Download as pdf
Download as pdf
You are on page 1of 48
ALGEBRA Lectures on rings and modules Bo Stenstrém 5 Wir 8 & ear Me SuphigeS % € % S Va eX Matematiska institutionen November 2001 Bo Stenstrém ALGEBRA: Lectures on rings and modules Contents! 1. Modules and algebras 2. Submodules and homomorphisms of modules 3. Free modules 4. Direct product and direct sum 5. The group of module homomorphisms 6. Exact sequences 7. Punetors and preservation of exactness 8. Bimodules 9. Noetherian rings and modules 10. Semisiple modules 11. Semisimple rings 12, Finitely generated modules over a PID 13. The Jordan canonical form 14. Tensor products 15. The tensor product as a functor . Algebras Flat modules Exterior powers ts and answers to some exercises 1. Modules and algebras ‘The concept of “module” will play a central role in this course. It comprises several important algebraic structures, as vector spaces, abelian groups, and ideals. The theory of modules may be viewed as linear algebra in a general sense, but with considerable features of ring theory. By a “ring” we shall always mean an associative ring with identity clement 1, and ring homomorphisms are required to preserve identity elements Let A be a ring. A (left) A-module is an abelian group M with a mapping Ax M —+ M, written as (a,2) + az, such that: 1) a(z+y) 2) (a+be 3) (ab)z = a(bz) 4) lease for all a,b € A and z,y € M. From axioms 1)-4) follow, as in the ease of rings, the rules: a) 0-2=0 and a-0=0, @ a(—z) = (~a)z = ~(az) ‘One usually writes 0 also for the module {0} consisting of only a 0 element Examples of modules 1. Every real ot complex vector space is an R-module or C-module respectively. In general one uses the term vector space for modules over division rings (because of their property of always having bases, as we shall later prove). 2, Every abelian group M is a Z-module, with nesttztert2 (n times) for n > 0, and with the obvious modification for n <0. Conversely, if M is a Z-module, then for n> O one has nea(ltl4e-¢)ealetlet--tleaztet- te, So we see that a Z-module is essentially the same thing as an abelian group. 3. Let A be an arbitrary ring. Then A may be viewed as an A-module by using the ring multiplication as the mapping A x A— A, More generally, if A is a subring of a ring B, then B may be viewed as an A-module in the corresponding way. 4. Let V be a real vector space with a given linear mapping 1: V —» V. ‘Then V may be module over the polynomial ring R[é), where R{é] x V —> V is defined as (ro tritt--trgt™)z = rox + mT (2) +--+ tT (2). wed as We have defined left A-modules. Similarly one defines a right A-module as an abelian group M with a mapping M x A— M, written as (z,a)+~+ za, such that: 1) (e+y)a=20+ya 2) (a+b) =2a426 3) 2(ab) = (2a)b 4) r-l=2 for all a, € A and z,y € M. Note that it is 3) that makes the basic difference betweeen left and right modules. If A is a commutative ring, then there is no real difference between left and right modules. In the general case we can introduce the opposite ring A°P, which has the same elements and the same addition as A, but where multiplication + is given by ab = b-a, where - is multiplication in A. Then a right A-module is the same thing as a left A°-module. In general we shall be working with left modules. We write 4M (resp. M,) to indicate that M is a left (or right) A-module. Algebras Let 2 be a commutative ring (in many applications it is either a field or the ring Z). An R-algebra is a ring A which also is an R-module, with the rule +(ab) = arb) = (ra)b for all r € R, a,b € A. (More precisely one might call it a “unital and associative linear R-algebra”) Equivalently one may regard an R-algebra as a ring A together with a ring homomorphism : R—+ A 2 whose image lies in the center of A. Often ¢ is injective (e.g. always when R is a field), and in that case one usually regards R as a subring of A by identifying r € R with r-1@ A. A module over an R-algebra may always be considered also as an R-module (ef. Exercise 2). Examples of algebras 5. Every ring is in a unique way a Z-algebra, 6. The polynomial ring Rf] is an Realgebra. 7. ‘The matrix ring Ma(I) over a commutative ring Ris an Realgebra. 8. Let G be a monoid (ie, a set with an associative multiplication having an identity element 1) For an arbitrary commutative ring R we define the monoid ring R{G] as a set of formal linear combinations RIG) = {kya | hy ER, by = 10 (“Almost all” means “all with the exception of finitely many”). Addition and multiplication in R{G] is given by Dot Deo= Dib eho Dee kee with = Tee. for almost all 9}. ‘The algebra axioms are easily verified for R{G]. ‘The multiplicative monoid G = {1,1,t?,...} is isomorphic to the additive monoid N by t" ++ n, and R[G] = {Tknt™ | kn € K, almost all ky = 0} is the polynomial ring R{fj. Similarly one obtains the polynomial ring R{ty,...,t,] as the monoid ring RG] with G =x N--- x N (r times) R{2] is the ring of “Laurent polynomials” S72, kyt” We shall as an application see how modules can be used in the theory of group representations. A ‘matrix representation of a group G is a group homomorphism 9: G+ GL(n,C), where GL(n,C) denotes the group of invertible n x n-matrices over C (for a certain fixed n). ‘The representation is faithful if is injective. Every quadratic n x n-matrix gives in a well-known way a linear mapping C” —» C”, so the representation gives an invertible linear mapping (g): C” — C” for every g € G. ‘Thus the group G operates on C" by g-t=9(9)2, 2EC", gE. One might say that C” in this way becomes a G-module. If we now “linearize” G by introducing the group algebra C[G], then €* becomes a C[G}-module: Dho2= Dk doz, — where ky CC, 9G, 2 EC” Suppose conversely that M is a C[G]-module which is fnite-dimensional as a vector space over C, s0 M&=C". Then M gives a representation ¢: G+ GL(n,C) by Wg): tr g-2, 2EM=C". We have thus found that matrix representations of G correspond to C[G}-modules which are finite- dimensional as complex vector spaces. Exercises 111 Show that if we let the matrix ring, Ma(R) operate to the left on IR" (whose elements are considered as column vectors), then IR" becomes a left module over Mn(IB). Show similarly that operation to the right on row vectors gives IR" the structure ofa right module over Mn(R). 1.2 Let @: A — B be a ring homomorphism. Show that every left B-module M may be considered as a left A-module by the rule ax = a(a)z for a € A, 2 M. 1.5 Write down the multiplication table for the elements of the froup ring Za[Gi, where G is a cyclic group of order three. 14 Show that a left R{G)-module M, for a group ring G, also may be viewed as a right R[G]-module with 2g= 92 for g€G,2€ M. 2. Submodules and homomorphisms of modules For every kind of algebraic structure one has a corresponding concept of homomorphism, meaning map- pings which respect the relevant structure. Let M och N be left A-modules. A (module) homomorphism e: M+ N is a mapping such that vz+y)=9(z)+¥(y), — e(az) = ap(z) for 2, € M, a € A. In particular g is a homomorphism of abelian groups. A homomorphism of A-modules is also called an A-linear mapping, ot a linear transformation of A-modules. Composition ve of homomorphisms yp: M —» Nand y: N — P is a homomorphism M —+ P. For every module M there is an identity homomorphism Idyy mapping x ++ z for every z € M. A module homomorphism is an isomorphism if it has an inverse (which then also is a homomorphism); as usual one has that a homomorphism is an isomorphism if and only if it is bijective. We write M ~ N to indicate that two A-modules M and NV are isomorphic. A homomorphism M — M is called an endomorphism, and an isomorphism M—+ M is called an automorphism, Let M be a left A-module. A subset L of M is a submodule of M if 1 is an additive subgroup of M and a € A, € L imply that az € L. In such cases 1 is itself a module, and the canonical mapping L— M isa module homomorphism. (A mapping. is informally called “canonical” if it is defined in the ‘way which in the context is the natural one). Let L be a submodule of M. Form the quotient group M/L and define Ax M/L + M/L as (a, 2]) ++ [a2]. ‘The mapping is well-defined, because if x and y are in the same coset, then [2] =) = 2-yeL + a(z-y) EL = [az] = [ay ‘The module axioms are easily verified for M/L, which is called a quotient module of M. ‘The canonical mapping M — M/L is a module homomorphism Examples ~ 1. Consider A as a left A-module. A submodule a of A is an additive subgroup of A such that a € A, z € a= az € a; we say that ais a left ideal in A. A subgroup of A which is both a left and a right ideal is a (two-sided) ideal. If ais a left ideal, we can form a left A-module A/a, with a. [2] well-defined for a,b € A. If ais a two-sided ideal, we can go further and consider A/a as a ring since {a} [8] then is defined. 2. Consider the situation in Sec. 1, Example 4. If V is regarded as an R{f]-module, then a submodule L of V is an additive subgroup of V such that Det" eR, ze L > Pk T*(2) € L. A submodule of V is thus a subspace such that T(L) C L. Let g: M — N be a homomorphism. We define its kerfiel as Ker = {2 € M | (ez) = 0} and its image as Imp = {y(z) | z © M). One verifies that Kerg is a submodule of M and that Im is a submodule of NV. One says that ¢ is a monomorphism if y is injective, i.e. Ker = 0, and that g is an epimorphism if ¢ is surjective, ic. Im = N. We shall now prove some important isomorphism theorems. A homomorphism yp: M —» N induces a homomorphism %: M/Ker p — Img defined as g: [2] ++ (2). Here @ is welldefined since [2] = [y] => 2—y € Kerp = (2) = p(y). We get a commutative diagram (ie. a diagram where the composed mappings will be the same whatever path one follows) | hal M/(Ker g) —5—+lmy ‘The mapping # is evidently bijective, soit is an isomorphism, and one gets: Theorem 2.A Every homomorphism y: M—+ N induces an isomorphism @: M/Kerp — Imp. Theorem 2.B If LC M CN are submodules, then M/L is a submodule of N/L, and (N/1)/(M/L) = N/M. Proof. We consider the homomorphism y: N/L—» N/M given as (z]s, > [z]ar, where indices L and M indicate that we take cosets with respect to L and M respectively. Note that ¢ is well-defined since LCM, We have Ker = {[e]c | lel = 0} = {[e]e |z €M)} = M/L. 4 So M/L is a submodule of N/L, and Theorem 2.A gives (N/L)(M/L) = (N/L)/Kere = N/M. Consider now instead the situation when we have two submodules L and M of NV. We then define L+M={z+y|zeLyeM)}, which is easily seen to be a submodule of W. ‘Theorem 2.C If L and M are submodules of N, then (L+ M)/M = L(LnM) Proof. We use the homomorphism y: L + (L-+M)/M given as z+ [z]. It is surjective since [2-+y] = [z] for € L, ye M. We have Kerp={z€L|[2]=0)={zeL|zeM}=LnM. From this follows that £1 M is a submodule of N, and ‘Theorem 2.A gives L[(LAM) = L/Kerp=(L+M)/M. Let M be a left A-module. For every subset S of M there is a least submodule L of M such that SCL. For define L as the set of all finite sums T>a;x; with a; € A and x € S (if S = 0 one takes L = 0). One easily verifies that L is a submodule of M, and $C L since 1+z € L for z € S. Every submodule of M containing S must also contain these linear combinations, so Lis the least submodule of M containing S. We write L = (S), and say that S is a set of generators for L. If in particular the ‘module M itself has a finite set of generators, then it is said to be finitely generated. A module generated by a single clement is cyclic (the terminology is known from group theory) Let us determine the cyclic A-modules. If M is a cyclic left A-module, generated by the element 2, then M = Az. Consider the module homomorphism A —+ M given by a ++ az, which of course is surjective. Its kernel is {a € A | az = 0}, which is a left ideal in A, called the annihilator of z and denoted by Ann(2). From Theorem 2.A follows that M ~ A/Ann(z). Conversely, for every left ideal a in A one has that A/a is a cyclic A-module, generated by [1]. Examples ~ 3. ‘The cyclic submodules of 4A are the principal left ideals in A. 4. IfV is a eyclic nonzero vector space over division ring D, then V © D since (0) is the only proper left ideal of D. 5. A submodule of a finitely generated module M is not necessarily finitely generated; consider e.g, left ideals in a polynomial ring with infinitely many indeterminates (see See. 9). On the other hand, if M has a finite set 5 of generators, then also every quotient module M/L is finitely generated, with {[2] | 2 € 5} asa set of generators. As an application of the preceding theory we shall consider the concept of torsion. Let A be a commutative ring and denote by $ the set of elements # 0 of A which are not zero-divisors. Note that the set S is closed under multiplication Let M be an A-module. An element z € M is a torsion clement if there exists s € S such that sr = 0. ‘The torsion elements form a submodule T(M) of M, because if sz = 0 and ty = 0, then st(z+y) = 0 with et € S; furthermore saz = 0 for any @€ A. Ia: M— N is Aclincar, then it maps torsion elements to torsion elements, so it restricts to a homomorphism T(M) —+ T(N), which we denote by T(a). ‘The module M is torsion-free if T(M) = 0. We note that the quotient module M/T(M) is torsion-free for any module M, because if s{z] = 0 in M/T(M) for some s € 5, then [sz] = 0, so sz € T(M), which means that {(s2) = 0 for some t € S, and hence z €T(M), ie. [2] =0. Exercises 21 Let G be a finite cyclic group of order n. Show that there is an isomorphism K{G] © K{t/(¢" — 1) of Kealgebras for every field K. 2.2 Show that if Land AM are submodules # 0 of the Zmodule Q, then A M #0. Also show that the corresponding is not true for Q/Z. 2.3 Show that if K, Land M are submodules of a module WV with M CK, then Kn (L+M) = (KN) +M (‘modular law”). 24 Show that if L is a submodule of a module N, then every submodule of N/L can be written as M/L for a uniquely determined submodule M of N with CM. 25 Let y: M —- N bea homomorphism of modules and La submodule of N. Show that: ) o71(L) isa submodule of I ¢ induces an injective homomorphism M/p"1(L) —+ N/L. (Hint: cf. the special case [= 0). 5 2.6 Suppose Lis a submodule of M and that y: M —+ N is a module homomorphism with Ker 2 L. Show that there exists a uniquely determined homomorphism M/L — N such that the diagram M——+MjL commutes, where Mf —+ M/L is the canonical homomorphisn 2.7 Let M be aleft A-module and «a two-sided ideal in A. Define aM asthe set ofall finite sums > aizs with 4 € Gand x; € M. Show that GM is a submodule of M, and that M/aM may be viewed ax « module over the quotient ring A/a by defining [a] [2] = [ax] (for the respective cosets) 2.8 Show that Q as a Z-module is generated by {1/p" | p prime, n > 0}, and that every finitely generated submodule of Q ia cyclic. 2.9 Let L be submodule of a module M. Show that if L and M/L are finitely generated, then M is finitely generated, 2.10 Let M be a module over a commutative ring A. For every subset X of M one defines Anna(X)=(a€ Al ax = 0,¥e € X}, and for every subset Sof A one defines Anny(S) = {z € M | az = 0,¥a € 5}. Show that: (@) Ann4(X) is an ideal in A, and Anny(S) ie « submodule of Af Gi) XC Anmas(Anna(X)) and $C Anna(Annas(S)). When does equality hold here? 2.11 Let M be a module over a commntative ring A. If K and L are submodules of M, let Z : aK CL}. Show that: () Le Kis an ideal in A. (i) lz eM, then LO Az =(L: Az). 