REPRESENTATIONS OF INFINITE DIMENSIONAL LIE ALGEBRAS AND DIRAC’S POSITRON THEORY
Jeremy S. Gresham
A Thesis Submitted to the University of North Carolina Wilmington in Partial Fulfillment Of the Requirements for the Degree of Master of Science
Department of Mathematics and Statistics
University of North Carolina Wilmington
2011
Approved by
Advisory Committee
Mark Lammers Gabriel Lugo
Dijana Jakeli´c
Chair
Accepted by
Dean, Graduate School TABLE OF CONTENTS
ABSTRACT ...... iv LIST OF SYMBOLS ...... v 1 INTRODUCTION ...... 1 2 Preliminaries ...... 3 3 Algebras ...... 8 3.1 Witt Algebra ...... 8 3.2 Virasoro Algebra ...... 9 3.3 Oscillator Algebra ...... 15 3.4 Algebras of Infinite Matrices ...... 16 3.5 Loop Algebras ...... 20 4 Representations of the Virasoro Algebra ...... 25 4.1 Hermitian Forms ...... 25 4.2 Highest Weight Representations of Vir ...... 26 4.3 Irreducible Positive Energy Representations ...... 30 5 Oscillator Representations ...... 39 5.1 Representations of the Oscillator Algebra ...... 39 5.2 Oscillator Representations of Vir ...... 45 6 Dirac Positron Theory ...... 53 6.1 Infinite Wedge Space ...... 53
6.2 Highest Weight Representations of gl∞ ...... 56
6.3 Representations ofa ¯∞ ...... 61 7 Some Physics ...... 64 7.1 Dirac Equation and First Quantization ...... 64 7.2 Second Quantization and Fock space ...... 74 APPENDIX ...... 76
ii A Triangular Decomposition ...... 76 B Tensor Products and Related Algebras ...... 76 C Geometry ...... 78 REFERENCES ...... 81
iii ABSTRACT
Representation theory and physics interact in complex and often unexpected ways, with one discipline building upon the work of the other. We present a number of rep- resentations for the Witt, Virasoro, and Heisenberg algebras building up to the Dirac theory for relativistic electrons, once from the representation theory perspective and then, less rigorously, from a physics perspective.
iv LIST OF SYMBOLS
• · | · - a symmetric bilinear form
• [·, ·] - a Lie bracket, usually the matrix commutator
• V , W - a vector space
• x, y, z, v, ψ - vectors
• gl∞,a ¯∞, g, X, d, etc - Lie algebras
• Vir - the Virasoro Algebra
•A - the oscillator (Heisenberg) algebra
• G, GLn, GL∞ - Lie groups
• Z - the integers
• i,j,k,l,m,n - integers
• R - the reals
• ~, , q - real numbers
• C - the complex numbers
• λ, µ, α, β - complex numbers
• λ¯ - the complex conjugate of λ
• B = C[x1, x2, ...] - the space of polynomials over C in infinitely many variables
• C[t, t−1] - the space of Laurent polynomials, polynomials in t and t−1
• ω - an antilinear anti-involution
v • π, φ, r,r ˆ, τ - algebra homomorphisms
• L, ⊕ - direct sum
• P - sum
• ⊗ - the tensor product
• ∧ - the antisymmetric (or wedge) product
• : aiaj : - the normal ordering of ai, aj (Section 4.2)
• I - identity matrix
• tr(·) - the trace
• exp(·) - the exponential map
• δi,j = δij - the Kronecker delta
• M - a manifold
• C∞(M) - smooth functions on a manifold M
• TxM - tangent space at a point x ∈ M
• X(M) - algebra of vector fields on M
• F - space of semi-infinite monomials
• Res0[·] - the residue at zero of some Laurent polynomial
• A† is the Hermitian conjugate of A
This list is not all-inclusive, but covers most of the symbols in their most common usage.
vi 1 INTRODUCTION
Representation theory has very strong connections to physics. In particular, infi- nite dimensional Lie algebras are important for conformal field theory and exactly solvable models. The Witt, Virasoro, and Heisenberg algebras all have interesting representations which can be used to describe Dirac’s positron theory. Dirac’s the- ory is an attempt to combine quantum mechanics and relativity. It contained the first prediction of antimatter, in the form of positrons. In this thesis we will describe some of these representations along with Dirac’s original theory with the aim of giving a taste of the interplay of representation theory and physics. We will also give some representations of loop algebras in the same context. The physics included will be presented in a more informal way than the mathe- matics – rather like a summary of relevant ideas. We point out that the Heisenberg or oscillator algebra defined here is an infinite dimensional algebra, and not directly related to the usual finite dimensional Heisenberg group. The Virasoro algebra, the unique central extension of the Witt algebra, is used for vertex operator algebras and has applications in conformal fields and string theory. Our main focus, however, is Dirac’s positron theory. This is an interesting model, but still simple enough to see some of the real interplay between physics and representation theory. Sections 3, 4, 5, 6 and 8 are devoted to representation theory and Dirac’s positron theory. These sections were based on material in Kac and Raina’s book “Bombay Lectures on Highest Weight Representations of Infinite Dimensional Lie Algebras” ([4]). Subsection 4.3 which covers positive energy and highest weight representa- tions of the Virasoro algebra follows also “Generalized Derivations on Algebras and Highest Weight Representations of the Virasoro Algebra” ([2]) by Jonas Hartwig. Section 7 addresses Dirac’s theory from a physics perspective, beginning with the Schr¨odingerequation and the relativistic energy equation. It draws material from Bernd Thaller’s “The Dirac Equation” ([6]), a physics text.
2 2 Preliminaries
Definition 1 A Lie algebra is a vector space g over a field K equipped with a multiplication [·, ·]: g × g → g having the following properties:
a) [·, ·] is K-bilinear,
b) [x, y] = −[y, x] for all x, y ∈ g
c) Jacobi identity: [[y, z], x] + [[z, x], y] + [[x, y], z] = 0, for all x, y ∈ g.
In this thesis we will mostly have K = C. Example 1 An example of a Lie algebra is the set of all linear transformations of a vector space V , denoted gl(V ), where [x, y] = xy − yx for x, y ∈ gl(V ). This bracket is called the commutator bracket. A common way of defining an algebra is by specifying a set of generators and relations among those generators. For Lie algebras this is done by giving a basis and specifying the Lie bracket on those elements.
Example 2 The Lie algebra sl2(C) is a vector space over C with basis {x, h, y} and Lie bracket defined by the relations
[x, y] = h
[h, x] = 2x
[h, y] = −2y.
This can be realized as a matrix algebra, taking
0 1 0 0 1 0 x = , y = , h = . 0 0 1 0 0 −1
3 Definition 2 A subalgebra of a Lie algebra g is a subspace e of g that is closed under the bracket, i.e., [x, y] ∈ e, ∀x, y ∈ e. A subalgebra e is abelian provided [x, y] = 0, ∀x, y ∈ e.
Definition 3 An ideal of a Lie algebra g is a subspace I such that for any x ∈ I and any y ∈ g we have [x, y] ∈ I.
The algebra sl2(C) defined above has no nontrivial (i.e. nonzero) proper ideals. sl2(C) can be seen as a proper ideal of gl2(C), the four dimensional Lie algebra of 2x2 matrices with the commutator bracket. Given a Lie algebra g and an ideal I of g, consider the quotient vector space g0 = g/I equipped with bracket [x + I, y + I] = [x, y] + I for all x, y ∈ g. Then g0 is a Lie algebra called the quotient algebra of g by I. Moreover, there is a surjective Lie algebra homomorphism φ : g → g0 such that kerφ = I. A triangular decomposition of a Lie algebra g consists of an abelian subal- gebra h 6= 0 and two subalgebras n+ and n− such that g = n− ⊕ h ⊕ n+ with a few conditions imposed on these subalgebras that we will not elaborate on. For a detailed definition, refer to Appendix A.
Definition 4 Given a Lie algebra g, the universal enveloping algebra of g is a pair (U(g), i), where U(g) is an associative algebra with unit (denoted by 1) over K, i : g → U(g) is a linear map satisfying
i([x, y]) = i(x)i(y) − i(y)i(x), ∀x, y ∈ g (1) and the following holds: for any associative algebra L over K with unit 1 and any linear map j : g → L satisfying (1) there exists a unique algebra homomorphism φ : U(g) → L with φ(1) = 1 such that φ ◦ i = j.
4 It can be proved that the map i is injective. We also state but do not prove the Poincar´e-Birkhoff-Witttheorem:
Theorem 1 Let {x1, x2, ···} be any ordered basis of g. Then the elements xi(1) ··· xi(m), m > 0, i(1) ≤ i(2) ≤ · · · ≤ i(m) along with 1 form a basis of U(g).
For an explicit construction of the universal enveloping algebra, please refer to Appendix B. A simple example of a universal enveloping algebra is the one associated with
j k l sl2(C). This is an associative algebra with a basis of elements x h y where j, k, l ∈ N.
Definition 5 A Lie algebra homomorphism is a map of Lie algebras φ : g1 → g2 such that φ is linear and
φ([x, y]) = [φ(x), φ(y)] ∀x, y ∈ g1.
Furthermore, a Lie algebra isomorphism is a homomorphism that is both injective and surjective.
Definition 6 A representation of a Lie algebra g is a Lie algebra homomorphism
π : g → gl(V ) where V is a vector space over K and gl(V ) the algebra of endomor- phisms of V .
Definition 7 A module for a given Lie algebra g is a vector space V over K along with an operation g × V → V sending (x, v) to x · v, usually denoted as xv. This operation must satisfy the following conditions:
(ax + by)v = a(xv) + b(yv)
x(av + bw) = a(xv) + b(xw)
[x, y]v = xyv − yxv
5 for all x, y ∈ g, v, w ∈ V and a, b ∈ K.
Note that if φ : g → gl(V ) is a representation of g, then V is a module for g by the action xv = φ(x)(v) and vice versa. We will use both terminologies interchangeably.
Definition 8 Two representations φ : g → gl(V ) and ψ : g → gl(W ) of a Lie algebra g are homomorphic if there exists a linear map T : V → W such that for all x ∈ g and v ∈ V we have T (φ(x)(v)) = ψ(x)(T (v)). If T is also invertible, the two representations are said to be isomorphic.
Definition 9 Let h be an abelian subalgebra of g, h∗ be its dual space, and let V be
∗ a representation of g. For λ ∈ h , the λ-weight space Vλ of V (relative to h) is defined as
Vλ = {v ∈ V |h · v = λ(h)v, ∀h ∈ h}.
If Vλ 6= 0, we say λ is a weight of V . Moreover, V is called a weight module if L V = λ∈h∗ Vλ.
Definition 10 A highest weight vector with highest weight λ (relative to h) of a g-module V is a nonzero vector v ∈ V such that
xv = 0, ∀x ∈ n+
hv = λ(h)v, ∀h ∈ h
where λ ∈ h∗. Furthermore, V is called a highest weight module (or representa- tion) provided V = U(g)v for some highest weight vector v ∈ V .
Definition 11 A Verma module L is a highest weight module for g with highest weight λ ∈ h∗ and highest weight vector v ∈ L with a universal property: for any
6 highest weight representation V of g with highest weight λ and highest weight vector w there exists a unique surjective homomorphism φ : L → V of g-modules which maps v to w.
7 3 Algebras
3.1 Witt Algebra
The Witt algebra is the complexification of the Lie algebra of (real) vector fields on the unit circle. More precisely, consider the unit circle in the complex plane S1 := {eiθ|θ ∈ [0, 2π]}. Obviously, S1 is a group under multiplication and it is a 1-manifold, so clearly S1 is a Lie group. For the definitions of a Lie group and
1 d a real vector field, please see Appendix C. Now consider X(S ) := {f(θ) dθ |f ∈ C∞(S1,S1), f(θ + 2π) = f(θ)}, the vector space of real vector fields on S1, where C∞(S1,S1) denotes the smooth (infinitely differentiable) functions from S1 to S1. The Lie bracket on X(S1) is given by:
d d d [f(θ) , g(θ) ] = (fg0 − f 0g)(θ) (2) dθ dθ dθ making X(S1) into a Lie algebra. Using Fourier series, a basis of the 2π - periodic
C∞(S1,S1)-functions is : {1, cos(nθ), sin(nθ)|n ∈ Z}. This in turn gives a basis for
1 d d d X(S ) as a vector space over R : { dθ , cos(nθ) dθ , sin(nθ) dθ |n ∈ Z}. The Witt algebra d is the complexification of X(S1), that is, the span of the above basis over C. Using Euler’s formula ( eiθ = cos(θ) + i sin(θ) ), a new basis
n+1 d iθ can be written as {dn|n ∈ Z} where dn = −z dz and z = e . This can be seen through a simple calculation:
dz = iz dθ d d i = −z . dθ dz
inθ d d d i d 1 Thus dn = ie dθ , while dθ = −id0, cos(nθ) dθ = − 2 (dn +d−n), sin(nθ) dθ = − 2 (dn − d−n). Taking the span of these vectors over C returns the algebra d. These basis
8 vectors satisfy the following commutation relations:
d d [d , d ] = [−iei(m+1)θ , −iei(n+1)θ ] m n dθ dθ d d = i3(n + 1)ei(m+n+1)θ − i3(m + 1)ei(m+n+1)θ dθ dθ d = (m − n) − iei(m+n+1)θ dθ
= (m − n)dm+n.
Definition 12 An antilinear involution on a Lie algebra g is a map ω : g → g such that
(a) ω(ω(x)) = x
(b) ω(λx) = λω¯ (x)
(c) ω([x, y]) = [ω(y), ω(x)]
∀x, y ∈ g and ∀λ ∈ C where λ¯ is the complex conjugate of λ.
An antilinear involution ω can be defined on d by setting
ω(dn) = d−n, n ∈ Z.
An easy calculation shows that the set of real elements of d, i.e., d ∩ X(S1), consists of the elements fixed under −ω.
3.2 Virasoro Algebra
Definition 13 Two elements x and y of a Lie algebra g are said to commute provided [x, y] = 0. A subalgebra e of g is called central if every element of e commutes with all elements of g.
Note that any central subalgebra is actually an ideal.
9 Definition 14 A central extension of a Lie algebra g is a Lie algebra A with a subalgebra e such that e is central and the quotient of A by e is g. Using short exact sequences, we have
0 → e → A → g → 0.
Theorem 2 The Witt algebra d has a unique nontrivial one-dimensional central extension d˜ = d ⊕ Cc¯ up to Lie algebra isomorphism. This extension has a basis
{c} ∪ {dn|n ∈ Z} where c ∈ Cc¯, such that the following relations are satisfied:
[c, dn] = 0 for n ∈ Z (3) m3 − m [d , d ] = (m − n)d + δ c for m, n ∈ . (4) m n m+n m,−n 12 Z
The extension d˜ is called the Virasoro algebra, and is denoted by Vir. Proof. To prove existence, it is enough to check that the relations (3)-(4) define a Lie algebra, which is easy. We give a proof of uniqueness. Suppose d˜ = d ⊕ Cc¯ is a ¯ nontrivial one-dimensional central extension of d. Let {dn|n ∈ Z} denote the basis elements of d from the last section, then we have
¯ ¯ ¯ [dm, dn] = (m − n)dm+n + a(m, n)¯c ¯ [¯c, dn] = 0 for m, n ∈ Z, where a : Z × Z → C is some function. Note that a(m, n) = −a(n, m) because d˜ is a Lie algebra and thus has anti-symmetric product:
¯ ¯ ¯ ¯ ¯ 0 = [dm, dn] + [dn, dm] = (m − n + n − m)dm+n + (a(m, n) + a(n, m))¯c.
10 Define new elements:
¯ d0 if n = 0 0 dn = ¯ 1 dn − n a(0, n)¯c if n 6= 0
c0 =c. ¯
0 0 ˜ Then {c } ∪ {dn|n ∈ Z} is a new basis for d. The new commutation relations are:
0 0 [c , dn] = 0
0 0 ¯ ¯ [dm, dn] = [dm, dn] ¯ = (m − n)dm+n + a(m, n)¯c
0 0 0 = (m − n)dm+n + a (m, n)c (5) for m, n ∈ Z where a0 : Z × Z → C is defined by
a(m, n) if m + n = 0 a0(m, n) = (6) m−n a(m, n) + m+n a(0, m + n) if m + n 6= 0.
Since a is antisymmetric, a0 is as well, and therefore a0(0, 0) = 0. From (6) it follows that a0(0, n) = 0 for any nonzero n. These facts together with (5) show that
0 0 0 [d0, dn] = −ndn.
11 Now, using the Jacobi identity in d˜, we have:
0 0 0 0 0 0 0 0 0 [[d0, dn], dm] + [[dn, dm], d0] + [[dm, d0], dn] = 0
0 0 0 0 0 0 0 0 [−ndn, dm] + [(n − m)dn+m + a (n, m)c , d0] − [dn, mdm] = 0
0 0 0 0 −(n + m)(n − m)dn+m − (n + m)a (n, m)c + (n − m)(n + m)dn+m = 0
(n + m)a0(n, m)c0 = 0 which shows that a0(n, m) = 0 unless n + m = 0 and n 6= 0, m 6= 0. Thus, setting b(m) = a0(m, −m), (5) can be rewritten as
0 0 [c , dn] = 0
0 0 0 0 [dm, dn] = (m − n)dm+n + δm+n,0b(m)c with b(0) = 0. Again using the Jacobi identity
0 0 0 0 0 0 0 0 0 [[dn, d1], d−n−1] + [[d1, d−n−1], dn] + [[d−n−1, dn], d1] = 0
0 0 0 0 0 0 [(n − 1)dn+1, d−n−1] + [(n + 2)d−n, dn] + [(−2n − 1)d−1, d1] = 0
0 0 0 0 (n − 1)(2(n + 1)d0 + b(n + 1)c ) + (n + 2)(−2nd0 + b(−n)c )
0 0 + (−2n − 1)(−2d0 + b(−1)c ) = 0
2 2 0 (2n − 2 − 2n − 4n + 4n + 2)d0 + {(n − 1)b(n + 1) − (n + 2)b(n)
+ (2n + 1)b(1)}c0 = 0, which is equivalent to
(n − 1)b(n + 1) = (n + 2)b(n) − (2n + 1)b(1). (7)
Next, b(m) = m and b(m) = m3 are shown to be two solutions of (7). First we show
12 b(m) = m is consistent with the (7):
(n − 1)(n + 1) = (n + 2)(n) − (2n + 1)(1)
n2 − 1 = n2 + 2n − 2n − 1
0 = 0 and now for b(m) = m3:
(n − 1)(n + 1)3 = (n + 2)(n)3 − (2n + 1)(1)
(n − 1)(n3 + 3n2 + 3n + 1) = n4 + 2n3 − 2n − 1
n4 + 2n3 − 2n − 1 = n4 + 2n3 − 2n − 1
0 = 0.
Since (7) is a second order linear recurrence relation in b and the above two solutions are linearly independent, then there are α, β ∈ C such that
b(m) = αm3 + βm.
Finally, set
α + β d = d0 + δ c0, n n n,0 2 and
c = 12αc0.
13 If α 6= 0, this is again a change of basis. Then,
0 0 [dm, dn] = [dm, dn]
0 3 0 = (m − n)dm+n + δm+n,0(αm + βm)c α + β = (m − n)d − (m − n)δ c0 + δ (αm3 + βm)c0 m+n m+n,0 2 m+n,0 α + β = (m − n)d − δ 2m c0 + δ (αm3 + βm)c0 m+n m+n,0 2 m+n,0 3 0 = (m − n)dm+n + δm+n,0(αm − αm)c m3 − m = (m − n)d + δ c. m+n m+n,0 12
The proof of uniqueness is finished. Notice also that α = 0 corresponds to the trivial extension. An antilinear involution ω can be defined similarly to the one defined for d, by requiring
ω(dn) = d−n ∀n ∈ Z
ω(c) = c.
We only need to check the following:
n3 − n [ω(d ), ω(d )] = [d , d ] = (−n + m)d − δ c n m −n −m −n−m −n,m 12 m3 − m = (m − n)ω(d ) + δ ω(c) m+n m,−n 12
= ω([dm, dn]).
Vir has the following decomposition into Lie subalgebras, called a triangular
14 decomposition of Vir:
V ir = n−⊕h ⊕ n+ where ∞ ∞ − M + M n = Cd−i, h = Cc ⊕ Cd0, n = Cdi. i=1 i=1
3.3 Oscillator Algebra
The oscillator (Heisenberg) algebra A is defined as the complex Lie algebra with generators ~ and an (n ∈ Z), and relations:
[~, an] = 0, ∀n ∈ Z
[am, an] = δm,−nm~, ∀m, n ∈ Z.
It is easy to see that A is well-defined. Note that [a0, an] = 0 for all n ∈ Z, so that a0 is central. An antilinear involution ω can be defined for A as follows:
ω(an) = a−n
ω(~) = ~.
The oscillator algebra has a triangular decomposition similar to the Virasoro algebra:
A = n−⊕h ⊕ n+ where ∞ ∞ − M + M n = Ca−i, h = C~ ⊕ Ca0, n = Cai. i=1 i=1
15 3.4 Algebras of Infinite Matrices
Let
M V = Cvj j∈Z be an infinite dimensional complex vector space with fixed basis {vj|j ∈ Z}. Identify
th vj with the column vector whose i entry is δij. Any vector in V is a finite sum of multiples of vj, j ∈ Z, and thus it has only a finite number of nonzero coordinates. This identifies V with C∞. Now we define two Lie algebras:
Definition 15 Let gl∞ be the vector space of matrices defined by:
gl∞ = {(aij)i,j∈Z|aij = 0 for all but finitely many i, j ∈ Z}
and let a¯∞ be the vector space defined by:
a¯∞ = {(aij)i,j∈Z|aij = 0 for |i − j| 0}.
Clearlya ¯∞ is a set of matrices with a finite number of nonzero diagonals and gl∞ ⊆ a¯∞. Note that the usual matrix multiplication is well-defined ina ¯∞: Let x, y ∈ a¯∞. Then xkl = 0 for |k − l| > M for some M ≥ 0, and ymn = 0 for
|m − n| ≥ N for some N > 0. Then for i, j ∈ Z,
X X (xy)ij = xikykj = xikykj k∈Z k∈Z |i−k|≤M |k−j|≤N
16 which is clearly a finite sum. Furthermore,
|i − j| = |i − k + k − j| ≤ |i − k| + |k − j| ≤ M + N
by the triangle inequality. This means that (xy)ij = 0 for |i − j| > M + N, thus the bracket [x, y] = xy − yx is an element ofa ¯∞, since each product is well-defined and each summand has finitely many nonzero diagonals.
Proposition 1 The vector spaces gl∞ and a¯∞ are Lie algebras, with gl∞ a subalge- bra of a¯∞.
Proof. This is shown by checking thata ¯∞ is an associative algebra, and thus a
Lie algebra under the commutator bracket, and by checking that gl∞ is closed under the commutator bracket. However, for illustration purposes, we demonstrate that gl∞ is a Lie algebra.
Let Eij be the matrix with 1 as the (i, j) entry and zeros elsewhere. The Eij obviously form a basis for gl∞. Clearly
Eijvk = δjkvi and
EijEmn = δjmEin.
Then the commutation relations (using the usual matrix commutator) are:
[Eij,Emn] = δjmEin − δniEmj.
17 This is clearly antisymmetric, and we can check the Jacobi identity. First, we have
[[Eij,Emn],Ekl] = δjm[Ein,Ekl] − δni[Emj,Ekl]
= δjm{δnkEil − δliEkn} − δni{δjkEml − δlmEkj}
= δjmδnkEil − δjmδliEkn − δniδjkEml + δniδlmEkj and then
[[Eij,Emn]Ekl] + [[Emn,Ekl],Eij] + [[Ekl,Eij],Emn] =
δjmδnkEil − δjmδliEkn − δniδjkEml + δniδlmEkj
+δnkδliEmj − δnkδjmEil − δlmδniEkj + δlmδjkEin
+δliδjmEkn − δliδnkEmj − δjkδlmEin + δjkδniEml
= 0.
Thus gl∞ is a Lie algebra.
Define the shift operators Λk by
Λkvj = vj−k.
Then we have
X Λk = Ei,i+k. i∈Z
Thus Λk is the matrix with 1 at each entry of the k-th diagonal and 0 elsewhere, so
Λk ∈ a¯∞. Now,
X ΛkΛj = Ei,i+k+j = ΛjΛk i∈Z
18 and so [Λj, Λk] = 0 for j, k ∈ Z, i.e., the Λk form an abelian subalgebra ofa ¯∞. For every pair of nonzero constants α and β there is an inclusion of the Witt algebra d ina ¯∞ as a subalgebra via
dn = (k − α − β(n + 1))Λn or, acting on Cn,
dn(vk) = (k − α − β(n + 1))vk−n.
The elements dn are clearly ina ¯∞. We show that these elements generate a subal- gebra ina ¯∞ isomorphic to d with the following calculation:
[dm, dn](vk) = dmdn(vk) − dndm(vk)
= dm(k − α − β(n + 1))vk−n − dn(k − α − β(m + 1))vk−m
= (k − α − β(n + 1))((k − n) − α − β((k − n) + 1))vk−n−m
− (k − α − β(m + 1))((k − m) − α − β((k − m) + 1))vk−m−n
= [−kn − βk(m + 1) + αn − β(n + 1)(k − n)
+ km + βk(n + 1) − αm + β(m + 1)(k − m)]vk−m−n
= (m − n)(k − α − β(m + n + 1))vk−n−m
= (m − n)dm+n(vk).
An alternate way to write these elements is
X dn = (k − α − β(n + 1))Ek−n,k. k∈Z
19 3.5 Loop Algebras
Definition 16 Let gln denote the Lie algebra of all n × n matrices with complex entries and let C[t, t−1] denote the ring of Laurent polynomials in t with complex
−1 coefficients. We define the loop algebra gle n to be gln(C[t, t ]), i.e., the complex Lie algebra of n × n matrices with Laurent polynomials as entries.
An element of gle n has the form
X k a(t) = t ak with ak ∈ gln k where k runs over a finite subset of Z. The standard basis {eij|0 ≤ i, j ≤ n} of gln yields a basis for gle n via
k eij;k = t eij (1 ≤ i, j ≤ n and k ∈ Z).
The elements of gle n form an associative algebra with multiplication defined on the basis by
k+l eij;kemn;l = t eijemn = δjmein;k+l.
Then the Lie bracket on gle n can be defined as the commutator:
[eij;k, emn;l] = δjmein;k+l − δniemj;k+l.
k+l Note that [eij;k, emn;l] = t [eij, emn].
n The Lie algebra gln has a natural representation on the space C given by matrix
n multiplication. Let {e1, . . . , en} denote the standard basis of C where eij is the n×1
−1 n column vector with entries (ej)i = δij, i, j ∈ {1, . . . , n}. The vector space C[t, t ]
20 consists of n×1 column vectors with Laurent polynomials in t as entries. The vectors
n −k vj;k = t ej with k ∈ Z, j ∈ {1, . . . , n}
−1 n −1 n ∞ n form a basis of C[t, t ] (over C). Thus C[t, t ] is identified with C via vj;k 7→ vnk+j. Now gle n acts on this space by
n vi;s−k j = l n eij;k( vl;k) = . 0 otherwise
n Using the identification of vi;j with vjn+k, this means
eij;kvkn+l = δjlvn(s−k)+i. (8)
For a(t) ∈ gle n denote the corresponding matrix ina ¯∞ by τ(a(t)). Then we have the following matrix representation of gle n ina ¯∞: