Group theoretical methods in classical mechanics and
continuum physics
Citation for published version (APA):
Mooren, L. C. J. M. (1979). Group theoretical methods in classical mechanics and continuum physics. Technische Hogeschool Eindhoven. https://doi.org/10.6100/IR63280
DOI:
10.6100/IR63280
Document status and date: Published: 01/01/1979
Document Version:
Publisher’s PDF, also known as Version of Record (includes final page, issue and volume numbers)
Please check the document version of this publication:
• A submitted manuscript is the version of the article upon submission and before peer-review. There can be important differences between the submitted version and the official published version of record. People interested in the research are advised to contact the author for the final version of the publication, or visit the DOI to the publisher's website.
• The final author version and the galley proof are versions of the publication after peer review.
• The final published version features the final layout of the paper including the volume, issue and page numbers.
Link to publication
General rights
Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights. • Users may download and print one copy of any publication from the public portal for the purpose of private study or research. • You may not further distribute the material or use it for any profit-making activity or commercial gain
• You may freely distribute the URL identifying the publication in the public portal.
If the publication is distributed under the terms of Article 25fa of the Dutch Copyright Act, indicated by the “Taverne” license above, please follow below link for the End User Agreement:
www.tue.nl/taverne
Take down policy
If you believe that this document breaches copyright please contact us at:
openaccess@tue.nl
GROUP THEORETICAL METHOOS IN
CLASSICAL MECHANICS AND CONTINUUM PHYSICS
GROUP THEORETICALMETHODS IN
CLASSICAL MECHANICS AND CONTINUUM PHYSICS
...
"TrIl: - Rek", c' n
ruming.
16-2-'~
,.-'
oren.
I
ec.
'::60
'4
dOss./
PROEFSCHRIFT
TER VERKRIJGING VAN DE GRAAD VAN DOCTOR IN DE TECHNISCHE WETENSCHAPPEN AAN DE TECHNISCHE HOGESCHOOL EINDHOVEN, OP GEZAG VAN DE RECTOR MAGNIFICUS, PROF. DR. P. VAN DER LEEDEN, VOOR EEN COMMISSIE AANGEWEZEN DOOR HET COLLEGE VAN DEKANEN IN HET OPENBAAR TE VERDEDIGEN OP
VRIJDAG 2 FEBRUARI 1979 TE 16.00 UUR
DOOR
LEONARDUS CAROLUS JOHANNES MARIA MOOR EN
GEBOREN TE 's-HERTOGENBOSCH
Dit proefschrift is goedgekeurd door de promotoren:
prof.dr. L.J.F. Broer dr. B.J. Verhaar
Contente
Abstract
o
Part I
I. Discrete systems
1.1 Canonical transformations 1
1.2 Poissonbracket and invariant canonical transformations 4
1.3 The canonical group 5
1.4 Relation between finite and infinitesimal canonical
transformations 9
1.5 Structure of the algebra of tbe canonical group 14
1.6 Constants of the motion 27
2. Continuous systems 32
2.1 Canonical transformations 32
2.2 Invariant canon ie al transformations 36
2.3 Special transformations. Campbell-Baker-Hausdorff formulas
2.4 Relation between finite and infinitesimal canonical trans format i ons
2.5 Examples of canonical transformations and canonical transformations as realizations of finite-parameter group
Part 11
3. Classification of quadratic constants of tbe motion for
37
41
44
linear systems 60
3.1 The equation qtt - -Vq• General Theory 60
3.2 Discrete examples 75
3.3 Examples of continuous systems, scalar fields 81
3.4 Pseudo-rotations, pseudo-translations 86
3.5 Constants of the motion with a density explicitly
depending ont t 10~
Appendix A Appendix B Appendix C Appendix D Appendix E Appendix F References Samenvatting Nawoord Levensloop 114 120 122 125 129 131 137 139 140 141
Abstract
This work consists of two parts.
In part I canonical transformations are treated for discrete- (finite number of degrees of freedom) as well as continuous systems. Relations between finite and infinitesimal canonical transformations are
explored. For discrete systems the algebra of the canonical group is constructed out of the algebra of several subgroups. A whole family of Fourier- and Laplace-like transformations can be constructed which are realizations of some classical finite parameter groups. The proce-dure followed in this thesis is somewhat different from,the "usual" one: first the realization of the algebra ofsome group is found, then from this a realization of that group is constructed.
Part I I of this work deals with the classification of quadratic
constants of the motion for linear systems. The relation between commu tators of the dynamical operator and constants of the motion is inves-tigated.
It turns out that a class of equations which does not possess rotation-symmetry nevertheless has a rotation-symmetryoperator with rotation-like pro-perties. We call this a pseudo-rotation operator. Together with this pseudo-rotation operator there can be found pseudo-translation opera-tors forming a Lie-algebra which is isomorphic with the conventional algebra consisting of rotation and translation operators (ISO(2) and ISO(3}l.
Eigenfunctions of the pseudo-rotation operator are constructed in Fourierlanguage. Unfortunately all attempts to find the eigenfunctions
PART I
I. Discrete systems
1.1 Canonical transformations
In chapter I we shall treat Hamiltonian systems. The Hamiltonian will be a function of coördinate- and momentum variables (q(t) aod
pet»~ and of t. The functions q(t), pet) are, in general, vector-functions, although we shall not write the indices in most cases. The Hamiltonian H (p,q,t) is a scalar function.
The Hamilton equations are equivalent withthe equationsof motion
of the system, they are
( 1,0
qt means :t q(t). If no confusion can arise we shall not write down the explicit t-dependence of the functions. The functions q and p are cal led a pair of canonically conjugated variables. Let there be a
transformation from q,p to the functions q,p;q = q(q,p,t),
P
=
p(q,p,t)We call such a trans format ion a canonical one if there exists a
function H(q,p,t) such that the equations we get from
H
0,2)
transformation q,p + q,p in (l,l) we get (1.2}. !here are two ways to construct a canonical transformation.
~) The first method is to use a generating function, a function of
one old and one new variabie (or in the n dimensional case: n old and n new variables). !here are four possibilities
We get the transformation in the usual way.
CJFl CJFl CJFl Fl P = oq , p
... - ar'
H • H+ ät
CJf2 oF:z oF:z Fz p.~ . q • q .. dp'H .. H +ät
oFa oFa oFa Fa q- Ft> •
p = -mr'
H"'H+m:
aF .. oF ... Clf .. F .. q ..- ä'P '
q"~ ~H ..
H + "li"i:A canonical transformation of a special kind is the one wbich
transforms the Hamilton equations into.
tlH
0 - = qtdP
aH
0 - = - P t dq!his is the case if
H
is a function of t alone. Generating thistransformation by Fl(q,q,tl we get:
aFl aFl ClFl
ii
=
H + "li"i: ' P = oq , p= -
~
(J ,3)
(I,4)
(1,5)
or
0.5)
\be Hamilton-Jacobi equation. Let H be a given functiOn, a possible
soluti~ F1 of tha H.J.-equation tben is tbeaction integral
F1
=
J
(p qt - H)dt ([2])o
*) tbe second method of construction is to start with an infinitesimal
canonical transformation (canonicalvariation). The identical
trans-formation is generated by F2 (p,q. t)::pq. A canonical variation can
then be generated by pq + 8g(P.q,t), g is called the generator of the
transformation (the variation as well as the fini te transformation ar~
constructed from tbis). The equation (1.3) gives
q
andp
8 a small parameter, so we may neglect second order terms in 8. That means we may change g(p,q.t) into g(P.q.t). So we get:
every function g(p,q,t} is the generator of a canonical trans forma-tion and tbe first order variaforma-tion in q and p is
A tinite transformation ean now be eonstrueted by performing sueees-sive infinitesimal transformations. We get in this way a whole family of transformations parametrised by a parameter s by integrating the equations
dq(s)
=
~ dp(s) = _ ~ p(O)=
P q(O)ds op(sJ • ds aq(s) • , q (1,7)
g is a funetion of q(s) and p(s}. See seetion 1,3 for more details.
1.2 Poissonbraeket and invariant eanonieal transformations
Given two funetions f(q,p) and g(q,pl. The Poissonbraeket
L.J
between f and g is defined in the following way
L
f,gJ af •2.K
_!! .
2.K
=
oq op ap aq (2.1)ramark: qand pare vectors. so the terms on the right hand side of the equation are inner produets.
A special case
l
q •• p.J
=
0 ~J" Let a eanonieal variation be generated~ J *
oy
g(q,p,tI aeeording to (1,6J. we then defineoH,
the variation ofH.
oH.· H.
(p.q,tl -
H. (p,q,tl (2,2)Up to firstorder in'~this ean ,be written as:
oH
aH
For
*"
we find an e.xpre.ssion, witl!.. q .and p evolved according to0.11.
dg .;. '.2.& + d t a t19,KJ
(2,n.
(2,4[. and (2,51 togat1ier: .~-
-'-
., - - '
E.
dt = H:(p,q,t[ - H(p,q,tl - H(p,q.tl + H(p,q,t) -+E.:~
-
H(p~Ci,tI
- Il(p,Ci,tI
Wa slia.ll caU tlie. trans format ion an invariant canonical trans forma-tion tf
i
is tlie. same. function ofp
andq
as K of p and q, or:ii(p,~;tL;' H(p,q;tL K(p,q;tI: == I:t(P.q,t)
We gat as a final re.sult: TIie. generator of an invariant canonical variation is a constant of t1ie. motien and vice versa.
1.3 The canonical group
Before going on we shall give fust a review of definitions of groups, . algeDras, representations and so on. We shall frequently use these
definitions and properties in the following sections (see for example
[3], [41).
A ~ is a set G of dements with one relation defined betwèen the
elements (relation • and with the properties:
a E G, bEG ~ a • b € G ; Va € G 3a-1 EG; 3 eEG with
-]
a • a = e. a • lol. "" e • a • a Va E G
A subgroup He G is a subset of G with a groupstructure itsel! and
the. same relation as G.
An
algebra A is a set of elements together with two relations (+ andx) between the elements. If a, b, c
€
A then a + b E A and a x b E A,further ax(b+c) • axb + axc.
An
algebra is an associative algebra if (axb)xc = ax(bxc). and aLie algebra ifaxa = Oand (axb)xc + (bxc)xa + (cxa)xb == 0 (the
Jacobi identity).
A subalgebra Be A is a subset of tha algebra Awith an algebra
struc-ture itself, a; strucstruc-ture with the same relations + and x. A
sub-algebra Be A is an invariant subalgebra ifaxb E B, Va E A and Yb € B.
Let A be an algebra and B a subalgebra of A: Two elements in A~ (A
mod B) are identified if their di'fference lies in B. This defines an e-quivalence-relation. The space A'\B is the space of equivslencè-classes. If B is invariant A\B is subalgebra of A. Tbe number çf linear inde-pendent elements of an algebra is cal led the dimension of the algebra. A representation R of a group G is a set of matrices R{a) with the
property R(a).R(b)
=
R(c) if a·b == c(a,b,c €G).
In an analogeous waythe representation of an algebra is defined, the mapping preserves .the
algebrastructure (the relations + and x).
of operators, preserving the algebra strueture. We sha11 not a1ways distingish between a representation and a rea1ization if no eonfusion
iA possib1e (a representation is a rea1ization). If by a trans
forma-tion T all matrices R(a) € R(G) ean be brought in bleek form the
representation Rwi11 be eal1ed reducible, otberwise irreducible.
The funetions g(P.q) (We suppose, for eonvenienee, g not explicitly
~epending on t) eonstitute a Lie algebra with tbe relations + and "poissonbraeket". The algebra bas dimension infinity and we use for
this algebra the symbol Al' We eau ehoose a basis for Al by taking
the homogeneous functions of degree n: qlpk. n
=
k+l, n integer~
O.For convenienee we sball treat the ene' dimensional case, a
generali-zation to more variables (n-tuple q.) is straightforward.
l.
To every function g(p,q) we ean associate a linear operator g(p,q) in the following way
g(P.q)
=
agL
_!&
aaq ap ap aq
g
operating on a function f(p,q} givesi .
f(p,ql •ig,fJ
special cases:
i .
q •-* '
g •
p ..~~
(3,1)
(3,21
remark: We llave restricted ourselves to functions g(p,q) wbich are analytie in a neighbourhood of tbe point (0,0).
This is sufficient for alocal theory. Even if g ean be eontinued analytically it is not sure that g generates global
transformations (transformations onto the whole phase space) see e.g. the example at the end of section 1.4.
The operators
g
constructed through (3,1) form a Lie algebra withrespect to + and the commutator [,J. This algebra Az is isomorphic
with[';;he algebra A111 (for isomorphism see e.g.)3], [4n~
Lg,fJ
=
h . .[g,IJ -
n
A lil means that we must throw away the central element 1 from Al
(the element 1 has Poissonbracket zero with all g € Al, an element
with that property is called central). An additiona1 constant has no
influence on the re1ation.
From the algebra Aa we can construct a group of canonica1
transfor-mations in q and p. First we construct a one-parameter subgroup,
generated by
g.
q(t) = [exp (a
8) •
q] (tlp(tl = [exp (a
g} •
pJ (t) (3,3)This is a canonica1 transformation from q,p to q,p.,The transformations of tnis kind form a commutative (Abe1ian} one-parameter group. tf
T is the transformation (3.3i tnen: T • Ta • qaT 0 • q, T
O is
a a ~ M~
tne identity. !he exponent in formu1a (3.31 is a forma1 power series
exp a
g
00I;' I n - n
L =r
a
gn=O n.
All possibie products of these one-parameter groups (of different g) constitute a group of transformations. the canonical group, an infinite-parameter group.
As already mentioned problems can arise if (3,3)does~ot hold on the
whoie phase-space. if the transformation becomes singular for some
value of a (or if the mapping q.P ~
q,p
is not surjective).In that case the one-parameter group under consideration is not a global Lie group. In a sufficientIy smail neighbourhood of the
identity (a "smaU enough") the transformation holds and we shall
speak of a Iocal Lie group. !his shows thet it is somewhat easier to consider the algebra of the canonical group rather than the group
itself. ComBining (3,3} and (3,4} gives:
1 2
[exp (a
8l •
q] (tl=
q(tl+a h.qJ +TI
a 19,1g,qJJ + ••• (3,S)In the following sections we shail meet exampies of transformations which become singular for certain a-values.
Remark.:
!he generator of' a transformation T .... a is given byg
='~
TI
as will Be clear from the foregoing.a a a=O
1.4 Relation lie.tween· finite and infinitesimal canonical transformations •
The description of canonical transformations described in sections
1.1 and 1.3~st Be equivalent in some sense.
If g(q,p} is given, We get a finite canonical trans format ion through
(3,5). and we can ask the que.stion: what is the generating function F of this transformation?
Another question is the inverse: given a generating funetion, say Fz, what is g(q,p)?
!Wo points must be mentioned:
l} The infinitesimal approach gives transformations whieh are
continuously connected with the identity (Lie group). The trans
forma-tions generated by Ft(q,q} and F4(p,p) are not connected with the
identity, transformations generated by Fz(q,p) and Fa(p,q} have that
property and we shall therefore use Fz and Fa.
2} Fi (i
=
1,2,3,4) is parameter independent. We shall eonnect withF. an one-parameter family Fr (-,',a} in sueh a way that F2(q,p) •
J. J.
F2'(q,p,aa) for a eertain value ao of a and F2(q,p,O) = qp, the
identity transformation (analogeous relations for Fa). Let F2(q,p} he
given, it is then clear thet we ean make the parametrization Fz'(q,p,a) in more than one way (in general} and the problem is not one-to-nne.
If. on the other band, g(q,p) is given we ean find the family Fz'(q,p,a) and the generating funetion Fz(q,p).
Let there be given a funetion g(q,p) whieh generates the transfor-mation
q(q,p,al = exp a g • q
p(q,p,al • exp ag- p (4,1)
Define a function F2(q,p,al (tba prime will be suppressed by now) such that
-
oF
ap(q,p,a) .. oq
- - OF2
q(q,p,a) ,. dP (4,2)
are equivalent with (4,1) (af ter inversion).
We shall now derive an equation for the function Fa. We make use of
the fact that the Poissonbracket between two functions can be calcu-lated with respect to any arbitrary pair of canonically conjugate
variables. For every function f(q,p) the equation
:~
= -
Lg,fJ holds.Especialy for q,p
(4,3)
g(q,p) is invariant under the transformation generated by itself, so
g(q,p) .. g(q,p) (from g(q,p) ,. exp a
g •
g(q,p) .. g(q,p».On the other hand we have:
d / d é) /
a
I
F )dp'I\"'" • ~ F2 + (,,- ;:;=r 2 d
op q,a oa op q,a op q,a a
Combining (4,3) and (4,4) we get
Cl / o F , .
~I
+ Clq/~
1_ ..
og1
'ijf q,a 3a 2 dP q dP q,a dq p dp q,a
In which g and q are regarded as functions of q,p and a,
g(q,p) .. g(q(q,p,a),p). In the same way we get
(4,.4)
and from this:
'2-1
aFz=
~/_aq p,a aa aq p,a
Cembining (4,5) and (4,6) gives the result
a
- I
.
Fz aa q,p(4,6)
(4,7)
The funetion ~(a) is not essential, every g + ~(a) (~ arbitrary) gives
the same trajeetory in phase-spaee, so we ean put ~
=
°
without anyrestrietion. We then arrive at the desired equation for Fz:
(4,S)
Equation (4,71 must oe aeeompanied by the initial eondition for a
°
-qp qp.
Ifa.=O, then
(4,8)readS~~Z
1
a=O g(q,p(q,p,01J=g(q,pl. It is possiblethat for a eertain value of a the transformation generated by
Fz(q,p,a} oeeomes singular. In that case we ean use F3(p,q,a), beeause'
Pz and 1"3 are,not singular simulta'neously «(2], see for instanee the
example at the end of this sectionl. When we take g = -H (The
Hamil-tonian~a.=t then(4,S[ is the Hamilton-Jaeobi-equation. We ean regard (4,Sr as a generalization of the H.J.-equation and so Fz as the
generalization of the action-integral. Tbis is the case because the
equations (4,3) are analogeous to the Hamilton equations, with a
instead oft and g instead of H. A canonical transformation from
begin -(tsO) to end-position (t) is generated by the action integral, a weIl chosen solution of the H.-J.-equation. With this in mind we
find an expression for Fz
a
F2
(q,p)
=J
(q
~
+g(q,p»
da + qpo
(4,9)
Tbe constant term on the right hand sidè of (4,9) is necessary to fullfi11 the initial conditions. Relations similar to (4,8) can be
derived for each Fk,' E.g. the relation for Fl(q,q} reads
= -
(4,10)Tbe boundary conditions must be chosen in a region where Fl is
non-singular, Dut this is not possi'Dle in tne neighbourhood of the
iden-tity. For tOa derivation of (4.31. (4,9) and (4,10) see also [5], [6].
Example: g(q,PL
=
pq2. !lie equation (4.81 now isaFz __ dF2
aa
az
and from tli.is F2
=
f(z-al h(p) + b(p}. Now usa the initialh = kp, b = 0 (k constant)
There is a singularity at a •
~.
for tbis value of a use F3- - -1
F!(p.<l.a)= -pq(l+a.q)· whicb is certainly not singular at a ,. -
.!.
q
Formally (4,8) determines tbe solution Fz. It is. of course,
some-times very bard to solve tbe equation.
1.5 Structure of tbe algebra of tbe canonical group.
The canonical group is thej?;roup cotresPQ.ndinl!lwitb::the,alge~ra with basis
1 k
q p ,land k integer ~ O. We shall form subalgebras and construct
tbe endre infinite-dimensional algebra frOlU finite-iIimensional
suh-algeDras. For tne canonical group of dimension n we use the symbol •
K(Znl. Th.e algelira of a group IC is L(Kl.
We shall treat tne algeBra L(IC(ZII.
!he set qn, n integer ~ 0 is an aoelian suoalgetira of L(IC(ZU
<lqn,Q1mJ ,. 0) generating the following transformations:
óq ,. 0
n-l óp ,. e:: n q
forma-tion of this kind has the property that the fir~t variation in tbe
Lagrangian L is a total time derivative, oL =
~t'l!
(q,q) (provided L.~
6
a
2Lhas the form q q - f(q) and
äqT
~ 0). Such a transformation leavesthe equations of motion invariant and L is "gauge-invarèant"
the transformation. The action Intp;'gral is invariant; 15
J
Ldtunder
'" 0.
h h n . 0 ' h I b O f
In t e same way t e set p , n ~teger ~ ~s t e a ge ra 0 the gauge
group in q, Y (an abelian group toa).
q
Another subgroup is generated by the al~ebra pqn, oq '" Eqn, op '"
n-I (' d" f ( ) ' 1
-E n pq ~n more ~mens~ons Pi q, L '" • • n, f a function
af of q and Qqi '" Ei f(q) , op. 0= -E.p. -,,-).
~ J J "qi .
This group is the group of coördinate transformations Qq (the vari-ation in q is a function of q only}. A realizvari-ation of tne algebra
L(Qq} is xna x (or Xln1 ••• X. ~ d , i
=
1, ••• , k ; k-dimensional).K X.
L
!he set qpn, n integer ~
°
forms an algebra. !he group correspondingherewith is the group of momentum transformations Q ;
p
~ n-I" n!he 1" k-d' . .
uq = E n q p ,up '" -E P • genera LzatLon to LmenSLOnS LS
as before. We can check immediatelythat Qp and Qp are groups,
for lpqn, pqmj
=
(n-rnJ pqn+m-I.an element of L(IC(2H. On the other hand: every element of L(K(2}} can
be written as a Poissonbracket la,bJ with a € L(Qql and b € L(~). We
can formulate this as follows:
We ean also very simply prove the following statement: L(K(2» is the algebrale closure of L(Y ) and L(Y ). Por the n-dimensional case
q p ,
the same is true, the proof is somewhat more complicated but still simple.
Finite dimensional subalgebras are
I) !q2, !p2. qp with relations liq2, !p2
J ...
-qp,L!q2,pqJ ... q2, llp2.qpJ ... _pi, forming ~n algebra isomorphic
with the algebra of SU(2i or SL(2,R}. Thisset is the algebra of
the group SP(2}. the group of linear canonical transformations,
I
'0 I,leaving the s;YlnPlectic matrix S'" -1'.0 invariant, this means
t
a S a ' " S. a € SP(2).
SU(2) is the group of unitary 2x2 matrices witb determinant equal
to one. SL(2,R) is the group of real 2x2 matrices with determinant
equal to 1.
2) p, pq, ~pq2, finite subalgebra of L(Q ):L(C ). q q
e
q stands for the group of conformal transformations in q. In this one-dimensional case: the translation, dilatation and the special conformal trans format ion (generated by !pq2). For more details see the end of this section.3) q, pq. !qp2, fini te subalgebra of L(Qp): L(Cp)' Cp the group of
conformal transformations in p.
L(K(2)} is the clusure of L(C ) and L(e ), everyelement
q p
a € L(K(2» can be formed by (repeated) Poissonbrackets between
e1e-ments of L(e ) q and L(e ). p We shall write K(2) ...
e*
q C for the groups p corresponding to the algebras, K(2) is the "product" of tbe groupsC and C •
1~e ean prove the foregoing statement by induction:
llpq , !qpJ c q p , lqn-l pn, q pJ • _2qnpn+l etc.
n n+1 k 1
From every q p we ean construct an element of the form q p by
forming lqnpn+l. q(p)J, and this several times.
Every g(q,p) generates a one-parameter group of transformations. If we take g2, g3 or f(g) resp. we obtain the same trajeetory in phase-space as generated by g, the only difference is the velocity with whieh the trajeetory is passed. (It is possible that for a certain value of the parameter the trajeetory stops, the transformation beeomes singular.) We ean see this immedia tly •
2"
~a ~a(g} 2g
g, g
=
dq
3P - apa,q
The set g,g2 ,g3, ••• is the generator set of an infinite parameter group of transformations in phase-spaee, all possessing the same
trajeetory. The group is abelian. we call this group the
trajeetory-f 1 ( ' ) -
a
a
group 0 g. Examp e: g q,p • qp, g .. p ap - q aq
exp a.
g
q..
q·e-a. q(a.)p p·ea p(a.)
exp a. g2
I:
q'e -2apqp·e2a.pQ
J q q2 q3 q" " "
..
" ..p qp q2p q3 p p2 qp2 q2pZ p3 qp3
p"
We ean now representa trajectory-group by a straight line through the central element 1. Every element on this line is a generator of the trajectory-group.
(3)
(g)
(I)
(I) is the trajectory-group of qP. (21 is the t-group of pqz. '(31 that
of q, etc. Every set qkpl
with~
a fixed rational number is the setgenerators of a trajectory-group.
Every element of the algebra L(K(2}} generates a mapping of the
alge-bra into itself. If we use Ig > for an element of L(K(2» the mapping
ean be defined in this way: h
!
g >=
I lh,gJ> (or h.g=
h.g). With respect: to tlie basi.s of the algebra consisting of the homogeneouspolynomial s the 1!lapping is a linear representation of the algebra
(infinite dimensional matrices). I f the mappingis anto h is an
automor-phism', moreprecj.sely an innerautomorphism: (an outer automorphism of the algebra is an automorphism due to operators not contained in the
alge-bra itself).
Remark: For groups too there is a "natural action" of the group into ieself: define a
I
b > =I
a'b> , a. b E K(2). This leads- to the known concept of adjoint representation for groups~ In
analogy we can speak of the adjoint representation of an alge-bra.
Examples of automorphisms of L(K(2) are
d + 0 0 0 0 0 f' h f '
-q oq p op. q op' p oq'
ä'P'
ä'q'
love operators, togee er ornung analgebra isomorpbic witb L(ISP(2» (ISp(2) is ebe inhomogeneous
symplec-tic group in two dimensions, the symplecsymplec-tic transformations together
with translations in p and q).
Tbree dimensional case.
We shall construct L(K(6l}. an infinite'parameeer algebra, from a
num-ber of finite-dimensional subalgebras with known properties and
reali-zations. Besides the algebralc closure we use a few more concepts:
Tbe direct product of two groups A and B is a group C. such that
every element of C can bewritten uniquely as a product of an element of A and one of B.
C .. A cS) B = {c
I
c - a'b " a € A" b EB" a'b .. b ·a}Tbe algebra L(C) is the direct sum of L(A} and L(B); L(e) ..
L(A}
e
L(B} means LL(Al. L(B>J=
0 and LL(A).L(A)J ~ L(A),LL(B}, L(B}J ~ L(B}. For semi-direct sum or product we use the
symbols
e ,
s ® • s C .. A <B> s B means A is an invariant subgroup of C, or-1
e a c E A,
va
E A and \Ic E C. Au element e € C ean not be written.as a'b in a unique way.
Tbe corresponding relation for the algebras is: LeC) .. L(A)
e
L(B),s '
We can construct subalgebras of L(K(6)) and so subgroups of K(6):
*) !he set Ti - Pi (i=I,2,3) translations in qi' generators of an
abelian three parameter group,
*)
{Ti}
=
{qi} translations in Pi' generators of a group isomorphicwith the three-dimensional translation group. subgroup of the Galilei-group.
*) D • Piqi' generator of the diiatation group. the group of scale
. -a a
transformat~ons qi + qie ,Pi + Pie ,
*) {R.} = {E"k q.Pk}, generators of the three-dimensional rotation
~ . ~J J
group 50(3), or more precisely: a canonical realization of this
group. !he defining, three-dimensional, irreducible representation of
this algebra is
I
' , '·1
' • 1
,,-1 • .~.
, ,-I 1 •
!he action on (q 1 ,q2 ,q3) is:
R· exp aRt' exp
B
R2, exp y~, col(q),q2,Q3)=
o
o
cose
0 -sin 6 cos ..y
sin y 0o
cos a sin ex 0o
-sin y cos y 0o
-sin ex cos a sinB
0 cos 6 0o
and the action on
p
is quite similar.p
+R'p.
We can choose anotherbut this is not :essential. The latter will be cal led a canonical
-parametrization. It is because R is an antihermitian operator that p
-transforms in,the same way as q.
*) {M •• }
~J {q.p.- ;. qlPlo .• }, just a's before the summation convention ~ J ~J
is used. M .. are the generators of the special linear group SL(3,R),
1.J
the group of rea I 3x3 matrices with determinant equal to one(without
the term qlPlOij we get the generator set of the group L(3,R), the
linear transformation)
*) {Ki} - {QkqkPi-2QiClkPk}' the generators of the special conformal
transformations in
q.
LK.~.J - 0, so the K. are the generators of an~ J 1.
abelian group.
*) {Ki} - {PkPkqi-2PiqkPk}' the generator-set of the special
confor-mal transformations in
p.
Ti' D, Ri and Ki form an algebra, the algebra of the conformal group
in three dimensions (a ten parameter group). Thecommutation relations
are: LR. 1. ,R. J
J
= 1;:. ' •• 1.J .... 1L -1<. LR.,DJ
=
LT.,DJ=
LT.,T.J ~ 1. ,1. J LR.T.J .. e:"k T_ 1. J 1.J l t LK. ,DJ .. K. 1. 1. LK.,K.J '" 0 1. JLKpll)
=
E:ijk ~ LIL;T.J .. -20 •• D -2e:"k R. ~ . J ~J ~J-1<.The infinitesimal transformations in qi and Pi are:
T. : Cli ... qi - e: 0 ••
J ~J
Pi ... Pi
R. : qi ..,. qi - e: jik qk J
D Pi -T Pi - E j ik Pk qi ->- qi - Eqi p. ->-1 p. 1 + EBi K. : qi ->- q. + E (2 qiqj - óij qkqk) J 1
Pi ->- Pi + E (2 qiPj - 2 qjPi - 2 óij qkPk)
Remark: The finite transformations generated byK. are: 1
,..". ... ,.,. ... ... "'-1
q. ->- exp (a.K.)·q. = (q. - a. q-q) (1-2a'q + a"a q"q)
1 JJ 1 1 1
The expression for exp (a,K.)·p. is fairly complicated, this J J 1
will be computed (just as the transformation of qi) in appendix A.
Quite similarlvTi, D, Ri' Ki are the generators of a group, isomorphic with the ten parameter group of conformal transformations in p, the
group Cp' The commutation relations 'are
LR. ,R.
J ..
E:" k R. 1 J 1.J-1<. LR. ,TlJ .. -E:. 'k Tk' 1. J 1.J LK!,DJ .. -Kl 1 1. LT!,T!J=
LR.
,DJ 1. J 1. LD,T!J=
T! 1. 1. LR. ,K!J - E:" k K-' 1. J lJ-1<. LK! ,T!J 2(ó .. D - E" k R.) 1. J 1.J l J - l t LK! .K!J ..
0 1. J!he infinitesimal transformations in q and pare:
Tl: q1.' ->- q1.'
J
Pi"" Pi + E Ó ij
Kl: ql' ->- ql' + E: (2p.q. - 2p.q. - 2ó··qkPk)
p. + p. + e (PkPko .• - 2p.p.)
1 1 1J J 1
D and R.: as before.
1
The finite transformations exp (a.·K!)·p. c.an be c.a1c.u1ated in the J --1 1
same way as donein appendix A.
... ... ... -1
exp (a.'K!)·p . • (p. + a. P'p) (1 + 2a.p. + a-a P'P)
J J 1 1 1 J J
The group C bas two Casi:Jnir operators (homogeneous po1ynomia1s,
. q
. bui1d from tne generators and commuting with all of them). A general
expression for the nth order Casimir operator is:
with
I.
the generators of the group and1
k
.Here the structur&<eotlstants c. of the group are defined by.
ij
The metric tensor of the group algebra is g with components gij' defined by
~ ~ i
ij
A sufficient condition for the existence of g is
11
g ••11
:(0 0~J
A group (algebra) is cal led simple if there exists no invariant proper
subgroup (subalgebra), and semi-simple if there is no invariant abelian proper subgroup (subalgebra).
lor simpie and semi simple groups Ilg .• 11 :(00 ([3]). The conformal
lJ
group has no invariant proper subgroups, so it is a simpie group and gij exists. The conformal group in three dimensions is 10caIIy
isomorphic with the group 80(4,1) (the group of orthogonal
trans-formations Ieaving the metric diag. (+1,+1,+1,+1,-1) invariant).
This ~roup has two independent Casimir opé~atQ~s (see e.g.
[3]). It tnrns out that tbe only non-identically zero operators are
C
2 andc
4•
C
2 can be simplified using the definition of gijC
-
=
ij 2 g X.X. 1 JWe find for
C
2 andC
4 af ter some calculation:
C
2 =
i<T.i.
1 1 +i.T.}- R.R.
1 1 1 1 +15 15
These operators are zero in this special realization. The two casimir
operators for Cp are
C
2
andè
4
with all generators replaced by theirprimed analogues, tbis means that C
2
and C4
are zero two in thecano-nieal realization. The conformal group in four dimensions (x,y,z,t) is treated e.g. in [8].
The algebrale closure of L(Cq) and L(SL(3» gives the set
Pl,qiPj,qiqjPl' ••• , qi ••• qkPl •••• , or the set f(q) Pi with f a polynomial in qi' This set is the generator set of the group of eoördinate transformations (or the "genera 1 eovariance group"; in the four dimensional "t.elativistie" case this
nO n1
transformations with generators ~ XI
is the group of coordinate
n
2 n3
x2 x3 0x.' [7],[8]).
l.
The closure of L(e ) and L(SL(3)} gives in the same way the algebra
p
g(P)qi' g a polynomial in Pi' The corresponding group will be called the "general contravariance group". For the proof ofthe two statements above see appendix B. The closure of the algebras of the general covariance - and general contravariance group gives the algebra L(K(6», this is analogeous to the one-dimensional case and the proof is straightforward.
We can set up a diagram for the algebra L(K(6}) or the group K(6). The dimension of every group (numher of generators) is given at the lower left corner of the boxes. The convention is in a semi-direct product
to write down first the invariant subgroup, so ISO(nl
=
Te
SO(n). ;s
(ISO(n) is the inhOmb"geneous ó~thogOnai group in n dimensions; T,the
n~dimensinnal'abelian translati6n:gro~p, SO(n)'the group of orthogonal
fransfórmat:ions}~l"Further;.A"" B !!' C means L(er is the closure ,of
L(A) and L(B).
Canonical group K(2n) K(2n) = COV * CONTR.
00
'COV: general covariance group C
*
SL(n) q 00 C :conformal group of q Cq=K*[T® (SO(n)~)] q s Hn+l) (n+2)CONTR.:general eontra~arianee group
C p
*
SL(n)linear C confo~l group of p.
. P=K*[T'® (SO(n)®D)]
p s
!~n+l)(n+2)
K:special conformall T®s(SO(n)®D) transformations 1 , - - - " , . K ":special 'éonformal transformations of p n Hn+l)n+l T: n ln(n-l)+l n
We now return to K(6). There are several other subgroups of K(6) with
physical importance. E.g. take the set H = !p.p.,R. = E .• kq.Pk'
L L L LJ J
P'i =Pi' Gi =qi' I= 1.
The commutationrelations are
lH,P.J = lH,R.J = 0
L L
lG.,p.J = 0 •• 1.
L J LJ
LH,G.J = -Po L L
The algebra formed by this set is isomorphic with the algebra of the
extended Galilei-group ([9],[10]), that is the Galilei-group extended
with the central element 1 (LG.,p.J = 0 .. I in stead of LG.,p.J = 0).
The interpretation is simple: H is the Hamiltonian of a free. nonrela-tivistic particle without spin, R. are the generators of the rotation
~
subgroup (orbital angular momentum) and P. are the components of the
. J
linear momentum of tbe partiele; Gi at last are the generators of the
pure Galilei-transformations.
1.6 Constants of the motion.
From two constants of tha motien we can build new ones by taking the Poissonbracket or the product. The set of cORstants of the motion forms
a Lie algebra with respect to
L,J
and also an associative algebrawith respect to x.
Given a Lie algebra A, by forming products of elements of A we get an
algebra Ä, an algebra with the relations
l.
j,
x (and +). We shallcall
Ä
(as usual) the universal enveloping algebra of A ([11]).Although tbe elassificatien of the constants of the motion will be treated in Part II, we shall give here a few simple examples: the two dimensional harmonic oscillator, isotropic and anisotropic ([12]). The three dimensional isotropic oscillator. is treated in [13).
*)
The two dimensional harmonie oscillator, isotropie, Mass andfrequenty are choosen equal to one for convenience.
There are four linear independent quadratic constants of the motion: Hand further
A
=
i(pt + q~) - i(p~ + q~)M = P2ql - Plq2 P
=
PIP2 + qlq2The four operators
H, A,
H,P
working on eol(ql,pl,q2,p2) have a four dimensional matrix representation (the defining representation ofSp (4);
H, A,
Mand Pare the generators of a subgroup of Sp (4)) •0 0 !6 -1 0 -1 0 0 0 !6 !6 -} 0 I· !6 !6 .,.} -1 0 0
o
-1o
o
o
o
o
-} -1o
H
generates rotations in ql,pl and q2,P2 plane simultaneously, Àgenerates rotations in the same planes, eloekwise and eountereloek-wise respeetively,H generates rotations in Ql,Q2 and Pl,P2 plane and
P at last rotations in QhP2 and Q2,Pl plane simultaneously. A,M,P are generators of a group isomorphie with 8U(2). The four dimensional representation of 8U(2) dedueed from the representation above is redueible and ean be redueed to two irredueible representations of
8U(2) (or U(2) if H is ineluded). The symmetry group of H is U(2).
Remark: The four dimensional representation of Sp(4) is irredueible. The algebra of eonstants of the motion has a basis AnI Mn2 pns, ~I,n2, ns integer,:: O. The transformations generated are regular in the whole phase spaee.
H2
=
A2 + M2 + p2. The eonstants AnI Mn2 pns are regular eonstants of the motion (the transformations are not singular) [12] [21].*) Two dimensional harmonie oscillator, anisotropie.
Two~ther constants of the motion are:
G • aretg Sl. -
1.
arctg wq:zPI W P2
Remark: arctg Sl. and
i(qt
+ pi> are the aetion-angle variables of thePI
one-dimensional harmonie oseiliator.
H and A are regular constants of the motion, G not:
~I
=
LG,qd ..
Pt;~f
:&1
=
lG,p;j -
pi!~i
~~2 = LG.o~J -q2
<IM ,C>&. p~+q~.'W2
(6.1 )
Gis the generator of a transformation whieh is a contraction in the
Q2,P2 plane and an expansion in the ql,pl plane. If we write
q.
=
q.(O),p. •
p.(O) for the starting values of tha trajeetory1 1 1 1
and from this
<h
and similar formulas for q2 and P2
pi
+ q1 = 2i + Pl + q1p~ + w2q~ = -2a •
P!
+ w2q~Solutions of (6.1) exists for the range
(6.2)
be zero (if a - ao say) and so ~he transformation (6.1) becomès
singular and the trajectory exp a
G .
col (ql,Pl,q2'PZ) stops. G isnot a regular constant of the motion ([2Il). With
arctg x =
h
ln[ (t+ix)(q-ix) -1] the function G can be written as:(6.3)
If K
=
(PI+iql)W (p2-iooqz) then both K and K* (complex eonjugatión)are constants of the motion. With this definition (6.3) ean be written as (2iw)-1 In K/K*. We ean in stead of K and K* also choose
the real functions Kl and K2 defined by
H,A,Kl,~ do not form a closed se.t, but we can form the closure of this set with respect to l,J and x, this results in an algebra of
constants of ~he motion.
• if wal we find constants of the mot ion of even order only.
• if wa2 we find H, A (of order two) and
Kl - qlPlP2-Q2P! + ~qt. ~ = P!PZ-Q!P2 +
<.
qlqzPl (of order three)af ter closure we get constants of arbitrary order.
• if 00-3 we get H and A (second order) and two fourth order constants
of the motion: Kl • Q1P2(3PI-Ql)-3qzPl(pt-3ql) >
K2
=
PIP2(pt-3qlJ + 3qlq2(3Pl-Ql}. The algebra formed af ter closureconsists of even order constants only. etc.
The closure of {H,A,Kl,K2} is, by definition, the algebra A. We can
extend this algebra A to an algebra
Ä
by adding formal power series.Fram this algebra Ä we can select four quantities with special
commutation relations. H A 00-1 Kl [!(ql + Pl)]-
-Z- •
Kl 00-1 K2 [!(ql + Pl)]- --2-- • KzThese quantities "fulfill the relations
LH,AJ
=
LH,Kd .. LH,K2J - 0LKl,KzJ - 2wA ~K2,AJ • 2wKl
We cbuld have used K and K* in stead of Kl and K2 but this is not es-sential. The set {Kl,K2,A} forms an algebra isomorphic with. the
alge-bra of SU(2). Aresult analogeous to the two dimensional isotropic
oscillato~. One remark should be made. The isomorphism is alocal one:
there always is a neighbourhood of the identity (parameter ~=O) where
the isomorphism holds. For the isotropic oscillator the transformations generated by H,A,P ,M are regular everywhere in phase-space, this is
not true for the anisotropic oscillator, for certain values of ~ the
transformation (generated by H,A.Kl.K2) become singular. The symmetry groups of H (isotropic) and H (anisotropic) are therefore not global isomorphic, only local.
2. Continuous systems.
2.1 Canonical transformations.
We shall treat systems which can be described by two canonically con~
jugated functions q(~.t), p(;,t),~ is the independent space variable
and q and p are vector functions in genera!.
The Hamiltonian H now is a functional of q and p. The equation of
motion is equivalent to the hamilton equations.
(I. 1)
ö
H{q.p}
J
d~
H(q,p,q .p ,q x. x. , .... ).X.X.
1 1 1 J
H is the Hami~tonian density of the system and is a function of
q,p,qx.'px.' etc. In general there can be n independent x-variables. 1 1
but we shall restrictourselves to n-l,2,3 which are the most natural situations. The Hamiltonian density H can depend explicitlyon; and
t, in that case H is functional of q and p and a function of t.
...
-
...
We can construct eanonical transformations q(x,t) + q(x.t) •
....
-
....
-
-p(x,t) + p(X,t), q and pare functionals of q and p.Just as in
section 1.1 the construction ean be done in two ways, with a
genera-ting funetional F (a functional of one "old" and one "new" funetion)
or via a generator of an infinitesimal'transformation, a functional of q and p.
*
Possible choises for F are:Fl{q,q,t} , Fa{q,p,t} , Fs{p,q,t} and F4{p,P,t}. We give, for com-pleteness, the equations whieh determine the transformations.
jj - H
=
~t
Fl {q.q,t}a
o
--r-F3{P.q.d Op.
1.
-
-!he new field variables q and p fulfill the relation
(1.2)
!he Poissonbraeket
l.
J
betweentwo funetionals 11 and G is defined inthe following way:
(1.3)
We have written here
~~Çi)
in stead of~~
to note that thefunetio-nal derivative is dependent on ~t we shall not do that in general,
unless confusion migbt arise. !he identieal transformation ean be
generated by F2
=
J
d~ q(~.t) p(~.t).
!he eanonical trans format ion whieh transforms theHamilton equations to the equations
oH
-!p - qt .. 0
oH
will again result in a kind of Hamilton~acobi equation. Let the
trans-formetion be generated by Fl and take
H.O.
the equations then are:a .
-H {p,q,t} • - ätF1{q.q,t}. (l.5)
taking together:
(1.6)
*
Tbe second metbod is to use a transfo'rmetion close to the identity.A generating functional is
f
d3; q(;.t) p(;.t} + eG {q,P.tl. € is a smeH parameter G.will be
cal led tbe generator of the transformation.
OG P = P - e: öq
-
öG
q=q+e:5'P
Tbe trans format ion is an infinitesimal one. sa we can treat everything
up to order €. In G we mey change
p
in p. Under the transformationabove tbe first order variations in the fields q and p become ([I]).
öq = €
* -
€ lq.GJ2.2 Invariant canonical transformations.
Under an infinitesimal transformation generated by G{q.P,t} the varia-tion in the Hamiltonian is defined as:
oH ... H{q,p,t} - H{q,p,t} (2.1)
This can be written
OB oH
OB ...
"8P
op
+oq oq.
(2.2)or, with (1.1) and (1.3)
(2.3)
With q and p solutions of theHamilton equations, we find for
:~
the expression dG ... dG + LG,HJ dt dt Combining (2.1)-(2.4} we get 8
!~
...
8~~
+"8 lG,HJ ... n{q,p,t} - H{q,P.t} - H{q,P,t} + H{q,P,t} ~ dG - - - __ 8 dt ... H{q,p,t} - H{q,p,t}. (2..4)The trans format ion is an invariant transformation if
n
is the sameH{q,p,t}
=
H{q,p,t}H{q,p,t} = H{q,p,t}.
remark: i f G doesn't depent explicitlyon t tben H{q,p} = H{q,pL So
the value of tbe functional is constant under tbe
transfor-mation q,p ~ q,p.
We can conclude (analogeous to tbe discrete transformations): The generator of an invariant infinitesimal canonical trans format ion
(invariant variation) is a constant of the motion and vice versa.
2.3 Special transformations. Campbell-Baker~Hausdorff-formulas.
The set of functionals G{q,p,t} with the operations + and
l.J
formsaninfinite,dimensional Lie algebra.
The structure is much more complicated tban in tbe discrete case
(section 1.3). Apart from transformations in the fields q and p only,
there are also transformations in the independent variables, so tbere is an enormeous increase of the degrees of freedom. A lattice as given in section ].5 seems not be poss,ible for tbe algebramentioned above.
Mncb of the theory given in chapter
i
can be used here, although withtbe right modifications.
We can associate an operator with every functional G{q,p,t} in a way analogeous to that of section 1.3.
M o ·OGo
q and p are vector functions andthe dot stands for an inner product.
The operator G is acting on functionals and functions. such as q and
p (we can interprete this functions as functionals if we use a delta-function).
G'F • lG,F
J
There is an isomorphism between the algebra of functionals with
operation
L,J,
and the algebra of operators with operation[,1.
(3.2)
Finite transformations ar~ constructed in the usual way by exponent
i-ating the infinitesimal transformation: exp a G, the exponent inter-preted as a formal power series:
~
V
generates a one-parameter subgroup of transformations exp a G.Let F be another generator of a one-parameter subgroup of trans
forma-tions exp ~
F.
The product of two elements of the subgroup is againa canonical transformation.
-
-exp a G • exp ~ F • exp L (3.4)
The relation between L and ehe operators G and F can be given by
~
-
~+ :2 a2f3 [G,[G,li']] + . . . . (3.5)
An expression 1ike this ean a1so be derived for the operators of
-ehapter 1. It is sometimes impossib1e to write L in c10sed form.
A variation of a special kind is the one generated by
G -
J
d3~
[BCLPa + Bf3QS +PCLB~
qS], with~
a 1inear differentia1::ct CL ai aij
operator BS • BS + BS di + BS didj + ••••
The coëfficients of ~ and Ba.
Ba
are funetions of the independentvariables. Greek indices are used for the field;eomponents ~nd latin
indices for the components of the x-variables. The summation conven-tion is used. G generates the most general inhomogeneous linear canoni-cal varia tions
(3.6)
!he trans format ion (3.61 with. all B except Ba zero is a generalized
trans lat ion in q-space (included in the Gauge transformations of q,
introduced in chapter
11.
if Ba '0,
all other B are zero thetrans-formation is a generalized translation in p-space (Gauge trans forma-don in p).
!he transformation (3.61 with Ba - B = 0 is the most gener al linear
a
coordinate transformation (oq depends on q alone, not p, and in a
linear wayl sometimes called a point transformation [I]. A subset of
these will be cal led internal transformations or local Gauge transfor-mations (of Gauge-theory), because they change only the fieldcomponents
(do not transform the independent variables). This kind of Gauge transformations is not the same as the kind introduced in chapter 1.
a
ex •
...
In the special caSe BB and B ~ndependent of x and t the trans
forma-tion is a global Gauge transformaforma-tion. Transformaforma-tions of the type
'i '
Xi is a function of (i,t), are transformations which aan be thought
as coming from transformations of the independent ~ variable:
- .... .. ai a i
x + x + X(x). In this case is BB
«3.6»
aqual to ~B X • Exampla:x.a gives ~qa - € x. qa, •• There are only first order derivatives,
~ x
j ~ J
so the subgroup of this kind of transformations is isomorphic with the
general covarianee group of chapter I (the general coordinate
trans-formations of the independent variabie
ilo
The generator of thetransformation
~qa _ Bai
B
+ Baije
+IJ (3 q , i f3 q. ij •••
(3.7)
can be written as
. (3,S)
~ is a function of the variable Xi and the differential operator ê
x.' If we ta ka =aB 13 • \,)13 ~a nl n2. na" Xl X2. X3 0x.' or all f ' poss~ble values of nlon2.. J
n3' we have the basis of the algebra of the general covarianee group.
There is an isomorphism between generators ,G, with Poissonbracket and
-a
.
the algebra of operators BB' ThlS can be seen by taking the
Poisson-bracket between two generators Gl and Ga of the type
(3.8).
Let ~13 and ~ be the corresponding operators, then af ter some calculation:
(3.9)
2.4 Relation between finite and infinitesimal canonical transformations.
Just as in section 1.4 we can ask for the relationship between the generating functional of a finite transformation and the generator of an infinitesimal one. Much of the problems here are similar to those of section 1.4. Oae of the problems is the non-existence of the trans format ion for certain parametervalues.
Suppose we are working with Fa{q,p,a}. If the trans format ion generated
by Fa becomes singular for a - ao we then can use F3{p.~.a} as the
ge-nerating functional in the neighbourhood of ao. Fg and F2 are related
by a Legendre transformation.
A functional F transforms under a finite transformation generated by
the generator G, according to
The Poissonbracket between two functionals can be taken with respect to any arbitrary pair of canonically conjugated variables,so
lF,GJ
q,p• lF,GJ- -
q,pIn the right hand side of (4.3) F and G must be written as
F{q{q,p}, p{q,p}}, G ·analogeous.
(4.3)
Let the generator of an infinitesimal transformation be given:
G{q,p}. We ask for tbe functional F2{·p,q,cÛ. q • q{q,p,cx} and
p •
p{q,p,a}(4.4)a
(4.4 )b
The indices of q,p and q,p will be ignored just as before, as far as possible. The formula (4.4)a is, more precisely:
l-
GJ ..
L
(oF2 ) +I
d3~dPiL(oF2)
q, <la - da -
-op
op.op
~
Th e f unct~ona . l d ' erlovatloves . -=- , -Ö Ö can e lontere ange , so b ' b d
öPi op .
l -
q.G • (laJ
<l (-=-)oF
2 +I
d:x3~l-
ppGJ
--=-,
oqiop
op
OQi 0 - --=-means-=
qi{q,P{q,p},a} I op op q.~ (4.5) (4.6)q can be written (in a formal way) as functional of q and p but also
of q and p. The transformation q,p + q,p is supposed to be invertible.
J
3" öG Oqi + d Xoq. / -
0- / L P P q,a (4.7) dThe left hand side can be written as 0 / :aFz
ol' N
'äcl
q,VI.
The right hand side
Tn the same way can be derived ([15])
(4.8)
Combining (4.7) and (4.8) we get
oFz {-{ - }-}
aa-
=
G q q,p,a ,p + function of a (4.9)or,
oFz = G{oFz
p}
=
G{q, OFZ}oa
op'
oq
(4.10),withG{q{q,p,a},p} = G{q,p{q,p,a}
The initial condition is
-
J -
3-Fz{q,p,O} q p d x.
From this condition: p(q,p,O) = p. The expression (4.10) is again a
generalization of the Hamilton-Jacobi equation for the continuous
case, G is the generalization of the Hamiltonian and Fz (the
functio-nal we search for) is a generalization of the action functiofunctio-nal and the parameter a plays the role of t.
2.5 Example of canonical transformations and canonical transformations as realizations of finite-parameter groups.
Let us at first stage take the simp lest case with one independent x-variable and one field variabie q. so q(x.t}.
The set xn3x' n integer ~ 0 is the generator set of the coördinate
transformations in tbe variabie x.
G
=
I
p(x) xn3 q(x) dx is the canonical realization of this setn x
(in the continuous case). A subalgebra of the set above is
-3 • x3 • x23 with commutation relations.
x x x [x3 , x -3 ] x = 3 x [x3 , x2a ] x2a x x x [x23 x· -a ] x 2xa x (5.1)
This algebra is isomorphic with the algebra J3, J! of the generators
of the group SL(2). (SL(2,R) and SL(2.~». or SU(2). respectively
the group of 2x2 real (or complex) matrices with determinant I and
the unitary 2x2 matrices, det = 1.
The operators in (5.1) are therefore a realization of L(SL(2,R» ,
the isomorphism is
-3
++1
0x. -1
:1
x3 x++
o
-1o
o
The one parameter subgroups are exp(-a3 ). exp (S Xd ) and
x x
exp (yx2a ). working on a function f(x) in the fOllowing way
exp (-aa ) • f(x) • f(x-a) x exp (~x3 ) • f(x) • f(x exp S)
x
exp (yxZax) • f(x) - f(l~x)
Tbis is a realization of L(SL(2» in one variable .~y Lie derivatives
(first order differential operators). One ean make, more general,
realizations by generalized Lie derivatives~ operators of tbe form
f(X) a + g(x). For tbe above situation we ean find a realization of x
L(SL(2» by "solving" the eommutation relations (5.1) (substituting
operators f(x) a + g(x) for -3 , xa and x2a ), see [3] and [16].
x x x x J3 ++ À + x3 X J+ ++ (2À + ~)x + x2
a
x (5.2)Tbe genera.lized Lie-derivatives give rise to a "loeal multiplier
re-presentation" of the corresponding group (SL(2) in this case), this
will be denoted by
[TV (g) • f] (x) - v(x;g) f(gx) (5.3)
In here g is an element of the group, v(x;g) is a local multiplier
depending on x and g}. TV is the representation with V the label of
the representation. Beeause of the groupproperties:
If J.
€
L(G). the algebra of the group G, the aetion of exp a.J. onJ J J
f(x) will then be given by solving the following equations (gener al i-zed to n • x-variables).
P.(x(a». J g
=
exp a.J.e:
G J J o = X. 1. (5.4)in here i and 1 runs from 1 till n, j from I to N. the number of
generators of the group G.
The trans format ion (5.3) of the fields q is generated by the
genera-ting functional.
- J
~ -1 -1F2{q,P} = dx peg x) v(g x;g) q(x)
(if gx runs through the whole integration area if g is fixed)
q(x)
=
v(x;g) q(gx)p(x) = v(g-lx;g) peg-lx)
g
e:
G, the group of transformations in x.In some cases G is alocal group of transformations, defined in a
neighbourhood of the identity but not for (some) finite parameter-values.
Au eXámple~ The operator set (5.2} with À
=
-in, n integer ~ 0 and~
=
O. A group element can be written as+
-ex yJ3 exp SJ exp aJ
e:
G.We use (5.4) to compute tbe action of f(x}.
[exp yJ3 exp SJ+ exp aJ-f] (x)"exp(-jny) [exp SJ+ exp aJ-f] (x exp y)
~ exp (-!ny) (1-8 x exp y)n [exp aJ-f]
(IS
exp y ).. I. ' (_1 ) _ Q
e
(1»n
f(x(1+il!3)eYJ
(lr) -ae,(-~x»
~exp 2X X ~ xp 2X exp
(-Ix -
Xa
expIx
'
The matr~x representation isexp
~X
0I I
1o
exp(-ix) •
0o ..
(1+a!3) exp6x
-a exp (-lr)i f we write this as
I: .
:1'
ad-be: = 1, then we find for the ae:tion on f(x) the expression(TV (g) fJ (x) .. (bx+d)n f(~~:a)
-a
exph
exp (-h)if n-o this is a frae:tional linear trans format ion in x, a non linear
realization of SL(2)
End of exampie.
If we drop the restrie:tion to first order differential operators we
e:an find other realizations of L(SL(2»:
J3 ++
-i(xa
x + AlJ+ ++
la
2 x J ++iX2
This is one possibiiity, however we prefer a set of anti-hermitean operators, the operator exp aJ then is a unitary operator for il reai.