2.12 Let R be a commutative ring. Show that every module M over Rt] is obtained by the construction in Example 4 of Sec. 1, ic. as an R-module equipped with an R-linear endomorphism T. {eed 3. Free modules We shall generalize the well-known concept of basis from finite-dimensional vector spaces to general modules. Let M be a left A-module. A subset S of M is linearly independent if az) + +--+ anzy = 0 implies that a; = 0 for i= 1,...,n, where 21,...,2q €S with x #2; fori fj and gC A. ISisa linearly independent set of generators for M, ic. if every element z 4 0in M can be written in a unique Haq te tanzy with a #0 in A, 4 ES, then $ is a basis for M. The module M is called free if it has a basis. Examples ~ 1. ‘Let A be an arbitrary ring We may consider AY = {(a1,...,dq) | a; € A} as left A-module, where addition and muliplication by scalars are defined component-wise. ‘This module is free with a basis consisting of the vectors ¢: = (1,0,...,0), ... 5 én = (0,--.,0)1), since (a1,---1da) = ares +...4neq. Conversely, if F is a free left A-module with a finite basis 21,...,2n, then F & A” under an isomorphism given by 2j ++ ¢; for i= 1,...,n. 2. If A is a commutative ring, then the polynomial ring Ald] is a free A-module, with infinitely many basis elements 1, ¢,02,¢3,.. 3. A free Z-module is the same thing as a free abelian group. The abelian group Q is torsion-free but not free since every linearly independent subset consists of at. most one element. 4. Let A be an arbitrary ring and $ an arbitrary set. ‘The free A-module on $ is defined as the set of all finite formal linear combinations >, a;s; with a; € A and s € S, where addition and ‘multiplication with scalars are defined formally in the obvious way. By construction this really becomes a free A-module, and the elements s € S can be identified with the elements 1- in the module, so $ ean be viewed as a basis for the module. A special case of this construction was used for the monoid ring in See. 1 Zorn’s lemma For the proof of existence of a basis for any vector space one needs some version of the axiom of choice, ‘This axiom is a set theoretical principle which is not intuitively obvious, and which in axiomatizations of set theory is added to the more evident axioms. Axiom of choice For every set $ there exists a function J such that f(z) € x for every nonempty reS, ‘The function f in question is called a choice function for S. Often one uses a choice function for the set $ = P(A) of subsets of a set A, in which case f is a function with {(B) € B for every nonempty subset B of A. In algebra the axiom of choice is most often used in the form of a maximal principle, called Zorn’s lemma. In order to formulate this we remind the reader about some terminology regarding ordered sets. A relation < on a set M is called an ordering (or sometimes a partial ordering) if it satisfies: 1) 2€M => 2 2<5; 3) 2 z 2=y orally € M. Zorn’s lemma Suppose M isa nonempty partially ordered set where every chain has an upper bound. ‘Then there exists a maximal element in M. Zorn’s lemma can be proved with the use of the axiom of choice, and the axciom of choice is conversely consequence of Zorn’s lemma (see any text in set theory). Intuitively one can motivate Zorn’s lemma through the use of the axiom of coice and transfinite recursion (i.e. recursion over the ordinal numbers) in the following way: suppose every chain in M has an upper bound but M does not have any maximal clement. Choose 2 € M. ‘There exists y such that y > 2, and then z such that z > y, etc. transfinitely This gives a chain C, which has an upper bound w € M. But w is then maximal in M for otherwise one could continue C with an clement > w. As a typical application of Zorn’s lemma we show: ‘Theorem 3.A Every proper left ideal in a ring is contained in a maximal left ideal. Proof. Suppose a is a left ideal. Let M be the set of those proper left ideals of A which contain a, partially ordered under inclusion. M is nonempty since a € M. Let € be a chain in M. Let 6 be the union of all ¢ € C. Then 6 is a left ideal, because if z,y € 6, then z € ¢and y € ¢ for some ¢,¢ € C; if now eg. ¢C ¢, then +y € ¢; further itis clear that az € 6. In order to know that 6 is an upper bound for € we must also show that 6 is a proper left ideal, ic. that 1 ¢ 6, which is an immediate consequence of 1 ¢ ¢for all cE C. Zom’s lemma can now be applied, and it gives the existence of a maximal element in M, ie. a maximal left ideal containing . A use of the axiom of choice or Zorn’s lemma means that a non-constructive feature is introduced in the reasoning, and therefore one should try to avoid them in cases where it is possible. But for a major part of modern mathematics they are indispensable tools. Example. Let A be the ring of bounded continuous real-valued functions on the real line. We have the well-known maximal ideals of functions which are 0 at a certain fixed point, but from ‘Theorem 3.A one sees that there are also other maximal ideals that are not accessible for an expli description. Bases in vector spaces Let X be a division ring. A K-module has been called a vector space over K. ‘The following theorem shows that every vector space V has a basis (by taking Z v): Theorem 3.B Let V be a left vector space over K. If $ is a set of generators for V and L CS is linearly independent, then there exists a basis B for V with LC BC S. Proof. Let M be the set of linearly independent sets L’ with LC L/ C $. Order M under inclusion Since L € M, we have M # 0. Let C be a chain in M. Define L as the union of all L € C. ‘Then LCLCS. Tose that L is an upper bound for ¢ it remains to show that L is linearly independent. Suppose 7}, aizi = 0 with 2; € L. Since only a finite number of 2; are involved here, and since C is ‘a chain, there exists L” € C such that 21,...,2n € L". But since 1" is linearly independent, we have a = dn = 0. So L € M and is an upper bound for C. Zorn’s lemma gives the existence of a maximal element B € M We now have to show that B generates V, ie. that every 2 € V is a linear combination of elements from B. We may assume z ¢ B. Because of the maximality of B we have that B U {z) is linearly dependent, so az + 5 aiz; = 0 for a # 0 and 2; € B. But from this follows that 2 = —S>a~!-a;xy as desired. ‘The following result should be well-known, but is included for the sake of completeness: Theorem 3.C Ifa vector space V has aset $ ofn generators, and 21,...,2m are m linearly independent elements in V, then m a,s € M correspond to (a, | s € S) in ACS). In particular we may more precisely than in Example 3.4 define the free A-module on a set Sas ACS) Let M be a module and (M; | i € I) a family of submodules of M. We define (My | i € 1) as consisting of all sums So, zs with 2; € Mj and z= 0 for almost all i. This is obviously a submodule of -M, and it can be described as the least submodule of M containing all My; one may also say that it the submodule of M generated by U; Mz (we have earlier in Sec. 2 discussed the sum My + Mz of two submodules). We can always define a homomorphism Qu-TM fer tal as (2; | € 1) + Dy ze. This homomorphism is always surjective, but it isin general not injective. When it is injective, and thereby an isomorphism, we say that 5°, M; is the inner direct. sum of the family (Mi; | i € 1). In practice one usually does not distinguish, neither notationally nor terminologically, between outer and inner direct sum, ‘Theorem 4.A The sum My + Mz is direct if and only if Mi Mz = 0. Proof. In this case we have that p: Mi @Mz > My+Mz sends (21,2) to 21+22, and Ker y = { (21,23) | 22) }, 60 we see that Kery=0 => MyM Mz =0. 0 A submodule L of a module M is called a direct summand if M = L@ L! for some submodule L! of M; the submodule L! is said to be a complement of L in M (it is of course not uniquely determined). A nonzero module M is indecomposable if it has no direct summands other than {0} and M itself. ‘Theorem 4.B Every subspace of a vector space is a direct summand. Proof. Suppose L is a subspace of V. Choose a basis B for L. Since B is linearly independent, there exists by Theorem 3.B a basis B’ for V with B CB’. If we now let BY = B! \ B and let L" be the subspace generated by BY, then V = L4L" and LOL" =0,s0V=L@L". § ‘The indecomposable vector spaces are thus the one-dimensional ones. Another important observation ‘Theorem 4.C Every module is a quotient module of a free module. Proof. Let $ be any set of generators for M (we could choose $= M, but that would be very uneco- nomical). Define a homomorphism y: A‘) + M as (a,),es ++ TD, dys. It becomes surjective since S generates M, and we obtain by Theorem 2.A that M = A()/Ker yp. (We may view Ker as a module of “relations between the generators in 5”). i Example ~ 2. Suppose a left ideal ais a direct summand of the ring A. So we have A = a@ 6 for some left ideal b, and correspondingly we can write 1=e+ f with e € aand f € 6. Multiplying by any a € awe get a= ae +af, and af = a—ac € afb = 0. Hence a = ae, and so ais a principal left ideal generated by ¢. We also see that ¢ is an idempotent, i.e. ¢ = e2 Conversely, if e is an idempotent, then A = Ae @ A(1 —e) because every 2 € A can be written = 2e-+2(1~e), and Aen A(1 —e) = 0 since ze = y(1 —e) gives ze = ze -e = y(1—e)e = 0. We may thus conclude that left ideals are direct summands if and only if they are of the form Ae for some idempotent e. Exercises 4.1 Show that Z/62.% 2/22 2/32, as Z-modules, and as ideals in the ting 2/62. 4.2 Suppose K CL are submodules of s module M. Show that if K ie a direct summand of M, then it is a direct summand also of 1. 4.3 Suppose there exist homomorphisms g: M+ N and ys N+ M such that go ¥ = Idyy. Show that: (i) pis an epimorphism and y is a monomorphism, (i) M=Kergoimy. 44 Let L be a submodule of the module M. Show that Lis a direct summand of M if and only if there exits ‘an idempotent endomorphism i of M such that Imp = 45 Let A be a ting and let Afa(A) denote the ring ofall n x n-matrices with coefficients in A. Show that if Mis an A-module, then M" may be viewed as an Ms(A)-module. 46 Let Lbeasubmodale ofa module M, and suppose Zina submodule of M which is maximal with LAL’ = (Exercise 3.2). Show that 4 L’ has nonzero intersection with every nonzero submodule of M A Let (Me|i€1) be family of submodles of « module Af. Show that the sum S>ycy Mi i direct if and only ifor every 4€ I one has M:N, 4, My =O. (This generalizes Theorem 4.A) 4.8 Suppose A '@ 6 for some two-sided ideals a@ and 6. Show that @ is generated by a central idempotent (jc. an idempotent belonging to the center of the ring). 4.9 Let e and f be idempotents ina ring A. Show that the left modules Ae and Af ate isom if the right modules @A and 4 are isomorphic: phic if and only 5. The group of module homomorphisms Let M and W be left modules over a ring A. We define Hom,(M,.N) to be the set of module homo- morphisms M —+ N. This set has the structure of an abelian group, because we can define a + 8 for af: M+ Nas a+B: z+ a(z)+((z); one easily verifies that a+ really is a homomorphism and that Hom,(M,N) in this way becomes an abelian group. A natural question to ask is whether Hom, (M,N) also may be considered as an A-module. For this we try to define aa fora € A and a: M+ N as aa: 2++a-a(2). It is easly seen that aa is additive, but we have for b€ A that (ea)(be) = a-a(te) = ab-a(2), while 8: (aa)(2) = ba -a(2). So if A is a commutative ring, we can say that Hom4(M,N) is an A-module, and more particularly, if Ais an R-algebra over a commutative ring R, then Hom4(M,N) is an R-module. But in the general case we only have that Hom,(M, V) is an abelian group. Examples ~ 1, If ais a let ideal in the ring A, then Homa(A/a,M) & {x € M | az =0}, with the isomorphism given by a ++ a({1]). For every homomorphism from A/a is uniquely determined by its value 2 on the generator [1], and it is welldefined if and only if az = 0 for all a € a. In particular wwe have Homa(A,M) = M. 2. Let F be a free module with basis {2; | i € I}. Then every homomorphism a: F —+ M ‘uniquely determined by its values on the basis elements by the rule that @ a(S ais) = Daa(x), 10 Conversely, every assignment of values on the basis elements defines a homomorphism from F since every element of F can be written as a linear combination of basis clements in a unique way. ‘Therefore we have an isomorphism Homa(F,M) ¥ MY, given by a ++ (a(zi) | i € 1). Note that this homomorphism is not canonical since it. depends on the choice of basis for F. 3. We specialize the preceding Example by assurning that also M is free and that the modules have finite bases {z; |i = 1,...,r} and { yj | j = 1,...,m} respectively. A homomorphism a: F + M is given by (1), where each a(2;) can be written so that a determines an r x m-matrix (a,;). Conversely, every such matrix determines a homomor- phisin according to the rules (1) and (2). ‘This gives an isomorphism Hom,(F,M) ¥ Mycm(A) of abelian groups, where Myxm(A) is the additive group of r x m-matrices over A. 4. Another way of generalizing Example 2 is to consider a family (1; | i € I) of modules. A homomorphism @;Z; > M to any module M is then uniquely determined by its values on each 1; since each element of @, L; can be written in a unique way as 2 =z; with each z € Ly, and it also follows that if homomorphisms 1; + M are given for each i € I, then these determine a homomorphism (, Li —+ M. We may conchude that there is an isomorphism (3) Homa (€B Ls, M) * T] Homa(Li,M), fer ier given by a+ (ai)ier, where 4: Lj —+ jer Le is the canonical embedding homomorphism. 2) (zy For every left A-module M we may give the set End4(M) of endomorphisms of M the structure of a ring. For we already know that it is an abelian group under the addition of endomorphismns, and we use the composition of endomorphisms as a multiplication. ‘The ring axioms are then easy to verify (associativity is a general property of the composition of mappings, so it remains to show the two distributivity axioms). ‘This ring is called the endomorphism ring of the module. Example ~ 5. We again consider a free module F with a finite basis {z; | i= 1,...,r}, with the corresponding isomorphism End 4(F) + Ma(A), where M,(A) denotes the ring of n x n-matrices. We know from the preceding that this is an isomorphism of abelian groups, and the question now arises if it is a ring isomorphism. So consider endomorphisms a and 8 given by a(z:)= Soaijzy and A(z) = J) 4525 Then (25) =D byales) = by Danze = (D byaye)ee, 7 j ae ied so that composition of endomorphisms corresponds to multiplication of matrices in the reverse order. Hence we have an isomorphism of rings End a(F) = Mn(A)®. ‘When the ring A is commutative, there is an anti-automorphism of M,(A) obtained by taking the ‘transpose at of the matrix a. Combining this with the previous anti-isomorphism, we obtain a8 3 result an isomorphism End4(F) = Ma(A). ‘The correspondence between endomorphisms and matrices is then given by a(2s) however that this works only when A is commutative. i524. Note We shall now investigate the dependence of Hom4(M,.N) on the “variables” M and N; we shall Jater formulate our findings in a functorial language. We begin with the first variable, and consider a homomorphism y: M’ + M (note the order!). For each module N we then get an induced mapping g°: Hom, (M,N) — Hom,(M’,N), given by a+ ap. Then (0+ A) = (a+ Bp =ay+ Ay = o"(a) +9"(6), ‘so g* is a homomorphism. If one also has a homomorphism y!: M“ — M’, then (4) yop" =(poy)’, 1 because for any a: M —+ N one gets y’* 0 p"(a) = y*(ay) = ayoy! = (poy')*(a) We next consider the second variable, and let y: N — N’ be a homomorphism. For each module M we then get an induced mapping. Ws: Hom (M,N) — Homa(M,N’), given by a++ pa. Then Val@-+ 8) =¥(a +9) = Hat 8 = v.(a) + ¥.