This action might not be possible to undo. Are you sure you want to continue?
PROBLEMS AND SOLUTIONS I N
MATHEMATICS
Major American Universities Ph.D. Qualifying Questions and Solutions
PROBLEMS AND SOLUTIONS IN
MTHEMTICS
Compiled by:
Chen JiXiu, Jiang GuoYing, Pan YangLian, Qin XeHu, Tong YuSun, Wu QuanShui and Xu ShengZhi
Edited by:
LI TATSIEN
Fudm University
orld Scientific
NewJersey London Singapore Hong Kong
Published by
World Scientific Publishing Co. Pte. Ltd.
5 Toh Tuck Link, Singapore 596224 USA ofice: Suite 202, 1060 Main Street, River Edge, NJ 07661 UX:office: 57 Shelton Street, Covent Garden, London WC2H 9HE
Library of Congress CataloginginPublication Data Problems and solutions in mathematics I [edited by] Li Ta Tsien. p. cm. Includes bibliographical references. ISBN 9810234791  ISBN 9810234805 (pbk) 1. Mathematics  Problems, exercises, etc. I. Li, Tachen. QA43.P754 1998 510,76dc21 9822020 CIP
British Library CataloguinginPublication Data A catalogue record for this book is available from the British Library.
First published 1998 Reprinted 2002,2003
Copyright 0 1998 by World Scientific Publishing Co. Pte. Ltd. All rights reserved. This book, or parts thereof; may not be reproduced in any form or by any means, electronic or mechanical, includingphotocopying, recording or any information storage and retrieval system now known or to be invented, without written permission from the Publisher.
For photocopying of material in this volume, please pay a copying fee through the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, USA. In this case permission to photocopy is not required from the publisher.
This book is printed on acidfree paper.
Printed in Singapore by UtoPrint
V
PREFACE
This book covers six aspects of graduate school mathematics: Algebra, Topology, Differential Geometry, Real Analysis, Complex Analysis and Partial Differential Equations. It contains a selection of more than 500 problems and solutions based on the Ph.D. qualifying test papers of a decade of influential universities in North America. The mathematical problems under discussion are kept within the scope of the textbooks for graduate students. Finding solutions to these problems, however, involves a deep understanding of mathematical principles as well as an acquisition of skills in analysis and computation. As a supplement t o textbooks, this book may prove to be of some help to the students in taking relevant courses. It may also serve as a reference book for the teachers concerned. It has to be pointed out that this book should not be regarded as an allpurpose troubleshooter. Nor is it advisable to take the book as an exemplary text and commit to memory all the problems and solutions and make an indiscriminate use of them. Instead, the students are expected to make a selective survey of the problems, take a doityourself approach and arrive at their own solutions which they may check against those listed in the book. It would be gratifying to see that the students can work out the problems on their own and come up with better solutions than those provided by the book. If the students fail to do so or their solutions may turn out to be incomplete, it may reveal the inadequacy of their knowledge or approach, thus spurring them to greater efforts to promote their skills. The very purpose of the authors in writing the book is just to help the students to discover the truth by trial and error. This book was inspired by Professor K. K. Phua’s proposals. We are particularly grateful to him for his support. We also wish to thank Dr. Xu Peijun, Professors Zhang Yinnan, Hong JiaXing and Chen Xiaoman for their painstaking efforts to collect testoriented data. For selecting problems and providing solutions, we wish to acknowledge the following professors respectively: Wu Quanshui (Part I), Pan Yanglian (Part II), Jiang Guoying (Part
vi
111), Tong Yusun, Xu Shengzhi (Part IV), Chen Jixiu (Part V) and &in Tiehu (Part VI). We are also indebted to Professor Guo Yutao for carefully reading and correcting the manuscript. Finally, we pay tribute t o Dr. Cai Zhijie for printing out the manuscript.
Li Tatsien Department of Mathematics Fudan University Shanghai 200433 China
vii
CONTENTS
Preface ................................................................ (v) Part I. Algebra ......................................................... (1) 1. Linear Algebra ................................................... (3) 2. Group Theory ................................................... (26)
3. Ring Theory ..................................................... (44) 4. Field and Galois Theory ......................................... (59) Part 11. Topology ...................................................... (81) 1. Point Set Topology .............................................. (83) 2. Homotopy Theory ............................................... (99) (118) 3. Homology Theory .............................................. Part 111. Differential Geometry ........................................ (151) 1. Differential Geometry of Curves ................................ (153) 2. Differential Geometry of Surfaces ............................... (171) 3. Differential Geometry of Manifold .............................. (194) Part IV. Real Analysis ................................................ (229) 1. Measurablity and Measure ...................................... (231) 2. Integral ........................................................ (256) 3. Space of Integrable Functions ................................... (283) 4. Differential ..................................................... (302) 5. Miscellaneous Problems ......................................... (322) Part V. Complex Analysis ............................................ (333) 1. Analytic and Harmonic Functions ............................... (335) 2. Geometry of Analytic Functions ................................ (360) 3. Complex Integration ............................................ (377) 4. The Maximum Modulus and Argument Principles ............... (413) 5. Series and Normal Families ..................................... (433) Part VI. Partial Differential Equations ................................ (455) 1. General Theory ................................................ (457) 2. Elliptic Equations .............................................. (472) 3. Parabolic Equations ............................................ (496) 4. Hyperbolic Equations ........................................... (513) Abbreviations of Universities in This Book ............................ (539)
Part I
Algebra
.
So c 3 5 dimR(V) = dimnR[X]/(f(X)) = degf(X). Prove that there is a nonzero subspace W of V . Thus a. . r n 2 ( t ) . Let S be a linear transformation of V into itself. Let W be an invariant subspace of V (that is. Make V into an B[X]module by defining for i=O w = T ( v ) for all w E V .3 Section 1 Linear Algebra 1101 Let V be a real vector space of dimension a t least 3 and let T E Endn(V).such that T ( W )C W . ml(t). Then V E E [ X ] / Ifor some maximal ideal of B[X]. such that T ( W ) W . 1102 Let V be a finite dimensional vector space over a field K . Since R [ X ] is a P. W # V .I. Now suppose V is a simple IR[X]module.D. W # V . This implies that we have an irreducible polynomial f ( A ) with degree 2 3 in B[X]. Hence V is not a simple B(X)module. Let m(t).Xi E B[X]and v E V It is clear that a subspace W of V is an B[X]submodule of V if and only if T (W ) W . that is. (Indian u) Solution.and rnz(t) be the minimal polynomial of S as linear transformation of V. W and V/W respectively.. This is a contradiction. there exists an irreducible polynomial f(X) of B[X]such that I = (f(X)). SW C W ) . there is a nonzero subspace W of V . (a) Prove that m(t)divides ml(t).
Then we have V = lR[X] . rnl(t) divides m(t). ml(t).m2(t) divides m(t). W = 0. (b) Since m(t)E AnnK[t]V G AnnK[t]W = (m1(t>).m2(t). V can be viewed as a module over the polynomial ring l R [ X ] via f ( X ) . a cyclic module. (a) Since m1(t). Then it is clear that ( m ( t ) = A n n ~ p l V ’(. m2(t)divides m(t). As usual. since m(t). W is a K[t]submodule of V.4 (b) Prove that if ml(t)and mZ(t)are relatively prime. v . since {Taw I i 2 0) spans V by (b). (c) Give an example of a case in which m(t)# ml(t). T/ can be viewed as a K[t]module via the linear transformation S . mz(t). (Indiana) Solution. + + 1103 Let V be a finite dimensional vector space over l and T : V + V be a R linear transformation such that (a) the minimal polynomial of T is irreducible and (b) there exists a vector u E V such that {TivI i 2 0) spans V. Since ml(t)and mz(t)are relatively prime. Similarly. z = f ( T ) (x).Then we have m(t)= ml(t).m l ( t ) ) A n n ~ p l W ) = and (m2(t)) = A%[t] V l W . (In i m a) d Solution. (c) Let W be a 2dimensional vector space over the field Q of rational numbers and S : W + W be a linear transformation with minimal polynomial t2 1. v c m1(t).for any f(A) E lR[X] and z E V. ) = Hence m(t)divides m l ( t ) m2(t). Then it is clear that m(t)= ml(t) = m2(t)= t 2 1. m2(t). then m(t)= m1(t)* m2(t). Since W is an Sinvariant subspace of V. So m(t)# ml(t)* m2(t) in this example. Let V = W @ W and S : V t V be the natural extenaion of S to V.ml(t)and mz(t)are all monic polynomials. Show that V doesn’t have proper Tinvariant subspace. ~ ( t m ( t ) E A n n ~ p l V (m(t)). Let m(X) be the minimal 3 .
(Inddana) Solution. is a Jordan block of order ei. . we have + V . Then JIB' = BIJl and Bl(#0 ) is nilpotent.Xn) commutes with P'BP. Then m(A) E Annwfxl(v). .l B ' P . Sufficiency. We have to show that A commutes with some nonzero nilpotent matrix. Necessity. then it is easy to see that A commutes with some nonzero nilpotent matrix). ) where Bi(i 2 2) = 0 E Me. 1104 Let A be an n x n matrix with entries in C...J t ) be the Jordan canonical form of A and PAP' = diag(J1.l .. Without loss of generality. So V is an irreducible R[A]module. Thus. Then there exists an invertible n x n matrix P such that PAP' = diag{ A 1 . . Let B' = diag(B1. B . . Taking B = P . So B = 0. Show that A has n distinct eigenvalues in C if and only if A commutes with no nonzero nilpotent matrix. An}..5 polynomial of the linear transformation T : V Since m(A) is irreducible.(C).. PAP' = diag{X1.. P .l B P is a diagonal matrix. . . . where P is an n x n invertible matrix and J. . . Suppose that A has n distinct eigenvalues A'. we may assume that el > 1 (If all the ei = 1. we have t o show B = 0. . Jz. . If A commutes with some nilpotent matrix B . =v R[X]/(m(A)) R[A]* E 21 (we may assume that V # 0). and A B = B A . A are distinct and ... . But the nilpontency of B implies that P ' B P is nilpotent. . . V does not have proper Tinvariant subspace. . A in C. Jt).. we have B # 0. B'. Jz . Then B' . . Since All Az. Let B1 be the Jordan block of order e l . PAP' = P A P . . Xa.. Let diag(J1.. Bz. with 0 on the diagonal. which is nilpotent. Hence we have P'BP = 0. Suppose that the characteristic polynomial of A has multiple roots. .
. U ~ . .A ) . z .>o.z = f(T).. A = ( u ~ ~ ) the ~ ~ of T relative to the base.vn}.{Ta?}. T E T }and { T * 8 } . the minimal polynomial m(X) of T is det(X1 .20 and { T Z 3 } . V = F[X]. s ) E V such that V = F [ X ] .6 1105 Suppose V is a finite dimensional vector space over a field K .. .T and 3 are any three nonzero vectors in V .A ) .(X)(d. be . r 0 (Stanford) Solution. In general.(X)} where the d.(F[X]) has the form diag{l. . .d.} and thus induce a linear transformation .(z) any z E V . U be a base for . V can be viewed as a module over the polynomial ring F[X] simply by f(X). . . 1106 Let V be a finite dimensional vector space over with basis {wl.A in M. .D. T : V + V a linear map such that the minimal polynomial of T coincides with the characteristic polynomial. By the structure theory of finite generated modules over P. > are nonzero submodules { . So at least two of them coincide. so m(X)= d.> coincide. Here. . .(X) are monic of positive degree and d.} of V over F . . matrix a normal form for X I . . + Hence s = 1 and v = F[X] z1 E?! f F[X]/(m(X)) is cyclic. z l $ F [ X ] . . Let CT be a permutation on {q.. 2 . . .(z = 1 ..(X) if i 5 j. (HU7Wrd) . . there exist z.. f ( A ) 6 F[X]. v. Obviously. then at least two of the three subspaces spaned by the sets (T*?it}z>~.(X) = det(X1. A on V . .) = where (d*(X)).Let { U ~ .Show that if %'. Ann(z.dl(X). according to the assumptions. ~ F [ X ] .zl has exactly two nonzero submodules.I>~ o of V over F[X]. .I. the three subspaces generated by the sets {T'Z'}. Show that A is diagonalizable.. which is the square of an irreducible polynomial in K [ T ] . . z z ~ . l. Since m(X) is the square of some irreducible polynomial..
1 is irreducible over the field Q of rational numbers.s 1 (io = 0. Prove that 3 divides the dimension of V. we may assume that CJ is the ncycle (v1.. . . Mj be the matrix of Alwj : Let Wi + Wj relative to the base { ~ i ~ ..~ + 1 . . . . . } . . . V ~ . Hence.7 Solution.w2.. vof~Wj overt. . 'u2. . . 1107 Let V be a finite dimensional vector space over the field of rational numbers..Ms+l} is the matrix of A relative to the base ( V ~ . .+ M = . . diag{l. Since T' = T2 T. Then the Wj are invariant subspaces of A and V = W1 @ W2 @ .. . .* .. ' we assume that when u is expressed as the product of disjoint cycles (This decomposition may have 1cycles). This completes the proof that A is diagonalizable. . By reordering the elements 211. .. . A be the matrix of T relative t o + + + ns some base of V . Let Wj be the subspace of V generated by {Kjl 1. V .. ) is . Obviously. 0 0 It is easy to see that the minimal polynomial of M is A" . .A A .. ~ ~ .. Suppose T is a nonsingular linear transformation of V such that T' = T2+T.2. . ..1. The matrix of A relative to the base { V ~ . 1. without loss of generality..(X)} be the normal form for X I . .1 is the minimal polynomial m(X) of T . So it suffices to prove that every M i is diagonalizable.. @ Ws+l.d. Now let n be the dimension of V over Q.. . then all such T's are similar.. Then M = i } diag{M1.1. and prove that if dimV = 3. . . . . .... 0 1 0 0 1 0 0 1 1 0 0 *. . ( H U 7 V d ) Solution.. . . is+l= n). vij} + for j = 1. l..dl(X). and thus M is diagonaliz able... . . .. V . .. .. T is annihilated by the polynomial X3 X2 .. X3 + X2 .u. Thus X3 X2 . .wn). V ~ .. . ~ ~ ~ .. . .
Then m(X) = 1 A" . It follows that all the 1108 Let Fq be a finite field with q = p" elements. .I.. II"'(u)} * is 0 1 0 .1 are in Fp. Suppose that 1 is diagonalizable as a linear map over Fp. . s = deg(det(XI .A ) ) = n. Let 1 : Fq + Fq be the Frobenius automorphism II(z) = z p .. The .. .l". 1.1. ..112. X3 rational canonical form for A (or 7') is T ' s are similar when dimV = 3. . ( X ) .1.1.. II(u). .). (Here is a misprint. If dimV = 3. . .A ) = d l ( X ) . By the Normal Base Theorem. 3 .M is diag{ 1 .1 = det(XE . The normal form for X . It is wellknown that Fp C Fq is a Galois extension with Gal(Fq/Fp) = {1. where p is a prime. .II"l(u)} is a base for Fq over F p . Since det(X1.8 where the &(A) are monk of positive degree and d i ( X ) l d j ( X ) if i 5 j.A is equivalent to diag(1. By the irreduciblity of &(A) = m(X)= X3 + X2 . On the other .1 has no multiple root. 11: .. Thus we have proved that 3 divides the dimension of V ..M ) . M= I. + X2  1).. then X I . It should be "n divides p . and all the roots of A" .:.1) and the minimal polyE nomial m(X)of M is A" .. we have &(A) = &(A) = .) (Hurvard) Solution.IIn1}. When II is considered as a linear map of Fq over Fp. . . Prove that II 1 considered as a linear map over Fp is diagonalizable if and only if n divides p" . the matrix of 1 relative to the base {u. = &(A) = X3 + X 2 . d . . there exists a u E Fpsuch that {u. 1 0 a . 1 II(u)....I12(u). I12(u). 1..11..A" .
So IT is diagonalizable as a linear map in over Fp. Show that the derivative of the determinant det A satisfies Then and . . (A .a2d). .  1 forms a subgroup of FJ = Fp\{O}.9 hand.1) . u ( ~ .a d .Lli)seelements are differentiable functions of real variable t .(A . . . Hence M is similar to diag{ 1.ad)(. 1109 Let A(t) be a nonsingular matrix r.a("l)d) where d = and a is the generator of the group F. A" .' ) ~ } Mn(Fp).1 has no multiple root and A"  1 = (A . all the root of An p .1.. Let A'(t) denote the matrix formed by the derivatives of the elements. if n divides p . Thus n divides Conversely..1. .\ .
A’). 41(z). trace(A’ . having degree 0. + det A.) .1. dt 1110 Let V be the vector space of polynomials p ( z ) = a coefficients a. Define an inner product on V by + bz + cz2 with real (a) Find a n orthonormal basis for V consisting of polynomials 4o(z). and q52(z). respectively.. . b.10 = det A1 + det A2 + . and 2. and c. Hence we have proved that d (det A) = det A . . (b) Use the answer to (a) to find the second degree polynomial that solves the minimization problem (Courant Inst.
9 and 175 cg = 0.x.p o ( z ) .$a(z:)) = ( p o ( z ) .x 2 to get an orthonormal basis of V with degree 0. Hence P O ( % ) = $ z [:ix3 .Zz.$i(.4i(. By an easy calculation. we get co = 0. 2 ) .))= C l (i = 0 . (b) Let and QO(Z) = z3 Po(.11 Solution. Then for any p(x) E V . Obviously. 8 3 . 3 . that is ) ( x 3 .)) =0 (i = 0.) such that ~ o ( z is orthogonal to V . 1 and 2 respectively as usual.5 z ) 2 d x = .x 2 is a base of V .) = z3 .1. c1 = and no(. (a) Since 1.2).x. 1 . rl ( . we can orthonormalization 1.
is minimal. a nn a2n +det anl . = det +s . ”’ .)... 3 det(A + szy*) a11 an1 = det + ..) Solution.. yn). (a) Show that there exist numbers a and b so that det(A + m y * ) = u + bs. . (a) Directly.. . and 1111 Suppose that A is a n n x n matrix and that 2 = ( :)... . Yn y*=(y1.. z. (b) Show that if det(A) # 0 then a = det(A) and b = det(A) y*Alz. X1Yn det a21 ... * . [[ Un1 . .**. . and y are real. Suppose that all the entries of A. SXIYI + ..... Xn Y = ( . ann * ZlYl ..12 Thus. * * aln ann + . + Sz1Yn SznY1 SXnYn .... when p ( z ) = gz.. ( c ) Is it true that a = 0 if det(A) = O? (Courant Inst.
. . An1 .. . Aln . a n n . we have a = det(A) and + + b = det ..13 Hence det(A + sxy*) = a bs for some a and b... .... (b) Since for any s . ... Ann All ) .... is the algebraic cofactor of Al Thus 1 det(A) ___ aij. . .. for any s. (c) If det(A) = 0.. * * An1 Aln ... ( If det A # 0... .... * . + det ... y*Alx.... ann + . n X n Yn n where A. . . ...). .. det(A + sxy*) = a bs as in (a).. .+ det n an11 a n . Ann = y*(det(A) ..In ... . . a = det(A) = 0.. .... . A')Z = det(A) .. All ... anl *....
. P' . ..O) . .l . and corresponding eigenvector V ~ . Since A1. 0 } . P isaninvertible matrix in illn(@) P . Then the matrix of T with respect to this basis is P ' . . . .fn} is a basis of ker(T).) = lirn X .O .. (Courant Inst. .... X ~ . ncc lim A'lw = = lim P d i a g { X . . .O) = P'P'l diag(l. e2. dim(ker(T)) = n . v 1200 P . diag{l.. Xz. (el.. Hence {fl} is a basis of Im(T). e n ) be the standard Let ncr. ... Let fn en Then ( f l . 5 j 5 n. . f n ) . v....) is a basis o f a n and diag(l...a. . P . f z . . lim Yn).. Assume that XI = .. .. .Xn aredistinct. ...O.. . Find the dimensions of the kernel and 1100 1 and that J A j l 1 for 2 < image of T and give basis for both.A2. {f2.. and dim(Im(T)) = 1. Let P = ( v l .1. v z .I A P = diag{A1. n+w ncc orthonormal basis of (En. 1113 Let A be a matrix.Y. w (Since lim (X.. Define 1 sin(A) = A . Define lirn T : C" + an by T ( v ) = lirn Anv.) Solution. f.l d i ag ( l ..O. f z . and For any v E a". XI. ... P ' . v n ).O.. V Z . . ..Az... Prove that noo Anv exists. A. For . . .14 1112 Let A be an n x n real matrix with distinct (possibly complex) eigenvalues.A3 3! 1 + A5 5! * .Xn} ..O)P'P'l is the matrix of T with respect t o the basis ( f l . . X ~ } ... .
15 express sin(A) in closed form.. ( 4 3o 0 ( 4 1 1 ) where . (Courant Inst.(4 0 10 ) .[. Hence sin(A)  ( . 1' )'=( 1 ). Denote B = ( 3 Then B is similar to ") . 103 O ( 1 1 ).) Solution.L 3!T ) 3 . 7 and 1 1 0 1 0 1 1 0bviously.
.. If B has more than two nonnegative eigenvalues. .B)Y >_ 0. (c) Obviously. (d) Show that B has 9 negative eigenvalues... we have Y T B Y = 5 5 X.) Solution. .I . (a) Show that (b) Use part (a) to show that A has all negative eigenvalues. . all the eigenvalues of A are negative. a . (a) A direct verification shows that (a) is true. . (Courant Inst. the symmetric matrix X.. Y T B Y 5 X. Y T B Y 5 0 and Y T B Y = 0 if and only if Y = 0. Let A1 and A 2 be two of them. z 1 .B is semidefinite positive. 1 A= **... 0 . Thus A .I . we have y1. . (b) Since z T A z 5 0 for all z E lR9 and z T A z = 0 if and only if z = 0. which agrees with A except in the first row and column: 1 3 3 2 B= 1 1 2 1 1 *. 2 1 1 2 All entries not shown are zero..2. Then for any Y E Elo. (c) Let B be the following 10 x 10 tridiagonal matrix.(B) > 1.. 0 . > 1. . 0 ) . Then for any Y E V1. Let A.YT(X. y2 E R I O such .. Show that A.YTY.16 1114 Let A be the 9 x 9 tridiagonal matrix '2 1 2 1 1 .(B) be the largest eigenvalue of B . ~ 9 ) I zi E IR} 5 Elo. (d) Let Vl = { ( O . . *.1 . .. Taking Y T = ( 1 .. this is.
. Az. . (Hyl = Iy1. (Courant Inst.Xn). A.. there exists an n x n orthogonal matrix P such that P'TP = diag(A1. . 2 ) . > A. we have min 2EV10) lTxl   min Z€VIO) IP'ldiag(Xl.y2). Since for any orthogonal matrix H and any y E B". en} be the standard orthonormal basis of Rn. * A 2 2 0. n x n matrix with distinct eigenvalues A 1 > A 2 > . Then min x €V ..A. .. Thus B has 9 negative eigenvalues. .to1 Idiag(Al. y E V1) and yTBy 2 0 (y E VZ). P . yTBy = a?A1 u.) 1x1 Idiag(Al. L ) z l 1x1 .A2. Show = by showing 5 and 2 separately. Show that A2 ITXI = max min v 2EV{O} 1x1 ' where V ranges over all two dimensional subspaces of Bnand 1x1 is the Euclidean norm of 2 . Then we have yTBy < 0 (y # 0. y1 y1 = yTy2 = 1 and Byi = Aiyi (i = 1 . .17 T that y1ly2. Let V = (yl... . .).a contradiction. ..x1 when V ranges over all two dimensional subspaces of lR". You may wish to express T in a basis of eigenvectors. e 2 . PlxI IP1x1 . . 1115 Let T be a real symmetric... and Now let {el. . Hence max v min Z€V{O} w = m a x 1x1 v min Z€V{O) Idiag(A1..l V is also twodimensional.A. PlxI 1x1  2EVIO) min If V is a 2dimensional subspace. V = (e1...) Solution.. z the subspace spaned by y1 and y2. z + + + 1 there exists 0 # y E V n VZ. By assumption.. Since dim V1+ dim V = 9 2 = 11 > 10.eZ).. Then for any y = ulyl uzy2 E V2. Hint. positivedefinite.) 1x1 .
we can find an orthogonal basis {fl. . A ). . . ... . . > A 2 > 0). = alf1 Then Q = ( q i j ) . ... It follows that I v 1x1 On the other hand. Let (fl. ez.. fz .fz). for any two dimensional subspace V of 1R".2 Az. and Thus when V ranges over all two dimensional subspace of B".fi.. e. lT. max min Z€V{O) ITX . and if 0 Idiag(A1. . x n is an orthogonal matrix. .fn} of IR" such that V = (fl. . fn IT = Q(e1.XI so. .18 = A2 (A. 1x1 #x + azfz E V .
(c) Use part (a) to compute the inverse of 1 A = [ ! i . r j m ( X ~ ) . . 1 0 0 (Courant Inst. . + z". Let A Az. we claim that the norms of all eigenvalues of P are < 1. * Then ~ m ( A l ) . < 00 k=O .19 1116 For any n x n matrix P . ~ are X . consider the sum R(P)= M c 00 Pk exist and R(P) ( I .) Solution. . Denote ( A1 e f .' .. then (IP)' < (b) Assume that IlPll < 1 in a matrix norm induced by a vector norm. r). * ) rjm(z)= 1 + z + * .. A. (a) First. = k=O (a) Prove that if k=O JIPklJJ00. + P" Since C k=O 00 IJPk(l co and the norms over vector space are all equivalent. QlpQ= be the Jordan form of P. Prove that ll(I .P ) . be the n eigenvalues of P and ' . ) n eigenvalues of m ( the sm= E + P + P 2 + . ~ ~ ~ .P)'ll 5 *.
P)' = I + P ( I .P). llull < llull for the corresponding vector norm. Hence pal < l ( 1 5 i 5 n). Thus all the eigenvalues of I . C m=O W is convergent (i = 1 . Since mw lim Q . The matrix I . Now suppose ( I .llPull 5 llPll . for any nonzero vector u.P)' = (IP)'.~ s . 2 .(IP) rn00 = IP"+'. m+m lim Sm = lim ( I . (b) By definition. that is. n ) .~ ( lim S.lim ~ + = nrn = (I .20 is convergent. = We must have u = 0. Since S.u 0 for some vector u.P)'.P"+')(I  PI' I .O)(I. moo lim dm(X.P ) is invertible.. . .) exists. mw 1 Q.1 .P is therefore invertible. Since llPll < 1. l l ~ l l =sup { #lv nonzero vector I . Q = Q .I J . .P are nonzero. Thus 00 ( l( I ) PI' R ( P )= k=O Pk = ( I . Then llull = IlIuII = IIPuII. and we can write ( I . Then we have . and (I.
. Then G' is isomorphic to a subgroup H of G L ( m .w + 0 * 01g E G}. there exists some P E GL. since there is a surjective homomorphism G + G' ( 5 EndF(V)). ( CoZumbia) Solution. It is well known that m j=1 and U= is a Sylow psubgroup of GL(m.F p n ) . the corresponding matrices of the linear transformations in G' are in U .(m. So 1117 Let G be a pgroup and G x V + V be a linear action of G on a finite vector space over a finite field Fan. Using Sylow theory.N . .21 Then A = I . which are unipotent matrices. if we change the base of V via P . By Sylow's Theorem. Then G' is a pgroup.Fan). prove that there exist a basis of V in which all the transformation v + a . V . where m = dirnp.Fpn)such that PHP' U . Thus. We fix a base of V over Fan. v(a E G) are unipotent matrices. Let G' = {V + V .
. where S is diagonal and N is strictly triangular. 1119 Let A be an n x n real matrix. Since the minimal polynomial of S is not equal to its characteristic polynomial. . dimF F [ S ]< dimF On the other hand. the dimension of the i vector spcae over F of matrices commutative with the matrix of S is v. there exists a t least one n x n real matrix B with B" = A . ) (Stunfod) Solution. Make use of the Jordan form S N of a conjugate of A . & ( A ) be the invariant factors # 1 of S and let n = degdi(X). . 0bviously. all of whose (complex) eigenvalues are real and positive. dimF F [ S ] = degrn(X). . (Stanford) + . Then as Fvector space. N 2 C j=1 8 nj = dimF V. where r n ( A ) is the minimal polynomial of S .. Show that for any integer rn 2 1. So there is a linear transformation T of V such that T commutes with S but T is not a polynomial in S.+akSk for some k and ai E F . Let F [ S ]= {f(S): V t V I f(X) E F [ X ] } . Hint. then by Frobenius theorem.22 1118 Let S be an endomorphism of a finite dimensional vector space V over a field F whose characteristic polynomial is not equal to its minimal polynomial.&(A).j=1 + 1)nj. Show that there is an endomorphism T of V so that T commutes with S but T is not a polynomial in S (T is a polynomial in S if T = aol+alS+a2S2+. . let d l ( X ) . S N = C 2s 2 j ( .
. any 1 5 i 5 k.. . J2. C. all of whose eigenvalues are real and positive. J k ) where Ji ( 1 5 i 5 k) are Jordan blocks with diagonal elements real and positive. As proved in the above. let b = m & B= 0 0 0 0 It is easy to see that *.(Aa)... .. Now let A be an n x n real matrix. .* b O 1 b Bm= Obviously. '.. Then there exists an n x n invertible real matrix P such that F I A P= diag(J1. A and Bm are similar.. for any integer m 2 1.}.Bk))m ... such that By = Ji (1 5 i 5 k). 1.. XEA and XEB" have the same invariant divisors (1. Hence PlAP = (diag(B1.B2. .lbmn+l cibm1 . .. Suppose first that A= be a Jordan block with a real and positive. chbml " b I* A = QlB"Q = (QlBQ)" for some invertible n x n real matrix Q. .. .. .. ~. Thus L : b " 0 Chb"l b " Cibm1 .2bmn+2 0 .. For any integer m and b 1 0 O b 1 2 1. .23 Solution.. there exists real matrix B.
+ V. V and elements of a. .(Xal)el'l . (a) Show that there are 2'invariant subspaces V .~ = ~ i ) ~ ~ j ) .j E V and A n n ( ~ . (Aan)ebe the elementary divisors of VqX]..* .Then . .j = K[A]wij. .a Z ) e z T .where a1. Let (A al)ell. (Xa2)e=. N ( T )E K [ T ] such that T = S ( T ) + N ( T ) . Hint. Denote V. is nilpotent. ( T ) : V + V is diagonalizable. a2. Use the Chinese Remainder Theorem.)enl. E K such that V is the direct sum of the V. . . ( Stanford) Solution. . and (T . (A transformation N is nilpotent if N" = 0. j ) ((A . * * * . Viewing V as a module over the polynomial ring K[A] via T. 0 where B k 1120 Let V be a finite dimensional vector space over an algebraically closed field K . .. (Aa. ( A .24 and B1 B2 0 P'] m = Bm.ail) : V .an are distinct and + By the structure theorem of finitely generated modules over PID. we may write where all the w. and let T : V 4 V be a linear transformation. .) (b) show that there are polynomials S ( T ) . . for some n 2 1.. and N ( T ) : V S V is nilpotent.
a i l ) : V.a 2 ) e 2 r 2 . waj for all wij and it is easy t o see that S ( T ) : V + V is diagonalizable. ( A . So the minimal polynomial m(X) I N ( X ) h for some positive integer h. Since ( A . ).S(X). . Thus (b) is proved. Let N ( X ) = X . + . Thus are pairwise relatively prime. by Chinese Remainder Theorem. there exists some S(X) E K[X]such that S(X) ai( mod ( A  ai)eir.al)e*'l (A . (1 5 i 5 n).:j (a) is proved. Thus N ( T ) : V + V is nilpotent.25 The K j ' s are Tinvariant subspace and ( T . . + vj is nilpotent. Obviously. T = S ( T ) N ( T ) and X . waj = Cya . Then S ( T ).a I N ( X ) for any i 1 5 i 5 n. .
let A be an invertible 2 x 2 matrix with real entries such that N=(d)?( if d = det A < 0. In case a). .) or NgN'= (1 1 ). Describe the set of conjugacy classes of elements of G. then N E G and NgN'= 1 0) A l ( . Let g E G. then A 1 = A 2 = 1 or A 1 = A 2 = 1. g = In case b) g = In case c). of determinant one.26 SECTION 2 GROUP THEORY 1201 Let G be the group of real 2 x 2 matrices. . Then A 1 A 2 = 1. (Hamud) Solution. i) If A 1 = A 2 . A 1 and A2 be the eigenvalues of g viewed in M2(@). g is similar to  in GL2(R).
M i f d = d e t M > 0 or N=(d)i( if det M = d 1 0) M ' < 0. Then N E G and Thus g is conjugate to classes ( x". AS in case c). So in this case. we have two conjugacy classes [ ( i1 yl')] and [ ( 0' 1. ) in A2 Xi [( i!i :)] [( i! .') a21 ( E:: :tz ) E G. g is similar to similar (conjugate) to is not conjugate to ( i11.')I* ( 0' 21 ) ( 0' ) ( i11. Thus in case d ) . ii) If X i # and E B (i = 1.) in G . then g is similar to ( 2 x". we have two conjugacy classes In case d). g is inGand ( i11' ) GL2(B). all = a22 and = 1 which is a contradiction. But (A i ) isnot in G. Thus in case c). There exists M E G L z ( R ) such that We take N = d4 .2). we have the conjugacy . ) in G . then we have and = 0.)].) or in G L ~ ( B ) .27 Hence g is conjugate to conjugate to ( ( q' ) 1 1 ) or (1 1 ) i n G . for if A*( where A = d e t A = a!l i :)A'=( .
B E M 2 ( B ) are similar in M2(~. we have the conjugacy classes . X2 11 < 8 8 # 0. where T.isino cos 8 sin8 cos8 ) in GL2(C). which is a contradiction. for if sin8 COSB sin8 cos8 such that cos8 sin8 ) Al = ( cos8 sin0 det A = (a:l + ail) = 1. g is similar to sin8 As it is well known. A2 = 1 and trace ( g ) = A 1 + A 2 E E ) . So in this case.28 iii) If XI # A 2 and X i E C\E (i = 1.7 implies that A and B are similar in M2(nZ). As in the above.  O i sin 8 ) in G L ~ ( C ) . we conclude that g is conjugate to cos$ sin6 or sin8 cos6 sin$ C O S ~ cosB sin8 in G.i s i n 8 (since XI . But cos8 sin$ and cos0 sin0 are not conjugate in G. that two matrices A . So there exists M E G L 2 ( E )such that cos8 sin0 MgMI= and also similar to ( cos8 ( cos8.2). we denote A1 = cos8 + i s i n 8 and = cos8 .
1 ways (a row of zeros not being allowed).q ways (no multiple of the first row being allowed). and so on. Thus we can conclude that the order of GL.q2 ways (no linear combination of the first two rows being allowed). the subgroup of upper triangular matrices with 1's on the diagonal a) Calculate the orders of G and U .) Xz and XI 3 XZ = 1 sin6 c0s6 cos6 1 0 (61 1202 Let G = G L n ( F q ) . a) When forming a matrix in G L n ( F q ) .(Fq) is + .1 Hint. E El " {[(i Y ) ]I b = * l } { [( )] < < . Deduce that U is a Sylow psubgroup of G.the group of invertible n x n matrices over the finite field Fq with q = p " . b) Deduce that every psubgroup of G is conjugate to a subgroup of U . the second row in qn .29 and cos6 sin6 sin6 cos6 To sum up. d) Show that g E G has ppower order iff g = I N with N n = 0. the third row in qn . c) Determine the number of Gconjugates of U. ) ] ~ A I . U = Un(Fq). e) Show that G contains elements of order qn . Make use of Fqn. p a prime.we may choose the first row in qn . (Columbia) Solution. the set of conjugate classes of elements of G is U { [ ( $ x".
So g N G ( U ) and we have N G ( U )= T .. 1 . n + We can define a map 19 : T + F. of matrices ( a i j ) E G with a. we get g i j = 0 if i > j Hence g E T = T n ( F q ) the set . U is a Sylow psubgroup of G. By this way... Suppose g = ( g i j ) n x n E N G ( U ) .. we get g n .3 ways and so on...' & U (by matrix multiplication). i=l b) It follows directly from Sylow's Theorem. we get gni = 0.we may choose the first row in q n . there exists some 1 U12 0 u=[. the number of Gconjugates of U is equal to [G : N G ( U ) ] where N G ( U ) is the normalizer of U in G. So gn. j = 0 if i < n . if g E T: It is easy to see that g U g . d ) Suppose g E G has ppower order.. the second row in q n P 2 ways: the third rows in q n . 0's elsewhere. Then for any 1 5 i < j 5 n.l M h = I + N where N = h . . . + . There exists an h E G such that h g h .. j E V such that g ( I E j j ) = u g where Eij is the matrix with a lone 1 in the (i.l E U . Denote h g h ..l = I + M where M is a matrix with zero on and below the diagonal.. .l ) . x .l ways. So the order of T is b Hence the number of Gconjugates of U is i=l n (qi n .l M " h = 0.I..or gUg' = U .30 When forming a matrix in U n ( F q ) . Uln ~ 2 3 ::: . Conversely.1)th row. . x F. Checking the last row of g ( l + E i j ) and u g .l . .l ) / ( q l)n.l M h and N" = h .j) place.. c) By Sylow's Theorem. by mapping a matrix onto its principal diagonal. matrix multiplication shows that 0 is a group epimorphism whose kernel is precisely U = Un(Fq). Then checking the ( n . = 0 if i < TI. Then ( 9 ) is a psubgroup and ( 9 ) is conjugate to a subgroup of U by b).. Hence the order of U n ( F q )is n (P'~ n SO U = Un(Fq) is a psubgroup of G = G L n ( F q ) . Obviously Mn = 0 and g = I h .1.j = 0 (0 5 j < i 5 n). Since p j [G : U ] = .
) Obviously. then we can obtain a group homomorphism (T : FLf. has an element with order q" . So the order of g is ppower. It follows that G contains element of order q" .1.1) + 1. the Jordan canonical form for N is a matrix with zero on and below the diagenal.1. We fix a base of Fqn over Fq. if g = I N with N" = 0. .31 Conversely. For any a(# 0) E Fqn. $ G = Gn(Fq). a l ( z ) = a z (over Fq) which is invertible. ~ ( u = at. (IHI ... f + 1203 [GI = IgEG U gHg' 5 [G : N ( H ) ]. It is well known that F. e) Consider the finite field Fqn as an extension of the field Fq and as a vector space over Fq. (T is injective. Thus g is similar (conjugate) to some matrix in U . we have a linear map al : Fqn FqR.
(1) = KOa K 1 a . we have H a G. Now. .i.l Ki/Kil (i = 1.1 u K . (b) Show that if H is simple. aK n P 1a K U G is a normal series of G with factors KiIKi1 (i = 1 . (a) < G.l a H . Obviously.. .. we have the isomorphism G / K N H . K a G and GIH is cyclic __of order 2. I c } N K 4 (Klein four group). n ) and G / K . l . Hence G is cyclic. . l Suppose H is simple. Let H be a maximal subgroup of G. (Columbia) Solution. (a) Give an example to show that G / K need not be isomorphic to H . then G I K N H . n .*i}. there exist subgroups Hi such that Hi I> H . (a) Let G = Q8 = { f l . Suppose that G is not cyclic. .. By the JordanHolder Theorem. = K be a composition series of . a H. a K. 3 t k } be the Hamilton’s Quaternion group and H = (i) = {3tl. . 1205 Prove that if G is a finitely generated infinite group then for each positive integer n. K = (1) = {*l} be the subgroups of G generated by i and 1 respectively. * * * . then (1) u H a H 1 a . On the other hand. so it is isomorphic to K .32 (b) First we claim that G is cyclic. K is not isomorphic to H .2. K . G has only finitely many subgroups of index n. . . . f j . = G N is a composition series of G with composition factors H and H i / H i . j . . n). This contradicts the maximality o f k . Since GIH N K . & i . .1). So (a) is contained in some maximal subgroup of G (since G is finite). n ) and Hi ~ H i + (i = O . Then for any a E G . UEG 1204 Suppose that H and K are normal subgroups of a finite group G and that GIH = K. Then we have G = U gHgl by assumption. But G I K = { 1. . .. By (a) G = H . H i I H N Ki ( i = O . it is easy to see that G is cyclic of prime power order since all of its maximal subgroup are conjugate. 2 . . . (b) Let (1) = KOa K1 4.
with HG as its kernel. Show that if H 5 G (finite index) then there exists a homomorphism of G/HG into S[G:XI. ker8= {( ) 16tnJN Obviously. where GIH denotes the set of all left cosets z H ( z E G ) . Show that G is solvable but not nilpotent. there are only finite number of homomorphisms from G to S. ~ ' ) Matrix multiplication shows that 0 is a homomorphism. G/HG has only finitely many subgroups of index n. Let H be any subgroup of G such that [G : H ] = n. 1206 Let G= {( a!l ) la E lR*. and . . g . n Hence the group GIHG is isomorphic to a subgroup of S. Hence G is a solvable group. It follows that the set g€G {HGIH < G . Let 0 be the map G + R* x lR* mapping a!l ) to its diagonal ( a . the symmetric group on n letters. denotes the symmetric group on n letters.lR). Let H 5 G and define HG = g€G glHg. (Columbia) Solution. ( z H ) = ( g x ) . N = ker0 (21 (lR. This induces a homomorphism G + S. Since G/HG is finite. ( Co2umbia) Solution. H defines an action of G on G I H . where S. Suppose . Then G x G I H + G / H .+))is an abelian subgroup of G.33 Hint. Thus G has only finitely many subgroups of index n. Since G is finite generated.. and GIN E x IR* is also abelian. The kernel of this action is HG = glHg. [G : H ] = n } is finite. E lR b 1 ( C_ GL(2.
p ) (p" . So G is not nilpotent. there is a basis of V with respect to which all of the elements of H are upper triangular. (a) Recall that a linear transformation is called semisimple if its minimal polynomial is separable. Prove that a transformation T E G is semisimple if and only if Tprnl= 1 for some positive integer m. Hence f(X) = fl(X).l = 1. Show that H can be simultaneously upper triangularized.. (b) Let H be a subgroup of G of order a power of p . that is. Then for any ( . Thus T p r n .l)(pn.(G) = a * * = G $ where Ci+l(G)/C. x!!l So ay+ bx' = xb+yal for any x Thus # 0. n2 'nk. Find a Sylow psubgroup of G.X over Fp and Ei = Fp[X]/(fi(X)) i 5 k ) . a = f l .l ) . G = G L F ~ ( V Recall that IGI = (p" ).= C. A direct discussion as above shows that C ( G / C ( G ) ) ( 1 ) . then f ( X ) = f l ( X ) f 2 ( X ) .. Let E be a splitting subfield of XPm . that is.l ) . (Indiana) Solution. f k ( X ) for some distinct irreducible polynomials fi ( 1 5 i 5 k ) . If T is semisimple.. Since ~i # 0 and 7. ) EG. (a) Let T E G and f ( X ) be the minimal polynomial of T over Fp. E. Denote n = degfi(X) i and m = n1 .p " .frnl = 1. Then X 1 f ( X ) . is isomorphic to a subfield of E . Let . Hint. y E IR.l .  . It follows that b = 0. f(X) is separable (remark: here separability means that f(X) has no multiple roots).(G) C ( G / C i ( G ) ) = (Co(G)= 1 ) .fi(X). Hence = C1(G)= C2(G)= * * . 1207 Let p be a prime and let V be an ndimensional vector space over Fp. f i ( A ) ~ ( A P r n . Since nilm. Then IEl = p" and Ei is a field of order (1 5 pni. ( 1 5 i 5 k ) .fi~(X)I(X~~~l) .. So E contains an element ~i whose minimal polynomial over Fp is f i ( x ) .34 the center of G.
So H K = K H is a subgroup of G. (a) Since K is the commutator subgroup of G. Now let H be a subgroup of G of order a power of p. the subgroup of upper triangular matrices with 1's on the diagonal.we may choose the first row in pn' ways.. 1208 Let p and q be distinct prime numbers. for any h E H and k E K . G = H K is abelian. (p" . By Sylow Theorem. Thus T is semisimple.p"') n n1 = p. ( c ) First we claim that H is not normal in G. Let H be a pSylow subgroup of G. Otherwise. Since (XP"'' .p " . (b) Let U = U.2 .p ) * * * (p" . = .p = pv (when forming a matrix in U .I). (Indiana) Solution. f ( X ) has no multiple roots.l 5 gKg' = U. q such that its commutator subgroup K is of order q. Hence G = H K and H 21 H / ( e ) = H / H n K 21 H K I K E G / K is abelian.1) . Since K is abelian and H is abelian. (b) Suppose that for any elements h E H and k E K holds hk = kh. Let G be a group of order p 3 . By Sylow's Theorem. Thus we have proved that H can be simultaneously upper triangularized. ( p .35 Next suppose TPm' = 1 for some positive integer m. (p"' . U is a Sylow psubgroup of G. Since )GI = (p" . hkhlkl E H n K = { e } and so hk = kh. (a) Show that H is abelian and G = H K . (c) From (b) show that p divides q .1.1)' = XP . Hence there exists an element g in G such that g H g .l)(pn . K is normal in G and G / K is abelian. This contradicts the fact that the commutator subgroup K of G is of order q. the second row in pn2 ways and so on). Then JUJ p"l . IHJ = p 3 .2 # 0. Then f(X)I(P"'' m 1)..(V).. . Since IK( = q . (b) Show that there are elements h E H and k 6 K such that hk # kh.  . H is contained in some Sylow psubgroup K of G and K is conjugate t o U . H n K = { e } and lHKl = p3 q = IGI.1). This proves (b).
= 0. So it is Again by Sylow Theorem. (a) Let rp be the number of Sylow psubgroups of G. it is easy t o see that rp = 1. Aq. 1200 (a) Let p . K be the subgroup of F generated by f1 = 9el 3ez 6e3. by Sylow’s Theorem. Let G be a group of order pq’. So G has a subgroup K of order q. (b) If q = 2. Denote + + + + 9 3 6 .1. 1210 Let A be the abelian group on generators el f. G also contains an element of order q . q be primes. and g. By Cauchy’s Theorem.1. So G has only one Sylow psubgroup H . Let F be the free abelian g r o u p g e l $Zez $Ze3. Obviously A E F / K . (Stanford) Solution. G may not contain a subgroup of order p .3f+6g Give a decomposition of A as a direct sum of cyclic groups of prime order or infinite order. = 0. Show G has a subgroup of order p q . A4 does not have a subgroup of order 6 . is a group of order 3 .3e2 6e3. Thus H . For example. (b) What can you say if q = 2 (and p > q ) ? (Indiana) Solution. subject to the relations 9e+3f+6g 3e = 0. K is a subgroup of order p . Then. the alternating group of degree 4. rPlqz and plrp . q . p > q > 2 . q are primes and p > q > 2 . + 3f 3e. q since H is normal in G.36 the number of pSylow subgroups of G is greater than 1 and divides q . fz = 3el 3ez and f3 = 3el . which is normal in G and of order p . 2 2 . Since p . q. plq .
$ Z e L and K = Zfl + Z f 2 +Zf3 = XCe.. so A S r" F / K = 2 e i $Zek $57eL/Uei 57/(3) $2/(6) @Z 2 2 $ 2 3 $573 @ Saei 2 $2. fL Then F =Zei $Ze'. f4)' = Q .6. e$>' = P . @ Uel. f3)'. 1211 Let M be an n x n matrix of integers. e3)' and (fi. e 2 . Let (e..0} ( P and Q are invertible matrices in M 3 ( Z ) ) .l ( e l . e i . View M as an endomorphism of 2". a) Show that 2"/MZn finite.6. is b) Show that the order of Zn/M57" is equal to the absolute value of the determinant of M . Suppose that M is invertible when viewed as a matrix of rational numbers. (fi. that there exists an n x n matrix N with rational entries so that M N equals the n x n identity matrix.37 It is easy t o get the normal form diag{ 3. (Stanford) .e.0} of M in M 3 ( Z ) and to find P = ( O1 0 1 ) 1 1 0 and 1 Q = ( O0 1 1 2 1 such that Q M P = diag{3.f 2 . i..
] Calculate the order of the additive group G = D2/K where K is the Dsubmodule of D2 generated by ( 2 t 11[ . e2. d z . . Since Q is invertible in M n ( Z ) . . .21)..dz.. f. 0bviously.f )' = Q . (fil f. .ek} is another base ofZ". det P = det M or . = det Q . .Solution. Then express G as a product of cyclic groups.) / I = diag{dl.. el. (Haward) + .dn} be the normal form of M in M. . . ... . ..l). . f n }.det M . en)' = M P (ei.fn )I I = Q M P (ei. ...(Z) be invertible matrices in M. where [MI is + the determinant of M .. .  Hence 1212 Let D = a [ [ with E = *.el.)' Then {ei.. e. . dz. . d. 1 * * f n )'. . Let (fl. Q E M. d. 1' 1 = Q . (fi f~r : . . . ([ + 2 . ( ..e...(Z) such that D = Q M P . It is also obvious from the facts that the map g : Zn/MZn Z " / l M p " .... . . det M . . .(Z). e n ) . Then M P is generated by { f l l f 2 . eh. e2.. b) Let D = diag{dl. . } . . . and P . . . . . Obviously det D = dl ... . . fn)' = M . n ..Denote (ei..f.ek)l = P'(el.. en)I . ..ez. M Z n can be generated by {f. . e n } be a base for the free module Z". . I / fz. .4) and (21. 2 . . 6 ( r ) = M * X is injective. (f. (ei. a) It follows directly from b). . M * is the adjoint of M . . . (fll f~.el.l). . eh . .} where f. .. (el. Let { e l . and the fact p n / l M p n l = ( I det MI)". 7 * * * ... . where the di # 0 and dildi+l (i = 1 . f.
e2 .1 2 1 1 A= 1 1 2 .39 Solution. for A in hf6x4(Z) and to find invertible matrices P E &(Z) and Q E 1M4(Z) such that PA& = N .l] 0 .+ 21 (el + 0 .21(el + 21(e2} is a generating +b(O e l subset of K as Zmodule (additive group).e2.2el + ( e l 4e2 + 5e2.(el . 2el.<ez) + (ez) + b(el+ (el . . It is easy t o see that {el + 2te1. e2 + 21 . Then + 0’= Del @ Dez = Z e l @Z(el @Zez @Z<e2. (e2). It follows that D 2 / K . D = Z [ t ] = Z@Z< (as additive group). It is routine to get the normal form 1 0 0 0 N = [ 0 .2(e2).e2 .2(e2.e2 .e2 . and {(1+2<)e1+(1+<)ez1(2+ ()el + (4 + () .5(e2.0 . e2} is a generating subset of the Dsubmodule K of 0’. .2/(3) @Z/(21)and the order of D 2 / K is 63.e2 + + + (e2) + b(2e1 + (4 + Oezl + (1 + t ) e ~ l . For any a bt E D . 0 21 . Let { e l l e 2 } be a base of the free Dmodule 0’.(el . + + (a bt)[(l+ = a ( e l + 2tel . el 21. 2 1 e l + 21e2. Denote 1 2 .5(ez).9 0 ~ 21 5 ~ . 21.e2 + 0 . Since ( is a root of the irreducible polynomial z2 z+ 1.  ( a b0[(2 ()el a(2el + < e l . 0 . {ell(el.(e2} i s a baseofD2asZmodule.el + (el .4ez + (a = a(21 .e2 + (ez. <el + 21 . e l + b()(2le1+ 21e2) + 0 . e2.
It is easy to know that if a = 1. Similarly. we claim that matrices in S L ( 2 . Z )is the group of 2 x 2 matrices with integral coefficients and determinant = 1. Z ) with the form ( 1) is a prodct of the for some b'.40 1213 where and S L ( 2 . a1 E Z such that 1 5 al < c and a = qc+al. .d' EZ. for some b' E Z. (Stanford) Solution.c'. Suppose that M = Prove that S L ( 2 . if c > a > 1. let c = ha + e l . then for some d' EZ. Since det M = ad .there exist some q . Z ) is generated by ( 1 rtl ) ) ( i1 If both a and c are nonzero. and if c = 1.be = 1. 1 5 c1 < a . If a Hence > c > 1. a and c are relatively prime.
It is easy to check this by the property of the elementary matrices. This completes the proof.41 According to the above discussions. X ) with the form for some b'.  1215 Let P be a Sylow subgroup of a finite group G. (Indiana) Solution. 1214 Let G be a finite group. So we have g = h .l . Let H be asubgroup of G. N . it suffices to prove that all are in (( 1 fl cc) . Let N = { x E G I xPx' = P } . H 2 N .l = H . ( i1 )) where b ' . Z ) = generated by ( 1 3zl ) ( and 0 1 ). So to prove that S L ( 2 . Hence h'gP(h'g)' = P and h'g E N .. we have 9Pgl C gKg' 2 K since K is normal in G. d' E Z. N . If N is the normalizer of P in G (i. c ' . h'g E K N . For any g E G. N = {g E G 1 gPg' = P } then show that G = K . By Sylow Theorem. K a normal subgroup of G and P a Sylow subgroup of K . then y E N. there exists h E K such that gPgl = h P h .e. c'. Thus we have G = K . Hence gPgl is also a Sylow subgroup of K . it can be derived that M = is a product of the matrices in S L ( 2 . d ' E Z . (Indiana) . Prove: If y E G such that y H y .
Hence h'yP (hly)' = P and h'y N .hlyEH. yPy' is also a Sylow subgroup of H .(1 1 2 1 . Since J{(a. Since yPy' & yHy' H.N cH. P is a Sylow subgroup of H . Obviously. then. So 1 y=h. 1216 Let G be a finite group.b)E G x G I ab = bu}l = P(G) = L .b) E G x G I ab = ba}J IG x GI (a) If G is not commutative. (b) Give an example such that P ( G ) is exactly 5/8. (Stanfod) Solution.42 Solution. So there exists an h E H such that yPy' = hPh'. (a) Let C(G) be the center of G and C G ( U= { b E G I ab = ba} be the ) centralizer of a in G for any a E G. If G is not commutative. prove that P ( G ) 6 5/8. The probability of G to be commutative is defined by P(G) = I{(a. IG/C(G)I L 4. obviously. + ) 4 =8 5 .
It is easy to see that P ( G ) = l =/ 518. Then IGJ = 8 and C(G) = (1. 3 5 l C ~ ( a )< 8. obviously. and so ~ C G ( U )4.C ( G ) . . 0 5 j 5 3. .43 (b) Let G be the dihedral group 0 4 = { z * y j I 0 5 i 5 1 .z2 = y4 = 1 zy = y3z}. y}. For any a E G .
Since = the integral divisor of yo + 2/2 or yo can only be f l . Hence Z [ m is Euclidean. 20 is not a prime element in Z[. 1) is the set of units of Z . Z [ G ] is a unique factorization domain..G) zg. So we have a map 6 :Z[]* + Z?. Then ab' = p v. we define 6 ( a ) = m2 2n2. Obviously T and 6(T) = 6 ( b ) * ( l p . Then w + + + a = b(p+vG) = b [ ( u + p. a H 6(a). b) By a).v)Q] +v E Z[] Q is in Z [] and r = a .bq = b(p . (Hurvard) Solution.44 SECTION 3 RING THEORY 1301 a) Prove the ring Z [ n ] is Euclidean.\/z where p and Y are rational numbers.n&? is also a n irreducible divisor of 2 0 . find all integer solutions to the equation y2 + 2 = x3. Suppose that a. . In the ring Z [ G ] . (1. b # 0 E a[]. ti = m . We can find integers u and v such that I .v)Q.u ) + b(v . If a = m nis an irreducible divisor of 2 0 .pl 5 1/2 u and l . a) It is readily known that Z [ n ] is a subring of a. [] Suppose (20.vI 5 1/2.yo) be an integer solution to the equation y2 2 = x3. For any a = m nz/2 E Z[z/2].\/z]. b) Using a). hence an integral ' domain.2112 5 6(b). (a + t) 2* + 21v  wl2) < 6(b).u ) = bq+r where Q = u + (v + v . we have (yo Q)(yo . and if alyo then + + + + n.
exactly (a . 2 . Show that for any element a E Z/$. M. On the ) are other hand.y 2 = i + j . 2 (q)2} y .(q)2} elements in S. : i 0 < i . Thus it is easy to conclude that the integer solution to the equation y 2 + 2 = z3 are y = 5 and x=3 { 1302 Z/pZsuch that a = b2 + c2. Since a .j # O ( O < i + j 5 p . i .. that is a = b2 c 2 . has exactly elements. n. for any A2 E S where 9 < n < p .p ) 2 = has exactly elements.S ) n S # 0.2n3 = (3m2. Let p be a prime.6mn2 1 = 3m2n. (b) Let J be a nilpotent ideal of R (J" = 0 for some m).T..1/2. {o. .a E R*. .. and GL.(R)*. the set a .1. yo is of the form ( m nJ2)3. R* denotes its multiplicative group of units.(R) = M.b2 I b E Z/@} has elements.z. Let S be the set { b 2 I b E a/$}.b2 = c 2 . When p = 2 . Assume that p # 2.22. n E Z.. .45  from which it follows that a31yo fl. ? 2 .S C Z/pZ and Z/pZ has only p elements. Then S = {o. (a) If a E R is nilpotent (am = 0 for some m) then 1 . it is trivial. .rn.0 I j 5 e. Now for any element a E Z/@.0 < j 5 q. . Hence + + + yo = m3 .S := {a .c3ly0 . 0 < p n 5 9 and A2 = ~  n2 = ( n .2n2). T . So without alyo loss of generality. Hence there exist 6. On one hand for any i # j ..T. c E (Harvard) Solution. 9 G .j < 9 if i > j .Thus 2 S' = { b 2 I b EZ/$} = { a . there exist b. ( q ) } 2 + 1303 For a ring R.(R) the ring of n x n matrices over R. c E a/@ such that a ..
s + 3. 1))  IA =I mod M . Obviously.(R) = I+M. ..T. b. . ( J ) } (c) By calculating the possibility of the row vectors of A in GL3(Fq).4 6 Show that GL. : F. (e) We consider the field extension 3 C 3.(R) t G L .". is it easy to see that U is a pSylow subgroup of GL3(3. % ( R / Jis surjective. 3: H az is an invertible linear . T.(J). . 2 = diag{i.} is a pSylow subgroup { A = ( a i j )I = { A E M.3. There exsits a E 3. . with kernel ) (d) Show that U = {( l a b 0 1 c 0 0 1 ) a.s.).c E F. Such that Fi3 = ( a ) .
1)(52.Then P/ ker cr ‘v U and JP] I kercr] .1 . a .1 .  P { A = ( a i j ) 3 x 3 E G&(Z/2U) 1 a ( A ) E U C GL3(Z/U)} where cr is the homomorphism G L 3 ( a / 2 u )+ G L 3 ( Z / U ) . Then J 2 = 0 and % = R / J N Z / U . we get an infinite set of integral solutions { ( a i .1 . Thus P is a 5Sylow subgroup of G L 3 ( Z / 2 U ) . and ” a = 2 4 . s 3 . v. say a’ = 2b.2b2 = 1).. = 3 .2y2 = 1. So if we take a. Let a = 2a’+ 1 and b = 2b’. II = 512. (Indiana) Solution. it is easy to see that if (ao. Suppose that (a. Obviously.Since . bo.1) = 2 7 .2y2 = 1 . then b = b. &om the above consideration. b = 2aobo) is another solution of z2. 512 31.bi) 1 i 2 0) of the pell’s equation x 2 . Obviously [ I M 3 ( J ) J= 59 IGL3(Z/U)I = 53(53 . b ) of the Pell’s equation z2 .1)(5 . Hence either both $ and a f 1 are square ’ numbers or a and ’ are square numbers. Then a must be b) odd and b even.1. Let A E GL3(Fq)be the matrix of T. b = f2aobo. = U 1304 Describe an infinite set of integral solutions ( a . bo = 2 and a.1.e.2y2 = 1 (i.2y2 = 1.31. the kernel of the surjective homomorphism G&3(Z/2w) GL3(Z/5Z) 3 is I and + M 3 ( J ) .lbil (i 2 l).. is an integral solution of x 2 . (ao. relative to some base for Fq. ( 3 . 3 .(a) = q3 . 3 .2y2 = 1 . . the order of A in GL3(Fq) q3 . If $ and a ’ 1 are square numbers. bi = 2a. then (al = 2 4 .2y2 = 1 ./Fq. bo) is a solution of z2 .bo) is a solution of z2.47 transformation over Fq. + so Let (GL3(Z/2U)I= 2 7 . By (b). for some integers ao.2y2 = 1. = 2aP1 . ad. Then we have bf2 = So must be a square number. a’ 1 = a. 2 ) is an integral solution of x 2 . + + . is ( f ) Let R = Z/2M and J = U/2U < R . b E Z such that a2 .
2y2 = 1. For any a (f 0) E R. Thus we have proved that R is isomorphic to a subring of a direct product of rings.c. RIP .c.a u(aP1.. if (ao. 1306 Let R be a commutative ring with 1. R not necessarily an integral domain). then ( a = 2ag + 1 . from the solution ( 7 . By (a).Z / p Z . Since p is a prime. If a' and are square numbers. Give example (without proof) of such a ring which is an integral domain but not a p. If R satisfies the a.d. each of which is isomorphic to a/$. the nil radical of R is 0. a'+ 1 = 2bg.1) = 0. a  P For any prime ideal P of R. b = 2 f 2 a o b o ) is an solution z2 . then (ao. for all ideals.. b = k2aobo. the intersection of all prime ideals of R is 0. (ascending chain condition) for finitely generated ideals then show that R satisfies the a.70) of z2 . say a' = ug. Let R be a commutative ring in which u p = a for all a E R. It follows that R has at most p elements.c. and for any a E R. Hence = all prime .2y2 = 1 and a = 2 a i + 1. since UP . u p = a. 5 ) of x 2 . For example. (a) Suppose that R is an integral domain.2y2 = 1. prove that R is isomorphic (as a ring) to (b) In the general case (i.z = 0.. prove that R is isomorphic (as a ring) to a subring of a direct product (not necessarily finite) of rings each of which is isomorphic to Z/pZ.e. we get a solution (99. RIP is an integral domain and for any a E RIP. (Indiana) Solution. (b) Since for any a E R.2y2 = 1.bo) is an solution of ' .i. (Indiana) . that is.bo) is an integer solution of z2 .2y2 = 1. (a) If R is an integral domain.p z . 1305 Let p be a prime. a is a root of the polynomial zP . Char(R) = p and R ZlpZ.48 Remark.c.a. ap' = 1. Conversely. Hence R is a field.
l R is a local ring.c. 1 = 1 + 0 E S . for all ideals.l R and those prime ideals P of R such that P n S = 0. (b) It is well known that there is an onetoone correspondence between the prime ideas of S .c. b E A. P n S # 0 if and only if P A # R here. for all ideals. and let + N be an Rhomomorphism. (b) Show there is a onetoone correspondence between the prime ideals of S . If R satisfies the a. what can you say about the structure of SlR? + + (Indiana) Solution.c. the So every elements in S is invertible in R.c. This is what we need to prove (b). then for any ideal I of R. + 1308 f :M Let I be a nilpotent ideal in a ring R. (c) If A is contained in every maximal ideal of R .49 Solution.l A is the unique maximal ideal of S . Hence R is a Noetherian ring. then S . X [ z ] . Obviously. and it is easy to see that S .l R is isomorphic to RA. but not a p.d.let M and N be Rmodules. For any 1+ a and 1+ b.l R and those prime ideals P of R such that P A # R. (a) Show that S = (1 a I a E A} is a multiplicatively closed subset of R.. So S .c. a . (d) If A is a maximal ideal of R. then A J ( R ) . Hence Jacobson radical of R. 1 is finitely generated. for finitely generated ideals. Let A be an ideal of R. 1307 Let R be a commutative ring. For otherwise. (a) Obviously. It follows that S is a multiplicatively closed subset of R. R satisfies the a.l R by (b).i. Show that if the induced map . (c) If A is contained in every maximal ideal of R. SlR = R. we can construct a strictly ascending chain of finitely generated subideals of I . what is S . i.satisfies the a.c. (1 + a ) ( l + b ) = 1 + (u + b+ab) 6 S.e. the localization of R at the maximal ideal A.the polynomial ring over Z.l R ? (d) If A is a maximal ideal of R.
VP(dt))). we have i) V.(f(t)) + % ( S W and 4 v. Let t be transcendental over F . = 0. So N = f ( M ) and f is surjective. then f is surjective. show that the set of a(t) E F ( t ) satisfying la(t)l 5 1 is a local ring. Let R = { a ( t )1 a ( t )E F ( t ) . Obviously.dt))= V.V p ( f ( t ) ) . Since f :M/IM + N/IN is surjective.1 E F [ t ] . I . Is(t)lPH 1 + m. 4 l IP(t>lP) .(f(t) + d t ) ) L fin(Vp(f(t)>. N / f ( M ) = 1 2 .(f(t). c ( t ) E F [ t ]and * ( / ( b ( t ) . we get an absolute value 1 .. : F ( t ) + I?i by %(O) = 00 and Vp(f(t)) = Ic if 0 # f ( t )= ~ ( t b) t )~c ( t ) .P(t)). 1 is nonarchimedean (If(t) W l P + Suppose F ( t ) is the completion with respect to 1 Ip. . Obviously. ii) V.where b ( t ) .(t). N / f ( M ) = * . (Columbia) Solution. P ( t ) ) = 1 = (. Solution. N/f(M) = I N Hence + f(M)/f(M)= N / f ( M ) . L m={I4t>Ip. Since I maXMt)l. 1309 Let F be a finite field containing 5 elements. 1 (called padic absolute value) on F ( t ) .50  f :M / I M t NIIN (Harvard) is surjective. If is the completion of F ( t ) with respect t o 1 1. We define V. * = I n . f ( M ) + I N = N . Explicitly construct one nonarchimedean absolute value I I on F ( t ) . = 2 . because I is nilpotent.Ia(t)lp5 1) and M = { a ( t )E R I la(t)l < 1).(f(t)) = 00 if and only if f ( t ) = 0. N / f ( M ) = I . Let p = p ( t ) = t . Then. It follows that I . . by defining I f ( t ) l .N / f ( M ) = .
. have to show that m 2 n.].X . . 2 . . . .. and X I . . . .2. = UilY1 * * * U l m Y m fa + + + for any X . are algebraically independent over a . . gj is a sum of monomials in so = 5(5 k=l uijbjk) x k + (terms in X I . ' m aijbjk j=1 = 6ik (i. Since X I . . .of degree 2 2) j=1 ( i = 1 ..X z .X .any For R is a subring of the field F ( t ) and M is an ideal of R.X .l E = F ( t ) is the inverse of a @ ) . .X. . in the polynomial ring a'[X1. . We We can write x. .Y. ] ( StanfonE) Solution. . . k = 1. ) of the ring C [ X 1 . b ( t > P ( t ) l P = l+)lPIP(t)lP>  . . where a i l . a ( t ) . . cannot be generated by fewer than n elements. . In the same way.51 and . . . . E a'. . . + * * * gj for j = 1 .So. . . n ). a j 2 . and f j is a sum of monomials in ' Y 1 . X . 2 .. X 2 . . ..* .. . a ( t )E R\M. Y2.X 2 . 1310 Prove that the ideal generated by X I . a i m E a . we also have y j = bjiXi + bj2X2 + + bjnX. . we have la(t)lP 1 and Ia(t)'lP= 1 where a ( t ) .. x .Y.. of degree two or greater. Suppose that { Y l .. . * * * . . bj. .n).rn. . where b j l .X 2 . of degree two or greater.l E R and a ( t ) is invertible in R. Thus R is a local ring with maximal ideal M . .X . .} is another generating subset of the ideal ( X I .
Prove that A N A1 x A2 for some (nontrivial) rings A1 and A2. " Then I1 and I2 are proper and I1 I2 = A . (b") = (a"b") = 0 and thus A N A/I1 x A/I2 where A/I1 and A/I2 nontrivial. By Chinese Remainder Theorem. = (Indiana) . and let I and J be two (proper) ideals such that every prime ideal of A contains either I or J but no prime ideal contains both I and J . 1312 Define f : [0. Find all z such that f(f(f(f(f(f(z)>>))>2. and since ab E I J C N ( A ) .we have I J N ( A ) (the nil radical of A ) and I + J = A . 22  if 0 5 if 1 5 22 22 < 1.52 1311 Let A be a commutative ring. There exist a E I . there exists an integer n such that (ab)" = an b = 0. 1. 1) + [0. Let 1 1 = (a") and 12 = (b"). Since every prime ideal of A contains either I or J but no prime ideal contains both I and J . b E J such that a + b = 1. we have I1 n 12 = 1 1 I2 = (a") . < 2. ( Stanford) Solution. 1) by f(x)= { 2x. since + 1= (a + b ) 2 n E (a") + (b") = I1 + 12.
) z if and only if z = = &. For any real number a . (2 * ( 2 .[ n .53 Solution. sinnz I n E nV}. (Columbia) Solution. for any z 6 [0. Notice that 2 = (642) = 642 . l).[nu .  0 ) .n[u] .n . Hence f(f(f(f(f(f(4 { n .[u]) . sin n2 0 .[n(u.([nu] n [ u ] ) = = nu . f(f(f(f(f(f(. ( 2 2 ) . By using the orthogonality of the set ( l . Since E [0.n[u]] nu . Define deg f(x) = N where U N or bN # 0. c o s n z . 0 5 k 5 62.{ u } } * * .) = (2 * (2 . Use this result to prove that T is an integral domain which is not a unique factorization domain. it is easy t o see that N f ( z ) = a0 + n=l a.[64 .[u] here. z] if and only if 63z = [64z]. cos nz + b. Since for any real number a and positive integer n.{ u } ] = n(u .[u])] = nu . Then it is clear that f ( z ) = 22 . Show that deg f . [u] denotes the largest integer We denote { u } = a . (2 . sin nz.[2z] = ( 2 2 ) 5 a (Gauss function). g = deg f + deg g.)))))) z = { 2 6 4 = (642).632 = [64z] = [63z + z] (to be an integer) if and only if 1 2 63 Thus f(f(f(f(f(f(z)). we have = n { u } . 1313 Let T be the ring of all real trigonometric polynomials N f(z) = a0 + n=l a. [a]. cos nz + b.[nu]= { n u } . 1).
ao. i. .. 2 and so aNCM = 0. Then we have a N = 0. Hence the units of T are {*1}.if and only if all the coefficients of f(z). a N c M ~ b N d ~ The coefficients of c o s ( N + M ) x and s i n ( N + M ) x in f(z). aN and bo. . bN = 0.e. g(z) = 1. which is contrary to aN or bN # 0.g(x)) = N + M = d e g f ( z ) + d e g g ( z ) . Suppose that deg f(z)= N and deg g(z) = M . .g(x) = 0 can happen only if either f(z)= 0 or g(z) = 0 by the above degree relation. sin nz + c M m=l + cm cosmz + d. then a N . Since (an cos n x + bn sin n z ) (cm cos m x + dm sin m z ) + mn=k c andm .b N C M . then the degree relation implies that d e g f ( z ) = 0 = degg(x). If both of them are zero. Thus we have proved that bN(c$ + d&) + = U N d M = 0. .bncm 2 ) sinkx] .. Since c h d& # 0 . . Let N .g(z) are and a N d M : b N C M respectively.bN f ( z ) = a0 and g(z) = c g + n= 1 an cos nx + b. dM = b N . are zero. d$ = . If f(z). sinmz. CM. b l . Hence deg(f(z). a l . So T is an integral domain. Now f ( z ) .
. Now. Hence B C Avl + Av2 .clznl + c n ~ z . 1 f f i s i n z are irreducible. First. Let p = Char(K) # and n = [ L : K ] .* u. j). it is easy to see that nLIK 0 # ( n ~ / K ( b = nb for any b E K ) .. n ) = 1. be the ~ +  minimal polynomial of a over K and j be the minimal positive integer such that p ] j and c j # 0. a n the set of the conjugates of a. u2. we get TrL. = K [ a ]for some a E L. u 2 . 1314 Let A be a Noetherian integral domain integrally closed in its field of fractions K . c. . * * ~ (l). we claim that the trace function ~ L I K 0. ( Stanfod) Solution. ( k i E B ) . a 2 . v 6 .cj # 0. . Ici = ~ L / K ( ~ u E ) A for any i. 3 has the form lclvl : k2v2 . outline a proof that B is Noetherian. . . Since A is Noetherian.55 Obviously. Let L be a finite separable field extension of K . Then. for any z E B . * knvn . The bilinear function ( 2 . K Let 211. E B .a s i n z). If p = 0 or ( p .&j) = (l)n+j+l * j . and cos z f sin z and 1 f d s i n z are not associates. suppose that p # 0 and p I n. (the elements in L satisfying T T L / ~ ( u ~ = ~ . Hence T is not a unique factorization domain. y) ) # be the dual basis of u l . 1. + Av. . + T r L / ~ ( z y is nondegenerate since ' I ~ L I K 0. Thus ~ L I # 0. Then by using Newton's identities on the elementary symmetric polynomials. Let ~ 1 . For any f(a)E L. Denote a = a 1 . be a basis of L over K contained in B. B is Noetherian. If B is the integral closure of A in L . 2 1 2. . . all the factors cosz f sin%. in T we have cos 22 = (cos z sin z)(cos z . . Since ) L _> K is a finite separable extension. + + Again by the degree relation.Let zn . . Thus cos 2% in T does not have an essentially unique factorization into irreducible elements. be nL/K(f(a)) = i=l C f(ai). . v... .sin z) = (1 a s i n z)(1.  + + + ) + . j for all i. Since xu.u.
Since A is commutative. . a0 a12 nilpotent and n=O c anzn is nilpotent. (Stanford) Solution. . is F ( z ) . 03 It is easy to see that f = 0 " + + . first we get that a0 is nilpotent. then it is easy to see that f = 0.e. and for n 2 1 let R. . If A is Noetherian. that is.. . E I.56 1315 If A is a commutative ring and A[[%]] the ring of formal power series over is A . is a Noetherian ring.. i. blbz . So. Suppose that f = implies that a? = 0. Conversely. that if all the elements ai are nilpotent. be the subring of the ring of polynomials F[x] consisting of polynomials f0 + f i x + fizz + . (b) Show that the field of fractions of R.b. (a) Show that R. Hence 2 n=k+l is nilpotent and so ak+l is nilpotent. then all the elements aa E A are 00 n=O nilpotent.. + f k x k E F [ z ] * such that f1 = fi = . then f is nilpotent. show that if f = c anzn is nilpotent. Let I be the ideal generated by {ai I 0 5 i < m}. * .b. = fn = 0. in its field of fractions.and determine the integral closure of R.  1316 Let F be a field. suppose that A is Noetherian and all the elements ai are nilpotent. (Stanford) . By induction. If I" = 0. + a k Z k Assume that is is nilpotent. = 0 for any b l . all the elements ai are nilpotent.. I is finitely generated and nilpotent. prove the converse. . " Hence f is nilpotent. ak is nilpotent. Since A is commutative Noetherian. a i . bz. ao..
by Let Obviously. E S). R.. (A subring contains 1 by definition.Then + 1. Hence Z s C R. a ( z ) / b ( z )E F [ z ] . E R. p On the other hand. P .57 Solution. q ) = 1 and pl q k = 1 for some I . for any q / p E R. On the other hand. It follows that R = Zs where 1 s={O#PEZ)ER}. Let Z s be the localization of Z at the multiplicatively closed subset S . z is integral over &. For any q / p E as ( q E Z. Then R _>a definition. we may assume that ( p . a ( x ) / b ( z )is integral over F [ z ] . Let R be a subrings of&. is a Noetherian ring by ArtinTate Lemma (or Eakin's Theorem). if a ( x ) / b ( x ) E F ( z ) is integral over R.1 + k * 4 E R. Hence F [ x ] is the integral closure of R.pl+qk .) ( Stanford) Solution.zn+' E R n [ X ] . in its field of fractions. (b') Observe that for any a ( x ) / b ( x )E F ( x ) . is F ( z ) . P P P So p E S and q / p E Zs. Hence F [ x ] is integral over R. = q . (a) Since F [ x ]i s Noetherian and s is finitely generated a &module.. Since F [ x ]is an integrally closed domian. Hence R C Zs. since z is a root of Xn+' . It is clear that the field of fractions of R. k E Z. 1317 Describe all subrings of &. Obviously. S is a multiplicatively closed subset of Z containing 1.
Thus (27s I S is a multiplicatively closed subset of Z} is the set of all subrings of& (S can be choosed as the complement in Z of the union of some prime ideals of a ) . .
( c ) If Card(X) is a prime number and if G acts transitively and contains a t least one transposition then G must be the whole permutation group of X.2 ( Cohmbia) is irreducible and determine its Galois group over &. (g(zl). G. x . ~(2.Z)(x.there exists g E G such that g ( x ) = y. 1) (a) By the defintion of . Hence m 2 n. x 2 . 2) Suppose f E &[z] is irreducible and has degree p .SECTION 4 FIELD AND GALOIS THEORY 1401 1) Let X be a finite set and G a subgroup of the group of permutations of X . (z2. (which interchanges z. (b) If G acts transitively. . Hence is an equivalence relation.  is reflexive and symmetric. Since G acts transitively on X .. we have    ( . . Show that the polynomial    (x2 + 4) . ( 2 1~ i ) g . Since (x.Y # z). If f has exactly p .' E . Solution. Show the following.u(zi))= ~ e . # Y. y 2 . 4z2). Define a relation on X by requiring x y if either x = y or the transposition (2. show that the Galois group of f over & is the symmetric group S. z ) . then all equivalence classes are distinct and contain the same number of elements. a prime number.4)(x2 .YY = ( z . we have n 2 m.16) . So all equivalence classes contain the same number of elements. Similarly.. on p symbols. Thus m = n. ~ ~ ~ .)) are n distinct elements belonging to the equivalence class determined by y1 = g(x1). (b) Let { z l . For any 1 < i 5 n. and }{ y 1 .. Thus { ~ ( z l ) . y E X and leaves all other elements y) fixed) is an element of G .Y)(Y. it is easy to see that is transitive. y m } two equivalence classes of zn be determined by z = x1 and y = y1 respectively. . .2 real roots and 2 complex roots. (a) is a n equivalence relation.
. .y)E G.. 2) Suppose  P f) ( . say n. Identify G = Gal(E/Q) with a permutation group of the set X = { T I . . Hence the restriction of the conjugation to E is an element of G and it is a transposition.1 + . Consider the conjugation automorphism onG. Since Card(X) is a prime. n. Let g(2) = (z2 + 4)2(z2. (z.e. Then.16).16) .y E X . . So Card(X) = m . Thus for any z.)/& into&(rj)/& by sending ri to r j .4)(z2. f(z)= (z2 + 4)2(z2 . Thus the Galois group G of f ( z ) over & is the symmetric group Sp on { T I .l.. and also over Q ( r j ) . This maps f(z) to itself. and n > 1. r p )is a splitting field of f(z) over & ( r i ) . r p } . all the equivalence classes contain the same number of elements. there exists an isomorphism of&(r. on X . by (b). . r P )is a splitting field of f(z) over & contained in a. z y.this isomorphism can be extended to an automorphism q of El&. . r j E X.60 (c) Suppose G acts transitively and contains a t least one transposition. which is generated by all the transpositions.f4. Let r1 and r2 be the two nonreal roots of f(z).and the graph of y = g(z) has the form Fig. Since E =&(rl.. since f(z) is irreducible and f ( r i ) = 0 = f ( r j ) . m = 1 and n = Card(X). = rI(z .4)(x2 . Then q E Gal(E/&) and q(ri) = ~ j which shows G acts transitively .. For any r i . It follows that G is the whole permutation group of X . @[XI. .Ti) i=l in So E = & ( T I . where m is the number of distinct equivalence classes determined by . i.2 is irreducible over &. f2. Thus the conjugation interchanges r1 and r2 = Fl and fixed all other real roots. . By Eisenstein criterion.r p } of the (distinct) roots by q + q1x. The real roots of g(z) are 0.
Use the generator CT : Cp + ( C p ) g of the Galois group of&(Cp) over &. a 2 ( ~ p ) . Hence the Galois group of f ( z ) over & is ST.Fix e.2 has five real roots and two nonreal roots.61 Since I g ( n ) l > 2 for any odd integer n.)g is an automorphism of &(GI over & and G = Gal(&(<. Now for any i. CT : Cp H (&. ~ ~ ~ . . Hint. = ~ } is < p Let H =< C T ~>< G.up’. vi generates a subfield of &(Cp) of degree e over &. C pP . for any i. g is a generator for (Z/pZ)*). Inv(H)/& is a Galois extension and Gal(Inv(H)/&) 21 G / H . a divisor of p .l)/e. Define f 1 . Obviously.it is easy to see that f(z)= g(z) .. Let Cp be a primitive pth root of unity and g a primitive root ( mod p ) (i. CT’. = @(cP): InvH] = IHI = f and [Inv(H) : &] = e .aP’ = 11. Since H is normal in G.. Then J H J f.1 and put f = ( p . { a ( < p ) .qa = C(<p)gej+i j =O Show that.e. Since g is a primitive root ( mod p ) . 1402 Let p be an odd prime.)/&) =< CT >= {c.1a(base)for&(<p)/&. (Columbia) Solution. .
that z is is + ... (a) Let A .AB 6 K . Hence j=O n (z q j ) is the minimal polynomial for qi over Q (for any  i).62 and f I j=O fI j=1 f1 . and let H denote the subgroup of G consisting of the substitutions 2 + 2 b with b E K . B E K [ X ] . Put G = Aut(K(z)/K). Inv(H) =@(qi) for any i. gcd(A.e. Show in succession that the polynomial A . It follows that 71.B) = 1. 1403 Let K be a field and z be an element of an extension of K such that 2 is transcendental over K .1 are distinct. e. i.. Thus qi generates a subfield of &(&) of degree e . is a primitive element for Inv(H)/Q.yB E K ( y ) [ X ] not 0. and put y = A(z)/B(z).
bc # 0.z)qZ+l.y B = 0 in K ( y ) [ X ] . I2 = {diag(a. which is contrary to the assumption. ( CoZumbia) Solution. Since A(z) . (c) By (b). Conclude that n.y B # 0.y B is irreducible in K ( y ) [ X ]and that .e. If y is algebraic over K . Find the fixed field of H .that y is transendental over K . polynomial of z over K ( y ) is the monic polynomial which is a multiple in K ( y ) of A . ] Thus A . B ) = 1. since gcd(A.63 algebraic over K ( y ) . then ordG = q 3 . Thus A .Suppose A .12.y B ) = max{degA. .y B is irreducible in A K [ y ] [ X= K [ X ] [ y ] . [ K ( z ): K ( y ) ]= max{deg A. then G is infinite and the fixed field of G is K . which is contrary to A B 6 K . Again.a) 1 a E K*}.y B ( z ) = 0.. deg B}. Find the fixed field of H . that A .y B is irreducible in K ( y ) [ X ] Thus the minimal . (b) Show that the elements of G are given by the fractional linear substitutions z +a z + b/cz d with a .yB and (c [ K ( z ): K ( y ) ]= deg(A .h(z) is a Galois extension of a given field L with lFq C L C Fq(x)if and only if L 2 l F q ( z ) . Hence y is transendental over K . c. Then It follows that degA = 0 and deg B = deg(A) = 0. y = A ( z ) / B ( z )is a generator of K ( z ) / K (i.bc # 0. (d) Show that when K = lFq and put + z (zq2 . E K [ X ]such that A S + B T = 1 T in K [ X ] . B ) = 1 in K [ X ] K ( y ) [ X ] ) .deg B } . . it is easy to see that G N G L z ( K ) / K *. (a) Since gcd(A. z is algebraic over K ( y ) . (c) Show that when K is infinite. deg B } = 1. z must be algebraic over K . there exist S.where K* .z)q+l/(z' . or if and only if y has the form az+ b/cz +d. d E K and ad .q and the fixed field of G is n. (b) By (a). K ( y ) = K ( z ) )if and only if max{ deg A. ad .h(z). b.
since H is also infinite. it is routine to check g(z) = g((zq2 . so is G. .bc # 0) in G.4.X)q2g. Similarly.x)q+l/(XP .Z)Q+1/(2~2)q2+1) 2. Thus we have Inv(G) = K . Then c and [F.q . (xq2 X)  1 (XP . .q . On the other hand.64 If K is infinite. (ad . Then gcd(A. which is contrary to the facts that G AutK(y)K(z) and G is infinite.X)q2+1 = ((xqx))q+l. K C Inv(G).X)n”n A = (1 + XP’ + X(P1)2+ . if there exists some y = f(z)/g(z)E Inv(G)\K. . we have (XP2 . (d) Suppose K = Fq. Hence AutK(y)K(z) is finite. B ) = 1 and max{deg A. On the other hand. Hence we have [Fq(z): Fq(z)] = q3 . For any g :2 H az + b/cz + d . K ( z ) / K ( y )is a finite dimensional simple extension. we have Inv(H) = K . . Thus Inv(G) = Fq(z) and Fq(z) is a Galois extension of a given field L with FP C L C F. deg B } = q3 . = Hence F ( z ) c Inv(G).(z) : Inv(G)] = q3 . Obviously. + X(Q1)Q)q+l B = (Xq.+ x ( n .1 1 2 + . + X(PQ?)q+l n1 (XP . .X)+P Denote and (1 + x . then by (a).(z) if and only if L _> F ( z ) . the fixed subfield of G.
E/Inv(G) is a Galois extension with Galois group G and [E : Inv(G)] = 2n.(z') = F q ( z q .(Y) = CY and T ( Y ) Y . . (a) Since Y is a generator of E over C. (c) Obviously.Y") E F[X]. Y is a root of X2" . (a) Show that there are unique automorphisms u and r of El@such that a ( Y ) = CY and T(Y) = Y .[X] and let z' = zq . G =< u . we have [Fq(z)Inv(H)] = q since IHI = q. Since u ( Z ) = u(Yn + Y") = Z and T(Z) = 2. Since ord(u) = n. It is readily verified that X2" ZX+ 1 is irreducible over F (Z is transendental over C and X2" .2 . 1404 Let E = C(Y) with Y an indeterminate. >rx D.y')(x .(Y') * (X  y1yl) . (b) Since X2" .X : q and B = 1 in F.. Let A = X . Then z' E Inv(H) and by (a).for example.z E Fq(z). the Dihedral group of "' ~ order 2n.Y)(X . Obviously. Hence c and T are automorphisms of E/C. ( X .the Dihedral group of order 2n. (b) Show that Y is a root of a polynomial of degree 2n with coefficients in < F. ord(7) = 2 and Tu = uT in G. = (T" = 1 = r 2 .65 Similarly. we have Z E Inv(G) and F = C(Z) Inv(G).l .. (c) Show that EIF is a Galois extension with Galois group G.Y")(X" .l . and that the subgroup G of Aut(E) which these generate is isomorphic to D.C Y ) . F = C ( 2 ) with Z = Y" + Y".ZX 1 since + + + x2n  zx + 1 = ( X .there are unique homomorphisms u and T of E to E over C such that .C"'y) (X. using 1403).z). x + 1 = X2" . (Columbia) Solution.Z X 1 is irreducible in C[Z][X].ZX I. Thus = Inv(H) = F.(Y"+Y") + 1 = (X" .. E is a splitting field of X2" . [Fq(z) : Fq(z')] q. and = earriIn.
there are exactly 3 subfields lying strictly between K and F which correspond to the three proper .1) = z It is easy to see that ( z . Let (x11)2 5. Consider the elements (T.1)’ &. Then&(q) 2 K F . On the other hand. K C_ F is a Galois extension and [F : K ] = IGI = 4.e. (Indiana) Solution. (c) How many subfields lie strictly between K and F ? How many of these are Galois over K ? Justify your answers. (b) If K is fixed field of G in F . Hence&(q) C F is a Galois extension and [ F : &(71)] = 4. Then f ( t ) is irreducible in & ( q ) [ t and F is a splitting field of f ( t ) since ] in F [ t ] .7 in A u b ( F ) (i..qt2 + 1€&(V)[t]. 2.1)2 + E K . 1405 Let F = &(z) be the field of rational polynomials in one variable x over & (i. (a) Since ( ~ ’ ( z ) = z. r 2 = 1 and UT = T U . find a finite subset S of F such that K = &(S).. It follows that (c) By the fundamental theorem of Galois theory. the quotient field of the polynomial ring&[z]). + f ( t )= t4 .1 (a) Find the subgroup G (of Aub(F)) generated by (T and 7.x and ~ ( . Hence where K 4 is the Klein 4group.z and = 7(z) = A. ?(z) = z and T ( T ( Z ) = = ( T T ( z ) .1) = 1 .e. the fixed field of G in F . It follows that E / F is Galois and [E : F ] = 2n. Hence we have F = Inv(G) and E / F is Galois with Galois group G. Denote 7 = (z . field automorphisms of F ) given by ( ~ ( z ) 2 .we have (T’ = 1.66 in E . (b) Obviously U ( Z .
So + . It is easy to see that r2 = 1. Hence. (Indiana) Solution. Let G be the subgroup of Aub(&(t)) generated by u and T . Since all the subgroups of K4 are normal. u4 = 1 and isomorphic to the Dihedral group (b) Since D4. :) + & ( t 2 + $) of is irreducible over & (t2 $) and Q ( t ) is a splitting field over separable polynomial f(z). all these 3 subfields are normal over K (finite dimensional and separable). (a) Obviously. = (2.) (.f . It follows that G is the fixed field of H . CT3 ( t ) = and a4(t) = T 3 (E) 3 t. and obviously [&(t) : InvH] = (HI 4. u and the Let be elements of Autq(&(t)) given by u ( t )= and ~ ( t=) t. T(T = C3T. = On the other hand.t ) ( z+ t ) (x z .67 subgroups of G K4. (b) Let H be the subgroup of G generated by u2 and T . Find a E @(t) such that & ( a )is the fixed field of H (Justify your answer). (a) Identify G. field of quotients of the polynomial ring @[t]. Thus all these 3 subfields are Galois over K.t 2 ) (2 .F ) 1 = ( . T 2 ( t ) = t . 1406 Consider &(t). u 2 ( t ) u = = .
[& + + + . It follows that that is. then f ( z ) has a factor with degree one in that is.rbe the other two roots of f ( 2 ) = z3 . Since f(2) = x3 . b) Let . the rational roots must be integral factors of 1.32 1 is reducible over &.68 is a Galois extension and [&(t) :& (t2 + $)I = 4. Hence /3 y = a and . ' c) Find a generator of the Galois group Gal(K/&).32 + 1 in a splitting field of f(x) over K. But fl are not roots of f(z).B. f(x) has a root in &. b) Prove that K/& is Galois. a) If f(z)= x3 . Thus f(x) is irreducible over &. let K = & ( a )where a is a root . (Indiana) + Solution. crPy = 1 and the discriminant A of f(2) is 34.2. f ( ) = x3 . of. Obviously.3x 1. Hint. cr + p + 7 = 0. This is a contradiction.3a: 1 is monic. Consider a . InvH =& (t2+ $) 1407 Let & denote the field of rational numbers. a) Prove that f(x) is irreducible over &.I.
(b) dimK K[x] = dimK K ( z ) = 00.zn & + anlxnl + + a12 + ' . 1 a E K } is linearly dependent over K . Let f E K ( z ) ..69 It follows that (Y. . Hence K / Q is Galois since Char(&) = 0.2 are the roots of f(x). (T: + A K=&(a)+K ak+(Y22 is a generator of Gal(K/&). So K is a splitting field of f ( x ) . E K such that + n Hence and i#i which is not true.(Y = a2 . . = a2 . K [ z ]and K ( z ) have what dimensions? (c) Let G denote the additive group of K. (We may check a2 . (a) Show that the rational functions f a = ( a E K ) are linearly independent over K. ( Y .K and some distinct eleE ments a l . Assume that K is infinite. it is easy to see that the Galois group Gal(K/&) F A3. (a) Suppose that {fa = There exist some nonzero elements ( Y ~ . .* a0 E K[z]. . . c) From a) and b).2 is a root of f ( x ) directly by using the hint). (Columbia) Solution. (c> Suppose f ( x ) = a.. u2. (b) As Kmodules. Show that the Gorbit of f spans a finite dimensional Kmodule if and only if f E K[z]. acting on K ( x ) by a E G sending z to z a.(Y 2 and . Thus { f a = I a 6 K } is linearly independent. a. . 1408 Let K be a field and 3: an indeterminate. .
