Orthogona l Polynomial s on the Uni t Circl e Par t 1: Classica l Theor y This page intentionally left blank http://dx.doi.org/10.1090/coll054.1

America n Mathematica l Societ y Colloquiu m Publication s Volum e 54 , Par t 1

Orthogona l Polynomial s on the Uni t Circl e Par t 1: Classica l Theor y

Barr y Simo n

America n Mathematica l Societ y Providence , Rhod e Islan d Editorial Boar d Susan J . Priedlander , Chai r Yuri Mani n Peter Sarna k

2000 Mathematics Subject Classification. Primar y 42C05 , 05E35 , 34L99 ; Secondary 47B35 , 30C85 , 30D55 , 42A10 .

For additiona l informatio n an d update s o n thi s book , visi t www.ams.org/bookpages/coll-54

Library o f Congres s Cataloging-in-Publicatio n Dat a Simon, Barry , 1946 - Orthogonal polynomial s o n the uni t circl e / Barr y Simon . p. cm. — (American Mathematical Societ y colloquium publications, ISS N 0065-925 8 ; v. 54 ) Contents: pt . 1 . Classica l theor y Includes bibliographica l reference s an d index . ISBN 0-8218-3446- 0 (par t 1 : alk . paper)—ISB N 0-8218-3675- 7 (par t 2 : alk . paper ) 1. Orthogona l polynomials . I . Title . II . Colloquiu m publication s (America n Mathematica l Society) ; v. 54 . QA404.5.S45 200 4 515'.55—dc22 200404621 9

AMS softcover ISB N 978-0-8218-4863- 0 (par t 1) ; 978-0-8218-4864-7 (par t 2) .

Copying an d reprinting . Individua l reader s o f thi s publication , an d nonprofi t librarie s acting fo r them , ar e permitted t o mak e fai r us e o f the material , suc h a s to cop y a chapter fo r us e in teachin g o r research . Permissio n i s grante d t o quot e brie f passage s fro m thi s publicatio n i n reviews, provided th e customar y acknowledgmen t o f the sourc e i s given . Republication, systemati c copying , or multiple reproduction o f any material i n this publicatio n is permitte d onl y unde r licens e fro m th e America n Mathematica l Society . Request s fo r suc h permission shoul d be addressed to the Acquisitions Department, America n Mathematica l Society , 201 Charle s Street , Providence , Rhod e Islan d 02904-2294 , USA . Request s ca n als o b e mad e b y e-mail to reprint-permissionQams.org . © 200 5 by the America n Mathematica l Society . Al l rights reserved . Reprinted b y the America n Mathematica l Society , 2009 . The America n Mathematica l Societ y retain s al l right s except thos e grante d t o the Unite d State s Government . Printed i n the Unite d State s o f America . @ Th e pape r use d i n this boo k i s acid-fre e an d fall s withi n th e guideline s established t o ensur e permanenc e an d durability . Visit th e AM S hom e pag e a t http://www.ams.org / 10 9 8 7 6 5 4 3 2 1 1 4 1 3 1 2 1 1 1 0 0 9 To my grandchildren an d thei r parent s This page intentionally left blank Contents

Preface t o Part 1 x i Notation xvi i Chapter 1 Th e Basic s 1 1.1 Introductio n 1 1.2 Orthogona l Polynomial s o n the Rea l Lin e 1 1 1.3 Caratheodor y an d Schu r Function s 2 5 1.4 A n Introduction t o Operato r an d Spectra l Theor y 4 0 1.5 Verbhmsk y Coefficient s an d the Szeg o Recurrence 5 5 1.6 Example s o f OPUC 7 1 1.7 Zero s and the First Proo f o f Verblunsky's Theore m 9 0 Chapter 2 Szego' s Theorem 10 9 2.1 Toeplit z s an d Verbhmsk y Coefficient s 10 9 2.2 Extrema l Properties , the Christofie l Functions , an d th e Christoffel-Darboux Formul a 11 7 2.3 Entrop y Semicontinuit y an d the Firs t Proo f o f Szego' s Theorem 13 6 2.4 Th e Szeg o Function 14 3 2.5 Szego' s Theorem Usin g the Poisso n Kernel 15 1 2.6 Khrushchev' s Proo f o f Szego's Theorem 15 6 2.7 Consequence s o f Szego' s Theorem 15 9 2.8 A Higher-Order Szeg o Theorem 17 2 2.9 Th e Relativ e Szeg o Function 17 8 2.10 Totik' s Workshop 18 4 2.11 Ries z Products an d Khrushchev' s Worksho p 18 9 2.12 Th e Workshop o f Denisov an d Kupi n 19 7 2.13 -Value d Measure s 20 6 Chapter 3 Tool s fo r Geronimus ' Theore m 21 7 3.1 Verblunsky' s Viewpoint : Proof s o f Verblunsky's an d Geronimus' Theorem s 21 7 3.2 Secon d Kin d Polynomial s 22 2 3.3 K W Pair s 23 9 3.4 Coefficien t Strippin g an d Associate d Polynomial s 24 5

Chapter 4 Matri x Representation s 25 1 4.1 Th e GG T Representatio n 25 1 4.2 Th e CM V Representatio n 26 2 4.3 Spectra l Consequence s o f the CM V Representation 27 4 4.4 Th e Resolven t o f the CM V Matrix 28 7 viii CONTENT S

4.5 Ran k Tw o Perturbations an d Decouplin g o f CM V Matrices 29 3 Chapter 5 Baxter' s Theore m 30 1 5.1 Wiener-Hop f Factorizatio n an d th e Inverse s o f Finite Toeplit z Matrice s 30 1 5.2 Baxter' s Proo f 31 3 Chapter 6 Th e Stron g Szeg o Theorem 31 9 6.1 Th e Ibragimo v an d Golinskii-Ibragimo v Theorem s 31 9 6.2 Th e Borodin-Okounko v Formul a 33 3 6.3 Representation s o f U(n) an d th e Bump-Diaconi s Proo f 34 6 6.4 Toeplit z Determinant s a s the Statistica l Mechanic s o f Coulomb Gase s and Johansson' s Proo f 35 2 6.5 Th e Combinatoria l Approac h an d Kac' s Proo f 36 8 6.6 A Second Loo k at Ibragimov' s Theore m 37 6 Chapter 7 Verblunsk y Coefficient s Wit h Rapi d Deca y 38 1 7.1 Th e Rate o f Exponential Deca y an d a Theorem o f Nevai-Totik 38 1 7.2 Detaile d Asymptotic s o f the Verblunsky Coefficient s 38 7 Chapter 8 Th e Densit y o f Zeros 39 1 8.1 Th e Densit y o f Zero s Measure vi a Potential Theor y 39 1 8.2 Th e Densit y o f Zero s Measure via the CM V Matrix 40 3 8.3 Rotatio n Number s 41 0 8.4 A Gallery o f Zero s 41 2 Bibliography 42 5 Author Inde x 45 7 Subject Inde x 46 3 Preface t o Part 2 x i Notation xii i Chapter 9 Rakhmanov' s Theore m an d Relate d Issue s 46 7 9.1 Rakhmanov' s Theore m vi a Polynomial Ratio s 46 7 9.2 Khrushchev' s Proo f o f Rakhmanov's Theore m 47 5 9.3 Furthe r Aspect s o f Khrushchev's Theor y 48 5 9.4 Introductio n t o MNT Theor y 49 3 9.5 Rati o Asymptotic s 50 3 9.6 Poincare' s Theore m an d Ratio Asymptotics 51 2 9.7 Wea k Asymptotic Measure s 52 1 9.8 Rati o Asymptotic s fo r Varyin g Measure s 53 0 9.9 Rakhmanov' s Theore m o n an Ar c 53 5 9.10 Wea k Limit s an d Relativ e Szeg o Asymptotics 53 8 Chapter 1 0 Technique s o f Spectral Analysi s 54 5 10.1 Aronszajn-Donoghu e Theor y 54 5 10.2 Spectra l Averagin g an d the Simon-Wolf f Criterio n 55 1 10.3 Th e Gordon-de l Rio-Makarov-Simo n Theore m 55 8 10.4 Th e Grou p U(l,l) 56 4 CONTENTS i x

10.5 Lyapuno v Exponent s an d th e Growt h o f Norms i n U(l, 1) 58 1 10.5A Appendix: Subshift s 60 0 10.6 Furstenberg' s Theore m an d Rando m Matri x Product s Prom U(l, 1 ) 60 6 10.7 Th e Transfe r Matri x Approac h to L 1 Verblunsk y Coefficient s 61 7 10.8 Th e Jitomirskaya-Las t Inequalitie s 63 1 10.9 Criteri a fo r A.C . Spectrum 63 9 10.10 Dependenc e o n the Tai l 64 8 10.11 Kotan i Theor y 65 2 10.12 Priife r Variable s 66 4 10.13 Modifyin g th e Measure : Insertin g Eigenvalue s an d Rational Functio n Multiplicatio n 67 3 10.14 Deca y o f CMV Resolvents an d Eigenfunction s 68 5 10.15 Countin g Eigenvalue s i n Gaps: Th e Birman-Schwinge r Principl e 69 0 10.16 Stochasti c Verblunsk y Coefficient s 70 1 Chapter 1 1 Periodi c Verblunsk y Coefficient s 70 9 11.1 Th e Discriminan t 71 0 11.2 Floque t Theor y 71 9 11.3 Calculatio n o f the Weigh t 72 4 11.4 A n Overvie w o f the Invers e Spectra l Problem 73 0 11.5 Th e Orthogona l Polynomial s Associate d t o Dirichle t Dat a 74 2 11.6 Wal l Polynomials an d the Determinatio n o f Discriminants 74 8 11.7 Abel' s Theore m an d the Invers e Spectra l Proble m 75 3 11.8 Almos t Periodi c Tor i 78 3 11.9 Quadrati c Irrationalitie s 78 8 11.10 Independenc e o f Spectral Invariant s an d Isospectra l Tor i 79 9 11.11 Isospectra l Flow s 80 1 11.12 Bound s o n the Green' s Functio n 80 8 11.13 Genericit y Result s 81 1 11.14 Consequence s o f Many Close d Gap s 81 2 Chapter 1 2 Spectra l Analysi s o f Specifi c Classe s of Verblunsky Coefficient s 81 7 12.1 Perturbation s o f Bounded Variatio n 81 7 12.2 Perturbation s o f Periodic Verblunsk y Coefficient s 82 6 12.3 Naboko' s Workshop : Dens e Point Spectru m i n the Szeg o Clas s 82 9 12.4 Generi c Singula r Continuou s Spectru m 83 4 12.5 Spars e Verblunsk y Coefficient s 83 8 12.6 Rando m Verblunsk y Coefficient s 84 5 12.7 Decayin g Random Verblunsk y Coefficient s 84 7 12.8 Subshift s 85 5 12.9 Hig h Barrier s 86 3 Chapter 1 3 Th e Connectio n to Jacob i Matrices 87 1 13.1 Th e Szeg o Mapping an d Geronimu s Relation s 87 1 13.2 CM V Matrices an d th e Geronimu s Relation s 88 1 13.3 Szego' s Theorem fo r OPRL : A First Loo k 88 9 13.4 Th e Denisov-Rakhmano v Theore m 89 2 13.5 Th e Damanik-Killi p Theore m 89 6 x CONTENT S

13.6 Th e Geronimo-Cas e Equation s 90 3 13.7 Jacob i Matrice s With Exponentiall y Decayin g Coefficient s 91 2 13.8 Th e P 2 Su m Rul e and Application s 92 0 13.9 Szego' s Theorem fo r OPRL : A Third Loo k 93 7 Appendix A Reader' s Guide : Topic s and Formula e 94 5 A.l What' s Don e Where 94 5 A. Schu r function s 94 5 B. Toeplit z matrice s an d determinant s 94 5 C. Szego' s theorem 94 6 D. Aleksandrov familie s 94 6 E. Zero s o f OPUC 94 6 F. Densit y o f zeros 94 6 G. CM V matrices 94 7 H. Periodic Verblunsky coefficient s 94 7 I. Stochasti c Verblunsk y coefficient s 94 7 J. Transfe r matrice s 94 8 K. Asymptotic s o f orthogonal polynomial s 94 8 A.2 Formula e 94 8 A. Basic objects 94 8 B. Recursio n 95 0 C. Bernstein-Szego approximatio n 95 5 D. Additional O P formula e 95 6 E. Additiona l Wal l polynomial formula e 95 6 F. Matri x representations 95 6 G. Aleksandro v familie s 95 9 H. Rotation o f measure 96 0 I. Sieve d polynomials 96 0 J. Toeplit z (se e also K) 96 1 K. Szego' s theory 96 1 L. Additional transfe r matri x formula e 96 4 M. Periodic Verblunsk y coefficient s 96 6 N. Connectio n t o Jacob i matrice s 96 7 Appendix B Perspective s 97 1 B.l OPR L vs . OPUC 97 1 B.2 OPU C Analog s o f the ra-function 97 3 Appendix C Twelv e Grea t Paper s 97 5 Appendix D Conjecture s an d Ope n Question s 98 1 D.l Relate d t o Extending Szego' s Theorem 98 1 D.2 Relate d to Periodi c Verblunsk y Coefficient s 98 1 D.3 Spectra l Theor y Conjecture s 98 2 Bibliography 98 3 Author Inde x 103 1 Subject Inde x 1039 Preface t o Par t 1

For an overvie w o f the subject o f these volumes , se e Section 1.1 .

Here i s how this boo k cam e to be . I'v e worke d i n part o n the spectra l theor y of Schrodinge r operator s fo r al l o f m y career . I n th e lat e 1970s , i t becam e clear , especially fo r those o f us trying to understand localizatio n i n random systems, tha t it wa s technicall y easie r t o stud y th e discret e Schrodinge r operator , whic h i n on e dimension read s

(hu)n = izn+i + w n_i + v nun (1 ) In abou t 1985 , I me t Moura d Ismai l wh o urge d m e t o loo k a t th e literatur e o n orthogonal polynomial s — and I should hav e paid mor e attention to hi s advice. At leas t I did loo k a little a t orthogona l polynomial s o n the rea l lin e (OPRL ) and learne d that theor y wa s essentially the stud y o f operators o f the for m

{Ju)n = a n+iwn+i + bn+iUn + a nun-i (2 ) where n = 0,1,2,... an d (oi , c&2,...) and (bi , &2 » • • •) are real with a n > 0 for n > 1 and a o = 0 . Th e discret e Schrodinge r cas e i s a n = 1 . One thin g I realize d earl y i s that th e Jacob i matri x (2 ) i s the natura l aren a for invers e theor y an d tha t i t seeme d difficul t t o clarif y wha t spectra l measure s corresponded to an = 1 with bn arbitrary. A s inverse problems became an important component o f my research, I worked o n what wa s essentially OPRL , eve n goin g s o far as writing a long review article [974 ] on moment problems that coul d be thought of as a primer on OPRL. But I didn't systematicall y loo k at literature fro m the O P community. My firs t rea l exposur e t o orthogona l polynomial s o n th e uni t circl e (OPUC ) came i n connectio n wit h m y rol e a s a n edito r o f Communication s i n Mathemati - cal Physics (CMP) , i n particular, wit h a submission o f Golinski i an d Neva i [464] . My sectio n i n CM P ha d accepte d a n earlie r pape r o n th e subjec t b y Geronim o and Johnso n [398] , but a t the time I hadn't pai d clos e attention to it. I personall y knew both those authors and the paper seemed to be studying some kind o f general difference equation , s o I sen t i t t o a n appropriat e refere e an d followe d th e recom - mendation. Th e Golinskii-Neva i pape r wa s different . I knew o f Nevai's reputatio n but didn't know either author. Th e paper was clearly about orthogonal polynomials and I wa s reluctant, give n CMP' s perpetua l pag e crunch , t o ope n u p th e journa l to a subjec t tha t I worrie d migh t b e disconnecte d t o eithe r physic s o r th e area s that w e normally cover . S o I spen t som e tim e carefull y readin g th e introduction , thinking abou t th e issues , and skimmin g the rest. I realized this paper wa s one o n spectral theor y relate d t o man y paper s i n CMP , s o I wa s comfortabl e sendin g i t out t o a referee an d followin g a positive recommendation .

xi xii PREFACE T O PAR T 1

The next aspects o f the story involve my own interests and those o f my research group. A major them e o f our wor k in the 1990 s concerned the spectral behavior o f discrete Schrodinge r operator s (1 ) an d th e continuum analo g i n one dimension : (Hu)(x) = -u"(x) + V(x)u(x) (3 ) It ha s bee n know n sinc e the 1930 s that V G L1 o r \V(x)\ 1 (equivalently, a n = 1 and b n el o r \b n\ < C(l + \n\)~ fo r the discret e case) i s a natural dividin g line. I f V G L1, the H give n by (3 ) has purely absolutel y continuous spectru m a t positiv e energie s an d scatterin g states . In m y ow n work, i t becam e clea r a = | wa s als o a natural dividin g line . Thi s appeared firs t i n wor k I did o n decaying rando m potential s [965] ; there wa s dens e point spectrum i f a < \ an d subsequent wor k [265, 268, 649] showed a.c. spectrum if a > \. I saw it agai n when I proved that Bair e generic potentials decayin g slowe r than n-1/2 hav e purely singula r continuou s spectru m [969] . I the n emphasize d tha t th e cas e \ > a > \ wa s open , an d Kiselev , i n hi s thesis don e unde r m e [637] , an d the n Christ-Kiselev-Remlin g [201 ] filled i n th e region showin g there wa s alway s a.c . spectrum i n this regime . Kiselev-Last-Simo n [639], motivate d b y thi s work , conjecture d tha t th e righ t borderlin e conditio n fo r a.c. spectru m wa s L 2. Rowa n Killip , a graduat e studen t I wa s supervisin g a t the time , go t intereste d i n thi s questio n an d h e aske d Perc y Deif t abou t i t whil e Percy wa s visiting . Perc y notice d tha t th e Kd V su m rul e involve d L 2 precisely , suggesting the sum rule could be relevant. The y then implemente d thi s idea [248] , not b y provin g the su m rul e i n this generalit y bu t usin g the su m rul e fo r nic e V' s and takin g suitable limits . At th e sam e time , I wa s encouragin g worker s i n th e field t o find example s o f mixed spectrum i n this 1 > a > \ region . I t wa s known one could have point spec - trum mixe d i n the a.c . spectrum [795 , 973], but th e issu e was singular continuou s spectrum an d what kind . I n honor o f the change o f millennium, I produced a list o f open problems in 199 9 and high in the list was mixed singular continuous spectru m with powe r decay . In January 2001,1 learned o f a paper by a young Russian, Serge y Denisov [271] . In it , h e showe d ther e existe d L 2 potential s V wit h essentiall y arbitrar y singula r part o n [0 , EQ] fo r EQ < oo. Technically , this didn't solv e the problem I stated sinc e I aske d fo r a power bound (thi s was later don e b y Kiselev [638]) , but I regarded i t as essentially a solution. Nic k Makarov an d I were so impressed b y this result tha t we arranged a postdoc positio n fo r Denisov , o f which mor e later . That summer , Rowa n Killi p visite d Caltec h an d w e tried t o understan d wha t made Denisov' s proo f work . W e weren't abl e to understand th e technical details o f his proo f becaus e h e use d Krei n systems , whic h w e had neve r see n an d fo r whic h the existing literature wa s spotty. Bu t w e understood that h e was using some kin d of invers e theor y construction . Initially , thi s puzzle d me . Fo r years , I' d bee n tol d that on e shoul d us e invers e spectra l theor y t o construc t mixe d spectra , an d m y response wa s to poin t ou t tha t th e Marchenk o theor y wouldn' t wor k i n such case s and tha t th e Gel'fand-Levita n theor y provide d n o informatio n abou t asymptotic s at infinity . Nevertheless, Deniso v prove d hi s potentia l wa s i n L 2. A t first, w e though t the magi c mus t b e i n Krei n systems , bu t the n w e realize d h e wa s gettin g th e L 2 PREFACE T O PAR T 1 Xlll property fro m a su m rule . S o w e focuse d o n usin g su m rule s goin g beyon d thei r use in Deift-Killip. I n connection with the earlie r wor k and it s followup, Killi p ha d learned o f the su m rule s o f Cas e [190 ] an d variou s revie w article s o f Nevai [814] . With these as background, w e were able to handle discrete eigenvalues in the Jacob i case an d prov e a general su m rul e that describe d exactl y whic h spectra l measure s 2 corresponded to Jacobi matrices with ^2 n{an — l) + 6 ^ < o o [633] . Alon g the way, I learne d tha t du e to wor k o f Szeg o an d Shohat , thi s proble m ha d bee n solve d i n the earl y part o f the twentiet h centur y unde r th e additiona l stron g conditio n tha t supp(d/i) C [—2,2] . Bu t I put asid e understanding ho w they ha d don e s o while w e worked ou t som e o f the detail s o f our work . In the fal l o f 2001, Denisov showed up at Caltec h and, with my encouragement , gave a course on Krein systems. Sinc e OPUC ar e a discrete analo g that motivate d Krein, Deniso v spen t th e firs t fe w week s presenting thei r theory , followin g i n par t the chapte r i n Akhiezer' s boo k [17 ] ( I had, o f course , rea d thi s boo k whe n I wa s writing m y articl e o n th e momen t proble m fo r R bu t I ha d skippe d thi s chapte r as irrelevan t t o me!) . I wa s struck b y the beaut y o f the subjec t an d bega n t o tr y to understan d i t better . I quickl y foun d th e proo f I giv e o f the Szeg o recursion , wondering wh y Akhiezer' s proo f wa s s o complex. I eventuall y foun d thi s proo f i n Atkinson [60 ] an d Landa u [672] , bu t i t wa s unknow n enoug h tha t I wa s abl e t o surprise on e expert wit h it . At the same time, in connection with my work with Killip, I realized what we' d found wa s an analog o f Szego's theorem, which I began to study. I also realized tha t OPUC wa s a n analo g o f Schrodinge r operators , a virtua l playgroun d fo r spectra l theorists. B y February o f 2002, I was convinced that th e sensibl e way to carry ove r all the Schrodinge r operator s t o OPU C wa s i n a singl e article , rathe r tha n lot s o f short ones . I als o realize d thoug h tha t OPU C ha d n o summary sinc e Geronimus ' book forty year s before, an d eve n that ha d focuse d o n only part o f the then know n subject. Th e ide a bega n t o for m o f writin g tw o revie w article s o n OPUC , eac h I estimated to be 120-15 0 pages long, the firs t reviewin g the classica l theory an d th e second o n . As I began planning the articles, I began to wonder who had first writte n abou t Szego's theorem a s a sum rule. I asked Paul Nevai , who replied that h e didn't kno w — but i t wa s a n interestin g questio n an d s o he sen t ou t a n e-mai l blas t t o abou t a doze n experts . On e fro m th e Russia n traditio n replie d tha t h e wasn' t sur e an d that h e hadn't see n the paper , bu t he' d hear d i t migh t b e " ..." an d h e gav e th e reference t o Verblunsk y [1067] . I got hol d o f this an d th e earlie r [1066 ] an d rea d them with fascination . I t wa s hard going , but a s I absorbed th e papers, i t becam e clear that ther e was an enormous number o f ideas in these papers that ha d becom e important, bu t the n forgotte n an d late r rediscovered ! S o added to the agend a wa s making sur e Verblunsky go t the credi t s o long denied him ! I began a n e-mai l correspondenc e wit h Golinski i an d Neva i about thei r paper , and i n discussin g th e GG T representation , Golinski i tol d m e abou t th e pape r o f Cantero, Moral , an d Velazque z [181 ] written i n 200 0 but onl y publishe d i n 2003 . He told m e the basi c ide a that orthonormalizin g {l,z,z _1,2:2,2:_2,...} lead s to a expressible in terms o f the usual OPUC and a five- expressibl e in term s o f th e Verblunsk y coefficients . Withi n a fe w hours , I ha d rederive d th e formulae an d then ove r a period o f a fe w weeks, Golinskii and I , via e-mail, worke d out th e detail s fo r genera l spectra l theor y tha t appea r a s Sectio n 4.3 . M y projec t xiv PREFAC E T O PAR T 1 had becom e large r a s I realized that presentin g the CM V matrix an d exploitin g i t was also a part o f the mix . As the summer o f 2002 finished, I realized I still had a lot to do and the idea of two 150-page articles wouldn't suffice . A book was needed, and when Sergei Gelfan d offered t o publis h i t i n th e sam e serie s wher e Szego' s boo k [1026 ] appeared , th e temptation t o accep t withou t consultin g other publisher s wa s s o great. Slowly the project too k on extra elements that caused it to grow. I had include d background section s on Schu r functions an d o n OPRL to make the book accessibl e to spectral theorists, but Doron Lubinsky convinced me to add a background section on operator theor y to make it accessibl e to workers with a background i n OP. Th e desire to reac h both group s produce d backgroun d materia l i n man y place s that, I hope, als o makes it accessibl e to graduate students . Other factor s cause d manuscrip t growth . I realized that ther e were results lik e the Bello-Lope z extensio n o f Rakhmanov' s theore m tha t I neede d t o includ e i f I wanted th e boo k t o b e a s comprehensiv e a s I desired . Perc y Deif t kep t pushin g me to include al l the main proof s o f the strong Szeg o theorem. Onc e the book ha d grown a s i t had , I foun d i t appropriat e t o ad d a chapte r o n use s o f OPUC i n th e theory o f OPRL . In this way , two 150-pag e articles gre w to on e book gre w to tw o volumes. The table o f contents and list o f notation cover both volumes. T o save trees, the author index , subjec t index , an d th e bibliograph y i n this volum e ar e onl y fo r thi s volume, althoug h ther e i s a complete bibliograph y an d complet e indice s i n Part 2 . For consistent numberin g betwee n volumes, the bibliograph y i n Part 1 has gaps i n its numbering . I considere d mor e elegan t referenc e referra l wit h autho r an d yea r but decide d agains t it . Whe n ther e ar e fou r o f fiv e "S " references , i t i s eas y t o locate [ST97] , but whe n ther e ar e ove r a hundred "S " references, i t i s difficult. I n the end, functionalit y wo n ove r elegance . I warn the reade r o f a personal quirk . I' m tol d that prope r usag e requires th e addition o f a period in a sentence that ends with a set-out equation. Bu t I find extr a dots i n suc h equation s confusing , s o I don' t us e punctuatio n i n set-ou t formulae , even i f proper gramma r say s they shoul d b e there. I doub t tha t thes e book s wil l hav e th e fou r edition s tha t Szeg o [1026 ] did , but i t seem s likel y ther e wil l b e late r editions . Tha t mean s I especiall y wel - come comments , corrections , missin g topic s an d references , an d informatio n o n new papers . Fo r th e latter , I muc h prefe r a lin k t o a n onlin e archiv e rathe r tha n that yo u sen d m e attachments . Addend a an d majo r correction s will be poste d a t http: //www. math. caltech. edu/opuc. html. M y email i s bsimonOcaltech. edu.

One perso n wh o learne d abou t thi s boo k wondere d ho w I coul d eve n Te X s o many page s in "onl y two years." Th e secret i s that I didn't. I wrote out lon g han d and Cheri e Galvez , m y super b secretary , TeXe d th e manuscript . I a m indebte d for he r har d wor k an d competence . Rowa n Killi p provide d advic e t o he r o n Te X technical issues , and I'm gratefu l fo r hi s help on this. Many mathematician s provide d usefu l input , bu t I should begi n with fou r de - serving specia l mention . Rowa n Killi p provide d insigh t o n man y o f th e issue s I studied here . Perc y Deif t wa s merciles s i n pushin g m e t o includ e materia l tha t worked out wel l and has added to the usefulness o f the book. Frit z Gesztesy taugh t me muc h abou t periodi c Jacob i matrices . Mos t o f all , Leoni d Golinski i spen t a PREFACE T O PAR T 1 x v huge amount o f time goin g through the manuscript. H e not onl y foun d typos , bu t places where the argument s weren' t quit e right . I also want to thank Richard Askey , Alexei Borodin, Albrech t Bottcher , Danie l Bump, Alicia Cachafeiro, Danny Calegari, Tiberiu Constantinescu, David Damanik, Sergey Denisov, Nathan Dunfleld, Harr y Dym, Jef f Geronimo , Dimitri Gioev, Peter Harremoes, Mourad Ismail, Alain Joye, Thomas Kailath, Victor Katsnelson, Serge y Khrushchev, Pete r Kuchment , Ar i Laptev , Yora m Last , Danie l Lenz , Guillerm o Lopez Lagomasino , Doro n Lubinsky , Russel l Lyons , Alphons e Magnus , Francisc o Marcellan, Andre i Martinez-Finkelshtein , Pete r Miller , Irin a Nenciu , Pau l Nevai , Olav Njastad , Fran z Peherstorfer , Yur i Safarov , Miha i Stoiciu , Gunte r Stolz , Alexander Teplyaev , Joh n Toland , Vilmo s Totik , Doro n Zeilberger , an d Andre j Zlatos fo r usefu l input . Finally, m y wif e Martha, whos e lov e makes it al l easier .

Barry Simo n Los Angeles, August 200 4 This page intentionally left blank Notation

This lis t o f notation i s broken int o the Gree k an d Roma n alphabet s an d the n nonalphabetic. Sinc e ther e ar e onl y a finite numbe r o f letters, som e symbol s ar e used i n differen t ways !

Greek Alphabe t i n the followin g order : a, /?, 7, T, > ^> ^

an Verblunsk y coefficients ; se e (1.1.8 ) Pf solutio n o f 0 + /T1 = £* wit h |/3 | > 1 ; see (13.8.21 ) (3(zo) Jitomirskaya-Las t ratio ; se e (10.10.13 ) d(32n,k(@) measur e i n varying ratio asymptotics ; se e (9.8.7 ) 7(2) Lyapuno v exponent , 7(2: ) = lim n_+oo \ lo g ||Tn(z)||; se e (10.5.14 ) ra?A essentia l spectru m fo r singl e arc , T a,\ — {z G <9 B | |arg(A^) | > 2arcsin(a)}; se e (9.9.2 ) re(x) functio n o f Damanik-Killip , r e(a;) = §[u(# ) — v(x)]; se e th e Note s t o Section 13. 1 m T0(x) functio n o f Damanik-Killip , V 0(x) = —^[u(x) + v(x)] , see th e Note s t o Section 13. 1 r(Bj) contou r surroundin g BJ; see (11.7.16 ) T(Gj) contou r surroundin g Gj\ se e (11.7.16 ) T{z) analyti c functio n whos e rea l par t i s th e Lyapuno v exponent , 7(2) ; se e (10.11.18) 50D relativ e Szeg o function; se e (2.9.6 ) A(z) Freu d function ; se e Theorem 2.2.1 4 A (2) discriminan t fo r periodi c Verblunsky coefficients ; se e (11.1.2 ) Qn poin t mas s approximations ; se e Theorem 2.2.1 2 dr)n(z) varyin g measure; se e (9.8.1 ) 9n(zo) Priife r angl e fo r OPUC ; se e (10.12.1 ) Qj CM V building block , 9, = (% _ % ); se e (4.2.20 ) n 1 nn leadin g coefficien t o f (p n(z) = n nz -f - • • •, K n = H^H"" ; se e (1.5.1 ) ^oo(C) Christoffe l function , Aoo(£ ) = inf{/|7r(e z6>)|2 d//(0) | 7 r a polynomial , TT(C) - 1} ; see (2.2.2 ) Xj(A) eigenvalue s o f a compact operato r A; see (1.4.27 ) 26> 2 ^n(C) approximat e Christoffe l function , X n(C) — min{/|7r(e )| d/x(0) | deg7r < n, TT(C ) = 1} ; see (2.2.1 ) 20 p An(£, d//;p) ^Christoffe l function , \ n((,d/j,',p) — min{/|7r(e )| d/x | deg7 r < n , TT(C) = 1} ; see (2.5.1 ) f^x\z) Schu r functio n o f an Aleksandro v family ; se e Subsection 1.3. 9 xviii NOTATIO N

F(x\z) Caratheodor y functio n o f an Aleksandrov family ; see (1.3.90 ) <3>^ OP s o f an Aleksandro v family ; se e (3.2.1 ) dp\ measure s o f an Aleksandro v family ; se e (1.3.91 ) A unitar y perturbation , A : PH - > PH, V = UAP + U(l - P) ; se e (4.5.10 ) pn(A) singula r value s o f an operator A, i.e., eigenvalue s o f \A\; see (1.4.33 )

d/iac absolutel y continuous part o f the measure d\i, i.e. , i f dp, = w^ - f d/xs, then

dpac = w^; se e Subsection 1.4. 6 see d/ipp pur e poin t par t o f a measure, dp pp = J2 X Mi^})^; Subsectio n 1.4. 6 dps(9) singula r par t o f a measure, dp s — dp — dpac; see Subsection 1.4. 6 dpsc singula r continuou s par t o f a measure, dp sc = dp s — dp pp; se e Subsec - tion 1.4. 6 dpfftf) G I approximations t o a measure; se e (6.1.24 ) v{n) Beurlin g weight ; se e (5.1.38 ) dv densit y o f zeros measure, dv = limdv n i f it exists ; se e Section 8. 1 dvn finite densit y o f zeros, dvn = ^ Yll=i ^ wher e Zj ar e the zero s o f $n(z) counting multiplicity ; se e Example 1.7.1 7

dvnj{6) approximation s use d i n Khrushchev's theory , dv n^(6) = . ( eie\\2 |f; see (9.3.14 )

En(z) vecto r recursio n fo r OPUC ; se e (2.2.44 ) 7r* extende d abelia n periods ; se e (11.8.2 ) 7rn decayin g CM V function, 7r n = Tn + F{z)xn\ se e (4.4.5 ) Hk(Bj) k-th periodi c associate d t o band, BJ; se e (11.7.16 ) Uk(Gj) fc-th periodi c associate d t o gap, Gj\ se e (11.7.16 ) 2 1 2 Pi p j = (l-\aj\ ) / ; se e (1.5.21 ) p(B) resolven t se t o f an operator B, p(B) = {z \ (B — z) 1 exists}; se e Subsec- tion 1.4. 2 p(Tn) Kotani-Ushiroy a ratio ; see (10.5.83 ) (Td(A) discret e spectru m o f an operator A; se e Subsection 1.4. 6 cress(^) essentia l spectru m o f a n operato r A, aess(A) = cr(A)\ad(A)] se e Subsec - tion 1.4. 6 n{ se e an hal f o f CM V basis, an = \2n — z~ P\ni (4.2.8) cr(B) spectru m o f an operator B, ^ lef t polynomial s i n the theory o f matrix-valued measures ; se e (2.13.17 ) n $* (2) dua l o f monic OPUC, z $n(l/z); se e (1.5.9 ) Xa{g) grou p character ; se e Section 6. 3 Xn CM V basi s element; se e (4.2.5 ) Xn fre e CM V basis element; see (4.2.4 ) NOTATION xi x

#n secon d kin d OPUC ; se e (3.2.3 ) Q± wav e operators; se e (10.7.46 ) an d (10.7.80 )

Roman Alphabe t an Jacob i parameter ; se e (1.2.13) an(x) continue d fractio n coefficients ; se e (11.9.1 ) ab(z) normalize d coefficien t i n quadratic equatio n fo r F\ se e (11.7.66) A suppor t o f the a.c. part; se e (10.10.4) A se t where j(E) = 0 for subshifts; se e Section 12.8 An Wal l polynomial ; se e (1.3.65) Avn(f) ergodi c average ; se e Theorem 10.5A. 2 A alphabe t fo r subshift; se e Appendix 10.5 l AR annulus , {z | R~ < \z\ < i?}; see (7.1.1) 21 Abe l map ; se e (11.7.26 ) 21 x extende d Abe l map ; se e (11.8.5) 2l„ two-side d Beurlin g algebra , {a | J^n(l + |n|)^|a n| < oo} ; see (5.1.44) 21+ positiv e Beurlin g algebra , {a G 21^ | a(n) = 0 for n < 0}; see (5.1.44) 2l~ negativ e Beurlin g algebra , {a £ 21 ^ | a(n) = 0 for n > 0}; see (5.1.44) bn Jacob i parameter ; se e (1.2.13) bni£ O P ratio error ; se e (9.1.1) 6n(z, d/i) invers e Schu r iterates ; se e (9.2.14) bb(z) normalize d coefficien t i n quadratic equatio n fo r F; see (11.7.67) B se t of z's in <9IE D for which lim r|i F(rz) doe s not exist; se e (10.10.5) B se t fo r whic h limsup:r n(2:) < o o in theory o f subshifts; se e (12.8.3 ) Bj band s fo r periodi c Verblunsk y coefficients , se e (11.1.6 ) Bj unio n o f touching bands; se e Section 11. 7 Bn Wal l polynomials; see (1.3.66 ) B(f) rati o D/D; se e (6.2.40 ) B(fji,v) bilinea r potential ; se e (8.1.9) B facto r i n CMV proof o f Geronimus relations ; se e (13.2.16 ) %nQ cn moment s o f a measure, J e~ d\i{&)\ se e (1.1.20 ) cn O P half o f solution o f the Geronimo-Case equations ; se e (13.6.10 ) cn>£ O P ratio; se e (9.1.13 ) cb(z) normalize d coefficien t i n quadratic equatio n fo r F; se e (11.7.68) cap(if) logarithmi c capacit y o f a compact se t K\ se e Section 8. 1 C unio n o f singula r continuou s support s ove r a n Aleksandro v family ; see (10.10.3) Cn normalize d solutio n o f Geronimo-Case equations ; se e (13.6.42 ) Cf0 th e contour s \z\ = 1 on the Rieman n surfac e