1: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2: \chapter{Lie algebras}\label{c.lie}
3:
4: Part III introduces the basics about Lie algebras and \at{.}
5: Lie groups,
6: with an emphasis on the concepts most relevant to the conceptual side
7: of physics.
8:
9: \bigskip
10: This chapter introduces Lie algebras together with the slightly richer
11: structure of a Lie $*$-algebra usually encountered in the mechanical
12: applications. We introduce tools for verifying
13: the Jacobi identity, and establish the latter both for the Poisson
14: bracket of a classical harmonic oscillator and, for quantum systems,
15: for the commutator of linear operators.
16:
17: Further Lie algebras arise as algebras of matrices closed under
18: commutation, as algebras of derivations in associative algebras,
19: as centralizers or quotient algebras, and by complexification.
20: An overview over semisimple Lie algebras and their classification
21: concludes the chapter.
22:
23: \bigskip
24: \at{adapt}
25: In finite dimensions, the relation is almost one-to-one, the ''almost''
26: being due to the fact that the so-called universal covering group of
27: a finite-dimensional Lie algebra (defined in Section \ref{univ.cover})
28: may have a nontrivial discrete normal subgroup.
29:
30: Many finite-dimensional Lie groups arise as groups of square
31: invertible matrices, and we discuss the most important families,
32: in particular the unitary and the orthogonal groups. We introduce
33: group representations, which relate groups of matrices (or linear
34: operators) to abstract Lie groups, and will turn out to be most
35: important for understanding the spectrum of quantum systems.
36:
37: Of particular importance for systems of oscillators are the Heisenberg
38: groups, the universal covering groups of the Heisenberg algebras.
39: Their product law is given by the famous Weyl relations, which are
40: an exactly representable case of the Baker--Campbell--Hausdorff
41: formula valid for many other Lie groups, in particular for
42: arbitrary finite-dimensional ones.
43: We also discuss the Poincar\'e group. This is the symmetry group of
44: space-time, and forms the basis for relativity theory.
45:
46:
47:
48:
49:
50:
51:
52:
53: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
54: \section{Basic definitions}\label{s.lie}
55:
56: We start with the definition of a Lie algebra over a field
57: $\Kz$, usually implicitly given by the context.
58: In our course, $\Kz$ is either the field $\Cz$ of complex
59: numbers, occasionally the field $\Rz$ of real numbers. Lie algebras
60: over other fields, such as the rationals $\Qz$ or finite fields $\Zz_p$
61: for $p$ prime, also have interesting applications in mathematics,
62: physics and engineering, but these are outside the scope of this book.
63: To denote the Lie product, we use the symbol $\lp$ introduced at the
64: end of Section \ref{s.thstat}. (This replaces other, bracket-based
65: notations common in the literature.)
66:
67: \begin{definition}~\\
68: (i) A {\bfi{Lie product}} on a vector space $\Lz$ over $\Kz$ is a
69: bilinear operation\index{$\lp$} on $\Lz$ satisfying
70:
71: (L1) $f\lp f = 0$,
72:
73: (L2) $f\lp (g\lp h) + g \lp (h \lp f)+ h \lp (f \lp g)=0$
74: for all $f,g,h\in \Lz$.
75:
76: Equation (L2) is called the {\bfi{Jacobi identity}}.
77:
78: (ii) For subsets $A,B$ of $\Lz$, we write
79: \[
80: A \lp B :=\{f\lp g\mid f\in A,g\in B\},
81: \]
82: and for $f,g\in\Lz$,
83: \[
84: A\lp g :=A\lp \{g\},~~~ f\lp B:=\{f\}\lp B.
85: \]
86: (iii)
87: A {\bfi{Lie algebra}} over $\Kz$ is a vector space $\Lz$ over $\Kz$
88: with a distinguished Lie product. Elements $f\in\Lz$ with $f\lp \Lz=0$
89: are called \bfi{(Lie) central}\index{central}\index{Lie central};
90: the set $Z(\Lz)$ of all these elements is called the {\bfi{center}}
91: of $\Lz$. A \bfi{real (complex)} Lie algebra is a Lie algebra over
92: $\Kz=\Rz$ (resp. $\Kz=\Cz$).
93: Unless confusion is possible, we use the same symbol $\lp$ for the Lie
94: product in different Lie algebras.
95: \end{definition}
96:
97: Clearly, if $f\lp g$ defines a Lie product of $f$ and $g$, so does
98: $f {\lp\!}_\iota g := \iota (f\lp g)$ for all $\iota\in \Kz$.
99: Thus the same vector space may be a Lie algebra in different ways.
100:
101: In physics, finite-dimensional Lie algebras are often defined in terms
102: of basis elements $X_k$ called \bfi{generators}\index{generator} and
103: {\bfi{structure constants}} $c_{jkl}$, such that
104: \lbeq{phys.struct}
105: X_j\lp X_k =\sum_l c_{jkl}X_l\,.
106: \eeq
107: By taking linear combinations and using the bilinearity of the Lie
108: product, the structure constants determine the Lie product completely.
109: Conversely, since the generators form a basis, the structure constants
110: are determined uniquely by the basis. They depend, however, on the
111: basis chosen. Frequently, there are distinguished bases with a physical
112: interpretation in which the structure constants are particularly simple,
113: and most of them vanish. If a basis and the structure constants are
114: given, man Lie algebra computations can be done automatically;
115: important software packages include LIE (\sca{van Leeuwen} et al.
116: \cite{vLeuCL}) and LTP (\sca{Torres-Torriti} \cite{TorT}).
117: In this book, we usually prefer a basis-free approach, resorting to
118: basis-dependent formulas only to make connections with traditional
119: physics notation.
120:
121:
122: As a consequence of (L1) (and in fact equivalent to it), we have the
123: following antisymmetry property:
124: \[
125: f\lp g = - g\lp f \,.
126: \]
127: This follows from observing that $f\lp 0 = 0 \lp f = 0$ and
128: \beqar
129: 0&=& (f+g)\lp (f+g)= f\lp
130: f
131: + f\lp g + g\lp f + g\lp g \nonumber\\
132: &=& f\lp g + g\lp f\,. \nonumber
133: \eeqar
134: Using the antisymmetry property of the Lie product one can write
135: the Jacobi identity in two other important forms, each equivalent
136: with the Jacobi identity:
137: \lbeq{jac.der}
138: f \lp (g\lp h) = (f \lp g)\lp h + g\lp (f\lp h)\,,
139: \eeq
140: \lbeq{jac.der2}
141: (f \lp g)\lp h = (f \lp h)\lp g+ f\lp (g\lp h)\,.
142: \eeq
143: These formulas say that one can apply the Lie product to a compound
144: expression in a manner familiar from the product rule for
145: differentiation.
146:
147: An important but somewhat trivial class of Lie algebras are the
148: {\bfi{abelian} Lie algebras}, where $f\lp g=0$ for all $f,g\in \Lz$.
149: It is trivial to check that (L1) and (L2) are satisfied.
150: Clearly, every vector space can be turned into an abelian Lie algebra
151: by defining $f\lp g=0$ for all vectors $f$ and $g$.
152: In particular, the field $\Kz$ itself and the center of any Lie
153: algebra are abelian Lie algebras.
154:
155: A subspace $\Lz'$ of a Lie algebra $\Lz$ is a {\bfi{Lie
156: subalgebra}} if it is closed under the Lie product, i.e., if
157: $f\lp g \in\Lz'$ for all $f,g\in\Lz'$. In this case,
158: the restriction of the Lie product $\lp$ of $\Lz$ to
159: $\Lz'$ turns $\Lz'$ into a Lie algebra. That is, a Lie subalgebra is a
160: subspace that is a Lie algebra with the same Lie product.
161: (For example, the subspace $\Kz f$ spanned by an arbitrary element $f$
162: of a Lie algebra is an abelian Lie subalgebra.)
163: A Lie subalgebra is \bfi{nontrivial} if it is not the whole Lie algebra
164: and contains a nonzero element.
165:
166:
167: \bigskip
168: The property $(L1)$ is usually easy to check.
169: It is harder to check the Jacobi identity $(L2)$ for a
170: proposed Lie product; direct calculations can be quite messy
171: when many terms have to be calculated before one finds that they
172: all cancel. Since we will encounter many Lie products that must be
173: verified to satisfy the Jacobi identity, we first develop some
174: technical machinery to make life easier, or at least more structured.
175: For a given binary bilinear operation $\circ$ on $\Lz$, we define
176: the {\bfi{associator}} of $f,g,h\in\Lz$ as
177: \lbeq{e.assoc}
178: [f,g,h]:= (f\circ g)\circ h - f\circ (g\circ h)\,.
179: \eeq
180:
181: \begin{prop}\label{assoclie}
182: If the associator of a bilinear operator $\circ$ on $\Lz$
183: satisfies
184: \lbeq{assoid}
185: [f,g,h]+[g,h,f]+[h,f,g]-[f,h,g]-[h,g,f]-[g,f,h]=0\,,
186: \eeq
187: then
188: \[
189: f\lp g := f\circ g - g\circ f\,
190: \]
191: defines a Lie product on $\Lz$.
192: \end{prop}
193:
194:
195: \bepf
196: Define
197: \[
198: J(f,g,h):= f\lp (g\lp h) + g \lp (h \lp f)+ h \lp (f \lp g)\,,
199: \]
200: and define
201: \[
202: S(f,g,h):= [f,g,h]+[g,h,f]+[h,f,g]-[f,h,g]-[h,g,f]-[g,f,h]\,.
203: \]
204: Writing out $S(f,g,h)$ and $J(f,g,h)$ with
205: $f \lp g :=f\circ g - g\circ f$, one sees $J(f,g,h)= - S(f,g,h)$
206: and hence if $S(f,g,h)=0$ for all $f,g$ and $h$, then the Jacobi
207: identity is satisfied for all $f,g$ and
208: $h$. The antisymmetry property $f\lp f= 0$ is trivial.
209: \epf
210:
211:
212: \begin{thm}\label{t.1dpoisson}
213: The binary operation $ \lp$ defined on the vector space
214: $C^\infty(\Rz\times\Rz)$ by
215: \[
216: f\lp g := f_p g_q - g_p f_q\,,
217: \]
218: where $f_p=\partial f/\partial p$ and $f_q=\partial f/\partial q$,
219: is a Lie product.
220: \end{thm}
221:
222: \bepf
223: We calculate the associator for the bilinear operator
224: $f\circ g = f_p g_q$. We have
225: \beqar
226: [f,g,h]&=& (f\circ g)_p h_q -f_p(g\circ h)_q \nonumber \\
227: &=& (f_pg_q)_p h_q - f_p(g_ph_q)_q \nonumber\\
228: &=& f_{pp}g_qh_q + f_pg_{qp}h_q - f_pg_{pq}h_q - f_p g_p h_{qq}
229: \nonumber\\
230: &=& f_{pp}g_qh_q - f_p g_p h_{qq}\nonumber
231: \eeqar
232: Writing the cyclic permutations we get
233: \beqar
234: [f,g,h]+[g,h,f]+[h,f,g]&=& f_{pp}g_qh_q +g_{pp}h_qf_q + h_{pp}f_q
235: g_q
236: \nonumber\\
237: &&-f_pg_p h_{qq} - g_ph_p f_{qq} - h_p f_p g_{qq}\,,\nonumber
238: \eeqar
239: which is symmetric in $f,g$; hence the identity
240: \gzit{assoid} is satisfied. Proposition \ref{assoclie} therefore
241: implies that $\lp$ is a Lie product.
242:
243: The reader is invited to prove this result also by a direct calculation.
244: \epf
245:
246: We end this section by introducing some concepts needed at various
247: later points but collected here for convenience.
248: If $\Lz$ and $\Lz'$ are Lie algebras we call a linear map $\phi:\Lz\to
249: \Lz'$ a {\bfi{homomorphism}} (of Lie algebras) if
250: \[
251: \phi(f\lp g)=\phi(f)\lp \phi(g)
252: \]
253: for all $f,g\in\Lz$. Note that the left-hand side involves the Lie
254: product in $\Lz$, whereas the right-hand side involves the Lie product
255: in $\Lz'$. An injective homomorphism is called an {\bf\idx{embedding}}
256: of $\Lz$ into $\Lz'$.
257: We call two Lie algebras $\Lz$ and $\Lz'$ {\bfi{isomorphic}}
258: if there is a homomorphism $\phi:\Lz\to \Lz'$ and a homomorphism
259: $\psi:\Lz'\to\Lz$ such that $\psi\circ\phi$ is the identity on $\Lz$
260: and $\phi\circ\psi$ is the identity on $\Lz'$. Then $\phi$ is called
261: an {\bfi{isomorphism}}, and $\psi$ is the inverse isomorphism.
262:
263:
264: Given a Lie algebra $\Lz$ and a subalgebra $\Lz'$, the
265: {\bfi{centralizer}} \idx{$C_{\Lz'}(S)$} in $\Lz'$ of a subset
266: $S\subset\Lz$ is defined by
267: \[
268: C_{\Lz'}(S) = \left\{ f\in \Lz' \mid f\lp g =0
269: \Forall g\in S\right\}\,.
270: \]
271: In words, $C_{\Lz'}(S)$ consists of all the elements in $\Lz'$ that Lie
272: commute with all elements in $S$. One may use the Jacobi identity to see
273: that $C_{\Lz'}(S)$ is a Lie subalgebra of $\Lz'$.
274:
275: An {\bfi{ideal}} of $\Lz$ is a subspace $I\subseteq \Lz$ such that
276: $f\lp g\in I$ for all $f\in \Lz$ and for all $g\in I$.
277: In other notation $\Lz \lp I=I\lp \Lz \subseteq I$.
278: Note that $0$ and $\Lz$ itself are always ideals;
279: they are called the trivial ideals. Also, the center of a Lie algebra
280: is always an ideal. A less trivial ideal is the
281: {\bfi{derived Lie algebra}\index{Lie algebra!derived} $\Lz^{(1)}$
282: of $\Lz$} consisting of all elements that can be
283: written as a finite sum of elements of $\Lz\lp\Lz$.
284: If $I\subseteq \Lz$ is an ideal in $\Lz$ one may form the
285: \idx{quotient Lie algebra}\index{Lie algebra!quotient} $\Lz/I$,
286: whose elements are
287: the equivalence classes $[f]$ of all $g\in\Lz$ such that $f-g\in I$,
288: with addition, scalar multiplication, and Lie product given by
289: \[
290: \alpha [g]:=[\alpha g]\,,~~~
291: [f] + [g]:=[f + g]\,,~~~
292: [f]\lp [g]:=[f\lp g]\,.
293: \]
294: It is well-known that the vector space operations are well-defined.
295: The Lie product is well-defined since $f'\in[f]$ implies $f'-f\in I$,
296: hence $(f-f')\lp g\in I$ and
297: $[f']\lp [g]=[f'\lp g]=[f \lp g + (f'-f)\lp g]=[f\lp g]$.
298:
299: If $\Lz$ and $\Lz'$ are Lie algebras, their {\bfi{direct sum}}
300: $\Lz\oplus \Lz'$ is the direct sum of the vector spaces equipped with
301: the Lie product defined by
302: \[
303: (x+x')\lp (y+y')=x\lp y + x'\lp y'
304: \]
305: for all $ x,y\in\Lz$ and all $x',y'\in\Lz'$. It is easily verified that
306: the axioms are satisfied.
307:
308:
309: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
310: \section{Lie algebras from derivations}\label{sec.calc.lie}
311:
312: Equation \gzit{jac.der},
313: \[
314: f \lp (g\lp h) = (f \lp g)\lp h + g\lp (f\lp h)\,.
315: \]
316: resembles the product rule for (partial) differentiation;
317: \[
318: \frac{\partial}{\partial x}(gh) = \frac{\partial g}{\partial x} h
319: + g\frac{\partial h}{\partial x}\,.
320: \]
321: To make the similarity more apparent we introduce for every
322: element $f\in \Lz$ a linear operator $\ad_f:\Lz\rightarrow \Lz$,
323: the {\bfi{derivative in direction} $f$},
324: given by\index{$\ad_f$}
325: \[
326: \ad_f g := f\lp g\,.
327: \]
328: The notation reflects the fact that the operator $\ad:\Lz\to\Lz$
329: defined by
330: \[
331: \ad f:=\ad_f
332: \]
333: is the {\bfi{adjoint representation}}\index{$\ad$} of $\Lz$; see
334: Sections \ref{fin-dim} and \ref{sec-fin-dim}. \at{check placement}
335:
336: Note that an element $f\in\Lz$ is in the center $Z(\Lz) = C_{\Lz}(\Lz)$
337: of $\Lz$ if and only if the linear operator $\ad_f$ is zero.
338:
339: \begin{expl}
340: For $f$ in the Lie algebra $C^\infty(\Rz\times\Rz)$ constructed in
341: Theorem \ref{t.1dpoisson}, we have
342: \lbeq{e.Xf}
343: \ad_f g= f\lp g= f_pg_q - f_qg_p
344: = \left( f_p\frac{\partial}{\partial q} -
345: f_q\frac{\partial}{\partial p}\right) g\,.
346: \eeq
347: The vector field \idx{$X_f$} on $\Rz\times \Rz$ defined by the
348: coefficients of $\ad_f$ \at{.} is
349: called the {\bfi{Hamiltonian vector field}} defined
350: by $f$; cf. Chapter 12. \at{.}
351: In particular, the Hamiltonian derivative operators with
352: respect to $p$ and $q$ take the explicit form
353: \[
354: X_p = \frac{\partial}{\partial q}\,, \ \ \
355: X_q = -\frac{\partial}{\partial p}\,,
356: \]
357: and we have
358: \[
359: \ad_f = f_p X_p +f_q X_q.
360: \]
361: \end{expl}
362:
363: With the convention that operators bind stronger than the Lie product,
364: the Jacobi identity can be written in the form
365: \[
366: \ad_f (g\lp h) = \ad_f g \lp h + g\lp \ad_f h\,.
367: \]
368: The Jacobi identity is thus equivalent to saying that the operator
369: $\ad_f$ defines for every $f$ a derivation of the Lie algebra.
370:
371:
372: \begin{dfn}~\\
373: (i) A {\bfi{derivation}} of a vector space $\Az$ with a
374: bilinear product $\circ$ is a linear map $\delta:\Az\rightarrow\Az$
375: satisfying
376: \[
377: \delta (f\circ g) = \delta f\circ g + f \circ \delta g\,,
378: \]
379: for all $f,g\in\Az$. We denote by \idx{$\der\Az$} the
380: set of all derivations of $\Az$. (In the cases of interest,
381: $\Az$ is an associative algebra with the associative product as $\circ$,
382: or a Lie algebra with the Lie product as $circ$.
383:
384: (ii) If $\Ez$ is an associative algebra $\Ez$, a
385: \bfi{(left) $\Ez$-module} is an additive abelian group $\Vz$
386: together with a multiplication mapping which assigns to
387: $f\in\Ez$ and $x\in\Vz$ a product $fx\in\Vz$ such that
388: \[
389: f(x+y)=fx+fy,~~~(f+g)x=fx+gx,~~~f(gx)=(fg)x
390: \]
391: for all $f,g\in\Ez$ and all $x,y\in\Vz$.
392: \end{dfn}
393:
394: \begin{prop} \label{p.der}
395: The commutator of two derivations is a derivation. In particular,
396: $\der \Ez$ is a Lie subalgebra of $\Lin
397: \Ez$ with Lie product
398: \[
399: \delta\lp \delta' := [\delta,\delta'].
400: \]
401: Moreover, if $\delta\in\der \Ez$ and $f\in\Ez$ then the \bfi{product}
402: $f\delta$ defined by
403: \[
404: (f\delta)g:=f(\delta g)
405: \]
406: is a derivation, and turns $\der \Ez$ into an $\Ez$-module.
407: \end{prop}
408: \bepf
409: Since $\der \Ez$ is a linear vector space, and since
410: the antisymmetry property and the Jacobi identity are already
411: satisfied in $\Lin\Ez$, we only need to check that the Lie
412: product of two derivations is again a derivation. We have:
413: \beqar
414: (\delta\lp\delta')(fg) &=& (\delta\delta')(fg) -
415: (\delta'\delta)(fg) \nonumber\\
416: &=& \delta ( (\delta'f)g + f(\delta'g) ) - \delta' ( (\delta f)g +
417: f(\delta g) ) \nonumber\\
418: &=& (\delta \delta' f)g +f\delta\delta' g - (\delta'\delta f)g -
419: f\delta\delta' g \nonumber\\
420: &=& (\delta \lp \delta' f)g + f(\delta\lp\delta' g)\nonumber
421: \eeqar
422: This proves the first part. The second part is straightforward.
423: \epf
424:
425: \begin{prop}\label{lem.der.cent}
426: The {\bfi{centralizer}} of a subset $S$ of $\Lin \Ez$, defined as
427: \[
428: C(S):=
429: \left\{ \delta \in\der \Ez \mid A\lp \delta=0\,\Forall A\in S\right\},
430: \]
431: is a Lie subalgebra of $\der\Ez$.
432: \end{prop}
433: \bepf As before, we only need to prove that the Lie product closes
434: within $C(S)$. If $\delta,\delta'\in C(S)$, the
435: Jacobi identity in the form \gzit{jac.der} implies
436: \[
437: A\lp (\delta\lp \delta') = (A\lp \delta)\lp \delta' +
438: \delta \lp (A\lp \delta') = 0\,.
439: \]
440: \epf
441:
442:
443: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
444: \section{Linear groups and their Lie algebras}\label{s.lLie}
445:
446: In quantum mechanics, linear operators play a central role;
447: they appear in two essentially different ways:
448: Operators describing time evolution and canonical transformations
449: are linear operators $U$ on a Hilbert space, that are \bfi{unitary}
450: in the sense that $U^*U=UU^*=1$, and hence bounded\footnote{
451: The bounded operators on a Hilbert space a so-called $C^*$-algebra;
452: see for example \sca{Rickart} \cite{Ric}, \sca{Baggett} \cite{baggett},
453: or \sca{Werner} \cite{werner}. But we do not use this fact.
454: }. % end footnote
455: The unitary operators form a group, which in many cases of interest is
456: a so-called Lie group.
457:
458: On the other hand, many important quantities in quantum mechanics are
459: described in terms of unbounded linear operators that are
460: defined not on the whole Hilbert space but only on a dense subspace.
461: Usually, the linear operators of interest have a common dense domain
462: $\Hz$ on which they are defined and which they map into itself.
463: $\Hz$ inherits from the Hilbert space the Hermitian inner product,
464: hence is a complex \bfi{Euclidean space}, and the Hilbert space can
465: be reconstructed from $\Hz$ as the \bfi{completion} $\ol\Hz$ of $\Hz$
466: by equivalence classes of Cauchy sequences, in the way familiar from
467: the construction of the real numbers from rationals. We therefore
468: consider the algebra \idx{$\Lin\Hz$} of continuous linear operators
469: on a Euclidean space $\Hz$, with composition as associative
470: product.
471:
472: In this section, we define the basic concepts relevant for a study of
473: groups and Lie algebras inside algebras of operators. Since for these
474: concepts neither the operator structure nor the coefficient field
475: matters in most cases - as long as the characteristic is not two -, we provide a slightly more general framework.
476: In the next section, we apply the general framework to the algebra
477: $\Cz^{n\times n} = \Lin \Cz^n$ of complex $n\times n$-matrices,
478: considered in the standard way as linear operators on the space
479: $\Cz^n$ of column vectors with $n$ complex entries.
480: Many of the Lie groups and Lie algebras arising in the applications
481: are naturally definied as subgroups or subspaces of this algebra.
482:
483: \bigskip
484: An {\bfi{(associative) algebra}} over a field $\Kz$
485: is a vector space $\Ez$ over $\Kz$ with a bilinear, associative
486: multiplication. For example, every $*$-algebra is an associative algebra
487: over $\Cz$. As traditional, the product of an associative algebra
488: (and in particular that of $\Lin\Hz$ and $\Kz^{n\times n}$) is
489: written by juxtaposition. An associative algebra $\Ez$ is called
490: \bfi{commutative} if $fg=gf$ for all $f,g\in\Ez$, and
491: \bfi{noncommutative} otherwise. In many cases we assume that such an
492: algebra has a unit element $1$ with respect to multiplication;
493: after the identification of the multiples of $1$ with the elements
494: of $\Kz$, this is equivalent to assuming that $\Kz\subseteq \Ez$.
495: If $\Ez$ and $\Ez'$ are associative
496: algebras over $\Kz$ with $1$, then a $\Kz$-linear map $\phi:\Ez\to \Ez'$ is an
497: \idx{algebra homomorphism} if $\phi(fg)=\phi(f)\phi(g)$ and
498: $\phi(1)=1$. Often we omit the reference to the ground field $\Kz$ and
499: assume a ground field has been chosen.
500:
501: We now show that every associative algebra has many Lie products,
502: and thus can be made in many ways into a Lie algebra.
503: For commutative Lie algebras, the construction is a bit cumbersome and
504: only leads to abelian Lie algebras.
505:
506: \begin{thm}\label{ass.lie.J}
507: Let $\Ez$ be an associative algebra. Then,
508: for every $J\in\Ez$, the binary operation $\lp_J$ defined on $\Ez$ by
509: \[
510: f{\lp\!\!}_J g := fJg-gJf\,
511: \]
512: is a Lie product. In particular ($J=1$), the binary operation
513: $\lp$ defined on $\Ez$ by
514: \[
515: f\lp g := [f,g]\,
516: \]
517: where
518: \[
519: [f,g]:=fg-gf
520: \]
521: denotes the {\bfi{commutator}} of $f$ and $g$, is a Lie product.
522: \end{thm}
523:
524: \bepf
525: We compute the associator \gzit{e.assoc} for the bilinear
526: operation $f\circ g := fJg$:
527: \[
528: [f,g,h]= (f\circ g)Jh - fJ(g\circ h) = fJgJh - fJgJh =0\,,
529: \]
530: by associativity. Hence the associator of
531: $\circ$ satisfies \gzit{assoid}. Hence ${\lp\!\!}_J$ is a Lie product.
532: \epf
533:
534: Note that $Jf \lp Jg =
535: J(f{\lp\!\!}_J g)$. Hence the corresponding
536: Lie algebras are isomorphic when $J$ is invertible.
537:
538:
539: If $\Ez$ and $\Ez'$ are two associative algebras with unity,
540: we may turn them into Lie algebras by putting $f\lp
541: g =[f,g]$ in both $\Ez$ and $\Ez'$. We denote by $\Lz$ and $\Lz'$ the
542: Lie algebra associated to $\Ez$ and $\Ez'$, respectively.
543: If $\phi$ is an algebra homomorphism
544: from $\Ez$ to $\Ez'$ then $\phi$ induces a Lie algebra
545: homomorphism between the Lie algebras $\Lz$ and $\Lz'$. Indeed
546: $\phi(f\lp g)=\phi(fg-gf)=\phi(f)\phi(g)-\phi(g)\phi(f)
547: =\phi(f)\lp \phi(g)$.
548:
549: Theorem \gzit{ass.lie.J} applies in particular to $\Ez=\Kz^{n\times n}$.
550: The Lie algebra $\Kz^{n\times n}$ with Lie product $f\lp g:=[f,g]$ is
551: called the {\bfi{general linear algebra}} $gl(n,\Kz)$ over $\Kz$.
552: If $\Kz=\Cz$, we simply write $gl(n)=gl(n,\Cz)$;
553: similar abbreviations apply without notice for the names of other
554: Lie algebras introduced later.
555:
556:
557: \begin{dfn}\label{d.llie}~\\
558: (i) A \bfi{Hausdorff $*$-algebra} is a $*$-algebra $\Ez$ with a
559: Hausdorff topology in which addition, multiplication, and conjugation
560: are continuous. An element $f\in\Ez$ is called \bfi{complete} if
561: the initial-value problem
562: \lbeq{e.complete}
563: \frac{d}{dt} U(t) = f U(t),~~~U(0)=1
564: \eeq
565: has a unique solution $U:\Rz\to\Ez$. Then the mapping $U$ is called a
566: \bfi{one-parameter group} with \bfi{infinitesimal generator} $f$,
567: and we write $e^{tf}:=U(t)$; this notation is unambiguous since
568: it is easily checked that $e^{t(sf)}=e^{(ts)f}$ for $s,t\in\Rz$.
569: An element $f\in\Ez$ is called \bfi{self-adjoint} if $f^*=f$ and
570: the product $if$ with the imaginary unit is complete.
571: We call an element $g \in \Ez$ \bfi{exponential} if it is of the form
572: $g=e^f$ for some complete $f\in\Ez$.
573: We call a Hausdorff $*$-algebra $\Ez$ an \bfi{exponential algebra}
574: if the set of exponential elements in $\Ez$ is a neighborhood of $1$.
575: \at{Is it enough to require that there is a path-connected
576: neighborhood of 1?}
577:
578: (ii) A \bfi{linear group} is a set $\Gz$ of invertible elements of some
579: associative algebra $\Ez$ such that $1\in\Gz$ and
580: \[
581: g,g'\in\Gz\implies g^{-1},gg'\in \Gz.
582: \]
583: If $\Ez$ is given with a topology in which its operations are
584: continuous, we consider $\Gz$ as a topological group with the
585: topology induced by calling a subset of $\Gz$ open or closed if it
586: is the intersection of an open or closed set of $\Ez$ with $\Gz$.
587:
588: (iii) A \bfi{linear Lie group} is a closed subgroup of the group
589: $\Ez^\times$ of all invertible elements of an exponential algebra $\Ez$.
590: A \bfi{Lie group} is a group $\widetilde\Gz$ with a
591: Hausdorff topology that is \bfi{isomorphic} to some linear Lie group
592: $\Gz$, i.e., for which there is an invertible mapping
593: $\phi:\widetilde\Gz\to\Gz$ such that $\phi$ and $\phi^{-1}$ are
594: continuous and $\phi(1)=1$, $\phi(gg')=\phi(g)\phi(g')$ for all
595: $g,g'\in\widetilde\Gz$.
596: \end{dfn}
597:
598: For all exponential algebras $\Ez$, the group $\Ez^\times$ is a linear
599: Lie group. Note that the law $e^fe^{f'}=e^{f+f'}$ holds if $f$ and $f'$
600: commute but not in general. In particular,
601: \[
602: e^fe^{-f}=e^{0}=1.
603: \]
604:
605: If $\Ez$ is a \bfi{Banach algebra}, i.e., if the topology of $\Ez$ is
606: induced by a norm $\|\cdot\|$ satisfying $\|fg\|\le \|f\|\,\|g\|$,
607: it is not very difficult to show that every $f\in\Ez$ is complete, and
608: we have for all $f\in\Ez$ the absolute convergent series expansion
609: \[
610: e^f = \sum_{k=0}^\infty \frac{f^k}{k!},
611: \]
612: and that $f=\log g$, where
613: \[
614: \log g := -\sum_{k=1}^\infty \frac{(1-g)^k}{k}\for \|1-g\|<1,
615: \]
616: provides an $f$ such that $g=e^f$.
617: Therefore every Banach algebra is exponential.
618: Note that the \bfi{exponential mapping}, which maps a matrix $f\in\Ez$
619: to $e^f\in\Ez^\times$, is usually not surjective. \at{counterexample?}
620:
621: The above applies to the case $\Ez=\Cz^{n\times n}$ with the maximum
622: norm, which is a Banach algebra, which covers all finite-dimensional
623: Lie groups.
624: In infinite dimensions, however, many interesting linear Lie groups
625: are not definable over Banach algebras.
626:
627:
628: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
629: \section{Classical Lie groups and their Lie algebras}\label{s.Liega}
630:
631:
632: This section is not yet in a good form.
633: \at{this section needs some polishing}
634:
635: \bigskip
636: A \bfi{matrix group} is a linear group in an algebra $\Kz^{n\times n}$.
637: In this section, we define the most important matrix groups and
638: the corresponding Lie algebras. Although we get Lie algebras no matter
639: which field is involved, we get Lie groups in the
640: sense defined above only when $\Kz$ is the field of real numbers of
641: the field of complex numbers\footnote{
642: For the real and fields, because exponentials can be defined,
643: the groups have a natural differential geometric
644: structure as differentiable manifolds; cf. Section \ref{s.Liegroups}.
645: For general fields, there are no exponentials, and one needs to
646: replace the differential geometric structure inherent in Lie groups
647: by an algebraic geometry structure, and may then interpret general
648: matrix groups as so-called \bfi{groups of Lie type}.
649: In particular, for finite fields, one gets the \bfi{Chevalley groups},
650: which figure prominently in the classification of finite simple groups.
651: }. % end footnote
652:
653: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
654: % The general linear group $GL(n,\Kz)$
655: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
656: \begin{expl}
657: The group
658: $GL(n,\Kz)$\index{$GL(n,\Kz)$}\index{$gl(n,\Kz)$} of all invertible
659: $n\times n$-matrices over $\Kz=\Rz$ or $\Kz=\Cz$ is a linear group,
660: The subgroup of $GL(n,\Kz)$
661: consisting of the matrices with unit determinant is denoted by
662: \idx{$SL(n,\Kz)$}. In other words, $SL(n,\Kz)$ is the kernel of the map
663: $\det:GL(n,\Kz)\to \Kz^*$, where \idx{$\Kz^*$} is the group of
664: invertible elements in $\Kz$. The Lie algebra of $SL(n,\Kz)$ is
665: denoted by \idx{$sl(n,\Kz)$} and consists of the traceless $n\times n$
666: matrices with entries in $\Kz$.
667: \at{show $\log GL(n,\Kz)=gl(n,\Kz)$.}
668: By Theorem \ref{ass.lie.J}, the algebra of $n\times n$-matrices
669: with entries in $\Kz$ is a Lie algebra the commutator as Lie product;
670: this Lie algebra is denoted by \idx{$gl(n,\Kz)$}.
671: The center of $gl(n,\Kz)$ is easily seen to be the
672: $1$-dimensional subalgebra spanned by the identity matrix,
673: $Z(gl(n,\Kz))=\Kz 1 = \Kz$.
674: \end{expl}
675:
676: Every subspace of a Lie algebra closed under the Lie product is again
677: a Lie algebra. This simple recipe provides a large number of useful
678: Lie algebras defined as Lie subalgebras of some $gl(n,\Kz)$.
679: Conversely, the (nontrivial) {\bfi{theorem of Ado}}, \at{ref? eg Jacobsen}
680: not proven here but see e.g. \sca{Jacobsen} \cite{jacobsen}, states that every finite-dimensional Lie algebra
681: is isomorphic to a Lie subalgebra of some $gl(n,\Rz)$.
682:
683: The group $GL(n,\Kz)$ is one of the most important finite-dimensional
684: linear groups and all finite-dimensional linear groups are isomorphic
685: tosubgroups of $GL(n,\Kz)$ for some $n$.
686: If $\Kz=\Rz$ or $\Kz=\Cz$ then every closed subgroup $\Gz$ of
687: $GL(n,\Kz)$ is a Lie group. These Lie groups have associated
688: Lie algebras $\Lz=\log\Gz$ of infinitesimal generators.
689: For any Lie subgroup $G$ of $GL(n,\Kz)$ one gets the Lie
690: algebra by looking at the vector space of those elements $X$ of
691: $gl(n,\Kz)$ such that
692: $e^{\eps X}$ is in $G$ for $\eps$ small enough. This criterion
693: is very useful since we can take $\epsilon$ so small that we only have
694: to look at the terms linear in $\epsilon$ so that we don't have to
695: expand the exponential series completely. If the subgroup $\Gz\subset
696: GL(n,\Kz)$ is connected and either compact or nilpotent, \at{defined?}
697: then the exponential map can be shown to be surjective, see
698: e.g. \sca{Knapp} \cite{knapp}.
699:
700:
701: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
702: % The special linear group $SL(n,\Kz)$
703: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
704:
705: The Lie algebra $sl(n,\Kz)$ is the Lie subalgebra of $gl(n,\Kz)$ given
706: by the traceless matrices. The dimension is $n^2-1$ and we have
707: \[
708: sl(n,\Kz) \cong gl(n,\Kz) / \Kz\,.
709: \]
710: The quotient is well defined and is a Lie algebra because $\Kz$
711: is the center and thus in particular an ideal.
712:
713: If $\Lz$ is a Lie algebra over $\Rz$ then by taking the tensor product
714: with $\Cz$ and extending the Lie bracket in a $\Cz$-linear way, one
715: obtains the \bfi{complexification} of $\Lz$, denoted $\Lz^\Cz$. The proces
716: of complexification is also called \bfi{extension of scalars}. In
717: particular, if we write $\Lz^\Cz=\Cz\otimes_\Rz \Lz$ then in $\Lz^\Cz$
718: the Lie bracket is given by $(\alpha\otimes x)\lp (\beta\otimes
719: y)=\alpha\beta\otimes (x\lp y)$. The reverse proces is called
720: \bfi{realization} or \bfi{restriction of scalars}; we clarify the
721: proces of restriction of scalars by an example.
722:
723:
724: \begin{example}\label{ex.real.sl}
725: Consider $\Lz=sl(2,\Cz)$. We wish to calculate \idx{$sl(2,\Cz)^\Rz$}.
726: A basis of $sl(2,\Cz)$ is given by the elements
727: \[
728: \pmatrix{1& 0\cr 0&-1}\,,~~\pmatrix{0&1\cr0&0}\,,~~\pmatrix{0&0\cr
729: 1&0}\,.
730: \]
731: This basis is as well a basis for $sl(2,\Rz)$; therefore we see
732: $sl(2,\Cz)^\Rz\cong sl(2,\Rz) \oplus_\Rz i\,sl(2,\Rz)$ as real vector
733: spaces. The Lie product of $f+ig$ and $f'+ig'$ for $f,f'\in sl(2,\Rz)$
734: and $ig,ig'\in i\,sl(2,\Rz)$ is given by
735: \[
736: (f+g)\lp (f'+g')= f\lp f' - g\lp g' + i(f\lp g' +f'\lp g)\,.
737: \]
738: The reader who has already some experience with Lie algebras is
739: encouraged to verify the isomorphism $sl(2,\Cz)^\Rz\cong so(3,1)$.
740: \end{example}
741:
742:
743: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
744: % The orthogonal groups $O(n,\Kz)$ and $O(p,q,\Kz)$
745: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
746:
747: \begin{expl}\label{ex.ortho}
748: Suppose we have a symmetric bilinear form $B$ on
749: $\Kz^n$. The Lie algebra $so(n,B;\Kz)$ is the subspace of all
750: $f\in sl(n,\Kz)$ satisfying
751: \lbeq{lie.sopq}
752: B(fv,w)=-B(v,fw).
753: \eeq
754: We leave it to the reader to show that if $f$ and $g$ satisfy
755: \gzit{lie.sopq}, then so does $fg-gf$; thus we have indeed a Lie
756: algebra. In the special case where $B(v,w)=v^Tw$, the
757: Lie algebra $so(n,B;\Kz)$ is called the
758: {\bfi{complex orthogonal Lie algebra}}\index{orthogonal Lie
759: algebra}\index{Lie algebra!orthogonal} \idx{$so(n,\Kz)$}. In
760: matrix language, $so(n,\Kz)$ is the Lie algebra of antisymmetric
761: matrices with entries in $\Kz$ and has dimension $n(n-1)/2$.
762:
763: An {\bfi{orthogonal matrix}} is a matrix $Q$ satisfying
764: \lbeq{e.orth}
765: Q^TQ=1.
766: \eeq
767: The orthogonal $n\times n$-matrices with coefficients in a field $\Kz$
768: form a subgroup of the group $GL(n,\Kz)$, the
769: {\bfi{orthogonal group}} $O(n,\Kz)$\index{$O(3)$}.
770: \at{this overlaps with the next example}
771: Since \gzit{e.orth} implies that $(\det Q)^2=1$, orthogonal matrices
772: have determinant $\pm 1$.
773: The orthogonal matrices of determinant one form a
774: subgroup of $O(n,\Kz)$, the {\bfi{special orthogonal group}}
775: \idx{$SO(n,\Kz)$}.
776: The corresponding Lie algebra is \at{why?} $so(n,\Kz)=\log O(n,\Kz) =
777: \log SO(n,\Kz)$, the Lie algebra of antisymmetric $n\times n$-matrices.
778: In particular, the group $SO(3)=SO(3,\Rz)$\index{$SO(3)$} consists of
779: the rotations in 3-space and was discussed in some detail in
780: Section \ref{s.mink}.
781: \end{expl}
782:
783: For a \bfi{nondegenerate}\index{bilinear form!nondegenerate} $B$
784: (i.e., one where $B(v,w)=0$ for all $v$ implies $w=0$) and
785: $\Kz=\Cz$ (or any algebraically closed field), we can always choose
786: a basis in which the bilinear form is represented as the identity
787: matrix. Therefore all $so(n,B;\Kz)$ with nondegenerate $B$ are
788: isomorphic to $so(n,\Kz)$.
789:
790: Over $\Kz=\Rz$, symmetric bilinear forms
791: are classified by their signature, i.e., the triple $(p,q,r)$
792: consisting of the number $p$ of positive, $q$ of negative, and $r$ of
793: zero eigenvalues of the symmetric matrix $A$ representing the bilinear
794: form $B$; $B(v,w)=v^TAw$. The form $B$ is nondegenerate if and only if $r=0$.
795: Bilinear forms with the same signature lead to
796: isomorphic Lie algebras. In particular, $so(p,q)$ denotes a
797: Lie algebra $so(p+q,B,\Rz)$ where $B$ is a nondegenerate symmetric
798: bilinear form $B$ on $\Rz^n$ of signature $(p,q,0)$.
799: The basis can always be chosen such that the representing matrix $A$
800: is
801: \[
802: I_{p,q}= \pmatrix{ 1_p & 0 \cr 0&- 1_q }\,,
803: \]
804: where $1_p$ and $1_q$ are the $p\times p$ and $q\times q$ identity
805: matrix, respectively. In this basis, the Lie algebra \idx{$so(p,q)$} is
806: the subalgebra of $gl(n,\Rz)$ consisting of elements $f$ satisfying
807: \[
808: f^T I_{p,q}+I_{p,q}f=0\,.
809: \]
810: Note that if $f\in so(p,q)$ then
811: \[
812: 0= \tr\left((f^T
813: I_{p,q}+I_{p,q}f)I_{p,q}\right)=2\tr(fI_{p,q}^2)=2\tr(f)
814: \]
815: and hence $so(p,q)\subset sl(n,\Rz)$.
816:
817: \begin{expl}\label{iso-ortho}
818: Let $V$ be a vector
819: space over a field $\Kz$. Suppose $V$ is equipped
820: with a symmetric or antisymmetric nondegenerate bilinear form $B$.
821: There is a symmetry group associated to the bilinear form consisting
822: of the linear transformations $Q:V\to V$ such that
823: \[
824: B(Qv,Qw)=B(v,w)
825: \]
826: for all $v,w$ in $V$. If $B$ is symmetric one calls the group of these
827: linear transformations an {\bfi{orthogonal group}} and denotes it by
828: \idx{$O(B,\Kz)$}. The associated Lie algebra is $o(B,\Kz)$. Indeed,
829: $e^{tf}$ transforms $x,y$ into
830: \[
831: \bary{lll}
832: B(e^{tf}x,e^{tf}y) &=& B(e^{tf}x,e^{tf}y)\\
833: &=& B((1+tf)x,(1+tf)y)+ O(t^2)\\
834: &=& B(x,y) +t B(fx,fy) + O(t^2).
835: \eary
836: \]
837: \at{but in the Lie algebra section, we used $x^TJy$ in place of
838: $B(x,y)$; adapt!}
839: \end{expl}
840:
841: \begin{expl}\label{ex.SOpq}
842: When $\Kz=\Rz$, one has for symmetric bilinear forms another
843: subdivision, since $B$
844: can have a definite \idx{signature} $(p,q)$ where $p+q$ is the dimension
845: of $V$. If $B$ is of signature $(p,q)$, this means that there exists
846: a basis of $V$ in which $B$ can be represented as
847: \[
848: B(v,w) = v^T A w\,,~~~\textrm{where}~~~A =
849: \textrm{diag}(\underbrace{-1,\dots, -1}_{p \mbox{\scriptsize ~times}},
850: \underbrace{1,\dots, 1}_{q\mbox{\scriptsize~times}})\,.
851: \]
852: The group of all linear
853: transformations that leaves $B$ invariant is denoted by \idx{$O(p,q)$}.
854: The subgroup of $O(p,q)$ of transformations with determinant one is
855: the so-called \bfi{special orthogonal group}\index{orthogonal
856: group!special} and is denoted by \idx{$SO(p,q)$}.
857: The associated real Lie algebra is denoted \idx{$so(p,q)$} and its
858: elements are linear transformations $A:V\to V$ such that
859: for all $v,w\in V$ we have $B(Av,w)+B(v,Aw)=0$. The Lie product
860: is given by the commutator of matrices.
861:
862: The group of all translations in $V$ generates together with $SO(p,q)$
863: the group of {\bfi{inhomogeneous special orthogonal
864: transformations}}, which is denoted \idx{$ISO(p,q)$}. One can obtain
865: $ISO(p,q)$ from $SO(p,q+1)$ by performing a contraction; that is,
866: by rescaling some generators with some parameter $\epsilon$ and then
867: choosing a singular limit $\epsilon\to 0$ or $\epsilon \to \infty$.
868: The group $ISO(p,q)$ can also be seen as the group of
869: $(p+q+1)\times (p+q+1)$-matrices of the form
870: \[
871: \pmatrix{ Q & b \cr 0 & 1 } ~~~ \textrm{with }~ Q\in SO(p,q)\,,~ b\in
872: V\,.
873: \]
874: The Lie algebra of $ISO(p,q)$ is denoted \idx{$iso(p,q)$} and can be
875: described as the
876: Lie algebra of $(p+q+1)\times (p+q+1)$-matrices of the form
877: \[
878: \pmatrix{ A & b \cr 0 & 0 } ~~~ \textrm{with }~ A\in so(p,q)\,,~ b\in
879: V\,.
880: \]
881: Again, the Lie product in $iso(p,q)$ is the commutator of matrices.
882: \end{expl}
883:
884:
885: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
886: % The symplectic group $Sp(2n,\Kz)$
887: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
888: We define the
889: {\bfi{symplectic Lie algebra}}\index{Lie algebra!symplectic}
890: \idx{$sp(2n,\Kz)$} as the
891: Lie subalgebra of $gl(2n,\Kz)$ given by the elements $f$ satisfying
892: \lbeq{symp.alg}
893: f^TJ+Jf=0\,,
894: \eeq
895: where $J$ is the $2n\times 2n$-matrix given by
896: \[
897: J=\pmatrix{ 0 & -1_n \cr 1_n& 0}\,.
898: \]
899: We leave it to the reader to verify that if $f$ and $g$ satisfy
900: \gzit{symp.alg},
901: then so does $fg-gf$. Another useful exercise is to prove
902: $sl(2,\Kz)\cong sp(2,\Kz)$. (Caution: The reader is warned that
903: in the literature there are different notational conventions
904: concerning the symplectic Lie algebras. For example, some people write
905: $sp(n,\Kz)$ for what we and many others call $sp(2n,\Kz)$.)
906:
907: If $B$ is antisymmetric in the example \ref{iso-ortho}, the group is
908: called a {\bfi{symplectic group}} and one writes \idx{$Sp(B,\Kz)$}.
909: The associated Lie algebras is $sp(B,\Kz)$.
910: If $V$ is of finite dimension $m$ one writes $Sp(B,\Kz)=Sp(m,\Kz)$.
911: Note that $m$ is necessarily even.
912:
913:
914: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
915: % The unitary group $U(n,\Kz)$
916: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
917: Other real Lie algebras that play a major role in many areas of
918: physics are the
919: {\bfi{unitary Lie algebras}}\index{Lie algebra!unitary} and the
920: {\bfi{special unitary Lie algebras}}\index{Lie algebra!special
921: unitary} -- called so because they are the
922: generating algebras of the groups of (special) unitary matrices,
923: a term that will be explained in Section \ref{s.Liegroups}.
924: The unitary Lie algebra \idx{$u(n)$} consists of all antihermitian
925: complex $n\times n$ matrices. The special unitary Lie algebra is
926: defined as the antihermitian $n\times n$ complex traceless matrices
927: and is denoted \idx{$su(n)$}. It is clear that $su(n)\subset u(n)$.
928: It might seem weird to
929: call a Lie algebra real if it consists of complex-valued
930: matrices. However, as a vector space the antihermitian complex
931: $n\times n$ matrices form a real vector space. If $f$ is a
932: antihermitian matrix, then $if$ is Hermitian. The dimension (as a
933: real vector space) of
934: $su(n)$ is $n^2-1$, and the dimension of $u(n)$ is $n^2$.
935: It is a good exercise to check that $so(3)\cong su(2)$ since these
936: two Lie algebras will return very often. A hint: $so(3)$
937: consists of anti-symmetric real $3\times 3$ matrices, so there are
938: only three. Choosing an obvious basis for both $su(2)$ and $so(3)$
939: will do the job.
940:
941: \begin{example}
942: A complex matrix $U$ is {\bfi{unitary}} if it satisfies
943: \[
944: UU^\dagger=1,
945: \]
946: where $(U^\dagger)_{ij}=\bar U_{ji}$. Since the inverse of a matrix is
947: unique, it follows that also $U^\dagger U=1$. By splitting all the
948: matrix entries into a real and imaginary part $U_{ij}=A_{ij}+iB_{ij}$
949: we see that the set of $n\times n$ unitary matrices makes up a
950: submanifold of $\Rz^{2n^2}$ of dimension $n^2 $. The linear group of
951: unitary $n\times n$ matrices is denoted \idx{$U(n)$}.
952: \at{adapt text; part of it also holds for $SL(n)$.}
953: \[
954: U=e^A=\sum_{k=0}^{\infty}\frac{1}{k!}A^k\,.
955: \]
956: Then multiply $A$ with a parameter $t$, take $t\to 0$ and keep only
957: the linear terms: $U=1+tA+O(t^2)$. Then
958: since $U$ has to be unitary, we obtain
959: \[
960: 1=(1+tA+O(t^2))(1+tA+O(t)^2)^\dagger=1+t(A+A^\dagger)+O(t)^2\,,
961: \]
962: implying that $A$ has to be antihermitian.
963: Thus the Lie algebra of infinitesimal generators of $U(n)$ is
964: \idx{$u(n)$}.
965:
966: The subgroup of $U(n)$ of all elements with determinant 1 is denoted
967: by \idx{$SU(n)$} and is called the {\bfi{special unitary group}}.
968: The dimension of $SU(n)$ is $n^2-1$. For the determinant we get
969: \[
970: \det U=1+\tr tA+O(t)^2\,,
971: \]
972: and thus the trace of infiniteesimal generators of $SU(n)$ has to
973: vanish, and we see that the corresponding Lie algebra is $su(n)$.
974: Note that the Lie algebra \idx{$u(n)$}
975: contains all real multiples of $i\cdot 1$, which commutes with all
976: other elements. Hence $u(n)$ has a center, whereas $su(n)$ does not.
977:
978: In the case $n=2$ it is a nice exercise to show that each special
979: unitary matrix $U$ can be written as
980: \[
981: U=\pmatrix{ x& y\cr -\bar y & \bar x }\,,~~~x,y\in \Cz\,,~
982: |x|^2+|y|^2=1\,.
983: \]
984: Writing $x=a+ib$ and $y=c+id$ for $a,b,c,d\in\Rz$ we see that
985: $a^2+b^2+c^2+d^2=1$. This implies that there is
986: a one-to-one correspondence between $SU(2)$ and the set of points
987: on the unit sphere $S^3$ in $\Rz^4$.
988: Thus $SU(2)$ is as a manifold homeomorphic to $S^3$.
989: In particular $SU(2)$ is compact.
990: Hence every element $U$ of $SU(2)$ can be
991: written as the exponent of a matrix $A$.
992:
993: Physicists prefer to work with Lie algebras defined by Hermitian
994: matrices, corresponding to Lie $*$-algebras. \at{clarify}
995: In the applications, distinguished real generators typically represent
996: important real-valued observables. Therefore they tend
997: to replace the matrix $A$ by $iA$ for a Hermitian matrix $A$. This is
998: one of the reasons why the structure constants for real algebras
999: appear in the physics literature with an $i$, as alluded at the end of
1000: Section \ref{sec.calc.lie}.
1001: % Although we specialized to $n=2$, the analysis goes through for $n>2$;
1002: % the Lie algebra $u(n)$ consists of the antihermitian matrices and
1003: % has a center. The Lie algebra $su(n)$ consists of all traceless
1004: % antihermitian matrices and has no center.
1005: \end{example}
1006:
1007:
1008: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1009: \section{Heisenberg algebras and Heisenberg groups}
1010:
1011: A {\bfi{Heisenberg algebra}} is a Lie algebra $\Lz$ with a
1012: 1-dimensional center and a distinguished Lie central element 1
1013: called {\bfi{one}} or {\bfi{identity}}, such that every
1014: $f\lp g$ is a multiple of 1 for all $f,g\in\Lz$. There is an
1015: embedding of $\Kz$ into $\Lz$ given by $\alpha\mapsto \alpha 1$
1016: which can be used to identify the multiples of 1 with the multipliers
1017: from the field, so that $\Kz=Z(\Lz)\subset \Lz$.
1018:
1019: When we divide out the center of a Heisenberg algebra we obtain an
1020: abelian Lie algebra. More generally, let $\Lz$ be any Lie algebra and
1021: let $\Lz'$ be another Lie algebra with a subalgebra $Z$ contained in the
1022: center of $\Lz'$. If $\Lz'/Z$ is isomorphic to $\Lz$, one calls $\Lz'$
1023: a {\bfi{central extension}} of $\Lz$.\footnote{
1024: %%%% footnote
1025: In more abstract terms, central extensions are conveniently described
1026: by short exact sequences. Let $A_i$ be a set of Lie algebras and suppose
1027: that there are maps $d_i:A_i\to A_{i+1}$;
1028: \begin{equation}
1029: \begin{diagram}[heads=LaTeX]
1030: \ldots &\rTo & A_{i-1}& \rTo^{d_{i-1}} & A_{i} & \rTo^{d_i}&
1031: A_{i+1} & \rTo\\
1032: \end{diagram}\,.
1033: \end{equation}
1034: We call the sequence {\bfi{exact}} if
1035: $\textrm{Ker~} d_i = \textrm{Im~} d_{i-1}$ for all $i$ where
1036: there are $d_{i-1}$ and $d_i$.
1037: As an exercise, the reader is invited to verify the following
1038: assertion: The sequence $0\to A\to B\to 0$ is exact if and only if
1039: $A\cong B$ and the isomorphism is the map from $A$ to $B$. A
1040: short exact sequence is a sequence of maps of the form
1041: \[
1042: 0\rightarrow A \rightarrow B \rightarrow C \rightarrow 0\,.
1043: \]
1044: A {\bfi{central extension}} of $\Lz$ is then a Lie algebra $\Lz'$
1045: such that there
1046: is an exact sequence $0\to Z\to \Lz' \to \Lz \to 0$ with $Z$ abelian.
1047: } %%%% end footnote
1048:
1049:
1050: Corresponding to any Heisenberg algebra there is an alternating
1051: bilinear form $\omega:\Lz\times\Lz\rightarrow \Kz$ given by
1052: \[
1053: f\lp g = \omega(f,g)\,.
1054: \]
1055: Conversely, given such a form on an arbitrary vector space $\Vz$
1056: not containing 1, this formula turns $\Lz:=\Kz\oplus\Vz$ into a
1057: Heisenberg algebra. If $\omega$ is nondegenerate on $\Vz$ it defines
1058: a \bfi{symplectic form} on $\Vz$.
1059:
1060: The Heisenberg algebra \idx{$h(n)$} is the special case where
1061: $\Kz=\Cz$, $\Vz=\Cz^{2n}$, and $\omega$ is nondegenerate. Thus $h(n)$
1062: is a central extension of the abelian Lie algebra $\Cz^{2n}$ and has
1063: dimension $2n+1$. We can find a
1064: basis of $\Vz$ consisting of vectors $p_k$ and $q_l$ for $1\leq k,l
1065: \leq n$ such that $\omega(p_k,p_l)=\omega(q_k,q_l)=0$ for all $k,l$ and
1066: $\omega(p_k,q_l)=\delta_{kl}$; that is, $\omega$ is then the standard
1067: symplectic form on $\Kz^{2n}$ represented by the matrix
1068: $\pmatrix{ 0 & -1 cr 1 \cr 0}$. Thus Heisenberg algebras encode
1069: symplectic vector spaces in a Lie algebra setting. Everything done here
1070: extends with appropriate definitions to general symplectic manifolds,
1071: and, indeed, much of classical mechanics can be phrased in terms of
1072: symplectic geometry, the geometry of such manifolds -- we refer the
1073: reader to the exposition by \sca{Arnold} \cite{arnold} on
1074: classical mechanics and symplectic geometry.
1075:
1076: \begin{expl}
1077: Let us write \idx{$t(n,\Kz)$} for the Lie subalgebra of
1078: $gl(n,\Kz)$ consisting of upper-triangular matrices and
1079: $n(n,\Kz)$ as the Lie subalgebra of $gl(n,\Kz)$
1080: consisting of strictly upper-triangular matrices, which have zeros
1081: on the diagonal.
1082:
1083: The Lie algebra $t(3,\Kz)$ of strictly upper triangular
1084: $3\times 3$-matrices is a Heisenberg algebra with
1085: \[
1086: 1=\pmatrix{ 0&0&1\cr 0&0&0\cr 0&0&0 }\,,
1087: \]
1088: since
1089: \[
1090: \pmatrix{ 0 & \alpha & \gamma \cr 0 & 0 & \beta \cr0&0&0 } \lp
1091: \pmatrix{ 0 & \alpha' & \gamma' \cr 0 & 0 & \beta'\cr 0&0&0 } =
1092: \pmatrix{ 0 & 0 & \alpha\gamma'-\gamma\alpha' \cr 0 & 0 & 0 \cr
1093: 0&0&0 } = \alpha\gamma'-\gamma\alpha'.
1094: \]
1095: The Lie algebra $t(3,\Cz)$ is called \bfi{the \idx{Heisenberg algebra}};
1096: thus if one talks about ''the'' (rather than ''a'') Heisenberg algebra,
1097: this Lie algebra is meant and is denoted $h(1)$.
1098: Introducing names for the special matrices
1099: \[
1100: p:=\pmatrix{ 0&1&0\cr 0&0&0\cr 0&0&0},
1101: ~~~
1102: q:=\pmatrix{ 0&0&0\cr 0&0&1\cr 0&0&0 },
1103: \]
1104: we find that $p,q$ and 1 form a basis of $t(3,\Cz)$, and we can
1105: express the Lie product in the more compact form
1106: \lbeq{e.heislp}
1107: (\alpha p+\beta q+\gamma) \lp (\alpha' p+\beta' q+\gamma')
1108: = \alpha\beta' - \beta\alpha'.
1109: \eeq
1110: Defining
1111: \[
1112: (\alpha p+\beta q+\gamma)^*:=\ol\alpha p+\ol\beta q+\ol\gamma
1113: \]
1114: turns the Heisenberg algebra into a Lie $*$-algebra in which
1115: R$p$ and $q$ are Hermitian.
1116: \at{definition of Hermitian in a Lie $*$-algebra? Real is better!}
1117: Note that here $*$ is {\em not} the conjugate transposition of matrices!
1118:
1119: \gzit{e.heislp} implies that $p$ and $q$ satisfy the so-called
1120: {\bfi{canonical commutation relations}}
1121: \lbeq{e.CCR}
1122: p \lp q =1\,,~~~ p \lp p=q\lp q=0.
1123: \eeq
1124: Since $f \lp 1 =0$ when $1$ is Lie central, \gzit{e.CCR}
1125: completely specifies the Lie product. The canonical commutation
1126: relations are frequently found in textbooks on quantum mechanics,
1127: but we see that they just characterize the Heisenberg algebra.
1128:
1129: The notation $q$ and $p$ is chosen to remind of position of momentum.
1130: Indeed, the canonical commutation relations arise naturally in
1131: classical mechanics. In the Lie algebra $C^\infty(\Rz\times \Rz)$
1132: constructed in Theorem \ref{t.1dpoisson},
1133: we consider the set of {\bfi{affine functions}}, that is, those that
1134: are of the form $f(p,q) = \alpha_f p + \beta_f q+\gamma_f$, with
1135: $\alpha_f,\beta_f,\gamma_f\in\Cz$. In particular, the constant functions
1136: are included with $\alpha_f=\beta_f=0$, and we identify them with
1137: the constants $\gamma_f\in\Cz$. Given another affine function
1138: $g(p,q)=\alpha_gp+\beta_gq+\gamma_g$, we find
1139: \[
1140: f\lp g = \alpha_f\beta_g - \beta_f\alpha_g \in\Cz\,.
1141: \]
1142: Since $f\lp g$ is just a complex number times the function that is
1143: $1$ everywhere, it is a central element, that is, it Lie
1144: commutes with all other algebra elements. Thus the affine functions
1145: form a Heisenberg subalgebra of $C^\infty(\Rz\times \Rz)$, and $p$ and
1146: $q$ satisfy the canonical commutation relations.
1147: \end{expl}
1148:
1149: \at{move to Poisson algebras!}
1150: Suppose that a commutative Poisson algebra $\Ez$
1151: contains two elements
1152: $p$ and $q$ satisfying the canonical commutation relations \gzit{e.CCR}.
1153: Then $\Ez$ contains a copy of the Heisenberg algebra.
1154: The algebra of polynomials
1155: in $p$ and $q$ is then a Poisson subalgebra of $\Ez$ in which
1156: \gzit{e.Xf} is valid. This follows from Proposition \ref{der.bracket}.
1157: Thus the canonical commutation relations capture the essence of the
1158: commutative Poisson algebra $C^\infty(\Rz\times \Rz)$.
1159: But getting the bigger algebra requires taking limits which
1160: need not exist in $\Ez$, since with polynomials alone, one does not get
1161: all functions.
1162:
1163: \at{Define the \bfi{oscillator algebra} $os(n)$.}
1164:
1165:
1166: \at{interlace this with the algebra part}
1167:
1168: \begin{expl}
1169: An upper triangular $n\times n$-matrix
1170: is called {\bfi{unit upper triangular}} if its elements on
1171: the diagonal are $1$, and {\bfi{strictly upper triangular}} if its
1172: elements on the diagonal are zero. It is straightforward to check
1173: that the unit upper triangular $n\times n$-matrices form a subgroup
1174: \idx{$T(n,\Kz)$} of the group $GL(n,\Kz)$, and the strictly
1175: upper triangular $n\times n$-matrices form a Lie subalgebra of
1176: $gl(n,\Kz)$, which we denote by \idx{$t(n,\Kz)$}.
1177: We have $t(n,\Kz)=\log T(n,\Kz)$. In the following we shall look more
1178: closely at the case $n=3$ which is especially important.
1179: \end{expl}
1180:
1181:
1182: \at{we need $H(n)$; otherwise the section should be called
1183: ''The....''}
1184:
1185: The {\bfi{Heisenberg group}} is the group
1186: \lbeq{heis.grp}
1187: T(3,\Cz)=\Big\{ \pmatrix{ 1 & a & c \cr 0 & 1 & b \cr 0&0&1 } \,
1188: \Big|\, a,b,c\in \Cz \Big\}
1189: \eeq
1190: of unit upper triangular matrices in
1191: $\Cz^{3\times 3}$; its corresponding Lie algebra is the Heisenberg
1192: algebra $t(3,\Cz)$. Since the Heisenberg group is defined in terms of
1193: matrices, it comes immediately with a representation, the defining
1194: representation. Note that the defining representation is not unitary.
1195: [att group representations not yet defined.]
1196:
1197: The relation between the Heisenberg algebra and the Heisenberg
1198: group is particularly simple since the exponential map
1199: has a simple form. Indeed, if $A\in \Cz^{n\times n}$ then
1200: \lbeq{exp.mat}
1201: e^A = \sum_{k=0}^{\infty}\frac{A^k}{k!}\,,
1202: \eeq
1203: where $A^0=1$ is the identity matrix and the series \gzit{exp.mat}
1204: is absolutely convergent. A note on the infinite-dimensional case:
1205: For linear operators $A$ on a Hilbert space $\Hz$, the series
1206: converges absolutely only when $A$ is bounded (and hence
1207: everywhere defined); for unbounded but self-adjoint $A$ (which are
1208: only densely defined), convergence holds in a weaker sense giving
1209: \lbeq{exp.mat2}
1210: e^A \psi= \sum_{k=0}^{\infty}\frac{A^k}{k!} \psi
1211: \eeq
1212: for a dense set of vectors $\psi\in\Hz$ that are analytic for $A$.
1213:
1214: If $A\in t(3,\Cz)$, then a direct calculation shows that $A^2$ is of
1215: the form
1216: \[
1217: \pmatrix{ 0&0&c\cr 0&0&0\cr 0&0&0}
1218: \]
1219: for some $c\in \Cz$. Hence $A^3=0$ and the exponential of $A$ is
1220: simply given by $ e^A = 1+A+\shalf A^2 $. Thus if $A$ is given by
1221: \[
1222: A=\pmatrix{ 0 & \alpha & \gamma \cr 0 & 0 & \beta \cr 0&0&0 }\,,
1223: \]
1224: the exponential is given by
1225: \[
1226: e^A = \pmatrix{ 1 & \alpha & \gamma+\shalf \alpha\beta \cr 0 & 1 &
1227: \beta \cr 0&0&1 }\,.
1228: \]
1229: The map $A\to e^A$ is clearly bijective. The inverse map is given
1230: by the logarithm, which is
1231: for matrices defined by
1232: \lbeq{log-taylor}
1233: \log(1 + X) =
1234: \sum_{k=1}^{\infty}\frac{(-1)^{k-1}}{k}X^k\,,
1235: \eeq
1236: so that for the Heisenberg group $\Gz$ we have
1237: \[
1238: \log(X) = (X-1) - \shalf (X-1)^2 = -2+ 2X - \shalf X^2 \,.
1239: \]
1240: We are thus in the situation that both $T(3,\Cz) = \exp t(3,\Cz)$
1241: and $t(3,\Cz) =\log T(3,\Cz)$. This is not special to the Heisenberg
1242: group, neither does
1243: it hold in general. But there is a class of groups for which
1244: this holds. For example, the exponential map is surjective for all
1245: connected Lie groups that are compact or nilpotent (see below),
1246: see, e.g., \sca{Helgason} \cite{helgason} or
1247: \sca{Knapp} \cite{knapp}.
1248: The Heisenberg group is a noncompact but nilpotent Lie group.
1249:
1250:
1251: Let us shortly repeat what it means when a group is \idx{nilpotent}.
1252: Given any group $G$, we can form the \idx{commutator subgroup}
1253: $G^{(1)}$, which is generated by all elements of the form
1254: $aba^{-1}b^{-1}$ for all $a,b\in G$. We can also consider
1255: the commutator subgroup of $G^{(1)}$ and denote it by $G^{(2)}$.
1256: Repeating this procedure we get a sequence of groups
1257: \[
1258: G\supseteq G^{(1)}\supseteq G^{(2)}\supseteq ...
1259: \]
1260: A group is nilpotent if the procedure ends in a finite number of
1261: steps with the trivial group $G^{(n)}=1$. It is easy to see that the
1262: Heisenberg group is two-step nilpotent since $G^{(2)}=1$.
1263:
1264:
1265: Since the exponential map is bijective for the Heisenberg group,
1266: there exists a binary operation $\oplus$ on $t(3,\Cz)$, where
1267: $A\oplus B$ is the element with
1268: \lbeq{e.oplus}
1269: e^A e^{B} = e^{A\oplus B}\,.
1270: \eeq
1271: It is not difficult to give an explicit formula for $A\oplus B$.
1272: Since $A$ and $B$ are strictly upper triangular, we have $A^p B^q=0$
1273: for $p+q\geq 3$. We thus have
1274: \[
1275: e^Ae^B = (1+A+\shalf A^2) (1+B+\shalf B^2)
1276: = 1+A+B +\shalf (A^2+B^2+2AB)\,.
1277: \]
1278: Applying \gzit{log-taylor} we find
1279: \beqar A\oplus B &=& \log\left(
1280: 1+A+B +\shalf
1281: (A^2+B^2+2AB)\right)\nonumber\\
1282: &=& A+B +\shalf (AB-BA)\,,\nonumber
1283: \eeqar
1284: hence
1285: \lbeq{e.weyl0}
1286: A\oplus B=A+B +\shalf A\lp B \,.
1287: \eeq
1288: Thus we get from \gzit{e.oplus} the formula
1289: $e^Ae^B=e^{A+B +\shalf A\lp B}$. Since $A\lp B$ is central, it behaves
1290: just like a complex number, and we find
1291: the {\bfi{Weyl relations}}
1292: \lbeq{e.weyl1}
1293: e^{A+B}=e^{-\half A\slp B}e^Ae^B\,.
1294: \eeq
1295: In fact this result is also a direct consequence of the famous
1296: (but much less elementary) \bfi{Baker--Campbell--Hausdorff (BCH)
1297: formula}\index{Baker--Campbell--Hausdorff formula}
1298: that gives for general matrix Lie groups a series expansion of
1299: $A\oplus B$ when $A$ and $B$ are not too large.
1300: Even more generally, the \idx{Baker--Campbell--Hausdorff
1301: formula} applies to abstract finite-dimensional Lie groups\footnote{
1302: In infinite dimensions, additional assumptions are needed for the
1303: BCH-formula to hold.
1304: }
1305: that are not necessarily matrix groups and says that for two fixed
1306: Lie algebra elements $A$ and $B$ and for small enough real numbers $s$
1307: and $t$ there is a function $C$ from $\Rz\times \Rz$ to the Lie algebra
1308: such that we have
1309: \[
1310: e^{sA}e^{tB} = e^{C(s,t)}\,.
1311: \]
1312: The function $P(s,t)$ is given by a (for small $s,t$ absolutely
1313: convergent) infinite power series, the first terms of
1314: which are given by
1315: \[
1316: P(s,t)= sA+tB +\frac{st}{2} A\lp B + \frac{st}{12}\left(s A\lp
1317: (A\lp B) - t B\lp (A\lp B)\right)+\ldots.
1318: \]
1319: In fact, this series expansion may be derived from a closed form
1320: integral expression.
1321:
1322: The Baker--Campbell--Hausdorff formula is of great importance in
1323: both pure and applied mathematics. It gives (where it applies; in
1324: particular in finite dimensions)
1325: the relation of a Lie group with the associated Lie algebra. It for
1326: example says that the product of $e^A$ and $e^B$ for some $A$ and $B$
1327: in the Lie algebra is again an element of the form $e^C$ with $C$ in
1328: the Lie algebra. Hence the exponents of the Lie algebra generate a
1329: subgroup of the corresponding Lie group.
1330:
1331: For infinite-dimensional Lie algebras and groups, one has to
1332: use a refined argument centering around the Hille--Yosida theorem.
1333: Let $U(t)$ denote a \at{merge with linear Lie group part. = motion?}
1334: one-parameter group of linear operators on a Hilbert space $\Hz$
1335: such that $t\to U(t)$ is {\bfi{strongly continuous}}, which
1336: means that $t\to U(t)\varphi$ is continuous for all
1337: $\varphi\in\Hz$. Then we can differentiate $U(t)$ to obtain the
1338: strong limit
1339: \[
1340: A= \lim_{t\rightarrow 0}\frac{U(t)-U(0)}{t}\,.
1341: \]
1342: The object $A$ is called {\bfi{infinitesimal
1343: generator}}\index{generator!infinitesimal} of the
1344: one-parameter group $U(t)$. It turns out that $A$ is a closed linear
1345: operator that is defined on a dense subspace in $\Hz$. The Hille--Yosida
1346: theorem gives a necessary and sufficient condition \at{state these
1347: and give a reference! Define selff-adjoint}
1348: for a closed linear
1349: operator $A$ to be the infinitesimal generator of some strongly
1350: continuous one-parameter semigroup
1351: \[
1352: U(t) = e^{tA} \,,
1353: \]
1354: since in general one might not get a group.
1355: The Hille--Yosida theorem is very useful for analyzing the solvability
1356: of linear differential equations
1357: \[
1358: \frac{d}{dt}\psi(t)= A\psi(t)\,, \ \ \ \psi(0)=\psi_0\,,
1359: \]
1360: examples of which are the Schr\"odinger equation or the heat equation.
1361: If the conditions of the Hille--Yosida theorem hold
1362: for $A$, the solution to this initial value problem takes the form
1363: \[
1364: \psi(t)= e^{tA}\psi(0)\,.
1365: \]
1366: For the (hyperbolic, conservative) Schr\"odinger equation,
1367: $A=-\frac{i}{\hbar}H$ with a self-adjoint Hamiltonian $H$,
1368: the solution exists for all $t$, and the $U(t)$ form a one-parameter
1369: group.
1370: For the (parabolic, dissipative) heat equation, $A=k\Delta$ is a
1371: positive multiple of the Laplacian
1372: $\Delta = \partial_x^2+\partial_y^2+\partial_z^2$, the solution exists
1373: only for
1374: $t\ge 0$, and we only get a semigroup.
1375:
1376:
1377:
1378:
1379: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1380: \section{Lie $*$-algebras}
1381:
1382: Many Lie algebras of interest in physics have an additional structure:
1383: an adjoint mapping compatible with the Lie product.
1384:
1385: \begin{definition}\label{d.liester}
1386: A {\bfi{Lie $*$-algebra}} is a Lie algebra $\Lz$ over $\Cz$
1387: with a distinguished
1388: element $1\ne 0$ called {\bfi{one}} and a mapping
1389: $*$ that assigns to every $f\in\Lz$ an {\bfi{adjoint}} $f^*\in\Lz$
1390: such that
1391: \[
1392: f\lp 1 =0,
1393: \]
1394: \[
1395: (f+g)^*= f^*+g^*,~~~
1396: (f\lp g)^* = f^*\lp g^*,
1397: \]
1398: \[
1399: f^{**}=f,~~~
1400: (\lambda f)^* = \bar \lambda f^*,~~~
1401: 1^*=1
1402: \]
1403: for all $f,g\in \Lz$ and $\lambda\in \Cz$. We identify the multiples
1404: of 1 with the corresponding complex numbers.
1405: \end{definition}
1406:
1407: The reason why we include the 1 into the definition of a Lie
1408: $*$-algebra is that many physically relevant
1409: Lie algebras are equipped with a distinguished central
1410: element\footnote{Many such Lie algebras are realized most naturally
1411: as central extensions of semisimple Lie algebras, corresponding to
1412: projective representations of semisimple Lie algebras.
1413: By including the 1 automatically we work directly in the central
1414: extension, and avoid the cohomological technicalities
1415: associated with the formal discussion of central extensions and
1416: projective representations.}.
1417: But the presence of $1$ is not a restriction, since one can always
1418: adjoin a central element $1$ to a Lie algebra $\Lz'$ without nonzero
1419: central element and form the direct sum $\Lz=\Lz'\oplus \Kz$.
1420:
1421: An important Lie $*$-algebra for nonrelativistic quantum mechanics
1422: is the algebra $\Ez=\Lin\Hz$ of linear operators of a Euclidean space
1423: $\Hz$ (usually a dense subspace of a Hilbert space $\ol \Hz$).
1424: The relevant Lie product is defined by Theorem \ref{ass.lie.J} with
1425: the choice
1426: \[
1427: J:=\frac{i}{\hbar}=\frac{i}{\hbar} 1_\Hz \in\Lin\Hz,
1428: \]
1429: where $1_\Hz$ is the identity operator on $\Hz$, and the conjugate
1430: of $f\in \Ez$ is given by the \bfi{adjoint} of $f$, defined as the
1431: linear mapping $f^*$ satisfying $\phi^* f\psi=(f\phi)^*\psi$ fir all
1432: $\phi,\psi\in\Hz$. Dropping the
1433: index $J$ in the Lie product, we get the {\bfi{quantum Lie product}}
1434: \lbeq{e.liequant}
1435: f\lp g = \frac{i}{\hbar}(fg-gf) =\frac{i}{\hbar}[f,g]
1436: \eeq
1437: of $f,g\in\Lin \Hz$, already familiar from \gzit{e.lp}.
1438: Note that the axioms require the purely imaginary factor in this
1439: formula, whereas the value of Planck's constant $\hbar$ is arbitrary
1440: from a purely mathematical point of view.
1441: In quantum field theory, a different choice of $J$
1442: is sometimes more appropriate.
1443:
1444: For any Lie $*$-algebra, the set
1445: \[
1446: \re \Lz :=\{f\in\Lz\mid f^*=f\}
1447: \]
1448: is a Lie algebra over $\Rz$. When describing symmetries, physicists
1449: often work with Lie algebras over the reals; the present Lie
1450: $*$-algebras are then the complexifications of these real algebras,
1451: with a central element 1 adjoined if necessary.
1452:
1453: The {\bfi{complexification}} of a real Lie algebra $\Lz$ is the
1454: Lie $*$-algebra \idx{$\Cz\Lz$} defined as follows. In case that
1455: a complex scalar multiplication is already defined on $\Lz$, one first
1456: replaces $\Lz$ by an isomorphic Lie algebra in which $if\not\in \Lz$
1457: if $f\in\Lz$ is nonzero. Then one defines
1458: \[
1459: \Cz\Lz=\Lz\oplus i\Lz,
1460: \]
1461: extending scalar multiplication in a natural way to the complex field.
1462: That is, any element $f\in\Cz\Lz$ is of the form
1463: \[
1464: f= f_1 + i f_2
1465: \]
1466: with $f_1,f_2\in\Lz$, and one defines
1467: \[
1468: \alpha(\beta f):=(\alpha\beta)f,~~~
1469: \alpha f\lp \beta g:= (\alpha\beta)f\lp g
1470: \]
1471: for all $f,g\in \Lz$ and $\alpha,\beta\in\Cz$. Conjugation is
1472: defined as
1473: \[
1474: (f_1 + i f_2)^*:=f_1 - i f_2\for f_1,f_2\in\Lz;
1475: \]
1476: The axioms for a Lie $*$-algebra are easily established if $1\in\Lz$.
1477: Note that the real dimension of $\Lz$ equals the complex dimension of
1478: $\Cz\Lz$. It is easy to check that
1479: \[
1480: \re \Cz\Lz \cong \Lz.
1481: \]
1482: Conversely, for a Lie $*$-algebra $\Lz$,
1483: \[
1484: \Cz \re \Lz \cong \Lz.
1485: \]
1486: If a complex Lie algebra $\Lz'$ is isomorphic to $\Cz\Lz$ as a Lie
1487: algebra, one says that $\Lz$ is a {\bfi{real form}} of the complex
1488: Lie algebra $\Lz'$.
1489:
1490: We leave it as an exercise to verify $\Cz su(n)\cong
1491: sl(n,\Cz)$ and $\Cz so(p,q) =so(p+q,\Cz)$.
1492: \at{describe more generally all classical $*$-algebras!}
1493: In general, a complex Lie
1494: algebra has more than one real form as we can see since for $p\neq
1495: q,n-q$ the Lie algebras $so(p,n-p)$ and $so(q,n-q)$ are not isomorphic.
1496:
1497: An {\bfi{involutive Lie algebra}}\index{Lie algebra!involutive}
1498: (\sca{Neeb}
1499: \cite{neeb}) is a Lie algebra
1500: $\tilde \Lz$ with Lie product $[\cdot,\cdot]$ and with an involutive,
1501: antilinear anti-automorphism $\sigma$, i.e., a mapping
1502: $\sigma:\tilde \Lz\to\tilde \Lz$ satisfying
1503: \[
1504: \sigma(\alpha f) = \alpha^* \sigma f,~~~
1505: \sigma(f\lp g) = \sigma g \lp \sigma f
1506: \]
1507: for $\alpha\in\Cz,f,g\in\tilde \Lz$.
1508: Associated to an involutive Lie algebra $\tilde \Lz$ is
1509: the \bfi{real form} $\tilde \Lz_\Rz=\Big\{ x\in \tilde \Lz\mid \sigma x
1510: = -x\Big\}$. Our definition of a Lie $*$-algebra is closely
1511: related and obtained as follows, after adjoining to $\tilde \Lz$
1512: a central element 1 if necessary. We define $\Lz$ as the vector space
1513: $\tilde \Lz$ equipped with the Lie product $\lp$ defined by
1514: $x\lp y = \frac{i}{\hbar} [x,y]$.
1515: Then the mapping $x\mapsto i\hbar x$, with $\hbar$ a positive
1516: real constant (in physical applications Planck's constant) is an
1517: isomorphism of Lie algebras. The map $\sigma$ induces the
1518: conjugation $x^* = -\sigma x$ and $\re \Lz = \tilde \Lz_\Rz$.
1519:
1520: \begin{rems}
1521: ~\at{give here the example of $so(3)$}\\
1522: (i) The nomenclature of Lie $*$-algebras is a bit tricky. If $\Lz$ is a
1523: Lie $*$-algebra, we therefore denote it (usually) with the name of the
1524: real Lie algebra $\re \Lz$. To avoid confusion,
1525: it is important to keep track of whether we are discussing real Lie
1526: algebras, complex Lie algebras or Lie $*$-algebras.
1527:
1528: (ii) In the physics literature, one often sees the defining relations
1529: \gzit{phys.struct} for real Lie algebras written in terms of complex
1530: structure constants,
1531: \[
1532: X_j \lp X_k =\sum_l ic_{jkl}X_l\,.
1533: \]
1534: where $i=\sqrt{-1}$ and the $c_{jkl}$ are real. That is, the Lie
1535: product takes values outside of the real Lie algebra!
1536: What is done by the physicists is that -- as in the above definition
1537: of a Lie $*$-algebra from an involutive Lie algebra -- they
1538: multiply all elements in the Lie algebra by $i$. The reasons for
1539: making this seemingly difficult construction mainly has historical
1540: reasons. One is that in some real algebras the elements are
1541: antihermitian matrices. By multiplying with $i$ one obtains
1542: Hermitian matrices and operators in quantum mechanics are represented
1543: as Hermitian operators.
1544: \end{rems}
1545:
1546: The converse process of complexification is {\bfi{realization}}.
1547: Given a complex Lie algebra $\Lz$, one regards it as a real Lie
1548: algebra \idx{$\Lz^\Rz$} by restricting scalar multiplication to
1549: real factors. Since $f$ and the imaginary scalar multiple $if$ are
1550: linearly independent over $\Rz$,
1551: the real dimension of $\Lz^\Rz$ is twice the complex dimension of $\Lz$.
1552: In the finite-dimensional case, a convenient way to obtain the
1553: realization is as follows:
1554: Choose a basis $t_1,\ldots,t_n$ of $\Lz$ and then form
1555: the elements $s_j=it_j$ for all $j$. All real linear combinations of
1556: $s_j$ and $t_j$ make up $\Lz^\Rz$. Given two elements $f,g$ in
1557: $\Lz^\Rz$ one calculates their Lie product as if they were elements
1558: of $\Lz$; the result can be written as
1559: \[
1560: f\lp g =\sum (\alpha_j +i\beta_j)t_j\,.
1561: \]
1562: The Lie product of $f$ and $g$ in $\Lz$ is then defined as
1563: \[
1564: f\lp g =\sum \alpha_jt_j +i \sum\beta_j s_j\,.
1565: \]
1566: See also Example \ref{ex.real.sl}. \at{\and/or the $so(3)$ example}
1567:
1568:
1569: