PDA

View Full Version : What is Complete Integrability?


John Baez
Aug18-04, 04:08 AM
<jabberwocky><div class="vbmenu_control"><a href="jabberwocky:;" onClick="newWindow=window.open('','usenetCode','toolbar=no, location=no,scrollbars=yes,resizable=yes,status=no ,width=650,height=400'); newWindow.document.write('<HTML><HEAD><TITLE>Usenet ASCII</TITLE></HEAD><BODY topmargin=0 leftmargin=0 BGCOLOR=#F1F1F1><table border=0 width=625><td bgcolor=midnightblue><font color=#F1F1F1>This Usenet message\'s original ASCII form: </font></td></tr><tr><td width=449><br><br><font face=courier><UL><PRE>\nIn article &lt;40e7fa99\\$1@news.sentex.net&gt;, &lt;tessel@tum.bot&gt; wrote:\n\n&gt;You raised another question to which I have yet to either figure out the\n&gt;answer on my own, or else to find answered in some book.\n&gt;\n&gt;Namely: what is "completely integrable" all about, and why oh why are\n&gt;there -two- distinct concepts known by the same name in one field?!\n\nAh, if there were only just *two*... wouldn\'t that make things\nlovely and simple?\n\nDream on!\n\nSounds like you need to look at this book:\n\nV. E. Zakharov, ed., What is Integrability?, Springer Verlag, Berlin, 1991.\n\nBut, don\'t expect *one* answer to the question! Expect *lots*.\n\nThe point is this: once upon a time a physics problem was said\nto be "completely integrable" if you could reduce the problem of\nsolving it down to the problem of computing a bunch of integrals.\n\nOther words for this notion include "reducible to quadratures"\nand - even vaguer - "exactly solvable".\n\nPeople invented lots of tricks for showing that problems were "completely\nintegrable", and a physical system susceptible to any of these tricks\ngot lumped in the class of completely integrable systems. You mention\ntwo tricks. There are a bunch more.\n\nThey also invented criteria for *guessing* that problems were completely\nintegrable without actually proving it - most notably the "Painleve test".\n\nSorting out the logical relations among all these tricks and criteria\nis a bit like cleaning up the Augean stables, the only difference being\nthat there is gold to be found here, not just horse ****.\n\nBut, it\'s a big mess.\n\nI\'ve seen a big diagram of different concepts in this subject\nand arrows relating them (unfortunately not morphisms in any\nwell-defined category), which takes up more than one page, and\nmainly serves to prove nobody knows how everything fits together!\n\nI wish I could find it on the web, just to intimidate you.\n\n&gt;So far, I have seen two definitions which certainly don\'t look equivalent,\n&gt;at least not to me, at least not yet, at least not without thought.\n\nI\'ve never seen anyone claim they\'re *equivalent*.\n\n&gt;The first definition goes by the long name "completely integrable in the\n&gt;sense of Liouville", and applies to a n-dimensional Hamiltonian system.\n&gt;If we can find 2n independent conserved integrals which are pairwise\n&gt;involutory (vanishing Poisson bracket), this system is completely\n&gt;integrable. (I think need some more conditions on these integrals, like\n&gt;analyticity.) This notion is at least recognizably a special case of\n&gt;"integrable system".\n\nRight, and there\'s a wonderful general theory of these, related\nto toric varieties, KAM theory, and other good things... so this\nis *my* favorite precise definition of complete integrability.\n\n(Analyticity is only necessary for certain things, like hooking\nup to algebraic geometry.)\n\n&gt;The second usage (I hesitate to say "definition") seems silly. Let me\n&gt;give it a long name too: a system is "completely integrable in the sense\n&gt;of Lax" if it is solvable by the IST.\n\nWhere "IST" means "inverse scattering transform".\n\nI think this *implies* complete integrability in the first\nsense, at least under some conditions. I can\'t imagine any\nway to get the converse.\n\nI wish I knew more about the Lax pair/inverse scattering transform/\nLie-Baecklund stuff, especially about how it relates to loop\ngroups and affine Lie algebras. Someone like Graeme Segal or\nHitchin wrote a book about this, which James Dolan was carrying\naround for a long time with little discernible effect. Whoever\nwrote it, it was some mathematician whom you\'d expect to be good\nat getting to the bottom of things!\n\n</UL></PRE></font></td></tr></table></BODY><HTML>');"> <IMG SRC=/images/buttons/ip.gif BORDER=0 ALIGN=CENTER ALT="View this Usenet post in original ASCII form">&nbsp;&nbsp;View this Usenet post in original ASCII form </a></div><P></jabberwocky>In article <40e7fa99$1@news.sentex.net>, <tessel@tum.bot> wrote:

>You raised another question to which I have yet to either figure out the
>answer on my own, or else to find answered in some book.
>
>Namely: what is "completely integrable" all about, and why oh why are
>there -two- distinct concepts known by the same name in one field?!

Ah, if there were only just *two*... wouldn't that make things
lovely and simple?

Dream on!

Sounds like you need to look at this book:

V. E. Zakharov, ed., What is Integrability?, Springer Verlag, Berlin, 1991.

But, don't expect *one* answer to the question! Expect *lots*.

The point is this: once upon a time a physics problem was said
to be "completely integrable" if you could reduce the problem of
solving it down to the problem of computing a bunch of integrals.

Other words for this notion include "reducible to quadratures"
and - even vaguer - "exactly solvable".

People invented lots of tricks for showing that problems were "completely
integrable", and a physical system susceptible to any of these tricks
got lumped in the class of completely integrable systems. You mention
two tricks. There are a bunch more.

They also invented criteria for *guessing* that problems were completely
integrable without actually proving it - most notably the "Painleve test".

Sorting out the logical relations among all these tricks and criteria
is a bit like cleaning up the Augean stables, the only difference being
that there is gold to be found here, not just horse ****.

But, it's a big mess.

I've seen a big diagram of different concepts in this subject
and arrows relating them (unfortunately not morphisms in any
well-defined category), which takes up more than one page, and
mainly serves to prove nobody knows how everything fits together!

I wish I could find it on the web, just to intimidate you.

>So far, I have seen two definitions which certainly don't look equivalent,
>at least not to me, at least not yet, at least not without thought.

I've never seen anyone claim they're *equivalent*.

>The first definition goes by the long name "completely integrable in the
>sense of Liouville", and applies to a n-dimensional Hamiltonian system.
>If we can find 2n independent conserved integrals which are pairwise
>involutory (vanishing Poisson bracket), this system is completely
>integrable. (I think need some more conditions on these integrals, like
>analyticity.) This notion is at least recognizably a special case of
>"integrable system".

Right, and there's a wonderful general theory of these, related
to toric varieties, KAM theory, and other good things... so this
is *my* favorite precise definition of complete integrability.

(Analyticity is only necessary for certain things, like hooking
up to algebraic geometry.)

>The second usage (I hesitate to say "definition") seems silly. Let me
>give it a long name too: a system is "completely integrable in the sense
>of Lax" if it is solvable by the IST.

Where "IST" means "inverse scattering transform".

I think this *implies* complete integrability in the first
sense, at least under some conditions. I can't imagine any
way to get the converse.

I wish I knew more about the Lax pair/inverse scattering transform/
Lie-Baecklund stuff, especially about how it relates to loop
groups and affine Lie algebras. Someone like Graeme Segal or
Hitchin wrote a book about this, which James Dolan was carrying
around for a long time with little discernible effect. Whoever
wrote it, it was some mathematician whom you'd expect to be good
at getting to the bottom of things!

Thomas Larsson
Aug19-04, 12:36 PM
<jabberwocky><div class="vbmenu_control"><a href="jabberwocky:;" onClick="newWindow=window.open('','usenetCode','toolbar=no, location=no,scrollbars=yes,resizable=yes,status=no ,width=650,height=400'); newWindow.document.write('<HTML><HEAD><TITLE>Usenet ASCII</TITLE></HEAD><BODY topmargin=0 leftmargin=0 BGCOLOR=#F1F1F1><table border=0 width=625><td bgcolor=midnightblue><font color=#F1F1F1>This Usenet message\'s original ASCII form: </font></td></tr><tr><td width=449><br><br><font face=courier><UL><PRE>baez@galaxy.ucr.edu (John Baez) wrote in message news:&lt;cfu08g\\$2qp\\$1@glue.ucr.edu&gt;...\n&gt; In article &lt;40e7fa99\\$1@news.sentex.net&gt;, &lt;tessel@tum.bot&gt; wrote:\n&gt;\n&gt; &gt;You raised another question to which I have yet to either figure out the\n&gt; &gt;answer on my own, or else to find answered in some book.\n&gt; &gt;\n&gt; &gt;Namely: what is "completely integrable" all about, and why oh why are\n&gt; &gt;there -two- distinct concepts known by the same name in one field?!\n&gt;\n&gt; Ah, if there were only just *two*... wouldn\'t that make things\n&gt; lovely and simple?\n\n&gt; &gt;The first definition goes by the long name "completely integrable in the\n&gt; &gt;sense of Liouville", and applies to a n-dimensional Hamiltonian system.\n\n&gt; &gt;The second usage (I hesitate to say "definition") seems silly. Let me\n&gt; &gt;give it a long name too: a system is "completely integrable in the sense\n&gt; &gt;of Lax" if it is solvable by the IST.\n&gt;\n&gt; Where "IST" means "inverse scattering transform".\n&gt;\n\n\nHere are two other, partly related, uses of the word\nintegrability:\n\nLattice integrability, which is more or less equivalent\nto solutions of the Yang-Baxter equation (in 2D) or\nZamolodchikov\'s tetrahedron equation (in 3D). In an\nintegrable lattice model the free energy can be computed\nexactly. This kind of integrability is closely related\nto quantum groups and Bethe\'s ansatz.\n\nIntegrability in the sense of the Kyoto school, pioneered\nby Sato and his students (Jimbo, Miwa, ...). This\napproach involves pseudo-differential operators\n(Laurent polynomials in d/dx) and Hirota\'s bilinear\nequations.\n\nSome of this stuff I once understood, but one problem\nwith increasing age is that the brain sees fit to purge\nunused knowledge.\n\n</UL></PRE></font></td></tr></table></BODY><HTML>');"> <IMG SRC=/images/buttons/ip.gif BORDER=0 ALIGN=CENTER ALT="View this Usenet post in original ASCII form">&nbsp;&nbsp;View this Usenet post in original ASCII form </a></div><P></jabberwocky>baez@galaxy.ucr.edu (John Baez) wrote in message news:<cfu08g$2qp$1@glue.ucr.edu>...
> In article <40e7fa99$1@news.sentex.net>, <tessel@tum.bot> wrote:
>
> >You raised another question to which I have yet to either figure out the
> >answer on my own, or else to find answered in some book.
> >
> >Namely: what is "completely integrable" all about, and why oh why are
> >there -two- distinct concepts known by the same name in one field?!
>
> Ah, if there were only just *two*... wouldn't that make things
> lovely and simple?

> >The first definition goes by the long name "completely integrable in the
> >sense of Liouville", and applies to a n-dimensional Hamiltonian system.

> >The second usage (I hesitate to say "definition") seems silly. Let me
> >give it a long name too: a system is "completely integrable in the sense
> >of Lax" if it is solvable by the IST.
>
> Where "IST" means "inverse scattering transform".
>


Here are two other, partly related, uses of the word
integrability:

Lattice integrability, which is more or less equivalent
to solutions of the Yang-Baxter equation (in 2D) or
Zamolodchikov's tetrahedron equation (in 3D). In an
integrable lattice model the free energy can be computed
exactly. This kind of integrability is closely related
to quantum groups and Bethe's ansatz.

Integrability in the sense of the Kyoto school, pioneered
by Sato and his students (Jimbo, Miwa, ...). This
approach involves pseudo-differential operators
(Laurent polynomials in d/dx) and Hirota's bilinear
equations.

Some of this stuff I once understood, but one problem
with increasing age is that the brain sees fit to purge
unused knowledge.

John Baez
Aug24-04, 07:28 AM
<jabberwocky><div class="vbmenu_control"><a href="jabberwocky:;" onClick="newWindow=window.open('','usenetCode','toolbar=no, location=no,scrollbars=yes,resizable=yes,status=no ,width=650,height=400'); newWindow.document.write('<HTML><HEAD><TITLE>Usenet ASCII</TITLE></HEAD><BODY topmargin=0 leftmargin=0 BGCOLOR=#F1F1F1><table border=0 width=625><td bgcolor=midnightblue><font color=#F1F1F1>This Usenet message\'s original ASCII form: </font></td></tr><tr><td width=449><br><br><font face=courier><UL><PRE>\nIn article &lt;cfu08g\\$2qp\\$1@glue.ucr.edu&gt;,\nJohn Baez &lt;baez@galaxy.ucr.edu&gt; wrote:\n\n&gt;In article &lt;40e7fa99\\$1@news.sentex.net&gt;,\n&gt;&lt;tessel@tum.bo t&gt; wrote:\n\n&gt;&gt;So far, I have seen two definitions which certainly don\'t look equivalent,\n&gt;&gt;at least not to me, at least not yet, at least not without thought.\n\n&gt;I\'ve never seen anyone claim they\'re *equivalent*.\n\nWell, now I have!\n\nThis weekend, in a local bookstore called Heffer\'s,\nI saw a Cambridge University Press book that looks really good:\n\nIntroduction to Classical Integrable Systems\nOlivier Babelon, Denis Bernard, Michel Talon\n\nIn it, they start out by showing how any system that\'s\ncompletely integrable in this sense:\n\n&gt;&gt;The first definition goes by the long name "completely integrable in the\n&gt;&gt;sense of Liouville", and applies to a n-dimensional Hamiltonian system.\n&gt;&gt;If we can find 2n independent conserved integrals which are pairwise\n&gt;&gt;involutory (vanishing Poisson bracket), this system is completely\n&gt;&gt;integrable.\n\n.... comes from a Lax pair in a clever tautologous sort of way.\n\nThe converse is even easier.\n\nI liked this book since it seemed intent on demystifying things.\nUnfortunately, it was too expensive to buy - about 70 pounds, or\nabout 120 US dollars.\n\n\n</UL></PRE></font></td></tr></table></BODY><HTML>');"> <IMG SRC=/images/buttons/ip.gif BORDER=0 ALIGN=CENTER ALT="View this Usenet post in original ASCII form">&nbsp;&nbsp;View this Usenet post in original ASCII form </a></div><P></jabberwocky>In article <cfu08g$2qp$1@glue.ucr.edu>,
John Baez <baez@galaxy.ucr.edu> wrote:

>In article <40e7fa99$1@news.sentex.net>,
><tessel@tum.bot> wrote:

>>So far, I have seen two definitions which certainly don't look equivalent,
>>at least not to me, at least not yet, at least not without thought.

>I've never seen anyone claim they're *equivalent*.

Well, now I have!

This weekend, in a local bookstore called Heffer's,
I saw a Cambridge University Press book that looks really good:

Introduction to Classical Integrable Systems
Olivier Babelon, Denis Bernard, Michel Talon

In it, they start out by showing how any system that's
completely integrable in this sense:

>>The first definition goes by the long name "completely integrable in the
>>sense of Liouville", and applies to a n-dimensional Hamiltonian system.
>>If we can find 2n independent conserved integrals which are pairwise
>>involutory (vanishing Poisson bracket), this system is completely
>>integrable.

.... comes from a Lax pair in a clever tautologous sort of way.

The converse is even easier.

I liked this book since it seemed intent on demystifying things.
Unfortunately, it was too expensive to buy - about 70 pounds, or
about 120 US dollars.