1: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2: \chapter{Fields, forms, and derivatives}\label{c.manifolds}
3:
4:
5: \at{adapt this introduction, and also the summary in Section 1.5
6: of QML}
7:
8: In this chapter we introduce basic material on manifolds,
9: the associated commutative algebra of scalar fields, and the Lie
10: algebra of vector fields. All manifolds
11: used in this book are arbitrarily often differentiable, real manifolds
12: whose dimension need not be finite. However, we are very brief and
13: sometimes incomplete in the technical details that need attention in
14: the infinite-dimensional case; on first reading, the reader may restrict
15: everything to the finite-dimensional case, where these details are not
16: required.
17:
18:
19:
20: We first recall some basics from differential geometry.
21: Our approach differs from standard introductions to differential
22: geometry since, consistent with the theme of the book, all definitions
23: are given in an algebraic way. As a side benefit, this prepares the
24: reader to noncommutative geometry, only briefly touched in this book,
25: where a manifold structure is no longer available and all geometry
26: enters in an algebraic way.
27: Among other applications, noncommutative geometry gives an interesting
28: geometric perspective to the quantum field theory of the standard model.
29: \at{ref Connes}
30:
31: Vector fields on a manifold $\Mz $ are essentially equivalent to
32: derivations on the commutative algebra $C^\infty(\Mz )$ of scalar
33: fields. However, to be able to use the traditional terminology,
34: where vector fields and the corresponding derivations (Lie derivatives)
35: are distinguished, we introduce an abstract set $\Wz=\vect \Mz $
36: of vector fields, whose elements are put into correspondence with
37: derivations by means of a mapping $d:\vect \Mz \to \der \Mz $ which is
38: applied at the right. In this way, the calculus on manifolds can be
39: formulated in a purely algebraic way, without any reference to the
40: manifold.
41:
42: We therefore formulate everything in terms of an arbitrary topological
43: commutative algebra $\Ez$ in place of $C^\infty(\Mz )$, and an
44: arbitrary set $\Wz$ in place of $\vect \Mz $.
45: However, the main situation that the reader should have in mind
46: is where $\Ez$ is an algebra of complex-valued, arbitrarily often
47: differentiable functions on a finite-dimensional manifold,
48: for example $C^\infty(\Rz^n)$. But $\Ez$ could also be the Schwartz
49: space of arbitrarily often differentiable functions all of whose
50: derivatives decay faster than polynomially at infinity.
51:
52: As a result, our presentation is completely coordinate-free, except in
53: some examples. For readers accustomed to differential geometry in
54: index notation but not to the coordinate-free Cartan notation,
55: we suggest that they translate the definitions and main results into
56: coordinates to understand their meaning, but to treat proofs as if the
57: concepts introduce new abstract algebraic notions.
58:
59:
60:
61:
62:
63: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
64: \section{Scalar fields and vector fields}\label{s.scalar}
65:
66: We introduce the objects, operators, and
67: operations needed for presenting the traditional differential calculus
68: in a purely algebraic framework: Lie derivatives applied to multilinear
69: forms, and exterior products and the exterior derivative of alternating
70: forms. As the most important special case, we consider manifolds and
71: associated geometric notions, in particular diffeomorphisms.
72:
73: Before giving the definitions, we discuss the letter conventions
74: and priority rules used in the formulas.
75:
76: We typically (i.e., when not forced by conflicts or tradition to do
77: otherwise) use lower case letters from the middle of the alphabet, such
78: as $f,g,h$, to denote scalar fields, capital letters from the end of
79: the alphabet, such as $X,Y,Z$ to denote vector fields, capital letters
80: from the beginning of the alphabet, such as $A$, for general
81: multilinear forms, but $z$, for linear forms, $\omega$ for alternating
82: bilinear forms, and $\eta$ for symmetric bilinear forms.
83:
84: We use the convention that a Lie derivative acts on the shortest
85: following expression which is syntactically a vector field or a
86: multilinear form. Similarly, the exterior derivative operator $d$ acts
87: either on the right on a vector field, or on the left
88: on the shortest following expression that is syntactically an
89: alternating form. The wedge product $\wedge$ has lower priority than
90: the operations written as juxtaposition, but higher priority than $+$
91: and $-$.
92:
93: %%%%%%%%%%%%%%%%%%%%%%%%%%%
94: \begin{dfn}~\\
95: (i) A \bfi{differential geometry} consists of a commutative algebra
96: $\Ez$ containing $\Cz$, a left $\Ez$-module $\Wz$ with an additional
97: Lie product $\lp$, both equipped with a topology such that all
98: operations are continuous,
99: and a continuous mapping $d$ (written on the right),
100: which maps $X\in\Wz$ to $Xd\in \der \Ez$, such that
101: \lbeq{e.dgeom}
102: (X+Y)d= Xd+Yd,~~~ (f X)d=f(Xd),~~~(X\lp Y)d = [Xd,Yd],
103: \eeq
104: for all $X,Y\in \Wz,f\in\Ez$.
105: The differential geometry is called \bfi{(non-)commutative} if
106: the multiplication in $\Ez$ is (non-)commutative.
107:
108: (ii) We refer to the elements of $\Ez$ as \bfi{scalar fields}, and to
109: the elements of $\Wz$ as \bfi{vector fields}.
110: The \bfi{Lie derivative} of a vector field $X$ is the linear
111: mapping $L_X$ which maps a scalar field $ f$ to\footnote{
112: As will become apparent in Section \ref{s.ext} (cf. Theorem \ref{t6.2}),
113: we may read the term
114: $Xd~f$ also as product of the vector field $X$ with the
115: exact linear form $df$. Until then, we shall write an explicit space
116: after $d$ to remind the reader of the correct way to group the
117: letters.
118: } % end footnote
119: \lbeq{e.ld6s}
120: L_X f:=Xd~ f,
121: \eeq
122: and a vector field $Y$ to
123: \lbeq{e.ld6}
124: L_XY ~:= ~ X \lp Y ~=~ -L_Y X ~.
125: \eeq
126: The scalar field $L_X f$ (resp. the vector field $L_XY$)
127: is called the \bfi{directional derivative} of the scalar field $ f$
128: (resp. the vector field $Y$) in the direction of the vector field $X$.
129: \end{dfn}
130:
131: %%%%%%%%%%%%%%%%%%%%%%%%%%
132: \begin{expl}
133: \bfi{(Canonical differential geometries)}\\
134: Let $\Ez$ be an arbitrary topological algebra containing $\Cz$.
135: We may give
136: $\Ez$ the structure of a differential geometry by picking an arbitrary
137: set $\Wz$ with the same cardinality as $\der \Ez$, and choosing an
138: arbitrary bijection $d$ from $\Wz$ to $\der \Ez$.
139: $\Wz$ inherits all properties of $\der\Ez$ by means of
140: the bijection $d$: We turn $\Wz$ into a Lie algebra and a
141: topological $\Ez$-module by defining
142: \[
143: X+Y:= (Xd+Yd)d^{-1},~~~ f X:=( f(Xd))d^{-1}\,,
144: \]
145: \[
146: X\lp Y := [Xd,Yd]d^{-1}.
147: \]
148: for $X,Y\in\Wz$ and $f\in\Ez$, and by calling a set $S\in\Wz$
149: closed if its image under $d$ is closed in the
150: topology of $\der \Ez$ induced by that of $\Ez$.
151: The result is a differential geometry.
152: We call differential geometries constructed in this way
153: \bfi{canonical}.
154: \end{expl}
155:
156: The following example is responsible for the naming.
157: Interpreting the set $\Mz $ in the example as the
158: domain of a chart of a finite-dimensional manifold, one can translate
159: everything said here to general finite-dimensional manifolds by a
160: process described in all books on differential geometry.
161: Thus the example gives essentially the full intuition for our
162: constructions, except for the complications that may arise in
163: infinite dimensions.
164:
165: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%
166: \begin{expl}\label{ex.Rn}
167: \bfi{(Differential geometry of open subsets in $\Rz^{\times n}$)}\\
168: Let $\Rz^{\times n}$\index{$\Rz^{\times n}$} denote the vector space
169: of row vectors\footnote{
170: The index notation corresponds to standard differential
171: geometric practice when working in a chart of a manifold (which is
172: essentially the situation we are discussing here).
173: The interpretation in terms of rows (row vectors = \bfi{rovectors},
174: indexed by upper indices = \bfi{roindices}) and columns
175: (column vectors = \bfi{covectors}, indexed by lower indices =
176: \bfi{coindices}) makes the transition to
177: standard linear algebra transparent.
178: } % end footnote
179: $x=(x^1,\dots,x^n)$ with $n$ real components $x^j$,
180: let $\Mz $ be a nonempty, open subset of $\Rz^{\times n}$, and let
181: $\Ez=C^\infty(\Mz )$ and $\Wz=C^\infty(\Mz ,\Cz^{\times n})$, equipped with
182: the weak topology. \at{define}
183: Thus scalar fields are real-valued functions, while vector fields are
184: row vector valued functions. In terms of the partial differential
185: operators $\partial_j$ defined by
186: \[
187: \partial_jf(x):=\partial f(x)/\partial x^j,
188: \]
189: we define the \bfi{gradient} $\partial f$ of a scalar field as the
190: column vector
191: with $n$ entries
192: \[
193: (\partial f)_j=\partial_j f.
194: \]
195: It is not difficult to show that a derivation on the algebra of
196: scalar fields can be uniquely expressed as a linear partial differential
197: operator of the form
198: \[
199: \delta = X\partial = \sum_{j=1}^n X^j \partial_j,
200: \]
201: with a vector field $X$. Such a $\delta$ acts on scalar fields $f$ as
202: \[
203: \delta f = X\partial f = \sum_{j=1}^n X^j \partial_j f.
204: \]
205: Thus the mapping $d$ which maps the vector field $X$ to the
206: differential operator $X\partial$ is a bijection of the type required
207: in the previous example. Thus we have a canonical differential geometry;
208: it is clearly commutative.
209: The reader is invited to check that the Lie derivative takes the form
210: \lbeq{e.ldc}
211: L_X f = X \partial f,~~~L_X Y = X\partial Y - Y \partial X.
212: \eeq
213: A second, noncanonical differential geometry results by using in the
214: above construction in place of $C^\infty(\Mz )$ the subalgebra
215: \idx{$C^\infty_0(\Mz )$} of scalar fields with compact support,
216: and in place of $C^\infty(\Mz ,\Cz^{\times n})$
217: the subspace \idx{$C^\infty_0(\Mz ,\Cz^{\times n})$} of vector fields with
218: compact support.
219: \end{expl}
220:
221:
222:
223: \begin{prop}
224: The \bfi{product rule}
225: \lbeq{e.ld3v}
226: L_X( f g)=(L_X f )g+ f (L_Xg),~~~
227: L_X( f Y)=(L_X f )Y+ f (L_XY),
228: \eeq
229: the \bfi{commutation rule}
230: \lbeq{e.ld4v}
231: [L_X \,, L_Y]=L_{X\!\lp Y},
232: \eeq
233: and the equations
234: \lbeq{e.ld27a}
235: L_{ f X}\, g = f L_X \, g,
236: \eeq
237: \lbeq{e.ld27}
238: L_{ f X}Y= f L_XY - XYd~f
239: \eeq
240: hold for $ f,g \in\Ez$ and $X,Y\in\Wz$,
241: \end{prop}
242:
243: \bepf
244: The first part of \gzit{e.ld3v} is trivial since in this case
245: $L_X = Xd$ is a derivation on $\Ez$. For the second part of
246: \gzit{e.ld3v}, we note that
247: \[
248: \bary{rcl}
249: \Big(L_X(f Y)\Big)d\;g &=& (X \lp fY)d\;g ~=~ [Xd \,, fYd\,] g
250: ~=~ Xd\Big(f(Yd~g)\Big) - fYd\;(Xd~g) \\
251: &=& (Xd~f)(Yd~g) + f Xd\;(Yd~g) -fYd\;(Xd~g) ~.
252: \eary
253: \]
254: Also,
255: \[
256: ((L_X f)Y)d~g ~=~ ((Xd~f)Y)d~g ~=~ (Xd~f)(Yd~g)
257: \]
258: and
259: \[
260: \bary{rcl}
261: \Big(f (L_XY)\Big)d~g &=& \Big(f (X \lp Y)\Big)d~g ~=~ f[Xd \,, Yd]g \\
262: &=& f\Big( Xd\;(Yd~g) - Yd\;(Xd~g) \Big) ~=~ fXd\;(Yd~g) - fYd\;(Xd~g)
263: \eary
264: \]
265: Putting these three pieces together proves the second part of
266: \gzit{e.ld3v}.
267:
268: To prove \gzit{e.ld4v}, note that $L_Y f = Yd~f$ is in $\Ez$,
269: so $L_X L_Y f = L_X(Yd~f) = Xd\;(Yd~f)$. Interchanging $X,Y$ we find
270: for the commutator:
271: \[
272: [L_X , L_Y]f ~=~ Xd~(Yd~f) - Yd~(Xd~f) ~=~ [Xd, Yd]f ~=~ (X\lp Y)d~f
273: ~=~ L_{X\slp Y} \, f ~.
274: \]
275: Formula \gzit{e.ld27a} is immediate from the definition, and
276: \gzit{e.ld27} follows from the product rule
277: $X\lp f Y = (Xd~ f )Y+ f (X\lp Y)$ by swapping $X$ and $Y$, using the
278: anticommutativity of $\lp$.
279: \epf
280:
281:
282: In the following, we develop the differential calculus for
283: commutative differential geometries only; thus, with exception of the
284: remarks on noncommutative geometry in Section \ref{s.mani},
285: \bfi{the algebra $\Ez$ of scalar fields is always assumed to be
286: commutative.} In this case, we extend the left module structure on
287: vector fields to a bimodule structure by putting
288: \[
289: Xf:=fX
290: \]
291: for $f\in\Ez$ and $X\in\Wz$.
292: Note that some authors treat vector fields as synonymous with
293: derivations and therefore write $X(f)$ for $Xd~f$. This should not
294: be confused with the present notation $Xf$ for multiplying the vector
295: field $X$ with the scalar field $f$.
296:
297:
298: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
299: \section{Multilinear forms}\label{s.mforms}
300:
301: Apart from scalar and vector fields, differential geometry makes
302: heavy use of multilinear forms and tensors, which we define
303: next.
304:
305: \begin{dfns}~\\
306: (i) A \bfi{linear form} $\zeta$ is a continuous, $\Ez$-linear mapping
307: $\zeta:\Wz\to\Ez$ (written on the right\footnote{Strictly speaking,
308: they should be called $\Ez$-linear forms, and a similar remark applies
309: later to multilinear forms. Talking about a {\em form} rather than a
310: mapping implies the assumption of continuity.\\
311: The standard notation for $X\zeta$ is $i_X\zeta=\zeta(X)$; the present
312: notation simply replaces $i_X$ by $X$. This way of writing the mapping
313: generalizes standard matrix calculus if we use the intuition gained
314: from Example \ref{ex.Rn} and think of vector fields as row vectors and
315: of linear forms as column vectors, an intuition that extends to matrix
316: fields. Since in the general situation, linear forms are often called
317: \bfi{covectors}, we shall occasionally use the analogous word {\bf
318: rovector} to denote a vector field, although, strictly speaking, one
319: should talk about covector fields and rovector fields. The same
320: ambiguity is traditionally maintained for multilinear forms on manifolds,
321: which refer both to the corresponding fields and to their values at a
322: particular point.
323: }\label{fn.iX}) % end footnote
324: which maps the vector field $X$ to the scalar field $X\zeta$. We
325: write $\Wz^*$ for the $\Ez$-module consisting of all linear forms,
326: (sometimes called the "$\Ez$-dual" of $\Wz$), with scalar
327: multiplication of $\zeta\in\Wz$ by $f\in\Ez$ defined via
328: \[
329: X( f \zeta):= f(X\zeta).
330: \]
331: (ii) A \bfi{$c$-linear} form $\phi$ is a mapping
332: $\phi : \Wz\times\dots\times\Wz \to \Ez$ (with $c$ factors of $\Wz$
333: in the Cartesian product) such that the image\footnote{
334: The traditional notation for $X_1\dots X_c \phi$ is
335: $i_{X_1}\dots i_{ X_c}\phi = \phi(X_c,\dots,X_1)$; as for linear forms,
336: the present notation simply replaces the $i_X$ by $X$.
337: Note the reverse order resulting in the arguments written in the
338: traditional way, needed
339: in order that \gzit{e.XA} together with our definition \gzit{e.ld8}
340: of insertion is consistent with the traditional definition
341: $(i_X\phi)(X_1,\dots,X_{c-1})=\phi(X,X_1,\dots,X_{c-1})$. In our
342: notation translates into
343: $X_{c-1}\dots X_1(i_X\phi)=X_{c-1}\dots X_1X\phi$.
344: } % end footnote
345: $X_1\dots X_c \phi$ of $(X_c,\dots,X_1)\in \Wz\times\dots\times\Wz$
346: depends $\Ez$-linearly on each argument $X_k$, i.e., if, for all
347: $X_j,Y,Z\in\Wz$ and $ f, g \in\Ez$,
348: \[
349: X_1\dots ( f Y+ g Z)\dots X_c \phi
350: ~=~ f X_1\dots Y\dots X_c \phi
351: ~+~ g X_1\dots Z\dots X_c \phi ~.
352: \]
353: Here the unindexed argument between the dots replaces the $k$th
354: argument $X_k$, for some $k$ in $1,\dots,c$.
355: In the degenerate case $c=0$, we consider the $0$-linear mappings
356: to be the scalar fields.
357:
358: (iii) We write\footnote{
359: Writing the ``$c$" in $\Wz_c$ as a subscript serves
360: as a reminder that when an element $\phi$ of $\Wz_3$ (say) is written
361: in index notation, it has 3 {\em lower} indices: ``$\phi_{ijk}$".
362: Indeed the ``$c$" is intended to be suggestive of ``covector". Note that
363: in terms of the direct products of $c$ factors $\Wz$ or $\Wz^*$,
364: there is a canonical isomorphism $\Wz^*\times\dots\times\Wz^*
365: \cong (\Wz\times\dots\times\Wz)^* =\Wz_c^*$.
366: } % end footnote.
367: $\Wz_c$ for the $\Ez$-module of continuous $c$-linear
368: mappings on $\Wz$. Scalar multiplication of $\phi\in\Wz_c$ by
369: $f\in\Ez$ is defined via
370: \[
371: X_1\dots X_c( f \phi) ~:=~ f(X_1\dots X_c \phi) ~.
372: \]
373: The elements of $\Wz_c$ are called \bfi{multilinear
374: forms}\MMt{\footnote{
375: Multilinear forms correspond to what {\sc
376: Penrose \& Rindler} \cite[Chapter 2]{PenRI} call a {\em type-I tensor}.
377: } % end footnote
378: } % end MMt
379: or \bfi{$c$-linear forms}; for $c=2$ also \bfi{bilinear forms}.
380: Note that $\Wz_0=\Ez$ consists of scalar fields (or 0-forms), and
381: $\Wz_1=\Wz^*$ consists of linear forms (or 1-forms).
382:
383: (iv) The \bfi{product} of a vector field $X\in\Wz$ and a $c$-linear
384: form $\phi\in\Wz_c$ is for $c=0$ the vector field $X \phi$ defined by
385: scalar multiplication with the scalar $\phi$, and for $c>0$ the
386: $(c-1)$-linear form $X \phi$ defined by
387: \lbeq{e.XA}
388: X_1\dots X_{c-1}(X \phi) ~:=~ X_1\dots X_{c-1}X \phi
389: ~,~\Forall X_1,\dots,X_{c-1}\in\Wz.
390: \eeq
391: The operator $i_X$ defined on multilinear forms $\phi$ by
392: \lbeq{e.ld8}
393: i_X\phi ~:=~ X\phi
394: \eeq
395: is traditionally called the \bfi{insertion} of $X\in\Wz$; cf.
396: footnote \ref{fn.iX}. \at{footnote reference comes out wrong}
397:
398: (v) A $c$-linear form $\phi$ is called \bfi{alternating} (or
399: a $c${\bf-form}) if either $c\le 1$ or $X\phi$ is alternating
400: and $XX\phi=0$ for all vector fields $X$.
401: $\phi$ is called \bfi{symmetric} if either $c\le 1$ or $X\phi$
402: is symmetric and $XY\phi=YX\phi$ for all vector fields $X,Y$.
403: We write $\Az_c$ and $\Sz_c$ for the space of alternating and
404: symmetric $c$-linear forms, respectively. In particular,
405: \[
406: \Ez_0=\Wz_0=\Ez,~~~ \Ez_1=\Wz_1,~~~ \Ez_2\oplus \Sz_2 = \Wz_2 ~.
407: \]
408: (vi) The \bfi{transpose} of a bilinear form $\phi$ is the bilinear form
409: $\phi^T$ defined by
410: \lbeq{e.transpose}
411: XY\phi^T:=YX\phi
412: \eeq
413: for all vector fields $X,Y$.
414: In particular, a bilinear form $\phi$ is symmetric iff $\phi^T=\phi$ and
415: alternating iff $\phi^T=-\phi$.
416: A bilinear form $ \phi$ is called \bfi{nondegenerate} if every
417: linear form $\zeta\in\Wz^*$ can be written as $\zeta=X\phi$ for a unique
418: vector field $X$; otherwise \bfi{degenerate}.
419: A bilinear form $\phi$ may be considered as a linear
420: mapping from $\Wz$ to $\Wz^*$ that maps the vector field $X$ to the
421: linear form $X\phi$. If $\phi$ is nondegenerate, this mapping is
422: invertible, and the inverse $\phi^{-1}$ is a linear
423: mapping from $\Wz^*$ to $\Wz$, which maps a linear form $\zeta$
424: to the vector field $\zeta\phi^{-1}$ in such a way that
425: \lbeq{e.Ainv}
426: \zeta\phi^{-1}\phi=\zeta.
427: \eeq
428: A nondegenerate bilinear form is called a \bfi{symplectic form}
429: if it is alternating, and a \bfi{metric} if it is symmetric.
430:
431: (vii) A $[c,r]${\bf-tensor field}\footnote{
432: An $[c,r]$-tensor is also called a tensor of $[{r \atop c}]$-valence
433: (\sca{Penrose \& Rindler} \cite{PenRI}).
434: With traditional index notation, a $c$-linear form is written with
435: $c$ lower (co)indices, and
436: a $[c,r]$-tensor is written with $r$ upper (ro)indices and
437: $c$ lower (co)indices. E.g., a $[3,2]$-tensor $T$ is written
438: ${T_{ijk}}^{mn}$, and the image $T\phi$ of a bilinear form is written
439: $(T\phi)_{ijk}={T_{ijk}}^{mn}\phi_{mn}$, using the traditional {\bf
440: Einstein summation convention} (which deletes the explicit indication
441: of the sum over $m$ and $n$ so as not to unnecessarily inflate the
442: formulas without conveying any more information). In the more modern
443: \bfi{abstract index notation} of \sca{Penrose \& Rindler} \cite{PenRI},
444: such repeated indices
445: denote instead an insertion (dual-pairing) {\em without} any implied
446: connotation of summation over basis-dependent components, and such
447: indices may be used to keep explicit track of the types of complicated
448: objects.
449: } % end footnote
450: $T$ is a continuous $\Ez$-linear mapping $T:\Wz_r \to \Wz_c$.
451: The space of $[c,r]$-tensor fields
452: is denoted by\footnote{
453: For the differential geometry of open subsets $\Mz $ of $\Rz^n$
454: (Example \ref{ex.Rn}), $\Wz[c,r] =\Wz_c^r$ is, in the traditional
455: terminology, the space of sections of the tensor bundle $T_c^r\Mz $.
456: } % end footnote
457: $\Wz[c,r] =\Wz_c^r=\Lin(\Wz_r,\Wz_c)$. A $[1,1]$-tensor
458: field is called a \bfi{matrix field}.
459:
460: \end{dfns}
461:
462: \begin{rems}
463: (i) Multilinearity implies
464: \[
465: XY\omega=-YX\omega
466: \mbox{~~~for alternating $c$-forms $\omega$ with $c>1$}.
467: \]
468: Thus, for $c\ge 2$,
469: \lbeq{e.ld10b}
470: i_X^2=0,~~i_Xi_Y =-i_Yi_X \mbox{~~~on alternating $c$-linear forms},
471: \eeq
472: whereas
473: \lbeq{e.ld10c}
474: i_Xi_Y =i_Yi_X \mbox{~~~on symmetric $c$-linear forms}.
475: \eeq
476: (ii) Note that there is a canonical identification
477: of $\Wz[c,0]$ with $\Wz_c$, and a canonical embedding of $\Wz$ into
478: $\Wz[0,1]$. In the case of finite-dimensional manifolds, we may also
479: identify $\Wz[0,1]$ with $\Wz$.
480:
481: (iii) The ordinary operator product of a $[c',c]$-tensor and an
482: $[c,r]$-tensor is well-defined, and is a $[c',r]$-tensor:
483: $\Wz[c',c]\Wz[c,s]\subseteq\Wz[c',s]$.
484: In particular, $\Wz[1,1]=\Lin \Wz_1$ is an algebra of matrix fields.
485: \end{rems}
486:
487: %%%%%%%%%%%%%%%%%%%%%%%%%%%
488: \begin{thm}\label{t.GenLieDeriv}
489: For every vector field $X$, the Lie derivative can be extended uniquely
490: to a linear operator $L_X$ mapping\footnote{
491: The Lie derivative can also be extended to tensors $T\in\Wz[c,r]$ by
492: defining
493: \[
494: (L_XT)B:=L_X(TB)-TL_XB \for B \in \Wz_r.
495: \]
496: We do not need such an extension for our limited
497: applications; it would be needed, however, in a treatment of general
498: relativity. The reader is invited to verify that $L_XT\in\Wz[c,r]$
499: and to formulate and prove the analogues to \gzit{e.ld3} and
500: \gzit{e.ld4}.
501: } % end footnote
502: vector fields to vector fields and $c$-linear forms to $c$-linear
503: forms, and satisfying the \bfi{product rule}
504: \lbeq{e.ld3}
505: L_X( f \phi)=(L_X f )\phi+ f (L_X\phi),~~~
506: L_X(Y\phi)=(L_XY)\phi+Y(L_X\phi).
507: \eeq
508: for $ f \in\Ez$, $Y\in\Wz$, and $\phi\in \Wz$ or $\phi\in \Wz_c$ .
509: The extended Lie derivative satisfies the commutation rules
510: \lbeq{e.ld4}
511: [L_X,L_Y]=L_{X\lp Y},
512: \eeq
513: \lbeq{e.ld9}
514: [L_X,i_Y]=i_{X\lp Y}
515: \eeq
516: for $X,Y\in\Wz$.
517: \end{thm}
518:
519: \bepf
520: We first assume that the product rule holds, and show that this
521: fixes the operation of $L_X$ on all multinear forms.
522: By the product rule \gzit{e.ld3},
523: \lbeq{e.ld21}
524: YL_X\phi = L_X(Y\phi)-(L_XY)\phi,
525: \eeq
526: This formula shows that $L_X$ is determined on $c$-linear forms
527: by its action on $(c-1)$-linear forms, and since it is given on
528: scalar fields, it is unique if it exists at all.
529:
530: Conversely, to show existence of the extension, we define $L_X$
531: recursively by \gzit{e.ld21},
532: starting with the known action of $L_X$ on scalar fields.
533:
534: Since
535: $( f Y)L_X\phi = L_X( f Y\phi)-L_X( f Y)\phi
536: = (L_X f )Y\phi+ f L_X(Y\phi)-(L_X f )Y\phi- f (L_XY)\phi
537: = f (L_X(Y\phi) - f (L_XY)\phi
538: = f (YL_X\phi)$, we see inductively that $YL_X\phi$ is
539: $\Ez$-linear in $Y$, so that $L_X\phi$ is indeed a tensor.
540:
541: The first part of the product rule holds since, by \gzit{e.ld21},
542: the equation
543: $YL_X(f\phi)=L_X(Yf\phi)-(L_XY)(f\phi)=L_X(fY\phi)-(L_XY)(f\phi)
544: = (L_Xf)Y\phi+fL_X(Y\phi)-(L_XY)f)\phi-f(L_XY)\phi
545: =Y(L_Xf)\phi+Yf(L_X\phi)$
546: holds for all vector fields $Y$.
547: The second part of the product rule follows directly from \gzit{e.ld21}.
548:
549: To prove \gzit{e.ld4}, we first note that by \gzit{e.ld21}, we have
550: $ZL_XL_Y\phi = L_X(ZL_Y\phi)-(L_XZ)L_Y\phi
551: =L_X(L_Y(Z\phi)-(L_YZ)\phi) - (L_XZ)L_Y\phi
552: =L_XL_Y(Z\phi) - (L_XL_YZ)\phi - (L_YZ)L_X\phi - (L_XZ)L_Y\phi$.
553: The last two terms are symmetric in $X,Y$, hence cancel when taking
554: the difference with $ZL_YL_X\phi$ in
555: $Z[L_X,L_Y]\phi=ZL_XL_Y\phi-ZL_YL_X\phi
556: =(L_XL_Y-L_YL_X)(Z\phi) - (L_XL_YZ-L_YL_XZ)\phi
557: =[L_X,L_Y](Z\phi)-([L_X,L_Y]Z)\phi$.
558: Since \gzit{e.ld4} is already known by \gzit{e.ld4v} to hold on vector
559: fields and on scalar fields (0-linear forms),
560: we assume that we know its validity for the action on $c$-linear forms.
561: Taking for $\phi$ a $(c+1)$-linear form, we may conclude that
562: $Z[L_X,L_Y]\phi=L_{X\lp Y}(Z\phi)-(L_{X\lp Y}Z)\phi = ZL_{X\lp Y}\phi$
563: by \gzit{e.ld21}. Since $Z$ was arbitrary, we conclude that
564: $[L_X,L_Y]\phi=L_{X\lp Y}\phi$ for $(c+1)$-linear forms $\phi$.
565: By induction, \gzit{e.ld4} holds in general.
566:
567: \gzit{e.ld9} follows from the product rule \gzit{e.ld3} since
568: \[
569: [L_X,i_Y]\phi=L_X(Y\phi)-Y(L_X\phi)=(L_XY)\phi=(X\lp Y)\phi
570: =i_{X\lp Y}\phi.
571: \]
572: \end{proof}
573:
574: The reader may wish to prove inductively that, for $c'$-linear forms
575: $\phi$ with $c'\ge c$,
576: \[
577: X_1\dots X_cL_X\phi
578: = L_X(X_1\dots X_c\phi)-\sum_{k=1}^c X_1\dots X\lp X_k\dots X_c\phi.
579: \]
580:
581: \begin{prop}
582: The Lie derivative of an alternating $c$-form is again an alternating
583: $c$-form.
584: \end{prop}
585:
586: \bepf
587: Indeed, for any alternating form $\omega$,
588: \[
589: \bary{rcl}
590: YYL_X\omega
591: &=& Y(L_X(Y\omega)-(L_XY)\omega) ~=~ Y L_X(Y\omega)-Y(L_XY)\omega \\
592: &=& Y L_X(Y\omega)+(L_XY)Y\omega = L_X(YY\omega) = 0 ~.
593: \eary
594: \]
595: \at{proof that $YL_X\omega$ is alternating is missing.}
596: \epf
597:
598:
599: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
600: \section{Exterior calculus}\label{s.ext}
601:
602: In the case of differential geometry in $\Rz^{\times n}$ (Example
603: \ref{ex.Rn}), the gradient operator $d=\partial$ behaves symbolically
604: similarly to a covector, except for the nontrivial behavior implied by
605: the Leibniz product rule. However, the gradient operator $d$, defined
606: there on scalar fields only, cannot be extended to a gradient operator
607: that associates with a general $c$-linear form $\phi$ a $(c+1)$-linear
608: form $d\phi$ such that $L_X\phi=Xd\phi$ for all vector fields $X$.
609: The existence of such an
610: extension would imply that $L_{ f X}\phi= f L_X\phi$. While this holds
611: by \gzit{e.ld27a} when $\phi$ is a scalar field, it fails already when
612: $\phi$ is a linear form. For a linear form $\zeta$, we have instead
613: \[
614: L_{ f X}\zeta= f L_X\zeta + d f X\zeta,
615: \]
616: which follows as a special case of \gzit{e.extb} below, or from
617: \[
618: L_X\zeta=X(\partial \zeta)^T + \partial X\zeta
619: \]
620: by substituting $fX$ for $X$. However, the
621: gradient can be generalized in a different way to alternating forms,
622: leading to the exterior derivative.
623:
624: The generalization is valid not only for Example \ref{ex.Rn}, but in
625: full generality. To define the exterior derivative we need some
626: preparations.
627:
628: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
629: \begin{thm}
630:
631: For every linear form $\zeta$, there is a unique $\Ez$-linear mapping
632: $\zeta\wedge$ mapping alternating $c$-forms to alternating
633: $(c+1)$-forms for $c=0,1,2,\dots$,
634: and vector fields to zero, such that
635: \lbeq{e.ld26}
636: (X\zeta)\omega = X(\zeta\wedge\omega) + \zeta\wedge X\omega
637: \eeq
638: for all vector fields $X$ and all alternating $c$-forms $\omega$.
639: $\zeta\wedge\omega$ is called the \bfi{exterior product} or
640: \bfi{wedge product}\footnote{
641: One can define an exterior product $\omega'\wedge\omega$
642: for arbitrary alternating forms $\omega,\omega'$,
643: but we do not need it.
644: } % end footnote
645: of $\zeta$ and
646: $\omega$. The exterior product satisfies the rules
647: \lbeq{e.exta}
648: \zeta\wedge \zeta' = - \zeta'\wedge \zeta,
649: \eeq
650: \lbeq{e.extb}
651: L_X(\zeta\wedge\omega) = L_X\zeta\wedge\omega +\zeta \wedge L_X\omega
652: \eeq
653: for alternating $c$-forms $\omega$, linear forms $\zeta,\zeta'$,
654: and vector fields $X$.
655: \end{thm}
656:
657: \bepf
658: A necessary and sufficient condition for \gzit{e.ld26} to hold is
659: that
660: \lbeq{e.ld26b}
661: X(\zeta\wedge\omega)=(X\zeta)\omega - \zeta\wedge X\omega;
662: \eeq
663: %
664: \MMt{
665: ($p=2$)
666: \[
667: \bary{rcl}
668: X^k \zeta_{[k} \omega_{ij]}
669: &=& \frac{1}{2} X^k \Big(\zeta_k\omega_{ij} - \zeta_i\omega_{kj}
670: - \zeta_j\omega_{ik} - \zeta_k\omega_{ji} \Big)
671: ~=~ \frac{1}{2} X^k \Big(2 \zeta_k\omega_{ij}
672: - \zeta_i\omega_{kj} - \zeta_j\omega_{ik} \Big)\\
673: &=& X^k \zeta_k\omega_{ij}
674: - \frac{1}{2}X^k(\zeta_i\omega_{kj} + \zeta_j\omega_{ik})
675: ~=~ (X\zeta)\omega_{ij}
676: - \frac{1}{2}(\zeta_i X^k\omega_{kj} - \zeta_j X^k\omega_{ki})\\
677: &=& (X\zeta)\omega_{ij}
678: - \frac{1}{2}(\zeta_i (X\omega)_j - \zeta_j(X\omega)_i)
679: ~=~ (X\zeta)\omega_{ij} - \zeta_{[i} (X\omega)_{j]} \\
680: &=& \Big((X\zeta)\omega\Big)_{ij}
681: - \Big(\zeta\wedge (X\omega)\Big)_{ij} \\
682: \eary
683: \]
684: }
685: %
686: in particular,
687: \lbeq{e.ld26a}
688: \zeta\wedge \omega =\zeta\omega \mbox{~~~for a 0-form~} \omega,
689: \eeq
690: This completely specifies the exterior product of a linear form
691: $\zeta$ and an alternating $(c+1)$-form $\omega$, given the exterior
692: product with an alternating $c$-form.
693: Therefore, if the exterior product exists, it is unique.
694:
695: To prove the existence of the exterior product, we have to
696: define the exterior product of a linear form $\zeta$ and an
697: alternating $c$-form $\omega$ to be the expression $\zeta\wedge\omega$
698: defined for $c=0$ by \gzit{e.ld26a}
699: and for $c>0$ recursively by \gzit{e.ld26b}.
700: To show that we really get an alternating $(c+1)$-form, we need to
701: show that $X(\zeta\wedge\omega)$ is alternating for $c>0$ and any
702: vector field $X$, and verify
703: \lbeq{e.ld23a}
704: ( f X)(\zeta\wedge\omega) = f (X(\zeta\wedge\omega))
705: \eeq
706: and
707: \lbeq{e.ld23b}
708: XX(\zeta\wedge\omega) =0.
709: \eeq
710: \at{proof missing, also proofs of \gzit{e.exta} and \gzit{e.extb}}
711:
712: \epf
713:
714:
715: \at{add sum formula for the general case?
716: The reader may wish to prove inductively that, for alternating
717: $c'$-forms $\omega$ with $c'\ge c$, ...}
718:
719:
720:
721: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
722: \begin{thm}\label{t6.2}
723: There is a unique linear mapping $d$ mapping vector fields to zero
724: and alternating $c$-forms to alternating $(c+1)$-forms
725: (for $c=0,1,2,\dots$) such that\footnote{
726: In particular, the relation $L_X=Xd$ valid on scalar fields
727: fails to hold for the extension of $L_X$ and $d$ to alternating forms.
728: } % end footnote
729: \lbeq{e.ld22}
730: L_X\omega = Xd\omega + d(X\omega)
731: \eeq
732: for all alternating $c$-forms $\omega$ and vector fields $X$.
733: The alternating form $d\omega$ is called
734: the \bfi{exterior derivative} of $\omega$, and satisfies the
735: \bfi{exactness relation}
736: \lbeq{e.ld25}
737: dd\omega = 0
738: \eeq
739: and the \bfi{product rules}
740: \lbeq{e.ld28}
741: d( f\omega)= f d\omega+d f \wedge\omega,
742: \eeq
743: \lbeq{e.ld28a}
744: d(\zeta\wedge\omega)=d\zeta\wedge\omega-\zeta\wedge d\omega,
745: \eeq
746: \lbeq{e.ld29}
747: L_{ f X}\omega = f L_X\omega +d f \wedge X\omega,
748: \eeq
749: for all alternating froms $\omega$, scalar fields $f$, linear forms
750: $\zeta$ and vector fields $X$.
751: Note the minus sign in \gzit{e.ld28a}!
752: \end{thm}
753:
754: \bepf
755: A necessary and sufficient condition for \gzit{e.ld22} to hold is
756: that
757: \lbeq{e.ld22b}
758: X(d\omega)=L_X\omega - d(X\omega);
759: \eeq
760: in particular,
761: \lbeq{e.ld22a}
762: X(d\omega)= Xd~\omega =L_X\omega \mbox{~~~for a 0-form~} \omega.
763: \eeq
764: This completely specifies the exterior derivative of an alternating
765: $c$-form.
766: Therefore, if the exterior derivative exists, it is unique.
767: To prove the existence of the exterior derivative, we have to
768: define the exterior derivative of an alternating $c$-form $\omega$
769: to be the expression $d\omega$ determined for $c=0$ by
770: \gzit{e.ld22a} and for $c>0$ recursively by \gzit{e.ld22b}.
771:
772: To show that we really get an alternating $(c+1)$-form, we need to
773: show that $X(d\omega)$ is alternating for $c>0$ and any
774: vector field $X$, and verify
775: \lbeq{e.ld30}
776: ( f X)d\omega = f (Xd\omega)
777: \eeq
778: which shows that $Xd\omega$ is $\Ez$-linear
779: in $X$, so that $d\omega$ is a $(c+1)$-linear form,
780: and
781: \lbeq{e.ld23}
782: XXd\omega =0
783: \eeq
784: which proves antisymmetry.
785: The proof of \gzit{e.ld30} is based on \gzit{e.ld29}.
786: To prove \gzit{e.ld29}, we \at{start missing} get inductively
787: $Y(L_{ f X}\omega- f L_X\omega-d f \wedge X\omega)
788: =YL_{ f X}\omega- f YL_X\omega-Yd f \wedge X\omega
789: =L_{ f X}(Y\omega)-(L_{ f x}Y)\omega
790: - f (L_X(Y\omega)-(L_XY)\omega)
791: -((Yd f ) X\omega-d f \wedge YX\omega)$.
792: Using the induction hypothesis on the first term and \gzit{e.ld28}
793: on the second term, one finds that all terms cancel.
794: From \gzit{e.ld29} one obtains
795: $( f X)d\omega = L_{ f X}\omega-d( f X\omega)
796: = f L_X\omega+d f \wedge X\omega -( f d(X\omega)+d f \wedge X\omega)
797: = f (L_X\omega-d(X\omega) = f (Xd\omega)$,
798: showing that \gzit{e.ld30} holds.
799: \gzit{e.ld23} follows inductively \at{check initial value} from
800: \[
801: XXd\omega = XL_X\omega - Xd(X\omega)
802: =XL_X\omega-L_X(X\omega)-d(XX\omega)
803: =-(L_XX)\omega-d(XX\omega)=0-0=0.
804: \]
805: To prove the product rule \gzit{e.ld28}, we
806: \at{initialize}, and then inductively
807: $Xd( f \omega)=L_X( f \omega)-d( f X\omega)
808: =(L_X f )\omega+ f L_X\omega-( f d(X\omega)+d f \wedge X\omega)
809: = f (L_X\omega-d(X\omega))+(Xd f )\omega-d f \wedge X\omega
810: = f Xd\omega+X(d f \wedge\omega)=X( f d\omega+d f \wedge\omega)$,
811: completing the induction.
812:
813: \at{proof of the other product rule?}
814:
815:
816: To prove the exactness relation \gzit{e.ld25}, we need the formula
817: \lbeq{e.ld24}
818: d(L_X\omega)=L_X(d\omega).
819: \eeq
820: Indeed,
821: $Yd(L_X\omega)=L_Y(L_X\omega)-d(YL_X\omega)
822: =LYL_X\omega-d(L_X(Y\omega)-(L_XY)\omega)$, whereas
823: $YL_X(d\omega)=L_X(Yd\omega)-(L_XY)d\omega
824: =L_X(L_Y\omega-d(Y\omega))-(L_{L_XY}\omega-d(L_XY)\omega)
825: =L_YL_X\omega+d((L_XY)\omega)-L_Xd(Y\omega)$
826: since $L_{X\lp Y}=L_{X\lp Y}=[L_X,L_Y]=L_XL_Y-L_YL_X$.
827: Comparing the two expressions, one finds that
828: $Yd(L_X\omega)-YL_X(d\omega) = d(L_X(Y\omega)-L_Xd(Y\omega)$,
829: which vanishes inductively.
830:
831: Now \gzit{e.ld25} follows inductively from the relation
832: $Xd(d\omega)=L_Xd\omega-d(Xd\omega)
833: =L_Xd_\omega-d(L_X\omega-d(X\omega))
834: =L_Xd\omega-d(L_X\omega)+dd(X\omega)=-dd(X\omega)$,
835: obtained by using \gzit{e.ld24}.
836: (For 0-forms, the $dd$-term is absent, which starts the induction.)
837:
838: \epf
839:
840:
841: In particular,
842: the exterior derivative of the linear form $\zeta$ is the alternating
843: bilinear form $d\zeta$ with
844: \[
845: YXd\zeta = YL_X\zeta-XL_Y\zeta+(X\lp Y)\zeta,
846: \]
847: since $YXd\zeta = Y(L_X\zeta-d(X\zeta)) = YL_X\zeta-L_Y(X\zeta)
848: = YL_X\zeta-((L_YX)\zeta+XL_Y\zeta)
849: = YL_X\zeta-XL_Y\zeta-(Y\lp X)\zeta$.
850:
851: \at{add sum formula for the general case?
852: The reader may wish to prove inductively that, for alternating
853: $c'$-forms $\omega$ with $c'\ge c$, ...}
854:
855: An alternating $c$-form $\omega$ is called \bfi{closed} if $d\omega=0$,
856: and \bfi{exact} if it can be written in the form $\omega=d\theta$
857: for some $(c-1)$-form $\theta$.
858: In particular, a linear form $\zeta$ is exact if it has the form
859: $\zeta=d f$
860: for some scalar field $ f$.
861:
862: By \gzit{e.ld25}, every exact $c$-form is closed.
863: The converse is not generally valid but holds in simple cases,
864: e.g., by the \bfi{Poincar\'e Lemma}, when $\Ez=C^\infty(\Mz )$, where $\Mz $
865: is a nonempty, open and convex subset of $\Rz^n$).
866:
867:
868:
869: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
870: \section{Manifolds as differential geometries}\label{s.dmani}
871:
872: A central notion for analysis on infinite-dimensional spaces is that
873: of a convenient vector space. This notion is discussed in detail in
874: \sca{Kriegl \& Michor} \cite{KriM}, and refines the notion of a
875: Hausdorff vector space, which is a vector space with the minimal amount
876: of topological structure to allow the definition of a meaningful limit.
877: A convenient vector space has in addition a meaningful notion
878: of differentiability of paths, a property essential for differential
879: geometry on manifolds.
880: (For more details on basic notions from topology and functional
881: analysis; see, for example, \sca{Rudin} \cite{rudin}.)
882:
883: \begin{dfn}
884: A vector space $\Fz$ over $\Rz$ is called \bfi{locally convex} if
885: there is a family $S$ of \bfi{seminorms}, i.e., mappings $s:\Fz\to\Rz$
886: such that
887: \[
888: s(\alpha x + \beta y)\le |\alpha| s(x) + |\beta| s(y)
889: \]
890: for all $x,y\in\Fz$ and all $\alpha,\beta\in\Rz$, with the property
891: that
892: \[
893: s(x)=0\mbox{``for all } s\in S \implies x=0.
894: \]
895: A locally convex vector space becomes a Hausdorff space by defining
896: a \bfi{neighborhood} of $x\in\Fz$ to be a set containing for
897: each $s\in S$ some set of the form $\{y\in\Fz\mid s(y-x)< r\}$
898: for some real number $r>0$. Thus a sequence $x_l$ ($l=0,1,2,\dots$)
899: in $\Fz$ converges to $x\in\Fz$ iff $s(x_l-x)\to 0$ for all $s\in S$.
900: A path $\pi$, i.e., a continuous mapping $\pi :\Rz\to\Fz$, is called
901: \bfi{smooth} or
902: \bfi{arbitrarily often differentiable} if there are paths
903: $\pi^{(k)} :\Rz\to\Fz$ ($k=0,1,2,\dots$) such that
904: \[
905: \pi(t+h)=\sum_{k=0^n} \frac{h^k}{k!}\pi^{(k)}(t) + O(h^{k+1})
906: \]
907: for all $t,h\in\Rz$ and all natural numbers $n$.
908: Clearly, $\pi^{(0)}=\pi$, and we write $\dot\pi:=\pi^{()}$.
909: \end{dfn}
910:
911: The reader should verify that, for any seminorm $s$ and
912: $\alpha\in\Rz$, $x\in\Fz$,
913: \[
914: s(0)=0,~~~s(\alpha x)=|\alpha|s(x)\ge 0,
915: \]
916: that in any locally convex vector space, addition and scalar
917: multiplication are continuous, and that differentiation satisfies
918: the traditional rules.
919:
920:
921: \begin{dfn}
922: (iv) A \bfi{convenient vector space} is a locally convex vector
923: space $\Fz$ over $\Rz$ such that every smooth path in $\Fz$ is the
924: derivative of another path in $\Fz$. A complex-valued function $f$ on
925: a nonempty and open subset $\Mz $ of $\Fz$
926: is called \bfi{smooth} or \bfi{arbitrarily often differentiable} if
927: the complex-valued function $f\circ \pi $ is smooth for every
928: smooth path $\pi :\Rz\to\Fz$. The space of smooth functions
929: $f:\Mz \to\Cz$ is denoted by $C^\infty(\Mz )$.
930: \end{dfn}
931:
932: The reader should verify that in a convenient vector space there is
933: a mapping $\partial:C^\infty(\Mz )\to Lin(C^\infty(\Mz ),\Fz^*)$,
934: called the \bfi{gradient} such that
935: \[
936: \frac{d}{dt} f(\pi(t)) = \dot\pi(t) \partial f(\pi(t))
937: \]
938: for all $f\in :C^\infty(\Mz )$, all arbitrarily often differentiable
939: paths $\pi:\Rz\to \Mz $ and all $t\in\Rz$.
940:
941:
942: \begin{expl}
943: The space $\Fz=\Rz^n$ is convenient, with $S$ consisting of
944: the Euclidean norm only. Other
945: examples of convenient vector spaces are Hilbert spaces and Schwartz
946: spaces; see \sca{Kriegl \& Michor} \cite{kriegl97convenient}.
947: \end{expl}
948:
949: \at{extend Example \ref{ex.Rn} to nonempty open subsets of a
950: convenient vector space, and give formulas for $L_X,\zeta\wedge \omega$
951: and $d\omega$ in terms of the gradient.}
952:
953: \begin{dfn}\label{d.pcm}
954: Let $\Ez$ be a differential geometry with Lie algebra $\Wz$ of
955: vector fields.
956:
957: (i) A \bfi{point} is a algebra homomorphism from $\Ez$ to
958: $\Cz$ which maps $1$ to $1$.
959: We write $\Mz (\Ez)$ for the set of points, and say that the point
960: $\xi$ maps the scalar field $f$ to the \bfi{value} $f(\xi)$ of $f$ at
961: $\xi$.
962:
963: (ii) If $\Fz$ is a convenient vector space, an \bfi{$\Fz$-chart}
964: is a homomorphism $C$ from $\Ez$ to $C^\infty(U(C),\Cz$ for some
965: nonempty, open subset $U(C)$ of $\Fz$.
966:
967: (iii) An \bfi{$\Fz$-manifold} is a differential geometry satisfying
968: the axioms
969:
970: (M1)~~ If $f(\xi)=0$ for all $\xi\in \Mz (\Ez)$ then $f=0$.
971:
972: (M2)~~ For all charts $C$ and all $x\in U(C)$, there is a unique
973: $\xi=\xi_C(x)\in \Mz (\Ez)$ such that
974: \[
975: f(\xi_C(x)) = (Cf)(x) \Forall f \in\Ez.
976: \]
977: (M3)~~ For all charts $C$ and all $\delta\in\der C^\infty_0(U(C))$,
978: there is a unique $\widehat \delta \in\der \Ez$ such that
979: \[
980: (\widehat \delta f)(\xi)=0 \Forall \xi\not\in \xi_C(U),
981: \]
982: \[
983: (\widehat \delta f)(\xi_C(x))=(\delta Cf)(x) \Forall x \in U.
984: \]
985: \end{dfn}
986:
987: Informally, property (M1) says that there are sufficiently many
988: points to separate scalar fields. It implies not only that $\Ez$
989: is commutative, since $(fg-gf)(x)=f(x)g(x)-g(x)f(x)=0$ for all points
990: $x$, but also excludes many other commutative algebras, such as
991: nontrivial quotients of the algebra of polynomials in a single variable.
992:
993: (M2) expresses that charts are
994: sufficiently large to represent scalar fields locally, and (M3) says
995: that there are sufficiently many dervations to reduce differentiation
996: locally to charts.
997:
998:
999: \at{Questions: How much of (M1)--(M3) must be assumed and what can
1000: already be proved?}
1001:
1002:
1003: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1004: \section{Manifolds as topological spaces}\label{s.mani}
1005:
1006: \begin{dfn}\label{d.mani}
1007: (i) Let $\Fz$ be a convenient vector space.
1008: A {\bfi{manifold} modeled on $\Fz$}
1009: (short {\bfi{$\Fz$-manifold}} or simply \bfi{manifold}\footnote{
1010: More precisely, this defines arbitrarily often differentiable, real
1011: manifolds whose dimension need not be finite. There are a number of
1012: other notions of a manifold which make somewhat different assumptions.
1013: } % end footnote
1014: if $\Fz$ is apparent from the context)
1015: is a set $\Mz $ whose elements are called \bfi{points} together with a
1016: family $\mathcal{C}$ of maps $\xi:U\to \Mz $ from a ($\xi$-dependent)
1017: nonempty open subset $U$ of $\Fz$ to $\Mz $ called {\bfi{charts}},
1018: with the properties
1019:
1020: (SM1) Every point of $\Mz $ is in the {\bfi{range}} $\xi[U]$ of some
1021: chart $\xi:U\to \Mz $;
1022:
1023: (SM2) A map $\xi:U\to \Mz $ is in $\mathcal{C}$ if and only if $\xi$ is
1024: injective and,
1025: for every nonempty open subset $V$ of $U$ and every chart
1026: $\xi':U'\to \Mz $ in $\mathcal{C}$ with $\xi[V]\subseteq \xi'[U']$,
1027: \[
1028: \xi'^{-1}\xi\big|_V \in C^\infty(V,\Fz).
1029: \]
1030: The manifold is canonically a topological space by declaring as
1031: open sets arbitrary unions of finite intersections of ranges of
1032: charts.
1033:
1034: (ii) The inverse of a chart is called a
1035: {\bfi{local coordinate system}}.
1036: An {\bfi{atlas}} is a family of charts whose ranges cover $\Mz $;
1037: the family $\cal C$ of all charts is the \bfi{universal atlas}.
1038:
1039: (iii) The dimension of $\Fz$ is called the {\bfi{dimension}} of $\Mz $.
1040: In particular, $\Mz $ is called \bfi{finite-dimensional ($d$-dimensional)}
1041: if $\dim \Fz<\infty$ (resp. $\dim\Fz =d$).
1042:
1043: (iv) A mapping $F$ from $\Mz $ to convenient some vector space $\Uz$ is
1044: called {\bfi{smooth}} (or \bfi{infinitely differentiable})
1045: if $F(\xi)\in
1046: C^\infty(U,\Uz)$ for every chart $\xi:U\to \Mz $. A {\bfi{scalar
1047: field}} on $\Mz $ is a smooth complex-valued function on $\Mz $; the algebra
1048: of all scalar fields on $\Mz $ with pointwise operations is denoted by
1049: $C^\infty(\Mz )$\index{$C^\infty(\Mz )$}. A {\bfi{derivation}} on $\Mz $
1050: is a mapping $\delta\in\Lin C^\infty(\Mz )$ satisfying
1051: \[
1052: \delta(fg) = (\delta f)g+f(\delta g)\Forall f,g\in C^\infty(\Mz ).
1053: \]
1054: Thus a derivation on $\Mz $ is an element of
1055: $\der C^\infty(\Mz )$, which we also denote by
1056: $\der \Mz $\index{$\der \Mz $}.
1057:
1058: (v) A canonical differential geometry whose scalar
1059: fields form the algebra $\Ez=C^\infty(\Mz )$ of a manifold $\Mz $
1060: is called a \bfi{differential geometry} of $\Mz $.
1061: \end{dfn}
1062:
1063: Note that a chart $\xi$ is injective, hence its inverse
1064: $\xi^{-1}$, the corresponding local coordinate system is
1065: well-defined on the range of the chart, and maps a nonempty, open
1066: subset of $\Mz $ to an open subset of $\Fz$. In many treatments of
1067: differential geometry, the local coordinate system $\xi^{-1}$
1068: rather than $\xi$ is called the chart.
1069:
1070: \begin{prop}~\\
1071: (i) The set $C^\infty(\Mz )$ of all scalar fields on $\Mz $ is a
1072: commutative $*$-algebra under pointwise multiplication.
1073:
1074: (ii) The set $\Lie \Mz :=\Der C^\infty(\Mz )$ of all derivations on $\Mz $
1075: with the commutator as Lie product is a Lie algebra.
1076: \end{prop}
1077:
1078: \bepf This is left to the reader as a straightforward exercise.
1079: \epf
1080:
1081: \at{add left module structure and topology for
1082: $\Der C^\infty(\Mz )$, if possible in a way
1083: that works for arbitrary $\der \Az$!}
1084:
1085:
1086: \begin{thm}
1087: $\Fz$-Manifolds in the sense of Definition \ref{d.pcm}(iii) and
1088: $\Fz$-manifolds in the sense of Definition \ref{d.mani}(i)
1089: are equivalent
1090: concepts.
1091: \end{thm}
1092:
1093: \bepf
1094: ~\at{give a more precise formulation of the theorem, and give a proof}
1095: \epf
1096:
1097: The motivating example defining the terminology is the surface of the
1098: earth, the \bfi{globe}, which may be regarded as a 2-dimensional
1099: manifold $\Mz $ with $\Fz=\Rz^{1\times 2}$, the vector space of
1100: 2-dimensional row vectors\footnote{
1101: It is convenient to think of points as row
1102: vectors; then tangent vectors are row vectors, too, and
1103: gradients of scalar fields are naturally column vectors.
1104: Thus later expressions like the directional derivative $Xdf$ of a
1105: scalar field $f$ in the direction of a vector field $X$ have a
1106: natural interpretation as ''scalar = row times colums'' in terms of
1107: ordinary matrix algebra.
1108: }. % end footnote
1109: Here the domain $U$ of a
1110: chart $\xi:U\to \Mz $ may be viewed as the paper on which
1111: the chart (a road map, say) is printed. The important points
1112: $q=(x,y)\in U$ correspond to Cartesian coordinates of marks on the
1113: road map labeled by towns. $\xi(q)$ denotes the location of the
1114: corresponding town on the globe. Note that our charts may have
1115: domains which are not neatly cut and may be disconnected or unbounded.
1116:
1117: The simplest examples of manifolds are open, nonempty subsets of
1118: $\Rz^n$. \at{find its general chart!}
1119:
1120: \begin{expl}
1121: We consider the concrete case where smooth manifolds
1122: modeled on the vector spaces $\Rz^d$ for some $d\in \Nz$
1123: are embedded into a bigger vector space $\Rz^n$, and where membership
1124: in the manifold is characterized by $m$ equations $F_k( x)=0$
1125: ($k=1,\dots,m$). For example, a {\bfi{$d$-sphere}} is the set of
1126: points $ x\in \Rz^{d+1}$ satisfying the single ($m=1$) equation
1127: $F( x):= x^T x-1=0$, where the superscript $T$ denotes the transpose.
1128:
1129: Let $\Mz _0$ be an open subset of $\Rz^n$ and let $F\in
1130: C^\infty(\Mz _0,\Rz^m)$. The {\bfi{gradient}} at $ x$ of $F$ is
1131: given by
1132: \[
1133: d F( x) = F'( x)^T \in \Rz^{n\times m}\,.
1134: \]
1135: This generalizes
1136: the traditional terminology for the case where $m=1$.
1137: The implicit function theorem implies that if the gradient has
1138: constant rank $m$, i.e.,
1139: $\rk dF( x)=m$ for all $ x\in \Mz _0$, then the set $\Mz $ given by
1140: \[
1141: \Mz = \left\{ x\in \Mz _0 \mid F( x)=0\right\}\,,
1142: \]
1143: is a $d$-dimensional manifold with $d=n-m$. If $\Mz $ defines a
1144: $d$-dimensional manifold given by an equation $F( x)=0$, then the
1145: tangent space at a point $ x\in \Mz $ is given by\index{$T_ x \Mz $}
1146: \[
1147: T_ x \Mz = \left\{ X \in \Rz^{1\times n} \mid X\cdot dF ( x) = 0
1148: \right\}\,.
1149: \]
1150: \at{prove this?}
1151: Thus the tangent space consists of those vectors perpendicular to
1152: the gradient, that is, the tangent vectors at $ x$ are tangent
1153: to $\Mz $ at $ x$. Hence the name tangent space. The vector fields of
1154: $\Mz $ are similarly given by\index{$\vect \Mz $}
1155: \[
1156: \vect \Mz = \left\{ X\in C^\infty(\Mz ,\Rz^{1\times n}) \mid X( x)\cdot
1157: dF( x) = 0, \, \Forall x\in \Mz \right\}\,.
1158: \]
1159: ~\at{find its general chart!}
1160: \end{expl}
1161:
1162: \bigskip
1163: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1164: % diffeomorphisms
1165: %%%%%%%%%%%%%%%%%
1166: Given an $\Fz$-manifold $\Mz $ and an $\Fz'$-manifold $\Nz$, we define
1167: $C^\infty(\Mz ,\Nz)$ as the set of maps $A:\Mz \to \Nz$ such that if
1168: $\xi:U\to \Mz $ is a chart on $\Mz $ and $\xi':V\to \Nz$ is a chart on
1169: $\Nz$ such that $A(\xi(U))\subseteq \xi'(V)$ implies
1170: $(\xi')^{-1}\circ A \circ \xi\in C^\infty(U,V)$.
1171: A {\bfi{diffeomorphism}} of $\Mz $ is an invertible
1172: mapping in $C^\infty(\Mz ,\Mz )$ with an inverse in $C^\infty(\Mz,\Mz)$;
1173: we write $A x$ for the image of a point $ x\in \Mz $ under a
1174: diffeomorphism $A$.
1175:
1176: We assume that the identity map
1177: on $\Mz $ is in $C^\infty(\Mz ,\Mz )$ and that the composition of $f\in
1178: C^\infty(\Mz ,\Nz)$ and $g\in C^\infty(\Nz,\Nz')$ is in
1179: $C^\infty(\Mz,\Nz')$, a condition\footnote{
1180: In technical terms this says that the modeling vector spaces should
1181: admit a category of smooth manifolds.
1182: }
1183: automatically satisfied in finite dimensions.
1184: \at{and now in general?}
1185: Then the set $\Diff \Mz $\index{$\Diff \Mz $} of all diffeomorphisms is
1186: a group under
1187: composition of maps. Additional conditions are needed to ensure that
1188: $\Diff \Mz $ is a $\vect \Mz $-manifold and hence (in the terminology ofSection ref{s.Liegroups} below) a Lie group;
1189: see, e.g., \sca{Neeb} \cite{Nee},
1190: where one can find a detailed discussion of pathologies that can
1191: arise in infinite-dimensional Lie groups.
1192:
1193: We define a {\bfi{motion}} on $\Mz $ as a mapping $A\in
1194: C^\infty([0,1],\fct{Diff}(\Mz ))$ such that $A(0)=1$ is the identity.
1195: The intuition is that the points $ x=A(0) x$ of an object
1196: (subset of $\Mz $) at time $t=0$, the start of the motion, is moved by
1197: the motion to the point $A(t) x$ at time
1198: $t\in[0,1]$, ending up in $A(1) x$ at the end of the motion.
1199: For every motion $A$ and for all
1200: $t\in[0,1]$ we define a vector field $\dot A(t)$ on $\Mz $ by
1201: \[
1202: \dot A(t)d f : x \to
1203: \frac{d}{ds} f(A(s)A(t)^{-1} x)\Big|_{s=t}
1204: \]
1205: for all scalar fields $ f$ and all $ x\in \Mz $. Since the product rule
1206: holds for
1207: smooth functions in $C^\infty(\Mz )$, the
1208: object $\dot A(t)d$ is indeed a derivation on $\Mz $, and hence
1209: $\dot A(t)$ is a vector field. From the definition of $\dot A(t)$
1210: we get the {\bfi{chain rule}}:
1211: \[
1212: \frac{d}{dt} f(A(t) x) = \dot A(t)d f(A(t) x)\,.
1213: \]
1214: If we are only interested in what happens in an infinitesimal
1215: neighborhood of a point $ x\in \Mz $, the vector fields in
1216: \[
1217: N( x) := \left\{ X_0 \in \vect \Mz \mid X_0d f ( x) = 0
1218: \Forall f\in C^\infty(\Mz )\right\}
1219: \]
1220: have no effect at $ x$. Since $N( x)$ is a vector space,
1221: we can form the quotient space
1222: \[
1223: T_ x \Mz = \vect \Mz / N( x)\,,
1224: \]
1225: called the {\bfi{tangent space}} or
1226: {\bfi{tangent (hyper-)plane}} at
1227: $ x$.
1228: We denote with
1229: \[
1230: X( x):=X+N( x),
1231: \]
1232: the equivalence class of $X$ with respect to the equivalence relation
1233: $X\sim Y \Leftrightarrow X-Y\in N( x)$. We call the equivalence
1234: class that contains the vector
1235: field $X$ the {\bfi{tangent vector} of $X$ at the
1236: point $ x$}.
1237:
1238: The union $T\Mz $ of all $T_ x\Mz$ is naturally a manifold called the
1239: {\bfi{tangent bundle}} of $\Mz$.
1240: \at{also in infinite dimensions?
1241: give charts or say we don't explain it.
1242: Define $T^*\Mz ,T_s\Mz ,T^r_s\Mz $}
1243:
1244: \at{Question: is there an algebraic construction of $C^\infty(T\Mz )$
1245: and $C^\infty(T^*\Mz )$ which does not mention points?}
1246:
1247:
1248: \bigskip
1249: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1250: \bfi{The Lie derivative in the traditional approach.}
1251: In the special case where $\Ez=C^\infty(\Mz )$ for some
1252: finite-dimensional manifold $\Mz $, there is
1253: an alternative, traditional route to the calculus on manifolds,
1254: using the following traditional definition of the Lie derivative.
1255:
1256: For any vector field $X$, the initial value problem
1257: \lbeq{e.ivp}
1258: \xi(0)=x,~~~\frac{d}{d\tau} \xi(\tau)= X(\xi(\tau))
1259: \eeq
1260: is solvable for every $x\in \Mz $, for $\tau$ in some $x$-dependent
1261: neighborhood of zero.
1262: \at{But I have not yet defined the tangent vector
1263: $\frac{d}{d\tau} \xi(\tau)\in T_{\xi(\tau)} \Mz $.}
1264: This follows from the standard theory of ordinary differential
1265: equations, since differentiable vector fields are locally Lipschitz.
1266:
1267: We denote by $e^{\tau X}$ the local diffeomorphism which maps $x$
1268: into the value $\xi(\tau)$ of the solution $\xi$ of \gzit{e.ivp}.
1269: Clearly, $e^0=1$ is the identity, but for fixed $\tau$, the map
1270: $e^{\tau X}$ need not be defined everywhere. The latter is the case
1271: only when \gzit{e.ivp} is solvable for all $\tau\in\Rz$; in this case,
1272: the vector field is called \bfi{complete}, and the $e^{\tau X}$
1273: form a 1-parameter group of diffeomorphisms. In general, we have,
1274: on the domain of definition,
1275: \[
1276: e^{\tau X}e^{\tau' X}=e^{(\tau+\tau') X},
1277: \]
1278: \[
1279: \frac{d}{d\tau} e^{\tau X}x = X(e^{\tau X}x).
1280: \]
1281: We define the \bfi{directional derivative} $L_X\phi$ of a tensor field
1282: $\phi$ with respect to the complete vector field $X$ by
1283: \[
1284: (L_X\phi)(x):= \frac{d}{d\tau} \phi(e^{\tau X}x)\Big|_{\tau=0}.
1285: \]
1286: This defines a linear differential operator $L_X$ mapping tensor
1287: fields to tensor fields of the same type $[c,r]$, called the
1288: \bfi{Lie derivative} of $X$. Clearly,
1289: \[
1290: (e^{\tau L_X}\phi)(x) = \phi(e^{\tau X}x).
1291: \]
1292: It is not difficult to show the \bfi{chain rule}
1293: \[
1294: \frac{d}{d\tau} \xi(\tau) =X(\tau)\xi(\tau) \implies
1295: \frac{d}{d\tau} \phi(\xi(\tau)) = L_{X(\tau)}\phi(\xi(\tau))
1296: \]
1297: for every smooth path $x:[0,1]\to \Mz $.
1298: Here $\dot \xi(\tau)$ is the tangent vector of the path at $\xi(\tau)$.
1299: The chain rule implies the product rule \gzit{e.ld3} for the Lie
1300: derivative.
1301: Therefore Theorem \ref{t.GenLieDeriv} implies that the traditional
1302: concept coincides with our algebraic concept when $\Ez$ is the algebra
1303: of scalar fields of a finite-dimensional manifold.
1304:
1305: In the infinite-dimensional case, this approach can also be carried
1306: through, although it requires considerable technicalities to account
1307: for the fact that initial-value problems for differential equations
1308: in infinite-dimensional spaces are not always solvable. For details,
1309: see \sca{Kriegl \& Michor} \cite{KriM}.
1310:
1311:
1312: \bigskip
1313: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1314: \section{Noncommutative geometry}
1315:
1316: In this short section, we indicate how things generalize
1317: to \idx{noncommutative geometry}, without giving details; the reader
1318: not familiar with the notions used may simply skip the section.
1319:
1320: In noncommutative geometry, position measurements are
1321: limited by uncertainty relations. The notion of a point therefore loses
1322: its meaning, and the evaluation of functions and vectors at a point
1323: is no longer well-defined.
1324: Thus, in noncommutative geometry, a manifold of points no
1325: longer exists, but in place of $C^\infty(\Mz )$ one has a noncommutative
1326: algebra $\Ez$ whose elements behave in a way analogous to scalar fields.
1327: All constructions based only on this algebra rather than a manifold
1328: generalize in an appropriate way to the noncommutative situation.
1329: Thus most geometric notions extend formally, but they can be matched
1330: with true geometric concepts only in certain commutative subalgebras.
1331: \at{match this with earlier stuff}
1332: The basic observation is that a point evaluation is a
1333: $*$-homomorphism of $C^\infty(\Mz )$ to $\Cz$, and conversely,
1334: all such homomorphisms of $C^\infty(\Mz )$ are obtained as point
1335: evaluations.
1336: Now, if $\Ez_0$ is a commutative normed $*$-subalgebra of $\Ez$ whose
1337: completion is a {\bfi{$B^*$-algebra}} (a term we shall not
1338: further use, and hence not introduce formally \at{but give ref's!})
1339: then one can reconstruct on $\Ez_0$
1340: a topological space $\Mz _0$ by calling the characters of $\Ez_0$
1341: points; the $B^*$-algebra is then canonically isomorphic to
1342: the algebra of bounded continuous functions on
1343: $\Mz _0$. If $\Ez_0$ admits sufficiently many derivations then $\Mz _0$
1344: is a (smooth) manifold. When $\Ez_1$ and $\Ez_2$ are two such
1345: commutative subalgebras that do not commute, then, in contrast to the
1346: commutative situation, the corresponding manifolds $\Mz _1$ and
1347: $\Mz _2$ are not naturally embedded into a bigger manifold.
1348: Thus there may be many maximal manifolds embedded in a single
1349: noncommutative geometry.
1350:
1351:
1352: \at{add here the procedure for constructing an algebra from an
1353: atlas of algebras, and invite the reader experienced in differential
1354: geometry to verify that this reduces to the traditional construction
1355: of manifolds by means of an atlas of charts.}
1356:
1357: \at{relate linear forms and Part II}
1358:
1359:
1360: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1361: \section{Lie groups as manifolds}\label{s.Liegroups}
1362:
1363: This section defines Lie groups in full generality.
1364: Differential equations defining the flow along a vector field
1365: naturally produce Lie groups and the exponential map, which relates
1366: Lie groups and Lie algebras. \at{polish the definition}
1367:
1368: \begin{dfn}~\\
1369: (i) A {\bfi{Lie group}} is a group $\Gz$ which is at the same time
1370: a manifold, such
1371: that multiplication and inversion are arbitrarily often differentiable.
1372: A Lie group is both a manifold and a group and the two structures
1373: are compatible. The identity element in a Lie group will always be
1374: written as 1.
1375:
1376: (ii) We canonically embed $\Gz$ into $\fct{Diff}(\Gz)$ by associating
1377: to $A\in\Gz$ the map $B\to AB$, which is a diffeomorphism.
1378: For the definition of the Lie algebra associated with a Lie group,
1379: it is important to know that the group $\Gz$ acts on $C^\infty(\Gz)$ by
1380: right multiplication, that is, to every $A\in\Gz$ we associate the
1381: map $R_A: C^\infty(\Gz)\to C^\infty(\Gz)$ given by
1382: \[
1383: (R_A\varphi)(B) := \varphi(BA)
1384: \]
1385: for all $B\in \Gz$, $\varphi\in C^\infty(\Gz)$. Of
1386: course, the group also acts by left-multiplication on
1387: $C^\infty(\Gz)$ but this action is not directly related to the Lie
1388: algebra.
1389:
1390: (iii) The Lie algebra $\vect \Gz$ contains the set
1391: \[
1392: \log \Gz =\left\{ X\in \vect \Gz \mid R_B\lp Xd = 0 \Forall
1393: B\in \Gz\right\}
1394: \]
1395: of {\bfi{invariant vector fields}}.
1396: \end{dfn}
1397:
1398: It is not difficult to show that every Lie group in the above sense
1399: is a Lie group in the sense of Definition \ref{d.llie}, since $\Gz$
1400: is canonically embedded into $\Lin C^\infty(\Gz)$. The converse is
1401: also valid but a bit more difficult to establish. \at{give details!}
1402:
1403:
1404: \begin{prop}
1405: The invariant vector fields $\log \Gz$ form a Lie algebra.
1406: \end{prop}
1407: \bepf
1408: To check the statement, we only need to show that
1409: the Lie product $X\lp Y$ of two invariant vector fields $X$ and $Y$
1410: is invariant. But this follows from Proposition \ref{lem.der.cent}
1411: since
1412: the invariant vector fields form the centralizer of the
1413: set $\left\{ R_A \mid A\in\Gz\right\}$.
1414: \epf
1415:
1416: \begin{prop}
1417: For any smooth motion $A\in C^\infty([0,1],\Gz)$, where $\Gz$
1418: is identified with a subset of $\fct{Diff}(\Gz)$, the vector field
1419: $\dot A(t)$ is an invariant vector field.
1420: \end{prop}
1421: \bepf We know that $\dot A(t)$ is a vector field. Hence we need to
1422: check that $\dot A(t)$ Lie commutes with $R_B$ for all $B\in\Gz$.
1423: For arbitrary $B\in\Gz$ and $\varphi\in C^{\infty}(\Gz)$ we have
1424: \beqar
1425: (R_B\lp \dot A(t)d)\varphi (A(t)\zeta) &=& R_B (\dot A(t)
1426: d\varphi)(A(t)\zeta) - \dot A(t) d (R_B \varphi)(A(t)\zeta) \nonumber\\
1427: &=&\dot A(t)d \varphi (A(t)\zeta B ) - \frac{d}{dt}(R_B
1428: \varphi)(A(t)\zeta) \nonumber \\
1429: &=& \frac{d}{dt}\varphi(A(t)\zeta B) - \frac{d}{dt}
1430: \varphi (A(t)\zeta B) =0\nonumber\,.
1431: \eeqar
1432: \epf
1433: \begin{rems}
1434: Note that an essential ingredient in the above proof is that the
1435: action of $\Gz$ on $C^\infty(\Gz)$ is defined from the right and
1436: the action of the vector field $\dot A(t)$ from the left.
1437: \end{rems}
1438:
1439: \begin{definition}\label{d.exp}
1440: A motion $A(t)$ is called a {\bfi{uniform motion}} if there exists a
1441: unique $f\in \log\Gz$ such that
1442: \lbeq{e.exp}
1443: \dot A (t) = f A(t) \Forall t\in [0,1].
1444: \eeq
1445: In this case we write $e^f$ for the group element $A(1)$ and call it
1446: the {\bfi{exponential}} of $f$. Conversely, $f$ is called the
1447: {\bfi{infinitesimal generator}}\index{generator!infinitesimal} of
1448: the motion.
1449: \end{definition}
1450:
1451: Formula \gzit{e.exp} is a linear differential equation with
1452: constant
1453: coefficients; the initial condition $A(0)=1$ is already part of the
1454: definition of a motion. In finite dimensions, such initial value
1455: problems are uniquely solvable; in infinite dimensions, unique
1456: solvability depends on additional conditions. It is easy to check that
1457: a uniform motion with infinitesimal generator $f\in\log\Gz$ is given
1458: by $A(t)=e^{tf}$.
1459:
1460:
1461: \begin{expl}
1462: In any associative algebra, the set of invertible elements is a group.
1463: In many cases, the group of invertible elements is a Lie group.
1464: \at{adapt}
1465: In particular, the group
1466: $GL(n,\Kz)$\index{$GL(n,\Kz)$}\index{$gl(n,\Kz)$} of all invertible
1467: $n\times n$-matrices over $\Kz=\Rz$ or
1468: $\Kz=\Cz$ is a Lie group, since it is the open set of points in
1469: $\Kz^{n\times n}$ where the determinant does not vanish, so that
1470: any point has an open neighborhood on which the identity is a
1471: chart. We can choose coordinates $x_{ij}$ for $1\leq i,j\leq n$ and
1472: then $GL(n,\Kz)$ is the open set where $\det (x_{ij})\neq 0$. Any
1473: derivation is of the form
1474: \[
1475: (Xf)(x_{ij}) = \sum_{1\leq i,j\leq n} X^{ij}\frac{\partial}{\partial
1476: x_{ij}}f(x_{ij})\,,
1477: \]
1478: for all $f\in C^{\infty}(GL(n,\Kz))$ and for some $X^{ij}\in \Kz$. One
1479: finds that
1480: $\log GL(n,\Kz)=gl(n,\Kz)$ is the Lie algebra of all
1481: $n\times n$-matrices over $\Kz$. It is easy to verify these properties
1482: by describing everything with matrices. The subgroup of $GL(n,\Lz)$
1483: consisting of the matrices with unit determinant is denoted by
1484: \idx{$SL(n,\Kz)$}. In other words, $SL(n,\Kz)$ is the kernel of the map
1485: $\det:GL(n,\Kz)\to \Kz^*$, where \idx{$\Kz^*$} is the group of
1486: invertible elements in $\Kz$. The Lie algebra of $SL(n,\Kz)$ is
1487: denoted by \idx{$sl(n,\Kz)$} and consists of the traceless $n\times n$
1488: matrices with entries in $\Kz$.
1489: \end{expl}
1490: