You are on page 1of 26

Oscar II's Prize Competition and the Error in Poincaré's Memoir on the Three Body Problem

Author(s): June Barrow-Green


Source: Archive for History of Exact Sciences, Vol. 48, No. 2 (June 1994), pp. 107-131
Published by: Springer
Stable URL: http://www.jstor.org/stable/41133989 .
Accessed: 15/06/2014 10:58

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp

.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.

Springer is collaborating with JSTOR to digitize, preserve and extend access to Archive for History of Exact
Sciences.

http://www.jstor.org

This content downloaded from 91.229.229.49 on Sun, 15 Jun 2014 10:58:35 AM


All use subject to JSTOR Terms and Conditions
Oscar IF s Prize Competition
and the Error in Poincaré's Memoir
on the ThreeBody Problem
June Barrow-Green
by Jesper Lützen
Communicated

Introduction

In theautumnof1890HenriPoincaré'smemoiron thethreebodyproblem
[1] was publishedin thejournalActa Mathematica as the winningentryin
theinternationalprizecompetition sponsored by Oscar II, KingofSwedenand
Norway, to mark his 60th on
birthday January21, 1889.Today Poincaré's
publishedmemoiris renownedboth for providingthe foundations for his
celebratedthree-volume Méthodes de la MécaniqueCéleste[2] and for
Nouvelles
containingthefirstmathematical ofchaoticbehaviorin a dynamical
description
system.
A combination of royalpatronageand carefully plannedpublicrelations
meantthatthecompetition achievedtheunusualdistinction ofgainingrecogni-
tionthatstretched well beyondthe worldof mathematics. However,despite
appearancesto the contrary, correspondence preservedat the InstitutMit-
revealsthatthe competition
tag-Leffler was in factbeleaguered by difficulties
throughout. In particular,it has emergedthat onlyweeksbeforethe prize-
winning memoirwas due to be publishedPoincarêdiscovered an errorin his
workwhichwas of suchgraveconsequencethathe was forcedto makevery
substantialchanges.Indeedit was onlyas a resultof correcting theerrorthat
he discoveredtheexistence of whattodayare knownas homoclinic points.As
a resultthememoir whicheventually appeared in Actawas remarkablydifferent
fromthe one whichhad actuallywon the prizealmosttwo yearsearlier.
The following is an accountof the troubledhistoryof the competition
together withan explanationof the errorin Poincaré'smemoir.

The competition

By thelate 19th mathematical


century, werea wellestab-
prizecompetitions
lishedmethodforseekingsolutionsto specificmathematical
problems.These
competitionsusuallyemanatedfromthenationalAcademies,
notablyBerlinand
Paris,and althoughtheprizesoffered
weregenerallyfinancial
in nature,they

This content downloaded from 91.229.229.49 on Sun, 15 Jun 2014 10:58:35 AM


All use subject to JSTOR Terms and Conditions
108 J.Barrow-Green

werevaluedfarmorein termsofacademicprestige. Nevertheless,althoughthe


existenceof a mathematical competition at this time was no novelty,this
particularone was somewhat unusual in thatits sponsor,King Oscar II, did
he chose ratherto linkit to an academic
not associateit withan institution;
journal,albeitone in whichhe had a personalinterest.ThatKingOscar should
have sponsoredsucha competition was not so surprising sincehe had himself
studiedmathematics and had thenbecomean activepatronof
at theuniversity
the subject,providingfinancialsupportforvariouspublishing in-
enterprises,
of
cludingthefounding Acta1, as well as making awards to individualmath-
ematicians.

of thecompetition
Organisation

From its beginnings in 1884,the competition was organisedby Gösta


Mittag-Leffler, then of
professor pure mathematics at thenewlyestablished
StockholmHögskola (later the University of Stockholm)and founderand
of Acta. Mittag-Lefflerhad obtainedhis doctoratefromthe
editor-in-chief
university Uppsalain 1872,and laterhad studiedunderCharlesHermitein
of
Paris, Ernst Scheringin Göttingenand Karl Weierstrassin Berlin.He
thereforehad first-hand experienceof lifewithinthe premiermathematical
communities in Europe and this,togetherwithhis involvement withActa,
meantthathe was wellplacedto promotetheidea ofan international competi-
tion.
Apart frombeing a talentedmathematician, Mittag-Lefflerwas also
a skilledcommunicator. He assiduously and
cultivated nurtured mathematical
contactsat home and abroad, both travelling and
extensively maintaining
a vigorouscorrespondence2. He had alreadyestablisheda good relationship
withKingOscar through thefounding ofActa and it is notclearwhetherthe
idea of holdingthecompetition camefromMittag-Leffler or fromtheKing
himself.However,sinceMittag-Lefflerenthusiastically embracedany oppor-
tunityto raisetheprofileofScandinavian mathematicsor indeedto enhancehis
own reputation withinthe mathematical community, it seemslikelythatthe
competition emergedas a consequenceof his initiative.
The firstknownreference to the competition occursin a letterfromMit-
tag-Lefflerto the Russianmathemetician Sonya Kovalevskayawrittenin
June1884,whosecontents showthatthetopicwas alreadyunderdiscussion. As
thefollowing extractoutlining the
its proposedformmakesclear, competition
was intendedto be one of pre-eminent importance in thefieldofmathematical
analysisand was originally intendedas a regularevent3:

1 The firstissue of Acta Mathematicaappeared at the end of 1882. See Domar [3].
2 Mittag-Leffler's considerable correspondence is preserved at the Institut
Mittag-Leffler and is describedin Grattan-Guinness [4].
3 Extract froma letterdated 7.6.1884,InstitutMittag-JLemer (tr. irom aweaisn oy
S. NORGAARD).

This content downloaded from 91.229.229.49 on Sun, 15 Jun 2014 10:58:35 AM


All use subject to JSTOR Terms and Conditions
Poincaréon theThreeBodyProblem 109

". . . / agree withWeierstrass thatifnoneof the answerson the set question


are worthyof theprize,thenthemedalmustbe awardedto themathematician
who withinrecentyears has made the best discoveriesin higheranalysis.
. . . we should not award our prize morefrequentlythan everyfourthyear.
Malmsten and theKing wanttheprizejury to be appointedby the King and
to consistof
1. The main editorof Acta Mathematica
2. A Germanor Austrianmathematician - = Weierstrass
3. - = Hermite
A Frenchor Belgian mathematician
4. An Englishor Americanmathematician - = Cayley? or Sylvester
5. A Russian or Italian mathematician- - the first time Brioschi or
Tschebychev,the second timeMrs Kovalevskaya.
Aftereach prize-givingtwo of theprizejudges shouldleave thejury and new
ones shouldbe appointedby King Oscar as long as he is alive - he mustbe
able to appoint [replacements']for both the leaving members.AfterKing
Oscar's death,the threeremaining mustappointtwo new membersbut always
in such a way as to fit the categoriesmentionedabove. ..."

In the event,Mittag-Leffler was unable to fulfilany one of King Oscar's


requirementsexactly.The difficulties
he faced are well illustratedby Kovalevs-
kaya's replywrittenwhile on holiday in Berlin4:

". . . In regardto the questionof theprize Weierstrass has promisedme that


he will writeyou his opinionon that in moredetail as soon as he receives
a letterfromyou. I did not informhimof what you wroteme in the letter
beforelast withregardto the choice ofjury,for I was sure in advance of his
completedisapproval.IndeedI believethatin thisway thethingpresentsmany
Just consider how one could hope that four famous
practical difficulties.
mathematicians, Weierstrass,Hermite,Cayley and Tschebychevwouldever
agree on the meritsof a memoir.I believe it is certain that each of the
four wouldrefuseto becomepart of thejury as soon as he learnedthe names
of the otherthree.As for Weierstrass, / am so sure of this that I didn't
even ventureto talk to him about it. In generalWeierstrass thinksthat it
will be quite difficultfor thejury to agree when theyhave no opportunity
to talk face to face. To do it by mail is considerablymore difficult; and
at bottom,whywouldtheseold gentlemen take so muchtroublefor us? There,
I fear, is a very great difficulty! As for the honour,quite the contrary,
each of thefour that you namedwill be outragedthat you chose the others
along withhim...."

4 Extract from
an undated letter,InstitutMittag-Leffler (tr. from French by R.
COOKE). The letteris reproducedin full(in translation)in The Mathematicsof Sonya
Kovalevskaya,R. COOKE, 1984, Springer-Verlag, New York.

This content downloaded from 91.229.229.49 on Sun, 15 Jun 2014 10:58:35 AM


All use subject to JSTOR Terms and Conditions
110 J.Barrow-Green

Whiletherewas an elementof melodramain Kovalevskaya'sletter(Hermite


and Weierstrasscertainly had a healthyrespectforeach other),forthemost
partherpresentiments provedto be wellfounded.The eventualoutcomewas
a commission comprisedof only three:Hermite,Weierstrass,and Mittag-
Leffler himself5.AlthoughMittag-Lefflerhad abandonedthe proposal
for a commissionof five members,he had managed to engage two of
the leadinganalystsof the day, one fromeach of the premiermathematical
nations,and, importantly, two mathematicians withwhom he had already
establishedwarmand productive friendships6. Nevertheless, despitethe reduc-
tion in the numberof people involved,the commissiondid still present
Apartfromtheobviousproblemsarisingfromthe
certainpracticaldifficulties.
differentgeographicallocations involved,Berlin,Paris and Stockholm, there
was an additionalcomplicationengendered by the lack of a common first
language.
Withregardto thecompetition King Oscar was keenthatit should
itself,
consistof a singlequestion,but his view was not sharedby thecommission
whofeltthata singlequestionwould both be too restrictiveand precludework
ofan innovatory nature.After an intensive correspondenceformat
a was finally
agreed:the competition would consistof fourquestionstogetherwiththe
opportunity to submitan entryon a self-selected topicbutwithpriority being
given to entriesattempting one of the proposed questions.
In mid-1885theofficial announcement ofthecompetition was publishedin
both Germanand Frenchin Acta,19as well as in severalotherlanguages
elsewhere.It gave detailsof the prize(a gold medal together witha sum of
2,500Crowns8), named the commission, listed the questions and stipulatedthe
conditionsof entry.
AlthoughMittag-Lefflerhad originallyproposed that Hermite and
Weierstrassshouldeach settwo questions, it seemsthatof thefourquestions
eventuallyset,the three
first were proposedby Weierstrassand thefourth by
Hermite9. The firstone addressed the well-known n-bodyproblem, reflecting

5 Part of the Weierstrass to Mittag-Leffler correspondenceis publishedin [5];


the Hermite to Mittag-Leffler correspondence1884-1891 is published in [6].
6 Not me Kings
only did the composition of the commission not accord witñ
but also the idea of making the competition a regular event was
original conception,
nevertaken any further. That the competitionwas held only once was probablydue to
in organisinga commissionand to the considerableproblems
the original difficulties
which it later encountered.
7 Acta
7, p. I-VI;
7a Nature, 30
July 1885 and nine others listed in Acta 10, p. 382.
8 For in 188¿ as
comparison: Domar [3] cites Mittag-Leffler's annual salary
as
a professorin Stockholm 7,000 Crowns, and Nature (February 21, 1889, 396), in the
announcementof the competitionresult,states that at that time it was equivalent to
£160. .. .
9 Letters from Mittag-Leffler to Hermite, 20.2.1885 and 0.3.I8W, insumí
Mittag-Leffler Nos. 356 and 374.

This content downloaded from 91.229.229.49 on Sun, 15 Jun 2014 10:58:35 AM


All use subject to JSTOR Terms and Conditions
Poincaréon theThreeBodyProblem 111

Weierstrass' longstandinginterestin the problem10;the second requireda de-


tailed analysis of Fuchs' theoryof differential equations; the third asked for
further investigationinto the firstorder non-linear
differential
equations studied
by Briot and Bouquet; and the last concernedthe of
study algebraicrelations
connecting Poincaré's Fuchsian functionswhich have the same automorphism
group.
The entrieswere to be sent to the chiefeditorof Acta beforeJune 1, 1888,
and, as was customaryin such competitions,theywere to be sent in anony-
mously,identifiableonly by a motto and accompanied by a sealed envelope
bearingthe motto and containingthe author's name and address. The entries
were not to have been previouslypublished and notice was given that the
winningentrywould be publishedin Acta.

Kronecker
's criticism

Unfortunately, the announcementdid not meet with universalapproval. It


provokedan angryreactionfromanotherprofessorat the Universityin Berlin:
Leopold Kronecker. Kronecker, apparentlyincensedat being leftout of the
commission,wrote to Mittag-Leffler with a catalogue of complaints11.But
since it was no secret that an intense rivalryexisted between himselfand
Weierstrass,it is likelythat he was more angryabout Weierstrass' inclusion
than he was about his own exclusion12.
Kronecker accused Mittag-Leffler of using the competitionas a vehicle
foradvertisingthe Acta. Why had the competitionnot been proposed by the
Swedish Academy? It was an accusation Mittag-Leffler could easily refute:
the King wished the competitionto be announced in Acta, not only because
Acta could claim a widermathematicalreadershipthan the transactionsof the
SwedishAcademy,but also because of his personal interestin thejournal. On
being challenged on the choice of members for the commission,Mittag-
Leffler explained to Kronecker that his instructionshad been to choose
a commissionof three,consistingof a representativefromeach of the
top
mathematicalnations,Germanyand France,witha thirdmemberfromSweden.

10 In a letterdated
Weierstrasstold Kovalevskaya thathe had con-
15.8.1878,
structeda formal seriesexpansionforsolutionsto theproblembutwas unableto prove
convergence and in 1880/81 he gavea seminaron theproblemsofperturbationtheoryin
astronomy. DespiteWeierstrass' own difficulties withthe problem,certainremarks
made by Dirichlet in 1858 had led him to believethata
completesolutionwas
possible,and hencehis choiceof the problemas one of the competition questions.
Weierstrass'interest in the problemis chronicled in [5].
The contents ofKronecker'sletterto Mittag-Lefflerhavebeenreconstructed
fromMittag-Leffler'sreplywritten in July1885,a copyof whichis at theInstitut
Mittag-Leffler.
Weierstrass believedthat Kronecker's avowed antipathyto the work of
Georg Cantor reflected Kronecker'sopposition to hisownwork.See Biermann[7].

This content downloaded from 91.229.229.49 on Sun, 15 Jun 2014 10:58:35 AM


All use subject to JSTOR Terms and Conditions
112 J.Barrow-Green

With regardto the Germanrepresentative, he told him that it had been


a straightchoicebetweenhimand Weierstrass, bothbeingequallysuitable, but
Weierstrass, being some 8 the
years elder, had been chosen on the groundsof
his "venerable"age. This may have mollified Kroneckerbut it is doubtful
whetherWeierstrasswould have been impressedby thisline of reasoning.
However,Kroneckerlevelledhis mostseriouschargeat Question4, theques-
tion set by Hermite.He maintainedthathe was the best personto judge
algebraicquestionsofthistypeand thathe had alreadyprovedthattheresults
requiredto resolvethisquestionwereimpossible to achieve,and he threatened
to tell the King as much13.As a defence, Mittag-Lefflercould onlyplead
ignorance on behalfofthecommission and concludedhis replywitha barrage
of flatterywell calculatedto appeal to Kronecker'svanity14.
Kronecker then let mattersrest,but indefinitely. In 1888 he launched
anotherassaultbut thistimeit was directedat thewordingof Question1. On
thisoccasionhe did not writeto Mittag-Lefflerbut insteadmade his com-
plaintat a meetingof the BerlinAcademyof Sciences[8]. Sincethistimehe
had chosento makehis viewspublicand theobjectof his censurewas one of
Weierstrass'questions, theattackappearedto be a further manifestation ofthe
rivalry between the two Germans as opposed to a critique of the commission.
Nevertheless, since Kroneckersteadfastly maintainedthathe did not know
who had composedthe question,it was difficult forthe commission to know
how bestto respondto him.Shouldtheydo so collectively and in thenameof
thecommission, or shouldWeierstrasspersonally take on the responsibility?
Hermitemade it quite plain that he did not wish to be involvedin the
dispute.Not onlydid he considerthematterto be an entirely Germanaffair
between the "two of
princes analysis", but also he considered it his patriotic
dutyto avoid doinganything that could be construed as having a national
connection. He was convincedthat Kronecker was a committed Francophobe
and so felttherewas nothingto be gainedby his intervention.
Weierstrassmadeit clearthatalthoughhe wouldhavehad no difficulty in
dealingwithKronecker'scomplaints he was reluctant to do so on his own since
he considered After
thetaska jointresponsibility. muchdeliberation thecommis-
siondecidedagainstan immediate response in the belief that it would be better
to waituntilthejudgingofthecompetition had been completed and the winning
It turnedout to be a wise decision. Not did
only subsequent
papers)published.
eventsovershadow theissuebuttheneedto replywas obviatedbyKronecker's
deathin 1891,shortly afterthe publication of the winning memoirs.

13 The 5' degree and it


question made referenceto solutions of equations of the
that this was what Kronecker objected to. See the letterfrom Mittag-
appears
Leffler to Hermite, 8.9.1885,InstitutMittag-Leffler No. 481.
14
Shortlyafterwards Hermite met Kronecker and told nun mat ne naa proposea
Question 4. He explainedhis in
intentions settingthe question, admittingthathe had set
satisfied
it specificallywith Poincarê in mind. Hermite's explanationseems to have
Kronecker as he did not pursue the issue. Perhaps it was that
sufficient Weierstrass
was not involved See [6, pp. 108-111].

This content downloaded from 91.229.229.49 on Sun, 15 Jun 2014 10:58:35 AM


All use subject to JSTOR Terms and Conditions
Poincaré on the Three Body Problem 113

The entriesin the competition

Despite the factthat the identityof entrantswas supposed to be secret,all


three membersof the commissionwere well aware that Poincaré meant to
enter.As early as July1887 Poincaré had made clear his intentionsto Mit-
tag-Leffler, explicitlymentioningQuestion I15. In October of the same year
Hermite told Mittag-Leffler that althoughhe knew Poincaré was working
on an entryfor the competition,he did not know whetherPoincaré would
submitit and, in any case, he was not sure whetherPoincaré was workingon
Question 1 or Question 4. Mittag-Leffler, still scarred from Kronecker's
original attack, admitted to Poincaré that he hoped he would provide an
answer to Question 4. In fact the selectionof topics for the competitionwas
such thatit would have been possibleforPoincaré to have workedon any one
of them.This raises the question: had theyall been chosen with Poincaré in
mind? Hermite freelyadmittedthat this was the case with his question and
perhaps Weierstrass too had designedhis questions to appeal particularlyto
Poincaré. CertainlyMittag-Leffler was an unquestionablechampionof Poin-
caré's work16.In the event,Poincaré chose to attemptQuestion 1, the most
difficultof the four.
By the closing date twelve entrieshad been received.Shortlyafterwards
a list of theirtitles,numberedin date order of submission,was published in
Acta withthe authorsidentifiedsolely by theirrespectivemottos17.Five of the
entrantshad attemptedQuestion 1, one had attemptedQuestion 3, and the
remainingsix had chosen theirown topics.
When PoiNCARÉ's entryarrivedit was clear that his readingof the regula-
tionshad been somewhatperfunctory. As requiredhe had inscribedhis memoir
with an epigraph18,but instead of enclosinga sealed envelope containinghis
name,he had writtenand signeda coveringletter,and had also senta personal
note to Mittag-Leffler. However,since he had already told Mittag-Leffler
and Hermite of his intentionto enter,and he knew that theywould recognise
his entryby its content- it was an explicitdevelopmentof his earlierwork on
equations - as well as by his handwriting,
differential it clearlywas not a deli-
berate attemptto floutthe procedures.

15 A selectionof
Poincaré's lettersto Mittag-Leffler concerningthe competition
are publishedin [91.
16 Mittag-Leffler
securedPoincaré's supportforthe launch of Acta, publishing
importantpapers by him in each of the firstfive volumes.
1 See Acta 11
(1888), 401-402. Apart from Poincaré's entryit has only been
possible to identify
positivelythreeof the othercontestants:Paul Appell (entryNo 8),
Guy De Longchamps (4) and Jean Ešcary nov
Nunquamprasscriptos transibuntsiderafines= Nothing exceeds the limitsof the
stars.

This content downloaded from 91.229.229.49 on Sun, 15 Jun 2014 10:58:35 AM


All use subject to JSTOR Terms and Conditions
114 J.Barrow-Green

Judgingthe entries

A large part of the judging was done by correspondence.Mittag-Leffler,


having receivedthe entriesin Stockholm,appointedone of the editorsof Acta,
Edvard Phragmén,to do the preliminary readingpriorto havingcopies of the
most significantentriesmade and sent to Hermite and Weierstrass. Within
a fortnightof the closing date Mittag-Leffler had decided that there were
only three entriesworth considering,two fromformerstudentsof Hermite
- PoiNCARÉand Appell - and one fromHeidelberg,19although none of the
threehad provided a complete solution to any of the given questions.
Mittag-Leffler spent August in Germanywith Weierstrass so that they
could studythe memoirstogether.The followingmonthhe wroteto Hermiteto
tell him that theythoughtthat Poincaré should win, withAppell being given
an honorablemention.He made the point that Poincaré had the advantagein
as much as he had at least attemptedone of the set questionswhereasAppell
had chosen his own topic: Poincaré had limitedhis investigations to a particu-
lar formof the three body problem(now known as the restrictedthreebody
problem) ratherthan the n body problem as specified in the question,while
Appell had considered the expansion of Abelian functions by trigonometric
series.MeanwhileHermite had also been studyingPoincaré's memoirand was
equally convinced of the importanceof the work.
The commissionhad quickly reached a unanimous decision but the hard
of
part of theirwork had not begun. It was one thingto recognisethe quality
Poincaré's work but quite anotherto understand it. Poincaré's entry was not
it
only extremelylong (when printedfor Acta it amounted to 158 pages) but
also contained many new ideas and results.Furthermore, as Hermite freely
admittedin a letterto Mittag-Leffler, the difficulties of comprehensionwere
compunded by Poincaré's customarylack of detail [6, p.147]20

". . . But it must be acknowledgedthat in this work as in almost all his


researches,Poincaré showstheway and givesthesigns,but leavesmuchto be
done to fill the gaps and completehis work.Picard has oftenasked himfor
enlightenment and explanationson veryimportant pointsin his articlesin the
ComptesRendus,withoutbeingable to obtain anythingexcept the response:
"it is so, it is like thať, so thathe seemslike a seer to whomtruthsappear in
a brightlight,but mostlyto him alone.

All three members of the commission struggledwith various parts of the


memoir,but it was Mittag-Leffler who, determinedthatthe versionsubmitted
to the King should be as complete as possible, enteredinto correspondence
with Poincaré - notwithstandingthe rules of the competition whereby
-
he should have been ignorant of the memoir's authorship appealing for

19 EntryNo 5.
20 Tr. JB-G.

This content downloaded from 91.229.229.49 on Sun, 15 Jun 2014 10:58:35 AM


All use subject to JSTOR Terms and Conditions
Poincaréon theThreeBodyProblem 115

clarificationon several issues. Poincaré responded by producinga series of


Notes which,when printed,added a further93 pages to the memoir.
Mtttag-Leffler may have had no qualms about his contactwithPoincaré
but Weierstrass was less happy about it and made a point of asking Mit-
tag-Leffler not to mention that he knew that Poincaré had entered the
competition.He told Mittag-Leffler thatit was almost an axiom in Germany
for prize papers to be published exactly in the form in which they were
submitted, althoughhe personallyconsideredthatthe propertimeforadditions
and correctionswas when the paper was beingeditedforpublication,providing
theywere clearlyacknowledged.
Having worked throughPoincaré's and Appell's memoirs,it now only
remainedfor the commissionto produce theirreports.Weierstrass had the
responsibilityfor writingthat on Poincaré's paper, Hermite the responsibility
forwritingthat on Appell's and Mittag-Leffler the responsibility forwriting
a general reporton the competitionas a whole.
On the 20thJanuary1889, the day beforethe King's 60thbirthday,Mit-
tag-Leffler wentto the palace (havingbelatedlyobtained fromPoincaré the
necessarysealed envelope) and the resultwas officiallyapproved. The general
report was published in the newspaper Posttidningenand Mittag-Leffler
wroteto Poincaré to tell him thathe would be receivingan officialcopy of the
report via the Swedish ambassador in Paris21. Almost everythinghad been
completedto the King's satisfaction,only Weierstrass' report on Poincaré's
memoirwas outstanding.Weierstrass,as Mittag-Leffler had feared,had been
too unwell to fulfilhis obligationwithinthe allotted time,although he gave
assurances that the reportwould soon be finished.

A prioritydispute?

Unfortunately for Mittag-Leffler, the publication of the general report


signalledthe startof a distressingpolemic betweenhimselfand the astronomer
Hugo Gyldén, a fellowlecturerat the StockholmHögskola and a memberof
the editorial board of Acta. Although the report only contained a
cursory
indicationof Poincaré's results,it was enough to convinceGyldén that he had
anticipatedPoincaré in a paper of his own [10] published some two years
earlier.
Mittag-Leffler appears to have had some idea of Gyldén's views almost
immediatelybecause only a matterof days afterthe reportwas published he
wrote to Poincaré to ask him forhis opinion on a resultin Gyldén's
paper.
This resultappeared to conflictwithsomethingthat Poincaré had written.The
particularpoint at issue concerned the convergenceof certain power series:
Gyldén had claimed that the serieswere definitely convergentwhile Poincaré

21 A Frenchtranslation
of the reportis publishedin Poincaré Œuvres 11
pp.286-289.

This content downloaded from 91.229.229.49 on Sun, 15 Jun 2014 10:58:35 AM


All use subject to JSTOR Terms and Conditions
116 J.Barrow-Green

had stated that the evidence for convergencewas inconclusive. Poincaré


responded swiftlyto Mittag-Leffler but he admitted that he had found
Gyldén's paper extremelyhard to read. In orderto give a definitive answerto
the convergencequestion he would have to make a much more detailed study
of it, a task he was reluctantto undertake.He was thereforeunable to say
whetherGyldén's method led to a proof of eitherconvergenceor divergence,
althoughhe believed divergencemore likely22.He was also unhappywith the
factthat Gyldén's methoddid not allow the successivetermsin the expansion
to be deduced recurrently but involved making choices at each stage of the
calculation, a featurewhich incorporatedan elementof chance into the process.
Meanwhile, Mittag-Leffler, at the King's request,had been due to give
a reviewof Poincaré's paper at the Februarymeetingof the SwedishAcademy
of Sciences.In the eventillnesspreventedhim fromattending,althoughhe had
expressedreluctanceto talk publiclyabout Poincaré's work withoutthe sup-
port of Weierstrass' report.Gyldén, on the otherhand, did attendthe meeting
and, moreover,did talk about Poincaré's memoir.He declared his own posi-
tion on Poincaré's resultsand effectively claimed priority.
Once again Mittag-Leffler was placed in an awkwardposition.The King
made it plain that he expected him to reply to Gyldén at the meeting
the followingmonth. Mittag-Leffler knew he could not rely on having
Weierstrass' report in time and so it became a matterof urgencyto have
detailed commentson Gyldén's paper fromPoincaré.
On hearingfromMittag-Leffler about Gyldén's position,Poincaré re-
sponded again and at length.He made the point that the disputebroughtinto
sharp focusthe difference betweenmathematiciansand astronomerswithregard
to theirinterpretation of convergence.He reasoned in detail against the rigour
of Gyldén's method,reiterating that he believedit to relyheavilyon questions
of judgement,and, in his final letteron the subject, showed clearly why he
believed that Gyldén's argumentactually led to divergentseries.
Despite the critical appearance of Poincaré's side of the correspondence,
Poincaré did in fact maintaina high regardfor Gyldén's work,appreciating
the flexibilityand practicaladvantagesof his methods.He had not intendedto
demolishGyldén but ratherto show how words such as proofand convergence
take on different meaningsdependingon whetherthe user is a mathematician
or an astronomer.Moreover,he was sensitiveto the fact that Gyldén's ap-
proach was coloured by a practicalinterestin the problemwhichhe himselfdid
not share.
Hermite and Weierstrass were also drawn into the polemic.Hermite,who
had firstheard about the disputefromKovalevskaya, thoughtGyldén's series
were asymptoticbut carefullyavoided drawinga directcomparisonbetweenthe
two memoirs.He had himselfreceiveda letterfromGyldén but since it was
writtenin Swedishhe had been unable to read it,althoughhe had deduced that
it concernedthe convergencequestion.

22 Poincaré's side of thecorrespondence


is publishedin [9].

This content downloaded from 91.229.229.49 on Sun, 15 Jun 2014 10:58:35 AM


All use subject to JSTOR Terms and Conditions
Poincaréon theThreeBodyProblem 117

Meanwhile,Mittag-Leffler gave his talk at the Academy,and wrote in


jubilation to Weierstrass and Poincaré, certain that those who had heard
him had been convinced that Poincaré deserved the prize. However Mit-
tag-Leffler's feelingof triumphwas shortlived. The academic communityin
Stockholmdecidedto weighin on the side of Gyldên, and, despitethe factthat
PoiNCARÉ'smemoir was not in the public domain, adopted the view that
Gyldén had indeed published proofs of everythingPoincaré had done. The
consensus was that Mittag-Leffler's denial of Gyldén's results had been
motivatedby jealousy; this idea was reinforcedby the mathematicianAlbert
Bäcklund who drew attentionto the fact that Gyldén's memoirhad recently
been awarded the St. Petersburgprize.
MeanwhileGyldén steadfastlymaintainedthat the values of the constants
forwhichhis seriesdivergedformedonlya countableset and so it was infinitely
unlikelythat the series was actually divergent.Mittag-Leffler continuedto
argue against him since,with Poincaré, he believedthat the serieswere diver-
gent not just fora countable set but fora perfectset in the neighbourhoodof
these constants.Moreover,he told Weierstrass that he thoughtGyldén not
enough of a mathematicianto understand.
With the publication of Poincaré's memoir not scheduled for several
months,the controversygraduallydied down. Nevertheless,when the memoir
finallyappeared,Gyldén did attemptto reopen the debate by writingdirectly
to Hermite. Possibly he thoughthe could count on Hermite's support since
Hermitewas knownto share his interestin the applicationsof ellipticfunction
theoryin celestialmechanics.But Hermite was not to be drawn. He stood by
thejudgementof the commission,declaringhis loyaltyto Mittag-Leffler and
Weierstrass.As he indicatedlaterto Mittag-Leffler [6, pp. 195-196], he was
not impressedby Gyldén's grasp of analysis,describingGyldên as a ghost
froma bygoneage, who had been leftbehind as the world of analysis trans-
formedabout him23.

Discoveryof the error

Mittag-Leffler's disputewithGyldên paled into insignificance


when com-
pared with the problemwhich subsequentlyemerged.Mittag-Leffler, having
allowed timeforediting,had hoped to have the volume of Acta containingthe
winningmemoirspublishedby October 1889. Apart fromWeierstrass' report,
forwhichhe had continuedto press,the actual printingwas completedby the

23 In the contextof Mittag-Leffler's


problemsover the competition,
it is of
interest
to recordthatin May 1889Gyldén metwithKronecker in Berlin,a meeting
whichMittag-Lefflerwouldsurelyhaveviewedwithsomemisgiving. In anycase,the
occasionprompted Mittag-Leffler to remarkto Weierstrassthatalthoughhe had
beenled to believethathis two adversaries
had understoodeach otherperfectly,
he
suspectedthat Gyldén reallyunderstoodas littleof Kronecker as Kronec'ker
understood of Gyldén.

This content downloaded from 91.229.229.49 on Sun, 15 Jun 2014 10:58:35 AM


All use subject to JSTOR Terms and Conditions
118 J.Barrow-Green

endofNovember. Butthevolumedid notappearuntiltheendofthefollowing


year and when it did,it did notcontaina replicaofthememoirPoincaréhad
submitted to the competition.
The firstglimmer thatanything was awryoccurred in July1889.Phragmén,
whowas editingPoincaré'smemoirforpublication, alertedMittag-Lefflerto
some passagesin it whichseemedto him a littleobscure.Thus prompted,
Mittag-Lefflerwroteto Poincaré foryetfurther clarification. Poincaré,in
the course of dealingwithPhragmén'squeries,realisedthat he had made
a seriouserrorin a different partof thepaper.At thebeginning of December,
he wroteto Mittag-Leffler.Makingno attempt to concealhisdistress, he told
himthat he had written to Phragmén to tellhim of the the
error, consequences
of whichweremorefar-reaching thanhe firstthought, and as a resulthe was
havingto make substantial changes to the memoir24.
This was most unwelcomenews for Mittag-Lefflerfor,althoughthe
volumeofActa had not beenpublished, a limitednumberofprintedcopiesof
the memoirhad been circulated. Once more Mittag-Leffler'scarefully nur-
turedmathematical reputation was in jeopardy.Despite his confidence in the
overallqualityof the memoir,he was only too conscious of the damagehe
personallymightsuffer shouldthescale of the error become public.Neverthe-
less,whiletotal secrecywas impossible, he knew that if he could securethe
returnof the printedcopies,then at least there would be no evidenceto
substantiateany rumourswhichmightcirculate. But while knewto whom
he
thecopieshad beensent,securing theirreturn wouldnotbe easysinceseveral
had beendispatched to foreign destinations. Apartfromthosesentto Hermite
and Weierstrass,the recipients includedKovalevskaya,Jordan,Von Dyck,
Gyldén, Lindstedt, Lie and Hill25.
Mittag-Leffleralso suggested to Poincaréthateverything concerning the
errorshouldbe keptbetweenthemselves, at leastuntilpublication of thenew
memoir.To safeguard himself stillfurther, he gave Poincaré detailedinstruc-
tionsaboutwhathe wantedin thecontents oftheintroduction to thereworked
memoirin orderto ensurethat no detailsof the errorwere included.In
addition,he also askedPoincaréto pay fortheprinting oftheoriginalversion
- a requestto whichPoincaréagreedwithoutdemur,despitethefactthatthe
billcameto just over3,500Crowns,whichwas some 1,000Crownsmorethan
the prizehe had won.
Phragmén'srole in settingPoincaré on the trailof an errorwhichhad
escapedtheattention oftheentirecommission was certainly worthy ofrecogni-
tion. However, and characteristically, Mittag-Leffler did not see it in his

24 The letteris given in full in the last section of this paper.


25 These names are listed by Phragmén m a note to MnrAu-jLti^LtK.
Mittag-Leffler appears to have been tirelessin his efforts
Despite the difficulties,
to retrievethese pre-publicationcopies of the memoir.In the libraryof the Institut
thereis one inscribedin Mittag-Leffler's hand with the phrase which
Mittag-Leffler
in Swedish reads "whole editiondestroyed".

This content downloaded from 91.229.229.49 on Sun, 15 Jun 2014 10:58:35 AM


All use subject to JSTOR Terms and Conditions
Poincaréon theThreeBodyProblem 119

intereststo make a public acknowledgement of Phragmén'sparticipation.Ne-


.he did ask Poincaré for writtensupport to help Phragmén in his
vertheless,
attemptto secure the chair in mechanicsat the universityin Stockholm,and
Phragménwas promotedto the editorialboard of Acta in the followingyear.
In Novemberthat year Phragmén himselfrevealed his interestin Poincaré's
memoirby publishinga paper [11] which showed that some of Poincaré's
resultscould be applied to dynamicalproblemsotherthan the restrictedthree
body problem.

Publicationof the winningentries

By the beginningof January1890 Poincaré had completedhis reworkingof


the memoirand had sent a copy to Phragmén for editing.Not only had he
made substantialalterationsto accommodate the correctionsarisingfromthe
errorbut also, whereappropriate,he had incorporatedthe explanatoryNotes
into the paper itself.Thus in two quite distinctways the memoir took on
a significantlydifferent appearance to that of its predecessor.
Althoughprintingbegan at the end of April that year,a backlog of other
work meant that it was not completeduntil the middle of November.When
Volume 13 of Acta eventuallyappeared it contained both Poincaré's and
Appell's memoirstogetherwith Hermite's reporton the latter.
Weierstrass' reporton Poincaré's revisedmemoirwas promisedfor a fu-
ture volume but it was never completed.Prior to the discoveryof the error
Weierstrass had got as faras writingthe introductionand in March 1889 had
sent it to Mittag-Leffler. However,it was only concernedwithgeneralissues
connectedwith the question as set, and made no specificreferencesto Poin-
caré's memoir26.
Given Mittag-Leffler's initialconcernover obtainingWeierstrass' report,
it mightseem surprisingthat he was not able to induce him to completeit.
However, afterthe discoveryof the error,there was a marked reductionin
Mittag-Leffler's interestin the report.Weierstrass had made it quite plain
that he felta moral obligation to make public the historyof the error,but
Mittag-Leffler's preoccupationwith his own reputationmeant that he was
extremely anxious to play down the error'simportanceand was thereforekeen
forWeierstrass to do likewise.It is temptingto assume that Mittag-Leffler
consideredit in his own best interestsforWeierstrass' reportneverto appear.
Thus, over a year later than Mittag-Leffler had originallyplanned,
the climax to the competition,the publication of the winning entries in
Acta, finallytook place. More than six years had elapsed since Mittag-
Leffler had writtenoptimistically to Kovalevskaya withthe originalplans for
the competition.Despite Kovalevskaya's foreboding,Mittag-Leffler could

26 Sincethe introductionwas not invalidated


by the discovery of theerror,Mit-
tag-Leffler laterselectedit to appearas part of an article[5] whichfocusedon
Weierstrass'interestin the n bodyproblem.

This content downloaded from 91.229.229.49 on Sun, 15 Jun 2014 10:58:35 AM


All use subject to JSTOR Terms and Conditions
120 J.Barrow-Green

scarcelyhave foreseenthe turbulent course of eventswhichwas to follow.


in the finalanalysisMittag-Leffler'sconsiderable
Nevertheless, efforts
were
rewarded.Once theActa volumewas in circulation, therumoursof theerror
fadedand the brillianceof Poincaré'smemoirwas freely Mit-
acknowledged.
tag-Leffler'shope thatthecompetition wouldresultin someimportant new
mathematics had been amplyfulfilled. Poincaré's memoirhad ensuredthat
King Oscar's 60thbirthday wouldnot be forgotten.
celebration

The errorin Poincaré'smemoir

In the publishedmemoir(henceforth referred to as [P2]; the firstprinted


versionbeing referred to as [PI]) Poincaré's only mentionof the erroris
a passingreference in the introduction which,althoughincluding an acknow-
ledgment to Phragmén, givesno indication ofitsnaturenoroftheextentofhis
alterations.Nevertheless, althoughPoincaré's originalcompetition entryre-
mainsuntraced, thediscovery ofhispersonally amendedcopyof[PI], whichis
preserved at the InstitutMittag-Leffler and whichin its alteredformcorres-
ponds almostexactlyto [P2], has made it possibleto followcompletely the
changeswhichPoincaré made.
In essencethe memoiris the culmination of severalstrandsof Poincaré's
workfromthepreviousdecade.In it he broughttogether and applieda whole
hostofideasand techniques whichhe had previously developed, manyofwhich
had originated in hispioneering research on thequalitative theory ofdifferential
equations[12]. Although in the firstpart of the memoir, which is concerned
withthe development of the underlying his
theory, approach was essentially
he a
geometrical, adopted methodology which incorporated the consideration of
complementary and
geometrical analytical theory, with the result that the error
too occurredin complementary forms:geometrically in thetheoryofinvariant
integralsand then analytically in the of
theory asymptotic solutions.Its full
implications become clear in the second part of the memoir whenPoincaré
the
dealswith application of the to
theory systems with two degreesoffreedom,
such as thosewhich describe the restricted three body problem.
Althoughthe areas of the memoirin whichthe erroroccurredcan be
pin-pointed quiteprecisely, sincethereare in theheartof Poincaré'smathema-
ticaltheory,a fewpreliminary remarksare helpfulin orderto put the error
properly into context.
First,what was the actual problemPoincaré was tryingto solve? The
competition questionhad askedfora solutionto then bodyproblem. Thatis,if
n particlesmovein space under their mutual gravitational attraction and with
is it
knowninitialconditions, possible to determine theirsubsequent motion?
Previousattempts to solve this had
problem generally centered on the specific
case ofthe three bodyproblem since this was the firstcase which could notbe
solvedexactly. The three body problem is in fact a complicated non-linear
problemand has a long and well documentedhistoryof its own (see for
exampleWhittaker[13] and Marcolongo [14]). Apartfromits intrinsic

This content downloaded from 91.229.229.49 on Sun, 15 Jun 2014 10:58:35 AM


All use subject to JSTOR Terms and Conditions
ontheThreeBodyProblem
Poincaré 121

appealas a problemsimpleto state,a strongmotivation foritsstudyhas been


its relationship withthe fundamental question of the of the solar
stability
a
system, question in which Poincaré himself was interested.
particularly
However,althoughPoincaré intendedto tackle the generalthree-body
problem, theinherent associatedwithit led himto focushis atten-
difficulties
tionon a particular form
simplified oftheproblemnowknownas therestricted
threebodyproblem27. In thiscase two of the bodies(the primaries)revolve
aroundtheircentreof mass in circularorbitsunderthe influence of their
mutualgravitational whilethethirdbody(theplanetoid),
attraction, assumed
masslesswithrespect to theothertwobodies,movesin theplanedefined bythe
two primaries and, whilebeinggravitationallyinfluenced by them, exertsno
influence ofits own.The restricted problemis thento ascertainthemotionof
theplanetoid.
In Poincarê'sformulation of theproblemthepositionof theplanetoidin
is
phase space describedby two linearand two angularvariables,x¿ and yt
respectively,y¡ being periodicwith period 2n, connectedby the integral
F(xl9x2,yuyi) = C. He put the differential equationsinto Hamiltonian form,
áx* 3F fy _8F_ (l
at dyt9 at ' '
3x,f
which,in accordancewithhisqualitative theory developedin [12], he regarded
as definingflowson a three-dimensional surface.
His brilliant
insight,introduced in [12], was to recognisethatratherthan
consideringtheflowin theentirethree-dimensional space,it was muchmore
convenient to considerthe firstreturnmap inducedby the flowon a two-
dimensional surfaceofsectionS transverse to theflow.Thismap is defined by
choosinga pointM on S at whichS is intersected by a flowline,theimageof
M underthemap is thenthepointM' at whichthatflowlinefirstintersects
S again.Thusin thethree-dimensional spacea periodicsolutioncorresponds to
a closedcurve,butunderthemap a 2n periodicsolutioncorresponds to a fixed
pointand a 2nkperiodicsolutioncorresponds to a cycleof periodk.
Of particularimportance in thememoiris thechapterin whichPoincaré
discusseshis theoryof periodicsolutionsand whichincludeshis discovery of
asymptoticsolutions.Theseare solutionswhichslowlyeitherapproachor move
awayfroman unstable(generating) periodicsolution.He showedthatin the
three-dimensionalsolutionspace of the restricted problem,theseasymptotic
solutionsgeneratefamiliesof curveswhichfillout surfaces and whichasymp-
approachthecurverepresenting
totically thegenerating periodicsolution,and
thatthesesurfaces correspond to curvesin thetransverse section.In orderto
to an of
try gain understanding the behaviourof theseasymptotic solutions,
Poincaréinvestigated thenatureofthecurveson thetransverse section;it was
thisinvestigation
whichrequiredhis theoryof invariant integrals.

Poincaréwasthefirst
to use theterm'problème
restreint'
[2, III, p. 69].

This content downloaded from 91.229.229.49 on Sun, 15 Jun 2014 10:58:35 AM


All use subject to JSTOR Terms and Conditions
122 J.Barrow-Green

With respectto the analyticpart of Poincaré's theory,the parameterwhich


he used to formthe power seriesexpansions of the solutionsto the differential
equations in the restrictedproblemwas ¿¿,the mass of the smallerof the two
primaries.Poincaré chose 'i as the parameterbecause when'i = 0 the problem
reducesto a pair of two body problems.This meantthathe was able to employ
the strategyof startingwitha particularsolutionforwhich'i = 0 and thenvary
'i analyticallyto see if solutionsexistedfor verysmall values of 'i.

Invariantintegrals

AlthoughPoincaré was not the firstto recognisethe existenceand value


of invariant integrals- they are earlier encounteredin the work of both
Liouville [15] and Boltzmann [16] - he was the firstto formalisea theory
centred on the concept. In an earlier paper [17] he had used the idea of
a particularinvariantintegralwithinthe contextof a problemconcerningthe
equations. In the memoirhe considered
stabilityof the solutionsof differential
the whole concept in a broader sense, developing a general theory which
revealedthat the existenceof an invariant integralis a fundamentalpropertyof
Hamiltoniansystemsof differential equations. In particular,he showed thatit is
a propertyof the systemof differential equations whichdescribesthe restricted
threebody problem.
To give a dynamicalinterpretation of the idea, Poincaré used the example
of the motion of an incompressiblefluid,where the motion of the fluid is
describedby the differential equations
dx dy -z
' ' ^ _ '
áí ¿t di

condition
togetherwith the incompressibility
ÔX ÕY 1 dZ = .
1 o.
dx dy dz
the flowis volumepreservingand so the volume,
As the fluidis incompressible,
which is given by the tripleintegral¡¡¡dxdydz, is an invariantintegral.
He concluded his discussionof invariantintegralswitha seriesof theorems
characterizedby theirgeometricnature.These theoremsincludeone of his most
celebrated results: the original formulationof his recurrencetheorem [1,
p. 314]28.

28 Poincaré consideredthecase wherethereare threedegreesof freedom


Briefly,
and the flowis volumepreserving,and he supposedthattherewas a movingpoint
P which remained bounded undertheflow.He thenproved,albeitsomewhat informally,
thatat somefuture close to its initialsituation
timethepointP willreturnarbitrarily
often.
and will do so infinitely

This content downloaded from 91.229.229.49 on Sun, 15 Jun 2014 10:58:35 AM


All use subject to JSTOR Terms and Conditions
Poincaréon theThreeBodyProblem 123

Figure1. Figure2.

This series of theoremsprovides the geometricalframeworkfor the later


analysis,the qualitativestudygivingan insightinto the global behaviourof the
system.It was in one of thesetheorems,the Corollaryto TheoremIII in [PI],
that the fundamentalerrorin his geometryoccurred.In effectPoincaré failed
to take proper account of the exact geometricnature of a particularcurve.
Beforediscussingthe relevanttheorems,that is Theorem III [PI] and its
Corollary and Theorem HI [P2], one definitionfrom [PI] (which does not
appear in [P2]) in needed. In Theorem HI [PI] Poincaré uses a term,quasi-
closed,whichis importantwithregardto the error,althoughunfortunately his
definitionis not altogetherclear. He said that an nthorder curve,by whichhe
meanta curvecoincidentwithits nthiterate,was quasi-closedif therewere two
pointsA and B on it whichwere separatedby a finitearc but whose distance
apart was verysmall of pthorder.
In Theorem III [PI] Poincaré proved that if an invariantcurve C was
quasi-closedsuch that the distancebetweenthe points A and B was also very
small of nthorder(verysmall of nthorderbeing preciselydefined)and therewas
an invariantintegral,then the distancebetweenthe point A and its iterateAx
and that betweenthe point B and its iterateBx were also very small of nth
order.In provingthe theoremPoincaré referredto Figures 1 and 2 and his
argument,which hinged on an application of the triangleinequality,showed
that the configuration givenin Figure 2, as opposed to that givenin Figure 1,
was correct.
Then in the Corollaryto Theorem IH [PI] Poincaré claimed that in the
case where an invariantcurve C is thoughtto be quasi-closed because the
distanceAB is known to be verysmall of nthorder at least and the distance
AAl is known to be finiteor small of (n - l)thorder at most,and thereis an
invariantintegral,thenin factthe curve C is not quasi-closed but it is actually
closed. He gave no proof,simplyobservingthat if the curve was only quasi-
closed thenthe distanceAAt would have to be of nthorder.What he did not
explore was the possibilitythat the curve ratherthan being closed mightbe
This was where he made his mistake.In essence he failed to
self-intersecting.
take into account the fullrange of possibilitiesconsistentwiththe constraintof

This content downloaded from 91.229.229.49 on Sun, 15 Jun 2014 10:58:35 AM


All use subject to JSTOR Terms and Conditions
124 j. Barrow-Green

Figure3.

area-preservation imposed by the existenceof the invariantintegral.Although


he knew that the area insidethe curvehad to remainconstantand independent
of the iterativeprocess,he focusedon a singleiterationand appears not to have
investigatedthe possible outcomesengenderedby the extensionof the iterative
process. As he later realised and showed in Theorem III [P2], the concept
consistentwith area preservationwas not closure but self-intersection.
PoiNCARÉset up Theorem III [P2] in essentiallythe same way as the
Corollary to Theorem III [PI]. He then showed that of the possible hypo-
theses:
1. The two arcs AAX and BBX intersecteach other.
2. The curvilinearquadrilateralAA1B1B is such that the four arcs which
compriseits sides do not have a point in common except for the four
cornersA, Au B, Bx (as in Figure 1).
3. The two arcs AB and AXBXintersecteach other at a point D (as in
Figure 2).
4. One of the arcs AB or AXBXintersectsone of the arcs AA1 or BBX' but
the arcs AAX and BBY do not intersecteach other,neitherdo the arcs
AB and AXBX.
He foundthat he could eliminatethe second and the fourthhypothesesbecause
theyfailedthe conditionof area preservation;in both cases the area AMB was
not equal to the area AxMBl9 althoughin orderto prove the lattercase he had
to do some additional juggling with the arcs. He also found that the third
hypothesiswas unacceptablebecause it impliedthat the distanceAAX must be
a quantityof nthorder(as in TheoremIII [PI]) and not a distanceof (n - l)th
orderat most as specified.Thus he could conclude thatthefirsthypothesis,that
the curve was self-intersecting,was the only one possible.
PoiNCARÉdid not include a diagram,but the correctformof the curve is
shown in Figure 3. The reason the correct curve has to take this slightly
complicated shape with two crossing points is to allow for more than one
iteration.If afterone iterationthe curve had only one crossingpoint,thenany
subsequentiterationwould violate the condition of area preservation.

This content downloaded from 91.229.229.49 on Sun, 15 Jun 2014 10:58:35 AM


All use subject to JSTOR Terms and Conditions
Poincaréon theThreeBodyProblem 125

solutions
Asymptotic

The errorwas thenreflected in Poincaré'sanalytical description of asymp-


toticsolutionswhichappearsat theend of thechapteron periodicsolutions.
A majorconcernin thischapteris the stability of the solutions;particularly
important in thisconnection is Poincaré's recognition of the significance of
certainconstantswhichhe called characteristic exponents. Poincaré's insight
was to realisethat it was the formof the characteristic exponentswhich
indicatedthestability of thesolutions.He saw thatifthecharacteristic expo-
nentswerepurelyimaginary, thenthe solutionwas stable,otherwise it was
unstable.However,as he discovered, theroleofcharacteristic exponents in the
generation of asymptotic solutionsmeantthattheyturnedout to have special
significance withregardto the error.
In [PI] whenPoincarécalculatedtheseriesexpansions fortheasymptotic
solutions,he claimedthatwhenthe differential equationsdependedon the
parameter ¡x,theseriescouldbe expandedin powersof¡1 or J~'i,accordingto
thecircumstances, althoughnowheredid he provethatsuchexpansionswere
actuallypossible.Furthermore, implicitin his claimwas thattheseserieswere
convergent, althoughagain he gave no proof.Particularly significant with
regardto theerroris thefactthathe did notdistinguish betweenautonomous
and nonautonomous systems ofdifferentialequations.He wrongly assumedthat
in bothcases the serieswereconvergent.
In [P2] Poincaré gave carefulconsideration to the convergence question,
and foundthatin thecase oftheseriesfortheasymptotic solutionstherewere
certainnecessary conditions forconvergence, one ofwhichwas thatwhen'i = 0,
all thecharacteristic exponents haveto be distinct. But whenhe calculatedthe
characteristicexponents as seriesexpansionsin powersof theparameter /x,he
discovered thatin theparticular case ofan autonomous Hamiltonian system, all
thecharacteristic exponents are zerowhen'i = 0, and furthermore, they cannot
be expandedin integer powersoffi,butinsteadhaveto be expandedin powers
of J~'l This necessarily implied,contrary to whathe had previously believed,
thattheseriesfortheasymptotic solutionsto therestricted threebodyproblem
weredivergent, and furthermore thattheywereseriesin powersof *J~'irather
than¡i29.He thenwenton to showthatthesedivergent seriesbelongto that
specialclassofserieswhichare nowknownas asymptotic series,and whichhe
himself had earlierdefinedin [19].

29 Poincaré was not thefirstto form


seriesin powersof the squareroot of the
parameter. As he acknowledged in hisintroductionto [P2], seriesof thistypeoccurin
Bohlin's paperof 1888 [18], wheretheyare used to overcometheproblemof small
divisorsin planetaryperturbationtheory.Poincaré latermadea detailedexamination
of Bohlin's seriesin the secondvolumeof his MéthodesNouvelles[2].

This content downloaded from 91.229.229.49 on Sun, 15 Jun 2014 10:58:35 AM


All use subject to JSTOR Terms and Conditions
126 J.Barrow-Green

Theerror'simplications

In PoiNCARÉ's geometric representation oftherestrictedproblem,a generat-


ing unstableperiodic solutionand its accompanying of
family asymptotic solu-
tionsare representedin thethreedimensional solutionspaceby a closedcurve
and two asymptotic surfaces.In orderto understand the behaviourof these
asymptotic solutions,Poincaré sought theexact equationsfortheseasymptotic
He firstnotedthatit was possibleto movefromone surfaceto the
surfaces.
otherbychangingthesignoftheparameter ¡xin theequationsforthesurfaces.
Thusby makingsucha signchangeit is possibleto generate thesecondsurface
fromthefirst.Furthermore, sincethesetwo surfaces cut another,theycan be
consideredtogether as twosidesofthesamesurface. Thissurfacewillthenhave
thespecialfeature ofa doublecurveand it is thisdoublecurvewhichidentifies
the particularserieswhichsatisfythe equationsforthe asymptotic surfaces.
The equationsof thesesurfacesare of the form

-=f(yuy2)
Xi

series
wherext and x2 are givenby the asymptotic

*i = sx(yl9y2,y/p), x2 = s2(yuy2,>//¿).
In orderto calculatetheseequationsexactly,Poincaré proceededin three
which,sincethe
stages.In thefirststagehe calculatedthefirsttwocoefficients
serieswerein powersofJ^ gavean approximation withan erroroftheorder
of/x.In thenextstagehe considered numberofcoefficients,
a larger,butfinite,
whichgivean erroroftheorder¡ipforanyfixedp, no matter howlarge.In the
finalstagehe calculatedthe exactequations.
He began by supposingthatthe seriescould be written

xt= x? + y/jJLxf
+ iixf + . . .

where the x¡ are functionsof y1 and y29which would give

x2_ x°2+ x' J~'i


*1
X' + X'4ÏL
as the firstapproximation forthe equationof the surfacetrajectories.
Havingcalculatedthe coefficients, he foundthatthisfirstapproximation,
whichwas of the forma + by/p>dependedupon the choice of a certain
constant,Cx and a certainfiniteperiodicfunction ofy2whichhe labelled[F{'.
In orderto identify theparticular surfacesdescribedby theapproximation
Poincaréconsidered theirintersectionswiththetransverse sectionS definedby
thesurface =
yt 0. Sincethepositionofa point P on thesurfaceS is definedby
thecoordinates (x2/*i)an(*3>2,whichare analogousto polar coordinates, the
curves(x2/*i)= constantare closed concentric curveson the surfaceS and
the positionof a pointP on S is unchangedwheny2 is increasedby In.

This content downloaded from 91.229.229.49 on Sun, 15 Jun 2014 10:58:35 AM


All use subject to JSTOR Terms and Conditions
Poincaréon theThreeBodyProblem 127

Figure4.

He therefore constructeda set of curvesfordifferent values of the constantCx,


the values for C1 having been determinedby his choice for the function'F{'.
As he illustratedin Figure 4[l,p. 419], since each one of these curves lies in
the plane y^ = 0, as y^ is varied from0 throughto In, each curve sweeps out
a surface.More precisely,if a line definedby the equations y2 = constant,
(x2/xi)= constantis drawn througheach point on an arbitraryone of these
curves,then the set of all these lines constitutesa closed surface.
The choice for the function[FJ also determinesthe number of periodic
solutionsto the differentialequations.In thisinstance,due to Poincaré's choice
for the function,thereare four,two stable and two unstable.On the surface
5 the stable solutions are representedby fixed points and the unstable by
double curves,and it is therefore the latterwhichrepresentthe firstapproxima-
tion.In Figure4, the stable solutionscorrespondto the two fixedpointsA and
B, and the unstableones correspondto the two double curves representedby
the unbrokenlines.
The purpose of the second approximation,which only appeared in [P2],
was to determinesome arbitrarybut finitenumberof coefficients of the series
forthe asymptoticsolutions.Since Poincaré had originallybelieved the series
to be convergentratherthan asymptotic, therewas no equivalentdiscussionin
[PI]. However, most of the second approximationwas in facttaken fromNote
F (entitledAsymptotic Surfaces)which Poincaré had added to [PI] because he
had wanted to include an analytic descriptionof the asymptoticsurfacesto
complementhis geometricone. In [PI], since he believed that the asymptotic
surfacescould be representedby series in J~'i which were convergentfor
arbitrary values of yx and y2,providingfiwas sufficientlysmall,he thoughtthat
his calculationsin Note F gave a fulldescriptionof the entireseries.In [P2] he
wrote the approximateequations for the asymptoticsurfaces as (divergent)
asymptoticseries.
However,it is in the finalapproximation- in which Poincaré constructed
the asymptoticsurfacesexactly,or rathertheirintersectionwith the transverse

This content downloaded from 91.229.229.49 on Sun, 15 Jun 2014 10:58:35 AM


All use subject to JSTOR Terms and Conditions
128 J.Barrow-Green

*U- - 'A'- ÍA
n*Y' Wo)
' v- -7
Figure5.

section}>i = 0 - thatthedifferences between[PI] and [P2] are mostdramatic.


In both[PI] and [P2] he used Figure5 [1,438] to illustrate his construction.
In [PI] the plain lines AO'B' and A'O'B represent the two asymptotic
surfaces whichcut thesurfaceyx= 0, and thedashedlinerepresents thecurve
=
j;1 y2 = 0. The dotted and dashed line,which is a closed curve with a double
pointat O, represents = =
the curvesyt 0, (x2/xi) a + by/p.The generating
(unstable)periodicsolutionis represented by a closed trajectory cuttingthe
surfaceyx= 0 at the point0', and the distance00' is of order'i.
PoiNCARÊ used his resultsfromtheend of thefirstapproximation to infer,
first,thatthecurveBO'B' was quasi-closed and,second,thatthedistanceofthe
point B to its iteratewas of the orderof y/jLAppealingto the (invalid)
Corollaryto TheoremIII he concluded(erroneously) thatthecurveBO'B' was
closed.Then,havinginvokedthe same argument to concludethatthe curve
AO'A'was closed,he was led to themistaken resultthattheasymptotic surfaces
were closed. Furthermore, inherentin thisresultwas the of
implication stability
in the sensethatthe solutionsremainedconfined to a givenregionof space.
Thus,in [PI] PoiNCARÉ believedhe had provedthatforsufficiently small
values of the parameter'i therewas, relativeto a given unstable periodic
solution,a set of asymptotic solutionswhichcould be considered stable,that
thesesolutionswerewell behaved,and thattheycould be completely under-
stood.His analysisin [P2] led to a verydifferent conclusion.
Althoughin [P2] he again referred to Figure5, he now consideredthe
dottedand dashedlineto represent thecurveswithequationsderivedfromthe
firstp termsin the seriesforthe asymptotic solutions.The questionhe then
asked was whetherthe curvesAO'B' and A'O'B werenecessarily closed.He
knewforcertainthattheywould have been if the series for the asymptotic
solutionshad beenconvergent. In thiscase,theplaincurveswouldhavediffered
as littleas requiredfromthedottedand dashedcurves,sincethedistancefrom
a pointon theformer to a pointon thelatterwouldhave tendedto zero as
p increased ButwerethecurvesAO'B' and A'O'B necessarily
indefinitely. closed
eventhough the serieswere divergent?Consideration ofthe specificexample of
a simplependulumweaklycoupledto a linearoscillatorgavehimtheanswer.

This content downloaded from 91.229.229.49 on Sun, 15 Jun 2014 10:58:35 AM


All use subject to JSTOR Terms and Conditions
Poincaréon theThreeBodyProblem 129

He found that in this particularcase the curves were not closed. He had
therefore shown that closure was not an inherentfeature.Nevertheless,as he
had learntfromhis workon invariantintegrals,lack of closuredid not rule out
Thus the question he now asked was whetherit
the possibilityof intersection.
was possibleforthe curvesO'B and O'B' to intersect.For if this should occur,
any trajectorywhich passed throughthe point of intersectionwould simulta-
neouslybelong to both sides of the asymptoticsurface.In otherwords if C is
the closed trajectorywhich passes throughthe point O' and representsthe
periodic solution,then a trajectorypassing throughthe point of intersection
would begin, when t is very large and negative,by being very close to the
closed trajectoryC, and it would then asymptoticallymove away, deviating
greatlyfromC, beforeasymptotically reapproachingC as t becomes verylarge
and positive.By showingthatthesystemsatisfiedthe conditionsof TheoremIII
[P2], Poincaré was able to show that such trajectories,
whichhe called doubly
asymptotic did indeed exist,and moreoverthattherewerein factan
trajectories,
infinitenumber of them. Poincaré later called these trajectorieshomoclinic
trajectoriesand the pointsof intersectionare now knownas homoclinicpoints30.
This is arguablythe firstmathematicaldescriptionof chaotic motionwith-
in a dynamicalsystem.AlthoughPoincaré drew littleattentionto the com-
plexityof the behaviour he had discovered and made no attemptto draw
a diagram, he was profoundlydisturbedby his discoveryas he revealed in
a letter(postmarked1.12.1889)to Mittag-Leffler31:

"J have writtenthismorningto M. Phragmén to tell himof an errorI have


made and doubtlesshe has shownyou myletter.But the consequencesof this
errorare moreseriousthanI firstthought.It is not truethatthe asymptotic
surfacesare closed,at least in the sense whichI originallyintended.What is
trueis thatifbothsides of thissurfaceare considered(whichI stillbelieveare
connectedto each other)theyintersectalong an infinite numberof asymptotic
(and moreoverthattheirdistancebecomesinfinitely
trajectories small of order
higherthan ¡ip howevergreat the orderof p).
I had thoughtall theseasymptotic curves,havingmovedawayfroma clo-
sed curverepresenting a periodicsolution,wouldthenasymptotically approach
the same closed curve.What is trueis that thereare an infinity whichenjoy
thisproperty.
I will not concealfromyou the distressthis has caused me. In thefirst
place, I do not know if you will still thinkthat the resultswhichremain,
namelytheexistenceof periodicsolutions,theasymptotic solutions,the theory
of characteristic
exponents,the non-existence of single-valuedintegrals,and
thedivergenceof Lindstedt's series,deservethe great rewardyou have given
them.

30 Poincaré firstused the


word 'homocline'in [14, III, p. 384].
31 Tr.JB-G.

This content downloaded from 91.229.229.49 on Sun, 15 Jun 2014 10:58:35 AM


All use subject to JSTOR Terms and Conditions
130 J. Barrow-Green

On the otherhand, many changes have become necessaryand I do not


know if you can begin to printthe memoir;I have telegraphedPhragmén.
In any case, I can do no morethanto confessmyconfusionto a friendas
loyal as you. I willwriteto you at lengthwhenI can see thingsmoreclearly."

Perhaps a further indicationof Poincaré's concernand confusionat the discov-


ery of the strangebehaviourexhibitedby these solutionscan be detectedin the
factthat when he describedthemin [P2] he made verylittlecommentabout
theircomplexity,nor did he draw attentionto them in his introduction.Of
course, this may well have been due to the fact that he feltunable to do so
withoutmentioningthe error (which Mittag-Leffler had asked him not to
do), or because he had had too littletimein whichto assess the implicationsof
his discovery.Nevertheless,it is notable that an intervalof almost ten years
elapsed beforethe publicationin 1899 of the thirdvolume of his Mécanique
Céleste [2] in which he reconsideredthe question again, this time adding the
now well known remark about these doubly asymptoticsolutions [2, III,
p. 389]32.

"When one tries to depict thefigureformedby these two curvesand their


infinity each of whichcorrespondsto a doubly asymptotic
of intersections,
solution,these forma kindof net,web, or infinitely
intersections tightmesh;
neitherof the two curvescan ever intersectitself,but mustfold back on itself
in a verycomplexway in orderto intersectall the linksof the meshinfinitely
often.
One is struck by the complexityof this figure which I am not even
attempting to draw.Nothingcan giveus a betteridea of thecomplexity of the
three-body problem and in generalall the problemsof dynamics where thereis
no single-valuedintegral and Bohlin's series diverge."

Acknowledgements. This paper derivesfromthe PhD thesisI preparedat The Open


University1989-1993 and I am very gratefulto JeremyGray whose constanthelp,
enthusiasmand supportin supervisingthe thesisensuredits completion.I also wish to
thank Jesper Lützen and John Fauvel for valuable help in editingthe manuscript,
for allowing me to use theirarchives.
and the InstitutMittag-Leffler

Keierences

1. H. POINCARÊ:'Sur le problémedes trois corps et les équations de la dynamique'


Acta Mathematica13 (1890), 1-270 = Œuvres 7, 262-479.
2. H. POINCARÊ:Les Méthodes Nouvelles de la Mécanique Celeste I-1I1, 1892-18^,
Gauthier Villars, Paris.
3. Y. DOMAR:'On the foundationof Acta Mathematica'Acta Mathematica148 (l^SZj,
3-8.

32 Tr. JB-G.

This content downloaded from 91.229.229.49 on Sun, 15 Jun 2014 10:58:35 AM


All use subject to JSTOR Terms and Conditions
Poincaré on the Three Body Problem 131

4. I. Grattan-Guinness: 'Materials forthe historyof mathematicsin the Instituteof


Mittag-Leffleť Isis 62 (1971), 363-374.
5. G. Mittag-Leffler: 'Zur Biographievon Weierstrass'Acta Mathematica38 (1912),
29-^5.
6. C. Hermite: 'Lettresde Charles Hermiteà Gosta Mittag-Leffler (1884-1891)' Ca-
hiersde Séminairesd'Histoiredes Mathématiques6 (1985), 79-217.
7. K.-R. Biermann: 'Die Mathematikund ihre Dozenten an der BerlinerUniversität
1810-1933' 1988, Akademie-Verlag,Berlin.
8. L. Kronecker: 'Bemerkungenüber Dirichleťs letzteArbeiten'Sitzungesberichte der
Königlich Preussischen Akademie der Wissenschaften1888, 439-442 = Werke V,
471^76.
9. H. Poincare: Lettresd Henri Pomcare a M. Mittag-Leffler concernantle mémoire
couronnédu prix de S.M. Le Roi Oscar II', Acta Mathematica38 (1921), 161-173.
10. H. Gyldén: 'Untersuchungen über die Convergenzder Reigen,welchezur Darstel-
lung der Coordinaten der Planeten angewendetwerden'Acta Mathematica9 (1887),
185-294.
11. E. PhragmêN: 'Poincaré'ska falletaf tekropparsproblemeť ( = On some dynamical
problemswhich are relatedto the restricted
threebody problem)K.V.À. Bihangtill
Handlingar15 (1) No. 13 (1889), 1-33.
12. H. Poincaré: 'Mémoire sur les courbes définiespar une équation différentielle1
Journalde Mathématiques(3) 7 (1881), 375-^22; 8 (1882), 251-296; Journal de
Mathématiques Pure et Appliquées(4) 1 (1885), 167-244; 2 (1886), 151-217 = Œuvres
T 3-44, 44-84, 90-161: 167-222.
13. E. T. Whittaker: 'Report on the progressof the solution of the problemof three
bodies', BAAS Report 1899, 121-159.
14. R. Marcolongo: 'II problema dei tre corpi da Newton (1686) al nostrigiorni',
1919, Hoepli, Milan.
15. J. Liouville: 'Note sur la théoriede la variationdes constantesarbitraires'Journal
de MathématiquesPures et Appliquées3 (1838), 342-349.
16. L. Boltzmann: 'Ueber die Druckkräfte,weiche auf Ringe wirksamsind, die in
bewegteFlüssigkeittauchen'Journalfür die Reine und AngewandteMathematik73
(1871), 111-134.
17. H. POINCARÉ:'Sur un méthode de M. Lindstedť BulletinAstronomique3 (1886),
57-61 = Œuvres 7, 546-560.
18. K. Bohlin: 'Zur Frage der Convergenz der Reihenentwickelungen in der Stör-
ungstheorie'Astronomische Nachrichten121 (1888), 17-24.
19. H. Poincaré: 'Sur les intégralesirrégulières des équations linéaires'Acta Mathema-
tica 8 (1886), 295-344 = Œuvres I, 290-332.

The Open University


Milton Keynes MK7 6AA
England

(ReceivedMay 3, 1994)

This content downloaded from 91.229.229.49 on Sun, 15 Jun 2014 10:58:35 AM


All use subject to JSTOR Terms and Conditions

You might also like