THEOREM 3.2. i.

If x = x(t), y = y(t) is a solution of the equations ~___a_~,OF ~___~_~OF for

IYl = i, F = 0, then the line ~(x(t), y(t)) is tangent to the surface f(x) = 0, and the point
of tangency $(t) of the line ~(x(t), y(t)) with the surface {f = 0} moves along a geodesic
of this surface.
4. Methods of Constructing Functions in Involution
4.1. The Method of Shift of Argument. Let f(x) be a function defined on a linear space
V, and let a C V be a fixed vector. On the space V we construct a family of functions f ~ ( x ) =
f(x~-%a), where % is an arbitrary number ( %~R if V is a linear space over R ) . We say that
the functions f%(x) are obtained from f(x) by t h e operation of shift of argument. If f% can
c o
' E n be expanded in % in a series [for example f(x) a polynomial] then f ~ ( x ) ~ - % f~,n(X) and the
n=0
shift operation generates an entire family of functions {fz,~(x)} from the function f(x).
In the theory of Hamiltonian systems shift of argument is used because of the following
theorem of Mishchenko and Fomenko [54]. This idea for the case of the Lie algebra so(n) ap-
peared in the work of M a n a k o v [53].
THEOREM 4.1.1 (see [65]). Let F, H be two functions on the space G* dual to the Lie
algebra G which are constant on orbits of t h e coadjoint representation (are invariants of the
coadjoint representation of the Lie group ~ corresponding to the Lie algebra G), let a ~ G *
be a fixed covector, and let ~, ~ER be arbitrary fixed numbers. Then F~(t)=F(t-~-~a) and H ~ ( t ) =
H ( L - ~ a ) are in involution on all orbits relative to the standard symplectic Kirillov struc-
:ture.
This theorem provides a complete involutive collection of functions for a rather broad
class of Lie algebras including semisimple Lie algebras (see [63, 65]).
A general construction of shift of basis functions in finite-dimensional representations
in the space of smooth functions on G* was proposed in the work of Trofimov [90]; see also
the survey [103]. We obtain Theorem 4.1.1 and also the construction of shifts of semiinvari-
ants of the coadjoint representation of the work [4] as a special case of this general con-
struction.
In the work [87] the operation of shift of invariants is generalized to the case of
arbitrary finite-dimensional algebras (not necessarily commutative or skew-commutative). A
construction of equations of hydrodynamic type for which these shifts are first integrals
is also given there.
In the construction of complete involutive families of functions it is also necessary
to use nonlinear shifts of argument. We present here one useful assertion of this sort con-
cerning nonlinear shifts.
T H E O R E M 4.1..2 (see [82, 163]). Let G be a Lie algebra equipped with a nondegenerate,
G-invariant bilinear form ($, N). Suppose functions f and h satisfy the condition [gradf(~),
~]--0, [grad h(~), ~ ] = 0 for all ~ G . We introduce the notation f~(~,~) = ~ ( ~ - ~ - ~ 2 e ) , h~(~,B) =
h(~-~N~-~2e), where ~ is a fixed element of the Lie algebra G, and % and ~ are arbitrary para-
meters. Then the functions [~(~,~) and h~(~,N) are in involution for any % and ~ relative to
the canonical Poisson bracket on orbits of the coadjoint representation of the Lie algebra
(O) = C | (~[x]/(x~i).
4.2. The Method of Chains of Subalgebras. Suppose in the Lie algebra G there is a
filtration of subalgebras ~ G ~ Q ~ . . . ~ O q . There then arises a chain of mappings O * - + O ~ - +
O~-+...-+Qq. Each of the mappings Oi-+O~+~ is the restriction to the subalgebra O~+~ of
linear functions defined on G i. The following assertion holds.
THEOREM 4.2.1 (see [88, 91]). Suppose there is a chain of s u b a l g e b r a s V m S. If func-
tions f, g on S* are in involution on all orbits of the representation A d ~ , where ~ is the
Lie group corresponding to the Lie algebra S, then f and g are in involution on all orbits
of the representation Ad~5, where f15 is the Lie group corresponding to the Lie algebra V and
f = f o z ~ , -g-----goz~, ~ ' V * - + S * is the restriction mapping.
One can become acquainted in more detail with various versions of the method of chains
of subalgebras, for example, in the survey [103].
We note that the result of Arkhangel'skii [4] can be strengthened by using Theorem 4.2.1.
2713
THEOREM 4.2.2. Let F, H be two semiinvariants ( F ( A d g * x ) = x ( g ) F ( x ) , x6G*, g6~J, where ~ is
a character of the Lie group ~ ) of the coadjoint representation of the Lie group ~, corre-
sponding to the Lie algebra G, let =E~* be a fixed covector, and let ~, ~ R . Then {f(tq-%a),
H ( t q - ~ a ) } = O on all orbits of the coadjoint representation relative to the standard symplectic
structure.
The assertion of Theorem 4.2.2 follows from Theorems 4.1.1 and 4.2.1, since any semi-
invariant is the continuation of an invariant of some smaller subalgebra.~
4.3. The Method of Tensor Extensions of Lie Algebras. This method was first proposed
by Trofimov [92, 93] and was then developed by Brailov [i0] and Le Ngok T'euen.
Let G be a Lie algebra, and let K be a commutative ring. Then O | K is a Lie algebra.
In the case where K-----R[xI,.. x n ] / ( x ~ ' + I " . , X m n + 1 ) . . . . n , we denote the Lie algebra G | by f ~ m ...... m n
(G). Let E i denote the image of x i in the ring K.
= ' ~e j , O~<j<r,
Let el,...,e r be a basis of the Lie algebra G. Then the vectors el ...
0,.<at~<~! , l < i ~ < n , form a basis of the Lie algebra ~ m ...... m n ( U ) . We denote the dual basis in
G* by el; it is defined from the relation < e l , e ] > =6~, where <f, x> is the value of f C O * o n
the vector x E U . We denote the coordinates in the spacelQm~,i. ,~n(G)* in the basis dual to the
basis ~ ' . . . e n a n e i, l - . < j - ~ < r , O ~ a t . . < r ~ i l . . < i . . < I z , by x(~1 .... . o~n) i, while the coordinates in G* in the
basis dual to el,-l~{-<.-Gwe denote by xi (IKi-.<r). In the space f ~ , n . . . . . . . n ( O ) * we introduce the
new variables
. . . . . ( 1 )
o~j+~j=mj
. l < j < t z
Description of algorithm (~) of V. V. Trofimov. Let F(xl, .... Xr) be an analytic function
on the space G*; we consider the function F(zl, .... z r) on the space f~m ...... mn(G)* with values
in the ring K. By expanding F(x I ..... x r) in a Taylor series and substituting z i from (i),
we obtain a finite sum, since sufficiently high powers of Ei are equal to z6ro. Thus, F
(z I .... ,z r) is a well defined function on the space ~m,,..~,mn(O)* with values in the ring K;
it can be represented in the form
F ( Z l , . . . . Zr ) ~ ~ 81~I... 8;nF a . . . . ~n (X ([~1 . . . . . ~,~t)t,
O~ctj<m]
s i n c e a ~l . . . a n % i s a b a s i s o f t h e a l g e b r a K~- I ~[ xl . . . . . Xn l / ( x [ ~'+' . . . . . xmn+r), F~ . . . . . . %(X(~31 . . . . .
~n)i )ER. The function F~, . . . %: Q, n .... mn( G) *. - +kwe call the homogeneous components of the function
F(zl, .... Zr). A l g o r i t h m (9~), by definition, takes a function F on the space G* into the col-
lection of homogeneous components ~(F)-----{f=,..=n} of the function F.
THEOREM 4.3.1 (see [93, 96, 97]). Suppose the functions F~(x) ..... FN(X) defined on G*
are in involution relative to the Kirillov form on all orbits of the coadjoint representation
of the Lie group ~ associated with the Lie algebra G. Then all functions 91(F~)U... U~I(FN)
coming out of algorithm (~D are in involution relative' to the Kirillov form on all orbits of
the coadjoint representation of the Lie group ~2rnt,..mn(~), associated with the Lie algebra
f ~ m .... m n ( O ) . Moreover, if F~, .... F N are functionally independent on G*, then all functions of
the family ~ ( F O U . . . U ~ ( F ~ ) a r e functionally independent on the space f ~ m .... a n ( O ) * .
Remark i. If F is an analytic function on an open subset U c G*, then the homogeneous
components F ~ . . . ~ n of the function F are also defined on some open subset L f ~ f ~ m .... m n ( O ) * .
The assertion of Theoem 4.3.1 hereby remains valid. This situation arises, for example, in
considering rational invariants of the coadjoint representation.
Remark 2. Regarding the connection of the method of tensor extensions with the general-
ized shift method of the work [90], see [98].
A partial generalization of algorithm (~) to the case of algebras with Poincar4 duality
was obtained in the work [i0]. Let A be a graded commutative a i g e b r a A = A o @ . . . @ A n , A i A i
A~+ i, i +j ~
~I~.
Definition 4.3.1. Let d i m A n = i, a n d let ~ be a linear . . . . . . . . function on A identically equal
to zero on A i for i-.</z--I and not equal to zero on A n . Let ~ ( g , b ) = o ( a b ) be a syn~metric bi-
linear form on A. If S is a nondegenerate form, then the algebra A is called an algebra w i t h
Poincar6 duality.
2714
It is easy to see that for such algebras A 0 = k. Let sl = i. We chose an arbitrary
basis ie2 ..... e7, in A I and then choose an arbitrary basis e7,+i ..... 87, in A 2 and so forth to
gradation n/2. In An/2 we choose a basis in which the matrix of the form $ is diagonal. In
the spaces Ai, i > n/2 we chose bases dual relative to ~ to those already chosen in An/2_ i .
As a result, w e obtain Sl, .... SN - a homogeneous basis of the algebra A which is self-adjoint
relative to 6 : A , ~ ( e , i , ej)----6~j, where * is permutation.
Let el,...,em be a basis in G*, and let xl,...,x m be the corresponding coordinates. The
linear functions xl, .... x m on G* are naturally considered elements of the Lie algebra G. Let
( T A - ~ - O | Then the elements x i | of the Lie algebra G A can be considered linear coordi-
nate functions on GI~ Let x { - ~ x ~ | be coordinates on G~. For a polynomial function P(x) on
G* given by
o o
k=0 1 , . . . , , l k
and I ~ ] ~ < N we define a polynomial function P(J)(x) on G ~ :
N
9 9 , . 8. l X ? j ' "
i , . . . . , t k
J ' . . . . . J k
here (ej, sj,.....e&) is the projection of the element ej~....'ei~ onto the basis vector ~j. Let
, and let A n n ( x ) = { g C O I a d ~ x - ~ - O } be the annihilator of the element x. The dimension of
the annihilator of an element xGfJ* of general position is called the index of the Lie algebra
G. Let ~ be a collection of polynomial functions on G*. The collection of polynomial func-
tions on G~ of the form p(i), PC~, l ~ < j ~ N , we denote by ~ A .
THEOREM 4.3.2 (see [i0]). If ~ is an involutive collection of functions on G*, then ~ A
is an involutive collection of functions on G~. If ~ is a complete invoiutive collection
and the number of independent polynomial invariants of the Lie algebra G is equal to the in-
dex, then ~ A is a complete involutive collection on G~. If the collection ~ contains a non-
degenerate quadratic form, then the collection,~ ~ also contains a nondegenerate quadratic
form.
Le Ngok T ' e u e n proved that the assertion of Theorem 4.3.2 holds only for algebras with "
P o i n c a r g duality. More precisely, let G be a Lie algebra, let A be a commutative algebra,
let ( J | be the tensor pr0duct of G and A, let ~i be a basis of the algebra A , and let
be a linear operator, ~ : e i - - + e o : t , such that ~2 = id. For f ~ O we set f ' ~ = f | 1 7 4 ~. If f
and g are polynomials on (O| then we define the function (fg)a by the formula ( f g ) a ~ _ ~
THEOREM 4.3.3. The equality { f ~ , g b } ={ f , g}ab holds if and only if the algebra k is an
algebra with Poincar~ duality.
4.4. The Method of Contraction of Lie Algebras. Let G be a Lie algbera. We suppose
that G = H + V w h e r e b y I/f, I-/]cH, [/-/,V]cV, IV, V]cff. Then G is called a Zg-graded algebra.
Let H*(V*) be the subspace in G* of all covectors annihilating V (respectively, H). Then
G*-----H*| *. For any x E O * and g C O everywhere below x H and x V denote the H*- and V*-compo-
nents of the element x, while gH, gv are the H- and V - c o m p o n e n t s of g. On G we define a new
commutator I x , b ' ] ~ . by the formula [ g , gq~.=[g~, g , v ] - - ] - [ g v , g v ] - ] - [ g ' v , g v ] - [ - - ~ , [ g v , gv]. The commutator [x, Y]X
satisfies the Jacobi identity. The Lie algebra obtained we denote by G k.
D e f i n i t i o n 4.4.1. Let G be a Z2-graded Lie algebra, and let G% be the series of Lie
algebras constructed above corresponding to the commutators [x, Y]X" The Lie algebra G O is
called a contraction of the Z2-graded Lie algebra G.
Let G be a Z2-graded Lie algebra. We denote the Poisson bracket for the contraction G o
by {f, g}0- In a manner similar to the way we define the contraction of the commutator [x,
y] above, for certain functions F on G* it is also possible to define something like a con-
traction.
Definition 4 . 4 . 2 . We suppose that the e x p a n s i o n of F in powers of V breaks off at some
term F o ( x F . ) - J r F 1 ( x ~ , X v ) ~ : - . . . 2 r F ~ ( x ~ , xK), w h e r e F ~ ( x H , x v ) for fixed x H is a k-form in x V. In
this case we set F ~ ( x ) = k T F x ~ , k 2 X v 9 We remark that F~(x) is also defined for X = 0,
2715
F0(x) = Fn(XH, xv). As A. V. Brailov noted, the operation of c o n t r a c t i o n of Lie algebras
and functions has the following properties.
THEOREM 4.4.1. Let G = H ~ V be a Z2-graded Lie algebra, and let F, F' be functions on G*
such that a) {F, F'} = 0, b) F% and Fi are defined (for example, F and F' are polynomials).
It is then asserted that {F~, F~} = 0, and, in particular, {F n, Fim}0 = 0, where F n, F 'm is
the maximal component in the expansion in homogeneous components in V.
4.5. The Method of Similar Functions. Let H be a semisimple Lie subaigebra of G. We
consider the adjoint action of the Lie algebra H in G. The Lie algebra G becomes an H-module.
Let V be an H-module in G not containing H. If in V we fix a basis Xl, .... x n, then on H there
is defined an n n matrix of I-forms. The value of this matrix on a vector h ~ H we denote
by W h. Thus, W h is the matrix of the transformation adh:V ~ V in the basis xl,...,x n. The
matrix elements mij of the matrix W are I-forms on H. Since the algebra H is semisimple, mij
may be considered to belong to H.
Let B be an invariant r-form relative to the action of the algebra H on rows (XI,. .,Xn) ,
X z ~ V . Such forms exist, since H is semisimple. Then B ( X W I~, .... X W It) is an H-invariant,
i.e., [ h , B ( X W i ~ I ...i )#Wtr)]=0, if we c o n s i d e r B ( X ~ v 6, .... x W t i ) as an element of the tensor
algebra S(G) over G. If we consider vectors of G as linear forms on G*, then this can be
reformulated in the form of the following assertion.
Proposition 4.5.1. Functions on G* of the form E ( X W 4, .... X W It) are in involution with
all the linear functions ~ii (I~<i, j~<n).
It may turn out that the H-module G has several submodules VI,...,Vm isomorphic to V.
In this case it makes sense to speak of functions of the form B ( X W 4 , X ~ W I ' , . . . ~ X ( m ) W t O. As
before, functions of this form are in involution with all the linear functions ~ii, I~i, j~<n.
Definition 4.5.1. Functions of the f o r m B ( X W 4, .... x ( m ) w tr) we call canonical H-invari-
ants.
n - - 1
According to the C a l e y - H a m i l t o n theorem W n = ~ a i ( W ) W i.
Definition 4.5.2. Powers of the matrix W are explicitly present in the notation for a
canonical H-invariant. We call two canonical H-invariants similar if one can be obtained
from the other by replacement of the matrix W n by W i whereby the power i is such that the
function ai(W) is not identically zero.
Definition 4.5.3. Let G be a Lie algebra, and let H be a subalgebra of it. We say that
the pair (G, H) satisfy the M-condition if the following conditions are satisfied. We sup-
pose that G = H ~ V I @ .. ~ V h is the decomposition of the H-module G into irreducible submodules.
It is required that a) H be a simple Lie subalgebra in G of classical type A n , Bn, C n, Dn;
b) either all V i are trivial H-modules or the representation of the Lie a~%ebra corresponding
to the submodule V i (i = l,...,k) is equivalent to the minimal representat~an.
There is the following basic theorem due to A. V. Belyaev.
THEOREM 4.5.1. Let G be a Lie algebra, and let H be a subalgebra Of it. We suppose
that the pair (G, H) satisfies the M-condition and F i = x'WiBX t is a canonical H-invariant.
Then the Poisson bracket {Fi, Fn} of similar functions Fi, F n (X may coincide with X') is
equal to zero.
4.6. The Restriction Theorem. Let (M, m) be a symplectic manifold, let h be a smooth
function on M, and let s g r a d h be the vector field dual relative to the form m to the differ-
ential dh. Let E be a group acting by symplectic diffeomorphisms on M, and suppose the func-
tion h is constant on orbits of the action of the group E. We then say that E is a group of
symmetries of the Hamiltonian system (M, m, h).
There is the following assertion due to A. V. Brailov which is useful in the construc-
tion of functions in involution.
THEOREM 4.6.1. Suppose Z is a compact group of symmetries of the Hamiltonian system
(M, m, h); ~ is a E-invariant algebra of integrals, ~z is a E-fixed subalgebra, N is the mani-
fold of fixed points, and ~ is the set of restrictions of integrals in ~ to N. It is then
asserted that a) (N, ~) is a symplectic manifold where ~ = m/N; b) the set ~ is closed
2716
relative to the Poisson bracket {f, g}N, and the restriction mapping is an epimorphism of Lie
algebras ( ~ z , { f , g } N ) - + ~ , { f , g } ~ ) ; c) ~ is an algebra of integrals of the Hamiltonian system
(N, ~, h), where h is the restriction of the function h to N.
Symplectic actions of groups arise, for example, in the following situation (A. V. Brai-
l o v ) .
Proposition 4.6.1. Let Z be a compact group acting by automorphisms on the Lie algebra
G, and let G n be the fixed subalgebra; let x E G n * c G * , let Oa(x) be the orbit of the represen-
tation Ad* of the Lie group corresponding to G passing through x, and let Gan(x ) be the orbit
for G n, Then a) Z acts by symplectic diffeomorphisms on Go(x); b) G o n ( x ) is open in the mani-
fold of E-fixed points of the orbit GQ(X); c) if ~, mn are the Kirillov forms of the Lie
algebras G and Gn, then ~ n = ~ i O n ~.
CHAPTER 3
COMPLETE LIOUVILLE INTEGRABILITY OF SOME HAMILTONIAN SYSTEMS ON LIE ALGEBRAS
i. Formulation of the Problem of Constructing a Complete Involutive
Collection of Functions
We shall distinguish here the three most important formulations of the problem indicated.
i) The problem of constructing a complete involutive collection of functions on Lie
algebras. Let G be a Lie algebra, let G* be the space dual to G, and let K be some class
of functions on G* which is closed relative to the Poisson bracket, i.e., K is a Lie sub-
algebra of C~(G *) (for example, for K it is possible to take all smooth functions on G*,
all analytic functions on G*, the space of all polynomials, rational functions, etc.).
It is required to construct a collection of functions fl,...,fs of the class K such
that a) { f i , fT }-----O i 1 - ~ < i , j-.<s (i.e., the subalgebra generated by fl ..... fs is Abelian);
b) s = i / 2 ( d i m G + i n d G ) ; c) fl, .... fs are functionally independent almost everywhere on G*.
Such collections of functions are important from various points of view.
a) We obtain a large reserve of regular examples of completely integrable Hamiltonian
systems. More precisely, each Hamiltonian vector field v i = sgradfi, i = l,...,s
defines a completely integrable Hamiltonian system on each orbit of general position
of the coadjoint representation.
b) Collections (fl ..... fs) define (locally) canonical coordinates on each orbit of gen-
eral position of the coadjoint representation.
c) We obtain a reduction of noncommutative integrability to commutative integrability.
d) We obtain a large reserve of algebraicizable Hamiltonian systems on the space O*_~_R N
(see Sec. 2, Chap. 2).
2) The problem of constructing a complete involutive family of functions on a single
orbit chosen individually of the coadjoint representation (in particular on singular orbits).
It is required to construct a collection of functions fl,...,fs on G* such that the restric-
tions fl]~ ..... f~l~ of the functions fl ..... fs to the given orbit ~ form a complete involu-
tire family of functions on G.
The importance of such collections is that we obtain regular examples of completely
integrable Hamiltonain systems on symplectic manifolds.
Problems i and 2 can be strengthened as follows: include in a complete involutive family
of functions a given Hamiltonian H which is interesting, for example, from a physical point
of view.
3) The problem of constructing a complete involutive family of functions on a given
symplectic manifold. Let (M 2n, m) be a symplectic manifold. It is required to construct a
collection of functions which almost everywhere are functionally independent and are pairwise
in involution.
if the manifold M 2n is homogeneous, then we return to problems i or 2, since any homo-
geneous symplectic manifold is an orbit of the coadjoint representation.
2717