(9), so ¥, is a homomorphism. If one also has a homomorphism "+N", then 6) Wow = Wow). similarly to the preceding ease. We combine these two cases by considering both variables simultaneously. So let g: M' — M and Ui N + N!, We then get a homomorphism Hom, (M,.) — Hom,(M',.N’) given by a + vay. The situation is described by the commutative diagrams Hom, (M,N) Hom,(M',N) Homa (M,.N’) ——+ Homa(M’, N') Exercises 5.1. Determine the following, groups of homomorphisms («) Homz(Q, 2); (0) Homs(Z/m2,, Z/nZ). 5.2 Determine the endomorphism ring of the Z-module Q. 5.8 Show that the ebelian group Hom a(M,1V) may be viewed asa right module over the ing End (MM) and asa left module over the ring End a(N) 5A Let (Mi |4€ 1) be a family of Amodules. Show that for every A-module L: there isan isomorphism Hom(L,[] Ms) & [] Homa(L,), given by a+ (xia)ier, where x: |], Jigs Mi + M; is the canonical projection, 6. Exact sequences A sequence of A-modules and homomorphisms a or Me tM gd Miya is called exact at My if Iman =Ker an, and it is exact if it is exact at every My where there is both ‘an in-coming and an out-going arrow (the sequence may be of finite or infinite length). ‘The sequence is more generally called a complex if Imay—1 C Ker an holds for all n, ie. if an 0.aq-1 = 0 for all n. ‘This later term emanates from algebraic topology, where one in the study of the homology groups of a space X considers complexes where My is the free abelian group on some kind of n-dimensional subsets of X (e.g. simplices), and an: My —+ M1 takes the boundaries of these sets; it is rather natural that a boundary itself has an empty boundary, 60 that one gets ay © an41 = 0 and hence a complex, and one can then define the homology groups of the space as H»(X) = Ker an/Imany1- Examples ~ 1. A sequence 0+ L-% M is exact ¢=3 Kera=0 = aris a monomorphisin 2M. N—Oisexact <> Imf=N <> fis an epimorphism. 3. Suppose L is a submodule of M. We then get an exact sequence 0—» L % M 2 M/L + 0, where a is the inclusion and @ is the canonical projection. Suppose conversely to Example 3 that we have an exact sequence @ 0-bSMAaNO0. 2 ‘Then a is a monomorphism and f is an epimorphism, and we have by ‘Theorem 2.A that N & M/Ker 9 = M/lma. The sequence is therefore of essentially the same type as in Example 3, which we may more precisely express by saying that the diagram 0 ——+ Ima —— M ——- M/Ima ——+0 Be oe is commutative. An exact sequence of the form (2) is usually called a short exact sequence. In such ‘a sequence we may thus without real Joss of generality assume that L is a submodule of M and that N=M/L. ‘The short exact sequences are of particular importance because every long exact sequence can be analyzed in terms of them. Consider namely an exact sequence (1). Setting In = Imay = Ker aq41, we get @ commutative diagram, In Ings Mg Me, oe. Mus where also the oblique sequences are exact. ‘The following is a useful technical result, called the “short five-lemma”, whose proof is an example of “diagram-chasing”. Lemma 6.a Suppose the diagram 0——1—*—M ae ut es is commutative, and that both rows are exact sequences. (i) If and v are monomorphisms, then jis a monomorphism. (ii) If) and v are epimorphisms, then 4 is an epimorphism. Proof. (i): Suppose 2 € M and p(z) = 0. Then vA(z) = f'y(2) = 0, and since v is a monomorphism, it follows that A(z) = 0. ‘The exactness of the upper row implies that z = a(y) for some y € L. ‘Then @!A(y) = way) = p(z) = 0, and since a and \ are monomorphisms, it follows that y = 0, and hence a(y) = 0. (ii): The proof is done by a “dual” kind of diagram chase, and is left to the reader as an exercise. ‘We shall in more detail consider the short exact sequences associated with a decomposition of a module into direct summands. If M = My @ Mp, then we get a diagram n 4 05 MSM. @ Ma M0, To Mi @ Ma 5 Ma where «1 and ¢2 are the inclusions, and ; and 72 are the projections. We obtain in this way two short exact sequences, one going forwards and the other backwards. Exact sequences of this kind are called split. More generally we say that a short exact sequence 0+ I, + M —+ N —» 0 is oplit if there is an 1B e * isomorphism M —+ L@N such that the diagram 0——1 M N—0 9 a t——-5-—_—_ non N——+0 is commutative, where the lower sequence is a split sequence of the kind we considered before. Note that it follows from Lemma 6.a that any M —+ L@ N making the diagram commutative is an isomorphism, Lemma 6.b The following assertions are equivalent for an exact sequence 0+ L *. M &. N +0: (a) The sequence is spit. (b) There exists x: M+ L such that xa (©) There exists x: N+ M such that fc Proof. (a) = (b) and (a) => (c) are clear from the preceding, (b)= (a): We define pw: M — L ® N as y(z) = (xz, Az). One easily verifies that the resulting diagram (3) is commutative (©) (a): In this case we define 9: L@ N+ M as n(z,y) = a(2)-+1(y). Lemma 6.c Every exact sequence 0+ L % M *. F +0, where F is a free module, is split. Proof. Let { 2; |i € 1} be a basis for F, and choose for each 2; an element ys in M such that f(y) = 25 (this requires the axiom of choice!). Now define ¢: F + M by taking ¢(z,) = y (Example 5.2). Then Bt=Idp, and we can use Lemma 6.b. Exercises 6.1 Construct two exact sequences of Z-modules OBB e0 and LBD 62 (The five-lemma) Co + a commutative diagram k—+L—-m—_- n+p fe lieth el ‘K' — 1! —~ M' +. N' +P’ with exact roms. Show that: (@) Ife is an epimorphism, and \ and v are monomorphism, then j is s monomorphism, i) If is a monomorphism, and A and v are epimorphisms, then jis an epimorphism, (iii) Ix, A, v and x are isomorphisms, then jis an isomorphism. 6.8 Consider a commutative diagram to se: Seeley 01! aM’ W's | J t 0—+b + —+N—0 Lie 1 OL" +m" —4N" 0 1 J t 0: 0 ema where the rows and the middle column are assumed to be exact. Show that the last column is exact if and only if the first column is exact. 6.4 Suppose that all rows and columns are exact in the diagram of the preceding Exercise. Show that from the diagram one can obtain an exact sequnce L@ M!—+ M—» N" 0, 6.5 Consider a commutative diagram x L—*M with exact rows. Show that if « is an epimorphism and v is a monomorphism, then Kerp=a(Kerd), Im =6"*(Imy). 7. Functors and preservation of exactness Let A and B be rings. A (covariant) functor F from left A-modules to left. B-modules is obtained as follows: — to each left A-module M it associates a left B-module F(M); — to cach homomorphism a: M — N it associates a homomorphism F(a): F(M) —» F(N) of left B-modules. It is required that F should satsify: 1) Fld) = Tdycar), where Id denotes the identity endomorphism of M; 2) Ifa: M— N and B: N —+ P are homomorphisms of left A-modules, then F(a) = F(8)F(a). Examples ~ 1. A functor from left A-modules to abelian groups (i.e. Z-modules) is obtained by taking a module M to itself considered just as an abelian group, and taking a module homomorphism to itself considered just as a group homomorphism. ‘This rather trivial kind of functor is called a “forgetful” functor since it forgets part of the module structure 2. Let A be a commutative ring. Taking the torsion submodule T(M) of a module M (Sec. 2) may be viewed as a functor from A-modules to A-modules; it is clear that properties 1) and 2) hold in this case. 3. A more important example is furnished by the Hom-functor. For a fixed left A-module L we consider Homa(L,-) as a functor F from left A-modules to abelian groups, defined as F(M)=Wom,(L,M), F(a) =a,: Homa(L,M) —+ Hom,(L,N) for a: MN. Property 1) is obvious, and 2) is just (5) of Sec. 6 We also have to consider functors which reverse the homomorphisms. A contravariant functor from left A-modules to left B-modules is obtained as follows: — to each left A-module M it associates a left B-module F(M); — to each homomorphism a: M — N it associates a homomorphism F(a): F(N) + F(M) of left B-modules, It is required that F should satisfy: V) F(ldae) = Tapas 2) Ifa: M+ N and 8: N—+ P are homomorphisms of left A-modules, then (82) = F(a) F(8) Example — 4. The most important example is again the Hom-functor. Por a fixed left A-module -M we considet Homa(:,M) as a functor F from left. A-modules to abelian groups, defined as Homa(E,M), F(a) =a": Homa(L,M) — Hom,(K,M) for a: K — L. now just (4) of See. 5. A functor F is called exact if it preserves exactness, i.e. carries each exact sequence into an exact sequence. It then in particular carries short exact sequences into short exact sequences. But also the converse holds: Lemma 7. Ifa functor F preserves exactness of short exact sequences, then it is exact Proof. We leave it as an exercise to show that F(0) = 0 (consider a short exact sequence with only zero modules). We next note that F preserves monomorphisms and epimorphisms. For in the case of a monomorphism a: L— M we have an exact sequence 0+ L—+ M — M/lma ~~ 0, whose exactness is preserved by P, 90 also F(a) is a monomorphism. ‘The argument for epimorphism goes dually. Consider now an exact sequence Ls M2, N, which is carried into a sequence P(E) p(ar) 2 P(N). We aa according to fees exact sequence nema thurs coacs sequences, What we need here are the factorizations « = :A and = mm described by tAima 4M % Imp +4, where A and 9 are epimorphisms, and 1 and v are monomorphisms. We have that In F(a) = Im F()F(A) = Im F(u) 16 since F(A) is an epimorphism. Similarly we get that Ker F(B) = Ker F(v)F(n) = Ker F(n) @ monomorphism. We have a short exact sequence 0+Ima4MSimp0 Ima, and since F preserves its exactness, we get that Im F(j1) = Ker F(7). It follows that er F(B). Exactness of the Hom-functors We shall study the exactness properties of the two Hom-functors, and we begin with the covariant functor Homa(L,-). Consider an exact sequence 0+ MANS N/M, where M is a submodule of N. We get an induced sequence a 0+ Homa (Z,M) “+ Homa (L, N) “+ Homa(L, N/M) +0. Here 1, is a monomorphism, because if 44(a) = 0 for some a: L + M, then a(x) = 0 for all 2 € L, and since jis a monomorphism, it follows that a(z) = Next we note that Im, = Kerr, because Kerv, = { since F(v) since Ker 8 Im F(a) y= {a:b N Imac M}=Imp,. L aN NM ° ‘The homomorphism v, in (1) is however not always au epimorphism. For if it were an epimorphism, then for every a: L + N/M there would exist : L + N such that vf = a, and this is not the case eg. if we let v be the canonical homomorphism Z —+ Z/2% of Z-modules and let a be the identity ‘endomorphism of 2/22. o—+ 2m, a" N— N/M Z—-4. ‘The sequence (1) is thus not in general exact, and we can only say that 0 Homa(L,M) *% Hom, (L,.V) **+ Homa(L,M/N) is exact. We say in general that a covariant functor F is left exact if it carries an exact sequence 0 —> L —> M — N ~ 0 into an exact sequence 0—+ F(L) > F(M) —» F(N). Similarly we say that F is right exact if the sequence F(L) -» F(M) —+ F(V) + 0 is exact. If the functor is both left and right exact, then it preserves exactness of short exact sequences and is therefore exact by Lemma 7.2. ‘We now study the contravariant functor Hom q(-, M) for a fixed module M. Consider an exact sequence 05K SLAL/K—O0. In this case we get an induced sequence @ 0 Homa(L/K,M) *> Homa(L,M) “+ Hom4(K,M) +0. Here A is a monomorphism, because if A*(a) = 0 for some a: L/K + M, then aA(z) = 0 for all x € L, and since is an epimorphism, it follows that a= 0. Further we have that Im A* = Ker", because (cf. Exercise 2.6) a: L-+M |an=0)}= (a: b+ M | «= Ba for some f: L/K + M} = 1m" Ker x* = 0 L/kK ——+0 ‘The homomorphism x* is however not always an epimorphism. For ifit were an epimorphism, then for every a: K —+ M there would exist an extension of « to a homomorphism 8: L —» M such that Ax = a, and this is not the case if we e.g. take x to be the inclusion homomorphism nZ + Z of Z-modules, and let a7: nZ— Z send nm € nd tom Z. ‘The sequence (2) is thus not in general exact, and we can only say that 0 — Hom4(L/K,M) 2+ Homa (L,M) + Homa(K,M) is exact. We therefore say that also the contravariant functor Hom 4(-, M) is left exact. ‘We have remarked in Sec. 6 that lack of exactness in complexes gives rise to homology groups. In hhomological algebra one similarly defines derived functors Ext(,-) of the Hom-functor, which measure the degree of inexactness of Homa(;,-). In particular one defines a let A-module P to be projective if the functor Hom,(P,-) is exact, i.e. if for every epimorphism p: M —+ N and every a: P + N there exists a lifting of a to 8: P—> M such that 43 =a. Similarly one defines M to be an injective module if for every monomorphism x: K — L and every a: K —+ M there exists an extension of a to 8: L—+ M such that x = a. We shall however not here be so much concerned with projective or injective modules (but see Exercises 2-5 below) Exercises 7.1 Show that both the covariant and the contravariant Hom-functors preserve split exactness of short exact sequences. 7.2 Show that every free module is projective. 1.3 Show that P = @je, P. is « projective module if and only if every P; it a projective module (cf. (3) of Sec. 5). What is the corresponding result for injective modules? 7.4 Show that the ring Z/6Z. can be written as the direct sum of two nonzero ideals, and that these are projective modules over 2/62 which are not free, 7.8 Show that the following assertions ate equivalent for a module P: (2) P is a projective module. (©) Every exact sequence 0+ + M— P— 0 splits, (¢) Pisa direct summand of a free module 16 Let gs A— B be ring homomorphism. Show that there isa functor from left B-modules to left A-modules obtained by considering each B-module as an A-module by means of y, and that this functor is exact. 117 Determine the exactness properties of the torsion submodule functor T. 7.8 Let A be an integral domain. An A-module D is divisible if D = aD for every nonzero « € A, ie. if for every 2 € D there exists y € D such that 2 = ay. (Typical examples: Q and Q/Z, as Z-modules). Show that: (i) IK is the field of quotients of A, then every vector space over K becomes torsion-free and divisible when itis viewed as a module over A. (ii) IED is a divisible module and a: D — M is an epimorphism, then M is divisible. Gii) Every module M contains a largest divisible submodule D(J). (iv) M ++ D(M) may be considered as a functor from A-modules to A-modules; determine its exactness properties (») Brery injective Z-module is a divisible module. (Hint: f. the example that was used to show that x* in 2) is not an epimorphism). (It can be shown thatthe injective Z-modules are precisely the divisible ones). 8. Bimodules Let A and B be rings. An A— B-bimodule is a left A-module M, which also is a right B-module, with the same additive structure in both cases, and satisfying a(2b) =(az)b forac A, 2€M,bEB. ‘We write in this ease 4Mp. In some cases it may be preferable to consider an A—B-bimodule with both A and B acting to the left, and the requirement then is that a(bz) = (az) in the previous notation, 7 Examples ~ 1. Every two-sided ideal of a ring A may be viewed as an A— A-bimodule. 2. Every left A-module may be viewed as an A—End,(M)-bimodule, with both A and End4(M) acting to the left. 3. If M and N are left. A-modules, then Hom, (M,N) may be viewed as an End.,(N)~ End 4(M)- bimodule, with the endomorphism acting by composition with homomorphisms M — N. We shall in particular consider how a bimodule structure gives a module structure on the Hom-group. Suppose we have bimodules «Mp and 4Nc, and consider the abelian group Hom.(M,WN). We may define an action of B to the left on Hom (M,N) by bp: 24+ 926) for be B, yp: M—N and 2 eM. Wis clear that bp: M —+ N is additive, and for a € A one has (bp)(az) = {(az)b) = p(a(2b)) = a( (2b) = a((bp)(z)), so by is a member of Hom,(M,N). This makes Hom4(M, WV) into a left B-module, because (bb!) = 4(0"p) as is easily verified We can also define an action of the ring to the right on Homa(M,N) by verze la)e force C,p: M—+N and ze M. Its again clear that ye: M —+ N is additive, and for a € A one has (ve)(az) = oaz)e= ap(2)e = alye)(z), s0 Homa(M, N) becomes a right C-module. Furthermore, for 6 € B and ¢ € C one has (by)e = Wipe) because ((be)eNz) = be)(z)e = e(zd)e = (ve)(zb) = (b(¢e))(2). Hence Hom, (M,N) is a B ~ C-bimodule Similarly one shows that in the situation »Ma, cNa one has the structure of — B-bimodule on Homa(M,N), with (ee)(z) = ep(z) and (p8)(z) = (bz). Note that in both cases the side of action of the ring remains the same for the covariant second variable, while it is changes for the contravariant first variable. Dual module Let M be a left A-module, By regarding A as an A— A-bimodule, we can by the preceding regard Homa(M, A) as a right A-module, with ga: x ++ y(z)a. This module is called the dual module of M, and we denote it by M*. Its elements are called “linear functionals” on M, and one often writes (z) as (2,9). By forming the dual (M*)* we get a left A-module, which is called the bidual of M, and is simply denoted by M**. There is a canonical homomorphism p: M — M** given by (n(2))(p) = 92), or in the alternative notation ay (eH(z)) = (2): ‘One easily verifies that this is a homomorphism of left A-modules. Bvery homomorphism a: M —+ NV of left A-modules induces by Sec. 5 a homomorphism a* : N* + M*, which is easily verified to be a homomorphism of right A-modules (cf. Exercise 4), and is called the transpose of a. It is given by (2,0"(y)) = (2,9) forz€M and EN’. We shall in particular consider the dual and bidual of a finitely generated free module M, with a basis z1,...2_. Every homomorphism y: M —+ A is then uniquely determined by its values on the basis elements since (2, a:2) = 5 axp(zi). We assert that a basis for M* is given by the functionals Fhy---y2) defined as 1 ifi=j ‘These are linearly independent, for if SD; aj2} = 0, then 0 = (SD, aj2})(z;) = a; for each i. ‘They generate M*, for if € M* with y(z;) =b, for each i, then for every 2 = Jo, aiz € M we get, (2) = D> asels) = Sabi = YD asbs25(2s) = D> by25(2), 7 7 i 7 Dy bj2}- Hence form a basis, which is said to be “dual” to the basis 2, A is commutative, this gives an isomorphism between M and M7, but this isomorphi since its definition depends on the choice of basis for M. However we have: 18 ‘Theorem 8.A The canonical homomorphism p: M —+ M** is an isomorphism when M is a finitely generated free module. Proof. Let 21,...,2n be a basis for M, with the dual basis 23,...,25. The definition (1) of jt shows that 21)... Haq is just the dual basis 2}°,...,25° of 2],...,2%. Since thus j1 carries a basis for M to a basis for M*, itis an isomorphism. In general one says that a module M is reflexive ify: M — is an isomorphism. Generalized triangular matrix rings Let A and B be rings, and let My be a bimodule. ‘The ring we shall define is denoted by AM o B) and its elements are matrices ($) with a € A, z € M and 6 € B. Addition and multiplication is defined to be the usual one for matrices, which means that multiplication goes like @ mG :) x G “12) os) \o v)=\o w One easily verifies that this really gives a ring. In more conventional notation one could have defined the ring to be the abelian group A@M@B, with multiplication given by (a,z,8)-(a', ys?) = (aa', ay +2’, bb’); this gives a more compact, but less conspicuous, notation. These generalized triangular matrix rings are interesting because they can be employed to give examples of relatively complicated noncommutative rings in which computations still can be made without too much difficulty. Example. We shall in more detail consider the ring KOK ox): where K is any field, and we shall determine its ideal structure (ie. its left, right or two-sided ideals) Suppose J is a proper left ideal in the ring. If F contains a matrix @ Ger) with a # 0 and 6 = 0, then it clearly contains also all matrices (2) with arbitrary a € K and b= 0 (cssentially because X has no proper ideals except 0). Thus one sees that ( ©) is a left ideal. If ‘one instead assumes that there is a matrix (2) in J with a = 0 and 6 # 0, then by multiplication from the left in the ring one a also matrices in J with nonzero entry in the northeastern corner, and one finds that T= (°). By this kind of argument one sees that a proper left ideal is either of the form (§ £), or it consists ofall matrices ({ §), where (a,c) belongs to a certain subspace of Kok. We shall alo determine the left modules over the ring (% %). Let V and W be vector spaces over K, with a given homomorphism gp: W + V. ‘Then V @W becomes a let module ovr (*) iff we define, using again matrix notation, ON eo ob) \y wy left to the reader to verify the module axioms. We shall now aoe that every module X over ( I can be obtained in this way. We define (cf. Exercise 2.7) Ko _(0 0 v=(E x, wa(28)x, which are subspaces of the vector space X over K. We have X = V @ W since each 2 € X can be written ( ‘)z=(1 r)e+(2 {)eavtw @) 19 It and clearly VA W = 0. We further note that for w € W we have ($5) w=(52)-(gi)wev, and we define Woe an wee ($5) w. For each z € X we then have by (3) that G ) = av + cp(w) + bw, as asserted. Exercises 8.1 Let A be a right B-module. Show that to give M the structure of an A— B-bimodule is equivalent to defining a ring homomorphism A — End a(M). 82 Let M be aleft A-module. Show that the isomorphism M —+ Hom 4(A,M) in Example 5.1 is an isomorphism of left A-modules (where the bimodule structure of A has been used). 8.3 Let ¢ be an idempotent in a ring A and let M be a left A-module. (i) Show that ede with e as an identity clement. (ii) Show that eM = {ex |z € M) is a left module over the ring eAe. (iii) Define g: ef — Homa(Ae, M) as y(ex): ae ++ acz. Show that this is a bijective mapping, and that it is an isomorphism of left eAe-modules (where Ae has been considered as an A~ eAe-bimodule) 8.4 Consider a homomorphism : M’ — Mf of left A-modules. Show for any bimodule aly that the induced mapping 9": Homa(M, N)—+ Hom,(M’, N) is a homomorphism of right B-modules. 8.5 Determine the ideal structure of the ri KV ox) where K is a field and V is a vector space over K. 9. Noetherian rings and modules ‘Most of the rings one uses in applications satisfy some kind of finiteness condition. ‘The most important of such conditions is that every (left) ideal of the ring is finitely generated. We begin our study of this condition by considering it in the more general situation of modules. Noetherian modules ‘Theorem 9.A The following conditions are equivalent for a left A-module M: (2) M and all its submodules are finitely generated. (b) Bvery strictly ascending chain My C Mz C ... of submodules of M is finite. (©) Every nonempty family of submodules of M has a maximal member. Proof. (a) =>(b): If a strictly ascending chain My C Mz C ... of submodules is given, let M’ be the union of its members. Then M’ is a submodule of M, so it has finite set of generators. Since they are finitely many, these generators all lie in a certain member M,, of the chain, and then the chain stops at My. (b)= (c): Let M be a family of submodules of M. Choose some Mi € M. If Mj is not maximal, there exists Mz € M such that M; C Mz. Continuing in this way, we get a strictly ascending chain, which must stop after finitely many steps, giving a maximal member of M. (Note that the axiom of choice is used in an essential way in this argument). (©)= (a): Consider any submodule L of M. Let M be the family of all finitely generated submodules of L. By (¢) there exists a maximal member M’ of M. If M’ # L, then let x € L\ M’. But then M’+ Ar is finitely generated submodule of L which strictly contains M, a contradiction. A module M satisfying these conditions is said to be noetherian (after Emmy Noether, who was leading, the development of the general theory of rings and modules in the 1920's). Examples ~ 1. A vector space is noetherian if and only if itis finite-dimensional. 2. Every finitely generated abelian group is noetherian, as we shall see later on. tis easy to see that every submodule and every quotient module of a noetherian module is noetherian. Conversely we have: 20 Lemma 9.0 If 0 +L M + N 0 is an exact sequence with L and N noctherian, then M is noetherian, Proof. We may as well assume that L is a submodule of M and that NV = M/L. We must show that every submodule M’ of M is finitely generated. Now M'(L is finitely generated since it is a submodule of 5 let 21,...2q be generators. We have an isomorphism M'/(M! OL) = (M! + L)/L by Theorem 2.C, and since (M' + L)/L is a submodule of the noetherian module M/L, it is finitely generated. So M'/(M'1 L) has finitely many generators [yi],..-[¥r] with all fy M’. We assert that 21,.--)2nstny-+ste together generate M’. Suppose z € M’. For [2] © M'/(M'N L) we have [2] = ails] +--+ + ar[ye], 80 2 — arm —---— ary € M'NL, and hence tam Grp = bizy +--+ + btn. is a linear combination of the elements 21,...,2m,Yiy.+- Ue: Ml In particular it follows from this result that if L and M are noetherian modules, then also L @ M is noetherian, From this it is seen that Noetherian rings A ring A is left noetherian if it is noetherian as a left A-module, ic. if all left ideals are finitely generated; equivalent conditions are by Theorem 9.A that every strictly ascending chain of left ideals is finite, or that every nonempty family of left ideals has a maximal member. If A is left noetherian, then every finitely generated free left. A-module is noetherian by the conclusion after Lemma 9.a, and by (the proof of) Theorem 4.C we obtain: ‘Theorem 9.B If A is a left noetherian ring, the every finitely generated left A-module is noetherian. Examples ~ 3. Every field is a noetherian ring. 4. The tings Z and K(f), for a field K, are noetherian since all their ideals are principal. Every finitely generated abelian group is therefore noetherian by ‘Theorem 9.B. 5. If A is a left noetherian ring, then A/a is left noetherian for every two-sided ideal a of A. For every left ideal of A/a is of the form 6/a, where 6 is a left ideal of A containing 4, and is therefore finitely generated. 6. The ring C{0, 1] of continuous real-valued functions on the interval (0, 1] is not noetherian, since eg. the ideal of functions which are zero at 1/2 is not finitely generated. ‘One of the most fundamental theorems in algebra is the following, known as Hilbert’s basis theorem (Hlitbert 1890) Theorem 9.C If A is a commutative noetherian ring, then Aft] is a noetherian ring. Proof. Let. a be any ideal of Alf]. We must show that itis finitely generated. For every n we put 4 = {a€ A | there exists a polynomial in a of degree n with leading coefficient a, or a = 0} It is easy to see that each gy is an ideal of A. Further we have oy C dys, because if a € ay with a polynomial f(t) € a with leading coefficient a, then tf(t) is a polynomial in eof degree n-+1 with leading coefficient a. So we have an ascending chain of ideals in A, and since A is noetherian, this chain becomes stationary after some no, ic. dy = Gn for all n > ng. For each i < no we choose finitely many generators 4; of the ideal a. ‘To these elements correspond polynomials fu; of degree i with leading coefficient aj;. We shall prove that these fij, which are finitely many, generate the ideal a. Suppose to the contrary that there exists a polynomial g € a which cannot be written as a linear combination of the fi's (with arbitrary polynomials as coefficients). Choose such a g of lowest possible degree. We write g(t) = bt" + lower degree terms (#0), where 9 € a implies that 6 € Gy. First assume m < ng. Since Gm has generators ams, we can write b=, bjamy with b € A. Now consider the polynomial A(t) = 9(t) — 9 bs Smal), which is a member of @ since g and fmj are so. ‘The leading coefficient of h is b— >, bjans = 0, $0 h has degree < m. But by the choice of g we can therefore write h as a linear combination of fij’s, and so we can then also do for g(t) = h(t) + D>, bj fnj(t), @ contradiction. a It remains only to consider the case when m > no. Then b = S>, bjanqj and we choose Nt) = a(t) — > bs Ines tt. By an argument similar to that in the first case we see that h has degree < m, and we may again conclude that g is a linear combination of fi;’s. By repeated use of the ‘Theorem, we obtain: Corollary If A is a commutative noetherian ring, then Alt,,...,tn) is a noetherian ring, Example ~ A commutative A-algebra B, over a commutative ring A, is a said to be finitely generated as an algebra if there exist finitely many elements 21,...,2 € B such that every element of B can be written as a polynomial expression in the 2j’s with coefficients in A, which means just that B & Ally... tn)/a for some ideal a. If A is a noetherian ring, then every finitely generated commutative A-algebra is noetherian by the Corollary and Example 5. Exercises 9.1 Let L be submodule of a module M. Show that if L and M/L are finitely generated, then M is fi generated, 9.2 Show that if G is a finite group and K is a field, then K{G] is a left and right noetherian ring. 9.3 Let A be a commutative noetherian ring. Show that if Zand M are finitely generated A-modules, then also Hom.(Z,M) is finitely generated as an A-module, 94 Let K be field and / an infinite set. Show that A‘ is an ideal in the ring K, and that this ideal is not finitely generated (so the ring K" is not noetherian). 9.5 Show that the generalized matrix ring 2Q 0 Q QR oR, is right noetherian, but not left noetherian. 97 Let M be a noetherian left A-module. Show that if an endomorphism : M — M is surjective, then it is an isomorphism, is right noetherian, but not left noetherian, 9.6 Show that the generalized mattix ring 10. Semisimple modules We shall consider class of rings, called “semisimple” rings, which generalize division rings, and a corresponding class of modules which may be viewed as generalizations of vector spaces. We begin by studying modules in this section. Let A be an arbitrary ring. A left A-module M # 0 is called simple (or irreducible) if it has no submodules except 0 and M. E.g, if mis a maximal left ideal of A, then A/m is clearly a simple module ‘over A. Suppose conversely that M is simple, and choose any 2 #0 in M. Then Az = M because of the simplicity, s0 every nonzero element in M is a generator of M. We have that Az ~ A/Ann(z) by Sec. 2, and Ann(z) must therefore be a maximal left ideal. (Note that different choices of x in M may give different Ann(z)). We thus have that a left A-module is simple if and only if it is of the form A/m for a maximal left ideal of A. We note an interesting property of simple module (known as Schur’s lemma): Lemma 10.a If M is a simple module, then End,(M) is a division ring. Proof. If y is a nonzero endomorphism of M, then Imp is a nonzero submodule of M, so it is equal to M. Similarly Kery # M, so Ker = 0. Hence has an inverse. A module is called semisimple if itis a direct sum of simple modules. (Also the module 0 is considered as semisimple) Examples ~ 1, Let D be a division ring. ‘The simple D-modules are the one-dimensional vector spaces over D, while all vector spaces over Date semisimple modules. 2. The simple Z-modules are the groups Z/p2 where p is a prime. 2 3. If K is a field, then K may be viewed as a simple module over the polynomial ring K(f], with tz = 0 for every z in the module K. 4. A left ideal in a ring A is a simple A-module if and only if itis a minimal left ideal (ie. minimal nonzero left ideal, ef. the definition of maximal left ideals). Many rings, as e.g. Z, do not have minimal ideals (see further Exercise 10.1), 5. If L is a submodule of a module M, then the quotient module M/Z is simple if and only if Z is ‘a maximal proper submodule of M. ‘The following equivalent characterizations of semisimple modules generalize properties of vector spaces: ‘Theorem 10.4 The following properties are equivalent for a module M- () M is semisimple. (b) M is a sum of simple submodules. (©) Every submodule of M is a direct summand. Proof. (¢)=» (b): Let S be the sum of all simple submodules of M. ‘Then M = $@ L for some submodule Lof M. We must show that £ = 0. Suppose L #0. It now suffices to show that L contains some simple submodule, which would contradict the fact that LMS = 0. Choose any cyclic submodule C # 0 of L. ‘Then G has a maximal proper submodule NV by Exercise 3.1. By (c) we have M = N @ P for some P. ‘Then C = N @(PNC) since NN(PNC) = (NNP)NC =0 and for each y € C we have y= u-+v with we N,v € P, which gives y—u=v CMP and hence y € N+(PMC). But since N is a maximal proper submodule of C, it is obvious that its complement PC must be simple (a) => (c) and (b) =(a) follow from the following more precise result. ‘Theorem 10.B Suppose M is the sum of a family (5; | i € I) of simple submodules S;, and that L is any submodule of M. Then there exists a subset J of I such that M = L ® (@jey Si). Proof. Since this theorem generalizes Theorems 3.B and 4.B, its proof is a refinement of the proofs of these previous results. Let M be the set of all subsets J of I for which the sum Sy = S7je, 5; is direct and LMS; =0. The set M is nonempty since @ belongs to it, and it is ordered under inclusion. Suppose Cis a chain in M. Let J’ be the union of all J €C. Then J” will be an upper bound of C if it belongs to M, To see that J’ € M, we mu:* show that L/S, = 0 and that the sum Sj¢,,5; is direct, and this is easily done because any element of Sr lies in some Sy with J € C. So this shows that J’ is an upper bound, and Zorn’s lemnia gives a inaximal J © M. Since we have shown that the sumis direct, it remains to see that + Dy¢y Sj = M for this maximal J, and for this it suffices to show that S; C L+ Sy for every i € I. Suppose this does not hold for i. Then 5; 9(L + Sz) = 0 because of the simplicity of Sj. If we adjoin i to J, then we still get a direct sum S;+(Syey Sj. We also have (S;-+S,)ML =0, because if z= y-+z with 2 € L, y € S; and z € Sy then y= 2-2 € 5; (L+ Sj) =0, 90 2 = y ELMS; =0. Altogether this contradicts the maximality of J If L is a submodule of a semisimple module M, then both L and M/L ate semisimple. For suppose M = @jer Si with simple 5,. Then by the preceding theorem we have that M = L ® (®je, Si) for some J CI. It follows directly that M/L ~ @j¢, Si is semisimple. ‘The theorem also gives that M = (@ies Si) ® (Bjex Si) for some K CI, and then L * M/@, Si ¥ @jex Sir 80 also L is semisimple. We let & denote the set of isomorphism classes of simple left A-modules, and for each semisimple mod- ule M and each w € M we let My denote the sum of all submodules of M belonging to the isomorphism class «. We say that M, is a homogeneous ot isotypical component of M. Lemma 10.b Every semisimple module is a direct sum of its isotypical components. Proof. 1t is obvious that a semisimple module M is the sum of its isotypical components. Suppose zy tes 2, =0 with 2; € My, for different w;. If 21 #0, then Avy C Mu, (Mu, +--+ Mu). But then a simple module in w, is by the preceding isomorphic to a submodule of a direct sum of simple modules in « for i= 2,...,n, which is impossible. If: M — M' is a homomorphism between semisimple modules, then y(M.) C Mj, since y(M,) is = quotient module of Mf, and hence is w-isotypical. For later use we note s converse result: 23 Lemma 10.¢ Let M be a semisimple module. If is a submodule of M such that (L) C L for every endomorphism p of M, then L is a direct sum of isotypical components of M Proof. We must show that if $ is a simple submodule of L and S&S" for some other simple submodule ‘S' of M, then there exists an endomorphism y of M with (S$) = S*. But this is easily done by taking to be 0 on a complement of Sin M. Tet us now in particular consider a homomorphism y: S$ @ S —+ $@ S for a simple module S. For (zu) €S@S we have vu) = (ent + Piven + ery) for homomorphisms yy : S—+ S. Thus y may be represented by a matrix ( pn 3) en en)” and since End4($) © D for some division ring D, we find that End 4($ @ S) © Mz(D). In the general case we have for a finitely generated semisimple module M a decomposition M = My, @-+-@ My, into isotypical components, and if S; is a simple module in w with End, (Sj) = Ds, and M,, is isomorphic to a direct sum of nj copies of S;, then End 4(M) © My, (D3) x -++ x Mp, (Dr) ‘Thus the endomorphism ring of a semisimple module is the direct product of matrix rings over division rings. Exercises 10.1 Determine the minimal left ideals in the rings 2x Q and (* ®) for 1022 Let @ be a minimal lef ideal in A (i) Show that if 4485 isa simple left A-module, then cither @and S are isomorphic as modules, or aS = 0, () Show that cither a = 0, or a= Ae for some idempotent ¢ € A 1038 Suppose my and my are two maximal left ideals in A. Show that the simple modules A/m and A/m, are Ssomorphic if and only if there exits « € AS my such that ma C ma. 104 Por what n is Z/nZ, a semisimple Z-module? 10.5 Show that if M ina semisimple module and + € M, then Ana(z) is an intersection of finitely many maximal left ideals 106 For each A-module M we define its socle S(M) to be the sum of all simple submodules of M. Show that M r+ S(M) gives functor from left A-modules to left A-modules. What are the exactness properties of this functor? 107 Determine the socle of the Z-module [],¢2Z/pZ, where P in the set of all primes. field K. 11. Semisimple rings A ring A is called semisimple if it is semisimple as a left A-module (we shall in a moment see that this definition actually is left-right symmetric). A semisimple ring A can be split into the direct sum of its isotypical components, as A= @ eq Aw, Where each A, is a direct sum of minimal left ideals which are morphic to each other. Now A, is not only a left ideal, but it is actually a two-sided ideal. To see this, we define for any a € A a mapping pa: A—+ A as pa(z) = 2a. Since p, clearly is a homomorphism of left A-modules, it follows from the discussion at the end of Sec. 10 that pa(A.) C Av, which means that A, is a right ideal. ‘Thus each A, is a two-sided ideal. We write the identity element of A as 1 = a, -+-+-+ aq, where a; € A,, for different isotypical components w,...,i%. For any a € A we then get Gy too +a dy E Ay, OB Aug. ‘This means that @) A= Ay BO Aug, and it also follows that © is finite, with © = {u1,...,0m). For if $ is a simple left A-module, then S = A/m for a maximal left idesl of A, and the semisimplicity of A implies that A= m@ a for some left ideal a; but a must then be simple as a left A-module, so 5’ A/m™ a, and S belongs to the same isomorphism class w as the left ideal a. ‘Thus we have: 4 Lemma 11.a If A is a semisimple ring, then there are only finitely many isomorphism classes of simple left A-modules. Each isotypical component of A is a two-sided ideal of A, and every two-sided ideal of A is a direct sum of some of these isotypical components. Proof. It only remains to show the last assertion. Let a be two-sided ideal of A. We shall use Lemma 10.¢, and therefore consider any endomorphism y of A as a left A-module. But y is then of the form 21+ 2a for some a € A, and since ais a right ideal, we have y(a) C a Thus ais a direct sum of isotypical components of A. i A ring A is said to be a simple ring if it is semi-simple and has only one isomorphism class of simple modules, of equivalently, if it has no proper two-sided ideals # 0. (Thus a simple ring A is not in general simple as an A-module). From the preceding Lemma follows that each semisimple ring A is a direct product of simple rings, which are the minimal two-sided ideals of A. In order to determine the semisimple rings we thus only have to consider the simple rings. Examples. Every commutative simple ring is a field since it has no proper nonzero ideals. Eve division ring is a simple ring. Theorem 11.A A ring A is sing D. mple if and only if it is isomorphic to a matrix ring My(D) over a division Proof. A matrix ring M,(D) is a direct sum of n left ideals aj, where a; consists of the matrices with all entries = 0 except those in the i-th column. Each ; is a minimal left ideal, because it is generated by any one of its nonzero elements. In fact, a nonzero element of a; can be described as a column vector ® £0 of elements of D, and for any other such vector w there is a linear mapping taking # to 1 this linear mapping corresponds to the multiplication to the left with an element of M,(D). ‘Thus Mz(D) is semisimple as a left module over itself. Since the minimal left ideals a; clearly ate isotypical, it follows that M,,(D) is a simple ring. Suppose conversely that A is any simple ring. ‘Then A is isomorphic as a left A-module to a direct sum of n copies of a simple module S. If we let End4(S) = D, which is a division ring, then by the preceding Section we have that End 4(4A) © M,(D). Now End (4A) © A°?, but on the other hand we also get Mn(D)° © My(D*?) by sending a matrix to its transpose (see Example 5.5). As a result we have an isomorphism A My(D°?). A simple ring A is thus isomorphic to a matrix ring over the opposite of the endomorphism ring of a simple left A-module. We see from this Theorem that the property of being a simple ring is left-right syinmotric, ie. we could equally well have defined it to be semisimple as a right module, without proper two-sided ideals #0. We can now more generally characterize semisimple rings: ‘Theorem 11.B The following conditions are equivalent for a ring A: (a) Ais a semisimple ring. (b) All left A-modules are semisimple. (©) Every left ideal of A is a direct summand. (@) A Mp (D3) X +++ x My, (Dr) for division rings Dy,...,De Proof. (a) = (b): If Ais semisimple, then clearly every free A-module is semisimple. Since every quotient module of semisimple module is semisimple, it follows that all modules are semisimple. (b) = (e)= (a) are special cases of Theorem 10.A. (a) + (@) follows from the preceding, in particular Lemma 11.2. ‘The equivalence of the conditions (a) and (d) is known as “Wedderburn’s structure theorem”. It was however first obtained by Molien (1893) in a particular case, and Wedderburn (1907) treated the case of a finite-dimensional algebra over an arbitrary field; the general case was obtained by Artin (1927) and Hopkins (1939). If A is a semisimple ring, then (1) gives the decomposition of A into simple rings Ay,, which are called the simple components of A. A semisimple ring is both left and right noetherian. In fact, each left ideal is principal let ideal, generated by an idempotent element (Exercise 4:3) ‘The importance of the concept of semisimplicity is enhanced by the following result (Maschke’s theo rem): 5 Theorem 11.C Let G be a finite group and let K be a field whase characteristic does not divide the order of G. Then K{G| is a semisimple ring. Before going into the proof, we shall consider some preliminaries. First we recall the definition of characteristic of a field. For any field K there is a unique ring homomorphism y: Z —+ K’ given by nis nl. Then Kerg is either 0 or pZ, for some prime p, and we then say that K has characteristic 0 or p respectively. In the first case we have that Kis an extension of a field isomorphic to Q, and in the second case of a field isomorphic to Zp. These smallest subfields of K are in both eases called the prime field of K. In the second case we have p- = 0 for any z € K, and y(n) is invertible in the prime field if and only if p does not divide n. Let G be a finite group and K any field. If M and N are modules over K(G] and a: M—+ N is a ‘K-linear mapping, then we can modify « into a K(GHinear mapping @ by setting Laas). 16 For @ is clearly K-linear, but is also K[G}-linear because for any go € G we have @(a0z) = J) 9a(9~*90z) = J) goha(h-*z) = 7 ™ (2) 0a(z). Proof of Theorem 11.C: Suppose now that the characteristic of K' does not divide the order of G. In order to show that K[G] is semisimple, we show that every left K[G-module M is semisimple, and this we do by showing that any submodule L of Mf is a direct summand. Now since L is a subspace of M as vector spaces over K, we have that L is a direct summand of M as a K-subspace. Thus there exists a K-linear projection x: M — L with |= Idp. We define an “averaging” ¢: M— L of x as ¢(2) = 12 where n is the order of the group G and + is the inverse of n in the prime field of K. We have seen that @ is K[Ghlinear, and for z € L we get #2)=1 Dox) =1 9-9 7 7 since g~"z € L. It follows from Exercise 4.2 that L is a direct summand of M. ‘Examples - 1. If G is a finite commutative group, then the ring K(G] is commutative. If then the characteristic of K does not divide the order of G, we know that K{G] is a direct product of fields (which are extensions of the field K). 2. Suppose A is a finite simple ring, consisting of m elements. Then A M,(D) for some finite division ring D. But in that case D is a field K (every finite division ring is a field by another famous theorem of Wedderburn). So A * Ma(GF(p*)) for some k, where p is the characteristic of K and GF(q) denotes the finite field with q elements. Note that K is isomorphic to the center of A, since that corresponds to the diagonal matrices with constant entries from K Exercises 11,1 Determine all left ideals, and their corresponding idempotent generators, of the ring Ma(D) for a division sing D. 112 Show that every quotient ring of & semisimple ring is semisimple. 11.3 Write the group ring Z(G}, for a cyclic group G of order three (Exercise 13) explicitly as a direct product of fies. Do the same for Za[G] when G is cyclic of order two. 114 Is the group ring Z[G] semisimple when Gin cyclic of order two? 26. 12. Finitely generated modules over a PID ‘Throughout this section we assume that A is a commutative principal ideal domain (PID). Recall that an element of A is called irreducible if it is noninvertible and cannot be written as a proper product of noninvertible factors. Every noninvertible element of A can be decomposed as a product of irreducible clement, and this decomposition is unique up to ordering and invertible factors. ‘The irreducible elements @ are those for which the ideal Aq is a maximal ideal of A, of equivalently, a nonzero prime ideal. ‘The irreducible clements are therefore also called primes. Our aim in this section is to describe the strucure of all finitely generated A-modules, and we begin by studying the torsion-free modules. Lemma 12.0 Every finitely gencrated torsion-free A-module is isomorphic to a submodule of a finitely generated free module, Proof. Let M be torsion-free with generators yi,...,tm, and suppose M is not free. ‘Take a maximal linearly independent subset of these generators; after reordering we can assume that it is {y1,..., Ya} for some n < m., For every 14 with i > n we then have some linear relation ajy:-+ ai1ys + ++*-+ GinYn = 0 with a; #0. We multiply these a; together, as a = dng, -...°Gm- Now F = Ay; +--+ Ayn is a free module, and ayy € F for all i, We can therefore use the mapping M+ M given by 2++ az, which is a monomorphism since M is torsion-free, and which has its image within F. Note that the proof of this result works for any commutative integral domain A. Theorem 12.A Every submodule of a finitely generated free A-module F is a free module whose dimension is < dimension of F. Corollary Every torsion-free finitely generated A-module is free. Proof of the Theorem. Suppose F is free on a basis z1,...,2q and that M is a submodule of F. We shall use induction to prove that M, = M (Az +++ Az,) is free for r sn. Now My = Mn Ary Since Az; ® A, we have that M; is isomorphic to an ideal of A, s0 it is either free or = 0. Suppose M, is free of dimension 0 in Z. ‘Then Z/nZ ~ Z/pi'Z x --- x Z/pZ is a primary decomposition. Clearly the same kind of decomposition holds for any cyclic module A/Aa over a PID A 2. For each prime number p we define Zp to be the p-primary component of Q/2. Clearly Z,= consists of the equivalence classes of elements in Q of the form m/p'. We have the primary decomposition Q/Z= Bye Zp If M is a torsion module and z € M, then the ideal Ann(z) is generated by some b € A, and we say that b is the period of the element z (it is determined only up to an invertible factor). In order to determine the torsion modules, it suffices by Theorem 12.B to consider p-primary modules, i.e. modules ‘M with M = M(p). If M is a finitely generated p-primary module, then clearly p’ M = 0 for some r. If = € M in this case, then since p” € Ann(z), we have that the period of z is p” for some n pe € Aa = Apb <=> cE Ab. Now Ab/Aa = Ab/Apb © A/Ap under the mapping A/Ap — Ab/Apb given by (z] ++ (2b), and we may conclude that Lp has dimension 1 as a vector space over A/Ap. If we apply this to the decomposition (1), we find that the dimension of My over A/Ap is equal to the number of a; with pla Now consider in particular a prime p dividing a. ‘Then p divides all a; in (1), 80 the dimension of Mp over A/Ap isn. Suppose we have another decomposition of M of the same kind, say M = A/Aby @-+-A/Abmy With b3[ba]..-[bm, then dim Mp = n implies that p divides the n last bi, so n f(a) = 0 for all z € L <= f(a) = 0, 50 the minimal polynomial for 7's in this case the usual minimal polynomial for a over K. 2. Let V = R?, with T: (z,y) + (~y,2). Then V is a simple Rf]-module since it is generated by by ‘anyone of its nonzero elements. In fact, if vo = (zo, vo) # 0, then tv = (—Yo,z0), and any(e,y) € V can be written as 2204 Yo qq = H0— 20 ate a+ So the module is cyclic over R{f], with invariant factor (? +1. Note that as a C[f]-module, it is no longer simple (2,y)=(4+ bm, with a= ‘Tosce what Theorem 13.A really says, we must find out what a cyclic K[f-module is. We suppose V is acyclic K{t}-module, with minimal polynomial gr, and we let v be a generator for V. So V © K(/(or) asa module, and gr(t)u = 0. Write gr = t#-+agit*!4.--+-at-+a9. The elements v, Tv, Tv, ...,T!40 form a basis for V over K. In fact, these elements are linearly independent because a linear relation between them would give {(TT) = 0 for some polynomial f of degree < d, which is imposssible. ‘They also generate V over K,, because every z € V can be written as z= f(T)» for some polynomial f, and we have f(t) = 9(ar(t) + r(t) with degr < d, s0 f(T) = r(T) and z = r(T)v. When we write T in matrix form, we can define its matrix (aij) relative to a basis 21,...,2, by T(z) = Dy aij2; since the ring K(q is commutative (Sec. 5). Relative to the basis v, Tv,T%v, ..., 74-1», the matrix of T has the form 000 ..00 ~a 100... 00 —« Mtr or. bevels, Boro 10s ans O10) Oo Ure ees ‘This matrix is the companion matrix C(¢r(t)) of the polynomial ¢r(t). In the general case described by Theorem 13.A we can choose a basis for V so that T' relative to this basis is represented by a block matrix Ea) 0 C(an(t)) 0 Gae(t)), ‘This matrix is called the rational canonical matrix for ‘T when q,|¢2|--- [gr Even if the rational canonical matrix for 7’ in many cases is useful, it has the drawback that it is in general not diagonal even when T’ may be represented by some diagonal matrix. We shall therefore consider a matrix for T, called the “Jordan canonical matrix”, which becomes diagonal whenever T has any diagonal representation. We first consider the case of a cyclic module V = K[tJ/(gr(t)) with gr(¢) = (t — a) for some a € K and d > 1. In that case we ean choose as basis over K for the cyclic K(f}-module V the elements v,(7~a)p,...,(T~ a)*"v; they are in fact linearly independent since a linear relation between them would give a polynomial 9 of degree < d with (7) = 0, and they must then form a basis since V has dimension d. Relative to this basis, the matrix of T° takes the form Oi j-2y FOND, 0 00 ae...) OO) (2) ie D)tcar 0 0 la since T(T — a)¥ = (1° a)##1 4 a(T'— a)! and (T— a)* =0. Now consider the general case, where V is not necessarily cyclic. ‘The minimal polynomial gp has factorization gr(t) = pf" -...- p{+ into prime powers, and we get a primary decomposition of V as V = V(p1) ®---@ V(pq), where each V(p,) is annihilated by pi and is a direct sum of eyclic Kt} ‘modules. Every submodule of ¥ is by definition stable under 7’ (ic. is mapped into itself by 7), and we can therefore choose a basis independently for each one of the eyelie summands of V. When into linear factors over K,, the matrix for T over such a cyclic summand takes the form (2). ‘Th together form a basis for V relatively to which the matrix of 7’ becomes block mat a 4 ® : ; te where each J; has the form (2). Each block corresponds to an elementary divisor (¢ — a,), and J; has diagonal entries all equal to a;. ‘The matrix (8) is the Jordan canonical form for T (and for a matrix representing 1’). We thus have: ‘Theorem 13.B If T: V ~+ V is a linear transformation of a finite-dimensional vector space V over K, and the minimal polynomial of T splits into linear factors over K’, then there is a basis for V relative to which Tis represented by a matrix in Jordan canonical form. Example ~ 3. Suppose gr = (ta)... (t~ ay) for different a;. Then V = Vi @ +++ Vp with each V; annihilated by T’—a; and therefore isomorphic to a direct sum of cyclic modules of the form K{O)/(¢ = aj). The Jordan normal form of Tis then a diagonal matrix cn ‘We shall finally see how the minimal polynomial is related to the characteristic polynomial. So let T 4s before be a linear transformation of a vector space V of finite dimension n over K. Let M be a matrix representing T'in some basis for V. The characteristic polynomial of T is Pht) = det(tiy ~ M) € K{Q. Ibis independent of the choice of basis since det(tl, —N~! MIN) = det(N~*(tI,—M)N) = det(tI,—M). We can choose a basis for V in which M is written in Jordan normal form (3). If we let Pr,(t) be the characteristic polynomial of the block Js, then Pr(t) = Pr.(t)-...- Pre(0). ‘Theorem 13.C If gu(t),..-y4e(6) are invariant factors for T, then P(t) = qu(t)-----ae(t)- Proof. We decompose V according to Theorem 12.D, as a direct sum V = Vi @-+-@ ¥; of eyc modules, corresponding to the invariant factors q1(t), K(t} =14e(t). For each V; we chose a basis such that. 31 TV; becomes represented by the companion matrix C(¢(t)). Then T gets represented by a block matrix composed by the C(qi(t)) and we have Pr(t) = det(tln ~ M) = T] det(tlm, — Clas(t))) = T] a(t) i i according to the following Lemma. Lemma 13.a If f(t) is a monic polynomial of degree n, then det(tl, — C(f(¢))) = F(t) Proof. Let f(t) = 0" + anit"! +----+ayt +a. Then 10g oe ay 1 t 0 0 0 a t o 0 a Un — CK) = GeO Orseee rw, 7 Che nyse le a are We compute the determinant of this matrix by expanding along the first row, using induction on n. We get det(tIn — CCFO) = tar + ant +--+ an at? $:1°71) + (-1)"Fao(—1)"4 sO. 8 ‘Theorem 13.C implies in particular that the minimum polynomial and the characteristic polynomial for T have the same irreducible factors. Another important consequence is the Cayley-Hamilton theorem: Theorem 13.D Pr(I)=0, ie. T satisfies the polynomial equation det(t! — 7) = 0. ‘This result actually holds for matrices over any commutative ring A, ie. uot only for fields. Example ~ 4. Let V be the real vector space of polynomials of degree <5, and let T: V + V be the linear transformation p(z) ++ p(z) + p/"(z). Relative to the basis {2°, 2*, 2°, 2?, 2,1} for V, the matrix for T becomes 100000 0 10000 01000 0 20100 0 06010 000201 lis characteristic polynomial is (t ~ 1)°, and hence the minimial polynomial of T is a divisor of (¢=1)®. We have that (r-1)25 = (-1)z* = 1227, (1): (7-1)? =2, (f-)z= (r-1)1=0. We see that (1’—1)° annihilates V, s0 it is the minimial polynomial gr(t). We also see that z° and 2* cach generates a cyclic module over K (J, both corresponding to invariant factors (t— 1). The Jordan canonical form of the matrix becomes 100000 110000 011000 000100] ooo0110 ooo011 relative to the basis (28,2028, 1202, 24, 1222, 24} 32 Exercises 18.1 Show that a matrix over a field K can be diagonalised if and only ifits minimal polynomial is a product of distinct linear factors in K{. 18.2 Determine the characteristic and minimal polynomials for the following matrices over C. Determine bases in which the matrices have the Jordan canonical form. ae 100 SAC yal On @ G a; @ [1 10); iy [9 7 aati i 188 Let M and W be quadratic matrices over a field. Show that the minimal polynomial for Mo ow is the least common multiple of the minimal polynomials for M and NV. 184 Let T be linear transformation of a finite-dimensional complex vector space V. Show that the following assertions are equivalent: (a) T=Idy +N, where NV is nilpotent, (b) There exists a basis for V in which T’is represented by a lower triangular matrix with all diagonal elements equal to 1. (©) All zeros of the characteristic polynomial for 7’ are 1 13.5 Let V be finite-dimensional vector space over the field K,, and let $ and 7’ be linear transformations of V. Denote by Vs and Vr the K{é}-module structures on V given by 3 and T respectively. Show that Vs and Vp are isomorphic K{}-modules if and only if $ and T are similar (ic. S = RT'R™? for some automorphism Rol V). 14. Tensor products We shall introduce the tensor product of two modules as a linearized product. To motivate it, we begin with an example. Example. Let V be any real vector space. We wish to define a “complexification” of V which makes it possible to operate on V with complex scalars but otherwise does not change V- more than necessary. To do this, it is natural to consider pairs (z,»), where z € C and v € V, with the operation of C given by 2/(z,v) = (2'z,v). But we do not want to consider the new space as the direct product Cx V, but rather want to have bilinear laws of addition like (z, v)+(z,v’) = (z,v-+v") and (2,v) + (2',v) = (z+ 2',v). Furthermore we wish to have (z,v) = (1,20) for € I since the reals should operate on the space as before. We accomplish this by “dividing out” with the desired relations, to get the tensor product C @x V. Let A be a ring, and let L be a right A-module, while M is a left A-module (note the different sides of operation by A). We shall as in the Example define L @q M, which in general will be only an abelian group and not an A-module, as a linearization of L x M which identifies (2, y + y/) with (z,y) + (z,1/) etc. This linearization can be characterized by a “universal mapping property”. We define a mapping J: Lx M— G, where G is an abelian group, to be balanced if Szu+y) = Sz,v) + fev) Le +2',u) = fley) + H(2,y) M(2a,y) = f(2,ay) for alla € A ‘The tensor product of L and M over A will be defined so that it is universal for balanced mappings from Lx M, ie. as an abelian group T with a balanced mapping r: L x M— T, such that: (+) for any abelian group G and any balanced mapping f: L x M —+ G there exists a unique group homomorphism gp: T + G with gr = f. LxM—1—T | G Note in particular the important requirement. that y should be unique with the stated property. This universal property uniquely determines the tensor product, up to canonical isomorphisms: 33 Lemma 14.a Suppose T and T’ are abelian groups with balanced mappings r: Lx M — T and 1: LX M +7" satisfying (+). ‘Then there exists a unique isomorphism a: T +1" such that ar = r!. Proof. By the universal properties of T and there exist a homomorphism a: T’—+ T° such that ar = 1’, and similarly for 7” there exists 8: 1’ + T with fr’ = r. We now use the uniqueness property in (+) for Ba: T + T, for which we know that far = fr’ = r. But also Id has the property that Idpr and uniqueness implies that Sa = Ip. Similarly we have that a = Id, so a is an isomorphism. Because of this uniqueness result we could have defined the tensor product by the universal mapping property, and then prove its existence. We shall however prefer to use an explicit construction as its definition. In order to construct the tensor product of L and M we first consider the free abclian group F on the set Lx M. So F consists of all finite linear combinations > n;(2;,y) with nj € Z, 2; € L and tk € M. We obtain the desired relations by taking the subgroup R generated by elements in F of the forms (z.u+¥)-(9)-@y) (e+2/,)-(9)-(@y) (za,y)—(z,ay) for ae A. Let T= F/R and define r: Lx MT as r(2,y) = [(2,y)]. ‘Then 7 is balanced since rut) = [ey ty)] = (2) +eV)) = leu) + 12), and similarly for the other two conditions. t We shall verify the universal property (#) for r. Suppose: L x M —+ G is any balanced mapping, Since the set Lx M is a basis for the free Z-module F, there is @ unique extension of f toa homomorphism a: FG. We assert that a =0 on R. Now f(z,y+y’) = f(z») + S(e,y/) implies that a(z,y + y) — a(z,y) ~ a(z,y) = 0, and similarly a is zero on the other generators of R, and therefore a(R) By Exercise 2.6 there exists a unique factorization of a over F/R = T, i. a unique homomorphism :T—G with gy =a. Then pr = f, and ¢ is unique with this property because of the uniqueness in the construction of e and ¢ (verify this!). LxM—++F—1.T= F/R NLA GC Weesball denote the tensor product we have just constructed by L@4M (the index A may be suppressed when it is clear from the context). This is an abelian group, generated by the elemente r(z,y), which we denote by 2@y. The elements of L@4 M can be written as finite sums > n,(2@ y), with relations (et+2)@y=z2@yt2@y, rO(yt+y)=zOyt+z0y, ra@y=20ay. When constructing some homomorphism from L @4 M, it suffices to define it on the generators x ® y, but one must then verify that the relations between these generators are respected; usually one does this by defining a balanced mapping on L x M and then uses the universal property (#) to obtain the homomorphism. Examples ~ 1. We have Z/nZ @z Q = 0, because it has generators P 2g ayatee injoC=(den- = imine ven Here we have used that 0@z =0, which holds because 0@ 2 +0@z=(0+0)@z=00z. 2. We have A@4 M = M, which are isomorphic as abelian groups. In fact, there is a mapping a: M ~ A@a M given by z++1@ 2, which clearly is a group homomorphism. In order to define 8: A@4M — M, we first define a mapping f: Ax M — M as (a,2) ++ az, and since this mapping clearly is balanced, it induces the desired 2, with f: a@z ++ az. We have Ba(z) = 6(1@ 2) = 2, while a8(a@ z) = a(az) = 1@az =a@z. Since the elements a@ z generate A@4 M, it follows that we have obtained an isomorphism M * A@, M. 3. For any family (1; |i € I) of right A-modules we have an isomorphisin (DL) Oa M = PlLi a M), 7 7 4 We define homomorphisms (@, Li) ®a M —+ @,(Li @a M) with (24); @ y + (24 ®@ v)r, and D(Li@aM) — (@, LiOaM with (2@ys)s + F(N (zi), where Aj: Li —+ @, Li canonically. It is clear that these mappings are inverses of each other, once we know that they exist and are homomorphisms. But this follows from the observation that ((z;)s,y) ++ (2: @ y)r and (zi,y) +» (As(2¢) @ y) are balanced mappings. AAs in the case of the Hom-groups, we can obtain module structures on the tensor product by considering bimodules. Suppose J is a B- A-bimodule, while M still is a left A-module. We ean make L®4 M into a left B-module by defining 6(z @ y) = (bz) @ y. This operation is well-defined, because the mapping (z,y) + (be,y) is balanced and therefore induces a homomorphism z @ y ++ bz ® y. Clearly it makes La M into a left B-module. Similarly we get a right C-module if we instead assume that B is an A-C-bimodule. If both L and M are bimodules aa above, then L@4 M becomes a B — C-bimodule. An important special case is when the ring A is commutative, in which case we find that L@4 M always is an A-module. We then also have an isomorphism L®4 MM @q L, given by z@ y+ y@z. Examples —4, The isomorphism A@4M & M considered in Example 2 is actually an isomorphism of left A-modules when A is considered as an A— A-bimodule. 5. Let F be a free right A-module. ‘Then F ¥ AC, and we get FOaM= AMO, M = QA @, M)= GM = MO, r r which is an isomorphism of abelian groups. But AY) @4 M = MU) is by the preceding an isomor- phism of A-modules. 6. If L and M are free modules over a commutative ring A, with bases (z4)r and (yj)y respectively, then L@q M is also free on a basis (2; @ yj)rxs- We namely have AW) @ AV) = AUX) by the preceding Example. The elements of @q M can thus be represented by matrices (aj) relative to the given bases. Suppose in particular that L and M have finite bases {z1,...,2m)} and {yi,...,ya}- ‘Then the dual modules L* and M" have the dual bases {23,...,2%,) and {yj,...,y,). On the other hand there is a dual basis 2; for (L@4 M)* for which z4(z, @ v,) has the value 1 if i =r and j = s, and 0 otherwise. There is an isomorphism (L@q M)* © L* @4 M*, which makes 2; correspond to 2]. yj. The details are left to the reader. 7. Suppose y: A—+ B is a ring homomorphism. Then we may consider B as a B — A-bimodule, so that we for each left A-module M may form 5 @4 M, which then becomes a left B-module. We call this the “extension of scalars” from A to B. 8. Using the embedding ¢: R —+ € for extension of scalars, we obtain the complexification C @a V of a real vector space V. Writing C * R@ R, with a+ ib r+ (a,b), we find that CoRV >V OV with (a+ ib) @ v ++ (av, bv). So V.@V may be regarded as a complex vector space by means of the operation (a+ ib)(2,y) = (az — by,be +ay). Exercises 14.1 Let A be an integral domain. Show that if Tis a torsion module and D is divisible module (Exercise 7.8), then T@4 D 14.2 Show that if B isan R-algebra over a commutative ring R, then multiplication in B induces a homomorphism B Gp B ~ B of B— B-dimodules, with b@ Wr bY. Is this an isomorphism when: a) Bis C as an R-algebra; b) B is Q as a Z-algebra? 143 Suppose f(t) is an ireducible polynomial over a ld K and that a is a root of f. Show that if J is any extension field of K, then K(a) @x L % £{8/(f). 144 For a commutative ring A one may consider A both as a right module V over the matrix ring My(A) and as aleft module W of M,(A). Determine V @,¢a) W- 14.5 Let A = K[X,Y] be the polynomial ring in two variables over a field K, and let I be the ideal (X,Y) of polynomials with constant coefficient = 0. Show that in 1@qT one has X @Y # YX, while XY(X@Y -Y@X)=0. 15. The tensor product as a functor Like the Hom-functor we may view the tensor product as a functor of two variables. Let A be a ring and suppose we have homomorphisms A: L — L! of right A-modules, and jt: M —+ M’ of left. A-modules. 35 ‘There is then induced a homomorphism a : L@4M — L’@4M! with z@y++ Az) p(y). Its existence follows from the fact that the mapping L x M — L! 4 M', given by (z,y) + \(z) ® u(y) is balanced, ly verified. We write (improperly) \@ ut: L@q M —+ L! @q M’ for this homomorphism. So we have that Q@n)(z ey) =Az) @n(y) ‘To.sce that this makes the tensor product into a functor in two variables (covariant in each), we consider homomorphisms pe, Mt 2M". We then have 22.8 p' = (X'@ p')(A.@ p), because on the generators of L@4 M we get (8 p'a)(e @ v) = NAz) @ H'n(y) = 2’ @ p')(A(z) @ lu) = (1 @ 2'\(A@ H)(z@ v), When A= 2’ = Iz, we have in particular that (Idz @ w')(Idz © 4) = Ir © wp, 80 L @a - is a functor from left A-modules to abelian groups. Similarly, «@4 M is a functor from right A-modules to abelian groups. ‘We shall study the exactness properties of these two functors, and by symmetry it suffices to consider one of them, say L@q -. From Example 14.3 follows that it preserves exactness of split short exact sequences. But in general this functor does not preserve monomorphisms: Example ~ 1. Let A= Z with L = 2/22, and p: Z— Z given by n+ 2n. In that case we have that Id @ p: 2/22 @x Z —+ 2/22 Gy Z sends [m] @ n to [m] © 2n = mn] @n = [0] @n = 0. But 2/22. @2 Z¥ 2/20 # 0, 80 1d ® pis not a monomorphism. ‘Thus the functor is in general not exact. But it is right exact: ‘Theorem 15.A_ The tensor product functor is right exact in both variables. Proof. Let 0 Ms M' “', M0 be an exact sequence. We shall prove that Wey, LoaM Loam’ 22H 5 9,M"—0 is an exact sequence. First we note that 1d@y is surjective, since every element of L@4 M” can be written as Da Ou = Dzi@y'(x) = (ld@y’)(S 219). Iie also easy to see that Im (Id) C Ker (Id@y’), because (Id @ p’)(Id @ p) = 14 ® (n'p) = 1d @0=0. It requires more work to show that Ker (Id @ y') C Im(Id @ 1). Consider the commutative diagram Wop Mop tem—2! rem L@M" 0 | : Lem’ (L@ M)/im(Id @ ») where the lower row is exact, and the existence of € follows from the fact that Im (Ids) C Ker (Id@y'). We wish to construct an inverse 9 to €. For this purpose we first define a mapping h: Lx M" —+ (L @ M')/Im (Id ® p) as h(z,y) = [2 @ 2] for some z € M“ with p/(z) = y. This mapping is well- defined, because if also p(x’) = y, then z— 2 € Kery! = Imp, 60 2 @ (z — 2’) € Im(Id @ p), and hence [2 @ (2 — 2')] = 0 and fz @ z] = [z@ 2’). Clearly h is balanced, so homomorphism n exists with nz © [2]) =[2 © 2, It is easily seen that 7 is an inverse of €. So € is an isomorphism, and the since the lower row of the diagram is exact, it follows that also the upper Tow is exact. Example ~ 2. We shall compute A/a@, M for a right ideal a. Now we have an exact sequence 0-a-2+A*. A/a 0, which induces an exact sequence ao 2214. 497 LOM siggy 0. So A/a@ M © (A@ M)/Ker(8 Ia) = (A @ M)/Im(a@ Id). Now A@ M ~ M, and with this isomorphism we have that @@ Id: Da; @ zi ++ Dai @ x = So the image of a @ Id in M is the subgroup aM. Hence A/a@, M © M/aM, with [a] @ 2 — [a ‘We have previously proved the commutativity of the tensor product over a commutative ring. We now prove the associativity: 36 ‘Theorem 15.B Given modules La, My and pI, there is a canonical isomorphism (L@qM)@_N = L@q(M Op N). Proof. The isomorphism shall of course be given by (z@y) ® z $3 x @ (y@ 2) on generators of the tensor products. It is obvious that these homomorphisms are the inverses of each other, if they are well-defined. We prove the existence for €: (L@4 M) @p N + L@a (M @p N). First we define for each 2 € N a homomorphism a,:.M —+ M @p N by a,(y) = y@z, which induces f, = Idy @a,:L@4 M L@a(M @p N). Next we define f: (L@a M) x N + L@q (M x N) as f(u,z) = B,(w). Thus in particular f(z ® y,z) = 8,(2 @y) = 28 (v@ 2). (L@sM)@nN (L@aM)xN L@a(M@pN) ‘The mapping f will induce the desired once we know that J is balanced. ‘The linearity of f in u is clear since B, is « homomorphism. For the linearity in z we first consider the case when u=2@ y, and get S(2@y,2+2') = 20(vO(e+#’)) = 20(vO2+ 82") = 2@(ve2)+2@(yO2') = f(2@y,2)+f(zOY,2’), and clearly we then have f(u,z+ 2/) = f(u,2) + f(u,2/) for general u by the linearity of 8. We finally note that F((@@ y)b, 2) = f(z @ yb, z) = 2 (ybo =) By this we have shown that f is balanced. 2@(y@bz) = f((z@y), bz). ‘We shall prove another kind of associativity result, which couples together the Homfunctor and the tensor product. We consider modules La, aMz and Ng. Then L@4 M will be a right B-module, while Homp(M, 1) will be a right A-module. ‘The following isomorphism corresponds to the set-theoretical formula S?*9 = ($9)? where f: P — 59 corresponds to the mapping P x Q —+ S given by (p,q) F(p\(a). ‘Theorem 15.C For any modules La, 4Mp and Np there is a canonical isomorphism Homp(L @4 M,N) % Homa(L, Hom (M,N) Proof. The B-linear homomorphisms L@4M —+ N correspond to the balanced mappings f: Lx M —+ NV for which f(z, yb) = f(z,y)b, and these correspond to A-linear homomorphisms ig: L —» Homp (M,N) by e(2): y+ f(z,y). The details are left to the reader. Example ~ 3. Let L and M be finitely generated free modules over a commutative ring A, and let * denote the operation of taking the dual module, Then (L@ M)* = Hom(L@ M, A) © Hom(L, Hom (M, A)) = Hom(L, M*). Using the canonical isomorphism L £**, we in particular obtain that L.@ M = (Hom(L,M*))*. Exercises 15.1 Determine Zm @x Zn. 15.2 Consider the monomorphism ¢: Zq ~+ Za, and determine Id @ 4: Za Oy Za -+ Za Ox Za 15.8 Show that if is a commutative ring, then A/a@a A/b = A/(a-+ 6) for any ideals @ and of A. 454 Show that if Land Af are simple left modules over a semisimple ring A, then L* @4 M % Hom a(L, M). 155 Let V and W be finite dimensional vector spaces over a field K. Show that there is a canonical (ie. not depending on a choice of bases) isomorphism V @ W & Hom (V*, W). 15.6 Show for any left A-modale M that there exists an “evaluation” homomorphism M” @)4 M —> A. 15.7 Let V be a finite-dimensional vector space over a field K. Consider the homomorphism tx: Hom (V,V) — V*@V —+ K obtained by composing the homomorphisms in the two preceding Exercises. Show that if an ‘element of Hom (V, V) is represented by a matrix (a:,), then tr(as;) = > ais (the ¢race of the matrix). 15.8 Show that for exact sequences K “+L 4. M — 0 and K! 1! 2. M’ — 0 of right (respectively left) ‘A-modules there is an exact sequence tceek (R41) (La K') —+ Lal’ op MOaM’—0. 37 16. Algebras We shall return to the study of algebras begun in Sec. 1. Let R be a commutative ring. All tensor products in this Section will be taken over R, and they will then be R-modules. We also note that the tensor produets over R ate universal for R-bilinear mappings, ie. mappings f: Lx M —+ N for modules L, M and N such that f is not only balanced but also satisfies f(az,y) = f(z,ay) = af(z,y). For each Realgebra A we have a bilinear mapping A x A — A given by muliplication, and it induces a linear mapping a: A@p A— A. The associativity of multiplication in A can be expressed by saying that the diagram A@A@ AMO, AQ 0) eo [x A@A—z— A is commutative, where we have identified (A @ A) @ A with A@(A@ A) because of the associativity (Theorem 15.B). We can actually formulate all of the algebra axioms in such diagrammatic form. ‘The distributivity is inherent in the properties of the tensor product by way of the mapping a. It remains to consider the existence of an identity element. Let «: R—+ A be the mapping ¢(r) =r 14, where 1a is the identity element of , and let ¢: R@ A—+ A and 7: A@ RA be the canonical isomorphisms. ‘The fact that 14 is a left and a right identity for multiplication can be expressed by saying that the two triangles in ou" @A swe 2) Re — eee oR A are commutative, It is mow clear that conversely we have: ‘Theorem 16.A If an R-module A is equipped with R-linear mappings a: A@A—+ A andi: RA such that the diagrams (1) and (2) commute, then A becomes an R-algebra with multiplication given by @ and with identity element (In). ‘The tensor algebra We shall shove together all the iterated tensor products of an A-module into an algebra, which will be called the “tensor algebra” of the module. It will be a graded algebra in the following sense. An Realgebra A, with multiplication a: A@ A —+ A, becomes a graded algebra when it is written as a direct sum A=@QAn, nex where each Ay is an R-module and a(Am @ An) CAmgn- In particular we have that Ag is a subring of A containing all elements r-1, and that itfurthermore is a two-sided ideal of A. ‘The elements of Ay are sald to be of degree: Cw Example ~ 1. The polynomial ring R{th,...,ts] is a graded R-algebra, where the elements of degree n are the homogeneous polynomials of degree n in the usual sense. let M be an module. The tensor algebra @ M of M is the graded algebra with (@ M)o = R and (@M)n = M@M@---@M (taken n times) for n > 1. We shall write @M for (@M)q. ‘The ‘multiplication is defined by linear extension from the isomorphisms (M) @ (SM) + "°M. ‘We shall in particular consider the “classical” case of the tensor algebra of a finite-dimensional vector space V over a field K. The elements of V are called contravariant vectors, while the elements of the dual space are covariant vectors. Let {z1,...,2n} be a basis for V. We shall here write the dual basis as {21,...,2"}. The space ¥;,. = (6V) @ (@V*), where r,s > 0 and with Vo = K, consists of tensors of type (r,8). Its elements ean be written in the form ra 24, BO xi, OL]... @ ahr 38. for uniquely determined scalars a'+’"4" (the sum is taken over all indices which appear once as an upper index and once as a lower index; this is the summation convention, by which one also often suppresses the summation sign). The mixed tensor algebra of V is the “bigraded” algebra @,,, Vr, with the obvious multiplication. 17. Flat modules Let A ba ring. A right A-module L is flat if the functor L @q - is exact, ic. if for every submodule M’ of a left A-module M it follows that that L @ M’ — L® M is a monomorphism. Lemma 17.a If (1; |i € 1) is a family of right A-modules, then @, [; is flat if and only if every L; is fat Proof. For every monomorphism M! —+ M we have a commutative diagram (@i)om Q@ijeom ; t Ql @ MH’) —- iti M), where the vertical arrows are isomorphisms by Example 14.3. It follows that the upper arrow is @ monomorphism if and ond only if the lower one is. § Examples — 1. A as a right A-module is flat by Example 14.2. From Lemma 17.a follows that every free module is flat, and then by Exercise 7.5 that every projective module is flat. 2. The Z-module 2/22 is not flat by Example 15.1. More generally it is clear that a nonzero finite abelian group can never be flat as a Z-module. 3. Still more generally we can show that if A is a commutative ring, then every flat module L must be torsion-free. For suppose s is not a zero-divisor, and consider the monomorphism a: A —+ A siven by a++ as, If L is flat, then Id @ a: L@ A — L@ A is a monomorphism. But under the identification of L@ A with L, this mapping is z—+ zs. Hence L has no torsion. ‘The following criterion for flatness is very useful since it reduces the task of testing preservation of left exactness to a very particular kind of monomorphism, ‘Theorem 17-A A right A-module L is flat if and only if the canonical homomorphism L®q a—+ L is a monomorphism for every finitely generated left ideal @ of A Proof. The homomorphism in question is Q) a:L@a—L@AXL given by 21 ® a; ++ SD 2iai. So it is clear that the condition is necessary for flatness. Suppose on the other hand that (1) is a monomorphism for all finitely generated left ideals a. We shall prove the flatness of L in steps 1) We first show that (1) is a monomorphism for all left ideals a. If y = 37 21a; € LO a with a(y) = 0, then the elements aj lie in a finitely generated left. ideal @ C a, and the composed homomorphism L@ e+ L@a— L is a monomorphism, which implies that > 2; @ a = 0 already in L@ @. 2) The next step is to show that L® M’ + L@ M is a monomorphism when M’ is a submodule of a finitely generated free module M. We do this by induction on the number n of generators of M. The case n= 1 is clear by step 1). Suppose M’C A". Consider the commutative diagram where the lower row is the splitting off of the first (say) component of A", and where we define M" = A(M') and a= M'MKer f= M'0.A. The vertical arrows are thus monomorphisms, and the upper row 39 is clearly exact. Tensoring this with L we get a commutative diagram Lea? .Lem’—_* Lem" 0 ior 0 LeA Lea Lea 0 with exact rows. Here A is a monomorphism by step 1), and 7 is a monomorphism by the induction hypothesis. ‘The proof of (i) in Lemma 6.a goes through also in this situation, and shows that pis a ‘monomorphism, 3) More generally we have that L @ M’ — L @ M is a monomorphism when M’ is a submodule of any free module M. For suppose 2 ® ys € L @ M! with 2; @ y = 0 in L@ M. The elements z; all lie in some finitely genererated free direct summand Mo of M. We then have a commutative diagram Lem’ Lem L@ (MoM M') + L ® Mo, where the lower map is a monomorphism by 2), and the right hand arrow is a monomorphism because ‘Mo is a direct summand of M. We therefore must have > 2 @ y; = 0 in L@ (MoM M’), and then itis 0 also in L@ M’ 4) Finally we consider the general case of a monomorphism M’—+ M. Write M as a quotient of a free module F, by means of a homomorphism @: F + M. Writing IT = @-1(M") C F, we get a commutative diagram 0 K—+H ——.’ —__. 0 K PF. M 0 B with exact rows, where the vertical arrows are monomorphisms. ' fensoring with L gives the commutative diagram Lok Len L@M’ 0 lee? ea Lek LoF LoM 0 with exact rows, where jis a monomorphism by 3). An easy dingram-chasing shows that 7 is a monomor- phism. Examples ~ 4. As a partial converse to Example 3 we can prove that if A is a PID, then a module is fiat if and only if it is torsion-free (this result does not extend to general integral domains, by Exercise 17.4), Suppose L is a torsion-free. By the Theorem it suflices to show that L@a— Lis a monomorphism for every ideal a. Now a= Aa for some a € A, and we can assume a # 0. Consider the homomorphisms A eee, where 8: b+-+ ba is an isomorphism and a is the inclusion. We then have L210 A929, 19 Ag 824, ,94~1, which maps 2 ++ 2 @a++ za for z € L. Since L is torsion-free, this is a monomorphism, and since 1d © 8 is an isomorphism, it follows that Id @ a is a monomorphism. 5. Suppose ¢: A — B is a ring homomorphism. Extension of scalars from A to B is a functor which to each left A-module M assigns a left B-module B @4 M. This functor is exact if and only if B is flat as a right A-module. For many applications this is an important instance of flatness Complexification (Example 14.8) is of course exact. 6. We shall consider one important case where we have flatness in the case of extension of scalars Tet A be an integral domain with field K of fractions. Then K is a flat A-module. For suppose we have an ideal a of A and an element u = >a; @ € a@ K that goes to 0 in K. ANter putting the fractions on a common denominator, and moving the nominators over to the left side of @, we can 40 write u a8 u= (SiH) 2, In Kit maps tothe element and hence u= 0 in a@ K° Extension of scalars form A to K is thus an exact functor, which to each A-module M assigns a vector space M @4 K. We shall determine the kernel of the canonical homomorphism M + M@ K, sends z to z@1. If z is a torsion element of M, with az = 0 for some a # 0, then 2@1=az@1/a=0, so the kernel contains all torsion elements. Suppose conversely that z@1= 0 in M@ K. Since K ia flat, we have 2@ 1 = 0 already in Az @ K, so Az @ K = 0. Let abe the annihilator ideal of z. Then Az & A/a, and by Example 15.2 we have that Az @ K & ‘K/aK, K = aK and hence a4 0. Thus 2 is the torsion submodule T(M). |, which implies that 57 a;4 = 0 torsion element, and we find that the kernel of M—+ M @ K ‘We shall finally show that flatness of a module means that every linear relation between the elements of the module comes from a similar relation between elements of the ring. For the proof of this result ‘we need the following criterion for when an clement of a tensor product is zero: Lemma 17-b Let there be given modules I. and 4M, and a family (y:), of generators for M. Suppose (=i); is a family of elements of L, with only finitely many x # 0. Then Pie 21® ye = 0 in L@q M if and only if there exists a finite family (uj); of elements of L and a family (ass) x1 of elements of A, with only finitely many ass # 0, such that Q) Days = 0 for all j € J, and 2; = 7 wja5s for all i€ 1. te ie Proof. Suppose the relations (2) hold. ‘Then Vaen= Dyas ew = uy oan Suppose on the other hand that 33,21 @ sequence Lye Daw =o. ra Since the sleennnin yy generate M, tere erlas an exact a 0— K—* Ao“ , My —9, with 8: e+ yi for the canonical basis elements ¢;. Since L@- is right exact, we get an exact sequence Lek 82,7640 408 194.0, with (Id @ 8) 5 2 @ ei = 24 @ yi = 0. The exactness of the sequence implies that Vee = (dea) Dy ez; = yj Gals) ta ja jo for some 2; € K. We can write a(z) = Sajiex since (¢:)r is a basis for AU), and Ba = 0 gives 0 = Ba(z;) = A(L; ajies) = Cy ajsw- We also get that Dea = Dy ea(s) = Dy @aner = Dyan Oe. 7 rr iF But 1.@ AW = 10 gives from thin that 2 =D, ujajs. Thus we have both relations in (2). ‘Theorem 17.B A right A-module L is flat if and only if it satisties the following condition: if DM, 21h: = 0 with 2 € L and by € A, then there exist t,...,tm € Land ay € A (G f= 1,-..ym) such that 3, ab; = 0 for all j, and 25 = 5°, wjajs forall i Proof. We first assume that the condition is satisfied, and show that L@ a—+ L is a monomorphism for every ideal a. Now suppose 2 = 52; @ ++ Dai = 0. With the uj and as, that exist by hypothesis, we get F= Dyan Obs = D (uj @ Daub) = oy @ a j 7 7 For the converse we assume that L is flat and that 77, 2b; = 0. Let a be the left ideal generated by b1,.-+:bm. Since L @ @—> L is a monomorphism, we have that )>, 2; ®bj = 0 in L@ a. We can now apply Lemma 17.b to obtain the desired condition. 41 Exercises 17a Show that if Ds @ we Luem holds already in L’@ M. 172 Suppose Lisa submodule ofa flat right A-module M. Show that M/L in flat if and only if Man L = La for every left ideal @ of A 17.3 A short exact sequence in L@ M, then there exists a finitely generated submodule L’ of L such that 6) o—+r—*& +2. y—s0 of left A-modules is said to be pure (and a(L) ia thea called « pure submodule of M) if the induced sequence “ 0 KesL+KOM—+K@N=0 is exact for every right A-module K. Show that: (i) If Lis a pure submodule of M, then L71 aM = al for every right ideal a of A. (Gi) ICN is a flat module, then every short exact sequence (3) is pure. ii) The sequence (3) is pure if and only if (4) is exact for every finitely generated module K. (iv) If A is a PID, then the current notion of purity coincides with the one introduced in Sec. 12, (CE. Exerc. 12.5). 17-4 Consider the polynomial ring K[X,] over a field K. Show that the ideal (X,Y) is not flat as a module over this ring. , 17.5 Show that the following conditions are equivalent for a ring A: (a) For every a € A there exists x € A such that @ = aza, (b) Every principal left ideal of A is a direct summand in A. (©) Every cyclic right A-module is flat. A ring satisfying these condi called von Neumann regular. 18. Exterior powers Exterior algebra is used e.g. for computation with differential forms f(z1,...,2n) zy A--+A din, where the operation A is bilinear and satisfies dz; Adzj = —dz; Adz; In this section we aaiume that A is a commutative ring. Let M and N be A-modules. A mapping 2M" + N is alternating if 1) f is multilinear, ie. is linear in each variable; 2) S(z1,-.-52,) = O if 25 = 2; for some i # j. In the following we let Sq denote the symmetric group of degree n, and for each o € Sq we let p(a) be the parity of o, Alternating mappings are antisymmetric Lemma 18.0 If f: M"—+ N is alternating and ¢ € Sp, then I(2o(1)s-+-)Zo(n)) = P(O)S(21,---2n)- Proof. Since every permutation is a product of transpositions, it suffices to consider a transposition o = (i,j). Singling out the positions é and j, we have WB b Bp es BEE Bee) = fle VIG. Sere e VPA eof Base) EL Go pase sso) Sass Bye Bjpee JEL Gres fy ess Bip). It follows that f(..-5255.-52)5.-.) = —Sle-sjy-o05 2 Conversely, antisymmetric multilinear mappings are alternating if there is no 2-torsion in N, i.e. if 2y=0 > y=0inN. ‘The exterior n-power of M will be defined so that it is universal for alternating mappings from M” i.e. as an abelian group P with an alternating mapping A: M" — P, such that: (+) for any A-module and any alternating mapping f: M" —+ N there exists a unique group homo- morphism g: P+ N with p\= f. As in the ease of tensor products we have that the universal property uniquely determines the exterior power, up to canonical isomorphisms. 42 ‘Theorem 18.A ‘There exists an exterior n-power of M for each n > 0. Proof. Let K be the submodule of @M generated by the elements x; @ ++-@ z_ with 2; = 2) for some i # 5, and consider P = (@M)/K. ‘There are canonical mappings r: M" + OM and x: &M — P, and let A= xr. Then 2 is alternating, because A(21,...,2n) = [#1 @ ---@ zn] = 0 if some a Suppose f: M" —+ N is some alternating mapping. Because of its multilinearity it has a factorization over #: @ MN, and since Y(21 @---@ zn) = J(21,..-)2n) = 0 if some 2 = zj, there is induced e:P—N. Clearly ois alternating, and its uniqueness ia also clear since Im A generates P. ‘The exterior n-power of M is written as AM, and we write M(r1,...,2,) = z1A-+-Azq. The elements f the module AM are called n-vectors, and they are sums of the decomposable n-vectors 21 A-:-A 2. ‘The operation A is multilinear, and 21A---Azq = Oif 2 that 2; for some i # j. The antisymmetry implies Q) Fay A+ Ate(ny = Plo)t1 Av Aza We have in particular that AM = A and that AM = M. ‘Taking the exterior n-th power is a functor, because if p: L — M is A-tinear, then simarly to the case of tensor products one will obtain an induced linear mapping Ap: AL — AM. The exterior algebra or Grassmann algebra of M is the graded A-algebra AM=@im, nen with multiplication AM @ Aat —+"A°M given on the generators by (21 Neen) @ (A Ap) = 1A Aen Ath Ae Ape We write the product of 2 and yin AM as Ay. We have that @) ZAy=(-1)"yAz for 2 €AM, ye AM since the moving of n elements p steps forward will mean np transpositions. Suppose M is finitely generated, with generators z1,...,2,. ‘Then AM is generated by elements of the form zi, A---Azi, with iy < ig < +++ < iy. For k > n there are no such elements, and hence AM for k > n, while AM is a cyclic module (possibly = 0), generated by the element 21 A---A zn From now on we assume that F is a finitely generated free A-module, with a basis 2),.. recall that @F is free module of dimension n*, with basis elements 21, @ xi, ® ---@ 21. Zp. We Theorem 18.B If F is a finitely generated froe A-module on a basis 21,...,2n, then AF is a free module of dimension (}), with a basis consisting of the elements x;, Axi, for which iy < iz < Proof. We first consider the case when k = n. Then the mapping A —+ AF, given by 1++ 21 A-+-A 2p, hhas an inverse which is induced by the mapping @F given by 2j, A---Azj, ++ p(a)- 1, where p(0) is the parity of the permutation ¢ = £2) and p(c) =0 when o is not a permutation. Thus AF is free of dimension 1. Suppose now that k ar. 2000 1200 oo20 00 0 2 relative to eg. a basis (0, 1,0,0), (1,0,1,1), (1, 1,0,0), (0,0,1,0) 14.5 First it is ear that X¥(X @Y -Y @X) =YX@XY—XY @YX =0. Next consider the bilinear mapping Tx I~ A/L& K given by (SayX'¥?, DO byX*V?) ++ aiobor. The induced homomorphism J @ + K sends X @Y r+ 1 and Y@X 0. (One can alternatively use Lemma 17.6). 15.5 Define a: V @x W + Homx(V*,W) as x @y + axy with azy(¢) = y(z)y- This is well-defined as a homomorphism from the tensor product since i K-linear, which gives dee! = Gey + @'yy Oe, ny + yy ad ais,y = ery. In order to show that itis an isomorphism, we choose a basis. (note however that the definition of a was made before the choice of basis). For A: V* — V we define BQ) = 5,21 @ M=2) € V @W,, where (23) is the dual basis for V*. We obtain Ba(e@y) = Blass) =) 21 @azy(23)= ) 21 O2i(zy=2@y since + = 50, 2{(2)zi. So aris the identity mapping on generators for V @ W and therefore on the whole of VG W. Hence a is injective, and since V @x W and Homx(V*,W) have the same dimension V - dim W, it follows that a is an isomorphism. (It is also easy to verify directly that af = Id). 11.5 a)=> bj: For a left ideal Aa, consider z such that a = aza, and put ¢ = xa. Then e? = ¢, and Aa = Ae. Use Example 4.2. b)-> a): If Aa = Ac with ¢? =, then ¢ = za and a = ye, 90a = ye yeaze = ara. 8) 42): Flatness of cyclic modules means that A/6@4 4+ A/b is injective for every right ideal b and left ideal a But A/b@q A & a/ba, so injectivity means that a71b = ba, Suppose (a) holds. Ifa € a6, with a= aza, then a= az-a€ ba Conversely, if (c) holds, then a €@A Aa = aa, 50 (a) follows, 18.6 AM = A, AM = M and AM = 0 for n > 3 since M is generated by two elements. The cave n = 2 requires ‘more work. Now AM = A(A/I @ A/J) is generated by clements (where we write [ ]; and [ ], to denote conets modulo J and J respectively of elements in A) (ols + (8)s) A (Cer + Eells) = (Cals A chr) + (Cals A Lely) + (U8) A Ede) + (B12 ALA2), where the first and the last terms are 0 since A/I and A/J are cyclic modules (note that the exterior products are here taken within the module M). So [als A [a]s) — (Lele A [6]z) = (ad — bo)fl}s A EAs. 45 If we now define AM —+ A/(I-+J) as u++ [ad—e]r4., then this is well-defined because ad — be is bilinear and alternating in (0,8) and (c, 4). The mapping is clearly surjective, and it in injctive since od-—be = 2-4 with 2 € I and y € J implies that w= (ad —bc)[lJ1 A [i]s = (lel A[1],) + ((1}r + [y]) = 0. Hence AM £A@(A/I1@ A/J) @ Al + J). 46 Corrections to “Algebra” Kompendium i Algebra fdk, Stockholm 1908 Some simple misprints without mathematical consequences have been omitted here. 3 Line 18: Change K to R. Line -2: Change “group ring” to “group” 5 Line 11: Change “submodule of N” to “submodule of L” 12 ‘The left commutative diagram: The dashed vertical arrow goes in the wrong direction. 18 Lines -6 to -3: Change to the following: ‘These ate linearly independent, for if 32, x34; generate M*, for if p € M* with g(a 0, then 0 = (32, x}4;)(2i) = a; for each i. They 4 for each i, then for every z= D>, asa, € M we get (2) = S avele) = Yo abi = Yair (aidby = Dx} Cebs, 1¢ 8: Change “b= 0" to “b= = 0". After line 14 one adds the following: ‘The right ideals have similar description, obtained by transposing the matrices. In particular there is only one nonzero proper two-sided ideal, namely (8 Line 8: Change “[y] € M'* to “ys € M". ‘Theorem 9.B: Change the” to “then” Line -3 in proof of Theorem 10.B: omit first (. Exercise 10.2(i): Omit first “left” Line 6 in proof of Theorem 11.4: Change “isotypical” to “isomorphic” Line -3: Change “Exercise 4.3” to “Example 4.2”, Last line in the proof: Change 4.2” to “1.3”. Line -6: Should be: a= u-pf? «---- pip for ... Last line before Lemma 12.b: Change “Exercise 4” to “Bxercise 5” line 2: Change m —1 to "n— 1”. Line 13: Change “pi:2” to pj” Line 18: Change first “A” to “Af Line 2 in Example 2: Change “x,y) € V" to “(z,y) €V”. Line 4 after the matrix (2): Change “K{t]~" to “K[f]—". In the last displayed matrix it should be macle clear that also ay and a, may have repetions in the diagonal, 34 Line 19: Change “p” to 4)" 35 Line 10: Change last. “B” to “M” Example 8: Last formula should read: Cx V = V @V. 36 Line 3 after the commutative diagram: Change “z € M" to “: € M’ ‘Three lines further below: First formula should read: n(x @ [2]) = [x ® 38 Two lines before Example 1: Change “it furthermore” to “@yo An”- 40 Line 9: Change “The elements 2,” to “The elements y,” In the last. commutative diagran toLoF. 42 In the definition of exterior power at the bottom of the page: Change “as an abelian group P” to “as an A-module P”, and change “exists a unique group homomorphism” to “exists a unique linear mapping” 43. Line -8 Change “F" to “AP”, S28 BENS es in the proof there is a “1” missing at the vertical arrow from LH

You might also like