assume that r ∈ R is a (non-zero) normal element of degree n.Then r induces the automorphism γ = γ r of the graded algebra R, by the formula
3.2. NON-CENTRAL PRIME ELEMENTS AND GHOSTS 63sr = rγ(s). Let M ∈ mod Z (R). Denote by M γ the object in mod Z (R), where Mand M γ coincide as abelian groups, and where the R-action on M γ is defined bym ·γ s def= m · γ(s).Then multiplication ·r : m ↦→ m · r defines a morphism between the graded rightmodules M and M γ (n), since(m · s) · r = m · (s · r) =m · (r · γ(s)) = (m · r) ·γ s.Since obviously γ(m) =m, the algebra automorphism γ gives rise to an automorphismγ ∗ by γ ∗ (M) =M γ on mod Z (R) and (denoted by the same symbol) alsoon H = modZ mod Z(R), such that σ ◦ γ∗ = γ ∗ ◦ σ; moreover,γ ∗ (˜M) =˜M γ . Sheafificationgives a homomorphism ˜M −→ ˜M γ (n), yielding a natural transformation0 ·r·r1 H −→ σ n ◦ γ ∗ = γ ∗ ◦ σ n . Obviously, γ gives rise to an isomorphism R −→ R γof graded right R-modules, hence γ ∗ leaves L fixed, that is, γ ∗ ∈ Aut(X). Notethat after identifying R with R γ via γ, the functor γ ∗ on mod Z (R) actsonelementsf· of Hom(R, R(m)) (in contrast to Hom(R γ ,R γ (m))!) like γ −1 , since(γ −1 ◦ (γ ∗ (f·)) ◦ γ)(r) = γ −1 (f) · r. We will also make use of the notationγ ∗ =(γ −1 ) ∗ .□The k-algebra automorphisms induced by normal elements are special cases ofa more general class of graded algebra automorphisms:Definition 3.2.2. An automorphism γ of the graded k-algebra R is calledprime fixing, if for all homogeneous prime ideals P (of height one) we have γ(P )=P . In other words, γ is prime fixing if and only if for each prime element π thereis a unit u ∈ R0 ∗ such that γ(π) =πu. DenotebyAut 0 (R) the subgroup of Aut(R)of all graded algebra automorphisms, which are prime fixing.Examples of prime fixing automorphism are the inner automorphisms ι u (u ∈R0)givenbyι ∗ u (r) =u −1 ru, defining the subgroup Inn(R), and automorphismsϕ a defined in the following way: Let a =(a i ) i≥1 be a sequence of elements a i ∈Z(R 0 ) ∗ with ra i = a i+1 r for all homogeneous elements r ∈ R of degree one. Thenϕ a (r) =a 1 · a 2 · ...· a n · r (for each homogeneous r ∈ R of degree n) defines anautomorphism ϕ a ∈ Aut 0 (R). (Recall that R is generated in degrees 0 and 1.) Wedenote the subgroup of Aut(R) generated by all ι u and all ϕ a by Inn(R) whichisa normal subgroup of Aut(R).□Let a =(a i ) be a sequence as above defining ϕ a . Since R contains centralhomogeneous elements (see 4.1.3) there is n ≥ 1 such that(3.2.1) a i+n = a i for all i ≥ 1.If ra 1 = a 1 r for some r of degree one then n = 1 can be chosen, and a 1 lies in thecentre of R. This is the case, for example, if there is a central element of degreeone in R.Proposition 3.2.3. (1) Let γ be a graded algebra automorphism of R. Thenthe induced automorphisms γ ∗ and γ ∗ on H are automorphisms of X. Moreover, γ ∗is trivial (that is, isomorphic to the identity) if and only if γ ∈ Inn(R). Hence theassignment γ ↦→ γ ∗ induces an injective group homomorphism from Aut(R)/Inn(R)into Aut(X).