. a n + l in K (note that K is infinite). a.. { f ( z + a ) 1 a E K } spans a finite dimensional Kmodule. .+1 . +an)(ai E K ) span the subspace < {f(x+u) I u E K } >. . " . . We write f(x) = g ( z ) / h ( x ) where g ( z ) .+1 a: a.. . Hence = P l f ( z + a l )+ P a f ( z + a 2 ) + ... . + P n + l f ( 2 + ~ 1 n + l ) . Then for any a E K .*f ./3n+l)K n t l to be the solution of the E equation system + + + + Then it is clear that i 1 1 a2  0 . . For any a E G.. there exist PI..01. ..P 2 ..f(x .. 1 ant1 a1 . . . = P ~ ( Z + $ I+)P~ ( x + a 2 ) " z and also (Z +. Let f ( x + a l ) . and the Gorbit o f f . . .f ( z + a 2 ) .f ( z a2).'.h ( z ) E K [ z ]and ( g ( z ) h ( z ) )= 1. suppose f ( z ) E K ( z ) .+a) This shows that { f(x + a i ) 1 1 5 i 5 n + 1) spans the subspace generated by the Gorbit of f . a. . f ( x an+l)spans the subspace generated by the Gorbit of f . . . f(. a.70 We claim that for any n 1 distinct elements a l . . a. .Pn in K such that f(x+a) = 9(" +a) h(x n + u) g(x+ai) = i=lP i y n h(x + ai) . f ( x a l ) . . a 2 . + Pn+l(z + Qn+1li for 1 5 i < n. . .)a =P ~ ( Z + alli + /32(z + a2)a + . we take (. . On the other hand.+Pn+l(~+an+l)~ + ./32.*. .. ..
Thus all the irreducible factors of f ( z ) over E are of the same degree. is a root of g z . Let a and a’ be roots of e(z) and e‘(z) in some extension field of E . It follows that Now. we can write E = F ( P ) .Show that all the irreducible factors of f ( z ) over E are of the same degree. we have h(z +. Then. ( c ozum bia) Solution. q(P) is a root of q ( q ( z ) ) = g(z). Hence 7 : F ( a ) + F ( a ’ ) can be extended to an isomorphism 7 of E(a) onto E(a’). u(e(z)) is an irreducible factor of f ( z ) over E . CY and a’ are roots of f(z).h ( z + u)) = 1. = we must have (vlE)(e(z)) = e’(z) and dege(z) = dege’(z). Since f ( z ) E F [ z ] . Then E(a) = F ( a ) ( P )and E(a’) = F ( a ’ ) ( P ) are splitting fields of g(z) over F ( a ) and F(a’) respectively. B () a separable irreducible polynomial over F . which is irreducible over F . For any (T E G = Gal(E/F). Since E / F is Galois. where . Then. Since K is infinite. Obviously E is a splitting field of g(z). for any u E G. we still denote u to be the isomorphism E[z] t E[z] which extends (T on E and maps x to z. Suppose e’(z) is another monk irreducible factor over E. 7 : E ( a ) + E(a’) is an isomorphism and ~ ( a )a’. a(f(z)) = f(z).)I i=l n n h ( z +(. an easy discussion in a splitting field of h ( z ) over F will lead to degh(z) = 0.a) for any a E F . since a and a’ are roots of the monic irreducible polynomial e(z) and e’(z) over E respectively. + . We prove in the following that all monic irreducible factors of f ( z ) over E arise in this way. Note that y ( P ) may not be P.Let e(z) be a monic irreducible factor of f(z) over E.71 Since (g(z + a ) . but :E E is in G. Hence we have an isomorphism 7 : F ( a ) + F(a’) which sends a to a (u E F ) and CY to a’. Thus we have 1409 Let E be a finite Galois extension of F and let f(z) be an irreducible polynomial in F [ z ] .
Hint. we obtain n .1 over Zg. . Comparing the coefficients of the term of degree n . + + + + in K [ t ] . u = 0 and there exist integers v and w such that w . Since p .u .c in K [ t ]is either irreducible or splits completely into p linear factors over K. Show that the polynomial tP .u .1.t c = f ( t ) .i2) * * * (t . Then for any m EZ. u = (v . 1 5 n 5 p . First prove that the zeroes of t62 .u .in). in E K . the prime field of K . g ( t ) in K [ t ]where f ( t ) is a monic polynomial of degree n. Hint. If u is a root of t P . Suppose that t p . n wp) u = v(n . Obviously.t .1 in F . Show that [ F :&I = 3. Thus we have tptc= (tum) +.t . Let E be a splitting field of tP . f ( t ) = (t .1)' = 62t6' = 2t61 # 0.. u il + i2 .1 form a cyclic group G of order 62. (b) Let G be all the zeroes of t62 . u)E K .c and let u E E be a root of this polynomial.72 1410 (a) Let K be a field of characteristic p > 0. [ F :as]2 3. .t .il)(t . (Indiana) S o htion.c then so is u+ 1. So we have n . n w p = 1. (b) Let F be the splitting field of the polynomial t62 . u E K . G is a subgroup of F" and G is cyclic of order 6 2 since (t6' It follows that .1.
If [L : KP] is finite. X Now for any finite field extension K I L . Hence. . when decomposed in nZ[a:]. Show that every finite field extension of L must in fact have degree equal to a power of 2. Let GI be its unique subgroup of order 62. and so K' = 1. Then f(a:) has no real root since f(a:) is irreducible in L [ z ] . Since L has no proper odd dimensional extension field. the subfield of 1R of numbers which are algebraic over &. For this purpose. The set {a:" J a: E K } is a subfield of K that is denoted by KP (no proof required). 1412 Let K be a field of characteristic p # 0. K 5 a' is of even degree. (b) Show that such a field L actually exists. & C_ L a'. Then all the elements of G' satisfies t6' . it is in fact a power of 2. Let f(a:) be the minimal polynomial of o over L.1 = 124. we may assume that K is Galois over L. By (a).1.1. then [ K : K'] = 2" and [K' : L] = m. (b) Let L be the field of real algebraic numbers. f ( z ) . If K' is the corresponding subfield of K I L . (a) Let K / L ( K C a ' ) be a finite field extension. We have to show [ K : L] is a power of 2. = 1411 (a) Suppose you are given a field L . So t6' . (Indiana) Solution. that is. So. (a) Let L be an intermediate field between Kp and K . By Sylow's Theorem. Then E* is a cyclic group of order 53 .73 Let E be a extension field of Z5 such that [E : Z5]= 3. [ K : L] = [ L ( a ): L] = d e g f ( s ) is even. Then & 2 L & ( and L/& is algebraic. G has a subgroup H of order 2n. Let G = GalK/L and JGJ= 2n . such that L/& is algebraic and every finite field extension K / L .1 splits in E and E is a splitting field of t62 . is a product of irreducible polynomials of degree 2. Thus we have E N F and [F :as] 3. and [K : L] = 2n. m where m is odd. prove that it is a power of p . there exists some element a: E K such that K = L ( o ) by Primitive Element Theorem. we must have m = 1. K a'.
and bi @ KP(bl. (b) If KP # K .. Since Bi is apindependent subset. there exist pindependent subsets.) : L] n = n [ K P ( b l . Without loss of generality. ... K contains a maximal pindependent subset. . . By Zorn’s Lemma. .. For example. b... .74 (b) A subset B of K is called pindependent if for any finite set bl. . then K contains a maximal pindependent subset B.. Since .. bi1) = KP when i = 1).bil). * .. .*. . of distinct elements of B .. there exists some i. . there exists a finite set of elements {bl.. .. (Indiana) Solution... * * .bp is irreducible in KP(b1.bn) L.. (c) Prove that the set B of part (b) satisfies K p ( B ) = K . . b i : KP(bl. B = { b } is a pindependent subset of K .l)] ) i=l = p”. we can assume that bi $ KP(bl. Then for any finite set { b l . Now suppose that {Bi I i E I} is a chain of pindependent subsets of K .b.. .bzl)] = p. b 2 ...b.. : It follows that B is pindependent... if 6 E K\Kp..b2. .b. (a) If [L : KP] is finite. b m )K P ) =pm. Hence [KP(bl. . tP . [ K P ( b l . Prove that if KP # K .} of distinct elements of B . . bz) : KP(b1.. b m } C Bi.. b2. . . .b2. .bil)[t].bn} such that = KP(b1..+ . . It follows that [L: KP] 1 [KP(bl. such that { b l .. Let B = U Bi. b 2 . b 2 . i€I .bi1) for any ! 1 5 i 5 n (KP(b1.
.*"l = p"* .bz. Then IK(a)l = prm and [ K ( a ): K ]= m. Since m I n.b. .p. then and [KP(bl. the independency of B U { b } contradicts the maximality of B . Let b E K\Kp(B). b 2 .... Let (Y be a root of the irreducible polynomial f ( t ) in some extension field of K. Then (El = prnand [E : K ] = n..t over K . . .b... Thus we have K P ( B ) = K .t . Since pp'" .b.b. If m is an integer which divides n and f(t) E K [ t ]is an irreducible polynomial of degree m. (Indiana) . for any distinct elements b l .) * 1 . . E contains a subfield L such that K & L E and L N K ( a ) . that is.. with E / F Galois and [K : L] = p . we have f ( t ) I ( t P r n .* . Hence there exists an element p E L C_ E such that f(P) = 0. . .. . bm) : KP] = p m .75 (c) Let B be a maximal pindependent subset of K . . then {bl. 1413 Let K be a finite field with pr elements ( p a prime) and n be a positive integer. .p = 0. b m } . 1414 Let K I F be a finite extension of fields and let L and E be intermediate fields. [KP(bl..b. . CB and we still : KP] : KP(b1. Let E be a splitting field of t p r n . . bml)] = [KP(bl.. a prime. show that f divides t P r n . in B U { b } .. The reason is that. b = b. Then B U { b } is pindependent.* and if b E {bl. . Prove that if p does not divide [E : F ] then E L. b 2 . f ( t ) is the minimal polynomial of p. Suppose K p ( B ) c K . .t ) . (Indiana) Solution.) ..}. b.* . if b @ { b l .. have [KP(bl.l} say.l) : KP] Since B C B U { b } ..
we have N_ G a l ( E . Hence E L is Galois over L. Ki+l is (0 €6' 10" E Ki for some n > 0). It is clear that i=l u Ki is a subfield of a'. Prove that deg(f) 2 5. Let f(z)be a separable irreducible polynomial over F such that E is its splitting field. Then E L = L(E) is a splitting field of f ( z ) over L. a p = IGal(E. contrary to the assumption. For any c E G a l ( E . L and n E)I [K : L] = p . L. If i the smallest subfield of @ containing the set 2 0. 1415 Let Ki be the subfields of a' defined as follows: KO = &. Then E . Let a E E be a root of f(z). L/L)I = IGal(E/L dividers IGal(E/F)I = [E : F ] . L/L) Gal(E/L n E ) . (1) Since KO C_ K1 C_ C_ Ki C_ Ki+l C_ M * * a is a chain of subfields of a'. (2) Let f(z) E K [ z ]be irreducible. since L Now suppose E prime. Let E * L = E(L) L ( E ) be the composite of E and L in K . L = K . Then E = F ( a ) and E . Thus we have E C L. + it is clear that g (E ~ Aut(E/L n E ) . L/L) = Gal(L(a)/L). It follows that C E . (Indiana) Solution. .76 Solution. Let i=O (1) Prove K is a field. L = L(E) = L(a). And further.
Let E = F(cr). where g ( n ) = p"l. Let f E K [ t ]be an irreducible polynomial of degree n. Suppose d e g f ( z ) 5 4.1 (Indiana) Solution. : Thus [Fp(a) F] is the smallest positive integer m such that : E Fp. if aPml I . cubic and quartic equations and Ki+l 1 (0 E(C 10" E K . for some n > O}. Contradicts the irreducibility of f(z). If m and n are coprime then show that f remains irreducible in F [t]. Let f(x) E K [ z ] be an irreducible polynomial with d e g f ( z ) > 1.Then IFp(a)I = p" and upn' = 1. a d n ) E Fp. Let a be an algebraic element in K . E Fp. Hence deg f ( z ) 2 5. Hence z over Fp and a E E . the field with p elements.x. (Indiana) Solution. 1416 Let K be an extension field of Fp. of . Let n = [ F p ( a ) Fp]. then On the other hand.77 (2) (Remark: d e g f ( z ) may be 1). P. Suppose that f(t) is reducible in F[t] let f ( t ) = g ( t ) .]([EpP]= m. By the formulas for the roots of quadratic.hence in K [ z ] . 1417 Let F 2 K be a field extension of finite degree m. for some positive integer m. Let E be a splitting field of F p ( a )C E . Then [E : Fp] = m and Fp n = [F. 1 5 k < n. Prove that [Fp(a) Fp] is the smaIlest positive integer : m such that E Fp. There exsits some i such that f(x) E Ki[x]. Since ( ~ ( g ( ~ ) ) P = : l apnl = 1. zPm  so u is a root of m zp .(&4)Pl = 1.(u) : F.h ( t in F[t] and ) where g ( t ) is a irreducible polynomial in F[t] degree k. f(z)splits in Ki+3[z].
n ) = 1 and 1 5 Ic < n. (Stanford) Solution. It is easy to see that IGal(&(td)/&)I = 4(n) and Gal(&(z.It follows that n I Ic .78 g ( t ) is irreducible in where cr is a root of g ( t ) in some extension field of F . Thus f ( t ) is irreducible in F [ t ] . for any positive integer n. Then [E : F ] = Ic since F [ t ] .z . E = & ( z 1 7 ) . d(n)is the Euler &function. is an nth primitive root of the unit. (Stanford) Solution. is an nth primitive root of the unit. It is easy to see that IGal(&(zi)/&)I = +(n) and Gal(&(z.)/&) E Aut( ) where G is the cyclic group of order n. [E : K ] = [E : K(cr)]. When n = 2" and m 2 3. the cyclotomic field & ( z n ) over & is a Galois extension and [ & ( z n ) : &] = d(n) where z .Zz @ . When n is prime. : Taking m = 7. then E is a Galois extension of & with Gal(E/&) cyclic of order 16. Aut(G) is cyclic of order n .)/&) N Aut(G) where G is the cyclic group of order n.So [E : K ] = [E : F ] [ F : K ] = Ic . . which contradicts ( m . the cyclotomic field &(zn) over Q is a Galois extension and [&(zn): Q]= +(n) where z. [K(a!) K ]= [E : K(cr)]. As in 1418. . then & E is a cyclic extension of order 32. 1419 Find a Galois extension E over & with Gal(E/&) cyclic of order 32. So if we take n = 17. On the other hand.Z2. Z p . if we take E = Inv(&). since cr is a root of f ( t ) and f ( t ) is irreducible in K [ t ] . 4(n) is the Euler $fur Ition.m . By the Fundamental Theorem of Galois Theory. For any positive integer n.1. : n. it is well known that Aut(G) . then & C E is a Galois extension and Gal(E/&) 1 .m.. . 1418 Find a Galois extension E over & with Gal(E/&) cyclic of order 16.
z).l } is a base for E / F .an(z)} in Mn(E). is the minimal polynomial of z over F.. G = Gal(E/F) and a E E .. . ~ .uz(. [E : F ] = n and Then .. and u 1 ( z ) .z n . (E M n ( F ) ) .79 1420 Let E / F be a finite Galois extension. M . Let z be a primitive element of E / F ..(z) = a z . . g n ( ( z )are distinct.a. Consider the Flinear map M a : E + E.znl}. . : E + E .. is similar to diag{ul(z).. 3c. a has the form since { l .1. the matrix of M .. Obviously.) = C g ( a ) where varies over G. M . for any a E E .+ . Anyway.. . relative to the base (1. Z ( z ) .@) in E. (Columbia) Solution. f(z) is also the minimal polynomial of the Flinear map M . we have  n i=l . z . (z) = z .say. . To prove that it suffices to prove that for any 1 5 k 5 n . Now. .. Show that its trace is given by Trp(M. Since f(z)has distinct roots C T ~ ( Z )g.
n This completes the proof.1. .and for any 1 5 i 5 n .
Part I1 Topology .
.
l). It means that A U B is connected. If A and B are path connected. there is a point of B which belongs to U .a3 SECTION 1 PO IN T S ET TOPOLOGY 2101 Let A and B be connected subspaces of a topological space X. Since A is connected. Since f is continuous. Let A = ((0. there exists a neighborhood of zo.O)) C R2 and 1 B = {(z.sin ) I 0 < z 5 1). Prove that there are open sets U c X and V c Y such that . Let zo E AnB. ~ We have f(zo) = 1. Without loss of generality. such that U c f'(l). say U . Let f be any continuous map from A U B to So = {1. need A U B be path connected? (Indiana) Solution. By the same reason. 2102 Suppose that A and B are compact subspaces of spaces X and Y respectively. Prove that A U B is connected. and that N is an open neighborhood of AxBcXxY. The following example shows that if A and B are path connected then A u B needs not to be path connected. Therefore we have B c f'(l). f ( must be constant. f l is also constant. we may assume ~ that A c f'(l). Hence f is not surjective. such that A n B # 8. But since zo E B. 3 : Then A and B are path connected and A n B = A . but A U B is not path connected.
We denote by X" the onepoint compactification of X. If X is compact.(z). ' Then by the Urysohn Lemma there exists a continuous function X + [0. there is asubcover {VYi(z). Does there exist a continuous function f : X [0. 2103 j= 1 Let X be a locally compact Hausdorff space. ( z ) c X and V. . It is easy to see that A and F = B u{m} are two disjoint closed subsets of X . For any y E B.84 Ax B cU x V cN. z E A } is an open cover of A . . Hence we may assume that X is noncompact. y) belongs to A x B and N is an open neighborhood of A x B c X x Y .y E B } covers B. there exist open sets U .il . ... Let X* = X U {m}. X is compact ' and Hausdorff. Since B is compact. U U U ( z j ) and V = j=1 q n V(zj). q } such that A Let U = sets of j=1 u U(zj). Therefore the family of open sets {Vy(z). Let A and B be disjoint subsets of X . j = l ..(z). It is easy to see that U and V are open q X and Y respectively and that A x B c U x V c N . then X is normal and the existence o f f is obvious.since (z. with A compact and B closed. .1] such that f l = 0 and f l = l? ~ ~ + ( Cincinnati) Solution.(4 cN. Since X is locally compact and Hausdorff. Let' z be a point of A .(z) x V. and I) n It is obvious that U ( z ) and V(z) are open i=l sets of X and Y respectively and that {z} x B i=l c U ( z )x V(z) cN. there exists asubcover { U ( z i ) . i=l P Let U ( z ) = fi Uyi(z) V(z) = 0 V.1] ' F: . (Indiana) Solution.(z) c Y such that ( Z C . . Since A is also compact. Y )E V. . . p } such that B = c U V. and consequently is also a normal space. c On the other hand. { U ( z ) .
only the correct answers to the question asked.satisfies = 0 and f l = 1.  2105 (a) Define quotient map.y.t’) if and only if { x = x’ and t = t’ = 0 or y = y’ and t = t’ = 1. By the quotient map ?r : X + X / we mean the map which assigns to x E X the equivalence   . Define by X / the space of equivalence classes under . Therefore. and S’ * So is homeomorphic to the unit sphere S2. (Indiana) Solution. If X and Y are topological spaces. Y is Hausdorff and f : X + Y is continuous and onto.1] by collapsing X x (0) and X x (1) to two points p and q respectively. (c) Show that i f f satisfies the condition of (b) then f is a quotient map. then f is a closed map.1]/ obtained from the cylinder X x [0. (b) Generally. (a) Let X be a topological space and be an equivalence relation on X. f. the restriction o f f on X .t) is equivalent to (d. (a) So * So is homeomorphic to the unit circle S’. (Indiana) Solution.85 such that f l = 0 and ~ the requirements that  f l = 1. (b) Show that if X is compact. are required for this problem. X * So is homeomorphic to the quotient space X x [0. ~ ~ 2104  No proofs. y. the join of X and Y is the quotient space * Y = ( X x Y x [ O ) 1])/ x ) where (z. What space ’ are they? (b) Describe X * So for a general space X . (a) So * So and S * So are homeomorphic to familiar spaces.
(c) Let be the equivalence relation on X associated to the map f. Let V = Y . we see that f f'(v) = x . Since X is compact. Since Y is Hausdorff. It follows from the continuity of f that f ( A ) is a compact subset of Y .       N 2106 Let f : X Y be a continuous function from a space X to a Hausdorff space Y .U .  (b) Let X = R and Y = S1. Since f is a continuous function. . It is clear that f = i o T . If Y is homeomorphic to X / under the map i : X/ + Y and f = i o T then we call f a quotient map. Under this topology.f ( W ) . Since W c f'(f(W)).1 map. (Indiana) Solution. By the result of (b). More generally. there is naturally associated an equivalence relation on X such that z1 5 2 if and only if f(x1) = f(z2). is obvious that It fyc)= {n I 72 EX}. Take C = (1) E S1. Then V is an open neighborhood of C in Y . f ( W ) is a compact set of Y . therefore. and.1 map from the compact space X / to the Hausdorff space Y . So f is a quotient map. is a homeomorphism. f ( A ) is closed in Y . Denote by [z] the equivalence class containing 2. W is compact too. and let U be an open neighborhood of f'(C) in X. (b) Give a counterexample to show that if X is not compact. : X + Y is the continuous function defined f by j ( t ) = earit for t E R. (a) Let W = X .fl(f(w))c x w = u. A is compact too. Hence i is a continuous 1 . Let C be a closed subspace of Y . The quotient space X / may be topologized by defining a subset U c X/ to be open if and only if r . then W is closed in X . (a) Prove that if X is compact then there is an open neighborhood V of C in Y such that f'(V) c U .' ( U ) is open in X . Since f is onto. and consequently is a closed set of Y because Y is a Hausdorff space. Then we define a map i : X/ 4 Y by i([x]) = f ( x ) . then there need not be such a neighborhood V . Since X is compact. i is obviously a 1 . Hence f is a closed map. r becomes a continuous map. if f : X + Y is a continuous map. the quotient space X / is compact. (b) Let A be a closed subset of X .86 class containing x.
(a) Show that the quotient space Y obtained by collapsing A t o a point is normal. Let U and V be two nonernpty closed sets in Y such that U n V = 0. Thus we see that R ( W ~ ) and R ( W ~ ) two disjoint open sets in Y and contain U and V respectively. are If yo E U (or V). (a) Let R : X Y be the identification map and yo E Y be the point which A collapses to. 2107 Let X be a normal topological space and A c X a closed subspace.=(+n. it is clear that there exist two disjoint open sets W1 and Wz in X containing R'(U) and R'(V) respectively such that WinA = 8 for i equal to 1 and 2. n  one can easily prove that there does not exist such a neighborhood V. If yo $ U U V. we only need to take the sets W1 and Wz without the restrictions that Wi n A = 0.n+). Then R'(U) and R . Let X be a connected space. (b) Let X be a regular space which is not a normal space. Then the quotient space Y obtained by collapsing A to a point yo is not regular. n n Since nw lim lUnl = 2 = 0. (b) Does this result hold if normality is replaced with regularity? (Indiana) Solution.' ( V ) are two nonempty disjoint closed sets in X . 2108 Let p : E B be a covering map with E locally path connected and simply connected.' ( C ) n €I in X . because one can prove that the point yo and the closed set R ( B )in Y cannot be separated by disjoint open sets in f Y. where 1 1 U. by the normality of X . let f : X +B be a continuous map. then.87 Let U = U Un be the open neighborhood of f . f . It means that there exist two disjoint closed sets A and B in X such that they cannot be separated by disjoint open sets in X .
X E A} is an open covering of X .  < 00. (Indiana) Solution. Thus. Then eo E p . i=l Choose another elementary neighborhood V(z) of z such that v ( ~cc ) np(w).(z). Therefore. Since p is an open map.' ( W ( z ) ) such that Zi E w.(z) n U i containing Zi. It is obvious that each & x is open and R(z) n (x) = 0 for () i # j . Prove that 2 is compact..88 and f l . we see that e l E pl(b0). 2109 Let p : X +X be an nsheeted covering projection. f2 : X + g :E +E E be two lifts of f . . By the assumptions p : E + B is the universal covering map. Let W ( z )be an elementary neighborhood. let p . Suppose that U = {UA.l ( b o ) . Let  where w ( )a path component of p . x I s f 1 ( z )= f 2 ( z ) } . Choose ix is a U . where bo = p ( e 0 ) . 2 . by the connectedness of X . Z n } . Suppose that (In dian a) Solution. i. n X is compact. Thus there exists a deck transformation g such that g(e0) = e l . Take a point zo E X and let f l ( z 0 ) = eo E E . For any point z E X.' ( z ) = ( 1 . W(x) is an pathconnected open neighborhood of x such that each path component of p . we see that A = X. It is not difficult to prove that A is bothopenandclosed in X. It is obvious that A is not empty. Let f2(20) = e l . Let A={ E . g f l ( z 0 ) = f z ( 2 0 ) .' ( W ( z ) ) is mapped topologically onto W ( z )by p . i=l fi p ( c ( z ) )is an open neighborhood of z in X . Prove that there is a deck transformation such that fi = g f 1 . Since p f 1 = p f 2 = f . E Ui. . E U such that Z .. Let c ( z )be the path component of %.e. which means fi = g f l : X + E . n .
Then h is a 1 . Take U = C n x and V = C n ( X . ) (Indiana) Solution. Prove that 7 p U. I). Suppose that 7 c U... continuous map.89 Let P'(v(zc>) = where that n (JW). We use the reduction to absurdity. 7 ) . Show that if C is a connected subset of X that intersects both A and X .can be covered by a finite number of V(zi).) (Indiana) Solution.. We use the reductio ad absurdum. we have proved that for any point z E X there exists an elementary neighborhood V(z) of 3: such that pl(V(z)) can be covered by afinite number of sets in U. 2111 Let X be a topological space and let A c X.A ) . U ) . Let U be any open neighborhood of 3:o in ( X . Thus h is a homeomorphism from (X. Since 7 c U . then C intersects BdA. . 6 Ui. (Recall that 7 c U means that every set in the topology 7 is contained in U. 7 ) and ( X . h : (X. Then it is easy to see is T(z)c E(z) c Ui for i = l .U) ( X .U is also an open neighborhood of z o in ( X . .1 map from + the compact space ( X .A ) c V .m. This contradicts the assumption. ' T ) is the identity map of X. U ) denote the topological spaces of X with respect t o 7 and U respectively. X i = 1. U which means U = 1.)to (X.+ . . (Recall that BdA = 2n (X. Since C n A c U and C n ( X . Hence h is continuous at 2 0 . Suppose that C n BdA = 0. i=l  2110 Let 7 and U be two different topology on X such that X is compact and Hausdorff with respect t o both.A . i=l c(z) the path component ofpl(V(z)) for any i.A ) . Let ( X . nIt follows that p'(V(z)) c . It is obvious that h(z0) = zo and h ( U ) = U . For any point 20 E X . and consequently X can be covered by a finite subcover of U. U ) to the Hausdorff space (X. That is. . Since X is compact. We claim that h is a 7).
< ( E .) 5 E.d ) < E . It is clear that C = U U V and both U and V are closed subsets of C . let O. ( A ) . Thus we have + which means 20 E C ..(A) = C.} in C .d(z0. Since A is compact. we give a counterexample as follows. Must C ( A ) be closed . ( A ) . that both U and V are open sets of C. + a for some a E A . 1). (b) Let zo be any cluster point of C e ( A ) . ( A ) is an open subspace of X . ( A ) . Then C a ( A )= [0. ( A ) is closed in X . Thus O . without loss of generality. I g) . we may assume that a. A : d(z. a ) > 0. ( A ) such that z # zo for any n and 2. (b) If A is compact. a ) < E for some a E A } . + 4 6 which means that 06/4(zo) c O . By the definition of C E ( A ) .e. . d ( z . (a) Let 20 be any point of O . a. we see that both U and V are nonempty subset of C. each z there exists an a. (a) Prove that O . z o ) + d ( z o . E A such that d(z. for . d ) be a metric space. Then. . a ) ~ d ( z . ( A ) is an open subspace of X . is not closed in X . Then there exists a sequence {z. consequently.(A) = {z E {z E A : d ( z . Take X = [0. there exists a point a E A such that d(z0. u ) < E . show that C ( A ) is closed in X . 6 = E . If A is not compact.. So C . But u nv = c n A n ( X  A ) = C n B ~ = 0. z o as n + 00. ( A ) . a ). i. for any 2 E 06/4(z0). A which is a contradiction t o the connectedness of C. for a general subspace A of X ? (Indiana) Solution. For any subspace A E cX and real number > 0.from the assumption. By the definition of O .2] c R and A = [0. 2112 Let ( X .a)5 E for some a E A } . and.
We give a counterexample as follows. Thus A would be a nonempty subset of X . a) This assertion is not correct. If Y were not connected. X . Y is not a Hausdorff space. It means A # X . f . c}. (Indiana) Solution. because Y .U ) would be an open set of X .' l v ( N ) are open in. f . Since X is dense in Y and Y . { b . X = U U V .b. (Indiana) Solution.Prove or give counterexamples to the following assertions: (a) If X is Hausdorff. But it is easy to see that the subspace { c } is dense in Y and is Hausdorff.91 2113 Suppose that X is a dense subspace of a topological space Y. then Y is connected. But U and V are open in X . b. A X . { c } . On the other hand. a) If U and V are open in X . Let Y = { a .A = X n (Y.' I u ( N ) and f . therefore. b) This assertion is true.0). Since any neighborhood of the point a always contains the point c.A is nonempty. show that f is continuous. The topology on Y is determined by the family of open sets 7 = { { a . then there would exist a nonempty proper subset U of Y which is bothopenandclosed in Y. We give a proof to it as follows. Since X is dense in Y and U is open in Y . and f : X + Y be a function so that flu and flv are continuous. which is bothopenandclosed in X. C l . a) Let N be an open set of Y. would be a nonempty open set of X. Let A = X n U . (b) If X is connected. respectively. b) Give an example where U and V are not open in X and f is not continuous.Since flu and fv are continuous. { a .' I u ( N ) and f . 2114 Let X and Y be topological spaces. el. This contracts the connectedness of X .U is open in Y .U is open in Y . c ) .l I v ( N ) are also open in X. U and V . Thus fF1(N) = flIv(N) uf'Iv(N) . then Y is Hausdorff.
( Coturnbia) Solution. Then f (u and flv are continuous. a) Suppose that X is not connected. O ) E R2 10 5 and 2 5 1) v = ((1.Y Y ) = (ql(y) n U ) u (ql(y) n V ) . 2115 Let q : X t Y be a quotient space projection from a topological space X t o a connected topological space Y. f : X +R is defined by f(x) = {x 2 XEU. and : X + Y be the map defined by q ( x . U and V such that X = UUV. () Since q is a quotient mapping.y) E V . V = q . b) The following example shows that the assumption that q is a quotient mapping is necessary for X to be connected.21.2]. and it is obvious that ql(y) n U and ql(y) n V are both nonempty open subsets of ql(y). Assume that ql(y) is connected for each y EY. Thus X must be connected. let y E q ( U ) n q ( V ) . Then q ( U ) n q ( V ) 0. and consequently. = For otherwise. b) Let X = [0. For each y E Y.l ( y ) is connected.l ( q ( V ) )and U = ql(q(U)). ) The topology of X is induced from the topology of R2. Then q is continuous but is not a quotient mapping. Therefore. where U = { ( x .1]. XEV. But X is not connected.92 is open in X . Let Y = [0. f is continuous. there exist two disjoint nonempty open subsets of X . which contradicts the connectedness of q'y. q ( U ) and q ( V ) are disjoint nonempty open subsets such that Y = q ( U ) U q ( V ) . . U = [0.y) E R2 11 5 y 5 2. y ) = 1 for (1. but f is not continuous. b) Is X necessarily connected if the map q is not assumed to be a quotient mapping? Justify your assertion. 1) and V = [l. Let X = U U V . the unit interval of R. q . a) Show that X is connected.which contradicts the connectedness of Y .q . 0) = x for ( x . O ) E U and q ( 1 . Thus.
Let p : X x Y + X be the usual projection onto the first factor. 2117 Let Y be a connected subset of the topological space. Thus X . we may assume that f ( Y ) = {l}.1} is the Odimensional sphere with discrete topology. According to the assumptions. Thus we have 2 c f'(l). = ) i=l n n Then W(z0) is an open neighborhood of z o in X . Since (W(z0)x Y )n U = 0..i. . Let U c X x Y be a closed subset and zo E X .. It means that there does not exist any continuous surjective map from 2 to So and therefore 2 is connected. Since U is closed. ( Cincinnati) Solution. Taking closures of these two sets with respect to 2 and noting that f'(l) is a closed subset of 2.e. p ( V ) is closed in X .p ( U ) is an open set of X .e.VYn (zo) such that n y = Let i=l u VYi(XO). which means that p is a closed map.Then.p ( U).(ZO) of Y such that ( Z O .p ( U ) . W(z0) C X .(~~). Since Y is connected. Y c f'(l). Let f : 2 + So be any continuous map. But c as wellknown.93 2116 Let X and Y be topological spaces.(ZO) and (Wzo(y)x V. i. we get (Y)z f'(l). . and it follows that f is not surjective.(y) of X and open set V. we see that W(z0) n p ( U ) = 0. ' . ( 2 0 . Show that p is a closed map. and consequently.(zo))nU = 0. we have Y c 2 c F. there exist an open set Wz. (Minnesotu) Solution. (Y)z= Y n 2 = 2. without loss of generality. Prove that 2 is connected. with Y compact. where So = {1. there must exist a finite number of V. Since Y is compact.y) 6 U . Y ) E Wz0(y) x V. for any y E Y .. w ( ~ ~w~. (zo). and let 2 be a set such that Y is a subset of 2 and 2 is a subset of the closure of Y .
then U U V is connected. Hence X\C must be connected.Since each V. C u U is connected because C is connected. without loss of generality. and if V is a bothopenandclosed subset with respect to X\U. 2119 1) A metric space X has property S if for every E > 0 there is a cover of X by connected sets each of which has diameter < E . and. By the above fact.. which contradicts the assumption that X is connected. b) Suppose X is a subset of a metric space. Let X\C = U U V where U and V are two disjoint nonempty bothopenandclosed subsets of X\C.94 2118 Let A be a connected subspace of a connected set X . Now suppose that X\C is not connected. consequently. for any E > 0. Then. So the above statement is proved.we see that C U U is a connected subset of X\A containing C . a) Suppose that X has a dense subset A which has property S. has diameter < ~ / 2 . Must X have property S? ( Cincinnati) Solution. Suppose the closure of X has property S. is also connected va . we may assume that U c W. which contradicts that C is a component of X\A. we may assume that A C V. Then let U U V = W1 U W2 where W1 and W2 are two disjoint nonempty bothopenandclosed subsets of U U V. a) Prove a metric space X has property S if X has a dense subset with property S. a bothopenandclosed subset of X\U. Hence W1 is a nonempty bothopenandclosed subset of (U U V )U (X\U) = X . Since C U U c X\A and CnU = 0. Since A c X\C and A is connected. U is a connected subset of X . there is a cover {Val& E I?} of A by connected sets such that each V. We first prove that if X is a connected space. show that X\C is connected. ( Cincinnati) Solution. Suppose that U U V is not connected. is connected. If C is a component of X\A. Thus W1 is a bothopenandclosed subset of V . Since U is connected.
Fig.95 and obviously has diameter < E. Let R denote the euclidean real line. ' Thus X has property S. . X be the set of all rational numbers. but it is easy to see that X does not have property S. s i n r / z ) 10 < z and 5 l} 6 = inf{z 1 (z7sin. Since = X and = U Val we see that aEr {ValQ: E I} is a cover of X by connected sets each of which has diameter < E . 22 (Toronto) Solution.lr/z) 10 < z 5 1) u ((0.sin. (ii) Let f : X + X be continuous. x 2120 Let X be the topologist's sine curve defined by X = {(z.rr/z) E A } . (i) X is shown in the Figure below. Show that either f ( X ) = X or else there exists 6  > 0 such that 3 3 < y 5 . b) We give a counterexample as follows. (i) Sketch X .1 (ii) Let A = f ( X ) n { ( z . Then = R has property S.2.} = 0. c R 2 .y) 11 5 y u{(z12) 5 2) I 5 5 '1 {(llY)7O 5 y 5 2.
( Toronto) Solution. . 2 2 2121 Let T = S1x S1denote the torus.2 Thus let U1 = X . UZ = X .I and U3 = X . 3 3 < y 5 } = 0. U U 3 . using the fact that f ( X ) is path connected. 2. we see that f ( X ) = X.Y) n 10 < 2 < 6.11. Hence in this case there exists a 60. we claim that f ( X ) = X . then it is obvious that f(x) {(z. 0 < 60 5 1. (See the Figure above. U2 and U3 are contractible open subsets and T = U ~ u U . To prove it.) Then U1. n a Fig. f ( X ) is also path connected. (i) It is wellknown that the torus T can be identified to the quotient space of a square X obtained by identifying opposite sides of the square X according to the directions indicated by the arrows as shown in the Figure below.Y) I 1 IY I 1) c f ( X ) * Once again. we first note that since X is path connected. (i) Show that T can be covered by 3 contractible open subsets.{ u U b } . If 6 > 0. such that Therefore it is easy to prove that the set {(O. (ii) Show that T cannot be covered by 2 contractible open subsets.96 I f 6 = 0.
there is an open neighborhood V . 2122 Let U = (U. . since U is locally finite.’ ( ( a .b) is any open interval of R and f E F } .(T) x Z @Z. Then let V = W n V1 n . say. It is clear that V is an open neighborhood of p such that V n ( up) = 0. . . nupi u u vp PEK P€K 2123 Let S be a set and let F be a family of real valued functions on S such that f ( s 1 ) = f ( s 2 ) for all f E F implies s1 = s2.. Prove that there exists a weakest topology in S amongst all those for which all members of F are continuous. only a finite number of Up for /3 E K . From the Van Kampen theorem it would follows that the fundamental group ?rl(T) = 0. . there is an open u PEK neighborhood W of p which intersects only a finite number of sets in U ..97 (ii) Suppose that T could be covered by 2 contractible open subsets. of p such that V . It contradicts the known fact that a.}aE~ an open cover of the space X.. Suppose that s1 and s2 are . Up. Since p $! r p . . Hence is closed. a) By definition. b ) ) 1 ( a . U is said to be locally finite. for each pi. ( Toronto) Solution. Show further that the resulting topological space satisfies the Hausdorff separation axiom. It is easy to see that 7 is the weakest topology on S amongst all those for which all members of F are continuous. = 0. b) For any point p $! vp.particularly. n V. be a) Give the definition for “U is locally finite”. . U PEK r p is closed. for any subset K c J . Then there exists a unique topology 7 on the set S of which U is the topology subbase. Up. (Hama4 Solution. Let U = { f . if each point p E X has a neighborhood which intersects only a finite number of U. . b) If U is locally finite show that.
is .7) such that si E U. So (S.l ( ( a z .b z ) ) are two disjoint open sets of (S.for i = 1.bl) ( U Z . 2 and (a1. bl)) and Uz = f . By the assumption there exists a t least a n f E F such that f(s1) # ~ ( s z ) .7) Hausdorff.2.bi) for i = 1 . Then n Ui = f'((al.Hence we may take two open intervals (a1. b z ) such that f ( S i ) E (ai.98 two distinct points of S.bl) and ( a z . bz) = 0.
99 SECTION 2 HOMOTOPY THEORY 2201 a) Give generators and relations for the fundamental groups of the torus and of the oriented surface of genus 2.3 (b). TI)). Let V be the image of the interior of the square under the identification. by the Van Kampen theorem. and let U = T . Let y be the center point of the square. I ) T is isomorphic to 7~1(U. Under the identification the sides a and b each become circles which intersect in the point 20. Solution.21) z1) containing the image +*(rl(UnV. b) Compute the fundamental group of the figure 8 and draw a piece of its universal covering space. a> As is wellknown. we can present the torus T as the space obtained by identifying the opposite sides of a square. Since V is simply connected. where q is the homomorphism induced L .{y}. modulo the smallest normal subgroup of ?rl(U.2. we conclude that ?rl(T. as shown in Fig.
a2. The smallest normal subgroup of a l ( U . where a . (See Fig. The following picture is a piece of its universal covering space. Hence a l ( U .) On the other hand.{ p } .20) is a free group on two generators a and p. we can see that the fundamental group of the oriented surface of genus 2 is a free group on four generators a1./31][a2. which are presented by circles a and b. it Fig . Changing to the base point 20. p 1 .~ . respectively. respectively.the equivalence class of a closed path c which circles around the point y once.p. respectively.2.p2].2.3 (a).20) ’ is a free abelian group on two generators a and fl.2. and. It is also clear that T ~ ( U .aa. (See Fig.2.100 by the inclusion map q5 : U n V + U . / 3 2 are presented by circles a l . P are presented by circles a and b. it is easy to see that x l ( U n V. we see that rl(T. Thus ?rl(T. respectively. By the Van Kampen theorem we can prove that a l ( X . each level segment is mapped on the circle a . In a similar way.21). ~ where 6 is the equivalence class of a path d from 20 to 21. consequently. and [ailpi] denotes the commutor aipiailpzrl.p2 with the single relation [a1. 2 1 ) ) is just the commutator subgroup of 7rl(U. It is easily seen that a u b is a deformation retract of U .2. V = X . Z is )a free group on two generators a’ = F1a6 and p’ = C1p6. Let U = X .~ f l ‘ .p1. a 2 .{ q } . where a and p are presented by circles a and b. as shown in Fig. 21) containing & ( a l ( U n V.zl) is an infinite cyclic group generated by 7 . z o ) is a free group on two generators a.) b) Fig .4 Let X denote the figure 8 space. where a l . that q5+(y)= c y ’ p I ~ y ’ .ba.4. bl. 1 ) is a free abelian group on 2 two generators a and p’.3 (b).5 Under the covering map T.
there exist disjoint open subsets of A . Therefore. E A . which is an open subset of A. Thus it is easy to see that f"(A) is compact. Noting that fn(A) is compact for every n and fn+'(A) c fn(A). is a compact subset. Wl and Wz such that U c W1 and V c W2. F(S' x I) is also closed in .W is closed in A and. Since F(S1x I) is compact and X . consequently. .. Since A . E f"(A) but 2.B be a homotopy between y and the constant map. there exists at least a limit point 20 of the sequence { x C n } . where U and V are disjoint nonempty closed subsets of B. Since A is obviously compact and Hausdorff. (Indiana) Solution. i) Since X is compact and A is closed. which is a contradiction. is ii) Let F : S1x I 3 X . and therefore. It is obvious that U and V are also closed subsets of A. Suppose that B is not connected. we see that B = is a nonempty closed subset of A . n=l ii) Suppose y : S1 X .f"(A) and such that the induced map y : S1+ X . Show that there exists a positive integer n such that y(S1) c X .W for every n. there would be a squence in A . @ W for every n. 2202 Let A be a connected. n times 00 f"(A) is connected. closed subspace of a compact Hausdorff space X . and each vertical segment is mapped on the circle b according t o the direction indicated by the double arrows. otherwise. A is a normal space. such that x. o ~ ( A ) . {xn}. and that A is compact. A is also compact.B is open. For each positive integer n let fn(A) = ~ o ~ o . We claim that there exists a positive integer N such that fN(A)c W .fn(A) ' (y'(s) = y(s) for s E S') is also nullhomotopic.101 according to the direction indicated by the arrow.. Then B = U u V . Noting that . For. n which contradicts the fact that fN(A) connected. and suppose f : A + A is a continuous map. Thus it follows that B c W1 U Wz = W . n=l n fn(A) m PYA) = ( f N W n Wl)u (fN(A) W2>. closed and connected.B is a nullhomotopic map. It means that x. can see we that xo E B c W . Thus f N ( A )c W .+'(A) c (i) Show that B = n f"(A) c A for any n.
. Define H : RP" x I + R" by  H(z.  + fl is null 2204 Let X be the quotient space obtained by collapsing { p t . for any continuous map f : RP" + T". Prove that any continuous map RP" + T m is nullhomotopic. It is also wellknown that "  ai(T") M Z e . c F(S1 x I ) c u c Particularly. say f. So homotopic and consequently f is also nullhomotopic.102 X .fN(A). Then H is a homotopy between f and the constant map PO. y(S') x .t). It is wellknown that R is a universal covering space of T". 2203 Let RP" be the real projective nspace and T" be the rntorus S1 x . } x to a point S1c S1 x S1 . Let P : " R + T" be the universal covering map.t) (1 . (rn times) and that a l ( R P n ) M Z2. cX . it follows that there exists a positive N such that fN(A) V . the induced homomorphism f+ : a 1 ( R T ) + nl(T") is trivial. there exists open sets U and V such that F(S1x I ) c U and B c V . . Thus there exists a lifting of f . (Indiana) Solution. Let po be a fixed point of R".. Since X is normal.fN(A). such that pf = f . x S1 ( m factors). Therefore. Therefore.T(z) tpo.The remainder of ii) is obvious. From the proof for i).
The conclusion is that H i ( X ) is an infinite cyclic group for i equal to 0 . where S is the equivalence class of a path d connecting 20 and z1. containing the 21) image q5.2.6 X may be identified with the space shown in Fig. where q5* is the homomorphism induced by the inclusion map q5 : U n V + U . 2205 (i) Suppose n 2 2.{y}. zl)is an infinite cyclic group on the generator SluS. Therefore. U = X . Therefore we conclude that . and V is simply connected. z l ) is isomorphic to the quotient group of a l ( U .(al(U n V. which is obtained by identifying the sides of a 2gon. 7 r l ( X . Does there exist a continuous map f : T + S1 which is not homotopic to a constant? ( Toronto) . by the Van Kampen theorem. Let y be the center of the 2gon.V).6) It is also clear that x l ( U n V. 2 1 ) modulo the smallest normal subgroup of R ~ ( U . a F ig. 1 . where y is the equivalence class of a closed path c which goes once round the point y. Then. To compute H . zl)). V.2. ( X ) . U . Does there exist a continuous map f : S" + S1 which is not homotopic to a constant? (ii) Suppose n 2 2.6. zl)is an infinite cyclic group on the generator y. and V be the interior of the 2gon. Does there exist a continuous map f : RP" + S1 which is not homotopic to a constant? (iii) Let T = S1 x S1 be the torus. It is easy to see that &(y) = lrl(u. U and V are open subsets. ( X ) . (Columbia) Solution.zl). we may apply the MayerVietoris sequence to the pair ( U . It is easy t o see that u is a deformation ratract of U . and U n V are path connected. al(U.l(X) =z.2. (See Fig.103 Compute x1(X) and H . 2 and is zero otherwise. Thus.
f must be homotopic to a constant. prove that p# : ?rl(E. (i) If F is path connected. + . Since al(S”)= 0 for n 2 2. t z 5 1. b o )where f : I + B is a closed path a t bo .any continuous map f : RPn + S1 induces a trivial homomorphism f* : x 1 ( R r n ) a l ( S 1 ) . there does not exist any continuous map f : Sn + S1which is not homotopic to a constant map. bo E B be a base point. Since R is contractible. Let i : F E denote the inclusion map. and eo E F . Let p :E B be a fibration. consequently. for any pair of continuous maps I t G : X x I + B and h : X x (0) t E such that p h = Glxx(oi there exists a continuous map H : X x I E such that H l x x i o ) = h and p H = G . By the same argument as in (i).104 Solution. 7 c 7: + It is easy t o see that the induced homomorphism f* : H1(T) + H1(S1)maps one generator of H1(T) to the generator of H1(S1). is exact. 2206 A continuous map of topological spaces: p : E B is called a fibration if it has the homotopy lifting property . respectively. Let G : I x I B be a continuous map defined by f . hence f is homotopic t o a o = C.rrl(B. (i) Let cr = [f] E ~ l ( B .eo) + Tl(E. (ii) Since a l ( R P n ) = Zz. has a lifting RP” + R such that a o f = f.and another generator to zero. e 2 x i t z ) = e27ritl E sl.bo) is surjeceo) tive. 20 5 t l . Define it f : T + S 1 by f ( p i t l .and.eo) + “1(B.. = and f : S” + S1be a continuous map. e 2 n ).e.l ( b o ) (the “fiber”). (i) Let a : R + S1be the universal covering map defined by ~ ( t )eZrrit. : S” + R such that T? = f.that is. ?must be homotopic to a constant map : S” + R. F = p . (ii) Prove that in general the 3term sequence + + + + Tl(F.which is also a constant map from S” to S1. we see that there is a lifting of f. Solution. Hence f* is not trivial. bo) (Indiana) in which the homomorphisms are i# and p # . (iii) Denote T = S1 x S1 by T = ( e 2 r r i t l . i. which means that f is not homotopic to a constant. which represents a.
and h : I x (0) + E be the constant map such that h(t. l which means that p# is surjective. G = = [c] = i#[c]. Noting that pgl. cg and c3 are paths in F and cl(0) = eo. Then pc = f . i#(a)= [pf] = [bo]. ~ ( 1 = c i 1 ( 0 ) and c i 1 ( l ) = eo.Then c1. we may choose two paths g1 and 92 in F such that gl(0) = e2. g2(0) = e l and gz(1) = e l . c kerp#.7 2207 Is the canonical map q : S2 + R P 2 (which identifies antipodal points of S2)nullhomotopic? Why or why not? (Indiana) . On the other hand. Then there exists a homotopy G : I x I + B between pTand the constant path bo such that G(t. Since F is pathconnected.eo). It is obvious that p# . It is easy to see that f is homotopic to C. 0) = bo. cg = H11. 0) = ( p T ) ( t ) . 1).O) eo. Hence im. Let c1 = H I { o i x r . (See Fig. It is obvious that p h = GII. s ) E bo for any s.Irl(F.) Therefore.e. 0) = f(t). s ) G ( l . Let c ( 0 ) = e l and c(1) = e2.105 G ( t .7. (ii) Let a = [f] E . Let C : I + E be a path in E defined by c ( t ) = H ( t . l I = b = [fl = a .{1} and c3 = H l { l } x ~ . where bo is the constant path a t bo. Therefore there exists a continuous map H : I x I + E such that H / I x { o }= h and p H = G .G(t. gl(1) = e l .{o}. i. ) ) Hence c = c1 x c2 x c i l is a closed path in F with base point eo. suppose that G = [f] E kerp#. Thus f = g2 * c * g1* gF1 is a closed path a t e l .2.  [fi Fig. we have P# rfi = bfi = bg21 .2. b 1 . By the homotopy lifting property.. t ) E I x I . It is clear that e l and e2 belong t o F. s ) = f ( s t ) for ( s . and G(0. 1) = f ( t ) . kerp# c imi#. Particularly we have p H ( t . bg11 . ~ ( 1 = cz(O). there exists a continuous map H : I x I + E such that P H = G and H ( t . pg2 and pgz' are all constant path at bo. . where gF1 is the inverse path of g2. Hence the sequence mentioned above is exact. h .
Let i : X t T 2be the inclusion map. there is a lifting of T o p .eo) such that p o ? = T o p. Hence degg = f l . ~ " . S1. (Here T n = T o T o .106 Solution. Then roi : X + X is . f   (Indiana) Solution.eo) + ( E .. T : ( E .      2209 Let T 2 = S1x S1. eo) since T" = Id. On the other hand. . Let T : B + B be a map so that T n = Id and so that T(b)= bfor some b E B. Choose eo E p . This is a contradiction. It is wellknown that q is also the universal covering map from S 2 to R P 2 .. denoted by r. We use the reduction to absurdity. Therefore.. < 2208 Let p : E B be an universal cover with E and B pathconnected and locally pathconnected. We have T o p ( e o ) = T ( b ) = b. we see that @ is a lifting of T" o p at the base point eo. Then q is homotopic to a constant map c : S 2 + R P 2 . Suppose that q is nullhomotopic. Since ?rl(E)is trivial... we have d e g c = degZ = 0. But it is obvious that f is the identity map or the antipodal map from S 2 to S2. Suppose that there is a retraction of T 2 to X .l = .) Show that there is a map ? : E + E so that p o T = T o p and Tn= Id. = Id.and therefore q has a lifting : S2 + S2which is also homotopic to the lifting of c . Prove (Indiana) Solution. Since I  T" = ( p T ) and . n times.l ( b ) and consider the map T o p : E + B . o T . = T ( e 0 ) = eo. = T~ O )  P T " ( e 0 ) = T"'(T(eo)) = T*'(eo) = . a constant map C : S 2 + S 2 . + ( E . let X c T 2 be the subset S1 x { 1) U (1) x and that there is no retraction of T 2 to X . On the other hand. .= l T ( p ~ n . it follows that identity map Id : (E. Thus we conclude that q is not nullhomotopic. by the uniqueness of lifting. we conclude that r?.eo) is obviously a lifting of T n p a t the base point eo.
It is wellknown that r1(T2)is abelian and that .{ c } and V = Y .: r l ( X ) + r l ( X ) is the identity homomorphism.0) and ( .Y. It is easy to see that U has the .Z) I (Y2 + z2)(x2+ 2)= O}.l .6 = 2.{pl.(2) = 3.A . b) Let e be a point on the arc ab different from a and b. Then we have Y = U U V . z) I x2 z 2 = 1) be a circular cylindrical surface. Hence ~. p ) is not abelian.{ e } . and let p = ( O .y.Therefore we have = which is a contradiction.8 a) Y has a structure of a graph with 4 vertices and 6 edges. b) Prove that r l ( R 3. + Fig. (Indiana) Solution. O . Take U = Y . 1). p l and p2 denote the points (1.8.2.A ) . Let X = {(x.1(X) is an nonabelian free group on two generators denoted by a and b.107 the identity map.2. Hence we have ab # ba and i*(a)i*(b) z*(b)i*(a). It is also clear that X . Then it is easy to see that X .{ p l . 2210 Let A c R3 be the union of the x and yaxis. A = {(X. Hence the rank of H I ( Y ) is equal to 1 . a) Compute H 1 ( R 3 . O .oi.A . 0) respectively.p2) is a deformations retract of R3 .0.A ) M H 1 ( Y ) z Y Z @YZ @Z. Therefore we conclude that H1(R3 . p z ) is homotopically equivalent to the space Y as shown in Fig. The Euler Characteristic K ( Y ) = 4 .
Then let fl .9. ZO) + ( Y .g) is a normal subgroup of 7rl(Y. (A 5. Let fo(zl) = y.g) + (Y. p*(.A . that is.y). that (R3 .y) be a regular covering space. (Indiana) .Since Since p is a regular covering space.A . Therefore we conclude that ?rl(R3 . p ) is not abelian. 2211 Let p : (?. by the Van Kampen theorem we see that ?rl(Y) is a free group generated by ?rl(V) and ?rl(V). : ( X . + Hence 5 is the lifting o f f we want. 2212 Let X be the identification space obtained from a unit 2disk by identifying points on its boundary if the arc distance between them on the boundary circle is Compute the fundamental group of X.108 same homotopy type as S1 and that V has the same homotopy type as the “eight figure space”. Suppose that f : X + Y is a continuous function from the pathconnected space X to Y with f ( z 0 ) = f(z1) = y. and that there is a lifting of f . consequently. 9.g. p ) M ?rl(Y) is a free group on three generators and. Since U n V is contractible. Show that there is a second lifting.= r o f o : x Y.) (Indiana) Solution.rrl(Y.z1) + need not be distinct from  5 TO (?. there is a deck transformation y such that y(y) = g. : (X.
) It is clear that r l ( U n V.zl)is an infinite cyclic group generated by y o the closed path class of a closed path c which circles around the point y once. (See Fig.9. Since V is simply connected.where 4 : (U n V ) + rl(U)is n the homomorphism induced by the inclusion 4 : U n V + U . r 1 ( X ) is isomorphic to the quotient group of r l ( U ) with respect to the least normal subgroup containing 4*(rl(U V ) ) . We may consider f t o be a continuous nonconstant map f : a +a. Then both U and V are path connected subsets of X and X = U U V . Note that If(z)l + 00 as JzI + 00.2. 2213 Sketch a proof of the Fundamental Theorem of Algebra (every nonconstant polynomial with complex coefficients has a complex zero) using techniques of algebraic topology.2. it is clear that wl(U. Since the circle a is a deformation retract of U . we may extend f ' . Fig. It is also clear that 4+(yc) 3Z. hence.109 Solution.) Let U = X . Therefore r l ( U 121) is an infinite cyclic group on ' generator Z = y'a'y. It follows that rl(X) NNZG/U n =as. where y is the path class of a path d from z o to z1. by the Van Kampen theorem. Hence the least normal subgroup containing = 4*(rl(U V ) ) is isomorphic to 3 2 .9.9 Take a point y in the open disk. Let denote the complex plane and f (2) be a polynomial of positive degree with complex coeficients. z o ) is an infinite cyclic group generated by u the closed path class of a.{y} and let V be the open disk. (Indiana) Solution.2. Take a point z1 E U n V as the base point. (See Fig.
It is obvious that . (b) Compute the integral homology groups of X . Furthermore. (a) Compute the fundamental group of X .(4).k > 0. we may prove that if f is any polynomial of degree k > 0 then the degree of the extension f : S2 S 2 is still equal to k . Then 1 we have U u V = X1 V X z .2. then the degree of the extension f : S 2 + S 2 is equal t o k. (See Fig. we take U = X 1 V X 2 . Then we may first prove that if f ( z ) = z k . we may prove the Fundamental Theorem of Algebra by means of the reduction to absurdity. and the portion.10 It is clear that X has the homotopy type of the one point union X 1 v X 2 where X 1 and X 2 are each homeomorphic to the union of the unit sphere and the portion of the z axis lying within S2.2. by the Van Kampen. a l ( X 1 v X 2 ) is a free product of the groups r l ( U ) and T ~ ( Vwith respect to ) the homomorphisms induced by the inclusion maps.{ p } and V = X 1 V X2 . (4 Fig. Since U n V is contractible. Noting the fact that if a continuous map f : S2 + S2 is not surjective then the degree o f f is zero. (Indiana) Solution.) To compute ~ l ( X v X 2 ) . of the z axis lying within S 2 .110 to a map of the onepoint compactification of ( X f : s2+ s2 by setting f ( m ) = 00.10. + 2214 Let X denote the subspace of R3 that is the union of the unit sphere S2. call it A. where 00 denotes the north pole. the unit disk D2 in the 2y plane.
f : X Y a continuous function. cac is a closed path at f ( z 0 ) . For any [a] E ? r I ( Y .2. f ( z l ) )where a is a closed path at f ( ~ l ) . 20.) (b) Applying the MayerVietoris sequence to the pair ( V . By the assumption. otherwise.111 Therefore. We can take as generators the closed path classes which are determined by the closed paths omp and omq respectively. f*([c'hc]) = [a].10. and E X . Since X is pathconnected. It means that f o h and Zu?' = are homotopic.V). . (See Fig. we conclude that i = 0.1] + X such that c(0) = 20 and c(1) = 2 1 . Hence  f* : m ( X . 2 1 ) is surjective. there is a closed path h at 20 such that f*([h]) [Z?']. there exists a path C : [0. Show that is also surjective. ? r l ( X )is a free product generated by two generators.f ( 2 1 ) ) . Thus f o (c'hc) and a are homotopic. (Indiana) Solution. consequently. and. n ( Y .2 and are zero otherwise. Then Z= f o c is a path connecting f(z0)and f(z1) I in Y . Suppose that the induced homomorphism f is surjective. we see that Noting that H i ( X 1 ) M Hi(X2) which are infinite cyclic for i equal to 0.1.2 1 Let X be a pathconnected space.
) Since the automorphism group A ( X .{f(eo)}. X has only two different types as shown in Fig. By the formula .{ f ( e O ) } is obviously not contractible.. Since X . but not necessarily homeomorphic. r ( x Y )= m ( X > m ( Y > .112 2216 Let B denote the “figure eight space”.{ e o } is contractible and Y .2.2. 2217 Calculate the fundamental group of the space R P 2 x S2. by considering the liftings of the circles a and b in X . Then X . we come to a contradiction.2.11. Let p : X + B and q : Y + B be 2fold covering maps. (See Fig.el}.11 are not homeomorphic. that. Then it is easy to see that they are homotopy equivalent to an 3leaved rose G3.2. it is not difficult to see. where both X and Y are connected. (Indiana) Solution.{ e o ) is homeomorphic to Y . (Indiana) Solution. let p .11. suppose that f : X + Y is a homeomorphism. Prove that X and Y are homotopy equivalent. .l ( z 0 ) = {eo. But the spaces X and Y shown in Fig. Otherwise.11 For any 2fold covering map p : X + B . rlX @ we see that 7rl(RP1x S2) x T1(RP2)@ “1(S2) a @ {O} M 2 =a. in substance. a m b Fig. p ) F Z & and X is connected.
113 2218 TI(.+ G a surjection.2. By the same reason.' ( Y ) contains exactly the points {eo. Z = X V Y . and p : 2 + 2 the *) corresponding cover.   (Indiana) Solution.) If A is a path component of p . Let x. what is the cardinality of the sat A n B?) (b) If G is a finite group.e3. e3} of p'(*).' ( Y ) . * is generated by 2 and y. where * denotes the vertex Fig. ( p * ( ~ 1 ( 2 .' ( X ) with a path component B of p .' ( * ) by p .e2. we see that the path component of p . X and Y circles.el. We denote p .e4} of p'(*) which corresponds to the elements {0. *) Let 2 denote the figure 8 space. Hence the covering space p : 2 + 2 is a 6fold cover.. how many intersection points of A and B are there? (i.!?.ez.' ( X ) and *)) B is a path component of p .l ( * ) . from h ( z ) = 2. Thus ~ l ( 2 .2.e5}. we conclude that A n B = {eo}. Y . Then.12 (a) Let h : ~ ' ( 2 .' ( Y ) divides the order of G . prove that the number of intersection points of a path component A of p . we see that the path component of p . (a) Since h ( z ) = 2 and h(y) = 3 and T ~ ( Z . ker h is isomor*)/ phic to the group Z/6Z= Zg.e4.' ( Y ) = p . Since p'(x) n p . and let p : 2 + 2 denote the covering space corresponding to the kernel of h.4}of& respectively. h : ~ ' ( 2 .' ( * ) = {eO. = ker(h).+ Z/62be the homomorphism satisfying h ( z ) = 2 and *) h(y) = 3.' ( X ) contains exactly the points {eO.e.y E be the elements in 2 defined by X . . it ) follows that the homomorphism h is surjective.
we conclude that $1 : . with proof. as shown in Fig. Under the identification the edges a become a circle through the point 20. 2219 Let X be the result of attaching a 2cell 0’to the circle S1by the map f : S1+ S1given in terms of complex numbers by z + z6. (Indiana) Solution. Let V be the image of the interior of the hexagon under the identification. (4 a d @ 2 0 a Fig. we see that the number of intersection points of A and B is equal to the number of elements in H1 n Ha. with proof. the homotopy class of a closed path c which circles around the point y once (see Fig. By Lagrange’s theorem. It is clear that U n V has the same homotopy type with S1. Then.2.13. that wl(U nV.21) is an epimorphism and its kernel is the smallest normal subgroup containing the image of the homomorphism 41: rl(u n v. the homology of the universal cover of X .13 Represent X as the space obtained by identifying the edges of a hexagon.114 (b) Since h is a surjection and G is a finite group.Irl(U.2. In a similar way as in (a). zl). and let U = X .zl) + r1(U.) Applying the Van Kampen theorem. V .rrl(X. Let y be the center point of the hexagon. (a) Compute. U . 2 1 ) + . it divides the order of G. and V is simply connected. (b) Compute. the corresponding cover p : 5 + 2 is a finite fold cover.2. . Let 2 1 be a point in U n V . and U n V are arcwise connected. 21) is a n infinite cyclic group and generated by y. Suppose that h ( z ) = r and h(y) = s. Then r generates a subgroup H1 of G and s generates a subgroup of H2 of G.13. U and V are open subsets.{y}. the fundamental group of X .
hence X ( 2 ) = 6. It is obvious that dl(r) = ar6. 7 r l ( U . i = 0. 2220 i = 1. Hi(%) i 2 3. 2 . So we conclude that .21)) is isomorphic t o U . S1. H I ( % ) = 0 and H 2 ( 2 ) x H 2 ( X . g : X + S1is defined by g)(z) = f ( z ) g(z) for any z E X . zo) is an infinite cyclic group generated by a and. Here ( f a  .S1] of homotopy classes of maps from X to S1 inherits a natural group structure. 1 where It is obvious that the circle a is a deformation retract of U. Then the multiplication of [X. (a) We denote the homotopy class of a map f : X + S1by [f] and write s1 as s1= { e 2 = i o ECT I o 5 8 5 1). consequently. is defined by [f] [g] = [f . But it is easy to see that X ( X ) = 1. g ] . where 7 is the homotopy class of a path d from zo to 21. and T 2= S1x S1. From H o ( 2 ) M I . S (Indiana) Solution. Thus ?rl(U. then it is known that the set [ X . It is wellknown that the Euler characteristic X ( 2 ) = 6 X ( X ) . (b) Compute this group explicitly for X = point. (a) Define this group operation explicitly and indicate the group identity and how inverses are formed. where 2 is the 1skeleton of 2.lrl(U n V .115 and 4 are homomorphism induced by inclusion maps. where the S1] map f . You do not need to prove your assertions. X ' ) . see ' we that H 2 ( 2 ) is a free Abelian group of rank 5. z l ) is an infinite cyclic group generated by a' = ?'a?. ( 0. Thus we conclude that ) (b) Since x is a finite 2dimensional cw complex and T ~ ( X= its universal covering space X is a 6fold covering space and is also a 2dimensional CW complex. If X is any topological space and S1 denotes the unit circle in the complex plane with its usual topology as a topological group with multiplication given by the multiplication of complex numbers. = I Z. Hence the smallest normal subgroup containing 41(.
Now we discuss the case of X = T 2 = S1x S1. It is clear that 4 is a epimorphism. Then by the facts that S2 is simply connected and the universal covering space of S1. where the map e : X + S1is defined by e ( z ) E eZni for any 3 E X. We have 4“fI . S1defined by t ) = f ( e 2 n i f 3 1e2xit+2xi(lt)+ 1.92)) = (deg fl + degg1. deg(f2 . [sl) = = = 4Uf.9)2) (deg(f1 * Sl). it is easy to see that each homotopy class [ f ] can be represented by a map f : S2 + S1which maps the northpole N of S2 to the point po = eZUiof S1.sl) = (deg(f * 911. 4 is a homomorphism from [X. So the group [ X . The inverse of [ f ] is the homotopy class of a map f. one can easily prove that [S2. 2.S1] %Z@Z. We conclude that 3 [X.S 1 ] to Z @Z.S1] is trivial.degfz). M r2(S1)= 0.S1]+Z@Z is defined by 4([f]) = (degf1. denoted by f1. Note that f l can be extended to a map from X to S1. p o )and fZ(6) = f ( p O l e z U i e )for any e2nie E S1. For the case of X = S1. and cosequently q is an isomorphism. which is : defined by f(3:) = l / f ( z ) for any 3: E X .S1] obviously has only one element. deg f 2 + deg 92) + (deg 91 deg 92) 1 = (deg fll deg f 2 ) 4(V1) + N g 1 )  Therefore. Then let 4 : [X. R is contractible. (b) If X = point. S l ] x “l(S1)xa. respectively. For the case of X = S2. deg(f . Thus one can easily prove that 4 is a monomorphism. by still which is homotopic to f under the homotopy map F : X x I + ~ ( ~ 2 r r ie0 n i G . then [ X . g ( z ) is defined by the multiplication of complex numbers.116 f ( z ) . can easily prove that one [S’. any S1] For map f from S1x S1+ S1we define two maps f1 and f 2 from S1+ S1 by fl(6) = f ( e Z n i e . The identity of this multiplication is [el. .
which is a contradiction. then ~ ~ ~ ( would ) a closed set of infinite points in X. show that ?rl(X. ?: 7 ..~ ( z o must have at least a limit point. What can you say if we just assume that X is path connected. i. the universal covering space of S1.e.e. f is homotopic to a constant map.20) is finite. Since X is locally path connected and simply connected. 2 0 be T . Denote the homotopy between and E by g. (Indiana) Solution. Let ?r : 2 + X be the universal cover of X. Let exp : R + 6 denote the exponential covering map. 7r1(X) = 0. Since R is simply connected. and the same conclusion holds. Thus it is ) easy to see that x is not a local homeomorphism at Z. =0 Hence there exists a lifting of f . So the above ) argument still works in this case. Since 2 is compact.such that n(Z) = 20. locally path connected and the fundamental group of X is finite? (Indiana) Solution. X + R such that exp(7) = f .117 2221 If X is a pathconnected space whose universal cover is compact. 20) were not finite. and f*(?rlT1(X)) for any map f : X + S1.  Prove that if X is locally path connected and simply connected then every map X 4 S1is homotopic to a constant.Then exp(2) is the homotopy between f and c . we see that f * ( ~ l ( X )must be trivial. i. Since ?rl(S1)% Z and f * ( ~ l ( X ) is a finite ) subgroup of ?rl(S1).. If n l ( X . say Z. Suppose that ?rl(X) is finite. f is homotopic to a constant map Z.
(Hamud) Solution.e.&(b) E kercl. there is a b E B such that c = b1(b).118 SECTION 3 HOMOLOGY THEORY 2301 Prove the 3 x 3 Lemma. Since b1 is surjective. an a E A3 such that a z ( a ) = a2. By ) 3 the commutativity we see . we show that 7 2 is injective. . Since P2(62(a) . Let c E C3 and 7 2 ( ~ = 0. Then since 2 ) c2(al(a2)) = Pl(EZ(a2)) = Pl(PZ(b)) = 0 and C 2 is injective. To prove the exactness at C. then the last row is also short exact. i. we see & ( a ) = b..b ) a2 E ker 01 = imcyz. Consider the following commutative diagram of abelian groups 0 1 0 t A3 62 1 B3 0 0 + 61 + 1 c 3 1 0 If all 3 columns and the first two rows are short exact.Pz(b) = 0 and f ~ 2 is injective.(b) = 6162(u) = 0. Hence 7 2 is injective. we have a l ( a 2 ) = 0. Thus there is = P262(a) .. Therefore c = S.h ( b ) = w 2 ( a ) . Hence there is a n a2 E A2 such that ~ ~ ( a= Pz(b).Pz(b) = E z ( a 2 ) .
Therefore there is a b3 E B3 such ) that p 2 ( b 3 ) = b .&). Since M is compact.X 1 = R P 2 x R P 2 .2. . M is 272orientable.~ 2 ( u 2 E kerpl. there is an a2 E A2 such that al(u2) = ul. Therefore X ( M ) = 0. and X4 = S2x S2. Then 72(61(b3))= E l h ( b 3 ) = &l(b). since ~1 is surjective. 2m+ 1 X ( M )= (l)i dimHi(M.i have different parity. Thus the exactness a t C is proved.3. Thus by the commutativity.~ 2 ( u 2 ) . since i and 2m 1 . we have b . Show examples of compact 4manifolds with X = 0. In a similar way we may prove that imy2 c ker 71. Therefore. they appear in the sum with opposite signs. Denote by u the connected sum of h h projective planes. there is an b E B2 such that c = ~ l ( b ) Thus it is easy to see that Pl(b) E ker C1. 2302 Prove that if M is a compact manifold of odd dimension. ( Columbia) Solution. then X ( M ) = 0.119 Now we prove that ker y1 c imy2. Hence c E imy2. 2 We leave the proof of the exactness a t C1 to the reader. Thus.1. i=O By the PoincarC duality theorem. Let X2 = U3 x 174. Then it is wellknown that X ( U h ) = 2h.E 1 & 2 ( a 2 ) = c. X3 = U3 x U s .4. 3 and 4. Let Xo = T2x T2. for any i. .2.Using the fact that + we see that X(X. Suppose that d i m M = 2m + 1. and consequently that there is an a1 E A1 such that &(u1) = Pl(b). Since a1 is surjective. For any c E ker 7 1 .1.) = i for i equal to 0.
~) + (T.l ( X ) denotes the reduced homology of X  2304 Consider the chain complex c d. = t {C"In>O} C" C.t)E C . Noting that imdn+l = {(O.s.t.{pl} and V = C X . s . (Illinois) Solution.{ p z } . Set U = E X . and E X = U UV.s.l defined by (s .t . Compute the homology of E X in terms of the homology of X . Then U and V are open sets of E X . . we have H .e. we have where H q . ( c ) = Z for even n. O ) =z3 (T. (Illinois) Solution. It is obviously to see that U and V are contractible spaces and X is a deformation retractor of U n V .s+t) nodd. When n is even. Compute H.120 2303 Let X be a topological space. ( c ) is isomorphic to Z $Z. Z . : C. } for even n.1] by identifying X x (1) to a point and X x { 1) to another point. ~ E Zn(c) if and only s = t .s+t. O . ( T . Let pl and pa be the identification point respectively. By MayerVietoris sequence.(C) for all n. For example the sphere S" is the suspension of S"' with the north and south poles corresponding to the two identification points. s .. So Zn(c) = { ( T . en}. s ) E ) i. The suspension E X of X is defined t o be the identification space obtained from X x [1.~) + n even (O.
( T . Therefore. ( c ) M Z for any n. + 2305 Construct a CW complex which has the following 2homology groups: ( Coz~mbia) Solution.+l = { ( ~ . . It is wellknown that H1(X1) = Z$Z.121 When n is odd.and noting imd. 0 . the image of one 2cell under f1 is homeomorphic to S 2 . X') T fi* 0 + T fils'+ a : H2(D2. ~. Therefore. one by the map f i : S' + 20 and the other by the map f2 : S' + a such that f2(z) = z2. the 2sphere. So Zn(c) is isomorphic to Z $Z. Since H ~ ( xx') = imfl. consider the following diagram 0 t HZ(X2) 5 H2(X2. CB imf2. Denote by X1 the "figure eight space" as shown below. M z CBZ. H . which has two generators a and b.S1) Hl(S1) + 0 The square is commutative and the level rows are exact. 0 )E C.X1) 3 + H'(X1) % H1(X2) .} M Z. s .HI(X2. Let X2 be the space obtained by attaching two 2cells to X I . ~ )E & ( c ) if and only if s t = 0.we have Hn(c)= z .
we may prove that gp(Sn(q)) is a free abelian group of rank q for p equal t o n and is zero otherwise. + + + Let a E S n ( q ) and b E S".. Then H p ( S o ( q ) )is a free abelian group of rank q for p equal to 0 and is zero otherwise.. Since H1(X2. Now let X be the space obtained by attaching two 3cells to X2. Then in a similar way as above. In fact S"(q + 1) = S"(q) v S". by the map z one and the other by the map z + z6.{ u } has the homotopy type of Sn(q) and V = Sn(q 1) . when n > 0. V S" When n = 0.{ b } also has the homotopy type of S"(q). ( Columbia) Solution. n 2 0. It is also easy to see that H z ( X 2 ) =z.X1) = 0. So(q) has J . Denote by S"(q) the space S" V S" V . It is easy to see that U = S n ( q 1) . by induction on q and the MayerVietoris sequence of the pair ( U . we may see that ima. + 2306 Compute H.. M 22 and kera. 2307 Build a CW complex X by adding two 2cells t o S1. It is also clear that Sn(q 1) = U U V and U n V has the homotopy type of S n ( q . we may conclude that the space X satisfies the requirements in the problem. V S"). = 0 and imf2Isi. = 2!J.(S" V S" V .122 imfllsi. Thus. z Y . for all p .. q times q + 1 points.1). where fip is the reduced homology group. we have H1(X2)M Z $ Z 2 . What is the homology of this space? + z4 (Indiana) . one by the map g1 : S 2 20 and the other by a map g2 : S2 + S 2 such that degg2 = 3. V ) .
where d. ( X ) M H n ( K ) . : and C n ( K )+ Cnl(K) is defined to be the composition of homomorphisms. Thus.we have H 1 ( K ) M 2/22 M &. Hi(X) = i = 0. since j 2 is injective. we see that H 2 ( K 2 ) imj2 M M ker 8 . KM ) ~ imf.123 Solution. is a n isomorphism. i = 2 . Then we have a chain complex K = {Cn(K). CB img. K1) 2 HI(K1) 2 H I ( K 1 . it follows that imd. imd2 M Z.(K) = 0 for n >_ 3. is isomorphic to the subgroup 22 of H I ( @ ) .l is the homomorphism induced by the inclusion map. where C n ( K )= Hn(Kn. is 6Z. Thus we conclude that ( zg3z2. K 2 is obtained from K 1 by attaching two 2cells via the maps f and g as indicated in Fig. It is clear that j is an isomorphism. Fig.KnA1)d. It is obvious that H n ( X ) = H.S1) 3 H l ( S 1 ) The square is commutative and it is well known that f. I 01 otherwise.14. is a monomorphism and 8. and that ker 3.dn}. we see that the image of ( f l S l ) * is 4Z. Noting f l s i is the map z + z4. It is wellknown that H . is isomorphic to Z @&. Since H ~ ( K ' .2. KO) T f* T (f IS^)* H2(E2. K 1 is obtained from KO by attaching one lcell. is the boundary operator of the pair (Kn. Therefore. Noting that 1 ker d l % H 1 ( K 1 .KO) M 2.14 X is a %dimensional CW complex with %skeleton K 2 = X . To compute H2(h). consider the following diagram: HZ(K1) t HZ(K2) 2 H2(K2. i = 1.In the same way we see that the image of (glsi). Let K" = K 2 for n 2 3. lskeleton K1 = S1 and Oskeleton KO = { p } .2.Knl) and j. M Z @&.. .
Similarly we obtain Hz(X. . otherwise. ( X . y.z) E R3 I zyz = 0). let Wl = .we see that b) Let z be any point of X different from the origin. So H. X . u .I(U. x . ( X . (O. I). x .{XI) = Z @Z @Z.X .(U . a) Let U = D3 n X .0)).(0). b) Prove that any homeomorphism h : X Solution. z ) E R3 I z2 + y2 + z 2 < 2). Let h : X + X be any homeomorphism.y.(U. Thus H . (Indiana) D3 = {(z. if z is in a coordinate axis.(f(0))). From the above results. where +X must leave the origin fixed. H 2 ( X . Wz). where S 2 is the unit 2sphere.X .(0)) g * ( S ) .124 2308 Let X = {(z. Since the local homology groups are invariant under homeomorphism. a) Compute H .((11 0. To compute Z*(S). 1)) = s and W2 = s .{ 0)).0. it follows that f(0) = 0.(1. It is easy to see that U is contractible.((O. Let S = S 2 n X .O.(0)) = H.(0)) M H2(X. Then S is a deformation retract of U .(0)) M H.(0)) for any q.O.0). Applying the MayerVietoris sequence of the pair (W1.
{ p } ) @ @. 2310 Let the real projective plane R P 2 be embedded in the standard way in the real projective 5space R P 5 . Since D is contractible..{ p } ) x Hq(S1).{ p } ) 3 Hq(S1x R1 .{ p } . ( R P 5 / R P 2 )x H q ( R P 5R P 2 ) .+ H q ( D. RZ which is split exact.. the nontrivial part of the MayerVietoris sequence is 0 + g1(D. (S1 = 1 . S1x R1 . we see that H.(D) f gq(S1X R1) + H g . A  b) Since S1 x R1 and D .{ p } .{ p } ) = otherwise. it follows that Hg(S1x R1)x H.{ p } ) 3 x R1 .sphere) b) Use a) to calculate the homology of S1x R1 .125 2309 a) Write down the MayerVietoris sequence in reduced homology which relates the spaces S1x R1.{ p } have the same homotopy type with S1. and a disk D in S1 x R1 about the point p. a) The MayerVietoris sequence is . (Indiana) Solution.(D . Thus we have %I iYi(S1 x R1 .{ p } ) + *.. (Indiana) Solution. Since R P 5 is compact Hausdorff and R P 2 is a strong deformation retract of a compact neighborhood of R P 2 in R P 5 .{ p } ) f gl(S1 x R1)+ 0 . . where R P 5 / R P 2 is the space obtained from R P 5 by identifying R P 2 to a point. for any q.l ( D . Compute Hq ( R P 5 / R P 2 ) .
Then the point 0 = (0.*. To compute the local homology groups Hi(Y x R.Y x R .(RP5)= { z. Therefore Hi(Y x R .1. &. 0. Solution. and H.5. = 1. Q q Thus.2.(RP2) = otherwise.  because U is contractible.*. (Indiana) Prove that Y x R is not homeomorphic to R 2 . 2 3 ) E R 3 1 2 1 2 2 = 0 . 2 22 0 ) . 0. we take an open neighborhood of the point 0. { z. where D3 is an open ball centered a t 0. 0.z2) E R2 I 2122 = 0 and 22 2 0).it follows that R H.{ 0 } ) . otherwise. q = 0. otherwise. = 0. &. from the exact sequence of the pair ( R P 5 . 4 = 0.5. P 2 ) .O) E Y x R. Let Y x R = { ( 2 1 .3.{0}). U = D3 n (Y x R ) . 2311 Let Y = {(z1. q = 1. (RP5/RP2) = a. i = 0. 0.126 It is wellknown that H. Let X be the space as shown in the following figure . 2 2 .(0)) M Hi(U. U  (0)) M Hil(U . Y x R .
Since B is a strong deformation retract of U . By the excision property.(0)) EZCBZ.(Y/B.(?r(U). X U C A is obtained from the subset X x (0) U A x [0.yo).(0) and that H I ( X ) X Z @Z. Since the local homology groups are isomorphic under homeomorphism. Let Y = X x (0) U A x [0. Then X U C A = Y / B . the inclusion map of pairs induces an isomorphism H * ( Y / B . 2312 Let A be a nonempty subset of X and X u C A be the union of X with the cone of A. Let V = A x [i. Y O ) = H*(Y/B.?r(U)) H.v.l(?r(U). ( X U C A ) is isomorphic to H.* + Hn(?r(U).1]and B = A x (1).u . *.YO)+ H * ( Y / B .(X. for every n.U ) .it follows from the exact sequence that .?r(U)).127 It is easy to see that X is a deformation retract of U .YO) * * * it follows that H.yo) = 0. Thus H z ( Y x R.R2 . (Indiana) Solution. Hence. it follows that Y x R is not homeomorphic to R2.( p } ) GZ. 1 Prove that the reduced homology group g . It is wellknown that for any point p E R2 the local homology group H 2 ( R 2 . U ) M H * ( Y . we have an isomorphism 11.1] c Y .?r(U). y o ) + + + H. that is. Then yo is a strong deformation retract of ?r(U). 1 of 1 X x [0. Thus.B ) = H * ( Y . in the exact sequence of the triple ( Y / B . Let ?r : Y + Y / B be the identification map and yo denote the point ?r(B)in Y / B .Y x R . 1 by identifying A x (1) to a point. H * ( Y . Let U = A x [$.V ) . H * ( Y . A ) .
u V). yo) Since Y admits obviously a strong deformation retraction onto which maps B onto A x { 0 } . we have X x (0) H.) t I f f : X + Y is a map.yo).: (Y  v. x Therefore. (Here Hi denotes reduced singular homology.) (Indiana) .n(V)) to give an isomorphism Now the restriction of the map n.128 Thus. ( X . + of (i) Prove that if f : X + Y is a homotopy equivalence then so is Cf.v. s) if s = t = 0 or s = t = 1. H. A ) for every 2313 n.gives a homeomorphism of pairs n.. where denotes the equivalence relation generated by requiring that (2. Dimension) for a homology theory prove that Hi (EX) naturally isois morphic to Eil(X). we have H * ( Y . (EX is the quotient space (X x [0. Exactness. For any topological space X let EX denote its (unreduced) suspension.u  V )+ ( Y / B .(Y. and so an isomorphism of their homology groups. it follows that H*(Y/B.(V)). Excision. ( X UCA). t ) (f(z). In a similar way the set T ( V )may be excised from the pair ( Y / B .T ( V ) T ( V ). ( X .(X U C A ) is isomorphic to H .   (ii) Using only the EilenbergSteenrod axioms (Homotopy. let Cf : C X + C Y be the map of suspensions induced by the map ( ~ . All of these combine to give an isomorphism H * ( Y / B .1]/ . ) (y. M H * ( Y . B ) M H .t) X x [0. from the exact sequence of the pair ( Y / B .yo)H * ( Y / B )= G . A ) . B )M H * ( Y .1].B ) . . On the other hand.
O) = f(z) and G ( z .(V) 0 for all i. Then v It is easy to see that U .t ) 1 t 2 2/3}. 3 It is obvious that U and V are both open sets of EX and that U and V are both contractible.O) or (z. Therefore Cf is homotopic to Cg. It follows from = the homology sequence of the pair (EX. By the definition we see that Cf = rIy 0 (f x id) 0 IIXl. if Z = po or p l .t)1 t > } 3 and 2 = {rIx(z. Let c U . u .'(Z) means any point (z. and by po. Thus we conclude that k i ( C X ) k*l(X). 1).W). It is also easy to see that H ( Z . and consequently it follows that if f : X + Y is a homotopy equivalence then so is Cf. for any (2. On the other hand. l). 0) = Cf(Z) and H ( E . (i) Denote by rIx : X x I + C X the quotient map. Suppose that g : X + Y is another continuous map which is homotopic to f by a homotopy G : X x I + Y such that G(z. By the excision property we have W = {IIx(z.z E X .W ) M E&'(U .i(U) = g. 1) = Eg(Z).1) = g(z) for any z E X . (ii) Let 1 u = {rIx(z. U ) . .t) 1 t < }. s) = E(II&'(%).W has the same homotopy type of X . U ) that fii(CX) M Hi(EX.s) E EX x I . Hence r?. where.W ) we see that Hi(V. where id : I + I is the identity map. 0) and (z. Then define : (X x I) x I + CY by c s) Let H : EX x I t CY be a map defined by H ( Z . Let f : X + Y be a continuous map. from the homolopy sequence of the pair (V. It is easy to check that H is welldefined and continuous. pl the equivalence classes of (z. lI. U .129 Solution. respectively.
H 1 ( U ) = 0. and the inclusion U n V + V induces a surjective homomorphism H1(UnV) + H 1 ( V ) .a 2 . It means that kerag = 0.a2 E kerP2.. (Here H denotes singular homology. and there is a bl E B1 such that P l ( b 1 ) = a(a1) . .) a) Prove that H 1 ( X ) = 0. Let a3 E A3 such that cyg(a3) = 0. Since PO is surjective. there is a bo E Bo such that Po(b0) = b l . Suppose that y and Po are surjective. Then from the exactness there is an a2 E A2 such that a z ( u 2 ) = a3. Since y is surjective. so 2315 Suppose that a topological space X is expressed as the union U U V of two open pathconnected subspaces such that U n V is path connected. (Indiana) Solution. which means that a l ( a 1 ) . Hence by the commutativity we have a l ( a 1 ) . i.a2 = 0. there is an a1 E Al such that y ( a 1 ) = /32(a2). Once again by the exactness we have 0 = PlPo(b0) = P l ( b l ) . Prove that a3 is injective. a3 is ) injective. Hence a3 = ~ 2 ( a 2 = a 2 a l ( a l ) = 0.e.130 2314 Consider the following commutative diagram of abelian groups in which the row and column are exact sequences.
H 1 ( X ) = 0. we see that 20 E D" . i.9 . a) Consider the MayerVietoris sequence of the pair ( U . ) n ) 5 ) n Since U n V is pathconnected. the unit 2sphere7 u = {(zclg. Suppose that f : D n t D" is a continuous map such that the restriction of f to S"l is a homeomorphism from S"l to s n .V ) A .e. z ) E s2I z < }. So the homomorphism H z ( U n V ) + H z ( V ) induced by the inclusion map is surjective.z) E s2I z > . Eo(U n V )= 0. We use the reduction to absurdity.. But H 2 ( X ) M Z. Suppose that there is a point 20 E D n such that zo @ f(D"). Prove that f is surjective.131 b) Give an example t o show that the analogous assertion becomes false in general if H is replaced everywhere by Hz.1)sphere. V and U n V are pathconnected and that H z ( U ) = H2(V)= 0. y. 1 (Indiana) Solution. It is obvious that U .5 ) and 1 1 v = {(z. b) Let X = S 2 .1 . 2 Then X = U U V . (Indiana) Solution. It means H 1 ( X ) = im$*. By the assumption. and denote the standard (n.l . .. By the hypothesis.. 2316 Let D" = { z E R" 1121 5 1) denote the standard unit ball in Euclidean space Rn. + H ~ ( x 3 H ~ ( U v)% H ~ ( u CB H ~ ( v )H ~ ( x + H ~ ( U v). d* is surjective so that Therefore im& = 0.
and consequently (g o i)* is trivial.l ( D n ) is t obviously a trivial homomorphism because of Hnl(Dn) = 0. + X ( X ) = rankHo(X) . X 1 )is free abelian. X 1 )= Z $ Z @ Z . and i.( 2 0 ) + Sn' be a retraction. Assume H1(X) x Z @Z/3.132 Therefore Sn' is a deformation retract of D" .X ' ) 3 H 1 ( X 1 ) 2. 2317 Let X be a connected CW complex with two 0cells.XO). Therefore rankHz(X) = 2. H2(X) Z Z $Z. Thus because H 2 ( X 3 . Therefore g o i : Sn' + Sn' is a homeomorphism. and H 2 ( X 2 . : Hn1(Sn') H n . H 2 ( X ) is also a free abelian group of rank 2.3 3 = 2. is the composition of homomorphisms H 2 ( X 2 .1 : 9 .rankHl(X) + rankH.X O ) } . i. Let T : D" . X 2 ) 0. which is a homeomorphism.l + S"' is just the restriction of f to Sn'. This is a contradiction. three 1cells. Hence (g 0 i)* : Hnl(S"') + H. three 2cells. The Euler characteristic of X . On the other hand.(X). where d. : H2(X2.. Then g = r o f is a continuous map from Dn to S"l and gIs.e. Therefore = H 2 ( X ) M ker{d. Let i : S"l + Dn be the inclusion map. But ( g o i)* = g* o i.X ' ) t H 1 ( X 1 .Compute the Euler characteristic of X and determine (with proof) all possibilities for H2(X). we see that H 3 ( X 3 . (Indiana) Solution. H1(X1. X ( X ) = 2 . . By the assumption. and no higherdimensional cells.'(S"') is an isomorphism.(20). Let X k denote the kskeleton of X .
Z) M ker dk+1/imdk+2. Thus we have d'+'(a) # 0. Therefore # 0. X k .x k . It follows that kerd"' = 0 and. and.133 2318 Let X be a CW complex with exactly one k+lcell. X k ) . H ' + l ( X k + l . Z )M ker dk c H k ( X k . (Indiana) Solution. Z )is not a nontrivial finite group and contradicts the hypothesis.l . We choose a generator a in H k + l ( X k + l . kerdk is trivial or free abelian. It means that H k ( X . Hk+l(X. i Y k ( X . But it is wellknown that H k ( X k . Z ) M ker dk/imd"' xk2). and by dk the composition of homomorphisms of pairs H k ( X k . consequently.X"'). and Hk+l(X.Z) is a nontrivial finite group then Hk+l(X. By the hypothesis.1 ) It is wellknown that 3 Hkl(Xkl) 5 H k T l ( X k . X k ) zz. 2319 Let and . therefore. We claim that &+'(a) Otherwise we would have imd'+l = 0.Z) = 0. Denote by Xk the Icskeleton of the C W complex X . H k ( X .Z) = 0.l ) is a free abelian group. Prove that if Hk(X.
2. 0. 0. O . other wise. (Indiana) Solution. From the homology sequence of the pair ( U . 0. (See Fig. 2: (See Fig. It is clear 1 that both C and D are homeomorphic to the unit sphere S2 and that D and C are deformation retracts of W1 and Wz respectively. 0.O)) .p2) and W2 = C U D . where and D = { ( ~ 1 . u . 2. O .( O .(O. Let 6 be the unit open ball centered at the point ( O . 2 2 . O . U . 2 4 )E R4 I 2 4 = 0.p4}. W2).O)) for all a. O . Therefore. applying the MayerVietoris 1 sequence to the pair (W1.O.x  (O. 0) in X . 0.O)) = Hi(U . 0. O .134 Let X = A u B.(0. O .15.{pl.O)) for all i. Then U = 6n X is an open neighborhood of ( O .2.u .O) has a deformation retract C U D. It is also clear that S1 is an deformation retract of W n Wz.O.) Then W1 and W2 are open subsets of C U D and W u W2 = C u D. Compute the relative homology groups H i ( X . It is obvious that U is contractible and that U . 2 3 . O ) in R4. Fig.(0. By the excision property we have &(X.{p3.O)) M H i ( U . O .15 Let Wi = C U D . 0)).(O.O. we see that Hi+l(U. 0.(0. we see that 2 = 0.15) + 2: + 2: = 1). X . 0.
Since ?rl(X. I ) for all i different from 0 and 2. where R : R2 x S 2 + Y is the covering map.Z) is trivial for i different from 0 and 2. Show that if al(X. Hence we have  7 f* = R. must . f(z)) be trivial.135 for all i 2 0. (Indiana) Solution.O. we may lift f t o a map : X + R2 x S 2 such that a o f = f . Hence we conclude that Hj(X. It is easy t o see that R2 x S 2 is the universal covering space of Y .1 otherwise. 0 f* : H l ( X . so is . then any continuous map f : X + Y induces the trivial map f* : H i ( X .O. Therefore. 0)) = i=o. Z ) +  + Hi(Y. Z ) for any i. But it is obvious that : H i ( X . we have It is clear that Rl(S1 x s1x S1) = “l(T2)@ “ l ( S 2 ) = a $a. Let Y = S 1 x S1x S2.X . 2320 Let X be a path connected and locally path connected space and let x E X .x) is finite. Z ) + H i ( Y . By the Kunneth theorem.(O.Z) Hi(R2 x S2. the homomorphism f* : R ~ ( Xz) + R ~ ( Y .z) is finite.
t F Prove exactness of (Indiana) Solution.136 2321 Call a commutative diagram of abelian groups “exact” if the sequence of groups and homomorphisms o + A (9)c ’‘ B c~ 5D It is an interesting fact that if A Z 6  o is exact. then so is A C ’a l . we have 6P = 7 a and &a’= 6’7. B P1 and C + D 17 B S E 71 1 E D .F 6’ are exact. . By the assumptions.a’a) 6’6P = .Therefore.6’ya = 0.&a’a= 6’ya .+ E 1 E PI 6’6 C .&)(P. we have (6’6.
) = b and p(a) = c . We define 1 1 Fn+l = f(Fn) inductively for n E Z+. i. and. Since f ( A ) = A iz xEA and zo E A .a’a) c ker(6’S . We claim that f ( A ) = A holds. i. there would exists an z1 E A such that f(z1) = 2 0 .137 It follows that im(P.E ) . we see that the subset A = n=l n F. consequently.a’). n E Z+} is a sequence of nonempty closed subsets in X and that Since X is compact. It is clear that { F n .1] and f : X + X is defined by f(z) = for z E X.e. (Indiana) Solution.S’). Let X = (0.. Give an example to show that compactness is essential for this assertion. p ) = ker(y . a’a)(u)E im(0. Let ( c . which is a contradiction. By the assumptions. Therefore. ( c . Therefore zo 2 z1 = 220.E ) c im(P. there exists an a E A such that a(.6). 20 1 = 521. Then there would exist an 2 0 E A such that z 5 zo for any z E A. Then F is a nonempty closed subset of X. 2322 Let X be a nonempty compact Hausdorff space and f : X . It means that ker(6’6 . Thus.y) = ker(c . We give an example to show that compactness is essential for this assertion.e. 00 is a nonempty closed subset of X . S’6(c) E ( e ) = 0. The exactness at C @ E is proved. a’a). we have i m ( a . Hence we have (b. Let F = f ( X ) .e ) = (0. Prove that there exists a nonempty closed subset A of X such that f ( A ) = A .E ) . . 6 ( c ) )E ker(E . e) E ker(6’6 .+ X be a continuous map. In fact z o = sup z. noting that ( e . a’a). c ) E ker(y . we see that there exists a b E B such that a’(b) = e and y(b) = 6 ( c ) .6) and im(a’.. Suppose that a nonempty closed A of X satisfies f ( A ) = A .
U and V are both contractible.O.d e g f .> I ZY = 01. Let f be a function from S3 to RP3. Denote by A and B the unit circle in 7rl and a2 respectively. Then we have f*(a)= 7r*(f*(a)) d e g f .o. l ) and V = A U B . (b) Suppose that there exists a homeomorphism f : X + R2. 0) is also a homeomorphism.o) : X . Take U = A U B . (b) Using part a . Since 7r1(S3)= { 0 } .( O . Applying the MayerVietoris sequence to the pair ( U . r .(O. we see that H1(X . Then it is easy to see that A U B is a deformation retract of X . (a) Compute H1(X . 0 ) + R2 . =  I because deg f E Z. Then it is clear that the degree of 7r is equal to 2. Let 7~ : S3 + RP3 denote the universal covering map. (Indiana) Solution. 1).Z) Z .O. (a) In fact.it is obvious that f* is not an isomorphism.(O. Z )x H3(RP3.O. Therefore. O . Denote by a and b the generators of H 3 ( S 3 ) and H3(RP3)respectively.(0. ( a ) = 2 . .138 2323 Recall that H 3 ( S 3 .(O.f ( O . there is a lifting of f. f : S3 + S3 such that 7 r f = f .) Prove that there is no function f : S3 + RP3 inducing an isomorphism on the third homology. O . 0 ) ) .V). (c) Prove or disprove: X is homotopy equivalent t o R2. 0 . O ) ) x H 1 ( A U B ) =Z $Z $Z.(0. show that X is not homeomorphic to R2. (Indiana) Solution. Then flx(o. M (RP" is ndimensional real projective space. 0). 2324 Let x = ((2. which . U U V = A U B and U n V has four path components which are all contractible. It is wellknown that S3 is the 2fold universal covering space of RP3. 0.Y. b . X = 7 ~ 1 7r2 where 7r1 is the plane y = 0 and 7r2 is the plane U z = 0.
d. = Z.l ) in which the level rows are exact and the squares are commutative.(X.l ) be the standard ball and sphere pair in R". (Indiana) Solution. + H. . 0)) to H1(R2 . and therefore It is wellknown that 8: is an isomorphism and H n .i(X) t f* a: t fl* i : + t f* Hnl(B") Hn(B".l : S"l + A is a homeomorphism. Since H . f1. O .l ) + ( X . Show that if & ( X ) = 0 then H . A ) is a continuous map and f l s n . Thus we have keri. 0)).S"1)+ H " . Consider the following diagram H . is an isomorphism.A ) M imd. Hence X is 2325 Let ( B " . n > 1. By the assumption.l ( S n . Then F is a deformation retraction of homotopy equivalent to R 2 . A ) = Z .S n .S n . X onto the plane 7r2. ( X ) 1?.H. = H . ( X .A) a. (c) Let F : X x I +X be a map defined as follows. O .139 wouid induce an isomorphism from H1(X .( O . ( X . But it is clear that It is a contradiction. .i(A) i* t H.l ( A ) = Z . Hence we have H . is a monomorphism. ( X ) = 0.l ( B n ) = 0. M ker i.f ( O . Suppose that f : ( B " .
Hence it is obvious that H i ( S 2 x S 5 ) is an infinite cyclic group for i equal to 0. (a) Let xo E S 2 . Then S2 is obtained by attaching a 2cell t o 2 0 .6.{ p 1 } a n d V = S 2 x S 5 . ( S 2 x S2)are (no proof is required). Applying is the MayerVietoris sequence to the pair (U. Therefore the CW structure on S2 x S 5 has a 0cell. and UUV = S 2 x S5.{ p l . ( S 2 x S2.140 2326 (a) Describe a CW structure on S2x S5 and use it t o compute the homology of s 2 x s . x 1 E S 5 . 5. (Indiana) Solution. (b) Let X = S 2 x S 5 .A)? (Indiana) Solution. p 2 } . and 7. which can be considered as the space obtained by attaching a 2cell and a 5cell to the point (20. V are both open sets of S2x S5 and UnV = X. . a 2cell. Then U . It is easy to see that S 2 x S5 is homeomorphic to the space obtained by attaching a 7cell to S2V S5. and S5 is obtained by attaching a 5cell to XI. It is easy to see that S2VS5 a deformation retract of both U and V . otherwise. 2327 Let A be a subspace of S 2 x S2 homeomorphic to the 2sphere S2. U = S 2 x S 5 . i = 1.2. Denote by S2 V S5 the one point union of S2 and S5. State what the homology groups H . By the Kiinneth formula we conclude that otherwise.5. and is zero otherwise. a 5cell and a 7cell. What can you say about the possibilities for the relative homology groups H . ez e z ez. 21) E S2 x S5. V) and using the fact that we conclude that i = 0. 2. 5 (b) Compute the homology of S2 x S5 with 2 points removed.{ p 2 } .
using the above result and the fact that H .2 and is zero otherwise. is the homomorphism induced by the inclusion map i : A S2 x S 2 .X”) conclude that Hi(X) infinite we is cyclic for i equal to 0. ( S 2 x S 2 . H.1.A ) M Z @Z/irni.A ) M keri. a. By the exactness. A ) .) Show that there is a natural isomorphism ax : H. (Indiana) Solution.l) in the xaxis}.A ) + H o ( A ) + Ho(S2 x S 2 ) + H o ( S 2 x S 2 . x we have H3(S2 x S 2 . 4 (Indiana) . X”) infinite cyclic for i equal to 1 and is zero is otherwise. where i. Let X’ = X . (This is the union of two “cones” on X. A ) +0.. 2329 Define the “unreduced suspension” CX of a space X to be the quotient space of I x X obtained by identifying (0) x X to one point and (1) x X to one point. ( A ) M Hi(S2). It is clear that Hi(X. A )+0 0 + H 1 ( S 2 x S 2 . A ) 4 H 2 ( A ) 5 H 2 ( S 2 x S 2 ) 5 H 2 ( S 2 x S 2 .(X) H i + l ( C X ) . Thus from the homology sequence of the pair ( X .we see that H .(S2 S 2 . Since &(A) = 0. A )M H i ( S 2 x S 2 ) for i 2 4 and that 4 the nontrivial parts of the sequence are 0 and H 3 ( S 2 x S 2 . and H 2 ( S 2 x S 2 . A ) = 0 for i equal to 0 and 1.   2328 Compute the homology groups of the space X obtained as the union of the 2sphere S2 and the xaxis in R3. for all i 2 0. It is easy to see that X’ has the homotopy type of 5’. Then the space X may be viewed as a space obtained from X* by attaching a 1cell.141 From the homology sequence of the pair ( S 2 x S 2 ..{the open interval (1.
) : V .. (a) Consider the homology sequence of the pair ( X . Consider the MayerVietories sequence of the pair (U.142 Solution.where z f * is the homomorphism induced by the adjunction map f : D 2 + X . ( X . d D 2 ) ) Z .2. It is wellknown that H . V s l +~ A ++ +  It is clear that U and V are contractible and that U n V has the homotopy type of X . Let (0)’x X and (1) x X be identified to the points 20 and 2 1 respectively.. Take U = EX . 0 4 H 2 ( T 2 )5 H 2 ( X ) 1 H 2 ( X ..) 20 Fig. The inverse of A * . T 2 ) . ( X ) .. T 2 ) . Then U and V are open sets of E X .(21). H .. 2330 Denote by X the union of the torus S1x S1with the disc D 2 .16 (a) Calculate H.a D 2 ) 4 w H l ( d D 2 )+ 0 . + ~ ( U ~ 3 )H i + l ( u > ~ H i9 H(~~ > ( E x ) H i ( u n v ) .( 2 0 ) and V = C X . CTX : & i ( X ) + H i + l ( C X ) is just what we are looking for. + a: H 1 ( T 2 )2 H 1 ( X ) + 0 Qt f* t fl* 0 + H2(D2. The naturality of ax can be derived from the naturality of the MayerVietories sequence. Thus the homomorphism A : k i + l ( C X ) + k i ( U n V ) is an isomorphism and we may obtain an isomorphism A* : H i + l ( C X ) t & . where 20 E S1. T 2 )= 0 for i # 2 and H 2 ( X .(X) for all n 2 0. Thus the nontrivial part of the sequence is as follows. (See below. and U u V = EX. T 2 ) M f * ( H 2 ( D 2 . (b) Is T 2 a retract of X ? Why or why not? (Indiana) Solution.. where D 2 is attached to T 2 by identifying dD2 with a meridian curve S1x (20) in the torus.
and. and ker& = ker f 1 + . Thus we see that H 1 ( X ) x Z @Z/Z M Z and H z ( X ) M H 2 ( T 2 )x Z . the pull back of h via g is the group g * ( A )= { ( c . + (a) Let g*(h) : g * ( A ) + C be the homomorphism obtained by restricting the projection onto the first factor C x A C to g * ( A ) . Identify the group g * ( A ) . 9*( A ) c 1 g*(h) A l h . we have T* o i.with explanation. We know that H 1 ( T 2 ) = Z @Z is generated by a and b. In other words. consequently.+ B 9 (b) Let A = Z / U . (See Fig.143 It is easy to see that ima: = irnfl. This is a contradiction. Let i : T2+ X be the inclusion map. a )E C x A 1 g ( c ) = h ( a ) } .2. Hence we conclude that (b) We claim that T 2 is not a retract of X. Prove that the kernel i of g*(A)is isomorphic to the kernel of h. = id : H 1 ( T 2 ) H 1 ( T 2 ) . B = 2/Z. there would exist a retraction map T : X .16) It is clear that f 1 * ( H l ( a D 2 ) )= 2 @ (0).+ T 2 such that ~ ( =pi d p . we have the following commutative diagram  because of H 1 ( T 2 ) = Z @I and H 1 ( X ) =Z. Then r o i = i&z : T 2 + T 2 . and let h : A + B be the surjection defined by h(1) = 1. Otherwise. Let C = 2/82 and let g : C + B be the surjection defined by g(1) = 1. (Indiana) . 2331 Given homomorphism h : A B and g : C + B .
a ) = a. (b) The homology of Y = R3 .2).4. B are open subsets in S3.0).C. I). and. and A U B = S 3 and A n B = R3 ..1>.~ .{m}. therefore.2}. then H i ( S 3C) is infinite cyclic for i equal to 0 or 1 and is zero otherwise. u ) E kerg*(h). h'(l) = {1. . (4. 2332 Recall that if C is a homeomorphic copy of the circle in S 3 .144 Solution. when C is a homeomorphic copy of the circle in R3.. the onepoint union of two circles. (2. (3. (a) Suppose that ( c .3}.0).0). + H.C ) = .1). and Hi(S3 . (a) Denote S3 = R3 U { }Let A = S3 .2)..C ) % Hi(S3 . g'(O) = {0.3).a) I a E kerh}.O). Let F : kerg*(h) + ker h be defined by F(l. ( C +  Noting that Ga(S3 . C = l.C.3).2..e.X . Then g*(h)(c. the identity of C.) (Indiana) Solution. By the definition of g*(A).R )~+ .7}.3). ( L 3 L (3.. where X c R3 is a homeomorphic copy of the "figureeight space" (i.+'(S3) 2 Hi(R3.(7. It is easy to see that F is an isomorphism. (4. (b) In this case.. otherwise. Assuming this fact compute (a) The homology of R3 . We have the following MayerVietoris sequence. we see that kerg*(h) = {(l.{ }= G@3) = 0 a ) for any i and that H i p 3 )= { z. (1.3. (6. (6. I>.. 0.B = S3 . Thus it is clear that the group g*(A) consists of the following 16 elements: (O. .C ) @ Hi(S3 ..2). we have h'(O) = {0.C.5.{co}) % Hi(s3)A H ~ . Then A and a . (5. (2. (7.2).6} and g'(l) = {1. (5. (0. i = 3.a) = l.
. Since the representative chain c of [c] can also be considered as a chain of (R3 . Take a sufficiently large T > 0 such that the sphere S 2 ( ~belongs to R3 . %A7 @Z  MZ + A We claim that the homomorphism is an epimorphism.V )is infinite cyclic for i equal to 1 or 2 and is zero otherwise.V ) .by the definition of ++ we have [c] = + + ( [ c ] . ~ ~ From the result of (a). Then we see that U and V have homotopy type of S 1 and that U n V is contractible.(u nv). The claim is proved.V )5 i i 2 ( ~ u n V > 2 U 3 3 %Z$Z H 1 ( Y ) 2 ii1(R3 .( ~ 2 ) . It is easy to see that iii(x3 (v n v))M iri(s2)>.U.2. otherwise. R3 .X .1. we have Y = R~ . Then S 2 ( r )is a ) deformation retract of R3 .145 we can see that  = {: i = 0.17 Represent X as shown in Fig.0).U ) @ ii1(R3 . Hence we can consider the generator of H 2 ( S 2 ( r ) ) .2. the nontrivial part of the MayerVietoris sequence of the pair (R3 .V ) is o + ii2(y) i i 2 ( ~ 3 ) CB i i 2 ( ~ . Thus we get H 1 ( Y ) = Z@Z and a split exact sequence   O + & ( Y ) 5ZCBZ5Z+O.{PI} and V = X .U ) and gi(R3.U n V).2. Let U = X .u) n ( R ~v) ~ and ( R .V ) + 0.U ) n (R3 . we see that Zi(R3.v )= R~ . Since U u V = X .[c].u )u ( R .x = ( R.U n V .17. Due to the above facts.as the generator of E2(R3 . Fig. which means imA = 0 too.
A ) M Hq(S3. Denote Y by Y = A U B.(S3 . i = 0. Therefore. . ii. ( S 3 . A  A  &(s3. H . . * ..ii. q#3. 4.X ) M H.B) H . . where A n B = 8 and both A and B are homeomorphic to S1. Using the fact that Z. .+l(~3) .B ) M H. (Indiana) Solution. H. 4=2.A ) U ( S 3 .Y where Y is a subspace Use of S3 homeomorphic to the disjoint union of two copies of S1. 4 = 0.H .Y )5 . 4 = 1.(S3) = we can easily see that { Z.146 It follows that HZ(Y) =Z. this fact to compute the homology of S 3 . ( s ~ .Y )=  { 2*Z.A ) n ( S 3 . i = 2.(S1) = and that 4 # 1. 3.(S1).B ) . ~ Z. S3 . . It is known that if X c S3 is homeomorphic t o S1 then H .Y ).B are open sets of S3 and ( S 3 . 4'1.~ ( s .Y = ( S 3 .A ) @ &(s3 .So we conclude that H l ( Y ) = {:w Z7 I 2333 i 2 3. Noting S3 . ( s ~ ) .A and S3 . i=l. 0.B)= S 3 .(S3 . in the MayerVietoris sequence we have *.
restricted on R. Thus. The connected 2fold covering of S1v S2 is shown as follows..147 2334 (a) Sketch pictures of the universal covering of the one point union S1v S2 and of the connected 2fold covering (no proofs required). The covering map T .{the antipodal point of p 2 } (see the above figure). and. the 2fold covering map: z + z 2 from S1to S1.2.18 The universal covering of the one point union S1VS2 is shown as in Fig.19 The covering map T . and. we take U = S2 V S1V S2 . restricted on each S2. (b) Compute the homology of the connected 2fold covering space of S1vS2.. b) To compute the homology of the connected 2fold covering space of S1v S 2 . is the exponential map: R + S1. Then it is easy to see that S1is a deformation retract of U n V .2. it is clear that . the homology of S2 V S1V S 2 .2. restricted on each S2. is the identity map.is the identity map. and that S1V S2 is a deformation retract of U and V. reis stricted on S1. (a) . (Indiana) Solution.e.18. Fig.{the antipodal point of p l } and V = S 2 V S1V S2.. Fig. i.
: H1(U n V) + k 1 ( U ) and I .(S1) = Therefore. 4 # 172. i. the nontrivial part of the MayerVietoris sequence is c Z.. It is clear that   H ~ ( v s1v s2) H ~ ( u ) kl(V)/im#. It means that H z ( S 2 V S 1 V S 2 ) ~ k e r A z i m $z H 2 ( U ) @ H z ( V ) = Z @ Z . s ~ M Hence. where A and B are each homeomorphic to S1and .148 and &(U) M &(V) x H.(S2   v s1v S 2 ) = 4 = 1. : G1(u n v)+ H1(u)@ H1(v) is injective. Hence the homomorphism   4. 4 = 1121 0. Thus imA = 0. we have H.20 It is easy to see that the space X = A u B is a deformation retract of the space H = S1x S1point. respectively. It is easy to see that the inclusionmaps k : unv + U and I : U n V + V induce injective homomorphisms k. ker 4%= 0. 2. 2335 Compute the homology of S1 x S1point.(S2) @ H. . otherwise.e. MZ. : H1(U n V) + H1(V). o + H2(u) H2(v) H2(s2 s v s ~A )H ~ ( U v) @ 5 v 1 + n 5 &(U) @ &(V) 2 &(S2 v s1v S 2 ) + 0. (Indiana) Solution. Fig.
b } is contractible. and 7.{ u } .20 Choose points a E A and b E B such that u # z o and b # 2 0 . (Indiana) Solution.P n ( b ) for any (a’. we have nL2. respectively. ( u ) . . ( d )= A(a’. we have j .. b ) = .2. @ B. such that i. Let U = X . Then 7 ’ o j L ( u ) E ker6. ( X ) = H .{ a . Thus Pn(b).. It is clear that A and B are deformation retracts of U and V . It means that u = P. and let V = X . b) E A.lojA). Let u E ker(6. . Due to the exactness .(a’) . and consequently.(b) = j k ( u ) .u.there exists a b E B. such that jn(b) = 7 ’ o j .u E kerj. b).(b) . ( A ) @ H . . is an isomorphism for all n.... o@.. and there exists an a’ E A. i BL + Anl + * * * Write out the proof of exactness at Bh.i . and that U n V = X . Applying MayerVietoris sequence t o U and V. j n ( b ) = jA(b). From the commutativity.(a’) = Pn(b). at C. + A . the rows are exact. Hn(S1 S1. +A’. 07. Construct an exact sequence .149 A n B = {zo} as shown in Fig.{ b } . 7. We give the proof of exactness at Bh as follows. @ B.point) = H . ( B ) = x 2336 Suppose the following diagram is commutative. The following sequence is exact: where A is defined by A(a’.
.3/4). that is.~ (n v > u (1) II Since X x (0) = Y / is homeomorphic to Y and is a deformation retract of U . H. is proved.1X. and each (z. o j .(V) E H. Thus the exactness a t B. we have H. 1) g(z). Then we have M = U U V and U n V = X x (1/2. In the same way. let V = X x (f. o Pn = 7.(b) for some a E Ah and 6 E B. from the assumption that Y is contractible.1 ( U ) @ H q .1] = Z/ ’.     . II 1 I Let U = X x [0. X x I and Z by identifying each (z. + A  H . . Prove that H..(b) = 0. it is easy to see that 6 0 7’ . i. o 7 ’ ojk).e. So k . ( Y ) = 0. . ( U n V ) Z & ( X ) . Suppose that u E imA. the equivalence relation is determined by (z.3/4] and Y .. (Indiana) Solution. we have the following exact sequence: ’  * . Suppose that X is a space and f : X + Y .+ H. u = ih(a’) . M is the identification space obtained from the disjoint union of Y . ’ Since j . ) C imA. V and U n V are open sets of M .k .P. It is obvious that U n V has the same homotopy type with X . o ih = 0 and j .O) with f (z). 0 jL(U) = . where the equivalence     relation is determined by (z.(u n V ) + H J U ) @ I?p(v)+ H. .3/4] = Y denotes the disjoint union of X x [0.l ( V ) + ’. where X x [0. g : X + 2 are two maps of X into contractible spaces Y and 2. from (l). Let M be the mapping cylinder o f f and g . 0) II f ( z ) . Now we prove imA c ker(6. .l ( X ) . 6 0 j.M Z H. o 7 ’ o j .3/4] = Y/.. Noting U . .(M) + H q . Thus. 2 In the same way.(U) H . 1) with g(z).(Z) = 0. ( M ) Z k q . by MayerVietoris sequence.150 So ker(6.
Part 111 Differential Geometry .
.
3. by Blaschke. (Indiana) Solution. "t Fig. where the oval is counterclockwise orientated and 8 denotes the oriented angle from the zaxis to I.(e) = p(e)sin 6 p'(0) cos 8. 5 2 (or some other constant independent of a . Define the diameter d.) with d. Prove that d. i) The following is an example of such an a ( s ) with k ( s ) 2 1 and d. we know that the simple closed curve a is an oval. > N . 2 is the best possible such constant).153 SECTION 1 DIFFERENTIAL GEOMETRY OF CURVES 3101 Let a ( s ) be a closed plane curve. i) For any N E Z + sketch an example of such an a(. of a ( s ) to be Now assume that the curvature k ( s ) 2 1 for all s. ii) Assume further that a is a simple closed curve. y(e) = p(e) cos e pye) sine. then the oval can be parameterized by 8 as follows .y) of the oval. if we take the origin 0 as shown in the figure and denote by p ( 0 ) the distance from 0 to the tangent 1 a t the point (z. For every oval.1 ii) From the hypothesis that k ( s ) 2 1 for all s. + + . > N .
From this we can conclude that. the curvatures of C and C1 satisfy the inequality k l ( P l ) 5 k ( P ) ..2 Now we can prove a more general result of Blaschke: Let two ovals C and C1 in a plane be internally tangent a t a point 0.(a). Suppose that. respectively. Let p ( 8 ) and p l ( 8 ) be the of C and C1. the relative curvature of the oval is k. a t every pair of points P and PI where C and C1 have the same tangent orientation. by direct computation. y(8)).(8) of C. and their support functions that the support value problem of because p ’ ( 8 ) is exactly the distance from 0 to the normal line of C at point (. Then the domain encircled by C1 must contain the domain encircled by C. take the tangent point 0 of C and C1 as the common tangent line as the zaxis. p ( 8 ) can be uniquely determined by k ( 8 ) = k.3. From the above.(8) = ( p ( 8 ) + (8) ) . we can say function p ( 8 ) must be the solution to the following initial ODE origin.1.3. Fig.154 p ( 8 ) is called the support function of the oval. . Hence..3 In fact. Fig.
then ii) follows immediately from the above assertion. For convenience. . we know that the curve a must be a circle (or a part of it). Therefore. we make the reflections of the oval C and C1 with respect t o their common tangent line a t 0 and reverse the orientation of the zaxis. Then . where c is a smooth function.155 Analogously. Therefore. Determine all such curves. we also have the similar expression of p l ( 8 ) . If we take a circle with radius 1 and centered at a ( s g ) N ( s o ) as C1.4)  I  + 3102 Let a be a regular C curve in lR3 with nonvanishing curvature. we assume that the parameter t is the arc length of the curve a. Thus we can immediately obtain k ( t ) = constant. we have. Thus. (Indiana) Solution. T ( t ) = 0 and c ( t ) = k. The assertion of Blaschke is proved. by the Frenet formula. by the fundamental theorem of the theory of curves. Suppose " the normal vector N ( t ) is proportional to the position vector. Hence.p ( 8 ) 2 0. Differentiating both sides of the equality N ( t ) = c(t)cr(t)with respect to t . and take a as C . N ( t ) = c ( t ) a ( t )for all t . we have sin(8 . Secondly. p l ( 8 ) .p ( 8 ) 2 0. Noticing that every oval is the envelope of all its tangent lines.can get we pl(8) . we know that As for the sign of sin(8 . where 8 and its corresponding original 8 satisfy 8 + 8 = 27~. if 0 5 8 5 x. if x 5 8 5 2x. we see that the domain encircled by C must be contained in the domain encircled by C1.b). then owing to 0 5 4 5 8. 2 0. we always have p l ( 8 ) 2p ( 8 ) . for the curve C1. By the hypothesis. that is. firstly.
Tsinws. we obtain L(u. 3104 Let y : (a. Denote the Frenet frame by { T . M ( u . (Indiana) Solution.v) = 0 for all ( u .w)du2 + 2 M ( u . c) Show that k / r = constant = k t a n a .v)dv2 = 0. where L(u.v). a = 0. v ) are the coefficients of the second fundamental form of S at point p ( u . Lz. every point of S is a planar point. If S is a triply ruled surface. there are three different asymptotic directions at every point of s.dv) satisfies L(u.b) 4 R3 be smooth with [y’l = 1 and curvature k and torsion 7. v ) .w) = M ( u . (Indiana) Solution. L3 lying in S such that L1 n L2 n L3 = { p } . a) Show that a circular helix is an example of such a curve. N ( u . v). Therefore. . b) Show that N . S must be a plane (or a part of it). by the hypothesis.v). Noticing that the above equation is of 2nd order with respect to du : dw and it has two roots. Determine all triply ruled surfaces.156 3103 A surface S c lR3 is called triply ruled if a t every p E S we can find three open line segments L1. Assume there exists a unit vector a E R3 with T .v)dudv + N ( u . B } . In other words. both nonvanishing. hws).v ) = N ( u . Observe that every asymptotic direction ( d u . a) Let a circular helix be parameterized as follows y(s) = (Tcosws. then. a = constant = cos a. N .
k ( s ) / ~ ( s ) & t a n a . and the image of y on the sector must be one of the following three cases: a generator. d m .0)) (Indiana) Solution. Then it is easy to verify that s is the arc length parameter. two rays which start from the generator I .157 where T . b ) . Noticing that the central angle of the sector is f i r .rwcosws.b ) . c) By the property of b). an obtuse angle. h. 3106 Let y : (a. we obtain k ( s ) N ( s ). we have 0 = (cos a k( s) f sin a . Assume k ( s ) # 0.hw). T(s ) ) N (s) . with curvature and torsion k ( s ) and ~ ( s ) . then T ( s ) . Hence T ( s )= (rwsinws.T(s) f s i n a . a straight line never intersecting the generator 1. and the geodesic y becomes a straight line on the developed infinite sector.O. Differentiating the above equality with respect to 3 s. and let T and N denote the unit tangent and normal vectors to . ~ ( s # 0 for all ) s E ( a . (z)y) E R2\{0. then the cone becomes an infinite sector without the vertex.i are all constants. then we can conclude that y never intersects itself.a = 0.b) 4 R3 be a C" curve parameterized by arc length. Thus. b) Differentiating T e a = constant with respect t o the arc length parameter s. w = ( r 2+ h 2 ) . we may assume that the constant vector a = cosa. a = hw = constant. l). for all s E ( a . B ( s ) . If we cut the cone along a generator 1 and develop it into a plane. from which follows N ( s ) . = 3105 Show that if y is a geodesic on the cone t = then y intersects itself at most a finite number of times.a = 0 for all s E ( a . If we take a = (O. b ) .
(b) From (a). dZ ds + + + Hence. The curve y is called a Bertrand curve if there exists a regular curve 7 : (a. 7 is called the Bertrand mate of y and we can write y(s) = y ( s ) + T N ( s )for some T = ~ ( sE) IR.158 y. ds dz ) Therefore. by (b). we have that r’(s) = 0. 1 r 1k( s ) r2k(s) + + ClT(9) = 1. (Indiana) Solution. namely (1 .C2 such that for all s E ( a .r ~ ( s ) = const.= a’(s)T(s) ( a ( s ) k ( s ) b ( s ) . In this case. b) + IR3 such that for each s E ( a . ) (c) Prove that if 7 has more than one Bertrand mate. then y is a circular + helix. (b) Prove that if y is a Bertrand curve (with r as above). . rg.b). Then. r ( s ) ) N ( s ) b’(s)B(s). + C27(5) = 1. we obtain  k ( s ) N ( s ) .b).r k ( s ) ) = const. (Note that s might not be an arc length parameter for 7.) (a) Prove that r is constant. (a) From y(s) = y(s) + r ( s ) N ( s )it follows by differentiation that Taking inner products at both sides with N ( s ) = fN(s). now we have  T ( s ) = (1 ds ds rk(s))T(s)  ds TT(S)B(S). C . there exists a constant C such that r k ( s ) C T ( S = 1 for all s E ( a . + then by differentiating with respect to s . ds dz . from N ( s ) = fN(s) we know that a’(s) = b’(s) = 0.b). b ) the normal lines of y and 7 a t s are equal. then there exists a constant C such that r k ( s ) C T ( S= 1 for all s E ( a . ds If we denote  T ( s )= a ( s ) T ( s ) b ( s ) B ( s ) . namely r = const. there exist constants T I . (c) Suppose that y1 and 7 2 are the Bertrand mates of y.
Differentiating (z(s). which means that b ( s ) = 2 At last.N ( s ) ) = 0. Assume that ~ ( s # ) 0 and k’(s) # 0 for all s. T ( s ) ) 0 with respect to s again.k’(s) k 2 ( S ) T ( S) * Conversely. Then we may assume that Z(S) = a ( s ) T ( s ) b ( s ) N ( s ) c ( s ) B ( s ) . Show that a necessary and sufficient condition for z(s) to lie on a sphere is that 1 +. the above system of linear algebraic equations has solution k ( s ) = const. c ( s ) are suitable functions to be ascertained later. where { T ( s ) . 12 are not equal. 3107 Let z(s) be a curve in R3 parameterized by arclength. t ( 8 ) . Hence y must be ) a circular helix. z(s)) = R2 with respect to s. ( s ) . and a ( s ) . Vs. N ( s ) ) = & with respect to s. Suppose that z(s) lies on a sphere centered at the origin. from which it follows that u ( s ) = 0. we have ( z ( s ) .159 Because the nonzero constants T I . Differentiating ) ( z ( s ) . we obtain namely. b ( s ) . c(s) = .B ( s ) } N is the Fkenet frame field along z(s). differentiating . T ( s ) = 0.P ( S ) 1 +(s) P ( S ) = constant. (Indiana) k4(s) S o ht ion. we obtain = + + 1 + k ( s ) ( 4 s ) . still differentiating (z(s). ~ ( s = const.
m . which is parallel to the zaxis. nor its principal normal. Observe that the geodesic curvature of a curve on M can be computed as follows k . . Then. the curve c ( t ) is not a geodesic on M . 1) (“top circle”). we have that vanishes identically because k’(s) # 0 and Then is a constant vector. because neither the top circle is a straight line.160 with respect to s. denoted by m. is parallel to the normal of M along c ( t ) . z ( s ) . y. 3108 obtained by rotating the circle ((0. Let M + c R3 be the torus (Indiana) Solution.m) = constant. and @(t) the angle between the is normal of M and the principal normal of the curve at the point corresponding to the parameter t . which is orthogonal to the zaxis. = f k ( t ) sin€J(t). and let c ( t ) = (2cost. Is this curve a geodesic on M ? Explain without long computations. z ) : (y 2)’ z’ = 1) around the zaxis. Hence (z(s) . z(s) lies on a sphere centered a t m. where k ( t ) is the curvature of the curve. namely. 2 sint.
in the direction of the circle of latitude. Namely. (a) 1.. f(zp)). Hint.O. Local coordinate computations are not necessary. f'(zp)= 0. (b) The vanishing curvature at point SO means ? ( S O ) = 0. that is. Since every circle of latitude on M is a line of curvature. From il the Frenet formula we know that the binormal vector field of y. f ( z ) ) : z E B}of f in the z z plane about the z axis.e. .161 3109 Let y : [0. (b) What happens if the curvature is allowed to vanish a t a point? (Indiana) Solution. = 1 implies that the curve is parameterized by arclength s. At P = ( z p 0. that is. the corresponding normal curvature where O(zp)is the angle between the normal of M and the principal normal of the circle of latitude at P. f ( z ) ) . 3110 Let f : lR + lR be positive and smooth. 0. (a) Show that y lies in a plane. is a principal curvature. Let M be the surface in lR3 obtained by rotating the graph {(z. then the corresponding normal curvature Ic. Thus. " . then it is an inflection point of the curve 7. y lies in a plane. $(y(s)B) = 0. (Indiana) Solution. Assume the torsion 7 = 0. which means that y ( s ) B = constant. Characterize in t e r m o f f the set of z such that && is a principal curvature of M a t (z. i. is constant. = implies that cos8(zP) = f l . the curve {(z. SO is a stationary point of y's tangent vector field. Thus.1] + lR3 be a C curve with 1il = 1 and nonvanishing curvature.f(z)) : z E IR} has a tangent parallel t o the z axis at P . If the point is the strict extreme value point of the tangent vector field. denoted by B . Ic. *$J .
) is always a linear combination of the binormal and normal vectors. then ( T ( s ) . Therefore. Suppose the curve a(.) c R3 be a smooth curve parameterized by arclength. we conclude that the set of L such that f of M a t (z. 0. we ) will get a ( s ) 5 0. the meridian passing P is also a line of curvature. show that a ( s ) is a closed curve.) is always a linear combination of the binormal and normal vector B ( s ) .) E M 2 is parameterized by arclength. the trivial case.162 Besides. T i) If k ( s ) > 0 and T ( S ) are curvature and torsion of a in E3. f(z)) is & is a principal curvature 3111 Let a(. Assume that the position vector a(. show that a ( s ) is a closed curve (kg is the geodesic curvature in M). a ( s ) = const.)) 0. (Here k . the second possible case is Therefore. Thus. If the position vector a(.) ii) If Icg(s) 1. Integrating the obtained equality. (Indiana) + Solution. if a ( s ) passes through the origin. (Indiana) Solution. i) The hypothesis of torsion ~ ( s ) 0 implies that the curve a is a plane curve.O. whereas the hypothesis of curvature k ( s ) > 0 implies that the plane R . a(. 3112 Let M 2 c IR3 be the cylinder z2 y2 = 1. N ( s ) of a ( s ) . Show that a ( s ) does not pass through 0 E lR3. Its curvature Ic a t P is just the other principal curvature of M . we have ( a ( s ) .
ii) If one develops M into a plane x. z ( s ) ) where z ( s ) is unknown. Therefore. ii) Given a C curve a ( s ) E B2 = ((z.e. d m  1 which means that it is a circle with radius 3113 Let T be a two dimensional distribution in B3 defined by i) Show that T is not involutive. i) That z& = and z&.y E IR} C IR3 and ” a ( s o ) . the curve a c M n x must be closed. iii) Show that if a ( s ) is a simple closed curve of length L bounding the region R in B2 then @(so + L ) . a must be closed. . ii) Let a ( s ) = (z(s).show that there exists a unique C curve p(s) E B3 such that . P ( S O ) = a ( s 0 ) and x(P(. By x ( p ( s ) ) = a ( s ) . where x(2. i. has the same geodesic curvature kg(s) 3 1.y(s).y. y(s). So.O).Y ) . then the corresponding plane curve of a . ds ’ ds ’ ds Hence Thus the problem of finding problem p(s) reduces to solving the following initial value .f3’(s) E ” q 3 ( $ ) . we may assume that p(s) = (z(s). Then .. denoted by &. are linearly independent shows that T is not involutive.O) : z.163 where the curve a lies does not parallel the generating line of the cylinder M . &A) where A = area of R.)) = a ( s > .O. Y.p ( s 0 ) = (O.B’(s) E Tp(. (Indiana) Solution.z(s)(:9 d2 dX [& & + I & { &. = (z. Thus the curvature of & is = 1 in x.) implies that d p ’ ( s ) = u(s)b(s) . 0). & + &} + + ”> for suitable functions a ( s ) and b(s).
. and i i= hence. Thus we have ) I” r ( s) ds .e.O. i) The hypothesis that l is always tangent to y means that (v. T N  = . 3114 Call a normal vector field Y along a space curve y parallel if l is always i tangent t o y. = (sine.P(s0) = (z(so + L ) .TB)+ s i n e . . Here s and L denote arclength and the length of y. B ( s ) .Y(SO>.y(so f L ) . without loss of generality. (Indiana) Solution.l ) 0. f ZdY) = (O.e. and v ( s ) = cos O(s) . i) Show that the angle through which a parallel normal vector field Y turns relative to the principal normal N along y is given by the total torsion of y. . N + case.the unknown function z ( s ) can be and so is the space curve P ( s ) . we may assume that (vI G 1. + i. N ( s ) sinB(s) .. Therefore. which measures the angle between ~ ( s and N ( s ) . iii) If P(s0 Q(S) is simple and closed. respectively.)sine. where the sign is determined by the orientation of the curve a. i. (ICT. ii) Show that the total torsion of any closed curve y which lies on a sphere in lR3 must vanish. B ) B +case. the normal vector field v has constant norm.z(so).fA). i.where e(s) is a smooth function globally defined on y.164 When the given plane curve uniquely determined by Q(S) is C”. then we easily have + L ) . N + (4  T )c o s ~B.cose k~ ( B ..
= ra T(S)dS +qo). fact that y is The closed means that f +)dS = e ( q . Furthermore. Also. where the integer n is just the degree of the map f. B ( L ). If k 5 0. fi = the constant map into PO. (Indiana) Solution. n = deg f. then C is naturally a geodesic on M . for every ytrwe have the corresponding map ft such that fo = f . finally we have I” T(S)~S = 27r .L]can be regarded as a lift of a certain differentiable map f : [O.hence the segment is also a geodesic one. Therefore. Let po E y and y contract to po smoothly on S 2 . l).. i. the smooth function e ( s ) . ii) For any closed curve y which lies on a sphere in El3. L] S1into R1. 3115 Let M be a surface in El3 and let P be a plane. Show that the intersection curve (parameterized by arclength) is a geodesic on M . along the segment where k ( s ) # 0.165 e(s) Noting that i/ is parallel to T .e(o) = 2n7F. t E [0.deg f1 = 0. we have the relation qS) J. On the other hand. we may assume that for every t E [0. Because the degree of a map is homotopically invariant.1].q o ) = J 0 L r(s)ds. the unit normal vector field Y of the sphere along y satisfies the above mentioned hypothesis. Suppose M and P intersect orthogonally. yt overlaps with y = yo about po = yl. Thus we get a family of curves { y t } . s E [O.e. Otherwise.we see that the above equality implies that = ~ ( s )and hence . . the normal of M is parallel to the principal normal of C. Let k ( s ) be the curvature of the intersection curve C = M n P .
If f ( s ) # 0.. then a”(s) 0.e. T(S) = 0. k’(s) = 0..) be a C2 curve in lR3 parameterized by arclength. Hence k ( s ) = const. a ( s ) is a straight line. i.r ( t ):= cy(t/c). a”(. and a ( s ) is a circle or part of it with radius 1 . then ( ( ~ ( s ) )= const. a ( s ) is a spherical curve. What can you deduce about f(s). for each t E lR. . (Indiana) Solution. Suppose that for some function f(s). If f ( s ) 0. differentiating with respect to s.e. From 1 f ( s ) a ( s ) a ’ ( s ) d(a’(s))2 = 0 = 2 we have f(s)d(a(s))2 = 0.) = f ( s ) a ( s ) . where c E lR is an arbitrary fixed constant. Derive the curvature function for this new curve. we have which implies that f(s)= k2(s).166 3116 Let a(. 0 3117 Suppose y(t) parameterizes a space curve with curvature function K ( t ) . i. Define a new curve 7 by setting. Further~ more. (Y(S)? (Indiana) Solution.
that is. then.0. the torsion is not defined at c. swith t ( s 1 ) = a . A counterexample is as follows. Hence. according to the above discussion. the plane ?r : ( p .b) + lR3 be a smooth curve with nonvanishing curvature. the connectedness of a and the smoothness of the Frenet frame field imply that T is unique. b ) .0) where the function f is defined by . . B Therefore.t E ( c . The condition k # 0 means that we can ~) 0 define the F’renet frame field { T . . but is assumed to be zero a t all other points. and a ( t ) . a ( t )E ?r for all t E ( a .167 Therefore. b ) must be on another plane a2. from which follows ( a ( t ( s ) )B ) = const = (a(t(so)). Then the hypothesis T implies that B is a constant vector field. maybe. i) Show that if the torsion of a vanishes identically then there is a plane ?r c lR3 containing a .t E ( a . $ ( a l l ? )= 0. Of course. from y ( t )= y ’ ( t / c ) and y”(t) = y ” ( t / c ) / c we have 3118 Let a : ( a . B } along a.Is ?r unique? ii) Is the conclusion of i) still true if the curvature is allowed to vanish at a single point c E (a. ( t .] (Indiana) Solution. Obviously.f ( t ) . f ( t ) ) if t > 0.a ( t ( s g ) ) B ) = 0 contains the curve a. (O. namely.N . ii) If the curvature k is allowed to vanish a t a single point c E ( a . s E ( ~ 1 . Let a(t)= { if t < 0.b).b)? [Of course. c ) must be on a plane T I . 0) if t = 0. ?rl # x2. t ( s g ) = b. ( t . i) Let a be reparameterized by arclength s.O. a ( t ) . ) . suppose a = a ( t ( s ) ) .
Let a be parameterized by arclength s .f3’(t)) 0. and let s . Differentiate the equation with respect t o s. ..Tl(sI))= 0. Firstly.. u2).e. duj o = d= . and M be locally parameterized by X(ul.. Differentiate the obtained expression for . (Recall that a is a generalized helix if for some constant vector u # 0. Therefore. 3120 Suppose that the unit normal vector to a surface M c lR3 is constant along c M . s1 represent their corresponding points.W d X n ds iJ=l 2 ds aut (Z) = W(a’). Using the Frenet formulas and the hypothesis (T(s).f3 is called an involute of a if for all t .s)T(s).. by the Weingarten formula.= . be parameterized by their arclength s and B s1 respectively. The curve .c ? . For convenience. Deduce that in this case.. = Show that every involute of a generalized helix a is a plane curve.) (Indiana) Solution. p is a plane curve. we have. Then we may assume . that a plane contains it. (Indiana) a regular curve a Solution. P ( t ) lies on the line tangent to a a t a ( t )and (a’(t).B successively.B(sl) = a ( s ) X ( s ) T ( s ) .a’(s)) const. Noting that a being a generalized hrelix implies + + we obtain by straightforward calculation i. a is an asymptotic (Note: A curve in a surface is called asymptotic if its acceleration is everywhere tangent to that surface. and that the Gauss curvature of M vanishes a t each point of a. let a and . E where s is arclength for a. .168 3119 Let a and p be two regular curves in lR3.) curve. along a . (u. we can ascertain that P ( s 1 ) = a ( s ) (so . the torsion of p vanishes everywhere.
2HII K I = 0 it follows that along Q + K =. Hence a" is everywhere tangent to the surface. is geodesic curvature.a ( s g ) .T . ifp's curvature function k ( s ) ( s = arclength) is odd. namely.0).a ( s o ) . from I11 . ( ( ~ ( s n) = const = (a(so).4 . along a . we have d ( a ( s ) .a')n kgn x a'. = where k.a')= 0. which means that the curve a is contained by a plane ( p . and has the following graph.:2 ) ( + 2H(W(a').O) andP'(0) = (1. we immediately deduce that. noting that n is constant. by the Gauss formula. n )= 0.169 Further. Secondly.3. Thirdly.n) = 0. n). ). n )= 0. Therefore. (Include an explanation with your sketch. ( a ( s ).) (Indiana) Fig. T]+ R2having p(0) = (0. satisfies r" Jo L(s)ds m 3 ~ 1 2 . 3121 Sketch the closed regular plane curve p : [ . along a 2 2 duj duk = kgn x a' + (W(a').
This assertion follows from I” k(s)ds = 3a J. too. O ) .3.B’(s) and p’(0). Fig.5. And yaxis is almost its “tangent” line at ( O .O). as shown in Fig.170 Solution.3. which has zaxis as its tangent line a t (0.” g d s = 6 ( ~) 6(0) x T . where 6 ( s ) is the oriented angle formed by . The image of the plane curve p is a “figure eight”.5 .
v). using the pull back of ds2 by (x')*.Compute interior of Q J k.) be a simple closed geodesic as described above and K ( z ) be the Gaussian curvature of the above metric at 2 E B2. 2 3 ) E B3 : 2 : xi 22 = l}. Thus.its coordinates with respect to the z1 and 2 2 axes are denoted by (u. For every p E R 2 . 2 2 . (Indiana) Solution. and B2 be the 21 o 2 2 plane. then their images are simple closed geodesics on ( E 2(xl)*ds2). Suppose that x : S2\{N} + B2 is the stereographic projection from the north pole of S 2 into the plane lR2.171 SECTION 2 DIFFERENTIAL GEOMETRY OF SURFACES 3201 i) Show that there exists a metric on the plane so that some geodesics are simple closed curves. i) Let S2 = { ( 2 1 .1 u2+v2+1' 212 + Then the metric of S2\{N} can be expressed by ds2 = + dv2) (22 + 212 + 1 ) 2 ' 4(du2 Now. the map x : ( S 2 \ { N } . we obtain 3 ) + + 21 = u2 + v2 + 1' 2u 22 = u2 + v2 + 1' 2v 23 = v2 .By direct calculation. ~ 2 ~ 2 E S2\{N} to (u. . which maps (21. d s 2 ) + . ii) Let a(. Here the integral is with respect to the Riemannian volume induced by the metric. we can obtain a 2dimensional Riemannian manifold ( B 2(x')*ds2). Since all great circles which do not pass through the north pole are closed geodesics on S2\{N}. .(x')*ds2) is an isometry. (B2.v) E B2.
. You may assume that the normal to the surface at x = 0 is ( O . (Indiana) Solution. 1). i.(O.(O. then its curvature at z = 0 is L ( 0 ) = lk. O . using the famous GaussBonnet formula. i) What is the normal curvature of a a t z = O? ii) What is the geodesic curvature of a a t x = O? iii) What is the Gaussian curvature of M 2 at x = O? iv) Sketch the surface near x = 0. a is a normal section.e.. &)I = 4.(O. &)= 0. and that in that basis the Weingarten map ax2 are L = ( 4 3 4 ”) * Let Q be the curve (near z = 0) obtained by intersecting M with the z1z3 plane. i) By the Meusnier theorem. or more directly. we know that the Gaussian curvature of M 2 at x = 0 is K ( 0 ) = 25. By the relation among k ( O ) .172 ii) Denote by D the simply connected domain encircled by the simple closed geodesic a. &)and the geodesic curvature k. we immediately have K ( x ) = 2 T X ( D ) = 2T. the normal curvature of Q at x = 0 is ii) Since a is obtained by intersecting M with 21x3 plane. Assume that and ax 1 tangent to M a t z = 0. &) we immediately obtain Ic. kn(O. K ( 0 ) = det ( y4 ) = 25. Then. iii) Because the eigenvalues of the Weingarten map L are f 5 . interior of J ( I 3202 Let M 2 c lR3 be a surface containing x = 0.
B . z = 2) and take (0. z ) E M : z > 21. we sketch the surface near x = 0 as follows. &)= 4 > 0. Then.(O. 1) and k. (Indiana) Solution. Choose the unit outward pointing normal as the orientation of M . z ) E E3 : x2 y2 _< 4 . 1) as its normal vector. Let P = {(z. by the Stokes’ formula.(zz y2)} (a paraboloid of revolution). we have that + l U p d w = 0.O. Noting that the normal to the surface at z = 0 is (O.r2)rdr = 16n.173 iv) From iii). y.0. Firstly.3. y. we know that x = 0 is a hyperbolic point of M 2 . Therefore. and w = yz2dx zzdy x2y2dz. s / Fig. Hence. z ) E E3 : z = 6 . D = {(x. Orient M and evaluate + + + LzdxAdy+dw. y.(z2 y2)]dx A dy = + 12r d 1 2 ( 6 . we have k [ 6 .6 3203 Let M = {(x.
3204 Let S be a surface diffeomorphic to the ordinary 2sphere. we have K d c = 27rx(D) = 0. x 2 )= ( r . J.y ) E lR2 : x2 y2 < and let 0 = (0. If y1 and yz do not intersect.0) = r2[1 .2r2 + r4 sin201'.z A d x + z d x A d y +zdz A dy + 2xy2dx A d z 2x2ydy A d z . Show that y1 and yz must intersect.O) E D. Here we assume the Jordan curve theorem.174 Secondly. . (Indiana) Solution. d y A d x = .& . Hence. Let ( z 1 . Suppose there is a C" metric of positive curvature on S for which there exist two simple closed geodesics y1 and 7 2 . 3205 Let D be the disk { ( x . r zdz A dy + dw = 8T. noticing that dw = z 2 d y A d x 2yzd. Applying the global GaussBonnet formula to the domain D c S and noting that y1 and y2 are geodesics of S. + + we have ~dw=~4dyAdx+2dxAdy=2 J. which contradicts the assumption of positive curvature. then y1 and 7 encircle a domain D such 2 that the Euler characteristic x ( D ) = 0 and d D = 71 uyz. 0 ) be the usual polar coordinates and define a metric on D\{O} by + i} where h(r.
= giidx2 When r 0. Use a smooth coordinate system. Hint. + Hence the metric of D\{O} may be rewritten as ds2 = dr2 + h(r.T sin adz).175 (a) Prove that this metric extends to a smooth metric on D. this metric can extend to a smooth metric az = % i +2 cos 8 sin 8[l . (b) Show that line segments in D which pass through the origin (suitably parameterized) are geodesics. g12 = o(T).+ 2lii2dxdy + li22.8 = O 0 } and l2 = { ( r .8)de2 + 4r4 sin2 8)2]dxdy +[sin2 8 + cos2 8( 1 . of I must be a continuous function of the coordinates. as a whole line segment. { Then = r. .(1 .d~~. we know that both l1 and 12 are geodesics in D\{O}.8 = B0 + T } . we can choose (x. y = r sin 8. and %o( l). 8 ) : 0 < T < :. the geodesic curvature k .2r2 + 4r4 sin2 8)2]dy2 . t 00.2r2 + = [cos28 + (1 . on D\{O}. we know that kgIo = 0. If we use the extended metric. Z22 = 1 + o ( r ) .2r2 + 4r4 sin219)~ 8)dz2 sin2 on D . Therefore. From { follows dx = cos Bdr .r sin 8d8 dy = sin 8dr T cos 8d8 + { dr = cos dx sin 8dy d8 = (cos 8dy . Because 11 and 12 are geodesics by either metric. y) as the coordinates. we see that &I = 1 + o ( r ) . (b) For any line segment I in D . (Indiana) Solution. 1 is a geodesic. we can express 2\{O} as the union of Il = { ( r . So. (a) Set x = rCOS8. $& = o( 1). Hence. Using ds2 = dr2 + h(r.8)d82. e ) : 0 < r < :.
e. Suppose M is inside a ball of radius R and suppose a point p E M is on the boundary sphere. Then there is a family of spheres that contain the surface M inside and have a fixed center. (Indiana) Solution. 3207 Let H 2 = {(z. Hence the Gauss curvature and of M at p is greater than or equal to ii) Suppose that M is a closed minimal surface in lR3. y) (a) Compute the components of the LeviCivita connection (i. there must be at least one common point of M and the sphere with radius R and centered a t the fixed point.y) E lR2.. K ( p ) 2 &. is the usual inner product on B2. l ) E H 2 . the boundary of the ball. Show that the Gauss curvature of M at p is 2 ii) Show that there is no closed minimal surface in lR3. &. have the same tangent plane a t p . But it contradicts the fact that . l ) be a tangent vector a t the point ( 0 . (b) Let V(0) = ( 0 . using the local canonical forms of the normal section curves M n 7~ and S 2 n7~ a t p . . Using the result of i). a t p . 6. Then. the Christoffel symbols). Ler R be the infimum of their radii. Let V(t) = (a(t). Then.176 3206 i) Let M be a surface (without boundary) in lR3. the curvature of M n T is not less than that of S2na = S1. i) The surface M and the sphere S 2 .y > 0) be the upper halfplane in lR2 and let H 2 have the Riemannian metric g such that where (2. one concludes that the Gauss curvature of M a t p . p is an elliptic point of M . for minimal surfaces. Let a be a normal plane of M and S 2 at p . one can easily show that.b( t) )be the parallel transport of V(0) along the curve z = t .
3208 Consider the hyperboloid S of one sheet x 2 + y2 .0). in the direction of the circle of latitude. (b) At P = (0. Therefore.vyO)) = ( 0 . the corresponding principal curvature Icl = f. 0 ) .1) = ( d ( t ) .z 2 = 4. (a) The Gauss curvature K of S at P is defined by K = k l k 2 . k2 are the two principal curvatures of S a t P. . Noticing the above expression of is equivalent to rik. Then b V ( t )must be the solution to the following initial value problem (vl(O). Hint. writing a ( t ) = cos@(t). Thus.along the curve. by taking the outer unit normal vector field as the orientation of S.2.u 2 ( t ) ) y) and the parallel transport vector field V ( t ) = ( a ( t ) . (a) From the hypothesis we have E = G = I F = 0 and hence ?I2 ' (b) For convenience. ( t ) ) = (wl(t). (Indiana) Solution. (a) Define the Gauss curvature K of S. denote the curve ( 2 .177 y = 1. Show that V ( t )makes an angle t with the direction of the yaxis in the clockwise direction. l ) . b ( t ) = sinO(t) where O ( t ) is the angle V ( t ) makes with the zaxis. v 2 ( t ) ) . where k l .b ( t ) = sin@(t). = ( t . we see that the problem = a(t) 9= b(t). the normal section is a circle x 2 + y2 = 4.we have immediately 0 = t. (b) Use the definition from part (a) to compute K at ( 0 . Write a ( t ) = cosO(t). (Indiana) Solution. 2 .
Thus. By the GaussBonnet formula.178 whereas in the direction of the meridian.& k i ) . locally. and s1 is the spherical cap encircled by C. 1 i..z 2 = 4. the normal section is a component of the hyperbola y2 .2. we take the lines of curvature as the parametric lines of M . ( M ) at P. := P . then by the Rodriques equation we have on P .(z) = x E Y ( X ) will immerse M provided I E ~ is the reciprocal of neither principal curvature k l or k2 at x E M .& k 1 ) ( 1  Ek2) ax dul x # 0 au2 ax . If. the Gauss curvature K = k l k 2 = a. with smooth unit normal vector field v. (Indiana) Solution. x Pc(z) P. In this case. thus the corresponding principal curvature k2 = Therefore. ( M ) Pc(z) dU' a = (1 .(z) dU1 au2 a a = (1 . in f + (Indiana) Solution.(x) terms of k l and k 2 . 3209 Calculate the total geodesic curvature of a circle of radius r on a sphere of radius R > r. : M R3defined by P. the total geodesic curvature is given by where C is the given circle. ax aui i = 1. express the principal curvatures of M . 3210 Let M be a compact surface embedded in R3. Show that the mapping P.
w ) for v . Assuming the Gauss curvature K of this metric is everywhere nonnegative. Suppose that.a ) . . the corresponding part of 2) is and the corresponding points of A .a .y) in lR2.. whose vertices are A ( . B ( a . Furthermore. . w E T(.1 and respectively. C ( a .sin 0 . respectively. use the GaussBonnet Theorem to deduce that in fact. D ( . B . ( M ) . C . we can express the metric of the Riemannian surface k2 ds2 = ex(y)(dx2 dy2). which shows that the mapping P. a ) respectively. w ) := ex(Y)(v. D .a ) . again by the Rodriques formula. which is obviously by conformal to that of the Euclidean plane B 2 ..lco = (do ds and . we can see that the geodesic curvatures of the segments of coordinate curves i g and EE are +  5.6 2 d z ax O=O. C . Using the Descartes coordinates (x.y= a k. (Indiana) Solution. Then. immerses M into R3.1. B .)lR2. in P . Take a positive number a and consider a square domain D in IR2.179 provided that I E ~ is the reciprocal of neither k1 or k2 at x E M .case + 2 O=n. D are A . and the corresponding principal curvatures are ~ k . we see that the u1 and u2 parametric lines are the lines of curvature l . by the Liouville formula. the surface is flat.a).a . 3211 Let X : lR + l be a smooth function with compact support. Applying the GaussBonnet . in lR2. and consider R the Riemannian surface obtained by equipping lR2 with a metric g of the form g(v. do 2dE 1 alnE aY coso+ 1 dlnG sin ' .y=a 1 dlnE 2~ aY 1 dlnG a ax whereas both 52are geodesics of k2.
dy2  d2X namely ’ dY is a decreasing function. we have < 0 . it leads to Besides. Thus. y=a y=a which combining the above equality shows that % = 0 in B1. z ) : z 2 0).Therefore. 2 3212 Let M 2 c R3be a smooth compact surface such that M 2 c ((2.) is an asymptotic curve on M 2 .180 formula to 5 c IR2. from the above integral equality we can obtain Because the number a is arbitrarily chosen. . z ) : z = 0) is a smooth curve a ( s ) . ii) Show that a’(s) is always a principal direction. letting a + 0. i. K 0.e. we have where the Gauss curvature Noting the area element d a = e’dxdy and the line element ds2 = ex(y)dz2 along i g and E E .. Assume that M 2 n ((2. R is flat.y.y .parameterized by arclength. by the hypothesis K 2 0. i) Show that a(.
i. we see that along a ( s ) . the plane {(z.illustrated in Fig. R3 with a (Indiana) Solution.3. Thus.z) : z = 0) consists of elliptic or parabolic points of M 2 . Hence. which means that the tangent vector of a ( s ) for every s is an asymptotic direction. i) The hypotheses imply that along the curve a ( s ) . (a) Consider the following surface S of revolution generated by a C vi" brating curve C.y.} such that length (ri)+ 0. and length (ci) 0. = 0. a ( s ) is an asymptotic curve on M 2 . along a ( s ) . + X Fig. a t which the plane {(z. If there exist closed geodesics c. then the set M 2 n {(z.i such . What kind of a set could M 2 n {(z. Then.. z ) : t = 0) is a curve. Let Pi denote the points where C has horizontal tangents. from which the Rodriques equation says that a ' ( s ) is always a principal direction with the principal curvature 0. rotating around the zaxis which is the asymptote of the curve. 3. Pi draw closed geodesics c.181 iii) Let's now drop the assumption that M 2 n ((2.y. y. z ) : z = 0) is tangent to M 2 .a ' ( s ) . = .7 (b) Suppose S is a compact surface. Therefore. 3213 (a) Construct an example of a noncompact C surface in " sequence of closed geodesics {c.O . t) : z = 0) be? (Indiana) Solution. y. the unit normal vector field n ( s ) of M 2 is constant along a ( s ) .z) : z = 0) is a tangent plane of the surface M 2 .7. on S. iii) If we drop the assumption that M 2 n { ( z ly l z ) : t = 0) is a curve. ii) Noticing the above fact. (b) Show that this is not possible if the surface is compact.y.
. T ( S ) sin 8. Thus.. z ( s ) ) a unit speed curve in B3 with T ( S ) > 0. we come t o a contradictory result 0 = 2 ~ . Consider be the surface of revolution (s.182 that length ( u i ) + 0.___ a G~ 8 rgl+ 8s 21ae 2~ as 2 ~ a . 3214 Let ( r ( s ) . On this surface ) compute the covariant derivatives V+& and V A & in terms of and 86 (Indiana) Solution.E~ a +. Because S is compact. setting i + 00 in the above equality. denotes the Christoffel symbol ( i . we may suppose R.a . 0 ) + ( ~ ( s COSO. z ( s ) ) .denotes the domain encircled by u.8~ 2 ~ a = .%+r2=+.r ( s ) ae’ e r.we have  a vs. The direct computation gives & 6.e) = (ul.a where as = l a a GI G2 r2. By the GaussBonnet formula. O . Ic = 1. K d a = 2 r x ( M Z ) = 4 ~1( 9 ) _< 0. (Indiana) Solution. u 2 ) .. is simply connected. aas 22ae 2 a ~ e a y 2 . 3215 Let M 2 c R3 be an embedded compact surface of genus 2 1. we have J I. then by the GaussBonnet formula we have where a.  . when i is sufficiently large. Show that the Gauss curvature of M must vanish somewhere on M ..ae 89 v8.j . = Denoting (s.2).T ( S ) T ’ ( S ) . I = d s 2 + r2(s)d02 Eds2 + Gd02.
0 < u. F = 0. (Indiana) Solution. there exists a domain U of P such that KIu > 0. y. 3210 Consider the torusofrevolution T obtained by rotating the circle ( z . Z) : (z2 y2 + + z 2 + u2  T ~ . we show that there exists another point Q with K(Q) < 0. L = T. by the connectedness. ( a + r c o s u ) s i n w . we would have a contradiction. ) ~ Parameterize this torus. A straightforward computation gives the coefficients of its first and second fundamental forms E = r 2 . N = COSU. G = (u + TCOSU)~. Finally. we claim that because of the compactness of M 2 . Therefore.w < 27r. Thus obtained torusofrevolution T can be parameterized by sT X(u. the continuous function K must vanish somewhere on M 2 . by the continuity.F 2 d u d v we have immediately KdA = = T(U + TCOS u)dudv.F 2 cos u T(U+ TCOSU)' Noting that the area element on T is dA = J E G . Otherwise. hence. .u ) ~ + z 2 = r 2 around the zaxis: T = {(z. compute its Gauss curvature function K. there must be a point P with K ( P ) > 0. ( u + TCOSU). and verify that KdA = 0 by explicit calculation. its Gauss curvature function K = L N . M = 0. r s i n u ) .w) = ( ( u + r c o s u ) c o s w . 12T 12' dv cos udu = 0. Secondly.183 Firstly.4u2(z2+ y2) = 0).M~ EG .
Thus. 2 2 E be a curve on S 2 c Ill3. the area of S 2 is not less than 47r. for any Riemannian metric on S 2 . 2 3219 Define “geodesics”.. if IKI Kdv = 2 x x ( S 2 ) = 47r. and characterize (with proof): i) All geodesics on the unit sphere S 2 c IR3.8 = 27r  (i. By the GaussBonnet formula. (corresponding to the central angle 8 = $t in the domain D). 1 cos(t). we obtain the result that. sin(t).184 3217 Let z(t)= (. Compute the parallel 21 a (Indiana) Solution. Consider the cone that is tangent t o the unit sphere S 2 along the curve z. for a displacement t of a moving point p along z starting from 0. q) 3218 Show that for any Riemannian metric on S2 with II K Gauss curvature. then 47r 5 Is IKldv 5 Area ( S 2 ) . 5 1. 5 1. “1 & . Because the parallel translation in the plane coincides with the normal Euclidean one. we have J. This cone minus one generator is isometric to an open set D c lR2 given in polar coordinates by 0 < p < +m. . 0 < 8 < AT. 7 0 5 t < 27r T(L fils2. the oriented angle formed by the tangent vector with the parallel translation vector X ( t ) of XO is given by 27r . Let X o = translation of Xo along z ( t ) . where K is the (Indiana) Solution.
185 ii) All geodesics on the surface ((2. y. What is the usual topological hypothesis that prevents this problem? (Indiana) .j ddsk = o ds j. every geodesic on the cylinder must be a helix. du u +Cr*. we have d ( S ) = Ic(s)N(s) where u1. E 0 or  d2ui ds2 . the principal normal vector N is parallel to n. Let a ( s ) be a curve on a surface M parameterized by arclength. draw a picture) to show that a connected surface can have two points which are not jointed by any geodesic.k i = i.u2 are local coordinates on M . k. every geodesic must become a straight line. (Indiana) Solution. Along a . or a straight generating line. and kg is the geodesic curvature of a. On the contrary. that starts from a given point and is tangent t o a given direction. z ) E R3 : z2 + y2 E 1).2. then if we develop the cylinder to a plane.. 3220 Give an example (e. owing t o the uniqueness of the initial value problem every geodesic on M . because along a . n is the unit normal vector field of M along a . i) On the unit sphere. every great circle is a geodesic. Therefore. a is a geodesic of M if and only if along a . must be a great circle. ii) Since geodesics are intrinsic objects. Then.g. or a circle of latitude.
cosu.sin w). Calculation is not necessary here. (Indiana) Solution.is minimal. Hint. w. i) From x “(sinu. Let K be a plane and p .8) = (z.O. 3221 Define “minimal surface in R3”. cosh(t) cos(8). y two points in K. w) = (cos Jz u . Let T be an interior point of the line segment p y .sinv.O).0. ii) Figure out the Clifford torus’ Gauss curvature function. and prove that the catenoid. A 1 X “a . “Completeness” is the usual topological hypothesis that prevents this problem. sin u. 3222 The “Clifford” torus in S3 can be parameterized using charts of the form 1 X ( u . Then the surface K\{T} is a case in point.(0.cosv) 1 . (Indiana) Solution. i) Compute the metric [gjj] on the Clifford torus for a coordinate chart of the indicated type. obtained by rotating the graph of y = cosh(z) around the zaxis in R3. A minimal surface is a surface with mean curvature H= 1E N . cosh(t) sin(8)) is a minimal surface in R3.2 F M + G L = 0. iii) Deduce from the result of ii) that the Euler characteristic of a torus is zero.186 Solution. cos where u and v are constrained to lie within intervals length < 2 ~ . 2 EGF2 The straightforward calculation shows that the catenoid X ( Z .
by straightforward calculation. ii) The Gauss equation implies that the Clifford torus’ Gauss curvature function K 0. a 2 jat ( O . Hint.187 we easily have 911 = 922 = . the principal curvatures of the graph z = f(Z1. O . iii) Using the GaussBonnet formula. O ) ) E B3 are the roots of the following equation . Then.g. we have Therefore. O ) ) E R3 are the same as the eigenvalues of the Hessian matrix [ a 2 f / a z . ] ii) Show that lR3 contains no compact embedded surface with strictly negative Gauss curvature at all its points. O . f ( 0 .. i) Show that the principal curvatures of the graph z = f(Z1. i) The hypothesis means that $ and vanish at (0. 2 1 g12 = 921 = 0.2) at ( O . . we see that the Euler characteristic of a torus T 3223 Let f : IR2 + l be a smooth function with a critical point (e. f ( O . Look a t the “lowest” point on the surface.22) at ( O .O). O ) . a minimum R or maximum) at the origin 2 1 = 2 2 = 0. (Indiana) Solution.
yo. where f(z. y ( u . )  .~ ( o = upn M . z ) for U .~ ( 2 . Since the inclusion map M + lR3 is an embedding. (Indiana) Solution. we can express this embedding as ( u . Y.188 ii) Let p ( z 0 . Therefore. the embedding can be expressed by ( 2 . there exists a neighborhood U of p in R3. ~ + ( z ( u . for each coordinate neighborhood V c M of p E M . such that V = U M . y ) .Y) (z. v ) . and in terms of the local coordinates (z.y). c V of p .f ( z . v ) ) .Y>. y. then T is the tangent plane of the surface a t p . { Y= 2 = 2(u.) Since the surface is above the plane T : z = zo and p is the common point of R and the surface. Y) Y)). z ) = z . (The existence of such a point follows from the compactness of the surface.v) Y(U. Thus the Gauss curvature of the surface a t p is not less than zero. if we take (O. z ( ~ .y). By observing the normal section at p . we can find a neighborhood Up c U of p in lR3. 1) as the unit normal of the surface at p. = Up n M . such that on V. Using the local coordinates ( u . Then.v ) for V and (z. 3224 Let M be a 2dimensional manifold smoothly embedded in lR3 with unit normal n. we have F . has smooth inverse { 2) =4ZlY) = v(2. such that V. it is easy to conclude that any normal curvature of the surface at p is greater than or equal t o zero. a) Prove that for each p E M there exist an open neighborhood U p of p in lR3 and a smooth function F : Up + B2 such that Fl(O) = Up n M . y . b) Find F if M is the graph of a smooth function f : lR2 + lR. ) Noticing that we know from the implicit function theorem that there exists a neighborhood V. f ( 2 . setting F : Up t R3 by F ( z . ~ ) .. zo) be the “lowest” point on the surface.O. = z ( u ( 2 . Y)).
For each p E M . y ) .. Xu x Xu Jacobi matrix the rank of the ( ffu ffv 3 /: ) = 2. z ) = z . [Recall that S C R3 is defined to be convex if for each two points of S the line segment joining these points lies in S.. 3225 Let M be a 2dimensional manifold smoothly embedded in R3 with unit normal n. Since nu x n. d) Show that J. prove that is a diffeomorphism onto S2. 211.) (Indiana) Solution.+ S2 is a local diffeomorphism. Assume that M is the boundary of a bounded convex open set. a) For each p E M .f ( z . . . we can take Up = R3. (You may assume b) and c) if you wish. You may use without proof the fact that M lies on one side of each of its tangent planes.v). define q ( p ) as the end point of the unit exterior normal n(p) after parallel translating it to the origin of E3.189 b) If M is the graph of a smooth function f : R2 + lR. b) and c).> rank = (4% . such that on U the Gauss map has the following expression q 0 h'(u.>> # 0.then for each p E M .P(. z ) : x 2 + y2 + z 2 = 1).and F : R3+ R defined by F ( z . = K X . 7) b) Prove that 7) is onetoone.] a) Define the Gauss (or sphere) map 7) :M $ s2= {(z. r(. We first prove that q : M . x Xu and K > 0. K ( P ) d P = 4?T.v) is the unit normal at X ( u . v) E C C M . y . there exist a coordinate neighborhood C c M of p and a coordinate map h : C + U c B 2 .).which is a diffeomorphism. c) Assuming that 71 is onetoone. = n ( u . where n(u. .. y. Assume that n is the exterior normal and that the Gauss curvature K of M is everywhere positive.
q l c .190 Thus. Thus q ( M ) is an open. q is an open map. Namely. : CQ + ~ ( C Qare diffeomorphisrns.Q respectively. we arrive at a contradiction. Now we prove that q : M t S2 is globally onetoone by leading to a contradiction. which implies q(M\U) = S2. On the other hand. q = (7o h') o h is a diffeomorphism. From the above argument we know that there are neighborhoods C p . = (vIc*)'(7(cP) v " rl(CQ)) Thus. connected and oriented. 5. : Cp 7 ( C p ) . . where da. Besides. set . 7 is also a closed map. such that 7Ic. Since 7 : M + S 2 is a local diffeomorphism. Therefore. closed and nonempty set of S 2 . noticing that differentiability is a local property. CQ of P. in the neighborhood of h ( p ) .hence 7 o h . 7 is a local diffeomorphism.respectively. The connectedness of S 2 implies that q ( M ) = S2. Q E M such that q ( P ) = q ( Q ) . In the end. C Q ) so small that C p n CQ = 0.1 ( 7 ( C P )n 7(CQ)). Then the GaussBonnet formula gives 1. I . the implicit function theorem says that 7 o hl has a smooth inverse on a neighborhood of h ( p ) (which may be smaller than U ) .l is a local diffeomorphism. Now take the inverse images of ~ ( C P ) ( C Q ) under nq q l ~ and q l cQ . Next. in M we can choose C ~ . because M is compact and S2 is a Hausdorff space. Suppose that there are two distinct points P. (by noting K > 0). we see that the globally onetoone. da = Hence Kda = 4n. it is easy to show that M is compact. u n v = 0 and q ( U ) = q(V). fi have local expressions Since Kda > 0. Because P # Q. we show that q : M S 2 is surjective. +  u = ( v I c p ) . Thus. on M . surjective local diffeomorphism is naturally a global diffeomorphism.
191 Remark. (Indiana) Solution. Since M is compact and S 2 is connected. defined using the covariant derivative on R3. they are called “extrinsic quantities”. a) Explain what is meant by intrinsic and extrinsic quantities on M . but extrinsic ones are not. as the above. and M c R3 is connected and hence path connected. consider a plane and a cylinder. Let Aw denote the angle variance caused by a tangent vector after parallel translating it around C once. b) The principal curvatures axe not intrinsic. Using the theory of covering map. Therefore. In fact. Aw J do so . Using the Hopf’s rotation index theorem and the GaussBonnet formula. we can show that the Gauss curvature of M a t p can be expressed by K ( p ) = lim D+P . Further. show that q : M t S2 is a local diffeomorphism. Firstly. then the local diffeomorphism q : M + S2 is also a covering map. 3220 Let M be a 2dimensional manifold smoothly embedded in R3with unit normal n.is intrinsic. b) Are the principal curvatures intrinsic? c) Discuss why the covariant derivative on M . we can simplify its proof as follows. a) The terminology “intrinsic quantities” means those geometrical quantities that are definable only by the first fundamental form of M and its derivatives. For example. c) Although the covariant derivative on M is defined by using the covariant derivative on R3. intrinsic quantities are those that are invariant under isometric correspondence. discuss why the Gauss curvature of M is intrinsic. last local expression of the covariant derivative of M the involves the tangent vector field and the Christoffel symbols of M . d) For each p E M . they are extrinsic. we know that the covering map q must be a homeomorphism and hence a global diffeomorphism. Otherwise. it is intrinsic. because S 2 is simply connected. In other words. d) Assuming c). this problem is the famous Hadamard theorem. let C be a simple closed curve encircling a simply connected domain D where p lies.
Thus the Euler characteristic number x ( M 2 )= 0. the uparameter curve X ( u . l).3. (Indiana) . Besides. too. 1)2+ N such that for each fixed wo E (0. b) Show it for arbitrary n. then the Gauss curvature is intrinsic.8 Solution. the boundary of M 2 consists of two circles which are just geodesics of M 2 . where K is the Gauss curvature and dA the area form. 3228 A surface C2 immersed in a Riemannian manifold N is said to be ruled if it can be parameterized near any point by a mapping X : (0. 3227 By revolving the curve y sketched below around the zaxis. we have immediately K d A = 0. a) When n = 3. because along these circles the normal vector of M 2 is parallel to the principal normal vector of the two circles respectively.) (Indiana) z I I Fig. by the famous GaussBonnet formula. show that the Gauss curvature of a ruled surface in Rfl is nowhere positive. on M . (Make sure to justify your answer.192 Since parallel translation is intrinsic. wo) is a geodesic in N. Compute JM K d A . The surface M 2 of revolution generated by the curve y is homeomorphic t o a section of a cylindrical surface. we get a surface of revolution M 2 c R3. Therefore.
l ) . we see F 0. ) dU2 f0. we can easily deduce that the first coefficient of the second fundamental form of i)l(w). for convenience.w) = a(w)+(uwhere Z(v) is the direction of the geodesic corresponding to w E ( 0 . we obtain the desired result. then C2 can be characterized locally by X ( u . . However.Z’(v)) + K = U2II’(V)12)dV2.iw2 <o. we may assume that v E ( 0 . suppose C2 can be characterized locally by where.F2 EG. l ) is the arclength of the curve a ( v ) . 1Z(v)1 SE 1.Z’(w))2 = la’(w)x I’(w)I2 2 0. through computation. Since the Gauss equation says * dz’ it suffices to show that 2 0. which shows that the Gauss curvature of the surface K= LN . we have Noting that 1l’(v)12 . Thus.(a’(v).F2  b) Similar to the above. that the first fundamental form of C2 is ds2 = d U 2 + (1+ 2u(a’(w). by a routine work.193 Solution. and (a’(u). a) Since every geodesic in N = R3 is a straight line. C2 d 2X L = (n.l(w)) Then. EG.
then 2 = aX by. Explain why it is a tensor.W ) . where R( ) is the Riemannian curvature tensor field of M " .V[X.respectively. Thus. X .X .( L X . Z . W. M. i) The sectional curvature is defined by K ( R ) = . Y .V y V x Z . Y ) = R ( X . where L is the Weingarten map.Y .R ( X . Y .Y)Z = V x V y Z .194 SECTION 3 DIFFERENTIAL GEOMETRY OF MANIFOLD 3301 Let M" be a Riemannian manifold and R a 2plane in T. we can conclude from straighforward computation that ( 1) + + is an orthogonal matrix. Use this basis to define the sectional curvature K ( R ) and show that it depends only on R and not the particular basis chosen. Y ) is another orthonormal basis of R . y. Y . E T'(M"). fW) = f R ( X . If ( 2 .Y]Z and the Riemannian curvature tensor field by R ( X . define the Riemannian curvature operator by R ( X . then R ( X . W)I.. Z. ? = cX d Y .Y ) Z .2.2.Y ) Z = ( L Y .Z ) L X . Y . we can easily have R ( X .W on M " . Y ) . Y } be an orthonormal basis of R . using the properties of covariant differentiation " and of the inner product. Y . fY.W ) = ( R ( X . W ) = R ( X . f Z . in terms of local coordinate frame field.Z ) L Y .X .y. 2. Y .Y. Noticing that field. (Indiana) Solution. one can show that V p E M " . Then for any C function f on M " . Z . Y ) . and R( ) is a 4th order covariant tensor  z. compute all sectional curvatures of the sphere {z E B4: 121 = 3). ii) For any vector fields X . is dependent only on X. R(fX. i) Let { X . which proves that K ( R )depends only on R .W ) . ) = R ( X . z.. R ( X . .. Using this or any other valid method. iii) Recall that if M" c En+' is a hypersurface.Y. W ) W = R ( X . Therefore. ii) Define the Riemann curvature tensor in terms of covariant differentiation.
X / 3 . Y } is any orthonormal basis of an arbitrary 2plane T in T p ( M n ) . where V denotes the covariant differentiation in lR4. " iii) For the sphere M n = { z E lR4. X ) ( L Y .Then by the original definition of : covariant differentiation. Y )= 1/9.1zl= 3 } .195 R( ) I p is a welldefined 4th order multilinear function.( R ( X . Suppose F : lR3 + lR4 is a 11 C immersion in some neighborhood of C. The image " F(C) is then a compact Riemannian submanifold of lR4 with boundary and therefore has a volume. X ) ( L X . if { X .V x g = . we can regard the position vector field 2 as the normal vector field of M n . the inner product and covariant differentiation are all C". { e i } be a local orthonormal Let frame field about 3 (as a point) of M n . 3302 Let C = {z E lR3 : 0 5 3ci 5 1) be the unit cube in B3.( L Y . Y ) = ( L X . one can easily show that V e i z= e i . Hence. the Weingarten map is L : X F+ L ( X ) = . Y ) X .    K(7r) = . Hence. R( ) thus defined is a C tensor field. Y ). Justify the following formula: . Besides. Therefore.
196
Let Au be the volume of the parallelopiped spanned by the vectors PzP;, P3Pl and in lR4. Then
0, Ax3 0. We take the principal is an infinitesimal when Ax1 + 0, Ax2 part of the infinitesimal Au as the volume element, namely, the volume element of F ( C ) is
+ f
Hence the desired formula follows immediately. Furthermore, if F ( x ) = ((1 XI)^, (1 ~ 2 )(1 , ~ 3 )(2 , ~ 3 ) ~then, by ~ ~ ), , denoting F(x) = F(z1,22,23) = (yl, y2, y3, ~ 4 ) we have that
+
+
+
F,
(&) =
2 ( 1 + ~ 1 ) a , aYl
= 2(1+x2),
a
aY2
= 2 ( 1 + ~ 3 ) a+ 3 ( 2 + ~ 3 ) dY3
a a Since   and ayi %a
2
.8
dY4
form an orthonormal frame, we obtain that
3303
There is no submersion from S3 into lR2.
(Indiana)
197
Solution. ( + ~ + Let S3 c E4be defined by { ( x ’ , x 2 , x 3 , x4) E E4; x ’ ) ~ ( x ~+) ( x 3 ) ) ” (2‘))” = 1). Observe the following three vector fields
where (xl, 2 , z 3 ,z‘) E S 3 . It is easy to see that they are nowherevanishing x tangent vector fields on S3. Since ( x ’ ) ~ (z2)>” ( 2 ’ ) ) ” ( x ~ = ~ then, ) 1, without loss of generality, we may assume x4 # 0. Hence, from the fact that
+
+
+
det
(
z2
x4
23
21
24
x3
24
x2
21
)
= ( x ‘ ) ~ x4(x3)’ + ~
+
~ ( 2 ’ ~) ~ ( 2 ’ x4 ~ ~ )
+
=
# 0,
we know that X I ,X2, X3 are three linearly independent vector fields. Furthermore, by direct calculation, we obain
Now we suppose that there is such a submersion a : S3 + El2. Then, a : T p ( S 3 + T T ~ p ~ (is a surjection for every point p E S 3 . Thus, we may , ) E2) assume that, for example, a*Xg,7r*X3 are linearly independent, and 7rIT,X1 = aa,Xz b7r,X3 a t p . Then, on the one hand,
+
on the other hand,
Therefore,
a,X3 = aba,Xz  b2a*X3
from which we get ab = 0 and b2 = 1. Similarly, from [ X s , X 1 ] = 2x2 we can deduce that a2 = 1 and ab = 0. They are all contradictory. So, there is no submersion from S3 into E z .
198
3304
Let M n and N k be Riemannian manifolds. Then M x N is naturally a Riemannian manifold with the product metric. If XI, . . .,x, are local coordinates on M and y1, * * * ,yk are local coordinates on N , the product metric in local coordinates ( X I, , x, , y1, . ,y k ) looks like

i) Let X be a vector field on M x N “along” M (i.e., in local coordinates no are present) and Y be a vector field along N . Show that DxY 0. ii) Show that a t z E M x N , some sectional curvatures always vanish. (e.g., product manifolds in the product metric never have strictly positive curvature.) (Indiana) Solution. i) Using the properties of the Riemannian connection on M x N , we only need to verify D e = 0. In fact, observing that the inverse matrix of g is of the form
yp and
&
&
and the first class of Christoffel symbols of M x N satisfy
we easily conclude that
Obviously, we also have D B y s azi = 0. a p
ii) By the definition of the curvature operator, using the result of i), we
= 0, Hence, for arbitrary X along M and Y This means that R along N , using the Cm(M x N)linearity of the curvature operator, we have
(&,&)
R ( X ,Y, Y ) = 0. X,
Therefore, the sectional curvature determined by X and Y always vanishes. Obviously, here X may involve yp, and Y may involve 2,.
3305
Let F : M t N be smooth, X and Y be smooth vector fields on M and N , respectively, and assume F,X = Y ; that is, that
a) Let w be a smooth 1form on N . Define the Lie derivative Lyw of w with respect to Y . (If you use local coordinates, you must verify independence of choice of coordinates.) b) Prove that F*(Lyw)= Lx(F*w).
c) Let Z be a smooth vector field on M such that F,Z = W , where W is a smooth vector field on N . Show that L y W = F , ( L x Z ) . (Indiana) Solution. a) L y w is defined by Lyw = d ( Y [ w ) + Y Ldw, where the symbol ''Y denotes the interior product of a vector field with a form, for example, if R is a pform, then Y [R is a ( p  1)form defined by
Thus, Lyw is a welldefined 1form on N . b) Let Z be a smooth vector field on it4 such that F,Z = W , where W is a smooth vector field on N . Then we have
200
c) Noticing that L y W = [Y,W]and [F,X,F,Z] = F , [ X , Z ] , we immediately obtain L y W = F , ( L x Z ) .
3306
Let
w =[
T,
8, z be the usual cylindrical coordinates in R3. Let
cos 8+5r2 sin28]d8+[2r cos 8+6zr sin 8]dz+[4zr2sin 8]dr.
~ T sin 8+3z2r Z
Let the curve y be given in rectangular coordinates by
y ( t ) = ( z ( t ) , y ( t ) , z ( t )= (cost,sint,4sin5t+sin2tcosst), )
o 5 t 5 27F.
(Indiana)
Evaluate
s,, w .
f be defined by
f(r,8,Z ) = 2rz cOs8
Solution. Let a function
+ 3z2rsin 8,
(T,
8,
.>
EB ~ .
Then
f is a C” function in lR3 and we have
df = [ ~ T s i n 8 + 3 z 2 r ~ ~ ~ 8 ] d 8 + [  2 r ~ ~ ~ 8 + 6 ~ ~ s i n 8 cos8+3z2 z Z ] d z + [  2 sin81dr.
It is easy to see that f I, is also a C” function, and by the invariance of the form of first order differentiation, the above expression of df is an exact lform on the closed curve y. Noticing that y is a closed curve, we have, by using the Stokes’ theorem,
w
=
I
(w  d f ) =
I
5r2sin2 8d8 + (4zr2sin 8 + 22 cos 8  3z2 sin 8)dr
Without loss of generality, we may assume 0
i w = lr5sin28d8
5 8 < 27r and T = 1. Then
d8 = 57~.
=5
201
3307
Let M be a C" Riemannian manifold. Assume the theorem that there is a unique C" mapping V : X ( M ) x X ( M ) + X ( M ) denoted by V : ( X , Y )+ VXY which has the following linearity properties: For all f , g E Cm(M)and X , X ' , Y, Y' E X ( M ) , we have
= f(vxy)+S(Vx,Y>, V x ( f Y gY') = f V x Y g v x y ' ( X f ) Y (Xg)Y', [ X ,Y ] = VXY  V Y X ,
VfX+gX, Y
+
+
+
+
X ( Y ,Y ' ) = (VXY,Y ' )
+ (Y,VXY').
a) Suppose a : [ q b ] + M is a smooth curve. Define what it means for a vector field Y on a to be parallel along a . Derive the differential equations that must be satisfied if Y is parallel along a. b) Let X , Y E X ( M ) . Let p E M and let a : [0, b] t M be an integral curve of X such that a(0) = p , da/dt = X ( a ( t ) ) . Show that
where P,,o,t : T,(o)M
+ T,(t)M
is the parallel transport along a. (Indiana)
Solution. a) As known, thus defined mapping V is the Riemannian connection. Using the properties of V , we can prove that V % Y is completely determined by a ( t ) and Y ( a ( t ) ) So V % Y is welldefined for every vector field along a. . Now we give a definition that a vector field Y on a is parallel along a , if and only if V % Y = 0, Vt E [a,b]. In order to derive the differential equations, we choose a coordinate neighborhood with local coordinates {xa}. Then
Denoting
202
we have, by the properties of V ,
Therefore, the desired differential equations are
b) Choose a basis { e l ,    , e , } of T,(M), where n = d i m M . Let
Since P,,o,t is a parallel isomorphism for every t , then { e l ( t ) , . . , e,(t)} is a basis of T a ( t ) ( M ) .Hence we can denote
Then, by the properties of the connection V and by the fact that e i ( t ) is parallel along a ( t ) ,i = 1,.. + n, we immediately have ,
On the other hand, because
is equivalent to
ei = p ~ h , t ( e i ( t ) ) ,
we can write
203
3308
Let M points p , q dependent identically be a Riemannian manifold with the property that given any two E M the parallel transport of a vector from T p M t o T,M is inof the curve joining p and q. Prove that the curvature of M is zero, i.e., R ( X ,Y ) Z = 0 for all X , Y ,Z E X ( M ) . (Indiana)
Solution. For a n arbitrary point p E M , take a coordinate neighborhood D of p with local coordinates zl, . ,zn . Let

