ebook img

Ziegler closures of some unstable tubes PDF

0.3 MB·
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview Ziegler closures of some unstable tubes

ZIEGLER CLOSURES OF SOME UNSTABLE TUBES 7 LORNAGREGORY 1 0 2 n a Abstract. We describe the modules in the Ziegler closure of ray and J coray tubes in module categories over finite-dimensional algebras. We 1 improveslightlyonKrause’sresultforstabletubesbyshowingthatthe 1 inverse limit along acoray in a ray or coray tubeis indecomposable, so inparticular,theinverselimitalongacorayinastabletubeisindecom- ] posable. In order to do all this we first describe the finitely presented T modulesoverandtheZieglerspectraofiteratedone-pointextensionsof R valuation domains. Finally we give a sufficient condition for the k-dual . of a Σ-pure-injectivemodule over a k-algebra to be indecomposable. h t Ray and coray tubes frequently occur in Auslander-Reiten quivers of a m finite-dimensional algebras. After stable tubes (also referred to as smooth [ tubes) they are the simplest form of Ringel and D’Este’s coherent tubes. Generalising Krause’s definition, [Kra98b, pg 20], of a generalised tube 1 we introduce the notions of a generalised coray tube and a generalised ray v 4 tube. Our aim here is to use these notions to describe the Ziegler closures 8 of unstable tubes containing no projective modules (dually, unstable tubes 1 containing no injective modules). 3 0 As in the case of stable tubes, we show that every ray tube has finitely . many non-finitely presented indecomposable pure-injectives in its Ziegler 1 0 closure each of which is either a direct limit along a ray, an inverse limit 7 along a coray or a generic module. Improving slightly on Krause’s results 1 we show that in any ray or coray tube (thus also in any stable tube) the : v inverse limit along a coray is indecomposable. This result was claimed in i X [Pre09, 15.1.10] but no proof is given and the proof indicated there does not r work. a In section 2, we describe the finitely presented modules over iterated one- point extensions of discrete valuation domains. This allows us, see section 3, to describe the indecomposable pure-injectives over iterated one-point extensions of discrete valuation domains. Using these results, we describe a functor from the module category of the iterated one-point extension of k[[x]] to a module category over a finite- dimensional k-algebra containing a generalised ray tube such that every Date: January 13, 2017. The author acknowledges the support of EPSRC through Grant EP/K022490/1 and thanksIvoHerzog and Mike Prest for useful discussions. This work was completed while theauthor was at theUniversity of Manchester. 1 2 LORNAGREGORY module in the closure of the generalised ray tube is a direct summand of some module in the image of this functor. In section 5 we introduce short embeddings. These are embeddings f : M → L between finite-length modules over an artin algebra such that for all a ∈ M, if ϕ generates the pp-type of a in M and ψ generates the pp- type of f(a) in L then the interval [ψ,ϕ] in the pp-1-lattice is finite-length. Equivalently, an embedding is short if the cokernel of (f,−) : (L,−) → (M,−) in (mod−R,Ab)fp is finite-length. Shortembeddingsallowustoinvestigatetheendomorphismringsofdirect limits along rays in ray and coray tubes. Using results in 5, we are able to showthatdirectlimitsalongraysinrayandcoray tubesareindecomposable and moreover that the canonical embedding of k into their endomorphism rings factored out by the radical is an isomorphism. In section 7 we will show that this implies that their k-duals are indecomposable. Inthefinalsection weputall thistogether todescribetheZiegler closures of ray and coray tubes for finite-dimensional algebras. We end by discussing some open questions. Throughout this paper we switch freely between the more concrete pp- formula formalism and the more abstract functor category formalism. Throughoutthispaper,ifRisaringthenMod−R(respectivelyR−Mod) denotes the category of right R-modules (respectively left R-modules) and mod−R (respectively R−mod) denotes that category of finitely presented right R-modules (respectively left R-modules). 1. Preliminaries Let R be a ring. A pp-n-formula is a formula in the language LR = (0,+,(·r)r∈R) of (right) R-modules of the form ∃y(x,y)H = 0 wherexisan-tupleofvariables andH isanappropriately sizedmatrix with entries in R. If ϕ is a pp-formula and M is a right R-module then ϕ(M) denotes the set of all elements m ∈ Mn such that ϕ(m) holds. Note that for any module M, ϕ(M) is a subgroup of Mn equipped with the addition induced by addition in M. A pair of pp-n-formulas ϕ/ψ is called a pp-pair if for all R-modules M, ϕ(M) ⊇ ψ(M). If we weaken our definition of a pp-formulato include all formulas (in one variable) in the language of (right) R-modules, L , which are equivalent R over the theory of R-modules, T , to a pp-formula then the T -equivalence R R classes of pp-n-formulas become a lattice under implication with the join of two formulas ϕ,ψ given by (ϕ+ψ)(x) := ∃y,z(x = y+z∧ϕ(y)∧ψ(z)) andthemeetgiven byϕ∧ψ. Givena(right)R-moduleM,wewritepp1(M) R for the lattice of pp-definable subgroups of M. ZIEGLER CLOSURES OF SOME UNSTABLE TUBES 3 Let M be an R-modules and m ∈ Mn. The pp-type of m in M, denoted ppM(m), is the set all pp-n-formulas ϕ such that m ∈ ϕ(M). R If M is finitely presented module and m ∈ Mn then there is a pp-n- formula ϕ which generates the pp-type of m in M, that is, for all pp- formulas ψ, ψ ≥ ϕ if and only if m ∈ ψ(M). Conversely, if ϕ is a pp- n-formula, then there exists a finitely presented module M and m ∈ Mn such that ϕ generates the pp-type of m in M. See [Pre09, Section 1.2.2] for proofs. Any pp-pair ϕ/ψ, gives rise to a finitely presented functor F : mod− ϕ/ψ R → Ab which sends a right R-module M to ϕ(M)/ψ(M) and conversely, any finitely presented functor is isomorphic to one of the form F , [Pre09, ϕ/ψ 10.2.30]. Given a lattice L, let ∼ be the congruence relation generated by the simple intervals in L. We define what it means for a lattice to have finite m-dimension. For the more general ordinal valued dimension, see [Pre09, section 7.2]. We say that L has m-dimension 0 if L/ ∼ is the one point lattice. We say that L has m-dimension n+1 if L/∼ has m-dimension L. We say that a module has m-dimension n if its lattice of pp-definable subgroups has m-dimension n. AnembeddingofR-modulesf :M → N ispureifforevery pp-1-formula ϕ, f(ϕ(M)) = ϕ(N)∩f(M). A right R-module M is pure-injective if it is injective over all pure-embeddings. In section 7, we will use that a module is pure-injective if and only if it is algebraically compact [Pre09, 4.3.11]. An R-module M is said to be algebraically compact if every system of equations over R in arbitrary manyvariables withparameters inM suchthat every finitesubsystemhasa solution in M, has a solution in M. Equivalently, a module is algebraically compact if every collection of cosets of pp-definable subgroups which has the finite intersection property has non-empty intersection. We say that a module is Σ-pure-injective if it has no descending chain of pp-definable subgroups. Every Σ-pure-injective module is pure-injective. The (right) Ziegler spectrum of a ring R is a topological space with set of points, pinj , the isomorphism classes of indecomposable pure-injectives R and basis of open sets given by (ϕ/ψ) := {N ∈ pinj | ϕ(N) 6= ψ(N)} R where ϕ/ψ is a pp-pair. Our descriptions of Ziegler closures and Ziegler spectras in this paper are all based on the Ziegler spectrum of a discrete valuation domain V. Theindecomposablepure-injectives over V aretheindecomposablefinite- length modules V/ml for l ∈ N, the Pru¨fer module E(V/m) (that is, the injective hullof V/m), thecompletion V of V andQ(V)thefield of fractions of V. See for instance [Zie84, 5.1]. b A subset X of Zg is closed if the following two properties hold: V 4 LORNAGREGORY (1) If X contains infinitely many finite-length modules then X contains E(V/m), V and Q(V). (2) If X contains V or E(V/m) then X contains Q(V). b This description of the topology can be extracted from [Pre09, section b 5.2.1]. A definable subcategory of Mod −R is a full subcategory closed under directlimits, productsandpure-submodules. IfD isadefinablesubcategory then there exists pp-pairs ϕ /ψ , λ ∈ I such that M ∈ D if and only if λ λ ϕ (M) = ψ (M) for all λ ∈ I and conversely, all subcategories of this form λ λ are definable subcategories. Adefinablesubcategory of Mod−R is determined by theindecomposable pure-injective it contains [Pre09, 5.1.4]. Thus there is a bijective correspon- dencebetween closed subsetof Zg and definablesubcategories of Mod−R. R Definition 1.1. Let ϕ be a pp-n-formula in the language of right R-modules of the form ∃y¯(x¯,y¯)H = 0 where x¯ is a tuple of n variables, y¯ is a tuple of l variables, H = (H′ H′′)T and H′ (respectively H′′) is a n×m (respectively l×m)matrixwithentriesinR. ThenDϕisthe pp-n-formulainthe language I H′ x¯ of left R-modules ∃z¯ = 0. (cid:18) 0 H′′ (cid:19)(cid:18) z¯ (cid:19) Similarly, let ϕ be a pp-n-formula in the language of left R-modules of x¯ the form ∃y¯H = 0 where x¯ is a tuple of n variables, y¯ is a tuple of (cid:18) y¯ (cid:19) l variables, H = (H′ H′′) and H′ (respectively H′′) is a m×n (respectively m×l)matrixwithentriesinR. ThenDϕisthe pp-n-formulainthe language I 0 of right R-modules ∃z¯(x¯,z¯) = 0. (cid:18) H′ H′′ (cid:19) Note that the pp-formula a|x for a ∈ R is mapped by D to a formula equivalent with respect to T to xa = 0 and the pp-formula xa = 0 for R a ∈ R is mapped by D to a formula equivalent with respect to T to a|x. R Theorem 1.2. [Pre88, Chapter 8][Pre09, 1.3.1] The map ϕ → Dϕ induces an anti-isomorphism between the lattice of right pp-n-formulae and the lat- tice of left pp-n-formulae. In particular, if ϕ,ψ are pp-n-formulae then D(ϕ+ψ) is equivalent to Dϕ∧Dψ and D(ϕ∧ψ) is equivalent to Dϕ+Dψ. Let R be a k-algebra. We denote that standard dual Hom(−,k) of an R-module M, respectively a morphism f, by M∗, respectively f∗. Lemma 1.3. Let R be a k-algebra and M an R-module. If ϕ(M) ≤ ψ(M) then Dψ(M∗)≤ Dϕ(M∗). Thus if M is an R-module with dcc (respectively acc) on pp-definable subgroups then M∗ has acc (respectively dcc) on pp-definable subgroups. So, in particular, if the pp-1-lattice of M is finite-length then so is the pp- 1-lattice of M∗. ZIEGLER CLOSURES OF SOME UNSTABLE TUBES 5 Another key tool in this paper is the use of additive functors F : Mod− R → Mod −S which commute with direct limits and arbitrary products. Such functors are called interpretation functors. These are exactly the functorswhicharefinitelypresentedwhencomposedwiththeforgetfulfunc- tor to Ab and thus given by pp-pairs. This is where there name comes from [Pre97]. See [Kra98a, 7.2] and [Pre11, 12.9] for the equivalence. We will need the following facts about interpretation functors F : Mod− R → Mod−S. (1) There exists an n ∈ N, such that for all M ∈ Mod−R, there is a lattice embedding pp1(FM) ֒→ ppn(M). S R (2) If M is pure-injective then FM is pure-injective. See [Pre97, 3.16] or [Kra98a, 6.1]. (3) If M is Σ-pure-injective (respectively has the acc on pp-definable subgroups)then FM is Σ-pure-injective (respectively has the acc on pp-definable subgroups). Follows from (1). (4) If pp1(M) has m-dimension α then pp1(FM) has m-dimension less R S than or equal to α. Follows from (1) plus the fact that the m- dimension of pp1(M) is equal to the m-dimension of ppn(M). R R (5) If D is a definable subcategory of Mod−R then after closing under direct summands and isomorphism, FD is a definable subcategory. See [Pre12, 3.8]. (6) If the m-dimension of pp1 is α then the m-dimension of the smallest R definable subcategory containing the image of F is less that or equal to α. 2. Iterated one-point extensions of discrete valuation domains LetRbearing,kafieldandLak−R-bimodule. Theone-pointextension of R by L is the ring R 0 R[L]:= . (cid:18) L k (cid:19) RightmodulesoverR[L]maybeviewedastriples(M ,M ,Γ )whereM 0 1 M 0 is a k-vector space, M is a right R-module and Γ : M → Hom (L,M ) 1 M 0 R 1 is a k-homomorphism. A morphism between two triples (N ,N ,Γ ) and 0 1 N (M ,M ,Γ ) is given by a pair (f ,f ) where f : N → M is a k-vector 0 1 M 0 1 0 0 0 space homomorphism, f : N → M is an R-module homomorphism and 1 1 1 the following diagram commutes N ΓN // Hom (L,N ) . 0 R 1 f0 Hom(L,f1) M(cid:15)(cid:15) ΓM // Hom ((cid:15)(cid:15)L,M ) 0 R 1 There are two full and faithful embeddings of Mod−R into Mod−R[L]: F :Mod−R → Mod−R[L] M 7→ (0,M,0) 0 6 LORNAGREGORY and F :Mod−R → Mod−R[L] M 7→ (Hom(L,M),M,Id ). 1 Hom(L,M) The forgetful functor r : Mod−R[L]→ R (M ,M ,Γ) 7→ M 0 1 1 is right adjoint to F and left adjoint to F . 0 1 Each of these additive functors commute with direct limits and arbitrary products and thus are interpretation functors. Throughout the rest of this section, let V be a discrete valuation domain with maximal ideal m and let R := V, L = V/m =:k, 0 0 R 0 R := n n+1 (cid:18) Ln k (cid:19) and L = F L . n+1 1 n Finally, for each n ∈ N, let T := (k,0,0) ∈Mod−R . n n The category of finitely presented modules over a discrete valuation do- mainV is Krull-Schmidtandtheindecomposablefinitely presentedmodules are V and V/mn where n ∈ N. Our goal for the rest of this section is to prove the following theorem and thus classify the finitely presented right R -modules. n Theorem 2.1. The category of finitely presented right modules over R is n Krull-Schmidt. The indecomposable finitely presented modules over R are n of the form T , FmFn−mN where 0 ≤ m ≤ n and N is an indecomposable n 0 1 finitely presented module over V and Fn−k−lFlT where k + l ≤ n and 0 1 k 0≤ k,l. Lemma 2.2. If M = (M ,M ,Γ) is finitely presented then M is finitely 0 1 1 presented and M is finite-dimensional. 0 Proof. SupposeRl −→f Rm → M → 0 is a presentation for M. Applying n+1 n+1 the exact functor r we get that rRl −r→f rRm → M → 0 is a presenta- n+1 n+1 1 tion for M . Since rR is finitely presented, M is finitely presented. 1 n+1 1 AsamoduleoveritselfR is(k,R ⊕L ,Γ)whereΓ : k → Hom(L ,R ⊕ n+1 n n n n L )take 1 ∈ k tothehomomorphismwhichsendsl ∈ L to (0,l) ∈R ⊕L . n n n n Thus kl −f→0 km → M → 0 is exact and M is finite-dimensional. 0 0 (cid:3) In order to prove 2.1, we prove the following two conditions by induction. Note that 2.1 follows from B by induction on n. n A : If M ∈ mod−R , M a finite-dimensional k-vector space and Γ : n n 0 M → Hom(L ,M) is aninjective k-vector spacehomomorphismthenthere 0 n is a basis v ,...,v for M and orthogonal idempotent endomorphisms e of 1 n 0 i M such that e Γ(v ) = Γ(v ) and e M is indecomposable. i i i i ZIEGLER CLOSURES OF SOME UNSTABLE TUBES 7 B : All finitely presented modules over R are direct sums of modules of n n the form F M,F M and T where M is a finitely presented indecomposable 1 0 n module over Rn−1 and dimHom(Ln,N) ≤ 1 for all indecomposable finitely presented modules N over R . n Lemma2.3. IfK ∈ Mod−R thenHom(F L ,F K)= 0. HenceF F K ∼= n 1 n 0 1 0 F F K. 0 0 Proof. Suppose(f ,f )isahomomorphismfromF L toF K. Thenf = 0 0 1 1 n 0 0 and hence f ◦1 = 0. So f = 0. (cid:3) 1 Ln 1 Remark 2.4. Suppose M ∈ mod − R , M a finite-dimensional k-vector n 0 space, Γ :M → Hom(L ,M) is an injective k-vector space homomorphism 0 n and α is an automorphism of M. There is a basis v ,...,v for M and 1 n 0 orthogonal idempotent endomorphisms e of M suchthat e Γ(v ) = Γ(v )and i i i i e M is indecomposable if and only if there is a basis v ,...,v for M and i 1 n 0 orthogonal idempotent endomorphisms e of M such that e αΓ(v )= αΓ(v ) i i i i and e M is indecomposable. i Lemma 2.5. If A holds then all finitely presented right R -modules are n n+1 direct sumsof modules of the form T := (k,0,0), (0,M ,0)and (k,M ,Γ) n+1 1 1 where M is a finitely presented indecomposable right R -module and Γ is 1 n an injective k-vector space homomorphism. Proof. Let (M ,M ,Γ) be an R -module. If (M ,M ,Γ) is finitely pre- 0 1 n+1 0 1 sented then M is finitely presented and M is finite-dimensional by 2.2. If 1 0 Γ is not injective then ∼ (M ,M ,Γ) =(kerΓ,0,0)⊕(M /kerΓ,M ,Γ). 0 1 0 1 So, without loss of generality, we may assume Γ is injective. By A there n exists a basis v ,...,v for M and orthogonal idempotent endomorphisms 1 n 0 e ,...,e of M such that e Γ(v ) = Γ(v ). For each 1 ≤ i≤ n, let t : M → 1 n i i i i 0 M be a k-linear map such that t (v ) = v and t (v ) = 0 if i 6= j. So 0 i i i i j (t ,e ),...,(t ,e ) are orthogonal idempotents for (M ,M ,Γ). Thus 1 1 n n 0 1 n n (M ,M ,Γ) = (0,(1− e )M ,0)⊕ (v k,e M ,Γ| ) 0 1 i 1 i i 1 vik Xi=1 Mi=1 as required. (cid:3) Lemma 2.6. For all n ∈ N, dimHom(L ,L ) = 1. n n Proof. SincedimHom(L ,L ) = 1andF isfullandfaithful,dimHom(L ,L ) = 0 0 1 n n 1 for all n ∈N. (cid:3) Lemma 2.7. Let M be a k-vector space, v ,...,v a basis for M and 0 1 m 0 Γ : M → Hom(L ,Tl) be an injective k-vector space homomorphism. 0 n n There exist e ,...e orthogonal idempotent endomorphism of Tl such that 1 m n e Γ(v )= Γ(v ) and e Tl is indecomposable. i i i i n 8 LORNAGREGORY Proof. Let (α ,0) = Γ(v ),...,(α ,0) = Γ(v ) where each α is a k-linear 1 1 m m i mapfromHom(Ln−1,Ln−1)tokl. SinceΓisinjectiveanddimHom(Ln−1,Ln−1)= 1, it follows that l ≥ m and α (Id ),...,α (Id ) ∈ kl are linearly 1 Ln−1 m Ln−1 independent. Let ǫ ,...,ǫ be the idempotent endomorphisms of kl such 1 m that ǫ α (Id )= α (Id )and dimimǫ = 1. Let e = (ǫ ,0),...,e = i i Ln−1 i Ln−1 i 1 1 m (ǫ ,0). By definition, e ,...,e have the required properties. (cid:3) m 1 m Proposition 2.8. For all n ≥ 1, An−1 and Bn imply An. Proof. Let M be a finitely presented module over R , M a k-vector space 1 n 0 andΓ :M → Hom(L ,M ) injective. By B , M = F N⊕F K⊕Tl where 0 n 1 n 1 1 0 n K,N ∈ mod−Rn−1 and l ∈ N0. By 2.3, Hom(F1Ln−1,F0K)= 0, so we may as well assume K = 0. Let v ,...,v be a basis for M such that 1 n 0 Γ(v ) = (f ,w ),...,Γ(v ) = (f ,w ) 1 1 1 m m m and Γ(v )= (0,w ),...,Γ(v ) = (0,w ) m+1 m+1 n n where f ,...,f ∈ Hom(L ,F N) are linearly independent over k and 1 m n 1 w ,...,w ∈ Hom(L ,Tl). 1 n n n Since F1 is full, there exist f1∗,...,fm∗ ∈ Hom(Ln−1,N) such that F1f1∗ = f ,...,F f∗ = f . 1 1 m m Let w∗,...,w∗ be such that w = (t ,0) and w∗ = t (Id ). Note that 1 n i i i i Ln−1 since Hom(Ln−1,Ln−1) is 1-dimensional, ti and hence wi is determined by w∗ = t (Id ). i i Ln−1 Here is a diagram for Γ(v ): i Hom(Ln−1,Ln−1) Id // Hom(Ln−1,Ln−1) g g 7− 7− → → (fi∗◦g,ti(g)) fi∗◦g (cid:15)(cid:15) (Id,0) (cid:15)(cid:15) Hom(Ln−1,N)⊕kl // Hom(Ln−1,N) Let α = (α ,α ) : F N ⊕ Tl → F N ⊕ Tl be such that α ◦ (f∗,0) = 0 1 1 n 1 n 0 i (f∗,−w∗) for i = 1,...,m, α ◦(0,w∗) = (0,w∗) for i = m+1...,n and i i 0 i i α = Id . 1 N Soα◦Γ(v ) = (f ,0),...α◦Γ(v ) = (f ,0)andα◦Γ(v ) = (0,w ),...,α◦ 1 1 m m m+1 m+1 Γ(v )= (0,w ). By 2.4, we may replace Γ by α◦Γ. n n Let M′ be the span of v ,...,v . By definition of M′, if u ∈ M′ then 0 1 m 0 0 Γ(u) = (Γ (u),0) ∈ Hom(L ,F N)⊕Hom(L ,Tl) = Hom(L ,F N ⊕Tl). 0 n 1 n n n 1 n Let ∆ : M0′ → Hom(Ln−1,N) be defined by setting F1∆(u) = Γ0(u). By An−1, there exists e1,...,em orthogonal idempotent endomorphisms of N such that e ∆(v ) = ∆(v ) and e N is indecomposable for 1 ≤ i ≤ m. Thus i i i i F (e )Γ(v ) = Γ(v ) and F (e )F (N)= F (e N) which is indecomposable. 1 i i i 1 i 1 1 i ZIEGLER CLOSURES OF SOME UNSTABLE TUBES 9 Let M′′ be the span of v ,...,v . By 2.7, there exist e ,...e 0 m+1 n m+1 n orthogonal idempotent endomorphism of Tl such that e Γ(v ) = Γ(v ) and n i i i e Tl is indecomposable. i n So σ = (e ,0),...,σ = (e ,0) and σ = (0,e ),...,σ = (0,e ) 1 1 m m m+1 m+1 n n are orthogonal idempotent endomorphismsof F N⊕Tl such that σ Γ(v ) = 1 n i i Γ(v ) and σ (F N ⊕Tl) is indecomposable. (cid:3) i i 1 n Proposition 2.9. For all n ∈N, B and A imply B . n n n+1 Proof. By 2.5, A implies that each finitely presented module over R is n n+1 a direct sum of modules of the form T ,(k,M ,Γ) and F M = (0,M ,0) n+1 1 1 1 1 where Γ is injective and M is an indecomposable R -module. So in order 1 n to show that the first clause of B is true, we need now consider modules n+1 of the form (k,M ,Γ). By B , M is either F N, F K or T where N,K 1 n 1 1 0 n are indecomposable Rn−1-modules. Since Hom(L ,F K) = 0, if M = F K then Γ is not injective. n 0 1 0 IfM = T then(k,T ,Γ)isisomorphictoF T sincedim Hom(L ,T )= 1 n n 1 n k n n 1 i.e. (Γ,Id ): (k,T ,Γ) → F T is an isomorphism. Tn n 1 n Now suppose that M1 = F1N for N an indecomposable Rn−1-module. Since Γ : k → Hom(L ,F N) is injective, dimHom(L ,F N) 6= 0. By n 1 n 1 Bn−1, dimHom(Ln−1,N) ≤ 1. So dimHom(Ln,F1N) = 1. Thus (Γ,IdN) : (k,N,Γ) → F N is an isomorphism. 1 It now remains to show that dimHom(L ,M) ≤ 1 for all indecom- n+1 posable M ∈ mod − R . If M = F K then dimHom(L ,M) = 0. n+1 0 n+1 If M = F N then dimHom(L ,M) = dimHom(L ,N) ≤ 1 by B . If 1 n+1 n n M = T then dimHom(L ,M) ≤ 1 follows from 2.6 and the definition n+1 n+1 of T . n+1 (cid:3) We now consider the base cases, A and B . 0 1 Lemma 2.10. Let V be a discrete valuation domain with maximal ideal m, M a finitely presented V-module and M ⊆ Hom(V/m,M) a V/m-vector 0 subspace. There exists v ,...,v a basis for M as a V/m-vector space and 1 n 0 orthogonal idempotent endomorphisms of M such that e M is indecompos- i able and e v = v for 1 ≤ i≤ n. i i i Proof. Since V is a principal ideal domain, M = M′ ⊕Vm where M′ is a torsion module. Since Hom(V/m,V) = 0, we may replace M by M′. Let W = {f(a) ∈ M |a ∈ V/m and f ∈M }. NotethatW isasubmoduleofM. 0 Let p ∈ V generate m and let L = {m ∈ M | 0 6= mpl ∈ W for some l ∈ N}. Since V is an RD-ring (see [Pre09, section 2.4.2]), L is pure in M. Since M and hence L is finite-length, L is pure-injective. Therefore L is a direct summand of M. The lemma now follows from the structure theorem for finitely generated module over principle ideal domains. (cid:3) 10 LORNAGREGORY Lemma 2.11. B holds. 1 Proof. We need to show that all finitely presented modules over R are 1 direct sums of modules of the form F M := (0,M,0), F M and T := 0 1 1 (k,0,0) whereM is a finitely presented indecomposable moduleover V. Let (M ,M,Γ) be an arbitrary finitely presented module over R . As usual, we 0 1 may assume Γ is injective. Letting e ,...,e be as in 2.10, we have that 1 n n n (M ,M,Γ) ∼= (0,(1− e )M,0)⊕ (v k,e M,Γ| ). 0 i i i vik Xi=1 Mi=1 Themodule(0,(1− n e )M,0) is equaltoF (1− n e )M andforeach i=1 i 0 i=1 i 1≤ i ≤ n, (vik,eiMP,Γ|vik) is isomorphic to F1eiM.P That Hom(L ,N) is 1-dimensional is proved exactly as in the proof of 1 2.9. (cid:3) Corollary 2.12. For all n ≥ 0 and m ≥ 1, A and B hold. n m Corollary 2.13. All indecomposable finitely presented modules over R n have local endomorphism rings. 3. Indecomposable pure-injectives and the Ziegler spectrum Throughout this section, let V be a valuation domain with maximal idea m. Lemma 3.1. All finitely presented indecomposable right R -modules are n+1 pure-injective except for Fn+1V. The module Fn+1V is pure-injective if and 0 0 only if V is pure-injective as a right module over itself. Proof. This follows directly from 2.1, the fact that the functors F and F 0 1 preserve pure-injectivity and that (k,0,0) is finite-length and hence pure- injective. (cid:3) Proposition 3.2. Every indecomposable pure-injective module over R n+1 is of the form F N, F N or T := (k,0,0) for some indecomposable pure- 0 1 n injective R -module N. n Proof. Since all finitely presented indecomposable modules over R are n+1 pure-injective except Fn+1V, the set of finitely presented pure-injective 0 modulestogether with Fn+1V isdensein Zg . SinceF andF commute 0 Rn+1 0 1 with direct limits and products, the images of F and F are definable sub- 0 1 b categories of Zg after closing under direct summands (see [Pre12, 3.8]). Rn+1 Thus Add(imF )∩Zg =: C and Add(imF )∩Zg =: C are closed 0 Rn+1 0 1 Rn+1 1 subsets of Zg . The point (k,0,0) is a closed point of Zg . Since Rn+1 Rn+1 C ∪ C ∪ {(k,0,0)} is a dense closed subset of Zg it is all of Zg . 1 2 Rn+1 Rn+1 Thus all indecomposable pure-injecitve R -modules are of the required n+1 form. (cid:3) Theorem 3.3. The indecomposable pure-injective modules over R are n

See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.