<?xml version="1.0" encoding="utf-8"?>
<?xml-stylesheet type="text/css" href="article.css"?>
<!DOCTYPE article PUBLIC "ISO 12083:1994//DTD Article//EN" "article.dtd">
<article>
<front>
<title>Non-Noether symmetries in singular dynamical systems</title>
<author>
	<fname>George</fname>
	<surname>Chavchanidze</surname>
</author>
<aff>
	<orgdiv>Department of Theoretical Physics</orgdiv>
	<orgname>A. Razmadze Institute of Mathematics</orgname>
	<street>1 Aleksidze Street</street>
	<city>Tbilisi</city>
	<postcode>0193</postcode>
	<country>Georgia</country>
</aff>
<abstract>
<title>Abstract</title>
<p>In the present paper geometric aspects of relationship
between non-Noether symmetries and conservation laws in Hamiltonian
systems is discussed. Case of irregular/constrained dynamical systems
on presymplectic and Poisson manifolds is considered.</p>
</abstract>
<supmatl>
<p>
	<emph>Keywords:</emph>
	<keyphras>Non-Noether symmetry</keyphras>,
	<keyphras>Conservation laws</keyphras>,
	<keyphras>Constrained dynamics</keyphras>
</p>
<p><emph>MSC 2000:</emph> 70H33, 70H06, 53Z05</p>
<!-- Georgian Math. J. 10 (2003) 057-061 -->
</supmatl>
</front>
<chapter>
<section id="intro">
<title>Introduction</title>
<p>
Noether's theorem associates conservation laws with particular continuous symmetries of
the Lagrangian. According to the Hojman's theorem <citeref rid="r1">[1]</citeref>-<citeref rid="r3">[3]</citeref> 
there exists the definite correspondence between
non-Noether symmetries and conserved quantities. In 1998 M. Lutzky showed that several integrals of
motion might correspond to a single one-parameter group of non-Noether transformations
<citeref rid="r4">[4]</citeref>. In the present paper, the extension of Hojman-Lutzky theorem to singular dynamical systems is considered.
</p>
<p>
First of all let us recall some basic knowledge of description of the regular dynamical systems
(see, e. g. <citeref rid="r5">[5]</citeref>).
In this case time evolution is governed by Hamilton's equation
<dformula>
i<inf>X<inf>h</inf></inf>ω + dh = 0,
</dformula>
where <formula>ω</formula> is the closed
(<formula>dω = 0</formula>) and non-degenerate
(<formula>i<inf>X</inf>ω = 0  ⇒ X = 0</formula>) 2-form,
<formula>h</formula> is the Hamiltonian and
<formula>i<inf>X</inf>ω</formula> denotes contraction of
<formula>X</formula> with <formula>ω</formula>.
Since <formula>ω</formula> is non-degenerate, this gives rise to an isomorphism between the vector
fields and 1-forms given by <formula>i<inf>X</inf>ω + α= 0</formula>.
The vector field is said to be Hamiltonian if it corresponds to exact form
<dformula>
i<inf>X<inf>f</inf></inf>ω + df = 0.
</dformula>
The Poisson bracket is defined as follows:
<dformula>
{f , g} = X<inf>f</inf> g = − X<inf>g</inf> f = i<inf>X<inf>f</inf></inf>
i<inf>X<inf>g</inf></inf>ω.
</dformula>
By introducing a bivector field <formula>W</formula> satisfying
<dformula>
i<inf>X</inf>i<inf>Y</inf>ω = i<inf>W</inf> i<inf>X</inf>ω ∧ i<inf>Y</inf>ω,
</dformula>
Poisson bracket can be rewritten as
<dformula>
{f , g} = i<inf>W</inf> df ∧ dg.
</dformula>
It's easy to show that
<dformula id="e6">
i<inf>X</inf>i<inf>Y</inf>L<inf>Z</inf>ω =
i<inf>[Z,W]</inf> i<inf>X</inf>ω ∧ i<inf>Y</inf>ω,
</dformula>
where the bracket <formula>[ · , · ]</formula> is actually a supercommutator,
for an arbitrary bivector field
<formula> W = <subform><roman>∑</roman><bottom>s</bottom></subform> V<sup>s</sup> ∧ U<sup>s</sup> </formula> we have
<dformula>
[X,W] = <subform><roman>∑</roman><bottom>s</bottom></subform>[X,V<sup>s</sup>] ∧ U<sup>s</sup>
+ <subform><roman>∑</roman><bottom>s</bottom></subform>V<sup>s</sup> ∧ [X,U<sup>s</sup>]
</dformula>
Equation <formref rid="e6">(6)</formref> is based on the following useful property of the Lie derivative
<dformula>
L<inf>X</inf>i<inf>W</inf>ω = i<inf>[X,W]</inf>ω +
i<inf>W</inf>L<inf>X</inf>ω.
</dformula>
Indeed, for an arbitrary bivector field
<formula>W = <subform><roman>∑</roman><bottom>s</bottom></subform> V<sup>s</sup> ∧ U<sup>s</sup> </formula> we have
<dformula>
L<inf>X</inf>i<inf>W</inf>ω = L<inf>X</inf><subform><roman>∑</roman><bottom>s</bottom></subform>i<inf>V<sup>s</sup> ∧ U<sup>s</sup></inf>ω =
L<inf>X</inf><subform><roman>∑</roman><bottom>s</bottom></subform> i<inf>U<sup>s</sup></inf>i<inf>V<sup>s</sup></inf>ω<break/>
= <subform><roman>∑</roman><bottom>s</bottom></subform> i<inf>[X,U<sup>s</sup>]</inf>i<inf>V<sup>s</sup></inf>ω +
<subform><roman>∑</roman><bottom>s</bottom></subform> i<inf>U<sup>s</sup></inf>i<inf>[X,V<sup>s</sup>]</inf>ω +
<subform><roman>∑</roman><bottom>s</bottom></subform>i<inf>U<sup>s</sup></inf>i<inf>V<sup>s</sup></inf>L<inf>X</inf>ω =
i<inf>[X,W]</inf>ω + i<inf>W</inf>L<inf>X</inf>ω
</dformula>
where <formula>L<inf>Z</inf></formula> denotes the Lie derivative along the vector field <formula>Z</formula>.
According to Liouville's theorem Hamiltonian vector field
preserves <formula>ω</formula>
<dformula>
L<inf>X<inf>f</inf></inf>ω = 0;
</dformula>
therefore it commutes with <formula>W</formula>:
<dformula>
[X<inf>f</inf> ,W] = 0.
</dformula>
In the local coordinates <formula> z<inf>s</inf> </formula> where
<formula>ω = <subform><roman>∑</roman><bottom>rs</bottom></subform>ω<sup>rs</sup>dz<inf>r</inf> ∧ z<inf>s</inf></formula> bivector field
<formula>W</formula> has the following form
<formula>W = <subform><roman>∑</roman><bottom>rs</bottom></subform>W<sup>rs</sup><fraction><num>∂</num><den>∂z<inf>r</inf></den></fraction> ∧ <fraction><num>∂</num><den>∂z<inf>s</inf></den></fraction></formula> where
<formula>W<sup>rs</sup></formula> is matrix inverted to <formula>ω<sup>rs</sup></formula>.
</p>

</section><section id="regular">
<title>Case of regular Lagrangian systems</title>

<p>
We can say that a group of transformations
<formula>g(z) = e<sup>zL<inf>E</inf></sup></formula> generated by the vector
field <formula>E</formula> maps the space of solutions of equation onto itself if
<dformula>
i<inf>X<inf>h</inf></inf>g<inf>*</inf>(ω) + g<inf>*</inf>(dh) = 0
</dformula>
For <formula>X<inf>h</inf></formula> satisfying
<dformula>
i<inf>X<inf>h</inf></inf>ω + dh = 0
</dformula>
Hamilton's equation.
It's easy to show that the vector field <formula>E</formula> should satisfy
<formula>[E , X<inf>h</inf>] = 0</formula>
Indeed,
<dformula>
i<inf>X<inf>h</inf></inf>L<inf>E</inf>ω + dL<inf>E</inf>h =
L<inf>E</inf>(i<inf>X<inf>h</inf></inf>ω + dh) = 0
</dformula>
since <formula>[E,X<inf>h</inf>] = 0</formula>. 
When <formula>E</formula> is not Hamiltonian,
the group of transformations <formula>g(z) = e<sup>zL<inf>E</inf></sup></formula> is non-Noether
symmetry (in a sense that it maps solutions onto solutions but does not preserve action).
</p>
<p><emph>Theorem 1.</emph> 
 (Lutzky, 1998) If the vector field <formula>E</formula> generates non-Noether symmetry, 
then the following functions are constant along solutions:
<dformula id="e15">
I<sup>(k)</sup> = i<inf>W<sup>k</sup></inf> ω<inf>E</inf><sup>k</sup>        k = 1...n,
</dformula>
where <formula>W<sup>k</sup></formula> and <formula>ω<inf>E</inf><sup>k</sup></formula> are outer
powers of <formula>W</formula> and <formula>L<inf>E</inf>ω</formula>.
</p>
<p><emph>Proof.</emph>  
We have to prove that <formula>I<sup>(k)</sup></formula> is constant along
the flow generated by the Hamiltonian. In other words, we should find that
<formula>L<inf>X<inf>h</inf></inf>I<sup>(k)</sup> = 0</formula> is
fulfilled. Let us consider
<formula>L<inf>X<inf>h</inf></inf>I<sup>(1)</sup></formula>
<dformula>
L<inf>X<inf>h</inf></inf>I<sup>(1)</sup>
= L<inf>X<inf>h</inf></inf>(i<inf>W</inf>ω<inf>E</inf>) =
i<inf>[X<inf>h</inf> , W]</inf>ω<inf>E</inf>
+ i<inf>W</inf>L<inf>X<inf>h</inf></inf>ω<inf>E</inf>,
</dformula>
where according to Liouville's theorem both terms
<formula>[X<inf>h</inf> , W] = 0</formula> and
<dformula>
i<inf>W</inf>L<inf>X<inf>h</inf></inf>L<inf>E</inf>ω =
i<inf>W</inf>L<inf>E</inf>L<inf>X<inf>h</inf></inf>ω =
0</dformula>
 since <formula>[E , X<inf>h</inf>] = 0</formula> and 
 <formula>L<inf>X<inf>h</inf></inf>ω = 0</formula> vanish.
In the same manner one can verify that
<formula>L<inf>X<inf>h</inf></inf>I<sup>(k)</sup> = 0</formula>
 ∎</p>
<p><!--Remark-->  
Theorem is valid for a larger class of generators <formula>E</formula> .
Namely, if <formula>[E , X<inf>h</inf>] = X<inf>f</inf></formula> where <formula>X<inf>f</inf></formula> is
an arbitrary Hamiltonian vector field, then <formula>I<sup>(k)</sup></formula> is still conserved. Such a
symmetries map the solutions of the equation
<formula>i<inf>X<inf>h</inf></inf>ω + dh = 0</formula>
on solutions of
<dformula>
i<inf>X<inf>h</inf></inf>g<inf>*</inf>(ω) +
d(g<inf>*</inf>h + f) = 0</dformula>
</p>
<p><!--Remark-->  
 Discrete non-Noether symmetries give rise to the conservation of
<formula>I<sup>(k)</sup> = i<inf>W<sup>k</sup></inf>g<inf>*</inf>(ω)<sup>k</sup></formula>
where <formula>g<inf>*</inf>(ω)</formula> is transformed <formula>ω</formula>.
</p>
<p><!--Remark-->  
 If <formula>I<sup>(k)</sup></formula> is a set of conserved quantities
associated with <formula>E</formula> and <formula>f</formula> is any conserved quantity, then the set of functions
<formula>{I<sup>(k)</sup> , f} </formula>
(which due to the Poisson theorem are integrals of motion) is associated with
<formula>[X<inf>h</inf> , E]</formula>. Namely it is easy to show by taking the Lie
derivative of <formref rid="e15">(15)</formref> along vector field <formula> E</formula> that
<dformula>
{I<sup>(k)</sup> , f} = i<inf>W<sup>k</sup></inf>ω<sup>k</sup><inf>[X<inf>f</inf> , E]</inf></dformula>
 is fulfilled.
As a result conserved quantities associated with Non-Noether symmetries form Lie algebra under
the Poisson bracket.
</p>
<p><!--Remark-->  
If generator of symmetry satisfies Yang-Baxter equation
<formula>[[E[E , W]]W] = 0</formula> Lutzky's conservation laws are in involution [7]
<formula>{Y<sup>(l)</sup> , Y<sup>(k)</sup>} = 0</formula>
</p>

</section><section id="irregular">
<title>Case of irregular Lagrangian systems</title>

<p>
The singular Lagrangian (Lagrangian with vanishing Hessian) leads to degenerate 2-form
<formula>ω</formula> and we no longer have isomorphism between vector fields and 1-forms.
Since there exists a set of "null vectors" <formula>u<inf>s</inf></formula> such that
<formula>i<inf>u<inf>s</inf></inf>ω = 0       s = 1,2 ... n − rank(ω),</formula>
every Hamiltonian vector field is
defined up to linear combination of vectors <formula>u<inf>s</inf></formula>. By identifying <formula>X<inf>f</inf></formula>
with <formula>X<inf>f</inf> + <subform><roman>∑</roman><bottom>s</bottom></subform>C<inf>s</inf>u<inf>s</inf>,</formula> we can introduce equivalence class
<formula> X<inf>f</inf><sup>∗</sup></formula> (then all <formula>u<inf>s</inf></formula> belong to
<formula>0<sup>∗</sup></formula> ).
The bivector field <formula>W</formula> is also far from being unique, but if
<formula>W<inf>1</inf></formula> and <formula>W<inf>2</inf></formula> both satisfy
<dformula>
i<inf>X</inf>i<inf>Y</inf> ω =
i<inf>W<inf>1,2</inf></inf> i<inf>X</inf>ω ∧ i<inf>Y</inf>ω,
</dformula>
then
<dformula>
i<inf>(W<inf>1</inf> − W<inf>2</inf>)</inf> i<inf>X</inf>ω ∧ i<inf>Y</inf>ω =  0        ∀X,Y
</dformula>
is fulfilled. It is possible only when
<dformula>
W<inf>1</inf> − W<inf>2</inf> = <subform><roman>∑</roman><bottom>s</bottom></subform>v<inf>s</inf> ∧ u<inf>s</inf>
</dformula>
where <formula>v<inf>s</inf></formula> are some vector fields and
<formula>i<inf>u<inf>s</inf></inf>ω = 0</formula>
(in other words when <formula> W<inf>1</inf> − W<inf>2</inf></formula> belongs to the class
<formula>0<sup>∗</sup></formula>)
</p>
<p><emph>Theorem 2.</emph> 
 If the non-Hamiltonian vector field <formula>E</formula>
satisfies <formula>[E , X<inf>h</inf><sup>∗</sup>] = 0<sup>∗</sup> </formula> commutation
relation (generates non-Noether symmetry), then the functions
<dformula>
I<sup>(k)</sup>
= i<inf>W<sup>k</sup></inf>ω<inf>E</inf><sup>k</sup>         k = 1...rank(ω)
</dformula>
(where <formula>ω<inf> E</inf> = L<inf>E</inf>ω</formula>) are constant along trajectories.
</p>
<p><emph>Proof.</emph> 
Let's consider <formula> I<sup>(1)</sup></formula>
<dformula>
L<inf>X<inf>h</inf><sup>∗</sup></inf>I<sup>(1)</sup>
= L<inf>X<inf>h</inf><sup>∗</sup></inf>(i<inf>W</inf>ω<inf>E</inf>)
= i<inf>[X<inf>h</inf><sup>∗</sup> , W]</inf>ω<inf>E</inf> +
i<inf>W</inf>L<inf>X<inf>h</inf><sup>∗</sup></inf>ω<inf>E</inf> = 0
</dformula>
The second term vanishes since <formula>[E , X<inf>h</inf><sup>∗</sup>] = 0<sup>∗</sup></formula> and
<formula>L<inf>X<inf>h</inf><sup>∗</sup></inf>ω = 0</formula>. The first one is
zero as far as <formula> [X<inf>h</inf><sup>∗</sup> , W<sup>∗</sup>] = 0<sup>∗</sup></formula> and
<formula>[E , 0<sup>∗</sup>] = 0<sup>∗</sup></formula> are satisfied. So
<formula>I<sup> (1)</sup></formula> is conserved.
Similarly one can show that <formula>L<inf>X<inf>h</inf></inf>I<sup>(k)</sup> = 0</formula> is
fulfilled.
 ∎</p>
<p><!--Remark--> 
 <formula>W</formula> is not unique, but <formula>I<sup>(k)</sup></formula> doesn't depend
on choosing representative from the class <formula>W<sup>∗</sup></formula>.
</p>
<p><!--Remark--> 
Theorem is also valid for generators <formula>E</formula> satisfying
<formula> [E , X<inf>h</inf><sup>∗</sup>] = X<inf>f</inf><sup>∗</sup></formula>
</p>
<p><emph>Example.</emph> 
Hamiltonian description of the relativistic particle leads to the following action
<dformula>
A = ∫ p<inf>0</inf>dx<inf>0</inf> + 
<subform><roman>∑</roman><bottom>s</bottom></subform>p<inf>s</inf>dx<inf>s</inf>
</dformula>
where
<formula>p<inf>0</inf> = (p<sup>2</sup> + m<sup>2</sup>)<sup>1/2</sup></formula>
with vanishing canonical Hamiltonian and degenerate 2-form defined by
<dformula>
p<inf>0</inf>ω = <subform><roman>∑</roman><bottom>s</bottom></subform>(p<inf>s</inf>dp<inf>s</inf> ∧ dx<inf>0</inf> + p<inf>0</inf>dp<inf>s</inf> ∧ dx<inf>s</inf>).
</dformula>
<formula>ω</formula> possesses the "null vector field"
<formula>i<inf>u</inf>ω = 0</formula>
<dformula>
u = p<inf>0</inf><fraction><num>∂</num><den>∂x<inf>0</inf></den></fraction> + <subform><roman>∑</roman><bottom>s</bottom></subform>p<inf>s</inf><fraction><num>∂</num><den>∂x<inf>s</inf></den></fraction>.
</dformula>
One can check that the following non- Hamiltonian vector field
<dformula>
E =p<inf>0</inf>x<inf>0</inf><fraction><num>∂</num><den>∂x<inf>0</inf></den></fraction>
+ p<inf>1</inf>x<inf>1</inf><fraction><num>∂</num><den>∂x<inf>1</inf></den></fraction> + ⋯ + p<inf>n</inf>x<inf>n</inf><fraction><num>∂</num><den>∂x<inf>n</inf></den></fraction>
</dformula>
generates non-Noether symmetry. Indeed, <formula>E</formula> satisfies
<formula>[E , X<inf>h</inf><sup>∗</sup>] = 0<sup>∗</sup></formula> because of
<formula>X<inf>h</inf><sup>∗</sup> = 0<sup>∗</sup></formula> and <formula>[E,u] = u</formula>.
Corresponding integrals of motion are combinations of momenta:
<dformula>
I<sup>(1)</sup> = <subform><roman>∑</roman><bottom>s</bottom></subform>p<inf>s</inf><break/>
I<sup>(2)</sup> = <subform><roman>∑</roman><bottom>r &gt; s</bottom></subform> p<inf>r</inf>p<inf>s</inf><break/>
 ⋯ <break/>
I<sup>(n)</sup> = <subform><roman>∏</roman><bottom>s</bottom></subform>p<inf>s</inf>
</dformula>
This example shows that the set of conserved quantities can be obtained from a single
one-parameter group of non-Noether transformations.
</p>
</section>
</chapter>
<back>
<ack>
<p><emph>Acknowledgements.</emph> 
Author is grateful to Z. Giunashvili and M. Maziashvili for
constructive discussions and particularly grateful to George Jorjadze for invaluable help.
This work was supported by INTAS (00-00561)
and Scholarship from World Federation of Scientists.
</p>
</ack>
<biblist>
<head>References</head>
<citation id="r1">
	<title>A new conservation law constructed without using either Lagrangians or Hamiltonians</title>
	<author>
		<fname>S.</fname>
		<surname>Hojman</surname>
	</author>
	<othinfo>J. Phys. A: Math. Gen. 25 L291-295</othinfo>
	<date>1992</date>
</citation>
<citation id="r2">
	<title>Geometric foundations of a new conservation law discovered by Hojman</title>
	<author>
		<fname>F.</fname>
		<surname>González-Gascón</surname>
	</author>
	<othinfo>J. Phys. A: Math. Gen. 27 L59-60</othinfo>
	<date>1994</date>
</citation>
<citation id="r3">
	<title>Remarks on a recent theorem about conserved quantities</title>
	<author>
		<fname>M.</fname>
		<surname>Lutzky</surname>
	</author>
	<othinfo>J. Phys. A: Math. Gen. 28 L637-638</othinfo>
	<date>1995</date>
</citation>
<citation id="r4">
	<title>New derivation of a conserved quantity for Lagrangian systems</title>
	<author>
		<fname>M.</fname>
		<surname>Lutzky</surname>
	</author>
	<othinfo>J. Phys. A: Math. Gen. 15 L721-722</othinfo>
	<date>1998</date>
</citation>
<citation id="r5">
	<title>Geometric Quantization</title>
	<author>
		<fname>N.M.J.</fname>
		<surname>Woodhouse</surname>
	</author>
	<othinfo>Claredon, Oxford</othinfo>
	<date>1992.</date>
</citation>
<citation id="r6">
	<title>Bi-Hamiltonian structure as a shadow of non-Noether symmetry</title>
	<author>
		<fname>G.</fname>
		<surname>Chavchanidze</surname>
	</author>
	<othinfo>math-ph/0106018</othinfo>
	<date>2001</date>
</citation>
</biblist>
</back>
</article>