Notes on Lie Groups, Lie algebras, and the Exponentiation Map Mitchell Faulk

Transcription

1 Notes on Lie Groups, Lie algebras, an the Exponentiation Map Mitchell Faulk 1. Preliminaries. In these notes, we concern ourselves with special objects calle matrix Lie groups an their corresponing Lie algebras, connecte via the exponentiation map. Our main goal is to establish that the Lie algebra corresponing to a matrix Lie group is a real vector space. Thus, in orer to further stuy a Lie group, one might more easily stuy its corresponing Lie algebra, since vector spaces are well-stuie. In these notes, we refer to the set of m n matrices over the fiel F by M m n (F ). We also use the notation R for fiel of real numbers an C for the fiel of complex numbers. We expect that the reaer is familiar with the exponential of a matrix. Namely, if A is an n n matrix, the matrix e A := i=0 A i i! (1) is the exponential of A, often calle e to the A. It can be shown that the exponential function exp : M n n (C) M n n (C) efine by exp(a) = e A is well-efine with omain M n n (C). In other wors, the series in (1) converges for each A M n n (C). 2. Matrix Lie Groups. Before we can answer the question of what a matrix Lie group is, we must first efine a group. Definition 1. A group is a set G together with a map : G G G [ (g 1, g 2 ) g 1 g 1 ] such that 1. (g 1 g 2 ) g 3 = g 1 (g 2 g 3 ), g 1, g 2, g 3 G; 2. There exists e G such that e g = g e = g for each g G; 3. For each g G there exists h G such that h g = g h = e. Example 2. We give several examples of matrix groups. In each case, the prouct is given by matrix multiplication. Recall that matrix multiplication is associative. 1

2 1. GL(n, R) := {A M n n (R) : A 1 exists}. The general linear group (over R). Note that (a) GL(n, R) is close with respect to matrix multiplication, since (AB) 1 = B 1 A 1. (b) The n n ientity matrix is an element of GL(n, R). (c) For each A, A 1 is an element of GL(n, R) satisfying property SL(n, R) := {A M n n (R) : et(a) = 1}. The special linear group (over R). (a) SL(n, R) is close with respect to matrix multiplication, since if et A = 1 an et B = 1, then et AB = 1. (b) The n n ientity is an element of SL(n, R). (c) For each A, A 1 exists an et A 1 = 1 as well. 3. U(n) := {A M n n (C) : AA = I}. The unitary group. 4. SU(n) := {A M n n (C) : AA = 1, et A = 1}. The special unitary group. 5. H. The Heisenberg group. The set of real matrices of the form 1 a b 0 1 c Note that H is close uner multiplication an that 1 a 1 b 1 a ac b 0 1 c = 0 1 c All of these examples are actually more precisely calle matrix Lie groups. We give the precise efinition of a matrix Lie group below. But first, we give a remark on Sophus Lie. Remark 3. Sophus Lie ( ) was a Norwegian mathematician. His major achievement was iscovering that such transformation groups as mentione above coul be better unerstoo by linearizing them an stuying the corresponing linear spaces, calle Lie algebras. Definition 4. Let A n be a sequence of complex matrices. We say that A n converges to a matrix A if each entry of A n converges to the corresponing entry of A. Definition 5. A matrix Lie group is any subgroup H of GL(n, C) with the property that if A n is a sequence of matrices in H an A n converges to a matrix A, then either A belongs to H or A is not invertible. 2

3 Essentially, what the efinition of a matrix Lie group requires is that H be a close subset of GL(n, C). Geometrically, one can think of matrix Lie groups as manifols embee in a space. Although we o not iscuss at length on this notion, it oes help motivate the following iscussion. 3. Connecteness. Definition 6. A matrix Lie Group G is connecte if for each E, F G, there is a continuous path A : [a, b] G, such that A(t) G for each t, A(a) = E, an A(b) = F. It turns out that most of the groups from Example 2 are connecte. We show now the case for U(n). In our proof, we will construct a path from any unitary matrix to the ientity matrix. It follows that one can construct a path between any two unitary matrices, by first traveling to the ientity, then out to the other unitary matrix. Proposition 7. The group U(n) is connecte for each n 1. Proof. Let U be unitary. Recall that if U is unitary, then U U = I = UU, so U is normal. By the spectral theorem for normal operators, there exists an orthonormal system of eigenvectors for U. Recall that each e-value of U satisfies λ = 1, so we may write e iθ e iθ2 0 U = U U 1 ( ) 0 0 e iθn where U 1 is unitary an θ i R. Conversely, one can show that a matrix of the form ( ) is unitary by noting that UU = I. For each t [0, 1] efine e i(1 t)θ e i(1 t)θ2 0 A(t) = U U e i(1 t)θn Then A(0) = U an A(1) = I. Furthermore, A(t) is unitary for every t. So each unitary matrix can be connecte to the ientity. So U(n) is connecte. We have constructe the following table isplaying whether each group from Example 2 is connecte. 3

4 Group Connecte? Components GL(n, C) Yes 1 GL(n, R) No 2 SL(n, C) Yes 1 SL(n, R) Yes 1 U(n) Yes 1 H Yes 1 Note 8. The group GL(n, R) is not connecte for the following reason. Suppose that et E > 0 an et F < 0. Let A : [0, 1] M n n (R) be a continuous path connecting E an F. Compose et A, a continuous function. By the intermeiate value theorem, there is a t (0, 1) such that (et A)(t) = 0, i.e., the path A leaves GL(n, R). 4. Lie Algebra of a Matrix Lie Group. As was mentione earlier, it is often easier to stuy the so-calle Lie algebra of a matrix Lie group than the group itself. This is because the algebra has some nice properties which we begin to iscuss now. Definition 9. Let G be a matrix Lie group. The Lie algebra of G, enote g, is the set of all matrices X such that e tx belongs to G for each real number t. Example 10. We now iscuss the corresponing Lie algebras of those groups in Example gl(n, R) := {X : e tx GL(n, R) t R}. Claim 0.1. gl(n, R) = M n n (R). Proof. Let X M n n (R). Then e tx is invertible an real for each real t, i.e., e tx GL(n, R). Suppose X is a matrix such that e tx is real for each t. Then X = t t=0e tx is real, i.e., X M n n (R). 2. sl(n, R) := {X M n n (R) : e tx SL(n, R) t R}. Claim 0.2. sl(n, R) = {X M n n (R) : tr(x) = 0}. Proof. We require the theorem et e X = e tr(x) (which we o not prove in these notes). Let X satisfy tr(x) = 0. Then tr(tx) = 0 for each t. So et e tx = 1 for each t as require. Conversely, suppose that et e tx = 1 for each t. Then e t tr(x) = 1 for each t. So t tr(x) = (2πi)k, k Z for each t. So tr(x) = 0. 4

5 3. u(n) := {X M n n (C) : e tx U(n) t R}. Claim 0.3. u(n) = {X M n n : X = X} (the set of skewsymmetric matrices) Proof. Recall that if e tx U(n), then (e tx ) = (e tx ) 1 = e tx. Note that (e tx ) = e t X = e tx. So X = X. Conversely, we can see that such X satisfy e tx U(n) for each t. 4. h := {X M n n (R) : e tx H t R} (The Heisenberg algebra). The Heisenberg algebra h consists of all matrices of the form 0 a b 0 0 c We want to iscuss some properties of a Lie algebra. But first, we require some knowlege about the matrix exponential an the matrix logarithm. 5. The Exponential an the Logarithm. We first make two observations about the matrix exponential function. Suppose we have n n matrices A an X where A is invertible. Then (AXA 1 ) m = AX m A 1. So e AXA 1 = Ae X A 1 (2) by looking at the series in (1) term by term. Now suppose X is an n n matrix an t is a real number. Then t etx = Xe tx = e tx X. (3) This follows by ifferentiating the series in (1) term-by-term an recognizing that X commutes with powers of itself. We now efine the matrix logarithm. Although we o not concern ourselves too much with the construction of this function, it can be shown that such a function is well-efine an satisfies the following properties. (In the theorem below, let enote the operator norm on n n matrices.) 5

6 Theorem 11. The function log : M n n (C) M n n (C) efine by log A = m=1 m+1 (A I)m ( 1) m is well-efine an continuous on the set of all n n matrices satisfying A I < 1. An if A is real, then log A is real. Furthermore, for each A with A I < 1, e log A = A. For each X satisfying X < log 2, we have e X I < 1 an log e X = X. Hence, for matrices of a certain istance from the ientity, the logarithm is the inverse of the exponential. Note that log is efine as we woul expect: an extension of the Taylor expansion for log to matrices. Lemma 12. There exists a constant c such that for all n n matrices A with A < 1/2, we have log(i + A) A c A 2. Proof. We note that log(i + A) A = ( 1) m Am m = A2 ( 1) m Am 2 m. A property of the operator norm is that XY X Y. A further property is that X + Y X + Y. So log(i + A) A A 2 A 2 A 2 ( 1) m Am 2 A m 2 m m (1/2) m 2 m. The series on the right converges to some c R by the ratio test. So we obtain as require. log(i + A) A c A 2 Theorem 13. Let X an Y belong to M n n (C). Then ( ) e X+Y = lim e 1 m X e 1 m Y m. m 6

7 Proof. Expaning the first few terms of the power series multiplication for the exponential, we obtain e 1 m X e 1 m Y = I + 1 m X + 1 m Y + C m where C m is a series such that C m k/m 2 for some k R. By choosing m large enough, we can make e 1 m X e 1 m Y close enough to I, thus making e 1 m X e 1 m Y in the omain of the logarithm. Therefore, for sufficiently large m, applying Lemma 12 gives ( ) ( log e 1 m X e 1 m Y = log I + 1 m X + 1 ) m Y + C m (4) where E m is an error term which satisfies = 1 m X + 1 m Y + C m + E m, (5) E m c 1 m X + 1 m Y + C m 2 c C m 2 ck m 2, for each m 2 an for some c R. Applying exp to both sies of (4) gives ( e 1 m X e 1 1 m Y = exp m X + 1 ) m Y + C m + E m. So ( e 1 m X e 1 m Y ) m = exp (X + Y + mcm + me m ). (6) Since both C m an E m are on the orer of 1/m 2 an since the exponential is a continuous function, we take the limit of both sies of (6) to obtain ( ) lim e 1 m X e 1 m Y m = exp(x + Y ), m as require. 6. Properties of a Lie Algebra Using our work in the previous section, we now obtain a nice result concerning Lie algebras, namely, that they are real vector spaces. First we require a small proposition. Proposition 14. Let G be a matrix Lie group an g its corresponing Lie algebra. Let X be an element of g an let A be an element of G. Then AXA 1 belongs to g. 7

8 Proof. By (2) of section 5, we see that e t(axa 1) = Ae tx A 1. Since A, A 1 an e tx belong to G for each t, we see that AXA 1 belongs to g by efinition. Theorem 15. Let G be a matrix Lie group an let g be its Lie algebra. Let X, Y g. Then 1. sx g for each s R. 2. X + Y g. Proof. For (1), if X g, then e tx G for each t. So e tsx G for each s since ts R for each t, s R. Therefore, sx belongs to g by efinition. For (2), let X an Y belong to g. Theorem 13 gives e t(x+y ) = lim m ( exp ( t m X ) exp ( t m Y )) m. (7) Since X an Y belong to g, exp ( t m X) an exp ( t m Y ) belong to G. Moreover, since G is a group, their prouct belongs to G, as oes any power of their prouct. To make sure that the limit is in G, we nee to make sure that the limit is invertible. Since the left han sie of (7) is invertible with inverse e t(x+y ), the limit is also invertible. Therefore, e t(x+y ) belongs to G. By efinition, X + Y belongs to g. Thus, we have establishe that the Lie algebra is a real vector space. Because of this, we immeiately get an interesting property about the Lie algebra: that it is close uner the so-calle bracket. The bracket of two elements X an Y in a matrix Lie algebra g, enote [X, Y ] is given by [X, Y ] = XY Y X. In a general Lie algebra, the bracket nee not be efine in this way. But in the case of matrices, this bracket efinition is commonly use. Corollary 16. Let G be a matrix Lie group an let g be its Lie algebra. Let X, Y g. Then [X, Y ] = XY Y X belongs to g. Proof. Recall the property of the exponential given by (3) in section 5: t etx = Xe tx = e tx X. By the Leibniz rule of ifferentiation, we have Evaluation at t = 0 gives t etx Y e tx = e tx Y ( e tx ) + ( e tx Y ) e tx t t = e tx Y Xe tx + e tx XY e tx. t t=0e tx Y e tx = Y X + XY = XY Y X. A property of vector spaces is that the erivative of any smooth curve lying in g must also be in g. Since by Proposition 14, e tx Y e tx belongs to g for each t, we see that XY Y X belongs to g. 8

9 References Brian C. Hall, Lie Groups, Lie Algebras, an Representations: An Elementary Introuction. Grauate Texts in Mathematics, Karin Ermann & Mark J. Wilon, Introuction to Lie Algebras. Springer,