Professional Documents
Culture Documents
ELEMENTS OF MATHEMATICS
$2
Springer-Vedag
NICOLAS BOURBAKI
ELEMENTS OF MATHEMATICS
Lie Groups
and Lie Algebras
Chapters 1—3
Springer-Verlag
Berlin Heidelberg NewYork
London Paris Tokyo
Originally published as
ELEMENTS DE MATHEMATIQUE,
GROUPES ET ALGEBRES DE LIE
© Bourhaki, 1971, 1972
1. This series of volumes, a list of which is given on page: ix and 2:, takes up
mathematics at the L ' ' and gives r' proofs. In ‘ ' '_ ' , it
requires no particular knowledge ofmathematics on the readers’ part, but only
a certain familiarity with mathematical reasoning and a certain capacity for
abstract thought. Nevertheless, it is directed especially to those who have a
good knowledge of at least the content of the first year or two of a university
mathematics course.
2. The method of exposition we have chosen is axiomatic and abstract, and
normally proceeds from the general to the particular. This choice has been
dictated by the main purpose of the treatise, which is to provide a solid
foundation for the whole body of modern mathematics. For this it is indis-
pensable to become familiar with a rather large number of very general ideas
and principles. Moreover, the demands of proofimpose a rigorously fixed order
on the subject matter. It follows that the utility of certain considerations will
not be immediately apparent to the reader unlas he has already a fairly
extended knowledge of mathematics; otherwise he must have the patience to
suspend judgment until the occasion arises.
3. In order to mitigate this disadvantage we have frequently inserted examples
in the text which refer to facts the reader may already know but which have
not yet been discussed in the series. Such examples are always placed between
two asterisks: * . . .*. Most readers will undoubtedly find that these examples
will help them to understand the text, and will prefer not to leave them out,
even at a first reading. Their omission would of course have no disadvantage,
from a purely logical point of view.
4. This series is divided into volumes (here called “Books“). The first six
Books are numbered and, in general, every statement in the text assumes as
known only those results which have already been discussed in the preceding
v
TOTEEREADER
volumes. This rule holds good within each Book, but for convenience of expo-
sition these Books are no longer arranged in a consecutive order. At the begin-
ning of each of these Books (or of these chapters), the reader will find a precise
indication of its logical relationship to the other Books and he will thus be
able to satisfy himself of the absence of any vicious circle.
5. The logical framework of each chapter consists of the definitions, the axionu,
and the theorem; of the chapter. These are the parts that have mainly to be
borne in mind for subsequent use. Las important results and those which can
easily be deduced from the theorems are labelled as “propositions”, “lemmas”, _
“corollaries”, “remarks”, etc. Those which may be omitted at a first reading
are printed in small type. A commentary on a particularly important theorem
appears occasionally under the name of “scholium”.
To avoid tedious repetitions it is sometimes convenient to introduce nota-
tions or abbreviations which are in force only within a certain chapter or a
certain section of a chapter (for example, in a chapter which is concerned only
with commutative rings, the word “ring” would always signify “commutative
ring”). Such conventions are always explicitly mentioned, generally at the
beginning of the chapter in which they occur.
6. Some passages in the text are designed to forewarn the reader against
serious errors. These passages are signposted in the margin with the sign
2 ("dangerous bend”) .
7. The Exercises are designed both to enable the reader to satisfy himself that
he has digested the text and to bring to his notice results which have no place
in the text but which are nonetheless of interest. The most difficult exercises
bear the sign 1}.
8. In general, we have adhered to the commonly accepted terminology,
except where there appeared to be good reasons for deviating from it.
9. We have made a particular effort always to use rigorously correct language,
without sacrificing simplicity. As far as possible we have drawn attention in the
text to alum: qf language, without which any mathematical text runs the risk of
pedantry, not to say unreadability.
10. Since in principle the text consists of the dogmatic exposition of a theory,
it contains in general no references to the literature. Bibliographical references
are gathered together in Historical Nam, usually at the end of each chapter.
These notes also contain indications, where appropriate, of the unsolved
problems of the theory.
The bibliography which follows each historical note contains in general
only those books and original memoirs which have been of the greatest impor-
tance in the evolution of the theory under discussion. It makes no sort of pre—
vi
TO THE READER
vii
CX)NTENTS
OF
THE ELEMENTS OF MATHEMATICS SERIES
I. THEORY or SE18
1. Description of formal mathematics. 2. Theory of sets. 3. Ordered sets;
cardinals; natural numbers. 4. Structures.
ll. ADGEBRA
l. Algebraic structures. 2. Linear algebra. 3. Tensor algebras, exterior
algebras, symmetric algebras. 4. Polynomials and rational fractions.
5. Fields. 6. Ordered groups and fields. 7. Modules over principal ideal
rings. 8. Semi-simple modules and rings. 9. Sesquilinear and quadratic
forms.
VI. Imomnon
COMMUTATIVE ALGEBRA
1. Flat modules. 2. Localization. 3. Graduations, filmtions, and topo-
logies. 4. Amociated prime ideals and primary decomposition. 5. Integers.
6. Valuations. 7. Divison.
SPECTRAL menu
1. Nomad algebras. 2. Locally compact groups.
Summary of results.
CONTENTS
To THEREADER ...............................................
Hmfl®mUIUImHU—I-I
§ 1. Definition of Lie algebras ......... .... ..
Algebras .........
wwvwwrwpr
§ 3. Representations .....
l. Representations ...................................
CONTENTS
esaweww
. Representations on homomorphism modules .
Examples .............................
Invariant elements . . .
Invariant bilinear forms .
Casimir element ......... .
Extension of the base ring ...........................
§4. Nilpotent Lie algebras ..................................
l. Definition of nilpotent Lie algebras
2. Engel’s Theorem .........................
3. The largest nilpotency ideal of a representation .
4. The largest nilpotent ideal in a Lie algebra . . . . ...
5. Extension of the base field ...........................
§ 5. Solvable Lie algebras ..................................
1. Definition of solvable Lie algebras .
2. Radical of a Lie algebra .........
3. Nilpotent radical of a Lie algebra .
4. A criterion for solvability .........
5. Further properties of the radical ...
6. Extension of the base field ...........................
§ 6. Semi-simple Lie algebras ................................
1. Definition of semi-simple Lie algebras
2. Semi-simplicity of representations ....................
3. Semi-simple elements and nilpotent elements in semi-
simple Lie algebras ...............................
4. Reductive Lie algebras ......
5. Application: a criterion for semi Simplicity of representa-
tions .............
. Subalgebras reductive m a Lie algebra . .
8 lbfluai
xii
CONTENTS
Exercisesfor§4 ..........
Exercisesfor§5 ...................................
Exercises for§6 ........................................... 102
Exercisesfor§7 ........................................... 109
xvii
CHAPTERI
Lie Algebras
l. ALGEBRAS
Let M be a unitary module over K with a bilinear mapping (x,y) >—>xy of
M x M into M. All the axioms for algebras are satisfied except associativity
of multiplication. By an abuse of language, M is called a not mterrarily asso-
ciative algebra over K, or sometimes, when no confusion can arise, an algebra
over K. In this no. we shall use the latter notation.
If the K-module M is given the multiplication (x,y) n an algebra is
obtained called the opposite of the above algebra.
A sub-K-module N of M which is stable under multiplication is given the
structure of an algebra over K in an obvious way. N is called a .mbalgebra of M.
N is called a left (resp. right) ideal of M if the conditions x e N, y e M imply
ya: 6 N (resp. zy e N). If N is both a left ideal and a right ideal of M, N is called a
two-sided ideal of M. In this case the multiplication on M enables us to define,
on passing to the quotient, a bilinear multiplication on the quotient module
M/N such that M/N has an algebra structure. M/N is called the quotient '
algebra of M by N.
1' The propositions proved in this chapter depend only on the properties estab-
lished in Books I to V1 and on certain results of Camrmualive Algzbva, Chapter III,
§ 2.
I LIE ALGEBRAS
Let M1 and M2 be two algebras over K and (I; a mapping of M1 into M,.
4) is called a homomorphism ifo is K-linear and (Mary) = ¢(x)d>(y) for x 6 M,,
y 6 M1. The kernel N of (1) is a two-sided ideal of M1 and the image of d) is a
subalgebra of M2. On passing to the quotient, 4) defines an isomorphism of the
algebra MllN onto the algebra ¢(Ml).
Let M be an algebra over Ki A mapping D of M into M is called a dm'mztz'nn
of M if it is K-linear and D(xy) = (Dx)y + x(Dy) for all x e M and y E M.
This definition generalizes Definition 3 of Algebra, Chapter IV, §4, no. 3. The
kernel of a derivation of M is a subalgebra of M. If D1 and D2 are derivations
of M, then Dn — D2D1 is a derivation of M (of. Algzbnz, Chapter IV, §4v,'
no. 3, Proposition 5: the proof of this proposition does not use the associativity
of the algebra).
Let M1 and M, be two algebras over K. On the product K-module
M = M, x M, we define a multiplication by writing
for all :41, y1 in M1, x2, y; in M3. The algebra thus defined is called the pruduct
algebra of M1 and M2. The mapping x, »—> (x), 0) (resp. x2 >—> (0, x,» is an
isomorphism of M1 (resp. M2) onto a. two-sided ideal of M. Under these iso-
morphisms MI and M2 are identified with two-sided ideals of M. The K-
module M is then the direct sum of M1 and MT Conversely, let M be an
algebra over K and M,, M, two two-sided ideals of M such that M is the direct
sum of M1 and M,. Then M,M3 C Ml n M2 = {0); then, ifzh _l/l belong to
M, and :3, y, to M2, then (x; + x¢)(y1 + ya) = x1y1 + x21”, so that M is
identified with the product algebra M1 x M,. Every left (resp. right, two-
sided) ideal of M1 is a. lefi (reap. right, two-sided) ideal of M. We leave to the
reader the task of formulating the analogous results in the case of an arbitrary
finite family of algebras.
Let M be an algebra over K and suppose that the K-module M admits a
basis (aflnh. There exists a unique system (YMV)0.,D.V)EL x L XL of demerits of K
such that aha“ = v nwav for all A, y. in L. The y,“ are mlled the constants bf
structure of M with respect to the basis (0,).
Let M be an algebra over K, K,3 a commutative ring with unit element and
p a homomorphism of K0 into K mapping unit element to unit element. Then
M can be considered as an algebra over K0 by writing out = p(ot) .x
for a: E K0, 1: e M. This is the case in particular when K0 is a subring of K
containing the unit element and p is taken to be the inclusion mapping of K0
into K.
Let M be an algebra over K, Kl a commutative ring with unit element and
a a homomorphism of K into K1 mapping unit element to unit element. Let
114““ u) = Mam be the Kl-module derived from M by extending the ring of
2
LIE “023m § 1.2
2. LE ALGEBRAS
DEFINITION 1. An algebra 5 over K i: called a Lie algebra over K {fits multiplication
(denoted by (x, y) I—> [x, y]) satisfies the identities:
(1) PM] = 0
(2) [’8 [% zll + [1b [1, *1] + [1: [#41] = 0
for all x, y, z in g.
The product [x, y] is called the bracket of x and y. Identity (2) is called the
Jacobi identity.
The bracket [x, y] is an alternating bilinear function ofx and y. We have the
identity:
3
I 113W
algebra over K canonically isomorphic to 31(11, K). Let (Eu) be the canonical
basis ofM,(K) (Algebra, Chapter II, § 10, no. 3). It follows easily that:
[EmEu]=O ifjaélc and i761
(5) [Em En] = Eu iff # I
[Em EM] = ‘51:; 1f! 9‘ k
[Em Eu] = Eu ‘ E11
The Lie subalgebra of M,(K) consisting of the triangular matrices (resp.
triangular matrices ofzero trace, mp. triangular matrices of zero diagonal) is
denoted by 1(n, K) (resp. “(11, K), reap. n(n, K)) (Algzbm, Chapter II, § 10,‘
no. 7).
*Example 3. Let V be an infinitely difl‘erentiable real manifold. The differen-
tial operators with infinitely differentiable real coefficients constitute an asso-
ciative algebra over R and hence, by Example 1, a Lie algebra A over R. The
bracket of two infinitely diflerentiable vector fields on V is an infinitely
difl'erentiable vector field and hence the infinitely difl'erentiable vector fields
on V constitute a Lie subalgebra f of A. IfV is a real Lie gimp, the left invariant
vector fields constitute a Lie subalgebra 9 of f called the Lie algebra of V. The
vector space 9 is identified with the tangent space to V at z (the identity element
of V). Let V’ be another real Lie group, e’ its identity element and 9' its Lie
algebra. Every analytic homomorphism of V into V’ defines a linear mapping
of the tangent space to V at 2 into the tangent space toV' at z’ ; this mapping is
a homomorphism of the Lie algebra 9 into the Lie algebra 9’. If V is the linear
group of a finite-dimensional real vector space E there exists a canonical iso-
morphism of 91(E) onto the Lie algebra g ofV, under which 9 is identified with
91(13):
DEmou 2. Let g b: a Lie algebra and x an element of 9. Th; linear "upping
y» [x,g] q into 9 isculledtheafioinllineaimappingqfx audit denotedby ad“): or
ad x.
PROPOSITION 1. Let 9 In a Lie algebra. For all x e 9, ad x it a derivation. 771: mapping
a: >—> ad at is a hmmorphirm ifthe Lie algebra 9 into the Lie algzbm b qfdm'ualions (J
g. IfD e b ands E 9, [D, ad x] = ad(Dx).
Identity (4) can be written:
(adXHy; 2] = [(34904) z] + U,(3dx)-21
or:
(adl-z = (adx).<(ady>.z) — (ady).<(ad~>.z)
whence the first two assertions. On the other hand, if D e b, x E 9, y e 9, then
[D, ad x] . y = D([x,y]) — [x, Dy] = [Dx,y] = (ad Dx) .y, whence the last
assertion.
4
ooumrrA'nvz LIE mums § 1.3
3. COMMUTATIVE LIE ALGEBRAS
DEFINITION 3. Two element: x, 31 cf a Lie algebra are said to be permutable if
[x, y] = 0. g is said ta be enmmutatiue ifany two qfils elements are pmrmtable.
Example I. Let L be an associative algebra and g the Lie algebra defined by
it (no. 2, Example 1). Two elemenm x, y are permutable in 9 if and only if
xy = get in L.
'Example 2. Ifa real Lie group G is commutative, its Lie algebra is commuta-
five,
Every K—module can obviously be given a unique commutative Lie algebra
structure over K.
If g is a Lie algebra, every monogenous submodule of g is a commutative Lie
subalgebra of g.
44 IDEALS
It follows from identity (3) that in a Lie algebra 9 there is no distinction
between left ideals and right ideals, every ideal being two-sided. We therefore
speak simply of ideals.
‘Example. Let G be a Lie group, 5 its Lie algebra and H a Lie subgroup of G.
Every left invariant vector field on H defines canonically a left invariant
vector field on G, whence there is a canonical injection of the Lie algebra I; of
H into 9; b is identified with a Lie subalgebra of 9 under this injection. If H is
normal in G, the canonical image of b in g is an ideal of 9.,
An ideal of g is a submodule of g which is stable under the inner derivations
of 9.
DEFINITION 4. A submadule 9f 9 which is stable under every derivation of 5 i: called
a characteristic ideal cf 9.
PROPOSITION 2. Let g be a Lie algebra, a an ideal (reap. a characteristic ideal) of g
and b a characteristic ideal Lfa. 77121: b is an ideal (resp. a chaiaetm'slic ideal) ofg.
Every inner derivation (resp. every derivation) of g lava a stable and
induces on a a derivation and hence leaves b stable.
LetgbeaLiealgebra. Ifaandbareidealsofg, a + bandan bareideals
of 9.
Let n and b be two submodule: ofg. By an abuse of notation, the submodule
of 9 generated by the elements of the form [x,y](x€a, ye b) is denoted by
[0, b]. We have [a, b] = [b, a] by identity (3). If 2 E g, [2, a], or [a, 2],
denotes the submodule [K1, (1] = (ad z) (a).
PROPOSITION 3. If a and b are ideals (resp. characteristic ideals) if g, [a, b] is an
ideal (resp. a characteristic ideal) cf 9.
l LIE AwEBRAS
6
muons § 1.7
ayix‘,
l
X9,/
is commutative (that is such thatfn )t = l', p.’ of = It). We show that such a
homomorphism is necasarily bxjeefivei First f is injective. For if x e g is such
thatf(x) = 0, then p(x) = y.’(f(x)) = O and hence x = My) for some y e a;
then A’(y) =f(x(y)) =f(x) = 0,hencey = Oand hence: = 0. On the other
hand, f is surjective. For p.’ of = p. is suljective and hencef(9) + A’(n) = 5’;
on the other handf(g) Df()\(a)) = A’(a).
It follows from this that the relation just defined between two extensions
of b by a is an equivalence relation.
Pxoposmou 6. Let
A II
a —> g —> b
be an extensionofbbyaandnitxkemel.
I LIE AmEBRAS
8. SEMI-DIRECT PRODUCTS
Let a and b be two Lie algebras over K. It is not easy to construct all the
extensions of b by 0. But we shall describe quite simply all the inmential
extensions of b by a.
Let g be an inessential extension of b by a. We identify a with an ideal of 9, b
with a subalgebra of g supplementary to a and the module g with the module
a x b. For all h e b, let (I), be the restriction to a ofad, b; this is a derivation ofn
and the mapping h » it» is a homomorphism of b into the Lie algebra of deriva-
tions of a. On the other hand, for a, a' in u and b, b' in b, we have:
(6) [(a, b), (a', b’)] = [a + h, a’ + b’]
= [4, 4’] + [41, ’1'] + lb, '1’] + [5, 1"]
= (la, 0'] + W“ - {Mb [5, 5'1)-
Conversely, let a and b be Lie algebras over K and h H 4:. a homomorphism
of 6 into the Lie algebra of derivations of a. On the product 9 ofthe K—module: a
and b we define the bracket of two elements by writing:
[(11,12). (a’,b')] = ([a, a’] + ¢.a’ — w, [1), b’])
sum-imma- pkonucis § 1.8
If n is identified with n undo“ I and b with v(b) under v, then, for a E n and
b E b:
(ad l0J1 = [(0, b), (a, 0)] = (4M, 0) = 4M-
tn (b = 0, g is the product Lie algebra of b and a. In the general case, 9 is
called the semi—direct product of b by a (corresponding to the homomorphism
b r—> 4)” of b into the Lie algebra of derivations of a).
We have therefore established the following proposition:
Pnorosmon 7. Let a and b be two Lie algebra: over K,
L II
n ——> g ——> b
«Wm
x9 1
i: cammutatiue, so that the two ekiom are equivalent.
Example 1. Let g be a Lie algebm over K and D a derivation of 3. Let I; be the
commutative Lie algebra K; The mapping x w ADO e K) is a homomorphism of
1) into the Lie algebra of derivations of g. \N’e form the corresponding semi-
direct product I ofb by 9. Let x0 be the element (0, 1) of I. For all x s g,
Dx = [x.,, x].
Example 2. Let g be a Lie algebra over K, M a K-module and p a homomor-
phism of 9 into gl(M). If M is considered as a commutative Lie algebra, the Lie
algebra of derivations of M is gl(M). We can therefore form the semi-direct
product [9 of 9 by M corresponding to p.
In particular, let 9 = 51(M) and p be the identity mapping of gl(M). The
semi-directproductofg byM is then denoted by af(M) (or af(n, K) ifM = K") .
An element of af(M) is an ordered pair (m, u), where m e M, u E 9|(M); and
the bracket is defined by
Him"). ("1" u')] = (“(M') - "'(M), [I‘d/l)-
CHANGE or BASE RING § 1.9
ll
I ma Mam
The centralize: ofa subset is the same in g and 9’. Hence fps = 7,9' for all
12
mwpmc ALGEBRA or A nooucr § 2.2
13
I LIE ALGEBRAS
l4
ENVELOPING Amman or THE OPPOSITE ALGEBRA §2.4
9—‘—>94>slb
1.15 if
v—L>U——>w
proves that i is zero on 6(b) and hence on R. Let 4) be the canonical homomor-
phism of U onto U/R. Them exists a homomorphism 4) of U/R
9—'—>sb
. 4*
U4» U/R $W
15
I LIE ALGEBRAS
The y“ for M E A,I form a system of generators of the vector K-space 5'". Now
their canonical images in S" constitute, by the above paragraph, a basis of S".
Hence (yM)ME A" is a basis of a supplementary subspace of I h T" in T”
(Algebra, Chapter II, § 1, no. 6, Proposition 4), which establishes our assertion.
Thus, when K is a field ofcharacteristic 0, the restriction to S”I of the canoni-
cal mapping T'| ~> S" is an isomorphism of the space 5'" onto the space S'| and
therefore has an inverse isomorphism. The inverse isomorphisms thus obtained
for each 71 define a canonical isomorphism of the space S onto the space
S’ = 120 S’" of symmetric tensors.
16
FILTRATION or THE ENVELOPING ALGEBRA §2.6
. U"
T"/
Uu \Gn
0.
kAsn/ml
I LIE ALGEBRAS
Remark. Let g’ be another Lie algebra over the ring K, U’ its enveloping
algebra, U,’, the set of elements of U’ of filtration $7; and U" (resp. U’") the
set of canonical images in U (resp. U’) of the homogeneous symmetric
tensor: of g (resp. g’) of order 11. Let n be a homomorphism of 3 into 9' and
let i be the corresponding homomorphism of U into U’. Then
18
THE Pomonm‘sqammOmm THEOREM § 2.7
We write I in the form ”21 XM" where the M. are sequences ofn elements of
A. The mapping 6 extends to a homomorphism of the algebra T into the
algebra $14?) (which we shall also denote by a), which is zero on J By Lemma
20
'n-nz rommkfiqmmomwm THEOREM § 2.7
.6”, in the notation ofdiagram (3), the restriction cf 1-,. to W i: an iramorphirm qfW onto
5", then the rerhictian af 4),, ta W ir an immorphirm qf W onto a supplement of Un_1
in U”.
The restriction to W often a 1',I is a bijection ofW onto C"; so is the restriction
6,, o 41,, to W. Hence the corollary.
COROLLARY 2. If g is aflee K-module, the cammieal mapping of 9 into iLr enuelaping
algebra ir injective.
This follows from Corollary 1 taking W = T1.
When 9 is a free K-module (in particular when K is a field), 9 is identified
with a submodule of U under the canonical mapping of 3 into U. This con-
vention is adopted from the following corollary onwards,
COROLLARY 3. If g admit: a tatally entered barir 09.)“ A, the element: xxlxh. . .xh
4f the enveloping algebra U, where (M, . . . , An) 1': an arbitrary increasingfinitc sequence
of element: qf A, firm a basis qf the K-module U.
Let A,‘ be the set of increasing sequences of n elements of A. For
M = (1b.. ., A") EA,” let yM = xx] ®"M ®~~~®xw Let W be the sub-
module of T’‘ with basis (yulus 1w Corollary 1 shows that the restriction of a
to W is an isomorphism of W onto a supplement of U,._l in U”. But
‘i’n(!lM) = "xfih- - J‘s“,
whence the corollary.
COROLLARY 4. Let S’” C T" be the set of homogeneous rymmetn'e tensor: qf arder n.
Suppose that K it afield characterim'c 0. Then the cmfiarite mapping If the canonical
mapping:
8" —> S"I —> U”
is an iramorphirm qf the vector space S“ onto a .mpplement q".I in U,,.
This follows from Corollary 1 taking W = 8’”.
Suppose henceforth that K is a field of characteristic 0. Let n, be the
mapping of S’| into U,I just defined. Let U’I = 7],,(S"). The vector space U is
the direct sum of the U". The n, define an isomorphism 'q of the vector space
S = Z S” onto the vector space U = 2: U", called the cammical iramnrlzhirm of
S antu U; this is nut an algebra isomorphism. We have the commutative
diagram:
un
4b- (in
S'n/ 7: \Gn
N sand
21
I LIE AwEBRAS
where each arrow represents a vector space isomorphism. If x1, x2, . . ., x" are
ing,-q,,mapsther ’ xlxa...x, ' ' rdinS,tothe '
we. . m:
which constitute a basis of U. Hence the corollary.
22
EXTENSION OF DERIVATIONS § 2.8
B. EXTENSION OF DERIVATIONS
23
1 LIE menus
= A(2
i=1
(10:1). . .c(Dx,) . . 43(4))
= ADu(c(x,) . . .o(x,.)).
Finally, let xe 9. Let A be the inner derivation y >—> c(x)y — yo(x) of U
(Algebra, Chapter IV, §4, no. 3, Exampb 2). Then, for z’ e g,
(A o c)(x’) = o(x)u'(z') — o‘(x’)a(x) = c([x, x’]) = (a o ad x)(x‘),
whence A a a = a o ad x. This completes the proof.
Applying Proposition 7 to the case of a commutative Lie algebra, it is seen
that every endomorphism u of a K-module can be extended uniquely to a
derivation of the symmetric algebra of this module; this derivation is derived
on passing to the quotient from the derivation of the tensor algebra which
extends it.
We again take a Lie algebra 3 over K and let D be a derivation of 9. We use
the earlier notation T, S, U, G. Let D1, D3 be the derivations of T, S which
extend D and let D" be the unique derivation of U such that a o D = D0 0 0'.
Since DU leaves the U" stable, Du defines on taking quotients a derivation
D5 of G. Since DH and D3 are derived from Dr when passing to quotients, the
commutative diagram (3) proves that DG can also be derived from Ds by the
24
REPREENTATIONS § 3.1
(x', y' in gm”). Clearly the canonical image ofjm, in Tan) is contained in J’.
To see that it is equal to J', it suffices to show that, ifx' and 3/ denote two ele-
ments of 9m): x’ 8) y' — y’ 8) x’ — [1', y’] belongs to this image. New
§ 3. REPRESENTATIONS
l. REPRESENTATIONS
DEFINITION 1. Let g be a Lie algebra over K and M a K-module. A homomorphirm of
9 into the Lie Algebra 91(M) i: called a representation of g on the module M. An
injective representation is calledfail/$11. IfK i: afield, the dimion (finite or White) of
M aver K it called the dimmion qflhe representation. The momenta/ion x >—> ad 7: (ft;
on the K-madule 9 it called the adjoin! reprermlalion 1J9.
A representation of g on M is thus a K-linear mapping 9 of 9 into the
endomorphism module of M such that
P([x,y])-m = P(*)P(!)-'" — 9(y)9(x)~m
forallxeg,yeg, meM.
25
I LE ALGEBRAS
*Example. Let G be a real Lie group, 9 its Lie algebra and 0 an analytic
representation of G on a finite-dimensional real vector space E. Then the
corresponding homomorphism of 9 into 91(E) is a representation of g on E.,,
Let U be the enveloping algebra of9. Proposition 1 of§ 2, no. 1 defines a one-
to—one correspondence between the set of representations of g on M and the set
of representations of U on M. On the other hand we know (Algebra, Chapter
VIII, § 13, no. 1) that there is an equivalence between the notion of repre-
sentation of the associative algebra U and that of left U-module.
DEFINITION 2. Let g be a Lie algebra aver K and U it: enveloping algebra. A unitary
left module aver U i: called a left g-rrwdule, or simply a 9-madule‘
If M is a g-module and x e U, X” will denote the homothety of M defined
by x (cf. Algebra, Chapter VIII, § 1, no. 2).
A unitary right module over U is called a right g-module. Such a module is
identified with a left U‘l-module, that is (§ 2, no. 4) with a left g°-modu1e.
Let 4) be the principal antiautomorphism of U. If M is a right g-module, a
left g-module structure is defined on M by writing 4.”: = m.¢(a) for m E M
and a e U.
The notions and results of the theory of modules can be translated into the
language of representations:
(1) Two representations 9 and p' of g on M and M' are called similar or
isomorphic if the g-modules M and M’ are isomorphic. For this it is necessary
and sufficient that there exist an isomorphism u of the K-module M onto the
K-module M' such that
9'(~) = u . m) ., w
for all x e g.
(2) For all i E I, let p; be a representation of g on M,. Let M be the g-module
the direct sum of the 9-modules M,. There is a corresponding representation p
ofg on M, called the dim! sum of the p. and denoted by g p‘ (or p1 + - - - + 9,,
in the case of n representations 91, . . ., 9”). If m = ("10.5, is an element of M
and x E 9, then PM” = (91(3) we.“
(3) A representation 9 of g on M is called simple or irreducible if the asso-
ciated g-module is simple. It amounts to the same to say that there exists no
sub-K-module of M (other than {0) and M) stable under all the p(x), x E g. A
class of simple g-modules (Algebra, Chapter VIII, § 3, no. 2) defines a clan Qf
simple representatian: of g.
(4) A representation 9 of g on M is called semi-simple or completely reducible if
the associated g-module is semi-simple. It amounm to the same to say that p is
similar to a direct sum ofsimple representations or that every sub«K-module of
26
REPRESENTATIONS § 3.1
M stable under the p(x) (x e g) has a supplement stable under the p(x) (x E g)
(cf. Algebra, Chapter VIII, § 3, no. 3).
(5) Let 8 be a class of simple representations of 9 corresponding to a class C
of simple g-modules. On the other hand let p be a representation of g on M.
The isotypical component MC ofspecies C of the g—module M (Algebra, Chapter
VIII, § 3, no. 4-) is also called the iralypical component of M afrpm'er 8. This com-
ponent is the sum of the sub-K-modules of M stable under the 9(x) and on
which the p(x) induce a representation of class 8; it is the direct sum of certain
of these submodules; if M0 is of length n, p is said to contain 8 n times. The sum
of the different M,J is direct; it is equal to M if and only if p is semi-simple.
(6) Let p, p’ be two representations of g. p’ is milled a mbnpmentatian (resp.
quotient reprerenlation) of p if the module of p' is a submodule (resp. quotient
module) of the module of 9.
Let M be a K-module. The zero representation of g on M defines on M a
g-module structure. With this structure M is called a trivial g-module.
Let M be a g-module. The quotient g—modules of the sub-g-modules of M are
also the sub-g—modules of the quotient modules of M: they are obtained by
considering two sub-g-modules U, U’ of M such that U D U’ and forming the
g-module U/U’. Then if all the simple modules of the above type are isomor-
phic to a given simple g-module N, M is called a pure g-module of specie: N.
If p and a are the representations of 9 corresponding to M and N, we also say
that p is pure cfrpeeies a.
Let M’ be a sub-g-module of M. For M to be pure ofspecies N, it is necessary
and sufficient that M’ and M/M’ be pure of species N. For the condition is
obviously necessary. Suppose that it holds and let U, U’ be sub- g-modules of
M such that U' C U and U/U’ is simple; let <1) be the canonical homomorphism
ofM onto M/M’; if MU) sé <1)(U’), U/U’ is isomorphism to d)(U)/¢(U’) and
hence isomorphic to N; if¢(U) = ¢(U’), then U C U' + M’, hence U/U’ is
isomorphic to a simple submodule of (U' + M’)/U’ and the latter module is
itself isomorphic to M’/(U’ n M’); hence U/U' is again isomorphic to N, so
that M is pure of species Ni
Henceforth let M be a g-module and suppose that the set of sub-g-modules
of M which are pure of species N admits a maximal element M’. Then every
submodule M” of M which is pure of species N is contained in M’. For
M’/(M’ n M”) and M’ are pure of species N, hence M' + M” is pure of
species N by the above and hence M' + M” C M'.
Suppose that the g-module M admits a Jordan-Holder series (Mao‘i‘n.
For M to be pure of species N, it is necessary and sufficient that Mn/Ml,
M1/M2, . . ., Mn_,/M,, be isomorphic to N; for the condition is obviously
necessary and its sufficiency follows immediately by induction on n from what
we have seen above.
PROPOSITION 1. Let g be a Lie algebra aver K and a an ideal q/g. Let M be a g-madule
27
I L“! ALGEBILAS
and N a simple n-module. Consider M a: an a-module and mppose that the set qfsuh—a—
modulu of M which an [21m qfrperies N admits a maximal damn: M’. Then M’ u a
mb-g-madule of M.
Let yeg. Let 4) be the canonical mapping of M onto M/M’ andf the
mapping m >—> (lg/Wm) of M’ into M/M’. It suffices to show thatf(M’) = {0}.
Let x e (I. Then, for In 6 M,
”mm-f0") = “India-m) = ‘l’lyuxu-m) + Ml’rsylu-ml'
Now [24, y] e a, whence ¢([x, y],..m) = 0; on the other hand,
¢(yuxfl'm) =f("u-'")-
Hence xM,Mr.f(m) = f(xMJn). It follows that f(M’) is a sub-o—module of
M/M’ isomorphic to a quotient of MI and hence pure of species N; hence
f(M’) = {0}-
COROLLARY. Let g be a Li: algebra over K and a an ideal of 9. Let M be a simple
g-maa'ule, qffinite length a: a K-madulo. Tiler: exist: :1 mph n-module N such that M
i: a pure a-madule qfspetie: N.
Since the a-module M is of finite length, there exists a minimal element
N in the set of sub-a-modules of M: it is a simple sub-a-module of M. The
largat sub—a-module of M which is pure of species N is therefore #{0} and
is a sub-g-module of M (Proposition 1) and is therefore identical with M.
28
REPRESENTATIONS on nonomonrmsm MODULES § 3.3
forallxeg, mleM,,...,mneM,..
The corresponding representation is called the tensor product of the given
representations of g on the M.
In particular, if M is a g-module, Proposition 2 defines a g-module structure
9
29
I LIE ALGEBRAS
or
(6) xwu = In," — uxul.
Combining these results with Proposition 2, we see that:
Pnorosmon 3. Let g be a Lie algebra aver K and MK a g-moduk (l S i S n + l).
n
Let N be the K-madul: 2AM”. . ., Mn; Mn“) ufmultilitwar mapping: #1:]; M.
into Mn“. The" exist: on: and only one g—module drum”: on N such that
n
(7) (z...u)<m.,...,m.) = —.Zlu(m.,....x....m.,...,m.)
+ xMMI.u(m,,...,m,,)
forallxeg,ueNtzmtlm‘eMl (l S {g n).
In particular, let g be a Lie algebra over K and M a g—module and consider
K as a trivial g-module. Proposition 3 defines a g—module structure on
S’K(M, K) = M“. The corresponding repmentation is called the dual repre-
sentation of the representation at >—> x“. We have:
30
EXAMPLE § 3.4
4. EXAMPLES
Exampl: I. Let g be a Lie algebra over K and M a 9-module. The g-module
structure on M and the trivial g-module structure on K define a g-module
structure on the K-module N = 2’(M, M; K) of bilinear forms on M. Then
(10) (In-BXMM') = ~B(xn-ma m') + lwm’)
forallxe g, m, m' in M,BEN. IfflisagivenelementofN, theset ofxegsnch
that xwfl = O is a subalgebra of 9.
Let M be a K-module and B a bilinear form on M. By the above, the set of
x e gI(M) such that
B(x.m, m’) + fl(m,x.m’) = 0
for all m E M and m' e M is a Lie subalgebra of gl(M). Suppose that K is a
field, M is finite-dimensional and B is non-degenerate. Then every xe gl(M)
admits a left adjoint x* (relative to p) which is everywhere defined and the
subalgebra in question is the set of x E gl(M) such that x“ = —x. By this pro-
cess we can construct two important examples of Lie algebras:
(3) Take M = K" and
WE], - . ~, Eu): (7)» - - and) = 5m + ~ ~ ~ + 5m..-
We canonically identify gl(K") with Mn(K). Then the Lie algebra obtained is
the Lie algebra of skew-symmetric matrices. *(When K = R, this algebra is
the Lie algebra of the orthogonal group 0(n, R)).“
31
l LIE ALGEBRAS
32
lNVARIANT BILIN'EAR FORMS §3.6
33
I LIE ALGEBRAS
3 4»
CASIMIR ELEMENT § 37
over K. Let a be an ideal q, B the Killingform ofg and p’ the Killingform qf n. Then
3' is the restriction (f [3 to a.
Let a be an endomorphism of the vector space g which leaves a stable. Let u
be the ratriction of a to u and w the endomorphism of the vector space g/u
derived from a when passing to the quotient. Then Tr a = Tr v + Tr w as is
seen by taking a basis (xi, . . . , x") of g ofwhich the firstp demerits form a basis
of G. Then let x E a, y e a and apply the above formula to the case where
u = (adv x)(adiy). Thenu = (add x)(aduy) and w = O. Henccfl(x,y) =fl'(x,y).
PROPOSITION 10. Suppose that K is afield and that the Lie algebra g isfinite-dimen-
:ianal over K. The Killingfirm {5 q is tompletely invariant.
Let D be a derivation of 5. There exists a Lie algebra {5' containing g as an
ideal of codimension 1 and an element no of 9' such that Dx = [5%, x] for all
xsg (§1, no. 8, Example I). Let B’ be the Killing form of 9'. For x, y in 9,
[We may) = w, DMD, that is MUM) + rm, Dy) = 0. Now the res-
triction of {1’ to g is [3 (Proposition 9). Hence the proposition.
7. CASIMIR ELEMENT
PROPOSITION 11. Let g be a Lie algebra over afield K, U its enveloping algebra, b a
finite-dimenrianal ideal of g and B an invariant bilinearfirm on 9, whose restriction to b
is nan-degenerate. Let 00“.“, (4)191“ be two base: of I) such that we” e;) = 8”.
n
Then the element 5 = ”:1 ele,’ of U belong: to the tentre of U and is independent to the
choice of basis (e,).
For x e 9 let x» be the restriction to I) ofad. 1. Then a: r-> in, is a representation
of g on the vector space I) and the ratriction [5' of B to b is invariant under this
35
1 LE AmEBRAS
36
EXTENSION 01? THE BASE RING § 3.8
7w 5 K1 for 1 S k S d, the matrix off = 1:21 )1]:fk with respect to these bases
has determinant which is a polynomial D0,, . . ., Ad) with coefficients in K.
Whenf = g, this determinant is non-zero and hence the coefficients of D are
not all non-zero. Therefore, if Q is the algebraic closure ofK, there exists (since
{I is infinite) elements W E Q (l < k < 11) such that D(u1, . . ., M) at 0
(Algebra, Chapter IV, § 2, no. 5, Proposition 8). If K2 is the algebraic extension
11
37
I LIE ALGEBRAS
Recall that henceforth K denotes a commativefield. In the rest of the chapter the Lie
algebra: are assumed It: befinite-dimensional over K.
38
ENGEL’s THEOREM §4.2
Definition 1 and Proposition 2 show that nilpotent Lie algebras are pre-
cisely the ' L " ' J from ive Lie ' L bya
of central extensions.
2. ENGEL’S THEOREM
Lemma 1. Lo! V be a vector xpare over K. If x i: a nilpalent :ndumorphirm qf V, [he
mapping y H [1, y] If 3’(V) inlu E (V) ix nilpalzut.
Iff denotes this mapping, f"(y) is a sum of terms of the form 1 x‘yx’ with
i+j = m. Ifx" = 0, thenf“"(y) = Ofm' ally.
THEOREM 1 (Engel). Let V be a vector space over K and g a finite-dimensianal rub-
algebra 13f 91(V) whore elements are nilpatenl endumorphiamr af V. If V aé (0}, there
exirh'u aé 0inVruchthalx.u = Oforallxeg.
39
l LIE ALGEB‘RAS
40
THE. mom mum-nanny IDEAL or A REPRESENTATION § 4.3
41
I LIE AmEBRAS
PROPOSITION 5. Let V b: a motor spice offinite dimension n over K and g a Lie rab-
algebra of gl(V) whose elements an nilpotmt mdomorphimu cf V. Then there exists a
decreasing some of vector subspaces V0, V1, . i . , V" of V, of dimensions
a,» — l,...,0,.mchthalx(V.) C V,+,j‘orallxegandi = 0,1,...,n — 1.
xir((Mi)<x.)) c (Ml+1)(Kx)
for all i and all x’ e um); hence xfil is nilpotent for x' e "am so that “am is con-
tained in the largest nilpotency ideal n’ for M'. We shall now see that, if K1
1‘: separable over K, then n' = "(Kw Let E be the associative K-algebra generated
by l and the cm (x E g), E' the associative K-algebra generated by l and the
x“: (x' e g') and R and R' the jacobson radicals of E and E’. The algebra E’ is
canonically identified with Eam- Then R’ = Run) (Algebra, Chapter VIII,
§ 7, no. 2, Corollary 2 (c) to Proposition 3). Then let y' E n’ and write
42
DEFINITION or SOLVABLE LIE menus § 5.1
Recall that K henceforth denotes afield of characteristic 0 and that all Lie algebras are
assumed to befinite—dimensional over K.T
PROPOSITION 1. Subalgebras and quotient algebras ofa solvable Lie algebra are solvable.
Every extension ofa solvable algebra by a solvable algebra is solvable. Eueryfinile product
of solvable algebras is solvable.
Let g be a Lie algebra, 9' a subalgebra, b an ideal of g, I = g/b and b the
canonical mapping of 9 onto !. If g is solvable then 9kg = {0} for some integer
k, hence 9’? C 9kg = {0}, 9"! = Mflkg) = {0} and hence 5’ and I are
solvable. If I) and I are solvable there exist integers s, t such that
1L The reader will note that the hypothesis on the characteristic of K is not
used in nos. 1 and 2 of this paragraph.
43
LIE ALGEBRAS
44
NILPOTENT RADICAL or A LIE ALGEBRA § 5.3
9(99 n r) = {0}.
COROLLARY 1. Let g be a :alvable Lie algebra. The nilpaten! radical of g i: 99.
If p i: a finite—dimenxianal :imfle reprerentaliaa y" 5, 9(9) is commutative and the
axsaeiative algebra L generated by l and 9(9) 1': afield 1yrfinite degree over K.
45
l LIE AIDEBRAS
Herer = g,whence5 = 99. Hence 9(99) = (0}, which shows that g' = p(g)
is commutative. Every element #0 of L is invertible by Schur’s Lemma;
hence L is a field.
COROLLARY 2 (Lie’s Theorem). Let g be a solvable Lie algebra; suppose that K
is algebraically closed. Let M be a g-module 4y" finite dimension aver K and let
(MQD‘K, be a Jordan-Holder series LfM. Then M‘-,/M, is (y’dimension 1 over
Kfor l S i S rand,far all x59, 2“,t = A.(x).l, where). is a linearfonn on
g whieh is zero on 99. In particular, every simple g-module offinite dimension over K
is infaet of dimension 1.
Let p, be the representation of g on M‘d/M‘. The associative algebra
L1 generated by 1 and p,(g) is a field, a finite extension of K and therefore equal
to K; and M‘d/M. is a simple Isl-module, whence dim M._1/M. = 1. The
rest of the corollary is obvious.
Remarks. (1) If (MONK, is replaced by another Jordan-Holder series of M,
the sequence (M, . . ., 1,) is replaced bya sequence of the fom‘i (Am), . . ., Mm):
where 1: is a permutation of {1, . . ., r}, as follows from the Jordan-Holder
Theorem.
(2) Let (en . . ., e,) be a basis ofM such that e, e M,_1, el (7? M, (l S i g r).
If x e g the endomorphism of M corresponding to x is represented with respect
to this basis by a triangular matrix whose diagonal coefficients are
110‘), - - -, 7‘1(")~
46
A CRITERION FOR SOLVABILITY § 5.4
t he the radical of 9. Every element at 61‘ such that p(x) is nilpotent belongs to the
largest nilpoleary ideal n of 9.
Let V be the space of 9; let (VOar be a Jordan-Holder series for the
t—module structure on V and let p, be the representation of r with space
V./V,.l (l < i S r). If p(x) is nilpotent, so is 9.0:); as for all i the algebra
generated by p,(x) is a field, 9.0:) = 0. Conversely, if p,(x) = 0 for all i,
p(x) = 0. This shows that the set a of are: such that p(x) is nilpotent is an
ideal of r. On the other hand, [9, a] C 99 n t C n n t C a and hence a is
an ideal of g. This proves that a C n.
COROLLARY 7. Let g be a Lie algebra and t its radical. The following four sets are
identical: (a) the largest nilpoteat ideal of g; (b) the largest nilpatent ideal y” t;
(c) the set If x e r such that ad. 2: is ailpatent; (d) the set of x e r such that ad, 2:
is ailpateat.
Let these sets be denoted by a, b, c, b. The inclusions n C b C b C c are
clear. c C a by Corollary 6 applied to the adjoint representation of 9.
47
I LIE ALGEBRAS
lei = f (1,)“. If (Eu) is the canonical basis of gl(M) defined by Ene,‘ = 81kt”
then
(ad ‘)Eu = (’1 — ADE"
(ad’)Eu = (f0!) '—f(7v))Eu~
There exists a polynomial P with no constant term and with coefficients in
Ksuch thatPOq —— 1,) =f()..) —f(A,) foralliandj(forif)q -— A, = In — A,”
thenfO‘t) —f(7v) =f(7w) -f(7w) and, int - A, = 0,1111) —f(7v) = 0)-e
Then ad t = P (ad J). On the other hand, ad .r is a polynomial with no constant
term in ad 1. Now (ad 2) (A) C B, whence also (ad t)(A) C B. By the
hypothesis 0 = Tr(zt) = 2 MO“), whence 0 =f(Tr(zt)) = 21%).)?
Since the f(N) are rational numbers, f = 0, which completes the proof.
THEOREM 2 (Cartan’s criterion). Let g be a Lie algebra, M a finite-dimensional
vector xpaee, g a representation of g on M and B the bilinearfarm on g awaciated with
9. Then 9(9) 1': tall/able If and only ifflg is orthngtmal to g with respect to {3.
It can obviously be reduced to the case where g is a Lie subalgebm of
gl(M) and p is the identity mapping. If g is solvable, 99 is contained in the
largest nilpotency ideal of the identity representation of 3 (Theorem 1) and
hence is orthogonal to g with respect to [3 (§ 4, Proposition 4 (11)). Suppose
that 99 is orthogonal to g with mpect to B. We prove that g is solvable.
Let T be the set of t e 91(M) such that [1, g] C 99. If [ET and x, y belong
to 9, then [13, x] e 99 and hence
Tr(t[x:yl) = MUM) = 0
whence by linearity Tr(tu) = 0 for all u 695. Also, clearly 95 C T. Hence
(Lemma 3) every element of 99 is nilpotent. It follows that 99 is nilpotent
(§ 4, Corollary 3 to Theorem 1) and hence that g is solvable (no. 3, Corollary
5 to Theorem 1).
48
EXTENSION OF THE BASE FIELD § 5.6
t D I D n D 5.
49
I LIE MEBRAS
Let r be the radical of 9. Then rm) 2‘: the radical (J Eam- For let; [3 be the
Killing form of 9. As t is the orthogonal of Fig with respect to (3 (Proposition
5 (b)), for.) is the orthogonal of (99%“, = 9(Sam) with respect to the form
derived from [3 by extension from K to K,, that is the Killing form of 90(1)
(§ 3, no. 8). Our assertion then follows from a further application of Proposi-
tion 5 (b).
Reeall that K denote: afield afehamcteristie 0 and that all Lie algebra: are assumed
to be finite-dimensional over K.
50
SEMI-SIMPLICITY or REPRESENTATIONS § 6.2
COROLLARY 1. Let g be a Lie algebra, {i it: Killing firm and o a semi-ample .mb-
algebra of g. The orthogonal b of a with respect to B is a supplementary subspace of
a in g and [(1, 1)] C I). (f a it an ideal of 9, so is I), which i: then the eentralizer of a
1n 9.
Let B’ be the restriction of B to a: it is the bilinmr form associated with the
representation at» ad,x of a on the space g. This representation is faithful
and hence fl' is non-degenerate (Proposition 1). Hence I) is supplementary
to uin 9. 0n theotherhand, ifx,yarein aand 26!), then
2. SEW-SIMPLE)!“ 0F REPRESENTATIONS
Lemma 1. Let g be a “mi-simple Lie algebra. The adjoin! representativn vf g is remi-
rimple. Every ideal and every quatien! algebra {If g is semi-simple.
Let a be an ideal of g. The orthogonal b of a in g with respect to the Killing
form is an ideal of g and a n b is a commutative ideal (§ 3, no. 6, Proposition
7) and hence zero. Hence b is supplementary to a in 9. Moreover, as the
Killing form of g is non-degenerate, so are its restrictions to a and 1: (Algebra,
Chapter IX, §4, no. 1, Corollary to Proposition 1) and hence a and b are
semi-simple (no. 1, Theorem 1 and § 3, no. 6, Proposition 9).
Lemma 2. Let g be a Lie algebra. Then thefollowing two conditivnr are equivalent:
(a) All 5‘ ‘ " ' 'linear , ' «#9 are ' p’.
(b) Given a linear representation p qf g on a finite-dimmmnal vector space V and
51
I LIE ALGEBRAS
a vector mbrpaee W bf eodimemian 1 such that p(x)(V) C W for all x-E-g, there
exists a supplementary line of W which it stable under 9(9) (and heme annihilated by
g .
Méleatly (a) implies (b). Suppose that (b) holds. Let a be a finite—dimen-
sional representation ofg on a vector space M and N a vector subspace which
is stable under 6(9). Let u be the representation of g on J (M) canonically
derived from a (§ 3, no. 3): recall that pL(x) = adgunfire). Let V (resp. W)
be the subspace of $(M) consisting of the linear mappings of M into N
whose restriction to N is a homothety (rap. zero); then W is of codimension '
l in V and Ma) (V) C W for all xe 9. By condition (b), there exists a EV
which is annihilated by [1(a) for all xeg and whose restricu'on to N is_ a
non-zero homothety. By multiplying u by a suitable scalar, it can be as-
sumed that u is a projector of M onto N. To say that y.(x) .u = 0 means that
u is permutable with 6(a). Hence the kernel of a is a supplement of N in M
which is stable under a(x) for all x e 5. Hence a is semi-simple.
52
SEMI-SIMPLICH‘Y or REPRESENTATIONS § 6.2
The simple idmls of a semi-simple Lie algebra are called the simple cam-
[uments of g.
COROLLARY 2. Let g, 9’ be two Lie algebras, v and r’ their radical: andf a homo-
morphixm q onto 9'. Then I’ =f(r).
As f(r) is solvable, f(r) C t". On the other hand, g/t is semi-simple (§ 5,
no. 2, Proposition 3), hence g’/f(r), which is isomorphic to a quotient of
g/t, is semi-simple (Lemma 1) and hence f(t) D I’ (§ 5, no. 2, Proposition
3).
Remarks. (1) Theorem 2 admits a converse: if every finite-dimensional repre-
sentation of 9 is semi-simple, g is semi-simple. For since the adjoint representa-
tion is semi-simple, every ideal of g admits a supplementary ideal and hence
can be considered as a quotient of g. If g is not semi-simple then 3 admits a
non-zero commutative quotient and therefore a quotient of dimension 1.
Now the Lie algebra K of dimension 1 admits non-semi-simple representa-
tions, for example
A 0
53
I LIE ALGEBRAS
54
SEMI-SIMPLE 12m AND mwo'nmr stamens § 6.3
55
I LIE ALGEBRAS
56
imnmmvn Lm menus § 6.4-
57
I LIE ALGEBRAS
COROLLARY. Let 9, 9’ be Lie algebras, b (resp. 9') the nilpatent radical of 9 (resp.
g’) arqa homomorphism ofg onto g'.
(3) Then 5’ = f(s).
(b) 9’ ix reductive if and only if the kernel off contains 9.
COROLLARY 1. Let g be a Lie algebra and p and 9' two finite-dimensional remi-
simple representations of 9. Then the tensor product of p and p' is semi-simple.
Let r be the radical of 9. For x e r, p(x) and p’(x) are semi-simple (Theorem
4), hence p(x) ® 1 + 1 ® p’(x) is semi-simple (Algebra, Chapter VIII, §9,
Corollary to Theorem 1) and hence the tensor product of p and 9’ is semi:
simple (Theorem 4).
58
sumwsnm REDUCTIVE IN A LIE ALGBBRA § 6.6
59
l LIE ALGEBRAS
X
=(3 a) he 3) H=<s -3)‘
Then X, Y, H form a basis ofsl(K‘) and
60
EXAMPLES or SEMI-SIMPLE LIE awaam § 6.7
61
I LIE ALGEBRAS
fi'
ell = exp 11 = “a 1‘”
Now let E be a vector space over the field K and u a m'lpotml endomorph-
. . u" .
mm of E. The series "20 '7 has only a finite number of non-zero terms and we
can therefore write
"n
e“ = exp" = Zn
This definition agrees with the above if K = R and if E is complete and
normed. If v is another nilpotent endomorphism of E which commutes with
u, then:
»»=(i"“)ti),i—
n-ofi
a,
«217! mp!
m
4—. (Z. ("W = Z; <u + W =
l q ) I
WW) = ”5:7
2 (f)u'(x)u‘(y)
62
'n-IE LEVI-MALGEV THEOREM § 6.8
= ”2,0 # # = mm)
m u’ x u‘
DEFINITION 7. Let 9 be a Lia algebra and r it: radical. A Levi .mbalgebm 4y“ 9 is
any rubalgebm a g mpplmwntary to r.
63
I LIE ALDEBRAS
64
THE LEVI-minev THEOREM § 6.8
Then 91 and 9 are two Levi subalgebras of b and by the induction hypothesis
there exists b e m such that e‘”(91) = 6‘ Hence a = ew”.e““'(o’). Finally, as
m is in the centre of [g,r], :“”’.¢“" = cam”) and b + ae[g,t], which
completes the proof.
COROLLARY 1. Le! a be a Levi subalgebm 9/ g and b a semi-simple :ubalgebm of g.
(3) There exist: a :pea'al automorphism of g mapping 1} onto a .mbalgebm 9f 5.
(b) b is contained in a Levi subalgebra (jg.
Let r be the radical of g and a = b + t, which is a subalgebra of g. Then
a]: is semi-simple and r is solvable, hence t is the radical of a and I; is a Levi
subalgebra of a. On the other hand, a n 5 = b' is a supplementary subalgebra
to r in a and hence also a Levi subalgebra of 0. Then there exists (Theorem 5)
a e [a, r] such that Nu" maps 13 onto b’. Now a e [9, t]; Nu" maps 1; onto
a subalgebra of e and e"°u"(5) is a Levi subalgebra of 9 containing I).
COROLLARY 3. Le! g be a Lie algebra and m an ideal qf 9 such that g/m is semi-
:imple. Then 5 contains a .mbalgebm supplementary to m in g. In uthcr words, every
extension If a semi-simple Lie algebra is inmential.
Let a be a Levi subalgebra of 9 (Theorem 5). Its canonical image in g/m
is a Levi subalgebra and therefore equal to g/m, hence g = a + m. Then an
ideal of a supplementary in s to the ideal in n a is a subalgebra of g supple-
mentary to m in g.
COROLLARY 4. Let g be a Lie algebra, r it: radical, a a Levi :ubalgebm y” g and m
an ideal cfg. Then m is the direct mm q n t which is its radical and m n 5 which
i: a Levi .mbalgebm of m.
We know that m n e is the radical of in (§ 5, no. 5, Corollary 3 to Proposi-
tion 5). Let b be a Levi subalgebra of m and 5' a Levi subalgebra of g contain-
ing 9 (Corollary 1). The algebra m n 3’ is an ideal of 9', is therefore semi-
simple, and contains I) and is therefore equal to 1). Hence m is the direct sum
65
1 LIE menus
«comm = Ma~(6(~)n),
whence (a). Assertion (b) follows from (a) (Algebra, Chapter VIII, §3, no.
4-, Proposition 10). IfN = P = M and a = r = p, assertion (b) gives asser-
tion (0) as a special case.
Lemma 5. Suppose further that a and 1' are semi-simple and hence N (resp. P) 1'; the
direct rum 4/ the N5 (resp. P5). For all n E N (resp. p E P), let nll (resp. p“) be it:
component in No (resp. Po). Let ma E MD. Thenfor all n EN, (rno.n)'q = mom“.
By linearity it suffices to consider the case where n E No. If 8 75 80, n"‘ = 0
and momePb (Lemma 4), hence (7710.71)“ = 0 = mom". If 8 = 80, n" = n
and mmneli'o (Lemma 4), hence (mom)?! = mom = mom".
THEOREM 6. Let g be a Lie algebra, V a :emi-n'mfle g-modale qffinite dimension
over K, S the symmetric algebra of V and res the derivatian of S which extend: xv
(III that it )—> A's is a representative! (if g on S).
(a) The algebra So of invariant: af S is generated by a finite number (if elements.
(b) For every class 8 If simple reprexentation: (f g offinite dimension aver K, let
Sa be the iratypical component «J S of species 8. Then S5 is a finitely generated So-
module.
Let S C S be the ideal of elements of S with no constant term. Let I be the
ideal of S generated by S“ n S and let (x1, :2, . . ., :p) be a finite system of
66
THE INVARXANTS THEOREM § 6.9
.t is <n. As I e I, s = 4;) 5,51, where the J,’ are elements of S which can be
assumed to be homogeneous, with deg(s") = deg(:) — deg(.r.) < 11. Lemma 5
can be applied, as the g-module S is semi-simple (no. 5, Corollary 2 to Theorem
4-); in the notation of this lemma,
1’ 9
The 5;“ are demerits of So which are homogenq and of degree <n (since
each Sn is a graded submodule). Hence theyare in 81 by the induction hypoth-
esis. Hence I 6 SI, which proves (a).
We now consider a simple representation of g of class 8 on a finite-dimen-
sional space M. Let L = 544M, S). For all 5 ES and all L, let {fbe the
element of L defined by (if/Mm) = :.f(m) (m E M); an S-module structure
is thus defined on L; as M is finite-dimensional over K, clearly L is a finitely
generated S-module and hence a Noetherian S-module since the ring S is
Noetherian. On the other hand, L has a canonical g-module structure. For
every integer n 2 0 let S" be the set of homogeneous elements of S of degree
n; then the g-module 3AM, S“) is semi-simple (no. 5, Corollary 3 to Theorem
4) and hence the g-module L is semi-simple. Moreover, for .t E S, f e L,
x e g and m e M,
67
1 Lu! ALGEBRAS
q
fn...,f¢. Then Ll C Lo. On the other hand, iffeLo, thenf= .2141];
with I. E S for all i and hence by Lemma 5 whose notation we adopt: 7
‘1 H a
68
THE EXTENSION THEOREM § 7.2
l. COEFFICIENTS OF A REPRESENTATION
Let U be an associative algebra with unit element over K, U“ the dual of
the vector space U and p a representation of U on a vector space E. For
2 e E and c’ e E‘, let 0(2, 3’) E U* be the corresponding mgfinlmt of 9 (Algebra,
Chapter VIII, § 13, no. 3). Recall that 0(e, :’)(x) = (900:, z’) and that the
mapping 9 >—> 0(2, 2') is for fixed 6’ a homomorphism of the U-module E into
the U-module U“ of the coregular representation of U ([00. 011., Proposition
1); therefore the vector subspace of U“ generated by the coefficients of p
(a subspace which we shall denote by C(p) in this paragraph) is a sub-U-
module of U*. If (51),“ is a family of elements generating E“ over K, the
mapping 2 ~—> (0(9, 22)) is an injertive U-homomorphism of E into C(p)‘, for
0(:, 2;) = 0 for all 1' implies (0,4) = <p(l)e,efi> = 6(5, 90(1) = 0 for all i
and hence a = 0.
In particular, if U is the enveloping algebra of a Lie algebra g and p is a
representation of 9 (identified with a representation of U) on an n-dimen-
sional vector space E, the g-module E is isomorphic to a sub—g-module of
(0(9))"-
2. TH}: EXTENSION THEOREM
Let g = I; + g' be a Lie algebra which is the direct sum of an ideal g’ and
a subalgebra I), U the enveloping algebra of g and U’ C U the enveloping
algebra of 9’. There exists one and only one 9-module structure on U’ such
that: (a) for x s 9' and u eU', xwu = —ux; (fl) for x5!) and u e U’,
xuru = 1m — wt (the latter element is certainly in U’ since the inner deriva-
tion of U defined by x leaves 9’ and hence U’ stable). For conditions ((7.) and
((3) define uniquely a linear mapping max", of g into S’K(U’). It therefore
sufiices to Verify that [x, 31h. = [arm guy]; it is only necessary to consider the
following cases:
(1) x E g’,y e 9’: then
69
I LIE ADGEBRAS
70
Aoo’s THEOREM § 7.3
Lemma L Let g = 9’ + l) be a Lie algebra which i: the rum qfan ideal g’ and a
mbalgebm I). Let a be a finite—dimensional representation cy‘ g. Suppose that u-(x) i:
nilpotentfor all x e g’ and all x E 9. Then 6(2) is nilpotentflzr all x e g.
Passing to the quotient by the kernel of a, a may be assumed to be faithful.
Then 5’ and l) are nilpotent and hence g, which is an extension of a. quotient
of b by g', is solvable. Then 1) and g’ are contained in the latgect nilpotency
ideal ofa‘ (§ 5, no. 3, Corollary 6 to Theorem 1).
3. ADO’S THEOREM
PROPOSITION 1. Let g be a Lie algebra, n its largest nilpatent ideal, a a nilpatent
ideal of g and p afinite—dimerm‘mal representation of a mh that every element of p(a)
Lt nilpotent. Then p admit: afinite-dimensional extension 0" to g such that every element
of 6(a) is nilpotent.
Let a = no C n, C - - - C n, = n be a sequence of subalgebras of“ such
that "(.1 is an ideal of n. of codimension l for 1 S i S [1 (§ 4, no. 1, Proposi-
tion 1 (e)). The algebra n, is therefore the direct sum of 1a-, and a l-dimen-
sional sub-algebra. As adnx is nilpotent for all x e II, it is possible (Theorem l)
to find one by one finite-dimensional extensions 9;, pg, . . ., pp = p’ of p to
n1, n2, . . ., II, = n such that every element of p’(n) is nilpotent.
Let r be the radical ofg and let tt = to C r, C - ~ ~ C I. = r be a sequence
of subalgebras of r such that 1}_1 is an ideal of I; of codimension 1 for
l s 1' $ q (§ 5, no. 1, Proposition 2 (d)). The algebra II is thus the direct
sum of r1_l and a l—dimensional subalgebra. As [r,t] C II, it is possible
(Theorem 1) to find one by one finite-dimensional extensions
9’1, pg, . . ., p; = p” of p' to r,, 12,. . ., ta = r such that every element of
p”(n) is nilpotent.
Finally 9 is the direct sum of r and a subalgebra 3 (§ 6, no. 8, Theorem 5).
As [3, r] c n, it is possible (Theorem I) to find a finite-dimensional exten-
sion 6 of p” to 9 such that every element of o(n) is nilpotent.
71
! LIE AMEBRAS
72
EXERCISES‘
§1
1. Let g be a Lie algebra and a and b ideals (rap. characteristic ideals)
of g. Then the set of: E 9 such that [3, b] C a is an ideal (resp. a characteristic
ideal) of 9 called the hampwter of b into a. Show that «“19 is the transporter
of g into «,9.
2. If g is an n-dimensional Lie algebra over a field K and the centre of
g is of dimension 2n — l, g is commutative.
3. Let M be a not necessarily associative algebra over a field K, (9‘)“,
a basis of the vector space M and cm: the constants of structure of M relative
to the basis (2‘). For M to be a Lie algebra, it is necessary and sufficient that
the ow satisfy the following conditions: (a) ”me = 0; (b) 5% = —c,.,‘; (c)
73
I LIE ALGEBRAS
74
EmCISFJ
11. Let g be a Lie algebra and I) an ideal of 9 such that 5% = b. Show that
I) is a characteristic ideal of g.
12. Let g be a Lie algebra.
(a) For a derivation D of g to commute with all the inner derivations of
9, it is necessary and sufficient that D map 9 into its centre.
(b) Suppose henceforth that g has centre zero, so that g is identified with
an ideal of the Lie algebra 9 of all derivations of 91 Show that the centralizer
of g in $9 is zero (use (11)). In particular, the centre of 9 is zero.
(0) Show that a derivation A of 9 such that A(g)= {0} is zero.
E[)A§@), 9] C Aug, 9]) + [9 MED] C A(£1) = {0} and hence A(‘9)= {0} by
((1) Show that, if further 9—
= 99, every derivation A of 9 is inner (use
(5) and Exercise ll).
134 Let g be a Lie algebra and a and b two submodules of 9. For every in-
teger i 2 0 we define the submodules m. = m‘(a, b) and m = "‘01, b) as
follows: m1 = b, rnH1 = [m‘, b], 111 = a, n“r1 = [11” b]. Show that
[0, m] C "m
For i = l, 2, . . . (argue by induction on i, observing that
"Ml“, 5], b) = "110‘. 1’)
and ME“, b]: b) = mm, b))-
1T 14-. Let a be a Lie algebra and b a subalgebra of a. A campadtion sm'u
joining a to b is a decreasing sequence (“Quin of subalgebras of a such that
:1n = a, c1,l = b, a‘+1 is an ideal of a, for O < i < n. b is called a :ubinvuriant
subalgebra of u if there exists a composition series joining a to b.
(a) Let b be a subinvariant subalgebra of a and 010“,“ a composition
series joining a to b. Deduce from Exercise 13 that [W‘wlg a] C ’fl’b by ob-
serving that [an b] c «”1 for l s i < m
(b) Deduce from (a) that the intersection @“b of the if"? (p = 0, l, 2, . . .)
is an ideal of a.
(o) Deduce from (b) that a subinvariant subalgebra c of a such that c = 9:
is an ideal of a.
(d) When K is a field and dimKn < +00, show that the intersection
9“!) of the 9% (p = 0, 1, 2, r r .) is an ideal of a. (Show that this intersection
is a subinvariant subalgebra of a and apply (5).)
11 15. (a) Let a be a Lie algebra, b a subalgebra of n, c an ideal of b and
3 the centralizer of c in n. Then [3, b] C 3. Deduce that b + A is a subalgebra
of a in which 3 is an ideal.
(b) Let g be a Lie algebra, a a subinvariant subalgebra of g and b an
ideal of a. If the centralizer of b in a and the centralizer of a in g are zero,
75
1 LE ALGEBRAS
76
EXERCISES
A,(x,y>= 2, (p — 0- g. .-1--.(adx.ady) y
belongs to the Lie subalgebra of L generated by x and y (“Ja€ab:on’:finnulaz”).
20. Let g be a Lie algebra over a ring K such that pK = {0} (p prime).
A mapping maxi" of 9 into itself is called a p—mapping if it satisfies the rela-
tions
ad 1“" = (ad x)’
(MOM = NW”
78
EXERNSE
a system (u,(x, y, 2)) I g, Q, ofn elements of.” will be denoted by u(x, y, z);
analogous notation will be used for formal power series containing only one
or two of the three systems x, y, 1. If u(x, y, z) EA,_,,_z and f = 0;), g =
(g), h = (11,) are three systems of n elements of Ax,” which are formal
power series with no constant term, we denote by u(f(x,y, z), g(x,y, z),
h(x, y, 1)) the {urinal power series obtained by substituting fi for X1, g, for
Y” )1, for Z, (1 < i < n) in It. For every system at = (on, . . ., at“) ofn integers
20, let x“ denote the monomial X}: . . .i; y“ and z" are defined simi-
larly‘ Let 2‘ denote the system a: for which at, = 0 ifj ye i, on = 1. We denote
by e the system (0, . . . , 0) of n elements of Am”.
(a) A foimal gnu}? [am over K (or, by an abuse of language, a formal group
over K) of dimension n is a system G = f(x, y) of n elements of A,” with
the following properties: (1) {(x, f(y, 1)) = f(f(x,y),z); (2) He, y) = y,
f(x, (5) = x. Show that necessarily fi(x, y) = XI + Yi + g.(x, y), where g.
contains only monomials of degree >2, each of which contains at least one
X, and at least one Y,-. Show that there exists one and only one system h(x)
of n elements of A, such that f(x, h(x)) = f(h(x), x) = e (Algzbra, Chapter
IV, § 5, no. 9, Proposition 10). G is called commutative if f(y, x) = f(x, y).
(11) For all u 5A,, let L,u denote the element u(f(y, x)) of Ax”. A deriva-
tion D of can be canonically extended to a derivation MA,” (also denoted
D by an abuse of notation) by the condition that D(Yi) = 0 for all 2' (Algebra,
Chapter IV, §5, no. 8, Proposition 6). D is said to be left invariant for the
formal group G in question if L,.D = BL," We denote by D“) the linear
mapping of Amy into A, defined by D‘°’(u) = (Du) (e, y); it maps A.x into
K and is determined by its restriction Ax. Show that, for D to be left invariant,
it is necessary and sufficient that for all u E A,, D‘°’(L,v) = (Du)( y).
Let D. be the derivation a/ax. of Ax (1 g 1' < u); there exists one and
only one left invariant derivation T‘ such that T5” = DP. Show that the
T, are linearly independent over K and that every left invariant derivation
is a linear combination of the T. with coeflicients in K. Deduce that, for
the bracket [D, D’] and the p-mapping D |—> D” (when [J > 0), the set 5; of
left invariant derivations is a Lie algzbra (a Lie p-ulgebnz if 1; > 0) called the
Lie algzbra 11f theformal Lie group G.
(6) Show that for every formal power series 14 2 AK:
where in the series as all the monomials are of degree 23 in the set of Y‘ and
2‘. Deduce that:
whose group law is defined byf,,(u, v) = —8‘, + 2:1 (8”,: + Um)(8k, + V”)
(8,, the Kronecker index; the fin2 indeterminate: of the general definition of
Exercise 24- are here denoted by U”, V”, Wu with 2 indices varying from 1 to
n). Show that, if D” = 6/6Uu, the left invariant derivations X” such that
Xi;’ = D}? are given by
81
I LIE ALGEBRAS
and let L be a field containing the ring of formal power series K[[Si,, S,’,, SM].
If we denote by S, S’, S” the matrices
2 sum, — En),
I<l «<12 Sam — fl), 2, saw“ — Efl),
then det(I —— R"S) aé 0 and the analogues for S’ and S” hold; ifwe write
U = (I — R“S)"(I + R'IS), U' = (I — R“S’)“(1 + R'ls’),
then also det(I + UU') aé 0. We write
F(S,S') = R(UU’ — I)(UU' + I)’1 = (fi,(S,S’));
the fi,(S, S’) belong to the ring of formal power series K[[S‘,, S{,]] and, by
consideringF (S, S’) as a. formal power series with coefficients in the algebra of
matrices of order n over K, we can write:
82
EXERCISES
trary (commutative) field and show that its Lie algebra is identified with the
subalgebra of 91(2n, K) consisting of the matrices X such that
0 1».
‘XA+AX=0,whe1-e A:( —I,| 0
27. Let K be a field of characteristic p > 0 and G the 2—dimensional formal
group defined b1(xx Y) = xi + Y: + X1Y1:f2(x) Y) = X: + Yaa + Xl')
Show that G is not commutative but that its Lie algebra is commutative.
§2
1. We adopt the notation T, J, T,I of Definition 1 and no. 6. Let t be a
homogeneous tensor of T of order p and a a permutation of{l, 2, . . . , [1}. Then
I — at EJ + T,_1. (Reduce it to the case where a is a transposition of two
consecutive integers.)
2. Suppose that K is a field. Let g be a Lie K-algebra and U its enveloping
algebra.
(a) Let u, v be in U. Ifu is of filtration > n, and u of filtration >11, then uv is
of filtration >71 + 1) (use Theorem 1).
(b) Deduce from (a) that the only invertible elemenm of U are the scalars.
(c) Deduce from (b) that the Jacobson radical of U is zero.
3. Suppose that K is a field. Let g be a Lie K-algebra and U its enveloping
algebra. The left regular representation of U corresponds to a representation p
of g on the space U. Show that the set U+ of elements of U with no constant
term is stable under p and that U+ has no supplement in U stable under 9. In
particular, 9 is not semi-simple. (This will be considered again in Theorem 2 of
83
1 LE ALGEBRAS
)t r—> A”) of 9 into the centre of U. Deduce that if (bA) is a basis of g, for there to
exist a p—mapping on 9, it is necessary and sufficient that, for all A, there exist
t,_ e 9 such that (ad bA 9 = ad 01; if this is so, there then exists one and only one
[I—mapping x >—> x!“ such that bi" = c. for all )t.
6. Let g be a Lie p—algebra over a ring K such that 11K = (0} (fl prime), U
its enveloping algebra and o' the canonical mapping of 9 into U. LetJ be the
two-sided ideal of U generated by the elements (c(x))" —< a-(xm) where 5: runs
through 9. The associative algebra U = U/_] is called the ”mined enveloping
algebra of g. The mapping cr defines when passing to the quotient a mapping 6-
(called canonical) of 9 into U, which is a p—homomorphism (when U is con-
sidered as a Lie p-algebra).
(a) The algebra U and the mapping 3 are solutions of a universal mapping
problem: for every p-homomorphismf of 5 into an associative algebra B over
K (considered as a Lie p-algebra), there exists one and only one K-homomor-
phism f' of U into B (with their associative algebra structures) such that
= ’ c 6-,
(b) Show that, if 01),,“ is a basis of 9 (where A is totally ordered), 5 is
injective and that, if x and 5(3) are identified for x e g, the elements l'lAbX‘
(where the A are in increasing order, the vi are zero except for a finite number
and 0 S v,_ < p for all A) form a basis of U. (Canonimlly identifying g with a
submodule of U, we write (Hz) = x’ —— x1" for all x 5 9. For every composite
index at = (05,) EN“), let at,‘ = 9,, + [In with 0 $ 5» < [I and let
Ta = (HibX*)(Ux(<l>(bx))“)~
Show that the Ta form a basis of U and the Ta such that Y = (7,.) 96 0 form a
basis of]; observe that the Mb») belong to the centre of U.)
7. Let g be a Lie p-algebra over a ring K such that [K = {0) (p prime). A
derivation D of g is called a p-derivalion if D(x’) = (ad x)"1. Dx for all x e 9.
Every inner derivation is a p—derivation.
(a) If L is an associative K-algebra, every derivation of L is a p—derivation
when L is considered as a Lie p-algebra. (Use formula. (2) of Exercise 19 of
§ 1.)
(1;) Suppose that g has a basis. For a derivation of 9 to be a fi-derivation, it
is necessary and sufficient that it can be extended to a derivation of the res-
tricted enveloping algebra of 9. Deduce that the p—derivations of 9 form a Lie
p-subalgebra of the p-algebra of derivations of g.
(c) If D is a p—derivation of g, the kernel of D is a p-subalgebra of g.
(d) For every derivation D of g, D(x’) — (ad x)”“. D): belongs to the
centre of g for all M; 9 (use formula (2) of Exercise 19 of § 1, applied to
L = 3%))-
8. Show that Theorem 1 remains valid if the module 9 is a direct sum of
84
EXERCISES
§3
I. Let g be a Lie algebra, p a representation of g on a K-module M and u-
the associated representation of g on the tensor algebra of M. Show that the
submodule of symmetric tensors and the submodule of skew-symmetric
tensors are stable under a.
2. Let g be a Lie algebra, p a representation ofg on a K—module M and a the
associated representation of g on Q = .9’(M, M; M). Forf e Qto be invariant
under a, it is necessary and sufficient that the 9(x) be derivations of M with the
multiplication defined byf.
3. Let V be a finite-dimensional vector space over a perfect field K.
(a) The identity representation of gl(V) defines canonically representations
9 a
85
I ”E AWEBRAS
86
mom
87
I LIE ALGEBRAS
+ Z (—1>”1(x.)..f<x., . . ., . . ., x...)
88
WISHES
$1H"(9a M) —>H*(9,N)
which is said to be associated with (I).
(f) Let L be a sub-g-module of M and N the quotient g-module M/L. The
canonical homomorphisms L —‘> M —p> N define homomorphisms
90
EXERCISES
by the characteristic of K, show that H”(g, M) = {0} for all 1;. (Using Proposi-
tion 12, show that I‘ is an automorphism of C‘ (9, M) and hence induces auto-
morphisms of Z’(g, M) and B’(g, M). For
§4
The conventions of § 4 remain valid unless otherwise mentioned.
1. Let g be a nilpotent Lie algebra and p (resp. q) the smallest integer such
that V’g = (0} (resp. Veg = 9). Show that}: = q + l and that fig 3 VV‘Q.
(Use the argument of Proposition 1.)
2. Let g be a semi-direct product of an algebra b of dimension 1 and a com-
mutative ideal 9’. Let x e I), x aé O, and u be the restriction ofad, x to 5’.
(a) For g to be nilpotent, it is necessary and sufficient that u be nilpotent.
(b) For the Killing form of g to be zero, it is necessary and sufficient that
Tr(u2 = O.
(c) )Deduce from (a) and (b) that there exist non-nilpotent Lie algebras
whose Killing forms are zero.
(:1) Deduce from (a) that in a nilpotent Lie algebra such that 9”"9 $5 (0},
(Pg = {0) it is possible that «19 96 V’"g.
3. (a) Let g be a nilpotent Lie algebra, 3 its centre and b a non—zero ideal
of 9. Show that 3 n b aé {0}. (Consider b as a 5-module by means of the adjoint
representation.)
(b) If in a Lie algebra g an ideal b is contained in (Sing but not in fig,
show that the ideals b n 9kg are all distinct for 0 s k g i + 1. (Apply
(0-)
4. Let g be a nilpotent Lie algebra.
(a) Every subalgebra of g is subinvariant. (Use Proposition 3.)
(17) Let b be a vector subspace of 9 such that b + 99 = 9. Show that the
subalgebra of 9 generated by b is 9. (Apply (a) to this subalgebra. Reduce it
thus to the case where b is an ideal of g and use Proposition 4 of § 1.) Deduce
that the minimum number of generators of g is dim 9/99.
91
[ LIE ALGEBRAS
92
EXERCISES
93
I LIE ALGEBRAS
94-
EXERCISES
(a) Let V be the vector space K[X]. Let D be the differential operator
with respect to X on V and M the operator of multiplimtion by X on V.
Show that if K is of characteristic 0 the mapping
M1+s+ixaHaD+PM+Y
is an infinite-dimensional simple representation 9 of g on V.
(b) If K is of characteristic [1 > 0, the ideal (X?) of K[X] is stable under
p(g). Under the quotient representation of g on K[X] /(X1’), no line is stable.
18. Let g be a 7—dimensional vector space over K with a basis (2,)“.fl.
An alternating bracket is defined on g by the fonnulae:
95
I LIE ALGEBRAS
and that the transporter ofWg into @‘g is :22 p Deduce that every deriva-
8
96
EXERCISE?
97
l LIE ALGEBRAS
24. Suppose that the field K is of characteristic 2. Show that in the nil-
potent Lie algebra g, of Exercise 9 there mtists no 2-mapping.
25. Let g be a Lie p—algebra. Show that the largest nilpotcnt ideal of g
is a p—ideal (cf. § 1, Exercise 22).
26. Let G be a group and let (H,,),,>1 be a decreasing sequence of sub-
groups of G; suppose that H1 = G and that ifwe write (x,y) = xyx"y",
the relations at E H‘, g e H, imply (x, y) e H”).
(a) Let G, = Ill/HM“ show that G‘ is commutative and that the mapping
1, y1—> (x, 1/) defines on passing to the quotient a Z-bilinear mapping of
G‘ X G, into G,”
98
EXERCISES
§5
The conventions of § 5 remain valid unless otherwise mentioned.
1. Let g be the 2-dimensional non—commutative solvable Lie algebra. Show
that the Killing form of g is non-zero, that every invariant bilinear form on g
is degenerate and that every derivation of g is inner.
2. (a) Show that in the 3-dimensional solvable Lie algebra over R defined
by the multiplication table [x,y] = z, [x, z] = —y, [3], z] = 0, there exists
no decreasing sequence of ideals of dimensions 3, 2, l, 0.
(b) Show that in the non-commutative 2-dimensional solvable Lie algebra
9 there exists a sequence of ideals of dimensions 2, l, 0 but that g is not nil-
potent.
3. Let g be a solvable Lie algebra such that the conditions x e g, y e g,
[[x,y], y] = 0 imply [94,31] = 0. Show that g is commutative. (Let k be the
largest integer such that 9""5 yé {0}, 9kg = {0}. Assuming k 2 2, show
first that [9"‘25, 9"“9] = {O} and then that [9"‘29, 9“”5] = {0}, whence
a contradiction.)
4-. Show that the centre of st(n, K) is zero and that that of n(n, K) is of
dimension 1.
5. Let g be a Lie algebra and (9°55, Q‘g, . . ., 9‘9) the sequence of derived
algebras of g (n 2 0, 9"“9 95 9’5). Then dim 9‘5/9‘“g 2 2"1 + l for
1 g i < n — 2. (Taking quotients by 9‘9, reduce it to the case where g
is solvable. Then use the fact that 99 is nilpotent and Exercise 7 (c) of § 4-.)
6. (a) Verify that the following multiplication table defines a 5-dimen-
sional solvable Lie algebra g:
["1: x2] = x5, ["n "al = 7‘s: [32: ”4] = *4
["n’fil = [xmxa] = [huh] = [xv 9] = 0~
(11) Show that the orthogonal of g with respect to the Killing form is
99 = K33 + Kx, + s. Deduce that 99 is the largest nilpotent ideal of g.
(2‘) Show that there exists no supplementary subalgebra of 99 in g. Deduce
that g is not the semi-direct product of a commutative algebra and a nil-
potent ideal. (Show that this nilpotent ideal would necessarily be 99.)
7. Let 9 be the 3-dimensional solvable Lie algebra with basis (at, y, z) such '
that [x,y] = z, [x, z] = y, [y, z] = 0. Show that the linear mapping which
maps x to —x, y to —z, z to y is an automorphism of g of order 4. Compare
this result with Exercise 2i (r) of § 4.
8. (a) Let go be a 3-dimensional solvable real Lie algebra such that Ego is
commutative and 2-dimensional. Let g be the algebra derived from 90 by
99
l LIE amen“
extending the base field from R to C. For x6 9, let u,‘ be the restriction of
adfix to 95;. Show that the eigenvalues of 11,, either have the same absolute
value or are linearly dependent over R. (We have x = M + z with z E 9g,
y E 90, A E C and hence ux = My. Now an is the C—Iinear extension to 95 of
an R-linear endomorphism of 990.)
(b) Show that there exism a 3—dimensional solvable complat Lie algebra
g, with 99 commutative and 2-dimensional, and an element x of g such that
the restriction of ad"): to 99 has eigenvalues which neither have the same
absolute value nor are linearly dependent over R. (Construct g as a semi-
direct product of a I-dimensional algebra and a 2~dimensional commutative
algebra.)
(a) Show that the algebra constructed in (1)) cannot be derived from a
real Lie algebra by extending the base field from R to C.
100
EXERCISES
is the direct sum of two non-trivial subspaces which are stable under 9(9).
Then argue by induction on the dimension of V.)
13. Let g be a Lie Algebra and 9 the Lie algebra of derivations of 3. For g
to be characteristically nilpotent, it is necessary and sufficient that 9 be
nilpotent and that dim g > 1. (To see that the condition is sufficient, write
g as a sum of subspaces 9", applying Exercise 12 to the identity mapping of
9. Show that [93", 9“] C 3““ and that each 3“ is an ideal of g. Deduce that
9" is commutative for A aé 0. Using again the fact that Q is nilpotent, show
that g = 9” if dimg > 1. Otherwise g = 9° x b, where b is commutative.
Show first that dim I) < 1. If dim b = 1, note that there would exist a deriva-
tion D of 9 such that D(g°) = {0) and D(b) is contained in the centre of 3‘7
(§ 4-, Exercise 8 (a)))
14. Let V be a finite-dimensional vector space over K and z an endo-
morphism of V. We adopt the notation z: of Exercise 3 of § 3. An endo-
morphism z’ of V is called a ”Mica of z if for all [I and 41 every zero of z: is a
zero of 2;”. Show that if Tr(zz') = 0 for every replica z’ of z, then 7. is nil-
potent. (Use the proof of Lemma 3. In the notation of this proof, prove in
particular that tis a replica of z.)
15. Let K be a field of characteristic 2. The identity representation of the
nilpotent Lie algebra 91(2, K) on K2 defines a semi—direct product is of
51(2, K) by K“. Show that b is solvable but that 90 is not nilpotent. Deduce
that 9 admin no faithful linear representation by triangular matrices. Show
also that the conclusions of Exercise 5 are false.
16. Let g be a Lie algebra over an arbitraIy field K, A a commutative
associative algebra over K and 9’ = g ®K A, which can be considered as a
Lie algebra over K.
(a) If D is a derivation ofA, show that there exists one and only one deriva-
tion D’ of 9’ such that D’(z ® a) = x ® Du for xeg, a GA.
(b) Let p be a prime number, G a cyclic group of order p and s a generator
of G. Suppose that K is of characteristic p and henceforth let A be the algebra
of G over K. Show that there exists one and only one derivation D of A such
that D0") = k:"'1 for all 1:52. Show that the K-linear combinations of
the x® (: — l)” (k =1,2,...,p —— l,xeg) form a solvable ideal I of 9’
and that 9’]: is isomorphic to g.
(0) Take 9 to be simple (cf. §6, no. 2, Definition 2). Then i: is the radical
ofg' butis not a characteristic ideal. (Observe thatD(x (8) (.r —- 1)) = 1 ® x.)1‘
17. Let g be a Lie algebra. Suppose that for every simple g-module M of
finite dimension over K the x” are permulable with one another. Show that
g is solvable. (Observe that 99 is contained in the nilpotent radical and hence
is solvable.)
101
l LIE ALGEBRAS
§6
The conventions of § 6 remain valid unless otherwise mentioned.
fl1.Letgbea " l.‘ Lie ‘3‘“ andpafi" " ' ‘repre-
sentation of g on M.
(a) If p is simple and non-zero, H’(g, M) = (0) for all [1. (Use Proposition
1 and Exercise 12 (j) of§ 3.)
(b) For all p, H’(g, M) = {0). (If p is simple and non-zero, apply (a).
If p is zero, use the fact that g = 95. In the general case, argue by induction
on the dimension of 9: if N is a sub—g-module of M distinct from {0} and M,
use the exact sequence:
H‘(5, N) -> H‘(9, M) -> H‘(9, MIN)
established in Exercise 12 (f) of § 3.) Hence recover Remark 2 of no. 2.
(c) Deduce from (b) a proof of Theorem 2. (Use Exercise 12 (g) of § 3.)
Deduce also from (b) that every derivation of g is inner. (Use Exercise 12 (h)
of § 3.)
(d) For all p, H’(g, M) = {0). (Arguing as for ([7), it suffices to consider
the case where p = 0. Let 56H2(9, M). Consider, following Exercise 12 (i)
of § 3, the Central extension b of g by M defined by r. The adjoint representa-
tion of 9 defines a representation of g on b. By Theorem 2, M admits in l)
a supplement which is stable under 5 and hence the extension is trivial.
Hence 0 = 0 by Exercise 12 (i) of§ 3.)
(e) Deduce from (b) and (d) a proof of Theorem 5. (As in the text, it can
be reduced to the case where the radical is commutative.)
2. Let g be a Lie algebra, r its radical and (a0, a,, . . .) a sequence of ideals
of 9 defined as follows: (1) “a = {0); (2) «(H/ui is a maximal commutative
ideal of 5/4.. Letp be the smallest integer such that a, = a,“ = - . ., Show
that r = a,. (Show that 9/0,, is semi-simple.)
3. For a Lie algebra g to be semi-simple, it is necessary and sufficient that
it be reductive in every Lie algebra containing 9 as a subalgebra. (If g satis-
fies this condition, let M be a g-module of finite dimension over K. Consider-
ing M as a commutative algebra, form the semi-direct product of g and M,
in which 9 is reductive. Dcduce that M is semi-simple.)
4. Let g be a semi-simple Lie algebra and p a non-zero simple representation
of g on a finite-dimensional space M. Let b be the corresponding semi-direct
product. Show that I) = 91), that the centre of b is zero and that I) is not
the product of a semi-simple algebra and a solvable algebra.
5. Let n be a Lie algebra and tits radical. If t has a decreasing sequence of
characteristic ideals r = to D t, D - - - D tn = {0} such that dim rl/r,+1 = 1
for 0 < i < n, g is the product ofa semi-simple algebra and a solvable algebra.
102
EXERCISES
7. (a) Let g be a Lie algebra, b a semi-simple ideal ofg and a the centralizer
of I; in 9, so that g is identified with I) x a. Show that, for every ideal I of 9,
then I = (I n b) X (1 n a)‘ (let ll be the canonical projection of 1 onto 1;;
it is an ideal of b and hence 9h = I]; deduce that II C t.)
(b) Let [3 be an invariant bilinear form on 9. Show that b and a are ortho-
gonal with respect to [3 (use the fact that I) = 9b and that {5 = (31 + {35,
where B, (resp. {52) is an invariant bilinear form whose restriction to a (resp.
b) is zero.
(a) Deduce from (a) that there mists in g a largest semi-simple ideal.
(Consider a maximal semi-simple ideal of g.)
8. Let g be a Lie algebra. An ideal I) of g is called minimal if b aé {0} and
every ideal of 5 contained in b is equal to {0} or b.
(a) Every simple ideal of g is minimal.
(b) Let b be a minimal ideal of g and t the radical of 9. Then either I) C r,
in which case b is abelian, or b n r = {0}, in which case I) is simple. (Use the
fact that the derived ideal of a Lie algebra and the simple components of a
semi-simple Lie algebra are characteristic ideals.)
9. For a Lie algebra g to be reductive, it is necessary and sufficient that its
centre c be equal to its largest nilpotent ideal. (If the condition holds, let I:
be the radical of 9; 9r is contained in the centre of t, hence I: is nilpotent
and hence r = c.)
10. Let g be a Lie algebra such that the conditions at e g, y e 9, [[1, y], y] = 0
imply Deg] = 0. Show that g is reductive. (Show that the radical r of 5 is
commutative using Exercise 3 of § 5. Then show that [9, r] = 0.)
ii 11. (a) Let g be a Lie algebra, 1: its radical, o a Levi subalgebra of g
and m an ideal of g containing r. There exists an idea] i of a, supplementary
to m in 9, such that [111,1] C r. '
([7) Let g be a Lie algebra and a a subinvariant subalgebra of 9. Then
there exists a composition series 9 = 9., :> 51 3 ~ - - D 9,: 2 n such that g, is
the direct sum of g”, and a subalgebra I)i which is either l-dimensional and
contained in the radical r(g,) ofg. or simple and such that [9,, g‘H] C t(g.H).
(Reduce it to the case where u is an ideal of 9. Let 9’ = g/a and r’ be the
radical of 9’. The algebra g'lr’ is the product of its simple ideals n1, «2, . . ., etc.
103
! LIB ALGEFRAS
104-
EXERCISES
105
I LIE ALGEBRAS
106
EXERCISES
algebra 9K2, K) = g the bilinear form associated with the identity repre-
sentation is non-degenerate, but that the centre 3 is contained in 99 and is
95 {0}-
(0) Let g be the 6<dimensional Lie algebra over K with basis (a, b, c, d, e,f)
and multiplication table [11, b] = —[b,a] = d, [11,0] = —[c, a] = 2,
[b, t] = — [5, b] = f and the other brackets are 0. Let B be the bilinear form
onasuchthatmd)=p(d,a)=1,3(a,f)=a(fia)=1,a(b,e)=p(a,b)= —1
and the other values of p at ordered pairs of the basis of g in question are 0.
Show that B is invariant, g nilpotent, a z 99 aé (0} and fl non-degenerate.
23. Let g be a 3«dimensional non-commutative Lie algebra over a field
K of arbitrary characteristic.
(a) Ifg has centres yé (0}, then dim 3 = 1 (§ ],Exercise 2) and dim 9g = 1.
If 3 # 9g, 5 is the product of 3 and the 2-dimensional non-commutative
algebra. If a = 95, g is the 3-dimensional non-commutative nilpotent algebra
(§ 4, Exercise 9 (a)).
(b) If a = {0} but there exists in g a 2-dimensional commutative sub-
algebra, this subalgebra is unique and is equal to 99; for all x¢99, the
restriction u of ad x to 99 is a bijection of this vector space, determined to
within a multiplicative constant. Conversely, every automorphism u of a
2-dimensional vector space Ka + Kb determines a Lie algebra structure on
g = Ka + Kb + K: by the conditions [a, I7] = 0, [0, a] = u(a), [2, b] = u(b).
For two Lie algebras thus defined by automorphisms ul, un to be isomorphic,
it is necessary and sufficient that the matrices of 141 and 112 be similar to within
a scalar factor.
(0) Suppose that there exists in 9 no commutative subalgebra of dimension
>1. Show that there then exists a E 9 such that a $‘(ad (1)9 (assuming that
an element 2: does not have this property, show, using the jacobi identity,
that another element has the desired property). There then exists a basis
(a, b, c) of 9 such that [a, b] = c, [a, t] = {517, [b, c] = 7a, with (3 ye 0 and
Y aé 0, and g is simple. Let 45 be a canonical isomorphism of 9 onto the ex-
2
107
l LIE ALGEBRAB
of index >0; 9 then admits a basis (a, b, t) such that [4, b] = b, [a, c] = —c,
[b, o] = a.
(e) If K is of characteristic 2, show that there is no 2-mapping on g and
therefore that g is not the Lie algebra of a formal group. Show that g admits
derivations which are not inner.
1[ 24-. Let K be a field of arbitrary characteristic p. We again adopt Defini-
tion 1.
(a) Show that, unless p = n = 2, the only non-trivial ideals of 91(71, K)
n
are the 1-dimensional centre 3 with basis ‘21 E“ and the Lie algebra 91(71, K)
consisting of the matrices of trace 0. (Except in the exceptional case indicated,
note that if an ideal 0 contains one of the Eu, it necessarily contains 51(n, K).
If a contains an element not belonging to a, by multiplying this element by
at most four suitably chosen elements E", a non-zero multiple of one of the
E” is obtained.) If n is not a multiple ofp, show that gl(n, K) is the direct
sum of a and sl(n, K) and sl(n, K) is simple. If on the contrary I: is a multiple
of}: and n > 2, show that al(n, K) [3 is simple (same methods).
(b) Show that, for n a multiple ofp and n > 2, gl(n, K)/3 has radical {0},
but admits the quotient gl(n, K) /e[(n, K) which is Abelian.
fl 25. Let K be a field of arbitrary characteristic [1. Let ap(2n, K) denote
the Lie algebra of the formal symplectic group in 2» variables over K (§ 1,
Exercise 26 (0)). Show that this Lie algebra, which is identified with a sub-
algebra of gl(2n, K), has basis consisting of the elements
Ht = Eu — Ewan” (1 $ iS ’1):
Ft] = Eu — E1+n.l+n (1 < if g "1" #j):
Gil = Eun- + Em“: Gil; = Et+n.; + Ei+ml (1<i<j < n),
Emu and Eli-ml (1 < i< ’0’
(a) Show that, if 1; ye 2 and n 2 l, the algebra np(2n, K) is simple (for
n = 1, 619(2, K) = 91(2, K)), (Same method as in Exercise 24.)
(b) If]: = 2 and n 2 3, show that the H” E), G}, and GI, form an ideal
a of dimension n(2n — l); the elements of a of the form
2 NH: + m
1-1 Z “41E; + R]
Z $11G]! 'l' «1793;;
2
the multiples of 2:1 H. an ideal c, the centre of sp(2n, K); b, c are the only
108
EXERCISES
§7
The conventions of § 7 remain valid unless otherwise mentioned.
1. (a) For a Lie K-algebra to be nilpotent, it is necessary and suflicient
that it be isomorphic to a subalgebra of an algebra n(n, K).
(b) Suppose that K is algebraically closed. For a Lie K-algebra to be
solvable, it is necessary and sufficient that it be isomorphic to a subalgebra
of an algebra t(n, K).
2‘ Let g be a Lie algebra and 11 its largest nilpotent ideal. There exist a
finite-dimensional vector space V and an isomorphism of 9 onto a subalgebra
of 91(V), which maps every element of n to a. nilpotent endomorphism of V.
(Use Ado’s Theorem and Exercise 5 of § 1.)
3. Let g be a Lie algebra and U its enveloping algebra.
(a) Show that for all u EU (u aé 0) there exists a finite-dimensional repre-
sentation 1: of U such that 1r(u) aé 0. (Use Exercise 2 and Exercise 5 (b) of
(b) Deduce from (a) that if g is send-simple there exists an infinity of in-
109
I LIE ADGEBRAS
llO
CHAPTER II
In this chaptnd’ the letter K denotes a rum-zero commutative ring. 771: unit alemm! 0f K
i: deflated by l. Unlm otherwise mlioned, all algebra, algebra: and bigebras, all
modules and all tmrm products are over K.
From §6, K will be maimed to be afield qf characteristic 0.
Throughout this paragraph, 9 will denote a Lie algebra over K, U(g) or simply
U its enveloping algebra (Chapter I, § 2, no. 1), a the canonical mapping of 9
into U(g) (lac. oil.) and (U,),,,., the canonical filtration of U ([00. 012., no. 6).
1' The results of Chapters II and III depend on the first six books, on Lie group:
and Lie Algebras, Chapter I, on Commutative Algebia and on Dtfirmfiable and Anulytir
Manifolds, Summary of Results; no. 9 of § 6 of Chapter III depends also on Spmml
Theories, Chapter I.
111
11 FREE LIE ALGEBRAS
The K-module E is the direct sum of E‘L and the submodule K. u which is free
with basis 11; let nu: E —> E + and 1;“: E —> Kru denote me projectors associated
with this decomposition. Then
(I) m) = x — 6(x).u. mm = swu-
DEFINITION 1. An element x of E i: called u-primitiw if
(2) c(x)=x®u+u®x.
The u-pn'mitive elements of E form a submodule of E, denoted by Pu(E).
Pnoposmon 1. Every uwprimz'tiue dawn! of E belong: In E”.
(2) impliesx = :(x).u + 501).): = e(x).u + x, whence e(x) = 0.
Remark. Ife and c(x) = x’ 80 u + u ®x’, where x’, x' are in E+, then
x = e(x’) .u + :(u) .z” = x"; similarlyx = x’ and x is u-pfimitive.
For all :6 E“, we write
(3) £:(z)=c(x)—x®u—u®x.
PROPOSITION 2. W: have
(4) (1n, ® 1:”) co = c: o 1:".
Let x be in E; then
("u ° "0000) = ((1 - m) 8) (1 - 7)u))(€(x))
= 0(3) - (1 ®m)(‘(*)) — (m ® 1)(€(*)) + (m ® 71000))-
As z is counit of E,
(1 ® m)(‘(x)) = ’6 ® '1’ (m ®1)(F(*)) = u ® 3
whence
(m ® 71000)) = (m ® 1)((1 ® 140900)) = 50‘) -'4 ® 11;
fiom this we conclude
(in ® m)(v(x)) = v(~) — x ® u — u 8:: + a(x).u ® a.
On the other hand,
mum) = 0(z) — x®u — 1403:: + s(x).u®u,
whence formula (4).
As E+ is a direct factor submodule of E, E"® E+ can be identified with a
direct factor submodule of E 8) E. With this identification, 12,, ® 1:“ is a pro-
jector ofE ® E onto E*® E+. Byformula (4), 5,: maps E+ into E*® E’r and
1vu is a mam/tint: of the cogebm (E, 5) int!) the algebra (E+, a: ).
112
PRIMJTIVE ELEMENTS OF A BIGEBRA § 1.2
113
11 FREE uE menus
3. FILTERED BIGEBRAS
DEFINITION 2. Let E be a bigebra with coproduct e. A filtration compatible with the
higebm :hueture an E is an iiu‘reasing sequence (E,,),Do ofsubmadule: of E such that
E°=K.l, E=nL)J°E,,
(6) EWE" C Em” for m 2 0,71 2 0
0(En) C H," Im(Ei ® E,) for n 2 0.1'
A bigehra with a filtration eampatible with its bigebra structure is called a filtered
bigebra.
Example. Let E be a graded bigebm (Algebra, Chapter III, § 11, no. 4, Defini-
1;
tion 3) and (E"),,,° its graduation. We write E" = l; E‘. The sequence (13,.) is
a filtration compatible with the bigebra structure on E.
PROPOSITION 6. Let E be afiltered bigebm and (E,),>0 it;filtration. For every integer
n2 0,letE,f = EnnE". TIME; = {0}aml
n— l
114
ENVELOPING BIGEBRA or A Ln: ALGEBRA § 1.4
such that c(o(x)) = c(x) ® 1 + 1 ® 60:) for x e g. This proves the uniqueness
assertion of Proposition 7.
(b) We shew that t‘ it coastatialive. The linear mappings e’ and c” of U into
U ® U (8 U defined by
e' = (e®Id,_,) at and t” = (Idu®e) or
are unital algebra homomorphism which coincide on u(g) since, for a E 5(9),
c’(a)=a®1®l+1®a®l+l®l®a=e'(a),
whence the result.
(c) We show that e is toeammutatiae. Let 'r be the automorphism of U ® U
such that 1(a ® b) = h ® a for a, h in U. The mappings r o c and e ofU into
U (8 U are unital algebra homomorphisms which coincide on 0(9), whence the
result.
(d) We thaw that e 1': a taunt! for e. The mappings (IdU ® a) a e and
115
11 FREE LIE ALGEBRAS
(8) 5(a1...a,,)=]‘.:I;t(a‘)=u(a,®l+l®a¢)
where I(i) denotes the set of permutations of [1, n] which are increasing in
each of the intervals [1, i] and [i + l, a]. As U,I is the K-module generated by
the products ofat most n elements ofa(g), formula (8) implies that the filtration
(Un) is compatible with the bigebra structure of (U, o).
Damn-non 3. The bigebra (U, c) i: called the enveloping bigebra of the Lie
algebra g.
116
STRUCTURE on THE COGEBRA U(5) IN CHARAm'ERxsnc 0 § 1.5
of the vector space 5(9) onto the vector space U (Chapter I, § 2, no. 7). Recall
thatit...,xnareing, then
1
(9) 7)("1- - J») = Ergo. “(find . - 59100).
=26).()x)‘®c(x-‘=§(?)n(¢)®n(x”"),
whence
(n ® ’2)(‘s("")) = MW»-
As the x", forx e g and n e N, generate the vector space 5(9), (74 ® 71) 0 cs = c a '4
and 1) is a cogebra isomorphism.
On the other hand, formula (10) shows thatn($"(g)) = U", which completes
the proof of (a) and (b) taking account of the fact that the graduation of 5(g)
is compatible with its cogebra structure.
The graduation (U"),,;., of U is called the mamfeal graduation.
COROLLARY. The canonical mapping 6 define: an immatphirm afg onto the Lie algebra
P(U) qn'mitiI/e element: of U.
As 6" is a graded homomorphism of degree 0,
a = (EM? = M(Z?«ag®ai'") = 0-
9:
Remarks. (1) Un = 420 U‘ (Chapter I, §2, no. 7, Corollary 4 to Theorem 1).
(2) The mapping n is the unique morphism of graded cogebras of 5(5) into
U such that 11(1) = l and not) = a(x) for x55. For if 11’ is a. morphism
satisfying these conditions, we prove by induction on n that n'(x") = TKX") for
n-l
(13) em = H “Wm.
is! m(i)!
The e“, for [at] < 1:, form a basis of the vector K-space U,I (Chapter I, §2,
no. 7, Corollary 3 to Theorem 1). Then
118
STRUCTURE or FILTERED BIGEBRAS 1N CHARACTERISTIC 0 § 1.6
119
[1 FREE LIE ALGEBRAS
(18) ‘I(Xu) = “2
~1=u
X, ‘81 X, for a 6N“),
sincefl; is a cogebra morphism.
We show thatfm i: iry'ecfive. This results from the following lemma:
Lemma 2. Let V be a vector space, E a mgebm andf: S(V) —> E a algebra morphirm.
If the restrirtian Iff to S°(V) + S‘(V) is injective, then f is irgiettive.
Let n 2 0; we write S,I = ‘2. S‘(V) and es for the coproduct of S(V) and
show by induction on n thatf | 5,, is injective. Since the assertion is trivial for
n = Oandn = l,wcassumethatn 2 Zandletuesnbe such thatf(u) = 0.
Then
0 = €n(f(“)) = (f®f7(€s(u))
=f('4) 81 + l ®f(“) + (f®fl(5§("))
= (f®f)(03 (1‘))
As e; (u) E 5,._] ® 5,.” by (11) the induction hypothesis shows that u is a
primitive element of S(V), hence is of degree I (no. 5, Remark 4) and hence
is zero, sincef ] S‘(V) is injective.
It follows in particular that the family (X‘,) isfree.
We .rlww that fE i: .mrjeatz'oe if E has a filtration compatible with its bigebra
structure. Let (E030 be such a filtration and write E: = E,| fl Ker(:z).
We show by induction on n that E; is contained in the image offs. As
E = K.1 + “go E}, this will imply the surjectivity offE. The assertion is
trivial for n = O and follows from the Corollary to Proposition 6 of no. 3 for
n = 1 ; suppose henceforth that n 2 2 and let x e E; . By Proposition 6 of no. 3,
"-1
cm) e 2 E: o EL.
l20
mum“ or FILTERED BIGEBRAS m cmmmsuc 0 § 1.6
and by the induction hypothesis there exist scalars I“, 5, where a, 3 are in NW,
which are zero except for a finite number, such that
(’;®Ide)(€;("))= Z u....xa®x5®x~
u. D: ’1 ¢ 0
121
11 FREE LIE ALGEBRAS
122
CONSTRUCTION OF THE FREE LEE ALGEBRA §2.2
I23
1] FREE LIE AIJGEBRAS
124-
FUNL'I‘ORIAL PROPERTIES § 2.5
the Q, for the T, in P, we obtain a Lie polynomial R = P((Q,),EJ) e L(I).
Then
(5) R((ti)isl) = P((QI((ti)lel))leJ)
for every family 1‘ = 0.),“ of elements ofa Lie algebra g, as is seen by operat-
ing by the homomorphism f, on the equation R = P((Q,),EJ) and using
(4).
Let g be a Lie algebra, I a finite set and P e L(I). Suppose that g is a free
K-module. The mapping
13: 9I —> 5
defined by 13%).“) = P((t‘)‘fl) is then palynomiaLT For the set F of mappings
of g‘ into 9 is a Lie algebra with the bracket defined by
5. FUNCTORIAL PROPERTIES
PROPOSITION 2. Let X and Y be two rm. Every mapping a: X—>Y can be extended
uniquely to a Lie algebra hammarphixm L(u): L(X) —>L(Y). For every mapping
v: Y—> Z, L(u o u) = L(v) o L(u).
The existence and uniqueness of L(u) follow from Proposition 1 of no. 2.
The homomorphisms L(v o u) and L(v) o L(u) have the same restriction to X
and hence are equal (Proposition 1).
COROLLARY. If u i: irg'zctz'vz (rap. mgjzth've, bijettiue), w it L(u).
Since the assertion is trivial for X = a, we assume X yé 125 . If u is injective
there exists a mapping 1; of Y into X such that u o u is the identity mapping of
X; by Proposition 2, L(v) o L(u) is the identity of automorphism of L(X) and
hence L(u) is injective. When u is suijective, there exists a mapping w of Y into
X such that u a w is the identity mapping ofY ; then L(u) o L(w) is the identity
mapping of L(Y), which proves that L(u) is suijective.
1’ Recall (Algsbya, Chapter IV, § 5, no. 10) the definition of polynomial mappings
of a free module M into a module N: if q is an integer 20, a mappingf: M —> N is
called harmgarwau: polynamial qf dag”: q if there exists a. multilinear mapping :4 of
M' into N such that
for) = u(x, . . ., x) for all xe M.
A mapping of M into N is called pnlgpmmial if it is a finite sum of homogeneous
polynomial mappings of suitable degrees.
125
1! FREE LIB AwEBRAS
Let X be a set and S a subset of X. The above corollary shows that the
canonical injection of S into X can be extended to an isomorphism on of L(S)
onto the Lie subalgebm L’(S) ofL(X) generated by S; we shall idmtyy L(S) and
L’(S) by means of on.
Let (Su)mel be a right directedfamily of subsets ofX with union S. The relation
S“ C SB implies L(Su) C L(SB) and hence the family of Lie subalgebras L(Su)
of L(X) is right directed. Therefore 9 = “Lg! L(Su) is a Lie subalgebra of
L(X); then S C 9, whence L(S) C 9, and, as L(S,) C L(S) for all eel,
g C L(S). Hence
PROPOSITION 4. Let S be a Mb!!! q/X. IfN‘s) is identified with it: canonical image in
NO“ (Algebra, Chapter I, § 7, no. 7), then L(S) = “Hm L“(X). Further, for all
u 2 NS), the homogeneous component of degree a under the multigmdualian an L(S) is
equal ta L“(X).
Let a e N‘fi’. The module L“(S) is generated by the images in L(X) of the
elements 10 in M(S) such that (Mm) = a, that is (Algebra, § 7, no. 9, formulae
(23) and (24)) the set ofw in M(X) such that ¢(w) = 0:. Hence L°‘(S) = L“(X).
The proposition follows from this and the relation L(S) = L‘(S).
ash“)
(16) L(fl
(u
fl L(S‘).
s.) = (a!
This follows from Proposition 4 and the obvious formula
(17) N“) = I0
Gl
we
where we have written S = 'Q S‘.
By formula (18), P" D ”2; [P,, P”_,] 3 [P1, Pnd] = P," whence (19).
When we take P = g, the sequence (Pu) is the lawer central series (Wm) of 9
(Chapter I, § 1, no, 5). Hence:
128
DERNA'HONS or me LIE mum §2.8
P12209110“ 6. Iztgbealieulgebmmd(?'g)_ntlulamermbalmi¢xq.
I. DERIVATIONS OF FREE IE mm
PROPosn'KONS. LetXbeam,MMbtmL(X)-mMe¢miI:ldb¢amppfigqf
XintoM. Warhkmaudmlymlinmmapptq(X)intoMmaflfig
dandmtigiringthenlafion:
(20) D([a, a']) = a.D(a') — a'.D(a) fa, a, u' in L(X).
We define a lie algebra g with underlying module M x L(X) by means of
the bracket
(21) [(M. a), (MC I1')] = (“'1' - B'Jn, [1.111).
ll FREE LIE ALGEBRAS
for a, a' in L(X) and m, m' in M (Chapter I, § 1, no. 8), Letfbe the homomor-
phism of L(X) into 9 such that f(x) = (d (x), x) for all x in X; let
f(a) = (D(a), u(u)) for all a in L(X). By formula (21), u is a homomorphism of
L(X) into itself; as u(x) = 1 for x in X, u(a) = a for all a in L(X), whence
9. ELIlVflNA’l'lON THEOREM
PROPOSITION 9. Let S1 and S, be [we disjoint sets and d a mapping of S1 x S2 into
L(S,). Let g be the quotient Lie algebra 9f L(S1 U Si) by the ideal generated by the
elements [51, :5] — d (3,, 5,) with :1 6 SJ, :2 E 8,; let 4/ be the canonical mapping qf
L(51 U 5,) onto 9.
(a) For i = l, 2, the restriction (In. 13/ q; to S, can be extended to an isomorphism
of L(Si) onto a subalgebra a. of g.
(b) g = a1 + a2,a1 ha, = {0}andu2isanidealafg.
For i = 1, 2, let 4;, denote the homomorphism of L(Si) into 5 which extends
(b, and 45 its image. Clearly ems.) generates «I.
Let :1 e 8,; we write D = ad (1)1(51). The derivation D ofg maps (1)2(82) into
an by the relation
Hid-‘1): ti>2(-'2)] = ‘i’nWUn 52)) {01” 52 5 52;
as the subalgebra a, of g is generated by 452(52), therefore D012) C n,. The set
of x e 9 such that ad 2: leaves «12 invariant is a Lie subalgebra. of g which con-
tains 471(51) by the above and hence also a, Hence
(23) [an “2] c “2‘
Therefore a1 + a3 is a Lie subalgebra of g and, as it contains the generating
53': ¢1(Si) U $262);
(24) a, + a2 = 9.
For all :1 E S there exists a derivation D,1 of L(Sg) such that
for all :1 in S2 (no. 8, Corollary to Proposition 8) The mapping :1 1—» D.l can
be extended to a homomorphism D ofL(Sl) into the Lie algebra of der1vations
ofL(S,) Let b be the semi-direct product ofL(S ) by L(Sa) corresponding to D
(tter I, § 1, no. 8). As a module 9 is equal to L(Sl) x L(Sz) and in par-
men
131
I! FREE LIE ALGEBRAS
132
HALL sen [N A FREE MAGMA § 2.10
For every subset S of X we identify the free magma M(S) with its canonical
image in M(X).
PROPOSITION 13. Let X be a finite set with at Iezut two elements. Let H deflate a
Hall ret relative to X. Then there exist a strictly Wearing bijectim p» to, if N
unto H and a sequence (P,),eN ofruhretr af H with thefallawing propertiet:
(a) P0 = X.
(b) For every integer p 2 0, w, e P,.
(c) For every integer n 2 I, there exists an integer [:(n) ruth that every element qf
P, 1': qflength > nfor all!) 2p(n).
(d) Far every integer]: 2 0, the re! P,“ with“: qfthe elements qftheform wpw,
wherei2 0, weP, andwgéwp.
As X is finite, each of the sets M"(X) is finite. Let H" = H n M"(X) for
all It 2 l. The Corollary to Proposition 12 shows that the finite set H,I is
non-empty. Let u" be the cardinal of H"; let v‘, = O and 0,, = 141 + - - - + u"
for n 2 1. As H" is a totally ordered finite set, there exists a strictly increasing
bijection p »—> w, of the interval (”n—1) v" — l] of N onto H... It is immediate
that p ~—> w, is a strictly increasing bijection of N onto H.
Let I“D = X and for every integer p 2 1 let P, be the set of elements w of
H such that w 2 w,, and, either w e X, or a:(w) < 10,, (note that if w is of
133
ll FREE LIE AIIGEBRAS
y = (“1.w)l)0.wEP,,w¢wp
as basic family. By Proposition 13 (d), the family (12)“1,” H is equal to 9" and
hence is a basic family of I), = g“). Hence L(X) = L, @T, Q gfil and,
35 Lp+l = L1: + T1» L(X) = Lp+1 @Spn and(u7fi,i1n..., Li’s—1, £71,) is a
basis of the module L1,”.
Let n be a positive integer. By Proposition 13 (0) there exists an integer
pot) such that Pp has only elements of length >n for p 2 [7(a) For [7 2 [7(n),
the Lie subalgebra g, of L(X) is generated by elements ofdegree > n and hence
L"(X) n g, = (0}. On the other hand, the elements a”), of L(X) are homo-
geneous and the family (10,)“,(1, is a basis of a supplementary module of 9,.
It follows immediately that the family of elements Lia of degree n is a basis
of the module L"(X) and that the sequence (12000 is a basis of the module
L(X).
(B) General case.
If S is a subset of X, recall that M(S) is identified with the submagna of
M (X) generated by S and L(S) is identified with the Lie subalgebra. of L(X)
generated by S; we have seen that ifw e M(S) is oflength 2 2 then ot(w) E M(S)
and {5(w) e M(S). It follows immediately that H n M(S) is a Hall set relative
to S.
For every finite subset (I) of H there exists a finite subset S of X such that
135
ll FREE LIE ALGEBRAS
(l) C M(S). Case (A) then shows that the elements :71 with w e 1D are linearly
independent in L(S) and hence in L(X). Therefore the family ("7)“): is free.
For every element a of L(X) there exists a finite subset S of X such that
a E L(S). By case (A), the subset ‘I’(H n M(S)) of‘l’(H) generates the module
L(S) and hence a is a linear combination of elements of ‘I’(H). Hence ‘I’(H)
generates the module L(X), which completes the proof.
COROLLARY. The module L(X) is free and :0 is each of the submadule; L“(X)
where a e No” and L"(X) where n E N. The modules L“(X) are offinite rank and :0
are the module: L"(X) if X isfinite.
There exists a Hall set H relative to X (Proposition 11). For all to E H, the
element ‘l’(w) of L(X) belongs to one of the modules L“(X) (with a 51“”)
and the module L(X) is the direct sum of the submodules L“(X). Further,
for all at ENG”, the set of elements of M(X) whose canonical image in N“)
is equal to a is finite; this shows that each of the modules L“(X) is free and
of finite rank and that L(X) is free. Now L“(X) = M“ L“(X) and hence
L“(X) is free; when X is finite, the set of ateN‘x’ such that Ia] = n is finite
and hence L"(X) is then of finite rank.
DEFINITION 3. A Hall basis qf a free Lie algebra L(X) it any basir of L(X) which
is the canonical image of a Hall Jet relative to X.
Remark. Suppose that X consists of two distinct elements 2: and y and let
L‘ '1’ be the submoclule of L(X) the sum of the L“(X) where at ENX and
a(y) = 1. It follows immediately from Theorem I and Proposition 12 of
no. 10 that the elements of (ad x)".y where n is an integer 20 form a basis
of the submodule L‘ "’. It follows that the restriction to L‘ "l of the mapping
ad 1 ix infective.
In this paragraph, A(X) = AK(X) denote: the fiee aubciative algebra Libas(X)
of the set X over the ring K (Algebra, Chapter III, § 2, no. 7, Definition 2). X is
identified with its canonical image in A(X) ; recall that the K—madule A(X) admit:
a: bad: the free manoid M0(X) derived from X; A*(X) deflate: the .mbrnadule of
A(X) generated by the nan-empty wordy.
136
ENVELOPING ALGEBRA or L(X) § 3.1
a: L(X) —> U(L(X)) he the canonical mapping of L(X) into it melaping algebra
and let p: U(L(X)) —>A(X) be the unique unital algebra hamomnrphixm mh that
[i 0 cr = at (Chapter I, §2, no. 1, Proposition 1). Then:
(a) at i: injective and a(L(X)) i: a direttfactar .mhmmtuk 11fA(X).
(b) {3 is biieetive.
Let B be a uniml K—algebra and (b a mapping of X into B; by Proposition
1 of § 2, no. 2, there exists a Lie algebra homomorphism Q}: L(X) —> B such
that L1] | X = (b; by Proposition 1 of Chapter 1, § 2, no. I, there exists a unital
algebra homomorphism 6: U(L(X)) ——>B such that 0 o a = q; and hence
such that (0 o a) [X = 4). As 6(X) generates the unital algebra U(L(X)),
the homomorphism 0 is the unique unital algebra homomorphism satisfying
the latter condition. This shows that the ordered pair (U(L(X)), o | X) is
a solution of the same universal mapping problem as A(X) ; taking (b to be
the canonical injection of X into A(X), we deduce that (5 is an isomorphism,
which proves (b).
Finally, as L(X) is a free K—module (§ 2, no. 11, Corollary to Theorem I),
u- is injective and U(L(X)) is a direct factor submodule of U(L(X)) (Chapter I,
§ 2, no. 7, Corollary 3 to Theorem 1). By (b), this proves (a).
COROLLARY 1. There exists on the algebra A(X) a unique taprodutt making A(X)
into a higebm mph that the elements {If X are primitive. Further, fl is an isomorphirm
of the higebra U(L(X)) unto A(X) with thir bigebra rtmtture.
This follows from assertion (b) of the theorem and the fact that X generates
the unital algebra A(X).
Henceforth A(X) is given this bigebra structure and L(X) is identified with
its image under at, that is with the Lie mbalgebra of A(X) generated by X.
COROLLARY 2. If K is afield ofcharatleristie 0, L(X) is the Lie algebra afprimitiue
element: q(X).
This follows fiom Corollary 1 and the Corollary to Proposition 9 of § 1l
no. 5.
Remarks. (1) Let K' be a commutative ring containing K. [I A(X), L(X)
and Lxr(X) are identified with subsets of AK»(X), we deduce from part (a)
of Theorem 1 the relation
(1) L(X) = LK.(X) n A(X).
(2) Corollary 2 to Theorem 1 remains valid if it only assumes that the
additive group of the ring K is torsion-free. For suppose first that K = Z;
every primitive element of A(X) is a primitive element of AQ(X) and hence
is in LQ(X) fl A(X) = L(X) (Corollary 2 and formula (1)). In the general
case, K is flat over Z and we apply Remark 2 of § 1, no. 2 and Proposition 3 of
§ 2, no. 5.
137
II FREE LIE ALGEBRAS
LI)—f‘—>A
: 1:
A(I)
where i is the canonical injection,f, is the Lie algebra homomorphism defined
by t and g, is the unital algebra homomorphism such that g,(i) = t, for i e I.
The diagram is commutative for g, a i andf, coincide on I. It follows that if
P EL(I), the element P((t,),“) defined in §2, no. 4, coincides with the ele-
ment Palms!) defined in Algtbra, Chapter III, §2, no. 8, Example 2.
138
PROJECTOR or A" (X) ONTO L(X) §3.2
139
11 FREE LIE ALGEBRAS
l_L
=m>a l—B-lg
—L!u"T“.
14-0
DIMENSION or me HouooEN'Eous couponem or L(X) §3.3
in the form ml, where m is an integer >1 and AeA. Condition (12) is
is equivalent to
"'_7‘ u
(14) Imam) = “2... W) (éyfi
d .
forallkeAandallm) l.
<18) Pm = (1 — 219‘“
I41
I! FREE LIE ALGEBRAS
(2) For all a ENx — (0), let (0,7,)lsygcm be a basis of L“(X) and give the
set I of ordered pairs (agj) such that or E Nx — (0) and l <j $ 5(1):) 3 total
ordering. By Theorem 1 of no. 1 and the Poincaré—Birkhoff—Witt Theorem
(Chapter I, § 2, no. 7, Corollary 3 to Theorem 1), the elements
Formula (1) implies Ga C G‘3 for (3 < a: and hence the family (Cu) is de-
Hearing. The filtration (GE) is called :efiamted if 0 G.I reduces to the identity
element and is called exhawtive ifG = g G“.
Remark. Let (G,l),,Ez be a decreasing sequence of subgroups of G. It is a
decreasing filtration in the sense ofCummumiue Algebra, Chapter III, § 2, no. 1,
Definition 1. For every integer n and all or in the interval )n — l, n] of R, we
write H“ = G”, in particular H,I = G”. It is immediate that we thus obtain
a real filtration (Hagan on G; such a filtration will be called an integral
filtration. Hence decreasing filtrations in the sense of Commutative Algebra,
Chapter III, § 2, can be identified with integral filtrations.
Let A be an algebra; a real filtration (Au) on the additive group of A is
called compatible with the algebra structure KAWAB C A“, for a, B in R and
K.Afl c A“ for an ER. If the filtration is exhaustive, (A0,) is a fundamental
system of neighbourhoods of 0 under the topology on A which is compatible
with the algebra structure. Let B be a unital algebra; a real filtration (Ba) on
the additive group of B is called compatible with the uniral algebra structure if
it is compatible with the algebra structure and 1 6 Ba.
2. ORDER FUNCTION
Let G be a group with identity element e. Let (Ga) be a. real filtration on G.
For all z in G let I, denote the set ofreal numbersa such that x E G,. Ifot e I,
and B < a, then (-1 E I, and hence 1,, is an interval (General Topalagy, Chapter
IV, §2, no. 4-, Proposition 1). Using relation (1), we see that I,, contains is
least upper bound when this is finite. Therefore Ix is ofthe form ] — co, v(x)] ran
with 0(x) e R; we have 0(a) = sup{u | x e G.)
The mapping v of G into R is called the mderfimlicm associated with the real
filtration (Ga) and u(x) is called the order of x. This mapping has the following
properties:
(a) Far I: e G and u E R, the relation: x e G and v(x)2 a: are equivalent
(b) For x, y in G,
(2) 00:”) = 11(x), 11(2) = +00.
(3) "(m > i“("0"): ”(14))-
Further, we have equality m (3) if 11(2) > u(y).
G(c) For all 016R, let G; denote the set Iy‘xeG mh that 0(x) > on. Then
=09“ GB and in particular G: i: a subgroup q.
GConversely, let u be a mapping of G into R satisfying relations (2) and (3).
For all an ER, let Gu be the set ofx e G such that u(x) 2 on. Then (Gm)asR is a
real filtration of G and v is the order function associated with this filtration.
143
11 FREE LIE ALGEBRAS
For the filtration (Ga) to be integral, it is necessary and sufficient that 1; map G
into Z U (+00, —oo}. For it to be exhaustive (resp. separated), it is necessary
and sufficient that 3/1(—oo) = :21 (resp. _v1(+oo) = (2)).
Let A be a K-algebra (resp. unital K-algebra). By the above, the relation
“xeAu ¢u(;r) 2 (l forxeA and ask”
definm a bijection of the set of exhaustive real filtrations (Aafin compatible
with the algebra (resp. unital algebra) structure on A onto the set of mappings .
v: A —> 1—! not taking the value —00 and satisfying axioms (4) to (7) (resp. (4-)
to (8)) below:
(4) ”(x + y) > him“): 001)) (fly in A)
(m m—x)=vu) (xeA)
(6) 00.1) 2 11(1) (A E K, x EA)
(7) ”(‘11) > ”(Z) + ”(y) (m in A)
(8) 0(1) 2 0.
Remark. If ”(x) is not everywhere equal to +00, conditions (7) and (8) imply
0(1) = 0.
m Q=H%
clearly G: is a subgroup of G“. We write gr,(G) = Cut/G;r and
g®=®mm men
The gradsd group associated with the filtered gum]: G is the group g'r(G) with its
natural graduation of type R.
Ramark. When the filtration (G8) is integral, gr¢(G) = {0) for non-integral at
and grn(G) = Gn/G".l for every integer n. The definition of the associated
graded group therefore coincides essentially with that of Commutative Algebra,
Chapter III, §2, no. 3.
Let A be an algebra (resp. unital algebra) and (Anna a real filtration com-
patible with the algebra (resp. unital algebra) structure (no. I). Then
Au'AB C A1141: Act-An + AuA; C ALB;
and the bilinear mapping of A“ x Afl into A“, the restriction of multiplica-
tion on A defines on taking quotients a bilinear mapping
gram x EMA) —>grm(A)-
144
comm. rum-mom ON A GROUP §4.4
146
AN EXAMPLE OF A (ENTRAL FILTRATION §4.5
147
II FREELIE ALGEBRAS
(ii) Let g be tlu group homomorphism of gr(G) into gr(A) what: restriction to
grm(G) is gufm' all a. The mapping g i: an irg'utiue hammphism qie Z-algtbmr.
(i) Let at > 0. By hypothesis, for all a in G“, 9(a) — l 6 Au; let A,I (a) denote
the class of 9(a) — 1 modulo A; 4 As A,“ C A: , relation (25) implies
[11(ab) = p¢(a) + paw). Then 11 EC; if and only if 9(a) — l 6A3; therefore
G; is the kernel of the homomorphism pm of G“ into gru(A). On passing to the
quotient, p, then defines an injective homomorphism gal of gra(G) into gru(A).
For at < 0, grm(G) : (0) and file only choice is gm : 0.
(ii) As g“ is injective for all real :1, g is injective. We show that g is a Lie
algebra homomorphism. As gra(G) = {0} for I: < 0, it suffices to establish the
formula
(27) mafia, 17)) = L0.(a),h(b)l
for a. > 0, [3 > 0, —> e G“ and b 6G5, which follows from (26).
6. INTEGRAL m FILTRATIONS
Recall (no. 1, Remark) that a filtration (G9!) on the group G is called integral
if Gm = G,‘ for every integer n and all on e]n — l, a]. To be given an integral
central filtration on a group G is equivalent to being given a sequence (0%)”)1
of subgroups of G satisfying the conditions
(i) G1 = G
(ii) G,,DGM1 foralIn>l
(iii) (Gm,G,,) C Gm" form 2 l andn 2 1.
For every integer n 2 l, G, is a normal subgroup of G and the quotient
grn(G) = Gn/Gnu is commutative. On taking quotients, the mapping
(x,y) »—>(x,y) = x'ly'lxy of G", x G,. into Gm” allows us to define on
gr(G) = Q grn(G) a graded Lie algebra structure of type N over the ring Z.
Recall (Algebra, Chapter I, §6, no. 3, Definition 5) that the lower antral
mic: of the group G is defined by
(28) (PC = G, C"+1 = (G, C”G) for a 2 1.
The corresponding filtration is called the lawn amtmlfiltratian of G.
PROPOSITION 4-. (i) The lawn central series of G is an integral central filtration on
G
(ii) If (Gang. is an integral central filtration on G, then C"G C CH for all
n e N*.
Assertion (i) has been proved in Algebra, Chapter I, §6, no. 3, formula
(7)
We prove (ii) by induction on n; C‘G = G = Cl; for n > 1,
CG = (G, C"“G) C (G,G,.-1) C G".
148
MAGNUS ALGEBRAS § 5.1
PROPOSITION 5. Let G be a group and gr(G) the graded Lie Z—algebm axsoeialed with
the lower centralfiltration on G. Then gr(G) it generated by grl(G) = G/(G, G).
Let L be the Lie subalgcbra of gr(G) generated by gr1(G) ; we show that
L D gr"(G) by induction on n, the assertion being trivial for n = 1. Suppose
that n > 1 and L 3 gr,_l(G). As C"G = (G, C““G), the construction ofthe
Lie algebra law on gr(G) shows immediately that
gr..(G) = mt(G),grn~1(G)] C L-
The above proof shows that the lower central series of the Lie algebra
gr(G) (§ 2, no. 7) is given by
l. MAGNUS ALGEBRAS
Let A(X) be the product module 1:; A”(X). We define on A(X) a multipli-
cation by the rule
(I) («.12). = Z m-
where a = (11,.) and b = (b,) are in A(X). We know (Commutative Algebra,
Chapter III, §2, no. 12, Example 1) that A(X) is an associative algebra and
that A(X) is identified with the subalgebra ofA(X) consisting of the sequence
all of whose terms are zero except for a finite number.
A(X) is given the product topology of the discrete topologies on the factors
149
11 FREE LIE ALGEBRAS
A"(X) ; this topology makes A(X) into a complete Hausdorfi‘ topological alge-
bra, when the ring K has the discrete topology, and A(X) is dense in A(X)
Let a = (a,,) e A(X) ; the family (a,,),,)n is summable and a = “2:0 an.
For every integer m 2 0, let AM(X) denote the ideal consisting of the series
a = ”m afl such that a,I eA”(X) for all n 2 m. This sequence of ideals is a
fundamental system of neighbourhoods ofO in A(X) and an integral filtration
on A(X). The order function associated with the above filtration is denoted by .
a); then 0(0) = +00 and m(a) = m il‘a = “Z“ an with a,eA"(X) for all
n 2 mandam 9e 0 (§4, nos. 1 and2).
A(X) is called the Magnu: algebra of the set X with coefficients in K. If there
is any ambiguity over K we write AK(X).
PROPOSITION I. Let B be a unilal associative algebra with a realfiltration (3,)“3 such
that B is Hausdozfl and complete (§ 4, nos. 1 and 2). Letf be a mapping uf X into B
such that there exist: A > Of»! whichf(X) C BX. Then f can be extended in one and
anly one way to a continuous unital homomorphismfofA(X) into B.
Letf’ be the unique unital algebra homomorphism ofA(X) into B extending
f (Algebra, Chapter 111, § 2, no. 7, Proposition 7). We show that f’ is continuous:
f’(A"(X)) C BM whence f'(A,,(X) n A(X)) C B”. Therefore f' can be
extended in one and only one way by continuity to a homomorphism
f: A(X) —> B.
We preserve the hypotheses and notation of Proposition 1 and let u E A(X).
The elementf(u) is denoted by u((f(x)) fix) and called the result ofsubstituting
thef(:r)far the z in a. In particular, u((x),,ex) = u. Now let u = (11,)e be a
family of elements ofA1(X) and let a E A(Y). The above allows us to define the
element a((u,,),,ey) e A(X) It is denoted by u o u. As u,((f(x))) E BA, the
elements uy((f(x))) can be substituted for the y in v. The mappings
v~—> (v ou)((f(x))) and a>—>v((uy((f(x))))) are then two continuous homo-
morphisms of unital algebras of A(X) into B taking the same value u,,((f(x)))
at the element 3/ 6 Y. Therefore (Proposition 1)
(2) (0 ° u)((f("))) = ”((MUO‘D)»
for all u E A(Y).
2. MAGNUS GROUP
For all a = (110,.” in A(X), the element ao of K will be called the constant Ienn
of a and denoted by 5(a). Formula (I) shows that z is an algebra homomorphism
ofA(X) into K.
Lemma 1. For an element a 13/A(X) (0 be invertible, it is necessary and mficienl that its
emtant term be invertible in K.
150
MAGNUS GROUP AND FREE GROUP §5.3
151
n ma LIE menus
bI 2 0 and (b1, . . ., h.) 9e (a1, . . ., a.) can be equal tox‘l“. . .4“. It follows that
152
rowan CENTRAL sums or A FREE GROUP §5.4
153
11 FREE LIE ADGEBRAS
COROLLARY. “n
)1 C"F(X) = {5}-
Remark. Let H be a Hall set relative to X (§ 2, no. 10). Let M be the magma
defined by the law of composition (x, y) »—> (x, y) = x'ly‘lxg/ on F(X) and let
(I) be the homomorphism of M(X) into M whose restriction to X is the
identity. The elements of ¢(H) are called the basic commutator: of F(X) asso-
ciated with the Hall set H. For every integer n 2 1, let H,| be the subset of H
consisting of the elements of length n; we know (§ 2, no. 11, Theorem 1) that
the canonical mapping «3a into LZ(X) is a basis of the Abelian group L; (X).
Moreover, 42(Hn) C C“; for all m E H,., let 4),.(m) denote the class mod. 0‘”
of (Mm) 6 C". Theorem 3 then shows that (bu i: a bijeetion oa onto a basis of the
Abelian grim]: C"/C"”. It follows immediately that, for all w E F(X) and all
i 2 1, there exists a unique element on; of 1(a) such that, for n 2 l,
n
(4) w = (1:! "1:; ¢(m)"4("" mod. CV“,
154
EXPON'ENTIAL AND LOGARm-IM IN FILTERED MEBRAS § 6.1
Then exp(:r) e l + m and the mapping exp: m —> 1 + m is called the apartm-
tial mapping ofA.
For all ye] + m, the family ((—l)"“(y — 1)"/")n>1 is summable. We
write
(2) logy = Z(—1>"-*(y — 1w».
155
11 FREE LIE menus
Then log yem and the mapping logzl + m—>m is called the logarithmic
mapping of A.
PROPOSITION l. The exponential mapping ir a homenmarphirm of III min 1 + m and
the logarithmic mapping is the invent: homeomorphism.
For x E A“, :15: EA”. It follows that the series defining the exponential con-
verges uniformly on each of the sets A“ for a > 0; as A“ is open in m and
m = “U0 A“, the exponential mapping is continuous. It can be similarly shown '
that the logarithmic mapping is continuous.
Let 2 and I be the formal power series with no constant term
XE
:(X) = an H' l(X)= n>1
Z(— I)“ ‘X"/n
We know (Algebra, Chapter IV, §G, no 9) that e(l(X)) = l (e(X)) = X on
A({X})= K[[X]]. By substitution (§ 5, no. 1), we deduce that
0001)) = 1(444)) = x
for x e m; as
exp): = :(x) +1, log(l + x) = l(x)
it follows immediately that
logexpx=x, explog(l+x)=l+x
for x in m, whence the proposition.
Remarks. (1) If x e m, y e III and x and y commute, then
aw + y) = expor) exm),
1
since the family (#311! ) is summable (cf. Algebra, Chapter IV, §6, no. 9,
LIEN
Proposition 11)
(2) As the series a and l are without constant term and A“ is a closed ideal
ofA, ep C l + A andlog(l + A“) r: Auwhenceepu = l + A‘and
log(1+ A“): Aviora > 0.
(3) Let B be a complete Hausdorfl' filtered unital associative algebra and
II C U" B. Letf be a continuous unital homomorphism of A into B such that
f(m) C0n. Then f(exp x) = expf(x) for xem and f(log:y)= logf(y) for
y E l + I11; we show for example the first of these formulae
f(mp1)=,,;,f(x”/n'n;f(x)"ln!=expf(x)-
156
HAUSDORFF GROUP § 6.2
1. HAUSDORFF GROUP
Let X be a set. We use the notation of § 5, nos. 1 and 2. The free Lie algebra
L(X) is identified with its canonical image in A(X) (§ 3, no. 1, Theorem I). We
denote by L(X) the closure of L(X) in A(X), that is the set of elements
of A(X) of the form a = 2:1 a,I such that aneL"(X) for all n 2 0; this is
filtered Lie subalgebra “A(X).
THEOREM 1. The mtriction qf Ihe exponential mapping y” A(X) to L(X) is a
bifmian 1y” L(X) arm: a closed subgroup affine Magma" graup I‘(X).
We write A(X) = A, A"(X) = A", A(X) = A, L"(X) = L", L(X) = L,
I‘(X) = I‘. Let B be the algebra A ® A with the graduation of type N defined
by B" = ”Z“ A‘ ® A’. Let B = I); B" be the associated complete filtered
algebra (Commutaliw Algtbra, Chapter III, §2, no. 12, Example 1). The co-
product 0: A—>A ®A defined in §3, no. 1, Corollary 1 to Theorem 1 is
graded of degree 0 and hence extends by continuity to a homomorphism
I: A —> 3 given by
it: an) = Z c(a,.) for a" 6A“.
)0 you
Z (an® 1),
812;) 1,.) = man 8120.1"): 2 (l ®a,,) for unsA".
u) nan
By Corollary 2 to Theorem 1 of § 3, no. 1, L” is the set of a,‘ e A" such that
:01") = 1,, ® I + 1 ® an. It follows that L is the set ofa EA such that
(a) 5(a) = 3'01) + 8'(a).
Let A be the set of 1: EA of constant term equal to l and satisfying the
relation
(4) 5(5) = 8’(b)~8'(b),
157
ll FREE LIE ALGEBRAS
in other words, the set of]; = 20 bn such that b" E A" for all n 2 0, £0 = I and
4b,) = ”/2" b, ® b, for n 2 0. The latter characterization shows that A is a
closed subset of I‘; as i, 8’ and 8" are ring homomorphism and every element of
SKA) commutes with every element of 8"(A), the restrictions to 1" of the
mappings o and 8’8” are group homomorphisms and A is a subgroup of I‘.
By Proposition 1 of no. 1, the exponential mapping ofA is a bijection of the
set A* of elements of A with no constant term onto I‘. Let aEA* and
b = exp a. As 5 is a continuous ring homomorphism,
3(1)) = 5(2):" o"/n!) = ”Z, £(a)"/n! = exp 5(a).
The relations .
8'(l’) = exp 3’01), 3'0) = exp 3'01)
are proved similarly and, as 8’(a) commutes with 8"(a) (no. 1, Remark 1),
8’(b) 8"(17) = exp (8'01) + 8'(a)).
Therefore 4 satisfies (3) ifand only if b satisfia (4) , which proves the theorem.
Remark. The above proof shows that expd.) is the subgroup A of I‘ consisting
of the b satisfying (4).
Hence the group law of A can be transported by the exponential mapping to
L. In other words, L is a complete topological group with the law of com-
position (a, b) >—> a H 1) given by
an I; = log(exp a.exp b).
The topological group thus obtained is called the Hausdorflgroup (derived from
X relative to K).
Let g be the homomorphism' of the free group F = F(X) into I‘ such that
g(x) = expx for xeX. As expx — l — x = unfl/n! is of order 22, g is
injective by Theorem 1 of § 5, no. 3. Therefore the mapping log a g is an injectiue
homomfirphfirn of F into the Hausdorfl" group whieh extend: the canonical injection
X —> .
For every integer 7n 2 l we denote by L," the set of elements of order 2 m in
L and by I‘m the set of u e I‘ such that u — 1 is of order 2m. Then
- l
L," = exp(I‘,,.) by Remark 2 of no. 1; as (1",!)m“ is an integral central
filtration on I‘ (§ 4, no. 5 Proposition 2), (I'm)m>1 it an integral central filtration
on the group L.
3. LIE FORMAL POWER serum
Lemma 1. Let g be a filtered Lie algebra (§ 4, no. 1), (5a)“: it: filtration and let
a: E R. Let P be a homogeneoux Lie polynomial of degree n in the indeterminate: (TI).E‘
(§ 2, no. 4). Then P((a,)) E g,“for everyfamily (Q‘s, Ly'elements of g“.
158
LIE FORMAL rowan SERIES §6.3
(5) :40.» = 2,
EN mo.»
Let a be a continuous homomorphism of 9 into a complete Hausdorff
filtered Lie algebra 9’ such that 9' = “go 9;. For everyfinite family t = 00‘“
ofelements ofg and all u E La),
(6) «(an)» = «(an)»,
for the homomorphism e of, is continuous and maps T, to 0-0.) for i e I.
Let u = (u,),u be afinile family of elements of L(I) and let ve LG); by
substituting the u, for the T, in v, we obtain an element w = I/((u,),e,) of L(I)
denoted by u o 14. Then
(7) WWW“) = ”(("I((1A)Iel))1ea)
for everyfinite family t = (1‘).“ of elements of g, as is seen by operating with
the continuous homomorphismfi on the equation w = v((u,),e,).
Let u = ‘12“ qLU), where qL"(l). The mapping :7 (li) >—> u((ti)) of V
g‘ into 9 is continuous: for in each of the open sets 9‘, with a. > 0 the family of
fin is uniformly summable and it suffices to prove that each 17,, is continuous,
which is immediate by induction on M.
T A Lie formal power series is not in general a formal power series in the sense
of Algebra, Chapter IV, §6.
159
XI FREE LIE ALGEBRAS
>1, Ihm
THEOREM? [f 1 and J an two positive integer: .mch that r +s>
H“ = H.’ + H1,” when
(9) (7 + M15.» =
z (—1)“ dU v,
m
l+u>1....,r:.-_‘x:'.:_x>1
((E —‘ n) —‘ .8 Vanni)“
”—1 adU '4 adV n
<10) <r+s)H¢..=
2‘;— I)" ’
m>1
Z
u+~~+r.—1=r-1
(“1:1<adrf)~(adsx>~)w).
Y‘+I];llt:::-Vll-l+’m— [)1
r .
In A.((U,V}), ep.cxpV =1+ w, where w = +2.11%; whence
H = ”‘21 (—1)’"“W"‘/m (no. 2), that is:
(11> H,,.=Z<_-1>““
own In
Z
In+~~ n."
fimfl
In"! A“!
m n. =3
u+s>1."r...+;..>1
The linear mapping Pu, defined by Pn(zl,.. ., x") = -:(fi (ad x0) (x,,) for
2 l and x1, . . ., x” in {U, V}, is a projector ofA°((U, V}) onto L”({U V})
(§3, no. 2, Corollary m Proposition 1); as 1-1,, belongs to w’aufvn,
H” = r+l<Hn.) NOW
U'V'
(12) Pm;( ‘ —‘)
r_! r!
(13) 1%(1—11"
£=1 n! E)
5,! = 7 + 5((1—1—(“111!U)" (—I—adv)")(—adm!
Ur” )(U)
when r,,l 2 l and I". = 0. Moreover, obviously (ad t)"1.t = 0 ifp 2 2 and
(ad t)°.t = t. It follows that the two sides of (12) are zero when :m 2 2 and
those of (13) are zero when 7,, 2 2. The theorem then follows since HL , is the
sum of the terms of type (12) and H1, is the sum of the terms of type (13).
Remarks. (1) We have defined (§ 3, no. 2, Remark) a projector QofA(X) onto
L(X) such that Q(n‘”) = 0 for aeL(X) and n1 2 2 and QC) = 0. Then
H = Q(exp H) = Q(exp U.exp V), whence immediately
[61
II FREE LIE ALGEBRAS
162
suns’n'ru'nons IN THE HAUSDORFF SEW §5.5
(3) Let g be a Lie algebra with the integral filtration (W9) defined by the
lower central series. Suppose that there exist: m 2 1 such that W9 = {0}.
With the topology derived from the filtration («v-9),“, the Lie algebra g is
Hausdorff, complete and even discrete. Then P(al, . . ., 4,) = 0 for al, . . ., a,
in g and for every homogeneous Lie polynomial P of degree >m; in particular,
H',(a,
, b) = 0 for r + s> m and the series H(a, b): 2 H, .(a, b) has only
a finite number of non-zero terms. The group law (a, b) »—> H(a, b) on g is
then a polynomial mapping (§ 2, no 4-).
PROPOSITION 5. La Km be tlw mmponml qf H(U + V, —U) of mullidagree
(7,5). Then
K..1(U, V) = (—fil—l—f. (ad U)"(V) for» 2 0
We write K(U, V) = H(U + V, -U), K,(U, V) = n; Kn',(U, V). We
denote by L (reap. R) left (resp. right) multiplication by U on A({U, V}).
We can write
Rt' -U = ”Z.—
U’V
p —U ) ‘
_!V (q!
_ 21 (m——R)") V)
— no 11—! mum!
and therefore
1 .
(22) zUVe'-v = go—,7! (ad U)”V.
We now calculate modulo the ideal ”2:0 '22 A’" "({U, V}) of A({U, V}.) For
all 71>, l,
163
1! FREE LIE ALGEBRAS
In this paragraph we assume that K is one of the fields R or C with its usual
absolute value. Recall that a normable algebra over K is a (not necessarily
associative) algebra over K with a topology .7” with the following properties:
(1) .7 can be defined by a norm:
(2) the mapping (x, y) >—> xy of A x A into A is continuous
A normed algebra over K is an algebra A over K with a norm such that
nxynWe< enote
M Mbyforg 211mm A.
a complete normable Lie algebra over K. We choose a
norm on 9 and a number M > 0 such that
164-
GROUP GERM DEFINED BY A COMPLETE Homo LIE § 7.2
f, g in P(g‘; 9), x 59‘; this Lie algebra structure can be extended by con-
tinuity to 15(31; 3) and makes it into a complete Hausdorfffiltered Lie algebra.
By proposition 2 of § 6, no. 3, there exists one and only one continuous Lie
algebra homomorphism ([21:11 >—> a of L(I) into f’(9‘; g) mapping the indeter- ‘
minate of index i to pr‘ for all ieI, since prieP(g‘;g). It follows that
fieP(g‘; g) for ueL(I); more precisely, when uEL(I), I? is just the poly-
nomial mapping (a) »—> u((t,)) of §2, no. 4-. On the other hand, clearly 4:, is
compatible with the multigmduations of L(I) and P(g‘; 9). If u = Viz“ u,,,
where u., e L"(I) for v e N’, then
veN
a= veNI
Z wherefivemss)
Let u = (11,)”, be a finite family of elements of L(I), let ueLU) and let
w = v on (§ 6, no. 3). We write i2 = (17,)msfi. Then
(2) iiofi=(vou)".
This follows by extending by continuity formula (7) of §6, no. 3 and Earn
Dg‘femttiable and Analytic Manifolds R, Appendix, no. 6.
where
Now let u and u be two positive real numbers such that u + v < log 2;
then 0 S sip(u + v) — l < l; the series derived from (5) and (6) by substi-
tuting u for U and 1/ for V are convergent and the above calculations imply
that
(9) 1.1)!)
mJu'v‘ = —log(2 — exp(u + 11)) < +00.
Let r, s 2 0 and let "fin,“ be the norm of the continuous-polynomial Id”
(Dg‘firmtiable and Analytic Mamfllds, R, Appendix, no. 2).
Lemma 1‘ 1|fi,,||<Mm-l ..
Letihs‘be 1aorl i< m,w1th:m— l wewriter=2rbs=254
and consider the following element of L({U, V}):
m— l
166
GROUP Gm DEFINED BY A COMPLETE NORMED LIE § 7.2
whence, by (8)
- M , 4S - 1
168
Exronanruu. m COMPLETE nomo ASSOCIATIVE ALGEBRAS §7.3
then i = 2": (L, where 12‘, is the polynomial mapping (tom, H u,,((t‘)).
Let u = (11,)“; be a finite family of elements of A(I), let ueAfl) and
write w = you (§ 5, no.1). Then
(18) (uou)”=17°fi.
This follows by extending by continuity formula (2) of §5, no. I and from
Diflmztiable and Analytic Mani/our, R, Appendix, no. 6.
In particular we take I = {U}, identify A and Au" and consider the images
2‘ and i of the series 2(U) = “Z! U‘In! and [(U) = n; (—- l)""U"/n in
15(A; A). Then “13"" S l for "x, . . .xn“ S “a“ . . . “an“ for x1, . . .,x,, in A.
Therefore the radius (fabsolulz convergent: 4J5 (resp. I) i: itgfinitz (resp. 2 1).
We shall denote by Q (resp. IA) the analytic mapping of A into A (resp. of
B into A, where B is the open unit ball of A) defined by the convergent series
5 (resp. i) and we shall write ep(z) = l + :A(x) (for s) and
1084*) = M" — 1)
(forzEA, “x — I" < 1). Then
(19) 1"
exp = "Z”! (xeA)
169
1! FREE LIE ALGEBRAS
(I) o = —-
170
p—Amc upmm neurons or 11-11: sums exp, log AND H § 8.1
(3) _ n — so»
”14"!) — fl
1.
v,(n!) = "21 1/,(i) and the number of integers 1' between 1 and n for which
7),.(5) 2 j is equal to the integral part [n/p’] of n/p’. Then
H = ,2 MU. V) e M)
be the Hausdorff series (§ 6, no. 4, Definition 1). Let 20,, be the local ring of
Z relative to the prime ideal (p) and (5,)“, a basis of 1.4,,(1) over Z (§ 2,
no. 11, Theorem 1). It is also a basis oq(I) over Q.
Let H = 2:0 Hme L({U, V)) be the Hausdorff series (§ 6, no. 4-, Definition
1). We shall show that the corresponding formal power series with continuous
components
(10) H = ns>0
2 11461361 x 9,9)
is convergent (Difl'emm'ahle and Analytic Manifuldx, R, 4.1.1).
172
GROUP DEFINED w A COMPLETE NORMED LIE ADGBBRA §8.3
(17) follows immediately from (16) and (16) is trivial for r = r = 0;if1 2 1,
then _
"17...“, y) [I S "Hull IIXII'Ilyll‘
X 7-1 I
g mg) (%)
< 1"";
we argue similarly if: > 1.
In particular, ]|h(x,y)|| < a“ for (my) 5G x G.
Pnoposmon 3. Let G be the ball {at e g] ”x” < a”). The analytic mapping
h: G x G —> G
makes G into a group in which 0 is the identity element and —x i: the inverse afxfw
alIzEG. Moreover, ifR i: a realnumhermh ”1010 < R < a”, the ball
{reel M < R}
(reap. {x 6 9| ||x]| S R}) i: an 0pm subgroup QfG.
As H(U, —U) = o and H(0, U) = H(U, 0) = U, h(x, —x) = o and
h(0, x) = h(x, 0) = x
for all x e G. It therefore remains to prove the associativity formula
(13) h<h<x,y),z) = h(x,h(y,2)) forx,y,zin G.
As
H(H(U,V),W) = H(U, H(V, W))
in L((U, V, W)) (§ 6, no. 5, Proposition 4), we have
(19) FIo(H >< Ida) =I.-Io(Id, x H)
in 15(9 x g x g; 9) (no. 2) and (19) implies (18) by (16) and Dgflerentiable
and Analytic Manyoldr, R, 4_1.5_
*In other words (Chapter III, §l), G with the Hausdorff function is a
Lie group.,,.
for x, y in A, and ample” with this norm. The results of the second and third
paragraphs of § 7, no. 3, remain valid.
We take I = {U} and consider the images 2 and iof the series 2(U)= :— U
101 71—!
andl(U)= Z (—1)"‘1U—in P(A- A). Then
mu
(21) (U7)~ S a'mp
PROPOSITION 5. Far 1, y in G,
(26) exPA-CXPAy = ep m": .11)-
2%" = emu-V) and hence
ma(1 +€,1+E)=(l+i)afl
in FHA x A;A) (where m denotes multiplication on A). The proposition
then follows from Proposition 2, Lemma 3 and Dyfirmtiabl: and Analytic
Manifolds, R, 4.1.5. .
APPENDIX
MOBIUS FUNCTION
Let n be an integer >1. If n is divisible by the square of a prime number, we
write y.(n) = 0. Ifn is not divisible by the square ofa prime number, we write
p.01) = (—l)", where k is the number of prime divisors of n. The function
u: N“ —> (— l, 0, 1} thus defined is called the Mfibiuxfirwtian.
Recall that given two integers n1 2 l, 712 2 1, we write n,|n2 if 711 divides
n2.
Pnopoamou. (i) The fimm'on y. i: the unique mapping If N“ into Z such that
y.(1) = 1 and
(1) “2M = o
for way integer n > 1.
(ii) Let r and t be two mapping: JN‘ into a cammutative group written additiuely.
In order that
(2) :(n) = gm!) firevay integern 2 I,
it i: mommy and mgficient that
(a) t(n) = 1% “(0:6) for my integer n 2 1.
The uniqueness assertion in (i) is obvious, for (1) allows us to determine
Mn) by induction on n. We show that the function y. satisfies (1). Let It be an
integer >1. Let P be the set of prime divisors of n and let u = 1—; 11%“) be
us
the decomposition of 7: into prime factors. If at is a divisor of n, then y.(d) = 0
unless a' is of the form E1), where H is a subset of P. Then
Z W) = Z (—1)°-""*
Iiln 3:?
176
APPENDIX
2:01) = an»
am 229(3):“)
m E = m
2:01) dun/u)
2 “(8) = :(n),
which completes the proof.
Formula (3) is called the Miz'bimr itwem‘mflrmulm
177
EXERCISES
§1
1. Suppose that K is a field of characteristic 0. Let E be a cocommutative
bigebra of finite rank over K. Show that P(E) = (0} (apply Theorem 1 of
no. 6).
2. Let E be a bigebra such that P(E) = {0} and let (En),.,,, be a filtration
of E compatible with its bigebra structure. Show by induction on n that
E; = {0} for all n 9 0 and deduce that E“ = {0}, Le. that E reduces to K.
3. Let G be a monoid and E = K[G] its algebra (Algebra, Chapter 111,
§2, no. 6).
(a) Show that there exists on E one and only one cogebra structure such
that :(g) = g ® g for all g e G; this structure is compatible with the algebra
structure on E and makes E into a cocommutative bigebra whose counit a
is such that e(g) = 1 for all g e G.
(b) Show that every primitive element of E is zero. Deduce that, if G eé {e},
the bigebra E admits no filtration compatible with its bigebra structure (apply
Exercise 2).
(6) Suppose that K is an integral domain. Show that the elements of G are
the only non-zero demerits x e E such that :(x) = x ® x. Show that E has an
inversion 1' (cf. Algebra, Chapter III, § ll, Exercise 4.) if and only if G is a
group and that in that case i(g) : g'1 for all g e G.
4. Let E be a cocommutative bigebra and let G be the set of g e E such
that 5(g) = l and 0(g) = g (8 g. Show that G is stable under multiplication
and that it is a group if E has an inversion. Show that, if K is a field, the
demerits of G are linearly independent over K.
5. Let E be a bigebm and let E' : Hom(E, K) be its dual with the algebra
structure derived by duality from the cogebra structure on E (Algebra, Chap-
ter III, §ll, no. I). Let m be the kernel of the homomorphism ut—>u(l) of
178
EXERCISES
E’ onto K; it is an ideal ofE’. Show that, ifu e m2 and x e P(E), then u(x) = 0.
When K is a. field show that P(E) is the orthogonal of m“ in E"; deduce that
dim P(E) S dim m/m2 and that we have equality if E is of finite rank over K.
11 ll. Let E be a bigebra and let I be a finite set. We are given a basis (4.)
l 79
II FREE LIE ALGEBRAS
2 new,
where x“ = 1:! X7“).
Show that u H d)“ is an isomorphism of E’ onto A and that this isomorphism
maps the topology of simple convergence on E’ to the product topology on A.
(17) Let a“, be the constants of structure of the algebra E relative to the
basis (cu). Then
‘oh = Z ‘uov‘v-
We write x instead of (X05, and y instead of (Y.)m, where the Y‘ are new
indeterminates. Show that there exists a. family f(x, y) = (fi(x,y)).“ of
formal power series in the variables x, y such that
1‘ The exercise in question assumes that K is a field, but nothing is lost if this
hypothesis is suppressed.
180
EXERCISES
restricted enveloping bigebra (Chapter I, §2, Exercise 6). Show that there
exiss on 0 one and only one bigebra structure which is compatible with its
algebra structure and for which the elements of g are primitive. Show that 1-} is
commutative and that Pa?) = g.
(b) If n is an integer 20, let C1,, denote the vector subspace of El generated
by the products x,. . .x” where x‘ e g for all 1'. Show that (73“),0‘, is a filtration
of (I compatible with its bigebra structure.
(a) Let E be a bigebra and let 5 = P(E) be the Lie algebra of its primitive
elements. The mapping x >—>x" leaves 9 stable and gives 9 a Lie p-algebra
structure. Show that the canonical injection 3 —> E an be extended uniquely
to a bigebra morphism U —> E and that this morphism is injective (use the
fact that, if (Mm is a basis of g with a total ordering, the monomials H2:
(0 S n. < 12) form a basis ofU (loo. oil.) and argue as in the proofofLemma 2).
§2
1. Let g be a free Lie algebra with basic family (x1, . . .,x,,. . .). Let 9,,
denote the subalgebra of 9 generated by x”. ..,x,, and let 5,, denote the
smallest ideal of g,‘ containing .7,"
(a) Show that [1,. is the submodule of 9,, generated by the
fl 2. Let X be a countable set with at least two elements and let 3 be the set
of subsets of M(X) which are Hall sets (CE Definition 2). Show that
Card(b) = 2‘0.
3. Let X be a set with a total ordering. Show that there exists a Hall set
relative to X such that H n M3(X) consists of the products z(yx) with y < x,
y S z and that H n M‘(X) consists of the products w(z(yz)) with w 2 z 2 y,
y < x and the products (ab)(cd) with a < b, c < dand either a < cora = c
and b < d.
4. (a) Show that the Lie algebra defined by the presentation
admits the basis (2:, y, [3, y]). Define an embedding of this algebra in a matrix
algebra.
(b) Do the same for the Lie algebra. defined by the presentation
(at: [:8 [m1] — 2x = [a [a y]] + 2y = 0}.
(1:) Show that the Lie algebra with presentation
{2,y,z;[x,y] -x = [mz] ~11 = [4*] - z = 0}
reduces to 0.
5. Let g be a free Lie algebra and let I be an ideal of 9. Suppose that
r = [g, I]. Show am 2 = {0} (use the fact that "Q“ m = {0}).
6. Let E be a module. Let ME be the flee algebra of E (Algebra, Chapter
111, §2, Exercise 13). Recall am ME = g; E," where
131:3, E,,=y£‘qD_nE,®Ea if ">2,
and the algebra structure on ME is defined by means ofthe canonical mappings
Eu 8 En -> BMW
(:1) LetJ he the smallest two-sided ideal of ME containing the elements of
the form xx and (IQ/)2 + (yz)x + (zx)y with x, y, z in E. Let LE = ME/J.
Show that] is a graded ideal of ME; deduce a graduation (L"E),k>1 of LE.
(b) Show that LE is a Lie algebra; it is called the free Li: algtbra of the
module E. Show that for every Lie algebra g and every linear mappingf: E —> g
there exists one and only one Lie algebra homomorphism F: LE —> g which
extends f.
(0) Define isomorphisms E —> LE and A 2E —> LE. Show that LSE is
identified with the quotient of E ® AaE by the submodule generated by
the elements
x®(y A z) +y®(z Ax) +z®(x Ay) forx,y,zinE.
(d) Show that, when E is a free module of basis X, LE is identified with the
free Lie algebra L(X) defined in no. 2.
(a) Show that the canonical mapping of E into the enveloping algebra
U(LE) of LE can be extended to an isomorphism of the tensor algebra TE
onto U(LE).
(f) Let 0' be the linear mapping of A 2E into TzE such that
a(xAy)=x®y—y®x.
Construct a module E for which a is not injective. Deduce that for this module
the canonical mapping of LE into U(LE) is not injective (compare with
Exercise 9 of Chapter I, § 2).
182
EXERCISES
7. Suppose that K is a. field. Let I be a free Lie algebra with basic family
(mm! and let M be an I-module.
(a) Show that H2(l, M) = {0}, of. Chapter I, §3, Exercise 12. (Use Corol-
lary 2 to Proposition 1 and part (i) of the Exercise in question.)
(b) Show that, for every family (711,)”, of elements of M, there exists a
cocycle 4): I——>M of degree 1 such that 420:.) = m. and that this cocycle is
unique. Deduce an exact sequence:
0—>H°(l, M) -—>M—>M"—>H‘(l, M) —>0.
If I is finite of cardinal n and M is of finite rank over K, then
rg H1(I, M) —— rg H°(l, M) = (n — l) rgM.
8. Suppose that K is a field. Let I be a free Lie algebra with basic family
(2),“, let r be an ideal of 1 contained in [l, I] and let 9 = l/r; let x, denote the
image of 2‘ in g.
Show the equivalence of the following properties:
(i) g is a flee Lie algebra.
(ii) g admits (2;) as basic family (i.e. t = (0}).
(iii) For every g-module M, H38!) M) = {0}.
(iv) If g operates trivially on K, then H’(g, K) = {0).
(The implications (ii) => (i) and (iii) 3 (iv) are obvious and (i) => (iii)
follows from Exercise 7. To prove that (iv) => (ii) it suffices to show that
r = [L r], cf. Exercise 5; otherwise take a hyperplane b of 0: containing [1, r]
and note that the extension r/b —> [/9 —> 1/: = g is essential.)
183
[1 FREE LIE ALGEBRAS
where the product is relative to the order relation given on Y. Let 6 denote the
subset of Y x M consisting of the ordered pairs (14, m) such that there exists
y e Y for which 3/ > u and m(g) 2 1. Show that the elements
0’"(u) for (u,m) 56 and 6’"(z) for (z, In) El x M
form a basic family of the Lie algebra fly.
(Reduce the problem to the case where Y is finite and argue by induction
on Card(Y). Use the Corollary to Proposition 10 applied to the greatest
element 3/ of Y and apply the induction hypothesis to Y - {y}.)
12. Let X = {1,31} be a set of two elements. Show that the derived algebra
of L(X) is a flee Lie algebra admitting as basic family the elements
((ady)“ a (ad z)’)(y), for]: 2 l, q 2 0. (Use Exercise ll.)
1[ 13. (In this exercise we assume that every projective module over K is
free; this is so for example if K is a principal ideal domain.)
Let I = "21 I,‘ be a graded Lie algebra admitting a basic family B consisting
of homogeneous elements and let 0 be a graded Lie subalgebra of l, which is a
direct factor of I considered as a module.
(a) For all i 2 0, let I“) be the graded subalgebra of I such that
1;" = b, ifj <21
1}”: I, ifj> 1'.
Then I = I“) D I“) D . - - D I). Show, arguing by induction on i, the existence
of a basic family B“) of [(0, consisting of homogeneous elements, such that the
elements of B“'” and of B“) of degree <i — l are the same. (Suppose that
B“‘” has been constructed Let m, be the intersection of I‘ = 6““ with the
subalgebra generated by the b, for j < i and let b; be the submodule of l,
generated by the elements of B““’ of degree 1'. The induction hypothesis
implies that I‘ = tnl ® 1).. As m, C bl, we can decompose has b‘ = 1). ® 3., so
that I). = m, ® 6:; by the hypothesis on K the modules 9‘ and a, are free;
184
EXERCISES
185
II FREE LIE ALGEBRAS
(0) Take 5 to be a free Lie algebra with basic family (X1, . . . , X,u H),
where n > 0, and take b to be the ideal of 9 generated by H. Show that g and
all) are free modules. Deduce that, in that case, in, = 11.. In particular, there
exism x e ”a such that (n(g))(x) = 6(Xl) . . .a(X,.)a(H).
(d) We return to the general case. Show that u, is generated over K by the
elements of the form 60:1). . .a(x,)o(h), with n 2 0, x) e g and h e 9 (note that
“a coincides with the left ideal generated by 1)). Show, using (0), that such an
element belongs to it, (use a suitable homomorphism of a free Lie algebra into
9). Deduce that fla = uh. v
(a) Show that, if n(5) is bijective, so is 11(g/b). Deduce finally that, for every
Lie algebra 9,1;(9) is bijzttiva (write g as the quotient of a free Lie algebra) and
that the canonical homomorphism
(015(9) —> grU(g) (cf. Chapter I, § 2, no. 6)
is an isamorphirm (“Poincaré—Birkhofi—Witt Theorem” for Lie algebras over
Q—algebras).
§3
The letter X denotes a set.
I. (a) Show that every element u E A"(X) can be written uniquely in the
form it = “x uzx, where ux e A(X).
(b) Show that {0) is the only submodule ofA+ (X) which is stable under all
the mappings u ~—> u,‘ (x e X); (Ifa is such a submodule and a aé {0), consider a
non-zero element of a of minimal degree.)
2. Let g be a Lie algebra which is a free module; it is identified by means of
a: g —> Us with its image in the enveloping algebra U9, Let U *9 be the kernel
of the canonical homomorphism Ug —> K. Let i be a mapping of X into 9 such
that i (X) generates g as a Lie algebra.
(a) Show that U+g is generated by i(X) as a left Ug-module.
(b) Show the equivalence of the following properties:
(i) g is free with basic family i: X —> 9.
(ii) The family i is a basis of the left Ug-module U*g.
(The implication (i) 9 (ii) follows from Exercise 1 (4) using the isomor-
phism Us; —> A(X). To show (ii) => (i), apply Exercise 1 (b) to the kernel of
the homomorphism A(X) —> Ug defined by L)
3. Let x e X. Show that the centralizer ofx in A(X) is the subalgebra gener-
ated by z. Deduce that the only elements of L(X) which commute with x are
the multiples ofz. In particular, the centre ofL(X) is (0) if Card(X) 2 2 and
the centre of L(X)/(n; L"(X)) reduces to the canonical image of L”(X).
186
EXERCISES
§4
In thefollowing exorcism the let!” G denote: a group.
1. Let (Gn) be an integral central filtration on G. Show that the Lie algebra
gr(G) is generated by gr1(G) if and only if G" = GM1.C”G for all n 2 1. In
that case, show that G" = 0”,.a for m > n and deduce that (0,.) = (C"G)
if there exists an integer m such that Gm = {2).
2. Let (Gm) be a real filtration on a group G and let u be the corresponding
order function. Let H be a subgroup of G.
(a) Let Hu = H n G“. Show that (H,) is a real filtration of H and that the
corresponding order function is the restriction of v to H. H: = H n G: and
gr(H) is identified with a graded subgroup of gr(G).
(1)) Suppose that H is normal and that ”(G) n R is a discrete subset ofR.We
write (G/H)“ = (GnH)/H. Show that ((G/HL) is a real filtration ofG/H and
that the corresponding order function 05,3 is given by the formula
”GIEO‘) = $3314!)-
Show that, for all at E R, there is an exact sequence
0 —> gram) -> grAG) —> gram/H) —> 0.
187
II FREE LIE ALGEBRAS
(6) With the hypotheses of (b), suppose that (G“) is a neutral filtration. Show
that so are the filtrations on H and G/H induced by (Ga) and that the Lie
algebra gr(G/H) is identified with the quotient of gr(G) by the ideal gr(H).
fl 3. Let (Hm) be a real filtration on a group H and let vH be the correspond-
ing order function. Suppose that G operates on H on the left and that, for all
g e G, the mapping h »—> g(}z) is an automorphism of the filtered group H.
(a) If a E R, we denote by Ga the set ofg E G such that
vn(h‘1g(h)) 2 1/301) + a for all h e H.
Show that (G) is a real filtration of G and that Gm = G if a: S 0. Let 0 denote
the corresponding order function.
(b) Suppose that the filtration (Ha) is central and that G; = G. Show that
(GE) is a central filtration of G. If E E grm(G) and n E grad-I), let g (resp. h) be a
representative of E (resp. 1;) in G“ (resp. Hg) 3 show that the image of h'1g(h) in
grm+ fl(H) is independent of the choice ofg and h; ifit is denoted by D301), show
that D: can be extended to a derivation of the Lie algebra g'r(H) of degree on
and that EH D: is a homomorphism of the Lie algebra gr(G) into the Lie
algebra of derivations of gr(H). If 123(H) n R is contained in a discrete sub-
group I‘ ofR, so is 0(G) n R and, for all a e R, the mapping EH Dz defined
above is injective.
4. Let H be a nilpotent group ofclass c and let G be the group of automor-
phisms of H which act trivially on H/(H, H). Show that G is nilpotent of class
<2 —- 1. (Apply Exercise 3 to the lower central series of H and note that G
operates trivially on gr(H).) Show that, if H is a finite p—group (p prime), so is
G (same method).
5. Let K be a commutative field and let L be a finite Galois extension of K
with Galois group G. Let u be a valuation of L with valud in Z (Commutative
Algebra, Chapter VI, § 3, no. 2) which is invariant under G. If g e G, we
write
188
MERGERS
K[G] = K («B I and I admits as basis the family ol'g — 1 wherege G-{z}.
(a) If n is an integer 2 0, we denote by I’I the ideal of K[G] the n-th power
of I. Let G" be the set ofg e G such that g — l e I". Show that (G) is an in-
tegral central filtration on G. In particular, G" 3 CG for all n.
(b) Show that if K = Z the mapping g >—> g — 1 defines on taking quotients
an isomorphism of G/(G, G) onto III“. Deduce that G, = C’G.‘|’
(5) Suppose that K is a field of characteristic zero and that G is finite.
Show that I" = I foralln 21.
(:1) Suppose that K is a field of characteristic p > 0 and that G is a p-group.
Show that I"(0} for n sufficiently large. (Show first, using Algebra, Chapter I,
§6, no. 5, Proposition 11, that every simple K[G]-module is isomorphic to K ;
deduce that I is the radial of K[G] and hence is nilpotent since K[G] is of
finite rank over K.)
(2) Suppose that K = Z and that G has the following property: for all
g e G such that g aé 1, there exists a prime number p, a p-group P and a homo-
morphisrnf: G—> P such that f(g) # a. Show that in that case 0 I” = {0}.
(Reduce it immediately to the case where G = P. By applying (d) to the field
17,, We see that there exists m such that I'" C p.Z[G] and, as I is a direct
factor in Z[G], this implia that I"l C p1, whence Q I'"" C Q p"I, which
reducu to 0 since I is a finitely generated Abelian group.)
7. Let G be given the filtration (C‘G) and suppose that gr, (G) = G/(G, G)
is cyclic. Show that grn(G)= {0} for n 2 2 (use Proposition 5) and deduce that
C"G= (G, G)forn2 2. '
s. LetG=SL2(Z)andletx=(01 1),y=(l
1 1 o1),w=(_lo 0).
l
(a) Verify the formulae w4= l, w = xy“x, warm" = y'.
I
1' We do not know whether G, = C”G for all n. It is true in any case when G is a
file group, cf. §5, Exercise 1.
1 It can be shown that E is nfnrdn 12.
189
X! FREE LIE AMEBRAS
(b) Let (11):“ be a generating family of G and let Ill 2 1. Suppose that, for
all (i,j) e I”, (x,, x,)"' 5 (PG. Show that, {or all q 2 2 and all u 6 04G, 14'" E C“ 1G
("56 (3))
10. Let x, y be in G and let i, 5 be two integers 21. Suppose that (x', y‘) = e.
(a) Show that, for all n 2 l, (x, y)”I e C”*’G. (It can be assumed that G is
generated by {x, y}; then apply Exercise 9 (b), noting that (x', y') E (x, y)"
mod. 03G.)
(b) Suppose that z’ = y’ = a; let t denote the g.c.d. of r and .t. Show that _
(x, y)‘ e CaG and deduce that (x,y)“ E C"”G for all n 2 l (same method).
11. Let H be a subgroup ofG and let in be an integer 2 1. Suppose that G is
generated by a family on)“, such that x2" 6 H for all i.
(a) Let H be given the filtration induced by the filtration (C’G) on G and
let gr(H) be identified with a graded Lie subalgebra of gr(G), of. Exercise 2.
Show that, for all n 2 0,
m".gr,,(G) C gr,(H).
Deduce that, for all z e H. CG, 2’” e H. O‘HG.
(b) If G is nilpotent, show that there exiss an integer N 2 0 (depending
only on the nilpotency class of G) such that 2"" e H for all z e G.
12. Suppose that G is nilpotent. Let H be a subgroup of G and let m be an
integer 2 l and let x, y be elements ofG such thatx’" E H and y'" e H. Show that
there exists an integer N 2 0 such that (xi/W" e H. (Apply Exercise 11 to the
group generated by {x, y) and its intersection with H.)
13. (a) Let F be a Bee group with basic family (7:, y) of two elements, let
c be an integer 22 and let m be an integer 21. We write Fc = F/C°F and de-
note by x, y the images of}, y in F‘. Let Ff,l be the subgroup of F” generated by
{#311}. Show that there exists an integer N 2 0 such that 2“" E F; for all
z e F0 (use Exercise 11).
([1) Let Ic (reap. 1;) be the normal subgroup of Fc (resp. Ff") generated by y.
Show that, if N is chosen as above and :5 1‘, then 2“" e I; (note that FCIIc
is an infinite cyclic group with generator the image of z and deduce that
Fin/If, —> F°/Ic is injective and hence that I; = Ff, n I”). In particular
xy’""x‘1 e If".
(6) Suppose that G is nilpotent. Let H be a subgroup of G and let L be a
normal subgroup ofH. Let g E G and let in 2 1 be such that g'" E H. Show that,
if N is sufficiently large, glm" g” e L for all I e L. (If G is of class < 1:, choose N
as in (a) and use the homomorphismf: F9 —> G such thatf(x) = g, f (y) :1;
note thatf(a) C H, f(1,3,) C L and apply ([1) above.)
1} 14. Let P be a set of prime numbers. An integer n is called a P-integn if it
is #0 and all its prime factors belong to P. An element x e G is called a P-
tam'on element if there exists a P—integer n such that x“ = e; G is called a P-
190
EXERCISES
192
EXERCISES
isomorphic to Z. Deduce, using (b) and (c), that j}(H) is induced by £(G).
(0) With the hypotheses of (d), let P' denote the complement of P in the set
of prime numbers. Show that the closure of H under 9;,(G) coincides with the
P’-saturation of H in G (Exercise 14) ; in particular, G is Hausdorff if and only
if it is P’-torsion-fiee.
18. The upper central series (ZLG) of the group G is defined inductively as
follows:
(i) Z,G = (e}i.fi< 0;
(ii) ZiG/Z¢_1G is the centre of G/ZhlG.
Then {a} = ZoG C Zc and ZIG is the centre of G; the LG are
characteristic subgroups of G.
(a) Show that G is nilpotent of class <5 if and only if G = ZbG.
(b) Show that (C"G, ZmG) c Zm_nG.
(0) Suppose that G is nilpotent and P-torsion-fi‘ee (where P is a set of
prime numbers, cf. Exercise 14). Show that the Z.G are P-saturated. (It
suffices to verify this for ZIG; ifn is a P-integer and g e G is such that g" e 21G,
then xg"x'1 = g" for all x e G, whence xgx'1 = g by Exercise 14 (r) and it
follows that g belongs to 21G.)
§5
In the following exercises we assume the hypotheses and notation of §5.
F denotes the free group F(X) and g the unique homomorphism of F into
the Magnus group I‘(X) such that g(::) = l + x for all xEX (cf. Theorem
1).
1. Let the algebra K[F] of F be given the filtration (1") consisting of the
powers of the augmentation ideal I (§ 4, Exercise 6).
(a) Let E: K[F] —> A(X) be the unique algebra homomorphism extending
g: F —>A(X)*. Show that 3 maps I" to the ideal A,(X) (cf. no. 1) and defines
when taking quotients an irorrwrphism E” of K[F]/1'I onto A(X)/A(X). (Define
an inverse homomorphism of g, by means of the homomorphism of A(X)
into K[F] which maps 1: to x — l for all xeX.) Deduce that A(X) is im-
morphic tn the Hausdoyftamplefim a[F] under the topology defined by (1").
(11) Suppose that K = Z. Show that the filtration (1") is separated (use
Proposition 3 and Exercise 6 (e) of § 4-) and that the filtration of F defined by
it (Exercise 6 (a) of §4) coincides with the filtration (C‘F). Deduce that -
E: Z[F] —> AZ(X) is injective.
2. Let G be a group, let K[G] be its algebra over K and let I be its aug-
mentation ideal (§ 4-, Exercise 6). Let 5 denote the canonical homomorphism
of K[G] onto K; then Ker(e) = I.
(a) Let M be a left K[G]-module and let Z(G, M) be the group of crossed
hommorfihimu of G into M (Algebra, Chapter I, §6, Exercise 7). If
[93
11 FREE LIE ALGEBRAS
n = so» + 2 MW — 1),
:EX
where D,(u) E K[F] and the D,,(u) are zero swept for a finite number. The
mapping u v—> D,(u) is called the partial derivative with respect to 1:. It is charac-
terized by the following properties:
(i) D, is a K—linear mapping of K[F] into K[F].
(ii) Doom) = u.Dx(II) + s(v).Dz(u) for u, v in K[F].
(iii) D,(x) = l and Dg(y) = 0 inX — {x}.
If n 2 1, then
D,(x") =1 +x+ +z"'1
D,‘(x"‘) ..__ —x"‘ _ x-IH-l _ . . . _ 3-1,
1‘ For more details on the D3, sec R. Fox, Ann qf Math, 57 (1953), pp. 547—560.
194
EXERCISES
(b) Let R be a normal subgroup ofF, let G = F/R and let] be the kernel
of the canonical homomorphism 7: K[F] ~> K[G]. Show thatj is generated
as a left (resp. right) ideal by the elements I ’— 1, where 16R. Prove the
exactness of the sequence
(*) o—>J/q.1)i>1/0.1)_”>K[G]-i>K—>o,
where at is derived by taking quotients from the inclusion of] in I and fl is
derived by taking quotients from the restriction ofy to I.
(c) For n e I, 9: ex, let 5,04) denge the image of Dx(u) in K[G] under
y. Show, using (a), that the family (Dmx defines on taking quotients an
immarfihirm of I/(J ‘1) onto K[G]“L
(d) For 76R, let 0(r) be the image off ~ 1 inJ/(JJ). Show that
9(r1’) = 9(1) + 9(7') and 9(y7y‘1) =y.0(r) forr,r’inR,yEF.
(Use the identities
rr’—l= (r— l)(r'-— 1) + (r—l) +(r’—l)
m" - 1 =y(r - 1)(y" - 1) +y(r - 1).)
Deduce that 6 defines a homomorphism of R/ (R, R) into J/(J .1) compatible
with the action of G and show that the image of this homomorphism generates
the K-module j/(JJ); when K = Z, show that an bamphiwn is thus ob-
tained of R/ (R, R) onto j/(JJ) (define the inverse homomorphism directly).
(5) Let (1a)“E A be a family of elements of R generating R as a normal
subgroup of F. Show that the 9(a) generate the K[G]-moduleJ/(J .I).
(f) The matrix (:I5,,,(rm)),‘exvc‘5A defines a homomorphism
p: K[G]‘A’ —> K[G]"n
of left K[G]-modules. Show that the sequence
T For more details on this exercise, see K. W. Gruenberg, 11mm Note; in Math,
no. 143, Chapter 3, and R. Swan, J. qlgcbm, 12 (1969), pp. 535—601.
[95
ll. FREE LIE ALGEBRAS
5. Let F = F(X) be the Magnus group on X and let (1}) be its natural
filtration (no. 2). Show that the graded Lie algebra gr(I‘) corresponding to
this filtration is isomorphic to the Lie subalgebra g An(X) ofA(X). IfX 96 0,
this Lie algebra is not generated by its elements of degree 1 ; deduce that
(I‘n) is not the lower central series of 1".
fl 6. Let P be a set of prime numbers, let 8,, be the set of P-integers (§ 4»,
Exercise l4) and let Z? = S; 1Z.
(a) Suppose that, for all 5681,, the mapping [Ct->5]: is a bijection of K
onto itself; this is equivalent to saying that K can be given a ZP-algebra
structure.
With the notation of Exercise 5, show that, for all 355?, the mapping
a n—> 11a of 1" into itself is bijective and that the same is true for each quotient
I‘ll",l for n 2 I. If! e ZP and a E F (resp. a e l‘/l",l , u‘ is defined as in Exercise
16 of § 4; writing a in the form l + a, where a EA,(X), show that
a‘ = <1 + «r = .2 0“"- I
(Note that the coefficients (fl) belong to 21,, cf. Exercise 4-.)
(b) Suppose now that K = 21,. Lets be an integer al. The group
F“ = F/CcF is identified with a subgroup of I‘ll"c by means of the homomorph-
196
memes
ism derived from g by taking quotients (notation of Theorem 2). The group
F/I‘L. is a P-torsion-free P—divisible group of class <r. Let F; be the P-satura—
tion of F” in l"/I‘c (§ 4, Exercise 14) and let i be the injection of F” into Ff,
The ordered pair (1', F?) is a P—ermlape of Fc (§ 4, Exercise 15). Deduce that
«my m'lpatmt gum]; ha: :2 Rem/clap: (note that every nilpotent group of class
<6 is the quotient of a group F“ for suitable X and use part (b) of Exercise
15 of § 4).
(c) If n < a, let Fi>.n be the intersection of Ff, with I‘n/I‘,; if n 2 c, we
write FE,“ = {a}. Show that F11”, is the P-saturation of C”F° in F/I‘c. The
filtration (F;.,.) is an integral central filtration of Ffs; let gr(F§) be the graded
module associated with this filtration. Show that, if n < r, the image of
grn(F§) in gr,.([‘) = AQIJX) is S;‘.L;(X) = L;,(X). Deduce that the Lie
algebra gr(F§) is generated by its elements of degree 1 and hence that
Ff“, = C"(Ff=) for all 7: (§ 4, Exercise 1). Show that the group Ff, is generated
by the x1" for x E X, .r 6 5; (note that the imagm of these elements in
gr,(Ff=) 2 Z?" generate the group gr,(Ff,) and apply Corollary 3 to Proposi-
tion 8 of Algebra, Chapter I, §6, no. 3).
(d) Let H be a Hall set relative to X (§ 2, no. 10) and let H(c) be the
subset of H consisting of the elements of length <5. For meH(c), let ¢o(m)
denote the image in F” of the basic commutator Mm) defined by m (of. no. 4,
Remark). Show that, for all w e F; there exists a unique element a 529“”)
such that
H «(mm
=nemn)
197
11 FREE LIE ALGEBRAS
The sequence (10,) is called typical if, for every ordered pair (It, a) the func-
tion wk“: 71 »—> wk'a(n) is a binomial polynomial nfdtgnz <15, cf. Exercise 4. This
condition is independent of the choice of the bases (61m)- Show that it is
equivalent to the existence of a sequence (49“,, of elements of Aq(X) such
that m(ak) 2 k for all k and
w, E I; “mum-'0 mod c
for all n e Z and all 5 2 1 (cf. no. 4, Remark). Show that (run) is typical ifand
only if, for all m e H, the function m—> a(m, n) is a binomial polynomial of
degree sl(m), where l (m) is the length of m.1‘
(f) A sequence (wn) is called l-Iypical if the ccrraponding functions wk‘,
are binomial polynomials of degree <1: —- 1. Show that, if (wn) is l-typical,
there existy. E C‘“F fori = 0,1, . . ., such that
for all n E Z and all i > 1. (Apply Exercise 8 (d) to the sequence of x‘"(x_1/)".)
Show that w; = y‘1(y‘1, x‘1)y E (x,y)'1 mod C’F.
(b) Let G be a nilpotent group of class <0 and let x, y be in G. Deduce
from the above the following formula (“Hull’iformula”);
c
(xy " = WU mm)“.
(c) Letpbeaprimenumbenshowthatthereexismu.EC‘F(2 < i <1: — l)
and w 5 OT such that
Qty)” = x’y’ug r . . u:_1w.
(Use (a), noting that (p) is divisible by 1: if 1 < i < p.) Let u') be the image
1
of w in gr,(F) = LflX) and let 1?), be the image of Liv in LEAK). Show that
u'), = ,(x,_t/), cf. Chapter I, § 1, Exercise 19. (Use the extension g of F to
A(X) and compare the terms of degree [I in g((xy)’) and g(z"y’u§ . A , u,’;_,m).
The first is equal to (x + y)” and the second is congruent mod. p to
x” + y” + rm. Hence the result.)
Show that there exist a E 03F and w’ 6 OT such that
(zya " = x’y’w’.
(Use the above formula for (:gr)’.) Deduce that in a nilpotent group of class
1; the [z-th powers form a subgroup.
(:1) Show that there exists a sequence ya, 11;, . . . of elements of F, where
0‘ e C‘F for all i and
n n
(m) a (mun?) . . . ultl mod CMF
for all n E N and all i 2 1. (Apply Exercise 8 (f) to the sequence of (x“,y).)
Deduce that, if p is a prime number, there exism t.eC‘F for 3 < i s p
and z e CH‘F such that
(fly) = (’6n - . : '51-
Show that the image 2, of z in 1.2:1(X) is (ad x)’(y). (Same method as for
(PM
11 10. Let p be a prime number and let G be a gmup with a real central
filtration (GK). Suppose that the relation xeGa implies z” e G”, in which
use the filtration (Go!) is called mm'md‘
(a) Show that the Lie algebra gr(G) associated with (G,) is such that
p.gr(G) = O and hence can be given an algzbm structure over F,.
(17) Let Es gru(G) and let x be a representative of E in G“. Show that the
image ofx’ in n(G) does not depend on the choice ofx (use Exercise 9 (0)).
If this image is denoted by Em, prove that E» EU” is a linear mapping of
gr,(G) into grw(G) and that (g + a)!" = E'" + Em + A,(E, g'). (Same
method.)
199
11 me LIE menus
200
EXERCISES
(b) Let at, g be elements of 9 such that A,(ax, by) = 0 for all a, b in 1:.
Show that, if M,” is the bihomogeneous component of A” of bidegree (r, r),
then A;"(x, y) = 0 for every ordered pair (7, r). Deduce (taking r = [I — l,
.\‘ = I) that (ad x)"1(y) = 0. In particular, if A,(x,y) = O for x, y in g, the
Lie algebra 9 satisfies Engel’s (p — l)-th condition.
(it) Let c be the centre of 9. Suppose that (ad x)” = 0 for all xeg. Show
that g/c satisfies Engel’s (p —- l)-th condition. (Show that A,(ad x, ad y) = 0
for x, y in g and apply (b) to the Lie algebra ad 9 = g/c.)
(4) Let G be a group such that z” = I for all x e G and let (Gm) be a real
central filtration of G. The filtration (G2) is remitted (Exercise 10) and gr(G)
is a Lie p—algebra whose p—mapping is zero. Deduce that gr(G) satisfies Engel’s
([1 — l)-th conditionif
§6
1' For properties of Lie algebras satisfying an Engel condition and their applica-
tions to the “restricted Burnside problem”, see A. Kostrikin, Izu. Akad. Nauk SSSR,
23 (1959), pp. 3734.
201
II FREE LIE ALGBBRAS
the Bemom'lli number: (Fumtiom'qf a Real Variable, Chapter VI, § 1, no. 4). Show
that
(—2“) . heed U) W)
H.(U, V) = (g(ad U))<V) = V + an V] + Z—
Derive the first few terms of the expansion of H1(U, V):
202
mncrsm
[U, V] = flung").
For example
U + V = U.V.(U,V)"f...
mw=mwm
(d) Let 9 be a nilpotent Lie Q—algebra with the Hausdorff group law. Let
u, v be in 9. Show that
u + u = flmx(u, 9)“)
mo=£lmomwu
where at and 5 are the families of rational numbers defined above (“inversion
9/ the Hamdorflformula”). (Use the continuous homomorphism 45: L —> g such
that (MU) = u, ¢(V) = u and note that it is also a homomorphism for the
Hausdorff group law.)
(2) Let G be a torsion-free divisible nilpotent group (i.e. P-torsion-free and
P-divisible). If u e G, t E Q, 11‘ is defined (§ 4-, Exercise 16). If u e G, v e G, we
define it + n and [11, v] by the formulae of (4') above. Show that G thus has a
nilpotent Lie nzbra structure and that the corresponding Hausdorff law is
the given group law on G (verify these assertions when G is a group Ff», cf.
(1:), and pass from this to the general case using homomorphisms of F; into G).
Let f be a mapping of G into a torsion-free divisible nilpotent group G’.
Show thatfis a (group) homomorphism if and only if it is a Lie algebra homo-
morphism. (The Hausdorfl law therefore defines an isomorphism of the “cate-
gory” of nilpotent Lie Qalgebras onto that of torsion-free divisible nilpotent
groups.)
5. Let 9 be a nilpotent Lie Q—algebra which we give the Hausdorff group
law, denoted by (x, y) H my.
(a) Show that, ifx e g, y e 9, then
203
II FREE LIE ALGEBRAS
(b) Let I) be a subset ofg. Show that b is a Lie subalgebxa (reap. an ideal) of 9
if and only if it is a saturated subgroup (resp. a saturated normal subgroup) of
the group 9. (Use the formulae of Exercise 4 (d) to pass from the group law to
the Lie algebra law.)
(a) Let I) be a subgroup of the group 3. Show that the P-saturation of b is
.1).
(11) Let b be a Lie subalgebra of 9. Show that the centralizer (resp. nor-
malizer) of b in the group 9 is the set of x e 9 such that (ad x)(b) = 0 (rap.
(ad ~)(b) C b)- '
(2) Show that the lower central series ofthe Lie algebra g coincides with that
of the group 9 and that the associated graded Lie algebra g'r(g) is the same
from the “group” point of view as from the “Lie algebra” point of view.
1T 6. Let G be a finitely generated torsion-free nilpotent group. Show that, if
n is sufficiently large, G can be embedded in the lower strict triangular group
of order 71 over Z. (Let (i, C) be a P-envelope of G, cf. § 5, Exercise 6; with its
canonical Lie algebra structure (Exercise 4), G is a finite-dimensional nil-
potent Lie Q—algebra. Apply Ado’s Theorem (Chapter I, § 7) to G and deduce
an embedding of G in a strict triangular group over Q. Pass from Qto Z by
conjugating by a suitable integral diagonal matrix.)
§7
1. Let g be a complete normed Lie algebra over K such that
| [*5 y] I < “I‘ll M
for x, y in 9. Let 6 denote the set ofx e 9 such that ][x][ < § log in For x, y in G),
h(x,y) e 0, cf. no. 2.
(a) We writef(T) = (l — a‘T)/T and g(T) = l/f(T). The seriesfcon-
verges throughout the complex plane and the series g on the open disc of
radius 21: (cf. Function: qfa Real Variable, Chapter VI, § 2, no. 3). Deduce that,
if z s 9, f(ad z) and g(ad z) are defined and that they are elements of $(g; g)
and inverses of one another.
(II) Let Dzh(x,y) be the second partial derivative of h at a point (x,y) of
G) x 9 (DWmntiable and Analytic Mani/”0141:, R, 1.6.2). It is an element of
3(9; 9). Show that
1321106 y) = 3(ad h(~.y)) °f(ady)-
(Use Exercise 3 (z) of § 6 to show this formula when x and y are sufficiently
close to zero and pass from this to the general case by analytic continuation.)
Show that the formula
f(ad Maw) ° Dsh("').'l) =f(ady)
is valid at eve_ry point of the domain of absolute convergence of the formal
power series H (cf. Proposition 1).
204
EXERCISES
§3
We suppose that the residual characteristic p of the field K is >0. We
denote by I); the ring of the valuation v of K.
1. (n) An element 1: ofK is called admisible ifu(1r) = 0 and 1:”-l E —1
(mod. wax).
(Note that 1:"1/p 5 DK since v(1:) = 9.)
Show that there exist fields K satisfying the conditions of the paragraph and
containing an admissible element (adjoin to Q, a (p — l)-th root of —p).
(b) Let 1: be an admissible element of K. Consider the following formal
power series with coefficients in K:
205
n mus): LIE means
7-1
Deduce, by means of the formula 1) = 1:! (w‘ —- 1), that 1: is adminiblz
(Exercise 1).
1] 4. Let a be an integer 2 l and let P be a set of prime numbers containing
all prime numbers $5. Let Z? = S; 1Z (§ 4, Exercise 16). Show that the terms
of degree $0 in the Hausdorfi series H(U, V) belong to LzF({IJ, V}). Deduce
TFcr more details on this exercise, see M. Lazard, Bull. 50:. Math. France,
91 (1963), pp. 435—451.
206
EXERCISE
APPENDIX
1. Let (I),l be the cyclotomic polynomial of index 1: (Algebra, Chapter V,
§ 11, no. 2). Using the formula.
x» — 1 = H (FAX),
“In
show that
(1),.(X) = 1;; (ad _ ”um.
(Apply the Mobius inversion formula to the multiplicative group of the field
(109-)
2. Let D be the total algebra of the monoid N‘ (Algebra, Chapter 111, § 2,
no. 10). Ifn E N“, let n“ denote its image in D, so that l” = land (7111:)“ = 71m
for n, m in N*.1' Every element f of D can be written uniquely as a series
(b) We write E = "2‘ n“ and y. = .121 y.(n)n“’. Show that C and y. are
inverses of one another. Deduce that, if 5 = 25(71):“ and t = 21007:" are
two elements of D, the relations .r = Li and t = y"; are equivalent (variant of
the Mobius inversion formula).
(6) Let P be the set of prime numbers. Show that the family of (l -— p“) for
p E P is multipliable in D and that
y.=Z(l—p")
per
and :JI ‘
use 1 _ 9°"
1 We often write —-.r instead of a; we then say that the elements of D are the
Dirichletfimal power rain with cocflicients in K.
207
CHAPTER III
Lie Groups
Throughout the chapter, K denotes either the valuedfield R ofreal numbers, or the valued
field C ofcomplex numbers, or a nan—discrete conialete ultrametric commutative field. We
assume that K is ofcharacteristic Ofiom § 4 onwards, that K = R or C in § 6, that K
is ultrametric in § 7. Unless otherwise ”motioned, all the manifolds, all the algebras and
all the vector spaces considered are over K. Recall that, when we speak y” a manifold of
classC’,IeNK,thatirr=mifKaéRand1<r< mifK=R.
The conventions on norms, nonnable spaces and normed spaces are those
of Diferentiable and Analytic Manyblds, R.
Recall that a normable algebra over K is a (not necasarily associative) algebra
A over K, with a topology .7 which possesses the following properties:
(I) .7 can be defined by a norm;
(2) the mapping (:4, y) )—> any of A x A into A is continuous.
The group of bicontinuous automorphisms of A is denoted by Aut(A).
Every finite-dimensional algebra over K is a normable algebra with the cano-
nical topology. A normed algebra over K is an algebra A oyer K with a norm such
that “any“ < “x“ H y“ for all x, y in A ; the algebra A with the topology defined by
this norm is a normable algebra. If A is a normable algebra, there exists a
norm on A defining its topology and making A into a normed algebra,
If G is a group, ea, or simply e, denotes the identity element of G. For
geG, ~{(g), 8(g) and Int(g) denote the mappings g’ Hgg’, g’ Hg’g‘l and
g’ Hgg’g“ of G into G. Iff is a mapping ofG into a set E,fdenotes the
mapping g»f(g") ofG into E.
209
III LI'E GROUPS
DEFINITION l. A Lie group over K is a set C with a group simmer: and an analytic
K—manifold itmoture such that these two structures are compatible.
A Lie group over R (resp. C, Q12) is called a real (resp. complex, [z-adic) Lie
group. .
Let G be a group with an analytic manifold stmcture. For g, h, g”, ha in G,
for all :5, :2, y,, y2 in W’. Writing at, = It, = y,, we obtain
211
ll! LIE GROUPS
212
uonpmsus or LIE GROUPS § 1.2
Hence E, with its additive group and analytic manifold structures, is a Lie
group.
In particular, K is a Lie group.
(2) Let A be a complete normable unital associative algebra over K. The
multiplication (x, y) r—>xy of A x A into A is bilinear and continuous and
hence analytic. Proposition 3 shows that the group A“ of invertible elements of
A is open in A (which also follows from General Topology, Chapter IX, §3,
Proposition 13) and that A* is a Lie group.
For example, let E be a complete normable space over K and let A = 3’(E)
(General Topology, Chapter IX, §3, Proposition 5). Then A* is the automor-
phism group GL(E) of E. This group therefore has canonically a Lie group structure
over K. More particularly, GL(n, K), with the manifold structure induced by
that on Mn(K), is a Lie group. For n = l, we see that the multiplicative group
K* is a Lie group with the manifold structure induced by that on K.
(3) Let G be a Lie group over K. Let K’ = R or C or a non-discrete com-
plete ultrametric field and a an isomorphism of the valued field K’ onto a
valued subfield of K. Then the group G, with the K'-manifold structure
obtained by restriction of scalars, is a Lie group over K', which is said to be
derivedfrom the Lie group G by rextriction ofsealar: (from K to K’ by means of a).
For example, every complex Lie group has canonically a. real Lie group
structure. Again, with every complex Lie group G is associated a complex Lie
group called the conjugate of G, derived from G by means of the automorphism
z >—> 2 of C.
213
In LIE GROUPS
no confusion can arise) is a morphism of the Lie group G into the Lie group
GL(E), in other words an analytic mapping 1: of G into GL(E) such that
7:(gg’) = 1:(g)1r(g’) for g, g’ in G. Suppose that E admits a. finite basis
(43,, e,, . . ., en) over K; let (cf, ta“, . . ., e:) be the dual basis; let p be a homo—
morphism of the group G into the group GL(E); then the following conditions
are equivalent:
(i) p is an analytic linear representation;
(ii) for all x e E and x’ e E’, the function g >—> (p(g)x, x’) on G is analytic;
(iii) for all 1' and j, the function g >—> <p(g)e., ef) on G is analytic. _
For the implications (i) 2 (ii) a (iii) are clear. On the other hand, the
functions u >—> (ueb cf) form a coordinate system on J(E); hence their res-
trictions to GL(E) form a coordinate system on GL(E), whence the implication
(iii) = (i).
Let G be a real Lie group, E a real complete normable space and p a
homomorphism of the group G into the group GL(E). We shall see in § 8,
Theorem 1 that, if p is continuous (when GL(E) has the topology derived
from the norm on L(E)), then 9 is analytic. But note that this notion of
‘ ‘ is "" from that ‘J ed in ' 6 ' , Chapter VIII,
§2, Definition l (ii) (Exercise 1).
(4-) Let G be a real Lie group and E a complex complete normable space. An
analytic linear represenlafian of G on E is a morphism of G into the underlying
real Lie group of GL(E).
PROPOSITION 4-. Let G and H be Lie group: andf a homomorphism 0f the group G
into the group H. Forf to be analytic, i! is necessary and :uficient that there exixl a non-
empty open subset U of G me}: thatf | U i: analytic.
The condition is obviously necessary. Suppose that it holds. For all x0 E G,
f(xox) = f(are) f(x) for all 2: EU and hence f | xoU is analytic. But the sets
xoU, where In E G, form an open covering of G.
Remark. Iff is an immersion at e (resp. a submersion at e), clearly f is an
immersion (resp. a submersion).
5. LEE SUHGROUPS
Let G be a Lie group and H a subgroup of G which is at the same time a sub—
manifold of G. Then the mapping (x, y) >—> xy" of H x H into G is analytic
and hence the mapping (:4, y) >—> xy“ of H x H into H is analytic. (Difl'eren-
liable and Analytic Manifolds, R, 5.8.5). Thus H, with the group and manifold
structures induced by those on G, is a Lie group.
DEFINITION 3. Let G be a Lie group. A subset H uf G is called a Lie rubgmup if
H is a rubgnmfi and a .mbmanifold of G.
214
max-mum vacuum or [m onoups § 1.4
2l5
III LIE GROUPS
216
QUOTlENT or A MANIFOLD m: A LIE GROUP § 1.5
kThis follows immediately from formulae (4) and (5) since T,,(8(g)) T,‘(-r(g))
and T,(y(g‘1)) are isomorphisms.
Comm“. Let x e x. IfK i: qfcharacterislic o and x irflnile—dinwnfianal, pm
i: a whimnmdon. [ffimher p(x) is injective, p(x) is an immem'an.
This follows from Proposition 9 and serenliable and Analytic Manifolds, R,
5.10.6.
Observe that, ifn denotes the mapping (3, x) H gr of G x X into X, then,
{or g E G, x s X, 14 e T,(G), v e T,‘(X),
217
to ma GROUPS
218
noncomous spams AND Qumran-r onoues § 1.6
and hence the inverse bijection old) is of class 0. Clearly (12(gg', w) = g¢(g', w)
for w e W, g e G, g' e G. Hence (X, G, X/G, 1:) is a principal left fibre
bundle.
Remark. With the above hypotheses, further let H be a manifold of class C’
and (x, h) >—> m(x, h) a mapping of class C’ of X x H into X such that
m(gx, h) = gm(x, h) for: e X,g e G, h e H. Letnbe themappingof(X/G) x H
into X/G derived from m by taking quotients. We show that n i: of class (3'.
Consider the diagram
x x H—">X
xx] x
(X/G) xH—"—->X/G
It is commutative, 1: o m is of class C' and 1: x l is a smjective submersion; it
then suffices to apply terenfiable and Analytic Mang'flzlds, R, 5.9.5.
Let G be a Lie group, X a manifold of class C' and (g, x) >—>zg a law of
right operation of class C' of G on X. Let 1:(g)x = p(x)g = xg for geG,
Jr 6 X. Then this time
219
I'll LIE GROUPS
Proposition 10 ofno. 5. The second follows from the Remark ofno. 5. Since 1: is a
submersion, the kernel of Tx(7r) is the tangent space at x to
-l
1! ("(X)) = ’60 = Y(X)(G)
and hence is obtained from T,(G) by T.(Y(x)).
Suppose that G is normal. Let m be the mapping (x, y) >—>xy" of
(X/G) x (X/G) into X/G. Then (m c (1: x 1r))(x,y) = «(z-y“) for all x, y in
X. Hence in o (1r x 1:) is analytic. As 1: x 1: is a surjective submersion, m is
analytic (Difierentioble and Analytie IbIonifoldx, R. 5.9.5), whence (ii). '
The homogeneous set X/G with the manifold structure defined in (i) is
called the quotient (left) Lie homogeneous spate of X by G. The (right) Lie homo-
geneous space G\X is defined analogously. When G is normal, the Lie group
X/G defined in (ii) is called the quotient Lie group Q/ X by G.
PROPOSITION 12. Let X be a Lie group and Y a non-empty analytic ”.11a with a
law of analytic left operation of X on Y. For ally e Y, let p(y) be the orbital mapping
by y and X” the stabilizer q in X. Thefollowing conditions are equivalent:
(i) there exist: y E Y such that 9(y) is a .rurjeetiue submersion;
(i’) for all y 6 Y, p(y) is a .rurieetive submersion;
(ii) there eximy E Y :t that X, is a Lie subgroup ofX and the canonical mapping
of XIX], into Y is a manifold isomyphiym;
(ii’) for all g e Y, X, i: a Lie subgroup of X and the canonical mapping of X/X,
into Y is a manifold isomorphism;
. (iii) the mapping (2, y) >—> (y, xy) of X x Y into Y x Y i: a rug'ectioe submer-
A10”.
220
HouooENEous SPACES AND QUO‘l'IENT GROUPS § LG
(i) Th: low of left operation (h, x) s—> hx of H on X satisfies conditions (a) and (b)
of Proposition 10 (so that we can consider the quotient ”moi/old: X/G and X/H).
(ii) The low of left operation of G on X defines on taking quotients o low If left
opemlion of [lass C" of G/H on X/H; this law satisfies condition: (a) and (b) of
Proposition 10 (so that we can considzr the quotient mong’old (X/H)/(G/H)).
(iii) The canonical mapping of X onto X/H defines on taking quotient: a bijection
of X/G onto (X/H) /(G/H). This hijation i: on isomorphism of manifold: 1y”olms 0’.
Clearly H operates freely on X; it operates properly by Gmrol Topology,
Chapter III, § 4, no. 1, Example 1. The orbital mappings of H on X are immer-
sions since the canonical injection of H into G is an immersion. This proves
(1).
The law of left operation of G on X obviously defines, on taking quotients, a
law of left operation of G/H on X/H. This law is of class C' by Diflzmitiabl: and
Analytic Manifolrit, R, 5.9.6. Letg e G and x e X be such that (Hg) (Hz) = Hr;
then H(gx) = H): and hence gx e H: and g e H; this proves that G/H operates
freely on X/H. The mapping 0: (g, x) -> (x, gx) of G x X into X X X is
closed; on the other hand, 6(Hg x Hx) = Hx x H(gx) ; it follows immediate-
ly that the mapping
(Hg, HI) H (HI: I{(90)
of (G/H) >< (X/H) into (X/H) >< (X/H) is closed; as moreover G/H operates
freely on X/H, Theorem I (c) of Central Topology, Chapter I, § 10, no. 2 proves
that G/H operates properly on X/H.
Let 1: be the canonical mapping of X onto X/H, a the canonical mapping
ofG onto G/H, :: an element ofX andy = 1r(x).
Gflx
G/HflX/H
The“ 7' ° 90‘) = PM) ° 6 and hence
Tx(1=)° T.(P(x)) = T.(P(y)) o T.(a).
Let u e TAG/H) be such that T,(p(y))u = 0. There exists 1/ E T¢(G) such that
u = T,(a-)o. Then T,(1r)(T,(p(x))o) = 0, hence T,(p(x))v is tangent to Hz
(Dyermtiable and Analytic Manifolds, R, 5.10.5) and therefore is of the form
T,(p(x) | H)u’ for some 11’ eT.(H). As T,(p(z)) is injective, it follows that
o = o’, whence v E T,(H) and therefore 14 = 0. Thus T.(p(y)) is injective. The
image of T,(p(y)) is equal to that of T,(7r) oT,(p(x)); now the image of
Te(p(x)) admits a topological supplement in T,(X) and contains the kernel of
Tx(1r). It is therefore seen that p(y) is an immersion, which completes the proof
of (ii).
221
III LIE GROUPS
Assertion (iii) follows from the above and Diflerentiable and Analytic Alangflalds,
R, 5.9.7.
COROLLARY. Let G be a Lie group and H and L normal Lie subgroups of G with
L C H. Then H/L is a normal Lie subgroup (if G/L and the canonical bijection of
G/H onto (G/L)/(H/L) is a Lie group isomorphism.
7. ORBITS
PROPosmON 14. let G be a Lie group, X an analytic manifold and (g, x) t—> gx a,
law of analytic left operation of G on X. let x E X. Suppose that the corresponding
orbital mapping p(x) it a subimmersian (which is always the case if K is ofcharacteristic
0 and X irfinite-dimmsional (Corollary to Proposition 9)). Let G,‘ be the stabilizer
o x in G.
f(i) G, is a Lie subgroup and T,(Gx = Ker T,(p(x)).
(ii) The canonical mapping ix ofthe homogeneous space G/Gx into X is an immersion
with image Git.
(iii) {ffilrther the orbit Ga is locally closed and the topology on G admits a countable
base, then Gx is a submanifold of X, ix is an isomorphism of the manifold G/G, onto
the manifold Oz and Tx(Gx) = Im T,(p(x)).
The inverse image of at under p(x) is G”. As p(x) is a subimmersion, G; is a
submanifold and, for all g5 G, the tangent space J to n = p(;,’(gx) at g
is Ker T,(p(x)) (Difierentiable and Analytic Manifolds, R, 5.10.5), whence (i).
Let 1:: G —> G/G, be the canonical mapping. Then i, o 1: = 90:). As G/G, is a
quotient manifold of G, this equality proves that iJr is analytic. Further, the
kernels of T,(p(x)) and Tg(1r) are both equal to J. Hence Tn(,)(i,) is injective.
The image of Tm,(i,‘) is equal to the image of T,(p(x)) and hence admits a
topological supplement. This proves (ii).
Suppose that Gx is locally closed. Every point of Gx then has a neigh-
bourhood in G1: which is homeomorphic to a closed subspace of a complete
metric space and hence is a Baire space. Hence Gx is a Baire space (General
Topology, Chapter IX, '9' 5, Proposition 4). IfG has a countable base, ix is there-
fore a homeomorphism of G/Gx onto 0:: (General Topology, Chapter IX, §5).
Then by (ii) and Diflerentiable and Analytic Manifolds, R, 5.8.3, ix is an isomor-
phism of the manifolds G/G,‘ onto the manifold Gx and
COROLLARY. Let G be a Lie group whose topology admits a countable base and X a
222
vecron BUNDLES wrm OPERATORS § 1.8
223
m LIE (mom
Examples. With the hypotheses and notation of Proposition 16, let F be a com-
plete normable space. Then .‘Z’((TX)’; F) is a vector G~bundle of class C"‘; so
is AIt"(TX; F) if K is of characteristic zero or X is finite-dimensional (cf.
Dzjfirmtiable and Analytic Mamfilds, R, 7.7, 7.8). if X is finite-dimensional,
(8)” (TX) ® (8)" (TX)* is a vector G-bundle of class 0‘1.
PROPOSITION 17. Let G be a Lie group, X a left Lie homogeneous space of G, are a
point of X, Go the stabilizer of x0 in G, E and 13' left vector Gbundlex of class Cr
and have space X, E, (rwp. E3) the fibre of E (resp. E’) at x0 andf an element of
$(Eo, E’D) such thatf(gu) = gf(u) for all a e E0 and g e G“. Then there exist: one
and only one morphism of E into E' compatible with the operation: of G and extendingf.
The uniqueness of this morphism is obvious. We prove its existence. Let g,
224
- VECTOR BUNDLES WITH OPERATORS § 1.8
225
III LU! GROUPS
226
LOCAL DEFINITION or A LIE GROUP § 1.9
HxH—">H
where m(x,y) = zy" (reap. n(x,y) = erg/'1) for x, y in G (resp. H). Then
71 e (f x f) is analytic, hence f o m is analytic and hence on is analytic since
f is an immersion. Therefore G is a Lie group.
The Lie group structure on G is allied the inverse image of the Lie group
structure on H underf.
COROLLARY. Let G be a topological group, N a discrete normal subgroup of G and
7: the canonical mapping of G onto G/N. Suppose that an analytic manifold structure
is given on G/N, compatible with the topolog'ual group structure on G/N. Then there
exists a unique manifold structure on G such that 1: is an immersion; this structure is
the inverse image under 1: {If the manifold structure on G/N. With this structure, 1: is
Itale, G is a Lie group and G/N is the quotient Lie group ofG by N.
Remark. Let H be a connected real or com lex Lie group, H its universal
oovering‘l‘ and re the canonical mapping of onto H. When we speak of H
TCf. General Topology, Chapter XI; whilst awaiting the publication of this
chapter, see for example L. S. Pontrjagin, Topological groups, 2nd edition trans-
lated from Russian, Gordon and Breach, 1966; or G. Hochschild, The structure of
Lie groups, Holden-Day, l965.
227
III LIE GROUPS
of
as a Lie group, we shall always mean with the inverse image structure
that on H under 1:.
10. GROUP GERMS
DEFINITION 5. A Lie group germ aver K is a system (G, e, 0, m) ratirfi/ing the
following conditions :
(i) G is an analyltk manifold over K;
(ii) e e G;
(iii) 6 Lr An analytic mapping if G into G;
(iv) m is an analytic mapping 0f an open subset Q q x G into G;
(V) fl" allge G, (M) e Q, (g, t) E 9, "1(2. 3) = “(3, I) = g;
(Vi) fordllge G, (g, 9(9) E 9’ (9(3),.g) G 9,"!(5', 9(3)) = m(°(g):£) = 6;
(vii) ifg, h, k are elements ofG such that (g, h) e (1, (h, k) e (2, (m(g, h), k) E Q,
(g: ”‘01: ’0) E 9, the“ m(m(£a ll): ’5) = "1(g, "‘0'; k)).
a is called the identity element of the group germ. We often write gh
instead of m(g, h) and (by an abuse of notation) g'l instead of 0(g).
A Lie group G is a Lie group gem with the obvious choice of e, 0, m.
Let G be a Lie group germ. Then :3" = I, that is
(8) e" = a.
For all g e G,
g = cg = ((g")"g'1)g = (g“)“(g‘ g) = (3")‘16,
that is
(9) (g'1)‘1 = g~
A subset of G invariant under the mapping g »—> g“ is called symmetric.
The manifold G, with the point e, the mapping g» g" and the mapping
(g, h) »—> hg is a Lie Group germ GV called the opposite of G.
The Lie group germ G is called commutative if, for all (g, It) 5G >< G
such that gh is defined, hg is defined and equal to git.
Let G be a Lie group germ. The set of (g,h) eG x G such that g}: is
defined is a neighbourhood of (e, 2). On the other hand, the mappings
(g, It) &—> gh and g ~—> g'1 are continuous. Hence (gh)k = g(hk) for g, h, k
sufficiently close to 2. Similarly, (h‘lg'l) (gh) = h"(eh) = h'1 h = t for g, h
sufiiciently close to c, whence multiplying on the right by (gh)",
(10) (3h) ‘ ‘ = It‘lg“ for g, h sufficiently close to e.
Pnorosrnou 20. Le! G b: a Lie group germ and g e G. Time zxist an open neigh-
bourhood U aft and an alien neighbourhwd V q with thefollowing properties:
(21) ug ir definedfor all u E U;
(h) vg'1 is defined/0r all u e V;
(c) the mappiflgi u H) 14.2. I! I~> 1g" 1 an inverse amtlylit isomorphisms of one
analher of U ml!) V and 0/ V unto U.
228
GROUP GERMS § 1.10
229
m LIE GROUPS
230
LAW CHUNKS or OPERATION § 1.1!
open neighbourhood V' of e in V such that the mapping 0 v—> all is an analytic
isomorphism of V’ onto an open neighbourhood of u in U (Proposition 20)
and at the same time onto an open neighbourhood of u in N. Hence U is
open in N and the identity mapping of U is an isomorphism with the given
manifold structure on U and the open submanifold structure on N ; in other
words, U is an open submanifold of N.
Finally, we consider an analytic manifold structure on N with properties
(i) and (ii) of the proposition and let N“ be the Lie group thus obtained. Then
the identity mapping of N into N* is étale at e and hence a Lie group iso-
morphism. This proves the uniqueness assertion of the proposition.
231
lIl LIE GROUPS
ls open in
that the set (1’ of (g, x) e Q n (G' x X’) such that 44g, 2:) EX’
.X)..Then
G’ x X’ (a condition which is always fulfilled if X’ is open in
is said
449’ is a law chunk of left operation of class C" of G' on X , Wth
to be derived from t]; by restriction to G’ and X’.
(3',
Pkoposmou 23. Let (G, e, 0, m) be a Lie group germ, X a manifold of clot:
x0 a point of X, 0 an open neighbourhood of (e, no) in G x X and q, a mapping of
9 into X with thefollowing properties:
(i) LIJ is ofolasx C';
(ii) Me, at) is equal to zfor x mfieiently close to x0;
(iii) ¢(m<g. g'). x) = Mg. Mg’, ow to g', x) mommy close to (e, ., n)-
Then there exist on open neighbourhood X’ of no in X and on open subset Q’ of
Q n (G X X') such that 449’ i: a low thunk qflefl operation ufclass C’ ofG
on X’.
There exist an open neighbourhood X’ of xn in X and an open neighbour-
hood G' of: in G such that Me, x) = x for all x e X, and
232
TANGENT uws or composmorz § 2.1
PROPOSITION 24. Let G be a Lie group gem, X a manifold of elm: C’ and (dx
a locallyfinite open covering of X. For all {9 I, let vb, be a law chard: of left operation
of clay: C’ g" G on X. Suppose that the underlying topological :pace If X is normal
and thal,for all (i,j) E I X I and all x EX, r\ X,, (It; and 4;, coincide on a neigh-
bourhood of (e, x). There exist: a law chunk 4/ of left operation of clay: C' of G on X
Inch that, for all i E I and all x e X‘, 4;. and q; coincide on a neighbourhood of (e, x).
For all (i,j) 61 x I and all xeX1 n X, choose an open neighbourhood
Vl,(x) of x in X, n X, such that 4;, and vb, are defined and equal on a neigh-
bourhood of {e} x V.,(x) in G x X. For all x e X choose an open neighbour-
hood V(x) of x in X such that conditions (a) and (b) of Lemma 3 are ful-
filled. Let I, be the set of i e I such that XEX‘. This is a finite set. Let U,‘
be the set of (g,y) SO x V(x) such that the q“ for £51,: are defined and
coincide on a neighbourhood of (g,y). Then U, is open and (47:) EU,”
The 41, for ie 1,, all have the same restriction to Ux. Let x, y be in X. If U”
and U” meet, V(x) and V(y) meet and hence there exists is I such that
233
m LIE (mom
236
CASE OF GROUP GERMS §2.3
237
III LIE GROUPS
This follows from Dyfmntinble and Analytic Manifolds, R, 13.4.1, 13.2.3 and
13.4.4.
Let .7‘“”(G) be the direct sum of the T?”(G) for g E G (cf. Difi’mntiable
and Analytic Manifoldr, R, 13.6.1). We define the convolution product in
31“”(G) as the bilinear mapping of .7‘”’(G) x 3"”)(G) into .7'“’(G)
extending the convolution product of Definition 1. We also denote it by a.
Thus .7“’(G) has an algebra structure filtered by the .7‘”(G). The subalgebra
.7‘°’(G) = 9?“ T‘,°)(G) is identified with the algebra K“) of the group G
over K.
Pkorosmou 2. The algebra 3"”)(G) is madaliw It is commutative if and only
If G is cumulative.
Let tEY“’(G), t’ 6.7““(G), !' 6.7“”(G). Then 1*(1’ t I”) is the image
of t®t’ ® I” under the mapping (g,g’,g”) Hg(g'g”) of G x G x G
into G and (t a: t’) t t” is the image of t ® I' ® t" under the mapping
(g,g’,g") v—> (gg’)g" ofG x G x G into G. Hence (t*t') t t' = H (1’ Ir 1").
It is seen similarly that, if G is commutative, H t' = t’ no: t. If the convolution
product is commutative, G is commutative by Proposition 1 (iii)
PROPOSITION 3. If! E .7“’(G) andg e G, thenflg)“ = e, at t, 8(g)*t = H 9,—1,
(Int g)“ = s, a t m 5,-1. In parlicular, c. i: the unit element f‘”’(G).
Consider the diagram
G 34> G x G L» G
where (I) is the mapping 1: v—> (g, h) and 44 is the mapping (h', h) »—> h'h. Then
Y(£) = 4! ° ‘1 and hence fig):t = 41.0140) But 4>*(').= =. 8 I and hcnce
¢.(¢*(t)) = c, t- t. The argument is similar for 8(g)*t. Fmally,
Intg = Hg) ° 8(.3)
and hence (1m .9). = fig): ° Mg).-
It is therefore seen that, for teT(G), e, t l and It s, are equal to g!
and 3g calculated in the group T(G) (§ 2, no. 2). But it should be noted
that, for t, l’ in T(G), the product It’ in the sense of § 2 is in general dif-
ferent from t n t'.
239
m LIE GROUPS
af the
DEFINITION 2. Let G be a Lie group. TIIL‘ .mbalgehra nffl"“"(G) consisting
distributions with support contained in a is denoted by U(G).
This algebra is filtered by the subspaces
U.(G) = U(G) n y"’(G) = Tl"(G)-
We write U+(G) = T‘,"”(G), U: (G) = U*(G) n U,(G) (cf. Difi'mntiable
and Analytic Manififlds, R, 13.2.1). Recall that Uo(G) is identified with K and
U: (G) with the tangent space T,(G), In U(G), U" (G) is a two-sided ideal -
supplementary to U0(G).
Example. Let E be a complete normable space considered as a Lie group.
Then the vector space U(E) is canonically identified with the vector space
TS(E) (Difimntiable and Analytic Manifolds, R, 13.2.4). Let m: E x E—>E
be addition on E. Then
1a,: TS(E x E) —>TS(E)
is equal to TS(m) (tnzntiable and Analytic Mani/blah, R, 13.2.4). For t, t’ in
U(E) = TS(E), the image t a t’ of the symmetric tensor product t ® t' under
m... is therefore TS (m) (t ® t'). By Algebra, Chapter IV, §5, no. 6, Proposition
7, this image is just the product tt’ in the algebra TS(E). Thus the algebra
U(E) is identified with the algebra TS(E).
PROPOSITION 4. Consider the bilinear mapping (:4, u) D—> a It 1/ (resp. (u, v) ~—> v a u)
of U(G) ® K“) into 9"”)(G). The corresponding linear mapping qf U(G) ® K‘“’
into 9"”)(G) is a 1mm spa“ ismrphiwn.
K“) is the direct sum of the K5,, for x e G. On the other hand, the mapping
u l—> a It a, (resp. u >—> c, e u) is an isomorphism of the vector space
U(G) = fl”7(G) onto the vector space 9:”)(0) by Proposition 3. Finally,
.7"°‘”(G) is the direct sum of the T§”)(G) for g E G.
Let X be a manifold of class C' (r 2 co) and xEX. We have defined
(Difmntiahle and Analytic Mamfilds, R, 13.3.1) a canonical filtration on the
vector space 9';°°’(X) and a canonical isomorphism ix”, of the associated
graded vector space onto the graded vector space TS(T,‘(X)). In particular,
let T,(G) = L; then i9”, is an isomorphism of the graded vector space gr U(G)
onto the graded vector space TS(L). But U(G) is a filtered algebra, from which
we obtain a graded algebra structure on gr U(G).
PROPOSITION 5. The isomorphism £5.92 gr U(G) —» TS(L) is an algebra iso-
morphism.
Let p be the mapping (t, t') >—>t® t’ of U(G) x U(G) into U(G x G).
Let c be the mapping (1, t') »—> t It I’ of U(G) x U(G) into U(G). Let m be
the mapping (g, g') H gg' of G x G into G. Then by Definition 1
(1) c = mop.
240
FUNCTORIAL “cram-ms § 3.2
2. FUNCTORIAL PROPERTIES
PROPOSITION 6. Let G, H b: Lie groups and 115 a morphism (If G into H. Fur t, t’
in 9"”(0): 4M! * V) = 4M!) * 4%")-
Consider the diagram
G x G —”‘> G
«no 0
H x H —"> H
where m(g, g') = g', 710:, h’) = hh’. This diagram is commutative. Hence
241
m 1m GROUPS
' P170:
(G1 x GI) x (Ga x Ga)
where :(g, g’) = (g', g). m(g, g’) = gg’, "(.43. g') = g’g for all g, g’ in G. This
diagram is commutative. Hence n.(! ® t') = m.(:*(t ® t')) = m.(t’ ® t).
This equality is precisely (i). Assertion (ii) follows fi‘om (i) and Proposition 6.
Pnoposmon 8. let G, H be Liegraupx andtb anwrphirm ofG in!» H. [ft 5 .7"“”(G),
thenLetMr) = (MW.
0 (resp. 0’) be the mapping gag—1 of G into C (rap. of H into H).
Then 4: ., e = e' . <1», whence MM!» = sum»-
PROPOSITION 9. Let G”- - -, G,I be Lie group: and G = GI x - - - x G". [fthe
vector spam 51"”(G) and 9"”)(G1) ® - ~ ~ ®9""”(G,,) are canonically identi-
fied, the algebraY‘“)(G) ixthetmmproduc!qfthealgebrax.7‘“’(G,), . . . . 3"”)(Gn).
{flief‘wKGwi = 1,. . .,n, then
(’1® ®tn V = li® @11-
It suffices m consider the case n = 2. Let I], I; be in 9"”)(Gl), I3, I; in
9"“)(69. We need to show that (11 ® :2) a: (t; ® 1;) = (t, : 1;) ® (ta I t,’.)
and that (11 ® 19V = l} ® 1;. Consider the diagram
G x G —‘> G x G
242
FUNUI‘ORJAL vacuums § 3.2
243
in ma GROUPS
Let G be a. Lie group. The restrictions of the convolution and the coproduct
to U(G) define a. bigebra structure on U(G). We have U(G)" = U(G).- If
(h: G—>H is a Lie group morphism, we denote by .U(¢) the mapping
tI—>¢,‘(t) of U(G) into U(H); this is a bigebra morphism. If 4;: H—>L 15
another Lie group morphism, then U(tl; a (b) = U(IJI) a U(o). If 4) 15 an im-
mexsion (resp. a submersion), U(rlz) is injective (resp. surjectwe) by Dz mn-
tiable and Analytic Manifolds, R, 13.2.3. In particular, if H is a Lie subgroup
of G, U(H) is identified with a subalgebra of U(G), the coproduct on U(H)
being the restriction of the coproduct on U(G). If H is open in G, then
U(H) = U(G). If G1, G1 are Lie groups, U(G1 x 0,) is identified with
U(G,) >< U(Gz). The primitive elements of U(G) are those of T,(G) (Dif-
fermtiable and Analytic Manifialds, R, 13.5.3).
Again let (I): G—>H be a Lie group morphism. If gr U(G) is identified
TS(T.(G)) and grU(H) with T$(T.(H)), then grU(¢) is identified with
TS(T,(¢)) (Difirmtiable and Analytic Mang'fizlds, R, 13.3.5). We apply this to
the isomorphism g» g" of G onto G"; then T,(¢) = —l and hence
244
convournou or pom-r ms-numn-Ions AND wucnons § 3.4
In particular, ifu eT(X) and IE T(G), e,” nu u and M em are equal to the
products gun and txo defined'In §2, no. 2.
PROPOSITION 15. Let G (resp. G’) be a Lie group and X (resp. X') a maniald of
clan C'. Suppox: that a law 12/ lefl 017mm,; qfclm: (3' cf G (rap. G’) on X (resp.
X’) is given. Let 21> be a morphixm If G into G' and (ll a tb—mmphism Qf X into X’.
Le! IEF‘WG), u eT‘"’(X) be such that: + s’ g 7. Then
‘Mt * U) = (Mt) * 4401)-
This can be proved as is Proposition 6 of no. 2.
Remark. Let f be a law of fight operation of class C' of G on X. 1ft 6 3"”(G)
and 1459'“”(X), with 3 + s’ < r, we denote by 11*! the image of u®t
underf,. Propositions l3, 14-, 15 go over to this situation in an obvious way.
PROPOSITION 16. Let G, G' be Lie groupx, X a mamfald 9)“ class C' and supper:
that G (resp. G’) operate: an X on Ike left (resp. right), with (gx)g’ = g(xg') fur
all xeX, geG, g’eG’. Let tefl'"’(G), I'ei"”(G’), t'eT‘"’(X), with
:+:’+:"<r. Then
(tu') u’ = 1* (t'u‘).
(t at t") t t’ (reap. H (t' H’)) is the image oft ® I” ® 1' under the map-
ping ( g, x, g’) »—> (gx)g’ (resp. g(xg’)) of G x X x G' into X.
245
m LIE GROUPS
248
mvnmm‘ 12mm OF POINT DISTRIBUTIONS § 3.6
249
III LIE GROUPS
if
In other words, a field of distributions g »—> A, on G is lefi invariant
A”, = Y(g)*l5n’ forg, g’ in G,
agam' if
or A", = e, e A, for g, g' in C.
It is right invariant if
Aw = 3(£"‘)*Ag forg,g'in G,
or A,,» = A, s em for g, g’ in G.
DEFINITION 5. Let G b: a Lie group and l e U(G). Let L, denote thefield of dirm'baL
lions-gas,ttanGandR,tfieldq/dirhibutiomgv—>t*e,MG.
In other words, L, (mp. R,) is the field of distributions defined by i and
G operating on G on the right (rap. left) by means of the mapping
(g, g’) >—> gg'. Let Q be an open subset of G and F a Hausdorff polynomed
space; iffe Wm, F), then L,f=f* t" s «”(9, F) and
R,f= 2V *fe@”(Q,F)
(no. 5). If G is finite-dimensional, the differential operators L, and R, are of
class C" (no. 5).
PRoposmON 23. (i) 7714 mapping tv—>L, (reap. I»R,) is an isomorphism qf
the vector rpacs U(G) onto the vector field qf lefl (resp. right) inunn'anl dimibuiim
an G
(ii) For t, t’ in U(G), L,.,, = L,o Ly, R, ,r = R, a R,, L, oRr = R,’ a L,
(with the abuse of notation firm. 5).
(iii) Ife is the mappinggHg‘l afG onto G, then e(L,) = Ry
(iv) 1ft sU<G> mtg e G, the» (L0. = (Rant-1)..
In G every right translation commutes with every left translation. By Proposi-
tion 21 of no. 5, L, is therefore left invariant, As (L,), = t, the mapping t>—> L,
is injective. Let A be a field of left invariant distributions on G; let 2 = A,;
then A and L, have the same value at e and are left invariant and hence
A = L,. This proves (i) for L, and the argument is similar for R,. The formulae
L,_,. = L,o Ly, RM = R, 0R, follow from (21) and (18). Let tsU,(G),
t' e U,,(G),fE W“), F), where Q is open in G and .t + .t' S r; then
mw= w m = (1” sf) u
=tW a(f*l") (PropositionQO)
= R1’L,f
and hence L, o Ry = Ru 0 L,. As 0 is an isomorphism of G onto GV, 0(L,)
is a field of right invariant distributions on G; its value at a is 0‘(t) = (V;
hence 9(L,) = R,.. Finally,
(L1); = en *t = (an * t * Ea") ‘ a, = (Rink-Iran-
250
LIE Amen“. or A LIE snoop §3.7
(be) H (11,), = 1* E:
of U,(G) X G into T‘"(G) is an isomorphism of analytic vector bundles; for
this mapping is bijective, linear on each fibre and analytic (no, 5); on the
other hand, let (I): T“’(G) —> U,(G) x G be the inverse bijection; if t e Tf,” (G),
then (Ht) = (t*e,—1,g) and hence (I) is analytic. The isomorphism 4) is
called the right trivialization of T‘"(G). Similarly, consider the mapping
(t, g) ~—> (L,), = ey *t of U'(G) x G into T‘"(G); the inverse isomorphism
is called the left trivialization of T‘”(G). By restriction we recover the right
and left trivializations ofT(G) (§ 2, no. 2).
251
"I LIE GROUPS
be the expansion Qf m at an integral series about (0, 0)? where ml: , ts a bxhomogeneou:
continuum-polynomial of bidegree (i, j) on E x E math valuex m E.
(i) m“D = “0,1 = Oforalliaé l and] #1.
(ii) ml'a(a, b) = a and mo,1(a, b) = bfar all a GE, b EE. .
(iii) let 4;: T.(G) —> E be the deflerential qfd) at e. For all a, I) m T,(G),
252
1.12 ALGEBRA or A us. GROUP § 3.7
253
II! LIE GROUPS
X X) is
Manifaln's, R, 8.1.2). The induced Vector bundle (T(G) X T(X))[({e)
identified with E = L(G) x T(X). Hence T(¢)|E is a vector bundle morph-
ism of class 0"]. For (a, x) e L(G) x X, T(¢)(a, x) = Dn(x), whence (i). .
The formula giving (D,f) (it) follows from §2, the end of no. 2, and Dif-
fnmtiable and Analytic Manifolds, R, 8.4.5. .
Suppose that r 2 2. Let a, b be in L(G) and f be a function of class C'
on an open subset of X. Then
Dimalf = Dh(Daf) — Du(Dbf) by (17)
= [D,, D,]f (Dif {‘9‘ Anal. Mam, R, 8.5.3).
Let I: 5 X. By takingf to be a chart on an open neighbourhood of x, it follows
that Dlnlbfix) = [DM D3] (1:), whence (iii). The argument is similar if G
operates on X on the right.
When 7 2 2, the mapping a H D, is called the law of infinitesimal opera-
tion associated with the given law of operation.
8. FUNCTORIAL PROPERTIES OF THE LIE ALGEBRA
Let G and H be Lie groups and (I; a morphism of G into H. The restriction
of U(<]>) to L(G), which is just T,(<]>), is a continuous morphism of L(G) into
L(H), which we denote by L(tb). If 4; is a morphism of H into a. Lie group,
then LOP ° (i) = M4!) ° L(tb)‘
For (I) to be an immersion, it is necessary and suflicient that L(o) be an
isomorphism of L(G) onto a subalgebra of L(H) admitting a topological
supplement. In particular, if G is a Lie subgroup of H and 4’ is the canonical
injection, L(G) is identified with a Lie subalgebra of L(H) by means of
L((lz). More particularly, if G is an open subgroup of H, L(G) = L(H).
If G is a Lie quasi-subgroup of H, L(G) is also identified with a closed Lie
subgroup of L(H).
For <15 to be a submersion, it is necessary and suflicient that L(o) be sur-
jecfive and that its kernel admit a topological supplement. In that case, the
kernel N of cl) is a Lie subgroup of G and L(N) = Ker L(o). In particular,
if H is the quotient Lie group of G by a normal Lie subgroup P, L(P) is an
ideal of L(G) and, if d) is the canonical surjection of G onto H, L(G/P) is
identified with L(G)/L(P) by means of the morphism derived from L(¢)
when passing to the quotient.
Let I be a finite set, (0.)“ a family of Lie groups, G their product and p.
the canonical morphism of G onto G‘. Then (L(p,))m is a morphism of the
Lie algebra L(G) into the Lie algebra El L(G‘) and is an isomorphism of
normable spaces. L(G) is therefore identified with EilL(G.) by means of
(L(Nhsp
254
FUNL'I‘ORIAL PROPERTIES OF THE LEE ALGEBRA § 3.8
PROPOSITION 28. Let G and H be Lie groups and <1; a morphism of G into H. Sup-
pose that K is qfeharaeteristie 0 and that H isfinite—dimensional.
(i) The kernel N oftb is a Lie subgroup ofG andL(N) = Ker L(zb).
. (ii) The morphism 4: of G/N into H derivedfrom ‘l’ when passing to the quotient
is an immersion.
(iii) If MG) is elosed in H and the topology of G has a countable base, MG) is
a Lie subgroup of H, vb is an isomorphism If the Lie group G/N onto the Lie group
<1>(G) andL(<l>(G)) = Im L(o).
Let G operate on H on the left by the mapping (g, h) I—> ¢(g)h. It suffices
to apply Proposition 14 of § 1, no. 7, to the orbit of e.
PROPOSITION 29. Let G and H be Lie groups and ti) a morphism (f G into H. Sup-
pose that K is of tharaeteristio 0 and that H itfinite—dirnemional. If H’ is a Lie sub-
group ofH, then G’ = q; "1(H') is a Lie subgroup q andL(G’) = L(o) '1(L(H')).
Let 1: be the canonical mapping ofH into the homogeneous space X = H(H’.
Let G operate on X on the left by the mapping (g, x) I» ¢(g)x. The stabilizer
oi'1:(e) is G’, which is therefore a Lie subgroup of G (§ 1, no. 7, Proposition
14). The orbital mapping of1r(e) is 7: o (I). By Proposition 14 of § 1, L(G’) is
the kernel of L(r: up) = T,(n) oL(<)>). The kernel of T,(n) is L(H') (§ 1,
not 6, Proposition 11 (i)) and hence Ker L(n o (b) = L(¢) “(L(H’)).
COROLLARY 1. Let G, H be Lie groups and 4); and (I), morphisms ofG into H.
Suppose that K is of characteristic 0 and that H is/inite-dimensional. The set ofg e G
such that ¢1(g) = ¢,(g) is a Lie subgroup G' ofG and L(G') is the set ofx e L(G)
such that L(¢,)x = L(¢,)x.
We write (Mg) = (¢,(g), ¢2(g)) for all geG, so that (I; is a morphism of
G into H x H. Let A be the diagonal subgroup ofH x H. Then G’ = ¢'1(A)
and L(¢)x = (L(¢1)x, L(¢2)x) for all x EL(G). It now suffices to apply
Proposition 29.
COROLLARY 2. Let G be a finite-dimensional Lie group and G, and G1 two Lie
subgroups efG. Suppose that K is qfoharaeteristio 0. Then G1 n G2 is a Lie subgroup
of G with Lie algebra L(Gl) n L(G,).
We apply Proposition 29 to the canonical injection of G1 into G and the
subgroup Ga.
COROLLARY 3. Let G, G’, H be Lie groups and (I): G—>H and o’: G’ ~>H Lie
group morphism. Suppose that K is Ifeharacteristio 0 and that H isfinite-dimensional.
Let F be the m qf(g,g') 50 x G’ such that (Hg) = (my). Then F is a Lie
subgroup of G x G' and L(F) is the set of (x,x') EL(G) X L(G') such that
L(¢)* = L(o')1'-
We apply Corollary 1 to the morphisms (g. e’) -> Mg) and (g, s’) H Ms’)
of G X G’ into H.
PROPOSITION 30. Let G be afinite—dirnem‘ionol Lie group with a countable base and
255
III LIB GROUPS
H and H' Li: subgroups hf G. Suppose that K it qfcllamcterim'c 0 and that HH' is
locally closed in G. I
(i) HH’ is a mbmmy'ald q/G andT,(HH') =. L(H) + L(H ). I I
(ii) Suppuxe that every element of H commutes with every element af H .I Then HH
1': a Lie subgroup 13/ G. Let 'i> he the mapping (h, h') r—>hh’ if H x H onto ‘HH .
The kernel of d; is the set if (m, m“) where m 6-H n H’ and the morphum af
(H X H’)/Ker 4) onto HH’ derivedfrom (I) by parxmg to the quotient u a Lie group
ismnorphirm.
Let H x H’ operate on G on the right by the mapping ( (h, h’), g) r—> hgh’ ’ 1.
The orbital mapping 9 of e is (h, h’) >—>hh"‘. By Proposition 14 (iii) of § 1,
no. 7, HH’ is a submanifold of G and T3(HH’) = Im T.(p). Now
256
LIE ALGEBRA § 3.9
258
LINEAR REPRESENTATIONS § 3.11
Let S (resp. S’) be the set ofy EA such thaty = y‘ (rap. y = —y‘). Then
S, S’ are closed vector subspaces of A. The formula
259
III LIE GROUPS
260
LINEAR REPRESENTATIONS § 3.11
(i) The mbrzprenntatiun 1:1 and th: quotient iqzrexznlation 1r, qf 1: defined by F are
analytic repm'mtatim.
(ii) F 1.: xtabl: under L(1r) (L(G)).
(iii) Let 91 and Pg b: the mbnpresentation and quotient representation qf L(1:)
dzfined by F. Than L(-n:,) = p], L(1:3)=
Let A be the set ofu e $(E) such that:2u(F) C F. Then A Is a closed vector
subspace of Y (E) and 1: take: us valua m A. By virtue of the hypotheses on K
and F, the mapping 3’: G —> A with the same graph as 1: is analytic (Dijfiren-
tiable and Analytin Manifolds, R, 5.8.5). The canonical mappings 91: A —> $(F)
and 62: A —> $(E/F) are continuous and linar and hence analytic. This
proves (i). The mappings T,(1r) and T,(1I:’) have the same graph and hence
L(1r)(L(G)) C A, which proves (ii). We have
TAM) = T49; W') = 91 °Ta("') = 91
Tim) = me. «- n') = 92 . .(n') = a.
PROPOSITION 41. Let G be a Lie group and 1:1, 1:2, . . ., 1:,” 7: analytic linear repre-
xmtatian cf G on complete mmnable spate: E1, E2, . . ., E,” E. Let
(x,, x,, . . ., x") HxIxz. . .x,
In a continuous mallilimar mapping @e1 x E, x ' - - x E" into E. Suppose that
"(5)0‘1’72- - '5‘») = ("1(3)”1)(1'2(g)”2) - - - ("n(£)‘n)
fatallgeG,x1eE,,...,xneE. Then
COROLLARY 3. Le! G be a Lie group and 1:1, . . ., 1r,I analytie linear representatians
262
LINEAR upmm‘emous § 3.11
E1 ® ‘ ‘ ‘ ® En-
By Proposition 41, we see that
263
Ill LIE GROUPS
—(d—"i<w(m»§o)) AI!)
= —(d%(n°¥)(l)) x-o a.
= —L(1r)a.Eo.
2G4
ADJOINT REPRESENTATION § 3.12
Pnoposmou 43. Let r, u be in y<°°)(G). Let 1-2) I‘ Q) t,’ be the image 0/ t under the
mpmduot. Then
n
tTu=i-21I,*utt(v.
By definition, t T u is the image of! 8) u under the mapping (g, g') .—> gg’g-i
of G x G into G. Now this mapping is obtained by composing the following
mappings:
a: (in!) » (g,g,g') ofG x (:1a x G x G
B: (g,g',g')->(g,g"1,g’) ofG x G x GintoG x G x G
Y: (“3g”) »—>gg’g’ ofG x G x G into G.
On the other hand:
l5#(‘2:1h®ti®u)=‘;h®tiv
Ye(‘:1’n®ti ®“)=‘:1h*“*tiv-u
COROLLARY]. Let u EL(G) andu’ ei‘"(G). Then a T u’ = u e u’ —- u’ n: u.
The image of 14 under the coproduet is u ® a, + 2, ® 14, whence
uTu' =Il*u'*€, +e,*u’*u" =usu’ — u’tu.
COROLLARY 2. Let tef‘mw) and geG. Then a, T I = ‘a*”%"‘ If
teL(G), then s, T! = glg'1 (when the latter product is evaluated in the group
T(G)).
The image of a, under the coproduct is 2, ® 5,.
COROLLARY 3. Le! a e L(G). The vector field defined by a and the lefl operation
gHIntgq onGisthefieldR, — L“.
The value of this field at g is
a Ta; = «*z, — even (Corollary 1)
: (Rn); — (L1,)? (Definition 5).
265
Ill LIE GROUPS
PROPOSITION 46. Let hsG, xsTn(G) and aeL(G). Let d) b: the mapping
(g,g’) Hgg’g‘l afG x G into C. The imagey qf(a, x) eT.(G) x TAG) under
T(-.n)(¢) uy = x + h((Adh“)a — a).
We have
y “(c.h)¢)(a®3n + 9c®")
=aTsn+eeTx
=ateh—znta+x
= h((Ad h“)a) — ha + x.
266
ADJOINT REPRESENTATION § 3.12
PROPOSITION 47. Let G be a Lie group, H and E Lie mbgt’aupx g"G and mppoie that
hEh‘1 = Efor all h EH. Then 3"”(H) T 3"”’(E) C 3"”)(E). In partieular,
Ad(H>(L(E)) c L(E) andIL<H),L(E)1 c L(E).
Iite 9"”)(H) and t' 6.7“”)(E), then t 87 t' 69"“‘(H x E) and the image
of H x E under the mapping (g, g’) l—> gg’g“ is contained in E.
PROPOSITION 48. Let G be a Lie group and H and E Lie .mbgroup: ifG Suppose that
G 1.3a: a Lie group, the term-direct produet ofH by E Let p be the linear representation
g r—> (Ad g) | L(E) of the Lu group G on L(E) (of Proposition 47) and let a be the
rextrietion of p to H. Then:
(i) L(G) i: the topological direct mm ofL(H) and L(E);
(ii) L(H) it a :ubalgebra ofL(G) and L(E) it an ideal of L(G) ;
(iii) L(o) it a linear reprexentation of L(H) on the Lie algebra of derivation: of
(iv; 13G) is the iemi-direet produet ofL(H) by L(E) Mined by L(o') (Chapter I,
§ 1, no.
(i) is obvious and (ii) follows li‘om Proposition 4-7. L(c): L(p) |L(H).
Now by Propositions 40 (no. 11) and 44 (no. 12), L(p) (t)1s, for all t E L(G),
the restriction ofadua, t to L(E). This proves (iii) Using (i) and (ii), this also
proves (iv).
COROLLARY. Let G be a Lie group. Let T,(G) be given it: unique commutative Lie
algebra structure. Let 1' be an adjoint representation of L(G). Then the Lie algebra of
T(G) is the .remi-direot product of L(G) by T.(G) defined by r. In other word:, for
x, x’ in L(G) andy,y’ in TAG),
COROLLARY. For all g e A‘, (at i(g) be the autarnarphiirn y 1—» gyg‘1 efA, to that
i is an analytic linear representation q’ an A. For all z e L(A’) = A, L(i)z i: the
inner derivation y ~—> zy — yz ofA.
This follows from Propositions 49 and 44.
267
m LIE GROUPS
268
MAURER—CARTAN FORMULAE § 3.14-
then obtain the differential form won G such that m, = T, (y(g‘ 1)) ; this difieren-
tial form is left invariant and analytic; it is called the lefl canonical dijfirmtial
form ofG. m,(t) = g"! for all teT,(G).
If F is again an arbitrary complete normable space and u e Y(T,(G), F),
then a)“ = u a m. In particular (taking F = K), the mapping v>—> u a m is a
linear bijection of the dual ofT,(G) onto the vector space of differential forms
of degree 1 with values in K which are left invariant under G.
Similarly, the differential form (0' on G such that a); = T,(B(g)) is called the
right canonical diflrmtialform of G. There are analogous properties to those ofco,
which we leave to the reader to state. The mapping g >—> g" 1 of G onto G trans-
forms on into m’.
269
III LIE GROUPS
dmk+z
4<1"”‘°"A to, = 0 (I: = l,2,...,n)
= figmflnu A my)¢(‘v‘r)-
The argument is similar for the w,’,.
270
mmonnue GROUP §3,16
15. CONSTRUCTION OF INVARIANT DIFFERENTIAL FORMS
Lemma 2. Let G be a Lie group, U a Jyrnmetric open neighbourhood of e in G, E a
complete normable space and c): U2 —> E an analytic mapping. For all g e U, let u), be
the diferential at the point g of the mapping h v—> ¢(g“h). Then (a iv the restriction to
U of the left invariant diferentialform on G whose value at e i: 434%
Clearly m, = 11,21). For allgeU and all tET,(G),
271
Ill LIE GROUPS
PROPOSITION 54. Let G be a Lie group «y’finile dimension n, m a left Marion! dg‘fl'eren-
tiolfonn qfdegree n on G and 11> on endomorphmn ofG. Then
<l>*(0) = (‘1‘:t L0H)!”-
We write L(¢) = u, a). =fand ¢‘(m), = g. For all x” . . ., x, in L(G),
g(x1, . . ., x“) =f(ux,, . . ., we”) = (detu)f(x1, . . ., x“)
and hence ¢‘(o>). = det L(<[>) .0)” On the other hand, ifg e G,
11> ° Y(:) = Y(<l?(g)) ° (l
and hence 7(g)*<]>*(m) = 41*(0). Thus (19(9)) is left invariant, whence the
proposition.
COROLLARY. For all g e G,
8(g)*m = (detAdg)m.
S(g)‘u = 3(g)‘y(g)*w = (Int g)*m and L(Int g) = Ad g.
Let G be a locally compact group and 4) an endomorphism of G. Suppose
that there exist open neighbourhoods V, V' of e such that ¢(V) = V' and
(I) | V is a local isomorphism of G into G. Let y. be a left Haar measure on G.
By Integration, Chapter VII, §l, Corollary to Proposition 9, there exists a
unique number a > 0 such that (My. | V) = fly. I V’. Clearly a is indepen-
dent of the choice of V, V’ and x-L. It is called the modulus ofcl> and is denoted
by moda <1) or simply mod (I). When <1) is an automorphism of G, we recover
Definition 4 of Integration, Chapter VII, § 1.
Pkoposmou 55. Suppose that K is locally tompad. Let y. be a Haor measure on the
additive group of K. Let G be a Lie group offinite dimension n.
(i) Let a) be n non-zero le/l invariant dferenlial form of degree n on G. Then the
meme: mod(w)u (Dyfmntiable and Analytic Manifoldx, R, 10.1.6) i: a lefl Haar
measure on G. If K = R and G ho: the orientation defined by m, the measure defined by
a: (Difirentioble and Analytic Manifoldr, R, 10.4.3) is a left Haar memre on G.
(ii) Let (I: be an elole endomorphim of G. Then mod <1) = mod det L(¢).
(i) is obvious. Let V, V’ be open neighbourhoods of e such that (NV) = V'
and (b ] V is a local isomorphism of G into G. Then
272
HMR MEASURE ON A LEE onour § 3.16
PROPOSITION 57. Let G be a Lie group of finite dimemian 71. Choose a basis for
/\"T,(G)*; by mean: nfthe right (resp. left) trivializatian Qf /\"T(G)‘, we can
273
m LIE onoups
identify this vector bundle with the trivial vector bundle G x K, .m that the transpose
of a scalar definenlial operator is ident'yied with a .realar dgfemifial operator.
Then, z'fu e U(G), the transpose qf L“ (rap. R1,) is L: (rap. RX).
We shall consider the case where A"T(G)“ has been trivialized using a right
invariant form to.
Suppose that the proposition has been proved for elemcnm ul, u, of U(G).
Then,
‘(qg = '(Lu, o Lug) (Proposition 23)
= ‘(L,,3) a ”(Lh) (Dz'fl. 69’ Anal. Mam, R, 14.3.3)
= Ll2 o L; by hypothesis
= Lin!“ (Proposition 23)
= n“), (Proposition 7)
and hence the proposition is true for u1 t 11;. It therefore suffices to prove the
proposition when u e T,(G). Now L,‘ is defined by G operating on G on the
right (no. 6) and hence Ohm = 0 since as is right invariant (Diflerentiabl: and
Analytic Manifolds, R, 8.4.5); therefore, iff is an analytic function in an open
neighbourhood of 2 with values in K, then 01m(ft») = (ea (Diflmntiable
and Analytic Manifolds, R, 8.4.8). Using the identifications made and Di mn-
liable and Analytic Mmg'fivldx, R, 14.4.1, the transpose of L" is —L,,, that is
L“.
COROLLARY. Le! G be a finitkdimemianal real Lie group, p. (resp. v) a left (resp.
right) Haar measure on G, I: an integer >0, ueUk(G) and f and g real-valued
flmdiam of class C" on G wilh men/Jae! support. Then
274-
LBFI‘DD’FERENTIAL §3.l7
denote byf ‘1 . if, and leave to the reader the task of translating the results for
the right differential.
Iff is the identity mapping of G, 1’"d is the canoniwl left difl‘erential
form a) of G. Returning to the general case of Definition 8,
(Jr-14.1mm = ‘"10») °T..(.f)
and hencef'1.¢‘1f=f*(m). This implies thatf‘1.dfis ofclass 0'".
Examples. (1) If G is the additive group of a. complete normable space and
To(E) is canonically identified with E, f "1.4;” is the difl‘erential 11f defined in
DWermIiable and Analytic Mani/Eddy, R, 8.2.2.
(2) Suppose that G is the multiplicative group A‘ associated with a com-
plete normable algebra A. Thenfcan be considered as a mapping of M into A
and hence the differential dfin the sense ofDiflnzntiable and Analytic Manijbldr,
R, 8.2.2 is defined and the product f ' luff in the sense of iflrmh'able and
Analytic Mani/"0M1; R, 8.3.2 is defined. Clearly the latter form is identical with
the left differential off.
Psoposmon 58. Let G and H be two Lie groups, M a mmfild nf clas: 0’, f a
mapping qfclas: C' q into G and}! a mtphitm q into H. Then
(I: °f)“-d(h =f) = we o (run = (h'1.dh)o mm
For all 16M and u eT,(M),
(ll °f)"-4’(ll °f)(") = ((h °f)(x))"-T(ll °f)(")«
The latter expression is equal, on the one hand, to
T(")(f(*)"-T(f) (14)) (§ 2, PI‘OPOSifion 5)
= .(h)((f“.df)("))
and, on the other hand, to
h(f(")) "T(h)(T(f)")
= (Ix-l mama)
PROPOSITION 59. Let G be a Lie group, M a manifizld (folly: 0’ f and g mapping:
qfclass C" qf M into G and]: the canonical my'zction of TM nnIaM
(1) (f5) 1410%) = (Ad °§ ° 1") 1 ° (fa-'1!) + Edi:-
(ii) Writing Mm) = f(m) " fur all m s M,
h-uh = —(Adnfop) nor-1.4;).
Assertion (i) follows from § 2, no. 2, Proposition 7. Assertion (ii) follows from
(i) by putting g = h.
COROLLARY 1. Let :EG and 5g be the mapping x»:g(x) If M into G. That
(Xg)“.d(5g) = (Mg.
275
m LIE GROUPS
276
Lm AMEBRA or A LIE axon? GERM § 3.18
277
m LIE GROUPS
18.6. Let G be a Lie group germ and t E U(G). L. denotes the field of point
distributions g >—> :, a: ton G and R‘ the field ofpoint distributions g >7> t A: e, on
G. Iffe WKG, F), then L,fe @“(G, F) and RJE W‘KG, F). For 1,! in U(G),
Lm' = L: ° Ln RM’ = Rl’ ° R!) L: ° R:’ = Rt“ ° Ln 0(L:) = R:-
278
PASSAGE FROM LIE ALGEBRA nonpmsms T0 LIE GROUP uoapmsms §4.l
279
III LE GROUPS
p;T(H). Let a be the graph off; it is a closed Lie subalgebra ofL(G) >< L(H)
which admits {0} x L(H) as topological supplement. For all (g, h) EG X H,
the graph off,‘ ,, is (g, h) .a, The union of these graphs is an integrable vector
subbundle of T(G x H) (Lemma 1). Then there exist (Dyfzrenttable and
Analytic Many’bldr, R, 9.3.7) an open neighbourt U of ‘5 in G and an
analytic mapping 4; of U into H such that (Meg) = on and T,(¢) = ft“,
for all g 5 U. In particular, T,o(d>) = f.
Let V be an open neighbourhood of ea in G such that, for (s, t) s V x V,
the products st and q>(:)¢(t) are defined and :16 U. Consider the mappings .
a“ an ofV x V into H defined by
280
PASSAGE mom LIE menus TO 1.11: (moms §4.2
281
III LIE GROUPS
group
THEOREM 2. Let L be a complete normable Liz algzbra. Than exirtr a Lie
germ G rash that L(G) is irornarphia to L. Two such Lie group germs on locally
isomorphic. .
The first assertion follows from Lemmas 2 and 3. The second assertion
follows from Corollary 1 to Theorem 1 of no. 1.
282
PASSAGE mom LIE AUGEBRAS 'ro LIE GROUPS §4.2
THEOREM 3. Let G be a Lie group gem and b a Lie suhalgebm of L(G) admitting
a Iapologicul supplement. There exist: a Lieruhgroup gem H of G such that L(H) = b.
If H1 and H, are Lie subgroup germs of G such that
L(H‘) = L(H,) = b,
Pnoposmou 1. Let G be a Lie group germ and H and H’ two Lie subgroup gen/u.
In order that L(H) => L(H’), it is memory andsufia'ent that H n H’ be open in H'.
If H n H’ is open in H', then L(H’) = L(H n H’) C L(H). Suppose that
L(H) D L(H'). Let i, i’ be the canonical injections of H, H’ into C. By
shrinking H’ if necessary, it can be assumed that there exists a morphism lb
of H' into H such that L(qa) is the canonical injection of L(H’) into L(H)
(no. 1, Theorem 1). Then L(i o 4;) = L(i’) and hence there exists a neigh-
bourhood V of:1T in H’ such that i 0 vi; and 1" coincide on V (Theorem 1).
Therefore V C H, hence VC HnH’ and HnH' is open in H' (§1, no.
10).
283
I" LIE GROUPS
3. EXPONENTIAL MAPPINGS
open neighbourhood
THEOREM 4. Let G be a Lie group germ, L it: Lie algebra, V an
= IdL.
of 0 in L, <1; an analytic mapping of V into G me}: that M0) = 0 and To(¢)
Thefollowing eonditionr are equivalent: .
(i) For all b e L, “(A + A’)b) = d»(lb)¢(1’b)for |x| and DJ| .mfinenlly mall.
(ii) For all b e L and every integer n > 0, <1>*(b") is homogeneous of degree n 1n
U(G) (T1,”(L) is identfiied with TS(L) and b" it enoluated in TS(L)).
(iii) The mapping 4’: of TS(L) into U(G) is compatible with the graduation: of
-
TS(L) and U(G).
(iv) The mapping 4": of TS(L) into U(G) is the canonical mapping of TS(L)
into the enveloping algebra y“ L.
(v) There exist a norm on L defining the topology of L and such that
"Biz/ll] S M M
for all x, y in L and an open subgroup germ W C V of the Lie group germ defined
by L (no. 2) such that (MW 1': an isomorphism of W onto an open Lie subgroup germ
a G.
f(v) => (i): obvious, for (110.0(1)) = (A + k')b in W for m and [NI suffi-
ciently small.
(i) => (ii): suppose that condition (i) is satisfied. Let b EL. Let L}; be the
restriction of 4; to V n Kb. By hypothsis, there exists a symmetric neighbour-
hood T of 0 in the additive Lie group Kb such that '4! [T is a morphism of the
Lie group germ T into G. Hence
454”") = (“TM“) = ((4111345)? = “MUD",
so that ¢*(b") is homogeneous of degree n in U(G).
(ii) => (iii): this follows from the fact that TS"(L) is the vector subspace of
TS(L) generated by the n-th powers of the demerits of L (Algebra, Chapter IV,
§ 5, Proposition 5).
(iii) r» (iv): the canonical mapping of TS(L) into the enveloping algebra
of L is the unique morphism of graded cogebras mapping 1 to 1 and extending
IdL (Chapter II, 13‘ 1, no. 5, Remark 3). Now (I). is a cogebra morphism and
¢*|L = IdL by hypothesis. If condition (iii) holds, it is seen that condition
(iv) also holds.
(iv) => (v): suppose that condition (iv) is satisfied. Choose a norm on L
defining the topology of L and such that ||[x, y] I} s “x“ H y“ for all x, y in L.
Let H be the Lie group germ defined by the normed Lie algebra L. By
Theorem 1, there exist an open subgroup germ S C V of H and an iso-
morphism o’ of S onto an open subgroup germ of G. As we know already that
(v) => (iv), the mapping o; of TS(L) into U(G) is the canonical mapping of
TS (L) into the enveloping algebra of L. Thus ¢,(t) = (1);“) for all t E T5,”)(L).
As <12 and 4)’ are analytic, d} and l” coincide on a neighbourt of 0.
284
EXPONENTIAL mpmcs § 4.3
DEFINITION 1. Let G be :1 Liz group gem and L its Lie algebra. An exponential
mapping if G is any analytic mapping 4: dqined on an open neighbourhood 1y” 0 in L,
with values in G and satisfying the conditions of Theamn 4.
Theorem 4 implies immediately that, for every Lie group germ G, there
exists an expzmential mapping of G and that two exponential mappings qf G mincidt
on a neighbourhood qfO.
Examples. (1) Let G be the additive group of a complete normable space E.
The canonical isomorphism of L(G) onto E satisfies condition (i) of Theorem
4- and is therefore an exponential mapping of G.
(2) Let A be a complete normed unital associative algebra. Let A* be the
Lie group consisting of the invertible demerits of A. We identify L(A*) with
A (§ 3, no. 9, Corollary to Proposition 33). If K = R or C, we know that
the mapping exp of A into A‘ defined in Chapter II, § 7, no. 3 satisfies condi-
tion (i) of Theorem 4- and hence is an exponential mapping. Now let K be
ultrametric. Let p be the characteristic of the residue field of K. If p 96 0,
let A = ”41””);i = 0, let A =1. Let U be the set ofxEA such that
[l < A. We know (Chapter II, §8, no. 4) that the mapping exp of U
into A‘ satisfies condition (i) ofTheorem4ahd henceis anexponential mapping.
Note that U is an additive subgroup of A.
This example explains the terminology adopted in Definition 1.
Let G be a Lie group germ and (1) an exponential mapping of G. Then
(l) is étale at 0 and hence there exists an open neighbourhood U of 0 in L(G)
such that (MU) is open in G and ¢|U is an isomorphism of the analytic mani-
fold U onto the analytic manifold ¢(U).
A cananical chart (qf thefirst species) on G is a chart qi'on the analytic mani-
fold G whose inverse mapping is an exponential mapping. If further G is
finite-dimensional and a basis ofL(G) is chosen, the coordinate system defined
by s]; and this basis in the domain of £1) is called a canonical coordinata system (of
thefirst species).
PnorosmON 3. Let G be a Lie group germ, L its Liz algebra and (I) an exponential
mapping of G. Let L1, . . ., L" b2 vector subspaces qf L suth that L is the tapologiml
dimt sum qf L1, . . ., L,,. The mapping
(b1) 11:, ~ ~ -, 17..) >-> 0(b1, 52. ~ ~ u bn) = #0044112) . . w.»
defintd an an open subset (y'L, >< Lz x - ~ ~ x Lm is analytic. The tangent mapping
at (0, 0, . . ., 0) to 0 is the canonical mapping (y‘L1 x - - - x L,I into L.
Let k. be the canonical injection of L, into L1 x L2 >< - ~ - x L... Then, for
all bELo (T<u.....u)e)(TokA)(b) = (Tol’xb) = b and hence (Tto.....a)e)lLi is
the canonical injection of Ll into L.
In particular, 0 is étale at (0, 0, . . ., 0). Its restriction to a sufficiently small
285
ll'l LIE GROUPS
a) x + y = ”33:3;303-14:-1(¢(m¢<xy))
(2) [w] = 3333:; r24»-I<¢<u)¢(3y)¢(—w¢(—Ay)>
(note am <1>“(d>(7»~)¢(7:y)) and ¢-‘(¢(M)¢<xy)¢<—m¢(—xy)) are defined
for |A| sufficiently small .
Let L = L(G) be given a norm defining the topology of L and such that
H [x, y] H < ”x” [| y[| for all x, y in L. Using Theorems 2 and 4:, it can be assumed
G is the Lie group germ defined by L and that 4) = Ida. Let
(x, y) >—> my denote the product in the group G. The formulae to be proved
can then be written
(a) x + y = uxlifiqflk"((kxl.(ky))
(4) [m1 = “333333-=<w>.w).(—m4-33)).
There exists an open neighbourhood V of 0 in K such that the function
lO) = (MHM)
is defined and analytic on V. By Chapter II, §6, no. 4, Remark 2, the expan-
sion off as an integral series about the origin is
W + y) + Why] +
and this proves (3). On the other hand, for u, v in G and flu", |[u|| sufficiently
small, 14.0 is an analytic function of (u, v) and the terms of degrees 1 and 2
in the expansion of this function as an integral series about the origin are
u + v + flu, 1)]. By Dgfirmtz'able and Analytic Manfildx, R, 3.2.7 and 4.2.3,
the terms of degrees 1 and 2 in the expansion of the function f (A) .f(—-)\)
as an integral series about the origin are the terms of degrees 1 and 2 in
, f0) +f(-7\) +l[f(7:):f(-7:)]
or also m
10' + y) + ‘Wlmyl - M” + y) + ll’Dw]
and this pm. (4), + W" + y): —x(~ + y)] = mm]
286
EXPONENT‘AL MAPPINGS § 4.3
m; m (“1 X) Y
l I
(Chapter II, § 6, no. 5, Proposition 5). The proposition then follows fiom
Dmrmtioble and Analytic Monifirlds, R, 3.2.4 and 4.2.3.
Let G be a Lie group germ and teK. A t-th power mopping if G is any
mapping, defined and analytic on an open neighbourhood of e, with values
in G and coinciding on a neighbourhwd of e with a mapping
.4 H “14> "(9)
where 4) is an injective exponential mapping of G.
Pnoposmon 7. (i) If 1 e2, a t-th power mapping m'oeide: on a neighbourhood of
e with the mapping g »—> g’.
(ii) The tangent mapping at e to 11 Mb power mapping is the homothrty 12f ratio t.
(iii) If h it a t-th power mapping and h' a t'-th power mapping «f G, h o h’ i: o
(tt’)-th power mapping and g 2—» h(g)h’(g) is o (t + t')-th power mapping,
(iv) [f h is a t-th power mapping and u e U"(G), then h,.(u) = fire.
It suffices to prove the proposition when G is the Lie group germ defined
by a complete normed Lie algebra and when the t-th power mappings con-
sidered are constructed using the exponential mapping 4) = Ida. But in that
use everything is obvious.
287
III [E GROUPS
fir all x e V.
9%)) = mam»)
This follows from Proposition 8 and Example 2 of no. 3.
COROLLARY 3. Let G be a Lie group and (b an exponential mapping of G.
(i) There ӣi a neighbourhood V of 0 in L(G) such that
Ad(¢(x)) = exp ad x.
fir all x e V.
(ii) Ifg e G, there exist: a neighbourhood W of O in L(G) such that
g¢(*)g“ = (Ads-x)
fir all n SW.
(i) follows from Corollary 2 and § 3, no. 12, Proposition 44.
(ii) follows fiom Proposition 8 applied to Int g.
288
smucruu maven) ON A summon-p §4.5
[01,a2] = (ada,).a,el)
since I) is closed in L(G). Hence we have proved (i). In the rest of the proof
we fix a Lie subgroup germ H' of G with Lie algebra b.
Let V, v0, f be as in (iii). Let Y be the left foliation of G associated with I)
(no. 1). For all y eH’, T,(H’) = yl). On the other hand, for all 1) EV, the
image of T,(V) under T.,(f ) is contained in 9m) = f(v)l) (by definition of
hm»). By Diflerentiabl: and Analytic Manifolds, R, 9.3.2, f is a morphism of V
into Y. As H’ is a leaf of Y (Diferentiable and Analytic Manifolds, R, 92.8),
f (v) e H’ for v sufficiently close to 00.
Let (a1, . . ., a,) be a basis of I). There exist an open neighbourhwd I of 0
in K and mappings f“ . . .,f, of class C' of I into G such that f,(0) = e,
f,(I) C H, (Tf,)1 = a, for all j. By (iii), £0.) eH’ for [A] sufficiently small.
Hence thef,(>.,)f,(l¢) . . .f.(}\,) constitute, for It‘ll, lhzl: . . ., IM sufiiciently
289
11! ME GROUPS
290
PRmmVPs 0P Dmum‘ml. FORMS §4.6
irmnmornmmmronmmmvummnm
ALGEBRA ,
Lemma 5. Let X be ammifizld 0felas: C', FandF' veetorbundlerofelar: C’ with
base:pmeXand¢amorphianq/FintoF’.FarallxeX,letbelhesetqf
(a, W» e F. (B F;
for a sh. Then the union 5 ofthe s,‘ is a veaormbbmdla ofF @F’.
Let 9 and 6’ be the mappings of F (-BF’ into itself defined as follows: if
(u, 11) 6F, ® F;, then
9(a) '1) = ("HI + Mu». 9’04?) = (14,” — $00)
By Dzfirentiable and Analytic Manifaldx, R, 7.7.1, 0 and 8’ are morphisms of
F (-BF’ into itself. Clearly 0 o 6' = 0' o 0 = Idfepr. Hence 0 and 0’ are
automorphisms of F G F’. Therefore, S = 0(F @(0}) is a vector subbundle
of F a F’.
Lemma 6. Let G be a Lie group germ, m the canonical lefl dzflerential form qf G
(§ 3, no. 18.9), M a mung/721:1 qfclm C’ (I 2 2) and a: a dg'fl’erenfialform ofelas:
G"1 and degree 1 on M with value: in L(G).
(i) The element: q(M x G) at which the dtferentialform
0 = prfa — lax-:0)
is zen; constitute a vector .mbbundle S ofT(M x G) qftlaxs 0'“.
(ii) For all (x,g) EM x G, T(pr1)|S(,,',, is an iromorphirm qf SUM, auto
T (M).
’‘(iii) If do: + [a 2 = 0 (cf. §3, no. 14-) the vector subbundle S is integrable.
If (2,5) 6 M x G and (u, u) eTx(M) X T,(G), then
e(x.n("x u) = 10‘) — g'l'“
Hence the kernel of 00",, is the set Sam, of (u, ga(u)) for u sTx(M), whence
(ii). We consider T(M x G) as the direct sum of two vector bundles F and
F' with E...) = TAM) x {0} and F2...) = {0} x T.(G) for an
(x, g) e M x G.
For it e Tx(M) x {0), we write (Mu) = (0, ga(u)). Using the left trivialization
of T(G), it is seen that (I; is a morphism of F into F', whence (i) (Lemma 5).
Finally, if act + [at]2 = 0, then
:16 prfldu) —— pram)
—%(pri‘ a A P’i a - Pr; 6 A Pr: a)
— —*l(Pri‘¢c — PT? 0’) A (Prl'a + Pr: 0*)
= —;e A (prra + pea)
and hence S is integrable (Diflererm'able and Analytic Manyaltir, R, 9.3.6).
291
m up. (moors
d¢
M5. LetGlzeuLic ou gem,Manmmfold¢folas:C' (r22)ari
in_L(G), such the:
33$;tial form if l‘lflfl‘ 0’5, ail degree 1 on M with values
Mined and
dot + [at]“ = 0. For all are M and all geG, there exutr a mappzngf,
that f (z) '= g
of elm: CY“ on an open neighbourhood of x, with values 1.". G, such
on a neigh-
and f ‘1.df = a. Two mopping: which satisfy there condition: coincide
.
ho hood 0 x.
"Itet x éfM and ge G. By Lemma 6 (whose nomtion we adopt) and Dif-
od
ferentiable and Analytic Manifolir, R, 9.3.7, there exist: an open neighbourho
U ofx in M and a mapping rm» Mm) = (rn,f(m)) ofclass C"1 ofU mto
M X G such thatf(x) = g and ¢*(0) = 0. Then
f".tlf=f‘(o) (§3,no.18.9)
= (P's ° 4’) ‘ (40) “Off = 9’2 ° i)
= win: a — 0) (Lem-a 6)
= (WW? 00 (for 4>*(°) = 0)
= at (for pr, 0 (l: = Id“).
292
PRIMITIVES or DIFFERENTIAL FORMS §4.6
whence (i). On the other hand, if c.) denotes the left canonial differential
form of G, then
”Jim". = (Tmf)_1°(f_1‘4f)(m)
= (Tmf)“°m(f(m)) °Tmf
and hence m r—> am'al o a", is the left: canonical differential form of M’.
Let M” be an open neighbourhood of mo, with a Lie group germ structure,
with identity element Inn and with the analogous properties to properties (i)
and (ii). Then an,“ is an isomorphism of L(M') onto 9 and also of L(M”)
onto 9 and hence L(M’) = L(M"). Therefore, shrinking M’ and M”, it can
be assumed that there exists an isomorphism (I) of the group germ M’ onto
the group germ M” (no. 1, Corollary 1 to Theorem 1). Then ¢‘1.dd> is the
canonical left differential of M’. On the other hand, let 4; be the canonical
injection of the manifold M’ h M' into the Lie group germ M”; clearly
ill-1.44; is a restriction of the canonical left differential of M”. Hence
(¢‘1.d¢)(m) = 09,701 0 at", = (¢“.d¢)(m) for all meM’ n M”. Therefore (it
and t]; coincide on a neighbourhood of mo (§ 3, 18.9). This proves the last
assertion of the proposition.
COROLLARY. Let M be an analytic many’old offinite dimension n. Let (0,, . . ., a),
be analytic dfirentialfim: 4y” degree 1 on M, with scalar values, which are linearly
independent at each point of M and such that, for all k = l, t . t, n, dink i: a linear
combination with constant coeflicient: of the to. A 0,. Then, for all rn0 EM, there
exist: an open neighbourhood M' of mo in M and a Lie group germ structure an M'
compatible with the manifizld structure on M', with identity element mo and such that
mllM’, . . ., mn|M’fimn a ban": qfthe space oflcfl invariant a'tfl'erentialflrm: on
M' of degree 1 with scalar values.
if M; and M; are two such group gem, M; and M; have a common open sub—
group germ.
Let X1, . . ., X,l be the vector fields on M such that, at each point m of M,
the (X,),,, constitute the basis of T,"(M) dual to ((co,),,,,. t ., (9%)...)- These
fields are analytic. By hypothesis, there exist cw, GK (1 $ 1', j, k S n) such
that cmc = —c,,,, and do», = (<2, cake), A 0),. By DMerentiable and Analytic
Manifolds, R, 8.5.7, formula (1 l),
293
III LIE GROUPS
= —du.
It then suffices to apply Proposition 10.
294
LAWS OF maximum arm-non §4~.7
Hence for 32X, the morphism gr—>a(g,x) and gr—tfl(g,x) are integrals
off; as ,
90, It) = M") = 1(6) x)
for all xeX, it follows from Dlflerentiable and Analytic Manifoldr, R, 9.3.7,
that a and fl coincide on a neighbourhood of (3, 1°). Hence the proposition.
COROLLARY. Let G be a Lie group gem and X a manifold of tlass C" Comider
two law chunks lfleft operation qfolass C' of G on X. Suppose that, for all a e L(G),
the corresponding vettorfield D,1 on X is the samefor both low chunks. Then these two
law thanks tointide on a neighbourhood of {e} x X.
THEOREM 6. Let G be a Lie group germ, X a mamfild of class C' (r 2 2) and x“
a point ofX. Let a >—>Dn be a law of left infinitesimal operation ofelass C"1 of
L(G) on X.
(i) There exists an open neighbourhood X’ of en in X and a law thunk of left
operation of class C"1 of G on X' such that the associated law of infinitesimal opera-
tion is o ~—> DQIX’.
(ii) Let there be two law [hunks of lefl operation of tlass C"1 of G on an open
neighbourhood X” of :0; if they admit a n—> D¢|X' as associated law of infinitesimal
operation, they coincide on a neighbourhood of (e, no).
Assertion (ii) follows from the Corollary to Proposition 11. We prove (i).
Ferall (g,x) eG x Xand all aeL(G),wewrite
0.441,") = (agyDa(~)) GTAG) X Tz(X)~
Let Sm,” be the set of QA(g, x) for a s L(G). By Lemma 5 of no. 6, the s“,
are the fibra of a vector subbundle S of T(G) x T(X)’. Let a, b be in L((}) ;
then
[Qm Qu](g, x) = “Ra: Rh] (5), [DD Do] (1‘))
= (-Rm.m(g), -Dm.u(’t)) (§ 3: 1&5)
= Q-la.o](g:*)
and hence S is integrable (Difl'ermtiahle and Analytic Manifolds, R, 9.3.3, (iv)).
By Dzfirentiable and Analytic Mam'fizlds, R, 9.3.7, there exist an open neigh-
bourhood G‘ of e in G, an open neighbourhood X1 ofxa in X and a mpping
(g,x) n of class C"1 ofGl x Xl into X such that ex = x for all xeX,
and
(6) (ag)x = D,(gx) for a e L(G), g e G], x 6 X1.
In particular
(7) ax = Duo).
Let G, be an open neighbourt ofe in GI and X2 an open neighbourhmd
of 3., in X1 such that g' is defined and belongs to CI for g, g’ in G2 and gx
295
III LIE GROUPS
is defined and belongs to Xl for (g, x) EGa X Xa- Consider the mappings
a1, :12 ofG, x (G, x X2) in“) X defined by
296
'n-m uoemcmms 5“,, § 5.1
Remark. Laws oflefl operation can of course be replaced, throughout this no.,
by laws of right operation.
Letf, g be two formal power series with coefficients in K in the same indeter-
minates, let )2 (resp. g.) be the homogeneous component off (rexp. g) of de-
gree i. We shall write
f E g mod deg p
iffi =g.fori <p.
In this §, C denotes a Lie group gem of finite dimension n; the base field
K is assumed to be of characteristic zero. We identify once and for all, by
means of a chart, an open neighbourhood of e in G with an open neighbour-
hood U ofO in K", so that e is identified with 0. For x,g in U and n 52, any
denotes the product of x and y and x‘“ the m-th power of x in C (When they
are defined). The coordinates of x e U are denoted by x1, x,, . . ., x".
1. ms COEFFICIENTS c...
Let Q be the set of (nay) e U x U such that mg is defined and belongs to U.
Then 9 is open in U x U and the mapping (x,y) why of 0 into U is
analytic. The coordinates zl, . i ., z,I of z = my therefore admit expansions as
integral series about the origin in the powers of xi, . . ., x,” 3/1, . . . , 3/". There-
fore, there exist well defined constants a,”"u,”‘gbwgflvbuuvn E K, such that
v v _ v B B
(1) 251l ' ' ' Zn" — a" "when clt1_...,m~,9h...,8.,v1,"”1”l - ' - 3:"!11‘ ‘ . 41.1”
for 7‘, . . .,y,, in N. Adopting the conventions of Dgferenh'ablc and Analytic
Mani/Hit, R, we shall write these formulae more briefly:
297
111 LIE GROUPS
component of
On the right hand side of (5), we consider the homogeneous
degree 2:
<t«.f> = 7*:-
In particular,
(t,, x”) = 8a,.
298
BRACKET m m up. ALGEBRA §5.2
Hence
(a, t ‘5: x') = (ea ® :5, (x.y)‘>
= (‘3 ® 0m “2 ‘u’a’fi‘'11")
299
I'll LIB GROUPS
and hence g,(x) = B(x, x). This proves (i). Then, using (i),
x.y.x"“ E (x + y + B(x,y)).(—x + B(x,x)) mod degB
E x + y + B(x,y) + (—x + B(x,x)) + B(x + y, -x) mod deg3
E y + B(x,y) — B(y, x) mod deg 3
—=— y + [16y] mod deg 3 (by (13))
whence (ii). The proof of (iii) is analogous. Combining (i) and (iii), we obtain
shag/["1114] E (—x + B(x,x)).(x + [x,y]) mod deg3
a —x + B(x,x) + x + [x,y] + B(—x,x) mod deg3
E [x, y] mod deg 3
whence (iv). The proof of (v) is analogous.
3. POWERS
Consider j points of G:
3(1) = (1(1):, “(1)2, . ~ ': 3(1),.)
’42) = (1(2):: 42):, ~ ~ -, 3(2)")
300
POWERS § 5.3
(We writs
(t) _ t(t — 1)...(t — i +,1)
i _ 1'!
far all t E K.)
Assertion (i) is obvious fi‘om the definition of the 49.
We prove (ii) for I an integer 20. By (14),
_ 1 #...+ I
(17) ”m — «(lwzwmm am).i...mu)"u“ "Q
For at = (1(1), . . ., «(0) e (N")’, let 6(a) denote the set ofje{l, 2, . . ., I}
such that «(1') aé 0. If, in the sum (17), we group together the terms for which
a(u) is the same, then
(18) gm = "Enh‘flm
where
Let a = {j,,j,, . . .,j,,}withj1 <j, < . . . < jg. In (14) (wherejis replaced by
t), we substitute 0 for x(k) for 1: ¢ 6; as 0 is the identity element ofG, we obtain
the expansion of AA) .x(j,) . . .x(j¢) as an integral series about the origin:
.).
(20) 4’4") = ,(;,a¢(i).....w> fU:)+---+w
By (19) and (20), we see that h,'°(x) = ‘lc-nl ,(x). Then (18) implim
301
Ill LIE GROUPS
act-u = Z, (—woo.
(3) The above expremion for 4;, and formula (6) prove that
(21) ¢2(x) E B(x, 1) mod deg 3.
Using Proposition 2, (i) and (ii), we see that
defined as above, the argument again proves that 1‘“ = ‘21 (I) ¢.(x) if
= I
teN.
302
EXPONBNTIAL § 5.4
4. EXPONENTIAL
Let E(x) be the expansion ofan uponential mapping of G as an integral series
about 0‘ Let L(x) be the expansion of the inverse mapping of an injective
exponential mapping of G as an integral series about 0. Since the tangent
mapping at 0 to any exponential mapping is the identity of L(G),
E(x) E x mod deg 2 and L(x) _=_ a: mod deg 2. Since E(L(x)) = L(E(x)) for x
sufficiently close to 0, the formal power series E and L are such that
E(L(X)) = L(E(X)) = X. An analogous argument shows that
EUX) = (1309)"), L(xm) = tL(X)
for t E K.
PROPOSmON 3.
__ -1
(27) L = §,( 2” 4a:
m l
(28) E = 21W“
(Recall that 41,. p is the homogmmu component (If 4;, qfdegm 1;.)
E(1") = (EGO)m
or, by (24-),
Em) = M; “Z“ Hemmer».
The two sides are formal power series in t and x. Equating the terms of first
degree in t, we obtain
(29) x = “2., 4mm».
Now the inversion of a system of formal power series, when it is possible, is
possible in only one way (Algebra, Chapter IV, § 6, Corollary Do Proposition 8).
Then
we = Z mot) by (29)
1100
= 2 flow) by (25)
Pull p
= Z ‘7)” m),
whence (i). Similarly, for 2 ye 0,
L0”) = 1L((Ix)“’1‘)
= IL(Z Z t'""<]>,_,,,(x)) by (24).
"we 10‘»:
303
in LIE GROUPS
a: = L(,?;:o (bu-"(6)
whence
E(x) 2 two)
m>o
Zofiivm.m(~) by (26).
= m;
PROPosmON 4. For Ihe chart of G mm! to b: omnital, it is nemrury and .mficimt
thattl;,=0forj2 2.
This is sufficient by Proposition 3. Suppose that the chart is canonical and
n
that 4,. = 0 for 2 g i < n. Then to: = gm = 20 (mix) = ux + Mg),
whence 41“ = 0. Hence 4;, = 0 for j 2 2 by induction on j.
304
LIB ALGEBRA MORPHISMS T0 LIE GROUP MORPHISMS § 6.1
and hence A(.t, l, U') C A(g, h, U). It then follows that the A(g, h, U) form
the base of a topology on G x H. We shall denote by Y the set G X H with
this topology and denote by [z the canonical projection of Y onto G, which is
open. The restriction ofp to A(g, h, U) is a homeomorphism ofA(g, h, U) onto
gU. Hence (Y, p) is a covering space of G. Let Yo be the subgroup of Y gene-
rated by A(e, e, W) and let 93 be the set of A(e, e, U). Clearly % satisfies con-
ditions (GV{) and (GVfi) of General Topology, Chapter III, § 1, no. 2. The set
Y5 ofy 5 Yo such that the mappings z ny' 1 and z Hy‘1zy o0 into Yo are
continuous at (e, e) is a subgroup OfYa- Let w E W. The mapping w' n—> ww’w"
of W into G is continuous and hence the mapping
("13f(w')) H (ww'w'fif(ll/“1'10"”
ofA (e, e, W) into Y is continuous at (e, e) . Nowf(ww'w ' l) = f(w)f(w')f(w ' 1)
and therefore (ww'w",f(ww'w‘1)) = (w,f(w))(w’,f(W’))(w,f(w))“-
As w“ EW, we see that (w,f(w)) 6Y3. Thus, A(e, e, W) C Y3, so that
Y", = Yo. The group Yo, with the filter base 3 therefore satisfies condition
(GVl’n) of General Topology, Chapter III, § 1, no. 2. As
(g, h) ~A(€, 6, U) = Mg, 1!, U),
Yo is a topological group which is connected since A(e, e, W) is connected.
Then MYo) is an open subgroup of G, whence 17(Yo) = G since G is connected.
The kernel of[z I Yo is discrete. As G is simply connected, [1 | Yo is a homeo-
morphism of Y0 onto G. Therefore Yo is the graph of a morphismf’ of G into
H- For g E W, (g,f(g)) 6 AG: 6 W) C Yo, whenceflg) =f'(s’)-
THEOREM 1. Let G and H be Lie groups and h a continuous morphism of L(G) into
L(H). Suppose that G is simply connected. Then there exists one and only one Lie group
morphism (I) of G into H such that It = L(<]>).
The existence of 4) follows from Lemma 1 and §4, not 1, Theorem 1 (i).
The uniqueness of <1) follows from § 4-, no. 1, Theorem 1 (ii) and the fact that G
is connected.
COROLLARY. Let G be a finite-dimensional simply connected Lie group. There exists
afinite-dimemional analytic linear representation of G whose kernel is discrete.
There exist (Chapter I, § 7, Theorem 2) a finite-dimensional vector space E
and an injective morphism h ofL(G) into the Lie algebra End(E). By Theorem
1, there exists a morphism (I: ofG into GL(E) such that L(¢) = h. Hence cl: is an
immersion and therefore its kernel is discrete.
Remarks. (1) There exist finite-dimensional simply connected Lie groups
which have no finite-dimensional injective analytic linear representation
(Exercise 2).
(2) There exist finite-dimensional connected Lie groups G such that every
finite-dimensional analytic linear representation of G has non-discrete kernel
(Exercises 3 and 4).
305
Ill LIE GROUPS
2. WTEGRAL SUBGROUPS
DEFINITION 1. Let G b: a Lie group. An integral subgroup ofG is a subgroup H with a
connected Lie group structure such that the canonical injection if H mm G u an
immersion.
A one-parameter subgroup of G is a l-dimensional integral subgroup of
306
mam suBGRovJPs § 6.2
are equal. This proves (ii) and also proves that the mapping considered in (i)
is injective.
Remark 1. Let H be an integral subgroup of G. Let Y be the lelt foliation of G
associated with L(H). If g e G, let gH be given the manifold structure derived
from that on H by y(g). By Difierentiable and Analytic Manifolit, R, 9.3.2, the
canonical injection ofgH into Y is a morphism. This morphism is étale. Hence
the maximal connected leaves of Y are the left oosets modulo H.
PROPOSITION 1. Let G and M be Lie groups, H an integral :ubgrnup of G and 4) a
morphism of M inla G web that L(<]>) (L(M)) C L(H). Suppose that M it connected.
Then ¢(M) C H and (5, considered a: a mapping qfM into H, i: a Lie grnup murphism.
In the notation of Remark 1, d) is a. morphism of M into Y (Difierentiabls and
Analytic Manifiilds, R, 9.3.2) and hence (HM) C H since M is connected.
COROLLARY 1. Let G and H be Lie groups, 4) a Lie group morphism ifG into H, N the
kernel of d) and h = L(<])). Suppose that G is connected and that H isfinite-dimenrianal.
(i) N is a Lie subgroup q and L(N) = Ker h.
(ii) Let H’ be the integral rubgruup 13f H with Lie algebra Im h. Then (MG) = H’.
(iii) The mapping (f GIN inta H’ derivedfrom 4; when passing to the quotient i: a
Lie gran]: itamarphixm.
(i) has already been proved (§ 3, no. 8, Proposition 28).
Let 4; be the Lie group morphism of G/N into H derived from :1) when
passing to the quotient; it is an immersion (§ 3, no. 8, Proposition 28). By
Proposition 1, u]; is a Lie group morphism of G/N into H’. This morphism is
étale and hence ¢(G/N) = H' since H’ is connected; this proves (ii). Then
4;: G/N —> H' is bijective and is a Lie group isomorphism, which proves
(iii).
COROLLARY 2. Let G be a Lie group and H1 and H2 integral subgroup: g" G. If
L(H2) C L(H1), then H, is an integral subgroup al.
Let i1: Hl —> G, i2: H2 —> G be the canonical injections. Then
L(i2)(L(Ha)) = L(H2) C L(H1)-
By Proposition 1, i2 is an analytic mapping ofH2 into H1 and even an immer-
sion of H2 into H1 since L62) is an isomorphism ofL(H2) onto a subalgebra of
L(H1) admitting a topological supplement.
COROLLARY 3. Let G be a finite-dimem‘imal Lie group and (HI)IE r a family 9/" Lie
subgroup; q. Then H = .01 H, i: a Lie subgroup ofG and
fl L(H.).
L(H) = tel
There exists a finite subset J of I such that ‘0‘, L(H‘) is equal to the inter-
307
III LIE GROUPS
section M of all the L(H,). We know that H“ = ‘0] H‘ is a Lie subgroup such
that L(H“) = M (§ 3, no. 8, Corollary 2 to Proposition 29). Let HD be the
identity component of H*. It is a Lie subgroup of G and L(Hu) = M. By
Corollary 2, Ho C H. for all i and hence Ho C H C H*, whence the corollary.
COROLLARY 4-. Let G be afinite-dimenrional connected Lie group. Thefullnwing con-
ditian: are equivalent:
(i) G i: unirncdular (Integration, Chapter VII, § I, no. 3, Definition 3);
(ii) detAdg = lfarallgeG;
(iii) Trad a = Ofnr all as L(G).
The mapping g »—> det Ad g is a morphism cl: of G into K“. By § 3, Proposi-
tions 35 (no. 10) and 44 (no. 12), L(¢)a = Tr add for all aeL(G). Clearly
Im L((b) = (0} or K. In the first (resp. second) case, Im q5 = {l} (resp.
In) 42 = K*) by Corollary I and hence G is unimodular (resp. not unimodular)
by § 3, no. 16, Corollary to Proposition 55.
PROPOSITION 2. Let G be afinite-dimenxianal Lie group and H an integral .mbgroup 13/
G. Thefollowing conditions are equivalent:
(i) H ix closed;
(ii) the tapology on H L: induced by that on G;
(iii) H i: a Lie subgroup of G.
(i) => (iii): this follows from § 1, Propositions 2 (iv) (no. 1) and 14 (iii)
(no. 7).
(iii) 3 (ii): obvious.
(ii) => (i): if the topology on H is induced by that on G, H is closed because
H is complete (§ 1, no. 1, Proposition 1).
PROPOSITION 3. Let G be a Lie group, H an integral subgroup of G, M a nan-empty
connected analytic mam'fiild,fa mapping qfM into C and r E NK. Consider thefillnwing
conditions:
(i) fix afclaxs C' andf(M) C H;
(ii) f(M) C H and]; considered a: a mapping ofM into H, is qfclaxs C';
(111) f u of class 0’, f(M) meek H and the image of TM(M) is contained in
f(m) .L(H)for all m E M.
Then (ii) ¢> (iii) 2 (i). If the tapolagy an H admit: a countable bare, the three can-
dih'anr are equivalent.
(ii) 2 (i) and (ii) 3 (iii): obvious.
(iii) :> (ii): suppose that condition (iii) holds. By Deflemm'able and Analytic
Manifolds, R, 9.2.8,f15 a morphism of class C' of M into the left foliation asso-
ciated with L(H). As M is connected,f(M) C H.
Ifthe topology on H admits a countable base, condition (i) implies thatfis a
mapplng of class C' of M into H (Diflbentiable and Analytic Manifoldr, R,
9.2.8); hence (i) :> (ii).
308
INTEGRAL summours § 6.2
309
I" LIE GROUPS
(iii) 2 (ii): suppose that condition (iii) is satisfied and let H be given the
structure induced by the Lie group structure on G. Let I) be the subalgebra
tangent to H at e. The identity component Hu of H is an integral subgroup of
G such that L(Ho) = I). We show that H = Ho. It suffices to prove the follow-
ing: let I be a connected open subset of K, f a mapping of class C’ of I into C
such thatf(I) C H and A and [1. two points 0“; ifffl) 9 Ho, thenf(p) EH0.
But, for all v e I, (v)(K) Cf(v)l} by definition of I), so that our assertion
follows from Proposition 3.
Remark 4-. If K = R, the integral subgroups of G can also be characterized as
the subgroups which, with the topology induced by that on G, are arcwise eon-
neoted (§ 8, Exercise 4). However, subgroups may be conneoted but not integral
(Commutative Algebra, Chapter VI, §9, Exercise 2).
COROLLARY. Let G be a [mite-dimensional Lie group and H1 and H, two integral
subgroups ofG. The subgroup of G generated by H1 and H2 and the subgroup (H1, H2)
of G are integral subgroups ofG.
The subgroup (G, G) of G is not always closed (§ 9, Exercise 6).
Recall (§ 3, no. 11, Corollary 5 to Proposition 41) that if a is a finite—dimen-
sional algebra, Aut(a) is a Lie subalgebra of GI.(a) and that L(Aut(n)) is the
Lie algebra of derivations of a.
DEFINITION 2. Let a be aflnite-dimemionol Lie algebra. Let Ad(a) or Int(a) denote
the integral subgroup ofAut(a) with Lie algebra ad(n). The elements of this group are
called inner automorphism of a.
By transport ofstructurc, ad(a) is invariant under Aut(a) and hence Int(a) is
normal in Aut(a). By § 4, no. 4, Corollary 1 to Proposition 8 and the fact that
Int(a) is connected, the elements of Int(a) are the finite products of auto-
morphisms of the form exp adx where tee a. In general, Int(a) is not a. Lie
subgroup ofAut(a) (Exercise 14).
3l0
EXPONENTIAL MAPPING § 6.4
4. EXPONENT’AL MAPPING
THEOREM 4. Let G be a Lie group. There exist: one and only one exponential mapping
(JG defined an L(G).
There exist a convex open neighbourhood U ofO in L(G) and an exponential
mapping ofG defined on U. It can be assumed, by choosing U sufficiently small,
that
(1) NO + 75)“) = MMMOJ“)
for a eL(G), A, 7" in K, M, Na, (A + flu in U.
Let a e L(G). There exists an integer n > 0 such that; a e U. Ifm is another
hence (49611))" = (4,61 a))m. There exists an extension 4,; L(G) —> G ofd)
such that Ma) = ($61))“ for aeL(G) and n an integer >0 such that
(()<)<)<))
By Proposition 7, this follows from Proposition 4 of § 4, no. 3, taking 1 = l-
n
PROPOSITION 9. Let G be a temple): Lie group and G' the underlying real Lie gimp.
Then expo = expuu
This follows from Proposition 5 of § 4, no. 3 and the analyticity ofexpa and
expuy.
PROPOSITION 10. Let G and H be Lie granp: and d; a morphism of G into H.
(9 it o em} ='eXPH . L(4>)-
(n) [f G I: an Integral subgroup of H, then exp} = expH | L(G).
312
EXPONENTIAL MAPPING § 6.4-
The two sides of the equality (i) are analytic mappings of L(G) into H
which coincide in a neighbourt of 0 (§ 4, Proposition 8, no. 4) and hence
are equal. Assertion (ii) is a special case of (i).
COROLLARY 1. Let G be a Lie group, G’ a Lie .mbgmup of G and a e L(G). The
following condition: are equivalent:
(i) a e L(G’) ;
(ii) exp(M) E G’for all 16 K and IA] Wendy mall;
(iii) expoia) e Gfot all k E K.
The argument is as in § 4, no. 4, Corollary 1 to Proposition 8.
COROLLARY 2. Let G be a Lie group, H an integral subgroup of G and a E L(G).
Conflder thefollowing conditions:
(i) a e um;
(ii) expG(M) e Hfor all A e K.
Then (i) = (ii). If the topology ofH admit: a countable base, then (i) 4: (ii).
Let i be the canonical injection of H into G. If a E L(H), then
fiPnW) = (CXPG ° L(0)011) = (i ° exPH)(M) GH‘
Hence (i) = (ii). The converse when H has a countable base follows from
Proposition 3.
313
"I LIE GROUPS
w Z4
(x) —— 700 (n + l)! (ad x)" I
The two sides are analytic functions of: and are equal for x sufficiently close
to 0 (§ 4, no. 3, Proposition 5).
Remork. 117(1) . (ad 2:) = exp ad x - 1. We write, by an abuse of notation,
_ exp ad x — 1.
a“) — ad x
COROLLARY. Let G be a complex Lie group and x e L(G). The tangent mapping at x
to expfi has kernel neZ—(D)
® Ker(ad x — 2i1rn).
. . l e’ — l .
The Integral function z~—> 20 m 1", equal to z for z 95 0, admits
as zeros the points of 21112 — (0}, which are all simple zeros. The corollary then
follows from Proposition 12 and the following lemma:
Lemma 2. Let E be a complex Banach space, u an element of .‘2’(E), S the xpectrum ofu
in g (E) (Spectral Theories, Chapter I, § 1, no. 2) and f a holomorphie complex
function on on open neighbourhood Q ofS. Suppose tho!fadmit: in (I only afinite number
of dixtinctrzemx 1,, . . ., z," of multiplicities h“ . . ., h". Then Kerf(u) i: the direct
rum :fthe Ker(u — z,)"lforl S i g n.
(For the definition off(u), see Spectral Theories, Chapter I, §4, no. 8.)
There exists a holomorphie function g on Q, everywhere non-zero, such that
M = (z — zoo. . .(z 2 2mm. mm gong-1(a) = g-'(u>g<u) = 1 and
hence Kerf(u) = Ker ‘1]: (u — 1,)”. Considering Kerf(u) as a C[X]—
module by means of the exnernal law (h, x) n—> h(u)x for h e C[X], x e Kerf(u),
314
APPLICATION TO LINEAR REPRESENTATIONS §6.5
we see that Kerf(a) Is the direct sum of the Ker(u — z,)"t, using Algebra,
Chapter VII, §2, no. 1, Proposition].
315
111 LEE GROUPS
p’ to be
COROLLARY 5. Suppose that p and p’ are finite-dimensionali [for p and
equivalent, it is necmary and .ruflicient that L(p) and L(p’) be equwalenl.
This is a special case of Corollary 4».
316
NORMAL 1mm suncnoues § 6.6
fore a Lie subgroup ofG. Let (I: be the morphism ofG/H into G’ derived from 4)
when passing to the quotient. This morphism is étale; since G is simply con<
nected, 4; is an isomorphism of G/H onto G'.
COROLLARY 1. Let G be afinite—dimemionol simply connected Lie group. Let m, I) be
Lie subalgebrax of L(G) such that L(G) is the semi-direct product of m by I). Let M, H
be the corresponding integral mbalgebrax of G. Then M and H are simply connected Lie
subgroups g"G and, as a Lie group, G i; the semi-direct product of M by H.
By Proposition 14-, H is a normal Lie subgroup of G and the Lie group G/H
is simply connected. Let 1: be the canonical morphism of G onto G/H. There
exists a morphism 6 ofG/H into M such that L(O) is the canonical isomorphism
ofL(G)/L(H) = L(G)/b onto L(M) = m. Then
L(u o 6) = L(n) o L(O) = Iduam,
and hence it o 0 = Ida/H. By no. 1, Corollary 1 to Proposition 1, 0(G/H) = M.
By Proposition 8 of § 1, no. 4, M is a Lie subgroup of G and the Lie group G is
the semi-direct product of M by H.
COROLLARY 2. Let G be a finite—dimensional simply connected Lie group, H a normal
connected Lie subgroup of G and 1c the canonical morphism of G onto G/H.
(i) There exists an analytic mapping 9 of G/H into G such that 1: a p = Ida/3.
(ii) For every mapping p with the properties of (i), the mapping (h, m) »—> hp(m) of
H x (G/H) into G is an isomorphism of analytic manzfoldx.
(iii) H and G/H are simply connected.
- Let n = dim G — dim H. The corollary is obvious for n = 0. We argue by
induction on n.
Suppose that there exists an ideal of L(G) containing L(H) distinct from
L(G) and L(H). Let H’ be the corresponding connected Lie subgroup of G.
Let 1:1: G —> G/H’ and 1:2: H’ —> H’IH be the canonical morphisms. By the
induction hypothesis, there exist analytic mappings pl: G/H’ —> G,
p]: H’/H —> H’ such that 1:, 9 (a1 = Idem], 1:, 0 p2 = Idflymi Let
1:5: G/H —> G/H'
be the canonical morphism. If x e G/H and y is a representative of x in C, y
and 91(n3(x)) have the same canonical image modulo H’ and hence
x'1n(pl(1ra(x))) e H'IH. Let
Pi“) = P;(Wa(x))92("(91("a(x)))"’0 5G-
Clearly p is an analytic mapping of G/H into C and
"(900) = "(tumor)”W2(92(W(Pi("a(x)))'50)
= "(Pi(1ra(*)))1r(91("s(x)))"I = x~
If new the only ideals of L(G) containing L(H) are L(G) and L(H), the Lie
317
I" LIE GROUPS
318
LAWS or INFINITESIMAI. OPERATION TO LAWS or opens-non §6.8
.n—>f1(:, x), :»—>f3(:,x) of G into X are continuous. Suppose that there exist: a
neighbourhood V of {e} X X in G x X Such that f1 andf, coincide on V. Then
f1 = f:-
Let x e X and A be the set ofg 6G such Ihatf,(g, x) =f,(g, x). Then A is
closed in G. On the other hand, let geA; we write y =f1(g, x) =f3(g, x).
There exists a neighbourhood U of e in G such thatf1 (t, y) = f,(t, y) for t e U,
in other words such thatf,(t', x) : f2(t’, x) for t’ E Ug (resp. gU). Hence A is
open in G and therefore A = G.
319
Ill LIE GROUPS
an open
compact, V can be assumed to be of the form U x X, where U is
neighbourhood of e in G. It then suflices to apply Lemma 5.
PROPOSITION 17. Let A be the set of 266 such that —1t < 1(1) < n and A’
the set ty’ 2 e C which are not real <0. let E be a complete normable space over C and
A (resp. A’) the set if): e Y(E) whose spectrum Sp(x) is contained in A (resp. A’). V
Then A (resp. A’) is an open subset of 2(E) (resp. GL(E)) and the mappings
expzA—>A’ and log:A'—>A (Spectral Theories, Chapter I, §4, no. 9) are
inverse isomorphisms of analytic manifolds.
This follows from Spectral Theories, Chapter I, §4, Proposition 10 and no. 9.
THEOREM 6. Let E be a real or complex Hilbert space and U the unitary group of E.
(i) The set H of Hermitian elements of 5?(E) is, with the real normed space struc-
ture, a closed oector subspace of 3(E) admitting a topological supplement.
(ii) The set H’ of elements 20 qL(E) is a real analytic submonifold of GL(E).
(iii) The restriction to H of the mapping exp is an isomorphism of real analytic
manifolds of H onto H’.
(iv) The mapping (h, u) »—> (exp h)u of H x U into GL(E) is an isomoephism
of real analytic manifolds.
Recall that, if x G Y(E), x“ denotes the adjoint of 1:. Let H1 be the set of
x6305) such that x“ = —x. The formula x = fix + x‘) + fix — 3*)
proves that, with its real normed space structure, 3’03.) is the topological
direct sum of H and H1, whence (i).
Suppose that K = C. In the notation of Proposition 17, H’ is the set of
h e H n A’ such that Sp h C Rt. As exp(R) = Rf“ (ii) and (iii) follow fiom
Proposition 17 and S , tral Theories, Chapter I, §4, Proposition 8 and §6,
no. 5. The mappinggh, u) >—> y = (exp h)u of H x U into GL(E) is bijec-
rive by Spectral Theories, Chapter I, §6, Proposition 15. It is real analytic by
the above. The mapping yr—>h = flog(yy") is real analytic and so also is
the mapping y i—> u = (exp h) “y. Hence (iv).
Suppose that K = R. Let E be the complexified Hilbert space of E and J
the mapping E + i-q H E — in (for E, 1; in E) ofE into E. Let H, H’, U de-
note the sets defined for E as were H, H', U for E. Then H (resp. H’, U) is
identified with the set of x E H (resp. H', U) such that ya“ = x. Properties
(ii), (iii) and (iv) then follow easily from (i) and the analogous properties in
the complex case.
PROPOSITION 18. Let E be a complete normable space over C, v e .Y(E) and g = exp 0.
Suppose that Sp(o) contains none of the points 2i1rn with n E Z — {0). Then, for all
x E E, the conditions on = 0 and gx = x are equivalent.
This follows from Lemma 2 of no. 4, applied to the function 2 »—> e‘ — l.
320
EXPONENTIAL MAPPING [N THE LINEAR GROUP § 6.9
COROLLARY 1. Let E be a rumplete mmmble :[mec aver C and F the xpaee of can-
timwus n-linear mapping: 4/ E" into E. For all u e 2(E), let 0(a) be the element qf
$(F) defined by
Then 11(1)) = ‘2; 1/, and the u‘ are pairwise permutable. Let A be the total
closed subalgebra of 2’(F) generated by the 11,; it is commutative (Spectral
Theories, Chapter I, §l, 1104 4) and Spgmu' = SpA v’ C ‘20 Sp 1), (Spectral
Theories, Chapter I, §3, Proposition 3 (ii)). Now, if A EC is such that u — A
is invertible, clearly the v. — A, are invertible and hence Sp 0‘ C Spy for
all 1'.
321
III LIE GROUPS
PROPOSITION 19. Let E be a tomplete normable space over C, v e 56’(E) and g = exp v.
Suppose that every zeSp v satisfies —n < 1(1) < 1:. Then, for every closed vector
subspace E1 of E, the conditions v(E’) C E’ and g(E’) = E’ are equivalent.
The condition o(E’) C E’ implies g(E') C E’ and g“(E’) C E’ and hence
g(E’) = E’. Suppose that g(E’) = E’. We use the notation A, A' of Proposi-
tion 17. Since Sp v is a compact subset of A, there exists a compact rectangle
Q = (a, b] x [a', b’] such that Sp o C QC A. The set A — Qis connected.
Hence Sp g C exp Q C A’, the set exp Qis compact and the set A’ — exp Q
is connected. The closure of the latter contains ]—00, 0] and hence
(A'—ep) U]—ao,0] : C—ep
is connected. Then exp Qis polynomially convex (Spectral Theories, Chapter I,
§ 3, Corollary 2 to Proposition 9) and hence the fiinction log, defined on A’,
is a limit in 0(exp Q) of polynomial functions (Spectral Theories, Chapter I,
§4, Proposition 3). Hence 1; = log g is the limit in $(E) of elements of the
form P(g), where P is a polynomial (Spectral Theories, Chapter I, §4, Theorem
3). As P(g)(E‘) C E', it follows that v(E') C E'.
COROLLARY. Let E be a complete normed space over C, o e $(E) and g = exp v.
Suppose that every 2 eSp v satisfies —1—2r < 1(2) < 7—5: Then, for every dosed vector
subspace M of 3 (E), the conditions gMg" = M and [17, M] C M are equivalent.
Let F = $(E), g' be the map}:oingfv—>gfg'l of F into F and v’ be the
mapping f l—> [v,f] of F into F. Then g’ = expv' (no. 4, Corollary 3 to
Proposition 10 and §3, no. 11, Corollary 1 to Proposition 41). Lemma 6
proves that — 1: < 1(1) < 1: for all z e Sp 11’. It then suffices to apply Proposi-
tion 19.
322
CDMPLEXIFICATION or A “NIKE-DIMENSIONAL REAL 1m GROUP § 6.10
. (iii) 1" it injecting, p’ i: .mjem’w, KerUz') = In_1(i’) and Ihefallowing diagram
15 mmmutatwa
A —‘> B .3.) C
(4-) r a in,
A’ LB’L’) 0.
(iv) [fbeBanda’eAfl [hm
.4:(’7)i'('1')g(b)‘1 = i'(‘»(”)a')-
(i) For an 12 in A, we have, in B”,
(f(01_ 1): z'(111))(f(42 ); 1.(012)) = (flail) (“(111)1'011», “405012”
=(f(flf )f (11“: an" ), “than”
= (f((”x”2)—l i(“1“2))
and hence a» (f(a“),”i(a)) is a homomorphism I: ofA into B” Let a GA,
a eA’, beB; then,inB’,
W4)b 1 = ”(VIM—1 = (“(b)f(a'1))(bab")
=f(ba“b‘1)(bab'1) = h(bab‘1)
a’h(a)a"1 = a’f(a")aa"‘ = aff(a")(m(a)a"‘)a
= «7(a“)f(u)a"V(a")a = 1'01)
and hence h(A) = D is normal in 13'.
(ii) For a e A,
(1’ ° 001(4)) = P(q(f(a")a)) = [1(0) = 6
and hence p o q is trivial on D.
(iii) Let a’ e A’ be such that i’(a’) = a; then a’ e D and hence there
exists a e A such that a' =f(a“)a; this implies a = 2, whence a’ = z; thus
1" is injective. As [1 and q are surjective, p’ is suzjectivc.
Let r denote the canonical morphism of B" onto B’. Let a' e A’, I; 53 and
b’ = r(a’b). If b’ e Im(i'), there exists a; e A’ such that b’ = r(a§); then there
exists a GA such that a’b = a;f(a“)a, whence b = :1 EA and
NV) =1’(q(a'b)) = [’(b) = I;
thus, Im(i') C Ker(p’). We preserve the notation a’, I], b’ but assume that
b’ e KcrUz’) ; then 2 = p’(b’) = p(q(a’b)) = p(b) and hence b E A, whence
0' = '(a'f(")f(b")”) = '(fl'f(b))€1m(i');
thus Ker([z') C Im(i’).
If a e A, then
If h E B, then
P'(g(b)) = M”)
and hence diagram (4) is commutative.
(iv) Let h e B, n' E A’. Then
g(b)i’(4’)t'(’>)‘1 = 700443701)" = ’(WV‘)
= '(NU’W) = 1"(@0003
PRoposmou 20 Let G be 11 finite—dimensional real Lie group.
(i) There exist a complex Lie group G and an R-nnnlytie morphism y of G into G
with the following properties: for every eomplex Lie group H and every R—onolyh'e
morphism 4) of G into H, there exists one and only one C-analylw morphirm q, of G
into H such that (11:4, 0 Y
(ii) If (day) has th_e some properties a: (G, 7), there exist: one and only one
isomorphism 6 ofG onto G’ such tha! 9 0 Y = y'
(iii) The C-linear mapping of L(G) ® G into L(G) which extend: L(Y) is our-
jeetioe; in particular dimc(G) < dimR(G).
Assertion (ii) is obvious. We prove the existence of an ordered pair (G, 7)
with properties (i) and (iii)
(21) Suppose first that G is connected. Let g =.L(G), g: = g ®R C be the
complexification of g, S (resp. S’) the simply connected real (resp. complex)
Lie group with Lie algebra g (rap. 9.) and o- the unique R—analytic morphism
of S into S’ such that L(o-) is the canonical injection of g into 99. Let 1: be the
unique R-analytic morphism of S onto G such that L(x) = Idua) and F =
Ker 1:4
1 -———>S
’1
1
6.
G ———> G¢
For every complex Lie group H and every R-analytic morphism (I) of G
into H, L(<Ii):g—>L(H) has a unique C-linmr extension to Sc and this
artension is of the form L(<{>*), where d)’ is a (fl-analytic morphism of S’ into
H. Then
L(i> ° 7') = 1-(4’) ° N") = L015) = L(<l>‘) ° 140) = LW‘ ° 6)
and hence 4’ o 11: = tb‘ n 6. Therefore ¢‘(5(F)) = d>(1r(F)) = {e}, whence
a(F) C Ker <1”.
Let P be the intersection of the Ker 42* for variable (I). This 15_a normal Lie
subgroup of S’ (no. 2, Corollary 3 to Proposition 1). LetG =S'[P and
324
courmxrrrmrrou or “NM-DIMENSIONAL REAL LIE GROUP § 6.10
326
PmGE mom us ALGEBRAS TO LIE onoups § 7.1
PROPOSITION 1. Let G be a Lie group gem with identity element e. There exists 4
fiendamental system of open neighbourhoods of e in G cam'sting of Lie subgroups of G.
Let L(G) be given a norm compatible with its topology and such that
‘|[(xG,y]|l < ”x” "y” for all x, y in L(G). Let G, be the Lie group defined by
).By §4, no. 2, Theorem 2, G and G, are locally isomorphic. Then it
sugces to apply § 4», no 2, Lemma 3 (iii)
THEOREM 1. Let L be a complete normable Lie algebra. There exists a Lie group G
such that L(G) is isomorphic to L. Two such groups are locally isomorphic.
The first assertion has been proved in §4, no. 2, Lemma 3, The second is
a. special case of § 4, no. 2, Theorem 2.
THEOREM 2. Let G be a Lie group and b a Lie subalgebra of L(G) admitting a topo-
logical supplement. There exists a Lie subgroup H of G such that L(H) = b. If H1
and H, are Lie subgroups y” C such that L(Hl) = L(Hz) = b, then H1 n H2 is
open in H1 and H2.
The first assertion follows from Proposition 1 and {$4, no. 2, Theorem 3.
The second is a special case of § 4, no. 2, Theorem 3.
THEOREM 3. Let G and H be Lie groups and h a continuous morphim of L(G) into
L(H).
(i) There exist an open subgroup G’ of G and a Lie group morphism b of G’ into
H such that h = L((b).
(ii) Let G“ G, be open subgroups ofG and (I). a morphism «MG, into H such that
h—- L((b‘). Then (I), and b2 coincide on an open subgroup ofG
By Proposition 1, this follows from § 4, no. 1, Theorem 1.
327
III LIE GROUPS
2. EXPONENTIAL MAPPINGS
PROPOSITION 3. Let G be a Lie group. There exists an exponential mapping 4; of G
with the following properties: -.
(i) (I: is defined on an open subgroup U J the adahtwe group L(G) ; . _
(ii) ¢(U) is an open subgroup ofG and (I; air an uomorphum qfthe analytic manJold
U onto the analytic mom'fild MU);
(iii) ¢(nx) = 4>(x)"for ollx EU andall n 52.
Let L(G) be given a norm compatible with its topology and such that.
[I Dr, y] [I < ”x“ "y“ for x, y in L(G). Let G, be the Lie group defined by L(G).
Let 4; = IdG‘, which is an exponential mapping of G1. For all p. > 0, let L"
be the set of x e L(G) such that ||x|| < It. Then, for y. sufficiently small, L,
is an open subgroup of the additive group L(G), ML”) is an open subgroup
of G1 (§ 4-, no. 2, Lemma 3), ML“ is an isomorphism of analytic manifolds
of L“ onto ML”) and Mm) = 440:)” for all x e L“ and all n E Z. The L“ form
a fundamental system of neighbourhoods of O in L(G). By Theorem 1, there
exist p. and an open subgroup G' of G such that ML“) and G’ are isomorphic,
whence the proposition.
PROPOSITION 4. Let G be a Lie group and (I; an injective exponential mapping of G.
Suppose thatp > O. For all x, y in L(G),
(1) x + y = ”5-3“ [2 ”11>“(d>(o"x)¢(o"y))
(2) [w] = 11mm P""<l>"(¢(p"x)<l>(t’"y)<l>(-P'x)<l>(-!I"y))-
These are Special 0333 of Proposition 4 of § 4, no. 3.
3. STANDARD GROUPS?
If S(Xl, X2, . . ., X“) is a formal power series with coefficients In A, then,
for all x), . . .,x, in m, the series S(x,,x2,.. ,x,) is convergent. More pre-
cisely, m x m x ~ ~ - x m is contained'm the domain of absolute convergence
of S (Dferenttable and Analytic Manifolds, R, 4- l 3).
DEFINITION 1. Let r be an integer >0. A standard group of dimention 7 over K is
a Lie group G with thefollowing propertier:
(i) the underlying analytic manfold of G u m x m x - - - x m (rfaotorx) ,
(ii) there exists a formal power series F in 21 variables with coefioient: in A',
without ootant term, ouch that x._1/—= F(x, 1]) for all x, y in G.
Then 0.0 = 0 and hence the identity element of G is the origin of
m x m - - - x m.
T The results of nos. 3 and 4 and their proofs remain true when the character-
istic of K is > 0.
328
STANDARD GROUPS § 7.3
L(G) will be identified with K'. By § 5, formula (13), the constants of struc-
ture of L(G) with respect to the canonical basis belong to A. We shall need,
in the same proof to consider the elements of m X - - - X m, now as elements
of G, now as elements of L(G).
where the series on the right hand side are convergent for x, y in G (§ 5,
no. 1). Let leK" and let the group law be transported from G to G’ = XG
by the homothety of ratio A. For s“, y’ in G’, the product x’.y’ and the in-
verse x’l‘“ evaluated in G' have coordinates
thatiscgmeAanddéeA;andontheotherhandG' 3m x m x x m.
Then m x m x x m isanopen subgroup ofG'andisastandard group.
329
Ill LIE GROUPS
330
POWERS 1N STANDARD GROUPS § 7.5
(fx 6 G(a) andy e C(b), then xi‘“.y['“.xiy, x.y.x““y"” and [x, y] baking to
G(c) and an congruent moduli) C(c').
By § 5, no. 2, Proposition 1, there exist cue A’ such that
-ll [—1] _ _
K‘ .1] .xy [x,y]—Mglncflmfl.
If x = 0 or y = 0, then xI'".y[‘“.x.y — [x,y] = 0; hence ”on = can = 0.
On the other hand, the conditions
xEGO‘): yEGU’), I'll > 1, [BI > 1, [“I + ”3| 9 3
imply
cmpx'y“ EG(a3b + ab“) c C(c’)
and hence xi'u.y“"x.y - [x,y] E G(c’). We see similarly that
x.y.x"“.y"“ -— [x,y] eG(c’).
Finally, by § 5, formula (13), [x,y] eG(ab) C C(c).
PROPOSITION 7. (i) The family (C(01)) i: a centralfiltmtian on G (Chapter II,
§4, no. 4-, Definition 2).
(ii) For AeRfi, G(a,_) = {xeG]m(x) 2 7r}, G(a,f)= {xeG|m(x) > )1}.
(ii) is obvious. We prove (i). Clearly cm): ('1 G(a )andG= AU Gm)
On the other hand, if x 6 0(1).) and y e G(au), then
zl‘n-yhu-“JEGU‘mQ
by Proposition 6 applied with a = ax, b = a“, c = c' = a“...
By Chapter II, §4, no. 4, we can form the group gr(G) associated with
the group G, with the central filtration (0(a)). Writing GA = G(a,.)/G(a,f)
for all 1 > 0, we obtain gr(G) = $190 Gk Recall (lac. n'L, Proposition 1)
that the commutator in G allows us to define a bracket in gr(G) with which
gr(G) is a Lie algebra, as follows: if? E G,‘ and? e G”, choose a representative
:1: of x in G(u,_) and a representative y of? in 0(0“); then [if] is the class of
x“ 11 .y“ 1‘ .x yeG(aMu) in GM". By Proposition 6, applied with a = a,”
b = a“, c = «Mu, c’ = ax”, we see that [fey] is also the class of [x,y] in
GM“. Thus, when G 15 considered as a Lie subalgebra of L(G)= K', filtered
by the G(ah), the associated graded Lie algebra (Chapter II, §4, no. 3) is
equal to gr(G).
331
'fll LIE GROUPS
x’=px+ 2 tax“
lulu
where c, E A' for all (X. Even for proving (i) it can be assumed that 110) = 1.
Then if m(x) 2 1, it follows that (”0:”) 2 m(x) + l and hence m(x"') tends
to +00 as It tends to +00; this proves (i). As (6) is divisible by p for
332
LOGARITHMIC MAPPING § 7.6
l S 1' < I — 1, Proposition 2 of§5, no. 3, proves that on. epA' for
2 s |u| <12 — 1 '
and hence
More“) > toque) = «3(3) + 1 for2 < M sp — 1.
On the other hand, if [an] 2 p, «(59%) > pm(:t) and park) > 01(x) + 1 if
l
o)(x) > P _ l « This proves (ii).
G. LOGARITHMIC MAPPING
Lemma 1. Suppose that p aé 0. let G be a Lie group, Gl an open subgroup of G which
is isomorphic to a standard group and x e G. Thefillowing condition: are equivalent:
(i) there exist: a power tfx which belongs to G1;
(ii) there exist: a strictly increasing sequence (n.) of integer: such that x": tend: to
e a: i tend: to +00.
(ii) = (i): obvious.
(i) =- (ii): suppose that y = 5:"e By Proposition 9 (i) of no. 5, y"
tends to e as n tends to +00, in other words rm” tends to e as It tends to +00.
PROPOSITION 10. Suppose that p aé 0. Let G be a finite-dimensional Lie group. Let
G, be the set of x e Gfir which there exist: a strictly inereming sequence (71.) qfintegm'
ruch that x"! tend: to e a: i tend: to +00.
(i) G, Lv open in G.
(ii) There exists one and only one mapping 4; of G, into L(G) with the following
properties:
(a) ¢(x") = n¢(x)for all x e G, and all 7152;
(b) there exist: an open neighbourhood V of e in G Mh that 411V i: the inverse
mapping of an injective exponential mapping.
(iii) The mapping 4; ix analytic.
There exists an open subgroup of G which is isomorphic to a standard
group (no. 3, Theorem 4-). Assertion (i) then follows from Lemma 1.
Let U be an open subgroup of L(G) and <1):U—><]>(U) an exponential
mapping of G with the properties of Proposition 3 of no. 2. U can be assumed
to be small enough for MU) C G,. Let xeG,. There exists meZ —{0}
such that x’" E ¢(U). The element-”11¢”(xm) does not depend on the choice
of m. For let m’ e Z be such that a“ e (NU). Then an“ e (MU) and
WWW”) = <l>“(x"‘"") = WWW),
whence our assertion. Let Her) = %¢"(fl)- Then MMU) = d)". On the
other hand, if n e Z, then
Hence I]; has properties (a) and (b) of the proposition. In a neighbourhood of
l
x, L]; is composed of the mappings x I—>:c’“, y» ¢'1(y) and z I»; 2; hence kl;
is analytic on 6,.
Finally, let da’ be a mapping of G, into L(G) and V' a neighbourhood of
e in G, such that ¢’(z") = nil/(x) for xEG, and a El and such that ¢’|V’
is the inverse mapping of an injective exponential mapping. Then i]; and 4/
coincide on a neighbourhood W of e. If are G,, there exists n 52 such that
x" s W. Then
"4"(x) = WI") = 440‘") = MN")
and hence x]; = 44’.
DEFINITION 2. The mapping 4! qmpoxition 10 is called the logarithmic mapping of
of G and is denoted by logo or simply log.
PROPOSITION ll. Sufioxe that p ;é 0. Let x, g be two permutable element: of G,.
Then xy E G, and log(xy) = log 1: + logy.
The fact that any E G, follows from Lemma 1. Let U be an open subgroup
of the additive group L(G) and (I): U —>¢(U) an exponential mapping of G
with the properties of Proposition 3 of no. 2; U can be assumed to be small
enough for loglMU) to be the invexse mapping 'of (b. For n E Z — {O} suitably
chosen, 1" E¢(U), y"e<]>(U). Let u = log :6", v = logy”, whence x” = (Ma),
y” = (My). By formula (2), [u, v] = 0. The Hausdorff formula proves then
that (1)001 + 12)) = 4>(M)4>(M) for N sufficiently small; hence, for every
sufficiently large integer 1',
MN“ + 0)) = ll>(1"u)4>(#”)
that is
["008 at” +10%") = 1°g(x""y’“")
or
np‘aogx + logy) = "1" 1°g(ry)-
PROPOSITION 12. Suppose that p 56 0. Let xeG,. The following conditions are
equivalent.-
(i) logx = 0;
(ii) x is offinite order in G.
If there exists an integer n > 0 such that x" = e, it follows that
nlogx = logx" = 0,
whence logx = 0. If log): = 0, let V be a neighbourhood of e in G, such
that loglV is the inverse mapping of an injective exponential mapping. There
exists an integer n > 0 such that x"eV; the equality log x" = 0 implies
z“ = e.
334
momma MAPPING § 7.6
m — l 1
By Chapter II, §8, no. 1, Lemma 1, ”(m!) g p _ 1- If m(x) >§_—1: we
see that m(x) — 0(m!) tends to +00 with m, whence
““2!“ l xl " s W W
l
336
CONTINUOUS uonpmsus § 8.!
337
In LIE GROUPS
defined by writing h(x) = %g(M) for all A such that MeV. Clearly I: is
continuous. We show that
(3) (x E L(G) and A E K) => 1104:) = M(x).
Let k’eK“ be such that )1e and A'MEV. Then
1 , 1 o 1 ,
110%) = 730 M) = ,7 M“) = Riga x) = M0!)-
Let x, y be in L(G). Then, by Proposition 4 of § 4, no. 3,
ll(it) + My) =A lim A‘“I"1(it(>~h(x))klt(>~h(y)))
= A6i133“x-‘o-1(¢(h<m)¢(h(zy))).
For |A| sufficiently small, Ax eV and Ag 6 V and hence the above expression
is equal to
”35;.” rlt"(f(4>(kx))f(4>(ly)))
= “£10 PW“ °f)(4>(M)<I>(7\y))
= “£13” k“§(¢"(4>(>\~)4>(1y)))
= ”£132“,h(’\“¢“(¢(l~)¢(7‘y)))
= M‘siimw l"<l>"(¢(M)¢(7‘y)))
= h(x + y).
Thus [1 is continuous and linear, hence g = h|V is analytic, hencef is analytic
on ¢(V) and hencef is analytic (§ 1, no. 10).
COROLLARY 1. Let G be a topological group. There exist: on G at most one analytic
manifold :tructurz over R (rcsp. Qp) compatible with the topological groufl :tmclur:
on G.
338
com-muons MORPnIsns § 8.1
339
ll! LIE GROUPS
340
CLOSED suncnoups § 8.2
Hence the mappingf: A »—> exp Ay, where Ag E }V,, mks its values in H and
(Tof) (l) = y. Hence y e I), which is absurd. The theorem is thus established in
the case of R.
HnV={e[f.(x) =f,(e)firollieI}.
(i) IfK = C,Hi.raLiembgrouqG.
(ii) IfK i: afim'te extension #9,, and I ixfim'le, H i: a Lie subgroup ofG.
(i) Suppose that K = C. We consider G as a real Lie group. Then H is a
real Lie subgroup of G (Theorem 2). Let a e L(H). There exists a connected
open neighbourhood W of 0 in C such that exp Aa E V for all A E W. Let i e I.
Then fi(exp Aa) =fi(e) if AERn W. Hence fl(exp Aa) =f.(e) if AEW by
analytic continuation. Thus, exp Aa e H for A e W and therefore pa 6 L(H) for
all pt E C. Therefore H is a Lie subgroup of the complex Lie group G (§ 4, no. 2,
Proposition 2).
(ii) Suppose that K is a finite extension of 1,2,. We consider G as a Lie group
over Qp. It is finite-dimensional and Theorem 2 implies that H is a p-adic Lie
subgroup ofG. Since I is finite, 1:! MI is a manifold and it can be assumed that
the family (fi) reduces to a single mappingf. Let a e L(G). Let (I) be an expo-
nential mapping of G. Then f(¢(Ao)) = f(e) for A E Q, and [AI sufficiently
341
ll] LIE GROUPS
small. Since f is K—analytie, it follows that f((1) (NO) = f(e) for A e K and [M
sufficiently small. Hence (MM) EH for A e K and |).[ suffiicently small and
therefore pm 6 L(H) for all p. E K. The proof is completed as in (i).
Corollary 2 (ii) is no longer true if we omit the hypothesis that I is finite.
342
oomuu-m-rons IN A LIE amour § 9.2
343
m LIE GROUPS
344
comm'm'rons IN A LIE GROUP § 9.2
[L(A),L(B)] = L(C)~
By Proposition 3, the Lie subalgebra tangent to (A’, B’) at 2 contains L(C).
Hence (A’, B') contains a Lie subgroup germ of G with Lie algebra L(C).
Therefore (A’, B’) is an open subgroup of C.
COROLLARY. Support that K it ultramelric. Let G b: a finite-dimensional Lie group
with Lie algzbm 9. There exist: an open :ubgroup G, of G such that, fir all i, D‘Go
(resp. CG“) {4' a Lie subgroup of G with Li: algzbm 9‘9 (resp. ($9).
(a) By Proposition 3 applied inductively, for every op'en subgroup G1 of
G and for all i, D‘Gl contains a Lie subgroup germ of G with Lie algebra
9‘9.
(b) Let G' be an open subgroup of G such that, for :< n, D‘G is a Lie
subgroup of G with Lie algebra 9‘9 By Proposition 6, there exist open sub-
groups H1, H2 of D"G’, such that (H1, H2) is a Lie subgroup with Lie algebra
9"“ 1g. Let G” be an open subgroup ofG' small enough for D"G' C H, A 1-1;.
Then D"“G” C (H1, Hz). The relations
DDG" C DUG’, D‘G” C D‘G’, . . . ,D"G" C D”G’, D"“G” C (H1, H,)
prove, using (a), that DO" 15, for l< n + l, a Lie subgroup of G with Lie
algebra 9‘9.
(c) There exists an integer p such that 9V9 = 9””9 : - - ~. By the above,
there exists an open subgroup Go of G such that D'Go is, for i < P, a Lie
subgroup of G with Lie algebra 9‘9. But, by (a), the same assertion remains
true fori > p since D’G0 D D‘Go fori > p.
(d) The argument is similar for the Cl.
345
Ill LIE GROUPS
3. CENTRALIZERS
Recall that two elements x, 3/ of a group are called permutable if (x, y) = e,
or (Int x)y = y, or (Int g)z = x; and that two elements a, II of a Lie algebra
are called permutable if [11, b] = 0, or (ad a) .b = 0, or (ad b) .a = 0, Let G
be a Lie group, x e G, a e L(G) ; x and a are called permutable if (Ad x) . a = a,
that is ifxa = are in T(G).
Let G be a Lie group, 9 its Lie algebra, A a subset of G and a a subset of 9.
Let ZG(A) (resp. ZG(a)) denote the set of elements of G which are permutable
with all the elements of A (resp. a). It is a closed subgroup of G. Let MA)
(rap. 3am) denote the set of elements of g which are permutable with all
the elements of A (resp. a). It is a closed Lie subalgebra of g.
PROPOSITION 7. Let G be a finik—dimemimal Lie group, 9 it: Lie algebra and a a
subset of 9. Then Za(a) i: a Lie subgroup qf G with Lie algebra 3,01).
This follows fmm § 3, Proposition 44 and Corollary 2 to Proposition 39.
PROPOSITION 8. Let G be a finite-dimensional real or complex Lie group, g it; Lie
algebra and A a subret bf G. Then ZG(A) i: a Lie subgroup qf G with Lie algebra
bn(A)'
Suppose that A consists of a single point a. Then 25(A) is the set of fixed
points of Int a; hence ZG(A) is a Lie subgroup of G and L(ZG(A)) is the set
of fixed points ofAd 0, that is MA) (§ 3, no. 8, Corollary 1 to Proposition 29).
The general case follows by means of § 6, no. 2, Corollary 3 to Proposition 1.
PROPOSITION 9. Le! G be a finite-dimenxiannl real 0r temple: Lie grbup, 9 it: Lie
algebra, A an integral subgroup of G and a = L(A). Then ZG(A) = Z501) a
MA) = M“) and ZG(A) is a Lie subgroup of G with Lie algebra Ma).
Let x e G. Then
’1 5 MA) ¢A C 29({u})
¢ a c L(Zu({"})) (§ 6, Corollary 2 to Proposition 3)
¢ ‘1 C a.((u}) (Proposition 7)
o u e 33(a)
and hence MA) = Ma). The last assertion then follows from Proposition 7
or Proposition 8.
346
NILPOTENT LIE GROUPS §9.5
4. NORMALIZERS
Let G be a Lie group, 9 its Lie algebra, A a subset of G and a a subset of g.
In this section, NG(A) will denote the set ofgeG such that gAg‘1 = A. It
is a subgroup of G, which is closed if A is closed. nu(a) will denote the set of
xe 9 such that [x, a] C 0 (cf. Chapter I, § 1, no. 4). It is a subalgebra of g,
which is closed if a is closed. N501) will denote the set of geG such that
gag‘1 = a.
PROPOSITION 10. Let G be a flaw-dimensional Lie group, 9 its Lie algebra and a
a vector subspace of 9. 77m: NG(a) is a Lie subgroup of G M Lie algebra n,(a).
This follows from § 3, Proposition 44 and Corollary 1 to Proposition 39.
Pkovosmmz 11. Let G be a finite-dimensional real or complex Lie group, 3 its Lie
algebra, A an integral subgroup of G and a = L(A). Then NG(A) = N503) and
NG(A) is a Lie subgroup if G containing A, with Lie algebra n,(a).
The equality NG(A) = NG(a) follows from § 6, no. 2, Corollary 2 to Proposi-
tion 3. By Proposition 10, NG(A) is then a Lie subgroup of G with Lie algebra
“‘01). Hence NG(A) is closed. As Nu(A) 3 A, NG(A) D A.
COROLLARY. If a = n,(a), A is a Lie subgmup of G and is the identity component
afNo<A).
This identity component is a Lie subgroup with Lie algebra n‘(a) (Proposi-
tion 11) and hence is equal to A by § 6, no. 2, Theorem 2 (i).
5. NILPOTENT LE GROUPS
PROPOSITION 12. Let G be afinile-dimensional Lie gmup. For, L(G) to be nilpoteut,
it is necessary and sufieimt that G have a nilpateut open subgroup.
Suppose that G has a nilpotent open subgroup Go. By the Corollaries to
Propositions 4- and 6, no. 2, W‘L(G0) = {0} for sufficiently large 1'. Hence
L(Gn) = L(G) is nilpotent.
Suppose that L(G) is nilpotent. If K = R or C, the identity component
Go of G is nilpotent by the Corollary to Proposition 4, no. 2, and G0 is open
in G. If K is ultrametric, the Corollary to Proposition 6, no. 2, proves that
there exist an open subgroup G, of G, an integer i > 0 and a neighbourhood
V of e in G such that C‘Gl n V = {3}. Then, if G“ is a sufficiently small sub-
group of G1, C‘Gu C V, hence C‘Go = {e} and Go is nilpotent.
Let g be a nilpotent Lie algebra. The Hausdorfl series H(X,Y) corres-
ponding to g has only a finite number of non-zero terms and we know (Chap-
ter II, § 6, no. 5, Remark 3) that the law ofcomposition (x, y) »—> H(x, y) defines
a group structure on 5;. Suppose further that g is complete and normable.
Clearly the law H is a continuous-polynomial (Dah‘able and Analytic Mom'-
folds, R, Appendix). Hence 9, together with the law H, is a Lie group G,
34-7
111 LEE onoups
348
NILPOTENT LIE GROUPS § 9.5
By (2) and (3), 0(§(x)) = 0(x) and hence y(x) e 9. Finally we write
u»—>expu= 2—:
that is
Tr(a(b—l))=Tr(ab—-l)—Tr(a—l)=0—0=O
for ab — 1 and a — l are nilpotent. As the vector subspace of $(V) gener-
ated by G is J(V) (Algebra, Chapter VIII, § 4, Corollary 1 to Proposition 2),
Tr(u(b — 1)) = 0 for all u e$(V) and hence b = 1. Thus G = {1}.
(b) We pass now to the general case. Let It be an algebraic closure of k,
V = V®k£ and G C GL(V) the set ofa® l for aeG. Let W (165?“ W’)
be the set of elements of V (resp. V) invariant under G (resp. G). Then
W’ =W®kk forW = 9Q}Ker(g— 1) and W' =‘Q}Ker(g— 1) @1. If
V1 denotes a minimal element in the set of non-zero vector subspaces of V
which are stable under 6, then V1 C W’ by part (a) of the proof; hence
W aé {0}, which proves (i).
(c) By induction on dim V, it follows from (i) that there exists an increas-
ing sequence (V1,V2, . . ., V") of vector subspaces of V which are stable
under G such that V,I = V and the automorphism group of V./V,_1 canonic-
ally derived from G reduces to {1} for all i (we make the convention that
V, = {0} for r < 0), This implies first (ii) and consequently (iii) (Chapter
II, §4, no. 6, Remark).
351
ll'l LIE GROUPS
at.“ E L(M) n g“
352
sownu: LIE (mom's § 9.6
353
m LIE GROUPS
sequence (V1, V3, ‘ . ., V,) of vector subspaces of V which are stable under
G such that dimVHllV. = 1fori< h and V,I = V; hence (it).
354
SEMI-SIMPLE LIE GROUPS §9.8
8. SEMI-SIMPLE LIE GROUPS
PROPOSITION 26. Let G be a finite-dimemiml connected real or complex Lie group.
Thefillcwing condition: are equivalent:
(i) L(G) is semi-simple;
(ii) the radical of G is {e};
(iii) every normal commutative integral subgroup of G is equal to {e}.
Condition (ii) means that the radical of L(G) is {0} and hence (i) c (ii)
(Chapter I, § 6, Theorem l). The equivalence of (i) and (iii) follows from § 6,
no. 6, Proposition 14.
DEFINITION 2. A connected real or complex Lie group is called semi-simple if it is
finite-dimem'onal and satisfies the conditions of Proposition 26.
Remark 1. Let G be a finite-dimensional connected real or complex Lie group.
If G is not semi-simple, G has a commutative connected Lie subgroup G’
which is invariant under every continuous automorphism, such that G’ 75 {e}.
For let u be the largest uilpotcnt ideal of L(G); then u eé {0} and the corre-
sponding analytic subgroup N is a Lie subgroup which is invariant under
every continuous automorphism of G (no. 7, Proposition 23); the centre G’
of N has the desired properties.
PROPOSITION 27. Let G be a finite—dimensional connected real or complex Lie group.
Thefollowing conditions are equivalent:
(i) L(G) is simple;
(ii) the only normal integral subgroups ofG are {e} and G andfirther G is not com-
mutative.
This follows from § 6, no. 6, Proposition 14.
355
m LIE (moors
G/Z'. Let 115 be the mapping gHgyg'1y‘1 of G into G. Then ¢(G) is con-
nected and contained in Z' and hence {>(G) = (I)({e}) = {2}. Therefore 11 E Z.
PROPOSITION 29. Let G be a semi-simple connected real or complex Lie group.
(i) G = (G, G). .
(ii) The centre Z of G i." dunete.
(iii) The centre t/Z i: {e}.
Assertion (1) follows from the Corollary to Proposition 4, no. 2 and Chap-
ter I,§6, Theorem 1. _
Assertion (ii) follows from §6, no. 4, Corollary 4 to Proposition 10 and
Chapter I, §6, no. 1, Remark 2.
Assertion (iii) follows from (ii) and Lemma 1.
PROPOSITION 30. (i) Let g be a semi-simple real or complex Lie algebra. Then
Int 9 is the identity component ofAut 9.
(ii) Let G be a moi-simple connected real or complex Lie group. 7725 adjoin! group
of G i: the identity component qllt L(G). It: tentre reduces to the identity element.
Every derivation of g is inner (Chapter I, §6, Corollary 3 to Proposition 1)
and hence L(Int g) = L(Aut g), which proves (i). The first assertion of (ii)
follows from (i). The second follows from Proposition 29 (iii) and §6, no. 4,
Corollary 4 (ii) to Proposition 10. .
Remark 2. Let g be a complex semi-simple Lie algebra and 90 its underlying
real Lie algebra. Then Aut(g) is open in Aut(go), for Int(go) C Aut(9).
PROPOSITION 31. Let G be a Jimply connected finite-dimensional real or complex Lie
group and R it: radical. There exists a semi-:imple simply connected Lie subgroup S of G
rueh that G, a: a Lie group, it the semi—direct product of S by R. lfS’ i: a :emi-rirnple
integral rubgraup of G, there exirtr x in the nilpatent radical of L(G) such that
(Ad exp x)(S’) c S.
This follows from §6, no. 6, Corollary 1 to Proposition 14 and Chapter I,
§ 6, Theorem 5 and Corollary 1.
Lemma 2. Let G be a group (reap. a topological group), G’ a normal subgroup of G,
V a finite-dimenrional vector space over a commutative field k (resp. over K), p a
linear representation (resp. a cantinuou: linear reprexentation) (JG on V and p' = FIG'.
(i) {f p is semioimple, p' i: semi-simple.
(ii) If p’ i: semi—simple and every finite-dimennlonal linear k-reprerentation (resp.
cantinuoru linear K-reprexenlation) of G/G’ (resp. G/G’) is semi-simple, then (a i:
:emi-rimple.
Suppose that p is semi-simple; we prove that 9’ is semi-simple. It sufiices
to consider the case where p is simple. Let V’ be a minimal non-zero sub-G’-
module of V. For all g e G, p(G’)p(g)V’ = p(g)p(G')V' = p(g)V', in other
356
SEMI-SIMPLE LIE GROUPS §9.8
357
"I LIE GROUPS
are
irritation: of G and n an integer 20. Then 9, ® 9,, T“p, 5’9, /\"p (Appendix)
:emi—rimple.
The semi-simplicity of 9, ® 92 follows from Theorem 1 and Chapter I,
§6, Corollary 1 to Theorem 4. The semi-simplicity of T"p, 5"p, A”p follows
from the semi-simplicity of pl ® 9,.
358
INFINITESIMAL Amazonian-115m § 10.1
I. MINITESIMAL AUTOMORPIIISMS
Lemma 1. Let G be a Liep and a a oeetorfieldon G. For allg 6G, let
Mo) = Mg)!1 6 L(G).
Thefillowing condition: are equivalent:
(i) a i: o homomorphirm of the group G into the group T(G) ;
(ii) for «11g, 47 in G mtg!) = «(5)! + g’;a(g)
010]?” 4’13: g m G 9(a)) = He) + (Ad 5W!)-
Condition (i) means that, for all g, g’ in G, we have In the group T(G):
l5(g)gfl(g‘)g’ = News!
0!‘
5(g)((Ad g)fl(g'))g£' = 9(gg')gg'~
But the product of 6(g) and (Ad g)B(g’) in T(G) 15 just the sum of [3(g) and
(Ad g){3(g) in L(G) (§2, no 1, Proposition 2). Hence (i)¢> ¢>(iii) On the
other hand condition (ii) may be written B(gg’)g = [3(g)gg’ + gB(g ) g , or
15(gg'= 3(3) + (Adgmg')
and hence (ii) 9 (iii).
DEFINITION 1. Let G be a Lie group. An irfiniflesimal automrphism g" G is my
analytic vetlurfield on G satixfi/ing the condition: 4fLemmo 1.
Lemma 2. Let K' be a non-discrete olored rub/field of K, A a K'-manifold, B and C
K-manifaldy and f a K’dmulytio mapping of A X B into C. Suppoxe that, fir all
359
Ill LIE GROUPS
4 EA, the mapping b Hf(a, b) q into C is K-analytic. Then, for all lETA,
the mapping :4 >—> (Tf) (I, u) «f TB into TC is K-analym. . I _
We fix teTA and write g(u) = (Tf)(t,u). Clearly g 15 K-analytlc. By
Diflerentiable and Analytic Manifolds, R, 5.14.6, it suffices to prove that the
tangent mappings to g are K-linear. It can be assumed that A, B, C are open
neighbourhoods of O in complete nonnable spaces E, F, .G over K , K, K
and that t is tangent to A at 0. We identify TA, TB, TC, th A x E, B x F,
C x G and twith an element of E. Then for all (x,y) 6TB = B x F.
360
mrmrrEsnnAL AUTOMORPr-nsux § 10.1
morphism 11f G. Time nix-t: a low of analytic operation (A, g) v—> (”(g) of K on G
with thoflllowing proportion
(1) if D i: the emaciated law of infinitesimal operation, than D0) = a;
(2) for all 7. e K, 4),, e Aut G.
(a) For all p. > 0, let KI, he the open ball of centre 0 and radius y. in K.
For all g e G, let i, be the set of analytic integral curves f of a defined in a
ball KI, and such that f (0) = g. By Diflmntiablc and Analytic Manifoldx, R,
9.1.3 and 9.1.5, Fig is non-empty and two elements of .97, coincide on the
intersection of their domains of definition; let p.(g) be the least upper bound
of the numbers it. such that there exists an element of 3“, defined in K“;
there exists a unique element 0119', defined in Km); we denote it byf,.
(b) Let g1, g2 be in G,fi 5.97,],f369’” withfi andf, defined on the
same ball Kn. Then f1fa: Ku—> G is analytic and (f1f2) (0) = g,g2. On the
other hand, for all A e K”,
(“($1.32) = ‘l’x(gil‘l>>.(£2)
361
III LIE GROUPS
362
AUTOMORPHISM GROUP or A LIE GROUP (REAL on courmx CASE) § l0.2
363
m LIE GROUPS
Suppose in the rest of the statement that Aut G has this structure.
(ii) Aut G i: afinite—dimensional Lie group.
(iii) The morphism (b: u »—> L(u) ofAut G into Aut L(G) is analytic.
(iv) If G is connected, 4: is an isomorphism of theLie group Aut G onto a Lie sub-
group of Aut L(G) ; this Lie rubgroup is equal to Aut L(G) if G is simply connected.
(V) Let a be the set of infinitesimal automorphism of G. Then a ir a Lie algebra
of vector field: and the law of infinitesimal operation associated with the mapping
(a, g) >—> u(g) of (Aut G) x G into G is an isomorphism ty‘L(Aut G) onto 0.
(vi) The topology of the Lie group Aut G is the topology .7}. .
(a) The uniqueness of the analytic structure considered in (i) is obvious.
(b) Suppose that G is connected. Let H be the universal covering space
of G, p the canonical morphism of H onto G and N = Ker p. We introduce
the notation 6, 11 and Aut(H, N) of Lemma 4. We transport the Lie group
structure of Aut L(G) to Aut H by means of 6. Then Aut H becomes a finite-
dimensional Lie group and Aut(H, N) a Lie subgroup ofAut H (Lemma 4 (ii)).
We transport the Lie group structure of Aut(H, N) to Aut G by means of
n'l. Then AutG becomes a finite—dimensional Lie group. Properties (ii),
(iii) and (iv) of the theorem are satisfied and the mapping (a, g) >—> u(g) of
(Aut G) x G into G is analytic (Lemma 4 (i)). Let M be an analytic mani-
fold, f a mapping of M into Aut G and (I) the mapping (m, g) >—>f (m) g of
M x G into G. Clearly, iff is analytic, (I) is analytic. Suppose that (I: is ana-
lytic. Then the To: TM x TG —> TG is analytic; its restriction to M x L(G),
that is the mapping (m, x) >—> L(f (m))x of M x L(G) into L(G) is therefore
analytic; as L(G) is finite-dimensional, it follows that the mapping
m >—> L(f (m)) of M into Aut L(G) is analytic and hence that f is analytic.
Thus (1) holds.
Let L(G) be given a norm. For all x > 0, let B,‘ be the open ball of centre
0 and radius A in L(G). We choose A > O sufficiently small for \l; = expulliA
to be an isomorphism of the analytic manifold BA onto the open submanifold
¢(B,‘) of G. Let G) be a filter on Aut G. For (D to converge to Ida, in Aut G,
it is necessary and sufficient that L((D) converge to Idug, in Aut L(G) and
hence that L((l))|BMa and L((D)"Bm converge uniformly to Idaw This
condition implies that (DIMBW) and ¢“]¢(Bm) converge unifomly to
Idwamr Conversely, suppose that (DHKBM) converges uniformly to Idmam).
There exists Med) such that, if ueM, then u(¢(BN2)) C (L(n); then
L(u) (BN2) is a connected subset ofL(G) whose image under expG is contained in
“Banal: hence L(u)(Bm) does not meet B,L — Bem and therefore
L(u)(Bm) C 13,; then the hypothesis that ¢|¢(BA,,) converges uniformly to
Elma”) implies that L((D)|Bm converges unifomly to Idam. It then follows
at:
((1) converga to Ida in Aut G) a ((1) convergs to 1d,; under .71.).
364
AUTOMORPH’ISM GROUP or A LIE GROUP (REAL on COMPLEX CASE) § 10.2
Let D be the law of infinitesimal operation associated with the law of left
operation ofAut(G) on G. By Propositions 1 and 2 of no. 1, D(L(Aut G)) = a.
Hence a is a Lie algebra of vector fields and D is a morphism of L(Aut G)
onto a. Let x1 and xa be elements of L(Aut G) such that D(x,) = D(x,).
Then the laws of operation (A, g) >—> (exp M1)g and (A, g) s—> (exp M2) g of
K on G have the same associated law of infinitesimal operation; hence, for
ll] sufficiently small, exp M1 and exp M, coincide on a neighbourhood of 3
(§ 4, no. 7, Theorem 6), whence exp M1 = exp 1x2. It follows that x1 = x,
and hence D is an isomorphism of L(Aut G) onto a.
The theorem has thus been completely proved for G connected.
(c) V\’e pass to the general case. By hypothesis, G is generated by Go and
a finite number of elements :1, :2, . . ., x”. Every 14 eAut G leaves Go stable.
Let Aut1G be the set of u eAutG which, on passing to the quotient, give
the identity automorphism of G/Go. This is a normal subgroup of Aut C.
By part (b) of the proof, Aut G0 has a canonical Lie group structure and the
mapping (£1; £2: - ~ -: gm u) H ("31: “£2: ' - u “321) °f Gii X Aut Go into G3
is analytic. Let P be the corresponding sernidirect product of Aut Go by G3;
it is a finite-dimensional Lie group (§ 1, no. 4-, Proposition 7).
Ifw E Aut1 G, we write
111., = lo eAut Go
w. = xi-lWO‘t) 5G0 (1 g ig n)
{(w) = «"11, - - u '0»): we) 5 P-
For all w, w’ in Aut,
C(wmw') = ((101: ~ - '3 wn)(w0(wl)’ . . -, "Mn/i), mow/6)
= ((wlw0(wi)a - - -. wnwiz(w5»)), Wolv'o)
= ((X'lw(xx)w(x1'IW’(xi))u - -, x; ’W(xn)W(x;‘W'(xu))), wawé)
= (((wu/h, - - ‘a (um/)n), (ww')o)
= {(ww’)
and hence C is a homomorphism of Aut,G into P. This homomorphism is
obviously injective.
We show that {(Aut) is closed in P. Let 0 be a filter on Aut such that
{(43) converges to a. point ((wh . . ., w”), mo) of P. Then 11> converges point-
wise to a mapping v of G into G. Clearly I) is an endomorphism of the group
G. Moreover, 11 leaves each coset modulo G0 stable and I)|G,J 2 mo. It follows
that v EAut. As {(v) = ((wl, . . ., w"), wg), we have shown that C(Aut)
is closed in P.
(d) In part (d) of the proof we assume that K = R. By § 8, no. 2, Theorem
2, {(Aut) is a Lie subgroup of P. We transport the real Lie group struc-
ture on E(Aut) to Aut by means of C“. Thus Aut becomes a finite-
dimensional Lie group.
365
III LIE GROUPS
366
Amouomrsu cxoup or A LIE amour (mm-mic cm) § 10.3
Clearly Aut is open in AutG with the topology 57],. Let w eAut G.
Let a be the automorphism u»—>wow'1 of Aut,Gr It is real analytic (§ 8,
no. 1, Theorem 1), L(a) is an R—automorphism of L(Aut) and
D a L(Aut,G) . 1)-1
is an R—automorphism of a. This automorphism is also the automorphism of
a derived from w by transport of structure; as w is K—analytic, we see that
L(u) is K-linear. Hence a is K-analytic (§ 3, no. 8, Proposition 32). By § 1,
no. 9, Proposition 18, there exists on Aut G one and only one Lie K—group
structure such that Aut,G is an open Lie subgroup of Aut G. The fact that
this structure has the properties of the theorem follows from the correspond-
ing properties for Aut.
COROLLARY I. Let G be ofinile-dimenxionol real Lie group and G0 it: identity com-
ponent. Suppose that G is generated by Go and afinite number of elements. Then Aut G
has the topology 9:, and is a finite-dimensional real Lie gimp.
COROLLARY 2. Let G be a semi-simple connected real or complex Lie group. The
group Int G is the identity component ofAut G.
The mapping u>—>L(u) is an isomorphism of AutG onto a Lie subgroup
of Aut L(G) (Theorem 1). The image of IntG under this isomorphism is
Ad G. But Ad G is the identity component of Aut L(G) (§ 9, no. 8, Proposi-
tion 30 (ii)).
367
m LE GROUPS
(c) We pass to the general case. By §7, no. 1, Proposition 1, there exists
an open compact subgroup Go of G which is of the type considered in (h).
Then G is generated by Go and a finite number of elements x1, x3, . . ., x".
Let Aut be the set of u E Aut G such that u(Go) = Go and u(x.GD) = #463,
for l S i S n. We define as in the proof of Theorem 1, part (c), a semi-direct
product P of Aut Gn by GE and an injective homomorphism Z of Aut1G into
P, whose image is closed in P.
(d), (e) : the argument is exacfly as in parts (d), (e) of the proof of Theorem
1 with R replaced by Q, and using Proposition 3 instead of Proposition 2. .
Remark. If K = Q” and the Lie group G is gmrated by a compact subset
(cf. Exercise 2), assertions (i), (ii), (iii) and (vi) of Theorem 1 are still true,
but not (v) (Exercise 3).
APPENDIX
OPERATIONS ON LINEAR REPRESENTATIONS
Let G be a group, It a commutative field, E1, Em. . ., E” vector spaces over
I: and 7': a linear representation of G on E. (l .< i < n). The mapping
g-—> "1(a) ® --~ ® Ms)
is a linear mapping of G into the vector space E, ® . - - ® Em called the
tensor pmduct of 1:1, . . ., 1:, and denoted by 1:, ® - - - ® 1:".
Let E be a vector space over k and 1: a linear representation of G on E.
For all g 6G, let T(g) (resp. 0(g), e(g)) be the unique automorphism of the
algebra T(E) (resp. S(E), /\(E)) which extends 1r(g) (Algebra, Chapter 111,
§5, no. 2, § 6, no. 2 and § 7, no. 2). Then 4.- (resp. a, e) is a linear representa-
tion of G on T(E) (resp. S(E), /\(E)) denoted by T(n) (resp. 5(rr), /\(1r)).
The subreprsentation of T(n) (resp. 5(a), /\(1r)) defined by T"(E) (resp.
5"(E), A"(E)) is called the n-th tensor (resp. symmetric, exterior) power of
1: and is denoted by T"(rr) (resp. S"(1r), A”(1:)). Then
T"(1r) = 1r®1r® ®1r
(7: factors). The representations 5(a), /\(1r) are quotient representations of
T(n) and hence S"(1r), A”(1r) are quotient representations of T"(7r).
Let g be a Lie algebra over It. The tensor product of a finite number of
representations of g has already been defined in Chapter I, §3, no. 2; it is
denoted by 1:1 ® - - - ® 1:”. Let E be a vector space over I: and 7: a representa-
tion of g on E. For all x e 9, let 1"(x) (resp. a’(x), s'(x)) be the unique deriva-
tion of the algebra T(E) (resp. 503), /\(E)) which extends 1v(x) (Algtbra,
Chapter 111, § 10, no. 9, Example I). Then 1’ (resp. a’, a') is a linear repre-
sentation of 9 on T(E) (resp. S(E), /\(E)) by Algebra, lac. cit., formula (35),
368
APPENDIX
369
EXERCISES
§1
1. Let G be a finite-dimensional connected real or complex Lie group.
Let H and L be Lie subgroups of G such that HL = G. Show that the canon-
ical mappings G—>G/H and G—>G/L define an isomorphism of analytic
manifolds
G/(H n L) —> (G/H) x (G/L).
2. Let P = {z E C|J(z) > 0}. Let G be the group of bijections z l—> uz + b
of P (a > O, b e R). Then G operates on P simply transitively, which allows
us to transport the complex analytic manifold structure on P to G. Show that
the structure obtained is invariant under left translations of G but not under
right translations.
3. Let R; be the additive group of real numbers with the discrete manifold
structure. Let Rd operate on the analytic manifold R by the law of operation
(1,3,) ~—> x + y. Then R, operata transitively on R but R is not a Lie homo-
geneous space of Rd.
4-. Let H be a compact real Lie group operating on a finite-dimensional
real manifold V; suppose that V is of class C', where reNR, and that the
action of H on V is of class C’. Let p eV be invariant under H; the group H
operates linearly on the tangent space T to V at 1). Show that there exist an
open neighbourhood U of [a which is stable under H and a C'—morphism
f: U —>T such that:
(a) f (p) = 0 and the tangent mapping to f at p is the identity;
(b) f commutes with the action of H on U and T.
(Choose first an fl, satisfying (a) and then definef by the formula
fix) = f haw-w,
where dli is the Haar measure on H with total mass 1.)
370
EXERCISES
1’ For more details, see R. S. PALAIS, On the existence of slices for actions of non-
compact Lie groups, Ann. 44' Math, 72 (1961), pp. 295—323.
1 For an example of such an ordered pair (E, F), cl‘. for example A. DOUADY, Le
probléme des modules pour les sous-espaces analytiques compacts d’un espace
analytique donné, Ann. Inst. Fourier, 16 (l966),:p. [6.
371
m LEE GROUPS
§3
1. Let G be a Lie group and G1 and G, Lie subgroups of G. Suppose that
G2 is of finite codimension in G and that K is of characteristic 0. Then
G1 fl G2 is a Lie subgroup of G with Lie algebra L(Gl) fl L(G2) (argue as
in Proposition 29 and its Corollary 2).
‘ 2. Suppose that K is locally compact. Let G be a Lie group of finite dimen-
Slon ’l.
372
EXERCISE
at“ 24B 3
(“Y 118 + M an) ‘
7’ 278 8’
(b) Show that m >—> det(m) is the non-degenerate quadratic form
q(a, b, t) = no — b2 .
on E and that p is a morphism of SL(2, C) onto SO(q) with kernel (I, —I}.
(For the surjectivity, observe that, in the matrix of an element of 80(q), the
lst column can be expressed in the form (an’, «y, y“) and the 3rd column in
the form (3‘, B8, 8“), where «’82 + BRY‘ — Quyfls = 1. Then observe that
2 elements of SO (q) which coincide on a non-isotropic plane are equal.)
(4) Deduce that the complex Lie group 50(3, C) is isomorphic to the
complex Lie group SL(2, C)/{I, -I}.
4-. (a) Let q be a quadratic form of signature (1, 2) on R“. Let P be the
set of m e R:3 such that q(m) > 0. Then P is the union of two disjoint oonvec
cones C and —C. IfsESO(q), then either 5(0) = C and .y(-—C) = -—C or '
:(C) = -—C and :(—C) = C. Let 50*(q) be the set of:eSO(q) such that
:(C) = G. Then 80*(q) is an open normal subgroup of 80(q) of index 2
in 50(q).
(b) By imitating the method of Exercise 3, define a morphism of SL(2, R)
onto SO+(q) with kernel {1, —1}. Deduce that the real Lie group SO+(q)
is isomorphic to the real Lie group SL(2, R) [(1, —I}.
373
“I LIE GROUPS
(5) For (Em) EC“ and (E', 71') EC“, let f((€,n)s (5': n'))1= Egl' — 70771
Show that the inner automorphism of SL(2, C) defined by 7—5 (_1. 7)
maps SU(f) to SL(2, R).
5. Let E be the real vector space of Hermitian complex matrices with 2
rows and 2 columns of zero trace. For all .r ESU(2, C), we define an auto-
morphism p(r) of E by p(:) (m) = m‘. We identify the element
( x y + £2)
y — 1'2 ——x
of E with the element (x, y, z) of R3. By imitating the method of Exercise 3,
show that p is a morphism of the real Lie group SU(2, C) onto the real Lie
group 50(3, R) with kernel {1, —I} and that 80(3, R) is isomorphic to
5U(2, GNU, -1}
1T 6. Let F be the vector space of Hermitian complex matrices with 2 rows
and 2 columns. For all .r e SL(2, C), we define an automorphism 6(5) of F by
o-(:) (m) = m*. We identify the element t : 1: y‘ :- 11) of F with the
element (t, x, y, z) of R4 and write
q(t,x,y,z) =1“ - I" -y’ — 2’.
Let C be the set of (t,x,y,z) eR4 such that t2 — x2 —y2 — 7.2 > 0, t > 0.
Let SO+ (g) be the set ofg E 50(4) such that g(C) = C; it is an open normal
subgroup of SO(q) of index 2 in SO(q) (cf. Exercise 4). Show that o' is a
morphism of the underlying real Lie group of SL(2, C) onto the real Lie
group SO+(q) with kernel (I, —I}r (For the suijectivity, use Exercise 5.)
Deduce that the real Lie group 80+(q) is isomorphic to the underlying rail
Lie group of SL(2, C)/{I, —I}, that is (Exercise 3) the underlying real Lie
group of SO(3, C).
7. (a) Let G be the set of quaternions of norm 1. This is a Lie subgroup of
the real Lie group 11*, homeomorphic to 5,, and hence simply connected,
cf. General Typology, Chapter XI.
(b) Let E be the real vector space of pure quaternions, identified with R3
by
(x,y, z)t—>xi +w' + 11:.
For all g e G, we define an automorphism p(g) of E by p(g)q = gqg“. Show
that p is a morphism of the Lie group G onto the real Lie group 50(3, R)
with kernel {1, — I}. Deduce that the Lie group 80(3, R) is isomorphic to
the Lie group G/{l, — l).
(5) Let C be identified with the subfield R + Ri of H. The mapping
374
EXERCISES
375
III LIE GROUPS
exists a greatest index I such that 1?, sé {0} and all the 2, of indices <7
are distinct. For [I 2 0,
. It +
dlm($p/$,H) $ (P +{)
§4
1. Let G be a Lie group, H a normal Lie quasi-subgroup of G and 1c the
canonical mapping of G onto GIH. There exism on GIH one and only one
Lie group structure with the following property: for a homomorphism 8
of G/H into a Lie group G' to be a Lie group morphism, it is necessary and
suflicient that 0 c 1: be a Lie group morphism. Moreover, L(G/H) is canonic-
ally isomorphic to L(G)/L(H). (Let Q be a Lie group germ such that
L(Q) = L(G)/L(H). Show that, by shrinking Q if necessary, Q can be
identified with an open neighbourhood of 0 in GIH and that Proposition 18
of § 1 can then be applied to G/H.)
H11 2. Let G and H be two Lie groups andf a Lie group morphism of G'into
376
EXERCISES
that (NO) = e, To(¢l>) = Idua, and 42((1 + x')b) = <1>(M)<]>()t’b) for M, Nb,
(A + A’)b in U. Then :1) coincides on a neighbourhood of 0 with an aponen-
tial mapping.
(b) Let k be a field of characteristic 0. Let K be the valued field k((X)),
which is of characteristic 0. Let G be the additive Lie group k[[X]]. Let ()2
be the continuous k-linear mapping of G into G such that (“X") = X” + X’"
for every integer n 2 0. Then (1) satisfies the conditions of (a), but coincides
on no neighbourhood of 0 with an arponential mapping.
5. Let (:1, e2, :3) be the canonical basis of 113‘ We give R“ the nilpotent
Lie algebra structure 9 such that [51,52] = 23, [91,33] = [65,93] = 0. Let
0”,, denote the constants of structure of 9 relative to the basis (:1, :3, :3).
(H) Show that on R3 the difierendal forms
‘01 = 1‘1, 02 = dim 0’s = “”1412 + (in
satisfy the relations
dmk = fig, We), A a), (k = 1,2, 3)
and are linearly independent at each point of R”.
(b) Deduce that there exists, on an open neighbourhood of 0 in R“, a
Lie group germ structure G such that L(G) = g and
(l) (‘11: “2: a3) (’51: ”a: *3) = (“'1 + “1; 7‘2 + “a: “a + '13 + “112)
for (al, a2, a3), (x1, #2, x3) sufliciently close to 0. Show that in fact formulae
(1) define a nilpotent Lie group structure on R“.
11 6. Let G be a finite-dimensional Lie group, 9 its Lie algebra, g‘ the dual
vector space of g, a the adjoint representation of G on g and p the representa-
tion ofG on 9* defined by p(g) = ‘a(g) '1 for all ge G.
(a) L(p)(:<) = —‘(ad x) for all x E g.
([7) Let fe 9*. Let G, denote the stabilizer off in G; it is a Lie subgroup
ofG and g, = L(G,) is the set ofx EG such that ‘(ad x)f= 0.
(c) For x, y in 9, we write B,(x, y) = f ([x, y]). Show that B, is an alter-
nating bilinear form on 9, that the mapping 5, of 9 into 9* left associated with
B, is the mapping x >—> '(ad 2:) f and that the orthogonal of g with respect to
B, is 9,. Let {3, denote the non—degenerate alternating bilinear form on 9/9,
(which is therefore of even dimension) derived from B, when passing to the
quotient. We denote by {3; the inverse form of B, on (g/g,)*.
((1) Let 9 denote an orbit of G on 9*; we give it the manifold structure
derived from that on GIG, by transport of structure, where f is an arbitrary
point of 9. Then G operates analytically on Q on the left. Let D be the
associated law of infinitesimal operation, For all a e g and all fe 9,
Du.(f) = —(‘ada)f-
377
[ll LIE GROUPS
The tangent subspace T,(Q) is the image of 5,, that is the orthogonal of g,
in g*, and is canonically identified with (g/g,)*, so that 5, defines a canonical
isomorphism r, of T,(Q) onto its dual. The fieldfn—> p; is an analytic differen-
tial form to of degree 2 on K], which is invariant under G. For f E Q, a E 5,
b e g, «(Dam Dbifl) =f([a: b1).
(9) Show that do) = 0.
(f) If at is a difl‘erential form of degree 1 on Q, the isomorphisms 7, allow
us to identify at with a vector field. Let A be the set of analytic functions on
Q with values in K. For d), '1? in A, we write [<1>, 4/] = um, (14‘) e A. Show that
A has thus a Lie algebra structure.
(g) For all a e 9, let v.11“ be the function f ~—>f (a) on Q. Show that a >—> up“
is a homomorphism of 9 into the Lie algebra A and that the form do, is
identified, by means of the isomorphisms 1,, with the vector field D“.
(/1) Let U be an open subset of g“ and d) an analytic function on U. For all
fe U, the differential d,<]> of d) atf is identified with an element of 5. Suppose
that X4; = 0 for every vector field X defined by the action of G on 9*. Show
that then, for allfE U, d,<lz belongs to the centre of 55,.
(2') For all f e 9‘, we write 1, = dim 9,. Let r = inf3,. Show that the set
i 5 I
V off e {1* such that r, = r is open and dense in 9‘. Show that, for all f e V,
g, is commutative. (Construct r functions 4: satisfying the conditions of (h)
in a neighbourhood off and such that their differentials at f are linearly
independent.)
11 7. (a) Let (A, x) ~—> L): be a law of continuous left operation of R on T.
Then one of the two following holds:
(1) either there exim a fixed point and the stabilizer of every point is either
R or {0};
(2) or there exists a point whose smbilizer is an infinite discrete subgmup
of R and R operates transitively on T. (If {Aelix = x} = {0}, the orbit
of x is homeomorphic to R and has frontier points Alim Lx which are fixed.
+ on
-
If {heRILx = x} = In with a $5 0, the orbit of x is homeomorphic to T
and R is transitive.)
(b) Iff is a homeomorphism of T onto itself and k an integer 21, a point
xeT is called periodic of period I: for f if f*(x) = x and f "(x) ;é a: for
1 < h < k. In the notation of (a), let f,L be the homeomorphism x»—> 1.1:.
Then the set of periodic points of period It > 1 forfi is empty or equal to T.
(If there exists a periodic point of period k > 1 for f,” its stabilizer is distinct
from {0} and R, hence R operates transitively on T and all the points of T
are periodic of period I: forfi.)
(1:) Let Q be a neighbourhood of e in Difl"“'(T) (Dzfirmtiable and Analytic
Mani/7211A, R, 15.3.8, Nole (1)). There exist an integer It > i and 9 with-
out fixed point such that the set of periodic points of period k for f is neither
empty nor equal to T. (Let p: R —> T bc the canonical mapping. Let p be
378
mm
11 $1100, 2—,: )
and mm) = p(x + m» for x 510, 3,? :where
0<).(x)<2TR—x;
can be chosen such that a- e 0'. Thenf = p a a has the required properties.)
Such an f cannot, by (b), be in the image of a continuous morphism of R
into Diff"u (T).
((1) Let Y be a compact difi‘erentiable manifold of dimension 21. Every
neighbourhood of z in Difi‘”(Y) contains an element It with the following
property: It belongs to the image of no continuous morphism ofR into Difl‘°(Y).
(The manifold Y contains an open submanifold of the form T x D, where
D is the open Eulcidean ball of radius 1 and centre 0 in R" (n = dimY — 1).
Let geDifl‘“(D) be such that g(0) = 0, ||g(x)]| > “x” for 0 < “x“ < 41- and
g(x) = x for ”x” 2 i and g very close to e in Difl‘”(D). Letf be as in (c),
and, for 0 < t < l,1etf,:T—>T be defined as follows: for 3 ET, 1/ E];"(x),
zep“(f(x)) and |z — y| < 1:, we write fi(x) =p(ty + (1 — t)z). Then
there exists h eDifl'“(Y) such that h(x,y) = (fzflyn(x),g(y)) for x ET, 3/ ED
and h(u) = u for u eY —- (T x D); moreover, we can choose f and g such
that h is arbitrarily close to e in Difl‘“(Y). The points of T X {0} are the
y e Y such that when It tends to +00, h""(x) tends to a point not fixed under h
(in fact it is periodic of period k). Therefore, every homeomorphism of Y
onto Y which commutes with It leaves T x (0) invariant. Then apply (0).)
(e) Let Y be a compact differentiable manifold of dimension 21. There
do not exist a Lie group G and continuous morphism Diff ”(Y) —v> G,
G —> Diff°(Y) whose composition is the canonical injection of Diff ”(Y) into
Difl‘°(Y). (Use (11).)
8. Let G be a finite-dimensional Lie group. Suppose that L(G) is simple.
Let A be a normal subgroup of G. If A is not open, A is discrete and its
centralizer in G is open. (Let n be the subalgebra tangent to A at a. Show that
a = {0). Let x be an element of A not belonging to the centre of G. Consider
the mapping y ny" of G into A. Deduce from the relation is = {0} that
the centralizer of x in G is open. Then, using an exponential mapping, show
that the centralize!" of x in G contains a neighbourhood of 1 independent
of x.)
379
[II LIE GROUPS
380
EXERCISES
has the property of Exercise 10, which is stable under inner automorphism
of U(n) and which is such that if x, y are non-permutable elements of V then
x and (x, y) are not permutable. (Use (c).)
(2) Let {3 be the normalized Haar measure of U(n). Let Vl be a symmetric
compact neighbourhood of e in U(n) such that V: C V. Let p be an integer
such that {3(V1) > Up. Show that, for every finite subgroup F of GL(n, C),
there exists a commutative normal subgroup A of F such that Card(F/A) S [7
(Jordan’s theorem). (Reduce it to the case where F C U(n). Take A to be the
subgroup of F generated by F n V and use (41).)
12. Let G be a finite-dimensional connected real or complex Lie group
and or a left invariant difi‘erential form of degree 9 on G. For a: to be right
invariant too, it is necessary and suflicicnt that do: = 0. (Observe that the
condition for right invariance can be written
/\”(‘Ad(I))-¢(¢) = w)
for all s e G and hence it is equivalent to the condition
I
,Z;<°‘(‘)’"l A A 14,.1 A [u,u,] A u,“ A A 11,) =0
for all u, 111,. . ., u, in L(G).) In particular, the left and right invariant dif-
ferential forms of degree 1 on G form a vector space of dimension
dim L(G) — dim[L(G), L(G)].
13. Let R" be given the usual scalar product ((5,), (111)) H‘; Em“ Let
I(n) be the group of isometric affine transformations of R". It is canonically
isomorphic to the Lie subgroup of GL(n + l, R) consisting of the matrices
S=(é’ f)
where U 6 C(71) and x is an arbitrary element of R" (matrix of type (It, 1)).
I(n) can be identified with the semi-direct product of 0(a) by R" defined by
the canonical injection of 0(a) into GL(n, R) (the matrix S is then denoted
by (U, x)). Let p: 1(a) ~> 0(n) be the canonical surjection, with kernel R".
(a) Let 1" be a discrete subgroup of 1(a). Consider the closure ATP) ofp(I‘)
in 0(a), Show that its identity component is commutative. (Let V be a neigh-
bourhood of I in 0(n) with the properties of Exercise 11 (d) and such that
further ||U — I H S i for all U EV (the norm on End(R”) being derived
from the Euclidean norm on R"). Argue by reductir ad abmrdum, assuming
that there exist SI = (U1, x1), S2 = (Umxz) such that U1 and (I; belong to
V and do not commute. Show that (SJ, S3) = ((U1, U1), y) with
M < Mllxill + “12“)-
38]
III LIE GROUPS
Then define the Sk inductively by S,‘ = (S,, Ska) for k 2 3. Using the choice
of V, show that this sequence has an infinity of distinct terms and is bounded
in I(n), which is absurd)
(b) l" is called crystallographic if I(n)/l" is compact. Show that if this is so,
for all x ER", the affine subspace L of R“ generated by D; is equal to R“.
(Otherwise, for all y e R", all the points of I‘y are the same distance from L
and, as this distance can be arbitrarily large I(n)/I‘ is not compact.)
(c) If I‘ is commutative and crystallographic, then I‘ C R“. (If
S = (U, x) E I‘ is such that the vector subspace V C R" consisting of the -
points invariant under U is not equal to R" and VJ- denotes the orthogonal
subspace of V, then, for all 5’ = ( U’, x’) e F, U' leaves V and Vi stable. On
the other hand, since UIVi has no non-zero invariant point, show that this
allows us to assume, by changing the origin, that e. Using (b), there
exists 5' = (U’, x’) e I‘ such that the orthogonal projection y’ of 1’ onto V‘-
is #0. By evaluating Sx’ by the formula S = S’SS"‘, deduce that U.y' = y’
contrary to the definition of V.)
(d) If I‘ is crystallographic, I‘ n R’‘ is a free commutative group of rank
n and I‘/(I‘ ft R“) is a finite group (Bieberbach’s Theorem). (Let W be the
vector subspace of R" generated by I‘ n R". The compact group W has
only a finite number of connected components.» If W = {0}, I‘ contains, by
(a), a commutative normal subgroup of finite index 1‘1; 1‘1 is then crystallo-
graphic, which implies a contradiction with (0). Hence W sé {0}. Show that
9(1") leaves W stable and that p(l")|W is finite; otherwise there would exist
a sequence (Sm) in I‘ such that if S", = (Um, x”), the U,,l are distinct and tend
to I; then form the (I, a,)S,,,(I, a,)‘1S;‘, where (a,) is a basis of I‘ n R”, and
obtain a contradiction with the hypothesis that l" is discrete. Finally, to see
that W = R”, show that otherwise the action of I‘ on R"/W would be that
of a crystallographic group containing no translation #0.)
§5
1. Suppose that K = R. Let r be an integer 21 or 00. A group of clam C'
is a set G with a group structure and a manifold structure of class C' such
that the mapping (my) ~—> xy“ ofG X G into C is of class C’.
(a) Suppose henceforth that r 2 2. A neighbourhood of e in G is identified
with a neighbourhood of 0 in a Banach space. Let xy = P(x, 3/), where P is
a mapping of class C2 on an open neighbourhood of (0, 0). Let
(D1132?) (0: 0) = B-
Show that
as (x, y) —> (0, 0). (Use an expansion of P of order 2 with integral remainder.)
(b) Show that
x“ = —x + B(x,z) + lx]’u(l)
m“ 1/ + Bony) - 13m”) + lxl IyW)
x‘ly‘ 1y = 130w) — B(y.x) + III [y|v(1)
as (my) tends to (0, 0).
2. Let G be a group ofclass C’ with r > 2.
(a) Let t 6.7"")(G), t’ 6.9"“)(G). Ifr + A" S r, define I: t’ EY"""’(G) as
in § 3, no. 1. If t and t’ are without constant term, It t' is without constant
term. The image of t under the mapping x »—> x“ of G into G is denoted by
t"; then If" 6.7'“‘(G). Iffis a function of class Cr on G with values in a
Hausdorff polynormed space, f): I denotes the function on G defined by
(f4: t)(x) = (5,, * t‘,f) for all xeG; this function is ofclass C"’ if: < 00.
Show, asin§3, no. 4, thatfa (tst') = (fst) *t'.
(b) If t e T,(G), the vector field x -> a, a: I on G is denoted by L‘. Show, as
in §3, no. 6, that, for t, t’ in T,(G), L”. = L, o L,. and hence that
Law—t": = [Lo Lu}
Therefore, H t' — t' t t is an element of T.(G) which we denote by [t, t'].
(0) Using the notation of Exercise 1 (u), if t eT,(G), then
L4") = (DaPOr; 0))(1)
for x sufficiently close to 0. Deduce that [t, I’] = BO, 1’) — B(t’, t).
(d) Show that T,(G), with the bracket (I, t’) >—> [t, t'], is a normable Lie
algebra. (To prove the Jacobi identity, use (c), Exercise 1 (b) and identity
(5) of Algebra, Chapter I, §6, no. 2.)
(For a sequel to this exercise, cf. §8, Exercise 6.)
§6
I. Let D be a set of elements of SL(2, C) of the form (3 a? 1) where
a > 0 and b E C. Show that D is a Lie subgroup of the underlying real Lie
group of SL(2, C) and that the mapping (u, d) »—> ml of SU(2, C) X D into
SL(2, C) is an isomorphism of real analytic manifolds. Deduce from this
and §3, Exercise 7 (a) that SL(2, C) is simply connected.
2. Let G be the universal covering of SL(2, R), 1: the canoncial morphism
of G onto SL(2, R) and N its kernel.
(:1) Arguing as in Exercise 1, show that there exists an isomorphism of the
real analytic manifold U x R2 onto the real analytic manifold SL(2, R).
Deduce that N is isomorphic to Z.
383
11! LE GROUPS
1’ Cf. N. H. KUIPER, The homotopy type of the unitary group of Hilbert space,
Tupulngy, 3 (1965), pp. 19—30. In this article it is even proved that G is con-
lractible.
384-
EXERCISES
number. Let 3 be the Lie subalgebra of L(G) x L(G) consisting of the (x, ax)
where xEL(Z). Let S be the corresponding integral subgroup of G x G.
Then 6 is an ideal of L(G) x L(G); however S is not closed and is dense in
Z x Z.
(b) Let g = (L(G) X L(G)) [5. There exists no Lie group with Lie algebra
9. (Let H be such a group. As G is simply connected, there exists a morphism
i“ G X G——>H such that L(o) is the canonical mapping of L(G) x L(G)
onto 3. LetN = Ker (b. ThenL(N) D 9, henceN D S and therefore N D Z x Z
by (0). Then L(N) 3 L(Z) x L(Z), which is absurd.)
9. Let X be a non-empty connected compact complex manifold of dimen-
sion n. Suppose that there exist holomorphic vector fields E1, . . ., E" on X
which are linearly independent at each point of X. Show that there exist a
complex Lie group G and a discrete subgroup D ofG such that X is difieomor-
phic to G/D. (Let cm: be the holomorphic functions on X such that
[Eu 5,] = 2 tag,“ The ‘uk are constants because X is compact. Take G to be
the simply connected complex Lie group whose Lie algebra admits the a,” as
constants of structure and use Theorem 5.)
10. Let G be a. Lie group with a finite number of connected components.
For G to be unimodular, it is necessary and sufficient that Tr ad a = 0 for all
a e L(G) .
11. Let E be a complete normable space over C, v e!(E) andg = exp(v).
Suppose that Sp(v) n 2i1:(Z — (0}) = 0. Let E1, E2 be closed vector sub-
spaces of E which are stable under 0, such that E; C E1. Suppose that the auto-
morphism of El/Ea derived from g is the identity. Then 0(E1) C E,.
1} 12. Consider a real or complex complete normable space E and a closed
vector subspace F of E such that F0 admits no topological supplement in E'.1‘
(a) Let A (resp. B) be the complete normable space of continuous endo-
morphisms of E (resp. F). Let C be the set ofu e A such that u(F) C F. Let at be
the mapping a >—> (u, u | F) of G into A x B. Then at is an isomorphism of the
complete normable space C onto a closed vector subspace of A X B and ot(C)
admits no topological supplement in A X B. (Let x e E be such that ,1: ¢ F. Let
1‘ Let 5.; (rap. 1 1, I m) be the Banach space ofreal or complex sequences (1:, :2, . . .)
satisfying lim x“ = 0 (resp. Z [2“] < +00, resp. sup [a < +oo) with the norm
a“: n
“at” = sup [xnl (rcsp. Z |x,.|, rcsp. sup l‘nl)- There exists a continuous morphism
n
1: of l1 gnto :0; let F be its kernel. Then (:0)’ = 11 IS identified with F”. But a
vector subspace of countable type of 1“ cannot be a topological direct factor in 1‘”
(cf. A. GKO ur les
K Su ' ’ , d’espaces
du type C(K), Can. J. Math, 5 (1953), p 169).
385
III LIE GROUPS
EE Fn be such that (x. Z) = 1. If 1) e 13', let i denote the element y H (y, 11)::
of A. Suppose that there eidsts a. projector 11: of A x B onto u(C). We define
17:: E’ —> E’ by 11(1)) = ‘(a‘1(n(fi, 0)))(E). Then 1': is a projector ofE’ onto F“,
which is absurd.)
(b) A unital real or complex complete normable algebra M is constructed
386
EXERCISES
387
Ill um GROUPS
«poo + y» _— “Lug-w
~
(exp ‘5u up ’1!) "
ll
. 1x ——tx — "’
eXPU’Dr. 11]) = ”1131” (61p ;~ap%~wp 7422:]: 71y)
uniformly for ta [0, 1].
20. Let G be a Lie group, H a Lie subgroup ofG and A an integral subgroup
of G such that L(H) n L(A) = {0}. Show that H n A is discrete in the Lie
group A.
21. Let G be a real Lie group and H a normal l-dimensional integral sub-
ST°UP~
(a) If H is not closed, Fl is compact (Spectral Theories, Chapter 11, § 2,
Lemma 1), hence isomorphic to T" and hence central in G if G is connected
(use General Topology, Chapter VII, §2, Proposition 5).
(b) Suppose that H is closed. Let a be an element of G and C(u) the set of
elements of G which commute with a. Suppose that H 4: (1(4).
IfH is isomorphic to R, then C(a) n H = {3). (Consider the automorphism
at: h>—>o"ha of H.) If H is isomorphic to T, C(a) n H has two elements.
(Consider on again and use General Topology, Chapter VII, § 2, Proposition 6.)
The second situation is impossible if G is connected (use (a)).
(a) With the hypotheses of (b), suppose further that G/H is commutative.
388
EXERCISES
389
ll! LIE GROUPS
390
EXERCISE
§7
1. Let G be a Lie group and (b: U -> G an exponential mapping such that
ZU C U and ¢(1x)= ¢(x)'fora.lleandallreZ. If]; > 0, :1) is ananalytic
isomorphism of U onto (MU). (If (Mac) = (My), then ¢(p”x) = (ME?) for all
n E N and hence x = y. Let W be an open neighbourhood ofO in L(G) such that
¢_1 is analytic on (MW). For all r e (MU), there exist I: e N and a neighbour-
hood V cfr in ¢(U) such that :eV => I" etI>(W).)
2. Let G be a Lie group, (I): U—>G an exponential mapping with the
properties of Proposition 3, V a neighbourhood of 0 in U such that ZV C V
and 4/: V —> G a tangent mapping at 0 to (I) such that Mm) = Ms)" for all
neV and all 7162. Up > 0, then 4’ = 42 |V. (Give L(G) anorm. Lete.
Then 9’s tends to 0 as 71 tends to +00, hence there exists an" > 0 such that an
tends to o and I <4>-1 ., mu) — M s anllwrll- But Wx) = W)“,
whence M" a door) — :41 s aux .)
3. Let U be the set of invertible elements of A and U’ = 1 + m C U.
(a) Show that U' C U,. (ID: = I + y withy e m, then x”" tends to l as It
tends to +00 by the binomial theorem.)
(b) Show that U, is the set of elements of U whose image in A/m is a root
of unity. (Use (11).) Hence recover the fact that U = U, when K is locally
compact (A/m is then finite).
39l
I]! LIE GROUPS
392
[—1[—][—]
EXERCISES
where the symbol [(1.] denotes the integral part of the real number at. Show
that, if x is an integer prime to I, then
0,((x" — l)(:c"‘1 — l). . .(x — 1)) 2 r(l,n)
and that there is equality if the image of x in (Z/I‘Z)* is a generator of this
group.
(5) In the notation of Exercise 5 (0), show that 170.») is the greatest power
of l which divides all the 111(0), for p prime #1 (or for all sufficiently large p,
which amounts to the same). (Use ([1) applied to x = p; then choose [a such
that its image in (Z/I’Z) “ is a generator of this group, which is possible by the
arithmetic progression theorem.1')
(11) Let I‘ be a finite subgroup of GL(n, Q) and let I ' be the greatest power of
I dividing the order of I‘. Show that e S r(l, n). (Use Exercise 5 to prove that
l' divides all the a,.([1) and then apply (c) above.)
(2) Conversely, show that there exists a finite l-subg‘roup I‘M of GL(n, Q)
whose order is 1"“). (Reduce it to the case where n is of the form l“(l — l),
with a 2 0. Decompose Q” as a sum of l“ oopia of OJ" and use this decom-
position to give an operation on Q" of the semi—direct product H,_,, of the
symmetric group 5” and the group (Z/IZ "‘. Take Pl.» to be a Sylow l-group
of H. ,,.
Show) that, if n is even, I‘M, is contained in a conjugate of the symplectic
group 51201, 2).
(f) *Let I" be a finite subgroup ofGL(n, Q) which is an l—group. Show that
l" is conjugate to a subgroup of l- (Show first, using a suitable reduction
modulo 1;, that the reduction of I‘ is a finite subgroup of the reduction of a con-
jugate of I‘M; then use the character of the representation of I‘ on Q”.) In
particular, every subgroup of GL(n, Q) of order l"""’ is conjugate to I‘M”
fl 7. In this exercise, let 122(a) denote the 2-adic valuation of an integer n.
(a) Let I‘ be a finite subgroup ofGL(n, Q). Show that there exists a positive
definite quadratic form with coefficients in Z, which is invariant under 1".
Deduce (by the same argument as in Exercises 4 and 5) that, for all sufficiently
large p, l" is isomorphic to a subgroup of an orthogonal group 0(a) over the
field F, (resp. a subgroup of 50(n) if I‘ is contained in SL(n, (2)).
(b) Suppose that I‘ is contained in SL(n, Q) and let 2‘ denote the greatest
power of 2 dividing the order of I‘. Show that, if n is odd, 2‘ divides all the
integers
bn(11)=(l”‘“ —1)(17”'° -1)---(202 - l)
for sufficiently large prime [7. (Use (a) and Exercise 13 of Algebra, Chapter IX,
w
1‘ For a proof of this theorem, see for example A. SELEERG, Ann. of Math, 50
(1949), pp. 297—304.
393
III LIE GROUPS
Let A be an integer 23. Show that 2‘3”"l is the greatest power of 2 which
divides all the bn(p) for p 2 A. (Same method‘l‘ as in Exercise 6; use the
existence of a prime number p Z A such that p E 5 (mod 8).) Deduce the
inequality 5 < 7(2, 71) —— l.
(d) Conversely, let C,I be the subgroup of GL(n, Z) generated by the
permutation matrices and the diagonal matrices with coefficients 11. The
order of C" is 2"nl and 1/,(2‘71!) = r(2,n). If F3”. denotes the intersection of
a Sylow 2-group of C,I with SL(n, Z), deduce that the order of F3,” is
gram-r
8. Let n be an integer 21. We write
M(") = U 1r(l.n),
where the product is taken over all prime numbers I and the 7(l, n) are defined
as in Exercises 6 and 7.
Then M0) = 2, M(2) = 25.3 = 24, M(3) = 2‘.3 = 48,
M(4) = 27.325 = 5760.
Deduce from Exercises 6 and 7 that the ltc.m. of the orders of the finite
subgroups of GL(n, Q) (or GL(n, 2), which amounts to the same) is equal
to M(n). Consider the same question for SL(n, Q) with M(n) replaced by
§M(n).
9. Suppose that K is locally compact. Let G = GL(u, K). Let G1 be the
set of g e G which leaves stable a lattice of K'I with respect to A. Let G2 be
the set of g e G which generate a relatively compact subgroup in G. Let Ga
be the set of gEG whose eigenvalues in an algebraic closure of K are of
T For more details on this exercise and on the preceding one, see H. mowsxr,
Garamm. AWL, Leipzig-Berlin, Teubner, 1911 (Ed I, S. 212—218) and W. Burmsins,
Theory 41/ groups qffmilz order (2nd «1.), Cambridge Univ. Pras, 1911 (pp. 479—
484).
394
mam
of (A1, A2,...) EG such that Neil for l S m g n. The G,| are closed Lie
395
Ill LIE GROUPS
g(~) = if(xx)9(x)"dseMn(R)~
Then govt) = g(z)p(t) for x e G and 16 H. IfV is sufficiently small,
30‘) 6 GM": R)
for x sufl‘iciently close to 9. Finally use the fact that the theorem to be proved
is true locally for GL(n, R) and p(H).
6. Let G be a group of class 0' (§ 5, Exercise 1) with r 2 2. There exists
on G one and only one real Lie group structure S such that the underlying
manifold structure of class C" of S is the given structure. (The uniqueness of
S follows from Corollary 1 to Theorem 1. Let L(G) be the normable Lie
algebra associated with G in Exercise 2 of § 5. There exist a real Lie group
genn G' and an isomorphism h of L(G’) onto L(G) (§ 4, Theorem 3). Verify
as in §4, no. 1 that there exist a symmetric open neighbourhood G" of 26
in G’ and a mapping ([2 of class C' of G" into G such that Tabb) = h and
¢(g1gfl) = ¢(g1)¢(gg) for g1, g2 in 0”. By shrinking G”, it can be assumed
that V = MG") is open in G and that (i) is an isomorphism of class C' of the
manifold G” onto the manifold V. There therefore exists on V a real Lie
group germ structure such that the underlying manifold structure of class
C" is the given structure. For all g E G, Int g defines as analytic mapping of
Vn(g'1Vg) onto (n") nV (Theorem 1). By Proposition 18 of § 1,
there exists on G a real Lie group structure S inducing the same analytic
structure as V on an open neighbourhood of e. By translation, the underlying
manifold structure of class C' of S on G is the given structure.)
7. Let G be a real Lie group and H a closed subgroup of G.
(a) Let b be the set of x6 L(G) such that exp(tx) EH for all teR. Then
I) is a Lie subalgebra of L(G). (Use Proposition 8 of § 6.)
396
EXERUSES
§9
1. Let G = 80(3, R). Let A1, A2 be two orthogonal lines in R3 and H.
the subgroup of G consisting of the rotations about A, (i = 1,2). Then
[L(Hl), L(Hfl)] is a l—dimensional Lie subalgebra of G distinct from the
Lie subalgebra tangent to (H1, H2) at a.
2. Suppose that K is ultrametric. Let G be a finite—dimensional Lie group,
9 its Lie algebra and A a finite subset of G. Then ZG(A) is a Lie subgroup of
G with Lie algebra MA). (Argue as in Proposition 8.)
3. Let G be a connected real or complex Lie group. The centre Z of G is
a Lie quasi-subgroup of G and L(Z) is the centre of L(G).
4. Suppose that K is ultrametric and p > 0 (in the notation of § 7). Let
G be a finite-dimensional Lie group, A a group of automorphisms of G and
B the corresponding group of automorphisms of L(G). Let GA (resp. L(G)“)
be the set of elements of G (reap. L(G)) which are fixed under A (resp. B).
Then GA is a Lie subgroup of G with Lie algebra L(G)“. (Use the logarithmic
mapping)
5. Let G be a finite—dimensional connected real or complex Lie group.
Let (G0, G1, . . .) be the upper central series of G (Chapter II, §4, Exercise
18) and let (go, 9,, . . .) be the upper central series of the Lie algebra L(G)
(Chapter I, § 1, no. 6). Then, for all i, G. is a Lie subgroup of G such that
L(G.) = 9..
6. Let 1" be the 3-dimensional simply connected nilpotent real Lie group
defined in Exercise 5 (b) of §4. Let ueR be an irrational number. Let P
be the discrete subgroup of I‘ x R consisting of the ((0, 0, at), out), where
It 62. Let G = (I‘ x R)/P. Show that (G, G) is not closed in G.
7. (a) Let G be a semi-simple connected real Lie group, Z its centre and 9 a
continuous linear representation of G on a finite—dimensional complex vector
space. There exists an integer p such that, for all z e Z, 9(2) is diagonalizable
and all the eigenvalues of 9(2) are p—th roots of unity. (It can be assumed
that 9 is irreducible. Then 9(2) is scalar by Schur’s Lemma. On the other
hand, det 9(g) = l for all g EG because G = 9G.)
(b) Deduce that, if G admits an injective finite-dimensional continuous
linear representation, Z is finite.
397
11! LE GROUPS
p = dimtuc), L(G)].
Let (G, G) be given its structure as an integral subgroup of G. There exists a
neighbourhood V of z in (G, G) such that every element of V is the product
ofp commutators of elements of G. (Let x1, y1, . . ., x,” y, be elements of L(G)
such that the [x., 3].] form a basis of L(G). For x, t in R, we write
Ma 1) = (64» w.) "(eXP Ix.) “(EXP 5%) (CXP Ix.)-
Apply the implicit function theorem to the mapping
(‘1: 1'1, - - -7 5w ’p) H Pi(—“n ’1) - - - 9903» ’11)
oil!” into (G, G).)
398
EXERCISES
12. Let (:1, 22, 9;) be the canonical basis of K3. Consider the nilpotent Lie
algebra structure on K3 such that [51, 3,] = 23, [01, c3] = [92, :3] = 0. With
the associated group law on K3 (no. 5),
Show that a solvable real Lie group G is thus obtained such that
DG = R“ x {0}. The centre Z of G is {0} x {0} x 21:1.
(1') For (x, y, z) E G, let n(x, y, 2) be the mapping
(x,p.)»—>(Acosz- psinz+x,ksinz+p.cosz+y)
of R2 into R“. Show that 1!: is a morphism of G onto a Lie subgroup G’ of the
affine group of R7, generated by the translations and rotations of R’. Show
that Ker r: = Z.
(a) We canonically identify L(G) = T(o.o.0)G with R”. Let (:1, 2,, ‘5) be
the canonical basis of R“. Show that [21, :2] = 0, [93, 21] = 2,, [em :9] = -—:,.
((1') Show that, for 5 9E 0,
Deduce that, for all u eL(G), Z C exp(Ru). Show that expo: L(G) —> G is
neither injective nor suijective.
14. (a) Let G = GL(n, C). Show that expa is surjective. (Use the holo-
morphic functional calculus of Spam! Theories, Chapter I, §4, no. 8.)
(b) Let G’ = SL(2,C). Show that, if uses, the element (‘01 :1) of
G’ does not belong to the image of expfiu
399
III LIE GROUPS
a: cos1/_.I+sm—‘:—/—Z—: “>0.
0 . .
(b) Let H = SL(2,R). Show that g = (0 )‘_1)EH is the image of exp];
if and only if 7‘ > 0 or A = -1. If A = 0, all the one-parameter subgroups
containing g are equal.
16. Let G be a finite-dimensional Lie group. Show that there exists a
basis (x1, . . ., x“) of L(G) such that, for all i, exp(Rx,) is a Lie subgroup of G.
(Use Spectral Theories, Chapter II, §2, Lemma 1.)
1[ 17. (a) Let c denote a real solvable Lie algebra with a basis (a, b, 0)
such that [a, b] = :, [a,c] = —b, [b,c] = 0. Let b denote a real solvable
Lie algebra with abasis (a, b, c, 11) such that [a, b] = c, [a, c] = —b, [17, t] = d,
[a, d] = [[7, d] = [0, d] = 0. Let g be a real solvable Lie algebra. If 9 con-
tains non-zero elements a, y, 1 such that [x,y] = z and [x, z] = —_i/, then
g contains a subalgebra isomorphic to c or D. (Write a, = y, b, = z, :1 = [y, z] ;
define do bu 0: Inductively by a‘= [‘11— 1: ‘I— 1]: bl = [53— 1) [I— 1]; 5i: [11,, [’41
Consider the smallest integer I: such that 6,, = 0.
(b) Let g and 5* be real solvable Lie algebras and (l) a homomorphism of
9 onto g“. If 3* contains a subalgebra isomorphic to c or b, g has the same
property.
(e) Let I) be a complex solvable Lie algebra. Consider a Jordan—Holder
series for the adjoint representation of I); the quotients of this series define
l-dimensional representations of I) and hence linear forms on [7. These linear
forms, which depend only on I), are called the root: of I). If I)’ is a real solvable
Lie algebra, the roots of l)’ are the restrictions to l)’ of the roots of b’ ®n C
Then let G be a finite—dimensional simply connected solvable real Lie
group and 9 its Lie algebra. Show that the following conditions are equiva-
lent: (u) expa is injective; (fl) expG is suijective; (7) e2t is bijective; (8) expG
is an isomorphism of the analytic manifold L(G) onto the analytic manifold
G; (e) L(G) contains no subalgebra isomorphic to c or b; (0 there exists no
quotient algebra of L(G) admitting a subalgebra isomorphic to c; (1;) every
root of g is of the form <1; + io’ where (l), 42’ are in 9* and di’ is proportional
to 4); (0) for all are 9, the only pure imaginary eigenvalue (in C) of adx
is 0.T
t For more details, of. M. SAITO, Sur certains groupes de Lie resolubles, Sci.
Paper: af the College nf General Educaliun, Univ. of Tokyo, 7 (1957), pp. 1—11 and
157468.
400
EXERCISES
18. Let G be a connected real Lie group, Z its centre, 20 the identity
component of Z, 9 the Lie algebra of G and a the centre of 9. Then Z and Z0
are Lie quasi-subgroups of G with Lie algebra 3 (Exercise 3). An acceptable
norm on g is a norm defining the topology of g and making 9 into a normed
Lie algebra.
(n) For all r > O and z e 3, there exists an acceptable norm on 9 whose
value at z is <r. (Let g be an acceptable norm on 9. Show that, for suitably
chosen A > 0, the function x» |[x|| = Aq(x) + 1nl'q(x + y) has the required
properties.)
(b) Show that the following conditions are equivalent:
(i) 20 is simply connected;
(ii) for every acceptable norm on g, the restriction of expg to the open
ball of centre 0 and radius 1: is injective;
(iii) there exists 1 > 0 such that, for every acceptable norm on g, the
restriction of exp‘; to the open ball of centre 0 and radius r is injecfive.
(To prove (iii) => (i), use (a). To prove (i) => (ii), suppose that x, y are
in a, M < w, M < 7‘, x #y, expax = upay- Then up adx = EXP Edy
and hence ad x = ad y by Proposition 17 of § 6 applied to a. complexification
of 9. Hence there exists a non-zero z in a such that expuz = e; it follows
that (i) is false.)
(6) By considering the group G' of Exercise 13 (b), show that the con-
clusion of (b) is no longer one if u is replaced by a number > 1:. (In the nota-
tion of Exercise 13, use the norm at; + baa + as v—> (a2 + I?2 + £3)i on L(G’).)
{I 21. (a) Let G be a simply connected solvable real Lie group with the
following properties: (a) L(G), which is n-dimensional, has a commutative
ideal of dimension n — 1, corresponding to a subgroup A of G; (9) there
exists an element a of the centre of G which does not belong to A. Show that
401
m LIE GROUPS
there exists an element x of L(G) such that expat = cr. Show that L(G) is
the product of a commutative ideal and an ideal with a basis
(7‘; an bl! - - u a,” bk)
“Ch that "1» bn - - a ”In bl: belong to L(A); [’8 at] = 2“”tbo ll") bl] = ‘2‘"7‘4'1t
(n1 6 Z — (0}) for all i. Generalize the results of Exercise 13 (d) to G.
(b) Let G be a finite-dimensional simply connected solvable real Lie group.
Let D be a discrete subgroup of the centre of G. There exist a basis
(x1, x2, . . ., x") of L(G) and an integer r g n with the following properties: .
(at) every element of G can be written uniquely in the form
(exp tlxl) . . t (exp tux")
where :1, . . ., t,. are in R; (6) x1, . . ., x, are pairwise permutable and
(exp #1: . . ., exp x.)
is a basis of the commutative group D. (Argue by induction on the dimension
of G. Let a be maximal commutative ideal of L(G). Let A be the corre-
sponding integral subgroup. Then A is closed in G and DA/A is a discrete
subgroup of the centre of G/A, to which the induction hypothesis can be
applied, whence there are elements xi“, . . ., x; of L(G/A) and an integer :.
For 1 < i < r, let oi be an element of D whose class modulo A is exp xf.
Construct one by one, using (a), representatives x1,” .,x,,, of :4,” .,x;
such that exp x, = a, and [x,, :9] = 0 for l S i,j S r.)1'
(c) Deduce from (b) that every finite—dimensional connected solvable real
Lie group is homomorphie to a. space R” x T'“ (m, n integers 20).
22. Let G be a finite-dimensional connected real Lie group such that
L(G) is reductive. Then G = (V x S)/N where V is a finite-dimensional
real vector space, where S is a simply connected semi-simple real Lie group
and N is a central discrete subgroup of V X S. Then G/fiG is isomorphic
to V/prl N. Hence G/D‘N is compact if and only if pr, N generates the
vector space V.
23. (a) Let G be a finite-dimensional commutative real Lie group with
only a finite number of connected components. For G to be compact, it is
necessary and suflicient that every finite-dimensional analytic linear repre-
sentation of G on a complex vector space he semi-simple. (Use the proof of
Proposition 32.) ,
(b) Let G be a finite-dimensional commutative complex Lie group with
only a finite number of connected components. Let N be the kernel of exp“.
For N to generate over C the vector space L(G), it is neceSary and sufficient
that every finite-dimensional analytic linear reprtsentation of G be semi-
simple. (Use (4), Lemma 1 and Proposition 32.)
24'. The Lie group SL(2,R) is connected and almost simple, but (I, —I)
is a normal commutative subgroup of SL(2, R).
25. Let G be an almost simple connected real or complex Lie group. Let
A be a normal subgroup of G. If A # G, A is discrete and central. (Use
Exercise 8 of § 4'.) Therefore the quotient of G by its centre is simple as an
abstract group.
1I26. Let G be a finite-dimensional connected real Lie group. Suppose
that G admits a finite-dimensional injective continuous linear representation
p. Then (G, G) is closed in G. (Let R be the radial of G and S a maximal
semi-simple integral subgroup of G. Using Exercise 7 (b), reduce it to the
case where G is the semi-direct product of S and R. By Chapter I, § 6, Proposi-
tion 6, p is unipotent on (G, R). Hence p((G. R)) is closed in the linear group
and therefore (G, R) is closed in G.) '
{[27. Let G be a finite-dimensional solvable simply connected real Lie
group and N the largest connected nilpotent normal subgroup of G. Then G
admits a finite-dimensional injective continuous linear representation which
is unipotent on N. (Argue by induction on the dimension of G. Use Proposi-
tion 20 and Chapter I, § 7, Theorem 1.)T
{I 28. (a) Let k be a commutative field of characteristic 0, L an n—dimen-
sional Lie algebra over k, L‘ an (n — l)-dimensional subalgebra containing
no non-zero ideal of L and a0 an element of L not belonging to L]. For
i = 2, 3, . . ., we define Li inductively to be the set of xeL._1 such that
[3, aa]eLi_1. We write L, = L for i < 0. Show that [L4,L,] C L”, (by
induction on i + j) and then that, for O < i S n, L‘ is a subalgebra of L of
codimension i (by induction on i).
([7) For 0 < 1' < 71, Choose in L. an element a, not belonging to L...1 such
that [anJ a.] E {111-1 (mod. L). Show, by induction on the ordered pairs
(1', j) ordered lexicographically, that, for 0 S i < j < n, i + j — l < n and
[0,, (1,] E (J ’ i)at+;—r(m°d-Ll+y)-
(c) Deduce that L is either of dimension 1, or of dimension 2 and non-
commutative, or isomorphic to 61(2, k).
(11) Let A be a finite-dimensional real Lie algebra. A subalgebra B of A
is called extendable if there exists a subalgebra B1 3 B such that
403
ll! LIE GROUPS
dimB1 = dim B + 1. Let R'(A) denote the intersection of all the non-
extendable subalgebras of A. Let R(A) denote the largest ideal of A with a
composition series as an A-module all of whose quotients are of dimension 1.
Show that R'(A) is a characteristic ideal of A. (Observe that R’ (A) is stable
under Aut(A).)
(2) Show that R'(A) D R(A) and that R’(A]R(A)) = R’(A)/R(A).
(f) Show that, if B is a subalgebra of A, then R’(B) D R’ (A) n B.
(g) If A is solvable, R’(A) 2: R(A). (It can be assumed that R(A) = {0},
by (5). Suppose that R’(A) aé {0}. By (d), there exists a minimal non-zero V
ideal I of A contained in R’(A). Using (f) and Chapter I, §5, Corollary 1
to Theorem 1, show that dim I = l, whence a contradiction.)
()1) Let P be the radical of R'(A), B a. semi-simple subalgebra of A and
C = P + B which is a subalgebra. of A by ((1). Using (f), show that ad
can be expressed in a triangular form with respect to a suitable basis of P.
Conclude that [P, B] = {0}.
(k) R(A) is the radical of R’(A). (Use (c), (f), (g), (h) and a Levi de-
composition of A.) Let R’(A) = R(A) 9 T be a Levi decomposition of R'(A).
By (h), R'(A) = R(A) x T. Then R(T) = {0}. Applying (a) to the simple
factors of T, conclude that T = {0}. It has therefore been shown that
R(A) = R’(A).
(I) Let G be a connected real Lie group with Lie algebra A. Let .9’ (G) be
the set of subgroups N of G such that there exists a decreasing sequence
(NM, NM“. . ., No) of subgroups with the following properties: m = N,
N0 = {5), each N. is a normal connected Lie subgroup of G and
dim Nt/N,_l = 1 for all i > 0. Show that the integral subgroup R(G) of
G with Lie algebra R(A) is the largest element of 3’(G). (Argue by induc-
tion on dim R(A). Let I be a l-dimensional ideal of A and N the correspond-
ing integral subgroup of G. Pass to the quotient by N. If N is not closed, use
Exercise 21 (a) of § 6.)
(m) A Lie subgroup H of G is called extendable if there exists a Lie sub-
group H1| D H such that dim Hl = dimH + 1. Let R'(G) denote the inter-
section of all the non-extendable connected Lie subgroups of G.
Let B be a non—extendable subalgebra of A. Show that the corresponding
integral subgroup of G is closed. (Use Proposition 5.) Deduce that
L(R’(G)) C R(A).
whence R'(G) C R(G).
(n) Show that R(G) = R'(G). (Reduce it to the case where R’(G) = {9}.
Then use Exercise 22 of § 6.)?
11 29. A real Lie group G is said to be of type (N) if it is finite—dimensional,
1‘ For more details, cf. J. Tris, Sur une classe de groupes de Lie résolubles,
Bull. Soc. Mam. B:lg., 11 (1959), pp. 100—1 15 and 14 (1962), pp. 196—209.
404
MERGERS
nilpotent, connected and simply connected. IfG is such a group and g is its Lie
algebra, the mapping exp: g—>G is an isomorphism when 9 is given the
group structure defined by the Hausdorff law (cf. Chapter II, §6, no. 5,
Remark 3). Let log: G —> g denote the 1nverse isomorphism.
(a) Let V be a vector Q—subspace ofg. Show the equivalence ofthe following
conditions.
(i) V is a Lie Q—subalgebra of 9;
(ii) exp(V) is a subgroup of G.
(Use Exercise 5 of § 6 of Chapter II.)
For a subgroup H of G to be of the form exp(V), where V is a vector Q-
suhspace of 5;, it is necessary and sufficient that H be saturated in G (Chapter
II, § 4-, Exercise 14), i.e. that the relations a: e G, x" e H, n sé 0 imply x e H (Inc.
at). Ifthis is so, show that H is an integral subgroup of G if and only iflog(H)
is a Lie R—subalgebra of g.
(12) Let V be a Lie qbalgebra of 9 of finite dimension m, let (tb. . ., cm)
be a basis of V over Qand let A be the subgroup of V generated by s,,. . ., 2",.
Using the fact that the Hausdorff law 15 polynomial show that there exists an
integer 11 > 1 such that, for every integer r which IS a non-zero multiple of J,
exp (rA) is a subgroup of G. Let :1 be such an integer and r a multiple ON. Show
that exp(1A) is discrete if and only if (:1, . . . , a...) is a free family over R, i.e. if
the canonical mapping of V ‘80 R into g is infective. Suppose that this is the
case; show that G/exp(rA) is compact if and only if V 80 R —> g is bijmive, i.e.
ifV i: a Qfarm 1yr 9 (To show that the condition is suflicient, argue by induc-
tion on the nilpotency class of g )
(6) Conversely, let I‘ be a discrete subgroup of G, let I‘ be 1ts saturation in
G (Chapter II, lac. cit.) and let 9“ = log(I‘)= Q. log([‘) be the corresponding
Lie Q;suba.lgebra. Suppose that R. g = g, i.e. that I‘ is Contained in no distinct
integral subgroup of G. Let 2 be an element 761 of the centre of I‘ and let
x = log (2). Show that it belongs to the centre ofg. If X = exp(Rx), show that
X/(I‘ n X) is compact and that the image of I‘ in G/X is discrete. Deduce,
arguing by induction on dim G, that G/I‘ is rampart and that g, is a G-fvrm ofg.
If A is a lattice of 9,, show that there exists an integer d sé 0 such that 1" con-
tains exp(dA) and that the index of exp(dA) in F is then finite.
Let H be an integral subgroup of G with Lie algebra b. Show that H/ (H n I‘)
is compact if and only if I) is rational with respect to the Qstructure g, (in
particular if I) is one of the terms of the lower—or upper—central series of g).
(41) Let I‘ be a discrete subgroup of G, let H be the smallest integral sub-
group of G containing I‘ and let I) be its Lie algebra. Show that H/I‘ is compact
and that g ®qR—>g is injective and has image I). (Apply (a) to the nil-
potent group H.)
(e) Let I‘ be a discrete subgroup of G. Show that there exists a basis
(x1, . . ., x“) of L(G) with the following properties:
(i) for alliE[l,q),Rx. +--- + Rx. is an ideal n, of._1 +~~~+ Rx,;
405
"1 LIE GROUPS
(ii) there exists [7 e [1, q] such that I‘ is the set of products
406
EXERCISES
407
m LIE GROUPS
408
EXERCISES
§1o
1. Let G be a finite-dimensional connected real Lie group. Show that the
canonical mapping of Aut(G) into Aut(L(G)) is not in general sutjective
(take G = T).
2. Suppose that K = Q7. Let G be a finiteudimensional Lie group. Show
that the following conditions are equivalent:
(a) there exist x1, x,, . . . , x" in G such that the subgroup of G generated by
{2:1, . . ., x”) is dense in G;
(b) G is generated by a compact subset.
(To prove that (b) implies (a), observe that, if (cl, . . ., 9,.) is a basis of L(G),
(exp 2,51) (exp 2,22) . . . (exp 2,2,.) is a neighbourhood of e.)
3. Let G be the set of (1,3,) E Q, X Q, such that |y| s 1. This is an open
subgroup of Q,1 x Q,.ThenL(G) = Q, X Q,.Letotbetheinfinitesimalauto—
morphism (x, y) )—> (O, x). Then the conclusion of Proposition 3 is false.
(4) Let K be a quadratic extension of QF and let oi E K — Q,” Let
G = Q, + 2,0), considered as a Lie group over K. The only automorphisms
of G are the (x, y) r—> (M, y) where )t is invertible in Zp. The set of these auto-
morphisms cannot be given a Lie gmup structure over K with the properties
of Theorem 1.
HISTORICAL NOTE
(Chapters I to III)
I. GENESIS
The theory, called for nearly a century “theory of Lie groups”, was essentially
developed by one mathematician: Sophus Lie.
Before embarking on the history, we summarize briefly various earlier
research which prepared the way.
(a) Tramflrmaliun group: (Klein-Lie, 1869—1872)
About 1860 the theory of permutation groups of afim‘tz set was developing and
beginning to be used (Semt, Kronecker, Mathieu, Jordan). On the other
hand, the theory of invariants, which was in full flight, was familiarizing
mathematicians with some infinite sets of geometric transformations stable
under composition (notably linear or projective transformations). But before
the work ofJordan [7] in 1368 on “groups ofmovements” (closed subgroups of
the group of displacements of 3-dimensional Euclidean space), no conscious
link seems to have been established between these two currents of ideas.
In 1869, the young Felix Klein (1849—1925), a pupil of Plficker, established
a friendship in Berlin with the Norwegian Sophus Lie (1842—1899), who was a
few years older, brought together by their common interest in Plficker’s “line
geometry” and in particular the theory of line complexes. It was about this
time that Lie conceived one of his most original ideas, the introduction of the
notion of invariant into Analysis and Differential Geometry; one of the sources
was his observation that the classical methods of integration “by quadratures”
of difl‘erential equations depended entirely on the fact that the equation is
invariant under a “continuous” family of transformations. 1869 is the date of
the first work (edited by Klein) where Lie used this idea ; there he studied the
“Reye complex” (set of four lines cutting the faces of a tetrahedron in four
points with a given cross—ratio) and the curves and surfaces which admit as
tangents lines of this complex [3 a]: his method depends on the invariance of
the Reye complex under the 3-parameter commutative group (maximal
410
HISTORICAL NOTE
torus of PGL(4, (3)) which leaves the vertices of the tetrahedron invariant.
The same idea dominated the work Klein and,Lie wrote together when
they were in Paris in the spring of 1870 [1, a]; there they essentially deter-
mined the commutative connected subgroups of the projective group of the
plane PGL(3, C) and studied the geometric properties of their orbits (under
the name of curves or surfaces V); this gave them, by a uniform procedure,
properties ofvarious curves, algebraic'or transcendental, such as y = ox” or the
logarithmic spirals. Both works served to underline the profound impression
made on them by the theories of Galois and Jordan (jordan’s commentary on
Galois had appeared in Math. Annalm in 1869; on the other hand, Lie had
heard talk of Galois theory as early as 1863). Klein, who in 1871 became
interested in non-Euclidean geometries, saw there the start of his research on a
classification principle for all known geometries which was to lead him in 1872
to the “Erlangen Programme”. For his part, Lie, in a letter of 1873 to A.
Mayer ([3], vol. V, p. 584-), dated his ideas on transformation groups from his
sojourn in Paris and in a work of 187] ([3 b] p. 208) he was already using the term
“transformation group” and explicitly posed the problem of the determination
of all subgroups (“oontinuous or discontinuous”) of GL(n, C). To be truthful,
Klein and Lie had both experienced some difficulty in entering this new
mathematical universe and Klein spoke ofJordan’s newly appeared “Treatise”
as a “book realzd with room Halo” ([2], p. 51); he wrote moreover concerning
[1 a] and [l b]: “To Lie belong: all the omiilfor the heuristic idea of a continuum;
group of r ' of dflmnttal equationr
and partial derivatives. All the notion: he developed later in his theory of continuous
group: were already there in embryo, but were however so little elaborated, that only after
long conversation: could I oorwince him of many details, for example to brgin with the
wry existence ofth: auras: V” ([2], p. 415).
(b) Iryinitmmal transfirmalion:
The concept of an “infinitely small” transformation goes back at least to the
beginnings of lnfinitesimal Calculus; we know that Descartes discovered the
instantaneous centre of rotation when admitting that “in the infinitely small”
every plane movement can be likened to a rotation: the elaboration of Analy-
tical Mechanics in the 13th century is entirely founded on similar ideas. In
1851, Sylvester, seeking to form invariants of the linear group GL(3, C) and
some ofits subgroups, gave the parameters appearing in his matrices “infinitely
small” increases of the forms nod! and expressed the fact that a functionf ((2,))
was invariant by writing the equationf((z, + u,(dt)) = f((z,)); this gave him
forf the linear partial differential equation Xf = 0, where
(1) Xf — af
_ 2,: a, 5?.
411
HISTORICAL NOTE
direction parameters at,” ([5], vol. 3, pp. 326 and 327); Sylvester seemed to
think that here was a general principle of considerable importance but appears
never to have returned to the question. A little later, Cayley ([6], vol. II,
pp. 164—178) proceeded similarly for the invariants of SL(2, C) under certain
representations of this group and showed that they are the solutions of two first
order partial differential equations Xf = 0, Yf 2 0, where X and Y are ob-
tained as above from “infinitely small” transformations
0 0 d 0 dt
(.1: 0) a" (0 0)‘
In modern terms, this is expressed by the fact that X and Y generate the Lie
algebra al(2, C); moreover Cayley calculated the bracket XY — YX explicitly
and showed that it was also derived from an “infinitely small” transformation.
In his memoir of 1868 on groups of movements [7],Jordan used from begin—
ning to end the concept of “infinitely small transformation”, but exclusively
from a geometric point of view. He is no doubt responsible for the idea of a
one-parameter group “generated” by an infinitely small transformation: for
Jordan, it was the set of transformations obtained by “suitably repeating” the
infinitely small transformation (lac. cit, p. 243). Klein and Lie, in their memoir,
used the same expression “repeated iyiniwly small transfinnatiun” [1 b], but the
context shows that they understood by that an integration of a differential
system. If the one-parameter group they considered consists of transformations
x’ = f(x, y, t), y' = g(x, y, t), the corresponding “infinitely small transfor-
mation" is given by
11* =P(x,y)dt: dy = 90w)!“
3
where Mm) = gm, to), «(m = 2% (m, to) and to corresponds to the
identity transformation ofthe group. As Klein and Lie knew the functionsfand
g explicitly, they had no difliculty in verifying that the functions
412
HISTORICAL NOTE
first order partial differential equations and the relations between these two
theories. We are not concerned here with the history ofthese questions and we
shall confine ourselves to mentioning a few points which seem to have played
an important role in the genesis of the theory of transformation groups.
The notion of contact transformation generalizes both point transformations
and inverse polar transformations. Roughly a contact transformation'l' on
C” is an isomorphism of an open subset Q of the manifold T'(C") of cotangent
vectors to C“ onto another open subset Q’ of T’(C") mapping the canonical
l-form of!) to that of Q’. In other words, if (x1, . . ., xmpl, . . .,p,,) denote the
canonical coordinates of T'(C"), a contact transformation is an isomorphism
(an, 9,) n—> (X‘, Pi) satisfying the relation ; 1’,X = ; p,dx.. Such transfor-
mations occur in the study of the integration ofpartial differential equations of
the form
31 32
(2) F(x1,x2,...,xu,a—n,...,wn) =0.
Lie became familiar during his research on these questions with the manipula-
tion of Poisson brackets
” afag age
(3) (fig)= giaTaE'a—méi
and the brackets: [X, Y] = XY — YX of differential operators of type (1) ; he
interprets the Poisson bracket (3) as the effect onf of a transformation of type
(1) associated with g and observes on this occasion that the Jacobi identity for
Poisson brackets means that the bracket of the differential operators corres-
ponding tofand g is associated with the bracket (g, h). Research into functions
g such that (F, g) = 0, which occurs in Jacobi’s method for integrating the
413
HIS'I‘OIUCAL NOTE
partial differential equation (2), became for Lie the study of infinitesimal con-
tact transformations which leave the given equation invariant. Finally,
Lie was led to study sets of functions (11,)1 “5,, of the x; and pl such that the
brackets (14,, uk) are functions of the uh and called these sets “groups” (they
had already essentially been studied by Jacobi) [3 c]i
Proceeding as in his memoir with Klein, Lie integrated the differential system
(7) d at = ...= d in = dt,
2 extra”. . ., at) Z ztxmel, . . ., an)
1' By this Lie meant that the f, could not be expressed in terms of less than 9‘
functions of the n,, or also that the jacobian matrix (afi/Ba,) is “in general” of
rank 7.
1 In his first notes, Lie thought that he could prove a priori the existence of the
identity and the inverse in the whole set of transformations (4) stable under
composition; he recognized later that his proof was incorrect and Engcl provided
him with a counterexample reproduced in [4-], vol. 1, § 44-. However, Lie showed
how “continuous" systems (4) stable under composition could be reduced to group
germs of transformations: such a system is of the form G o h, where G is a group
germ of transformations and h a transformation of the system ([4], vol. 1, Theorem
26, p. 163 and vol. 3, Theorem 46, p. 572).
4-14-
HISTORICAL NOTE
which gave him, for each point (:1, . . ., 2,) a one—parameter group
(8) “”‘i=gli‘nn-axmlnnuZn‘) (1 <i<nl
such that g‘(xl, . . ., x,“ 11, . . ., 2,, 0) = x, for all i. He showed by an ingenious
method, using the fact that the transformations (4) form a set which is stable
under composition, that the one-parameter group (8) is a subgroup of the given
group [3 d]. The new idea, the key to the whole theory, is to take the Taylor
expansions of the functions (4) to the second aid”. The progress of his argument
was quite confused and heuristic ([3 d] and [3], vol. 5, pp. 600—601) ; it can be
presented as follows. For sufficiently small 1,, let t = l in (8); thus new para-
meters zl, . . ., z, are obtained for the transformations of the group (this is in
fact the first appearance of the “canonical parameters”). Then by definition,
using (7),
flag—t. =2 zkxktofiau -: xvi)
whence
0’
3f;‘= gab—S:—“(A ,4) 3x’37’
= 22* 6X377% a4.)(zz»xuo4,...,xn))
7
which gives
+ <§~( knit!
Z zkzn—
5:“(run ., x,)X,,,(x,, . . ., x,,))ti + - - -,
where the terms omitted are not linmr in u or 1;. Transforming (10), using (9)
and (11) and then comparing the terms in uhnk of the two sides. Lie obtained
the relations
" '9i 5X“) .
(12) ’21 (XM ax, — H ex, = 1;: cmxu (1 g h,k < 7,1 g 1 s 71).
His experience with the theory of partial differential equations led him to
write these conditions in a simpler form: following the pattern of (l), he
associated with each of the infinitely small transformations obtained by setting ‘
z): = l, 2,, = 0 for h aé k in (6), the differential operator
.
af
Arm = 2 xi. a7;
(13)
and rewrote conditions (12) in the form
(14) mm = Z c...A.,
the corner stone of his theory. Until then he used the terms “infinitely small
transformation" and “infinitesimal transformation” indiflermfly (e.g. [3 e]);
the simplicity of relations (14) led him to call the operator (13) the “symbol”
of the infinitesimal transformation dx‘ = dt (l < i S n) [3 11] and very
soon he called the operator (13) itself the “iry‘initesimal tramfirmafiun” ([3 d] and
[3], vol. 5, p. 589).
He then became aware of the close links which united the theory of “con-
tinuous groups” with his earlier research on contact transformations and par-
tial differential equations. This bringing together filled him with enthusiasm:
“My earlier workr were a: it were all ready there waiting to found the new theory 13f
transfuimatian groups” he wrote to Mayer in 1874- ([3], vol. 5, p. 586).
In the following years, Lie pursued the study of transformation groups.
Besides the general theorems summarized below (§ III), he obtained some more
special results: the determination of the transformation groups of the line and
plane, the subgroups of low oodimension in the projective groups, the groups
in at most 6 parameters, etc. He did not abandon diflerential equations for
long. In fact, it seems that, for him, the theory of transformation groups was a
tool for integrating differential equations, where the transformation group
played a role analogous to that of the Galois group of an algebraic equations!"
T This research had little influence on the general theory of differential equa-
tions, since the automorphism group of such an equation is usually trivial. As a
compensation, for certain types of equations (for example linear equations),
interesting results have been obtained later by Picard, Vasiot and, more recently,
Ritt and Kolchin.
416
HISTORICAL NOTE
We note that this research led him equally to the introduction of certain trans-
formation sets in an infinity of parameters which he called “infinite continuous
groups”1; he reserved the name “finite continuous groups” for transformation
groups in a finite number of parameters of type (4) above.
1 Today they are called “Lie pseudo-groups”; care should be taken not to con-
fuse them with the “Banach” Lie groups defined in this volume.
§ From 1886 to 1898, Lie occupied the chair at Leipzig left vacant by Klein
and had Engel as his assistant; this circumstance favoured the production of an
active mathematical school as well as the diHusion of Lie’s ideas, so little known
until then (notably because of the fact that his first memoirs were usually written
in Norwegian and published in the Comptes Rendus de l’Académie cle Christiania,
which was little used elsewhere). Thus at a time when it was unusual for young
French mathematicians to go to Germany for instruction, E. Vessiot and A. Tresse
spent a year studying at Leipzig with Sophus Lie.
1' The analogous notion for permutation groups had been introduced and
studied by Jordan in his “Truths”.
4-17
HISTORICAL NOTE
of the results and they explicitly reaffirmed the “local” point ofview whenever
necessary (cf. loo. cit, p. 168 or 189 for example or flirt, vol. 3, p. 2, note at the
bottom of the page); in other words, the mathematical object they studied is
close to what we call in this treatise a law chunk of operation. They did not
refrain, on occasions, from considering global groups, for example the 4 series
of classical groups ([4], vol. 3, p. 682), but do not appear to have asked them-
selves the question of what in general constitutes a “global group”; they were
content to obtain, for the “parameters” of the classical groups (the “variables”
of these groups introduced no difficulty, since the transformations in question
are linear transformations ofC") “local” systems of parameters in a neighbour-
hood of the identity transformation, without worrying about the domain of
validity of the formulae they were writing clown. They however set themselves
a problem which arises neatly out of the local theoryi: the study of “mixed”
groups, that is groups with a finite number of connected components, such as
the orthogonal group ([4], vol. 1, p. 7). They presented this study as a study of
a set of transformations stable under composition and passage to the inverse
which is the union of sets H, each of which is described by systems of functions
(f‘m) as in (4-); the number of (essential) parameters of each H, is even a [mini
assumed to depend on j, but they showed that in fact this number is the same
for all the H,. Their principal result was then the existence ofa finite continuous
group G such that H, = G n It, for some 11, e H, and for all j; they also estab-
lished that G is normal in the mixed group and noted that the determination
of the invariants of the latter reduces to that of the invariants of G and a dis-
continuous group ([4], vol. 1, Chapter 18).
The general theory developed in [4] ended (without the authors saying so
very systematically) by achieving a “dictionary” translating the properties of
“finite continuous” groups into those of the set of their infinitesimal transfor-
mations. It is based on the “three theorems ofLie”, each ofwhich consists of an
assertion and its converse.
Thefirst theorem ([4], vol. 1, pp. 33 and 72 and vol. 3, p. 563) affirms in the
first place that if in (4) the parameters are effective, the functions f; satisfy a
system of partial differential equations of the form
418
IHS’I‘ORICAL NOTE
funetionsfi have this property, formulae (4) define a group germ of trans-
formations.
The second theorem ([4], vol. I, pp 149 and 158, and vol. 3, p. 590) gives rela-
tions between the E“ on the one hand and the 41,, on the other: the con-
ditions on the in can be written in the form
419
HISTORICAL NOTE.
1' A similar evolution can be pointed out in the theory of “abstract” groups, in
particular finite groups. They were first defined as transformation groups, but
Gayley had already noted that what is essential is the way the transformations
are composed with one another and not the nature of the concrete representation
of the group as a permutation group of particular objects.
420
HISTORICAL NOTE
T Lie had already stated without proof a result of this type ([3 g], no. 7). He had
been led there by his research on the foundations of geometry (“Helmholz’s prob-
lem"), where he had remarked that the analyticity hypotheses are not natural.
The result of F. Schur was to lead Hilbert in 1900 to ask if the same conclusion
remained true if it was only assumed that the f. were continuous (“Hilbert’s 5th
problem”). This problem has stimulated much research. The most complete
result along this line is the following theorem, proved by A. Gleason, D. Mont-
gomery and L. Zippin; every locally compact topological group has an open sub-
group which is an inverse Limit of Lie groups; it implies that every locally Euclidean
group is a Lie group. For more details on this question, cf. D. MONTGOMERY and
L. ZIPPIN [41].
421
HISTORICAL NOTE
T The term “Lie algebra" was introduced by H. Weyl in 1934: in his work of
1925, he had used the expression “infinitesimal group”. Earlier, mathematicians had
spoken simply of the “infinitesimal transformations X, f, . , ., X, f ” of the group,
which Lie and Engel frequently abbreviated by saying “the group Xx]; . . ., Xff” !
3 Almost at the beginning of his research, Lie had encountered solvable linear
groups and even in fact nilpotent linear groups [3 h].
422
HISTORICAL NOTE
§ Up to the fact that he found two exceptional groups of dimension 52 but did
not notice that they were isomorphic. (He only considered complex simple Lie
algebras for the more general problem had not been considered at that time;
Killing’s methods are in fact valid for any algebraically closed field of charac-
teristic 0).
1‘ Cartan’s approach consisted of studying the non-trivial extension Lie algebras
of a simple Lie algebra and a (commutative) radical of minimal dimension.
423
HISTORICAL NOTE
1' H. Weyl remarked on this occasion that the construction given by E. Cartan
for irreducible representations implicitly uses this property.
1 The first use of difi‘erential operators of higher order generated by the Xl
was no doubt the use of the “Casimir operator” for the proof of the complete
reducibility theorem. After 1950, the research of Gelfand and his school and
Harish-Chandra, on infinite-dimensional linear representations, has brought
these operators into the forefront.
424
I‘USTDRICAL NOTE
point of view. Engel also showed that the exponential is surjective for the
other classical groups augmented by homothetis; this work was taken up and
pursued by Maurer, Study and others, without producing substantial new
results.
In 1899, H. Poincare ([14], vol. 3, pp. 169—172 and 173—212) embarked
upon the study of the exponential mapping from a difl‘erent point of view. His
memoirs seem to have been hastily edited, for in several places be affirmed that
every element of a connected group is an exponential, whereas he gave
examples to the contrary elsewhere. His results were mainly concerned with the
adjoint representation: he showed that a semi-simple element of such a group
G may be the exponential of an infinity of elements of the Lie algebra L(G),
whereas a non-semi-simple element may not be an exponential at all. If
ad(X) has no eigenvalue which is a non-zero multiple of 2113', then exp is étale
at X. He also proved that, if U and V describe loops in L(G) and W is defined
by continuity such that c".e" = W, it is possible not to return to the original
value of W. He used a residue formula which amounts essentially to
_1 ¢(E)dE
d>(adX)_2—1;f————E_adx
where ad(X) is a semi-simple element whose non-zero eigenvalues are ofmulti-
plicity l, (D is an integral series with sufficiently large radius ofconvergence and
the integral is taken over a contour surrounding the eigenvalues of ad X; he
also studied what happens as X tends to a transformation with multiple eigen-
values.
Research into expressions for W as a function of U and V in the formula
2".” = :W had already, just before Poincaré’s work, been the object of two
memoirs by Campbell [13]. As Baker wrote a little later “. . . Liz theury rugged:
in an abvious way that the product :"zV i: theform ,w where W is a series ry‘altemnznlr in
U and V . . .”. The later works on this subject aimed at making this assertion
precise and giving an explicit formula (or a method of construction) for W
(“Hausdorff formula”). After Campbell and Poincare, Pascal, Baker [15] and
Hausdorff [16] returned to the question ; each considered that the proofs of his
predecessors were not convincing; the principal difficulty resided in what is
meant by “alternants”: are they special elements of the Lie algebra in ques-
tion, or universal ”symbolic” expressions? Neither Campbell, nor Poincare,
nor Baker expressed himself clearly on this point. Hausdorff’s memoir, on the
other hand, is perfectly precise; he worked first on the algebra of (non-
commutative) associative formal power series in a finite number of indeter-
minates and considered U, V, W as elements of this algebra. He proved the
existence of W by a differential equation argument analogous to that of his
predecessors. He used the same argument to prove the convergence of the
series when the indeterminates are replaced by elements ofa finite—dimensional
Lie algebra. As Baker had remarked, and Poincaré independently, this result
425
HMORICAL NOTE
could be used to give a proof of Lie’s third theorem; he clarified the correspon-
dence between Lie groups and Lie algebras, for example where the commuta-
tor subgroup is concerned. _ ‘
In 1947, Dynkin [39] again took up the question and obtained the exphcrt
coefficients of the Hausdorfl" formula, by considering from the outset a normed
Lie algebra (of finite dimension or otherwise, over R, C or an ultrametric
field).T
T In the ultrametric case, the classical method of upper bounds cannot be ex-
tended without precautions hecause of the asymptotic behaviour of the padic
absolute value of 1 In as It tends to infinity.
: H. Weyl did not define explicitly this nation, with which he had been familiar
since the editing of his course on Riemann surfaces (1913). It was 0. Sehreier
[22] who in 1926—1927 gave, for the first time, the definition of a topological group
and that of a “continuous” group (i.e. locally homeomorphic to a Euclidean
space), and the construction of the universal covering of such a group.
4-26
HISTORICAL NOTE
428
HISTORICAL NOTE
VIILFREELIEALGEBRAS
It remains to speak of a series of works on Lie algebra; where the connection
with the theory of Lie group: is very tenuous; this research has on the other
hand important applications to the theory of “abstract” groups and more
especially nilpotent groups.
Its origin is the work of P. Hall [24], which appeared in 1932. However Lie
algebras are not discussed here: P. Hall had in mind the study of a certain
class ofp-groups, those which he called “regular”. But this led him to examine
in detail iterated commutators and the lower central series of a group; on this
subject he established a version oftheJacobi identity (cf. Chapter II, § 4-, no. 4,
formula (20)) and the “Hall formula”
(xi/yI = fly“(x, y)’“‘"‘”’. . . (cf. Chapter II, § 5, Exercise 9).
Almost immediately (in 1935—1937) appeared the fundamental works of W.
Magnus ([25 a] and [25 b]) and E. Witt [30]. In [25 a] Magnus used the same
algebra offormal power series A as Hausdorff (since called “Magnus algebra") ;
he embedded the free group F in it and used the natural filtration of A to
obtain a decreasing sequence (Fn) of subgroups of F; it is one of the first
examples of a filtratiun. He conjectured that the F“ coincide with the terms of
the lower central series of F. This conjecture was proved in his second memoir
[25 b]; also in this work he showed explicitly the close relation between his
ideas and those of P. Hall and defined the free Lie algebra L (as a subalgebra
of A) which he showed essentially to be identified with the graded algebra
of F. In [30], Witt completed this result on various poinm. He showed notably
that the enveloping algebra of L is a free associative algebra and deduced
immediately the rank of the homogeneous components ofL (“Witt formulae”).
As for the proof of the basis of L known as the “Hall basis” (cf. Chapter 11,
§ 2, no. 11), it seems that it only appeared in 1950 in a note of M. Hall [40],
although it was implicit in the works of P. Hall and W. Magnus quoted
above.
429
BIBLIOGRAPHY
1. F. KLEIN and S. Luz: (a) Sur une certaine famille de courbes et surfaces,
C. R. Ami. Sci, 70 (1870), pp. [222—1226 and 1275—1279 (=[2], pp.
416—420 and [3], vol. 1, pp. 78—85); (b) Uber diejenigen ebenen Kurven,
welche dutch tin geschlossenes System von einfach unendlich vielen
vertauschbaren linearen Transformationen in sich fibergehen, Math.
Arm, 4 (1871), pp. 50—84 (=[2], pp. 424—459 and [3], vol. 1, Abh. XIV,
pp. 229—266).
2. F. KLEIN, Gemmmelle mathematiyclle Abhandlungen, Ed. I, Berlin (Springer),
1921. '
3. S. LIE, Gesammlte Abhandlungm, 7 VOL, Leipzig (Teubner): (a) Uber die
Reziprozitfitsverhfiltnisse des Reyeschen Komplexes, vol. I, Abh. V,
pp. 68—77 (=Gétl. Natl]. (1870), pp. 53—66); (b) Uber einc Klasse
geometrischer Transformationen, vol. I, Abh. XII, pp. 153—214
(=Chn‘5tiana For. (1871), pp. 182—245); (c) Uber partielle Differential-
gleichungen erster Ordnung, vol. III, Abh. VII, pp. 32—63 (=Chn'xlz'anu
Far. (1873), pp. 16—51); ((1) Theorie der Transformationsgruppen 11,
vol. V, Abh. III, pp. 42—75 (=Arthivf. Math., 1 (1876), pp. 152—193);
(J) Ueber Gruppen von Transformationen, v01. V, Abh. I, pp. 1—8
(=Gb‘tt. Nathr. (1874), pp. 529—542); (f) Allgcmcine Untersuchun-
gen fiber Difl'erentialgleichungen, die eine kontinuierliche endliche
Gruppe gestatten, vol. VI, Abh. III, pp. 139—223 (=Math. Ann., 25
(1885), pp. 71451); (g) Beitréige zur ailgemeincn Transformationens-
theorie, vol. VI, Abh. V, pp. 230436 (=Lez'pzign B”. (1888), pp,
14—21); (h) Theorie der Transformationsgruppen III, vol. V, Abh.
IV, pp. 78—133 (=Archiuf. Malh., vol. III, (1887), pp. 93—165).
4. S. LIE and F. ENGEL, Themie der Transfummlionsgruppm, 3 vol., Leipzig
(Teubner), 1888—1893.
5. J. J. SYLVESTER, Collated Mathematical Paperx, 4 vols., Cambridge, 1904—
1911.
6. A. CAYLEY, Callntm' Mathematical Papers, 1?} vols., Cambridge, 1889~
1898.
430
BIBLIOGRAPHY
431
BIBLIOGRAPHY
pp. 271—309; I], ibid., 24 (1926), pp. 328—376; III, ibid., 24 (1926),
pp. 377—395 (= Wake, vol. II, pp. 543—647).
22. O. SCHREIER: (a) Abstmkte konlinuierliche Gruppcn, Abh. math. Sm.
Hamburg, 4 (1926), pp. 15—32; ([1) Die Verwandschafi: stetiger Gruppen
in grossen, MM, 5 (1927), pp. 233—244.
23. J. VON NEUMANN, Zur Theorie der Darstellung kontinuierlicher Gruppen,
Sitzungsber. Berlin, 1927, pp. 76—90 (=Callmed Works, vol. I, pp. 134—148).
24. P. HALL, A contribution to the theory of groups of prime power order,
Proc. Landon Math. Son, (3) 4 (1932), pp. 29—95. ‘
25. W. MAGNUS: (a) Beziehungen zwischen Gruppen und Idealen in einem
speziellen Ring, Math. Ann, 111 (1935), pp. 259—280; (1;) Uber Bezie-
hungen zwischen hfiheren Kommutatoren, J. Crelle, 177 (1937), pp.
105—1 15.
26. j. H. C. Wmuznmz (It) On the decomposition of an infinitesimal group,
Pm. Camb. Phil. Sea, 32 (1936), pp. 229—237 (=Mathematiml Works, I,
pp. 281—289); (b) Certain equations in the algebra of a semi-simple
infinitesimal group, Quart. Jaum. nfMaUL, (2) 3 (1937), pp. 220—237
(=Mathematz'ml War/cs, I, pp. 291—308).
27. I. ADO: (a) Note on the representation of finite continuous groups by
means of linear substitutions (in Russian), Bull. Phys. Math. Soc. Kazan,
7 (1935), pp. 3-43; (b) The representation of Lie algebras by matrices
(in Russian), Uspeln’ Mat. Nauk, 2 (1947), pp. 159—173 (English trans-
lation: Amer. Math. Sat. TramL, (1) 9, pp. 308-327).
28. N. JAoonsouz (a) Rational methods in the theory of Lie algebras, Ann. 13/
Math, 36 (1935), pp. 875—881; (b) Classes of restricted Lie algebras of
characteristic [1, II, Dulcz Math. Journal, 10 (1943), pp. 107-121.
29. G. Bum-mm: (a) Continuous groups and linear spaces; Rec. Math.
Moment, 1 (1936), pp. 635—642; (b) Representability of Lie algebras and
Lie groups by matrica, Arm. ty'Math., 38 (1937), pp. 526—532.
30. E. WITT, Treue Dmtellung Lieschen Ringe, J. Crelle, 177 (1937), pp.
152—160.
31. R. BRAUER, Eine Bcdingung fur vollstindige Reduzibiljtfit von Darstellun—
gen gewc'ihnlicher und infinitesimaler Gruppen, Math. Zeilschr., 41
(1936), pp. 330—339.
32. H. CASIMIR—B. L. VAN DER WAERDEN, Algebraischer Beweis der voll-
stindigen Reduzibilitit der Darstellungen halbeinfacher Liescher
Gruppen, Maih. Ann., 111 (1935), pp. 1—12.
33. A. WEIL, Sur les fonctions elliptiquesp—adiques, C. R. Ami. Sci, 203 (1935),
p. 22.
34-. E. LU'rz, Sur l’équation y‘ = x3 — Ax — B dans les corps [z-adiques, J.
Crelle, 177 (1937), pp. 23744-7.
35. C. CHABAUTY, Sur les points rationels des courbes algébriques de genre
supérieur a l’unité, C. R. Ami. Sal, 212 (1941), pp. 882—884.
432
BIBLIOGRAPHY
433
INDEX OF NOTATION
435
INDEX OF NOTATION
E, 13*: 11.1.1.
P(E), 1r, 1‘, 6*: 11.1.6.
5(9), c,, n: 11.1.5.
fig: U(P(E)) —> E: 11.1.6.
M(X), [(w), Lib(X) = LibK(X): 11.2.1.
L(X) = LK(X): 11.2.2.
9: X )) L(X): 11.2.2.
(2, r):11.2.3.
L(u): 11.2.5.
Lib°(X), L5(X), L"(X): 11.2.6.
P": 11.2.7.
W19): 11.2.7.
H, dv: 11.2.10.
E = ‘1’(w): 11.2.11.
A(X) = AK(X), A*(X), Mo(X): 11.3.
n: 11.3.2.
(Ga), (GD: 11.4.1.
II: I .42.
gr(G), gra(G): 11.4.3.
F(X), A(X), A"(X): 11.5.
A(X), to: 11.5.1.
:(a): 11.5.2.
1(x), exp(x), 103(31): 11.6.1.
:(X), [(X): 11.6.1.
£(X): 11.6.2.
a H b: 116.2.
H, H," H": 11.6.4.
H, Q: 11.7.2.
A, ep, logA, 15(AI, A): 11.7.3.
0,6: I, _ 1’ 11.8.
8(a): 11.8.1.
h(;r,y): 11.8.3.
Gflz'IIflA.
{L(u): 11.App. Y
‘1 9G, fig), 3(g), Int(g),f: 111.Conventions.
GL(E), GL(n, K): 111.].l, 111.3.10.
GV: 111.1.2.
1(g), p(x): 11115.
(G, g, 0, m):111.1.10.
T(m): 111.2.1.
T(G), T(q:): 111.2.2.
Ht':111.3.1, 111.3.18.
436
INDEX OF NOTATION
437
INDEX OF TERMINOLOGY
The reference numbers indicate respectively the chapter, paragraph and num-
ber (or, occasionally, exercise)
Adjoint group of a real or complex Lie group: III.6.4-.
linear mapping of an element of a Lie algebra: 1.1.2.
representation of a Lie algebra: 1.3.1.
representation of a Lie group: 111.3.12.
Ado’s Theorem: 1.7.3.
Algebra (not necessarily associative): 1.1.1.
derived from an algebra by extending the scalars: 1.1.1.
enveloping, of a Lie algebra: 1.2.1.
Lie: 1.1.2.
opposite: 1.1.1.
product. 1.1.1.
quotient. 1.1 1
mtricted enveloping, of a Lie p—algebra: 1.2, Exercise 6.
symmetric, of a module. 1.2.5.
Almost simple Lie group: 111.9.8.
Alternants of degree 7:: 11.2.6.
Antiautomorphism, principal, of the enveloping algebra of a Lie algebra: 1.2.4.
Associated (bilinear form) with a g-rnodule (a representation): 1.3.6.
(law of infinitesimal operation) with a law of operation: 111.3.7.
Associative algebra, free: 1113.
Autonmr'phism, special, of a Lie algebra: 1.6.8.
Basic commutators: 11.5.4.
family ofa Lie algebra: 11.2.3.
Bieberbach’s Theorem: 111.4, Exercise 13.
Bigebra: 11.1.2.
Biinvariant section: 111.3.13.
Binomial polynomial: 11.5, Exercise 4.
Bracket: 1.1.2.
439
INDEX OF TERMINOLOGY
440
mDEX 0F TERMINOIDGY
4-41
INDEX OF TERMINOLOGY
442
INDEX OF TERmNOLDGY
443
INDEX OF TERMINOLOGY
444
INDEX OF TERMINOLOGY
4-45
INDEX OF TERMINOLOGY
44-6
INDEX OF TERMINOLOGY
447
SUMMARY
Qfeerlain properties qffinite-dimemional Lie algebra: am afield qfeharaeteristir 0.
Let g be a Lie algebra, tits radical, 11 its largest nilpotent idea], 5 its nilpotent
radical and l the orthogonal of 9 relative to the Killing form. Then it, n, b, I are
characteristic ideals and t D I D n D a.
44-9
SUMIMARY
450
NICOLAS BOURBAKI
Elements of Mathematics
Graded algebras
Tensor products of algebras
Tensor algebra. Tensors
Symmetric algebras
. Exterior algebras
. Determinants
Norms and traces
§ 10. Derivations
§ 11. Cogebras, products of multilinear forms, inner products and duality
Appendix. Alternative algebras. Octonions
Exercises
IlADE AND PRINTED' IN GREAT BRITAIN BY
WILLIAII CLOWES & SONS, LIMITED
LONDON, 330cm AND cowl-mm
ISBN 3-540-50218-1
ISBN 0—387-50218-1