Algebras and Representations MATH 3193 Andrew Hubery [email protected] An algebra is a set A which at the same time has the structure of a ring and a vector space in a compatible way. Thus you can add and multiply elements of A, as well as multiply by a scalar. One example is the set of matrices of size n over a field. A representation is an action of an algebra on a vector space, similar to how matrices of size n act on an n-dimensional vector space. It is often the case thatinformationaboutthealgebracanbededucedfromknowingenoughabout its representations. An analogy might be that one can begin to understand a complicated function by computing its derivatives, or more generally a surface by computing its tangent planes. There are lots of interesting examples of algebras, with applications to mathe- matics and physics. In this course we will introduce some of these algebras, as well as some of the general theory of algebras and their representations. 1 Contents I Algebras 4 1 Quaternions 4 2 Algebras 9 3 Homomorphisms and Subalgebras 16 4 Ideals 21 5 Quotient Algebras 25 6 Presentations of Algebras 28 II Representations 35 7 Representations 35 8 Modules 40 9 Homomorphisms and Submodules 43 10 Quotient Modules 47 11 More on homomorphism spaces 49 III Two Examples 52 12 Modules over the polynomial algebra 52 13 The Weyl Algebra 54 IV Semisimple Algebras 56 14 Semisimple Modules and Schur’s Lemma 56 15 Wedderburn’s Structure Theorem 59 2 16 Maschke’s Theorem 62 17 The Jacobson Radical 65 18 Clifford Algebras 69 19 The Jordan-H¨older Theorem 76 20 Division algebras and Frobenius’s Theorem 80 V Some More Examples 86 21 K[x]/(x2)-modules 86 22 Symmetric Groups (non-examinable) 87 VI Appendix 91 A Review of Some Linear Algebra 91 B Rotations 106 C Presentations of Groups (non-examinable) 111 3 Part I Algebras 1 Quaternions 1.1 Complex Numbers SincetheworkofWessel(1799)andArgand(1806)wethinkofcomplexnumbers as formal expressions z =x+yi with x,y ∈R, which we can add and multiply by expanding out, substituting i2 = −1, and collecting terms. In other words, we have a two-dimensional real vector space C with basis {1,i}, on which we have defined a multiplication C×C→C satisfying the following properties for all a,b,c∈C and λ∈R: Associative a(bc)=(ab)c. Unital there exists 1∈C with 1a=a=a1. Bilinear a(b+λc)=ab+λac and (+λb)c=ac+λbc. Commutative ab=ba. Complex numbers have wonderful properties, for example: • The conjugate of z = x + yi is z¯ = x − yi, and its absolute value is p |z| = x2+y2. Thus |z¯| = |z| and zz¯= |z|2. Also, |zw| = |z||w| for all complex numbers z,w. • Every non-zero complex number z =x+yi has an inverse z−1 =(x−yi)/(x2+y2)=z¯/|z|2, so they form a field. • Rotations of the plane correspond to multiplication by complex numbers of absolute value 1. 1.2 Quaternions Trying to find a way to represent rotations in three dimensions, Sir William RowanHamiltoninventedthequaternionsin1843. Heneededfourrealnumbers, not three, and also had to drop commutativity. Quaternions are expressions a+bi+cj+dk with a,b,c,d ∈ R. They add and subtract in the obvious way, and multiply by first expanding out (being careful 4 with the ordering), making the following substitutions i2 =−1 ij =k ik =−j ji=−k j2 =−1 jk =i ki=j kj =−i k2 =−1 and then collecting terms. For example (2+3i)(i−4j)=2i−8j+3i2−12ij =−3+2i−8j−12k. The set of all quaternions is denoted H. In other words we have a four-dimensional real vector space H with basis {1,i,j,k}, on which we have defined a multiplication H×H→H satisfying the following properties for all a,b,c∈H and λ∈R: Associative a(bc)=(ab)c. Unital there exists 1∈H with 1a=a=a1. Bilinear a(b+λc)=ab+λac and (a+λb)c=ac+λbc. Remark. Multiplication of quaternions is not commutative: ij =k but ji=−k, so ij 6=ji. 1.3 Basic Definitions Let q =a+bi+cj+dk be a quaternion. We say that the real part of q is a, and the imaginary part is bi+cj+dk. A pure quaternion is one whose real part is zero. The conjugate of q is q¯=a−bi−cj−dk. Thus if a is a real number and p is a pure quaternion, then the conjugate of a+p is a−p. √ The absolute value of q is |q|= a2+b2+c2+d2. Note that |q¯|=|q|. 1.4 Elementary properties 1. If p is a pure quaternion then p2 =−|p|2. For, (bi+cj+dk)(bi+cj+dk) =b2i2+c2j2+d2k2+bc(ij+ji)+bd(ik+ki)+cd(jk+kj) =−(b2+c2+d2). 2. Conjugation is a linear map, so p+λq =p¯+λq¯ for all p,q ∈H and λ∈R. 5 3. Conjugation satisfies q¯¯=q and pq =q¯p¯ for all p,q ∈H. The first property is clear, so we just need to prove the second. Since multiplication is bilinear and conjugation is linear, we can reduce to the case when p,q ∈ {1,i,j,k}. If p = 1 or q = 1, then the result is clear, so we may assume that p,q ∈{i,j,k}. If p = q, then p2 = −1 = p¯2. Otherwise p and q are distinct elements of {i,j,k}. Letr bethethirdelement. Thenp¯=−p,andsimilarlyforq and r. Using the multiplication rules we see that pq =±r and qp=∓r, so q¯p¯=(−q)(−p)=qp=∓r =±r¯=pq as required. 4. We have qq¯=q¯q =|q|2. For,writeq =a+pwitha∈Randpapurequaternion,sothatq¯=a−p. Then ap=pa, so qq¯=(a+p)(a−p)=a2+pa−ap−p2 =a2−p2 =(a−p)(a+p)=q¯q. Using p2 =−|p|2 we get qq¯=a2−p2 =a2+|p|2 =|q|2. 5. For any two quaternions p and q we have |pq|=|p||q|. For |pq|2 =pqpq =pqq¯p¯=p|q|2p¯=|q|2pp¯=|q|2|p|2. 1.5 Lemma Any non-zero quaternion has a multiplicative inverse; that is, H is a division algebra. Proof. The inverse of q is q−1 =q¯/|q|2. 1.6 Lemma Every quaternion can be written in the form q =r(cid:0)cos(1θ)+sin(1θ)n(cid:1) 2 2 where r,θ ∈ R with r = |q| ≥ 0 and θ ∈ [0,2π], and n is a pure quaternion of absolute value 1. The use of 1θ is traditional; the reason will become clear later. 2 6 Proof. If q = 0, then this is clear; otherwise q/|q| is a quaternion of absolute value 1. Now let q have absolute value 1 and write q = a+p with a ∈ R and p a pure quaternion. Then 1=|q|2 =a2+|p|2, so we can write a = cos(1θ) for some unique θ ∈ [0,2π], whence |p| = sin(1θ). 2 2 Finally, if θ = 0,2π, then p = 0 so we can take n to be arbitrary; otherwise n=p/|p| is a pure quaternion of absolute value 1. 1.7 Lemma We may identify the set of pure quaternions P = {bi+cj +dk : b,c,d ∈ R} with R3 such that i,j,k correspond respectively to the standard basis vectors e ,e ,e . We equip R3 with the usual dot product and cross product. Then 1 2 3 pq =−p·q+p×q ∈H for all p,q ∈P. Note that the dot product of two elements of P is in R, and the cross product is in P, so the sum makes sense in H. Proof. Each operation is bilinear, so it suffices to check this for p,q ∈ {i,j,k}. 1.8 Theorem If q is a quaternion of absolute value 1, then the linear transformation R : P →P, R (p):=qpq−1 q q is a rotation. Explicitly, if q = cos(1θ) + sin(1θ)n, then R = R is the 2 2 q n,θ rotation about axis n through angle θ. Two quaternions q,q0 of absolute value 1 give the same rotation if and only if q0 =±q. Proof. Recall that an ordered basis (f ,f ,f ) of R3 is called a right-handed 1 2 3 orthonormal basis provided that f ·f =δ and f =f ×f . i j ij 3 1 2 Also, if n∈R3 has length 1, then the rotation about axis n through angle θ is the linear map n7→n, u7→cos(θ)u+sin(θ)v, v 7→−sin(θ)u+cos(θ)v where (n,u,v) is any right-handed orthonormal basis. Now let q be a quaternion of absolute value 1. By Lemma 1.6 we can write q = cos(1θ)+sin(1θ)n with θ ∈ [0,2π] and n a pure quaternion of absolute 2 2 7 value 1. Let (n,u,v) be a right-handed orthonormal basis for P. The previous lemma tells us that nu=−n·u+n×u=v and un=−u·n+u×n=−n×u=−v. Similarly uv =n=−vu and vn=u=−nv. For simplicity set c := cos(1θ) and s := sin(1θ). Now, since n2 = −|n|2 = −1 2 2 we have qnq−1 =(c+sn)n(c−sn)=(c+sn)(cn−sn2)=(c+sn)(cn+s) =c2n+cs+csn2+s2n=(cs−cs)+(c2+s2)n=n. Also quq−1 =(c+sn)u(c−sn)=(c+sn)(cu−sun)=(c+sn)(cu+sv) =c2u+csv+csnu+s2nv =(c2−s2)u+2csv =cos(θ)u+sin(θ)v and hence qvq−1 =qnuq−1 =qnq−1quq−1 =cos(θ)nu+sin(θ)nv =−sin(θ)u+cos(θ)v. This is the rotation claimed. Any particular rotation occurs in exactly two ways, as the rotation about axis n through angle θ, and as the rotation about −n through angle 2π −θ. The latter corresponds to the quaternion cos(π− 1θ)+sin(π− 1θ)(−n)=−cos(1θ)−sin(1θ)n=−q. 2 2 2 2 1.9 Remarks (non-examinable) The set of quaternions of absolute value 1 form a group under multiplication, denotedSp(1),anditisnothardtoseethatthemapq 7→R fromtheprevious q theorem defines a surjective group homomorphism R: Sp(1)→SO(3,R) to the groupofrotationsofR3. Thisgrouphomomorphismisadouble cover, meaning that there are precisely two elements of Sp(1) mapping to each rotation in SO(3,R). In fact, we can say more. A quaternion q =a+bi+cj+dk has absolute value 1 precisely when a2+b2+c2+d2 =1, so Sp(1) can be thought of as a 3-sphere S3 ={(a,b,c,d)∈R4 :a2+b2+c2+d2 =1}. SimilarlySO(3,R)⊂M (R)∼=R9. Thereforebothofthesesetshaveaninduced 3 topology on them, and both the multiplication and inversion maps are contin- uous, so they are topological groups. In this set-up the group homomorphism Sp(1)→SO(3,R) is also continuous. Let us fix a pure quaternion n of absolute value 1. Then, as θ increases from 0 to 2π, we get a sequence of rotations starting and ending at the identity. The sequenceofquaternions,however,startsat1butendsat−1. Wethereforeonly 8 get 4π-periodicity for the quaternions. This is relevant in quantum mechanics for ‘spin 1/2’ particles like electrons. We can visualise this by rotating a book held in a hand: a 2π rotation returns the book to its original position, but a 4π rotation is needed to return both the book and the hand to their original positions. You can read about the ‘quaternion machine’ in J. Conway and R. Guy, The book of numbers. 2 Algebras 2.1 Definition Fix a base field K, for example R or C. An algebra over K, or K-algebra, consists of a K-vector space A together with a multiplication A×A→A, (a,b)7→ab, satisfying the following properties for all a,b,c∈A and λ∈K: Associative a(bc)=(ab)c. Unital there exists 1∈A such that 1a=a=a1. Bilinear a(b+λc)=ab+λ(ac) and (a+λb)c=ac+λ(bc). Aside. If you have seen the definition of a ring, then you will see that the distributivity axiom has been replaced by the stronger bilinearity axiom. We can do this since our algebra is a priori a vector space. AnalternativedescriptionwouldthereforebethatAisbothavectorspaceand a ring, and that these structures are compatible in the sense that scalars can always be brought to the front, so a(λb)=λ(ab). 2.2 Remarks 1. In the literature, the algebras we consider might be called unital, as- sociative algebras. There are other types: Banach algebras are usually non-unital; Lie algebras and Jordan algberas are non-associative. 2. Recall that a vector space V is finite dimensional if it has a finite basis. Not all of our algebras will be finite dimensional. 3. There is a very rich theory of commutative algebras, where one assumes thatthemultiplicationiscommutative, soab=baforalla,b∈A. Thisis relatedto,amongstotherthings,algebraicgeometryandalgebraicnumber theory. Inthiscoursewewillbeconcernedwithgeneral,non-commutative, algebras. 9 2.3 Examples 1. K is a 1-dimensional algebra over itself. 2. C is a 2-dimensional R-algebra with basis {1,i} as a vector space over R. (It is also a 1-dimensional C-algebra, as in Example 1.) 3. H is a 4-dimensional R-algebra with basis {1,i,j,k}. Even though it containsacopyofC,forexamplewithbasis{1,i},itcannotbeconsidered as a C-algebra since i does not commute with j and k. 2.4 Division algebras A division algebra is a non-zero algebra A in which every non-zero element has a multiplicative inverse; that is, for all a 6= 0 there exists a−1 such that aa−1 =1=a−1a. 2.5 Lemma A division algebra has no zero divisors: ab=0 implies a=0 or b=0. Proof. If ab=0 and a6=0, then 0=a−1(ab)=(a−1a)b=1b=b, so b=0. 2.6 Example A field is the same as a commutative division algebra. The quaternions form a division algebra, but are not commutative so do not form a field. 2.7 Characteristic The characteristic of a field K is the smallest positive integer n such that n·1=0∈K. If no such n exists, we define the characteristic to be zero. Examples. 1. The fields Q, R and C all have characteristic zero. 2. If p∈Z is a prime, then F =Z is a field of characteristic p. p p 3. The characteristic of a field is either zero or a prime number. For, if char(K)=n and n=ab with 1<a,b<n, then ab=n=0∈K, whence either a=0∈K or b=0∈K, contradicting the minimality of n. 2.8 Algebras given by a basis and structure coefficients Oneimmediatequestionwhicharisesishowtodescribeanalgebra. Weshallgive two answers to this question, both having their advantages and disadvantages. 10