(Modern Birkhäuser Classics) Noga Alon, Jean Bourgain, Alain Connes, Mikhael Gromov, Vitali D. Milman - Visions in Mathematics - GAFA 2000 Special Volume, Part I Pp. 1-453-Birkhäuser Basel (2010)

Modern Birkhäuser Classics
Many of the original research and survey monographs in pure and

applied mathematics published by Birkhäuser in recent decades
have been groundbreaking and have come to be regarded as found-
ational to the subject. Through the MBC Series, a select number of
these modern classics, entirely uncorrected, are being re-released
in paperback (and as eBooks) to ensure that these treasures remain
accessible to new generations of students, scholars, and resear-
chers.
Visions in Mathematics
GAFA 2000 Special Volume, Part I
N. Alon
J. Bourgain
A. Connes
M. Gromov
V. Milman
Editors
Reprint of the 2000 Edition
Birkhäuser
Editors:
N. Alon J. Bourgain
School of Mathematical Sciences School of Mathematics
University of Tel Aviv Institute for Advanced Study
Tel Aviv 69978 Princeton University
Israel Princeton, NJ 08540
e-mail: nogaa@math.tau.ac.il USA
e-mail: bourgain@math.ias.edu
A. Connes
Collège de France M. Gromov
3, rue d‘Ulm Institut des Hautes Études Scientifiques
75231 Paris cedex 05 35, Route de Chartres
France 91440 Bures-sur-Yvette
e-mail: alain@connes.org France
e-mail: gromov@ihes.fr
V. Milman
Department of Mathematics
University of Tel Aviv
Tel Aviv 69978
Israel
e-mail: milman@post.tau.ac.il
Originally published under the title “Visions in Mathematics. Towards 2000”

as Special Volume, Part I, pp. 1-453 of the Journal “GAFA Geometric And
Functional Analysis” by Birkhäuser Verlag, Switzerland, ISSN 1016-443X
© 2000 Birkhäuser Verlag, P.O. Box 133, CH-4010 Basel, Switzerland
2000 Mathematics Subject Classification 35-02, 37-02, 00A05, 00A79, 01A67,

17B37, 42B20, 53C23, 76F55, 82C03
Library of Congress Control Number: 2010920640
Bibliographic information published by Die Deutsche Bibliothek

Die Deutsche Bibliothek lists this publication in the Deutsche Nationalbibliografie;
detailed bibliographic data is available in the Internet at <http://dnb.ddb.de>.
ISBN 978-3-0346-0421-5
This work is subject to copyright. All rights are reserved, whether the whole or
part of the material is concerned, specifically the rights of translation, reprinting,
re-use of illustrations, broadcasting, reproduction on microfilms or in other ways,
and storage in data banks. For any kind of use whatsoever, permission from the
copyright owner must be obtained.
© 2010 Birkhäuser, Springer Basel AG

P.O. Box 133, CH-4010 Basel, Switzerland
Part of Springer Science+Business Media
Printed on acid-free paper produced of chlorine-free pulp. TCF ∞
ISBN 978-3-0346-0421-5 e-ISBN 978-3-0346-0422-2
987654321 www.birkhauser.ch
4ABLE OF #ONTENTS
&OREWORD
0ROGRAM OF THE #ONFERENCE
3 "LOCH AND ( %SNAULT

'AUy[-ANIN DETERMINANT CONNECTIONS AND PERIODS OF
IRREGULAR CONNECTIONS
* "OURGAIN
0ROBLEMS IN (AMILTONIAN 0$%S
HLICH 4 0 4SAI AND ( 4 9AU
* &RO
/N A CLASSICAL LIMIT OF QUANTUM THEORY AND THE
NON LINEAR (ARTREE EQUATION
74 'OWERS
2OUGH STRUCTURE AND CLASSIlCATION
- 'ROMOV
3PACES AND QUESTIONS
6 +AC
#LASSIlCATION OF INlNITE DIMENSIONAL SIMPLE GROUPS OF
SUPERSYMMETRIES AND QUANTUM lELD THEORY
$ +AZHDAN
'EOMETRIZATION IN REPRESENTATION THEORY
! "ERENSTEIN AND $ +AZHDAN

'EOMETRIC AND UNIPOTENT CRYSTALS
! "RAVERMAN AND $ +AZHDAN

l FUNCTIONS OF REPRESENTATIONS AND LIFTING
!PPENDIX BY 6 6OLOGODSKY
3 +LAINERMAN
0$% AS A UNIlED SUBJECT
! +UPIAINEN
,ESSONS FOR TURBULENCE
%( ,IEB AND * 9NGVASON

4HE MATHEMATICS OF THE 3ECOND ,AW OF 4HERMODYNAMICS
sSZ
, ,OVA
$ISCRETE AND CONTINUOUS TWO SIDES OF THE SAME
9 .EEMAN
0YTHAGOREAN AND 0LATONIC CONCEPTIONS IN
88TH #ENTURY PHYSICS
3 .OVIKOV
#LASSICAL AND MODERN TOPOLOGY 4OPOLOGICAL
PHENOMENA IN REAL WORLD PHYSICS
9A' 3INAI
3OME PROBLEMS IN THE THEORY OF DYNAMICAL SYSTEMS AND
MATHEMATICAL PHYSICS
%- 3TEIN
3OME GEOMETRICAL CONCEPTS ARISING IN HARMONIC ANALYSIS
&OREWORD
4HE MEETING <6ISIONS IN -ATHEMATICS [ 4OWARDS TOOK PLACE
MAINLY AT 4EL !VIV 5NIVERSITY IN !UGUST 3EPTEMBER WITH
A FEW DAYS AT THE 3HERATON -ORIAH (OTEL AT THE $EAD 3EA (EALTH
2ESORT 4HE MEETING INCLUDED ABOUT LECTURES BY SOME OF THE LEADING
RESEARCHERS IN THE WORLD IN MOST AREAS OF MATHEMATICS AND A NUMBER
OF DISCUSSIONS IN DIkERENT DIRECTIONS ORGANIZED IN VARIOUS FORMS
4HE GOALS OF THE CONFERENCE AS DElNED BY THE SCIENTIlC COMMIT
TEE CONSISTING OF . !LON * "OURGAIN ! #ONNES - 'ROMOV AND
6 -ILMAN WERE TO DISCUSS THE IMPORTANCE METHODS FUTURE AND
UNITYDIVERSITY OF MATHEMATICS AS WE ENTER THE ST #ENTURY TO CON
SIDER THE RELATION BETWEEN MATHEMATICS AND RELATED AREAS AND TO DIS
CUSS THE PAST AND FUTURE OF MATHEMATICS AS WELL AS ITS INTERACTION WITH
3CIENCE
! NEW FORMAT OF MATHEMATICAL DISCUSSIONS DEVELOPED BY THE END
OF THE #ONFERENCE INTO AN INTERESTING ADDITION TO THE MORE STANDARD
FORM OF LECTURES AND QUESTIONS
7E BELIEVE THAT THE MEETING SUCCEEDED IN GIVING A WIDE PANORAMA
OF MATHEMATICS AND MATHEMATICAL PHYSICS BUT WE DID NOT TOUCH UPON
THE INTERACTION OF MATHEMATICS WITH THE EXPERIMENTAL SCIENCES
4HIS IS THE lRST PART OF THE PROCEEDINGS OF THE MEETING 4HE SECOND
PART WILL APPEAR LATER THIS YEAR
)T IS A PLEASURE TO THANK -RS -IRIAM (ERCBERG FOR HER GREAT TECH
NICAL HELP IN THE PREPARATION OF THIS MANUSCRIPT
. !LON * "OURGAIN
! #ONNES - 'ROMOV
6 -ILMAN
GAFA2000
Tel Aviv, Israel
W. T. Gowers
S. Bloch M. Gromov
J. Bourgain V. Kac
J. Fröhlich D. Kazhdan
S. Klainerman Y. Ne’eman
A. Kupiainen S. Novikov
E. H. Lieb Ya. G. Sinai
L. Lovász E. M. Stein
6ISIONS IN -ATHEMATICS [ 4OWARDS
3CIENTIlC 0ROGRAM
7EDNESDAY !UGUST
,EV !UDITORIUM 4EL !VIV 5NIVERSITY
/PENING REMARKS

'EOMETRY #HAIR , 0OLTEROVICH
- 'ROMOV 'EOMETRY AS THE ART OF ASKING QUESTIONS
( (OFER (OLOMORPHIC CURVES AND REAL THREE DIMENSIONAL
DYNAMICS
9 %LIASHBERG 3YMPLECTIC lELD THEORY
%RGODIC 4HEORY $YNAMIC 3YSTEMS #HAIR 3 -OZES

( &URSTENBERG $YNAMICAL METHODS IN DIOPHANTINE PROB
LEMS
' -ARGULIS $IOPHANTINE APPROXIMATION LATTICES AND
mOWS ON HOMOGENEOUS SPACES
9 3INAI /N SOME PROBLEMS IN THE THEORY OF DYNAMICAL
SYSTEMS AND MATHEMATICAL PHYSICS
$ISCUSSION WITH INTRODUCTION BY 9 3INAI
4HURSDAY !UGUST
-ATHEMATICAL 0HYSICS #HAIR ! *AkE

* &ROHLICH ,ARGE QUANTUM SYSTEMS
9 .EEMAN 0HYSICS AS GEOMETRY 0LATO VINDICATED
! #ONNES .ON COMMUTATIVE GEOMETRY
#OMPUTER 3CIENCE #HAIR . !LON
0 3HOR -ATHEMATICAL PROBLEMS IN QUANTUM INFORMATION

THEORY
! 2AZBOROV #OMPLEXITY OF PROOFS AND COMPUTATION
! 7IGDERSON 3OME FUNDAMENTAL INSIGHTS OF COMPUTA
TIONAL COMPLEXITY
- 2ABIN 4HE MATHEMATICS OF TRUST AND ADVERSITY
&RIDAY !UGUST
-ORIAH (OTEL $EAD 3EA
#HAIR ! #ONNES
! *AkE -ATHEMATICS OF QUANTUM lELDS
3 .OVIKOV 4OPOLOGICAL PHENOMENA IN REAL PHYSICS
$ISCUSSION ON -ATHEMATICAL 0HYSICS WITH INTRODUCTION BY

! #ONNES
$ISCUSSION ON 'EOMETRY WITH INTRODUCTION BY - 'ROMOV
3UNDAY !UGUST
-ORIAH (OTEL $EAD 3EA
$ISCUSSION ON -ATHEMATICS IN THE 2EAL 7ORLD IMAGE AP

PLICATIONS ETC WITH INTRODUCTION BY 2 #OIFMAN
$ISCUSSION ON #OMPUTER 3CIENCE AND $ISCRETE -ATHEMAT
ICS WITH INTRODUCTION BY - 2ABIN
-ONDAY !UGUST
!NALYSIS #HAIR ! /LEVSKItp
2 #OIFMAN #HALLENGES IN ANALYSIS

0 *ONES #HALLENGES IN ANALYSIS
% 3TEIN 3OME GEOMETRICAL CONCEPTS ARISING IN HARMONIC
ANALYSIS
$ISCUSSION
.UMBER 4HEORY #HAIR : 2UDNICK

( )WANIEC !UTOMORPHIC FORMS IN RECENT DEVELOPMENTS
OF ANALYTIC NUMBER THEORY
0 3ARNAK 3OME PROBLEMS IN NUMBER THEORY AND ANALYSIS
$ :AGIER /N <Q OR <#ONNECTIONS BETWEEN MODULAR
FORMS COMBINATORICS AND TOPOLOGY
$ISCUSSION <4HE UNREASONABLE EkECTIVENESS OF MODULAR
FORMS INTRODUCED BY 0 3ARNAK
4UESDAY !UGUST
$ISCRETE -ATHEMATICS #HAIR ! 7IGDERSON
, ,OVsASZ $ISCRETE AND CONTINUOUS TWO SIDES OF PHENOM

ENA
. !LON 0ROBABILISTIC AND ALGEBRAIC METHODS IN DISCRETE
MATHEMATICS
' +ALAI !N INVITATION TO 4VERBERGS THEOREM
$ISCUSSION
!NALYSIS #HAIR % 3TEIN
4 'OWERS 2OUGH STRUCTURE AND CRUDE CLASSIlCATION

* "OURGAIN 3OME PROBLEMS IN (AMILTONIAN 0$%S
6 -ILMAN 4OPICS IN GEOMETRIC ANALYSIS
$ISCUSSION
7EDNESDAY 3EPTEMBER
!LGEBRA #HAIR 6 +AC
3 "LOCH #HARACTERISTIC CLASSES FOR LINEAR DIkERENTIAL EQUA

TIONS
6 6OEVODSKY -OTIVIC HOMOTOPY TYPES
$ +AZHDAN 4HE LIFTING PROBLEMS AND CRYSTAL BASE
$ISCUSSION
!LGEBRA #HAIR $ +AZHDAN
! "EILINSON !ROUND GEOMETRIC ,ANGLANDS

* "ERNSTEIN %QUIVARIANT DERIVED CATEGORIES
6 +AC #LASSIlCATION OF INlNITE DIMENSIONAL SIMPLE GROUPS
OF SUPERSYMMETRIES AND QUANTUM lELD THEORY
$ISCUSSION WITH INTRODUCTION BY $ +AZHDAN
4HURSDAY 3EPTEMBER
#HAIR - 'ROMOV
2 -AC0HERSON /N THE APPLICATIONS OF TOPOLOGY
- +ONTSEVICH 3MOOTH AND COMPACT
$ 3ULLIVAN 3TRING INTERACTIONS IN TOPOLOGY
$ISCUSSION
#HAIR ' +ALAI

) !UMANN -ATHEMATICAL GAME THEORY ,OOKING BACK
WARD AND FORWARD
% (RUSHOVSKI ,OGIC AND GEOMETRY
$ISCUSSION 4HE ROLE OF HOMOTOPICAL ALGEBRA IN PHYSICS
WITH INTRODUCTIONS BY $ 3ULLIVAN AND - +ONTSEVICH
&RIDAY 3EPTEMBER
-ATHEMATICAL 0HYSICS AND 0$% #HAIR 6 :AKHAROV

4 3PENCER 5NIVERSALITY AND STATISTICAL MECHANICS
% ,IEB 4HE MATHEMATICS OF THE SECOND LAW OF THERMO
DYNAMICS
! +UPIAINEN ,ESSONS FOR TURBULENCE
3 +LAINERMAN 3OME GENERAL REMARKS CONCERNING NONLIN
EAR 0$%S
$ISCUSSION
#LOSING REMARKS
#ONTINUATION OF $ISCUSSION
'!&! 'EOM FUNCT ANAL C "IRKH
l AUSER 6ERLAG "ASEL
3PECIAL 6OLUME [ '!&! [
83 '!&! 'EOMETRIC !ND &UNCTIONAL !NALYSIS
'!5y[-!.). $%4%2-).!.4 #/..%#4)/.3 !.$

0%2)/$3 &/2 )22%'5,!2 #/..%#4)/.3
sLE
3PENCER "LOCH AND (E rNE %SNAULT
!BSTRACT
'AUy[-ANIN DETERMINANT CONNECTIONS ASSOCIATED TO IRREGULAR CONNEC
TIONS ON A CURVE ARE STUDIED 4HE DETERMINANT OF THE &OURIER TRANSFORM
OF AN IRREGULAR CONNECTION IS CALCULATED 4HE DETERMINANT OF COHOMOL
OGY OF THE STANDARD RANK +LOOSTERMAN SHEAF IS COMPUTED MODULO
TORSION 0ERIODS ASSOCIATED TO IRREGULAR CONNECTIONS ARE STUDIED IN THE
VERY BASIC EXPF CASE AND ANALOGIES WITH THE 'AUy[-ANIN DETERMI
NANT ARE DISCUSSED
%VERYTHINGS SO AWFUL REGLAR A BODY CANT STAND IT

4HE !DVENTURES OF 4OM 3AWYER
-ARK 4WAIN
)NTRODUCTION
! VERY CLASSICAL AREA OF MATHEMATICS AT THE BORDERLINE BETWEEN APPLIED
MATHEMATICS ALGEBRAIC GEOMETRY ANALYSIS MATHEMATICAL PHYSICS AND NUM
BER THEORY IS THE THEORY OF SYSTEMS OF LINEAR DIkERENTIAL EQUATIONS CON
NECTIONS 4HERE IS A VAST LITERATURE FOCUSING ON REGULAR SINGULAR POINTS
0ICARD &UCHS DIkERENTIAL EQUATIONS $ELIGNES 2IEMANN (ILBERT CORRESPON
DENCE AND ITS EXTENSION TO $ MODULES AND MORE RECENTLY VARIOUS INDEX
THEOREMS IN GEOMETRY
4HE CLASSIlCATION OF IRREGULAR SINGULAR POINTS IN TERMS OF 3TOKES STRUC
TURES IS EXPLAINED IN ;- CHAP )6= )T INVOLVES THE 2IEMANN (ILBERT CORRE
SPONDENCE FOR HOLONOMIC $ MODULES WITH REGULAR SINGULAR POINTS TOGETHER
WITH A lLTRATION RELATED TO THE RATE OF GROWTH OF LOCAL SOLUTIONS
.ONETHELESS THERE ARE SOME VERY MODERN THEMES IN THE SUBJECT WHICH
REMAIN VIRTUALLY UNTOUCHED /N THE ARITHMETIC SIDE ONE MAY ASK FOR EX
AMPLE HOW IRREGULAR CONNECTIONS CAN BE INCORPORATED IN THE MODERN THEORY
OF MOTIVES (OW DEEP ARE THE APPARENT ANALOGIES BETWEEN WILD RAMIlCATION
IN CHARACTERISTIC P AND IRREGULARITY #AN ONE DElNE PERIODS FOR IRREGULAR
CONNECTIONS )F SO DO THE RESULTING PERIOD MATRICES HAVE ANYTHING TO DO
3 ",/#( !.$ ( %3.!5,4 '!&!
WITH n FACTORS FOR @ ADIC REPRESENTATIONS /N THE GEOMETRIC SIDE ONE CAN
ASK FOR A THEORY OF CHARACTERISTIC CLASSES CI % R FOR IRREGULAR CONNECTIONS
SUCH THAT C IS THE RANK OF THE CONNECTION AND C IS THE ISOMORPHISM CLASS
OF THE DETERMINANT 7ITH THESE ONE CAN TRY TO ATTACK THE 2IEMANN 2OCH
PROBLEM
)N THIS NOTE WE DESCRIBE SOME CONJECTURES AND EXAMPLES CONCERNING
WHAT MIGHT BE CALLED A FAMILIES INDEX THEOREM FOR IRREGULAR CONNECTIONS
,ET
F 83
BE A SMOOTH PROJECTIVE FAMILY OF CURVES OVER A SMOOTH BASE 3 ,ET $
BE AN EkECTIVE RELATIVE DIVISOR ON 8 ,ET % BE A VECTOR BUNDLE ON 8 R
% %ij8 $ AN INTEGRABLE ABSOLUTE CONNECTION WITH POLES ALONG $ 4HE
RELATIVE DE 2HAM COHOMOLOGY 2Fb jb8_$ i% INHERITS A CONNECTION 'AUy[
-ANIN CONNECTION AND THE FAMILIES INDEX OR 2IEMANN 2OCH PROBLEM IS TO
DESCRIBE THE ISOMORPHISM CLASS OF THE LINE BUNDLE WITH CONNECTION
DET 2Fb jb8_$ i %
4HE 'AUy[-ANIN CONSTRUCTION IS FAIRLY STANDARD AND WE DO NOT RECALL
IT IN DETAIL "Y WAY OF EXAMPLE WE CITE TWO CLASSICAL FORMULAS 'AUy
HYPERGEOMETRIC AND "ESSEL FUNCTIONS RESPECTIVELY
:
`B `C _ B
& A B C Z UB_ _ U C_B_ _ UZ _A DU
`C
: q r
_N Z DU
*N Z U EXP U_ 3 CIRCLE ABOUT
xI 3 U U
)N BOTH CASES THE INTEGRAND IS A PRODUCT OF A SOLUTION OF A RATHER SIMPLE
DEGREE DIkERENTIAL EQUATION IN U THE SOLUTION BEING EITHER
_ `
UB_ _ U C_B_ _ UZ _A OR U_N EXP Z U U
WITH AN ALGEBRAIC FORM DU OR DUU 4HE INTEGRAL IS TAKEN OVER A CHAIN IN
THE U PLANE 4HE RESULTING FUNCTIONS & A B C Z AND *N Z SATISFY 'AUy[
-ANIN EQUATIONS WHICH ARE MUCH MORE INTERESTING DEGREE EQUATIONS IN Z
)T IS NOT OUR PURPOSE TO GO FURTHER INTO THE CLASSICAL THEORY BUT TO
UNDERSTAND THE ROLE OF THE DETERMINANT WE REMARK THAT IN EACH OF THE
ABOVE CASES THERE IS A SECOND PATH AND A SECOND ALGEBRAIC FORM SUCH
THAT THE TWO INTEGRALS SAY F Z AND F Z SATISFY THE SAME SECOND ORDER
EQUATION 4HE 7RONSKIAN DETERMINANT
k k
k F F k
k DF k
k DF k
DZ DZ
6)3)/.3 ). -!4( )22%'5,!2 #/..%#4)/.3
SATISlES THE DEGREE EQUATION GIVEN BY THE DETERMINANT OF 'AUy[-ANIN

7E ARE WORKING HERE WITH ALGEBRAIC DE 2HAM COHOMOLOGY !NALYTICALLY
IE PERMITTING COORDINATE AND BASIS TRANSFORMATIONS WITH ESSENTIAL SINGU
LARITIES ON $ THE BUNDLE % AN J8_$ CAN BE TRANSFORMED LOCALLY ON 3 TO
HAVE REGULAR SINGULAR POINTS ALONG $ BUT THE ALGEBRAIC PROBLEM WE POSE IS
MORE SUBTLE
)F $ OR MORE GENERALLY IF R HAS REGULAR SINGULAR POINTS THE ANSWER
IS KNOWN
_ `
DET 2Fb jb8_$ i% Fb DET % ? _ DET R ` C 83
)N THE CASE OF REGULAR SINGULAR POINTS THE C HAS TO BE TAKEN AS A RELATIVE
CLASS ;34= ;"%=
)N A RECENT ARTICLE ;"%= WE PROVED AN ANALOGOUS FORMULA IN THE CASE
WHEN % WAS IRREGULAR AND RANK &OR A SUITABLE $ THE RELATIVE CONNECTION
INDUCES AN ISOMORPHISM
_ m `
R83$ %J$ w %J$ i $
IE A TRIVIALIZATION OF $ J$ 4HE CONNECTION PULLS BACK FROM A RANK
CONNECTION % R% ON THE RELATIVE 0ICARD SCHEME 0IC8 $ AND THE 'AUy[
-ANIN DETERMINANT CONNECTION IS OBTAINED BY EVALUATING % R% AT THE
PRIVILEDGED POINT $ R83$ 0IC8 $
7E WANT NOW TO CONSIDER TWO SORTS OF GENERALIZATIONS &IRST WE FOR
MULATE AN ANALOGOUS CONJECTURE FOR HIGHER RANK CONNECTIONS WHICH ARE AD
MISSIBLE IN A SUITABLE SENSE 7E PROVE TWO SPECIAL CASES OF THIS CONJECTURE
COMPUTING THE DETERMINANT OF THE &OURIER TRANSFORM OF AN ARBITRARY CON
NECTION AND UP TO TORSION THE DETERMINANT OF COHOMOLOGY OF THE BASIC
RANK +LOOSTERMAN SHEAF
3ECOND WE INITIATE IN THE VERY SIMPLEST OF CASES % / R DF THE
STUDY OF PERIODS FOR IRREGULAR CONNECTIONS ,ET M DEG F 7E ARE LED TO
A STATIONARY PHASE INTEGRAL CALCULATION OVER THE SUBVARIETY OF 0IC0 M `
CORRESPONDING TO TRIVIALIZATIONS OF M ` AT 4HE SUBTLE PART OF THE
INTEGRAL IS CONCENTRATED AT THE SAME POINT
_ `
M ` R83M` 0IC0 M `
MENTIONED ABOVE
/UR HOPE IS THAT THE GEOMETRIC METHODS DISCUSSED HERE WILL CARRY OVER
TO THE SORT OF ARITHMETIC QUESTIONS MENTIONED ABOVE 7E EXPECT THAT THE
DISTINGUISHED POINT GIVEN BY THE TRIVIALIZATION OF $ PLAYS SOME ROLE IN
THE CALCULATION OF n FACTORS FOR RANK @ ADIC SHEAVES AND THAT THE HIGHER
RANK CONJECTURES HAVE SOME @ ADIC INTERPRETATION AS WELL
3 ",/#( !.$ ( %3.!5,4 '!&!
7E WOULD LIKE TO THANK 0IERRE $ELIGNE FOR SHARING HIS UNPUBLISHED LET
TERS WITH US 4HE BASIC IDEA OF USING THE RELATIVE *ACOBIAN TO STUDY n FACTORS
FOR RANK SHEAVES WE LEARNED FROM HIM 7E ALSO HAVE GOTTEN CONSIDERABLE
INSPIRATION FROM THE WORKS OF 4 3AITO AND 4 4ERASOMA CITED IN THE BIB
LIOGRAPHY 4HE MONOGRAPH ;+= IS AN EXCELLENT REFERENCE FOR +LOOSTERMAN
SHEAVES AND THE SUBJECT OF PERIODS FOR EXPONENTIAL INTEGRALS IS DISCUSSED
BRIEmY AT THE END OF ;+O=
4HE #ONJECTURE
,ET 3 BE A SMOOTH SCHEME OVER A lELD K OF CHARACTERISTIC 7E CONSIDER
A SMOOTH FAMILY OF CURVES F 8 3 AND A VECTOR BUNDLE % WITH AN
ABSOLUTE INTEGRABLE CONNECTION R % % i j8 b$ (ERE $ y 8 IS A
DIVISOR WHICH IS SMOOTH OVER 3 7E ARE INTERESTED IN THE DETERMINANT OF
THE 'AUy[-ANIN CONNECTION
D LOG
DET 2Fb jb8_$ 3 i % 0ICR 3 ( 3 /3a _ j3
w
`3 j3 D LOG /3a
0ROPOSITION ,ET + K3 BE THE FUNCTION lELD OF 3 4HEN THE
RESTRICTION MAP 0ICR 3 0ICR 3PEC+ IS AN INJECTION
0ROOF )N VIEW OF THE INTERPRETATION OF 0ICR AS `j D LOG /a THE PROPO
SITION FOLLOWS FROM THE FACT THAT FOR A MEROMORPHIC FUNCTION G ON 3 WE
HAVE G REGULAR AT POINT S 3 IF AND ONLY IF DGG IS REGULAR AT S c
4HUS WE DO NOT LOSE INFORMATION BY TAKING THE BASE 3 TO BE THE SPEC

TRUM OF A FUNCTION lELD 3 3PEC+ 7E SHALL RESTRICT OURSELVES TO THAT
CASE
,ET
8
$ MX X
X$+
BE AN EkECTIVE DIVISOR SUPPORTED ON $ 3UPPOSE THAT THE RELATIVE CONNECTION

HAS POLES OF ORDER BOUNDED BY $ IE
R83 % % i $
AND THAT $ IS MINIMAL WITH THIS PROPERTY 7E FREQUENTLY WRITE IN PLACE
OF j83 7E VIEW $ AS AN ARTINIAN SCHEME WITH SHEAF OF FUNCTIONS /$
AND RELATIVE DUALIZING SHEAF $ $ 4O SIMPLIFY WE DO NOT USE THE
NOTATION $+
6)3)/.3 ). -!4( )22%'5,!2 #/..%#4)/.3
0ROPOSITION $ElNE %$ % i /$ 4HEN R83 INDUCES A FUNCTION

LINEAR MAP
R83$ %$ %$ i $
0ROOF 3TRAIGHTFORWARD c
,ET J 8 _ $ 8 4HERE ARE NOW TWO RELATIVE DE 2HAM COMPLEXES

WE MIGHT WISH TO STUDY
Jb J b % Jb J b % i
% % i $
4HE lRST IS CLEARLY THE CORRECT ONE &OR EXAMPLE ITS RELATIVE COHOMOLOGY
CARRIES THE 'AUy[-ANIN CONNECTION /N THE OTHER HAND THE SECOND COM
PLEX IS SOMETIMES EASIER TO STUDY AS THE RELATIVE COHOMOLOGY OF THE SHEAVES
INVOLVED HAS lNITE DIMENSION OVER + 3INCE WE ARE PRIMARILY INTERESTED
IN THE IRREGULAR CASE THAT IS WHEN POINTS OF $ HAVE MULTIPLICITY t THE
FOLLOWING PROPOSITION CLARIlES THE SITUATION
0ROPOSITION ,ET NOTATION BE AS ABOVE AND ASSUME EVERY POINT OF
$ HAS MULTIPLICITY t 4HEN THE NATURAL INCLUSION OF COMPLEXES
g h g h
r % % i $ Jb J b % Jb J b % i
IS A QUASIISOMORPHISM IF AND ONLY IF R83$ %$ %$ i $ IN 0ROPOSI
TION IS AN ISOMORPHISM
0ROOF ,ET $ H BE A LOCAL DElNING EQUATION FOR $ 7RITE $ $ $
WHERE $ IS THE REDUCED DIVISOR WITH SUPPORT SUPP$ 7E CLAIM lRST
THAT THE MAP r IS A QUASIISOMORPHISM IF AND ONLY IF FOR ALL N t THE MAP
' DElNED BY THE COMMUTATIVE DIAGRAM
R83$ _NÌDi DH _ `
%%_$ ___________H %$ %_$ i $

w w
Y<`H_N Y<`H_N
m _ ` ' _ `
%N$ % N _ $ ___ %N$ $ %N _ $ $ i$
GIVEN BY H_N E H_N R83$ E _ NH_N_ E i DH IS A QUASI ISOMORPHISM
4HIS FOLLOWS BY CONSIDERING THE COKERNEL OF r
Jb J b %% Jb J b % i % i $
AND lLTERING BY ORDER OF POLE 4HE ASSERTION OF THE PROPOSITION FOLLOWS BE
CAUSE NH_N_ EiDH HAS A POLE OF ORDER STRICTLY SMALLER THAN THE MULTIPLICITY
OF N $ AT EVERY POINT OF $ c
3 ",/#( !.$ ( %3.!5,4 '!&!
7E WILL CONSIDER ONLY THE CASE

R83$ %$ w
%$ i $
7E NOW CONSIDER THE SHEAF Jb J b0
j8
OF ABSOLUTE IE RELATIVE TO K FORMS
ON 8 WITH POLES ON $ ,ET $ MX X BE AN EkECTIVE DIVISOR SUPPORTED
ON $ AS ABOVE AND WRITE $ $ _ $
$EFINITION 4HE SHEAF jP8 F$G y Jb J b jP8 IS DElNED LOCALLY AROUND A
POINT X $ WITH LOCAL COORDINATE Z BY
jP8 F$GX jP8 $ X jP_ DZ
8X > Z MX
,
4HE GRADED SHEAF P jP8 F$G IS STABLE UNDER THE EXTERIOR DERIVATIVE AND
INDEPENDENT OF THE CHOICE OF LOCAL COORDINATES AT THE POINTS OF $ /NE HAS
EXACT SEQUENCES
F b jP3 $ jP8 F$G jP_
3 $ i $
$EFINITION !N INTEGRABLE ABSOLUTE CONNECTION ON % WILL BE CALLED
ADMISSIBLE IF THERE EXISTS A DIVISOR $ SUCH THAT R % % i j8 F$G AND
SUCH THAT R83$ %$ w %$ i $
2EMARK 7HEN % HAS RANK THERE ALWAYS EXISTS A $ SUCH THAT R IS
ADMISSIBLE FOR $ ;"% ,EMMA = )N HIGHER RANK THIS NEED NOT BE TRUE
EVEN IF R83$ IS AN ISOMORPHISM FOR SOME $ &OR EXAMPLE LET p j+
BE A CLOSED FORM ,ET N t BE AN INTEGER AND LET C K C 4HE
CONNECTION MATRIX q CDZ r
p
Z M Z N
!
CDZZM _ Z
NDZ
SATISlES D! ! > ! FOR ALL M N . BUT THE RESULTING INTEGRABLE

CONNECTION IS NOT ADMISSIBLE FOR N M ALTHOUGH R83 IS AN ISOMORPHISM
FOR C N IF M .OTE IN THIS CASE IT IS POSSIBLE TO CHANGE BASIS TO GET AN
ADMISSIBLE CONNECTION 7E DONT KNOW WHAT TO EXPECT IN GENERAL 4HERE
DO EXIST CONNECTIONS FOR WHICH R83$ IS NOT AN ISOMORPHISM FOR ANY $
FOR EXAMPLE IF ONE TAKES A SUM OF RANK CONNECTIONS WITH DIkERENT MX
AS ABOVE SEE NOTATIONS AND A LOCAL BASIS ADAPTED TO THIS DIRECT SUM
DECOMPOSITION
(ENCEFORTH 3 3PEC+ IS THE SPECTRUM OF A FUNCTION lELD AND WE
CONSIDER ONLY INTEGRABLE ADMISSIBLE CONNECTIONS R % % i j8 F$G WITH
$ )N SECTIONS AND WE WILL SEE MANY IMPORTNAT EXAMPLES &OURIER
TRANSFORMS +LOOSTERMAN SHEAVES OF ADMISSIBLE CONNECTIONS "Y ABUSE OF
NOTATION WE WRITE
b
_ `
($23 % (b 8 % % i $
6)3)/.3 ). -!4( )22%'5,!2 #/..%#4)/.3
b %J
7E ASSUME 0ROP THIS GROUP COINCIDES WITH ($2 8_$ 4HE ISO
b
MORPHISM CLASS OF THE 'AUy[-ANIN CONNECTION ON THE + LINE DET ($23 %
IS DETERMINED BY AN ELEMENT
b
DET ($23 % j+K D LOG+ a
WHICH WE WOULD LIKE TO CALCULATE
3UPPOSE lRST THAT % IS A LINE BUNDLE 4WISTING % BY /m FOR SOME
DIVISOR m SUPPORTED ON THE IRREGULAR PART OF THE DIVISOR $ WE MAY ASSUME
DEG % )N THIS CASE THE RESULT THE MAIN THEOREM IN ;"%= IS THE
FOLLOWING 3INCE % HAS RANK R83$ %$ w %$ i $ CAN BE INTERPRETED
AS A SECTION OF $ WHICH GENERATES THIS SHEAF AS AN /$ MODULE 4HE EXACT
SEQUENCE
$ $
YIELDS AN ELEMENT R83$ ( 8 w + WHICH IS KNOWN TO EQUAL
DEG % 4HUS WE CAN lND SOME S ( 8 $ LIFTING R83$ 7E
WRITE S FOR THE DIVISOR OF S AS A SECTION OF $ SO S IS DISJOINT FROM $
4HEN THE RESULT IS
_ `
DET ( b % w _Fb S ` %
$23
7HEN S IS A DISJOINT UNION OF + POINTS THE NOTATION ON THE RIGHT SIMPLY
MEANS TO RESTRICT % WITH ITS ABSOLUTE CONNECTION TO EACH OF THE POINTS
AND THEN TENSOR THE RESULTING + LINES WITH CONNECTION TOGETHER .OTICE
THAT UNLIKE THE CLASSICAL 2IEMANN 2OCH SITUATION EG THE DIVISOR S
DEPENDS ON % R83
!NOTHER WAY OF THINKING ABOUT WILL BE IMPORTANT WHEN WE CONSIDER
PERIODS )T TURNS OUT THAT THE CONNECTION % R PULLS BACK FROM A RANK
CONNECTION % R% ON THE RELATIVE 0ICARD SCHEME 0IC8 $ WHOSE POINTS
ARE ISOMORPHISM CLASSES OF LINE BUNDLES ON 8 WITH TRIVIALIZATIONS ALONG $
4HE PAIR $ R8$$ DETERMINE A POINT T 0IC8 $ + AND IS
EQUIVALENT TO
DET ( b % w
$23 _% R% JT
a
,ET $y $ BE THE SUBSET OF ELEMENTS GENERATING $ AS AN /$ MODULE
,ET $ ( 8 + $ElNE " ^ a < _ /NE HAS A
$
a
NATURAL ACTION OF + a AND THE QUOTIENT $ + a IS IDENTIlED WITH ISOMOR
PHISM CLASSES OF TRIVIALIZATIONS OF $ J$ AND HENCE WITH A SUBVARIETY OF
0IC8 $ /NE HAS
^ a y a + a y 0IC8 $
T " "+
$
3 ",/#( !.$ ( %3.!5,4 '!&!
4HE RELATION BETWEEN T " % IS THE FOLLOWING %J" w

/" SO THE CONNECTION
R$ J" IS DETERMINED BY A GLOBAL FORM d 4HEN
dT j"+ i +T
)NDEED T IS THE UNIQUE POINT ON " WHERE THE RELATIVE FORM d+ VANISHES
CF ;"% ,EMMA =
.OW SUPPOSE THE RANK OF % IS 7E WILL SEE WHEN WE CONSIDER
EXAMPLES IN THE NEXT SECTION THAT DET ($23b
% DEPENDS ON MORE THAN
JUST THE CONNECTION ON DET % 2EMARK 4HUS IT IS HARD TO IMAGINE A
SIMPLE FORMULA LIKE )NDEED THERE IS NO OBVIOUS WAY OTHER THAN BY
TAKING THE DETERMINANT TO GET RANK CONNECTIONS ON 8 FROM % 4HE TRULY
SURPRISING THING IS THAT IF WE REWRITE ALGEBRAICALLY WE lND A FORMULA
WHICH DOES ADMIT A PLAUSIBLE GENERALIZATION 7E SUMMARIZE THE RESULTS
OMITTING PROOFS WHICH ARE GIVEN IN DETAIL IN ;"%= &OR EACH XI $
CHOOSE A LOCAL SECTION SI OF j8 F$G WHOSE IMAGE IN $ GENERATES AT XI
7RITE THE LOCAL CONNECTION MATRIX IN THE FORM
pI
!I GI SI MI _
ZI
WITH ZI A LOCAL COORDINATE AND pI F b j3 ,ET S BE A MEROMORPHIC SECTION

OF $ WHICH IS CONGRUENT TO THE FSI G MODULO $ $ElNE
_ `
C $ FSI G S 0IC8 $
!S IN WE CAN DElNE
_ `
Fb C $ FSI G ` DET% R j+ D LOG + a
7E FURTHER DElNE
g h _ ` 8
C $ R Fb C $ FSI G ` DET% R _ RES 4RDGI GI_ !I
I
(ERE RES REFERS TO THE MAP
TRANSFER
j8 F$G j+ i $ j+ i $ _ j+
#ONJECTURE ,ET R % % i j8 F$G
BE AN ADMISSIBLE CONNECTION
AS IN $ElNITION 4HEN
b
g h
DET ($23 8 _ $ % _ C $ R j+ D LOG+ a i: 1
/UR MAIN OBJECTIVE HERE IS TO PROVIDE EVIDENCE FOR THIS CONJECTURE /F
COURSE ONE SURPRISING FACT IS THAT THE RIGHT HAND SIDE IS INDEPENDENT OF
CHOICE OF GAUGE ETC !GAIN THE PROOF IS GIVEN IN DETAIL IN ;"%= BUT WE
REPRODUCE TWO BASIC LEMMAS 4HERE ARE FUNCTION LINEAR MAPS
_ `
R8$ %$ %$ i j8 F$Gj8 R83$ %$ %$ i $
6)3)/.3 ). -!4( )22%'5,!2 #/..%#4)/.3
AND IT MAKES SENSE TO CONSIDER THE COMMUTATOR

_ `
;R8$ R83$ = %$ %$ i j8 F$Gj8 i $
,EMMA ;R8$ R83$ =
0ROOF 7ITH NOTATION AS IN WE TAKE SI DZI ZIMI WHERE ZI IS A LOCAL
COORDINATE )NTEGRABILITY IMPLIES
o p
pI
D!I DGI > ZDZ
MI D
I
MI _
DZI
!I ;pI GI = MI _
n
I ZI ZI
WITH n j+ i +8 -ULTIPLYING THROUGH BY ZIMI

WE CONCLUDE THAT
;pI ZIMI _ GI = IS REGULAR ON $ WHICH IS EQUIVALENT TO THE ASSERTION OF THE
LEMMA c
4HE OTHER LEMMA WHICH WILL BE USEFUL IN EVALUATING THE RIGHT HAND TERM
IN IS
,EMMA 7E CONSIDER THE SITUATION FROM AND AT A lXED
XI $ &OR SIMPLICITY WE DROP THE I FROM THE NOTATION !SSUME DS
4HEN _ p `
RES 4RDGG _ ! RES 4R DGG _ ZM_
0ROOF 7E MUST SHOW RES 4RDGS 5SING AND 4R;G p= WE
REDUCE TO SHOWING RES 4RDpZ _M j+ 3INCE p F b j+ WE MAY DO
THE COMPUTATION FORMALLY LOCALLY AND REPLACE D BY DZ 4HE DESIRED VANISHING
FOLLOWS BECAUSE AN EXACT FORM HAS NO RESIDUES c
7ITH ,EMMA WE CAN FORMULATE THE CONJECTURE IN A MORE INVARIANT
WAY IN TERMS OF AN !$ COCYCLE ON 8 2ECALL ;%=
D LOG
!$ 8 ( 8 + _ j8 w
( 8 j8 D LOG +
4HE !$ GROUPS ARE THE CONES OF CYCLE MAPS FROM #HOW GROUPS TO (ODGE
COHOMOLOGY AND AS SUCH THEY CARRY CLASSES FOR BUNDLES WITH CONNECTIONS
4HERE IS A GENERAL TRACE FORMALISM FOR THE !$ GROUPS BUT IN THIS SIMPLE
CASE THE READER CAN EASILY DEDUCE FROM THE RIGHT HAND ISOMORPHISM IN
A TRACE MAP
Fb !$ 8 !$ 3 j+ D LOG + a
7HEN THE CONNECTION R HAS NO POLES OR MORE GENERALLY WHEN IT HAS REGULAR
SINGULAR POINTS IT IS POSSIBLE TO DElNE A CLASS
n C ` C % R !$ 8

WITH Fb n ;DET ($23 8 % = 2EMARKABLY THOUGH IT NO LONGER HAS THE
PRODUCT DESCRIPTION ONE CAN ASSOCIATE SUCH A CLASS TO ANY ADMISSIBLE
CONNECTION
3 ",/#( !.$ ( %3.!5,4 '!&!
&IX THE DIVISOR $ AND CONSIDER TUPLES F% R , uG WHERE % R IS AN

ADMISSIBLE ABSOLUTE CONNECTION , IS A LINE BUNDLE ON 8 AND u %$ w
%$ i ,$ 7E REQUIRE
_ `
;u R8$ = %$ %$ i ,$ i j8 F$Gj8
/F COURSE THE EXAMPLE WE HAVE IN MIND USING ,EMMA IS
g h
F% RG % R $ R83$
4O SUCH A TUPLE SATISFYING WE ASSOCIATE A CLASS n% R , u
a
!$ 8 AS FOLLOWS #HOOSE COCHAINS CIJ ',R /8 FOR % tIJ /8

FOR , uI ',R /$ FOR u AND I - R a R j8 F$G FOR R #HOOSE
LOCAL LIFTINGS uÎ ',R /8 FOR THE uI
0ROPOSITION 4HE #ECH HYPERCOCHAIN
_ `
FtIJ DETCJK G D LOG tIJ > 4RJ 4R_DuÎ uÎ _ > I
REPRESENTS A CLASS
_ `
n% R , u ( 8 + j8 F$G j8 F$Gj8 w !$ 8
4HIS CLASS IS WELL DElNED INDEPENDENT OF THE VARIOUS CHOICES 7RITING
n% R n% R $ R83$ WE HAVE
g h
Fb n% R C $ R
WHERE THE RIGHT HAND SIDE IS DElNED IN
0ROOF !GAIN THE PROOF IS GIVEN IN DETAIL IN ;"%= AND WE OMIT IT c
!S A CONSEQUENCE WE CAN RESTATE THE MAIN CONJECTURE

#ONJECTURE ,ET R BE AN INTEGRABLE ADMISSIBLE ABSOLUTE CONNEC
TION AS ABOVE 4HEN
b
DET ($23 % R _Fb n% R
4O lNISH THIS SECTION WE WOULD LIKE TO SHOW THAT BEHIND THE QUITE TECH
NICAL COCYLE WRITTEN IN 0ROPOSITION THERE IS AN ALGEBRAIC GROUP PLAYING
A ROLE SIMILAR TO 0IC8 $ IN THE RANK CASE ,ET ' BE THE ALGEBRAIC GROUP
WHOSE + POINTS ARE ISOMORPHISM CLASSES , u WHERE , IS AN INVERTIBLE
SHEAF AND u %$ %$ i ,$ IS AN ISOMORPHISM COMMUTING WITH R8$
)T IS ENDOWED WITH A SURJECTIVE MAP Q ' 0IC8 !S NOTED ' CONTAINS
THE SPECIAL POINT $ R83$ 4HE COCYLE OF 0ROPOSITION DElNES A
CLASS IN ( 8 a+ ' + j8a' F$ a 'G j8a' F$ a 'Gj8a' 4AK
ING ITS TRACE ONE OBTAINS A CLASS IN ,% R% !$ ' THAT
IS A RANK ONE CONNECTION ON ' 4HEN Fb n% R IS SIMPLY THE RESTRICTION OF
,% R% TO THE SPECIAL POINT $ R83$
6)3)/.3 ). -!4( )22%'5,!2 #/..%#4)/.3
.OW WE WANT TO SHOW THAT THIS SPECIAL POINT AS IN THE RANK CASE HAS
A VERY SPECIAL MEANING "Y ANALOGY WITH WE DElNE
_ RES 4R `
^ +ER(OM%$ %$ i $ __
" _ + < )SOM%$ %$ i $

^
" "+ a
y '
7E OBSERVE THAT ,EMMA SHOWS THAT G " #HOOSING A LOCAL TRIVIAL
IZATION $ w /$ ZDZ
M AND A LOCAL TRIVIALIZATION OF %$ WE WRITE R83 J$ AS
DZ
A MATRIX G ZM WITH G ',R /$ b IS THEN IDENTIlED WITH A TRANSLATION
INVARIANT FORM ON THE RESTRICTION OF SCALARS 2ES$+ ',R
_ ` _ `
b ,% R'+ % RES 4R Duu_ G ZDZM
4HE ASSUMPTION THAT G " IMPLIES THAT b DESCENDS TO AN INVARIANT FORM
ON 2ES$+ ',R 'M z ' WHERE ' F/ u 'G "Y INVARIANCE IT
GIVES RISE TO A FORM ON THE ' TORSOR '$ F$ u 'G 7E HAVE
" y '$ y '
,ET 3 y ' BE THE SUBGROUP OF POINTS STABILIZING "
0ROPOSITION bJ" VANISHES AT A POINT T " IF AND ONLY IF T LIES IN
THE ORBIT G ` 3
0ROOF 7RITE TH G 7RITE THE UNIVERSAL ELEMENT IN 2ES$+ ',R AS A
0M_ K
MATRIX 8 K 8IJ IJ Z K 4HE ASSERTION THAT bJ" VANISHES IN THE lBRE
AT T MEANS
o8_ ` p o8 p
K M_
RES 4R D8IJ IJ Z K H ZDZM T A D8II T
K I
FOR SOME A + .OTE THIS IS AN IDENTITY OF THE FORM
8 K K K
CIJ D8IJ j'+ i +T CIJ +
7E lRST CLAIM THAT IN FACT THIS IDENTITY HOLDS ALREADY IN j'+ 4O SEE THIS
K
WRITE ' 2ES$+ ',R .OTE j' IS A FREE MODULE ON GENERATORS D8IJ
!LSO ' y ' IS DElNED BY THE EQUATIONS
s M_
8 K M_
8 K t
8IJ IJ Z K GIJ IJ ZK
K K
s M_
8 M_
8 t
K K
8IJ IJ Z K pIJ IJ Z K
K K
3 ",/#( !.$ ( %3.!5,4 '!&!
WHICH ARE OF THE FORM

8 K K K
BIJP 8IJ P - BIJP +
IJK
K
THAT IS ARE LINEAR EQUATIONS IN THE 8IJ WITH + COEnCIENTS 4HUS WE HAVE
AN EXACT SEQUENCE
. ? j'+ i /' j'+

K
WHERE . ? IS GENERATED BY + LINEAR COMBINATIONS OF THE D8IJ 7E HAVE
THEREFORE A REDUCTION OF STRUCTURE OF THE SEQUENCE FROM /' TO +
AND THEREFORE j'+ w j i+ /' WHERE j y j'+ IS THE + SPAN OF
K K
THE D8IJ (ENCE ANY + LINEAR IDENTITY AMONG THE D8IJ WHICH HOLDS AT A
POINT ON ' HOLDS EVERYWHERE ON ' !S A CONSEQUENCE WE CAN INTEGRATE TO
AN IDENTITY
_ ` o8 p
M_
RES 4R 8H ZDZM A 8II s
I
WITH s + )F WE SPECIALIZE 8 T WE lND

_ ` _ `
RES 4R G ZDZM A ` RES 4R T ZDZM s s
7E CONCLUDE FROM AND THAT H 3 c
4HE &OURIER 4RANSFORM

)N THIS SECTION WE CALCULATE THE 'AUy[-ANIN DETERMINANT LINE FOR THE &OURIER
TRANSFORM OF A CONNECTION
0 ON 0 _ $ AND SHOW THAT IT SATISlES THE #ONJEC
TURE ,ET $ Mj j BE AN EkECTIVE K DIVISOR ON 0K ,ET % gR / BE
A RANK R FREE BUNDLE ON 0K AND LET i % % i $ BE A K CONNECTION
ON % ,ET , d DENOTE THE RANK CONNECTION ON 0 a 0 WITH POLES ON
F G a F G GIVEN BY , /0 a0 AND d DZT (ERE Z T ARE THE
COORDINATES ON THE TWO COPIES OF THE PROJECTIVE LINE ,ET + KT 7E
HAVE A DIAGRAM
0Z _ $ a 0T 0Z _ $ a 3PEC+

YP YP
0Z _ $ 3PEC+
6)3)/.3 ). -!4( )22%'5,!2 #/..%#4)/.3
4HE 'AUy[-ANIN DETERMINANT OF THE &OURIER TRANSFORM IS GIVEN AT THE

GENERIC POINT BY
b
_ `
DET ($2+ 0Z _ $ + Pb % i i , d
b
_ `
DET ($2+ 0Z _ $ + % R
WITH % Pb % i ,J0Z _$ + AND R i i i d 7E HAVE THE FOLLOWING
EASY
2EMARK 7RITE
88 Mj
GIj DZ
i I
DG Z GM

_
Z M _
j
Z _ j
I
WHERE GIj - R a R K 4HEN
DZ ZDT
Ri T _ T

IS ADMISSIBLE IF AND ONLY IF EITHER
I M s AND GM j IS INVERTIBLE FOR ALL j OR
j
j
II M t GMj IS INVERTIBLE FOR ALL j AND GM IS INVERTIBLE
_
4HEOREM 4HE CONNECTION % R SATISlES #ONJECTURE

0ROOF 7E lRST CONSIDER THE CASE WHEN i HAS A POLE OF ORDER s AT INlNITY
SO THE GI IN ! BASIS FOR
o o8 pp
( 0+ % i Mj j
IS GIVEN BY
DZ
EJ i DZ EJ i s I s Mj s J s R
Z _ j I

0
($2+ AND ($2+ COKER( % ( % i Mj j
HAS BASIS
DZ
EJ i s I s Mj s J s R
Z _ j I
4O COMPUTE
0 THE 'AUy[-ANIN CONNECTION0 WE CONSIDER THE DIAGRAM HERE

$ Mj j AND $ $ _ $ Mj _ j
( % ( %

R R
Y 8 Y 83
A
_ ( %$ i j+ _ ( % i j0 F$G _ ( % i $ _

R R
Y 83 i Y 8
w

( %$ i $ i j+ _ ( % i j0 F$ $ G
3 ",/#( !.$ ( %3.!5,4 '!&!
/NE DEDUCES FROM THIS DIAGRAM THE 'AUy[-ANIN CONNECTION

R

($2+ % w '-
COKERR83 ___ ($2+ % i j+ W R8 A_ W

_ DZ
` DZ
7E MAY CHOOSE A_ EJ i Z_j I EJ i Z_j I SO BY
q r q r
DZ DZ ZDZ > DT
R'- EJ i R8 EJ i EJ i
Z _ j I Z _ j I Z _ j I T

)N ($2+ w
COKERR83 WE HAVE THE IDENTITY
EJ i DZ _TiEJ
7E CONCLUDE
q r
DZ
R'- EJ i
Z _ j I
o p
EJ i DZ
jE J i DZ
> DT
s I s Mj
Z_j I_
p Z_j
I T
o
_TiEJ jEJ i DZ DT
Z_j > T I
)N PARTICULAR THE DETERMINANT CONNECTION WHICH IS GIVEN BY 4RR'- CAN

NOW BE CALCULATED
8 8 Gj DT
RMj jDT
4RR'- T
_ 4R
T
j j
7E COMPARE THIS WITH THE CONJECTURED VALUE WHICH IS THE NEGATIVE OF
$ElNE
8 'Z
& Z Mj
_ S & Z DZ
j
Z _ j Z _ j Mj
/NE HAS
q y zr
DZ
C $ DZ '
Z _ j Mj
THE DIVISOR OF ZEROES OF ' 7E NEED TO COMPUTE DET % DET R J' 7E HAVE
89 9
'Z Z _ k Mk _ Z _ j Mj
j kj j
0 o8 p 0
Mj
_Z Mj j Fj J Mj G Z Mj _
.OTE THAT THE COEnCIENTS OF ' DO NOT INVOLVE T SO THE DZ PART OF THE
6)3)/.3 ). -!4( )22%'5,!2 #/..%#4)/.3
CONNECTION DIES ON ' AND WE GET

k 8
4RRJ' _ RZTDT k' _ RTDT
k
k
'k
o8 p
_ RTDT
Mj j Fj J Mj G
)T REMAINS TO EVALUATE THE CORRECTION TERMS RES 4RDGG _ ! OCCURRING IN

)N THE NOTATION OF _ pZ M_
` _ZDTT AND BY ,EMMA WE
_ _ ZDT
HAVE RES 4RDGG ! _ RES 4R DGG T #LEARLY THE ONLY CONTRIBUTION
COMES AT Z 4AKE U Z _ !T THE CONNECTION IS
q88 r
GIj UI DU DT
!_ I

_
j
_ Uj T U UT
I
7E REWRITE THIS IN THE FORM ! GS Up AS IN WITH S AS IN
AND p _DTT 7E lND
0 GIj UI _
Ij _Uj I T s
G 0 UMj
j _Uj Mj _ V
DElNING s AND V TO BE THE NUMERATOR AND DENOMINATOR RESPECTIVELY
4HEN
_ `
RES 4RDGG_ ! _ RES 4R DGG _ UT
DT

_ `
_ RES 4RDss_ U_ R ` RES 4RDVV _ U_ DT
T
o 8 p
_T 4RGj _ RFj J Mj G DTT

#OMBINING AND WE CONCLUDE
_ `
4RR'- _ 4RRJ' _ RES 4RDGG _ !
WHICH IS THE DESIRED FORMULA
7E TURN NOW TO THE CASE WHERE i HAS A POLE OF ORDER t AT INlNITY
7E WRITE
88 Mj
GIj DZ
i I
G DZ G G GM

Z M _
j
Z _ j
I
R i DZ ZDT
T _ T

0
! BASIS FOR `0 Mj j M IS GIVEN BY
DZ
EJ i s I s Mj EJ i Z I DZ s I s M _
Z _ j I
3 ",/#( !.$ ( %3.!5,4 '!&!
! BASIS FOR THE 'AUy[-ANIN BUNDLE IS GIVEN BY OMITTING EJ i Z M _ DZ !S

IN [ THE 'AUy[-ANIN CONNECTION IS
DT
W ZW > T

4O COMPUTE THE TRACE NOTE THAT IN ($2+ WE HAVE IF M t
q8
DZ
EJ i Z M _ DZ _GM
_
GIj EJ i

Z _ j I
Ij
r
G EJ GM

_
EJ Z M _
T_
E J i DZ
)F M
q8 r
DZ
EJ i DZ _G T _ _
GIj EJ i
Z _ j I
Ij
)T FOLLOWS THAT
_ 0 ` DT
_
j RMj j _ 4RGM GM _ T M t
4RR'- _ 0 0
`
_ _ j DT
j RMj j _ j 4RG T G T M
4O COMPUTE THE RIGHT HAND SIDE IN #ONJECTURE WE TAKE AS TRIVIALIZING

SECTION
8 DZ 'Z DZ
S Mj
_ Z M _ DZ 1 Mj

j
Z _ j j Z _ j
WHERE
89 9
'Z Z _ k Mk _ Z M _ Z _ j Mj
j kj j
7E HAVE S ' THE DIVISOR1OF ZEROES OF THE POLYNOMIAL ' !GAIN '
DOES NOT INVOLVE T SO IF ' _ Z _ AK WE HAVE
RAP DT
4RRJZAP _ T

4HUS
0
_ 0 RMj jR`Fj
T
J Mj G DT
M
4RRJS RMj jDT
_ T
M t
&INALLY WE HAVE TO DEAL WITH THE CORRECTION TERM RES 4RDGG_ _ZDTT
4HERE IS NO CONTRIBUTION EXCEPT AT 7E PUT U Z _ AS BEFORE AND
6)3)/.3 ). -!4( )22%'5,!2 #/..%#4)/.3
G sV WITH
Mj M _I j
88 U G
s I
G T_ UM _ GM

_ jU I
j I
8 UM _Mj
V _
j
_ jU Mj
!SSUME lRST M 4HEN
_ `
RES 4RDGG _ ! _ RES 4R DGG _ UT
DT

_ RES 4RDZ ss_ U_ DT _ _ DT

T R ` RES 4RDZ VV U T
o 8 p
_4R G T_ _ Gj DT
T
_ R ` Fj J Mj G DT
T

)N THE CASE M t WE lND
_ _ `
RES 4RDGG _ ! _4R GM
GM _ DT
T
4HE THEOREM FOLLOWS BY COMPARING AND c
2EMARK 4HE PRESENCE OF NONLINEAR TERMS IN THE GIj IN MEANS
b
THAT THE CONNECTION ON DET ($2+ % IS NOT DETERMINED BY DET % ALONE
+LOOSTERMAN 3HEAVES
)N THIS SECTION WE SHOW THAT THE MAIN CONJECTURE HOLDS AT LEAST UP TO
TORSION FOR THE BASIC RANK +LOOSTERMAN SHEAF ;+= 4HE BASE lELD K IS #
&IX A B + a IN FACT ONE CAN WORK OVER + #;A B A_ B_ = ,ET
j k # _ : AND ASSUME ALSO j _ k # _ : #ONSIDER TWO CONNECTIONS
,I RI ON THE TRIVIAL BUNDLE ON 'M GIVEN BY jD LOGT DAT AND
kD LOGU DBU WHERE T U ARE THE STANDARD PARAMETERS ON TWO COPIES
OF 'M ,ET 8 'M a 'M AND CONSIDER THE EXTERIOR TENSOR PRODUCT
CONNECTION ON 8
, , b, /8 R R jD LOGT kD LOGU DATBU
.OTE ALL THE ABOVE ARE INTEGRABLE ABSOLUTE CONNECTIONS
0ROPOSITION
I , + w + I
($2 I
I

($2 , + i ( , + P
P
($2 ,+ w
$2
P
3 ",/#( !.$ ( %3.!5,4 '!&!
0ROOF 4HE DE 2HAM COMPLEX OF GLOBAL SECTIONS ON 8 FOR , R IS THE
TENSOR PRODUCT OF THE CORRESPONDING COMPLEXES FOR THE RI SO FOLLOWS
R+
FROM &OR WE HAVE EG THE COMPLEX OF GLOBAL SECTIONS / _

jD LOGT ADT )N ($2 THIS GIVES FOR ALL N :
ATN DT p _j N TN_ DT
!SSERTION FOLLOWS EASILY c
7E NOW COMPUTE 'AUy[-ANIN 7E WILL ABUSIVELY USE SHEAF NOTATION

WHEN WE MEAN GLOBAL SECTIONS OVER 'M OR 8 !LSO WE WRITE R FOR EITHER
ONE OF THE RI OR THE EXTERIOR TENSOR CONNECTION R+ IS THE CORRESPONDING
RELATIVE CONNECTION /NE HAS THE DIAGRAM
/ /

R
Y R Y +
z
_
/ i j+ ___ j ___

R i
Y + YR
r
i j+ ___ j &
w

(ERE z IS THE OBVIOUS FUNCTION LINEAR MAP EG FOR R R zTN DT TN DT
AND & y j IS THE SUBGROUP OF FORMS COMING FROM THE BASE 4HIS LEADS
TO THE 'AUy[-ANIN DIAGRAM
R_zR
+
/ _____ / i j+

R R i
Y + Y +
_r_ Rz
_____ i j+
&OR EXAMPLE WHEN R R WE GET ON ($2
_ `
R'- TN DT _r_ R > TN DT _r_ TDA > TN DT TN DT i DA
_j
3INCE TDT p A DT WE GET
R'- DT _j DT i D LOGA p _jDT i D LOGA MOD D LOG+ a
, ( , i( ,
/N THE RANK TENSOR PRODUCT CONNECTION ($2 $2 $2
THE 'AUy[-ANIN DETERMINANT CONNECTION IS THEREFORE
_ jD LOGA _ kD LOGB
.OTE THAT WE COMPUTED THIS DETERMINANT HERE BY HAND BUT WE COULD
HAVE AS WELL APPLIED DIRECTLY 4HEOREM OF ;"%= THE DETERMINANT OF
6)3)/.3 ). -!4( )22%'5,!2 #/..%#4)/.3
, IS JUST THE RESTRICTION OF R TO THE DIVISOR OF 0 DElNED BY THE

($2
TRIVIALIZING SECTION jD LOGT ADT OF THAT IS BY AT j 4HUS
THE DETERMINANT IS _jD LOGA j+ D LOG + a AND SIMILARLY FOR ($2 ,

THE DETERMINANT IS _kD LOGB j+ D LOG + a
4HE IDEA NOW IS TO RECALCULATE THAT DETERMINANT CONNECTION USING THE

,ERAY SPECTRAL SEQUENCE FOR THE MAP x 8 'M xT U TU 7RITE V
FOR THE COORDINATE ON THE BASE SO xb V TU
0ROPOSITION 7E HAVE
I
2I xb$2 ,
RANK BUNDLE ON 'M I
0ROOF ,ET Rx BE THE RELATIVE CONNECTION ON , WITH RESPECT TO THE MAP x
AND TAKE T TO BE THE lBRE COORDINATE FOR x 7RITE U VT 4HEN
Rx jD LOGT kD LOGU ADT BDU j _ k D LOGT ADT _ BV DT
T
SO IN 2 xb$2 WE HAVE
p Rx TN j _ k N TN_ DT ATN DT _ BVTN_ DT
)T FOLLOWS THAT 2 xb$2 , HAS RANK GENERATED EG BY DT AND DTT AND
THE OTHER 2I AS CLAIMED c
$ElNE
% 2 xb$2 , R R'- % % i j'M
b 'M % SATISlES
4HEOREM 4HE 'AUy[-ANIN CONNECTION ON ($2+
#ONJECTURE UP TO TORSION
2EMARK )N FACT WE WILL SEE THAT R IS NOT ADMISSIBLE IN THE SENSE
OF $ElNITION BUT ITS INVERSE IMAGE VIA A DEGREE COVERING IS 3INCE
THE NEW DETERMINANT OF DE 2HAM COHOMOLOGY OBTAINED IN THIS WAY IS TWICE
THE OLD ONE WE LOSE CONTROL OF THE TORSION 7E DO NOT KNOW WHETHER THE
CONJECTURE HOLDS EXACTLY IN THIS CASE OR NOT
0ROOF 7E CAN NOW CALCULATE THE CONNECTION R R'- ON % JUST AS BEFORE
7E HAVE THE 'AUSy[-ANIN DIAGRAM zTN DT TN DT
R_zR
x
/8 _____ /8 i j'M

R R i
Y x Y x
_r_ Rz
x _____ x i j'M
(ERE
_ _
R j _ k DTT k DV
V ADT TDA T DBV BVDT
3 ",/#( !.$ ( %3.!5,4 '!&!
7E GET
_ DT ` _ `
R'- T _r R > DTT DTT i k DV
V DT i DA
DT
T
i DBV
_
__ DV _
` `
R'- DT _r k V TDA T DBV > DT
DT i k DV
V TDT i DA
DT
T i DBV
7E CAN NOW SUBSTITUTE
_ `
DT
T p BV _ j _ k DTT ADT
BV DT j _ k
TDT p _ DT
A T A
GETTING lNALLY
_ ` _ ` _ `
R'- DTT DT
i j _ k DB
T
DV DB DV
B j V DT i DA A B A V
_ ` DT _ DA `
R'- DT DT i k DV DA
V _ j _ k A T i BV A DBV
&OR CONVENIENCE DElNE
q DA DB DV
A B V
2EPRESENTING AN ELEMENT IN OUR RANK TWO BUNDLE AS A COLUMN VECTOR
RS RDT S DTT
THE MATRIX FOR THE CONNECTION ON % BECOMES
q r
kq _ j DA
A _ k DB
B Aq
! DB
BVq jq _ j DA
A _k B
4HE CORRESPONDING CONNECTION HAS A REGULAR SINGULAR POINT AT V AND
AN IRREGULAR ONE AT V %XTENDING % TO / ON 0+ WE CAN TAKE $
BUT THE MATRIX G IS NOT INVERTIBLE AT )N ORDER TO P
REMEDY THIS
MAKE THE BASE CHANGE Z _ V AND ADJOIN TO + THE ELEMENT AB .OTICE
THE BASE CHANGE MODIlES THE 'AUy[-ANIN DETERMINANT COMPUTATION ,ET
US IGNORE THIS FOR
P A WHILE AND CONTINUE WITH THE DETERMINANT CALCULATION
$ElNE l ABZ -AKE THE CHANGE OF BASIS
!NEW
q r
_Z
B k _ l kq _ j DA DB
A _k B Aq
Z B
jq _ j DA DB
B Z q A _k B
q r
q r
k _ l _ZB k _ l _Dl
a B
Z Z
B
D ZB
4HIS WORKS OUT TO
q r
lq _ j DA DB
A _k B j k lq _ kj q
!NEW q
j k q _ lq _ j DA DB
A _k B
6)3)/.3 ). -!4( )22%'5,!2 #/..%#4)/.3
DA DB
(ERE OF COURSE q A B _ DZ
Z
.OTE
4R!NEW j k q _ j DA DB
A _ k B
p k _ j DA DB DZ
A j _ k B _ j k Z MOD D LOG+ a
!T Z THE POLAR PART OF !NEW LOOKS LIKE
P P
AB DA DB DZ AB DA DB DZ DZ
Z A B _ Z k j Z A B _ Z kj Z
_DZ
P
_ AB DA DB DZ DZ

Z Z A B _ Z _ j k Z
7RITING !POL G DZ
Z
pZ THE MATRIX FOR G WITH COEnCIENTS IN #;Z=Z
IS q P P r
_ AB _j Pk AB kj Z
G
_Z AB _ j k Z
!LSO qP _ DA DB ` P _ DA DB `r
AB A B j kP AB A` B
p _
_ AB DA A B
DB
7ITH RESPECT TO THE TRIVIALIZATION D LOGZ_ THE MATRIX G AT Z IS

q r
kj
G
_ _j k
.OTICE THAT THE MATRICES FOR G ARE INVERTIBLE BOTH AT AND 7RITING
!POL G DZ _
Z p THE CONTRIBUTION RES 4R DG G p IS OF COURSE VANISH
ING AS WELL AS THE CONTRIBUTION AT OBTAINED BY CHANGING THE TRIVIALIZATION
DZZ TO UNIT ` DZZ !T WE GET v G G MOD Z
_ _ p
` _ ` _ `
RES 4R DGG Z RES 4R Dv G Gv_ pZ DA DB
A B a
q rq r_ q r
kj _ _k j k j
4R
_ _j k
q r
_ ` DA DB
DA DB
A B k j p j k MOD D LOG+ a
A B
.OW WE COMPARE WITH THE CONJECTURAL FORMULA
_ `
DET($2b
% _ C $ S ` DETR _ RES 4R DGG _ pZ
(ERE S CAN BE TAKEN TO BE THE DIVISOR DElNED BY THE TRIVIALIZING SECTION
DZ
Z
_ DZ
Z OF THE SHEAF ` THAT IS Z 4HUS ONE HAS
4R!NEW k _ j DA DB
A j _ k B
&URTHER WE HAVE TO WRITE
DZ
S Z
W
3 ",/#( !.$ ( %3.!5,4 '!&!
a p
WHERE W _ Z /8 3INCE 4Rp RES 4R DW
W Z AS WELL THUS
THE LOCAL CONTRIBUTION AT IS GIVEN BY
4HE CONJECTURE GIVES WRITING m 0 0 X X
b _ DA DB `
DET($2 mb % _ k _ j DA DB
A j _ k B _ j k A B
_j DA DB
A _ k B
P
.OTICE WE HAVE ADJOINED AB TO + SO WE HAVE LOST SOME TORSION
"EARING IN MIND THAT % 2 xb$2 WHICH INTRODUCES A MINUS SIGN IN THE
DETERMINANT CALCULATIONS AND COMPARING WITH OUR EARLIER CALCULATION
ABOVE WE lND THAT WHAT WE NEED TO lNISH IS
0ROPOSITION 4HE 'AUy[-ANIN DETERMINANT FOR DE 2HAM COHOMOLOGY
OF mb % IS TWICE THE CORRESPONDING DETERMINANT FOR %
0ROOF !GAIN WE USE SHEAF NOTATION FOR WORKING WITH MODULES 2ECALL FOR
THE PULLBACK WE SUBSTITUTED Z W V _ 7E CAN WRITE mb % % g Z%
7E HAVE
mb RZE ZRE ZE i _V
DV

SO WITH RESPECT TO THE ABOVE DECOMPOSITION WE CAN WRITE
_ `
mb % mb R % R g % R _ DV
V
4HE SECOND TERM ON THE RIGHT IS THE CONNECTION OBTAINED BY TENSORING %
2 xb$2 , b , WITH / _ DVV 5SING THE PROJECTION FORMULA AND THE
INVARIANCE OF THE LATTER CONNECTION THIS IS THE SAME AS THE CONNECTION ON
2 xb$2 , _ DTT b , _ DU
U IE IT AMOUNTS TO REPLACING j k BY

j _ k _ 5SING THIS CHANGES THE 'AUy[-ANIN DETERMINANT BY
P
D LOG AB WHICH IS TRIVIAL )T FOLLOWS THAT THE 'AUy[-ANIN DETERMINANT OF
mb % IS TWICE THAT OF % WHICH IS WHAT WE WANT c
4HIS CONCLUDES THE PROOF OF 4HEOREM c
0ERIODS
,ET 8# BE A SMOOTH COMPLETE CURVE 7E CONSIDER A CONNECTION RELATIVE
TO # R % % i 8 $ ,ET % BE THE CORRESPONDING LOCAL SYSTEM ON
8# _ $ .OTICE THAT WE DO NOT ASSUME R HAS REGULAR SINGULAR POINTS SO
% DOES NOT DETERMINE % R &OR EXAMPLE IT CAN HAPPEN THAT % IS A TRIVIAL
LOCAL SYSTEM EVEN THOUGH R IS HIGHLY NONTRIVIAL )N THIS SECTION WE CONSIDER
b 8 _ $ % 7E WORK WITH
THE QUESTION OF ASSOCIATING PERIODS TO DET ($2
ALGEBRAIC DE 2HAM COHOMOLOGY IN ORDER TO CAPTURE THE IRREGULAR STRUCTURE
6)3)/.3 ). -!4( )22%'5,!2 #/..%#4)/.3
4HE lRST REMARK IS THAT IT SHOULD BE POSSIBLE USING 3TOKES STRUCTURES ;-= TO
WRITE DOWN A HOMOLOGICAL DUAL GROUP ( 8 b % AND PERFECT PAIRINGS % ?
IS THE DUAL LOCAL SYSTEM
( 8 b % ? a ($2

8 _ $ % #
(ERE 8b IS SOME MODIlCATION OF THE 2IEMANN SURFACE 8 4HE POINT IS THAT
EG IN THE EXAMPLE WE GIVE BELOW THE DE 2HAM GROUP CAN BE LARGE WHILE
THE LOCAL SYSTEM % IS TRIVIAL AND 8 _ $ ! ,ET & y # BE A SUBlELD
AND ASSUME WE ARE GIVEN I AN & STRUCTURE ON % IE AN & LOCAL SYSTEM %&
AND AN IDENTIlCATION %& i # w % II ! TRIPLE 8 $ % DElNED OVER
& AND AN IDENTIlCATION OF THE EXTENSION TO # OF THESE DATA WITH 8 $ %
7HEN EG % R SATISlES THE CONDITION OF 0ROPOSITION ONE HAS
b
DET ($2 8 _ $ %
_ `_
w
# i& DET ( b 8 % i DET ( b 8 % i $
SO THE DETERMINANT OF DE 2HAM COHOMOLOGY GETS AN & STRUCTURE EVEN IF
R IS NOT NECESSARILY ITSELF DElNED OVER & /F COURSE I DETERMINES AN
& STRUCTURE ON (b 8 b % #HOOSING BASES FPJ G FpK G COMPATIBLE WITH
2 THE
& STRUCTURE AND TAKING THE DETERMINANT OF THE MATRIX OF PERIODS PJ pK
YIELDS AN INVARIANT
0ER% R %& #a & a
-ORE GENERALLY ONE CAN CONSIDER TWO SUBlELDS K & y # WITH A REDUCTION
OF % TO & AND A REDUCTION OF % TO K 4HE RESULTING DETERMINANT LIES IN
& a N #a Ka )N THE CASE OF REGULAR SINGULAR POINTS THESE DETERMINANTS
HAVE BEEN STUDIED IN ;34=
.OTICE THAT THE PERIOD INVARIANT DEPENDS ON THE CHOICE OF AN & STRUCTURE
ON THE LOCAL SYSTEM % KERRAN 7HEN % R ARE <MOTIVIC IE COME
FROM THE DE 2HAM COHOMOLOGY OF A FAMILY OF VARIETIES OVER 8 THE COR
RESPONDING LOCAL SYSTEM OF "ETTI COHOMOLOGY GIVES A NATURAL 1 STRUCTURE
ON % "Y A GENERAL THEOREM OF 'RInTHS THE CONNECTION R IN SUCH A CASE
NECESSARILY HAS REGULAR SINGULAR POINTS )N A NON GEOMETRIC SITUATION OR
EVEN WORSE IN THE IRREGULAR CASE THERE DOESNT SEEM TO BE ANY CANONICAL
SUCH 1 OR & STRUCTURE &OR EXAMPLE THE EQUATION F _ F HAS SOLUTION
SPACE # ` EX )S THE 1 REDUCTION 1 ` EX MORE NATURAL THAN 1 ` EX /F
COURSE IN CASES LIKE THIS WHERE THE MONODROMY IS TRIVIAL THE CHOICE OF %&
IS DETERMINED BY CHOOSING AN & POINT X 8 _ $ AND TAKING %& %X
%VEN IF THERE IS NO CANONICAL & STRUCTURE ON % ONE MAY STILL ASK FOR A
FORMULA ANALOGOUS TO #ONJECTURE FOR 0ER% R %& )N THIS lNAL SECTION
3 ",/#( !.$ ( %3.!5,4 '!&!
WE DISCUSS THE VERY SIMPLEST CASE

8 0 $ M ` % /8 R DF DAM_ XM_ A X

0ERIOD DETERMINANTS IN THIS CASE AND MORE GENERAL CONmUENT HYPERGEOMET
RIC CASES WERE COMPUTED BY A DIkERENT ARGUMENT IN ;4= 7E STRESS THAT
OUR OBJECTIVE HERE IS NOT JUST TO COMPUTE THE INTEGRAL BUT TO EXHIBIT THE
ANALOGY WITH FORMULA 7E WOULD LIKE ULTIMATELY TO lND A FORMULA FOR
PERIODS OF HIGHER RANK IRREGULAR CONNECTIONS WHICH BEARS SOME RELATION TO
#ONJECTURES AND 7E CONSIDER THE SITUATION WITH AM_
4HEN ($2 AND ( w ( 0 / M` HAS AS BASIS THE CLASSES
$2
OF Z DZ s I s M _ /NE HAS % ? # ` EXPF Z TRIVIAL LOCAL SYSTEM
I
SO WE TAKE THE OBVIOUS 1 STRUCTURE WITH BASIS EXPF 7E WANT TO COMPUTE

THE DETERMINANT OF THE PERIOD MATRIX
q: r
J_
EXPF Z Z DZ
zI IJM_
FOR CERTAIN CHAINS zI ON SOME 8 b ,ET 3 BE A UNION OF OPEN SECTORS ABOUT
INlNITY ON 0 WHERE 2EF IS POSITIVE IE 3 IS A UNION OF SECTORS OF THE
FORM HERE . | AND n { ARE lXED
y
k _ARGAM_ K _ _ n x
3K REIq k . R qK
M_
z
_ARGAM_ K n x

M_
SO 8 b 0 _ 3 w 0 _ FP PM_ G WHERE THE PK ARE DISTINCT POINTS
)N PARTICULAR ( 8 b :M_ $ElNE zK lK _ l WHERE
y z
k _ARGAM_ K x
k
lK R EXPIq s R q
M_
4HE zK s K s2 M _ FORM A BASIS FOR ( 8 b :
7RITE 0IJ zI EXPF Z Z J_ DZ
,EMMA 7E HAVE
DET0IJ sIJsM_
: 9
EXPF Z F ZM_ ZJ _ZI DZ > > DZM_
z a``àzM_ IJ
0ROOF 4HE ESSENTIAL POINT IS THE EXPANSION
9 8 A _ A _ AM_ _
ZJ _ ZI _ SGNA Z Z ` ` ` ZM_
IJ A
6)3)/.3 ). -!4( )22%'5,!2 #/..%#4)/.3
WHERE A RUNS THROUGH PERMUTATIONS OF F M _ G c

7E WILL EVALUATE BY STATIONARY PHASE CONSIDERATIONS PRECISELY PAR
ALLEL TO THE TECHNIQUES DESCRIBED IN SECTION AND ;"%= )NDEED THE DEGREE
M _ PART * M_ 0 M ` y *0 M ` OF THE GENERALIZED JACOBIAN
a TORSOR a
IS SIMPLY THE /M` M` OF TRIVIALIZATIONS OF M` M `
MODULO MULTIPLICATION BY A CONSTANT IN #a 7RITING U Z _ WE MAY
IDENTIFY THIS TORSOR WITH
g DU k h
B BM_ DUM
k BM_
U U
4HE QUOTIENT OF SUCH TRIVIALIZATIONS UP TO GLOBAL ISOMORPHISM IS
a
g h g h
M` #a SM_ DU DU DU
U S UM_ UM SM_ S

a #a BE DElNED BY S a
,ET " y M` M_ ,ET `0 M ` DENOTE
THE SPACE OF SECTIONS WHICH GENERATE M ` AT 7E HAVE
DIV _ à
!M_ p 3YMM_ ! w ` 0 M ` #a w a
" y M` #a

,ET Z ZM_ BE AS IN AND ADD AN EXTRA VARIABLE ZM_ 4AKE
SK Z ZM_ TO BE THE K TH ELEMENTARY SYMMETRIC FUNCTION SO EG
SM_ Z Z ` ` ` ZM_ 7E HAVE A COMMUTATIVE DIAGRAM
!M_ ___ "

Z _
YM Y
ZSM_ Z S Z
!M_ _____________ M`
a #a
.OTICE THAT
9
ZJ _ ZI DZ > > DZM_ DS > > DSM_
IJ
,ET PK Z ZM_ ZK ZM_

K BE THE K TH POWER SUM OR K TH
.EWTON CLASS $ElNE
& S SM_ F Z F ZM_ A P AM_ PM_

.OTICE THAT ALTHOUGH THE RIGHT HAND EXPRESSION MAKES SENSE ON ALL OF
a #a WE THINK OF & AS DElNED ONLY ON " S
M` M_ ,ET i BE
THE DIRECT IMAGE ON " OF THE CHAIN z a ` ` ` a zM_ ON !M_ 4HE INTEGRAL
BECOMES
:
_ `
EXP & S SM_ DS > > DSM_
i
3 ",/#( !.$ ( %3.!5,4 '!&!
,EMMA ,ET B " 3YMM_ ! CORRESPOND TO THE DIVISOR OF ZEROES
OF DF F DZ A A Z M _ AM_ Z M_ DZ 4HEN D& VANISHES
AT B AND AT NO OTHER POINT OF "
0ROOF 4HE DIkERENTIAL FORM
p A DP Z DZM_ AM_ DPM_ Z ZM_
a
ON M` #a IS TRANSLATION INVARIANT )NDEED TO SEE THIS WE MAY TRIVIALIZE
THE TORSOR AND TAKE THE POINT S SM_ TO BE THE IDENTITY
)NTRODUCING A FORMAL VARIABLE 4 WITH 4 M THE GROUP STRUCTURE IS THEN
GIVEN BY S g S S WITH
_ `_ `
_ S 4 _ M_ SM_ 4 M_ _ S 4 _ M_ SM_ 4 M_
_ `
_ S 4 _ M_ SM_ 4 M_
3INCE _ LOG _ S 4 _ M_ SM_ 4 M_ P 4 PM_ 4 M_
IT FOLLOWS THAT THE PI ARE ADDITIVE WHENCE p IS TRANSLATION INVARIANT
.OTE THAT D& pJ" $ElNE x ! " y M` a #a BY x b P Z K K s
K
M _ x SM_ 4HEN x p DF )N PARTICULAR x b p VANISHES AT THE
b b
ZEROES OF DF F DZ )T FOLLOWS THAT SINCE B F 3YMM_ ! WE HAVE

$& JB AS WELL 4HE PROOF THAT B IS THE UNIQUE POINT WHERE pJ" VANISHES
IS GIVEN IN ;"% ,EMMA = 7E SHALL OMIT IT HERE c
.OTE THAT
q r
A A M _ AM_
B B BM_
M _ AM_ M _ AM_ M _ AM_
IN THE S COORDINATE SYSTEM ON " 3ET TI SI _ BI AND WRITE & S
& B 'T SO 'T TM_ HAS NO CONSTANT OR LINEAR TERMS
,EMMA 4HERE EXISTS A NON LINEAR POLYNOMIAL CHANGE OF VARIABLES OF
THE FORM
TJ TJ "J T TJ_
SUCH THAT " AND 'T 1T WHERE 1 IS HOMOGENEOUS OF
DEGREE
0ROOF 4HE PROOF IS CLOSE TO ;"% ,EMMA = 7E WRITE ABUSIVELY PK S
FOR THE POWER SUM ZK ZM_ K TAKEN AS A FUNCTION OF THE ELEMENTARY SYM
METRIC FUNCTIONS S SK 4HE QUADRATIC MONOMIALS SI SM__I ALL OCCUR
WITH NONZERO COEnCIENT IN PM_ "Y CONSTRUCTION & S A P S
AM_ PM_ S WITH AM_ )F WE THINK OF SI AS HAVING WEIGHT I PK S IS
PURE OF WEIGHT K SO SI SM__I OCCURS WITH NONZERO COEnCIENT IN & S 3INCE
6)3)/.3 ). -!4( )22%'5,!2 #/..%#4)/.3
THE WEIGHT M _ IS MAXIMAL TI TM__I WILL OCCUR WITH NONZERO COEnCIENT

IN 'T AS WELL 4HUS WE HAVE
'T 1T TM_ (T
WHERE 1 IS QUADRATIC AND CONTAINS TI TM__I WITH NONZERO COEnCIENT AND
( HAS NO TERMS OF DEGREE &URTHER ( HAS NO TERMS OF WEIGHT M _
)N PARTICULAR THE VARIABLE TM_ DOES NOT OCCUR IN ( )F WE REPLACE TM_
BY TM_ TM_ !M_ T TM_ FOR A SUITABLE POLYNOMIAL !M_
WE CAN ELIMINATE T FROM ( COMPLETELY 'T 1T TM_ TM_
^ TM_ 4HE WEIGHT AND DEGREE CONDITIONS ON (
(T ^ ARE THE SAME AS
^
THOSE ON ( SO WE CONCLUDE THAT ( DOES NOT INVOLVE TM_ !LSO THIS CHANGE
DOES NOT AkECT THE MONOMIALS TI TM__I IN 1 FOR I t 4HUS WE MAY WRITE
' b T T ^ TM_ (T
bb T TM_ 1T ^ TM_
M_
SINCE bb WE MAY CONTINUE IN THIS FASHION WRITING TM_ TM_

!M_ T TM_ ETC c
4HE CONSTANT TERM CAN BE WRITTEN

8
& B F k
k
F k
4HE NONLINEAR CHANGE OF VARIABLES T T HAS JACOBIAN !LSO THE QUADRATIC

FORM ABOVE IS NECESSARILY NONDEGENERATE OTHERWISE & WOULD HAVE MORE THAT
ONE CRITICAL POINT /NE HAS
0ROPOSITION
9 :
DET0IJ sIJsM_ EXPF k EXP1T DT > > DTM_
b
k
F k
WHERE
1 T TM_
IS THE STANDARD NONDEGENERATE
2 QUADRIC ON " AND b IS SOME N _ CHAIN
ON " -OREOVER b IS DETERMINED UP TO 1a MULTIPLE ON PURELY GEOMETRIC
GROUNDS
3INCE THE SHAPE OF THE INTEGRAL IS OBVIOUSLY COMING FROM
2 THE CHANGE OF
COORDINATES T T WE HAVE TO UNDERSTAND THE MEANING OF b
,ET 7 y 0N a 0 BE THE FAMILY OF QUADRICS OVER 0 DElNED BY
5 13 3M_ _ 6 4
3 ",/#( !.$ ( %3.!5,4 '!&!
(ERE 3 3M_ 4 ARE HOMOGENEOUS COORDINATES ON 0M_ 5 6 ARE HO

MOGENEOUS COORDINATES ON 0 AND 13 3M_ IS A NONDEGENERATE
QUADRIC 7E HAVE 7EIL DIVISORS 9 5 4 : 1 4 IN 7 .OTE
9 AND : ARE SMOOTH AND 7SING 9 < : ,ET x 7 7 BE THE BLOWUP
OF 7 ALONG THE 7EIL DIVISOR 9 ,ET 9 y 7 BE THE EXCEPTIONAL DIVISOR
,EMMA I 7 IS SMOOTH
II 4HE STRICT TRANSFORM : OF : IN 7 IS ISOMORPHIC TO : AND
9 < : y 9SMOOTH

0ROOF ,ET 0 ",9 y 0N a 0 BE THE BLOWUP 4HEN 7 IS THE STRICT
TRANSFORM OF 7 IN 0 3INCE 9 < : IS THE #ARTIER DIVISOR 5 IN : IT
FOLLOWS THAT THE STRICT TRANSFORM OF : IN 0 OR 7 IS ISOMORPHIC TO :
7E CONSIDER THE STRUCTURE OF 7 LOCALLY AROUND THE EXCEPTIONAL DIVISOR
7E MAY ASSUME SOME 3I IS INVERTIBLE AND WRITE SJ 3J 3I T 4 3I U
56 Q 13I 4HE LOCAL , DElNING EQUATION FOR 7 IS UQS _ T
P
4HINKING OF 7 AS 0ROJ Pt ) /7 WITH ) U T WE HAVE OPEN SETS
5 T^ AND 5 U
^ 4HE TILDE INDICATES WE VIEW THESE AS PROJECTIVE
COORDINATES ON THE 0ROJ 7E HAVE THE FOLLOWING COORDINATES AND EQUATIONS
FOR 7 AND 9
5 U T U 7 U QS _ T 9 T

5 T U T 7 QS _ UT 9 U
4HE STRICT TRANSFORM : OF : LIES IN THE LOCUS T^ AND SO DOES NOT MEET 5
"OTH DElNING EQUATIONS FOR 7 ARE SMOOTH AND 9 IS SMOOTH ON 5 &INALLY
: < 9 Q T U IS ALSO SMOOTH c
7RITE
7 7 _ : 9 9 _ 9 < :
7E WANT TO SHOW THAT THE CHAINS OVER WHICH WE INTEGRATE CAN BE UNDER
STOOD AS CHAINS ON THE TOPOLOGICAL PAIR 7 _ 5 9 FOR SOME OPEN 5 CF
,EMMA BELOW )N Z COORDINATES WE DEAL WITH CHAINS lK WHICH
ARE PARAMETRIZED jK RK EIqK s RK FOR lXED qK .OTE THAT FOR R |
THE REAL PART OF F ON lK WILL _ "Y ABUSE OF NOTATION WE WRITE lJ ALSO
FOR THE CLOSURE OF THIS CHAIN ON 0 IE INCLUDING THE POINT R
7RITE & Z & B 13 3M_ )T IS EASY TO CHECK BY LOOKING AT
WEIGHTS THAT J3K J /JRJK AS JRJ /N THE OTHER HAND BECAUSE THE
PATHS ARE CHOSEN SO THE REAL PARTS OF AM_ jKM_ ARE ALL NEGATIVE WE lND THERE
EXIST POSITIVE CONSTANTS # # SUCH THAT #JRJM_ s J& Z J J13Z J s
# JRJM_ )N HOMOGENEOUS COORDINATES
3K 4 5 6
6)3)/.3 ). -!4( )22%'5,!2 #/..%#4)/.3
THE POINT ASSOCIATED TO A POINT WITH COORDINATES Z ON OUR CHAIN IS

3K 3K Z /JRJK s K s M _ 4

5 6 13Z t #JRJM_
7ITH REFERENCE TO THE COORDINATES IN WE SEE THAT
JUJ J56 J s # _ JRJ_M JTJ J4 3I J t # JRJ_I
JT J JTUJ t # JRJM__I
)N PARTICULAR THE LIMIT AS JRJ DOES NOT LIE ON 5 3INCE NEAR
: y 5 WE CONCLUDE OUR CHAINS STAY AWAY FROM : AT INlNITY
7E lX n { AND . | AND DElNE A CONNECTED SIMPLY CONNECTED
DOMAIN $ y ! y 0 BY
N k O
$ REIq k R . _ x _ n q x n
THUS $ IS AN OPEN SECTOR AT INlNITY AND EXPZ IS RAPIDLY DECREASING AS
JZJ IN THE COMPLEMENT OF $ )N WHAT FOLLOWS LET G 7 0 BE THE
PROJECTION
,EMMA 4HE ASSIGNMENT
:
_ `
l EXP 13 4 3M_ 4 D3 4 > > D3M_ 4
l
DElNES A FUNCTIONAL (M_ 7 _ G _ $ 9 1 #

0ROOF 7RITE { FOR THE ABOVE INTEGRAND ,ET - BE SOME NEIGHBORHOOD OF :
4HEN { IS RAPIDLY DECREASING ON 7 _ G_ $ NEAR 9 _ - < 9 WHERE
THE SIZE IS DElNED BY SOME METRIC ON THE HOLOMORPHIC M _ FORMS ON 7
3INCE THE CHAINS ARE COMPACT A CHAIN l ON 7 WILL BE SUPPORTED ON 7 _-
FOR A SUnCIENTLY SMALL NEIGHBORHOOD - OF : 4HUS INTEGRATION DElNES A
FUNCTIONAL _ `
#M_ 7 _ G _ $ 9 #
2
)T REMAINS TO SHOW ` { FOR AN M _ CHAIN ` ,ET - BE AN OPEN
NEIGHBORHOOD OF : NOT MEETING ` ,ET 2 BE AN OPEN NEIGHBORHOOD OF 9
IN 7 _2- 7RITE ` ` ` WHERE ` y 2 v AND ` < 2 3INCE { IS
CLOSED ` { /N THE OTHER HAND THE VOLUME OF ` CAN BE TAKEN TO
2
BE BOUNDED INDEPENDENT OF 2 )T FOLLOWS THAT ` { c
,EMMA (M_ 7 _ G _ $ 9 1 w

1
0ROOF ,ET P $ BE A POINT 7E lRST SHOW
_ `
(M_ 7 _ G _ P 9 1 w 1
3 ",/#( !.$ ( %3.!5,4 '!&!
7E CALCULATE ( b 7 _ G _ P USING THE ,ERAY SPECTRAL SEQUENCE &OR

X G _ X IS A SMOOTH AnNE QUADRIC OF DIMENSION M _ 3O
2P Gb 1 AWAY FROM FOR P M _ AND 2M_ Gb 1J0 _FG
IS A RANK LOCAL SYSTEM 4HE MONODROMY ABOUT AND IS INDUCED BY
3 4 3 _4 AND 3 4 _3 4 RESPECTIVELY "OTH ACTIONS GIVE _
ON THE lBRES )T FOLLOWS THAT WRITING J 0 _ F G 0 WE HAVE
w Jb 2M_ Gb 1J0 _FG
J 2M_ Gb 1J0 _FG w 2Jb 2M_ Gb 1J0 _FG

)T FOLLOWS THAT THE NATURAL MAP 2N_ Gb 1 Jb 2N_ Gb 1J0 _FG IS SUR
JECTIVE AND WE GET A DISTINGUISHED TRIANGLE IN THE DERIVED CATEGORY
0 2N_ Gb 1 2Jb 2N_ Gb 1J0 _FG
WHERE 0 IS A SHEAF SUPPORTED OVER )N PARTICULAR
_ `
w ( 0 _ F G 2N_ Gb 1
( 0 2M_ Gb 1
WHERE THE VANISHING COMES BY IDENTIFYING WITH GROUP COHOMOLOGY OF : ACT
ING ON 1 WITH THE GENERATOR ACTING BY _ !N EASY 'YSIN ARGUMENT YIELDS
_ ` _ `
( 0 _ FPG 2M_ Gb 1 w ( 0 _ F PG 2M_ Gb 1 1
)T ALSO FOLLOWS FROM THAT ( 0 2M_ Gb 1 4HE SPECTRAL
SEQUENCE THUS GIVES
( M_ 7 1 w
2M_ Gb 1 FG
4O COMPUTE THESE STALKS NOTE THE lBRE OF 7 0 OVER IS A SIN
GULAR QUADRIC WITH SINGULAR POINT 3 3M_ 4 AWAY
FROM : 4HUS THE lBRE OF G 7 0 OVER IS THE HOMOGENEOUS AnNE
QUADRIC 13 WHICH IS CONTRACTIBLE &URTHER BECAUSE : MEETS THE
lBRE OF 7 SMOOTHLY ONE HAS BASECHANGE FOR THE NON PROPER MAP G SO
2M_ Gb 1 !T INlNITY WE HAVE SEEN AGAIN THAT : MEETS THE lBRE
SMOOTHLY SO AGAIN ONE HAS BASECHANGE FOR G ,ET H 7 _ 9 7 )T
FOLLOWS THAT
_ M_ `
2 Gb H 1 FG
#OMBINING YIELDS
4O lNISH THE PROOF OF THE LEMMA WE MUST SHOW THE INCLUSION
_ ` _ `
7 _ G _ $ 9 7 _ G _ P 9
IS A HOMOTOPY EQUIVALENCE 7E CAN DElNE A HOMOTOPY FROM $ _ FPG TO
$ BY mOWING ALONG AN OUTWARD VECTOR lELD V %G IF P AND ONE
D D
HAS CARTESIAN COORDINATES X Y ONE CAN TAKE V X DX Y DY 3INCE 7 0 IS
6)3)/.3 ). -!4( )22%'5,!2 #/..%#4)/.3
SMOOTH OVER $ ONE CAN LIFT V TO A VECTOR lELD W ON G _ $ 3INCE : MEETS
THE lBRES OF G SMOOTHLY OVER SOME LARGER $ z $ WE CAN ARRANGE FOR W
TO BE TANGENT TO : ALONG : ,ET H BE A SMOOTH FUNCTION ON 0 WHICH IS
POSITIVE ON $ AND VANISHES ON 0 _ $ 7E VIEW G b H W AS A VECTOR lELD ON
v
7 _ G _ P &LOWING ALONG G b H W LIFTS THE mOW ALONG HV CARRIES G _ $
_ v
INTO G $ AND STABILIZES 7 _ : OVER $ 4HIS IS THE DESIRED HOMOTOPY
EQUIVALENCE c
2EMARK )T FOLLOWS FROM 4HEOREM IN ;4= THAT

: q rM_
x
EXP1 DT ` 1
b M _ AM_
2EFERENCES
;"%= 3 "LOCH ( %SNAULT ! 2IEMANN 2OCH THEOREM FOR mAT BUNDLES WITH
VALUES IN THE ALGEBRAIC #HERN 3IMONS THEORY !NNALS OF -ATHEMATICS TO
APPEAR
;"%= 3 "LOCH ( %SNAULT 'AUy[-ANIN DETERMINANTS FOR RANK IRREGULAR
CONNECTIONS ON CURVES PREPRINT
;$= 0 $ELIGNE LETTER TO * 0 3ERRE FsEVRIER
;%= ( %SNAULT !LGEBRAIC DIkERENTIAL CHARACTERS 0ROCEEDINGS OF THE #ON
FERENCE ON -OTIVES AT THE ,ANDAU )NSTITUTE IN *ERUSALEM -ARCH
"IRKHAUSER 6ERLAG IN PRINT
;+= . +ATZ 'AUy 3UMS +LOOSTERMAN 3UMS AND -ONODROMY 'ROUPS !N
NALS OF -ATH 3TUDIES 0RINCETON 5NIVERSITY 0RESS
;+O= - +ONTSEVICH 0ERIODS PREPRINT
;,= ' ,AUMON 4RANSFORMATION DE &OURIER CONSTANTES DsEQUATIONS FONCTION
NELLES ET CONJECTURE DE 7EIL 0UBL -ATH )(%3 [
;-= s
" -ALGRANGE %QUATIONS DIksERENTIELLES rA COEnCIENTS POLYNOMIAUX
0ROGRESS IN -ATH "IRKH AUSER
;34= 4 3AITO 4 4ERASOMA $ETERMINANT OF PERIOD INTEGRALS *OURN !-3
[
;4= 4 4ERASOMA #ONmUENT HYPERGEOMETRIC FUNCTIONS AND WILD RAMIlCATION
*OURN OF !LG [
3PENCER "LOCH $EPT OF -ATH 5NIVERSITY OF #HICAGO #HICAGO ), 53!

BLOCH MATHUCHICAGOEDU
(EsLr
ENE %SNAULT -ATHEMATIK &" 5NIVERSIT
AT %SSEN %SSEN 'ERMANY
ESNAULT UNI ESSENDE
02/",%-3 ). (!-),4/.)!. 0$%3
* "OURGAIN
)NTRODUCTION
4HE PURPOSE OF THIS EXPOSsE IS TO DESCRIBE A LINE OF RESEARCH AND PROBLEMS
WHICH ) BELIEVE WILL NOT BE BY ANY MEANS COMPLETED IN THE NEAR FUTURE
!S SUCH WE CERTAINLY HOPE TO ENCOURAGE FURTHER INVESTIGATIONS 4HE LIST OF
TOPICS IN THIS lELD IS FAIRLY EXTENSIVE AND ONLY A FEW WILL BE COMMENTED ON
HERE 4HEIR CHOICE WAS MAINLY DICTATED BY PERSONAL RESEARCH INVOLVEMENT
)T SHOULD ALSO BE MENTIONED THAT THE DIkERENT GROUPS OF RESEARCHERS MAY
HAVE VERY DIkERENT STYLES AND AIMS !S A SCIENCE CLAIMS AND RESULTS RANGE
FROM PURE EXPERIMENTATION TO RIGOROUS MATHEMATICAL PROOFS !LTHOUGH MY
PRIMARY INTEREST IS THIS LAST ASPECT ) HAVE NO DOUBT THAT NUMERICS OR HEURIS
TIC ARGUMENTATION MAY BE EQUALLY INTERESTING AND IMPORTANT 4HE HISTORY
OF THE +ORTEWEG DE 6RIES EQUATION FOR INSTANCE IS A STRIKING EXAMPLE OF HOW
A PROBLEM MAY EVOLVE THROUGH THESE DIkERENT INTERACTING STAGES TO EVEN
TUALLY CREATE A BEAUTIFUL THEORY !S A MATHEMATICIAN ) FEEL HOWEVER THAT
IT IS ESSENTIAL ONE REMAINS FULLY AWARE OF WHAT IS RIGOROUS MATHEMATICAL
ARGUMENTATION AND WHAT IS NOT &AILURE TO DO SO WOULD RESULT IN GENERAL
CONFUSION ABOUT THE NATURE OF THE STATEMENTS AND A GREAT LOSS OF CHALLENGING
MATHEMATICS
3OME OF MY COWORKERS BELIEVE TODAYS AVAILABILITY OF POWERFUL COMPU
TATIONAL MEANS IS PARTLY RESPONSIBLE FOR A DECLINING INTEREST IN THE OFTEN
DInCULT RIGOROUS WORK !N AMAZING COMMENT HERE IS THAT THEORETICAL COM
PUTER SCIENCE HAS BEEN TO THE CONTRARY MATHEMATICALLY INVIGORATING WITH NO
CONSENSUS PROBLEM ABOUT RIGOR )T IS ALSO TRUE THAT EVIDENCE OF CERTAIN PHE
NOMENA GATHERED FROM EXTENSIVE COMPUTATION IS OFTEN RECEIVED BY THE PURE
MATHEMATICIAN WITH CERTAIN SCEPTICISM OR DISMISSED AS UNRELIABLE !T THIS
POINT THERE DOES NOT SEEM TO BE SUCH A THING AS A TRULY CERTIlED NUMERICAL
0$% EXPERIMENT
4HE DISCUSSION BELOW IS PURELY MATHEMATICALLY ORIENTED 7E MAINLY
AIM TO HIGHLIGHT A SET OF PROBLEMS THAT ARE BOTH PHYSICALLY IMPORTANT AND
OkER AN ANALYTIC CHALLENGE 4HESE PROBLEMS RELATE TO THE TIME EVOLUTION IN
(AMILTONIAN 0$%S AND SOME OF THE THEMES ARE
6ISIONS IN -ATH 02/",%-3 ). (!-),4/.)!. 0$%3
I %XISTENCE AND BREAK DOWN OF SOLUTIONS TO THE #AUCHY PROBLEM FOR DATA
IN VARIOUS THRESHOLDS IN PARTICULAR THE EXISTENCE OF GLOBAL CLASSICAL
SOLUTIONS
II )N CASE OF SINGULARITY FORMATIONS SUCH AS BLOWUP OF LOCAL SOLUTIONS
ONE MAY INVESTIGATE BLOWUP SPEED PROlLE ETC
II !SSUMING GLOBAL SOLUTIONS EXIST FOR A CLASS OF DATA ONE MAY ASK FOR
FURTHER DETAILS ON THE DYNAMICS ! DISTINCTION NEEDS TO BE MADE
HERE BETWEEN PROBLEMS ON UNBOUNDED SPATIAL DOMAINS SUCH AS 2D
AND PROBLEMS ON BOUNDED DOMAINS FOR INSTANCE PERIODIC OR $IRICHLET
BOUNDARY CONDITIONS )N THE lRST CASE ONE OFTEN ENCOUNTERS A SCAT
TERING PHENOMENON WITH AN ASYMPTOTICALLY LINEAR mOW )N THE SECOND
CASE NO DISPERSION IS POSSIBLE AND IT IS NATURAL TO LOOK FOR CERTAIN RE
CURRENCE IN THE DYNAMICS SUCH AS INVARIANT MEASURES ON +!- TYPE
BEHAVIORS
7E WILL CENTER THE DISCUSSION AROUND TWO TYPES OF MODEL EQUATIONS
NAMELY THE NONLINEAR WAVE EQUATIONS .,7 AND NONLINEAR 3CHR ODINGER
EQUATIONS .,3 4HE .,7 CASE IS A MODEL WITH lNITE SPEED PROPAGATION
THE .,3 CASE IS NOT )N CERTAIN ISSUES SUCH AS THE #AUCHY PROBLEM INlNITE
SPEED PROPAGATION MAKES THE PROBLEM HARDER )N OTHER ISSUES SUCH AS THE
STABILITY OF QUASI PERIODIC SOLUTIONS ON BOUNDED DOMAINS FOR INSTANCE THE
STRONGER SEPARATION PROPERTIES OF NORMAL MODES IN THE .,3 CASE PLAYS A POS
ITIVE ROLE !LTHOUGH THE .,7 AND .,3 MODELS DESCRIBED BELOW HAVE BEEN
EXTENSIVELY STUDIED OVER PREVIOUS YEARS THEY STILL OkER A LARGE COLLECTION
OF UNSOLVED BASIC PROBLEMS 0AST RESEARCH HAS ALSO SHOWN THAT ARGUMENTS
AND METHODS DISCOVERED IN THIS CONTEXT USUALLY APPLY TO BROADER CLASSES
OF EQUATIONS 7E DO HOWEVER EXCLUDE FROM OUR DISCUSSION THE FEW ISOLATED
INSTANCES OF INTEGRABLE EQUATIONS THAT DO PRESENT LOTS OF ATYPICAL STRUCTURE
AND FEATURES
! COMPREHENSIVE TREATMENT OF THE TOPICS BROUGHT UP BELOW IS CERTAINLY
BEYOND THE SCOPE OF THE PAPER 4HE READER INTERESTED IN THE SUBJECT MAY
CONSULT FOR INSTANCE THE RECENT BOOKS ;3U3= ;"O= AND REFERENCES CITED 4HE
REFERENCE LIST IN THE PRESENT PAPER IS VERY INCOMPLETE AND ONLY COMPLEMENTS
THIS SHORT EXPOSsE
4WO (AMILTONIAN -ODELS

7E CONSIDER .,7 OF THE FORM
YTT _ aY yY e Y P_
* "/52'!). '!&!
WHERE y t 4HE (AMILTONIAN FORMULATION OF IS

YT _V _ (V

YT _aY yY e Y P_ (Y
WITH (AMILTONIAN
: H I

(
JRYJ V yY e P Y P
AND CANONICAL COORDINATES Y POSITION AND V SPEED

.OTICE THE POSSIBLE SIGNS DEFOCUSING CASE AND _ FOCUSING
CASE
)N THE CLASSICAL .,3 TYPE EQUATION
IUT aU f UJUJP_
THE FUNCTION U IS COMPLEX VALUED 4HE (AMILTONIAN FORMAT OF IS THUS
(
IUT U
WITH (AMILTONIAN : H I
( JRUJ e P JUJP
AND CANONICAL COORDINATES 2E U )M U !GAIN THERE IS THE DEFOCUSING RESP
FOCUSING CASE CORRESPONDING TO RESP _ SIGN IN
4HUS IN THE DEFOCUSING CASE FOR .,7 RESP .,3 CONSERVATION
2 OF THE
(AMILTONIAN IMPLIES IN PARTICULAR AN A PRIORI BOUND ON JRYJ V RESP
2
JRUJ ASSUMING INITIAL DATA Y V Y Y
? IN ( a ( RESP
S
U ( 7E DENOTE HERE BY ( THE USUAL ORDER S 3OBOLEV SPACE IE
K}K( S K S }K
)N THE FOCUSING CASE THE (AMILTONIAN MAY NOT BE BOUNDED FROM BELOW
FOR SUnCIENTLY LARGE P DEPENDING ON DIMENSION D IN THE NONLINEARITY AND
SMOOTH SOLUTIONS MAY DEVELOP SINGULARITIES IN lNITE TIME
"ESIDES THERE ARE OTHER IMPORTANT EXAMPLES OF .,7 AND .,3
TYPE EQUATIONS INVOLVING NONLOCAL NONLINEARITIES OR NONLINEARITIES CONTAINING
DERIVATIVES 7E DO NOT INTEND TO DISCUSS THEM HERE
4HE #AUCHY 0ROBLEM

7E WILL MAINLY CENTER THE DISCUSSION AROUND THE .,3
IUT aU tUJUJP_
6ISIONS IN -ATH 02/",%-3 ). (!-),4/.)!. 0$%3
4HERE IS A PARALLEL THEORY FOR .,7 4HE READER MAY CONSULT THE EXPOSsE OF
3 +LAINERMAN IN THIS VOLUME ;+= FOR A REVIEW OF RECENT PROGRESS IN THE lELD
OF HYPERBOLIC EQUATIONS
/UR SPATIAL DOMAIN WILL BE 2D IN THIS SECTION /BSERVE THAT THE EQUATION
IS INVARIANT UNDER THE SCALING

U UA X T A P_ UAX A T
WHICH ALSO LEAVES THE HOMOGENEOUS 3OBOLEV SPACE ( S WITH

P_
D _ S
INVARIANT .OTICE THE PARTICULAR CASES

P D THE CONFORMALLY INVARIANT .,3 WHERE S
AND

P D t WHERE S
D_
4HUS IN THE , NORM WHICH IS A CONSERVED QUANTITY FOR .,3 EQUATIONS
OF THE FORM IUT aU U F JUJ IS THE SCALE INVARIANT SPACE )N CASE
WE GET THE SPACE ( AND THE ( NORM MAY BE A PRIORI CONTROLLED
FROM THE (AMILTONIAN IN THE DEFOCUSING CASE
7E WILL SAY THAT THE )60 )NITIAL VALUE PROBLEM

IUT aU tUJUJP_

U } ( S
IS LOCALLY WELLPOSED PROVIDED HAS A UNIQUE SOLUTION U #( S ; 4 ; ON
A NONTRIVIAL TIME INTERVAL 4 AND U U} DEPENDS CONTINUOUSLY ON THE
INITIAL DATA } IN MOST POSITIVE RESULTS THE DEPENDENCE OF U ON } WILL BE AT
LEAST ,IPSCHITZ
)F ; 4 ; 2 THE TIME 4 MAY BE TAKEN ARBITRARILY LARGE WE CALL THE
PROBLEM GLOBALLY WELLPOSED
%SSENTIALLY SPEAKING ONE SHOULD NOT EXPECT A WELLPOSEDNESS THEORY BE
LOW THE SCALE INVARIANT THRESHOLD )N THIS RESPECT THE SITUATION FOR .,3 IS
WELL UNDERSTOOD LOCAL IN TIME
4HEOREM !SSUME IN THAT S t AND S t S DElNED BY
!SSUME ALSO P _ ;S= IF P IS NOT AN EVEN INTEGER A SMOOTHNESS COM
PATIBILITY CONDITION BETWEEN S AND THE NONLINEARITY 4HEN THERE IS 4 b
ST IS WELLPOSED ON ; 4 b ; IN THE SENSE DESCRIBED ABOVE -OREOVER
4 b 4 b K}K( S IF S S
)F K}K( S IS SUnCIENTLY SMALL THEN 4 b
* "/52'!). '!&!
3EE ;#= OR ;3U3= FOR DETAILS

2EMARKS I 4HE RESULT IS DEPENDENT ON WHETHER THE EQUATION IS FOCUSING
t OR DEFOCUSING t
II 4HE KEY INGREDIENTS IN THE PROOF ARE ESTIMATES ON THE LINEAR GROUP
EITa -ORE PRECISELY
[ 4HE DECAY ESTIMATE
KEITa }K, s #JTJ_D K}K,
[ 3TRICHARTZ INEQUALITY WHICH IS GLOBAL IN SPACE TIME
KEITa }K D DX DT #K}K
, D
&ROM THE , CONSERVATION

KUT K K}K
THERE IS THE FOLLOWING IMMEDIATE CONSEQUENCE OF 4HEOREM
#OROLLARY )F P D THEN IS GLOBALLY WELLPOSED FOR } ,
AND ALSO FOR } , < ( S S )N THIS LAST CASE WE GET A SOLUTION
U #( S 2
)T IS WELL KNOWN THAT FOR P t D IN THE FOCUSING CASE SUnCIENTLY
LARGE SMOOTH INITIAL DATA MAY LEAD TO SOLUTIONS BLOWING UP IN lNITE TIME
SEE NEXT SECTION /NE CONJECTURES HOWEVER THAT IN THE DEFOCUSING CASE
THE LOCAL SOLUTION GIVEN BY 4HEOREM EXTENDS TO A GLOBAL ONE 4HUS
0ROBLEM !SSUME t IN #AN ONE TAKE 4 b IN 4HEOREM
!SSUMING THE EQUATION DEFOCUSING t LET US SUMMARIZE SOME POS
ITIVE RESULTS AND MORE SPECIlC PROBLEMS 7E WILL CALL THE PROBLEM (
SUBCRITICAL RESP CRITICAL SUPERCRITICAL IF S RESP S S
4HUS IN DIMENSION D THE PROBLEM IS ALWAYS SUBCRITICAL 4HE NEXT
STATEMENT IS A CONSEQUENCE OF 4HEOREM AND THE ENERGY CONCENTRATIONS
#OROLLARY )N THE DEFOCUSING ( SUBCRITICAL CASE THE )60 IS GLOB
ALLY WELLPOSED FOR DATA IN ( < ( S S t )N PARTICULAR CLASSICAL SOLUTIONS
EXIST FOR ALL TIME
)N THE CRITICAL CASE THERE IS THE FOLLOWING PARTIAL RESULT SEE ;"O= ;'R=

4HEOREM ,ET D t AND P D_ DEFOCUSING CASE 4HEN THERE
IS GLOBAL WELLPOSEDNESS OF THE )60 FOR RADIAL } ( < ( S S t
6ISIONS IN -ATH 02/",%-3 ). (!-),4/.)!. 0$%3
2EMARKS I /BSERVE THAT THE LOCAL RESULT 4HEOREM PROVIDES IN THE
CRITICAL CASE S S AN EXISTENCE TIME 4 b 4 b } NOT ONLY DEPENDENT ON
K}K( S 4HUS THE DInCULTY IS TO DEAL WITH POSSIBLE ENERGY CONSERVATIONS
ON SMALL BALLS /NE OF THE ADDITIONAL INGREDIENTS INVOLVED HERE IS -ORAWETZ
DISPERSIVE INEQUALITY D t
::
JUX T JP
DX DT
JXJ
WHICH IS ALSO THE BASIS OF GLOBAL SCATTERING
4HE ANALOGUE FOR DEFOCUSING .,7
YTT _ aY Y D
YTT _ aY Y D
ETC
IS UNRESTRICTED TO RADIAL DATA 4HUS THERE IS WELLPOSEDNESS FOR ARBITRARY DATA
} ( < ( S S t 4HE PARTICULAR CASE OF RADIAL DATA IS DUE TO - 3TRUWE
;3T= AND THE GENERAL CASE WAS SOLVED BY - 'RILLAKIS ;'R= 4HIS IS AN
INSTANCE WHERE THE INlNITE SPEED PROPAGATION IN .,3 MAKES THE PROBLEM
OF CONTROLLING ENERGY CONCENTRATIONS HARDER
0ROBLEM $OES 4HEOREM REMAIN VALID FOR GENERAL DATA
II /BSERVE THAT SINCE WE DO NOT HAVE AT OUR DISPOSAL ANY CONSERVED
QUANTITY STRONGER THAN THE ENERGY THE STATEMENT IN 4HEOREM REMAINS
NONTRIVIAL FOR CLASSICAL IE SMOOTH SOLUTIONS "OTH PROOFS ;"O= ;'R=
CONTAIN IN FACT AN IMPORTANT COMPONENT WHICH IS PURELY ( 4HE SAME
CONSIDERATION OF LACK OF CONSERVED QUANTITIES EXPLAINS ALSO WHY BOTH FOR
SUPERCRITICAL .,7 AND .,3 NO GLOBAL RESULTS SEEM AVAILABLE EXCEPT FOR
SMALL DATA CF 4HEOREM 4HUS THE FOLLOWING ISSUE IS WIDELY OPEN
0ROBLEM 'LOBAL EXISTENCE OF CLASSICAL SOLUTIONS FOR DEFOCUSING SUPERCRIT
ICAL .,7 .,3
III 3OME COMMENTS ABOUT SCATTERING IN THE ENERGY SPACE
#ONSIDER THE DEFOCUSING .,3
IUT aU _ UJUJP_ U } ( < ( S S t
!SSUME

P AND P IF D t
D D_
$ENOTING THE WAVE MAP OPERATOR
:
_ `
j I EI{ a UJUJP_ { D{

* "/52'!). '!&!
ONE HAS THAT

j ( < ( S
l ITa l
lE j _ UT l S T
_
( <(
&OR D t THESE ARE CLASSICAL RESULTS GOING BACK TO 'INIBRE 6ELO ;'6= AND

,IN 3TRAUSS ;,I3= )N THE RADIAL CASE THE RESULT REMAINS VALID FOR P D_
D SEE ;"O= 2ECENTLY SCATTERING IN THE ENERGY SPACE WAS ALSO
PROVEN BY .AKANISKI ;.= IN DIMENSION D P D
0ROBLEM )S THERE GLOBAL SCATTERING IN THE ENERGY SPACE FOR P D
3EE ALSO ;#= FOR OTHER RESULTS ON SCATTERING
IV 7E LIKE TO SKETCH THE THEORETICAL POSSIBILITY FOR COMPUTER ASSISTED
PROOFS OF GLOBAL EXISTENCE AND SCATTERING FOR A GIVEN DATA } #ONSIDER FOR
INSTANCE THE $ SUPERCRITICAL PROBLEM

IUT aU _ UJUJ

U }
WHERE } IS A GIVEN SMOOTH FUNCTION 7E DO EXPECT A GLOBAL SMOOTH SOLUTION
SCATTERING &OR THIS TO HOLD IT IS SUnCIENT TO SHOW THAT FOR SOME TIME
4
A HAS A SMOOTH SOLUTION ON ; 4= %QUIVALENTLY 4 b 4 WHERE
4 b REFERS TO 4HEOREM
B 4HE NORM KEIT_4 a U4 K, ,
X
m
Tt4
WHERE m IS SOME NUMERICAL CONSTANT WE DO NOT EXPLAIN THE ROLE OF

THE , NORM HERE !BOUT STEP A )F WE lX A TIME 4 ONE MAY ESTAB
LISH THE RESULT NUMERICALLY 4O DO THIS ONE lRST GATHERS SUnCIENTLY MANY
DISCRETE DATA AND INTERPOLATES THEM WITH A SMOOTH FUNCTION V VX T
T 4 !SSUMING HAS INDEED A SMOOTH SOLUTION THE FUNCTION V WILL
EVENTUALLY NECESSARILY SATISFY THE EQUATION APPROXIMATELY IE

IVT aV _ VJVJ X T

V }
(ERE MAY BE MADE ARBITRARILY SMALL IN ANY CHOSEN SPACE TIME NORM
T s 4 BY PUSHING THE NUMERICS FAR ENOUGH $ENOTING U THE <TRUE SOLU
TION THE DIkERENCE W U _ V SATISlES AN EQUATION

IWT aW _ WJWJ 0 W W X T

W
6ISIONS IN -ATH 02/",%-3 ). (!-),4/.)!. 0$%3
WHERE 0 IS A POLYNOMIAL IN W W OF DEGREE s NO CONSTANT TERMS AND

COEnCIENTS DEPENDING ON V 4HE WELLPOSEDNESS OF ON A GIVEN INTERVAL
; 4= RESULTS THEN FORM THE SAME ANALYSIS AS THE SMALL DATA THEORY PROVIDED
X T IS SUnCIENTLY SMALL DEPENDING ON THE SIZE OF 4
.EXT CONSIDER STEP B )T MAY BE DERIVED FROM 3TRICHARTZ INEQUALITIES
THAT l I{ a l l l
lE U4 _ l s # lU4 _ l
, X{ (
SO THAT THE ISSUE IS STABLE UNDER APPROPRIATE APPROXIMATION &OR GIVEN
SMOOTH THERE IS THE STATIONARY PHASE EVALUATION FOR {
:
JXJ _ X`
>
EI{ a X EIXw{ JwJ w w J{ J_ E_I { > _ {
HENCE k I{ a k k _ `k
kE X k J{ J_ k> _ X k
{
)T SUnCES THEREFORE CLEARLY TO EVALUATE THE OSCILLATORY INTEGRAL IN ON A
LARGE ENOUGH lNITE SPACE TIME REGION /BSERVE THAT IN THIS SECOND STEP THE
FUNCTION V4 IS AN EXPLICITLY GIVEN OBJECT GOTTEN IN PARTICULAR FROM
THE STEP A CONSTRUCTION PRACTICALLY THE FUNCTIONS V AND MAY BE GIVEN
BY lNITE WAVELET EXPANSIONS
3INGULARITY &ORMATION FOR .,3

4HE ONLY KNOWN MECHANISMS IN THE .,3 CASE FOR SINGULARITY FORMATION
RELATE TO FOCUSING .,3
IUT aU UJUJP_ Pt
D
)N THIS CASE CLASSICAL SOLUTIONS OF LARGE , NORM
MAY BLOWUP IN lNITE
TIME 4 ! GENERAL CRITERION RESULT FROM 'LASSEYS VIRIEL INEQUALITY ;'L=
s: t
D
JUT J JXJ DX s C(} FOR SOME C
DT
FOR DATA } WITH (} /NE DEDUCES INDEED EASILY FROM THAT THERE
HAS TO BE A SINGULARITY FORMATION IN lNITE TIME 4 b

!SSUMING P D_ IT FOLLOWS IN PARTICULAR FROM THE LOCAL EXISTENCE
THEORY THAT
LIM

KUT K(
T 4 b
0ROBLEM 7HAT MAY BE SAID ABOUT THE BLOWUP SPEED IE THE GROWTH OF

KRUT K FOR T 4 b
* "/52'!). '!&!
,ET US CONSIDER THE CRITICAL NONLINEARITY P D 2ETURNING TO THE

LOCAL EXISTENCE THEOREM 4HEOREM FOLLOWING FACT MAY BE SHOWN ;"O=
&ACT )F THE SOLUTION U FOR AN , DATA } GIVEN BY 4HEOREM HAS MAXIMAL
EXISTENCE TIME ; 4 b ; 4 b THERE IS THE FOLLOWING , CONCENTRATION
PHENOMENON :
LIM SUP SUP JUT J DX C
)
T 4 b )y2D
J)J4 b _T
WHERE C IS A CONSTANT 7E DENOTE HERE BY ) SOME D DIMENSIONAL CUBE
AND J)J ITS SIZE
)N THE CASE OF A CLASSICAL DATA THIS CONCENTRATION PHENOMENON HAD BEEN
OBSERVED EARLIER IN PARTICULAR IN THE WORKS OF #AZENAVE 7EISSLER -ERLE AND
4SUTSUMI )N THIS CASE IMPLIES THAT
LIM 4 b _ T KUT K( C
T4 b
)N FACT IT IS KNOWN THAT IN THIS CASE THERE HAS TO BE AN , CONCENTRATION

OF AT LEAST K1K ON ARBITRARY SMALL BALLS FOR T 4 b (ERE 1 REFERS TO THE
GROUND STATE IE THE POSITIVE RADIAL SOLUTION OF
a1 _ 1 1D
)N THE $ CASE FOR INSTANCETHERE IS QUITE A BIT OF NUMERICAL AND META
MATHEMATICAL WORK RELATED TO THE CUBIC .,3
IUT aU UJUJ
3ELF SIMILAR BLOWUP SOLUTIONS ARE SUGGESTED WITH VARIOUS BLOWUP SPEEDS
J@N4 b _ T J ` 4 b _ T _ J@N4 b _ T Jl ` 4 b _ T _ FOR CERTAIN l
a @NJ@N4 b _T J b
4 b _T AMONG OTHERS 4HE READER MAY CONSULT ;3U3= FOR A MORE
DETAILED DISCUSSION AND RELATED LITERATURE
4HE FOLLOWING CONJECTURE SEEMS CONSISTENT WITH ALL HEURISTIC PREDICTIONS
0ROBLEM 0ROVE THAT IF U BLOWS UP AT TIME 4 b THEN NECESSARILY
LIM

4 b _ T KUT K(
T 4 b
4HIS WOULD BE A STRENGTHENING OF INEQUALITY

)N THE CONFORMAL CASE P D THE .,3 HAS AN ADDITIONAL SYMMETRY
GIVEN BY THE SO CALLED PSEUDO CONFORMAL TRANSFORMATION
JXJ _ `
U #UX T T_D E IT U XT _ T T
6ISIONS IN -ATH 02/",%-3 ). (!-),4/.)!. 0$%3
MAPPING A SOLUTION U TO A SOLUTION #U )N THE FOCUSING CASE PERMITS

US TO CONSTRUCT EXPLICIT BLOWUP SOLUTIONS FOR THE .,3
IUT aU UJUJD
,ET U X T EIT 1X BE THE GROUND STATE SOLUTION 1 GIVEN BY 4HEN
BY
JXJ _ _ `
UX T T_D E IT 1 XT
YIELDS AN EXPLICIT SOLUTION OF BLOWING UP AT TIME T
4HUS IN WE HAVE KUT K K1K
&ACT )F U IS A CLASSICAL SOLUTION OF BLOWING UP IN lNITE TIME THEN
KUT K t K1K
"LOWUP SOLUTIONS FOR WHICH KUT K K1K HAVE BEEN EXTENSIVELY STUD
IED BY & -ERLE ;-E= WHO PROVED IN PARTICULAR THE FOLLOWING ROUGHLY
STATED UNIQUENESS PROPERTY
4HEOREM )F KUT K K1K AND U BLOWS UP IN lNITE TIME THEN U
COINCIDES WITH UP TO THE SYMMETRIES OF THE EQUATION
3TATEMENT IS A CONSEQUENCE OF THE EARLIER STATED FACT THAT THE
AMOUNT OF THE , NORM ABSORBED IN THE BLOWUP IS AT LEAST K1K 4HERE IS
THE FOLLOWING QUESTION ABOUT A POSSIBLE QUANTIZATION
0ROBLEM !SSUME U A SOLUTION OF BLOWING UP AT TIME 4 )S THE
AMOUNT OF , NORM ABSORBED IN THE BLOWUP NECESSARILY OF THE FORM KK1K
WITH K :
2EMARKS I -ERLE ;-E= HAS CONSTRUCTED EXAMPLES OF BLOWUP SOLUTIONS
WITH AN ARBITRARY NUMBER OF BLOWUP SITES AT A lXED TIME 4HESE EXAMPLES
ARE ESSENTIAL OBTAINED BY SUPERPOSITION OF TRANSLATES OF AND SATISFY
THUS THE QUANTIZATION PROPERTY CITED ABOVE
II )N ;"O7= EXAMPLES ARE CONSTRUCTED OF SOLUTIONS U OF BLOWING
UP AT SOME TIME 4 WITH ONLY PART OF KUT K ABSORBED IN THE BLOWUP
%SSENTIALLY THE SOLUTIONS OBTAINED IN ;"O7= ARE OF THE FORM

D JXJ _ _ `
UX T T_ E IT 1 XT ~;T= VX T
WHERE V IS SMOOTH FOR ALL TIME !LSO SOLUTIONS ARE CONSTRUCTED WITH
SEVERAL DISTINCT BLOWUP TIMES
III )T IS CLEAR THAT IN
_`
KUT K( T FOR T
* "/52'!). '!&!
#LEARLY THESE MINIMUM , NORM BLOWUP SOLUTIONS ARE UNSTABLE (OWEVER

ONE MAY CONJECTURE A STABILITY PROPERTY IN THE FOLLOWING RESTRICTED SENSE WE
FORMULATE IT AS A PROBLEM ON THE GROUND STATE SOLUTION AND MAY BE CON
VERTED TO A STABILITY STATEMENT OF BY APPLYING THE PSEUDO CONFORMAL
TRANSFORMATION
0ROBLEM 4HERE IS A lNITE CODIMENSIONAL MANIFOLD - IN A NEIGHBORHOOD

OF 1 SUCH THAT IF } - THEN THE SOLUTION U OF THE )60
IUT aU UJUJD U }
IS OF THE FORM
UX T EIT 1X DISPERSIVE
!N IMPORTANT ELEMENT HERE IS THE WORK OF - 7EINSTEIN ;7= ON THE
STABILITY ANALYSIS OF THE LINEARIZATION OF EQUATION AT THE GROUNDSTATE
SOLUTION IE THE NON SELFADJOINT LINEAR PROBLEM

IVT _ ,V

V }
WHERE _ `
,V _aV V _ D 1D V _ D 1D Vv
)T IS ALSO THE MAIN INGREDIENT IN ;"O7= MENTIONED ABOVE
0ERIODIC "OUNDARY #ONDITIONS

,ET US REPLACE THE SPATIAL DOMAIN 2D BY A BOUNDED DOMAIN SUCH AS THE
D TORUS 4D 4 2: !S MENTIONED BEFORE THE ABSENCE OF DISPERSION
HERE GIVES AN ENTIRELY DIkERENT OUTLOOK ON THE TIME DYNAMICS )N THIS AND
THE NEXT SECTIONS WE PLAN TO REVIEW SOME EXPLORATIONS IN THIS CONTEXT
3TARTING WITH THE #AUCHY PROBLEM ESPECIALLY FOR .,3 THERE IS THE ISSUE OF
UNDERSTANDING THE PERIODIC ANALOGUES OF 3TRICHARTZ INEQUALITY
KEITa }K D s #K}K
, D DX DT
FOR THE LINEAR GROUP EITa WHICH IS THE BASIC INGREDIENT 2ECALL THAT
IS A GLOBAL INEQUALITY IN SPACE TIME )N THE PERIODIC SETTING THE OSCILLATORY
INTEGRAL IN IS REPLACED BY AN EXPONENTIAL SUM
8
> INXJNJ T
EITa } X }N E
N:D
WHICH IS NATURAL TO STUDY AS A FUNCTION ON 4D

6ISIONS IN -ATH 02/",%-3 ). (!-),4/.)!. 0$%3
0ROBLEM #ONJECTURE 4HE FOLLOWING INEQUALITIES HOLD

D
I KEITa }K,Q 4D s #K}K, 4D FOR Q D
D >
II KEITa }KQ { . K}K FOR Q D SUPP } y " .
2EMARKS I )NEQUALITY FAILS AT THE ENDPOINT Q D
D AND
SEEMS THE PROPER MODIlCATION
II 3TATEMENT IS CORRECT FOR D Q s AND STATEMENT FOR
D
4HESE FACTS RESULT FROM ELEMENTARY ARITHMETICS
#OMBINING EXPONENTIAL SUM ESTIMATES AND TECHNIQUES FROM &OURIER !NAL
YSIS ONE MAY OBTAIN THE FOLLOWING DISTRIBUTIONAL INEQUALITY RELATED TO
a k b D
MES X T 4D k JEITa }J t { . t_ D FOR t . D
4HE RESTRICTION ON t IN IS REMINISCENT AND RELATED TO THE RESULTS AND
PROBLEMS ON LARGE VALUES ESTIMATES FOR $IRICHLET SUMS IN NUMBER THEORY
SEE ;-O=
!S A CONSEQUENCE OF THE PRECEDING THE LOCAL AND GLOBAL #AUCHY PROBLEM
FOR PERIODIC .,3
IUT aU tUJUJP_

U } ( S 4D

U PERIODIC IN X
IS NOT AS WELL UNDERSTOOD AS THE 2D CASE ESPECIALLY WHEN D t (OWEVER
WHAT IS KNOWN FOR INSTANCE IS THE GLOBAL WELLPOSEDNESS FOR THE PERIODIC
CUBIC .,3
IUT aU _ UJUJ
IN DIMENSION D FOR DATA } ( S S t THE D CASE CORRESPONDS
TO THE INTEGRABLE :AKHAROV 3HABAT EQUATION AND SHOULD BE OMITTED FROM THE
DISCUSSION BELOW 4HUS THERE IS A SOLUTION U #( S 2 WHICH DEPENDS
REAL ANALYTICALLY ON THE DATA AND SATISlES MOREOVER AN ESTIMATE
KUT K( S s JTJ#S_ K}K( S FOR T
/BSERVE THAT FOR S THE KUT K( S IS NOT SUBJECT TO AN A PRIORI BOUND
EXCEPT FOR D
0ROBLEM )S THERE A POSSIBLE GROWTH OF KUT K( S FOR T )F SO HOW
FAST MAY THIS GROWTH BE
4HUS THE ISSUE HERE IS HOW FAST ENERGY STORED IN LOW &OURIER MODES MAY
ESCAPE TO HIGHER MODES WHEN T
* "/52'!). '!&!
2EMARKS I 4HIS PROBLEM MAY ALSO BE ADDRESSED IN CERTAIN 2D MODELS

(OWEVER IF SCATTERING IN THE ( S NORM OCCURS IT ALSO IMPLIES THAT
SUPT KUT K( S
II %STIMATES OF THE FORM MAY BE DERIVED FOR A LARGE CLASS OF (
SUBCRITICAL EQUATIONS INCLUDING .,7 /NE SHOULD OBSERVE THAT POWER LIKE
GROWTH OF HIGHER 3OBOLEV NORMS MAY ACTUALLY OCCUR IN CLOSE RELATIVES OF THE
$ CUBIC .,7 IE EXAMPLES OF THE FORM
YTT " Y Y
P
WHERE " _a IS GIVEN BY A SELFADJOINT &OURIER MULTIPLIER 7E
CONJECTURE HOWEVER THAT IN .,3 EQUATIONS SAY OF THE FORM AN ESTIMATE
KUT K( S { T K}K( S T FOR ALL
HOLDS IMPROVING ON 4HE REASON FOR THIS IS THE STRONGER ARITHMETIC
SEPARATION PROPERTIES OF NORMAL FREQUENCIES uN JNJ FOR .,3 COMPARED
WITH .,7 uN JNJ THAT DO FORCE A CERTAIN LOCALIZATION OF THE ENERGY OVER
LONG TIME SCALES )N THIS RESPECT INEQUALITY WAS PROVEN FOR LINEAR
TIME DEPENDENT 3CHRODINGER EQUATIONS OF THE FORM
IUT aU 6 X T U
WITH 6 AN ARBITRARY REAL BOUNDED SMOOTH X PERIODIC POTENTIAL AMAZINGLY
NO SPECIlED BEHAVIOR IN T SEEMS NEEDED BESIDES SMOOTHNESS
&OR MORE DETAILS AND RELEVANT LITERATURE ON THE MATERIAL PRESENTED IN
THIS SECTION THE READER MAY CONSULT ;"O=
)NVARIANT -EASURES
)N THE UNDERSTANDING OF LONG TIME DYNAMICS IN BOUNDED SPATIAL DOMAINS
EXISTENCE OF INVARIANT MEASURES ON VARIOUS PHASE SPACES IS A NATURAL THING
TO LOOK FOR &ROM A CLASSICAL POINT OF VIEW THE MOST DESIRABLE SITUATION
SHOULD BE TO CONSTRUCT SUCH A MEASURE ON SMOOTH FUNCTIONS %XCEPT FOR THE
INTEGRABLE MODEL OF THE $ CUBIC .,3 IUT UXX e UJUJ WHETHER THIS
MAY BE DONE OR NOT IS UNKNOWN
%SSENTIALLY SPEAKING IN THE OTHER CASES THE ONLY KNOWN EXAMPLES OF
INVARIANT MEASURES ARE EITHER LIVING ON SPACES OF ROUGH DATA OR lELDS THOSE
PRODUCED FROM THE 'IBBS MEASURE OR LIVE ON lNITE DIMENSIONAL OR INlNITE
DIMENSIONAL TORI WITH VERY STRONG COMPACTNESS PROPERTIES OBTAINED FROM
+!- TORI 2ESULTS ALONG THIS LINE WILL BE DESCRIBED IN THIS AND NEXT SECTION
6ISIONS IN -ATH 02/",%-3 ). (!-),4/.)!. 0$%3
)N lNITE DIMENSIONAL PHASE SPACE IT FOLLOWS FROM ,IOUVILLES THEOREM

THAT THE ,EBESGUE MEASURE ON 2N IS INVARIANT UNDER THE (AMILTONIAN mOW
( (
P? I _ QI Q?I PI I N
WITH (AMILTONIAN (P PN Q QN (ENCE SO IS THE 'IBBS MEASURE
Dv E_k( eNI DPI DQI
#ONSIDER THE CASE OF .,3 WITH PERIODIC BOUNDARY CONDITIONS IN DIMENSION D
(
IUT U
7RITING A &OURIER EXPANSION
8
UT } B
}N E INX

N:D
WE CHOOSE
B QN )M }N
PN 2E }N B
AS CANONICAL COORDINATES FOR OUR INlNITE DIMENSIONAL PHASE SPACE 4HIS
CHOICE RATHER THAN 2E U )M U HAS THE ADVANTAGE THAT IT IS EASY TO PASS
TO lNITE DIMENSIONAL MODELS BY PROJECTING ON lNITELY MANY &OURIER MODES
4HE FORMULA MAY BE WRITTEN AS

2 2
2
J}JP
JR}J J}JP
Dv Ef P E_ eD } Ef P Du
WHICH HAS THE TENTATIVE INTERPRETATION OF A WEIGHTED 7IENER MEASURE 4HIS
MEASURE IS FORMALLY INVARIANT UNDER THE mOW OF )N ORDER TO DEVELOP
THIS IDEA RIGOROUSLY THE TWO PROBLEMS TO RESOLVE ARE THE NORMALIZATION OF
AND NEXT THE CONSTRUCTION OF A WELL DElNED DYNAMICS ON THE SUPPORT
OF THAT MEASURE 4HIS PROGRAM WAS INITIATED IN THE PAPERS ;,23= 4HE
PROBLEM IS STRONGLY DIMENSIONAL DEPENDENT AND ALSO REmECTS CERTAIN ASPECTS
OF THE #AUCHY PROBLEM ESPECIALLY IN THE FOCUSING CASE SUCH AS BLOWUP
BEHAVIOR

)N DIMENSION
2 D 7IENER MEASURE LIVES ON FUNCTIONS OF CLASS ( _ e
AND HENCE J}JP DX IS ALMOST
2
SURELY lNITE FOR ANY P 4HUS IN THE DEFOCUSING
P
J}J
CASE WITH WEIGHT E_ P IN THE FORMULA PRODUCES TRIVIALLY A WELL
DElNED MEASURE THAT IS ABSOLUTELY CONTINUOUS WRT 7IENER MEASURE Du
/NE OF THE MAIN RESULTS IN ;,23= IS THAT IN THE FOCUSING CASE
2
MAY BE
P
J}J
NORMALIZED FOR P s BY RESTRICTING THE , NORM 4HUS E P IS REPLACED
BY 2
P
E P J}J ~;K}K "=

/BSERVE ALSO THAT THESE MEASURES ARE SIGNIlCANTLY DIkERENT FROM THOSE CONSTRUCTED
FOR DIkUSIVE 0$%S
* "/52'!). '!&!
WHERE " IS AN ARBITRARY CUTOk FOR P AND TAKEN SUnCIENTLY SMALL FOR
P !S MENTIONED IN SECTION P D IS CRITICAL IN THE SENSE OF
POSSIBLE BLOWUP BEHAVIOR FOR LARGE DATA IN THE CLASSICAL THEORY )T TURNS OUT
INDEED THAT THE RESTRICTIONS IN THE PREVIOUS STATEMENT ARE NECESSARY
)N ;"O= THE AUTHOR ESTABLISHED A UNIQUE DYNAMICS 4HIS AMOUNTS TO
PROVING GLOBAL WELLPOSEDNESS OF THE #AUCHY PROBLEM

IUT UXX e UJUJP_

U }
FOR ALMOST ALL DATA } IN THE SUPPORT OF THE 7IENER MEASURE 4HIS TASK IN FACT
REDUCES TO VERIFYING THE RESULT LOCAL IN TIME SINCE THE MEASURE INVARIANCE
MAY BE EXPLOITED SIMILARLY TO A CONSERVATION2 LAW
&OR D AND D THE EXPRESSION J}JP DIVERGES 4HE PROBLEM
IS RESOLVED BY THE WELL KNOWN PROCESS OF 7ICK ORDERING 2 ASSUMING J}JP A
2POLYNOMIAL IN } } HENCE P AN EVEN INTEGER 4HUS J}JP IS REPLACED BY
P
J}J WHICH IS ALMOST SURELY lNITE -OREOVER IN THE $ DEFOCUSING CASE
THE FORMULA 2 P
E_ J}J Du
STILL DElNES A WEIGHTED 7IENER MEASURE WITH DENSITY IN <P , u P
#ONSTRUCTION OF THE DYNAMICS FOR P WAS PERFORMED IN ;"O=

#ONSIDERING THE $ FOCUSING CASE EQUATION
IUT aU UJUJ
A NATURAL CONSTRUCTION WOULD CONSIST OF CONSIDERING MEASURES
2
J}J
~;2 J}J "= E Du
OBTAINED BY RESTRICTION OF THE 7ICK ORDERED , NORM
: 8 JGN J _
J} J
N
JNJ y
AND GENERATING 7IENER MEASURE BY THE RANDOM &OURIER SERIES
8 GN
} P EINX

JNJ y
N:
(ERE FGN G REFERS TO A SYSTEM OF STANDARD

2 INDEPENDENT COMPLEX 'AUSSIAN
RANDOM VARIABLES /BSERVE INDEED THAT J}J
2 AS IF D WHILE
J}J AS FOR D
4HIS IDEA WAS EXPLORED BY ! *AkE ;*= FOR CUBIC NONLINEARITIES 5NFOR
TUNATELY THE CONSTRUCTION BARELY MISSES THE QUARTIC NONLINEARITY )T DOES
6ISIONS IN -ATH 02/",%-3 ). (!-),4/.)!. 0$%3
APPLY HOWEVER IN THE CONTEXT OF (ARTREE EQUATIONS INVOLVING A NONLOCAL

NONLINEARITY OBTAINED BY CONVOLVING WITH A SUITABLE POTENTIAL #ONSIDER
THE EQUATION _ `
IUT aU JUJ b 6 U
)N $ WITH POTENTIAL 6 SATISFYING ANY CONDITION OF THE FORM
J6B N J JNJ_ FOR JNJ
7ICK ORDERING OF THE NONLINEARITY AND RESTRICTION OF THE 7ICK ORDERED ,
NORM AS IN PRODUCES THEN AN INVARIANT NORMALIZED MEASURE FOR
3IMILAR CONSTRUCTIONS WERE PERFORMED IN ;"O= FOR THE $ CASE
#OMING BACK TO THE $ NORMALIZATION PROBLEM FOR FOCUSING .,3 THE
PRECEDING DISCUSSION LEADS NATURALLY TO THE QUESTION HOW TO 7ICK ORDER OR
DEVELOP A SUBSTITUTE FOR 7ICK ORDERING IN THE CASE OF POWERS J}JP WHERE P
IS NOT AN EVEN INTEGER OR }P P : IN THE REAL CASE 4HOSE ATTEMPTS
SEEM ONLY TO PRODUCE THE FREE MEASURE ON PHASE SPACE HOWEVER ;3= WHICH
IS PHYSICALLY NOT VERY INTERESTING
0ROBLEM 7HAT ARE THE ERGODICITY PROPERTIES OF THE 'IBBS MEASURES IN
THE TYPICAL IE NON INTEGRABLE CASE
0ROBLEM %XISTENCE OF INVARIANT MEASURES ON SMOOTH PHASE SPACE FOR
INSTANCE FOR DEFOCUSING .,3
IUT aU _ UJUJ
IN D D
0ROBLEM %XISTENCE OF A LIMITING INVARIANT 'IBBS MEASURE DYNAMICS WHEN
THE PERIOD TENDS TO INlNITY
!BOUT THE LAST PROBLEM IN THE CASE OF THE $ WAVE EQUATION RESULTS
WERE OBTAINED BY -C+EAN 6ANINSKI ;-6= 4HE PROBLEM FOR .,3 IN THIS
CONTEXT IS AGAIN THE LACK OF lNITE SPEED PROPAGATION )N ;"O= RESULTS ARE
OBTAINED IN $ FOR THE DEFOCUSING EQUATION IUT UXX _ UJUJP_ WITH
P s
+!- 4ORI AND .EKHOROSHEV 3TABILITY

/VER RECENT YEARS THERE HAS BEEN A CERTAIN AMOUNT OF RESEARCH RELATED TO THE
EXISTENCE AND PERSISTENCY OF INVARIANT TORI FOR (AMILTONIAN PERTURBATIONS OF
LINEAR OR INTEGRABLE 0$%S 4HE BASIC MODEL WOULD BE THAT OF A PERTURBED
LINEAR EQUATION OF THE FORM
(
IUT !U U
v
* "/52'!). '!&!
WHERE ! IS A SELFADJOINT OPERATOR CONTAINING PARAMETERS %XAMPLES RELATED

TO THE 3CHRODINGER EQUATION WOULD BE
! &OURIER MULTIPLIER WITH !N w JNJ FOR N
! _a 6 WITH 6 A POTENTIAL
'IVEN A NONLINEAR EQUATION PARAMETERS MAY OFTEN BE EXTRACTED FROM THE
NONLINEARITY USING PARTIAL "IRKHOk NORMAL FORMS AND AMPLITUDE FREQUENCY
MODULATION 4HIS HAS BEEN PURSUED FOR INSTANCE IN ;+U0= FOR $ .,3 OF
THE FORM
IUT aU MU UJUJ HIGHER ORDER
AND IN ;"O= FOR THE $ EQUATION 7E SHALL GIVE SOME MORE DETAILS AT THE
END OF THIS SECTION
! BASIC DIkERENCE WITH THE CORRESPONDING TOPIC IN THE THEORY OF CLASSICAL
SMOOTH DYNAMICAL SYSTEMS ARE NEW RESONANCE PROBLEMS DUE TO AN INlNITE
DIMENSIONAL PHASE SPACE 4HESE DInCULTIES ARE PARTICULARLY SEVERE IN SPACE
DIMENSION $ t
3EVERAL METHODS HAVE BEEN DEVELOPED )N ;+U= THE CLASSICAL +!- AND
-ELNIKOV SCHEME BASED ON ELIMINATING THE PERTURBATION BY CONSECUTIVE
CANONICAL TRANSFORMATIONS HAS BEEN REWORKED IN THE CONTEXT OF lNITE DI
MENSIONAL TORI IN INlNITE DIMENSIONAL PHASE SPACE PROVIDED THERE ARE NO
MULTIPLICITIES OR NEAR MULTIPLICITIES IN THE NORMAL FREQUENCIES )N THE CONTEXT
OF THIS RESTRICTION LIMITS THE APPLICATION TO $ MODELS WITH $IRICHLET OF
D
.EUMANN BCS 2ECALL THAT THE PERIODIC SPECTRUM tN OF _ DX 6 APPEARS
TYPICALLY IN PAIRS tN_ tN WHERE ROUGHLY
JtN _ tN_ J w J6B N J
RAPIDLY FOR N
)N HIGHER DIMENSION UNBOUNDED MULTIPLICITIES RESULT FROM A POSSIBLE
LARGE NUMBER OF LATTICE POINTS ON A CIRCLE OR SPHERE OF RADIUS 2 WHEN
2
/NE OF THE MOST INTERESTING DEVELOPMENTS IN THIS RESEARCH IS THE EMERG
ING OF NEW TECHNIQUES TO OVERCOME THIS PROBLEM 4HE PIONEERING PAPER
HERE IS THAT OF 7 #RAIG AND % 7AYNE ;#R7= IN WHICH THE SIMPLEST CASE
PRESENTING THE MULTIPLICITY ISSUE OF TIME PERIODIC SOLUTIONS OF THE $ WAVE
EQUATIONS WITH PERIODIC BC OF THE FORM
YTT _ YXX yY Y HIGHER ORDERX Y
IS INVESTIGATED 4HE METHOD USED IS A VARIANT OF THE .ASH -OSER SCHEME
AND IS A PRIORI INDEPENDENT OF (AMILTONIAN STRUCTURE )T IS BASED ON A
6ISIONS IN -ATH 02/",%-3 ). (!-),4/.)!. 0$%3
,IAPOUNOV 3CHMIDT DECOMPOSITION AND A TREATMENT OF SMALL DIVISOR PROB

LEMS BY MULTISCALE ANALYSIS VERY REMINISCENT OF THE ARGUMENTS USED IN LOCAL
IZATION THEORY FOR LATTICE 3CHR
ODINGER OPERATORS )N PARTICULAR IN THE CONTEXT
OF A QUASI PERIODIC POTENTIAL SUCH AS THE ALMOST -ATHIEU OPERATOR
a COSNt z
CF ;&37I=
4HIS TECHNIQUE HAS LEAD TO CONSIDERABLE PROGRESS 4IME PERIODIC SOLU
TIONS IE DIM TORI FOR MAY BE PRODUCED IN ANY DIMENSION AND THE
FULL THEORY OF QUASI PERIODIC SOLUTIONS IS AVAILABLE FOR $ .,3 WITH
PERIODIC BC CF ;"O=
3EE ALSO ;#R= FOR AN UPDATED SURVEY
0ROBLEM #ONSTRUCTION OF QUASI PERIODIC SOLUTIONS FOR $ .,3 WITH PERI
ODIC BC
2EMARK 4HE DIkERENCE BETWEEN THE $ AND $ CASE IS THE LARGER MUL
TIPLICITY OF NORMAL FREQUENCIES )N $ THESE BASICALLY CORRESPOND TO THE
INTEGER SOLUTIONS OF AN EQUATION
N N 2
WITH lXED 2 4HE SET OF THOSE LATTICE POINTS IS AT MOST OF SIZE
EXP LOGLOGLOG22 { 2 THE DIVISOR FUNCTION AND MAY BE PARTITIONED IN 2
DISTANT SETS SAY CONTAINING AT MOST SITES )N $ IS REPLACED BY THE
EQUATION
N N N 2
WHICH MAY HAVE MORE THAN 2 SOLUTIONS
0ROBLEM #ONSTRUCTION OF QUASI PERIODIC SOLUTIONS FOR $ .,7 WITH PE

RIODIC BC
P
2EMARK 4HE NORMAL FREQUENCIES HERE ARE ESSENTIALLY uN N N AND
FuN _ uN G IS CLEARLY A DENSE SET IN 2 4HIS FACT SEEMS TO REQUIRE SUBSTANTIAL
IMPROVEMENT OF THE KNOWN TECHNOLOGIES
0ROBLEM (OW ABUNDANT ARE THE QUASI PERIODIC SOLUTIONS IN PHASE SPACE
4HE QUESTION MAY BE POSED IN THE CONTEXT OF ANY NONINTEGRABLE (AMILTONIAN
0$%
2EMARK 4HE METHODS OF CONSTRUCTION ARE PERTURBATIVE AND TYPICALLY
REQUIRE US TO PROCEED IN A NEIGHBORHOOD OF THE ORIGIN 4HE FOLLOWING PAR
TICULAR ISSUE FOR INSTANCE IS UNKNOWN
* "/52'!). '!&!
0ROBLEM #ONSIDER AN .,3

IUT UXX UJUJ HIGHER ORDER TERMS
AND DENOTE 10 THE UNION OF ALL SMOOTH QUASI PERIODIC TRAJECTORIES IN THE
PHASE SPACE )S 10 DENSE FOR THE WEAK TOPOLOGY
,ET US ALSO MENTION THAT IN FACT THE lRST NEW APPLICATION OF THE METHODS
DEVELOPED FOR 0$%S RELATES TO THE CLASSICAL -ELNIKOV THEOREM IN lNITE
DIMENSIONAL PHASE SPACE WHICH REMAINS VALID IN THE PRESENCE OF MULTIPLE
NORMAL FREQUENCIES
2EMARK ,ET US POINT OUT AN IMPORTANT DIkERENCE BETWEEN THE STANDARD
+!- SITUATION WHERE DIM INVARIANT TORI DIM PHASE SPACE AND THE
-ELNIKOV PROBLEM WHERE DIM INVARIANT TORI DIM PHASE SPACE )N THE
lRST SITUATION THE ADMISSIBLE FREQUENCIES FOR THE PERTURBED TORI CONTAINS
IN PARTICULAR SETS OF FREQUENCIES CHARACTERIZED BY DIOPHANTINE CONDITIONS
INDEPENDENT OF THE PERTURBATION AND ONE OBTAINS ALSO FAMILIES DEPENDING
SMOOTHLY ON THE PERTURBATION PARAMETER 4HIS IS NOT THE CASE IN THE
-ELNIKOV SETTING AND ONE lNDS SETS OF ADMISSIBLE FREQUENCIES DEPENDING ON
THE PERTURBATION 4HIS PROBLEM WAS OBSERVED IN THE WORK OF -OSER ;-OS=
%LIASSON ;%= AND IN ;"O=
0$%S ALSO PROVIDE A NATURAL SETTING FOR PROBLEMS OF EXISTENCE OF INlNITE
DIMENSIONAL INVARIANT TORI 4HERE HAS BEEN RESEARCH ON THIS TOPIC BY VARI
OUS AUTHORS 3OME OF THE MODELS CONSIDERED IN MATHEMATICAL PHYSICS ONLY
PRESENT lNITE RANGE INTERACTIONS CF ;&37= )N OTHER MORE RECENT INVES
TIGATIONS FREQUENCIES ARE CHOSEN SUnCIENTLY LACUNARY TO AVOID DIOPHANTINE
PROBLEMS -Y PARTICULAR INTEREST HERE GOES TO +!- TORI CONSTRUCTED ON THE
FULL SET OF FREQUENCIES ONLY OF QUADRATIC GROWTH FOR AN .,3 OF THE FORM
(
IUT UXX _ 6 X U U
(ERE THE NONLINEARITY IS lXED AND 6 IS TAKEN TO BE A <TYPICAL POTENTIAL
#ONSTRUCTION OF SUCH +!- TORI WAS PERFORMED INDEPENDENTLY BY * 0 OSCHEL
AND THE AUTHOR USING DIkERENT METHODS SEE ;0= ;"O= )N BOTH CON
STRUCTIONS THERE IS A VERY RAPIDLY HARDLY EXPLICIT DECAY OF THE AMPLITUDES
HENCE LEADING TO STRONG COMPACTNESS PROPERTIES )MPOSING SUCH FAST DECAY
APPEARS HERE AS TRADE Ok FOR THE SEVERE SMALL DIVISOR ISSUES 2ElNING THE
TECHNIQUES TO REACH A MORE REALISTIC DECAY CONDITION ON THE ACTION VARIABLES
IS CERTAINLY A MOST INTERESTING PROJECT
7HILE PROVING ALMOST PERIODIC BEHAVIORS FOR LARGE CLASSES OF DATA SEEMS
DInCULT AND PROBABLY IS UNTRUE ONE MAY IN SEVERAL INTERESTING INSTANCES
6ISIONS IN -ATH 02/",%-3 ). (!-),4/.)!. 0$%3
EXPECT SUCH APPROXIMATIVE BEHAVIOR OVER VERY LONG TIME SCALES )N THE CLAS
SICAL lNITE DIMENSIONAL THEORY THIS ISSUE IS ADDRESSED BY THE .EKHOROSHEV
STABILITY THEORY AND OUR AIM HERE IS TO DISCUSS SOME INVESTIGATIONS IN THIS
SPIRIT FOR (AMILTONIAN 0$%S
#ONSIDER A (AMILTONIAN OF THE FORM
8
(Q Qv tN )N & )N ( Q Qv
WHERE )N JQN J THUS A PERTURBATION OF THE INTEGRABLE CASE
2EMARK THAT WE HAVE NOT INTRODUCED ACTION ANGLE VARIABLES IN ( TO
AVOID THE SINGULARITY WHEN JQN J FOR JNJ !SSUME JtN J t &
BOUNDED FROM BELOW AND ( Q Qv REAL ANALYTIC IN THE TOPOLOGY INDUCED BY
THE UNPERTURBED (AMILTONIAN /BVIOUSLY
k: k
k k k 4 D)N k
k)N 4 _ )N k k k
k DT k

k: 4 k
k k
kk 2E QvN Q?N kk

k: 4 k
k ( k
kk )M QvN k
QN k
: 4
s J(Q Qv J 4

3O THAT STABILITY FOR TIME SCALES IS CLEAR
,ET US RECALL TWO MECHANISMS LEADING TO STABILITY FOR LONGER TIME SCALES
l
4 { ! ! LARGE OR EVEN 4 E l 4HE lRST IS THAT OF A
PERTURBATION OF A NONRESONNANT LINEAR SYSTEM THUS WITH (AMILTONIAN OF THE
FORM 8
( tN )N ( Q Qv
v
WHERE THE FREQUENCY VECTOR t tN HAD <GOOD DIOPHANTINE PROPERTIES
)N THIS CASE ONE AIMS TO REMOVE BY THE STANDARD MECHANISM OF SYMPLECTIC
TRANSFORMATIONS THE PERTURBATION IN UP TO HIGH ORDER IN 4HESE
SYMPLECTIC TRANSFORMATIONS ` ARE OBTAINED AS TIME SHIFT FOR THE (AMILTO
NIAN mOW WITH (AMILTONIAN & OF THE FORM
8 ANv
& Q Qv Q v Q v Qv Q v ` ` `
v tN v tN ` ` ` N N N N
v e Q p Q Q _ p Qv
g

h
( ` ( F( & G F( & G & ` ` `
0 _ ( `
WHERE F( ( G N ( (
QN QvN _ QvN
(
QN IS THE 0OISSON BRACKET
* "/52'!). '!&!
)N HIGH OR INlNITE DIMENSIONAL PHASE SPACE THE RESONANCE PROBLEMS MAY

REQUIRE US TO
1ALLOW IN THE REMAINDER TERM ALSO CERTAIN EXPRESSIONS INVOLVING
v
MONOMIALS J QNJJ OF LARGE DEGREE AND CERTAIN MONOMIALS INVOLVING QN S WITH
N LARGE BESIDES ! TERMS !SSUMING THE QN SMALL WITH RAPID DECAY AT
INlNITY THESE ADDITIONAL TERMS TURN OUT ALSO TO HAVE AN ! EkECT ! LARGE
4HUS IN THE NEW COORDINATES THE (AMILTONIAN HAS THE FORM
8
( ^ N )N !
t
WITH t^ N tN THE MODULATED FREQUENCIES AND WE GET STABILITY FOR

TIMES 4 _! )N THIS SETUP THE STABILITY RESULT MAY RELATE TO TOPOLO
GIES DIkERENT FROM THE (AMILTONIAN ONE &OR INSTANCE IN ;"O= RESULTS
ALONG THIS LINE WERE OBTAINED FOR (AMILTONIAN PERTURBATIONS OF NONRESON
NANT LINEAR 3CHRODINGER AND WAVE EQUATIONS $ WITH $IRICHLET BC OF THE
FORM
(
IUT UXX 6 X U U
v
YTT _ YXX yY Y F Y
,ET US RECALL HERE THAT IN 0$% ISSUES THE (AMILTONIAN ( Q Qv IN
DOES NOT INVOLVE ONLY SHORT RANGE INTERACTIONS
!S MENTIONED BEFORE ONE MAY SOMETIMES REDUCE THE (AMILTONIAN TO THE
FORM WITH A DIOPHANTINE FREQUENCY VECTOR t v BY EXPLOITING NONLINEAR
PARTS OF THE (AMILTONIAN AND AMPLITUDE FREQUENCY MODULATION 4HE TYPICAL
EXAMPLE IS THAT OF A "IRKHOk NORMAL FORM
8 8
( tN )N )N ( Q Qv
$ElNE
*N )N _ )N
SO THAT 8 8
( CONST ^ N *N
t *N ( Q Qv
WITH MODULATED FREQUENCIES
^ N tN )N
t
&OR TYPICAL DATA Q )N JQN J
ONE MAY THUS EXPECT CERTAIN DIO
PHANTINE PROPERTIES FROM THE NEW FREQUENCIES WHICH REDUCES THE QUESTION
TO CASE
4HERE ARE 0$% MODELS PARTICULARLY SUITABLE TO THIS APPROACH #ONSIDER
FOR INSTANCE THE $ .,3
8
IUT UXX e UJUJ HIGHER ORDER UX T QN T EINX
6ISIONS IN -ATH 02/",%-3 ). (!-),4/.)!. 0$%3
4HE NORMAL FORM TRANSFORMATION OBSERVED BY +UKSIN 0 OSCHEL ;+U0= PER

MITS US TO EXPRESS THE (AMILTONIAN OF THIS MODEL IN THE FORM
8 8
(Q Qv N y JQN J v JQN J HIGHER ORDER
!SSUMING KQK APPROPRIATE RESCALING LEADS THUS TO A NORMAL FORM

,ONGTIME STABILITY RESULTS ALONG THESE LINES WERE OBTAINED BY "AMBUSI
;"= CONSIDERING TYPICAL DATA Q THAT ARE ALMOST FULLY LOCALIZED ON A lNITE SET
OF &OURIER MODES THUS
8
Q QN EINX
N&
WHERE & " LOG
)N ;"O= ONE CONSIDERS TYPICAL DATA Q IN ( S S LARGE OF SIZE AND
OBTAIN STABILITY TIMES ! WITH ! !S FOR S
4HERE IS A SECOND MECHANISM FOR STABILITY SOMETIMES REFERRED TO AS
THE <GEOMETRIC .EKHOROSHEV THEOREM 4HE (AMILTONIAN CONSIDERED IS FOR
INSTANCE OF THE FORM
8 8
( tN )N )N ( Q Qv
NO DIOPHANTINE ASSUMPTION ON t (ERE THE NONLINEARITY CONVEXITY IN
CASE BUT WEAKER ASSUMPTIONS MAY BE EXPLOITED TOO ;.E= IS USED
TO DERIVE STABILITY OF THE ACTION VARIABLES IN THE (AMILTONIAN TOPOLOGY
4HERE IS AN ELEGANT ARGUMENT DUE TO ,OCHAK ;,O= BASED ON APPROXIMATION
BY PERIODIC ORBITS )T WAS ALSO REWORKED IN 0$% CONTEXT RELATED TO .,3
TO OBTAIN EXPONENTIALLY LONG STABILITY TIMES CONSIDERING AGAIN DATA
OF THE FORM SEE ;"=
0ROBLEM 4O WHAT EXTENT DOES THE GEOMETRIC .EKHOROSHEV THEOREM APPLY
TO THE 0$% CONTEXT
7HAT SEEMS CLEAR IS THAT IN BOTH STABILITY MECHANISMS DESCRIBED ABOVE
STRONG RESTRICTIONS NEED TO BE IMPOSED ON THE METRICAL ENTROPY OF THE UN
PERTURBED TORUS COMPARED WITH THE SIZE OF THE PERTURBATION )N THIS
RESPECT IT IS POSSIBLE TO CONSTRUCT EXAMPLES IN . DIM PHASE SPACE
OF (AMILTONIANS
8. 8
( tN JQN J JQN J ( Q Qv
N
tN w WITH ( A POLYNOMIAL BOUNDED IN @. # NORM SUCH THAT FOR
4 w k k
k JQ 4 J _ JQ J k
* "/52'!). '!&!
WITH FREQUENCIES tN CHOSEN RANDOMLY 7HAT HAPPENS HERE IS THAT THE

UNPERTURBED TORUS IS TOO LARGE IN THE SENSE OF @ ENTROPY COMPARED WITH
4HE CONSTRUCTION MAY PROBABLY BE IMPROVED IN THIS RESPECT BUT IT CERTAINLY
DOES NOT COME CLOSE TO EXPLAIN FULLY THE STRONG LIMITATIONS OF THE POSITIVE
RESULTS AVAILABLE SO FAR
)N THE CLASSICAL THEORY A DRIFT OF ACTION VARIABLES OVER EXPONENTIALLY LONG
TIMES IN IS EXPLAINED BY THE MECHANISM OF !RNOLD DIkUSION INVOLVING A
MOTION ON TRAJECTORIES ESSENTIALLY CONTAINED IN ALTERNATING STABLE AND UNSTA
BLE MANIFOLDS OF +!- TORI /F COURSE ONE MAY EXPECT THAT THIS MECHANISM
WILL ALSO BE PRESENT IN INlNITE (AMILTONIAN MODELS /BSERVE HERE THAT THIS
METHOD REQUIRES A CERTAIN ABUNDANCE OF INVARIANT +!- TORI WHERE OUR
UNDERSTANDING IS CERTAINLY MORE LIMITED IN THE 0$% SETTING AT THIS POINT
)T IS HOWEVER MORE NATURAL TO LOOK HERE FOR MECHANISMS OF DIkUSION THAT
ARE DUE TO LARGE SPACE DIMENSION AND INTERACTIONS WITH MANY MODES )N
THE EXAMPLE MENTIONED ABOVE FOR INSTANCE Q INTERACTS WITH ALL THE MODES
Q Q. AND THE INSTABILITY TIME IS MUCH SHORTER )T IS LIKELY THE
PRECISE NATURE OF THE NONLINEARITY FOR INSTANCE SHORT RANGE INTERACTIONS
PLAYS A ROLE HERE
2EFERENCES
;"= $ "AMBUSI ,ONG TIME STABILITY OF SOME SMALL AMPLITUDE SOLUTIONS IN
NONLINEAR 3CHRODINGER EQUATIONS #-0 [
;"= $ "AMBUSI .EKHOROSHEV THEOREM FOR SMALL AMPLITUDE SOLUTIONS IN NON
LINEAR 3CHRODINGER EQUATIONS -ATH : [
;"E&'= ' "ENETTIN * &RO HLICH ! 'IORGILLI ! .EKHOROSHEV TYPE THE
OREM FOR (AMILTONIAN SYSTEMS WITH INlNITELY MANY DEGREES OF FREEDOM
#-0 [
;"O= * "OURGAIN 'LOBAL SOLUTIONS OF NONLINEAR 3CHRODINGER EQUATIONS !-3
#OLLOQUIUM 0UBLICATIONS
;"O= * "OURGAIN 'LOBAL WELLPOSEDNESS OF DEFOCUSING $ CRITICAL .,3 IN THE
RADIAL CASE * !-3
;"O= * "OURGAIN 2ElNEMENTS OF 3TRICHARTZ INEQUALITY AND APPLICATIONS TO
$ .,3 WITH CRITICAL NONLINEARITY )-2. . [
;"O= * "OURGAIN .ONLINEAR 3CHRODINGER EQUATIONS IN (YPERBOLIC EQUATIONS
AND FREQUENCY INTERACTIONS 0ARK #ITY 54 )!30ARK #ITY
-ATH 3ER !-3
;"O= * "OURGAIN 0ERIODIC NONLINEAR 3CRODINGER EQUATIONS AND INVARIANT
MEASURE #OMM -ATH 0HYS [
;"O= * "OURGAIN )NVARIANT MEASURES FOR THE $ DEFOCUSING NONLINEAR
6ISIONS IN -ATH 02/",%-3 ). (!-),4/.)!. 0$%3
3CHRODINGER EQUATION #-0 [

;"O= * "OURGAIN )NVARIANT MEASURES FOR THE 'ROSS 0IATEVSKII EQUATION *
-ATH 0URES ET !PPLIQUsEES 4 &ASE [
;"O= * "OURGAIN )NVARIANT MEASURES FOR .,3 IN INlNITE VOLUME PREPRINT
TO APPEAR IN #-0
;"O= * "OURGAIN 1UASI PERIODIC SOLUTIONS OF (AMILTONIAN PERTURBATIONS OF
$ LINEAR 3CHRODINGER EQUATIONS !NNALS OF -ATH [
;"O= * "OURGAIN /N -ELNIKOVS PERSISTENCY PROBLEM -ATH 2ESEARCH ,ET
TERS [
;"O= * "OURGAIN #ONSTRUCTION OF APPROXIMATIVE AND ALMOST PERIODIC SO
LUTIONS OF PERTURBED LINEAR 3CHR ODINGER AND WAVE EQUATIONS '!&!
[
;"O= * "OURGAIN /N DIkUSION IN HIGH DIMENSIONAL (AMILTONIAN SYSTEMS AND
0$% TO APPEAR IN * !NALYSE *ERUSALEM
;"O7= * "OURGAIN 7 7ANG #ONSTRUCTION OF BLOWUP SOLUTIONS FOR THE
NONLINEAR 3CHRODINGER EQUATION WITH CRITICAL NONLINEARITY !NNALI 3CUOLA
.ORM 3UP 0ISA 6OL DEDICATED TO $E 'IORGIO [
;#= 4 #AZENAVE !N INTRODUCTION TO NONLINEAR 3CHR ODINGER EQUATION 4EX
TOS DE -sETODOS -ATHEMs ATICOS )NSTITUTO DE -ATEMATICA 5&2* 2IO
DE *ANEIRO
;#R= 7 #RAIG 0ROBLsEMES DE PETITS DIVISEURS DANS LES EQUATIONS AUX DsERIVIES
PARTIELLES PREPRINT
;#R7= 7 #RAIG # 7AYNE .EWTONS METHOD AND PERIODIC SOLUTIONS OF NON
LINEAR WAVE EQUATIONS #OMM 0URE !PPL -ATH [
;%= ,( %LIASSON 0ERTURBATIONS OF STABLE INVARIANT TORI !NN 3CUOLA .ORM
3UP 0ISA #L 3CI [
;&37= * &RO HLICH 4 3PENCER % 7AYNE ,OCALIZATION IN DISORDERED NON
LINEAR DYNAMICAL SYSTEMS * 3TAT 0H [
;&37I= * &RO HLICH 4 3PENCER 0 7ITTWER ,OCALIZATION FOR A CLASS OF
ONE DIMENSIONAL QUASI PERIODIC 3CHRODINGER OPERATORS #-0
[
;'6= * 'INIBRE ' 6ELO 3CATTERING THEORY IN THE ENERGY SPACE FOR A CLASS OF
NONLINEAR 3CHRODINGER EQUATIONS * -ATH 0URE !PPL [
;'L= 2 'LASSEY /N THE BLOWING UP OF SOLUTIONS TO THE #AUCHY PROBLEM FOR
NONLINEAR 3CHRODINGER OPERATORS * -ATH 0HYS [
;'R= - 'RILLAKIS 2EGULARITY AND ASYMPTOTIC BEHAVIOUR OF THE WAVE EQUA
TION WITH A CRITICAL NONLINEARITY !NNALS OF -ATH [
;'R= - 'RILLAKIS #LASSICAL SOLUTIONS FOR THE NONLINEAR 3CHRODINGER EQUATION
PREPRINT
;*= ! *AFFE 0RIVATE COMMUNICATION
;+= 3 +LAINERMAN 0$% AS A UNIlED SUBJECT IN THESE 0ROCEEDINGS
;+U= 3 +UKSIN .EARLY INTEGRABLE INlNITE DIMENSIONAL (AMILTONIAN SYSTEMS
* "/52'!). '!&!
3PRINGER ,ECTURE .OTES IN -ATH

;+U0= 3 +UKSIN * 0O SCHEL )NVARIANT #ANTOR MANIFOLDS OF QUASI PERIODIC
OSCILLATIONS FOR A NONLINEAR 3CHRODINGER EQUATION !NNALS OF -ATH
[
;,23= * ,EBOWITZ 2 2OSE % 3PEER 3TATISTICAL MECHANICS OF THE NONLINEAR
3CHRODINGER EQUATION * 3TAT 0HYS [
;,I3= * ,IN 7 3TRAUSS $ECAY AND SCATTERING OF A SOLUTION OF A NONLINEAR
3CHRODINGER EQUATION *&!
;,O= 0 ,OCHAK #ANONICAL PERTURBATION THEORY VIA SIMULTANEOUS APPROXIMA
TION 5SPEKHI -ATH .AUK [
;-6= ( -C+EAN + 6ANINSKI 3TATISTICAL MECHANICS OF NONLINEAR WAVE EQUA
TIONS IN @3TOCHASTIC !NALYSIS 0ROC 3YMP 0URE -ATH [

;-E= & -ERLE $ETERMINATION OF BLOW UP SOLUTIONS WITH MINIMAL MASS FOR
NONLINEAR 3CHRODINGER EQUATION WITH CRITICAL POWER $UKE -ATH *
[
;-E= & -ERLE #ONSTRUCTION OF SOLUTIONS WITH EXACT K BLOW UP POINTS FOR
THE 3CHR ODINGER EQUATION WITH CRITICAL POWER NONLINEARITY #OMM -ATH
0HYS [
;-O= ( -ONTGOMERY 4EN ,ECTURES ON THE )NTERFACE "ETWEEN !NALYTIC .UM
BER 4HEORY AND (ARMONIC !NALYSIS #"-3 2EGIONAL #ONF 3ER IN -ATH

;-OS= * -OSER /N THE THEORY OF QUASI PERIODIC MOTIONS 3IAM 2EVIEW
[
;.= . .AKANISHI %NERGY SCATTERING FOR NONLINEAR +LEIN['ORDON AND
3CHRODINGER EQUATIONS IN SPATIAL DIMENSIONS AND PREPRINT 54-3

;.E= . .EKHOROSHEV !N EXPONENTIAL ESTIMATE OF THE TIME OF STABILITY OF
NEARLY INTEGRABLE (AMILTONIAN SYSTEMS 5SPEKHI -ATH .AUK
[
;0= * 0O SCHEL /N THE CONSTRUCTION OF ALMOST PERIODIC SOLUTIONS FOR A NON
LINEAR 3CHRODINGER EQUATION PREPRINT
;3= 4 3PENCER 0RIVATE COMMUNICATION
;3T= - 3TRUWE 'LOBALLY REGULAR SOLUTIONS TO THE U +LEIN 'ORDON EQUATION
!NN 3CUOLA .ORM 3UP 0ISA 3ER [
;3U3= # 3ULEM 0 , 3ULEM 4HE NONLINEAR 3CHR ODINGER EQUATION 3ELF
&OCUSING AND 7AVE #OLLAPSE 3PRINGER !PPLIED -ATH 3CIENCES
;7= - 7EINSTEIN -ODULATIONAL STABILITY OF GROUND STATES OF NONLINEAR
3CHRODINGER EQUATIONS 3IAM * -ATH !NAL [
*EAN "OURGAIN )NSTITUTE FOR !DVANCED 3TUDY 0RINCETON .* 53!
AND 5NIVERSITY OF )LLINOIS 5RBANA ), 53!
BOURGAIN IASEDU
/. ! #,!33)#!, ,)-)4 /& 15!.45- 4(%/29

!.$ 4(% ./. ,).%!2 (!242%% %15!4)/.
RG &RO
*U HLICH 4AI 0ENG 4SAI AND (ORNG 4ZER 9AU
'ENERAL 0ERSPECTIVE $ESCRIPTION OF THE 0ROBLEM

(ISTORICALLY THE DESCRIPTION OF .ATURE IN THEORETICAL PHYSICS BEGAN WITH
AN ENCODING OF RATHER SPECIAL SIMPLE NATURAL PHENOMENA EXHIBITING HIGH
APPROXIMATE SYMMETRY IN SPECIAL UNSTABLE MATHEMATICAL STRUCTURES &OR
EXAMPLE THE SHAPE OF THE ORBITS OF PLANETS AROUND THE SUN WAS CONCEIVED
TO BE CIRCULAR !S THE AMBITION TO DESCRIBE MORE GENERAL CLASSES OF NATURAL
PHENOMENA GREW THOSE SPECIAL MATHEMATICAL STRUCTURES HAD TO BE DEFORMED
AND ENLARGED INTO MORE GENERAL MORE STABLE MATHEMATICAL STRUCTURES +E
PLER DISCOVERED THAT THE IDEA OF ELLIPTIC ORBITS WITH THE SUN IN ONE OF THE
FOCI PROVIDES A BETTER MODEL FOR THE MOTION OF PLANETS
7ITH THE ADVENT OF THE IDEA OF DYNAMICAL ,AWS OF .ATURE ONE DISCOVERED
THAT IT IS THESE LAWS THAT MAY EXHIBIT HIGH SYMMETRY RATHER THAN SPECIAL
CONSEQUENCES OF THESE LAWS .EWTONIAN MECHANICS LED TO A lRST MODEL OF
SPACE AND TIME 4HE ORBITS OF PLANETS AND MOONS COULD BE UNDERSTOOD AS
TRAJECTORIES IN .EWTONIAN SPACE TIME WHICH ARE SOLUTIONS OF THE .EWTONIAN
EQUATIONS OF MOTION 4HESE EQUATIONS OF MOTION EXHIBIT HIGH SYMMETRY
)N ALL INERTIAL FRAMES OF REFERENCE THEY TAKE THE SAME FORM THE TRANSITION
FROM ONE TO ANOTHER INERTIAL FRAME BEING GIVEN BY A 'ALILEI TRANSFORMATION
4HESE SYMMETRIES ARE ACCOMPANIED BY CONSERVATION LAWS
$URING THE TWENTIETH CENTURY OUR UNDERSTANDING OF .ATURE UNDERWENT
REVOLUTIONARY DEVELOPMENTS AND CHANGES WHICH LED TO AT LEAST THREE FUN
DAMENTAL DEFORMATIONS OF THE MATHEMATICAL STRUCTURE UNDERLYING PHYSICAL
THEORY 4HE lRST ONE ORIGINATED FROM THE STUDY OF THE LAWS OF BLACK BODY
RADIATION AND RESULTED IN THE DISCOVERY OF QUANTUM THEORY !S EARLY AS
IT UNCOVERED NEW FUNDAMENTAL CONSTANTS OF .ATURE THE ELEMENTARY ELECTRIC
CHARGE E THE "OLTZMANN CONSTANT K" 0LANCKS CONSTANT Y [ ALL RELATED TO
.OTES ACCOMPANYING A LECTURE PRESENTED BY * & AT THE CONFERENCE <6ISIONS IN -ATH
EMATICS 4EL !VIV !UGUST

* &2/(,)#( 4 0 43!) !.$ ( 4 9!5 '!&!
THE ATOMIC CONSTITUTION OF MATTER [ AND THE 0LANCK LENGTH @0 WHICH IS

CONNECTED WITH GRAVITY 0LANCKS LAW OF BLACK BODY RADIATION COMBINED
WITH PREVIOUSLY DISCOVERED LAWS OF THERMODYNAMICS AND STATISTICAL PHYSICS
ENABLED HIM TO DETERMINE THE APPROXIMATE VALUES OF E K" AND Y AND FROM
THE SPEED OF LIGHT C AND .EWTONS LAW OF GRAVITATIONAL ATTRACTION HE COULD
THEN INFER THE APPROXIMATE VALUE OF @0
4HE SECOND REVOLUTION AROSE FROM ATTEMPTS TO lND A UNIlED DESCRIPTION
OF MECHANICAL AND ELECTROMAGNETIC PHENOMENA AND RESULTED IN %INSTEINS
SPECIAL THEORY OF RELATIVITY AND IN THE DEFORMATION OF .EWTONIAN SPACE TIME
TO -INKOWSKI SPACE TIME AND OF THE 'ALILEI GROUP TO THE 0OINCARsE GROUP
4HE UNDERLYING DEFORMATION PARAMETER IS THE SPEED OF LIGHT C 4HE THIRD
ONE AROSE FROM TRYING TO lND A RELATIVISTIC THEORY OF GRAVITATION AND AN
INVARIANT COORDINATE INDEPENDENT FORMULATION OF FUNDAMENTAL DYNAMICAL
LAWS AND RESULTED IN %INSTEINS GENERAL THEORY OF RELATIVITY )N THIS THEORY
-INKOWSKI SPACE TIME IS DEFORMED TO A DYNAMICAL SPACE TIME DESCRIBED AS
A ,ORENTZIAN MANIFOLD WHOSE TOPOLOGY AND METRIC SHAPE ARE DETERMINED
IN PART BY THE DISTRIBUTION OF MATTER 'LOBAL SPACE TIME SYMMETRIES ARE
PROMOTED TO LOCAL SYMMETRIES GAUGE SYMMETRIES OF THE SECOND KIND 4HE
UNDERLYING DEFORMATION PARAMETER IS @0
4HE DEFORMATION OF THE MATHEMATICAL STRUCTURE UNDERLYING (AMILTONIAN
MECHANICS TO THE ONE UNDERLYING NON RELATIVISTIC QUANTUM MECHANICS WAS
ACCOMPLISHED LAST AND HAS REMAINED PERHAPS THE MOST MYSTERIOUS ONE
4HE UNDERLYING DEFORMATION PARAMETER IS 0LANCKS CONSTANT Y )T REPRE
SENTS THE lRST EXAMPLE OF A DEFORMATION OF CLASSICAL GEOMETRY TO NONCOM
MUTATIVE GEOMETRY NAMELY THE DEFORMATION OF SYMPLECTIC GEOMETRY TO THE
NONCOMMUTATIVE GEOMETRY OF QUANTIZED PHASE SPACES
4HE THREE DEFORMATIONS MENTIONED ABOVE AND THEIR DIVERSE FUNDAMEN
TAL PHYSICAL CONSEQUENCES HAVE SET MUCH OF THE AGENDA FOR THE THEORETICAL
PHYSICS OF THE TWENTIETH CENTURY AND ARE LIKELY TO OCCUPY AN IMPORTANT PLACE
IN OUR AGENDA FOR MANY YEARS TO COME
&ROM THE EXPERIENCE GAINED FROM THE DEFORMATIONS OF .EWTONIAN TO REL
ATIVISTIC MECHANICS OF SPECIAL TO GENERAL RELATIVITY AND OF (AMILTONIAN TO
QUANTUM MECHANICS ONE MAY INFER THAT DIMENSIONFUL FUNDAMENTAL CON
STANTS OF .ATURE C Y @0 PLAY THE ROLE OF DEFORMATION PARAMETERS TO BE
DISTINGUISHED FROM SCALE HIERARCHY PARAMETERS SUCH AS THE 2YDBERG THE
1#$ SCALE OR THE MASS OF THE : BOSON AND FROM DIMENSIONLESS COUPLING
CONSTANTS SUCH AS THE FEINSTRUCTURE CONSTANT j OR 9UKAWA COU
PLINGS 4O <TURN ON C_ OR Y OR @0 NECESSITATES A FUNDAMENTAL CHANGE IN
6ISIONS IN -ATH /. ! #,!33)#!, ,)-)4 /& 15!.45- 4(%/29
THE MATHEMATICAL STRUCTURE OF PHYSICAL THEORY THE PASSAGE FROM ONE FAMILY
TO A NEW FAMILY OF THEORIES WHILE RATIOS OF ENERGY SCALES AND DIMENSION
LESS COUPLING CONSTANTS CAN BE VIEWED AS MODULI PARAMETRIZING A FAMILY OF
THEORIES $IMENSIONFUL FUNDAMENTAL CONSTANTS OF NATURE PERMIT US TO IN
TRODUCE A NATURAL SYSTEM OF UNITS IN WHICH THEY TAKE THE VALUE IE C
Y AND @0 IN NATURAL UNITS WHILE RATIOS OF ENERGY SCALES AND DIMEN
SIONLESS COUPLING CONSTANTS ARE PURE NUMBERS INDEPENDENT OF OUR CHOICE OF
UNITS
$EFORMING ONE FAMILY OF THEORIES TO A NEW FAMILY OF THEORIES IS THE SUBJECT
OF DEFORMATION THEORY WHILE THE PASSAGE FROM AN EkECTIVE DESCRIPTION OF
NATURAL PHENOMENA AT A HIGH ENERGY SCALE TO AN EkECTIVE DESCRIPTION OF
PHENOMENA AT A LOWER ENERGY SCALE WITHIN ONE FAMILY OF EkECTIVE THEORIES IS
THE SUBJECT OF RENORMALIZATION GROUP THEORY $EFORMATION THEORY IS ALGEBRA
RENORMALIZATION GROUP THEORY IS ANALYSIS 0ROGRESS IN THEORETICAL PHYSICS
APPEARS TO RELY ON PROGRESS IN DEFORMATION [ AND IN RENORMALIZATION GROUP
THEORY
4HE SKETCH IN &IG MAY CONVEY A VERY ROUGH IMPRESSION OF WHERE WE
STAND AT PRESENT
! .EWTONIAN MECHANICS %UCLIDIAN GEOMETRY (AMILTONIAN MECHANICS
SYMPLECTIC GEOMETRY
" 3PECIAL RELATIVISTIC CLASSICAL lELD THEORY -INKOWSKI GEOMETRY INlNITE
DIMENSIONAL SYMPLECTIC GEOMETRY
# .ON RELATIVISTIC QUANTUM MECHANICS NONCOMMUTATIVE GEOMETRY OF
QUANTIZED PHASE SPACES
$ 3PECIAL RELATIVISTIC QUANTUM lELD THEORY NONCOMMUTATIVE GEOMETRY
OF INlNITE DIMENSIONAL QUANTIZED PHASE SPACES FUNCTIONAL INTEGRATION
"234 AND "6 COHOMOLOGY
% .EWTONIAN GRAVITY AND ITS REFORMULATION AS A GEOMETRICAL THEORY MAK
ING USE OF 2IEMANNIAN GEOMETRY
& 'ENERAL 2ELATIVITY GENERAL ,ORENTZIAN GEOMETRY
' <.ON RELATIVISTIC QUANTUM GRAVITY NONCOMMUTATIVE 2IEMANNIAN
GEOMETRY
- <1UANTUM GRAVITY OR <- THEORY INlNITE DIMENSIONAL NONCOM
MUTATIVE DIkERENTIAL GEOMETRY
4HE QUESTION MARKS INDICATE THAT THE CORRESPONDING TERMS DO NOT HAVE
A MATHEMATICALLY PRECISE DElNITION YET 4HE <%SCHERIAN ASPECT OF THE
CUBE IN &IG IS INTENDED TO INDICATE THAT WE ARE STILL FACING OBSTACLES OR

* &2/(,)#( 4 0 43!) !.$ ( 4 9!5 '!&!
&IGURE
PERHAPS OBSTRUCTIONS IN ATTEMPTING TO ADD THE MISSING VERTICES ' AND -

AND A LINK BETWEEN THEM TO THE CUBE
/UR DISCUSSION MAKES IT CLEAR THAT THE FOLLOWING HAVE BEEN AND WILL
REMAIN IMPORTANT TASKS FOR THEORETICAL PHYSICISTS
&IND SUITABLE STABLE MATHEMATICAL STRUCTURES AS FOUNDATIONS FOR
PHYSICAL THEORY BY DEFORMING MORE SPECIAL LESS STABLE MATHEMATICAL STRUC
TURES KNOWN TO UNDERLY PHYSICAL THEORY IN VARIOUS ASYMPTOTIC REGIMES
&OR THE FUTURE @0 &IND THE DEFORMATIONS LEADING TO VERTEX ' AND
MOST IMPORTANTLY TO VERTEX - <- THEORY
/NCE A PHYSICAL THEORY FOUNDED ON A GENERAL STABLE MATHEMATI
CAL STRUCTURE EG NON RELATIVISTIC QUANTUM MECHANICS HAS BEEN FOUND
ONE MUST UNDERSTAND IN WHICH ASYMPTOTIC REGIME IT REDUCES TO ONE OF ITS
ANCESTOR THEORIES EG (AMILTONIAN MECHANICS FOUNDED ON A LESS STABLE
STRUCTURE FROM WHICH THE FORMER HAS ARISEN BY DEFORMATION
&OR THE FUTURE LENGTH SCALES | @0
6ISIONS IN -ATH /. ! #,!33)#!, ,)-)4 /& 15!.45- 4(%/29
SPECIAL RELATIVISTIC
<- THEORY QUANTUM lELD THEORY
COMPACT DIMENSIONS
$ BRANES
7E DONT REALLY KNOW PRECISELY WHAT <- THEORY IS AND WE DONT UNDER
STAND HOW AS YET UNOBSERVED SYMMETRIES IN PARTICULAR SPACE TIME SUPER
SYMMETRY ARE BROKEN AND HOW A HIERARCHY OF LOW ENERGY SCALES IS GENERATED
&OR NOW C | TYPICAL VELOCITIES
SPECIAL RELATIVISTIC NON RELATIVISTIC

I QUANTUM lELD THEORY QUANTUM MECHANICS
WEAK INTERACTION STRENGTHS BETWEEN

COMPOUNDS SATURATION OF FORCES
j SMALL V { C
&OR SOME RESULTS SEE ;$=
NON RELATIVISTIC
II QUANTUM MECHANICS .EWTONIAN MECHANICS
MANY PARTICLES WITH

ATTRACTIVE INTERACTIONS
WEAK INTERACTIONS
NON RELATIVISTIC
CLASSICAL lELD THEORY
! SIMPLE EXAMPLE OF A SOLUTION OF PROBLEM II IS THE MAIN SUBJECT OF THESE

NOTES
7ARNING 3INCE IN NATURAL UNITS C Y @0 THE PROBLEMS
FORMULATED ABOVE CANNOT BE SOLVED BY LETTING C_ OR Y OR @0 TEND TO

* &2/(,)#( 4 0 43!) !.$ ( 4 9!5 '!&!
4HE PASSAGE FROM MICROSCOPIC THEORIES AT SHORT DISTANCE [ AND TIME

SCALES TO EkECTIVE THEORIES AT LARGE DISTANCE [ AND TIME SCALES WHICH HOPE
FULLY ARE EASIER TO ANALYZE NECESSITATES A DEEPER AND MATHEMATICALLY MORE
PRECISE UNDERSTANDING OF RENORMALIZATION GROUP METHODS 3EE THE NOTES OF
4 3PENCER AND REFS GIVEN THERE
)N THESE NOTES WE DESCRIBE SOME RECENT RESULTS WHICH PROVIDE AN EXAMPLE
OF A SOLUTION OF PROBLEM II 7E CONSIDER A SYSTEM OF VERY MANY NON
RELATIVISTIC BOSONS WITH VERY WEAK TWO BODY INTERACTIONS AND EXHIBIT A CLASS
OF STATES WITH THE PROPERTY THAT THE TIME EVOLUTION OF EXPECTATION VALUES
OF CERTAIN OPERATORS IN THESE STATES IS APPROXIMATELY DESCRIBED BY CLASSI
CAL (AMILTONIAN EQUATIONS OF MOTION MORE SPECIlCALLY BY THE NON LINEAR
(ARTREE EQUATIONS 7E THEN SKETCH HOW CERTAIN SOLUTIONS OF THE NON LINEAR
(ARTREE EQUATIONS APPROACH SOLUTIONS OF .EWTONIAN EQUATIONS OF MOTION FOR
POINT PARTICLES WHEN THE TWO BODY FORCES ARE ATTRACTIVE AT SHORT DISTANCES
$ETAILS MAY BE FOUND IN ;&R49= 4HE READER IS INVITED TO CONSULT THE NOTES
OF * "OURGAIN FOR MATHEMATICAL RESULTS ON RELATED (AMILTONIAN PARTIAL DIF
FERENTIAL EQUATIONS
,ARGE 3YSTEMS OF .ON 2ELATIVISTIC "OSONS

)N NON RELATIVISTIC QUANTUM MECHANICS PURE STATES ARE UNIT RAYS IN A SEPA
RABLE (ILBERT SPACE ( MIXED STATES ARE DENSITY MATRICES ON ( ! SPECIlC
PHYSICAL SITUATION IS ENCODED IN THE CHOICE OF SOME CONCRETE b ALGEBRA !
OF BOUNDED OPERATORS CONTAINED IN OR EQUAL TO THE ALGEBRA "( OF ALL
BOUNDED OPERATORS ON ( 4HE TIME EVOLUTION OF AN OPERATOR A ! IN THE
(EISENBERG PICTURE IS GIVEN BY
B B
! A AT EI(TYA E_I(TY
WHERE ( B IS A SELFADJOINT OPERATOR DENSELY DElNED ON ( WHOSE SPECTRUM IS

USUALLY ASSUMED TO BE BOUNDED FROM BELOW
!S AN EXAMPLE WE CONSIDER A SYSTEM OF . SPINLESS NON RELATIVISTIC
IDENTICAL QUANTUM MECHANICAL POINT PARTICLES IN PHYSICAL SPACE % MOVING
UNDER THE INmUENCE OF AN EXTERNAL FORCE WITH POTENTIAL 7 AND OF TWO BODY
INTERACTIONS WITH POTENTIAL _h !S AN ALGEBRA ! WE MAY EG CHOOSE
THE ALGEBRA OF SMOOTH BOUNDED FUNCTIONS AX X. ON 2. WHICH ARE
TOTALLY SYMMETRIC IN THEIR . ARGUMENTS )TS SPECTRUM CORRESPONDS TO THE
SPACE OF CONlGURATIONS OF . IDENTICAL POINT PARTICLES
.
4HE (ILBERT SPACE OF STATE VECTORS ( (z IS CHOSEN TO BE A SUBSPACE
6ISIONS IN -ATH /. ! #,!33)#!, ,)-)4 /& 15!.45- 4(%/29
OF THE SPACE , 2. e.

J D XJ OF SQUARE INTEGRABLE FUNCTIONS ON 2
. OF
A SPECIlC SYMMETRY TYPE z UNDER PERMUTATIONS OF THEIR . ARGUMENTS

.
<QUANTUM STATISTICS IE A FUNCTION (z TRANSFORMS UNDER AN IRRE
DUCIBLE REPRESENTATION z OF THE GROUP OF PERMUTATIONS OF F . G WHEN ITS
ARGUMENTS ARE PERMUTED 7HEN z IS THE TRIVIAL REPRESENTATION IS TOTALLY
SYMMETRIC IN ITS . ARGUMENTS AND WE SPEAK OF "OSE[%INSTEIN STATISTICS
WHEN z IS THE ALTERNATING REPRESENTATION IS TOTALLY ANTI SYMMETRIC IN ITS
. ARGUMENTS AND WE SPEAK OF &ERMI[$IRAC STATISTICS (IGHER DIMENSIONAL
REPRESENTATIONS z ARE SOMETIMES CALLED PARASTATISTICS
4HE (AMILTONIAN OF THE SYSTEM IS GIVEN BY
8. q r 8
B . Y
( p( _ J 7 XJ _ h XI _ XJ

J sIJs.
AND WE HAVE CHOSEN UNITS SO mTHAT THE MASS OF A PARTICLE IS

0
)N EQ J A XAJ AND XJ XJ XJ XJ DENOTES THE PO
SITION OF THE J TH PARTICLE IN % IN A lXED SYSTEM OF #ARTESIAN COORDINATES
)F 7 IS EG A SMOOTH POSITIVE FUNCTION ON % AND h IS <SMALL RELATIVE TO
_ IN THE SENSE OF +ATO THEN ( . IS ESSENTIALLY SELFADJOINT ON A NATURAL
DOMAIN DENSE IN , 2. eD XJ 3INCE ( . IS SYMMETRIC UNDER PERMU
TATIONS OF THE PARTICLE LABELS . THE TIME EVOLUTION EXP_I( . TY
.
LEAVES THE SUBSPACES (z OF , 2. eD XJ INVARIANT FOR ARBITRARY z
4HERE ARE STRAIGHTFORWARD GENERALIZATIONS OF THIS CLASS OF SYSTEMS TO
SYSTEMS OF SEVERAL SPECIES OF IDENTICAL PARTICLES SOME OF WHICH MAY HAVE
NON ZERO QUANTUM MECHANICAL SPIN EG TO #OULOMB SYSTEM SEE ;,=
4WO IMPORTANT PROBLEMS IN THE ANALYSIS OF SUCH SYSTEMS ARE SEE ;,=
! 3TABILITY IE THE ENERGY PER PARTICLE IS BOUNDED FROM BELOW BY A
CONSTANT _ UNIFORMLY IN . ( . t _. . AND
" %XISTENCE OF THERMODYNAMIC LIMIT CHOOSE 7 7c SO AS TO CONlNE
THE PARTICLES TO A BOUNDED REGION c y % KEEP THE PARTICLE DENSITY
y .JcJ lXED WHERE JcJ IS THE VOLUME OF c AND TRY TO PASS TO
THE LIMIT c % [ ! NECESSARY CONDITION FOR THE EXISTENCE OF A
THERMODYNAMIC LIMIT IS STABILITY
-UCH OF THEORETICAL CONDENSED MATTER PHYSICS IS THE STUDY OF LARGE SYS
TEMS OF IDENTICAL NON RELATIVISTIC QUANTUM MECHANICAL POINT PARTICLES FOR
WHICH ! AND " HOLD 3UCH SYSTEMS EXHIBIT FASCINATING PHYSICAL PHE
NOMENA SUCH AS "OSE[%INSTEIN CONDENSATION z TRIVIAL REPRESENTATION
TWO BODY INTERACTIONS REPULSIVE SUPERCONDUCTIVITY ,ANDAU AND MARGINAL

* &2/(,)#( 4 0 43!) !.$ ( 4 9!5 '!&!
&ERMI LIQUIDS (ALL LIQUIDS DIkERENT FORMS OF MAGNETISM z ALTERNAT

ING REPRESENTATION 0RECISE MATHEMATICAL UNDERSTANDING OF SOME OF THESE
PHENOMENA IS ONLY BEGINNING TO EMERGE SEE EG ;&34= AND ;#&3= FOR A
REVIEW
4HE TOPIC OF THESE NOTES IS A STUDY OF DILUTE SYSTEMS OF A LARGE NUMBER
OF NEUTRAL ATOMS WITH "OSE[%INSTEIN STATISTICS z TRIVIAL REP AND WITH
WEAK TWO BODY INTERACTIONS )F IN THE INITIAL STATE THE MEAN DISTANCE
BETWEEN ATOMS IS SUnCIENTLY LARGE AND THE AVERAGE KINETIC ENERGY PER ATOM
IS LARGE THEN ONE MAY ARGUE THAT ON SUnCIENTLY SHORT TIME SCALES SHORT
RANGE REPULSION BETWEEN THE ATOMS CAN BE NEGLECTED 4HEN THE SYSTEM CAN
BE DESCRIBED BY A (AMILTONIAN ( . OF THE FORM WHERE _h IS EG THE
POTENTIAL OF A LONG RANGE ATTRACTIVE FORCE )N THIS IDEALIZATION STABILITY !
IS HOWEVER USUALLY VIOLATED
7E PROPOSE TO IDENTIFY A CLASS OF INITIAL STATES FOR SUCH SYSTEMS WITH
THE PROPERTY THAT THE TIME EVOLUTION OF THE EXPECTATION VALUES OF CERTAIN
OPERATORS IN SUCH STATES APPROACHES ONE DESCRIBED BY SOME CLASSICAL NON
RELATIVISTIC (AMILTONIAN 0$% IN THE LIMIT WHERE THE STRENGTH s OF THE
TWO BODY INTERACTION TENDS TO WITH . s_
)N ORDER TO ANALYZE SUCH SYSTEMS IT IS CONVENIENT THOUGH DISPENSABLE
TO MAKE USE OF THE FORMALISM OF <SECOND QUANTIZATION 7E IMAGINE THAT THE
NUMBER . OF ATOMS MAY mUCTUATE EG BECAUSE THE SYSTEM IS COUPLED TO
A RESERVOIR OF ATOMS NOT INCLUDED IN OUR DESCRIPTION )T IS THEN CONVENIENT
TO INTRODUCE THE (ILBERT SPACE

-
& (. ( #
.
.
WHERE (. (zTRIVAL REP WHICH IS COMMONLY CALLED &OCK SPACE ! VECTOR
& IS A SEQUENCE
F. J . (. G
.
WITH

8
K K K . K
.
4HE SCALAR PRODUCT OF TWO VECTORS AND IS GIVEN BY

8
H I . .
.
6ISIONS IN -ATH /. ! #,!33)#!, ,)-)4 /& 15!.45- 4(%/29
WHERE ` ` IS THE USUAL , SCALAR PRODUCT ,ET F BE A TEST FUNCTION ON %

7E DElNE AN <ANNIHILATION OPERATOR AF ON & BY SETTING
AF F
WHERE
:
. P
F X X. . D X F X . X X X.
4HE OPERATOR AF IS UNBOUNDED BUT CLOSABLE 4HE <CREATION OPERATOR

Ab F IS DElNED TO BE THE ADJOINT OF AF ON & )T IS AN UNBOUNDED CLOSABLE
OPERATOR )T IS EASY TO VERIFY THAT
a b a b
AF AG Ab F Ab G
a b
AF Ab G F G L
FOR ARBITRARY TEST FUNCTIONS F AND G %QS ARE CALLED CANONICAL COMMU
TATION RELATIONS 4HE (AMILTONIAN (B ON &OCK SPACE IS GIVEN BY

-
B
( ( . WITH (
.
)N TERMS OF ANNIHILATION AND CREATION OPERATORS ( B CAN BE EXPRESSED AS

FOLLOWS
:
a b
B
( D X RAb X RA X 7 X Ab X AX

: :

_ D X D YAb X Ab Y hX _ Y AY AX

AND WE NOW USE UNITS SUCH THAT Y 4HE TWO BODY POTENTIAL _h IS
CHOSEN AS FOLLOWS
h shC s {
WHERE hC , , WITH ,P ,P 2 D X DEPENDING ON THE ANALYTICAL
METHODS EMPLOYED ONE MAY NEED TO ALSO IMPOSE SOME REGULARITY PROPERTIES
ON lRST AND SECOND DERIVATIVES OF h SEE EG ;(= !N EXAMPLE FOR hC IS
hC X JXJ_ GJXJ_ G {
FOR JXJ LARGE COMPARED TO THE DIAMETER OF AN ATOM AND hC X IS REGULAR NEAR
X 4HE lRST TERM IN hC DESCRIBES VAN DER 7AALS ATTRACTION THE SECOND
TERM GRAVITATIONAL ATTRACTION BETWEEN ATOMS 4HE EXTERNAL POTENTIAL 7 IS
ASSUMED TO BE A SMOOTH POSITIVE FUNCTION

* &2/(,)#( 4 0 43!) !.$ ( 4 9!5 '!&!
)T IS EASY TO CHECK THAT

( . t _#s.
FOR SOME lNITE CONSTANT # &OR . s_ KINETIC AND POTENTIAL ENERGY
TERMS IN ( . BOTH SCALE LIKE s_
.EXT WE INTRODUCE RESCALED ANNIHILATION AND CREATION OPERATORS
B B
!s F T s EI(T AF E_I(T
Y AND
!bs F T !s F T b
,ET BE A VECTOR IN &OCK SPACE & WITH A lNITE NUMBER OF PARTICLES AND
SMOOTH K PARTICLE WAVE FUNCTIONS OF COMPACT SUPPORT FOR ALL K 7E DElNE
A VECTOR q s BY SETTING
qs #s Ab .s
WHERE IS A SMOOTH FUNCTION OF COMPACT SUPPORT WHICH FOR CONVENIENCE
WE CHOOSE TO HAVE K K
.s ;s_ =
WHERE ;A= DENOTES THE INTEGER PART OF A POSITIVE NUMBER A AND THE CONSTANT
#s IS CHOSEN SUCH THAT Hqs q s I
,ET X T BE A SOLUTION OF THE NON LINEAR (ARTREE EQUATION
_ `
I T _ 7 _ hC b JJ
WITH INITIAL CONDITION
X X
7E DElNE :
F T D XF X X T
AND
b F T F T
WHERE F IS A TEST FUNCTION
4HE FOLLOWING THEOREM IS PROVEN IN ;(=
4HEOREM 5NDER THE HYPOTHESES DESCRIBED ABOVE
{ 9 M | 9M
LIM q s !bs FJ TJ !s GJ SJ qs b FJ TJ GJ SJ
s
J J
6ISIONS IN -ATH /. ! #,!33)#!, ,)-)4 /& 15!.45- 4(%/29
FOR ARBITRARY M ARBITRARY TEST FUNCTIONS F FM G GM

AND lNITE TIMES T TM S SM
4HUS A CLASSICAL LIMIT DESCRIBED BY THE NON LINEAR (ARTREE EQUATION
IS APPROACHED IN SYSTEMS OF . w s_ BOSONS INTERACTING THROUGH TWO BODY
FORCES OF STRENGTH s AS s TENDS TO FOR AN APPROPRIATE CHOICE OF THE INITIAL
STATE OF THE SYSTEM
)N THE FOLLOWING WE SHOW THAT THE NON LINEAR (ARTREE EQUATION IS AN
EXAMPLE OF A (AMILTONIAN 0$% AND THEN EXPLORE A .EWTONIAN POINT PARTICLE
LIMIT OF THE NON LINEAR (ARTREE EQUATION FOR TWO BODY POTENTIALS WHICH ARE
ATTRACTIVE AT SHORT DISTANCES
4HE THEOREM QUOTED ABOVE IS BUT ONE EXAMPLE OF A RESULT ON THE AP
PROACH OF QUANTUM DYNAMICS TOWARDS SOME CLASSICAL (AMILTONIAN DYNAMICS
IN A SUITABLE LIMIT !NOTHER EXAMPLE WOULD BE A QUANTUM SPIN SYSTEM WHOSE
DYNAMICS IS EXPECTED TO APPROACH THE ONE DESCRIBED BY THE ,ANDAU[,IFSHITZ
EQUATION IN THE LIMIT WHERE THE LOCAL MAGNETIZATION BECOMES LARGE
!N UNSATISFACTORY ASPECT OF RESULTS OF THIS TYPE IS THAT ONE MUST CHOOSE
RATHER SPECIAL SEQUENCES OF STATES Fq s G IN ORDER TO APPROACH A CLASSICAL
LIMIT 3TATES LEADING TO A CLASSICAL LIMIT DO NOT FORM A LINEAR SPACE WHICH
MAY EXPLAIN WHY THERE ARE NO PUZZLES RELATED TO THE SUPERPOSITION PRIN
CIPLE 7E DO NOT UNDERSTAND MUCH YET ABOUT A GENERAL CLASS OF STATES
LEADING TO CLASSICAL LIMITS AND IN WHICH PHYSICAL SITUATIONS SUCH STATES AP
PEAR NATURALLY
4HE MATERIAL IN THE FOLLOWING SECTIONS IS MOTIVATED BY THE ABOVE THE
OREM EQ 7E STUDY THE ., (ARTREE EQUATION AND THE QUESTION
UNDER WHAT CIRCUMSTANCES IT REDUCES TO THE .EWTONIAN MECHANICS OF POINT
PARTICLES 7E BEGIN WITH A BRIEF REVIEW OF THE (AMILTONIAN NATURE OF THE
NON LINEAR (ARTREE EQUATIONS
., (ARTREE %QUATIONS AS A (AMILTONIAN 3YSTEM

4HE PHASE SPACE FOR THE NON LINEAR (ARTREE EQUATION IS THE ENERGY
3OBOLEV SPACE
` ( 2
v
WITH COMPLEX COORDINATES X X 4HE SYMPLECTIC TWO FORM IS GIVEN BY
I
D > D v
)T LEADS TO THE FOLLOWING 0OISSON BRACKETS

g h g h
v
X Y X v
Y
g h
v
X Y ImX _ Y

* &2/(,)#( 4 0 43!) !.$ ( 4 9!5 '!&!
4HE (AMILTON FUNCTIONAL ( LEADING TO THE ., (ARTREE EQUATIONS IS

GIVEN BY
:
v a b
( JRJ 7 JJ _ hC b JJ JJ

2 2
WITH X ` 2 D X` AND b DENOTES CONVOLUTION &OR hC
, , ( IS WELL DElNED ON ` 3OBOLEV AND (OLDER INEQUALITIES AND
IT IS BOUNDED BELOW ON THE SPHERES
g k h
v .
3. k ` .
IN PHASE SPACE ` WHERE
:
v
. JJ
SEE EG ;'6= (AMILTONS EQUATIONS OF MOTION FOR ARE

g h
? T (
X v X T
AND ARE EASILY SEEN TO BE EQUIVALENT TO THE ., (ARTREE EQUATION &ROM
THE FORM OF ( WE INFER THE FOLLOWING SYMMETRIES AND CORRESPONDING CONSER
VATION LAWS
'AUGE INVARIANCE OF THE lRST KIND
X EIq X v
X v
E_Iq X
LEAVES ( INVARIANT IS A SYMPLECTIC mOW GENERATED BY THE (AMILTONIAN
v 4HUS .
VECTOR lELD ASSOCIATED WITH THE FUNCTIONAL . v MUST
BE CONSERVED BY THE TIME EVOLUTION GENERATED BY THE (AMILTONIAN VECTOR
lELD DETERMINED BY ( WHICH MEANS THAT
F( . G
)T FOLLOWS THAT THE SPHERES 3. DElNED IN ARE INVARIANT UNDER THE TIME
EVOLUTION
)F 7 AND IF hC IS ROTATION INVARIANT THEN ARBITRARY 'ALILEI TRANS
FORMATIONS ARE SYMMETRIES OF ( 3PACE TRANSLATIONS ARE GENERATED BY THE
MOMENTUM FUNCTIONAL
:
0v I v
R

AND ROTATIONS BY THE ANGULAR MOMENTUM FUNCTIONAL
:
v I v > R
, X

6ISIONS IN -ATH /. ! #,!33)#!, ,)-)4 /& 15!.45- 4(%/29
!LL COMPONENTS OF 0 AND , 0OISSON COMMUTE WITH ( 6ELOCITY TRANS

FORMATIONS BOOSTS IE X X _ VT ARE REPRESENTED ON TIME DEPENDENT
TRAJECTORIES X T IN ` BY
T
X T V X T X _ VT T EIV`X_V
4HESE TRANSFORMATIONS MAP SOLUTIONS OF THE EQUATIONS OF MOTION TO

OTHER SOLUTIONS OF )T FOLLOWS THAT THE <CENTER OF MASS MOTION OF A
SOLUTION X T OF IS INERTIAL
)NSTEAD OF THE (AMILTONIAN FORMALISM WE MAY MAKE USE OF THE ,A
GRANGIAN FORMALISM TO STUDY THE (ARTREE EQUATIONS 4HESE EQUATIONS
ARE THEN OBTAINED AS THE %ULER[,AGRANGE EQUATIONS CORRESPONDING TO THE
ACTION FUNCTIONAL
: T s : t
I _ `
v
3 DT v v
` T ` T _ ( ` T ` T
T T
BY VARIATION WITH RESPECT TO v WITH m `

v TI FOR I
%XISTENCE AND REGULARITY OF GLOBAL SOLUTIONS OF THE #AUCHY PROBLEM AS
SOCIATED WITH THE NON LINEAR (ARTREE EQUATION UNDER THE CONDITIONS ON hC
DESCRIBED ABOVE IS KNOWN )F _hC IS RADIALLY SYMMETRIC DECREASING RE
PULSIVE TWO BODY FORCES AND OF SUnCIENTLY SHORT RANGE THEN THE SCATTERING
MAP ASSOCIATED WITH THE NON LINEAR (ARTREE EQUATIONS EXISTS AND SATISlES
ASYMPTOTIC COMPLETENESS 4HESE RESULTS ARE REVIEWED IN ;'6= )N THE SITUA
TION OF INTEREST IN THESE NOTES _hC RADIALLY SYMMETRIC BUT INCREASING IE
FOR ATTRACTIVE TWO BODY FORCES EXISTENCE AND REGULARITY ARE KNOWN FROM THE
RESULTS REVIEWED IN ;'6= BUT THE SCATTERING THEORY IS MUCH MORE SUBTLE
AND ONLY VERY PRELIMINARY RESULTS ARE AVAILABLE /UR MAIN INTEREST IS IN THE
.EWTONIAN POINT PARTICLE LIMIT OF THE ., (ARTREE EQUATIONS FOR TWO BODY
FORCES WHICH AT SMALL DISTANCES ARE ATTRACTIVE
4HE 0OINT 0ARTICLE ,IMIT OF THE ., (ARTREE %QUATIONS

WITH !TTRACTIVE 4WO "ODY &ORCES
&OR REASONS THAT WILL BECOME APPARENT SHORTLY WE START BY TRYING TO lND
LOCAL AND ABSOLUTE MINIMA OF THE (AMILTON FUNCTIONAL ( RESTRICTED TO A
SPHERE 3. AS DElNED IN FOR A VANISHING EXTERNAL POTENTIAL 7
)N THIS SITUATION THE MOMENTUM AND ANGULAR MOMENTUM FUNCTIONALS 0
AND , ARE CONSERVED 7E SHOULD THEREFORE STUDY THE GENERALIZED ENERGY

* &2/(,)#( 4 0 43!) !.$ ( 4 9!5 '!&!
FUNCTIONAL
_ `
v v x t (
% v v v
. _ .
_ ` _ `
v t ` , _ ,
x ` 0 _ 0 v
WHERE v x AND t ARE ,AGRANGE MULTIPLIERS 6ARIATION WITH RESPECT TO
v v x AND t YIELDS THE FOLLOWING VARIATIONAL EQUATIONS

_ `
_ _ hC b JJ _ v _ Ix ` R _ It ` X > R
AS WELL AS THE COMPLEX CONJUGATE EQUATION AND
v .
. v 0
0 v ,
,
7E DO NOT KNOW MUCH ABOUT THE GENERAL SOLUTIONS TO THESE EQUATIONS "UT
FOR THE PURPOSES OF THESE NOTES THE FOLLOWING SPECIAL CASE IS PARTICULARLY
IMPORTANT 7E SEARCH FOR AN ABSOLUTE MINIMUM 1. OF ( v J3 WHICH
.
IS ROTATION INVARIANT AND HAS ZERO MOMENTUM 4HE FUNCTION 1. IS THE
SOLUTION OF THE EQUATIONS
_ `
_ _ hC b JJ v
v .
.
AND WE HENCEFORTH SET v % )T FOLLOWS THAT
:
a b
% JR1. J _ hC b J1. J J1. J
.
WHICH IS STRICTLY NEGATIVE FOR A NON TRIVIAL MINIMUM 1.
&AIRLY STANDARD METHODS OF VARIATIONAL CALCULUS SHOW THAT THERE EXISTS
SOME lNITE .b .b hC SUCH THAT FOR . .b EQS HAVE A NON
POSITIVE SOLUTION 1. X WHICH DECAYS EXPONENTIALLY IN JXJ
TRIVIAL SMOOTH P
WITH DECAY RATE _% &ROM THE THEORY OF BINDING IN 3CHR ODINGER QUANTUM
MECHANICS ONE INFERS THAT .b hC IF hC X IS POSITIVE AND DECREASES
IN JXJ MORE SLOWLY THAN JXJ_ BUT .b hC IF hC IS OF SHORT RANGE
#OVARIANCE UNDER TRANSLATIONS AND VELOCITY TRANSFORMATIONS THEN IMPLIES
THAT
V
. V A X T 1. X _ A _ VT EIV`X_;
%=T

SOLVES THE NON LINEAR (ARTREE EQUATIONS FOR AN ARBITRARY VELOCITY V2
)T DESCRIBES A <SOLITON TRAVELING WITH
k VELOCITY V
,ET ( DENOTE THE (ESSIAN OF (k3 AT 1. 4HEN ONE lNDS THAT
.

1. ( 1. 1. 1.
6ISIONS IN -ATH /. ! #,!33)#!, ,)-)4 /& 15!.45- 4(%/29
)N FACT ( HAS PRECISELY ONE EIGENVECTOR } CORRESPONDING TO A NEGATIVE

EIGENVALUE &URTHERMORE
( XA 1. FOR A
WHICH FOLLOWS FROM THE TRANSLATION INVARIANCE OF ( AND THE CONTINUOUS
SPECTRUM OF ( COVERS THE INTERVAL ;_% WITH _%
!SSUMPTIONb )T WILL BE ASSUMED HENCEFORTH THAT hC CAN BE CHOSEN SUCH
THAT THE MULTIPLICITY OF THE EIGENVALUE OF ( IS EQUAL TO
4HE EIGENSPACE OF ( CORRESPONDING TO THE EIGENVALUE IS THEN SPANNED
BY THE VECTORS X 1. X 1. X 1. AND IT FOLLOWS THAT
} ( } t } } s _%
FOR ARBITRARY } , 2 WITH } 1. XA 1. A

7E ARE NOW PREPARED TO DISCUSS THE .EWTONIAN POINT PARTICLE LIMIT OF THE
NON LINEAR (ARTREE EQUATION /UR DISCUSSION IS HEURISTIC FOR MATHEMATICAL
RESULTS SEE ;&R49=
7E CHOOSE AN EXTERNAL POTENTIAL 7 OF THE FORM
7 X p 7 X 6 X
WHERE 6 IS A POSITIVE SMOOTH FUNCTION ON 2 AND { IS A PARAMETER
&URTHERMORE THE TWO BODY POTENTIAL _hC IS CHOSEN TO BE
hC X hS X h@ X
WHERE hS X IS A ROTATION INVARIANT SMOOTH FUNCTION DECAYING RAPIDLY IN
R JXJ AS R AND WITH THE PROPERTIES THAT
hS R
R FOR R
AND THAT !SSUMPTIONb ABOVE HOLDS FOR hC hS 4HE POTENTIAL h@ IS
ROTATION INVARIANT AND MAY BE OF LONG RANGE EG
Jh@ X J w R_ AS R
&OR TECHNICAL CONVENIENCE WE ALSO ASSUME THAT h@ IS ONCE CONTINUOUSLY
DIkERENTIABLE WITH J h@ RJ UNIFORMLY BOUNDED IN R
.EXT WE PICK A POSITIVE INTEGER K AND K POSITIVE NUMBERS . .K
WITH .J .b hS FOR ALL J &OR AND . .b hS WE DElNE
S :
m. . _ D X1. X X

* &2/(,)#( 4 0 43!) !.$ ( 4 9!5 '!&!
7E CONSIDER A CONlGURATION OF K SOLITONS 1.J X_QJ QJ 2 J K

WITH THE FOLLOWING PROPERTIES %ACH 1.J X IS A SOLUTION OF EQUATIONS
WITH . .J FOR hC hS h@ IE AND
o po p
MAX m.J MIN JQI _ QJ J s
JK sIJsK
/UR GOAL IS TO CONSTRUCT A SOLUTION OF THE NON LINEAR (ARTREE EQUA

TION OF THE FORM
K
8 _ `
X T 1.J X _ QJ T EIqJ XT H X T
J
WHERE QJ QJ Q?J VJ J K WITH THE FOLLOWING PROPERTIES

4HERE IS A POSITIVE CONSTANT 4 SUCH THAT FOR ALL TIMES T WITH JTJ 4
I KH ` T K( O
a b
II qJ X T Q?J T ` X _ QJ T J T
WHERE J T IS INDEPENDENT OF X THE TRAJECTORIES Q T QK T AND THE
PHASES T K T WILL TURN OUT TO BE DETERMINED BY THE NON LINEAR
(ARTREE EQUATIONS AND PROPERTIES I AND II
)N ORDER TO CONVINCE OURSELVES THAT THE ANSATZ WITH PROPERTIES
I AND II IS PLAUSIBLE WE EVALUATE THE ACTION FUNCTIONAL 3 v INTRO
v v
DUCED IN EQ ON THE ANSATZ FUNCTIONS WITH AS
IN 7E SHALL REGARD THE SOLITON TRAJECTORIES Q T QK T THE PHASES
T K T AND THE SMALL AMPLITUDE WAVE H AS VARIATIONAL PARAMETERS
OF THE ANSATZ v !FTER A MODERATELY LENGTHY CALCULATION ONE lNDS
THAT FOR A CERTAIN CHOICE OF J T
: { 8K s
.J
3v DT Q?J T _ .J 6 QJ T
_{
J
8 t
_ `
.I .J h@ QI T _ QJ T O
IIJ
FOR { 4 5P TO ERROR TERMS w O WHICH ARE THE SOURCE OF A VARIETY
OF MATHEMATICAL SUBTLETIES THIS IS THE ACTION FOR A SYSTEM OF K .EWTONIAN
POINT PARTICLES WITH MASSES . .K MOVING UNDER THE INmUENCE OF AN
EXTERNAL ACCELERATION lELD WITH POTENTIAL 7 AND OF TWO BODY FORCES WITH
POTENTIAL _.I .J h@ QI _ QJ s I J s K QJ T IS THE POSITION OF THE
J TH PARTICLE AT TIME T
6ISIONS IN -ATH /. ! #,!33)#!, ,)-)4 /& 15!.45- 4(%/29
%XPRESSION FOR 3v SUGGESTS THAT FOR 3

v TO HAVE A CRITICAL
v v
POINT AT WITH GIVEN BY AND SATISFYING PROPERTIES I
AND II THE PARTICLE TRAJECTORIES FQJ T GKJ MUST BE SOLUTIONS OF .EWTONS
EQUATIONS OF MOTION
H 8 _ Ì
QJ T _ R6 QJ T _ .I Rh@ QI T _ QJ T
IJ
AJ T J K
WHERE THE <FRICTION FORCE AJ T SATISlES A BOUND JAJ T J O FOR JTJ 4
4HESE IDEAS ARE MADE PRECISE IN THE FORM OF MATHEMATICAL THEOREMS
IN ;&R49= USING METHODS PARTLY BASED ON THE WORK OF 7EINSTEIN ;7= FOR
JUST ONE SOLITON IN AN EXTERNAL POTENTIAL 7 "UT THE METHODS IN ;&R49=
CAN BE EXTENDED TO GENERAL K
!PPARENTLY WE HAVE GAINED INSIGHT INTO HOW .EWTONIAN POINT PARTICLE
MECHANICS WITH SMALL FRICTION FORCES CAN EMERGE FROM THE NON LINEAR (ARTREE
EQUATION WITH ATTRACTIVE TWO BODY FORCES OF SHORT RANGE hS AND WEAK AT
TRACTIVE OR REPULSIVE TWO BODY FORCES OF VERY LONG RANGE h@ ` BY CON
SIDERING THE EVOLUTION OF CONlGURATIONS OF FAR SEPARATED SOLITONS OVER LARGE
lNITE INTERVALS OF TIME 4HE FRICTION FORCES AJ T APPEARING IN .EWTONS
EQUATIONS OF MOTION ORIGINATE IN THE EMISSION OF SMALL AMPLITUDE DISPER
SIVE RADIATION FROM THE SOLITONS 4HIS IS AN EXAMPLE OF A GENERAL AND QUITE
FUNDAMENTAL PHENOMENON FRICTION OR DISSIPATION THROUGH EMISSION OF
RADIATION IN CONSERVATIVE CLASSICAL AND QUANTUM MECHANICAL SYSTEMS OF
INlNITELY MANY DEGREES OF FREEDOM
&OR SIMILAR CONSIDERATIONS IN THE CONTEXT OF VORTEX DYNAMICS IN A 'INZ
BURG[,ANDAU SUPERmUID SEE ;/3=
4HE RESULTS DISCUSSED ABOVE APPEAR TO BE A SPECIAL CASE OF MORE GENERAL
RESULTS ON THE .EWTONIAN LIMIT OF THE NON LINEAR (ARTREE EQUATION /NE
MIGHT IMAGINE THAT IT IS POSSIBLE TO CONSTRUCT <ROTATING BODY SOLUTIONS OF
THE NON LINEAR (ARTREE EQUATIONS WITH NON ZERO ANGULAR MOMENTUM RELATIVE
TO THEIR CENTER OF MASS 3UCH SOLUTIONS ARE IN GENERAL UNSTABLE .EVER
THELESS ONE COULD ATTEMPT TO CONSTRUCT SOLUTIONS OF THE NON LINEAR (ARTREE
EQUATION WHICH OVER A LARGE lNITE INTERVAL OF TIME DESCRIBE A CONlGURATION
OF lNITELY MANY ROTATING BODIES WHOSE MOTION IS DESCRIBED BY A SYSTEM OF
%ULER TYPE EQUATIONS WITH SMALL FRICTION TERMS EMITTING SMALL AMPLITUDE
DISPERSIVE RADIATION
4HE RESULTS REVIEWED IN EQS THROUGH AND THE SPECULATIONS JUST
DESCRIBED RAISE THE ISSUE OF ASYMPTOTIC PROPERTIES OF THE DYNAMICS DETER

* &2/(,)#( 4 0 43!) !.$ ( 4 9!5 '!&!
MINED BY THE NON LINEAR (ARTREE EQUATION IN THE LIMIT AS TIME TENDS TO
e 3UCH PROBLEMS BELONG TO THE REALM OF NON LINEAR SCATTERING THEORY
)N ;&R49= A RESULT ON THE SCATTERING OF SMALL AMPLITUDE WAVES Ok A SINGLE
SOLITON HAS BEEN ESTABLISHED )T REQUIRES 7 X AND hC X TO BE SMOOTH AND
OF SHORT RANGE 7E CONSIDER AN <ASYMPTOTIC PROlLE GIVEN BY
%=T
AS X T 1. X _ VT EIX`V_;V HX T
WHERE H IS A SOLUTION OF THE FREE PARTICLE 3CHR
ODINGER EQUATION
I T HX T _ HX T
> 4HE CLAIM
WITH HX ( 2 < 7 2 JXJ D X AND HV
IS THAT THERE ARE SOLUTIONS e X T OF THE NON LINEAR (ARTREE EQUATION
SUCH THAT
e X T _ AS X T
T
IN ( 2
4HUS NON LINEAR -|LLER WAVE MAPS je EXIST AS SYMPLECTIC MAPS
ON ASYMPTOTIC PROlLES OF THE FORM SEE ALSO ;'6= FOR BACKGROUND
MATERIAL ON NON LINEAR SCATTERING THEORY /F COURSE THE RESULTS DESCRIBED
HERE DO NOT BRING US ANYWHERE CLOSE TO AN UNDERSTANDING OF ASYMPTOTIC
COMPLETENESS OF THE SCATTERING MAP
-ORE !BOUT /PEN 0ROBLEMS #ONCLUSIONS

)N THIS LAST SECTION WE DRAW THE READERS ATTENTION TO SOME INTERESTING OPEN
PROBLEMS
#LEARLY A DEEPER UNDERSTANDING OF THE NON LINEAR SCATTERING THEORY
OF SOLITONS AND WAVES FOR THE NON LINEAR (ARTREE EQUATION WOULD BE HIGHLY
DESIRABLE )N PARTICULAR IT WOULD BE IMPORTANT TO GAIN MORE DETAILED INSIGHT
INTO THE DOMAIN OF DElNITION AND THE IMAGE OF THE -|LLER WAVE MAPS je
! COMPARATIVELY SIMPLE SITUATION HAS BEEN STUDIED BY 3OkER AND 7E
INSTEIN ;37= 4HEY CONSIDER A SMOOTH EXTERNAL POTENTIAL 7 OF SHORT RANGE
WITH EXACTLY ONE QUANTUM MECHANICAL BOUND STATE IE THE SPECTRUM OF THE
OPERATOR _ 7 CONSISTS OF ONE SIMPLE EIGENVALUE AND OF PURELY
ABSOLUTELY CONTINUOUS SPECTRUM COVERING ; /NE MUST ALSO ASSUME
THAT THERE IS NO ZERO ENERGY RESONANCE 4HEY CONSIDER A LOCAL NON LINEARITY
., 3CHRODINGER EQUATION BUT THEIR TECHNIQUES CAN BE EXTENDED TO THE .,
(ARTREE EQUATION WITH hC OF SHORT RANGE 4HEY CONSTRUCT SMALL AMPLITUDE
6ISIONS IN -ATH /. ! #,!33)#!, ,)-)4 /& 15!.45- 4(%/29
NON LINEAR BOUND STATES STATIONARY SOLUTIONS OF BY STANDARD BIFUR

CATION THEORY TECHNIQUES 4HEY THEN EXHIBIT A CLASS OF INITIAL CONDITIONS
WITH THE PROPERTY THAT THE ASYMPTOTIC BEHAVIOR OF SOLUTIONS OF WITH
THESE INITIAL CONDITIONS IS THAT OF A NON LINEAR BOUND STATE PLUS A PURELY
DISPERSIVE WAVE WHICH TENDS TO ZERO POINTWISE AT THE FREE DISPERSION RATE
AS T e 4HEY ALSO SHOW THAT INITIAL CONDITIONS CLOSE TO A SMALL
AMPLITUDE NON LINEAR BOUND STATE HAVE THIS PROPERTY
/BVIOUSLY ONE WOULD LIKE TO EXTEND THESE RESULTS TO EXTERNAL POTENTIALS
7 WITH MORE THAN ONE BOUND STATE AND UNDERSTAND THE RADIATIVE DECAY OF
INITIAL CONDITIONS CLOSE TO AN EXCITED BOUNDSTATE OF _ 7 TO A NON LINEAR
GROUNDSTATE PLUS DISPERSIVE RADIATION /NE WOULD ALSO LIKE TO GO BEYOND RE
SULTS FOR SMALL AMPLITUDE INITIAL CONDITIONS )N THIS CASE IT CAN HAPPEN
THAT THE RESTRICTION OF THE (AMILTON FUNCTIONAL ( v SEE EQ OF THE
NON LINEAR (ARTREE EQUATION TO A SPHERE 3. IN PHASE SPACE ` SEE
HAS SEVERAL DISTINCT LOCAL QUADRATIC MINIMA PROVIDED . IS LARGE ENOUGH
CHOOSE 7 TO HAVE SEVERAL MINIMA SEPARATED BY LARGE BARRIERS AND _hC
ATTRACTIVE /NE WOULD THEN LIKE TO UNDERSTAND THE SHAPE OF THE <BASINS
OF ATTRACTION IN PHASE SPACE OF THESE LOCAL MINIMA ;4HE FORWARD BACK
WARD
k BASIN OF ATTRACTION OF A ONE PARAMETER FAMILY OF LOCAL MINIMA OF
(k3 PARAMETRIZED BY . CONSISTS OF ALL INITIAL CONDITIONS IN PHASE SPACE
.
` WHICH APPROACH AN ELEMENT OF THIS FAMILY PLUS DISPERSIVE RADIATION DE
CAYING TO ZERO AS T T _ RESPECTIVELY -ORE GENERALLY ONE
MAY INTRODUCE A NOTION OF A CENTER MANIFOLD OF ASYMPTOTICALLY ATTRACTING
CONlGURATIONS OF SOLITONS FREELY MOVING FOR 7 TO WHICH SOLUTIONS OF
THE NON LINEAR (ARTREE EQUATION WITH INITIAL CONDITIONS SUnCIENTLY CLOSE TO
THE CENTER MANIFOLD CONVERGE AS T OR T _ )N THIS CONNECTION
SEE EG ;07==
#ONSIDER A CONlGURATION OF SAY JUST TWO FAR SEPARATED SOLITONS
WITH INITIAL POSITIONS Q Q AND INITIAL VELOCITIES V Q? V Q? AS
DESCRIBED IN EQ WITH H X ,ET US ASSUME THAT THE TWO BODY
FORCE WITH POTENTIAL _h@ IS PURELY ATTRACTIVE AND OF SHORT RANGE 7E CHOOSE
THE INITIAL CONDITIONS . Q V AND . Q V IN SUCH A WAY THAT THE
TWO SOLITONS FORM A BOUND SYSTEM IN THE SENSE THAT
. .
_ `
V V _ . . h@ Q _ Q
7E WOULD LIKE TO UNDERSTAND THE FATE OF THIS CONlGURATION AS T 4O
BEGIN WITH WE WOULD LIKE TO ESTIMATE THE POWER 02 T OF DISPERSIVE RA
DIATION THROUGH A SPHERE OF RADIUS 2 | MAXJQ J JQ J SEE ;/3= FOR SOME
RESULTS ON A RELATED PROBLEM &URTHERMORE WE WOULD LIKE TO SHOW THAT

* &2/(,)#( 4 0 43!) !.$ ( 4 9!5 '!&!
A CONlGURATION OF TWO BOUND SOLITONS COLLAPSES TO A SINGLE SOLITON MOVING

THROUGH SPACE WITH A CONSTANT VELOCITY FOR 7 PLUS A DISPERSIVE OUT
GOING WAVE TENDING TO / POINTWISE AT THE FREE DISPERSION RATE <RADIATIVE
COLLAPSE OF A BINARY SYSTEM
)T APPEARS REASONABLE TO IMAGINE THAT IF THE POTENTIAL h@ IS CHOSEN
TO BE GIVEN BY
h@ X JXJ_
FOR JXJ LARGE hS AS IN SECT AND 7 THE NON LINEAR (ARTREE EQUA
TION CAN BE USED TO DESCRIBE .EWTONIAN GRAVITY OF A GIANT CLOUD OF
BOSONIC ATOMS WITH BUILT IN DISPERSIVE FRICTION )T WOULD BE INTERESTING TO
UNDERSTAND HOW ON INTERMEDIATE TIME SCALES INHOMOGENEITIES IN THE INI
TIAL CONDITIONS GROW TO FORM A STRUCTURE CONSISTING OF ROTATING BODIES PLUS
DISPERSIVE RADIATION <STRUCTURE FORMATION BEFORE THE SYSTEM COLLAPSES
INTO A NUMBER OF FAR SEPARATED SOLITONS ESCAPING FROM EACH OTHER 0ERHAPS
ONE MAY GAIN SOME INSIGHT INTO SUCH PROBLEMS WITH THE HELP OF NUMERICAL
SIMULATIONS SEE EG ;-+.:=
)T IS OF INTEREST TO STUDY TRANSPORT PHENOMENA DESCRIBED BY THE
NON LINEAR (ARTREE EQUATION WITH A RANDOM EXTERNAL POTENTIAL 7 WHOSE
DISTRIBUTION IS HOMOGENEOUS AND HAS CORRELATIONS OF SHORT RANGE &OR SOME
RESULTS IN THIS DIRECTION SEE ;!&=
0ERHAPS MOST IMPORTANTLY WE WOULD LIKE TO UNDERSTAND MUCH MORE
ABOUT HOW NON LINEAR EQUATIONS OF CLASSICAL PHYSICS EMERGE AS DESCRIPTIONS
OF QUANTUM SYSTEMS IN CERTAIN ASYMPTOTIC REGIMES
\\\\\\\\\\\\\
0ROGRESS IN THEORETICAL PHYSICS IS ACHIEVED IN MANY DIkERENT WAYS )F

ONE IS VERY LUCKY AND INGENIOUS ONE MAY DISCOVER NEW ,AWS OF .ATURE OR
NEW PHYSICAL THEORIES OR A NEW MATHEMATICAL STRUCTURE USEFUL TO FORMULATE
SUCH THEORIES -ORE MODESTLY ONE MAY ATTEMPT TO IMPROVE THE MATHEMAT
ICAL UNDERSTANDING OF KNOWN PHYSICAL THEORIES OR DERIVE NEW CONSEQUENCES
OF SUCH THEORIES )T APPEARS SAFE TO PREDICT THAT WE WILL BE SAILING WELL INTO
THE ST CENTURY WITH PLENTY OF INTERESTING QUESTIONS ABOUT PHYSICAL THEO
RIES DISCOVERED DURING THE PAST THREE HUNDRED YEARS TO WORRY ABOUT AND
THAT <ANALYSIS AND PHYSICS WILL BE A WORTHY PARTNER OF THE CURRENTLY MORE
POPULAR <GEOMETRY AND PHYSICS
!CKNOWLEDGMENTS 7E THANK * "OURGAIN AND ! 3OkER AND ESPE
CIALLY )- 3IGAL AND 4 3PENCER FOR VERY HELPFUL DISCUSSIONS 4HE WORK
6ISIONS IN -ATH /. ! #,!33)#!, ,)-)4 /& 15!.45- 4(%/29
OF (4 9AU HAS BEEN PARTIALLY SUPPORTED BY THE 53 .3& UNDER GRANT
$-3
2EFERENCES
;!&= # !LBANESE * &RO HLICH 0ERIODIC SOLUTIONS OF SOME INlNITE
DIMENSIONAL (AMILTONIAN SYSTEMS ASSOCIATED WITH NON LINEAR PAR
TIAL DIkERENCE EQUATIONS ) #OMMUN -ATH 0HYS [
# !LBANESE * &RO HLICH 4 3PENCER 0ERIODIC SOLUTIONS
OF SOME INlNITE DIMENSIONAL (AMILTONIAN SYSTEMS ASSOCIATED WITH
NON LINEAR PARTIAL DIkERENCE EQUATIONS )) #OMMUN -ATH 0HYS
[
;#&3= 4 #HEN * &RO HLICH - 3EIFERT 2ENORMALIZATION GROUP
METHODS ,ANDAU &ERMI LIQUID AND "#3 SUPERCONDUCTOR IN 0ROC
OF 3ESSION ,8)) OF THE ,ES (OUCHES SUMMER SCHOOLS [ <&LUCTUATING
'EOMETRIES IN 3TATISTICAL -ECHANICS AND &IELD 4HEORY & $AVID
0 'INSPARG * :INN *USTIN EDS %LSEVIER !MSTERDAM .EW 9ORK

;$= * $IMOCK 4HE NON RELATIVISTIC LIMIT OF 0 } QUANTUM lELD THE
ORIES TWO PARTICLE PHENOMENA #OMMUN -ATH 0HYS
[ 1%$ IN THE #OULOMB GAUGE !NN )NST ( 0OINCARsE
[
;&34= * &ELDMAN - 3ALMHOFER % 4RUBOWITZ 2ENORMALIZATION
OF THE &ERMI SURFACE IN 0ROC OF THE 8))TH )NTL #ONGRESS OF -ATH
0HYSICS [ )NTERNATIONAL 0RESS #AMBRIDGE -! AND REF
ERENCES TO PREVIOUS WORK GIVEN THERE
;&R49= * &RO HLICH 4 0 4SAI ( 4 9AU 4HE POINT PARTICLE LIMIT OF
THE NON LINEAR (ARTREE EQUATION TO APPEAR
;'6= * 'INIBRE ' 6ELO /N A CLASS OF NONLINEAR 3CHR ODINGER EQUA
TIONS WITH NONLOCAL INTERACTION -ATH : [ SEE
ALSO * 'INIBRE ' 6ELO 3CATTERING THEORY IN THE ENERGY SPACE
FOR A CLASS OF NONLINEAR 3CHRODINGER EQUATIONS * -ATH 0URE !PPL
[ 3CATTERING THEORY IN THE ENERGY SPACE FOR A CLASS
OF (ARTREE EQUATIONS PREPRINT
;(= + (EPP 4HE CLASSICAL LIMIT FOR QUANTUM MECHANICAL CORRELATION
FUNCTIONS #OMMUN -ATH 0HYS [ SOME OF THE
IDEAS IN THIS PAPER CAN BE TRACED BACK TO % 3CHRO DINGER $ER

STETIGE 5BERGANG VON DER -IKRO ZUR -AKROMECHANIK $IE .ATUR

* &2/(,)#( 4 0 43!) !.$ ( 4 9!5 '!&!
WISSENSCHAFTEN [

;,= %( ,IEB 4HE 3TABILITY OF -ATTER &ROM !TOMS TO 3TARS 3ELECTA
OF %LLIOTT ( ,IEB 3PRINGER 6ERLAG "ERLIN (EIDELBERG .EW 9ORK

;-+.:= !6 -IKHAILOV %! +UZNETSOV !# .EWELL 6% :A
KHAROV EDS <4HE .ONLINEAR 3CHRODINGER EQUATION 0ROC OF CON
FERENCE IN #HERNOGOLOVKA 0HYSICS $ .ORTH (OLLAND !MS
TERDAM .EW 9ORK
;/3= 9U. /VCHINNIKOV )- 3IGAL $YNAMICS OF LOCALIZED STRUC
TURES 0HYSICA ! [ AND REFERENCES GIVEN THERE
EG 9U. /VCHINNIKOV )- 3IGAL 'INZBURG ,ANDAU EQUA
TION ) 'ENERAL DISCUSSION IN <0$%S AND THEIR !PPLICATIONS
, 3ECO ET AL EDS #2- 0ROCEEDINGS AND ,ECTURE .OTES
[
;07= # ! 0ILLET #% 7AYNE )NVARIANT MANIFOLDS FOR A CLASS OF DIS
PERSIVE (AMILTONIAN PARTIAL DIkERENTIAL EQUATIONS * $Ik %QUA
TIONS [
;37= ! 3OFFER -) 7EINSTEIN -ULTICHANNEL NONLINEAR SCATTERING
FOR NONINTEGRABLE EQUATIONS #OMMUN -ATH 0HYS
[ -ULTICHANNEL NONLINEAR SCATTERING FOR NONINTEGRABLE EQUA
TIONS )) THE CASE OF ANISOTROPIC POTENTIALS AND DATA * $Ik %QUA
TIONS [
;7= -) 7EINSTEIN -ODULATION STABILITY OF GROUNDSTATES OF NON
LINEAR 3CHRODINGER EQUATIONS 3)!- * -ATH !NNAL
[
RG &RO
*U HLICH 4HEORETICAL 0HYSICS %4( (ONGGERBERG #([ :
URICH
3WITZERLAND
4AI 0ENG 4SAI AND (ORNG 4ZER 9AU #OURANT )NSTITUTE OF -ATHEMATI
CAL 3CIENCES -ERCER 3TREET .EW 9ORK 5NIVERSITY .EW 9ORK .9
53!
2/5'( 3425#452% !.$ #,!33)&)#!4)/.
74 'OWERS
)NTRODUCTION
7HEN ) WAS lRST ASKED TO SPEAK AT THE <6ISIONS IN -ATHEMATICS CONFERENCE
) HAD WHAT ) BELIEVE WAS A TYPICAL REACTION ) WANTED TO TRY TO EMULATE
(ILBERT A CENTURY AGO BUT SINCE ) KNEW THAT ) COULD NOT POSSIBLY MATCH
HIS BREADTH OF VISION ) WAS FORCED TO MAKE SOME SORT OF COMPROMISE )N
THIS PAPER ) SHALL DISCUSS SEVERAL OPEN PROBLEMS NOT ALWAYS IN AREAS ) KNOW
MUCH ABOUT BUT THEY ARE NOT INTENDED AS A LIST OF THE MOST IMPORTANT
QUESTIONS IN MATHEMATICS OR EVEN THE MOST IMPORTANT QUESTIONS IN THE
AREAS OF MATHEMATICS THAT ) HAVE WORKED IN 2ATHER THEY ARE A PERSONAL
SELECTION OF PROBLEMS THAT FOR ONE REASON OR ANOTHER HAVE CAPTURED MY
ATTENTION OVER THE YEARS
4HE TITLE OF THIS PAPER REFERS TO A GENERAL THEME THAT LINKS SEVERAL UN
SOLVED PROBLEMS IN COMBINATORICS ) BELIEVE THAT WHAT MAKES THEM DInCULT
IS IN MANY CASES THE LACK OF ADEQUATE <ROUGH CLASSIlCATION THEOREMS )
SHALL EXPLAIN WHAT ) MEAN BY THIS IN X AND GIVE EXAMPLES OF PROBLEMS FOR
WHICH CLASSIlCATION MIGHT VERY WELL BE USEFUL )N X ) GIVE A MORE MISCELLA
NEOUS LIST OF QUESTIONS WITH BRIEF ACCOUNTS OF WHY ) lND THEM INTERESTING
-OST BUT NOT ALL OF THE QUESTIONS ON THIS LIST ARE VERY WELL KNOWN
!S THE CONFERENCE DREW NEARER ) DECIDED THAT ) COULD NOT RESIST AT LEAST
SOME ATTEMPT AT <THE VISION THING ;"U= ) AM NOT IN FAVOUR OF PREDICTING
WHICH AREAS OF MATHEMATICS ARE LIKELY TO BE FASHIONABLE IN lFTEEN LET ALONE
lFTY YEARS TIME BUT IN X ) HAVE MADE A MORE GENERAL PREDICTION ABOUT
MATHEMATICS AS A WHOLE "RIEmY MY CONTENTION IS THAT DURING THE NEXT
CENTURY COMPUTERS WILL BECOME SUnCIENTLY GOOD AT PROVING THEOREMS THAT
THE PRACTICE OF PURE MATHEMATICAL RESEARCH WILL BE COMPLETELY REVOLUTION
IZED QUITE POSSIBLY TO THE POINT WHERE WE WOULD BE RELUCTANT TO RECOGNIZE
4HE RESEARCH FOR THIS ARTICLE WAS PARTIALLY SUPPORTED BY THE #LAY -ATHEMATICS
)NSTITUTE
74 '/7%23 '!&!
IT AS MATHEMATICAL RESEARCH AT ALL -Y REASONS FOR THIS OPINION ARE OUTLINED

IN THE NEXT SECTION
) DO NOT CLAIM ANY ORIGINALITY FOR MY VIEWS ON THIS TOPIC )NDEED BY THE
TIME OF MY LECTURE AT THE CONFERENCE SOME OF THE IDEAS HAD BEEN TOUCHED
ON BY OTHER SPEAKERS AND DURING DISCUSSION SESSIONS -OREOVER INTEREST
IN AUTOMATIC THEOREM PROVING HAS EXISTED FOR MANY YEARS AND SINCE THE
CONFERENCE ) HAVE BEEN REFERRED TO SIMILAR VIEWS EXPRESSED BY PEOPLE WHO
WORK IN THE AREA /NE REFERENCE IS GIVEN AT THE END OF THE NEXT SECTION
7ILL -ATHEMATICS %XIST IN

&OR AT LEAST A CENTURY MATHEMATICIANS HAVE THOUGHT ABOUT THE POSSIBILITY
OF AUTOMATING MATHEMATICS (ILBERT FAMOUSLY ASKED IN HIS TENTH PROBLEM
WHETHER THERE WAS AN ALGORITHM FOR SOLVING $IOPHANTINE EQUATIONS AND
LATER EXTENDED THIS TO THE QUESTION OF WHETHER THERE WAS AN ALGORITHM WHICH
WOULD lND A PROOF OF ANY MATHEMATICAL STATEMENT THAT HAD ONE )N
4URING EQUALLY FAMOUSLY FORMALIZED THE NOTION OF ALGORITHM AND SOON AFTER
WARDS DEMONSTRATED THE INSOLUBILITY OF THE HALTING PROBLEM THUS SHOWING
THAT NO SUCH ALGORITHM EXISTED 7HILE THIS RESULT AND LATER DEMONSTRA
TIONS THAT SEVERAL NATURAL AND WELL KNOWN PROBLEMS IN MATHEMATICS WERE
ALSO IMPOSSIBLE TO SOLVE SYSTEMATICALLY MAY HAVE INITIALLY SEEMED SOMEWHAT
NEGATIVE THEY ALSO HAD A POSITIVE SIDE 4HE IDEA THAT ALL OUR CREATIVITY AND
INSIGHT MIGHT BE REDUCED TO SOMETHING MECHANICAL WAS AFTER ALL NOT VERY
APPEALING 4URINGS RESULT THEREFORE CAME AS A RELIEF SINCE IT LEFT MATHE
MATICIANS WITH SOMETHING TO DO
-ORE RECENTLY WITH THE RISE OF THE THEORY OF COMPUTATIONAL COMPLEXITY
AND .0 COMPLETENESS IT HAS BEEN WIDELY UNDERSTOOD THAT THIS RELIEF WAS
SOMEWHAT MISPLACED BECAUSE AS 'ODEL IN FACT APPRECIATED AN ALGORITHM
FOR lNDING PROOFS IS NOT MUCH USE IF IT TAKES AN UNFEASIBLY LONG TIME &AR
MORE THREATENING TO MATHEMATICIANS WOULD BE AN ALGORITHM WHICH COULD
DETERMINE FOR ANY GIVEN MATHEMATICAL STATEMENT WHETHER OR NOT IT HAD A
PROOF OF LENGTH AT MOST N AND DO THIS IN POLYNOMIAL TIME IDEALLY WITH A
POLYNOMIAL OF SMALL DEGREE 4HE EXISTENCE OF AN ALGORITHM OF THIS KIND IS
ONE OF THE MAJOR OPEN PROBLEMS OF MATHEMATICS SINCE IT IS EQUIVALENT TO
WHETHER 0.0 BUT MANY MATHEMATICIANS TAKE COMFORT FROM THE FACT THAT
NOBODY HAS FOUND A SHORT ALGORITHM FOR SOLVING ANY .0 COMPLETE PROBLEM
AND MOST EXPERTS BELIEVE THAT NO SUCH ALGORITHM EXISTS
(OWEVER THIS COMFORT EVEN IF YOU ARE CONVINCED THAT 0.0 IS STILL
6ISIONS IN -ATH 2/5'( 3425#452% !.$ #,!33)&)#!4)/.
MISPLACED BECAUSE SEARCHING FOR SHORTISH PROOFS OF ARBITRARY MATHEMATI

CAL STATEMENTS IS A VERY BAD MODEL OF WHAT MATHEMATICIANS ACTUALLY DO
)NDEED THERE IS ABSOLUTELY NO EVIDENCE THAT WHERE MATHEMATICIANS SCORE
OVER COMPUTERS IS IN THEIR ABILITY TO SOLVE .0 COMPLETE QUESTIONS QUICKLY
)F SOMEBODY WERE TO UNEARTH AN IDIOT SAVANT WHO COULD DETERMINE IN HIS OR
HER HEAD WHETHER COLLECTIONS OF "OOLEAN FORMULAE WERE SATISlABLE OR EVEN
FACTORIZE LARGE INTEGERS THEN ONE WOULD BE FORCED TO ASK SOME PROFOUND
QUESTIONS BUT THAT WILL NOT HAPPEN A PREDICTION ) MAKE WITH GREAT CON
lDENCE 4HEREFORE THE ONLY POSSIBLE EXPLANATION FOR THE GREAT SUCCESS OF
MATHEMATICIANS OVER THE CENTURIES IS THAT OUR SEARCH FOR PROOFS OF THEOREMS
IS VERY RESTRICTED )NDEED THIS IS OBVIOUSLY TRUE [ WE TEND TO PREFER QUES
TIONS THAT ARE INTERESTING COMPREHENSIBLE AND SEEMINGLY NOT COMPLETELY
OUT OF REACH AND WE LIKE OUR PROOFS TO PROVIDE EXPLANATIONS RATHER THAN
JUST FORMAL GUARANTEES OF TRUTH
4HIS LINE OF THOUGHT LEADS TO THE FOLLOWING QUESTION CAN THE NOTION OF A
<GOOD PROOF BE USEFULLY FORMALIZED )F SO THEN THE IMPACT ON MATHEMATICS
COULD WELL BE COMPARABLE TO THE IMPACT OF THE FORMALIZING OF THE NOTIONS
OF PROOF AND ALGORITHM )T IS PERHAPS UNLIKELY THAT A DElNITION WILL EMERGE
WHICH IS AS INDISPUTABLY CORRECT AS THAT OF A RECURSIVE FUNCTION OR 4URING
MACHINE BUT IT MIGHT BE REALISTIC TO TRY TO ISOLATE CERTAIN PROPERTIES OF
WHAT WE COMMONLY REGARD AS GOOD PROOFS AND FORMALIZE THEM (ERE IS AN
INCOMPLETE LIST OF FEATURES OFTEN ASSOCIATED WITH GOOD PROOFS
I 0ROOFS ARE USUALLY CLEARER IF THEY HAVE A HIERARCHICAL STRUCTURE &OR
EXAMPLE IF THE MAIN THEOREM FOLLOWS FROM THREE LEMMAS THAT ALL HAVE CLEAR
STATEMENTS THEN ONE CAN ISOLATE PARTS OF THE PROOF THINK ABOUT THEM SEP
ARATELY AND UNDERSTAND THE ARGUMENT WITHOUT HAVING TO HOLD TOO MUCH IN
ONES HEAD AT ONCE
II -ANY GOOD PROOFS ARE VARIANTS OF EXISTING BETTER KNOWN ARGUMENTS
4HIS MAKES THEM EASIER TO UNDERSTAND BECAUSE ALL ONE HAS TO DO IS CONCEN
TRATE ON THE PARTS THAT ARE NEW
III )T IS EASIER TO UNDERSTAND A PROOF THAT CAN BE REDUCED TO A FEW BASIC
IDEAS .EEDLESS TO SAY IT IS NOT EASY TO FORMALIZE WHAT ONE MEANS BY A
<BASIC IDEA
IV 3OMETIMES THE FORM OF A PROOF IS DICTATED TO SOME EXTENT BY THE
EXISTENCE OF CERTAIN EXAMPLES AND COUNTEREXAMPLES 4HESE CLARIFY THE PROOF
BY SUGGESTING APPROACHES THAT MIGHT WORK AND DEMONSTRATING THAT OTHER
APPROACHES CANNOT WORK
V ! GOOD DElNITION CAN GREATLY IMPROVE A PROOF )N EXTREME CASES
74 '/7%23 '!&!
CHOOSING THE RIGHT DElNITION CAN REDUCE A DInCULT PROBLEM TO A SIMPLE

EXERCISE IN CHECKING
VI -ANY OF THE BEST PROOFS OR AT LEAST THE BEST WRITTEN PROOFS COME
WITH SOME INDICATION OF HOW THEY WERE DISCOVERED 4O PUT THAT THE OTHER
WAY ROUND PROOFS THAT DEPEND ON MAGIC TRICKS WHILE THEY MAY BE IMPRES
SIVE AND ELEGANT ARE NOT COMPLETELY SATISFACTORY AS EXPLANATIONS
4HE ABOVE FEATURES ARE INTERRELATED AND THEY ALL HAVE IN COMMON THAT
A PROOF WITH ONE OF THE FEATURES IS OTHER THINGS BEING EQUAL EASIER TO
UNDERSTAND THAN ONE WITHOUT ) WOULD LIKE TO CONCENTRATE ON VI AND
ADOPT IT AS A WORKING DElNITION OF <EXPLANATION AS OPPOSED TO <GUARANTEE
OF TRUTH 4HAT IS ) SHALL THINK OF AN EXPLANATION AS A PROOF TOGETHER
WITH AN ACCOUNT OF HOW THE PROOF WAS DISCOVERED 4HIS ACCOUNT SHOULD
NOT BE SOMETHING LIKE <) WAS SITTING IN THE BATH AND SUDDENLY HAD THE
FOLLOWING BRILLIANT IDEA BUT RATHER A DEMYSTIlCATION OF THE PROOF THAT IS
A CONVINCING DEMONSTRATION OF HOW A MERE MATHEMATICALLY TRAINED HUMAN
BEING COULD SEARCH FOR A PROOF AND lND THIS ONE WITHOUT RELYING ON EXCESSIVE
AMOUNTS OF LUCK
) AM AWARE THAT THIS DElNITION WHICH WILL BE MADE MORE PRECISE IN A
MOMENT DOES NOT CAPTURE ALL OUR ASSOCIATIONS WITH THE WORD <EXPLANATION
)N PARTICULAR IT MIGHT BE EASIER TO SHOW HOW ONE CAME UP WITH A BRUTE FORCE
ARGUMENT THAN TO SHOW HOW ONE THOUGHT OF A NEAT DElNITION WHICH AFTER
ONE HAD SEEN IT MADE THE RESULT OBVIOUS .EVERTHELESS IF WE WISH TO TEACH
COMPUTERS TO lND PROOFS IT IS LIKELY TO BE A GOOD IDEA TO REmECT ON HOW WE
DO SO OURSELVES
/NE WAY OF EXPLAINING A PROOF IN THE ABOVE SENSE IS TO SHOW THAT IT
CAN BE GENERATED USING WHAT ONE MIGHT THINK OF AS STANDARD MATHEMATICAL
REmEXES 3OME OF THESE MIGHT BE VERY GENERAL METHODS OF APPROACH AND
OTHERS MIGHT BE TRYING TO APPLY WELL KNOWN TECHNIQUES FROM THE AREA OF
MATHEMATICS IN QUESTION (ERE ARE A FEW EXAMPLES OF WHAT ) MEAN BY GEN
ERAL METHODS OF APPROACH &OR MY LECTURE ) DREW mOW CHARTS BUT MY 4E8
SKILLS ARE VERY BASIC SO ) SHALL REPRESENT THEM HERE AS PSEUDO ALGORITHMS !
MUCH MORE DETAILED AND SOPHISTICATED DISCUSSION OF MATHEMATICAL REmEXES
CAN BE FOUND IN 0OLYAS CLASSIC BOOKS SUCH AS ;0 VOLS ) AND ))=
! 'ENERALIZATION
)S P A SPECIAL CASE OF Q
)F IT SEEMS NOT TO BE THEN GOTO ELSE GOTO
#AN Q BE APPROPRIATELY MODIlED
)F IT CAN THEN GOTO ELSE GOTO
6ISIONS IN -ATH 2/5'( 3425#452% !.$ #,!33)&)#!4)/.
)S Q TRUE
)F IT MIGHT BE THEN GOTO ELSE GOTO
4RY TO PROVE Q INSTEAD OF P
4RY A DIkERENT APPROACH
" -ODIlCATION OF +NOWN !RGUMENT
&IND A KNOWN STATEMENT Q THAT RESEMBLES P
#AN THE PROOF OF Q BE EASILY MODIlED TO GIVE P
)F YES THEN %.$ ELSE GOTO
)DENTIFY A STATEMENT R FORMING PART OF THE PROOF OF Q SUCH THAT
THE ANALOGOUS STATEMENT S WHICH ONE WOULD LIKE AS PART OF THE
PROOF OF P IS NOT EASY TO PROVE
)F S IS OBVIOUSLY FALSE THEN GOTO ELSE GOTO
4RY TO PROVE S
# 3POTTING 0ATTERNS IN 3PECIAL #ASES
&IND A SPECIAL CASE R OF P
)F R IS TRIVIAL THEN GOTO ELSE GOTO
#AN ) IDENTIFY THE NEXT SIMPLEST SPECIAL CASE
)F SO CALL IT R AND GOTO ELSE GOTO
#AN ) SEE HOW TO PROVE R
)F SO THEN GOTO ELSE GOTO
$OES THE PROOF OF R GENERALIZE TO A PROOF OF P
)F SO THEN %.$ ELSE GOTO
#AN ) lND A DIkERENT MORE INSTRUCTIVE PROOF OF R
4RY TO PROVE R INSTEAD OF P
$ (ABITUATION OR 'ETTING "ORED
(AVE ) MANAGED TO THINK OF ANY INTERESTING IDEAS CONNECTED WITH
P IN THE LAST COUPLE OF HOURS
4HINK ABOUT ONE OF THE INTERESTING IDEAS
-AKE A NICE CUP OF COkEE AND THINK ABOUT SOMETHING ELSE FOR A
WHILE
!S FOR REmEXES OF A MORE SPECIlC KIND ) MEAN THINGS LIKE TRYING TO AP
PLY WELL KNOWN THEOREMS OR TECHNIQUES &OR THIS ONE NEEDS TO HAVE SOME
ABILITY TO CLASSIFY PROBLEMS SO THAT ONE HAS AN IDEA OF WHICH THEOREMS OR
74 '/7%23 '!&!
TECHNIQUES MAY BE HELPFUL 2ATHER THAN GIVING SEVERAL EXAMPLES OF THE

USE OF STANDARD METHODS TO SOLVE PROBLEMS LET ME RETURN TO THE QUESTION
OF AUTOMATING MATHEMATICS AND PRESENT AN IMAGINED DIALOGUE BETWEEN A
MATHEMATICIAN AND A COMPUTER IN TWO OR THREE DECADES TIME 4HE IDEA
OF THE DIALOGUE IS THAT THE COMPUTER IS VERY HELPFUL TO THE MATHEMATICIAN
WHILE NOT DOING ANYTHING PARTICULARLY CLEVER 4HIS REPRESENTS AN UNTHREAT
ENING INTERMEDIATE STAGE BETWEEN WHAT WE HAVE NOW COMPUTERS THAT ACT
AS SLAVES DOING UNBELIEVABLY BORING CALCULATIONS FOR US AND FULL AUTOMATION
OF MATHEMATICS ) HAVE WRITTEN THE DIALOGUE IN %NGLISH BUT THIS IS SUP
POSED TO BE A TRANSLATION OF A MORE FORMAL LANGUAGE WHICH HAS NOT YET BEEN
INVENTED ) SHALL DISCUSS THIS POINT A LITTLE MORE LATER
-ATHEMATICIAN )S THE FOLLOWING TRUE ,ET m 4HEN FOR . SUnCIENTLY
LARGE EVERY SET ! y F . G OF SIZE AT LEAST m. CONTAINS A SUBSET OF
THE FORM FA A D A DG
#OMPUTER 9ES )F ! IS NON EMPTY CHOOSE A ! AND SET D
- !LL RIGHT ALL RIGHT BUT WHAT IF D IS NOT ALLOWED TO BE ZERO
# (AVE YOU TRIED INDUCTION ON . WITH SOME m m. TENDING TO ZERO
- 4HAT IDEA IS NO HELP AT ALL 'IVE ME SOME EXAMPLES PLEASE
# 4HE OBVIOUS GREEDY ALGORITHM GIVES THE SET
F G
) NOTICE THAT LARGE PARTS OF THE SET ARE TRANSLATIONS OF OTHER PARTS )N
FACT THIS SET IS VERY LIKE THE #ANTOR SET SO THIS GIVES A BOUND OF m
. LOG LOG _
!S FOR RANDOM METHODS LET US CHOOSE EACH ELEMENT OF F . G TO
BE IN ! WITH PROBABILITY m THE CHOICES BEING INDEPENDENT 4HE EXPECTED
NUMBER OF SUBSETS OF ! OF THE FORM FA A D A DG IS m . TO WITHIN AN
ABSOLUTE CONSTANT (ENCE IF m . _ IT IS POSSIBLE FOR THERE TO BE NONE
4HE EXPECTED SIZE OF ! IS m. AND STANDARD ESTIMATES TELL US THAT ! IS VERY
LIKELY INDEED TO BE OF ABOUT THIS SIZE
!PPLYING ONE FURTHER STANDARD IDEA WE NOTE THAT IF THE NUMBER OF SETS
IN ! OF THE FORM FA A D A DG IS AT MOST HALF THE NUMBER OF POINTS
IN ! THEN WE CAN DELETE ONE ELEMENT FROM EACH OF THEM AND STILL BE LEFT
WITH A SUBSTANTIAL PROPORTION OF ! 4HIS TELLS US THAT WE CAN CHOOSE ANY
m THAT SATISlES #m . m. SO IN PARTICULAR WE CAN lND SOME m OF THE
FORM C. _
- 7ELL RANDOM METHODS OFTEN GIVE THE BEST ANSWER FOR PROBLEMS LIKE
THIS ,ETS TRY TO PROVE THAT C. _ IS BEST POSSIBLE
6ISIONS IN -ATH 2/5'( 3425#452% !.$ #,!33)&)#!4)/.
# ;0AUSES FOR SECONDS= !CTUALLY IT ISNT "EHREND FOUND A MUCH

BETTER BOUND IN ;$OWNLOADS PAPER=
- /H DEAR )M OUT OF IDEAS THEN #OULD YOU GIVE ME A SUGGESTION BY
ANY CHANCE
# 7E HAVE A SET ! 7E WANT TO PROVE THAT A SUBSET OF A CERTAIN FORM
EXISTS 4HE BEST WAY OF PROVING EXISTENCE IS OFTEN TO COUNT
- ;)NTRIGUED= 9ES BUT WHAT WOULD THAT MEAN FOR A PROBLEM LIKE THIS
# (ERE WE WISH TO COUNT THE NUMBER OF SOLUTIONS X Y Z OF THE SINGLE
LINEAR EQUATION Y X Z ! STANDARD TOOL IN SUCH SITUATIONS IS &OURIER
ANALYSIS EXPONENTIAL SUMS THE CIRCLE METHOD WHATEVER YOU LIKE TO CALL IT
- 7HERE IS THE GROUP STRUCTURE
# )DENTIFY F . G WITH :. : )T0ISNT PRETTY BUT IT SOMETIMES
WORKS ) WRITE :. FOR :. : AND !R > FOR S! EXPxIRS. 4HEN THE
0 > !_R
>
NUMBER OF TRIPLES A A D A D ! IS . _ R:. !R
- 7E ARE TRYING TO SHOW THAT THAT EXPRESSION CANNOT BE ZERO UNLESS m IS
VERY SMALL
# ;4RAINED TO HUMOUR MATHEMATICIANS= 4HAT IS INDEED ALMOST EQUIVALENT
TO OUR ORIGINAL PROBLEM 4HANK YOU .OTICE THAT . _ ! > m.
EXACTLY THE NUMBER WE HAD BEFORE WHEN ! WAS CHOSEN RANDOMLY 4HIS
NUMBER IS LARGE AND POSITIVE [ A GOOD SIGN
- 3O WE WOULD LIKE THE REST OF THE SUM TO BE SMALL ENOUGH NOT TO CANCEL
OUT THE ZERO TERM
# 9ES ) AM TRYING A FEW OBVIOUS IDEAS SUCH AS THE #AUCHY 3CHWARZ
INEQUALITY 5NFORTUNATELY THERE DOES NOT SEEM TO BE ANY REASON FOR THE
REST OF THE SUM TO BE SMALL
- 0LEASE SHOW ME YOUR CALCULATIONS
# (ERE THEY ARE ;$ISPLAYS THEM= 4HEY DO ENABLE ME OR RATHER US TO
>
PROVE A PARTIAL RESULT )T STATES THAT IF MAXR J!R J Cm . THEN THE
IMAGE OF ! IN :. CONTAINS MANY SETS OF THE DESIRED FORM
4HIS DIALOGUE COULD BE CONTINUED TO SHOW WHAT TO DO WHEN !R > IS LARGE
FOR SOME R (OWEVER THE COMPUTER HAS ALREADY GUIDED THE MATHEMATICIAN
TO AN IMPORTANT INSIGHT WHICH IS THE BASIS FOR 2OTHS PROOF OF HIS THEOREM ON
ARITHMETIC PROGRESSIONS ;2O= &INISHING THE PROOF IS NOT ALL THAT DInCULT
THOUGH PERHAPS IT MIGHT BE FOR THE MATHEMATICIAN ABOVE WHO SEEMS A BIT
TOO RELIANT ON HIS COMPUTER
.OTICE THAT THE COMPUTER IS AT EVERY STAGE TRYING STANDARD IDEAS IN
DUCTION A GREEDY ALGORITHM RANDOM METHODS COUNTING TAKING THE &OURIER
TRANSFORM AND APPLYING #AUCHY 3CHWARZ TO A SUM OF PRODUCTS 7HAT MAKES
74 '/7%23 '!&!
IT THINK OF THESE STANDARD IDEAS RATHER THAN SOME OTHER COMPLETELY INAP
PROPRIATE ONES 0ART OF THE ANSWER LIES IN HOW THE PROBLEM IS INITIALLY PUT TO
THE COMPUTER ) WOULD ENVISAGE NOT THE FORMAL STATEMENT GIVEN AT THE BE
GINNING OF THE DIALOGUE BUT SOMETHING MORE INTERACTIVE &OR EXAMPLE THIS
STAGE OF THE PROCESS COULD BE MENU DRIVEN )NITIALLY ONE WOULD CHOOSE #OM
BINATORICS OR PERHAPS #OMBINATORIAL .UMBER 4HEORY 4HIS WOULD BRING UP
ANOTHER MENU FROM WHICH ONE WOULD CHOOSE %XTREMAL 0ROBLEMS &OR THE
NEXT MENU ONE WOULD CHOOSE THE WORD -AXIMIZE 4HE NEXT CHOICE WOULD
BE /VER 3ETS ! OF )NTEGERS 4HEN ONE WOULD CHOOSE #ARDINALITY OF ! 4HEN
A MENU WOULD COME UP WITH A CHOICE OF CONSTRAINTS OR PERHAPS ONE WOULD
HAVE TO TYPE THEM IN ! y F . G AND ! CONTAINS NO ARITHMETIC
PROGRESSION OF LENGTH THREE
!T THE END OF A PROCESS LIKE THIS THE COMPUTER WOULD HAVE MANY IDEAS
ABOUT HOW THE PROBLEM WAS CONVENTIONALLY CLASSIlED &OR EXAMPLE IT WOULD
BE NATURAL FOR THE COMPUTER TO THINK OF THE CIRCLE METHOD BECAUSE MANY
PROBLEMS IN ADDITIVE NUMBER THEORY WOULD INVOLVE CONCEPTS SUCH AS THE
NUMBER OF ARITHMETIC PROGRESSIONS OF LENGTH THREE IN A SET ! 4HE COMPUTER
COULD THEN LOOK FOR A LIST OF STANDARD TRICKS AND lND THE SLOGAN IN THE ABOVE
DIALOGUE TO DO WITH COUNTING SOLUTIONS OF A LINEAR EQUATION
.OTICE THAT IT IS IMPORTANT FOR THE COMPUTER THAT IT HAS ACCESS TO A
MATHEMATICAL DATABASE WHICH IS MUCH MORE SOPHISTICATED THAN ANYTHING WE
HAVE AT PRESENT /NE MIGHT WONDER HOW THE COMPUTER COULD HAVE DISCOVERED
"EHRENDS PAPER SO QUICKLY SINCE UNDERSTANDING ORDINARY TEXT IS NOTORIOUSLY
DInCULT FOR COMPUTERS ) WAS IMAGINING THAT A GREAT DEAL OF WORK HAD GONE
INTO CLASSIFYING PAPERS SO THAT "EHRENDS PAPER APPEARED ON THE DATABASE
NOT AS 0ROC .AT !CAD 3CI [ BUT RATHER AS SOMETHING
LIKE -AXIMIZE [ 3IZE OF ! [ 3UBJECT 4O [ ! y F . G !
CONTAINS NO !0 [ ,OWER "OUND !S WELL AS THE PAPER THERE P WOULD BE A
SUMMARY OF THE MAIN RESULT [ ,OWER "OUND !T ,EAST N EXP_C LOG N
) DO NOT SEE ANY OBSTACLE TO CREATING A USEFUL DATABASE OF THE ABOVE KIND
OTHER THAN THE TIME AND EkORT THAT WOULD BE INVOLVED 4HE STRUCTURE OF THE
DATABASE WOULD BE VERY WELL SUITED TO THE INTERNET WHICH WOULD ALSO MAKE
IT EASIER TO DISTRIBUTE THE WORK AMONGST MANY PEOPLE SO PERHAPS EVEN THIS
OBSTACLE IS NOT ALL THAT GREAT 4HERE ARE OF COURSE PARTS OF MATHEMATICS THAT
DO NOT LEND THEMSELVES PARTICULARLY EASILY TO CLASSIlCATION OF THE ABOVE KIND
BUT THIS IS NO REASON TO DISMISS THE WHOLE IDEA %VEN A DATABASE COVERING
ONLY CERTAIN AREAS OF MATHEMATICS WOULD BE USEFUL
%NCODING AND CLASSIFYING GENERAL METHODS TRICKS RULES OF THUMB AND
6ISIONS IN -ATH 2/5'( 3425#452% !.$ #,!33)&)#!4)/.
SO ON WOULD OBVIOUSLY BE HARDER /NE OF THE MAIN REASONS FOR THIS IS THAT
THEY ARE OFTEN NOT PRECISELY FORMULATED AND DEALING WITH VAGUE STATEMENTS
IS A WELL KNOWN HARD PROBLEM IN COMPUTER SCIENCE (OWEVER HARD DOES
NOT MEAN IMPOSSIBLE AND THERE HAS ALREADY BEEN IMPORTANT PROGRESS ON IT
-OREOVER AT LEAST SOME OF THE VAGUE PRINCIPLES OF MATHEMATICS DO NOT
APPEAR TO PRESENT SERIOUS DInCULTIES #ONSIDER FOR EXAMPLE THE FOLLOWING
SIMPLE BUT EXTREMELY USEFUL CONSEQUENCE OF THE #AUCHY 3CHWARZ INEQUALITY
q8 N r 8N
JAI J N JAI J
I I
4HIS IS A USEFUL INEQUALITY BECAUSE THE @ NORM IS OFTEN EASIER TO HANDLE
THAN THE @ NORM (OWEVER IN MANY CONTEXTS THE INEQUALITY IS FAR TOO
WEAK SO ONE TENDS TO USE IT ONLY WHEN THERE IS SOME REASON TO BELIEVE THAT
THE AI ARE OF ROUGHLY EQUAL SIZE 4HIS SOUNDS LIKE A VAGUE IDEA AND MANY
MATHEMATICIANS WHEN THINKING ABOUT A PROBLEM WILL NOT NEED TO CLARIFY IT
FURTHER BUT FURTHER CLARIlCATION IS PERFECTLY POSSIBLE &OR EXAMPLE ANOTHER
VERY SIMPLE BUT SURPRISINGLY USEFUL INEQUALITY IS
N
8 N
8
JAI J MAX JAI J JAI J
IN
I I
WHICH TELLS US THAT THE lRST

0 INEQUALITY WILL BE SHARP TO WITHIN A CONSTANT #
IF MAXIN JAI J #N_ NI JAI J 4HIS IS NOT A NECESSARY CONDITION BUT IT
COULD FORM ONE OF A NUMBER OF TESTS THAT A COMPUTER COULD APPLY OR RATHER
ATTEMPT TO APPLY WHENEVER IT USED THE lRST INEQUALITY TO SEE WHETHER IT
WAS LIKELY TO BE EXPENSIVE
)F THERE WERE TO BE A SERIOUS ATTEMPT TO CODIFY THE MANY PRINCIPLES AND
WORKING METHODS OF MATHEMATICIANS THEN IT WOULD BE AN IMMENSELY USEFUL
EXERCISE TO RETURN TO THE MATHEMATICS WE ALREADY KNOW AND TRY TO REWRITE
THE PROOFS AS EXPLANATIONS IN THE RESTRICTED SENSE THAT ) DElNED EARLIER
-ANY TEXTBOOKS BEGIN WITH DElNITIONS THAT JUSTIFY THEMSELVES LATER BY LEAD
ING TO INTERESTING RESULTS THIS WOULD BE FORBIDDEN IN THE KIND OF TEXTBOOK )
HAVE IN MIND )NSTEAD EVERY DElNITION SHOULD BE PRESENTED AS ARISING NATU
RALLY IN THE SEARCH FOR THE SOLUTION OF SOME PROBLEM /F COURSE THE PROBLEM
ITSELF SHOULD ALSO BE JUSTIlED OR ALTERNATIVELY ONE COULD TAKE AS ONES START
ING POINT THAT THE PROBLEM IS WORTH STUDYING LEAVING THE JUSTIlCATION TO
SOMEBODY ELSE 4HE QUESTION OF JUSTIFYING OPEN PROBLEMS WAS DISCUSSED
BY 'ROMOV AT THE CONFERENCE 3EE HIS CONTRIBUTION TO THIS VOLUME ESPE
CIALLY THE ENDS OF X AND X &OR THE SAKE OF BREVITY ) HAVE IGNORED MANY
ASPECTS OF MATHEMATICAL PRACTICE SUCH AS MAKING CONJECTURES AND PROPOS
74 '/7%23 '!&!
ING DElNITIONS 4HESE MUST ALSO BE TAKEN SERIOUSLY IF MATHEMATICS IS TO BE

AUTOMATED
%VEN IF THE AUTOMATION OF MATHEMATICS TURNS OUT TO BE MUCH MORE DIF
lCULT THAN ) HAVE SUPPOSED THIS EXERCISE OF REWRITING COULD HAVE ENORMOUS
BENElTS BECAUSE UNSURPRISINGLY THE QUESTION OF HOW WE MIGHT TEACH MATH
EMATICS TO COMPUTERS IS CLOSELY RELATED TO THE QUESTION OF HOW WE SHOULD
TEACH IT TO HUMAN BEINGS ) HAVE NO TIME FOR THOSE WHO BELIEVE THAT TALENT
FOR MATHEMATICAL RESEARCH IS A MYSTERIOUS UNTEACHABLE QUALITY WHICH EITHER
YOU HAVE OR YOU DONT /F COURSE THERE IS SUCH A THING AS NATURAL ABILITY
BUT MOST MATHEMATICIANS DEVELOP THEIR ABILITY AS THEY MATURE !T THE MO
MENT THEY TEND TO DO THIS WITH VERY LITTLE EXTERNAL GUIDANCE AND VERY LITTLE
HELP FROM THE LITERATURE ! SERIES OF BOOKS AIMED AT MAKING TRANSPARENT
THE THOUGHT PROCESSES THAT LED TO CERTAIN RESULTS COULD SHOW TO YOUNG MATH
EMATICIANS WHY RESEARCH IS NOT AS IMPOSSIBLE AS IT LOOKS AND CONVEY MUCH
MORE EnCIENTLY THE TRICKS USED BY THOSE WITH MORE EXPERIENCE
4WO EXAMPLES OF WHAT ) MEAN ABOUT THE JUSTIlCATION OF DElNITIONS ARE
PROVIDED BY SIMPLICIAL HOMOLOGY AND THE 2IEMANN ZETA FUNCTION 7HEN )
WAS TAUGHT SIMPLICIAL HOMOLOGY ) WAS PRESENTED WITH A COMPLICATED AS IT
SEEMED AT THE TIME DElNITION WHICH THEN MIRACULOUSLY LED TO INTERESTING
AND FAR MORE CONCRETE RESULTS SUCH AS lXED POINT THEOREMS 3IMILARLY A TYP
ICAL INTRODUCTION TO THE ZETA FUNCTION BEGINS WITH %ULERS PRODUCT FORMULA
WHICH LEADS RAPIDLY TO INTERESTING THEOREMS ABOUT THE PRIMES BUT IS NOT PRE
SENTED AS THE INEVITABLE RESULT OF THINKING ABOUT THE PRIMES &OR THESE TWO
CONCEPTS ) CAN GIVE PLAUSIBLE BUT COMPLETELY AHISTORICAL AND BY NO MEANS
UNIQUE ACCOUNTS OF HOW THEY MIGHT HAVE BEEN INVENTED ) INTEND TO PUT
THEM ON MY PERSONAL WEB PAGE HTTPWWWDPMMSCAMACUK^WTG
BUT DO NOT EXPECT TO GET ROUND TO THIS BEFORE ABOUT *ANUARY ) ALSO
HOPE TO GIVE SOME EXAMPLES OF HOW APPARENT BRIGHT IDEAS CAN BE EXPLAINED
AS THE NATURAL RESULT OF FAMILIAR METHODS OF SEARCHING FOR PROOFS
) AM FULLY AWARE THAT WHAT ) HAVE DESCRIBED SO FAR IS A LONG WAY FROM THE
FULL AUTOMATION OF MATHEMATICS )N PARTICULAR ) HAVE IGNORED SOME NOTO
RIOUS PROBLEMS IN ARTIlCIAL INTELLIGENCE ) HAVE BRIEmY DISCUSSED VAGUENESS
BUT THE EXAMPLE ) GAVE WAS A SIMPLE ONE AND DOES NOT OBVIOUSLY GENER
ALIZE ) HAVE NOT MENTIONED THE DInCULTY OF PATTERN RECOGNITION ALTHOUGH
RECOGNISING SIMILARITY IS A VERY IMPORTANT PART OF MATHEMATICAL RESEARCH
7HAT ) HAVE TRIED TO SHOW IS THAT THESE DInCULTIES ARE NOT A GOOD REASON FOR
DOING NOTHING 2EmECTING ON ONES THOUGHT PROCESSES DOES NOT REQUIRE ANY
EXPERTISE IN COMPUTING AND IS POTENTIALLY VERY BENElCIAL TO OTHERS AND A
6ISIONS IN -ATH 2/5'( 3425#452% !.$ #,!33)&)#!4)/.
COMPUTER WITH A GREAT DEAL OF INFORMATION BUT ONLY VERY MODEST INTELLIGENCE
COULD STILL BE EXTREMELY USEFUL
) HAVE ALSO NOT DISCUSSED THE MANY EkORTS THAT HAVE ALREADY BEEN MADE
IN THE GENERAL AREA OF AUTOMATIC THEOREM PROVING AN AREA WHICH AS IS
PROBABLY OBVIOUS ) KNOW LITTLE ABOUT -Y IMPRESSION IS THAT MOST OF THE
PROGRESS SO FAR HAS BEEN FROM THE BOTTOM UP [ TRYING TO GET COMPUTERS
TO DISCOVER SIMPLE BUT INTERESTING THEOREMS STARTING ALMOST FROM SCRATCH
7HAT ) HAVE TALKED ABOUT IS MORE OF A TOP DOWN PROJECT TO WHICH ANY
MATHEMATICIAN COULD CONTRIBUTE )T MIGHT BE THAT EVENTUALLY SUCH A TOP
DOWN APPROACH COULD REDUCE THE GENERAL PROBLEM OF AUTOMATING MATHE
MATICS TO A LARGE NUMBER OF SMALLER MORE MANAGEABLE PROBLEMS SO THAT
PERHAPS THE MATHEMATICAL COMMUNITY COULD MEET THE AUTOMATIC THEOREM
PROVERS HALF WAY /NE WAY TO LEARN ABOUT THE STATE OF THE ART IN AUTOMATIC
THEOREM PROVING IS TO VISIT THE WEBSITE HTTPWWWMATHTEMPLEEDU^ZEIL
BERG/0).)/.3HTML AND FOLLOW LINKS ESPECIALLY FROM THE RATHER EXTREME
OPINION
&INALLY THEN HERE IS MY GUESS ABOUT THE IMPACT OF COMPUTERS ON MATH
EMATICS OVER THE NEXT CENTURY !LTHOUGH SOME OF THE DInCULTIES INVOLVED
IN AUTOMATIC THEOREM PROVING ARE FORMIDABLE A CENTURY IS A VERY LONG TIME
IN COMPUTING SO ) EXPECT COMPUTERS TO BE BETTER THAN HUMANS AT PROVING
THEOREMS IN ) DO NOT FEEL PARTICULARLY HAPPY ABOUT THIS BUT ) EXPECT
THE IMPACT ON MY OWN MATHEMATICAL LIFE TO BE ENTIRELY POSITIVE BECAUSE A
SEMI INTELLIGENT DATABASE OF THE KIND ) HAVE DESCRIBED WOULD BE A WONDERFUL
RESOURCE AND WOULD TAKE A GREAT DEAL OF THE DRUDGERY OUT OF RESEARCH 4HERE
MIGHT EVEN BE A GOLDEN AGE WHEN COMPUTERS WERE GOOD AT EXERCISES BUT NOT
YET GOOD AT HAVING DEEP INSIGHTS 4HEN INSTEAD OF WASTING A WEEK NOT
NOTICING THAT A HOPED FOR LEMMA HAD A SIMPLE COUNTEREXAMPLE ONE COULD
GET THE COMPUTER TO CHECK IT )N OTHER WORDS COMPUTERS WOULD STILL DO THE
BORING BITS FOR US BUT THESE WOULD NOT BE QUITE AS BORING AS THEY ARE NOW
(OWEVER SUCH A GOLDEN AGE IF IT OCCURS IS UNLIKELY TO LAST FOR LONG 4HE
NEXT STAGE MIGHT BE ONE WHERE ONLY A VERY FEW OUTSTANDING MATHEMATICIANS
COULD DISCOVER PROOFS THAT WERE INACCESSIBLE TO COMPUTERS 0ERHAPS OTHERS
WOULD CONCENTRATE ON TEACHING BUT COMPUTERS WOULD PROBABLY BECOME BET
TER THAN US AT THIS AS WELL AT LEAST IF THEY THEMSELVES HAD BEEN TAUGHT IN
ANYTHING LIKE THE WAY ) HAVE OUTLINED )N THE END THE WORK OF THE MATH
EMATICIAN WOULD BE SIMPLY TO LEARN HOW TO USE THEOREM PROVING MACHINES
EkECTIVELY AND TO lND INTERESTING APPLICATIONS FOR THEM 4HIS WOULD BE
A VALUABLE SKILL BUT IT WOULD HARDLY BE PURE MATHEMATICS AS WE KNOW IT
TODAY
74 '/7%23 '!&!
2OUGH 3TRUCTURE AND #LASSIlCATION AND 2ELATED

0ROBLEMS
)N HIS TALK 'ROMOV DREW ATTENTION TO THE FACT THAT NOT MANY WAYS WERE
KNOWN OF DElNING MANIFOLDS 3EE THE SUBSECTION <2ANDOMIZATION IN X OF
HIS CONTRIBUTION TO THIS VOLUME &OR A DIkERENT REASON SOMETHING SIMILAR
IS TRUE FOR GRAPHS 4HIS MIGHT AT lRST SEEM A RIDICULOUS ASSERTION [ ALL YOU
HAVE TO DO TO DElNE A GRAPH IS TAKE THE SET OF ALL PAIRS OF ELEMENTS OF THE SET
F NG AND PASS TO AN ARBITRARY SUBSET (OWEVER WHILE IT IS TRUE THAT
A MATHEMATICAL OBJECT DOES NOT HAVE TO SATISFY DELICATE CONSTRAINTS IN ORDER
TO BE A GRAPH ONE STILL FACES THE PURELY LOGICAL PROBLEM THAT THE NUMBER
OF COMPREHENSIBLE DESCRIPTIONS IS LIMITED -OREOVER WHILE GRAPH THEORISTS
OFTEN TALK AS THOUGH THEY ARE CONSIDERING INDIVIDUAL GRAPHS MOST PROBLEMS
IN GRAPH THEORY ARE STRICTLY SPEAKING ABOUT FAMILIES 'N N OF GRAPHS
WHERE THE NUMBER OF VERTICES OF 'N IS USUALLY N OR PERHAPS SOME SIMPLE
FUNCTION OF N THAT TENDS TO INlNITY
(ERE IS A LIST OF A FEW COMMON WAYS OF DElNING FAMILIES OF GRAPHS
I 2EPEATING PATTERNS 3OME EXAMPLES OF THIS ARE
A +N THE COMPLETE GRAPH OF ORDER N
B 0N THE PATH WITH N VERTICES
C #N THE CYCLE WITH N VERTICES
D +N THE <STAR WITH ONE VERTEX JOINED TO N OTHERS AND NO
FURTHER EDGES
E F GN THE DISCRETE N DIMENSIONAL CUBE
F A DYADIC TREE WITH A ROOT AND N FURTHER LAYERS
II 3IMPLE CONSTRUCTIONS FROM KNOWN GRAPHS 4HESE INCLUDE PRODUCTS
QUOTIENTS DISJOINT UNIONS REMOVAL OF A FEW EDGES AND SO ON
III 5SE OF ALGEBRA AND NUMBER THEORY 4WO NOTABLE EXAMPLES ARE THE
0ALEY GRAPH AND THE 2AMANUJAN GRAPHS OF ,UBOTZKY 0HILLIPS AND
3ARNAK ;,03= )N GENERAL #AYLEY GRAPHS OFTEN HAVE INTERESTING PROP
ERTIES
IV 2ANDOM GRAPHS )T IS NOT QUITE CLEAR IN WHAT SENSE THIS PHRASE DElNES
A FAMILY OF GRAPHS SEE THE DISCUSSION BELOW
4HIS LIST COULD UNDOUBTEDLY BE CONTINUED BUT ) WOULD BE SURPRISED IF SOME
BODY COULD GIVE ME A FURTHER TEN GENUINELY DISTINCT AND INTERESTING CLASSES
OF DElNITIONS
,ET US CONSIDER FOR A MOMENT WHAT IT MEANS TO SAY THAT A GRAPH IS
<RANDOM 7E WOULD LIKE TO MAKE STATEMENTS SUCH AS <! RANDOM GRAPH ON
6ISIONS IN -ATH 2/5'( 3425#452% !.$ #,!33)&)#!4)/.
_ `
N VERTICES WITH EDGE PROBABILITY HAS APPROXIMATELY N EDGES 3OME
TIMES ONE CAN MAKE SENSE OF SUCH A STATEMENT BY INSERTING THE WORDS <WITH
HIGH PROBABILITY IN THE APPROPRIATE PLACE (OWEVER THERE ARE CONTEXTS IN
WHICH THIS IS NOT ADEQUATE &OR EXAMPLE THERE IS A CLASS OF GRAPHS KNOWN
AS EXPANDERS WHICH HAVE DIRECT PRACTICAL APPLICATIONS IN THE DESIGN OF CER
TAIN ALGORITHMS )T IS KNOWN THAT GRAPHS CHOSEN RANDOMLY ARE WITH HIGH
PROBABILITY GOOD EXPANDERS /N THE OTHER HAND FACED WITH A PARTICULAR
RANDOMLY CHOSEN GRAPH HOW CAN ONE BE SURE THAT IT IS NOT ONE OF THE EX
CEPTIONS 4HIS IS PERHAPS NOT A GENUINE PRACTICAL PROBLEM SINCE A GRAPH
THAT WORKS WITH PROBABILITY __ IS JUST AS USEFUL AS ONE THAT WORKS WITH
COMPLETE CERTAINTY THOUGH ) HAVE HEARD THAT PEOPLE IN INDUSTRY WHO BUY
COMPUTER SYSTEMS ARE VERY SUSPICIOUS OF RANDOM ELEMENTS IN THE DESIGN
%VEN SO A VERY IMPORTANT DEVELOPMENT IN GRAPH THEORY WAS THE REALIZATION
THAT SEVERAL PROPERTIES SHARED BY ALMOST ALL GRAPHS ARE EQUIVALENT AND SERVE
AS A USEFUL DElNITION OF <RANDOMNESS
4HESE PROPERTIES APPLY TO A GRAPH ' WITH N VERTICES SUCH THAT ALMOST
EVERY VERTEX HAS DEGREE APPROXIMATELY N 4HEY SHOW THAT ' RESEMBLES
A GRAPH WHERE EACH EDGE IS CHOSEN INDEPENDENTLY WITH PROBABILITY
3IMILAR RESULTS HOLD FOR OTHER PROBABILITIES AS WELL (ERE ARE THREE OF
THEM
I &OR ANY TWO LARGE SETS ! AND " OF VERTICES THE NUMBER E' ! " OF
PAIRS X Y ! a " SUCH THAT XY IS AN EDGE OF ' IS J!JJ"J ON
AND THUS ROUGHLY THE NUMBER ONE WOULD EXPECT IF ' WAS A RANDOM
GRAPH
II &OR ANY lXED SMALL GRAPH ( THE NUMBER OF INDUCED SUBGRAPHS OF
' ISOMORPHIC TO ( IS ROUGHLY WHAT THE EXPECTED NUMBER WOULD HAVE
BEEN IF ' HAD BEEN CHOSEN RANDOMLY
III $ElNE THE ADJACENCY MATRIX ! OF ' BY SETTING !X Y TO BE IF XY
IS AN EDGE AND OTHERWISE 4HEN THE SECOND LARGEST EIGENVALUE IN
MODULUS OF ! IS ON
)N CONNECTION WITH PROPERTY I ABOVE DElNE THE DENSITY m' ! " TO BE
E' ! " J!JJ"J !NOTHER WAY OF STATING THE PROPERTY IS TO SAY THAT m' ! "
IS APPROXIMATELY FOR ALL LARGE SETS ! AND " )NTERESTINGLY ONE CAN RE
PLACE II BY THE WEAK SEEMING ASSUMPTION THAT ' CONTAINS ROUGHLY THE
EXPECTED NUMBER OF CYCLES OF LENGTH FOUR
/NE OF THE lRST PAPERS IN WHICH IT WAS SHOWN THAT THE SMALLNESS OF THE
SECOND LARGEST EIGENVALUE IMPLIED RANDOMNESS PROPERTIES WAS BY !LON AND
-ILMAN ;!-= 4HE lRST PAPER DEALING EXPLICITLY WITH THE IDEA OF PSEUDO
74 '/7%23 '!&!
RANDOM GRAPHS WAS BY 4HOMASON ;4= 4HE PAPER OF #HUNG 'RAHAM AND
7ILSON ;#'7= MADE THE WHOLE AREA MUCH MORE SYSTEMATIC GIVING A LONG
LIST OF EQUIVALENT RANDOMNESS PROPERTIES 4HEY CALLED A GRAPH SATISFYING
ANY ONE AND HENCE ALL OF THE PROPERTIES QUASIRANDOM PERHAPS TO MAKE
CLEAR THE DISTINCTION BETWEEN THIS IDEA AND THE RATHER DIkERENT CONCEPT OF
PSEUDORANDOMNESS IN COMPUTER SCIENCE
4HERE ARE TWO MORALS ) WOULD LIKE TO DRAW FROM THE ABOVE RESULTS 4HE
lRST IS THAT IN MANY CONTEXTS THEY ALLOW ONE TO GIVE A CLEAR MEANING TO THE
CONCEPT OF A RANDOM GRAPH EVEN WHEN THE GRAPH HAS BEEN EXPLICITLY CHOSEN
&OR EXAMPLE THE 0ALEY GRAPH IS KNOWN TO BE QUASIRANDOM IN THIS SENSE
4HE SECOND IS THAT THEY SUPPORT OUR INTUITION THAT TWO RANDOM GRAPHS ARE
BASICALLY THE SAME /F COURSE THEY WILL NOT BE ISOMORPHIC BUT THEY SHARE
ALL THE PROPERTIES THAT INTEREST US SO THE DIkERENCE BETWEEN THEM IS NOT
IMPORTANT
) CAN NOW EXPLAIN WHAT ) MEAN BY <ROUGH CLASSIlCATION )T IS CLEAR THAT
THERE ARE TOO MANY GRAPHS FOR IT TO BE SENSIBLE TO TRY TO CLASSIFY THEM ALL
UP TO ISOMORPHISM (OWEVER AS JUST COMMENTED WE OFTEN LIKE TO REGARD
NON ISOMORPHIC GRAPHS AS BASICALLY THE SAME )MPLICIT IN THAT DISCUSSION
WAS A METRIC WHICH CAN BE DElNED AS FOLLOWS 'IVEN TWO GRAPHS ' AND (
BOTH WITH N VERTICES AND A BIJECTION } BETWEEN 6 ' AND 6 ( DElNE
k k
D} ' ( N_ MAX kE' ! " _ E( }! }" k
!"y6 '
AND LET D' ( BE THE MINIMUM OF ALL THE D} ' ( )F ' AND ( ARE CLOSE
IN THIS METRIC IT MEANS THAT THERE IS A ONE TO ONE CORRESPONDENCE BETWEEN
THEIR VERTICES SUCH THAT THE DENSITIES m' ! " IN ' ARE APPROXIMATELY THE
SAME AS THE CORRESPONDING DENSITIES IN ( AT LEAST WHEN ! AND " ARE LARGE
! GRAPH IS QUASIRANDOM IF AND ONLY IF IT IS CLOSE TO A TYPICAL RANDOM GRAPH
IN THIS METRIC
"Y A ROUGH CLASSIlCATION OF GRAPHS ) MEAN A CHOICE OF A SMALL NUMBER
OF REPRESENTATIVE GRAPHS SUCH THAT EVERY GRAPH IS CLOSE IN THE ABOVE METRIC
TO ONE OF THE REPRESENTATIVES 3UCH A CLASSIlCATION WILL BE USEFUL IF THE
REPRESENTATIVE GRAPHS CAN BE EASILY DESCRIBED AND ANALYSED )T IS NOT AT ALL
OBVIOUS THAT A USEFUL CLASSIlCATION SHOULD EXIST BUT AN IMPORTANT RESULT
3ZEMERsEDIS REGULARITY LEMMA ;3= SHOWS THAT IT DOES
4O STATE THE RESULT LET US INTRODUCE A FURTHER DElNITION )F ' IS A GRAPH
AND ! AND " ARE SETS OF VERTICES IN ' LET US SAY THAT THE PAIR ! " IS
n REGULAR IF WHENEVER ! AND " ARE SUBSETS OF ! AND " WITH J! J nJ!J AND
J" J nJ"J WE HAVE JD' ! " _ D' ! " J n )F ! AND " ARE DISJOINT
6ISIONS IN -ATH 2/5'( 3425#452% !.$ #,!33)&)#!4)/.
THIS CONDITION WHICH IS VERY SIMILAR TO THE lRST PROPERTY OF QUASIRANDOM

GRAPHS LISTED ABOVE STATES THAT THE EDGES BETWEEN ! AND " BEHAVE VERY
LIKE THOSE OF A RANDOM BIPARTITE GRAPH ON ! AND " WITH EDGE PROBABILITY
D' ! " 3ZEMERsEDIS REGULARITY LEMMA IS AS FOLLOWS
,EMMA ,ET n 4HEN THERE EXISTS A POSITIVE INTEGER K Kn SUCH THAT
GIVEN ANY lNITE GRAPH ' THE VERTICES OF ' CAN BE PARTITIONED INTO_ K`CLASSES
6 6K ALL OF SIZE BNKC OR DNKE SUCH THAT ALL BUT AT MOST n K OF THE
PAIRS 6I 6J WITH I J ARE n REGULAR
! COMMON WAY OF DElNING GRAPHS IS TO CHOOSE A FEW SETS 6 6K AND
A SYMMETRIC ; = VALUED MATRIX PIJ KIJ AND TO JOIN X 6I TO Y 6J
WITH PROBABILITY PIJ ALL CHOICES BEING INDEPENDENT 3ZEMERsEDIS REGULARITY
LEMMA STATES THAT EVERY GRAPH IS LIKE THAT AT LEAST UP TO CLOSENESS IN THE
METRIC DElNED EARLIER 3INCE RANDOM GRAPHS ARE QUITE WELL UNDERSTOOD THIS
GIVES US A USEFUL CLASSIlCATION
/NE CONSEQUENCE OF THIS RESULT IS THAT IF ' IS A GRAPH AND ( IS AN IN
DUCED SUBGRAPH OF ' FORMED BY CHOOSING HALF THE VERTICES OF ' AT RANDOM
THEN ( LOOKS VERY SIMILAR TO ' 4HE REASON IS THAT WITH HIGH PROBABILITY
THE VERTEX SET OF ( WILL BE MADE UP OF ROUGHLY HALF OF EACH OF THE SETS 6I
IN THE PARTITION GIVEN BY 3ZEMERsEDIS LEMMA AND THE REGULARITY OF A PAIR
6I 6J IS INHERITED BY LARGE SUBSETS 6I 6J AT THE COST OF A SLIGHT WORSEN
ING OF THE PARAMETER n 4HIS GENERAL PRINCIPLE ABOUT RANDOM SUBGRAPHS IS
HARD TO DEMONSTRATE WITHOUT THE USE OF SOME RESULT ALONG THE LINES OF THE
REGULARITY LEMMA THOUGH THERE ARE WEAKER VERSIONS THAT SUnCE
! MAJOR DRAWBACK WITH THE REGULARITY LEMMA IS THAT THE DEPENDENCE OF
K ON n IS VERY BAD [ A TOWER OF TWOS OF HEIGHT PROPORTIONAL TO n_ -OREOVER
IT HAS BEEN SHOWN THAT THIS UNFORTUNATE SITUATION IS NECESSARY )N ;'= A
GRAPH IS CONSTRUCTED WHICH GIVES A LOWER BOUND FOR K OF A TOWER OF TWOS OF
HEIGHT PROPORTIONAL TO n_ 4HIS LIMITS THE USE OF THE LEMMA TO SITUATIONS
WHERE THE DEPENDENCE ON n IS NOT IMPORTANT
"EFORE ) MOVE ON TO TALK ABOUT OPEN PROBLEMS LET ME BRIEmY DISCUSS THE
OTHER HALF OF MY TITLE ROUGH STRUCTURE -ATHEMATICIANS VARY CONSIDERABLY
IN THE AMOUNT OF STRUCTURE THEY LIKE AND THIS ALLOWS THEM TO BE ROUGHLY
CLASSIlED IN A USEFUL WAY !T ONE END OF THE SPECTRUM ARE THOSE WHO START
WITH SO MANY ASSUMPTIONS THAT EVEN lNDING A NEW OBJECT THAT SATISlES THE
ASSUMPTIONS IS A GREAT ACHIEVEMENT 4HOSE WHO MANAGE TO DO SO ARE OFTEN
REWARDED WITH THE DISCOVERY OF BEAUTIFUL AND DETAILED PATTERNS !T THE
OTHER END ARE MATHEMATICIANS WHO LIKE MORE ELBOW ROOM 4HEY OPERATE
UNDER FEWER CONSTRAINTS SO THE OBJECTS THEY STUDY EXIST IN GREAT PROFUSION
74 '/7%23 '!&!
AND THE PROBLEMS THAT ARISE ARE OF A VERY DIkERENT CHARACTER

!S MY DISCUSSION OF CLASSIlCATION ALREADY INDICATES THIS DISTINCTION
THOUGH IT UNDOUBTEDLY EXISTS IS NOT QUITE AS GREAT AS IT MIGHT SEEM )
WOULD LIKE TO BLUR IT A LITTLE FURTHER BY SHOWING THAT PATTERNS CAN EMERGE
UNEXPECTEDLY EVEN IN VERY COMBINATORIAL PROBLEMS
4HERE IS A THEOREM OF &REIMAN ;&R= THAT ILLUSTRATES THIS EXTREMELY WELL
,ET ! BE A SET OF INTEGERS AND WRITE ! ! FOR THE SET FX Y X Y !G
)T IS AN EASY EXERCISE TO SHOW THAT IF J!J N THEN N _ J! !J

NN AND THAT BOTH BOUNDS ARE TIGHT &REIMANS THEOREM GIVES A
COMPLETE DESCRIPTION OF ! WHEN ! ! IS SMALL OR TO BE PRECISE WHEN
J! !J #N FOR SOME lXED CONSTANT # !N OBVIOUS WAY TO CONSTRUCT
SUCH A SET IS TO LET 0 BE AN ARITHMETIC PROGRESSION OF SIZE AT MOST #N
AND TO LET ! BE AN ARBITRARY SUBSET OF 0 OF SIZE N (OWEVER THIS IS NOT
THE ONLY SOURCE OF EXAMPLES ONE CAN TAKE A <TWO DIMENSIONAL ARITHMETIC
PROGRESSION WHICH IS A SET OF THE FORM
FA R D R D R M R M G
)T CAN BE CHECKED THAT IF ! IS A TWO DIMENSIONAL PROGRESSION THEN J!!J
J!J -ORE GENERALLY ! ! WILL BE SMALL IF ! IS A LARGE SUBSET OF A D
DIMENSIONAL ARITHMETIC PROGRESSION FOR SOME SMALL D 4HE DElNITION SHOULD
BE OBVIOUS
! FURTHER EASY EXERCISE IS TO SHOW THAT IF J! !J N _ THEN ! MUST
BE AN ARITHMETIC PROGRESSION OF LENGTH N (ARDER THAN THIS BUT STILL NOT TOO
HARD IS TO SHOW THAT IF J! !J IS JUST A LITTLE LARGER THAN N AT MOST N
SAY THEN THERE IS AN ARITHMETIC PROGRESSION OF LENGTH SLIGHTLY GREATER THAN
N THAT CONTAINS ! 4HE STRONGEST RESULT OF THIS KIND IS DUE TO &REIMAN
4HEOREM ,ET ! BE A SET OF INTEGERS OF SIZE N )F B N _ AND
J! !J N _ B THEN ! IS CONTAINED IN AN ARITHMETIC PROGRESSION OF
LENGTH N B
4HIS RESULT SHOWS THAT ! IS FORCED TO BE ONE DIMENSIONAL IF J! !J
N _ 4HIS BOUND IS TIGHT BECAUSE IF M IS ANY INTEGER LARGER THAN N _
AND ! IS THE SET F N _ MG THEN J! !J N _
&REIMANS MAIN THEOREM IS MUCH DEEPER AND TELLS US WHAT HAPPENS
WHEN WE GRADUALLY RELAX THE CONDITION ON THE SIZE OF THE SUMSET )T STATES
THAT THE ONLY SETS WITH SMALL SUMSETS ARE LARGE SUBSETS OF LOW DIMENSIONAL
ARITHMETIC PROGRESSIONS
4HEOREM &OR EVERY # THERE EXIST CONSTANTS + +# AND D D#
SUCH THAT IF ! IS ANY SET OF N INTEGERS AND J! !J #N THEN ! IS A
6ISIONS IN -ATH 2/5'( 3425#452% !.$ #,!33)&)#!4)/.
SUBSET OF A GENERALIZED ARITHMETIC PROGRESSION OF DIMENSION AT MOST D AND

CARDINALITY AT MOST +N
.OTICE THAT THE ASSUMPTION MADE IN &REIMANS THEOREM IS COMBINATORIAL
BUT THE CONCLUSION IS VERY DElNITELY STRUCTURAL IN AN APPROPRIATE ROUGH
SENSE 4HIS IS NOT PARTICULARLY SURPRISING WHEN # IS CLOSE TO THE EXTREMAL
VALUE OF WHEN ! IS JUST AS ONE MIGHT EXPECT A SMALL PERTURBATION OF
SOME EASILY DESCRIBED EXTREMAL SET BUT FOR LARGER VALUES OF # THE RESULT
IS MUCH LESS TO BE EXPECTED AND IS NOT AT ALL EASY TO PROVE ! BEAUTIFUL
AND COMPLETELY DIkERENT PROOF OF &REIMANS THEOREM WAS OBTAINED BY 2UZSA
;2U= AND A CONSIDERABLY CLEANED UP VERSION OF &REIMANS ORIGINAL ARGUMENT
HAS RECENTLY BEEN GIVEN BY "ILU ;"I=
) HAVE PRESENTED 3ZEMERsEDIS REGULARITY LEMMA AND &REIMANS THEOREM IN
A VERY POSITIVE WAY AND THEY BOTH HAVE IMPORTANT APPLICATIONS (OWEVER
THEY DO NOT SOLVE EVERYTHING &OR THE REST OF THIS SECTION ) SHALL DISCUSS
PROBLEMS WHICH APPEAR TO BE HARD TO SOLVE WITHOUT SOME SORT OF ROUGH
CLASSIlCATION THEOREM )N EACH CASE THE CLASSIlCATION THEOREM WOULD HAVE
TO BE BETTER THAN ANYTHING WE HAVE AT PRESENT ) BEGIN WITH THREE PROBLEMS
IN ELEMENTARY 2AMSEY THEORY
"OUNDS FOR DIAGONAL 2AMSEY NUMBERS 2AMSEYS THEOREM AS
SERTS THAT FOR EVERY POSITIVE INTEGER K THERE EXISTS A POSITIVE INTEGER . SUCH
THAT IF THE EDGES OF THE COMPLETE GRAPH ON . VERTICES ARE COLOURED RED AND
BLUE THEN THERE MUST BE K OF THE VERTICES WITH ALL EDGES JOINING THEM OF
THE SAME COLOUR 4HE GRAPH SPANNED BY THESE VERTICES IS CALLED MONOCHRO
MATIC 4HE SMALLEST . WITH THIS PROPERTY IS DENOTED BY 2K K )T IS
KNOWN THAT 2 AND 2 BUT 2K K IS NOT KNOWN FOR
ANY K )T IS UNLIKELY THAT A REASONABLE FORMULA FOR 2K K EXISTS SO A
MUCH MORE INTERESTING QUESTION IS HOW IT BEHAVES ASYMPTOTICALLY 3IMPLE
ARGUMENTS SHOW THAT K 2K K K BUT DESPITE STRENUOUS EkORTS
ON THE PART OF MANY MATHEMATICIANS NO SIGNIlCANT IMPROVEMENT ON EITHER
BOUND HAS BEEN OBTAINED "Y SIGNIlCANT ) MEAN A LOWER BOUND OF jK WITH
j OR AN UPPER BOUND OF kK WITH k
4HERE IS AN EQUIVALENT DElNITION OF 2K K WHICH ) SHALL SOMETIMES USE
WITHOUT FURTHER COMMENT )T IS THE SMALLEST . SUCH THAT ANY GRAPH WITH
. VERTICES CONTAINS A CLIQUE OR INDEPENDENT SET WITH K VERTICES THAT IS K
VERTICES EITHER COMPLETELY JOINED OR NOT JOINED AT ALL
#OVERING BY TRIANGLE FREE GRAPHS 4HE 2AMSEY NUMBER 2
WHERE OCCURS R TIMES IS THE SMALLEST INTEGER . SUCH THAT IF THE COM
PLETE GRAPH ON . VERTICES IS COLOURED WITH R COLOURS THEN THERE MUST BE
74 '/7%23 '!&!
A TRIANGLE WITH ALL ITS EDGES OF THE SAME COLOUR 2ELATIVELY STRAIGHTFORWARD
ARGUMENTS SHOW THAT THIS NUMBER LIES BETWEEN ER AND R R )T IS AN OPEN
QUESTION WHETHER THERE IS AN UPPER BOUND OF THE FORM # R FOR SOME ABSOLUTE
CONSTANT #
!N EQUIVALENT FORMULATION OF THIS PROBLEM IS AS FOLLOWS IF THE COMPLETE
GRAPH ON N VERTICES IS TO BE WRITTEN AS A UNION OF K TRIANGLE FREE GRAPHS
HOW LARGE MUST K BE /NE CAN DO IT IF K DLOG NE )S IT NECESSARY TO USE
AT LEAST C LOG N GRAPHS
4HERE IS A FURTHER EQUIVALENT FORMULATION OF THE PROBLEM IN TERMS OF
3HANNON CAPACITY [ SEE THE CONTRIBUTION OF !LON IN THIS VOLUME FOR A DIS
CUSSION OF THIS CONCEPT
2AMSEY NUMBERS UNDER ADDITIONAL CONSTRAINTS 4HE FOLLOWING
IS A BEAUTIFUL PROBLEM OF %RD]
OS AND (AJNAL ,ET ( BE A lXED lNITE GRAPH )F
' IS ASSUMED NOT TO CONTAIN ANY INDUCED COPY OF ( THEN BY HOW MUCH DOES
THE BOUND FOR 2K K IMPROVE )N PARTICULAR DOES THERE EXIST A CONSTANT
# DEPENDING ON ( ONLY SUCH THAT EVERY GRAPH WITH AT LEAST K# VERTICES
CONTAINING NO INDUCED ( HAS A COMPLETE OR EMPTY SUBGRAPH WITH K VERTICES
7HAT IS IT ABOUT THESE THREE PROBLEMS THAT MAKES THEM HARD 4HE
MOST IMPORTANT REASON SEEMS TO BE THAT IT IS NOT OBVIOUS WHAT THE EXTREMAL
GRAPHS SHOULD BE &OR THE lRST PROBLEM THE LOWER BOUND 2K K K
WAS PROVED BY %RD] OS ;%= IN (IS PROOF CAN BE SUMMARIZED AS FOLLOWS
COLOUR EACH EDGE RED OR BLUE RANDOMLY AND INDEPENDENTLY WITH PROBABILITY
ONE HALF AND CHECK THAT THE EXPECTED NUMBER OF MONOCHROMATIC COMPLETE
SUBGRAPHS WITH K VERTICES IS LESS THAN ONE 4O PUT THAT EVEN MORE LOOSELY
A RANDOM COLOURING WORKS
/NCE ONE HAS SEEN THIS PROOF IT COMES TO SEEM SO NATURAL THAT IT IS
TEMPTING TO CONJECTURE THAT THE BEST POSSIBLE BOUND OR SOMETHING CLOSE
TO IT IS ATTAINED BY A RANDOM GRAPH AND INDEED MANY WHO HAVE THOUGHT
ABOUT THE PROBLEM BELIEVE THIS TO BE SO (OWEVER AN UNSETTLING RESULT OF
4HOMASON ;4= SUGGESTS THAT THE PROOF OF SUCH A FACT WILL NOT BE EASY !
NATURAL APPROACH TO ESTIMATING 2K K IS TO CONSIDER THE FOLLOWING MORE
GENERAL PROBLEM ,ET K AND N BE POSITIVE INTEGERS AND SUPPOSE THAT THE
EDGES OF A COMPLETE GRAPH ON N VERTICES ARE COLOURED RED AND BLUE 7HAT
IS THE SMALLEST POSSIBLE NUMBER OF MONOCHROMATIC COMPLETE SUBGRAPHS OF
ORDER K %RD] OS CONJECTURED THAT A RANDOM COLOURING SHOULD BE BEST POSSI
K _ `
BLE WHICH WOULD SUGGEST A BOUND OF APPROXIMATELY _ NK 4HIS RESULT
IS TRUE WHEN K BUT 4HOMASON SHOWED THAT IT IS FALSE WHEN K
)T IS IMPORTANT TO EMPHASIZE THE SENSE IN WHICH RANDOM GRAPHS ARE NOT
6ISIONS IN -ATH 2/5'( 3425#452% !.$ #,!33)&)#!4)/.
BEST 3OME GRAPHS SUCH AS THE REMARKABLE 2AMANUJAN GRAPHS IMPROVE ON

RANDOM CONSTRUCTIONS BY BEING <MORE RANDOM THAN RANDOM &OR EXAMPLE
2AMANUJAN
P GRAPHS ARE D REGULAR AND HAVE A SECOND LARGEST EIGENVALUE OF
D _ WHICH IS ASYMPTOTICALLY AS SMALL AS POSSIBLE !LTHOUGHP RANDOM
GRAPHS ALSO TEND TO HAVE A SECOND LARGEST EIGENVALUE OF SIZE # D _ THE
CONSTANT # IS LARGER THAN 4HIS IS NOT THE NATURE OF THE NON RANDOMNESS
OF 4HOMASONS EXAMPLE AND INDEED CANNOT BE BECAUSE ANY GRAPH WHICH
IS BETTER THAN RANDOM WILL BE QUASIRANDOM IN THE SENSE DElNED EARLIER AND
THEREFORE IT AND ITS COMPLEMENT WILL CONTAIN ABOUT THE SAME NUMBER OF
COPIES OF + AS A RANDOM GRAPH
4HOMASONS EXAMPLE IS CONSTRUCTED AS FOLLOWS (E lRST CHOOSES A lXED
GRAPH ( WITH K VERTICES /NE EXAMPLE THAT WORKS IS A CERTAIN PRODUCT OF
SEVEN TRIANGLES (E THEN PARTITIONS THE VERTICES OF A COMPLETE GRAPH INTO
EQUAL SIZED SETS 6 6K AND COLOURS AN EDGE BETWEEN X 6I AND Y 6J
BLUE IF IJ IS AN EDGE OF ( AND RED OTHERWISE $ETAILS CAN BE FOUND IN ;4=
.OTICE THAT 3ZEMERsEDIS REGULARITY LEMMA AND THE FACTS ABOUT QUASIRANDOM
GRAPHS IMPLY THAT ANY EXAMPLE IS NECESSARILY SIMILAR TO 4HOMASONS IN THAT
IT WILL INVOLVE PARTITIONING THE VERTICES INTO A FEW SETS AND CHANGING THE
EDGE PROBABILITIES BETWEEN THE SETS FROM TO SOMETHING ELSE ALTHOUGH
IT IS QUITE SURPRISING THAT 4HOMASON MANAGES WITH PROBABILITIES OF ZERO
AND ONE
!S FOR THE SECOND PROBLEM ONE OBVIOUS WAY TO COVER N EDGES WITH
DLOG NE TRIANGLE FREE GRAPHS IS TO USE BIPARTITE ONES /NE CAN LET THE VERTICES
OF ' BE ALL SEQUENCES n nK OF LENGTH K AND FOR EACH I DElNE A
BIPARTITE GRAPH (I BY JOINING n nK TO p pK IF AND ONLY IF nI pI
/NCE AGAIN WE HAVE A VERY NATURAL EXAMPLE WHICH LOOKS AS THOUGH IT MIGHT
BE BEST POSSIBLE (OWEVER IT IS NOT
)N ORDER TO DEMONSTRATE THIS WE MAKE A SIMPLE OBSERVATION 3UPPOSE
THAT ' IS A COMPLETE GRAPH ON M VERTICES WHICH CAN BE COVERED BY R TRIANGLE
FREE GRAPHS ( (R .OW LET K AND CONSIDER THE COMPLETE GRAPH
WITH VERTEX SET 'K &OR I R AND J K DElNE A GRAPH (IJ BY
JOINING X XK TO Y YK IF AND ONLY IF XJ YJ IS AN EDGE OF (I )T IS
EASY TO CHECK THAT THE GRAPHS (IJ ARE TRIANGLE FREE AND COVER ALL EDGES OF 'K
4HE NUMBER OF GRAPHS (IJ IS RK AND 'K HAS MK VERTICES )F N MK THEN
RK LOG NLOG M R SO TO BEAT THE lRST BOUND ALL WE NEED TO DO IS lND
AN EXAMPLE WHERE LOG M R IS GREATER THAN LOG /NE SUCH EXAMPLE EXISTS
WHEN M 4HE COMPLETE GRAPH ON lVE VERTICES CAN BE COVERED WITH
TWO lVE CYCLES AND LOG LOG ! BETTER EXAMPLE STILL IS OBTAINED
74 '/7%23 '!&!
AS FOLLOWS ,ET ' BE THE COMPLETE GRAPH + AND IDENTIFY THE VERTICES OF
' WITH THE INTEGERS MODULO 4HE QUARTIC RESIDUES FORM A SUBGROUP 1
OF THE MULTIPLICATIVE GROUP OF & 4HIS SUBGROUP HAS INDEX FOUR AND IT
CAN BE CHECKED THAT NO TWO OF ITS ELEMENTS DIkER BY ONE 7E NOW COLOUR
THE EDGES XY OF ' WITH FOUR COLOURS ACCORDING TO THE COSET OF X _ Y 3INCE
_ IS A QUARTIC RESIDUE THE COSET OF X _ Y EQUALS THE COSET OF Y _ X )T
IS EASY TO SEE THAT A MONOCHROMATIC TRIANGLE WOULD IMPLY THE EXISTENCE OF
FOUR NON ZERO ELEMENTS A U V W OF & SUCH THAT AU AV AW 4HIS IN
TURN WOULD IMPLY THE EXISTENCE OF NON ZERO P AND Q SUCH THAT P Q
WHICH WERE CHECKED NOT TO EXIST 4HEREFORE EACH COLOUR CLASS REPRESENTS A
TRIANGLE FREE GRAPH AND IN THIS CASE LOG M R IS LOG WHICH IS LARGER
THAN LOG
%VEN THIS IS NOT THE BEST KNOWN EXAMPLE WHICH AS ) LEARNED FROM .OGA
!LON AT THE CONFERENCE IS RELATED TO A CERTAIN #AYLEY GRAPH WITH OVER THREE
HUNDRED VERTICES AND WAS DISCOVERED BY A COMPUTER
&OR THE THIRD PROBLEM IF ( IS A MODERATELY COMPLICATED GRAPH WITH
SEVEN VERTICES SAY THEN THERE IS NO HOPE OF GUESSING WHAT THE EXTREMAL
GRAPH ' WILL BE 4HE RESULT IS PLAUSIBLE BECAUSE THE CONDITION THAT '
CONTAINS NO INDUCED COPY OF ( RULES OUT ALMOST ANY USE OF RANDOMNESS
WHICH SEEMS TO BE VERY IMPORTANT IF A GRAPH IS NOT TO CONTAIN LARGE CLIQUES
OR INDEPENDENT SETS
)T IS INTERESTING TO THINK ABOUT WHY FOR SOME EXTREMAL PROBLEMS IT IS EASY
TO IDENTIFY THE BEST POSSIBLE EXAMPLES WHILE FOR OTHERS IT IS MUCH HARDER
4HE THREE PROBLEMS ABOVE ARE ALL HARD BECAUSE THERE ARE TWO COMPLETELY
DIkERENT TECHNIQUES FOR PRODUCING EXAMPLES AND ONE CAN PRODUCE A WHOLE
SPECTRUM OF FURTHER EXAMPLES BY MIXING THEM &OR EXAMPLE THE FOLLOWING
GRAPH ' SHOWS THAT 2K K K ,ET THE VERTICES OF ' BE 6 ; ` ` ` ; 6K
WHERE THE 6I ARE DISJOINT SETS OF SIZE K *OIN X 6I TO Y 6J IF AND ONLY
IF I J !T ONE TIME 4URs AN BELIEVED THAT THE BOUND OF K WAS SHARP AND
ALTHOUGH WE NOW KNOW THAT HE WAS COMPLETELY WRONG 4HOMASONS EXAMPLE
SHOWS THAT IT IS DInCULT TO DISMISS EXAMPLES THAT MIX THIS SORT OF EXPLICIT
APPROACH WITH A PROBABILISTIC ONE 4HIS ALSO ACCOUNTS FOR THE DInCULTY WITH
THE THIRD PROBLEM ! SIMILAR PHENOMENON OCCURS WITH THE SECOND IN THAT
THERE IS A SPECTRUM OF METHODS OF CONSTRUCTING TRIANGLE FREE GRAPHS !T ONE
END IS A COMPLETE BIPARTITE GRAPH AND AT THE OTHER IS A RANDOM GRAPH WITH
APPROPRIATE EDGE PROBABILITIES AND A FEW DELETIONS )N BETWEEN ONE CAN
CONSTRUCT A TRIANGLE FREE GRAPH ( ON R VERTICES USING RANDOM METHODS AND
THEN BLOW IT UP TO A GRAPH ' ON N VERTICES BY PARTITIONING THE VERTICES OF
6ISIONS IN -ATH 2/5'( 3425#452% !.$ #,!33)&)#!4)/.
' INTO SETS 6 6R AND JOINING X 6I TO Y 6J IF AND ONLY IF IJ IS AN

EDGE OF (
)T IS BECAUSE THE EXTREMAL GRAPHS FOR THESE PROBLEMS ARE NOT OBVIOUS
THAT THERE SEEMS TO BE A NEED FOR SOME SORT OF CLASSIlCATION 2ETURNING TO
THE PROBLEM OF ESTIMATING 2K K ALTHOUGH THE RANDOM GRAPH MAY WELL
GIVE THE RIGHT ANSWER IT APPEARS TO BE NECESSARY TO TAKE ACCOUNT OF GRAPHS
SUCH AS THE ONE CONSTRUCTED BY 4HOMASON EVEN IF ONLY TO RULE THEM OUT
/NE MIGHT IMAGINE THAT 3ZEMERsEDIS REGULARITY LEMMA WAS IDEALLY SUITED
TO THE PURPOSE BUT THE POOR BOUNDS IN THE LEMMA MAKE IT FAR TOO CRUDE
)N FACT EVEN THE DElNITION OF QUASIRANDOM GRAPHS IS TOO CRUDE BECAUSE
IT IS PERFECTLY POSSIBLE FOR A GRAPH ' ON N VERTICES TO BE N REGULAR AND
P
HAVE A SECOND LARGEST EIGENVALUE PROPORTIONAL TO N BUT TO CONTAIN A COM
P
PLETE SUBGRAPH WITH N VERTICES )N OTHER WORDS THE DISTINCTION BETWEEN
TWO QUASIRANDOM GRAPHS AS PREVIOUSLY DElNED CAN BE VERY IMPORTANT IN
THE CONTEXT OF 2AMSEYS THEOREM 4HIS SUGGESTS THE NEED FOR A STRONGER
DElNITION AS FAR AS ) KNOW NOTHING SUITABLE HAS BEEN PROPOSED
) HAVE CONCENTRATED SO FAR ON GRAPH THEORY BUT SIMILAR DInCULTIES ARISE
IN MANY OTHER CONTEXTS (ERE ARE TWO PROBLEMS IN COMBINATORIAL NUMBER
THEORY
2OTHS THEOREM ON ARITHMETIC PROGRESSIONS )N 3ZEMERsEDI
;3= PROVED THE FOLLOWING THEOREM WHICH HAD BEEN CONJECTURED BY %RD]OS
AND 4URsAN IN ;%4=
4HEOREM ,ET K BE AN INTEGER AND LET m 4HEN IF . IS SUnCIENTLY
LARGE EVERY SUBSET ! OF F . G OF CARDINALITY AT LEAST m. CONTAINS AN
ARITHMETIC PROGRESSION OF LENGTH K
4HE CORRECT ASYMPTOTIC DEPENDENCE OF . ON K AND m IS STILL A MAJOR OPEN
PROBLEM BUT THE PROBLEM ) WISH TO CONCENTRATE ON HERE IS MORE SPECIlC )T
IS TO DETERMINE THE DEPENDENCE OF . ON m WHEN K
4HE lRST PROGRESS TOWARDS THE %RD] OS 4URsAN CONJECTURE WAS DUE TO 2OTH
;2O= WHO SHOWED THAT THE CONJECTURE WAS TRUE WHEN K AND THAT
. COULD BE TAKEN TO BE EXP EXP#m 4HIS WAS LATER IMPROVED TO
EXPm # BY 3ZEMERsEDI ;3= AND (EATH "ROWN ;(E= 6ERY RECENTLY IT HAS
BEEN SHOWN BY "OURGAIN ;"O= THAT # CAN BE n -ORE PRECISELY IF
m C LOG LOG . LOG . _ THEN ! MUST CONTAIN AN ARITHMETIC PROGRESSION
OF LENGTH THREE
4HE BEST KNOWN BOUND IN THE OTHER DIRECTION IS GIVEN BY AN INGENIOUS
EXAMPLE DISCOVERED BY "EHREND IN ;"= )T SHOWS P THAT THE INEQUALITY
. EXP# LOGm _ OR EQUIVALENTLY m EXP_C LOG . IS NOT ENOUGH
74 '/7%23 '!&!
TO GUARANTEE AN ARITHMETIC PROGRESSION OF LENGTH THREE

4HIS SITUATION IS INTERESTING FOR A NUMBER OF REASONS &IRST THE BOUNDS
LEAVE OPEN THE QUESTION OF WHETHER A DENSITY m LOG . _ IS ENOUGH 4HIS
IS AN IMPORTANT QUESTION BECAUSE LOG . _ IS THE DENSITY OF THE PRIMES IN
THE INTERVAL F . G ! RESULT OF VAN DER #ORPUT BASED ON TECHNIQUES
OF 6INOGRADOV SHOWS THAT THE PRIMES CONTAIN INlNITELY MANY ARITHMETIC
PROGRESSIONS OF LENGTH THREE )T IS STILL POSSIBLE HOWEVER THAT THERE IS A
PROOF OF THIS FACT WHICH DEPENDS ONLY ON THE DENSITY OF THE PRIMES AND NOT
ON MORE DETAILED FACTS ABOUT THEIR DISTRIBUTION
! SECOND REASON IS THAT ALTHOUGH "EHRENDS EXAMPLE GIVES A VERY WEAK
BOUND COMPARED WITH WHAT IS KNOWN IN THE OTHER DIRECTION IT DOES AT LEAST
SHOW THAT PURE RANDOM METHODS DO NOT GIVE AN EXTREMAL SET 4HIS POINT
WAS DISCUSSED BY THE MATHEMATICIAN AND THE COMPUTER IN THE DIALOGUE IN
THE PREVIOUS SECTION
"OUNDS FOR &REIMANS THEOREM )N THE STATEMENT OF &REIMANS
THEOREM GIVEN EARLIER NO MENTION WAS MADE OF THE DEPENDENCE OF + AND D
ON # 4HIS IS STILL FAR FROM BEING COMPLETELY RESOLVED 3EE THE ARTICLE OF "ILU
;"I= FOR DETAILS ABOUT WHAT IS KNOWN )MPROVING THE KNOWN BOUNDS WOULD
HAVE IMPORTANT APPLICATIONS 4O GIVE SOME IDEA OF WHAT IS NOT KNOWN HERE
ARE TWO SPECIlC QUESTIONS )F ! IS A SET OF N INTEGERS AND J! !J #J!J
CAN ANYTHING BE SAID ABOUT THE STRUCTURE OF ! IF # Nj FOR SOME j
OR EVEN IF # LOG N 4HE EXISTING PROOFS GIVE NO INFORMATION AT ALL IN
THIS SITUATION &OR THE SECOND QUESTION LET ! SATISFY THE SAME CONDITIONS
$OES ! HAVE A SUBSET " OF SIZE CN SUCH THAT " IS CONTAINED IN A GENERALIZED
ARITHMETIC PROGRESSION 0 OF CARDINALITY + AND DIMENSION D WHERE C AND +
HAVE SOME REASONABLE DEPENDENCE ON # AND D ! LOG # FOR SOME ABSOLUTE
CONSTANT ! &OR APPLICATIONS IT WOULD STILL BE VERY INTERESTING IF D COULD
BE BOUNDED ABOVE BY LOG # !
4HESE TWO PROBLEMS RAISE DInCULTIES THAT ARE SIMILAR IN SPIRIT TO THE
DInCULTIES IN THE GRAPH THEORETIC PROBLEMS "EHRENDS EXAMPLE SHOWS THAT
THERE IS NOT AN OBVIOUS EXTREMAL SET FOR 0ROBLEM AND ONCE AGAIN THERE IS
A CONmICT BETWEEN RANDOM AND EXPLICIT TECHNIQUES WHICH PARTIALLY ACCOUNTS
FOR THIS SITUATION 3IMILARLY THE SETS THAT ARISE IN ANY CONSIDERATION OF
&REIMANS THEOREM ARE VERY VARIED 3O FOR BOTH PROBLEMS A CLASSIlCATION
WOULD ALMOST CERTAINLY BE USEFUL
-ORE SPECIlCALLY FOR BOTH PROBLEMS ONE lNDS ONESELF WANTING TO UN
DERSTAND SUBSETS ! y :. OF SIZE j. WHERE j IS A CONSTANT OR PERHAPS
A NUMBER THAT TENDS TO ZERO VERY SLOWLY AS . TENDS TO INlNITY (ERE :.
6ISIONS IN -ATH 2/5'( 3425#452% !.$ #,!33)&)#!4)/.
IS THE GROUP OF INTEGERS MODULO . ! GOOD WAY TO DO THIS IS TO LOOK

AT THE &OURIER
0 TRANSFORM OF THE CHARACTERISTIC FUNCTION OF ! DElNED BY
>
!R S! ERS. WHERE EX IS NOTATION FOR EXPxIX 4HE COEn
>
CIENT ! >
IS JUST THE CARDINALITY OF ! )F !R IS SIGNIlCANTLY SMALLER THAN

j . FOR EVERY NON ZERO R THEN ! BEHAVES IN MANY WAYS LIKE A RANDOM SUB
SET OF :. WHERE EVERY ELEMENT IS CHOSEN WITH PROBABILITY j 7E THEREFORE
HAVE A GOOD DElNITION OF QUASIRANDOMNESS FOR SUBSETS OF :. 3EE ;#'= FOR
MORE DETAILS ABOUT THIS
.OTICE THAT THE INFORMATION CONVEYED BY &OURIER COEnCIENTS ABOUT SUB
SETS OF :. IS VERY SIMILAR TO THE INFORMATION CONVEYED BY EIGENVALUES ABOUT
GRAPHS 4HIS IS NOT A COINCIDENCE IF ! IS A SUBSET OF :. ONE CAN DElNE
A DIRECTED GRAPH '! WITH VERTEX SET :. BY HAVING AN EDGE FROM X TO Y IF
AND ONLY IF Y _ X ! AND IT IS EASY TO CHECK THAT THE EIGENVALUES OF '!
ARE THE &OURIER COEnCIENTS OF ! 4HE ANALOGIES GO FURTHER *UST AS ONE CAN
TRY TO CLASSIFY GRAPHS ONE CAN ALSO TRY TO CLASSIFY SUBSETS OF :. &OR THIS
ONE NEEDS A NOTION OF <NOT INTERESTINGLY DIkERENT IT TURNS OUT THAT THE
DIkERENCE BETWEEN TWO SETS ! AND " IS FOR MANY PURPOSES UNIMPORTANT IF
> _ "R
!R > IS SMALL FOR EVERY R
&OURIER ANALYSIS HAS BEEN AN ESSENTIAL PART OF THE BEST ARGUMENTS KNOWN
FOR ANALYSING SUBSETS ! OF :. IN CONNECTION WITH 0ROBLEMS AND 4HE
REASON IS THAT BECAUSE PERTURBING THE &OURIER COEnCIENTS OF A SET ! DOES
NOT MAKE AN IMPORTANT DIkERENCE WE ONLY HAVE TO WORRY ABOUT THE LARGE
COEnCIENTS 4HERE ARE VERY FEW OF THESE SO WE HAVE THE FOLLOWING EnCIENT
WAY OF ENCODING THE ESSENTIAL INFORMATION ABOUT ! CHOOSE A CONSTANT k
>
LET + FR J!R J k. G AND FOR R + LET F! R !R > )T MIGHT
SEEM AS THOUGH MAPPING SETS ! TO THEIR FUNCTIONS F! IS A READY MADE
CLASSIlCATION SCHEME BUT ) WOULD REGARD <SEMI CLASSIlCATION AS A BETTER
DESCRIPTION BECAUSE IT IS NOT EASY TO USE ALL THE INFORMATION CONTAINED
IN F! 4HE KNOWN ARGUMENTS TEND TO THROW AWAY ALL INFORMATION ABOUT THE
SET + EXCEPT FOR ITS CARDINALITY BUT THE STRUCTURE OF + IS VERY IMPORTANT
AS WELL )N PARTICULAR THE BEHAVIOUR OF ! DEPENDS VERY MUCH ON WHETHER
THERE ARE SIMPLE ARITHMETICAL RELATIONSHIPS BETWEEN THE ELEMENTS OF + SUCH
AS TWO OF THEM ADDING TO A THIRD 5NFORTUNATELY IT IS NOT EASY TO SEE HOW
TO INCORPORATE THIS INFORMATION INTO ARGUMENTS AND THIS SEEMS TO BE A
BARRIER TO FURTHER PROGRESS ON THE TWO PROBLEMS /NCE AGAIN WHAT APPEARS
TO BE MISSING IS A BETTER CLASSIlCATION SCHEME 3IMILAR REMARKS APPLY TO
EIGENVALUES OF GRAPHS 5SEFUL THOUGH THEY ARE FOR MANY PROBLEMS IT IS
IMPORTANT TO KNOW HOW THE CORRESPONDING EIGENVECTORS lT TOGETHER AND
74 '/7%23 '!&!
THIS INFORMATION IS HARD TO USE

4HE NEXT TWO PROBLEMS ARE DESIGNED TO SHOW THAT DInCULTIES SIMILAR TO
THE ONES DISCUSSED SO FAR CAN ARISE IN MANY CONTEXTS
! PROBLEM ABOUT SUBSETS OF THE N SPHERE $ENOTE BY 3N_ THE
%UCLIDEAN UNIT SPHERE OF 2N ,ET M N LET n AND LET ! BE A SUBSET OF
3N_ THAT INTERSECTS EVERY M DIMENSIONAL SUBSPACE OF 2N 7RITE !n FOR THE
SET FX DX ! nG 4HERE ARE INTERESTING QUESTIONS ONE CAN ASK ABOUT
THE SET !n &OR EXAMPLE HOW SMALL CAN ITS MEASURE BE &OR GIVEN VALUES
OF M AND n WHAT IS THE LARGEST R SUCH THAT !n IS FORCED TO CONTAIN THE UNIT
SPHERE OF AN R DIMENSIONAL SUBSPACE
4HESE QUESTIONS ARE HARD BECAUSE THERE ARE TWO NATURAL WAYS OF CON
STRUCTING A SET ! WITH THE GIVEN PROPERTY 4HE lRST IS TO TAKE THE UNIT
SPHERE OF AN N _ M DIMENSIONAL SUBSPACE AND THE SECOND IS TO TAKE A
UNION OF RANDOMLY PLACED SPHERICAL CAPS OF SOME GIVEN RADIUS !GAIN WE
ARE IN A SITUATION WHERE MANY OTHER CONSTRUCTIONS CAN BE DEVISED THAT MIX
THESE TWO IDEAS FOR EXAMPLE INSTEAD OF CAPS WHICH ARE EXPANSIONS OF ZERO
DIMENSIONAL SETS ONE COULD TAKE EXPANSIONS OF K DIMENSIONAL SETS FOR SOME
SMALL K )T SEEMS LIKELY THAT IF ! IS REQUIRED TO BE CENTRALLY SYMMETRIC
THEN THE MEASURE OF !n IS MINIMIZED WHEN ! IS THE SPHERE OF AN N _ M
DIMENSIONAL SUBSPACE /N THE OTHER HAND IT DOES NOT SEEM NECESSARY FOR
!n TO CONTAIN THE SPHERE OF SUCH A LARGE SUBSPACE 0ERHAPS ONE CAN DO
BETTER BY ALLOWING !n TO BE LARGER BUT LESS REGULAR
&OR SOME VALUES OF M AND n A ROUGH CLASSIlCATION OF SUBSETS OF 3N_
MIGHT HELP TO DETERMINE A GOOD BOUND FOR R /NE WOULD LIKE TO EXPLOIT
THE FACT THAT SETS ! FOR WHICH !n IS SMALL HAVE SOME STRUCTURE AND DEDUCE
FROM THIS THAT R IS LARGER THAN IS IMPLIED BY MEASURE THEORETIC CONSIDERATIONS
ALONE
/NE POSSIBLE METHOD OF PROVIDING UPPER BOUNDS FOR R IS TO LET " BE
THE UNIT SPHERE OF AN N _ M DIMENSIONAL SUBSPACE LET } BE A CONTINUOUS
ANTIPODAL MAP FROM 3N_ TO ITSELF AND LET ! }" 4HIS MAKES ! A SORT
OF TOPOLOGICAL SPHERE OF DIMENSION N _ M 5NFORTUNATELY IT IS HARD TO THINK
OF A CHOICE FOR } WHICH WILL PROVIDE A GOOD BOUND AND BE EASY TO ANALYSE
! PROBLEM ABOUT LAYERS OF THE DISCRETE CUBE ,ET R S
T N BE POSITIVE INTEGERS WRITE ;N=S FOR THE SET OF ALL SUBSETS OF ;N=
F NG OF SIZE S AND LET ! BE A SUBSET OF ;N=S 4HAT IS ! IS A
COLLECTION OF SETS OF SIZE S 7RITE R ! FOR THE COLLECTION OF ALL SETS " ;N=R
SUCH THAT " y ! FOR SOME ! ! AND T ! FOR THE COLLECTION OF ALL SETS
# ;N=T SUCH THAT ! y # FOR SOME ! ! 3UPPOSE THAT EVERY SET IN
6ISIONS IN -ATH 2/5'( 3425#452% !.$ #,!33)&)#!4)/.
;N=T CONTAINS SOME SET IN ! SO T ! ;N=T (OW SMALL CAN R ! BE

4HE NATURE OF THIS PROBLEM DEPENDS A GREAT DEAL ON THE VALUES OF R S T
AND N .OTICE THAT ONCE AGAIN THERE ARE TWO COMPLETELY DIkERENT WAYS
OF CONSTRUCTING A SET ! THAT SATISlES THE CONDITIONS OF THE PROBLEM )F
ONE WISHES TO MINIMIZE THE CARDINALITY OF ! THEN IT IS BEST TO CHOOSE IT
RANDOMLY (OWEVER AMONGST ALL SETS ! OF A GIVEN CARDINALITY RANDOM ONES
ARE THOSE FOR WHICH R ! IS LARGEST !N ALTERNATIVE CONSTRUCTION EXPLOITS THE
PIGEONHOLE PRINCIPLE IN A SIMPLE WAY &OR CONVENIENCE LET US SUPPOSE THAT S
DIVIDES T AND THAT K TS DIVIDES N 0ARTITION THE SET ;N= INTO K EQUAL SIZED
SUBSETS 6 ; ` ` ` ; 6K AND LET ! BE THE COLLECTION OF ALL SETS ! OF SIZE R SUCH
THAT ! y 6I FOR SOME I %VERY SET OF SIZE T INTERSECTS AT LEAST ONE OF THE 6I
IN AT LEAST NK S POINTS SO ! SATISlES THE REQUIRED CONDITION
,ET US SUPPOSE FOR A MOMENT THAT S IS MUCH SMALLER THAN T )F R IS ONLY
VERY SLIGHTLY SMALLER THAN S THEN IT LOOKS AS THOUGH A RANDOM CHOICE OF !
IS BEST SINCE EACH SET OF ! WILL CONTAIN ONLY A FEW SETS IN ;N=R )F ON THE
OTHER HAND R IS MUCH SMALLER THAN S THEN THE EXPLICIT CONSTRUCTION TURNS
OUT TO BE BETTER "ETWEEN THESE TWO EXTREMES THERE IS A RANGE OF VALUES OF R
WHERE IT IS NOT OBVIOUS WHICH CONSTRUCTION IS BETTER OR WHETHER SOMETHING
<IN BETWEEN IS MORE SUITABLE )N FACT ) DO NOT EVEN KNOW HOW TO PROVE
WHAT APPEARS TO BE THE CASE IN THE EXTREME SITUATIONS [ THAT THE EXAMPLES
) HAVE SKETCHED ARE BEST POSSIBLE
4HIS PROBLEM IS PERHAPS SLIGHTLY ARTIlCIAL BUT IT IS A SIMPLE EXAMPLE
OF A TYPE OF DInCULTY ) HAVE MET SEVERAL TIMES WHEN THINKING ABOUT OTHER
QUESTIONS AND ) DOUBT WHETHER ) AM ALONE IN THIS 7HATEVER ONE THINKS
ABOUT THE INTEREST OR OTHERWISE OF THE PROBLEM IT IS HARD TO IMAGINE A
SOLUTION BEING DULL /NCE AGAIN IT LOOKS AS THOUGH A ROUGH CLASSIlCATION
WOULD BE VERY USEFUL THIS TIME OF SUBSETS OF ;N=S 4HERE DO EXIST VERSIONS
OF 3ZEMERsEDIS REGULARITY LEMMA WHEN S IS lXED BUT THEY WOULD NOT HELP
P
IF ONE TOOK S N FOR EXAMPLE
&INDING ROUGH CLASSIlCATIONS !RISING FROM THE PREVIOUS PROB
LEMS IS THE GENERAL PROBLEM OF lNDING NEW AND USEFUL ROUGH CLASSIlCATIONS
!MONGST THE OBJECTS THAT ONE MIGHT LIKE TO CLASSIFY ARE THE FOLLOWING
I 'RAPHS 4HE PROBLEM HERE IS TO lND A USEFUL REPLACEMENT OF 3ZE
MERsEDIS REGULARITY LEMMA "ECAUSE THE BAD BOUND IN THE LEMMA IS
IN A VERY WEAK SENSE TIGHT REPLACING THE LEMMA WOULD MEAN RE
PLACING THE DESCRIPTION IT GIVES OF A GENERAL GRAPH BY SOMETHING MORE
COMPLICATED IN RETURN FOR MUCH BETTER BOUNDS !NOTHER INTERESTING
PROJECT IS TO lND ALTERNATIVE PROOFS FOR RESULTS THAT USE THE LEMMA
74 '/7%23 '!&!
JUST AS IT IS INTERESTING TO REPLACE A PROOF THAT RELIES ON THE CLASSIlCA

TION OF lNITE SIMPLE GROUPS BY ONE THAT DOES NOT 3EE ;+O3= ;'R22=
AND ;' 0ROP = FOR SOME EXAMPLES OF HOW IT CAN BE AVOIDED &OR
MORE GENERAL INFORMATION ON THE REGULARITY LEMMA AND ITS VARIANTS
SEE ;+O3=
II 4RIANGLE FREE GRAPHS 4WO WAYS OF CONSTRUCTING THESE ARE TO USE RAN
DOM METHODS OR TO TAKE A GIVEN TRIANGLE FREE GRAPH ( WITH VERTICES
X XK AND TO DElNE A GRAPH ' WITH VERTEX SET PARTITIONED INTO
SETS 6 6K BY JOINING X 6I TO Y 6J IF AND ONLY IF XI XJ IS AN EDGE
OF ( #OMBINING THESE METHODS AND OTHER SIMPLE OPERATIONS SUCH
AS SUBGRAPHS AND DISJOINT UNIONS YIELDS A LARGE CLASS OF TRIANGLE FREE
GRAPHS $O TECHNIQUES SUCH AS THESE GET ANYWHERE NEAR TO GENERATING
ALL TRIANGLE FREE GRAPHS
III 3PARSE GRAPHS )T HAS LONG BEEN RECOGNISED THAT A MAJOR DRAWBACK
WITH 3ZEMERsEDIS REGULARITY LEMMA IS THAT THE ONLY GRAPHS FOR WHICH IT
GIVES
_ ÀNY INFORMATION ARE THOSE WITH A NUMBER OF EDGES PROPORTIONAL
TO N WHILE MANY PROBLEMS OF GREAT INTEREST CONCERN GRAPHS WITH
SAY Nj EDGES FOR SOME j 3OME VARIANTS OF THE REGULARITY
LEMMA DO EXIST FOR SUCH SPARSE GRAPHS BUT THEY FALL WELL SHORT OF
ANYTHING THAT ONE MIGHT WISH TO CALL A CLASSIlCATION
IV 3UBSETS OF !BELIAN GROUPS ESPECIALLY :. AND : WITH TWO SETS RE
GARDED AS SIMILAR IF THEIR &OURIER TRANSFORMS ARE CLOSE
V 3UBSETS OF :. WITH TWO SETS REGARDED AS SIMILAR IF THEIR CHARACTERISTIC
FUNCTIONS ARE CLOSE IN THE FOLLOWING NORM
8 9
KF K . _ Fn X _ n A _ n B _ n C
XABC:. nFG
(ERE Fn DENOTES F IF n n n IS EVEN AND THE COMPLEX CONJUGATE

OF F OTHERWISE )T TURNS OUT THAT WHILE THE LARGE &OURIER COEnCIENTS
OF A SET ! CONTAIN ALL THE INFORMATION ABOUT ARITHMETIC PROGRESSIONS
OF LENGTH THREE THEY DO NOT DO SO FOR LONGER PROGRESSIONS )F ! AND
" ARE CLOSE IN THE ABOVE NORM THEN THEY WILL CONTAIN APPROXIMATELY
THE SAME NUMBER OF ARITHMETIC PROGRESSIONS OF LENGTH FOUR [ HENCE ITS
IMPORTANCE &OR MORE DETAILS SEE ;'= ! CLASSIlCATION WITH RESPECT
TO THIS NORM WOULD NECESSARILY BE MUCH MORE SOPHISTICATED THAN A
CLASSIlCATION WITH RESPECT TO CLOSENESS OF &OURIER COEnCIENTS
VI 3UBSETS OF METRIC SPACES SUCH AS ; = AND 3N_ WITH TWO SETS RE
GARDED AS SIMILAR IF THEIR METRIC PROPERTIES ARE ROUGHLY THE SAME
4HERE ARE MANY WAYS OF MAKING THIS QUESTION PRECISE SOME OF WHICH
6ISIONS IN -ATH 2/5'( 3425#452% !.$ #,!33)&)#!4)/.
HAVE BEEN STUDIED EXTENSIVELY /NE POSSIBILITY IS TO SAY THAT SUBSETS

! AND " OF A METRIC SPACE 8 ARE SIMILAR IF THERE IS A SMALL n AND
THERE ARE ,IPSCHITZ MAPS } ! 8 AND " 8 SUCH THAT
EVERY POINT OF " IS WITHIN n OF SOME POINT OF }! EVERY POINT OF !
IS WITHIN n OF SOME POINT OF " AND DX }X AND DY }Y ARE AT
MOST n FOR EVERY X ! AND Y "
VII 3UBSETS OF COMBINATORIAL STRUCTURES SUCH AS THE DISCRETE CUBE F GN
AND ;N=R
! &URTHER 3ELECTION OF 0ROBLEMS

%VERY PROBLEM IN THIS SECTION HAD TO SATISFY TWO CRITERIA &IRST ) MYSELF
SHOULD HAVE SPENT AT LEAST AN HOUR TRYING TO SOLVE IT AND SECOND ) SHOULD
HAVE AT LEAST SOME INTENTION OF SPENDING MORE TIME ON IT IN THE FUTURE 4HESE
WERE ENOUGH TO RULE OUT MOST OF THE INTERESTING PROBLEMS IN MATHEMATICS
LEAVING ME WITH A MANAGEABLE LIST
4HE 2IEMANN HYPOTHESIS 4HIS IS SO WELL KNOWN THAT ) SHALL NOT
EVEN BOTHER TO STATE IT HERE ) INCLUDE IT BECAUSE ) DO NOT WISH TO MAKE
SOME KIND OF STATEMENT BY NOT DOING SO (OWEVER ) HAVE NOTHING MUCH
TO SAY ABOUT IT )NSTEAD ) REFER THE READER TO THE VAST LITERATURE ON THE
SUBJECT INCLUDING PARTS OF THE ARTICLES OF #ONNES )WANIEC AND 3ARNAK IN
THIS VOLUME
4HE +AKEYA PROBLEM 4HIS IS CLOSELY RELATED TO THE 2IEMANN
HYPOTHESIS )T HAS MANY FORMULATIONS OF WHICH HERE ARE THREE
I ,ET ! BE A SUBSET OF 2D SUCH THAT FOR EVERY LINE , IN 2D ! CONTAINS A
LINE , PARALLEL TO , 3UCH A SET IS OFTEN CALLED A +AKEYA SET OR A "ESICOVITCH
SET "ESICOVITCH SHOWED THAT THE MEASURE OF ! CAN BE ZERO BUT ALL KNOWN
EXAMPLES HAVE (AUSDORk DIMENSION D "OURGAIN ;"O= HAS SHOWN THAT THE
(AUSDORk DIMENSION OF ! MUST BE AT LEAST D
WHICH IS A SIGNIlCANT
IMPROVEMENT ON A RELATIVELY EASY BOUND OF D 6ERY RECENTLY THIS BOUND
WAS IMPROVED FURTHER BY +ATZ AND 4AO TO D
;+A4= )T IS KNOWN THAT
THE DIMENSION MUST BE D WHEN D BUT FOR D THE PROBLEM IS OPEN
4HE lRST NON TRIVIAL LOWER BOUND WHEN D WAS DISCOVERED BY "OURGAIN
;"O= (IS ARGUMENT SHOWED THAT THE DIMENSION OF A THREE DIMENSIONAL
+AKEYA SET WAS AT LEAST 4HIS WAS IMPROVED BY 7OLk TO ;7=
4HIS BOUND WAS RECENTLY IMPROVED VERY SLIGHTLY BY +ATZ ,ABA AND 4AO
;+A,4= AN INTERESTING RESULT AS WAS A NATURAL LIMIT IN 7OLkS PROOF
AND SURPASSING IT WAS VERY DInCULT
74 '/7%23 '!&!
II ,ET 3 BE THE SQUARE FX Y Z 2 Z X Y G AND

LET 0 BE THE POINT .OW DIVIDE 3 INTO N SUBSQUARES 3IJ OF
SIDELENGTH N_ IN THE OBVIOUS WAY &OR EACH OF THESE SUBSQUARES FORM A
CONE #IJ BY TAKING THE CONVEX HULL OF 3IJ AND 0 4RANSLATE EACH CONE #IJ IN
AN XY DIRECTION OR IN OTHER WORDS SLIDE ITS BASE TO SOMEWHERE ELSE IN THE
XY PLANE 4HEN HOW SMALL CAN THE VOLUME
AND CALL THE TRANSLATED CONE #IJ
3N
OF IJ #IJ BE )F THE SMALLEST POSSIBLE VOLUME BEHAVES LIKE N_j FOR SOME

j THEN THERE IS A +AKEYA SET IN 2 WITH (AUSDORk DIMENSION LESS THAN

THREE 4HE CONVERSE IS ALSO TRUE 4HE OBVIOUS D DIMENSIONAL ANALOGUE OF
THIS EQUIVALENCE HOLDS AS WELL
III ,ET M AND R BE INTEGERS AND FOR EVERY D BETWEEN AND R LET 0D y :
BE AN ARITHMETIC PROGRESSION
3 OF LENGTH M AND COMMON DIkERENCE D (OW
SMALL CAN THE UNION RD 0D BE )N PARTICULAR IF R M# FOR SOME lXED
INTEGER # THEN CAN ONE IMPROVE ON THE TRIVIAL BOUND OF MR BY A FACTOR
OF Mj FOR SOME j /NCE AGAIN A CONSTRUCTION THAT ACHIEVED THIS COULD
BE TRANSLATED INTO AN EXAMPLE OF A +AKEYA SET WITH LESS THAN FULL DIMENSION
4HE THIRD VERSION OF THE PROBLEM GIVES SOME CLUE THAT IT HAS A NUMBER
THEORETICAL CONTENT ! CLOSE CONNECTION WITH A CONJECTURE OF -ONTGOMERY
SEE ;-O #HAPTER = AND HENCE WITH THE ZEROS OF THE ZETA FUNCTION WAS
UNCOVERED BY "OURGAIN IN HIS MODESTLY TITLED PAPER ;"O= %VER SINCE A
FAMOUS PAPER OF &EkERMAN ;&= IT HAS BEEN REALIZED THAT +AKEYA SETS ARE
ALSO OF GREAT IMPORTANCE IN HARMONIC ANALYSIS ! SOLUTION TO THE +AKEYA
PROBLEM WOULD THEREFORE BE A MAJOR ADVANCE FOR SEVERAL DIkERENT REASONS
$OES 0 EQUAL .0 ) HAVE ALREADY MENTIONED THIS PROBLEM IN X
3INCE NOT ALL MATHEMATICIANS FEEL COMFORTABLE STATING IT LET ME GIVE WHAT
) lND THE CLEANEST DElNITION OF THE CLASS OF .0 PROBLEMS &IRST IT IS CON
VENIENT TO THINK NOT ABOUT PROBLEMS BUT ABOUT "OOLEAN FUNCTIONS THAT IS
FUNCTIONS } F GN F G ! PROBLEM SUCH AS THE TRAVELLING SALESMAN
PROBLEM CAN EASILY BE REGARDED AS SUCH A FUNCTION _ ` ENCODE YOUR GRAPH '
WITH N VERTICES AS A SEQUENCE p' OF LENGTH N IN AN OBVIOUS WAY AND
DElNE }p' TO BE IF ' CONTAINS A (AMILTON CYCLE AND OTHERWISE !
"OOLEAN FUNCTION } BELONGS TO 0 IF THERE IS A POLYNOMIAL TIME ALGORITHM FOR
COMPUTING } 3TRICTLY SPEAKING ONE SHOULD TALK ABOUT A SEQUENCE }N OF
FUNCTIONS WHERE }N HAS F N VARIABLES AND THE TIME TAKEN TO COMPUTE }N
IS BOUNDED ABOVE BY A POLYNOMIAL FUNCTION OF F N )T BELONGS TO .0 IF IT
IS A PROJECTION OF A FUNCTION IN 0 IN THE FOLLOWING SENSE THERE IS AN INTEGER
M BOUNDED ABOVE BY A POLYNOMIAL IN N AND A FUNCTION F GM a F GN
WHICH BELONGS TO 0 SUCH THAT }p IF AND ONLY IF THERE EXISTS w F GM
6ISIONS IN -ATH 2/5'( 3425#452% !.$ #,!33)&)#!4)/.
SUCH THAT w p

&OR THE TRAVELLING SALESMAN PROBLEM w WILL REPRESENT AN ORDERING OF THE
VERTICES p A GRAPH AND w p IF AND ONLY IF THE ORDERING OF THE VERTICES
GIVEN BY w DElNES A (AMILTON CYCLE IN THE GRAPH GIVEN BY p 4HIS IS VERY
EASY TO CHECK SO BELONGS TO 0 AND THE FUNCTION } DElNED EARLIER WHICH
IS IF AND ONLY IF w p FOR SOME w IS IN .0
/NE FURTHER DElNITION WHICH ) SHALL NOT GIVE PRECISELY IS THAT OF AN
.0 COMPLETE FUNCTION !N IMPORTANT DISCOVERY OF #OOK ;#O= AND +ARP
;+= WHICH EXPLAINS THE INTEREST IN THE 0.0 PROBLEM IS THAT MANY .0
PROBLEMS OR CORRESPONDING "OOLEAN FUNCTIONS ARE UNIVERSAL IN THE SENSE
THAT A POLYNOMIAL TIME ALGORITHM FOR SOLVING ONE OF THEM CAN BE CONVERTED
INTO A POLYNOMIAL TIME ALGORITHM FOR SOLVING AN ARBITRARY PROBLEM IN .0
3UCH A PROBLEM IS CALLED .0 COMPLETE )N FACT HUNDREDS OF SUCH PROBLEMS
ARE NOW KNOWN AND MANY OF THEM ARE OF DIRECT PRACTICAL IMPORTANCE
4HE 0.0 PROBLEM MAY LOOK Ok PUTTING BECAUSE THE SET OF ALL FUNCTIONS
4URING COMPUTABLE IN POLYNOMIAL TIME IS SOMEHOW NOT VERY MATHEMATICAL
) HAVE SOMETIMES SEEN IT WRITTEN THAT THE PROBLEM IS DInCULT BECAUSE IT IS
HARD TO RULE OUT ALL POSSIBLE POLYNOMIAL TIME ALGORITHMS (OWEVER THERE
IS A VERY CLEAN AND COMPLETELY MATHEMATICAL REFORMULATION OF THE PROBLEM
IN TERMS OF SO CALLED CIRCUIT COMPLEXITY WHICH SHOWS THAT THIS IS NOT REALLY
THE DInCULTY AT ALL ) SHALL GIVE THE DElNITION IN AN EQUIVALENT FORM
,ET } F GN F G BE A "OOLEAN FUNCTION AND LET ! Fp }p G
&OR EACH I N LET %I Fp pI G AND CALL THE SETS %I AND THEIR
COMPLEMENTS ELEMENTARY SETS 4HE CIRCUIT COMPLEXITY OF } OR ! IS THE
LENGTH . OF THE SHORTEST SEQUENCE ! ! !. SUCH THAT !. ! AND
EACH !I IS EITHER AN ELEMENTARY SET THE COMPLEMENT OF SOME EARLIER !J OR
THE UNION OR INTERSECTION OF TWO EARLIER !J S
! PROBLEM WHICH IS ALMOST BUT NOT QUITE EQUIVALENT TO WHETHER 0.0
IS THE FOLLOWING ,ET } BE A "OOLEAN FUNCTION IN .0 -UST THE CIRCUIT COM
PLEXITY OF } BE BOUNDED ABOVE BY A POLYNOMIAL IN THE NUMBER OF VARIABLES
! POLYNOMIAL TIME ALGORITHM FOR COMPUTING } CAN BE CONVERTED INTO A SE
QUENCE ! !. OF POLYNOMIAL LENGTH SO A SUPERPOLYNOMIAL BOUND FOR
THE CIRCUIT COMPLEXITY WOULD IMPLY THAT 0 AND .0 ARE NOT EQUAL )T IS THEO
RETICALLY POSSIBLE THAT THE CONVERSE TO THIS IS FALSE BUT IT IS HARD TO ENVISAGE
A CONSTRUCTION OF A SEQUENCE ! !. WITH SO LITTLE REGULARITY THAT IT OR
SOME ADAPTATION OF IT COULD NOT BE EnCIENTLY GENERATED ALGORITHMICALLY )
AM THEREFORE CONVINCED THAT CIRCUIT COMPLEXITY IS THE RIGHT APPROACH TO THE
PROBLEM 4HIS IS THE STANDARD VIEW OF THEORETICAL COMPUTER SCIENTISTS BUT
74 '/7%23 '!&!
NOT OF ALL MATHEMATICIANS

5NFORTUNATELY IT DOES NOT MAKE THE PROBLEM EASY AND THE BEST KNOWN
LOWER BOUND FOR THE CIRCUIT COMPLEXITY OF AN .0 FUNCTION IS STILL ONLY LIN
EAR ! BEAUTIFUL AND IMPORTANT RECENT PAPER OF 2AZBOROV AND 2UDICH ;22=
GOES A LONG WAY TOWARDS EXPLAINING WHY THE PROBLEM IS DInCULT AND IS RE
QUIRED READING FOR ANYBODY WHO WOULD LIKE TO PROVE THAT 0.0 ! NATURAL
APPROACH TO PROVING LOWER BOUNDS FOR CIRCUIT COMPLEXITY IS TO DElNE SOME
MEASURE OF SIMPLICITY PROVE THAT SETS ! WITH SMALL CIRCUIT COMPLEXITY MUST
BE SIMPLE AND PROVE THAT THE SET ! CORRESPONDING TO SOME .0 FUNCTION
IS NOT SIMPLE 2AZBOROV AND 2UDICH DElNE A NOTION OF <NATURAL PROOF
WHICH BASICALLY MEANS A PROOF RESULTING FROM A <NATURAL MEASURE OF SIM
PLICITY 4HEY SHOW THAT ALL THE KNOWN BOUNDS IN CIRCUIT COMPLEXITY HAVE
BEEN ACHIEVED USING NATURAL PROOFS IN THEIR SENSE AND THEY THEN SHOW THAT
ANY NATURAL PROOF THAT 0.0 COULD BE CONVERTED INTO A POLYNOMIAL TIME
ALGORITHM FOR FACTORIZING LARGE INTEGERS 4HIS IS CONVINCING EVIDENCE THAT A
PROOF THAT 0.0 WOULD HAVE TO BE VERY STRANGE INDEED &OR MORE DETAILS
SEE THE CONTRIBUTIONS OF 2AZBOROV AND 7IGDERSON IN THIS VOLUME OR THE
PAPER OF 2AZBOROV AND 2UDICH ITSELF WHICH IS WELL WRITTEN AND NOT HARD TO
READ
4HE FOLLOWING PROBLEM MENTIONED BY 7IGDERSON AT THE CONFERENCE
HAS A SIMILAR mAVOUR TO THE 0.0 PROBLEM BUT HAS NOTHING TO DO WITH
COMPUTERS AND APPEARS AT lRST GLANCE TO BE MUCH SIMPLER ,ET - BE AN
NaN INVERTIBLE MATRIX OVER & 'AUSSIAN ELIMINATION TELLS US THAT - CAN BE
CONVERTED INTO THE IDENTITY MATRIX BY A SEQUENCE OF AT MOST N ELEMENTARY
ROW OPERATIONS -OREOVER A SIMPLE COUNTING ARGUMENT THERE ARE FEWER

THAN N ELEMENTARY ROW OPERATIONS AND ALMOST N NON SINGULAR MATRICES
SHOWS THAT ALMOST ALL MATRICES NEED AT LEAST CN LOG N OPERATIONS 4HE
PROBLEM IS TO GIVE AN EXPLICIT EXAMPLE OF A MATRIX THAT NEEDS A SUPERLINEAR
NUMBER <%XPLICIT DOES NOT MEAN <UNAMBIGUOUSLY DElNED OR ONE COULD
JUST ENUMERATE ALL THE MATRICES AND TAKE THE lRST ONE THAT WORKED /NE
POSSIBLE INTERPRETATION IS THAT IT SHOULD BE POSSIBLE TO GENERATE AN EXPLICIT
MATRIX OR RATHER SEQUENCE OF MATRICES WITH A POLYNOMIAL TIME ALGORITHM
4HIS PROBLEM REMAINS STUBBORNLY OPEN AND A SOLUTION MIGHT WELL SHED LIGHT
ON THE 0.0 PROBLEM ITSELF
!N EkECTIVE VERSION OF 2OTHS THEOREM "Y 2OTHS THEOREM )
NOW MEAN NOT HIS THEOREM ON ARITHMETIC PROGRESSIONS BUT HIS MUCH MORE FA
MOUS THEOREM ON RATIONAL APPROXIMATION OF ALGEBRAIC NUMBERS ;2O= 4HIS
STATES THAT FOR EVERY ALGEBRAIC NUMBER j AND EVERY n THERE EXISTS
6ISIONS IN -ATH 2/5'( 3425#452% !.$ #,!33)&)#!4)/.
C SUCH THAT FOR ANY PAIR P Q OF POSITIVE INTEGERS WE HAVE THE INEQUALITY
Jj _ PQJ CQ n 4HIS IS A VERY CONSIDERABLE STRENGTHENING OF ,IOUVILLES
THEOREM 4HE PROBLEM IS THAT THE PROOF GIVES NO INFORMATION AT ALL ABOUT
THE DEPENDENCE OF C ON j AND n 4HE SAME IS TRUE OF EARLIER WEAKER RESULTS
OF 4HUE AND 3IEGEL
4HE KNOWN RESULTS IN THIS DIRECTION ARE VERY MODEST &OR EXAMPLE
"AKER ;"= HAS PROVED AN EkECTIVE VERSION OF 2OTHS THEOREM IN THE CASE
j OBTAINING THE INEQUALITY J _ PQJ _ Q WHICH SHOULD
BE COMPARED WITH THE LOWER BOUND FROM ,IOUVILLES THEOREM OF CQ WHERE
C CAN BE TAKEN TO BE
! CLOSELY RELATED PROBLEM MENTIONED BY &URSTENBERG IN HIS TALK AT THIS
CONFERENCE CONCERNS THE CONTINUED FRACTION EXPANSION OF ALGEBRAIC NUMBERS
)F j IS A QUADRATIC IRRATIONAL THEN EVERYTHING IS KNOWN AND THE PARTIAL
QUOTIENTS ARE EVENTUALLY PERIODIC 2OTHS THEOREM IMPLIES THAT THE PARTIAL
QUOTIENTS OF AN ALGEBRAIC NUMBER CANNOT GROW TOO QUICKLY BUT EXCEPT IN
THE QUADRATIC CASE THERE IS NO EXAMPLE FOR WHICH IT IS KNOWN WHETHER THEY
ARE BOUNDED
,ITTLEWOODS CONJECTURE 4HIS PROBLEM WAS DISCUSSED BY -AR
GULIS AT THE CONFERENCE ,ITTLEWOOD CONJECTURED THAT IF j AND k ARE ANY TWO
REAL NUMBERS AND IF KXK DENOTES THE DISTANCE FROM X TO THE NEAREST INTEGER
THEN LIM INF N. N KNjK KNkK
4O UNDERSTAND THE PROBLEM CONSIDER A QUADRATIC IRRATIONAL j ,IOU
VILLES THEOREM IS EQUIVALENT TO THE STATEMENT THAT KNjK IS ALWAYS BOUNDED
BELOW BY CN_ /N THE OTHER HAND USUALLY KNjK IS MUCH BIGGER THAN THIS
/NE CAN THEREFORE ASK WHETHER IT IS POSSIBLE TO lND TWO NUMBERS j AND
k NEITHER OF WHICH CAN BE WELL APPROXIMATED BY RATIONALS WITH THE FUR
THER PROPERTY THAT KNkK IS LARGE WHENEVER KNjK IS SMALL AND VICE VERSA
,ITTLEWOODS CONJECTURE IS A PRECISE FORMULATION OF THIS QUESTION
4HE EXISTENCE OF SUCH A PAIR OF NUMBERS j k WOULD HAVE THE MORE
COMBINATORIAL CONSEQUENCE THAT FOR EVERY N ONE COULD lND A SET 8 OF N
POINTS
1 IN ; = AND AN _ABSOLUTE CONSTANT C SUCH THAT THE INEQUAL
ITY I JXI _ YI J CN HELD FOR ANY TWO DISTINCT POINTS X Y 8
)NDEED WRITING HTI FOR THE FRACTIONAL PART OF T A SUITABLE SET WOULD BE
FMN_ HjMI HkMI M NG !S FAR AS ) KNOW THE EXISTENCE OF SUCH
A SET 8 IS AN OPEN QUESTION SO IT IS POSSIBLE THAT ,ITTLEWOODS CONJECTURE
IF IT IS TRUE HAS A COMBINATORIAL PROOF ! SIMPLE ARGUMENT SHOWS THAT ONE
CAN lND N LOG N POINTS WITH THE GIVEN PROPERTY !NY IMPROVEMENT WOULD
BE INTERESTING
74 '/7%23 '!&!
)F ,ITTLEWOODS CONJECTURE P
IS FALSE
P THEN A GOOD CANDIDATE FOR A COUN
TEREXAMPLE WOULD BE THE PAIR )T SEEMS TO BE VERY HARD TO ANALYSE
CONCRETE EXAMPLES SUCH AS THIS AND PERHAPS THIS IS THE MAIN DInCULTY WITH
THE PROBLEM
3ELF AVOIDING RANDOM WALKS )N COMMON WITH MANY MATHEMATI
CIANS ) AM FASCINATED BY THE LONG LIST OF RESULTS IN STATISTICAL PHYSICS THAT
HAVE BEEN OBTAINED BY PHYSICISTS USING NON RIGOROUS ARGUMENTS /NE OF
THE MOST APPEALING SOURCE OF PROBLEMS IS THE AREA OF SELF AVOIDING RANDOM
WALKS ! SELF AVOIDING WALK OF LENGTH N IS A PATH X X XN OF VERTICES
IN THE LATTICE :D WHERE X XI_ AND XI ARE ADJACENT FOR I N AND
THERE ARE NO REPEATS )F ONE PLACES THE UNIFORM DISTRIBUTION ON THE SET OF ALL
SELF AVOIDING WALKS OF LENGTH N IN : OR : THEN ONE OBTAINS AN IMPORTANT
PROBABILISTIC MODEL ABOUT WHICH A GREAT DEAL IS KNOWN TO PHYSICISTS AND
ALMOST NOTHING TO MATHEMATICIANS ) WILL MENTION JUST A FEW PROBLEMS
.EARLY ALL THE MOST INTERESTING QUESTIONS ARE SPECIAL CASES OF THE FOLLOW
ING VAGUE ONE WHAT DOES A TYPICAL SELF AVOIDING WALK OF LENGTH N LOOK LIKE
0ERHAPS THE MOST OBVIOUS QUESTION IS THE EXPECTED DISTANCE FROM THE ORIGIN
TO XN )N TWO DIMENSIONS THE ARGUMENTS OF PHYSICISTS SHOW THAT THE ANSWER
OUGHT TO BE N BUT NO UPPER BOUND OF THE FORM N_n HAS BEEN PROVED
RIGOROUSLY 4HE SITUATION FOR LOWER BOUNDS IS EVEN WORSE /NE MIGHT EXPECT
A TRIVIAL ARGUMENT TO SHOW A LOWER BOUND OF CN SINCE A SELF AVOIDING WALK
SURELY TRAVELS FURTHER THAN A PURE RANDOM WALK (OWEVER IT IS NOT KNOWN
HOW TO PROVE RIGOROUSLY THAT A TYPICAL SELF AVOIDING WALK DOES NOT TEND TO
FOLD BACK ON ITSELF /F COURSE MOST POINTS OF A SELF AVOIDING WALK MUST HAVE
DISTANCE CN FROM THE ORIGIN BY A TRIVIAL COUNTING ARGUMENT SO THIS LAST
QUESTION IS SPECIlCALLY TO DO WITH THE END POINT
!NOTHER QUESTION CONCERNS THE NUMBER F N OF SELF AVOIDING WALKS OF
LENGTH N )T IS EASY TO SEE THAT F IS SUBMULTIPLICATIVE SO A STANDARD ARGU
MENT SHOWS THAT F N # N GN FOR SOME CONSTANT # KNOWN AS THE CONNEC
TIVE CONSTANT AND SOME FUNCTION G THAT GROWS SUBEXPONENTIALLY 4HE EXACT
VALUE OF # IS NOT KNOWN BUT THIS IS NOT A PARTICULARLY INTERESTING PROBLEM
FOR TWO REASONS &IRST # CAN IN PRINCIPLE BE DETERMINED TO ANY ACCURACY
BY AN EXACT COUNTING OF ALL WALKS OF SOME SUITABLE LENGTH N ALTHOUGH THIS
RAPIDLY BECOMES IMPOSSIBLE IN PRACTICE -ORE IMPORTANTLY # DOES NOT
HAVE A PROPERTY KNOWN AS UNIVERSALITY
/NE CAN CHANGE THE MODEL OF SELF AVOIDING WALKS IN MANY WAYS SUCH AS
CHANGING THE LATTICE IN WHICH IT LIVES )T IS KNOWN BY PHYSICISTS THAT CERTAIN
PARAMETERS ARE UNAkECTED BY SUCH CHANGES &OR EXAMPLE THE EXPECTED END
6ISIONS IN -ATH 2/5'( 3425#452% !.$ #,!33)&)#!4)/.
TO END DISTANCE IS ALWAYS PROPORTIONAL TO N 4HESE UNCHANGING PARAME

TERS ARE CALLED UNIVERSAL AND ARE REGARDED AS HAVING PHYSICAL SIGNIlCANCE
4HE CONNECTIVE CONSTANT IS NOT UNIVERSAL IN THIS SENSE AS IT CERTAINLY DOES
DEPEND ON THE UNDERLYING LATTICE
/N THE OTHER HAND THE FUNCTION G DOES SEEM TO BE UNIVERSAL /NCE
AGAIN MUCH MORE IS KNOWN TO PHYSICISTS THAN TO MATHEMATICIANS THEY
TELL US THAT GN GROWS LIKE N /NE MIGHT AT lRST WONDER HOW ONE
COULD HOPE TO STUDY THE FUNCTION G WHEN THE CONNECTIVE CONSTANT # IS NOT
KNOWN (OWEVER IT IS A SIMPLE EXERCISE TO SHOW THAT G HAS THE FOLLOWING
INTERPRETATION GN GROWS LIKE Nj IF AND ONLY IF THE PROBABILITY THAT TWO SELF
AVOIDING RANDOM WALKS OF LENGTH N DO NOT INTERSECT EXCEPT AT THE ORIGIN
BEHAVES LIKE N_j 4HIS SHOWS THAT G DOES HAVE SOME SORT OF INDEPENDENT
SIGNIlCANCE &ROM A PURE MATHEMATICAL POINT OF VIEW IT IS A VERY INTERESTING
OPEN PROBLEM TO PROVE THAT G IS BOUNDED ABOVE BY A POLYNOMIAL 4HE BEST
KNOWN BOUND IN TWO DIMENSIONS IS DUE TO (AMMERSLEY AND 7ELSH ;(A7=
WHO SHOWED BY AN INGENIOUS ELEMENTARY ARGUMENT THAT GN EXPCN
4HE REASON THE HYPOTHESIS OF UNIVERSALITY IS PLAUSIBLE IS THAT ONE WOULD
EXPECT THE MACROSCOPIC BEHAVIOUR OF A VERY LONG SELF AVOIDING WALK NOT TO
DEPEND TOO CRITICALLY ON THE PRECISE BEHAVIOUR AT VERY SMALL MAGNITUDE
SCALES #ONNECTED WITH THIS BELIEF IS THE VIEW THAT THERE SHOULD BE SOME
SORT OF APPROPRIATELY SCALED LIMIT OF SELF AVOIDING RANDOM WALKS PLAYING THE
ROLE THAT "ROWNIAN MOTION AND 7IENER MEASURE PLAY FOR ORDINARY RANDOM
WALKS 4HIS IS ANOTHER IMPORTANT OPEN PROBLEM
4HESE PROBLEMS ARE ALL AT THEIR MOST INTERESTING IN TWO AND THREE DIMEN
SIONS 3INCE THE (AUSDORk DIMENSION OF A "ROWNIAN PATH IS TWO ONE DOES
NOT EXPECT THE SELF AVOIDANCE CONSTRAINT TO BE ALL THAT SIGNIlCANT IN FOUR DI
MENSIONS OR MORE 4HIS INTUITION HAS IN FACT BEEN JUSTIlED RIGOROUSLY WHEN
THE DIMENSION IS AT LEAST lVE BY (ARA AND 3LADE ;(AR3= %VEN THOUGH
THEIR RESULTS ARE NOT VERY SURPRISING THEY ARE A MAJOR ACHIEVEMENT AND THE
PROOFS ARE NOT AT ALL EASY &OR MUCH MORE DETAIL ON ALL THESE PROBLEMS AND
RESULTS SEE THE EXCELLENT BOOK OF -ADRAS AND 3LADE ;-3=
5NTYING THE UNKNOT 4HERE HAS BEEN A GREAT DEAL OF EXCITEMENT
ABOUT KNOT THEORY IN THE LAST TWENTY YEARS INCLUDING THE DISCOVERY OF IM
PORTANT NEW KNOT INVARIANTS (OWEVER THE FOLLOWING BASIC PROBLEM IS OPEN
DOES THERE EXIST A POLYNOMIAL TIME ALGORITHM FOR RECOGNISING A DIAGRAM OF
THE UNKNOT AND FOR PRODUCING A SEQUENCE OF 2EIDEMEISTER MOVES THAT UN
TIES IT 4HE TIME IS A FUNCTION OF THE NUMBER OF CROSSINGS IN THE DIAGRAM
)F SUCH AN ALGORITHM COULD BE FOUND IT WOULD OPEN UP A COMPLETELY NEW
74 '/7%23 '!&!
APPROACH TO RECOGNISING AND CLASSIFYING KNOTS ) KNOW AT LEAST ONE EXPERT

IN KNOT THEORY WHO STRONGLY BELIEVES IN THE EXISTENCE OF SUCH AN ALGORITHM
)T IS BY NO MEANS TRIVIAL TO lND ANY ALGORITHM FOR RECOGNISING THE UNKNOT
BUT THIS HAS BEEN DONE BY (AKEN ;(= ) MYSELF HAVE PERFORMED THE FOLLOW
ING SCIENTIlC EXPERIMENT ) TAKE A LONG PIECE OF STRING TIE THE TWO ENDS
TOGETHER WITH A SMALL KNOT AND THEN ENTANGLE THE RESULTING LOOP TRYING TO
MAKE IT HARD TO DISENTANGLE BUT WITHOUT PULLING ANYTHING TIGHT ) HAVE
NEVER MANAGED TO CAUSE MYSELF THE SLIGHTEST DInCULTY
4HE +NASTER HYPOTHESIS 4HIS IS THE FOLLOWING QUESTION ,ET
F 3N 2 BE CONTINUOUS AND LET X XN BE ANY N POINTS IN 3N
-UST THERE EXIST AN ORTHOGONAL MAP j SUCH THAT THE VALUES F jXI ARE ALL
EQUAL 7HEN N A POSITIVE ANSWER FOLLOWS TRIVIALLY FROM THE INTERMEDIATE
VALUE THEOREM ) THINK THE CONJECTURE IS KNOWN TO BE TRUE WHEN N BUT
OPEN FOR N )T IS ALSO KNOWN FOR CERTAIN VERY SPECIAL SETS OF POINTS
X XN
/NE MOTIVATION FOR THE +NASTER HYPOTHESIS ASIDE FROM THE INTRINSIC
APPEAL OF THE QUESTION IS THAT AS OBSERVED BY -ILMAN ;-I= A POSITIVE
ANSWER WOULD PROVIDE A COMPLETELY NEW PROOF OF ONE OF THE CENTRAL RESULTS
OF "ANACH SPACE THEORY [ $VORETZKYS THEOREM ;$= 4HIS STATES THAT FOR
ANY K AND n THERE EXISTS N SUCH THAT EVERY N DIMENSIONAL NORMED
SPACE HAS A K DIMENSIONAL SUBSPACE WHOSE "ANACH -AZUR DISTANCE FROM A
(ILBERT SPACE IS AT MOST n 'EOMETRICALLY IT SAYS THAT AN N DIMENSIONAL
CENTRALLY SYMMETRIC CONVEX BODY HAS AN ALMOST ELLIPSOIDAL K DIMENSIONAL
CROSS SECTION $VORETZKYS THEOREM FOLLOWS FROM THE +NASTER HYPOTHESIS AS
FOLLOWS #ONSIDER 3N_ AS THE UNIT SPHERE OF @N LET X XN BE A m NET OF
THE UNIT SPHERE OF A K DIMENSIONAL SUBSPACE 6 AND LET F BE THE RESTRICTION
OF AN ARBITRARY NORM KK ON 2N TO 3N_ "Y THE +NASTER HYPOTHESIS THERE IS
A ROTATION j SUCH THAT ALL THE VALUES KjXI K ARE EQUAL 3TANDARD ARGUMENTS
THEN SHOW THAT ALL POINTS X IN THE SUBSPACE j6 HAVE ROUGHLY THE SAME
RATIO KXK KXK 4HIS ARGUMENT WHEN MADE PRECISE GIVES THE BEST POSSIBLE
ESTIMATE FOR THE ASYMPTOTIC DEPENDENCE OF N ON n FOR lXED K WHICH IS STILL
AN OPEN PROBLEM
#ERTAIN NATURAL APPROACHES TO THE +NASTER HYPOTHESIS TURN OUT TO FAIL
FOR GOOD REASONS AND MANY PEOPLE HAVE THE FEELING THAT IT MAY WELL BE
P
FALSE (OWEVER EVEN A CONSIDERABLE WEAKENING SUCH AS PROVING IT FOR N
POINTS IN 3N WOULD BE USEFUL FOR $VORETZKYS THEOREM
4HE DENSITY (ALES *EWETT THEOREM 4HE (ALES *EWETT THEOREM
CONCERNS COLOURINGS OF THE SET F KG. )F X BELONGS TO THIS SET ! IS A
6ISIONS IN -ATH 2/5'( 3425#452% !.$ #,!33)&)#!4)/.
SUBSET OF F . G AND J K THEN LET US WRITE X g J! FOR THE POINT

Y OBTAINED FROM X BY CHANGING XI TO J IF I ! AND LEAVING IT UNALTERED
OTHERWISE ! COMBINATORIAL LINE IS A SET OF THE FORM FX g J! J KG
&OR EXAMPLE WHEN K AND . THEN THE POINT X AND
THE SET F G GIVE RISE TO THE COMBINATORIAL LINE
F G
.OTICE THAT IF THE SET F KG. IS IMAGINED AS AN . DIMENSIONAL GRID
THEN COMBINATORIAL LINES ARE SPECIAL CASES OF GEOMETRICAL LINES THAT IS SETS
OF K COLLINEAR POINTS
4HE (ALES *EWETT THEOREM STATES THAT FOR ANY K AND R THERE EXISTS .
SUCH THAT IF THE SET F KG. IS COLOURED WITH R COLOURS THEN THERE
MUST BE A COMBINATORIAL LINE CONSISTING OF POINTS OF THE SAME COLOUR 4HIS
THEOREM IS EASILY SEEN TO BE A STRENGTHENING OF VAN DER 7AERDENS THEOREM
AND LIKE VAN DER 7AERDENS THEOREM IT HAS A DENSITY VERSION FOR EVERY K
AND EVERY m THERE EXISTS . SUCH THAT EVERY SUBSET ! y F KG.
WITH AT LEAST mK . ELEMENTS CONTAINS A COMBINATORIAL LINE
4HIS DENSITY THEOREM IS DUE TO &URSTENBERG AND +ATZNELSON ;&U+= AND
MAKES HEAVY USE OF ERGODIC THEORY -OREOVER NO PROOF IS KNOWN THAT DOES
NOT USE THESE TECHNIQUES )T WOULD BE VERY INTERESTING TO lND A NON ERGODIC
PROOF EVEN IN THE CASE K &ROM THE THEOREM IN THIS CASE IT IS A SIMPLE
EXERCISE TO DEDUCE THE FOLLOWING RESULT &OR EVERY m THERE EXISTS . SUCH
THAT EVERY SUBSET OF THE TWO DIMENSIONAL GRID F . G OF SIZE AT LEAST
m. CONTAINS A TRIPLE OF THE FORM FX Y X Y D X D Y G THAT IS AN
ALIGNED ISOSCELES RIGHTANGLED TRIANGLE %VEN THIS STATEMENT LACKS ANY PROOF
THAT GIVES A REASONABLE UPPER BOUND FOR . IN TERMS OF m
$ANZERS PROBLEM )S IT POSSIBLE TO lND A SET 8 OF POINTS IN 2
SUCH THAT EVERY CONVEX BODY OF AREA CONTAINS AT LEAST ONE POINT OF 8 AND
NOT MORE THAN # POINTS WHERE # IS AN ABSOLUTE CONSTANT )F + IS A CONVEX
BODY OF AREA THEN THERE ARE HOMOTHETIC RECTANGLES 2 AND 2 SUCH THAT
2 y + y 2 AND SUCH THAT THE AREA OF 2 IS AT MOST FOUR TIMES THAT OF
2 )T FOLLOWS THAT IT IS ENOUGH TO CONSIDER RECTANGLES IN THE STATEMENT OF
THE PROBLEM
4HERE IS A SMALL CONNECTION BETWEEN THIS PROBLEM AND SOME OF THE QUES
TIONS IN THE PREVIOUS SECTION BECAUSE THE OBVIOUS RANDOM AND EXPLICIT
METHODS FOR CONSTRUCTING SUCH A SET 8 BOTH FAIL BUT FOR VERY DIkERENT
REASONS
$ANZERS ACTUAL QUESTION IS WEAKER THAN THE ABOVE (E ASKS FOR A SET 8
WHICH INTERSECTS EVERY CONVEX BODY OF AREA AND ALSO SATISlES THE INEQUALITY
74 '/7%23 '!&!
J8 < 2$J #2 FOR EVERY 2 WHERE $ IS THE UNIT DISC ABOUT THE ORIGIN
! BEAUTIFUL PAPER OF "AMBAH AND 7OODS ;"7= SHOWS THAT 8 CANNOT BE A
lNITE UNION OF TRANSLATES OF LATTICES THUS RULING OUT WHAT AT lRST APPEARS
TO BE A PROMISING SOURCE OF POTENTIAL EXAMPLES
"ILLIARDS IN POLYGONS 4HERE IS A FAMOUS EXAMPLE KNOWN AS
+AFKAS STUDY WHICH SHOWS THAT ONE CAN DESIGN A CONNECTED BUT NON
CONVEX ROOM SUCH THAT EVEN IF THE WALLS ARE ENTIRELY LINED WITH MIRRORS
NO SINGLE LIGHT SOURCE CAN ILLUMINATE THE WHOLE OF THE ROOM 4HE EXAMPLE
EXPLOITS THE FACT THAT A BEAM OF LIGHT WHICH PASSES BETWEEN THE TWO FOCI
OF AN ELLIPSE WILL RETURN BETWEEN THE TWO FOCI AFTER IT HAS BEEN REmECTED Ok
THE BOUNDARY OF THE ELLIPSE )T IS NOT KNOWN WHETHER THERE IS A POLYGONAL
EXAMPLE
! CLOSELY RELATED QUESTION IS THE FOLLOWING 3UPPOSE AN INlNITESIMALLY
SMALL BILLIARD BALL IS SET MOVING ON A FRICTIONLESS AND POCKETLESS POLYGONAL
TABLE 7ILL IT RETURN ARBITRARILY CLOSE TO ITS STARTING POINT AND DIRECTION
4HE ANSWER IS KNOWN TO BE POSITIVE IF THE ANGLES BETWEEN ALL THE SIDES OF
THE POLYGON ARE RATIONAL MULTIPLES OF x AS THEN THERE ARE ONLY lNITELY MANY
DIRECTIONS TO CONSIDER AND THE PROBLEM CAN EASILY BE REDUCED TO A KNOWN
AND EASY RESULT ABOUT SO CALLED INTERVAL EXCHANGE MAPS
,ET ; BE PARTITIONED INTO M HALF OPEN INTERVALS IN TWO DIkERENT WAYS
) ; ` ` ` ; )M AND * ; ` ` ` ; *M IN SUCH A WAY THAT )R AND *R HAVE THE SAME
LENGTH FOR EVERY R M ,ET }R BE THE TRANSLATION OF )R TO *R ,ET } BE
THE MAP WHICH TAKES X )R TO }R X !N INTERVAL EXCHANGE MAP FROM ;
TO ; IS ANY MAP OF THIS KIND )T IS QUITE STRAIGHTFORWARD TO PROVE THAT
SUCH MAPS ARE RECURRENT IN THE SENSE THAT FOR EVERY X AND EVERY n THERE
EXISTS . SUCH THAT THE DISTANCE FROM }. X TO X IS AT MOST n
) HAVE ASKED AROUND AND FOUND NOBODY WHO CAN TELL ME WHETHER THE
OBVIOUS TWO DIMENSIONAL GENERALIZATION OF THE ABOVE RESULT IS TRUE OR FALSE
4HAT IS SUPPOSE ONE CHOPS A SQUARE INTO RECTANGLES AND REARRANGES THEM
BY TRANSLATION SO THAT THEY FORM THE ORIGINAL SQUARE AGAIN )S THE RESULTING
RECTANGLE EXCHANGE MAP RECURRENT ) BELIEVE THAT IT IS BUT SOME EXPERTS
IN DYNAMICAL SYSTEMS TO WHOM ) HAVE ASKED THE PROBLEM HAVE THE OPPOSITE
INSTINCT
2EFERENCES
;!-= . !LON 6$ -ILMAN %IGENVALUES EXPANDERS AND SUPERCONCENTRATORS
0ROC TH !NNUAL &/#3 3INGER )SLAND &, )%%% .EW 9ORK [

6ISIONS IN -ATH 2/5'( 3425#452% !.$ #,!33)&)#!4)/.
;"7= 20 "AMBAH !# 7OODS /N A PROBLEM OF $ANZER 0ACIlC * -ATH

[
;"= &! "EHREND /N SETS OF INTEGERS WHICH CONTAIN NO THREE IN ARITHMETIC
PROGRESSION 0ROC .AT !CAD 3CI [
;"I= 9 "ILU 3TRUCTURE OF SETS WITH SMALL SUMSET !STsERISQUE [

;"O= * "OURGAIN 2EMARKS ON -ONTGOMERYS CONJECTURES ON $IRICHLET SERIES
'EOMETRIC !SPECTS OF &UNCTIONAL !NALYSIS 3PRINGER ,ECTURE
.OTES IN -ATHEMATICS [
;"O= * "OURGAIN "ESICOVITCH TYPE MAXIMAL OPERATORS AND APPLICATIONS TO
&OURIER ANALYSIS '!&! [
;"O= * "OURGAIN /N TRIPLES IN ARITHMETIC PROGRESSION '!&!
[
;"O= * "OURGAIN /N THE DIMENSION OF +AKEYA SETS AND RELATED MAXIMAL IN
EQUALITIES '!&! [
;"U= 'EORGE "USH )NTERVIEW 4IME TH *AN
;#'= &2+ #HUNG 2, 'RAHAM 1UASI RANDOM SUBSETS OF :N * #OMB
4H ! [
;#'7= &2+ #HUNG 2, 'RAHAM 2- 7ILSON 1UASI RANDOM GRAPHS
#OMBINATORICA [
;#O= 3! #OOK 4HE COMPLEXITY OF THEOREM PROVING PROCEDURES 0ROC RD
!NNUAL !#- 3YMPOSIUM ON THE 4HEORY OF #OMPUTING [
;$= ! $VORETZKY 3OME RESULTS ON CONVEX BODIES AND "ANACH SPACES 0ROC
3YMP ON ,INEAR 3PACES *ERUSALEM [
;%= 0 %RDO ] S 3OME REMARKS ON THE THEORY OF GRAPHS "ULL !MER -ATH 3OC
[
;%4= 0 %RDO ] S 0 4URA s N /N SOME SEQUENCES OF INTEGERS * ,ONDON -ATH
3OC [
;&= # &EFFERMAN 4HE MULTIPLIER PROBLEM FOR THE BALL !NNALS OF -ATH
[
;&U+= ( &URSTENBERG 9 +ATZNELSON ! DENSITY VERSION OF THE (ALES *EWETT
THEOREM * $!NALYSE -ATH [
;&R= '! &REIMAN &OUNDATIONS OF A 3TRUCTURAL 4HEORY OF 3ET !DDITION IN
2USSIAN +AZAN 'OS 0ED )NST +AZAN
;&R= '! &REIMAN &OUNDATIONS OF A 3TRUCTURAL 4HEORY OF 3ET !DDITION
4RANSLATIONS OF -ATHEMATICAL -ONOGRAPHS !MER -ATH 3OC 0ROV
IDENCE 2) 53!
;'= 74 'OWERS ,OWER BOUNDS OF TOWER TYPE FOR 3ZEMERsEDIS UNIFORMITY
LEMMA '!&! [
;'= 74 'OWERS ! NEW PROOF OF 3ZEMERsEDIS THEOREM FOR ARITHMETIC PRO
GRESSIONS OF LENGTH FOUR '!&! [
;'R22= 2, 'RAHAM 6 2O DL ! 2UCIN
s SKI /N GRAPHS WITH LINEAR 2AMSEY
74 '/7%23 '!&!
NUMBERS PREPRINT
;(= 7 (AKEN 4HEORIE DER .ORMALmACHEN EIN )SOTOPIEKRITERIUM F UR DEN
+REISNOTEN !CTA -ATH [
;(A7= *- (AMMERSLEY $*! 7ELSH &URTHER RESULTS ON THE RATE OF CON
VERGENCE TO THE CONNECTIVE CONSTANT OF THE HYPERCUBICAL LATTICE 1UART *
-ATH /XFORD 3ER [
;(AR3= 4 (ARA ' 3LADE 3ELF AVOIDING WALK IN lVE OR MORE DIMENSIONS
#OMM -ATH 0HYS [
;(AR3= 4 (ARA ' 3LADE 4HE LACE EXPANSION FOR SELF AVOIDING WALK IN lVE OR
MORE DIMENSIONS 2EV -ATH 0HYS [
;(E= $2 (EATH "ROWN )NTEGER SETS CONTAINING NO ARITHMETIC PROGRESSIONS
* ,ONDON -ATH 3OC [
;+= 2- +ARP 2EDUCIBILITY AMONG COMBINATORIAL PROBLEMS #OMPLEXITY OF
#OMPUTER #OMPUTATIONS 0ROC 3YMPOS )"- 4HOMAS * 7ATSON 2ES
#ENTR 9ORKTOWN (EIGHTS .9 2% -ILLER *7 4HATCHER EDS
0LENUM 0RESS .EW 9ORK [
;+A,4= .( +ATZ ) ,ABA 4 4AO !N IMPROVED BOUND ON THE -INKOWSKI
DIMENSION OF "ESICOVITCH SETS IN 2 !NNALS OF -ATH TO APPEAR
;+A4= .( +ATZ 4 4AO ! NEW BOUND ON PARTIAL SUM SETS AND DIkERENCE SETS
AND APPLICATIONS TO THE +AKEYA CONJECTURE SUBMITTED
;+O3= * +OMLO s S - 3IMONOVITS 3ZEMERsEDIS 2EGULARITY ,EMMA AND ITS AP
PLICATIONS IN 'RAPH 4HEORY IN <#OMBINATORICS 0AUL %RD] OS IS 6OL
"OLYAI 3OCIETY -ATH 3TUDIES [ +ESTHELY (UNGARY "U
DAPEST
;,03= ! ,UBOTZKY 2 0HILLIPS 0 3ARNAK %XPLICIT EXPANDERS AND THE
2AMANUJAN CONJECTURES 0ROCEEDINGS OF THE TH !#- 3YMPOSIUM ON THE
4HEORY OF #OMPUTING [ ALSO #OMBINATORICA [

;-3= . -ADRAS ' 3LADE 4HE 3ELF !VOIDING 2ANDOM 7ALK "IRKHAUSER
"OSTON
;-I= 6$ -ILMAN ! FEW OBSERVATIONS ON THE CONNECTIONS BETWEEN LOCAL THE
ORY AND SOME OTHER lELDS IN 'EOMETRIC !SPECTS OF &UNCTIONAL !NALYSIS
)SRAEL SEMINAR '!&! * ,INDENSTRAUSS 6$ -ILMAN EDS
3PRINGER ,.- [
;-O= (, -ONTGOMERY 4EN ,ECTURES ON THE )NTERFACE "ETWEEN !NALYTIC
.UMBER 4HEORY AND (ARMONIC !NALYSIS #"-3 2EGIONAL #ONFERENCE 3E
RIES IN -ATH !-3
;0= ' 0OLYA -ATHEMATICS AND 0LAUSIBLE 2EASONING 6OLS ) AND )) 0RINCETON
5NIVERSITY 0RESS
;22= !! 2AZBOROV 3 2UDICH .ATURAL PROOFS IN TH !NNUAL !#- 3YM
POSIUM ON THE 4HEORY OF #OMPUTING 34/# -ONTREAL 01
ALSO * #OMPUT 3YSTEM 3CI [
6ISIONS IN -ATH 2/5'( 3425#452% !.$ #,!33)&)#!4)/.
;2O= + 2OTH /N CERTAIN SETS OF INTEGERS * ,ONDON -ATH 3OC

[
;2O= + 2OTH 2ATIONAL APPROXIMATIONS TO ALGEBRAIC NUMBERS -ATHEMATIKA
[ WITH CORRIGENDUM P
;2U= ): 2UZSA 'ENERALIZED ARITHMETIC PROGRESSIONS AND SUMSETS !CTA -ATH
(UNGAR [
;3= % 3ZEMERs EDI 2EGULAR PARTITIONS OF GRAPHS #OLLOQUES )NTERNATIONAUX
#.23 [ 0ROBLrEMES #OMBINATOIRES ET 4HsEORIE DES 'RAPHES /RSAY
[
;3= % 3ZEMERs EDI /N SETS OF INTEGERS CONTAINING NO K ELEMENTS IN ARITHMETIC
PROGRESSION !CTA !RITH [
;3= % 3ZEMERs EDI )NTEGER SETS CONTAINING NO ARITHMETIC PROGRESSIONS !CTA
-ATH (UNGAR [
;4= !' 4HOMASON 0SEUDO RANDOM GRAPHS 0ROCEEDINGS OF 2ANDOM
'RAPHS 0OZNs AN - +ARONSKI ED !NNALS OF $ISCRETE -ATHEMATICS
[
;4= !' 4HOMASON ! DISPROOF OF A CONJECTURE OF %RD] OS IN 2AMSEY THEORY
* ,ONDON -ATH 3OC [
;4= !' 4HOMASON 'RAPH PRODUCTS AND MONOCHROMATIC MULTIPLICITIES
#OMBINATORICA [
;7= 4( 7OLFF !N IMPROVED BOUND FOR +AKEYA TYPE MAXIMAL FUNCTIONS
2EVISTA -AT )BEROAMERICANA [
74 'OWERS #ENTRE FOR -ATHEMATICAL 3CIENCES 7ILBERFORCE 2OAD #AMBRIDGE

#" 7! 5+
30!#%3 !.$ 15%34)/.3
-ISHA 'ROMOV
$AWN OF 3PACE
/UR %UCLIDEAN INTUITION PROBABLY INHERITED FROM ANCIENT PRIMATES MIGHT
HAVE GROWN OUT OF THE lRST SEEDS OF GEOMETRY IN THE MOTOR CONTROL SYS
TEMS OF EARLY ANIMALS WHO WERE BROUGHT UP TO SEA AND THEN TO LAND BY THE
#AMBRIAN EXPLOSION HALF A BILLION YEARS AGO 4HE PRIMATES BRAIN HAD BEEN
IDLING FOR MILLION YEARS 3UDDENLY IN A mASH OF ONE MILLION YEARS IT
EXPLODED INTO GROWTH UNDER THE RELENTLESS PRESSURE OF SEXUAL SOCIAL COMPE
TITION AND SPROUTED A MASSIVE NEOCORTEX NEURONS IN HUMANS WITH AN
INEXPLICABLE CAPABILITY FOR LANGUAGE SEQUENTIAL REASONING AND GENERATION OF
MATHEMATICAL IDEAS 4HEN -AN CAME AND LAID DOWN SPACE ON PAPYRUS IN A
STRING OF AXIOMS LEMMAS AND THEOREMS AROUND "# IN !LEXANDRIA
0ROJECTED TO WORDS THE BRAINS MODEL OF SPACE BEGAN TO EVOLVE BY DROP
PING MODIFYING AND GENERALIZING ITS AXIOMS 4HE 0ARALLEL 0OSTULATE FELL
lRST 'AUSS 3CHWEIKART ,OBACHEVSKI "OLYAI WHO ELSE CAME TO THE
CONCLUSION THAT THERE IS A UNIQUE NON TRIVIAL ONE PARAMETER DEFORMATION OF
THE METRIC ON 2 KEEPING THE SPACE FULLY HOMOGENEOUS
)T IS BELIEVED THAT 'AUSS WHO CONVINCED HIMSELF OF THE VALIDITY OF HY
PERBOLIC GEOMETRY SOMEWHERE BETWEEN AND WAS DISCONCERTED BY
THE ABSENCE OF A %UCLIDEAN REALIZATION OF THE HYPERBOLIC PLANE ( "Y THAT
TIME HE MUST HAVE HAD A CLEAR PICTURE OF THE GEOMETRY OF SURFACES IN 2
EXPOSED IN HIS <$ISQUISITONES CIRCA SUPERlCIES CURVAS IN WHERE THE
INTRINSIC DISTANCE BETWEEN TWO POINTS ON A SURFACE IS DElNED AS THE LENGTH
OF THE SHORTEST BETTER TO SAY <INlMAL CURVE IN THE SURFACE BETWEEN THESE
POINTS 4HIS IDEA MUST HAVE BEEN IMPRINTED BY .ATURE IN THE BRAIN AS
MOST ANIMALS ROUTINELY CHOOSE SHORTEST CUTS ON RUGGED TERRAINS 'AUSS DIS
COVERED THE FOLLOWING POWERFUL AND EnCIENT CRITERION FOR ISOMETRY BETWEEN
SURFACES DISTINGUISHING FOR EXAMPLE A PIECE OF A ROUND SPHERE 3 y 2

)NCIDENTALLY THE lRST MATHEMATICS TEACHER OF 'AUSS x *OHANN -ARTIN "ARTELS
LATER ON BECAME THE TEACHER OF ,OBACHEVSKI x IN +AZAN

! METRIC SPACE 8 IS FULLY HOMOGENEOUS IF EVERY PARTIAL ISOMETRY 8 z a a y 8
EXTENDS TO A FULL ISOMETRY OF 8 AS FOR %UCLIDS TRIANGLES WITH EQUAL SIDES IN 2
6ISIONS IN -ATH 30!#%3 !.$ 15%34)/.3
FROM AN ARBITRARILY BENT SHEET OF PAPER RETAINING ITS INTRINSIC %UCLIDEANNESS

UNDER BENDINGS
-AP A SURFACE 3 y 2 TO THE UNIT SPHERE 3 BY TAKING THE VECTORS
vS 3 PARALLEL TO THE UNIT NORMAL VECTORS vS S 3 )F 3 IS # SMOOTH
S O(s')
s'
O(s') O(s)
S2 s
O(s)
&IGURE
THE 'AUSS MAP ' 3 3 FOR S vS IS # AND ITS *ACOBIAN IE THE
INlNITESIMAL AREA DISTORTION COMES WITH A NON AMBIGUOUS SIGN SINCE THE
DIRECTIONS OF vS GIVE COHERENT ORIENTATIONS TO 3 AND 3 AND SO 3 APPEARS
WITH A REAL FUNCTION CALLED 'AUSS CURVATURE +S *AC 'S
DEF
4HEOREMA %GREGIUM %VERY ISOMETRY BETWEEN SURFACES SAY F 3 3

PRESERVES 'AUSS CURVATURE +F S +S FOR ALL S 3
&OR EXAMPLE THE PLANE HAS + p AS THE 'AUSS MAP IS CONSTANT AND
SO IT IS NOT EVEN LOCALLY ISOMETRIC TO THE UNIT SPHERE WHERE + p FOR THE
'AUSS MAP IS IDENTITY ON 3 -ORE GENERALLY NO STRICTLY CONVEX SURFACE
IS LOCALLY ISOMETRIC TO A SADDLE SURFACE SUCH AS THE GRAPH OF THE FUNCTION
Z XY FOR INSTANCE SINCE STRICT CONVEXITY MAKES + WHILE SADDLE POINTS
HAVE + s
'AUSS WAS WELL AWARE OF THE FACT THAT THE HYPERBOLIC PLANE ( WOULD
HAVE CONSTANT NEGATIVE CURVATURE IF IT WERE REALIZED BY A SURFACE IN 2
"UT HE COULD NOT lND SUCH A SURFACE )N FACT THERE ARE RELATIVELY SMALL
PIECES OF SURFACES WITH + _ IN 2 INVESTIGATED BY "ELTRAMI IN
AND IT IS HARD TO BELIEVE 'AUSS MISSED THEM BUT HE DElNITELY COULD NOT
REALIZE THE WHOLE ( BY A # SURFACE IN 2 AS IS PRECLUDED BY A THEOREM
OF (ILBERT 4HIS COULD BE WHY BESIDES HIS TIMIDITY IN THE FACE OF THE
-)3(! '2/-/6 '!&!
+ANTIAN GUARDS OF 4RILOBITES INTUITION 'AUSS REFRAINED FROM PUBLISHING

HIS DISCOVERY
0ROBABLY 'AUSS WOULD HAVE BEEN DELIGHTED TO LEARN MAYBE HE KNEW
IT THAT THE mAT ,ORENZ -INKOWSKI <METRIC DX DY _ DZ ON 2 2
INDUCES A TRUE POSITIVE METRIC ON THE SPHERE

3_ FX Y Z J X Y _ Z _G
WHERE EACH OF THE TWO COMPONENTS OF 3_ ONE IS WHERE Z AND THE OTHER

WITH Z IS ISOMETRIC TO ( AND WHERE THE ORTHOGONAL GROUP / IE
THE LINEAR GROUP PRESERVING THE QUADRATIC FORM X Y _ Z ACTS ON THESE
TWO ( S BY HYPERBOLIC ISOMETRICS
4HERE IS NO COMPARABLE EMBEDDING OF ( INTO ANY 2. THOUGH (
ADMITS A RATHER CONTORTED ISOMETRIC # IMMERSION TO 2 TO 2 AND
INCREDIBLY AN ISOMETRIC # EMBEDDING INTO 2 BUT IT ADMITS AN EMBEDDING
INTO THE (ILBERT SPACE SAY F ( 2 WHERE THE INDUCED INTRINSIC METRIC
IS THE HYPERBOLIC ONE WHERE ALL ISOMETRIES OF ( UNIQUELY EXTEND TO THOSE
OF 2 AND SUCH THAT
P
DIST2 F X F Y DIST( X Y mDISTX Y
WITH BOUNDED FUNCTION mD WHERE ONE CAN lND F WITH mD BUT THIS
WILL BE NOT ISOMETRIC IN OUR SENSE AS IT BLOWS THE LENGTHS OF ALL CURVES IN (
TO INlNITY 3IMILAR EMBEDDINGS EXIST FOR METRIC TREES AS WELL AS FOR REAL AND
COMPLEX HYPERBOLIC SPACES OF ALL DIMENSIONS BUT NOT FOR OTHER IRREDUCIBLE
SYMMETRIC SPACES OF NON COMPACT TYPE 4HIS IS EASY FOR TREES ARRANGE A
GIVEN TREE IN 2 SUCH THAT ITS EDGES BECOME ALL MUTUALLY ORTHOGONAL AND
HAVE PRESCRIBED LENGTHS
3UMMARY 3URFACES IN 2 PROVIDE US WITH A LARGE EASILY ACCESSIBLE POOL
OF METRIC SPACES TAKE A DOMAIN IN 2 SMOOTHLY MAP IT INTO 2 AND VOILr A
YOU HAVE THE INDUCED 2IEMANNIAN METRIC IN YOUR LAP 4HEN STUDY THE
ISOMETRY PROBLEM FOR SURFACES BY LOOKING AT METRIC INVARIANTS CURVATURE IN
THE ABOVE DISCUSSION RELATE THEM TO STANDARD SPACES 2. 2 2 AND
CONSIDER INTERESTING TO WHOM CLASSES OF SURFACES EG THOSE WITH +
AND WITH +
2EMARKS AND REFERENCES A )T SEEMS TO ME THAT THE REVERENCE FOR
HUMAN INTUITION AND INTROSPECTIVE SOUL SEARCHING STAND IN THE WAY OF ANY
ATTEMPT TO UNDERSTAND HOW THE BRAIN DOES MATHEMATICS (OPEFULLY THE

/F ALL PEOPLE HAD HE BEEN ALIVE +ANT HIMSELF COULD HAVE BEEN ABLE TO ASSIMILATE IF
NOT ACCEPT THE NON %UCLIDEAN IDEA
6ISIONS IN -ATH 30!#%3 !.$ 15%34)/.3
EXPERIENCE OF NATURAL SCIENTISTS MAY LEAD US TO A MEANINGFUL MODEL A PRO

VISIONAL ONE AT THIS STAGE SAY IN THE SPIRIT OF +ANERVAS IDEA OF DISTRIBUTED
MEMORY SEE ;+=
B /UR ALLUSIONS TO THE HISTORY OF MATHEMATICS ARE BORROWED FROM ;+L=
;.= AND ;6=
C ,ITTLE IS KNOWN OF WHAT KIND OF MAPS 3 3 CAN SERVE AS 'AUSS
MAPS ' OF COMPLETE SURFACES IN 2 &OR EXAMPLE GIVEN A DOMAIN 5 y 3
ONE MAY ASK WHETHER THERE EXISTS AN ORIENTED CLOSED IMMERSED IE WITH
POSSIBLE SELF INTERSECTIONS SURFACE 3 y 2 WITH '3 y 5 WHERE 5
3 FORCES 3 TO BE TOPOLOGICALLY THE TORUS 4HIS NOW APPEARS TO ME A
TYPICAL MISGUIDED <NATURAL QUESTION YET ) HAVE NOT LOST HOPE IT MAY HAVE
A REVEALING SOLUTION COMPARE IN ;'=
D )T IS UNKNOWN IF EVERY SURFACE WITH A # SMOOTH 2IEMANNIAN METRIC
CAN BE ISOMETRICALLY # IMMERSED INTO 2 !NOTHER <NATURAL QUESTION
"UT ISOMETRIC IMMERSIONS INTO HIGH DIMENSIONAL SPACES ARE PRETTY WELL UN
DERSTOOD SEE [ AND 0ART IN ;'=
E 4HE ABOVE EQUIVARIANT EMBEDDING ( 2 TELLS US THAT THE ISOM
ETRY GROUP )SO( 0 3, 2 IS A 4 MENABLE OPPOSITE TO +AZHDANS PROP
ERTY 4 DElNED IN ! OF X 6 SATISlED BY THE MAJORITY OF GROUPS ! 4
MENABILITY GENERALIZES AMENABILITY 4HIS PROPERTY WAS lRST RECOGNIZED BY
(AAGERUP ) PRESUME WHO USED DIkERENT TERMINOLOGY
3PIRIT OF 2IEMANN
4HE TRIANGLE INEQUALITY IS NOT ALWAYS EASY TO VERIFY FOR A GIVEN FUNCTION
IN TWO VARIABLES D 8 a 8 2 AS IT IS A NON LOCAL PROPERTY OF D ON
8 THUS ONE CANNOT CREATE METRICS AT WILL 9ET AN ARBITRARY METRIC D ON
A CONNECTED SPACE CAN BE MADE <LOCAL BY REPLACING IT BY THE SUPREMAL
METRIC D AGREEING WITH D ON AN <ARBITRARY lNE COVERING OF 8 AS WE PASS
FOR EXAMPLE FROM THE RESTRICTION OF THE %UCLIDEAN METRIC ON A SUBMANIFOLD
EG A SURFACE IN 2. TO THE INDUCED 2IEMANNIAN OR INTRINSIC METRIC
-ORE GENERALLY FOLLOWING 2IEMANN (ABILITATIONSSCHRIFT *UNE ONE
STARTS WITH AN ARBITRARY lELD G OF %UCLIDEAN METRICS ON A DOMAIN 5 y 2N
IE A CONTINUOUS MAP U GU FROM 5 TO THE NN DIMENSIONAL SPACE ' OF
POSITIVE DElNITE QUADRATIC FORMS ON 2N /NE MEASURES DISTANCES IN SMALL
NEIGHBOURHOODS 5 U y 5 U 5 BY SETTING DU U U KU _ U KU
GU U _ U U _ U FOR U U 5 AND THEN DElNES THE 2IEMANNIAN
GEODESIC DISTANCE DISTG ON 5 AS THE SUPREMAL METRIC FOR THESE DU AS
-)3(! '2/-/6 '!&!
&INALLY 2IEMANNIAN MANIFOLDS 6 APPEAR AS METRIC SPACES LOCALLY

ISOMETRIC TO THE ABOVE 5 S 4HE LATTER STEP IS SLICK BUT HARD TO IMPLEMENT
4RY FOR EXAMPLE TO SHOW THAT THE INDUCED INTRINSIC METRIC ON A SMOOTH
SUBMANIFOLD 6 y 2. IS 2IEMANNIAN IN OUR SENSE
4HE MAGIC POWER OF THIS DElNITION IS DUE TO THE INlNITESIMAL KINSHIP OF
<2IEMANNIAN TO <%UCLIDEAN )F 6 IS SMOOTH IE ALL GS ON 5 S ARE AT LEAST
# THEN LOCALLY NEAR EACH POINT V ONE CAN REPRESENT 6 BY A NEIGHBOURHOOD
5 OF THE ORIGIN 2N WITH V SO THAT THE CORRESPONDING G ON 5 AGREES
0N THE %UCLIDEAN IE CONSTANT IN V METRIC G G X Y HX YI
WITH
I XI YI UP TO THE lRST ORDER
N q r
8 G
GU G UI UJ
UI UJ
IJ
IE WITH THE lRST ORDER 4AYLOR TERMS MISSING AND WHERE MOREOVER ONLY
N N _ _ ` G
TERMS AMONG NN
SECOND DERIVATIVES UIuvUJ DO NOT VANISH

4HE RESULTING N N_ FUNCTIONS ON 5 WHEN PROPERLY ORGANIZED MAKE THE
2IEMANN CURVATURE TENSOR OF 6 WHICH REDUCES TO THE 'AUSS CURVATURE FOR
N MEASURING THE DEVIATION OF 6 G FROM mATNESS IE %UCLIDEANNESS
4HE POLYLINEAR ALGEBRAIC STRUCTURE BUILT INTO G ALLOWS A FULLY mEDGED
ANALYSIS ON 6 G SUCH AS THE ,APLACE (ODGE OPERATOR POTENTIAL THEORY ETC
4HIS TURNED OUT TO BE USEFUL FOR PARTICULAR CLASSES OF MANIFOLDS DISTINGUISHED
BY ADDITIONAL GLOBAL LOCAL OR INlNITESIMAL SYMMETRY WHERE THE MAJOR
ACHIEVEMENTS COMING TO ONES MIND ARE
(ODGE DECOMPOSITION ON THE COHOMOLOGY OF + AHLER MANIFOLDS 6 AND
A SIMILAR NON LINEAR STRUCTURE ON THE SPACES OF REPRESENTATIONS OF x 6
%XISTENCE OF %INSTEIN METRICS ON +AHLER MANIFOLDS WITH ALGEBRA
GEOMETRIC CONSEQUENCES
3PECTRAL ANALYSIS ON LOCALLY SYMMETRIC "RUHAT 4ITS AND ADELIC AS WELL
AS 2IEMANNIAN SPACES LEADING FOR INSTANCE TO VARIOUS COHOMOLOGY VAN
ISHING THEOREMS 4 PROPERTY WITH APPLICATIONS TO EXPANDERS AND AFTER
DELINEARIZATION TO SUPER RIGIDITY OF LATTICES IN SEMI SIMPLE ,IE GROUPS
4HE LINEAR ANALYSIS ON GENERAL 2IEMANNIAN MANIFOLDS PIVOTS AROUND THE
!TIYAH 3INGER $IRAC OPERATOR AND THE INDEX 2IEMANN 2OCH THEOREMS
4HESE ORIGINATED FROM 'ELFANDS QUESTION RAISED IN THE LATE lFTIES AIMING
AT AN EXPLICIT TOPOLOGICAL FORMULA FOR THE INDEX OF AN ELLIPTIC OPERATOR WHICH
IS EASILY SEEN TO BE DEFORMATION INVARIANT AND BECAME A CENTRAL THEME IN
MATHEMATICS STARTING FROM THE PAPER BY !TIYAH AND 3INGER
6ISIONS IN -ATH 30!#%3 !.$ 15%34)/.3
4HE NON LINEAR 2IEMANNIAN ANALYSIS ON GENERAL 6 S FOLLOWED FOR THE

MOST PART THE CLASSICAL TRADITION CONCENTRATING AROUND ELLIPTIC VARIATIONAL
PROBLEMS WITH MAJOR ADVANCES IN THE EXISTENCE AND REGULARITY OF SOLUTIONS
MINIMAL SUBVARIETIES HARMONIC MAPS ETC 4HE MOST VISIBLE <EXTERNAL AP
PLICATION IN MY VIEW CONCERNS MANIFOLDS WITH POSITIVE SCALAR CURVATURE [
THE SUBJECT MOTIVATED BY PROBLEMS AND IDEAS COMING FROM GENERAL RELA
TIVITY [ RESOLVED BY 3CHOEN AND 9AU WITH A USE OF MINIMAL HYPERSURFACES
-ANIFOLDS OF EACH DIMENSION TWO THREE AND FOUR MAKE WORLDS OF THEIR
OWN RICHER IN STRUCTURE THAN ALL WE KNOW SO FAR ABOUT N t
)N DIMENSION TWO WE POSSESS THE #AUCHY 2IEMANN EQUATIONS AND ARE
GUIDED BY THE BEACON OF THE 2IEMANN MAPPING THEOREM THE CROWN JEWEL
OF DIkERENTIAL GEOMETRY
4HE FOUR DIMENSIONAL PECULIARITY STARTS WITH ALGEBRA THE ORTHOGONAL
GROUP / LOCALLY DECOMPOSES INTO TWO / 4HIS ALLOWS ONE TO SPLIT OR
RATHER TO SQUARE ROOT CERTAIN NATURAL FOR THE / SYMMETRY NON LINEAR
SECOND ORDER OPERATORS IN A WAY SIMILAR TO HOW WE EXTRACT THE #AUCHY
2IEMANN FROM THE BORINGLY NATURAL SELFADJOINT ,APLACIAN a ON 2 4HE
RESULTING lRST ORDER OPERATORS MAY HAVE NON ZERO INDICES AND SATISFY A KIND
OF NON LINEAR INDEX THEOREM DISCOVERED BY $ONALDSON IN FOR THE 9ANG
-ILLS AND THEN EXTENDED TO THE 3EIBERG 7ITTEN EQUATION "OTH EQUATIONS
WERE lRST WRITTEN DOWN BY PHYSICISTS ACCORDING TO THE TH CENTURY LORE
-ANIFOLDS OF DIMENSION THREE BORROW FROM THEIR TWO AND FOUR DIMEN
SIONAL NEIGHBOURS 4HURSTONS CONSTRUCTION OF HYPERBOLIC METRICS ON BASIC
MANIFOLDS RELIES ON GEOMETRY OF SURFACES WHILE &LOER HOMOLOGY DESCENDS
FROM 9ANG -ILLS
3HALL WE EVER REACH SPACES BEYOND 2IEMANNS IMAGINATION
2EMARK AND REFERENCES )T WILL NEED HUNDREDS OF PAGES TO ACCOUNT FOR
THE ABOVE FORTY LINES (ERE WE LIMIT OURSELVES TO A FEW POINTS
A 4HE 2IEMANNIAN METRIC G NATURALLY IE FUNCTORIALLY DElNES PAR
ALLEL TRANSPORT OF VECTORS ALONG SMOOTH CURVES IN 6 WHICH IS DUE TO THE
ABSENCE OF lRST DERIVATIVES IN AN APPROPRIATE 4AYLOR EXPANSION OF G 4HIS
CAN BE SEEN CLEARLY FOR 6 REALIZED IN SOME 2. WHICH IS NOT A HINDRANCE AC
CORDING TO THE #ARTAN *ANET "URSTIN .ASH ISOMETRIC EMBEDDING THEOREM
WHERE A FAMILY 8T OF TANGENT VECTORS IS PARALLEL IN 6 ALONG OUR CURVE l
PARAMETRIZED BY T 2 Ik THE ORDINARY %UCLIDEAN DERIVATIVE D8T
DT 2.
IS NORMAL TO 6 AT lT 6 FOR ALL T 4HIS IS INDEPENDENT OF THE ISOMETRIC
-)3(! '2/-/6 '!&!
&IGURE
EMBEDDING 6 2. )F A CURVE l ; = 6 COMES BACK MAKING A LOOP

IE l l EVERY TANGENT VECTOR 8 8 4V 6 V l 6
DEF
TRANSFORMS TO lb 8 8 4V 6 AND WE OBTAIN A HOMOMORPHISM
FROM THE <GROUP OF LOOPS AT V TO THE LINEAR GROUP OF ISOMETRIC AUTOMOR
PHISMS OF THE TANGENT SPACE 4V 6 2. IE TO THE ORTHOGONAL GROUP /N
ITS IMAGE ( y /N IS CALLED THE HOLONOMY GROUP OF 6 WHICH IS INDEPEN
DENT OF V FOR CONNECTED 6 'ENERICALLY ( /N 3/N FOR ORIENTABLE
6 BUT SOMETIMES ( HAS POSITIVE CODIMENSION IN /N &OR EXAMPLE
DIM ( Ik 6 IS LOCALLY %UCLIDEAN 0ARALLEL 0OSTULATE IS EQUIVALENT TO
( FIDG AND IF 6 6 a 6 THEN ( ( a ( y /N a /N y /N

FOR N N 4HEN THERE ARE SEVERAL DISCRETE SERIES OF SYMMETRIC SPACES
[ MONUMENTAL LANDMARKS TOWERING IN THE VASTNESS OF ALL 2IEMANNIAN MET
RICS 2N 3 N ( N #0 N 3,N 3/N )T IS NATURAL TO THINK THAT THESE
ARE ESSENTIALLY ALL 6 S WITH SMALL HOLONOMY SINCE CODIM ( IMPLIES A
RATHER OVER DETERMINED SYSTEM OF 0$% FOR G FOR EXAMPLE DIM (

CURVATURE G IE N N_ EQUATIONS AGAINST MERE NN
COMPONENTS
OF G 9ET mAT METRICS EXISTS "UT LO AND BEHOLD LOTS OF EVEN DIMENSIONAL
MANIFOLDS CARRY + AHLER METRICS WHERE ( y 5 N y 3/N *UST TAKE A

COMPLEX ANALYTIC SUBMANIFOLD 6 IN #. OR IN #0 . AND OBSERVE WHICH IS
OBVIOUS ONCE BEING SAID THAT THE PARALLEL TRANSPORT IN THE INDUCED MET
RIC PRESERVES THE COMPLEX STRUCTURE IN THE TANGENT SPACES 4HIS MAY BE
NOT SO STRIKING PERHAPS FOR THOSE WHO HAVE ABSORBED THE IMPACT OF HOLO
MORPHIC FUNCTIONS OVER DETERMINED BY #AUCHY 2IEMANN IN MANY VARIABLES
BUT THERE ARE LESS EXPECTED BEAUTIFUL EXOTIC HOLONOMY BEASTS PREDICTED BY
"ERGERS CLASSIlCATION AND BROUGHT TO LIFE BY "RYANT SEE ;"R=
4HE +AHLER WORLD IS TIGHTLY KNIT UNLIKE THE FULL 2IEMANNIAN UNIVERSE
6ISIONS IN -ATH 30!#%3 !.$ 15%34)/.3
WITH DEEP FUNCTORIAL LINKS BETWEEN GEOMETRY AND TOPOLOGY &OR EXAMPLE THE
lRST COHOMOLOGY OF A COMPACT + AHLER 6 COMES BY THE WAY OF A HOLOMORPHIC
MAP TO SOME COMPLEX TORUS 6 #D LATTICE FOR D RANK( 6 4HIS
EXTENDS TO NON !BELIAN REPRESENTATIONS x 6 ',N FOR N t 3IU
#ORLETTE 3IMPSON SEE ;!"#+4= FURNISHING SOMETHING LIKE AN <UNRAM
IlED NON !BELIAN +AHLERIAN CLASS lELD THEORY IN THE SPIRIT OF ,ANGLANDS
PROGRAM BUT WE HAVE NO NOT EVEN CONJECTURAL PICTURE OF THE <TRANSCEN
DENTAL PART OF x 6 KILLED BY THE PROlNITE COMPLETION OF x #AN
FOR EXAMPLE x + AHLER HAVE AN UNSOLVABLE WORD PROBLEM )S THERE AN
INTERNAL STRUCTURE IN THE CATEGORY OF +AHLER FUNDAMENTAL GROUPS FUNCTORI
ALLY REmECTING THE GEOMETRY OF THE +AHLER CATEGORY !LL KNOWN COMPACT
+AHLER MANIFOLDS CAN BE DEFORMED TO COMPLEX PROJECTIVE ONES AND IT MAY
BE PREFERABLE TO STAY WITHIN THE COMPLEX ALGEBRAIC CATEGORY WITH NO FEAR
OF RAMIlCATIONS SINGULARITIES AND NON PROJECTIVENESS
B 4HERE EXISTS A UNIQUE UP TO NORMALIZATION SECOND ORDER DIkERENTIAL
OPERATOR 3 FROM THE SPACE OF POSITIVE DElNITE QUADRATIC DIkERENTIAL FORMS
2IEMANN METRICS G ON 6 TO THE SPACE OF FUNCTIONS 6 2 WITH THE
FOLLOWING TWO PROPERTIES
3 IS $Ik EQUIVARIANT FOR THE NATURAL ACTION OF DIkEOMORPHISMS OF 6 ON
BOTH SPACES
3 IS LINEAR IN THE SECOND DERIVATIVES OF G BEING A LINEAR COMBINATION OF
COMPONENTS OF THE FULL CURVATURE TENSOR
4HEN 3G OR 36 IS NAMED THE SCALAR CURVATURE OF 6 G WITH THE
CUSTOMARY NORMALIZATION 3 3 N NN _
)F N 3 COINCIDES WITH 'AUSS CURVATURE IT IS ADDITIVE FOR PRODUCTS
36 a 6 36 36 AND IT SCALES AS G _ IE 3tG t_ 3G FOR
t
4HE FOLLOWING QUESTION PROVED TO BE MORE TO THE POINT THAN ONE COULD
EXPECT
7HAT IS THE GEOMETRIC AND TOPOLOGICAL STRUCTURE OF MANIFOLDS WITH 3
4HIS COMES FROM GENERAL RELATIVITY AS 3 ON WORLD SHEETS REmECTS
POSITIVITY OF ENERGY
4HE CONDITION 3 APPEARS QUITE PLASTIC FOR N t WHERE ONE CAN
RATHER FREELY MANIPULATE METRICS G KEEPING 3G EG PERFORMING GE
OMETRIC SURGERY BESIDES EVERY COMPACT 6 TURNS INTO 6 WITH 36
WHEN MULTIPLIED BY A SMALL ROUND SPHERE OF DIMENSIONS t 9ET THIS PLAS
TICITY HAS ITS LIMITS ,ICHNEROWICZ FOUND IN A RATHER SUBTLE TOPOLOGICAL
-)3(! '2/-/6 '!&!
B IF 6 IS SPIN WITH THE USE OF THE INDEX THEOREM

OBSTRUCTION !6
4HEN IN 3CHOEN AND 9AU APPROACHED THE PROBLEM FROM ANOTHER ANGLE
LINKED TO IDEAS IN GENERAL RELATIVITY AND PROVED AMONG OTHER THINGS THAT
N TORI AT LEAST FOR N s CARRY NO METRICS WITH 3 THUS ANSWERING
A QUESTION BY 'EROCH )NSPIRED BY THIS WE REVIVED WITH "LAINE ,AWSON
IN ,ICHNEROWICZ IDEA COMBINED IT WITH THE ,USZTIG -ISTCHENKO AP
PROACH TO THE .OVIKOV CONJECTURE ON HOMOTOPY INVARIANCE OF 0ONTRYAGIN
CLASSES OF NON SIMPLY CONNECTED MANIFOLDS AND FOUND OUT THAT THE BULK OF
THE TOPOLOGICAL OBSTRUCTIONS FOR 3 COMES FROM A <LIMIT ON GEOMETRIC
SIZE OF 6 INDUCED BY THE INEQUALITY 3 SIMILAR TO BUT MORE DELICATE
THAN THAT FOR + SEE )))
9ET THE ABOVE QUESTION REMAINS OPEN WITH AN EXTRA MYSTERY TO SETTLE
WHAT DO MINIMAL HYPERSURFACES AND THE $IRAC OPERATOR HAVE IN COMMON
3EEMINGLY NOTHING AT ALL BUT THEY LEAD TO ALMOST IDENTICAL STRUCTURE RESULTS
FOR 3 SEE ;'= FOR AN INTRODUCTION TO THESE ISSUES
)T APPEARS THAT AN ESSENTIAL PART OF THE DInCULTY IN UNDERSTANDING 3
AND THE .OVIKOV CONJECTURE IS LINKED TO THE FOLLOWING SIMPLE MINDED
QUESTION WHAT IS THE MINIMAL t SUCH THAT THE UNIT SPHERE 3 . IN
. .
THE "ANACH SPACE @ 2 KXK SUPI JXI J ADMITS A t ,IPSCHITZ MAP INTO
THE ORDINARY N SPHERE 3 . IN 2. @. WITH NON ZERO DEGREE 0ROBABLY
.
t FOR . EVEN IF WE STABILIZE TO MAPS 3 a 3 - 2
3 . a- WITH ARBITRARILY LARGE - AND 2 AND THIS MIGHT INDICATE NEW
WAYS OF MEASURING <SIZE OF 6 IN THE CONTEXT OF 3 AND THE .OVIKOV
CONJECTURE
<3OFT AND HARD 'EOMETRIC AND SOME NON GEOMETRIC SPACES AND CAT
EGORIES OF MAPS TENSORS METRICS SUB VARIETIES CAN BE RANKED ALBEIT
AMBIGUOUSLY ACCORDING TO PLASTICITY OR mEXIBILITY OF THE TOTALITY OF THEIR
MEMBERS 4OPOLOGY COULD APPEAR mABBY AND STRUCTURELESS TO 0OINCARsES
CONTEMPORARIES BUT WHEN FACTORED BY HOMOTOPIES THE VERY SOURCE OF mEX
IBILITY IT CRYSTALLIZES TO A RIGID ALGEBRAIC CATEGORY AS HARD AND SYMMETRIC
AS A DIAMOND
2IEMANNIAN MANIFOLDS AS A WHOLE ARE SHAPELESS AND mEXIBLE YET
THEY ABIDE <CONSERVATION LAWS IMPOSED BY THE 'AUSS "ONNET #HERN IDENTI
TIES
$EEPER RIGIDITY APPEARS IN THE PRESENCE OF ELLIPTIC OPERATORS EXTRACTING l
NITE DIMENSIONAL STRUCTURE OUT OF INlNITE DIMENSIONAL DEPTH OF FUNCTIONAL
SPACES !LSO WE START SEEING STRUCTURAL RIGIDITY EG #HEEGER COMPACTNESS
BY lLTERING METRICS THROUGH THE GLASSES OF SAY SECTIONAL CURVATURE
6ISIONS IN -ATH 30!#%3 !.$ 15%34)/.3
+AHLER METRICS AND ALGEBRAIC VARIETIES SEEM STRAIGHT AND RIGID IN

THE 2IEMANNIAN LANDSCAPE NEVER MIND A DENSE SET OF 2IEMANNIAN SPACES
APPEARING AS REAL LOCI OF COMPLEX ALGEBRAIC ONES BUT THEY LOOK SOFTISH TO THE
EYE OF AN ALGEBRAIC GEOMETER (ESHE REINFORCES RIGIDITY WITH THE #ALABI
9AU !UBIN THEOREM TURNING + AHLER TO %INSTEIN +AHLER .OTHING OF THE
KIND SEEMS PLAUSIBLE IN THE FULL 2IEMANNIAN CATEGORY FOR N
(OMOGENEOUS ESPECIALLY SYMMETRIC SPACES STAND ON THE TOP OF THE
GEOMETRIC RIGIDITY HIERARCHY TEMPTING ONE TO Q DEFORM THEM AND SOME
TIMES HIDDEN SYMMETRIES GOVERN INTEGRABLE REGARDED RIGID SYSTEMS 3OFT
NESS IN DYNAMICS IS ASSOCIATED WITH HYPERBOLICITY
,ATTICES ` IN SEMI SIMPLE ,IE GROUPS GROW IN RIGIDITY WITH DIMENSION
PASSING THE CRITICAL POINT AT ` y 3, # WHERE THEY mOURISH IN 4HURSTONS
HYPERBOLIC LAND ! GEOMETER UNHAPPY WITH -OSTOW OVER RIGIDITY FOR N
IS TEMPTED TO SWITCH FROM LATTICES TO LESS CONDENSED SUBGROUPS WITH INl
NITE COVOLUMES AND MORE BALANCED PRESENTATIONS TO THE DISMAY OF A NUMBER
THEORIST THRIVING ON THE FULL ARITHMETIC SYMMETRY OF ` -OST mEXIBLE AMONG
ALL GROUPS ARE THE GENERALIZED SMALL CANCELLATION ONES FOLLOWED BY HIGHER
DIMENSIONAL HYPERBOLIC GROUPS WHILE LATTICES AND lNITE SIMPLE GROUPS ARE
MOST RIGID 6AGUELY SIMILARLY THE RIGIDITY OF ,IE ALGEBRAS INCREASES WITH
DECREASE OF THEIR GROWTH CULMINATING IN +AC -OODY AND lNITE DIMENSIONAL
ALGEBRAS
(OLOMORPHIC FUNCTIONS ON 3TEIN MANIFOLDS 6 ARE RELATIVELY SOFT #AR
TANS THEORY AS WELL AS HOLOMORPHIC MAPS F 6 7 FOR HOMOGE
NEOUS AND ELLIPTIC IE WITH A KIND OF EXPONENTIAL SPRAY 7 BY THE GEN
ERALIZED 'RAUERT THEOREM ALLOWING A HOMOTOPY OF EVERY CONTINUOUS MAP
F 6 7 TO A HOLOMORPHIC ONE (OLOMORPHIC MAPS MODERATED BY
BOUNDS ON GROWTH BECOME MORE RIGID EG FUNCTIONS OF lNITE ORDER HAVE
ESSENTIALLY UNIQUE 7EIERSTRASS PRODUCT DECOMPOSITION !LGEBRAIC MAPS
ORDINARILY RIGID SOMETIMES TURN SOFT EG FOR HIGH DEGREE MAPS OF CURVES
TO 0 BY THE 3EGAL THEOREM !ND 6OEVODSKI THEORY IF ) INTERPRET CORRECTLY
WHAT LITTLE ) UNDERSTOOD FROM HIS LECTURE SOFTENS THE CATEGORY OF ALGEBRAIC
VARIETIES BY INJECTING SOME KIND OF HOMOTOPIES INTO THESE
"IG THREE 4HERE ARE THREE OUTSTANDING INSTANCES WHERE STRIK
ING STRUCTURAL PATTERNS EMERGE FROM LARGE AND mEXIBLE GEOMETRIC SPACES
SYMPLECTICCONTACT DIMENSION AND 3 CONDUCTED IN ALL THREE CASES
BY <2IEMANNIAN AND <ELLIPTIC 7E MET 3 EARLIER WHICH SEEMS LEAST
CONCEPTUALLY UNDERSTOOD AMONG THE BIG THREE SYMPLECTIC AND CONTACT BE
LONG TO %LIASHBERG AND (OFER AT THIS MEETING WITH <SOFT VERSUS <HARD
-)3(! '2/-/6 '!&!
DISCUSSED IN ;'= AND NOBODY ALAS GAVE US A PANORAMA OF N

H 0RINCIPLE 'EOMETERS BELIEVED FROM TILL SINCE #AUCHY
ALMOST PROVED RIGIDITY OF CLOSED CONVEX POLYHEDRAL SURFACES IN 2 THAT
ISOMETRIC IMMERSIONS ARE ESSENTIALLY RIGID 4HEN .ASH DElED EVERYBODYS
INTUITION BY SHOWING THAT EVERY SMOOTH IMMERSION OF A 2IEMANNIAN MANI
FOLD F 6 2. CAN BE DEFORMED FOR . _ t N DIM 6 TO A # SMOOTH
NOT # ISOMETRIC F 6 2. WITH LITTLE LIMITATION FOR THIS DEFORMATION
ALLOWING ONE IN PARTICULAR TO FREELY # DEFORM ALL 6 y 2. KEEPING THE IN
DUCED INTRINSIC METRIC INTACT 4HIS IS SHEER MADNESS FROM A HARD MINDED
ANALYSTS POINT OF VIEW AS THE . COMPONENTS OF F SATISFY NN PARTIAL DIF
FERENTIAL EQUATIONS COMPRISING AN OVER DETERMINED SYSTEM FOR . NN
WHERE ONE EXPECTS NO SOLUTIONS AT ALL 4HE FOLLOWING YEAR +UIPER
ADJUSTED .ASHS CONSTRUCTION TO . N THUS DISPROVING # RIGIDITY OF
CONVEX SURFACES IN 2
.EXT IN 3MALE STUNNED THE WORLD BY TURNING THE SPHERE 3 y 2
INSIDE OUT (E DID IT NOT BY EXHIBITING A PARTICULAR REGULAR HOMOTOPY
THIS WAS DONE LATER AND ONLY CHOSEN FEW ARE ABLE TO FOLLOW IT THROUGH
BUT BY DEVELOPING THE HOMOTOPY THEORETIC APPROACH USED BY 7HITNEY FOR
IMMERSIONS OF CURVES INTO 2 4HEN (IRSCH INCORPORATED 3MALE INTO THE
OBSTRUCTION THEORY AND SHOWED THAT A CONTINUOUS MAP F 6 7 CAN BE
HOMOTOPED INTO AN IMMERSION IF THE OBVIOUS NECESSARY CONDITION IS SATISlED
F LIFTS TO A lBERWISE INJECTIVE HOMOMORPHISM OF TANGENT BUNDLES 4 6
4 7 WITH THE EXCEPTION OF THE CASE OF CLOSED EQUIDIMENSIONAL MANIFOLDS
6 AND 7 WHERE THE PROBLEM IS BY FAR MORE SUBTLE
)T TURNED OUT THAT MANY SPACES 8 OF SOLUTIONS OF PARTIAL DIkERENTIAL
EQUATIONS AND INEQUALITIES ABIDE THE HOMOTOPY PRINCIPLE SIMILAR TO THAT OF
.ASH 3MALE (IRSCH AND 'RAUERT EVERY SUCH 8 IS CANONICALLY HOMOTOPY
EQUIVALENT TO A SPACE OF CONTINUOUS SECTIONS OF SOME JET BUNDLE NATURALLY
ASSOCIATED TO 8 &OR EXAMPLE THE SPACE OF IMMERSIONS 6 7 IS HOMO
TOPY EQUIVALENT TO THE SPACE OF lBERWISE INJECTIVE MORPHISMS 4 6 4 7
BY THE (IRSCH THEOREM
4HE GEOMETRY UNDERLYING THE PROOF OF THE H PRINCIPLE IS SHAMEFULLY SIM
PLE IN MOST CASES ONE CREATES LITTLE ESSENTIALLY DIMENSIONAL rA LA 7HITNEY
WRINKLES IN MAPS X 8 WHICH ARE SPREAD ALL OVER BY HOMOTOPY AND RENDER
8 SOFT AND mEXIBLE "UT THE OUTCOME IS OFTEN SURPRISING AS SEEN IN -ILNORS
INTUITIVELY INCONCEIVABLE TWO DIkERENT IMMERSED DISKS IN THE PLANE WITH
COMMON BOUNDARY WHICH COME UP WITH LOGICAL INEVITABILITY IN %LIASHBERGS
FOLDING THEOREM
6ISIONS IN -ATH 30!#%3 !.$ 15%34)/.3
&IGURE
$ESPITE THE GROWING ARRAY OF SPACES SUBJUGATED BY THE H PRINCIPLE SEE

;'RO=0$2 ;3= WE DO NOT KNOW HOW FAR THIS PRINCIPLE AND SOFTNESS IN GEN
ERAL EXTENDS EG FOR 'AUSS MAPS WITH A PREASSIGNED RANGE SEE ) !RE
THERE SOURCES OF SOFTNESS NOT ISSUING FROM DIMENSION ONE 3OME ENCOURAGING
SIGNS COME FROM 4HURSTONS WORK ON FOLIATIONS 'AO ,OHKAMP H PRINCIPLE
FOR METRICS WITH 2ICCI AND ESPECIALLY FROM $ONALDSONS CONSTRUCTION
OF SYMPLECTIC HYPERSURFACES WHERE <SOFTNESS IS DERIVED FROM A KIND OF
<AMPLENESS NOT DISSIMILAR IN SPIRIT TO 3EGALS THEOREM SEE
! TANTALIZING WISH IS TO lND NEW INSTANCES BESIDES THE BIG THREE WHERE
SOFTNESS REACHES ITS LIMITS WITH SOMETHING GREAT HAPPENING AT THE BOUNDARY
)S THERE YET UNDISCOVERED LIFE AT THE EDGE OF CHAOS !RE WE FOR EVER BOUND
TO ELLIPTIC EQUATIONS )F SO WHAT ARE THEY 4HERE ARE FEW GLOBALLY ELLIPTIC
NON LINEAR EQUATIONS AND NO GENERAL CLASSIlCATION "UT EVEN THOSE WE KNOW
COMING FROM (ARVEY ,AWSON CALIBRATIONS REMAIN MAINLY UNEXPLORED !ND
IF THIS WISH DOES NOT COME TRUE WE STILL CAN MAKE A LIVING IN SOFT SPACES
EXPLORING THEIR GEOMETRY THEIR TOPOLOGY IS COMPLETELY ACCOUNTED FOR BY
THE H PRINCIPLE AS WE DO IN ANISOTROPIC SPACES SEE )))
/UR <SOFT AND HARD ARE NOT MEANT TO REVEAL SOMETHING PROFOUND
ABOUT THE NATURE OF MATHEMATICS BUT RATHER TO PREDISPOSE US TO ACCEPTANCE
OF GEOMETRIC PHENOMENA OF VARIOUS KINDS "ESIDES IT IS OFTEN MORE FRUITFUL
TO REGARD <NUMBERS <SYMMETRIC SPACES <'AL11 <3,N ! 3,N 1
AS <TRUE MATHEMATICAL ENTITIES RATHER THAN DESCENDANTS OF OUR GENERAL
<SPACES <GROUPS <ALGEBRAS ETC "UT ONE CANNOT HELP WONDERING HOW
THESE PERFECT ENTITIES COULD ORIGINATE AND SURVIVE IN THE SOFTLY STRUCTURED
BRAIN HASTILY ASSEMBLED BY BLIND EVOLUTION 3OME BASIC POINT SCIENTIlC
NOT PHILOSOPHICAL SEEMS TO COMPLETELY ELUDE US
-)3(! '2/-/6 '!&!
.ATURE AND NATURALITY OF QUESTIONS (ERE ARE BRIEF INCOMPLETE PER

SONAL AND AMBIGUOUS REMARKS INTENDED TO MAKE CLEARER AT LEAST TERMINO
LOGICALLY THE ISSUES RAISED DURING DISCUSSIONS WE HAD AT THE MEETING
<.ATURAL MAY REFER TO THE STRUCTURE OR NATURE OF MATHEMATICS GRANTED
THIS EXISTS FOR THE SAKE OF ARGUMENT OR TO <NATURAL FOR HUMAN NATURE 7E
DIVIDE THE FORMER INTO PURE MATHEMATICS LOGIC AND PHILOSOPHY AND THE
LATTER ACCORDING TO INTERNAL OR EXTERNAL REWARD STIMULI INTO INTELLECTUAL
EMOTIONAL AND SOCIAL %MOTIONAL PLAYS THE UPPER HAND IN HUMAN DE
CISION AND OPINION EXCEPT FOR A SINGLE MAN YOU MIGHT HAVE THE PRIVI
LEGE TO TALK MATHEMATICS TO AND IN SOME PEOPLE &ERMAT 2IEMANN 7EIL
'ROTHENDIECK I E NATURALLY CONVERGES TO M L P "UT FOR MOST OF US IT IS
NOT EASY TO PROBE THE FUTURE BY CONJECTURALLY EXTRAPOLATING MATHEMATICAL
STRUCTURES BEYOND THE PRESENT POINT IN TIME (OW CAN WE TRUST OUR MIND
OVERWHELMED BY I E S IDEAS TO COME UP WITH TRUE M L P QUESTIONS!N E
S MINDED SOCIOLOGIST WOULD SUGGEST LOOKING AT TRENDS IN FUND DISTRIBUTION
COMPARABLE WEIGHTS OF AUTHORITIES OF SCHOOLS AND INDIVIDUALS AND COULD BE
ABLE TO PREDICT THE INmUENTIAL ROLE OF (ILBERTS PROBLEMS AND "OURBAKI FOR
EXAMPLE WITHOUT BOTHERING TO READ A SINGLE LINE IN THERE !ND <I E S
NATURAL DOES NOT MAKE <A STUPID QUESTION THE COLOR PROBLEM BY ITS
SHEER DInCULTY AND EXPECTATION FOR A STRUCTURALLY REWARDING PROOF HAS
FOCUSED ATTENTION ON GRAPHS WHILE THE SOLUTION HAS CLARIlED THE PERSPECTIVE
ON THE ROLE OF COMPUTERS IN MATHEMATICS "UT THIS BEING UNPREDICTABLE
AND UNREPEATABLE CANNOT HELP US IN M L P EVALUATION OF CURRENT PROBLEMS
WHICH MAY LOOK I E DECEPTIVELY COLORED !S FOR MYSELF ) LOVE UNNATURAL
CRAZILY UNNATURAL PROBLEMS BUT YOU STUMBLE UPON THEM SO RARELY
+ AND /THER -ETRIC 3TORIES

7HAT ARE THE <MOST %UCLIDEAN 2IEMANNIAN MANIFOLDS
7E HAVE ALREADY BEEN ACQUAINTED WITH THE FULLY HOMOGENEOUS SPACES
ALSO CALLED FOR A GOOD REASON COMPLETE SIMPLY CONNECTED OF CONSTANT
CURVATURE + THE ROUND 3 N WITH + THE mAT 2N WITH + AND
THE HYPERBOLIC ( N WITH + _b /BSERVE THAT t3 N AND t( N CONVERGE
DEF
TO 2N FOR t IN A NATURAL SENSE WHERE t8 DIST 8t DIST AND

+t8 t_ +8 AS IS CLEARLY SEEN FOR EXAMPLE FOR t SCALED SURFACES
8 y 2 .OW SOMEWHAT PERVERSELY WE BRING IN TOPOLOGY AND ASK FOR
b
4HE FOURTH AND THE LAST FULLY HOMOGENEOUS 2IEMANNIAN SPACE IS 0 N 3 N FeG
6ISIONS IN -ATH 30!#%3 !.$ 15%34)/.3
COMPACT MANIFOLDS WITH CONSTANT CURVATURE IE LOCALLY ISOMETRIC TO ONE

OF THE ABOVE 3 N 2N OR ( N ,ETTING 3 N GO WE START WITH THE mAT IE
+ CASE AND CONlRM THAT COMPACT LOCALLY %UCLIDEAN MANIFOLDS EXIST
JUST TAKE A LATTICE c IN 2N EG c :N AND LOOK AT THE TORUS 4 2N cN
%SSENTIALLY THERE IS LITTLE ELSE TO SEE
& THEOREM %VERY COMPACT mAT MANIFOLD 8 IS COVERED BY A TORUS WITH

THE NUMBER OF SHEETS BOUNDED BY A UNIVERSAL CONSTANT KN
4HIS SOUNDS DRY BUT IT HIDES A LITTLE ARITHMETIC GERM ON THE BOTTOM
THERE IS NO REGULAR K GON WITH VERTICES IN A LATTICE c EG c : IN 2
FOR K t OR FOR K )NDEED TRANSPORTED EDGES OF SUCH K GON 2 WOULD
MAKE A SMALLER REGULAR 2 y c AND THE CONTRADICTION FOLLOWS BY ITERATION
2 2
R
R'
&IGURE
4HE TORI 4 THEMSELVES STOP LOOKING mAT AS THEY ALL TOGETHER MAKE THE
MARVELLOUS MODULI SPACE 3/N N3,N 2 3,N : OF ISOMETRY CLASSES OF
4 S WITH 6OL 4 LOCALLY ISOMETRIC TO 3,N 2 3/N APART FROM MILD
ORBIFOLD SINGULARITIES DUE TO ELEMENTS OF lNITE ORDER IN 3,N :
4URNING TO + _ WE MAY START WONDERING IF SUCH SPACES EXIST IN

A COMPACT FORM AT ALL 4HEN FOR N WE OBSERVE THAT THE ANGLES OF
SMALL REGULAR K GONS 2 y ( ARE ALMOST THE SAME AS IN 2 WHILE LARGE
2 y ( HAVE ALMOST ZERO ANGLES THUS BY CONTINUITY FOR EVERY K t THERE
EXISTS 2c y ( WITH m ANGLES 7E REmECT ( IN THE LINES EXTENDING
THE SIDES OF 2c AND TAKE THE SUBGROUP ` y )SOM( GENERATED BY THESE
K REmECTIONS 4HIS ` IS DISCRETE ON ( WITH 2c SERVING AS A FUNDAMENTAL
DOMAIN SIMILARLY TO THE CASE OF THE SQUARE 2c y 2 AND THE QUOTIENT SPACE
( ` EQUAL 2 BECOMES AN HONEST MANIFOLD RATHER THAN ORBIFOLD IF WE
TAKE INSTEAD OF ` A SUBGROUP ` y ` WITHOUT TORSION WHICH IS NOT HARD TO
lND
-)3(! '2/-/6 '!&!
&IGURE
4HE SAME IDEA WORKS FOR DODECAHEDRA IN ( AND SOME OTHER CONVEX
POLYHEDRA IN ( N FOR SMALL N BUT THERE ARE NO COMPACT HYPERBOLIC RE
mECTION GROUPS FOR LARGE N BY A DInCULT THEOREM OF 6INBERG 4HE ONLY
KNOWN SOURCE OF HIGH DIMENSIONAL ` COMES FROM ARITHMETICS ESSENTIALLY
BY INTERSECTING 3/N SOMEHOW EMBEDDED INTO 3,. 2 WITH 3,. :
WHERE THE ORTHOGONAL GROUP 3/N DOUBLY COVERS THE ISOMETRY GROUP
0 3/N OF ( N y 2N .ON ARITHMETIC ` ARE ESPECIALLY PLENTIFUL FOR
N BY 4HURSTONS THEORY AND OFTEN HAVE UNEXPECTED FEATURES EG SOME
6 ( ` lBER OVER 3 WHICH IS HARD TO IMAGINE EVER HAPPENING FOR LARGE
N -OREOVER THE TOPOLOGICAL MANIFOLDS lBERED OVER 3 GENERICALLY IE
FOR PSEUDO !NOSOV MONODROMY ADMIT METRICS WITH + _ 5NBELIEVABLE
[ BUT TRUE BY 4HURSTON WHO HIMSELF DOES NOT EXCLUDE THAT lNITE COVERS
OF MOST ATORICAL 6 lBER OVER 3 YET THIS REMAINS OPEN EVEN FOR 6 WITH
+ _
!LEXANDROVS SPACES 7HAT ARE THE MOST GENERAL CLASSES OF SPACES
SIMILAR TO THOSE WITH + e
!LEXANDROV SUGGESTED AN ANSWER IN BY INTRODUCING SPACES WITH
+ t WHERE THE GEODESIC TRIANGLES HAVE THE SUM OF ANGLES t x AND
THOSE WITH + s WHERE AT LEAST SMALL TRIANGLES HAVE IT s x "UT WE
TAKE ANOTHER MORE FUNCTORIAL ROUTE DEPARTING FROM THE FOLLOWING
%UCLIDEAN + THEOREM %VERY ,IPSCHITZ IE DISTANCE NON INCREASING
MAP F FROM A SUBSET a y 2N TO SOME 2M ADMITS A ,IPSCHITZ EXTENSION
F 2N 2M FOR ALL M N s
4HIS IS SHOWN BY CONSTRUCTING F POINT BY POINT AND LOOKING AT THE WORST
CASE AT EACH STAGE WHERE EXTENDABILITY FOLLOWS FROM AN OBVIOUS GENERALIZA
TION OF THE PRETTY LITTLE LEMMA
,EMMA ,ET a AND a IN 3 N_ y 2N BE THE SETS OF VERTICES OF TWO SIMPLICES
INSCRIBED INTO THE SPHERE 3 N_ WHERE THE EDGES OF a ARE CORRESPONDINGLY
s THAN THOSE OF a 4HEN a IS CONGRUENT TO a IE s PROVIDED a
IS NOT CONTAINED IN A HEMISPHERE
6ISIONS IN -ATH 30!#%3 !.$ 15%34)/.3
.OW WE SAY THAT A METRIC SPACE 8 HAS + t IF EVERY PARTIAL ,IPSCHITZ

MAP 8 z a 2M EXTENDS TO A ,IPSCHITZ F 8 2M FOR ALL M WHILE
+ s IS DElNED WITH SUCH EXTENSIONS FOR 2N z a 8 WHERE FOR + s
WE ADDITIONALLY REQUIRE UNIQUENESS OF MIDDLE POINTS IN 8 SEE BELOW
4O MAKE THIS WORTHWHILE ONE ADDS THE METRIC COMPLETENESS OF 8 AND
THE LOCALITY PROPERTY DISTX X SHOULD EQUAL THE INlMAL LENGTH OF CURVES IN
8 BETWEEN X AND X %QUIVALENTLY THERE IS A MIDDLE POINT Y 8 WHERE
DISTX Y DISTY X DISTX X 4HEN ONE ARRIVES AT THE FOLLOWING
ELEGANT PROPOSITION JUSTIFYING THE DElNITIONS
+ 4HEOREM )F +8 t AND +9 s THEN EVERY PARTIAL ,IPSCHITZ

MAP 8 z a 9 ADMITS A ,IPSCHITZ EXTENSION 8 9
4O APPLY THIS ONE NEEDS EXAMPLES OF SPACES WITH + AND THESE ARE
EASIER TO OBSERVE WITH !LEXANDROVS DElNITION &ORTUNATELY BOTH DElNITIONS
ARE EQUIVALENT AND WE HAVE
#OMPLETE EG CLOSED CONVEX HYPERSURFACES IN 2N HAVE + t WHILE
SADDLE SURFACES IN 2 HAVE AT LEAST LOCALLY + s 3YMMETRIC SPACES OF
COMPACT TYPE HAVE + t WHILE THOSE OF NON COMPACT TYPE HAVE + s
4AKE A DIMENSIONAL POLYHEDRON 6 ASSEMBLED OF CONVEX %UCLIDEAN K
GONS AND OBSERVE THAT THE LINK , OF EACH VERTEX IN 6 IS A GRAPH IE
COMPLEX WITH A LENGTH ASSIGNED TO EVERY EDGE E EQUAL THE K GONAL ANGLE
CORRESPONDING TO E 4HEN +6 t EACH , IS ISOMETRIC TO A SEGMENT
s x OR TO A CIRCLE NO LONGER THAN x
+6 s ALL CYCLES IN ALL , ARE LONGER THAN x AND 6 IS SIMPLY
CONNECTED
&INALLY IF 8 HAS LOCALLY + t THEN THE SAME IS TRUE GLOBALLY WHILE
GLOBALLY + s LOCALLY + s x
4HE x CONDITION BREAKS THE HARMONY ) GUESS IT WAS UPSETTING
!LEXANDROV AND BRINGS CONFUSION TO THE NOTION OF + s AS THE <LOCAL
AND <GLOBAL MEANINGS DIVERGE "UT AT THE END OF THE DAY THE x RIPPLE
MAKES THE GEOMETRY OF + s MUCH RICHER AND SOFTER THAN ALL WE KNOW
OF + t SINCE THERE ARE LOTS OF SPACES 6 WITH +LOC s AS ALREADY SEEN
IN THE DIMENSIONAL POLYHEDRA WHERE THE GROUP THEORETIC STUDY OF x 6
MAY RELY ON GEOMETRY FOR EXAMPLE IN THE .OVIKOV CONJECTURE "ESIDES
THE GLOBAL DElNITION OF + s CAN BE RELAXED BY ROUGHLY ALLOWING t
,IPSCHITZ EXTENSION OF PARTIAL ,IPSCHITZ MAPS WITH s t s CONST
BRINGING ALONG LARGER CLASSES OF HYPERBOLIC LIKE SPACES AND GROUPS WHERE
GEOMETRY AND ALGEBRA ARE ENGAGED IN A MEANINGFUL CONVERSATION
-)3(! '2/-/6 '!&!
!NISOTROPIC SPACES 4HERE IS A CLASS OF METRICS WHICH CAN BE ANA

LYTICALLY GENERATED WITH THE SAME EASE AS THE 2IEMANNIAN ONES BESIDES WE
lND AMONG THEM SPACES 8 IN SOME WAY CLOSER TO 2N THAN 3 N AND ( N THESE
8 ARE METRICALLY HOMOGENEOUS AS WELL AS SELF SIMILAR IE t8 IS ISOMETRIC
TO 8 FOR ALL t )N THE 2IEMANNIAN CATEGORY THERE IS NOTHING BUT 2N
LIKE THAT
! POLARIZATION ON A SMOOTH MANIFOLD 6 EG 2N OR A DOMAIN IN 2N
IS A SUBBUNDLE ( OF THE TANGENT BUNDLE 4 6 IE A lELD OF M PLANES ON
6 WHERE FOR s M s N _ N DIM 6 IN THE CASE AT HAND "ESIDES (
WE NEED AN AUXILIARY 2IEMANNIAN METRIC G ON 6 BUT WHAT MATTERS IS THE
RESTRICTION OF G ON ( 7E DElNE DISTV V DIST(G V V BY TAKING THE
INlMUM OF G LENGTHS OF PIECEWISE SMOOTH CURVES BETWEEN V AND V WHICH ARE
CHOSEN AMONG ( HORIZONTAL CURVES IE THOSE WHICH ARE EVERYWHERE TANGENT
TO ( )T MAY HAPPEN THAT THIS DISTANCE IS INlNITE EVEN FOR CONNECTED 6
IF SOME POINTS ADMIT NO HORIZONTAL CONNECTING PATHS BETWEEN THEM AS IT
HAPPENS FOR INTEGRABLE ( WHERE DISTV V V AND V LIE IN THE
SAME LEAF OF THE FOLIATION INTEGRATING ( 4HIS IS NOT SO BAD AS IT SEEMS
BUT WE WANT DIST AT THE MOMENT AND SO WE INSIST ON THE EXISTENCE OF
A HORIZONTAL PATH BETWEEN EVERY TWO POINTS IN 6 )T IS NOT HARD TO SHOW
THAT GENERIC # SMOOTH POLARIZATIONS ( DO HAVE THIS PROPERTY FOR M t
WHERE <GENERIC IMPLIES IN PARTICULAR THAT THE SPACE OF <GOOD (S IS OPEN
AND DENSE IN THE SPACE OF ALL # POLARIZATIONS ON 6
! PRACTICAL WAY FOR CHECKING THIS IS TO TAKE M t M VECTOR lELDS TANGENT
TO ( AND SPANNING IT THESE ALWAYS EXIST SAY 8 8M 4HEN THE
SUnCIENT CRITERION FOR OUR ( CONNECTIVITY ALSO CALLED CONTROLLABILITY IS AS
FOLLOWS 4HE SUCCESSIVE COMMUTATORS 8I ;8I 8J = ;;8I 8J = 8K = OF THE
lELDS SPAN THE TANGENT BUNDLE 4 6
4HE SIMPLEST INSTANCE OF THE ABOVE IS THE PAIR OF THE lELDS 8I X AND
8 X X X IN 2 WHERE ;8 8 = _ X (ERE ( CAN BE PRESENTED
AS THE KERNEL OF THE STANDARD CONTACT FORM DX X DX AND IN FACT CONTACT
lELDS ( ARE ( CONNECTED FOR ALL CONTACT MANIFOLDS OF DIMENSION t
.EXT LOOK AT A LEFT INVARIANT POLARIZATION ( ON A ,IE GROUP ' DElNED
BY THE LEFT TRANSLATES OF A LINEAR SUBSPACE H IN THE ,IE ALGEBRA , ,'
)T IS NOT HARD TO SEE THAT THE COMMUTATORS CONDITION HOLDS THERE IS
NO ,IE SUBALGEBRA IN , CONTAINING H BESIDES , ITSELF 4AKE FOR INSTANCE
THE DIMENSIONAL (EISENBERG GROUP ' HOMEOMORPHIC TO 2 WITH ,
,' GENERATED BY X X X WHERE ;X X = X AND X COMMUTES WITH
6ISIONS IN -ATH 30!#%3 !.$ 15%34)/.3
X AND X (ERE ONE TAKES H SPANNED BY X X AND OBSERVE THAT THIS H

IS INVARIANT UNDER AUTOMORPHISMS OF , DElNED BY X t X X tX
X tX FOR ALL t 4HEN FOR EACH LEFT INVARIANT METRIC G ON ' THE
CORRESPONDING AUTOMORPHISMS !t ' ' PRESERVES ( AND SCALE G ON (
BY t 4HUS ' IS SELFSIMILAR BUT QUITE DIkERENT FROM 2 &OR EXAMPLE THE
(AUSDORk DIMENSION OF ' DIST(G IS RATHER THAN
7HAT ARE NATURAL MAPS BETWEEN THE ABOVE NON ISOTROPIC SPACES ,IPS
CHITZ MAPS DO NOT SERVE AS WELL HERE AS IN THE 2IEMANNIAN CATEGORY SINCE
OUR NEW SPACES ARE USUALLY NOT MUTUALLY BI ,IPSCHITZ EQUIVALENT /N THE
OTHER HAND TYPICAL SPACES 8 AND 9 OF THE SAME TOPOLOGICAL DIMENSION ARE
LOCALLY (
OLDER HOMEOMORPHIC WITH SOME POSITIVE EXPONENT j WITH A
BOUND j jN "UT THE OPTIMAL VALUE OF j FOR GIVEN SPACES OR CLASSES
OF THOSE REMAINS UNKNOWN ! SIMILAR PROBLEM IS THAT OF lNDING A SHARP
LOWER BOUND FOR THE (AUSDORk DIMENSION OF A SUBSET 9 y 8 8 DIST(G
BY THE TOPOLOGICAL DIMENSION IN TERMS OF COMMUTATION PROPERTIES OF lELDS
SPANNING ( 4HIS IS QUITE EASY IF 9 IS A SMOOTH SUBMANIFOLD IN 8
4HE LARGE SCALE GEOMETRY OF 8 DIST(G IS RATHER CLOSE TO THE 2IEMAN
NIAN GEOMETRY OF 8 DISTG AND SO ( DOES NOT MATTER MUCH #ONVERSELY
THE LOCAL GEOMETRY ESSENTIALLY DEPENDS ON ( AND VERY LITTLE ON G AND THE
MAIN OPEN PROBLEMS ARE BASICALLY LOCAL
#ONCENTRATED MM SPACES ,ET 8 BE A METRIC SPACE WHICH IS ALSO GIVEN
A "OREL MEASURE u OFTEN ASSUMED TO BE A PROBABILITY MEASURE IE u8
4HIS u MAY COME OUT OF THE METRIC EG THE 2IEMANNIAN MEASURE
SOMETIMES NORMALIZED TO HAVE u8 BUT OFTEN u HAS LITTLE TO DO WITH
THE ORIGINAL DISTANCE AS FOR EXAMPLE THE 'AUSSIAN MEASURE ON 2N 7E WANT
TO STUDY 8 DIST u IN A PROBABILISTIC FASHION BY THINKING OF FUNCTIONS F
ON 8 AS RANDOM VARIABLES BEING CONCERNED WITH THEIR DISTRIBUTIONS IE
PUSH FORWARD MEASURES Fb u (ERE IS OUR
"ASIC PROBLEM 'IVEN A MAP F 8 9 RELATE METRIC PROPERTIES OF F
TO THE STRUCTURE OF THE MEASURE Fb u ON 9
(ERE <METRIC REFERS TO HOW F DISTORTS DISTANCE EXPRESSED FOR INSTANCE
BY THE ,IPSCHITZ CONSTANT tF ? WHERE WE DISTINGUISH THE CASE OF FUNC
TIONS IE OF 9 2 /N THE PROBABILISTIC SIDE WE SPEAK OF <STRUCTURE OF
F b u EXPRESSED ENTIRELY IN TERMS OF 9 AND WHERE A TYPICAL QUESTION IS HOW
CONCENTRATED Fb u IS IE HOW CLOSE IT IS TO A POINT MEASURE IN 9
'AUSSIAN EXAMPLE 4AKE 2. FOR 8 WITH THE MEASURE EXP _KXK DX
4HEN EVERY ,IPSCHITZ FUNCTION F 2. 2 IS AT LEAST AS MUCH CONCEN
-)3(! '2/-/6 '!&!
TRATED AS THE ORTHOGONAL PROJECTION F 2. 2 4HERE IS A ,IPSCHITZ

SELF MAPPING 2 2 PUSHING FORWARD Fb u TO Fb u
! PARALLEL EXAMPLE WHERE THE GEOMETRY IS SEEN MORE CLEARLY IS 8 3 .
WITH THE NORMALIZED 2IEMANNIAN MEASURE (ERE AGAIN EVERY ,IPSCHITZ
FUNCTION F 3 . 2 IS CONCENTRATED AS MUCH AS THE LINEAR ONE CONSE
QUENTLY
P F b u CONVERGES FOR . TO A m MEASURE ON 2 WITH THE RATE
x .
4HE ESSENCE OF THE ABOVE CONCENTRATION IS THE SHARP CONTRAST BETWEEN THE
SPREAD OF THE ORIGINAL MEASURE ON 8 EG THE DISTANCE BETWEEN u RANDOM
POINTS IN 3 . IS x x FOR LARGE . AND STRONG LOCALIZATION OF Fb u PON
9 EG THE CHARACTERISTIC DISTANCE ON 2 WITH RESPECT TO Fb u IS x .
IN THE SPHERICAL CASE 4HE FOLLOWING DElNITION IS AIMED TO CAPTURE THIS
PHENOMENON IN THE LIMIT FOR . BY INTERBREEDING METRIC GEOMETRY
WITH THE ERGODIC THEORY NOT QUITE AS IN THE ERGODIC THEOREM WHERE F F.
APPEARS AS THE AVERAGE OF . TRANSFORMS OF A GIVEN F
,ET 8 BE A PROBABILITY MEASURE SPACE AND D 8 a 8 2 ;
A FUNCTION SATISFYING THE STANDARD METRIC AXIOMS EXCEPT THAT WE ALLOW
DX X )N FACT WE ARE KEEN FOR THE APPARENTLY ABSURD SITUATION OF
D ALMOST EVERYWHERE ON 8 MODERATED BY THE
%RGODICITY AXIOM &OR EVERY 9 y 8 WITH u8 THE DISTANCE TO 9
D9 X INF DX Y
DEF Y9
IS MEASURABLE AND AE lNITE ON 8
%XAMPLE ,ET 8 BE A FOLIATED MEASURE SPACE WHERE EACH LEAF IS MEASUR
ABLY IN X 8 ASSIGNED A METRIC 4HEN WE DElNE
y
IF X Y ARE NOT IN THE SAME LEAF
DX Y
DX Y DIST, X Y IF X AND Y LIE IN SOME LEAF ,
AND OBSERVE THAT OUR ERGODICITY AXIOM AMOUNTS HERE TO THE ORDINARY ERGOD
ICITY
.EXT WE DISTINGUISH CONCENTRATED SPACES BY INSISTING ON THE UNIVERSAL
BOUND ON THE DISTANCES BETWEEN SUBSETS IN 8 IN TERMS OF THE MEASURES OF
THESE SUBSETS (ERE
DEF
DIST9 9 INF DY Y INF D9 X
X9
AND THE BOUND IS GIVEN BY A FUNCTION #A A A A WHERE # MAY GO
TO INlNITY FOR A A SO THAT DIST9 9 s #u9 u9
%XAMPLES )F 8 IN THE PREVIOUS EXAMPLE IS FOLIATED IE PARTITIONED INTO
THE ORBITS OF AN AMENABLE GROUP ' ACTING ON 8 THEN THE RESULTING D ON
6ISIONS IN -ATH 30!#%3 !.$ 15%34)/.3
8 IS ESSENTIALLY NEVER CONCENTRATED "UT IF ' HAS PROPERTY 4 THEN IT IS

CONCENTRATED
,ET 8 8 BE A SEQUENCE OF 2IEMANNIAN MANIFOLDS AND 8 BE THE
INlNITE #ARTESIAN PRODUCT 8 a 8 a WHERE THE <METRIC BETWEEN
INlNITE SEQUENCES X X X AND Y Y Y IS THE 0YTHAGOREAN
ONE
DEF _ `
DISTX Y DIST X Y DIST X Y
7HAT GOES WRONG HERE IS THAT DISTX Y FOR MOST X AND Y BUT WE
CAN TOLERATE THIS IN THE PRESENCE OF THE PRODUCT MEASURE u ON 8 COMING
FROM NORMALIZED 2IEMANNIAN MEASURES ON 8I )F THE lRST SOMETIMES CALLED
SECOND EIGENVALUES OF THE ,APLACE OPERATOR a ON 8I ARE SEPARATED FROM
ZERO t 8I t THEN THE PRODUCT IS CONCENTRATED AND MOREOVER ONE
CAN MAKE A MEANINGFUL ANALYSIS ON FUNCTIONS AND OFTEN ON FORMS ON 8
&URTHERMORE IF t 8I FOR I THEN a HAS DISCRETE SPECTRUM ON 8
WITH lNITE MULTIPLICITY COMPUTED BY THE USUAL FORMULA FOR PRODUCTS &OR
INSTANCE IF 8I ARE NI SPHERES OF RADII 2I THEN t 8I NI 2I_ AND THEIR
P P
PRODUCT IS CONCENTRATED FOR 2I s NI WITH EXTRA BENElTS FOR NI 2I
/NE CAN DEFORM AND MODIFY PRODUCTS RETAINING CONCENTRATION EG FOR
PROJECTIVE LIMITS OF SOME TOWERS OF SMOOTH lBRATION SUCH AS THE INlNITELY
ITERATED UNIT TANGENT BUNDLES OF 2IEMANNIAN MANIFOLDS
3IMILARLY TO PRODUCTS THE SPACES 8 OF MAPS BETWEEN 2IEMANNIAN MANI
FOLDS ! " CARRY MANY DIkERENT <FOLIATED (ILBERT MANIFOLD STRUCTURES
WHICH IN THE PRESENCE OF MEASURES EG 7IENER MEASURE FOR DIMENSIONAL
! ALLOW ANALYSIS ON 8
.OW COMES A PAINFUL QUESTION ARE THESE 8 GOOD FOR ANYTHING $O
THEY POSSESS A STRUCTURAL INTEGRITY OR JUST ENCOMPASS MANY BUT SO WHAT
EXAMPLES ! CONVINCING THEOREM IS TO BE PROVED YET
2EMARKS AND REFERENCES A 4HE & THEOREM MAKES THE CORE OF
"IEBERBACHS SOLUTION TO A PROBLEM ON (ILBERTS LIST . WHERE THE N
DIMENSIONAL HYPERBOLIC CASE IS ALSO MENTIONED AND DISMISSED AS ADDING LITTLE
NEW TO THE RESULTS AND METHODS OF &RICKE AND +LEIN
B /UR DElNITION OF + IS MOTIVATED BY ;,3= WHERE THE AUTHORS
PROVE THE + THEOREM FOR MAPS FROM SPACES WITH + t t TO THOSE WITH
+ s t IN THE SENSE OF !LEXANDROV UNDER A MILD RESTRICTION RULING OUT FOR
EXAMPLE MAPS 3 N 3 M FOR M N WHERE + PROPERTY OBVIOUSLY FAILS TO
BE TRUE BUT UNFORTUNATELY MISSING MAPS 3 N 3 M FOR M t N WHERE IT IS
KNOWN TO BE TRUE A CONCLUSIVE VERSION SEEMS NOT HARD BUT NO PUBLISHED
PROOF IS AVAILABLE !LSO THERE IS A ,IPSCHITZ EXTENSION RESULT FROM ARBITRARY
-)3(! '2/-/6 '!&!
METRIC SPACES TO THOSE WITH + s DUE TO ,ANG 0AVLOVIC AND 3CHROEDER

C 4HE THEORY OF SPACES WITH + t AS WELL AS WITH + s t IS
BY NOW WELL DEVELOPED SEE ;0= AND INTRINSICALLY ATTRACTIVE 9ET IT SUkERS
FROM THE LACK OF A SYSTEMATIC PROCESS OF GENERATING SUCH SPACES APART FROM
CONVEX HYPERSURFACES AND DESPITE SEVERAL GENERAL CONSTRUCTIONS PRODUCTS
QUOTIENTS SPHERICAL SUSPENSION !LSO THERE IS NO SERIOUS LINK WITH OTHER
BRANCHES OF GEOMETRY NOT EVEN WITH THE LOCAL THEORY OF "ANACH SPACES AND
THERE ARE FEW THEOREMS + THEOREM IS A HAPPY EXCEPTION WHERE THE CON
CLUSION IS HARDER TO VERIFY IN THE AVAILABLE EXAMPLES THAN THE ASSUMPTIONS
AS IS UNFORTUNATELY FREQUENT IN GLOBAL 2IEMANNIAN GEOMETRY ! POSSIBLE
WAY OF ENRICHING AND SOFTENING + t IS LETTING N AND TAKING
N SERIOUSLY AS IS DONE FOR "ANACH SPACES 3EE ;"ER= FOR A BROADER
PERSPECTIVE ;0E= FOR A MOST RECENT ACCOUNT AND ;'= FOR A PEDESTRIAN GUIDE
TO CURVATURE
D 2ICCI !NALYTICALLY SPEAKING THE MOST NATURAL OF CURVATURES IS THE
2ICCI TENSOR THAT IS THE QUADRATIC DIkERENTIAL FORM ON 6 ASSOCIATED TO G VIA
A ESSENTIALLY UNIQUE $Ik INVARIANT SECOND ORDER QUASI LINEAR DIkERENTIAL
OPERATOR LIKE 3 DENOTED 2I 2IG -ANIFOLDS WITH POSITIVE DElNITE 2I
GENERALIZE THOSE WITH + t BUT THEY ADMIT NO SIMPLE METRIC DESCRIPTION
AS THEIR ESSENTIAL FEATURES INVOLVE THE 2IEMANNIAN MEASURE ON 6 EG 2
BALLS IN 6 WITH 2I t HAVE SMALLER VOLUMES THAN THE %UCLIDEAN ONES )F
+ t IS MOTIVATED BY CONVEXITY THEN 2I t CAN BE TRACED TO POSITIVE
MEAN CURVATURE OF HYPERSURFACES 4HUS IT REMAINS UNCLEAR HOW FAR THE IDEA
OF 2I t EXTENDS BEYOND THE 2IEMANNIAN CATEGORY WHAT ARE ADMISSIBLE
SINGULARITIES AND WHAT HAPPENS FOR N 3EE ;'= AND ;'= FOR AN
INTRODUCTION AND ;##= FOR THE PRESENT STATE OF THE ART
%NCOURAGED BY 2I t ONE TURNS TO 2I s FORMALLY GENERALIZING + s
BUT THE NAIVE LOGIC DOES NOT WORK EVERY METRIC CAN BE APPROXIMATED
BY THOSE WITH 2I BY THE ,OHKAMP H PRINCIPLE AND NO HARD STRUCTURAL
GEOMETRY EXIST
E %INSTEIN AND THE FORLORN QUEST FOR THE BEST METRIC )T IS
THE GEOMETERS DREAM lRST ARTICULATED BY (EINZ (OPF ) BELIEVE TO lND A
CANONICAL METRIC GBEST ON A GIVEN SMOOTH MANIFOLD 6 SO THAT ALL TOPOLOGY OF
6 WILL BE CAPTURED BY GEOMETRY 4HIS HAPPENED TO COME TRUE FOR SURFACES
AS ALL OF THEM CARRY ALMOST UNIQUE METRICS OF CONSTANT CURVATURE AND IS
PREDICTED FOR N BY 4HURSTONS GEOMETRIZATION CONJECTURE !LSO THERE
IS A GLIMPSE OF HOPE FOR N %INSTEIN SELF DUAL BUT NO TRACE OF GBEST HAS
EVER BEEN FOUND FOR N t 7HAT IS THE REASON FOR THIS ,ET US TAKE SOME
6ISIONS IN -ATH 30!#%3 !.$ 15%34)/.3
ENERGY FUNCTION % ON 2THE SPACE ' OF METRICS SAY BUILT OF THE FULL CURVATURE
N
TENSOR SOMETHING LIKE #URVG DV WHERE THE EXPONENT N MAKES THE
INTEGRAL SCALE INVARIANT )MAGINE THE GRADIENT mOW OF % BRINGS ALL OF '
TO A <NICE SUBSPACE 'BEST y ' IDEALLY A SINGLE POINT OR SOMETHING NOT
VERY LARGE ANYWAY 4HEN THE GROUP $Ik 6 WOULD ACT ON 'BEST AS ALL WE DO
SHOULD BE $Ik INVARIANT WITH COMPACT ISOTROPY SUBGROUPS WE ASSUME 6 IS
COMPACT AT THE MOMENT EG IF 'BEST CONSISTED OF A SINGLE POINT THEN $Ik
6 WOULD ISOMETRICALLY ACT ON 6 'BEST "UT THE HIGH DIMENSIONAL TOPOLOGY
UNKNOWN TO (OPF TELLS US THAT THE SPACE $Ik 6 IS TOO VAST SOFT AND UNRULY
TO BE CONTAINED IN SOMETHING NICE AND COSY LIKE THE DESIRED 'BEST $Ik IS
GOVERNED BY 7ALDHAUSEN + THEORY BRINGING LOTS OF HOMOTOPY TO $Ik WHICH
HARDLY CAN BE ACCOUNTED FOR BY A RIGID GEOMETRY !ND PRIOR TO 7ALDHAUSEN
OUR DREAM WAS SHATTERED BY -ILNORS SPHERES RULING OUT SMOOTH CANONICAL
DEFORMATIONS OF GENERAL METRICS ON THE SPHERES 3 N N t TO THE STANDARD
ONES
"ESIDES TOPOLOGY THERE IS A GEOMETRIC REASON WHY WE CANNOT FREELY NAV
IGATE IN THE RUGGED LANDSCAPES OF SPACES LIKE $Ik 4O SEE THE IDEA LET US
LOOK AT THE SIMPLER PROBLEM OF lNDING THE <BEST CLOSED CURVE IN A GIVEN
FREE HOMOTOPY CLASS OF LOOPS IN A 2IEMANNIAN MANIFOLD 6 )F +6 s
FOR INSTANCE THE GRADIENT mOW OF THE ENERGY FUNCTION ON CURVES HAPPILY
TERMINATES AT A UNIQUE GEODESIC THIS IS THE BEST WE COULD HOPE FOR "UT
SUPPOSE x 6 IS COMPUTATIONALLY COMPLICATED IE THE WORD PROBLEM CAN
NOT BE SOLVED BY A FAST ALGORITHM SAY UNSOLVABLE BY ANY ALGORITHM AT ALL
/UR mOW DISCRETIZED IN AN OBVIOUS WAY IS A PARTICULAR ALGORITHM WE KNOW
IT MUST BADLY FAIL AND THE ONLY WAY FOR IT TO FAIL IS TO GET LOST AND CONFUSED
IN DEEP LOCAL MINIMA OF THE ENERGY 4HUS OUR 6 MUST HARBOR LOTS AND LOTS OF
LOCALLY MINIMAL GEODESICS IN EACH HOMOTOPY CLASS IN PARTICULAR INlNITY OF
CONTRACTIBLE CLOSED GEODESICS DISRUPTING THE ROUTE FROM TOPOLOGY TO SIMPLE
GEOMETRY
7ELL ONE MAY SAY LET US ASSUME x 6 "UT WHAT THE HELL DOES
<ASSUME MEAN 'IVEN A 6 PRESENTED IN ANY CONCEIVABLE GEOMETRIC FORM
REMEMBER WE ARE GEOMETERS HERE NOT SHAPE BLIND TOPOLOGISTS THERE IS
NO WAY TO CHECK IF x 6 SINCE THIS PROPERTY IS NOT ALGORITHMICALLY
VERIlABLE #ONSEQUENTLY THERE ARE INNOCUOUSLY LOOKING METRICS ON SUCH
MANIFOLDS AS 3 N FOR N t TEEMING WITH SHORT CLOSED CURVES WHICH NO
HUMAN BEING CAN CONTRACT IN A GIVEN STRETCH OF TIME )N FACT A PREDOMINANT
MAJORITY OF METRICS ARE LIKE THIS FOR N t ! SIMILAR PICTURE ARISES FOR
HIGHER DIMENSIONAL EG MINIMAL SUBVARIETIES WITH EXTRA COMPLICATION FOR
-)3(! '2/-/6 '!&!
LARGE DIMENSION AND CODIMENSION EVEN FOR SIMPLEST 6 SUCH AS mAT TORI
WHERE THE TRICKERY OF MINIMAL SUBVARIETIES WAS DISCLOSED BY "LAINE ,AWSON
AND BY THE WORK OF !LEX .ABUTOVSKI THE SPACES LIKE '6 $Ik6 HARBOR
THE SAME KIND OF COMPLEXITY ROOTED IN THE 'ODEL 4URING THEOREM
&OLLOWING !LEX WE ) SPEAK FOR MYSELF ARE LEAD TO THE PESSIMISTIC CON
CLUSION THAT THERE IS NO CHANCE FOR A DISTINGUISHED GBEST OR EVEN 'BEST y '
FOR N t AND THAT <NATURAL METRICS EG %INSTEIN ' WITH 2IG tG FOR
t MUST BE CHAOTICALLY SCATTERED IN THE VASTNESS OF ' WITH NO MEANINGFUL
LINK BETWEEN GEOMETRY AND TOPOLOGY 4HIS DOES NOT PRECLUDE BUT RATHER
PREDICTS THE EXISTENCE OF SUCH METRICS EG %INSTEIN ON ALL 6 OF DIMENSION
t THE PROBLEM IS THERE MAY BE TOO MANY OF THEM
/N THE OPTIMISTIC SIDE WE CONTINUE SEARCHING FOR GBEST IN SPECIAL DOMAINS
IN ' EG FOLLOWING (AMILTONS 2ICCI mOW SAY FOR 2ICCI t OR + s OR
STICK TO DIMENSIONS THREE WHERE -ICHAEL !NDERSON MAKES HIS THEORY OR
FOUR WHERE 4AUBS DISCOVERED CERTAIN SOFTNESS IN SELFDUALITY !LTERNATIVELY
WE CAN ENLARGE RATHER THAN LIMIT THE CATEGORY AND LOOK FOR EXTREMAL POS
SIBLY SINGULAR VARIETIES WITH ONLY PARTIALLY SPECIlED TOPOLOGY IE WITH A
PRESCRIBED VALUE OF A CERTAIN TOPOLOGICAL INVARIANT SUCH AS A CHARACTERISTIC
NUMBER OR THE SIMPLICIAL VOLUME &OR EXAMPLE EACH DECENT TOPOLOGICAL
SPACE 8 IS RATHER CANONICALLY ACCOMPANIED BY METRIC SPACES HOMOTOPY
EQUIVALENT TO IT SUCH AS A SUITABLY SUBDIVIDED SEMISIMPLICIAL MODEL OF 8
HOM
WHICH IS AN INlNITE DIMENSIONAL SIMPLICIAL COMPLEX CALL IT 8a w 8 WHERE
THE METRIC ON 8a COMES FROM A CHOICE OF A STANDARD METRIC ON EACH K
SIMPLEX K 4HIS 8a IS TOO LARGE TO PLEASE OUR EYE BUT IT MAY
CONTAIN SOME DISTINGUISHED SUBVARIETIES EG MINIMIZING HOMOLOGY CLASSES
IN (b 8a (b 8 AT LEAST FOR CERTAIN SPACES 8 SEE ( IN ;'= AND
REFERENCES THEREIN
F !NISOTROPIC METRICS APPEAR UNDER A VARIETY OF NAMES <NON HOLONOM
IC <CONTROL <SUB ELLIPTIC <SUB 2IEMANNIAN <#ARNOT #ARATHEODORY
BEARING THE TRACES OF THEIR ORIGIN 4HEY WERE EXTENSIVELY STUDIED BY ANALYSTS
SINCE (ORMANDERS WORK ON HYPO ELLIPTIC OPERATORS BUT ) DO NOT KNOW WHERE
AND WHEN THEY WERE PROMOTED FROM TECHNICAL DEVICES TO FULL MEMBERSHIP
IN THE METRIC COMMUNITY .OW BESIDES 0$% THEY HELP GROUP THEORISTS
IN THE STUDY OF DISCRETE NILPOTENT GROUPS ` SINCE THE LOCAL GEOMETRY OF
ESPECIALLY SELF SIMILAR METRICS ON NILPOTENT ,IE GROUPS ' z ` ADEQUATELY
REmECTS THE ASYMPTOTIC GEOMETRY OF ` 4HE LOCAL GEOMETRY OF DIST( REDUCES
IN TURN TO THAT OF THE POLARIZATION ( y 4 ' WHICH IS RATHER SOFT AS FAR AS
LOW DIMENSIONAL ( HORIZONTAL IE TANGENT TO ( SUBVARIETIES IN ' ARE CON
6ISIONS IN -ATH 30!#%3 !.$ 15%34)/.3
CERNED AS FOLLOWS FROM THE .ASH IMPLICIT FUNCTION THEOREM 4HIS PROVIDES
SOME INFORMATION ON DIST( MINIMAL SUBVARIETIES IN ' ALAS ONLY THE CASE
OF SURFACES IS UNDERSTOOD AND ALLOWS US TO EVALUATE THE $EHN FUNCTION OF `
SEE ;'= ;'= AND REFERENCES THEREIN !LSO OBSERVE THAT HOWEVER MEAGER
RESULTS AND PROBLEMS IN THE ANISOTROPIC GEOMETRY MAY SERVE AS MODEL FOR
OTHER SOFT SPACES OF SOLUTIONS OF 0$% AS ( HORIZONTALITY FOR (
G )T IS HELPFUL TO THINK OF A METRIC MEASURE SPACE 8 u AS A HIGH
DIMENSIONAL CONlGURATION SPACE OF A PHYSICAL SYSTEM WHICH IS INDEED OF
TEN THE CASE (ERE F 8 9 IS AN OBSERVABLE PROJECTING 8 TO A LOW
DIMENSIONAL <SCREEN OUR 9 SUCH AS THE SPACE TIME 2 FOR INSTANCE WHERE
Fb u IS WHAT WE SEE ON THE SCREEN 4HE CONCENTRATION OF Fb u UBIQUITOUS
IN THE PROBABILITY THEORY AND STATISTICAL MECHANICS WAS BROUGHT TO GEOM
ETRY STARTING FROM "ANACH SPACES BY 6ITALI -ILMAN FOLLOWING THE EARLIER
WORK BY 0AUL ,EVY 4HE ,EVY -ILMAN CONCENTRATION PHENOMENON HAS BEEN
OBSERVED FOR A WIDE CLASS OF EXAMPLES WHERE BESIDES THE MERE CONCENTRA
TION ONE IS CONCERNED WITH LARGE DEVIATIONS AND mUCTUATIONS 3EE ;-I=
;'= AND REFERENCES HEREIN 5NFORTUNATELY OUR DElNITION OF CONCENTRATED
SPACES DOES NOT CAPTURE LARGE DEVIATIONS WHICH ARE MORE FUNDAMENTAL THAN
mUCTUATIONS 0OSSIBLY THIS CAN BE HELPED BY SOMEHOW ENRICHING THE STRUC
TURE "ESIDES ONE CAN PROCEED BY ALLOWING VARIABLE MEASURES STATES u
AS IN THE CUBE F G. WITH THE PRODUCT uP OF . COPIES OF THE P _ P
MEASURE ON F G AND IN 'IBBS MEASURES PARAMETRIZED BY TEMPERATURE
4HEN INSPIRED BY PHYSICS ONE WONDERS WHAT SHOULD BE GENERAL OBJECTS
RESPONSIBLE FOR CONCENTRATION IN QUANTUM STATISTICAL MECHANICS WHERE THE
CONCENTRATION IS LIMITED BY THE (EISENBERG PRINCIPLE AND lNALLY ONE MAY
TURN TO NON METRIC STRUCTURES WHICH MAY COME IN PROBABILITY AND GEOMETRY
ALONG WITH CONCENTRATION
,IFE 7ITHOUT -ETRIC

)T IS TOO EASY TO CONCOCT INVARIANTS OF A METRIC SPACE 6 EG BY LOOKING
AT THE RANGES OF #ARTESIAN POWERS OF THE DISTANCE FUNCTION MAPPING 6 K TO
KK_
2 FOR
DK V VI VK FDISTVI VJ GsIJsK
4HE DIAMETER OF 6 APPEARS FOR EXAMPLE AS THE MAXIMAL SEGMENT IN THE
IMAGE OF DK IN 2 FOR K 4HEN THERE ARE VARIOUS OFTEN POSITIVE
<ENERGY FUNCTIONS % ON SPACES OF SUBVARIETIES 7
2 IN 8 AND MAPS F 6
6 SUCH AS 7 6OLM 7 FOR M DIM 7 AND KDF K DV FOR 2IEMANNIAN
-)3(! '2/-/6 '!&!
6 AND 6 %ACH % GENERATES INVARIANTS OF 6 FOR INSTANCE INlMA OF %

ON GIVEN CLASSES OF 7 S OR F S EG INF 6OLM 7 FOR ;7 = H (M 6
OR MORE GENERALLY THE FULL -ORSE LANDSCAPE OF % INCLUDING THE SPECTRUM
OF THE CRITICAL VALUES OF % EG THE SPECTRUM OF THE ,APLACE OPERATOR ON
FUNCTIONS AND FORM ON A 2IEMANNIAN 6 4HESE INVARIANTS ) PROVIDE US
WITH THE RAW MATERIAL FOR ASKING QUESTIONS AND MAKING CONJECTURES WHAT
ARE POSSIBLE VALUES OF ) )% AND RELATIONS BETWEEN )S FOR DIkERENT %S
7HAT ARE SPACES 6 WITH A GIVEN BEHAVIOUR OF AN ) ETC "UT SPACES WITHOUT
METRICS BECOME RATHER SLIPPERY HARD TO GRASP AND ASSESS *UST LOOK AT A
FOLIATION OR DYNAMICAL SYSTEM & ON A MANIFOLD 6 4HE ESSENTIAL INVARIANTS
ENTROPY ASYMPTOTICS OF PERIODIC ORBITS OFTEN CHANGE DISCONTINUOUSLY UNDER
DEFORMATIONS PERTURBATIONS OF THE STRUCTURE AND ARE HARD TO EVALUATE EVEN
APPROXIMATELY FOR A GIVEN & !ND IT IS NOT EASY AT ALL TO COME UP WITH NEW
NUMERICAL INVARIANTS MAKING SENSE FOR ALL OBJECTS IN THE CATEGORY 4HIS IS DUE
TO THE FACT THAT THE LOCAL GROUP OF APPROXIMATE AUTOMORPHISMS OF SUCH A
STRUCTURE AT A POINT V 6 IS POTENTIALLY NON COMPACT #ONSEQUENTLY THE
ACTION OF $Ik6 ON THE SPACE & OF OUR STRUCTURES & ON 6 MAY HAVE NON
TRIVIAL DYNAMICS EG NON COMPACT ISOTROPY GROUPS )S& !UT6 & &
& MAKING THE QUOTIENT SPACE & $Ik 6 WHERE OUR INVARIANTS ARE SUPPOSED
TO LIVE NON SEPARABLE )NTUITIVELY INVARIANTS SHOULD BE INDEPENDENT OF
OBSERVERS ATTACHED TO DIkERENT COORDINATE SYSTEMS IN 6 IF THERE ARE NON
COMPACTLY MANY EQUIVALENT OBSERVERS IT BECOMES DInCULT TO RECONCILE THEIR
VIEWS AS IS IN THE SPECIAL AND GENERAL RELATIVITY FOR EXAMPLE
.OW WE GLANCE AT A COUPLE OF ( STRUCTURES FOR INTERESTING NON COMPACT
SUBGROUPS ( y ',N 2 WHERE THE COMPACT CASE OF ( /N CORRESPONDS
TO THE 2IEMANNIAN GEOMETRY
# STRUCTURES 4HESE CUSTOMARILY CALLED ALMOST COMPLEX STRUCTURES * ON

6 ARE lELDS OF # LINEAR STRUCTURES *V IN 4V 6 V 6 3UCH A * MAY BE
EXPRESSED
P BY AN ANTI INVOLUTION ALSO CALLED * 4 6 4 6 CORRESPOND
ING TO _ WHERE THE PERTINENT ( IS NON COMPACT ',M # y ',N 2
FOR N M -ORPHISMS CALLED # MAPS F 6 6 WHERE THE DIkEREN
TIAL $F 4 6 4 6 IS # LINEAR IE COMMUTE WITH * ARE RARE FOR
NON INTEGRABLE 6 AND 6 SINCE THE CORRESPONDING ELLIPTIC 0$% SYSTEM
IS OVER DETERMINED FOR DIM2 6 3O WE STICK TO # CURVES MAPS OF 2IE
MANN SURFACES 3 6 6 * ALSO CALLED * CURVES IF THIS MATTERS 7E
MARK EACH 3 WITH A POINT S 3 AND THEN THE TOTALITY OF # CURVES MAKES
A HUGE SPACE 3 36 * FOLIATED INTO SURFACES WHERE EACH LEAF IN 3 IS
REPRESENTED BY A lXED 3 6 WITH VARIABLE MARKING S 3
6ISIONS IN -ATH 30!#%3 !.$ 15%34)/.3
7HAT IS THE POSSIBLE GLOBAL GEOMETRY EG DYNAMICS OF 3 AND HOW CAN
IT BE READ FROM *
3TART WITH THE SUBFOLIATION # y 3 OF CLOSED LEAVES IN 3 CORRESPONDING TO
CLOSED 2IEMANN SURFACES 3 6 AND TRY TO MIMIC THE GEOMETRY OF CURVES
IN COMPLEX ALGEBRAIC lRST OF ALL PROJECTIVE VARIETIES 6 4HIS # IS lLTERED
BY THE DEGREES D OF 3 y 6 PLAYING THE ROLE OF PERIODS OF CLOSED ORBITS SAY
FOR ACTIONS OF : ON SOME SPACE WHERE CURVES OF DEGREE D MAY DEGENERATE TO
SEVERAL CURVES OF LOWER DEGREE THUS COMPACTIFYING EACH MODULI SPACE #D
BY LOW DIMENSIONAL STRATA BUILT OF #I WITH I D #ONVERSELY ONE CAN OFTEN
FUSE LOWER DEGREE CURVES TO HIGHER DEGREE D BY DEFORMING THEIR REDUCIBLE
UNIONS TO IRREDUCIBLE 3 #D
7HAT HAPPENS TO THIS WEB OF ALGEBRAIC CURVES OF DEGREE D WHEN WE SLIGHTLY
PERTURB THE UNDERLYING *
4HE ANSWER DEPENDS ON THE VIRTUAL DIMENSION OF THE SPACE #D IE THE
&REDHOLM INDEX OF THE ELLIPTIC SYSTEM DElNING 3 #D &OR EXAMPLE THE
CURVES IN ABELIAN VARIETIES #M c ARE UNSTABLE UNDER EVEN INTEGRABLE DE
FORMATIONS OF * BUT THE CURVES IN CERTAIN 6 S WITH SUnCIENTLY AMPLE
ANTI CANONICAL BUNDLES SUCH AS #0 M REMAIN ESSENTIALLY INTACT UNDER SMALL
AND LARGE AS WE SHALL SEE BELOW DEFORMATIONS OF * YET THE SHAPE AND
POSITION OF #D IN 3 MAY BE A PRIORI GREATLY DISTORTED BY DEFORMATIONS
OF * FOR D LARGE COMPARED TO _ !S WE FOLLOW A DEFORMATION *T MOV
ING IT FURTHER AWAY FROM THE ORIGINAL * A CURVE 3 y 6 PERSISTENT FOR
SMALL T MAY EVENTUALLY PERISH BY BLOWING UP TO SOMETHING NON COMPACT AS
T REACHES SOME CRITICAL VALUE TC 7HAT IS NEEDED TO KEEP 3 ALIVE AS A CLOSED
*T CURVE IS AN A PRIORI BOUND ON THE AREA OF 3 6 * *T MEASURED
WITH SOME BACKGROUND 2IEMANNIAN METRIC G IN 6 WHERE A SPECIlC CHOICE
OF G IS NOT IMPORTANT AS WE DEAL HERE WITH COMPACT 6 3UCH A BOUND IS
GUARANTEED BY THE FOLLOWING TAMENESS ASSUMPTION OF * WHICH LIMITS THE AREA
OF EVEN APPROXIMATELY * HOLOMORPHIC CURVES 3 BY THEIR TOPOLOGY NAMELY
BY ;3= ( 6
#ALL * TAMED BY A CLOSED FORM ON 6 IF IS POSITIVE ON ALL * CURVES IN
6 IE { *{ FOR ALL NON ZERO VECTORS { 4 6 )F SO * IS TAME WITH
RESPECT TO THE COHOMOLOGY CLASS H ;= ( 6 THE AREA OF EACH CLOSED
ORIENTED <APPROXIMATELY * HOLOMORPHIC CURVE 3 y 6 IS BOUNDED IN TERMS
OF THE VALUE H;3= 2 4O MAKE IT PRECISE DENOTE BY 3 y 3 THE SET OF POINTS
S 3 y 6 WHERE THE PLANE 4S 3 y 43 6 IS CLOSE TO A # LINE FOR A lXED
BACKGROUND METRIC 4HEN <TAME MEANS THE EXISTENCE OF m # SUCH
-)3(! '2/-/6 '!&!
THAT THE INEQUALITY AREA 3 t _ m AREA 3 IMPLIES H3 t # _ AREA 3 FOR

ALL CLOSED ORIENTED SURFACES 3 y 6 #LEARLY < TAME <H TAME BUT THE
CONVERSE H REMAINS QUESTIONABLE /NE MAY TRY THE (AHN "ANACH
THEOREM ESPECIALLY FOR DIM 6
)F 6 * IS A +AHLER EG ALGEBRAIC MANIFOLD THEN * IS TAMED BY
THE SYMPLECTIC +AHLER FORM AND AS FAR AS *T REMAINS ;W= TAME WE HAVE
DECENT MODULI SPACES OF *T HOLOMORPHIC CURVES IN 6 * WHERE * *T
MAY BE QUITE FAR FROM * EG * ! * FOR AN ARBITRARY SYMPLECTIC AU
TOMORPHISM 4 6 4 6 &OR EXAMPLE IF WE START WITH THE STANDARD
#0 M * AND * * IS JOINED WITH * BY A HOMOTOPY OF ; = TAME STRUC
TURE *T FOR THE STANDARD SYMPLECTIC FORM ON #0 M THEN #0 M *
ADMITS A RATIONAL * CURVE 3 OF DEGREE THROUGH EACH PAIR OF POINTS WHERE
MOREOVER 3 IS UNIQUE FOR M 4HIS REMAINS TRUE FOR ALL TAME STRUC
TURES ON #0 WITH NO A PRIORI ASSUMPTION BY THE WORK OF 4AUBS
AND $ONALDSON "UT WHAT HAPPENS TO CLOSED # CURVES AT THE lRST MOMENT
TC WHEN *T LOSES TAMENESS 7HAT KIND OF SUBFOLIATION 3D y 3 IS FORMED BY
THE LIMITS OF 3 #D FOR T TC )T SEEMS AT LEAST FOR DIM 6 EG FOR
#0 AND 3 a3 THAT MOST OF THE CLOSED # CURVES BLOW UP SIMULTANEOUSLY
FORMING A REGULAR FOLIATED LIKE STRUCTURE IN 6 4HIS IS REMINISCENT OF HOW
+LEINIAN GROUPS DEGENERATE REMAINING DISCRETE AND BEAUTIFUL AT THE VERGE
OF EXTINCTION
!RE THERE NON TAME 6 * WITH RICH MODULI SPACES OF CLOSED ESPECIALLY
RATIONAL CURVES SAY HAVING SUCH A CURVE PASSING THROUGH EACH PAIR OF POINTS
IN 6 )F A DIMENSIONAL 6 * HAS MANY * CURVES IT IS TAME BY AN EASY
ARGUMENT /N THE OTHER HAND THE MAJORITY OF HIGHER DIMENSIONAL 6 *
CONTAIN ISOLATED POCKETS OF * CURVES WITH RATHER SHAPELESS AND USELESS #D
LIKE CLOSED GEODESICS IN MOST 2IEMANNIAN MANIFOLDS LOST IN ACCIDENTAL WELLS
OF ENERGY
4URNING TO NON CLOSED # CURVES WE lND A PREREQUISITE FOR THE .EVAN
LINNA THEORY AS THEY SHARE THE PRINCIPLE SYMBOL OF WITH ORDINARY HOLO
MORPHIC FUNCTIONS AND MAPS 4HIS IS ALSO CRUCIAL FOR THE STUDY OF CLOSED
# CURVES &OR EXAMPLE WE CAN DElNE HYPERBOLIC 6 * WHICH RECEIVE NO
NON CONSTANT * MAPS # 6 AND THESE 6 WE ASSUME COMPACTNESS CARRY
A NON DEGENERATE +OBAYASHI METRIC IE THE SUPREMAL METRIC FOR WHICH THE
# MAPS ( 6 ARE ,IPSCHITZ 4HIS HYPERBOLICITY HAS A POINT IN COM
MON WITH TAMENESS THE SPACE OF # MAPS F 3 6 WITH F RANGING
IN A COMPACT SET IS COMPACT FOR HYPERBOLIC 6 4HIS IS ALSO IMPLIED BY
THE <TAME BOUND AREA 3 s CONST;3= FOR APPROXIMATELY * HOLOMORPHIC
6ISIONS IN -ATH 30!#%3 !.$ 15%34)/.3
SPHERES IN 6 PROVIDED THERE IS NO * HOLOMORPHIC SPHERES IN 6 #ONSE

QUENTLY FOR EACH # STRUCTURE ON 7 6 a 3 COMPATIBLE WITH * ON THE
lBERS 6 a S S 3 THERE IS A RATIONAL IE SPHERICAL # CURVE IN 7 PASS
ING THROUGH A GIVEN POINT W 7 THAT CONTRACTIBLY PROJECTS TO 7 4HIS
REMAINS VALID FOR IRRATIONAL CURVES 3 IF THE 4EICHM ULLER SPACE OF 3 IS IN
CORPORATED INTO 7 !NOTHER LINK BETWEEN <TAME AND <HYPERBOLIC IS
EXPRESSED BY THE FOLLOWING EASY TO PROVE TOPOLOGICAL CRITERION FOR HYPER
BOLICITY ,ET 6E BE A 'ALOIS COVERING OF 6 AND @^ BE A FORM WITH SUBLINEAR
^ V K DIST^
GROWTH K@^ V V^ FOR V^ AND WITH INVARIANT UNDER THE
DECK TRANSFORMATION GROUP DIkERENTIAL W ^ )F * IS TAMED BY THE CORRE
^ D@
SPONDING CLASS ;W= ( 6 THEN 6 * IS HYPERBOLIC &OR EXAMPLE IF * IS
TAME x 6 AND x 6 IS HYPERBOLIC THEN 6 * IS HYPERBOLIC
!RE THERE FURTHER TOPOLOGICAL CRITERIA FOR NON HYPERBOLICITY WHERE
x 6 IS NOT SO LARGE &OR EXAMPLE ARE THERE TAME HYPERBOLIC STRUCTURES
ON THE TORUS 0ARABOLIC CURVES IE # MAPS INTO SUCH TORUS 4 W *
COULD HELP US TO STUDY WHERE THE ULTIMATE GOAL IS TO SHOW THAT EVERY
SYMPLECTIC STRUCTURE ON 4 IS ISOMORPHIC TO THE STANDARD ONE AS IS KNOWN
FOR #0 FOR INSTANCE BY THE WORK OF 4AUBS AND $ONALDSON
7HAT ARE ESSENTIAL METRIC PROPERTIES OF 6 DIST+OB FOR HYPERBOLIC 6
.OTHING IS KNOWN ABOUT IT (OW MUCH CAN ONE DEFORM A HYPERBOLIC *
KEEPING HYPERBOLICITY )T IS CLEAR BY "RODYS ARGUMENT THAT SMALL PERTUR
BATIONS DO NOT HURT 4AKE FOR INSTANCE A NECESSARILY SINGULAR * CURVE 3
IN AN TAME #0 * WHERE THE FUNDAMENTAL GROUP OF THE COMPLEMENT IS
HYPERBOLIC EG FREE NON CYCLIC WHICH IS THE CASE FOR 3 CONSISTING OF D
RATIONAL CURVES OF DEGREE WHERE EXACTLY D OF THEM MEET AT A SINGLE POINT
AND D t 4HEN EVERY NON CONSTANT * MAP # #0 * MEETS 3 AND IF
WE SLIGHTLY DEFORM 3 THE COMPLEMENT OF THE RESULTING COMPLEMENT #0 N3
REMAINS HYPERBOLIC PROVIDED EACH IRREDUCIBLE COMPONENT OF 3 MINUS THE
REMAINING COMPONENTS WERE HYPERBOLIC AS HAPPENS IF THE ABOVE D LINES
ARE AUGMENTED BY ANOTHER ONE MEETING EACH AT A SINGLE POINT 4HUS FOR
EVERY D t THERE IS A NON EMPTY OPEN SUBSET (D y #D OF SMOOTH * CURVES
3 IN #0 * OF DEGREE D WHERE EVERY NON CONSTANT * MAP # #0 MEETS
3 0ROBABLY THIS (D IS DENSE AS WELL AS OPEN FOR LARGE D POSSIBLY DEPEND
ING ON * WHERE THE CASE OF CURVES WITH MANY DEPENDING ON * IRREDUCIBLE
COMPONENTS SEEMS WITHIN REACH ! SIMILAR OBSERVATION CAN BE MADE FOR
OTHER COMPACT 6 AND DIkERENCES OF THESE EG FOR THOSE ASSOCIATED TO TORI
BUT NOWHERE DOES ONE COME CLOSE TO WHAT IS KNOWN IN THE CLASSICAL ALGEBRAIC
CASE (ERE IS ANOTHER KIND OF TEST QUESTION WITH NO CLASSICAL COUNTERPART
-)3(! '2/-/6 '!&!
'IVEN A HYPERBOLIC 6 * WHEN AND HOW CAN ONE MODIFY * INSIDE A ARBI
TRARILY SMALL NEIGHBOURHOOD 5 y 6 SUCH THAT THE RESULTING 6 * ADMITS
A PARABOLIC OR EVEN RATIONAL CURVE THROUGH EACH POINT V 6
)F 6 * IS HYPERBOLIC THEN THE SPACE ( OF * MAPS ( 6 IS COMPACT
AND PROJECTS ONTO 3 WITH CIRCLE lBERS 4HE GROUP ' 0 3, 2 )SO (
NATURALLY ACTS ON ( AND PERIODIC IE COMPACT ORBITS CORRESPOND TO CLOSED
* CURVES 3 y 6 WHERE GENUS 3 t CONST!REA 3 )F 6 * IS ALGEBRAIC
THEN PERIODIC ORBITS ARE DENSE IN ( BESIDES THERE ARE MANY lNITE DI
MENSIONAL INVARIANT SUBSETS IN ( CORRESPONDING TO SOLUTIONS OF ALGEBRAIC
/$% OVER 6 !RE lNITE DIMENSIONAL INVARIANT SUBSETS DENSE IN ( FOR ALL
TAME 6 7HEN DOES 6 CONTAIN A METRICALLY COMPLETE * CURVE 3 y 6
PREFERABLY WITH LOCALLY BOUNDED GEOMETRY 7HAT ARE IF ANY INVARIANT MEA
SURES IN ( OUT OF ( FOR TAME * IN THE ABSENCE OF CLOSED * CURVES
4AKE FOR EXAMPLE AN ALGEBRAIC 6 * WITH A NICE FOLIATION EG A mAT
CONNECTION OVER A 2IEMANN SURFACE OR THE STANDARD FOLIATION ON A (ILBERT
MODULAR SURFACE (OW DOES SUCH A FOLIATION FARE UNDER TAME HOMOTOPIES *T
)F 6 * IS NON HYPERBOLIC THEN THE MAIN ISSUE IS TO UNDERSTAND THE
SPACE 0 y 3 OF PARABOLIC LEAVES OR EQUIVALENTLY THE SPACE 0 OF NON
CONSTANT * MAPS # 6 WITH THE ACTION OF ' !k # /NE KNOWS FOR
THE STANDARD #0 N THAT RATIONAL MAPS ARE DENSE IN 0 AND HENCE IN (
BY THE 2UNGE THEOREM 4HIS PROBABLY IS TRUE FOR ALL TAME 6 * WITH
<SUnCIENTLY MANY RATIONAL CURVES BUT THIS SEEMS UNKNOWN EVEN FOR RATIO
NALLY CONNECTED ALGEBRAIC 6 ,ET FOR INSTANCE 6 #0 N * WHERE * IS
TAMED BY THE STANDARD 4HEN BY FUSING A SEQUENCE OF RATIONAL CURVES
ONE CAN OBTAIN A PARABOLIC ONE CONTAINING A GIVEN COUNTABLE SUBSET IN 6
-OST LIKELY ONE CAN PROVE THE 2UNGE THEOREMS FOR 6 IN THIS WAY FOLLOWING
$ONALDSONS APPROACH TO 2UNGE FOR 9ANG -ILLS
$ENOTE BY " y 0 THE SPACE OF NON CONSTANT ,IPSCHITZ # MAPS # 6
AND OBSERVE THAT THE CLOSURE OF EVERY ' ORBIT IN 0 MEETS " BY THE "LOCH
"RODY PRINCIPLE )F 6 CONTAINS MANY RATIONAL CURVES THEN " IS QUITE LARGE
EG IF 6 #0 N *STAND THEN THERE IS A MAP F " INTERPOLATING FROM AN
ARBITRARY m SEPARATED SUBSET a y # PROBABLY THIS REMAINS TRUE FOR ALL *
TAMELY HOMOTOPIC TO THE STANDARD ONE AS WELL AS FOR MORE GENERAL <RATIONALLY
CONNECTED 6 #ONVERSELY A STRENGTHENED VERSION OF THIS INTERPOLATION
PROPERTY IS LIKELY TO IMPLY THE EXISTENCE OF MANY RATIONAL CURVES IN 6
!PART FROM RATIONAL CURVES ONE CAN SOMETIMES PRODUCE # MAPS # 6
BY PRESCRIBING SOME ASYMPTOTIC BOUNDARY CONDITIONS EG BY PROVING NON
6ISIONS IN -ATH 30!#%3 !.$ 15%34)/.3
COMPACTNESS OF THE SPACE OF * DISKS WITH BOUNDARIES ON A NON COMPACT

FAMILY OF ,AGRANGIAN SUBVARIETIES 4HIS WORKS FOR MANY * ON THE STANDARD
SYMPLECTIC 2M AND YIELDS FOR INSTANCE PARABOLIC CURVES FOR * ON TORI
TAMED BY THE STANDARD (OW LARGE IS " FOR THESE TORI (OPEFULLY THE
MEAN ORDINARY DIMENSION OF " FOR THE ACTION OF # y ' ON " IS lNITE
AND THE NATURAL MAP FROM " TO THE PROJECTIVIZED ( 6 IS NON AMBIGUOUS
AND SOMEHOW REPRESENTS THE HOMOLOGY CLASS OF #0 M_ CORRESPONDING TO
# LINES IN ( 6 #M 2M
0SEUDO 2IEMANNIAN MANIFOLDS 'IVEN AN ( STRUCTURE G ON A MANIFOLD
6 FOR A NON COMPACT GROUP ( y ',N 2 N DIM 6 ONE MAY RIGIDIFY
6 G BY REDUCING ( TO A MAXIMAL COMPACT SUBGROUP + y ( IE BY
CONSIDERING A 2IEMANNIAN METRIC G+ ON 6 COMPATIBLE WITH G &OR EXAMPLE
IF G * IS A # STRUCTURE THEN + 5 M y ',M # M N AND
G+ IS AN (ERMITIAN METRIC IF G IS PSEUDO 2IEMANNIAN IE A QUADRATIC
DIkERENTIAL FORM ON 6 OF TYPE P Q WITH P Q N THEN + /P a
/Q y /P Q AND G+ IS A 2IEMANNIAN METRIC SUCH THAT THERE IS A FRAME
{ {P {_ {Q_ AT EACH POINT WHERE BOTH FORMS G AND G+ BECOME
DIAGONAL WITH G+ {I {I G{I {I AND G+ {J_ {J_ _G{J_ {J_
7HAT PROPERTIES INVARIANTS OF G CAN BE SEEN IN AN INDIVIDUAL G+ ANDOR
IN THE TOTALITY '+ '+ G OF ALL G+ )T MAY HAPPEN THAT TWO STRUCTURES G
AND G ARE VIRTUALLY INDISTINGUISHABLE IN THESE TERMS NAMELY WHEN FOR EACH
THERE EXIST G+ AND G+ WHICH ARE BI ,IPSCHITZ EQUIVALENT 4HE
$Ik ORBIT OF '+ MIGHT BE # DENSE IN THE SPACE OF ALL 2IEMANNIAN METRICS
UNDER THE WORST SCENARIO 4HERE ARE FEW KNOWN CASES WHERE '+ TELLS YOU
SOMETHING USEFUL ABOUT G !N EXCEPTIONALLY PLEASANT EXAMPLE IS GIVEN BY
CONFORMAL STRUCTURES G WITH '+ G TELLING YOU EVERYTHING ABOUT G IE THE
# CLOSURES OF THE $Ik ORBITS OF '+ AND '+ ARE ESSENTIALLY DISJOINT UNLESS

G AND G ARE ISOMORPHIC FURTHERMORE THE ACTION OF $Ik 6 ON THE SPACE
OF CONFORMAL STRUCTURES IS PROPER APART FROM THE STANDARD CONFORMAL 3 N
.EXT IF ( PRESERVES AN EXTERIOR R FORM ON 2N AND THE CORRESPONDING FORM
j jG ON 6 IS CLOSED THEN G+ IS BOUNDED FROM BELOW BY VOLUMES OF R
CYCLES # y 6 WITH j# AND AT LEAST $Ik '+ IS NOT DENSE &INALLY FOR
CERTAIN SYMPLECTIC STRUCTURES G THE MINIMAL G+ AREAS OF SOME CYCLES
# REALIZABLE BY # CURVES EQUAL # FOR ALL G+ THUS LIMITING GEOMETRY OF
6 G+ IN A MORE SIGNIlCANT WAY
(AVING FAILED WITH ROBUST ,IPSCHITZ GEOMETRY OF G+ ONE RESORTS TO
CURVATURE AND LOOKS AT THE SUBSETS '+ s y '+ OF ADAPTED METRICS G+
G+ G WITH SOME NORM OF CURVATURE BOUNDED BY s 2 (ERE THE MERE
-)3(! '2/-/6 '!&!
FACT OF '+ s BEING EMPTY FOR A GIVEN s GIVES ONE A NON TRIVIAL COMPLEXITY
BOUND ON G &OR EXAMPLE ONE MAY STUDY THE INlMAL s 2 SUCH THAT G
ADMITS G+ WITH THE SECTIONAL CURVATURES BETWEEN _s AND s WITH NORMALIZED
VOLUME IF ( y 3,N 2 !ND AGAIN ONE IS TEMPTED TO SEARCH FOR THE <BEST
G+ ADAPTED TO G WITH NON ZERO EXPECTATION FOR <MILDLY NON COMPACT (
INSPIRED BY CONFORMAL METRICS OF CONSTANT SCALAR CURVATURE DELIVERED BY THE
3CHOEN !UBIN SOLUTION TO THE 9AMABE PROBLEM
.OW WE TURN TO THE CASE AT HAND PSEUDO 2IEMANNIAN G OF TYPE P Q ON
A TYPICALLY COMPACT MANIFOLD 6 HAVING CLOSER KINSHIP TO 2IEMANNIAN GE
OMETRY THAN GENERAL G 7HAT ARE MOST GENERAL MORPHISMS IN THE PSEUDO
2IEMANNIAN CATEGORY COMPARABLE TO <,IPSCHITZ FOR 2IEMANNIAN MANIFOLDS
3INCE WE CAN COMPARE METRICS ON A lXED MANIFOLD BY G s G FOR G _ G
BEING POSITIVE SEMIDElNITE WE MAY SPEAK OF LONG MAPS F 6 6
WHERE F b G t G (OWEVER UNLIKE THE 2IEMANNIAN CASE WHERE SHORT RATHER
THAN LONG MAPS ARE USEFUL FOR ALL 6 AND 6 THIS MAKES SENSE ONLY FOR P Q
NOT BEING TOO SMALL COMPARED TO P Q IF P P THEN EVERY ISOMETRIC
IMMERSION F 6 6 CAN BE A LITTLE # PERTURBED TO SOME F WITH F b G
BEING AS LARGE AS YOU WANT ACTUALLY EQUAL TO A GIVEN G Fb G HOMOTOPIC
TO G BY THE .ASH +UIPER ARGUMENT AND IF ALSO Q Q EVERYTHING BECOMES
SOFT AND ONE GETS ALL METRICS ON 6 HOMOTOPIC TO G BY # IMMERSIONS
ARBITRARILY # CLOSE TO F 4HUS WE STICK TO P P AND START WITH POSITIVE
SLICES IN 6 IE IMMERSED P DIMENSIONAL 7 y 6 WITH GJ7 )F 7
IS CONNECTED WITH NON EMPTY BOUNDARY WE SET 27 W DISTW 7
FOR THE INDUCED 2IEMANNIAN METRIC SUP 2W 7 IS CALLED <IN RADIUS OF
W
7 AND DElNE 2 6 V AS THE SUPREMUM OF 2V 7 FOR ALL POSITIVE SLICES
THROUGH V #ONNECTED 2IEMANNIAN MANIFOLDS WITH NON EMPTY BOUNDARIES
HAVE 2 2 AT ALL THEIR POINTS WHILE CLOSED MANIFOLDS HAVE 2
IF AND ONLY IF x "UT EVEN FOR x FEG THERE IS NO EkECTIVE BOUND ON

2 DUE TO NON DECIDABILITY OF x FEG -OREOVER THERE ARE RATHER SMALL
METRICS ON 3 WITH ALMOST NEGATIVE CURVATURE AND THUS WITH ARBITRARILY
LARGE 2 .EXT WE DUALIZE BY TAKING LONG MAPS F FROM 6 V TO
2IEMANNIAN 7 W OF DIMENSION P AND DENOTE BY 2 6 V THE INlMUM OF
2 7 W OVER ALL POSSIBLE 7 W AND F WHERE CLEARLY 2 t 2
! LONG MAP 6 7 IS NECESSARILY A NEGATIVE SUBMERSION IE WITH
G NEGATIVE lBERS #ONVERSELY GIVEN A NEGATIVE SUBMERSION F PROPER IF 6
IS NON COMPACT OF 6 TO A SMOOTH MANIFOLD 7 ONE CAN lND A 2IEMANNIAN
METRIC ON 7 MAKING F LONG 4HERE IS A UNIQUE SUPREMAL &INSLER METRIC ON
6ISIONS IN -ATH 30!#%3 !.$ 15%34)/.3
7 MAKING F LONG WHICH CAN BE MINORIZED BY A 2IEMANNIAN ONE 4HUS

FOR EXAMPLE IF A COMPACT POSSIBLY WITH A BOUNDARY 6 ADMITS A NEGATIVE
SUBMERSION INTO A CONNECTED MANIFOLD WITH lNITE x OR WITH NON EMPTY
BOUNDARY THEN 2 6
-ANIFOLDS WITH 2 6 V s CONST FOR ALL V 6 ARE KIND OF HYPER
BOLIC IN THE DIRECTIONS EG THIS CONDITION IS # STABLE )F P AND
6 IS CLOSED THEN IT ALWAYS HAS A CIRCULAR POSITIVE SLICE AND 2 6 V
FOR ALL V 6 BUT IF P t THEN EVERY G G ADMITS A DEFORMATION GT WITH
2 FOR T TAKE A GENERIC AND THUS NON INTEGRABLE P PLANE lELD
3 y 4 6 WITH GJ3 AND MAKE GT G TG_ WHERE G_ IS THE NEGATIVE
PART OF G FOR THE NORMAL SPLITTING 4 6 3 g3 4HIS WORKS EVEN LOCALLY
AND SHOWS THAT THE MAJORITY OF G HAVE 2 SMALL IN SOME REGIONS OF 6 AND
INlNITE IN OTHER REGIONS
)T SEEMS BY FAR MORE RESTRICTIVE TO REQUIRE THAT BOTH 2 AND 2_ IE 2
FOR _6 ARE BOUNDED ON 6 WHERE A SUnCIENT CONDITION FOR COMPACT 6 IS THE
EXISTENCE OF A NEGATIVE SUBMERSION 6 7 AS WELL AS A POSITIVE SUBMERSION
6 7_ WHERE 7 AND 7_ HAVE x OR NON EMPTY BOUNDARIES 4HIS
CAN BE SLIGHTLY GENERALIZED BY ALLOWING SOMEWHAT MORE GENERAL PAIRS OF e
FOLIATIONS WITH UNIFORMLY COMPACT lBERS EG COMING FROM SUBMERSION TO
SIMPLY CONNECTED OR TO BOUNDED ORBIFOLDS AND SOMETIMES ONE SUBMERSION
SUnCES &OR EXAMPLE START WITH A NEGATIVE SUBMERSION F 6 7 WHERE
7 IS SIMPLY CONNECTED OR BOUNDED AND THE NEGATIVE lBERS OF F ARE
ALSO SIMPLY CONNECTED OR BOUNDED $EFORM THE ORIGINAL G ON 6 TO G
AGREEING WITH G ON THE lBERS WHILE BEING VERY POSITIVE NORMALLY TO THE
lBERS 4HEN ALL NEGATIVE SLICES IN 6 G KEEP # CLOSE TO THE lBERS OF F
AND SO 2_ 6 G IS BOUNDED AS WELL AS 2 0ROBABLY THERE ARE MORE
SOPHISTICATED SAY CLOSED MANIFOLDS 6 WHERE THE BOUND ON 2e 6 V COMES
FROM DIkERENT SOURCES AT DIkERENT V 6
$ESPITE SOME HYPERBOLIC FEATURES e BOUNDED PSEUDO 2IEMANNIAN MET
RICS ON CLOSED MANIFOLDS ARE REMINISCENT OF POSITIVE CURVATURE EG THEY ARE
ACCOMPANIED IN KNOWN EXAMPLES BY CLOSED POSITIVE OR NEGATIVE SLICES WITH
x ARE THESE INEVITABLE AND SEEM HARD TO MAKE ON ASPHERICAL 6
"ESIDES TAMING G BY e FOLIATIONS ONE MAY TRY PAIRS OF DIkERENTIAL
FORMS ON 6 WHERE A CLOSED P FORM IS SAID TO BE STRICTLY TAME G
IF IT IS STRICTLY NON SINGULAR ON THE POSITIVE SLICES &OR EXAMPLE IF 6 IS
METRICALLY SPLIT 6 6 a 6_ THE PULL BACK OF THE VOLUME FORM OF 6
STRICTLY TAMES G SIMILARLY _ STRICTLY _ TAMES G WHILE _ IS
-)3(! '2/-/6 '!&!
e TAMING ALBEIT NON STRICTLY (OW MUCH DO CLOSED POSITIVE SLICES PERSIST
UNDER STRICTLY TAME HOMOTOPIES OF G )N PARTICULAR WHAT HAPPENS TO
AREA MAXIMIZING CLOSED POSITIVE SLICES IN Q MANIFOLDS UNDER STRICTLY
TAME HOMOTOPIES
4HE TANGENT BUNDLE TO THE SPACE 3_ OF NEGATIVE SLICES IN 6 CARRIES A
NATURAL POSITIVE ,R NORM WE USE R AS P IS OCCUPIED SINCE G IS POSITIVE
NORMALLY TO 3 y 6 FOR ALL 3 3_ AS WELL AS FOR ALL 3P Q y 6 WITH s
P P "UT THE ASSOCIATED PATH METRIC IN 3_ MAY A PRIORI DEGENERATE
AND EVEN BECOME EVERYWHERE ZERO 9ET THERE ARE SOME POSITIVE SIGNS
,ET 6 METRICALLY SPLIT 6 ; = a 6_ FOR 6_ CLOSED 4HEN THE ,
DISTANCE BETWEEN a 6_ AND a 6_ EQUALS THE VOLUME OF 6 AND HENCE
#ONSEQUENTLY ALL ,R DISTANCES ARE
,ET 6 6 a 6_ WITH CLOSED 6_ AND ARBITRARY 2IEMANNIAN 6 4HEN
a 6 D DIST V V SINCE THE PROJECTION
DIST, V a 6_ V _ 6
6 6_ TOGETHER WITH A ,IPSCHITZ RETRACTION

6 ; D= ;V V =
GIVE US A SHORT IE _ LONG MAP 6 ; D= a 6_ AND THE ABOVE
APPLIES ONLY TO , NOT NECESSARILY TO OTHER ,R
,ET 6 lBER OVER A COMPACT P MANIFOLD WITH CLOSED NEGATIVE lBERS 4HEN
THE , DISTANCE IS BETWEEN EVERY TWO DISTINCT lBERS ESSENTIALLY BY THE
SAME ARGUMENT ALSO YIELDING POSITIVITY OF , DISTANCES BETWEEN GENERAL
CLOSED NEGATIVE SLICES ISOTOPIC TO THE lBERS
(OW MUCH DOES THE METRIC GEOMETRY OF 3_ DIST, AND OF 3 CAPTURE
THE STRUCTURE OF G (OW DEGENERATE CAN DIST, BE FOR GENERAL 6 G 4HIS
DIST, IS VAGUELY SIMILAR TO (OFERS METRIC IN THE SPACE OF ,AGRANGIAN SLICES
IN SYMPLECTIC 6 WHICH ALSO SUGGESTS VANISHING OF OUR DIST,R R FOR
MOST 6
!S WE MENTIONED EARLIER A GENERAL COMPLICATION IN THE STUDY OF (
STRUCTURES G WITH NON COMPACT ( y ',N 2 IS A POSSIBLE NON STABILITY OR
RECURRENCE OF G DUE TO CERTAIN UNBOUNDEDNESS OF THE SET OF DIkEOMORPHISMS
OF 6 MOVING G OR A SMALL PERTURBATION OF G CLOSE TO G 4HE SIMPLEST
MANIFESTATION OF THAT IS NON COMPACTNESS OF THE AUTOMORPHISMS ISOMETRY
GROUP OF 6 G WHICH MAY HAVE DIkERENT NATURE FOR DIkERENT STRUCTURES G
&OR EXAMPLE NON COMPACTNESS OF THE CONFORMAL TRANSFORMATIONS F OF 3 N
IS SEEN IN THE GRAPHS `F y 3 N a 3 N AS DEGENERATION OF THESE TO UNIONS
OF TWO lBERS S a 3 N ; 3 N a S WITH AN UNIFORM BOUND ON 6OL `F FOR
6ISIONS IN -ATH 30!#%3 !.$ 15%34)/.3
F /N THE OTHER HAND GRAPHS OF ISOMETRICS F OF P Q MANIFOLDS 6

ARE REPRESENTED BY TOTALLY GEODESIC ISOTROPIC WHERE THE METRIC VANISHES
N MANIFOLDS `F y 6 a _6 AND THEIR VOLUMES AS WELL AS IN RADII BOTH
MEASURED WITH RESPECT TO SOME BACKGROUND 2IEMANNIAN METRIC IN 6 a 6
GO TO INlNITY FOR F WHILE THEIR LOCAL GEOMETRY REMAINS BOUNDED
UNLIKE THE CONFORMAL CASE 7ITH THIS IN MIND WE CALL G STABLE IF IT ADMITS
A # NEIGHBOURHOOD 5 IN THE SPACE ' OF ALL GS SUCH THAT THE GRAPHS OF
ISOMETRICS F 6 G 6 G WITH G G 5 HAVE 6OLN `F s CONST
WHERE THE BACKGROUND METRIC IS NOT ESSENTIAL AS WE ASSUME 6 COMPACT
%XAMPLE 3TART WITH A METRICALLY SPLIT 6 6 a 6_ 4HE ISOMETRICS
HERE ARE ESSENTIALLY THE SAME AS THOSE OF THE 2IEMANNIAN MANIFOLD 6
6 a _6_ SINCE 6 a _6 6 a 6_ a _6 a _6_ 6 a _6_ a
_6 a 6_ 6 a _6 AND SO 6 a _6 AND 6 a _6 SHARE
THE SAME ISOTROPIC SUBMANIFOLDS )F 6 IS CLOSED SIMPLY CONNECTED OR F
DOES NOT MIX UP x 6 AND x 6_ TOO MUCH ONE SEES BY LOOKING AT
LOCAL ISOMETRIES OF 6 THAT ALL ISOTROPIC SUBMANIFOLDS IN 6 a _6 HAVE
BOUNDED IN RADII AS WELL AS VOLUMES IF THEY ARE CLOSED AND SO )SO6 G IS
COMPACT /NE SEES EQUALLY WELL THAT G G g G_ IS STABLE AND MOREOVER
ONE GETS A GOOD CONTROL OVER THE STABILITY DOMAIN 5 OF G .AMELY TAKE
t AND LET 5t 5t G CONSIST OF THOSE G WHERE THE lBERS 6 a V_ AND
V a 6_ ARE G POSITIVE AND G NEGATIVE CORRESPONDINGLY AND THE PROJECTIONS
OF THESE lBERS TO 6 G AND 6_ G_ ARE t BI ,IPSCHITZ FOR G RESTRICTED TO
THESE lBERS 4HEN AT LEAST FOR x 6 THE GRAPHS OF DIkEOMORPHISMS
F 6 36 WITH 5t < F 5t ARE UNIFORMLY BI ,IPSCHITZ TO 6 AND SO
ALL G t 5t ARE STABLE
7HAT ARE THE MOST GENERAL STABLE G !RE SIMPLY CONNECTED MANIFOLDS
OF TYPE Q ALWAYS STABLE /F COURSE GENERIC G ARE STABLE BUT WE ARE
CONCERNED WITH EXCEPTIONAL 6 G EG WITH NON COMPACT GROUP )SO6 G
4HE ABOVE MOTIVATES THE IDEA OF ISO STABILITY FOR 6 OF TYPE P Q WITH
P Q EG 6 6 g _6 LIMITING THE SIZE OF n ISOTROPIC SUBMANIFOLDS IN
6 4HIS IS ENHANCED IN THE PRESENCE OF TAMING P FORMS ON 6 WHICH DO
NOT VANISH ON THE ISOTROPIC P PLANES IN 4 6 &OR EXAMPLE IF 6 a _6 IS
TAME WITH DEG N ON THIS OCCASION THEN 6 IS STABLE WITH RESPECT TO
THE DIkEOMORPHISMS WITH THE GRAPHS HOMOLOGOUS TO THE DIAGONAL IN 6 a 6
AS IS THE CASE FOR THE ABOVE G 5t AND FOR 6 G TAMED BY e FOLIATIONS
WITH P AND Q VOLUME PRESERVING OR JUST UNIFORMLY BOUNDED HOLONOMIES
2EMARKS AND REFERENCES A #LOSED # CURVES IN TAME MANIFOLDS EX
HIBIT A WELL ORGANIZED STRUCTURE WITH INTRICATE INTERACTION BETWEEN MODULI
-)3(! '2/-/6 '!&!
SPACES #D FOR DIkERENT D AND REGULAR ASYMPTOTICS FOR D QUANTUM MUL

TIPLICATION MIRROR SYMMETRY ETC SEE ;-3= 4HIS ALSO APPLIES TO NON CLOSED
CURVES WITH PRESCRIBED &REDHOLM BOUNDARY OR ASYMPTOTIC CONDITION EG
* MAPS OF 2IEMANN SURFACES WITH BOUNDARIES 3 3 6 7 FOR A GIVEN
TOTALLY REAL 7 y 6 WHERE EVERYTHING GOES AS IN THE BOUNDARY FREE CASE IN
CLUDING +OBAYASHI METRIC "LOCH "RODY ETC ,ESS OBVIOUS CONDITIONS COME
UP IN THE STUDY OF lXED POINTS OF (AMILTONIAN TRANSFORMATIONS AND RELATED
PROBLEMS &LOER HOMOLOGY ! CATEGORIES CONTACT HOMOLOGY OF %LIASHBERG
AND (OFER "UT IT IS UNCLEAR ONLY TO ME WHAT IS THE MOST GENERAL &RED
HOLM CONDITION IN THE # GEOMETRY
B 4HE QUESTIONS CONCERNING UNBOUNDED # CURVES WHICH PARALLEL .EVAN
LINNA KIND COMPLEX ANALYSIS RATHER THAN ALGEBRAIC GEOMETRY REMAIN AS
WIDELY OPEN AS WHEN ) COLLECTED THEM FOR THEN EXPECTED CONTINUATION OF
;'= $O THE SPACES 3 ( 0 " POSSESS GEOMETRIC STRUCTURE COMPARABLE TO
AND COMPATIBLE WITH WHAT WE SEE IN # ;D #D 7HAT IS THE RIGHT LANGUAGE
TO DESCRIBE SUCH A STRUCTURE IF IT EXISTS AT ALL
%VEN IN THE CLASSICAL CASE OF ALGEBRAIC 6 BOASTING OF LOTS AND LOTS OF DEEP
DInCULT THEOREMS THERE IS NO HINT OF THE GLOBAL PICTURE IN SIGHT NOT EVEN
A CONJECTURAL ONE SEE ;-C= FOR THE LATEST IN THE lELD
C (YPERBOLICITY OF 6 * CAN SOMETIMES BE DERIVED FROM NEGATIVITY
OF A SUITABLE CURVATURE OF A 2IEMANN OR &INSLER METRIC ADAPTED TO *
EITHER ON 6 ITSELF OR ON SOME JET SPACE OF # CURVES IN 6 ! MOST GENERAL
SEMI LOCAL HYPERBOLICITY CRITERION IS EXPRESSED BY THE LINEAR ISOPERIMETRIC
INEQUALITY IN # CURVES 3 y 6 )F SUCH AN INEQUALITY HOLDS TRUE ON RELATIVELY
SMALL * SURFACES 3 y 6 THEN IT PROPAGATES TO ALL 3 IN THE SAME WAY AS THE
REAL HYPERBOLICITY DOES SEE ;'= WHICH MAKES IT IN PRINCIPLE VERIlABLE FOR
COMPACT 6 4HE LINEAR INEQUALITY SEEMS TO FOLLOW FROM HYPERBOLICITY BY
A RATHER ROUTINE ARGUMENT BUT ) FAILED TO CARRY IT THROUGH 0OSSIBLY ONE
SHOULD LIMIT ONESELF TO CLOSED 6 AND INTEGRABILITY OF * MAY BE ALSO HELPFUL
)T WOULD BE AMUSING TO lND A SUnCIENTLY GENERAL POSITIVE CURVATURE
CONDITION FOR THE EXISTENCE OF MANY RATIONAL CURVES IN 6 * ENCOMPASS
ING COMPLEX HYPERSURFACES 6 WITH DEG 6 MUCH SMALLER THAN DIM 6 FOR
EXAMPLE AND ALLOWING SINGULAR SPACES IN THE SPIRIT OF !LEXANDROVS + t
#ONVERSELY IN THE PRESENCE OF MANY CLOSED ESPECIALLY RATIONAL CURVES ONE
EXPECTS EXTRA LOCAL STRUCTURES ON 6 EG TAMING FORMS WHICH EASILY COME
FROM CLOSED CURVES FOR DIM 6 SEE ;'=
D /NE CAN SOMETIMES MAKE FOLIATIONS OR AT LEAST LAMINATIONS OUT OF
6ISIONS IN -ATH 30!#%3 !.$ 15%34)/.3
PARABOLIC CURVES IN 6 AS IS DONE IN ;"= FOR * ON TORI TAMED BY THE STANDARD

SYMPLECTIC WHEREAS THE ORIGINAL QUESTION AIMED AT EVENTUALLY PROVING
THAT IS STANDARD
E (OW MUCH DO WE GAIN IN GLOBAL UNDERSTANDING OF A COMPACT 6 * BY
ASSUMING THAT THE STRUCTURE * IS INTEGRABLE IE COMPLEX )T SEEMS NOTHING
AT ALL THERE IS NO SINGLE RESULT CONCERNING ALL COMPACT COMPLEX MANIFOLDS
)F DIM 6 THEN THE +ODAIRA CLASSIlCATION TELLS US QUITE A BIT SAY FOR
EVEN B t ESPECIALLY IF THERE ARE ELEMENTS IN ( 6 WITH NON ZERO
PRODUCT YIELDING A lNITE MORPHISM OF 6 ONTO # c 4HIS SUGGESTS THE
PRESENCE OF UNREACHABLE POCKETS OF MODULI SPACES OF INTEGRABLE *S
WITH WEIRD PROPERTIES LIKE THOSE PRODUCED BY 4AUBS ON MANIFOLDS BUT
THERE IS NO GENERAL EXISTENCE THEOREM FOR COMPLEX STRUCTURES EITHER NOT EVEN
FOR OPEN 6 S COMPARE P IN ;'RO=0$2 AND EVEN WORSE NO SYSTEMATIC
WAY TO PRODUCE THEM 3O FAR #/-0!#4 #/-0,%8 -!.)&/,$3 HAVE
NOT STOOD UP TO THEIR FAMEb
F # CURVES DElNED BY RESTRICTING THEIR TANGENT PLANES TO THE SUBVARI
ETY % #0 M y 'R 2M OWE THEIR BEAUTY AND POWER TO THE ELLIPTICITY OF
% THERE IS A SINGLE PLANE E % THROUGH EACH LINE IN 2M /NE CAN DE
FORM % BY KEEPING THIS CONDITION THUS ARRIVING AT GENERALIZED # STRUCTURES
WHERE THE RESULTING % CURVES ARE SIMILAR TO # CURVES AND WHERE THE PICTURE
IS THE CLEAREST FOR M SEE ;'= AND REFERENCES THEREIN )N GENERAL A
lELD % OF SUBSETS %V y 'RK 4V 6 'RK 2N N DIM 6 DElNES A CLASS OF
K DIMENSIONAL % SUBVARIETIES IN 6 SAID <DIRECTED BY % EG 7 y 2N WITH
'AUSS IMAGE IN % WHICH SEEM MOST INTRIGUING UNDER ELLIPTICITY ASSUMP
TIONS ON % 4O FORMULATE THESE LET & & % % y 'RK 2N DENOTE
THE SPACE OF PAIRS E H FOR E % AND ALL HYPERPLANES H y E AND LOOK AT
THE TAUTOLOGICAL MAP x & 'RK_ 2N )DEALLY & IS A SMOOTH CLOSED
MANIFOLD AND x IS STABLY ONE TO ONE AND ONTO IE REMAINS SUCH UNDER SMALL
PERTURBATIONS OF % AS IN THE # CASE 3UCH % DO NOT COME CHEAPLY ALL
KNOWN INSTANCES OF THEM APPEAR AS DEFORMATIONS OF ,IE GROUP ORBITS IN
'RK 2N EG IN (ARVEY ,AWSON CALIBRATED GEOMETRIES WHERE A MOST TANTA
LIZING % y 'R 2 IS ASSOCIATED TO RATIONAL * CURVES IN 3 FOR THE STANDARD
' INVARIANT # STRUCTURE * ON 3 /NE GAINS MORE EXAMPLES BY DROP
PING <ONTO THUS ARRIVING AT OVER DETERMINED ELLIPTIC SYSTEMS <ISOTROPIC
IN PSEUDO 2IEMANNIAN FOR INSTANCE WHICH NEED INTEGRABILITY IN ORDER TO
HAVE SOLUTIONS AS <SPECIAL ,AGRANGIAN OF (ARVEY ,AWSON !LSO ONE MAY
b
&EDIA "OGOMOLOV SUGGESTED TO LOOK AT MANIFOLDS APPEARING AS SPACES OF LEAVES OF
FOLIATIONS IN PSEUDOCONVEX BOUNDED DOMAINS IN #. WITH ALGEBRAIC TANGENT BUNDLES
-)3(! '2/-/6 '!&!
ALLOW % AND x TO HAVE SOME SINGULARITIES SIMILAR TO THOSE PRESENT IN 9ANG

-ILLS IN A DIkERENT SETTING BUT IN ALL CASES ONE IS STUCK WITH TWO PROBLEMS
WHAT ARE POSSIBLE ELLIPTIC % AND WHAT ARE GLOBAL AND LOCAL ANALYTIC PROP
ERTIES ESPECIALLY SINGULARITIES OF THE CORRESPONDING % SUBVARIETIES 7 IN
6 % F%V G )F % IS OVER DETERMINED ONE LOOKS AT 7 S DIRECTED BY A
SMALL NEIGHBOURHOOD % z %
G 4HE RADII 2e ARE LESS USEFUL THAN 2e AS THEY MAKE SENSE ONLY FOR
RATHER SPECIAL PSEUDO 2IEMANNIAN MANIFOLDS 6 6 G 9ET 2e CAN BE
USED FOR CHARACTERIZATION OF SUCH 6 EG THE EQUALITY 2e 2e SEEMS
TO ALMOST DISTINGUISH METRICALLY SPLIT MANIFOLDS !LSO ONE CAN GENERALIZE
2e BY CONSIDERING SUBMERSIONS 6 7e BEING e LONG NORMALLY TO THE
lBERS WHERE s DIM 7 s P AND s DIM 7_ s Q AND WHERE G IS e
DElNITE ON THE BUNDLE OF VECTORS NORMAL TO THE lBERS 4HEN THE RESULTING
RADII 2
e 6 SATISFY 2e t 2e t 2e
H 4HE lNITENESS OF 2e 6 DOES NOT ENSURE STABILITY OF 6 FOR DIM 6 t

AS SIMPLE EG SPLIT EXAMPLES SHOW BUT THIS SEEMS TO <LIMIT INSTABILITY TO
CODIMENSION TWO #AN ONE GO FURTHER WITH STRONGER RADII TYPE INVARIANTS
I "ESIDES THE IN RADIUS THERE ARE OTHER 2IEMANNIAN INVARIANTS TO GAUGE
PSEUDO 2IEMANNIAN METRICS SUCH AS THE MACROSCOPIC DIMENSION SEE ;'=
OF COMPLETE POSITIVE SLICES IN 6 OR THE MAXIMAL RADIUS OF A %UCLIDEAN
P BALL IN 6 4HE %UCLIDEAN METRIC ON 2P DOMINATES OTHER G THERE IS A
LONG MAP 2P 2P G FOR EVERY 2IEMANNIAN METRIC G ON 2P "UT ONE CAN
GO BEYOND 2P BY ADMITTING SLICES WITH NON TRIVIAL TOPOLOGY 3OME 6 MAY
CONTAIN LOTS OF THESE EG SOME 6 OF LARGE DIMENSION SUPPORT P Q METRICS
G SO THAT EVERY 2IEMANNIAN P MANIFOLD ADMITS AN ISOMETRIC IMMERSION INTO
6 G &URTHERMORE ONE MAY LOOK AT HOMOTOPIES AND EXTENDABILITY OF
SLICES WITH CONTROLLED SIZE THUS GETTING EXTRA INVARIANTS OF 6 !CTUALLY THE
MERE TOPOLOGY OF THE SPACE OF SAY CLOSED POSITIVE SLICES CAN BE IMMENSELY
COMPLICATED ENCOURAGING US TO SEEK CONDITIONS LIMITING THIS COMPLEXITY EG
IN THE SPIRIT OF DIAGRAM GROUPS SEE ;'U3=
J )F 6 IS COMPACT ONE MAY DISTINGUISH COMPLETE SLICES FOR THE METRIC
INDUCED FROM SOME 2IEMANNIAN BACKGROUND H IN 6 AND THEN COMPARE THESE
INDUCED 2IEMANNIAN METRICS WITH THOSE INDUCED FROM OUR G OF TYPE P Q
"ESIDES COMPLETENESS H BRINGS FORTH OTHER CLASSES OF SLICES EG THOSE WITH
SOME BOUNDS ON CURVATURE
K 4HE ISOMETRICS OF NON STABLE 6 G HAVE ATTRACTIVE GEOMETRY AND
DYNAMICS SEE ;$'= FOR AN INTRODUCTION AND REFERENCES WITH MANY ELEMEN
6ISIONS IN -ATH 30!#%3 !.$ 15%34)/.3
TARY QUESTIONS REMAINING OPEN EG DOES EVERY ISOMETRY OF THE INTERIOR OF
6 EXTEND TO THE BOUNDARY 4HIS COMES FROM RELATIVITY ) GUESS
L -OST OF THE CURRENT PSEUDO 2IEMANNIAN RESEARCH IS LINKED TO GENERAL
RELATIVITY FOCUSED ON THE %INSTEIN EQUATION SEE;"E%%=
3YMBOLIZATION AND 2ANDOMIZATION

! COMMON WAY TO GENERATE QUESTIONS NOT ONLY IN GEOMETRY IS TO CONFRONT
PROPERTIES OF OBJECTS SPECIlC TO DIkERENT CATEGORIES WHAT IS A POSSIBLE TOPOL
OGY EG HOMOLOGY OF A MANIFOLD WITH A GIVEN TYPE OF CURVATURE (OW IS
THE DYNAMICS OF THE GEODESIC mOW CORRELATED WITH TOPOLOGY ANDOR GEOME
TRY (OW FAST CAN A HARMONIC FUNCTION DECAY ON A COMPLETE MANIFOLD WITH
A CERTAIN ASYMPTOTIC GEOMETRY EG CURVATURE (OW MANY CRITICAL POINTS
MAY GEOMETRICALLY DElNED ENERGY HAVE ON A GIVEN SPACE OF MAPS OR SUBVA
RIETIES 7HAT ARE POSSIBLE SINGULARITIES OF THE EXPONENTIAL MAP OR OF THE
CUT LOCUS $OES EVERY ALMOST COMPLEX MANIFOLD OF DIM t SUPPORT A
COMPLEX STRUCTURE ETC 4HESE SEDUCE US BY SIMPLICITY AND APPARENT NATU
RALITY SOMETIMES LEADING TO NEW IDEAS AND STRUCTURES TANGENTIALLY RELATED
TO THE ORIGINAL QUESTIONS AS IN THE -ORSE ,USTERNIK 3CHNIRELMANN THEORY
MOTIVATED BY CLOSED GEODESICS BUT OFTEN THE MIRAGE OF NATURALITY LURES US
INTO A FEATURELESS DESERT WITH NO CLEAR PERSPECTIVE WHERE THE SOLUTION EVEN
IF FOUND DOES NOT QUENCH OUR THIRST FOR STRUCTURAL MATHEMATICS %XAMPLES
ARE LEFT TO THE READER
!NOTHER APPROACH CONSISTS IN INTERBREEDING RATHER THAN INTERSECTING
CATEGORIES AND IDEAS 4HIS HAS A BETTER CHANCE FOR A SUCCESSFUL OUTCOME
WITH QUESTIONS FOLLOWING RATHER THAN PRECEDING CONSTRUCTION OF NEW OB
JECTS *UST LOOK AT HOW IT WORKS SYMBOLIC DYNAMICS ALGEBRAIC ARITHMETIC
AND NON COMMUTATIVE GEOMETRY QUANTUM COMPUTERS DIkERENTIAL TOPOLOGY
RANDOM GRAPHS P ADIC ANALYSIS .OW WE WANT TO CONTINUE WITH SYMBOLIC
GEOMETRY AND RANDOM GROUPS
'IVEN A CATEGORY OF <SPACES 8 WITH lNITE #ARTESIAN PRODUCTS WE CON
SIDER FORMAL INlNITE PRODUCTS 8 a 8I WHERE THE INDEX SET HAS AN AD
I)
DITIONAL DISCRETE STRUCTURE EG BEING A GRAPH OR A DISCRETE GROUP ` )N
THE LATTER CASE WE ASSUME THAT ALL 8I ARE THE SAME 8 8 ` CONSISTS OF
FUNCTIONS ~ ` 8 AND ` ACTS NATURALLY ON 8 .OTHING HAPPENS UNLESS
WE START LOOKING AT MORPHISMS h 8 9 OVER A lXED ` 3UCH A h IS GIVEN
BY A lNITE SUBSET a y ` OF CARDINALITY D AND A MAP 8 a 8 D 9
-)3(! '2/-/6 '!&!
IE A FUNCTION Y X XD WHERE h~ l IS DElNED AS THE VALUE OF

ON THE RESTRICTION OF ~ TO THE l TRANSLATE la y ` FOR ALL l ` 4HUS
WE ENRICH THE ORIGINAL CATEGORY BY MAKING SINGLE VARIABLE ` EQUIVARIANT
FUNCTIONS h~ OUT OF FUNCTIONS X XD IN SEVERAL VARIABLES
4AKE A PARTICULAR CATEGORY OF 8S EG ALGEBRAIC VARIETIES SMOOTH SYM

PLECTIC OR 2IEMANNIAN MANIFOLDS SMOOTH DYNAMICAL SYSTEMS WHATEVER
YOU LIKE AND START TRANSLATING BASIC CONSTRUCTIONS NOTIONS AND QUESTIONS
INTO THE <SYMBOLIC LANGUAGE OF 8 S 4HIS IS PURSUED IN ;'= AND ;'=
WITH AN EYE ON CONTINUOUS COUNTERPARTS TO 8 EG SPACES OF HOLOMORPHIC
MAPS # 8 FOR ALGEBRAIC VARIETIES 8 WITH A HOPE TO MAKE <ALGEBRAIC
SOMEHOW REmECTED IN SUCH SPACES ) HAVE NOT GONE FAR A SYMBOLIC VER
SION OF THE !X MAPPING THEOREM FOR AMENABLE ` SIMILAR TO THE 'ARDEN OF
%DEN IN CELLULAR AUTOMATA AND A NOTION OF MEAN DIMENSION DElNED FOR ALL
COMPACT ` SPACES 8 WITH AMENABLE ` IN THE SPIRIT OF TOPOLOGICAL ENTROPY
RECAPTURING DIMTOP 8 FOR 8 8 ` APPLICABLE TO SPACES LIKE " OF * MAPS
# 8 * FOR INSTANCE 4HATS ABOUT IT 4HE READER IS MOST WELCOME TO
THESE 8 IF ANYTHING THERE IS NO LACK OF OPEN QUESTIONS YET NO GUARANTEE
THEY WOULD LEAD TO A NEW GRAND THEORY EITHER
2ANDOMIZATION 2ANDOM LIES AT THE VERY SOURCE OF MANIFOLDS AT LEAST
IN THE SMOOTH AND THE ALGEBRAIC CATEGORIES GENERAL SMOOTH MANIFOLDS 6
APPEAR AS PULL BACK OF SPECIAL SUBMANIFOLDS UNDER GENERIC OR RANDOM
SMOOTH MAPS F BETWEEN STANDARD MANIFOLDS EG ZEROS OF GENERIC FUNC
TIONS F 2. 2. _N OR PROPER GENERIC MAPS FROM 2. TO THE CANONICAL
VECTOR BUNDLES 7 OVER 'RASSMANN MANIFOLDS 'R. _N 2- 4HOM CONSTRUC
TION WHERE 6 COME AS F _ FOR THE ZERO SECTION 'R. _N 2- y 7
/THER CONSTRUCTIONS IN DIkERENTIAL TOPOLOGY AMOUNT TO A LITTLE TINKERING
WITH 6 S CREATED BY GENERICITY 3IMILARLY THE BULK OF ALGEBRAIC MANIFOLDS
COMES FROM INTERSECTING AMPLE GENERIC HYPERSURFACES IN STANDARD MANIFOLDS
EG IN #0 . AND THE FULL LIST OF KNOWN CONSTRUCTIONS OF SAY NON SINGULAR
ALGEBRAIC VARIETIES IS DISMALLY SHORT
/NE MAY OBJECT BY POINTING OUT THAT EVERY COMBINATORIAL MANIFOLD

CAN BE ASSEMBLED OUT OF SIMPLICES )NDEED IT IS EASY TO MAKE POLYHEDRA
BUT NO WAY TO RECOGNIZE MANIFOLDS AMONG THEM AS EVENTUALLY FOLLOWS FROM
UNDECIDABILITY OF TRIVIALITY FOR lNITELY PRESENTED GROUPS (ERE IS ANOTHER
BASIC PROBLEM LINKED TO <NON LOCALITY OF TOPOLOGY (OW MANY TRIANGU
LATIONS MAY A GIVEN SPACE 8 EG A SMOOTH MANIFOLD SAY THE SPHERE 3 N
HAVE .AMELY LET T8 . DENOTE THE NUMBER OF MUTUALLY COMBINATORI
6ISIONS IN -ATH 30!#%3 !.$ 15%34)/.3
ALLY NON ISOMORPHIC TRIANGULATIONS OF 8 INTO . SIMPLICES $OES THIS T GROW

AT MOST EXPONENTIALLY IN . IE WHETHER T8 . s EXP #8 . .OTICE
THAT THE NUMBER OF ALL 8 BUILT OF . N DIMENSIONAL SIMPLICES GROWS SUPER
EXPONENTIALLY ROUGHLY AS NN AND THE MAJOR DInCULTY FOR A GIVEN 8 COMES
FROM x 8 AND POSSIBLY BUT LESS LIKELY FROM ( 8 WHERE THE ISSUE
IS TO COUNT THE NUMBER OF TRIANGULATED MANIFOLDS 8 WITH A lXED x 8 OR
( 8
4HESE QUESTIONS COMING FROM PHYSICISTS WORKING ON QUANTIZATION OF
GRAVITY HAVE AN ESSENTIALLY EQUIVALENT COMBINATORIAL COUNTERPART WE
STUMBLED UPON WITH !LEX .ABUTOVSKI EVALUATE THE NUMBER T, . OF CON
NECTED VALENT IE DEGREE s GRAPHS 8 WITH . EDGES SUCH THAT CYCLES
OF LENGTH s , NORMALLY GENERATE x 8 OR AT LEAST GENERATE ( 8 )S
T, . AT MOST EXPONENTIAL IN . FOR A lXED SAY , 4HE QUESTIONS
LOOK JUST GREAT AND NO IDEA HOW TO ANSWER THEM
(ERE IS A SOMEWHAT SIMILAR BUT EASIER QUESTION WHAT DOES A RANDOM
GROUP RATHER THAN A SPACE LOOK LIKE !S WE SHALL SEE THE ANSWER IS MOST
SATISFACTORY AT LEAST FOR ME <NOTHING LIKE WE HAVE EVER SEEN BEFORE .O
BIG SURPRISE THOUGH TYPICAL OBJECTS ARE USUALLY ATYPICAL
2ANDOM PRESENTATION OF GROUPS 'IVEN A GROUP & EG THE FREE GROUP
&K WITH K GENERATORS ONE MAY SPEAK OF RANDOM QUOTIENT GROUPS ' &;2=
WHERE 2 y & IS A RANDOM SUBSET WITH RESPECT TO SOME PROBABILITY MEASURE
u ON ' AND ;2= STANDING FOR THE NORMAL SUBGROUP GENERATED BY 2 4HE
SIMPLEST WAY TO MAKE A u IS TO CHOOSE WEIGHTS Pl ; = FOR ALL l ` AND
TAKE THE PRODUCT MEASURE u IN ' IE 2 IS OBTAINED BY INDEPENDENT CHOICES
OF l & WITH PROBABILITIES Pl 4HIS IS STILL TOO GENERAL WE SPECIALIZE TO
Pl PJlJ WHERE JlJ DENOTES THE WORD LENGTH OF l FOR A GIVEN SAY lNITE
SYSTEM OF GENERATORS IN & ! PRETTY SUCH P IS P Pq l CARDFl
& J Jl J JlJG _q q )F THE <TEMPERATURE q IS CLOSE TO ZERO P l
DECAYS SLOWLY AND RANDOM 2 IS SO LARGE THAT IT NORMALLY GENERATES ALL &
MAKING ' FEG WITH PROBABILITY ONE PROVIDED & IS INlNITE 0 &OR EXAMPLE
IF & &K THIS HAPPENS WHENEVER Pl @ &K IE
l P l
4HIS IS EASY BUT IT IS NOT SO CLEAR IF ' MAY BE EVER NON TRIVIAL FOR LARGE q
(OWEVER IF & &K OR A GENERAL NON ELEMENTARY WORD HYPERBOLIC GROUP
ONE CAN SHOW THAT ' IS INlNITE WITH POSITIVE PROBABILITY FOR q qCRIT &
AND qCRIT &K PROBABLY EQUALS IE Pq l @ CARD' WITH NON
ZERO PROBABILITY SEE ;'= FOR A SLIGHTLY DIkERENT Pl WHERE THE CRITICAL
COMES AS THE %ULER CHARACTERISTIC OF 3 VIA THE SMALL CANCELLATION THEORY
2ANDOM GROUPS 'q LOOK VERY DIkERENT FOR DIkERENT q )T SEEMS THAT 'q
-)3(! '2/-/6 '!&!
CANNOT BE EMBEDDED EVEN IN A MOST GENEROUS SENSE OF THE WORD INTO 'q
FOR q q AS THE <DENSITY OF RANDOM 'q DECREASES WITH <TEMPERATURE
&URTHERMORE GENERIC SAMPLES OF 'q FOR THE SAME LARGE q ARE PROBABLY
MUTUALLY NON ISOMORPHIC NOT EVEN QUASI ISOMETRIC WITH PROBABILITY ONE
YET THEIR <ELEMENTARY INVARIANTS ARE LIKELY TO BE THE SAME )T IS CLEAR FOR
ALL q THAT 'q AS HAVE NO lNITE FACTOR GROUPS AND THEY MAY SATISFY
+AZHDANS PROPERTY 4 4 IS MORE PROBABLE FOR SMALL q WHERE IT IS HARDER
FOR 'q TO BE INlNITE
,ET US MODIFY THE ABOVE PROBABILITY SCHEME BY CONSIDERING A RANDOM
HOMOMORPHISM FROM A lXED GROUP ( TO & WITH ' &;( = 4O BE
SIMPLE LET ( x a FOR A DIRECTED GRAPH a AND BE GIVEN BY RANDOM
ASSIGNMENT OF GENERATORS OF & TO EACH EDGE E IN a INDEPENDENTLY FOR ALL
EDGES $ENOTE BY . , THE NUMBER OF NON ORIENTED CYCLES IN a OF LENGTH
s , AND OBSERVE THAT FOR LARGE . , THE GROUP ' IS LIKELY TO BE TRIVIAL "UT
WE CARE FOR INlNITE ' AND THIS CAN BE GUARANTEED WITH POSITIVE PROBABILITY IF
. , s EXP ,k FOR LARGE k t kCRIT & FOR FREE GROUPS & &K K t AND
NON ELEMENTARY HYPERBOLIC GROUPS & IN GENERAL ) HAVE CHECKED THIS SO FAR
UNDER AN ADDITIONAL LACUNARITY ASSUMPTION ALLOWING FOR EXAMPLE a BEING
THE DISJOINT UNION OF lNITE GRAPHS aI OF CARDINALITIES DI I WITH
DI t EXP DI AND SUCH THAT THE SHORTEST CYCLE IN aI HAS LENGTH t kCR LOG DI
FOR ALL I
4O HAVE AN INlNITE RANDOM ' '& a WITH INTERESTING FEATURES WE
NEED A SPECIAL a 7E TAKE a SUCH THAT IT CONTAINS ARBITRARILY LARGE t
EXPANDERS WITH A lXED POSSIBLY SMALL t 3UCH a DO EXIST IN FACT
RANDOM a IN OUR CATEGORY CONTAIN SUCH EXPANDERS SEE ;,U= 4HEN RANDOM
GROUPS ' AS ENJOY THE FOLLOWING PROPERTIES
! ' ARE +AZHDAN 4 IE EVERY AnNE ISOMETRIC ACTION OF SUCH ' ON
THE (ILBERT SPACE 2 HAS A lXED POINT &URTHERMORE EVERY ACTION OF '
ON A POSSIBLY INlNITE DIMENSIONAL COMPLETE SIMPLY CONNECTED 2IEMANNIAN
MANIFOLD 6 WITH +6 s HAS A lXED POINT IF DIM 6 THEN THERE IS
NO NON TRIVIAL ACTION AT ALL
" &OR EVERY ,IPSCHITZ FOR THE WORD METRIC MAP F ' 2 THERE
ARE SEQUENCES GI GI ' WITH DIST' GI GI AND DIST2 F GI F GI s
CONST AND THE SAME REMAINS TRUE FOR THE @P SPACES FOR P
/NE MAY THINK THE ABOVE <PATHOLOGIES ARE DUE TO THE FACT THAT ' ARE
NOT lNITELY PRESENTED "UT ONE CAN SHOW THAT SOME <QUASI RANDOM GROUPS
AMONG OUR ' ARE RECURSIVELY PRESENTED AND SO EMBED INTO A lNITELY PRESENTED
6ISIONS IN -ATH 30!#%3 !.$ 15%34)/.3
GROUP ' WHICH THEN AUTOMATICALLY SATISlES " AND CAN BE CHOSEN WITH
AN ASPHERICAL PRESENTATION BY A RECENT UNPUBLISHED RESULT BY )LIA 2IPS
AND -ARK 3APIR WHERE 4 CAN BE PRESERVED BY ADDING EXTRA RELATIONS
4HEN ONE CAN ARRANGE ' x 6 FOR A CLOSED ASPHERICAL MANIFOLD 6
OF A GIVEN DIMENSION N t WHICH BESIDES " HAS MORE <NASTY FEATURES
ARRESTING FOR EXAMPLE ALL KNOWN AS FAR AS ) CAN TELL ARGUMENTS FOR PROVING
STRONG .OVIKOVS CONJECTURE FOR ' 3EE ;'= BUT ) COULD NOT RULE OUT
HYPERSPHERICITY YET ) FEEL RANDOM GROUPS ALTOGETHER MAY GROW UP AS
HEALTHY AS RANDOM GRAPHS FOR EXAMPLE
4HERE ARE OTHER POSSIBILITIES TO DElNE RANDOM GROUPS EG BY FOLLOWING
THE <SYMBOLIC APPROACH WHERE COMBINATORIAL MANIPULATIONS WITH lNITE SETS
ARE REPLACED BY PARALLEL CONSTRUCTIONS IN A GEOMETRIC CATEGORY &OR EXAMPLE
WE MAY GIVE SOME STRUCTURE TOPOLOGY MEASURE ALGEBRAIC GEOMETRY TO THE
GENERATING SET " OF FUTURE ' WITH RELATIONS BEING GEOMETRIC SUBSETS IN
THE #ARTESIAN POWERS OF " 4HEN DEPENDING ON THE STRUCTURE ONE MAY
SPEAK OF <RANDOM OR <GENERIC GROUPS ' WITH A POSSIBLE RETURN TO lNITELY
GENERATED GROUPS VIA A MODEL THEORETIC REASONING ,OOKS PROMISING BUT )
HAVE NOT ARRIVED AT A POINT OF ASKING QUESTIONS
) CONCLUDE BY THANKING 3TEPHANIE !LEXANDER AND -IRIAM (ERCBERG FOR
THEIR HELP IN MY STRUGGLE WITH %NGLISH
2EFERENCES
;!"#+4= * !MORO s S - "URGER + #ORLETTE $ +OTSCHICK
$ 4OLEDO &UNDAMENTAL 'ROUPS OF #OMPACT + AHLER -ANIFOLDS -ATHE
MATICAL 3URVEYS AND -ONOGRAPHS !MERICAN -ATHEMATICAL 3OCIETY 0ROV
IDENCE 2)
;"= 6 "ANGERT %XISTENCE OF A COMPLEX LINE IN TAME ALMOST COMPLEX TORI
$UKE -ATH * [
;"E%%= *+ "EEM 0% %HRLICH +, %ASLEY 'LOBAL ,ORENTZIAN 'EOME
TRY ED -ONOGRAPH 4EXTBOOKS IN 0URE AND !PPL -ATHEMATICS
-ARCEL $EKKER )NC .9
;"ER= - "ERGER 2IEMANNIAN GEOMETRY DURING THE SECOND HALF OF THE TWENTIETH
CENTURY *AHRBERICHT DER $EUTSCHEN -ATH 6EREINIGUNG [
;"R= 2 "RYANT 2ECENT ADVANCES IN THE THEORY OF HOLONOMY 3sEM . "OURBAKI
VOL JUIN
;##= * #HEEGER 4 #OLDING /N THE STRUCTURE OF SPACES WITH 2ICCI CURVATURE
BOUNDED BELOW ))) PREPRINT
;$'= ' $!MBRA - 'ROMOV ,ECTURES ON TRANSFORMATION GROUPS 'EOME
TRY AND DYNAMICS 3URVEYS IN $IkERENTIAL 'EOMETRY [
-)3(! '2/-/6 '!&!
;'= - 'ROMOV WITH !PPENDICES BY - +ATZ 0 0ANSU AND 3 3EMMES

-ETRIC 3TRUCTURES FOR 2IEMANNIAN AND .ON 2IEMANNIAN 3PACES BASED
ON 3TRUCTURES -sETRIQUES DES 6ARIsETsES 2IEMANNIENNES * ,A&ONTAINE
0 0ANSU EDS %NGLISH 4RANSLATION BY 3EAN - "ATES "IRKHAUSER
"OSTON["ASEL["ERLIN
;'= - 'ROMOV !SYMPTOTIC INVARIANTS OF INlNITE GROUPS 'EOMETRIC GROUP
THEORY 6OL 0ROC 3YMP 3USSEX 5NIV "RIGHTON *ULY
,ONDON -ATH 3OC ,ECTURE .OTES .IBLO 2OLLER ED #AMBRIDGE
5NIV 0RESS #AMBRIDGE [
;'= - 'ROMOV #ARNOT #ARATHEODORY SPACES SEEN FROM WITHIN SUB
2IEMANNIAN GEOMETRY 0ROC *OURNsEES NONHOLONOMES 'sEOMsETRIE SOUS
RIEMANNIENNE THsEORIE DU CONTR>OLE ROBOTIQUE 0ARIS *UNE [*ULY
! "ELLAICHE ED "IRKH
AUSER "ASEL 0ROG -ATH [
;'= - 'ROMOV %NDOMORPHISMS OF SYMBOLIC ALGEBRAIC VARIETIES * %UR
-ATH 3OC [
;'= - 'ROMOV (YPERBOLIC GROUPS IN <%SSAYS IN 'ROUP 4HEORY -ATHEMAT
ICAL 3CIENCES 2ESEARCH )NSTITUTE 0UBLICATIONS [ 3PRINGER
6ERLAG
;'= - 'ROMOV 0OSITIVE CURVATURE MACROSCOPIC DIMENSION SPECTRAL GAPS
AND HIGHER SIGNATURES IN <&UNCTIONAL !NALYSIS ON THE %VE OF THE ST
#ENTURY 'INDIKIN 3IMON ET AL EDS 6OLUME )) IN HONOR OF THE EIGHTIETH
BIRTHDAY OF )- 'ELFAND 0ROC #ONF 2UTGERS 5NIV .EW "RUNSWICK .*
53! /CT "IRKH AUSER "ASEL 0ROG -ATH [
;'= - 'ROMOV 0ARTIAL $IkERENTIAL 2ELATIONS 3PRINGER 6ERLAG %RGEB
DER -ATH &OLGE "D
;'= - 'ROMOV 2ANDOM WALK IN RANDOM GROUPS IN PREPARATION
;'= - 'ROMOV 3IGN AND GEOMETRIC MEANING OF CURVATURE 2END 3EM -ATH
&IS -ILANO [
;'= - 'ROMOV 3OFT AND HARD SYMPLECTIC GEOMETRY 0ROC )#- "ERKE
LEY !-3 [
;'= - 'ROMOV 4OPOLOGICAL INVARIANTS OF DYNAMICAL SYSTEMS AND SPACES OF
HOLOMORPHIC MAPS TO APPEAR IN *OURN OF 'EOMETRY AND 0ATH 0HYSICS
;'U3= 6 'UBA - 3APIR $IAGRAM 'ROUPS -EMOIRS OF THE !-3 .OVEMBER

;+= 0 +ANERVA 3PARSE $ISTRIBUTED -EMORY #AMBRIDGE -ASS -)4 0RESS

;+L= & +LEIN 6ORLESUNGEN U BER DIE %NTWICKLUNG DER -ATHEMATIK IM
*AHRHUNDERT 4EIL "ERLIN 6ERLAG VON *ULIUS 3PRINGER
;,3= 5 ,ANG 6 3CHROEDER +IRSZBRAUNS THEOREM AND METRIC SPACES OF
"OUNDED #URVATURE '!&! [
;,U= ! ,UBOTZKY $ISCRETE GROUPS EXPANDING GRAPHS AND INVARIANT MEASURES
"IRKHAUSER 0ROGRESS IN -ATHEMATICS
6ISIONS IN -ATH 30!#%3 !.$ 15%34)/.3
;-3= $ -C$UFF $ 3ALAMON * HOLOMORPHIC #URVES AND 1UANTUM #OHO

MOLOGY 5NIV ,ECT 3ERIES Nm !-3 0ROVIDENCE
;-C= - -C1UILLAN (OLOMORPHIC CURVES ON HYPERPLANE SECTIONS OF FOLDS
'!&! [
;-I= 6$ -ILMAN 4HE HERITAGE OF 0 ,sEVY IN GEOMETRICAL FUNCTIONAL ANALYSIS
IN <#OLLOQUE 0 ,sEVY SUR LES 0ROCESSUS 3TOCHASTIQUES !STsERISQUE [

;.= *2 .EWMAN 4HE 7ORLD OF -ATHEMATICS 6OL 3IMON AND 3CHUSTER
.9
;0= ' 0ERELMAN 3PACES WITH CURVATURE BOUNDED BELOW 0ROC )#-
3 #HATTERJI ED "IRKH
AUSER 6ERLAG "ASEL [
;0E= 0 0ETERSEN !SPECTS OF GLOBAL 2IEMANNIAN GEOMETRY "!-3
[
;3= $ 3PRING #ONVEX )NTEGRATION 4HEORY "IRKH AUSER 6ERLAG "ASEL["OSTON[
"ERLIN
;6= !6 6ASILIEV .IKOLAI )VANOVICH ,OBACHEVSKI -OSCOW .AUKA IN
2USSIAN
-IKHAIL 'ROMOV )(%3 ROUTE DE #HARTRES "URES SUR 9VETTE &RANCE
AND 4HE #OURANT )NSTITUTE -ATHEMATICAL 3CIENCES .95 -ERCER 3TREET .9
53!
#,!33)&)#!4)/. /& ).&).)4% $)-%.3)/.!,

3)-0,% '2/503 /& 350%239--%42)%3 !.$
15!.45- &)%,$ 4(%/29
6ICTOR ' +AC
)NTRODUCTION
4HIS WORK WAS MOTIVATED BY TWO SEEMINGLY UNRELATED PROBLEMS
,IES PROBLEM OF CLASSIlCATION OF <LOCAL CONTINUOUS TRANSFORMATION
GROUPS OF A lNITE DIMENSIONAL MANIFOLD
4HE PROBLEM OF CLASSIlCATION OF OPERATOR PRODUCT EXPANSIONS /0%
OF CHIRAL lELDS IN DIMENSIONAL CONFORMAL lELD THEORY
) SHALL BRIEmY EXPLAIN IN X HOW THESE PROBLEMS ARE RELATED TO EACH OTHER
VIA THE THEORY OF CONFORMAL ALGEBRAS ;$+= ;+[= 4HIS CONNECTION LED TO
THE CLASSIlCATION OF lNITE SYSTEMS OF CHIRAL BOSONIC lELDS SUCH THAT IN THEIR
/0% ONLY LINEAR COMBINATIONS OF THESE lELDS AND THEIR DERIVATIVES OCCUR
;$+= WHICH IS BASICALLY WHAT IS CALLED A <lNITE CONFORMAL ALGEBRA
)T IS OF COURSE WELL KNOWN THAT A SOLUTION TO ,IES PROBLEM REQUIRES QUITE
DIkERENT METHODS IN THE CASES OF lNITE AND INlNITE DIMENSIONAL GROUPS
4HE MOST IMPORTANT ADVANCE IN THE lNITE DIMENSIONAL CASE WAS MADE BY
7 +ILLING AND % #ARTAN AT THE END OF THE TH CENTURY WHO GAVE THE CEL
EBRATED CLASSIlCATION OF SIMPLE lNITE DIMENSIONAL ,IE ALGEBRAS OVER # 4HE
INlNITE DIMENSIONAL CASE WAS STUDIED BY #ARTAN IN A SERIES OF PAPERS WRITTEN
IN THE BEGINNING OF THE TH CENTURY WHICH CULMINATED IN HIS CLASSIlCATION
OF INlNITE DIMENSIONAL <PRIMITIVE ,IE ALGEBRAS ;#A=
4HE ADVENT OF SUPERSYMMETRY IN THEORETICAL PHYSICS IN THE S MO
TIVATED THE WORK ON THE <SUPER EXTENSION OF ,IES PROBLEM )N THE lNITE
DIMENSIONAL CASE THE LATTER PROBLEM WAS SETTLED IN ;+= (OWEVER IT TOOK
ANOTHER YEARS BEFORE THE PROBLEM WAS SOLVED IN THE INlNITE DIMENSIONAL
CASE ;+= ;#H+= !N ENTERTAINING ACCOUNT OF THE HISTORICAL BACKGROUND
OF THE FOUR CLASSIlCATIONS MENTIONED ABOVE MAY BE FOUND IN THE REVIEW ;3T=
! LARGE PART OF MY TALK XX[ IS DEVOTED TO THE EXPLANATION OF THE
FOURTH CLASSIlCATION THAT OF SIMPLE INlNITE DIMENSIONAL LOCAL SUPERGROUPS
OF TRANSFORMATIONS OF A lNITE DIMENSIONAL SUPERMANIFOLD 4HE APPLICATION
6ISIONS IN -ATH ).&).)4% 3)-0,% 350%239--%42)%3 !.$ 1&4
OF THIS RESULT TO THE SECOND PROBLEM THAT OF CLASSIlCATION OF /0% WHEN

FERMIONIC lELDS ARE ALLOWED AS WELL OR EQUIVALENTLY OF lNITE CONFORMAL
SUPERALGEBRAS IS EXPLAINED IN X
) AM CONVINCED HOWEVER THAT THE CLASSIlCATION OF INlNITE DIMENSIONAL
SUPERGROUPS MAY HAVE APPLICATIONS TO <REAL PHYSICS AS WELL 4HE MAIN
REASON FOR THIS BELIEF IS THE OCCURRENCE IN MY CLASSIlCATION OF CERTAIN EX
CEPTIONAL INlNITE DIMENSIONAL ,IE SUPERGROUPS THAT ARE NATURAL EXTENSIONS
OF THE COMPACT ,IE GROUPS 35 a 35 a 5 AND 35 )N X ) FORMULATE A
SYSTEM OF AXIOMS WHICH VIA REPRESENTATION THEORY OF THE CORRESPONDING ,IE
SUPERALGEBRAS SEE X PRODUCE PRECISELY ALL THE MULTIPLETS OF FUNDAMENTAL
PARTICLES OF THE 3TANDARD MODEL
) WISH TO THANK + 'AWEDZKI $ &REEDMAN , -ICHEL , /KUN 2 0EN
ROSE 3 0ETCOV ! 3MILGA ( 3TREMNITZER ) 4ODOROV FOR DISCUSSIONS AND
CORRESPONDENCE ) AM GRATEFUL TO 0 ,ITTLEMANN AND * VAN DER *EUGT FOR
PROVIDING TABLES OF CERTAIN BRANCHING RULES
,IES 0ROBLEM AND #ARTANS 4HEOREM

)N HIS <4RANSFORMATION GROUPS PAPER ;,= PUBLISHED IN ,IE ARGUES AS
FOLLOWS ,ET ' BE A <LOCAL CONTINUOUS GROUP OF TRANSFORMATIONS OF A lNITE
DIMENSIONAL MANIFOLD 4HE MANIFOLD DECOMPOSES INTO A UNION OF ORBITS SO
WE SHOULD lRST STUDY HOW THE GROUP ACTS ON AN ORBIT AND WORRY LATER HOW
THE ORBITS ARE PUT TOGETHER )N OTHER WORDS WE SHOULD lRST STUDY TRANSITIVE
ACTIONS OF ' &URTHERMORE EVEN FOR A TRANSITIVE ACTION IT MAY HAPPEN THAT
' LEAVES INVARIANT A lBRATION BY PERMUTING THE lBERS "UT THEN WE SHOULD
lRST STUDY HOW ' ACTS ON lBERS AND ON THE QUOTIENT MANIFOLD AND WORRY
LATER HOW TO PUT THESE ACTIONS TOGETHER 7E THUS ARRIVE AT THE PROBLEM OF
CLASSIlCATIONS OF TRANSITIVE PRIMITIVE IE LEAVING NO INVARIANT lBRATIONS
ACTIONS
.EXT ,IE ESTABLISHES HIS FAMOUS THEOREMS THAT RELATE THE ACTION OF ' ON
A MANIFOLD - IN A NEIGHBORHOOD OF A POINT P TO THE ,IE ALGEBRA OF VECTOR
lELDS IN THIS NEIGHBORHOOD GENERATED BY THIS ACTION !CTUALLY HE TALKS
ABOUT FORMAL VECTOR lELDS IN A FORMAL NEIGHBORHOOD OF P HENCE ' GIVES
RISE TO A ,IE ALGEBRA , OF FORMAL VECTOR lELDS AND ITS CANONICAL lLTRATION
BY SUBALGEBRAS ,J OF , CONSISTING OF VECTOR lELDS THAT VANISH AT P UP TO
THE ORDER J 4HEN TRANSITIVITY OF THE ACTION OF ' IS EQUIVALENT TO THE
PROPERTY THAT DIM ,, M DIM - AND PRIMITIVITY IS EQUIVALENT TO THE
PROPERTY THAT , IS A MAXIMAL SUBALGEBRA OF ,
6' +!# '!&!
4HE lRST BASIC EXAMPLE IS THE ,IE ALGEBRA OF ALL FORMAL VECTOR lELDS IN
M INDETERMINATES y8 z
M
7M 0I X
XI
I
0I X ARE FORMAL POWER SERIES IN X X XM ENDOWED WITH THE FOR
MAL TOPOLOGY ! SUBALGEBRA , OF 7M IS CALLED TRANSITIVE IF DIM ,, M
WHERE , 7M < , ! RIGOROUS STATEMENT OF ,IES PROBLEM IS AS FOLLOWS
ST FORMULATION #LASSIFY ALL CLOSED TRANSITIVE SUBALGEBRAS , OF 7M
SUCH THAT , IS A MAXIMAL SUBALGEBRA OF , UP TO A CONTINUOUS AUTOMOR
PHISM OF 7M
#ARTAN PUBLISHED A SOLUTION TO THIS PROBLEM IN THE INlNITE DIMENSIONAL
CASE IN ;#A= 4HE RESULT IS THAT A COMPLETE LIST OVER # CONJECTURED BY
,IE IS AS FOLLOWS M t
7M
3M F8 7M J DIV 8 G M t
#3M F8 7M J DIV 8 CONSTG M t 0
(M F8 7M J 8S G M K WHERE S KI DXI > DXKI
IS A SYMPLECTIC FORM
#(M F8 7M J 8S CONST S G M K
0MK F8 7M J 8C F C G M K WHERE C DXM
+
I XI DXKI IS A CONTACT FORM AND F IS A FORMAL POWER SERIES DE
PENDING ON 8
4HE WORK OF #ARTAN HAD BEEN VIRTUALLY FORGOTTEN UNTIL THE SIXTIES !
RESURGENCE OF INTEREST IN THIS AREA BEGAN WITH THE PAPERS ;3I3= AND ;'3=
WHICH DEVELOPED AN ADEQUATE LANGUAGE AND MACHINERY OF lLTERED AND GRADED
,IE ALGEBRAS 4HE WORK DISCUSSED IN THE PRESENT TALK ALSO USES HEAVILY THE
IDEAS FROM ;7= ;+= AND ;'=
4HE TRANSITIVITY OF THE ACTION IMPLIES THAT , CONTAINS NO NON ZERO IDEALS
OF , 4HE PAIR , , IS CALLED PRIMITIVE IF , IS A PROPER MAXIMAL SUB
ALGEBRA OF , WHICH CONTAINS NO NON ZERO IDEALS OF , 5SING THE 'UILLEMIN
3TERNBERG REALIZATION THEOREM ;'3= ;"L= IT IS EASY TO SHOW ;'= THAT THE ST
FORMULATION OF ,IES PROBLEM IS EQUIVALENT TO THE FOLLOWING MORE INVARIANT
FORMULATION
ND FORMULATION #LASSIFY ALL PRIMITIVE PAIRS , , WHERE , IS A
LINEARLY COMPACT ,IE ALGEBRA AND , IS ITS OPEN SUBALGEBRA
2ECALL THAT A TOPOLOGICAL ,IE ALGEBRA IS CALLED LINEARLY COMPACT IF ITS
UNDERLYING SPACE IS ISOMORPHIC TO A TOPOLOGICAL PRODUCT OF DISCRETELY TOPOL
OGIZED lNITE DIMENSIONAL VECTOR SPACES THE BASIC EXAMPLES ARE lNITE DIM
6ISIONS IN -ATH ).&).)4% 3)-0,% 350%239--%42)%3 !.$ 1&4
ENSIONAL SPACES WITH DISCRETE TOPOLOGY AND THE SPACE OF FORMAL POWER SERIES
IN X WITH FORMAL TOPOLOGY
!NY LINEARLY COMPACT ,IE ALGEBRA , CONTAINS AN OPEN HENCE OF lNITE
CODIMENSION SUBALGEBRA , ;'= (ENCE IF , IS SIMPLE CHOOSING ANY MAX
IMAL OPEN SUBALGEBRA , WE GET A PRIMITIVE PAIR , , /NE CAN SHOW
THAT THERE EXISTS A UNIQUE SUCH , AND THIS LEADS TO THE FOUR SERIES
AND 4HE REMAINING SERIES AND ARE NOT SIMPLE THEY ACTUALLY ARE THE
,IE ALGEBRAS OF DERIVATIONS OF AND BUT THE CHOICE OF , IS AGAIN UNIQUE
5SING THE STRUCTURE RESULTS ON GENERAL TRANSITIVE LINEARLY COMPACT ,IE
ALGEBRAS ;'= IT IS NOT DInCULT TO REDUCE IN THE INlNITE DIMENSIONAL CASE
THE CLASSIlCATION OF PRIMITIVE PAIRS TO THE CLASSIlCATION OF SIMPLE LINEARLY
COMPACT ,IE ALGEBRAS CF ;'= 3UCH A REDUCTION IS POSSIBLE ALSO IN THE ,IE
SUPERALGEBRA CASE BUT IT IS MUCH MORE COMPLICATED FOR TWO REASONS
A A SIMPLE LINEARLY COMPACT ,IE SUPERALGEBRA MAY HAVE SEVERAL MAXIMAL
OPEN SUBALGEBRAS
B CONSTRUCTION OF ARBITRARY PRIMITIVE PAIRS IN TERMS OF SIMPLE PRIMITIVE
PAIRS IS MORE COMPLICATED IN THE SUPERALGEBRA CASE
)N THE NEXT SECTIONS ) WILL DISCUSS IN SOME DETAIL THE CLASSIlCATION OF INlNITE
DIMENSIONAL SIMPLE LINEARLY COMPACT ,IE SUPERALGEBRAS
3TATEMENT OF THE -AIN 4HEOREM

4HE <SUPERIZATION BASICALLY AMOUNTS TO ADDING ANTICOMMUTING INDETERMI
NATES )N OTHER WORDS GIVEN AN ALGEBRA ASSOCIATIVE OR ,IE ! WE CONSIDER
THE 'RASSMANN ALGEBRA !HNI IN N ANTICOMMUTING INDETERMINATES w wN
OVER ! 4HIS ALGEBRA CARRIES A CANONICAL :: GRADATION CALLED PARITY
DElNED BY LETTING
P! PwI ::
&OR EXAMPLE #HNI IS THE 'RASSMANN ALGEBRA IN N INDETERMINATES OVER # )F
/M DENOTES THE ALGEBRA OF FORMAL POWER SERIES OVER # IN M INDETERMINATES
THEN /M HNI IS THE ALGEBRA OVER # OF FORMAL POWER SERIES IN M COMMUTING
INDETERMINATES X X XM AND N ANTICOMMUTING INDETERMINATES w
w wN
XI XJ XJ XI XI wJ wJ XI wI wJ _wJ wI
2ECALL THAT A DERIVATION $ OF PARITY P$ :: OF A :: GRADED
ALGEBRA IS A VECTOR SPACE ENDOMORPHISM SATISFYING CONDITION
$AB $A B _ P$ PA A$B
6' +!# '!&!
&URTHERMORE THE SUM OF THE SPACES OF DERIVATIONS OF PARITY AND IS CLOSED

UNDER THE <SUPER BRACKET
;$ $ = $$ _ _ P$ P$ $ $
4HIS <SUPER BRACKET SATISlES <SUPER ANALOGS OF ANTICOMMUTATIVITY AND
*ACOBI IDENTITY HENCE DElNES WHAT IS CALLED A ,IE SUPERALGEBRA
&OR EXAMPLE THE ALGEBRA !HNI HAS DERIVATIONS wI OF PARITY DElNED
BY
wI A FOR A ! wI wJ mIJ
AND THESE DERIVATIONS ANTICOMMUTE SO THAT ; wI wJ =
4HE <SUPER ANALOG OF THE ,IE ALGEBRA 7M IS THE ,IE SUPERALGEBRA DE
NOTED BY 7 MJN OF ALL CONTINUOUS DERIVATIONS OF THE :: GRADED ALGEBRA
/M HNI N : WITH THE DElNED ABOVE <SUPERBRACKET
y8 M N
8 z
7 MJN 0I X w 1J X w
XI wJ
I J
WHERE 0I X w 1J X w /M HNI )N A MORE GEOMETRIC LANGUAGE THIS IS THE

,IE SUPERALGEBRA OF ALL FORMAL VECTOR lELDS ON A SUPERMANIFOLD OF DIMEN
SION MJN
4HERE IS A UNIQUE WAY TO EXTEND DIVERGENCE FROM 7M TO 7 MJN SUCH
THAT THE DIVERGENCELESS VECTOR lELDS FORM A SUBALGEBRA
o8 8 p 8 8
0I 1
DIV 0I XI 1J wJ XI _ P1J wJJ
I J J J
AND THE <SUPER ANALOG OF 3M IS

g k h
3MJN 8 7 MJN k DIV 8
)N ORDER TO DElNE <SUPER ANALOGS OF THE (AMILTONIAN AND CONTACT ,IE
ALGEBRAS (M AND +M INTRODUCE A <SUPER ANALOG OF THE ALGEBRA OF DIkER
ENTIAL FORMS ;+= 4HIS IS AN ASSOCIATIVE ALGEBRA OVER /M HNI DENOTED BY
jMJN ON GENERATORS DX DXM Dw DwN AND DElNING RELATIONS
DXI DXJ _DXJ DXI DXI DwJ DwJ DXI DwI DwJ DwJ DwI
AND THE :: GRADATION DElNED BY
PXI PDwJ PwJ PDXI
4HE ALGEBRA jMJN CARRIES A UNIQUE CONTINUOUS DERIVATION D OF PARITY
SUCH THAT
DXI DXI DwJ DwJ DDXI DDwJ
6ISIONS IN -ATH ).&).)4% 3)-0,% 350%239--%42)%3 !.$ 1&4
4HE OPERATOR D HAS ALL THE USUAL PROPERTIES EG

8 8 F
F
DF XI DX I wJ DwJ FOR F /M HNI AND D
I J
!S USUAL FOR ANY 8 7 MJN ONE DElNES A DERIVATION r8 CONTRACTION

ALONG 8 OF THE ALGEBRA jMJN BY THE PROPERTIES
Pr8 P8 r8 XJ r8 DXJ _ P8 8XJ
4HE ACTION OF ANY 8 7 MJN ON /M HNI EXTENDS IN A UNIQUE WAY TO THE
ACTION BY A DERIVATION OF jMJN SUCH THAT ;8 D= 4HIS IS CALLED ,IES
DERIVATIVE AND IS USUALLY DENOTED BY ,8 BUT WE SHALL WRITE 8 IN PLACE OF
,8 UNLESS CONFUSION MAY ARISE /NE HAS THE USUAL #ARTAN FORMULA FOR THIS
ACTION ,8 ;D r8 =
5SING THIS ACTION ONE CAN DElNE SUPER ANALOGS OF THE (AMILTONIAN AND
CONTACT ,IE ALGEBRAS FOR ANY N :
g k h
(MJN 8 7 MJN k 8S
0 0
WHERE S KI DXI > DXKI NJ DwJ
g k h
+MJN 8 7 MJN k 8C F C
0 0
WHERE C DXM KI XI DXKI NJ wJ DwJ AND F /M HNI
.OTE THAT 7 JN 3JN AND (JN ARE lNITE DIMENSIONAL ,IE SUPER
ALGEBRAS 4HE ,IE SUPERALGEBRAS 7 JN AND 3JN ARE SIMPLE Ik N t
AND N t RESPECTIVELY (OWEVER (JN IS NOT SIMPLE AS ITS DERIVED ALGE
BRA ( JN HAS CODIMENSION IN (JN BUT ( JN IS SIMPLE Ik N t
4HUS IN THE ,IE SUPERALGEBRA CASE THE LISTS OF SIMPLE lNITE AND INlNITE
DIMENSIONAL ALGEBRAS ARE MORE CLOSELY RELATED THAN IN THE ,IE ALGEBRA CASE
4HE FOUR SERIES OF ,IE SUPERALGEBRAS ARE INlNITE DIMENSIONAL IF M t IN
WHICH CASE THEY ARE SIMPLE EXCEPT FOR 3JN 4HE DERIVED ALGEBRA 3 JN
HAS CODIMENSION IN 3JN AND 3 JN IS SIMPLE Ik N t
)N MY PAPER ;+= ) CONJECTURED THAT THE FOUR SERIES CONSTRUCTED ABOVE
EXHAUST ALL INlNITE DIMENSIONAL SIMPLE LINEARLY COMPACT ,IE SUPERALGEBRAS
2EMARKABLY THE SITUATION TURNED OUT TO BE MUCH MORE EXCITING
!S WAS POINTED OUT BY SEVERAL MATHEMATICIANS THE 3CHOUTEN BRACKET
;3+= MAKES THE SPACE OF POLYVECTOR lELDS ON A M DIMENSIONAL MANIFOLD INTO
A ,IE SUPERALGEBRA 4HE FORMAL ANALOG OF THIS IS THE FOLLOWING lFTH SERIES OF
SUPERALGEBRAS CALLED BY PHYSICISTS THE "ATALIN 6ILKOVISKI ALGEBRA
g k h
(/MJM 8 7 MJM k 8OS
6' +!# '!&!
0
WHERE OS M I DXI DwI IS AN ODD SYMPLECTIC FORM &URTHERMORE UNLIKE
IN THE (MJN CASE NOT ALL VECTOR lELDS OF (/MJN HAVE ZERO DIVERGENCE
WHICH GIVES RISE TO THE SIXTH SERIES
g k h
3(/MJM 8 (/MJM k DIV 8
4HE SEVENTH SERIES IS THE ODD ANALOG OF +MJN ;!,3=
g k h
+/MJM 8 7 MJM k 8OC F OC
0
WHERE OC DwM M I wI DXI XI DwI IS AN ODD CONTACT FORM /NE
CAN TAKE AGAIN THE DIVERGENCE VECTOR lELDS IN +/MJM IN ORDER TO
CONSTRUCT THE EIGHTH SERIES BUT THE SITUATION IS MORE INTERESTING )T TURNS
OUT THAT FOR EACH k # ONE CAN DElNE THE DEFORMED DIVERGENCE DIV k 8
;+O= ;+= SO THAT DIV DIV AND
g k h
3+/MJM k 8 +/MJM k DIVk 8
IS A SUBALGEBRA 4HE SUPERALGEBRAS (/MJM AND +/MJM ARE SIM
PLE Ik M t AND M t RESPECTIVELY 4HE DERIVED ALGEBRA 3(/ MJM
HAS CODIMENSION IN 3(/MJM AND IT IS SIMPLE Ik M t 4HE DE
RIVED ALGEBRA 3+/ MJM k IS SIMPLE Ik M t AND IT COINCIDES WITH
3+/MJM k UNLESS k OR M_ M WHEN IT HAS CODIMENSION
3OME OF THE EXAMPLES DESCRIBED ABOVE HAVE SIMPLE <lLTERED DEFORMA
TIONS ALL OF WHICH CAN BE OBTAINED BY THE FOLLOWING SIMPLE CONSTRUCTION
,ET , BE A SUBALGEBRA OF 7 MJN WHERE N IS EVEN 4HEN IT HAPPENS IN
THREE CASES THAT
,w eNJ wJ ,
IS DIkERENT FROM , BUT IS CLOSED UNDER BRACKET !S A RESULT WE GET THE
FOLLOWING THREE SERIES OF SUPERALGEBRAS 3 w JN ;+= 3(/ w MJM ;#H+=
AND 3+/w MJM M M ;+O= THE CONSTRUCTIONS IN ;+O= AND ;#H+= WERE
MORE COMPLICATED 7E THUS GET THE NINTH AND THE TENTH SERIES OF SIMPLE
INlNITE DIMENSIONAL ,IE SUPERALGEBRAS
3(/w MJM M t M EVEN
q r
w M
3+/ MJM M t M ODD
M
)T IS APPROPRIATE TO MENTION HERE THAT THE FOUR SERIES 7 JN 3JN
3 w JN AND ( JN ALONG WITH THE CLASSICAL SERIES S@MJN AND OSPMJN
STRANGE SERIES PN AND QN TWO EXCEPTIONAL SUPERALGEBRAS OF DIMENSION
AND AND A FAMILY OF DIMENSIONAL EXCEPTIONAL SUPERALGEBRAS ALONG
WITH THE MARVELOUS lVE EXCEPTIONAL ,IE ALGEBRAS COMPRISE A COMPLETE LIST
OF SIMPLE lNITE DIMENSIONAL ,IE SUPERALGEBRAS ;+=
6ISIONS IN -ATH ).&).)4% 3)-0,% 350%239--%42)%3 !.$ 1&4
! SURPRISING DISCOVERY WAS MADE IN ;3H= WHERE THE EXISTENCE OF THREE

EXCEPTIONAL SIMPLE INlNITE DIMENSIONAL ,IE SUPERALGEBRAS WAS ANNOUNCED
4HE PROOF OF THE EXISTENCE ALONG WITH ONE MORE EXCEPTIONAL EXAMPLE WAS
GIVEN IN ;3H= !N EXPLICIT CONSTRUCTION OF THESE FOUR EXAMPLES WAS GIVEN
LATER IN ;#H+= 4HE lFTH EXCEPTIONAL EXAMPLE WAS FOUND IN THE WORK
ON CONFORMAL ALGEBRAS ;#H+= AND INDEPENDENTLY IN ;3H= 4HE ALLEGED
SIXTH EXCEPTIONAL EXAMPLE %J OF ;+= TURNED OUT TO BE ISOMORPHIC TO
3+J ;#H+=
.OW ) CAN STATE THE MAIN THEOREM
4HEOREM ;+= 4HE COMPLETE LIST OF SIMPLE INlNITE DIMENSIONAL LINEARLY
COMPACT ,IE SUPERALGEBRAS CONSISTS OF TEN SERIES OF EXAMPLES DESCRIBED ABOVE
AND lVE EXCEPTIONAL EXAMPLES %J %J %J %J AND %J
)T HAPPENS THAT ALL INlNITE DIMENSIONAL SIMPLE LINEARLY COMPACT ,IE AL
GEBRAS , HAVE A UNIQUE TRANSITIVE PRIMITIVE ACTION ;'= 4HIS IS CERTAINLY
FALSE IN THE ,IE SUPERALGEBRA CASE (OWEVER IF , IS A SIMPLE LINEARLY COM
PACT ,IE SUPERALGEBRA OF TYPE 8MJN IT HAPPENS THAT M IS MINIMAL SUCH
THAT , ACTS ON A SUPER MANIFOLD OF DIMENSION MJN IE , y 7 MJN N
IS MINIMAL FOR THIS M AND , HAS A UNIQUE ACTION WITH SUCH MINIMAL MJN
)NCIDENTALLY IN ALL CASES THE GROWTH OF , EQUALS M 2ECALL THAT GROWTH
IS THE MINIMAL M FOR WHICH DIM ,,J IS BOUNDED BY 0 J WHERE 0 IS A
POLYNOMIAL OF DEGREE M
,ET ME NOW DESCRIBE THOSE LINEARLY COMPACT INlNITE DIMENSIONAL ,IE
SUPERALGEBRAS , THAT ALLOW A TRANSITIVE PRIMITIVE ACTION ,ET 3 BE A SIMPLE
LINEARLY COMPACT INlNITE DIMENSIONAL ,IE SUPERALGEBRA AND LET 3HNI DENOTE
AS BEFORE THE 'RASSMANN ALGEBRA OVER 3 WITH N INDETERMINATES 4HE ,IE
SUPERALGEBRA $ER3HNI OF ALL DERIVATIONS OF THE ,IE SUPERALGEBRA 3HNI IS
THE FOLLOWING SEMI DIRECT SUM
$ER3HNI $ER 3 HNI 7 JN
&OR A DESCRIPTION OF $ER 3 SEE ;+ 0ROPOSITION = $ENOTE BY ,3 N THE
SET OF ALL OPEN SUBALGEBRAS , OF $ER3HNI THAT CONTAIN 3HNI AND HAVE THE
PROPERTY THAT THE CANONICAL IMAGE OF , IN 7 JN IS A TRANSITIVE SUBALGEBRA
5SING A DESCRIPTION OF SEMI SIMPLE LINEARLY COMPACT ,IE SUPERALGEBRAS
SIMILAR TO THE ONE GIVEN BY 4HEOREM FROM ;+= CF ;#H= ;'= AND ;"L=
AND 0ROPOSITION FROM ;'= IT IS EASY TO DERIVE THE FOLLOWING RESULT
0ROPOSITION )F A LINEARLY COMPACT INlNITE DIMENSIONAL ,IE SUPERALGE
BRA , ALLOWS A TRANSITIVE PRIMITIVE ACTION THEN , IS ONE OF THE ALGEBRAS OF
THE SETS ,3 N
6' +!# '!&!
%XAMPLE #ONSIDER THE SEMIDIRECT SUM , 3HNI 2 WHERE 2 IS A

TRANSITIVE SUBALGEBRA OF 7 JN 4HEN ,, 3 HNI 2 IS A PRIMITIVE
PAIR IF 3 IS A MAXIMAL OPEN SUBALGEBRA OF 3 AND THESE ARE ALL PRIMITIVE
PAIRS IN THE CASE WHEN 3 $ER 3 /NE CAN ALSO REPLACE IN THIS CONSTRUCTION
3 BY $ER 3 AND 3 BY A MAXIMAL OPEN SUBALGEBRA OF $ER 3 HAVING NO NON
ZERO IDEALS OF $ER 3
%XPLANATION OF THE 0ROOF OF 4HEOREM

3TEP )NTRODUCE 7EISFEILERS lLTRATION ;7= OF , &OR THAT CHOOSE A MAX
IMAL OPEN SUBALGEBRA , OF , AND A MINIMAL SUBSPACE ,_ SATISFYING THE
PROPERTIES
,_ , ;, ,_ = y ,_
'EOMETRICALLY THIS CORRESPONDS TO A CHOICE OF A PRIMITIVE ACTION OF , AND AN
INVARIANT IRREDUCIBLE DIkERENTIAL SYSTEM 4HE PAIR ,_ , CAN BE INCLUDED
IN A UNIQUE lLTRATION
, ,_D z ,_D z ` ` ` z ,_ z , z , z ` ` `
/F COURSE IF , LEAVES INVARIANT NO NON TRIVIAL DIkERENTIAL SYSTEM THEN THE
<DEPTH D AND THE 7EISFEILER lLTRATION COINCIDE WITH THE CANONICAL
lLTRATION )NCIDENTALLY IN THE ,IE ALGEBRA CASE D ONLY FOR +N WHEN
D BUT IN THE ,IE SUPERALGEBRA CASE D IN THE MAJORITY OF CASES
4HE ASSOCIATED TO 7EISFEILERS lLTRATION : GRADED ,IE SUPERALGEBRA IS OF
THE FORM 'R, eJt_D =J AND HAS THE FOLLOWING PROPERTIES
' DIM =J SINCE CODIM ,
' =_J =J_ FOR J t BY MAXIMALITY OF ,
' ;X =_ = FOR X =J J t X BY SIMPLICITY OF ,
' = MODULE =_ IS IRREDUCIBLE BY CHOICE OF ,_
7EISFEILERS IDEA WAS THAT PROPERTY ' IS SO RESTRICTIVE THAT IT SHOULD
LEAD TO A COMPLETE CLASSIlCATION OF : GRADED ,IE ALGEBRAS SATISFYING ' [
' )NCIDENTALLY THE INlNITE DIMENSIONALITY OF , AND HENCE OF 'R, SINCE
, IS SIMPLE IS NEEDED ONLY IN ORDER TO CONCLUDE THAT = 4HIS INDEED
TURNED OUT TO BE THE CASE ;+= )N FACT MY IDEA WAS TO REPLACE THE CONDITION
OF lNITENESS OF THE DEPTH BY lNITENESS OF THE GROWTH WHICH ALLOWED ONE TO
ADD TO THE ,IE #ARTAN LIST SOME NEW ,IE ALGEBRAS CALLED NOWADAYS AnNE
+AC -OODY ALGEBRAS
(OWEVER UNLIKE IN THE ,IE ALGEBRA CASE IT IS IMPOSSIBLE TO CLASSIFY ALL
lNITE DIMENSIONAL IRREDUCIBLE FAITHFUL REPRESENTATIONS OF ,IE SUPERALGEBRAS
6ISIONS IN -ATH ).&).)4% 3)-0,% 350%239--%42)%3 !.$ 1&4
/NE NEEDED A NEW IDEA TO MAKE THIS APPROACH WORK

3TEP 4HE MAIN NEW IDEA IS TO CHOOSE , TO BE INVARIANT WITH RE
SPECT TO ALL INNER AUTOMORPHISMS OF , MEANING TO CONTAIN ALL EVEN AD
EXPONENTIABLE ELEMENTS OF , ! NON TRIVIAL POINT IS THE EXISTENCE OF SUCH ,
4HIS IS PROVED BY MAKING USE OF THE CHARACTERISTIC SUPERVARIETY WHICH IN
VOLVES RATHER DInCULT ARGUMENTS OF 'UILLEMIN ;'= THAT UNFORTUNATELY )
WAS UNABLE TO SIMPLIFY
.EXT USING A NORMALIZER TRICK OF 'UILLEMIN ;'= ) PROVE FOR THE ABOVE
CHOICE OF , THE FOLLOWING VERY POWERFUL RESTRICTION ON THE = MODULE =_
' ;= X= =_ FOR ANY NON ZERO EVEN ELEMENT X OF =_
3TEP #ONSIDER A FAITHFUL IRREDUCIBLE REPRESENTATION OF A ,IE SUPER
ALGEBRA F IN A lNITE DIMENSIONAL VECTOR SPACE 6 4HIS REPRESENTATION IS
CALLED STRONGLY TRANSITIVE IF
F ` X 6 FOR ANY NON ZERO EVEN ELEMENT X 6
.OTE THAT PROPERTY ' ALONG WITH ' ' FOR J AND ' SHOWS
THAT THE = MODULE =_ IS STRONGLY TRANSITIVE
)N ORDER TO DEMONSTRATE THE POWER OF THIS RESTRICTION CONSIDER lRST THE
CASE WHEN F IS A ,IE ALGEBRA AND 6 IS PURELY EVEN 4HEN THE STRONG TRAN
SITIVITY SIMPLY MEANS THAT 6 NFG IS A SINGLE ORBIT OF THE ,IE GROUP 0 COR
RESPONDING TO F )T IS RATHER EASY TO SEE THAT THE ONLY STRONGLY TRANSITIVE
SUBALGEBRAS F OF G@6 ARE G@6 S@6 SP6 AND CSP6 4HESE FOUR CASES LEAD TO
'R, WHERE , 7N 3N (N AND +N RESPECTIVELY
)N THE SUPER CASE THE SITUATION IS MUCH MORE COMPLICATED &IRST WE
CONSIDER THE CASE OF <INCONSISTENT GRADATION MEANING THAT =_ CONTAINS A
NON ZERO EVEN ELEMENT 4HE CLASSIlCATION OF SUCH STRONGLY TRANSITIVE MOD
ULES IS RATHER LONG AND THE ANSWER CONSISTS OF A DOZEN SERIES AND A HALF
DOZEN EXCEPTIONS SEE ;+ 4HEOREM = 5SING SIMILAR RESTRICTIONS ON
=_ =_ WE OBTAIN A COMPLETE LIST OF POSSIBILITIES FOR
'R,s gJs =J
IN THE CASE WHEN =_ CONTAINS NON ZERO EVEN ELEMENTS )T TURNS OUT THAT ALL
BUT ONE EXCEPTION ARE NOT EXCEPTIONS AT ALL BUT CORRESPOND TO THE BEGIN
NING MEMBERS OF SOME SERIES !S A RESULT ONLY %J <SURVIVES BUT THE
INFAMOUS %J DOES NOT
3TEP .EXT WE TURN TO THE CASE OF A CONSISTENT GRADATION IE WHEN
=_ IS PURELY ODD "UT THEN = IS AN <HONEST ,IE ALGEBRA HAVING A FAITHFUL
IRREDUCIBLE REPRESENTATION IN =_ CONDITION ' BECOMES VACUOUS !N
6' +!# '!&!
EXPLICIT DESCRIPTION OF SUCH REPRESENTATIONS IS GIVEN BY THE CLASSICAL #ARTAN

*ACOBSON THEOREM )N THIS CASE ) USE THE <GROWTH METHOD DEVELOPED IN
;+= AND ;+= TO DETERMINE A COMPLETE LIST OF POSSIBILITIES FOR 'R,s 4HIS
CASE PRODUCES MAINLY THE REMAINING FOUR EXCEPTIONS
3TEP IS RATHER LONG AND TEDIOUS ;#H+= &OR EACH 'R,s OBTAINED
IN 3TEPS AND WE DETERMINE ALL POSSIBLE <PROLONGATIONS IE INlNITE
DIMENSIONAL : GRADED ,IE SUPERALGEBRAS SATISFYING ' WHOSE NEGATIVE
PART IS THE GIVEN 'R,s
3TEP )T REMAINS TO RECONSTRUCT , FROM 'R, IE TO lND ALL POSSIBLE
lLTERED SIMPLE LINEARLY COMPACT ,IE SUPERALGEBRAS , WITH GIVEN 'R, SUCH
AN , IS CALLED A SIMPLE lLTERED DEFORMATION OF 'R, /F COURSE THERE IS A
TRIVIAL lLTERED DEFORMATION 'R, eJt_D =J WHICH IS SIMPLE Ik 'R, IS
)T IS PROVED IN ;#H+= BY A LONG AND TEDIOUS CALCULATION THAT ONLY
3(/MJM FOR M EVEN t AND 3+/MJM M M FOR M ODD t HAVE A
NON TRIVIAL SIMPLE lLTERED DEFORMATION WHICH ARE THE NINTH AND TENTH SERIES
)T WOULD BE NICE TO HAVE A MORE CONCEPTUAL PROOF 2ECALL THAT 3(/MJM
IS NOT SIMPLE THOUGH IT DOES HAVE A SIMPLE lLTERED DEFORMATION .OTE ALSO
THAT IN THE ,IE ALGEBRA CASE ALL lLTERED DEFORMATIONS ARE TRIVIAL
#ONSTRUCTION OF %XCEPTIONAL ,INEARLY #OMPACT ,IE

3UPERALGEBRAS
)N ORDER TO DESCRIBE THE CONSTRUCTION OF THE EXCEPTIONAL INlNITE DIMENSIONAL
,IE SUPERALGEBRAS GIVEN IN ;#H+= ) NEED TO MAKE SOME REMARKS ,ET
jM jMJ BE THE ALGEBRA OF DIkERENTIAL FORMS OVER /M LET jKM DENOTE
THE SPACE OF FORMS OF DEGREE K AND jKMC@ THE SUBSPACE OF CLOSED FORMS &OR
ANY t # THE REPRESENTATION OF 7M ON jKM CAN BE <TWISTED BY LETTING
8 ,8 t DIV 8 8 7M
TO GET A NEW 7M MODULE DENOTED BY jKM t THE SAME CAN BE DONE FOR
7MN /BVIOUSLY jKM t jKM WHEN RESTRICTED TO 3M 4HEN WE HAVE THE
FOLLOWING OBVIOUS 7M MODULE ISOMORPHISMS jM jM
M _ AND jM
jM M &URTHERMORE THE MAP 8 r8 DX > ` ` ` > DXM GIVES THE FOLLOWING
7M MODULE AND 3M MODULE ISOMORPHISMS
M_ M_
7M jM _ 3M jMC@
7E SHALL IDENTIFY THE REPRESENTATION SPACES VIA THESE ISOMORPHISMS
4HE SIMPLEST IS THE CONSTRUCTION OF THE LARGEST EXCEPTIONAL ,IE SUPER
ALGEBRA %J )TS EVEN PART IS THE ,IE ALGEBRA 3 ITS ODD PART IS THE
6ISIONS IN -ATH ).&).)4% 3)-0,% 350%239--%42)%3 !.$ 1&4
SPACE OF CLOSED FORMS jC@ 4HE REMAINING COMMUTATORS ARE DElNED AS

FOLLOWS FOR 8 3 jC@
;8 = ,8 ; = > jC@ 3
0
%ACH QUINTUPLE OF INTEGERS A A A SUCH THAT A I AI IS EVEN
DElNES A : GRADATION OF %J BY LETTING
DEG XI _ XI AI DEG DXI AI _ A
4HE QUINTUPLE DElNES THE ONLY CONSISTENT : GRADATION WHICH
HAS DEPTH %J eJt_ =J AND ONE HAS
= S@ AND =_ c # =_ #b AS = MODULES
&URTHERMORE eJt =J IS A MAXIMAL OPEN SUBALGEBRA OF %J THE ONLY
ONE WHICH IS INVARIANT WITH RESPECT TO ALL AUTOMORPHISMS 4HERE ARE THREE
OTHER MAXIMAL OPEN SUBALGEBRAS IN %J ASSOCIATED TO : GRADATIONS
CORRESPONDING TO QUINTUPLES AND AND
ONE CAN SHOW THAT THESE FOUR ARE ALL UP TO CONJUGACY MAXIMAL OPEN SUB
ALGEBRAS CF ;#H+=
!NOTHER IMPORTANT : GRADATION OF %J WHICH IS UNLIKE THE PRE
VIOUS FOUR BY INlNITE DIMENSIONAL SUBSPACES CORRESPONDS TO THE QUINTU
PLE AND HAS DEPTH %J ett_ =t /NE HAS =
%J AND THE =t FORM AN IMPORTANT FAMILY OF IRREDUCIBLE %J MODULES
;+2U= 4HE CONSISTENT : GRADATION OF %J INDUCES THAT OF = %J
eJt_ 7J WHERE
7 S@ g S@ g G@ 7_ # b # b # 7_ # b # b #
! MORE EXPLICIT CONSTRUCTION OF %J IS AS FOLLOWS ;#H+= THE EVEN PART
IS 7 j i S@ THE ODD PART IS j _ i # WITH THE OBVIOUS ACTION OF
THE EVEN PART AND THE BRACKET OF TWO ODD ELEMENTS IS DElNED AS FOLLOWS
; i U i V= > i U > V D > > D i U ` V
(ERE THE IDENTIlCATIONS j _ 7 AND j j _ ARE USED
4HE GRADATION OF %J CORRESPONDING TO THE QUINTUPLE
HAS DEPTH AND ITS TH COMPONENT IS ISOMORPHIC TO %J CF ;#H+=
4HE CONSTRUCTION OF %J IS ALSO VERY SIMPLE ;#H+= 4HE EVEN PART
IS 7 THE ODD PART IS j _ AND THE BRACKET OF TWO ODD ELEMENTS IS
; = D > > D j _ 7
4HE CONSTRUCTION OF %J IS SLIGHTLY MORE COMPLICATED AND WE REFER TO
;#H+= FOR DETAILS
6' +!# '!&!
#LASSIlCATION OF 3UPERCONFORMAL !LGEBRAS

3UPERCONFORMAL ALGEBRAS HAVE BEEN PLAYING AN IMPORTANT ROLE IN SUPERSTRING
THEORY AND IN CONFORMAL lELD THEORY (ERE ) WILL EXPLAIN HOW TO APPLY
4HEOREM TO THE CLASSIlCATION OF <LINEAR SUPERCONFORMAL ALGEBRAS "Y
A <LINEAR SUPERCONFORMAL ALGEBRA ) MEAN A ,IE SUPERALGEBRA = SPANNED
BY COEnCIENTS OF A lNITE FAMILY & OF PAIRWISE LOCAL lELDS SUCH THAT THE
FOLLOWING TWO PROPERTIES HOLD
FOR A B & THE SINGULAR PART OF /0% IS lNITE IE
8
;AZ BW = CJ W WJ mZ _ W A lNITE SUM
J
WHERE ALL CJ W #; W =&

= CONTAINS NO NON TRIVIAL IDEALS SPANNED BY COEnCIENTS OF lELDS FROM
A #; W = SUBMODULE OF #; W =&
4HIS PROBLEM GOES BACK TO THE PHYSICS PAPER ;23= SOME PROGRESS IN ITS
SOLUTION WAS MADE IN ;+= AND A COMPLETE SOLUTION WAS STATED IN ;+=
! COMPLETE CLASSIlCATION EVEN IN THE <QUADRATIC CASE SEEMS TO BE A MUCH
HARDER PROBLEM SEE ;&,= FOR SOME VERY INTERESTING EXAMPLES 4HE SIM
PLEST EXAMPLE IS THE LOOP ALGEBRA E = #;X X_ = i = CENTERLESS AnNE
+AC -OODY SUPER ALGEBRA g WHERE 0 = IS A SIMPLE lNITE DIMENSIONAL
h ,IE
SUPER ALGEBRA 4HEN & AZ N: XN iA Z _N_ A= AND ;AZ BW =
;A B=W mZ _ W 4HE NEXT EXAMPLE IS THE ,IE ALGEBRA 6ECT #a OF REG
ULAR VECTOR lELDS ON #a CENTERLESS
0 6IRASORO
_ D ÀLGEBRA & CONSISTS OF ONE
lELD THE 6IRASORO lELD ,Z _ N: XN DX Z _N_ AND ;,Z ,W =

W ,W mZ _ W ,W mW Z _ W
/NE OF THE MAIN THEOREMS OF ;$+= STATES THAT THESE ARE ALL EXAMPLES IN
THE ,IE ALGEBRA CASE 4HE STRATEGY OF THE PROOF IS THE FOLLOWING ,ET Z
AND CONSIDER THE lNITELY GENERATED #; = MODULE 2 #; =& $ElNE THE
<t BRACKET 2 i 2 #;t= i 2 BY THE FORMULA
8
;At B= tJ CJ
J
4HIS SATISlES THE AXIOMS OF A CONFORMAL SUPER ALGEBRA SEE ;$+= ;+=
SIMILAR TO THE ,IE SUPER ALGEBRA AXIOMS
I ; At B= _t;At B= ;At B= t ;At B=
II ;At B= _;B_t_ A=
III ;At ;Bu C== ;;At B=tu C= _ PA PB ;Bu ;At C==
6ISIONS IN -ATH ).&).)4% 3)-0,% 350%239--%42)%3 !.$ 1&4
4HE MAIN OBSERVATION OF ;$+= IS THAT A CONFORMAL SUPER ALGEBRA IS COM

PLETELY DETERMINED BY THE ,IE SUPER ALGEBRA SPANNED BY ALL COEnCIENTS OF
NEGATIVE POWERS OF Z OF THE lELDS AZ CALLED THE ANNIHILATION ALGEBRA ALONG
WITH AN EVEN LOCALLY NILPOTENT SURJECTIVE DERIVATION OF THE ANNIHILATION ALGE
BRA &URTHERMORE APART FROM THE CASE OF CURRENT ALGEBRAS THE COMPLETED
ANNIHILATION ALGEBRA TURNS OUT TO BE AN INlNITE DIMENSIONAL SIMPLE LINEARLY
COMPACT ,IE SUPER ALGEBRA OF GROWTH 3INCE IN THE ,IE ALGEBRA CASE THE
ONLY SUCH EXAMPLE IS 7 THE PROOF IS lNISHED
)N THE SUPERALGEBRA CASE THE SITUATION IS MUCH MORE INTERESTING SINCE
THERE ARE MANY INlNITE DIMENSIONAL SIMPLE LINEARLY COMPACT ,IE SUPERALGE
BRAS OF GROWTH "Y 4HEOREM THE COMPLETE LIST IS AS FOLLOWS
7 JN 3 JN +JN AND %J
4HE CORRESPONDING SUPERCONFORMAL ALGEBRAS IN THE lRST THREE CASES ARE DE
lNED IN THE SAME WAY EXCEPT THAT WE REPLACE / HNI BY #X HNI DENOTE
THEM BY 7N 3N AND +N RESPECTIVELY 4HE SUPERCONFORMAL ALGEBRAS
7N AND +N ARE SIMPLE FOR N t EXCEPT FOR + WHICH SHOULD BE RE
PLACED BY ITS DERIVED ALGEBRA +
AND 3N IS SIMPLE FOR N t 4HE UNIQUE
SUPERCONFORMAL ALGEBRA CORRESPONDING TO %J IS DENOTED BY #+ )TS
CONSTRUCTION IS MORE DInCULT AND MAY BE FOUND IN ;#H+= OR ;+=
(OWEVER THE SUPERCONFORMAL ALGEBRA WITH THE ANNIHILATION ALGEBRA

3 JN IS NOT UNIQUE SINCE THERE ARE TWO UP TO CONJUGACY EVEN SURJECTIVE
LOCALLY NILPOTENT DERIVATIONS OF 3 JN )N ORDER TO SHOW THIS WE MAY
ASSUME THAT BOTH AND 3 JN ARE IN 7 JN 5SING A CHANGE OF INDE
TERMINATES WE MAY ASSUME THAT X BUT THEN THE STANDARD VOLUME
FORM V THAT DElNES DIV MAY CHANGE TO 0 w V SINCE IT MUST BE ANNIHI
LATED BY X &URTHER CHANGE OF INDETERMINATES BRINGS THIS FORM TO V OR
TO w wN V 4HIS GIVES US ANOTHER SUPERCONFORMAL ALGEBRA WITH THE
ANNIHILATION ALGEBRA 3 JN N EVEN WHICH IS THE DERIVED ALGEBRA OF
g k _ ` h
3EN 8 7N k DIV w wN 8
4HE SITUATION IS MORE INTERESTING IN THE CASE N SINCE THE ALGEBRA
OF OUTER DERIVATIONS OF 3 J IS DIMENSIONAL ;0= BUT THIS GIVES NO NEW
SUPERCONFORMAL ALGEBRAS /NE ARGUES SIMILARLY IN THE CASE +JN 4HE
CASE %J IS CHECKED DIRECTLY 7E THUS HAVE ARRIVED AT THE FOLLOWING
THEOREM
4HEOREM ! COMPLETE LIST OF SUPERCONFORMAL ALGEBRAS CONSISTS OF LOOP
ALGEBRAS E
= WHERE = IS A SIMPLE lNITE DIMENSIONAL ,IE SUPERALGEBRA AND OF

THE FOLLOWING ,IE SUPERALGEBRAS N : 7N 3N 3EN
N EVEN
6' +!# '!&!
AND #+
+N N +
.OTE THAT THE lRST MEMBERS OF THE ABOVE SERIES ARE WELL KNOWN SUPER
ALGEBRAS 7 + IS THE 6IRASORO ALGEBRA + IS THE .EVEU 3CHWARZ
ALGEBRA + 7 IS THE . ALGEBRA + IS THE . ALGEBRA

3 IS THE . ALGEBRA 4HESE ALGEBRAS ALONG WITH 7 AND #+
ARE THE ONLY SUPERCONFORMAL ALGEBRAS FOR WHICH ALL lELDS ARE PRIMARY WITH
POSITIVE CONFORMAL WEIGHTS ;+= )T IS INTERESTING TO NOTE THAT ALL OF THEM
ARE CONTAINED IN #+ WHICH CONSISTS OF lELDS THE EVEN ONES ARE THE
6IRASORO lELDS AND CURRENTS THAT FORM SOE AND THE ODD ONES ARE AND
lELDS OF CONFORMAL WEIGHT AND RESPECTIVELY (ERE IS THE TABLE OF
INCLUSIONS WHERE IN SQUARE BRACKETS THE NUMBER OF lELDS IS INDICATED
#+ ;= z 7 ;= z 7 + ;= z + ;= z 6IR
; ;
+ ;=
3 ;=
!LL OF THESE ,IE SUPERALGEBRAS HAVE A UNIQUE NON TRIVIAL CENTRAL EXTENSION
EXCEPT FOR #+ THAT HAS NONE 4HE ONLY OTHER SUPERALGEBRAS LISTED BY
4HEOREM WHICH HAVE NON TRIVIAL CENTRAL EXTENSIONS ARE 3E
WITH ONE SUCH

EXTENSION AND + WITH TWO INEQUIVALENT CENTRAL EXTENSIONS 4HE PRES
ENCE OF A CENTRAL TERM IS NECESSARY FOR THE CONSTRUCTION OF AN INTERESTING
CONFORMAL lELD THEORY
2EPRESENTATIONS OF ,INEARLY #OMPACT ,IE 3UPERALGEBRAS

"Y A REPRESENTATION OF A LINEARLY COMPACT ,IE SUPERALGEBRA , WE SHALL MEAN
A CONTINUOUS REPRESENTATION IN A VECTOR SPACE 6 WITH DISCRETE TOPOLOGY
THEN THE CONTRAGREDIENT REPRESENTATION IS A CONTINUOUS REPRESENTATION IN
A LINEARLY COMPACT SPACE 6 b &IX AN OPEN SUBALGEBRA , OF , 7E SHALL
ASSUME THAT 6 IS LOCALLY , lNITE MEANING THAT ANY VECTOR OF 6 IS CONTAINED
IN A lNITE DIMENSIONAL , INVARIANT SUBSPACE THIS PROPERTY ACTUALLY OFTEN
IMPLIES THAT 6 IS CONTINUOUS 4HESE KINDS OF REPRESENTATIONS WERE STUDIED
IN THE ,IE ALGEBRA CASE BY 2UDAKOV ;2U=
)T IS EASY TO SHOW THAT SUCH AN IRREDUCIBLE , MODULE 6 IS A QUOTIENT OF AN
INDUCED MODULE )ND, , 5 5 , i5 , 5 WHERE 5 IS A lNITE DIMENSIONAL
IRREDUCIBLE , MODULE BY A UNIQUE IN GOOD CASES MAXIMAL SUBMODULE
4HE INDUCED MODULE )ND, , 5 IS CALLED DEGENERATE IF IT IS NOT IRREDUCIBLE !N
IRREDUCIBLE QUOTIENT OF A DEGENERATE INDUCED MODULE IS CALLED A DEGENERATE
IRREDUCIBLE MODULE SUCH A MODULE IS CALLED A TOP MODULE IF THE CORRESPOND
6ISIONS IN -ATH ).&).)4% 3)-0,% 350%239--%42)%3 !.$ 1&4
ING INDUCED MODULE CANNOT BE MAPPED INTO ANOTHER INDUCED MODULE

/NE OF THE MOST IMPORTANT PROBLEMS OF REPRESENTATION THEORY IS TO DE
TERMINE ALL DEGENERATE REPRESENTATIONS AND AMONG THEM ALL TOP REPRESEN
TATIONS ) WILL STATE HERE THE RESULT FOR , %J WITH , eJt 7J
SEE X SO THAT THE lNITE DIMENSIONAL IRREDUCIBLE , MODULES ARE ACTUALLY
7 S@ g S@ g G@ MODULES WITH eJ 7J ACTING TRIVIALLY 7E SHALL NOR
MALIZE THE GENERATOR 9 OF G@ BY THE CONDITION THAT ITS EIGENVALUE ON 7_ IS
_ 4HE lNITE DIMENSIONAL IRREDUCIBLE 7 MODULES ARE LABELED BY TRIPLES
MN B 9 WHERE MN RESP B ARE LABELS OF THE HIGHEST WEIGHT OF AN IRRE
DUCIBLE REPRESENTATION OF S@ RESP S@ SO THAT M AND M LABEL 3 M #
AND 3 M #b RESP B LABELS 3 B # AND 9 IS THE EIGENVALUE OF THE CENTRAL ELE
MENT 9 3INCE IRREDUCIBLE %J MODULES ARE UNIQUE QUOTIENTS OF INDUCED
MODULES THEY CAN BE LABELED BY THE ABOVE TRIPLES AS WELL
4HEOREM ;+2U= 4HE COMPLETE LIST OF IRREDUCIBLE DEGENERATE %J
MODULES IS AS FOLLOWS M B : B IF M IN THE SECOND SERIES
_ ` _ ` _ ` _ `
M B _B_ M_ M B B_ M M B _B M M B B M
4HE TOP %J MODULES ARE PRECISELY THOSE THAT ARE CONTRAGREDIENT TO THE
=_ AND = FROM THE : GRADATION OF DEPTH OF %J DESCRIBED IN X
4HEIR LABELS ARE _ FOR =_ AND FOR =
&UNDAMENTAL 0ARTICLE -ULTIPLETS

)N ORDER TO EXPLAIN THE CONNECTION OF REPRESENTATION THEORY OF LINEARLY COM
PACT ,IE SUPERALGEBRAS TO PARTICLE PHYSICS LET ME PROPOSE THE FOLLOWING
AXIOMATICS OF FUNDAMENTAL PARTICLES
! 4HE ALGEBRA OF SYMMETRIES IS A LINEARLY COMPACT ,IE SUPERALGEBRA ,
WITH AN ELEMENT 9 CALLED THE HYPERCHARGE OPERATOR SUCH THAT
I AD 9 IS DIAGONALIZABLE AND NORMALIZED SUCH THAT ITS SPECTRUM IS
BOUNDED BELOW y : AND y :
II THE CENTRALIZER OF 9 IN , IS 7 S@ S@ #9 ONE MAY WEAKEN
THIS BY REQUIRING z IN PLACE OF
" ! PARTICLE MULTIPLET IS AN IRREDUCIBLE SUBREPRESENTATION OF 7 IN A
DEGENERATE IRREDUCIBLE REPRESENTATION OF , 0ARTICLES IN A MULTIPLET
_ ARE
`
LINEARLY INDEPENDENT EIGENVECTORS OF THE S@ GENERATOR ) _
#HARGE 1 OF A PARTICLE IS GIVEN BY THE 'ELL -ANN .ISHIJIMA FORMULA

1 ) EIGENVALUE HYPERCHARGE
6' +!# '!&!
# &UNDAMENTAL PARTICLE MULTIPLET IS A PARTICLE MULTIPLET SUCH THAT

I J1J s FOR ALL PARTICLES OF THE MULTIPLET
II ONLY THE DIMENSIONAL THE TWO FUNDAMENTAL REPRESENTATIONS OR
THE ADJOINT REPRESENTATION OF S@ OCCUR
5SING 4HEOREM IT IS EASY TO CLASSIFY ALL FUNDAMENTAL MULTIPLETS WHEN
THE ALGEBRA OF SYMMETRIES , %J ;+2U= 4HE ANSWER IS GIVEN IN THE
LEFT HALF OF 4ABLE 4HE RIGHT HALF CONTAINS ALL THE FUNDAMENTAL PARTICLES
OF THE 3TANDARD MODEL SEE EG ;/K= THE UPPER PART IS COMPRISED OF THREE
GENERATIONS OF QUARKS AND THE MIDDLE PART OF THREE GENERATIONS OF LEPTONS
THESE ARE ALL FUNDAMENTAL FERMIONS FROM WHICH MATTER IS BUILT AND THE
LOWER PART IS COMPRISED OF FUNDAMENTAL BOSONS WHICH MEDIATE THE STRONG
AND ELECTRO WEAK INTERACTIONS %XCEPT FOR THE LAST LINE THE MATCH IS PERFECT
MULTIPLETS CHARGES PARTICLES

_U, ` _C, ` _T, `
_ D, S, B,
_UE2 ` _EC2 ` _ET2 `
_ _ DE2 SE2 EB2
_ _ U

E, E
C, E
T2
U2 C2 T2
_ _ D2 S2 B2
DE, SE, EB,

_v, ` _vu, ` _v{ , `
_ _ E, u, {,
_vE2 ` _vEu2 ` _vE{ 2 `
E2 u
E {E2
2
E, u
E, {E,
_ _ E2 u2 {2
GLUONS
_ 7 7 _ : GAUGE BOSONS
l PHOTON
e e [
4ABLE
6ISIONS IN -ATH ).&).)4% 3)-0,% 350%239--%42)%3 !.$ 1&4
3PECULATIONS AND 6ISIONS

!S THE TITLE OF THE CONFERENCE SUGGESTS EACH SPEAKER IS EXPECTED TO PROPOSE
HIS OR HER VISIONS IN MATHEMATICS FOR THE ST CENTURY 4HIS IS AN OBVIOUS
INVITATION TO BE IRRESPONSIBLY SPECULATIVE 3OME OF THE ITEMS PROPOSED
BELOW ARE OF THIS NATURE BUT SOME OTHERS ARE LESS SO
)T IS CERTAINLY IMPOSSIBLE TO CLASSIFY ALL SIMPLE INlNITE DIMENSIONAL ,IE
ALGEBRAS OR SUPERALGEBRAS 4HE MOST POPULAR TYPES OF CONDITIONS THAT
HAVE EMERGED IN THE PAST YEARS AND THAT ) LIKE MOST ARE THESE
A EXISTENCE OF A GRADATION BY lNITE DIMENSIONAL SUBSPACES AND
lNITENESS OF GROWTH ;+= ;-=
B TOPOLOGICAL CONDITIONS ;'= ;+= XX[
C THE CONDITION OF LOCALITY ;$+= ;+= X
0ROBLEM A IN THE ,IE ALGEBRA CASE HAS BEEN COMPLETELY SOLVED IN
;-= BUT AN ANALOGOUS CONJECTURE IN THE ,IE SUPERALGEBRA CASE ;+,= IS
APPARENTLY MUCH HARDER
#ONCERNING B LET ME STATE A CONCRETE PROBLEM ,ET , #X N
WHERE #X IS THE SPACE OF FORMAL ,AURENT SERIES IN X WITH FORMAL
TOPOLOGY %XAMPLES OF SIMPLE TOPOLOGICAL ,IE ALGEBRAS WITH THE UN
DERLYING SPACE , ARE THE COMPLETED CENTERLESS AnNE AND 6IRASORO
ALGEBRAS !RE THERE ANY OTHER EXAMPLES
)NCIDENTALLY AFTER GOING TO THE DUAL 4HEOREM GIVES A COMPLETE
CLASSIlCATION OF SIMPLE ,IE CO SUPERALGEBRAS
)N X ) EXPLAINED HOW TO USE CLASSIlCATION OF SIMPLE LINEARLY COMPACT
,IE SUPERALGEBRAS OF GROWTH IN ORDER TO CLASSIFY SIMPLE <LINEAR /0%
OF CHIRAL lELDS IN DIMENSIONAL CONFORMAL lELD THEORY 7ILL #+
PLAY A ROLE IN PHYSICS OR IS IT JUST AN EXOTIC ANIMAL !RE THE LINEARLY
COMPACT ,IE SUPERALGEBRAS OF GROWTH IN ANY WAY RELATED TO /0%
OF HIGHER DIMENSIONAL QUANTUM lELD THEORIES
%ACH OF THE FOUR TYPES 7 3 ( + OF SIMPLE PRIMITIVE ,IE ALGEBRAS
, , CORRESPOND TO THE FOUR MOST IMPORTANT TYPES OF GEOMETRIES
OF MANIFOLDS ALL MANIFOLDS ORIENTED MANIFOLDS SYMPLECTIC AND CON
TACT MANIFOLDS 3INCE EVERY SMOOTH SUPERMANIFOLD OF DIMENSION MJN
COMES FROM A RANK N VECTOR BUNDLE ON A M DIMENSIONAL MANIFOLD IT IS
NATURAL TO EXPECT THAT EACH OF THE SIMPLE PRIMITIVE ,IE SUPERALGEBRAS
CORRESPONDS TO ONE OF THE MOST IMPORTANT TYPES OF GEOMETRIES OF VECTOR
BUNDLES ON MANIFOLDS &OR EXAMPLE THE lVE EXCEPTIONAL SUPERALGE
BRAS HAVE ALTOGETHER UP TO CONJUGACY MAXIMAL OPEN SUBALGEBRAS
6' +!# '!&!
4HEY CORRESPOND TO IRREDUCIBLE : GRADATIONS LISTED IN ;#H+= FOR

%J FOR %J FOR %J FOR %J AND FOR %J AS
3HCHEPOCHKINA POINTED OUT WE MISSED TWO : GRADATIONS OF %J
AND IN NOTATION OF ;#H+= 4HERE
SHOULD BE THEREFORE EXCEPTIONAL TYPES OF GEOMETRIES OF VECTOR BUN
DLES ON MANIFOLDS WHICH ARE ESPECIALLY IMPORTANT
4HE MAIN MESSAGE OF X OF MY TALK IS THE FOLLOWING PRINCIPLE
.ATURE LIKES DEGENERATE REPRESENTATIONS
4HERE ARE SEVERAL THEORIES WHERE THIS PRINCIPLE WORKS VERY WELL &IRST
IS THE THEORY OF DIMENSIONAL STATISTICAL LATTICE MODELS ESPECIALLY
THE MINIMAL MODELS OF ;"0:= WHICH ARE FOR C NOTHING ELSE
BUT THE TOP DEGENERATE REPRESENTATIONS OF THE 6IRASORO ALGEBRA AND
THE 7:7 MODELS INCLUDING THE CASE OF FRACTIONAL LEVELS WHICH ARE
BASED ON DEGENERATE TOP MODULES OVER THE AnNE +AC -OODY ALGEBRAS
SEE ;+= FOR A REVIEW ON THESE MODULES 3ECOND IS THE THEORETICAL
EXPLANATION OF THE QUANTUM (ALL EkECT BY ;#4:= BASED ON DEGENERATE
TOP MODULES OF 7 SEE ;+2=
)N VIEW OF THE DISCUSSION IN XX AND IT IS NATURAL TO SUGGEST THAT
THE ALGEBRA SU SU U OF INTERNAL SYMMETRIES OF THE 7EINBERG
3ALAM 'LASHOW 3TANDARD MODEL EXTENDS TO %J ) AM HOPEFUL
THAT REPRESENTATION THEORY WILL SHED NEW LIGHT ON VARIOUS FEATURES
OF THE 3TANDARD MODEL INCLUDING THE +OBAYASHI -ASKAWA MATRIX
)T TURNS OUT ;+2U= THAT ALL DEGENERATE %J 6ERMA MODULES HAVE A
UNIQUE NON TRIVIAL SINGULAR VECTOR 4HIS SHOULD LEAD TO SOME CANONICAL
DIkERENTIAL EQUATIONS ON THE CORRELATION FUNCTIONS CF ;"0:=
) lND IT QUITE REMARKABLE THAT THE 35 'RAND UNIlED MODEL OF 'EORGI
'LASHOW COMBINES THE LEFT MULTIPLETS OF FUNDAMENTAL FERMIONS IN PRE
CISELY THE NEGATIVE PART OF THE CONSISTENT GRADATION OF %J SEE
X 4HIS IS PERHAPS AN INDICATION OF THE POSSIBILITY THAT AN EXTENSION
FROM SU TO %J ALGEBRA OF INTERNAL SYMMETRIES MAY RESOLVE THE
DInCULTIES WITH THE PROTON DECAY
/NE OF COURSE MAY TRY OTHER lNITE OR INlNITE DIMENSIONAL ,IE SUPER
ALGEBRAS &OR EXAMPLE * VAN DER *EUGT HAS TRIED RECENTLY , S@J
AND IT WORKED RATHER NICELY BUT OSPJ HAS BEEN RULED OUT
,ET ME END WITH THE MOST IRRESPONSIBLE SUGGESTION 3INCE 7 IS ON
THE ONE HAND THE ALGEBRA OF SYMMETRIES OF %INSTEINS GRAVITY THEORY
AND ON THE OTHER HAND THE EVEN PART OF %J IT IS A NATURAL GUESS
THAT %J IS THE ALGEBRA OF SYMMETRIES OF A NICE SUPER EXTENSION OF
6ISIONS IN -ATH ).&).)4% 3)-0,% 350%239--%42)%3 !.$ 1&4
GENERAL RELATIVITY /NE KNOWS THAT THE ALGEBRA OF SYMMETRIES OF THE

MINIMAL . SUPERGRAVITY THEORY IS 3J ;/3=
2EFERENCES
;!,3= $ !LEXSEEVSKI $ ,EITES ) 3HCHEPOCHKINA %XAMPLES OF SIMPLE
,IE SUPERALGEBRAS OF VECTOR lELDS #2 !CAD "UL 3CI [

;"0:= !! "ELAVIN !- 0OLYAKOV !- :AMOLODCHIKOV )NlNITE CON
FORMAL SYMMETRY OF CRITICAL mUCTUATIONS IN TWO DIMENSIONS * 3TAT 0HYS
[
;"L= 2* "LATTNER )NDUCED AND PRODUCED REPRESENTATIONS OF ,IE ALGEBRAS
4RANS !MER -ATH 3OC
;"L= 2* "LATTNER ! THEOREM OF #ARTAN AND 'UILLEMIN * $Ik 'EOM
[
;#4:= ! #APPELLI #! 4RUGENBERGER '2 :EMBA 3TABLE HIERARCHAL
QUANTUM (ALL mUIDS AS 7 MINIMAL MODELS .UCL 0HYS "
[
;#A= % #ARTAN ,ES GROUPES DES TRANSFORMATIONS CONTINUsES INlNIS SIMPLES
!NN 3CI %COLE .ORM 3UP [
;#H= 3 * #HENG $IkERENTIABLY SIMPLE ,IE SUPERALGEBRAS AND REPRESENTATIONS
OF SEMISIMPLE ,IE SUPERALGEBRAS * !LGEBRA [
;#H+= 3 * #HENG 6' +AC ! NEW . SUPERCONFORMAL ALGEBRA #OM
MUN -ATH 0HYS [
;#H+= 3 * #HENG 6' +AC 'ENERALIZED 3PENCER COHOMOLOGY AND lLTERED
DEFORMATIONS OF : GRADED ,IE SUPERALGEBRAS !DV 4HEOR -ATH 0HYS
[
;#H+= 3 * #HENG 6' +AC 3TRUCTURE OF SOME : GRADED ,IE SUPERALGEBRAS
OF VECTOR lELDS 4RANSFORMATION 'ROUPS [
;$+= ! $!NDREA 6' +AC 3TRUCTURE THEORY OF lNITE CONFORMAL ALGEBRAS
3ELECTA -ATHEMATICA [
;&,= %3 &RADKIN 69A ,INETSKY #LASSIlCATION OF SUPERCONFORMAL AND
QUASISUPERCONFORMAL ALGEBRAS IN TWO DIMENSIONS 0HYS ,ETT "
[
;'= 67 'UILLEMIN ! *ORDAN ( OLDER DECOMPOSITION FOR A CERTAIN CLASS OF
INlNITE DIMENSIONAL ,IE ALGEBRAS * $Ik 'EOM [
;'= 67 'UILLEMIN )NlNITE DIMENSIONAL PRIMITIVE ,IE ALGEBRAS * $Ik
'EOM [
;'3= 67 'UILLEMIN 3 3TERNBERG !N ALGEBRAIC MODEL OF TRANSITIVE DIF
FERENTIAL GEOMETRY "ULL !MER -ATH 3OC [
;+= 6' +AC 3IMPLE IRREDUCIBLE GRADED ,IE ALGEBRAS OF lNITE GROWTH -ATH
5332 )ZVESTIJA [
6' +!# '!&!
;+= 6' +AC ,IE SUPERALGEBRAS !DV -ATH [

;+= 6' +AC -ODULAR INVARIANCE IN MATHEMATICS AND PHYSICS IN <-ATHE
MATICS INTO THE ST #ENTURY !-3 #ENTENNIAL 0UBL [
;+= 6' +AC 6ERTEX !LGEBRAS FOR "EGINNERS 5NIVERSITY ,ECTURE 3ERIES
!-3 0ROVIDENCE 2) 3ECOND EDITION
;+= 6' +AC 4HE IDEA OF LOCALITY IN <0HYSICAL !PPLICATIONS AND -ATHEMAT
ICAL !SPECTS OF 'EOMETRY 'ROUPS AND !LGEBRAS ($ $OEBNER ET AL
EDS 7ORLD 3CI 3INGAPORE [
;+= 6' +AC 3UPERCONFORMAL ALGEBRAS AND TRANSITIVE GROUP ACTIONS ON
QUADRICS #OMM -ATH 0HYS [
;+= 6' +AC #LASSIlCATION OF INlNITE DIMENSIONAL SIMPLE LINEARLY COMPACT
,IE SUPERALGEBRAS !DV -ATH [
;+,= 6' +AC * VAN DER ,EUR /N CLASSIlCATION OF SUPERCONFORMAL ALGE
BRAS IN <3TRINGS 3* 'ATES ET AL EDS 7ORLD 3CI [
;+2= 6' +AC !/ 2ADUL 2EPRESENTATION THEORY OF 7 4RANSFORMA
TION 'ROUPS [
;+2U= 6' +AC !. 2UDAKOV )RREDUCIBLE REPRESENTATIONS OF LINEARLY COM
PACT ,IE SUPERALGEBRAS IN PREPARATION
;+O= 9U +OCHETKOFF $sEFORMATIONS DE SUPERALGsEBRES DE "UTTIN ET QUANTIl
CATION #2 !CAD 3CI 0ARIS SER ) [
;,= 3 ,IE 4HEORIE DER TRANSFORMATIONS 'RUPPEN -ATH !NN [

;-= / -ATHIEU #LASSIlCATION OF SIMPLE GRADED ,IE ALGEBRAS OF lNITE GROWTH
)NVENT -ATH [
;/3= 6) /GIEVETSKII %3 3OKACHEV 4HE SIMPLEST GROUP OF %INSTEIN SU
PERGRAVITY 3OV * .UCL 0HYS [
;/K= , /KUN 0HYSICS OF %LEMENTARY 0ARTICLES .AUKA IN 2USSIAN
;0= % 0OLETAEVA 3EMI INlNITE COHOMOLOGY AND SUPERCONFORMAL ALGEBRAS
PREPRINT
;23= 0 2AMOND *( 3CHWARZ #LASSIlCATION OF DUAL MODEL GAUGE ALGEBRAS
0HYS ,ETT " [
;2U= !. 2UDAKOV )RREDUCIBLE REPRESENTATIONS OF INlNITE DIMENSIONAL ,IE AL
GEBRAS OF #ARTAN TYPE -ATH 5332 )ZVESTIJA [
;3+= *! 3CHOUTEN 7 VAN DER +ULK 0FAkS 0ROBLEM AND ITS 'ENERAL
IZATIONS #LARENDON 0RESS
;3H= ) 3HCHEPOCHKINA .EW EXCEPTIONAL SIMPLE ,IE SUPERALGEBRAS #2 "UL
3CI [
;3H= ) 3HCHEPOCHKINA 4HE lVE EXCEPTIONAL SIMPLE ,IE SUPERALGEBRAS OF VEC
TOR lELDS PREPRINT
;3I3= )- 3INGER 3 3TERNBERG /N THE INlNITE GROUPS OF ,IE AND #ARTAN )
* !NALYSE -ATH [
;3T= 3 3TERNBERG &EATURED REVIEW OF <#LASSIlCATION OF INlNITE DIMENSIONAL
6ISIONS IN -ATH ).&).)4% 3)-0,% 350%239--%42)%3 !.$ 1&4
SIMPLE LINEARLY COMPACT ,IE SUPERALGEBRAS BY 6ICTOR +AC -ATH 2EVIEWS

M
;7= "9 7EISFEILER )NlNITE DIMENSIONAL lLTERED ,IE ALGEBRAS AND THEIR CON
NECTION WITH GRADED ,IE ALGEBRAS &UNCT !NAL !PPL [
6ICTOR +AC $EPARTMENT OF -ATHEMATICS -)4 #AMBRIDGE -! 53!

KAC MATHMITEDU
'%/-%42):!4)/. ). 2%02%3%.4!4)/. 4(%/29
$ +AZHDAN
4HE COMMON THEME OF THE ARTICLES <l FUNCTION OF 2EPRESENTATIONS AND

,IFTING AND <'EOMETRIC AND 5NIPOTENT #RYSTALS IS THE GEOMETRIZATION OF
FAMILIAR OBJECTS )N THE lRST ARTICLE WE INTERPRET SOME SPECIAL FUNCTIONS
APPEARING IN THE THEORY OF REPRESENTATIONS AS MATERIALIZATIONS OF GEOMET
RIC OBJECTS WHICH WE CALL ALGEBRAIC GEOMETRIC DISTRIBUTIONS IN THE SECOND
ONE WE INTERPRET ELEMENTS OF THE CRYSTALLINE BASIS WHICH PARAMETRIZE THE
SPECIAL BASIS OF lNITE DIMENSIONAL REPRESENTATIONS OF REDUCTIVE GROUPS AS
MATERIALIZATIONS OF GEOMETRIC CRYSTALS -OREOVER THESE TWO TOPICS ARE VERY
MUCH INTERWOVEN SINCE THE THEORY OF GEOMETRIC CRYSTALS IS NECESSARY FOR THE
CONSTRUCTION OF OUR ALGEBRAIC GEOMETRIC DISTRIBUTIONS
"Y DElNITION AN ALGEBRAIC GEOMETRIC DISTRIBUTION h ON AN ALGEBRAIC VA
RIETY 8 OVER A lELD & IS QUADRUPLE h 9 P F WHERE 9 IS AN & VARIETY
P 9 8 IS A RATIONAL MORPHISM F IS RATIONAL FUNCTION AND IS A TOP FORM
ON 9 )F & IS A LOCAL lELD AND IS A NON TRIVIAL ADDITIVE CHARACTER OF & THEN
WE CAN ASSOCIATE WITH AN ALGEBRAIC GEOMETRIC DISTRIBUTION h 9 P F
A DISTRIBUTION h P F JJ ON 8& WHERE JJ IS A MEASURE ON THE SET
9& CORRESPONDING TO THE TOP FORM AND P F JJ IS THE PUSH FORWARD
OF THE COMPLEX VALUED MEASURE F JJ
)F 7 IS A lNITE GROUP ACTING ON 8 THEN A LIFT OF THIS ACTION TO h IS A
RATIONAL ACTION OF 7 ON 9 COMPATIBLE WITH P AND PRESERVING F AND 'IVEN
AN ACTION OF 7 ON h WE CAN DElNE AN ALGEBRAIC GEOMETRIC DISTRIBUTION
h 7 97 P F WHICH WE CALL THE DESCENT OF h AND THEREFORE A
DISTRIBUTION h7 ON 87 & 7E CALL IT <THE DESCENT OF h /F COURSE
THE DESCENT IS NOT DETERMINED BY THE DISTRIBUTION h AND AN ACTION OF 7 ON
8 BUT DEPENDS ON THE CHOICE OF THE ACTION OF 7 TO h
/NE OF THE TWO MAIN IDEAS OF THE ARTICLE <l FUNCTIONS OF REPRESENTATIONS
AND LIFTING IS TO FORMULATE A CONJECTURE THAT FOR ANY REDUCTIVE GROUP '
OVER A LOCAL lELD & AND A REPRESENTATION y OF THE ,ANGLANDS DUAL GROUP '?
OF THE GROUP ' THE DISTRIBUTION h' y ON '& RESPONSIBLE FOR THE LIFTING
Ly OF y CAN BE OBTAINED AS A DESCENT OF THE ANALOGOUS DISTRIBUTION h4 y FOR
THE MAXIMAL TORUS 4 OF ' )T IS EASY TO CONSTRUCT THE DISTRIBUTION h4 y AS
4
A MATERIALIZATION OF AN ALGEBRAIC GEOMETRIC DISTRIBUTION hy AND THEREFORE
6ISIONS IN -ATH '%/-%42):!4)/. ). 2%02%3%.4!4)/. 4(%/29
A CONJECTURAL DESCRIPTION OF THE LIFTING Ly OF y IS REDUCED TO AN ALGEBRA

GEOMETRIC QUESTION ABOUT CONSTRUCTION OF AN ACTION OF 7 ON h 4 y
4HE ARTICLE <'EOMETRIC AND UNIPOTENT CRYSTALS DEALS WITH A GEOMETRIZA
TION OF THE CONCEPT OF CRYSTALS WHICH WE BELIEVE IS NECESSARY FOR A CONSTRUC
TION OF AN ACTION OF 7 ON h4 y
)F YOU WISH TO CONSTRUCT A REPRESENTATION OF A GROUP WE HAVE lRST OF
ALL TO CONSTRUCT THE SPACE 6 OF THIS REPRESENTATION "UT HOW TO CONSTRUCT
A VECTOR SPACE 4HE SIMPLEST WAY IS TO REALIZE 6 AS A SPACE OF COMPLEX
VALUED FUNCTIONS ON A SET " )N OTHER WORDS ONE HAS TO lND A BASIS VB
B " OF 6 )T WAS DISCOVERED BY ,USZTIG AND +ASHIWARA ;,U= ;+= THAT
IN THE CASE WHEN 6 IS AN IRREDUCIBLE REPRESENTATION OF A SEMISIMPLE GROUP
' ONE CAN CONSTRUCT A BASIS VB OF 6 SUCH THAT THE PARAMETER SPACE " HAS
AN INTERESTING CRYSTALLINE STRUCTURE
3INCE THE THEORY OF CRYSTALS IS NOT WIDELY KNOWN ) WILL GIVE WITH A SHORT
INTRODUCTION TO THIS THEORY
,ET xN 3, # %ND 2N N t BE AN IRREDUCIBLE REPRESENTATION
OF DIMENSION N 4HEN WE CAN WRITE ANY lNITE DIMENSIONAL REPRESENTATION
6 OF 3, # AS A DIRECT SUM 6 gNt 6N WHERE 6N 2N i (OM2N 6
,ET VB B " BE A BASIS OF 6 $ElNE &N gsMsN 6M &OR ANY B "
WE DENOTE BY VvB THE IMAGE OF THE VECTOR VB &N &N_ WHERE N NB
IS THE SMALLEST NUMBER SUCH THAT VB &N 7E DENOTE BY 2B y &N &N_
THE MINIMAL 3, # INVARIANT SUBSPACE OF &N &N_ CONTAINING VvB "Y THE
CONSTRUCTION 2B IS A SPACE OF A REPRESENTATION OF 3, #
7E SAY THAT A BASIS VB IS COMPATIBLE WITH THE ACTION OF THE ,IE GROUP
3, # IF
A &OR ANY B " VECTOR VB y 6 IS A ( EIGENVECTOR WHERE ( y 3, #
IS THE DIAGONAL SUBGROUP 7E DENOTE BY l^B THE CORRESPONDING CHAR
ACTER OF (
B &OR ANY N t THERE EXISTS A SUBSET "N y " SUCH THAT THE SUBSPACE
&N IS EQUAL TO THE SPAN OF VB B "N
C &OR ANY B "N THE SPACE 2B IS SPANNED BY THE IMAGES OF ELEMENTS OF
"N IN &N &N_
D &OR ANY B " THE REPRESENTATION OF 3, # ON 2B IS IRREDUCIBLE
,ET " y 3, # RESP 5 y 3, # BE THE SUBGROUP OF UPPER
TRIANGULAR RESP UNIPOTENT UPPER TRIANGULAR MATRICES AND LET SL BE THE
,IE ALGEBRA OF 3, # #HOOSE A NON ZERO ELEMENT E IN THE ,IE ALGEBRA
OF 5
'IVEN A lNITE DIMENSIONAL REPRESENTATION 6 OF 3, # AND A BASIS VB
$ +!:($!. '!&!
B " OF 6 COMPATIBLE WITH THE ACTION OF 3, # WE DElNE A PARTIAL

BIJECTIONS E^ ON " IN THE FOLLOWING WAY 'IVEN B " WE SAY THAT E^B IS
DElNED IF EvVB )N SUCH A CASE IT IS EASY TO SEE THAT THERE EXISTS A UNIQUE
ELEMENT E^B " SUCH THAT VE^B &NB AND THAT Ev VB IS PROPORTIONAL
TO VvE^B
,ET ' BE A SEMISIMPLE GROUP OVER # 4 y ' BE A MAXIMAL TORUS 8 b 4
THE GROUP OF CHARACTERS OF 4 AND " y ' A "OREL SUBGROUP OF ' CONTAINING 4
7E DENOTE BY ) THE SET OF VERTICES OF THE $YNKIN DIAGRAM OF ' &OR ANY
I ) WE DENOTE BY ,I y ' THE SUBGROUP GENERATED BY THE ROOT SUBGROUPS
5I AND 5_I !S IS WELL KNOWN ,I IS QUOTIENT OF 3, #
,ET 6 BE A lNITE DIMENSIONAL REPRESENTATION OF ' AND VB B " BE A
BASIS OF 6 7E SAY THAT A BASIS VB IS COMPATIBLE WITH THE ACTION OF THE ,IE
GROUP ' IF FOR ANY I ) THE BASIS VB IS COMPATIBLE WITH THE ACTION ON ,I
!S FOLLOWS FROM ;"'= FOR ANY lNITE DIMENSIONAL REPRESENTATION 6 OF ' THERE
EXISTS A BASIS VB B " OF 6 COMPATIBLE WITH THE ACTION OF '
,ET 6 BE A lNITE DIMENSIONAL REPRESENTATION OF ' AND VB B " BE A
BASIS OF 6 COMPATIBLE WITH THE ACTION OF ' 3INCE THE #ARTAN SUBGROUPS
(I y ,I GENERATE 4 FOR ANY B " THE VECTOR VB IS A 4 EIGENVECTOR 7E
DENOTE BY l^ B y 8 b 4 THE CORRESPONDING CHARACTER OF 4 "Y DElNITION FOR
ANY I ) THE BASIS VB IS COMPATIBLE WITH THE RESTRICTION TO ,I 7E DENOTE BY
EÎ THE CORRESPONDING PARTIAL BIJECTION ON " AND BY FÎ THE INVERSE BIJECTION
)T IS EASY TO CHECK THAT FOR ANY B " I ) WE HAVE l^ ^ EI B l^ B jI
WERE jI 8 b 4 IS THE SIMPLE ROOT CORRESPONDING TO I 4HE DATA CONSISTING
OF l^ " 8 b 4 AND PARTIAL BIJECTIONS EÎ I ) ON " DElNE A STRUCTURE OF
A PRE CRYSTAL ON " SEE DETAILS IN THE APPENDIX TO <'EOMETRIC AND 5NIPOTENT
#RYSTALS !S FOLLOWS FORM THE CONSTRUCTION IN ;"'= IT IS NATURAL TO CALL THE
DATE l^ EÎ I ) ON " A PRE CRYSTAL STRUCTURE ASSOCIATED WITH THE ,ANGLANDS
DUAL GROUP '? OF '
)T WOULD BE VERY INTERESTING TO CONSTRUCT AN ANALOG OF A CRYSTAL ASSOCI
ATED WITH INlNITE DIMENSIONAL REPRESENTATIONS OF REDUCTIVE GROUPS OVER LOCAL
lELDS 7E DO NOT KNOW HOW TO lND SUCH A CONSTRUCTION BUT WE CAN TRY TO
ANALYZE THE PROBLEM &INITE DIMENSIONAL REPRESENTATIONS 6 OF REDUCTIVE
GROUPS HAVE DIkERENT SIZES AND THEREFORE THE SET " "6 PARAMETRIZ
ING BASES OF 6 ARE DIkERENT /N THE OTHER HAND THE MAJORITY OF INlNITE
DIMENSIONAL REPRESENTATIONS OF REDUCTIVE GROUPS OVER LOCAL lELDS ARE OF THE
SAME SIZE 4HEREFORE IT IS NATURAL TO LOOK FOR SOME <UNIVERSAL INlNITE SETS
" WITH CRYSTAL STRUCTURE 3UCH INlNITE <FREE CRYSTAL APPEARED EARLIER
IN THE WORKS OF +ASHIWARA /NE COULD HOPE THAT THE SPACE OF A GENERIC
6ISIONS IN -ATH '%/-%42):!4)/. ). 2%02%3%.4!4)/. 4(%/29
REPRESENTATION OF A REDUCTIVE GROUP '& WHERE & IS A LOCAL lELD CAN BE

REALIZED IN THE SPACE OF FUNCTIONS ON & b i "
4HE MAIN PURPOSE OF THE PAPER <'EOMETRIC AND UNIPOTENT CRYSTALS IS A
CONSTRUCTION OF A GEOMETRIC ANALOG OF THE NOTION '? CRYSTAL )T CONSISTS OF
AN ALGEBRAIC VARIETY 8 WITH A RATIONAL MORPHISM l 8 4 AND A FAMILY EI
I ) OF BIRATIONAL ACTIONS OF THE MULTIPLICATIVE GROUP 'M ON 8 SATISFYING
SOME COMPATIBILITY CONDITIONS 4HE SET " IS ISOMORPHIC TO THE SET OF
CONNECTED COMPONENTS OF THE SPACE ,8 OF FORMAL LOOPS ON A PARTICULAR
ALGEBRAIC VARIETY 8 AND PARTIAL BIJECTIONS EÎ COME FROM THE ACTIONS EI OF THE
GROUP 'M ON 8
4O OUR SURPRISE THESE GEOMETRIC CRYSTALS ARE IMPORTANT FOR lNDING THE
ACTION OF THE 7EYL GROUP 7 CENTRAL FOR THE CONSTRUCTION OF THE DESCENT OF
ALGEBRAIC GEOMETRIC DISTRIBUTIONS
2EFERENCES
;"'= ! "RAVERMAN $ 'AITSGORY #RYSTALS VIA THE AnNE 'RASSMANNIAN
PREPRINT MATH !'
;+= - +ASHIWARA #RYSTAL BASES OF MODIlED QUANTIZED ENVELOPING ALGEBRA
$UKE -ATH * [
;,U= ' ,USZTIG #ANONICAL BASES ARISING FROM QUANTIZED ENVELOPING ALGEBRAS *
!MER -ATH 3OC [
$AVID +AZHDAN $EPARTMENT OF -ATHEMATICS (ARVARD 5NIVERSITY #AMBRIDGE

-! 53!
'%/-%42)# !.$ 5.)0/4%.4 #2934!,3
!RKADY "ERENSTEIN AND $AVID +AZHDAN
#ONTENTS
)NTRODUCTION
$ElNITIONS AND -AIN 2ESULTS ON 'EOMETRIC #RYSTALS

'ENERAL NOTATION
'EOMETRIC PRE CRYSTALS AND GEOMETRIC CRYSTALS
4HE GEOMETRIC CRYSTAL 8W
0OSITIVE GEOMETRIC CRYSTALS AND THEIR TROPICALIZATION
4RIVIALIZATION OF GEOMETRIC CRYSTALS
5NIPOTENT #RYSTALS

$ElNITION OF UNIPOTENT CRYSTALS AND THEIR PRODUCT
&ROM UNIPOTENT ' CRYSTALS TO UNIPOTENT , CRYSTALS
&ROM UNIPOTENT ' CRYSTALS TO GEOMETRIC , CRYSTALS
0OSITIVE UNIPOTENT CRYSTALS AND DUALITY
$IAGONALIZATION OF THE PRODUCTS OF UNIPOTENT CRYSTALS
5NIPOTENT ACTION OF THE 7EYL GROUP
%XAMPLES OF 5NIPOTENT #RYSTALS

3TANDARD UNIPOTENT CRYSTALS
-ULTIPLICATION OF STANDARD UNIPOTENT CRYSTALS
7 INVARIANT FUNCTIONS ON STANDARD UNIPOTENT ' CRYSTALS
0OSITIVE STRUCTURES ON STANDARD UNIPOTENT CRYSTALS
$UALITY AND SYMMETRIES FOR STANDARD UNIPOTENT CRYSTALS
4HE RESEARCH OF BOTH AUTHORS WAS SUPPORTED IN PART BY .3& GRANTS

6ISIONS IN -ATH '%/-%42)# !.$ 5.)0/4%.4 #2934!,3
0ROOFS
0ROOFS OF RESULTS IN SECTION
0ROOF OF RESULTS IN SECTION
0ROJECTIONS OF "RUHAT #ELLS TO THE 0ARABOLIC 3UBGROUPS

'ENERAL FACTS ON PROJECTIONS
%XAMPLE OF THE PROJECTION FOR ' ',MN
!PPENDIX #OMBINATORIAL 0RE CRYSTALS AND 7 CRYSTALS
)NTRODUCTION
,ET ' BE A SPLIT SEMISIMPLE ALGEBRAIC GROUP OVER 1 = BE THE ,IE ALGEBRA OF '
AND 5Q = BE THE CORRESPONDING QUANTIZED ENVELOPING ALGEBRA ,USZTIG HAS
INTRODUCED IN ;,U= CANONICAL BASES FOR lNITE DIMENSIONAL 5Q = MODULES
!BOUT THE SAME TIME +ASHIWARA INTRODUCED IN ;+= CRYSTAL BASES AS A NATU
RAL FRAMEWORK FOR PARAMETRIZING BASES OF lNITE DIMENSIONAL 5Q = MODULES
)T WAS SHOWN IN ;,U= THAT +ASHIWARAS CRYSTAL BASES ARE THE LIMITS AS Q
OF ,USZTIGS CANONICAL BASES ,ATER IN ;+= +ASHIWARA INTRODUCED A NEW
COMBINATORIAL CONCEPT [ CRYSTALS +ASHIWARAS CRYSTALS GENERALIZE THE CRYS
TAL BASES AND PROVIDE A NATURAL FRAMEWORK FOR THEIR STUDY
)N THIS PAPER WE STUDY GEOMETRIC CRYSTALS AND UNIPOTENT CRYSTALS WHICH
ARE ALGEBRO GEOMETRIC ANALOGUES OF RESPECTIVELY THE CRYSTALS AND THE CRYSTAL
BASES
/UR APPROACH OF THE <GEOMETRIZATION OF COMBINATORIAL OBJECTS COMES
BACK TO THE WORKS OF ,USZTIG /N THE ONE HAND ,USZTIG CONSTRUCTED FOR
EACH REDUCED DECOMPOSITION I OF THE LONGEST ELEMENT W OF THE 7EYL GROUP
7 OF ' A PARAMETRIZATION I OF THE CANONICAL BASIS " BY THE CONE :t L
WHERE L IS THE LENGTH W /N THE OTHER HAND IN HIS STUDY OF THE POSITIVE
ELEMENTS OF '2 ,USZTIG CONSTRUCTED FOR EACH I A BIRATIONAL ISOMORPHISM
x I !L F
5 WHERE 5 IS THE UNIPOTENT RADICAL OF THE "OREL SUBGROUP OF '
-OREOVER ,USZTIG OBSERVED THAT FOR ANY TWO DECOMPOSITIONS I I OF W

THE PIECEWISE LINEAR TRANSFORMATION I _ m I OF :t L IS SIMILAR TO THE

BIRATIONAL TRANSFORMATION x I _ m x I OF !L SEE ;,U= ;"&:=
4HIS ANALOGY WAS FURTHER DEVELOPED IN A SERIES OF PAPERS ;"&:= ;"3=
;&:= 4HE MAIN IDEA OF THESE PAPERS IS TO INTRODUCE THE NOTION OF THE <TROP
ICALIZATION [ A SET OF FORMAL RULES FOR THE PASSAGE FROM BIRATIONAL ISOMOR
PHISMS TO PIECEWISE LINEAR MAPS AND TO STUDY PIECEWISE LINEAR AUTOMOR
! "%2%.34%). !.$ $ +!:($!. '!&!
PHISMS OF " USING THE <TROPICALIZATION OF EXPLICIT BIRATIONAL ISOMORPHISMS

x I !L F
5
!NOTHER GEOMETRIC APPROACH TO CONSTRUCTING CRYSTAL BASES IS SUGGESTED IN
THE RECENT WORK OF "RAVERMAN AND 'AITSGORY ;"R'= 7ITHIN THIS APPROACH
THE CRYSTAL BASES AND IN FACT ACTUAL BASES IN lNITE DIMENSIONAL ' MODULES
ARE CONSTRUCTED IN TERMS OF PERVERSE SHEAVES ON THE AnNE 'RASSMANNIAN
OF '
)N THE PRESENT PAPER WE GO IN THE OPPOSITE DIRECTION [ WE STUDY GEOMETRIC
CRYSTALS IN THEIR OWN RIGHT AND <GEOMETRIZE SOME OF THE PIECEWISE LINEAR
STRUCTURES OF THE CRYSTAL BASES MENTIONED ABOVE
,ET ) BE THE SET OF VERTICES OF THE $YNKIN DIAGRAM OF ' AND 4 BE A
MAXIMAL TORUS OF ' 4HE STRUCTURE OF A GEOMETRIC CRYSTAL ON AN ALGEBRAIC
VARIETY 8 CONSISTS OF A RATIONAL MORPHISM l 8 4 AND A COMPATIBLE
FAMILY EI 'M a 8 8 I ) OF RATIONAL ACTIONS OF THE MULTIPLICATIVE
GROUP 'M ON 8 7E SHOW THAT SUCH A STRUCTURE INDUCES A RATIONAL ACTION OF
THE 7EYL GROUP 7 ON 8 3URPRISINGLY MANY INTERESTING RATIONAL ACTIONS OF
7 COME FROM GEOMETRIC CRYSTALS &OR EXAMPLE THE NATURAL ACTION OF 7 ON
'ROTHENDIECKS SIMULTANEOUS RESOLUTION ' ^ ' COMES FROM A STRUCTURE OF
^
A GEOMETRIC CRYSTAL ON ' !LSO ALL THE EXAMPLES OF THE ACTION OF 7 IN ;"R+=
COME FROM GEOMETRIC CRYSTALS !NOTHER APPLICATION OF GEOMETRIC CRYSTALS
IS A CONSTRUCTION FOR ANY 3,N CRYSTAL BUILT ON 8 l OF A TRIVIALIZATION [ A
7 EQUIVARIANT ISOMORPHISM 8 F l _ E a 4
)T IS ALSO INTERESTING THAT THE ,ANGLANDS DUAL GROUP , ' EMERGES WHEN
WE RECONSTRUCT +ASHIWARAS COMBINATORIAL CRYSTALS OUT OF POSITIVE GEOMETRIC
CRYSTALS SEE SECTION 4HE PRESENCE OF , ' IN THE <CRYSTAL WORLD HAS
BEEN NOTICED IN ;,U= 4HE COMBINATORIAL RESULTS OF ;":= ALSO INVOLVE , '
! NUMBER OF STATEMENTS IN THE PAPER ARE ALMOST IMMEDIATE 7E CALL
THEM LEMMAS 0ROOFS OF ALL THE LEMMAS FROM SECTIONS AND ARE LEFT TO
THE READER
4HE MATERIAL OF THE PAPER IS ORGANIZED AS FOLLOWS
n )N SECTION WE INTRODUCE GEOMETRIC CRYSTALS AND FORMULATE THEIR MAIN
PROPERTIES
n )N SECTION WE INTRODUCE UNIPOTENT CRYSTALS AS AN ALGEBRO GEOMETRIC
ANALOGUE OF CRYSTAL BASES FOR 5Q , = MODULES WHERE , = IS THE ,IE ALGEBRA
OF , ' ! UNIPOTENT CRYSTAL BUILT ON AN IRREDUCIBLE ALGEBRAIC VARIETY 8 CON
SISTS OF A RATIONAL ACTION OF 5 ON 8 AND A RATIONAL 5 EQUIVARIANT MORPHISM
F 8 '5 4HE UNIPOTENT CRYSTALS HAVE A NUMBER OF PROPERTIES CHARAC
TERISTIC OF THE CRYSTAL BASES /NE OF THE MAIN PROPERTIES IS A NATURAL PRODUCT
6ISIONS IN -ATH '%/-%42)# !.$ 5.)0/4%.4 #2934!,3
OF UNIPOTENT CRYSTALS WHICH IS AN ANALOGUE OF THE TENSOR PRODUCT OF CRYSTAL

BASES 7E PROVE THAT THE CATEGORY OF UNIPOTENT CRYSTALS IS STRICTLY MONOIDAL
WITH RESPECT TO THIS PRODUCT 7E ALSO DElNE THE NOTION OF A DUAL UNIPOTENT
CRYSTAL /NE OF THE MAIN RESULTS OF THE SECTION IS A CONSTRUCTION FOR ANY
UNIPOTENT CRYSTAL ON 8 OF AN INDUCED GEOMETRIC CRYSTAL 8 7E ALSO GIVE A
CLOSED FORMULA FOR THE ACTION OF 7 ON THIS INDUCED GEOMETRIC CRYSTAL 8
n )N SECTION WE STUDY THE STANDARD UNIPOTENT CRYSTALS WHICH ARE THE
5 ORBITS "W"" y '" 7E OBTAIN A DECOMPOSITION OF ANY STANDARD
CRYSTAL "W"" INTO THE PRODUCT OF DIMENSIONAL CRYSTALS CORRESPONDING
TO ANY REDUCED DECOMPOSITION OF W 7 4HIS IS A GEOMETRIZATION OF
+ASHIWARAS RESULTS 7E ALSO CONSTRUCT 7 INVARIANT FUNCTIONS ON CERTAIN
STANDARD UNIPOTENT CRYSTALS 4HESE FUNCTIONS ARE IMPORTANT FOR THE STUDY
OF l FUNCTIONS OF REPRESENTATIONS , ' SEE ;"R+=
n )N SECTION WE COLLECT PROOFS OF THE RESULTS FROM SECTIONS AND
n )N SECTION WE STUDY THE RESTRICTIONS TO A STANDARD ,EVI SUBGROUP
, y ' OF STANDARD UNIPOTENT ' CRYSTALS "W"" )N PARTICULAR FOR EACH
W 7 WE CONSTRUCT A RATIONAL MORPHISM PW "W"" ,5 < , )N
THE CASE WHEN PW IS A BIRATIONAL ISOMORPHISM WITH ITS IMAGE WE STUDY THE
DIRECT IMAGE OF 7, INVARIANT FUNCTIONS UNDER PW )N THE SPECIAL CASE WHEN
' ',MN , ',M a',N AND W W, `W IT IS POSSIBLE TO WRITE
EXPLICITLY THE CORRESPONDING ACTION OF THE GROUP 7, ON THE IMAGE OF PW
AND THE 7, EQUIVARIANT TRIVIALIZATION 4HIS SIMPLIlCATION OF THE STRUCTURE
OF THE CRYSTAL ON "W"" IS USED IN ;"R+= FOR THE PROOF THAT 0IATETSKI
3HAPIROS l FUNCTION TO ',M a ',N IS EQUAL TO THE ONE INTRODUCED IN
;"R+= 7E EXPECT THAT IN GENERAL THE ACTION OF 7, ON "W"" AND THE
TRIVIALIZATION OF THE CORRESPONDING CRYSTAL WOULD BE EASY TO DESCRIBE AND
THAT THIS DESCRIPTION CAN HELP IN THE STUDY OF l FUNCTIONS ON ,
n )N THE !PPENDIX WE COLLECT THE NECESSARY DElNITIONS AND RESULTS RELATED
TO +ASHIWARAS CRYSTALS 7E REFER TO THESE CRYSTALS AS COMBINATORIAL IN ORDER
TO DISTINGUISH BETWEEN THEM AND THEIR GEOMETRIC COUNTERPARTS
4HE NOTION OF A GEOMETRIC CRYSTAL WAS INTRODUCED BY THE lRST AUTHOR WHO
NOTICED THAT THE FORMULAS FOR THE 7 ACTION WHICH APPEAR IN THE DElNITION
OF l FUNCTIONS FOR THE GROUP ', a ', SEE ;"R+= CAN BE INTERPRETED
AS A <GEOMETRIZATION OF THE 7 ACTION FOR THE FREE COMBINATORIAL CRYSTAL
DElNED EARLIER BY THE lRST AUTHOR 7E WANT TO EXPRESS OUR GRATITUDE TO
! "RAVERMAN FOR HIS HELP AT DIkERENT STAGES OF THIS WORK AND IN PARTICULAR
FOR THE DElNITION OF THE NOTION OF DEGREE IN SECTION 7E ARE ALSO GRATEFUL
TO 9 &LICKER FOR NUMEROUS REMARKS ABOUT THE PAPER
! "%2%.34%). !.$ $ +!:($!. '!&!
$ElNITIONS AND -AIN 2ESULTS ON 'EOMETRIC #RYSTALS

'ENERAL NOTATION ,ET ' BE A SPLIT REDUCTIVE ALGEBRAIC GROUP OVER
1 AND 4 y ' A MAXIMAL TORUS 7E DENOTE BY c? AND c THE LATTICES
OF CO CHARACTERS AND CHARACTERS OF 4 AND BY H` Ì THE EVALUATION PAIRING
c? a c :
,ET " BE A "OREL SUBGROUP CONTAINING 4 $ENOTE BY ) THE SET OF VERTICES
OF THE $YNKIN DIAGRAM OF ' FOR ANY I ) DENOTE BY jI c THE SIMPLE
ROOT jI 4 'M AND BY jI? c? THE SIMPLE COROOT jI? 'M 4
,ET " _ y ' BE THE "OREL SUBGROUP CONTAINING 4 SUCH THAT " <" _ 4
$ENOTE BY 5 AND 5 _ RESPECTIVELY THE UNIPOTENT RADICALS OF " AND " _
&OR EACH I ) WE DENOTE BY 5I y 5 AND 5I_ y 5 _ THE CORRESPONDING
SIMPLE ROOT SUBGROUPS AND DENOTE BY wI 5 5I wI_ 5 _ 5I_ THE
CANONICAL PROJECTIONS
7E lX A FAMILY OF ISOMORPHISMS XI 'A F 5I_ I ) SUCH
5I YI 'A F
THAT q r q r
A A
XI A YI A YI jI
?
AA
XI
AA AA
#LEARLY EACH ISOMORPHISM YI IS UNIQUELY DETERMINED BY XI
2EMARK %ACH PAIR XI YI DElNES A HOMOMORPHISM }I 3, '
q r oCp q r
A B ? B
}I YI jI A XI
C D A A
$ENOTE BY 5 B AND 5 B _ RESPECTIVELY THE SPACES (OM5 'A AND
_
(OM5 'A &OR EACH I ) DElNE ~I 5B ~_ 5
B _ BY
I
~I U X_
I wI U ~_ _ _
I U_ YI wI U_
FOR U_ 5 _ U 5 "Y DElNITION THE FAMILY ~I I ) IS A BASIS IN THE
VECTOR SPACE 5 B AND THE FAMILY ~_ I ) IS A BASIS IN THE VECTOR SPACE 5 B _
I
B
&OR ANY ~ 5 AND ANY ~ _ 5B _ DElNE FUNCTIONS ~ ~ _ 5 _ ` 4 ` 5 '
A
BY
~ _ U_ TU ~_ U_ ~U_ TU ~U
4HE 7EYL GROUP 7 .ORM' 4 4 OF ' IS GENERATED BY SIMPLE RE
mECTIONS SI 7 I ) 4HE GROUP 7 ACTS ON LATTICES c c? BY SI t
t _ HjI? tI jI FOR t c SI t? t? _ Ht? jI I jI? FOR t? c?
,ET L 7 :t W LW BE THE LENGTH FUNCTION &OR ANY SEQUENCE
I I IL ) L WE WRITE WI SI ` ` ` SIL ! SEQUENCE I ) L IS CALLED
REDUCED IF THE LENGTH OF WI IS EQUAL TO L &OR ANY W 7 WE DENOTE BY
2W THE SET OF ALL REDUCED SEQUENCES I SUCH THAT WI W 7E DENOTE BY
W 7 THE ELEMENT OF THE MAXIMAL LENGTH IN 7
6ISIONS IN -ATH '%/-%42)# !.$ 5.)0/4%.4 #2934!,3
&OR I ) DElNE SI ' BY

SI XI YI _ XI
%ACH SI BELONGS TO .ORM' 4 AND IS A REPRESENTATIVE OF SI 7 )T
IS WELL KNOWN ;"O= THAT THE ELEMENTS SI I ) SATISFY THE BRAID RELA
TIONS 4HEREFORE WE CAN ASSOCIATE TO EACH W 7 ITS STANDARD REPRESENTATIVE
W .ORM' 4 IN SUCH A WAY THAT FOR ANY I IL 2W WE HAVE
W SI ` ` ` SIL
'EOMETRIC PRE CRYSTALS AND GEOMETRIC CRYSTALS ,ET 8 AND
9 BE ALGEBRAIC VARIETIES OVER 1 $ENOTE BY 28 9 THE SET OF ALL RATIONAL
MORPHISMS FROM 8 TO 9
&OR ANY F 28 9 DENOTE BY DOMF y 8 THE MAXIMAL OPEN SUBSET
OF 8 ON WHICH F IS DElNED DENOTE BY FREG DOMF 9 THE CORRESPOND
ING REGULAR MORPHISM 7E DENOTE BY RANF y 9 THE CLOSURE OF THE CON
STRUCTIBLE SET FREG DOMF IN 9 &OR ANY REGULAR MORPHISM F 8 9
WHERE 8 y 8 IS A DENSE SUBSET WE DENOTE BY ;F = 8 9 THE CORRESPOND
ING RATIONAL MORPHISM .OTE THAT ;FREG = F AND DOMFREG DOMF FOR
ANY F 28 9
)T IS EASY TO SEE THAT FOR ANY IRREDUCIBLE ALGEBRAIC VARIETIES 8 9 : AND
RATIONAL MORPHISMS F 8 9 G 9 : SUCH THAT DOMG INTERSECTS
RANF NON TRIVIALLY THE COMPOSITION F G F m G IS WELL DElNED AND IS A
RATIONAL MORPHISM FROM 8 TO :
7E DENOTE BY 6 THE CATEGORY WHOSE OBJECTS ARE IRREDUCIBLE ALGEBRAIC
VARIETIES AND ARROWS ARE DOMINANT RATIONAL MORPHISMS
&OR ANY ALGEBRAIC GROUP ( WE CALL A RATIONAL ACTION j ( a 8 8
UNITAL IF DOMj z FEG a 8
$EFINITION ,ET 8 /B6 AND l BE A RATIONAL MORPHISM 8 4 ! GE
OMETRIC ' PRE CRYSTAL OR SIMPLY GEOMETRIC PRE CRYSTAL ON 8 l IS A FAMILY
EI 'M a 8 8 I ) OF UNITAL RATIONAL ACTIONS OF THE MULTIPLICATIVE
GROUP 'M C X ECI X SUCH THAT lECI X jI? C lX
2EMARK 'EOMETRIC PRE CRYSTALS ARE ANALOGUES OF FREE COMBINATORIAL PRE
CRYSTALS SEE THE !PPENDIX 5NDER THIS ANALOGY THE VARIETY 8 CORRESPONDS
TO THE SET " THE MAXIMAL TORUS 4 CORRESPONDS TO THE LATTICE c? THE RATIONAL
MORPHISM l 8 4 CORRESPONDS TO THE MAP l E " c? AND RATIONAL
ACTIONS EI OF 'M ON 8 CORRESPOND TO BIJECTIONS EÎ " " 7E WILL MAKE
THIS ANALOGY PRECISE IN SECTION
'IVEN A GEOMETRIC PRE CRYSTAL 8 AND A REDUCED SEQUENCE II IL ) L
! "%2%.34%). !.$ $ +!:($!. '!&!
WE DElNE A RATIONAL MORPHISM EI 4 a 8 8 BY

j T jL T
T X ETI X EI m ` ` ` m EIL X
WHERE jK SIL SIL_ ` ` ` SIK jIK K L ARE THE ASSOCIATED POSITIVE
ROOTS
$EFINITION ! GEOMETRIC PRE CRYSTAL 8 ON 8 l IS CALLED A GEOMETRIC
CRYSTAL IF FOR ANY W 7 AND ANY I I 2W ONE HAS
EI EI
,EMMA 4HE RELATIONS ARE EQUIVALENT TO THE FOLLOWING RELATIONS
BETWEEN EI EJ FOR I J )
ECI ECJ ECJ ECI
IF HjI? jJ I
ECI ECJ C ECI ECJ ECI C ECJ
IF HjI? jJ I HjJ? jI I _
C C C C
ECI EJ ECI C ECJ ECJ ECI C EJ ECI
IF HjI? jJ I _ HjJ? jI I _
C C C C C C C C C C C C
ECI EJ EI EJ ECI C ECJ ECJ ECI C EJ EI EJ ECI
IF HjI? jJ I _ HjJ? jI I _
2EMARKS 4HE RELATIONS [ ARE MULTIPLICATIVE ANALOGUES OF
THE 6ERMA RELATIONS IN THE UNIVERSAL ENVELOPING ALGEBRA 5 , = SEE ;,U
0ROPOSITION =
!N ANALOGUE OF THE RELATIONS FOR A COMBINATORIAL ',N
CRYSTAL WAS CONSIDERED IN ;"+ 4HEOREM =
,EMMA ,ET 8 BE A GEOMETRIC PRE CRYSTAL &OR ANY t c I )
tlX
THE FORMULA X EI X DElNES A RATIONAL MORPHISM 8 8 4HIS IS
A BIRATIONAL ISOMORPHISM 8 F 8 IF AND ONLY IF HjI? tI F_ G )N THE
LATTER CASE THE INVERSE MORPHISM IS GIVEN BY THE FORMULA

EtlX X IF Hj? tI _
I I
X
E tlX X IF Hj? tI
I I
,ET 8 BE A GEOMETRIC CRYSTAL &OR EACH W 7 WE DElNE A RATIONAL

MORPHISM EW 4 a 8 8 BY THE FORMULA EW EI FOR ANY I 2W
SEE
6ISIONS IN -ATH '%/-%42)# !.$ 5.)0/4%.4 #2934!,3
0ROPOSITION 4HE CORRESPONDENCE 7 a 8 8 DElNED BY

_
W X WX ElX
W X
IS A RATIONAL UNITAL ACTION OF 7 ON 8

j lX
2EMARK 4HE FORMULA SI X EI I X IS A MULTIPLICATIVE ANALOGUE OF
IN THE !PPENDIX
4HE GEOMETRIC CRYSTAL 8W ,ET "W _ 5 W 5 < " _ 4HE NAT

_
URAL INCLUSION "W ' INDUCES THE OPEN INCLUSION J "W _ '" ,ET

l '" 4 BE THE RATIONAL MORPHISM DElNED BY l PR4 m ;J =_ WHERE
PR4 " _ 4 " _ 5 _ IS THE NATURAL PROJECTION
&OR I ) LET I '" 'A BE THE RATIONAL FUNCTION GIVEN BY I
~I_ m ;J =_ WHERE ~I_ " _ ` 5 'A IS THE REGULAR FUNCTION DElNED IN
SECTION
,EMMA &OR EACH I ) WE HAVE
A I p
B I XIA `X IX A
C lXI A ` X jI? AI X lX
&OR EACH I ) DElNE A RATIONAL MORPHISM EI 'M a '" '" BY
THE FORMULA q r
C_
ECI X XI ` X
I X
2EMARK ,EMMA IMPLIES THE EQUALITY I ECI X C_ I X FOR I )
4HEOREM 4HE MORPHISMS EI I ) DElNE A GEOMETRIC CRYSTAL
ON '" l
7E DENOTE THIS CRYSTAL BY 8W 4HIS IS ONE OF OUR MAIN EXAMPLES OF
GEOMETRIC CRYSTALS
2EMARK 4HE GEOMETRIC CRYSTAL 8W IS AN ANALOGUE OF THE FREE COMBI
NATORIAL 7 CRYSTAL "W AND THE RATIONAL FUNCTIONS I '" 'M ARE
ANALOGUES OF THE FUNCTIONS _EI ON "W SEE !PPENDIX
0OSITIVE GEOMETRIC CRYSTALS AND THEIR TROPICALIZATION ,ET 4
BE AN ALGEBRAIC TORUS OVER 1 7E DENOTE BY 8 4 AND 8 4 RESPECTIVELY
THE LATTICES OF CHARACTERS AND CO CHARACTERS OF 4 AND BY H` Ì THE CANONICAL
PAIRING 8 4 a 8 4 :
,ET ,4 BE THE SET OF FORMAL LOOPS } 'M 4 )N PARTICULAR ,'M
,C WHERE ,C IS THE SET OF ALL ,AURENT SERIES IN THE VARIABLE C
! "%2%.34%). !.$ $ +!:($!. '!&!
$ENOTE BY , 4 THE SET FORMAL DISKS THAT IS THE SET OF ALL } , 4

SUCH THAT FOR ANY u 8 4 WE HAVE u m } , C WHERE , C y , C
IS THE SET OF ALL INVERTIBLE 4ALYOR SERIES IN THE VARIABLE C
#LEARLY , 4 HAS A NATURAL STRUCTURE OF AN IRREDUCIBLE PRO ALGEBRAIC
VARIETY
,EMMA 4HE MULTIPLICATION MAP 8 4 a , 4 ,4 IS A BIJEC
TION
4HIS DElNES A SURJECTIVE MAP DEG4 ,4 8 4
2EMARKS )F 4 'M THEN DEG'M ,C : IS THE VALUATION MAP
WHICH ASSOCIATES TO A NON ZERO ,AURENT SERIES }C ITS LOWEST DEGREE "Y
DElNITION , C DEG_
'M
&OR ANY CO CHARACTER t 8 4 THE PRE IMAGE DEG_
4 t y ,4 IS

NATURALLY ISOMORPHIC TO , 4 (ENCE IT MAKES SENSE TO TALK ABOUT GENERIC
POINTS OF DEG_
4 t
&OR ANY F 24 4 AND ANY t 8b 4 u 8b 4 LET 5F t u
BE THE SET OF ALL } DEG_ _
4 t SUCH THAT F m } DEG4 u
,EMMA ,ET F 24 4 &OR ANY t 8b 4 THERE IS A UNIQUE
u 8b 4 SUCH THAT THE SET 5F t u IS DENSE IN DEG_
4 u
4HIS ALLOWS US TO DElNE A MAP DEGF 8b 4 8 4 BY DEGF t
u WHERE u 8b 4 IS DETERMINED IN ,EMMA
7E CALL THIS MAP THE DEGREE OF F )T IS EASY TO CHECK THAT DEGF IS A
PIECEWISE LINEAR MAP .OTE THAT DEGF IS LINEAR IF F IS A GROUP HOMOMOR
PHISM
$EFINITION A ! RATIONAL FUNCTION F ON A TORUS 4 IS CALLED POSITIVE IF IT
CAN BE WRITTEN AS A RATIO F F F WHERE F AND F ARE LINEAR COMBINATIONS
OF CHARACTERS WITH POSITIVE INTEGER COEnCIENTS
B &OR ANY TWO ALGEBRAIC TORI 4 4 WE CALL A RATIONAL MORPHISM
F 24 4 POSITIVE IF FOR ANY CHARACTER u 4 'M THE COMPOSI
TION u m F IS A POSITIVE RATIONAL FUNCTION ON 'M
$ENOTE BY 2 4 4 THE SET OF POSITIVE RATIONAL MORPHISMS FROM 4
TO 4
2EMARK !NY HOMOMORPHISM OF ALGEBRAIC TORI IS POSITIVE
,EMMA &OR ANY TWO POSITIVE MORPHISMS F 2 4 4 G
2 4 4 THE COMPOSITION G m F 2 4 4 IS WELL DElNED
4HEREFORE WE CAN CONSIDER A CATEGORY 4 WHOSE OBJECTS ARE ALGEBRAIC
TORI AND THE ARROWS ARE POSITIVE RATIONAL MORPHISMS
6ISIONS IN -ATH '%/-%42)# !.$ 5.)0/4%.4 #2934!,3
&OR ANY F 24 4 AND ANY t 8b 4 u 8b 4 LET 5F t BE THE

SET OF ALL } DEG_ 4 t SUCH THAT F m } ,4

7E SAY THAT A FORMAL LOOP },4 IS POSITIVE RATIONAL IF }2 'M 4

,EMMA ,ET F 4 4 BE A POSITIVE MORPHISM 4HEN FOR ANY

t 8 4 AND ANY FORMAL POSITIVE RATIONAL LOOP } DEG_ 4 t WE HAVE
DEG4 mF m } DEGF t
#OROLLARY &OR ANY ALGEBRAIC TORI 4 4 4 AND ANY F 2 4 4
G 2 4 4 WE HAVE
DEGG m F DEGG m DEGF
4HIS IMPLIES THAT IS A FUNCTOR 4ROP 4 3ET SUCH THAT 4ROP4
8 4 AND 4ROPF 4 4 DEGF 8 4 8 4
&OLLOWING ;"&:= WE CALL THE FUNCTOR 4ROP TROPICALIZATION
2EMARK )F F 4 4 IS NOT POSITIVE THEN THE ASSERTION OF ,EMMA
IS NOT TRUE EVEN IF F IS A BIRATIONAL ISOMORPHISM -OREOVER ONE CAN lND A
BIRATIONAL ISOMORPHISM F F 4 F 4 DOES NOT HOLD FOR F G WHERE
G F 4 4 #ONSIDER FOR EXAMPLE THE CASE WHEN 4 4 'M
_
F C C _ GC C
$EFINITION ,ET 8 BE A GEOMETRIC PRE CRYSTAL ON 8 l ! BIRATIONAL
ISOMORPHISM q 4 F 8 FOR SOME ALGEBRAIC TORUS 4 IS CALLED A POSITIVE
STRUCTURE ON 8 IF THE FOLLOWING CONDITIONS ARE SATISlED
4HE RATIONAL MORPHISM l m q 4 4 IS POSITIVE
&OR EACH I ) THE RATIONAL MORPHISM EIq 'M a 4 4 GIVEN BY
EIq C T q_ ECI qT
IS POSITIVE
2EMARK )N ALL OUR EXAMPLES OF POSITIVE STRUCTURES q THE COMPOSITION
l m q 4 4 IS A GROUP HOMOMORPHISM
7E SAY THAT TWO POSITIVE STRUCTURES q q ARE EQUIVALENT IF THE RATIONAL
MORPHISMS q_ m q AND q _ m q ARE POSITIVE
2ECALL FROM THE !PPENDIX THAT A COMBINATORIAL PRE CRYSTAL CONSISTS OF A
SET " A MAP l E " c? AND A COMPATIBLE COLLECTION OF PARTIAL BIJECTIONS
EÎ " "
&OR ANY POSITIVE STRUCTURE q ON A GEOMETRIC PRE CRYSTAL 8 BUILT ON 8 l
DElNE FOR I ) THE : ACTION E^nI : a 8 4 8 4 BY THE FORMULA
E^nI 4ROPEIq

WHERE EIq 'M a 4 4 IS THE POSITIVE MORPHISM DElNED BY
!LSO DENOTE lETROP 4ROPl m q 8 4 c?
! "%2%.34%). !.$ $ +!:($!. '!&!
/BVIOUSLY THE MAP EÎ 8 4 8 4 IS A BIJECTION FOR I ) 4HE

BIJECTIONS EÎ EÎ I ) DElNE A FREE COMBINATORIAL PRE CRYSTAL 4ROPq 8
ON 8 4 l
Eq
4HEOREM ,ET 8 BE A GEOMETRIC CRYSTAL AND LET q BE A POSITIVE
STRUCTURE ON 8 4HEN 4ROPq 8 IS A FREE COMBINATORIAL 7 CRYSTAL
7E CALL THIS FREE COMBINATORIAL PRE CRYSTAL 4ROPq 8 THE TROPICALIZATION
OF 8 WITH RESPECT TO THE POSITIVE STRUCTURE q
&OR A REDUCED SEQUENCE I I IL DElNE A MORPHISM qI 'M L
'" BY
qI C CL XI C SI ` ` ` XIL C SIL ` "
4HE FOLLOWING THEOREM WAS PROVED IN ;":= IN A SLIGHTLY DIkERENT FORM
4HEOREM &OR EACH I 2W THE MORPHISM qI IS A POSITIVE
STRUCTURE ON THE GEOMETRIC CRYSTAL 8W AND THE TROPICALIZATION OF 8W WITH
RESPECT TO qI IS EQUAL TO THE FREE COMBINATORIAL 7 CRYSTAL "I !LL THESE
POSITIVE STRUCTURES qI ARE EQUIVALENT TO EACH OTHER
2EMARK )N ;,U= ,USZTIG INTRODUCED MORPHISMS qI 'M LW '"
qI C CL YI C ` ` ` YIL CL ` "
AND PROVED THAT THESE MORPHISMS ARE RELATED TO EACH OTHER IN THE SAME WAY
AS THE CORRESPONDING PARAMETRIZATIONS OF THE CANONICAL BASIS FOR 5Q , =
7E DISCUSS SIMILAR MORPHISMS x I IN SECTION
)T WAS SHOWN IN ;":= THAT THE MORPHISMS qI 'M LW '"
I 2W ARE ALSO POSITIVE STRUCTURES ON 8W AND MOREOVER THESE POSITIVE
STRUCTURES ARE EQUIVALENT TO qI I 2W
4RIVIALIZATION OF GEOMETRIC CRYSTALS ,ET 8 BE A GEOMETRIC '
CRYSTAL BUILT ON 8 l 7ITHOUT LOSS OF GENERALITY WE MAY ASSUME THAT l
IS A REGULAR SURJECTIVE MORPHISM 8 4 $ENOTE BY 8 l _ E THE lBER
OVER THE UNIT E 4
"Y DElNITION OF THE ACTION 7 ON 8 SEE 0ROPOSITION WX X
FOR X 8 W 7
$EFINITION ! TRIVIALIZATION OF A GEOMETRIC CRYSTAL 8 IS A 7 INVARIANT
RATIONAL PROJECTION { 8 8
)T IS EASY TO SEE THAT THE FOLLOWING FORMULA DElNES A TRIVIALIZATION FOR
ANY GEOMETRIC 3, CRYSTAL 8

lX
{ X EI X
WHERE ) FG AND I IS THE ONLY FUNDAMENTAL WEIGHT OF 4
6ISIONS IN -ATH '%/-%42)# !.$ 5.)0/4%.4 #2934!,3
)N THE CASE WHEN ' 3, WE CAN CONSTRUCT TWO DIkERENT TRIVIALIZATIONS
{ { 8 8 FOR ANY GEOMETRIC 3, CRYSTAL 8 4HESE TRIVIALIZATIONS ARE
GIVEN BY
lX
lX lX lX
{ X E E E X
lX
{ X E
lX lX lX
E E X
WHERE c ARE THE FUNDAMENTAL WEIGHTS

lX lX
7ARNING 4HE MORPHISM 8 8 X E E X IS NOT 7
INVARIANT
!CTUALLY FOR ANY R ANY GEOMETRIC ' 3,R CRYSTAL 8 WE CAN
CONSTRUCT TWO TRIVIALIZATIONS { { 8 8 OF 8 IN THE FOLLOWING WAY ,ET
R c BE THE FUNDAMENTAL WEIGHTS ORDERED IN THE STANDARD WAY
AND I X I lX FOR I R
&OR EVERY GEOMETRIC 3,R CRYSTAL 8 DElNE A MORPHISM { 8 8
BY THE FORMULA
o po R X R X p o X X po X p
X R_ X X X X
{ X ER X ` ` ÈRR X E R_ ` ` ÈR_ ` ` ` E E E X
4HEOREM 4HIS MORPHISM { 8 8 IS A TRIVIALIZATION OF 8
4HE FORMULA FOR THE SECOND TRIVIALIZATION { 8 8 IS OBTAINED FROM
{ BY APPLYING THE AUTOMORPHISM OF THE $YNKIN DIAGRAM WHICH EXCHANGES
I WITH R _ I
7E DO NOT KNOW WHETHER TRIVIALIZATIONS OF GEOMETRIC ' CRYSTALS EXIST
FOR OTHER REDUCTIVE GROUPS 7E EXPECT THAT FOR THE UNIPOTENT CRYSTALS ONE
CAN lND A TRIVIALIZATION
5NIPOTENT #RYSTALS
$ElNITION OF UNIPOTENT CRYSTALS AND THEIR PRODUCT !S WE
HAVE SEEN GEOMETRIC CRYSTALS ARE GEOMETRIC ANALOGUES OF FREE COMBINATORIAL
7 CRYSTALS /UR NEXT TASK IS TO INTRODUCE A GEOMETRIC ANALOGUE OF CRYSTAL
BASES
$EFINITION ! 5 VARIETY 8 IS A PAIR 8 j WHERE 8 /B6 AND
j 5 a8 8 IS A RATIONAL UNITAL 5 ACTION ON 8 SUCH THAT Ea8 y DOMj
WHERE E IS THE UNIT OF 5
&OR 5 VARIETIES 8 9 WE SAY THAT A RATIONAL MORPHISM F 8 9 IS A
5 MORPHISM IF IT COMMUTES WITH THE 5 ACTIONS
! "%2%.34%). !.$ $ +!:($!. '!&!
)T IS WELL KNOWN THAT THE MULTIPLICATION IN ' INDUCES A BIRATIONAL ISO

MORPHISM " _ a 5 F
' $ENOTE BY G THE INVERSE BIRATIONAL ISOMORPHISM
"_ a 5
G 'F
_ "_
,ET x ' AND x ' 5 BE RATIONAL MORPHISMS DElNED BY
x _ PR m G x PR m G
"Y DElNITION DOMx _
DOMx "_
` 5
"_
0ASSING TO THE QUOTIENT WE OBTAIN A BIRATIONAL ISOMORPHISM '5 F
4HEREFORE THE NATURAL LEFT ACTION OF 5 ON '5 DElNES A LEFT 5 ACTION
j" _ 5 a " _ " _ 4HIS ACTION SATISlES FOR U 5 B " _
_ `_
j" _ U B x _ U ` B U ` B ` xU ` B
)N PARTICULAR THE PAIR "_ " _ j" _ IS A 5 VARIETY
,EMMA &OR U 5 _ T 4 ONE HAS
_ ` _ `
x XI A ` U ` T XI A_ ~_
I U
_
` jI T_
,EMMA A &OR B U ` T U 5 _
T 4 A 'A AND I ) WE HAVE
q r
_ ` A
j" _ XI A B XI A ` B ` XI _
A~_ I U jI T
B %VERY 5 ORBIT IN "_ IS THE INTERSECTION OF " _ WITH A 5 a 5 ORBIT
IN '
$EFINITION ! UNIPOTENT CRYSTAL IS A PAIR 8 F WHERE 8 IS A 5 VARIETY
AND F 8 "_ IS A 5 MORPHISM
7E DENOTE BY 5 #RYST THE CATEGORY WHOSE OBJECTS ARE UNIPOTENT '
CRYSTALS AND ARROWS ARE DOMINANT RATIONAL MORPHISMS
&OR ANY 8 F8 9 F9 /B5 #RYST DElNE A RATIONAL MORPHISM
j 5 a 8 a 9 8 a 9 BY THE FORMULA
_ ` _ `
j U X Y UX xU ` F8 X Y
WHERE UX jU X 7E WILL OFTEN WRITE UX Y INSTEAD OF jU X Y
4HEOREM A 4HE MORPHISM j 5 a 8 a 9 8 a 9 DElNED ABOVE
IS A RATIONAL 5 ACTION ON 8 a 9
B ,ET M " _ a " _ " _ BE THE MULTIPLICATION MORPHISM ,ET F
F8a9 8 a9 " _ BE THE RATIONAL MORPHISM DElNED BY F MmF8 aF9
4HEN F8a9 IS A 5 MORPHISM
7E DENOTE THE 5 VARIETY 8 a 9 j8a9 BY 8 aF 9 !CCORDING TO THE
THEOREM THE PAIR 8 aF 9 F8a9 IS A UNIPOTENT CRYSTAL 7E CALL IT THE
PRODUCT OF 8 F8 AND 9 F9 AND DENOTE IT BY 8 F8 a 9 F9
6ISIONS IN -ATH '%/-%42)# !.$ 5.)0/4%.4 #2934!,3
2EMARK 4HE PRODUCT OF UNIPOTENT CRYSTALS IS ANALOGOUS TO THE TENSOR

PRODUCT OF +ASHIWARAS CRYSTALS DElNED IN ;+= 4HIS ANALOGY IS MADE
PRECISE IN SECTIONS AND BELOW
0ROPOSITION 4HE PRODUCT OF UNIPOTENT CRYSTALS IS ASSOCIATIVE
2EMARK 4HE ABOVE RESULTS DElNE A STRICTLY MONOIDAL STRUCTURE ON THE
CATEGORY 5 _ #RYST
4HE PRODUCT OF UNIPOTENT CRYSTALS IS NOT COMMUTATIVE IN 1 GENERAL &OR
ANY FAMILY 8K FK /B5 #RYST K L WE DENOTE BY LK 8K FK
THE PRODUCT 8 FK a ` ` ` a 8L FL
%XAMPLE $ENOTE BY 5 THE PAIR 5 j5 WHERE j5 5 a 5 5 IS THE
LEFT ACTION ,ET PR5 _
E 5 " BE THE PROJECTION ON THE UNIT E #LEARLY
5 PR5E /B5 #RYST 3IMILARLY DENOTE BY ' THE PAIR ' j' WHERE
j' 5 a ' 5 IS THE LEFT ACTION #LEARLY ' x _ /B5 #RYST
0ROPOSITION 4HE BIRATIONAL ISOMORPHISM G INDUCES THE ISOMORPH
ISM IN 5 #RYST
' x_ F
"_ ID" _ a 5 PR5
E
-ORE GENERALLY TAKING THE RIGHT QUOTIENT BY ANY SUBGROUP 5 y 5 WE
OBTAIN THE BIRATIONAL ISOMORPHISM G5 '5 F
" _ a 55 WHICH INDUCES
THE ISOMORPHISM IN 5 #RYST

'5 x5_ F
"_ ID" _ a 55 PR55
E
WHERE x5_ PR m G5
4HE FOLLOWING RESULT SHOWS THAT THE UNIPOTENT CRYSTAL ' x _ IS UNIVER
SAL
,EMMA &OR EVERY 5 EQUIVARIANT RATIONAL MORPHISM F5 5 " _
THERE IS AN ELEMENT W
^ .ORM' 4 SUCH THAT
F5 U x _ U ` W
^
FOR EVERY U 5
&ROM UNIPOTENT ' CRYSTALS TO UNIPOTENT , CRYSTALS 4HROUGH
OUT THIS SECTION WE lX A SUBSET * OF ) ,ET , ,* BE THE ,EVI SUBGROUP
OF ' GENERATED BY 4 AND BY EACH 5J 5J_ J *
,ET 0 , ` 5 AND 0 _ 5 _ ` , "Y DElNITION 0 AND 0 _ ARE PARABOLIC
SUBGROUPS OF ' SUCH THAT 0 z " 0 _ z " _ AND 0 < 0 _ ,
,ET 50 AND 50_ BE RESPECTIVELY THE UNIPOTENT RADICALS OF 0 AND 0 _ ,ET
5, , < 5 5,_ , < 5 _ 4HEN 5, AND 5,_ ARE THE OPPOSITE UNIPOTENT
RADICALS OF ,
! "%2%.34%). !.$ $ +!:($!. '!&!
$ENOTE ",_ " _ <, AND ", "<, 4HE OPEN INCLUSION ",_ ,5,
INDUCES A RATIONAL ACTION OF 5, ON ",_ WHICH WE DENOTE BY j, 5, a ",_
",_ ,ET P_ P_ _ _
, " ", BE THE CANONICAL PROJECTION "Y DElNITION
_
P, COMMUTES WITH THE RATIONAL ACTION OF 5, )N PARTICULAR "_ P_
/B5, #RYST
&OR ANY 5 VARIETY 8 WE DENOTE BY 8J, THE 5, VARIETY OBTAINED BY THE
RESTRICTION OF THE 5 ACTION j 5 a 8 8 TO 5,
,EMMA 4HE MAPPING 8 F8 8 F8 J, 8J, P_, m F8 DElNES
A FUNCTOR OF MONOIDAL CATEGORIES J, 5 _ #RYST _ 5, _ #RYST
7E CALL THE UNIPOTENT , CRYSTAL 8 F8 J, RESTRICTION OF 8 F8 TO ,
&ROM UNIPOTENT ' CRYSTALS TO GEOMETRIC , CRYSTALS &OR EACH
UNIPOTENT ' CRYSTAL 8 F DElNE A MORPHISM l l8 8 4 BY l
PR4 mF &OR EACH I ) DElNE THE FUNCTION I 8
I BY
I ~I_ m F8
,ET SUPP8 F BE THE SET OF ALL THOSE I ) FOR WHICH 8
Ip 7E CALL
THIS SET THE SUPPORT OF THE UNIPOTENT CRYSTAL 8 F8 &OR I SUPP8 F
DElNE THE MORPHISM EI 'M a 8 8 BY
q r
C C_
EI X XI X
I X
)T IS EASY TO SEE THAT EACH EI IS A RATIONAL ACTION OF 'M ON 8
4HEOREM &OR ANY 8 F /B5 #RYST THE ACTIONS EI 'M a
8 8 I SUPP8 F DElNE A GEOMETRIC ,* CRYSTAL ON 8 l8 WHERE
* SUPP8 F
7E DENOTE THIS GEOMETRIC ,* CRYSTAL BY 8IND AND CALL IT GEOMETRIC CRYSTAL
INDUCED BY 8 F
2EMARK &OR THE UNIPOTENT CRYSTAL '" ;J =_ SEE SECTION 4HEO
REM SPECIALIZES TO 4HEOREM
%XAMPLES &OR THE UNIPOTENT CRYSTAL "_ ID" _ THE ACTIONS EI
'M a " _ " _ I ) ARE GIVEN BY
q r q r
C C_ C_ _
EI B XI ` B ` XI
I B I B jI lB
&OR THE UNIPOTENT CRYSTAL ' x _ THE ACTIONS EI 'M a ' '
I ) ARE GIVEN BY
q r
C C_
EI B XI `G
~I G
6ISIONS IN -ATH '%/-%42)# !.$ 5.)0/4%.4 #2934!,3
,EMMA &OR 8 F8 9 F9 /B5 #RYST PUT : F: 8 F8 a
9 F9 WHERE : 8 a 9 AND LET :IND BE THE GEOMETRIC CRYSTAL ON : l:
INDUCED : F: 7E HAVE
A l: M m l8 m l9 WHERE M 4 a 4 4 IS THE MULTIPLICATION
MORPHISM
B &OR EACH I ) X Y :
9I Y
: 8
I X Y I X
jI l8 X
WHICH IMPLIES THAT SUPP: F: SUPP8 F8 ; SUPP9 F9
C &OR ANY I SUPP: F: THE ACTION EI 'M a : : IS GIVEN BY THE
FORMULA ECI X Y ECI X ECI Y WHERE
CI X jI lX I Y I X jI lX I Y
C C
I X jI lX I Y I X jI lX C_ I Y
2EMARK 4HE FORMULA FOR THE ACTION OF EI ON 8 a 9 IS ANALOGOUS
TO THE FORMULA IN ;+ 3ECTION = FOR THE TENSOR PRODUCT OF +ASHIWARAS
CRYSTALS
0OSITIVE UNIPOTENT CRYSTALS AND DUALITY ,ET 8 F
/B5 #RYST AND LET 4 BE AN ALGEBRAIC TORUS OF THE SAME DIMENSION AS 8
! BIRATIONAL ISOMORPHISM q 4 F 8 IS CALLED A POSITIVE STRUCTURE ON 8 F
IF THE FOLLOWING TWO CONDITIONS ARE SATISlED
4HE ISOMORPHISM q IS A POSITIVE STRUCTURE ON THE INDUCED GEOMETRIC
,* CRYSTAL 8IND WHERE * SUPP8 F
&OR ANY I * THE FUNCTION 8
I m q ON 4 IS POSITIVE
4HEOREM ,ET 8 F8 9 F9 /B5 #RYST AND q8 4 8
q9 4 9 BE RESPECTIVELY THE POSITIVE STRUCTURES 4HEN THE BIRA
TIONAL ISOMORPHISM q8a9 q8 a q9 IS A POSITIVE STRUCTURE ON THE PRODUCT
8 F8 a 9 F9
&OR A GEOMETRIC PRE CRYSTAL 8 ON 8 l DENOTE BY l b 8 4 THE
MORPHISM l b X lX _ AND CONSIDER THE DUAL GEOMETRIC PRE CRYSTAL
_
8 b ON 8 l b BY DElNING EbI C X ECI X
'IVEN A GEOMETRIC PRE CRYSTAL 8 ON 8 l FOR ANY MORPHISM q 4 8
DElNE A MORPHISM qb 4 8 AS THE COMPOSITION OF q WITH THE INVERSE
_ 4 4
4HE FOLLOWING FACT IS OBVIOUS

,EMMA &OR ANY GEOMETRIC CRYSTAL 8 THE DUAL GEOMETRIC PRE CRYSTAL
8 b IS A GEOMETRIC CRYSTAL &OR ANY POSITIVE STRUCTURE q ON 8 THE MORPHISM
qb IS A POSITIVE STRUCTURE ON 8 b
! "%2%.34%). !.$ $ +!:($!. '!&!
2EMARK $UALITY IN GEOMETRIC CRYSTALS IS AN ANALOGUE OF DUALITY IN

+ASHIWARAS CRYSTALS SEE ;+= AND THE !PPENDIX BELOW -ORE PRECISELY
THE TROPICALIZATION OF 8 b WITH RESPECT TO THE POSITIVE STRUCTURE qb IS THE FREE
COMBINATORIAL 7 CRYSTAL DUAL TO THE TROPICALIZATION 8 WITH RESPECT TO q
,ET 8 F BE A UNIPOTENT CRYSTAL $ElNE A MORPHISM F b 8 " BY
F b X F X _ AND A RATIONAL MORPHISM jb 5 a 8 8 BY
_ `
jb U X x U ` F b X X
0ROPOSITION 4HE MORPHISM jb IS A RATIONAL ACTION OF 5 ON 8 AND
F b IS A 5 MORPHISM WITH RESPECT TO THIS ACTION
$ENOTE THE PAIR 8 jb BY 8b 4HEN THE PAIR 8b F b IS A UNIPOTENT
CRYSTAL 7E CALL IT THE DUAL UNIPOTENT CRYSTAL OF 8 F AND DENOTE IT BY
8 F b
2EMARK )T FOLLOWS FROM THE DElNITION THAT THE MAPPING 8 F 8 F b
IS AN INVOLUTIVE FUNCTOR b 5 #RYST 5 #RYST
4HEOREM &OR ANY UNIPOTENT CRYSTALS 8 F8 9 F9 THE PERMU
TATION OF THE FACTORS 8 a 9 9 a 8 INDUCES THE ISOMORPHISM OF
UNIPOTENT CRYSTALS
_ `b
8 F8 a 9 F9 F 9 F9 b a 8 F8 b
2EMARK 4HIS THEOREM IMPLIES THAT THE FUNCTOR b 5 #RYST 5 #RYST
REVERSES THE MONOIDAL STRUCTURE ON 5 #RYST
4HEOREM ,ET 8 F BE AN UNIPOTENT CRYSTAL AND 8 BE THE INDUCED
GEOMETRIC CRYSTAL 4HEN THE GEOMETRIC CRYSTAL INDUCED BY 8 F b IS EQUAL
TO 8 b
$IAGONALIZATION OF THE PRODUCTS OF UNIPOTENT CRYSTALS ,ET
V BE THE REGULAR MORPHISM "W " 5 DElNED BY VUTW U UU FOR ANY
U U 5 T 4
"Y DElNITION V IS TWO SIDED 5 EQUIVARIANT
2EMARK &OR ' 3, THE MAP V ' 5 IS GIVEN BY
q r q r
A B AD
C
V
C D
7E CALL A UNIPOTENT CRYSTAL 8 F8 NON DEGENERATE IF RANF8 INTERSECTS
"W " NON TRIVIALLY
&OR ANY NON DEGENERATE UNIPOTENT CRYSTAL 8 F8 /B5 #RYST AND
ANY 9 F9 /B5 #RYST DElNE THE RATIONAL MORPHISM & &8a9
8 a 9 8 a 9 BY & X Y X VX Y WHERE VX VF8 X #LEARLY &
6ISIONS IN -ATH '%/-%42)# !.$ 5.)0/4%.4 #2934!,3
IS A BIRATIONAL ISOMORPHISM OF 8 a 9 F 8 a 9 AND ITS INVERSE IS GIVEN BY

& _ X Y X VX _ Y
$ENOTE BY m m8a9 5 a 8 a 9 8 a 9 THE DIAGONAL ACTION OF 5
_ ` _ `
U X Y UX UY
0ROPOSITION ,ET 8 F BE A NON DEGENERATE UNIPOTENT CRYSTAL
AND 9 F9 BE ANY UNIPOTENT CRYSTAL 4HEN
& m j m m ID5 a &
)N OTHER WORDS THE ACTION j j8a9 5 a 8 a 9 8 a 9 IS DIAGONALIZED
BY &
5NIPOTENT ACTION OF THE 7EYL GROUP &OR EACH * y ) LET 7* BE

THE SUBGROUP OF 7 GENERATED BY SJ J * "Y DElNITION 7* IS THE 7EYL
GROUP OF THE STANDARD ,EVI SUBGROUP , ,* 3O WE SOMETIMES DENOTE 7*
BY 7,
,EMMA ,ET 8 F BE A UNIPOTENT ' CRYSTAL * SUPP8 F AND
8IND BE THE INDUCED GEOMETRIC ,* CRYSTAL 4HEN THE FORMULA DElNES
RATIONAL ACTION OF 7* ON 8
7E CALL THIS ACTION OF 7, ON 8 THE UNIPOTENT ACTION OF 7* ON 8 F8
OR SIMPLY UNIPOTENT ACTION OF 7* ON 8
2ECALL FROM SECTION THAT "_ ID" _ IS A UNIPOTENT CRYSTAL
,EMMA 4HE UNIPOTENT ACTION OF 7 ON " _ SATISlES FOR I )
SI B XI A ` B ` XI A _
_jI lB
WHERE A I B jI lB
2EMARK 4HIS LEMMA IMPLIES THAT DOMSI FB " ~I_ B G

&OR EACH W 7 LET SUPPW BE THE MINIMAL SUBSET * y ) SUCH THAT
W 7* &OR EXAMPLE SUPPSI FIG
0ROPOSITION &OR EACH W 7 THERE IS A RATIONAL MORPHISM UW
" _ 5 SUCH THAT
A &OR ANY W 7 SUCH THAT SUPPW z SUPPW EACH RATIONAL 5 ORBIT
OF THE FORM T ` 5W 5 < " _ T 4 INTERSECTS DOMUW NON TRIVIALLY
B 4HE UNIPOTENT ACTION OF 7 ON " _ IS GIVEN BY
W B WB UW B ` B ` UW B _
&OR ANY UNIPOTENT CRYSTAL 8 F8 AND ANY W 7 WITH SUPPW y
SUPP8 F8 WE DElNE A RATIONAL MORPHISM U8 8
W 8 5 BY UW UW m F8
! "%2%.34%). !.$ $ +!:($!. '!&!
4HEOREM ,ET 8 F8 /B5 #RYST WITH SUPP8 F8 * 4HEN

THE CRYSTAL ACTION OF 7* ON 8 IS GIVEN BY
WX U8
W X X
FOR W 7*
0ROPOSITION ,ET 8K F8K /B5 #RYST K L AND LET
* ;K SUPP8K F8K 4HEN THE UNIPOTENT ACTION OF 7* ON 8 8 a ` ` ` a
8L IS GIVEN BY _ `
WX XL UX UXL
FOR W 7 WHERE U U8 W X XL IS AS IN
%XAMPLES OF 5NIPOTENT #RYSTALS

3TANDARD UNIPOTENT CRYSTALS &OR W 7 LET /W 5 W55
#LEARLY /W IS A LEFT 5 ORBIT IN '5 $ElNE 5 W 5 < W5W _
,EMMA 4HE MAPPING 5 /W DElNED BY U UW5 FOR U 5 IS LEFT
5 EQUIVARIANT AND SURJECTIVE )T INDUCES THE ISOMORPHISM OF HOMOGENEOUS
5 VARIETIES
p^W 55 W F
/W
2EMARK 5NDER THE NATURAL MAP ' '" EACH ORBIT /W IS IDEN
TIlED WITH THE CORRESPONDING 3CHUBERT CELL )N PARTICULAR DIM /W
DIM 55 W LW
&OR W 7 DElNE 5 W 5 < " _ W _ " _ AND "W
_
5 W5 < " _
%XAMPLE &OR ' 3, W W THE SETS 5 W AND "W
_ CONSIST RESPECTIVELY
OF THE MATRICES OF THE FORM

q r q r
C C

C_
C 'M
$ENOTE BY JW 5 W 55 W THE MORPHISM INDUCED BY THE NATURAL
INCLUSION 5 W 5 AND BY JW "W
_ /W THE MORPHISM INDUCED BY THE
_
NATURAL INCLUSION "W 5 W5
"Y DElNITION THE RESTRICTIONS OF x AND x _ TO " _ ` 5 ARE THE REGULAR
PROJECTIONS " _ `5 5 AND " _ `5 " _ RESPECTIVELY 3INCE " _ W _ " _ ` W
y " _ ` 5 WE CAN DElNE A REGULAR MORPHISM p W 5 W " _ BY
p W U x _ UW
FOR U 5 W
6ISIONS IN -ATH '%/-%42)# !.$ 5.)0/4%.4 #2934!,3
0ROPOSITION A 4HE MORPHISMS JW AND JW ARE OPEN INCLUSIONS

_ W
"W /W AND 5 55 W RESPECTIVELY
B 4HE MORPHISM p W IS A BIREGULAR ISOMORPHISM 5 W F _ 4HE INVERSE
"W
W _
ISOMORPHISM pW p "W F _ W
5 IS GIVEN BY
pW B xW ` B_ _
C 4HE FOLLOWING DIAGRAM IS COMMUTATIVE
p^W
55 W ___ /W
X X
J W J
W
pW
5W ___ "W _
4HE BIRATIONAL ISOMORPHISMS ;JW = AND ;JW = DElNE FOR ANY W 7 RATIONAL
UNITAL 5 ACTIONS jW 5 a 5 W 5 W AND jW 5 a "W _ " _ RESPECTIVELY
7E DENOTE BY 5W 5 W jW AND "_ W "W jW THE CORRESPONDING

5 VARIETIES
.OTE THAT 5W p W AND "_W IDW ARE UNIPOTENT ' CRYSTALS 7E CALL
THESE UNIPOTENT CRYSTALS STANDARD
,EMMA &OR EACH W 7 THE ISOMORPHISM p W INDUCES THE ISOMOR
PHISM OF UNIPOTENT CRYSTALS 5W p W F
"_
W IDW
.OTE THAT THE 5 ACTION ON "_ W IS GIVEN BY )N ORDER TO DESCRIBE
THE 5 ACTION ON 5W WE NEED TO INTRODUCE MORE NOTATION
"Y 0ROPOSITION THE MULTIPLICATION MORPHISM 5 a 5 5 INDUCES
THE BIRATIONAL ISOMORPHISM 5 W a 5 W F 5 $ENOTE BY x W 5 5 W
THE COMPOSITION OF THE INVERSE OF THIS ISOMORPHISM WITH PR
"Y DElNITION THE ACTION 5 a 5 W 5 W U X UX IS GIVEN BY
_ `_
UX U ` X ` x W U ` X
,EMMA &OR ANY W W 7 SUCH THAT LWW LW LW AND

X 5 W Y 5 W U 5 WE HAVE
_ `
x WW U ` X ` Y x W x W U ` X ` Y

OR EQUIVALENTLY UX ` Y UX x W U ` X Y
.EXT WE COMPUTE THE ACTION EXPLICITLY IN TERMS OF THE ISOMOR
PHISM pW
0ROPOSITION &OR EACH W 7 U 5 AND 5 W WE HAVE
_ `
xW U ` X x !D WxU ` B ` x _ W ` B_
WHERE B p W X
! "%2%.34%). !.$ $ +!:($!. '!&!
-ULTIPLICATION OF STANDARD UNIPOTENT CRYSTALS $ElNE A NEW

ASSOCIATIVE MULTIPLICATION ON THE SET 7 BY
WW WW
IF LWW LW LW FOR W W 7 AND SI SI SI FOR ALL I
5NDER THIS NEW OPERATION 7 IS IDENTIlED WITH THE STANDARD MULTIPLICA
TIVE MONOID OF THE DEGENERATE (ECKE ALGEBRA OF 7 7E DENOTE THIS MONOID
BY 7
,EMMA &OR ANY W W 7 THE MULTIPLICATION MORPHISM ' a ' '
INDUCES A DOMINANT RATIONAL MORPHISM

x W W 5 W a 5 W 5 WW

4HIS IS A BIRATIONAL ISOMORPHISM IF AND ONLY IF WW WW
)T IS WELL KNOWN SEE EG ;&: 4HEOREMS = THAT FOR ANY W W 7
SUCH THAT WW WW THE MULTIPLICATION MORPHISM ' a ' ' INDUCES
THE OPEN INCLUSION
_ _ _
"W a "W "WW
4HE FOLLOWING RESULT DEALS WITH A GENERALIZATION OF TO ANY PAIR
W W
0ROPOSITION &OR ANY W W 7
_ ` " _ WITH " _ _
A 4HE INTERSECTION "W W WW IS A DENSE SUBSET OF "WW
B 4HERE EXISTS AN ALGEBRAIC SUB TORUS 4EWW y 4 SUCH THAT THE RESTRICTION
OF THE MULTIPLICATION MORPHISM 'a' ' TO "W _ a" _ IS A DOMINANT
W
RATIONAL MORPHISM
"W_ _
a "W _ E
"WW ` 4WW
4HIS IS A BIRATIONAL ISOMORPHISM IF AND ONLY IF LWW
LW LW _ DIM 4EWW

C 4EWW FEG IF AND ONLY IF W W WW
D 4ESI SI jI? 'M FOR EACH I )
E &OR ANY W W W 7 ONE HAS 4EWW W `4EW W 4EWW W `W _ 4EWW
#OROLLARY &OR ANY W W 7 THE MORPHISM INDUCES THE MOR
PHISM IN 5 #RYST "_ _ _ E
W IDW a "W IDW "WW ` 4WW ID" _ `4E

WW WW
.EXT WE COMPUTE A LOWER BOUND FOR EACH 4EWW

&OR EACH W 7 LET US CONSIDER A HOMOMORPHISM OF TORI 4 4 DElNED
BY T WT ` T_ $ENOTE BY 4W THE IMAGE OF THIS HOMOMORPHISM #LEARLY
4W IS A SUB TORUS OF 4 SUCH THAT
<
8 4W cW gI :jI?
6ISIONS IN -ATH '%/-%42)# !.$ 5.)0/4%.4 #2934!,3
WHERE cW Ft c Wt tG 4HIS IMPLIES THAT 8 4W HAS A : BASIS

OF CERTAIN NOT NECESSARILY SIMPLE COROOTS
0ROPOSITION &OR ANY W W 7 WE HAVE 4EWW z 4WW _ `W`W
&ROM NOW ON WE WILL FREELY USE THE NOTATION OF SECTION &OR ANY
STANDARD ,EVI SUBGROUPS , y , y ' DElNE THE ELEMENTS W, , W,, 7
BY

W, , W, W, W,, W, W,
#LEARLY W, , ` W, , E AND LW, , LW,, LW _ LW, ,
.OTE THAT 5, W,, 5, AND 5, W, , !D W, , 5,

)T TURNS OUT THAT FOR W W, , THE FORMULA SIMPLIlES WHEN U 5,
0ROPOSITION ,ET , y , BE STANDARD ,EVI SUBGROUPS OF ' 4HEN
FOR ANY U 5, X 5 W, , WE HAVE
_ `
x W, , U ` X !D W, , xU ` p W, , X
&OR EACH W 7 LET )W BE THE SET OF ALL I ) SUCH THAT WjI jI
FOR SOME I ) .OTE THAT AN ELEMENT W 7 BELONGS TO 7)W IF AND ONLY
_
IF !D W5 W 5 WW W
,EMMA ,ET W 7 W W 7)W 4HEN
_
!D W m x W J5 W x WW W J5 WW W_ m !D W
4HEOREM ,ET , y , BE STANDARD ,EVI SUBGROUPS OF ' AND LET
W 7 BE ANY ELEMENT SUCH THAT W, , _ ` W CENTRALIZES 5, 4HEN FOR ANY
W 7)W WE HAVE

A 4HE MORPHISM FWW 5 W, , a5 W 5 GIVEN BY X Y X`!D WY
_
IS A DOMINANT RATIONAL MORPHISM 5 W, , a 5 W 5 WW W W, ,
4HIS IS A BIRATIONAL ISOMORPHISM IF AND ONLY IF LWW W_ W, ,
LW LW, ,
B )F FWW IS A BIRATIONAL ISOMORPHISM THEN IT INDUCES AN ISOMORPHISM
IN THE CATEGORY 5, _ #RYST
_ W _ W
5W, , p W, , J, a 5W p W J, 5WW W , , p WW W , , J,
,ET ' ',MN M N 7E USE THE STANDARD LABELING OF THE
$YNKIN DIAGRAM OF TYPE !MN_ ) F MN_G ,ET * )NFMG
#LEARLY ,* ,MN ',M a ',N y ',MN $ENOTE ,M ',M a FEG
,N FEG a ',N SO THAT ,MN ,M ` ,N w
,M a ,N ,ET WMN W,MN '
AND WNM W',MN WMN _
)T IS EASY TO SEE THAT

WMN SM SM_ ` ` ` S SM SM ` ` ` S ` ` ` SMN_ SMN_ ` ` ` SN
! "%2%.34%). !.$ $ +!:($!. '!&!
)N PARTICULAR DIM 5 WNM MN $ENOTE #M S ` ` ` SM AND #N

SMN_ ` ` ` SM .OTE THAT 5WNM p WNM 5#M p #M AND 5#N p #N ARE
UNIPOTENT ',MN CRYSTALS
#OROLLARY A 4HE RESTRICTION 5WNM p WNM J,M IS ISOMORPHIC IN
5,M #RYST TO
9N
5#M p #M J,M
L
B 4HE RESTRICTION 5WNM p WNM J,N IS ISOMORPHIC IN 5,N #RYST TO
9M
k
5#N p #N k,
N
K
7 INVARIANT FUNCTIONS ON STANDARD UNIPOTENT ' CRYSTALS &OR
EACH W 7 LET 5 B W BE THE SET OF ALL ~ 5
B SUCH THAT ~W _ UW ~U
FOR ALL U 5 W
B E 5
#LEARLY 5 B
B W 5
&OR ANY ~ 5 B DElNE A REGULAR FUNCTION ~W "W" 'A BY
~W G ~W _ G
? 4 AD 4 AD
,ET y 8 BE THE CO CHARACTER OF 4 :' SUCH THAT
Hy? jI I FOR I ) $ElNE THE ANTI AUTOMORPHISM r OF ' BY rG
!D y? _ G _
"Y DElNITION rT T_ FOR T 4 AND r m XI XI r m YI YI FOR ALL
I )
)T IS EASY TO SEE THAT rW W_ FOR ANY W 7 AND ~ m r ~ FOR ANY
~5 B
&OR ANY ~ ~ 5 B AND ANY W 7 DElNE A REGULAR FUNCTION F W ON
~~
"W" BY
W W W_
F~~ ~ ~ m r
0 ,ET , B ,* BE, ANYB STANDARD ,EVI SUBGROUP OF ' &OR ANY ~

I) AI ~I 5 LET ~ 5 BE DElNED BY
8
~, AI ~I
I)N*
"Y DElNITION ~, J5, p

4HEOREM &OR ANY ~ 5 B W,' AND ANY T :, THE RESTRICTION
W', _
OF THE FUNCTION F~~, TO T ` "W,' IS INVARIANT UNDER THE UNIPOTENT ACTION OF
7 ON T ` "W_
,'
6ISIONS IN -ATH '%/-%42)# !.$ 5.)0/4%.4 #2934!,3
.EXT WE DESCRIBE FOR EACH W 7 A BASIS OF THE SUBSPACE 5 B W y 5 B &OR

_
EACH W 7 DElNE A BIJECTIVE MAP oW )W )W BY WjI joW I
4HIS oW IS A PARTIAL BIJECTION ) ) SEE THE !PPENDIX WITH DOMo )W
AND RANo )W _
&OR EACH I ) WE DElNE THE SET OW I AS FOLLOWS &OR I DOMo ;RANo
WE PUT
g h
OW I oW _ I oW _ I I oW I oW I
WHERE FOR EACH K : THE POWER oWK IS CONSIDERED TO BE A PARTIAL BIJECTION
) ) &OR I DOMo ; RANo WE DElNE OW I FIG
&OR ANY * y ) LET ~* 5 B BE DElNED BY ~* 0
J* ~J
,EMMA &OR EACH I ) THE FUNCTION ~OW I BELONGS TO 5 B W AND THE
SET OF ALL ~O I I ) IS A BASIS IN THE VECTOR SPACE 5 B W
W
2ECALL THAT THE UNIPOTENT ACTION OF 7* WAS DElNITED IN SECTION

4HEOREM A &OR EACH ~ 5 B W', THE RESTRICTION OF ~ TO 5 W',
IS INVARIANT UNDER THE UNIPOTENT ACTION OF 7* ON THE UNIPOTENT , CRYSTAL
5W', p W', J,
B &OR EACH ~ 5 B W,' THE RESTRICTION OF THE FUNCTION ~W,' TO " _
W,'
IS INVARIANT UNDER THE UNIPOTENT ACTION OF 7* ON THE UNIPOTENT , CRYSTAL
"_ W,' IDW,' J,
0OSITIVE STRUCTURES ON STANDARD UNIPOTENT CRYSTALS &OR ANY

I ) DElNE THE MORPHISM xI 'M " _ BY THE FORMULA xI C
YI C jI? C #LEARLY xI IS A BIREGULAR ISOMORPHISM 'M F "S_I 4HE ISO
MORPHISM xI 'M F _
"WI DElNES A STRUCTURE OF 5 VARIETY ON 'M AND A
UNIPOTENT CRYSTAL ON 'M WHICH WE DENOTE BY 'M xI
&OR EACH SEQUENCE I I IL DElNE THE REGULAR MORPHISM xI
'M L " _ BY THE FORMULA
xI C CL xI C ` ` ` xIL CL
&OR ANY SEQUENCE I I IL ) L DElNE W I SI ` ` ` SIL
0ROPOSITION &OR ANY SEQUENCE I I IL ) L WE HAVE
A 4HERE EXISTS A SUB TORUS 4EI y 4 SUCH THAT ;xI = IS A DOMINANT RATIONAL
MORPHISM 'M L "W _
I
` 4EI 4HE MORPHISM ;xI = IS A BIRATIONAL
ISOMORPHISM IF AND ONLY IF
L LW I DIM 4EI
B 4HE RATIONAL MORPHISM ;xI = INDUCES THE MORPHISM IN 5 #RYST
_ `
'M xI a ` ` ` a 'M xIL "_ E
W I ` 4I ID" _ `4EI
W I

! "%2%.34%). !.$ $ +!:($!. '!&!
C )F THE EQUALITY HOLDS THEN ;xI = IS A POSITIVE STRUCTURE ON THE

UNIPOTENT CRYSTAL "_ E
W I ` 4I ID" _ `4EI

W I
D 4HERE IS AN INDUCTIVE FORMULA FOR THE COMPUTATION OF 4EI 4EI jI? 'M
AND FOR ANY SEQUENCE I I IL ) L AND ANY I ) ONE HAS

SI 4EI IF LW I SI LW I
4EII
E E
SI 4I ` 4I IF LW I SI LW I _
2EMARKS &OR EACH W 7 I 2W THE TROPICALIZATION OF THE
GEOMETRIC CRYSTAL INDUCED BY "_
W IDW WITH RESPECT TO qI IS EQUAL TO THE
CORRESPONDING FREE COMBINATORIAL CRYSTAL "I WHICH WAS CONSTRUCTED IN ;+=
AS A PRODUCT OF DIMENSIONAL CRYSTALS "I "IL "Y THE CONSTRUCTION "I
IS A FREE 7JIJ CRYSTAL WHERE JIJ FI IL G
&OR ANY DOMINANT t? c? THE IMAGE OF +ASHIWARAS 0 EMBEDDING
CAN BE DESCRIBED AS FOLLOWS SEE EG ;":= ,ET ~ I) ~I 5 B
&OR ANY I 2W LET qI ID4 a ;xI = 4 a 'M F L _
4 ` "W BE THE POSITIVE
STRUCTURE ON 4 `"W _ 4HEN THE IMAGE OF THE EMBEDDING "6 ? " :L
t I
IS g h
W
B "I 4ROPqI F~~ t? B t
&OR EACH I I IL ) L DElNE THE REGULAR MORPHISM x I 'M L
5 BY
xI C CL XI C ` ` ` XIL CL
0ROPOSITION &OR ANY SEQUENCE I I IM ) L WE HAVE
A 4HE MORPHISM xI INDUCES A DOMINANT RATIONAL MORPHISM 'M L
_
5 W I )F I 2W FOR SOME W 7 THEN x I IS AN OPEN INCLUSION
'M LW 5 W
B )F THE SEQUENCE I BELONGS TO 2W FOR SOME W 7 THEN THE BIRATIONAL
_
ISOMORPHISM ;x I = 'M LW F
5 W IS A POSITIVE STRUCTURE ON THE
_ _
UNIPOTENT CRYSTAL 5W p W
$UALITY AND SYMMETRIES FOR STANDARD UNIPOTENT CRYSTALS )T
IS EASY TO SEE THAT THE INVERSE `_ " _ " _ INDUCES THE ISOMORPHISM
_F _
"W "W _ FOR EACH W 7
,EMMA 4HE INVERSE IN " _ INDUCES THE ISOMORPHISM OF UNIPOTENT

CRYSTALS
"_ b
W IDW F "_
W_
IDW_
FOR EACH W 7 )N PARTICULAR THE UNIPOTENT CRYSTAL "_
W IDW IS SELF DUAL
IF W E
6ISIONS IN -ATH '%/-%42)# !.$ 5.)0/4%.4 #2934!,3
0ROOFS
0ROOFS OF RESULTS IN SECTION
0ROOF OF 4HEOREM 3INCE 4HEOREM IS A PARTICULAR CASE OF 4HEO
REM WE REFER TO THE PROOF OF 4HEOREM IN SECTION c
0ROOF OF 0ROPOSITION #LEARLY SI 8 8 IS A BIRATIONAL INVOLUTION

4HUS IT SUnCES TO PROVE THAT FOR ANY W W 7 SATISFYING LW W
LW LW WE HAVE
W W X W W X
!S FOLLOWS FROM THE DElNITION OF EW 4 a 8 8 WE HAVE
ETW W EW
W
T
m ETW
FOR ALL W W 7 SATISFYING LW W LW LW .OTE THAT FOR ANY
W 7 WE HAVE
lETW X T ` WT_ lX
lX _
4HIS IMPLIES THAT lWX lEW X WlX 4HEREFORE FOR ANY
W W 7 SUCH THAT LW W LW LW WE HAVE
_ _ lX _ `
W W X ElW
W
X
EW X
W lX _ _ _
EW
EW X lX ElX
W W X W W X
0ROPOSITION IS PROVED c
0ROOF OF 4HEOREM ,ET 7 a 4 4 BE THE ACTION OF 7 OBTAINED

FROM THE ACTION BY TWISTING WITH q THAT IS W T q_ WqT
)T IS EASY TO SEE THAT 7 ACTS ON 4 BY A POSITIVE BIRATIONAL ISOMORPHISM
4HEREFORE APPLYING THE FUNCTOR 4ROP TO THE ACTION 7 a 4 4 WE OBTAIN
AN ACTION OF 7 ON 8 4
4HEOREM IS PROVED c
0ROOF OF 4HEOREM "Y DElNITION FOR I J ) F RG WE HAVE

I ECJ X CmIJ I C
4HUS { EC X { X AND THEREFORE { S X { X
,ET US COMPUTE { ECJ X FOR J 4HE COMPUTATION IS BASED ON THE
FOLLOWING OBVIOUS STATEMENT
,EMMA &OR EACH J R WE HAVE
{ ECJ X EXJR m {J ECJ X
! "%2%.34%). !.$ $ +!:($!. '!&!
WHERE
o JZ Z J Z
Z
p o Z Z po Z p
Z Z Z
{J Z E J ` ` ` EJ J ` ` ` E E E Z
o R X R X p o J X
X J X
p
J X
EXJR E R X
` ` ` ER R X
` ` ` E J
` ` ` EJ
AND WE USE THE CONVENTION R X p AND EXRR p
J_ X J X
3UBSTITUTING C jJ lX J X INTO WE OBTAIN
{ SJ X EJR X m {J SJ X
4HIS IMPLIES THAT ALL WE HAVE TO DO TO PROVE 4HEOREM IS TO CHECK THAT
{J SJ X {J X
FOR J R
,ET US lRST COMPUTE {J ECJ X J t 3UBSTITUTING Z ECJ X IN
AND USING THE EQUALITIES K Z CmJK J X AND {J_ m ECJ ECJ m {J_ WE
OBTAIN
o C
J X
X
J X
C X
po CJ X
X
CJ X
p
J_ X _ C `
{J ECJ X E J ` ` ` EJ J E J_ ` ` ` EJ_ EJ m {J_ X
J_ X J X
3UBSTITUTING C J X INTO WE OBTAIN
o J X
J_ X
J X
J_ X
po J X
J X
J X
J X
p J_ X J X
J X
{J SJ X E ` ` ` EJ E ` ` ` EJ_ EJ {J_ X
)N ORDER TO lNISH THE PROOF OF 4HEOREM WE WILL USE THE FOLLOWING
RESULT
,EMMA &OR ANY J R THE FOLLOWING RELATION HOLDS
C C
EC ` ` ` ECJ EC ` ` ` ECJ_ EJ EC ` ` ` ECJ EC ` ` ` ECJ_
C C C
0ROOF )NDUCTION IN J &OR J THE IDENTITY E IS TRUE EC E
.OW LET J 5SING THE COMMUTATION OF EJ WITH EACH OF E EJ_
WE REWRITE THE LEFT HAND SIDE OF AS
C C
EC ` ` ` ECJ_ EC ` ` ` ECJ_ ECJ ECJ_ EJ
!PPLYING TO THE ABOVE EXPRESSION THE BASIC RELATION WRITTEN IN THE FORM
C C C C
ECJ ECJ_ EJ EJ_ ECJ ECJ_ WE SEE THAT THE LEFT HAND SIDE OF EQUALS
C C
EC ` ` ` ECJ EC ` ` ` ECJ_ EJ_ ECJ ECJ_
&INALLY APPLYING THE INDUCTIVE HYPOTHESIS WITH J _ WE OBTAIN

EC ` ` ` ECJ_ EC ` ` ` ECJ_ ECJ ECJ_ EC ` ` ` ECJ EC ` ` ` ECJ_
4HE LEMMA IS PROVED c
6ISIONS IN -ATH '%/-%42)# !.$ 5.)0/4%.4 #2934!,3
J X
7E SEE THAT IS A SPECIAL CASE OF ,EMMA WITH C J_ X
X
C J
J X

4HEOREM IS PROVED c
0ROOF OF THE RESULTS IN SECTION

0ROOF OF 4HEOREM 0ROVE A 7E START WITH THE FOLLOWING RESULT
,EMMA &OR ANY UNIPOTENT CRYSTAL 8 F THE FOLLOWING IDENTITY HOLDS
_ ` _ ` _ `
x U ` F UX ` x U ` F X x U U ` F X

FOR ANY U U 5 X 8
0ROOF $ENOTE B F X .OTE THAT F UX UB U ` B ` xU ` B _ 4HEN
THE IDENTITY CAN BE REWRITTEN AS
_ `
x U ` U ` B ` xU ` B _ ` xU ` B xU U ` B
4HIS IDENTITY IS TRUE BECAUSE THE MORPHISM x ' 5 IS RIGHT 5 EQUIVARIANT

/NE CAN EASILY SEE THAT IMPLIES THAT U U X Y U UX Y FOR
ALL U U 5 X Y 8 a 9
4HUS THE FORMULA DElNES AN 5 ACTION
0ART A IS PROVED 0ROVE B 7E NEED THE FOLLOWING FACT
,EMMA &OR U 5 B B " _ WE HAVE
_ `
xU ` B ` B x xU ` B ` B
0ROOF )T SUnCES TO TAKE U XI A )N THIS CASE ,EMMA IMPLIES THAT
FOR U U 5 _ T T 4 WE HAVE
_ ` _ `
x XI A ` U ` TU ` T x XI A ` U ` TU T_ ` TT
_ `
XI A_ ~_ _ _
I UTU T ` jI TT _
_ `
XI A_ ~_ _
I U jI T ~I U
_
` jI T_ ` jI T _
_ _ _ ` _ `
XI A ~_ _
I U jI T x XI A ` U T
WHERE A A_ ~_ I U
_ ` j T_ 4HE LEMMA IS PROVED
I c
0ART B IS PROVED AND 4HEOREM IS ALSO PROVED c
0ROOF OF 0ROPOSITION ,ET 8 F8 8 F8 8 F8 BE UNIPOTENT
CRYSTALS AND LET
_ `
8 F8 8 F8 a 8 F8 a 8 F8
_ `
8 F8 8 F8 a 8 F8 a 8 F8
&OR ANY XK 8K K WE HAVE
_ ` _ `
F8 X X X F8 X ` F8 X ` F8 X F8 X X X
! "%2%.34%). !.$ $ +!:($!. '!&!
THAT IS F8 F8

)T SUnCES TO PROVE THAT 8 8 THAT IS THESE TWO 5 ACTIONS ON
8 a 8 a 8 ARE EQUAL
&OR XK 8K K DENOTE BK F8K XK
,ET US WRITE THE 5 ACTIONS RESPECTIVELY ON 8 8
_ ` _ `
U X X X UX X xU ` B ` B X
_ `
UX xU ` B X x U ` B ` B X
_ ` _ `
U X X X UX xU ` B X X
_ `
UX xU ` B X xxU ` B ` B X
)T FOLLOWS FROM ,EMMA THAT UX X X UX X X
4HIS PROVES 0ROPOSITION c
0ROOF OF 0ROPOSITION )T FOLLOWS FROM THE DElNITION OF G AND THE
DElNITION OF x THAT FOR G ' U 5 ONE HAS
GU ` G F' U ` G ` xU ` G
.OTE THAT F' U ` G U ` F' G AND xUG xUF' G ` xG 4HUS
0ROPOSITION FOLLOWS FROM 4HEOREM c
0ROOF OF 4HEOREM )N VIEW OF ,EMMA IT SUnCES TO PROVE THE

RELATIONS [
&IRST WE ARE GOING TO PROVE THE RELATIONS AND &OR F _G
LET
g h
) a ) I J ) a ) I J AND HjI? jJ I HjJ? jI I
.OTE THAT IF ' IS SIMPLY LACED THEN ) a ) ; ) a ) _ ; a) ) a )
,EMMA A &OR ANY I J ) a ) WE HAVE
XI A XJ A XJ A XI A
B &OR I J ) a ) _ WE HAVE
q r q r
A A A A
XI A XJ A XI A XJ XI A A XJ
A A A A
0ROOF #LEAR c
0ROPOSITION ,ET 8 F BE A UNIPOTENT ' CRYSTAL AND LET 8IND BE THE
INDUCED GEOMETRIC CRYSTAL &OR ANY I J ) a ) ; ) a ) _ SUCH THAT
I p J p ONE HAS
A )F I J ) a ) THEN
I ECJ X I X I ECJ X I X
6ISIONS IN -ATH '%/-%42)# !.$ 5.)0/4%.4 #2934!,3
B )F I J ) a ) _ THEN THERE EXIST RATIONAL FUNCTIONS IJ IJ ON 8

SUCH THAT
I X J X IJ X JI X
AND
IJ ECJ X IJ X IJ ECI X C_ IJ X

JI ECI X JI X JI ECJ X C_ JI X
C 7E HAVE
J X I ECJ X CIJ X JI X I X J ECI X CJI X IJ X

C C_
IJ X JI X
I ECJ ECI X I X
C JI X IJ X

C C_
JI X IJ X
J ECI ECJ X J X
C IJ X JI X
0ROOF )T SUnCES TO PROVE THE LEMMA IN THE ASSUMPTION THAT ) FI JG
THAT IS WHEN ' IS SEMISIMPLE OF TYPES ! a ! AND ! RESPECTIVELY 0ART
A FOLLOWS
,ET US PROVE B )T SUnCES TO ANALYZE THE CASE WHEN ' ', $UE TO
,EMMA IT SUnCES TO PROVE THE STATEMENT ONLY FOR 8 ' F8 x _ )N
THIS CASE ) FI JG AND WE SET I J IN THE STANDARD WAY
a G
)T IS EASY TO SEE THAT K G ~K_ G aKK G FOR K WHERE
a G G a G G
q r q r
G G G G
a G DET a G DET
G G G G
&URTHERMORE IS EASY TO SEE THAT THE ACTIONS E E 'M a " _ " _ ARE
GIVEN BY THE FORMULA
o p
ECK G XK C _ aK G
a G ` G K
FOR K
$ElNE THE FUNCTIONS ON ' BY
a a
G
G
a G a G
WHERE q r
G G
a G G a G DET
G G
a G a G a G a G a G a G
! "%2%.34%). !.$ $ +!:($!. '!&!
4HIS IDENTITY IMPLIES &URTHERMORE IT IS EASY TO SEE THAT

aK ECL G CmKL aK G aK ECL G aK G
FOR K L F G AND aK ECK G aK G FOR K 4HIS IMPLIES
0ART B IS PROVED
0ART C EASILY FOLLOWS FROM B c
)N ORDER TO PROVE THE RELATIONS FOR ANY I J ) a ) WE COMPUTE

THE LEFT HAND SIDE AND THE RIGHT HAND SIDE OF "Y DElNITION
ECI ECJ X XI A XJ A X
C _ C _ C _
WHERE A C
I EJ X
I X AND A J X !NALOGOUSLY
ECJ ECI X XJ A XI A X
4HUS USING THE RELATION WE SEE THAT THE LEFT AND THE RIGHT HAND SIDES
OF ARE EQUAL 4HIS PROVES ALL THE RELATIONS
4O PROVE THE RELATIONS FOR ANY I J ) a ) _ WE COMPUTE THE
LEFT HAND SIDE AND THE RIGHT HAND SIDE OF "Y DElNITION
ECI ECJ C ECI X XI A XJ A XI A X
WHERE
C _ C C _ I X C C _
A A C
I X J EI X C JI X IJ X
AND
C _ C _ C JI X IJ X
A C C C
I EJ EI X I X C IJ X JI X
3IMILARLY
ECJ ECI C ECJ X XJ A XI A XJ A X
WHERE
C _ C C _ J X C C _
A A C
J X I EJ X C IJ X JI X
AND
C _ C _ C IJ X JI X
A
J ECI C ECJ X J X C JI X IJ X
A A A A A A A A A A A
4HUS IT FOLLOWS FROM THAT
ECI ECJ C ECI X XI A XJ A XI A X
XJ A XI A XJ A X ECI ECJ C ECI X
4HIS PROVES ALL THE RELATIONS
6ISIONS IN -ATH '%/-%42)# !.$ 5.)0/4%.4 #2934!,3
4HUS WE HAVE PROVED 4HEOREM FOR ALL SIMPLY LACED REDUCTIVE
GROUPS ' )N PARTICULAR THE RELATIONS HOLD FOR SUCH GROUPS
)T REMAINS TO PROVE THE RELATIONS AND )NSTEAD OF DOING THE
COMPUTATIONS DIRECTLY WE WILL PROVE THE RELATIONS FOR ANY GROUP ' BY
DEDUCING THESE RELATIONS FROM THE RELATIONS FOR A CERTAIN SIMPLY LACED
GROUP ' CONTAINING '
7ITHOUT LOSS OF GENERALITY WE MAY ASSUME THAT ' IS ADJOINT SEMISIMPLE
)T IS WELL KNOWN THAT THERE IS AN ADJOINT SEMISIMPLE SIMPLY LACED GROUP
' AN OUTER AUTOMORPHISM z ' ' AND AN INJECTIVE GROUP HOMOMOR

PHISM F ' ' SUCH THAT F ' ' z )N WHAT FOLLOWS WE IDENTIFY '
WITH ITS IMAGE F '
-OREOVER ONE CAN ALWAYS CHOOSE AN OUTER AUTOMORPHISM z IN SUCH A
WAY THAT z PRESERVES A CHOSEN "OREL SUBGROUP " y ' AND A MAXIMAL TORUS
4 y " AND SATISlES " " z 4 4 z ,ET " _ BE THE OPPOSITE
"OREL SUBGROUP CONTAINING 4 4HEN " _ IS ALSO z INVARIANT AND " _
" _ z !LSO THE AUTOMORPHISM z INDUCES AN INJECTION z c? c ?
,ET ) BE THE $YNKIN DIAGRAM OF ' )T IS EASY
0 TO SEE THAT FOR ANY I )
THERE EXISTS A SUBSET { I ) SUCH THAT jI? I { I j ?
I AND THAT FOR ANY
I J { I THE SUBGROUPS 5I 5J COMMUTE 4HIS IMPLIES THAT THE 7EYL
GROUP 7 IS THE SUBGROUP OF THE 7EYL GROUP 7 OF ' AND ITS GENERATORS SI
I ) ARE GIVEN BY 9
SI SI
I { I
/NE CAN ALWAYS CHOOSE THE HOMOMORPHISMS XI 'M 5I YI 'M 5I
I ) IN SUCH A WAY THAT
9 9
XI A XI A YI A YI A
I { I I { I
1
FOR ANY I ) 4HIS IMPLIES THAT SI I { I S I
$UE TO ,EMMA WE MAY ASSUME WITHOUT LOSS OF GENERALITY THAT
8 ' ,ET 8IND BE THE GEOMETRIC ' CRYSTAL INDUCED BY ' ID' AND
LET 8IND BE THE GEOMETRIC ' CRYSTAL INDUCED BY ' ID' 7E DENOTE BY
EI 'M a ' ' I ) THE ACTIONS OF 'M AND BY EI 'M a' '
I ) THE CORRESPONDING ACTIONS OF 'M !S FOLLOWS FROM THE TRANS

FORMATIONS EI C COMMUTE WITH EJ C FOR ANY I ) I J { I
&OR EACH I ) LET US lX A LINEAR ORDERING { I OF { I )T IS EASY TO
SEE THAT FOR ANY REDUCED SEQUENCE I I IL ) L THE SEQUENCE { I
{ I { IL IS ALSO REDUCED
,ET EI 4 a ' ' AND E{ I 4 a ' ' BE MORPHISMS AS IN
! "%2%.34%). !.$ $ +!:($!. '!&!
4HE FOLLOWING LEMMA IS OBVIOUS

,EMMA &OR EACH REDUCED SEQUENCE I I IL ) L WE HAVE
I DOMEI DOME{ I < 4 a '
II E{ I J4 a' EI
4HE LEMMA IMPLIES THAT FOR ANY REDUCED SEQUENCES I Î ) L SATISFYING
WI WÎ WE HAVE EI E{ I J4 a' E{ Î J4 a' EÎ
4HIS PROVES ALL THE RELATIONS FOR GEOMETRIC ' CRYSTALS
4HEOREM IS PROVED c
0ROOF OF 4HEOREM )T FOLLOWS FROM ,EMMA THAT

A 4HE MORPHISM l8a9 m q8a9 4 a 4 4 IS POSITIVE
B %ACH FUNCTION }8a9
I IS POSITIVE
C %ACH ACTION EI 'M a 8 a 9 8 a 9 IS POSITIVE
4HEOREM IS PROVED c
0ROOF OF 0ROPOSITION 7ITHOUT LOSS OF GENERALITY WE MAY ASSUME

THAT 8 IS A SUBSET OF " _ OR EVEN 8 " _ SO THAT F8 IS THE IDENTITY MAP
8 " _ 4HEN IT IS EASY TO SEE THAT jb U B UB_ _ 4HIS IMPLIES
THAT jb IS AN ACTION OF 5 &URTHERMORE F b B B_ 4HUS
F b jb U B UB_ UF b B
0ROOF OF 0ROPOSITION )T IS EASY TO SEE THAT THE FORMULA IS

EQUIVALENT TO THE EQUATION
_ `
UVF X V F UX xF X
FOR X " _ U 5 V IS TWO SIDED 5 EQUIVARIANT THIS IDENTITY IS EQUIVALENT
TO _ ` _ `
V U ` F X V F UX ` xF UX
4HIS IDENTITY IS TRUE SINCE
F UX xF UX UF X xF UX U ` F X
0ROOF OF 4HEOREM 7E NEED THE FOLLOWING OBVIOUS RESULT

,EMMA &OR ANY UNIPOTENT CRYSTAL 8 F LET bI ~I_ F8 X _ BE
THE FUNCTION ON 8 DElNED BY FOR THE DUAL UNIPOTENT CRYSTAL 8 F b
I ) 4HEN bI X _I X jI lX
6ISIONS IN -ATH '%/-%42)# !.$ 5.)0/4%.4 #2934!,3
5SING WE OBTAIN jb XI A X XI A_ bI X _ jI l b X _ X

3UBSTITUTING A C_ b
b X INTO THIS EXPRESSION FOR j WE OBTAIN THE ACTION
I
OF THE CORRESPONDING MULTIPLICATIVE GROUP OF THE INDUCED DUAL CRYSTAL /N
THE OTHER HAND SIMPLIFYING THE RIGHT HAND SIDE WE OBTAIN
q q r r q r q _ r
b C_ C_ C _ _
j XI X XI X XI X ECI X
bI X _CbI X jI l b X I X
WHICH PROVES 4HEOREM c
0ROOF OF 0ROPOSITION $ENOTE BY 4R THE SET OF REGULAR ELEMENTS OF 4

AND BY "RSS_ THE PRE IMAGE PR_ WHICH IS THE SET OF ALL REGULAR SEMISIMPLE
4
ELEMENTS IN " _ 4HIS IS AN OPEN DENSE SUBSET .OTE THAT "RSS _
INTERSECTS
_
NON TRIVIALLY EACH "W T T 4R
_ 5 W INDUCTIVELY $ElNE
7E DElNE THE RATIONAL MORPHISM UW "RSS
UE PRE q r
_ jI lB
USI B XI
I B jI lB
FOR I ) AND FOR ANY W W 7 SUCH THAT LW W LW LW DElNE

_ `
UW W B UW UW B ` B ` UW B _ ` UW B
,EMMA &OR EACH W 7 THE MORPHISM UW DOES NOT DEPEND ON THE
CHOICE OF THE EXPRESSION AND SATISlES
0ROOF )T IS EASY TO SEE THAT UW SATISlES 3INCE THE CENTRALIZER IN 5 OF
ANY ELEMENT B "RSS_ IS TRIVIAL WE SEE THAT U DOES NOT DEPEND ON A CHOICE
W
OF THE EXPRESSION
&URTHERMORE WE CAN REPEAT THE SAME DElNITIONS FOR ANY SUB VARIETY
_ _
OF THE FORM "W T < "RSS )N THIS CASE FOR ANY I SUPPW THE RATIONAL
_ _ 5 SI IS WELL DElNED AND IS GIVEN BY THE
MORPHISM USI "W T < "RSS
ANALOGOUS FORMULA

0ROOF OF 0ROPOSITION &OLLOWS FROM ;": 4HEOREM = WITH U E
V W c
0ROOF OF 0ROPOSITION ,ET US EXPRESS THE ACTION OF 5 ON 5 W AS THE

_ WITH THE ISOMORPHISM p
CONJUGATION OF THE ACTION OF 5 ON "W W
UX pW UB
! "%2%.34%). !.$ $ +!:($!. '!&!
WHERE B p W X 5SING AND THE RIGHT 5 EQUIVARIANCY OF x WE

OBTAIN
_ `_ _ `_
UX x W ` xU ` B ` B_ U_ U ` xW ` xU ` B ` B_
5SING IN THE LEFT HAND SIDE OF THE ABOVE IDENTITY WE OBTAIN
_ ` _ `
x W U ` X x W ` xU ` B ` B_ ` X x W ` xU ` B ` B_ ` X
&INALLY
_ `_
B_ X B_ pW B B_ ` xW ` B_ W _ ` x _ W ` B_
4HE PROPOSITION IS PROVED c
0ROOF OF 0ROPOSITION 0ROVE A AND B )F W W WW

THEN THE
STATEMENT FOLLOWS FROM 4HE GENERAL STATEMENT REDUCES TO THE CASE
WHEN W W SI FOR I ) ,ET US USE THE FOLLOWING IDENTITY IN 3,
q r q r q r q r
C C D D
` `
C_ C _ D_ D _
FOR ANY C C D D 'M SATISFYING D C DD DD CC _ 4HIS
IMPLIES THAT "S_I ` "S_I p "S_I ` jI? 'M AND THAT "S_I ` "S_I CONTAINS THE SET
"S_I N FxI _ G WHERE xI 'M F "S_I IS THE BIREGULAR ISOMORPHISM DElNED
IN SECTION
0ARTS A AND B ARE PROVED 0ARTS C E EASILY FOLLOW
0ROOF OF 0ROPOSITION &OR W 7 DElNE AN ACTION OF 4 ON ' BY

T G !DW TG WT ` G ` T_
,EMMA &OR ANY W W 7 ONE HAS !DW 4 5 W 5 5 W 5 `4W _ W
0ROOF )T SUnCES TO SHOW THAT !DW 4 W W ` 4W _ W "Y DElNITION
WE HAVE
!DW TW W T ` W ` T_ W ` W _ W T ` T_
FOR T 4 4HE LEMMA IS PROVED c
&URTHERMORE WE HAVE _ `" _

y 5 W5 `5W 5 4HUS
"W _ _
!DW`W 4 "W `"W
W
_ _ _ _
"W ` "W /N THE OTHER HAND !DW`W 4 "WW "WW ` 4WW _ `W`W
_ _
4HEREFORE BY 0ROPOSITION A "WW ` 4WW _ `W`W y "WW E
` 4WW
0ROOF OF 0ROPOSITION $ENOTE TEMPORARILY W W, , B p W X

x_ XW AND U !D W, , xU ` B
&URTHERMORE WE HAVE BY
_ `
x W U ` X x U ` x _ W ` B_
6ISIONS IN -ATH '%/-%42)# !.$ 5.)0/4%.4 #2934!,3
)T IS EASY TO SEE THAT U 5 BECAUSE xU ` B 5, .OTE THAT

W `B_ W5 W _ 5 4HUS x _ W `B_ W5, W _ <5 _ W50 <, W _ y
W50 W _ 4HE LATTER SET IS THE UNIPOTENT RADICAL OF W0 W _ 4HIS IMPLIES
THAT !D U x _ W ` B_ 5 _ AND
_ `
x W U ` X x !D U x _ W ` B_ ` U U
0ROOF OF 4HEOREM 0ART A FOLLOWS FROM ,EMMA 0ROVE B )T

SUnCES TO SHOW THAT FWW COMMUTES WITH THE ACTION OF 5, "Y DElNITION
THE ACTION OF 5, ON THE PRODUCT 5W, , p W, , a 5W p W IS GIVEN BY
SEE _ `
UX Y UX xU ` p W, , X Y

&OR X 5 W, , Y 5 W AND U 5, WE HAVE
_ `
U FWW X Y UX ` Y UX ` U Y
WHERE Y !D WY AND U x W, , U ` X "Y 0ROPOSITION U
!D W, , U WHERE U xU ` p W X
U ` Y W, , ` U ` W, , _ ` W ` Y ` W _ !D WU ` Y
BECAUSE W, , _ ` W COMMUTES WITH U 5,
&URTHERMORE
_
U Y U ` Y ` x WW W U Y _ !D WU Y
_ _
BECAUSE x WW W U Y x WW W !D WU Y !D Wx W U Y BY
,EMMA APPLIED WITH W W, W
&INALLY
_ ` _ `
U FWW X Y UX `!D WU Y FWW UX U X Y FWW UX Y
4HEOREM IS PROVED c
0ROOF OF 4HEOREM &OR EACH W 7 DENOTE 6 W 5 < W5 _ W _

)T IS EASY TO SEE THAT 5 W ` 6 W 5 5 W < 6 W FEG AND
5 W5 6 W W5 y W5 _ ` 5
B WE HAVE SEE
,EMMA &OR ANY W 7 ~ 5
~W U GU ~W _ T_ U TW ~U ~W G
FOR ALL G T ` 5W5 T 4 U .ORM5 6 W AND U 5
0ROOF .OTE THAT ~W U GU ~W U G ~U FOR ALL U 3O WE PROVE THE
LEMMA UNDER THE ASSUMPTION THAT U
! "%2%.34%). !.$ $ +!:($!. '!&!
,ET G 5 W5 %XPRESS G AS G U TWU WHERE U 6 W U 5

4HEN ~W G ~W _ U TWU ~U BECAUSE W _ 6 W W y 5 _ &URTHER
MORE FOR EACH U .ORM5 6 W WE HAVE U G U U TWU U U TWU
WHERE U !D U U 6 W &INALLY
~W U G ~W _ U G ~W _ U U TWU ~W _ U TW ~U
0ROPOSITION &OR ANY ~ 5 B W,' AND ANY T :, THE FUNCTION
W,'
F~~, SATISlES
W ,' W
,'
F~~, U GU ~U ~U F~~, G
FOR U U 5 G T ` 5 W', 5 )N PARTICULAR F~~, IS INVARIANT UNDER THE

ADJOINT ACTION OF 5 ON T ` 5 W', 5
0ROOF 4HROUGHOUT THE PROOF WE DENOTE FOR SHORTNESS W W,' )N THIS
CASE WE HAVE 5 W 5, 6 W 50 WHERE 0 5 ` " WHICH IMPLIES
THAT .ORM5 6 W,' 5
4HE FOLLOWING RESULT IS OBVIOUS
,EMMA ,ET ~ 5 B W 4HEN FOR ANY U 5 AND T :, WE HAVE
_ _ `
~, W_ ` T_ U ` W_ ~W _ UW ~U _ ~, U
&URTHERMORE LET T :, G T ` 5 W5 U U 5 4HEN
W,' W , W_
_ `
F~~, U GU ~ U GU ~ rU rG rU
!PPLYING ,EMMA TWICE WE OBTAIN
W,' _ _
F~~, U GU ~W T U TW ~U ~W G
_ _ ` _
~, W_ ` TrU T_ ` W_ ~, rU ~, W rG
_
"Y ,EMMA WE HAVE ~, W_ rU W_ AND ~W _ T_ U TW
~U _ ~, U 4AKING INTO ACCOUNT THAT ~, rU ~, U WE OBTAIN
W,' W W _
F~~ , U GU ~U ~ G ~U ~ rG
4HEOREM IS PROVED c
0ROOF OF 4HEOREM 2ECALL THAT 7, 7* DENOTES THE 7EYL GROUP OF

, ,*
0ROPOSITION &OR ANY ~ 5 B W,' T 4 WE HAVE
~W,' U GU ~T_ U T ~U ~W,' G
FOR U 5, U 5 G T ` 5 W,' 5 )N PARTICULAR ~W,' IS INVARIANT UNDER
THE ADJOINT ACTION OF 5, ON T ` 5 W,' 5 FOR T :,
6ISIONS IN -ATH '%/-%42)# !.$ 5.)0/4%.4 #2934!,3
0ROOF 4AKING INTO ACCOUT THAT .ORMU 6 W,' 5 WE REWRITE

~W,' U GU ~W,' _ UTW,' ~U ~W,' G
~T_ U T ~U ~W,' G
0ROOF OF 4HEOREM B 7E HAVE SUPPW,' ) "Y FOR EACH

J * AND GENERIC B_ "W _ WE HAVE SJ B_ !D UB_ FOR SOME U 5I
,'
4HIS IMPLIES THAT FOR ALL J * THE RESTRICTION ~W,' J"W_ IS SJ INVARIANT
,'
4HEOREM B IS PROVED c
2ECALL FROM SECTION THAT THE 5 VARIETY 5W',

EQUIPPED WITH THE
ISOMORPHISM p W', 5 W', F _
"W',
IS A UNIPOTENT ' CRYSTAL
0ROOF OF 4HEOREM A 4HE PROOF IS BASED ON THE FOLLOWING PROPERTY
_F
OF THE BIREGULAR ISOMORPHISM pW "W 5 W WHICH IS DElNED IN SECTION
,EMMA &OR ANY ~ 5 B AND ANY W 7 WE HAVE
_
~pW B ~W rB
FOR B _
"W
0ROOF )T IS EASY TO SEE THAT FOR ANY ~ 5 B AND U 5 WE HAVE ~U_
B G 5 _ ` 4 ` 5 WE HAVE
_~U )T IS ALSO CLEAR THAT FOR ANY ~ 5
~T GT_
_~G
WHERE T y? _ 4 AD AS IN SECTION
&URTHERMORE BY DElNITION OF W ONE HAS
_
T WT_
W
_
#OMPOSING pW_ WITH ~ WE OBTAIN FOR B "W _ ~ 5 B
~pW B ~xWB_ _ _~xWB_ _~WB_
_ _ ` _
~T WB_ T_
~ W
_ T B_ T_
~W rB
.OW WE APPLY ,EMMA WITH W W', ~5 B W', 5SING

_
W,'
0ROPOSITION AND THE PROPERTY OF THE MAPPING B rB
r5W', 5 5 W,' 5 r!D 5, G !D 5, rG
WE SEE THAT THE RIGHT HAND SIDE OF IS AN !D 5, INVARIANT FUNCTION ON
"W_ 4HEREFORE BY THE RIGHT HAND SIDE OF IS INVARIANT UNDER
',
THE UNIPOTENT 7, ACTION
3INCE BY ITS DElNITION pW', IS AN ISOMORPHISM OF UNIPOTENT ' CRYSTALS
"_ 5W', pW',
W', IDW', F
! "%2%.34%). !.$ $ +!:($!. '!&!
THE RESTRICTION ~J5 W', IS ALSO INVARIANT UNDER THE UNIPOTENT ACTION OF 7*
4HEOREM A IS PROVED c
0ROOF OF 0ROPOSITION 0ART A FOLLOWS FROM PART B FOLLOWS

FROM #OROLLARY PART C FOLLOWS FROM 4HEOREM AND PART D IS
CLEAR
0ROOF OF 0ROPOSITION 0ART A IS IMMEDIATE 3O WE PROVE B $UE

TO ,EMMA AND 0ROPOSITION C IT SUnCES TO PROVE THAT FOR ANY
_
I 2W AND I 2W_ THE BIRATIONAL ISOMORPHISM ;xI =_ m p W m x I AND
ITS INVERSE ARE POSITIVE ISOMORPHISMS 4 F
4 WHERE 4 'M LW 4HIS
FOLLOWS FROM THE RESULTS OF ;": 3ECTIONS AND =
0ROJECTIONS OF "RUHAT #ELLS TO THE 0ARABOLIC 3UBGROUPS

'ENERAL FACTS ON PROJECTIONS 2ECALL THAT WE HAVE DElNED IN
SECTION THE MONOID STRUCTURE 7 ON 7 #LEARLY FOR ANY STANDARD
,EVI SUBGROUP , ,* THE MONOID 7, IS A SUB MONOID OF 7 UNDER THE
OPERATION DElNED IN SECTION AND THIS SUB MONOID 7, IS GENERATED
BY ALL SJ J *
,EMMA 4HE CORRESPONDENCE

SI IF I *
SI
;SI =
E IF I ) N *
EXTENDS TO THE HOMOMORPHISM OF MONOIDS ;`= 7 7,
3IMILARLY TO THE PROJECTION P_, "
_ " _ DElNED IN SECTION LET
,

P P, BE THE NATURAL PROJECTION 5 5,
4HERE IS A NATURAL EQUIVALENCE RELATION BETWEEN CONSTRUCTIBLE SUBSETS OF
' 7E SAY THAT TWO SUCH SUBSETS 3 3 y ' ARE EQUIVALENT IF THEIR INTERSEC
TION 3 < 3 IS DENSE IN EACH 3 AND 3 )N THIS CASE WE DENOTE 3 p 3
,EMMA &OR EACH W 7 ONE HAS P 5 W p 5 ;W=
0ROOF 7E PROCEED BY THE INDUCTION IN LW )F LW THAT IS W SI FOR
SOME I ) THEN

5 SI IF I *
P 5 SI
FEG IF I ) N *
6ISIONS IN -ATH '%/-%42)# !.$ 5.)0/4%.4 #2934!,3
,ET US NOW ASSUME THAT LW 5SING 0ROPOSITION AND THE FACT
THAT P IS A GROUP HOMOMORPHISM WE OBTAIN FOR ANY I I IL 2W
P 5 W p P 5 SIL 5 SIL_ ` ` ` ` ` ` ` 5 SI
P 5 SIL ` P 5 SIL_ ` ` ` ` ` ` ` P 5 SI p 5 ;SI =```;SIL = p 5 ;W=
,EMMA IS PROVED c
4HE COMPUTATION OF P_ _
, "W IS A LITTLE MORE COMPLICATED
0ROPOSITION &OR ANY W 7 WE HAVE
A 4HE INTERSECTION OF P_ _ _
, "W WITH ";W= IS A DENSE SUBSET OF ";W=
_
_ E
B 4HERE IS AN ALGEBRAIC SUB TORUS 4EW y 4 SUCH THAT P_ _
, "W p "W ` 4W
C 4HE TORUS 4EW SATISlES 4EW FEG IF AND ONLY IF W 7, 4ESI jI? 'M
FOR I ) N * AND 4EWW ;W =_ 4EW ` 4^W ` 4E;W=;W = FOR ANY W W SUCH
THAT W W WW
0ROOF )T FOLLOWS FROM 0ROPOSITION THAT FOR ANY W W 7 SATISFYING
W W WW WE HAVE
P_ " _ p P_ "W
, WW
_
,` " _ P_ "W
_
` P_ " _ p " _ ` 4EW ` " _ ` 4EW
W , , W ;W= ;W =
p _
";W= ` _
";W =
_
` ;W = 4^W ` 4EW ` 4E;W=;W =
4HIS PROVES B AND C ,ET US PROVE A NOW E _
$ENOTE " ;W=
P_ _ _
, "W < ";W= 7E WILL PROCEED BY INDUCTION IN LW )F W E THE
STATEMENT IS OBVIOUS .OW LET W E ,ET US EXPRESS W W W WHERE
W E W E AND LW LW LW 4HUS THE INDUCTIVE HYPOTHESIS
IMPLIES THAT " ^ _ IS DENSE IN " _ AND " E _ P_ " _ < " _ IS DENSE
;W = ;W = ;W = , W ;W =
_
IN ";W =
4HEN IMPLIES THAT

^_ ` "
P_ " _ z P_ " _ ` " _ z P_ " _ ` P_ " _ z " E _
, WW , W W , W , W ;W = ;W =
"UT "E_ ` "

E _ IS A DENSE SUBSET OF " _ ` " _ &INALLY 0ROPOSITION A
;W = ;W = ;W = ;W =
_ _
IMPLIES THAT THE LATTER SET CONTAINS A DENSE SUBSET OF ";W =;W = ";W W =
4HIS PROVES PART A 4HE PROPOSITION IS PROVED c
)T IS WELL KNOWN THAT THE SET ' 50_ ` 0 IS OPEN IN ' $ENOTE BY
P P, THE NATURAL PROJECTION ' 0 "Y DElNITION P COMMUTES WITH
THE ACTION OF ", a " AND THE RESTRICTION OF P TO " _ y ' IS THE NATURAL
PROJECTION P_ _
, " ",
_
/UR NEXT TASK IS TO DESCRIBE THE IMAGE OF EACH REDUCED "RUHAT CELL 5 W5
UNDER THE PROJECTION P
! "%2%.34%). !.$ $ +!:($!. '!&!
2ECALL THAT IN SECTION WE DElNED FOR EACH W 7 THE SUB TORUS
4W y 4
A 5 ;W=5 ` 4;W=_ W y P5 W5 < '
B 4HE RESTRICTION PW P,W OF P TO 5 W5 IS A DOMINANT RATIONAL MOR
PHISM
PW 5 W5 5;W=5 ` 4EW
E
C 4;W=_ W y 4W
0ROOF 0ROVE B 4HE MULTIPLICATION IN ' INDUCES THE OPEN INCLUSIONS
_ a 5 5 W5 < ' " _ a 5 5 W5 < ' &URTHERMORE
"W W , ,
_
P"W ` 5 P_ _
, "W ` 5
_
P"W ` 5, P_ _
, "W ` 5,
4HEREFORE IT FOLLOWS FROM ,EMMA THAT
P5W5 p P_ _ _ E _ E E
, "W ` 5 p ";W= ` 4W ` 5 ";W= ` 5 ` 4W p 5 ;W=5 ` 4W
0ART B IS PROVED
,ET US NOW PROVE A 2ECALL THAT P IS 5, a5 EQUIVARIANT AND 5 ;W=5
5, ;W=5 .OTE THAT P_ _ _
, "W INTERSECTS ";W= NON TRIVIALLY 4HUS
5 ;W=5 y P5 W5 < '

2ECALL THAT THE ACTION !DW 4 a ' ' IS DElNED IN THE PROOF OF
0ROPOSITION #LEARLY BOTH 0 AND THE REDUCED "RUHAT CELL 5 W5 ARE
INVARIANT UNDER THIS ACTION AND THE MORPHISM P ' 0 COMMUTES WITH
THIS ACTION 4HUS APPLYING ,EMMA TO THE ABOVE IDENTITY WE OBTAIN
!DW 4 5 ;W=5 5 ;W=5 ` 4;W=_ W y P5 W5 < '
0ART A IS PROVED 0ART C FOLLOWS
_
$ENOTE 50_W
50_ <5W5 `;W= 4HE SET 50_W IS NOT EMPTY BECAUSE
;W= P5 W5 < ' DUE TO 0ROPOSITION A .OTE ALSO THAT 50_W IS
CLOSED IN 50_ &OR ANY U_ 50_W LET QU_ 5 ;W=5 ` 4;W=_ W 5 W5 BE THE
MORPHISM DElNED BY
_ `
QU_ U ` ;W= ` U ` T U ` U_ ` ;W= ` U ` T
FOR U 5 U 6 ;W= AND T 4;W=_ W
&OR ANY W 7 LET 4 W FT 4 WT TG #LEARLY DIM 4 W ` 4W
DIM 4
6ISIONS IN -ATH '%/-%42)# !.$ 5.)0/4%.4 #2934!,3
A FOR EACH U_ 50_W THE MORPHISM QU_ IS A SECTION OF PW THAT IS

PW m QU_ ID5 ;W=5 `4
;W=_ W
_
B 4HE VARIETY 50_W IS INVARIANT UNDER THE ADJOINT ACTION OF 4 ;W= `W `
5, ;W=
C ,ET Q BE A MORPHISM 50_W a 5 ;W=5 ` 4;W=_ W 5 W5 DElNED BY
QU_ G QU_ G 4HEN Q IS AN INCLUSION
0ROOF 0ROVE A )NDEED 6 ;W= y 5, 4HUS QU_ IS A SECTION OF PW
BECAUSE PW IS 5, a " EQUIVARIANT
0ROVE B )T IS EASY TO SEE THAT 5, ;W= . ORM5, ;W= 4HUS FOR U_
_ _
50 W AND ANY U .ORM5, ;W= T 4 ;W= `W WE HAVE !DW TUÙ_ `;W=Ù_
5W5 /N THE OTHER HAND !DW TU ` U_ ` ;W= ` U_ TU ` U_ ` TU _ ` ;W=
4HIS IMPLIES THAT TU ` U_ ` TU _ 50_W
0ROVE C 7E PROCEED BY THE CONTRADICTION !SSUME THAT Q IS NOT
INJECTIVE 4HAT IS WE ASSUME THAT THERE ARE ELEMENTS U_ U_ 50_W SUCH
THAT U_ U_ AND
QU_ ;W= QU_ U ;W=UT
FOR SOME U 6 ;W= U 5 AND T 4 )N OTHER WORDS
U_ ` ;W= U ` U_ ` ;W= ` U ` T
$ENOTE U ^_ !D U U_ #LEARLY U ^_ 50_ BECAUSE U 5, &URTHERMORE
LET US FACTORIZE !D ;W=U U ` U_ WHERE U 5 ;W= AND U_ 5,_
4HUS THE ABOVE IDENTITY CAN BE REWRITTEN AS
U_ _ ` U_ U ` U ` U_ ` ;W=T
^
4HIS IMPLIES THAT T E U_ E U Ù E AND U_ U ^_ )N PARTICULAR
U U _ 6 ;W= < 5 ;W= 5, ;W= FEG

)N ITS TURN THIS IMPLIES THAT U_ U_ WHICH CONTRADICTS THE ORIGINAL
ASSUMPTION c
$EFINITION 7E CALL AN ELEMENT W 7 , SPECIAL OR SIMPLY SPECIAL IF

LW L;W= DIM 4;W=_ W
4HEOREM &OR ANY SPECIAL ELEMENT W 7 WE HAVE
A 4HE SET 50_W CONSISTS OF A SINGLE ELEMENT U_ U_
0 W 4HIS ELEMENT
_
U_ IS CENTRALIZED BY THE GROUP 4 ;W= `W ` 5, ;W= 4HE INCLUSION QU_
IS DENSE
B PW ;QU_ =_ )N PARTICULAR PW 5 W5 5 ;W=5 ` 4;W=_ W IS A
BIRATIONAL ISOMORPHISM
C 4EW 4;W=_ W
! "%2%.34%). !.$ $ +!:($!. '!&!
D 4HE COMPOSITION P_ , m p
W IS A BIRATIONAL ISOMORPHISM
W _
5 F ";W= ` 4;W=_ W
0ROOF &OLLOWS FROM 0ROPOSITION c
%XAMPLE OF THE PROJECTION FOR ' ',M N ,ET US lX TWO

INTEGERS M s N ,ET ' ',MN , ,MN ',M a ',N y ',MN
)N THIS CASE 0 0MN , ` " IS THE CORRESPONDING MAXIMAL PARABOLIC
SUBGROUP WHICH CONSISTS OF THOSE MATRICES G ',MN WHICH HAVE ZEROS IN
THE M a N RECTANGLE IN THE LOWER LEFT CORNER
7E USE THE STANDARD LABELING OF THE $YNKIN DIAGRAM OF TYPE !MN_
) F M N _ G ,ET * ) N FMG 4HEN THE 7EYL GROUP 7 IS
NATURALLY IDENTIlED WITH THE SYMMETRIC GROUP 3MN VIA SI I I
.OTE THAT THE LONGEST ELEMENT W OF 7 ACTS ON SIMPLE ROOTS BY W jI
_jIb FOR I M N _ WHERE Ib M N _ I
4HE CORRESPONDING ELEMENT W,' WMN 7 ADMITS A FACTORIZA
TION
&OR I s J DENOTE BY SIJ THE REmECTION ABOUT THE ROOT jI ` ` ` jJ
#LEARLY SIJ IS THE TRANSPOSITION I J 7 3MN
$ENOTE zMN ;WMN =_ WMN )T IS EASY TO SEE THAT
M
9
zMN SIIb
I
THAT IS zMN IS A PRODUCT OF EXACTLY M COMMUTING REmECTIONS IN 7 4HIS
IMPLIES THAT DIM 4zMN M LWMN _ L;WMN = 4HUS WMN IS SPECIAL
,ET QMN BE THE DENSE INCLUSION 5, ;WMN =5 `4zMN 5 WMN 5 PRESCRIBED
BY 4HEOREM
,ET MN BE THE FUNDAMENTAL WEIGHTS FOR ',MN &OR EACH
W 3MN THE WEIGHT WK IS NATURALLY IDENTIlED WITH A K ELEMENT SUBSET
OF F M NG
&OR ANY W W 3MN LET aW I W I ' ! BE THE MINOR LO
CATED IN THE INTERSECTION OF ROWS INDEXED BY W I AND COLUMNS INDEXED
BY W I
0ROPOSITION 4HE VARIETY 5, ;WMN =5, ` 4zMN CONSISTS OF ALL G
", ;WMN =", SATISFYING FOR I M N _

aM M G IF I ;M N=
a;WMN =I I G
a;WMN =Ib Ib G IF I
;M N=
6ISIONS IN -ATH '%/-%42)# !.$ 5.)0/4%.4 #2934!,3
0ROOF ,ET US DESCRIBE THE SUB TORUS 4zMN OF 4 7E HAVE

9M
4zMN {SIIb
I
4HAT IS 4zMN CAN BE DESCRIBED IN 4 BY THE EQUATIONS I M N _

aM M T IF I ;M N=
aI I T
aIb Ib T IF I ;M N=
/N THE OTHER HAND FOR ANY W 7* T 4 ONE HAS
5, W5, T FG ", W", aWI I G aI I T I M N _ G
0ROPOSITION &OR ANY ~ 5 B WMN U 5, U 5 T 4 WE HAVE

_ `
~WMN QMN U ;WMN = ` zMN T ` T_ ` U
~U ~U _ mMN jM T ` ~XM
0ROOF &OR ANY s I s M s J W 7* LET SÎJ .ORM' 4 BE THE
REPRESENTATIVE OF SIJ DElNED BY
SEIJ W _ SM W
FOR ANY W 7* SUCH THAT W_ SM W
SIJ
7E NEED THE FOLLOWING OBVIOUS RElNEMENT OF
_ 1
,EMMA ;WMN = WMN M I SG
IIb
&OR ANY s I s M CHOOSE AN ELEMENT WI 7 SUCH THAT SIIb
WI_ SM WI "Y WE HAVE
_
SG
IIb U _
I UI UI
WHERE U _ _ _
I W I XM _ W I UI WI YM WI
#LEARLY UI 50 < !D WMN _ 50_ AND U_
_
I 50 < !D WMN
_ 5
0

)T IS ALSO CLEAR THAT UI COMMUTES WITH EACH UJ WHENEVER I J WHERE
F _G
1
$ENOTE U M
I UI FOR F _G
,EMMA 4HE ONLY ELEMENT OF 50WMN IS EQUAL TO !D WMN U _
AND
U_ ` ;WMN = WMN ` U_ ` U
0ROOF ,ET U_ !D WMN U _ #LEARLY U_ 50_
.EXT LET US REWRITE THE EQUATION FROM ,EMMA WMN _ ;WMN =
U Ù_ Ù OR EQUIVALENTLY ;WMN = WMN Ù Ù_ Ù U_ _ Ù
WMN Ù
_
WHERE U !D WMN U #LEARLY U 50
! "%2%.34%). !.$ $ +!:($!. '!&!
_
4HUS U_ 50_ < 5WMN 5 ;WMN = 50WMN
,EMMA IS PROVED c
3INCE QMN IS 5, a 5 EQUIVARIANT IT SUnCES TO PROVE IN THE CASE

WHEN U U E 5SING ,EMMA WE OBTAIN
_ ` _ `
~WMN QMN !DWMN T;WMN = ~WMN !DWMN TQMN ;WMN =
_ `
~WMN !DWMN TWMN U_ U ~U_ U ` T_ ~TU T_
,ET US COMPUTE ~TU T_ USING THE FACT THAT U I ;5 5 = FOR I
M _
M
8
~TU T_ ~TU _ _
I T ~TUM T _mMN ~XM T
I
4HE LAST EQUALITY HOLDS BECAUSE TU
MT
_ ;5 5 = FOR M N AND TU T_
M
_
TXM _ T XM _jM T IF M N
4HE PROPOSITION IS PROVED c
,ET , ,N_M 'M M a ',N_M y ,MN "Y DElNITION ;WMN =

W, ,MN )T IS EASY TO SEE THAT ;WMN =4zMN y :,
2ECALL FROM 0ROPOSITION THAT FOR ANY ~ 5 B THE FUNCTION F ;WMN = ON
~~,
:, ` 5, ;WMN =5, DElNED BY IS PROPER UNDER THE ACTION OF 5, a 5,
#OROLLARY &OR ANY ~ 5 B WMN < 5
B ;WMN = AND G 5, ;WMN =5, `
T T 4zMN ONE HAS
;W =
~WMN QMN G F~~MN
,
G _ mMN AM ` T_
M
0MN_
WHERE AM IS THE COEnCIENT OF ~M IN THE EXPANSION ~ I AI ~I
2EMARK )T IS WELL KNOWN THAT 50MN IS ABELIAN IN THE ABOVE EXAMPLE /NE
CAN EXPECT THAT FOR ANY ' AND ANY STANDARD PARABOLIC SUBGROUP 0 , ` 50
THE ELEMENT W,' IS SPECIAL IF 50 IS ABELIAN
/N THE OTHER HAND THERE IS AN EXAMPLE OF A STANDARD PARABOLIC SUBGROUP
, IN THE SIMPLE GROUP ' OF THE TYPE $ WITH NON ABELIAN 50 AND NON SPECIAL
W,'
7E USE THE FOLLOWING LABELING OF THE $YNKIN DIAGRAM OF TYPE $ )
F G WHERE STANDS FOR THE <CENTER OF THE $YNKIN DIAGRAM ,ET *
F G AND , ,* 4HEN W, S S S AND W,' S S S S S S S S S
)T IS EASY TO SEE THAT ;W,' = W, 4HUS LW,' _ _ L;W,' = _
DIM 4;W,' =_ W,' WHICH IMPLIES THAT W,' IS NOT SPECIAL
-OREOVER PW IS NOT A BIRATIONAL ISOMORPHISM BECAUSE DIM 5 W,' 5
DIM 5, ;W,' =5 AND DIM 4EW,' s
6ISIONS IN -ATH '%/-%42)# !.$ 5.)0/4%.4 #2934!,3
!PPENDIX #OMBINATORIAL 0RE CRYSTALS AND 7 CRYSTALS

)N THIS SECTION WE RECALL +ASHIWARAS FRAMEWORK SEE ;+=
$EFINITION ! PARTIAL BIJECTION OF SETS F ! " IS A BIJECTION ! "
OF SUBSETS ! y ! " y " 7E DENOTE THE SUBSET ! BY DOMF AND THE
SUBSET " BY RANF
2EMARK ! PARTIAL BIJECTION F ! " IS AN EMBEDDING ! " IF AND
ONLY IF DOMF !
4HE INVERSE F _ OF A PARTIAL BIJECTION F ! " IS THE INVERSE BIJECTION
RANF DOMF #OMPOSITION G m F OF PARTIAL BIJECTIONS F ! "
G " # IS NATURALLY A PARTIAL BIJECTION WITH DOMG m F DOMF <
F _ RANF < DOMG AND RANG m F GRANF < DOMG )N PARTICULAR
FOR ANY PARTIAL BIJECTION F " " AND N : THE N TH POWER F N IS A
PARTIAL BIJECTION " "
.OTE THAT FOR ANY PARTIAL BIJECTION F ! " THE COMPOSITION F _ m F
IS THE PARTIAL IDENTITY BIJECTION IDDOMF ! !
$EFINITION ,ET " BE A SET l E " c? BE A MAP 7E CALL A FAMILY OF
PARTIAL BIJECTIONS COMBINATORIAL PRE CRYSTAL OR SIMPLY PRE CRYSTAL ON " l
E
IF
l EB jI?
EI B l
E^
FOR B DOM^ EI I )
7ITH ANY PRE CRYSTAL " WE DElNE PARTIAL BIJECTIONS SÎ " " I )
BY
_HE
l B jI I
SÎ B EÎ B
)N FACT SÎ ARE PARTIAL INVOLUTIONS SÎ IDDOM^SI
%XAMPLES &IX I ) $ENOTE BY l EI : c? THE MAP N NjI?
7E DElNE A PRE CRYSTAL "I ON : l EI IN THE FOLLOWING WAY 7E DElNE A
PARTIAL BIJECTION EÎ : : BY EÎ N N AND FOR ANY J I WE DElNE
DOM^ EJ RAN^ EJ 4HIS DElNES A PRE CRYSTAL ON : l EI WHICH WE
DENOTE BY "I SEE ;+ %XAMPLE =
4HE LATTICE c? IS A PRE CRYSTAL WITH l E IDJc? AND EÎ t? t? jI?
FOR t? c? I )
&OR ANY PRE CRYSTAL " ON " l E WE DENOTE BY l E THE MAP c? a" c?
GIVEN BY l ? ?
E t B t E l B /NE CAN DElNE A PRE CRYSTAL ON c? a" l Ec?
? ?
BY EÎ t B t EÎ B
&OR ANY PRE CRYSTAL " ON " l E PUT l b _l $ElNE ALSO THE PARTIAL
_
BIJECTIONS EÎ EÎ 4HE COLLECTION EÎb I ) DElNES THE STRUCTURE OF A
b
! "%2%.34%). !.$ $ +!:($!. '!&!
PRE CRYSTAL ON " l b 4HIS PRE CRYSTAL IS CALLED DUAL TO " AND IS DENOTED
BY "b
2EMARKS )N ;+= THE PARTIAL BIJECTION EÎ_ WAS DENOTED BY FÎ
/UR PRE CRYSTALS CORRESPOND TO +ASHIWARAS CRYSTALS FOR , =
$EFINITION ! PRE CRYSTAL " IS CALLED FREE IF EACH EÎ IS A BIJECTION " "
$EFINITION ! MORPHISM OF PRE CRYSTALS F " " IS A PARTIAL BIJECTION
F " " SUCH THAT DOMF " l
ElE m F AND THE PARTIAL BIJECTIONS

EÎ " " I ) ARE OBTAINED FROM THE PARTIAL BIJECTIONS EI " "
I ) BY EÎ F _ m EÎ m F
$EFINITION 7E SAY THAT A PRE CRYSTAL " IS A COMBINATORIAL 7 CRYSTAL OR
SIMPLY 7 CRYSTAL IF FOR ANY I ) WE HAVE DOM^ SI " AND THE INVOLUTIONS
SÎ SATISFY THE BRAID RELATIONS
)T IS CLEAR THAT A STRUCTURE OF A 7 CRYSTAL ON " lE DElNES AN ACTION OF
7 ON " IN SUCH A WAY THAT FOR ANY I ) SI 7 ACTS BY SÎ " "
2EMARKS 4HE CONDITION THAT DOM^ SI " IS EQUIVALENT TO
g N
h i j g h
INF N B DOM^
EI s _ lEB jI s SUP N B DOM^ EIN
FOR ALL B " I )
4HE PRE CRYSTAL ON c? IDc? DElNED ABOVE IS A FREE 7 CRYSTAL
&OR ANY 7 CRYSTAL THE STRUCTURE MAP l E " c? IS 7 EQUIVARIANT
7E DENOTE BY 0RE #RYST THE CATEGORY SUCH THAT THE OBJECTS ARE PRE
CRYSTALS AND ARROWS ARE MORPHISMS OF PRE CRYSTALS 7E DENOTE BY 7 #RYST
THE FULL SUB CATEGORY OF 0RE #RYST WHOSE OBJECTS ARE 7 CRYSTALS .OTE THAT
EACH MORPHISM F " " IN 7 #RYST IS AN INJECTIVE 7 EQUIVARIANT MAP
" "
%XAMPLE ,ET 6 BE A lNITE DIMENSIONAL 5Q , = MODULE AND LET "
"6 BE A CRYSTAL BASIS FOR 6 SEE ;+= $ENOTE BY l E " c? THE WEIGHT
GRADING 4HEN THE PARTIAL BIJECTIONS EÎ FÎ " " DElNE THE STRUCTURE OF
A PRE CRYSTAL ON " l )T IS A DEEP RESULT OF ;+= THAT THIS IS A 7 CRYSTAL
7E DENOTE THIS 7 CRYSTAL BY "6
2EMARK &OR " "6 ALL EI E_ I ARE NILPOTENT THAT IS DOM^ EIN
_N
DOM^ EI FOR SOME . AND _HE EIN G
l B jI I INFFN B DOM^
SUPFN B DOM^ N
EI G FOR B "6 I ) #LEARLY THIS IDENTITY IMPLIES
SI "6 FOR EACH I )
THE INEQUALITIES THAT IS DOM^
4HERE ARE EXAMPLES OF 7 CRYSTALS FOR WHICH LI LI_ p EG THERE
ARE EXAMPLES OF FREE 7 CRYSTALS !MONG THEM ARE THE FREE 7 CRYSTALS "I
6ISIONS IN -ATH '%/-%42)# !.$ 5.)0/4%.4 #2934!,3
FROM ;+ 3ECTION = BUILT ON :LW l EI I 2W +ASHIWARA HAS

CONSTRUCTED EACH "I AS THE TENSOR PRODUCT OF "I "IL SO THAT "I IS
EQUIPPED WITH THE FAMILY OF FUNCTIONS EII :LW : I ) SATISFYING
I
EI ^
EI B I LW
EI B FOR B : I )
,ET 2W BE THE CATEGORY WHOSE OBJECTS ARE REDUCED SEQUENCES
I 2W AND FOR EACH PAIR OF OBJECTS I I 2W THERE IS EXACTLY ONE
ARROW I I
! FUNCTOR & 2W _ 7 _ #RYST SUCH THAT & I "I WAS DElNED

IN ;":= -OREOVER IT WAS SHOWN IN ;":= THAT EII II m & I I FOR
ANY I I 2W 4HEREFORE THE FREE 7 CRYSTAL "W IS WELL DElNED AND
IS EQUIPPED WITH THE FAMILY OF FUNCTIONS ^ "W : I ) SUCH THAT
E EI B B
^ E
2EMARK ,ET 6t? BE THE SIMPLE 5Q , = MODULE WITH THE LOWEST WEIGHT
t EG t? IS AN ANTI DOMINANT CO WEIGHT AND LET "6t? BE THE CORRE
SPONDING COMBINATORIAL CRYSTAL )T FOLLOWS FROM THE RESULTS OF +ASHIWARA
THAT THERE IS A MORPHISM OF 7 CRYSTALS
Ft? "6t? c? a "W
_N
SUCH THAT FE t? B SUPFN B DOM^ EI G FOR B "6t? I )
2EFERENCES
;"&:= ! "ERENSTEIN 3 &OMIN ! :ELEVINSKY 0ARAMETRIZATIONS OF CANON
ICAL BASES AND TOTALLY POSITIVE MATRICES !DV IN -ATH [
;"+= ! "ERENSTEIN !N +IRILLOV 'ROUPS GENERATED BY INVOLUTIONS
'ELFAND 4SETLIN PATTERNS AND COMBINATORICS OF 9OUNG TABLEAUX IN 2US
SIAN !LGEBRA I !NALIZ [ 4RANSLATION IN 3T 0ETERSBURG
-ATH * [
;":= ! "ERENSTEIN ! :ELEVINSKY 4OTAL POSITIVITY IN 3CHUBERT VARIETIES
#OMMENT -ATH (ELV [
;":= ! "ERENSTEIN ! :ELEVINSKY 4ENSOR PRODUCT MULTIPLICITIES CANONICAL
BASES AND TOTALLY POSITIVE VARIETIES PREPRINT MATH 24
;"O= . "OURBAKI ,IE GROUPS AND ,IE ALGEBRAS #HAPTERS [ 4RANSLATED FROM
THE &RENCH %LEMENTS OF -ATHEMATICS 3PRINGER 6ERLAG "ERLIN .EW 9ORK

;"R'= ! "RAVERMAN $ 'AITSGORY #RYSTALS VIA THE AnNE 'RASSMANNIAN
PREPRINT MATH !'
;"R+= ! "RAVERMAN $ +AZHDAN l FUNCTIONS OF REPRESENTATIONS AND LIFTING
'!&! IN THIS ISSUE
;&:= 3 &OMIN ! :ELEVINSKY $OUBLE "RUHAT CELLS AND TOTAL POSITIVITY *
!MER -ATH 3OC [
! "%2%.34%). !.$ $ +!:($!. '!&!
;+= - +ASHIWARA #RYSTALIZING THE Q ANALOGUE OF UNIVERSAL ENVELOPING ALGE

BRAS #OMM -ATH 0HYS [
;+= - +ASHIWARA 4HE CRYSTAL BASE AND ,ITTELMANNS RElNED $EMAZURE
CHARACTER FORMULA $UKE -ATH * [
;+= - +ASHIWARA #RYSTAL BASES OF MODIlED QUANTIZED ENVELOPING ALGEBRA
$UKE -ATH * [
;,= 0 ,ITTELMANN ! ,ITTLEWOOD 2ICHARDSON RULE FOR SYMMETRIZABLE +AC
-OODY ALGEBRAS )NVENT -ATH [
;,U= ' ,USZTIG #ANONICAL BASES ARISING FROM QUANTIZED ENVELOPING ALGEBRAS
* !MER -ATH 3OC [
;,U= ' ,USZTIG )NTRODUCTION TO QUANTIZED ENVELOPING ALGEBRAS 0ROGRESS IN
-ATH * 4IRAO . 7ALLACH EDS "IRKH
AUSER "OSTON [
;,U= ' ,USZTIG )NTRODUCTION TO 1UANTUM 'ROUPS "IRKH AUSER "OSTON
!RKADY "ERENSTEIN $EPARTMENT OF -ATHEMATICS (ARVARD 5NIVERSITY #AM

BRIDGE -! 53! ARKADIY MATHHARVARDEDU
$AVID +AZHDAN $EPARTMENT OF -ATHEMATICS (ARVARD 5NIVERSITY #AMBRIDGE
-! 53! KAZHDAN MATHHARVARDEDU
l &5.#4)/.3 /& 2%02%3%.4!4)/.3 !.$ ,)&4).'
! "RAVERMAN AND $ +AZHDAN

!BSTRACT
,ET & BE A LOCAL NON !RCHIMEDEAN lELD AND LET & #a BE
A NON TRIVIAL ADDITIVE CHARACTER OF & 4O THIS DATA ONE ASSOCIATES
A MEROMORPHIC FUNCTION l x l x ON THE SET OF IRREDUCIBLE
REPRESENTATIONS OF THE GROUP ',N & IN THE FOLLOWING WAY #ONSIDER
THE INVARIANT DISTRIBUTION h TRG J DETG JN JDGJ ON ',N &
WHERE JDGJ DENOTES A (AAR DISTRIBUTION ON ',N & !LTHOUGH THE
SUPPORT OF h IS NOT COMPACT IT IS WELL KNOWN THAT FOR A GENERIC
IRREDUCIBLE REPRESENTATION x OF ',N & THE ACTION OF h IN x IS WELL
DElNED AND THUS IT DElNES A NUMBER l x 4HESE GAMMA FUNCTIONS
AND THE ASSOCIATED , FUNCTIONS WERE STUDIED BY * 4ATE FOR N AND
BY 2 'ODEMENT AND ( *ACQUET FOR ARBITRARY N
.OW LET ' BE THE GROUP OF POINTS OF AN ARBITRARY QUASI SPLIT REDUC
TIVE ALGEBRAIC GROUP OVER & AND LET '? BE THE ,ANGLANDS DUAL GROUP
,ET ALSO y '? ',N # BE A lNITE DIMENSIONAL REPRESENTATION
OF '? 4HE LOCAL ,ANGLANDS CONJECTURES PREDICT THE EXISTENCE OF A
NATURAL MAP Ly )RR' )RR',N & !SSUMING THAT A CERTAIN
TECHNICAL CONDITION ON y GUARANTEEING THAT THE IMAGE OF Ly DOES NOT
LIE IN THE SINGULAR SET OF l IS SATISlED ONE CAN CONSIDER THE MERO
MORPHIC FUNCTION ly ON )RR' SETTING ly x l Ly x
4HE MAIN PURPOSE OF THIS PAPER IS TO PROPOSE A GENERAL FRAMEWORK
FOR AN EXPLICIT CONSTRUCTION OF THE FUNCTIONS ly .AMELY WE PROPOSE
A CONJECTURAL SCHEME FOR CONSTRUCTING AN INVARIANT DISTRIBUTION hy
ON ' WHOSE ACTION IN EVERY x )RR' IS GIVEN BY MULTIPLICATION
BY ly 3URPRISINGLY THIS TURNS OUT TO BE CONNECTED WITH CERTAIN
GEOMETRIC ANALOGS OF - +ASHIWARAS CRYSTALS CF ;"E+=
7E WORK OUT SEVERAL EXAMPLES IN DETAIL !S A BYPRODUCT WE OB
TAIN A CONJECTURAL FORMULA FOR THE LIFTING Ly q WHERE q IS A CHARACTER
OF AN ARBITRARY MAXIMAL TORUS IN ',N & )N SOME OF THESE EXAM
PLES WE ALSO GIVE A DElNITION OF THE CORRESPONDING LOCAL , FUNCTION
AND STATE A CONJECTURAL y ANALOGUE OF THE 0OISSON SUMMATION FORMULA
FOR &OURIER TRANSFORM 4HIS CONJECTURE IMPLIES THAT THE CORRESPOND
ING GLOBAL , FUNCTION HAS MEROMORPHIC CONTINUATION AND SATISlES A
FUNCTIONAL EQUATION
! "2!6%2-!. !.$ $ +!:($!. '!&!
)NTRODUCTION AND 3TATEMENT OF THE 2ESULTS

4HE ,ANGLANDS PROGRAM AND THE LIFTING ,ET & BE EITHER A LOCAL
OR lNITE lELD AND & ITS SEPARABLE CLOSURE 7E WILL DENOTE ALGEBRAIC VARIETIES
OVER & AND MAPS BETWEEN THEM BY BOLDFACE LETTERS 4HE CORRESPONDING
ORDINARY LETTERS WILL DENOTE THE ASSOCIATED SETS OF & POINTS AND THE INDUCED
MAPS BETWEEN THEM
,ET ' BE A CONNECTED REDUCTIVE ALGEBRAIC GROUP OVER & ' '&
ITS GROUP OF & POINTS )RR' THE SET OF ISOMORPHISM CLASSES OF COMPLEX
IRREDUCIBLE REPRESENTATIONS OF ' &OR SIMPLICITY WE SHALL ASSUME IN THE
INTRODUCTION THAT ' IS SPLIT ALTHOUGH MOST OF OUR RESULTS GENERALIZE EASILY
TO QUASI SPLIT GROUPS 4O ' ONE ASSOCIATES A CONNECTED COMPLEX ALGEBRAIC
GROUP '? CALLED THE ,ANGLANDS DUAL GROUP
%XAMPLE )F ' ',N THEN '? ',N #
4HE LOCAL ,ANGLANDS CONJECTURES CAN NOW BE SUMMARIZED AS FOLLOWS 4O
THE lELD & ONE ASSOCIATES A PRO ALGEBRAIC GROUP =& OVER # WHICH IS VERY
CLOSELY RELATED WITH THE 'ALOIS GROUP 'AL& & IN THE CASE WHEN & IS A
LOCAL lELD IT IS CALLED THE 7EIL $ELIGNE GROUP
4HE ,ANGLANDS CONJECTURE SAYS THAT
4HERE EXISTS A lNITE TO ONE MAP FROM )RR' TO THE SET OF ALL '? ORBITS
ON (OM=& '?
)F ' ',N & THEN THE ABOVE MAP IS A BIJECTION
2EMARK 4HIS CONJECTURE IS KNOWN FOR A lNITE lELD DUE TO ,USZTIG )T
IS KNOWN FOR LOCAL lELDS WHEN ' IS A TORUS LOCAL CLASS lELD THEORY AND FOR
',N CF ;,A23= ;(= AND ;(4=
.OW LET y '? ',N # BE AN ALGEBRAIC REPRESENTATION OF '?
4HEN THE ABOVE CONJECTURE IMPLIES THAT THERE EXISTS A MAP Ly )RR'
)RR',N &
/UR MAIN TASK IS TO TRY TO PROPOSE A CONJECTURE FOR AN EXPLICIT DESCRIP
TION OF THE MAP Ly AND TO FORMULATE A CERTAIN y ANALOGUE OF THE 0OISSON
SUMMATION FORMULA
)N THE REST OF THIS PAPER UNTIL SECTION WE ASSUME THAT & IS A LOCAL
lELD -OREOVER WE ASSUME THAT THE CHARACTERISTIC OF & IS GOOD WITH RESPECT
TO ' AND THAT THE ,IE ALGEBRA = OF ' POSSESSES A NON DEGENERATE INVARIANT
FORM 7E DENOTE BY (' THE (ECKE ALGEBRA OF ' "Y DElNITION WHEN
& IS NON !RCHIMEDEAN RESP !RCHIMEDEAN THIS IS THE ALGEBRA OF LOCALLY
CONSTANT RESP # COMPACTLY SUPPORTED DISTRIBUTIONS ON ' 7E SHALL
CHOOSE A (AAR MEASURE JDGJ AND THUS IDENTIFY (' WITH THE SPACE #C '
6ISIONS IN -ATH l &5.#4)/.3 /& 2%02%3%.4!4)/.3 !.$ ,)&4).'
OF LOCALLY CONSTANT RESP # WHEN & IS !RCHIMEDEAN COMPACTLY SUPPORTED

FUNCTIONS ON '
l FUNCTIONS 7E WILL NOT TRY TO CONSTRUCT EXPLICITLY A REPRESENTA
TION Ly x x )RR' BUT GIVE AN INDIRECT DESCRIPTION OF THE LIFTING Ly
&OR THIS PURPOSE WE DElNE A COMPLEX VALUED FUNCTION l ON )RR',N&
#HOOSE ONCE AND FOR ALL A NON TRIVIAL ADDITIVE CHARACTER & #a
,ET JDGJ DENOTE THE UNIQUE (AAR MEASURE ON ' ',N & SUCH THAT THE
&OURIER TRANSFORM } &} DElNED BY THE FORMULA
:
&} Y }X TRXY J DETX JN JDXJ
'
IS UNITARY
#ONSIDER THE DISTRIBUTION h G TRG J DETG JN JDGJ ON ',N &
#LEARLY THE DISTRIBUTION h h IS INVARIANT UNDER THE ADJOINT ACTION .OW
LET x 6 BE AN IRREDUCIBLE REPRESENTATION OF ' #ONSIDER THE INTEGRAL
:
xh xG hG
'
/NE CAN SHOW CF ;'O*= THAT THE INTEGRAL IS CONVERGENT ALTHOUGH
NOT ABSOLUTELY CONVERGENT FOR GENERIC x )RR' 4HEREFORE FOR GENERIC
x THE OPERATOR xh IS WELL DElNED AND IT IS EQUAL TO MULTIPLICATION BY A
SCALAR lx /NE CAN CONSIDER l AS A MEROMORPHIC FUNCTION ON )RR' CF
SECTION FOR THE DElNITION OF THIS NOTION
.OW LET ' AND y BE AS ABOVE AND ASSUME THAT WE KNOW THE LIFTING MAP Ly
4HEN WE DElNE A FUNCTION ly ly ON THE SET )RR' BY ly Lyb l )N
OTHER WORDS ANY IRREDUCIBLE REPRESENTATION x OF ' WE DElNE
ly x lLy x
4HE FUNCTION ly IS A WELL DElNED MEROMORPHIC FUNCTION ON )RR' PRO
VIDED THAT Ly x DOES NOT LIE IN THE SINGULAR SET OF l FOR GENERIC x )RR'
)N ORDER TO GUARANTEE THIS WE SHALL ALWAYS EXCEPT FOR SECTION ASSUME
THAT y SATISlES THE FOLLOWING CONDITION THERE EXISTS A COCHARACTER z 'M
:'? OF THE CENTER :'? OF THE GROUP '? SUCH THAT y m z ? )D .OTE
THAT z CAN BE REGARDED AS A CHARACTER OF '
)T IS EASY TO SHOW ASSUMING THAT Ly SATISlES CERTAIN NATURAL PROPERTIES
THAT THERE EXISTS AN AD INVARIANT DISTRIBUTION hy hy' ON ' SUCH THAT
ly x xhy ` )D
7E SEE FROM THAT A LIFTING Ly OF A REPRESENTATION y OF THE DUAL
GROUP '? DETERMINES AN INVARIANT DISTRIBUTION hy ON ' 3INCE THE MAP Ly
! "2!6%2-!. !.$ $ +!:($!. '!&!
COLLAPSES THE , PACKETS WE ASSUME THAT THE FUNCTION hy' IS STABLE /UR
MAIN PURPOSE IS TO GIVE A CONJECTURAL DESCRIPTION OF THE FUNCTION hy' )N
GENERAL WE CAN ONLY PROPOSE A FRAMEWORK FOR SUCH A DESCRIPTION 3OMETIMES
WE CAN MAKE OUR SUGGESTION PRECISE AND IN SOME SIMPLEST CASES WHEN THE
LIFTING IS KNOWN WE CAN CHECK THAT OUR DElNITION IS CORRECT
4HE &OURIER TRANSFORM &y !SSUME NOW THAT WE KNOW THE LIFTING
Ly AND THEREFORE CAN CONSTRUCT THE DISTRIBUTION hy ON ' '& )N
SECTION WE DElNE CERTAIN &OURIER TYPE TRANSFORM OPERATOR &y ACTING IN
THE SPACE OF FUNCTIONS ON ' .AMELY FOR } (' WE DElNE
&y } JzJ_L_ hy b r }
WHERE r }G }G _ AND L IS THE SEMI SIMPLE RANK OF '

)N THE CASE WHEN ' ',N AND y IS THE STANDARD REPRESENTATION OF
'? ',N THE OPERATOR &y COINCIDES WITH THE USUAL &OURIER TRANSFORM
IN THE SPACE OF FUNCTIONS ON THE SPACE -N & OF N a N MATRICES OVER &
7E CONJECTURE THAT FOR A &y EXTENDS TO A UNITARY OPERATOR ACTING IN THE
SPACE ,y ' , ' JzJL JDGJ WHERE JDGJ IS A (AAR MEASURE ON '
.OTE THAT IN SECTION WE GIVE A DElNITION OF hy AND THUS OF &y IN
CERTAIN CASES WHERE LP IS NOT KNOWN
,OCAL , FUNCTIONS 5NTIL NOW WE HAVE DISCUSSED ONLY THE LOCAL
LIFTING "UT THE ONLY KNOWN FORMULATION OF THE LOCAL LIFTING CONJECTURE IS
TO INTERPRET IT AS A LOCAL PART OF THE GLOBAL LIFTING CONJECTURE 4HERE ARE
TWO APPROACHES TO A PROOF OF THE GLOBAL LIFTING CONJECTURE 4HE lRST AP
PROACH IS BASED ON THE 4RACE FORMULA AND THE SECOND ONE ON THE STUDY OF
, FUNCTIONS 3O IT IS NATURAL TO TRY TO GIVE A DIRECT DElNITION OF THE ,
FUNCTION ,Ly x S FOR x )RR' IN TERMS OF y AND x !SSUME THAT WE
KNOW THE DISTRIBUTION hy )N SUCH A CASE WE PRESENT A CONJECTURE FOR A DIRECT
DElNITION OF THE , FUNCTION ,Ly x S 4O ANY REPRESENTATION y OF '? AS
ABOVE WE ASSOCIATE A CERTAIN SUBSPACE 3y OF THE SPACE OF FUNCTIONS ON ' SUCH
THAT 3y CONTAINS THE SPACE 3' OF SMOOTH COMPACTLY SUPPORTED FUNCTIONS
ON ' 7E CONJECTURE THAT 3y IS INVARIANT UNDER &y &OR ANY REPRESENTATION
x )RR' WE DElNE THE FUNCTION ,Ly x S 2AS THE COMMON DENOMINATOR OF
RATIONAL OPERATOR VALUED FUNCTIONS xS } G' }G JzG JS xG JDGJ WHERE
} 3y )N THE CASE WHEN ' ',N AND y IS THE STANDARD REPRESENTATION
OF '? ',N THE SPACE 3y COINCIDES WITH THE SPACE OF 3CHWARTZ "RUHAT
FUNCTIONS ON THE SPACE -N OF N a N MATRICES WITH ENTRIES IN & AND OUR
DElNITION OF THE , FUNCTION COINCIDES WITH THE DElNITION FROM ;'O*=
6ISIONS IN -ATH l &5.#4)/.3 /& 2%02%3%.4!4)/.3 !.$ ,)&4).'
#ONJECTURAL APPLICATIONS TO AUTOMORPHIC , FUNCTIONS .OW

LET + DENOTE A GLOBAL lELD AND LET ALSO !+ BE THE CORRESPONDING ADELE RING
&OR A PLACE F OF + WE DENOTE BY +F THE CORRESPONDING LOCAL COMPLETION
OF + )N ;'O*= 'ODEMENT AND *ACQUET DElNED THE , FUNCTION ,x S
FOR EVERY AUTOMORPHIC REPRESENTATION x OF ',N !+ 4HIS , FUNCTION IS
DElNED AS THE PRODUCT OF THE CORRESPONDING LOCAL , FUNCTIONS OVER ALL PLACES
OF + )T IS SHOWN IN ;'O*= THAT ,x S IS MEROMORPHIC AND HAS A FUNCTIONAL
EQUATION 4HE MAIN TOOLS IN THE CONSTRUCTION AND THE PROOF ARE THE SPACE
3-N & OF 3CHWARTZ "RUHAT FUNCTIONS ON N a N MATRICES THE &OURIER
TRANSFORM ACTING IN THIS SPACE AND THE 0OISSON SUMMATION FORMULA
!SSUME NOW THAT WE ARE GIVEN A GROUP ' AND NON SINGULAR A REPRESEN
TATION y OF '? 2 ,ANGLANDS CF ;,= CONJECTURED THAT ONE COULD DElNE
AN , FUNCTION ,y x S ATTACHED TO EVERY AUTOMORPHIC REPRESENTATION x OF
'!+ AS A PRODUCT OF LOCAL FACTORS 7E CONJECTURE THAT THESE LOCAL FACTORS
ARE EQUAL TO ONES DElNED IN SECTION )N SECTION WE DElNE A GLOBAL
VERSION 3y '!+ AND DElNE A GLOBAL y ANALOGUE &y OF THE &OURIER TRANS
FORM "Y THE DElNITION. THE SPACE 3y '!+ IS THE SPAN OF FUNCTIONS y
WHICH ARE PRODUCTS } }F OF LOCAL FACTORS 7E CONJECTURE THE EXISTENCE
OF AN &y INVARIANT DISTRIBUTION y ON 3y '!+ SUCH THAT

8
y } }G
G'+
IF SOME LOCAL FACTOR }F HAS COMPACT SUPPORT ON 'F

&OR ANY } 3y '!+
y } y &y }
%XISTENCE OF SUCH CAN BE THOUGHT OF AS A y ANALOGUE OF THE 0OISSON
SUMMATION FORMULA )N THE CASE WHEN ' ',N AND y IS THE STAN
DARD REPRESENTATION OF '? ',N OUR CONJECTURE SPECIALIZES TO THE USUAL
0OISSON SUMMATION FORMULA 4HE VALIDITY OF OUR y ANALOGUE OF THE 0OISSON
SUMMATION FORMULA WOULD IMPLY THAT THE AUTOMORPHIC , FUNCTION ,y x S
HAS MEROMORPHIC CONTINUATION WITH ONLY A lNITE NUMBER OF POLES AND SAT
ISlES A FUNCTIONAL EQUATION
!PPLICATION TO LIFTING )N SECTION WE GIVE A CONSTRUCTION OF THE
DISTRIBUTION hy IN THE CASE WHEN ' ',N a ',M FOR ARBITRARY N AND
M AND WHEN y IS EQUAL TO THE TENSOR PRODUCT OF THE STANDARD REPRESENTATIONS
OF ',N # AND ',M # !S A BYPRODUCT WE GET A CONJECTURAL FORMULA
FOR Ly q WHERE q IS A CHARACTER OF AN ARBITRARY MAXIMAL TORUS IN ',N &
! "2!6%2-!. !.$ $ +!:($!. '!&!
)DEA OF THE CONSTRUCTION OF hy )N SECTION WE CHECK THAT

IN THE CASE WHEN ' IS A SPLIT TORUS 4 THE LIFTING AND THE DISTRIBUTION
hy4 ON 4 ARE WELL DElNED FOR ANY REPRESENTATION y4 OF 4? SATISFYING
THE CONDITION FROM SECTION -OREOVER ONE CAN CONSTRUCT AN ALGEBRAIC
VARIETY 9y4 A MAP P4 9y4 4 A FUNCTION F4 ON 9y4 AND A TOP
FORM 4 `jTOP 9y4 SUCH THAT THE DISTRIBUTION hy4 IS EQUAL TO THE
PUSH FORWARD P4 F4 J4 J WHERE J4 J IS THE MEASURE ON 9y4 ASSOCI
ATED WITH THE FORM 4 CF ;7E= 7E SAY THAT hy4 9y4 Fy4 y4
IS AN ALGEBRO GEOMETRIC DISTRIBUTION REPRESENTING THE DISTRIBUTION hy4 AND
SAY THAT THE DISTRIBUTION hy4 IS A MATERIALIZATION OF AN ALGEBRO GEOMETRIC
DISTRIBUTION hy4
.OW LET ' BE AN ARBITRARY REDUCTIVE GROUP 7E CONJECTURE THAT THE !D
INVARIANT DISTRIBUTION hy IS STABLE AND THEREFORE COMES FROM A DISTRIBUTION
^ y ON THE SET '!D& OF & RATIONAL POINTS OF THE GEOMETRIC QUOTIENT
h
'!D 3PEC/' ' OF ' BY THE ADJOINT ACTION 3INCE '!D 47
WHERE 4 y ' IS A MAXIMAL SPLIT TORUS AND 7 IS THE 7EYL GROUP OF ' WE CAN
CONSIDER h^ y AS A DISTRIBUTION ON THE SET 47 & 7E CONJECTURE THAT THE
DISTRIBUTION h^ y IS A MATERIALIZATION OF AN ALGEBRO GEOMETRIC DISTRIBUTION hy
WHICH IS A <DESCENT OF h y4 4 WHERE y4 IS THE RESTRICTION OF y ON 4? -ORE
PRECISELY WE CONJECTURE THAT THERE EXISTS AN ACTION OF THE 7EYL GROUP 7 ON
h y4 SUCH THAT THE DISTRIBUTION h^ y ON THE SET 47 & IS THE MATERIALIZATION
OF THE ALGEBRO GEOMETRIC DISTRIBUTION hy ON 47 WHICH IS A <DESCENT OF
h y4 4 WE SHALL EXPLAIN THE <DESCENT PROCEDURE CAREFULLY IN SECTION
)N SECTION WE GIVE EXPLICIT FORMULAS FOR THIS ACTION IN A NUMBER OF
CASES
4HIS PAPER IS ORGANIZED AS FOLLOWS )N SECTION WE FORMULATE THE

,ANGLANDS LIFTING CONJECTURE AND DISCUSS SEVERAL EXAMPLES )N SECTION WE
DElNE l FUNCTIONS AND FORMULATE THEIR CONJECTURAL PROPERTIES )N SECTION
WE GIVE AN EXPLICIT CONSTRUCTION OF THE LIFTING FOR THE CASE WHEN OUR RE
DUCTIVE GROUP IS A SPLIT TORUS 4 AND WE WRITE AN EXPLICIT FORMULA FOR THE
CORRESPONDING DISTRIBUTION hy4 3ECTION AND SECTION ARE DEVOTED RE
SPECTIVELY TO THE DElNITION OF THE 3CHWARTZ SPACE 3y AND AN ANALOGUE OF THE
0OISSON SUMMATION FORMULA FOR THE OPERATOR &y )N SECTION WE DEVELOP
THE NOTION OF AN ALGEBRO GEOMETRIC DISTRIBUTION 5SING THIS NOTION WE RE
DUCE CONJECTURALLY THE PROBLEM OF CONSTRUCTING THE DISTRIBUTION hy TO A
PURELY ALGEBRAIC QUESTION
3ECTION IS DEVOTED TO THE DISCUSSION OF THIS QUESTION IN THE FOLLOWING
6ISIONS IN -ATH l &5.#4)/.3 /& 2%02%3%.4!4)/.3 !.$ ,)&4).'
CASE ,ET M N BE TWO POSITIVE INTEGERS $ElNE THE GROUP

g h
'M N ! " ',M a ',N J DET! DET"
4HE DUAL GROUP 'M N ? IS THE QUOTIENT OF ',M # a ',N # BY THE
SUBGROUP CONSISTING OF ALL PAIRS OF MATRICES T)DM T_ )DN FOR T #a (ENCE
THE TENSOR PRODUCT yM iyN OF THE STANDARD REPRESENTATIONS OF ',M # AND
',N # DESCENDS TO A REPRESENTATION y OF 'M N ? 4HE CORRESPONDING
l FUNCTION IS ESSENTIALLY CONSTRUCTED IN ;*03=
)N SECTION WE CONSTRUCT hy IN THIS CASE 5SING ;"E+= ONE CAN CHECK
THAT OUR DElNITION GIVES RISE TO ALMOST THE SAME l FUNCTION AS THE ONE
DElNED IN ;*03= !S AN APPLICATION WE GIVE CONJECTURAL FORMULAS FOR THE
LIFTING PROBLEM DISCUSSED IN SECTION
)N SECTION WE EXPLAIN HOW TO CONSTRUCT hy IN THE CASE WHEN THE lELD
& IS lNITE -ORE PRECISELY IN THIS CASE WE CONSTRUCT A PERVERSE SHEAF h y
ON ' SUCH THAT hy IS OBTAINED FROM IT BY TAKING TRACES OF THE &ROBENIUS
MORPHISM IN THE lBERS
&INALLY THE !PPENDIX BY 6 6OLOGODSKY CONTAINS A PROOF OF A RESULT
ABOUT NON !RCHIMEDEAN OSCILLATING INTEGRALS USED THROUGHOUT THE PAPER
.OTATION %VERYWHERE EXCEPT FOR SECTION & DENOTES A LOCAL lELD
)F & IS NON !RCHIMEDEAN THEN WE DENOTE BY /& y & ITS RING OF INTEGERS
AND BY Q THE NUMBER OF ELEMENTS IN THE RESIDUE lELD OF &
)F 8 IS EITHER A TOTALLY DISCONNECTED TOPOLOGICAL SPACE OR A # MANIFOLD
WE DENOTE BY #8 THE SPACE OF # VALUED CONTINUOUS FUNCTIONS ON 8 7E
DENOTE BY #C 8 THE SPACE OF ALL COMPACTLY SUPPORTED # VALUED FUNCTIONS
ON 8 WHICH ARE
LOCALLY CONSTANT IF 8 IS TOTALLY DISCONNECTED
# IF 8 IS A # MANIFOLD
7E SHALL DENOTE ALGEBRAIC VARIETIES OVER & BY BOLDFACE LETTERS EG
' 8 4HE CORRESPONDING ORDINARY LETTERS ' 8 WILL DENOTE THE
CORRESPONDING SETS OF & POINTS
,ET 8 BE A SMOOTH ALGEBRAIC VARIETY OVER & AND `8 jTOP 8
WHERE jTOP DENOTES THE SHEAF OF DIkERENTIAL FORMS OF DEGREE DIM8 ON 8
!CCORDING TO ;7E= TO ONE CAN ASSOCIATE A DISTRIBUTION JJ
)F ' IS A TOPOLOGICAL GROUP RESP A ,IE GROUP THEN WE DENOTE BY )RR'
THE SET OF ISOMORPHISM CLASSES OF IRREDUCIBLE ALGEBRAIC CF ;"ER:= REP
RESENTATIONS OF ' RESP THE SET OF EQUIVALENCE CLASSES OF CONTINUOUS IRRE
DUCIBLE "ANACH REPRESENTATIONS OF ' WITH RESPECT TO INlNITESIMAL EQUIVA
LENCE CF ;7=
! "2!6%2-!. !.$ $ +!:($!. '!&!
!CKNOWLEDGMENTS 7E ARE GRATEFUL ! "ERENSTEIN * "ERNSTEIN

6 $RINFELD $ 'AITSGORY - (ARRIS - +ONTSEVICH 3 2ALLIS AND 6 6OL
OGODSKY FOR VERY HELPFUL AND INTERESTING DISCUSSIONS ON THE SUBJECT 7E ARE
ALSO GRATEFUL TO 9 &LICKER FOR NUMEROUS REMARKS ABOUT THIS PAPER
,IFTING
5NRAMIlED LIFTING ,ET & BE A NON !RCHIMEDEAN LOCAL lELD AND '
A QUASI SPLIT GROUP OVER /& WHICH CAN BE SPLIT OVER AN UNRAMIlED EXTENSION
OF & OF DEGREE K WE ASSUME THAT K IS MINIMAL WITH THIS PROPERTY 4HEN
THE CYCLIC GROUP :K: ACTS NATURALLY ON '? 7E DENOTE THE ACTION OF THE
GENERATOR OF :K: BY G sG
4HE GROUP '? ACTS ON ITSELF BY G X GXsG _ ,ET 3' DENOTE THE
SET OF CLOSED ORBITS ON '? WITH RESPECT TO THE ABOVE ACTION
,ET )RRUN ' DENOTE THE SET OF ISOMORPHISM CLASSES OF UNRAMIlED REPRE
SENTATIONS OF ' IE THE SET OF IRREDUCIBLE REPRESENTATIONS OF ' WHICH HAVE A
NON ZERO '/& INVARIANT VECTOR 4HEN ONE HAS THE NATURAL IDENTIlCATION
)RRUN ' 3'
,ET y '? J :K: ',N BE A REPRESENTATION 4HE MAP X X s
GIVES RISE TO A WELL DElNED MAP FROM 3' TO THE SET OF SEMISIMPLE CONJUGACY
CLASSES IN '? J :K: )F WE COMPOSE THIS MAP WITH y WE GET A MAP FROM
3' TO 3',N
#ONSIDER NOW THE COMPOSITE MAP
)RRUN ' 3' 3',N )RRUN ',N &
7E WILL DENOTE THIS MAP BY LyUN AND CALL IT THE UNRAMIlED LIFTING
4HE GLOBAL LIFTING ,ET + BE A GLOBAL lELD AND !+ ITS RING OF
ADELES 7E DENOTE BY 0+ THE SET OF PLACES OF + &OR EVERY F 0+ WE
LET +F BE THE CORRESPONDING LOCAL COMPLETION OF +
.OW LET ' BE A QUASI SPLIT REDUCTIVE GROUP OVER + WHICH SPLITS OVER A
lNITE SEPARABLE EXTENSION OF + WITH 'ALOIS GROUP ` !S IS WELL KNOWN ANY
IRREDUCIBLE REPRESENTATION x OF THE GROUP '!+ CAN BE UNIQUELY WRITTEN
AS A RESTRICTED TENSOR PRODUCT iF0+ xF WHERE xF )RR'+F AND xF
)RRUN '+F FOR ALMOST ALL F
7E DENOTE BY !UT' & THE SET OF IRREDUCIBLE AUTOMORPHIC REPRESENTA
TIONS OF '!+
4HE GROUP ` ACTS NATURALLY ON '? (ENCE WE CAN CONSIDER THE SEMIDIRECT
PRODUCT '? J ` &OR ALMOST ALL F 0+ THE GROUP ' CAN BE SPLIT OVER
6ISIONS IN -ATH l &5.#4)/.3 /& 2%02%3%.4!4)/.3 !.$ ,)&4).'
AN UNRAMIlED EXTENSION OF +F OF DEGREE KF AND IN THIS CASE THE GROUP

'? J :KF: IS NATURALLY EMBEDDED IN '? J `
(ENCE EVERY REPRESENTATION y '? J` ',N DElNES A REPRESENTATION
OF '? J :KF: FOR ALMOST ALL F
$EFINITION ,ET x ixF BE AN AUTOMORPHIC REPRESENTATION OF '!+
4HE GLOBAL LIFTING Ly x IS AN AUTOMORPHIC REPRESENTATION Ly x iLy x F OF
THE GROUP ',N !+ SUCH THAT FOR EVERY PLACE F 0+ SUCH THAT xF IS
UNRAMIlED ONE HAS Ly x F LyUN xF WHERE LyUN IS AS IN SECTION
2EMARK )T FOLLOWS FROM THE STRONG MULTIPLICITY ONE THEOREM FOR ',N
CF ;0= THAT IF Ly x EXISTS THEN IT IS UNIQUE
#ONJECTURE &OR EVERY ' x AND y AS ABOVE THE LIFTING Ly x EXISTS
,ET ' BE A CONNECTED QUASI SPLIT REDUCTIVE GROUP OVER A LOCAL lELD &
WHICH SPLITS OVER A NORMAL SEPARABLE EXTENSION ,& SUCH THAT 'AL,& `
4HEN FOR EVERY REPRESENTATION y '? J ` ',N # THERE EXISTS A MAP
Ly& )RR' )RR',N & SUCH THAT FOR ANY GLOBAL lELD + A PLACE
F 0+ SUCH THAT & +F A QUASI SPLIT GROUP ' OVER + AS IN SECTION
SUCH THAT 'F ' AND EVERY x !UT' + x ixF THE F TH COMPONENT
OF THE AUTOMORPHIC REPRESENTATION Ly x IS EQUAL TO Ly& xF
7E SAY THAT y IS LIFTABLE IF #ONJECTURE HOLDS FOR y
2EMARK )T IS EASY TO SEE THAT THE LOCAL LIFTING Ly IS UNIQUELY DETERMINED
BY THE CONDITIONS OF #ONJECTURE
%XAMPLE /NE OF THE EXAMPLES THAT WE WOULD LIKE TO CONSIDER IN THIS
PAPER IS THE FOLLOWING ,ET %& BE A SEPARABLE EXTENSION OF & OF DEGREE N
,ET 4% 2ES%& 'M% WHERE 2ES%& DENOTES THE FUNCTOR OF RESTRICTION OF
SCALARS ,ET `& DENOTE THE ABSOLUTE 'ALOIS GROUP OF & `& ACTS NATURALLY ON
THE SET OF ALL EMBEDDINGS OF % INTO & WHICH IS A lNITE SET WITH N ELEMENTS
(ENCE WE GET A HOMOMORPHISM j% `& 3N WHICH IS DElNED UNIQUELY
UP TO 3N CONJUGACY ,ET ` )Mj% "Y THE DElNITION ` IS A SUBGROUP OF
3N WHICH IS ALSO A QUOTIENT OF `& -OREOVER IT IS CLEAR THAT THE TORUS 4%
SPLITS OVER THE 'ALOIS EXTENSION ,& WHERE 'AL,& `
4HE DUAL TORUS 4? IS ISOMORPHIC TO #a N AND THE GROUP ` ACTS ON IT
BY THE RESTRICTION OF THE STANDARD ACTION OF 3N ON #a N TO ` (ENCE THE
STANDARD EMBEDDINGS OF #a N AND 3N TO ',N # GIVE RISE TO A HOMO
MORPHISM y4 4? J ` ',N # /NE OF THE PURPOSES OF THIS PAPER IS
TO GIVE A CONJECTURAL CONSTRUCTION OF Ly IN THIS CASE
! "2!6%2-!. !.$ $ +!:($!. '!&!
'AMMA FUNCTIONS AND "ERNSTEINS #ENTER

"ERNSTEINS CENTER ,ET ' BE A REDUCTIVE ALGEBRAIC GROUP OVER &
' '& 7E DENOTE BY -' THE CATEGORY OF SMOOTH REPRESENTATIONS
OF '
2ECALL THAT THE "ERNSTEIN CENTER :' OF ' IS THE ALGEBRA OF ENDOMOR
PHISMS OF THE IDENTITY FUNCTOR ON THE CATEGORY -' )T WAS SHOWN BY
"ERNSTEIN THAT :' IS ISOMORPHIC TO THE DIRECT PRODUCT OF ALGEBRAS OF THE
FORM /j WHERE j IS A lNITE DIMENSIONAL IRREDUCIBLE COMPLEX ALGEBRAIC
VARIETY 7E SHALL DENOTE THE SET OF ALL j AS ABOVE BY # "Y 3PEC:' WE
SHALL MEAN THE DISJOINT UNION OF ALL j # )T IS SHOWN IN ;"ER$= THAT THE
NATURAL MAP )RR' 3PEC:' SENDING EVERY IRREDUCIBLE REPRESENTATION
TO ITS INlNITESIMAL CHARACTER IS lNITE TO ONE AND MOREOVER IS GENERICALLY ONE
TO ONE
4HE FOLLOWING RESULT ALSO DUE TO * "ERNSTEIN GIVES A DIkERENT INTERPRE
TATION OF :' ,ET h BE A DISTRIBUTION ON ' 7E SAY THAT h IS ESSENTIALLY
COMPACT IF FOR EVERY FUNCTION } (' ONE HAS h b } (' 'IVEN
TWO ESSENTIALLY COMPACT DISTRIBUTIONS h h ONE DElNES THEIR CONVOLUTION
h b h BY SETTING
h b h } h h b }
)T IS EASY TO SEE THAT h b h IS AGAIN AN ESSENTIALLY COMPACT DISTRIBUTION
,EMMA 4HE ALGEBRA :' IS NATURALLY ISOMORPHIC TO THE ALGEBRA OF
INVARIANT ESSENTIALLY COMPACT DISTRIBUTIONS
%XAMPLE ,ET ' ',N ' ',N & &IX ANY A & a AND A
NON TRIVIAL ADDITIVE CHARACTER & #a ,ET 'A FG ' J DETG AG
4HEN A CHOICE OF A (AAR MEASURE ON 3,N & DElNES A MEASURE ON 'A FOR
EVERY A & a SINCE 'A IS A PRINCIPAL HOMOGENEOUS SPACE OVER 3,N &
7E DENOTE THIS MEASURE BY DA G $ElNE A DISTRIBUTION hA ON ' SUPPORTED
ON 'A BY
:
hA } }TRG DA G
'A
FOR ANY } #C '

/NE CAN VERIFY THAT hA IS ESSENTIALLY COMPACT
2ATIONAL FUNCTIONS ON )RR' ,ET 1 +' DENOTE THE <lELD OF

FRACTIONS OF :' IE THE ASSOCIATIVE ALGEBRA +j WHERE j RUNS OVER
CONNECTED COMPONENTS OF 3PEC:' AS ABOVE AND +j IS THE lELD OF
6ISIONS IN -ATH l &5.#4)/.3 /& 2%02%3%.4!4)/.3 !.$ ,)&4).'
FRACTIONS OF /j 3INCE GENERICALLY 3PEC:' AND )RR' ARE IDENTIlED

WE SHALL REFER TO ELEMENTS OF +' AS RATIONAL FUNCTIONS ON )RR'
z REGULAR CENTRAL ELEMENTS &OR AN ASSOCIATIVE ALGEBRA ! WE LET
-' ! DENOTE THE CATEGORY OF ' MODULES ENDOWED WITH AN ACTION OF !
BY ENDOMORPHISMS OF THE ' MODULE STRUCTURE WE CALL IT THE CATEGORY OF
' ! MODULES
,ET ! #;T T_ = " #T !SSUME THAT WE ARE GIVEN AN & RATIONAL
CHARACTER z ' 'M 4HEN WE CAN DElNE AN ACTION OF ' ON ! AND " BY
REQUIRING THAT
G ` T JzG JT
FOR EVERY G '
$ElNE THE FUNCTOR &! -' ! -' " SUCH THAT &! 6
6 i! " 3ET :z ' %ND &! THE %ND IS TAKEN IN THE CATEGORY OF FUNCTORS
FROM -' ! TO -' "
7E WOULD LIKE NOW TO IDENTIFY :z ' WITH A SUBSPACE OF +' 4HE
ASSIGNMENT x x i JzJLOGQ Z GIVES RISE TO AN ACTION OF #b ON )RR' HERE
Z #b )T IS EASY TO SEE THAT THIS ACTION DESCENDS TO A WELL DElNED ALGEBRAIC
ACTION OF #a ON 3PEC:' &OR EVERY F +' WE CAN WRITE
8
F Z ` X FN X Z N
FOR SOME FN +'
,EMMA :z ' IS NATURALLY ISOMORPHIC TO THE SPACE OF ALL F +'
SUCH THAT FN :' FOR EVERY N :
,EMMA EXPLAINS THE TERM <z REGULAR
z COMPACT DISTRIBUTIONS .OW LET h BE A DISTRIBUTION ON ' AND
LET 'N z _ x N /b 'N IS AN OPEN SUBSET OF ' &OR EVERY N : WE DElNE
A NEW DISTRIBUTION hN ON ' BY hN ~N h WHERE ~N IS THE CHARACTERISTIC
FUNCTION OF 'N
$EFINITION 7E SAY THAT h IS z COMPACT IF THE FOLLOWING TWO CONDI
TIONS ARE SATISlED
hN IS ESSENTIALLY COMPACT FOR EVERY N :
&OR EVERY } (' THERE EXISTS A RATIONAL FUNCTION &} # ('
SUCH THAT FOR EVERY G ' ONE HAS
8
Z N hN b } G &} Z G
N:
! "2!6%2-!. !.$ $ +!:($!. '!&!
4HE FOLLOWING LEMMA IS STRAIGHTFORWARD FROM THE DElNITIONS AND

,EMMA
,EMMA 4HE SPACE OF ALL z COMPACT DISTRIBUTIONS IS NATURALLY ISOMOR
PHIC TO :z '
%XAMPLE ,ET ' ',N & #HOOSE A NON TRIVIAL ADDITIVE CHAR
ACTER & #a AND SET h G TRG J DETG JN JDGJ WHERE JDGJ IS A
(AAR MEASURE ON ' ,ET ALSO z ',N 'M BE GIVEN BY zG DETG
0ROPOSITION h IS z COMPACT
0ROOF ,ET DA G J DET JN JDGJ BE THE <ADDITIVE MEASURE ON ' AND LET ALSO
3-N DENOTE THE SPACE OF 3CHWARTZ "RUHAT FUNCTIONS ON N a N MATRICES
-N OVER & 7E SHALL REGARD 3-N AS A SUBSPACE OF #' #ONSIDER THE
&OURIER TRANSFORM & 3-N 3-N GIVEN BY
:
&} Y }X TRXY DA X
'
)T IS EASY TO SEE THAT &} h b r } J DET J_N WHERE r }X }X_
,ET US NOW SHOW THAT CONDITION HOLDS 7ITHOUT LOSS OF GENERALITY
WE MAY ASSUME THAT SUPP } y ' 4HEN Q _N hN b r} &} ~_N 3INCE
&} 3-N IT FOLLOWS THAT &} ~_N (' 4HE VERIlCATION OF THE
SECOND CONDITION IS LEFT TO THE READER c
7E SHALL DENOTE BY l THE RATIONAL FUNCTION ON )RR',N & WHICH

CORRESPONDS TO h
,IFTING AND THE DISTRIBUTIONS h y .OW LET y '? ',N #
BE A lNITE DIMENSIONAL REPRESENTATION OF THE ,ANGLANDS DUAL GROUP '?
4HE LOCAL LIFTING CONJECTURE PREDICTS THE EXISTENCE OF THE NATURAL MAP Ly
)RR' )RR',N & !SSUMING THAT WE KNOW Ly WE CAN TRY TO DElNE
A RATIONAL FUNCTION ly BY SETTING ly x l Ly x (OWEVER IT MIGHT
HAPPEN THAT THE IMAGE OF Ly LIES ENTIRELY INSIDE THE SINGULAR SET OF Ly 4O
GUARANTEE THAT THIS DOES NOT HAPPEN WE INTRODUCE THE FOLLOWING NOTION
$EFINITION 7E SAY THAT y IS ADMISSIBLE IF THE FOLLOWING CONDITIONS
HOLD
+ERy IS CONNECTED
4HERE EXISTS A CHARACTER z ' 'M DElNED OVER & SUCH THAT
y m z )D
WHERE WE REGARD z AS A COCHARACTER OF THE CENTER :'? OF '? AND
)N ',N # DENOTES THE IDENTITY MATRIX
6ISIONS IN -ATH l &5.#4)/.3 /& 2%02%3%.4!4)/.3 !.$ ,)&4).'
)T IS EASY TO SEE THAT IF WE ASSUME THAT y IS AN ADMISSIBLE AND LIFTABLE

THEN ly SATISlES THE CONDITIONS OF ,EMMA 4HUS WE CAN CONSTRUCT THE
CORRESPONDING z COMPACT DISTRIBUTION hy' /UR MAIN TASK WILL BE TO TRY
TO CONSTRUCT THE DISTRIBUTION hy' EXPLICITLY WE WILL SOMETIMES DROP THE
INDICES AND ' WHEN IT DOES NOT LEAD TO CONFUSION
#OMPATIBILITY WITH INDUCTION ,ET 0 y ' BE A PARABOLIC SUB
GROUP OF ' DElNED OVER & AND - THE CORRESPONDING ,EVI SUBGROUP ,ET m0
BE THE DETERMINANT OF THE ACTION OF - ON THE ,IE ALGEBRA OF THE UNIPOTENT
RADICAL 50 OF 0
&OR EVERY } (' WE CAN CONSTRUCT A NEW FUNCTION R0 } ON '50
:
R0 } G }U JDUJ
50
.OTE THAT - ACTS ON '50 ON THE RIGHT
.OW LET y BE A REPRESENTATION OF '? AS ABOVE 3INCE -? IS NATURALLY A
SUBGROUP OF '? WE CAN REGARD y AS A REPRESENTATION OF -? 7E CONJECTURE
THAT THE FOLLOWING PROPERTY HOLDS FOR THE DISTRIBUTIONS hy' AND hy-
#ONJECTURE
_ `
R0 hy' b } R0 } b hy- Jm0 J_
!SSUME THAT y y g y IS A DIRECT SUM OF TWO ADMISSIBLE REP
RESENTATIONS SUCH THAT THE CORRESPONDING CHARACTER z IS THE SAME FOR BOTH
REPRESENTATIONS 4HE FOLLOWING RESULT CAN BE EASILY DEDUCED FROM THE DEl
NITIONS
,EMMA /NE HAS
hy hy b hy
l FUNCTIONS DETERMINE THE LIFTING ,ET ' ',N_ a',N
AND THE CHARACTER z BE GIVEN BY zGN_ GN DETGN_ )N ;'+ #H =
A RATIONAL FUNCTION lN_N WAS DElNED ON THE SUBSET OF NON DEGENERATE
REPRESENTATIONS OF THE GROUP ' /NE CAN SHOW THAT lN_N EXTENDS TO A
z REGULAR CENTRAL ELEMENT FOR ' (ENCE THERE EXISTS A z REGULAR DISTRIBUTION
hN_N ON THE GROUP ',N _ & a ',N & SUCH THAT lN_N lhN_N
)N ;*03= THE ANALOGOUS FUNCTION lMN ON )RR',M & a ',N & WAS
DElNED )T IS ALSO EASY TO SEE THAT IT COMES FROM A z REGULAR DISTRIBUTION
hMN ON THE GROUP ',M & a ',N &
!S SHOWN IN ;'+= FOR ANY NON DEGENERATE REPRESENTATION x OF

',N & THE RATIONAL FUNCTION lx lN_N x ON )RR ',N_ & DETER
MINES UNIQUELY THE REPRESENTATION x -OREOVER ;'+= CONTAINS AN EXPLICIT
! "2!6%2-!. !.$ $ +!:($!. '!&!
RECIPE FOR THE CONSTRUCTION OF THE REPRESENTATION x IN TERMS OF THE FUNCTION

lx )T IS CLEAR FROM THIS RECIPE THAT WE HAVE TO KNOW THE FUNCTION lx ONLY
UP TO MULTIPLICATION BY A CONSTANT
.OW LET ' BE ARBITRARY AND y '? ',N # BE AN AD
MISSIBLE REPRESENTATION &OR ANY POSITIVE INTEGER M WE DENOTE BY yM
'? a ',M # ',MN # THE REPRESENTATION GIVEN BY yM G ? R
yG? i R !SSUME THAT BOTH y AND yM ARE LIFTABLE &OR ANY x )RR'

WE DElNE RATIONAL FUNCTIONS lx lx ON )RR ',M & BY lx z lyM x z

lx z lNM Ly x z
0ROPOSITION &OR ANY x )RR' THE TWO RATIONAL FUNCTIONS lx AND
lx ON )RR ',M & COINCIDE
4HIS PROPOSITION FOLLOWS FROM THE MAIN RESULT OF ;(=
#OROLLARY &OR ANY y '? ',N # SUCH THAT THE REPRESEN
TATION yN_ IS LIFTABLE THE LIFTING Ly IS DETERMINED BY THE KNOWLEDGE OF THE
FUNCTION lyM x ON )RR ',N _ & -OREOVER IT IS SUnCIENT TO KNOW
THE FUNCTION lyM x ONLY UP TO A MULTIPLICATION BY A CONSTANT
4HE #ASE OF 3PLIT 4ORI

)N THIS SECTION WE ASSUME THAT & IS NON !RCHIMEDEAN
,IFTING FOR SPLIT TORI !LMOST THE ONLY CASE WHEN Ly IS KNOWN FOR
EVERY REPRESENTATION y IS THE CASE WHEN ' IS EQUAL TO A SPLIT TORUS 4 )N
THIS SECTION WE SHOW HOW TO CARRY OUT OUR PROGRAM IN THIS EASY CASE
7E START WITH AN EXPLICIT DESCRIPTION OF THE LIFTING Ly
7E CAN WRITE y IN THE FORM
N
-
y tI
I
WHERE tI ARE CHARACTERS OF 4? "Y THE DElNITION OF 4? WE CAN REGARD

tI AS COCHARACTERS tI 'M 4 OF 4 ,ET c 8b 4 BE THE GROUP OF
COCHARACTERS OF 4 7E DElNE SUPPy y c AS THE UNION OF tI s I s N
,ET q BE A CHARACTER OF 4 AND ~I qmtI BE THE CORRESPONDING CHARACTERS
OF & a 4HEN FOR EVERY I N THERE EXISTS ZI 2 SUCH THAT J~I T J
V T
ZI & FOR EVERY T & a
,ET US ASSUME THAT THE ORDERING t tN IS CHOSEN IN SUCH A WAY THAT
JZ J s JZ J s ` ` ` s JZN J
6ISIONS IN -ATH l &5.#4)/.3 /& 2%02%3%.4!4)/.3 !.$ ,)&4).'
,ET "N BE THE SUBGROUP OF UPPER TRIANGULAR MATRICES IN ',N & 7E

CAN DElNE THE CHARACTER ~ "N #a BY SETTING
N
9
~B ~I BII
I
WHERE BII ARE THE DIAGONAL ENTRIES OF B

',N&
#ONSIDER THE UNITARY INDUCED REPRESENTATION I"N ~ )T IS WELL KNOWN
CF ;"ER:= THAT THIS REPRESENTATION HAS UNIQUE IRREDUCIBLE QUOTIENT 7E
DElNE THIS QUOTIENT TO BE Ly q !S FOLLOWS FROM THE THEORY OF %ISENSTEIN
SERIES THIS DElNITION SATISlES THE CONDITIONS OF #ONJECTURE
l FUNCTIONS FOR SPLIT TORI 7E WILL GIVE NOW AN EXPLICIT DESCRIP
TION OF THE DISTRIBUTION hy4 ,ET 4? N
y 'M BE THE CORRESPONDING TORUS IN
',N 4HUS WE GET A HOMOMORPHISM Py 4? 4? ?
y ,ET 4y BE THE SPLIT
TORUS OVER & DUAL TO 4? y THUS 4y & a N 7E THEN GET AN & RATIONAL
MAP Py 4y 4
#ONSIDER A TOP DEGREE DIkERENTIAL FORM y DT DTN ON 4y AND A
FUNCTION Fy 4y ! WHERE Fy T TN T TN
&OR EVERY R 2t WE DENOTE BY 4y R THE SET FT 4y J JFy T J s RG
0ROPOSITION &OR EVERY R 2t AND FOR EVERY OPEN COMPACT
SUBSET # y 4 THE INTEGRAL
:
R
hy # Fy T Jy J
TP_
y # <4y R
IS ABSOLUTELY CONVERGENT
&OR EVERY # AS ABOVE THE LIMIT
LIM hRy #
R
EXISTS 7E DENOTE THIS LIMIT BY hy # 7E ALSO DENOTE BY hy THE CORRE

SPONDING DISTRIBUTION ON 4
&OR EVERY CHARACTER q OF 4 THERE EXISTS S q 2 SUCH THAT FOR EVERY
S # SUCH THAT 2ES S q THE CONVOLUTION hy b qJzJS IS ABSOLUTELY
CONVERGENT AND
hy b qJzJS ly4 qJzJS
hy IS z REGULAR
/UR NEXT GOAL IS TO DESCRIBE EXPLICITLY THE SPACE 3y IN THE CASE OF A TORUS
! "2!6%2-!. !.$ $ +!:($!. '!&!
4HE &OURIER TRANSFORM &y 7E ASSUME NOW THAT THE REPRESEN
TATION y IS FAITHFUL 4HIS ASSUMPTION WILL BE KEPT UNTIL THE END OF SECTION
#ONSIDER THE MAP &y #C 4 # 4 GIVEN BY
} JzJ_ hy b r }
WHERE r }X }X_
,ET ALSO ,y 4 , 4 JzJDb T 4HE FOLLOWING LEMMA IS STRAIGHTFORWARD
,EMMA &y EXTENDS TO A UNITARY AUTOMORPHISM OF ,y 4
7E NOW SET
3y 4 #C 4 &y #C 4 y ,y 4
THE READER SHOULD COMPARE THIS DElNITION WITH THE DElNITION OF A 3CHWARTZ
SPACE 38 IN ;"R+=
4HE FUNCTION #y ,ET US GIVE AN EXAMPLE OF A FUNCTION IN 3y 4
,ET 4 y 4 BE THE MAXIMAL COMPACT SUBGROUP OF 4 4HEN WE CAN IDENTIFY
THE QUOTIENT 4 4 WITH THE LATTICE c OF COCHARACTERS OF 4
,ET
-N
y tI
I
BE AN ADMISSIBLE REPRESENTATION OF 4? 7E DENOTE THE SET OF ALL tI ABOVE
BY SUPPy
,ET US DElNE A 4 INVARIANT FUNCTION 3y ON 4 IE A FUNCTION ON 4 4
c BY SETTING
g k h
#y u A AN :N k A t AN tN u
,EMMA #y 3y 4
4HE SEMIGROUP 4y 7E NOW WANT TO EXHIBIT CERTAIN LOCALITY PROP
ERTIES OF THE SPACE 3y 4 &OR THIS WE lRST HAVE TO INTRODUCE ADDITIONAL
NOTATION ,ET cy BE THE SUB SEMIGROUP OF c GENERATED BY t SUPPy /y
BE THE GROUP ALGEBRA OF cy OVER & AND 4y 3PEC /y )T IS EASY TO SEE THAT
4y IS AN & SEMI GROUP WHICH CONTAINS 4 AS AN OPEN DENSE SUBGROUP
0ROPOSITION ,ET } BE A FUNCTION ON 4 4HEN } 3y 4 IF AND ONLY
IF } SATISlES THE FOLLOWING CONDITIONS
4HE CLOSURE OF SUPP} IN 4 y IS COMPACT
&OR EVERY X 4 y THERE EXISTS A NEIGHBOURHOOD 5X OF X IN 4 y AND A
FUNCTION } 3y 4 SUCH THAT
}J5X } J5X
6ISIONS IN -ATH l &5.#4)/.3 /& 2%02%3%.4!4)/.3 !.$ ,)&4).'
0ROPOSITION SAYS THAT ONE CAN DETERMINE WHETHER A FUNCTION } LIES
IN 3y 4 BY LOOKING AT ITS LOCAL BEHAVIOUR AROUND POINTS OF 4 y 7E WILL
DISCUSS THE NOTION OF A LOCAL SPACE OF FUNCTIONS IN MORE DETAIL IN SECTION
&OR EVERY X 4y WE DENOTE BY :X THE STABILIZER OF X IN 4a4 )F X 4 y
WE DElNE AN :X MODULE 3yX AS THE QUOTIENT 3y ' 3yX 4 WHERE 3 4
yX
THE SPACE OF ALL FUNCTIONS FROM 3y ' WHICH VANISH IN SOME NEIGHBOURHOOD
OF X
,EMMA 4HERE EXISTS A UNIQUE :X INVARIANT FUNCTIONAL X 3yX #
SUCH THAT X #y
4HUS GIVEN } 3y 4 WE CAN PRODUCE A FUNCTION } ON 4 SETTING } X
X } (OWEVER THE FUNCTION } IS NOT LOCALLY CONSTANT IN GENERAL
-ELLIN TRANSFORM 7E ARE NOW GOING TO GIVE YET ANOTHER EQUIV
ALENT DElNITION OF 3y 4 USING THE -ELLIN TRANSFORM ON 4 ,ET 84
(OM4 #a 4HEN 84 HAS THE NATURAL STRUCTURE OF AN ALGEBRAIC VA
RIETY WHOSE IRREDUCIBLE COMPONENTS ARE PARAMETRISED BY (OM4 #a
(OM4 3 #HOOSE A (AAR MEASURE Db T ON 4
,ET c BE THE LATTICE OF COCHARACTERS OF 4 !S BEFORE ONE CAN IDENTIFY
c WITH 4 4 WHERE 4 y 4 IS THE MAXIMAL COMPACT SUBGROUP OF 4 7E
DENOTE BY 8 UN 4 y 84 THE SUBSET OF UNITARY CHARACTERS OF 4
7E DENOTE BY 8 4 THE COMPONENT OF 84 CONSISTING OF CHARAC
TERS WHOSE RESTRICTION TO 4 IS TRIVIAL /NE HAS THE NATURAL IDENTIlCATION
8 4 4? WHERE 4? DENOTES THE DUAL TORUS TO 4 OVER # &OR ANY
t SUPPy AND CHOOSE A LIFT t OF t TO 4 4HEN WE DENOTE BY Pt THE
REGULAR FUNCTION ON 84 DElNED BY

~t IF ~ m tJ/b
Pt ~
OTHERWISE
#LEARLY THIS DElNITION DOES NOT DEPEND ON THE CHOICE OF t
&OR } #C 4 WE DElNE THE -ELLIN TRANSFORM - } OF } AS A FUNCTION
ON 84 GIVEN BY
:
- } ~ }T ~T Db T
4
4HE FOLLOWING LEMMA IS STANDARD
,EMMA - DElNES AN ISOMORPHISM BETWEEN #C 4 AND /84
WHICH EXTENDS TO AN ISOMORPHISM BETWEEN , 4 AND , 8 UN 4
,ET y BE AN ADMISSIBLE REPRESENTATION OF 4? AND &y DENOTE THE OP
ERATOR DElNED ABOVE 4HE FOLLOWING RESULT IS PROVED BY A STRAIGHTFORWARD
! "2!6%2-!. !.$ $ +!:($!. '!&!
CALCULATION
4HEOREM
1 ! FUNCTION } ,y 4 LIES IN 3y 4 IF AND ONLY IF
sIsN PtI _ - } IS A REGULAR FUNCTION ON 84
,ET US NOW DISCUSS THE CONNECTION OF THE SPACE 3y 4 WITH THE CORRE
SPONDING LOCAL , FUNCTIONS
,EMMA &OR EVERY } 3y 4 AND EVERY CHARACTER ~ OF 4 THE
INTEGRAL
:
:} ~ S }T ~G JzT JS Db T
4
IS ABSOLUTELY CONVERGENT FOR ES |
:} M S HAS A MEROMORPHIC CONTINUATION TO THE WHOLE OF # AND
DElNES A RATIONAL FUNCTION OF QS
4HE SPACE OF ALL :} M S AS ABOVE IS A lNITELY GENERATED NON ZERO
FRACTIONAL IDEAL OF THE RING #;Q S Q _S = 7E SHALL DENOTE THIS IDEAL BY *~ ,ET
ALSO ,y S ~ TO BE THE UNIQUE GENERATOR OF *~ OF THE FORM 0 Q _S _ WHERE
0 IS A POLYNOMIAL1SUCH THAT 0
,~ S sIsN ,~ m tI S WHERE ,~ m tI S DENOTES THE CORRE
SPONDING 4ATES , FUNCTION CF ;4=
,OCAL 3PACES OF &UNCTIONS

7E NOW WANT TO GENERALIZE SOME CONSTRUCTIONS OF THE PRECEDING SECTION TO
THE CASE OF ARBITRARY REDUCTIVE GROUP ' ,ET US REMEMBER THAT WE ASSUME
NOW THAT THE REPRESENTATION y IS ADMISSIBLE AND FAITHFUL
)N THIS SECTION WE ASSUME THAT WE ARE GIVEN AN ADMISSIBLE REPRESENTA
TION y OF '? AND THE CORRESPONDING DISTRIBUTION hy WHICH SATISlES THE
ASSUMPTIONS OF SECTION
3ATURATIONS ,ET 8 BE A TOPOLOGICAL SET AND 8 y 8 AN OPEN DENSE
SUBSET 'IVEN A SPACE , OF FUNCTIONS 8 WE SAY THAT , IS LOCAL IF THERE EXISTS
A SHEAF , ON 8 AND AN EMBEDDING , Ib #8 SUCH THAT , `C 8 ,
WHERE #8 IS THE SHEAF OF CONTINUOUS FUNCTIONS ON 8
)T IS EASY TO SEE THAT WHEN THE SPACE 8 IS TOTALLY DISCONNECTED THEN
LOCALITY OF , IS EQUIVALENT TO THE FOLLOWING TWO CONDITIONS
#C 8 y , y #8
, IS CLOSED UNDER MULTIPLICATION BY ELEMENTS OF #8
'IVEN A SUBSPACE 6 y #8 WE DENOTE BY 06 THE PRESHEAF ON 8 SUCH
THAT `5 06 IS A SUBSPACE OF } #5 < 8 CONSISTING OF ALL } #5 < 8
6ISIONS IN -ATH l &5.#4)/.3 /& 2%02%3%.4!4)/.3 !.$ ,)&4).'
SUCH THAT THERE EXISTS V 6 FOR WHICH VJ5 <8 } ,ET ,6 BE THE ASSOCIATED
SHEAF )T IS CLEAR THAT WE HAVE THE NATURAL EMBEDDING ,6 Ib #8 7E
SET 6 `C 8 ,6 AND CALL 6 THE SATURATION OF 6 WITH RESPECT TO 8
# VERSION .OW LET 8 BE A CLOSED SUBSET OF A # MANIFOLD 9
,ET D` ` BE A METRIC ON 8 COMING FROM A 2IEMANNIAN METRIC ON 9
,ET I 8 8 BE AN OPEN DENSE EMBEDDING OF # MANIFOLD 8 INTO 8
,ET ALSO , y Ib # 8 BE A SUBSHEAF 7E DENOTE BY ÀS
C 8 , THE SPACE OF

# FUNCTIONS } ON 8 SUCH THAT FOR ANY X 8 AND ANY . THERE EXISTS
AN OPEN NEIGHBOURHOOD 5 OF X AND A SECTION L `5 , SUCH THAT FOR ANY
Y 5 ONE HAS k k
k}Y _ LY k s #DX Y .
WHERE # IS A CONSTANT IE # DOES NOT DEPEND ON Y
%XAMPLE ,ET 8 8 AND LET , BE THE SHEAF OF POLYNOMIAL
FUNCTIONS ON 8 4HEN ÀS
C 8 , #C
'IVEN A SUBSET 6 OF # 8 WE CAN DElNE ITS SATURATION 6 AS ÀS
C 8 ,6
WHERE ,6 IS DElNED AS IN SECTION
4HE &OURIER TRANSFORM &y )N THIS SUBSECTION WE DElNE A CERTAIN
<TWISTED ANALOGUE OF THE &OURIER TRANSFORM IN THE SPACE OF FUNCTIONS OF '
ATTACHED TO EVERY y AS ABOVE !S BEFORE WE DENOTE BY ,y ' THE SPACE
OF , FUNCTIONS ON ' WITH RESPECT TO THE MEASURE JzJL JDGJ WHERE L IS THE
SEMI SIMPLE RANK OF ' AND JDGJ IS A (AAR MEASURE ON '
&OR EVERY } #C WE DElNE A NEW FUNCTION &y }
&y } JzJ_L hy b r }
WHERE r }G
}G _
)N WHAT FOLLOWS WE WILL ASSUME THE VALIDITY OF THE
FOLLOWING CONJECTURE
#ONJECTURE &y EXTENDS TO A UNITARY OPERATOR ON THE SPACE ,y '
AND 3y IS &y INVARIANT
%XAMPLE ,ET ' ',N & AND y BE THE STANDARD REPRESENTATION OF
'? ',N )N THIS CASE ,y ' IS THE SAME AS , -N WITH RESPECT TO
THE ADDITIVE MEASURE ON -N AND &y COINCIDES WITH THE &OURIER TRANSFORM
ON -N WHERE WE IDENTIFY -N WITH THE DUAL VECTOR SPACE BY MEANS OF THE
FORM H! "I TR!"
4HE SEMIGROUP 'y 7E WOULD LIKE TO EMBED ' AS AN OPEN SUBSET
IN A LARGER AnNE VARIETY 'y IN A ' a ' EQUIVARIANT WAY
4HE VARIETY 'y IS AN ALGEBRAIC SEMIGROUP CONTAINING ' AS AN OPEN
DENSE SUBGROUP )T IS WELL KNOWN CF ;6= THAT IN ORDER TO DElNE SUCH
! "2!6%2-!. !.$ $ +!:($!. '!&!
A SEMIGROUP ONE NEEDS TO EXHIBIT A SUBCATEGORY OF THE CATEGORY OF lNITE

DIMENSIONAL ' MODULES CLOSED UNDER SUBQUOTIENTS EXTENSIONS AND TENSOR
PRODUCTS
7E WILL DElNE NOW THE CATEGORY OF y POSITIVE REPRESENTATIONS OF ' WHICH
WILL SATISFY THE ABOVE PROPERTIES ,ET t 4? 'M BE A CHARACTER WHICH
HAS NON ZERO MULTIPLICITY IN y 7E CAN REGARD t AS A COCHARACTER OF 4 7E
SAY THAT A ' MODULE 6 IS y POSITIVE IF FOR EVERY t AS ABOVE THE REPRESEN
TATION y m t OF 'M IS ISOMORPHIC TO A DIRECT SUM OF CHARACTERS OF THE FORM
T TI FOR I t )T IS CLEAR THAT THE CATEGORY OF y POSITIVE REPRESENTATIONS
SATISlES THE PROPERTIES DISCUSSED ABOVE AND THUS DElNES A SEMI GROUP 'y
4HE SPACE 3y ' ,ET
6y #C ' &y #C ' y ,y '
5NLIKE IN THE CASE OF THE TORUS ONE CAN SHOW THAT THE SPACE 6y IS ALMOST
NEVER LOCAL THIS IS NOT SO FOR EXAMPLE WHEN ' ', AND y IS THE
STANDARD REPRESENTATION OF '? ', # 7E DElNE 3y ' TO BE THE
SATURATION OF 6y THE DElNITION MAKES SENSE BOTH FOR !RCHIMEDEAN AND
NON !RCHIMEDEAN &
4HE FUNCTION #y ,ET US NOW EXHIBIT A CERTAIN EXPLICIT ELEMENT IN
3y 4 )N WHAT FOLLOWS WE CHOOSE A SQUARE ROOT Q OF Q
4HE 3ATAKE TRANSFORM ,ET + '/ AND LET (+
#C +N'+ BE THE CORRESPONDING (ECKE ALGEBRA 2ECALL THAT THERE EXISTS
?
THE NATURAL ISOMORPHISM 3 (+ /4? 7 /'? ' WHERE /4? 7
IS THE ALGEBRA OF 7 INVARIANT REGULAR FUNCTIONS ON 4? 4HIS ISOMORPHISM IS
CHARACTERIZED AS FOLLOWS /NE CAN IDENTIFY 4? WITH THE GROUP OF UNRAMIlED
CHARACTERS OF THE TORUS 4 &OR EVERY t 4? LET It DENOTE THE CORRESPOND
ING REPRESENTATION OF ' OBTAINED BY NORMALIZED INDUCTION OF THE CHARACTER
t FROM " TO ' ,ET Vt It BE THE UNIQUE UP TO A CONSTANT NON ZERO
+ INVARIANT VECTOR IN It 4HEN FOR EVERY } #C +N'+ ONE HAS
}JDGJ ` Vt 3} t Vt
,ET US ALSO DElNE THE TWISTED 3ATAKE ISOMORPHISM 3E BY REQUIRING THAT

E q _Q Hqm' I 3CHq
3CH
WHERE CHq DENOTES THE CHARACTER OF THE IRREDUCIBLE REPRESENTATION OF '?
WITH HIGHEST WEIGHT q AND m' IS THE SUM OF ALL POSITIVE ROOTS OF '
4HE FUNCTION #y ,ET y BE AS ABOVE
6ISIONS IN -ATH l &5.#4)/.3 /& 2%02%3%.4!4)/.3 !.$ ,)&4).'
3ET
_
FIy t TR t 3YMI y
FOR EVERY I :
7E NOW DElNE

8
#y 3E_ FIy
I
)T IS EASY TO SEE THAT THE NON DEGENERACY ASSUMPTIONS ON y IMPLY THAT #y IS

A WELL DElNED LOCALLY CONSTANT FUNCTION ON ' WHICH HOWEVER DOES NOT HAVE
COMPACT SUPPORT
,EMMA #y 3' AND &y # #y
4HE LEMMA FOLLOWS EASILY FROM #ONJECTURE OF SECTION
#ONJECTURE 4HERE EXISTS A UNIQUE UP TO A CONSTANT :X INVARIANT
FUNCTIONAL X 3yX #
%XAMPLE ,ET ' ',N AND TAKE y TO BE THE STANDARD REPRESENTATION
OF '? ',N 4HEN 'y -N IS THE SEMI GROUP OF N a N MATRICES !LSO
ONE HAS 3y ' 3-N THE SPACE OF 3CHWARTZ "RUHAT FUNCTIONS ON -N
)N THIS CASE 3yX IS ONE DIMENSIONAL FOR EVERY X AND THE FUNCTIONAL X IS
EQUAL TO THE EVALUATION OF A LOCALLY CONSTANT FUNCTION AT X
,OCAL , FUNCTIONS (ERE WE GIVE A DElNITION OF THE LOCAL , FACTOR

FOR EVERY ADMISSIBLE NON SINGULAR REPRESENTATION y
,ET x BE AN IRREDUCIBLE REPRESENTATION OF ' 7E DENOTE BY - x y
# ' THE SPACE OF ALL MATRIX COEnCIENTS OF x
#ONJECTURE &OR EVERY } 3y ' AND EVERY M - x THE
INTEGRAL
:
:} M S }G MG JzG JS JDGJ
'
IS ABSOLUTELY CONVERGENT FOR ES |
:} M S HAS A MEROMORPHIC CONTINUATION TO THE WHOLE OF # AND
DElNES A RATIONAL FUNCTION OF Q S
4HE SPACE OF ALL :} M S AS ABOVE IS A lNITELY GENERATED NON ZERO
FRACTIONAL IDEAL OF THE RING #;Q S Q _S = 7E SHALL DENOTE THIS IDEAL BY *x
7E NOW DElNE ,y S x TO BE THE UNIQUE GENERATOR OF *x OF THE FORM
0 Q _S _ WHERE 0 IS A POLYNOMIAL SUCH THAT 0
! "2!6%2-!. !.$ $ +!:($!. '!&!
4HE 0OISSON 3UMMATION &ORMULA

!S BEFORE LET + BE A GLOBAL lELD AND 0+ BE THE SET OF PLACES OF + 7E
ALSO DENOTE BY !+ THE ADELE RING OF +
,ET ' BE A SPLIT REDUCTIVE ALGEBRAIC GROUP OVER + AND y '?
',N # BE AS BEFORE &OR EVERY PLACE F 0+ WE CAN CONSIDER THE
3CHWARTZ SPACE 3yF 3y '&F WITH THE DISTINGUISHED FUNCTION #yF IN
IT 7E DElNE THE SPACE 3y '!+ AS THE RESTRICTED TENSOR PRODUCT OF THE
SPACES 3yF WITH RESPECT TO THE FUNCTIONS #yF
#HOOSE NOW A NON TRIVIAL CHARACTER !+ #a SUCH THAT J& IS
TRIVIAL 4HEN DElNES AN ADDITIVE CHARACTER F OF +F FOR EVERY F 4HE
&OURIER TRANSFORM
9
&y &yF
F0&
ACTS ON THE SPACE 3y '!+

%VERY } 3y '!+ GIVES RISE TO A FUNCTION ON '!+
#ONJECTURE A 4HERE EXISTS A '+ INVARIANT FUNCTIONAL
3y '!+ # SUCH THAT
} &y } FOR ANY } 3y '!+
1
,ET } }F 3y '!+ !SSUME THAT THERE EXISTS A PLACE F
0+ SUCH THAT }F (' 4HEN
8
} }G
G'+
B 4HE FUNCTIONAL IS SUPPORTED AT 'y + )N OTHER WORDS ASSUME

THAT WE ARE GIVEN } 3y '!+ SUCH THAT FOR EVERY G 'y + THERE
EXISTS A NEIGBOURHOOD 5G OF G SUCH THAT }X FOR EVERY X 5G < '+
4HEN }
7HEN ' ',N AND y IS THE STANDARD REPRESENTATION THE STATEMENT
OF #ONJECTURE IS THE STANDARD 0OISSON SUMMATION FORMULA )N THIS CASE
} IS A SMOOTH FUNCTION ON THE SPACE -N !+ OF ADELIC N a N MATRICES AND
ONE HAS
8
} }G
G-N +
7E DO NOT KNOW ANY EXPLICIT FORMULA FOR IN GENERAL

.OW LET x BE AN IRREDUCIBLE AUTOMORPHIC REPRESENTATION OF '!+ AND
6ISIONS IN -ATH l &5.#4)/.3 /& 2%02%3%.4!4)/.3 !.$ ,)&4).'
xF ITS COMPONENT AT F 03 /NE CAN CONSIDER THE GLOBAL , FUNCTION

9
,y x S ,y xF S
F03
5SING THE ARGUMENTS OF ;'O*= IS EASY TO SEE THAT THE VALIDITY OF THE ABOVE
CONJECTURES IMPLIES THE MEROMORPHIC CONTINUATION AND FUNCTIONAL EQUATION
OF THE CORRESPONDING AUTOMORPHIC , FUNCTIONS
<!LGEBRAIC )NTEGRALS OVER ,OCAL &IELDS

)N THE NEXT TWO SECTIONS WE ASSUME THAT THE LOCAL lELD & HAS CHARACTERISTIC
,ET 8 BE AN ALGEBRAIC VARIETY OVER & OF DIMENSION D 8 8& 2ECALL
THAT CF ;7E= ANY `8 jTOP DElNED OVER & DElNES A MEASURE JJ
ON 8 &OR A PROPER MORPHISM F 8 9 AND AS ABOVE WE DENOTE BY
F JJ THE CORRESPONDING PUSH FORWARD MEASURE ON 9
,OCALLY INTEGRABLE FUNCTIONS 7E SAY THAT A FUNCTION } 8 #
IS LOCALLY INTEGRABLE FOR EVERY POINT X 8 AND ANY TOP FORM `5 jD
DElNED IN A COMPACT NEIGHBOURHOOD 5 y 8 OF X THE INTEGRAL
:
J}Y JJY J
5
IS CONVERGENT
4HEOREM 6 6OLOGODSKY ,ET F 8 ! BE A PROPER MORPHISM
AND LET `8 jD 3UPPOSE THAT BOTH ARE DElNED OVER & 4HEN THERE
EXIST AN OPEN COMPACT SUBGROUP + OF /&a AND A + INVARIANT FUNCTION jX
SUCH THAT DISTRIBUTION F JJ ON & IS EQUAL TO jX DX FOR JXJ |
4HEOREM IS PROVED IN THE !PPENDIX
#OROLLARY ,ET & #a BE A NON TRIVIAL CHARACTER 4HEN THE
INTEGRAL
: : :
F JJ F JJ ` LIM F JJ
8 & . JXJ.
IS CONVERGENT
!LGEBRAIC GEOMETRIC DISTRIBUTIONS ,ET AS BEFORE 8 BE AN ALGE
BRAIC VARIETY OVER &
$EFINITION "Y AN ALGEBRO GEOMETRIC DISTRIBUTION ON 8 WE MEAN A
QUADRUPLE h 9 F P WHERE
9 IS A SMOOTH ALGEBRAIC VARIETY OVER &
! "2!6%2-!. !.$ $ +!:($!. '!&!
P 9 8 F 9 ! ARE MORPHISMS DElNED OVER &

`9 jTOP
9
"Y AN ISOMORPHISM BETWEEN TWO ALGEBRO GEOMETRIC DISTRIBUTIONS h
9 F P AND h 9 F P WE SHALL MEAN A BIRATIONAL ISOMORPHISM
BETWEEN 9 AND 9 PRESERVING ALL THE STRUCTURE
'IVEN AN ALGEBRO GEOMETRIC DISTRIBUTION h WE CAN TRY TO DElNE A USUAL
DISTRIBUTION h h ON 8 BY WRITING
_ ` _ `k
h P F JJ LIM P F JJ k FY9 JJF Y Js. G

.
#OROLLARY GUARANTEES THAT IN A NUMBER OF CASES MAKES SENSE &OR
EXAMPLE ONE CAN SHOW THAT THE FOLLOWING RESULT HOLDS
0ROPOSITION ,ET AN ALGEBRO GEOMETRIC DISTRIBUTION h AS ABOVE BE
GIVEN !SSUME THAT THERE EXISTS AN OPEN DOMINANT EMBEDDING 9 9 E
E E
SUCH THAT 9 IS SMOOTH AND SUCH THAT P F AND EXTEND TO 9 !SSUME
FURTHERMORE THAT THE MORPHISM P a F 9 E 8 a ! IS PROPER 4HEN h
GIVEN BY IS A WELL DElNED DISTRIBUTION ON 8
7E SAY THAT AN ALGEBRO GEOMETRIC DISTRIBUTION h IS lNITE IF IT SATISlES
THE CONDITIONS OF 0ROPOSITION !LSO WE CALL h WEAKLY lNITE IF IT IS
lNITE OVER THE GENERIC POINT OF 8 7E ALSO SAY THAT h IS ANALYTICALLY lNITE
IF IT IS WEAKLY lNITE AND FOR ANY lNITE EXTENSION %& THE CORRESPONDING
DISTRIBUTION h DElNED A PRIORI ON A DENSE OPEN SUBSET OF 8% IS LOCALLY
, ON THE WHOLE OF 8% )N THIS CASE WE CALL h THE MATERIALIZATION OF h
,EMMA ,ET 8 4 BE A SPLIT TORUS OVER & AND LET y BE AN ADMISSIBLE
REPRESENTATION OF 4? ,ET h y4 4y Fy Py y BE GIVEN BY THE FORMULAS
OF SECTION 4HEN h y4 IS ANALYTICALLY lNITE AND THE DISTRIBUTION hy4 IS
THE MATERIALIZATION OF h y4
2EDUCTION OF ALGEBRO GEOMETRIC DISTRIBUTIONS ,ET h
9F P BE AN ALGEBRO GEOMETRIC DISTRIBUTION AND LET 6 BE A VECTOR GROUP
DElNED OVER & 7E SAY THAT 6 ACTS ON h IF WE ARE GIVEN A FREE ACTION OF 6
ON 9 SUCH THAT
P AND ARE 6 INVARIANT
,ET Q 9 : 96 BE THE QUOTIENT MAP 4HEN FOR ANY Z : THE
RESTRICTION OF F TO Q_ Z IS AN AnNE FUNCTION
,ET 6b DENOTE THE DUAL VECTOR SPACE TO 6 7E DENOTE BY S : 6b
THE MAP SENDING EVERY Z : TO THE LINEAR PART OF F JQ_ Z 3ET
g k h
9 Z : k SZ
6ISIONS IN -ATH l &5.#4)/.3 /& 2%02%3%.4!4)/.3 !.$ ,)&4).'
7E SAY THAT AN ACTION OF 6 ON h IS NON DEGENERATE IF FOR ANY Y 9 THE

DIkERENTIAL DS 4Y 9 6b IS ONTO )N THIS CASE THE VARIETY 9 IS SMOOTH
AND FOR ANY Y 9 WE GET THE NATURAL ISOMORPHISM 4Y :4Y 9 6b (ENCE
FOR ANY Y 9 AND ANY Y Q_ Y WE HAVE THE NATURAL ISOMORPHISMS
cTOP 4Y 9 cTOP 4Y : icTOP 6 cTOP 4Y 9 icTOP 4Y :4Y 9 icTOP 6
cTOP 4Y 9 i cTOP 6b i cTOP 6 cTOP 4Y 9
4HEREFORE FOR ANY Y 9 Y Q_ Y THE RESTRICTION OF TO 4Y 9 DElNES
AN ELEMENT Y c4Y 9 b 3INCE IS 6 INVARIANT IT FOLLOWS THAT Y DOES
NOT DEPEND ON THE CHOICE OF Y Q_ Y )T IS EASY TO SEE THAT THERE EXISTS
A TOP DEGREE FORM ON 9 SUCH THAT FOR ANY Y 9 ONE HAS Y Y
"Y THE DElNITION THE RESTRICTION OF F TO Q_ 9 IS 6 INVARIANT 3O
F JQ_ 9 F m Q FOR SOME FUNCTION F ON 9 !LSO SINCE P IS 6 INVARIANT WE
HAVE PJQ_ 9 P m Q FOR SOME MORPHISM P 9 8
7E CALL THE ALGEBRO GEOMETRIC DISTRIBUTION h 9 P F THE REDUCTION
OF h
,ET US SEE THE EkECT OF THIS REDUCTION PROCEDURE ON MATERIALIZATIONS
!SSUME THAT h IS ANALYTICALLY lNITE AND DENOTE BY h ITS MATERIALIZATION
&OR SIMPLICITY WE SHALL ASSUME THAT & IS NON !RCHIMEDEAN ,ET 6N BE A
SEQUENCE OF OPEN COMPACT SUBGROUPS OF 6 SUCH THAT

;
6 6N
N
,ET ALSO 9N BE AN INCREASING COVERING OF 9 BY 6N INVARIANT OPEN COMPACT
SUBSETS #ONSIDER THE SEQUENCE hN N OF DISTRIBUTIONS ON 8
WHERE
:
hN } F Pb } JJ
9N
,EMMA 4HE SEQUENCE hN WEAKLY CONVERGES TO h

7HEN DIM 6 THIS IS PROVED IN ;'+=X 4HE GENERAL CASE IS TREATED
SIMILARLY
7 EQUIVARIANT DISTRIBUTIONS .OW LET ' BE A REDUCTIVE ALGE
BRAIC GROUP OVER & AND LET 4 BE ITS #ARTAN GROUP 4HE 7EYL GROUP
7 ACTS NATURALLY ON 4 AND WE HAVE THE NATURAL & RATIONAL MORPHISM
S ' 47 ,ET 'R DENOTE THE SET OF REGULAR ELEMENTS IN ' 4HEN
THE RESTRICTION OF S TO 'R IS A SMOOTH MAP )N THE REST OF THIS SECTION S WILL
ALWAYS DENOTE THE RESTRICTION OF S TO 'R
! "2!6%2-!. !.$ $ +!:($!. '!&!
.OW LET h 4 94 F4 P4 4 BE AN ALGEBRO GEOMETRIC DISTRIBUTION

ON 4 7E ASSUME THAT P4 IS A DOMINANT MAP "Y A 7 EQUIVARIANT STRUC
TURE ON h WE SHALL MEAN A BIRATIONAL ACTION OF 7 ON 4 SUCH THAT
F4 IS INVARIANT UNDER THIS ACTION
P4 INTERTWINES THE ACTION OF 7 ON 94 AND ON 4
/NE HAS Wb 4 _ LW WHERE L 7 : IS THE LENGTH FUNCTION
3INCE 7 IS A lNITE GROUP WE CAN ALWAYS lND AN OPEN DENSE SUBSET OF 9
ON WHICH THE ACTION 7 IS BIREGULAR 7E CAN ALSO ASSUME THAT THIS SUBSET IS
CHOSEN IN SUCH A WAY THAT IT IS CONTAINED IN THE SET OF REGULAR SEMI SIMPLE
ELEMENTS .OTE THAT BIRATIONAL MODIlCATIONS OF 94 SUCH AS PASSING TO
AN OPEN SUBSET DO NOT CHANGE THE DISTRIBUTION h GIVEN BY 7E ALSO
ASSUME IN THE SEQUEL THAT P4 IS A SMOOTH MAP
!SSUME NOW THAT WE ARE GIVEN A 7 EQUIVARIANT ALGEBRO GEOMETRIC DISTRI
BUTION h4 AS ABOVE 7E ASSUME ALSO THAT 4 z ` Da T WHERE z ' 'M
IS A CHARACTER OF ' AND Da T A TRANSLATION INVARIANT TOP DEGREE DIkEREN
TIAL FORM ON 4 4HEN WE MAY CONSTRUCT AN ALGEBRO GEOMETRIC DISTRIBUTION
h 9' F' P' ' OVER 'R IN THE FOLLOWING WAY 7E MAY ASSUME THAT
7 ACTS BIREGULARLY ON 94 4HEN WE DElNE
9' 'R a 94 7
47
7E LET P' BE THE PROJECTION ON THE lRST MULTIPLE #ONDITION ABOVE IMPLIES
THAT F4 DESCENDS TO A FUNCTION ON 94 7 AND HENCE DElNES A FUNCTION F'
ON 9'
4HE CHOICE OF Da T GIVES RISE TO A CHOICE OF A LEFT INVARIANT TOP DEGREE
DIkERENTIAL FORM DG ON ' 7E SET ' z L DG WHERE L AS BEFORE DENOTES
THE SPLIT RANK OF '
#ONJECTURE ,ET ' BE A SPLIT REDUCTIVE GROUP OVER & AND LET
y '? ',N # BE AN ADMISSIBLE REPRESENTATION OF '? ,ET h y4 BE
THE ALGEBRO GEOMETRIC DISTRIBUTION AS IN ,EMMA 4HEN THERE EXISTS A
7 EQUIVARIANT STRUCTURE ON hy4 SUCH THAT THE CORRESPONDING DISTRIBUTION
h y' IS ANALYTICALLY lNITE AND THE UNDERLYING DISTRIBUTION ON ' IS EQUAL TO
hy' INTRODUCED IN SECTION
,ET 0 y ' BE A PARABOLIC SUBGROUP AND LET , BE THE CORRESPONDING
,EVI FACTOR 4HE REPRESENTATION y DElNES CANONICALLY A REPRESENTATION y,
OF ,? WITH THE SAME RESTRICTION TO 4? .OTE THAT THE 7EYL GROUP 7, IS
NATURALLY A SUBGROUP IN 7
4HE RESTRICTION OF THE 7 ACTION ON h y4 TO 7, IS EQUAL TO THE CORRE
SPONDING 7, ACTION ON hy, 4
6ISIONS IN -ATH l &5.#4)/.3 /& 2%02%3%.4!4)/.3 !.$ ,)&4).'
2EMARK 7E SHALL SEE SOME EXAMPLES OF THE ABOVE 7 EQUIVARIANT STRUC

TURE IN SOME EXAMPLES 3URPRISINGLY THE CORRESPONDING 7 ACTION ON 4y
TURNS OUT TO BE RATHER COMPLICATED AND IT IS RELATED TO THE THEORY OF GEO
METRIC CRYSTALS CF ;"E+=
4HE #ASE OF 'M N AND ITS !PPLICATIONS

,ET ' 4 AND y BE AS ABOVE )N SECTION WE HAVE CONSTRUCTED A lNITE
ALGEBRO GEOMETRIC DISTRIBUTION h y4 4y Fy Py y ON 4 )T IS EXPLAINED
IN SECTION THAT IN ORDER TO GET FROM IT A lNITE ALGEBRO GEOMETRIC DISTRI
BUTION ON hy' ONE HAS TO DElNE A BIRATIONAL ACTION OF THE 7EYL GROUP 7
ON 4y SATISFYING CERTAIN CONDITIONS 7E DO NOT KNOW HOW TO CONSTRUCT THIS
ACTION IN GENERAL )N THIS SECTION WE GIVE AN EXPLICIT CONSTRUCTION OF THIS
ACTION FOR THE GROUP ' 'N M CF SECTION
4HE CASE M 7E NOW WANT TO DElNE THE ACTION OF 7 ON 4y
IN THE CASE WHEN M 4HE 7EYL GROUP 7 OF ' ' N IS ISOMORPHIC
TO : a 3N )N THIS SUBSECTION WE ARE GOING TO GIVE AN EXPLICIT FORMULA FOR
THE ACTION OF THE lRST FACTOR BY USING PART OF #ONJECTURE
,ET 4 BE AS BEFORE A MAXIMAL TORUS OF ' )T CAN BE IDENTIlED WITH THE
VARIETY OF ALL COLLECTIONS A A B BN WHERE AI BJ 'M AND A A
B BN
4HE TORUS 4y CONSISTS OF ALL MATRICES T TIJ WHERE I AND
J N 4HE MAP Py IS GIVEN BY
Py TIJ T T TN T T TN T T T T TN TN
0
4HE FUNCTION Fy SENDS THE 1MATRIX TIJ TO IJ TIJ AND THE DIkERENTIAL FORM y
IS UP TO A SIGN EQUAL TO DTIJ
4O DElNE : ACTION ON h y4 WE NEED TO DElNE AN INVOLUTION { ON 4y
THAT PRESERVES Fy SENDS y TO _y AND HAS THE FOLLOWING TWO PROPERTIES
&OR EVERY T 4y AND EVERY J N ONE HAS TJ TJ { T J { T J
&OR EVERY T 4y T TN { T { T N AND T TN { T { T N
)T IS EASY TO SEE THAT IN THE CASE WHEN N THERE EXIST MANY WAYS TO
DElNE AN INVOLUTION ON 4y WHICH SATISFY THE COMPATIBILITY CONDITIONS
AND 7E WILL USE THE CONJECTURAL PROPERTIES OF THE DISTRIBUTION hy'N
TO OBTAIN THE FORMULA FOR THE INVOLUTION {
,ET 4N 'NM AND LET . 4N 'M BE GIVEN BY . T TN
T TN $ElNE
g h
'N G T ', a 4N J DETG . T
! "2!6%2-!. !.$ $ +!:($!. '!&!
4HE GROUP 'N CAN BE NATURALLY REGARDED AS A ,EVI SUBGROUP OF ' N

(ENCE THE DUAL GROUP '? ?
N IS A ,EVI SUBGROUP OF ' N AND WE CAN RESTRICT
THE REPRESENTATION y TO IT 4HIS RESTRICTION CAN BE EXPLICITLY DESCRIBED AS
FOLLOWS 4HE GROUP '? a N
N IS ISOMORPHIC TO THE QUOTIENT OF ', # a# BY
THE SUBGROUP CONSISTING OF ELEMENTS OF THE FORM X ` )D X_ X_ X_
&OR EVERY I N WE CAN DElNE A TWO DIMENSIONAL REPRESENTATION yI OF
'?N SENDING G Z ZN TO GZI 4HEN IT IS EASY TO SEE THAT
N
-
y yI
I
4HE #ARTAN GROUP OF 'N IS NATURALLY ISOMORPHIC TO THE #ARTAN GROUP
4 OF ' N CONSIDERED IN SECTION 4HE 7EYL GROUP OF 'N IS EQUAL TO
THE lRST FACTOR OF THE 7EYL GROUP 7 OF ' N
!S FOLLOWS FROM AND THE DISTRIBUTION hy'N IS EQUAL TO THE
CONVOLUTION
hy'N G B BN hB b ` ` ` b hBN JDB DBN J
WHERE DISTRIBUTIONS hB B & a ON ',M & ARE DElNED IN SECTION )N
OTHER WORDS hy'N CAN BE THOUGHT OF AS A MATERIALIZATION OF THE ALGEBRO
GEOMETRIC DISTRIBUTION h E 9 E E E
F P E WHERE
E
n 9 ', N
n EF G GN TRG TRGN

n PE G GN G GN DETG DETGN
n E DETG GN DG DGN
WHERE DG DENOTES A TRANSLATION INVARIANT TOP DEGREE DIkERENTIAL FORM
ON ',
-ORE PRECISELY FOR EVERY . DElNE AN OPEN SUBSET 9E . OF 9E BY
g k h
9E . _ G GN ', N k JGI jk J s .
E. DENOTE THE RESTRICTION OF THE DISTRIBUTION FE JE
,ET u J TO 9E .
,EMMA
hy'N LIM P E
u.
.
&IX A MAXIMAL UNIPOTENT SUBGROUP 5 IN ', 4HEN 5 'A &OR
EVERY G ', THE FUNCTION U TRGU _ TRG IS LINEAR IN U 7E SET
TRGU _ TRG G_ U THUS G G_ IS A FUNCTION ON ',
,ET 6 5N_ E
$ElNE AN ACTION OF 6 ON 9 BY SETTING
U UN_ G GN G U_ _
U G U UN_ GN
6ISIONS IN -ATH l &5.#4)/.3 /& 2%02%3%.4!4)/.3 !.$ ,)&4).'
E AND
4HIS ACTION PRESERVES P E WE HAVE
E AND FOR ANY G GN 9
N_
8 _ `
E
F UG _ E
F G UI GI _ _ GI _
I
4HUS WE OBTAIN AN ACTION OF 6 ON h E 9E E E

F P E ,ET h5
E
9 P F BE THE 5 REDUCTION OF h CF SECTION !S FOLLOWS FROM
,EMMA THE MATERIALIZATION h v 5 IS EQUAL TO hy
,ET " BE THE VARIETY OF "OREL SUBGROUPS OF ', &OR ANY G ',
WE DENOTE BY "G y " THE SUBVARIETY OF "OREL SUBGROUPS CONTAINING G
,EMMA &OR ANY REGULAR G ', THE CENTRALIZER OF G IN 0',
ACTS SIMPLY TRANSITIVELY ON THE VARIETY " _ " G
#OROLLARY &OR ANY TWO UNIPOTENT SUBGROUPS 5 5_ y ', WE
HAVE A CANONICAL ISOMORPHISM h5 h5_ 4HEREFORE WE CAN WRITE h
INSTEAD OF h 5
,ET
g k h
E N " G T " a 'N k G "
'
7E HAVE THE NATURAL MAPS R ' E N 'N SENDING " G T TO G AND M
'E N 4 SENDING " G T TO G MOD 5" WHERE 5" DENOTES THE UNIPOTENT
RADICAL OF " 4HEN R IS A RAMIlED DOUBLE COVERING AND FOR ANY G T 'N
ONE HAS R_ G T "G 7E DENOTE BY h E 9
E PE E
F
E THE LIFT OF h TO 'E N
"Y THE DElNITION 9 E 9 a' ' E N
N
3INCE THE GENERIC lBER OF M CARRIES A CANONICAL TOP FORM SEE SECTION
WE CAN DElNE AN ALGEBRO GEOMETRIC DISTRIBUTION Mb h hy4 ON 'E N
7E WRITE M hb
hy4 94 P4 F4 4 WHERE 94 9y4 a4 ' E N 7E WILL
E
CONSTRUCT NOW AN ISOMORPHISM u h M h b hy4
&OR ANY " " SUCH THAT 5 y " THE MULTIPLICATION MAP GIVES RISE
TO BIRATIONAL ISOMORPHISMS " a 5 ', AND 5 a " ',
4HEREFORE THE SUBSET 9 E "N_ a ', y 9 E IS A SECTION OF THE ACTION
E E
OF 6 ON 9 &IX ANY ELEMENT G T " 'N SUCH THAT G IS REGULAR AND
SEMISIMPLE 3INCE BY DElNITION G " WE SEE THAT 9 E 4P E _ G T " y "N
3O WE CAN DESCRIBE THE VARIETY P E _ G T " AS THE PREIMAGE S" _ WHERE
S" IS THE RESTRICTION OF THE MAP S 9 E 6b TO "N 4 P_ G T WHERE S IS AS
IN SECTION )N OTHER WORDS WE CAN IDENTIFY THE VARIETY 9 E WITH S" _
,ET 5 y " BE THE UNIPOTENT RADICAL OF " AND LET 4G y " BE THE CEN
TRALIZER OF G 4HEN " 4G 5 AND 4G IS CANONICALLY ISOMORPHIC TO 'M
4HEREFORE 4NG IS CANONICALLY ISOMORPHIC TO 4y
! "2!6%2-!. !.$ $ +!:($!. '!&!
,EMMA A 4HE NATURAL PROJECTION " 4G DElNES AN EMBEDDING

u S_ _
" 4y AND THE IMAGE OF I IS EQUAL TO Py4 G T
)N OTHER WORDS WE HAVE CONSTRUCTED AN ISOMORPHISM u 9 E 9
4
B u DElNES AN ISOMORPHISM OF ALGEBRO GEOMETRIC DISTRIBUTIONS
4HE NATURAL INVOLUTION q OF 'E N OVER 'N INDUCES AN INVOLUTION qE OF h E

AND WE CAN DElNE AN INVOLUTION q OF M hb hy4 BY q um qEmu )T IS EASY
_

TO SEE THAT THE INVOLUTION q IS ', INVARIANT AND THEREFORE IT IS INDUCED
BY AN INVOLUTION { ON h y4
,ET US GIVE AN EXPLICIT FORMULA FOR { &OR EVERY K N WE DElNE
THE FUNCTION aK ON 4y BY
aK T T TK_ T TK_ TK ` ` ` T TK
BY DElNITION a T
7E ALSO DElNE A RATIONAL FUNCTION p ON 4y SETTING
T TN _ T TN
pT
aN T
,EMMA 4HE INVOLUTION { SATISlES
{ T { T { T K T TK aK T pT
)T IS CLEAR THAT { IS UNIQUELY DETERMINED BY AND BY THE CONDITIONS
AND ABOVE
!SSUME FOR EXAMPLE THAT N 4HEN ONE CAN COMPUTE THE ABOVE
ACTION EXPLICITLY .AMELY IN THIS CASE WE HAVE
g k h
4 A B A B 'M k A B A B
,ET W 7 BE THE INVOLUTION WHICH INTERCHANGES A AND B ,ET { {W
4HEN AN EXPLICIT CALCULATION SHOWS THAT
q r q T T T r
T T T T T T TT T
{
T

T
T T T TT
T
T TT
T
-OREOVER IT IS EASY TO SEE THAT IN THIS CASE { GIVEN BY IS THE UNIQUE
BIRATIONAL INVOLUTION OF 4y WHICH SATISlES OUR REQUIREMENTS
4HE CASE OF 'N M ,ET US NOW CONSIDER THE GROUP 'N M FOR
ARBITRARY N AND M )N THIS CASE ONE CAN IDENTIFY 4y WITH THE VARIETY TIJ
I M J N OF M a N MATRICES WITH NON ZERO ENTRIES 7E
WISH TO DElNE A BIRATIONAL ACTION OF THE 7EYL GROUP 7 3M a 3N ON 4y
&OR EVERY j M _ WE DElNE A BIRATIONAL INVOLUTION {j ON 4y IN
THE FOLLOWING WAY
6ISIONS IN -ATH l &5.#4)/.3 /& 2%02%3%.4!4)/.3 !.$ ,)&4).'
!LL ROWS OF {j T TIJ EXCEPT FOR THE j AND j ST ARE EQUAL TO THE
CORRESPONDING ROWS OF T
4HE j AND j ST ROWS OF {j T ARE OBTAINED FROM THOSE OF T BY
MEANS OF
3IMILARLY FOR EVERY k N _ WE DElNE {k T {k T WHERE T IS
THE TRANSPOSED MATRIX TO T 4HE FOLLOWING LEMMA IS STRAIGHTFORWARD
,EMMA 4HE INVOLUTIONS {j {k COMMUTE WITH Py PRESERVE Fy AND MAP
y TO _y
4HE FOLLOWING THEOREM IS PROVEN IN ;"E+ X =
4HEOREM 4HE INVOLUTIONS {j AND {k DElNE A BIRATIONAL ACTION OF
THE GROUP 3M a 3N ON 4y )N PARTICULAR {j COMMUTES WITH {k FOR ANY j
AND k
$ESPITE THE FACT THE FORMULAS ARE QUITE EXPLICIT WE DID NOT MANAGE
TO PROVE 4HEOREM DIRECTLY 4HE PROOF GIVEN IN ;"E+= USES THE MACHINERY
OF GEOMETRIC CRYSTALS
l FUNCTIONS ,ET US STILL ASSUME THAT ' 'N M 2ECALL THAT
WE HAVE THE NATURAL CHARACTER z ' 'M WHICH SENDS THE PAIR ! "
OF MATRICES TO DET! DET" 5SING 4HEOREM WE CAN DElNE AN
ALGEBRO GEOMETRIC DISTRIBUTION hy' OVER THE GENERIC POINT OF ' 7E WOULD
LIKE TO CHECK THAT IT GIVES THE <CORRECT ANSWER
,ET xM xN BE GENERIC REPRESENTATIONS OF RESPECTIVELY ',M & AND
',N & 4HEN IN ;*03= ( *ACQUET ) 0IATETSKII 3HAPIRO AND * 3HALIKA
DElNE A RATIONAL FUNCTION `xM xN S OF ONE COMPLEX VARIABLE S WHICH DE
PENDS ON THE CHOICE OF A CHARACTER 4HE FOLLOWING RESULT CAN BE DEDUCED
FROM ;"E+=
4HEOREM 4HE MATERIALIZATION hy' OF h y' DElNES A z COMPACT
DISTRIBUTION ON THE WHOLE OF ' 7E DENOTE BY ly THE CORRESPONDING RATIONAL
FUNCTION ON THE SET )RR'
!SSUME THAT N t M ,ET xM xN BE AS ABOVE AND LET x BE AN
IRREDUCIBLE REPRESENTATION OF ' SUCH THAT (OM' x xM i xN 4HEN
`xM xN S SIGNxN ly xJ ` JS
WHERE SIGNx IS THE VALUE OF THE CENTRAL CHARACTER OF xN AT THE MATRIX
_)DN ',N &
,IFTING FROM NON SPLIT TORI ,ET US RECALL THE NOTATION OF SEC
TION LET %& BE A SEPARABLE EXTENSION OF & OF DEGREE N ,ET j%
'AL& & 3N BE THE CORRESPONDING HOMOMORPHISM 7E DENOTE BY
! "2!6%2-!. !.$ $ +!:($!. '!&!
. 4% 'M THE MORPHISM OF ALGEBRAIC GROUPS COMING FROM THE NORM

MAP . % a & a
,ET 4% 2ES%& 'M% WHERE 2ES%& DENOTES THE FUNCTOR OF RESTRICTION
OF SCALARS &OR EVERY M WE SET
g k h
'%M T G 4% a ',M k .T DETG
!S IS EXPLAINED IN SECTION IN ORDER TO DElNE THE LIFTING L% q
)RR',N & OF A CHARACTER q % a #a IT IS ENOUGH TO DElNE THE
l FUNCTION lL% q xM AS A FUNCTION ON )RR',M & -OREOVER IT IS
ENOUGH TO KNOW THIS FUNCTION ONLY UP TO A CONSTANT
7E ARE NOW GOING TO GIVE A CONJECTURAL DElNITION OF lL% q xM BY
CONSTRUCTING A WEAKLY lNITE ALGEBRAIC GEOMETRIC DISTRIBUTION h %M ON '%M
IN FACT WE WILL DElNE IT ONLY UP TO A CONSTANT MULTIPLE 7E CONJECTURE THAT
THIS DISTRIBUTION IS ANALYTICALLY lNITE AND THAT THE CORRESPONDING FUNCTION
ON )RR% a a )RR',M & IS EQUAL TO l%M
#HOOSE D & a SUCH THAT ITS IMAGE IN & a & P a IS EQUAL TO THE DIS
CRIMINANT OF % ,ET US ALSO CHOOSE ITS SQUARE ROOT D IN &

,ET
g k h
'%M T TN G 'NM a ',M k T TN DETG
5SING SECTION WE DElNE AN ALGEBRO GEOMETRIC DISTRIBUTION h
9 P F ON 'NM a ',M -OREOVER WE HAVE THE NATURAL 3N ACTION
ON 9 WHICH IS COMPATIBLE WITH P AND LEAVES F RESP INVARIANT RESP
SKEW INVARIANT
$ElNE A NEW 'AL& & ACTION ON 9 BY
G NEW Y j% G OLD Y

4HIS ACTION DElNES A NEW & RATIONAL STRUCTURE ON 9 7E DENOTE BY 9 THE
CORRESPONDING & VARIETY )T IS CLEAR THAT THE MAP P GIVES RISE TO AN &
RATIONAL
P MORPHISM P 9 '%M !LSO THE FUNCTION F AND THE DIkERENTIAL
FORM D GIVE RISE TO A & RATIONAL FUNCTION F AND A DIkERENTIAL FORM
ON 9 4HUS WE SET h %M 9 P F TO BE THE REQUIRED ALGEBRO GEOMETRIC
DISTRIBUTION 7E DENOTE BY h%M THE P MATERIALIZATION OF h %M
#LEARLY FOR DIkERENT CHOICES OF D THE RESULTING MATERIALIZATIONS h%M
WILL DIkER ONLY BY A MULTIPLICATION BY C #b (ENCE THE ABOVE CONSTRUCTION
SUnCES IN ORDER TO DETERMINE THE LIFTING UNIQUELY
3INCE THE LOCAL ,ANGLANDS CONJECTURE IS KNOWN SEE ;(= ;(4= AND ;,A23=
WE CAN ASK WHETHER OUR DElNITION OF LIFTING COINCIDES WITH ONE WHICH IS
IMPLIED BY ;(= ;(4= AND ;,A23= -ORE PRECISELY LET ~ BE A CHARACTER
OF THE GROUP 4 4& 4HE LOCAL CLASS lELD THEORY ASSOCIATES TO ~ A
6ISIONS IN -ATH l &5.#4)/.3 /& 2%02%3%.4!4)/.3 !.$ ,)&4).'
v 'AL& &
HOMOMORPHISM q~ 'AL%% #a ,ET b~ )ND'AL%%
v q~ 4HEN
b~ IS AN N DIMENSIONAL REPRESENTATION OF THE GROUP 'AL& & 3INCE THE
LOCAL ,ANGLANDS CONJECTURE FOR ' ',N IS KNOWN ONE ASSOCIATES WITH
b~ AN IRREDUCIBLE REPRESENTATION x~ )RR',N & 4HEREFORE WE CAN
CONSIDER A FUNCTION `xM x~ S ON THE SET )RR ',M & /NE CAN ASK
WHETHER THERE EXISTS C #a SUCH THAT FOR ANY xM )RR ',M & WE HAVE
lxM x~ C~ i xM h%M
4HE #ASE OF &INITE &IELDS

)N THIS SECTION WE SHALL A GIVE AN EXPLICIT CONJECTURAL CONSTRUCTION OF hy'
FOR ANY ' AND FOR ALMOST ANY y IN THE CASE WHEN THE lELD & IS lNITE )T
TURNS OUT THAT IN THIS CASE THE RELEVANT TOOL FROM ALGEBRAIC GEOMETRY WHICH
ALLOWS US TO GO FROM THE CASE OF A TORUS TO THE CASE OF AN ARBITRARY GROUP
IS NOT THE LANGUAGE OF ALGEBRO GEOMETRIC DISTRIBUTIONS BUT THAT OF @ ADIC
PERVERSE SHEAVES 4HIS TOOL ALLOWS US TO AVOID CONSTRUCTING AN ACTION OF 7
ON 4y
.OTATION )N THIS SECTION WE lX A lNITE lELD & WITH Q ELEMENTS
A PRIME NUMBER L DIkERENT FROM THE CHARACTERISTIC OF & AND A NON TRIVIAL
a
CHARACTER & 1L !LL REPRESENTATIONS DISCUSSED IN THE NOTE ARE OVER
THE lELD 1L 7E DENOTE BY , THE !RTIN 3CHREIER SHEAF ON !&
,ET ' BE A REDUCTIVE ALGEBRAIC GROUP OVER & AND LET 4 BE ITS ABSTRACT
#ARTAN GROUP 4HE TORUS 4 COMES EQUIPPED WITH A CANONICAL & RATIONAL
STRUCTURE )N THIS SECTION WE ASSUME FOR SIMPLICITY THAT ' IS SPLIT 4HEN
THIS & RATIONAL STRUCTURE IS SPLIT TOO 7E WILL DENOTE BY & R 4 4 THE
CORRESPONDING &ROBENIUS MORPHISM ,ET 7 DENOTE THE 7EYL GROUP OF '
&OR EVERY W 7 WE SET & R W W m & R 4HE MORPHISM & RW INDUCES A NEW
& RATIONAL STRUCTURE ON 4 AND WE WILL DENOTE THE CORRESPONDING ALGEBRAIC
TORUS OVER & BY 4W )T IS WELL KNOWN THAT THERE EXISTS AN EMBEDDING
4W ' AND THAT IN THIS WAY WE GET A BIJECTION BETWEEN CONJUGACY CLASSES
OF ELEMENTS IN 7 AND CONJUGACY CLASSES OF & RATIONAL MAXIMAL TORI IN '
a
&OR A CHARACTER q 4W 1L WE DENOTE BY 2qW THE CORRESPONDING
$ELIGNE ,USZTIG REPRESENTATION
&OR TWO @ ADIC COMPLEXES ! AND " ON ' WE CAN DElNE THE CONVOLUTION
COMPLEX ! " AND BY SETTING
! " M ! b " ;DIM '=
WHERE M ' a ' ' IS THE MULTIPLICATION MAP
! "2!6%2-!. !.$ $ +!:($!. '!&!
,ET 8 BE AN ALGEBRAIC VARIETY OVER & AND LET & BE A COMPLEX OF @ ADIC
SHEAVES ON IT "Y A 7EIL STRUCTURE ON & WE SHALL MEAN AN ISOMORPHISM
w & R b & & WHERE & R DENOTES THE GEOMETRIC &ROBENIUS MORPHISM ON 8
)N THIS CASE WE CAN DElNE A FUNCTION 4R& ON 8 8& BY SETTING
8 _ `
4R& X _ I TR wX ( I &X ( I &X
I
HERE ( I &X
DENOTES THE I TH COHOMOLOGY OF THE lBER OF & AT X
,ET US CHOOSE A SQUARE ROOT Q OF Q 4HEN FOR ANY 7EIL SHEAF & ON 8
AND ANY HALF INTEGER N WE CAN CONSIDER THE 4ATE TWIST &N OF & "Y THE
DElNITION ONE HAS
4R&N 4R& Q_N
l FUNCTIONS FOR ',N
4HE MAIN RESULT ,ET x 6 BE AN IRREDUCIBLE REPRESENTATION
a
OF ' ',N & #HOOSE A NON TRIVIAL ADDITIVE CHARACTER & 1L AS
ABOVE AND CONSIDER THE OPERATOR
8
TRG xG _ N Q _N %ND' 6
G'
3INCE x IS IRREDUCIBLE THIS OPERATOR TAKES THE FORM l' x ` )D6 WHERE
l' x 1L WE WILL OMIT THE SUBSCRIPTS ' AND WHEN IT DOES NOT LEAD TO
CONFUSION 4HE NUMBER lx l' x IS CALLED THE GAMMA FUNCTION OF THE
REPRESENTATION x 4HE PURPOSE OF THIS SUBSECTION IS TO COMPUTE EXPLICITLY
THE GAMMA FUNCTIONS OF ALL IRREDUCIBLE REPRESENTATIONS OF '
,ET 7 3N DENOTE THE 7EYL GROUP OF '
a
&IX W 7 &OR A CHARACTER q 4W 1L WE SET
8
lW q _ NLW Q _N TRT qT 1L
T4W
%XAMPLE !SSUME THAT W 3N IS A CYCLE OF LENGTH N 4HEN 4W % a

WHERE % IS THE UNIQUE UP TO ISOMORPHISM EXTENSION OF & OF DEGREE N
)N THIS CASE lW q l% q FOR ANY CHARACTER q OF % a WHERE BY l% q WE
DENOTE THE l FUNCTION DElNED AS IN FOR THE GROUP ', % % a
4HEOREM !SSUME THAT AN IRREDUCIBLE REPRESENTATION x 6 APPEARS
IN 2qW FOR SOME W AND q AS ABOVE 4HEN
lx _ LW lW q
WHERE 2 DENOTES THE SET OF POSITIVE ROOTS OF ' )N PARTICULAR lx lx
IF x AND x APPEAR IN THE SAME VIRTUAL REPRESENTATION 2qW
6ISIONS IN -ATH l &5.#4)/.3 /& 2%02%3%.4!4)/.3 !.$ ,)&4).'
4HE REST OF THIS SUBSECTION IS OCCUPIED WITH THE PROOF OF 4HEOREM
#HARACTER SHEAVES ,ET ' BE AN ARBITRARY REDUCTIVE ALGE
BRAIC GROUP OVER & ,ET US RECALL ,USZTIGS DElNITION OF SOME OF THE
E DENOTE THE VARIETY OF ALL PAIRS " G WHERE
CHARACTER SHEAVES ,ET '
n " IS A "OREL SUBGROUP OF '
n G"
/NE HAS NATURAL MAPS j ' E 4 AND x ' E ' DElNED AS FOLLOWS
&IRST OF ALL WE SET x" G G .OW IN ORDER TO DElNE j LET US RECALL
THAT FOR ANY "OREL SUBGROUP " OF ' ONE HAS CANONICAL IDENTIlCATION u"
"5" F 4 WHERE 5" DENOTES THE UNIPOTENT RADICAL OF " IN FACT THIS IS
HOW THE ABSTRACT #ARTAN GROUP 4 IS DElNED .OW WE SET j" G u" G
,ET , BE A TAME LOCAL SYSTEM ON 4 7E DElNE +, x jb , ;DIM '=
/NE KNOWS CF ;,U= ;,A= THAT THE SHEAF +, IS PERVERSE
!SSUME NOW THAT FOR SOME W 7 THERE EXISTS AN ISOMORPHISM ,
b
& RW , AND LET US lX IT )T WAS OBSERVED BY ' ,USZTIG IN ;,U= THAT lX
ING SUCH AN ISOMORPHISM ENDOWS +, CANONICALLY WITH A 7EIL STRUCTURE
,USZTIGS DElNITION OF THIS 7EIL STRUCTURE WAS AS FOLLOWS
,ET J 'RS ' DENOTE THE OPEN EMBEDDING OF THE VARIETY OF REGULAR
SEMISIMPLE ELEMENTS IN ' INTO '
,EMMA
+, Jb +, J'RS
(ERE Jb DENOTES THE 'ORESKY -AC0HERSON INTERMEDIATE EXTENSION CF
;""$=
4HE LEMMA FOLLOWS FROM THE FACT THAT THE MAP x IS SMALL IN THE SENSE
OF 'ORESKY AND -C0HERSON
4HE LEMMA SHOWS THAT IT IS ENOUGH TO CONSTRUCT THE 7EIL STRUCTURE
ONLY ON THE RESTRICTION OF +, ON 'RS 4HE LATTER NOW HAS A PARTICULARLY
SIMPLE FORM .AMELY LET ' E RS DENOTE THE PREIMAGE OF 'RS UNDER x AND
LET xRS DENOTE THE RESTRICTION OF x TO ' E RS 4HEN IT IS EASY TO SEE THAT
E RS 'RS IS AN UNRAMIlED 'ALOIS COVERING WITH 'ALOIS GROUP 7 )N
xRS '
PARTICULAR 7 ACTS ON 'E RS AND THIS ACTION IS COMPATIBLE WITH THE ACTION OF
7 ON 4 IN THE SENSE THAT THE RESTRICTION OF j ON ' E RS IS 7 EQUIVARIANT
b
.OW AN ISOMORPHISM , & RW , GIVES RISE TO AN ISOMORPHISM
jb , W m & R b jb ,
E
HERE BOTH W AND & R ARE CONSIDERED ON THE VARIETY ' 3INCE xRS IS A
'ALOIS COVERING WITH 'ALOIS GROUP 7 IT FOLLOWS THAT ONE HAS CANONICAL
! "2!6%2-!. !.$ $ +!:($!. '!&!
IDENTIlCATION
_ ` _ `
xRS & Rb jb , xRS W m & R b jb ,
(ENCE FROM AND WE GET THE IDENTIlCATIONS
_ ` _ `
& Rb xRS jb , xRS & Rb jb , xRS W m & R b jb , xRS jb ,

WHICH GIVES US A 7EIL STRUCTURE ON xRS jb , +, J'RS (ENCE WE HAVE
DElNED A CANONICAL 7EIL STRUCTURE ON +,
a
.OW LET q 4W 1L BE ANY CHARACTER )T IS WELL KNOWN THAT ONE
CAN ASSOCIATE TO q A ONE DIMENSIONAL LOCAL SYSTEM ,q TOGETHER WITH AN
ISOMORPHISM ,q & RbW ,q 4HIS LOCAL SYSTEM IS CONSTRUCTED AS FOLLOWS
#ONSIDER THE SHEAF & RW b 1L )T IS CLEAR THAT THIS SHEAF HAS A NATURAL
lBERWISE ACTION OF THE GROUP 4W 4HUS WE SET ,q TO BE THE DIRECT SUMMAND
OF & RW b 1L ON WHICH 4W ACTS BY MEANS OF THE CHARACTER q 4HE FOLLOWING
RESULT IS DUE TO ' ,USZTIG
4HEOREM
4R+,q _ DIM ' CH2qW
.OW WE ASSUME AGAIN THAT '',N 3ET h TRb , ;N =N
a
)N THIS CASE IT IS WELL KNOWN THAT FOR EVERY CHARACTER q 4W 1L THE
VECTOR SPACE SPANNED BY THE CHARACTERS OF ALL IRREDUCIBLE CONSTITUENTS OF
2qW COINCIDES WITH THE VECTOR SPACE SPANNED BY THE FUNCTIONS OF THE FORM
4R! WHERE ! RUNS OVER ALL DIRECT SUMMANDS OF +,q (ENCE 4HEOREM
FOLLOWS FROM THE FOLLOWING RESULT
0ROPOSITION ,ET TR4 4 ! DENOTE THE RESTRICTION OF THE TRACE
MORPHISM FROM ' TO 4 4HEN FOR EVERY , AS ABOVE ONE HAS

_ ` IF I N
DIM (CI TRb4 , i ,
IF I N
7E SET (, (CN TRb4 , i , N
a
,ET q 4W 1L BE A CHARACTER FOR SOME W 7 4HEN THE NATURAL
ISOMORPHISM & RbW ,q ,q GIVES RISE TO A NATURAL ENDOMORPHISM OF (,q
WHICH BY ABUSE OF LANGUAGE WE WILL ALSO DENOTE BY & R W 4HEN
& RW J(,q _ LW lW q ` )D
/NE HAS
h +, (, i +,
6ISIONS IN -ATH l &5.#4)/.3 /& 2%02%3%.4!4)/.3 !.$ ,)&4).'
-OREOVER IF , IS ENDOWED WITH AN ISOMORPHISM & RbW , , THEN IS

& RW EQUIVARIANT
4HE PROOF FOLLOWS EASILY FROM THE STANDARD PROPERTIES OF THE &OURIER
$ELIGNE TRANSFORM 4HE DETAILS ARE LEFT TO THE READER
4HE CASE OF ARBITRARY GROUP ,ET ' BE AN ARBITRARY CONNECTED
REDUCTIVE ALGEBRAIC GROUP OVER & &OR SIMPLICITY WE WILL ASSUME THAT ' IS
SPLIT ,ET '? DENOTE THE ,ANGLANDS DUAL GROUP OF ' WHICH IS A REDUCTIVE
ALGEBRAIC GROUP OVER 1L ,ET y '? !UT% BE A REPRESENTATION OF
'? WHERE % IS A VECTOR SPACE OVER 1L OF DIMENSION N 7E WOULD LIKE
TO ASSOCIATE TO IT A CERTAIN FUNCTION x ly x ON THE SET OF ISOMORPHISM
CLASSES OF IRREDUCIBLE REPRESENTATIONS OF ' 4HIS IS DONE AS FOLLOWS
4HE GROUP '? COMES EQUIPPED WITH A CANONICAL MAXIMAL TORUS 4? ,ET
US DIAGONALIZE yJ4? ,ET t tN BE THE CORRESPONDING CHARACTERS OF 4?
IN y WITH MULTIPLICITIES ,ET ALSO -? y DENOTE THE MINIMAL ,EVI SUBGROUP
CONTAINING y4? SUCH THAT THE ACTION OF -? y IS MULTIPLICITY FREE 7E DENOTE
BY 7y THE GROUP .ORM!UT% -? y - ? 7E ALSO DENOTE BY 7 THE 7EYL
y y
GROUP OF !UT%
,EMMA 7y IS NATURALLY A SUBQUOTIENT OF 7y
0ROOF 3TANDARD c
,ET 4? N
y 'M1 BE THE #ARTAN GROUP OF ',N 4HUS WE GET A NATURAL
L
MAP P? ?
y 4 4y
? SENDING EVERY T TO t T tN T
.OW LET 4y 'M& DENOTE THE DUAL TORUS TO 4? y OVER & AND LET Py
4y 4 DENOTE THE MAP WHICH IS DUAL TO P? y %XPLICITLY ONE HAS
Py X XN t X tN XN
4HE REPRESENTATION y DElNES THE NATURAL HOMOMORPHISM FROM 7 TO 7y
WHICH BY ABUSE OF THE LANGUAGE WE WILL ALSO DENOTE BY y
.OW LET x BE AN IRREDUCIBLE REPRESENTATION OF ' !SSUME THAT x APPEARS
a
IN SOME 2qW FOR SOME q 4W 1L ,ET W BE ANY LIFT OF yW TO 7y 4HEN
P INDUCES AN & RATIONAL MAP PW 4yW 4W HENCE A HOMOMORPHISM
PW 4yW 4W $ElNE
ly x lx
WHERE x IS ANY IRREDUCIBLE REPRESENTATION OF 'y WHICH APPEARS IN 2PbW q W
"Y 4HEOREM ONE HAS

ly x _ LW l4W PbW q
,EMMA 4HE DElNITION OF ly x DOES NOT DEPEND ON THE CHOICE OF W
! "2!6%2-!. !.$ $ +!:($!. '!&!
.OW LET hy DENOTE THE UNIQUE CENTRAL FUNCTION ON ' SUCH THAT FOR EVERY
IRREDUCIBLE REPRESENTATION x 6 OF ' ONE HAS
8
hy G xG ly x ` )D6
G'
7E WOULD LIKE TO COMPUTE THIS FUNCTION EXPLICITLY 7E WILL BE ABLE TO SOLVE
ONLY A SLIGHTLY WEAKER PROBLEM .AMELY WE ARE GOING TO CONSTRUCT EXPLICITLY
AN @ ADIC PERVERSE SHEAF h y ON ' SUCH THAT
4Rh Ey
hy h
E y IS A FUNCTION WHICH SATISlES THE FOLLOWING CONDITION
WHERE h
8
E y G xG ly x ` )D
h
G'
FOR EVERY IRREDUCIBLE <GENERIC REPRESENTATION x OF ' CF 4HEOREM FOR
THE PRECISE STATEMENT
4HE PERVERSE SHEAF hy )N WHAT FOLLOWS WE ASSUME THAT
THE ZERO WEIGHT DOES NOT APPEAR IN y ,ET TRy 4y ! BE GIVEN BY
TRy X XN X XN
#ONSIDER THE COMPLEX !y Py TRby , ;N=N
ON 4 )T IS EASY TO SEE
THAT THIS COMPLEX IS PERVERSE 7E WOULD LIKE TO ENDOW THIS COMPLEX WITH A
7 EQUIVARIANT STRUCTURE
#HOOSE W 7 7E NEED TO DElNE AN ISOMORPHISM rW Wb !y F !y
,ET AS ABOVE W BE ANY LIFT OF yW TO 7y 4HEN ONE HAS
Py W T WPy T
4HE SHEAF TRby , IS OBVIOUSLY 7y EQUIVARIANT 4HIS TOGETHER WITH
GIVES RISE TO AN ISOMORPHISM rW Wb !y F !y 7E NOW DElNE rW

_ LW _LW rW
0ROPOSITION 4HE ISOMORPHISM rW DOES NOT DEPEND ON THE CHOICE
OF W 4HE ASSIGNMENT W rW DElNES A 7 EQUIVARIANT STRUCTURE ON THE
SHEAF !y
4HE ABOVE 7 EQUIVARIANT STRUCTURE ON !y GIVES RISE TO THE SHEAF "y
Q !y 7 ON THE QUOTIENT 47 WHERE Q 4 47 IS THE CANONICAL MAP
,ET 'R BE THE SET OF REGULAR ELEMENTS OF ' AND LET I BE ITS EMBEDDING
TO ' ,ET S 'R 47 BE THE MORPHISM COMING FROM THE IDENTIlCATION
47 /' ' 7E DElNE
q r
b DIM ' _ DIM 4
h y Ib S "y ;DIM ' _ DIM 4=

6ISIONS IN -ATH l &5.#4)/.3 /& 2%02%3%.4!4)/.3 !.$ ,)&4).'
7E ALSO DElNE
E y 4Rh
h hy
"Y THE DElNITION THE COMPLEX hy ON ' IS ' EQUIVARIANT WITH RESPECT TO
E y IS CENTRAL
THE ADJOINT ACTION AND THEREFORE THE FUNCTION h
4HE ACTION OF h E y IN 2qW 2ECALL THAT A LOCAL SYSTEM , ON
4 IS CALLED QUASI REGULAR IF FOR EVERY COROOT j? 'M 4 THE LOCAL SYSTEM
a
j? b , IS NON TRIVIAL 7E SAY THAT A CHARACTER q 4W 1L IS QUASI REGULAR
IF THE LOCAL SYSTEM ,q IS QUASI REGULAR
4HEOREM ,ET x BE AN IRREDUCIBLE REPRESENTATION OF ' WHICH AP
PEARS IN THE $ELIGNE ,USZTIG REPRESENTATION 2qW FOR A QUASI REGULAR CHAR
a
ACTER q 4W 1L 4HEN
8
E y G xG ly x
h
G'
4HE BASIC CONJECTURE )N ORDER TO PROCEED WE WILL HAVE TO

ASSUME THAT THE FOLLOWING CONJECTURE HOLDS
#HOOSE A MAXIMAL UNIPOTENT SUBGROUP IN 5 y ' $ENOTE BY " THE
"OREL SUBGROUP OF ' WHICH NORMALIZES 5 ,ET 8 '5 AND LET R
' 8 DENOTE THE NATURAL PROJECTION "Y THE DElNITION ONE HAS CANONICAL
IDENTIlCATION 4 "5 (ENCE 4 IS NATURALLY EMBEDDED INTO 8
#ONJECTURE 4HE COMPLEX R "y VANISHES OUTSIDE OF 4
/NE CAN SHOW THAT #ONJECTURE IMPLIES THAT HOLDS FOR ANY q

.OTATION )N THIS APPENDIX WE ASSUME THAT & IS A lNITE EXTENSION
OF 1P 7E DENOTE BY C THE MAXIMAL IDEAL OF /&
,ET # BE A SMOOTH QUASI PROJECTIVE CURVE OVER & AND C BE A & POINT
OF # ,ET # #NC
,ET F 8 # BE A SMOOTH PROPER MORPHISM OF RELATIVE DIMENSION N
AND BE A DIkERENTIAL FORM jTOP
8# OF THE TOP DEGREE

&OR AN & POINT C # & WE DENOTE BY JJ THE MEASURE ASSOCIATED
2 TO
ON THE SPACE OF & POINTS OF THE lBER 8C OVER C 0UT 6 C 8C + JJ
#HOOSE A COORDINATE T ON A NEIGHBORHOOD OF C WITH TC
4HEOREM 4HERE EXIST INTEGERS N AND R SUCH THAT 6 TT 6 T FOR
ALL T CN T /&b R
! "2!6%2-!. !.$ $ +!:($!. '!&!
4HIS THEOREM EASILY IMPLIES 4HEOREM

0ROOF &IRST WE CAN MAKE USE OF .AGATAS THEOREM TO CONSTRUCT A PROPER
INTEGRAL SCHEME 8 OVER # SUCH THAT 8 a# # 8 "Y (IRONAKAS THEOREM
WE CAN BLOW UP 8 TO BUILD E F 9 # SUCH THAT THE UNION OF THE lBER
9C OVER C AND THE CLOSURE THE ZERO LOCUS OF THE DIkERENTIAL FORM ON
9N9C IS A NORMAL CROSSING DIVISOR )T FOLLOWS THAT FOR ANY & POINT A OF
9C THERE EXIST LOCAL COORDINATES XI I N ON A NEIGHBORHOOD :A
OF A SATISFYING THE FOLLOWING CONDITIONS
F UeI XNI I FOR SOME FUNCTION U ON :A WITH UA AND NI t
I TE
II VeI XM I
I DX DX DX DXN_ ON :A N9C FOR SOME FUNCTION V ON :A
WITH VA AND MI :
)T IS CLEAR THAT FOR A POINT C # & SUnCIENTLY CLOSE TO C THE lBERS
8C & AND 9C & HAVE THE SAME MEASURE (ENCE WE CAN REPLACE 8 BY
9N9 C "Y THE IMPLICIT FUNCTION THEOREM WE CAN CHOOSE A NEIGHBORHOOD
:A OF A IN P ADIC TOPOLOGY A :A y :A SUCH THAT COORDINATES XI GIVE AN
ISOMORPHISM OF SETS :A CL a CL a ` ` ` a CL WHERE L IS A POSITIVE INTEGER
_
-OREOVER IF WE CHOOSE L SUnCIENTLY LARGE THE FUNCTION UA U N IS WELL
_
DElNED ON :A HENCE CHANGING X FOR UA U N X WE CAN SUPPOSE THAT
ON :A ONE HAS TF^ UA eI XNI I 7ITHOUT LOSS OF GENERALITY WE CAN ASSUME
THAT JVJ IS CONSTANT ON :A 3INCE 9C & IS COMPACT IT CAN BE COVERED BY
lNITELY MANY OPEN SETS :A SATISFYING THE PROPERTIES STATED ABOVE )N FACT
ONE CAN CHOOSE THEM TO BE DISJOINT 4O PROVE THIS WE NEED THE FOLLOWING
LEMMA
,EMMA ,ET 1 BE AN OPEN COMPACT SUBSET OF THE AnNE SPACE & N 7E
CLAIM THAT 1 CAN BE COVERED BY lNITELY MANY DISJOINT BALLS CONTAINED IN 1
! BALL IS A SUBSET OF THE FORM "RA FX XN + N J XI _ AI J RG
3
7E APPLY THE LEMMA TO SETS :Q JQ :A J Q K "Y CONSTRUC
TION EACH OF THEM IS IDENTIlED WITH AN OPEN COMPACT SUBSET OF & N
0ROOF 3INCE 1 IS COMPACT WE CAN COVER IT BY lNITELY MANY BALLS CONTAINED
IN 1 ,ET R BE THE SMALLEST RADIUS OF THESE BALLS 0ICK AN INTEGER L SUCH
THAT _ LOGQ R L #ONSIDER THE PROJECTION P 1 &ML N 4HE PRE
IMAGE OF EACH ELEMENT OF )M P IS A BALL 4HESE BALLS CONSTITUTE THE DESIRABLE
COVERING
.OW THE PROPOSITION FOLLOWS FROM THE FOLLOWING SIMPLE LEMMA
,EMMA ,ET STAND FOR DIkERENTIAL FORM
9 M
XI I DX DX DX DXN_
I
6ISIONS IN -ATH l &5.#4)/.3 /& 2%02%3%.4!4)/.3 !.$ ,)&4).'
WHERE MI : AND :
6 T N
JJ
eI XI I TXI M
4HEN 6 T ` T 6 T FOR ANY T C T /&b N c
2EFERENCES
;""$= ! "EILINSON * "ERNSTEIN 0 $ELIGNE &AISCEAUX PERVERS IN <!NAL
YSIS AND 4OPOLOGY ON 3INGULAR 3PACES ) ,UMINY !STrERISQUE
[
;"E+= ! "ERENSTEIN $ +AZHDAN 'EOMETRIC AND UNIPOTENT CRYSTALS
'!&! IN THIS ISSUE
;"ER$= * "ERNSTEIN 0 $ELIGNE ,E CENTRE DE "ERNSTEIN 4RAVAUX EN #OURS
2EPRESENTATIONS OF 2EDUCTIVE 'ROUPS OVER A ,OCAL &IELD [ (ERMANN
0ARIS
;"ER:= * "ERNSTEIN ! :ELEVINSKY 2EPRESENTATIONS OF THE GROUP ',N &
WHERE & IS A LOCAL NON !RCHIMEDEAN lELD IN 2USSIAN 5SPEHI -AT .AUK
[
;"ER:= * "ERNSTEIN ! :ELEVINSKY )NDUCED REPRESENTATIONS OF REDUCTIVE P
s
ADIC GROUPS ) !NN 3CI %COLE .ORM 3UP [
;"O*= ! "OREL ( *ACQUET !UTOMORPHIC FORMS AND AUTOMORPHIC REPRESEN
TATIONS WITH A SUPPLEMENT </N THE NOTION OF AN AUTOMORPHIC REPRESEN
TATION BY 20 ,ANGLANDS 0ROC 3YMPOS 0URE -ATH 888))) !UTOMOR
PHIC FORMS REPRESENTATIONS AND , FUNCTIONS 0ROC 3YMPOS 0URE -ATH
/REGON 3TATE 5NIV #ORVALLIS /RE 0ART !MER -ATH 3OC
0ROVIDENCE 2) [
;"R+= ! "RAVERMAN $ +AZHDAN /N THE 3CHWARTZ SPACE OF THE BASIC AnNE
SPACE 3ELECTA -ATH .3 [
;''0= ) 'ELFAND - 'RAEV ) 0IATETSKII 3HAPIRO 2EPRESENTATION 4HEORY
AND !UTOMORPHIC &UNCTIONS 0HILADELPHIA 0! ,ONDON 4ORONTO
;'+= )- 'ELFAND $ +AZHDAN 2EPRESENTATIONS OF THE GROUP ',N +
WHERE + IS A LOCAL lELD IN <,IE 'ROUPS AND 4HEIR 2EPRESENTATIONS 0ROC
3UMMER 3CHOOL "OLYAI *ANOS -ATH 3OC "UDAPEST (ALSTED .EW
9ORK [
;'E+= 3 'ELFAND $ +AZHDAN #ONJECTURAL ALGEBRAIC FORMULAS FOR REPRESEN
TATIONS OF ',N IN <3IR -ICHAEL !TIYAH A 'REAT -ATHEMATICIAN OF THE
4WENTIETH #ENTURY !SIAN * -ATH [
;'O*= 2 'ODEMENT ( *ACQUET :ETA FUNCTIONS OF SIMPLE ALGEBRAS 3PRINGER
,ECTURE .OTES IN -ATHEMATICS
;(= - (ARRIS 4HE LOCAL ,ANGLANDS CONJECTURE FOR ',N OVER A P ADIC lELD
N P )NVENT -ATH [
! "2!6%2-!. !.$ $ +!:($!. '!&!
;(4= - (ARRIS 2 4AYLOR /N THE GEOMETRY OF COHOMOLOGY OF SOME SIMPLE

3HIMURA VARIETIES PREPRINT
;*03= ( *ACQUET ) 0IATETSKII 3HAPIRO * 3HALIKA 2ANKIN 3ELBERG CON
VOLUTIONS !MER * -ATH [
;+= $ +AZHDAN /N LIFTING ,IE GROUP REPRESENTATIONS )) #OLLEGE 0ARK -D
3PRINGER ,ECTURE .OTES IN -ATH [
;+= $ +AZHDAN &ORMS OF THE PRINCIPLE SERIES FOR ',N IN <&UNCTIONAL !NAL
YSIS ON THE %VE OF THE ST #ENTURY 6OL .EW "RUNSWICK .*
"IRKHAUSER 0ROGR -ATH [
;+= $ +AZHDAN !N ALGEBRAIC INTEGRATION IN <-ATHEMATICS &RONTIERS AND
0ERSPECTIVES 6 !RNOLD - !TIYAH 0 ,AX " -AZUR EDS !-3
;,= 20 ,ANGLANDS 0ROBLEMS IN THE THEORY OF AUTOMORPHIC FORMS IN <,EC
TURES IN -ODERN !NALYSIS AND !PPLICATIONS ))) 3PRINGER ,ECTURE .OTES
IN -ATH [
;,A= ' ,AUMON &AISCEAUX CHARACTrERS DAPRrES ,USZTIG IN &RENCH 3sEMINAIRE
"OURBAKI 6OL !STsERISQUE [
;,A23= ' ,AUMON - 2APOPORT 5 3TUHLER $ ELLIPTIC SHEAVES AND THE
,ANGLANDS CORRESPONDENCE )NVENT -ATH [
;,U= ' ,USZTIG #HARACTER SHEAVES ) !DV IN -ATH [
;0= ) 0IATETSKII 3HAPIRO -ULTIPLICITY ONE THEOREMS IN <!UTOMORPHIC
&ORMS 2EPRESENTATIONS AND , FUNCTIONS 0ROC 3YMPOS 0URE -ATH /RE
GON 3TATE 5NIV #ORVALLIS /RE 0ART 0ROC 3YMPOS 0URE -ATH
888))) !MER -ATH 3OC 0ROVIDENCE 2) [
;6= %" 6INBERG /N REDUCTIVE ALGEBRAIC SEMI GROUPS IN <,IE 'ROUPS AND
,IE !LGEBRAS %" $YNKINS 3EMINAR !MER -ATH 3OC 4RANSL 3ER
[
;4= * 4ATE &OURIER ANALYSIS IN NUMBER lELDS AND (ECKES ZETA FUNCTIONS !L
GEBRAIC .UMBER 4HEORY 0ROC )NSTRUCTIONAL #ONF "RIGHTON
[
;7= . 7ALLACH 2EAL 2EDUCTIVE 'ROUPS ) 0URE AND !PPLIED -ATHEMATICS
!CADEMIC 0RESS )NC "OSTON -!
;7E= ! 7EIL !DsELES AND !LGEBRAIC 'ROUPS "IRKHAUSER 0ROGRESS IN -ATHE
MATICS
! "RAVERMAN $EPARTMENT OF -ATHEMATICS -)4 -ASS !VE #AMBRIDGE

-! 53!
$ +AZHDAN $EPARTMENT OF -ATHEMATICS (ARVARD 5NIVERSITY #AMBRIDGE -!
53!
6 6OLOGODSKY $EPARTMENT OF -ATHEMATICS (ARVARD 5NIVERSITY #AMBRIDGE
-! 53!
0$% !3 ! 5.)&)%$ 35"*%#4
3ERGIU +LAINERMAN
)NTRODUCTION
'IVEN THAT ONE OF THE GOALS OF THE CONFERENCE IS TO ADDRESS THE ISSUE OF THE
UNITY OF -ATHEMATICS ) FEEL EMBOLDENED TO TALK ABOUT A QUESTION WHICH HAS
KEPT BOTHERING ME ALL THROUGH MY SCIENTIlC CAREER )S THERE REALLY A UNIlED
SUBJECT OF -ATHEMATICS WHICH ONE CAN CALL 0$% !T lRST GLANCE THIS SEEMS
EASY WE MAY DElNE 0$% AS THE SUBJECT WHICH IS CONCERNED WITH ALL PARTIAL
DIkERENTIAL EQUATIONS !CCORDING TO THIS VIEW THE GOAL OF THE SUBJECT IS TO
lND A GENERAL THEORY OF ALL OR VERY GENERAL CLASSES OF 0$%S 4HIS <NATURAL
DElNITION COMES DANGEROUSLY CLOSE TO WHAT - 'ROMOV HAD IN MIND )
BELIEVE WHEN HE WARNED US DURING THE CONFERENCE THAT OBJECTS DElNITIONS
OR QUESTIONS WHICH LOOK NATURAL AT lRST GLANCE MAY IN FACT <BE STUPID
)NDEED IT IS NOW RECOGNIZED BY MANY PRACTITIONERS OF THE SUBJECT THAT THE
GENERAL POINT OF VIEW AS A GOAL IN ITSELF IS SERIOUSLY mAWED 4HAT IT EVER HAD
ANY CREDIBILITY IS DUE TO THE FACT THAT IT WORKS QUITE WELL FOR LINEAR 0$%S
WITH CONSTANT COEnCIENTS IN WHICH CASE THE &OURIER TRANSFORM IS EXTREMELY
EkECTIVE )T HAS ALSO PRODUCED SIGNIlCANT RESULTS FOR SOME GENERAL SPECIAL
CLASSES OF LINEAR EQUATIONS WITH VARIABLE COEnCIENTS )TS WEAKNESS IS MOST
EVIDENT IN CONNECTION TO NONLINEAR EQUATIONS 4HE ONLY USEFUL GENERAL RESULT
WE HAVE IS THE #AUCHY +OWALEVSKY THEOREM IN THE QUITE BORING CLASS OF
ANALYTIC SOLUTIONS )N THE MORE RESTRICTIVE FRAMEWORKS OF ELLIPTIC HYPERBOLIC
OR PARABOLIC EQUATIONS SOME IMPORTANT LOCAL ASPECTS OF NONLINEAR EQUATIONS
CAN BE TREATED WITH A CONSIDERABLE DEGREE OF GENERALITY )T IS THE PASSAGE
FROM LOCAL TO GLOBAL PROPERTIES WHICH FORCES US TO ABANDON ANY GENERALITY
AND TAKE FULL ADVANTAGE OF THE SPECIAL FEATURES OF THE IMPORTANT EQUATIONS
4HE FACT IS THAT 0$%S IN PARTICULAR THOSE THAT ARE NONLINEAR ARE TOO
SUBTLE TO lT INTO A TOO GENERAL SCHEME ON THE CONTRARY EACH IMPORTANT

,INEAR EQUATIONS WITH VARIABLE COEnCIENTS APPEAR NATURALLY BY LINEARIZING NONLINEAR
EQUATIONS AROUND SPECIlC SOLUTIONS 4HEY ALSO APPEAR IN THE STUDY OF SPECIlC OPERATORS
ON MANIFOLDS IN 3EVERAL #OMPLEX 6ARIABLES AND 1UANTUM -ECHANICS 4HE INTERACTION
BETWEEN THE v OPERATOR IN 3#6 AND ITS NATURAL BOUNDARY VALUE PROBLEMS HAVE LED TO VERY
INTERESTING LINEAR EQUATIONS WITH EXOTIC FEATURES SUCH AS LACK OF SOLVABILITY
3 +,!).%2-!. '!&!
0$% SEEMS TO BE A WORLD IN ITSELF -OREOVER GENERAL POINTS OF VIEW OFTEN

OBSCURE THROUGH UNNECESSARY TECHNICAL COMPLICATIONS THE MAIN PROPERTIES
OF THE IMPORTANT SPECIAL CASES ! USEFUL GENERAL FRAMEWORK IS ONE WHICH
PROVIDES A SIMPLE AND ELEGANT TREATMENT OF A PARTICULAR PHENOMENON AS IS
THE CASE OF SYMMETRIC HYPERBOLIC SYSTEMS IN CONNECTION TO THE PHENOMENON
OF lNITE SPEED OF PROPAGATION AND THE GENERAL TREATMENT OF LOCAL EXISTENCE
FOR NONLINEAR HYPERBOLIC EQUATIONS 9ET EVEN WHEN A GENERAL FRAMEWORK IS
USEFUL AS SYMMETRIC HYPERBOLIC SYSTEMS CERTAINLY ARE ONE WOULD BE WRONG
TO EXPAND THE FRAMEWORK BEYOND ITS NATURAL ROLE 3YMMETRIC HYPERBOLIC
SYSTEMS TURN OUT TO BE SIMPLY TOO GENERAL FOR THE STUDY OF MORE RElNED
QUESTIONS CONCERNING THE IMPORTANT EXAMPLES OF HYPERBOLIC EQUATIONS
!S THE GENERAL POINT OF VIEW HAS LOST ITS APPEAL MANY OF US HAVE ADOPTED
A PURELY PRAGMATIC POINT OF VIEW OF OUR SUBJECT WE CHOSE TO BE CONCERNED
ONLY WITH THOSE 0$%S OR CLASSES OF 0$%S WHICH ARE CONSIDERED IMPORTANT
!ND INDEED THE RANGE OF APPLICATIONS OF SPECIlC 0$%S IS PHENOMENAL MANY
OF OUR BASIC EQUATIONS BEING IN FACT AT THE HEART OF FULLY mEDGED lELDS OF
-ATHEMATICS OR 0HYSICS SUCH AS #OMPLEX !NALYSIS 3EVERAL #OMPLEX 6ARI
ABLES -INIMAL 3URFACES (ARMONIC -APS #ONNECTIONS ON 0RINCIPAL "UN
DLES +AHLERIAN AND %INSTEIN 'EOMETRY 'EOMETRIC &LOWS (YDRODYNAM
ICS %LASTICITY 'ENERAL 2ELATIVITY %LECTRODYNAMICS .ONRELATIVISTIC 1UAN
TUM -ECHANICS ETC /THER IMPORTANT SUBJECTS OF -ATHEMATICS SUCH AS
(ARMONIC !NALYSIS 0ROBABILITY 4HEORY AND VARIOUS AREAS OF -ATHEMATICAL
0HYSICS ARE INTIMATELY TIED TO ELLIPTIC PARABOLIC HYPERBOLIC OR 3CHR
ODINGER
TYPE EQUATIONS 3PECIlC GEOMETRIC EQUATIONS SUCH AS ,APLACE["ELTRAMI AND
$IRAC OPERATORS ON MANIFOLDS (ODGE SYSTEMS 0SEUDO HOLOMORPHIC CURVES
9ANG[-ILLS AND RECENTLY 3EIBERG[7ITTEN HAVE PROVED TO BE EXTRAORDINAR
ILY USEFUL IN 4OPOLOGY AND 3YMPLECTIC 'EOMETRY 4HE THEORY OF )NTEGRABLE
SYSTEMS HAS TURNED OUT TO HAVE DEEP APPLICATIONS IN !LGEBRAIC 'EOMETRY
THE SPECTRAL THEORY ,APLACE["ELTRAMI OPERATORS AS WELL AS THE SCATTERING
THEORY FOR WAVE EQUATIONS ARE INTIMATELY TIED TO THE STUDY OF AUTOMORPHIC
FORMS IN .UMBER 4HEORY &INALLY !PPLIED -ATHEMATICS TAKES AN INTER
EST NOT ONLY IN THE BASIC PHYSICAL EQUATIONS BUT ALSO ON A LARGE VARIETY OF
PHENOMENOLOGICAL 0$%S OF RELEVANCE TO ENGINEERS BIOLOGISTS CHEMISTS OR
ECONOMISTS
7ITH ALL ITS OBVIOUS APPEAL THE PRAGMATIC POINT OF VIEW MAKES IT DInCULT
TO SEE 0$% AS A SUBJECT IN ITS OWN RIGHT 4HE DEEPER ONE DIGS INTO THE STUDY
OF A SPECIlC 0$% THE MORE ONE HAS TO TAKE ADVANTAGE OF THE PARTICULAR
FEATURES OF THE EQUATION AND THEREFORE THE CORRESPONDING RESULTS MAY MAKE
6ISIONS IN -ATH 0$% !3 ! 5.)&)%$ 35"*%#4
SENSE ONLY AS CONTRIBUTIONS TO THE PARTICULAR lELD TO WHICH THAT 0$% IS

RELEVANT 4HUS EACH MAJOR EQUATION SEEMS TO GENERATE ISOLATED ISLANDS OF
MATHEMATICAL ACTIVITY -OREOVER A PARTICULAR 0$% MAY BE STUDIED FROM
LARGELY DIkERENT POINTS OF VIEW BY AN APPLIED MATHEMATICIAN A PHYSICIST
A GEOMETER OR AN ANALYST !S WE LOSE PERSPECTIVE ON THE COMMON FEATURES
OF OUR MAIN EQUATIONS WE SEE 0$% LESS AND LESS AS A UNIlED SUBJECT 4HE
lELD OF 0$% AS A WHOLE HAS ALL BUT CEASED TO EXIST EXCEPT IN SOME OLD
FASHIONED TEXTBOOKS 7HAT WE HAVE INSTEAD IS A LARGE COLLECTION OF LOOSELY
CONNECTED SUBJECTS
)N THE END ) lND THIS VIEW NOT ONLY SOMEWHAT DISCONCERTING BUT ALSO
INTELLECTUALLY AS UNSATISFACTORY AS THE lRST 4HERE EXISTS AFTER ALL AN IM
PRESSIVE GENERAL BODY OF KNOWLEDGE WHICH WOULD CERTAINLY BE INCLUDED UNDER
THE FRAMEWORK OF A UNIlED SUBJECT IF WE ONLY KNEW WHAT THAT WAS (ERE
ARE JUST A FEW EXAMPLES OF POWERFUL GENERAL IDEAS
7ELL POSEDNESS &IRST INVESTIGATED BY (ADAMARD AT THE BEGINNING OF
THIS CENTURY WELL POSED PROBLEMS ARE AT THE HEART OF THE MODERN THEORY OF
0$% 4HE ISSUE OF WELL POSEDNESS COMES ABOUT WHEN WE DISTINGUISH BETWEEN
ANALYTIC AND SMOOTH SOLUTIONS 4HIS IS FAR FROM BEING AN ACADEMIC SUBTLETY
WITHOUT SMOOTH NON ANALYTIC SOLUTIONS WE CANNOT TALK ABOUT lNITE SPEED
OF PROPAGATION THE DISTINCTIVE MARK OF RELATIVISTIC PHYSICS 0ROBLEMS ARE
SAID TO BE WELL POSED IF THEY ADMIT UNIQUE SOLUTIONS FOR GIVEN SMOOTH INI
TIAL OR BOUNDARY CONDITIONS 4HE CORRESPONDING SOLUTIONS HAVE TO DEPEND
CONTINUOUSLY ON THE DATA 4HIS LEADS TO THE CLASSIlCATION OF LINEAR EQUA
TIONS INTO ELLIPTIC HYPERBOLIC AND PARABOLIC WITH THEIR SPECIlC BOUNDARY
VALUE PROBLEMS 7ELL POSEDNESS ALSO PLAYS A FUNDAMENTAL ROLE IN THE STUDY
OF NONLINEAR EQUATIONS SEE A DETAILED DISCUSSION IN THE LAST SECTION OF THIS
PAPER 4HE COUNTERPART OF WELL POSEDNESS IS ALSO IMPORTANT IN MANY AP
PLICATIONS )LL POSED PROBLEMS APPEAR NATURALLY IN #ONTROL 4HEORY )NVERSE
3CATTERING ETC WHENEVER WE HAVE A LIMITED KNOWLEDGE OF THE DESIRED SO
LUTIONS 5NIQUE CONTINUATION OF SOLUTIONS TO GENERAL CLASSES OF 0$%S IS
INTIMATELY TIED TO ILL POSEDNESS

) FAILED TO MENTION IN THE FEW EXAMPLES GIVEN ABOVE THE DEVELOPMENT OF TOPOLOGICAL
METHODS FOR DEALING WITH GLOBAL PROPERTIES OF ELLIPTIC 0$%S AS WELL AS SOME OF THE IM
PORTANT FUNCTIONAL ANALYTIC TOOLS CONNECTED TO (ILBERT SPACE METHODS COMPACTNESS THE
IMPLICIT FUNCTION THEOREMS ETC ) ALSO FAILED TO MENTION THE LARGE BODY OF KNOWLEDGE WITH
REGARD TO SPACES OF FUNCTIONS SUCH AS 3OBOLEV 3CHAUDER "-/ AND (ARDY ETC OR THE
RECENT IMPORTANT DEVELOPMENTS IN NONLINEAR WAVE AND DISPERSIVE EQUATIONS CONNECTED TO
RESTRICTION THEOREMS IN &OURIER !NALYSIS &OR A MORE IN DEPTH DISCUSSIONS OF MANY OF THE
IDEAS MENTIONED BELOW AND THEIR HISTORY SEE THE RECENT SURVEY ;"RE"=
3 +,!).%2-!. '!&!
! PRIORI ESTIMATES BOOT STRAP AND CONTINUITY ARGUMENTS ! PRIORI ESTI

MATES ALLOW US TO DERIVE CRUCIAL INFORMATION ABOUT SOLUTIONS TO COMPLICATED
EQUATIONS WITHOUT HAVING TO SOLVE THE EQUATIONS 4HE BEST KNOWN EXAMPLES
ARE ENERGY ESTIMATES MAXIMUM PRINCIPLE OR MONOTONICITY TYPE ARGUMENTS
#ARLEMAN TYPE ESTIMATES APPEAR IN CONNECTION TO ILL POSED PROBLEMS 4HE
A PRIORI ESTIMATES CAN BE USED TO ACTUALLY CONSTRUCT THE SOLUTIONS PROVE
THEIR UNIQUENESS AND REGULARITY AND PROVIDE OTHER QUALITATIVE INFORMATION
4HE BOOT STRAP TYPE ARGUMENT IS A POWERFUL GENERAL PHILOSOPHY TO DERIVE A
PRIORI ESTIMATES FOR NONLINEAR EQUATIONS !CCORDING TO IT WE START BY MAKING
ASSUMPTIONS ABOUT THE SOLUTIONS WE ARE LOOKING FOR 4HIS ALLOWS US TO THINK
OF THE ORIGINAL NONLINEAR PROBLEM AS A LINEAR ONE WHOSE COEnCIENTS SATISFY
PROPERTIES CONSISTENT WITH THE ASSUMPTIONS 7E MAY THEN USE LINEAR METH
ODS A PRIORI ESTIMATES TO TRY TO SHOW THAT THE SOLUTIONS TO THE NEW LINEAR
PROBLEM BEHAVE AS WELL OR BETTER THAN WE HAVE POSTULATED ! CONTINUITY
TYPE ARGUMENT ALLOWS US TO CONCLUDE THE ORIGINAL ASSUMPTIONS ARE IN FACT
TRUE 4HIS <CONCEPTUAL LINEARIZATION OF THE ORIGINAL NONLINEAR EQUATION LIES
AT THE HEART OF OUR MOST IMPRESSIVE RESULTS FOR NONLINEAR EQUATIONS
2EGULARITY THEORY FOR LINEAR ELLIPTIC EQUATIONS 7E HAVE SYSTEMATIC
METHODS FOR DERIVING POWERFUL REGULARITY ESTIMATES FOR LINEAR ELLIPTIC EQUA
TIONS 4HE , ESTIMATES ARE COVERED BY 3CHAUDER THEORY 4HE MORE RElNED
,P THEORY OCCUPIES AN IMPORTANT PART OF MODERN 2EAL AND (ARMONIC !NAL
YSIS 4HE THEORY OF SINGULAR INTEGRALS AND PSEUDODIkERENTIAL OPERATORS ARE
INTIMATELY TIED TO THE DEVELOPMENT OF ,P REGULARITY THEORY
$IRECT VARIATIONAL METHODS 4HE SIMPLEST EXAMPLE OF A DIRECT VARIA
TIONAL METHOD IS THE $IRICHLET 0RINCIPLE 4HOUGH lRST PROPOSED BY $IRICHLET
AS A METHOD OF SOLVING THE 0OISSON EQUATION a} F AND LATER USED BY 2IE
MANN IN HIS CELEBRATED PROOF OF THE 2IEMANN -APPING 4HEOREM IN COMPLEX
ANALYSIS IT WAS ONLY PUT ON A lRM MATHEMATICAL GROUND IN THIS CENTURY
4HE METHOD HAS MANY DEEP APPLICATIONS TO ELLIPTIC PROBLEMS )T ALLOWS
ONE TO lRST SOLVE THE ORIGINAL PROBLEM IN A<GENERALIZED SENSE AND THEN
USE REGULARITY ESTIMATES TO SHOW THAT THE GENERALIZED SOLUTIONS ARE IN FACT
CLASSICAL 4HE ULTIMATE KNOWN EXPRESSION OF THIS SECOND STEP IS EMBODIED
IN THE $E 'IORGI[.ASH METHOD WHICH ALLOWS ONE TO DERIVE FULL REGULARITY
ESTIMATES FOR THE GENERALIZED SOLUTIONS OF NONLINEAR SCALAR ELLIPTIC EQUA
TIONS 4HIS PROVIDES IN PARTICULAR THE SOLUTION TO THE FAMOUS PROBLEM OF
THE REGULARITY OF MINIMAL HYPERSURFACES AS GRAPHS OVER CONVEX OR MEAN
CONVEX DOMAINS IN ALL DIMENSIONS /THER IMPORTANT APPLICATIONS OF THE
$E 'IORGI[.ASH METHOD WERE FOUND IN CONNECTION WITH SUCH DIVERSE SITUA
6ISIONS IN -ATH 0$% !3 ! 5.)&)%$ 35"*%#4
TIONS AS THE #ALABI PROBLEM IN +AHLER 'EOMETRY 2 (AMILTONS 2ICCI mOW

AND FREE BOUNDARY VALUE PROBLEM ARISING IN #ONTINUUM -ECHANICS
%NERGY TYPE ESTIMATES 4HE ENERGY ESTIMATES PROVIDE A VERY GENERAL
TOOL FOR DERIVING A PRIORI ESTIMATES FOR HYPERBOLIC AND OTHER EVOLUTION EQUA
TIONS 4OGETHER WITH 3OBOLEV INEQUALITIES WHICH WERE DEVELOPED FOR THIS
REASON THEY ALLOW US TO PROVE LOCAL IN TIME EXISTENCE UNIQUENESS AND CON
TINUOUS DEPENDENCE ON THE INITIAL DATA FOR GENERAL CLASSES OF NONLINEAR HYPER
BOLIC EQUATIONS SUCH AS SYMMETRIC HYPERBOLIC SIMILAR TO THE CLASSICAL LOCAL
EXISTENCE RESULT FOR ORDINARY DIkERENTIAL EQUATIONS ! MORE GENERAL TYPE
OF ENERGY ESTIMATES BASED ON USING THE SYMMETRIES OF THE LINEAR PART OF
THE EQUATIONS ALLOWS ONE TO ALSO PROVE GLOBAL IN TIME PERTURBATION RESULTS
SUCH AS THE GLOBAL STABILITY OF THE -INKOWSKI SPACE IN 'ENERAL 2ELATIVITY
-ICROLOCAL ANALYSIS PARAMETRICES AND PARADIkERENTIAL CALCULUS /NE OF
THE FUNDAMENTAL DInCULTIES OF HYPERBOLIC EQUATIONS CONSISTS OF THE INTERPLAY
BETWEEN GEOMETRIC PROPERTIES WHICH CONCERN THE PHYSICAL SPACE AND PROP
ERTIES INTIMATELY TIED TO OSCILLATIONS WHICH ARE BEST SEEN IN &OURIER SPACE
-ICROLOCAL ANALYSIS IS A GENERAL STILL DEVELOPING PHILOSOPHY ACCORDING TO
WHICH ONE ISOLATES THE MAIN DInCULTIES BY CAREFUL LOCALIZATIONS IN PHYSI
CAL OR &OURIER SPACE OR IN BOTH !N IMPORTANT APPLICATION OF THIS POINT
OF VIEW IS THE CONSTRUCTION OF PARAMETRICES AS &OURIER INTEGRAL OPERATORS
FOR LINEAR HYPERBOLIC EQUATIONS AND THEIR USE IN PROPAGATION OF SINGULARI
TIES RESULTS 4HE PARADIkERENTIAL CALCULUS CAN BE VIEWED AS AN EXTENSION
OF THIS PHILOSOPHY TO NONLINEAR EQUATIONS )T ALLOWS ONE TO MANIPULATE THE
FORM OF A NONLINEAR EQUATION BY TAKING ACCOUNT OF THE WAY LARGE AND SMALL
FREQUENCIES INTERACT TO ACHIEVE REMARKABLE TECHNICAL VERSATILITY
'ENERALIZED SOLUTIONS 4HE IDEA OF A GENERALIZED SOLUTION APPEARS AL
READY IN THE WORK OF $!LEMBERT SEE ;,U= IN CONNECTION WITH THE ONE
DIMENSIONAL WAVE EQUATION VIBRATING STRING ! SYSTEMATIC AND COM
PELLING CONCEPT OF GENERALIZED SOLUTIONS HAS DEVELOPED IN CONNECTION WITH
THE $IRICHLET PRINCIPLE MORE GENERALLY VIA THE DIRECT VARIATIONAL METHOD
4HE CONSTRUCTION OF FUNDAMENTAL SOLUTIONS TO LINEAR EQUATIONS LED ALSO TO
VARIOUS TYPES OF SUCH SOLUTIONS 4HIS AND OTHER DEVELOPMENTS IN LINEAR
THEORY LED TO THE INTRODUCTION OF DISTRIBUTIONS BY , 3CHWARTZ 4HE THE
ORY OF DISTRIBUTIONS PROVIDES A MOST SATISFACTORY FRAMEWORK TO GENERALIZED
SOLUTIONS IN LINEAR THEORY 4HE QUESTION OF WHAT IS A GOOD CONCEPT OF A GEN
ERALIZED SOLUTION IN NONLINEAR EQUATIONS THOUGH FUNDAMENTAL IS FAR MORE
MURKY &OR ELLIPTIC EQUATIONS THE SOLUTIONS DERIVED BY THE DIRECT VARIATIONAL
METHODS HAVE PROVED VERY USEFUL &OR NONLINEAR ONE DIMENSIONAL CONSER
3 +,!).%2-!. '!&!
VATION LAWS THE CONCEPT OF A GENERALIZED SOLUTION HAS BEEN DISCUSSED QUITE
EARLY IN THE WORKS OF ** 3TOKES SEE ;3T= 2ANKINE (UGONIOT 2IEMANN
ETC &OR HIGHER DIMENSIONAL EVOLUTION EQUATIONS THE lRST CONCEPT OF A WEAK
SOLUTION WAS INTRODUCED BY * ,ERAY ) CALL WEAK A GENERALIZED SOLUTION FOR
WHICH ONE CANNOT PROVE ANY TYPE OF UNIQUENESS 4HIS UNSATISFACTORY SITUA
TION MAY BE TEMPORARY DUE TO OUR TECHNICAL INABILITIES OR UNAVOIDABLE IN
THE SENSE THAT THE CONCEPT ITSELF IS mAWED ,ERAY WAS ABLE TO PRODUCE BY
A COMPACTNESS METHOD A WEAK SOLUTION OF THE INITIAL VALUE PROBLEM FOR THE
.AVIER[3TOKES EQUATIONS 4HE GREAT ADVANTAGE OF THE COMPACTNESS METHOD
AND ITS MODERN EXTENSIONS WHICH CAN IN SOME CASES CLEVERLY CIRCUMVENT
LACK OF COMPACTNESS IS THAT IT PRODUCES GLOBAL SOLUTIONS FOR ALL DATA 4HIS
IS PARTICULARLY IMPORTANT FOR SUPERCRITICAL OR CRITICAL NONLINEAR EVOLUTION
EQUATIONS WHERE WE EXPECT THAT CLASSICAL SOLUTIONS DEVELOP lNITE TIME SIN
GULARITIES 4HE PROBLEM HOWEVER IS THAT ONE HAS VERY LITTLE CONTROL OF THESE
SOLUTIONS IN PARTICULAR WE DONT KNOW HOW TO PROVE THEIR UNIQUENESS 3IM
ILAR TYPES OF SOLUTIONS WERE LATER INTRODUCED FOR OTHER IMPORTANT NONLINEAR
EVOLUTION EQUATIONS )N MOST OF THE INTERESTING CASES OF SUPERCRITICAL EVOLU
TION EQUATIONS SUCH AS .AVIER[3TOKES THE USEFULNESS OF THE TYPE OF WEAK
SOLUTIONS USED SO FAR REMAINS UNDECIDED
3CALING PROPERTIES AND CLASSIlCATION OF NONLINEAR EQUATIONS %SSENTIALLY
ALL BASIC NONLINEAR EQUATIONS HAVE WELL DElNED SCALING PROPERTIES 4HE RELA
TIONSHIP BETWEEN THE NONLINEAR SCALING AND THE COERCIVE A PRIORI ESTIMATES
OF THE EQUATIONS LEADS TO AN EXTREMELY USEFUL CLASSIlCATION BETWEEN SUBCRIT
ICAL CRITICAL AND SUPERCRITICAL EQUATIONS 4HE DElNITION OF CRITICALITY AND
ITS CONNECTION TO THE ISSUE OF REGULARITY WAS lRST UNDERSTOOD IN THE CASE
OF ELLIPTIC EQUATIONS SUCH AS (ARMONIC -APS THE EUCLIDEAN 9ANG[-ILLS OR
9AMABE PROBLEM 4HE SAME ISSUE APPEARS IN CONNECTION WITH GEOMETRIC
HEAT mOWS AND NONLINEAR WAVE EQUATIONS
'IVEN THAT SOME 0$%S ARE INTERESTING FROM A PURELY MATHEMATICAL
POINT OF VIEW WHILE OTHERS OWE THEIR RELEVANCE TO PHYSICAL THEORIES ONE
OF THE PROBLEMS WE FACE WHEN TRYING TO VIEW 0$% AS A COHERENT SUBJECT IS
THAT OF THE FUNDAMENTAL AMBIGUITY OF ITS STATUS IS IT PART OF -ATHEMATICS OR
0HYSICS OR BOTH )N THE NEXT SECTION ) WILL TRY TO BROADEN THE DISCUSSION BY
CONSIDERING SOME ASPECTS OF THE GENERAL RELATIONSHIP BETWEEN -ATHEMATICS

,ERAY WAS VERY CONCERNED ABOUT THIS POINT 4HOUGH LIKE ALL OTHER RESEARCHERS AFTER
HIM HE WAS UNABLE TO PROVE UNIQUENESS OF HIS WEAK SOLUTION HE SHOWED HOWEVER THAT IT
MUST COINCIDE WITH A CLASSICAL ONE AS LONG AS THE LATTER REMAINS SMOOTH

3EE THE SECTION <4HE 0ROBLEM OF "REAKDOWN FOR A MORE THOROUGH DISCUSSION
6ISIONS IN -ATH 0$% !3 ! 5.)&)%$ 35"*%#4
AND 0HYSICS RELEVANT TO OUR MAIN CONCERN ) WILL TRY TO ARGUE THAT WE CAN
REDRAW THE BOUNDARIES BETWEEN THE TWO SUBJECTS IN A WAY WHICH ALLOWS US
TO VIEW 0$% AS A CORE SUBJECT OF -ATHEMATICS WITH AN IMPORTANT APPLIED
COMPONENT )N THE THIRD SECTION ) WILL ATTEMPT TO SHOW HOW SOME OF THE
BASIC PRINCIPLES OF MODERN PHYSICS CAN HELP US ORGANIZE THE IMMENSE VARIETY
OF 0$%S INTO A COHERENT lELD %QUALLY IMPORTANT IN THE FOURTH SECTION
) WILL ATTEMPT TO SHOW THAT OUR MAIN 0$%S ARE NOT ONLY RELATED THROUGH
THEIR DERIVATION THEY ALSO SHARE A COMMON FUNDAMENTAL PROBLEM REGULARITY
OR BREAKDOWN ) HAVE TRIED TO KEEP THE DISCUSSION OF THE lRST FOUR SECTIONS
AS GENERAL AS POSSIBLE AND HAVE THUS AVOIDED GIVING MORE THAN JUST A FEW
REFERENCES ) APOLOGIZE TO ALL THOSE WHO FEEL THAT THEIR CONTRIBUTIONS ALLUDED
TO IN MY TEXT SHOULD HAVE BEEN PROPERLY MENTIONED )N THE LAST SECTION
OF THE PAPER ) CONCENTRATE ON A TOPIC OF PERSONAL RESEARCH INTEREST TIED
TO THE ISSUE OF REGULARITY CONCERNING THE PROBLEM OF WELL POSEDNESS FOR
NONLINEAR WAVE EQUATIONS -Y MAIN GOAL HERE IS TO DISCUSS THREE PRECISE
CONJECTURES WHICH ) FEEL ARE IMPORTANT DInCULT AND ACCESSIBLE TO GENERATE
FUTURE DEVELOPMENTS IN THE lELD %VEN IN THIS SECTION HOWEVER ) ONLY
PROVIDE FULL REFERENCES TO WORKS DIRECTLY CONNECTED TO THESE CONJECTURES
-ANY OF THE IMPORTANT POINTS ) MAKE BELOW SUCH AS THE UNIlED GEOMET
RIC STRUCTURE OF THE MAIN 0$%S THE IMPORTANCE OF THE SCALING PROPERTIES
OF THE EQUATIONS AND ITS CONNECTION TO REGULARITY AND WELL POSEDNESS HAVE
BEEN DISCUSSED IN SIMILAR WAYS BEFORE AND ARE SHARED BY MANY OF MY FRIENDS
AND COLLABORATORS -Y ONLY CLAIM TO ORIGINALITY IN THIS REGARD IS THE FORM
IN WHICH ) HAVE ASSEMBLED THEM 4HE IMPERFECTIONS ERRORS AND OMISSIONS
ARE CERTAINLY MY OWN
) WOULD LIKE TO THANK MY FRIENDS ( "RsEZIS ! #HANG $ #HRISTODOULOU
# $AFERMOS 0 $EIFT 7EINAN % ' (UISKEN * +OHN % 3TEIN 0 3ARNAK
9 3INAI - 3TRUWE * 3TALKER AND MY WIFE !NCA FOR READING PREVIOUS
VERSIONS OF THE PAPER AND SUGGESTING MANY CORRECTIONS AND IMPROVEMENTS
"ETWEEN -ATHEMATICS AND 0HYSICS

)N SEARCH OF A UNIlED POINT OF VIEW FOR OUR SUBJECT IT PAYS TO LOOK AT THE
BROADER PROBLEM OF -ATHEMATICS AS A WHOLE )SNT -ATHEMATICS ALSO IN
DANGER OF BECOMING A LARGE COLLECTION OF LOOSELY CONNECTED SUBJECTS /UR
CHERISHED INTELLECTUAL FREEDOM TO PURSUE WHATEVER PROBLEMS STRIKE OUR IMAG
INATION AS WORTHWHILE IS A GREAT ENGINE OF INVENTION BUT IN THE ABSENCE
OF UNIFYING GOALS IT SEEMS TO LEAD TO AN ENDLESS PROLIFERATION OF SUBJECTS
3 +,!).%2-!. '!&!
4HIS IS PRECISELY ) BELIEVE WHAT 0OINCARsE ;0= HAD IN MIND IN THE FOLLOW
ING PASSAGE CONTAINED IN HIS ADDRESS TO THE lRST )NTERNATIONAL #ONGRESS OF
-ATHEMATICIANS MORE THAN A HUNDRED YEARS AGO
< 4HE COMBINATIONS THAT CAN BE FORMED WITH NUMBERS AND SYMBOLS
ARE AN INlNITE MULTITUDE )N THIS THICKET HOW SHALL WE CHOOSE THOSE THAT ARE
WORTHY OF OUR ATTENTION 3HALL WE BE GUIDED ONLY BY WHIMSY ;4HIS=
WOULD UNDOUBTEDLY CARRY US FAR FROM EACH OTHER AND WE WOULD RAPIDLY
CEASE TO UNDERSTAND EACH OTHER "UT THAT IS ONLY THE MINOR SIDE OF THE
PROBLEM .OT ONLY WILL PHYSICS PERHAPS PREVENT US FROM GETTING LOST BUT
IT WILL ALSO PROTECT US FROM A MORE FEARSOME DANGER TURNING AROUND
FOREVER IN CIRCLES (ISTORY ;SHOWS THAT= PHYSICS HAS NOT ONLY FORCED US TO
CHOOSE ;FROM THE MULTITUDE OF PROBLEMS WHICH ARISE= BUT IT HAS ALSO IMPOSED
ON US DIRECTIONS THAT WOULD NEVER HAVE BEEN DREAMED OF OTHERWISE
7HAT COULD BE MORE USEFUL
4HE FULL TEXT OF ;0= IS A MARVELOUS ANALYSIS OF THE COMPLEX INTERACTIONS
BETWEEN -ATHEMATICS AND 0HYSICS 0OINCARsE ARGUES NOT ONLY THAT 0HYSICS
PROVIDES US WITH A GREAT SOURCE OF INSPIRATION AND COHESIVENESS BUT THAT
ITSELF IN RETURN OWES ITS LANGUAGE SENSE OF BEAUTY AND ORDER TO -ATHEMAT
ICS 9ET 0OINCARsES VIEWPOINT CONCERNING THE IMPORTANCE OF CLOSE RELATIONS
WITH 0HYSICS WAS LARGELY IGNORED DURING MOST OF THIS CENTURY BY A LARGE SEG
MENT OF THE MATHEMATICAL COMMUNITY /NE REASON IS CERTAINLY DUE TO THE
FACT THAT TRADITIONAL AREAS OF -ATHEMATICS SUCH AS !LGEBRA .UMBER 4HE
ORY AND 4OPOLOGY HAVE OR SEEMED TO HAVE RELATIVELY LITTLE TO GAIN FROM
DIRECT INTERACTIONS WITH 0HYSICS !NOTHER MORE SUBTLE REASON MAY HAVE TO
DO WITH THE REMARKABLE AND UNEXPECTED EkECTIVENESS OF PURE MATHEMATICAL
STRUCTURES IN THE FORMULATION OF THE MAJOR PHYSICAL THEORIES OF THE CENTURY
3PECIAL AND 'ENERAL 2ELATIVITY 1UANTUM -ECHANICS AND 'AUGE 4HEORIES
4HIS HAS LED TO THE POPULAR POINT OF VIEW COINED BY 7IGNER ;7I= AS <4HE
UNREASONABLE EkECTIVENESS OF -ATHEMATICS ACCORDING TO WHICH MATHEMAT
ICAL OBJECTS OR IDEAS DEVELOPED ORIGINALLY WITHOUT ANY REFERENCE TO 0HYSICS
TURN OUT TO BE AT THE HEART OF SOLUTIONS TO DEEP PHYSICAL PROBLEMS %INSTEIN
HIMSELF WROTE THAT ANY IMPORTANT ADVANCE IN 0HYSICS WILL HAVE TO COME IN
THE WAKE OF MAJOR NEW DEVELOPMENTS IN -ATHEMATICS 4HIS VERY SEDUCTIVE
PICTURE HAS EMBOLDENED US MATHEMATICIANS TO BELIEVE THAT ANYTHING WE DO
MAY TURN OUT EVENTUALLY TO HAVE REAL APPLICATIONS AND HAS THUS PARADOX
ICALLY CONTRIBUTED TO THE PROBLEM OF IGNORING THE PHYSICAL WORLD 0OINCARsE

4HE SITUATION HAS CHANGED DRAMATICALLY IN THE LAST YEARS WITH THE ADVENT OF 'AUGE
lELDS AND 3TRING 4HEORIES
6ISIONS IN -ATH 0$% !3 ! 5.)&)%$ 35"*%#4
HAS WARNED US AGAINST

"UT THIS IS ONLY A MINOR PARADOX BY COMPARISON TO THE ONE WHICH SEEMS
TO ARISE FROM THE ABOVE DISCUSSION RELATIVE TO THE REMARKABLE SYMBIOSIS BE
TWEEN 0HYSICS AND -ATHEMATICS /N ONE HAND AS 0OINCARsE ARGUES VERY
CONVINCINGLY IN ;0= -ATHEMATICS NEEDS TO KEEP ITSELF TOGETHER UNIFYING
GOALS AND PRINCIPLES 0HYSICS DUE ) GUESS TO THE PERCEIVED UNITY OF THE
0HYSICAL 7ORLD IS IN A PERFECT POSITION TO PROVIDE THEM FOR US /N THE
OTHER HAND 0HYSICS OWES TO -ATHEMATICS THE VERY TOOLS WHICH MAKES IT
POSSIBLE TO UNCOVER AND FORMULATE THE UNIlED FEATURES OF PHYSICAL REALITY
IT IS INDEED THE SEARCH FOR A SELFCONSISTENT MATHEMATICAL FORMALISM WHICH
SEEMS TO BE AT THE CORE OF THE CURRENT ATTEMPTS TO lND THAT UNIlED THE
ORY OF EVERYTHING WHICH AS THEORETICAL PHYSICISTS OFTEN DECLARE IS 0HYSICS
ULTIMATE GOAL 4HE PARADOX IS DUE OF COURSE TO THE ARTIlCIAL DISTINCTIONS
WE MAKE BETWEEN THE TWO SUBJECTS 7E IMAGINE THEM AS SEPARATED WHEN
IN FACT THEY HAVE A NONTRIVIAL INTERSECTION #AN WE IDENTIFY THAT INTERSEC
TION 4HE NAIVE PICTURE WOULD BE OF TWO SETS WHICH INTERSECT IN AN AREA
SOMEWHAT PERIPHERAL TO BOTH WHICH WE MIGHT CALL -ATHEMATICAL 0HYSICS
"UT THIS PICTURE DOES NOT HELP TO SOLVE THE PARADOX WE HAVE MENTIONED
ABOVE WHICH CONCERNS THE CORE OF BOTH SUBJECTS ! CENTRAL INTERSECTION
HOWEVER COULD IMPLY SOME FORM OF EQUALITY OR INCLUSION BETWEEN THE TWO
SUBJECTS WHICH IS DElNITELY NOT THE CASE -ATHEMATICS PURSUES GOALS WHICH
ARE NOT NECESSARILY SUGGESTED BY THE PHYSICAL SCIENCES ! RESEARCH DIRECTION
IS DEEMED IMPORTANT BY MATHEMATICIANS IF IT LEADS TO ELEGANT DEVELOPMENTS
AND UNEXPECTED CONNECTIONS 0HYSICS ON THE OTHER HAND CANNOT ALLOW IT
SELF THE LUXURY OF BEING CARRIED AWAY BY ELEGANT MATHEMATICAL THEORIES IN
THE lNAL ANALYSIS IT HAS TO SUBJECT ITSELF TO THE TOUGH TEST OF REAL EXPERI
MENTS -OREOVER THE DIkERENCE BETWEEN THE WORK PRACTICE AND PROFESSIONAL
STANDARDS OF MATHEMATICIANS AND MODERN THEORETICAL PHYSICISTS CANNOT BE
MORE STRIKING 7E MATHEMATICIANS lND OURSELVES CONSTRAINED BY RIGOR AND
ARE OFTEN RELUCTANT TO PROCEED WITHOUT A SYSTEMATIC ANALYSIS OF ALL OBSTACLES
IN OUR PATH )N THEIR QUEST FOR THE ULTIMATE TRUTH THEORETICAL PHYSICISTS HAVE
NO TIME TO WASTE ON UNEXPECTED HURDLES AND UNPROMISING TERRITORY #LEARLY
THE RELATIONSHIP BETWEEN THE TWO SUBJECTS IS FAR MORE COMPLEX THAN MAY
SEEM AT lRST GLANCE
4HE TASK OF DElNING 0$% AS A UNIlED SUBJECT IS TIED TO THAT OF CLAR
IFYING SOMEHOW THIS AMBIGUOUS RELATIONSHIP BETWEEN -ATHEMATICS AND
0HYSICS 4HE VERY CONCEPT OF PARTIAL DIkERENTIAL EQUATIONS HAS ITS ROOTS IN
0HYSICS OR MORE APPROPRIATELY -ATHEMATICAL 0HYSICS THERE WERE NO CLEAR
3 +,!).%2-!. '!&!
DISTINCTIONS AT THE TIME OF $!LEMBERT %ULER 0OISSON ,APLACE BETWEEN

THE TWO SUBJECTS 2IEMANN WAS THE lRST ) BELIEVE TO SHOW HOW ONE CAN USE
0$%S TO ATTACK PROBLEMS CONSIDERED PURE MATHEMATICAL IN NATURE SUCH AS
CONFORMAL MAPPINGS IN #OMPLEX !NALYSIS 4HE REMARKABLE EkECTIVENESS
OF 0$%S AS A TOOL TO SOLVE PROBLEMS IN #OMPLEX !NALYSIS 'EOMETRY AND
4OPOLOGY HAS BEEN CONlRMED MANY TIMES DURING THIS CENTURY
/NE CAN SEPARATE ALL MATHEMATICIANS AND OTHER SCIENTISTS CONCERNED WITH
THE STUDY OF 0$%S INTO FOUR GROUPS ACCORDING TO THEIR MAIN INTERESTS )N
THE lRST GROUP ) INCLUDE THOSE DEVELOPING AND USING 0$% METHODS TO ATTACK
PROBLEMS IN $IkERENTIAL 'EOMETRY #OMPLEX !NALYSIS 3YMPLECTIC 'EOME
TRY 4OPOLOGY AND !LGEBRAIC 'EOMETRY )N THE SECOND ) INCLUDE THOSE WHOSE
MAIN MOTIVATION IS THE DEVELOPMENT OF RIGOROUS MATHEMATICAL METHODS TO
DEAL WITH THE 0$%S ARISING IN THE PHYSICAL THEORIES )N THE THIRD GROUP )
INCLUDE MATHEMATICIANS PHYSICISTS OR ENGINEERS INTERESTED IN UNDERSTAND
ING THE MAIN CONSEQUENCES OF THE PHYSICAL THEORIES GOVERNED BY 0$%S
USING A VARIETY OF HEURISTIC COMPUTATIONAL OR EXPERIMENTAL METHODS )T IS
ONLY FAIR TO DElNE YET A FOURTH CATEGORY WHICH INCLUDE ALL THOSE LEFT OUT OF
THE GROUPS DElNED ABOVE !CCORDING TO THE COMMON PRECONCEPTIONS ABOUT
THE PROPER DELIMITATIONS BETWEEN -ATHEMATICS AND 0HYSICS ONLY THE lRST
GROUP BELONGS UNAMBIGUOUSLY WITHIN -ATHEMATICS 4HE THIRD GROUP IS CON
SIDERED CORRECTLY IN MY VIEW AS BELONGING EITHER TO !PPLIED -ATHEMATICS
OR !PPLIED 0HYSICS 4HE SECOND GROUP HOWEVER HAS AN ILL DElNED IDENTITY
3INCE THE ULTIMATE GOALS ARE NOT DIRECTLY CONNECTED TO SPECIlC APPLICATIONS
TO THE TRADITIONAL BRANCHES OF -ATHEMATICS MANY VIEW THIS GROUP AS PART
OF EITHER !PPLIED -ATHEMATICS OR -ATHEMATICAL 0HYSICS 9ET APART FROM
THE ORIGINAL MOTIVATION IT IS HARD TO DISTINGUISH THE SECOND GROUP FROM THE
lRST "OTH GROUPS ARE DEDICATED TO THE DEVELOPMENT OF RIGOROUS ANALYTIC
TECHNIQUES 4HEY ARE TIED BY MANY SIMILAR CONCERNS CONCEPTS AND METHODS
4HEY ARE BOTH INTIMATELY TIED TO SUBJECTS CONSIDERED PURE MAINLY 2EAL AND
&OURIER !NALYSIS BUT ALSO 'EOMETRY 4OPOLOGY AND !LGEBRAIC 'EOMETRY
)N VIEW OF THE ABOVE AMBIGUITIES IT HELPS TO TAKE A CLOSER LOOK AT THE ROLE
PLAYED BY -ATHEMATICS IN DEVELOPING THE CONSEQUENCES OF THE ESTABLISHED

-Y CLASSIlCATION IS MAINLY RHETORICAL 4HERE ARE OF COURSE MANY MATHEMATICIANS
WHO CAN CROSS THESE ARTIlCIAL BOUNDARIES ) WILL IN FACT ARGUE BELOW THAT THE lRST TWO
GROUPS SHOULD BE VIEWED AS ONE

4HIS INCLUDES IN PARTICULAR 0$%S APPEARING IN "IOLOGY OR %CONOMICS %XOTIC
0$%S NOT NECESSARILY CONNECTED WITH ANY SPECIlC APPLICATION SHOULD ALSO BE INCLUDED
IN THIS CLASS
6ISIONS IN -ATH 0$% !3 ! 5.)&)%$ 35"*%#4
PHYSICAL THEORIES ) HAVE HEARD THEORETICAL PHYSICISTS AND ALSO ALAS MATH
EMATICIANS EXPRESSING THE VIEW THAT THE CONSEQUENCES OF AN ESTABLISHED
PHYSICAL THEORY ARE OF LESSER IMPORTANCE AND MAY PROPERLY BE RELEGATED
TO %NGINEERING OR #HEMISTRY .OTHING IN MY MIND CAN BE FURTHER FROM
THE TRUTH 4HE lRST SUCCESSFUL PHYSICAL THEORY THAT OF SPACE WAS WRITTEN
DOWN BY %UCLID MORE THAN TWO THOUSAND YEARS AGO 5NDOUBTEDLY %U
CLIDEAN 'EOMETRY WAS USED BY ENGINEERS TO DESIGN LEVERS PULLEYS AND MANY
OTHER MARVELOUS APPLICATIONS BUT DOES ANYBODY VIEW THE FURTHER DEVELOP
MENT OF THE SUBJECT AS %NGINEERING 'EOMETRY IS THE PRIMARY EXAMPLE OF
A <PHYSICAL THEORY DEVELOPED FOR CENTURIES AS A PURE MATHEMATICAL DISCI
PLINE WITHOUT TOO MUCH NEW INPUT FROM THE PHYSICAL WORLD WHICH GREW TO
HAVE DEEP MYSTERIOUS COMPLETELY UNEXPECTED CONSEQUENCES TO THE POINT
THAT PRE EMINENT PHYSICISTS TALK TODAY OF A COMPLETE <GEOMETRIZATION OF
MODERN PHYSICS SEE ;.E=
"UT THIS IS NOT ALL THE 0RINCIPLE OF ,EAST !CTION WAS DEVELOPED BY MATH
EMATICIANS SUCH AS &ERMAT ,EIBNITZ -AUPERTIUS THE "ERNOULLI BROTHERS
AND %ULER FROM THE ANALYSIS OF SIMPLE GEOMETRIC AND PHYSICAL PROBLEMS SEE
;(4= FOR A VERY GOOD PRESENTATION OF THE EARLY HISTORY OF THE PRINCIPLE
4HEIR WORK LED TO A COMPREHENSIVE REFORMULATION OF THE LAWS OF -ECHANICS
BY ,AGRANGE WHO SHOWED HOW TO DERIVE THEM FROM A SIMPLE 6ARIATIONAL
0RINCIPLE 4ODAY THE ,AGRANGIAN POINT OF VIEW TOGETHER WITH ITS (AMIL
TONIAN REFORMULATION AND THE FAMOUS RESULT OF % .OETHER CONCERNING THE
RELATION BETWEEN THE SYMMETRIES OF THE ,AGRANGIAN AND CONSERVATION LAWS
IS A FOUNDATIONAL PRINCIPLE FOR ALL 0HYSICS #ONNECTED TO THESE ARE THE CON
TINUOUS GROUPS OF SYMMETRIES ATTACHED TO THE NAME OF 3 ,IE
&OURIER !NALYSIS WAS INITIATED IN WORKS BY $!LEMBERT %ULER AND
$ "ERNOULLI IN CONNECTION WITH THE STUDY OF THE INITIAL BOUNDARY VALUE
PROBLEM FOR THE ONE DIMENSIONAL WAVE EQUATION VIBRATING STRING "ER
NOULLIS IDEA OF APPROXIMATING GENERAL PERIODIC FUNCTIONS BY SUMS OF SINES
AND COSINES WAS LATER DEVELOPED BY * &OURIER IN CONNECTION WITH THE (EAT
%QUATION &URTHER MATHEMATICAL DEVELOPMENTS MADE THE THEORY INTO A FUN

) DISTINGUISH BETWEEN THE QUEST TO UNCOVER THE BASIC LAWS OF .ATURE WHICH DElNES
THE CORE OF THEORETICAL 0HYSICS AND THE SCIENTIlC ACTIVITIES CONCERNED WITH DERIVING THE
CONSEQUENCES OF A GIVEN ESTABLISHED THEORY WHICH INVOLVE APPLIED PHYSICISTS ENGINEERS
CHEMISTS APPLIED MATHEMATICIANS AND AS ) ARGUE BELOW <PURE MATHEMATICIANS .EED
LESS TO SAY MATHEMATICIANS HAVE OFTEN HAD DIRECT FUNDAMENTAL CONTRIBUTIONS TO THEO
RETICAL 0HYSICS "UT MORE OFTEN ) BELIEVE THE MOST IMPRESSIVE CONTRIBUTIONS CAME FROM
INNER DEVELOPMENTS WITHIN -ATHEMATICS OF SUBJECTS WITH DEEP ROOTS IN THE PHYSICAL WORLD
SUCH AS 'EOMETRY .EWTONIAN -ECHANICS %LECTROMAGNETISM 1UANTUM -ECHANICS ETC
3 +,!).%2-!. '!&!
DAMENTAL TOOL THROUGHOUT ALL OF 3CIENCE

4HERE ARE PLENTY OF OTHER EXAMPLES ) SUSPECT THAT MANY IF NOT MOST OF
THE EXAMPLES OF THE<UNREASONABLE EkECTIVENESS OF MATHEMATICS ARE IN FACT
OF THIS TYPE 4HERE ARE ALSO MANY OTHER EXAMPLES OF IDEAS WHICH ORIGINATE
IN -ATHEMATICAL 0HYSICS AND TURN OUT TO HAVE A DEEP MYSTERIOUS IMPACT
ON THE TRADITIONAL SUBJECTS OF -ATHEMATICS SUCH AS 4OPOLOGY 'EOMETRY OR
EVEN .UMBER 4HEORY
!LL THIS SEEMS TO POINT TO THE FACT THAT THE FURTHER DEVELOPMENT OF THE
ESTABLISHED PHYSICAL THEORIES OUGHT TO BE VIEWED AS A GENUINE AND CENTRAL
GOAL OF -ATHEMATICS ITSELF )N VIEW OF THIS ) THINK WE NEED TO REEVALUATE OUR
CURRENT PRECONCEPTION ABOUT WHAT SUBJECTS WE CONSIDER AS BELONGING PROP
ERLY WITHIN -ATHEMATICS 7E MAY GAIN CONSISTENT WITH 0OINCARsES POINT OF
VIEW CONSIDERABLY MORE UNITY BY ENLARGING THE BOUNDARIES OF -ATHEMATICS
TO INCLUDE ON EQUAL FOOTING WITH ALL OTHER MORE TRADITIONAL lELDS PHYSICAL
THEORIES SUCH AS #LASSICAL AND 1UANTUM -ECHANICS AND 2ELATIVITY 4HEORY
WHICH ARE EXPRESSED IN CLEAR AND UNAMBIGUOUS MATHEMATICAL LANGUAGE 7E
MAY THEN DEVELOP THEM IF WE WISH ON PURE MATHEMATICAL TERMS ASKING
QUESTIONS WE CONSIDER FUNDAMENTAL WHICH MAY NOT COINCIDE AT ANY GIVEN
MOMENT WITH THOSE PHYSICISTS ARE MOST INTERESTED IN AND PROVIDING FULL
RIGOR TO OUR PROOFS /F COURSE THIS HAS HAPPENED TO A CERTAIN EXTENT -ATH
EMATICAL 0HYSICS AND PARTS OF !PPLIED -ATHEMATICS FULlLL PRECISELY THIS
ROLE 9ET THEIR STATUS REMAINS AMBIGUOUS AND SOMEWHAT PERIPHERAL -ANY
MATHEMATICIANS ASSUME THAT SUBJECTS LIKE #LASSICAL 'ENERAL 2ELATIVITY OR

! CLEAR EXAMPLE OF THIS TYPE THIS CENTURY IS THE DISCOVERY OF THE SOLITON AND THE
<INTEGRABLE METHOD 4HOUGH BOTH EMERGED IN CONNECTION WITH SIMPLE NONLINEAR PARTIAL
DIkERENTIAL EQUATIONS THE INTEGRABILITY METHOD HAS FOUND DEEP APPLICATIONS WAY BEYOND
THE ORIGINAL 0$% CONTEXT 4HERE ARE OTHER EXAMPLES WHICH DO NOT QUITE lT INTO MY
DESCRIPTION 4HE EXTRAORDINARY ROLE PLAYED BY COMPLEX NUMBERS IN THE FORMULATION OF
1UANTUM -ECHANICS IS CERTAINLY ONE WHICH HAS ITS ROOTS IN !LGEBRA RATHER THAN 'EOMETRY
OR -ATHEMATICAL 0HYSICS

4HE FORMULATION OF 'ENERAL 2ELATIVITY BY ! %INSTEIN FOLLOWING THE WORK OF 'AUSS
AND 2IEMANN IN 'EOMETRY AND THAT OF ,ORENTZ 0OINCARsE %INSTEIN AND -INKOWSKI ON
SPECIAL RELATIVITY CAN BE VIEWED AS ONE OF THE MOST IMPRESSIVE TRIUMPHS OF -ATHEMATICS
&OLLOWING THE RECENT EXPERIMENTS WITH DOUBLE PULSARS '2 IS CONSIDERED THE MOST ACCU
RATE OF ALL PHYSICAL THEORIES 2ESEARCH IN 'ENERAL 2ELATIVITY INVOLVES IN A FUNDAMENTAL
WAY ALL ASPECTS OF TRADITIONAL MATHEMATICS $IkERENTIAL 'EOMETRY !NALYSIS 4OPOLOGY
'ROUP 2EPRESENTATION $YNAMICAL SYSTEMS AND OF COURSE 0$%S !SSUMING THAT THE
FURTHER DEVELOPMENT OF THE SUBJECT IS COVERED BY PHYSICS DEPARTMENTS IS MISLEADING MOST
THEORETICAL PHYSICISTS VIEW CLASSICAL '2 AS A COMPLETELY UNDERSTOOD PHYSICAL THEORY THEIR
MAIN GOAL NOW IS TO DEVELOP A QUANTUM THEORY OF GRAVITY 'IVEN THEIR LACK OF INTEREST
AND THE RICH MATHEMATICAL CONTENT OF THE SUBJECT IS THERE ANY REASON WHY WE SHOULD NOT
6ISIONS IN -ATH 0$% !3 ! 5.)&)%$ 35"*%#4
1UANTUM -ECHANICS BELONG PROPERLY TO 0HYSICS DEPARTMENTS WHILE 0HYSI

CISTS OFTEN CONSIDER THEM AS PERFECTLY WELL UNDERSTOOD CLOSED SUBJECTS
4HEY ARE INDEED CLOSED OR SO IT SEEMS IN SO FAR AS THEORETICAL PHYSICISTS ARE
CONCERNED &ROM THEIR PERSPECTIVE 'EOMETRY MAY HAVE BECOME A CLOSED
BOOK MORE THAN YEARS AGO WITH THE PUBLICATION OF %UCLIDS %LEMENTS
"UT THEY PRESENT US MATHEMATICIANS WITH WONDERFUL FUNDAMENTAL CHAL
LENGES FORMULATED IN THE PUREST MATHEMATICAL LANGUAGE 3HOULD WE RELEGATE
SUBJECTS SUCH AS #LASSICAL AND 1UANTUM -ECHANICS OR 'ENERAL 2ELATIVITY
TO THE PERIPHERY OF -ATHEMATICS DESPITE THEIR WELL DElNED AND RICH MATHE
MATICAL STRUCTURES ONLY BECAUSE THEY HAPPEN TO DESCRIBE IMPORTANT ASPECTS
OF THE PHYSICAL WORLD )S IT REASONABLE TO HESITATE TO INCLUDE 'ENERAL 2EL
ATIVITY AS A SUBJECT OF -ATHEMATICS SIMPLY BECAUSE IT CONCERNS ITSELF WITH
,ORENTZ RATHER THAN 2IEMANNIAN METRICS /R BECAUSE IT DOES NOT SEEM TO
HAVE ANY APPLICATIONS TO 4OPOLOGY 4HERE ARE IN FACT PROPOSALS TO TIE '2
TO THE GEOMETRIZATION CONJECTURE OF $ MANIFOLDS SEE ;&-=
-Y PROPOSAL IS NOT JUST TO ACCEPT THESE DISCIPLINES AS SOME APPLIED AP
PENDICES TO PURE -ATHEMATICS BUT TO GIVE THEM THE CENTRAL ROLE THEY
DESERVE 4HIS WOULD FORCE US TO BROADEN OUR OUTLOOK AND WOULD GIVE US
FRESH ENERGY AND COHESION IN THE SPIRIT ENVISIONED BY 0OINCARsE )T WOULD
HELP US IN PARTICULAR TO CLARIFY THE AMBIGUOUS STATUS OF SUBJECTS SUCH AS
0$%S AND -ATHEMATICAL 0HYSICS AND THEIR RELATIONS WITH !PPLIED -ATH
EMATICS )T WOULD ALSO SET MORE NATURAL BOUNDARIES BETWEEN -ATHEMATICS
AND 0HYSICS !S THEORETICAL PHYSICISTS ARE PRIMARILY INTERESTED IN UNDER
STANDING NEW PHYSICAL PHENOMENA THE FURTHER MATHEMATICAL DEVELOPMENTS
OF A CONlRMED PHYSICAL THEORY BECOMES ONE OF OUR TASKS 4HOUGH OUR
PURE MATHEMATICAL CONSIDERATIONS MAY LEAD US INTO SEEMINGLY ESOTERIC DI
RECTIONS WE SHOULD HOLD OUR GROUND FOR WITH TIME PHYSICISTS MAY COME TO
ADMIT ONCE MORE TO THE UNEXPECTED EkECTIVENESS OF OUR 3CIENCE
&INALLY ) WANT TO DISTINGUISH MY PROPOSAL FROM ANOTHER MORE RADICAL
POINT OF VIEW DISCUSSED IN THIS CONFERENCE ACCORDING TO WHICH -ATHEMAT
ICS OUGHT TO BECOME FULLY ENGAGED WITH THE GREAT PROBLEMS OF #HEMISTRY
"IOLOGY #OMPUTING %CONOMICS AND %NGINEERING 4HOUGH ) STRONGLY SUS
PECT THAT ONE DAY SOME STILL TO BE DISCOVERED DEEP MATHEMATICAL STRUCTURE
TAKE THE OPPORTUNITY AND EMBRACE IT FULLY AS OUR OWN

!N EASY STEP WHICH WILL GO A LONG WAY IN THIS DIRECTION WOULD BE TO ADD AS A REQUIRE
MENT FOR MATHEMATICS MAJORS OR GRADUATE STUDENTS A COURSE CONTAINING A COMPREHENSIVE
DISCUSSION OF THE MATHEMATICAL STRUCTURES WHICH UNDERLY THE MAIN PHYSICAL THEORIES

4HIS DOES NOT EXCLUDE THE POSSIBILITY THAT THE SAME SUBJECT MAY BE PURSUED IN
DIkERENT WAYS IN BOTH -ATHEMATICS AND 0HYSICS DEPARTMENTS
3 +,!).%2-!. '!&!
WILL HELP EXPLAIN SOME OF THE IMPORTANT FEATURES OF COMPLEX BIOLOGICAL SYS
TEMS WE ARE VERY FAR FROM THAT )T IS CERTAINLY TO BE HOPED THAT INDIVIDUAL
MATHEMATICIANS WILL MAKE SIGNIlCANT CONTRIBUTIONS TO THESE lELDS BUT IT IS
UNREALISTIC TO THINK THAT -ATHEMATICS CAN FULLY EMBRACE THESE AREAS WHILE
MAINTAINING ITS INNER CONTINUITY COHERENCE AND FUNDAMENTAL COMMITMENT
TO RIGOR 7E HAVE TO DISTINGUISH BETWEEN THE CORE OF -ATHEMATICS WHERE
) BELIEVE THE BASIC PHYSICAL THEORIES OUGHT TO BELONG AND VARIOUS PROBLEMS
OF 3CIENCE AND %NGINEERING WHERE MATHEMATICIANS CAN PLAY A VERY USEFUL
ROLE
4HE -AIN %QUATIONS

4O RETURN TO 0$% ) WANT TO SKETCH A WAY OF LOOKING AT THE SUBJECT FROM
SIMPLE lRST PRINCIPLES WHICH HAPPEN TO COINCIDE WITH SOME OF THE UNDERLYING
GEOMETRIC PRINCIPLES OF MODERN 0HYSICS )T TURNS OUT THAT MOST OF OUR BASIC
0$%S CAN BE DERIVED IN THIS FASHION 4HUS THE MAIN OBJECTS OF OUR SUB
JECT TURN OUT TO BE IN NO WAY LESS <PURE MATHEMATICAL IN NATURE THAN THE
OTHER FUNDAMENTAL OBJECTS STUDIED BY MATHEMATICIANS NUMBERS FUNC
TIONS AND VARIOUS TYPES OF ALGEBRAIC AND GEOMETRIC STRUCTURES "UT MOST
IMPORTANTLY THESE SIMPLE PRINCIPLES PROVIDE A UNIFYING FRAMEWORK FOR OUR
SUBJECT AND THUS HELP ENDOW IT WITH A SENSE OF PURPOSE AND COHESION )T ALSO
EXPLAINS WHY A VERY SMALL NUMBER OF LINEAR DIkERENTIAL OPERATORS SUCH AS
THE ,APLACIAN AND $!LEMBERTIAN ARE ALL PERVASIVE THEY ARE THE SIMPLEST
APPROXIMATIONS TO EQUATIONS NATURALLY TIED TO THE TWO MOST FUNDAMENTAL
GEOMETRIC STRUCTURES %UCLIDEAN AND -INKOWSKIAN 4HE (EAT EQUATION IS
THE SIMPLEST PARADIGM FOR DIkUSIVE PHENOMENA WHILE THE 3CHR ODINGER EQUA

3OME PURE MATHEMATICIANS DISTRUST THE BASIC PHYSICAL 0$%S AS PROPER OBJECTS OF
-ATHEMATICS ON THE SPURIOUS NOTION THAT THEY ARE JUST IMPERFECT APPROXIMATIONS TO
AN ULTIMATE PHYSICAL REALITY OF WHICH WE ARE STILL IGNORANT /N THE BASIS OF THIS ANALYSIS
GROUPS # b ALGEBRAS TOPOLOGICAL VECTOR SPACES OR THE OPERATOR ARE PERFECT MATHEMATICAL
OBJECTS AS LONG AS THEY HAVE NO DIRECT RELATIONS TO 0HYSICS WHILE (AMILTONIAN SYSTEMS
THE -AXWELL %ULER 3CHR ODINGER AND %INSTEIN EQUATIONS ARE NOT

4HE SCHEME ) PRESENT BELOW IS ONLY AN ATTEMPT TO SHOW THAT IN SPITE OF THE ENOR
MOUS NUMBER OF 0$%S STUDIED BY MATHEMATICIANS PHYSICISTS AND ENGINEERS THERE ARE
NEVERTHELESS SIMPLE BASIC PRINCIPLES WHICH UNITE THEM ) DONT WANT BY ANY MEANS TO
IMPLY THAT THE EQUATIONS DISCUSSED BELOW ARE THE ONLY ONES WORTHY OF OUR ATTENTION )T
WOULD BE ALSO FOOLISH TO PRESUME THAT WE CAN PREDICT WHICH 0$%S ARE GOING TO LEAD TO
THE MOST INTERESTING DEVELOPMENTS #ERTAINLY NOBODY COULD HAVE PREDICTED YEARS AGO
THE EMERGENCE ON THE SCENE OF THE %INSTEIN AND 9ANG[-ILLS EQUATIONS OR THE REMARKABLE
MATHEMATICAL STRUCTURE BEHIND THE SEEMINGLY PEDESTRIAN +D6 EQUATION
6ISIONS IN -ATH 0$% !3 ! 5.)&)%$ 35"*%#4
TION CAN BE VIEWED AS THE .EWTONIAN LIMIT OF A LOWER ORDER PERTURBATION

OF THE $!LEMBERTIAN 4HE GEOMETRIC FRAMEWORK OF THE FORMER IS 'ALILEAN
SPACE WHICH ITSELF IS SIMPLY THE .EWTONIAN LIMIT OF THE -INKOWSKI SPACE
SEE ;-=
3TARTING WITH THE %UCLIDEAN SPACE 2N THE ,APLACIAN a IS THE SIMPLEST
DIkERENTIAL OPERATOR INVARIANT UNDER THE GROUP OF ISOMETRIES OR RIGID TRANS
FORMATIONS OF 2N 4HE HEAT 3CHRODINGER AND WAVE OPERATORS T _ a

I T _ a AND T _ a ARE THE SIMPLEST EVOLUTION OPERATORS WHICH WE CAN
FORM USING a 4HE WAVE OPERATOR c _ T a HAS A DEEPER MEANING
IT IS ASSOCIATED TO THE -INKOWSKI SPACE 2N IN THE SAME WAY THAT a IS
ASSOCIATED TO 2N -OREOVER THE SOLUTIONS TO THE EQUATION a} CAN BE
VIEWED AS SPECIAL TIME INDEPENDENT SOLUTIONS TO c} 4HE 3CHR ODINGER
EQUATION CAN ALSO BE OBTAINED BY A SIMPLE LIMITING PROCEDURE FROM THE
+LEIN['ORDON OPERATOR c _ M !PPROPRIATE INVARIANT AND LOCAL DElNI
TIONS OF SQUARE ROOTS OF a AND c OR c _ M CORRESPONDING TO SPINORIAL
REPRESENTATIONS OF THE ,ORENTZ GROUP LEAD TO THE ASSOCIATED $IRAC OPERATORS
)N THE SAME VEIN WE CAN ASSOCIATE TO EVERY 2IEMANNIAN OR ,ORENTZIAN
MANIFOLD - G THE OPERATORS aG RESP cG OR THE CORRESPONDING $IRAC
OPERATORS 4HESE EQUATIONS INHERIT IN A STRAIGHTFORWARD WAY THE SYMME
TRIES OF THE SPACES ON WHICH THEY ARE DElNED 4HERE EXISTS A MORE GENERAL
UNREASONABLY EkECTIVE SCHEME OF GENERATING EQUATIONS WITH PRESCRIBED SYM
METRIES 4HE VARIATIONAL 0RINCIPLE ALLOWS US TO ASSOCIATE TO ANY ,AGRANGIAN
, A SYSTEM OF PARTIAL DIkERENTIAL EQUATIONS CALLED THE %ULER[,AGRANGE EQUA
TIONS WHICH INHERIT THE SYMMETRIES BUILT IN , )N VIEW OF .OETHERS PRIN
CIPLE TO ANY CONTINUOUS SYMMETRY OF THE ,AGRANGIAN THERE CORRESPONDS A
CONSERVATION LAW FOR THE ASSOCIATED %ULER[,AGRANGE 0$% 4HUS THE 6ARIA
TIONAL 0RINCIPLE GENERATES EQUATIONS WITH DESIRED CONSERVATION LAWS SUCH AS
%NERGY ,INEAR AND !NGULAR -OMENTA ETC 4HE GENERAL CLASS OF ,AGRANGIAN
EQUATIONS PLAYS THE SAME SELECTED ROLE AMONG ALL 0$%S AS THAT PLAYED BY
(AMILTONIAN SYSTEMS AMONG /$%S #ALCULUS OF 6ARIATIONS IS BY ITSELF A
VENERABLE AND VAST SUBJECT OF -ATHEMATICS 4HE MAIN EQUATIONS OF INTEREST
IN BOTH 'EOMETRY AND 0HYSICS HOWEVER ARE NOT JUST VARIATIONAL THEY

4HERE ARE HOWEVER IMPORTANT GEOMETRIC PROBLEMS SUCH AS PRESCRIBED CURVATURE AND
ISOMETRIC EMBEDDINGS IN 2IEMANNIAN 'EOMETRY OR ,EWY mAT SURFACES IN #OMPLEX 'EOM
ETRY WITHOUT AN OBVIOUS VARIATIONAL STRUCTURE 4HE REAL AND COMPLEX -ONGE !MPERE
EQUATIONS ARE TYPICAL EXAMPLES 4HE 0SEUDO HOLOMORPHIC #URVES USED BY 'ROMOV IN
THE STUDY OF SYMPLECTIC MANIFOLDS PROVIDES ANOTHER EXAMPLE .EVERTHELESS THESE EQUA
TIONS HAVE A RICH GEOMETRIC STRUCTURE AND SHARE WITH THE VARIATIONAL 0$%S MANY COMMON
CHARACTERISTICS -OREOVER ON CLOSER INSPECTION THEY MAY TURN OUT TO HAVE A NONTRIVIAL
3 +,!).%2-!. '!&!
ARE OBTAINED FROM ,AGRANGIANS CONSTRUCTED FROM SIMPLE GEOMETRIC OBJECTS

SUCH AS
,ORENTZ OR 2IEMANNIAN METRICS /N A ,ORENTZIAN MANIFOLD - G THE
,AGRANGIAN GIVEN BY THE SCALAR CURVATURE 2G OF THE METRIC LEADS THROUGH
VARIATIONS OF THE METRIC TO THE %INSTEIN 6ACUUM %6 EQUATIONS OF 'ENERAL
2ELATIVITY ! SIMILAR PROCEDURE LEADS TO %INSTEIN METRICS
2
IN 2IEMANNIAN GE
OMETRY 4HE RESTRICTION OF THE %INSTEIN FUNCTIONAL 2G DVG TO A CONFORMAL
CLASS OF METRICS LEADS TO THE WELL KNOWN 9AMABE EQUATION
#ONNECTIONS ON A PRINCIPAL BUNDLE 4HE QUADRATIC SCALAR INVARIANT
FORMED BY THE CURVATURE OF A CONNECTION DElNES THE 9ANG[-ILLS ,AGRANGIAN
4HE 9ANG[-ILLS 9- EQUATIONS ARE OBTAINED THROUGH VARIATIONS OF THE CON
NECTION 4HE -AXWELL EQUATIONS CORRESPOND TO THE CASE OF A TRIVIAL BUNDLE
OVER THE -INKOWSKI SPACE WITH STRUCTURE GROUP 5 4HE STANDARD MODEL
OF PARTICLE PHYSICS CORRESPONDS TO THE GROUP 35 a 35 a 5 4HE
9- EQUATIONS USED IN 4OPOLOGY CORRESPOND TO 2IEMANNIAN CONNECTIONS
WITH NONABELIAN GROUP 35
3CALAR EQUATIONS !RE DERIVED FOR SCALAR FUNCTIONS } - 2 # 4HE
,AGRANGIAN IS , Guv }u }v 6 } WITH 6 } t 7HEN 6 WE
DERIVE aG } IN THE 2IEMANNIAN CASE AND cG } IN THE ,ORENTZIAN
CASE 4HE CASE 6 } M J}J CORRESPONDS TO THE +LEIN['ORDON EQUATION
6 } J}J LEADS TO THE WELL KNOWN CUBIC WAVE EQUATION 7E WILL REFER
TO THIS TYPE OF EQUATIONS AS NONLINEAR SCALAR WAVE EQUATIONS .37%
-APPINGS BETWEEN TWO MANIFOLDS #ONSIDER MAPPINGS } - G
. H BETWEEN THE PSEUDORIEMANNIAN DOMAIN MANIFOLD - OF DIMENSION
D AND 2IEMANNIAN TARGET . OF DIMENSION N ,ET }b H BE THE SYM
METRIC TENSOR ON - OBTAINED BY TAKING THE PULL BACK OF THE METRIC H
OF . ,ET t t tD BE THE EIGENVALUES OF }b H RELATIVE TO THE METRIC
G AND 3 3 3D THE CORRESPONDING ELEMENTARY SYMMETRIC POLYNOMIALS
IN t t tD !NY SYMMETRIC FUNCTION OF t t tD OR EQUIVALENTLY
ANY FUNCTION ,3 3 3D CAN SERVE AS A ,AGRANGIAN "Y VARYING THE
2
ACTION INTEGRAL - ,DVG RELATIVE TO } WITH DVG THE VOLUME ELEMENT OF THE
METRIC G WE OBTAIN A VAST CLASS OF INTERESTING EQUATIONS (ERE ARE SOME
EXAMPLES
HIDDEN VARIATIONAL STRUCTURE 4HIS IS THE CASE FOR EXAMPLE OF THE -ONGE !MPERE EQUA
TIONS SEE ;"R=

) WANT TO THANK $ #HRISTODOULOU FOR HIS HELP IN THE PRESENTATION OF THIS SECTION
-OST OF THE EXAMPLES BELOW AND MUCH MORE ARE DISCUSSED IN DETAIL IN HIS BOOK ;#HR=
6ISIONS IN -ATH 0$% !3 ! 5.)&)%$ 35"*%#4
I 4HE (ARMONIC AND 7AVE -APS 7- ARE OBTAINED IN THE PARTICULAR
CASE , TRG }b H 4HE ONLY DISTINCTION BETWEEN THEM IS DUE TO THE
CHARACTER 2IEMANNIAN RESPECTIVELY ,ORENTZIAN OF THE METRIC G
II 4HE BASIC EQUATIONS OF #ONTINUUM -ECHANICS ARE OBTAINED FROM A
GENERAL ,AGRANGIAN AS DESCRIBED ABOVE IN THE PARTICULAR CASE WHEN
G IS ,ORENTZIAN N D AND THE ADDITIONAL ASSUMPTIONS THAT } HAS
MAXIMAL RANK AT EVERY POINT AND THE CURVES }_ P ARE TIME LIKE FOR
ALL P . 3INCE THE DIMENSION OF . IS ONE LESS THAN THE DIMENSION OF
- ONE OF THE EIGENVALUES SAY t IS IDENTICALLY ZERO %LASTICITY CORRE
SPONDS TO GENERAL CHOICES OF , AS A SYMMETRIC FUNCTION OF t tD
&LUID -ECHANICS CORRESPONDS TO THE SPECIAL CASE WHEN , DEPENDS ONLY
ON THE PRODUCT t ` t ` ` ` ` tD /NE CAN ALSO DERIVE THE EQUATIONS OF
-AGNETO HYDRODYNAMICS -($ BY ASSUMING AN ADDITIONAL STRUCTURE
ON . GIVEN BY A FORM j 4HE FORM & }b DElNES THE ELECTRO
MAGNETIC lELD ON - 4HE ,AGRANGIAN OF -($ IS OBTAINED BY ADDING
THE -AXWELL ,AGRANGIAN &uv ` & uv TO THE mUID ,AGRANGIAN DESCRIBED
ABOVE
III P
4HE MINIMAL mP
SURFACE EQUATION IS DERIVED FROM THE ,AGRANGIAN ,
b
DETG } H DETG IN THE CASE WHEN G IS 2IEMANNIAN AND M
D N 4HE CASE WHEN G IS ,ORENTZIAN LEADS TO A QUASILINEAR
WAVE EQUATION
,AGRANGIAN LEADING TO HIGHER ORDER EQUATIONS 7HILE THE MAIN EQUA
TIONS OF 0HYSICS ARE ALL lRST OR SECOND ORDER THERE IS NO REASON WHY ONE
SHOULD AVOID HIGHER ORDER EQUATIONS FOR APPLICATIONS TO 'EOMETRY )T IS NAT
URAL FOR EXAMPLE TO CONSIDER EQUATIONS ASSOCIATED TO CONFORMALLY INVARIANT
,AGRANGIANS -ANY OF THE KNOWN ,AGRANGIANS WHICH LEAD TO SECOND ORDER
EQUATIONS SUCH AS (ARMONIC -APS ARE CONFORMALLY INVARIANT ONLY IN DIMEN
SION 4O PRODUCE A LARGER CLASS OF CONFORMALLY INVARIANT EQUATIONS IN EVEN
DIMENSIONS IT PAYS TO LOOK FOR HIGHER ORDER THEORIES SUCH AS BIHARMONIC
MAPS IN $ SEE ;#79= 4HE VARIATIONAL PROBLEM ASSOCIATED TO THE ZETA
FUNCTIONAL DETERMINANT OF THE ,APLACE["ELTRAMI OPERATOR OF A HIGHER DI
MENSIONAL 2IEMANNIAN METRIC ALSO LEADS TO HIGHER ORDER EQUATIONS &INALLY
THE 7ILLMORE PROBLEM FOR CLOSED SURFACES IN 2 PROVIDES ANOTHER INTERESTING
EXAMPLE OF A FOURTH ORDER EQUATION
#OMPOSITE ,AGRANGIANS "Y ADDING VARIOUS ,AGRANGIANS WE DERIVE OTHER
BASIC EQUATIONS 4HIS IS TRUE MOST REMARKABLY FOR THE GRAVITATIONAL ,A
GRANGIAN GIVEN BY THE SCALAR CURVATURE OF THE METRIC )N COMBINATION
WITH THE ,AGRANGIAN OF A MATTER THEORY IN FACT ANY OTHER RELATIVISTIC ,A
3 +,!).%2-!. '!&!
GRANGIANS DESCRIBED ABOVE IT LEADS TO THE FAMOUS %INSTEIN &IELD %QUATIONS

2uv _ Guv 2 x4uv WITH 4 THE ENERGY MOMENTUM TENSOR OF THE MATTER
,AGRANGIAN 4HE ,AGRANGIAN OF THE 3EIBERG[7ITTEN EQUATIONS ARE OBTAINED
BY COUPLING THE ,AGRANGIAN OF THE -AXWELL THEORY WITH THAT OF THE $IRAC
EQUATION
4HE EQUATIONS DERIVED BY THE ABOVE GEOMETRIC CONSTRUCTIONS ARE ELLIPTIC
IF THE METRIC G ON - IS 2IEMANNIAN AND HYPERBOLIC IF G IS ,ORENTZIAN )N
THE HYPERBOLIC CASE WE DISTINGUISH BETWEEN THE &IELD 4HEORIES FOR WHICH THE
ONLY CHARACTERISTICS OF THE CORRESPONDING 0$%S ARE GIVEN BY THE ,ORENTZ
METRIC G AND THE OTHER EQUATIONS &LUIDS #ONTINUUM -ECHANICS -($
ETC WHICH HAVE ADDITIONAL CHARACTERISTICS 4HE 9- 7- AND THE %6 ARE
ALL lELD THEORIES IN THE SENSE WE HAVE JUST DElNED 4HE %6 EQUATIONS IS
DISTINGUISHED FROM THE OTHER lELD THEORIES BY BEING THE ONLY ONE FOR WHICH
THE METRIC G ITSELF IS THE SOLUTION 4HIS FACT GIVES THE %6 EQUATIONS A
QUASILINEAR CHARACTER &OR ALL OTHER lELD EQUATIONS SINCE THE METRIC G IS
lXED THE EQUATIONS ARE SEMILINEAR
7ITH THE EXCEPTIONAL CASE OF %6 WHICH DOES NOT HAVE LOCAL CONSERVATION
LAWS ALL EQUATIONS DESCRIBED ABOVE HAVE ASSOCIATED TO THEM A WELL DElNED
ENERGY MOMENTUM TENSOR 4 WHICH VERIlES THE POSITIVE ENERGY CONDITION )
RECALL THAT THE ENERGY MOMENTUM TENSOR OF A ,AGRANGIAN THEORY IS A RANK
SYMMETRIC TENSOR 4uv VERIFYING THE LOCAL CONSERVATION LAW $u 4uv 7E
SAY THAT 4 SATISlES THE POSITIVE ENERGY CONDITION IF 4 8 9 t FOR ALL
TIME LIKE FUTURE ORIENTED VECTORlELDS 8 9
-ANY OTHER FAMILIAR EQUATIONS CAN BE DERIVED FROM THE FUNDAMENTAL
ONES DESCRIBED ABOVE BY THE FOLLOWING PROCEDURES
A 3YMMETRY REDUCTIONS !RE OBTAINED BY ASSUMING THAT THE SOLU
TIONS WE ARE LOOKING FOR HAVE CERTAIN CONTINUOUS SYMMETRIES 4HEY LEAD TO
MUCH SIMPLER EQUATIONS THAN THE ORIGINAL OFTEN INTRACTABLE ONES !NOTHER
SOMEWHAT MORE GENERAL WAY OF OBTAINING SIMPLER EQUATIONS IS TO LOOK FOR
SOLUTIONS WHICH VERIFY A CERTAIN ANSATZ SUCH AS STATIONARY SPHERICALLY SYM
METRIC EQUIVARIANT SELF SIMILAR TRAVELING WAVES ETC
B 4HE .EWTONIAN APPROXIMATION AND OTHER LIMITS 7E CAN DERIVE A
LARGE CLASS OF NEW EQUATIONS FROM THE BASIC ONES DESCRIBED ABOVE BY TAKING
ONE OR MORE CHARACTERISTIC SPEEDS TO INlNITY 4HE MOST IMPORTANT ONE IS
THE .EWTONIAN LIMIT WHICH IS FORMALLY OBTAINED BY LETTING THE VELOCITY
OF LIGHT GO TO INlNITY !T THE LEVEL OF THE SPACE TIME MANIFOLD ITSELF THIS
LIMIT DESCRIBED IN THE SEMINAL PAPER OF -INKOWSKI ;-= TAKES A ,ORENTZ
MANIFOLD TO THE 'ALILEAN SPACE TIME OF .EWTONIAN MECHANICS !S WE HAVE
6ISIONS IN -ATH 0$% !3 ! 5.)&)%$ 35"*%#4
MENTIONED ABOVE THE 3CHRODINGER EQUATION ITSELF CAN BE DERIVED IN THIS

FASHION FROM THE LINEAR +LEIN['ORDON EQUATION )N THE SAME WAY WE CAN
FORMALLY DERIVE THE ,AGRANGIAN OF NONRELATIVISTIC %LASTICITY SEE ;:= &LUIDS
OR -($ EQUATIONS 4HE FORMAL .EWTONIAN LIMIT OF THE FULL %INSTEIN lELD
EQUATIONS LEADS TO THE VARIOUS CONTINUUM MECHANICS THEORIES IN THE PRESENCE
OF .EWTONIAN GRAVITY 4HE .EWTONIAN POTENTIAL IS TIED TO THE LAPSE FUNCTION
OF THE ORIGINAL SPACE TIME METRIC
7E SHOULD NOT BE SURPRISED THAT THE BETTER KNOWN NONRELATIVISTIC EQUA
TIONS LOOK MORE MESSY THAN THE RELATIVISTIC ONES 4HE SIMPLE GEOMETRIC
STRUCTURE OF THE ORIGINAL EQUATIONS GETS LOST IN THE LIMIT 4HE REMARKABLE
SIMPLICITY OF THE RELATIVISTIC EQUATIONS IS A POWERFUL EXAMPLE OF THE IMPOR
TANCE OF 2ELATIVITY AS A UNIFYING PRINCIPLE
/NCE WE ARE IN THE FAMILIAR WORLD OF .EWTONIAN PHYSICS WE CAN PERFORM
OTHER WELL KNOWN LIMITS 4HE FAMOUS INCOMPRESSIBLE %ULER EQUATIONS ARE
OBTAINED BY TAKING THE LIMIT OF THE GENERAL NONRELATIVISTIC mUID EQUATIONS
AS THE SPEED OF SOUND TENDS TO INlNITY 6ARIOUS OTHER LIMITS ARE OBTAINED
RELATIVE TO OTHER CHARACTERISTIC SPEEDS OF THE SYSTEM OR IN CONNECTION WITH
SPECIlC BOUNDARY CONDITIONS SUCH AS THE BOUNDARY LAYER APPROXIMATION IN
mUIDS 4HE EQUATIONS OF %LASTICITY FOR EXAMPLE APPROACH IN THE LIMIT WHEN
ALL CHARACTERISTIC SPEEDS TEND TO INlNITY TO THE FAMILIAR EQUATIONS OF A RIGID
BODY IN #LASSICAL -ECHANICS !NOTHER IMPORTANT TYPE OF LIMIT LEADING TO
THE WELL KNOWN (AMILTON[*ACOBI EQUATIONS OF #LASSICAL -ECHANICS IS THE
HIGH FREQUENCY OR THE GEOMETRIC OPTICS APPROXIMATION
-ANY OF THESE VERY SINGULAR LIMITS REMAIN PURELY FORMAL 7HILE SOME
OF THEM HAVE BEEN RIGOROUSLY DERIVED MANY MORE PRESENT SERIOUS ANALYTIC
DInCULTIES
C 0HENOMENOLOGICAL ASSUMPTIONS %VEN AFTER TAKING VARIOUS LIMITS
AND MAKING SYMMETRY REDUCTIONS THE EQUATIONS MAY STILL REMAIN UNYIELD
ING )N VARIOUS APPLICATIONS IT MAKES SENSE TO ASSUME THAT CERTAIN QUANTITIES
ARE SMALL AND MAY BE NEGLECTED 4HIS LEADS TO SIMPLIlED EQUATIONS WHICH
COULD BE CALLED PHENOMENOLOGICAL IN THE SENSE THAT THEY ARE NOT DERIVABLE
FROM lRST PRINCIPLES 4HEY ARE USED TO ILLUSTRATE AND ISOLATE IMPORTANT PHYS
ICAL PHENOMENA PRESENT IN COMPLICATED SYSTEMS ! TYPICAL WAY OF GENERATING
INTERESTING PHENOMENOLOGICAL EQUATIONS IS TO TRY TO WRITE DOWN THE SIMPLEST
MODEL EQUATION WHICH DESCRIBES A PARTICULAR FEATURE OF THE ORIGINAL SYSTEM

) USE THIS TERM HERE QUITE FREELY IT IS TYPICALLY USED IN A SOMEWHAT DIkERENT CON
TEXT !LSO SOME OF THE EQUATIONS WHICH ) CALL PHENOMENOLOGICAL BELOW EG DISPERSIVE
EQUATIONS CAN BE GIVEN FORMAL ASYMPTOTICS DERIVATIONS BY !PPLIED -ATH TECHNIQUES
3 +,!).%2-!. '!&!
4HUS THE SELF FOCUSING PLANE WAVE EkECTS OF COMPRESSIBLE mUIDS OR ELASTIC
ITY CAN BE ILLUSTRATED BY THE SIMPLE MINDED "URGERS EQUATION UT UUX
.ONLINEAR DISPERSIVE PHENOMENA TYPICAL TO mUIDS CAN BE ILLUSTRATED BY THE
FAMOUS +D6 EQUATION UT UUX UXXX 4HE NONLINEAR 3CHR ODINGER
EQUATIONS PROVIDE GOOD MODEL PROBLEMS FOR NONLINEAR DISPERSIVE EkECTS IN
/PTICS 4HE 'INZBURG[,ANDAU EQUATIONS PROVIDE A SIMPLE MODEL EQUATION
FOR SYMMETRY BREAKING PHASE TRANSITIONS 4HE -AXWELL[6LASOV EQUATIONS IS
A SIMPLIlED MODEL FOR THE INTERACTIONS BETWEEN %LECTOMAGNETIC FORCES AND
CHARGED PARTICLES USED IN 0LASMA 0HYSICS
7HEN WELL CHOSEN A MODEL EQUATION LEADS TO BASIC INSIGHTS INTO THE
ORIGINAL EQUATION ITSELF &OR THIS REASON SIMPLIlED MODEL PROBLEMS ARE ALSO
ESSENTIAL IN THE DAY TO DAY WORK OF THE RIGOROUS 0$% MATHEMATICIAN 7E
ALL TEST OUR IDEAS ON SUCH CAREFULLY SELECTED MODEL PROBLEMS )T IS CRUCIAL
TO EMPHASIZE THAT GOOD RESULTS CONCERNING THE BASIC PHYSICAL EQUATIONS ARE
RARE A VERY LARGE PERCENTAGE OF IMPORTANT RIGOROUS WORK IN 0$% DEALS WITH
SIMPLIlED EQUATIONS SELECTED FOR TECHNICAL REASONS TO ISOLATE AND FOCUS OUR
ATTENTION ON SOME SPECIlC DInCULTIES PRESENT IN THE BASIC EQUATIONS
)T IS NOT AT ALL A SURPRISE THAT THE EQUATIONS DERIVED BY SYMMETRY RE
DUCTIONS VARIOUS LIMITS AND PHENOMENOLOGICAL ASSUMPTIONS HAVE ADDITIONAL
SYMMETRIES AND THEREFORE ADDITIONAL CONSERVATION LAWS )T IS HOWEVER RE
MARKABLE THAT SOME OF THEM HAVE INlNITELY MANY CONSERVED QUANTITIES OR
TURN OUT TO BE EVEN INTEGRABLE 4HE DISCOVERY OF THE INTEGRABILITY OF THE +D6
EQUATION AND LATER THAT OF OTHER INTEGRABLE 0$%S IS ONE OF THE MOST IM
PRESSIVE ACHIEVEMENTS OF THE lELD OF 0$%S IN THIS CENTURY )T REMAINS ALSO
THE MODEL CASE OF A BENElCIAL INTERACTION BETWEEN NUMERICAL EXPERIMENTS
HEURISTIC APPLIED MATHEMATICS ARGUMENTS ALGEBRA AND RIGOROUS ANALYSIS 4O
GETHER THEY HAVE LED TO THE CREATION OF A BEAUTIFUL MATHEMATICAL THEORY WITH
EXTENSIVE AND DEEP APPLICATIONS OUTSIDE THE lELD OF 0$%S WHERE THEY HAVE
ORIGINATED FROM 7E HAVE TO BE AWARE HOWEVER OF THE OBVIOUS LIMITATIONS
OF INTEGRABLE SYSTEMS WITH FEW EXCEPTIONS THE +0 ) AND +0 )) EQUATIONS
ARE SORT OF DIMENSIONAL ALL KNOWN INTEGRABLE EVOLUTION EQUATIONS ARE
RESTRICTED TO ONE SPACE DIMENSION
)N ALL THE ABOVE DISCUSSION WE HAVE NOT MENTIONED DIkUSIVE EQUATIONS
SUCH AS THE .AVIER[3TOKES 4HEY ARE IN FACT NOT VARIATIONAL AND THEREFORE
DO NOT lT AT ALL IN THE ABOVE DESCRIPTION 4HEY PROVIDE A LINK BETWEEN
THE MICROSCOPIC DISCRETE WORLD OF .EWTONIAN PARTICLES AND THE CONTINUOUS
MACROSCOPIC ONE DESCRIBED BY #ONTINUUM -ECHANICS 0ASSING FROM DIS
CRETE TO CONTINUOUS INVOLVES SOME LOSS OF INFORMATION HENCE THE CONTINUUM
6ISIONS IN -ATH 0$% !3 ! 5.)&)%$ 35"*%#4
EQUATIONS HAVE DIkUSIVE FEATURES 4HE BEST KNOWN EXAMPLES OF DIkUSIVE

EkECTS ARE THE <HEAT CONDUCTION WHICH APPEARS IN CONNECTION WITH THE
DISSIPATION OF ENERGY IN COMPRESSIBLE mUIDS AND <VISCOSITY CORRESPONDING
TO DISSIPATION OF MOMENTUM IN &LUIDS !NOTHER EXAMPLE IS THAT OF <ELECTRI
CAL RESISTIVITY FOR THE ELECTRODYNAMICS OF CONTINUUM MEDIA 4HE .AVIER[
3TOKES EQUATION APPEARS IN THE INCOMPRESSIBLE LIMIT 4HE INCOMPRESSIBLE
%ULER EQUATIONS ARE THE FORMAL LIMIT OF THE .AVIER[3TOKES EQUATIONS AS THE
VISCOSITY TENDS TO ZERO "ECAUSE OF THE LOSS OF INFORMATION INVOLVED IN THEIR
DERIVATION THE DIkUSIVE EQUATIONS HAVE PROBABILISTIC INTERPRETATIONS
$IkUSIVE EQUATIONS TURN OUT TO BE ALSO VERY USEFUL IN CONNECTION WITH
GEOMETRIC PROBLEMS 'EOMETRIC mOWS SUCH AS MEAN CURVATURE INVERSE MEAN
CURVATURE (ARMONIC -APS 'AUSS #URVATURE AND 2ICCI mOWS ARE SOME OF
THE BEST KNOWN EXAMPLES 3OME CAN BE INTERPRETED AS THE GRADIENT mOW FOR
AN ASSOCIATED ELLIPTIC VARIATIONAL PROBLEM 4HEY CAN BE USED TO CONSTRUCT
NONTRIVIAL STATIONARY SOLUTIONS TO THE CORRESPONDING STATIONARY SYSTEMS IN
THE LIMIT AS T OR TO PRODUCE FOLIATIONS WITH REMARKABLE PROPERTIES
SUCH AS THAT USED RECENTLY IN THE PROOF OF THE 0ENROSE CONJECTURE
2EMARK 4HE EQUATIONS WHICH ARE OBTAINED BY APPROXIMATIONS OR BY PHE
NOMENOLOGICAL ASSUMPTIONS PRESENT US WITH AN INTERESTING DILEMMA 4HE
DYNAMICS OF SUCH EQUATIONS MAY LEAD TO BEHAVIOR WHICH IS INCOMPATIBLE
WITH THE ASSUMPTIONS MADE IN THEIR DERIVATION 3HOULD WE CONTINUE TO TRUST
AND STUDY THEM NEVERTHELESS FOR PURE MATHEMATICAL REASONS OR SHOULD WE
ABANDON THEM IN FAVOR OF THE ORIGINAL EQUATIONS OR A BETTER APPROXIMATION
7HATEVER ONE MAY FEEL ABOUT THIS IN A SPECIlC SITUATION IT IS CLEAR THAT
THE PROBLEM OF UNDERSTANDING RIGOROUSLY THE RANGE OF VALIDITY OF VARIOUS
APPROXIMATIONS IS ONE OF THE FUNDAMENTAL PROBLEMS IN 0$%
4HE 0ROBLEM OF "REAKDOWN

4HE MOST BASIC MATHEMATICAL QUESTION IN 0$% IS BY FAR THAT OF REGULARITY
)N THE CASE OF ELLIPTIC EQUATIONS OR SUBELLIPTIC IN #OMPLEX !NALYSIS THE
ISSUE IS TO DETERMINE THE REGULARITY OF THE SOLUTIONS TO A GEOMETRIC VARIA
TIONAL PROBLEM )N VIEW OF THE MODERN WAY OF TREATING ELLIPTIC EQUATIONS
ONE lRST CONSTRUCTS A GENERALIZED SOLUTION BY USING THE VARIATIONAL CHARACTER
OF THE EQUATIONS 4HE ORIGINAL PROBLEM THEN TRANSLATES TO THAT OF SHOWING
THAT THE GENERALIZED SOLUTION HAS ADDITIONAL REGULARITY 4HIS IS A COMMON
TECHNIQUE FOR BOTH LINEAR AND NONLINEAR PROBLEMS -OREOVER THE TECHNIQUE
CAN BE EXTENDED TO SCALAR FULLY NONLINEAR NONVARIATIONAL PROBLEMS SUCH AS
3 +,!).%2-!. '!&!
-ONGE !MPERE EQUATIONS WITH THE HELP OF THE VISCOSITY METHOD )N LINEAR
CASES AS WELL AS IN SOME FAMOUS NONLINEAR CASES SUCH AS THE MINIMAL HYPER
SURFACES AS GRAPHS OVER CONVEX DOMAINS OR MORE GENERALLY MEAN CONVEX
ONE CAN SHOW THAT THE GENERALIZED SOLUTIONS ARE SMOOTH 4HE SOLUTIONS TO
THE GENERAL 0LATEAU PROBLEM HOWEVER MAY HAVE SINGULARITIES )N THIS CASE
THE MAIN ISSUE BECOMES THE STRUCTURE OF THE SINGULAR SET OF NONSMOOTH SOLU
TIONS 'EOMETRIC -EASURE 4HEORY PROVIDES SOPHISTICATED ANALYTICAL TOOLS TO
DEAL WITH THIS PROBLEM 3INGULARITIES ARE ALSO KNOWN TO OCCUR IN THE CASE OF
HIGHER DIMENSIONAL HARMONIC MAPS FOR POSITIVELY CURVED TARGET MANIFOLDS
SUCH AS SPHERES
)N THE CASE OF EVOLUTION EQUATIONS THE ISSUE IS THE POSSIBLE SPONTANEOUS
lNITE TIME IN VIEW OF RESULTS CONCERNING LOCAL IN TIME EXISTENCE THE BREAK
DOWN CAN ONLY OCCUR AFTER A SHORT TIME INTERVAL BREAKDOWN OF SOLUTIONS
CORRESPONDING TO PERFECTLY NICE INITIAL CONDITIONS 4HIS IS A TYPICAL NONLIN
EAR MULTIDIMENSIONAL PHENOMENON )T CAN BE BEST ILLUSTRATED IN THE CASE
OF THE "URGERS EQUATION UT UUX 2 $ESPITE THE PRESENCE OF INlNITELY
MANY POSITIVE CONSERVED QUANTITIES JUT X JK DX K . ALL SOLUTIONS COR
RESPONDING TO SMOOTH COMPACTLY SUPPORTED NONZERO INITIAL DATA AT T
BREAKDOWN IN lNITE TIME 4HE BREAKDOWN CORRESPONDS PHYSICALLY TO THE
FORMATION OF A SHOCK WAVE 3IMILAR EXAMPLES OF BREAKDOWN CAN BE CON
STRUCTED FOR COMPRESSIBLE mUIDS OR %LASTICITY SEE ;*= ;3I= 3INGULARITIES ARE
ALSO KNOWN TO FORM IN SOME SPECIAL CASES FOR SOLUTIONS TO THE %INSTEIN lELD
EQUATIONS IN 'ENERAL 2ELATIVITY -OREOVER ONE EXPECTS THIS TO HAPPEN IN
GENERAL IN THE PRESENCE OF STRONG GRAVITATIONAL lELDS )T IS ALSO WIDELY EX
PECTED THAT THE GENERAL SOLUTIONS OF THE INCOMPRESSIBLE %ULER EQUATIONS IN
THREE SPACE DIMENSIONS MODELING THE BEHAVIOR OF INVISCID mUIDS BREAKDOWN
IN lNITE TIME 3OME SPECULATE THAT THE BREAKDOWN MAY HAVE SOMETHING TO
DO WITH THE ONSET OF TURBULENCE FOR INCOMPRESSIBLE mUIDS WITH VERY HIGH
2EYNOLDS NUMBERS 4HESE mUIDS ARE IN FACT DESCRIBED BY THE .AVIER[3TOKES
EQUATIONS )N THIS CASE THE GENERAL CONSENSUS IS THAT THE EVOLUTION OF ALL
SMOOTH lNITE ENERGY INITIAL DATA LEAD TO GLOBAL IN TIME SMOOTH SOLUTIONS
4HE PROBLEM IS STILL WIDELY OPEN )T IS CONCEIVABLE THAT THERE ARE IN FACT
PLENTY OF SOLUTIONS WHICH BREAK DOWN BUT ARE UNSTABLE AND THUS IMPOSSIBLE
TO DETECT NUMERICALLY OR EXPERIMENTALLY
"REAKDOWN OF SOLUTIONS IS ALSO AN ESSENTIAL ISSUE CONCERNING NONLINEAR

&OR SMOOTH ONE DIMENSIONAL (AMILTONIAN SYSTEMS WITH POSITIVE ENERGY SOLUTIONS
ARE AUTOMATICALLY GLOBAL IN TIME 4HIS THE CASE FOR EXAMPLE OF THE NONLINEAR HARMONIC
D
OSCILLATOR DT X 6 X 6 t
6ISIONS IN -ATH 0$% !3 ! 5.)&)%$ 35"*%#4
GEOMETRIC mOWS SUCH AS THE MEAN AND INVERSE MEAN CURVATURE mOWS 2ICCI
mOW ETC !S SINGULARITIES DO ACTUALLY FORM IN MANY IMPORTANT GEOMETRIC
SITUATIONS ONE IS FORCED TO UNDERSTAND THE STRUCTURE OF SINGULARITIES AND lND
WAYS TO CONTINUE THE mOW PAST THEM 5SEFUL CONSTRUCTIONS OF GENERALIZED
mOWS CAN LEAD TO THE SOLUTION OF OUTSTANDING GEOMETRIC PROBLEMS AS IN THE
RECENT CASE OF THE 0ENROSE CONJECTURE ;(U)=
4HE PROBLEM OF POSSIBLE BREAKDOWN OF SOLUTIONS TO INTERESTING NON
LINEAR GEOMETRIC AND PHYSICAL SYSTEMS IS NOT ONLY THE MOST BASIC PROBLEM
IN 0$% IT IS ALSO THE MOST CONSPICUOUS UNIFYING PROBLEM IN THAT IT AF
FECTS ALL 0$%S )T IS INTIMATELY TIED TO THE BASIC MATHEMATICAL QUESTION
OF UNDERSTANDING WHAT WE ACTUALLY MEAN BY SOLUTIONS AND FROM A PHYSICAL
POINT OF VIEW TO THE ISSUE OF UNDERSTANDING THE VERY LIMITS OF VALIDITY OF THE
CORRESPONDING PHYSICAL THEORIES 4HUS IN THE CASE OF THE "URGERS EQUATION
FOR EXAMPLE THE PROBLEM OF SINGULARITIES CAN BE TACKLED BY EXTENDING OUR
CONCEPT OF SOLUTIONS TO ACCOMMODATE <SHOCK WAVES IE SOLUTIONS DISCON
TINUOUS ACROSS CURVES IN THE T X SPACE /NE CAN DElNE A FUNCTIONAL SPACE OF
GENERALIZED SOLUTIONS IN WHICH THE INITIAL VALUE PROBLEM HAS UNIQUE GLOBAL
SOLUTIONS 4HOUGH THE SITUATION FOR MORE REALISTIC PHYSICAL SYSTEMS IS FAR
LESS CLEAR AND FAR FROM BEING SATISFACTORILY SOLVED THE GENERALLY HELD OPIN
ION IS THAT SHOCK WAVE TYPE SINGULARITIES CAN BE ACCOMMODATED WITHOUT
BREAKING THE BOUNDARIES OF THE PHYSICAL THEORY AT HAND 4HE SITUATION OF
SINGULARITIES IN 'ENERAL 2ELATIVITY IS RADICALLY DIkERENT 4HE TYPE OF SIN
GULARITIES EXPECTED HERE IS SUCH THAT NO CONTINUATION OF THE SOLUTIONS IS
POSSIBLE WITHOUT ALTERING THE PHYSICAL THEORY ITSELF 4HE PREVALING OPINION
IN THIS RESPECT IS THAT ONLY A QUANTUM lELD THEORY OF 'RAVITY COULD ACHIEVE
THIS
/NE CAN FORMULATE A GENERAL PHILOSOPHY TO EXPRESS OUR EXPECTATIONS
WITH REGARD TO REGULARITY 4O DO THAT WE NEED TO CLASSIFY OUR MAIN EQUA
TIONS ACCORDING TO THE STRENGTH OF THEIR NONLINEARITIES RELATIVE TO THAT OF THE
KNOWN COERCIVE CONSERVATION LAWS OR OTHER A PRIORI ESTIMATES !MONG THE
BASIC CONSERVATION LAWS THAT PROVIDED BY THE %NERGY IS COERCIVE BECAUSE
IT LEADS TO AN ABSOLUTE LOCAL SPACE TIME BOUND ON THE SIZE OF SOLUTIONS OR
THEIR lRST DERIVATIVES 4HE OTHERS SUCH AS THE LINEAR AND ANGULAR MOMEN
TUM DO NOT PROVIDE ANY ADDITIONAL INFORMATION CONCERNING LOCAL REGULARITY
&OR THE BASIC EVOLUTION EQUATIONS DISCUSSED IN THE PREVIOUS SECTION THE EN
ERGY INTEGRAL PROVIDES THE BEST POSSIBLE A PRIORI ESTIMATE AND THEREFORE THE
CLASSIlCATION IS DONE RELATIVE TO IT 4HIS RAISES A QUESTION OF FUNDAMENTAL
IMPORTANCE ARE THERE OTHER STRONGER LOCAL A PRIORI BOUNDS WHICH CANNOT BE
3 +,!).%2-!. '!&!
DERIVED FROM .OETHERS 0RINCIPLE 4HERE ARE METHODS WHICH CAN RULE OUT
THE EXISTENCE OF SOME EXACT CONSERVED QUANTITIES DIkERENT FROM THE PHYS
ICAL ONES YET THERE IS NO REASON ) BELIEVE TO DISCOUNT OTHER MORE SUBTLE
BOUNDS ! WELL KNOWN -ORAWETZ MULTIPLIER METHOD LEADS FOR SOME CLASSES
OF NONLINEAR WAVE EQUATIONS TO BOUNDED SPACE TIME QUANTITIES WHICH DO
NOT CORRESPOND TO ANY CONSERVATION LAW 4HE -ORAWETZ QUANTITY HOWEVER
HAS THE SAME SCALING PROPERTIES AS THE ENERGY INTEGRAL IT ONLY PROVIDES
ADDITIONAL INFORMATION IN THE LARGE
2EMARK 4HE DISCOVERY OF ANY NEW BOUND STRONGER THAN THAT PROVIDED
BY THE ENERGY FOR GENERAL SOLUTIONS OF ANY OF OUR BASIC PHYSICAL EQUATIONS
WOULD HAVE THE SIGNIlCANCE OF A MAJOR EVENT
)N OTHER CASES WHEN THERE ARE ADDITIONAL SYMMETRIES ONE OFTEN HAS
BETTER A PRIORI ESTIMATES &OR MANY ELLIPTIC EQUATIONS FOR EXAMPLE ONE
CAN MAKE USE OF THE MAXIMAL PRINCIPLE OR SOME MONOTONICITY ARGUMENTS TO
DERIVE FAR MORE POWERFUL A PRIORI ESTIMATES THAN THOSE GIVEN BY THE ENERGY
INTEGRAL )NTEGRABLE EQUATIONS SUCH AS +D6 ALSO HAVE ADDITIONAL COERCIVE
CONSERVATION LAWS !S EXPLAINED ABOVE THE "URGERS EQUATION HAS INlNITELY
MANY POSITIVE CONSERVED QUANTITIES 4HE INCOMPRESSIBLE %ULER EQUATIONS
IN DIMENSION N HAVE IN ADDITION TO THE ENERGY A POINTWISE A PRIORI
ESTIMATE FOR THE VORTICITY )T IS FOR THIS REASON THAT WE CAN PROVE GLOBAL
REGULARITY FOR $ %ULER EQUATIONS )N ALL THESE CASES THE CLASSIlCATION HAS
TO BE DONE RELATIVE TO THE OPTIMAL AVAILABLE A PRIORI ESTIMATE
)N WHAT FOLLOWS ) WILL RESTRICT MYSELF TO THE CASE ) lND PERSONALLY MOST
INTERESTING THAT OF THE BASIC EVOLUTION EQUATIONS FOR WHICH THERE ARE NO
BETTER KNOWN A PRIORI ESTIMATES THAN THOSE PROVIDED BY THE %NERGY INTEGRAL
4HESE INCLUDE ALL RELATIVISTIC lELD THEORIES &LUIDS #ONTINUUM -ECHANICS
AND -AGENTOmUIDYNAMIC IN THREE SPACE DIMENSIONS AND THE ABSENCE OF
ANY ADDITIONAL SYMMETRIES )N THESE CASES THE CLASSIlCATION IS DONE BY
MEASURING THE SCALING PROPERTIES OF THE ENERGY INTEGRAL RELATIVE TO THOSE OF
THE EQUATIONS 4O ILLUSTRATE HOW THIS IS DONE CONSIDER THE NONLINEAR SCALAR

EQUATION c} _ 6 } WITH 6 } P J}JP 4HE ENERGY INTEGRAL IS
2 _
`
P T X DX )F WE ASSIGN TO THE SPACE TIME
GIVEN BY J }T X J J}J
VARIABLES THE DIMENSION OF LENGTH , THEN c HAS THE DIMENSION OF ,_

AND } ACQUIRES FROM THE EQUATION THE DIMENSION , _P 4HUS THE ENERGY

INTEGRAL HAS THE DIMENSION ,E E N _ _P 7E SAY THAT THE EQUATION IS
SUBCRITICAL IF E CRITICAL FOR E AND SUPERCRITICAL FOR E 4HE SAME
ANALYSIS CAN BE DONE FOR ALL THE OTHER BASIC EQUATIONS 9- IS SUBCRITICAL
FOR N s CRITICAL FOR N AND SUPERCRITICAL FOR N 7- IS SUBCRITICAL
6ISIONS IN -ATH 0$% !3 ! 5.)&)%$ 35"*%#4
FOR N CRITICAL FOR N AND SUPERCRITICAL FOR ALL OTHER DIMENSIONS

4HE SAME HOLDS TRUE FOR THE %INSTEIN 6ACUUM EQUATIONS -OST OF OUR
BASIC EQUATIONS SUCH AS %6 %ULER .AVIER[3TOKES #OMPRESSIBLE %ULER
%LASTICITY ETC TURN OUT TO BE SUPERCRITICAL IN THE PHYSICAL DIMENSION N
! 0$% IS SAID TO BE REGULAR IF ALL SMOOTH lNITE ENERGY INITIAL CONDITIONS
LEAD TO GLOBAL SMOOTH SOLUTIONS
4HE GENERAL PHILOSOPHY IS THAT SUBCRITICAL EQUATIONS ARE REGULAR WHILE
SUPERCRITICAL EQUATIONS MAY DEVELOP SINGULARITIES #RITICAL EQUATIONS ARE
IMPORTANT BORDERLINE CASES &OR THE PARTICULAR CASE OF lELD THEORIES AS
DElNED IN THE PREVIOUS SECTION ONE CAN FORMULATE A MORE PRECISE CONJECTURE
'ENERAL CONJECTURE I !LL BASIC SUBCRITICAL lELD THEORIES ARE REGULAR
FOR ALL SMOOTH DATA
II 5NDER WELL DElNED RESTRICTIONS ON THEIR GEOMETRIC SET UP THE CRITICAL
lELD THEORIES ARE REGULAR FOR ALL SMOOTH DATA
III <3UnCIENTLY SMALL SOLUTIONS TO THE SUPERCRITICAL lELD THEORIES ARE
REGULAR 4HERE EXIST SOLUTIONS CORRESPONDING TO LARGE SMOOTH lNITE ENERGY
DATA WHICH DEVELOP SINGULARITIES IN lNITE TIME
4HE PART III OF THE #ONJECTURE IS THE MOST INTRIGUING 4HE FACT THAT
ALL SMALL SOLUTIONS ARE REGULAR SEEMS TO BE TYPICAL TO lELD THEORIES IT MAY
FAIL FOR mUIDS OR THE GENERAL ELASTICITY EQUATIONS 4HE ISSUE OF EXISTENCE OF
SINGULAR SOLUTIONS FOR SUPERCRITICAL EQUATIONS IS ALMOST ENTIRELY OPEN )N
THE CASE OF SUPERCRITICAL DEFOCUSING .37% c} _ 6 } FOR POSITIVE
POWER LAW POTENTIAL 6 MOST ANALYSTS FAMILIAR WITH THE PROBLEM EXPECT
THAT GLOBAL REGULARITY STILL PREVAILS .UMERICAL CALCULATIONS SEEM TO SUPPORT
THAT VIEW )T IS HOWEVER ENTIRELY POSSIBLE THAT SINGULAR SOLUTIONS EXIST BUT ARE
UNSTABLE AND THEREFORE DInCULT TO CONSTRUCT ANALYTICALLY AND IMPOSSIBLE TO
DETECT NUMERICALLY ! SIMILAR PHENOMENON MAY HOLD TRUE IN THE CASE OF THE
$ .AVIER[3TOKES EQUATIONS WHICH WOULD CONTRADICT THE ALMOST UNIVERSAL
ASSUMPTION THAT THESE EQUATIONS ARE GLOBALLY REGULAR
)F THIS WORST CASE SCENARIO IS TRUE THE BIG CHALLENGE FOR US WOULD BE TO
PROVE THAT ALMOST ALL SOLUTIONS TO SUCH EQUATIONS ARE GLOBALLY REGULAR !T
THE OPPOSITE END OF POSSIBLE SITUATIONS IS THAT FOR WHICH ALMOST ALL SOLUTIONS

3OME OF THE MOST EXCITING ADVANCES IN 'EOMETRIC 0$%S IN THE LAST TWENTY lVE YEARS
INVOLVE THE STUDY OF 0$%S WHICH ARE CRITICAL RELATIVE TO THE OPTIMAL AVAILABLE A PRIORI
ESTIMATES 4HIS IS THE CASE OF THE 9AMABE PROBLEM RELATED TO THE CRITICAL EXPONENT OF
THE 3OBOLEV INEQUALITY 7EAKLY (ARMONIC -APS IN $ 9ANG[-ILLS CONNECTIONS IN $
THE 7ILMORE PROBLEM IN $ 3EE ;3= FOR A BEAUTIFUL SURVEY AND ;9= FOR HIS UPDATED LIST OF
PROBLEMS IN 'EOMETRY
3 +,!).%2-!. '!&!
FORM SINGULARITIES 4HE $ INCOMPRESSIBLE %ULER EQUATIONS ARE A GOOD CAN

DIDATE FOR THIS SITUATION -OREOVER IT IS NOT INCONCEIVABLE THAT THIS MOST
UNSTABLE OF ALL KNOWN EQUATIONS WOULD EXHIBIT THE FOLLOWING PERVERSE SCE
NARIO 4HE SET OF ALL INITIAL DATA WHICH LEAD TO GLOBAL REGULAR SOLUTIONS HAS
MEASURE ZERO YET IT IS DENSE IN THE SET OF ALL REGULAR INITIAL CONDITIONS RELA
TIVE TO A REASONABLE TOPOLOGY 3UCH A POSSIBILITY WHICH CANNOT BE RULED OUT
WOULD CERTAINLY EXPLAIN WHY IT IS SO DInCULT TO MAKE ANY PROGRESS ON THE
$ %ULER EQUATIONS WITH OUR PRESENT TECHNIQUES )T WOULD ALSO EXPLAIN IN
PARTICULAR WHY IT IS SO DInCULT TO PRODUCE SPECIlC EXAMPLES OR NUMERICAL
EVIDENCE OF THE WIDELY EXPECTED lNITE TIME BREAKDOWN OF SOLUTIONS
2EMARK 4HE DEVELOPMENT OF METHODS WHICH WOULD ALLOW US TO PROVE
GENERIC GLOBAL RESULTS MAY BE VIEWED AS ONE OF THE GREAT CHALLENGES FOR THE
SUBJECT OF 0$%S IN THE NEXT CENTURY
)T IS EXPECTED THAT THE GLOBAL STRUCTURE OF SINGULARITIES IN 'ENERAL 2EL
ATIVITY WILL HAVE TO BE PHRASED IN TERMS OF GENERIC CONDITIONS SEE ;!-=
AND ;7= FOR UP TO DATE SURVEYS CONCERNING #OSMIC #ENSORSHIP AND RECENT
MATHEMATICAL PROGRESS ON IT 5NDERSTANDING THE PROBLEM OF TURBULENCE
FOR THE .AVIER[3TOKES EQUATIONS WOULD ALMOST CERTAINLY REQUIRE A STATISTICAL
APPROACH 4HE EkECTIVENESS OF MANY GEOMETRIC mOWS IS HINDERED BY THE
PRESENCE OF BAD SEEMINGLY NONGENERIC TYPE OF SINGULARITIES 3O FAR THE
SUBJECT OF NONLINEAR 0$%S HAS BEEN DOMINATED BY METHODS WELL SUITED FOR
THE STUDY OF INDIVIDUAL SOLUTIONS WE HAVE HAD VERY LITTLE SUCCESS IN DEALING
WITH FAMILIES OF SOLUTIONS "Y COMPARISON IN THE CASE OF lNITE DIMENSIONAL
(AMILTONIAN SYSTEMS THE NATURAL ,IOUVILLE MEASURE DElNED IN THE SPACE
PHASE ALLOWS ONE TO PROVE NONTRIVIAL GENERIC RESULTS SUCH AS 0OINCARsES
RECURRENCE THEOREM
4HE 0ROBLEM OF 7ELL POSEDNESS FOR .ONLINEAR %QUATIONS

7ITH THE EXCEPTION OF THE A PRIORI ESTIMATES DERIVED FROM CONSERVATION LAWS
OR MONOTONICITY AND MAXIMUM PRINCIPLE FOR ELLIPTIC OR PARABOLIC EQUATIONS
ALMOST ALL METHODS CURRENTLY USED TO DEAL WITH NONLINEAR 0$%S DEPEND ON
ELABORATE COMPARISON ARGUMENTS BETWEEN SOLUTIONS TO THE ORIGINAL SYSTEM
AND THOSE OF AN APPROPRIATE LINEARIZATION OF IT )T IS ESSENTIAL TO HAVE VERY
PRECISE ESTIMATES FOR THE LINEAR SYSTEM IN TUNE WITH THE A PRIORI ESTIMATES

4HERE EXIST SOME INTERESTING GENERIC RESULTS IN 0$% ALSO BASED ON THE CONSTRUCTION OF
'IBBS MEASURES ON THE SPACE OF SOLUTIONS SEE ;"= 5NFORTUNATELY THE CLASS OF EQUATIONS
FOR WHICH SUCH MEASURES CAN BE CONSTRUCTED IS EXTREMELY LIMITED
6ISIONS IN -ATH 0$% !3 ! 5.)&)%$ 35"*%#4
AND THE SCALING PROPERTIES OF THE NONLINEAR EQUATIONS )N THE CASE OF ELLIPTIC
AND PARABOLIC PROBLEMS WE HAVE A LARGE AND POWERFUL ARSENAL OF SUCH ESTI
MATES ALMOST ALL DEVELOPED DURING THE COURSE OF THIS CENTURY SEE ;"RE"=
/UR KNOWLEDGE OF LINEAR ESTIMATES FOR HYPERBOLIC AND DISPERSIVE EQUATIONS
IS FAR LESS SATISFACTORY
4HE NEED FOR A WELL ADAPTED LINEAR THEORY FOR EVOLUTION EQUATIONS CAN
BE BEST UNDERSTOOD FROM THE PERSPECTIVE OF THE PROBLEM OF OPTIMAL WELL
POSEDNESS )N WHAT FOLLOWS ) WILL LIMIT MY DISCUSSION TO lELD THEORIES SUCH
AS THE NONLINEAR SCALAR WAVE EQUATION .37% 9ANG[-ILLS 9- 7AVE
-APS 7- AND THE %INSTEIN 6ACUUM %6 EQUATIONS -Y GOAL IS TO WRITE
DOWN THREE SPECIlC CONJECTURES 70[70 WHICH ARE ) FEEL JUST BEYOND
THE BOUNDARY OF WHAT CAN BE OBTAINED WITH PRESENT DAY TECHNIQUES 4HEY
ARE THUS BOTH ACCESSIBLE AND IMPORTANT TO GENERATE INTERESTING MATHEMATICS
4HE INITIAL VALUE PROBLEM FOR AN EVOLUTIONARY SYSTEM OF EQUATIONS IS SAID
TO BE WELL POSED 70 RELATIVE TO A "ANACH FUNCTIONAL SPACE 8 IF FOR ANY
DATA IN 8 THERE EXIST UNIQUELY DElNED LOCAL IN TIME SOLUTIONS BELONGING TO
8 FOR T AND DEPENDING CONTINUOUSLY ON THE DATA 4HE PROBLEM IS SAID
TO BE STRONGLY 70 IF THE DEPENDENCE ON THE DATA IS ANALYTIC AND WEAKLY
70 IF THE DEPENDENCE IS MERELY CONTINUOUS OR DIkERENTIABLE )N THE CASE OF
HYPERBOLIC EQUATIONS ESPECIALLY QUASILINEAR THERE IS A NATURAL APPARENTLY
UNIQUE CHOICE FOR 8 ,OCALLY IT HAS TO COINCIDE WITH THE 3OBOLEV SPACE
( S 2N 4HIS IS DUE TO THE FACT THAT ,P NORMS ARE NOT PRESERVED BY THE
LINEAR EVOLUTION IN DIMENSION N WHILE NORMS DElNED IN &OURIER SPACE
ARE MEANINGLESS FOR QUASILINEAR EQUATIONS 4AKING INTO ACCOUNT THE SCALING
PROPERTIES OF THE BASIC lELD EQUATIONS AND PROCEEDING IN THE SAME MANNER
AS IN THE PREVIOUS SECTION ONE CAN DElNE THE CRITICAL 70 EXPONENT SC TO
BE THAT VALUE OF S FOR WHICH THE ( S NORM OF INITIAL DATA IS DIMENSIONLESS
7ITH THIS DElNITION WE CAN FORMULATE THE FOLLOWING
'ENERAL 70 CONJECTURE I &OR ALL BASIC lELD THEORIES THE INITIAL VALUE
PROBLEM IS LOCALLY STRONGLY WELL POSED FOR ANY DATA IN ( S S SC
II 4HE BASIC lELD THEORIES ARE WEAKLY GLOBALLY WELL POSED FOR ALL INITIAL
DATA WITH SMALL ( SC NORM
III 4HERE CAN BE NO WELL DElNED SOLUTIONS FOR S SC

7E TALK OF A SPACE ( S RATHER THAN A PAIR ( S ( S_ 4HUS IN THE CASE OF THE )60 FOR
THE WAVE EQUATION c} } F T } G F G ( S MEANS F ( S G ( S_

7EAK SOLUTIONS MAY EXIST BELOW THE SC THRESHOLD BUT ARE COMPLETELY UNSTABLE AND
HAVE WEIRD PROPERTIES )N OTHER WORDS WEAK SOLUTIONS CORRESPONDING TO S SC ARE
MATHEMATICAL <GHOSTS
3 +,!).%2-!. '!&!
4HE PROOF OF THE 70 CONJECTURE FOR S t SC WILL PROVIDE US WITH AN

ESSENTIAL TOOL FOR THE PROBLEM OF REGULARITY DISCUSSED IN THE PREVIOUS SEC
TION 3O FAR THE CONJECTURE WAS PROVED ONLY IN THE CASE OF .37% SEE ;+=
AND ;3H3= IT IS BASED ENTIRELY ON 3TRICHARTZ TYPE INEQUALITIES 3EMILINEAR
EQUATIONS WHOSE NONLINEAR TERMS INVOLVE DERIVATIVES SUCH AS 9- AND 7-
ARE FAR MORE DInCULT SEE DISCUSSION BELOW 4HE CASE S SC IS INTERESTING
FOR A PHILOSOPHICAL REASON 4HERE ARE SUPERCRITICAL CASES IN THE CASE OF
THE SUPERCRITICAL .37% SEE ;3TR= FOR THE CASE OF 7- SEE ;3H= ;- U3= FOR
WHICH ONE CAN PROVE THE EXISTENCE OF A WEAK SOLUTION CORRESPONDING TO ANY
lNITE ENERGY INITIAL CONDITIONS 0ART III OF THE ABOVE CONJECTURE ASSERTS
THAT THESE SOLUTIONS ARE UNSTABLE IT IS EASY IN FACT TO SEE THAT THEY ARE
LINEARLY UNSTABLE AND THEREFORE NOT PARTICULARLY USEFUL )T IS INTERESTING TO
REMARK IN THIS RESPECT THE RECENT REMARKABLE RESULT OF 3CHAEkER ;3= SEE
ALSO ;3HN= 3CHAEkER HAS CONSTRUCTED EXAMPLES OF WEAK SOLUTIONS FOR THE $
%ULER EQUATIONS WHICH ARE COMPACTLY SUPPORTED IN SPACE TIME 4HE RESULT
IS REMINISCENT OF THE FAMOUS RESULT OF .ASH ;.= SEE ALSO +UIPER ;+U= ON #
ISOMETRIC IMBEDDINGS WHICH TURN OUT TO BE PLENTIFUL DENSE IN THE SET OF ALL
SMOOTH FUNCTIONS AND A LOT MORE PLIANT THAN THE MORE REGULAR ONES !N
OTHER REMARKABLE EXAMPLE OF HOW BAD WEAK SOLUTIONS CAN SOMETIMES BE IS
THAT OF 2IVIrERE CONCERNING WEAK HARMONIC MAPS FROM A THREE DIMENSIONAL
SPACE TO 3 WITH A DENSE SET OF SINGULARITIES ;2= 4HIS IS IN SHARP CONTRAST
TO THE CASE OF MINIMIZERS ;3CH5= OR STATIONARY SOLUTIONS ;%= FOR THE SAME
EQUATIONS ) SUSPECT THAT SIMILAR UNACCEPTABLE PROPERTIES OF WEAK SOLUTIONS
TYPE RESULTS CAN BE PROVED FOR SOLUTIONS TO NONLINEAR WAVE EQUATIONS BELOW
THE CRITICAL REGULARITY -OREOVER SHORT OF ADDITIONAL REGULARITY ASSUMPTIONS
ON THE INITIAL DATA THERE MAY EXIST NO ENTROPY TYPE CONDITIONS WHICH WOULD
STABLIZE THE SOLUTIONS
)N THE CASE OF SUBCRITICAL EQUATIONS FOR WHICH THE ENERGY NORM IS STRONGER
THAN ( SC PART I OF THE CONJECTURE WOULD IMPLY WELL POSEDNESS IN THE EN
ERGY NORM AND THEREFORE BY ENERGY CONSERVATION GLOBAL WELL POSEDNESS
AND REGULARITY )N OTHER WORDS THE SOLUTIONS PRESERVE THE ( S REGULARITY OF
THE DATA FOR ANY S SC 4HIS WOULD THUS SETTLE THE lRST PART OF THE 'ENERAL
#ONJECTURE STATED IN THE PREVIOUS SECTION
)N THE CASE OF CRITICAL EQUATIONS PART II OF THE 70 CONJECTURE WILL IMPLY
THE FOLLOWING
3MALL ENERGY CONJECTURE &OR ALL BASIC CRITICAL lELD THEORIES ALL SMALL

4HIS PHENOMENON HAS BEEN CALLED THE H 0RINCIPLE AND DISCUSSED IN A VERY GENERAL
SET UP BY - 'ROMOV SEE ;'R=
6ISIONS IN -ATH 0$% !3 ! 5.)&)%$ 35"*%#4
ENERGY SOLUTIONS ARE GLOBALLY REGULAR

4HE SMALL ENERGY CONJECTURE IS AN ESSENTIAL STEP IN THE PROOF OF THE GEN
ERAL REGULARITY CONJECTURE FOR CRITICAL lELD THEORIES )N THE CASE OF WAVE
EQUATIONS WHOSE NONLINEARITIES DO NOT DEPEND ON DERIVATIVES OR IN THE CASE
OF SPHERICAL SYMMETRIC SOLUTIONS ONE CAN PROVE IT DIRECTLY )N THE CASE
OF EQUATIONS LIKE 9- OR 7- WITH DERIVATIVES APPEARING IN THE NONLINEAR
TERMS IT IS NOW BELIEVED THAT THE ONLY WAY TO SETTLE THE SMALL ENERGY CON
JECTURE IS TO PROVE THE MUCH STRONGER PART II OF THE 70 CONJECTURE )N
WHAT FOLLOWS ) WILL GIVE A MORE PRECISE FORMULATION OF IT FOR THE SPECIAL CASE
OF THE 7- AND 9- EQUATIONS
#ONJECTURE 70 4HE 7AVE -APS EQUATION DElNED FROM THE -INKOWSKI
2N TO A COMPLETE 2IEMANNIAN TARGET MANIFOLD IS GLOBALLY WELL POSED
FOR SMALL INITIAL DATA IN ( N N t
#ONJECTURE 70 4HE 9ANG[-ILLS EQUATION FOR 3/. 35 . STRUC
N_
TURE GROUPS IS GLOBALLY WELL POSED FOR SMALL INITIAL DATA IN ( N t
4O UNDERSTAND THE DInCULTIES INVOLVED IN 70 ) WILL SUMMARIZE BELOW
WHAT ARE THE MOST SIGNIlCANT KNOWN RESULTS IN CONNECTION TO IT
4HE CONJECTURE IS TRUE IN THE CASE OF EQUIVARIANT WAVE MAPS SEE ;3H:=
IN WHICH CASE THE NONLINEAR TERMS DO NOT DEPEND ON DERIVATIVES )N ;#HR:=
THE SMALL ENERGY CONJECTURE WAS PROVED FOR THE SPECIAL CASE OF SPHERICALLY
SYMMETRIC SOLUTIONS 4HEIR APPROACH AVOIDS THE PROOF OF THE 70 CONJEC
TURE WHICH IS STILL NOT KNOWN EVEN IN THE SPHERICALLY SYMMETRIC CASE BY
PROVING DIRECTLY IN THIS CASE THE SMALL ENERGY CONJECTURE )N THE GENERAL
CASE IT DOES NOT SEEM POSSIBLE TO PROVE THE SMALL ENERGY CONJECTURE INDE
PENDENT OF #ONJECTURE 70 4HIS HAS TO DO WITH THE LACK OF ANY SPACE TIME
,P P lRST DERIVATIVE ESTIMATES SEE ;7O= FOR SOLUTIONS TO c} &
)N ;+L-= AND ;+L3= ONE PROVES LOCAL WELL POSEDNESS FOR ALL DATA IN
( S S SC N N t SEE ALSO ;+E4= FOR N 4HE RESULT DEPENDS
HEAVILY ON BILINEAR ESTIMATES 4HIS WAS FURTHER IMPROVED IN ;4= WHO HAS
ESTABLISHED WELL POSEDNESS HIS RESULT IS IN FACT GLOBAL IN TIME IN VIEW OF THE
N
SCALING PROPERTIES OF THE EQUATIONS FOR SMALL DATA IN THE "ESOV SPACE "
"OTH ABOVE MENTIONED RESULTS FAIL TO TO TAKE INTO ACCOUNT THE COMPLETENESS
OF THE TARGET MANIFOLD
7E KNOW FROM SIMPLE EXAMPLES THAT WE MAY NOT HAVE ( N WELL
POSEDNESS IF THE TARGET MANIFOLD IS NOT COMPLETE
4HE DEPENDENCE OF SOLUTIONS ON THE DATA WITH RESPECT TO THE ( N
NORM CANNOT BE TWICE DIkERENTIABLE
3 +,!).%2-!. '!&!
4HE METHODS WHICH HAVE BEEN USED TO TACKLE THE CASE S SC DEPEND
HEAVILY ON AN ITERATIVE PROCEDURE IN WHICH ONE ESTIMATES THE ( Sm NORM OF
EACH ITERATE FOR S N m IN TERMS OF THE ( Sm NORMS OF THE PREVIOUS
ITERATES 4HESE NORMS DElNED WITH RESPECT TO THE SPACE TIME &OURIER TRANS
FORM ARE INTIMATELY TIED TO THE SYMBOL OF c AND TO BILINEAR ESTIMATES SEE
;+L-= ;+L3= AND ;&O+= 3IMILAR NORMS WHERE INTRODUCED BY * "OURGAIN
;"= SEE ALSO ;+EN06= IN CONNECTION WITH NONLINEAR DISPERSIVE EQUATIONS
4O TREAT THE CRITICAL CASE ONE NEEDS TO OVERCOME TWO DInCULTIES 4HE
lRST HAS TO DO WITH IMPROVING THE ESTIMATES AT EACH ITERATIVE STEP TO MAKE
THEM OPTIMAL 4HE SECOND IS AN IMPORTANT CONCEPTUAL DInCULTY WHICH HAS
TO DO WITH THE ITERATIVE PROCESS ITSELF !NY ITERATIVE PROCEDURE IF SUCCESS
FUL WOULD IMPLY NOT ONLY WELL POSEDNESS BUT ALSO ANALYTIC DEPENDENCE ON
THE DATA IN THE ( N NORM 4HIS IS HOWEVER WRONG ACCORDING TO THE OB
SERVATION ABOVE 4O UNDERSTAND THIS EkECT CONSIDER THE (ILBERT SPACE
8 ( N 2N U A FUNCTION IN 8 AND LET hT EITU )T IS KNOWN THAT
hT IS A # FUNCTION OF T WITH VALUES IN 8 BUT SINCE 8 IS NOT CLOSED UNDER
MULTIPLICATIONS IT IS NOT IN # SEE ;+E4= 4HE REASON EITU 8 IS DUE
TO THE FACT THAT THE FUNCTION EIU IS BOUNDED IT CANNOT BE GUESSED BY JUST
0
CONSIDERING THE 4AYLOR EXPANSION EIU Nt N IU N IN WHICH ALL TERMS ARE
DIVERGENT
)N THE CASE OF THE 7- EQUATIONS ANY ITERATIVE PROCEDURE LOSES THE CRUCIAL
INFORMATION ABOUT THE COMPLETENESS OF THE TARGET MANIFOLD AND THEREFORE
LEADS TO LOGARITHMIC DIVERGENCES 4O SEE THAT CONSIDER 7- SOLUTIONS OF THE
FORM } lU WHERE cU WITH DATA IN ( N AND l IS A GEODESIC OF THE
TARGET MANIFOLD - 3INCE THE , NORM OF U IS NOT CONTROLLED lU MAKES
SENSE ONLY IF THE GEODESIC IS GLOBALLY DElNED ! STANDARD ITERATION FAILS TO
DISTINGUISH BETWEEN COMPLETE AND INCOMPLETE GEODESICS
4HIS SITUATION SEEMS TO CALL FOR A <RENORMALIZATION PROCEDURE -ORE
PRECISELY ONE MAY HOPE THAT BY UNDERSTANDING THE NATURE OF THE LOGARITHMIC
DIVERGENCES OF EACH ITERATE WE CAN OVERCOME THEM BY A CLEVER REGULARIZA
TION AND LIMITING PROCEDURE )N VIEW OF THE SIMPLE MINDED MODEL PROBLEM
STUDIED IN ;+L-= ONE MAY HOPE THAT SUCH AN APPROACH IS NOT IMPOSSIBLE
) WILL ONLY MAKE A FEW REMARKS CONCERNING THE 70 CONJECTURE 4HE
OPTIMAL KNOWN RESULT IN DIMENSION N t IS SMALL DATA WELL POSEDNESS FOR
S SC SEE ;+L-= AND ;+L4= )N THE CASE S SC IT CAN BE SHOWN THAT ANY
ITERATION PROCEDURE LEADS TO LOGARITHMIC DIVERGENCES 4HE SITUATION SEEMS
THUS SIMILAR TO THAT DESCRIBED IN THE PREVIOUS CONJECTURE )N DIMENSION
N WE HAVE GLOBAL WELL POSEDNESS IN THE ENERGY NORM S SEE ;+L-=
6ISIONS IN -ATH 0$% !3 ! 5.)&)%$ 35"*%#4
AND THE DISCUSSION IN CONNECTION TO 70 BELOW AND LOCAL WELL POSEDNESS
FOR S )T IS NOT AT ALL KNOWN WHAT HAPPENS FOR SC S s
4HE CASE OF THE %INSTEIN 6ACUUM EQUATIONS IS FAR MORE DInCULT THAN
THAT OF 7- OR 9- 7RITTEN RELATIVE TO WAVE COORDINATES THE %6 EQUATIONS
TAKE THE FORM G uv u v Gjk . G G WHERE G IS A ,ORENTZ METRIC AND .
IS A NONLINEAR TERM QUADRATIC IN THE lRST DERIVATIVES OF G 4HIS FORM OF THE
%INSTEIN EQUATION LEADS TO THE STUDY OF QUASILINEAR WAVE EQUATIONS OF THE
FORM
cG} } `} 1R } }
WITH G} A ,ORENTZ METRIC DEPENDING SMOOTHLY ON } ` SMOOTH FUNCTION OF
} AND 1R } } QUADRATIC IN } /THER TYPES OF QUASILINEAR WAVE EQUA
TIONS SUCH AS THOSE APPEARING IN %LASTICITY OR #OMPRESSIBLE &LUIDS DEPEND
ING ONLY ON } CAN BE WRITTEN AS SYSTEMS OF WAVE EQUATIONS OF TYPE
5SING ENERGY ESTIMATES AND 3OBOLEV INEQUALITIES ONE CAN PROVE THE <CLAS
SICAL LOCAL EXISTENCE RESULT OR LOCAL WELL POSEDNESS FOR ( S INITIAL DATA WITH
S SC N 4HIS RESULT LEADS IN THE CASE OF THE %6 EXPRESSED
RELATIVE TO WAVE COORDINATES TO THE WELL KNOWN LOCAL EXISTENCE RESULT OF
9# "RUHAT "RUHATS RESULT SEE ;"RU= REQUIRES IN FACT MORE DERIVATIVES
OF THE DATA 4HE OPTIMAL DIMENSIONAL RESULT S SC WAS
PROVED IN ;&-=
'ETTING CLOSE TO THE CRITICAL EXPONENT S SC IS ENTIRELY OUT OF
REACH ) BELIEVE HOWEVER THAT THE INTERMEDIATE RESULT S IS BOTH VERY
INTERESTING AND ACCESSIBLE
#ONJECTURE 70 4HE %INSTEIN 6ACUUM EQUATIONS ARE STRONGLY LOCALLY
WELL POSED FOR INITIAL DATA SETS f G K FOR WHICH 2ICG , f AND
K ( f
4HE CONJECTURE CAN BE VIEWED IN A SENSE AS A FAR MORE DInCULT ANALOGUE
OF THE WELL POSEDNESS RESULT SEE ;+L-= FOR THE 9- EQUATIONS IN THE
ENERGY NORM 7RITING THE 9- EQUATIONS IN THE ,ORENTZ GAUGE WHICH IS THE
PRECISE ANALOGUE OF WAVE COORDINATES ONE IS LED TO A SYSTEM OF EQUATIONS OF
THE FORM
c} 1R} } #}
WITH 1R QUADRATIC IN } } AND LINEAR IN } AND # CUBIC IN } )N THIS
CASE THE SCALING EXPONENT IS SC N_ 4HE CLASSICAL LOCAL EXISTENCE RESULT
BASED ON ENERGY ESTIMATES AND THE ( z y , z N 3OBOLEV ESTIMATE

f G IS A 2IEMANNIAN $ MANIFOLD AND K A SYMMETRIC TENSOR VERIFYING THE CON
STRAINT EQUATIONS
3 +,!).%2-!. '!&!
REQUIRES DATA IN ( S S SC /NE CAN IMPROVE THE RESULT TO S SC

FOR N AND S SC IN HIGHER DIMENSIONS BY USING THE CLASSICAL
3TRICHARTZ TYPE INEQUALITIES FOR SOLUTIONS TO THE INHOMOGENEOUS STANDARD
WAVE EQUATION c} & -OREOVER ONE CAN SHOW SEE ;,= THAT FOR N THE
RESULT S SC IS OPTIMAL FOR GENERAL EQUATIONS OF TYPE 4HEREFORE
TO PROVE THE ( WELL POSEDNESS RESULT FOR THE 9ANG[-ILLS EQUATIONS ONE
NEEDS TO TAKE ADVANTAGE OF SOME ADDITIONAL CANCELLATIONS PRESENT IN THE
NONLINEAR TERMS /NE CAN DO THAT BY USING THE <GAUGE COVARIANCE OF THE
9ANG[-ILLS EQUATIONS ACCORDING TO WHICH A SOLUTION OF 9- IS A CLASS OF
EQUIVALENCE OF SOLUTIONS RELATIVE TO GAUGE TRANSFORMATIONS )N VIEW OF THIS
ONE IS FREE TO PICK THE PARTICULAR GAUGE CONDITIONS BEST SUITED TO THE PROBLEM
AT HAND )N ;+L-= THE CHOICE OF THE #OULOMB GAUGE LEADS TO A COUPLED
SYSTEM OF ELLIPTIC HYPERBOLIC EQUATIONS WHICH SATISlES THE <NULL CONDITION
4HIS MEANS VERY ROUGHLY THAT THE HYPERBOLIC PART OF THE 9- #OULOMB
SYSTEM HAS THE FORM
c} 1} } BETTER BEHAVED TERMS
WITH 1} } A NONLOCAL <NULL QUADRATIC FORM 4O DEAL WITH THE CANCELLA
TIONS PRESENT IN THE NULL QUADRATIC FORMS 1 ONE HAS DEVELOPED THE SO CALLED
BILINEAR ESTIMATES SEE ;+L-= ;&O+=
)N TRYING TO IMPLEMENT A SIMILAR STRATEGY TO %6 ONE ENCOUNTERS FUN
DAMENTAL DInCULTIES DUE THE QUASILINEAR CHARACTER OF THE %INSTEIN EQUA
TIONS &OR EXAMPLE TO IMPROVE "RUHATS CLASSICAL LOCAL EXISTENCE RESULT
FROM S SC TO S SC IN WAVE COORDINATES ONE NEEDS TO PROVE A
VERSION OF THE CLASSICAL 3TRICHARTZ ESTIMATES FOR c REPLACED WITH THE WAVE
OPERATOR cG WHERE G IS A ROUGH ASSUMING WE lX } THE METRIC G} WILL
HAVE THE <EXPECTED REGULARITY OF } ,ORENTZ METRIC
5NTIL RECENTLY THIS SEEMED TO BE AN INTRACTABLE PROBLEM )N FACT IT IS
KNOWN THAT IF THE COEnCIENTS OF A LINEAR WAVE EQUATION HAVE LESS REGULARITY
THAN # SOME OF THE MAIN 3TRICHARTZ INEQUALITIES MAY FAIL SEE ;3M3=
( 3MITH SEE ;3M= WAS ALSO ABLE TO SHOW THAT ALL THE 3TRICHARTZ TYPE
INEQUALITIES HOLD TRUE IF THE COEnCIENTS ARE AT LEAST # AND N s SEE ;4
PARTS )) AND )))= FOR N t 4HE # CONDITION HOWEVER IS MUCH TOO STRONG
TO APPLY TO NONLINEAR EQUATIONS
2ECENTLY *9 #HEMIN AND ( "AHOURI SEE ;#H"= HAVE SUCCEEDED IN
DERIVING THE lRST IMPROVEMENT OVER THE CLASSICAL RESULT 4HEY HAVE PROVED

4HE 3TRICHARTZ TYPE INEQUALITIES ARE INTIMATELY TIED TO RESTRICTION RESULTS IN &OURIER
!NALYSIS 4OGETHER WITH THE MORE RECENT BILINEAR ESTIMATES THEY EXEMPLIFY THE STRONG
MODERN TIES BETWEEN (ARMONIC !NALYSIS AND NONLINEAR WAVE EQUATIONS
6ISIONS IN -ATH 0$% !3 ! 5.)&)%$ 35"*%#4
LOCAL 70 FOR EQUATIONS OF TYPE PROVIDED THAT S SC FOR N t AND

S SC FOR N 4HE SAME RESULT WAS PROVED ALSO BY $ 4ATARU ;4=
USING A SOMEWHAT DIkERENT METHOD "OTH #HEMIN "AHOURI AND 4ATARU HAVE
LATER OBTAINED SOME FURTHER IMPROVEMENTS BUT FALL SHORT OF THE EXPECTED
OPTIMAL RESULT 4HE OPTIMAL KNOWN RESULT S SC FOR N t AND
S SC IS PROVED IN ;4 PART )))= )N DIMENSION N WE ALSO HAVE
EXAMPLES DUE TO ( ,INBLAD ;,= WHICH SHOW THAT ONE CANNOT HAVE WELL
POSEDNESS IN GENERAL FOR S s
%VEN IF THE 3TRICHARTZ BASED METHODS INITIATED BY #HEMIN "AHOURI AND
4ATARU CAN BE MADE OPTIMAL THEY WILL STILL FALL SHORT OF PROVING THE DESIRED
( RESULT CONJECTURED BY 70 4O OBTAIN SUCH A RESULT ONE NEEDS TO TAKE
INTO ACCOUNT THE <NULL STRUCTURE OF THE %6 EQUATIONS 7E KNOW INDIRECTLY
FROM THE PROOF OF STABILITY OF THE -INKOWSKI SPACE ;#HR+= THAT WRITTEN
IN APPROPRIATE FORM IE USING THEIR GENERAL COVARIANCE THE EQUATIONS MUST
EXHIBIT SUCH A STRUCTURE 9ET THE INDIRECT METHOD OF ;#HR+= BASED ON THE
"IANCHI IDENTITIES AND A CAREFUL DECOMPOSITION OF ALL GEOMETRIC COMPONENTS
APPEARING IN THE EQUATION RELATIVE TO A NULL FRAME CANNOT BE USED IN THIS
CASE /NE NEEDS INSTEAD A METHOD SIMILAR TO THE ONE WE HAVE SKETCHED
ABOVE FOR 9- )N OTHER WORDS WE NEED A <GAUGE CONDITION SIMILAR TO THE
#OULOMB ONE IN 9- RELATIVE TO WHICH ALL QUADRATIC TERMS OF THE %INSTEIN
EQUATIONS EXHIBIT A NULL BILINEAR STRUCTURE /NCE THIS IS DONE WE NEED TO
DEVELOP TECHNIQUES TO PROVE BILINEAR ESTIMATES SIMILAR TO THOSE OF ;+L-=
;&O+= IN A QUASILINEAR SET UP ! GOOD WARM UP PROBLEM IN THIS RESPECT
WOULD BE THE STUDY OF THE -INKOWSKI SPACE ANALOGUE OF THE MINIMAL SURFACE
EQUATION FOR WHICH THE NULL STRUCTURE IN THE SENSE OF ;+L= ;#HR= IS
OBVIOUS
4O SUMMARIZE THE STUDY OF #ONJECTURE 70 REQUIRES
4O DEVELOP NEW ANALYTIC TECHNIQUES TO IMPROVE THE RESULTS OF #HEMIN
"AHOURI TO THE OPTIMAL REGULARITY POSSIBLE FOR 3TRICHARTZ BASED METH
ODS
4O INVESTIGATE QUASILINEAR EQUATIONS WHICH VERIFY THE NULL CONDITION
AND DEVELOP BILINEAR ESTIMATES FOR LINEAR EQUATIONS WITH VERY ROUGH
COEnCIENTS
4O INVESTIGATE IN A DIRECT WAY THE NULL STRUCTURE OF THE %INSTEIN
EQUATIONS

4HE BILINEAR ESTIMATES OF ;+L-= HAVE BEEN RECENTLY DERIVED BY 3MITH AND 3OGGE
;3M3= FOR # COEnCIENTS
3 +,!).%2-!. '!&!
2EFERENCES
;!-= , !NDERSON 6 -ONCRIEF 4HE GLOBAL EXISTENCE PROBLEM IN GENERAL
RELATIVITY PREPRINT
;"= * "OURGAIN .ONLINEAR 3CHRODINGER EQUATIONS IN <.ONLINEAR 7AVE %QUA
TIONS AND &REQUENCY )NTERACTIONS !-3 SERIES 0ARK #ITY
;"= * "OURGAIN (ARMONIC ANALYSIS AND NONLINEAR 0$%S 0ROCEEDINGS OF
)#- :URICH
;"= * "OURGAIN &OURIER TRANSFORM RESTRICTION PHENOMENA FOR CERTAIN LAT
TICE SUBSETS AND APPLICATIONS TO NONLINEAR WAVE EQUATIONS ) 3CHRODINGER
EQUATIONS )) 4HE +D6 EQUATION '!&! [ [
;"R= 9 "RENIER -INIMAL GEODESICS ON GROUPS OF VOLUME PRESERVING MAPS
#OMM 0URE !PPL -ATH [
;"RE"= ( "REZIS & "ROWDER 0ARTIAL DIkERENTIAL EQUATIONS IN THE TH CEN
TURY %NCYCLOPEDIA )TALIANA IN ITS SERIES ON THE HISTORY OF THE TWENTIETH
CENTURY TO APPEAR
;"RU= 9#H "RUHAT 4HEOREMES DEXISTENCE POUR CERTAINS SYSTEMES DEQUATIONS
AUX DERIVEE PARTIELLES NONLINEAIRES !CTA -ATH [
;#79= ! #HANG , 7ANG 0 9ANG ! REGULARITY OF BIHARMONIC MAPS
#0!- ,)) [
;#H"= *9 #HEMIN ( "AHOURI %QUATIONS DONDES QUASILINEAIRES ET EkECT
DISPERSIF !*- TO APPEAR
;#HR= $ #HRISTODOULOU 4HE !CTION 0RINCIPLE AND 0$%S !NNALS OF -ATH
3TUDIES
;#HR= $ #HRISTODOULOU 'LOBAL SOLUTIONS OF NONLINEAR HYPERBOLIC EQUATIONS
FOR SMALL INITIAL DATA #0!- [
;#HR+= $ #HRISTODOULOU 3 +LAINERMAN 4HE 'LOBAL .ONLINEAR 3TABILITY
OF THE -INKOWSKI 3PACE 0RINCETON -ATHEMATICAL 3ERIES
;#HR:= $ #HRISTODOULOU !34 :ADEH /N THE REGULARITY OF SPHERICALLY
SYMMETRIC WAVE MAPS #OMM 0 !PPL -ATH [
;%= ,# %VANS 0ARTIAL REGULARITY FOR HARMONIC MAPS INTO SPHERES !RCH 2AT
-ECH !NAL [
;&-= ! &ISCHER * -ARSDEN 4HE %INSTEIN EVOLUTION EQUATIONS AS A lRST
ORDER QUASILINEAR SYMMETRIC HYPERBOLIC SYSTEM #OMM -ATH 0HYS
[

7ITH FEW EXCEPTIONS MOST OF THE REFERENCES GIVEN BELOW ARE IN REGARD WITH THE
LAST SECTION OF THE PAPER MORE PRECISELY IN CONNECTION TO THE CONJECTURES 70[70 )
APOLOGIZE FOR GIVING ONLY A VERY LIMITED NUMBER OF REFERENCES IN CONNECTION TO THE lRST
FOUR SECTIONS
6ISIONS IN -ATH 0$% !3 ! 5.)&)%$ 35"*%#4
;&-O= ! &ISCHER 6 -ONCRIEF 4HE %INSTEIN mOW THE SIGMA CONSTANT AND
THE GEOMETRIZATION OF THREE MANIFOLDS PREPRINT
;&O+= $ &OSCHI 3 +LAINERMAN /N BILINEAR ESTIMATES FOR SOLUTIONS TO THE
WAVE EQUATION !NNALES %.3 TO APPEAR
;''+-= #3 'ARDNER *- 'REEN -$ +RUSKAL 2- -IURA -ETHOD
FOR SOLVING THE +D6 EQUATION 0HYS 2EV ,ETT [
;'R= - 'ROMOV 0ARTIAL $IkERENTIAL 2ELATIONS 3PRINGER 6ERLAG "ERLIN
;(4= 3 (ILDEBRANDT ! 4ROMBA -ATHEMATICS AND /PTIMAL &ORM 3CIEN
TIlC !MERICAN ,IBRARY
;(U)= ' (UISKEN 4 )LMANEN 4HE INVERSE MEAN CURVATURE mOW AND THE 0EN
ROSE CONJECTURE *$' TO APPEAR
;*= & *OHN &ORMATION OF SINGULARITIES IN ELASTIC WAVES 3PRINGER ,ECTURE
.OTES IN 0HYS [
;+= , +APITANKY 'LOBAL AND UNIQUE WEAK SOLUTIONS OF NONLINEAR WAVE EQUA
TIONS -ATH 2ES ,ETT
;+E4= - +EEL 4 4AO ,OCAL AND GLOBAL WELL POSEDNESS OF WAVE MAPS ON 2
FOR ROUGH DATA )-2. [
;+EN06= # +ENIG ' 0ONCE , 6EGA 4HE #AUCHY PROBLEM FOR THE +D6
EQUATION IN 3OBOLEV SPACES OF NEGATIVE INDICES $UKE -ATH *
[
;+L= 3 +LAINERMAN ,ONG TIME BEHAVIOR OF SOLUTIONS TO NONLINEAR WAVE EQUA
TIONS 0ROC )#- 7ARSZAWA [
;+L= 3 +LAINERMAN 4HE NULL CONDITION AND GLOBAL EXISTENCE TO NONLINEAR WAVE
EQUATIONS !-3 ,ECTURES IN !PPLIED -ATHEMATICS [
;+L-= 3 +LAINERMAN - -ACHEDON 3PACE TIME ESTIMATES FOR NULL FORMS AND
THE LOCAL EXISTENCE THEOREM #OMM 0URE !PPL -ATH [

;+L-= 3 +LAINERMAN - -ACHEDON &INITE ENERGY SOLUTIONS FOR THE 9ANG[
-ILLS SOLUTIONS IN 2 !NNALS OF -ATH [
;+L-= 3 +LAINERMAN - -ACHEDON 3MOOTHING ESTIMATES FOR NULL FORMS
AND APPLICATIONS $UKE -ATH * [
;+L-= 3 +LAINERMAN - -ACHEDON /N THE ALGEBRAIC PROPERTIES OF THE
(N SPACES )-2. [
;+L-= 3 +LAINERMAN - -ACHEDON /N THE OPTIMAL LOCAL REGULARITY FOR
'AUGE lELD THEORIES $Ik )NTEGR %QS [
;+L3= 3 +LAINERMAN 3 3ELBERG 2EMARK ON THE OPTIMAL REGULARITY FOR EQUA
TIONS OF WAVE MAPS TYPE #OMM 0$% [
;+L4= 3 +LAINERMAN $ 4ATARU /N THE OPTIMAL LOCAL REGULARITY FOR THE
9ANG[-ILLS EQUATIONS * !-3 [
;+U= .( +UIPER /N # ISOMETRIC EMBEDDINGS ) 0ROC +ONINKL .EDER !K
7ET ! [
;,= ( ,INBLAD #OUNTEREXAMPLES TO LOCAL EXISTENCE FOR SEMILINEAR WAVE EQUA
3 +,!).%2-!. '!&!
TIONS !*- [

;,= ( ,INDBLAD #OUNTEREXAMPLES TO LOCAL EXISTENCE FOR QUASILINEAR WAVE
EQUATIONS -2, TO APPEAR
;,U= * ,UTZEN 4HE 0REHISTORY OF THE 4HEORY OF $ISTRIBUTIONS 3PRINGER
6ERLAG
;-= ( -INKOWSKI 3PACE AND 4IME %NGLISH TRANSLATION OF THE ORIGINAL ARTICLE
IN <4HE -EANING OF 2ELATIVITY $OVER
;- U3= 3 -U LLER - 3TRUWE 'LOBAL EXISTENCE FOR WAVE MAPS IN DIMEN
SIONS FOR lNITE ENERGY DATA PREPRINT
;.= * .ASH # ISOMETRIC EMBEDDINGS !NN OF -ATH [
;.E= 9 .EEMAN 0YTHAGOREAN AND 0LATONIC CONCEPTIONS IN 88TH #ENTURY
PHYSICS IN THIS ISSUE
;0= ( 0OINCARs E 3UR LES RAPPORTS DE LANALYSE PURE ET DE LA PHYSIQUE MATH
EMATIQUE &IRST )NT #ONGRESS OF -ATHEMATICIANS :URICH
;2= 4 2IVIr ERE !PPLICATIONS HARMONIQUE DE " DANS 3 PARTOUT DISCONNUES
#2!CAD 3CI 0ARIS [
;3= 6 3CHAEFFER !N INVISCID mOW WITH COMPACT SUPPORT IN SPACE TIME
*OURN 'EOM !NAL [
;3C= 2 3CHOEN ! REPORT ON SOME RECENT PROGRESS ON NONLINEAR PROBLEMS IN
GEOMETRY 3URVEYS IN $IkERENTIAL 'EOMETRY [
;3C5= 2 3CHOEN + 5HLENBECK ! REGULARITY THEORY FOR HARMONIC MAPS *
$Ik 'EOM [
;3H= * 3HATAH 7EAK SOLUTIONS AND DEVELOPMENT OF SINGULARITIES IN THE 35
z MODEL #OMM 0URE !PPL -ATH [
;3H:= * 3HATAH !34 :ADEH /N THE #AUCHY PROBLEM FOR EQUIVARIANT WAVE
MAPS #0!- [
;3HN= ! 3HNIRELMAN /N THE NONUNIQUENESS OF WEAK SOLUTIONS FOR %ULER EQUA
TIONS #0!- [
;3H3= * 3HATAH - 3TRUWE 7ELL POSEDNESS IN ENERGY SPACE FOR SEMILINEAR
WAVE EQUATIONS WITH CRITICAL GROWTH )NT -ATH 2ES .OT [

;3I= 4 3IDERIS &ORMATION OF SINGULARITIES IN THREE DIMENSIONAL 6OMPRESSIBLE
mUIDS #OMM -ATH 0HYS [
;3M= ( 3MITH ! PARAMETRIX CONSTRUCTION FOR WAVE EQUATIONS WITH # COEF
lCIENTS !NNALES DE ,)NSTITUT &OURIER [
;3M3= ( 3MITH # 3OGGE /N 3TRICHARTZ AND EIGENFUNCTION ESTIMATES FOR LOW
REGULARITY METRICS PREPRINT
;3M3= ( 3MITH # 3OGGE .ULL FORM ESTIMATES FOR SYMBOLS AND
LOCAL EXISTENCE FOR A QUASILINEAR $IRICHLET WAVE EQUATION !NN 3CI %.3
TO APPEAR
;3O= # 3OGGE 0ROPAGATION OF SINGULARITIES AND MAXIMAL FUNCTIONS IN THE
PLANE )NV -AT [
6ISIONS IN -ATH 0$% !3 ! 5.)&)%$ 35"*%#4
;3T= ** 3TOKES /N A DInCULTY IN THE THEORY OF SOUND 0HILOSOPHICAL -AGA

ZINE [
;3TR= 7 3TRAUSS 7EAK SOLUTIONS FOR NONLINEAR WAVE EQUATIONS !NNAIS !CAD
"RAZIL #IENCIAS [
;4= $ 4ATARU ,OCAL AND GLOBAL RESULTS FOR WAVE MAPS ) #0$%
[ PART )) TO APPEAR IN !*-
;4= $ 4ATARU 3TRICHARTZ ESTIMATES FOR OPERATORS WITH NON SMOOTH COEn
CIENTS AND THE NONLINEAR WAVE EQUATION !*- TO APPEAR PART )) AND )))
PREPRINTS
;7= 2 7ALD 'RAVITATIONAL #OLLAPSE AND #OSMIC #ESORSHIP GR
QC
;7I= % 7IGNER 4HE UNREASONABLE EkECTIVENESS OF MATHEMATICS IN THE NATURAL
SCIENCES #0!- [
;7O= 4 7OLFF 2ECENT 7ORK #ONNECTED TO THE +AKEYA 0ROBLEM 0ROSPECTS IN
-ATH !-3 0RINCETON
;7O= 4 7OLFF ! SHARP BILINEAR RESTRICTION ESTIMATE )-2- TO APPEAR
;9= 34 9AU /PEN PROBLEMS IN GEOMETRY PREPRINT
;:= !34 :ADEH 2ELATIVISTIC AND NONRELATIVISTIC ELASTODYNAMICS WITH SMALL
SHEAR STRAIN !NN )NST ( 0OINCARsE 0HYSIQUE 4HEORIQUE [

3ERGIU +LAINERMAN $EPARTMENT OF -ATHEMATICS 0RINCETON 5NIVERSITY 0RINCE

TON .* 53!
,%33/.3 &/2 452"5,%.#%
!NTTI +UPIAINEN
.AVIER 3TOKES %QUATIONS AND 4URBULENCE

4URBULENCE IS ONE OF THE MAJOR UNSOLVED PROBLEMS OF CLASSICAL PHYSICS 4HE
REASON FOR ITS CONTINUED INTEREST AMONG PHYSICISTS AND MATHEMATICIANS NOT
TO MENTION ENGINEERS IS EASY TO UNDERSTAND /N THE ONE HAND IT HAS
ENORMOUS PRACTICAL CONSEQUENCES RANGING FROM AIRPLANE AND SHIP DESIGN
TO WEATHER FORECASTS /N THE OTHER HAND IT IS BELIEVED TO BE A PROPERTY
OF SOLUTIONS OF A RELATIVELY SIMPLE PARTIAL DIkERENTIAL EQUATION THE .AVIER
3TOKES .3 EQUATION THE SIMPLICITY OF THE EQUATION STANDS IN CONTRAST WITH
THE COMPLEXITY OF THE PHENOMENA
,ET US CONSIDER A mUID IE A GAS OR A LIQUID CONlNED IN A REGION j y 2D
WHERE THE SPATIAL DIMENSION D IS IN APPLICATIONS USUALLY OR $ENOTING THE
VELOCITY OF THE mUID AS OBSERVED AT A POINT X AND AT TIME T BY UT X 2D
THE .3 EQUATION IS

TU U ` R U _ vR U y F _ RP
WHERE v IS THE VISCOSITY OF THE mUID y ITS DENSITY F IS THE EXTERNAL FORCE
ACTING ON THE mUID AND P IS THE PRESSURE )N MOST APPLICATIONS ONE MAY
ASSUME THAT THE mUID IS INCOMPRESSIBLE IE THAT y IS CONSTANT AND THAT U
IS DIVERGENCE FREE R ` U )T FOLLOWS THEN BY TAKING THE DIVERGENCE OF
BOTH SIDES OF %Q THAT R P _R ` U ` R U SO THE PRESSURE IS ACTUALLY
A FUNCTION OF VELOCITY lELD )TS EkECT IS TO PROJECT THE NONLINEAR TERM INTO
A SPACE OF DIVERGENCELESS VECTOR lELDS ON j
%Q IS A NONLINEAR PARABOLIC 0$% ACTUALLY AN INTEGRODIkERENTIAL
EQUATION DUE TO THE PRESSURE TERM 4HE vR U TERM IN THE .3 EQUATION
REPRESENTS THE EkECT OF FRICTION AND TENDS TO SMOOTHEN AND DRIVE DOWN THE
MOTION OF THE mUID 4HE MOTION IS MAINTAINED BY THE EXTERNAL FORCE F
WHEREAS THE FUNCTION OF THE NONLINEAR TERM IS MORE SUBTLE AND WILL BE DIS
CUSSED BELOW
4O HAVE A CLEAN FORMULATION OF THE PROBLEM OF TURBULENCE IT IS IMPORTANT
TO UNDERSTAND EXACTLY WHAT SORT OF FORCES ONE SHOULD CONSIDER IN )N EX
PERIMENTAL SITUATIONS THE FORCING MAY BE PRODUCED BY BOUNDARY CONDITIONS
6ISIONS IN -ATH ,%33/.3 &/2 452"5,%.#%
LIKE IN THE CASE OF mOW IN A PIPE OR AROUND AN OBSTACLE EG AIRPLANE WING
WHERE THE VELOCITY lELD VANISHES AT THE BOUNDARY AND TAKES A CHARACTERISTIC
VALUE FAR ENOUGH AWAY FROM IT THE RELATIVE VELOCITY OF THE AIR AND THE PLANE
OR THE VELOCITY WITH WHICH THE mUID ENTERS THE PIPE !NOTHER POSSIBILITY
IS TO USE AN EXPLICIT STIRRING MECHANISM LIKE IN RECENT EXPERIMENTS ;0*4=
WHERE ELECTRIC CURRENT IS LED INTO A SOLUTION OF WATER AND SALT AND STIRRING
IS ACHIEVED BY APPLYING A TIME DEPENDENT MAGNETIC lELD )N THESE SITUA
TIONS THE FORCE IS APPLIED ON A CHARACTERISTIC LENGTHSCALE THE DIAMETER OF
THE PIPE OR THE WING THE SIZES OF THE MAGNETS ,ET US CALL THIS SCALE , )N
EQUATION WE COULD EG CONSIDER F WHOSE &OURIER TRANSFORM IN X B F T w
HAS COMPACT SUPPORT NEAR JwJ ,_ BELOW WE WILL DISCUSS A MORE EXPLICIT
EXAMPLE
)T IS INTUITIVELY CLEAR THAT THE RELATIVE SIZES OF v AND F CONTRIBUTE TO THE
STRENGTH OF THE NONLINEARITY F TENDS TO INCREASE U AND v DRIVE IT DOWN !
MORE PRECISE FORMULATION OF THIS INTUITION INVOLVES THE 2EYNOLDS NUMBER A
MEASURE OF THE STRENGTH OF THE NONLINEAR EkECTS THAT TAKES INTO ACCOUNT TWO
SYMMETRIES OF SCALE AND 'ALILEAN INVARIANCE 4HE SCALING CORRESPONDS
TO SIMPLE CHANGES IN SPATIAL AND TEMPORAL SCALES SETTING
U T X { S_ U{ T SX
THEN U SOLVES THE .3 EQUATION WITH VISCOSITY v
{ S_ v
AND FORCE F

{ S_ F { T SX 4HE NEW FORCE IS ON THE SCALE ,S 4HUS A SCALE INVARIANT
MEASURE OF THE RELATIVE STRENGTH OF THE FORCING VS VISCOSITY IS GIVEN BY THE
2EYNOLDS NUMBER
,mU
2E
v
WHERE mU IS A CHARACTERISTIC SIZE OF VELOCITY DIkERENCES PRODUCED BY THE
FORCE 4HE REASON WHY VELOCITY DIkERENCES AND NOT ABSOLUTE VELOCITIES MAT
TER IS THAT THE .3 EQUATION IS INVARIANT MODULO BOUNDARY CONDITIONS UNDER
'ALILEAN TRANSFORMATION UT X UT X VT _ V WHERE V IS A CONSTANT
VELOCITY &OR THE mOW IN THE PIPE OR AROUND THE WING WE MAY TAKE mU AS THE
DIkERENCE OF THE VELOCITY AT THE BOUNDARY ZERO ACTUALLY AND FAR AWAY FROM
IT )N PRACTICAL TERMS THE SCALE INVARIANCE IS IMPORTANT ONE CAN SUBSTITUTE
SAY WATER FOR AIR BY CHANGING THE SIZE OF THE SYSTEM ANDOR STRENGTH OF THE
FORCE
4HE BASIC PHENOMENOLOGICAL FACTS ABOUT HYDRODYNAMIC mOWS ARE AS FOL
LOWS )F 2E { ONE ENCOUNTERS REGULAR <LAMINAR mOWS &OR 2E BETWEEN
w AND w A SERIES OF BIFURCATIONS OCCUR LEADING TO EVER MORE COM
PLICATED mOWS <TRANSITION TO CHAOS &INALLY FOR 2E | A PECULIAR
!.44) +50)!).%. '!&!
CHAOTIC STATE TURBULENCE IS REACHED THAT SEEMS TO EXHIBIT FOR VERY LARGE 2E
<DEVELOPED TURBULENCE UNIVERSAL FEATURES IE FEATURES THAT ARE INDEPEN
DENT ON THE DETAILED NATURE OF THE FORCING 7E WILL DISCUSS THESE FEATURES
BELOW BUT LET US lRST MENTION SOME MATHEMATICAL DEVELOPMENTS
4HE )NITIAL 6ALUE 0ROBLEM

-UCH OF THE MATHEMATICAL WORK ON TURBULENCE HAS CONCENTRATED ON THE
STUDY OF THE INITIAL VALUE PROBLEM FOR IE EXISTENCE AND UNIQUENESS
OF SOLUTIONS FOR ALL TIMES GIVEN U X )MPORTANT ROLE HERE IS PLAYED BY
CONSERVATION LAWS 4HE KINETIC ENERGY OF THE mUID IS GIVEN BY THE , NORM
OF U WE SET FOR SIMPLICITY THE DENSITY TO AND IT IS FORMALLY CONSERVED BY
THE v mOW WITHOUT FORCING #ORRESPONDINGLY SUPPOSING U IS A SMOOTH
SOLUTION TO VANISHING AT THE BOUNDARY OF j ONE GETS BY INTEGRATING THE
SCALAR PRODUCT OF WITH U OVER j THE ENERGY CONSERVATION EQUATION
: : :
D
DT
U _v RU F ` U
j j j
0HYSICALLY THE EQUATION STATES THAT 2 THE RATE OF CHANGE OF mUID ENERGY IS
EQUAL TO THE ENERGY INJECTION RATE F ` U WORK OF THE EXTERNAL
2 FORCES PER
UNIT TIME MINUS THE ENERGY DISSIPATION PER UNIT TIME v RU DUE TO THE
VISCOUS FRICTION .OTE THAT THE NONLINEAR TERM DOES NOT CONTRIBUTE HERE
!N IMMEDIATE CONSEQUENCE OF IS AN A PRIORI BOUND FOR SMOOTH SOLU
TIONS THE , j NORM OF U STAYS BOUNDED FOR ALL TIMES SAY FOR A BOUNDED
FORCE IN A COMPACT j AS DOES THE , ; T= a j NORM OF RU 4HIS A PRIORI
BOUND IS ENOUGH TO PROVE THE EXISTENCE OF WEAK IE DISTRIBUTIONAL SOLUTIONS
TO THE .3 EQUATION AS WAS REALIZED BY ,ERAY IN ;,ER= (E APPROXI
MATED THE EQUATION BY A lNITE DIMENSIONAL PROBLEM FOR WHICH THE APRIORI
BOUND GUARANTEES A UNIQUE SOLUTION "Y COMPACTNESS SUCH APPROXIMATIONS
YIELD A WEAK SOLUTION TO .3
4HE UNIQUENESS AND SMOOTHNESS OF THESE WEAK SOLUTIONS HAS EVER SINCE
BEEN AN OPEN PROBLEM IN THE CASE OF SPATIAL DIMENSION 3UPPOSE FOR
SIMPLICITY j IS COMPACT AND THE FORCE IS ZERO SO CALLED DECAYING TURBULENCE
AND WE START WITH SMOOTH INITIAL DATA &OR 2 THE 2EYNOLDS NUMBER IN THIS
CASE WE MAY TAKE WITH mU ,JjJ_ j JRUJ WHERE , IS SAY THE
DIAMETER OF j 4HEN FOR SMALL ENOUGH 2E THE VISCOUS TERM DOMINATES IN THE
.3 EQUATION AND ONE MAY PROVE UNIQUENESS AND SMOOTHNESS OF SOLUTIONS
FOR ALL TIMES !S 2E OF THE INITIAL DATA INCREASES ONE LOSES CONTROL OF THE (
NORM OF THE SOLUTION IE THE TIME T 2EYNOLDS NUMBER AND SMOOTHNESS CAN
6ISIONS IN -ATH ,%33/.3 &/2 452"5,%.#%
BE PROVED ONLY FOR A TIME THAT TENDS TO ZERO AS 2E TENDS TO INlNITY 4HE
A PRIORI ENERGY CONSERVATION IMPLIES ONLY THAT RU IS SQUARE INTEGRABLE IN
SPACE AND TIME AND THUS DOESNT RULE OUT SINGULARITIES /NLY UPPER BOUNDS
FOR THE (AUSDORk DIMENSION OF THE SINGULARITY SET CAN BE DEDUCED ;3=
!S WE WILL SEE BELOW PHYSICALLY ONE EXPECTS THE TYPICAL IN THE SENSE OF
ERGODIC THEORY SOLUTIONS TO BE SMOOTH AND TO LOSE THEIR SMOOTHNESS ONLY
IN THE LIMIT OF INlNITE 2E
)N THE CASE OF DIMENSION TWO THE %ULER EQUATION HAS A SECOND CONSERVED
QUANTITY THE ENSTROPHY WHICH IS THE , NORM OF THE VORTICITY XV _ XV
4HE .3 EQUATION IMPLIES THE FOLLOWING EQUATION FOR

T U ` R _ vR F _ F
WHICH IN TURN IMPLIES A RELATION ANALOGOUS TO FOR ENSTROPHY G XF _ XF
: : :
D
DT
_v R G
j j j
YIELDS AN A PRIORI BOUND FOR THE , ; T= ( j AND , ; T= ( j
NORMS OF U 4HESE SUnCE TO YIELD UNIQUENESS AND SMOOTHNESS OF SOLUTIONS
FOR ALL TIMES NOW WE HAVE AN APRIORI BOUND FOR THE TIME T 2EYNOLDS NUMBER
3TATISTICAL (YDRODYNAMICS
)T IS FAIR TO SAY THAT THE STUDY OF THE GLOBAL REGULARITY OF THE SOLUTIONS OF THE
.3 EQUATION HAS SHED RELATIVELY LITTLE LIGHT ON THE PROBLEM OF TURBULENCE
4HERE HAS BEEN HOWEVER ANOTHER LINE OF APPROACH TO TURBULENCE PIONEERED
BY !. +OLMOGOROV ;+= BASED ON IDEAS FROM ERGODIC THEORY THAT HAS BEEN
MORE SUCCESSFUL
)N THIS APPROACH ONE VIEWS AS A DYNAMICAL SYSTEM IN SOME SPACE OF
VECTOR lELDS ON j AND ASKS QUESTIONS ON THE BEHAVIOUR OF TYPICAL TRAJECTORIES
WITH RESPECT TO AN INVARIANT MEASURE /F COURSE TO REALLY CARRY OUT THIS
STRATEGY A NECESSARY PREREQUISITE WOULD BE AN EXISTENCE AND UNIQUENESS
THEOREM FOR SOLUTIONS BUT ONE MAY ALSO ADOPT A PRAGMATIC ATTITUDE AND
INQUIRE WHAT CAN BE SAID ABOUT THIS MEASURE IF IT EXISTS
!N ERGODIC INVARIANT MEASURE u WOULD SATISFY
: 4 :
LIM 4 _ & UT DT & U uDU
4
IE THE TIME AVERAGE OF SUITABLE FUNCTIONS OF THE VELOCITY lELD ARE GIVEN
BY THE ENSEMBLE AVERAGES 4HUS THE INVARIANT MEASURE CAN BE ACCESSED BY
TAKING THE AVERAGE OF A TIME SERIES FROM AN EXPERIMENT OR AN AVERAGE OVER
<SNAPSHOTS OF SEVERAL EXPERIMENTS
!.44) +50)!).%. '!&!
)N THE CONTEXT OF SMOOTH HYPERBOLIC DYNAMICAL SYSTEMS SUCH A <PHYSI

CAL MEASURE EXISTS )T IS THE 3INAI 2UELLE "OWEN MEASURE AND IT IS SELECTED
AS THE ONE THAT IS STABLE UNDER SMALL NOISE 7ITH THIS IN MIND IT IS REA
SONABLE TO CONSIDER THE .3 EQUATION WITH A RANDOM FORCE -ORE PRECISELY
CONSIDER A STOCHASTIC 0$%
_ `
DU U ` R U _ vR U RP DT D&
WHERE &T X IS A SUITABLE 7IENER PROCESS 4O AVOID WORRYING ABOUT THE
BOUNDARIES LET US TAKE THE SPATIAL REGION TO BE THE WHOLE 2D ALTHOUGH
IN TWO DIMENSIONS AS WE WILL SEE THIS IS NOT EXPECTED TO BE AN INNOCENT
ASSUMPTION 7E GET A MODEL OF HOMOGENEOUS AND ISOTROPIC TURBULENCE IE
INVARIANT UNDER SPATIAL TRANSLATIONS AND ROTATIONS WITH FORCING AT SCALE ,
BY TAKING & WITH COVARIANCE
%&j T X &k T Y T#jk X _ Y ,
WITH # HAVING A &OURIER TRANSFORM #w B CJwJ AND CK SMOOTH WITH
_
COMPACT SUPPORT NEAR K , 7ITH GOOD ENOUGH CONTROL OF THE SOLUTIONS
OF ONE COULD THEN DElNE A -ARKOV CHAIN IN A SUITABLE SPACE 6 OF VECTOR
lELDS ON 2D BY THE TRANSITION PROBABILITIES PU 5 0ROBFU 5 J
U UG 4HE STATIONARY MEASURES u OF THIS -ARKOV CHAIN WOULD THEN
GIVE THE TIME AVERAGES IN
4HE MATHEMATICAL STATUS OF IS SIMILAR TO THE NONRANDOM CASE 7EAK
STATISTICAL SOLUTIONS EXIST ;6&= IF WE REPLACE 2D BY A COMPACT REGION AND IN
TWO DIMENSIONS THEY ARE UNIQUE 5NIQUENESS OF INVARIANT MEASURE IS OPEN IN
TWO DIMENSIONS EXISTENCE IS EASY FOR THE KIND OF FORCING CONSIDERED ABOVE
&OR THE NONRANDOM CASE EXISTENCE OF A lNITE DIMENSIONAL ATTRACTING SET TWO
DIMENSIONS COMPACT REGION IS KNOWN ;,= ;#O&4= AND SIMILAR RESULTS HOLD
FOR THE -ARKOV CHAIN DElNED ABOVE ;"R+,=
5NIVERSALITY
4HE CONTRAST BETWEEN THEORETICAL UNDERSTANDING OF TURBULENCE AND EXPER
IMENTAL DATA ON IT IS STRIKING 2ATHER ACCURATE EXPERIMENTAL INFORMATION
ON THE NATURE OF THE HIGH 2EYNOLDS NUMBER INVARIANT MEASURE IS AVAILABLE
)N PARTICULAR FOR LARGE 2E ANOTHER LENGTH SCALE p SEEMS TO EMERGE IN THE
PROBLEM IN ADDITION TO THE FORCING SCALE , AND THE INVARIANT MEASURE SEEMS
TO POSSESS UNIVERSAL FEATURES FOR SCALES BETWEEN p AND , !S 2E THE
SCALE p AND IN THIS LIMIT THE SO CALLED INERTIAL RANGE ;p ,= EXTENDS ALL
THE WAY TO ZERO 7E SHALL NOW DISCUSS MORE EXPLICITLY SOME EXAMPLES OF
CORRELATIONS AND THE NATURE OF THE CONJECTURED UNIVERSALITY
6ISIONS IN -ATH ,%33/.3 &/2 452"5,%.#%
%NERGY SPECTRUM 4HE ENERGY CONSERVATION EQUATION BECOMES AN

ENERGY BALANCE RELATION FOR THE INVARIANT MEASURE
:

v RUX uDU TR# p n
4O DERIVE NOTE THAT BY AN )TO FORMULA IMPLIES DU U ` vR U _

U ` R U _ RP DT U ` D& TR# 4AKING AVERAGES OVER THE INVARIANT
MEASURE AND NOTING THAT BY SPATIAL TRANSLATION INVARIANCE THE NONLINEAR
TERM WILL NOT CONTRIBUTE THE CLAIM FOLLOWS 4HE PHYSICAL INTERPRETATION OF
IS THAT AT THE STATIONARY STATE THE ENERGY INJECTION RATE TR# IS BAL
ANCED BY THE ENERGY DISSIPATION RATE WHOSE COMMON VALUE WE DENOTE BY n
4HE FORMER IS A PROPERTY OF THE FORCING AND EXPERIMENTALLY CONTROLLABLE THE
LATTER CAN BE DEDUCED FROM MEASUREMENTS OF VELOCITY CORRELATION FUNCTION
,ET : :
UX ` UY uDU 'X _ Y B
EIwX_Y 'w Dw
2D
4HEN THE AVERAGE ENERGY DENSITY OF THE mUID IS GIVEN AS
: : :
B

UX uDU 'w Dw EK DK
2D
B
WHERE WE DElNED THE ENERGY SPECTRUM EJwJ 'w JwJ D_ )N THREE DIMEN
SIONS EK SEEMS TO HAVE A POWER LAW BEHAVIOUR FOR K IN THE INERTIAL RANGE
;,_ p _ = CONSISTENT WITH THE CONJECTURE
_ `
EK #K _l OpK ,K _
4HE EXPONENT l IS CLOSE TO

lw
&OR K p _EK DECREASES FAST PRESUMABLY EXPONENTIALLY &ROM THIS
EXPERIMENTAL INFORMATION WE MAY DRAW INTERESTING CONCLUSIONS ON THE STA
TIONARY STATE
&IRST WRITING THE ENERGY BALANCE RELATION IN TERMS OF E WE HAVE WE
B
SUPPOSE THE FORCING IS ISOTROPIC IE #w ONLY DEPENDS ON JwJ
: :
B
v K EK DK TR# TR#K K DK

4HE ,(3 GETS MOST OF ITS CONTRIBUTION NEAR K p _ SINCE K EK w K
WHEREAS THE INTEGRAND ON THE 2(3 IS SUPPORTED IN A NEIGHBOURHOOD OF ,_
4HIS IS AN INDICATION OF THE SO CALLED CASCADE PICTURE ACCORDING TO WHICH
THE STATIONARY STATE IS AN ENERGY mUX STATE WHERE ENERGY IS INJECTED TO THE
SYSTEM AT LARGE SCALES AND TRANSPORTED BY THE NONLINEARITY TO SMALL SCALES
!.44) +50)!).%. '!&!
WHERE IT IS DISSIPATED BY THE VISCOUS FORCES .OTE ALSO THAT THE LEFT HAND
SIDE BEHAVES AS vp _ AND THE RIGHT HAND SIDE IS v INDEPENDENT 4HUS AS
THE VISCOSITY IS TAKEN TO ZERO THE VISCOUS SCALE p TENDS TO ZERO AS v )T
SHOULD BE STRESSED HOWEVER THAT THERE IS NO REAL DERIVATION OF THE CASCADE
PICTURE FROM THE .3 EQUATION
4HE EXPONENTIAL DECREASE OF EK AS K WOULD GUARANTEE THAT FOR
v THE INVARIANT MEASURE SITS ON SMOOTH VELOCITIES (ENCE IF THERE WAS
lNITE TIME BLOWUP IN .3 IT WOULD HAVE ZERO PROBABILITY IN THE STATIONARY
STATE AND THUS IT WOULD BE HARD TO SEE NUMERICALLY "UT OF COURSE ONE
CANNOT DEDUCE FROM EXPERIMENTS FOR SURE THE LARGE K ASYMPTOTICS OF EK
(OWEVER AS v THE EXPECTATION OF RUX DIVERGES (ENCE IN
THE INVISCID LIMIT2 THE TYPICAL SOLUTIONS ARE NOT DIkERENTIABLE IN FACT SINCE
%JRJj UX Kj EK DK WE EXPECT ONLY (OLDER CONTINUITY WITH EXPO
NENT APPROXIMATELY IN THAT LIMIT 4HE INVISCID LIMIT OF .3 IS THE %ULER
EQUATION 4HUS THE TYPICAL SOLUTIONS OF THE .3 EQUATION BECOME AT MOST
WEAK SOLUTIONS OF THE %ULER EQUATION SOLUTIONS THAT DISSIPATE ENERGY 4HE
OBSERVATION THAT SOLUTIONS OF THE %ULER EQUATION HAVE TO HAVE LOW REGULARITY
IN ORDER TO DISSIPATE ENERGY GOES BACK TO /NSAGER IN THE LATE FORTIES ;/=
)N TWO DIMENSIONS THE PREVIOUS DISCUSSION NEEDS INTERESTING MODIlCA
TIONS DUE TO THE ADDITIONAL CONSERVED QUANTITY ENSTROPHY 4HE CONSERVATION
LAW IMPLIES FOR THE STATIONARY STATE
:

v RX uDU TRa#
WITH THE INTERPRETATION AS EQUALITY OF ENSTROPHY DISSIPATION RATE AND EN

STROPHY INJECTION RATE 4HUS IN ADDITION TO E HAS TO SATISFY
: :
v
K EK DK B
TR#K K
DK

.OW WE SEEM TO HAVE A DILEMMA IF AS BEFORE THE MAIN CONTRIBUTION TO
AND COMES FROM K NEAR THE VISCOUS SCALE ONLY THE ENSTROPHY DISSIPATION
RATE CAN STAY NONZERO AS v )NDEED EXPERIMENTS ;0*4= INDICATE THAT
NOW EK HAS THE SCALING BEHAVIOUR WITH THE EXPONENT GAMMA
l
(ENCE ENSTROPHY IS DISSIPATED NEAR THE VISCOUS SCALE THAT NOW IS PROPOR
TIONAL TO p w v 4HUS THE DISSIPATION RATE OF ENERGY ON THE INTERVAL
K ;,_ p _ = TENDS TO ZERO AS v AND CANNOT HOLD 4O UN
DERSTAND WHAT ACTUALLY HAPPENS CONSIDER lRST OUR INlNITE VOLUME MODEL
SYSTEM 3TARTING WITH SAY INITIAL CONDITION U ENSTROPHY mOWS TOWARDS
6ISIONS IN -ATH ,%33/.3 &/2 452"5,%.#%
LARGE K IE THE K _ SPECTRUM BUILDS UP ON THE INTERVAL ;,_ p _ = 4HIS IS

CALLED THE DIRECT ENSTROPHY CASCADE /N THE OTHER HAND ENERGY STARTS TO
mOW TOWARDS SMALL K ,_ IN WHAT IS CALLED THE INVERSE CASCADE BUILDING
A SECOND POWER LAW REGION WITH THE EXPONENT l .OTE THAT SUCH A
POWER LAW FOR EK ON ; ,_ = WOULD MEAN INlNITE TOTAL ENERGY DENSITY
IE UX WOULD HAVE INlNITE EXPECTATION 7HAT SHOULD ACTUALLY HAPPEN
IS THAT AT TIME T THE INITIAL CONDITION HAS EVOLVED TO A MEASURE WITH TOTAL
ENERGY PROPORTIONAL TO T AND ENERGY SPECTRUM HAVING THE LAW DOWN TO
K w T_ U IS NOT A WELL DElNED RANDOM VARIABLE WITH RESPECT TO THE
INVARIANT MEASURE ONLY THE VORTICITY IS )N lNITE VOLUME THE ENERGY CAN
NOT ESCAPE TO MODES WITH SMALLER AND SMALLER K AND WILL BE DISSIPATED !
STATIONARY STATE THEN EXISTS ALSO FOR U 4HE INVERSE CASCADE WAS POSTULATED
IN THE S BY +RAICHNAN AND "ATCHELOR ;+R= ;"=
-ULTISCALING 4HE DISCUSSION ABOVE INDICATES THAT THE TURBULENT STATE
SEEMS TO HAVE SCALE INVARIANCE FOR LENGTH SCALES MUCH SMALLER THAN THE
FORCING SCALE , %XACT SCALE INVARIANCE WOULD MEAN THAT UX _ UY
t_o UtX _ UtY IN LAW AND THE ENERGY SPECTRUM INDICATES THAT o w
4HINGS ARE MORE INTERESTING HOWEVER /NE CAN GET HOLD OF THE ( OLDER
PROPERTIES OF THE VELOCITY BY STUDYING SO CALLED STRUCTURE FUNCTIONS NAMELY
MOMENTS OF VELOCITY DIkERENCES IN THE INVARIANT MEASURE 4HE LONGITUDINAL
STRUCTURE FUNCTIONS ARE DElNED BY
:
_X `N
3N R JXJ ` UX Y _ UY uDU
WHERE R JXJ AND TRANSLATIONAL AND ROTATIONAL INVARIANCE OF THE INVARIANT
MEASURE WAS ASSUMED
%XPERIMENTALLY THE STRUCTURE FUNCTIONS SEEM TO BEHAVE AS
3N R RoN
FOR R IN THE INERTIAL RANGE ;p ,= ( OLDER CONTINUITY WITH EXPONENT
WOULD INDICATE THAT oN WERE N 4HIS INDEED IS THE PREDICTION MADE BY
+OLMOGOROV IN
4O SEE HOW THIS RESULT COULD ARISE LET US CONSIDER THE MODEL 4AKE
FOR DElNITENESS THE LARGE SCALE FORCING COVARIANCE AS #X , C ,X WITH C
A lXED RAPIDLY DECREASING FUNCTION 3N DEPENDS ON THE PARAMETERS IN THE
EQUATION NAMELY v , C -AKING THE CHANGE OF VARIABLES WE GET
_ `N o
p
3N R v , C {L 3N RL L{ v ,L {L C
3UPPOSE NOW THAT THE LIMIT LIM, LIMv 3N R v , C 3N R C EXISTS

4HEN TAKING L R AND {L n_ RECALL n TRC IS THE ENERGY INJECTION
!.44) +50)!).%. '!&!
RATE WE GET
3N R C !N nR N
WHERE !N 3N R Cn +OLMOGOROV MADE THE FURTHER ASSUMPTION THAT
!N IS INDEPENDENT ON C AND THUS THE STRUCTURE FUNCTIONS ARE UNIVERSAL
DEPENDING ONLY ON R AND ON FORCING ONLY THROUGH THE AMOUNT OF ENERGY
PUMPED IN PER UNIT TIME AND VOLUME
)N REALITY DOESNT QUITE lT WITH EXPERIMENTS /NLY o AGREES WELL
WITH THE VALUE o 4HE EXPERIMENTAL VALUES FOR THE lRST FEW EXPONENTS
ARE ;"EN#"2=
o o o
o o
WITH THE +OLMOGOROV VALUES IN THE PARENTHESIS FOR COMPARISON 4HESE VALUES
INDICATE THAT THE LIMIT ARGUMENT LEADING TO IS mAWED )NDEED IT IS
BELIEVED THAT THE THE v LIMIT SHOULD EXIST BUT NOT THE , LIMIT
4HEN THE RIGHT ASYMPTOTICS SHOULD LOOK LIKE
N _ _ ``
3N R "N RoN , _oN O pR ,R
WITH "N NONUNIVERSAL CONSTANTS DEPENDING ON THE DETAILS OF FORCING IE
C IN OUR MODEL WHEREAS THE oN ARE BELIEVED TO BE UNIVERSAL "Y ( OLDER
INEQUALITY THE EXPONENTS oN ARE CONCAVE FUNCTIONS OF N FOR EVEN N 4HE
DATA INDICATES THAT THEY ARE STRICTLY CONCAVE NOT LINEAR 4HUS 3N R 3N R
GROWS WITH N MORE AND MORE AS R THIS PHENOMENON IS REFERRED TO
AS INTERMITTENCY 4HE VELOCITY DIkERENCES BEHAVE VERY DIkERENTLY FROM
GAUSSIAN RANDOM VARIABLES AND MORE SO WITH DIMINISHING DISTANCE 4HE
ORIGIN OF THESE EXPONENTS AND DERIVATION OF THEIR VALUES IS A MAJOR THEORETICAL
PROBLEM OF TURBULENCE
)N TWO DIMENSIONS THINGS SEEM TO BE SIMPLER &OR THE DIRECT ENSTRO
PHY CASCADE IE DISTANCES LESS THAN THE INJECTION SCALE THE +RAICHNANS
K_ ENERGY SPECTRUM SEEMS EXACT BY RECENT EXPERIMENTS ;0*4= AND THE
+OLMOGOROVS K_ SPECTRUM CORRESPONDINGLY FOR THE INVERSE ENERGY CAS
CADE IE DISTANCES LARGER THAN , BY EXPERIMENTS ;04= AS WELL AS NUMERICAL
SIMULATIONS ;"O#6= (ENCE TWO DIMENSIONAL TURBULENCE MIGHT BE MORE
ACCESSIBLE TO MATHEMATICAL UNDERSTANDING THAN THE STRONGLY INTERMITTENT
THREE DIMENSIONAL ONE
4O SUMMARIZE THERE IS VERY LITTLE MATHEMATICAL UNDERSTANDING ON THE
BASIC FEATURES OF TURBULENCE THE ENERGYENSTROPHY CASCADE PICTURE OF THE
STATIONARY STATE AND THE PHENOMENON OF INTERMITTENCY 4HERE IS ALMOST
NO THEORETICAL UNDERSTANDING OF THE LATTER AS A PHENOMENON ARISING FROM
6ISIONS IN -ATH ,%33/.3 &/2 452"5,%.#%
.AVIER 3TOKES EQUATIONS IN SPITE OF SEVERAL DECADES OF ATTEMPTS %VEN WORSE

THERE DOESNT SEEM TO BE ANY REASONABLE APPROXIMATION TO .3 EQUATIONS
WHERE INTERMITTENCY CAN BE STUDIED 5NTIL VERY RECENTLY THERE WAS NO OTHER
MATHEMATICAL MODEL WHERE INTERMITTENCY WOULD ALLOW A MATHEMATICAL ANAL
YSIS -UCH OF THE RECENT PROGRESS IN THE lELD INVOLVES A MODEL OF ADVECTION
ORIGINALLY INTRODUCED BY +RAICHNAN ;+R= IN THE SIXTIES )T TURNS OUT THIS
PROBLEM HAS VERY SIMILAR PHENOMENOLOGY AS THE .3 EQUATION
[ #ASCADE PICTURE OF THE STATIONARY STATE THAT IS SUPPORTED ON WEAK EN
ERGY DISSIPATING SOLUTIONS OF THE ZERO DIkUSIVITY LIMIT OF THE EQUATION
[ ! RANGE OF SCALES WHERE SCALING IS PRESENT AND A +OLMOGOROV LIKE
SCALING THEORY THAT IS VIOLATED BY THE PHENOMENON OF INTERMITTENCY
4HE DIkERENCE WITH .3 IS THAT IN THE +RAICHNAN PROBLEM THESE FEATURES
CAN BE THEORETICALLY UNDERSTOOD
2ANDOM !DVECTION
0ASSIVE SCALAR 4HE VELOCITY lELD UT X GIVES RISE TO A mOW NAMELY DIF
FEOMORPHISMS }TS ON j BY SOLVING THE /$% X{ ? U{ X{ OF SO CALLED
,AGRANGIAN TRAJECTORIES FROM TIME S TO TIME T 0HYSICALLY X{ DESCRIBES
MOTION OF <mUID PARTICLES 4HE mOW ALLOWS US TO TRANSPORT GEOMETRICAL OB
JECTS SCALARS DENSITIES VECTORS &OR INSTANCE EQUATION FOR THE VORTICITY
OF THE TWO DIMENSIONAL mUID TELLS US THAT FOR ZERO VISCOSITY AND FORCE IS
TRANSPORTED BY U AS A SCALAR
_ `
T X }_
T X
3IMILARILY THE VORTICITY OF A THREE DIMENSIONAL mUID IS TRANSPORTED AS A
VECTOR &OR NONZERO VISCOSITY IS BOTH TRANSPORTED AND DIkUSED -ANY
PHYSICAL QUANTITIES OF INTEREST ARE TRANSPORTED AND DIkUSED BY THE mUID
MOTION %G THE TEMPERATURE 4 T X OF THE mUID AT X AT TIME T DEPENDS
ON THE MOLECULAR MOTION AND IN MANY SITUATIONS CAN BE ASSUMED NOT TO
AkECT THE mUID MOTION TAKING PLACE IN MUCH LARGER SCALES THE TEMPERATURE
DEPENDENCE OF VISCOSITY CAN OFTEN BE NEGLECTED 4HEN 4 IS TRANSPORTED BY
THE mUID VELOCITY AND IN ADDITION IT DIkUSES VIA MOLECULAR MOTION )T IS AN
EXAMPLE OF A PASSIVE SCALAR SATISFYING THE EQUATION
T4 U ` R 4 _ sR 4 F
WHERE F T X IS A SOURCE FOR 4
,ET US RECALL BRIEmY THE SOLUTION OF IN TERMS OF PARTICLE MOTION 7E
SUPPOSE lRST THAT U IS SMOOTH AND PROCEED IN STEPS
!.44) +50)!).%. '!&!
s F 4HIS IS JUST TRANSPORT AND SOLVED BY OR IN OTHER WORDS

_ `
4 T X 4 S XTX S
WHERE XTX S IS THE ,AGRANGIAN TRAJECTORY WITH XTX T X
s F 4HE DIkUSION TERM IN ARISES FROM ADDING A "ROWNIAN
MOTION TO THE PARTICLE MOTION ,ET k BE THE STANDARD "ROWNIAN MOTION
#ONSIDER THE STOCHASTIC /$%
_ `
DXS U S XS DS s DkS
WITH XT X NOTE THAT THIS EQUATION IS SOLVED BACKWARD IN TIME 4HEN
4 IS GIVEN AS AN AVERAGE OVER THE TRAJECTORIES
_ `
4 T X %k 4 X
)N OTHER WORDS GIVES RISE TO A -ARKOV PROCESS WITH STATE SPACE 2D AND
TRANSITION PROBABILITY KERNEL 0TS X Y J U THAT IS THE FUNDAMENTAL SOLUTION
TO THE EQUATION THE DERIVATIVES ACT ON THE VARIABLES X AND T AND SO
READS
:
4 T X 0T X Y J U 4 Y DY
F 4HE GENERAL CASE IS THEN GIVEN BY

: : T:
4 T X 0T X Y J U 4 Y DY 0TS X Y J U F S Y DY DS

4HE +RAICHNAN MODEL )N PRACTICAL APPLICATIONS THE VELOCITY lELD U
IN IS A POSSIBLY TURBULENT VELOCITY lELD SATISFYING THE .3 EQUATIONS
)N THAT CASE IDEALLY ONE WOULD LIKE TO STUDY THE STATISTICS OF 4 ARISING
FROM THE STATISTICS OF U )N THE ABSENCE OF THE KNOWLEDGE OF THE LATTER
IT IS STILL OF INTEREST TO CONSIDER SOME EXPLICIT ENSEMBLES OF U AND TRY TO
DEDUCE STATISTICAL PROPERTIES FOR 4 )N THE +RAICHNAN MODEL UT X ARE TAKEN
'AUSSIAN RANDOM VARIABLES THAT ARE ( OLDER CONTINUOUS IN X 4HE FORMER
ASSUMPTION IS FAR FROM TRUE IN THE CASE OF REALISTIC TURBULENT VELOCITIES
BUT AS WE HAVE SEEN THE LATTER MIMICS THE EXPECTED PROPERTIES OF INlNITE
2EYNOLDS NUMBER mOWS
-ORE EXPLICITLY CONSIDER THE 7IENER PROCESS 5T X WITH COVARIANCE
%5j T X 5k T Y T$jk X _ Y
WHERE $ IS (OLDER CONTINUOUS !N EXAMPLE IS THE &OURIER TRANSFORM OF
o p
$B jk K mjk _ Kj Kk ~,JKJ JKJ_D_w
JKJ
6ISIONS IN -ATH ,%33/.3 &/2 452"5,%.#%
WHERE THE lRST FACTOR ASSURES INCOMPRESSIBILITY OF THE VELOCITY AND ~ IS

SMOOTH CUTOk FUNCTION THAT VANISHES IN A NEIGHBOURHOOD OF K 4HE VE
LOCITY U IS THE TIME DERIVATIVE OF 5 IE WE MUST NOW INTERPRET THE lRST TERM
ON THE 2(3 OF ALSO AS A STOCHASTIC DIkERENTIAL UT X DT D5T X )T
IS ALMOST SURELY (OLDER CONTINUOUS IN X WITH ANY EXPONENT LESS THAN w ,
IS THE <INTEGRAL SCALE OF U !S WE WILL SEE BELOW THE PRODUCT D5 ` R4 IN
HAS TO BE INTERPRETED IN THE 3TRATONOWICH SENSE OR EQUIVALENTLY IN )TO
SENSE WITH THE TERM _ $jk j k ADDED
4O COMPLETE THE DElNITION OF THE +RAICHNAN MODEL WE CONSIDER THE
SOURCE F IN ANALOGOUSLY TO WHAT WE DID WITH THE STOCHASTIC .3 TO BE
RANDOM AND ON LARGE SCALE IN X F T X DT D& T X WITH
_ `
%& T X & T Y T# X_Y
,
THE , HERE NEED NOT BE THE SAME AS IN THE VELOCITY COVARIANCE BUT WE
SUPPOSE SO FOR SIMPLICITY OF NOTATION
DETERMINES THE RANDOM VARIABLES 4 T X AS FUNCTIONS OF U AND F
(ENCE WE MAY INQUIRE ABOUT THE SAME QUESTIONS AS IN THE CASE OF .3 NAMELY
THE LONG TIME BEHAVIOUR OF 4 AND THE EXISTENCE OF STATIONARY MEASURE FOR
IT .OTE THAT 4 IS A NONLINEAR FUNCTION OF U IE THE STATISTICS OF 4 MAY BE
NONTRIVIAL EVEN IF U IS GAUSSIAN )NDEED THAT IS WHAT HAPPENS
,ET US lRST SUMMARIZE THE RESULTS ,ET u DENOTE THE INVARIANT MEASURE
OF 4 )T SATISlES THE <ENERGY BALANCE RELATION AS IN .3 WE CALL THE , NORM
OF 4 STILL BY ENERGY
:

s R4 X uD4 #
4HE ENERGY SPECTRUM IS DElNED AS IN .3 AND WE HAVE

_ `
EK #KD_w O@K ,K _
WHERE THE DISSIPATION SCALE L AS s 4HUS THERE IS SCALING ON THE
<INERTIAL RANGE ;@ ,= AND ONE MAY SHOW ;'+= THAT THE INVARIANT MEASURE
EXHIBITS A CASCADE OF <ENERGY
-OREOVER AS s THE INVARIANT MEASURE IS SUPPORTED ON WEAK SOLU
TIONS OF THE s EQUATION AND THESE SOLUTIONS ARE ONLY ( OLDER CONTINUOUS
.EXT CONSIDER THE STRUCTURE FUNCTIONS OF 4
:
_ `N _ `
4 X _ 4 Y uD4 p 3N JX _ YJ
4HE FOLLOWING ASYMPTOTICS HOLDS

_w _ _ ``
3N R !N RoN ,N
_oN
O R@ ,R
!.44) +50)!).%. '!&!
WITH oN STRICTLY CONCAVE IN N 4HUS THE PASSIVE SCALAR EXHIBITS JUST THE SORT
OF MULTISCALING THAT IS EXPECTED IN THE .3 TURBULENCE )T SHOULD BE NOTED
THAT THE <+OLMOGOROV SCALING THEORY HERE ACTUALLY DUE TO #ORSIN WOULD
PREDICT oN N _w

4URBULENCE VS CHAOS )T IS CLEAR FROM THAT IN ORDER TO CONTROL THE
CORRELATION FUNCTIONS OF 4 WE NEED TO UNDERSTAND THE JOINT STATISTICS OF THE
PROBABILITY KERNELS 0 5 &OR INSTANCE TAKING 4 X WE HAVE FOR THE
POINT FUNCTION
: T: o p
_ X _X
%4 T X 4 T X % 0TS X X 0TS X X # , DX DX DS

AND SIMILAR EXPRESSIONS FOR EXPECTATIONS OF HIGHER POINT FUNCTIONS 4HE EX
PECTATION ON THE 2(3 HAS AN INTERESTING INTERPRETATION IN TERMS OF STATISTICS
OF PARTICLE TRAJECTORIES #ONSIDER THE EQUATIONS FOR TWO TRAJECTORIES
P
DXI S D5S XI S sDkI S XI T XI I
WHERE kI ARE INDEPENDENT "ROWNIAN MOTIONS AND D5 IS OUR VELOCITY lELD
UDT 4HEN THE EXPECTATION ON THE 2(3 OF IS THE AVERAGE OVER THE
VELOCITY REALIZATIONS OF THE PROBABILITY DENSITY OF PARTICLES MOVING FROM
XI T TO XI S
7HAT CAN ONE SAY ABOUT THE RANDOM KERNELS 0 5 ,E *AN AND 2AI
MOND HAVE PROVED ;,E2= THAT FOR ALMOST ALL REALIZATIONS OF 5 THESE KERNELS
DElNE A -ARKOV PROCESS EVEN AS s 4HIS RANDOM -ARKOV PROCESS HAS
MARKEDLY DIkERENT BEHAVIOUR DEPENDING ON WHETHER 5 IS ,IPSCHITZ OR ONLY
(OLDER ,ET US SET s "Y AN )TO FORMULA ONE GETS

T %0TS X X $jk j k %0TS X X p - %0TS X X
)N OUR CASE - IS JUST A MULTIPLE OF ,APLACIAN AND THUS INDIVIDUAL TRAJEC
TORIES DIkUSE 4HIS IS NO SURPRISE AS OUR VELOCITY lELD IS DERIVATIVE OF A
"ROWNIAN (OWEVER A SIMILAR APPLICATION OF )TO FORMULA YIELDS
_ ` _ `
% 0TS X X 0TS X X ET_S - X X X X
WITH - THE OPERATOR
- - i i - $jk X _ X Xj Xk
,ET US CONSIDER THE SEPARATION OF TWO PARTICLE TRAJECTORIES 8S
X S _ X S 4HE BEHAVIOUR OF 8S DEPENDS CRUCIALLY ON THE SPATIAL
SMOOTHNESS OF THE VELOCITY )TS TRANSITION PROBABILITY DENSITY AVERAGED OVER
THE VELOCITY IS
%0TS 8 8 ET_S - 8 8
6ISIONS IN -ATH ,%33/.3 &/2 452"5,%.#%
WITH - GIVEN BY RESTRICTING - TO THE SPACE OF FUNCTIONS DEPENDING ON

THE DIkERENCE VARIABLE X _ X
_ `
- $jk _ $jk 8 8j 8k
.OTE THAT THE SYMBOL OF THIS ELLIPTIC OPERATOR IS JUST THE VELOCITY POINT
STRUCTURE FUNCTION )T VANISHES WHEN 8 AS J8Jw
&OR w THE SEMIGROUP BEHAVES VERY MUCH LIKE THE HEAT KERNEL OF

THE ,APLACIAN IT GIVES RISE TO SUPERDIkUSION 8T w T _w )N PARTICULAR
8T SUPER DIkUSES EVEN IF 8 4HUS THE RELATIVE MOTION OF THE
PARTICLES IS NOT DETERMINISTIC FOR A TYPICAL REALIZATION OF THE VELOCITY lELD
)N CONTRAST FOR w t DETERMINISM HOLDS
ET- 8 m8
IE IDENTICAL INITIAL CONDITIONS FOR THE PARTICLES INDUCE IDENTICAL TRAJECTORIES
(OWEVER THERE IS SENSITIVE DEPENDENCE ON INITIAL CONDITIONS 8T GROWS
EXPONENTIALLY WITH T 4HERE IS A SIMILAR PICTURE FOR THE JOINT PROBABILITIES
OF N PARTICLE TRAJECTORIES
iN
%0TS ET_S -N
WHERE -N IS A SINGULAR ELLIPTIC OPERATOR GIVING RISE EITHER TO A DIkUSIVE OR
DETERMINISTIC BEHAVIOUR DEPENDING ON WHETHER w IS LESS OR GREATER OR EQUAL
TO
)T SHOULD BE STRESSED THAT THE DIkUSIVEDETERMINISTIC NATURE OF THE TRA
JECTORIES HOLDS FOR A TYPICAL REALIZATION OF THE VELOCITY lELD AND DEPENDS ON
WHETHER THE lELD IS ,IPSCHITZ OR JUST (OLDER .OTE THAT THIS IS PRECISELY
THE DISTINCTION THAT DETERMINES WHETHER SOLUTIONS OF /$%S ARE UNIQUE
)N OUR CASE A GIVEN TYPICAL (OLDER VELOCITY lELD GIVES RISE NOT ONLY TO
NONUNIQUENESS BUT A WHOLE DIkUSION PROCESS FOR THE PARTICLE SEPARATIONS
4YPICAL HIGH 2EYNOLDS NUMBER TURBULENT VELOCITIES ARE (OLDER FOR SCALES
LARGER THAN THE VISCOUS SCALE SO WE EXPECT THEM TO GIVE RISE TO SUPER
DIkUSIVE PARTICLE MOTION INSTEAD OF CHAOTIC BUT DETERMINISTIC 4HIS INDEED
IS THE CASE AS WAS KNOWN EXPERIMENTALLY ALREADY TO ,& 2ICHARDSON IN
WHO MEASURED THE GROWTH OF THE SEPARATION BETWEEN BALLOON PROBES IN THE
ATMOSPHERE TO BE J8T J w T 4HUS IT APPEARS THAT TURBULENT TRANSPORT
IS VERY DIkERENT FROM CHAOTIC TRANSPORT
,ET US lNALLY BRIEmY DISCUSS THE INTERMITTENCY 4HE STATIONARY STATE
OF THE SCALAR CAN BE EXPRESSED IN TERMS OF THE OPERATORS -N %G FOR THE
TWO POINT FUNCTION WE GET FROM
:
_``
' X Y p 4 X 4 Y uD4 -_ # , X _ Y
!.44) +50)!).%. '!&!
AND SIMILAR EXPRESSIONS CAN BE WRITTEN FOR THE N POINT CASE IN TERMS OF THE
OPERATORS -M M s N ;'+= 4HE VARIOUS CLAIMS ON THE NATURE OF THIS
STATE MENTIONED ABOVE CAN THEN BE EXPLICITLY ANALYZED &OR INSTANCE THE
ASYMPTOTICS OF THE STRUCTURE FUNCTIONS RESULTS AS FOLLOWS 'N ARISES FROM
THE LONG TIME BEHAVIOUR OF THE DIkUSION PROCESS WITH THE GENERATOR -N
4HIS IN TURN IS DOMINATED BY THE ZERO MODES OF -N 3INCE THE OPERATOR IS
HOMOGENEOUS UNDER SCALING OF X THE ZERO MODES ARE ALSO /NE OBTAINS AN
ASYMPTOTIC EXPANSION
: 9 .
.
4 XN uD4 ". , _w _o. . X
N
WHERE . X ARE HOMOGENEOUS ZERO MODES OF THE OPERATORS -. OF DEGREE
o. THAT CAN BE EXPLICITLY ANALYZED FOR w SMALL ;'+= WITH THE RESULT
. . _
o. .
_ w _ D w /w
7E HAVE DENOTED IN ONLY THE DOMINANT TERM IN THE LIMIT ,
THAT CONTRIBUTES TO THE STRUCTURE FUNCTION 3N 4HIS IS RESPONSIBLE FOR THE
BEHAVIOUR
4HIS SMALL w RESULT FOLLOWS FROM A NON RIGOROUS PERTURBATIVE ANALYSIS
BUT IT HAS BEEN CONlRMED BY ANOTHER FORMAL PERTURBATIVE ANALYSIS IN POWERS
OF THE INVERSE DIMENSION ;#&+,= AS WELL AS BY NUMERICAL SIMULATIONS ;&-6=
-ANY OTHER PHENOMENA HAVE BEEN UNCOVERED IN THE PASSIVE SCALAR PROB
LEM &OR INSTANCE IF THE INCOMPRESSIBILITY PROPERTY OF THE VELOCITY lELD IS
RELAXED AND ONE INTRODUCES A PARAMETER INTERPOLATING BETWEEN DIVERGENCE
LESS AND POTENTIAL lELDS A PHASE TRANSITION OCCURS AS THE PARAMETER IS VARIED
TO A STATE WHERE THERE IS AN INVERSE CASCADE OF ENERGY TO LARGE DISTANCES AND
NO ANOMALOUS SCALING ;#+6 '6=
,ESSONS AND PROSPECTS 4HE MAIN DIkERENCE BETWEEN THE +RAICHNAN
MODEL DISCUSSED ABOVE AND THE .3 EQUATION IS THAT THE FORMER IS EXPLICITLY
SOLVABLE 4HE CORRELATION FUNCTION 'N X XN SATISlES THE EQUATION
8 o p
X _X
-N 'N 'N_ # I , J
IJ
WHERE THE ARGUMENTS XI XJ ARE ABSENT IN THE 'N_ 4HIS ALLOWS FOR THE
ITERATIVE CONSTRUCTION OF ALL 'N IN TERMS OF # AND THE 'REEN FUNCTIONS OF
THE OPERATORS -N 4HE CORRESPONDING EQUATION FOR THE STATIONARY STATE
OF .3 THE SO CALLED (OPF EQUATION INVOLVES 'N ON THE 2(3 DUE TO THE
NONLINEAR TERM IN .3 4HUS THE EQUATIONS DONT CLOSE AND THERE HAS BEEN A
LONG AND UNSUCCESSFUL HISTORY OF lNDING CLOSURES IE APPROXIMATIONS WHERE
6ISIONS IN -ATH ,%33/.3 &/2 452"5,%.#%
ONLY A lNITE NUMBER OF EQUATIONS NEED TO BE CONSIDERED (ENCE VERY LITTLE

CAN BE LEARNED FOR SOLVING .3 FROM THE ACTUAL SOLUTION OF THE +RAICHNAN
MODEL (OWEVER SOME OF THE QUALITATIVE FEATURES UNCOVERED IN THE PASSIVE
SCALAR COULD HAVE THEIR CORRESPONDENCE IN THE .3 TURBULENCE
/NE SUCH FEATURE IS THE THE MECHANISM FOR ENERGY DISSIPATION IN THE
INlNITE 2EYNOLDS NUMBER LIMIT )N THE +RAICHNAN MODEL THE ,AGRANGIAN
TRAJECTORIES BECOME STOCHASTIC ALREADY IN A lXED REALIZATION OF THE VELOCITY
lELD DUE TO THE ROUGHNESS OF THE LATTER 3IMILAR PHENOMENON SHOULD OCCUR
IN MORE GENERAL VELOCITY ENSEMBLES THUS PROVIDING A MECHANISM FOR THE
ENERGY DISSIPATION IN WEAK SOLUTIONS OF THE %ULER EQUATION 4HE %ULER
EQUATION IS KNOWN TO HAVE ENERGY NONCONSERVING SOLUTIONS ;3= BUT THESE
ARE UNPHYSICAL THE ENERGY MAY ALSO INCREASE )T WOULD BE VERY INTERESTING
TO HAVE A NATURAL CLASS OF DISSIPATING WEAK SOLUTIONS 4HE RECENT WORK
BY 3HNIRELMAN ;3H= AND BY $UCHON AND 2OBERT ;$2= IS BUILDING TO THIS
DIRECTION
)N THE +RAICHNAN MODEL INTERMITTENCY IS RELATED TO NONTRIVIAL ZERO
MODES OF THE OPERATORS -N 4HIS CAN BE INTERPRETED IN TERMS OF THE
,AGRANGIAN TRAJECTORIES AS CONSERVED STRUCTURES IN THE N PARTICLE MOTION
;"ER'+= 4HIS AGAIN IS A FEATURE THAT SHOULD BE STUDIED IN MORE GENERAL
VELOCITY ENSEMBLES
7HILE THE INTERMITTENCY PROBLEM IN D TURBULENCE PROBABLY IS BEYOND
CURRENT IDEAS FOR SOME INTERESTING WORK IN THAT DIRECTION HOWEVER SEE
;"E,00= THE TWO DIMENSIONAL PROBLEM COULD BE ACCESSIBLE 4HE CONJEC
TURED AND EXPERIMENTALLY AND NUMERICALLY SUPPORTED ABSENCE OF INTERMIT
TENCY IN BOTH THE INVERSE AND THE DIRECT CASCADE IS INTRIGUING AND AN EXACT
ANALYSIS IS NOT EXCLUDED )N THE ABSENCE OF THAT IT WOULD STILL BE VERY
INTERESTING TO GET SOME MATHEMATICAL HOLD OF THE INVERSE CASCADE &OR IN
STANCE IT WOULD BE NICE TO SEE SOME GROWTH OF NORMS MEASURING SHORT WAVE
NUMBER ACTIVITY IN THE MANNER OF ;+U "OU= WHO IN THE CASE OF THE NON
LINEAR 3CHRODINGER EQUATION STUDIED THE INCREASE OF HIGHER 3OBOLEV NORMS
SIGNALING A DIRECT CASCADE
&INALLY IT WOULD BE NICE TO HAVE A NONLINEAR MODEL WHERE A MATHE
MATICAL CONTROL OF THE CASCADE PICTURE COULD BE CARRIED OUT ! CANDIDATE
FOR THIS IS THE SO CALLED <WEAK TURBULENCE PHENOMENON ;:,&= ! PRIME
EXAMPLE OCCURS IN THE NONLINEAR 3CHRODINGER EQUATION A (AMILTONIAN SYS
TEM 4HIS EQUATION IS KNOWN TO HAVE AN INVARIANT MEASURE GIVEN BY THE
EXPONENTIAL OF THE (AMILTONIAN ;"OU= (OWEVER THIS EQUILIBRIUM MEASURE
IS NOT THE ONE RELEVANT FOR TURBULENCE BUT ANOTHER ONE OBTAINED BY ADDING
!.44) +50)!).%. '!&!
A SMALL DISSIPATIVE TERM AND A LARGE SCALE FORCING TO THE EQUATION 5NLIKE
IN .3 HOWEVER THE NONLINEAR 3CHRODINGER EQUATION HAS A COUPLING CONSTANT
MULTIPLYING THE NONLINEARITY AND THE TURBULENT STATE IS BELIEVED TO BE AC
CESSIBLE BY PERTURBATION THEORY AROUND A CLOSURE ANSATZ )T WOULD BE VERY
INTERESTING TO CARRY OUT SUCH AN ANALYSIS ON A MORE MATHEMATICAL LEVEL
2EFERENCES
;"= '+ "ATCHELOR #OMPUTATION OF THE ENERGY SPECTRUM IN HOMOGENEOUS
TWO DIMENSIONAL TURBULENCE 0HYS &LUIDS 3UPPL )) [
;"E,00= 6) "ELINCHER 6 ,VOV ! 0OMYALOV ) 0ROCACCIA #OMPUTING
THE SCALING EXPONENTS IN mUID TURBULENCE FROM lRST PRINCIPLES $EMONSTRA
TION OF MULTISCALING * OF 3TAT 0HYS [
;"EN#"2= 2 "ENZI 3 #ILIBERTO # "AUDET ' 2UIZ #HAVARIA /N THE
SCALING OF THREE DIMENSIONAL HOMOGENEOUS AND ISOTROPIC TURBULENCE 0HYS
ICA $ [
;"ER'+= $ "ERNARD + 'AWE xDZKI ! +UPIAINEN 3LOW MODES IN PASSIVE
ADVECTION * 3TAT 0HYS
;"O#6= ' "OFFETTA ! #ELANI - 6ERGASSOLA )NVERSE CASCADE IN TWO
DIMENSIONAL TURBULENCE DEVIATIONS FROM 'AUSSIANITY CHAO DYN
;"OU= * "OURGAIN )NVARIANT MEASURES FOR THE $ DEFOCUSING NONLINEAR
3CHRODINGER EQUATION #OMMUN -ATH 0HYSICS [
;"OU= * "OURGAIN /N GROWTH IN TIME OF 3OBOLEV NORMS OF SMOOTH SOLUTIONS OF
NONLINEAR 3CHRODINGER EQUATIONS * D!NALYSE -ATH [
;"R+,= * "RICMONT ! +UPIAINEN 2 ,EFEVERE 0ROBABILISTIC ESTIMATES FOR
THE TWO DIMENSIONAL STOCHASTIC .AVIER 3TOKES EQUATIONS * 3TAT 0HYS TO
APPEAR
;#&+,= - #HERTKOV ' &ALKOVICH ) +OLOKOLOV 6 ,EBEDEV .ORMAL
AND ANOMALOUS SCALING OF THE FOURTH ORDER CORRELATION FUNCTION OF A RAN
DOMLY ADVECTED SCALAR 0HYS 2EV % [
;#+6= - #HERTKOV ) +OLOKOLOV - 6ERGASSOLA )NVERSE VERSUS DIRECT
CASCADES IN TURBULENT ADVECTION 0HYS 2EV ,ETT [
;#O&4= 0 #ONSTANTIN # &OIAS 2 4EMAM /N THE DIMENSION OF THE ATTRAC
TORS IN TWO DIMENSIONAL TURBULENCE 0HYS $ [
;$2= * $UCHON 2 2OBERT )NERTIAL ENERGY DISSIPATION FOR WEAK SOLUTIONS OF
INCOMPRESSIBLE %ULER AND .AVIER 3TOKES EQUATIONS .ONLINEARITY TO APPEAR
PREPRINT
;&-6= 5 &RISCH ! -AZZINO - 6ERGASSOLA )NTERMITTENCY IN PASSIVE
SCALAR ADVECTION 0HYS 2EV ,ETT [
;'+= + 'AWE xDZKI ! +UPIAINEN 5NIVERSALITY IN TURBULENCE AN EXACTLY SOLV
ABLE MODEL IN <,OW $IMENSIONAL -ODELS IN 3TATISTICAL -ECHANICS AND
1UANTUM &IELD 4HEORY ( 'ROSSE , 0ITTNER EDS 3PRINGER "ERLIN
6ISIONS IN -ATH ,%33/.3 &/2 452"5,%.#%
[
;'+= + 'AWE xDZKI ! +UPIAINEN !NOMALOUS SCALING OF THE PASSIVE SCALAR
0HYS 2EV ,ETT [
;'6= + 'AWE xDZKI - 6ERGASSOLA 0HASE TRANSITION IN THE PASSIVE SCALAR
ADVECTION COND MAT TO APPEAR IN 0HYSICA $
;+= !. +OLMOGOROV 4HE LOCAL STRUCTURE OF TURBULENCE IN INCOMPRESSIBLE
VISCOUS mUID FOR VERY LARGE 2EYNOLDS NUMBERS #2 !CAD 3CI 5233
[
;+R= 2( +RAICHNAN )NERTIAL RANGES IN TWO DIMENSIONAL TURBULENCE 0HYS
&LUIDS [
;+R= 2( +RAICHNAN 3MALL SCALE STRUCTURE OF A SCALAR lELD CONVECTED BY
TURBULENCE 0HYS &LUIDS [
;+U= 3 +UKSIN 3PECTRAL PROPERTIES OF SOLUTIONS FOR NONLINEAR 0$%S IN THE
TURBULENT REGIME '!&!
;,= /! ,ADYZHENSKAYA /N DYNAMICAL SYSTEM GENERATED BY THE .AVIER
3TOKES EQUATION * 3OVIET -ATH
;,E2= 9 ,E *AN / 2AIMOND 3OLUTION STATISTIQUES FORTES DES sEQUATIONS
DIksERENTIELLES STOCHASTIQUES #2 !CAD 3CI 0ARIS 3ER ) -ATH
[
;,ER= * ,ERAY %SSAI SUR LE MOUVEMENT DUN LIQUIDE VISQUEUX REMPLISSANT
LESPACE !CTA -ATHEMATICA [
;/= , /NSAGER .UOVO #IM 3UPPL [
;0*4= * 0ARET -# *ULLIEN 0 4ABELING 6ORTICITY STATISTICS IN THE TWO
DIMENSIONAL ENSTROPHY CASCADE COND MAT
;04= * 0ARET 0 4ABELING )NTERMITTENCY IN THE $ INVERSE CASCADE OF EN
ERGY EXPERIMENTAL OBSERVATIONS 0HYS &LUIDS [
;3= 63CHEFFER (AUSDORk MEASURE AND THE .AVIER 3TOKES EQUATION #OM
MUN -ATH 0HYS [
;3= 6 3CHEFFER !N INVISCID mOW WITH COMPACT SUPPORT IN SPACE TIME *
'EOM !NAL [
;3H= ! 3HNIRELMAN 7EAK SOLUTIONS WITH DECREASING ENERGY OF INCOMPRESSIBLE
%ULER EQUATIONS )(%3- PREPRINT
;6&= -* 6ISHIK !6 &URSIKOV -ATHEMATICAL 0ROBLEMS OF 3TATISTICAL (Y
DRODYNAMICS +LUWER $ORDRECHT
;:,&= 6% :AKHAROV 63 ,VOV ' &ALKOVICH +OLMOGOROV 3PECTRA OF
4URBULENCE ) 3PRINGER "ERLIN
!NTTI +UPIAINEN -ATHEMATICS $EPARTMENT (ELSINKI 5NIVERSITY 0/"OX

(ELSINKI &INLAND
4(% -!4(%-!4)#3 /& 4(% 3%#/.$ ,!7 /&

4(%2-/$9.!-)#3
%LLIOTT ( ,IEB AND *AKOB 9NGVASON
!BSTRACT
4HE ESSENCE OF THE SECOND LAW IS THE @ENTROPY PRINCIPLE WHICH STATES
THAT ADIABATIC PROCESSES CAN BE QUANTIlED BY AN ENTROPY FUNCTION ON
THE SPACE OF ALL EQUILIBRIUM STATES WHOSE INCREASE IS A NECESSARY AND
SUnCIENT CONDITION FOR SUCH A PROCESS TO OCCUR )T IS ONE OF THE FEW
REALLY FUNDAMENTAL PHYSICAL LAWS IN THE SENSE THAT NO DEVIATION HOW
EVER TINY IS PERMITTED AND ITS CONSEQUENCES ARE FAR REACHING 3INCE
THE ENTROPY PRINCIPLE IS INDEPENDENT OF MODELS STATISTICAL MECHANICAL
OR OTHERWISE IT OUGHT TO BE DERIVABLE FROM A FEW LOGICAL PRINCIPLES
WITHOUT RECOURSE TO #ARNOT CYCLES IDEAL GASES AND OTHER ASSUMPTIONS
ABOUT SUCH THINGS AS @HEAT @HOT AND @COLD @TEMPERATURE @REVERSIBLE
PROCESSES ETC AS IS USUALLY
0 DONE 4HE WELL KNOWN FORMULA OF STATISTI
CAL MECHANICS 3 _ P LOG P IS IRRELEVANT FOR THIS PROBLEM )N THIS
PAPER THE FOUNDATIONS OF THE SUBJECT AND THE CONSTRUCTION OF ENTROPY
FROM A FEW SIMPLE AXIOMS WILL BE PRESENTED 4HE AXIOMS BASICALLY
ARE THOSE OF A PREORDER EXCEPT FOR AN IMPORTANT ADDITIONAL PROPERTY
CALLED @THE COMPARISON HYPOTHESIS WHICH WE ANALYZE IN DETAIL AND
DERIVE FROM OTHER AXIOMS )T CAN BE SAID THAT THIS THEORY ADDRESSES
THE QUESTION @7HEN IS A PREORDER ON A SET EQUIVALENT TO A MONOTONE
FUNCTION ON THE SET !S SUCH IT COULD CONCEIVABLY BE USEFUL IN OTHER
AREAS OF MATHEMATICS &INALLY WE CONSIDER SOME OPEN PROBLEMS AND
DIRECTIONS FOR FURTHER STUDY
&OREWORD
!S PART OF THE CONFERENCE <6ISIONS IN -ATHEMATICS 4OWARDS 4EL
!VIV ONE OF US %, CONTRIBUTED A TALK WITH THE ABOVE TITLE )T
%,S WORK PARTIALLY SUPPORTED BY 53 .ATIONAL 3CIENCE &OUNDATION GRANT 0(9
*9S WORK PARTIALLY SUPPORTED BY THE !DALSTEINN +RISTJANSSON &OUNDATION 5NI
VERSITY OF )CELAND
C
l BY THE AUTHORS 2EPRODUCTION OF THIS ARTICLE BY ANY MEANS IS PERMITTED FOR
NON COMMERCIAL PURPOSES
6ISIONS IN -ATH 3%#/.$ ,!7 /& 4(%2-/$9.!-)#3
WAS A REVIEW OF OUR ATTEMPT IN ;,I9= AT UNDERSTANDING THE MATHEMATICAL

STRUCTURE OF ONE OF THE MOST FAR REACHING AND EXACT LAWS OF PHYSICS 4O
SOME EXTENT WE SUCCEEDED BUT NOTHING IS PERFECT AND OPEN PROBLEMS ABOUT
THE MEANING OF ENTROPY REMAIN 4HE CONFERENCE ORGANIZERS REQUEST FOR A
RECORD OF THE TALK AND AN ADDENDUM ABOUT OPEN PROBLEMS COULD BEST BE
MET WE BELIEVE BY REPRODUCING A SUMMARY OF ;,I9= THAT APPEARED IN THE
!-3 .OTICES ;,I9= AND BY ADDING A NEW SECTION ABOUT OPEN PROBLEMS
4HUS THE lRST PART IS THE .OTICES ARTICLE WITH SOME MODIlCATIONS !
SHORTER SUMMARY APPEARS IN ;,I9= 4HE OPEN PROBLEMS SECTION IS HIGHLY
SPECULATIVE AND IS HERE LARGELY TO SHARE THE SENSE THAT THE 3ECOND ,AW IS
STILL OPEN TO DISCUSSION AND CREATIVE IDEAS 3OME OF THE PROBLEMS IN PART
WERE MENTIONED IN ;,I= ! RECENT SCHOLARLY PEDAGOGICAL REVIEW OF VARIOUS
HISTORICAL APPROACHES TO CLASSICAL THERMODYNAMICS AND THE 3ECOND ,AW IS
IN ;5=
! 'UIDE TO %NTROPY AND THE 3ECOND ,AW OF

4HERMODYNAMICS ;,I9=
4HIS ARTICLE IS INTENDED FOR READERS WHO LIKE US WERE TOLD THAT THE SECOND
LAW OF THERMODYNAMICS IS ONE OF THE MAJOR ACHIEVEMENTS OF THE NINETEENTH
CENTURY THAT IT IS A LOGICAL PERFECT AND UNBREAKABLE LAW [ BUT WHO WERE
UNSATISlED WITH THE @DERIVATIONS OF THE ENTROPY PRINCIPLE AS FOUND IN TEXT
BOOKS AND IN POPULAR WRITINGS
! GLANCE AT THE BOOKS WILL INFORM THE READER THAT THE LAW HAS @VARIOUS
FORMULATIONS WHICH IS A BIT ODD AS IF TO SAY THE TEN COMMANDMENTS HAVE
VARIOUS FORMULATIONS BUT THEY ALL LEAD TO THE EXISTENCE OF AN ENTROPY FUNC
TION WHOSE REASON FOR EXISTENCE IS TO TELL US WHICH PROCESSES CAN OCCUR AND
WHICH CANNOT #ONTRARY TO CONVENTION WE SHALL REFER TO THE EXISTENCE OF
ENTROPY AS THE SECOND LAW 4HIS AT LEAST IS UNAMBIGUOUS 4HE ENTROPY WE
ARE TALKING ABOUT IS THAT DElNED BY THERMODYNAMICS AND NOT SOME ANA
LYTIC QUANTITY USUALLY INVOLVING EXPRESSIONS SUCH AS _P LN P THAT APPEARS IN
INFORMATION THEORY PROBABILITY THEORY AND STATISTICAL MECHANICAL MODELS
7HY ONE MIGHT ASK SHOULD A MATHEMATICIAN BE INTERESTED IN THE SEC
OND LAW OF THERMODYNAMICS WHICH HISTORICALLY HAD SOMETHING TO DO WITH
ATTEMPTS TO UNDERSTAND AND IMPROVE THE EnCIENCY OF STEAM ENGINES 4HE
ANSWER AS WE PERCEIVE IT IS THAT THE LAW IS REALLY AN INTERESTING MATHEMAT
ICAL THEOREM ABOUT ORDERINGS ON SETS WITH PROFOUND PHYSICAL IMPLICATIONS
4HE AXIOMS THAT CONSTITUTE THIS ORDERING ARE SOMEWHAT PECULIAR FROM THE
%( ,)%" !.$ * 9.'6!3/. '!&!
MATHEMATICAL POINT OF VIEW AND MIGHT NOT ARISE IN THE ORDINARY RUMINA
TIONS OF ABSTRACT THOUGHT 4HEY ARE SPECIAL BUT IMPORTANT AND THEY ARE
DRIVEN BY CONSIDERATIONS ABOUT THE WORLD WHICH IS WHAT MAKES THEM SO IN
TERESTING -AYBE AN INGENIOUS READER WILL lND AN APPLICATION OF THIS SAME
LOGICAL STRUCTURE TO ANOTHER lELD OF SCIENCE
#LASSICAL THERMODYNAMICS AS IT IS USUALLY PRESENTED IS BASED ON THREE
LAWS PLUS ONE MORE DUE TO .ERNST WHICH IS MAINLY USED IN LOW TEMPERATURE
PHYSICS AND IS NOT IMMUTABLE LIKE THE OTHERS )N BRIEF THESE ARE
4HE :EROTH ,AW WHICH EXPRESSES THE TRANSITIVITY OF EQUILIBRIUM

AND WHICH IS OFTEN SAID TO IMPLY THE EXISTENCE OF TEMPERATURE
AS A PARAMETRIZATION OF EQUILIBRIUM STATES 7E USE IT BELOW BUT
FORMULATE IT WITHOUT MENTIONING TEMPERATURE )N FACT TEMPER
ATURE MAKES NO APPEARANCE HERE UNTIL ALMOST THE VERY END
4HE &IRST ,AW WHICH IS CONSERVATION OF ENERGY )T IS A CONCEPT
FROM MECHANICS AND PROVIDES THE CONNECTION BETWEEN MECHANICS
AND THINGS LIKE FALLING WEIGHTS AND THERMODYNAMICS 7E DIS
CUSS THIS LATER ON WHEN WE INTRODUCE SIMPLE SYSTEMS THE CRUCIAL
USAGE OF THIS LAW IS THAT IT ALLOWS ENERGY TO BE USED AS ONE OF
THE PARAMETERS DESCRIBING THE STATES OF A SIMPLE SYSTEM
4HE 3ECOND ,AW 4HREE POPULAR FORMULATIONS OF THIS LAW ARE
#LAUSIUS .O PROCESS IS POSSIBLE THE SOLE RESULT OF
WHICH IS THAT HEAT IS TRANSFERRED FROM A BODY TO A
HOTTER ONE
+ELVIN AND 0LANCK .O PROCESS IS POSSIBLE THE SOLE
RESULT OF WHICH IS THAT A BODY IS COOLED AND WORK IS
DONE
#ARATHsEODORY )N ANY NEIGHBORHOOD OF ANY STATE THERE
ARE STATES THAT CANNOT BE REACHED FROM IT BY AN ADIA
BATIC PROCESS
!LL THREE ARE SUPPOSED TO LEAD TO THE ENTROPY PRINCIPLE DElNED BELOW
4HESE STEPS CAN BE FOUND IN MANY BOOKS AND WILL NOT BE TRODDEN AGAIN
HERE ,ET US NOTE IN PASSING HOWEVER THAT THE lRST TWO USE CONCEPTS SUCH
AS HOT COLD HEAT COOL THAT ARE INTUITIVE BUT HAVE TO BE MADE PRECISE BEFORE
THE STATEMENTS ARE TRULY MEANINGFUL .O ONE HAS SEEN @HEAT FOR EXAMPLE
4HE LAST WHICH USES THE TERM <ADIABATIC PROCESS TO BE DElNED BELOW
PRESUPPOSES SOME KIND OF PARAMETRIZATION OF STATES BY POINTS IN 2N AND
THE USUAL DERIVATION OF ENTROPY FROM IT ASSUMES SOME SORT OF DIkERENTIABILITY
6ISIONS IN -ATH 3%#/.$ ,!7 /& 4(%2-/$9.!-)#3
SUCH ASSUMPTIONS ARE BESIDE THE POINT AS FAR AS UNDERSTANDING THE MEANING
OF ENTROPY GOES
4HE BASIC INPUT IN OUR ANALYSIS OF THE SECOND LAW IS A CERTAIN KIND OF
ORDERING ON A SET AND DENOTED BY
}
PRONOUNCED @PRECEDES )T IS TRANSITIVE AND REmEXIVE AS IN ! ! BELOW
BUT 8 } 9 AND 9 } 8 DOES NOT IMPLY 8 9 SO IT IS A @PREORDER 4HE BIG
QUESTION IS WHETHER } CAN BE ENCODED IN AN ORDINARY REAL VALUED FUNCTION
ON THE SET DENOTED BY 3 SUCH THAT IF 8 AND 9 ARE RELATED BY } THEN
38 s 39 IF AND ONLY IF 8 } 9 4HE FUNCTION 3 IS ALSO REQUIRED TO BE
ADDITIVE AND EXTENSIVE IN A SENSE THAT WILL SOON BE MADE PRECISE
! HELPFUL ANALOGY IS THE QUESTION 7HEN CAN A VECTOR lELD 6 X ON 2
BE ENCODED IN AN ORDINARY FUNCTION F X WHOSE GRADIENT IS 6 4HE WELL
KNOWN ANSWER IS THAT A NECESSARY AND SUnCIENT CONDITION IS THAT CURL 6
/NCE 6 IS OBSERVED TO HAVE THIS PROPERTY ONE THING BECOMES EVIDENT AND
IMPORTANT )T IS NECESSARY TO MEASURE THE INTEGRAL OF 6 ONLY ALONG SOME
CURVES [ NOT ALL CURVES [ IN ORDER TO DEDUCE THE INTEGRAL ALONG ALL CURVES 4HE
ENCODING THEN HAS ENORMOUS PREDICTIVE POWER ABOUT THE NATURE OF FUTURE
MEASUREMENTS OF 6 )N THE SAME WAY KNOWLEDGE OF THE FUNCTION 3 HAS
ENORMOUS PREDICTIVE POWER IN THE HANDS OF CHEMISTS ENGINEERS AND OTHERS
CONCERNED WITH THE WAYS OF THE PHYSICAL WORLD
/UR CONCERN WILL BE THE EXISTENCE AND PROPERTIES OF 3 STARTING FROM
CERTAIN NATURAL AXIOMS ABOUT THE RELATION } 7E PRESENT OUR RESULTS WITHOUT
PROOFS BUT FULL DETAILS AND A DISCUSSION OF RELATED PREVIOUS WORK ON THE
FOUNDATIONS OF CLASSICAL THERMODYNAMICS ARE GIVEN IN ;,I9= 4HE LITERATURE
ON THIS SUBJECT IS EXTENSIVE AND IT IS NOT POSSIBLE TO GIVE EVEN A BRIEF ACCOUNT
OF IT HERE EXCEPT FOR MENTIONING THAT THE PREVIOUS WORK CLOSEST TO OURS IS
THAT OF ;'I= AND ;"U= SEE ALSO ;#O= ;$= AND ;2,= 4HESE OTHER APPROACHES
ARE ALSO BASED ON AN INVESTIGATION OF THE RELATION } BUT THE OVERLAP WITH OUR
WORK IS ONLY PARTIAL )N FACT A MAJOR PART OF OUR WORK IS THE DERIVATION OF
A CERTAIN PROPERTY THE <COMPARISON HYPOTHESIS BELOW WHICH IS TAKEN AS
AN AXIOM IN THE OTHER APPROACHES )T WAS A REMARKABLE AND LARGELY UNSUNG
ACHIEVEMENT OF 'ILES ;'I= TO REALIZE THE FULL POWER OF THIS PROPERTY
,ET US BEGIN THE STORY WITH SOME BASIC CONCEPTS
n 4HERMODYNAMIC 3YSTEM 0HYSICALLY THIS CONSISTS OF CERTAIN SPECIlED
AMOUNTS OF CERTAIN KINDS OF MATTER EG A GRAM OF HYDROGEN IN A
CONTAINER WITH A PISTON OR A GRAM OF HYDROGEN AND A GRAM OF OXYGEN
IN TWO SEPARATE CONTAINERS OR A GRAM OF HYDROGEN AND TWO GRAMS OF
%( ,)%" !.$ * 9.'6!3/. '!&!
HYDROGEN IN SEPARATE CONTAINERS 4HE SYSTEM CAN BE IN VARIOUS STATES

WHICH PHYSICALLY ARE EQUILIBRIUM STATES 4HE SPACE OF STATES OF THE
SYSTEM IS USUALLY DENOTED BY A SYMBOL SUCH AS ` AND STATES IN ` BY
8 9 : ETC
0HYSICAL MOTIVATION ASIDE A STATE SPACE MATHEMATICALLY IS JUST A SET
[ TO BEGIN WITH LATER ON WE WILL BE INTERESTED IN EMBEDDING STATE SPACES
IN SOME CONVEX SUBSET OF SOME 2N IE WE WILL INTRODUCE COORDINATES
!S WE SAID EARLIER HOWEVER THE ENTROPY PRINCIPLE IS QUITE INDEPENDENT OF
COORDINATIZATION #ARATHsEODORYS PRINCIPLE NOTWITHSTANDING
#OMPOSITION AND SCALING OF STATES 4HE NOTION OF #ARTESIAN PRODUCT
` a` CORRESPONDS SIMPLY TO THE TWO OR MORE SYSTEMS BEING SIDE BY
SIDE ON THE LABORATORY TABLE MATHEMATICALLY IT IS JUST ANOTHER SYSTEM
CALLED A COMPOUND SYSTEM AND WE REGARD THE STATE SPACE ` a ` AS
THE SAME AS ` a ` 0OINTS IN ` a ` ARE DENOTED BY PAIRS 8 9 AS
USUAL 4HE SUBSYSTEMS COMPRISING A COMPOUND SYSTEM ARE PHYSICALLY
INDEPENDENT SYSTEMS BUT THEY ARE ALLOWED TO INTERACT WITH EACH OTHER
FOR A PERIOD OF TIME AND THEREBY ALTER EACH OTHERS STATE
4HE CONCEPT OF SCALING IS CRUCIAL )T IS THIS CONCEPT THAT MAKES OUR
THERMODYNAMICS INAPPROPRIATE FOR MICROSCOPIC OBJECTS LIKE ATOMS OR
COSMIC OBJECTS LIKE STARS &OR EACH STATE SPACE ` AND NUMBER t
THERE IS ANOTHER STATE SPACE DENOTED BY `t WITH POINTS DENOTED BY
t8 4HIS SPACE IS CALLED A SCALED COPY OF ` /F COURSE WE IDENTIFY
` ` AND 8 8 7E ALSO REQUIRE `t u `tu AND ut8
ut 8 4HE PHYSICAL INTERPRETATION OF `t WHEN ` IS THE SPACE OF ONE
GRAM OF HYDROGEN IS SIMPLY THE STATE SPACE OF t GRAMS OF HYDROGEN
4HE STATE t8 IS THE STATE OF t GRAMS OF HYDROGEN WITH THE SAME
@INTENSIVE PROPERTIES AS 8 EG PRESSURE WHILE @EXTENSIVE PROPERTIES
LIKE ENERGY VOLUME ETC ARE SCALED BY A FACTOR t BY DElNITION
&OR ANY GIVEN ` WE CAN FORM #ARTESIAN PRODUCT STATE SPACES OF THE TYPE
`t a `t a ` ` ` a `t. 4HESE WILL BE CALLED MULTIPLE SCALED COPIES OF `
4HE NOTATION `t SHOULD BE REGARDED AS MERELY A MNEMONIC AT THIS
POINT BUT LATER ON WITH THE EMBEDDING OF ` INTO 2N IT WILL LITERALLY BE
t` Ft8 8 `G IN THE USUAL SENSE
!DIABATIC ACCESSIBILITY .OW WE COME TO THE ORDERING 7E SAY 8 } 9
WITH 8 AND 9 POSSIBLY IN DIkERENT STATE SPACES IF 9 IS ADIABATICALLY
ACCESSIBLE FROM 8 ACCORDING TO THE DElNITION BELOW
7HAT DOES THIS MEAN -ATHEMATICALLY WE ARE JUST GIVEN A LIST OF PAIRS
6ISIONS IN -ATH 3%#/.$ ,!7 /& 4(%2-/$9.!-)#3
8 } 9 4HERE IS NOTHING MORE TO BE SAID EXCEPT THAT LATER ON WE WILL

ASSUME THAT THIS LIST HAS CERTAIN PROPERTIES THAT WILL LEAD TO INTERESTING
THEOREMS ABOUT THIS LIST AND WILL LEAD IN TURN TO THE EXISTENCE OF AN
ENTROPY FUNCTION 3 CHARACTERIZING THE LIST
4HE PHYSICAL INTERPRETATION IS QUITE ANOTHER MATTER )N TEXT BOOKS A
PROCESS TAKING 8 TO 9 IS USUALLY CALLED ADIABATIC IF IT TAKES PLACE IN @THER
MAL ISOLATION WHICH IN TURN MEANS THAT @NO HEAT IS EXCHANGED WITH THE
SURROUNDINGS 3UCH CONCEPTS HEAT THERMAL ETC APPEAR INSUnCIENTLY
PRECISE TO US AND WE PREFER THE FOLLOWING VERSION WHICH IS IN THE SPIRIT OF
0LANCKS FORMULATION OF THE SECOND LAW ;0= AND AVOIDS THOSE CONCEPTS /UR
DElNITION OF ADIABATIC ACCESSIBILITY MIGHT AT lRST SIGHT APPEAR TO BE LESS
RESTRICTIVE THAN THE USUAL ONE BUT AS WE HAVE SHOWN IN ;,I9= IN THE END
ANYTHING THAT WE CALL AN ADIABATIC PROCESS MEANING THAT 9 IS ADIABATICALLY
ACCESSIBLE FROM 8 CAN ALSO BE ACCOMPLISHED IN @THERMAL ISOLATION AS THE
CONCEPT IS USUALLY UNDERSTOOD /UR DElNITION HAS THE GREAT VIRTUE AS DIS
COVERED BY 0LANCK THAT IT AVOIDS HAVING TO DISTINGUISH BETWEEN WORK AND
HEAT [ OR EVEN HAVING TO DElNE THE CONCEPT OF HEAT 7E EMPHASIZE HOWEVER
THAT THE THEOREMS DO NOT REQUIRE AGREEMENT WITH OUR PHYSICAL DElNITION OF
ADIABATIC ACCESSIBILITY OTHER DElNITIONS ARE CONCEIVABLY POSSIBLE 7E EM
PHASIZE ALSO THAT WE DO NOT CARE ABOUT THE TEMPORAL DEVELOPMENT INVOLVED
IN THE STATE CHANGE WE ONLY CARE ABOUT THE NET RESULT FOR THE SYSTEM AND
THE REST OF THE UNIVERSE
! STATE 9 IS ADIABATICALLY ACCESSIBLE FROM A STATE 8 IN SYMBOLS
8 } 9 IF IT IS POSSIBLE TO CHANGE THE STATE FROM 8 TO 9 BY
MEANS OF AN INTERACTION WITH SOME DEVICE CONSISTING OF SOME
AUXILIARY SYSTEM AND A WEIGHT IN SUCH A WAY THAT THE AUXIL
IARY SYSTEM RETURNS TO ITS INITIAL STATE AT THE END OF THE PROCESS
WHEREAS THE WEIGHT MAY HAVE RISEN OR FALLEN
4HE ROLE OF THE @WEIGHT IN THIS DElNITION IS MERELY TO PROVIDE A PARTICU
LARLY SIMPLE SOURCE OR SINK OF MECHANICAL ENERGY .OTE THAT AN ADIABATIC
PROCESS PHYSICALLY DOES NOT HAVE TO BE GENTLE OR @STATIC OR ANYTHING OF THE
KIND )T CAN BE ARBITRARILY VIOLENT 4HE @DEVICE NEED NOT BE A WELL DElNED
MECHANICAL CONTRAPTION )T CAN BE ANOTHER THERMODYNAMIC SYSTEM AND
EVEN A GORILLA JUMPING UP AND DOWN ON THE SYSTEM OR A COMBINATION OF
THESE [ AS LONG AS THE DEVICE RETURNS TO ITS INITIAL STATE
!N EXAMPLE MIGHT BE USEFUL HERE 4AKE A POUND OF HYDROGEN IN A
CONTAINER WITH A PISTON 4HE STATES ARE DESCRIBABLE BY TWO NUMBERS ENERGY
AND VOLUME THE LATTER BEING DETERMINED BY THE POSITION OF THE PISTON
%( ,)%" !.$ * 9.'6!3/. '!&!
3TARTING FROM SOME STATE 8 WE CAN TAKE OUR HAND Ok THE PISTON AND LET
THE VOLUME INCREASE EXPLOSIVELY TO A LARGER ONE !FTER THINGS HAVE CALMED
DOWN CALL THE NEW EQUILIBRIUM STATE 9 4HEN 8 } 9 1UESTION )S 9 } 8
TRUE !NSWER .O 4O GET FROM 9 TO 8 ADIABATICALLY WE WOULD HAVE TO USE
SOME MACHINERY AND A WEIGHT WITH THE MACHINERY RETURNING TO ITS INITIAL
STATE AND THERE IS NO WAY THIS CAN BE DONE 5SING A WEIGHT WE CAN INDEED
RECOMPRESS THE GAS TO ITS ORIGINAL VOLUME BUT WE WILL lND THAT THE ENERGY
IS THEN LARGER THAN ITS ORIGINAL VALUE
,ET US WRITE
8 }} 9 IF 8}9 BUT NOT 9 } 8 WRITTEN 9 } 8
)N THIS CASE WE SAY THAT WE CAN GO FROM 8 TO 9 BY AN IRREVERSIBLE ADIABATIC
PROCESS )F 8 } 9 AND 9 } 8 WE SAY THAT 8 AND 9 ARE ADIABATICALLY
EQUIVALENT AND WRITE
!
8w9
!
%QUIVALENCE CLASSES UNDER w ARE CALLED ADIABATS
#OMPARABILITY 'IVEN TWO STATES 8 AND 9 IN TWO SAME OR DIkERENT
STATE SPACES WE SAY THAT THEY ARE COMPARABLE IF 8 } 9 OR 9 } 8
OR BOTH 4HIS TURNS OUT TO BE A CRUCIAL NOTION 4WO STATES ARE NOT
ALWAYS COMPARABLE A NECESSARY CONDITION IS THAT THEY HAVE THE SAME
MATERIAL COMPOSITION IN TERMS OF THE CHEMICAL ELEMENTS %XAMPLE
3INCE WATER IS ( / AND THE ATOMIC WEIGHTS OF HYDROGEN AND OXYGEN
ARE AND RESPECTIVELY THE STATES IN THE COMPOUND SYSTEM OF
GRAM OF HYDROGEN AND GRAMS OF OXYGEN ARE COMPARABLE WITH STATES
IN A SYSTEM CONSISTING OF GRAMS OF WATER BUT NOT WITH GRAMS
OF WATER OR GRAMS OF OXYGEN
!CTUALLY THE CLASSIlCATION OF STATES INTO VARIOUS STATE SPACES IS DONE
MAINLY FOR CONCEPTUAL CONVENIENCE 4HE SECOND LAW DEALS ONLY WITH STATES
AND THE ONLY THING WE REALLY HAVE TO KNOW ABOUT ANY TWO OF THEM IS WHETHER
OR NOT THEY ARE COMPARABLE 'IVEN THE RELATION } FOR ALL POSSIBLE STATES OF
ALL POSSIBLE SYSTEMS WE CAN ASK WHETHER THIS RELATION CAN BE ENCODED IN AN
ENTROPY FUNCTION ACCORDING TO THE FOLLOWING
%NTROPY PRINCIPLE 4HERE IS A REAL VALUED FUNCTION ON ALL STATES OF ALL
SYSTEMS INCLUDING COMPOUND SYSTEMS CALLED ENTROPY AND DENOTED BY 3
SUCH THAT
A -ONOTONICITY 7HEN 8 AND 9 ARE COMPARABLE STATES THEN
8}9 IF AND ONLY IF 38 s 39
6ISIONS IN -ATH 3%#/.$ ,!7 /& 4(%2-/$9.!-)#3
B !DDITIVITY AND EXTENSIVITY )F 8 AND 9 ARE STATES OF SOME

POSSIBLY DIkERENT SYSTEMS AND IF 8 9 DENOTES THE CORRESPOND
ING STATE IN THE COMPOUND SYSTEM THEN THE ENTROPY IS ADDITIVE FOR
THESE STATES IE
38 9 38 39
3 IS ALSO EXTENSIVE IE FOR OR EACH t AND EACH STATE 8 AND ITS
SCALED COPY t8 `t DElNED IN ABOVE
3t8 t38
! FORMULATION LOGICALLY EQUIVALENT TO A NOT USING THE WORD @COMPARA
BLE IS THE FOLLOWING PAIR OF STATEMENTS
!
8 w 9 38 39 AND
8 }} 9 38 39
4HE LAST LINE IS ESPECIALLY NOTEWORTHY )T SAYS THAT ENTROPY MUST INCREASE
IN AN IRREVERSIBLE ADIABATIC PROCESS
4HE ADDITIVITY OF ENTROPY IN COMPOUND SYSTEMS IS OFTEN JUST TAKEN FOR
GRANTED BUT IT IS ONE OF THE STARTLING CONCLUSIONS OF THERMODYNAMICS &IRST
OF ALL THE CONTENT OF ADDITIVITY IS CONSIDERABLY MORE FAR REACHING THAN
ONE MIGHT THINK FROM THE SIMPLICITY OF THE NOTATION #ONSIDER FOUR STATES
8 8 9 9 AND SUPPOSE THAT 8 } 9 AND 8 } 9 /NE OF OUR AXIOMS
! WILL BE THAT THEN 8 8 } 9 9 AND CONTAINS NOTHING NEW
OR EXCITING /N THE OTHER HAND THE COMPOUND SYSTEM CAN WELL HAVE AN
ADIABATIC PROCESS IN WHICH 8 8 } 9 9 BUT 8 } 9 )N THIS CASE
CONVEYS MUCH INFORMATION )NDEED BY MONOTONICITY THERE WILL BE MANY
CASES OF THIS KIND BECAUSE THE INEQUALITY 38 38 s 39 39
CERTAINLY DOES NOT IMPLY THAT 38 s 39 4HE FACT THAT THE INEQUALITY
38 38 s 39 39 TELLS US EXACTLY WHICH ADIABATIC PROCESSES ARE
ALLOWED IN THE COMPOUND SYSTEM AMONG COMPARABLE STATES INDEPENDENT
OF ANY DETAILED KNOWLEDGE OF THE MANNER IN WHICH THE TWO SYSTEMS INTERACT
IS ASTONISHING AND IS AT THE HEART OF THERMODYNAMICS 4HE SECOND REASON
THAT IS STARTLING IS THIS &ROM ALONE RESTRICTED TO ONE SYSTEM THE
FUNCTION 3 CAN BE REPLACED BY 3 AND STILL DO ITS JOB IE SATISFY
(OWEVER SAYS THAT IT IS POSSIBLE TO CALIBRATE THE ENTROPIES OF ALL SYSTEMS
IE SIMULTANEOUSLY ADJUST ALL THE UNDETERMINED MULTIPLICATIVE CONSTANTS
SO THAT THE ENTROPY 3 FOR A COMPOUND ` a ` IS 3 8 9 3 8
3 9 EVEN THOUGH SYSTEMS AND ARE TOTALLY UNRELATED
7E ARE NOW READY TO ASK SOME BASIC QUESTIONS
%( ,)%" !.$ * 9.'6!3/. '!&!
1 7HICH PROPERTIES OF THE RELATION } ENSURE EXISTENCE AND ESSENTIAL

UNIQUENESS OF 3
1 #AN THESE PROPERTIES BE DERIVED FROM SIMPLE PHYSICAL PREMISES
1 7HICH CONVEXITY AND SMOOTHNESS PROPERTIES OF 3 FOLLOW FROM THE
PREMISES
1 #AN TEMPERATURE AND HENCE AN ORDERING OF STATES BY <HOTNESS AND
<COLDNESS BE DElNED FROM 3 AND WHAT ARE ITS PROPERTIES
4HE ANSWER TO QUESTION 1 CAN BE GIVEN IN THE FORM OF SIX AXIOMS
THAT ARE REASONABLE SIMPLE @OBVIOUS AND UNEXCEPTIONABLE !N ADDITIONAL
CRUCIAL ASSUMPTION IS ALSO NEEDED BUT WE CALL IT A @HYPOTHESIS INSTEAD OF
AN AXIOM BECAUSE WE SHOW LATER HOW IT CAN BE DERIVED FROM SOME OTHER
AXIOMS THEREBY ANSWERING QUESTION 1
!
! 2EmEXIVITY 8 w 8
! 4RANSITIVITY )F 8 } 9 AND 9 } : THEN 8 } :
! #ONSISTENCY )F 8 } 8 AND 9 } 9 THEN 8 9 } 8 9
! 3CALING )NVARIANCE )F t AND 8 } 9 THEN t8 } t9
!
! 3PLITTING AND 2ECOMBINATION 8 w _ t 8 t8 FOR ALL
t .OTE THAT THE TWO STATE SPACES ARE DIkERENT )F 8 ` THEN
THE STATE SPACE ON THE RIGHT SIDE IS `_t a `t
! 3TABILITY )F 8 : } 9 : FOR SOME : : AND A SEQUENCE
OF S TENDING TO ZERO THEN 8 } 9 4HIS AXIOM IS A SUBSTITUTE FOR
CONTINUITY WHICH WE CANNOT ASSUME BECAUSE THERE IS NO TOPOLOGY YET
)T SAYS THAT @A GRAIN OF DUST CANNOT INmUENCE THE SET OF ADIABATIC
PROCESSES
!N IMPORTANT LEMMA IS THAT ! [! IMPLY THE CANCELLATION LAW
WHICH IS USED IN MANY PROOFS )T SAYS THAT FOR ANY THREE STATES 8 9 :
8 : } 9 : 8 } 9
4HE NEXT CONCEPT PLAYS A KEY ROLE IN OUR TREATMENT
#( $ElNITION 7E SAY THAT THE #OMPARISON (YPOTHESIS #( HOLDS
FOR A STATE SPACE ` IF ALL PAIRS OF STATES IN ` ARE COMPARABLE
.OTE THAT ! ! AND ! AUTOMATICALLY EXTEND COMPARABILITY FROM A
SPACE ` TO CERTAIN OTHER CASES EG 8 } _ t 9 t: FOR ALL s t s IF
8 } 9 AND 8 } : /N THE OTHER HAND COMPARABILITY ON ` ALONE DOES NOT
ALLOW US TO CONCLUDE THAT 8 IS COMPARABLE TO _ t 9 t: IF 8 } 9 BUT
: } 8 &OR THIS ONE NEEDS #( ON THE PRODUCT SPACE `_t a `t WHICH
IS NOT IMPLIED BY #( ON `
4HE SIGNIlCANCE OF ![! AND #( IS BORNE OUT BY THE FOLLOWING THEO
REM
6ISIONS IN -ATH 3%#/.$ ,!7 /& 4(%2-/$9.!-)#3
4HEOREM %QUIVALENCE OF ENTROPY AND ! ! GIVEN #( 4HE FOLLOWING

ARE EQUIVALENT FOR A STATE SPACE `
I 4HE RELATION } BETWEEN STATES IN POSSIBLY DIkERENT MULTIPLE SCALED
COPIES OF ` EG `t a `t a ` ` ` a `t. IS CHARACTERIZED BY AN
ENTROPY FUNCTION 3 ON ` IN THE SENSE THAT
t 8 t 8 } t 8 t 8
IS EQUIVALENT TO THE CONDITION THAT
8 8
tI 38I s tJ 38J
I J
WHENEVER 8 8
tI tJ
I J
II 4HE RELATION } SATISlES CONDITIONS ! [! AND #( HOLDS FOR
EVERY MULTIPLE SCALED COPY OF `
4HIS ENTROPY FUNCTION ON ` IS UNIQUE UP TO AnNE EQUIVALENCE IE
38 A38 " WITH A
4HAT I II IS OBVIOUS 4HE PROOF OF II I IS CARRIED OUT BY AN
EXPLICIT CONSTRUCTION OF THE ENTROPY FUNCTION ON ` [ REMINISCENT OF AN OLD
DElNITION OF HEAT BY ,APLACE AND ,AVOISIER IN TERMS OF THE AMOUNT OF ICE
THAT A BODY CAN MELT
"ASIC CONSTRUCTION OF 3 0ICK TWO REFERENCE POINTS 8 AND 8 IN ` WITH
8 }} 8 )F SUCH POINTS DO NOT EXIST THEN 3 IS THE CONSTANT FUNCTION
4HEN DElNE FOR 8 `
g h
38 SUP t _ t 8 t8 } 8
2EMARKS !S IN AXIOM ! TWO STATE SPACES ARE INVOLVED IN "Y AXIOM
!
! 8 w _ t 8 t8 AND HENCE BY #( IN THE SPACE `_t a `t 8 IS
COMPARABLE TO _ t 8 t8 )N WE ALLOW t s AND t t BY USING
THE CONVENTION THAT 8 _9 } : MEANS THAT 8 } 9 : AND 8 9 8
&OR WE ONLY NEED TO KNOW THAT #( HOLDS IN TWO FOLD SCALED PRODUCTS OF
` WITH ITSELF #( WILL THEN AUTOMATICALLY BE TRUE FOR ALL PRODUCTS )N
THE REFERENCE POINTS 8 8 ARE lXED AND THE SUPREMUM IS OVER t /NE CAN
ASK HOW 3 CHANGES IF WE CHANGE THE TWO POINTS 8 8 4HE ANSWER IS THAT
THE CHANGE IS AnNE IE 38 A38 " WITH A
4HEOREM EXTENDS TO PRODUCTS OF MULTIPLE SCALED COPIES OF DIkERENT
SYSTEMS IE TO GENERAL COMPOUND SYSTEMS 4HIS EXTENSION IS AN IMME
DIATE CONSEQUENCE OF THE FOLLOWING THEOREM WHICH IS PROVED BY APPLYING
%( ,)%" !.$ * 9.'6!3/. '!&!
4HEOREM TO THE PRODUCT OF THE SYSTEM UNDER CONSIDERATION WITH SOME

STANDARD REFERENCE SYSTEM
4HEOREM #ONSISTENT ENTROPY SCALES !SSUME THAT #( HOLDS FOR ALL
COMPOUND SYSTEMS &OR EACH SYSTEM ` LET 3` BE SOME DElNITE ENTROPY
FUNCTION ON ` IN THE SENSE OF 4HEOREM 4HEN THERE ARE CONSTANTS A` AND
"` SUCH THAT THE FUNCTION 3 DElNED FOR ALL STATES OF ALL SYSTEMS BY
38 A` 3` 8 "`
FOR 8 ` SATISlES ADDITIVITY EXTENSIVITY AND MONOTONICITY IN
THE SENSE THAT WHENEVER 8 AND 9 ARE IN THE SAME STATE SPACE THEN
8}9 IF AND ONLY IF 38 s 39
4HEOREM IS WHAT WE NEED EXCEPT FOR THE QUESTION OF MIXING AND
CHEMICAL REACTIONS WHICH IS TREATED AT THE END AND WHICH CAN BE PUT ASIDE
AT A lRST READING )N OTHER WORDS AS LONG AS WE DO NOT CONSIDER ADIABATIC
PROCESSES IN WHICH SYSTEMS ARE CONVERTED INTO EACH OTHER EG A COMPOUND
SYSTEM CONSISTING OF VESSEL OF HYDROGEN AND A VESSEL OF OXYGEN IS CONVERTED
INTO A VESSEL OF WATER THE ENTROPY PRINCIPLE HAS BEEN VERIlED )F THAT IS
SO WHAT REMAINS TO BE DONE THE READER MAY JUSTIlABLY ASK 4HE ANSWER
IS TWOFOLD &IRST 4HEOREM REQUIRES THAT #( HOLDS FOR ALL SYSTEMS AND
WE ARE NOT CONTENT TO TAKE THIS AS AN AXIOM 3ECOND IMPORTANT NOTIONS OF
THERMODYNAMICS SUCH AS @THERMAL EQUILIBRIUM WHICH WILL EVENTUALLY LEAD
TO A PRECISE DElNITION OF @TEMPERATURE HAVE NOT APPEARED SO FAR 7E
SHALL SEE THAT THESE TWO POINTS IE THERMAL EQUILIBRIUM AND #( ARE NOT
UNRELATED
!S FOR #( OTHER AUTHORS ;'I= ;"U= ;#O= AND ;2,= ESSENTIALLY POSTULATE
THAT IT HOLDS FOR ALL SYSTEMS BY MAKING IT AXIOMATIC THAT COMPARABLE STATES
FALL INTO EQUIVALENCE CLASSES 4HIS MEANS THAT THE CONDITIONS 8 } : AND
9 } : ALWAYS IMPLY THAT 8 AND 9 ARE COMPARABLE LIKEWISE THEY MUST
BE COMPARABLE IF : } 8 AND : } 9 "Y IDENTIFYING A @STATE SPACE WITH
AN EQUIVALENCE CLASS THE COMPARISON HYPOTHESIS THEN HOLDS IN THESE OTHER
APPROACHES BY ASSUMPTION FOR ALL STATE SPACES 7E IN CONTRAST WOULD LIKE
TO DERIVE #( FROM SOMETHING THAT WE CONSIDER MORE BASIC 4WO INGREDIENTS
WILL BE NEEDED 4HE ANALYSIS OF CERTAIN SPECIAL BUT COMMONPLACE SYSTEMS
CALLED @SIMPLE SYSTEMS AND SOME ASSUMPTIONS ABOUT THERMAL CONTACT THE
@:EROTH LAW THAT WILL ACT AS A KIND OF GLUE HOLDING THE PARTS OF A COMPOUND
SYSTEMS IN HARMONY WITH EACH OTHER
! 3IMPLE 3YSTEM IS ONE WHOSE STATE SPACE CAN BE IDENTIlED WITH SOME
OPEN CONVEX SUBSET OF SOME 2N WITH A DISTINGUISHED COORDINATE DENOTED
6ISIONS IN -ATH 3%#/.$ ,!7 /& 4(%2-/$9.!-)#3
BY 5 CALLED THE ENERGY AND ADDITIONAL COORDINATES 6 2N CALLED WORK

COORDINATES 4HE ENERGY COORDINATE IS THE WAY IN WHICH THERMODYNAMICS
MAKES CONTACT WITH MECHANICS WHERE THE CONCEPT OF ENERGY ARISES AND IS
PRECISELY DElNED 4HE FACT THAT THE AMOUNT OF ENERGY IN A STATE IS INDEPEN
DENT OF THE MANNER IN WHICH THE STATE WAS ARRIVED AT IS IN REALITY THE lRST
LAW OF THERMODYNAMICS ! TYPICAL AND OFTEN THE ONLY WORK COORDINATE
IS THE VOLUME OF A mUID OR GAS CONTROLLED BY A PISTON OTHER EXAMPLES
ARE DEFORMATION COORDINATES OF A SOLID OR MAGNETIZATION OF A PARAMAGNETIC
SUBSTANCE
/UR GOAL IS TO SHOW WITH THE ADDITION OF A FEW MORE AXIOMS THAT #(
HOLDS FOR SIMPLE SYSTEMS AND THEIR SCALED PRODUCTS )N THE PROCESS WE
WILL INTRODUCE MORE STRUCTURE WHICH WILL CAPTURE THE INTUITIVE NOTIONS OF
THERMODYNAMICS THERMAL EQUILIBRIUM IS ONE
&IRST THERE IS AN AXIOM ABOUT CONVEXITY
! #ONVEX COMBINATION )F 8 AND 9 ARE STATES OF A SIMPLE SYSTEM
AND T ; = THEN
_ `
T8 _ T 9 } T8 _ T 9
IN THE SENSE OF ORDINARY CONVEX ADDITION OF POINTS IN 2N ! STRAIGHT
FORWARD CONSEQUENCE OF THIS AXIOM AND ! IS THAT THE FORWARD SECTORS
!8 F9 ` 8 } 9 G
OF STATES 8 IN A SIMPLE SYSTEM ` ARE CONVEX SETS
!NOTHER CONSEQUENCE IS A CONNECTION BETWEEN THE EXISTENCE OF IRRE
VERSIBLE PROCESSES AND #ARATHsEODORYS PRINCIPLE ;#= ;"= MENTIONED ABOVE
,EMMA !SSUME ! [! FOR ` y 2. AND CONSIDER THE FOLLOWING
STATEMENTS
A %XISTENCE OF IRREVERSIBLE PROCESSES &OR EVERY 8 ` THERE IS A 9 `
WITH 8 }} 9
B #ARATHsEODORYS PRINCIPLE )N EVERY NEIGHBORHOOD OF EVERY 8 `
THERE IS A : ` WITH 8 } :
4HEN A B ALWAYS )F THE FORWARD SECTORS IN ` HAVE INTERIOR POINTS
THEN B A
7E NEED THREE MORE AXIOMS FOR SIMPLE SYSTEMS WHICH WILL TAKE US INTO
AN ANALYTIC DETOUR 4HE lRST OF THESE ESTABLISHES A ABOVE
! )RREVERSIBILITY &OR EACH 8 ` THERE IS A POINT 9 ` SUCH THAT
8 }} 9 4HIS AXIOM IS IMPLIED BY ! BELOW BUT IS STATED HERE
SEPARATELY BECAUSE IMPORTANT CONCLUSIONS CAN BE DRAWN FROM IT ALONE
%( ,)%" !.$ * 9.'6!3/. '!&!
! ,IPSCHITZ TANGENT PLANES &OR EACH 8 ` THE FORWARD SECTOR

!8 F9 ` 8 } 9 G HAS A UNIQUE SUPPORT PLANE AT 8 IE
!8 HAS A TANGENT PLANE AT 8 4HE TANGENT PLANE IS ASSUMED TO BE
A LOCALLY ,IPSCHITZ CONTINUOUS FUNCTION OF 8 IN THE SENSE EXPLAINED
BELOW
! #ONNECTEDNESS OF THE BOUNDARY 4HE BOUNDARY !8 RELATIVE TO
THE OPEN SET ` OF EVERY FORWARD SECTOR !8 y ` IS CONNECTED 4HIS
IS TECHNICAL AND CONCEIVABLY CAN BE REPLACED BY SOMETHING ELSE
!XIOM ! PLUS ,EMMA ASSERTS THAT EVERY 8 LIES ON THE BOUNDARY
!8 OF ITS FORWARD SECTOR !LTHOUGH AXIOM ! ASSERTS THAT THE CONVEX SET
!8 HAS A TRUE TANGENT AT 8 ONLY IT IS AN EASY CONSEQUENCE OF AXIOM !
THAT !8 HAS A TRUE TANGENT EVERYWHERE ON ITS BOUNDARY 4O SAY THAT THIS
TANGENT PLANE IS LOCALLY ,IPSCHITZ CONTINUOUS MEANS THAT IF 8 5 6
THEN THIS PLANE IS GIVEN BY
8N
5 _ 5 0I 8 6I _ 6I

WITH LOCALLY ,IPSCHITZ CONTINUOUS FUNCTIONS 0I 4HE FUNCTION 0I IS CALLED
THE GENERALIZED PRESSURE CONJUGATE TO THE WORK COORDINATE 6I 7HEN 6I IS
THE VOLUME 0I IS THE ORDINARY PRESSURE
,IPSCHITZ CONTINUITY AND CONNECTEDNESS IS A WELL KNOWN GUARANTEE THAT
THE COUPLED DIkERENTIAL EQUATIONS
5
_ `
6J 6 _0J 5 6 6 FOR J N
NOT ONLY HAVE A SOLUTION SINCE WE KNOW THAT THE SURFACE !8 EXISTS BUT
THIS SOLUTION MUST BE UNIQUE 4HUS IF 9 !8 THEN 8 !9 )N SHORT
THE SURFACES !8 FOLIATE THE STATE SPACE ` 7HAT IS LESS OBVIOUS BUT VERY
IMPORTANT BECAUSE IT INSTANTLY GIVES US THE COMPARISON HYPOTHESIS FOR ` IS
THE FOLLOWING
4HEOREM &ORWARD SECTORS ARE NESTED )F !8 AND !9 ARE TWO FORWARD
SECTORS IN THE STATE SPACE ` OF A SIMPLE SYSTEM THEN EXACTLY ONE OF THE
FOLLOWING HOLDS
!
A !8 !9 IE 8 w 9
B !8 y )NTERIOR!9 IE 9 }} 8
C !9 y )NTERIOR!8 IE 8 }} 9
)T CAN ALSO BE SHOWN FROM OUR AXIOMS THAT THE ORIENTATION OF FORWARD
SECTORS WRT THE ENERGY AXIS IS THE SAME FOR ALL SIMPLE SYSTEMS "Y CON
VENTION WE CHOOSE THE DIRECTION OF THE ENERGY AXIS SO THAT THE THE ENERGY
6ISIONS IN -ATH 3%#/.$ ,!7 /& 4(%2-/$9.!-)#3
ALWAYS INCREASES IN ADIABATIC PROCESSES AT lXED WORK COORDINATES 7HEN

TEMPERATURE IS DElNED LATER THIS WILL IMPLY THAT TEMPERATURE IS ALWAYS
POSITIVE
4HEOREM IMPLIES THAT 9 IS ON THE BOUNDARY OF !8 IF AND ONLY IF 8 IS
!
ON THE BOUNDARY OF !9 4HUS THE ADIABATS IE THE w EQUIVALENCE CLASSES
CONSIST OF THESE BOUNDARIES
"EFORE LEAVING THE SUBJECT OF SIMPLE SYSTEMS LET US REMARK ON THE CON
NECTION WITH #ARATHsEODORYS DEVELOPMENT 4HE POINT OF CONTACT IS THE FACT
THAT 8 !8 7E ASSUME THAT !8 IS CONVEX AND USE TRANSITIVITY AND ,IP
SCHITZ CONTINUITY TO ARRIVE EVENTUALLY AT 4HEOREM #ARATHsEODORY USES
&ROBENIUSS THEOREM PLUS ASSUMPTIONS ABOUT DIkERENTIABILITY TO CONCLUDE
THE EXISTENCE [ LOCALLY [ OF A SURFACE CONTAINING 8 )MPORTANT GLOBAL INFOR
MATION SUCH AS 4HEOREM ARE THEN NOT EASY TO OBTAIN WITHOUT FURTHER
ASSUMPTIONS AS DISCUSSED EG IN ;"=
4HE NEXT TOPIC IS THERMAL CONTACT AND THE ZEROTH LAW WHICH ENTAILS
THE VERY SPECIAL ASSUMPTIONS ABOUT } THAT WE MENTIONED EARLIER )T WILL
ENABLE US TO ESTABLISH #( FOR PRODUCTS OF SEVERAL SYSTEMS AND THEREBY
SHOW VIA 4HEOREM THAT ENTROPY EXISTS AND IS ADDITIVE !LTHOUGH WE
HAVE ESTABLISHED #( FOR A SIMPLE SYSTEM ` WE HAVE NOT YET ESTABLISHED
#( EVEN FOR A PRODUCT OF TWO COPIES OF ` 4HIS IS NEEDED IN THE DElNITION
OF 3 GIVEN IN 4HE 3 IN IS DETERMINED UP TO AN AnNE SHIFT AND WE
WANT TO BE ABLE TO CALIBRATE THE ENTROPIES IE ADJUST THE MULTIPLICATIVE
AND ADDITIVE CONSTANTS OF ALL SYSTEMS SO THAT THEY WORK TOGETHER TO FORM A
GLOBAL 3 SATISFYING THE ENTROPY PRINCIPLE 7E NEED lVE MORE AXIOMS 4HEY
MIGHT LOOK A BIT ABSTRACT SO A FEW WORDS OF INTRODUCTION MIGHT BE HELPFUL
)N ORDER TO RELATE SYSTEMS TO EACH OTHER IN THE HOPE OF ESTABLISHING
#( FOR COMPOUNDS AND THEREBY AN ADDITIVE ENTROPY FUNCTION SOME WAY
MUST BE FOUND TO PUT THEM INTO CONTACT WITH EACH OTHER (EURISTICALLY WE
IMAGINE TWO SIMPLE SYSTEMS THE SAME OR DIkERENT SIDE BY SIDE AND lX THE
WORK COORDINATES EG THE VOLUME OF EACH #ONNECT THEM WITH A <COPPER
THREAD AND WAIT FOR EQUILIBRIUM TO BE ESTABLISHED 4HE TOTAL ENERGY 5 WILL
NOT CHANGE BUT THE INDIVIDUAL ENERGIES 5 AND 5 WILL ADJUST TO VALUES THAT
DEPEND ON 5 AND THE WORK COORDINATES 4HIS NEW SYSTEM WITH THE THREAD
PERMANENTLY CONNECTED THEN BEHAVES LIKE A SIMPLE SYSTEM WITH ONE ENERGY
COORDINATE BUT WITH SEVERAL WORK COORDINATES THE UNION OF THE TWO WORK
COORDINATES 4HUS IF WE START INITIALLY WITH 8 5 6 FOR SYSTEM
AND 8 5 6 FOR SYSTEM AND IF WE END UP WITH 8 5 6 6
FOR THE NEW SYSTEM WE CAN SAY THAT 8 8 } 8 4HIS HOLDS FOR EVERY
%( ,)%" !.$ * 9.'6!3/. '!&!
CHOICE OF 5 AND 5 WHOSE SUM IS 5 -OREOVER AFTER THERMAL EQUILIBRIUM

IS REACHED THE TWO SYSTEMS CAN BE DISCONNECTED IF WE WISH AND ONCE MORE
FORM A COMPOUND SYSTEM WHOSE COMPONENT PARTS WE SAY ARE IN THERMAL
EQUILIBRIUM 4HAT THIS IS TRANSITIVE IS THE ZEROTH LAW
4HUS WE CANNOT ONLY MAKE COMPOUND SYSTEMS CONSISTING OF INDEPEN
DENT SUBSYSTEMS WHICH CAN INTERACT BUT SEPARATE AGAIN WE CAN ALSO MAKE
A NEW SIMPLE SYSTEM OUT OF TWO SIMPLE SYSTEMS 4O DO THIS AN ENERGY CO
ORDINATE HAS TO DISAPPEAR AND THERMAL CONTACT DOES THIS FOR US !LL OF THIS
IS FORMALIZED IN THE FOLLOWING THREE AXIOMS
! 4HERMAL CONTACT &OR ANY TWO SIMPLE SYSTEMS WITH STATE SPACES `
AND ` THERE IS ANOTHER SIMPLE SYSTEM CALLED THE THERMAL JOIN OF `
AND ` WITH STATE SPACE
g h
a 5 6 6 5 5 5 WITH 5 6 ` 5 6 `

-OREOVER
_ `
` a ` 5 6 5 6 } 5 5 6 6 a
! 4HERMAL SPLITTING &OR ANY POINT 5 6 6 a THERE IS AT LEAST
ONE PAIR OF STATES 5 6 ` 5 6 ` WITH 5 5 5
SUCH THAT
!
5 6 6 w 5 6 5 6
!
)F 5 6 6 w 5 6 5 6 WE SAY THAT THE STATES 8 5 6
AND 9 5 6 ARE IN THERMAL EQUILIBRIUM AND WRITE
4
8w9
4 4
! :EROTH LAW OF THERMODYNAMICS )F 8 w 9 AND IF 9 w : THEN
4
8 w :
! AND ! TOGETHER SAY THAT FOR EACH CHOICE OF THE INDIVIDUAL WORK
COORDINATES THERE IS A WAY TO DIVIDE UP THE ENERGY 5 BETWEEN THE TWO
SYSTEMS IN A STABLE MANNER ! IS THE STABILITY STATEMENT FOR IT SAYS THAT
JOINING IS REVERSIBLE IE ONCE THE EQUILIBRIUM HAS BEEN ESTABLISHED ONE CAN
CUT THE COPPER THREAD AND RETRIEVE THE TWO SYSTEMS BACK AGAIN BUT WITH A
SPECIAL PARTITION OF THE ENERGIES
4HIS REVERSIBILITY ALLOWS US TO THINK OF THE THERMAL JOIN WHICH IS A SIMPLE
SYSTEM IN ITS OWN RIGHT AS A SPECIAL SUBSET OF THE PRODUCT SYSTEM ` a `
WHICH WE CALL THE THERMAL DIAGONAL )N PARTICULAR ! ALLOWS US TO PROVE
4
EASILY THAT 8 w t8 FOR ALL 8 AND ALL t
! IS THE FAMOUS ZEROTH LAW WHICH SAYS THAT THE THERMAL EQUILIBRIUM
IS TRANSITIVE AND HENCE AN EQUIVALENCE RELATION /FTEN THIS LAW IS TAKEN TO
6ISIONS IN -ATH 3%#/.$ ,!7 /& 4(%2-/$9.!-)#3
MEAN THAT THERE THE EQUIVALENCE CLASSES CAN BE LABELED BY AN @EMPIRICAL

TEMPERATURE BUT WE DO NOT WANT TO MENTION TEMPERATURE AT ALL AT THIS
POINT )T WILL APPEAR LATER
4WO MORE AXIOMS ARE NEEDED
!
! REQUIRES THAT FOR EVERY ADIABAT IE AN EQUIVALENCE CLASS WRT w
4
THERE EXISTS AT LEAST ONE ISOTHERM IE AN EQUIVALENCE CLASS WRT w CON
TAINING POINTS ON BOTH SIDES OF THE ADIABAT .OTE THAT FOR EACH GIVEN 8
ONLY TWO POINTS IN THE ENTIRE STATE SPACE ` ARE REQUIRED TO HAVE THE STATED
PROPERTY 4HIS ASSUMPTION ESSENTIALLY PREVENTS A STATE SPACE FROM BREAKING
UP INTO TWO PIECES THAT DO NOT COMMUNICATE WITH EACH OTHER 7ITHOUT IT
COUNTEREXAMPLES TO #( FOR COMPOUND SYSTEMS CAN BE CONSTRUCTED ! IM
PLIES ! BUT WE LISTED ! SEPARATELY IN ORDER NOT TO CONFUSE THE DISCUSSION
OF SIMPLE SYSTEMS WITH THERMAL EQUILIBRIUM
! IS A TECHNICAL AND PERHAPS CAN BE ELIMINATED )TS PHYSICAL MOTI
VATION IS THAT A SUnCIENTLY LARGE COPY OF A SYSTEM CAN ACT AS A HEAT BATH
FOR OTHER SYSTEMS 7HEN TEMPERATURE IS INTRODUCED LATER ! WILL HAVE
THE MEANING THAT ALL SYSTEMS HAVE THE SAME TEMPERATURE RANGE 4HIS POS
TULATE IS NEEDED IF WE WANT TO BE ABLE TO BRING EVERY SYSTEM INTO THERMAL
EQUILIBRIUM WITH EVERY OTHER SYSTEM
! 4RANSVERSALITY )F ` IS THE STATE SPACE OF A SIMPLE SYSTEM AND IF
4
8 ` THEN THERE EXIST STATES 8 w 8 WITH 8 }} 8 }} 8
! 5NIVERSAL TEMPERATURE RANGE )F ` AND ` ARE STATE SPACES OF
SIMPLE SYSTEMS THEN FOR EVERY 8 ` AND EVERY 6 BELONGING TO THE
PROJECTION OF ` ONTO THE SPACE OF ITS WORK COORDINATES THERE IS A
4
9 ` WITH WORK COORDINATES 6 SUCH THAT 8 w 9
4HE READER SHOULD NOTE THAT THE CONCEPT @THERMAL CONTACT HAS APPEARED
BUT NOT TEMPERATURE OR HOT AND COLD OR ANYTHING RESEMBLING THE #LAUSIUS
OR +ELVIN 0LANCK FORMULATIONS OF THE SECOND LAW .EVERTHELESS WE COME TO
THE MAIN ACHIEVEMENT OF OUR APPROACH 7ITH THESE AXIOMS WE CAN ESTABLISH
#( FOR PRODUCTS OF SIMPLE SYSTEMS EACH OF WHICH SATISlES #( AS WE ALREADY
KNOW &IRST THE THERMAL JOIN ESTABLISHES #( FOR THE SCALED PRODUCT OF
A SIMPLE SYSTEM WITH ITSELF 4HE BASIC IDEA HERE IS THAT THE POINTS IN THE
PRODUCT THAT LIE ON THE THERMAL DIAGONAL ARE COMPARABLE SINCE POINTS IN
A SIMPLE SYSTEM ARE COMPARABLE )N PARTICULAR WITH 8 8 8 AS IN !
THE STATES _ t 8 t8 AND _ t 8 t8 CAN BE REGARDED AS STATES
OF THE SAME SIMPLE SYSTEM AND ARE THEREFORE COMPARABLE 4HIS IS THE KEY
POINT NEEDED FOR THE CONSTRUCTION OF 3 ACCORDING TO 4HE IMPORTANCE
OF TRANSVERSALITY IS THUS BROUGHT INTO FOCUS
%( ,)%" !.$ * 9.'6!3/. '!&!
7ITH SOME MORE WORK WE CAN ESTABLISH #( FOR MULTIPLE SCALED COPIES
OF A SIMPLE SYSTEM 4HUS WE HAVE ESTABLISHED 3 WITHIN THE CONTEXT OF ONE
SYSTEM AND COPIES OF THE SYSTEM IE CONDITION II OF 4HEOREM !S LONG
AS WE STAY WITHIN SUCH A GROUP OF SYSTEMS THERE IS NO WAY TO DETERMINE THE
UNKNOWN MULTIPLICATIVE OR ADDITIVE ENTROPY CONSTANTS 4HE NEXT TASK IS TO
SHOW THAT THE MULTIPLICATIVE CONSTANTS CAN BE ADJUSTED TO GIVE A UNIVERSAL
ENTROPY VALID FOR COPIES OF DIkERENT SYSTEMS IE TO ESTABLISH THE HYPOTHESIS
OF 4HEOREM 4HIS IS BASED ON THE FOLLOWING
,EMMA %XISTENCE OF CALIBRATORS )F ` AND ` ARE SIMPLE SYSTEMS THEN
THERE EXIST STATES 8 8 ` AND 9 9 ` SUCH THAT
8 }} 8 AND 9 }} 9
AND
!
8 9 w 8 9
4HE SIGNIlCANCE OF ,EMMA IS THAT IT ALLOWS US TO lX THE MULTIPLICATIVE
CONSTANTS BY THE CONDITION
3 8 3 9 3 8 3 9
4HE PROOF OF ,EMMA IS COMPLICATED AND REALLY USES ALL THE AXIOMS !
TO ! 7ITH ITS AID WE ARRIVE AT OUR CHIEF GOAL WHICH IS #( FOR COMPOUND
SYSTEMS
4HEOREM %NTROPY PRINCIPLE IN PRODUCTS OF SIMPLE SYSTEMS 4HE COM
PARISON HYPOTHESIS #( IS VALID IN ARBITRARY SCALED PRODUCTS OF SIMPLE SYS
TEMS (ENCE BY 4HEOREM THE RELATION } AMONG STATES IN SUCH STATE
SPACES IS CHARACTERIZED BY AN ENTROPY FUNCTION 3 4HE ENTROPY FUNCTION IS
UNIQUE UP TO AN OVERALL MULTIPLICATIVE CONSTANT AND ONE ADDITIVE CONSTANT
FOR EACH SIMPLE SYSTEM UNDER CONSIDERATION
!T LAST WE ARE NOW READY TO DElNE TEMPERATURE #ONCAVITY OF 3 IM
PLIED BY ! ,IPSCHITZ CONTINUITY OF THE PRESSURE AND THE TRANSVERSALITY
CONDITION TOGETHER WITH SOME REAL ANALYSIS PLAY KEY ROLES IN THE FOLLOWING
WHICH ANSWERS QUESTIONS 1 AND 1 POSED AT THE BEGINNING
4HEOREM %NTROPY DElNES TEMPERATURE 4HE ENTROPY 3 IS A CON
CAVE AND CONTINUOUSLY DIkERENTIABLE FUNCTION ON THE STATE SPACE OF A SIMPLE
SYSTEM )F THE FUNCTION 4 IS DElNED BY

_ 3`
4 5 6
4 4
THEN 4 AND 4 CHARACTERIZES THE RELATION w IN THE SENSE THAT 8 w 9
IF AND ONLY IF 4 8 4 9 -OREOVER IF TWO SYSTEMS ARE BROUGHT INTO
6ISIONS IN -ATH 3%#/.$ ,!7 /& 4(%2-/$9.!-)#3
THERMAL CONTACT WITH lXED WORK COORDINATES THEN SINCE THE TOTAL ENTROPY
CANNOT DECREASE THE ENERGY mOWS FROM THE SYSTEM WITH THE HIGHER 4 TO THE
SYSTEM WITH THE LOWER 4
4HE TEMPERATURE NEED NOT BE A STRICTLY MONOTONE FUNCTION OF 5 INDEED
IT IS NOT SO IN A @MULTIPHASE REGION )T FOLLOWS THAT 4 IS NOT ALWAYS CAPA
BLE OF SPECIFYING A STATE AND THIS FACT CAN CAUSE SOME PAIN IN TRADITIONAL
DISCUSSIONS OF THE SECOND LAW [ IF IT IS RECOGNIZED WHICH USUALLY IT IS NOT
-IXING AND CHEMICAL REACTIONS 4HE CORE RESULTS OF OUR ANALYSIS HAVE
NOW BEEN PRESENTED AND READERS SATISlED WITH THE ENTROPY PRINCIPLE IN THE
FORM OF 4HEOREM MAY WISH TO STOP AT THIS POINT .EVERTHELESS A NAGGING
DOUBT WILL OCCUR TO SOME BECAUSE THERE ARE IMPORTANT ADIABATIC PROCESSES IN
WHICH SYSTEMS ARE NOT CONSERVED AND THESE PROCESSES ARE NOT YET COVERED
IN THE THEORY ! CRITICAL STUDY OF THE USUAL TEXBOOK TREATMENTS SHOULD
CONVINCE THE READER THAT THIS SUBJECT IS NOT EASY BUT IN VIEW OF THE MANIFOLD
APPLICATIONS OF THERMODYNAMICS TO CHEMISTRY AND BIOLOGY IT IS IMPORTANT
TO TELL THE WHOLE STORY AND NOT IGNORE SUCH PROCESSES
/NE CAN FORMULATE THE PROBLEM AS THE DETERMINATION OF THE ADDITIVE
CONSTANTS "` OF 4HEOREM !S LONG AS WE CONSIDER ONLY ADIABATIC PRO
CESSES THAT PRESERVE THE AMOUNT OF EACH SIMPLE SYSTEM IE SUCH THAT %QS
AND HOLD THESE CONSTANTS ARE INDETERMINATE 4HIS IS NO LONGER
THE CASE HOWEVER IF WE CONSIDER MIXING PROCESSES AND CHEMICAL REACTIONS
WHICH ARE NOT REALLY DIkERENT AS FAR AS THERMODYNAMICS IS CONCERNED )T
THEN BECOMES A NONTRIVIAL QUESTION WHETHER THE ADDITIVE CONSTANTS CAN BE
CHOSEN IN SUCH A WAY THAT THE ENTROPY PRINCIPLE HOLDS /DDLY THIS DETER
MINATION TURNS OUT TO BE FAR MORE COMPLEX MATHEMATICALLY AND PHYSICALLY
THAN THE DETERMINATION OF THE MULTIPLICATIVE CONSTANTS 4HEOREM )N TRA
DITIONAL TREATMENTS ONE USUALLY RESORTS TO GEDANKEN EXPERIMENTS INVOLVING
STRANGE NONEXISTENT OBJECTS CALLED @SEMIPERMEABLE MEMBRANES AND @VAN
T(OkT BOXES 7E PRESENT HERE A GENERAL AND RIGOROUS APPROACH WHICH
AVOIDS ALL THIS
7HAT WE ALREADY KNOW IS THAT EVERY SYSTEM HAS A WELL DElNED ENTROPY
FUNCTION EG FOR EACH ` THERE IS 3` AND WE KNOW FROM 4HEOREM THAT
THE MULTIPLICATIVE CONSTANTS A` CAN BEEN DETERMINED IN SUCH A WAY THAT THE
SUM OF THE ENTROPIES INCREASES IN ANY ADIABATIC PROCESS IN ANY COMPOUND
SPACE ` a ` a 4HUS IF 8I Ì AND 9I Ì THEN
0 0
8 8 } 9 9 IF AND ONLY IF I 3I 8I s J 3J 9J
WHERE WE HAVE DENOTED 3Ì BY 3I FOR SHORT 4HE ADDITIVE ENTROPY CONSTANTS
%( ,)%" !.$ * 9.'6!3/. '!&!
DO NOT MATTER HERE SINCE EACH FUNCTION 3I APPEARS ON BOTH SIDES OF THIS
INEQUALITY )T IS IMPORTANT TO NOTE THAT THIS APPLIES EVEN TO PROCESSES THAT
IN INTERMEDIATE STEPS TAKE ONE SYSTEM INTO ANOTHER PROVIDED THE TOTAL
COMPOUND SYSTEM IS THE SAME AT THE BEGINNING AND AT THE END OF THE PROCESS
4HE TASK IS TO lND CONSTANTS "` ONE FOR EACH STATE SPACE ` IN SUCH
A WAY THAT THE ENTROPY DElNED BY
38 3` 8 "` FOR 8`
SATISlES
38 s 39
WHENEVER
8}9 WITH 8 ` 9 `
!DDITIONALLY WE REQUIRE THAT THE NEWLY DElNED ENTROPY SATISlES SCALING AND
ADDITIVITY UNDER COMPOSITION 3INCE THE INITIAL ENTROPIES 3` 8 ALREADY SAT
ISFY THEM THESE REQUIREMENTS BECOME CONDITIONS ON THE ADDITIVE CONSTANTS
"`
t t
"` a ` t "` t "`
FOR ALL STATE SPACES ` ` UNDER CONSIDERATIONS AND t t 3OME RE
mECTION SHOWS US THAT CONSISTENCY IN THE DElNITION OF THE ENTROPY CONSTANTS
"` REQUIRES US TO CONSIDER ALL POSSIBLE CHAINS OF ADIABATIC PROCESSES LEAD
ING FROM ONE SPACE TO ANOTHER VIA INTERMEDIATE STEPS -OREOVER THE AD
DITIVITY REQUIREMENT LEADS US TO ALLOW THE USE OF A @CATALYST IN THESE PRO
CESSES IE AN AUXILIARY SYSTEM THAT IS RECOVERED AT THE END ALTHOUGH A
STATE CHANGE WITHIN THIS SYSTEM MIGHT TAKE PLACE 7ITH THIS IN MIND WE
DElNE QUANTITIES & ` ` THAT INCORPORATE THE ENTROPY DIkERENCES IN ALL
SUCH CHAINS LEADING FROM ` TO ` 4HESE ARE BUILT UP FROM SIMPLER QUANTI
TIES $` ` WHICH MEASURE THE ENTROPY DIkERENCES IN ONE STEP PROCESSES
AND %` ` WHERE THE @CATALYST IS ABSENT 4HE PRECISE DElNITIONS ARE AS
FOLLOWS &IRST
g h
$` ` INF 3` 9 _ 3` 8 8 ` 9 ` 8 } 9
)F THERE IS NO ADIABATIC PROCESS LEADING FROM ` TO ` WE PUT $` `
.EXT FOR ANY GIVEN ` AND ` WE CONSIDER ALL lNITE CHAINS OF STATE SPACES
` ` ` `. ` SUCH THAT $Ì Ì FOR ALL I AND WE DElNE
g h
%` ` INF $` ` ` ` ` $`. _ `.
WHERE THE INlMUM IS TAKEN OVER ALL SUCH CHAINS LINKING ` WITH ` &INALLY
WE DElNE g h
& ` ` INF %` a ` ` a `
6ISIONS IN -ATH 3%#/.$ ,!7 /& 4(%2-/$9.!-)#3
WHERE THE INlMUM IS TAKEN OVER ALL STATE SPACES ` 4HESE ARE THE @CATA
LYSTS
4HE IMPORTANCE OF THE & S FOR THE DETERMINATION OF THE ADDITIVE CON
STANTS IS MADE CLEAR IN THE FOLLOWING THEOREM
4HEOREM #ONSTANT ENTROPY DIkERENCES )F ` AND ` ARE TWO STATE
SPACES THEN FOR ANY TWO STATES 8 ` AND 9 `
8}9 IF AND ONLY IF 3` 8 & ` ` s 3` 9
!N ESSENTIAL INGREDIENT FOR THE PROOF OF THIS THEOREM IS %Q
!CCORDING TO 4HEOREM THE DETERMINATION OF THE ENTROPY CONSTANTS
"` AMOUNTS TO SATISFYING THE INEQUALITIES
_& ` ` s "` _ "` s & ` `
TOGETHER WITH THE LINEARITY CONDITION )T IS CLEAR THAT CAN ONLY
BE SATISlED WITH lNITE CONSTANTS "` AND "` IF & ` ` _ 4O
EXCLUDE THE PATHOLOGICAL CASE & ` ` _ WE INTRODUCE OUR LAST AXIOM
! WHOSE STATEMENT REQUIRES THE FOLLOWING DElNITION
$EFINITION ! STATE SPACE ` IS SAID TO BE CONNECTED TO ANOTHER STATE
SPACE ` IF THERE ARE STATES 8 ` AND 9 ` AND STATE SPACES ` `.
WITH STATES 8I 9I Ì I . AND A STATE SPACE ` WITH STATES
8 9 ` SUCH THAT
8 8 } 9 8I } 9I I . _ 8. } 9 9
! !BSENCE OF SINKS )F ` IS CONNECTED TO ` THEN ` IS CONNECTED
TO `
4HIS AXIOM EXCLUDES & ` ` _ BECAUSE ON GENERAL GROUNDS ONE
ALWAYS HAS
_& ` ` s & ` `

(ENCE & ` ` _ WHICH MEANS IN PARTICULAR THAT ` IS CONNECTED
TO ` WOULD IMPLY & ` ` IE THAT THERE IS NO WAY BACK FROM `
TO ` 4HIS IS EXCLUDED BY !XIOM
4HE QUANTITIES & ` ` HAVE SIMPLE SUBADDITIVITY PROPERTIES THAT ALLOW
US TO USE THE (AHN "ANACH THEOREM TO SATISFY THE INEQUALITIES WITH
CONSTANTS "` THAT DEPEND LINEARLY ON ` IN THE SENSE OF %Q (ENCE
WE ARRIVE AT
4HEOREM 5NIVERSAL ENTROPY 4HE ADDITIVE ENTROPY CONSTANTS OF ALL
SYSTEMS CAN BE CALIBRATED IN SUCH A WAY THAT THE ENTROPY IS ADDITIVE AND
EXTENSIVE AND 8 } 9 IMPLIES 38 s 39 EVEN WHEN 8 AND 9 DO NOT
BELONG TO THE SAME STATE SPACE
%( ,)%" !.$ * 9.'6!3/. '!&!
/UR lNAL REMARK CONCERNS THE REMAINING NON UNIQUENESS OF THE CON
STANTS "` 4HIS INDETERMINACY CAN BE TRACED BACK TO THE NON UNIQUENESS
OF A LINEAR FUNCTIONAL LYING BETWEEN _& ` ` AND & ` ` AND HAS TWO
POSSIBLE SOURCES /NE IS THAT SOME PAIRS OF STATE SPACES ` AND ` MAY NOT
BE CONNECTED IE & ` ` MAY BE INlNITE IN WHICH CASE & ` ` IS ALSO
INlNITE BY AXIOM ! 4HE OTHER IS THAT THERE MIGHT BE A TRUE GAP IE
_& ` ` & ` `
MIGHT HOLD FOR SOME STATE SPACES EVEN IF BOTH SIDES ARE lNITE
)N NATURE ONLY STATES CONTAINING THE SAME AMOUNT OF THE CHEMICAL ELE
MENTS CAN BE TRANSFORMED INTO EACH OTHER (ENCE & ` ` FOR MANY
PAIRS OF STATE SPACES IN PARTICULAR FOR THOSE THAT CONTAIN DIkERENT AMOUNTS
OF SOME CHEMICAL ELEMENT 4HE CONSTANTS "` ARE THEREFORE NEVER UNIQUE
&OR EACH EQUIVALENCE CLASS OF STATE SPACES WITH RESPECT TO THE RELATION OF
CONNECTEDNESS ONE CAN DElNE A CONSTANT THAT IS ARBITRARY EXCEPT FOR THE
PROVISO THAT THE CONSTANTS SHOULD BE ADDITIVE AND EXTENSIVE UNDER COMPO
SITION AND SCALING OF SYSTEMS )N OUR WORLD THERE ARE CHEMICAL ELEMENTS
OR STRICTLY SPEAKING A SOMEWHAT LARGER NUMBER . SINCE ONE SHOULD COUNT
DIkERENT ISOTOPES AS DIkERENT ELEMENTS AND THIS LEAVES US WITH AT LEAST
FREE CONSTANTS THAT SPECIFY THE ENTROPY OF ONE GRAM OF EACH OF THE CHEMICAL
ELEMENTS IN SOME SPECIlC STATE
4HE OTHER POSSIBLE SOURCE OF NON UNIQUENESS A NONTRIVIAL GAP FOR
SYSTEMS WITH THE SAME COMPOSITION IN TERMS OF THE CHEMICAL ELEMENTS
IS AS FAR AS WE KNOW NOT REALIZED IN NATURE .OTE THAT THIS ASSERTION
CAN BE TESTED EXPERIMENTALLY WITHOUT INVOKING SEMIPERMEABLE MEMBRANES
(ENCE ONCE THE ENTROPY CONSTANTS FOR THE CHEMICAL ELEMENTS HAVE BEEN lXED
AND A TEMPERATURE UNIT HAS BEEN CHOSEN TO lX THE MULTIPLICATIVE CONSTANTS
THE UNIVERSAL ENTROPY IS COMPLETELY lXED
7E ARE INDEBTED TO MANY PEOPLE FOR HELPFUL DISCUSSIONS INCLUDING &RED
!LMGREN 4HOR "AK "ERNARD "AUMGARTNER 0IERLUIGI #ONTUCCI 2OY *ACK
SON !NTHONY +NAPP -ARTIN +RUSKAL -ARY "ETH 2USKAI AND *AN 0HILIP
3OLOVEJ
3OME 3PECULATIONS AND /PEN 0ROBLEMS

!S WE HAVE STRESSED THE PURPOSE OF THE ENTROPY FUNCTION IS TO QUAN
TIFY THE LIST OF EQUILIBRIUM STATES THAT CAN EVOLVE FROM OTHER EQUILIBRIUM
STATES 4HE EVOLUTION CAN BE ARBITRARILY VIOLENT BUT ALWAYS 38 s 39
IF 8 } 9 )NDEED THE EARLY THERMODYNAMICISTS UNDERSTOOD THE MEANING OF
6ISIONS IN -ATH 3%#/.$ ,!7 /& 4(%2-/$9.!-)#3
ENTROPY AS DElNED FOR EQUILIBRIUM STATES /F COURSE IN THE REAL WORLD ONE IS
OFTEN CLOSE TO EQUILIBRIUM WITHOUT ACTUALLY BEING THERE AND IT MAKES SENSE
TO TALK ABOUT ENTROPY AS A FUNCTION OF TIME AND EVEN SPACE FOR SITUATIONS
CLOSE TO EQUILIBRIUM 7E DO A SIMILAR THING WITH RESPECT TO TEMPERATURE
WHICH HAS THE SAME PROBLEM THAT TEMPERATURE IS ONLY STRICTLY DElNED FOR
A HOMOGENEOUS SYSTEM IN EQUILIBRIUM !T SOME POINT THE THOUGHT AROSE
AND WE CONFESS OUR IGNORANCE ABOUT HOW IT AROSE AND BY WHOM THAT IT
OUGHT TO BE POSSIBLE TO DElNE AN ENTROPY FUNCTION RIGOROUSLY FOR MANIFESTLY
NON EQUILIBRIUM STATES IN SUCH A WAY THAT THE NUMERICAL VALUE OF THIS FUNC
TION WILL INCREASE WITH TIME AS A SYSTEM GOES FROM ONE EQUILIBRIM STATE TO
ANOTHER
$ESPITE THE FACT THAT MOST PHYSICISTS BELIEVE IN SUCH A NON EQUILIBRIUM
ENTROPY IT HAS SO FAR PROVED TO BE IMPOSSIBLE TO DElNE IT IN A CLEARLY SAT
ISFACTORY WAY &OR EXAMPLE "OLTZMANNS FAMOUS ( 4HEOREM SHOWS THE
STEADY INCREASE OF A CERTAIN FUNCTION CALLED ( 4HIS HOWEVER IS NOT THE
WHOLE STORY AS "OLTZMANN HIMSELF KNEW FOR ONE THING ( 3 IN EQUILIB
RIUM EXCEPT FOR IDEAL GASES AND FOR ANOTHER NO ONE HAS SO FAR PROVED THE
INCREASE WITHOUT MAKING SEVERE ASSUMPTIONS AND THEN ONLY FOR A SHORT TIME
INTERVAL CF ;,= %VEN TODAY THERE IS NO UNIVERSAL AGREEMENT ABOUT WHAT
PRECISELY ONE SHOULD TRY TO PROVE AS AN EXAMPLE OF THE ONGOING DISCUSSION
SEE ;,E02=
)T IS NOT CLEAR IF ENTROPY CAN BE CONSISTENTLY EXTENDED TO NON EQUILIBRIUM
SITUATIONS IN THE DESIRED WAY !FTER A CENTURY AND A HALF OF THOUGHT INCLUD
ING THE RISE OF THE SCIENCE OF STATISTICAL MECHANICS AS A PARADIGM WHICH WAS
NOT AVAILABLE TO THE EARLY THERMODYNAMICISTS AND THEREFORE OUTSIDE THEIR
THOUGHTS WE ARE FAR FROM SUCCESS )T HAS TO BE ADDED HOWEVER THAT A GREAT
DEAL OF PROGRESS IN UNDERSTANDING THE PROBLEM HAS BEEN MADE RECENTLY EG
;'=
)F SUCH A CONCEPT CAN BE FORMULATED PRECISELY WILL IT HAVE TO INVOLVE THE
NOTION OF ATOMS AND STATISTICAL MECHANICAL CONCEPTS OR CAN IT BE DElNED IR
RESPECTIVE OF MODELS AS WE HAVE DONE FOR THE ENTROPY OF EQUILIBRIUM STATES
4HIS IS THE QUESTION WE POSE
4HERE ARE SEVERAL MAJOR PROBLEMS TO BE OVERCOME AND WE LIST TWO OF
THEM
A 4HE PROBLEM OF TIME REVERSIBILITY )F THE CONCEPT IS GOING TO DEPEND
UPON MECHANICAL MODELS WE HAVE TO ACCOUNT FOR THE FACT THAT BOTH
CLASSICAL AND QUANTUM MECHANICS ARE TIME REVERSIBLE 4HIS MAKES IT
DInCULT TO CONSTRUCT A MECHANICAL QUANTITY THAT CAN ONLY INCREASE
%( ,)%" !.$ * 9.'6!3/. '!&!
UNDER CLASSICAL OR QUANTUM MECHANICAL TIME EVOLUTION )NDEED THIS

PROBLEM USUALLY OCCUPIES CENTER STAGE IN MOST DISCUSSIONS OF THE SUB
JECT BUT IT MIGHT ULTIMATELY NOT BE THE MOST DInCULT PROBLEM AFTER
ALL
B )N OUR VIEW OF THE SUBJECT A KEY ROLE IS PLAYED BY THE IDEA OF A MORE
OR LESS ARBITRARY PEACEFUL OR VIOLENT INTERACTION OF THE SYSTEM UNDER
DISCUSSION WITH THE REST OF THE UNIVERSE WHOSE lNAL RESULT IS A CHANGE OF
THE SYSTEM THE CHANGE IN HEIGHT OF A WEIGHT AND NOTHING MORE (OW
CAN ONE MODEL SUCH AN ARBITRARY INTERACTION IN A MECHANICAL WAY "Y
MEANS OF AN ADDITIONAL TERM IN THE (AMILTONIAN 4HAT HARDLY SEEMS
LIKE A REASONABLE WAY TO MODEL A SLEDGEHAMMER THAT HAPPENS TO FALL
ON THE SYSTEM OR A GORILLA JUMPING UP AND DOWN !LL DISCUSSIONS
OF ENTROPY INCREASE REFER TO THE EVOLUTION OF ISOLATED SYSTEMS SUBJECT
TO (AMILTONIAN DYNAMICAL EVOLUTION 4HIS CAN HARDLY EVEN COPE WITH
DESCRIBING A STEAM ENGINE MUCH LESS A RANDOM VIOLENT EXTERNAL FORCE
!S A MATTER OF FACT MOST PEOPLE WOULD RECOGNIZE A AS THE IMPORTANT
PROBLEM NOW AND IN THE PAST )N B WE INTERJECT A NEW NOTE WHICH TO US
IS POSSIBLY MORE DInCULT 4HERE ARE SEVERAL PROPOSALS FOR A RESOLUTION OF
THE IRREVERSIBILITY PROBLEM SUCH AS THE LARGE NUMBER x OF ATOMS
INVOLVED OR THE @SENSITIVE DEPENDENCE ON INITIAL CONDITIONS ONE CAN SHOOT
A BALL OUT OF A CANNON BUT IT IS VERY DInCULT TO SHOOT IT BACK INTO THE
CANNONS MOUTH 0ROBLEM B IN CONTRAST HAS NOT RECEIVED NEARLY AS MUCH
ATTENTION
!N ESSENTIAL ROLE IN OUR STORY WAS PLAYED BY AXIOMS ! AND !
WHICH REQUIRE THE POSSIBLITY OF HAVING ARBITRARILY SMALL SAMPLES OF A GIVEN
MATERIAL AND THAT THESE SMALL SAMPLES BEHAVE IN EXACTLY THE SAME WAY AS A
KILOGRAM SAMPLE 7HILE THIS ASSUMPTION IS MADE IN EVERYONES FORMULATION
OF THE SECOND LAW WE HAVE TO RECOGNIZE THAT ABSURDITIES WILL ARISE IF WE
PUSH THE CONCEPT TO ITS EXTREME %VENTUALLY THE ATOMIC NATURE OF MATTER
WILL REVEAL ITSELF AND ENTROPY WILL CEASE TO HAVE A CLEAR MEANING 7HAT
PROTECTS US IS THE HUGE POWER OF TEN EG THAT SEPARATES MACROSCOPIC
PHYSICS AND THE REALM OF ATOMS
,IKEWISE A HUGE POWER OF TEN SEPARATES TIME SCALES THAT MAKE PHYSI
CAL SENSE IN THE ORDINARY MACROSCOPIC WORLD AND TIME SCALES SUCH AS
SECONDS TIMES THE AGE OF THE UNIVERSE WHICH ARE NEEDED FOR ATOMIC
mUCTUATIONS TO UPSET THE TIME EVOLUTION ONE WOULD OBTAIN FROM MACROSCOPIC
DYNAMICS /NE MIGHT SAY THAT ONE OF THE HIDDEN ASSUMPTIONS IN OUR AND
EVERYONE ELSES ANALYSIS IS THAT } IS REPRODUCIBLE IE 8 } 9 EITHER HOLDS OR
6ISIONS IN -ATH 3%#/.$ ,!7 /& 4(%2-/$9.!-)#3
IT DOES NOT AND THERE ARE NO HIDDEN STOCHASTIC OR PROBABILISTIC MECHANISMS

THAT WOULD MAKE THE LIST OF PAIRS 8 } 9 @FUZZY
/NE OF THE BURGEONING AREA OF PHYSICS RESEARCH IS MESOSCOPICS WHICH
DEALS WITH THE INTERESTING PROPERTIES OF TINY PIECES OF MATTER THAT MIGHT
CONTAIN ONLY A MILLION ATOMS A CUBE OF ATOMS ON A SIDE OR LESS !T
SOME POINT THE SECOND LAW HAS TO GET FUZZY AND A SIGNIlCANT OPEN PROBLEM
IS TO FORMULATE A FUZZY VERSION OF WHAT WE HAVE DONE IN ;,I9= /F COURSE
NO AMOUNT OF INGENUITY WITH MESOSCOPIC SYSTEMS IS ALLOWED TO VIOLATE THE
SECOND LAW ON THE MACROSCOPIC LEVEL AND THIS WILL HAVE TO BE TAKEN INTO
ACCOUNT /NE POSSIBILITY COULD BE THAT AN ENTROPY FUNCTION CAN STILL BE
DElNED FOR MESOSCOPIC SYSTEMS BUT THAT } IS FUZZY WITH THE CONSEQUENCE
THAT ENTROPY INCREASES ONLY ON @THE AVERAGE BUT IN A TOTALLY UNPREDICTABLE
WAY [ SO THAT THE OCCASIONAL DECREASE OF ENTROPY CANNOT BE UTILIZED TO VIOLATE
THE SECOND LAW ON THE MACROSCOPIC LEVEL
4HERE ARE OTHER PROBLEMS AS WELL ! SIMPLE SYSTEM SUCH AS A CONTAINER
OF HYDROGEN GAS HAS STATES DESCRIBED BY ENERGY AND VOLUME &OR A MESO
SCOPIC QUANTITY OF MATTER THIS MAY NOT SUnCE TO DESCRIBE AN EQUILIBRIUM
STATE !NOTHER PROBLEM IS THE MEANING OF EQUILIBRIUM AND THE IMPLICIT
ASSUMPTION WE MADE THAT AFTER THE VIOLENT ADIABATIC PROCESS IS OVER THE
SYSTEM WILL EVENTUALLY COME TO SOME EQUILIBRIUM STATE IN A TIME SCALE THAT
IS SHORT COMPARED TO THE AGE OF THE UNIVERSE /N THE MESOSCOPIC LEVEL
THE ACHIEVEMENT OF EQUILIBRIUM MAY BE MORE DELICATE BECAUSE A MESOSCOPIC
SYSTEM MIGHT NEVER SETTLE DOWN TO A STATE WITH INSIGNIlCANT mUCTUATIONS
THAT ONE WOULD BE PLEASED TO CALL AN EQUILIBRIUM STATE
4O SUMMARIZE WE HAVE LISTED TWO AND THERE ARE SURELY MORE AREAS
IN WHICH MORE THOUGHT BOTH MATHEMATICAL AND PHYSICAL IS NEEDED THE
EXTENSION OF THE SECOND LAW AND THE ENTROPY CONCEPT TO NON EQUILIBRIUM
SITUATIONS AND MESOSCOPIC AND EVEN ATOMIC SITUATIONS /NE MIGHT OB
JECT THAT THE PROBLEMS CANNOT BE SOLVED UNTIL THE @RULES OF THE GAME ARE
MADE CLEAR BUT DISCOVERING THE RULES IS PART OF THE PROBLEM 4HAT IS SOME
TIMES INHERENT IN MATHEMATICAL PHYSICS AND THAT IS ONE OF THE INTELLECTUAL
CHALLENGES OF THE lELD
2EFERENCES
;"= *' "OYLING !N AXIOMATIC APPROACH TO CLASSICAL THERMODYNAMICS 0ROC
2OY 3OC ,ONDON ! [
%( ,)%" !.$ * 9.'6!3/. '!&!
;"U= (! "UCHDAHL 4HE #ONCEPTS OF #LASSICAL 4HERMODYNAMICS #AMBRIDGE

5NIVERSITY 0RESS #AMBRIDGE
;#= # #ARATHs EODORY 5NTERSUCHUNG U BER DIE 'RUNDLAGEN DER 4HERMODY
NAMIK -ATH !NNALEN [
;#O= *," #OOPER 4HE FOUNDATIONS OF THERMODYNAMICS *OUR -ATH !NAL
AND !PPL [
;$= ** $UISTERMAAT %NERGY AND ENTROPY AS REAL MORPHISMS FOR ADDITION
AND ORDER 3YNTHESE [
;'= ' 'ALLAVOTTI 3TATISTICAL -ECHANICS ! 3HORT 4REATISE 3PRINGER 4EXTS
AND -ONOGRAPHS IN 0HYSICS
;'I= 2 'ILES -ATHEMATICAL &OUNDATIONS OF 4HERMODYNAMICS 0ERGAMON /X
FORD
;,= /% ,ANFORD ))) 4IME EVOLUTION OF LARGE CLASSICAL SYSTEMS 3PRINGER
,ECTURE .OTES IN 0HYSICS * -OSER ED [
;,E02= *, ,EBOWITZ ) 0RIGOGINE $ 2UELLE 2OUND TABLE ON IRREVERSIBIL
ITY IN <3TATISTICAL 0HYSICS 88 .ORTH (OLLAND

;,I= %( ,IEB 3OME PROBLEMS IN STATISTICAL MECHANICS THAT ) WOULD LIKE TO
SEE SOLVED )50!0 "OLTZMANN 0RIZE ,ECTURE 0HYSICA !

;,I9= %( ,IEB * 9NGVASON 4HE PHYSICS AND MATHEMATICS OF THE 3ECOND
,AW OF 4HERMODYNAMICS 0HYSICS 2EPORTS [ !USTIN -ATH
0HYS ARCHIVE [ ,OS !LAMOS ARCHIVE COND MAT
;,I9= %( ,IEB * 9NGVASON ! GUIDE TO ENTROPY AND THE 3ECOND ,AW OF 4HER
MODYNAMICS .OTICES OF THE !MER -ATH 3OC [ !USTIN
-ATH 0HYS ARCHIVE [ ,OS !LAMOS ARCHIVE COND MAT
;,I9= %( ,IEB * 9NGVASON ! FRESH LOOK AT ENTROPY AND THE 3ECOND ,AW OF
4HERMODYNAMICS 0HYSICS 4ODAY [ MP ARC AR8IV
MATH PH
;0= - 0LANCK 5BER DIE "EGRUNDUNG DES ZWEITEN (AUPTSATZES DER 4HER
MODYNAMIK 3ITZUNGSBER 0REUSS !KAD 7ISS 0HYS -ATH +L
[
;2,= &3 2OBERTS 2$ ,UCE !XIOMATIC THERMODYNAMICS AND EXTENSIVE
MEASUREMENT 3YNTHESE [
;5= * 5FFINK "LUk YOUR WAY IN THE SECOND LAW OF THERMODYNAMICS PREPRINT

%LLIOT ( ,IEB $EPTS OF -ATHEMATICS AND 0HYSICS 0RINCETON 5NIVERSITY *AD

WIN (ALL 0/ "OX 0RINCETON .* 53!
*AKOB 9NGVASON )NSTITUT FUR 4HEORETISCHE 0HYSIK 5NIVERSIT
AT 7IEN "OLTZMAN
NGASSE ! 6IENNA !USTRIA
$)3#2%4% !.$ #/.4).5/53

47/ 3)$%3 /& 4(% 3!-%
sSZLO
,A s ,OVA
sSZ
!BSTRACT
(OW DEEP IS THE DIVIDING LINE BETWEEN DISCRETE AND CONTINUOUS MATHE
MATICS "ASIC STRUCTURES AND METHODS OF BOTH SIDES OF OUR SCIENCE ARE
QUITE DIkERENT "UT ON A DEEPER LEVEL THERE IS A MORE SOLID CONNECTION
THAN MEETS THE EYE
)NTRODUCTION
4HERE IS A RATHER CLEAR DIVIDING LINE BETWEEN DISCRETE AND CONTINUOUS MATH
EMATICS #ONTINUOUS MATHEMATICS IS CLASSICAL WELL ESTABLISHED WITH A RICH
VARIETY OF APPLICATIONS $ISCRETE MATHEMATICS GREW OUT OF PUZZLES AND THEN
IS OFTEN IDENTIlED WITH SPECIlC APPLICATION AREAS LIKE COMPUTER SCIENCE OR
OPERATIONS RESEARCH "ASIC STRUCTURES AND METHODS OF BOTH SIDES OF OUR
SCIENCE ARE QUITE DIkERENT CONTINUOUS MATHEMATICIANS USE LIMITS DISCRETE
MATHEMATICIANS USE INDUCTION
4HIS STATE OF MATHEMATICS AT THIS PARTICULAR TIME THE TURN OF THE MILLEN
NIUM IS A PRODUCT OF STRONG INTRINSIC LOGIC BUT ALSO OF HISTORICAL COINCIDENCE
4HE MAIN EXTERNAL SOURCE OF MATHEMATICAL PROBLEMS IS SCIENCE IN PAR
TICULAR PHYSICS 4HE TRADITIONAL VIEW IS THAT SPACE AND TIME ARE CONTINUOUS
AND THAT THE LAWS OF NATURE ARE DESCRIBED BY DIkERENTIAL EQUATIONS 4HERE
ARE OF COURSE EXCEPTIONS CHEMISTRY STATISTICAL MECHANICS AND QUANTUM
PHYSICS ARE BASED ON AT LEAST A CERTAIN DEGREE OF DISCRETENESS "UT THESE
DISCRETE OBJECTS MOLECULES ATOMS ELEMENTARY PARTICLES &EYNMAN GRAPHS
LIVE IN THE CONTINUUM OF SPACE AND TIME 3TRING THEORY TRIES TO EXPLAIN THESE
DISCRETE OBJECTS AS SINGULARITIES OF A HIGHER DIMENSIONAL CONTINUUM
!CCORDINGLY THE MATHEMATICS USED IN MOST APPLICATIONS IS ANALYSIS THE
REAL HARD CORE OF OUR SCIENCE "UT ESPECIALLY IN NEWLY DEVELOPING SCIENCES
DISCRETE MODELS ARE BECOMING MORE AND MORE IMPORTANT
/NE MIGHT OBSERVE THAT THERE IS A lNITE NUMBER OF EVENTS IN ANY lNITE
DOMAIN OF SPACE TIME )S THERE A PHYSICAL MEANING OF THE REST OF A FOUR OR
TEN DIMENSIONAL MANIFOLD $OES <THE POINT IN TIME HALF WAY BETWEEN TWO
CONSECUTIVE INTERACTIONS OF AN ELEMENTARY PARTICLE MAKE SENSE 3HOULD
s
, ,/6!3: '!&!
THE ANSWER TO THIS QUESTION BE YES OR NO DISCRETE FEATURES OF SUBATOMIC

WORLD ARE UNDENIABLE (OW FAR WOULD A COMBINATORIAL DESCRIPTION OF THE
WORLD TAKE US /R COULD IT HAPPEN THAT THE DESCRIPTIONS OF THE WORLD AS
A CONTINUUM OR AS A DISCRETE BUT TERRIBLY HUGE STRUCTURE ARE EQUIVALENT
EMPHASIZING TWO SIDES OF THE SAME REALITY
"UT LET ME ESCAPE FROM THESE UNANSWERABLE QUESTIONS AND LOOK AROUND
ELSEWHERE IN SCIENCE "IOLOGY TRIES TO UNDERSTAND THE GENETIC CODE A GIGAN
TIC TASK WHICH IS THE KEY TO UNDERSTANDING LIFE AND ULTIMATELY OURSELVES
4HE GENETIC CODE IS DISCRETE SIMPLE BASIC QUESTIONS LIKE lNDING MATCHING
PATTERNS OR TRACING CONSEQUENCES OF mIPPING OVER SUBSTRINGS SOUND MORE
FAMILIAR TO THE GRAPH THEORIST THAN TO THE RESEARCHER OF DIkERENTIAL EQUA
TIONS 1UESTIONS ABOUT THE INFORMATION CONTENT REDUNDANCY OR STABILITY OF
THE CODE MAY SOUND TOO VAGUE TO A CLASSICAL MATHEMATICIAN BUT A THEORETI
CAL COMPUTER SCIENTIST WILL IMMEDIATELY SEE AT LEAST SOME TOOLS TO FORMALIZE
THEM EVEN IF TO lND THE ANSWER MAY BE TOO DInCULT AT THE MOMENT
%CONOMICS IS A HEAVY USER OF MATHEMATICS \ AND MUCH OF ITS NEED IS
NOT PART OF THE TRADITIONAL APPLIED MATHEMATICS TOOLBOX 0ERHAPS THE MOST
SUCCESSFUL TOOL IN ECONOMICS AND OPERATIONS RESEARCH IS LINEAR PROGRAMMING
WHICH LIVES ON THE BOUNDARY OF DISCRETE AND CONTINUOUS 4HE APPLICABILITY
OF LINEAR PROGRAMMING IN THESE AREAS HOWEVER DEPENDS ON CONDITIONS OF
CONVEXITY AND UNLIMITED DIVISIBILITY TAKING INDIVISIBILITIES INTO ACCOUNT FOR
EXAMPLE LOGICAL DECISIONS OR INDIVIDUAL AGENTS LEADS TO INTEGER PROGRAM
MING AND OTHER MODELS OF COMBINATORIAL NATURE
&INALLY THE WORLD OF COMPUTERS IS ESSENTIALLY DISCRETE AND IT IS NOT
SURPRISING THAT SO MANY DISCRETE MATHEMATICIANS ARE WORKING IN THIS SCIENCE
%LECTRONIC COMMUNICATION AND COMPUTATION PROVIDES A VAST ARRAY OF WELL
FORMULATED DInCULT AND IMPORTANT MATHEMATICAL PROBLEMS ON ALGORITHMS
DATA BASES FORMAL LANGUAGES CRYPTOGRAPHY AND COMPUTER SECURITY 6,3)
LAYOUT AND MUCH MORE
)N ALL THESE AREAS THE REAL UNDERSTANDING INVOLVES ) BELIEVE A SYNTHESIS
OF THE DISCRETE AND CONTINUOUS AND IT IS AN INTELLECTUALLY MOST CHALLENGING
GOAL TO DEVELOP THESE MATHEMATICAL TOOLS
4HERE ARE DIkERENT LEVELS OF INTERACTION BETWEEN DISCRETE AND CONTINU
OUS MATHEMATICS AND ) TREAT THEM ) BELIEVE IN THE ORDER OF INCREASING
SIGNIlCANCE
7E OFTEN USE THE lNITE TO APPROXIMATE THE INlNITE 4O DISCRETIZE A
COMPLICATED CONTINUOUS STRUCTURE HAS ALWAYS BEEN A BASIC METHOD\FROM
THE DElNITION OF THE 2IEMANN INTEGRAL THROUGH TRIANGULATING A MANIFOLD IN
6ISIONS IN -ATH $)3#2%4% !.$ #/.4).5/53 47/ 3)$%3 /& 4(% 3!-%
SAY HOMOLOGY THEORY TO NUMERICALLY SOLVING A PARTIAL DIkERENTIAL EQUATION

ON A GRID
)T IS A SLIGHTLY MORE SUBTLE THOUGHT THAT THE INlNITE IS OFTEN OR PERHAPS
ALWAYS AN APPROXIMATION OF THE LARGE lNITE #ONTINUOUS STRUCTURES ARE
OFTEN CLEANER MORE SYMMETRIC AND RICHER THAN THEIR DISCRETE COUNTERPARTS
FOR EXAMPLE A PLANAR GRID HAS A MUCH SMALLER DEGREE OF SYMMETRY THAN
THE WHOLE EUCLIDEAN PLANE )T IS A NATURAL AND POWERFUL METHOD TO STUDY
DISCRETE STRUCTURES BY <EMBEDDING THEM IN THE CONTINUOUS WORLD
3OMETIMES THE KEY STEP IN THE PROOF OF A PURELY <DISCRETE RESULT
IS THE APPLICATION OF A PURELY <CONTINUOUS THEOREM OR VICE VERSA 7E ALL
HAVE OUR FAVORITE EXAMPLES ) DESCRIBE TWO IN SECTION
)N SOME AREAS OF DISCRETE MATHEMATICS KEY PROGRESS HAS BEEN ACHIEVED
THROUGH THE USE OF MORE AND MORE SOPHISTICATED METHODS FROM ANALYSIS
4HIS IS ILLUSTRATED IN SECTION BY DESCRIBING TWO POWERFUL METHODS IN DIS
CRETE OPTIMIZATION ) COULD NOT lND ANY AREA OF <CONTINUOUS MATHEMATICS
WHERE PROGRESS WOULD BE ACHIEVED AT A SIMILAR SCALE THROUGH THE INTRODUC
TION OF DISCRETE METHODS 0ERHAPS ALGEBRAIC TOPOLOGY COMES CLOSEST
#ONNECTIONS BETWEEN DISCRETE AND CONTINUOUS MAY BE THE SUBJECT
OF MATHEMATICAL STUDY ON THEIR OWN RIGHT .UMERICAL ANALYSIS MAY BE
THOUGHT OF THIS WAY BUT DISCREPANCY THEORY IS THE BEST EXAMPLE )N THIS
ARTICLE WE HAVE TO RESTRICT OURSELVES TO DISCUSSING TWO CLASSICAL RESULTS IN
THIS BLOOMING lELD SECTION WE REFER TO THE BOOK OF "ECK AND #HEN
;"E#= THE EXPOSITORY ARTICLE OF "ECK AND 3sOS ;"E3= AND THE RECENT BOOK OF
-ATOUtSEK ;-=
4HE MOST SIGNIlCANT LEVEL OF INTERACTION IS WHEN ONE AND THE SAME
PHENOMENON APPEARS IN BOTH THE CONTINUOUS AND DISCRETE SETTING )N SUCH
CASES INTUITION AND INSIGHT GAINED FROM CONSIDERING ONE OF THESE MAY BE
EXTREMELY USEFUL IN THE OTHER ! WELL KNOWN EXAMPLE IS THE CONNECTION
BETWEEN SEQUENCES AND ANALYTIC FUNCTIONS PROVIDED BY THE POWER SERIES
EXPANSION )N THIS CASE THERE IS A <DICTIONARY BETWEEN COMBINATORIAL
ASPECTS RECURRENCES ASYMPTOTICS OF THE SEQUENCE AND ANALYTIC PROPERTIES
DIkERENTIAL EQUATIONS SINGULARITIES OF ITS GENERATING FUNCTION
)N SECTION ) DISCUSS IN DETAIL THE DISCRETE AND CONTINUOUS NOTION OF
<,APLACIAN CONNECTED WITH A VARIETY OF DISPERSION TYPE PROCESSES 4HIS
NOTION CONNECTS TOPICS FROM THE HEAT EQUATION TO "ROWNIAN MOTION TO RAN
DOM WALKS ON GRAPHS TO LINKLESS EMBEDDINGS OF GRAPHS
!N EXCITING BUT NOT WELL UNDERSTOOD FURTHER PARALLELISM IS CONNECTED
WITH THE FACT THAT THE ITERATION OF VERY SIMPLE STEPS RESULTS IN VERY COMPLEX
s
, ,/6!3: '!&!
STRUCTURES )N CONTINUOUS MATHEMATICS THIS IDEA COMES UP IN THE THEORY OF

DYNAMICAL SYSTEMS AND NUMERICAL ANALYSIS )N DISCRETE MATHEMATICS IT IS
IN A SENSE THE BASIS OF MANY ALGORITHMS RANDOM NUMBER GENERATION ETC
! SYNTHESIS OF IDEAS FROM BOTH SIDES COULD BRING SPECTACULAR DEVELOPMENTS
HERE
/NE MIGHT MENTION MANY OTHER AREAS WITH SUBSTANTIAL DISCRETE AND CON
TINUOUS COMPONENTS GROUPS PROBABILITY GEOMETRY )NDEED MY STARTING
OBSERVATION ABOUT THIS DIVISION BECOMES QUESTIONABLE IF WE THINK OF SOME OF
THE RECENT DEVELOPMENTS IN THESE AREAS ) BELIEVE THAT FURTHER DEVELOPMENT
WILL MAKE IT TOTALLY MEANINGLESS
!CKNOWLEDGEMENT ) AM INDEBTED TO SEVERAL OF MY COLLEAGUES IN PAR
TICULAR TO !LAN #ONNES AND -IKE &REEDMAN FOR THEIR VALUABLE COMMENTS
ON THIS PAPER
$ISCRETE IN #ONTINUOUS AND #ONTINUOUS IN $ISCRETE

-ARRIAGE AND MEASURES ! STRIKING APPLICATION OF GRAPH THEORY
TO MEASURE THEORY IS THE CONSTRUCTION OF THE (AAR MEASURE ON COMPACT
TOPOLOGICAL GROUPS 4HIS APPLICATION WAS MENTIONED BY 2OTA AND (ARPER
;2O(= WHO ELABORATED UPON THE IDEA IN THE EXAMPLE OF THE CONSTRUCTION
OF A TRANSLATION INVARIANT INTEGRAL FOR ALMOST PERIODIC FUNCTIONS ON AN AR
BITRARY GROUP 4HE PROOF IS ALSO RELATED TO 7EILS PROOF OF THE EXISTENCE
OF (AAR MEASURE 4HERE ARE OTHER PERHAPS MORE SIGNIlCANT APPLICATIONS
OF MATCHINGS TO MEASURES THAT COULD BE MENTIONED HERE FOR EXAMPLE THE
THEOREM OF /RNSTEIN ;/= ON THE ISOMORPHISM OF "ERNOULLI SHIFTS CF ALSO
;$U= ;,O-= ;,O0= ;3T=
(ERE WE SHALL DESCRIBE THE CONSTRUCTION OF A TRANSLATION INVARIANT INTE
GRAL FOR CONTINUOUS FUNCTIONS ON A COMPACT TOPOLOGICAL GROUP EQUIVALENT
TO THE EXISTENCE OF THE (AAR MEASURE ;,O0= !N INVARIANT INTEGRATION FOR
CONTINUOUS FUNCTIONS ON A COMPACT TOPOLOGICAL GROUP IS A LINEAR FUNCTIONAL
, DElNED ON THE CONTINUOUS REAL VALUED FUNCTIONS ON ' WITH THE FOLLOWING
PROPERTIES
A ,jF kG j,F k,G LINEARITY
B )F F t THEN ,F t MONOTONICITY
C )F L DENOTES THE IDENTITY FUNCTION THEN ,L NORMALIZATION
D )F S AND T ARE IN ' AND F AND G ARE TWO CONTINUOUS FUNCTIONS SUCH
THAT GX F SXT FOR EVERY X THEN ,G ,F DOUBLE TRANSLATION
INVARIANCE
6ISIONS IN -ATH $)3#2%4% !.$ #/.4).5/53 47/ 3)$%3 /& 4(% 3!-%
4HEOREM &OR EVERY COMPACT TOPOLOGICAL GROUP THERE EXISTS AN INVARI

ANT INTEGRATION FOR CONTINUOUS FUNCTIONS
0ROOF ,ET F BE THE FUNCTION WE WANT TO INTEGRATE 4HE IDEA IS TO APPROXI
MATE THE INTEGRAL BY THE AVERAGE
8

F ! J!J F A
A!
WHERE ! IS A <UNIFORMLY DENSE lNITE SUBSET OF GROUP ' 4HE QUESTION IS
HOW TO lND AN APPROPRIATE SET !
4HE KEY DElNITION IS THE FOLLOWING ,ET 5 BE AN OPEN NON EMPTY SUBSET
OF ' THINK OF IT AS <SMALL ! SUBSET ! y ' IS CALLED A 5 NET IF !<S5 T
FOR EVERY S T ' )T FOLLOWS FROM THE COMPACTNESS OF ' THAT THERE EXISTS A
lNITE 5 NET
/F COURSE A 5 NET MAY BE VERY UNEVENLY DISTRIBUTED OVER THE GROUP AND
ACCORDINGLY THE AVERAGE OVER IT MAY NOT APPROXIMATE THE INTEGRAL 7HAT
WE SHOW IS THAT THE SIMPLE TRICK OF RESTRICTING OURSELVES TO 5 NETS WITH
MINIMUM CARDINALITY MINIMUM 5 NETS FOR SHORT WE GET A SUnCIENTLY
UNIFORMLY DENSE lNITE SET
4O MEASURE THIS UNIFORMITY WE DElNE
g k h
m5 SUP JF X F Y J k X Y S5 T FOR SOME S T '
4HE COMPACTNESS OF ' IMPLIES THAT IF F IS CONTINUOUS THEN IT IS ALSO UNI
FORMLY CONTINUOUS IN THE SENSE THAT FOR EVERY n THERE EXISTS A NON EMPTY
OPEN SET 5 SUCH THAT m5 n
,ET ! AND " BE MINIMUM 5 NETS 4HE FOLLOWING INEQUALITY IS THE HEART
OF THE PROOF
k k
kF ! _ F " k s m5
4HE COMBINATORIAL CORE OF THE CONSTRUCTION LIES IN THE PROOF OF THIS IN
EQUALITY $ElNE A BIPARTITE GRAPH ( WITH BIPARTITION ! " BY CONNECTING
X ! TO Y " IF AND ONLY IF THERE EXISTS S T ' SUCH THAT X Y S5 T 7E
USE THE -ARRIAGE THEOREM TO SHOW THAT THIS BIPARTITE GRAPH HAS A PERFECT
MATCHING 7E HAVE TO VERIFY TWO THINGS
) J!J J"J 4HIS IS TRIVIAL
)) %VERY SET 8 q ! HAS AT LEAST J8J NEIGHBORS IN " )NDEED LET 9 BE
THE SET OF NEIGHBORS OF 8 7E SHOW THAT 4 9 ; ! N 8 IS A 5 NET
,ET S T ' WE SHOW THAT 4 INTERSECTS S5 T 3INCE ! IS A 5 NET THERE
EXISTS AN ELEMENT X ! < S5 T )F X
8 THEN X 4 AND WE ARE DONE
/THERWISE WE USE THAT THERE EXISTS AN ELEMENT Y " < S5 T 4HEN XY
IS AN LINE OF ( AND SO Y 9 AND WE ARE DONE AGAIN
s
, ,/6!3: '!&!
4HUS 4 IS A 5 NET 3INCE ! IS A 5 NET WITH MINIMUM CARDINALITY WE

MUST HAVE J4 J t J!J WHICH IS EQUIVALENT TO J9 J t J8J
4HUS ( HAS A PERFECT MATCHING FA B AN BN G 4HEN
k N k
k k k 8_ `k
kF ! _ F " k k F AI _ F BI kk
Nk
I
N
8 kk k _ `
s F AI _ F BI k s Nm5 m5
N N
I
4HE REST OF THE PROOF IS RATHER ROUTINE &IRST WE NEED TO SHOW THAT
AVERAGING OVER MINIMUM 5 NETS FOR DIkERENT 5 S GIVES APPROXIMATELY THE
SAME RESULT -ORE EXACTLY LET ! BE A MINIMUM 5 NET AND " A MINIMUM
6 NET THEN
k k
kF ! _ F " k s m5 m6
)NDEED FOR EVERY B " !B IS ALSO A 5 NET WITH MINIMUM CARDINALITY SO
BY k k
kF ! _ F !B k s m5
(ENCE
k k k 8 k 8 kk k
kF ! _F !" k kkF ! _ k
F !B k s F ! _F !B k s m5
J"J J"J
B" B"
3IMILARLY k k
kF " _ F !" k s m6
WHENCE FOLLOWS
.OW CHOOSE A SEQUENCE 5N OF OPEN SETS SUCH THAT m5N AND LET
!N BE A MINIMUM 5N NET 4HEN IMPLIES THAT THE SEQUENCE F !N TENDS
TO A LIMIT ,F WHICH IS INDEPENDENT OF THE CHOICE OF 5N AND !N AND SO IT
IS WELL DElNED #ONDITIONS A [D ARE TRIVIAL TO VERIFY c
$ISJOINT SUBSETS AND TOPOLOGY .OW WE TURN TO EXAMPLES DEMON

STRATING THE CONVERSE DIRECTION APPLICATIONS OF CONTINUOUS METHODS TO A
PURELY COMBINATORIAL PROBLEMS /UR lRST EXAMPLE IS A RESULT WHERE AL
GEBRAIC TOPOLOGY AND GEOMETRY ARE THE ESSENTIAL TOOLS IN THE PROOF OF A
COMBINATORIAL THEOREM
4HEOREM ,ET US PARTITION THE K ELEMENT SUBSETS OF AN N ELEMENT SET INTO
N _ K CLASSES 4HEN ONE OF THE CLASSES CONTAINS TWO DISJOINT K SUBSETS
4HIS RESULT WAS CONJECTURED BY +NESER ;+= AND PROVED IN ;,O= 4HE
PROOF WAS SIMPLIlED BY "sARs
ANY ;"= AND WE DESCRIBE HIS VERSION HERE
6ISIONS IN -ATH $)3#2%4% !.$ #/.4).5/53 47/ 3)$%3 /& 4(% 3!-%
0ROOF 7E lRST INVOKE A GEOMETRIC CONSTRUCTION DUE TO 'ALE 4HERE EXISTS A

SET OF K D VECTORS ON THE D SPHERE 3 D SUCH THAT EVERY OPEN HEMISPHERE
CONTAINS AT LEAST K OF THEM &OR D TAKE THE VERTICES OF A REGULAR
K GON
#HOOSING D N _ K WE THUS GET A SET 3 OF N POINTS 3UPPOSE THAT ALL
THE K SUBSETS ARE PARTITIONED INTO D N _ K CLASSES 0 0 0D
,ET !I BE THE SET OF THOSE UNIT VECTORS H 3 D FOR WHICH THE OPEN HEMISPHERE
CENTERED AT H CONTAINS A K SUBSET OF 3 WHICH BELONGS TO 0I #LEARLY THE SETS
! ! !D ARE OPEN AND BY THE DElNITION OF 3 THEY COVER THE WHOLE
SPHERE
.OW WE TURN FROM GEOMETRY TO TOPOLOGY BY THE "ORSUK 5LAM THEOREM
ONE OF THE SETS !I CONTAINS TWO ANTIPODAL POINTS H AND _H 4HUS THE
HEMISPHERES ABOUT H AND _H BOTH CONTAIN A K SUBSET FROM 0I 3INCE THESE
HEMISPHERES ARE DISJOINT THESE TWO K SUBSETS ARE DISJOINT c
4HERE ARE NUMEROUS OTHER EXAMPLES WHERE METHODS FROM ALGEBRAIC TOPOL
OGY HAVE BEEN USED TO PROVE PURELY COMBINATORIAL STATEMENTS SEE ;"J= FOR
A SURVEY
/PTIMIZATION $ISCRETE ,INEAR 3EMIDElNITE

/UR NEXT EXAMPLE ILLUSTRATES HOW A WHOLE AREA OF IMPORTANT APPLIED MATH
EMATICAL PROBLEMS NAMELY DISCRETE OPTIMIZATION RELIES ON IDEAS BRINGING
MORE AND MORE SOPHISTICATED TOOLS FROM MORE TRADITIONAL CONTINUOUS OPTI
MIZATION
)N A TYPICAL OPTIMIZATION PROBLEM WE ARE GIVEN A SET 3 OF FEASIBLE SOLU
TIONS AND A FUNCTION F 3 2 CALLED THE OBJECTIVE FUNCTION 4HE GOAL
IS TO lND THE ELEMENT OF 3 WHICH MAXIMIZES OR MINIMIZES THE OBJECTIVE
FUNCTION
)N A TRADITIONAL OPTIMIZATION PROBLEM 3 IS A DECENT CONTINUUM IN A
EUCLIDEAN SPACE AND F IS A SMOOTH FUNCTION )N THIS CASE THE OPTIMUM CAN
BE FOUND BY CONSIDERING THE EQUATION RF OR BY AN ITERATIVE METHOD
USING SOME VERSION OF STEEPEST DESCENT
"OTH OF THESE DEPEND ON THE CONTINUOUS STRUCTURE IN A NEIGHBORHOOD
OF THE OPTIMIZING POINT AND THEREFORE FAIL FOR A DISCRETE OR COMBINATORIAL
OPTIMIZATION PROBLEM WHEN 3 IS lNITE 4HIS CASE IS TRIVIAL FROM A CLASSICAL
POINT OF VIEW <IN PRINCIPLE ONE COULD EVALUATE F FOR ALL ELEMENTS IN 3 AND
CHOOSE THE BEST "UT IN MOST CASES OF INTEREST 3 IS VERY LARGE IN COMPARISON
WITH THE NUMBER OF DATA NEEDED TO SPECIFY THE INSTANCE OF THE PROBLEM &OR
s
, ,/6!3: '!&!
EXAMPLE 3 MAY BE THE SET OF SPANNING TREES OR THE SET OF PERFECT MATCHINGS
OF A GRAPH OR THE SET OF ALL POSSIBLE SCHEDULES OF ALL TRAINS IN A COUNTRY OR
THE SET OF STATES OF A SPIN GLASS )N SUCH CASES WE HAVE TO USE THE IMPLICIT
STRUCTURE OF 3 RATHER THAN BRUTE FORCE TO lND THE OPTIMIZING ELEMENT
)N SOME CASES TOTALLY COMBINATORIAL METHODS ENABLE US TO SOLVE SUCH
PROBLEMS &OR EXAMPLE LET 3 BE THE SET OF ALL SPANNING TREES OF A GRAPH '
AND ASSUME THAT EACH EDGE OF ' HAS A NON NEGATIVE <LENGTH ASSOCIATED
WITH IT )N THIS CASE A SPANNING TREE WITH MINIMUM LENGTH CAN BE FOUND
BY THE GREEDY ALGORITHM DUE TO "OR|UVKA AND +RUSKAL WE REPEATEDLY
CHOOSE THE SHORTEST EDGE THAT DOES NOT FORM A CYCLE WITH THE EDGES CHOSEN
PREVIOUSLY UNTIL A TREE IS OBTAINED
)N MANY IN A SENSE MOST CASES SUCH A DIRECT COMBINATORIAL ALGORITHM
IS NOT AVAILABLE ! GENERAL APPROACH IS TO EMBED THE SET 3 INTO A CONTINUUM
3 OF SOLUTIONS AND ALSO EXTEND THE OBJECTIVE FUNCTION F TO A FUNCTION F
DElNED ON 3 4HE PROBLEM OF MINIMIZING F OVER 3 IS CALLED A RELAXATION
OF THE ORIGINAL PROBLEM )F WE DO THIS RIGHT SAY 3 IS A CONVEX SET AND
F IS A CONVEX FUNCTION THEN THE MINIMUM OF F OVER 3 CAN BE FOUND BY
CLASSICAL TOOLS DIkERENTIATION STEEPEST DESCENT ETC )F WE ARE REALLY LUCKY
OR CLEVER THE MINIMIZING ELEMENT OF 3 WILL BELONG TO 3 AND THEN WE HAVE
SOLVED THE ORIGINAL DISCRETE PROBLEM )F NOT WE MAY STILL USE THE SOLUTION
OF THE RELAXATION TO OBTAIN A BOUND ON THE SOLUTION OF THE ORIGINAL OR TO
OBTAIN AN APPROXIMATE SOLUTION 3EE LATER
0OLYHEDRAL COMBINATORICS 4HE lRST SUCCESSFUL REALIZATION OF THIS
SCHEME WAS WORKED OUT IN THE S AND S WHERE TECHNIQUES OF LINEAR PRO
GRAMMING WERE APPLIED TO COMBINATORICS ,ET US ASSUME THAT OUR COMBI
NATORIAL OPTIMIZATION PROBLEM CAN BE FORMULATED SO THAT 3 IS A SET OF _
VECTORS AND THE OBJECTIVE FUNCTION IS LINEAR 4HE SET 3 MAY BE SPECIlED BY
A VARIETY OF LOGICAL AND OTHER CONSTRAINTS IN MOST CASES IT IS QUITE EASY TO
TRANSLATE THESE INTO LINEAR INEQUALITIES
g h
3 X F GN A4 4
X s B AM X s BM
4HE OBJECTIVE FUNCTION IS
N
8
4
F X C X CI XI
I
WHERE THE CI ARE GIVEN REAL NUMBERS 3UCH A FORMULATION IS TYPICALLY EASY
TO lND
4HUS WE HAVE TRANSLATED OUR COMBINATORIAL OPTIMIZATION PROBLEM INTO
A LINEAR PROGRAM WITH INTEGRALITY CONDITIONS )T IS QUITE EASY TO SOLVE THIS
6ISIONS IN -ATH $)3#2%4% !.$ #/.4).5/53 47/ 3)$%3 /& 4(% 3!-%
IF WE DISREGARD THE INTEGRALITY CONDITIONS THE REAL GAME IS TO lND WAYS TO

WRITE UP THESE LINEAR PROGRAMS IN SUCH A WAY THAT DISREGARDING INTEGRALITY
CONDITIONS IS JUSTIlED
! NICE EXAMPLE IS MATCHING IN BIPARTITE GRAPHS 3UPPOSE THAT
' IS A BIPARTITE GRAPH WITH NODE SET FU UN V VN G
WHERE EVERY EDGE CONNECTS A NODE UI TO A NODE VJ 7E WANT TO
lND A PERFECT MATCHING IE A SET OF EDGES COVERING EACH NODE
EXACTLY ONCE
3UPPOSE THAT ' HAS A PERFECT MATCHING - AND LET

IF UI VJ -
XIJ
OTHERWISE
4HEN THE DElNING PROPERTY OF PERFECT MATCHINGS CAN BE EX
PRESSED AS FOLLOWS
8N 8 N
XIJ FOR ALL J XIJ FOR ALL I
I J
#ONVERSELY WE lND A SOLUTION OF THIS SYSTEM OF LINEAR EQUATIONS
WITH EVERY XIJ OR THEN WE HAVE A PERFECT MATCHING
5NFORTUNATELY THE SOLVABILITY OF A SYSTEM OF LINEAR EQUATIONS
EVEN OF A SINGLE EQUATION IS .0 HARD 7HAT WE CAN DO IS TO
REPLACE THE CONDITION XIJ OR BY THE WEAKER CONDITION
XIJ t
4O SOLVE A LINEAR SYSTEM LIKE IN NON NEGATIVE REAL NUMBERS
IS STILL NOT TRIVIAL BUT DOABLE EnCIENTLY IN POLYNOMIAL TIME
USING LINEAR PROGRAMMING
7E ARE NOT DONE OF COURSE SINCE IF WE lND THAT [ HAS
A SOLUTION THIS SOLUTION MAY NOT BE IN INTEGERS AND HENCE IT
MAY NOT <MEAN A PERFECT MATCHING 4HERE ARE VARIOUS WAYS
TO CONCLUDE EXTRACTING A PERFECT MATCHING FROM A FRACTIONAL
SOLUTION OF [ 4HE MOST ELEGANT IS THE FOLLOWING 4HE SET OF
ALL SOLUTIONS FORMS A CONVEX POLYTOPE .OW EVERY VERTEX OF THIS
CONVEX POLYTOPE IS INTEGRAL 4HE PROOF OF THIS FACT IS AMUSING
USING #RAMERS 2ULE AND BASIC DETERMINANT CALCULATIONS 3EE
EG ;,O0=
3O WE RUN A LINEAR PROGRAMMING ALGORITHM TO SEE IF HAS A
SOLUTION IN REAL NUMBERS )F NOT THEN THE GRAPH HAS NO PERFECT
MATCHING )F YES MOST LINEAR PROGRAMMING ALGORITHMS AUTO
MATICALLY GIVE A BASIC SOLUTION IE A VERTEX 4HIS IS AN INTEGRAL
s
, ,/6!3: '!&!
SOLUTION OF [ AND HENCE CORRESPONDS TO A PERFECT MATCH

ING
-ANY VARIATIONS OF THIS IDEA HAVE BEEN DEVELOPED BOTH IN THEORY AND
PRACTICE 4HERE ARE WAYS TO AUTOMATICALLY GENERATE NEW CONSTRAINTS AND ADD
THEM TO IF THESE FAIL THERE ARE MORE SUBTLE MORE EnCIENT METHODS TO
GENERATE NEW CONSTRAINTS IN FOR SPECIAL PROBLEMS THERE ARE WAYS TO HANDLE
THE SYSTEM EVEN IF IT GETS TOO LARGE TO BE WRITTEN UP EXPLICITLY
3EMIDElNITE OPTIMIZATION 4HIS NEW TECHNIQUE OF PRODUCING RE
LAXATIONS OF DISCRETE OPTIMIZATION PROBLEMS MAKES THE PROBLEMS CONTINUOUS
IN A MORE SOPHISTICATED WAY 7E ILLUSTRATE IT BY THE -AXIMUM #UT PROB
LEM ,ET ' 6 % BE A GRAPH 7E WANT TO lND A PARTITION 3 3 v OF 6
FOR WHICH THE NUMBER - OF EDGES CONNECTING 3 TO 3v IS MAXIMUM
4HIS PROBLEM IS .0 HARD SO WE CANNOT HOPE TO lND AN EnCIENT POLY
NOMIAL TIME ALGORITHM THAT SOLVES IT )NSTEAD WE HAVE TO SETTLE FOR LESS
WE TRY TO lND A CUT THAT IS CLOSE TO BEING OPTIMAL
)T IS EASY TO lND A CUT THAT PICKS UP HALF OF THE EDGES *UST PROCESS
THE NODES ONE BY ONE PLACING THEM IN THE SET 3 OR 3v WHICHEVER GIVES THE
LARGER NUMBER OF NEW EDGES GOING BETWEEN THE TWO SETS 3INCE NO CUT CAN
HAVE MORE THAN ALL EDGES THIS SIMPLE ALGORITHM OBTAINS AN APPROXIMATION
OF THE MAXIMUM CUT WITH AT MOST RELATIVE ERROR
)T TURNED OUT TO BE QUITE DInCULT TO IMPROVE UPON THIS SIMPLE FACT UNTIL
'OEMANS AND 7ILLIAMSON ;'7= COMBINED SEMIDElNITE OPTIMIZATION WITH
A RANDOMIZED ROUNDING TECHNIQUE TO OBTAIN AN APPROXIMATION ALGORITHM
WITH A RELATIVE ERROR OF ABOUT /N THE OTHER HAND IT WAS PROVED BY
(wASTAD ;(= TIGHTENING EARLIER RESULTS OF !RORA ET AL ;!R,-33= THAT NO
POLYNOMIAL TIME ALGORITHM CAN PRODUCE AN APPROXIMATION WITH A RELATIVE
ERROR LESS THAN ABOUT UNLESS 0 . 0
7E SKETCH THE 'OEMANS[7ILLIAMSON ALGORITHM ,ET US DESCRIBE ANY
PARTITION 6 ;6 OF 6 BY A VECTOR X F_ G6 BY LETTING XI _ Ik I 6
0
4HEN THE SIZE OF THE CUT CORRESPONDING TO X IS IJ XI _ XJ (ENCE
THE -!8 #54 PROBLEM CAN BE FORMULATED AS MAXIMIZING THE QUADRATIC
FUNCTION
8

XI _ XJ
IJ
OVER ALL X F_ G6 4HE CONDITION ON X CAN ALSO BE EXPRESSED BY

QUADRATIC CONSTRAINTS
XI I 6
6ISIONS IN -ATH $)3#2%4% !.$ #/.4).5/53 47/ 3)$%3 /& 4(% 3!-%
2EDUCING A DISCRETE OPTIMIZATION PROBLEM TO THE PROBLEM OF MAXIMIZING

A QUADRATIC FUNCTION SUBJECT TO A SYSTEM OF QUADRATIC EQUATIONS MAY SOUND
GREAT AND ONE MIGHT TRY TO USE ,AGRANGE MULTIPLICATORS AND OTHER CLASSICAL
TECHNIQUES "UT THESE WONT HELP IN FACT THE NEW PROBLEM IS VERY DInCULT
.0 HARD AND IT IS NOT CLEAR THAT WE GAIN ANYTHING
4HE NEXT TRICK IS TO LINEARIZE BY INTRODUCING NEW VARIABLES YIJ XI XJ
s I J s N 4HE OBJECTIVE FUNCTION AND THE CONSTRAINTS BECOME LINEAR IN
THESE NEW VARIABLES YIJ
8

MAXIMIZE YII YJJ _ YIJ
IJ
SUBJECT TO Y ` ` ` YNN
/F COURSE THERE IS A CATCH THE VARIABLES YIJ ARE NOT INDEPENDENT )NTRO
DUCING THE SYMMETRIC MATRIX 9 YIJ NIJ WE CAN NOTE THAT
9 IS POSITIVE SEMIDElNITE
AND
9 HAS RANK
4HE PROBLEM OF SOLVING [ WITH THE ADDITIONAL CONSTRAINTS AND
IS EQUIVALENT TO THE ORIGINAL PROBLEM AND THUS .0 HARD IN GENERAL "UT
IF WE DROP THEN WE GET A TRACTABLE RELAXATION )NDEED WHAT WE GET
IS A SEMIDElNITE PROGRAM MAXIMIZE A LINEAR FUNCTION OF THE ENTRIES OF
A POSITIVE SEMIDElNITE MATRIX SUBJECT TO LINEAR CONSTRAINTS ON THE MATRIX
ENTRIES
3INCE POSITIVE SEMIDElNITENESS OF A MATRIX 9 CAN BE TRANSLATED INTO
INlNITELY MANY LINEAR INEQUALITIES INVOLVING THE ENTRIES OF 9
V49 V t FOR ALL V 2N
SEMIDElNITE PROGRAMS CAN BE VIEWED AS LINEAR PROGRAMS WITH INlNITELY
MANY CONSTRAINTS (OWEVER THEY BEHAVE MUCH NICER THAN ONE WOULD EX
PECT !MONG OTHERS THERE IS A DUALITY THEORY FOR THEM SEE EG 7OLKOWITZ
;7= )T IS ALSO IMPORTANT THAT SEMIDElNITE PROGRAMS ARE POLYNOMIAL TIME
SOLVABLE UP TO AN ARBITRARILY SMALL ERROR NOTE THAT THE OPTIMUM SOLUTION
MAY NOT BE RATIONAL )N FACT THE ELLIPSOID METHOD ;'R,3= AND MORE IM
PORTANTLY FROM A PRACTICAL POINT OF VIEW INTERIOR POINT METHODS ;!L= EXTEND
TO SEMIDElNITE PROGRAMS
#OMING BACK TO THE -AXIMUM #UT PROBLEM LET 9 BE AN OPTIMAL SO
LUTION OF AND LET - b BE THE OPTIMUM VALUE OF THE OBJECTIVE
FUNCTION 3INCE EVERY CUT DElNES A SOLUTION WE HAVE - b t -
s
, ,/6!3: '!&!
4HE SECOND TRICK IS TO OBSERVE THAT SINCE 9 IS POSITIVE SEMIDElNITE WE

CAN WRITE IT AS A 'RAM MATRIX IE THERE EXIST VECTORS UI IF 2N SUCH THAT
U4I UJ 9IJ FOR ALL I AND J )N PARTICULAR WE GET THAT JUI J AND
8
UI _ UJ - b
IJ
.OW CHOOSE A UNIFORMLY DISTRIBUTED RANDOM HYPERPLANE ( THROUGH THE
ORIGIN 4HIS DIVIDES THE UI INTO TWO CLASSES AND THEREBY DElNES A CUT IN '
4HE EXPECTED WEIGHT OF THIS CUT IS
8
#IJ 0 ( SEPARATES UI AND UJ
IJ
"UT IT IS TRIVIAL THAT THE PROBABILITY THAT ( SEPARATES UI AND UJ IS x

TIMES THE ANGLE BETWEEN UI AND UJ )N TURN THE ANGLE BETWEEN UI AND UJ
IS AT LEAST UI _ UJ WHICH IS EASILY SEEN BY ELEMENTARY CALCULUS
(ENCE THE EXPECTED SIZE OF THE CUT OBTAINED IS AT LEAST
8
UI _ UJ - b t -
IJ
4HUS WE GET A CUT WHICH IS AT LEAST ABOUT PERCENT OF THE OPTIMAL
%XCEPT FOR THE LAST PART THIS TECHNIQUE IS ENTIRELY GENERAL AND HAS
BEEN USED IN A NUMBER OF PROOFS AND ALGORITHMS IN COMBINATORIAL OPTI
MIZATION ;,O3=
$ISCREPANCY 4HEORY
)N SECTION WE STARTED WITH A DISCRETE PROBLEM AND TRIED TO lND A GOOD
CONTINUOUS APPROXIMATION RELAXATION OF IT ,ET US DISCUSS A REVERSE PROB
LEM A BIT ONLY TO THE EXTENT OF A COUPLE OF CLASSICAL EXAMPLES 3UPPOSE
THAT WE HAVE A SET WITH A MEASURE HOW WELL CAN IT BE APPROXIMATED BY A
DISCRETE MEASURE
)N OUR lRST EXAMPLE WE CONSIDER THE ,EBESGUE MEASURE t ON THE UNIT
SQUARE ; = 3UPPOSE THAT WE WANT TO APPROXIMATE THIS BY THE UNIFORM
MEASURE ON A lNITE SUBSET 4 /F COURSE WE HAVE TO SPECIFY HOW THE ERROR
OF APPROXIMATION IS MEASURED HERE WE DElNE IT AS THE MAXIMUM ERROR ON
AXIS PARALLEL RECTANGLES
k k
a4 SUPkJ4 < 2J _ t2 J4 Jk
2
WHERE 2 RANGES OVER ALL SETS OF THE FORM 2 ;A B= a ;C D= WE SCALED UP BY
J4 J FOR CONVENIENCE 7E ARE INTERESTED IN lNDING THE BEST SET 4 AND IN
6ISIONS IN -ATH $)3#2%4% !.$ #/.4).5/53 47/ 3)$%3 /& 4(% 3!-%
DETERMINING THE <DISCREPANCY

aN INF a4
J4 JN
4HE QUESTION CAN BE RAISED FOR ANY FAMILY OF <TEST SETS INSTEAD OF RECT
ANGLES CIRCLES ELLIPSES TRIANGLES ETC 4HE ANSWER DEPENDS IN AN INTRICATE
WAY ON THE GEOMETRIC STRUCTURE OF THE TEST FAMILY SEE ;"E#= FOR AN EXPOSI
TION OF THESE BEAUTIFUL RESULTS
4HE QUESTION CAN ALSO BE RAISED IN OTHER DIMENSIONS 4HE DIMENSIONAL
CASE IS EASY SINCE THE OBVIOUS CHOICE OF 4 FN NN G IS
OPTIMAL "UT FOR DIMENSIONS LARGER THAN THE ORDER OF MAGNITUDE OF aN
IS NOT KNOWN
P P
2ETURNING TO DIMENSION THE OBVIOUS lRST CHOICE IS TO TRY A N a N
P
GRID FOR 4 4HIS LEAVES OUT A RECTANGLE OF SIZE ABOUT N AND SO IT
P
HAS a4 x N 4HERE ARE MANY CONSTRUCTIONS THAT DO BETTER THE MOST
ELEMENTARY IS THE FOLLOWING ,ET N K 4AKE ALL POINTS X Y WHERE BOTH
X AND Y ARE MULTIPLES OF _K AND EXPANDING THEM TO K BITS IN BINARY WE
GET THESE BITS IN REVERSE ORDER X B B BK AND Y BN BN_ B )T
IS A NICE EXERCISE TO VERIFY THAT THIS SET HAS DISCREPANCY a4 x K LOG N
)T IS HARD TO PROVE THAT ONE CANNOT DO BETTER EVEN TO PROVE THAT aN
WAS DInCULT ;!= 4HE LOWER BOUND MATCHING THE ABOVE CONSTRUCTION WAS
lNALLY PROVED BY 3CHMIDT ;3=
aN bLOG N
4HIS FUNDAMENTAL RESULT HAS MANY APPLICATIONS
/UR SECOND EXAMPLE CAN BE INTRODUCED THROUGH A STATISTICAL MOTIVATION
3UPPOSE THAT WE ARE GIVEN A _ SEQUENCE X X A XN OF LENGTH N 7E
WANT TO TEST WHETHER IT COMES FROM INDEPENDENT COIN mIPS
/NE APPROACH RELATED TO VON -ISESS PROPOSAL FOR THE DElNITION OF A RAN
DOM SEQUENCE COULD BE TO COUNT S AND S THEIR NUMBERS SHOULD BE ABOUT
THE SAME 4HIS SHOULD ALSO BE TRUE IF WE COUNT BITS IN THE EVEN POSITIONS
OR ODD POSITIONS OR MORE GENERALLY IN ANY lXED ARITHMETIC PROGRESSION OF
INDICES
3O WE CONSIDER THE QUANTITY
k8 k
k k
aX MAX k XI _ J!Jk
!
I!
WHERE ! RANGES THROUGH ALL ARITHMETIC PROGRESSIONS IN F NG %LEMEN
TARY PROBABILITY THEORY TELLS US THAT IF X IS GENERATED BY INDEPENDENT COIN
P
mIPS THEN aX x N WITH HIGH PROBABILITY 3O IF aX IS LARGER THAN
s
, ,/6!3: '!&!
THIS WHICH IS THE CASE OF MOST NON RANDOM SEQUENCES ONE THINKS OF THEN
WE CAN CONCLUDE THAT IT IS NOT GENERATED BY INDEPENDENT COIN mIPS
"UT CAN X FAIL THIS TEST IN THE OTHER DIRECTION BEING <TOO SMOOTH )N
OTHER WORDS WHAT IS
aN MIN aX
X
WHERE X RANGES OVER ALL _ SEQUENCES OF LENGTH N 7E CAN THINK OF THE
QUESTION AS A PROBLEM OF MEASURE APPROXIMATION WE ARE GIVEN THE MEA
SURE ON F NG IN WHICH EACH ATOM HAS MEASURE 4HIS IS A DISCRETE
MEASURE BUT NOT INTEGRAL VALUED WE WANT TO APPROXIMATE IT BY AN INTE
GRAL VALUED MEASURE 4HE TEST FAMILY DElNING THE ERROR OF APPROXIMATION
CONSISTS OF ARITHMETIC PROGRESSIONS
P
)T FOLLOWS FROM CONSIDERING RANDOM SEQUENCES THAT aN / N 4HE
lRST RESULT IN THE OPPOSITE DIRECTION WAS PROVED BY 2OTH ;2OT= WHO SHOWED
THAT aN jN )T WAS EXPECTED THAT THE BOUND WOULD BE IMPROVED TO
P
N EVENTUALLY SHOWING THAT RANDOM SEQUENCES ARE THE EXTREME AT LEAST IN
THE ORDER OF MAGNITUDE IN MANY SIMILAR INSTANCES OF COMBINATORIAL EXTREMAL
PROBLEMS FOR EXAMPLE IN 2AMSEY 4HEORY RANDOM CHOICE IS THE BEST "UT
SURPRISINGLY 3s ARKOZY SHOWED THAT aN /N WHICH WAS IMPROVED BY
"ECK ;"E= TO THE ALMOST SHARP aN /N LOG N AND EVEN THE LOGARITHMIC
FACTOR WAS RECENTLY REMOVED BY -ATOUtSEK AND 3PENCER ;-3= 4HUS THERE
ARE SEQUENCES WHICH SIMULATE RANDOM SEQUENCES TOO WELL
4HE ,APLACIAN
4HE ,APLACIAN AS WE LEARN IT IN SCHOOL IS THE DIkERENTIAL OPERATOR
8
X

I
I
7HAT COULD BE MORE TIED TO THE CONTINUITY OF EUCLIDEAN SPACES OR AT LEAST OF
DIkERENTIAL MANIFOLDS THAN A DIkERENTIAL OPERATOR )N THIS SECTION WE SHOW
THAT THE ,APLACIAN MAKES SENSE IN GRAPH THEORY AND IN FACT IT IS A BASIC
TOOL -OREOVER THE STUDY OF THE DISCRETE AND CONTINUOUS VERSIONS INTERACT
IN A VARIETY OF WAYS SO THAT THE USE OF ONE OR THE OTHER IS ALMOST A MATTER
OF CONVENIENCE IN SOME CASES
2ANDOM WALKS /NE OF THE FUNDAMENTAL GENERAL ALGORITHMIC PROB
LEMS IS SAMPLING GENERATE A RANDOM ELEMENT FROM A GIVEN DISTRIBUTION OVER
A SET 6 )N NON TRIVIAL CASES THE SET 6 IS EITHER INlNITE OR lNITE BUT VERY
LARGE AND OFTEN ONLY IMPLICITLY DESCRIBED )N MOST CASES WE WANT TO GEN
ERATE AN ELEMENT FROM THE UNIFORM DISTRIBUTION AND THIS SPECIAL CASE WILL
6ISIONS IN -ATH $)3#2%4% !.$ #/.4).5/53 47/ 3)$%3 /& 4(% 3!-%
BE ENOUGH FOR THE PURPOSE OF OUR DISCUSSIONS

)LL USE AS EXAMPLES TWO SAMPLING TASKS
A 'IVEN A CONVEX BODY + IN 2N GENERATE A UNIFORMLY DIS
TRIBUTED RANDOM POINT OF +
B 'IVEN A GRAPH ' GENERATE A UNIFORMLY DISTRIBUTED RANDOM
PERFECT MATCHING IN '
0ROBLEM A COMES UP IN MANY APPLICATIONS VOLUME COMPUTA
TION -ONTE #ARLO INTEGRATION OPTIMIZATION ETC 0ROBLEM B
IS RELATED TO THE )SING MODEL IN STATISTICAL MECHANICS
)N SIMPLE CASES ONE MAY lND ELEGANT SPECIAL METHODS FOR SAMPLING &OR
EXAMPLE IF WE NEED A UNIFORMLY DISTRIBUTED RANDOM POINT IN THE UNIT BALL
IN 2N THEN WE CAN GENERATE N INDEPENDENT COORDINATES FROM THE STANDARD
NORMAL DISTRIBUTION AND <NORMALIZE THE OBTAINED VECTOR APPROPRIATELY
"UT IN GENERAL SAMPLING FROM A LARGE lNITE SET OR AN INlNITE SET IS
A DInCULT PROBLEM AND THE ONLY GENERAL APPROACH KNOWN IS THE USE OF
-ARKOV CHAINS 7E DElNE USING THE STRUCTURE OF 3 AN ERGODIC -ARKOV
CHAIN WHOSE STATE SPACE IS 3 AND WHOSE STATIONARY DISTRIBUTION IS UNIFORM
)N THE CASE OF A lNITE SET 3 IT MAY BE EASIER TO THINK OF THIS AS A CONNECTED
REGULAR NON BIPARTITE GRAPH ' WITH NODE SET 6 3TARTING AT AN ARBITRARY
NODE STATE WE TAKE A RANDOM WALK ON THE GRAPH FROM A NODE I WE STEP
TO A NEXT NODE SELECTED UNIFORMLY FROM THE SET OF NEIGHBORS OF I !FTER A
SUnCIENTLY LARGE NUMBER 4 OF STEPS WE STOP THE DISTRIBUTION OF THE LAST
NODE IS APPROXIMATELY UNIFORM
)N THE CASE OF A CONVEX BODY A RATHER NATURAL -ARKOV CHAIN TO
CONSIDER IS THE FOLLOWING STARTING AT A CONVENIENT POINT IN + WE
MOVE AT EACH STEP TO A UNIFORMLY SELECTED RANDOM POINT IN A BALL
OF RADIUS m ABOUT THE CURRENT POINT IF THE NEW POINT IS OUTSIDE
+ WE STAY WHERE WE WERE AND CONSIDER THE STEP <WASTED
4HE STEP SIZE m WILL BE CHOSEN APPROPRIATELY BUT TYPICALLY IT IS
P
ABOUT N
)F WE WANT TO SAMPLE FROM PERFECT MATCHINGS IN A GRAPH '
THE RANDOM WALK TO CONSIDER IS NOT SO OBVIOUS *ERRUM AND
3INCLAIR CONSIDER A RANDOM WALK ON PERFECT AND NEAR PERFECT
MATCHINGS MATCHINGS THAT LEAVE JUST TWO NODES UNMATCHED )F
WE GENERATE SUCH A MATCHING IT WILL BE PERFECT WITH A SMALL BUT
NON NEGLIGIBLE PROBABILITY NCONST IF THE GRAPH IS DENSE 3O
WE JUST ITERATE UNTIL WE GET A PERFECT MATCHING
s
, ,/6!3: '!&!
/NE STEP OF THE RANDOM WALK IS GENERATED BY PICKING A RANDOM

EDGE E OF ' )F THE CURRENT MATCHING IS PERFECT AND CONTAINS E
THEN WE DELETE E IF THE CURRENT MATCHING IS NEAR PERFECT AND E
CONNECTS THE TWO UNMATCHED NODES WE ADD IT IF E CONNECTS AN
UNMATCHED NODE TO A MATCHED NODE WE ADD E TO THE MATCHING
AND DELETE THE EDGE IT INTERSECTS
!NOTHER WAY OF LOOKING AT THIS IS DO A RANDOM WALK ON A <BIG
GRAPH ( WHOSE NODES ARE THE PERFECT AND NEAR PERFECT MATCH
INGS OF THE <SMALL GRAPH ' 4WO NODES OF ( ARE ADJACENT IF
THE CORRESPONDING MATCHINGS ARISE FROM EACH OTHER BY CHANGING
DELETING ADDING OR REPLACING A SINGLE EDGE ,OOPS ARE ADDED
TO MAKE ( REGULAR .OTE THAT WE DO NOT WANT TO CONSTRUCT (
EXPLICITLY ITS SIZE MAY BE EXPONENTIALLY LARGE COMPARED WITH '
4HE POINT IS THAT THE RANDOM WALK ON ( CAN BE IMPLEMENTED
USING ONLY THIS IMPLICIT DElNITION
)T IS GENERALLY NOT HARD TO ACHIEVE THAT THE STATIONARY DISTRIBUTION OF

THE CHAIN IS UNIFORM AND THAT THE CHAIN IS ERGODIC 4HE CRUCIAL QUESTION
IS WHAT IS THE MIXING TIME IE HOW LONG DO WE HAVE TO WALK 4HIS
QUESTION LEADS TO ESTIMATING THE MIXING TIME OF -ARKOV CHAINS THE NUMBER
OF STEPS BEFORE THE CHAIN BECOMES ESSENTIALLY STATIONARY &ROM THE POINT
OF VIEW OF PRACTICAL APPLICATIONS IT IS NATURAL TO CONSIDER lNITE -ARKOV
CHAINS [ A COMPUTATION IN A COMPUTER IS NECESSARILY lNITE "UT IN THE
ANALYSIS IT DEPENDS ON THE PARTICULAR APPLICATION WHETHER ONE PREFERS TO USE
A lNITE OR A GENERAL MEASURABLE STATE SPACE !LL THE ESSENTIAL AND VERY
INTERESTING CONNECTIONS THAT HAVE BEEN DISCOVERED HOLD IN BOTH MODELS )N
FACT THE GENERAL MATHEMATICAL ISSUE IS DISPERSION WE MIGHT BE INTERESTED
IN DISPERSION OF HEAT IN A MATERIAL OR DISPERSION OF PROBABILITY DURING A
RANDOM WALK OR MANY OTHER RELATED QUESTIONS
! -ARKOV CHAIN CAN BE DESCRIBED BY ITS TRANSITION MATRIX - PIJ
WHERE PIJ IS THE PROBABILITY OF STEPPING TO J GIVEN THAT WE ARE AT I )N
THE SPECIAL CASE OF RANDOM WALKS ON A REGULAR UNDIRECTED GRAPH WE HAVE
- D ! WHERE ! IS THE USUAL ADJACENCY MATRIX OF THE GRAPH ' .OTE
THAT THE ALL VECTOR IS AN EIGENVECTOR OF - BELONGING TO THE EIGENVALUE
)F z IS THE STARTING DISTRIBUTION WHICH CAN BE VIEWED AS A VECTOR IN 26
THEN THE DISTRIBUTION AFTER T STEPS IS - T z &ROM THIS IT IS EASY TO SEE THAT THE
SPEED OF CONVERGENCE TO THE STATIONARY DISTRIBUTION WHICH IS THE EIGENVECTOR
N DEPENDS ON THE DIkERENCE BETWEEN THE LARGEST EIGENVALUE AND
THE SECOND LARGEST EIGENVALUE t THE SPECTRAL GAP 7E COULD ALSO DElNE THE
6ISIONS IN -ATH $)3#2%4% !.$ #/.4).5/53 47/ 3)$%3 /& 4(% 3!-%
SPECTRAL GAP AS THE SMALLEST POSITIVE EIGENVALUE OF THE MATRIX , - _ )

7E CALL THIS MATRIX , THE ,APLACIAN OF THE GRAPH &OR ANY VECTOR
F 26 THE VALUE OF ,F AT NODE I IS THE AVERAGE OF F OVER THE NEIGH
BORS OF I MINUS FI 4HIS PROPERTY SHOWS THAT THE ,APLACIAN IS INDEED A
DISCRETE ANALOG OF THE CLASSICAL ,APLACE OPERATOR
3OME PROPERTIES OF THE ,APLACE OPERATOR DEPEND ON THE lNE STRUCTURE
OF DIkERENTIABLE MANIFOLDS BUT MANY BASIC PROPERTIES CAN BE GENERALIZED
EASILY TO ANY GRAPH /NE CAN DElNE HARMONIC FUNCTIONS THE HEAT KERNEL
PROVE IDENTITIES LIKE THE ANALOGUE 'REENS FORMULA
8_ `
FI ,G I _ GI ,F I
I
AND SO ON -ANY PROPERTIES OF THE <CONTINUOUS ,APLACE OPERATOR CAN BE
DERIVED USING SUCH EASY COMBINATORIAL FORMULAS ON A GRID AND THEN TAKING
LIMITS 3EE #HUNG ;#HU= FOR AN EXPOSITION OF SOME OF THESE CONNECTIONS
"UT MORE SIGNIlCANTLY THE DISCRETE ,APLACIAN IS AN IMPORTANT TOOL IN
THE STUDY OF VARIOUS GRAPH THEORETIC PROPERTIES !S A lRST ILLUSTRATION OF
THIS FACT LET US RETURN TO RANDOM WALKS
/FTEN INFORMATION ABOUT THE SPECTRAL GAP IS DInCULT TO OBTAIN THIS IS
THE CASE IN BOTH OF OUR INTRODUCTORY EXAMPLES /NE POSSIBLE REMEDY IS TO
RELATE THE DISPERSION SPEED TO ISOPERIMETRIC INEQUALITIES IN THE STATE SPACE
4O BE MORE EXACT DElNE THE CONDUCTANCE OF THE CHAIN AS
8 xI PIJ
h MAX
y3y6 x3 x6 N 3
I3J6 N3
4HE NUMERATOR CAN BE VIEWED AS THE PROBABILITY THAT CHOOSING A RANDOM

NODE FROM THE STATIONARY DISTRIBUTION AND THEN MAKING ONE STEP THE lRST
NODE IS IN 3 AND THE SECOND IS IN 6 N3 4HE DENOMINATOR IS THE SAME PROBA
BILITY FOR TWO INDEPENDENT RANDOM NODES FROM THE STATIONARY DISTRIBUTION
4HE FOLLOWING INEQUALITY WAS PROVED BY *ERRUM AND 3INCLAIR ;*3=
4HEOREM
h
s _ t s h
4HIS MEANS THAT THE MIXING TIME WHICH WE HAVE TO USE INFORMALLY
HERE SINCE THE EXACT DElNITION DEPENDS ON HOW WE START HOW WE MEASURE
CONVERGENCE ETC IS BETWEEN h AND h
)N THE CASE OF RANDOM WALKS IN A CONVEX BODY THE CONDUCTANCE
IS VERY CLOSELY RELATED TO THE FOLLOWING ISOPERIMETRIC THEOREM
;,O3I= ;$Y&=
s
, ,/6!3: '!&!
4HEOREM ,ET + BE A CONVEX BODY IN 2N OF DIAMETER $

,ET THE SURFACE & DIVIDE + INTO TWO PARTS + AND + 4HEN
VOL+ VOL+
VOLN_ & t
$ VOL+
! LONG THIN CYLINDER SHOWS THAT THE BOUND IS SHARP
&ROM 4HEOREM IT FOLLOWS THAT AFTER APPROPRIATE PREPROCESS
ING AND OTHER TECHNICAL DInCULTIES WHICH ARE NOW SWEPT UNDER
THE CARPET ONE CAN GENERATE A SAMPLE POINT IN AN N DIMENSIONAL
CONVEX BODY IN /b N STEPS
)N THE CASE OF MATCHINGS *ERRUM AND 3INCLAIR PROVE THAT FOR
DENSE GRAPHS THE CONDUCTANCE OF THE CHAIN DESCRIBED IS BOUNDED
FROM BELOW BY NCONST AND HENCE THE MIXING TIME IS BOUNDED
BY NCONST
(OW TO ESTABLISH ISOPERIMETRIC INEQUALITIES 4HE GENERIC METHOD IS TO

CONSTRUCT EXPLICITLY OR IMPLICITLY mOWS OR MULTICOMMODITY mOWS 3UPPOSE
THAT FOR EACH SUBSET y 3 y 6 WE CAN CONSTRUCT A mOW THROUGH THE GRAPH
SO THAT EACH NODE I 3 IS A SOURCE THAT PRODUCES AN AMOUNT OF xI x6 N 3
OF mOW AND EACH J 6 N 3 IS A SINK THAT CONSUMES xJ x3 AMOUNT OF mOW
3UPPOSE FURTHER THAT EACH EDGE IJ CARRIES AT MOST +xI PIJ mOW 4HEN A
SIMPLE COMPUTATION SHOWS THAT THE CONDUCTANCE SATISlES

ht +
)NSTEAD OF CONSTRUCTING A mOW FOR EACH SUBSET 3 OF NODES ONE MIGHT PREFER
TO CONSTRUCT A MULTICOMMODITY mOW IE A mOW OF VALUE xI xJ FOR EACH I
AND J )F THE TOTAL mOW THROUGH EACH EDGE IJ IS AT MOST +xI PIJ THEN THE
CONDUCTANCE IS AT LEAST +
(OW GOOD IS THE BOUND ON THE CONDUCTANCE OBTAINED BY THE MULTICOM
MODITY mOW METHOD !N IMPORTANT RESULT OF ,EIGHTON AND 2AO ;,2= IMPLIES
THAT FOR THE BEST CHOICE OF THE mOWS IT GETS WITHIN A FACTOR OF /LOG N /NE
OF THE REASONS ) BRING THIS UP IS THAT THIS RESULT IN TURN CAN BE DERIVED
FROM THE THEOREM OF "OURGAIN ;"O= ABOUT THE EMBEDABILITY OF lNITE METRIC
SPACES IN @ SPACES WITH SMALL DISTORTION #F ALSO ;,I,2= ;$E,=
4O CONSTRUCT THE <BEST mOWS OR MULTICOMMODITY mOWS IS OFTEN

THE MAIN MATHEMATICAL DInCULTY )N THE CASE OF THE PROOF OF
4HEOREM IT CAN BE DONE BY A METHOD USED EARLIER BY 0AYNE
AND 7EINBERGER ;07= IN THE THEORY OF PARTIAL DIkERENTIAL EQUA
TIONS 7E ONLY GIVE A ROUGH SKETCH
6ISIONS IN -ATH $)3#2%4% !.$ #/.4).5/53 47/ 3)$%3 /& 4(% 3!-%
3UPPOSE THAT WE WANT TO CONSTRUCT A <mOW FROM + TO + "Y

THE <(AM[3ANDWICH THEOREM THERE EXISTS A HYPERPLANE THAT
CUTS BOTH + AND + INTO TWO PARTS WITH EQUAL VOLUME 7E CAN
SEPARATELY CONSTRUCT THE mOWS ON BOTH SIDE OF THIS HYPERPLANE
2EPEATING THIS PROCEDURE WE CAN CUT UP + INTO <NEEDLES SO
THAT EACH NEEDLE IS SPLIT BY THE PARTITION 3 6 N 3 IN THE SAME
PROPORTION AND HENCE IT SUnCES TO CONSTRUCT A mOW BETWEEN
+ AND + INSIDE EACH NEEDLE 4HIS IS EASILY DONE USING THE
"RUN[-INKOWSKI THEOREM
&OR THE RANDOM WALK ON MATCHINGS THE CONSTRUCTION OF THE MUL
TICOMMODITY mOW CALLED CANONICAL PATHS BY *ERRUM AND 3IN
CLAIR ALSO GOES BACK TO ONE OF THE OLDEST METHODS IN MATCHING
THEORY 'IVEN SAY A PERFECT MATCHING - AND A NEAR PERFECT
MATCHING - WE FORM THE UNION 4HIS IS A SUBGRAPH OF THE
<SMALL GRAPH ' THAT DECOMPOSES INTO A PATH WITH EDGES AL
TERNATING BETWEEN - AND - SOME CYCLES AGAIN ALTERNATING
BETWEEN - AND - AND THE COMMON EDGES OF - AND - .OW
IT IS EASY TO TRANSFORM - INTO - BY WALKING ALONG EACH CYCLE
AND THE PATH AND REPLACING THE EDGES ONE BY ONE 7HAT THIS
AMOUNTS TO IS A PATH IN THE <BIG GRAPH ( CONNECTING NODES
- AND - )F WE USE THIS PATH TO CARRY THE mOW THEN IT CAN BE
SHOWN THAT NO EDGE OF THE GRAPH ( IS OVERLOADED PROVIDED THE
GRAPH ' IS DENSE
4HE #AGE THEOREM AND CONFORMAL MAPPINGS )T WAS PROVED
BY 3TEINITZ THAT EVERY CONNECTED PLANAR GRAPH CAN BE REPRESENTED AS THE
SKELETON OF A DIMENSIONAL POLYTOPE )N FACT THERE IS A LOT OF FREEDOM
IN CHOOSING THE GEOMETRY OF THIS REPRESENTING POLYTOPE !MONG VARIOUS
EXTENSIONS THE MOST INTERESTING FOR US IS THE CLASSICAL CONSTRUCTION GOING
BACK TO +OEBE ;+O= lRST PROVED BY !NDREEV ;!N= CF ALSO ;4=
4HEOREM 4HE #AGE THEOREM ,ET ( BE A CONNECTED PLANAR GRAPH
4HEN ( CAN BE REPRESENTED AS THE SKELETON OF A DIMENSIONAL POLYTOPE ALL
WHOSE EDGES TOUCH THE UNIT SPHERE
7E MAY ADD THAT THE REPRESENTING POLYTOPE IS UNIQUE UP TO A PROJECTIVE
TRANSFORMATION OF THE SPACE THAT PRESERVES THE UNIT SPHERE "Y CONSIDERING
THE <HORIZON FROM EACH VERTEX OF THE POLYTOPE WE OBTAIN A REPRESENTATION
OF THE NODES OF THE GRAPH BY OPENLY DISJOINT CIRCULAR DISKS IN THE PLANE SO
THAT ADJACENT NODES CORRESPOND TO TOUCHING CIRCLES
4HE #AGE THEOREM MAY BE CONSIDERED AS A DISCRETE FORM OF THE 2IEMANN
s
, ,/6!3: '!&!
MAPPING THEOREM IN THE SENSE THAT IT IMPLIES THE 2IEMANN MAPPING THE
OREM )NDEED SUPPOSE THAT WE WANT TO CONSTRUCT A CONFORMAL MAPPING OF
A SIMPLY CONNECTED DOMAIN + IN THE COMPLEX PLANE ONTO THE UNIT DISK $
&OR SIMPLICITY ASSUME THAT + IS BOUNDED #ONSIDER A TRIANGULAR GRID IN
THE PLANE AN INlNITE REGULAR GRAPH AND LET ' BE THE GRAPH OBTAINED
BY IDENTIFYING ALL GRIDPOINTS OUTSIDE + INTO A SINGLE NODE S #ONSIDER THE
+OEBE REPRESENTATION BY TOUCHING CIRCLES ON THE SPHERE WE MAY ASSUME
THAT THE CIRCLE REPRESENTING THE NODE S IS THE EXTERIOR OF $ 3O ALL THE OTHER
CIRCLES ARE INSIDE $ AND WE OBTAIN A MAPPING OF THE SET OF GRIDPOINTS IN
+ INTO $ "Y LETTING THE GRID BECOME ARBITRARILY lNE IT WAS SHOWN BY
2ODIN AND 3ULLIVAN ;23= THAT IN THE LIMIT WE GET A CONFORMAL MAPPING OF
+ ONTO $
3O THE #AGE THEOREM IS INDEED A BEAUTIFUL BRIDGE BETWEEN DISCRETE MATH
EMATICS GRAPH THEORY AND CONTINUOUS MATHEMATICS COMPLEX ANALYSIS
"UT WHERE IS THE ,APLACIAN 7ELL SEE ONE CONNECTION IN THE NEXT SECTION
!NOTHER ONE OCCURS IN THE WORK OF 3PIELMANN AND 4ENG ;3P4= WHO USE THE
#AGE THEOREM TO SHOW THAT THE EIGENVALUE GAP OF THE ,APLACIAN OF A PLANAR
GRAPH IS AT MOST N 4HIS IMPLIES AMONG OTHERS THAT THE MIXING TIME
OF THE RANDOM WALK ON A PLANAR GRAPH IS AT LEAST LINEAR IN THE NUMBER OF
NODES
#OLIN DE 6ERDIr ERES INVARIANT )N #OLIN DE 6ERDIrERE ;#O=
INTRODUCED A PARAMETER u' FOR ANY UNDIRECTED GRAPH ' 2ESEARCH CON
CERNING THIS PARAMETER INVOLVES AN INTERESTING MIXTURE OF IDEAS FROM GRAPH
THEORY LINEAR ALGEBRA AND ANALYSIS
4HE EXACT DElNITION GOES BEYOND THE LIMITATIONS OF THIS ARTICLE ROUGHLY
SPEAKING u' IS THE MULTIPLICITY OF THE SMALLEST POSITIVE EIGENVALUE OF THE
,APLACIAN OF ' WHERE THE EDGES OF ' ARE WEIGHTED SO AS TO MAXIMIZE THIS
MULTIPLICITY
4HE PARAMETER WAS MOTIVATED BY THE STUDY OF THE MAXIMUM MULTIPLICITY
OF THE SECOND EIGENVALUE OF CERTAIN ,APLACIAN TYPE DIkERENTIAL OPERATORS
DElNED ON 2IEMANN SURFACES (E APPROXIMATED THE SURFACE BY A SUnCIENTLY
DENSELY EMBEDDED GRAPH ' AND SHOWED THAT THE MULTIPLICITY OF THE SECOND
EIGENVALUE OF THE OPERATOR CAN BE BOUNDED BY THIS VALUE u' DEPENDING
ONLY ON THE GRAPH
#OLIN DE 6ERDIrERES INVARIANT CREATED MUCH INTEREST AMONG GRAPH THE
ORISTS BECAUSE OF ITS SURPRISINGLY NICE GRAPH THEORETIC PROPERTIES !MONG
OTHERS IT IS MINOR MONOTONE SO THAT THE 2OBERTSON[3EYMOUR GRAPH MINOR
THEORY APPLIES TO IT -OREOVER PLANARITY OF GRAPHS CAN BE CHARACTERIZED BY
6ISIONS IN -ATH $)3#2%4% !.$ #/.4).5/53 47/ 3)$%3 /& 4(% 3!-%
THIS INVARIANT u' s IF AND ONLY IF ' IS PLANAR

#OLIN DE 6ERDIrERES ORIGINAL PROOF OF THE <IF PART OF THIS FACT WAS
MOST UNUSUAL IN GRAPH THEORY BASICALLY REVERSING THE ABOVE PROCEDURE
HE SHOWED HOW TO RECONSTRUCT A SPHERE AND A POSITIVE ELLIPTIC PARTIAL DIF
FERENTIAL OPERATOR 0 ON IT SO THAT u' IS BOUNDED BY THE DIMENSION OF THE
NULL SPACE OF 0 AND THEN INVOKED A THEOREM OF #HENG ;#= ASSERTING THAT
THIS DIMENSION IS AT MOST
,ATER VAN DER (OLST ;(O= FOUND A COMBINATORIAL PROOF OF THIS FACT 7HILE
THIS MAY SEEM AS A STEP BACKWARD AFTER ALL IT ELIMINATED THE NECESSITY OF
THE ONLY APPLICATION OF PARTIAL DIkERENTIAL EQUATIONS IN GRAPH THEORY ) KNOW
OF IT DID OPEN UP THE POSSIBILITY OF CHARACTERIZING THE NEXT CASE 6ERIFYING
A CONJECTURE OF 2OBERTSON 3EYMOUR AND 4HOMAS IT WAS SHOWN BY ,OVsASZ
AND 3CHRIJVER ;,O3= THAT u' s IF AND ONLY IF ' IS LINKLESSLY EMBEDABLE
IN 2
#AN ONE GO BACK TO THE ORIGINAL MOTIVATION FROM u' AND lND A <CON
TINUOUS VERSION OF THIS RESULT )N WHAT SENSE DOES A LINKLESSLY EMBEDDED
GRAPH APPROXIMATE THE SPACE 4HESE QUESTIONS APPEAR VERY DInCULT
)T TURNS OUT THAT GRAPHS WITH LARGE VALUES OF u ARE ALSO QUITE INTERESTING
&OR EXAMPLE FOR A GRAPH ' ON N NODES WITH u' t N _ THE COMPLEMENT
OF ' IS PLANAR UP TO INTRODUCING <TWIN POINTS AND THE CONVERSE OF THIS
ASSERTION ALSO HOLDS UNDER REASONABLY GENERAL CONDITIONS 4HE PROOF OF THE
LATTER FACT USES THE +OEBE[!NDREEV REPRESENTATION OF GRAPHS
3O THE GRAPH INVARIANT u' IS RELATED AT ONE END OF THE SCALE TO ELLIPTIC
PARTIAL DIkERENTIAL EQUATIONS #HENGS THEOREM ON THE OTHER TO 2IEMANNS
THEOREM ON CONFORMAL MAPPINGS
2EFERENCES
;!= 4 VAN !ARDENNE %HRENFEST /N THE IMPOSSIBILITY OF A JUST DISTRI
BUTION .EDERL !KAD 7ETENSCH 0ROC [ )NDAGATIONES
-ATH [
;!L= & !LIZADEH )NTERIOR POINT METHODS IN SEMIDElNITE PROGRAMMING WITH
APPLICATIONS TO COMBINATORIAL OPTIMIZATION 3)!- * ON /PTIMIZATION
[
;!LO3= . !LON * 3PENCER 4HE 0ROBABILISTIC -ETHOD 7ITH AN APPENDIX BY
0AUL %RD]OS 7ILEY .EW 9ORK
;!N= % !NDREEV /N CONVEX POLYHEDRA IN ,OBACHEVSKY SPACES -AT 3BORNIK
.OV 3ER [
;!R,-33= 3 !RORA # ,UND 2 -OTWANI - 3UDAN - 3ZEGEDY 0ROOF
VERIlCATION AND HARDNESS OF APPROXIMATION PROBLEMS 0ROC RD &/#3
s
, ,/6!3: '!&!
[
;"= s RA
) "A s NY ! SHORT PROOF OF +NESERS CONJECTURE * #OMBIN 4HEORY !
[
;"E= * "ECK 2OTHS ESTIMATE OF THE DISCREPANCY OF INTEGER SEQUENCES IS NEARLY
SHARP #OMBINATORICA [
;"E#= * "ECK 7 #HEN )RREGULARITIES OF $ISTRIBUTION #AMBRIDGE 5NIV 0RESS

;"E3= * "ECK 64 3O s S $ISCREPANCY 4HEORY #HAPTER IN <(ANDBOOK OF
#OMBINATORICS 2, 'RAHAM - 'R OTSCHEL , ,OVs ASZ EDS .ORTH
(OLLAND !MSTERDAM
;"J= ! "JORNER 4OPOLOGICAL METHODS IN <(ANDBOOK OF #OMBINATORICS
2, 'RAHAM , ,OVs ASZ - 'ROTSCHEL EDS %LSEVIER !MSTERDAM
[
;"O= * "OURGAIN /N ,IPSCHITZ EMBEDDING OF lNITE METRIC SPACES IN (ILBERT
SPACE )SR * -ATH [
;#= 39 #HENG %IGENFUNCTIONS AND NODAL SETS #OMMENTARII -ATHEMATICI
(ELVETICI [
;#H= !* #HORIN 6ORTICITY AND 4URBULENCE 3PRINGER .EW 9ORK
;#O= 9 #OLIN DE 6ERDIr ERE 3UR UN NOUVEL INVARIANT DES GRAPHES ET UN CRITrERE
DE PLANARITsE *OURNAL OF #OMBINATORIAL 4HEORY 3ERIES " [
;#HU= &2+ #HUNG 3PECTRAL 'RAPH 4HEORY #"-3 2EG #ONF 3ERIES
!MER -ATH 3OC
;$0= # $ELORME 3 0OLJAK #OMBINATORIAL PROPERTIES AND THE COMPLEXITY
OF MAX CUT APPROXIMATIONS %UROP * #OMBIN [
;$E,= - $EZA - ,AURENT 'EOMETRY OF #UTS AND -ETRICS 3PRINGER 6ERLAG

;$U= 2- $UDLEY $ISTANCES OF PROBABILITY MEASURES AND RANDOM VARIABLES
!NN -ATH 3TAT [
;$Y&= - $YER ! &RIEZE #OMPUTING THE VOLUME OF CONVEX BODIES A CASE
WHERE RANDOMNESS PROVABLY HELPS IN<0ROBABILISTIC #OMBINATORICS AND )TS
!PPLICATIONS "sELA "OLLOBsAS ED 0ROCEEDINGS OF 3YMPOSIA IN !PPLIED
-ATHEMATICS 6OL [
;$Y&+= - $YER ! &RIEZE 2 +ANNAN ! RANDOM POLYNOMIAL TIME ALGO
RITHM FOR APPROXIMATING THE VOLUME OF CONVEX BODIES *OURNAL OF THE !#-
[
;'7= -8 'OEMANS $0 7ILLIAMSON !PPROXIMATION ALGORITHMS FOR
-!8 #54 AND -!8 3!4 0ROC TH !#- 3YMP ON 4HEORY OF #OM
PUTING
;'R,3= - 'RO TSCHEL , ,OVA sSZ ! 3CHRIJVER 'EOMETRIC !LGORITHMS AND
#OMBINATORIAL /PTIMIZATION 3PRINGER
;(= * (wASTAD 3OME OPTIMAL IN APPROXIMABILITY RESULTS 0ROC TH !#-
3YMP ON 4HEORY OF #OMP [
6ISIONS IN -ATH $)3#2%4% !.$ #/.4).5/53 47/ 3)$%3 /& 4(% 3!-%
;(O= ( VAN DER (OLST ! SHORT PROOF OF THE PLANARITY CHARACTERIZATION OF

#OLIN DE 6ERDIrERE *OURNAL OF #OMBINATORIAL 4HEORY 3ERIES "
[
;*3= - *ERRUM ! 3INCLAIR !PPROXIMATING THE PERMANENT 3)!- * #OM
PUTING
;+= - +NESER !UFGABE .O *BER $EUTSCH -ATH 6ER
;+O= 0 +OEBE +ONTAKTPROBLEME DER KONFORMEN !BBILDUNG "ERICHTE UBER DIE
6ERHANDLUNGEN D 3ACHS !KAD D 7ISS -ATH[0HYS +LASSE
[
;,2= &4 ,EIGHTON 3 2AO !N APPROXIMATE MAX mOW MIN CUT THEOREM FOR
UNIFORM MULTICOMMODITY mOW PROBLEMS WITH APPLICATIONS TO APPROXIMA
TION ALGORITHMS 0ROC TH !NNUAL 3YMP ON &OUND OF #OMPUTER 3CIENCE
)%%% #OMPUTER 3OC
;,I,2= . ,INIAL % ,ONDON 9 2ABINOVICH 4HE GEOMETRY OF GRAPHS AND
SOME OF ITS ALGEBRAIC APPLICATIONS #OMBINATORICA [
;,O= , ,OVA sSZ +NESERS CONJECTURE CHROMATIC NUMBER AND HOMOTOPY *
#OMB 4HEORY !
;,O= , ,OVA sSZ /N THE 3HANNON CAPACITY OF GRAPHS )%%% 4RANS )NFORM
4HEORY [
;,O-= , ,OVA sSZ 0 -AJOR ! NOTE ON A PAPER OF $UDLEY 3TUDIA 3CI -ATH
(UNG [
;,O0= , ,OVA sSZ -$ 0LUMMER -ATCHING 4HEORY !KADsEMIAI +IADsO .ORTH
(OLLAND "UDAPEST
;,O3= , ,OVA sSZ ! 3CHRIJVER #ONES OF MATRICES AND SET FUNCTIONS AND
OPTIMIZATION 3)!- * ON /PTIMIZATION [
;,O3= , ,OVA sSZ ! 3CHRIJVER ! "ORSUK THEOREM FOR ANTIPODAL LINKS AND A
SPECTRAL CHARACTERIZATION OF LINKLESSLY EMBEDDABLE GRAPHS 0ROCEEDINGS OF
THE !-3 [
;,O3I= , ,OVA sSZ - 3IMONOVITS 2ANDOM WALKS IN A CONVEX BODY AND AN
IMPROVED VOLUME ALGORITHM 2ANDOM 3TRUCTURES AND !LG [

;-= * -ATOUt SEK 'EOMETRIC $ISCREPANCY 3PRINGER 6ERLAG
;-3= * -ATOUt SEK * 3PENCER $ISCREPANCY IN ARITHMETIC PROGRESSIONS *
!MER -ATH 3OC [
;/= $ /RNSTEIN "ERNOULLI SHIFTS WITH THE SAME ENTROPY ARE ISOMORPHIC !D
VANCES IN -ATH [
;07= ,% 0AYNE (& 7EINBERGER !N OPTIMAL 0OINCARsE INEQUALITY FOR CON
VEX DOMAINS !RCH 2AT MECH !NAL [
;23= " 2ODIN $ 3ULLIVAN 4HE CONVERGENCE OF CIRCLE PACKINGS TO THE 2IE
MANN MAPPING * $IkERENTIAL 'EOM [
;2O(= ' # 2OTA ,( (ARPER -ATCHING THEORY AN INTRODUCTION IN <!D
VANCES IN 0ROBABILITY 4HEORY AND 2ELATED 4OPICS 0 .EY ED 6OL )
s
, ,/6!3: '!&!
-ARCEL $EKKER .EW 9ORK [

;2OT= +& 2OTH 2EMARK CONCERNING INTEGER SEQUENCES !CTA !RITH
[
;3= 7 3CHMIDT )RREGULARITIES OF DISTRIBUTION 1UART * -ATH /XFORD
[
;3P4= $! 3PIELMAN 3 ( 4ENG 3PECTRAL PARTITIONING WORKS PLANAR GRAPHS
AND lNITE ELEMENT MESHES 0ROC TH !NN 3YMP &OUND OF #OMP 3CI
)%%% [
;3T= 6 3TRASSEN 4HE EXISTENCE OF PROBABILITY MEASURES WITH GIVEN MARGINALS
!NN -ATH 3TATIST [
;4= 7 4HURSTON 4HE 'EOMETRY AND 4OPOLOGY OF 4HREE MANIFOLDS 0RINCE
TON ,ECTURE .OTES #HAPTER 0RINCETON
;7= ( 7OLKOWITZ 3OME APPLICATIONS OF OPTIMIZATION IN MATRIX THEORY ,IN
EAR !LGEBRA AND ITS !PPLICATIONS [
sSZLO
,A s ,OVA
sSZ -ICROSOFT 2ESEARCH /NE -ICROSOFT 7AY 2EDMOND 7!
53! LOVASZ MICROSOFTCOM
094(!'/2%!. !.$ 0,!4/.)# #/.#%04)/.3 ).

884( #%.4529 0(93)#3
9UVAL .EEMAN
)NTRODUCTION 7HY 3TART IN 'REECE

0YTHAGOREANISM THE SYNCRETISTIC PHILOSOPHY EXPOUNDED BY
0YTHAGORAS CHIEmY DISTINGUISHED BY ITS DESCRIPTION OF REALITY
IN TERMS OF ARITHMETICAL RELATIONSHIPS ;@!MERICAN (ERITAGE DIC
TIONARY OF THE %NGLISH LANGUAGE=
<0LATO ALLEGES THAT 'OD FOREVER GEOMETRIZES ;0LUTARCH 1UAEST
0LAT=
)N A REEVALUATION OF THE MATHEMATICAL MOLD ADOPTED BY &UNDAMENTAL
0HYSICS AT THE END OF THE 88TH #ENTURY WE lND IT INSTRUCTIVE TO START
OUR RETROSPECTIVE WITH A SEARCH FOR <ROOTS IN 'REEK CULTURE IN THE 6) TH
#ENTURY "# [ THEN MAKE A LEAP FORWARD AND CONCENTRATE ON THE EVOLUTION
OF THE MATHEMATICAL FORMULATION THROUGHOUT THE 88TH #ENTURY ITSELF )N
WHAT FOLLOWS WE MOTIVATE THIS CHOICE
3CIENCE CONSISTS OF REPRESENTATIONS OF SECTORS OF REALITY EACH BASED UPON
A SMALL NUMBER OF POSTULATES FROM WHICH ALL OBSERVABLE QUALITATIVE AND
QUANTITATIVE FEATURES OF THAT SECTOR SHOULD BE DERIVABLE BY LOGICAL INFERENCE
%UCLIDS FORMATTING OF 'EOMETRY IN TERMS OF AXIOMS THEOREMS AND COROL
LARIES HAS PROVIDED US EVER SINCE WITH A MODEL OF WHAT ANY SCIENCE SHOULD
BE MADE TO LOOK LIKE
(ISTORICALLY THIS VIEW REPLACED RELIGIOUS OR MYTHOLOGICAL INTERPRETATIONS
OF REALITY )T THUS HAD TO PRE SUPPOSE THAT ANY IMPACT UPON PHYSICAL REALITY [
IF THERE BE ONE [ ORIGINATING IN HYPOTHETICAL RELIGIOUS OR ETHICAL DIMENSIONS
IS ALREADY INCLUDED WITHIN THE SAME FRAMEWORK CONSTRAINING REALITY AND
THUS DOES NOT DISTURB THE REQUIREMENT OF A GOOD lT WITH OBSERVATIONS AS
THE KEY CRITERION FOR A THEORYS VALIDITY )N OTHER WORDS [ FOR SCIENCE TO BE
OBSERVATIONALLY VERIlABLE <MIRACLES IF ANY ALSO HAVE TO OCCUR WITHIN THE
LAWS OF NATURE
!NOTHER ASSUMPTION IS THAT THE PATCHY NATURE OF THE SCIENTIlC COVERAGE
OF REALITY IS A TEMPORARY FEATURE WITH A GRADUAL STEP BY STEP UNIlCATION
9 .%%-!. '!&!
PROCESS COMING EVER CLOSER TO THE IDEAL OF A FULLY UNITED AND COMPLETE SCI
ENTIlC DESCRIPTION OF EXISTENCE %VEN AREAS SUCH AS ,IFE #ONSCIOUSNESS OR
#REATION WHICH AT THE BEGINNING OF THE 88TH #ENTURY TO MANY STILL AP
PEARED TO LIE BEYOND THE REACH OF SCIENCE HAVE SINCE YIELDED OR AT LEAST HAVE
BEGUN TO YIELD TO THE SCIENTIlC TREATMENT IN THE COURSE OF THIS CENTURY
"UT WHY START IN 'REECE 4RUE SCIENCE ORIENTED ACTIVITIES HAD INDEED
STARTED IN PREHISTORY AND DEVELOPED SOMETIMES INDEPENDENTLY IN MOST
EARLY CULTURES [ %GYPT 3UMER AND !KKAD )NDIA #HINA [ AND LATER IN
THOSE OF THE !MERICAN CONTINENT 4HIS GENERALLY INCLUDED FOR EXAMPLE THE
OBSERVATION OF THE HEAVENS AND THE SEPARATION OF THE OBJECTS OBSERVED IN
TWO CATEGORIES [ THE ROTATING TAPESTRY OF THOUSANDS OF <lXED STARS [ AND
SEVEN OBJECTS MOVING WITH RESPECT TO THAT FRAME OBJECTS STILL HONORED BY
THE SEVEN DAY WEEK AND HAVING ITS DAYS NAMED AFTER THEM
!ND YET IT IS ONLY IN 'REECE THAT WE CAN POINT TO SCIENCE IN THE SENSE WE
HAVE DElNED %VERYWHERE ELSE IT WAS EITHER AN IMPROVEMENT ON MAGIC OR
AT BEST AN EkORT TO ACHIEVE A BETTER UNDERSTANDING AS TO THE WAY THE 'ODS
WERE RUNNING @THEIR WORLD WITHOUT THE SEARCH FOR COMPLETENESS AND WITHOUT
THE IDEA THAT IT COULD ALL BE DESCRIBED AND EXPLAINED BY A LOGICAL INFERENCE
NETWORK !MONG THE 3EMITIC CULTURES OF THE -IDDLE %AST BOTH %ASTERN
!KKAD AND 7ESTERN #ANAANITE (EBREW #ODES OF ,AW EVOLVED DURING
THE 3ECOND -ILLENIUM "# (AMMURABIS IN !KKAD -OSES IN )SRAEL 7HEN
THIS MOVEMENT SPREAD TO 'REECE AT THE END OF THE 6))TH #ENTURY PRODUCING
3OLONS REFORMS AND HIS CODE OF LAWS ITS IMPACT INCLUDED SOMETHING THAT
HAD NOT HAPPENED ELSEWHERE NAMELY THE IDEA THAT <THE WORLD MIGHT ALSO
OBEY A CODE OF LAWS ;= [ BEYOND AND ABOVE THE ATTEMPTS TO EXPLAIN NATURE
IN TERMS OF ANTHROPOMORPHIC DEITIES
'REEK SCIENCE mOURISHED BETWEEN "# AND !$ REACHING REMARK

ABLE PEAKS $ECAY STARTED WHEN THE INTELLECTUAL ELITE BEGAN TO BE DRAWN
AWAY FROM SCIENCE BY THE SPREAD OF *UDEO #HRISTIAN ETHICS WITH ITS EMPHASIS
ON HUMANISM AND SOCIAL JUSTICE IN ADDITION TO ITS RAISING MILLENARY EXPECTA
TIONS 4HE RISE OF THE #HRISTIAN %STABLISHMENT [ IN WHOSE EYES THE SCIENTIlC
COMPONENT OF 'REEK NEO 0LATONIST PHILOSOPHY WAS UNJUSTLY CONSIDERED
INSEPARABLE FROM )DOLATRY [ THEN COMPLETED THE ERADICATION OF ALL RESEARCH
CENTERS UNTIL THE -EDITERRANEAN COUNTRIES WHERE 'REEK SCIENCE HAD GROWN
AND PROSPERED WERE SWEPT CLEAN OF ANY TRACE OF IT THE !CADEMY IN !THENS
WAS ALL THAT WAS LEFT BY !$ AND WAS lNALLY ALSO CLOSED IN !$ 3O
MUCH FOR THE HISTORY ,ET US NOW VIEW THE CONTENTS
6ISIONS IN -ATH 094(!'/2%!. !.$ 0,!4/.)# #/.#%04)/.3 2%!,):%$
'REEK <-ETA SCIENCE AS $ISTINGUISHED FROM 'REEK

3CIENCE
)N RETROSPECT ) SUGGEST THAT IT IS USEFUL WITHIN THE 'REEK CONTRIBUTION TO
TREAT SEPARATELY ON THE ONE HAND THE CLEAR CUT SCIENTIlC ACHIEVEMENTS [ AND
ON THE OTHER SOME IDEAS WHICH THOUGH THEY TOO MUST HAVE BEEN TRIGGERED
BY SOME CONTEMPORARY PHENOMENOLOGY NEVERTHELESS COULD NOT BE REALIZED
IN THOSE EARLY STAGES OF THE DEVELOPMENT OF SCIENCE AND YET WERE PROFOUND
AND PREGNANT WITH CONCEPTUAL ADVANCES lNALLY ACHIEVING FULL VINDICATION IN
MODERN TIMES 7E THEREFORE CLASSIFY THESE AS ACHIEVEMENTS IN META SCIENCE
RATHER THAN IN SCIENCE IN -ETAPHYSICS RATHER THAN IN 0HYSICS NOTE THAT )
USE THE PRElX META IN THE SENSE OF FOUNDATIONS OF NOT IN THE SENSE OF EITHER
PSEUDO OR SPIRITUAL WHICH IS NOW OFTEN THE INTENDED MEANING OF METAPHYSICS
%XAMPLES OF SCIENTIlC ACHIEVEMENTS IN -ATHEMATICS THE 0YTHAGORE
ANS PROOF OF THE EXISTENCE OF IRRATIONAL NUMBERS w"# %UCLIDS 'REEK
%UKLIDES GEOMETRY w"# OR IN 0HYSICS !RCHIMEDES LAW OF BUOY
ANCY OR HIS CONCEPTION OF ANGULAR MOMENTUM TO USE THE MODERN DElNITION
IN HIS STUDY OF THE LEVER w"# IN 0HYSICAL 'EOGRAPHY %RATHOSTENES
PRECISION MEASUREMENT OF THE RADIUS OF THE %ARTH w"# IN !S
TRONOMY (IPPARCHOS PRECISION MEASUREMENT OF THE DISTANCE TO THE
-OON w"# (ERONS INVENTION OF THE STEAM ENGINE !$ ETC 4HERE
WERE ALSO SOME FAILURES SUCH AS THE COMPLETELY WRONG RESULT FOR THE DISTANCE
TO THE 3UN
.OW AS A lRST EXAMPLE OF WHAT ) DESCRIBED AS METAPHYSICAL CONCEPTS
WE TAKE THE IDEA OF ATOMISM AS LAUNCHED BY $EMOKRITOS OF !BDERA AROUND
"# INmUENCED BY THE 0YTHAGOREAN 0HILOLAOS OF #ROTON AND INCORPORAT
ING SOME IDEAS ABSORBED BY $EMOKRITOS FROM THE TEACHINGS OF HIS MENTOR
,EUCIPPOS 4RUE $EMOKRITOS ALSO INCLUDED IN HIS PRESENTATION A PREMATURE
REALIZATION OF HIS IDEA BASED UPON THE POOR STATE OF #HEMISTRY IN HIS TIME
WHICH WAS LAGGING MUCH BEHIND THE ADVANCES IN 0HYSICS OR -ATHEMATICS
4HIS LED HIM TO SUGGEST HIS FOUR ELEMENT THEORY SOIL WATER AIR AND lRE
7HAT WAS REQUIRED AND MISSING WAS THE WORK OF SUCH AS *OSEPH 0RIESTLEY
OR OF !NTOINE ,AVOISIER IN THE %IGHTEENTH #ENTURY ACHIEVING A BETTER UN
DERSTANDING OF THE SIMPLEST CHEMICAL REACTIONS EG THE DISCOVERY OF OXYGEN
AND THE REJECTION OF THE PHLOGISTONCALORIC APPROACH [ WHICH MADE IT THEN
POSSIBLE FOR *OHN $ALTON TO CONCEIVE THE lRST SERIOUS SCIENTIlC REALIZATION OF

) HAVE TAKEN THE LIBERTY OF OFTEN REVERTING TO THE 'REEK FORM FOR 'REEK NAMES INSTEAD
OF THE USUAL ,ATINIZED OR !NGLICIZED FORMS EG %UKLIDES INSTEAD OF %UCLID $EMOKRITOS
RATHER THAN $EMOCRITUS 0LATON INSTEAD OF 0LATO !RISTOTELES RATHER THAN !RISTOTLE ETC
9 .%%-!. '!&!
$EMOKRITOS METAPHYSICAL IDEA 7E SHALL RETURN TO THE FOUR ELEMENT MODEL

IN OUR DISCUSSION OF 'EOMETRY SINCE $EMOKRITOS IN HIS MODEL RELATED THE
FOUR ELEMENTS TO FOUR OF THE lVE <PERFECT POLYHEDRONS WHILE RELATING THE
lFTH TO THE #OSMOS .OTE THAT THE LINKAGE BETWEEN ATOMS AND THE #OSMOS
IS BY ITSELF YET ANOTHER FRUITFUL METAPHYSICAL IDEA WHICH HAS HAD ITS SUCCESSES
LATELY
) HAVE RECENTLY ENUMERATED lVE IMPORTANT METAPHYSICAL CONTRIBUTIONS
PROFOUND IDEAS ABOUT THE SCIENCE OF 0HYSICS AND YET IDEAS WHICH COULD
TRULY NOT BE REALIZED IN ANTIQUITY ;= WHATEVER THE PHENOMENA AND CIR
CUMSTANCES WHICH INITIALLY TRIGGERED THEIR CONCEPTION [ BUT HAVE ALL SINCE
BEEN VINDICATED THE LAST TWO ONLY IN THE 88TH #ENTURY THOUGH IN A VERY
IMPRESSIVE WAY &OLLOWING EACH OF THE lVE NOTIONS ) LIST THE KEY RUNGS ON
THE LOGICAL LADDER LINKING THEM TO PRESENT PHYSICS 4HE REMAINING SECTIONS
OF THIS ARTICLE THEN REVIEW THE LAST TWO NOTIONS AS ALSO IMPLIED BY OUR TITLE
6ARIATIONAL 0RINCIPLES WERE INTRODUCED BY !RISTOTELES ;= AND AP

PLIED BY (ERON AND lFTEEN CENTURIES LATER BY &ERMAT [ WITH ,EIBNIZ THEN
DEVELOPING BOTH THE PHYSICAL IDEA AND THE MATHEMATICAL TOOLS 4HE IDEA
WAS THEN TAKEN UP BY THE BROTHERS *EAN AND *ACQUES "ERNOULLI AND TREATED
BY THEM AND BY THEIR STUDENT ,EONHARD %ULER THEN BY *OSEPH ,AGRANGE
WITH 0IERRE ,OUIS DE -AUPERTUIS lNALLY ARRIVING AT THE 0RINCIPLE OF ,EAST
!CTION ,ATER ADDITIONS ARE DUE TO 72 (AMILTON AND +' *ACOBI ALSO
TO &ELIX +LEINS AND 3OPHUS ,IES INTRODUCTION OF ALGEBRAIC VARIATIONS WITH
THEIR <%RLANGEN PROGRAM ;= TO %MMY .OETHER WITH HER TWO THEOREMS ;=
RELATING ALGEBRAIC INVARIANCE TO PHYSICAL CONSERVATION LAWS AND TO 2ICHARD
&EYNMAN WITH HIS VERSION OF 0ATH )NTEGRALS AND APPLICATIONS OF &UNCTIONAL
!NALYSIS /UR WORK IN REFS ;= HAS COVERED THE EVOLUTION OF THE 6ARIA
TIONAL 0RINCIPLES
4HE 6ACUUM AN ABSTRACT NOTION CONCEIVED BY 0LATON DENIED BY
!RISTOTELES EkECTIVE APPROACH REVIVED BY 'ALILEO AND AN ESSENTIAL ELEMENT
FOR HIS DElNITION OF INERTIA [ WITH 4ORRICELLI THEN PROVIDING A PHYSICAL REAL
IZATION 4HE CONCEPT WAS ESSENTIAL TO .EWTONS POSTULATION OF HIS LAWS OF
MECHANICS [ THIS IS THE CLASSICAL STORY ;= 4HE CONCEPT ACQUIRES A MAJOR
PHYSICAL ROLE IN THE 88TH #ENTURY WITH 1UANTUM -ECHANICS DISCOVERING
VACUUM ENERGY 'ENERAL 2ELATIVITY BRINGING IN THE #OSMOLOGICAL #ONSTANT
AND )NmATIONARY #OSMOGONY UNIFYING THE TWO PICTURES IN THE STORY OF #RE
ATION ;= 4HE 0LATON TO .EWTON PHASE OF THE <STORY OF THE VACUUM IS
DISCUSSED IN REF ;=
6ISIONS IN -ATH 094(!'/2%!. !.$ 0,!4/.)# #/.#%04)/.3 2%!,):%$
!TOMISM FROM $EMOKRITOS A BIG LEAP TO THE TIMES OF *OHN $ALTON

3INCE THEN ** 4HOMSON % 2UTHERFORD AND * #HADWICK DISCOVER THE
CONSTITUENTS OF $ALTONS ATOM FOLLOWED BY THE DISCOVERY lRST IN #OSMIC
2AYS AND THEN IN PARTICLE ACCELERATORS OF HUNDREDS OF HADRONS 4HIS LEAD
US TO THE DISCOVERY OF THE QUARK LAYER IN THE ONION OF MATTER AND THE
STORY HAS NOT ENDED ;= .OTE ALSO THAT THE OBSERVATIONS BY 4HOMSON
2UTHERFORD ETC AT THE BEGINNING OF THE 88TH #ENTURY ALSO DISPOSED OF THE
POSITIVIST <AS IFIC VERSION AS ESPOUSED BY % -ACH AND OTHERS REGARDING
PHYSICAL REALITY NAMELY THAT ATOMS DO NOT <REALLY EXIST THEY JUST REPRESENT
A <MODEL FOR THE DESCRIPTION OF PHENOMENA IE THINGS BEHAVE <AS IF THERE
ARE ATOMS 7E SHALL NOT REVIEW THE STORY OF !TOMISM HERE
4HE 'EOMETRIZATION OF PHYSICS 0LATON UNEXPECTED AND YET FULLY
REALIZED IN THE 88TH #ENTURY lRST IN THE WORKS OF %INSTEIN
AND THEN AGAIN BY THE EMERGENCE OF THE <3TANDARD -ODEL
4HIS WILL BE REVIEWED HERE THOUGH IT WAS DISCUSSED IN SOME DETAIL IN REFS
;= !S TO 0LATONS BELIEF IN 'EOMETRY AS A VEHICLE FOR 0HYSICS [ AS
DESCRIBED BY 0LUTARCH IN THE PASSAGE WE QUOTE THE USUAL GUESS BASED UPON
0LATONS PRESENTATION IN THE 4IMMEOS DIALOGUE OF THE GEOMETRICAL FEATURES
OF $EMOKRITOS MODEL INSPIRED BY 0HILOLAOS IS THAT THIS IS INDEED WHAT HE
HAD IN MIND 7E SHALL SEE IN OUR DISCUSSION OF PRESENT TRENDS THAT SOMETHING
OF THAT NATURE HAS INDEED COME TO PLAY A ROLE IN THE CANDIDATE FUTURE THEORY
AT THE END OF THE 88TH #ENTURY
0YTHAGOREANISMS OR <TETRACYTES [ SEE THE DICTIONARY QUOTATION
ABOVE 4HIS MUSIC INSPIRED AND ARITHMETICALLY ORIENTED PRESCRIPTION ALSO
MATERIALIZED IN THE 88TH #ENTURY EXCLUSIVELY FOR REASONS ) SHALL EXPLAIN )
HAVE TREATED THIS ISSUE BEFORE AND AM THUS REVIEWING IT HERE IN SOME DETAIL
0YTHAGOREANISMS AS INDUCED BY 1UANTUM -ECHANICS

7E ARE DElNITELY UNCERTAIN ABOUT THE ORIGINS OR JUSTIlCATION OF 0LATONS
CONlDENCE WITH RESPECT TO THE GEOMETRIZATION OF 0HYSICS BUT THERE IS NO
SUCH DOUBT IN THE MATTER OF THE 0YTHAGOREAN CLAIM THAT NATURE SHOULD
BE DESCRIBABLE BY SIMPLE ARITHMETICAL RELATIONS EG <TETRACYTES
(ERE IT IS THE ROLE PLAYED BY -USIC IN THE EYES OF 0YTHAGORAS AND OF HIS
3CHOOL WHICH CAUSED THE CONCEPTUAL <PHASE TRANSITION
4HE EMERGENCE OF DIMENSIONLESS SIMPLE RATIOS IN -USIC ;= IS CAUSED BY
THE WAVE NATURE OF 3OUND AS REALIZED BY STANDING WAVES IN MUSICAL INSTRU
MENTATION 4HE ARITHMETIC EMERGES THROUGH AN ARTIlCIAL <QUANTIZATION
9 .%%-!. '!&!
PROCESS 0YTHAGORAS AND HIS 3CHOOL UNCOVERED ALL THESE FACTS WATCHING A
MONOCORD THEY ALSO NOTICED THE EMERGENCE OF HARMONICS EMBELLISHING EACH
BASIC NOTE [ AND REALIZED THAT THE ORIGINATING MECHANISM IN THAT CASE IS THE
DIVISION IN OF THE VIBRATING STRING IN A VIOLIN OR A LYRE OR A SIMI
LAR DIVISION OF THE AIR COLUMN IN A mUTE OR IN ANY OTHER WIND INSTRUMENT
/NE THUS FACES A DISCRETE SYSTEM OF FREQUENCIES OR OF WAVELENGTHS [ OR IN
OTHER WORDS A 3PECTROSCOPY THOUGH ARTIlCIAL IN THE CASE OF MUSIC &OR
0YTHAGORAS GENERALIZATION TO BE REALIZED IE BEYOND MUSIC TWO ELEMENTS
WERE THEREFORE NECESSARY A lELD IN WHICH THE PHENOMENA INCLUDE A SPEC
TROSCOPY NATURAL THIS TIME AND THE MATHEMATICAL TOOLS [ IE 'ROUP AND
'ROUP 2EPRESENTATION 4HEORY WITH PERHAPS THE 0LANCHEREL FORMULA REPLAC
ING THE SIMPLE COUNT OF THE &OURIER EXPANSION
4HERE WERE ONE OR TWO FALSE STARTS *OHANNES +EPLER ONCE THOUGHT HE
HAD UNDERSTOOD THE SYSTEMATICS OF THE PLANETARY ORBITS HAVING WRONGLY
GUESSED THAT THEY CORRESPOND TO THE SEQUENTIAL BABUSHKA DOLL LIKE EMBED
DING OF THE lVE PERFECT SOLIDS )N THE 88TH #ENTURY 3IR !RTHUR %DDINGTON
!STRONOMER 2OYAL AND GREAT PROPONENT OF 'ENERAL 2ELATIVITY WAS ALSO A
SWORN 0YTHAGOREAN AND AFTER A FEW MINOR CASES TRIED IN VAIN TO EXPLAIN
THE < OF THE DIMENSIONLESS j%- E xYC
4HE MAIN MATHEMATICAL TOOLS NAMELY 'ROUP 4HEORY AND 'ROUP 2EP
RESENTATION 4HEORY WERE LAUNCHED AROUND BY TWO YOUNG MEN IN THEIR
TEENS .IELS (ENRYK !BEL AND %VARISTE 'ALOIS WITH
FURTHER DEVELOPMENTS BY 3OPHUS ,IE %LIE #ARTAN
AND OTHERS 0HENOMENOLOGICALLY IT WAS AT THE END OF THE 8)8TH #EN
TURY THAT ATOMIC SPECTROSCOPY STARTED BEING USED AND REVEALED SYSTEMS
OF LINES WHICH WERE ORGANIZED AND SYSTEMATIZED 7RITING t FOR THE WAVE
LENGTH N AND K FOR THE ,OWER AND THE 5PPER 1UANTUM .UMBERS RESPEC
TIVELY AND 2 FOR THE 2YDBERG CONSTANT WE HAVE THE 2YDBERG FORMULA
t_ 2;N_ _ K _ = FROM PHENOMENOLOGY )N (YDROGEN WE HAVE THE
SERIES
N K
,YMAN ;56=
"ALMER ;VISIBLE56=
0ASCHEN ;)2=
"RACKETT ;)2=
0FUND ;)2=
FOR WHICH .IELS "OHR FOUND AN AD HOC DERIVATION AND 1UANTUM -ECHANICS
AN EXACT ONE ;=
4HE SAME OCCURRED IN THE 0HYSICS OF ATOMIC NUCLEI WHERE -ARIA 'OEP
6ISIONS IN -ATH 094(!'/2%!. !.$ 0,!4/.)# #/.#%04)/.3 2%!,):%$
PERT -AYER AND ( *ENSEN IDENTIlED THE SHELL MODEL STRUCTURE ;= WITH
ONE OF THE MAIN CLUES SUPPLIED BY THE OBSERVATION OF THE SO CALLED MAGIC
NUMBERS DOESNT THIS SOUND 0YTHAGOREAN &URTHER WORK REVEALED MORE
STRUCTURE THE <COLLECTIVE MODEL ;= AND SOME FEATURES CORRESPONDING TO
SHAPE PULSATIONS QUADRUPOLE EXCITATIONS ;= AND lNALLY FEATURES LEADING
TO A MORE UNIlED PICTURE NAMELY THE )NTERACTING "OSON -ODEL ;= DUE TO
! !RIMA AND & )ACHELLO
4HE <4HIRD 3PECTROSCOPY IS A NAME GIVEN TO THE SPECTRUM OF THE
HADRONS WHICH WE IDENTIlED IN AS GIVEN BY THE ZERO TRIALITY UNITARY
REPRESENTATIONS OF 35 ;= OR IN OTHER WORDS BY THE UNITARY IRREDUCIBLE
REPRESENTATIONS <UNIRREPS OF THE QUOTIENT GROUP 35 : ;= (ERE
TOO STRUCTURAL DECIPHERING THE <1UARK -ODEL ;= AND DYNAMICAL UN
DERSTANDING 1UANTUM #HROMODYNAMICS ;= FOLLOWED BUT AS THIS SEQUEL IS
CAST INTO A GEOMETRICAL MOLD OF THE LOCAL TYPE IT WILL BE DISCUSSED UNDER
THAT CATEGORY
7HAT BETTER ILLUSTRATION OF 0YTHAGOREANISMS THAN THE RESULTS OBTAINED
FROM THE STATIC APPROXIMATION CONSISTING OF THE COMBINATION OF THE ABOVE
HADRON UNITARY SYMMETRY 35 H WITH SPIN IN 35 STATIC YIELDING FOR EX
AMPLE FOR THE RATIO BETWEEN THE MAGNETIC MOMENTS OF THE PROTON AND NEU
uP
TRON uN _ OR FOR THE RATIO BETWEEN THE ASYMPTOTIC TOTAL CROSS
SECTIONS FOR A MESON PROJECTILE TO THAT OF A NUCLEON PROJECTILE BOTH BEING
zx8
HURLED AT THE SAME TARGET 8 IS z.8 ETC
.OTE THAT THERE ARE NOW NEW QUESTIONS RELATING TO SUCH <MUSIC OF THE
SPHERES 4HE 3TANDARD -ODEL DOES NOT EXPLAIN THE PATTERN OF MASSES
AT THE LEVEL OF THE FUNDAMENTAL MATTER lELDS AT THIS STAGE THE QUARK AND
LEPTON lELDS 7HY <GENERATIONS ;U D C S T B FOR THE QUARKS EACH
IN COLORS AND vE Evu uv{ { FOR THE LEPTONS= -OREOVER WHAT IS THE
MEANING OF THE MASS OR ENERGY SPECTRUM THEY DISPLAY 4HIS QUESTION HAS
BECOME ESPECIALLY <HOT AFTER THE DISCOVERY OF THE TOP QUARK WITH A MASS
OF SOME 'E6 WHEREAS THE OTHER QUARKS AND ALL LEPTONS HAVE MASSES
SMALLER THAN 'E6 7ILL THE PATTERN DISPLAYED BY THIS FOURTH SPECTROSCOPY
PROVE TO BE AS READABLE BY ALGEBRAIC MEANS AS THE PRECEDING ONES
! REMARK REGARDING THE 0YTHAGOREANS WHO ARE GENERALLY TREATED AS A
<CULT )F THIS CLASSIlCATION IS DUE TO THEIR BIZARRE UNIVERSAL NAMELY A BELIEF

)N ) EXPERIENCED A PERSONALIZED ILLUSTRATION OF THE SIMILARITIES BETWEEN THE
DISCOVERY OF THE REGULARITIES IN ATOMIC AND HADRONIC SPECTROSCOPIES ) HAD JUST PUBLISHED
A NON TECHNICAL ARTICLE ABOUT THE DISCOVERY OF THE 35 CHARGES AND CURRENTS WHEN )
RECEIVED A LETTER FROM A LADY IN 3WEDEN WHO WAS 2YDBERGS DAUGHTER AND GAVE ME AN
EYEWITNESS DESCRIPTION OF THE SIMILARITIES BETWEEN THE TWO STORIES
9 .%%-!. '!&!
IN THE POSSIBILITY OF DESCRIBING NATURE IN TERMS OF ARITHMETICAL RELATIONS

THEY SHOULD INSTEAD BE CONSIDERED AS THE MOST SCIENTIlC AND NONRELIGIOUS
INTELLECTUAL GROUPING OF ALL )N THEIR INTEREST IN MUSIC THEY HAD UNDERSTOOD
AND RECONSTRUCTED A LIMITED SECTOR IN THE WORKINGS OF NATURE [ AND THEIR
CREDO AMOUNTED TO A FEELING THAT THIS SECTOR WAS TYPICAL 7E NOW KNOW THAT
IT HAD NO ANALOG IN MACROSCOPIC PHYSICS BUT WITH THEIR GUESS THEY HIT THE
JACKPOT ONCE WE HAD CROSSED INTO THE QUANTUM WORLD
4HE &IRST 'EOMETRIZATION 2ELATIVITY 'LOBAL

AND ,OCAL 0HASES
)N WRITING HIS *UNE lRST 2ELATIVITY PAPER </N THE %LECTRODYNAMICS
OF -OVING "ODIES ;= %INSTEIN HAD NO IDEA HE WAS TOUCHING ANYTHING
HAVING TO DO WITH GEOMETRY )T WAS HIS FORMER TEACHER AT THE %4( (ER
MANN -INKOWSKI NOW A PROFESSOR AT ' OTTINGEN WHO REALIZED AND POINTED
OUT THAT %INSTEINS 0RINCIPLE OF 2ELATIVITY WITH ITS POSTULATED INVARIANCE
OF THE VELOCITY OF LIGHT WAS IN FACT A STATEMENT ABOUT THE DIMENSIONAL
GEOMETRY OF SPACETIME BEING PSEUDO EUCLIDEAN - IE WITH METRIC Guv
DIA_ X CT INSTEAD OF THE 'ALILEAN 2 a 2 -INKOWSKI EX
PLAINED THIS RESULT AT THE .ATIONAL #ONGRESS OF 'ERMAN 0HYSICIANS AND
.ATURE RESEARCHERS IN #OLOGNE IN ;= .O DOUBT THAT THIS INTERPRE
TATION WAS USEFUL IN SIMPLIFYING THE APPLICATION OF ,ORENTZ AND 0OINCARsE
GROUP TRANSFORMATIONS BUT THIS WOULD NOT HAVE BEEN CRUCIAL HAD %INSTEIN
NOT BEEN WORKING ON HIS PROGRAM OF HARMONIZING .EWTONIAN 0HYSICS IN
VARIANT UNDER THE 'ALILEO GROUP WITH -AXWELLS %LECTRODYNAMICS AND 3PE
CIAL 2ELATIVITY INVARIANT UNDER THE 0OINCARsE AND ,ORENTZ GROUPS 5SING
% MC WHICH HE HAD MEANWHILE DERIVED IN HIS SECOND 3PECIAL 2ELATIV
ITY PAPER ;= AND ASSUMING GRAVITY TO BE COUPLED TO ENERGY THROUGH ITS
M p %C INVERSE HE EVALUATED THE EkECTS OF A GRAVITATIONAL lELD ON A BEAM
OF LIGHT AND GOT THE DEmECTION HALF THE ONE RESULTING FROM 'ENERAL 2ELA
TIVITY ;= 'ONE WAS THE INVARIANCE OF THE SPEED OF LIGHT )T WAS AT THIS
POINT THAT %INSTEIN WAS FORCED TO LOOK FOR A MORE COMPREHENSIVE APPROACH
[ AND FOUND THE 'EOMETRICAL LINE UNIQUE IN BEING ABLE TO ACCOMMODATE
CHANGES IN THE METRIC [ PROVIDED IT WOULD GO BACK TO -INKOWSKIS ON THE
TANGENT AT ANY POINT OF SPACETIME 4HIS WAS 2IEMANNIAN 'EOMETRY %IN
STEIN LEARNED FROM -ARCEL 'ROSSMANN HIS FRIEND AND FORMER SCHOOLMATE AT
THE %4( %INSTEIN WORKING lRST WITH 'ROSSMANN ;= AND THEN PURSUING
ON HIS OWN ARRIVED AT THE <'ENERAL 4HEORY OF 2ELATIVITY IN ;=
6ISIONS IN -ATH 094(!'/2%!. !.$ 0,!4/.)# #/.#%04)/.3 2%!,):%$
.OTE THE STORY OF (ILBERTS GETTING THERE lRST HAS RECENTLY BEEN lNALLY
PUT TO REST ;= 4HE GEOMETRY OF 'ENERAL 2ELATIVITY POSSESSES TWO LOCAL
INVARIANCES PASSIVELY INVARIANCE UNDER THE LOCAL DIkEOMORPHISMS OF 2
WHILE ACTIVELY IT IS INVARIANT UNDER THE ,ORENTZ GROUPS ACTION ON A LOCAL
FRAME AN ANHOLONOMIC FEATURE 3UCH LOCAL INVARIANCES REQUIRE THE PRESENCE
;A
OF CONNECTION lELDS `yzu X FOR THE HOLONOMIC DIkEOMORPHISMS AND ù B=
NECESSARY FOR THE ANHOLONOMIC LOCAL ,ORENTZ INVARIANCE #ONNECTION lELDS
ENTER INTO THE REALIZATION OF PARALLEL TRANSPORT THROUGH A COVARIANT DERIVA
TIVE $u u _ ù X WHERE THE ANHOLONOMICAL INDICES IN THE CONNECTION
lELD HAVE BEEN CONTRACTED WITH THESE CORRESPONDING MATRICES IN THE ,IE
ALGEBRA !SIDE FROM THIS ONE HAS TO USE THE FRAMES ALGEBRAIC VECTOR OR
TANGENT lELDS AND THEIR COTANGENT INVERSES FORMS IN DElNING THE OVERALL
COVARIANT DERIVATIVE $A EuA $u
%XTENDING THE 'EOMETRY SO AS TO )NCLUDE ,OCAL

<)NTERNAL 3YMMETRIES #REATION OF THE -EANS IN AN
%ARLY !TTEMPT AT 5NIlCATION
-UCH HAS BEEN SAID AND WRITTEN ABOUT %INSTEINS 5NIlCATION 0ROGRAM
LAUNCHED IN IN WHICH THE AIM WAS TO UNITE 'RAVITY WITH %LECTROMAG
NETISM .OTE THAT THE EXISTENCE OF OTHER INTERACTIONS ONLY BECAME CLEAR
AROUND WITH THE DISCOVERY OF THE NEUTRON [ WHOSE STRONG BINDING IN
NUCLEI IS TIMES STRONGER THAN GRAVITY &ROM THE STRENGTH OF THE BINDING
BETWEEN TWO PROTONS WITHIN AN ALPHA PARTICLE (ELIUM NUCLEUS ONE CAN
ALSO ESTIMATE THAT IT IS SEVERAL HUNDRED TIMES STRONGER THAN THE #OULOMB
REPULSION BETWEEN THESE PROTONS !NYHOW BEFORE THE ONLY CONTEM
PLATED UNIlCATION WAS WITH -AXWELLS lELD 4HREE METHODS GOT FURTHEST IN
THIS PROGRAM
(ERMANN 7EYL ARRIVED AT A &IBER "UNDLE 'EOMETRY [ KNOWN IN
0HYSICS AS A LOCAL GAUGE THEORY (IS lRST ATTEMPT TRIED TO IDENTIFY THE
ABELIAN GROUP IN THE lBER AS THE GROUP OF SCALE TRANSFORMATIONS 4HIS WAS
WRONG AND WAS REPLACED TEN YEARS LATER BY 5 AS THE &IBER GROUP ;=
ACTING ON THE COMPLEX PHASE OF THE ELECTRON WAVE FUNCTION IN 1UANTUM
-ECHANICS )N #. 9ANG AND 2, -ILLS CONSTRUCTED THE GENERAL
IZATION OF THIS GEOMETRY FOR THE CASE OF A NON ABELIAN ,IE GROUP ;= 4HE
0RINCIPAL &IBER "UNDLE "- ' x ? IS GIVEN BY THE BASE SPACE - MOSTLY
SPACETIME LAID QUASI HORIZONTALLY THE VERTICAL lBER HERE THE ,IE GROUP
5 OR 35 ETC THE VERTICAL PROJECTION x FROM ANY POINT 0 X ONTO
9 .%%-!. '!&!
X AND THE DOT PRODUCT IS A MAP FROM THE ABSTRACT ,IE ALGEBRA OF ' ONTO
OPERATORS LYING WITHIN THE TANGENT OF THE "UNDLE MANIFOLD AS SUCH ;=
0REEMPTING OUR COMING DESCRIPTION OF ITS DEVELOPMENT WE NOTE HERE THAT
THE ENTIRE 3TANDARD -ODEL ;= AS COMPLETED IN IS A 9ANG -ILLS &"
WITH SPACETIME - AS BASE SPACE AND ' 3;5 COLOR a 5 WEAK =
!LBERT %INSTEIN HIMSELF TRIED TO EXTEND THE GEOMETRY OVER DIMENSIONS
WITH AN ANTISYMMETRIC METRIC UNSUCCESSFULLY !FTER THE #04 THEOREM AND
WHAT WE HAVE LEARNED ABOUT QUANTUM lELD THEORY IT IS CLEAR THAT THE SUB
SPACE WITH THE ANTISYMMETRICAL COORDINATES IE SYMPLECTIC METRIC SHOULD
HAVE INVOLVED SPINORIAL COORDINATES WITH HALF INTEGER SPIN 4HIS IS SUPERSPACE
AND THE NATURAL SUPERALGEBRA OF ISOMETRY IS THE /RTHO SYMPLECTIC SUPERALGE
BRA (OWEVER SPACETIME WITH ITS 0OINCARsE ISOMETRY ALGEBRA CAN SUPPORT A
SMALLER SUPERALGEBRA NAMELY SUPERSYMMETRY AS INTRODUCED BY 'OLFAND AND
,IKHTMAN IN AND BY * 7ESS AND " :UMINO IN ;= 7ORK ON THE
(IERARCHY ISSUE HAS SHOWN THAT SUPERSYMMETRY IS ESSENTIAL AND SHOULD BE
OBSERVED IN THE HADRON SPECTRUM UNDER 4E6 3UPERSYMMETRY ALSO ALLOWS
ALMOST UNIQUELY FOR THE UNIlCATION OF SPACETIME AND INTERNAL SYMMETRY
GROUPS [ AND THAT OF GRAVITY WITH OTHER GAUGE GROUPS WITHIN SUPERGRAVITY
4HEODOR +ALUZA AND /SKAR +LEIN ;= DEVISED A DIkERENT MERGER WITH
SPACETIME IN WHICH ONE ASSUMES ADDITIONAL DIMENSIONS GENERALLY OF ANGU
LAR TYPE THE NEW ANGLES PROVIDING SPACETIME WITH NEW GAUGE SYMMETRIES
AND THE ENTIRE EMBEDDING MANIFOLD OBEYING THE 2IEMANNIAN CONSTRAINTS
)N RECENT YEARS THE ISSUE HAS BEEN REVIVED FOR DIkERENT AND RATHER MORE
COMPELLING REASONS IN TWO DIkERENT CONTEXTS
A <MAXIMAL SUPERGRAVITY ;= WHICH IS CONSTRUCTED WITH . A SIN
GLE SUPERSYMMETRY IN DIMENSIONS YIELDING UNDER DIMENSIONAL REDUC
TION . SUPERSYMMETRIES IN < DIMENSIONS THE MAXIMUM ALLOWED
BY THE AXIOMS !N EXAMPLE OF SPONTANEOUS COMPACTIlCATION OF THE ADDED
DIMENSIONS HAS BEEN WORKED OUT ;=
B SUPERSTRINGS ONLY EXIST IN N DIMENSIONS FOR THE CANCELLATION
OF THE DILATIONAL 1UANTUM ANOMALY AND IN PARTICULAR IN DIMENSIONS FOR
THE CANCELLATION OF A TACHYON IN ITS EXCITATIONS SPECTRUM ;= 3UPERSTRING
THEORY AND ITS NEWEST VERSION <- THEORY ;= HAVE BEEN DEVELOPED AS
CANDIDATE POST 0LANCK ENERGY PHYSICAL THEORIES UNIFYING 'RAVITY AND THE
3TANDARD -ODEL ALL QUANTIZED IE REPRODUCING A lNITE OR RENORMALIZABLE
1UANTUM 'RAVITY THEORY ;=
3UMMING UP WE NOTE THAT THE THREE MAIN ROUTES TO UNIlCATION DRAWN
IN WHEN -AXWELLS ELECTRODYNAMICS WAS THE ONLY CANDIDATE FOR
6ISIONS IN -ATH 094(!'/2%!. !.$ 0,!4/.)# #/.#%04)/.3 2%!,):%$
UNIlCATION WITH GRAVITY ARE ALSO THE MAIN ROUTES NOW UTILIZED BOTH IN
CONSTRUCTING THE 3TANDARD -ODEL IE PREPARING THE ANALOG TO -AXWELLS
THEORY IN A 7EYL LIKE GEOMETRIZED MOLD AND IN UNIFYING IT WITH GRAVITY [
AND ALSO IN GOING BEYOND IT WHETHER IN A 1UANTUM &IELD 4HEORY FORMU
LATION OR WITHIN 3UPERSTRINGS THEORY AND ITS 3UPERMEMBRANE EXTENSIONS
;= ETC
4HE 3ECOND 'EOMETRIZATION 'AUGE 4HEORIES 'LOBAL

AND 1UASI LOCAL 0HASE
4HE THIRTY YEARS STARTING WITH THE BIRTH OF 1UANTUM -ECHAN
ICS WERE EXTREMELY NON GEOMETRICAL IN THEIR FUNDAMENTAL PHYSICS OUTPUT
4HIS NON PLATONIC ERA CAN BE CONSIDERED TO HAVE ENDED AROUND AS A
RESULT OF EXPERIMENTS INVOLVING THE HADRONS IE PARTICLES PARTAKING IN THE
3TRONG )NTERACTIONS BUT PROBED HERE THROUGH BOTH THE 7EAK AND THE %LEC
TROMAGNETIC )NTERACTIONS !LREADY IN %NRICO &ERMI HAD SHOWN THAT
THE EkECTIVE 7EAK )NTERACTIONS (AMILTONIAN COULD BE WRITTEN AS A PRODUCT
OF TWO CURRENTS SOMEWHAT LIKE THE INTERACTION BETWEEN TWO CONDUCTORS IN
ELECTROMAGNETISM &OR INSTANCE NUCLEAR BETA DECAY N P E_ vE CAN
BE DESCRIBED AS THE INTERACTION BETWEEN A NUCLEON CURRENT N P AND A
LEPTON CURRENT IN THE <PAIR CREATION MODE vE E_ [ JUST AS WE WOULD HAVE
DESCRIBED A PROTON CURRENT EMITTING AN ELECTRON POSITRON PAIR 4HE
7EAK )NTERACTIONS EXPERIMENTS MOSTLY HAILED AT THE TIME AS PROOF OF
THE NON CONSERVATION OF 3PATIAL 2EmECTION 3YMMETRY <0ARITY CONSISTED
OF PROBING FOR THE ,ORENTZ BEHAVIOR OF THE &ERMI CURRENTS IE IDENTIFY
ING EXPERIMENTALLY WHICH AMONGST THE lVE ALLOWED $IRAC SPINOR BILINEARS
3CALAR 0SEUDO SCALAR 6ECTOR !XIAL VECTOR ANTI SYMMETRIC 4ENSOR IS IN
VOLVED AND THE ANSWER WAS <6 ! OR Ju X l v u ; l =X )N A
MASSLESS $IRAC SPINOR COMPLEX COMPONENTS THIS PROJECTS OUT A <LEFT
CHIRAL [ WITH COMPLEX COMPONENTS 7EYL SPINOR [ A SEPARATION WHICH
EXISTS FOR A -INKOWSKI METRIC IN D N DIMENSIONS N AN INTEGER [ OR IN
D N AS A REAL COMPONENTS -AJORANA 7EYL SPINOR 4HE WEAK CHIRAL
ELECTRICALLY CHARGED CURRENTS NO ELECTRICALLY NEUTRAL WEAK TRANSITIONS HAD
BEEN OBSERVED AT THAT TIME WOULD BE PRECISELY CONSERVED IN THE LIMIT OF
MASSLESS FERMION MATTER lELDS [ AND THUS lT INTO A 9ANG -ILLS TYPE IN
TERACTION WITH HYPOTHETICAL VECTOR POTENTIALS OR INDUCED lELDS ENSURING
PARALLEL TRANSPORT IN THE lBER BUNDLE GEOMETRY AND [ PHYSICALLY [ MEDIAT
ING THE INTERACTION BETWEEN TWO CURRENTS 4HE 7EAK )NTERACTION APPEARED
9 .%%-!. '!&!
TO TAKE A TURN TOWARDS GEOMETRY THOUGH THE PRECISE GROUP STRUCTURE WASNT
CLEAR YET
-EANWHILE 2OBERT (OFSTADTER WAS PROBING THE ELECTROMAGNETIC STRUC
TURE OF THE NUCLEON (IS RESULTS WERE EXPLAINED BY 9 .AMBU AND
OTHERS WHO SHOWED THAT THEY INDICATED THAT THE NUCLEONS PRIMORDIAL SUR
ROUNDING MESON CLOUD IS GENERATED BY ISOSPIN AN ABSTRACT 35 TRANSFOR
MATION IN (ILBERT SPACE TURNING A PROTON INTO A NEUTRON ) AND )
SPIN PARITY * 0 _ MASSIVE IE SHORT RANGED VECTOR POTENTIALS [ RATHER
THAN BEING CONSTITUTED OF THE ) * 0 _ 9UKAWA PIONS AS PREVIOUSLY
THOUGHT 3OME _ SECONDS AFTER THEIR EMISSION THESE VECTOR MESONS DE
CAY INTO AND PIONS RESPECTIVELY FACTS WHICH WERE SOON VERIlED DIRECTLY
WITH INERTIAL MASS PLOTS IN BUBBLE CHAMBER EXPERIMENTS ** 3AKURAI THEN
CONVINCINGLY ARGUED THAT THE 3TRONG )NTERACTIONS SHOULD BE DESCRIBED
BY A 9ANG -ILLS TYPE (AMILTONIAN GENERATED BY ,OCAL )NVARIANCE UNDER
;35 ) i 5 9 =
i 5 " 5 )9 i 5 "
:
INVOLVING THE LOCAL CONSERVATION OF ISOSPIN ) HYPERCHARGE 9 AND BARYON
NUMBER " 4HE INCLUSION OF THIS LAST 5 " CHARGE CORRESPONDING TO BARYON
NUMBER OR !TOMIC -ASS .UMBER FOLLOWED %DWARD 4ELLERS ORIGINAL SUGGES
TION AS A POSSIBLE MECHANISM EXPLAINING THE OBSERVED HARD CORE LIMITING THE
COMPRESSIBILITY OF NUCLEAR MATTER 4HE lELD EMITTED BY THE BARYONIC CHARGE
WOULD PRODUCE A #OULOMB LIKE REPULSION AMONG ANY TWO NUCLEONS 3AKURAI
NOW POINTED TO CORROBORATING EXPERIMENTAL EVIDENCE
)T IS AT THIS STAGE THAT ) BECAME PERSONALLY INVOLVED IN THIS STORY IN MY
SEARCH FOR AN ALGEBRAIC DESCRIPTION OF THE HADRON SPECTRUM SEE LAST PARA
GRAPH OF SECTION 4HIS WAS A NEW APPROACH CONSIDERED AS <ABSTRACT
BECAUSE IT WAS NOT PRE SUPPOSING ANY DYNAMICAL OR STRUCTURAL MODEL [
WHEREAS MOST RESEARCHERS IN 0ARTICLE 0HYSICS WERE TRYING TO GUESS AT SOME
HYPOTHETICAL INITIAL CONDITIONS EITHER MECHANICAL EG A SPINNING TOP OR
STRUCTURAL SUCH AS THE 3AKATA MODEL IN WHICH ALL HADRONS ARE MADE OF
P N t <MORE FUNDAMENTAL PARTICLES )N THE ABSTRACT APPROACH ONE
WAS SEARCHING FOR A SIMPLE GROUP OF RANK R SO THAT IT WOULD CON
TAIN THE ISOSPIN HYPERCHARGE 5 4HE RANK SEMI SIMPLE ,IE ALGEBRAS IN
#ARTANS CLASSIlCATION ARE ! " # $ ' AND VARIOUS EXPERIMENTAL OB
SERVATIONS SELECTED ! THE GENERATOR ALGEBRA FOR 35 OR 3, 2 4HE
REQUIREMENTS OF lNITE MULTIPLETS CORRESPONDING TO UNITARY IRREDUCIBLE REPRE
SENTATIONS THEN SELECTS 35 4HE INCLUSION 35 z 5 )9 GAVE SENSIBLE
RESULTS INCLUDING THE ADDITION OF NEW VECTOR MESONS <+ b AND THEIR ANTI
6ISIONS IN -ATH 094(!'/2%!. !.$ 0,!4/.)# #/.#%04)/.3 2%!,):%$
PARTICLES WHICH WERE OBSERVED EXPERIMENTALLY "ARYON NUMBER ENTERS VIA

AN EXTENSION TO THE NON SIMPLE 5 4HE RESULT WAS THUS BOTH A CLASSIlCA
TION OF ALL HADRONS A <LAW OF FORCE REPRESENTED BY THE #LEBSCH 'ORDAN CO
EnCIENT IN THE INVARIANT COUPLING BETWEEN MULTIPLETS CONSERVED CURRENTS
OR COUNTING BARYON NUMBER BY .OETHERS lRST THEOREM APPROXIMATE
UNIVERSAL COUPLING OF THESE CURRENTS [ IE EQUAL TO THE MATRIX ELEMENTS OF
THE CORRESPONDING CHARGES BETWEEN THE RELEVANT STATES [ TO THE g VECTOR
MESONS [ THUS AN APPROXIMATE 9ANG -ILLS )NTERACTION FOR 5
4HE NEXT STEP WAS STRUCTURAL 7ITH (AIM 'OLDBERG ) SUGGESTED A
STRUCTURAL MODEL ;= .ORMALIZING THE 5 OF THE BARYON NUMBER TO "
MEANT THAT ALL BARYONS SHOULD BE IN REPRESENTATIONS APPEARING IN THE
TRIPLE DIRECT SELF PRODUCT OF THE DElNING REPRESENTATION OF 35 NAMELY
i i g g g WHEREAS MESONS SHOULD CONSIST IN THE PRODUCT
b i g 4HIS ALGEBRAIC MODEL WAS FURTHER DEVELOPED BY - 'ELL -ANN
WHO COINED THE TERM QUARKS FOR THE BASIC TRIPLET AND BY 'EORGE :WEIG
.OTE THAT QUARKS HAVE FRACTIONAL ELECTRICAL CHARGES U D_ S_ )T
WAS SUGGESTED [ AND LATER VERIlED EXPERIMENTALLY [ THAT QUARKS HAVE SPIN
*
7E SAW THAT THE 7EAK )NTERACTION INVOLVES QUASI CONSERVED CURRENTS
[ MORE PRECISELY A FULLY CONSERVED VECTOR HALF AND A PARTIALLY CONSERVED
AXIAL VECTOR HALF 4HE LATTERS <LEAK u Au #})x IS COMPENSATED BY
THE PION lELD #ONSERVATION LAWS INDICATE THE PRESENCE OF SYMMETRY [
AND SUCH A PARTIAL CONSERVATION CORRESPONDS TO A SPONTANEOUS SYMMETRY
BREAKDOWN GENERATED BY A 'OLDSTONE BOSON 4HIS AMOUNTS TO RELEGATING
THE SYMMETRY BREAKING AGENT TO THE BOUNDARY CONDITIONS HERE THE (ILBERT
SPACE OF QUANTUM STATES WITH A NON TRIVIAL ASSIGNMENT OF THE VACUUM STATE
[ AND THE 'OLDSTONE STATES CONSTITUTING THE REMAINING STATES IN THIS VACUUM
CONTAINING MULTIPLET THUS ALSO FORCING THEIR MASSES TO VANISH IN A LIMIT IN
WHICH THE COMPENSATION TO THE BROKEN CONSERVATION LAW BECOMES PRECISE
7ITH THE PION AND OTHER MEMBERS OF THE * 0 _ 35 OCTET PLUS SINGLET
MESONS NO MORE TAKEN AS FUNDAMENTAL THE EXPRESSION USED HAS CHANGED TO
PSEUDO 'OLDSTONE STATES
4HE 3ECOND 'EOMETRIZATIONS ,OCAL 0HASE A 1#$

-EANWHILE THE MYSTERY OF THE PARADOXICAL NON OBSERVATION OF STATES BE
LONGING IN UNIRREPS WITH NON ZERO TRIALITY DEEPENED 4HE 4RIALITY ;= OF A
9 .%%-!. '!&!
REPRESENTATION OF 35 IS THE NUMBER OF TRIPLET FACTORS NEEDED TO REACH

THAT REPRESENTATION IN A DIRECT PRODUCT i i i MOD 4HE REPRESEN
TATIONS OF 35 : ARE THE 4 REPRESENTATIONS OF 35 AS : THE
CENTER OF 35 IS A GROUP OF THREE BY MATRICES A WITH A THE THREE
COMPLEX CUBIC ROOTS OF THE IDENTITY [ SO THAT IN 4 UNIRREPS THEY ARE
ALL REPRESENTED BY THE IDENTITY !LSO THE ANTI QUARK UNIRREP b HAS TRIALITY
4 p _ AS CAN BE SEEN IN THE REDUCTION OF THE PRODUCT i g b
INTO ITS SYMMETRIC AND ANTISYMMETRIC PIECES 4HE OBSERVED UNIRREPS CAN
BE REACHED EITHER BY A PRODUCT OF QUARKS OR BY MULTIPLYING A QUARK BY
AN ANTI QUARK i b g IE IN EITHER CASE 4 -OREOVER EVEN
VERY HIGH ENERGY SCATTERING NEVER RESULTED IN A QUARK <BREAKING Ok ANY
EXTRA ENERGY JUST BROUGHT ABOUT THE EMISSION OF MESONS THUS PRESERVING
THE TRIALITY OF THE ORIGINAL HADRON 4HIS FEATURE CAME TO BE KNOWN AS QUARK
CONlNEMENT
7ORKING WITH QUARKS IN A STATIC APPROXIMATION INDUCES COUPLED SPIN

UNITARY SPIN INVARIANCE IE AN INVARIANCE UNDER THE EXCHANGE OF THE SIX
BASIC QUARK HELICITY STATES OR AN 35 STATIC SYMMETRY ;= )T YIELDED
SEVERAL HUNDRED VERIlABLE RESULTS [ WHICH WERE INDEED VALIDATED SEE OUR
DISCUSSION OF 0YTHAGOREANISMS (OWEVER USING 35 FURTHER COMPLICATES
THE SITUATION BECAUSE ALL GOOD RESULTS CONlRM AN ASSIGNMENT WHICH PUTS THE
GROUND STATE HADRONS IN UNIRREPS SYMMETRICAL IN THE ENTIRE SYSTEM OF QUARK
1UANTUM .UMBERS THUS APPEARING TO VIOLATE THE SPIN STATISTICS THEOREM
4HE ONLY POSSIBLE RESOLUTION OF THE PARADOX IS TO ASSUME THE EXISTENCE OF AN
ADDITIONAL CHARGE OR 1UANTUM .UMBER CARRIED BY QUARKS AND IN WHICH
THESE MATTER MULTIPLETS ARE ANTISYMMETRICAL 3INCE ALL THESE MULTIPLETS ARE
QUARK STATES THE NEW 1. SHOULD HAVE A TOTALLY ANTISYMMETRICAL STATE
IN A TRIPLE PRODUCT [ AND IS THEREFORE MOST PROBABLY YET ANOTHER 35
-OREOVER TO GUARANTEE THAT ALL <GROUND STATE HADRONS DO COINCIDE WITH
THE ANTISYMMETRICAL ASSIGNMENT FOR THE NEW 1. AND VICE VERSA IT HAS TO
BE DYNAMICAL IE THE AGENT INVOLVED IN GLUING QUARKS INTO HADRONS 4HERE
WAS A lRST VERSION OF THE NEW 35 COLOR DUE TO (AN AND .AMBU IN WHICH
THE QUARKS DO NOT HAVE FRACTIONAL CHARGES AND THESE APPEAR AS AN AVERAGE
OF THREE TYPES OF INTEGER CHARGE QUARKS BUT EXPERIMENTAL RESULTS REJECTED
THIS RATHER ELEGANT MODEL AND DElNITELY POINTED TO A STRAIGHTFORWARD SOLU
TION IN WHICH THE GENERATOR ALGEBRAS OF THE TWO 35 GROUPS ;= COM
MUTE ;SU HADRON SU COLOR = 3INCE IT HAS TO BE DYNAMICAL THE 35
INVARIANCE IS LOCAL [ IE A 9ANG -ILLS MODEL 4HIS IS 1UANTUM #HROMO
DYNAMICS <1#$ )T WAS POSTULATED IMMEDIATELY AFTER THE PUBLICATION
6ISIONS IN -ATH 094(!'/2%!. !.$ 0,!4/.)# #/.#%04)/.3 2%!,):%$
OF A MATHEMATICAL PROOF OF ASYMPTOTIC FREEDOM A FEATURE ALMOST UNIQUE TO

9ANG -ILLS THEORIES AND CONSISTING OF A GRADUAL AND MONOTONIC WEAKENING
OF THE THEORYS UNIVERSAL COUPLING AT EVER SHORTER DISTANCES 4HIS EXPLAINS
YET ANOTHER MYSTERY NAMELY THE DETECTABILITY OF THE QUARKS INSIDE A HADRON
[ IN WHAT IS SUPPOSED TO BE A VERY 3TRONG )NTERACTION )T ALSO RAISED HOPES OF
A DYNAMICAL DERIVATION OF COLOR CONlNEMENT WITH THE COUPLING WEAKENING
AT SHORT DISTANCES IT MUST BECOME STRONGER WITH INCREASING DISTANCE &ULL
CONlNEMENT WOULD IMPLY THAT AT A CERTAIN lNITE DISTANCE THE VALUE OF THE
COUPLING GOES TO INlNITY WHICH IS WHY QUARKS CANNOT ESCAPE .OTE THAT AS
OF THE YEAR [ ALL EXISTING MATHEMATICAL PROOFS OF COLOR CONlNEMENT
EITHER USE THE APPROXIMATION OF REPLACING SPACETIME BY A lNITE INTERVAL LAT
TICE [ OR INVOLVE ADDITIONAL MECHANISMS WHICH WE DO NOT HAVE IN A <PURE
9ANG -ILLS THEORY SUCH AS 1#$ -OSTLY MODELS APPLY THE (IGGS MECHA
NISM )N THE MOST ELEGANT OF SUCH PROOFS ;= THERE IS AN ADDED ASSUMPTION
OF SUPERSYMMETRY WHICH BRINGS IN SCALAR lELDS ACCOMPANYING THE MATTER
lELD IN ITS SUPERMULTIPLETS /NE OF THESE SCALARS BECOMES THE (IGGS lELD
)N THAT CASE CONlNEMENT IS JUST THE EkECT OF A SHORT RANGED PENETRATION
DUE TO THE CHROMODYNAMICAL lELD BECOMING MASSIVE !S IT IS IN FACT THE
<CHROMO ELECTIC COMPONENT WHICH HAS TO ACQUIRE MASS WE HAVE A <DUAL
TO THE PICTURE IN SUPERCONDUCTIVITY WHERE THE -EISSNER EkECT RELATES TO THE
MAGNETIC lELD
4HE GEOMETRICAL NATURE OF THE 1#$ lBER BUNDLE HAS BEEN EXHIBITED IN
THE APPLICATION OF THE THEORYS TOPOLOGICAL SOLUTIONS SUCH AS THE INSTANTONS
;= WHICH CAN BE SEEN TO PLAY AN IMPORTANT ROLE IN THE DYNAMICS !NOTHER
<STRONGLY GEOMETRICAL FEATURE WAS REVEALED BY * 4HIERRY -IEG ;= IN HIS
INDENTIlCATION OF THE UNITARITY PRESERVING GHOST lELDS WITH THE <VERTICAL
COMPONENTS OF THE CONNECTION WITH THE <"234 EQUATIONS CORRESPONDING
TO #ARTANS STRUCTURAL EQUATION FOR BUNDLE STATING THE <HORIZONTALITY OF
THE CURVATURE -OREOVER THE SAME SOLUTIONS HAVE BEEN EXPLOITED BY THE
GEOMETRICIANS ;= AND HAVE PRODUCED THE lRST SERIOUS ADVANCE IN MANY
YEARS IN THE EXPLORATION AND CLASSIlCATION OF DIMENSIONAL MANIFOLDS 4HIS
STUDY HAS PRODUCED ENTIRELY NEW TYPES AND <EXOTIC MANIFOLDS ;= &URTHER
PROGRESS HAS BEEN ACHIEVED IN THIS EXPLORATION BY EXPLOITING THE METHODS
DEVELOPED IN ;=
4HE 3ECOND 'EOMETRIZATIONS ,OCAL 0HASE B THE

%LECTROWEAK 4HEORY
7E HAVE SEEN THAT [ UP TO A FORM OF SPONTANEOUS SYMMETRY BREAKDOWN [ THE
CHARGED WEAK CURRENTS ARE BASICALLY CONSERVED AND COUPLED UNIVERSALLY INDI
9 .%%-!. '!&!
CATING THE PRESENCE OF A LOCAL GAUGE SYMMETRY IE A 9ANG -ILLS lBER BUNDLE
GEOMETRY [ EXCEPT THAT THE GROUP WAS NOT YET IDENTIlED 4HERE WAS FOR A
WHILE SOME CONFUSION CAUSED BY THE COEXISTENCE OF RELATED TRANSITIONS AS IN
THE BETA DECAYS OF THE NEUTRON ) N PE vE AND OF THE LAMBDA HYPERON
)) c P E vE AS AGAINST MUON DECAY ))) u_ E_ vu vE WHERE
THE MEASURED COUPLINGS APPEAR IN THE PROPORTION WHEN NOR
MALIZED TO ))) 4HE PROBLEM WAS SOLVED BY . #ABIBBO ;= WHO SHOWED
THAT THE ANSWER LAY IN THE DElNITION OF THE DECAYING STATE WHICH SHOULD BE
N ;N COS q c SIN q= P E vvE WITH #ABIBBO ANGLE q !PPLYING
THE QUARK MODEL MEANS REPLACING D N S c D N &ROM THE POINT
OF VIEW OF 2ELATIVISTIC 1UANTUM &IELD 4HEORY IT IS THE STATE D SEEN BY THE
7EAK )NTERACTION WHICH CORRESPONDS TO THE ORIGINAL AND AS YET <UNMIXED
MATTER lELDS HERE THE PRIMED <DOWN QUARK lELD OR ND CURRENT QUARK STATE
AND IT IS THE 3TRONG )NTERACTION WHICH MIXES D AND THE ORTHOGONAL S PRO
DUCING OUR LOW ENERGY 3) VIRTUAL CONSTITUENT QUARK EIGENSTATES D AND S
-ORE CORRECTLY THE 3) RESULTS IN N AND c HADRONS WHICH ARE THEN SEEN AS
COMPOUNDS AND USED TO DElNE THE CONSTITUENT QUARKS .OTE THAT WE NOW
HAVE SIX SUCH QUARK mAVORS WITH FURTHER SUCH 3) MIXING PARAMETRIZED IN THE
#ABIBBO[+OBAYASHI[-ASKAWA MATRIX ;= GENERALIZING THE ANGLE qCABIBBO [
AND AMPLE EXPERIMENTAL CONlRMATION FOR THIS PICTURE
/NCE THE TRANSITIONS WERE CORRECTLY UNDERSTOOD AND SEEING THAT ALL THERE
WAS CONSISTED OF CURRENT DENSITIES W FOR THE CHARGE RAISING AND W_ FOR THE
CHARGE LOWERING OPERATORS WITHIN AN SU WEAK GENERATOR ALGEBRA ONE COULD
ADD THE MISSING NEUTRAL TRANSITIONS THIRD COMPONENT W AND SOME ADDI
TIONAL HYPOTHETICAL PIECE WY 3UCH A PIECE IS CALLED FOR IN THE %LECTROWEAK
CONJECTURE BY 3 7EINBERG AND ! 3ALAM ;= IN WHICH ONE ASSUMES THAT THE
ELECTROMAGNETIC CURRENT ARISES AS A LINEAR COMBINATION OF AW BWY WEM
;= IE AN SU a U ORIGINALLY SUGGESTED BY 'LASHOW ;= 4HE OR
THOGONAL COMBINATION _B W AWY IS THEN A PREDICTION OF THE EXISTENCE
OF ELECTRICALLY NEUTRAL WEAK CURRENTS 4HESE NEW CURRENTS WERE OBSERVED
EXPERIMENTALLY IN 4HE GAUGE GROUP IS THUS 35 a 5 EXCEPT
THAT WE STILL HAVE TO ACCOUNT FOR THE BREAKING OF THIS SYMMETRY GROUP AS
EXHIBITED BY THE MASSLESSNESS OF THE ELECTROMAGNETIC lELD VERSUS THE VERY
MASSIVE MEDIATOR BOSONS THE MASSES WERE PREDICTED AND INDEED FOUND AT
THE PREDICTED VALUE
4HE ANSWER TO THE SYMMETRY BREAKING HERE HAS BEEN GIVEN BY 0 (IGGS
;= 4 +IBBLE AND OTHERS AND IS INSPIRED BY THE ,ANDAU 'INZBURG METHOD
FOR PHASE TRANSITIONS )T REPRESENTS YET ANOTHER TYPE OF SPONTANEOUS SYMME
6ISIONS IN -ATH 094(!'/2%!. !.$ 0,!4/.)# #/.#%04)/.3 2%!,):%$
TRY BREAKDOWN THIS TIME FOR A LOCAL SYMMETRY IT ALSO DESTROYS ASYMPTOTIC
FREEDOM 4HE ENSUING <(IGGS -ODEL INVOLVES A SCALAR lELD hX WITH A
<-EXICAN (AT POTENTIAL 6 _u th WHOSE MINIMIZATION PRODUCES
NON ZERO VALUES hMIN e;u t=_ 4HIS HAS LED TO CORRECT PREDICTIONS
WITH RESPECT TO THE MASSES OF THE VECTOR MESONS CONNECTIONS AND MANY
OTHER FEATURES 4HE SPONTANEOUSLY BROKEN 9ANG -ILLS THEORY HAS TOPOLOG
ICAL SOLUTIONS THE MAIN ONE BEING THE MONOPOLE ;= APPEARING WHEN THE
RESIDUAL INVARIANT SUBGROUP IS 5 TAKING THE SECOND HOMOTOPY GROUP AS
A MAPPING UNTO THE ASYMPTOTIC REGIONS AND THE VACUUM AND APPLYING THE
REDUCTION THEOREM e '( e (
-ORE RECENTLY $ 1UILLENS SUPERCONNECTION ;= AND ! #ONNES NON
COMMUTATIVE GEOMETRY ;= .#' TWO RELATED NEW BRANCHES OF GEOMETRY
HAVE BEEN APPLIED IN LIEU OF THE (IGGS MODEL OF 3PONTANEOUS 3YMMETRY
"REAKDOWN 33" #ONNES AND ,OTT ;= HAVE REPRODUCED THE 7EINBERG
3ALAM THEORY AS A LOCAL SYMMETRY IN .#' AND A MODEL DEVELOPED BACK IN
BY ME ;= AND BY $" &AIRLIE ;= INDEPENDENTLY AND INVOLVING THE
SUPERGROUP 35 IS AN EXAMPLE OF A SUPERCONNECTION ;= AND WAS ALSO
SHOWN TO BE DERIVABLE FROM .#' ;= ) HAVE RECENTLY APPLIED A SIMILAR
TECHNIQUE IN REPRODUCING %INSTEINIAN 'RAVITY FROM 33" OF AN !nNE 'E
OMETRY ;= IN THE FORM OF A SUPERCONNECTION WITH 0 2 AS .#' LOCAL
SYMMETRY
4HE 3TANDARD -ODEL AND "EYOND 4OWARDS A 4HIRD

'EOMETRIZATION
)N THE ABOVE GAUGE THEORIES WERE CONSOLIDATED IN THE 3TANDARD -ODEL
;= 4HIS IS AGAIN A lBER BUNDLE WITH 'AUGE GROUP 3;5 i 5 = 4HE
<3 FOR <SPECIAL CORRELATES THE TWO ORIGINAL SYSTEMS THE LEPTONS HAVE
INTEGER 9W AND NULL IE EVEN TRIALITY 35 COLOR WHEREAS THE QUARKS HAVE
FRACTIONAL 9W AND ODD TRIALITY 7ITHIN THE 5 THE HALF INTEGER 35 W
LEFT CHIRAL LEPTONS HAVE ODD EIGENVALUES OF 9W WHILE LEFT CHIRAL QUARKS
HAVE AN ODD NUMBER OF THIRDS OF THE UNIT FOR 9W WHEREAS THE RIGHT CHIRAL
LEPTONS AND QUARKS IE 35 W SCALARS HAVE AN EVEN NUMBER OF 9W UNITS
FOR LEPTONS AND OF THIRDS OF A UNIT FOR THE QUARKS
)N THE 3- SOME CONSTRAINTS ARE <NATURALLY FULlLLED EG THE VANISHING
OF THE DILATIONAL ANOMALIES OCCURS BECAUSE THE SUM OF THE ELECTRICAL CHARGES
OF THE QUARKS AND LEPTONS IN EACH GENERATION VANISHES 4HIS IS MOST
PROBABLY AN INDICATION THAT THERE IS A LARGER ALGEBRAIC STRUCTURE A '54 [
9 .%%-!. '!&!
GAUGE UNIlED THEORY IN WHICH THIS IS A BASIC POSTULATE !NOTHER FEATURE

WHICH SEEMS TO POINT TO A LARGER SYMMETRY RESULTS FROM THE RENORMALIZATION
GROUP CALCULATIONS SHOWING THAT ALL THREE COUPLINGS 3TRONG [ AND THE TWO
%LECTROWEAK <MEET SOMEWHERE AROUND 'E6 -OREOVER THIS VALUE IS
RATHER CLOSE TO THE 0LANCK ENERGY WHERE GRAVITY DOMINATES [ WHICH MIGHT
HINT AT FURTHER UNIlCATION
4HE NEXT STEP IS INDEED EXPECTED TO BRIDGE BETWEEN THE TWO GEOMETRIZED
THEORIES WE MENTIONED WITH GRAVITY lRST UNDERGOING A FEW ADAPTATIONS )N
SUCH THEORIES GRAVITY lRST HAS TO BE QUANTIZED SO AS TO BE MATCH ABLE WITH
THE 3- A QUANTIZED THEORY FORMULATED AS A 21&4 .OTE THAT EVEN THOUGH
EVERYTHING UP TO 4 E6 IS DESCRIBED BY EITHER '2 OR THE 3- THERE ARE
INDICATIONS OF POSSIBLE NEW INTERACTIONS &OR INSTANCE THE BARE MASSES OF
LEPTONS AND QUARKS ARE CONSIDERED AS INPUT PARAMETERS IN THE 3- THOUGH
THEY MAY BE <THE TIP OF THE ICEBERG OF NEW INTERACTIONS ETC
4WO SETS OF MODELS HAVE BEEN CONSIDERED AS CANDIDATE FRAMEWORKS FOR
FURTHER UNIlCATION SUPERGRAVITY ;= AND A FORMALISM GOING BEYOND 21&4
NAMELY THEORIES OF QUANTIZED EXTENDED BODIES 4HE LATTER ENTERED THE SCENE
WITH THE STRING AND ITS SUPERSYMMETRIC VERSION THE SUPERSTRING ;= AND
HAVE SINCE BEEN GENERALIZED INTO MEMBRANES AND SUPERMEMBRANES AND P
DIMENSIONAL <P BRANES AND <SUPER P BRANES ;= A JOKE ) TRIED
TO lGHT IN A PURIST SPIRIT BY INTRODUCING INSTEAD D DIMENSIONAL <EXTENDONS
WITH THEIR <SUPEREXTENDONS BUT FAILED BADLY )N THESE THEORIES THE RELA
TIVISTIC QUANTUM lELDS CORRESPOND TO A LOW ENERGY TRUNCATION OF THE SUPER
EXTENDON AND THE PARTICLE STATES REPRESENT RESONATING FREQUENCIES OF THESE
GEOMETRICAL ELEMENTS )SNT THIS IN THE SPIRIT OF 0HILOLAOS $EMOKRITOS AND
0LATONS DISCUSSION ;= .OTE THAT THIS THEORY IS FULLY QUANTIZED AND PRO
VIDES A PERTURBATIVE SOLUTION TO THE 1UANTUM 'RAVITY RIDDLE
3UPERGRAVITY THEORIES AS RELATIVISTIC QUANTUM lELD THEORIES 21&4 OF
GRAVITY HAVE YET TO BE QUANTIZED [ THOUGH THIS NEGATIVE STATUS MIGHT NOT
YET REPRESENT THE lNAL VERDICT 4HE SUPEREXTENDON THEORIES HOWEVER ARE
FULLY QUANTIZED AND CONTAIN GRAVITY THUS ALREADY IN QUANTIZED FORM EM
BEDDED IN SOME SUPERGRAVITY FRAMEWORK ALSO QUANTIZED 4HE PRESENTLY
POPULAR SUPEREXTENDON MODEL KNOWN AS <- THEORY ;= CONTAINS '2
EMBEDDED WITHIN THE MOST COMPREHENSIVE SUPERGRAVITY MODEL ;= 4HUS
REPLACING THE 21&4 BY AN <EXTENDED STRUCTURE LIMITED LOCALIZATION PRO
VIDES A DIMENSIONAL CUT Ok BY REPLACING A POINT BY A REGION )N THE OPPOSITE
DIRECTION THE DIVERGENCES ENTER THROUGH THE TRUNCATION PROCESS
3HOULD THIS EXTENDED DIRECTION PROVE TO WIN AS THE NEXT THEORETICAL
6ISIONS IN -ATH 094(!'/2%!. !.$ 0,!4/.)# #/.#%04)/.3 2%!,):%$
CONSOLIDATION AND ADVANCE WE WOULD HAVE COME CLOSE TO THE SPIRIT OF THE
IDEAS WHICH ACCORDING TO THAT CONJECTURE TRIGGERED AND MOTIVATED 0LATONS
FAITH IN GEOMETRY AS A KEY TO PHYSICS NAMELY $EMOKRITOS MODEL INCORPO
RATING 0HILOLAOS IDEA OF THE VARIOUS ATOMS AS MANIFESTATIONS OF FOUR OF THE
lVE PERFECT POLYHEDRA IE MATTER ITSELF BORN OUT OF GEOMETRY SOMEWHAT LIKE
"OTICELLIS 6ENUS EMERGING FROM SEA WAVE RIPPLES )N THE <POST - THEORY
THE BASIC OBJECTS ARE ALSO GEOMETRICAL EXCEPT THAT THEY ALL LIVE IN A SUPER
MANIFOLD WITH $ BOSONIC DIMENSIONS [ AND ARE DISTINGUISHED EACH
BY ITS DIMENSIONALITY NAMELY D FOR THE STRING D FOR THE MEMBRANE
ETC AND TOPOLOGICAL FEATURES SUCH AS <OPEN AND <CLOSED FOR THE STRING
THE GENUS FOR THE MEMBRANE ETC 4HERE IS IN ADDITION AN OVERLAP WITH
NON COMMUTATIVE GEOMETRY !LTOGETHER A THIRD GEOMETRIZATION APPEARS TO
BE IN THE MAKING VERY CLOSE IN SPIRIT TO THE IDEAS LAUNCHED BY 0HILOLAOS
AND DEVELOPED BY $EMOKRITOS AND 0LATON 4HE FORMULATION OF THE THEORY
IS AS YET RATHER AD HOC HOWEVER SOME NEW AND FUNDAMENTAL INVARIANCE
POSTULATE MAY BE MISSING
-EANWHILE THERE IS AN INDEPENDENT EkORT WHICH ALSO APPEARS TO HAVE
ACHIEVED A FORMULATION ;= OF 1UANTUM 'RAVITY BASED ON $IRACS IDEA ;=
OF (AMILTONIAN QUANTIZATION A NON PERTURBATIVE MODE WHICH IS THEREFORE
HARDER TO JUXTAPOSE WITH THE 3- THEORIES 4HE KEY GAUGE INVARIANT VARI
ABLE IS THE 7ILSON LOOP INTEGRAL ALREADY PRESENT IN ( 7EYLS WORK WHICH
SERVES AS A TRANSFORM 2WHICH <INSCRIBES THE CONNECTION !X IN SPACETIME
4! l . _ 4 REXP l ! 4HIS BRINGS IN THE DIkEOMORPHISM INVARIANCE
OF THE LOOP SO THAT ONLY KNOT THEORY CAN PROVIDE THE CLASSIlCATION OF SUCH
1' STATES ;= .OTE THE RESEMBLANCE TO STRINGS IS IT FORTUITOUS !ND
YET THIS APPROACH HAS ALSO BEEN SHOWN TO LEAD TO THE 1UANTIZATION OF 'E
OMETRY [ AREAS AND VOLUMES ;= (ERE TOO THERE IS AN OVERLAP WITH
NON COMMUTATIVE GEOMETRY AND THERE IS ALSO SOME RESEMBLANCE WITH THE
EXTENDONS OF THE SUPERSTRINGS SUCCESSORS - THEORY AND THE LIKE
2ETURNING TO THE ELEMENTS OF A 4HIRD 'EOMETRIZATION WE NOTE THAT
THESE ALREADY APPEAR TO PROVIDE A VIEW BEYOND <0LANCK DIMENSIONS BOTH
AS GEOMETRY AND AS PHYSICS 7ITH CLASSICAL '2 AND <NON RELATIVISTIC
1- WE ALREADY GET A LIMIT ON LOCALIZABILITY WHEN THE 3CHWARZSCHILD RA
DIUS EQUALS THE #OMPTON WAVELENGTH NAMELY AT 0LANCK DIMENSIONS 4HIS
IS FURTHER CLARIlED ;= WHEN APPLYING THE DECOHERENCE INTERPRETATION FOR THE
MEASUREMENT PROCESS IN 1- ;= 7ITH SUPEREXTENDONS IN - THEORY ONE
ACHIEVES A FURTHER INSIGHT 4HE THEORY HAS A SET OF DUALITY SYMMETRY TRANS
FORMATIONS SIMILAR TO BOTH (ODGE DUALITY IN THE 'RASSMANN ALGEBRA AS IN
9 .%%-!. '!&!
ELECTRIC MAGNETIC INVERSIONS AND BOOTSTRAP DUALITY BETWEEN THE DYNAMICAL

VARIABLES AND THE SOLITONIC SOLUTIONS 7ITH MAGNETIC MONOPOLES FULlLLING
$IRACS CONDITION EG N WE OBSERVE THAT A STRENGTHENING OF E THE ELECTRIC
CHARGE WEAKENS THE MAGNETIC CHARGE G AND VICE VERSA 4HIS IMPLIES THAT
THE EkECTIVE COUPLINGS BEHAVE MUTUALLY ACCORDING TO E p R G p R
AND VICE VERSA 4HUS GOING TO SMALLER DIMENSIONS THAN R0 L WILL TAKE US
BACK INTO LARGER DIMENSIONS IN THE DUAL SYSTEM 4HIS IS JUST A lRST GLIMPSE
AT THE THIRD GEOMETRIZATION TWO OTHER RECENT SUCCESSES HAVE CONSISTED OF
THE RECOVERY OF THE MACROSCOPICALLY DERIVED "EKENSTEIN FORMULA ;= FOR
BLACK HOLE ENTROPY [ FROM A STATISTICAL ANALYSIS OF THE DENSITY OF STATES
OF THE GRAVITATIONAL lELD IN THIS 1UANTUM 'RAVITY MODEL ;= AND A
CALCULATION OF 1#$ COLOR CONlNEMENT PARAMETERS FROM SUPERGRAVITY IN
DIMENSIONAL ANTI $E 3ITTER SPACETIME ;= BY APPLYING A TRANSFORMATION
RELATED TO THE 7IGNER )NONU GROUP CONTRACTION /NE MAY GUESS THAT MUCH
FURTHER GEOMETRICAL BLISS AWAITS THE 88)ST #ENTURYS PHYSICISTS
!CKNOWLEDGEMENT ) WOULD LIKE TO THANK MY FRIEND $R -AURICE #O
HEN FOR HIS OUTSTANDING SUPPORT OF A STUDY OF SOME GEOMETRICAL FEATURES OF
PRESENT GAUGE THEORIES BOTH <INTERNAL AND <EXTERNAL
2EFERENCES
;= # 0REAUX IN <#ONNAISSANCE 3CIENTIlQUE ET 0HILOSOPHIE !CADEMIE 2OYALE
DE "ELGIQUE 0ROC #OLLOQUIUM COMMEMORATING TH !NN OF 2OYAL
!CAD OF "ELGIUM "RUXELLES [
;= 9 .EEMAN )NSIGHT AND FORESIGHT IN PHYSICS FROM 0LATO AND !RISTOTLE TO
%INSTEIN AND THE STANDARD MODEL IN <.UCLEAR -ATTER (OT AND #OLD 0ROC
3CHOOL OF 0HYSICS AND !STRONOMY 4EL !VIV 5NIVERSITY !PRIL $ !SHERY
* !LISTER EDS TO APPEAR
;= 7 9OURGRAU 3 -ANDELSTAM 6ARIATIONAL 0RINCIPLES IN $YNAMICS AND
1UANTUM 4HEORY $OVER PUB .EW 9ORK PP
;= (! +ASTRUP IN <3YMMETRIES IN 0HYSICS -' $ONCEL ET
AL EDS 5NIVERSITY OF "ARCELONA 0UB "ARCELONA [
;= 9 .EEMAN 4HE IMPACT OF %MMY .OETHERS THEOREMS ON 88)ST #ENTURY
PHYSICS IN <)SRAEL -ATHEMATICS #ONF 0ROC [
;= 9 .EEMAN $E LAUTOGEOMETRISATION DE LA PHYSIQUE IN <,ES 2ELATIONS EN
TRE LES -ATHEMATIQUES ET LA 0HYSIQUE 4HEORIQUE &ESTSCHRIFT FOR THE TH
!NNIVERSARY OF THE )(%3 [
;= 9 .EEMAN )NmATION THE TOP AND ALL THAT LECTURES AT %RICE .!4/
!DVANCED 3TUDY )NSTITUTE -AY IN <#URRENTS IN (IGH %NERGY !STRO
PHYSICS [
6ISIONS IN -ATH 094(!'/2%!. !.$ 0,!4/.)# #/.#%04)/.3 2%!,):%$
;= 9 .EEMAN 4HE %TERNAL )NmATIONARY 5NIVERSE IN (EBREW 7EIZMANN )N

STITUTE 9OUTH !CTIVITIES 0UBLICATIONS PP
;= 9 .EEMAN 7HERETO ,AMBDA PAPER READ ON !PRIL AT THE 3YM
POSIUM IN !STROPHYSICS AND #OSMOLOGY HELD ON THE OCCASION OF THE TH
!NNIVERSARY OF 7OLFGANG 0RIESTER AT THE "ONN 2ADIOASTRONOMICAL /BSERVA
TORY
;= 9 .EEMAN 9 +IRSCH 4HE 0ARTICLE (UNTERS ND EDITION #AMBRIDGE
5NIVERSITY 0RESS PP 3EE #HAP
;= 9 .EEMAN 0LATO AND THE SELF GEOMETRIZATION OF PHYSICS IN THE TWENTIETH
CENTURY &UNCTIONAL $IkERENTIAL %QUATIONS [
;= %% (ELM 3C !M $EC [
;= % 3CHROEDINGER !NN D 0HYSIK 7 0AULI :EIT 0HYSIK

;= -' -AYER 0HYS 2EV / (AXEL *($ *ENSEN
(% 3UESS 0HYS 2EV
;= *0 %LLIOTT 0ROC 2OY 3OC ,ONDON !
;= 9 $OTHAN - 'ELL -ANN 9 .EEMAN 0HYS ,ETT [
t C
9 .EEMAN $J 3IJA tKI 0HYS ,ETT " [
;= ! !RIMA & )ACHELLO !NN 2EV .UCL 3C
;= 9 .EEMAN .UCL 0HYS - 'ELL -ANN #ALTECH PREPRINT
#43, UNPUB
;= '% "AIRD ,# "IEDENHARN 0ROC ST #ORAL 'ABLES #ONF ON
3YMMETRY 0RINCIPLES AT (IGH %NERGY " +URSUNUGLU ET AL EDS 7(
&REEMAN 0UB
;= ( 'OLDBERG 9 .EEMAN .UOVO #IM - 'ELL -ANN
0HYS ,ETT ' :WEIG #%2. REP 4( UNPUB
;= 3 7EINBERG 0HYS 2EV ,ETT ( &RITZSCH - 'ELL
-ANN IN <0ROC 86)TH )#(%0 #HICAGO
;= ! %INSTEIN !NN D 0HYSIK
;= ( -INKOWSKI ,ECTURE AT TH !SSEMBLY OF 'ERMAN RESEARCHERS IN .ATURAL
3CIENCE AND PHYSICIANS #OLOGNE
;= ! %INSTEIN !NN D 0HYSIK AND
;= ! %INSTEIN *AHRBUCH D 2ADIOACTIVITAET UND %LECTRONIK
;= ! %INSTEIN - 'ROSSMANN :EIT F -ATH U 0HYSIK
;= ! %INSTEIN !NN D 0HYSIK
;= , #ORRY * 2ENN * 3TACHEL 3CIENCE
;= ( 7EYL :EIT & 0HYSIK
;= #. 9ANG 2, -ILLS 0HYS 2EV
;= 3 7EINBERG 0HYS 2EV $
;= 9! 'OLFAND %0 ,IKHTMAN *%40 ,ETTERS * 7ESS
" :UMINO .UCL 0HYS "
9 .%%-!. '!&!
;= 4H +ALUZA 3ITZ 0REU !KAD 7ISS / +LEIN .ATURE

;= % #REMMER " *ULIA 0HYS ,ETT " AND .UCL 0HYS "

;= 0'/ &REUND -! 2UBIN 0HYS ,ETT "
;= *( 3CHWARZ ED 3UPERSTRING 4HEORY VOLS 7ORLD 3CIENTIlC 0UB 3IN
GAPORE PP -" 'REEN *( 3CHWARZ % 7ITTEN 3UPER
STRING 4HEORY VOLS #AMBRIDGE 5 0 PP
;= 0+ 4OWNSEND 0HYS ,ETT "
;= 9 .EEMAN 1UANTIZING GRAVITY AND SPACETIME 7HERE DO WE STAND IN
!NNAL DER 0HYSIK [
;= 9 .EEMAN % %IZENBERG -EMBRANES AND /THER %XTENDONS @P BRANES
7ORLD 3CIENTIlC 3INGAPORE PP
;= & 'U RSEY , 2ADICATI 0HYS 2EV ,ETT SEE ALSO
& $YSON 3YMMETRY 'ROUPS IN .UCLEAR AND 0ARTICLE 0HYSICS 7! "EN
JAMIN PUB .EW 9ORK PP
;= 9 .EEMAN 0HYS 2EV " AND 4HE ORIGINAL lFTH INTERAC
TION IN <%LEMENTARY 0ARTICLES AND THE 5NIVERSE %SSAYS IN HONOR OF -URRAY
'ELL -ANN 0ROC !NIV #ONF 0ASDENA *AN *( 3CHWARTZ ED
[
;= . 3EIBERG % 7ITTEN .UCL 0HYS " IDEM "

;= !- 0OLYAKOV 0HYS ,ETT " ' T (OOFT 0HYS 2EV ,ETT

;= * 4HIERRY -IEG * -ATH 0HYS
;= 3 $ONALDSON * $Ik 'EOM - &RIEDMAN * $Ik 'EOM

;= 2& 'OMPF * $Ik 'EOM
;= . #ABIBBO 0HYS 2EV ,ETT - +OBAYASHI + -ASKAWA
0ROG 4HEOR 0HYS
;= 3 7EINBERG 0HYS 2EV ,ETT ! 3ALAM IN <%LEMENTARY
0ARTICLE 4HEORY . 3VARTHOLM ED !LMQUIST 6ORLAG !" 3TOCKHOLM
;= 3, 'LASHOW .UCL 0HYS
;= 0 (IGGS 0HYS ,ETT & %NGLERT 2 "ROUT 0HYS 2EV
,ETT
;= ' T (OOFT .UCL 0HYS " !- 0OLYAKOV *%40 ,ETT

;= $ 1UILLEN 4OPOLOGY
;= ! #ONNES .ONCOMMUTATIVE 'EOMETRY !CADEMIC 0RESS .9 PP
;= ! #ONNES * ,OTT .UCL 0HYS " 0ROC 3UPPL
;= 9 .EEMAN 0HYS ,ETT "
;= $ &AIRLIE 0HYS ,ETT "
6ISIONS IN -ATH 094(!'/2%!. !.$ 0,!4/.)# #/.#%04)/.3 2%!,):%$
;= 3 3TERNBERG 9 .EEMAN 3UPERCONNECTIONS AND INTERNAL SUPERSYMMETRY

DYNAMICS 0ROC .AT !CAD 3CI 53! [
;= 2 #OQUEREAUX 2 (A USSLING .! 0APADOPOULOS & 3CHECK )NT
* -OD 0HYS ! 2 #OQUEREAUX ' %SPOSITO &ARs ESE
& 3CHECK )NT * -OD 0HYS !
;= 9 .EEMAN 0HYS ,ETT " [
;= $: &REEDMAN 0V .IEUWENHUIZEN 3 &ERRARA 0HYS 2EV $
3 $ESER " :UMINO 0HYS ,ETT "
;= 0! #OLLINS 07 4UCKER .UCL 0HYS "
;= 9 .EEMAN $ 3IJAt C tKI 0HYS ,ETT [ 9 .EEMAN
t C
$ 3IJA tKI "
;= 0LATO 4IMMEOS
;= % #REMMER " *ULIA .UCL 0HYS "
;= ! !SHTEKAR 0HYS 2EV ,ETT
;= 0!- $IRAC #AN * -ATH 0ROC 2OYAL 3OC !

;= * "AEZ ED +NOTS AND 1UANTUM 'RAVITY /XFORD 3CIENCE PUB #LARENDON
0RESS /XFORD
;= # 2OVELLI , 3MOLIN GR QC
;= 9 .EEMAN 0HYS ,ETT !
;= ($ :EH &OUND 0HYS
;= * "EKENSTEIN ,ETT .UOV #IMENTO
;= ! 3TROMINGER # 6AFA 0HYS ,ETT "
;= * -ALDACENA 4HE LARGE . LIMIT OF SUPERCONFORMAL lELD THEORIES AND SU
PERGRAVITY HEP TH 7ILSON LOOPS IN LARGE . lELD THEORIES HEP TH

9UVAL .EEMAN 3ACKLER &ACULTY OF %XACT 3CIENCES 0HYSICS 4EL !VIV 5NIVERSITY
4EL !VIV )SRAEL
AND
#ENTER FOR 0ARTICLE 0HYSICS 5NIVERSITY OF 4EXAS !USTIN 48 53!
#,!33)#!, !.$ -/$%2. 4/0/,/'9

4/0/,/')#!, 0(%./-%.! ). 2%!, 7/2,$
0(93)#3
3ERGEY 0 .OVIKOV
!CCORDING TO THE !NCIENT 'REEKS THE FAMOUS REAL AND MYTHICAL FOUNDERS OF
-ATHEMATICS AND .ATURAL 0HILOSOPHY SUCH AS 0YTHAGORAS !RISTOTLE AND OTHERS IN
FACT BORROWED THEM FROM THE %GYPTIAN AND -IDDLE %ASTERN CIVILIZATIONS (OWEVER
WHAT HAD BEEN TOLD BEFORE IN THE HIDDEN MYSTERIES 'REEK SCIENTISTS TRANSFORMED INTO
WRITTEN INFORMATION ACCEPTABLE TO EVERYBODY )MMEDIATELY AFTER THAT THE DEVELOP
MENT OF SCIENCE IN THE MODERN SENSE STARTED AND HAD ALREADY REACHED A VERY HIGH
LEVEL YEARS AGO 4HEREFORE YOU MAY SAY THAT THE FREE EXCHANGE OF INFORMA
TION AND MAKING IT CLEAR TO PEOPLE WERE THE 'REEKS MOST IMPORTANT DISCOVERIES
) WOULD SAY IT IS THE BASIS OF OUR SCIENCE NOW !S YOU WILL SEE ANY VIOLATION OF
THIS FUNDAMENTAL RULE DOES SERIOUS HARM TO OUR SCIENCE AND INEVITABLY LEADS TO ITS
DECAY
#LASSICAL AND -ODERN 4OPOLOGY

0REHISTORY 4HE lRST lFTY YEARS OF 4OPOLOGY 4HE lRST IMPORTANT
TOPOLOGICAL IDEAS WERE OBSERVED BY FAMOUS MATHEMATICIANS AND PHYSICISTS
SUCH AS %ULER 'AUSS +ELVIN -AXWELL AND THEIR PUPILS DURING THE 86)))TH
AND 8)8TH #ENTURIES !S EVERYBODY KNOWS IT WAS 0OINCARsE WHO REALLY
STARTED 4OPOLOGY AS A BRANCH OF -ATHEMATICS IN THE LATE 8)8TH #ENTURY
-ANY TOP CLASS MATHEMATICIANS PARTICIPATED IN THE DEVELOPMENT OF 4OPOL
OGY IN THE lRST HALF OF OUR CENTURY ! HUGE NUMBER OF MUTUALLY CONNECTED
FUNDAMENTAL NOTIONS WERE INVENTED DEGREE OF MAPS AND SINGULARITIES OF VEC
TOR lELDS HOMOTOPY AND HOMOLOGY GROUPS DIkERENTIAL FORMS AND SMOOTH
MANIFOLDS THE FUNDAMENTAL IDEA OF TRANSVERSALITY THE SIMPLICIALCELL#7
AND SINGULAR COMPLEXES AS TOOLS FOR STUDYING TOPOLOGICAL INVARIANTS BRAIDS
KNOT INVARIANTS AND MANIFOLDS COVERINGS lBRE BUNDLES AND CHARACTERIS
TIC CLASSES AND MANY OTHERS $EEP CONNECTIONS WITH 1UALITATIVE !NALYSIS
#ALCULUS OF 6ARIATIONS #OMPLEX 'EOMETRY AND $YNAMICAL 3YSTEMS WERE
4HIS WORK IS SUPPORTED BY .3& 'RANT $-3
6ISIONS IN -ATH 4/0/,/'9
ESTABLISHED IN THIS PERIOD #OMBINATORIAL 'ROUP 4HEORY AND (OMOLOGICAL

#ALCULUS STARTED FROM TOPOLOGICAL SOURCES ! GREAT NEW lELD OF TOPOLOGICAL
OBJECTS UNKNOWN TO THE CLASSICAL MATHEMATICS OF THE 8)8TH #ENTURY AP
PEARED lNALLY IN THE S !T THAT TIME THIS NEW AREA WAS KNOWN TO A
FEW MATHEMATICIANS ONLY (OWEVER THERE WAS A VERY HIGH DENSITY OF REALLY
OUTSTANDING SCIENTISTS AMONG THEM
S AND S 4HE 'OLDEN !GE OF #LASSICAL 4OPOLOGY 4HE FUN
DAMENTAL SET OF ALGEBRAIC IDEAS UNIFYING ALL THESE BRANCHES OF MATHEMATICS
APPEARED IN THE S A NEW ERA STARTED ABOUT 3PECTRAL SEQUENCES OF
lBRE BUNDLES SHEAVES HIGHLY DEVELOPED HOMOLOGICAL ALGEBRA OF THE GROUPS
ALGEBRAS AND MODULES (OPF ALGEBRAS AND COALGEBRAS WERE INVENTED AND
HEAVILY USED FOR THE CALCULATION OF TOPOLOGICAL INVARIANTS NEEDED FOR THE SO
LUTION OF THE FUNDAMENTAL PROBLEMS OF TOPOLOGY ,ET ME POINT OUT THAT IN
MANY CASES IT WAS A COMPLETELY NEW TYPE OF CALCULATIONS BASED ON THE DEEP
COMBINATION OF THE VERY GENERAL <CATEGORIAL PROPERTIES OF THESE QUANTITIES
WITH VERY CONCRETE GEOMETRIC ALGEBRAIC OR ANALYTICAL STUDY OF A COMPLETELY
NEW TYPE )N THE PREVIOUS PERIOD PEOPLE HAD NOT EVEN DREAMT ABOUT HOW
THEY COULD BE CALCULATED 2EGULAR METHODS WERE BUILT TO CALCULATE HOMO
TOPY GROUPS FOR EXAMPLE )T WAS ONE OF THE MOST DInCULT PROBLEMS OF
TOPOLOGY ! LOT OF THEM WERE COMPUTED COMPLETELY OR PARTIALLY INCLUDING
THE HOMOTOPY GROUPS OF SPHERES ,IE GROUPS AND HOMOGENEOUS SPACES 4HE
TOPOLOGICALLY IMPORTANT COBORDISM RINGS WERE COMPUTED AND USED IN MANY
TOPOLOGICAL INVESTIGATIONS 4HE FAMOUS SIGNATURE FORMULA FOR DIkERENTIABLE
MANIFOLDS WAS DISCOVERED )T HAS INNUMERABLE OF APPLICATIONS IN THE TOPOL
OGY OF MANIFOLDS "ESIDES THAT THIS FORMULA PLAYED A KEY ROLE IN THE PROOF
OF THE SO CALLED 2IEMANN 2OCH THEOREM IN !LGEBRAIC 'EOMETRY AND LATER
IN THE STUDY INDEX THE FAMOUS HOMOTOPY INVARIANT OF &REDHOLM OPERATORS
4HE MUTUAL INmUENCE OF 4OPOLOGY AND !LGEBRAIC 'EOMETRY DURING THAT
PERIOD LED TO THE BROAD EXTENSION OF THE IDEAS OF HOMOLOGY THE EXTRAORDI
NARY CO HOMOLOGY THEORIES LIKE + THEORY AND COBORDISMS APPEARED 4HEY
BROUGHT A NEW TYPE OF TECHNIC TO TOPOLOGY WITH MANY APPLICATIONS 2EP
RESENTATION THEORY AND COMPLEX GEOMETRY OF MANIFOLDS DEEPLY UNIlED WITH
HOMOLOGICAL ALGEBRA AND (OPF ALGEBRAS 4HE TECHNIC OF FORMAL GROUPS AP
PEARED HERE )T HAS BEEN APPLIED IN PARTICULAR FOR THE IMPROVEMENT OF THE
CALCULATIONS OF STABLE HOMOTOPY GROUPS OF SPHERES !S EVERYBODY KNOWS
DURING THIS PERIOD TOPOLOGY SOLVED THE MOST FUNDAMENTAL PROBLEMS IN THE
THEORY OF MULTIDIMENSIONAL SMOOTH MANIFOLDS
.ONTRIVIAL DIkERENTIABLE STRUCTURES ON MULTIDIMENSIONAL SPHERES WERE
30 ./6)+/6 '!&!
DISCOVERED ON THE BASIS OF THE RESULTS OF ALGEBRAIC TOPOLOGY COMBINED WITH A

NEW UNDERSTANDING OF THE GEOMETRY OF MANIFOLDS AND BUNDLES 4HE MULTI
DIMENSIONAL ANALOGUE OF THE 0OINCARsE #ONJECTURE AND ( COBORDISM THEO
REM WERE PROVED #OUNTEREXAMPLES TO THE SO CALLED <(AUPTVERMUTUNG DER
4OPOLOGIE WERE FOUND ! CLASSIlCATION THEORY FOR THE MULTIDIMENSIONAL
SMOOTH AND FOR 0, MANIFOLDS AS WELL WAS COMPLETELY CONSTRUCTED 4HE
ROLE OF THE FUNDAMENTAL GROUP IN THIS THEORY LED TO THE DEVELOPMENT OF A NEW
BRANCH OF ALGEBRA THE ALGEBRAIC + THEORY 4OPOLOGICAL INVARIANCE OF THE
MOST FUNDAMENTAL CHARACTERISTIC CLASSES WAS lNALLY PROVED 4HE SO CALLED
<!NNULUS #ONJECTURE WAS PROVED .O MATTER HOW ELEMENTARY THESE RESULTS
CAN BE FORMULATED NOBODY HAS SUCCEEDED TO AVOID THE USE OF A WHOLE BUNCH
OF RESULTS AND TOOLS OF ALGEBRAIC AND DIkERENTIAL TOPOLOGY IN THE PROOF 4HE
CLASSIlCATION THEORY FOR THE IMMERSIONS OF MANIFOLDS WAS CONSTRUCTED 4HE
THEORY OF MULTIDIMENSIONAL KNOTS WAS CONSTRUCTED 3EVERAL CLASSICAL PROB
LEMS OF THE THEORY OF MANIFOLDS ALSO WERE SOLVED DURING THAT PERIOD THE
SO CALLED $EHNS PROGRAM WAS lNISHED AFTER A YEAR BREAK THE ALGORITHM
FOR RECOGNIZING THE TRIVIAL KNOT IN THREE SPACE HAS BEEN THEORETICALLY CON
STRUCTED AS A PART OF THE DEEP UNDERSTANDING OF THE STRUCTURE OF MANIFOLDS
AND THE SURFACES IN THEM !S A BY PRODUCT OF TOPOLOGY THE FUNDAMENTAL
BREAKTHROUGH IN THE TOPOLOGICAL UNDERSTANDING OF GENERIC DYNAMICAL SYS
TEMS WAS REACHED ! NEW GREAT PERIOD STARTED IN THIS AREA 1UALITATIVE
THEORY OF FOLIATIONS HAS BEEN CONSTRUCTED WITH ESPECIALLY DEEP RESULTS FOR
MANIFOLDS
!S A SUMMARY ) WOULD LIKE TO ADD ONE MORE VERY IMPORTANT CHARACTER
ISTIC OF THE TOPOLOGICAL COMMUNITY IN THE GOLDEN AGE OF CLASSICAL TOPOLOGY
!LL IMPORTANT WORKS HAVE BEEN CAREFULLY CHECKED )F SOME
THEOREM HAD NOT BEEN PROVED IT IMMEDIATELY BECAME KNOWN TO
EVERYBODY
3O YOU CAN lND A FULL SET OF PROOFS IN THE LITERATURE 5NFORTUNATELY A
FULL SET OF TEXTBOOKS COVERING ALL THESE DEVELOPMENTS HAS NOT
BEEN WRITTEN YET -ANY MODERN TEXTBOOKS ARE WRITTEN IN A VERY ABSTRACT
WAY %VEN IF THEY COVER SOME PIECES FORMALLY IT IS MORE DInCULT TO READ
THEM THAN THE ORIGINAL PAPERS ,ET ME RECOMMEND TO YOU THE %NCYCLOPEDIA
ARTICLE ;= WRITTEN EXACTLY FOR THE EXPOSITION OF THESE IDEAS
S 0ERIOD OF DECAY )N MY OPINION THE PERIOD OF THE S CAN
BE CHARACTERIZED AS A PERIOD OF DECAY FOR CLASSICAL TOPOLOGY 4HERE ARE
MANY INDICATIONS FOR THAT 3EVERAL LEADING SCIENTISTS LEFT TOPOLOGY FOR NEW
AREAS SUCH AS ALGEBRA AND NUMBER THEORY 2IEMANNIAN AND SYMPLECTIC GE
OMETRY DYNAMICAL SYSTEMS AND COMPLEXITY THEORY FUNCTIONAL ANALYSIS AND

REPRESENTATIONS 0$%S AND DIkERENT BRANCHES OF MATHEMATICALTHEORETICAL
PHYSICS )T IS CERTAINLY A GOOD CHARACTERIZATION OF THE COMMUNITY IF IT
COULD GENERATE SUCH A mUX OF SCIENTISTS IN MANY DIkERENT AREAS AND BRINGING
TO THEM COMPLETELY NEW IDEAS !NYWAY THIS COMMUNITY DISPERSED
7HAT CAN WE SAY ABOUT THE TOPOLOGICAL COMMUNITY AFTER THAT
&IRST OF ALL SOME IMPORTANT NEW IDEAS APPEARED IN THE S EG LOCAL
IZATION TECHNIQUE IN HOMOTOPY TOPOLOGY THE NICELY ORGANIZED THEORY OF THE
RATIONAL HOMOTOPY TYPE HYPERBOLIC TOPOLOGY OF MANIFOLDS WERE DEVELOPED
BY 3ULLIVAN AND 4HURSTON AMONG OTHERS (OWEVER A HUGE INFORMATIONAL
MESS WAS CREATED IN THE S ,ET ME POINT OUT THAT A SERIES OF FUN
DAMENTAL RESULTS OF THAT PERIOD WAS NOT WRITTEN WITH FULL PROOF UNTIL NOW
,ET ME GIVE YOU A LIST
3ULLIVANS (AUPTVERMUTUNG THEOREM WAS ANNOUNCED lRST IN EARLY
!FTER THE CAREFUL ANALYSIS MADE BY "ILL "ROWDER AND MYSELF IN 0RINCETON
OF THE lRST VERSION IN -AY BEFORE PUBLICATION HIS THEOREM WAS COR
RECTED A NECESSARY RESTRICTION ON THE TORSION OF THE GROUP ( - : WAS
MISSING 4HIS GAP WAS FOUND AND RESTRICTION WAS ADDED &ULL PROOF OF
THIS THEORY HAS NEVER BEEN WRITTEN AND PUBLISHED )NDEED NOBODY KNOWS
WHETHER IT HAS BEEN lNISHED OR NOT 7HO KNOWS WHETHER IT IS COMPLETE OR
NOT 4HIS QUESTION IS NOT CLARIlED PROPERLY IN THE LITERATURE -ANY PIECES
OF THIS THEORY WERE DEVELOPED BY OTHER TOPOLOGISTS LATER THEY SOMETIMES
USED DIkERENT IDEAS .OBODY HAS UNIlED THEM UNTIL NOW )NDEED THESE
RESULTS WERE USED BY MANY OTHERS LATER )N PARTICULAR THE lNAL +IRBY
3IEBENMANN CLASSIlCATION OF TOPOLOGICAL MULTIDIMENSIONAL MANIFOLDS THERE
FORE IS NOT PROVED YET IN THE LITERATURE
4HE SECOND STORY IS THE THEORY OF ,IPSCHITZ STRUCTURES ON MANIFOLDS
)N THE MID SEVENTIES 3ULLIVAN DISTRIBUTED A PREPRINT CONTAINING THE IDEA OF
HOW TO PROVE EXISTENCE AND UNIQUENESS OF SUCH STRUCTURES ON THE MANIFOLD
. N N 4HIS IDEA OBVIOUSLY INCLUDED FOR UNIQUENESS THE DIRECT USE OF
THE !NNULUS #ONJECTURE AND THEREFORE OF ALL IDEAS AND TECHNIQUES NEEDED
IN THE PROOF OF TOPOLOGICAL INVARIANCE OF THE RATIONAL 0ONTRYAGIN #LASSES
INSIDE 0ROOF OF THE ,IPSCHITZ THEORY HAS NEVER BEEN PUBLISHED )NDEED
MANY YEARS LATER ALREADY IN THE S SOME BRILLIANT YOUNGER SCIENTISTS
DEVELOPED A VERY NICE THEORY OF &REDHOLM ELLIPTIC OPERATORS ON ,IPSCHITZ
MANIFOLDS !S A COROLLARY THEY CLAIMED THAT A NEW PROOF OF TOPOLOGICAL
INVARIANCE OF RATIONAL 0ONTRYAGIN CLASSES HAS BEEN OBTAINED FROM !NALYSIS
THIS WAS A PROBLEM POSED BY 3INGER IN THE S 9OUNG SCIENTISTS MADE
30 ./6)+/6 '!&!
A <LOGICAL CIRCLE BELIEVING IN THE CLASSICAL RESULTS .OBODY TOLD THEM THAT
THE CORRESPONDING THEOREMS HAD NEVER BEEN PROVED (OW COULD IT HAPPEN
4HIS FUNNY STORY SHOWS THE MODERN STATE OF INFORMATION IN THE TOPOLOGICAL
COMMUNITY
!NOTHER INFORMATIONAL MESS HAS BEEN CREATED IN $ HYPERBOLIC TOPOLOGY
4HIS BEAUTIFUL AREA WAS STARTED BY 4HURSTON IN THE MID S &OR MANY YEARS
PEOPLE COULD NOT lND OUT WHAT WAS PROVED HERE )N THIS AREA THE SITUATION
HAS BEEN lNALLY RESOLVED IT HAS BEEN ACKNOWLEDGED THAT THESE METHODS
LEAD TO THE PROOF OF THE ORIGINAL CLAIM THE SO CALLED 'EOMETRIZATION #ON
JECTURE ONLY FOR THE SPECIAL CLASS OF (AKEN MANIFOLDS 4HE 'EOMETRIZATION
#ONJECTURE MEANS MORE OR LESS THAT IN THE CASE OF CLOSED MANIFOLDS THE
FUNDAMENTAL GROUP CAN BE REALIZED AS A DISCRETE SUBGROUP ACTING IN THE $
HYPERBOLIC SPACE IF TRIVIAL NECESSARY CONDITIONS ARE SATISlED ALL ITS ABELIAN
SUBGROUPS ARE CYCLIC AND x (OWEVER IT IS DInCULT TO lND OUT WHO
ACTUALLY PROVED THIS THEOREM )T SEEMS FOR ME THAT THE YOUNGER MATHEMATI
CIANS WHO MANAGED TO lNISH THIS PROGRAM DID NOT RECEIVE PROPER CREDIT
) WOULD LIKE TO MENTION THAT THIS KIND OF INFORMATIONAL MESS HAS HAP
PENED SINCE NOT ONLY IN TOPOLOGY &OR EXAMPLE THE FAMOUS RESULTS OF
+!- IN THE THREE BODY PROBLEM KNOWN SINCE THE EARLY S WERE FOUND RE
CENTLY UNPROVED )T WAS ANNOUNCED FOR THE lRST TIME AT THE "ERLIN #ONGRESS
LAST YEAR SEE THE PLENARY TALKS OF - (ERMANN AND * -OSER +OLMOGOROV
AND -OSER NEVER CLAIMED SOLUTION OF THIS SPECIlC DEGENERATE PROBLEM IN
THEIR WORKS )N THIS CASE SOME WORKS SUPPOSEDLY CONTAINING FULL PROOF
WERE PUBLISHED IN THE lRST HALF OF THE S $OES THIS MEAN THAT NOBODY
ACTUALLY READ THEM FOR AT LEAST YEARS
$O YOU THINK THAT ALGEBRA IS BETTER ,ET ME TELL YOU AS A CURIOUS
REMARK THAT ALL WORKS OF THE 3TEKLOV )NSTITUTE IE 3HAFAREVICHS SCHOOL IN
ALGEBRAIC NUMBER THEORY ALGEBRAIC GEOMETRY AND THEORY OF lNITE P GROUPS
AWARDED BY THE HIGHEST ,ENIN AND 3TATE PRIZES IN THE FORMER 3OVIET 5NION
SINCE DID NOT CONTAIN FULL PROOF 4HE GAPS IN THE PROOFS WERE FOUND
MANY YEARS LATER .OT ALL THESE GAPS WERE REALLY DEEP (OWEVER SOME OF
THESE AUTHORS KNEW THEIR MISTAKES MANY YEARS BEFORE THEY BECAME PUBLICLY
KNOWN AND COULD NOT CORRECT THEM 4HEY MANAGED TO lLL GAPS AFTER MANY
YEARS USING MUCH LATER TECHNICAL ACHIEVEMENTS MADE BY OTHER PEOPLE $OES
IT MEAN THAT IN THE CORRESPONDING TIME DESPITE MANY PUBLIC PRESENTATIONS
NOBODY IN FACT READ THESE GREAT WORKS #AN WE SAY THAT ALL PROOFS ARE
KNOWN NOW IN ALL THESE CASES
4HERE ARE MUCH WORSE CASES IN MODERN ALGEBRA INDEED (OW MANY OF
YOU KNOW THAT THE SO CALLED CLASSIlCATION OF SIMPLE lNITE GROUPS DID NOT
EXIST AS A MATHEMATICAL THEOREM UNTIL NOW )N THIS CASE WE CAN EVEN SAY
THAT IN FACT AS A FEW REAL EXPERTS HAVE KNOWN SINCE NOT ONE WORK
EXISTED CLAIMING THAT THIS PROBLEM WAS lNISHED IN THIS WORK !LL PUBLIC
OPINION HAS BEEN BASED ONLY ON THE <.EW 9ORK 4IMES 4HEOREM FOR THE
PAST YEARS
S AND S 0ERIOD OF RECOVERY 4HE ROLE OF 1UANTUM &IELD

4HEORY )T BECAME CLEAR ALREADY IN THE LATE S THAT MODERN QUANTUM
lELD THEORY STARTED TO GENERATE NEW IDEAS IN TOPOLOGY )T GAVE SEVERAL
NEW ALTERNATIVE WAYS TO CONSTRUCT TOPOLOGICAL INVARIANTS 0ATH INTEGRAL FOR
METRIC INDEPENDENT ACTIONS ON MANIFOLDS WAS USED FOR THE lRST TIME 4HE
FAMOUS SELF DUALITY EQUATION APPEARED lRST IN THE WORKS OF PHYSICISTS )T
WAS APPLIED IN THE S FOR THE SOLUTION OF FUNDAMENTAL TOPOLOGICAL PROBLEMS
IN THE THEORY OF MANIFOLDS 1UANTUM STRING THEORY BROUGHT IN THE EARLY
S NEW DEEP RESULTS IN THE THEORY OF THE CLASSICAL &UCHSIAN GROUPS AND
MODULI SPACES !T lRST PHYSICISTS LIKE T(OOFT AND 0OLYAKOV WERE NOT
VERY INTERESTED IN SUCH BY PRODUCTS OF THEIR ACTIVITY 4HEY ALWAYS SAID THAT
THEY WERE DOING PHYSICS OF THE REAL WORLD NOT PURE MATHEMATICS (OWEVER
THE NEXT WAVE OF BRILLIANT PHYSICISTS LIKE 7ITTEN 6AFA AND OTHERS STARTED
TO SOLVE PROBLEMS OF PURE MATHEMATICS 3UCH PURELY TOPOLOGICAL SUBJECTS
AS -ORSE THEORY AND COBORDISM THEORY ASSOCIATED WITH ACTION OF COMPACT
GROUPS ON MANIFOLDS WERE DEVELOPED IN THE S FROM THE COMPLETELY NEW
POINT OF VIEW 3YMPLECTIC 4OPOLOGY REACHED A VERY HIGH LEVEL IN THE LATE
S 7E ARE FACING NOW IMPRESSIVE DEVELOPMENT OF #ONTACT 4OPOLOGY
#ERTAINLY 1UANTUM 4HEORY BROUGHT NEW BEAUTIFUL IDEAS "ESIDES THAT
THE FUNDAMENTAL NEW INVARIANTS OF KNOTS WERE DISCOVERED IN THE S BY THE
TOPOLOGISTS WHO CAME FROM FUNCTIONAL ANALYSIS AND THEORY OF # b ALGEBRAS
4HESE INVARIANTS ALSO RECEIVED QUANTUM TREATMENT IN THE LATE S 4HE
BEAUTIFUL CONNECTION OF THE SPECIlC &EINMANN DIAGRAMS WITH SURFACES WAS
BORROWED FROM PHYSICS LITERATURE )T BECAME A VERY EkECTIVE TOOL FOR THE
SOLUTION OF SEVERAL TOPOLOGICAL PROBLEMS 5NFORTUNATELY ONLY A FEW MATH
EMATICIANS LEARNED THIS TECHNIQUE AND STARTED TO APPLY IT IN TOPOLOGY )
KNOW ONLY 3INGER +ONTSEVICH AND A VERY SMALL NUMBER OF OTHERS %VEN IF
YOU WILL ADD HERE THE NAMES OF PURE MATHEMATICIANS WHO LEARNED THIS WITH
THE INTENTION TO DO REAL PHYSICS THIS LIST WILL INCREASE INESSENTIALLY ) DO NOT
COUNT HERE PEOPLE WHO WERE TRAINED ORIGINALLY IN THE PHYSICS COMMUNITY
! LARGE NUMBER OF THEM MOVED INTO PURE MATHEMATICS WITH THE INTENTION
TO PROVE RIGOROUS THEOREMS ABOUT THE MODELS SERVING IN THEIR OPINION AS
30 ./6)+/6 '!&!
AN IDEALIZATION OF THEORETICAL PHYSICS 4HEY CALL THIS AREA -ATHEMATICAL

0HYSICS BUT NOT EVERYBODY AGREES WITH SUCH A DElNITION OF MATHEMATICAL
PHYSICS 4HIS COMMUNITY DOES NOT DO TOPOLOGY
) WOULD LIKE TO MAKE A REMARK HERE CONCERNING A BEAUTIFUL WORK OF
+ONTSEVICH CALCULATING CERTAIN #HERN NUMBERS ON THE PUNCTURED MODULI
SPACES OF 2IEMANN 3URFACES THROUGH THE SPECIAL SOLUTION TO THE +D6 HIER
ARCHY 4HIS FORMULA HAS BEEN KNOWN AS A 7ITTEN #ONJECTURE 9OU HAVE TO
SPECIFY FOR THIS SOME COMPACTIlCATION OF THE MODULI SPACES OF PUNCTURED
2IEMANN SURFACES OTHERWISE IT MAKES NO SENSE +ONTSEVICH ACTUALLY PROVED
THIS FORMULA FOR ONE SPECIlC <3TROBEL 0ENNER COMPACTIlCATION IN
7HAT ABOUT THE STANDARD $ELIGNE -UMFORD COMPACTIlCATION +ONTSEVICH
CLAIMED IN IN HIS WORK IN )NVENTIONES THAT IT IS TRUE (OWEVER NO
PROOF HAS BEEN PRESENTED UNTIL NOW 3O THIS PROBLEM IS OPEN 4HERE WAS A
MISTAKABLE STATEMENT ABOUT THIS AT THE "ERLIN #ONGRESS
,ET ME POINT OUT THAT THE PHYSICS COMMUNITY DID NOT CREATE ANY INFOR
MATIONAL MESS IN TOPOLOGY !CCORDING TO THEIR TRAINING TRADITION THEORETICAL
WORK PRODUCES #ONJECTURES WHICH SHOULD BE PROVED ONLY BY SOME KIND OF
EXPERIMENT 3TARTING TO DO BEAUTIFUL NONRIGOROUS MATHEMATICS THEY DO
NOT CLAIM THAT THEY <PROVED SOMETHING 4HEY ARE SAYING THAT THEY <PRE
DICTED THIS FACT )N THE CASE OF PURE MATHEMATICS THE lNAL PROOF DONE BY
PURE MATHEMATICIANS THESE PEOPLE MAY TREAT AS AN <EXPERIMENTAL CONlRMA
TION )N THE PAST TEN YEARS SEVERAL DEEP RESULTS HAVE BEEN OBTAINED IN THE
$ TOPOLOGY 7E CANNOT SAY THIS ABOUT $ TOPOLOGY QUANTUM INVARIANTS
HERE CREATED SOME SORT OF <INVARIANTOLOGY A LOT OF PEOPLE ARE CONSTRUCTING
TOPOLOGICAL INVARIANTS BUT NO ONE NEW TOPOLOGICAL RESULT HAS BEEN OBTAINED
FOR ALMOST YEARS )NDEED THESE IDEAS LOOK BEAUTIFUL IN SOME CASES )N
MY OPINION NEW DEEP RESULTS WILL APPEAR AFTER BETTER UNDERSTANDING OF THE
RELATIONSHIP OF NEW INVARIANTS WITH CLASSICAL TOPOLOGY
4OPOLOGICAL 0HENOMENA IN 2EAL 7ORLD 0HYSICS

4OPOLOGICAL IDEAS IN PHYSICS IN THE PERIOD OF THE EARLY S ) SPENT
ABOUT YEARS LEARNING DIkERENT PARTS OF -ODERN 4HEORETICAL 0HYSICS IN
THE S AND S !FTER JOINING THE PHYSICS COMMUNITY IE ,ANDAU SCHOOL
IN THE EARLY S ) FOUND OUT THAT MOST PHYSICISTS DID NOT KNOW AT ALL THE
NEW AREAS OF MATHEMATICS LIKE TOPOLOGY DYNAMICAL SYSTEMS AND ALGEBRAIC
GEOMETRY INCLUDING ANALYSIS ON 2IEMANN SURFACES 4HE QUANTUM PEOPLE
KNEW SOME EXTRACTS FROM GROUP THEORY AND REPRESENTATIONS BECAUSE THEY
NEEDED IT IN 3OLID 3TATE 0HYSICS AS WELL AS IN %LEMENTARY 0ARTICLES 4HEORY

SINCE THE S ! LOT OF THEM KNEW SOMETHING ABOUT 2IEMANNIAN 'EOM
ETRY BECAUSE OF THE %INSTEINIAN 'ENERAL 2ELATIVITY (OWEVER THESE PEOPLE
HAD ALREADY HEARD SOMETHING ABOUT THE NEW MATHEMATICS OF THE 88TH CEN
TURY AND BADLY WANTED TO lND ITS REALIZATION IN PHYSICS 9OU HAVE TO TAKE
INTO ACCOUNT THAT AMONG THEM THERE WAS A GREAT NUMBER OF EXTREMELY TAL
ENTED PEOPLE AT THAT TIME WITH VERY GOOD TRAINING IN PRACTICAL MATHEMATICS
)N SOME CASES ) WAS ABLE TO HELP PHYSICISTS LIKE 0OLYAKOV $ZYALOSHINSKI
+HALATNIKOV AND SOME OTHERS ESPECIALLY AT THE ,ANDAU )NSTITUTE TO LEARN
AND TO USE TOPOLOGY IN THE S ) WORKED THIS PERIOD IN 'ENERAL 2ELATIVITY
(OMOGENEOUS #OSMOLOGICAL -ODELS AND 0ERIODIC +D6 4HEORY WITH MY
PUPILS AND COLLABORATORS 7E FOUND COMPLETELY NONSTANDARD APPLICATIONS
OF $YNAMICAL 3YSTEMS AND !LGEBRAIC 'EOMETRY IN THESE AREAS (OWEVER
UNTIL THE LATE S ) DID NOT PRODUCE ANY NEW TOPOLOGICAL IDEAS -Y VERY
lRST TOPOLOGICAL WORK IN PHYSICS WAS MADE IN SEE ;= ) STARTED TO
USE IN THE SPECTRAL THEORY OF THE 3CHRODINGER OPERATORS IN PERIODIC LATTICE
AND MAGNETIC lELD THE IDEA OF TRANSVERSALITY APPLIED TO THE FAMILIES OF (ER
MITIAN MATRICES OR ELLIPTIC OPERATORS ON THE TORUS 4HIS IDEA LED TO THE
DISCOVERY OF THE SERIES OF TOPOLOGICAL INVARIANTS #HERN .UMBERS OF $ISPER
SION 2ELATIONS 4HEY ARE WELL DElNED FOR THE GENERIC OPERATORS ONLY 4HE
CLASSICAL 3PECTRAL 4HEORY IN MATHEMATICS NEVER CONSIDERED SUCH QUANTITIES
BECAUSE THEY ARE NOT DElNED FOR EVERY OPERATOR WITH PRESCRIBED ANALYTICAL
PROPERTIES OF COEnCIENTS 4HE IDEOLOGY OF TRANSVERSALITY IS IMPORTANT HERE
4HIS WORK WAS NOT UNDERSTOOD BY MY COLLEAGUES PHYSICISTS AT THAT TIME
THE VICE EDITOR OF *%40 DID NOT WANT TO PUBLISH IT AS <NONPHYSICAL SO
) PUBLISHED IT IN THE MATH LITERATURE 0EOPLE THOUGHT THAT THE IMPORTANT
INTEGER VALUED OBSERVABLE QUANTITIES IN 3OLID 3TATE 0HYSICS MAY COME FROM
SYMMETRY GROUPS ONLY )NDEED THE )NTEGRAL 1UANTUM (ALL PHENOMENON
WAS DISCOVERED SOON 3OME FAMOUS THEORETICAL PHYSICISTS REDISCOVERED MY
MATHEMATICAL IDEA AFTER THAT )T IS CERTAINLY A SUM OF THE #HERN CLASSES OF
DISPERSION RELATIONS BELOW THE &ERMI LEVEL
-Y NEXT TOPOLOGICAL DISCOVERY WAS MADE IN THE JOINT WORK WITH STUDENT
) 3CHMELZER IN DEDICATED TO THE VERY SPECIAL PROBLEM OF CLASSICAL ME
CHANICS AND HYDRODYNAMICS SEE ;= ) IMMEDIATELY REALIZED ITS VALUE FOR
MODERN THEORETICAL PHYSICS AS WELL AS FOR MATHEMATICS AND DEVELOPED THIS
IDEA IN SEVERAL DIRECTIONS IN THE SAME YEAR ;= 4HE SERIES OF WORKS IN
THE 4HEORY OF .ORMAL -ETALS WHICH ) AM GOING TO DISCUSS TODAY IS ALSO
ONE OF BY PRODUCTS OF THAT DISCOVERY $OING THE (AMILTONIAN FACTORIZATION
30 ./6)+/6 '!&!
PROCEDURE FOR THE TOP SYSTEMS ON THE PHASE SPACES LIKE 4 b 3/ BY THE
ACTION OF 3 YOU ARE COMING TO THE SYSTEMS MATHEMATICALLY EQUIVALENT TO
THE MOTION OF THE CHARGE PARTICLE ON THE SPHERE 4HIS SPHERE IS EQUIPPED
BY SOME NONTRIVIAL 2IEMANNIAN METRIC 7HAT IS IMPORTANT AND HAS BEEN
MISSED BY THE GOOD EXPERTS IN ANALYTICAL MECHANICS LIKE +OZLOV AND +HAR
LAMOV IS THAT THE EkECTIVE MAGNETIC lELD LIKE $IRAC MONOPOLE APPEARS HERE
FOR THE NONZERO VALUES OF THE <AREA INTEGRAL ASSOCIATED WITH 3 ACTION 4HIS
MEANS PRECISELY THAT THE MAGNETIC mUX ALONG THE SPHERE IS NONZERO 4HE
REASON FOR THIS IS THAT THE SYMPLECTIC 0OISSON STRUCTURE AFTER FACTORIZA
TION IS TOPOLOGICALLY NONTRIVIAL )N TERMS OF MODERN SYMPLECTIC GEOMETRY
THE MAGNETIC lELD IS EQUIVALENT TO THE CORRECTION OF THE SYMPLECTIC STRUC
TURE 4HIS FACT IS NOT WIDELY KNOWN IN THE GEOMETRIC COMMUNITY EVEN NOW
4HE APPEARANCE OF THE TOPOLOGICALLY NONTRIVIAL SYMPLECTIC STRUCTURES AFTER
3 FACTORIZATION OF SYMPLECTIC MANIFOLDS WAS INDEPENDENTLY DISCOVERED AND
FORMULATED IN GEOMETRIC NONPHYSICAL TERMINOLOGY IN IN THE BEAUTIFUL
WORK ;= FOR DIkERENT GOALS CALCULATING OF INTEGRALS
)T HAS BEEN REALIZED IN ; = THAT THE ACTION FUNCTIONAL FOR SUCH SYSTEMS
IS IN FACT A CLOSED FORM ON THE SPACES OF LOOPS 4HESE FUNCTIONALS HAVE BEEN
IMMEDIATELY GENERALIZED FOR HIGHER DIMENSIONS TO THE SPACES OF MAPPINGS
& OF Q MANIFOLDS IN SOME TARGET SPACE - WHERE A CLOSED Q FORM IS GIVEN
INSTEAD OF MAGNETIC lELD 7E ARE COMING lNALLY TO THE ACTION FUNCTIONAL
WELL DElNED AS A CLOSED FORM ON THE MAPPING SPACES & 4HE TOPOLOGICAL
QUANTIZATION CONDITION FOR SUCH ACTIONS WAS FORMULATED IN ; = AS A
CONDITION THAT THIS CLOSED FORM SHOULD DElNE AN INTEGRAL COHOMOLOGY CLASS
IN ( & : )T IS NECESSARY AND SUnCIENT FOR THE &EINMANN AMPLITUDE TO
BE WELL DElNED AS A CIRCLE MAP
EXPFI3HG & 3
&OR THE CASE Q THE ORIGINAL $IRAC REQUIREMENT WAS BASED ON A DIF
FERENT IDEA THE MAGNETIC lELD SHOULD BE A #HERN CLASS FOR THE LINE BUNDLE
WHOSE SPACE OF SECTIONS SHOULD SERVE AS A (ILBERT SPACE OF STATES FOR OUR
1UANTUM -ECHANICS 4HEREFORE IT SHOULD BE INTEGRAL IN ( - :
)N PURE TOPOLOGY AND IN THE #ALCULUS OF 6ARIATIONS THESE IDEAS LED TO THE
CONSTRUCTION OF THE -ORSE TYPE THEORY FOR THE CLOSED FORMS ON THE lNITE
AND INlNITE DIMENSIONAL MANIFOLDS ,ET ME REFER TO THE LAST PUBLICATION OF
THE PRESENT AUTHOR WITH 0 'RINEVICH IN THIS DIRECTION ;= WHERE THE SURVEY
OF RESULTS AND PROBLEMS IS DISCUSSED ) WOULD LIKE TO POINT OUT THAT FOR
THE COMPACT SYMPLECTIC MANIFOLDS THE ACTION FUNCTIONAL FOR ANY NONTRIVIAL
(AMILTONIAN SYSTEM IS MULTIVALUED 4HE COHOMOLOGICAL CLASS OF SYMPLECTIC
FORM CANNOT BE TRIVIAL HERE ) DO NOT KNOW OF SUCH CASES IN REAL PHYSICS
WHERE THE SYMPLECTIC MANIFOLD IS COMPACT (OWEVER EVEN IN THE COMMUNITY
OF SYMPLECTIC GEOMETERS NOBODY PAID ATTENTION TO SUCH PROPERTIES OF ACTION
FUNCTIONAL UNTIL THE S
!FTER THAT ) STARTED TO THINK ABOUT DIkERENT ASPECTS OF THE (AMILTONIAN
4HEORY WHERE THE CLASS OF ONE VALUED FUNCTIONS NATURALLY CAN BE EXTENDED TO
THE CLASS OF ALL CLOSED FORMS &OR EVERY SYMPLECTIC 0OISSON MANIFOLD -
WITH ( - WE MAY CONSIDER (AMILTONIAN 3YSTEMS GENERATED BY THE
CLOSED FORM D( WHERE THE FUNCTION ( IS MULTIVALUED )NSTEAD OF ENERGY
LEVELS ( CONST WE HAVE TO CONSIDER NONTRIVIAL CODIMENSION FOLIATION
D( WITH -ORSE OR -ORSE "OTT SINGULARITIES 7E ARE COMING TO THE
TOPOLOGICAL PROBLEMS OF STUDYING SUCH FOLIATIONS )T HAS BEEN POSED IN ;=
3EVERAL PARTICIPANTS OF MY SEMINAR ! :ORICH ,E 4U 4HANG , !LANIA
HAVE MADE VERY IMPORTANT CONTRIBUTIONS TO THE STUDY OF THIS SUBJECT )N
TERESTING QUASIPERIODIC STRUCTURE APPEARS HERE )T IS NOT REVEALED FULLY IN
MY OPINION SEE REFERENCES AND DISCUSSIONS IN THE ARTICLE ;=
-ULTIVALUED (AMILTONIANS IN REAL PHYSICS ) STARTED TO LOOK AROUND
IN ASKING THE FOLLOWING QUESTION CAN YOU lND SUCH SYSTEMS IN REAL
PHYSICS WHERE (AMILTONIAN OR SOME OTHER IMPORTANT INTEGRAL OF MOTION IS
MULTIVALUED IE D( IS WELL DElNED AS A CLOSED FORM -UCH LATER PEOPLE
REALIZED THAT IN THE THEORY OF THE SO CALLED ,ANDAU ,IFSHITZ EQUATION WHICH
IS A WELL KNOWN PHYSICAL INTEGRABLE SYSTEM WITH ZERO CURVATURE REPRESEN
TATION ELLIPTIC IN THE SPECTRAL PARAMETER THE MOMENTUM IS A MULTIVALUED
FUNCTIONAL
!T THAT TIME ) FOUND ONLY ONE SUCH SYSTEM DESCRIBING MOTION OF
THE QUANTUM <"LOCH ELECTRON IN THE SINGLE CRYSTAL $ DIMENSIONAL NORMAL
METAL $ UNDER THE INmUENCE OF THE HOMOGENEOUS MAGNETIC lELD "
7E ARE WORKING HERE WITH ONE PARTICLE APPROXIMATION FOR THE SYSTEM OF
&ERMI PARTICLES WHOSE TEMPERATURE IS LOW ENOUGH &OR THE ZERO TEMPERATURE
OUR ELECTRONS lLL IN ALL ONE PARTICLE QUANTUM "LOCH STATES P BELOW THE SO
CALLED <&ERMI ,EVEL n s n& )TS VALUE DEPENDS ON THE NUMBER OF ELECTRONS
IN THE SYSTEM )T IS THE INTRINSIC CHARACTERISTIC OF OUR METAL 4HE INDEX P
HERE MAY BE CONSIDERED lNALLY AS A POINT IN THE TORUS 4 $ DElNED BY THE
RECIPROCAL LATTICE DUAL TO THE CRYSTALLOGRAPHIC ONE
P 4$ 4 $ 2 `b
4HERE IS A -ORSE FUNCTION nP 4 $ 2 DISPERSION RELATION SUCH

THAT THE DOMAIN n s n& IN THE TORUS 4 $ IS lLLED IN BY "LOCH ELECTRONS
30 ./6)+/6 '!&!
)TS BOUNDARY n n& IS A CLOSED SURFACE -& y 4 $ FOR $ 7E CALL IT

A &ERMI 3URFACE )T IS HOMOLOGOUS TO ZERO IN THE GROUP ( 4 : &OR
lNITE BUT VERY SMALL TEMPERATURE ALL ESSENTIAL EVENTS ARE HAPPENING NEARBY
THE &ERMI 3URFACE
!DD NOW A HOMOGENEOUS MAGNETIC lELD TO OUR SYSTEM IE PUT METAL
IN THE MAGNETIC lELD " .OBODY SUCCEEDED IN CONSTRUCTING A SUITABLE
WELL FOUNDED THEORY FOR THE EXACT DESCRIPTION OF ELECTRONS IN THE MAGNETIC
lELD AND LATTICE )RRATIONAL PHENOMENA APPEAR IN THE SPECTRAL THEORY OF
3CHRODINGER OPERATORS AND DESTROY THE WHOLE GEOMETRIC PICTURE (OWEVER
SINCE THE LATE S PHYSICISTS HAVE USED SOME SORT OF ADIABATIC APPROXIMATION
WHICH THEY CALL <SEMICLASSICAL ,ET ME WARN YOU THAT THIS APPROXIMATION
HAS NOTHING TO DO WITH THE STANDARD UNDERSTANDING OF SEMICLASSICAL AP
PROXIMATION 7E TAKE DISPERSION RELATION nP AS A FUNCTION ON THE TORUS 4
EXTRACTED FROM THE EXACT SOLUTION OF THE ONE PARTICLE 3CHRODINGER OPERATOR
IN THE LATTICE WITHOUT MAGNETIC lELD 7E CONSIDER A PHASE SPACE 4 a 2
WITH COORDINATES PI XJ I J AND 0OISSON BRACKET OF THE FORM
FXJ XK G FXJ PK G mKJ FPJ PK G "JK
WHERE "JK X ARE COMPONENTS OF THE MAGNETIC lELD " TREATED AS A FORM
/UR SPACE 2 IS %UCLIDEAN SO WE CAN TREAT THE MAGNETIC lELD AS A VECTOR
" WITH COMPONENTS " J
4AKE NOW THE FUNCTION nP AS A (AMILTONIAN )T GENERATES THROUGH THE
0OISSON STRUCTURE ABOVE A (AMILTONIAN SYSTEM IN THE PHASE SPACE 4 a 2
&OR THE HOMOGENEOUS IE CONSTANT MAGNETIC lELD WE CAN SEE THAT OUR
PHASE SPACE PROJECTS ON THE TORUS 4 WITH 0OISSON BRACKET FPJ PK G "JK
4HIS 0OISSON BRACKET HAS A #ASIMIR !NNIHILATOR #" P nIJK PI "JK " I PI
0 I#ASIMIR IS MULTIVALUED IT IS DElNED BY THE CLOSED FORM "
4HIS
I " DPI ON THE TORUS !S YOU WILL SEE THIS IS THE MAIN REASON FOR THE
APPEARANCE OF NONTRIVIAL TOPOLOGICAL PHENOMENA IN THIS PROBLEM
/UR (AMILTONIAN ( nP DEPENDS ON THE VARIABLE P ONLY 4HEREFORE
ALL IMPORTANT INFORMATION CAN BE EXTRACTED FROM THE (AMILTONIAN SYSTEM
ON THE TORUS WITH 0OISSON BRACKET DElNED BY THE MAGNETIC lELD 4HE
ELECTRON TRAJECTORIES FOR THE LOW TEMPERATURE CAN BE DESCRIBED AS A CURVES
IN THIS TORUS SUCH THAT
nP n& #" P CONST
(OWEVER THE LEVELS OF THE #ASIMIR ON A &ERMI SURFACE ARE IN FACT LEAVES
OF FOLIATION GIVEN BY THE CLOSED FORM RESTRICTED ON THE &ERMI SURFACE

8
" J-& " I DPI J-&
I
3OME PEOPLE IN ERGODIC THEORY STUDIED IN FACT THE MOST GENERIC ERGODIC
PROPERTIES OF <FOLIATIONS WITH TRANSVERSAL MEASURE ON THE 2IEMANN SURFACES
)N A SENSE OUR SITUATION IS A PARTIAL CASE OF THAT (OWEVER OUR PICTURE IN
TORUS IS NONGENERIC IN THAT SENSE 7E CANNOT APPLY ANY RESULTS OF THAT
THEORY 7E HAVE TO WORK WITH FOLIATIONS OBTAINED IN THE TORUS BY THIS
SPECIAL PROCEDURE ONLY /UR USE OF WORD <GENERIC HERE IS RESTRICTED BY THAT
REQUIREMENT !S WE SHALL SEE ERGODICITY IS A NONGENERIC PROPERTY WITHIN
THIS PHYSICALLY REALIZABLE SUBCLASS OF FOLIATIONS SURFACES GIVEN BY THE CLOSED
FORM 7HAT IS INTERESTING IS THAT ERGODIC EXAMPLES EXIST IN OUR PICTURE
BUT THEY OCCUPY A MEASURE ZERO SUBSET ON THE SPHERE OF DIRECTIONS OF THE
MAGNETIC lELDS IF GENERIC &ERMI SURFACE IS lXED
!S ) REALIZED IN SEE ;= THIS PICTURE LEADS TO NONTRIVIAL DIMEN
SIONAL TOPOLOGY AND ) POSED IT AS A PURELY TOPOLOGICAL PROBLEM TO MY STU
DENTS 4HE lRST BEAUTIFUL TOPOLOGICAL OBSERVATION WAS MADE BY ! :ORICH ;=
FOR THE MAGNETIC lELDS CLOSED TO THE RATIONAL ONE !FTER NEW DISCUSSION AND
RECONSIDERING ALL CONJECTURES SEE ;= ) $YNNIKOV MADE A DECISIVE BREAK
THROUGH IN THE TOPOLOGICAL UNDERSTANDING OF THIS PROBLEM FOR THE GENERIC
DIRECTIONS OF MAGNETIC lELDS SEE ;= 3 4SAREV CONSTRUCTED IN THE
lRST NONTRIVIAL ERGODIC EXAMPLES LATER IMPROVED BY $YNNIKOV SEE ;=
(OWEVER SEVERAL YEARS PASSED BEFORE SOME PHYSICAL RESULTS WERE OBTAINED
SEE THE lRST REMARK ABOUT THE POSSIBILITY OF THAT IN MY ARTICLE ;= 7E
MADE A SERIES OF JOINT WORKS WITH ! -ALTSEV SEE ; = DEDICATED TO
PHYSICAL APPLICATIONS %SSENTIALLY WE BORROWED TOPOLOGICAL RESULTS FROM THE
WORKS OF :ORICH AND $YNNIKOV (OWEVER THE NEEDS OF APPLICATIONS REQUIRED
THAT WE NOT APPLY THEIR THEOREMS DIRECTLY BUT EXTRACT THE KEY POINTS FROM
THE PROOFS AND REFORMULATE THEM 3O THE MODERN TOPOLOGICAL FORMULATIONS OF
THESE RESULTS ARE BY PRODUCTS OF THESE WORKS WITH APPLICATIONS SEE THE MOST
MODERN SURVEY IN ;= ,ET ME FORMULATE HERE OUR MAIN PHYSICAL RESULTS
AND AFTER THAT EXPLAIN THE TOPOLOGICAL BACKGROUND AND GENERALIZATIONS
4HIS PICTURE HAS BEEN EXTENSIVELY USED IN SOLID STATE PHYSICS SINCE THE
LATE S 4HE LEADING THEORETICAL SCHOOL IN THAT AREA HAS BEEN THE +HARKOV[
-OSCOW SCHOOL OF ) ,IFSHITZ AND HIS PUPILS LIKE - !ZBEL - +AGANOV
6 0ESCHANSKI ! 3LUDSKIN AND OTHERS 9OU MAY lND ALL PROPER QUOTATIONS
TO PHYSICS LITERATURE IN THE SURVEY ARTICLE ;= 4HE FOLLOWING FUNDAMEN
TAL 'EOMETRIC 3TRONG -AGNETIC &IELD ,IMIT WAS FORMULATED BY THAT
30 ./6)+/6 '!&!
SCHOOL AND FULLY ACCEPTED LATER BY THE PHYSICS COMMUNITY

!LL ESSENTIAL PHENOMENA IN THE CONDUCTIVITY OF NORMAL METALS IN A STRONG
MAGNETIC lELD SHOULD FOLLOW FROM THE GEOMETRY OF THE DYNAMICAL SYSTEM
DESCRIBED ABOVE
(OW TO UNDERSTAND THIS PRINCIPLE 9OU HAVE TO TAKE INTO ACCOUNT THAT
THIS PICTURE CERTAINLY WILL BE DESTROYED BY THE <VERY STRONG MAGNETIC lELD
WHERE QUANTUM PHENOMENA OF THE MAGNETIC ORIGIN ARE IMPORTANT )T
SHOULD HAPPEN FOR SUCH MAGNETIC lELDS THAT MAGNETIC mUX THROUGH THE
ELEMENTARY LATTICE CELL IS COMPARABLE WITH QUANTUM UNIT (OWEVER THE
LATTICE CELL IN SOLID STATE PHYSICS IS SO SMALL THAT YOU NEED FOR THAT MAGNETIC
lELD THE ORDER OF MAGNITUDE " w 'AUSS OR " w T WHERE T IS EQUAL
TO 'AUSS
4HEREFORE WE ARE COMING TO THE CONCLUSION THAT EVEN FOR THE <REAL STRONG
MAGNETIC lELDS LIKE T THIS PICTURE STILL WORKS WELL
&OR OUR GOAL WE NEED TO CONSIDER SUCH METALS THAT A &ERMI SURFACE IS
TOPOLOGICALLY NONTRIVIAL )T MEANS PRECISELY THAT THE IMBEDDING HOMOMOR
PHISM OF FUNDAMENTAL GROUPS
x -& x 4 :
IS ONTO !S PEOPLE HAVE KNOWN ALREADY FOR MANY YEARS THE NOBLE METALS
LIKE COPPER GOLD PLATINUM AND OTHERS SATISFY THIS REQUIREMENT 0ROBABLY
THE VERY lRST TIME THIS PROPERTY WAS FOUND WAS BY 0IPPARD IN FOR
COPPER -ANY OTHER MATERIALS WITH REALLY COMPLICATED &ERMI SURFACES ARE
KNOWN NOW
"Y DElNITION THE ELECTRON ORBIT IS COMPACT IF IT IS PERIODIC AND HOMO
TOPIC TO ZERO IN 4 4HEREFORE IT REMAINS COMPACT ON THE COVERING SURFACE
IN 2 WHERE 2 IS A UNIVERSAL COVERING SPACE OVER THE TORUS 4 !LL OTHER
TYPES OF TRAJECTORIES WILL BE CALLED NONCOMPACT
.ORMALLY ALL PICTURES IN PHYSICS LITERATURE ARE DRAWN IN 2 BUT EVERY
BODY KNOWS THAT QUASIMOMENTUM VECTORS P P 2 SUCH THAT P _ P
BELONGS TO THE RECIPROCAL LATTICE `b ARE PHYSICALLY IDENTICAL
4HE ,IFSHITZ GROUP STARTED TO STUDY THIS DYNAMICAL SYSTEM ABOUT
AND MADE THE lRST IMPORTANT PROGRESS &OR EXAMPLE ,IFSHITZ AND 0ESCHAN
SKI FOUND SOME NONTRIVIAL EXAMPLES OF NONCOMPACT ORBITS STABLE UNDER THE
VARIATION OF THE DIRECTION OF MAGNETIC lELD )T LOOKS LIKE NOBODY COULD UN
DERSTAND THEM PROPERLY IN THE PHYSICS COMMUNITY AT THAT TIME )T WAS
SEVERAL DECADES BEFORE THIS COMMUNITY STARTED TO UNDERSTAND THE GEOMETRY
OF DYNAMICAL SYSTEMS 4HE ,IFSHITZ GROUP WAS AHEAD OF ITS TIME 4HEY
MADE SOME MISTAKES LEADING TO WRONG CONCLUSIONS AND INVESTIGATIONS WERE
STOPPED 9OU MAY lND THE DETAILED DISCUSSION IN OUR SURVEY ARTICLE ;=
4HEIR MISTAKES HAVE BEEN FOUND ONLY NOW BECAUSE THEY CONTRADICTED OUR
lNAL RESULTS DESCRIBING THE CONDUCTIVITY TENSOR
/UR MAIN RESULTS #ONSIDER PROJECTION OF THE CONDUCTIVITY TENSOR ON THE
DIRECTION ORTHOGONAL TO MAGNETIC lELD 4HIS IS A a TENSOR z" !PPLYING
ANY WEAK ELECTRIC lELD % ORTHOGONAL TO " WE GET CURRENT J )TS PROJECTION
z" % ORTHOGONAL TO " IS ONLY WHAT IS INTERESTING FOR US NOW 7E CLAIM
THAT FOR THE STRONG MAGNETIC lELD J"J OF THE GENERIC DIRECTION IN 3
ONLY TWO TYPES OF ASYMPTOTICS ARE POSSIBLE
4OPOLOGICALLY 4RIVIAL 4YPE
z" J"J
-ORE EXACTLY WE HAVE z" /J"J_ FOR THE TOPOLOGICALLY TRIVIAL TYPE !LL
DIRECTIONS WITH TRIVIAL TYPE OCCUPY A SET 5 OF MEASURE EQUAL TO u ON THE
TWO SPHERE 5 y 3
4OPOLOGICALLY .ONTRIVIAL 4YPE

z" z" /J"J_
(ERE a TENSOR z" IS A NONTRIVIAL LIMIT FOR THE CONDUCTIVITY TENSOR 7E
CLAIM THAT IT HAS ONLY ONE NONZERO EIGENVALUE ON THE PLANE ORTHOGONAL TO "
,ET US DESCRIBE THE TOPOLOGICAL PROPERTIES OF THIS LIMITING CONDUCTIVITY TEN
SOR )T HAS EXACTLY ONE EIGEN DIRECTION p p" WITH EIGENVALUE EQUAL TO
ZERO #ONSIDER ANY SMALL VARIATION " OF THE MAGNETIC lELD " &OR THE NEW
lELD " WE HAVE AN ANALOGOUS PICTURE IF PERTURBATION IS SMALL ENOUGH 7E
HAVE A NEW a TENSOR z" WITH ONE ZERO EIGEN DIRECTION p" p /UR
STATEMENT IS THAT THE PLANE Ap Bp A B 2 GENERATED BY THIS PAIR OF DI
RECTIONS IS LOCALLY STABLE UNDER THE VARIATIONS OF MAGNETIC lELD 4HIS PLANE
IS INTEGRAL IE GENERATED BY TWO RECIPROCAL LATTICE VECTORS )T CONTAINS ZERO
EIGEN DIRECTIONS p" FOR ALL SMALL VARIATIONS OF THE MAGNETIC lELD " )T CAN
BE CHARACTERIZED BY RELATIVELY PRIME INTEGER NUMBERS M M M M
4HIS TRIPLE OF INTEGER NUMBERS IS A MEASURABLE TOPOLOGICAL INVARIANT OF THE
CONDUCTIVITY TENSOR !N OPEN SET OF DIRECTIONS 5M y 3 WITH MEASURE uM
CORRESPONDS TO THIS TYPE 4HE TOTAL MEASURE OF ALL THESE TYPES IS FULL
8
u uM x
M:
7E STARTED TO LOOK IN THE OLD EXPERIMENTAL DATA OBTAINED IN THE +APITZA

)NSTITUTE IN THE S BY 'AIDUKOV AND OTHERS SEE REFERENCES IN ;= 4HEY
30 ./6)+/6 '!&!
MEASURED RESISTANCE FOR THE SINGLE CRYSTAL GOLD SAMPLES IN THE MAGNETIC lELD
ABOUT T T FOLLOWING THE SUGGESTION OF ,IFSHITZ #ONlRMING THE IDEAS OF THE
,IFSHITZ GROUP SEVERAL DOMAINS WITH NONISOTROPIC BEHAVIOR OF CONDUCTIVITY
WERE FOUND AND MANY SUSPICIOUS <BLACK DOTS MAYBE DOMAINS OF SMALL
SIZE ON THE SPHERE 3 )T IS NOT HARD TO SEE EVEN NOW THAT SEVERAL LARGER
DOMAINS IN THESE DATA WITH NONISOTROPIC CONDUCTIVITY SHOULD CORRESPOND TO
THE SIMPLEST STABLE TOPOLOGICAL TYPES LIKE e e e e e e
UP TO PERMUTATION IN THE NATURAL BASIS OF THIS CUBIC LATTICE (OWEVER
FOR GOOD CHECKING IT WOULD BE NICE TO INCREASE MAGNETIC lELD TO T T
FOR A MORE DECISIVE CONCLUSION 4HE BLACK DOTS EITHER CORRESPOND TO THE
SMALLER DOMAINS WITH LARGER VALUES OF THE TOPOLOGICAL INTEGERS OR TO SOME
ERGODIC REGIMES OCCUPYING MEASURE ZERO SET ON THE SPHERE &OR THE lNAL
DECISION THESE EXPERIMENTS SHOULD BE REPEATED AND INCREASED ABOUT TIMES
MAGNETIC lELD AND SMALLER TEMPERATURE LIKE _ +
,ET ME EXPLAIN NOW THE TOPOLOGICAL BACKGROUND OF THESE RESULTS #ON
SIDER THE GENERIC -ORSE FUNCTION n 4 2 AND ITS GENERIC NONSINGULAR
LEVEL -& y 4 n n& IN THE TORUS AND IN THE COVERING SPACE - y 2 7E
CALL THE SURFACE - y 2 A PERIODIC SURFACE !PPLY NOW GENERIC MAGNETIC
lELD " AND MAKE THE FOLLOWING CONSTRUCTION
2EMOVE ALL NONSINGULAR COMPACT TRAJECTORIES .#4 FROM THE PERIODIC
SURFACE - AND ITS IMAGE -& IN THE TORUS 4HE REMAINING PART IS EXACTLY
SOME SURFACE WITH BOUNDARY IF IT IS NONEMPTY
;
-& N. #4 -I
IE &ERMI SURFACE MINUS ALL .#4 IS EQUAL TO THE UNION OF SURFACES WITH
BOUNDARY 7E CALL THESE SURFACES -I AND THEIR CLOSURE BELOW THE #ARRI
ERS OF /PEN 4RAJECTORIES !LL BOUNDARY CURVES ARE THE SEPARATRIX TYPE
TRAJECTORIES HOMOTOPIC TO ZERO IN 4 4HEY BOUND DISCS IN THE CORRESPOND
ING PLANES ORTHOGONAL TO MAGNETIC lELD " ,ET US lLL THEM BY THESE DISCS IN
THE PLANES 7E GET CLOSED PIECEWISE SMOOTH SURFACES - v I 7E DENOTE THEIR

HOMOLOGICAL CLASSES BY ZI ( 4 :
7E USE THE FOLLOWING EXTRACT FROM THE PROOFS OF THE MAIN THEOREMS OF
:ORICH AND $YNNIKOV SEE ; = THEIR THEOREMS HAVE NOT BEEN FORMULATED
IN THAT WAY BUT YOU MAY EXTRACT THESE KEY POINTS FROM THE PROOFS
)N THE GENERIC CASE ALL THESE HOMOLOGY CLASSES ARE NONTRIVIAL
AND EQUAL TO EACH OTHER UP TO SIGN ZI eZ ( 4 : WHERE
Z IS SOME INDIVISIBLE CLASS IN THIS GROUP !LL THESE CLOSED SURFACES
HAVE A GENUS EQUAL TO
!S YOU MAY SEE THIS STATEMENT MEANS IN FACT SOME KIND OF THE <4OPO
LOGICAL #OMPLETE )NTEGRABILITY OF OUR SYSTEMS ON THE &ERMI SURFACES FOR THE
GENERIC MAGNETIC lELD
&OR OBTAINING OUR lNAL RESULT ON THE CONDUCTIVITY TENSOR WE NEED TO USE
THE +INETIC %QUATION FOR THE QUASI PARTICLES BASED ON "LOCH WAVES NEARBY
THE &ERMI LEVEL 4HIS EQUATION HAS BEEN USED A LOT BY SOLID STATE PHYSICISTS
FOR THE PAST YEARS &OR THE SMALL BUT NONZERO TEMPERATURE STRONG
MAGNETIC lELD AND APPROPRIATE GENERAL ASSUMPTIONS ON THE IMPURITIES THE
MOTION OF QUASI PARTICLES CONCENTRATES ALONG THE ELECTRON TRAJECTORIES ABOVE
4HIS FACT LEADS TO OUR CONCLUSIONS $ESPITE THE FACT THAT THIS THEORY IS
CONSIDERED A WELL ESTABLISHED ONE ALREADY FOR MANY YEARS IN THE PHYSICS
COMMUNITY ANY ATTEMPT TO PROVE SUCH THINGS AS THE RIGOROUS MATHEMATICAL
THEOREMS WOULD BE A HUGE MESS !S WE SEE OUR lNAL CONCLUSION IS SEPARATED
FROM ALL THEOREMS BY SOME GAP WHICH CANNOT BE ELIMINATED ,ET ME POINT
OUT THAT IT IS ALWAYS SO <2IGOROUS PROOFS IN MATHEMATICAL PHYSICS NEVER
PROVE ANYTHING IN REAL WORLD PHYSICS
7HAT ABOUT NONGENERIC TRAJECTORIES 4SAREV AND $YNNIKOV CONSTRUCTED
VERY INTERESTING EXAMPLES WHERE GENUS OF CARRIERS OF THE OPEN TRAJECTORIES IS
LARGER THAN SEE;= 7E CALL SUCH CASES STOCHASTIC 3OMETIMES WE CALL
THEM ERGODIC 4HERE WERE SOME ATTEMPTS TO EXTRACT FROM THEIR PROPERTIES
HIGHLY NONTRIVIAL ASYMPTOTICS OF THE CONDUCTIVITY TENSOR IN THE STRONG MAG
NETIC lELD ;= (OWEVER THESE ATTEMPTS NEED A BETTER UNDERSTANDING OF THE
PROPERTIES OF SUCH TRAJECTORIES 7E HAVE TO ANSWER THE FOLLOWING QUESTIONS
(OW MANY DIRECTIONS OF THE MAGNETIC lELD ON THE SPHERE 3
ADMIT ERGODIC TRAJECTORIES
!CCORDING TO MY CONJECTURE FOR THE GENERIC &ERMI SURFACE THIS SET OF
DIRECTIONS HAS A (AUSDORF DIMENSION NOT GREATER THAN SOME NUMBER A
ON THE SPHERE 3 &OR THE SPECIAL &ERMI SURFACES n OF THE EVEN FUNCTIONS
LIKE COS P COS P COS P WE EXPECT TO HAVE ERGODIC TRAJECTORIES FOR
THE SET OF DIRECTIONS WITH (AUSDORF DIMENSION LIKE A $YNNIKOV
STARTED TO INVESTIGATE THIS EXAMPLE IN HIS 4HESIS AND PROVED SEVERAL GENERAL
PROPERTIES 2ECENTLY 2 $ELEO INVESTIGATED SUCH EXAMPLES MORE CAREFULLY
AND PERFORMED MORE DETAILED CALCULATIONS ;= (IS RESULTS CONlRM OUR
CONJECTURES (OWEVER THE (AUSDORF DIMENSION OF THIS SET HAS BEEN UNKNOWN
IN THIS EXAMPLE UNTIL NOW
7HICH GEOMETRIC PROPERTIES DOES A <TYPICAL ERGODIC TRA
JECTORY HAVE
!CCORDING TO THE CONJECTURE OF -ALTSEV THESE TRAJECTORIES ARE TYPICALLY
30 ./6)+/6 '!&!
THE <ASYMPTOTICALLY SELF SIMILAR PLANE CURVES IN THE NATURAL SENSE (IS IDEA
IF IT IS TRUE LEADS TO THE INTERESTING UNUSUAL PROPERTIES OF THE ASYMPTOTIC
CONDUCTIVITY TENSOR !NYWAY THIS PROBLEM IS VERY INTERESTING
$YNNIKOV INVESTIGATED ALSO THE DEPENDENCE OF THESE INVARIANTS ON THE
LEVEL n& OF THE DISPERSION RELATION SEE ;= 4HESE RESULTS ARE USEFUL FOR THE
RIGHT UNDERSTANDING OF OUR CONJECTURES
-ULTIDIMENSIONAL GENERALIZATIONS #ONSIDER THE FOLLOWING PROBLEM
7HAT CAN BE SAID ABOUT TOPOLOGY OF THE LEVELS F X Y CONST OF THE
QUASIPERIODIC FUNCTIONS WITH M PERIODS ON THE PLANE X Y
&OR THE CASE M THIS PROBLEM EXACTLY COINCIDES WITH OUR SUBJECT
ABOVE "Y DElNITION A QUASIPERIODIC FUNCTION ON THE PLANE IS A RESTRICTION
ON THE PLANE 2 y 2M OF THE M PERIODIC FUNCTION /UR SPACE 2 WAS A
SPACE OF QUASI MOMENTA MORE PRECISELY ITS UNIVERSAL COVERING /UR PLANE
WAS ORTHOGONAL TO THE MAGNETIC lELD #AN THIS THEORY BE GENERALIZED TO
THE CASE M !CCORDING TO MY CONJECTURE IT CAN BE GENERALIZED TO THE
CASE M ) THINK THAT FOR SMALL PERTURBATIONS OF THE RATIONAL DIRECTIONS
THIS THEORY CAN BE GENERALIZED TO ANY VALUE OF M 7E CONSIDER NOW ANY
PERIODIC FUNCTION F 2 4 2 AND PAIR OF THE RATIONAL DIRECTIONS
L L CORRESPONDING TO SOME LATTICE : IN 2
,ET ME FORMULATE THE FOLLOWING THEOREM
4HEOREM 4HERE EXIST TWO NONEMPTY OPEN SETS 5 5 ON THE SPHERE 3
CONTAINING THE RATIONAL DIRECTIONS L L CORRESPONDINGLY SUCH THAT
&OR EVERY PLANE 2L y 2 FROM THE FAMILY GIVEN BY EQUATIONS L
CONST L CONST THE QUASIPERIODIC FUNCTIONS FL HAVE ONLY THE FOLLOWING TWO
TYPES OF CONNECTIVITY COMPONENTS OF THE LEVELS FL CONST ON THE PLANE 2L
4HE CONNECTIVITY COMPONENT OF THE LEVEL IS A COMPACT CLOSED CURVE ON
THE PLANE
4HE CONNECTIVITY COMPONENT OF THE LEVEL IS AN OPEN CURVE LYING IN THE
STRIP OF lNITE WIDTH BETWEEN PARALLEL STRAIGHT LINES WITH THE COMMON
DIRECTION p 4HIS SITUATION IS STABLE IN THE FOLLOWING SENSE !FTER ANY
SMALL VARIATIONS OF THE DIRECTIONS L 5 L 5 OF FUNCTION F ON
4 OR THE LEVEL WE STILL HAVE SUCH OPEN COMPONENT WITH DIRECTION p
&OR ALL POSSIBLE PERTURBATIONS THIS SET OF DIRECTIONS p p BELONG TO
SOME INTEGRAL HYPERPLANE IN 2
4HIS PROPERTY CAN BE FORMULATED IN TERMS OF THE INTEGRAL HOMOLOGY CLASS
IN THE GROUP ( 4 : AND OF THE TORICAL TOPOLOGY OF THE CARRIERS OF THE
OPEN TRAJECTORIES 4HE IDEA OF THE PROOF WAS RECENTLY PUBLISHED BY THE
AUTHOR IN ;=
7E MAY REFORMULATE THIS PROBLEM IN TERMS OF (AMILTONIAN SYSTEMS ,ET

THE CONSTANT 0OISSON "RACKET "IJ BE GIVEN ON THE TORUS 4 M WHOSE RANK IS
EQUAL TO !NY (AMILTONIAN F GENERATES SUCH SYSTEMS WHOSE TRAJECTORIES
ARE EQUAL TO THE LEVELS OF F ON THE PLANES /UR THEOREM MEANS THAT IN
THESE CASES THIS (AMILTONIAN SYSTEM IS #OMPLETELY )NTEGRABLE IN THE SPECIlC
TOPOLOGICAL SENSE DESCRIBED ABOVE
2EFERENCES
;= 3 .OVIKOV 4OPOLOGY ) %NCYCLOPEDIA OF -ATHEMATICAL 3CIENCES VOL
3PRINGER 6ERLAG "ERLIN (EIDELBERG .EW 9ORK
;= 3 .OVIKOV "LOCH FUNCTIONS IN A MAGNETIC lELD AND VECTOR BUNDLES 4YPICAL
DISPERSION RELATIONS AND THEIR QUANTUM NUMBERS $OKLADY !. 3332
[
;= 3 .OVIKOV ) 3CHMELZER 0ERIODIC SOLUTION OF THE +IRCHHOk EQUATIONS
&UNCTIONAL !NALYSIS !PPL [
;= 3 .OVIKOV -ULTIVALUED FUNCTIONS AND FUNCTIONALS $OKLADY !. 3332
[
;= 3 .OVIKOV 4HE (AMILTONIAN FORMALISM AND MULTIVALUED ANALOG OF THE
-ORSE THEORY 2USSIAN -ATH 3URVEYS [
;= * $UISTERMAAT ' (ECKMANN /N THE VARIATION IN THE COHOMOLOGY OF
THE SYMPLECTIC FORM OF THE REDUCED PHASE SPACE )NVENTIONES -ATH
[
;= 0 'RINEVICH 3 .OVIKOV .ONSELlNTERSECTING MAGNETIC ORBITS ON THE
PLANE 0ROOF OF THE OVERTHROWING OF THE CYCLES PRINCIPLE !MERICAN -ATH 3O
CIETY 4RANSLATIONS 3ER 4OPICS IN 4OPOLOGY AND -ATH 0HYSICS
EDITED BY 3 .OVIKOV [
;= ! :ORICH .OVIKOVS PROBLEM ON THE SEMICLASSICAL ELECTRON IN THE HOMOGE
NEOUS MAGNETIC lELD 2USSIAN -ATH 3URVEYS [
;= 3 .OVIKOV 1UASIPERIODIC STRUCTURES IN TOPOLOGY 0ROCEEDINGS OF THE #ONFER
ENCE <4OPOLOGICAL -ETHODS IN -ODERN -ATHEMATICS 3TONYBROOK*UNE
3TONYBROOK 5NIVERSITY
;= ) $YNNIKOV 0ROOF OF .OVIKOV CONJECTURE ON THE SEMICLASSICAL ELECTRON
-ATH :AMETKI [
;= 3 .OVIKOV 4HE SEMICLASSICAL ELECTRON IN A MAGNETIC lELD AND LATTICE 3OME
PROBLEMS OF LOW DIMENSIONAL <PERIODIC TOPOLOGY 'EOMETRIC !ND &UNCTIONAL
!NALYSIS [
;= 3 .OVIKOV ! -ALTSEV 4OPOLOGICAL QUANTUM CHARACTERISTICS OBSERVED OF
THE CONDUCTIVITY IN NORMAL METALS ,ETTERS OF *%40 [
4RANSLATED INTO %NGLISH BY THE !MERICAN )NSTITUTE OF 0HYSICS
;= 3 .OVIKOV ! -ALTSEV 4OPOLOGICAL PHENOMENA IN THE NORMAL METALS
5SPEKHI 0HYS .AUK [ TRANSLATED IN %NGLISH BY THE
30 ./6)+/6 '!&!
!MERICAN )NSTITUTE OF 0HYSICS

;= ) $YNNIKOV 'EOMETRY OF STABILITY ZONES IN THE .OVIKOV PROBLEM ON THE
SEMICLASSICAL MOTION OF AN ELECTRON 2USSIAN -ATH 3URVEYS [

;= ) $YNNIKOV ! -ALTSEV *%40 *OURN %XPERIMENTAL AND 4HEORETICAL
0HYSICS [
;= 2 $ELEO %XISTENCE AND MEASURE OF ERGODIC LEAVES IN .OVIKOVS PROBLEM ON
THE SEMICLASSICAL MOTION OF AN ELECTRON 2USSIAN -ATH 3URVEYS
;= 3 .OVIKOV 4HE LEVELS OF 1UASIPERIODIC FUNCTIONS ON THE PLANE AND TOPOLOGY
2USSIAN -ATH 3URVEYS
3ERGEY 0 .OVIKOV -ATH $EPARTMENT AND )034 5NIVERSITY OF -ARYLAND #OLLEGE

0ARK -$ 53!
AND
,ANDAU )NSTITUTE FOR 4HEORETICAL 0HYSICS -OSCOW +OSYGINA
3/-% 02/",%-3 ). 4(% 4(%/29 /& $9.!-)#!,

3934%-3 !.$ -!4(%-!4)#!, 0(93)#3
9A' 3INAI
4HE PROBLEMS CONSIDERED BELOW ARE NOT NEW 4HEY HAVE BEEN DISCUSSED
BEFORE IN MANY PUBLICATIONS AND IMPORTANT RESULTS HAVE BEEN OBTAINED
(OWEVER IT SEEMS THAT AN ADEQUATE UNDERSTANDING IS STILL LACKING AND SE
RIOUS PROGRESS IS EXPECTED WHICH COULD BE EVEN MORE IMPORTANT THAN THE
INITIAL GOALS
! BRIEF LOOK AT THE HISTORY OF THE THEORY OF DYNAMICAL SYSTEMS OVER THE
LAST lFTY YEARS SHOWS THAT IT DEVELOPED IN BURSTS AROUND THE GREAT DISCOVER
IES WHEN NEW IDEAS METHODS AND PHENOMENA APPEARED ,ET ME MENTION
THESE DISCOVERIES EXPLICITLY IN MORE OR LESS CHRONOLOGICAL ORDER
+!- THEORY
4HE WORKS OF 3MALE !NOSOV AND OTHERS ON THE STRUCTURAL STABILITY
AND HYPERBOLICITY OF DYNAMICAL SYSTEMS STRANGE ATTRACTORS AND THE
,ORENZ MODEL AND HYPERBOLIC BILLIARDS
%NTROPY THEORY WHICH STARTED WITH THE CLASSICAL WORK BY +OLMOGOROV
AND ENDED WITH THE FAMOUS WORKS OF /RNSTEIN AND HIS COAUTHORS
&EIGENBAUM UNIVERSALITY IN PERIOD DOUBLING BIFURCATIONS
2EMARKABLE RESULTS WERE OBTAINED IN OTHER DIRECTIONS SUCH AS NON LOCAL
BIFURCATIONS HOLOMORPHIC DYNAMICS mOWS ON SURFACES ERGODIC THEORY ON
HOMOGENEOUS SPACES BILLIARDS BUT IT SEEMS THAT [ WERE DOMINATING EVENTS
&IFTY YEARS AGO IT WAS ABSOLUTELY IMPOSSIBLE TO ANTICIPATE ANY OF THESE
FOUR BREAKTHROUGHS 4HIS UNPREDICTABILITY IS ONE OF THE ATTRACTIVE FEATURES
OF THE THEORY OF DYNAMICAL SYSTEMS &OR THIS REASON ON THE OCCASION OF THIS
#ONFERENCE ONE CAN ONLY HOPE TO INDICATE SOME DIRECTIONS WHERE NEW AND
SERIOUS PROGRESS LOOKS POSSIBLE 4HE TEXT BELOW PRESENTS SOME SUGGESTIONS
'ENERICALLY 1UASI PERIODIC AND (YPERBOLIC $YNAMICS

ARE THE /NLY 3TRUCTURALLY 3TABLE 4YPES OF $YNAMICS
4HE PROBLEM GOES BACK TO +OLMOGOROV )N THE CASE OF (AMILTONIAN DY
NAMICS A POSITIVE ANSWER WOULD MEAN THAT THE PHASE SPACE CAN BE DECOM
POSED ONTO SUBSETS OF POSITIVE MEASURE CONSISTING OF INVARIANT TORI WITH
9!' 3).!) '!&!
QUASI PERIODIC MOTIONS AND STOCHASTIC LAYERS WHERE AT LEAST SOME ,YAPUNOV
EXPONENTS ARE NON ZERO 4HE STATEMENT OF THIS TYPE IS IMPORTANT FOR APPLI
CATIONS IN DIkERENTIAL GEOMETRY THE STRUCTURE OF GEODESIC mOWS ASTRONOMY
STATISTICAL MECHANICS AND OTHER lELDS USING (AMILTONIAN DYNAMICS
&OR DISSIPATIVE SYSTEMS SOME VERSION OF THIS PROBLEM WAS FORMULATED
RECENTLY AS THE WHOLE PROGRAM BY * 0ALIS SEE ;0=
4HERE ARE SOME RESULTS WHICH CAN BE CONSIDERED AS RELATED TO THIS PROB
LEM
&IRST OF ALL ) MEAN ONE DIMENSIONAL DYNAMICS WHERE DUE TO THE IM
PORTANT RESULTS OF *AKOBSON ;*= "ENEDICKS AND #ARLESON ;"#= ,YUBICH
;,Y= 'RACZYK AND 3WIATEK ;'R3= WE KNOW THAT IN ANALYTIC FAMILIES OF
ONE DIMENSIONAL MAPS THE SPACE OF PARAMETERS CAN BE DECOMPOSED ONTO
A SUBSET WHERE THE DYNAMICS HAS A STABLE PERIODIC ORBIT A SUBSET WHERE
THE DYNAMICS HAS AN ABSOLUTELY CONTINUOUS INVARIANT MEASURE AND IS CHAOTIC
AND A NEGLIGIBLE SUBSET OF MEASURE ZERO
+!- THEORY GIVES THE CONDITIONS UNDER WHICH THE INVARIANT TORI WITH
QUASI PERIODIC DYNAMICS PERSIST IN (AMILTONIAN SYSTEMS )N SPITE OF A RECENT
ACTIVITY AROUND THE CONSTRUCTION OF THESE TORI IT IS HARD TO EXPECT NEW DIS
COVERIES BASED ON THIS TECHNIQUE EVEN IN ITS MODERNIZED FORM /N THE OTHER
HAND MANY PROBLEMS WHICH GREW OUT OF +!- THEORY REMAIN COMPLETELY
OPEN 4HE MOST NOTABLE ONE IS THE PROBLEM OF CLASSIlCATION OF +!- ISLANDS
)T IS HIGHLY NON TRIVIAL EVEN WITHIN THE CLASS OF TWO DIMENSIONAL TWIST MAPS
IE SYMPLECTIC MAPS OF THE TWO DIMENSIONAL CYLINDER 4 Z Z
Z Z K6 Z (ERE 6 IS A SMOOTH PERIODIC FUNCTION OF
K IS A PARAMETER )T IS WIDELY BELIEVED THAT THE FAMILIES OF TWIST MAPS
ARE THE NEXT CANDIDATES FOR DEEP STUDY AFTER THE FAMILIES OF ONE DIMENSIONAL
MAPS /NE OF THE NON DIRECT CONSEQUENCES OF +!- THEORY IS THE EXISTENCE
FOR GENERIC TWIST MAPS OF +!- ISLANDS IE INVARIANT OR PERIODIC DOMAINS
BOUNDED BY INVARIANT OR PERIODIC CURVES .EAR THE ISLANDS ONE CAN lND
SIMILAR ISLANDS OF A SMALLER SCALE AND SO ON SEE &IGURE 7E CLEARLY SEE A
HIERARCHICAL STRUCTURE OF THE ISLANDS WITH SOME SCALE INVARIANCE 4HE BASIC
PROBLEMS HERE ARE THE FOLLOWING
TO ESTABLISH THE EXISTENCE OF THIS HIERARCHICAL STRUCTURE
TO DESCRIBE ROTATION NUMBERS FOR THE INDUCED MAPS ON THE BOUNDARIES
OF THE ISLANDS
TO ESTIMATE THE GLOBAL MEASURE OF +!- ISLANDS AS K
!PPARENTLY CAN BE STUDIED WITH THE HELP OF RENORMALIZATION GROUP
TECHNIQUE BECAUSE IN TYPICAL SITUATIONS ASYMPTOTICALLY THE STRUCTURE OF THE
6ISIONS IN -ATH $9.!-)#!, 3934%-3 !.$ -!4(%-!4)#!, 0(93)#3
&IGURE
CORRESPONDING POWER OF THE MAP NEAR ANY SMALL ISLAND DEPENDS ONLY VERY
WEAKLY ON THE STRUCTURE OF THE INITIAL MAP AND BELONGS TO SOME CLASS OF
UNIVERSALITY 4HESE IDEAS WERE PROPOSED IN THE WORKS OF ,ICHTENBERG ;,=
:ASLAVSKY ;:= -ELNIKOV ;-E= AND OTHERS 4HE RENORMALIZATION GROUP AP
PROACH TO THE CONSTRUCTION OF +!- TORI CAN BE FOUND IN ;!+= ;+3= IN ;!+=
ONE CAN lND OTHER RELATED REFERENCES
0ROBLEM IS CONNECTED WITH ANOTHER IMPORTANT QUESTION #ONSIDER AN
INVARIANT CURVE ` GIVEN BY THE EQUATION Z F` )F ` IS A +!- CURVE
THEN IT IS NOT ISOLATED AND IS IN A NATURAL SENSE A DENSITY POINT OF OTHER
+!- CURVES 4HEREFORE BOUNDARY CURVES WHICH SEPARATE STOCHASTIC LAYERS
FROM +!- CURVES SHOULD BE NON SMOOTH /N THE OTHER HAND THEY ARE
<LAST CURVES WITH GIVEN ROTATION NUMBERS AND THIS IS THEIR CHARACTERISTIC
FEATURE SEE &IGURE 7ITH THE CHANGE OF PARAMETER THEY BIFURCATE INTO
!UBRY -ATHER SETS 4HUS ANOTHER PROBLEM ALONG THESE LINES IS
TO CONSTRUCT OR TO PROVE THE EXISTENCE OF NON SMOOTH INVARIANT CURVES
OF TWIST MAPS
!T THE MOMENT THE ONLY HOPE TO ATTACK THIS PROBLEM IS AGAIN THE RENOR
MALIZATION GROUP THEORY #HIRIKOV ;#H= STRESSES THAT ROTATION NUMBERS OF
INDUCED MAPS ON THESE CURVES SHOULD BE SPECIAL )N THIS CONNECTION ONE CAN
FORMULATE THE NEXT PROBLEM
9!' 3).!) '!&!
&IGURE
7HY THE ROTATION NUMBER OF THE LAST INVARIANT CURVE IS A GOLDEN MEAN
)T IS EASY TO SHOW THAT FOR LARGE ENOUGH K TWIST MAPS HAVE NO INVARIANT
CURVES ` DESCRIBED BY THE EQUATIONS Z F` SEE FOR EXAMPLE ;3I=
4HEREFORE THERE EXISTS A CRITICAL VALUE KCR SUCH THAT FOR K KCR THERE ARE
NO SUCH CURVES )T WAS A NUMERICAL DISCOVERY BY * 'REENE ;'RE= THAT FOR
SOME TWIST MAPS THE ROTATION NUMBER ON THE LAST CURVE IS A GOLDEN MEAN
2 -AC+AY ;-= PROPOSED A RENORMALIZATION GROUP TECHNIQUE TO STUDY THESE
CURVES
! SATISFACTORY ANSWER TO THE LAST PROBLEM CAN EXPLAIN WHY THE GOLDEN
MEAN APPEARS SO OFTEN IN MATHEMATICS AND NOT ONLY IN MATHEMATICS AR
CHITECTURE CONSTRUCTION ART
0ROBLEM IS CONNECTED WITH THE FOLLOWING WELL KNOWN HYPOTHESIS
(YPOTHESIS &OR ANY K THE TWIST MAP HAS POSITIVE METRIC ENTROPY
(YPOTHESIS &OR GENERIC 6 THE SET OF K FOR WHICH 4 HAS NO ISLANDS AT

ALL AND IS ERGODIC HAS DENSITY
4HIS HYPOTHESIS IS MOTIVATED BY ANALOGY WITH THE *AKOBSON THEOREM
IN ONE DIMENSIONAL DYNAMICS AND "ENEDICKS[#ARLESON RESULTS FOR (ENON
MAP )T MEANS THAT FOR K IN (YPOTHESIS THE MAP 4 IS HYPERBOLIC BUT
THE HYPERBOLICITY SHOULD BE VERY WEAK )N PARTICULAR STABLE AND UNSTABLE
MANIFOLDS ARE NOT TRANSVERSAL EVERYWHERE AND CAN HAVE TANGENCY UP TO SOME
ORDER )T WOULD BE INTERESTING TO SHOW THAT TYPICALLY THE TANGENCY IS OF lNITE
ORDER (OWEVER EVEN THE PRECISE FORMULATION OF AN EXACT STATEMENT CAN BE
NON TRIVIAL /NE SHOULD MENTION THE RESULT BY $UARTE ;$= WHO SHOWED THAT
THE SET OF K FOR WHICH 4 HAS +!- ISLANDS IS OPEN AND EVERYWHERE DENSE
4HE EXTENSION OF THESE PROBLEMS TO THE MULTIDIMENSIONAL SETTING IS CER
TAINLY VERY IMPORTANT BUT APPARENTLY VERY DInCULT
6ISIONS IN -ATH $9.!-)#!, 3934%-3 !.$ -!4(%-!4)#!, 0(93)#3
1UALITATIVE 4HEORY OF 0$% AND )NlNITE $IMENSIONAL

3YSTEMS
)T SEEMS THAT THIS THEORY WILL BE DIkERENT FROM ITS lNITE DIMENSIONAL COUN
TERPART STARTED BY 0OINCARsE $ENJOY ,YAPUNOV AND OTHERS 3O FAR THE CON
CEPT OF 0$% AS A VECTOR lELD ON A TOPOLOGICAL MANIFOLD WAS NOT SO FRUITFUL
PROBABLY BECAUSE THE TOPOLOGY OF THE PHASE SPACE DID NOT PLAY AN ESSENTIAL
ROLE /N THE OTHER HAND EXCEPT GENERAL QUESTIONS OF EXISTENCE UNIQUENESS
AND SMOOTHNESS OF SOLUTIONS THE EXISTING THEORY IS MUCH MORE CONCRETE
4HERE ARE RELATIVELY FEW 0$% WHICH ARE IMPORTANT FOR APPLICATIONS THE
.AVIER[3TOKES AND OTHER HYDRODYNAMICAL SYSTEMS THE +ORTEWEG[DE 6RIES
EQUATION THE +0 EQUATION THE 'INZBURG[,ANDAU EQUATION THE +URAMOTO[
3IVASHINSKI EQUATION REACTION DIkUSION EQUATIONS THE EQUATION OF GENERAL
RELATIVITY MEAN CURVATURE mOWS AND A FEW OTHERS "ASIC QUALITATIVE AND NU
MERICAL RESULTS ARE CONCENTRATED AROUND THE CONSTRUCTION OF VARIOUS TYPES
OF SPACE PATTERNS WITH DIkERENT GEOMETRY ROLLS DISLOCATIONS AND OTHERS
4HE TRANSITION FROM ONE TYPE OF PATTERN TO ANOTHER SEEMS TO BE A NEW TYPE
OF BIFURCATION 4HE BASIC PROBLEM CAN BE FORMULATED AS FOLLOWS
TO CLASSIFY SPATIAL PATTERNS WHICH ARE DISPLAYED BY 0$% AND TO DE
SCRIBE POSSIBLE TYPES OF THEIR BIFURCATIONS 3OME RESULTS IN THIS DIREC
TION CAN BE FOUND IN THE BOOK BY 'OLUBITSKY AND 3CHAEkER ;'3=
)N PREVIOUS YEARS THERE WERE MANY EXAMPLES OF COMPUTER ASSISTED
PROOFS IN THE THEORY OF DYNAMICAL SYSTEMS 4HE MOST NOTABLE ONE WAS
THE PROOF BY ,ANFORD OF THE EXISTENCE OF A &EIGENBAUM[#OLLET[4RESSER
lXED POINT IN THE THEORY OF PERIOD DOUBLING BIFURCATIONS 2ECENTLY
4UCKER GAVE A COMPUTER ASSISTED PROOF OF THE EXISTENCE OF ,ORENZ AT
TRACTOR SEE ;4= (OWEVER IN GENERAL THIS DIRECTION DID NOT ATTRACT
MUCH ATTENTION 0EOPLE USUALLY DO NOT WANT TO PROVIDE DETAILED MATH
EMATICAL ARGUMENTS IF NUMERICAL EVIDENCE IS CONVINCING ENOUGH )T IS
QUITE POSSIBLE THAT IN THE CASE OF 0$% RIGOROUS RESULTS WILL GO TOGETHER
WITH SERIOUS NUMERICAL STUDIES )T IS ALREADY CLEAR THAT COMPLETE PROOFS
REQUIRE IN MANY CASES THE MOST POWERFUL COMPUTERS
.AVIER[3TOKES EQUATIONS DESERVE SPECIAL ATTENTION 4HE EXISTENCE
PROBLEM OF STRONG SOLUTIONS FOR A $ .AVIER[3TOKES SYSTEM STILL RE
MAINS AN OPEN OUTSTANDING PROBLEM IN SPITE OF THE EkORTS OF MANY
RESEARCHERS HAVING DIkERENT OPINIONS ABOUT THE ANSWER )F WE ADOPT
THE POINT OF VIEW THAT BLOW UPS IN lNITE TIME ARE POSSIBLE EVEN IN THE
PRESENCE OF VISCOSITY THEN WE CAN TAKE THE NEXT STEP AND ASK WHICH
mOWS WITH lNITE ENERGY AND INlNITE ENSTROPHY CAN BE LIMITS OF LOCAL
9!' 3).!) '!&!
SOLUTIONS OF A $ .AVIER[3TOKES SYSTEM 0ROBABLY THIS FORMULATION OF

THE BASIC PROBLEM IS MORE AMENABLE TO DIRECT ATTACKS SINCE IT IS LOCAL
4HE PROBLEM OF TURBULENCE WILL BE IN THE CENTER OF INTEREST OF MANY
MATHEMATICIANS PHYSICISTS AND ENGINEERS &OR MATHEMATICIANS ONE
OF THE MAIN PROBLEMS IS TO FORMALIZE NOTIONS WHICH ARE WIDELY USED
BY PHYSICISTS ENGINEERS ETC (ERE IS ONE EXAMPLE
4O PROPOSE DElNITIONS OF SOLUTIONS DISPLAYING THE mOW OF ENERGY OVER
THE SPECTRUM AND TO PROVE THE EXISTENCE OF SUCH SOLUTIONS
)T IS BELIEVED THAT ONE OF THE MATHEMATICAL MODELS OF TURBULENCE IS
.AVIER[3TOKES SYSTEM WITH RANDOM FORCING 3OME RESULTS CONCERNING
THE EXISTENCE OF INVARIANT MEASURE FOR THIS MODEL WERE OBTAINED BY
&LANDOLI AND -ASLOWSKI SEE ;&-= AND OTHERS (OWEVER BASIC QUES
TIONS REMAIN UNANSWERED 4HE MAIN PROBLEM CAN BE FORMULATED AS
FOLLOWS
4O CONSTRUCT -ARKOV RANDOM lELDS FOR .AVIER[3TOKES SYSTEMS WHERE
FORCING IS A WHITE IN TIME RANDOM lELD AND TO STUDY THE PROPERTIES OF
THEIR CORRELATION FUNCTIONS IN THE LIMIT OF ZEROTH VISCOSITY
0ROGRESS IN THIS DIRECTION CAN LEAD TO THE MATHEMATICAL JUSTIlCATION
OF +OLMOGOROVS THEORY OF TURBULENCE
1UANTUM #HAOS AND !NDERSON ,OCALIZATION

1UANTUM CHAOS AS A PART OF PHYSICS APPEARED ABOUT TWENTY YEARS AGO &OR
MATHEMATICIANS ITS MAIN PURPOSE WAS TO DESCRIBE PROPERTIES OF QUANTUM
SYSTEMS WHICH DEPEND ON THEIR CLASSICAL LIMITS 0ROBLEMS OF QUANTUM CHAOS
ARE DISCUSSED ALSO BY 0 3ARNAK AT THIS CONFERENCE 4HE lRST MATHEMATICAL
RESULT IN QUANTUM CHAOS WAS PROVEN BY 3HNIRELMAN SEE ;3= WHO SHOWED
THAT IF A GEODESIC mOW ON A SMOOTH COMPACT 2IEMANNIAN MANIFOLD IS ERGODIC
THEN THE SQUARES OF CORRESPONDING EIGENFUNCTIONS OF ,APLACIAN ARE IN SOME
SENSE UNIFORMLY DISTRIBUTED
!CTUALLY PROBLEMS OF QUANTUM CHAOS IN MATHEMATICS STARTED EVEN EAR
LIER !SSUME THAT 1 IS A COMPACT SMOOTH D DIMENSIONAL 2IEMANNIAN MAN
IFOLD AND . t =FtI s tG WHERE tI ARE EIGENVALUES OF THE ,APLACIAN
7EYLS FAMOUS ASYMPTOTICS TELLS US THAT
. t CtD Nt
WHERE C A DIMENSION DEPENDENT CONSTANT AND Nt IS A REMAINDER

)T IS WIDELY BELIEVED THAT THE BEHAVIOR OF Nt IS NON UNIVERSAL AND IS
6ISIONS IN -ATH $9.!-)#!, 3934%-3 !.$ -!4(%-!4)#!, 0(93)#3
DETERMINED BY ERGODIC PROPERTIES OF THE GEODESIC mOW ON THE UNIT TANGENT

BUNDLE OF 1 )T FOLLOWS BASICALLY FROM THE WORKS OF #OLIN DE 6ERDIERE SEE
;#= THE TWO DIMENSIONAL CASE IN ;+O-3= THAT IN THE INTEGRABLE CASE THE
ANALYSIS OF Nt CAN BE REDUCED TO NUMBER THEORETIC PROBLEMS AND THE RESULT
D_
ESSENTIALLY HAS THE FORM Nt t 1t N t WHERE 1t IS A QUASI
PERIODIC FUNCTION AND N t HAS A SMALLER ORDER OF MAGNITUDE
)N THE CASE OF MANIFOLDS OF NEGATIVE CURVATURE WHERE THE GEODESIC mOW IS

STRONGLY MIXING NUMERICAL AND QUALITATIVE RESULTS INDICATE THAT Nt GROWS
ONLY AS LOG t l FOR SOME l 4HIS RESEMBLES THE ASYMPTOTICS OF THE NUM
BER OF ZEROS OF THE 2IEMANN o FUNCTION /THER RESULTS SUGGEST THAT THE
LIMITING PROBABILITY DISTRIBUTION OF SPACINGS BETWEEN THE NEAREST EIGENVAL
UES BEHAVES SIMILARLY TO THE DISTRIBUTION OF SPACINGS IN SOME ENSEMBLES OF
RANDOM MATRICES OF GROWING DIMENSION ,ET ME MENTION IN THIS CONNECTION
RESULTS BY (EJHAL AND 2ACKNER SEE ;(2= WHO CONVINCINGLY SHOWED THAT THE
PROBABILITY DISTRIBUTIONS OF EIGENFUNCTIONS OF ,APLACIANS ON SUCH MANIFOLDS
ARE 'AUSSIAN IN THE LIMIT t !LL THESE RESULTS LEAD TO THE FOLLOWING
PROBLEM
7HY ,APLACIANS ON COMPACT MANIFOLDS OF NEGATIVE CURVATURE BEHAVE

AS ELEMENTS OF 7IGNER ENSEMBLES OF RANDOM MATRICES
3O FAR ASYMPTOTIC EXPRESSIONS FOR EIGENFUNCTIONS OF ,APLACIANS ARE
KNOWN FOR THE ABOVE MENTIONED CASES OF 1 WITH INTEGRABLE GEODESIC
mOWS 3CHNIRELMANS THEOREM SHOWS THAT IN THE CASE OF MIXING mOWS
THEY ARE MUCH MORE RANDOM
2ANDOM EIGENFUNCTIONS APPEAR IN A QUITE DIkERENT PART OF MATHEMAT
ICAL PHYSICS THE SO CALLED !NDERSON LOCALIZATION !, !, IS A PROP
ERTY OF 3CHROEDINGER OPERATORS WITH RANDOM POTENTIALS AND MEANS
THAT WITH PROBABILITY ALL EIGENFUNCTIONS DECAY EXPONENTIALLY AT IN
lNITY !, IS BASED ON QUANTUM RESONANCES IE OF THE APPEARANCE OR
DISAPPEARANCE OF APPROXIMATE EIGENFUNCTIONS WITH CLOSE EIGENVALUES
/NE OF THE MAIN OPEN PROBLEMS HERE IS
TO CONSTRUCT FOR D t A NON LOCALIZED EIGENFUNCTION BELONGING TO THE

CONTINUOUS SPECTRUM
)T WOULD BE INTERESTING TO CONNECT THE THEORY OF !, AND QUANTUM
CHAOS )T LOOKS PLAUSIBLE THAT THE ASYMPTOTIC OF EIGENFUNCTIONS OF
,APLACIANS IN MIXING CASES HAS SOMETHING COMMON WITH !,
9!' 3).!) '!&!
2EFERENCES
;!+= * !BAD ( +OCH 2ENORMALIZATION AND PERIODIC ORBITS FOR (AMILTON mOWS
#OMM -ATH 0HYSICS SUBMITTED
;"#= - "ENEDICKS , #ARLESON 4HE DYNAMICS OF THE -ENON MAP !NN
-ATH [
;#= " #HIRIKOV 0RIVATE COMMUNICATION
;#O= 9 #OLIN DE 6ERDIs ERE 3PECTRE CONJOINT DOPERATEURS PSEUDO DIkERENTIELS
QUI COMMUTENT ,E CAS INTEGRABLE -ATH :EITSCHRIFT [
;$= 0 $UARTE 0LENTY OF ELLIPTIC ISLANDS FOR THE STANDARD FAMILY !NN )NST
(ENRI 0OINCARsE 0HYS 4HEOR [
;&-= & &LANDOLI " -ASLOWSKI %RGODICITY OF THE $ .AVIER[3TOKES EQUATION
UNDER RANDOM PERTURBATION #OMM -ATH 0HYS [
;'3= - 'OLUBITSKY $ 3CHAEFFER 3INGULARITIES AND 'ROUPS IN "IFURCATION
4HEORY 6OLS 3PRINGER 6ERLAG .EW 9ORK
;'R3= * 'RACZYK ' 3WIATEK 'ENERIC HYPERBOLICITY IN THE LOGISTIC FAMILY
!NN OF -ATH [
;'RE= * 'REENE ! METHOD FOR DETERMINING A STOCHASTIC TRANSITION * -ATH
0HYS [
;(2= $ (EJHAL " 2ACKNER /N THE TOPOGRAPHY OF -AASS WAVEFORMS FOR 03,
: %XPERIMENT -ATH [
;*= - *ACOBSON !BSOLUTELY CONTINUOUS INVARIANT MEASURE FOR CERTAIN MAPS OF
AN INTERVAL #OMM -ATH [
;+3= + +HANIN 9A 3INAI 2ENORMALIZATION 'ROUP -ETHOD AND THE +!-
4HEORY IN <.ONLINEAR 0HENOMENA IN 0LASMA 0HYSICS AND (YDRODYNAMICS
2: 3AGDEEV ED -OSCOW -I2
;+O-3= $ +OSYGIN ! -INASOV 9A 3INAI 3TATISTICAL PROPERTIES OF SPECTRE
OF ,APLACE "ELTRAMI OPERATORS ON ,IONVILLE SURFACES 2USSIAN -ATH 3URVEYS
[
;,= !* ,ICHTENBERG 5NIVERSALITY FROM RESONANCE RENORMALIZATION OF (AMIL
TONIAN MAPS 0HYSICA $ [
;,Y= - ,YUBICH 2EGULAR AND STOCHASTIC DYNAMICS IN THE REAL QUADRATIC FAMILY
0ROC .ATIONAL !CADEMY OF 3CIENCES 53! [
;-= 2 -AC+AY 2ENORMALIZATION IN !REA 0RESERVING -APS 7ORLD 3CIENTIlC

;-E= 6 -ELNIKOV IN <4RANSPORT #HAOS AND 0LASMA 0HYSICS ND ED
3 "ENKADDA & $OVEIL 9 %LSKENS EDS 7ORLD 3CIENTIlC
;0= * 0ALIS ! GLOBAL VIEW OF DYNAMICS AND A CONJECTURE ON THE DENSENESS OF
lNITUDE OF ATTRACTORS PREPRINT )-0! "RAZIL
;3= ! 3HNIRELMAN /N THE ASYMPTOTIC PROPERTIES OF EIGEN FUNCTIONS IN THE RE
GIONS OF CHAOTIC MOTION IN <+!- THEORY AND SEMI CLASSICAL APPROXIMATIONS
TO %IGEN FUNCTIONS 6& ,AZUTKIN ED 3PRINGER 6ERLAG "ERLIN
6ISIONS IN -ATH $9.!-)#!, 3934%-3 !.$ -!4(%-!4)#!, 0(93)#3
;3I= 9A 3INAI 4OPICS IN %RGODIC 4HEORY 0RINCETON 5NIVERSITY 0RESS
;4= 7 4UCKER 4HE ,ORENZ ATTRACTOR EXISTS PREPRINT $EPARTMENT OF -ATHE
MATICS 5PPSALA 5NIVERSITY 3 5PPSALA 3WEDEN
;:= ' :ASLAVSKY 0HYSICS OF #HAOS IN (AMILTONIAN 3YSTEMS )MPERIAL #OLLEGE
0RESS
9A' 3INAI $EPARTMENT OF -ATHEMATICS 0RINCETON 5NIVERSITY 0RINCETON .*

53!
AND
,ANDAU )NSTITUTE OF 4HEORETICAL 0HYSICS
3/-% '%/-%42)#!, #/.#%043 !2)3).' ).

(!2-/.)# !.!,93)3
%LIAS - 3TEIN
4HAT GEOMETRIC CONSIDERATIONS ENTER IN A DECISIVE WAY IN MANY QUESTIONS

OF HARMONIC ANALYSIS IS BY NOW A WELL KNOWN FACT )N EXPLICIT FORM SUCH IDEAS
AROSE lRST IN THE ESTIMATION OF THE &OURIER TRANSFORM OF SURFACE CARRIED MEA
SURE THEY HAVE SINCE PLAYED A KEY ROLE IN AVERAGES OVER LOWER DIMENSIONAL
VARIETIES RESTRICTION THEOREMS IN CONNECTION WITH THE STUDY OF OSCILLATORY
INTEGRALS AND &OURIER INTEGRAL OPERATORS AND IN APPLICATION TO LINEAR AND
NON LINEAR DISPERSIVE EQUATIONS
7HAT WE WANT TO CONSIDER IN THIS PAPER HOWEVER IS ANOTHER FUNDAMEN
TAL OCCURRENCE OF GEOMETRIC CONCEPTS IN THIS SUBJECT (ERE OUR FOCUS WILL
BE ON SINGULAR INTEGRALS AND THE GEOMETRY THAT DESCRIBES THE SINGULARITIES
OF THE KERNELS OF THESE OPERATORS AND WHICH CONTROLS THE RELEVANT ESTIMATES
THAT ARE MADE
4HE KIND OF GEOMETRY THAT ARISES IS <LOCAL IN NATURE AND IS BASED IN
THE lRST INSTANCE ON A <DISTANCE DX Y 4HIS METRIC CONTROLS WHAT
HAPPENS WHEN Y IS NEAR X HOWEVER THE EXACT SIZE OF DX Y IS NOT CRUCIAL
BUT WHAT MATTERS IS THE ORDER OF MAGNITUDE OF D AS Y X
!CCORDING TO PAST EXPERIENCE FOR EACH KIND OF SINGULAR INTEGRAL THAT
AROSE ONE COULD lND RELATED TO IT SUCH A METRIC MORE PRECISELY AN EQUIV
ALENCE CLASS OF SUCH METRICS WHICH WAS USED IN DESCRIBING THE UNDERLYING
FRAMEWORK AND EXPRESSING ALL THE ESTIMATES THAT AROSE 4HIS WAS A VERY
SATISFACTORY STATE OF AkAIRS (OWEVER MORE RECENTLY WE HAVE BEGUN TO REAL
IZE IN VIEW OF SOME INTERESTING EXAMPLES WHICH OCCUR IN SUB ELLIPTIC PDE
AND SEVERAL COMPLEX VARIABLES THAT MATTERS ARE NOT ALWAYS THAT SIMPLE
)N FACT IT MAY WELL HAPPEN THAT THERE APPEAR TWO OR MORE DISTINCT AND
CONmICTING METRICS EACH ARISING NATURALLY FROM THE CONTEXT AT HAND AND
EACH DESCRIBING DIkERENT ASPECTS OF THE SINGULAR INTEGRAL IN QUESTION 3UCH
SINGULAR INTEGRAL OPERATORS ARE BY THEIR NATURE MORE COMPLEX THAN THOSE
HITHERTO STUDIED )N PARTICULAR THE #ALDERs ON[:YGMUND PARADIGM USED TO
PROVE ,P ESTIMATES IS NO LONGER APPLICABLE 4HUS THE PROBLEM THAT NEEDS
TO BE ADDRESSED IS HOW WE CAN REFORMULATE OUR EARLIER VIEWS ABOUT SINGULAR
INTEGRALS TO TAKE THESE MORE GENERAL CIRCUMSTANCES INTO ACCOUNT
6ISIONS IN -ATH '%/-%42)#!, #/.#%043 ). (!2-/.)# !.!,93)3
)N ITS MOST GENERAL FORM A SINGULAR INTEGRAL OPERATOR IS GIVEN AS A LINEAR

MAPPING F 4 F OF THE KIND
:
4 F X +X Y F Y DY
-
(ERE - IS A MANIFOLD AND THE MEASURE DY HAS A SMOOTH SENSITY 4HE
KERNEL + IS A DISTRIBUTION ON - a - WHICH IS <SINGULAR AT LEAST ON THE
DIAGONAL AND POSSIBLY ON OTHER SUB VARIETIES OF - a - BUT IS OTHERWISE
SMOOTH 7HILE THE STUDY OF SUCH OPERATORS REPRESENTS AN EXTREMELY BROAD
AND DIVERSE SUBJECT OUR ATTENTION HERE WILL BE MORE NARROWLY FOCUSED ON THE
PROBLEMS ALLUDED TO ABOVE 4HE ISSUES WE WANT TO DEAL WITH ARE CONNECTED
WITH THE FOLLOWING QUESTIONS
) (OW DO SUCH OPERATORS ARISE IN APPLICATIONS
)) 7HAT IS THE UNDERLYING <GEOMETRY
))) (OW ARE THESE OPERATORS AND THEIR KERNELS DESCRIBED BY THIS GEOM
ETRY
)6 (OW DOES ONE PROVE , OR ,P REGULARITY FOR THESE OPERATORS
7E SHALL ORGANIZE OUR EXPOSITION BY DESCRIBING THREE DIkERENT STAGES OF
DEVELOPMENT OF THESE QUESTIONS &IRST A REVIEW OF SOME PAST RESULTS NEXT
OUR PRESENT POINT OF VIEW IN TERMS OF SEVERAL EXAMPLES OF INTEREST WHERE
WE SHALL DESCRIBE SOME JOINT WORK WITH ! .AGEL AND & 2ICCI WHICH IS
IN PROGRESS AND lNALLY WE WILL INDULGE IN SPECULATION ABOUT SOME FUTURE
DIRECTIONS OF THE SUBJECT
<#LASSICAL 4HEORY
4HIS THEORY CAN BE THOUGHT AS ORIGINATING ABOUT A HALF CENTURY AGO WITH
WORK OF -IHLIN #ALDERs ON AND :YGMUND )T HAS SINCE UNDERGONE NUMEROUS
DEVELOPMENTS /F PARTICULAR NOTE ARE THE FURTHER IDEAS INTRODUCED TO STUDY
THE #AUCHY INTEGRAL ON ,IPSCHITZ CURVES AND THE SUBSEQUENT GENERAL FOR
MULATION REFERRED TO AS THE <4 THEOREM 4HE ESSENCE OF THAT THEOREM
CAN BE STATED IN A SOMEWHAT SIMPLIlED FORM AS FOLLOWS 7E CONSIDER AN
OPERATOR 4 AS IN WHERE - 2N DY ,EBESGUE MEASURE AND THE
DISTRIBUTION + IS # AWAY FROM THE DIAGONAL 7E MAKE TWO ASSUMPTIONS
A $IkERENTIAL INEQUALITIES
k j k k
k X Y +X Y k JX _ YJ_N_JjJ_JkJ

3EE THE EXPOSITIONS IN -EYER ;-E= 3TEIN ;3= ALSO $AVID *OURNsE AND 3EMMES ;$*3=

# REGULARITY CAN BE RELAXED TO # OR EVEN LESS
%- 34%). '!&!
B #ANCELLATION CONDITIONS
K4 K, RN AND K4 b K, RN
4HIS IS TO HOLD FOR ANY WHICH IS A <NORMALIZED BUMP FUNCTION ASSOCIATED
TO A BALL " FX 2N JX _ X J RG OF RADIUS R IE WE REQUIRE ONLY THAT
IS SUPPORTED IN " AND SUP" JJ R SUP" JRJ
.OTE THAT HERE THE UNDERLYING GEOMETRY IS THE OBVIOUS ONE [ IT IS GIVEN
BY THE %UCLIDEAN DISTANCE DX Y JX _ YJ
4HE MAIN RESULT FOR SUCH OPERATORS IS THEIR , AND ,P P
BOUNDEDNESS 4HIS IS PROVED AS IS WELL KNOWN IN TWO STEPS &IRST ONE
DEALS WITH THE , REGULARITY )N THE TRANSLATION INVARIANT SETTING IE WHEN
+X Y +X _ Y THE &OURIER TRANSFORM IS THE NATURAL TOOL TO USE 4HIS
IS ALSO THE CASE WHEN ONE CONSIDERS <VARIABLE COEnCIENT VERSIONS OF THESE
OPERATORS IE PSEUDO DIkERENTIAL OPERATORS (OWEVER FOR THE MORE GENERAL
SITUATION ENVISAGED ABOVE OTHER TOOLS MUST BE USED IN PARTICULAR ARGUMENTS
INVOLVING <ALMOST ORTHOGONALITY
/NCE THE , BOUNDEDNESS IS ESTABLISHED THE ,P REGULARITY IS THEN PROVED
IN A SERIES OF STEPS WHICH IS COMMONLY REFERRED TO AS THE #ALDERsON[:YGMUND
PARADIGM 4HIS APPROACH INVOLVES A COVERING ARGUMENT FOR DISJOINT BALLS
AND A DECOMPOSITION OF A GIVEN FUNCTION INTO A <GOOD PART AND <BAD PART
SUPPORTED ON THE UNION OF THESE BALLS
4HESE RESULTS CAN BE LOCALIZED AND EXTENDED TO THE SITUATION WHERE -
IS A COMPACT MANIFOLD AND THE GEOMETRY IS THAT GIVEN BY A SUITABLE
2IEMANNIAN METRIC 4HIS EXTENSION IS PARTICULARLY RELEVANT FOR APPLICATIONS
7E SHALL BRIEmY RECALL TWO OF THE MOST FUNDAMENTAL EXAMPLES
A 4HE #AUCHY INTEGRAL IN # (ERE - l IS A SUITABLY SMOOTH SIMPLE
CLOSED CURVE IN # WHICH DIVIDES THE PLANE INTO TWO REGIONS j THE INTERIOR
AND j_ THE EXTERIOR &OR F GIVEN ON l WE OBTAIN A PAIR OF HOLOMORPHIC

2 F T
FUNCTIONS &e DElNED IN je RESPECTIVELY BY & Z xI l T_Z DT 4AKING
THE BOUNDARY RADIUS OF &e WE OBTAIN A PAIR OF SINGULAR INTEGRAL OPERATORS
F 4e F 2ECALL THAT IF l IS THE REAL AXIS WITH j THE UPPER HALF SPACE
THEN 4 )I( WHERE ( IS THE (ILBERT TRANSFORM
B %LLIPTIC PDE 7E LIMIT OURSELVES TO THE SIMPLEST SITUATIONS &IRST
<INTERIOR ESTIMATES FOR A SECOND ORDER ELLIPTIC OPERATOR , GIVEN BY
8
,U AIJ X XI UXJ
IIJ JN

! GENERAL FORMULATION OF THIS KIND OF ARGUMENT IS IN #OIFMAN AND 7EISS ;#O7= AND
3TEIN ;3 #HAP =
6ISIONS IN -ATH '%/-%42)#!, #/.#%043 ). (!2-/.)# !.!,93)3
WITH FAIJ X G A SMOOTH REAL SYMMETRIC POSITIVE DElNITE MATRIX 5SING A

FUNDAMENTAL SOLUTION OPERATOR + IE WHERE +, ) % WITH % INlNITELY

SMOOTHING ONE CAN THEN SHOW THAT THE OPERATORS 4 XI XJ + ARE SINGULAR
INTEGRALS OF THE ABOVE KIND
3INGULAR INTEGRALS ALSO ARISE WHEN STUDYING COERCIVE BOUNDARY VALUE
PROBLEMS FOR , 7E MENTION HERE A VERY SPECIAL INSTANCE BECAUSE A VARIANT
WILL ARISE LATER IN X 7E TAKE - j WHERE j IS A BOUNDED DOMAIN IN 2N
WITH SMOOTH BOUNDARY 7E THEN CONSIDER THE <$IRICHLET
k TO .EUMANN OPER
ATOR & $F DElNED AS FOLLOWS $F Uv k- FOR U SATISFYING ,U
WITH UJ- F WHERE v IS THE NORMAL DERIVATIVE AT THE BOUNDARY 4HE
OPERATOR $ CAN BE INVERTED MODULO AN INlNITELY SMOOTHING OPERATOR 7RIT
ING $_ FOR SUCH A RELATIVE INVERSE THEN THE OPERATOR 8$_ IS A SINGULAR
INTEGRAL WITH CONDITIONS ANALOGOUS TO AND WHENEVER 8 IS A
SMOOTH VECTOR lELD ON -
%XTENSIONS 4HE 3ETTING

3TARTING IN THE LATE S AN INCREASING NUMBER OF SUGGESTIVE EXAMPLES
CAME TO LIGHT WHICH LEAD TO A RECASTING OF THE THEORY SO AS NOT TO BE BOUND
BY THE <ELLIPTIC OR 2IEMANNIAN FRAMEWORK DESCRIBED ABOVE &ROM OUR
PARTICULAR POINT OF VIEW THIS DEVELOPMENT CAN BE SCHEMATIZED AS FOLLOWS
2N NILPOTENT GROUPS WITH DILATIONS GEOMETRY ASSOCIATED TO A COL
LECTION OF VECTOR lELDS
4HE HEADING 2N SIGNIlES THE TRANSLATION INVARIANT THEORY IN 2N ALLUDED
TO IN X 4HE NILPOTENT GROUPS ARE ,IE GROUPS EQUIPPED WITH DILATIONS ONE
PARAMETER GROUP OF AUTOMORPHISMS THESE GROUPS WERE ORIGINALLY REFERRED
TO AS <HOMOGENEOUS OR <GRADED 4HE COLLECTION OF VECTOR lELDS ARE GIVEN
IN SOME NEIGHBORHOOD IN 2N AND THE ,IE ALGEBRA THEY GENERATE IS ASSUMED
TO SPAN 4HE RESULTING GEOMETRY IS THEN THE <CONTROL GEOMETRY WE SHALL
DESCRIBE BELOW IN X
7E NOW LIST A NUMBER OF EXAMPLES OF THE ASSOCIATED SINGULAR INTEGRALS
A )NTERTWINING OPERATORS 7E LET ' BE A SEMISIMPLE ,IE GROUP OF NON
COMPACT TYPE WITH )WASAWA DECOMPOSITION ' +!. AND ASSOCIATED SYM
METRIC SPACE '+ ! DISTINGUISHED PART OF ITS BOUNDARY CAN BE IDENTIlED
WITH THE NILPOTENT GROUP . $ILATIONS ON . ARE OF THE FORM N AT NA_ T
WHERE FAT G IS AN APPROPRIATE ONE PARAMETER SUB GROUP OF ! 4HE INTERTWIN
ING OPERATORS OF THE PRINCIPAL SERIES OF REPRESENTATIONS OF ' CAN BE REALIZED

-ORE RECENTLY THEY HAVE BEEN DUBBED <#ARNOT[#ARATHEODORY GROUPS
%- 34%). '!&!
AS <SINGULAR INTEGRALS ON . AND IN THIS SETTING THEIR , STUDY CAME TO

THE FORE (OWEVER THE INTEREST OF THE OPERATORS WAS NOT LIMITED TO REPRE
SENTATION THEORY BUT WAS ALSO RELEVANT IN COMPLEX ANALYSIS )N FACT WHEN
' 35 N THE CORRESPONDING SYMMETRIC SPACE IS THE COMPLEX UNIT
BALL IN #N AND . IS THE (EISENBERG GROUP (N WHICH CAN BE NATURALLY
IDENTIlED WITH THE BOUNDARY OF THE BALL WITH THE EXCEPTION OF ONE POINT
WHICH CORRESPONDS TO THE POINT OF INlNITY FOR (N )T THEN TURNS OUT THAT
THE #AUCHY[3ZEG O PROJECTION OPERATOR IS A PARTICULAR EXAMPLE OF SUCH AN
INTERTWINING OPERATOR 4HUS THE ,P REGULARITY OF THIS OPERATOR COULD BE
SUBSUMED IN THE MORE GENERAL THEORY OF SINGULAR INTEGRALS
B !NALYSIS ON STRONGLY PSEUDO CONVEX DOMAINS 4HE INSIGHTS REGARDING
THE COMPLEX BALL AND THE (EISENBERG GROUP PROVIDE IMPORTANT HINTS FOR THE
STUDY IN THE MORE GENERAL CASE OF STRONGLY PSEUDO CONVEX DOMAINS j IN
#N (ERE ONE TAKES - j AND THE RELEVANT GEOMETRY IS THE CONTROL
GEOMETRY DETERMINED LOCALLY BY VECTOR lELDS F8J 9J GNJ SO THAT : J
8J I9J J N ARE A BASIS OF THE TANGENTIAL #AUCHY[2IEMANN VECTOR
lELDS ON j 4HE FOCUS OF ATTENTION IS THEN THE B COMPLEX THE #AUCHY[
3ZEGO PROJECTION OPERATOR # AND THE INVERSE OF THE +OHN ,APLACIAN cB
b b
B B B B
! RESULT OF THAT ANALYSIS IS THAT THE OPERATOR # AS WELL AS 18 9 c_ B
ARE SINGULAR INTEGRALS ATTACHED TO THE CONTROL METRIC HERE 18 9 IS A
QUADRATIC POLYNOMIAL IN 8 8N 9 9N
0K
C 4HE SUB ,APLACIAN , J 8J 4HE ( ORMANDER OPERATOR IS
DETERMINED BY REAL VECTOR lELDS 8 8 8K WHOSE COMMUTATORS ARE
ASSUMED TO SPAN !S A RESULT , IS HYPOELLIPTIC 4HE FURTHER OPTIMAL ES
TIMATES FOR THE REGULARITY OF , ARE CONSEQUENCES OF THE FOLLOWING /NE CAN
CONSTRUCT A PARAMETRIX 3 AN APPROXIMATE INVERSE FOR , ESSENTIALLY 3,
AND ,3 DIkER FROM ) BY A SMOOTHING OPERATOR 4HEN THE SHARP ESTIMATES
FOLLOW FROM THE FACT THAT THE OPERATORS 8I 8J 3 ARE SINGULAR INTEGRALS ASSO
CIATED TO THE CONTROL METRIC DETERMINED BY F8 8 8K G
7E NOW TURN TO A DESCRIPTION OF THIS CONTROL GEOMETRY AND THE CHARAC

3EE +NAPP AND 3TEIN ;+3=

3EE +ORs
ANYI[6sAGY ;+O6=

%ACH cB IS AN OPERATOR ON Q FORMS AND IS INVERTIBLE IF Q N _ THE CASE TO
WHICH WE LIMIT OURSELVES

3EE &EkERMAN ;&= &OLLAND AND 3TEIN ;&3T=

3EE (
ORMANDER ;(=

3EE 2OTHSCHILD AND 3TEIN ;23= .AGEL 3TEIN AND 7AINGER ;.37= AND &EkERMAN
AND 3ANCHEZ[#ALLE ;&3=
6ISIONS IN -ATH '%/-%42)#!, #/.#%043 ). (!2-/.)# !.!,93)3
TERIZATION OF THE ATTACHED SINGULAR INTEGRALS
#ONTROL -ETRIC AND !SSOCIATED 3INGULAR )NTEGRALS

3UPPOSE 8 8K ARE K GIVEN REAL AND SMOOTH VECTOR lELDS DElNED IN
SOME NEIGHBORHOOD - IN 2N WHOSE COMMUTATORS SPAN 7E THEN DE
lNE THE CONTROL DISTANCE D AS DX Y INF 40 SO THAT THERE IS A CURVE
K
lT WITH l X l4 Y AND lT ? J AJ T 8J lT WHERE
0K
J AJ T 4HE IMPORTANT FACT lRST REALIZED IN SPECIAL CASES BUT
PROVED IN #HOW ;#H= IS THAT ANY TWO NEARBY POINTS X AND Y COULD BE JOINED
BY A CURVE WHOSE TANGENT POINTS IN A DIRECTION SPANNED BY THESE VECTOR
lELDS 4HAT IS
DX Y FOR Y NEAR X
7E NEXT DElNE "X m TO BE THE BALL IN THIS METRIC OF RADIUS m CENTERED
AT X g h
"X m Y DX Y m
!PPLICATION OF THE #ALDERs ON[:YGMUND PARADIGM REQUIRES THE FOLLOWING
PROPERTIES OF THIS METRIC STATED IN TERMS OF THOSE BALLS
A 4HE ENGULlNG PROPERTY "X m < "Y m } "Y m y "X
m 4HIS FOLLOWS IMMEDIATELY WITH C SINCE D IS A METRIC
B 4HE DOUBLING PROPERTY J"X m J #J"X m J
4O PROVE B AND FOR THE FURTHER ANALYSIS BELOW ONE NEEDS AN ANALYTIC
DESCRIPTION OF THESE BALLS 4O THIS WE NOW TURN 7ITH 8 8K OUR
GIVEN VECTOR lELDS WE DElNE A COMMUTATOR OF DEGREE R TO BE THE VECTOR
lELD 8) ;8I ;8I ;` ` ` 8IR === WHERE ) I I IR IJ K AND
DEGREE ) J)J R /UR ASSUMPTION OF THE SPANNING OF THE COMMUTATORS
IS THEN THE EXISTENCE OF AN INTEGER M SO THAT THE COLLECTION F8) G J)J M
SPAN
.EXT A FRAME F IS DElNED TO BE ANY N TUPLE OF THE 8) F
F8) 8) 8)N G 7E DElNE DEGREE F JF J J) J J) J ` ` ` J)N J
AND DET F DET8) 8) 8)N X 7ITH THESE WE FORM THE BASIC <IN
VARIANT cX m WHICH FOR EACH X - IS A POLYNOMIAL IN m OF DEGREE M

)T HAS BEEN SHOWN BY .AZAROV 4REIL AND 6OLBERG ;.A46= THAT IN MANY CIRCUMSTANCES
ONE DOES NOT NEED TO ASSUME THIS PROPERTY (OWEVER BECAUSE OF THE NATURE OF OUR
FRAMEWORK IN PARTICULAR BELOW THIS PROPERTY STILL ARISES IN AN ESSENTIAL WAY IN
WHAT FOLLOWS

&OR DETAILS OF THE FACT GIVEN HERE SEE .AGEL 3TEIN AND 7AINGER ;.37= ! DIkERENT
ANALYSIS HAVING INTERESTING CONNECTIONS WITH ASYMPTOTICS OF EIGEN VALUES IS IN &EkERMAN
AND 0HONG ;&0= AND &EkERMAN AND 3ANCHEZ[#ALLE ;&3=
%- 34%). '!&!
GIVEN BY 8 k k
cX m kDETF X km JF J
JF JM
&OR EACH X THE BEHAVIOR OF cX m AS m DETERMINES THE LOCAL GEOM
ETRY NEAR X )N FACT CHOOSE FOR EACH X m A FRAME F SO THAT JDETF Jm JF J
CcX m FOR SOME SMALL BUT lXED CONSTANT 4HIS CAN BE DONE BY CHOOSING
A LARGEST TERM IN 4HEN ONE HAS
0ROPOSITION N o8 p O
"X m x EXP AJ 8)J X JAJ J m J)J J
)J F
k k
k"X m k x cX m
#ONCLUSION IMMEDIATELY IMPLIES THE DOUBLING PROPERTY 4HE ASSER
TION STATES THAT THE BALL "X m HAS AXIS LENGTH m J)J J IN THE DIRECTION
OF 8)J )T IS IMPORTANT TO NOTE THAT EVEN IF X IS lXED THE COORDINATE SYS
TEM IE CHOICE OF FRAME USED TO DESCRIBE THE BALL MAY DEPEND ON m IN
FACT lNITELY MANY DIkERENT CHOICES OF COORDINATE SYSTEMS MAY BE NEEDED
4HESE CIRCUMSTANCES ARE ILLUSTRATED BY THE FOLLOWING SIMPLE BUT SUGGESTIVE
EXAMPLE
(ERE - 2 FX Y G AND WE TAKE TWO VECTOR lELDS 8 X 9
X Y 4HE ONLY COMMUTATION RELATION IS : ;8 9 = Y )T FOLLOWS THAT
cX m JXJm m AND HENCE
g h
EXPA8 B9 X Y JAJ m JBJ m IF JXJ m
"X m x g h
EXPA8 C: X Y JAJ m JCJ m IF JXJ m
7E NOW TURN TO THE ANALYTIC CONSEQUENCES OF THIS GEOMETRIC FRAMEWORK
WITH - ENDOWED WITH THE METRIC D ARISING AS ABOVE WE CONSIDER THE OPER
ATORS OF THE FORM :
4 F X +X Y F Y DY
-
4HE KERNEL + WILL BE ASSUMED TO BE SMOOTH AWAY FROM THE DIAGONAL
AND THE FOLLOWING ADDITIONAL PROPERTIES WITH R lXED THE <NON ISOTOPIC
SMOOTHING ORDER OF 4
k j k k R_JjJ_JkJ
k8X 8Y +X Y k DX Y
6 X Y

4HIS FUNDAMENTAL FACT SEEMS TO HAVE BEEN MISUNDERSTOOD BY SOME LATER WRITERS &OR
INSTANCE 'ROMOV ;'R P = MISTATES REQUIRING ONLY ONE FRAME THE PROOF OF THE
DOUBLING PROPERTY P IS THEREFORE INCORRECT "ELLA]pCHE ;"= DEALS WITH THE EXAMPLE
BELOW BUT DOES NOT ALLUDE TO THE FACT THAT THERE TWO COORDINATE SYSTEMS ARE NEEDED
6ISIONS IN -ATH '%/-%42)#!, #/.#%043 ). (!2-/.)# !.!,93)3
(ERE 6 X Y IS THE VOLUME OF THE BALL "X m WITH m DX Y SO
6 X Y x cX m !LSO 8 j 8j 8j ` ` ` 8KjK JjJ j j ` ` ` jK
/NE ALSO ASSUMES
K4 K, K4 b K, m R J"X m J
4HE FUNCTION IS ANY NORMALIZED BUMP FUNCTION FOR THE BALL "X m
IE IS SUPPORTED IN "X m AND SATISlES
y 8K z
SUP JJ m J8J J
" J
/N THE BASIS OF AND ONE CAN PROVE THE FOLLOWING
0ROPOSITION A )F R THE OPERATOR 4 IS BOUNDED ON ,P P
B 4HE PARAMETRIX 3 FOR THE SUB ,APLACIAN , DISCUSSED IN X IS AN
OPERATOR OF THE ABOVE TYPE WITH R
C 4HE OPERATORS 8I 8J 3 ARE OF THE ABOVE TYPE WITH R
)T IS USEFUL TO REMARK THAT IN ALL THE ABOVE THE METRIC D CAN BE REPLACED
BY AN EQUIVALENT METRIC D IN THE FORMULATIONS AND CONCLUSIONS STATED 4HE
EQUIVALENCE OF D AND D IS THE ASSERTION THAT C DDXY
XY C FOR A PAIR
OF POSITIVE CONSTANTS C AND C /NE CAN EVEN EXTEND THIS TO A WIDER FORM
OF EQUIVALENCE IN WHICH C uD D C WHERE u IS A SUITABLE STRICTLY
MONOTONE FUNCTION 4HEN FOR EACH X THE FAMILY OF BALLS F"X m G AND
F" X m G ARE ESSENTIALLY THE SAME THEY DIkER ONLY INSOFAR THAT F" X m G
ARISES ESSENTIALLY FROM "X m G BY REPARAMETRIZING THE RADIUS m
4HE #ONmICT OF -ETRICS &IRST %XAMPLE

4HE SITUATION DESCRIBED ABOVE [ WHERE IN A GIVEN ANALYTIC CONTEXT THERE
ARISES A NATURAL METRIC WHICH DETERMINES THE RELEVANT CLASS OF SINGULAR IN
TEGRALS [ REPRESENTS A VERY SATISFACTORY STATE OF AkAIRS 5NFORTUNATELY THERE
ARE NOW AN INCREASING NUMBER OF EXAMPLES WHICH DO NOT FALL UNDER THE SCOPE
OF THESE IDEAL CIRCUMSTANCES THERE MAY IN FACT BE TWO DIkERENT AND CONmICT
ING METRICS EACH INHERENT IN THE PARTICULAR SITUATION AND EACH DESCRIBING
DIkERENT FEATURES OF THE RELEVANT SINGULAR INTEGRALS THAT ARISE 7E SHALL GIVE
TWO EXAMPLES OF THIS
)N THE lRST INSTANCE WE TAKE j TO BE A SMOOTH AND BOUNDED DOMAIN IN
#N AND - j 5NDER THE CONDITIONS ON j BELOW WE ARE LEAD TO TWO
DIkERENT METRICS

/F COURSE THEN THE CONDITIONS AND MUST BE APPROPRIATELY REINTERPRETED
%- 34%). '!&!
I THE <CONTROL METRIC

II THE <3ZEGO METRIC
&OR I WE ASSUME THAT j IS OF lNITE TYPE IN THE SENSE THAT THE ,IE
ALGEBRA GENERATED BY THE TANGENTIAL #AUCHY[2IEMANN VECTOR lELDS AND THEIR
CONJUGATES SPAN THE TANGENT SPACE OF - 4HEN WE DElNE THE CONTROL METRIC
D# AS IN X WHERE 8 8N ARE THE REAL AND IMAGINARY OF THE #AUCHY[
2IEMANN VECTOR lELDS
4O DElNE II WE ASSUME THAT j IS CONVEX AND OF lNITE TYPE IN THE SENSE
THAT ANY COMPLEX LINE HAS AT MOST A lNITE ORDER CONTACT WITH - j 4HE
BALLS OF THIS METRIC D3 ARE DETERMINED AS FOLLOWS &OR EACH X j WE LET
vX DENOTE THE UNIT INWARD NORMAL AT X 4HEN "X m IS THE <CAP ON j
<CUT OUT BY INSERTING A MAXIMAL POLYDISC IN j CENTERED AT X mvX )T
TURNS OUT THAT IN THIS lNITE TYPE CONVEX SITUATION THE 3ZEGO METRIC IS THE
ONE THAT CONTROLS THE #AUCHY[3ZEG O PROJECTION AS A SINGULAR INTEGRAL
7E NOW MAKE THE FOLLOWING REMARKS
/BSERVATION 7HEN j IS STRONGLY PSEUDO CONVEX EG THE UNIT BALL
IN #N THEN DC x D3 SO THAT THESE METRICS ARE COMPARABLE IN THE
WIDE SENSE DESCRIBED IN X AND THEREFORE GIVE ESSENTIALLY THE SAME FAMILY
OF BALLS
/BSERVATION )N GENERAL HOWEVER THIS IS NOT THE CASE 4AKE FOR
EXAMPLE j y # FZ Z W G WITH j F)MW JZ JN JZ JM G

WHERE N AND M ARE INTEGERS AND N M AND M )F T 2EW THE

POINTS OF - j CAN BE PARAMETRIZED BY Z Z T AND A NATURAL PAIR OF
TANGENTIAL #AUCHY[2IEMANN RADIUS ON - IS GIVEN BY
: Z _ INJZ JN_ Z T 8 I9
: Z _ IMJZ JM_ Z T 8 I9
)N THIS CASE THE DISTANCES OF A POINT Z Z T TO THE ORIGIN ARE GIVEN
ESSENTIALLY BY
D# Z Z T x JTJN JZ J JZ J
D3 Z Z T x JTJ JZ JN JZ JM
! COMPARISON OF THESE FORMULAE HIGHLIGHTS THE DIkERENT NATURES OF THE
METRICS D# AND D3 .AMELY THE CONTROL METRIC D# WEIGHS THE Z AND Z

4HESE BALLS ARE DESCRIBED IN -C.EAL ;-C= THERE IS AN EARLIER VERSION IN THE REAL CASE
CONSIDERED IN "RUNA .AGEL AND 7AINGER ;"R.7= )T IS REMARKABLE THAT IN THE LATTER
SITUATION THE BALLS ARE INTIMATELY CONNECTED WITH THE DECAY OF THE &OURIER TRANSFORM OF
SMOOTH MEASURES CARRIED ON j

&OR THIS SEE -C.EAL AND 3TEIN ;-C3=
6ISIONS IN -ATH '%/-%42)#!, #/.#%043 ). (!2-/.)# !.!,93)3
DIRECTIONS EQUALLY AND WEIGHS THE T DIRECTION ACCORDING TO THE LEAST NUMBER
OF COMMUTATORS NEEDED TO SPAN THAT DIRECTION (OWEVER THE 3ZEGO MET
RIC D3 ASSIGNS DIkERENT WEIGHTS TO Z AND Z CALIBRATED IN TERMS OF THEIR
DIkERING RELATION TO THE T DIRECTION
&OR THE DOMAIN j AND ITS BOUNDARY - THERE ARE THREE BASIC AND RELATED
OPERATORS OF INTEREST
# #AUCHY[3ZEG
O PROJECTION

_cB : : : :
, 8 9 8 9
.OW # IS CONTROLLED BY THE METRIC D3 AS NOTED ABOVE !LSO THE INVERSE

OF , IS CONTROLLED BY D# SEE X (OWEVER _cB , lRST ORDER TERMS
AND WOULD SEEM TO REmECT BOTH D# AND D3 !DDED TO THIS IS THE REMARKABLE
FACT POINTED OUT IN 2OTHSCHILD ;2= THAT AS OPPOSED TO THE STRONGLY PSEUDO
CONVEX SITUATION THE OPERATOR cB IS NOT MAXIMALLY SUB ELLIPTIC ! QUESTION
THAT ARISES IS
0ROBLEM 7HAT IS THE NATURE OF c_ B THE RESULTING SINGULAR INTEGRALS
AND THEIR REGULARITY
3OME OF THESE QUESTIONS IN THE CASE WHERE N M HAVE BEEN
CONSIDERED BY -ACHEDON ;-= )N THAT CASE BESIDES THE USUAL CONTROL METRIC
HE CONSTRUCTS A SECOND <CONTROL METRIC BASED ON DIkERENT VECTOR lELDS
WHICH HAPPENS TO BE EQUIVALENT WITH THE 3ZEGO METRIC DElNED ABOVE (E
ALSO SHOWS THAT THE KERNEL FOR c_
B SATISlES CERTAIN DIkERENTIAL INEQUALITIES
INVOLVING BOTH METRICS (OWEVER THE CONSTRUCTION FOR THE SECOND METRIC
AND RESULTING ANALYSIS SEEMS TO BE APPLICABLE ONLY IN THE CASE N
4HE #ONmICT OF -ETRICS 3ECOND %XAMPLE

4HIS EXAMPLE ARISES WHEN WE CONSIDER THE .EUMANN PROBLEM 7E PRESENT
A MODIlED AND SIMPLIED VERSION OF THE PROBLEM
7E TAKE j TO BE A BOUNDED SMOOTH DOMAIN IN #N 2N WHICH WE
ALSO ASSUME IS OF lNITE TYPE #ONSIDER THE INHOMOGENEOUS ,APLACE EQUATION
aU F IN j

3TRICTLY SPEAKING THIS REPRESENTS ONLY ONE COMPONENT AT cB ON FORMS BUT IS
NEVERTHELESS TYPICAL

)N THIS VERSION THE PROBLEM IS FOR A SCALAR VALUED FUNCTIONS INSTEAD OF A CORRESPONDING
SYSTEM
%- 34%). '!&!
TAKEN WITH THE COMPLEX .EUMANN LIKE BOUNDARY CONDITION

_ ` k
k
v I { U A ` U j
(ERE v DENOTES THE INWARD UNIT NORMAL { * { WHERE *
IS THE INVOLUTION OF THE COMPLEX STRUCTURE OF #N THUS v I { IS THE
RESTRICTION TO THE BOUNDARY OF A TRANSVERSE #AUCHY[2IEMANN VECTOR lELD
A IS AN APPROPRIATE COMPLEX VALUED FUNCTION
3OLUTION OF THIS PROBLEM CAN BE REDUCED TO THE INVERSION OF A CORRE
SPONDINGm BOUNDARY OPERATOR A <$IRICHLET TO COMPLEX .EUMANN OPERATOR
& $& DElNED AS FOLLOWS &OR & GIVEN ON j LET m 5 BE THE SOLUTION
OF THE $IRICHLET PROBLEM a5 5 J j & 4HEN $ & IS DElNED TO BE
_ ` k
v I { A 5 j k
m
.OW $ IS A lRST ORDER PSEUDO DIkERENTIAL OPERATOR !S FAR AS ITS INVERT
IBILITY AND THE NATURE OF THIS INVERSE NOT MUCH IS KNOWN IN GENERAL (OW
EVER IT IS VERY REASONABLE TO EXPECT THAT TWO DIkERENT METRICS MUST COME
INTO PLAY IN THESE QUESTIONS 4HERE IS lRST THE RESTRICTION OF THE %UCLIDEAN
METRIC OF 2N TO THE BOUNDARY WE CALL THIS METRIC D% IT REmECTS THE
%UCLIDEAN STRUCTURE INHERENT IN THE ,APLACIAN a 4HIS INTERIOR INmUENCE
COMPETES WITH THE CONTROL METRIC D# WHICH REmECTS THE INTERPLAY OF THE
BOUNDARY WITH THE COMPLEX STRUCTURE OF #N
"ESIDES THESE
m GENERAL OBSERVATIONS WE DO NOT KNOW A GREAT DEAL ABOUT
THE INVERSE OF $ EXCEPT WHEN EITHER j IS STRONGLY PSEUDO CONTEXT OR WHEN
N AND j IS PSEUDO CONVEX OF lNITEm TYPE 4HAT SITUATION HAS BEEN
STUDIED IN DETAIL AND IT IS KNOWN THAT $ IS INVERTIBLE WHEN THE RANGE OF A
EXCLUDES CERTAIN FORBIDDEN VALUES )N THE RELATED SPECIAL CASE CORRESPONDING
TO THE UNIT BALL WHEN THE BOUNDARY IS REALIZED AS (N THE INVERSE OPERATOR
IS A CONVOLUTION OPERATOR F 4 F F + WHOSE KERNEL0CAN BE DESCRIBED
RATHER EXPLICITLY 2EMARKABLY IT CAN BE WRITTEN + K@ WHERE
K@ +
EACH SUMMAND + K@ IS ITSELF THE PRODUCT OF HOMOGENEOUS FUNCTIONS WITH
RESPECT TO THE TWO DILATIONS WITH m
Z T _ mZ mT
Z T _ mZ m T
4HE lRST IS THE NON AUTOMORPHIC ISOTROPIC DILATIONS CONNECTED WITH
THE <%UCLIDEAN METRIC D% 4HE SECOND ARE THE AUTOMORPHIC DILATIONS CON
NECTED WITH THE CONTROL METRIC D# ON (N -ORE PRECISELY
+ K@ Z T %K Z T (@ Z T

3EE 'REINER AND 3TEIN ;'3= 4REVES ;4= AND #HANG .AGEL AND 3TEIN ;#.3=
6ISIONS IN -ATH '%/-%42)#!, #/.#%043 ). (!2-/.)# !.!,93)3
WHERE %K IS HOMOGENEOUS OF DEGREE _K WITH RESPECT TO AND (@ IS

HOMOGENEOUS OF DEGREE _@ WITH RESPECT TO AND BOTH ARE SMOOTH
AWAY FROM THE ORIGIN
4HE MOST INTERESTING <CRITICAL CASE OF SUCH OPERATORS ARISES WHEN K
N @ ! TYPICAL EXAMPLE IS
WZ T
+Z T N
`
JZJ T JZJ IT
WHERE IS HOMOGENEOUS OF DEGREE WITH RESPECT TO IS SMOOTH WHEN
JZJ T AND HAS VANISHING MEAN VALUE ON THE SPHERE JZJ T
)T IS CLEAR THAT THE KERNEL DISPLAYS FEATURES OF BOTH METRICS THE
lRST FACTOR IS CONTROLLED BY D% THE SECOND IS CONTROLLED BY D# 4HE QUESTION
THEN BECOMES
0ROBLEM (OW ARE WE TO UNDERSTAND THE CLASS OF CONVOLUTION OPERATORS
WHOSE KERNELS ARE TYPIlED BY (OW DOES ONE PROVE THEIR , AND ,P
REGULARITY
! .EW 6IEWPOINT
7HAT WE WANT TO DO IN THE REST OF THIS ESSAY ARE THREE THINGS
A 0RESENT AN INITIAL FRAME WORK IN TERMS OF WHICH WE CAN HOPE TO DEAL
WITH THESE FURTHER CLASSES OF SINGULAR INTEGRALS
B %XPLAIN MORE CONCRETELY HOW THIS POINT OF VIEW ALLOWS ONE TO RESOLVE
THE QUESTIONS RAISED IN X AND X
C 3PECULATE ABOUT SOME OF THE FEATURES OF A GENERAL THEORY THAT MIGHT
ENCOMPASS THESE IDEAS
4URNING TO THE lRST POINT WE SHALL TRY TO EXPLAIN MATTERS IN TERMS OF
WHAT MIGHT BE CALLED THE <TWO FACTOR CASE 4HE PICTURE THAT SHOULD BE
KEPT IN MIND IS THE FOLLOWING
- a -

x
Y
-
(ERE - IS OUR UNDERLYING SPACE [ WHICH IS THE WORLD WE ACTUALLY SEE
!BOVE IT WHAT WE DONT SEE BUT IS MERELY A CONSTRUCT IS A PRODUCT OF SPACES

3EE 0HONG AND 3TEIN ;03=
%- 34%). '!&!
- a - 7HAT IS GOING ON WILL BE ACCOUNTED FOR BY THIS PRODUCT AND THE

PROJECTION x "ESIDES <PRODUCTS AND <PROJECTIONS THE OTHER KEY WORDS
ATTACHED TO THIS VIEW POINT ARE <LIFTINGS AND <mAGS
7E WILL ILLUSTRATE THIS lRST IN TERMS OF THE OPERATOR ARISING ON THE (EISEN
BERG GROUP IN X WHOSE KERNEL IS GIVEN BY
Z T
+Z T `
JZJ T N JZJ IT
/NE PROCEEDS BY INTRODUCING AN ADDITIONAL VIRTUAL SINGULARITY IN +
THAT IS WE DEAL WITH IT AS IF IT WERE THE KERNEL + Z T GIVEN BY
Z
+ Z T N
`
JZJ JZJ IT
4HE SINGULARITIES OF + ARE NOW REPRESENTED BY A mAG STRUCTURE
FG y 3 y (N
WHERE 3 IS THE CENTRAL SUB GROUP F T G 4HE KERNEL + HAS ITS HEAVIEST
SINGULARITY AT THE ORIGIN IT HAS ALSO A SINGULARITY ON 3 [ BUT TO A LESSER
DEGREE ON THE COMPLEMENT OF 3 IT IS REGULAR 4HE DIkERENTIAL INEQUALITIES
SATISlES BY + AND A FORTIORI BY + ARE
k j k k _ `
k Z + Z T k JZJ_N_JjJ JZJ JTJ __k EQNO
T
4HE KERNELS ALSO SATISFY APPROPRIATE CANCELLATION CONDITIONS INVOLVING

SEPARATE INTEGRATIONS IN THE Z AND T VARIABLES
7E NOW OUTLINE HOW WE CAN DEAL WITH SINGULAR INTEGRALS WHICH HAVE SUCH
KERNELS EG + OR +
n 4HE PICTURE BECOMES IN THIS CASE
(N a 2

x
Y
(N
m
WITH xZ T U Z T U 4HIS REALIZES (N AS (N a 2 ` WHERE ` IS
THE SUB GROUP F U _U G
n 7E NEXT LIFT THE KERNEL + ON (N TO A KERNEL + E ON (N a 2 SO THAT
E
+ IS A <PRODUCT KERNEL !MONG OTHER THINGS THIS MEANS + E SHOULD SATISFY

(ERE WE DESCRIBE THE TREATMENT GIVEN IN -
ULLER 2ICCI AND 3TEIN ;-
U23=
6ISIONS IN -ATH '%/-%42)#!, #/.#%043 ). (!2-/.)# !.!,93)3
THE DIkERENTIAL INEQUALITIES

k j k l k _ `
_N__JjJ_k
k E k ` JUJ__l
Z T U +Z T U ` JZJ JTJ
AND APPROPRIATE CANCELLATION CONDITIONS INVOLVING SEPARATE INTEGRATIONS IN
THE Z T AND U VARIABLES
n 4HE LIFTED KERNEL +E IS CONSTRUCTED SO AND THIS IS THE MAIN POINT THAT
IT <PROJECTS TO + IE
:
+Z T E T _ U U DU
+Z
_
TAKEN IN THE SENSE OF DISTRIBUTIONS
n /PERATORS 4E CORRESPONDING TO PRODUCT KERNELS + E CAN ULTIMATELY BE
DEALT WITH IN TERMS OF TENSOR PRODUCTS OF OPERATORS ACTING EACH VARIABLE
SEPARATELY AND IN THIS WAY ONE CAN OBTAIN THEIR ,P BOUNDEDNESS
n &ROM THE ,P REGULARITY OF 4E WE OBTAIN THAT OF 4 BY TRANSFERENCE USING
THE PROJECTION x IE
! .EW 6IEWPOINT ))
4HE lRST STEP IN A GENERAL THEORY IS A PROOF THAT THE RESULTS FOR THE (EISEN
BERG GROUP EXTEND TO ANY NILPOTENT ,IE GROUP WITH DILATIONS
7E ASSUME . IS SUCH A GROUP WHICH WE IDENTIFY WITH ITS ,IE ALGEBRA D
! mAG FOR . WILL BE AN INCREASING SET OF HOMOGENEOUS SUB GROUPS
FG y 3 y 3 ` ` ` y 3@ .
7E IDENTIFY 3J WITH ITS ,IE ALGEBRA IJ OBTAINING A lLTRATION OF SUB ALGEBRAS
FG y I y I ` ` ` y I@ D
/UR ASSUMPTION ABOUT THE mAG IS THAT THERE IS A GRADATION IJ
IJ_ g WJ OF SUBSPACES WJ J K SO THAT ;WJ WK = IF J K
4HIS IS EQUIVALENT WITH THE EXISTENCE OF ANOTHER GROUP . E WHICH IS THE
E
DIRECT PRODUCT . . a . ` ` ` a .@ WHERE EACH .J IS HOMOGENEOUS AND
A HOMORPHISM x OF . E ONTO . SO THAT
3J x. a . ` ` ` a .J a FG ` ` ` FG J @
)N ANALOGY WITH THE ABOVE WE CAN DElNE mAG KERNELS WITH RESPECT TO THE
mAG !S IN THE CASE OF (N EACH SUCH KERNEL CAN BE LIFTED TO A PRODUCT

4HIS PROCEDURE WHICH USES SQUARE FUNCTIONS AND ,ITTLEWOOD[0ALEY THEORY WAS USED
IN THE CONTEXT OF 2N IN &EkERMAN AND 3TEIN ;&3T=

4HIS IS THE SUBJECT OF A FORTHCOMING PAPER OF .AGEL 2ICCI AND 3TEIN ;.23=
%- 34%). '!&!
KERNEL ON . E 4HE PROPERTIES OF THE CONVOLUTION OPERATOR ON . ARE THEN

CONSEQUENCES OF THE CORRESPONDING PROPERTIES ON . E
4HE THEORY JUST DESCRIBED HAS A NUMBER OF FURTHER APPLICATIONS )N THE
CASE WHERE . 2N IT CAN BE USED TO OBTAIN AN EXPLICIT DESCRIPTION OF THE
SINGULARITIES AND ,P REGULARITY OF THE #AUCHY[3ZEG O PROJECTION FOR TUBE
DOMAINS OVER POLYGONAL CONES !LSO THE CASE WHEN . IS A TWO STEP NILPO
TENT GROUP ALLOWS ONE TO GENERALIZE THIS TO THE #AUCHY[3ZEGO PROJECTIONS
ASSOCIATED TO #2 SUB MANIFOLDS OF #. WHICH ARE GIVEN IN TERMS OF SIMUL
TANEOUSLY DIAGONALIZABLE QUADRATIC FORMS 4HE SINGULAR INTEGRALS FOR THE
CORRESPONDING mAGS MAKE IT POSSIBLE TO STUDY THE RELATIVE REGULARITY OF cB
FOR THESE SUB MANIFOLDS
4HE GENERAL POINT OF VIEW DESCRIBED HERE ALSO APPLIES TO THE EXAMPLE
IN X 7E SKETCH VERY BRIEmY THE SET UP
3TARTING WITH j y # GIVEN BY j F)MW h Z h Z G
h Z JZ JN h Z JZ JM 7E LET
g h
- j )MW h Z h Z
AND SET -J jJ F)MWJ hJ ZJ G y # J (ERE x
- a - - IS GIVEN BY xW Z W Z W W Z Z
/NE AGAIN STUDIES mAG TYPE OPERATORS (ERE THERE ARE TWO ALTERNATIVE
mAG STRUCTURES EACH CORRESPONDING TO A VARIABLE X -
FXG y 6XI y - I

(ERE 6X x- aFv X G WHERE v
X X
v IS ANY POINT SO THAT xv
X X
v X

4HERE IS A SIMILAR DElNITION OF THE ALTERNATE MANIFOLD 6X )T THEN TURNS
OUT THAT THE OPERATOR c_B CAN BE EXPRESSED IN TERMS OF APPROPRIATE KERNELS
ATTACHED TO THESE mAGS 4HEIR ANALYSIS LEADS TO SHARP ESTIMATES AND ALSO
ACCOUNTS FOR THE RESTRICTED SUB ELLIPTICITY MENTIONED IN X
7E NOW TAKE UP AGAIN THE ISSUE OF THE OCCURRENCE OF A MULTIPLICITY
OF METRICS ON - WHICH MOTIVATED OUR CONSIDERATIONS IN X AND X 7E
READILY SEE THAT THIS PHENOMENON CAN BE SEEN AS ARISING FROM THE ABOVE
FRAME WORK 4HE STARTING POINT IS THE PREMISE THAT EACH FACTOR -J CARRIES
ITS OWN NATURAL METRIC 4HESE INDIVIDUAL METRICS CAN BE COMBINED IN A
VARIETY OF DIkERENT WAYS TO GIVE A METRIC ON THE PRODUCT - a - 5SING
THE PROJECTION x - a - - WE THEN OBTAIN A NUMBER OF DIkERENT
METRICS ON -

&OR FURTHER DETAILS SEE .AGEL 2ICCI AND 3TEIN ;.23=

4HIS IS THE SUBJECT OF A FORTHCOMING PAPER OF .AGEL AND 3TEIN ;.3=
6ISIONS IN -ATH '%/-%42)#!, #/.#%043 ). (!2-/.)# !.!,93)3
)N THE EXAMPLE WHERE - (N - (N - 2 THE NATURAL METRIC

ON (N AND 2 ARE RESPECTIVELY GIVEN BY D JZJ JTJ D JUJ HERE
D AND D DENOTE THE DISTANCE OF Z T (N AND U 2 FROM THE ORIGIN

#ONSIDER THE METRICS D^% D D AND D^# D D ON - a - 4HEN
UNDER x D^% AND D^# PROJECT TO METRICS EQUIVALENT WITH D% AND D# IN ANY
BOUNDED NEIGHBORHOOD OF THE ORIGIN
3OME !SPECTS OF A 'ENERAL 4HEORY

4HE VARIOUS EXAMPLES WE HAVE DEALT WITH AND THE RESULTS ALREADY OBTAINED
FOR HOMOGENEOUS GROUPS RAISE THE QUESTION OF lNDING A GENERAL FORMULATION
FOR SINGULAR INTEGRALS ENCOMPASSING THESE IDEAS !T THIS STAGE WE CAN HAVE
NO CLEAR PICTURE BUT WE CAN NEVERTHELESS SPECULATE ABOUT THE POSSIBLE SHAPE
OF THE OUTCOME )N TRYING TO DO THIS WE CAN ENVISAGE A FRAMEWORK BASED ON
CERTAIN GEOMETRIC DATA WHICH WE NOW DESCRIBE
&IRST WE SUPPOSE WE ARE GIVEN FOR EACH X - A mAG OF SUBMANIFOLDS
CENTERED AT X OF INCREASING DIMENSIONS
FXG y 6X y 6X ` ` ` y 6X@ - X-
J
(ERE EACH 6X IS A SUB MANIFOLD DEPENDING SMOOTHLY ON X
3ECOND CONNECTED WITH THIS THERE IS ASSUMED TO BE AN INCREASING COL
LECTION OF VECTOR lELDS
` y ` ` ` ` y `@
WHERE `@ CONSISTS OF VECTOR lELDS 8 8 8K AND THEIR COMMUTATORS UP
TO A CERTAIN ORDER )T WOULD ALSO BE SUPPOSED THAT THE VECTORS IN `J SPAN
J
THE TANGENT SPACE OF 6X AT X
4O SUCH AN UNDERLYING GEOMETRIC STRUCTURE WE SHOULD LIKE TO ASSOCIATE
SINGULAR INTEGRALS
:
4 F X +X Y F Y DY
-
4HE CONNECTION WITH THE DATA AND WE HAVE IN MIND THAT THE
J
DISTRIBUTION +X Y WOULD BE ALLOWED TO BE SINGULAR WHEN Y 6X J @
AND THAT THE <DEGREE OF THAT SINGULARITY MEASURED IN TERMS OF THE VECTOR
lELDS IN ` WOULD DECREASE AS J INCREASES
4HIS RATHER VAGUE DESCRIPTION MUST OF COURSE BE MADE PRECISE 4O DO
THIS A NUMBER OF ISSUES WOULD HAVE TO BE ADDRESSED AND RESOLVED
7HAT ARE THE FURTHER GEOMETRIC PROPERTIES OF THE 6 J AND `J THAT
NEED TO BE ASSUMED
%- 34%). '!&!
7HAT ARE THE DIkERENTIAL INEQUALITIES AND CANCELLATION CONDITIONS THAT

SHOULD BE IMPOSED ON +
(OW DOES ONE PROVE THE , AND ,P BOUNDEDNESS OF THESE OPERATORS
$O THESE OPERATORS FORM AN ALGEBRA
4O ANSWER THESE QUESTIONS FULLY LOOKS AT PRESENT TO BE QUITE A DAUNTING
TASK 4HERE IS HOWEVER A FURTHER EXAMPLE THAT MIGHT PROVIDE SOME USEFUL
HINTS THE <SINGULAR 2ADON TRANSFORMS
)N TERMS OF THE FORMULATION SKETCHED ABOVE THESE OPERATORS ARISE WHEN
WE ARE DEALING WITH A TWO STEP mAG
FXG y 6X y -
WHERE EACH 6X IS K DIMENSIONAL WITH K N AND WHEN WE RESTRICT
OURSELVES TO THE SPECIAL CASE WHEN THE KERNEL + IN IS ENTIRELY SUPPORTED
ON FY 6X G
,ET US TRY TO MAKE THIS CONNECTION A LITTLE MORE PRECISE /NE REALIZES
6X IN PARAMETRIZED FORM 6X FlT X T 5 G WITH X T lT X A SMOOTH
MAPPING OF - a 5 - WHERE 5 IS AN OPEN NEIGHBORHOOD OF THE ORIGIN
IN 2K WE ALSO SUPPOSE l X p X 3UCH MAPPINGS lT HAVE AN ASYMPTOTIC
EXPANSION o8 p
lT X w EXP 8j Tj j X AS T
j
FOR APPROPRIATE VECTOR lELDS F8j G )T IS CONVENIENT TO SET 8*
;8j ;8j ;` ` ` 8jJ === AND ASSIGN WEIGHT Jj J Jj J ` ` ` JjJ J TO 8* WHERE
* j j jJ 4HE ASSUMPTION THAT IS MADE ON lT IS THAT THE F8* G
SPAN THE TANGENT SPACE OF - ! SINGULAR 2ADON TRANSFORM IS THEN OF THE
FORM :
2F X KX T F lT X DT
5
WHERE FOR EACH X KX T IS A #ALDERs ON[:YGMUND KERNEL ON 2K DEPENDING
SMOOTHLY ON X WITH COMPACT SUPPORT IN -
7E NOW INDICATE A CLASS OF OPERATORS OF THE CLASS [ THAT
WOULD SEEM TO SUBSUME THE SINGULAR 2ADON TRANSFORM &OR THIS CON
SIDER THE 8J J K WHICH ARE COEnCIENTS OF THE LINEAR MONOMIALS
APPEARING IN #HOOSE ALSO 8* 8* 8*N_K SO THAT 8 8
8K 8* 8* 8*N_K IS A BASIS OF THE TANGENT SPACE OF - $ElNE l
ET S

&OR A GENERAL TREATMENT OF THESE OPERATORS SEE #HRIST .AGEL 3TEIN AND 7AINGER
;#HR.37=
6ISIONS IN -ATH '%/-%42)#!, #/.#%043 ). (!2-/.)# !.!,93)3
BY
q N_K
8 r
l
ET S X EXP SJ 8*J ` lT X
J
WHERE T S LIES IN A NEIGHBORHOOD OF THE ORIGIN IN 2K a 2N_K 7E THEN TAKE
KX T S FOR EACH X - TO BE A mAG KERNEL ON 2N RELATIVE TO THE mAG
FG y 2K y 2N 2K FT G
AND HOMOGENEITY m T S mT mT mTK m A S m AN_K SN_K (ERE
AJ IS THE WEIGHT OF 8*J WE ALSO SUPPOSE THAT X KX IS SMOOTH AND HAS
COMPACT SUPPORT 4HEN THE KIND OF OPERATOR WE WISH TO CONSIDER IS
:
4 F X KX T S F E l T S X DT DS
WHICH IS OF THE FORM WITH

+X Y KX T S * _ X Y Yl
ET S X
AND WHERE *X Y IS THE *ACOBIAN OF THE MAPPING T S Y l ET S X FOR
EACH lXED X 7E NOTE THAT IN THE SPECIAL CASE KX T S KX T mS WHERE m
IS THE $IRAC DELTA FUNCTION AT THE ORIGIN IN 2N_K OUR OPERATOR 4 BECOMES
ESSENTIALLY THE OPERATOR 2 IN
2EFERENCES
;"= ! "ELLA]pCHE 4HE TANGENT SPACE IN SUB 2IEMANNIAN GEOMETRY IN <3UB
2IEMANNIAN 'EOMETRY 0ROGRESS IN -ATH "IRKHAUSER [
;"R.7= * "RUNA ! .AGEL 3 7AINGER #ONVEX HYPERSURFACES AND &OURIER
TRANSFORMS !NN OF -ATH [
;#.3= $# #HANG ! .AGEL % 3TEIN %STIMATES FOR THE .EUMANN PROB
LEM IN PSEUDOCONVEX DOMAINS OF lNITE TYPE IN # !CTA -ATH
[
;#H= 7 #HOW 5BER SYSTEME VON LINEARER PARTEILLEN DIkERENTIALGLEICHUNGEN
ENRTER ORDNUNG -ATH !NN [
;#HR.37= - #HRIST ! .AGEL % 3TEIN 3 7AINGER 3INGULAR AND MAXI
MAL 2ADON TRANSFORMS ANALYSIS AND GEOMETRY !NN OF -ATH TO APPEAR
;#O7= 2 #OIFMAN ' 7EISS !NALYSE HARMONIQUE NON COMMUTATIVE SUR CER
TAINS sESPACES HOMOGENES 3PRINGER ,ECTURE .OTES IN -ATH
;$*3= ' $AVID *, *OURNs E 3 3EMMES /PsERATEURS DE #ALDERs ON[:YGMUND
FONCTIONS PARA ACCRsETIVES ET INTERPOLATION 2EV -ATH )BEROAMER
[
;&= # &EFFERMAN 4HE "ERGMAN KERNEL AND BIHOLOMORPHIC MAPPINGS OF
PSEUDO CONVEX DOMAINS )NVENT -ATH [
%- 34%). '!&!
;&0= # &EFFERMAN $ 0HONG 3UBELLIPTIC EIGENVALUE PROBLEMS IN <#ONFER

ENCE ON (ARMONIC !NALYSIS IN HONOR OF !NTONI :YGMUND "ECKNER ET
AL 7ADSWORTH [
;&3= # &EFFERMAN ! 3ANCHEZ[#ALLE &UNDAMENTAL SOLUTIONS OF SECOND
ORDER ELLIPTIC OPERATORS !NN OF -ATH [
;&3T= 2 &EFFERMAN % 3TEIN 3INGULAR INTEGRALS ON PRODUCT SPACES !DV IN
-ATH [
;&O3= ' &OLLAND % 3TEIN %STIMATES FOR THE B COMPLEX AND ANALYSIS ON THE
(EISENBERG GROUP #OMM 0URE AND !PPL -ATH [
;'3= 0 'REINER % 3TEIN %STIMATES FOR THE .EUMANN PROBLEM -ATH
.OTES 0RINCETON 5NIVERSITY 0RESS
;'R= - 'ROMOV #ARNOT[#ARATHEODORY SPACES SEEN FROM WITHIN IN <3UB
2IEMANNIAN 'EOMETRY 0ROGRESS IN -ATH "IRKHAUSER [

;(= , (O RMANDER (YPOELLIPTIC SECOND ORDER DIkERENTIAL EQUATIONS !CTA
-ATH [
;+3= ! +NAPP % 3TEIN )NTERWINING OPERATORS FOR SEMI SIMPLE GROUPS !NN
OF -ATH [
;+O6= ! +ORA s NYI 3 6A s GY 3INGULAR INTEGRALS IN HOMOGENEOUS SPACES AND
SOME PROBLEMS IN CLASSICAL ANALYSIS !NN 3CUOLA .ORM 3UP 0ISA
[
;-= - -ACHEDON %STIMATES FOR THE PARAMETRIX OF THE +OHN[,APLACIAN ON
CERTAIN DOMAINS )NV -ATH [
;-C= * -C.EAL %STIMATES ON THE "ERGMAN KERNELS OF CONVEX DOMAINS !DV
-ATH [
;-C3= * -C.EAL % 3TEIN 4HE 3ZEGO PROJECTION ON CONVEX DOMAINS -ATH
:EIT [
;-E= 9 -EYER /NDELETTES ET /PsERATEURS )) (ERMANN
;-U23= $ -U LLER & 2ICCI %- 3TEIN -ARCINKIEWICZ MULTIPLIERS AND
MULTI PARAMETER STRUCTURE ON (EISENBERG TYPE GROUPS ) )NV -ATH
[
;.3= ! .AGEL % 3TEIN IN PREPARATION
;.23= ! .AGEL & 2ICI % 3TEIN &LAG SINGULARITIES AND ANALYSIS ON CERTAIN
QUADRATIC #2 MANIFOLDS IN #. IN PREPARATION
;.37= ! .AGEL % 3TEIN 3 7AINGER "OUNDARY BEHAVIOR OF FUNCTIONS
HOLOMORPHIC IN DOMAINS OF lNITE TYPE .AT !CAD 3CI 53!
[
;.37= ! .AGEL % 3TEIN 3 7AINGER "ALLS AND METRICS DElNED BY VECTOR
lELDS !CTA -ATH [
;.A46= & .AZAROV 3 4REIL ! 6OLBERG 7EAK TYPE ESTIMATES AND #OTLAR
INEQUALITIES FOR #ALDERsON[:YGMUND OPERATORS IN NON HOMOGENEOUS SPACES
)NT -ATH 2ES .OTICES [
6ISIONS IN -ATH '%/-%42)#!, #/.#%043 ). (!2-/.)# !.!,93)3
;03= $ 0HONG % 3TEIN 3OME FURTHER CLASSES OF PSEUDODIkERENTIAL AND SIN

GULAR INTEGRAL OPERATORS ARISING IN BOUNDARY VALUE PROBLEMS !MER *OUR
-ATH [
;2= , 2OTHSCHILD .ONEXISTENCE OF OPTIMAL , ESTIMATES FOR BOUNDARY
,APLACIAN OPERATOR ON CERTAIN WEAKLY PSEUDOCONVEX DOMAINS #OMM 0ART
$Ik %Q [
;23= , 2OTHSCHILD % 3TEIN (YPOELLIPTIC DIkERENTIAL OPERATORS AND NILPO
TENT GROUPS !CTA -ATH [
;3= % 3TEIN (ARMONIC !NALYSIS 2EAL VARIABLE -ETHODS /RTHONGONALITY
AND /SCILLATORY )NTEGRALS 0RINCETON 5NIVERSITY 0RESS
;4= & 4REVES )NTRODUCTION TO 0SEUDO $IkERENTIAL AND &OURIER )NTEGRAL /P
ERATORS ) 0LENUM 0RESS
%LIAS - 3TEIN $EPARTMENT OF -ATHEMATICS 0RINCETON 5NIVERSITY 0RINCETON

.* 53!

(Modern Birkhäuser Classics) Noga Alon, Jean Bourgain, Alain Connes, Mikhael Gromov, Vitali D. Milman - Visions in Mathematics - GAFA 2000 Special Volume, Part I Pp. 1-453-Birkhäuser Basel (2010)

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

(Modern Birkhäuser Classics) Noga Alon, Jean Bourgain, Alain Connes, Mikhael Gromov, Vitali D. Milman - Visions in Mathematics - GAFA 2000 Special Volume, Part I Pp. 1-453-Birkhäuser Basel (2010)

Uploaded by

Copyright:

Available Formats

Modern Birkhäuser Classics

Many of the original research and survey monographs in pure and

Reprint of the 2000 Edition

Originally published under the title “Visions in Mathematics. Towards 2000”

2000 Mathematics Subject Classiﬁcation 35-02, 37-02, 00A05, 00A79, 01A67,

Library of Congress Control Number: 2010920640

Bibliographic information published by Die Deutsche Bibliothek

© 2010 Birkhäuser, Springer Basel AG

ISBN 978-3-0346-0421-5 e-ISBN 978-3-0346-0422-2

0ROGRAM OF THE #ONFERENCE

3 "LOCH AND ( %SNAULT

! "ERENSTEIN AND $ +AZHDAN

! "RAVERMAN AND $ +AZHDAN

%( ,IEB AND * 9NGVASON

E. H. Lieb Ya. G. Sinai

,EV !UDITORIUM 4EL !VIV 5NIVERSITY

  /PENING REMARKS

%RGODIC 4HEORY $YNAMIC 3YSTEMS #HAIR 3 -OZES

,EV !UDITORIUM 4EL !VIV 5NIVERSITY

-ATHEMATICAL 0HYSICS #HAIR ! *AkE

  0 3HOR  -ATHEMATICAL PROBLEMS IN QUANTUM INFORMATION

-ORIAH (OTEL $EAD 3EA

  ! *AkE -ATHEMATICS OF QUANTUM lELDS

  3 .OVIKOV  4OPOLOGICAL PHENOMENA IN REAL PHYSICS

  $ISCUSSION ON -ATHEMATICAL 0HYSICS WITH INTRODUCTION BY

  $ISCUSSION ON 'EOMETRY WITH INTRODUCTION BY - 'ROMOV

-ORIAH (OTEL $EAD 3EA

 $ISCUSSION ON -ATHEMATICS IN THE 2EAL 7ORLD IMAGE AP

,EV !UDITORIUM 4EL !VIV 5NIVERSITY

!NALYSIS #HAIR ! /LEVSKItp

  2 #OIFMAN #HALLENGES IN ANALYSIS 

.UMBER 4HEORY #HAIR : 2UDNICK

,EV !UDITORIUM 4EL !VIV 5NIVERSITY

$ISCRETE -ATHEMATICS #HAIR ! 7IGDERSON

  , ,OVsASZ  $ISCRETE AND CONTINUOUS TWO SIDES OF PHENOM

  4 'OWERS 2OUGH STRUCTURE AND CRUDE CLASSIlCATION

,EV !UDITORIUM 4EL !VIV 5NIVERSITY

!LGEBRA #HAIR 6 +AC

  3 "LOCH #HARACTERISTIC CLASSES FOR LINEAR DIkERENTIAL EQUA

!LGEBRA #HAIR $ +AZHDAN

  ! "EILINSON !ROUND GEOMETRIC ,ANGLANDS

,EV !UDITORIUM 4EL !VIV 5NIVERSITY

#HAIR ' +ALAI

,EV !UDITORIUM 4EL !VIV 5NIVERSITY

-ATHEMATICAL 0HYSICS AND 0$% #HAIR 6 :AKHAROV

'!5y[-!.). $%4%2-).!.4 #/..%#4)/.3 !.$

%VERYTHINGS SO AWFUL REGLAR A BODY CANT STAND IT

SATISlES THE DEGREE  EQUATION GIVEN BY THE DETERMINANT OF 'AUy[-ANIN

4HUS WE DO NOT LOSE INFORMATION BY TAKING THE BASE 3 TO BE THE SPEC

BE AN EkECTIVE DIVISOR SUPPORTED ON $ 3UPPOSE THAT THE RELATIVE CONNECTION

0ROPOSITION  $ElNE %$  % i /$  4HEN R83 INDUCES A FUNCTION

,ET J  8 _ $ 8 4HERE ARE NOW TWO RELATIVE DE 2HAM COMPLEXES

7E WILL CONSIDER ONLY THE CASE

SATISlES D! ! > !   FOR ALL M N  . BUT THE RESULTING INTEGRABLE

4HE RELATION BETWEEN T " % IS THE FOLLOWING %J" w

WITH ZI A LOCAL COORDINATE AND pI  F b j3  ,ET S BE A MEROMORPHIC SECTION

AND IT MAKES SENSE TO CONSIDER THE COMMUTATOR

WITH n  j+ i +8  -ULTIPLYING THROUGH BY ZIMI

&IX THE DIVISOR $ AND CONSIDER TUPLES F% R , uG WHERE % R IS AN

!S A CONSEQUENCE WE CAN RESTATE THE MAIN CONJECTURE

WHICH ARE OF THE FORM

 . ? j'+ i /' j'+   

WITH s  + )F WE SPECIALIZE 8 T WE lND

3 "LOCH AND ( %SNAULT

! "ERENSTEIN AND $ +AZHDAN

! "RAVERMAN AND $ +AZHDAN

%( ,IEB AND * 9NGVASON

/PENING REMARKS

%RGODIC 4HEORY $YNAMIC 3YSTEMS #HAIR 3 -OZES

-ATHEMATICAL 0HYSICS #HAIR ! *AkE

0 3HOR -ATHEMATICAL PROBLEMS IN QUANTUM INFORMATION

! *AkE -ATHEMATICS OF QUANTUM lELDS

3 .OVIKOV 4OPOLOGICAL PHENOMENA IN REAL PHYSICS

$ISCUSSION ON -ATHEMATICAL 0HYSICS WITH INTRODUCTION BY

$ISCUSSION ON 'EOMETRY WITH INTRODUCTION BY - 'ROMOV

$ISCUSSION ON -ATHEMATICS IN THE 2EAL 7ORLD IMAGE AP

!NALYSIS #HAIR ! /LEVSKItp

2 #OIFMAN #HALLENGES IN ANALYSIS

.UMBER 4HEORY #HAIR : 2UDNICK

$ISCRETE -ATHEMATICS #HAIR ! 7IGDERSON

, ,OVsASZ $ISCRETE AND CONTINUOUS TWO SIDES OF PHENOM

4 'OWERS 2OUGH STRUCTURE AND CRUDE CLASSIlCATION

!LGEBRA #HAIR 6 +AC

3 "LOCH #HARACTERISTIC CLASSES FOR LINEAR DIkERENTIAL EQUA

!LGEBRA #HAIR $ +AZHDAN

! "EILINSON !ROUND GEOMETRIC ,ANGLANDS

#HAIR ' +ALAI

-ATHEMATICAL 0HYSICS AND 0$% #HAIR 6 :AKHAROV

%VERYTHINGS SO AWFUL REGLAR A BODY CANT STAND IT

SATISlES THE DEGREE EQUATION GIVEN BY THE DETERMINANT OF 'AUy[-ANIN

BE AN EkECTIVE DIVISOR SUPPORTED ON $ 3UPPOSE THAT THE RELATIVE CONNECTION

0ROPOSITION $ElNE %$ % i /$ 4HEN R83 INDUCES A FUNCTION

,ET J 8 _ $ 8 4HERE ARE NOW TWO RELATIVE DE 2HAM COMPLEXES

SATISlES D! ! > ! FOR ALL M N . BUT THE RESULTING INTEGRABLE

4HE RELATION BETWEEN T " % IS THE FOLLOWING %J" w

WITH ZI A LOCAL COORDINATE AND pI F b j3 ,ET S BE A MEROMORPHIC SECTION

WITH n j+ i +8 -ULTIPLYING THROUGH BY ZIMI

&IX THE DIVISOR $ AND CONSIDER TUPLES F% R , uG WHERE % R IS AN

!S A CONSEQUENCE WE CAN RESTATE THE MAIN CONJECTURE

. ? j'+ i /' j'+

WITH s + )F WE SPECIALIZE 8 T WE lND

7E CONCLUDE FROM AND THAT H 3 c

4HE &OURIER 4RANSFORM

4HEOREM 4HE CONNECTION % R SATISlES #ONJECTURE

CONNECTION DIES ON ' AND WE GET

)T REMAINS TO EVALUATE THE CORRECTION TERMS RES 4RDGG _ ! OCCURRING IN

! BASIS FOR THE 'AUy[-ANIN BUNDLE IS GIVEN BY OMITTING EJ i Z M _ DZ !S

4O COMPUTE THE RIGHT HAND SIDE IN #ONJECTURE WE TAKE AS TRIVIALIZING

_ RES 4RDZ ss_ U_ DT _ _ DT

7E NOW COMPUTE 'AUy[-ANIN 7E WILL ABUSIVELY USE SHEAF NOTATION

, IS JUST THE RESTRICTION OF R TO THE DIVISOR OF 0 DElNED BY THE

7ITH RESPECT TO THE TRIVIALIZATION D LOGZ_ THE MATRIX G AT Z IS

4HIS CONCLUDES THE PROOF OF 4HEOREM c

SO WE TAKE THE OBVIOUS 1 STRUCTURE WITH BASIS EXPF 7E WANT TO COMPUTE

WHERE A RUNS THROUGH PERMUTATIONS OF F M _ G c

,ET PK Z ZM_ ZK ZM_

ZEROES OF DF F DZ )T FOLLOWS THAT SINCE B F 3YMM_ ! WE HAVE

THE WEIGHT M _ IS MAXIMAL TI TM__I WILL OCCUR WITH NONZERO COEnCIENT

SINCE bb WE MAY CONTINUE IN THIS FASHION WRITING TM_ TM_

4HE NONLINEAR CHANGE OF VARIABLES T T HAS JACOBIAN !LSO THE QUADRATIC

(ERE 3 3M_ 4 ARE HOMOGENEOUS COORDINATES ON 0M_ 5 6 ARE HO

DElNES A FUNCTIONAL (M_ 7 _ G _ $ 9 1 #

,EMMA (M_ 7 _ G _ $ 9 1 w

7E CALCULATE ( b 7 _ G _ P USING THE ,ERAY SPECTRAL SEQUENCE &OR

2EMARK )T FOLLOWS FROM 4HEOREM IN ;4= THAT

3PENCER "LOCH $EPT OF -ATH 5NIVERSITY OF #HICAGO #HICAGO ), 53!

02/",%-3 ). (!-),4/.)!. 0$%3

4WO (AMILTONIAN -ODELS

WHERE y t 4HE (AMILTONIAN FORMULATION OF IS

AND CANONICAL COORDINATES Y POSITION AND V SPEED

4HE #AUCHY 0ROBLEM

3EE ;#= OR ;3U3= FOR DETAILS

&ROM THE , CONSERVATION

WHERE m IS SOME NUMERICAL CONSTANT WE DO NOT EXPLAIN THE ROLE OF

WHERE 0 IS A POLYNOMIAL IN W W OF DEGREE s NO CONSTANT TERMS AND

3INGULARITY &ORMATION FOR .,3

,ET US CONSIDER THE CRITICAL NONLINEARITY P D 2ETURNING TO THE

)N FACT IT IS KNOWN THAT IN THIS CASE THERE HAS TO BE AN , CONCENTRATION

4HIS WOULD BE A STRENGTHENING OF INEQUALITY

MAPPING A SOLUTION U TO A SOLUTION #U )N THE FOCUSING CASE PERMITS

#LEARLY THESE MINIMUM , NORM BLOWUP SOLUTIONS ARE UNSTABLE (OWEVER

0ROBLEM 4HERE IS A lNITE CODIMENSIONAL MANIFOLD - IN A NEIGHBORHOOD

0ERIODIC "OUNDARY #ONDITIONS