be n linearly independent vectors in T p M . Using the hypothesis that the parallel transport of a vector from T,M to T,M is independent of the curve joining p and q , then for q E D, we can transport every Vpi from T p M to T,M. Thus, we can define n linearly independent vector fields V1, . . . ,V , in D. Obviously, all V;’s are welldefined, and if we transport along special coordinate curves, then
namely
k wll,i(z) = 0 i, j , k = 1, * * * ,n.
Now, by the Ricci identity, we have
k k vil,jrn(zC) v i l , m j ( z ) =

C v11(z>Rfjrn(z)= 0.
I
Noting the linear independence of
K’S, we immediately
have, in D,
Rljm= 0 k , l , j , m = l ,   . , n , k
i.e., the curvature of M is identically zero.
204
3309 Let M be an ndimensional Riemannian manifold. Suppose there are n orthonormal vector fields XI,. . * , X, that commute with each other (i.e., [Xi, Xj] = 0; i,j = 1, .,n), show that the sectional curvature of M is identically zero. (Indiana)
Solution.
Set
k
Then, from (Xi,Xj) = 6;j it follows that
(vxkxiIxj)
+ (xi> 'Xkx'j)
= O!
i.e., T i i + rtj = 0; i,j,k = 1,. . ,n. On the other hand, from [ X i ,Xj] = 0 it follows that vx;xj  VX,Xi = [Xa,Xj] = 0, i.e., rk.  r!. = 0; i,j,k = 1,.. ,n. Namely, the Christoffel symbols '3 3% are antisymmetric with respect t o k,j, and symmetric with respect to i , j . Therefore,
rk. = rJ
'3
ak
= rii = rfj= rfk= rk. = r.. k
3'
a3
= 0; i, j , k = 1, . . , n. Hence the sectional curvature of M is which means identically zero.
3310
Let X be a smooth vector field on a Riemannian manifold M . The divergence of X, denoted by div(X), is defined by the function trace ( V X ) . i) If M is closed (i.e., compact without boundary), show that
/M
.
div(X)dv = 0.
ii) If M is compact with boundary dM, show that div(X)dw =
JM
1,
(X, N ) d s ,
205
where N is the outer normal vector field of a M . Hint. Consider w E h l ( M ) defined by w(Y) = ( X , Y ) , and try to use the Stokes' Theorem. (Indiana) S o ht ion. In fact, this problem is the famous Green theorem. The outline of the proof is as follows. Firstly, one can show that, for example, by means of the normal coordinate system about p E M , thus defined div(X) satisfies div(X)QM = d(i(X)Qm), where Q M is the volume form of M and i( ) is the interior product operator. Next, for p E d M , choose an oriented orthonomal frame field about p {el,...,e,} such that, at p , el = N p . Let { w l ,   . , w n } be the dual frame field of {el,. . , en}. Then the volume elements of M and d M are respectively
dv = n ~ ( p=) w
Observing that, a t p ,
~ A * * * A w ~ s,= R a M ( p ) d
= W 2
A*..Awn.
i(X)n,
= i(X)(w1 A . . . A w n ) = w1(x)w2 A . . A w n + (terms involve w ' )
=
(x,N ) Q ~ M (terms involve wl), +
d(i(X)QM) = J,,(X,
N)%M
and along d M w1 s 0, one can obtain, by Stokes' theorem,
/M
div(X)dv =
= J,,IX,
Wds.
If M is compact without boundary, the right hand side of the above formula vanishes naturally.
3311
Let w l , .. * , wk be oneforms. Show that {wi}fZl are linearly independent if and only if w1A w2 A . A wk # 0. (Indiana) Solution. Let wl, ., wk be defined on an ndimensional manifold with n 2 Ic. Suppose w1 A . . . A w k # 0. If wl, ,wk are linearly dependent, then without loss of generality, we may suppose that
1 9 . .
W k =Ulwl$"'+Uklw
k1
206
with suitable functions a l ,
W1A
* * *
,a k  1 .
* * *
Thus we have
(UlW'
A
WWk
W1A
A Wk' A
+ . . . + ak1Wkl)
T
0,
which contradicts the above hypothesis. . On the contrary, if wl,.. , w k are linearly independent, then we can extend Thus them to a basis {W~,...,W~,W~'~,...,W~}.
w1A
implies that
... A w k A wk+l A ... A w k # o
3312
Let M be a Riemannian manifold. Let p E M (a) Show that there exists S > 0 such that exp, : Bb(0) c T p M + M is a diffeomorphism onto its image. (b) Show that there exists E > 0 such that exp,(B,(O)) is a convex set. Hint. let d ( z ) = distance from z to p . Show that d2 is convex in a neighborhood of p . (Indiana) Solution. (a) This is just the existence of the normal neighborhood of p . Use the fact that d exp, is nonsingular at p, and then the implicit function theorem. (b) The existence of convex neighborhoods is a classic result due to J. H. C. Whitehead. Refer to every standard textbook on differential geometry.
3313
Show that (i) A is a differentiable manifold.
207
(ii) A is a Lie group with the standard matrix multiplication as a product. (Indiana) Solution. Define a map F : Gl(2,LR) + Gl(1,LR) by F ( X ) = d e t X . Then F is a smooth homomorphism between Lie groups, and the rank of F is constant. Therefore, the kernel of F k e r F = F'(l) = A is a closed regular submanifold of GZ(2, El) and thus a Lie group.
3314
(a) Let f be a smooth function on a Riemannian manifold M. Let gradf be the vector field defined by the equation
(gradf, w ) = dpf v , ~
w E TpM.
Let (z',. . . , 2") be local coordinates around p . Find the expression for gradf in terms of z', ,2". (b) For a vector field X define the divergence of X, div(X) as the trace of the operator Y + DyX where D is the LeviCivita connection. Find the expression for the divergence of X in a local coordinate system (z', . . . , P). (c) Use (a) and (b) to find the expression for the Laplacian A in local coordinates, where A acting on a smooth function f is defined by Af = div(Vf). (Indiana) Solution. For convenience, we omit the suffix p . (a) Let

grad f = Then, from
c .
1
a1 8 ax1
it follows that
k
208
Thus,
(b) Denote
x = c x i  a: . ax
i
Then, by the definition of divergence, we have
Af
= div(gradf) =
c
i
ik
af
3315 Let M be a compact connected Riemannian manifold without boundary. Let f be a smooth function satisfying Af = 0. Show that f = const. Hint. Use the definitions in Problem to show that div( f V f) = IV f 12+ f A f . (Indiana) Solution. For any function f , by straight calculation, we have div(fVf) = lVfI2
+ fAf.
Now, noticing the hypothesis of this problem, by Green’s theorem we have
J, lVf l2dW = 0,
which implies df = 0 everywhere, that is, f = const.
209
Suppose F : M N is a smooth map between differentiable manifolds, and is homotopically trivial. Show that in this case, F*w will be exact whenever w is a closed 1form on N . (Note: F is homotopically trivial if it extends to a smooth mapping : M x [0,1] N such that B(z, = F ( z ) , for all z E M , 0) while F(z, 1) q E N ( q constant) for all z.) (Indiana) Solution. Consider the map G : M + { q } and the inclusion i : { q ) N . Then the map F : M + N is homotopic to the composition ioG. Furthermore, we know that F* and (i o G)* = G*o i* induce the same homomorphism
+
f

3316
F** = ( i o G ) * * H 1 ( N , d ) . *

H 1 ( M , d).
Since i * ( Z 1 ( N d)) c Z1({q),d ) = 0, the induced homomorphism F** = (i o , G)** is a zero homomorphism, i.e., F * ( H 1 ( N , d ) )= 0. In other words, for every w E Z 1 ( N , d ) ,F*w E B 1 ( M , d ) ,namely, F*w is exact.
3317
Regard lR9 as the space of all 3 x 3 matrices with real entries. Does the subset C = {A E Rg: det(A) = 0) form a smooth submanifold of R9?
(Indiana) Solution. Observe that {A E lRg : rankA 5 l), the union of all axis in Rg, is closed. Then M := R9\{A E R9 : rankA 5 1) is an open submanifold of Rg. Define a map F : M + B1by F ( A ) = det(A), A E M . Noting that dF = ( A l l , A12, A13, A21,A22, A23, A31, A32, A33) where Aij is the algebraic complement of the corresponding entry aij of A, we see that rankdF = 1 on M . Therefore, Fl(O) = {A E IR9 : det(A) = 0) is a closed regular submanifold of M . Hence, it is also a submanifold of lR9.
210
3318
Let
Compute w , where S3 = { x E IR4 : 1x1 = l), oriented as the boundary of the unit ball (assume standard orientation on B'). (Indiana) Solution. Denote D4 = { x E IR4 : 1x1 5 1). Then, by the Stokes theorem we have
ss3
= J,,4w=J,4dw =
L
(dxl A d22 A d23 A d24
A dxl A d22 A d24)
+ d22 A dxl A d23 A d24
. 2
x2
+dx3
=
J,. dxl
A d22 A d23 A d24 = v01(D4) =
3319
Prove that
is a threedimensional submanifold of IR4.
(Indiana)
Solution. Define a map F from E4\{(0,, O , 0)) into B1by O
" Then F is a C map with ( 2 4  2 3  2 2 x l ) as its Jacobi matrix which has constant rank 1 on IR4\{(0, 0 , 0 , 0)). Thus
(21, 22,23,24)
: rank
( X . a axl a ax4 i) Is there a 2dimensional submanifold M 2 of lR4 such that for each p E M2.(23 2 4 ) . Y ) = 01. Noting that E4\{(0. 0)) with dimension 3. is orientable. Y) = ( f l . a ax2 ax3 x2=+22. is a smooth compact embedded submanifold of IR6 and be identified with the unit tangent bundle of 5 ' .Y E Solution.X2] = . O . (Indiana) :X.y E lR3} and define a map F : R6+ IR3 by F ( x .asJ does not satisfy the Frobenius condition. Xl(P). a i) No.M2? ii) Is there a nonconstant function f in the neighborhood of 0 E lR4 such that X1f E 0 and X2 f O? (Indiana) Solution. The reason is that the bracket [Xl. llYll = 1. i) Identify lR6 with {(x. 3320 Let vector fields XI.O.IIXII = 1.we see that is also a threedimensional submanifold of lR4.X2(P) E T. & + 3321 Let M = { ( X . f3) = (11x112. we can set f = 21x2 .f 2 . there is not. for example. Y ) i) Show that M explain how M can ii) Show that M B3.O . ii) Yes. (x. X2 on lR4 be defined by x1=  a + 21.21 1 is a regular submanifold of R4\{(0. 0)) is an open submanifold of B 4 . It is easy to verify that the Jacobi matrix of the C map F " .y) : x.Y)). O.llY1I2.
y) = = 1 0.E . so is N and F is a covering map. ( S 2 ) . i.. x I. and the image of its end point under F must be y.. If (U. ( U .y)II = (llzllz Ilyl12)i = fi which means M is bounded in IR6. If y is a limit point of F ( M ) in N .& x $ is . Because F is a local isometry.(S2) with E 1)yI)= 1. y) E M . F ( M ) is also closed in N . since II(z. the Hopffinow theorem means that N is complete. #Jp x $ p ) # 0. E M . then we can identify M with the unit tangent bundle of S 2 . Thus. if we regard z E S2 and y E T. = (0. Besides. Thus. z has the local coordinates (ui. x I.O) = M is a closed embedded submanifold of B6. has the following J acobian + + which means that { ( V a x I. a coordinate neighborhood of (2. we show that F is a covering map. f2 = ( ( ~ ( =~ 1 and f3 = (z.ui) U. 1. F’(l. 0. # J ~ ) }of coherently oriented coordinate neighborhoods. for every (z. take S > 0 so small that exp.e... where B’(6) = {v E . By using identification of M with the unit tangent bundle of S 2 . S2 has a covering { ( U .) neighborhoods. Besides. 4 x 4.212 has constant rank three when fl = 11z1(2 1.. : B’(S) + Bi is a diffeomorphism. F ( M ) is open in N . M is orientable. F is surjective. y) to 0. . . ) au. M must be compact. then the transition function .. Naturally. Next. (Indiana) Solution. y) E ) with E being a suitable positive real number and 4 is the map from (2. and y is uniquely determined by the oriented angle 0.y) M . since F maps every geodesic of M into a geodesic of N .dol x 4) form a covering of coherently oriented . too. Therefore. 3322 Let F : M 4 N be a local isometry between connected Riemannian manifolds M and N . then there is a point z E M such that there exists a geodesic in N connecting F ( z ) and y. a t z E from 4 to the unit tangent vector y E T . ii) Because S 2 is orientable. where I. Hence. . The local isometry of F and the completeness of M imply that the above geodesic can be uniquely lifted to a geodesic in M starting from z. Therefore. the connectedness of N implies that F ( M ) = N .. For every x’ E N . Show that if M is complete. for every (z.) n (Up x I p .
ya. z = y a ( l ) = yp(1) = zcp. Thirdly. The uniqueness implies ya = yp. i. and F : B p 6 B. f ( N ) : IvI < S} and Bi = {y’ E N : d ( y ’ . In fact. Fl(z’) is discrete. Then. expXm also a diffeomorphism.) = y = F ( y p ) . which . if z E F’(B. In particular. we have F(y. Since F is a local isometry. we have the following commutative diagram and expZe is surjective. M is complete. are isometric. then 0 there is a unique geodesic y : [0. In other words. . So. uBr c Fl(B. contradicts a # p. z a ) < a}. Since F is a local isometry. Secondly.213 T . if there is z E B. are diffeomorphisms. 2’.. i) Show that each component of X = {z E M : F ( z ) = z } is an embedded totally geodesic submanifold of M . we claim : that Bi is an admissible neighborhood of z’ and F is a covering map. we claim that F . F o expZm = expzf odF is a diffeomorphism and hence exp.“ + B. Fl(B. L ( y ) = L(y) < S means that z = T ( 0 ) E B r .. Hence F = exp. Let y be the unique geodesic in Bi connecting F ( z ) and .. F : BO + Bi is a diffeomorphism.) 0 a is obvious.8. z ’ ) < 5).. o d F o (exp.1] B.1] + M such that F ( T ( t ) ) = y ( t ) . exp. Firstly. Besides. Since . Because both F : B. there exists a geodesic y : [0.e. V t .yp are the lifts of y through z. is is an immersion. then BF n Bf = 8. Thus the claim is proved. Denote Fl(z’) = {z.) c U B r . such that y(0) = F ( z ) and y(1) = 2’.l ( B i ) = U B ? .} c M and set Ba(S) = {v E T. Therefore. Otherwise. Hence F(y(1)) = z’ and y(1) = z for some a .)’ is a diffeomorphism. we know that d F . we say that for any a . yp connecting z with .” n B f . + Let F : M + M be an isometry of a Riemannian manifold M . .). Besides. we claim that if Q # . z zp respectively. then there exist unique geodesics y. On the + other hand.((M) lvl < 6) and Br = {y E M : d ( y ..
z = 0) u {(z.(V n B(6)).e. in a neighborhood of S O . d F ( j ( 0 ) ) = j(O).(VnB(G)) is obviously a submanifold of M . Thus the uniqueness implies F ( y ) = y.v = y. = 01 u {(z. V is a subspace of T. where a 6 is so small that exp. Hence. F is an isometry. d F ( v ) = v. we show that every geodesic y : ( a . .(M). In order to prove the claim. Thus. let <(s) be a geodesic of M such that <(so) = SO). F ( z . z ) = (z. Let y : [0. = 0). b ) .1] 4 M be the unique shortest geodesic y(t) = exp. <~ o . From d F ( v ) = v follows d F ( T ( 0 ) )= j ( 0 ) .(V n B(6)). Since F ( < ( s o ) ) = (‘(SO). y ) < 6}. i) First. ~ . v E V which means that y E exp. which means that y E X n Bg. Use exponential coordinates.z ) . Then M is a 2dimensional manifold. and F is an isometry. : B(6) + B is a diffeomorphism.. X is totally geodesic. Therefore.On the other hand. = y.. i.. In particular. i(so) = ? ( S O ) . In particular.(VnB(6)). For any SO E ( a . namely.y < 0}. Because so is arbitrarily chosen.. y. ( M ) : d F ( v ) = v}.i. suppose that v E V n B ( 6 ) and g y = exp.y E X . we can assert that X has submanifold structures. Therefore. y is a geodesic of M .y. is naturally a geodesic of X.i. then F ( < ) and are two geodesics of M which satisfy the same initial conditions. Thus. we show that X has submanifold structures. Next. i. Then. Besides. Let the geodesic y : [0. Then we claim that XnBa = exp.z) E ~3 : > o . For z E X c M . Since z.214 Hint. (Indiana) Solution. F ( y ) = F(y(1)) = y(1) = y. because exp. ( M ) : IvI < 6) and B = {y E M : d ( z .y.z) ~ E 1123 : =o . and x = {(2. X n B c exp. v.e..e. If this is proved. b ) + X parameterized by arclength is also a geodesic of M . lies in X . y . set B(6) = {v E T .(V n B ( 6 ) )c X n Ba.(tv) connecting z and y.(tv). exp. we first assume that y = X n & and v E B(6) such that exp.1] 4 M be defined by y ( t ) = exp. ii) Give an example in which the components of X have different dimensions.e. ii) Let < < < < M = {(qy. F ( < ) = <. and F be a reflection with respect to the plane z = 0.z> E IR3 : 2 = 0.y. F ( y ) is also a shortest geodesic connecting F ( z ) = z and F ( y ) = y. Define V = {v E g T .) E IR3 : y > 0 . that F is an isometry implies F(y) = y. d F ( ( ( s 0 ) )= ((so) and F is an isometry.
Let d6 be its dual nonvanishing 1form on S'(Caution: Here d6 is only a formal symbol. n is globally welldefined on S1.d) = {f E Coo(S1.d) = Cm(S1.A'(S1)). without citing the de Rham Theorem. d ) . A'(S')).CdB is exact. := {z I/ u. (Indiana) Solution. observe that Z'(S'. d ) = Zo(S'.z1 < E for some z E X } .d)= 0 and S' is connected. define a function & Because R(0) = R(27r). Do so directly. then H * ( S ' .d) = Z'(S'. since there are no nonvanishing kforms on S'.v E N .Cd6 = do.d) = {df I f E Coo(S1. H1(S1.x + :2 . B1(S1. i. E ) := {y E En: Iy .. I ' For H k ( S 1 . denoting we see that w .d) = (Cd6 I C E Rl}Z R1. we d) have Zk(S1.d)/B'(S'. Besides. Let 6 be the polar coordinate characterizing S1. = 0 and hence H k ( S 1 .215 3324 Compute the de Rham cohomology groups of the circle S'. Hence. Le.R1)df = 0) E B . B ( X . For every w = g(O)d6 E C"(S'. k> 1.R1)}. Therefore.. and it is not exact. w . since Bo(S1. 3325 Let X denote a submanifold of Euclidean space E n . because 6 in usual sense is not a globally welldefined function on S'). d )= 0.e. X . set and E x. IvI < E } . Then is a nonvanishing vector field on S1.
either compact or complete. E )for all E . . t ) satisfies . j ( z ) = g (&.216 Show that U. i. and 2 = 2 ( z .E l . z m .g.X = + B ( X .y > > S(JhX. t ) such that m i. In local coordinates (zl. ) Can you give conditions which imply equality? (Indiana) Solution. Call a vector field 2 on M a killing vector field if 2 generates a 1parameter group of isometries of M . ii) Show that the expression (**) above is equivalent to where V denotes the LeviCivita connection for ( M . the end points. i) Show. E ) . Here L z denotes the Lie derivative along z.e. &).. E ) .. Z ( g ( X . (Consider examples of 1dimensional submanifold in E 2 . Y ) g ( X . let the 1parameter group of isornetries generated by a vector field 2 = C zi& i=l m be expressed by 2 = 2(z1. t ):= 2 ( z .e.g ) . 3326 Consider a Riemannian manifold ( M .  a . Setting y = z v immediately implies that V E Xc B ( X . Let X be an open line segment in E 2 . If X has no boundary. g ) . L Z Y ) = + (**) for all vector fields X and Y on M . Then considering the boundary of X .j=l m k. Show that the two are not generally equal.9 ) . z m ) of ( M .that when 2 is Killing. i. we have Lzg = 0 . can show that V E X is a proper subset of B ( X .l=l where g . then U. e..X c B ( X . (Indiana) Solution.
or equivalently g (V a Z .i + S i k zkj $9 ( l . 2 . = Y.i = 0. + 3327 Let M 2 be a connected Riemannian manifold and X .i V 0. l) j e er3 Z ) =o. i) Show that the integral curves of X are curves of constant geodesic curvature. (Indiana) Solution. Conclude that ( X . or equivalently. we easily obtain (w). X ) . Y complete vector fields on M . From this follows what we desire in ii). j = 1 . Y ) and ( Y . Because the flow X t of X for every t is an isometry. Y ) g ( X . ( X . + .o X t . &)= 0. Y )are constant on M . the vector field X should satisfy the Killing equation Xi. i) For every p E M . ii) Assume that X and Y are linearly independent at all 2 E M 2and that their flows commute X t o Y. take a local coordinate neighborhood about p such that the coordinate curves are the integral curves of X and their orthogonal trajectories. V Z Y ) = and the Lie derivative satisfies L z X = [Z. Namely. Assume that the flows Xtand Yt are isometries of M for all t. X I . i.j Xj.217 Differentiating the obtained equality with respect to t and then setting t = 0. we may assume that X = X I & and g 1 2 = (&. we obtain which can be written as k gkjz.Y)) g ( V z X . Noting that the LeviCivita connection V satisfies Z ( g ( X .
X = Y = G. j . respectively. we obtain a922 agll.8x1 a x 2 dx2 dXl = O . taking i = j = 1 and i = j = 2.e.0 . X1 = X1(xl). Ic = 1 . i. we obtain 911% following coordinate transformation = 0. if a adopting the original notations. we can obtain = 0. i. a ( X .g 2 2 ( ~ ~ ) ( d ~ ~ ) ~ . then the vector field X = from the corresponding killing equation. = const. along every integral curve of X .218 Taking i = 1. (Y. about p . Then. j = 2.. Because their flows are commutative X t o Y. Ic. Therefore. Now. all gij's are constants. make the Hence If we still adopt the original notations. ii) Analogously.. = Y. Then. using the Liouville formula. Again. we can make a suitable coordinate transformation and then. we can compute the geodesic curvature of the xlcoordinate curve as follows: where 0 = 0.Y ) = g12 = const. that is equivalent to lay2 d ax1 d [ X . we obtain d=Yol2. i. we have locally X = X I & . =o 8x1 dxi ..Y ) = 922 = const. ak = 0. ( X . Y = Y2&. 2 .e. we immediately have x ax2 Therefore. which means that the metric of M 2about p can be written as ds2 = g11(x2)(dx1)2 &.e.Y 2 . Y ]= . take the integral curves of X and Y as the x1 and x2 coordinate curves. by the corresponding Killing equations. o X t . . +.. ax = 0. i.X ) = gll = const. Noting the expressions of X and Y . dX1 &.
e.f A f. f ] .219 3328 Let M be a compact Riemannian manifold without boundary. e i ) f l = C [ e r ( e f f ) ( v e .n. j = 1. may suppose that e i = C a ! e r i = l . . For any f E Cm(M).C [ e i ( e i f ). define O f E X ( M ) and Af E C “ ( M ) as follows: At any p E M . they do not depend on the choice of orthonormal frame) and then show that (Indiana) Solution..( v e . we can easily obtain i i C [ e i ( e i f ). . e i } be another orthonormal frame field around p .. choose an orthonormal frame field { e l .. Then Green’s theorem implies . Noting (ei . i . e . . Using the definition of the Laplacian here. e : ) . i i Hence Of and Af are all well defined.e3) = S i j . we have div( f V f ) = llVf112 . . Then we Let { e . j for suitable functions a!. . .( v e i e i ) f I = i i Verify first that V f and A f are well defined (i. . } around p and then define (Vf)(p) = C ( e i f ) ( e i ) and (Af>(p> .. we have j Using these equalities and the properties of Riemannian connection. . .1n  e . through direct claculation.e j ) = (e: . .
If a = 0.4a 2 0 is the prerequisite condition. x : + 2 : = b or 2. 1. we can show that they are all 2dimensional submanifolds of R4.f3 are two roots of the equation X 2 . 3. we can easily show that &. (Indiana) Solution. = 2. Using the theorem of closed regular submanifolds proved by rank theorem.b is a 1dimensional submanifold of R 4 .4a > 0 and when b2 . Let X be a smooth vector field on M . Thus S0. x 3 = 2 4 = 0). (i) Show by an example that d F ( X ) may not be a vector field on N .b = ( 2 E E4I b + 2. b # 0.o is a 0dimensional manifold.f3 = x i + x z .4a = 0 Sa. + 2.2 : = b . Denote a = xf + x i and . Then C Y ~ a and CY + p = b = means that a and . Analogously. 3330 " " Let F : M + N be a C map between two C manifolds. b2 .220 3320 For what a . Assume that F is onto.bX + a = 0. is Sa. b # 0.b a manifold? Explain. then = { Z E B4 1 x1 = 2 2 = 0 . If a = b = 0 . then when b2 . . i . = 2}. 2. b 2 0. If a # 0. then 21 = 2 2 = 2 3 = 2 4 = 0. Therefore.
x2. Therefore.. (iii) Using local coordinates.YZ]. by direct computation. Show that d F ( [ X l .). d(F o a) (i) = ( d F o da) (g) = d F ( X ) = Y. y(t) has the following expression  21 1 t.x n ) be the local coordinates about p and set p : (xk. X Z l )= [Yl. 1 and X z . i. Hence. Show that F takes integral curves of X into integral curves of Y.e.y E IR}.) be an arbitrary integral curve of X . (Indiana) Solution.yz)). If y ( t ) is the ith coordinate curve that passes p . z (iii) Suppose XI. XJ 1 x 3 0 (. Y are related as X and Y in (ii) above. (ii) Let a(. Then..N = {x : x E E}. # then. as a vector field. (Indiana) Solution.. one will obtain the desired equality. we have d F ( X ( x . there is a unique vector field V f (called the gradient of f on M ) such that whenever y is a smooth curve in M with y(t) = p .y l ) ) # d F ( X ( z . which means that F takes integral curves of X into integral curves of Y . d F ( X ) is not well defined in every point of N .. = (xz + y 2 ) & . .y) = x.i 4. Show that whenever f : M + lR is a smooth function. Then.221 (ii) Suppose Y = d F ( X ) is a smooth vector field on N . . we have da ($) = X . if y1 # yz. . along a . (i) Let M = {(x. Let (xl. 3331 Let M" be a Riemannian manifold. by . Y.y) : z. x and F : M + N be defined by F(z.
3332 The space ELnxn of n x n real matrices forms an n2 dimensional Euclidean space.z.. Let Sn' be the unit sphere in IR". . and r : an RnX" be defined by (21.zn) E S"l. z i.) in S"l to the symmetric matrix ~ ( x= I [ z i z j ] . 5 i=1 1 " (u(x). ) Jz i) Show that u maps S"' into the sphere of radius centered a t the origin in R" n. n i) Let x = (zl.j b i j . which means that u(x)is on the sphere of radius 2 centered at the origin in 45 nnxn + ii) Let i : S"I + Rnbe the standard inclusion map.and define u : S n w 1 E n x n the map + as sending x = (21. .. . Hence ..e..B ) := i=l j=1 a. ii) Prove that cr is a local isometry (i.222 we have the expression where gij's are the components of the matrix [gijl' = [(&. the pullback via g of the dotproduct metric defined above on RnX" the standard one on S"'). a . It is easy to verify that the above expression of V f (p)is independent of the choice of local coordinates.. .j=1 c = 1 5.. is (Indiana) Solution. Then C z = 1. in which the dot product between A = [a.j]and B = [ b i j ] is given by n n ( A .u(x))= z.) + [yij] = &[zizj]. 2. Suppose that .
where f : IR" IR is a smooth function. k E { 1 . a metric of the form [sij] = e 2 f [ S i j ] . i) Show that for arbitrary indices i..j=1 (dXj + VjXi)(WiXj +WjXi) = 3333 c n i=l viwi = (v. i. .w). n} = & f ii) Show that when n = 2.223 belong to Tx(Sn'). the sectional curvature of M along an plane a t p is given by e1. Let ei = denote the standard coordinate basis vector fields.e2 (Indiana) . and S i j is the Kronecker delta.e.Then we have n n i=l i= 1 Therefore Hence we have (dn(v). 2 .dn(w))=  1 . Let M be the Riemannian manifold obtained by equipping B" with a metric conformal t o its usual one. j . c n i.
224 Solution. the complete vector field Y is Finvariant. F.Y]= 0.invariant. W E R. Y ]= L X Y = lim [Y . i) Suppose that X.Y ]= 0. is a diffeomorphism for each s.x. Observe that for a diffeomorphism F : M +M . o r. = yt o X.. Yt be the flows induced by them. Then . o yt = yt o X . Thus 1 [ X .. i) Show that X . i. t E R.t E IR implies [ X . . from the Gauss equation we obtain 3334 Let X . for all s. 3 4 0 s ii) Now suppose that [ X . for all s. Noting that E = G = e 2 f . Y be complete vector fields on a manifold M and let X . ii) Prove the converse to i). i) Set From gij = e 2 f & it follows that ii) The sectional curvature of M along an e l . e2 plane at p is just the Gauss curvature of M a t p .Y = Y . Y ]= 0. (Indiana) Solution. then Y is X.Since X .e. F = 0. if and only if F o yt = yt o F .
4(exp tv) = exp tw. namely. namely.z ) and (y'. ox Hence 3335 If 4 : G1 + G2 is a Lie group homomorhism. we obtain X .invariant. t E l is a 1parameter subgroup of G1 generated by R ' v E LG1. Assuming that wel " a =cvi. . we have (Indiana) Solution. Since 4 is a Lie group homomorphism.for all s . ( X s * Y ) p f ( X o * Y ) p f= Y p f . . 4(exp(u)) = exp(+*v). respectively. = yt O X . . we have for all s E R. Note that exptw. i=l axi we2 = a E w e aye ' n e=1 .225 Hence for each p E M and any f E C"(p). . X. Y = Y . Y is X. Let ( z l . 4 can be expressed by and exp tv and exp tw can be denoted by respectively. . then 4(exptv) is also a 1parameter subgroup of G2 generated by a suitable w E LG2. . yn) be local coordinates of G1 and G2 about " the identities el and e2. = Since f is arbitrary. Therefore. t E R. Locally. . show that for all w E LG1.
24)..21.226 we have t=O t=o Therefore. Taking t = 1 completes the proof. i.24y3 + 23Y4 = 0.: a) Is the rank2 distribution defined by these two vector fields in U completely integrable? b) Is the rank2 distribution orthogonal to these two vector fields completely integrable? (Indiana) Solution. by left translation. (22.w. d*w = w. Hence 4(exptw) = expt4. we have d $ ( v ) = w. y 3 and Y4 I Z l Y l + 22y2 22y1+ + 23Y3 + 24Y4 = 0. a) Direct calculation gives [r.23) v. 23. y2. 2 4 ) E B4 are given by rx : U := B4\0 = = (21. b) Construct the following linear algebraic equations about y 1 . Thus the Frobenius theorem guarantees that the rank2 distribution defined by r and v in U is completely integrable. 2 3 . a 1 t=O aY" Furthermore.. ZlYZ .v] = 0.22.e. 24. = 2 (Z) (= I =we. 2 2 . 3336 Consider the linearly independent vector fields r and v on whose values at x = ( 2 1 . which are equivalent to .
. 2224 p = . we obtain 2123 2223 +2224 . k 2.:). o).2223 Setting and respectively.2124 2124 2224 + 2123 . j = 1. ~ = 2(2: ] zi)v which does not belong to the distribution defined by CY and p. ).[ e i . we characterize the LeviCivita connection on (G.e. + (. by the vector fields Veiej. a) For arbitrary smooth vector fields X and Y on G. Thus the F’robenius theorem tells us that this distribution is not completely integrable. . (Indiana) Solution. e j ] 2 for all i. b) Show that when G is a Lie group with a biinvariant metric ( . ~ 2 . i. a) Show that any connection on G is competely determined by its effect on the frame field. we may assume 2 1 # 0. 4 ) E B 4 \ 0 = U. Therefore.j = 1. we have [ c Y .s+.n. . ( 7 )) by setting.2223.i. (2122 222322124. we obtain the following two linearly independent vector fields CY and p which define a rank2 distribution orthogonal t o the above one and whose values at are given by = (2123+2224. By a long but straightforward calculation.n. ( : .))..227 Because ( 2 1 . 0. 2 32. + 3337 Let G be a Riemannian manifold with a global framefield {ei}y=l.without loss of generality. 2123. 1 Veiej = . and the framefield is leftinvariant..we have  n n .
we obtain V e i e j= . e j ] . Noting that g ( t ) is a geodesic of G and Xg(tl = we have obviously 9 y. where e is the unit element of G. we can well define x'e.228 Then. i. then V e i + e j ( e i e j ) = 0 implies that Veiej Veje. because the metric of G is biinvariant. b) For every left invariant vector field X on G.j=l n Thus defined operator V certainly satisfies all properties of a linear connection on G. if X = ei e j . . + + + On the other hand. 2 1 .(g)ej + C zigveiej.ej . Thus. Especially. we know that V x X = 0 is valid everywhere. = X .[ e i . vx.V e j e i= [ e i . the LeviCivita connection V satisfies V. = 0. let g ( t ) denote the unique 1parameter subgroup of G such that g(0) = e .x =  t=O Besides. e j ] . motivated by the properties of connection..
Part IV Real Analysis .
.
pn(z)= 2 .. 10n 10 +. For any z E [0.3 101 + . l ) Then 00 1 3n. V i < n .2 .) # 3). p n ( z ) # 2 and p n ( z ) # 3) and B = {z E [ O . 3rn 2 n : p .. # 3) E [0. 1) I h P n ( z ) # 2 and P() n . V i < n .+ . * p s Let = (1) U{z u{ . E LO. p n ( z ) = 2 . .3 10n1 kl + . (Stanford) Solution. Since c y ) n . p i ( z ) # 2 and pi(.1] be the set defined by the property that z E S if and only if in the decimal representation of z the first appearance of the digit 2 precedes the first appearance of the digit 3. 2 kl ..1 kn1 lo*1 + 4 1uR) and B = C u [$+.231 SECTION 1 MEASURABLITY AND MEASURE 4101 Let S c [0. (2) Vn. Prove that S is Lebesgue measurable and find its measure.. A = {z E [O. ( z ) 58 and (3) z = n=l . 10" l o + 3 It follows that both A and B are measurable and therefore is S.+ kn1 n = l kj#2.1) there is a unique sequence {pn(z))r'l of integers satisfying the following properties (1) 0 Then 5 p n ( z ) 5 9. 1) I V n .+ .1) I 3 n . p i ( z ) # 2 and pi(z) # 3).k . kn1 n=lki#2.
+ If T . Then [z] = [y] [r].y E &.1) n& are such that ( E [TI) n ( E [ s ] ) # 0 there are [z]. Consider on S1the equivalence relation: [z] [y] # z . If E is measurable then + + + T €[O. I E <. ( StanfoTd) Solution. s E [0. For . For any [z] E S1there is by the construction of E a unique element [y] E E such that z . [z] ++[. + 2 4102 Define S1= IR/Z endowed with the natural Lebesgue measure. 1) m(E+[r]) = C T m(E).e. i. and let E c S1be the set of these points. It follows that { E + [TI 1 T E [0.V[ E S'/ }. S1is an abelian group under the binary operation ([z].1)W 00.where T E [0.s < 1. a contradiction. It follows that z [z] = [y] by the construction of E .. [y] E E y mod& and therefore such that [z] [T] = [y] [ s ] .  E = {F Show that E is not measurable. a w €P. If m ( E ) = 0 then m(S1) = 0. each ( E S1/ choose a representative f E S1. 1) n&} is a partition of S1.1) n& is such that r G z .y m o d Z .I.y E &. which then implies that T = s since 1 < r .y E IR and [ ] denoting the class in S1.232 and we have 1 m ( S )= m(A) m ( B ) = . Let m denote the natural Lebesgue measure on S1such that m(S1) = 1.l)W + m(S1) = T€ C [O. Otherwise m ( S 1 ) = contradiction. . Thus [r] = [ s ] . for z. [y]) H [z + y ] and the inverse operation. too. So we conclude that S1= U ( E + [TI).
D c a c P ( X ) such that R is closed under the following operations: (i) finite disjoint unions. A )< T) 2 . y E B } > 0. (ii) differences A\B. x C R. let a be the smallest system.233 4103 Let X be a set and D c ' P ( X ) . R = R since for any A E R. (Iowa) Solution.'D closed under finite intersection. B subsets of I such R that inf{ 1 . (Iowa) Solution.B ) > 0. Prove that a = R. We have since the former is a ring containing D.yI I z E A . Let U = {Z E I d ( z . By equality (2). F'uthermore. Also we have since by the equality (1) the former is a ring containing 2). Let T = d ( A . 2 Prove or disprove that p * ( AU B ) = p * ( A ) +p*(B). 4104 Let p* be the Lebesgue outer measure on IR and A . We will show the equality. Denote ' by R the ring generated by D. Obviously. B c A.A = A n A.
We have \ o o m / n . that is for each B E B. A C WI C U . p ) with a finite. B C W2 2 V } i inf(p(W1) 1 open. that B is the Bore1 aalgebra.} be countably many measurable sets which are 00 mutually disjoint. Prove that p is countably additive. a) The sufficiency. Prove that p is countably additive if and only if it satisfies the following condition: If A. inf(X+Y)=infX+infY. finitely additive measure p. positive. T00 b) Now that suppose X is a locally compact Hausdorff space. is a decreasing sequence of sets with empty intersection then lim p(An) = 0. B C W2 C V} P * ( A )+ P * ( B ) . = i=n+l U Bi. Then n=l n An = 0. 4105 a) Consider a measurable space (X. and that p is a finite. Let {B.A U B inf(p(W1) C W} : inf{p(W) I W o p e n . We have p * ( Au B ) = = = inf{p(W) I W open. Let A. positive.234 Then U and V axe disjoint open sets containing A and B respectively. p ( B ) = sup{p(K) I K 5 B and K is compact}. finitely additive measure on B. X. Suppose moreover that p is regular. A U B C W C U U V ) E = + p ( W 2 ) I W open. A 5 cU } ~1 WI +inf{p(W2) I W2 open. Finite additivity means that whenever { B i } is a finite collection of mutually disjoint measurable sets. then p(UBi) = C p ( B i ) .Y where the equality (1) follows from the following equality GR. (Iowa) Solution.
( IndianaPurdue) Solution.} of measurable sets with empty intersection such that . 1 where M is any positive number. If m ( E ) = 0.1] + B. 03 # 0. let E C { z I f'(z) exists}.235 Therefore Y is a measure. If(y) . by a) there is a decreasing sequence {A. such that which implies that and therefore i=l n Thus { i=l Ki I n E N}is a decreasing sequence of nonempty compact subsets n=l in the compact space K1.21 if I Y . lf'(4I < MI.f(z>II MIY.too lim p(A. Denote F ={ . which contradicts the fact that n=l 4106 For f : [0. show that m(f(E)) = 0. The necessity is obvious. n For each n there is a compact Kn contained in A .21 < 1. . Set Fn = I z E F . So n K. It suffices to prove m ( f ( F ) )= 0. n .) = inf p ( A n ) > 0. b) If /I is not countably additive. 12 E E .
m ( F n ) = 0.y E Fn f l I n $ . Prove that m ( A )= 0. (Iowa) Solution.n + 1) such that + 1)) # 0. we have Therefore 5 k m * ( f ( F nf' I n . .n + 1) C U C (n. . + 1) n . If m ( A ) # 0 there is an n such that m ( A n (n. k ) < ME 5 M Thus m*(f(Fn))0. which implies that m ( f ( F ) )= 0. lim For any E > 0. There is an A n (n. k ) ) m ( I n .236 Then F1 C F2 C * . take a sequence {In$} of open intervals such that For 2 .a < b. and f(F)c uf(Fn) = ncu f(Fn). b E IR.n open subset U in ( n . = 4107 Suppose A c IR is Lebesgue measurable and assume that for any a.
We are left with the set K x . 1 Remove from each of them their middle parts of lengths X l l i j l . There are at most countably many disjoint intervals ( u j . Claim. we are left with two intervals 1 and 12. Then the total length of intervals removed in the k 1th step is + k i=l . For any n E W .n 1)). n + 1) = U A n ( a j . and keep doing this ad infinitum. Assume that it holds for any n 5 k . (Stanford) Solution. etc. j We have which deduces that m ( A n (n.237 where E < m(An (n.1] its middle part of length A. + 1))< E.n a contradiction. the total length of intervals removed in the nth step is X ( l The claim holds for n = 1. Prove that the set Kx has Lebesgue measure zero. 4108 Choose 0 < X < 1 and construct the Cantor set Kx as follows: Remove from [0. b j ) . bj)’s such that + Then A n ( n .
01. or p(z z + Y1 = 4 0 .lim p((.A y . n+oo n and therefore p is the Lebesgue measure.11) = P([O. Yes. = P(Y. Moreover. we conclude that g ( z ) = zg(1). From the rightcontinuity of g. p is the Lebesgue measure. y E IR with z < y. Y1.O]) n+oo n 1 = 1 .238 By induction the claim holds for any n E JV. y ] = g(y) . 4109 Suppose p is a positive Borel measure on lR such that (i) p([O. and (ii) p ( E ) = p ( z E) for any Borel set E of 2R and every z E B. However. Y L0 <0 + Y. I. Does this imply that p is the Lebesgue measure? Justify your answer.P ( { W 1 = 1 . y E B. (Iowa) Solution. For any z . It follows that the Lebesgue measure of Kx is 1 c n=l X ( 1 . It follows that g ( z ) 2 z . Y we have g(z + Y) = g ( z ) + d Y ) . It follows that the measure p is induced by g. g(1) = p((0. For any z E IR define + Then g : lR + l is nondecreasing and rightcontinuous. 11) . from either A 7 .1 = 0.l]) = 1.g()) = 1.lim (g(O) . for any R z .g(z). p ( z .
4111 Let X be Lebesgue measure on R. there is a Lebesgue measurable set A c E with X(A) = !.239 4110 Let U be a aalgebra of subsets of a set X and p n : U + lR be signed measures such that p ( E ) = lim p n ( E ) exists for every E E U. Then . for each E E U .G. by the VitaliHahnSaks Then k=l which shows that p is a generalized measure and therefore a signed measure. (Iowa) Solution.Show that for any Lebesgue measurable set E c 1R with X(E) = 1. is a finite measure and pn’s are absolutely continuous with respect to Z . ndoo is a signed measure. Given any mutually disjoint measurable sets En’s. Prove that u . Let. j (Iowa) .
Thus . . S is dense in C([O.f(Y)I x+ w 5 1 2 . co). It is continuous by the following inequality If(.Y E E . = {z Then m(A. Let S = span{ePa5 I n E lN u (0)) then S is a selfadjoint subalgebra of C([0. Show that if r w for all n = 0 . .. 20 lim f(z) = 0 and lim f(z) = 1. . .zo]. Put A = E n (co. .) Since f(z0) x+w . Show that there 1 exist two points z/ # z” in A with 2’ . It follows from the StoneWeierstrass Theorem. + T. Denote all the rational numbers in [1.T.YI. which then implies that p = 0. . u (Iowa) Solution. Therefore for any f E C([O.240 Solution. 1 by 1 X(E n (m. . A . 1 be a measurable set of positive measure.) = m(A) > 0. . c [I.oo]).7 3 . 4112 Let p be a complex Bore1 measure on [0.1] by rl. 1 .m]) under the supremum norm topology.. there is a point E E such that = f. co])separating the points of [0. 21. ( IndianaPurdue) Solution.z]). 4113 Let A c [O. .. . Define the function f : IR f(2) = +[0. co]. Z. = 0.z” rational. Denote A. 2 . 2 E E. then . which is required. 12 E A}.
. Prove that E is Lebesgue measurable with Lebesgue measure 0. (Iowa) Solution..+.1].T.+ + * a +  3"1 The converse inclusion is obvious..1).+1.. E is measurable and since + ["3 + 2. n Am = 0 if n # m. 3"' 3 [? 2. Then we c a n find z'... 1)\E = u { 00 Indeed. 4114 2. Thus  = 'I'm .241 U An c [1721. We will show next that * *  [O. for any 2 E [O. Then C m(An) I m([1. Also show that E + E = [0. 5 zl. Let n = min{k I 2 .l)\E.xnlI 1 and therefore 2 21 + . = 0 or 1 . I 2 and for any n there is an m 2 n such that 2 .1 . n=l 00 Suppose that A. . 3 31  3n1 xnl 7 = 2. 2 = n=l ? $ (where 0 I z. Therefore there must be some m.21) n=l 00 = 3. Take 2 E An n Am. there is at least one n such that 2. Obviously..+ .. By equality ( l ) .1 2 2 + + 3" .. n such that A n n Am # 0. =2 I1 + T. 2 E A such that ' ' z =2 ' + r.. which contradicts m ( A ) > 0. . E c [0.= 2) then 0 1 a:€ I l).
1 gl(z). 4115 Let f : It2 + IR+ be measurable. Take {r. limr.glllcx. [If I{. {zIg(z)= r } = {z [If gllco = . \ { z g z = T } \ = 0. () . = 2.gl)loo 5 E . there is an m 2 n such that # 2).T } \ = 0. . = +m. I r n . ~'$2'' Then z' = C 0 0 3 and .. = Since \ ( r c U{zI z E Enlarcctgz = r. n E N. and for any n. and z . < T. T.+~ r .1 < fi(z) I ~n}. n= 1 This completes the proof. Obvior. L E.242 m ( E )= 0. . Show that there exists g : I22 + El+ measurable such that (i) . Set fl(z)= f z + arcctgz. and let E > 0. CF n= 1 03 5 z .=l 00 X" $ belong to E so that = z..E + E [0. We have g 2 0 and For every r E B.1) say z = 2 .  Set g(z) = gl(z) . I g ( 4 = .sly. (IndianaPurdue) Solution.arcctgz. . n \ arcctgrc = T . Denote () En = { z Set g1 = TnXE. Since E C [0. . I 2. for all n. .arcctgz = T } TI.} such that 0 < r1 < r2 < . For any z E [0. (where 0 Define i].gllm 5 E (ii) for every T E B. 1z . 13. < E . < . llfl Then g ( )2a c c t g z .}I = 0. 2 ' ' = C .
2 uu (0 . b) Evaluate 1 1 f(z)dz. By the definition. a 1 . O . 1]\K = p 2 1 ai=0.243 4116 Let f be the function on [0. we have which then is measurable. . (b) lim f ( z ) = 0. Then [O.U p . Let K denote the Cantor ternary set. * a p . . a 1 a 2 . a) Prove that f is measurable. h0 lim m ( ( E+ h ) n E ) = m ( E ) . (Stanford) Solution. We will show first that for any Lebesgue measurable set E . z++w + (Illinois) Solution.1 1 .1] defined as follows f(z) = 0 if z is a point on the Cantor ternary set and f ( z ) = if 2 is in one of the complementary intervals of length 3 . . .1 2 ) . and 4117 Let E be a Lebesgue measurable subset of lR with m ( E ) < 03 and let f ( ~= m ( ( ~2 ) n E ) .P . ) Show that (a) f ( z ) is a continuous function on R.
h0 In general. so hO lirn m ( ( E + h)nE ) = m(E). 3 ~ .244 If E is of the form . then n 5 hm(En(E+h)) h+O <_ h+O lim m ( E n ( E + h ) ) 5 m ( E ) .) = m ( E ) . there is an increasing sequence { E n } of compact sets such that En C E and lirn m(E. n+mh+O n+m lim m ( E n )= lim lirn m ( ( E n n+m h10 h+O . ) o i )where ) r=l n (q. n+m Then m(E) = 5 + h) n En) lim lim m ( ( E + h) n E ) = lim m ( ( E + h) n E ) . for any E > 0. an open set E’ of the above form such that E c E’ and m(E’\E) < E .@i)’s are finitely many mutually disjoint open intervals of finite length. If E is a compact set.5 5 h+Om ( ( E +h ) n E ) + 3~ 5 m ( E )3 . h+O It follows that m ( E ) = lim m ( ( E+ h ) n E ) . then there is. Let F = E’\E one has m(E ) < m( El) = lim m((E’ hiO + h) n E’) = l i m m ( ( ( E + h ) n E ) u ( ( E + h ) n F ) U ( ( F + h ) nE ) G ( ( F + h ) n F ) ) h0 G 5 h m ( ( E + h ) n E ) + 3.G( a .
m ( ( E + (z . then for any n have n U ( z a i . If B is a bounded open set.z)\E) m ( E (z . z b i ) (ai.f(.bj)’s i<m u are mutually disjoint open intervals. ncc lim m ( ( E z) X++OO = = zr+cc lim ncc lim x++m + nE ) m ( ( E+ z) n E ) . .bi) n i=l <m and any z >0 we c ZB nB. say (a*. (z as y + z. In general there is an increasing sequence { E n } of compact sets such that En E and lim m ( E n )= m ( E ) . the line y = mz has a nonvoid m 1 intersection with A x A .245 so h0 lim m ( ( E + h)nE ) = m(E).)I = Im(((E+ Y)\(E + + Y) ) l 5 m ( ( E+ Y)\(E + = m ( E (Y .m ( ( E + z)\(E (a) For any y.m((En+ n E n ) + m((En + n E n ) lim m ( ( ( E+ z) n E)\((En z) n E n ) ) + m( (En + z) n En) + Z) Z) 5 nrn lim ( m ( ( E . 4118 Let A c l be a set of positive R Lebesgue measure. (b) If E is compact. z E 2R If(Y) . where 2 5 m 5 03 and ( a j .m ( E (Y .z)) n E ) m ( ~ ) + + + + M ( ( E + z)\(E + Y)) + + . Use this result to prove that there exists an E > 0 such that for any m E 2R with I .1 < E .)\(En + + z)) + m(E\En)) = 0.b i ) . then there is an T > 0 such that for any z > T . (Iowa) Solution. The claim follows. Then + E ) n E = 0. Prove that cp(z) = J X & ) X A ( t ) d t is continuous a t z = 1. n E ) .y ) \ E ) = m ( E ).y)) n E ) + 0.
(1) where p is the Lebesgue measure.} of compact sets such that U K .\K)) we have by (1) + P(Bn\K) In general.} of bounded open sets such that K = n=l n B. n n ..a) 5 ~ P ( z B ~ B ) 21 5 X’lim ~ l ( Z nB ) B 5~L(B). If K is a compact set.) aJ = 0. C A and p(A\ n=l 00 n= 1 u K. there is a decreasing sequence {B. By (2) we have Since .246 We have p(B) = sup n<m a = 1 C(bi .p ( z K n K ) I P(Z(B. n Bn) . 00 Since p ( z B . there is an increasing sequence {K.
p(S)]. ( A n A ) = 0. n03 a contradiction. 4119 Let p be a countably additive measure on a set S with p(S) < $00 that is without atoms. if A is a measurable set with p ( A ) > 0. Let A = n=l  P = { A I A measurable and p ( A ) < t o } / . this implies that m . Given a totally ordered set Q of P . then 03 = .  u A n . ) = P 5 to. Let where is an equivalence relation: A B if p(A\B) = p(B\A) = 0.) Solution. E B such that Imn . and therefore p ( A ) = lim p ( m n ( An A ) ) = 0. However. i. ] } of Q with { p ( A . if [ B ] 5 [An] for some n. then for 2 1 any n E mT there is an m. [B]= [A] since P(A\B) 5 C p(An\B) n=l 03 =0 .247 If there is no E > 0 with the property mentioned in the question.. ) } increasing to p. then [B]5 [ A ] .e.1 < and the line y = mnx has a void intersection with A x A . Then P 5 t o and there is an increasing sequence { [ A .p(S)) not in the range of p. then there is a measurable set B c A such that 0 < p ( B ) < p ( A ) . For any [ B ] E Q. (Courant Inst.B = sup{p(A) I [A] E Q } .403 lim p ( A .otherwise. If there is a t o E (O. Then P is a partially ordered set: [41 5 [B]if p(A\B) = 0. SO p ( A ) = /3 < t o and therefore [A] E 7'. let . Prove that the range of p is the closed interval [0.
It follows that [ B ]is a lower bound of S in R.p ( B n C ) = a = 0. if [Bn]I [C]for some n. (Iowa) . Let n=l then So p ( B ) = a > 0.p ( A ) and there is a decreasing sequence { [Eln]}of S with { p ( B . Given a = inf{p(B) I [B]E S}. by the assumption that there is a subset of COof Bo with 0 < ~ ( C O ) ( B o[CO] [Bo] [CO] [Bo]. contradiction. p(S\A) > t o .p ( A 0 ) whenever B C S\Ao and p ( B ) > 0.P(B) > O}/ .p ( A 0 ) . For any [Cl E S. then cr 2 t o . Let where is defined as above. Let  R ={BI B c S\Ao. then [B]5 [C].g. There is by Zorn’s Lemma a minimal element [Bo] R. Consider the calgebra 2 generated 3 by the compact G sets. (e. It follows that p ( B ) > t o . 5 as above. ) } decreasing t o a.otherwise [B]= [C]since n=l and ~ ( B \ C )= p ( B ) . Therefore p ( B ) > t o . However. Equip 72 with the partial order a totally ordered set S of R. 5 and # a 4120 Let X be a compact Hausdorff space.p ( A 0 ) ) .248 It follows that [A] is an upper bound of Q in P. [B]E R. There is by Zorn’s Lemma a maximal element [Ao]of P. <~ ). Show that any positive measure p on B which is finite a on compact sets is automatically regular.
Then we have V). .K ? compact and K E B } . 5 P(B\L) < P(K\L) p(K\L) < p ( K +E. Then 7E is a ring such that 23 = By definition. For any E > 0 there are a B E 0 and an M E X such that K C B and p(B\K) < E and such that M C B\L and p(B\L) < p ( M ) E . + + + K\L K nM c B\L. C K\L.U (Ki\Li) I K i .\E. Let K and 0 denote the classes of compact sets and open sets in 2 .V open and V E 23) p ( E ) = sup{p(K) 1 K C E. there are B1. Thus K\L is regular.. If { E . L in K . M) + E. E i = 1. . Let us preceed in five steps to show the regularity of p. Consequently. Step 1. Let us show each set U in 0 is inner regular. respec3 tively. . i= 1 n} is a finite class of mutually disjoint. Let a = { ¶=1 . Also recall that E is outer regular if p ( E ) = inf{p(V) 1 E and E is inner regular if C V.249 Solution.. Step 2. For any pair K . there is a V in 0 such that X\U C V and p ( V ) < p(X\V) E . . 7 ~ . B. Recall that sets in 23 are called Baire sets and each compact Baire set K is a Ga set. Then X\V 2 U and p ( U ) < p(X\V) E . regular Obviously For any E > 0. I 1 5 i ? sets. each set in K is outer regular. . Li E K}. where the symbol 0 means disjoint union. .) < . . 6 i=l Ei C ( Bi J i=l . then U Ei is regular. K\L is regular. For any c > 0. E 0 such that E C Bi with i p(B.
while the inner regularity follows from the following inequalities co Step 4. Obviously 00 n=l p ( E ) I inf{p(V) I E For any E G V E O}.250 and which shows the outer regularity of the following inequalities u Ei. If { E n } is a decreasing sequence of regular sets. The inner regularity follows from n i= 1 Step 3. If { E n 03 In E IN} is an increasing sequence of regular sets then (J En is regular. n=l n En is ... En > 0 there are 03 V1. n=l Let E = u En. then regular. V. . . then n= which shows the outer regularity of E . . Let V = U1 Vn E 0. . . E 0 such that C Vn and p(Vn) < p(En) + &.
S contains a. By steps 1 and 2. > n> . E ?E such that n= 1 and which shows the inner regularity of E .251 Let E = n En. 00 For any E > 0. By steps 3 and 4. Let Assume f ( ~is finite for all ) T > 0 and assume Prove that p is identically 0.. t monotone class. Let S = { E E 23 1 E is regular }. It 4121 Suppose that p is a nonnegative Bore1 measure on En.S is follows that S = B. The outer regularity follows from the following inequalities.u(En) = p ( E n 5 V E 0) Step 5.there is a K . n E lN. inf{p(V) I E V E 0} 5 inf inf{p(V) I En = inf.. (Indianu) .
l) %k + z B(Xk31)k+m Hence . a s ) ) & .1))where B ( z . E Given a compact set K of IR". K c u k i=l C(z"7) c C(0.1 be a finite Bore1 measure defined on lRn. T ) c B ( z ...s).&?Fir). 4122 Let 1. in K . Then T k i= 1 k i=l k = Z A ( C ( z $ T ) ) & 5 X(C(0. let s > 0 be such that K c C (0 . ) ' S are mutually disjoint. we get that p ( K ) = 0 and therefore p is identically 0. It is easy to show that Assume without loss of generality that B(z.2s) and C(z*. . (Indiana) Solution. Let K be any nonempty compact set of 1R" and let a = i n f f ( K ) . and define a function f on Enby f ( z ) = p ( B ( z . i= 1 Letting E + 0.252 Since C(z. Prove that f attains its minimum on each compact set of IR". 1) denotes the open ball centered at z of radius 1. There exists an sequence { z k } of K such that k+oo lim f(zk) =a. For any > 0 there exists an T > 0 such that g ( T ) < ( ~ T ) ~ EThere exist finitely many . zl. zk E IRn such that .
. . 4124 Let f : IR + R" be a function such that for all z. write z = 0.ala2.Ulaz...2800. ) Hence a is finite and f attains its minimum on K ..~~18)~l.= 0. 1 0 ... For any z E [0..611a2 * * * Un17 = O.an17.2'. l]\E.. .aia2 * * * ~~1699. E E . (ii) Determine whether E is a Borel set. 2 (*> Show that if E c IR is a measurable set with m l ( E ) = 0 then m . .alaz. 3 3 3 ..UlU2.  1 27 28 (Indiana) Solution. Let n = min{k 2 1 I ak = 7). Thus.~~lS. The reverse inclusion is obvious. ( f ( E ) ) = 0.}. . . It follows that 0.1] which can be written in a decimal expansion with no sevens appearing.. .l]\EU{(O. .2699..0.a. 1) = Q.yI.. If ak = 0 for all k 3 : > n then = 0. 4123 Let E be the set of all numbers in [0...anl 7: n = 1. y E R If(. 3 ' 100 ' 100 (i) Compute the Lebesgue measure of E .. So E is Borel measurable and W m ( E )= 1 n=l 9"l x = 1  1 10" 1 = 0. .= 0.. . . .a1a2 .) . E a contradiction. ..f(y)ln 5 e'zl+lyI 1 . E.253 which shows that P ( B ( X . (Indiana) . and therefore I # [O.Un17< x < O.
then f k is measurable.ail < CneZr&. It follows that r n . ( f ( ~ ) _< ) C C. Since the Lebesgue measure is complete f is measurable.I 2 E l ) = m * ( { zE uk I If(x)I 2 E } ) 5 C. (eZrlbi. For any E > 0. there is an open set U with X(U) < E such that f is continuous on E"\U (in the relative topology). Hence mn(f(E)) = 0. Assume that r A = A for every nonzero rational number r. T ) and rnl(U\E) < E . Let fk = fXEn\u. 1 It follows that {fk} converges t o f in measure.e2'Ibi .. For any E > 0 there exists an open set U such that E U C ( .254 Solution.bi)'s are mutually disjoint. 4125 Let f : R" + lR be an arbitrary function having the property that for each E > 0. Prove that f is measurable. a which implies that rn:(f(E)) = 0 and therefore f ( E ) is measurable. Prove that either A or R\A has Lebesgue measure zero.a i l ) ' ) . Write U = U(uilbi). (Indiana) . where (ai. 4126 T Let A c R be a Lebesgue measurable set and let r A = { r A I z E A } where is a real number.bi)) i c ~(f(ai). Then condition (*) implies that f((ai. (Indiana) Solution.r). Let u k be an open set such that x ( u k ) < and f is continuous on R n \ u k . m * ( { zI lfk f )(. Let E 2 (r. Assume without loss of generality that E is bounded.T .
define function f : lR* + [0. dY XK(Y)XLl(Ylz). For any compact subset L of B . If m(lR*\B) = m(R\A) > 0 there exists a compact subset K of E*\B with positive Lebesgue measure. we conclude that m ( L . There exists a sequence {En}r=p=lmutually disjoint measurable sets of of finite and positive measure such that X = let n=l u En.m).l ) = 0 and therefore m ( L ) = 0. T.. m ) ( ~ ( z = 1) such that p is absolutely ) continuous with respect to v . Y 2 E a*. and v is absolutely continuous with respect to p. . c )by m f(2) = 1. 4127 Let p be a cfinite measure on the measure space ( X . Then f is continuous and for any nonzero rational number Hence f( z) = 0 for any z E R* Since .255 Solution. Let lR* = R\{O} and 3 = A\{O}. (Indiana) Solution. For each measurable set E QJ then Y is the desired probability measure. It follows that m ( B ) = 0 and therefore m ( A ) = 0. Prove that there exists a probability measure Y on ( X . Then r B = B and T(R*\B) = R*\B for every nonzero rational number T .
There is a measurable set E of measure zero such that any x E (0. 2 ( 0 . (Illinois) Solution. l ) such that E > 0. For example. 1)\E i s a Lebesgue point o f f .256 SECTION 2 INTEGRAL 4201 Prove or disprove that the composition of any two Lebesgue integrable functions with compact support f . 1). 1 = x{o..) < E . m(J. (> = X { 0 } ( 4 and d . Assume that for any x E ( 0 . since g o f (z) 5 1. let f. and fdm = 0. l ) and every an open interval J. Prove that for every open interval I E ( 0 . g : IR + IR is still integrable.e. there is x EJ.. . (Stanford) Solution. Then f and g are integrable functions with compact support. the function g o f is not integrable. l ) fdm = 0 .l}(x). 4202 Let f E &(O. It is not true. i. However.
) < and It follows by equality (1) that f ( z ) = 0. Therefore 4203 Let (X. l ) such that z E J.257 For any 2 E (0.. Show that a M measurable function f : X + [0. 03) is integrable (i. m(J. one has Jx f d p < CO) if and only if the series n=O c 00 I f (z) 2 n}) (Iowa) converges. f ( z ) = 0 a. i. Then W .e. Suppose that f is integrable...e. p ) be a positive measure space with p ( X ) < 03.e. Solution. l)\E and for any n there is an open interval Jn & ( 0 . ..
(a) Yes. (Iowa) Solution. / p(t)dt > 0 4205 Let E be a Banach space.258 Conversely. (b) Is there a Borel measure p (positive or complex) on B with the property that J.p ) a probability space. + E . let ‘p 2 0 be a continuously differentiable function of compact support.?r. which shows that f is integrable. (X. taking value one on [1. f d p = f’(0) + (c for all continuously differentiable f : R of compact support? Justify. Then a contradiction occurs from the following limits JL% and Irn ( p ( t ) e t d t= . fdp = f (0) for all continuous f : B + (c of compact support? Justify. Let p ( E ) = x ~ ( 0for any Borel set E . 4204 (a) Is there a Borel measure p (positive or complex) on IR with the property that J . ) (b) No. and f : X such that g o f is pintegrable for every g E E‘. If there were such a Borel measure.1].
Define the linear operator T : E’ + L1(X). 4206 Let ( X .259 Define L : E‘ + IR.and let f and Show that either (a) f = g a. T is bounded. p H p of. (Iowa) Solution.e. p ) be a positive measure space with p(X) g be realvalued measurable functions with < co. By the closed graph theorem. If (b) does not hold.. Assume that pn + ‘p in E’ and T(pn) + h in L1(X). or (b) there exists an E E M such that (Iowa) Solution. It is ture that L E E”. M . It follows that pn o f converges t o p o f everywhere and to h in measure and therefore h = p o f in L 1 ( X ) . Does L E E”? Justify your answer. Since . then for any E E M . We have So L is bounded. L ( g ) = I g 0 fdp.
260 for any E E M . and S a closed set ina'.If .) = 0. 4207 Let ( X . > 0 the sets are measurable.A 1 5 E ~ } ' s . From the equalities and we conclude that p ( E + ) = p ( E . Show that { z E X 0. Since a' is second countable. I f(z)@ S } has measure (Iowa) Solution. Suppose that 1 JE ) P( E f@€S whenever E E M and p ( E ) > 0. .M . J. For any E fdP = J. It follows that Therefore (a) holds. one finds that a'\S is the union of countably many closed balls { z E C I Iz . d P . p ) be a positive measure space.
Thus .X. Therefore 2 +).X n I F E n ) ) # O * I ( But then E = {x E X & J”f d p belongs to { z E (J‘ I Iz .a contradiction. Take an N such that Suppose that E C [0.1] + (0.) . m) and let 0 < a 5 1. where 5 E ~ } ) . Denote El = E (f Then m(E2) < 5 .not to S.261 there is a t least an n such that P ( { ~ E Xf I ~ ) . EZ = E\E1. 0bviously. Show that where inf is extended over all measurable E c [0. so m(E1) > $. I 5 E ~ } .1].1] with m ( E ) 2 a . (IndianaPurdue) Solution. m ( E ) > a.XnI I If(. 4208 Let f : [0.
(Indiana) Solution. Since fn + f almost everywhere with respect to Lebesgue measure.f k .l ) ( t ) k=2 converges. a measurable set E of measure zero such that for any t E B\E. Let .262 4209 Let {fn} be a sequence of realvalued functions in L 1 ( B )and suppose that for some f E L1(IR). It follows that fn + f almost everywhere.and define Show that f ( z ) is finite a. (I Zin o is) l there is. Prove that Solution.e. n fn(t) = fl(t) k C ( f k . 4210 Let /I be a finite measure on lR. with respect to the Lebesgue measure on lR. n Therefore for any t E lR\E. by Levi’s Lemma.
where dv(z) = 1 dx +22’ by h b i n i ’ s Theorem. The conclusion follows from that the measure v and the Lebesgue measure are equivalent. 4211 Let ( X . the function is finite a.e. [p]. f n : X + [0. without loss of generality. d v ) .M . with respect to the measure v. .e. that p is totally afinite. and f : X t [0. CQ] a sequence of integrable functions. p ) be a positive measure space. i.e.and that Prove that (Iowa) Solution. m] an integrable function. Suppose that fn +f a. Assume..263 then g E L 1 ( R .
f n d p } is bounded and therefore admits a convergent subsequence. For any A E M . equivalent t o p. Let d v ( 2 ) = g ( z ) d p ( z ) . It follows that and therefore r r For any E > 0 there is a 6 > 0 such that Xn(E) < E whenever v ( E ) < 6. by the VitaliHahnSaks Theorem. There is by Egroff’s Theorem an E with v ( E ) < 6 such that converges t o f uniformly on X\E. Let Then g is integrable. If f n . d p } is any such subsequence. 71 E m* Then An << v. Let d X n ( z ) 1f n ( z ) d p ( z ) . Then v is a finite measure on X .264 n=l where Xn’s are mutually disjoint measurable sets of finite and positive measure. {s. Then 9 { $} . then by Fatou’s Lemma one has {s.
and that P ( E ) = 0 ep ( E ) = 0 for E E M . Show also that r is a positive. (b) Show that if f is a complex function on X which is measurable with respect to M. P . Conversely. If p ( E ) = 0 then P(E) = 0.}be a disjoint sequence of M such that X = U X . as desired. then (Iowa) Solution. ) > 0. The set function is of course a positive. l ) . (a) Let {X. positive measure on a ualgebra M in a set X. Let W=C1 O0 n=l 2 1 . and n=l 03 p ( X . + p(Xn) xx. finite measure on M .265 4212 Let p be a afinite. finite measure on M . (a) Show that there exists W E L1(p) which takes its values in the open interval ( 0 .
if f is integrable with respect to the measure p. Accordingly. then both sxf d j i and sxfWdp equal to 2 aiii(Ei).f is integrable with respect to p. . . i.. and since are lim n+w Jx\jnwfmWldp= lim n+w m+cc ~f~fmldji=~. .fdF=J X fWdP Conversely. n f =CaixEi! i=l where p ( E i ) i= 1 < +a. if fW is integrable with respect to pl assuming without loss of generality that f is positivevalued. there is a sequence of simple functions { f n } such that lfnl 5 f and lim fn = f .= i l . Since and lim fn n+cc = f.l n . Then fnW is integrable with respect noo to p and hence fn is integrable with respect to p. we see that fW is integrable with respect to p. there is a sequence { fn} of simple functions integrable with respect to p such that Then fnW’s integrable with respect to p . m+oo ncc lim fnW = f W . n In general.e. . We have and therefore J.266 (b) If f is a simple function.
There is a continuous function g : [0. fndm= 0.267 4213 Let m denote the Lebesgue measure on [O. Then we have = n+m lilfnXAdml It follows that Jiir J. For any E > 0.1] t R such that (3) 0 5 g 5 1. 4214 Let { fn} be a sequence of nonnegative measurable functions in LP( R) for some 1 < p < 00. (Iowa) Solution. ( 4 ) g = 1 on K and (5) g = 0 outside U .fndml % hdm < E whenever m(E) < 6. IJ.1]. fndm + 0 for each Bore1 subset A [0.Suppose that (i) fngdm + 0 for each g E C([O. For such a 6 there are a compact set K and an open set U such that (1) K A C U and (2) m(U\K) < 6. if : ( IndianaPvrdm) . 1 and let (fn) be a sequence in 3 L 1 ( m )and h a nonnegative element of L1(m). there is a 6 > 0 such that IE . Show that fn 4 f(P) and only if f 4 fP(L1).11) and (ii) Ifn I h for all n. Show that J.
take .” Then fn f p l . it follows that +  Jfi For any c N1 such that .268 Then f n 5 f and IL . Therefore Conversely.f I L Ifn . Just as above. 2 f. Suppose that J If.fl which implies that IIE . +  Since IJu~ > 0.f (Ip + 0. we have Since T:+X SO =f.”+fP. f > 0.  PI 0.fP)/ I J tfi Jfp.+0.
4215 Let 1 5 p < 00. Take N such that N 2 a n d m ( l f n . llfnllp and llfn . UP Ifn flp) for any n >N. Denote 7 = &/m(A)f. a measurable set. Wehave N1 (J. < EP e if m ( e ) < S and n > N1. Take A .. 5 M . < f P &PP Then J. Thus Jf. Vn ..fllp f . such that m ( A ) < +m and J. f: > N1. (a) Give an example where { f n } converges to f pointwise. All parts refer to Lebesgue measure on R.269 for n > N1 and any measurable set E .. Take S > 0 such that < EP for n if m ( e ) < 6. 0.f l > q ) < S i f n > N .
llfllp 5 M . i ) .1].fJPdz = 0. then +  llfn .e.fllp 0 (IndianaPurdue) Solution. Then gn 2 0. Then zE lim f n ( z )= f ( z ) = 0. we have Therefore no3 I f n .l f n . on [0. (c) Set gn = 2"(lfn Ip + Ifl") .fll = llfnll = 1. Set ncc fn =n ~ ( ~ . Using the Fatou lemma. and n+m lim g n ( z ) = 2 p + ' l f l p pointwise. (c) Show that if { f n } converges to f pointwise and llfnllp + IlfllP.13. 4216 Suppose f n is a sequence of measurable functions on [0. what can you conclude about Ilfllp? Justify this conclusion.270 (b) If { f n } converges to f pointwise and l ] f n l l p + M < $00.1] with 1' and f n + lfn(z)12dz 5 10 0 a. (a) Consider L[O.f l P . By the Fatou's lemma. Prove 1' lfnldz t 0 . [O. 1 1 and (b) We conclude that llfn .
1 We have by the CauchySchwartz Inequality and therefore Let E + 0. Show that i f f is an integrable function with values in [1. then lim P10 (J.1] such that (1) m([O.27 1 Hint.1 I E ) < E and (2) f n converges to 0 uniformly on [O. Otherwise. p ) be a probability space (a positive measure space with p ( X ) = 1).log f d p ) . (Iowa) Solution. Use Egorov and CauchySchwartz. 1]\E. It suffices to show that If f = 1 a.M . equality (1) holds. For any E > 0 there is by Egorov’s Theorem a measurable set E c [0. for any 0 < p < 1 and any 2 EX . (Stanford) Solution.f p d p ) ’ = exp ( J . and we have rl 4217 Let (X.e.00)..
4218 Let { a n } n > 2 be a sequence of real numbers with lan[ < logn. . 1+ ( f P n Pn  lim n00 ( log(1 + JJfP" J. n fn(x) = j+k5. Consider 00 n=2 a) Prove that this series converges in L1[2.(fP"   1)dP) J#"  1)dP Pn = Llogfdp. l ) decreasing to 0. (Stanford) Solution.m). k=2 x 2 2. a) Since The series b) Let C n=2 00 ann2 converges in L1[2.272 By the Dominated Convergence Theorem we have nco lim log(Jx f P n d P ) Pn = = n lim 00 log(J. where { p n } is any sequence of ( 0 . 00) b) Prove that where the sum on the right is the pointwise limit (you need not prove that this pointwise limit exists).
we have = X(S). By the RiemannLebesgue Lemma. (Iowa) Solution. 1 2 . 2 Theorem. and by the Dominated Convergence 4219 Let S be a bounded Lebesgue measurable set in IR. Show that where X is Lebesgue measure. and let { c n } be a sequence in lR.273 Then (i) nw fn(z) = lim 00 cc C aklc" k=2 and (ii) Ifn(z)I 5 C lanln". k=2 00 By a). H IakIk" k=2 is integrable.
Y)ldXdY rdrde drde (b) We have . be explicit!) (Iowa) Solution. Y ?M2+Y211) If(.274 4220 (a) Is the function Lebesgue integrable on the unit square 0 5 x 5 1. (a) The function f is not integrable on the unit square. (c) Does the integral exist? (If you use a theorem. for ) b .. 0 5 y 5 l? (b) Compute the repeated integrals in the two orders.
p ) be a measure space and f E L ( p ) . Define the function H .275 and 4221 Let ( X . (Iowa) Solution. It is easy to show that for any x 20 2.l).M . In( 1 + x’) 5 It follows that for any n E lN the function n In( 1 !$) is dominated by and therefore integrable. By the Dominated Convergence Theorem + If I 4222 e t z f ( t ) is in Suppose that f is a measurable real valued function on l such that t R L 1 ( E )for all x E (1. Evaluate ncc lim L n l n ( l + (!)’) dp.
(x 4 and t A 2 Iy .l). for any y ) > z. 2 0 and x < y < H < 0 and x < y < A' lzz. and we conclude that t H e " t f ( t ) is also in L1(E). l. It kkz By the same method as above. Since for any x E (1. l ) .276 for all x E (. < z < y) = \ e t z " t 2 f ( t ) ( z. For any fixed z E (1. I). Next we show that dx E (1.Prove that 'p is differentiable on (. +m etztf(t)dt. e"f(t)dt = m J_.1.x ) ~ (x < z' < z ) = I(ett  e")tf(t)l. t t Iy .l). we see that t follows that 4223 Evaluate . (Iowa) Solution.1 .zlet"t2f(t). et"t2f(t) is integrable.z l e ' z X t f(t).
)n =ex. Since for any n E IV and z 2 0. (Stanfod) Solution. . Hint.. we have (I+ x ) n s e x . i=l )(l&).. By the Dominated Convergence Theorem = brn e"dx = 1. First show that (1 E)n 5 e" for x + 2 0.277 justifying any interchange of limits you use.(lS Z! i=l + and ncc lim ( 1 + .
A. and Indeed. Under the condition that f In f is integrable we conclude that both f and In f are integrable. let Xi = nV (the natural numbers). otherwise and 0 5 In f 5 f In f we see that both f and In f are integrable. Let Mi = 2N (the galgebras of all subsets of nV) and let pi be the counting measure.2. For the function f : X I x X z + B defined by . Since l n t 5 t l n t for any t > 0.278 4224 Let ( X . p ) be a probability space and f : X function. since ' '{ f ( x ) 5 el "f'ln f . Prove or disprove: + [I. 4225 For i = 1.co) a measurable (Iowa) Solution.
j ) = 2). and for X > 0 define Cp(4 = P({Z E Q I f ( Z ) > XI) . For f E L1(Q. f ( i . the two iterated integrals exist Li€N and coincide by the Fubini'i Theorem. i H = 0. 2i.p ) be any measure space.. otherwise. j ) is integrable and Since f is integrable ( C I f l ( i . For any i E X I . j )= { 279 . 4226 Let (0.2i. j = i. j ) d p z ( j )= 2i + 2i Therefore For any j E X 2 . j = is 1. Compute the iterated integrals and How do you reconcile your answers with Fubini's Theorem? (Iowa) Solution. j H f(i. p ) . f ( i .j)is integrable and J.f ( i .
it may be helpful to consider f positive first. .I). (Iowa) Solution. 2 and II. On one hand. ~ ) ( l f l ( zA) is measurable. The function (z. The measurablity of 'p and follows from that they are monotone. 4227 Let f E L 2 ( 0 . Set .) + but on the other. X . By Fubini's Theorem we have provided that either side exists. A) H x ~ O . 0 Hint. are Bore1 measurable and that llflll = J m ( ' p ( A ) + +(A))dA. As usual.280 and 4(A)= P ( { Z Show that the functions 'p E c I f(z) > A>).
. l ] . We have ( I J1h 0 F(z + h) . (UC. y E [ O . Imine) Solution.281 Show that for each h. 0 < h < 1 where c is a positive number independent of f. 1) and assume that f(z)g(y) = f ( y ) g ( z ) for all z.g E L1(O.F ( z ) h l2 dz) 4228 Let f.
282 where A = {(Z.Y) I 0 5 2 < Y 511 (Iowa) and dA denotes the planar Lebesgue measure. .
p > 1 and be positive numbers c.m). is a nondecreasing function of p E (1. . . for i E { 1. i = 1 .n}. * f n E Cp(/l)? (Iowa) Justify. q E (1.. p ) be a fixed measure space. w) with p < q .=. n 1 1 i=l If. .w)valued measurable function f on X . let M such that p.. 4302 Let (X.n. Under what circumstances is I ( p ) strictly increasing as a function of p? (Iowa) Solution. we have + which shows that I ( p ) is a nondecreasing function of p E (1..283 SECTION 3 SPACE OF INTEGRABLE FUNCTIONS 4301 Let X be a positive measure space of total measure 1. m). . . . The function I is strictly increasing if and only if f is not almost everywhere equal to the constant function. let CY = and = then $ = 1. > 1. . By Holder’s inequality. . fi E CPi ( p ) . For any p . Show that for any [0.. must it be the case that fifi .
c}. f k ) f k + l E LP and which completes the proof. Yes. Then for n=k+l. Assume that the conclusion and the inequality (1) hold for n = k . 1 o < +.fnllp 5 llflllpl .and L"O(p)? Justify your answers....*. f i * ..284 Solution. L z ( p ) . By inductive assumption f i e . 4303 Let X = { a .1  which shows that ppk c+k12p > Pl P k ? ' Pk+l PPk 1 pk+l?p < ' 9 1. (1) In case n = 2. fk belongs to 1 L Pk+:P ( P ) and Then by the case n = 2 we conclude that (fl . . . and by H6lder's inequality. What are the dimensions of the vector spaces L 1 ( p ) . .1 . b..1. let M be the cralgebra of all subsets of X . (Iowa) . P({b)) = 1 1 and P({. .)> = 00.+ 1= . Moreover llfl. IIfnIlp. lfilp E L"(p) and PI If2Ip E L F ( p ) . and let p be the measure determined by P ( { a ) ) = 0. fn E LP(p).
4304 lo Show that f d m = 0. Let f be a nonnegative function in P ( E )for some 1 5 p T > 0. c } we have dimLW(p) = 2. Since each function f of L 1 ( p )or L 2 ( p )vanishes at c and two functions taking the same value at b coincide in L 1 ( p ) or L 2 ( p ) . Denoting we have 4305 T +s =p. we conclude that dimL'(p) = dimL2(p) = 1. + < oc) and let . Also let f h ( z ) = f ( z h).285 Solution. ( IndianaPurdue) Solution. s > 0. Since two functions are equal in L"(p) if and only if they are identical on { b . We have dimL1(p) = dimL2(p) = 1 and dimLW(p) = 2.
then z h 4 E whenever z E E. (i) Obviously follows that fh E Lp(R). Set p = 1 + f .N . N ] . From f[fS L f . it E L1(R). f" E L f . Therefore + 4306 Let f : R +R+ be measurable. For any E > 0. Since we have . Then p > 1 and f + & = 1. and let 0 < T < 00. (ii) We claim that lim lhl+m 11 f i f s I I 1 = 0.286 (i) Show that fif s E L1(R). (ii) Investigate what happens to 11 fifslll as fi E Ihl + co. take N such that JEc f P < E where E = [ . Show that (IndianaPPurdue) Solution. and P +P = 1. (IndianaPurdue) Solution. Now if Ihl > 2 N .
l). : Therefore . Prove that (IZZinois) Solution.287 which implies 4307 Let f be a bounded measurable function on (0. llfllm = I I Ilf llm. For any Then E > 0.
Let f(z)= However for any t > 0 & .  1 (1+t)2  < 2.1] satisfying Determine those values of p . (Illinois) Solution. 1 . 1 l+t 4309 With Lebesgue measure on [0. m M it follows that f P is integrable.1 I llfllcc I 1) 1 is not compact in L1[O. then f E LP. 1 5 p < 00 for which f E LP and find the minimum value of p for which f may fail to be in P. If 1 5 p < 2.288 4308 Let f be a nonnegative measurable function on [0. prove that s = {f E C[O.1. for any p E [l.let . Then f 2 1 2 is not integrable. The minimum value of p for which f may fail to be in LP is 2. Indeed. 11). 3 (Iowa) Solution. 2). For each n E IN. It suffices to show that S is not closed in L1([O.13.
.289 4310 Let 'H be Hilbert space L 2 ( 0 .w) = 1 2= u(z)w(z)dz.e. for every w E 'H. n = 1 .1 such that 1 .1]. the Hilbert space consisting of functions f E L2[0.2r) by Holder's inequality. in the strong (i. with inner product defined by 7) ( u .w E L1(0.}. i.. By RiemannLebesgue Lemma. but not compact. + 0 as n + 00 in the weak topology of X .. w E 'H.  Consider the elements u. (a) Obviously..(z) = sin(nz) for z E (0.". 2 .too lim (un.i. .}rZl . (b) For any w E X . (Stanford) Solution. norm) topology of 2. . For any m. the set {u.e. (b) u. defined by u.w) = 0. Since it admits no convergent subsequence. n 2 1 with m # n which shows that {urn I m E IN} is closed.~ is bounded. u . E 'H.  {u. it is not compact. 4311 Let H1 be the Sobolev's space on the unit interval [0. Show that (a) the set is closed and bounded.2 1 .271).e.
1] too the series n=w C f^(n)e2ninzconverges to f(z) both in H1 and in L"O. Solution. Suppose For each h E lR define the function fh by f h ( z ) = f ( z . 2 a ) . 2 T ~ belongs to ~ ~ ( O .h). (i) Give the Fourier expansion of fh.290 n=cc where are Fourier coeffents o f f . Show that there exists constant C > 0 such that (Stanford) I l f I l P I Cllflll.fllz in terms of the a. and h. (Stanford) . (iii) Prove that unless f is constant almost everywhere. 4312 Let f be a periodic function on lR with period 2 a such that f 1 [ 0 . It follows from (1) that IlfllP I Cllflll. Since for any z E [O. (ii) Find the &norm llfh .
}span a dense subspace of L 2 ( 0 . 1 n = 1.f l l 2 = 0. there is an n # 0 such that an # 0. Then 4313 Let f n ( x ) be an orthonormal family of functions in the Hilbert space L2(0. (Iowa) Solution. n # 0.1]. and that this inequality is sharp (equality) if and only if {f. = 0. then a.. By Bessel’s inequality . Prove that for all 2 E [0. I f f is not constant almost everywhere.1).. It follows that fh = f and 11th . (i) We have n=w +w n=w (ii) We have n=w (iii) I f f is constant almost everywhere.1)..291 Solution.
fllz.fllL2(nz) = 0.then {fn} is a n orthonomal basis of L2(0.1). Then L t+O lim llft . Conversely. there is a sequence {fn} of continuous functions with compact support converging to f in L2(IR).fnll2  + llfn .1) while is a closed subspace of L2(0.fllz) i t0 2 l l f n W . In general. I~0] 5 z I 1) = s p a n { ~ ( ~ .292 If {fn I R. since s ~ a n { x ( o . let ft be the translate ft(z) = f ( t + z).l).l). {fn} span a dense subspace of L2(0.1) and therefore (1) is sharp.f n t l l 2 + llfnt . containing span{x(o.1] = { z y 12 E K . then t+O lim llft . E nV} span a dense subspace of L 2 ( 0 . y E [1. it follows and therefore.z] 1 0 5 z 5 l}. We have where K = supp(f) and . Prove that i f f is square integrable with respect to Lebesgue measure. (Stanfonl) Solution.1 b l Ia Ib 5 I} O is a dense subspace of L2(0.fllz + llfnt fnllz) = 2llfn . l]} is compact. Suppose that f : lR + a' be a continuous function with compact support.fllz 5 S(1lft . Then ft converges to f uniformly. .[1. 4314 I f f is a function on IR.
00. 4315 A trigonometric polynomial is a function p on lR of the form N n=N for some Cn €a'. t0 lim lift . we have . there exists a trigonometric polynomial N n=N such that Ilf and PI103 <E ICnl 5 lan] for all n.and let Show that for every E > 0.fll2 = 0. Then where a ( d z ) is the Harr measure on S1such that a(S') = 27r. Regard f as a continuous function on the unit circle S1 of a'.293 Letting n . Hint. You may wish t o think about harmonic functions on the unit disk. Let u be the harmonic extension o f f to the unit disc D = { z 1 IzI 5 1). Then . Suppose f is continuous and 21rperiodic on El. N 2 0. (Stanford) Solution.
1 ) I f ( z ) = 0 or 1.f(0) = 0).1 ) is not 3 dense in Loo([O. (ii) False.S is of cardinal N. n E m. (ii) Lw([O. S is not separable. Since C([O. 3 n+1 n Then as a subset of L”([O. C([O. fngn + f g in measure. 11). since the distance between any two elements in S is 1. 4317 Let be a sequence in LP([O. If lim f n = f in l.11) is not. Let 1 1 S = {f E L”([0.11) is separable while Lw([O. since .294 There is an N such that where r is fixed so that Put N n=N which is required. However.l]) (with Lebesgue measure) and = 1. l]).P {gn}T=l a sequence in Lq([O.l]). < 2 5 .1] is dense in Loo([O. 4316 Prove or disprove the following statements: (i) the set of continuous functions on the interval [0. (i) False.is it true that f n g n nw nw + f g in measure? Justify. 11) ( Stanfo rd) Solution. 13). where {fn}r=l + and lim gn = g in L Q . Yes. (Iowa) Solution. is a separable metric space. Indeed.
f) 2 0 and d(f. and that if and only if fn + f in measure. as usual. d(f.295 we have for any E >0 where A is the Lebesgue measure. h E S.e. iff .e. 4318 Let X be a measure space with measure p and suppose that p ( X ) < Let 00. (Here. = d(g. we have &  0 a. Obviously. g ) = p(f . fn + f in this metric (Stanford) Solution. two measurable complex functions are equivalent if they agree a. For f . S = {(equivalence class) of measurable complex functions on X}.1 For f E S. define Show that d ( f . g ..g) is a metric on S.g) = 0 iff g) f = g in S.
. If d(fn. . then %0 in measure by the equality f 1+f + { 2 EX I 1 + Ifn . f(z)g(z)dz. By the BanachSteinhaus Theorem. Then gn E Lq(IR) and for any f E L p ( l R ) . * . By the Dominated Convergence Theorem d(fn. f ++ f(z)g(z)dz is continuous on P ( R ) and therefore there is an h E L q ( R )such that . 5 n and 11 2 5 n.f) & 1+E + 0 as n t m. ( st u n f o d ) Solution.* L lsnfl 5 .s which then implies that g "2. Prove that there is a constant M such that .s all f E Lp(IR). then  d(fn.296 Therefore d is a metric on S. Define for each n E B a measurable function gn V gn(2) = { g!x" 5 hfl 5 if Ig(z)I else.f) +0. 4319 Let g be a measurable function such that lfgl < 00 for every f E P ( B ) (fixed p >_ 1). lim gn f = gf. h E L q ( R ) . Moreover. 191fl .(z)dz on P ( I Rconverges to f H . ) J pointwise. . If fn + f in measure. It follows that the sequence of bounded linear ntm functionah f H J f(z)g.fl Ifn  fl > &} =  { 2 E XI Ifn  f l 2 lE '} (O<&<l).f ) + 0.5 Isfl.
Y P Y . *k 1 k E L2(S1)} is (Iowa) Solution. We have ~ ~is an orthonormal basis of } ~ ~ a Therefore {f * k I k E L 2 ( S 1 ) }contains L2( S1).297 4320 Let S1 denote the unit circle (the set of complex numbers with modulus one.(O) = elne then { L2(S1). or the real numbers modulo 2a). where the integral is with respect to Lebesgue measure. Let for any n E Z.x.Do not neglect to check the measurablity of f *!l. The convolution of two functions on S1 is f *g(a) =  2'. . too f(y)g(z . g E L 2 ( R ) . {xn I n €27) and therefore is dense in 4321 For functions f .define the convolution f * g by f * 9 = J_. Suppose that f is an element of L2(S1) with the property that its Fourier coefficient is nonzero for all n E Z. I'" f(%7(ct  wo. a) Show that f * g E L o o ( R ) .Show that the linear space { f dense in L2(S1).
Show that Tf is a bounded linear operator. By the Closed Graph Theorem. Tfis bounded. (Iowa) Solution. Hint.9 / 1 2 we see that h = f + 0. oo  YVdY) . By (l).define + (1) + i ([r If(y)l2dY) (Jt" Id. Indeed. Since by Holder's inequality If(y)g(z .298 b) Suppose that f has the property that for all g E L2(BZ) the convolution f * g is also an element of L2(IR). Since l(f * 9 n . Define Tf : L 2 ( R )+ L2(BZ)by T f ( g ) f * g . b) Let { g n } be a sequence of L2(IR) such that gn t g and Tfgn+ h in L 2 ( R ) . f * g(z) is well defined. 4322 I f f E L 1 ( R ) . a) We will show that f * g is uniformly continuous. I r+m I = 0.f * g)(z)l I llfllzllgn . = It is evident that Tf is a linear operator. 2 E lR * g = T f g . you need not check this.f * g is bounded and therefore belongs to L m ( B ) .Y)ldY II l f l l 2 l l s l l 2 . closed graph theorem.
(Y let . co) be an essentially bounded function.1] + [0. for any f E L1(IR).299 Prove that. llflloo > 0. ?(() Solution. have we 4323 Let f : [0. ( Stanford) (1) In general. Show (Indiana) Solution. there is a sequence {fn} of functions of the form (1) such that Then Letting n + +a. If f is simple. say +0 as + 00. For any a with 0 < < Ilflloo.
300 and F. p ) be a measure space for which p ( X ) < 00. For any k E N . by = 0. Let 1 < p < Suppose that { f k } is a sequence in p ( x )such that ~upIIfkI(~ 00 and < k fk(Z) n+m lim = f ( z ) exists for pa .e . then X(E. 0 Let ( X . we get that 4324 0 .l]\E. . where X is the Lebesgue measure. Hence Letting a 1 llflloo. and Lp(x) reflexive there exists a subsequence is converging to If1 weakly in I. = [O. also fill > 0. Prove that (Indiana) Solution. Since { l f k l } is bounded in P ( X ) {(fk. M . such that lim km > 0 there exists asubsequence of {fk}. If on the contrary k+m lim Ilfkflll llfk denoted {fk}. 2. I} of {lfkl Ik E m}.) > 0. the Dominated Convergence Theorem.p(X).
Then Letting E + 0.301 By the VitaliHahnSaks Theorem. E > 0 there exists a 6 > 0 such r r whenever p ( E ) < 6. a. . for any that for k = 1. For such S > 0 there exists by Egroff’s Theorem a set E such that p ( E ) < S and {fk} converges to f uniformly on X\E. ..2. we get that a contradiction.
Z) L y’ 4(Y) L g(z1y) L By (1) we have an inclusion (1) while the latter set is a t most countable. Y E (071)). Show that lim g(z. since by (l). y).f’(z) = 4(z) whenever # is continuous at z.f(. y) = +(z) exists except in a countable Y+X + set and that for 0 5 z 5 y 5 1 we have Hint.) L g(Y1 Z)(Y  ( for all 2. v 5 y g(u..13 + lR and g : [0. (*) Assume also that g is nondecreasing in each variable. z) then 4(z) 4 is nondecreasing and L !J(Y. i. u 5 z. 1 + lR satisfy the inequality 3 f(Y) .e. So lim g(z. However.1] x [O. . Let #(z) = g(z. Again by (1) f is Lipschitz and therefore absolutely continuous. Observe that (*) is equivalent to (Iowa) Solution.302 SECTION 4 DIFFERENTIAL 440 1 Let f : [0. y) Y+X # 4(z) exists except in a countable set. v) 5 g(z.
1].1] there is a sequence { g n } of continuous functions Ign(z)l rl IM. with for all 2. 1 1 For any n E A . Conversely.x E [0.y E [0.303 4402 Prove that a function f : [0. (Stanford) Solution.1] + R is Lipschitz. 11. Since I ‘ z I f() such that and = f(0) + i= 2 f’(x)dz. . one is to use the Fundamental Theorem of Calculus. (11) fn(x) + f ( x ) for all 1 E [0. If f is Lipschitz. by the Mean Value Theorem This completes the proof. 5 M. then f is differentiable almost everywhere and moreover for any x E [O. 1 1 f) ( . E [O. 1 + R such that 1 (I) [fA(x)l I M for all x E [O. if and only if there is a sequence of continuously differentiable functions f n : [O. There are several different ways to do this problem. define T which is required. : Hint.1].
Let n=l in L1\O. 1 . 1 . (IndianaPurdue) Solution.1] such that (a) f(z)= n=l C 00 fn(z) converges for every 2 E [O. If both f and f’ are in L1(R) show that (i) f’ = 0. Then 1 It follows that f is absolutely continuous. (nlinois) Solution. (ii) f ( z ) + 0 as 121 + 00. 4404 Assume that f E A C ( I ) for every I c R. Since f E A C ( I ) . 3 Show that f is absolutely continuous on [0.304 4403 Let {fn} be a sequence of absolutely continuous realvalued functions on [0.1]. SR .for every I c R.
In the same way. we have lim f(z)= 0 and x+w Thus L f‘(z)dz = J_. fn(z) = 1’ fXW.f(O) = f(o).13. L’ which implies f E AC( [0.show that there is f E AC([O.}is Cauchy ( L l ) . 1 ( IndianaPurdue) Solution.1 ) such that 1 f n + f uniformly on [0.305 Since f’ E L1. 4405 Let { f n } C AC([O.l]). there exists g E L1 such that f : + g. Therefore we f’(z)dz = Iim f(z) .1]. Since x++w f E L1. lim x+w sox’ ( t ) d t exists. must have lim f ( z ) = 0. which means f x+w x++w lim f(z)exists. 0 f’(z)dz + I’” f ’ ( z ) d z = 0.1 ) such that fn + f uniformly on [0. 11).fn(0) = 0 for every n. Since fn E AC( [0. Thus So there exists an f E C([O. Then Moreover. If {f. l]). .
(2) 3 (1). Then Suppose that 22 . lGnl I . Take 6 > 0 such that If(bj)f(aj)l 5 E for any Ij = [ u j . (b) Show that the following two statements are equivalent. Suppose that 22 . :2n 1 Let . bj] c IIj I 5 6. (1) f : I + R is Lipschitz (2) E > 0 3 35 = S ( E ) > 0 such that { I j = [ a j .21) < 5. be open subsets of [0.. 4407 Let Gn.306 4406 (a) Assume that f E A C ( I ) and f’ E L”(I). For any 2 1 . < C N = 2 2 . such that Then N l f ( z 1 ) .f(zl)l < 1. E I . Take { c j } such that x1 = cg < c1 < .1)1 I Therefore we find a Lipschitz constant 1 $.1] such that GI 3 G 2 3 * * * . Take N such that f 5 < 6. we give the following fact. .2 E I ..2 1 < 5. I. Show that f is Lipschitz. .f(. I f ( 4 . n = 1 . 2 1 < 2 2 . Take N 2 0. 2 (b) (1) =+ (2) obviously. So we have 4< N(z2 .1 < 6. 2 2 . 2 .b j ] } c I with (IndianaPurdue) Solution.2 1 2 6. (a) For any zl. c.c.
Take x E GK. b. take N such that C N+1 M & < 5.= &.11).f(x’)l = C IG. (ii) Suppose that there exists no such that Gn = 2‘‘ 2 2 8 for n 2 n o . (ii) If(x’) .al.a i ) < 6. Then . b E (x . Solution. x” E [0. If(x”) . &  2 which means f E AC([O. there is k’ > K .S. i=l n=l K n=l K 1 = Klb.1] iff there exists no such that (IndianaPurdue) E > 0. suppose that for any natural number K .6. If {(ailbi)} is a sequence of disjoint open intervals in [0.x‘’]~5 no1x” 1 no Conversely. 6 > 0 such that (x .x 6) c GK. (i) For any (9 f 5 MIX’.1] and C ( b i . then < & + N C(bi U i ) < & + . Then GK # 0.11). for > x’. n 2 no.x”1. Take a.307 Show that AC([O. . n [x’. Set 6 = &.f(x”)I Gn = 0. a < b. x Then + 6) c GK = K Gi. x’. + a .2’1. Gkt # 0.
f is Then 1 E 1 1 . 1 and 1 E > 0. (Iowa) Solution.let us first establish the following equality provided that either side is finite. Indeed. The function f is of bounded variation if and only if To show (l).l).308 4408 For what values ‘a’ and ‘b’ is the function (i) of bounded variation in (1.1]. (ii) defferentiable a t ‘0’. for any of bounded variation on [ E . since f is continuously differentiable on (0.
> Ibl xb . If a + b 5 0. there is a 6 that 1bl2 Let N = > 0 such that 0 < x 5 6 implies d3 . Then sup O<E<' If(&)I < iff a + b 2 0. b = 0. b < 0. 1& + 1. since and . while a 2 0 implies that sup O<e<l J' lf'(x)ldx = sup J' . Then b1 If a + b < 0.1]. for Since a + 6 = 0 and a + b > 0 imply respectively that and ' the function f is integrable on (0. Case 111. 0<€<1 00 E Case 11.309 Case I. Then sup 0<€<1 If (&)I < 00 iff a 2 0. b > 0.2 I4 .:l!al € dx < oc).
27r] and f’(z) = 0 a. 4409 Prove or disprove a) Let f be a real function on [0. 2 ~ with f’(z) = 0 a. then and and therefore f’ is integrable. .e. a) The conclusion is ture. we have rl If a + b > 0. (Stanford) Solution.yI. all 2 2 .27r].27r] satisfying ..e.)ld.) f is absolutely continuous.310 I’ and If’(. Then f must be constant. on [0. f is constant. Since by the condition that If(.y E [0. (ii) f is differentiable a t 0 if and only if { %>l or { b<o a > 1. b) Let f be a real valued function of bounded variation on [ 0 . = +m. ] Then f must be constant.f(y)l <: 1 .
( t ) d t cc oo . 4410 Let f . g E L1(IR). Since the function is of bounded variation. Can you deduce that f’(x) = g(z) a. a+t g(z)dz +c a. Show a1 b) if a+h h+O f(z)dz = c.e. c such that f ( u + t )= J. T ) and f ( x ) = 1 on [T. (S t a n f o d ) Solution.. Then f is of bounded variation with f’(z) = 0 a. a) For any E > 0 there is a continuous function : lR + lR such that xL f(t)dt = j z f+ ( t ) d t j z f. without loss of generality. c ) if then there are constants a .31 1 b) is false. but f is not constant. it is differentiable almost everywhere by Lebesgue’s Theorem. 27~1. Let f(x) = 0 on [0.e.e. Assume.? State explicitly the theorems you use. that f and g are realvalued.
. we see that and therefore b) We have by a) h0 lim f(' a + h.f ( x ) d Z h = f ( a + t ) . . c) By a). there is an a E BZ and a c such that I ra+h J a k.e.c a. f ( z ) d x = c. t E IR.312 We have Letting c t 0.
t E IR.b) + Dl be measurable. (2) Show that the smallest constant M in (b) is equal to Ilfll.e. < 2 of ( a . . 1. we have = 0. (1) Prove that the following two statements are equivalent: (a) There is an f E L2(a.e. (b) There is an M > 0 such that for any finite partition zo < z 1 < ..b) such that F(z)= lz fdm ( m the Lebesgue measure). (Iowa) ..313 By b) and by the assumption.e.. It follows that f’(z) = g(z) a. a.b). 4411 Let F : ( a .’.
b ) . By the CauchySchwarz inequality (b) 3 (a) Obviously. a t most countable).b). Since a F is absolutely continuous. .. n E lN. We will show that if E l . . .we have on)of ( a .En are any mutually disjoint measurable sets. say Ei = 3EAi .b ) E such that f F(x) = LZ f(t)dt + c.. for any mutually disjoint intervals (a. o ( a i j .314 Solution. then for any E > 0 If Ei's are open sets of ( a . & ) (A. . 2 E (a. (1) (a) 3 (b).b). and therefore there is an integrable function f : ( a .
. I 65 f ( z ) < 1f ( z )5 and E. . ) . In general. . there is for each i an increasing sequence { E . Let E = {z ’ . . Applying (3) to the 300 mutually disjoint compact sets Elj. If Ei’s are compact sets there is for each i a decreasing sequence { E i j } E 1 of open sets of ( a . j } E 1of compact sets such that j=1 u Eij 00 E. Enj and passing to the limits we see that (3) holds for E l .=I j € A i P i j . . . En. .a i j 5 M. To show that f is square integrable. .Enj and passing to the limits we see that (3) holds in general. . . . Applying (3) to the mutually disjoint open sets Elj. = {x Then M I < 6}. j ) = m ( E . f I 2 . b) such that Ei = j=1 Eij and Eil n E k l = 0 (i # k ) . . it suffices to show that 00 where En = {z In 5 If(z)12 < n + 1). and lim p ( E .315 then by inequality (2) I 2 00 It:.
a ) ) such that + Then Illf1I2. It follows that sequence {fn}rz1 b] such that 11 of C[a. then 11.11 f E C[a. For any E (zip'. .b].b ) . For any fn  f E L 2 ( a .316 5 M. Moreover 5 l l f l l n by (1). so We have Ilfnll2 = Ilf112. there is is a norm on L 2 ( a . (2) Define.b]. IlfII = nlimO llfnll = 40 llfll = Ilfll2 for any f E C[a. z i ) (xn = a k ( b . for each f E L 2 ( a .b ) there is a fll2 f 0. This completes the proof. b )by (3).
We have and 4413 Let X be the Lebesgue measure on R2 and R defined by: 2 /I and v the Borel measures on . . A. t E E e X(E) = 0 e q5(E) = o . (Iowa) Solution. where X the Lebesgue measure.Y and q5 on M by 1 E Ida and $ ( E )= IdP.e. Does 4~ exist? Does exist? Compute the value of the RadonNikodym d? derivatives that exist. 2 p ( E )= and ] 2tdt r E + ( E )= We have 1 E 3t2dt.1]. Let a ( t ) = t 2 . t E E @ X(E) = 0 e 3 t 2 = 0 a.e.317 4412 p(t) = t 3 and define measures p ( E )= Let X = [0. Justify. M the calgebra of Borel subsets of X. For any Borel subset E of X . p(E)= 0 It follows that p 4  e 2t = 0 a.
Define for any Lebesgue measurable subset E of [0. m].e. Since and we have " 1 du n=l and dv d X Therefore (1) holds.. i. = n=l c " 1 $~(n.in]. << p2 or/and p2 << pl? (Iowa) . 4414 Let p be the Lebesgue measure on [0.2n)n A ) = 0 and therefore p ( ( n . X((n. let A = 21 and then v ( A )= 0 but p ( A ) = We show next that (g. i.318 Is p << v or u << p? Find the corresponding derivaties if they exist. find the corresponding derivatives. however. It is true that v << p and it is false that p << v.pn) n A ) = 0. Is p If so. (Iowa) Solution. m]. Because p ( A ) = 0 implies for all n E N . v ( A ) = 0.
. that q5(tg(. Moreover From p2([0.1)) = 1..e. and assume g E L(lR). Define f(t) = J dJ(tg(z))dz. t E IR. Since r 1 we have It follows that both p1 << p2 and p 2 << p1.39 1 Solution. p << p2 does not hold. continuously differentiable function with a R R bounded derivative.)) E L(lR) for all t E R). 4415 Let C#J : l + l be a bounded. l R a) What additional assumption on 4 will insure that f is well defined? (i.1)) = 0 and p([O.
320 b) Under the additional assumption in a) above.) = 1 . I). (Indiana) Solution. There are sequences {a. f’(t) exists for all t E IR. i. Prove that . z there exists an s with 0 < s < 1 such that Since b. show that f is differentiable. 4416 Let ( f k } denote a sequence of nondecreasing functions defined on ( 0 .} such that 0 < a. We will show that Indeed.e. nco < 1. for any t . (Indiana) Solution.} and {b.a. for any s. l ) with the property that lim f k ( Z ) = 1 for almost every z E (0. lim (bn . Indeed. t E R (s # t). and The conclusion follows from the Dominated Convergence Theorem. a.fi(z) = O for almost every lim krn k+oo z E (0. Assume 4(0) = 0 then f is well defined.l). < b..
.l).321 and k+m lim fk(a.) = lim kr 00 fk(bn) =1 It follows that lim krm f. lim fL(z) = 0 for almost every z E (0. is integrable and I’ ks Hence k+co f i ( z ) d z = 0.
". We have u ma 00 : = 1 z € UI%P. there must be a (Y such that m x E la. x can not be in 1. Show that is countable. 4502 Let { I a } .Q E r.).322 SECTION 5 MISCELLANEOUS PROBLEMS 4501 Let S c R be a set of real numbers with the property that IS1+. ( IndianaPurdue) Solution. Show that S is countable. n=l where cyn.?. .I for every finite subset {!I. .pnTU I:. Since S n (i. . Take n such that but : . 51 (IndianaPardae) Solution. Obviously u1 : aEr is an open set. . Sn}c S. For any r UIa\UI. .+S. S n (n. . be a collection of closed intervals in R. i) be finite and n ) and must S is countable.
there are sequences {kn} and {I.27r] n+co lim sin(nz) = 1. i H ai of I to D is injective and therefore I is countable.where D is a countable dense subset of X .21 I Ic. Let 2 A = {x E (0. For each i E I there is an ai E D such that a E Ui.is irrational}. 6 Icy.which is a contradiction. Moreover. P n )then x . c ( a n .) = . 4503 Show that any infinite set of nonempty.. It follows that the map. (Stanford) Solution. Therefore a a n which implies uIa\ u1 : a a is countable. since {Ic: . for any x E A . I E a } is a dense subgroup of lR. mutually disjoint. open sets in a separable metric space X is countable. Indeed for any x E A . 7~ 2 . Let {Ui I i E I } be such an infinite set.} o f Z such that nm 2 1 lirn (kn.27r) nm I .. lim sin(nx) = 1.323 If I . 7F Then A is a measurable set of measure 27r. Then ai # aj i whenever i # j .21. Thus we must have x = on or P. ( st anford) Solution. 4504 Prove that for almost every x E [0.
n+cc Otherwise lirn (3kA) = +co and ncc lirn ((3kh)n+cc 2 T 1 + 2(31. Show that there is a sequence of odd polynomials p n ( z ) with pn + f uniformly on [O. + 1))= 2 and therefore n+m lirn sin(3kAz) = lirn sin(3khx ncc + 2(31. Then f ( O ) = lim pn(O) = 0.I EZ}.} either increasing to t 0 0 or decreasing to m. If lirn Ic. nw . ( IndianaPurdue) Solution.'for every z E I .} admits a subsequence {k.11). + 1)w) = 1. take a polynomial pn such that 4506 Let f E C([O.2 1 . Show that there exists a sequence of polynomials { p n } such that pn + f uniformly on I and p l ( z ) 5 p 2 ( z ) 5 . ~ )= 1. . {k.324 Since f 4 { k : . ( IndianaPurdzle) Solution.1] iff f(0) = 0. = $00 then nco ncc lim sin(khz) = lim sin(khz . 4505 Let f E C ( I ) . + f .2 1 I k. For any n. Suppose that there is a sequence of odd polynomials pn with p .
Take a sequence of polynomials p .1 1 1 1 r+’ is a contracting mapping on C[O. Set f ( z ) = f(z). We have uniformly on [0.1]. 2 E [1.1] satisfying the conditions: { f(0)( z=) = &f Solution. ex + zf(x2). b) Show that there exists one and only one smooth function f on [0. b) By the Banach’s Theorem. + f uniformly on [1. Then Pn is an odd polynomial for any n. (Stanford) 1. ( z ) with p .325 Conversely. 12 f(t)dt. Set 1 pn(x) = [pn(x) 2 ~n(x)]. 4507 a) Show that the mapping I : C[O.. Then f is a continuous function on [1. I]. suppose that f(0) = 0. with the supremum norm. f(z) = es + f J.1].l]). there is only one function f E C[O. 11. i.1 + C[O. (1) .0]. a) For any f and g in C([O.e.1 such 1 that I(f) = f.
g = f. by the uniqueness of f..3. .326 By (1) f is smooth and satisfies the conditions: f(0) = 1. then i.e. E L1 such that (i) llKnlll 5 M < 00. take Nz such that . .)l < 4M & E R any IyI < 6.2. lKnl Show K .. . by (ii) there is an N1 such that If n > N1. Thus Take 6 > 0 such that lf(z .. For the above 6. K .61 + O as n + 00 for all 6 > 0. 4508 Given f : R + R bounded and uniformly continuous and {K . ..+ o o . n = 1. (UC. . Take M1 > 0 such that If(z)l < M I for all z E R. . (iii) J12:121. . I ( g ) = g . n = 1 . (ii) ~ K . }.+ l a s n .Y) holds for all 3:  f(. 3 . * f + f uniformly. 2 . Irvine) Solution. For any E > 0. If g is another smooth function satisfying (2).
(b) Show: If I is an open interval contained in S.+ M .E < 1.m ( z ) 2 1).oo) m ( z )= liminf f(y). there exists E > 0 such that f(z0)  m(z0) < 1 . Let S = { z I f(z) .&z ..327 & < . (c) Show that S is nowhere dense.(2) that < holds for all z E R if n & & +=& 2 2 >N 4509 Let f : R + (03. It follows from (l). .N2). co) be upper semicontinuous and define m : R [oo. (UC. Irvine) Solution. & 4 4M 2 Set N = max(N1. by YX + (a) Show S is closed. then m ( z ) = aon I.. (a) For any zo E S". .
m(5)< 1.6). 4510 Prove that any topological metric space is homeomorphic to a bounded metric space. From (b). Denote i + i.. By the definition of rn(z). () (c) Suppose that S is not nowhere dense. By induction. 2 E S".e. .f(.l .6) such that f(5)< rn(z0) Therefore f: . we obtain a sequence of open intervals { I n } such that 7. Take an open interval I1 such that 7 c A l . Since S is closed. Thus f) ( . 6) c I . Therefore m ( z )= liminf f ( y ) Y+Z 2 rn(zo).S).d ) to ( X . (b) Suppose that there is zo E I such that rn(z0) > m. Z ) .for 2 E O(z0.there must be F E O(z0. ? which is a contradiction. As in (a). .& & holds for z E O(z0.m(. we can find 6 > 0 such that m ( z ) > rn(z0).o) + . i.e.< 1 2 2 & & +& = 1 if z E O(zo.) < 1). (Stanford) Solution.. i.6). A1 is a nonempty open set.m(2)< f(20) + . . 5 6 S which contradicts % E I c S. we can take an open interval I2 satisfying 7 2 c I1 and f(z) < 2 for z E 7 2 .328 Take 6 > 0 such that for 2 E O(z0. c I n . In the 1 same way. there is an open interval I c S.d ) is a metric space.. Define another metric d o n X by It is easy to show that the identity mapping of X is a homeomorphism of ( X . f(z)< n for z E 7 Obviously f ( z ) = oo for 2 E n . Suppose that ( X . . m ( z )= m for z E I . Al = {Z E I. So S" is open.
too. ) = +00. + t B be a continuous convex 00. is a periodic map of order 3. z) E C x 2R The isometry A : M +M I 1 . ~ 1+ z2 = 13. l ) and (0. (i) Since A' = A 2 . z) E M . A o A o A = Id and d ( A z .. periodic of order 3.329 4511 Let M be a metric space with distance function d. i.e. There is a sequence {zn} of H such that nco lim f(zn)= a . 4512 Let H be a Hilbert space and let f : H function such that f ( z n ) + oa whenever llznll minimum. Let (Y = inf f ( H ) . Stony Brook) Solution. (i) Show that A is a continuous bijective isometry.Ay) 5 d ( z .z) = (e. z). A' is distance nonincreasing. (ii) Give a n example of a complete metric space M and an isometry A on M . If SUP llznll = $00. then a = lim f ( z n . defined by A ( z . (Stanford) Solution.}F=~ that such krco l/znk (1 f +m. n there is a subsequence {Z~. which has exactly two fixed points.*'z. Therefore A is an isometry and therefore continuous. a distance nonincreasing periodic map of order 3. . suppose A : M + M is . 1). (ii) Let M be the 2dimensional sphere M = { ( z . y). ~ ( z . Prove that f attains a (SVNY. having exactly two fixed points ( 0 .
) (Stanford) Solution.e. (a) Indeed. Then n+m {fn} converges to f in measure and therefore by F.11 Therefore and for any y E cov{z. Recall that the adjoint A* is the unique bounded linear operater on H such that ( A z . (b) Show that AA* . (a) Show that IIA*ll = IIAll.y)I llYll51 114151 5 IlAll . 4513 fzI A is the subset of L1(BZ)consisting of all functions f satistying I ( ) 5 1 a. where IlAll is the norm of A. h Since X E n * cov{z. on lR. there is a subsequence { z n k } For any p > a there is an N E lT such that for any k > N . y ) = (z. (You may wish to use (a) prove this.Riesz Theorem there is a subsequence {fnk} converging to f almost everywhere. I k > N } . 1 k >N} W = coV{znk I k > N } 11. f ( y ) < .y E H . Prove that A is closed in the norm topology of L1(lR). exists in L1(BZ).e. It follows that I ( ) 5 fzI 1 a.A*A cannot be the identity on H. A*y) for all z. By the weak compactness of the closed converging weakly to a point z E H . So sup llzn[l < +m.B. one has f ( z ) 5 p.330 a contradiction. since llA*11 =  SUP IlYll51 llA*~ll = I l Y l l 5 1 114151 SUP SUP I(x7A*y)I SUP SUP I(Az. ball of H . If {fn} is a sequence of A such that f = lim f. So A is closed... ( StunfoTd) Solution. It follows that a is finite and f attains its mininum at z. 4514 Let A be a bounded linear operator on Hilbert space H . on aZ. f ( x n k ) < 0..
B c 0. m ( B ) > contains an interval of positive Lebesgue measure. we will show that Indeed equalities (1) follow from the following inequalities If for some bounded linear operator A on H . (Indiana) Solution. AA* . Assume without less of generality that A and B are compact sets. (b) First. 4515 Suppose that A . a contradiction. with m(A) > 0. .33 1 and a fortiori IlAll = 11A*\\. Show that IR are Lebesgue measurable. Since = m(A)m(B) > 0.A*A = I then = + It follows that I)A))2 JJAJ12 1.
B ) n A ) > 0) CA + B. and (3) p (2 each n E lN a realvalued continuous function f : X .B ) n A > 0) is nonempty and open.. 4516 Let X be a compact Hausdorff topological space and let p be a finite regular Bore1 measure on X . .e. Is it true that if f : X + IR is pmeasurable then there exists a sequence {fn}~'l of continuous realvalued functions such that lim f = f a. Then (1) i=l is for f = f on X.) f is continuous on X. n0 (Iowa) Solution.e. For any n E l there is a compact subset F.) < and f is continuous on F. Since z H m((zB)nA) is continuous. Then the conclusion follows from the following inclusion { Z I m((z . Let X ..) p(X\X. the set { z I m((z . . = 00 u Fi. N Yes. . + l R such that b]. . Then lim f = f a.332 there exists an zo E IRsuch that m((zoB)nA) > 0. [p]? Justify. nco < i. of X such that p(X\F..
Part V Complex Analysis .
.
vy exist at z = 0 with u. A counterexample is f(z) = sin&. if false. quote a relevant theorem or reason.335 SECTION 1 ANALYTIC AND HARMONIC FUNCTIONS 5101 TrueFalse. then f'(0) exists. z n has radius of convergence exactly R. then C n 3 a . and the zeros have no limit point in R. uy. (a) if f(z) = u iv is continuous at z = 0. then If(z)I 2 1 ( z E R). But f ( z ) is not bounded in R.v. and continuous on with If(iy)I E 1. 6. then either f E 0 or g 0. (e) True. give a counterexample or other justification. A counterexample is f(z) = e". then both f and g have a t most countably many zeros in R.. a n+m . g ( z ) E 0 in R. Rez > 0 } . which is analytic in R.and has zeros z = 1 n = 1. (b) False. (c) True. continuous on with If(iy)I 5 1 (co < y < +co). but f'(0) doesn't exist. (c) if f and g are analytic in R and f(z) . f satisfies Cauchy(a) False. Then f(z) .. a t z = 0. (b) if f(z) is analytic in R and has infinitely many zeros in R .2. then f E 0.. and the partials u. (IndianaPurdue) Solution. (d) False. y ) = Riemann equations a t z = 0. (f) sin f i is an entire function.g ( z ) is not identically zero in R. But f is not identically zero in R. (d) if f(z) is analytic in R = { z . . A counterexample is f ( z . + m. If the assertion is true. (e) if C a n z n has radius of convergence exactly R. Because lim G = it follows from 1. f is analytic in R = { z : IzI < l}. = vy and 1ly = 21. If neither of f and g is identically zero in R .
then F ( z ) is analytic in 1. z = 0 is a branch point of sin &. sin f i is not analytic at z = 0. If both f ( z ) and z f ( z ) are harmonic in a domain Q. Show that If(0)l 5 u2. Actually. (c) Determine all entire function f that < lf(z)I. (Stunf07d) Solution. What can you conclude if this holds with equality. 1+az hence . When 17iz IzI = 1. (b) Suppose that f is analytic with If(z)I < 1 in IzI < 1 and that f(iu) = 0 where u is a complex number with 0 < la1 < 1.( < 1). (a) It is well known that the Laplacian can be written as [). 5102 (a) Let f(z) be a complexvalued function of a complex variable.336 (f) False. prove that f is analytic there.(’fI and that which implies that f ( z ) is analytic in R.
then for any closed curve c c G.+ . ay2 We also have ap . (a) Show that . we obtain If(0)l I bI2. y) and Q ( x .=aQ  ay ax axay azay = 0. the integral 2 i g . I $$I + % 5103 Let G be a region in a‘ and suppose u : G t lR is a harmonic function. (b) Show that u has a harmonic conjugate on G if and only if has a primitive (antiderivative) on G.)[ 5 1 for IzI < 1. elel which is equivalent to When it holds with equality. Hence f (z) = c c’e‘”. (a) Let 2 i e Because u is a harmonic function. 1+zz (c) From lf’(z)l < lf(z)I. where c and c’ are constants and JcI< 1. hence P+iQ=i is analytic on G.i% is an analytic function on G. we have F ( z ) f ( z ) = a. Take z = 0. Integrating on both sides.y ) satisfy the CauchyRiemann equations. (Indiana) Solution. (b) If has a primtive. we obtain logf(z) = cz d. we have ap ax aQ ay  aZu + ax2 ~ = 0. which implies that fIIl is also an entire function.’ zu 1Tiz z + a . we know that f has no zero in a‘. f(2) I I < 1.. au ax au ay aZU azu So P ( x . It follows from < 1 that = c.337 which implies that IF(.
we have = dx dU dU + dy. and the integral is taken along any curve connecting zo and z in G . z E G. ay ax For any closed curve c C = ld(u Hence + iv) = 0. y) is a harmonic conjugate of u(x.338 It follows that Kdx dU du +z d y =0 holds for any closed curve c c G. av au ay ax’ we know that v(x. + (IndianaPurdue) Solution. Because av . Prove that u iv must be analytic on 0.__ ax dy’ .i aY has a primitive sz: k 5104 Suppose that u and v are real valued harmonic functions on a domain R such that u and v satisfy the CauchyRiemann equations on a subset S of R which has a limit point in R. 2 . The reason lies on the fact that the real and imaginary parts of fi and f2 satisfy the CauchyRiemann equations respectively (see 5103 (a)). if u has a harmonic conjugate v on G.ik and f2 = aY aY are analytic functions on a.au _ . y) on G. Hence we can define a singlevalued function where zo. fi = . On the contrary.iau Because u and v are harmonic functions. % . . then dv = dx av + dy aV ax aY G .
(Indiana) Solution.1]. au av _. Suppose that f(z f(z + 1). we know that f1 = if2 holds for all z E Q. by Cauchy integral theorem. ax dy’ du dy av __ ax when z E S c Q. Hence fi = & . and let f be holomorphic in a neighborhood of Q .f ( z ) is real and + i) . which implies that u + iw is analytic on Q.1] x [0. It follows from fi = i f 2 for z E R that for z E Q.i $ ) bu & when z E S. we have As .339 By the assumption of the problem.f ( z ) is real and 2 0 for z E [O. Because the subset S has a limit point in Q.i] 2 0 for z E [0. 5105 Let Q = [0.1] c a‘ be the unit square. Because f is holomorphic on the closed unit square Q. by the uniqueness theorem of analytic functions. Show that f is constant.i a u = if 2 = i ( g .
we obtain that f ( z + i ) = f ( z ) for 0 5 z 5 1and f ( l + y i ) = f ( y i ) for 0 5 y _< 1. by comparing the real and imaginary parts in the above identity. Hence we obtain Re(f(z)dz) = 0 on dS which implies ReF(z) = 0 when z E dS. 5106 Let f be continuous on the closure 3 of the unit square s = { z = z + i y E @ ’ : 0 < 2 < l . we obtain IG(z)I 1 for z E S.f ( Y 4 2 0 for 0 I y 5 1. so l/G(z) is also analytic in S and 11/G(z)l = 1 when z E dS. Apply the maximum modulus principle to both G ( z ) and l/G(z). . and let f be analytic on S. Let G ( z ) = e F ( * ) .and if If = 0 on S n ({z = 0) U ( 2 = l}). and on sn ({z = 0) u { z = 1)) we have v = 0 and dz = 0. Because G ( z )has no zeros in S. + f(z)dz = (udz . For z E we choose the integral path on dS and consider the differential form f(z)dz in the integral. It follows from G(z) = eF(’) that F ( z ) is also a constant.Then G(z) is analytic in S and IG(z)I = 1 when z E 85’. Then F ( z ) is analytic in S and continuous on S.vdy) + i(wdz + u d y ) . Hence f ( z ) = F’(z) G 0. Define F ( z ) = f(z)dz.prove that f = 0 everywhere on S. o < y < l}. then si as. which is holomorphic in a and satisfies ‘ lf(z)I I %${If (. Hence f ( z ) can be analytically extended to a doubleperiodic function by f(z) = f ( z 5 1 and + 1) = f ( z + i ) . where the integral is taken along any curve in which has endpoints 0 and z. which implies that G ( z ) is a constant of modulus 1. If Rf = 0 on sn ({y = O } U {y = l)). Let f = u i w .)I1 < This shows that f ( z ) must be a constant. (Indiana) Solution. On 3 n ({y = 0) U {y = 1)) we have u = 0 and dy = 0.340 for 0 5 z f(1+ Yi) .
1) = lim =  2 4 1 (z4 . hence f ( z ) is holomorphic on an open set containing { z : IzI 5 I}.lz12  lz1 1 0. we have A.4)’ 24 23 23 lim 21 4z3 + 3z2 + 2z + 1 1 1 10’ we know that z = 1 is a simple pole of a 4 + a s + 21 + 2 . (Iowa) Solution. and the equalities hold if and only if z = 1.4 in { z : I z J5 l}. &. .+ +1 + z . we obtain that f ( z ) has no singular point in { z : IzI 5 l}.341 5107 (a) Find the constant c such that the function 1 f(z)= z4+z3+22+24  C z1 is holomorphic in a neighborhood of z = 1. z 2 . lz4 + z3 + + z2 . A “partial fractions” expression of $ has the form: . (b) When IzI 5 1. . + 5108 Let P ( z ) be a polynomial of degree d with simple roots z1. (b) Show that the function f is holomorphic on an open set containing the closed disk { z : IzI 5 1). . which shows that z = 1 is the only zero of z 4 + z 3 z2 + z . zd. . (a) As 241 lim(z .41 2 4 .4 with residue equal t o z Hence when c = f ( z ) is holomorphic in a neighborhood of z = 1.lz4 + 2 + + z~ 2 4 z2 144  lz1 3 .4 22 1 + + 22 + z . By (a).
P(. hence (c) Denote the Taylor expansion of P ( z ) at z = z1(= z2) by P(2)= c 00 a n ( z .342 (a) Give a direct formula for cn in terms of P. p. f(z) is I 4 0 0 identically equal to zero.) (b) Let Then f ( z ) is analytic o n c and lim f ( z ) = 0. By Liouville’s theorem.) Solution. (4 c n = lim z+.zm z .a t z = z. () (c) Give a formula similar to (*) that works when z1 = z2 but all other roots are simple. (b) Show that 2 really has a representation of the form (*).Z1)n.) .. (Courant Inst.zn  1 P(z) P’(zn)’ which is the residue of 2. Then the Laurent expansion of 2 a t z = z1 is P ( Z ) Hence 1 the form: has P(.
then B1 is real and  Az2 + Bz + z ( z . K2. Kn be pairwise disjoint disks in C. Assume K1 = { z . and j=1 U Kj. fn such that (b) f ( z ) = C fj(z) for z €C\ U Kj.343 5109 Suppose f is meromorphic in a neighborhood of D (D = { ( z (< 1)) whose only pole is a simple one a t z = a E D. . It is obvious that g ( z ) is analytic on D and g ( a D ) C lR. such that C1 = { Z.. Suppose g ( z ) B1. j=1 n (Indiana) Solution.ZB1. ( a . If f ( a D ) C IR.Tiz) * Solution.) = ‘(z . n n f1. .a ) ( l . B = (ZAl + a z l ) + B1(1+ (at2)E lR. show that there is a complex constant A and a real constant B such that Az2 + Bz + z (Indiana) f(.Aizz 1z z. g ( z ) can be extended to an analytic function on the Riemann sphere hence g ( z ) must be a constant. .z l ( 5 rl}.a ) ( l . Choose c1 > 0 sufficiently small. 5110 Let K1. Assume that the residue o f f at z = a is Al. fz. ..~1 < + ~ 1 C ) C\ UKj j= 1 n * ..Tia)’ where A = 2 1 . Define g(z) = f ( z )   Ai za . By the reflection principle. TI < )Z . and let f be an analytic function i n C \ (a) f j Show that there exist functions j=1 is analytic in C\Kj..
~ (2) 2 ) f2(z) ~ . f ( z ) .fl(z) is analytic in @\ Assume K 2 u Kj.f ( z ) has the Laurent expansion f(z) = c +m up(z .344 In C1. j=1 n 5111 Recall that a divisor Df of a rational function f(z) on @ is a set { p E of zeros and poles p of f(z) (including the point co). Set Then we have n j=1 where fj(z) is analytic in@\Kj. Iz .221 5 72).f n .I)? k=m Set fl(Z) = c 0 k=oo u p ( z . Repeat the above procedure n . and the above identity holds for z E @\ U Kj.* * * .fl(z)fZ(Z) .l ( z ) . Choose ~2 n = { z . such Laurent expansion 0 Set f2(z) = k=m uk ( z . @ u {m}}. Because f ( z ) . we get a function f ( z ) .fi(z) has an analytic continuation j=2 to K1.z1)5 fl(z) is analytic in @\K1.1 times. > 0 sufficiently small. which is analytic in @\Kn. is analytic in@\KZ. consisting . Because f(z) K2. fl(z)f2(z) has an analytic continuation to in @\ f(z)fl(z)f2(z) is analytic ij j=3 ~ j .
m) be all the zeros and poles of g ( z ) m with multiplicities mqjrespectively. Let f and g be two rational functions with disjoint divisors. . It should be noted that p. By the property of rational functions.345 counted with their multiplicities np E 2. Prove that (SUNY. i=l n Similarly. Stony Brook) Solution.2. Let pi (i = 1. . 2 .p i ) n p i and m Then n n n m and m m m n . . then we have j=1 mqj= 0. . First we assume that the point and g can be represented by 00 is not a zero or a pole o f f or g. C npi= 0. . we have  1 . n) be all the zeros and poles of f(z) with multiplicities npirespectively. let q j ( j = 1 . is a zero of f when npi> 0 and a pole of f when npi < 0. then f n f ( z ) = A JJ(z .
(b) Find the radius of convergence R of the Taylor series T ( z ) of f ( z ) at 4 3i. = without loss of generality.) j=1 = g(m) = B . Then n1 00 f(z) = A n ( z i=l and g(z) m =B n(~ m  4j)"'j. Hence n. we may assume p . 3i. 5112 Let f ( z ) be the "branch" of logz defined off the negative real axis so that f( 1) = 0. j=1 Since C mpj = 0. simplifying the (a) Find the Taylor polynomial of f of degree 2 a t 4 coefficients.346 In case the point 00 is a zero or a pole of f or g. we may assume that g(p.1 i=l n1 m and i = l j=1 which completes the proof of the problem. + + .
1 by a. co = log 5 c1 where 3 + i(7r .arcsin ). (Minnesota) Solution. the radius of convergence R of T ( z ) is equal to the distance between z = 4 3i and z = 0.. and has no other singular point. or a general impossibility? Explain and/or give examples.. 5 4 =  + 3i 25 ' 25 24i 1250 ' and c2 = + (b) Denote the Taylor series of f(z) at 4 3i by T ( z ) .3i)] z+43i log( 4 + 3i) + log 1 + [3 z4+ 4+3ii 3 log 5 + i(7r . When z E Cl = { z : Iz 4 .5. (a) When z is in the neighborhood of zo = 4 f ( z ) = logz + 3i. If there are no such points.Imz < 0 } . is this something special to this example. Because log z has only z = 0 and z = 00 as its branch points. T ( z ) # f(z). while f(z) in R is the continuation of log z a t 4 3i in the slit plane @\(m.3i) + cg(z + 4 . Hence + + + + + + .347 (c) Identify on a picture any points z where T ( z ) converges but T ( z ) # f(z).3il < 5}. Hence R = 5. It is because T ( z ) in R is the continuation of logz at 4 3i in the disk { z : Iz 4 .3 4 2 . we have = l0g[(4 + 3i) + (Z + 4 .4 + 3 i I ( ) 2 +. and describe the relationship between f and T at such points.3i( < 5.* Hence the Taylor polynomial of co f of degree 2 at 4 + 3i is + c1(z + 4 . (c) Denote the shaded domain shown in Fig.01.arcsin ) + 5 4+3i 1 z+43i 5 .
Express g in terms of f .$(.5. f. Then g(z) = Ig(z)leia%dz) 1 = Arargz 3 = ~() I.e. i. T(z) = f ( z ) + 27ri. Fig.41 < 4) so that f ( z ) = z i ( z + 1)i in A and f ( z ) is positive for 0 < 2 < 8. We have Arargf(z) 1 1 1 + Ararg(z + 1) = (27~)+ 2( .)..5. An analytic function g in A is obtained from f by analytic continuation along the path starting and ending at z = 4 (see Fig.l(argf(z)+Arargf(zr)). and denote the change of $ ( z ) when z goes along r from the start point to the end point by A.5. (Indiana) Fig.2 by I?.348 the difference is 2ai. Denote the closed path in Fig.2).2 Solution.5.2 ~ ) 3 2 Hence .1 5113 Let f be the analytic function defined in the disk A = { z : Iz .
l ) ~ ( e = ~ enn). 2 . When z is in the small neighborhood of z = 0. . 2 .349 5114 Define eJ. sin f i which is still f(z). Hence z = 0 and z = 00 are no longer the branch points of f(z). : (b) We have which shows that z = n2r2are simple poles of f ( z ) with residues 2 n ~ ( . Let r = { z : IzI = T } . . (c) Evaluate f(z)dz.eJ. f(z) can be represented by 03  2 n =c zn O c 00 > O n z n n=O (%+I)! which implies that z = 0 is a removable singular point of f ( z ) .=25 (Indiana) Solution. .e& sin(&)  e f i . and when z goes along I’ once in the counterclockwise sense.6 . . Hence f ( z ) is singlevalued and analytic inC\{z = n2r2 n = 1 . (a) It is known that z = 0 and z = 00 are the branch points of function &. 4x. f i is changed to fi. } .) are poles of f ( z ) . efi f(z) = (a) Where is sin& * f singlevalued and analytic? (b) Classify the singularities o f f . . It is obvious that z = n2R2 (n = 1 . . . while f(z) is changed to e .
arg(l 2 +z) 0 27r  1 arg(l 2 (1 (1  z)] =0 1 1 Ar[arg(l + z ) . For (b) g ( z ) = can we choose singlevalued branches which are holomorphic in fl. and hence is a nonisolated singular point of f(z). Hence f(z)dz = 2 ~ i R e s ( fr 2 ) . It is ' known that f and g can be represented by and  g(z)  4 1  . y = 0) deleted. (c) f ( z ) has only one pole z = 7r2 in the disk { z : IzI < 25). L < 2 5 5115 Let fl be the plane with the segment (1 which of the multivalued functions (a) f(z) = 1 5 z 5 1 . and denote by A.z ) ] = 2 2 5 z 5 1 . y = 0) not inside r 5 z 5 1 . Let I? be an arbitrary simple closed curve in a. it is the limit point of the poles of f(z). y = 0) inside I? .arg(l .350 As to z = 00.{+log(l+z)+log(ll)} Because and 1 Ar[argz . Why? (IndianaPurdue) Solution.r#J(z)the change of r#J(z)when z goes continuously along I counterclockwise once. Which of these branches are (is) the derivative of a singlevalued holomorphic function in 0.
y = 0) is not inside I?. it is obvious that f (z)dz = 0 and Sr g(z)dz = 0. give a precise meaning to the integral J'(1og ~ ) ~ and find all possible values which can be assigned to it.. Hence both f ( z ) and g ( z ) have singlevalued branches which are holomorphic in Q. 1. Because . f(z) g(2) = *i(l z  )5 22 1 1 = f i (1 + a2 + 7 + * * 7 a4 . (b) Let y denote the unit circle.. The singlevalued branch of logz is defined by loge = 1.*1. i = *(1 )T 22 1 1 1 b3 b5 =fi . Because e E (e'. sr ( 1 szt + 5116 (a) Let D c C be the complement of the simply connected closed set (eetie 1 6 E R}U (0).) .. where the integral is taken along any curve connecting zo and z in 0. z 23 zi It follows that f(z)dz = 0 and lrg(z)dz = f 2 a . (a) The set {ee+ie I 6 E IR} U (0) is a spiral which intersects the positive real axis at {e2nn : n = 0.y = 0) is inside I?. Hence loge15 = loge A r l o g z . But neither of the branches of g is the derivative of a singlevalued holomorphic function in a. (b) Define the lift mapping by w = logz which lifts the unit circle y onetoone onto a segment with length 2ir on the imaginary axis of wplane. It follows that A r l o g z = Aplog IzI + iArargz. and each of f and g has two singlevalued branches. + + + .+ . Hence we obtain that both of the singlevalued branches of f are the derivatives of singlevalued holomorphic functions in Q. Find log el5. e2=). we consider the integrals 1 f(z)dz and 1 g(z)dz. d z (HUT7NZrd) Solution. If the segment (1 5 z 5 1.1 = 14. By lifting the integration to a n appropriate covering space. If the segment (1 5 z 5 . a ) . oriented counterclockwise. and the primitives are f(z)dz c.+ T +. and A r l o g IzI = 15 . where I is a continuous ' curve connecting z = e and z = el5 in D and A r log z is the change of log z when z goes continuously along r from z = e to z = el5. Hence log el5 = 1+ (14 4ai) = 15 4ai. Let log be a branch of the logarithm on D such that loge = 1. Justify your answer. e6=).351 we have A r f ( z ) = 0 and Apg(z) = 0. we know that when I connects e and el5 in D. el5 E (e4A. In order t o know which of f and g has a singlevalued primitive in R.*2. . we consider the Laurent expansion of f and g about z = 03: s . ' Arargz must be 4a.
we have reason to assume that the function f (with real variable t ) has continuous derivatives f’(t) and f”(t). z E C\O. where (Y. + 5117 Find the most general harmonic function of the form f(lzI). Since log IzI has no singlevalued harmonic conjugate inC\O.i(t+ 27r)). Because f ( I . Hence the most general harmonic function of the form f(1). + . has a singlevalued harmonic conjugate in C\O.1) have a single valued harmonic conjugate? (Indiana) Solution. Hence we have J(W2dz 7 = lt i(t+2*) w2ewdw= (w2ew)lit i(t+2x)  21t i(t+27r) weWdw ewdw i(t = eit(47rt . it must be a constant.Pare two real constants. we know that when f(1). Which of these f(I. Note that the Laplacian A=+=4ax2 ay2 a2 a2 a2 azaz7 and we obtian where t = IzI. 1 ) is harmonic. where t can be any real number.352 both the starting point of y and the singlevalued branch of log z on y can be arbitrarily chosen. which implies that the set of values being assigned to the integral J (log ~ ) ~ d z 7 is a spiral (443 i)ezs : s E R}. the segment on wplane can be denoted by [it.(2wew)lit+ Z U ) + 2 1 t i(t+la) = 47~(t+ ?r + i)eit = 47r(t + 7~ + i ) e i ( t + 7 r ) . This differential equation is easy to solve. in C\O is Q log IzI 0.47r2) . and the solution is f ( t ) = Q log t + 0.
.. Denote the interior domain of the pentagram shown in Fig. By the symmetry of 0 on the rest of the boundary by domain D. . Fig. we have It follows from in . .3 Then denote the harmonic function on D with boundary values 1 on l k and u k ( z ) . Let u be the harmonic function in the interior of the pentagram which has boundary values 1 on the two segments shown and 0 on the rest of the boundary. 2 . and the ten segments of the boundary by I l l I z .5. . 10. . k = 1 . . .5.3 by D . (Stanford) Solution.353 5118 Consider the regular pentagram centered at the origin in the complex plane. 110. put in order of counterclockwise. What is the value of u at the origin? Justify your claim.
does this implies that h is harmonic on G? (Iowa) Solution.l] holomorphic. and h ( z ) is continuous o n C . Suppose that u and v are harmonic in D = { z : IzI < 1) and continuously differentiable on D u y. because z = 0 and z = 1 are branch points of Remark. 11. A counterexample is h ( z ) = Red. If the boundary values satisfy u = v on y and the radial derivatives satisfy = on y. 5120 Let 7 be an arc of the unit circle. h l . then h must be holomorphic on G. Since d m is analytic in C\[O. q p l ] is harmonic.' r d . prove that u = v in D. h ( z ) is harmonic there. Let u* be a conjugate harmonic function of u in D and v* be a conjugate harmonic function of v in D. The answer is No. dmlz=z lim z=x+yi+z &zij=&cT)i. and h : G + lR is continuous. Y<O Hence h ( z ) = 0 when z = 2. If the problem is changed to h : G + C is continuous and hl~\[o. is d m .1] c G.354 5119 Suppose G is a region in C. When 0 5 2 5 1. Y>O lim z=x+yi+x d m = 4i. But h ( z ) is not harmonic on&. where the singlevalued branch of d m is chosen by = 4. 0 5 2 5 1. (Indiana) Solution. [0. We know that a variation of CauchyRiemann equations for f = u + i * and g = v + iv* are u du r=dr du* ae  au dU* ae .
such that f = g on y.) Solution. Without loss of generality. which implies u = v in D.355 and It follows from the continuous differentiability of u and v on D U y that u* and v* can be continuously extended to D U y and then are also continuously differentiable on D U y.g. Hence s i U ( z ) = u(i) z1 zS1 = arg() T 2 z+l z1 . Then for z E y. F can be analytically extended to an analytic function on D u y U D*. we obtain F 0 on D U y U D*. = lz :dB = v*(z). Let F = f . The boundary correspondence is that the negative real axis {w : oo < w < 0) corresponds to the arc I’l = { z = eie : 0 < 8 < T ) and the positive real axis {w : 0 < w < +a} corresponds to the arc rz = { z = eie : T < d < 2 ~ } . Since F = 0 on y. Hence we obtain two functions f = u iu* and g = v iv* which are analytic in D and continuous on D Uy. Give the correct determination of any multiplevalued functions appearing in your answer. Then by the reflection principle. It is well known that u(w) = targw . where D* = {z : IzI > 1). It is easy to know that w = is a conformal mapping of the unit disc D = { z : IzI < 1) onto the upper half plane H = {w : Imw > 0). (Courant Insf. we may assume that u * ( z o ) = v*(zo) = 0.and for z E y denote the subarc of y from zo to z by y z . + + 5121 Use conformal mapping to find a harmonic function U ( z ) defined on the unit disc { Iz1 < 1) such that r+l lim U(reie)= +1 f o r O < d < a 1 for 7r < 8 < 27r. Let zo be a fixed point on y.1 is a harmonic function in H and assume $1 on the negative real axis and 1 on the positive real axis. .2.
log 2.dy and A = h z l = r * d u .dz u.{2arg(C . where A is a real number not necessarily zero.. This problem can be solved directly from the Poisson formula as follows: 2) 1 = A.argc) . Let * d u = u. 2R A .1 A  2 z+l arg2..z ) .where the singlevalued branch of arg( is defined by arg( %)lr=o = a. Suppose u ( z ) is a continuous function on (0 < IzI 5 1) which is harmonic on (0 < IzI < 1) and identically zero on { I z I = 1). (Minnesota) Solution. then w(z) is the conjugate harmonic function of u ( z ) .(Z)) . is a harmonic function in D = { z : I1 < 1) with the boundary values $1 on t and 1 on I'2. Define + f ( z ) = (u(2) + i. Denote w(z) = Jz * d u . but may be not singlevalued. Remark. R 21 5122 Determine all continuous functions on { z E C : 0 < 1z1 5 l} which are harmonic on { z : 0 < IzI < 1) and which are identically 0 on { z EC : IzI = 1).
= 1).zn A + log 21r Iz].too < Iz1 < 1) and Ref(z) is n=m 5  u. Jensen's formula states that Prove this. then . . From the above discussion. Define g(z) = n=m C 00 bnzn. hence f ( z ) = g ( z ) ia. and If(0)l = 1. Then f ( z )  g(z) = n=O c. and IF(z)I = 1 on (1. Consider P ( z ) = e f ( ' ) . counted with multiplicity. .( < 1) and Re(f(z) .g ( ' ) which is analytic and does not assume zero in { IzI < l}. by the maximum and minimum modulus principles. = bn for n = 0. 2 .2. Then g ( z ) is an analytic function on (0 < IzI < +m}. where b. .357 then f ( z ) is a singlevalued analytic function on (0 identically zero on { IzI = 1). we have F ( z ) i~ e i a . .1 = l}. 2 . = a _ .1. . . .zn is an analytic function in 1. show that . satisfying 6.g ( z ) ) is identically zero on (1. we finally obtain + +cc u ( z ) = Re n=m b. 2. let n(t)be the number of (j= 1 . 5123 (a) Let f ( z ) be a holomorphic function in the disc IzI 5 T whose zeros in this disc are given by a l l a2.. 5 1. + . When IzI = 1. Suppose further that luj I < T for all j = I. a .n ) such that ( u j ( 5 t . . for n = 1 . . it follows from Rebo = 0 and 1 n=cc 03 03 n=l n=l 00 n=l that Reg(z) = 0. uj . and b.zn be the Laurent expansion of f ( z ) on (0 ncc < IzI < lim = 0 and lim 0 .. Using Jensen's formula. n. (b) With the hypotheses and notations of (a). Let f ( z ) = l}. = bn and  n+cc lim = 0.
we obtain n ( r )5 M(R)/log R . (a) Let then F ( z ) is holomorphic and has no zero in the disc ( 1 . By the mean value theorem of harmonic functions. n(t)we have (b) It is obvious that log 2 = la3 I h:j. ~ 1 5 r } . we have & Lzx 0<8<2R log Jf(Re2')ldB= R Denote max log If(Re")J by M ( R ) .358 (c) For r < R deduce an estimate on n ( r ) in terms of max log If(Reie)l. By the definition of the function which shows that the identity holds. .)/ is harmonic in 1.( 5 r } . 0<852x (d) What can be said about the zeros of bounded holomorphic functions in the unit disc? (Hamard) Solution. which implies that log IF(. (c) Apply the identity in (b). $.
~ 1 = T } .359 (d) Let f ( z ) be a bounded holomorphic function in { z : IzI < 1)..and hence Let T + I. we can choose T such that T > lun[. Suppose z = 0 is a zero of f ( z ) of multiplicity m 2 0 with = a. we Since f(z) is bounded. Apply Jensen's formula in (a) to F(z) = with 0 < T < 1 such that there is no zero of f on ( 1 . and let the other zeros be ordered by 0 < lull 5 la21 I . We know that f(z) can have countably many zeros.. we obtain n which implies that the series C (1 . j=1 00 . Obviously Ian1 + 1. we assume For any n.luj I) is convergent.
< l} ont ( S U N Y .{[0.O 5 Rew < 1). 1(.SECTION 2 GEOMETRY OF ANALYTIC FUNCTIONS 5201 Find a onetoone holomorphic map from the unit disk slit disk {IwI < 1) . (Illinois) Solution. We construct the map by the following steps: 2 1 = 41(z) = 2. 5202 (a) Find a function f that conformally maps the region { z : largzl < 1) onetoone onto the region {w : 1wI < l}. 43(22) = 22” : {z2 : 1221 < 1 and Imz2 > 0} + {w : Iw( < l}\{w : Imw = 0. z + 1 : { z : IzI < 1) z1 f (21 : Imzl < O}. if No. (b) Is it possible to require that f(1) = 0 and f(2) = $? If yes.1)}. Show that the function you have found satisfies the required conditions. Then w = 43 0 4 2 o 41(z) = f ( z ) is a o n e .b o n e holomorphic map from the unit disk {IzI < l} onto the slit disk {IwI < 1}\{[0. {zi : Imzl w 1 < 0) t {z2 : 1.~21< 1 and Imzz > O}. (a) = f l ( z ) = z: = . give an explicit map. Hence w < 5 . and w = f 2 ( ( ) = is a conformal map of {< : Re< > 0) onto { w : II < l}. explain why not. 5 l o g r (log 1 = 0) is a conformal map of { z : largzl < l} onto {< : Re( > 0}.1)). Stony B Solution.
b] (0 < < 1.bc > 0.y) : x 2 y2 < 1 and y < 0).i a a is z=e 4 2 + iy : 2 > required conformal map. Then w = 4 2 o &(z) = d n + z 2 . [1. ~ ): 2 (1) Find one 11 onto conformal map f that sends the open quadrant > 0 and y > 0) onto the open lower half disc {(z. d where 1 = . Let w = 4 2 ( ( ) = (=a <+ dm+(. 2r+1 . c . b. = . ( 6 1 which implies that If(._ (b) Suppose W = f ( z ) is an arbitrary conformal map of {z : largzl < 1) onto {W : 11< 1) with f(1) = 0. Then w = F(6)= foY'(G) is a conformal map 6 of { G : 1 < 1) onto {w : II < 1) with F ( 0 ) = 0.)( = lY(z)I for every z E { z : (argzl < 1). Hence f can be written as 4 2 o II. (2) I f f is an arbitrary conformal map satisfying the condition of (l). (1) Let C = 41(z) = z 2 .0 < b < 1) conformally on the full disk {IwI < 1) by means of a function . s. d E IR.i is a conformal a.then 4T1 o f o 4F1(C) is a conformal map of the upper half plane onto itself. 5204 a Map the disk { Izl < 1) with slits along the segments [a. where d 1 =. which can be represented by $([) = where a. It is a conformal map of { z = 0 and y > 0) onto {C = iv : 77 > 0).361 is a conformal map of { z : largzl < 1) onto {w : IwI < 1) with f(1) = 0 and f(2) = 2q1. we have both [ F ( G ) 5 1 and I F ( w ) l 5 Iw(.' ( w ) 6 1 w is a conformal map of {w : II < 1) onto {W : 11 < l} with F ( 0 ) = 0. and W = F(w)= f o f .11. Since    we cannot require that f(2) =  i. (2) Find all such f . ( Toronto) + Solution. It map of {C = < + iq : 7 > 0) onto {w = u + iw : u2 + v2 < 1 and w < 0). . 5203 ( ( 2 . By w 6 Schwarz's lemma. ad . 0 & ( z ) .
> 0) {w : ll < 1). Compute f'(0) and the lengths of the arcs (Haruard) Solution. 42(.rs+i that 44(i) = 0 and d k ( i ) = f. ( i ) z 1 = 4 1 ( z ) = z + i : { I z I < 1}\{[a. f'(0) corresponding to the slits.. (ii) z2 = 4 2 ( 4 = the point correspondences z i + ( b + ") Zl+(aP+): @\{[6 .i... We construct the conformal mapping by the following steps.b]}+~\{[b~. and 4 2 (2) and it is easy to know that ( 4 2 0 41)'(0) = (a 1 + 1 + b + ) b < 0.1]U[1. It is obvious w . > 0). > 0. +m) + (z3 : Imz3 and For the convenience of computation. 4 l ( a ) = a +. + ) It has a .a + i]} C\[O.. 41(b) = b . It has the point (iii) z3 = 43(~2) = correspondences fi : C\[O.) = 1.362 w = f(z) with f(0) = 0.a+~]}. 1 + + #1(1)= 2 and &(1) = 2. let A= L4 &+&' i. B= 24' 6 t &+4 We also know that +i(l)= (iv) w = 44(z3) = ~3: { z 3 : Imz3 . It has the point correspondences #q(O) = 00.
The lengths of the two arcs are and containing 2 I1 Ai A+i = arg. f'(0) > 0.arg Bi 1 $4 = 4arctg. b] and [a. We are going to find the map f by the following steps: . (Stanford) Solution.E } and let fir be Let 0 < E < a.. we know that f maps the unit disk with slits [1. > I}.= 4arctg B+z B A+& 5205 denote the arc {eit : E 5 t 5 27r .1 conformally 1 onto the unit disk with f(0) = 0 and What correspond to the slits are the arc with endpoints containing point z = 1 and the arc with endpoints and point z = 1. let the complement of yE in the Riemann sphere.arg= 4arctgA = 4arctg A+z Ai T+& B+i B2 5 4 and I2 = xg . f(0) = 0. I f f is the conformal map of the unit disk onto a. From the above discussion. describe the part of the unit disc that f maps onto 1(.363 Now we define w = f(z) = 4 4 o 4 3 o 4 2 o & ( z ) .
and the symmetric domain of D1 with respect t o I.4). then w = F ( C )= f o g'(C) is a conformal map of {C : 1" < r 1 ) onto {w : Iwl < r } + c . and if a = 1 i.. 5206 Suppose that w = f ( z ) maps a simply conncted region G onetoone and conformally onto a circular disk DT with center w = 0. is D 2= (1 < 11 l}\n1. After considering the boundary correspondence. Hence the conformal map f in the problem is nothing but the inverse of @.( > 1) is D2. a ) of D. Since the symmetric domain of 1. is uniquely determined by G and a. C . we know that 1.( < 1) with respect to arc { z = eit : It( < E } is 1(. arg$$(eT*) = f 5 .4 Let + ( z ) = $3 o $2 o # q ( z ) . and the domain f maps onto 1. and an circular arc I. (2) Determine r ( G . then @ maps { z : IzJ < l} conformally onto D 1 with @(O) = 0.T a in {/<I 5 1). with q53(e+a) = 0.which is bounded by circular arcs I. and {C = eie : (01 5 i}. > l}. which is C } c. + Fig.5. radius r . (1)Suppose = g ( z ) is another conformal map of G onto a circular disk D.5. y > 0) and the positive axes. (Indianu ) Solution.. the reflection principle. 5 B 5 27r . where D1 is a domain bounded by { = eie. corresponds t o the arc { z = eit : It(< E } under the map @. a ) if G is the region between the hyperbola zy = 1 ( z > 0. (1) Prove that the radius r = r ( G .364 (shown in Fig. orthogonal t o = 1) and connects points e f * and e . W(0) > 0. such that g ( a ) = 0 and Ig'(a)l = 1. with center C = 0 and radius T I . such that f ( a ) = 0 and If'(a)l = 1 for some point a E G. @ ( z ) can be extended t o a conformal map by of fie onto {C : ICI < 1).
2 . in the following steps: 2. apply Schwarz's lemma to F . l#l z1 = & ( z ) = z2 : G + { q : 0 < Imzl < 2}. (a) Assume that z1.if T ( z ) = ./ = A. a ) = * f {w : JwI < &}. 3 .i = 1. Because the cross ratio is invariant under Mobius transformations.. . i. we have . (a) Let a E a' be a point different from z1. = 4 2 ( z 1 ) = eHZ1: ( 2 1 : o < Imzl < 2 ) =fey. T is uniquely determined by G and a. which implies T I = T . z T ( z ) . (2) We construct a conformal map of G onto a circular disk D.365 with F ( 0 ) = 0 and IF'(0)l = = 1. For the same reason.e. Hence r ( G .z2 (b) Use (a) to find an expression for T " ( z ) . then w = f ( z ) : G f ( a ) = 0 . Irj5i(iegx)I = 2 = d%lret=* Define f ( z ) = 4 3 ..2.($ + i)l with 43(ie%") 0.z 1 z = c. 11#:(1+ f). z2 E a' are two distinct fixed points for T . 4 2 o 41(z). If'(a)I = 1. T ( z i ) = zi. z2. .' ( w ) and we have T 5 T I . n = 1 . Id. 23 1 . + (22 : Imzz with 42(: + i) = icy.32 (Iowa) Solution. with 5207 Let T ( z ) = be a Mobius transformation. > 01.z2 21 . In other words. Show that there exists a constant c such that 3 T ( z ). with 41(1+ 22 f) = $ + i. Apply Schwarz's lemma to F(C) and we have IF'(0)l 5 hence TI 5 T .
(2n . = z .z1 = c. It follows from T"(z)l 21 = 2"Tn(z)+l z+l that T"(z) = (2" (2" + 1). (b) Prove that such confocal conics intersect orthogonally. if a t all.z 1 T ( z ). When a2 < X < b2. we obtain that z = f l are two fixed points of T. CT"l(z) . T ( a ).1) + 1) .l ( z ) . then T ( a ) = 5.366 which is Denoting T ( a ).z1 = czT"2(z) .22 T"2(z) . the curves form .z2 When T ( z )= by solving the equation = z. (a) Without loss of generality.22 z .3 2 we obtain T ( z ).. . (c) Show that the transformation 20 = $(z+:) carries straight lines through the origin and circles centered at the origin into a family of confocal conics. Choose a = 2. while when X > b2.2 2  T " .3 2 . 2 ..z1 . easy to have it is T"(3) 31 T " ( z ).  2 z z1 z zZ (b) Since T " ( z ) = T(T"l(z)).31 .a .z1 Cn . we assume a > b > 0. (Humurd) Solution. the curves form a family of hyperbolas.~  c.(2" .Y2 b2+X 1 form a family of confocal conics". a . hence c = 51 : 21 5+1 2+1 s..1)z' 5208 (a) Justify the statement that "the curves 22 a2+X +.
) 1 . 4 + Al)(U2 + A.A () When A = J(a2 + A) + [(b2 + A)] = m.+X2  u2+Ai Y2 =1 b2+Xi and L 2 : .367 a family of ellipses. &).yo) is t i = (&.T sin 8. if at all.) + ( b 2 + A l ) ( b 2 + A.+X2 a2+ A2 b2+ Y2 A2 . where A1 # X 2 . we have (u2 Y.) and which implies that the confocal conics intersect orthogonally. (b) Suppose (xO. the tangent vector of L2 at (xo.yo) is 7 2 = ($&. Noting that the tangent vector of L1 at (xo. When a2 <A <4 2 .1. . Suppose the focuses of the conics are ( ~ c ( X ) . O ) . (c) Let z = Tez'. > b2. and w =u + iv = 1z + 1 ( ) 2 Z = 1 (T 2 1 i + ) cos 8 + (r 2 T . *). Hence the curves 22 a 2 + A Y2 +=1 b2+A form a family of confocal conics.yo) is the intersection point of L 1 : . It followfrom and that = O.
the focuses of the hyperbolas are ( f l .l) = { z : IzJ < 1) fa' is an analytic function which satisfies f(0) = 0. The image of circles centered at the origin is which are ellipses in wplane. the focuses 1 w = z(z + ). and if IRef(z)l < 1 for all z E D(0.:)' of the ellipses are ( f l . prove that (Indiana) Solution. Because cos2 e + sin2 e = I.0). 5209 If f : D(0. Hence w = F ( z ) = g o f ( z ) is analytic in . l). Because a(r :)2 . Hence the transformation 0).The image of straight lines through the origin is which are hyperbolas in wplane. + = 1.1 carries straight lines through the origin and circles centered at the origin into a family of confocal conics. It is easy to know that eYc +1 is a conformal mapping of the domain {C : lReCl < 1) onto the unit disk {w : lwl < 1) with g(0) = 0.$ ( r .
Find (Indiana) Solution. IF(C>l ICI. By Schwarz's lemma. For any analytic function w = f ( z ) : fl +a' such that If1 < 1 and f ( 0 ) = 0. we have it follows from f(0) = 0 that If'(0)l I . . where 8 is a real number.we have eT+1 E . 1 < Imz functions f : R +a' such that < If1 l}. we consider the composite function w = F(C) = f o f{'(C). The equality holds if and only if F(C) = e"C. F(C) is analytic in the unit disk such that IF(C)I < 1 and F ( 0 ) = 0. which implies and the supremum is attained by f ( z ) = e"fo(z). I Choose (0 = w.369 D(0. 7r 4 5210 Let R = { z E a'.l) and satisfies F ( 0 ) = 0 and IF(. By Schwarz's lemma. and let F be the family of all analytic < 1 on R and f(0) = 0.)[ IF'(0)l 5 1. Because < 1. is a conformal mapping of fl onto the unit disk with the origin fixed.
and suppose that 1 f ( z ) I < 1+ I f ( z ) I whenever J z J 1. ) Set g(w) = (i f t = 2). w=1 I(s O f)’(O)lL 1. Let R = C\{W = u + iv : u 2 0 and w = 0). 1 < (Indiana) Solution. Because w = 23 1c (that is the inverse of C = disc {C : i.2i + 2i’ f2(z) Then F is a holomorphic function on the unit disc with F ( 0 ) = 0 and IF(z)I < 1. Describe E explicitly. Let f E P and define {f(i) < = F ( z ) = f2(z).370 5211 Let f be an analytic function on D = { z . on D with g o f ( 0 ) = 0 and 1g o f ( z ) I < 1. 5212 Let P be the set of holomorphic function f on the open unit disc so that (i) Both the real and imaginary parts of f ( z ) are positive for IzI < 1. By Schwarz’s lemma. It follows from 1 f ( z ) I < 1 1 If(z)I that f ( 0 C R. we obtain If’(O>l L 4. which implies I F ( $ ) ]5 It should be noted that when f changes in P. (ii) f ( 0 ) = 1 + i. IzI < 1) such that f ( 0 ) = 1. Then g o f ( z ) is an analytic function + + + E. (Minnesota) Solution. By Schwarz’s lemma. Prove that If’(0)l 5 4. we have IF(z)I 5 121. F ( i ) can take any value in the 1+c is a 5 f}. Let E = : f E P}. s) .
This upper bound can be reached if we let f = q5l which belongs to family 7 and satisfies 4l(8) = 4 So we obtain a.2 5 3 + 4 5 0). fE3 sup If(8)l = 4  a. r 4 .&)I 54 a. + 4)) a) IF(.4.y p s i n 4 4 = 0 by PI(+). 3 5 $}. f€3 and you should prove your assertion. Hence If(8)l = IF(4 .. Define w = + ( z ) = 2(5+ :).)I 5 1. . W+CC R = {w = u + i w : u 2 + 52 w2 > l}. p 2 ( 4 ) where pl(q5) p z ( 4 ) and Iq5 .r2sin28 8 2 4 . find sup If(8)l. it is easy to know that w = 4 ( z ) is a conformal map of D = { z : IzI < 1) onto R with d(0) = 00 and 4(4 = 8. 2 5 3 1 l r 1 4 20 E = {f() : f E P } = {reio : 1 .1 < arcsin . If we denote the two roots of p2 . we obtain that {w:Iwz~< Hence }={w=pe*+: 8 3 lr 4 20 Id1<arcsin. (Indiana) Solution.$i w ' l the set : f E P} is equal to {f2(i) 10.371 conformal mapping of { C : 5 :} onto {w : I .p~psind+4<0). Then F ( z ) = f o 4 ( z ) = f ( 2 ( 5 is analytic in D and satisfies F ( 0 ) = 0 and IF(z)I 5 1. By Schwarz's lemma. the set E can also be represented by < + 5213 Let 3 If F is the family of all analytic function on R such that If1 5 1 in R and lim f ( w ) = 0.$1 5 arcsin $. Your answer should be an explicit number.
b. it can be written as f ( z ) = where a. Since A = (d . we know that f ( z ) = z has two conjugate roots. f is injective).a ) z . So f has a unique fixed point inside D. Since R is convex. ( S U N Y . Thus we have a + d = 0 and the inequality a d . t E [0. we have f Hence > 0 for z E Q. Now we consider the equation f ( z ) = = z .bc > 0.b = 0. 1 3 is the line segment connecting z1 and 22. which is equivalent to cz2 + (d . Since f is a conformal map of D onto itself. which implies . L c il. L : z ( t ) = zl +t(zz . Hence adbc > 0 and a2+bc = d2+bcimpies f = id. one in the upperhalf plane and the other in the lowerhalf plane. which contradicts the condition f # id. S t o n y Brook) Solution. we know that f ( z 1 ) # f ( z 2 ) whenever z1 # 2 2 . + bc)z + b(a + d) + d)z + d2 + bc * f(z)= C(U d2 If a + d # 0.e. Prove that f is onetoone in 51 (i.372 5214 Let D be the upperhalf and let f # id be a conformal map of D onto itself such that f o f = id. Let z1 # z2 be two arbitrary points in R. Since Ref‘(z) Jif‘(z(t))dt # 0. + bc # 0.a ) 2 + 4bc is equal to 4bc + 4a2 < 0. d E IR and ad .z l ) . (Indiana) Solution. then b = c = 0.bc > 0 can be written as bc a2 < 0.. open subset of a‘ and let f : R a‘ be an analytic function satisfying Ref’(z) > 0. z E 51. It follows from f o f = id that b(a + d) = c ( a + d) = 0 and a2 + bc = + 5215 Let fl be a convex. c. Then (a2 s. Prove that f has a unique fixed point inside D.
( z .. (b) For any nonvanishing polynomial P ( z ) . z 2 . z. . (a) Let z1.z1)(z . .l).z 2 ) . be the zeros of P ( z ) ...l ) ~ .zl)(z . not identically zero..21 z . and P ( z ) = u ( z . is onetoone in the unit disk + + + n1 (l)nlna. then [nu.[ 1. we have n + > 1.n). Stony Brook) Solution. we obtain Iul = n..2. ~ n1 j=1 n1 4 lal' 5 1. By assumption. then Izj 2 1 for j = 1 .z. (Courant Inst.n . . Since a1 = 1. + P ( z ) z . IzI < 1 and a1 = 1.) Solution. In other words. .( < 1). . .. . j=1 zj = al. 2 . 5217 Let P ( z ) be a polynomial on the complex plane.). Let z1. ( z .22 z . It follows that 1 1 P'k) + 1 (logP(z))' = . Rezj > 0 ( j = 1.373 5216 Show that if the polynomial P ( z ) = a. .l z " .2 2 ) ( z . . . .1. by comparing the constant terms. z z . Because P ' ( z ) can also be written as nu. . It follows from the univalence of P ( z ) in { IzI < 1) that P'(z) = nu. (a) If all roots of P ( z ) lie in H .z. z.. + 2 u ~ z + u l# 0 for all z E 1. .+ . .z. use the result in (a) to show that the convex hull of the roots of P ( z ) contains the roots of d P / d z . . 5 ( S U N Y .1 be the roots of P ' ( z ) . show that the same is true for the roots of dP/dz. let H = { z : Rez > 0).^^^ + (n.zn un1zn' alz ao.~ + . the roots of P ' ( z ) are all situated outside the open unit disk. .
we obtain that the zeros of P’(z) lie on the convex hull of the zeros of P ( z ) . we have Re{eae(z . Hence Re < arg(zzj) < $.374 When z E { z : Rez 5 0}.) or equivalently. Re& < C j=1 n & < 0. (a) Show that there exists on { z : IzI = 1) with < (b) Characterize those functions with (Minnesota) Solution. . then 0. which shows can not be zero on { z : Rez 5 0). When z is on the left side of I. (a) Let f ( z ) = too n=cu C bnzn.z j ) ) < 0. Denote the intersectional angle from the positive direction of the imaginary axis to I by 8. z. . and I is a directed straight line passing through two zeros zk and zi such that the other zeros are on the right side of I (including on I ) . . z 2 . . be the zeros of P ( z ) . 5218 Let f ( z ) be a Laurent series centered a t 0. p(. then Hence . with residue b at z = 0. (b) Let z1. After considering all the directed straight lines passing through two of the zeros of P ( z ) such that the other zeros are on the right side of the line. convergent inC\{O}. Hence which shows that the zeros of P ’ ( z ) do not lie on the left side of I .
It follows fiom that which implies that (p(8)= 8. .<l1 < J b .l)ei+(e). 5219 Assume f is analytic in a neighborhood of 0.P I = Ib .we obtain f(z) = z EC\{O}. Let f(() . and f maps d D into a l l .<l = ( b . Apply the discreteness of zeros for analytic functions to f ( z ) .f’(z) # 0. (b) Show that &[argf(e”)] > 0 for 8 in R. b1 p. then which is a contradiction.1)holds for all C with I(I = 1. and hence f ( 0 . p. where ( = eie and 4(8) is a continuous realvalued function. it follows from that If(C) .cl = 7 holds on {C : 1[1 = 1). ( a ) Show that Vz E aD.11. f maps D into D . Hence there exists ( with = 1 such that (b) If max ICI=1 If(<)  = Ib . where D = { z : IzI < 1).1 1 holds for all C with I(I = 1.375 If If() .
z o ) + b z ( z . t .W O ) = nArarg(z . g ( e i 6 ) < 0 implies that R > 1 when r < 1. ( IndianaPurdue) Solution. Hence f(z) has no zero in (0 < I1 < I}. then = at point ei6.ZO) 3T n+ Arargg(z) > . Show that f(z) # 0 whenever 0 < IzI < 1. which implies that g ( e i 6 ) = f'(ei0) = 0. Then where n 2 2. 2 3 which implies that f(z) assumes values outside the disk when z E I. we know that % ( e i 6 ) = 0. Hence f'(z) # 0 for all 2 € aD. Let I be an arc in ' defined by I = { z E % : ) z . . with bo # 0. Since R = 1 when r = 1.. Let f(z0) = wo E all. ' and denote by Ar#(Z) the change of # ( z ) when z goes along the arc I in ' the counterclockwise sense. it follows from = aR that %(e i6 ) < 0. It is demanded that r is sufficiently small such when z E I?. g(z) = b o + b l ( z .. This is also impossible. It follows from that Ararg(z . Hence we obtain 2 2 7s Rg (c) Because f l a ~ a twetoone map from 8 0 onto a D . z = 0 is a zero of f of multiplicity m = 2. which implies that f ( z ) has two zeros (counted by multiplicity) in D. (a) Assume f'(z0) = 0. = = = 0. where zo E d o .~ 0 ) 2 + .zo) > and Ig(z) . . . (b) Let z = rei6.376 (c) Assume that f(0) = f'(0) = 0 and f l a ~ a twotoone map from a D is onto dD.bol < f ( z ) .z o ) " g ( z ) ( n >_ 2). Since f(0) = f'(0) = 0.> K . It is ' a contradiction to the fact that f maps D into D.wo = ( z . If g ( e i e )1 0 .A variation of the CauchyRiemann equations for analytic function w = f(z) is Since f maps a D into a l l . &Alzl=largf(z) = is 2. If g ( e i e ) < 0 . that Ararg(f(z) . and w = f(z) = Rei@.zoI = r ) . But from (a) it is impossible.
( n.= e.. + By the residue theorem.. . and its Laurent expansion The coefficient of the term $ in the above development ~ is 1 1 + 1 + 2! + . Function e e z is analytic in { z : 0 around z = 0 is: < IzI < +co}. (Indiana) .377 SECTION 3 COMPLEX INTEGRATION 5301 Evaluate the integral (Indiana) Solution.l)! 5302 Evaluate where y is the positively oriented circle { Jz1= 1). we obtain 1 +.
.. Hence we obtain a = 2. 1 U a ( .+. its and F(z) = (.... The Laurent expansion of around z = 0 can be obtained as follows: &  1 sin3 z  1 (z  k. By the . i. and with z = 0 as a pole. . Function F ( z ) = ( $ 5 ) c o s z is analytic in { z : 0 < IzI Laurent expansion around z = 0 is: + < +a}.378 Solution... the coefficient of the term in the above development is zero. + 5) +.)cosz= 2+('2)P+(. is $.) (: coszdz (Indiana) Solution.2aiRes 5303 For what value of a is the function f ( z )= singlevalued? LZ(. = 1 The necessary and sufficient condition for f ( z ) to be singlevalued is that the residue of F ( z ) a t z = 0 is zero. . 1 1 1 2 t4!z4..)Z+. sin z =~ i . we have p in the above development ( 3 1O ) .23 + $25 1  . It is obvious that is analytic in { z : 0 < IzI 5 l}.e. I ~ Hence the coefficient of the term residue theorem.
et  A (t . For any r < e . Hence the largest possible P is equal t o e. It is easy to see that when t + 1.IzI < 2).rt and the convergence of the integral that is uniformly convergent in any compact subset of { z : IzI < e}.et dt.1)2’ where A = $ cos 1. + What is the largest possible P so that h ( z ) is analytic for IzI < P? ( IndianaPurdue) Solution. et . which implies that the integral is divergent.379 5304 Define h ( z ) = L m ( l ztet)let cos(t2)dt. Hence P can not be larger than e . let IzI 5 T . we know that h ( z ) is analytic in { z : IzI < e } . Show that . By Weierstrass theorem. When z = e. 5305 Let f ( z ) be analytic in S = { z €67. dt It follows from let + ztl 2 et . Consider the integral et + z t dt. cos(t2) et .
Stony Brook) Solution.380 (Iowa) Solution. 1 ( S U N Y . Let f ( z ) = u ( z ) iw(z). 1 . It is easy to see that It follows from the above three equalities that 5300 'u Suppose that the realvalued function u is harmonic in the disk {IzI is its harmonic conjugate and u(0) = v(0) = 0. dz dz z where y(t) = eZnit.t E [0. and we have + . Show that < 2}. Then f ( z ) is analytic in { z : IzI < 2}.
IzI 5 11.381 It follows from u ( z )= and f(z) +m 2 f f (> . prove that > 0 for all points (Indianu ) . z E D ( 0 . l).l) = { z . w(z) = that  22 5307 Let f be an analytic fqnction on an open set containing D(0. (a) Prove that d”f(0) = dzn (b) If f(0) = 1. and if Ref(z) /2T 0 enie[Ref (eie)]dO.
2 0. we have = 2(n!). Hence (b) Because Ref(z) is harmonic on D(0. (a) Assume that 00 k=O we have By Cauchy Integral Formula. kI'" Noting that Ref(eis) Ref(e")de = Ref(0) = 1. l). . by the meanvalue formula of harmonic functions.382 Solution.
383 5308 I f f is analytic in the unit disk and its derivative satisfies IfWI I (1. = 1 f)odz. (0 < T < 1). 5309 Let f = u + iw be an entire function. : (1 + )I n1 1 < e.14)1. . (Stanford) Solution. It follows from f ( z >= 00 C anzn n=O that f’(2) = c 00 nunznl. It is obvious that lull = I’0l I 1 < e . where e is the base of natural logarithms. n=l where nu. we choose T = 1 . show that the coefficients in the expansion n =O satisfy lunl < e for n 2 1.. f() For n > 1.
"(. Then it follows from Schwarz's theorem that when IzI < R. then f ( z ) is a polynomial of degree bounded by h. ( Stanford) Solution. Let R be an arbitrary positive number. 5 A+BR~ R" . . (Indiana) Solution. (b) Let For any integer n > h.'(. (a) Let F ( z ) = e . F ( z ) must be a constant. Letting R + +m. which implies that f ( z ) is a constant.).(ua(z)v2(z))Ziu(L)v(r) Then F ( z ) is an entire function with IF(z>I = e (.384 (a) Show that if u 2 ( z ) 2 wz(z) for all z EC. Show that f is a polynomial of degree at most n. = 0. 5310 Let f be an entire function that satisfies IRe{f(z)}I 5 lzln for all z . then f must be a constant. (b) Show that if If(z)I 5 A+BIzlh for all z € @ w i t h some positive numbers A .f 2 ( z ) = .)) 5 1. By Liouville's theorem. which implies that f ( z ) is a polynomial of degree bounded by h. h. where n is a positive integer. we obtain that a. B .
Let +B where Hence when k > n.385 Especially when IzI = +. which shows that f(z) is a polynomial of degree at most n. 3R" * 2n + IIm{f(O))l = 3R" If(. 5311 Compute the double integral 1J . (Iowa) Solution. B such that If(z>I 5 AlzY holds for all z EC.)l L 1 + IIm{f(O))lr which implies that there exist constants A . First we have the following complex forms of Green's formula: . cos z d x d y where D is the disk given by { z = x + i y E a' : x 2 + y 2 < 1).
1 sin zd( ) Z 5312 n=O be analytic in D = (1.386 The problem can be solved directly by either one of the above two forms: o r JJDcoszdxdy = sin zd? = . (Indiana) Solution. (a) Express A in terms of the coefficients an. (b) Prove that for z E D. < 1) and assume that the integral is finite.sin z dz = R. (a) B Y n=l we have .
Let the Laurent expansion of f in {z : 0 < IzI < l} be 00 From .387 Noting that we obtain that . The answer to the problem is Yes. we have 5313 Let f be analytic in (0 < IzI < 1) and in L2 with respect to planar Lebesque measure. ( Stanford) Solution. Is 0 a removable singularity? Proof or counterexample. 1 00 00 (b) By Cauchy’s inequality.
the left side of the above inequality will tend to infinity. must be zero when n 5 1. Otherwise. Hence n=O which shows that z = 0 is a removable singularity of f . and then . 5314 Evaluate the integral (Indiana) Solution. which leads to a contradiction. while the right side of the inequality is finite. 12r2n+1dr< Then a. a = rei+.388 we have Let E < 1 be a small positive number. .21r 1' la. let E + +0. Let z = peie.
Then we have I(a) = dx 2dx dx . (Columbia) Solution. When r > p.When r < p. Denote It is obvious that I ( u ) is an analytic function in { u : la1 lr > 1). 5315 Evaluate by the method of residues.
=l (z  idz  z1)(z . 1 = 0 by z1 and z2. 1221 < 1. we may assume that ll > 1.  2dm' 0. Since z1 z2 = 1.z2 7T 27r d(Zl t 2 2%& * ~ ~ ) ~ 42122 .6)de when ( z ( < 1.2(2u 1).=l + + 22  idz 2(2u + l ) z + 1' =  iz.390 Let z = eisl then dx cosx and = = z dz iz ' + 21 2 ' = Denote the two roots of z 2 .27T z2) z1 . By the residue theorem we have zl iz. It should be noted that branch of is also analytic in { u : la1 > l}. (b) Integrate the Taylor expansion n=O term by term to find the coefficients in the Taylor expansion . and the d m should be chosen by a 5316 r g d w Consider the function (a) Use the residue theorem to find an explicit formula for f(z) = 1 2n g(z.
(Courant Inst.y . Because IC1 .zk k=O 00 where Uk 2r = (1)' sink ode. (a) Let c = ele.( (2 < 1) is defined =i(dnl). m in 1. k=O (1.1 ) ' s i n k 8 .e) = C ( . and f(z)=$' (g(l)ksink8.391 (c) Verify directly that (a) and (b) agree when IzI < 1. Z 2 " n=O .) Solution. the integration with respect to 8 can be taken term by term.eiO sin8 = 22  C2 1 2ic ' and where of d ci = $(d=l). It is easy to obtain that uznl = 0 and 2R (c) In order to verify that (a) and (b) agree when IzI function f(z) in (a) into a power series: < 1.i0 . z k . and the singlevalued branch by 4 I r = ~ = 1.l ) " C I f . C21 = 1. we develop the f(z) = 2K ~ 00  dC7 27r(l z"i = 2 7 r . < 1). Then . Since the series converges uniformly for all 8 E [0.27r]. (b) It follows from I sin81 5 1 and 00 IzI < 1 that g(z.
' Define f ( z ) = e" and choose the contour of integration I = r3 u r4 as shown in Fig.5.392 we know that the results in (a) and (b) agree when J z J< 1.5. (UC.5. 5317 If a is real. by Cauchy integral theorem. ' .5 As f ( z ) is analytic inside I. Irvine) Solution. First we have It follows from the existence of hllRex2dz R that exists. rl U I'2 U Fig. show that exists and is independent of a.
0 R Fig . f ( z ) has one simple pole z = eEi inside I'. dY it follows from the facts that eR2 + 0 ( R + CQ) and 5318 Let n 2 2 be an integer.6 Let f ( z ) = &. By the residue theorem. .393 ieR2 J. we have f(z)dz = 2~iRes(f. and select the integral contour r as shown in Fig.6. Compute (Iowa) Solution.5.' ey2+2Ryi = 0.5. Letting R + 00.e:*).
5.z 2 .5. (Minnesota) Solution. Fig. 3 j=1 . and choose the contour of integration f(z) = have eIZ r= 3 rj j=1 as shown in Fig.7 Define f(z)= e .7. we ' . B caus is analytic on I' and inside I by Cauchy integral theorem.394 and 5319 Evaluate JD cos (x') dx with full justification.
sin z2)dz = 2 ' which implies 5320 Evaluate (Iowa) . we have For the integral of f ( z ) on r3. then where 72 = { w :lwl= R2.395 For the integral of f ( z ) on r2. L m ( c o s z2 f(z)dz = I" / 0 ex2dz + 2 f fi when R +00. 2 7F By Jordan's lemma. Hence we obtain by letting R co that + sin x2)dz + i 00 & (cos z2 .O 5 argw 5 }. we make a change of variable by w = z2. we have It is well known that l.
8. Because f(z) is analytic inside r. Fig.396 Solution. by Cauchy integral theorem.5.8 Define and select the integral contour as shown in Fig.5. where It is easy to see that lim Rr 00 I' iReief (Reie)d6= 0 and .
2 7r 5321 Let f ( z ) be holomorphic in the unit disk IzI 5 1. we know that Res(f. ( S U N Y . and the singlevalued branch of logz be chosen by argzl.5. J. Fig. .397 Since the Laurent expansion of f about z = 0 is 00 where a1 = 2i.9. Prove that where respective integration goes along the straight line from 0 to 1 and along the positively oriented unit circle starting from the point z = 1. The branch of log is chosen to be real for positive z. by Cauchy integral theorem. Since f(z)logz is holomorphic inside the contour I?.9 ' Let the contour of integration I be shown as in Fig.5. Stony Brook) Solution.we obtain dx = .f ( z )logzdz = 0.=I = T . 0) = 22. Letting E + 0 and R + ca.
. 5322 where a and b are complex constants. then d4 = and + g. First we assume la1 > lbl. and then the multivalued analytic function log(a bz) has singlevalued branch on { z : IzI 5 1). we obtain where the integration contour IzI = 1 has starting point and end point z = 1. and the value of logz at the starting point z = 1 is defined as 0.398 where It is easy t o see that Letting E + 0. I 2r log la + be'@]& = Re = R e ( 2 ~ l o g a )= 2rlog lal. not both equal to zero. Take ei@ = z . (Hamud) Solution.
10.5.399 When la1 < lbl. let b = ae". Since f ( z ) is analytic on I' and inside r. Because + s .10 In order t o evaluate the integral we define f. we have JR I*log 11+ ei@1d4 = Re JR IR log(1 + e*+)d4 . f(z)dz = 0. () = log( 1 z +z) 7 where the singlevalued branch is defined by log(1 z ) lz=o= 0. Choose a contour of integration = r. as shown in Fig. by Cauchy integral theorem. we have In the case la1 = lbl. Uy. Then Fig.5.
Fig.5.. The singlevalued branch of logz is chosen by argzI.log lbl}. f(z)dz = 2?riRes(f.1 = ?r. l).11 Let and select the integral path I? as shown in Fig. we have J. By the residue theorem.400 Hence we obtain J.'" Evaluate log la + beiGldq5 = 2 7 max(l0g la[. ~ 5323 (Iowa) Solution.5.11. .
1) = 27ri + 2 2 . Hence 2aiRes( f .ai. Comparing the imaginary parts on the two sides of the above identity. ice"f (&eie)d6.l). it turns out that 4ai log x + 4a2 dx = 2ai + 2 2 . Rrm lim J. we obtain .'" iReief(Reie)d6 = 0 and In order to find Res(f. AS E + 0 and R + 00.401 where 2X iRegef (Reie)d6 + (logx + 2ai)' dx rO + It is obvious that 1. we consider the Laurent expansion of f about z = 1: = where a1 n=3 c a n ( z + lyl = 1 .
12 Define 1 The singlevalued branch for log(z + 1) is chosen by log(z + l ) l z = o = 0.4 ) l o g ( r + l ) (the integration is over the imaginary axis). 2). (a) Fig.402 5324 Evaluate the following integrals: +im at (a) Sim ( t 2 .12. we have f ( z ) d z = 2 ~ i R e sf(.5. where and Because . (b) (Courant Inst. &dz for CY in the range 1 < QI < 2. As f ( z ) is analytic on and ' inside J? except a simple pole at z = 2. and the contour I of integration is shown in Fig.) Solution.5.
I F where and Res(f.4)log(z + 1) = (14 Ti ). we have f ( z ) d z = 2 ~ i R e s ( f e:')).5. and the contour 'I of integration is shown in Fig.13 Define The singlevalued branch for zm is chosen by argzIr=z. +ice dz ( z 2 .O = 0. e f i 3e +a .efa) f ( z ) z+e:' . As f ( z ) is analytic on and inside r except .5.2 log3 Fig. we obtain ..403 and by letting E + 0 and R + 00.1  3e f ( 2 .efa) = lim ( z .m ) i ' . . a simple pole at z = e:'..13./.
14 where . Let ' where (argza)z=z>O = 0. What happens if a = l? Solution. for 1 00 Xu 7r( 1 . Fig. we have f ( z ) d z = 2aiRes(f(z). f(Reie)iReied8 = 0 2a  when a < 2 and lilil ' + f(&eie)iseied8= 0 + co.5. 5.14.a ) ( 1 + x 2 ) 2 d x = 4 cos( y )' < a < 3. i ) .we 7r when a > 1. By the residue theorem. a # 1.404 Because lirn R U C 2z . by letting E 0 and R obtain dx = 5325 Show that J. and select the integral path I as shown in Fig.
" iEeie f (&eie)d8 and Res(f(z).+ (1 when a = 1. When a # 1.a(1 . and from a > 1 that Letting E + 0 and R $ 00. I.a ) x y d X= 4 cos( y) ' X T(1a)  1 5326 (a) Prove that converges if 0 <a < 1 . 2 " al a) ( .O iReief (Re")dB = 0 .i) = lim [ ( ~ 2'2 2 ta i)2 + 1'9  e"i 2 .405 + J. we obtain XQ a(1a) 2 e 1" 2 . It follows from a < 3 that lim RC .
406
(b) Use complex integration to show that
?fa x P a cos x d x = sin  . I( a ? 2
+ 1).
(Hamad)
S o h tion.
(4
1
03
eixxudx =
(
1
J!
x  cos x d x ~
+
x P a cos x d x )
+i
03
x  sin x d x . ~
It follows from a < 1 that
z  cos x d x ~
is convergent. It is also obvious that
lLA
cos x d x
1 LA I
5 2,
sinxdx
5 2,
xPa is monotonic decreasing and
lim x  =~O X++03 for
(Y
> 0.
x  sin x d x are ~
By Dirichlet’s criterion, we know that 2  O cos x d x and e i z x  a d x is convergent when 0 also convergent. Hence
s;”
< a < 1.
(b)
zRD
iC o
c
R
Fig.5.15 Let f(z) = z  a e  z , and the contour of integration r is chosen as shown in Fig.5.15. The singlevalued branch of f ( z ) on I? is definde by zaIz=2>o > 0.
407
By Cauchy integral theorem,
+L
0
i&e"f(&eie)de 0. =
It follows from a
< 1 that
l i i
L
0
ice'e f (&eie)d8 0, =
and from a
> 0 and Jordan's lemma that
im l iReief (Reie)de= 0.
a
.* 1
Letting
E +
0 and R +
00,
we have
Multiplying both sides by e y i , and comparing the imaginary parts, we obtain
I"
x  cos xdx = sin I?( ~ 2
Ta
a
+ 1).
5327
Use a change of contour to show that
dt 7
provided that a and p are positive. Define the left side as a limit of proper integral and show that the limit exists. (Courant Inst.)
Solution. Since
408
is monotonic with respect to x and
1 lim = 0, x p
x++a,
+
the convergence of the integral
dx
follows from Dirichlet's criterion. Define
and choose the contour of integration I' = I'l U I'2 U
I ' 3
as shown in Fig.5.16.
Fig.5.16 By Cauchy integral theorem, we have
=
1
R
ax
'(

x2+pz
+
r
l2 1
f(z)dz
R
axi
b
d
x = 0.
It follows from Jordan's lemma that
Letting R +
00
and considering the real part in the above identity, we obtain
409
5328
(a) Let c be the unit circle in the complex plane, and let f be a continuous a'valued function on c . Show that
is a holomorphic function of z in the interior of the unit disk. (b) Find a continuous f on c which is not identically zero, but so that the associated function F is identically zero. (Minnesota) Solution. (a) Let zo be an arbitrary point in the unit disk. Then 1  lzol = p > 0. Choosing 6 > 0 such that 6 < p, we prove that
has a power series expansion in { Iz  zo I 5 6). It is clear that
when Iz  z0I 5 6 and continuous on c . Thus
C E c.
We can also assume
I f (<)I 5
1
M because f is
fo=
cz
(C  20)  ( z  zo)
f( 0

f 0 . (
<  zo
1 C  2 0 zlo
As
and
n=O
(f)n is convergent, the series C
n=O
00
( ) z "
converges uni
formly for all
E c . Hence termwise integration is permissible, and we obtain
410
Since zo is arbitrarily chosen in the unit disk, F f ( z ) is holomorphic in { IzI (b) Take f(C) = f = 1). Then
(Icl
< 1).
F f ( z )=
In fact,
/
C
1 < d
=
( 1

t)
1 dC = ;(2ai

2ai) = 0.
can be taken as for any positive integer n and fixed zo E { z : IzI < 1). When E c ,
f(C)
c
5329
Let [a,b] be a finite interval in B and define, for z in
D = C  [a,bll
f(z) =
J." tA. z
Show that f(z) is analytic in D. Given c, a < c < b, calculate the limit of f ( z ) as z tends to c from the upper half plane and as z tends to c from the lower half plane. (UC, v i n e ) I Solution. For any zo E D , choose 6 > 0 sufficiently small such that { z : Iz  201 5 6 } n { z = a : + i y : y = O , a L z I , b ) = 8 . Whenlzzol < b , a < t < b , w e h a v e 1 1 tz (tzo)(zzo)


. 1
t0
1 1 t  % a
00
n=O
and the series converges uniformly for t with a I t ,
5 b.
Hence
41 1
holds for z E { z : Iz 201 < b } , which implies f(z)is analytic in { z : )z  zoJ < 6). Since zo is an arbitrary point in D ,we obtain that f(z) is analytic in D. For z E D, f ( z ) can also be represented explicitly by
where the singlevalued branch is defined by arg r = s o > b = 0. Let rl and r be two continuous curves connecting z = xo > b and z = c in the upper 2 half plane and the lower half plane respectively. Then the limit of f ( z ) as z tends t o c from the upper half plane is log  +iAr,arg
[::I
ab
za
 log
1
cb
ca
+ *i,
while the limit of f ( z ) as z tends to c from the lower half plane is
5330 For each z E U = { z : Imz
> 0)
define
g(z) =
C l ,ta
1
sin2 t
dt.
Determine which points a E B have the following property: there exist E > 0 and an analytic function f on D ( u , E such that f(z) = g ( z ) for all z E U n )
D(a,E l .
(Indiana)
Solution. Let I be the half unit circle in the lower half plane whose direction is ' defined from point z = 1 to point z = 1, and define a function
It follows from the Cauchy integral theorem that when z E U ,f ( z ) 5 g ( z ) . With a similar reason as in problem 5328, f ( z ) is analytic in the complement of I?. Hence we obtain that for any a E a, # f l , there exists E > 0 a
412
( E < min{lall, la+ll}) such that f ( z ) is analyticin D ( u , E= { z : Izal < E } ) and f ( z ) = g ( z ) for all z E U n D(u,E). When a = 3 ~ 1such a f(z) does not exist. The reason is as follows: As ,

sin2t  sin2t  sin2z tz tz
+sin2z' tz
where sin2 tsina tx know that
x
is an analytic function of two variables for ( t ,z ) E a' x a', we
h ( z )= 2 ~ i
' 'J'
sin2 t  sin2 z dt tz
is analytic for z E C. But sin2z dt = 2Ti
Il'
sin2z z 1 dlog(t  z ) = 2Ti log z+l'
which has branch points z = f l , hence g ( z ) can not be analytically continued to D ( & l , & ) .
413
SECTION 4 THE MAXIMUM MODULUS AND ARGUMENT PRINCIPLES
5401
Let u €67, JuJ5 1, and consider the polynomial
U P ( z ) =  + (1  ("I.> 2
 z
a 2
2
.
(Indiana)
Show that ( P ( z ) (5 1 whenever
12)
5 1.
Solution.
~ ( z )=
=
When IzI = 1,
+ (1 IaIz)z  a 2 z 2 2 l a z [ ( l  la12)+ (  izz)]. 2 2
U


U a u a Re(   Z z ) = Re[  (Cz)] = Re[  1 = 0, z Z z z
IIm(  az)I _< 21al.
2
a
Hence when ( z (= 1,
l a lp(z)12 = (1  Ia12)>" (Im[(   ~ i z ) ] ) ~ 2 2 <_ (1  2[u(' + la14) [aI2= 1  luI2 + laI4 5 I.
+
+
By the maximum modulus principle, IP(z)( 5 1 whenever IzJ 5 1.
5402
Let f be holomorphic in the unit disk 1(. < l}, continuous in 1.( 5 1) and If(.)/ = 1 whenever (zJ= 1. Prove that f is a rational function. ( S U N Y , Stony Brook)
414
Solution. If f ( z ) has infinite many zeros, by the isolatedness of the zeros of holomorphic functions, the zeros must have limit points on the boundary of the unit disk. But it will violate the fact that f is continuous in ( 1 . ~ 1 5 1) and If ( . ) / =1 whenever IzI = 1. Hence f has only finite zeros in the unit disk. Denote all these zeros by zlr 22,. . . ,z,, multiple zeros being repeated, and define
Then F ( z ) is holomorphic in 1.( < 1). continuous in 1(. 5 1) and IF(z)I = 1 when I z J = 1. By the maximum modulus principle, IF(z)I 5 1 in (1. 1 5 1). Since F ( z ) has no zero in (IzI 5 1) L is also holomorphic in 1.( < l}, F(z)
= 1 when IzI = 1. Application of the yields IF(z)I 2 1 in 1(. 5 1). Hence maximum modulus principle to IF(z)I = 1 holds in ( 1 . ~ 1 5 l}, which implies F ( z ) = eia with Q: a real number. So we obtain
continuous in 1) and
(1.~1 5
I&JI
&
5403
Let f be a continuous function on = { z : IzI 5 1) such that f is analytic in U. If f = 1 on the halfcircle y = (eie : 0 5 8 5 a}, prove that f = 1 everywhere in (Indiana) Solution. Define F ( z ) = (f(z)  l ) ( f (  z )  l), then F ( z ) is also continuous on V and analytic in U. When z E d U , we have either f ( z )  1 = 0 or f(z)  1 = 0. Hence F ( z ) = 0 holds for all z E which implies either f ( z )  1 0 or 0. Since f(z)  1 0 is equivalent to f(z)  1 0, we obtain f(z)  1 f ( z ) 1 for all z E V . Remark. The condition that “f = 1 on the halfcircle y” can be weakened t o that “f = 1 on an arc y = (eie : 0 5 8 5 f } , where n is a natural number”. In this case, the proof is the same except that F ( z ) is defined by
v
v.
v,
~ ( z = ( f ( z >  l)(j(ze:z) )

l ) ( f ( z e + i )  ~ ) . . . ( f ( z e ? ~ a )  1).
2n 1
415
5404
Let S denote the sector in the complex plane given by S = { z : 5 < argz < $}. Let 3 denote the closure of S. Let f be a continuous complex function on 3 which is holomorphic in S. Suppose further (1) If(z)I 5 1 for all z in the boundary of S; (2) ~ f ( z i y ) ~ e f i for all z i y E S. 5 Prove that If(z)J 5 1 for all z E S. (SVNY, Stony Brook) Solution. Let F ( z ) = e  ' " f ( z ) , where E > 0 is an arbitrary fixed number. Then F ( z ) is also continuous on 3 and analytic in S. When z is on the boundary of S, J F ( z ) I = e'"If(z)I 5 1. When IzI + $00 < argz < %), IF(z)I 5 e€" . e f i + 0. By the maximum modulus principle, we have IF(z)I 5 1 for all z E S, which implies If(z)I 5 lea'[ = ear for all z E S. Because E > 0 can be arbitrarily chosen, letting E + 0, we obtain lf(z)I 5 1 for all z E S.
+
+
(4
5405
Let K be a compact, connected subset of@ containing more than one point and let f be a onetoone conformal map of@\K onto A = { z : J z J 1) with < f(00) = 0. If p is a polynomial of degree n for which Ip(z)I 5 1 for z E K , prove that M z ) I 5 If(z)I" for z EG\K.
(Indiana) Solution. Because f is a onetoone conformal map of@\K onto A with f(w) = 0, it has a simple zero a t z = 00. Since p is a polynomial of degree n, it has a pole of order n a t z = M. Hence the function F ( z ) = p(z)fn(z) is analytic in G\K which contains point z = 00. As f ( z ) maps@\K onto A = { z : IzJ< l}, we have lim lf(z)J = 1. Together with Ip(z)I 5 1 for z E K, we know that the
%+K
limit of IF(.)\ when z tends to K can not be larger than 1. Apply the maximum modulus principle to F ( z ) o n @ \ K , we obtain IF(z)I 5 1 for z E G \ K , which implies Ip(z)I 5 If(z)I" for all z E@\K.
416
5406
Suppose f and g (nonconstant functions) are analytic in a region G and continuous on the closure of the region. Assume that E is compact. Prove that ( g ( achieves its maximum value on the boundary of G. (Iowa) Solution. Assume that 191 achieves its maximum value c ( c > 0) at zo E we prove that if zo E G , then f and g must be constants. Let If(zo)I = f(zo)e*+’, Ig(zo)I = g(zo)ei+2.
If1 +
If1 +
z,
Then for fixed
4 1
and
42,
~ ( z = f(z)ei+l )
+ g(z)e”Z
is analytic in G and continuous on
G. It follows from
= ~f(zo)I
IF(.)I
~ ( 2 0 )
5 If(.>I+ Is(z)I L c ,
= f ( z o ) e i + l + g(zO)ei”
~ ( z= f ( z ) e i + l )
+ Ig(zo)I = c
and zo E G that
+ g(z)ei+2
must be the constant c . Without loss of generality, we assume that f is not a constant, and try to lead to a contradiction. Since the image of an open set { z : Iz  201 < 6) c G under f is an open set which contains point f(zo), f(z) assumes all the values f ( z ) = f(z0) &ei+ for small E > 0 and 0 5 4 < 2 r in { z : Iz  z0I < 6). Then 7 when 4 4 # 0, ?r, we have 1
+
+
lf(.)I + lg(z)l
= =

lf(.)I + Ic  f ( z ) e i 9
lf(z0)
+
+ ~c f(zo)ea+l
 &ei(+++l)l
l,cei(++h) + f(zo)e”l f(zo)ei+l
I + l&ei(+++l)  g(zo)ei@21
>
+ g(zO)ei+Z = c,
If1 + lgl.
Hence f must
which contradicts that zo is a maximum value point of be a constant, which also implies g is a constant too.
417
5407
Suppose f(z) is an entire function with
Show that f(z) is identically 0.
(Iowa)
Solution. For any R > 0, consider function
g(z)
= (Z  Ri>(z R i ) f ( z ) .
+
When IzI = R, and Imz 2 0, denote by 0 the angle between the line perpendicular to the imaginary axis and the line passing through z and Ri. Then 0 5 0 5 5, and
When Iz( = R, and Imz < 0, denote by 8 the angle between the line perpendicular to the imaginary axis and the line passing through z and Ri. Then 050< and
t,
It follows from the above discussion that when J I R, z=
By the maximum modulus principle, when IzI < R,
Now fixing z, and letting R + +m, we obtain f(z) = 0. Since R can be arbitrarily large, we have f(z) = 0 for all t EC.
418
5408
Suppose f is analytic on
(2;
0 < IzI
< 1) and
Show that f E 0.
(Indiana)
Solution. Denote the Laurent expansion o f f on { z ; 0
< IzI < 1) by
where
It follows that
When n < 0, letting r + 0, we have
a, = O
( n = 1,2,...),
which implies z = 0 is a removable singularity of f . In other words, f can be extended to an analytic function of the unit disk. = 0 when IzI = 1. By the maximum modulus principle, we Since log 121 obtain f 0.
5409
Let f be an analytic function on D = { z : IzI < l},f ( D ) E D and f(0) = 0. (a) Prove that f (  z ) I 5 2)zI2 for all z in D and if equality occurs for some nonzero z in D , then f(z) = eiaz2.
).(fI
+
. is a real constant. the other coefficients must be zero. By the maximum modulus principle.) f(z)I 5 21zI2 holds for all z E D . Let M n=l it follows from that a2 f ( z >+ f ( . then F(O) = 0. If(. If equality occurs for some nonzero z in D . Because If(z)I < 1 for z E D . we have + I I f(z) + f ( . limit of the Hence can not be larger than 2.z ) = 2eioz2.z ) = 2eiaz2 = eio. we have Since a2 = e i a . a4 = a6 = . and when z tends to dD. is analytic in D . (b) . = 0. where a.419 (b) Prove that (Indiana) Solution.. which implies f ( z ) = e i a z 2 . (a) Let F ( z ) = f(z) + f(z).
which implies that f ( z ) has a fixed point in { z : IzI < 1).Z. prove that f ( z ) has a fixed (Rutgers) Solution. X > 1. Since G ( z )has only one simple zero in { z : IzI < 1))we conclude that f ( z ) .e” and G ( z )= X When z E I?.z . Hence f(z) takes value X exactly once in the right half plane. ?r Take a closed ?r I’ = { z = t + i y : t = 0. is the point necessarily real? Justify. where > 2A. Because R can be arbitrarily large.420 5410 I f f is analytic and lf(z)I point. 5411 Let f ( z ) = z + e . By RouchC’s theorem) F ( z ) and G ( z ) have the same number of zeros in { z : IzI < 1). R 5 y 5 R} u { z : Izl = R. Let F ( z ) = f(z) .z has one zero in { z : IzI < l}. IF(z) .z and G ( z )= z. .’ .< argz 5 }. IF(z) . it follows from RouchC’s theorem that ’ F ( z ) has exactly one zero inside I?. . Prove or disprove: f ( z ) takes the value X exactly once in the right halfplane. < 1 on { z : IzI 5 l}. (Iowa) Solution.G(z)l = le”l 5 1 < IG(z)I.G ( z ) /= lf(z)I < 1 = IG(z)l. If the answer is yes. Let R be a sufficiently large real number such that R curve I’ on the right halfplane. F ( z ) has exactly one zero in the right halfplane. . 2 2 Define F ( z ) = X . Since G ( z )has exactly one zero inside I. When IzI = 1.
Let C denote z the unit circle traversed in the counterclockwise sense. We have F ( x ) = X .z.42 1 Take z = x 2 0.e". Since F ( x ) is continuous and F(O) > 0. n . . the point z in the right halfplane such that f ( z ) = X is necessarily real.z2. +. (Iowa) Solution. Suppose that and Find the location of the zeros o f f in the open unit disc {z : II z < 1). Assume f is nonzero on the unit circle {z : II = 1). multiple zeros f(z) = !dz) j=1 nc. are the zeros of f ( z ) in {z : II . We have .x .Zj). x++w lim F ( z ) = m. Assume z1. such that F ( z 0 ) = 0. Then < l). In other words. which is a realvalued function of real variable x. 0 < xo < $00. where g(z) is analytic and has no zero in {z : Il z 5 1). there must exist 20. z being repeated. 5412 Suppose f is analytic in a region which contains the closed unit disc {z : ( z ( 5 1).
422 It follows from (1) that n = 2.z z ) g ( z ) .y = 0) U { z : II = R.2 = ki.z1)(z . f ( z ) has two zeros in the unit disk.O 5 argz 5 } z 2 R}. U{z=z+iy:z=O.e. Then for f(z) = ( z . and which show that unit disc. lr Choose a closed curve ' = I { z = z +iy. z1.O<yL . Hence z = 4 ~ 2 the only zeros of f(z) in the are 5413 (a) How many roots does this equation z4+z+5=0 have in the first quadrant? (b) How many of them have argument between f and $? (IndianaPurdue) Solution.O 5 z 5 R. i. (a) Let R be sufficiently large such that when IzI = R. Set f ( z ) = z4 +z +5 and g(z) = z4 + 5..
the numbers of the zeros of f and g inside I' are equal. (b) Let R be sufficiently large such that when la1 = R.423 It is obvious that If(.) .)l holds when z E I?. i) f (a$*) E {w : ?r E < argw < where E > 0 is very small. and f ( ~ E {w : o < argw < E } . Set f(z)=z4+z+5 and 9is approximately and r3 = { z : IzI = R. we know that ' . We also know that Ar. By RouchC's theorem. zero. 2 A It is easy t o see that Imf(z) > 0 when f(0) = 5. Since g has only one zero inside I?. Ap. where Ar. .argf(z) = Ar. z4 z + +5 = 0 has one root in the first quadrant.argz4 + Ar.mgf(z) denotes the change of argf(z) when z goes continuously from Re<' to R along I3 It is obvious that i '.< argz 4  A 5 }.argz4 = A.arg (1 + F) . while . f has also one zero inside I Noting that R can be arbitrarily large.g(z)I < Is(z.
z and z = z + i y . Let f(z) = sinz . it follows 3 from the above discussion that By the argument principle. then f ( z ) can be written as For any fixed natural number n. choose a positive number t >> logn and a closed contour I' = rl U I'2 U r3 U r4 in the counterclockwise sense. the number of the roots of f ( z ) = 0 inside equal to is Hence has exactly one root in the domain {z : f(z) = z4 z + 5 = 0 x 4 + < argz < }.424 is very small. 2 x 5414 Prove that the equation sin z = z has infinitely many solutions in C. Let I' = I'l U r2 U r is taken once counterclockwise. (Indiana) Solution. where .
Hence the winding number of f(r)around w = 0 is 1. and denote all the zeros by z1.z has one zero inside the contour I?. By Hadamard’s theorem.et ) . o 5w 5 1 . * .(et 2 .zg) and g( z) is a polynomial. : t = 2(n+ i f(I’3) ~ ) T . Remark. By the argument principle. f ( z ) = sinz . Assume that f ( z ) = sinz . with the direction downwards.t ) lies in the annulus starting from w = 2(n + 1)n + i in the counterclockwise sense. . Since n is arbitrarily chosen. we conclude that sinz = z has infinitely many solutions in c. multiple zeros being repeated. f ( z ) can be written as f(z) = e g ( z ) k ) ) P where P(Z> n = k=l . f(rz) {w = u + i v = with the direction upwards. This problem can also be proved by Hadamard’s theorem.z n .z has only finite zeros in (X. .2 7 ~ v ) 0) ~= with the direction from the right to the left.425 Then we consider the image of r under w = f (2): f(rl) {w = u + i~ = : 2(n +1 ) 5~u 5 . z z .
5415 (a) Let f be a nonconstant analytic function in the annulus (1 < IzI < 2) and suppose that If I = 5 on the boundary. where T03 which implies that g ( z ) must be a polynomial of degree 1. Let g(C) be a conformal map of {(=<+irl:$+$<l. By letting y + +oo and y + oo respectively.where rl = { z : IzI = 2) t is in the counterclockwise sense. . It might occur that f(rl) and f(r2)are two subarcs of {w : IwI = 5). (b) I f f is meromorphic in the annulus.argf(z) 2 1. (a) Let D = { z : 1 < I1 < 2) and d D = rlU r2. Because f is nonconstant analytic in D and If1 = 5 when z E dD. The following is a counterexample. which shows by the argument principle that f has at least two zeros in D. Let z = z+iy and z be fixed. the statement in (a) is not true. we obtain that Imu < 0 in the former case and that Imu > 0 in the latter case. Hence we have sin z . Show that f has at least two zeros. f has no zero in D.z = e a r + b p ( z ) . Hence &Ap. and I'2 = { z : IzI = 1) is in the clockwise sense.we know that both f(I'1) and f(rz) must be {w : lwl = 5) in the counterclockwise sense. In the latter case. This contradiction implies that sin z = z has infinite many solutions in C.argf(z) 2 1 and &Ar. and comparing the increasing order on both sides. is the statement in part (a) still true? (Stanfod) Solution. or both f(r1) and f ( r 2 ) are {w : IwI = 5) in the clockwise sense. In other words.426 It is obvious that f ( z ) is an entire function of order X = 1. (b) If f is meromorphic in D.
and aZn # 0.y = 01. where rl = { Z = TP: o 5 8 L ) . Then is a nonconstant meromorphic function in D with f has no zero in D. and suppose that Prove that P has exactly TI roots (counted with multiplicity) in the open upper half plane { z E 67 : Imz > 0). 5416 I f I = 5 when t E LID. and let P be a polynomial of exact degree 2n: P ( Z )= a0 + alz + a2z2 + + aZntzn. But Let n be a positive integer.. Suppose that there is no real number x such that P ( z ) = 0. Then the number of zeros of P ( z ) inside I is equal to ' It is already known that .427 onto {w : IwI > 5) with the normalization g(0) = M. > Take a closed contour I? = I'l U I'2 in the counterclockwise sense. and r2= { z = X + i y : T 5 T. Let T > 0 be sufficiently large such that when IzI = T . e . g'(0) > 0. 0 where each aj €67.. (Indiana) Solution.+azn_1z2"11. ( a 2 n ~ 2 n ( (ao+alz+.
428 We also have d log P(z) = G A r . +.+  z 2!z2 1 1 + .arg 2a (u2nz2n) =n we obtain that P has exactly n roots (counted with multiplicity) in the open upper half plane.+gcn. * . (a) Let 1 F(C) = f() = 1 + c c 1 1 1 + 3c2+ .e3+ * .) Solution.zn n! (a) What does the integral count? (b) What is the value of the integral for large n and fixed r? (c) What does this tell you about the zeros of f ( z ) for large n? (Courant Inst. 5417 Consider the function 1 1 1 f(z) = 1 + . . From . argP(z) 1 Note that and 1 Ap.
when n is sufficiently large. 5418 (a) Suppose that f ( z ) is analytic in the closed disk IzI 5 R. Show that this solution is given by the formula W (b) Show that. ICl=? then m > 0.429 we know that the negative of 11 represents the number of zeros of F(C) in (1 < :}. all the n zeros of f ( z ) are in { IzI 5 T } . When n is sufficiently large. C are equal. (b) When n + 00. lF(C) . (c) From the above discussion. simple solution z1 of the equation f(z) = w in 1(. if the integer n is sufficiently large. < R}. which is just the number of zeros of f ( z ) in 1(. we obtain {/(I < :} for fixed T and large n. and that there is a unique. > T}. . Let min leCl = m. we conclude that for any fixed T > 0. the equation has exactly one solution with ( z ( < 2. In other words. Since ec has no zero in a'.ec I < m I 1ecI for ICI = : which implies the numbers of zeros for F(C) and e in . there is no zero of f ( z ) in { IzI > T } . F(C) converges to ec uniformly in any compact subset of@.
E Hence fn(z) and g ( z ) have the same number of zeros in { IzI < 2 . Then f 1 f’(z) = [log(f(z) .(z) is a continuous realvalued function for 1 5 2 5 $. show that lim (zl.g ( z ) l = 11 5 z n = (1  5)” < 1 . Hence we have z?’ E (1. we choose E = 2 . ncc 2 (Courant Inst. $). z > 0 sufficiently small such that when E re.E L Ig(z>I.E } .€1. (a) Let f ( z ) .w = ( z . where Q ( z ) is analytic and has no zero in { IzI < R}. When n is sufficiently large. Ifn(z) .430 (c) If z1 is the solution in (b).1). Since E can be arbitrarily small. the equation z = 1 + (5)” has exactly one solution (denoted by z p ’ ) in 1(.) 1 1 Solution.zi)&(z).= . we have fn (1) < 0 and fn > 0.& ’ ( z ) = ’ w zzi Q(z)* + + Hence (b) Let fn(z) = z .1 and rE = For fixed large n. < 2).z ~ ) l o g & ( ~ ) ]= .w)]’ [log(z . It follows from (z) that . and the number is 1. (c) f.
f of order n and z = f is a zero of f(z) . Since 7 .nr2) 1 2 1 f(i) is homologous to zero (mod Q). (Indiana) Solution.) will not happen. Since f is analytic in R and bounded by lf(z)I < 1. f(rl) and 1 f(y2) will tend to either 20 = 5 or 2u = . the points z = =kf must be the removable singularities o f f .ny2 is not homologous to zero (mod R).$ of order m. Thus we obtain either f() 1 1 =2 2' f() =  1 2 1 2 . because otherwise. then f * y is not homologous to zero (mod R). f * yl and 1 f *y2 are also not homologous to zero (mod R). which is a contradiction.l)\{. and the directions of 7 and 7 2 are both in the counter1 clockwise sense. then f * (my1 . 2 2 Find all analytic functions f : S2 + R with the following property: if y is any cycle in R which is not homologous to zero (mod R). }. As E tends to zero. while my1 . f * 7 or f * y2 1 will be homologous to zero (mod R). Let where E > 0 is small. Hence we obtain f(*) = f2. for example. 1 1 1 f(1 = f() = 5. If.7 2 are not homologous to zero (mod R).43 1 which implies 5419 Let 1 1 R = D(O. Now we claim that the case that = f(. . 2 2 we assume that z = f is a zero of f(z) .
51 1 1.i f ( z ) 2 1. = i)= 1 we consider the function f(z)+i G(z) = 1 + if(.l) and satisfies IF(z)I 5 1.t . 2 2 1 1 we consider the function F(z)= fk). 1 2 f() = .t z 1 which is analytic in D(0. which implies that f(z) = z. Thus we conclude that the functions which satisfy the requirements of the problem are f(z) = z and f(z) = 2.zz which is also analytic in D(0.l) and satisfies IG(z)I 5 1. which implies that f ( z ) = z. It follows from G ( .)   z . It follows from F ( that F ( z ) 1. 1.432 or f() 1 2 = .i ) = 1 that G(z) 1. = .
which implies It follows that .z . on I1 = T . . Prove that lim an/an+l = zo. In other words. n+w (Indiana) Solution. the Taylor expansion of F ( z ) a t z = 0 has a radius of convergence larger than T . which is a simple pole. and define F(z)= f ( z ) A .433 SECTION 5 SERIES AND NORMAL FAMILIES 5501 Let 00 nO = have a radius of convergence T and let the function f ( z ) t o which it converges t have exactly one singular point zo. Hence the power series F(z) = n=O A Cantn. Assume that the residue of f ( z ) at zo is A .zo z Then F ( z ) is analytic on { z : 1zJ5 T } .zo 00 is convergent a t z = zo.
C $ is convergent and C n>l n>l E C with la1 = 1. we have 1 1Z . by Dirichlet’s criterion we know that converge. (1) Let a = eit. c. which shows that .a ) for such a.( < 1).434 and and we obtain lim = zo. Differentiating term by term. (2) Show that this series converges to log(1 . (2) Let f(z) =  n>l c n> 1  zn n 1. 2 ~ )then .=c an n>l n>l cos nt + i sin nt n For t E ( 0 . (Minnesota) Solution. t E ( 0 . 2 ~ we have ) and Because both C for 1 # a n>l tends to zero monotonically. n+co % + I an 5502 (1) Show that the series n> 1 is convergent for 1 # a E C with la/ = 1.
C n an n> = = C. I n= 1 Show that the series converges to a holomorphic function on the open unit disk centered at origin..z ) . ( Columbia) Solution. for (a1 < 1.e.e i t ) = log(1 . Suppose the series : is + . we obtain f(z) = log(1.a). z = Teat where 0 < T < 1. First of all. we prove the following proposition: If the radius of convergence of n =O is equal to 1 and an 2 0 for all n. Let Q = eit. . Prove that the boundary of the disk is the natural boundary of the function. then for fixed 2 E ( 0 . i. z = 1 is a regular point of f. 0 < t < 2a.= n> 1 r+l eint limCr+1n>l (Teit)n 1 n lim log(1 T e i t ) = log(1. Assume the proposition is false. 5503 Consider a power series O01 C nz n .435 Integrating both sides on the above identity. It follows from Abel's limit theorem that . then a = 1 is a singular point of f(z). l ) there exists a small real number 6 > 0 such that the power series expansion of f at point 3 is convergent at z = 1 6.
. c . For any natural numbers p and q . z = 1 is a singular point of F ( z ) . and hence the order of summation can be changed. .k + k=On=k is convergent at z = 1 6 .. &I n.". we assert that when z = 1 6 . Noting that when z = 1 6 the right side in the above identity is a convergent double series with positive terms.Ic + 1 )an(z . Now we return to the power series It follows from that the radius of convergence of " 1 zn! n n=l is equal to 1. a ( .1) .z ) k z n . By the above proposition.436 Thus 00 w w x b k ( z z)k = k=O n(n . n=O which contradicts the statement that the radius of convergence of n=O is equal to 1. + + n=O W k=O .
we have [).and if If(z)I = 1.p r i : p . it follows from f(0) = 0 and If(z)l < 1 that 5 IzI for all z E U .1 for some z # 0. Let 0 < r < 1. In other cases. we conclude that the unit circle {IzI = 1) is the natural boundary of F ( z ) ..437 Since z = 1 is a singular point of " 1 I zn. prove that either g ( z ) = 0 or g ( z ) = z . In the case when f ( z ) = eiaz. q = 1.. which implies that f ( z ) = z and g ( z ) = t. 5504 z E Suppose f is analytic in U = 1(.. Remark.. By the above discussion. (IndianaPurdue) Solution. If the sequence { f n } is defined by composition < 1 for all and f n ( z >4 g(z) for all z E U .2. Then . Since f n ( z ) is convergent. Since the set { e . the boundary of the unit disk is also the natural boundary of the function c C 00 $zn! although the series is absolutely n=l and uniformly convergent on the closure of the unit disk. z = e?4. then f(z) = einz where a is a real number. By Schwarz's lemma. we obtain a = 0.} is dense on {It1= l}. f n ( z ) = einaz.i is a singular aq point of F ( z ) . n n=p it is also a singular point of F ( z e 9 " ' ) . < 1) with f(0) = 0 and If(z)I U .(fI for all z E U . In other words.
f(z>I< m 5 lf(z>I holds on {z : Iz . Then m > 0.. we obtain g ( z ) = 0 for all z E U . . Hence fn(z) converges to zero uniformly in{lzl 5 arbitrarily chosen. such that f(z) # 0 when z E {Z : 0 < IZ 5 T } C D. (UC. which is a contradiction to the assumption that fn(z) # 0 for all z E D.z0I < T } . Ifn(z>. Suppose f is pot identically zero. . As {fn} converges to f(z) uniformly on compact subsets of D . (b) If each fn is onetoone on D .(fI on { z : Iz . we have .438 For all z E { IzI 5 T } .z01 = T}.z0I = T } . Since zo is a zero of f . we know that for sufficiently large n. Irvine) Solution. It follows from RouchC’s theorem that fn and f have the same number of zeros in { z : Iz . show that f is either constant or onetoone on D . we know from Weierstrass’ theorem that f is analytic on D ... but has a zero point zo E D . T}.z0I < T } . there exists T > 0. Since the zeros of a nonzero analytic function are isolated. fn must have a zero in { z : Iz . (a) Prove that if fn(z) # 0 for all n 2 1 and z E D . then either f is identically zero in D or f(z) # 0 for all z E D. (a) First of all. {fn}r=l ZOI Let m be the minimum value of /). Since 0 < T < 1 is 5505 Let be a sequence of analytic functions in a domain D which converges uniformaly on compact subsets of D t o a function f on D .
such that and f ( z ) . In other words.z2( < T } .) .a ) holds on that f n (z) . which implies fn(z{) = fn(zi) ( : # 24).zll = T or Iz .439 (b) Suppose f is not a constant.z11 < T } and { z : (z .a1 on { z : IZ .a and f(z) . and is not onetoone on D.a have the same number of zeros in {z : ) z.z 2 ) = T } . Then there exist zl. Let rn be the minimum value of If(. With the same reason as in (a).u = 0.)I = Ifn(z) . I(fn(z) .f(z)I < m L If(. Then m > 0. This is a contradiction to the assumption that f n is onetoone on 2 D.z 2 ( = T } .z2 E D (z1 # zz)..z l ( 5 T } U {z : 0 < Iz . when n is sufficiently large.a # 0 in {z : 0 < ( z .221 < T } respectively. Choose T > 0 sufficiently small. such that f(z1) = f(z2) (denote it by u ) .al { z : Iz . such that fn(Z{) .) . and let { f n } be a sequence of analytic automorphisms of D such that n+m lim f n ( a ) = b E 8D for some point u E D.a = 0 and fn(zi) . 5506 Let D c a! be a bounded domain. (Indiana) . Prove that n+m lim f n ( z ) = b for every z E D.221 5 T } .211 <T} and z: E { z : (z . It follows from RouchC’s theorem E { z : Iz .(f(z).zll = T or (z . there exists 2 : .
If {fn(ao)} does not converge t o b. f(0) = 0. 5507 Which of the following families are normal.Re{g} > 0. C E S}. For any two fixed points z. then m = min{lf(z) . 1 1 on { z : Iz . diam f ( D ) 5 2) (b) B = { g : g is analytic in D. Without loss of generality. which is a contradiction to the fact that fnk does not assume the value b E d D in D because f n k is an automorphism of D . which obviously satisfies the conditions that f(z) is analytic in D and f(0) = 0. Since { f n k } converges uniformly to f on { z : ( z . Since { f n ( z ) }is a normal family. nP .m n+ca lim fn(ao) = bo # b.CI : z .g(O) = l. f(z) is a nonconstant analytic function of D .a1 < T } .b has no zero in { z : 0 < I z . when Ic is sufficiently large. fnk(z) . Because f(u) # f ( a o ) .it follows from f(0) = 0 and diam f ( D ) 5 2 that lf(z)I 5 2. (a) 3 = {f : f is analytic in D . Take a0 E D . Then there exists a subsequence { f n k } converging uniformly in compact subsets of D to f(z). (a) For any f E F.C E D . and which is compact? Justify your answers. we assume lim fn(u) = b 6 d D .a1 = T } > 0. ao # a.a = T } . we have . Let T be sufficiently small such that f ( z ) .440 Solution.bl : )z . Let {fn} be a sequence of functions in F. diam g ( D ) 2 1). By Rouche’s theorem. which shows that 7 is normal.a = T } . there is a subsequence { f n k ( z ) ) converging uniformly on compact subsets of D to f(z). (Indiana) S o htion.b has zero(s) in { z : Iz .u [ 5 T } c D . there exists a subsequence of {fn(ao)} converging to bo # b. Here the diameter of a set S is diam S = sup{lz .
= {w : I ..(z) # 0 for all z E D . hence f(z) E 7 . ( D ) 5 2.f(0l 52.( .} with respect to w = 1. then the constant is el because ~ ( 0 = lim ) egn(') nvw = e1. Hence we can define an analytic function g ( z ) = log G(z). a} i. We choose a compact subset K c D such that z. First we can choose a sequence of functions g n ( z ) in Q as follows: g n ( z ) is a conformal mapping of D onto R. which shows that family Q is normal.} converges to R = {w : ( w .31 < 1) U {w : IImwl < . Then for Gn(z) = egn('). such a mapping gn exists and is unique. Hence there exists a subsequence {Gnk} converging uniformly in compact subsets of D to a function G(z) which is either a constant or a nonconstant analytic function in D . since G.which shows that F is also compact. 5508 Suppose that 1 5 p < 00 and c 2 0 is a real number. } be any sequence of functions in Q.1 < 1 which is called the kernel of {R.441 because diam f n . where the singlevalued branch is chosen by g(0) = . 1 < Rew < 3) w 1 w 4 1 1 n satisfying gn(0) = 1. gL(0) > 0. and it is obvious that g n satisfies all the conditions required by the family Q.) . by Caratheodory's theorem. we obtain diam f ( D ) 5 2. we have IGn(z)l < 1.log G(0) = 1. Because the domain sequence {a. It follows from the uniform convergence of {f. If G(z) is a constant. and we conclude that g n. which shows that is not compact. But family Q is not compact.1 < } U {w : I . { g n ( z ) } converges uniformly in compact subsets of D to g ( z ) which is a conformal mapping of D onto 52. By the Riemann mapping theorem. (b) Let { g . by Hurwitz's theorem. c E D can be arbitrarily chosen. Let 7 be the set of all analytic functions f on { IzI < 1) such that .C E K .) = logGn. Since diam g(D) = g ( z ) does not belong to the family 0. if G(z) is nonconstant analytic. Since z .(z) converges uniformly in compact subsets of D to g ( z ) . we have G(z) # 0 for all z E D .} on K that If(.
If it is not the case. The contradiction implies that 3 is a normal family. while the right side of the inequality is a constant.201 Cauchy integral formula. P q 1 1 Then As n + co. Show that 3 is a normal family. then there exist zn E D .442 sup O<r<l /'^ o If(reioe)lPde 5 c. It suffices to prove that the functions in 3 are uniformly bounded on every compact set of {zI < 1).f n E F such that zn + zo E D and fn(Zn) + 00. < r.lzol = 3r. B y Hence where +=l. the left side of the above inequality tends to infinity. Izn . (Illinois) Solution. . Then when n is sufficiently large. Let 1 . We prove the assertion by contradiction.
and apply this to find the explicit (b) Find the Taylor's expansion of g(w). series expansion of the root of the equation z3+ 32 . f(z) # 0 for 0 < IzI < r 5 R. Solution. Show that 1 tf'(t)dt g(w)= 2?ri f(t)w J. .443 5500 (a) Let f be holomorphic for 121 < R and satisfy f(0) = 0. which implies that g(w)is holomorphic in {w : II w < m}. and that z = g(w)is the unique solution of that tends to zero with w. Hence tf'(t)dt g(w)= GJ. Let C be the circle IzI = p where p < T .w = 0 that tends t o zero with w. f'(0) # 0.w f 1 =5(&J. (a) It follows from IwI <m= rp lf(pe")l n that when t E C. define a holomorphic function of w for 1I w < m = m p If(pe")l. ( t ).p n=O t f ' ( t ) dt) w n .
Because lwl the winding number < m = min If(pe”)l.w). 9 @.w)]’ = [log(t . we assert that the unique solution g(w) tends to zero together with w . (b) Let f(z) = z3 32. we know that f(z) = w has a unique solution in { z : 1x1 < p } .w have the same number of zeros in { z : IzI < p } . and * f’(t) = pog(f(t) . As the constant term in the Taylor expansion of g(w) is which is obviously zero.444 Let r be the image of C under f where C is the circle { z : IzI = p } taken once counterclockwise. Since z = 0 is the only simple zero of f in { z : IzI < p } . which shows that f(z) and f ( z ) .zi)Q(t) where Q(t) is analytic and has no zero in {t : I1 t < p } .O) = n(r.z1) f(t) .w Hence 1 Q’(t> + log Q(t)]’ = + t z i Q(t) which shows that g(w) is just the unique solution of f(z) = w.w = (t . then f(t). + then where . Denote the unique solution by z l .
2.. By MittagLeffler's theorem. * . . when IzI 5 N . Prove in detail that your function has all the required properties.2 (:)k2 k + '2kk .2 . . 2N2 Hence converges uniformly in { IzI In addition. .. So f ( z ) is analytic in 1.( < N } \ { .. . 5 N } to a function 2N1 n=l which is analytic in { IzI <N}. 2k t 5510 Find an explicit formula for a meromorphic function f whose only singularities are simple poles at 1.N + 1 } .. n 2 2N. z2 is a meromorphic function whose only singularities in { IzI < N } are simple poles at z = 1...:) . .445 After some computation. 2. 2. (Illinois) Solution. it is obvious that f ( z ) has all the required properties of the problem. a n d z = 1. + .l . we obtain a2k1 a2k = 0 and =  zl 1 33k2 32klt2k1 [ 2 '2k . . ... .N 1 with residue n at z = n. we construct For any natural number N . . Because N can be chosen arbitrarily large.1 (5)k'] dt (3ck2c".with residue n a t z = n.3 .N+lareitssimple poles with residue n at z = n.
(a) No. it follows from the holds for all z E { Izl < 2 ) . there exists N > 0 such that when n > N . give an explicit formula for g. Multiply both sides by )zI2.2. explain why not. The contradiction follows by taking z = 0 in the inequality.446 5511 (a) Does there exist a sequence of polynomials {P..} such that P n ( z ) i uniformly on the annulus 1 < IzI < 2? If Yes.3. give an explicit formula for the Pn .we have 5 $1 a).1 is an analytic function in maximum modulus principle that 1. (b) Yes. (Illinois) Solution. (b) Does there exist an entire function g whose zeroset is {+( 1+ i) : n = 0. IPn(z) < E holds for all z E (1 < Izl < 2).( < 2}. if No. )? If Yes.1. For any R > 0. The function g can be chosen as where an = f i ( 1 i). It is easy to see that . If there exists asequence of polynomials {P.1 < 2 } . 5 Because z 2 P n ( z ). then for any E E (0. if No.} such that P n ( z ) + uniformly on (1 < 1. . explain why not. Then when n 2N . let (z( R and choose N 5 + > R2.
We prove the assertion by contradiction. If for each positive number n there exists an entire function fn such that then for 15 IzI 5 2. and 5 e 1 ~ ~ ~ ( zeRefm(z) < 2e > l =  .447 Because C n=N 00 5 converges. and its zeros in { z : 1z1< R} are 0. Define Fn(z) = efn(')... (q 5512 State whether the following statement is True or False. Hence g(z) is analytic in { z : IzI < R}.. which implies m is analytic in { z : IzI < R } .. we see that g(z) is a n entire function with the required zereset. we have 1 5 Refn(z) 5 1 + l o g 2 . For each positive integer n there exists an entire function fn such that (Indiana) Solution.al.ak . and prove your assertion. we know that is analytic in {z : IzI < R}. Since R can be arbitrarily large. Then Fn(z) are entire functions with no zeros.1 5 IC < $). False.a2.
For example. which implies that F ( z ) = cyz with = 1 in { z : I1 < 2). This is a 5513 Let G = D\(1.0]. where D = { z : IzI < 1). (b) Choose where n > 2.(z)I 5 2e for IzI 5 2. IF. F ( z ) cannot be identically zero. (a) zi is defined by eil0gz. (a) Give a singlevalued definition for zi in G.48 4 for 15 IzI 5 2. t contradiction to the fact that F ( z ) has no zero in { z : IzI < 2). C Kn+lr and nrco lim K. Because the complement of K. In domain G.(z)} is a normal family in { z : ( z ( < 2). singlevalued branch of logz can be chosen. is connected and contains z = 00. But we have for 1 5 IzI < 2. we know by Runge's theorem that there exists a sequence of polynomials which converges . such that nrm lim P. = G . 5 (Indiana) Solution. a singlevalued branch of zi in G can be defined by argzlo<2<l= 0. By the maximum modulus principle. and by Hurwitz's theorem F ( z ) has no zero in { z : 1x1 < 2). Then K. Since IFn(z)l 2 for 1 5 IzI 5 2. Hence {F. (b) Why should there exist a sequence of polynomials P.(z) = z* for all t in G? (c) Can the polynomials be chosen so that there exists a constant M with IPn(z)l M for all z E G and all n? Justify your answer. and there exists a subsequence {Fnk(z)} converging locally uniformly to an analytic function F ( z ) in { z : IzI < 2).
(c) No. IPn(z)l 5 M for all z E D and all n. (c) Develop a cot az in a Laurent series about the origin directly and by use of (b). Since Pn(z) converges t o zi in G. = 1.} converges to zi uniformly on n compact subsets of G.) is . then because Pn(z) are continuous on D. It is obvious impossible. In other words. 5514 (a) Prove that (b) Use this to show that acotaz = Z l o o + c2z 22 .. so the contradiction is obtained. and there exists a subsequence P. (a) Let f(z) = G. Hence {Pn(Z). It follows that {Pn(z)} is a normal family in D .Now we consider (2n)2.2..449 uniformly on K . hence zi = f ( z ) for z E G. If there exists M with IPn(z)l5 M for all z E G and all n. .n2 ' n=l Justify your steps. a2 1 The singular part of f at z = n (n = 0. * . fl. the series  . with enough terms to find the values of n=l C 5 and C 00 00 1 n=l (Hamad) Solution. . f2. which implies that z' can be extended to a singlevalued analytic function in D. we can find a polynomial Pn(z) such that for all z E K.. to z i . ( z ) which converges uniformly on compact subsets of D to an analytic function f ( z ) in D .
we restrict z in the strip { z :0 < Rez 5 1). It follows from the convergence of C 5 n=2N 00 and C 5 that C ~~ M cQ n=m & is analytic in Because N can be arbitrarily large. It is obvious that lim f ( z ) = 0. . Imt+f00 As the convergence of is uniform for lIm4 2 1.450 For any natural number N and la1 5 N. As f(z) and are both periodic functions with period equal to 1. Let Then g ( z ) is an entire function. we obtain the result that 00 1 n=cc is a meromorphic function which has the same singulaxities as f(z). Hence g ( z ) is a bounded entire function. the limit of the series for Imy + fco can be obtained by taking the limit in each term and the limit is also zero. holds for n 2 2N and n n=2N 5 2N.
. It is obvious that the constant must be zero. For z2 n=l . Differentiating both sides of the above identity. Thus we obtain the identity (b) Let F ( z ) = 7rctg7rz. Hence we obtain 7rcot?fz= +c1 " 22 z2 n=l . we know that is a meromorphic function which has the same singularities as F ( z ) . it follows from the fact that F ( z ) and 1 .n2 + c. we consider Now 22 With similar discussion to that in (a). + n=l c 22 .n2 + Then G ( z ) is an entire function. The singular part of F at z = n (n = 0.)is the series 1 " n=l  A."+ E n 22 n=l are both odd functions that c = 0.45 1 which implies that g ( z ) is a constant.G'(z) Comparing this identity with (l). . Let 22 1 " F(z)= .&2. we obtain T2 1 " n=l 1 1 . have G'(z) = 0 which implies that G = c we ( c is a constant).n2' ..G(z). f l .
. = ( z .2 with f (zj) = uj if and only if n (c) Given a sequence of complex numbers z1. f of degree . rz x . does there exist an entire function f with f(zj) = aj? Can you write this function down?  (Hamad) .. Take z = 0. Zn be distinct complex numbers. (a) Show that (b) Show that there exists a polynomial of degree 5 n .452 (c) The Laurent expansion of 7rcot 7rz about the origin is j cot l .z2.tl) 9 * .. we can also obtain 5515 Let z1.23 .z 1 z 7r2 3 7r4 . . .( z ..2 and g(z) Let f and g be polynomials.. 45 (3) It follows from (2) and (3) that around the origin. we obtain After differentiating (4) on both sides. 5 n .zn). such that lzn 1 + 00...
2..2 with f(zj) = uj (j = If u l . Since . while f ( z ) is of degree 5 n . ."'. and the coefficient of zn' is 1.z.an are n complex numbers such that n we construct the function f ( z ) by For each j .q ) ( z .Z. . .l ) ( z . z ) g( z .( z .2. Because g ( z ) = ( z .Z2. (a) Take R sufficiently large such that Zl.Z n ) is of degree n. ( z .z 2 ) * .Z j + l ) .( z . . < RS.z 1 ) .453 Solution. u 2 . j=1 (b) If f ( z ) is a polynomial of degree 1.) is a polynomial of degree n .. . then by (a).1. . we have 5 n . E 1(. we obtain ?#=O..zj . dz dz = lim = 0.(2 .. n).. .z j .
(c) For the given sequence z1. which implies that f ( z ) is an entire function satisfying the required condition.1 I +00.  where 7. Then we define  a. Hence > 0.I > 2R 2 00 1 u n ( a ) converges uni5 R when n 2 N .I + 00. we can construct an entire function g ( z ) with simple zeros z l ..454 the coefficient of znl of f(z) is zero.n). while for k #j. Since R can be formly for all IzI 5 R. .( holds for all IzI arbitrarily large. f ( z ) is an entire function. z2. In other words. Because Iz.2. Because lim z+rj . In other words.2. f ( z ) satisfies the condition f(zj) = aj ( j = 1.Zj) . is chosen such that when IzI 5 k$. f(z) is a polynomial of degree 5 n . . such that .’ajz j ) g ( dz) ~ ( z . n=l < R}. uk(zn) = 0. It is easy to see that while for k # n.aj. so that f(z) is analytic in 1.. by the Weierstrass theorem about the canonical product of entire functions. z2. . there exists N > 0 such that Iun(z>l 5 Iz. for any R when n 2 N .
Part VI Partial Differential Equations .
.
.j)&=l be a real matrix. ) is the dot product of vectors in R" and wn is the area of the unit sphere in En.457 SECTION 1 GENERAL THEORY 6101 a) Let A = (a. a) It is not difficult to verify that aijxjxjdx = 0. Show that ( 2 . b) Show that u E C:(R") implies (Iowa) Solution. =  . Ax)dx = Wn L<l n(n + 2) trace A. lxl<l Vi # j and Therefore.trace A Wn 1 n n(n + 2) trace A. . Here (.
T' = 0 if and only if T(q5')= 0.pdx = 1. the conclusion is true for u E Ci(IRn). Vq5 E C r ( B ) . i.e. It is easy to verify that a function q5 in Cr(nZ) is the derivative of a function in C r ( B )if and only if q5dx = 0. R (Indiana) Solution. g d x = 0.. we get immediately As Cr(lRn)is dense in Hi(IRn). show that there is a number a such that T(q5)= / aq5dx for all q5 E C r ( l R ) . Show that T = const.458 b) First let u E Cr(IRn). Then it is easy to verify that for any q E C r ( I R ) 5 satisfies the condition ? J. 6102 Let T be a distribution on IR and suppose that T' = 0 on IR.By applying Green's formula. . IR Take a function p E C r ( l R )such that J.
Therefore.a / 2 E L2(Rn) if s > n/2. 14+w ( cincinnat2) Solution. Let {fk} be a sequence in D((0. Then we have a compact interval [a. defined above. as Ic + 00.b] c ( 0 . It is clear that u. l).bl. and support your answer. is a linear functional in D((0. l ) such that SUPPfk c [a. u E H s ( R n ) if and only if (1 lt12)'/2G E L 2 ( R n ) . And it is clear that (1 1[ 1 2) . vk: . In fact.l)) such that fk + 0 in D ( ( 0 . Show that lim u ( z ) = 0. Then we get immediately + + 6104 Let u be defined for f E D((0. We show first that 6 E L 1 ( B n )if u E H s ( E n ) with s > n/2.459 Hence T($) 0 and = 6103 Let u E Hs(En) with s > n/2.l)). l)). (Indiana) Solution.l)) by Determine whether u is a distribution on (0. we have 6 E L 1 ( R n ) .
and is a distribution on (0.< a. o uniformly on [u.#dxdy . u. Let uy(x7 y. u#. ) 5 belong to WIJ’(B)? (Iowa) Solution.dxdy = J.l). 1 N Then we have Therefore.uy E L1(B). that is J.460 and jp). XY 2 ( . + y2)3/2 * It is clear that us. b] as for any nonnegative integer n. 6105 Let B = {(x.l)). + oo. Let N be a positive integer such that . in the sense of distributions. y) I x2 For which p 2 1 does the function + Y2 < 1. u is continuous in D ( ( 0 .
x)ds for 1 5 p < 2. = aB. ' BY It is not difficult to verify that + J. Derive a necessary and sufficient condition in terms of v +. Assume furthermore that v lies in both Co(?&) and C0((n2)(but not necessarily Co((IRn)!!) with v+(xo) lim v(x). v is a distribution solution of (*) in En and only if if . u~cos(?~'.dzdy =  lc ux4dxdy + 1. 0 in the above inequality. Therefore. And we can verify that for which means that u @ W1yP(B) p = 2. Letting E . v. Denote by v = v(x) the unit normal to C pointing into R 2 . vu(xo) lim v(x) xxo xDO +En1 XEn* for each xo E C . Suppose v : R" + R" solves the equation divv(x) = b(x) for x E R" . u E W1ip(B)for 1 5 p < 2... Let RE = B\BE for any E C r ( B ) . 6106 Q .JB uy#dxdy r for all radius E E Cr(B). Similarly L u = 2ly.. (Indiana) Solution. we get &u = u.and v for ZI to be a distribution solution of (*) on all of R". and Let C be a smooth hypersurface dividing R" into two disjoint open regions R2.C ( *I in the classical (C')sense where b is a continuous function.u4. Then we have < 1. we denote the ball centred at the origin with the By and S.461 and r JB u+ydxdy = . B.
By Green’s formula..462  b#dx. (Indiana) Solution. 2. Recall the definition of the curl operator in two dimensions acting on a vector field U(z.. Therefore v is a distribution solution t o (*) on all of B* if and only if (w+ .v) .. V# E C. 2).y). V+dx = .v(x.y) = (u(z.y)): vxuu~2Lyvu. and U1 and U2 are (different) constant vectors. (b) Suppose U takes the form 2 2 1 2 where 5 1 and a are open sets such that a U n = B2.Y = o 6107 on C. describe as completely as possible the curve r.J. (a) Let u. V#dx = IRn V# E CF(lRn). #dx.I’ z d a l n a 5 2 2 is a smooth curve. vds = 0. Hence the equation (1) is equivalent to #(v+ . J. Then curl U = 0 .... (a) Give a definition of what it means for a (not necessarily continuous) vector field U to have curl zero in the sense of distributions.v) uds + lRn divv .V) . If U has curl zero in lR2 in the sense of distributions.v E Lloc(B2).(IRn). we have Ln 2. V+dx =  +(v+ . V#dx  J.v .
and 6 is the Dirac distribution defined by 6 4 = 4(0). v< E {< : < R}. there exists a positive constant R such that {< : P(i<) = 0) c {< : 1 < R).6 E S. we have J. where Then curl U = 0 if and only if u1.%)4dz + (v1 .vz). As the set {< : p ( i < ) = 0) is bounded. Integrating by parts. (Indiana) Solution. w1. and assume that the set {z : p ( i z ) = 0) is bounded. u2 and 712 are constants. where S’(W) denotes the space of tempered distributions on B”. 6108 Let p be a polynomial in n variables. Therefore.(ul .(lR”. Here S is the Schwartz class of rapidly decreasing functions. Prove that there exists a tempered distribution E on lRn such that p ( D ) E . U2 = (uz. that and E . I’ is a straight line defined by (u1 . v4 E C.v2)dy = 0. WI). 4  Construct a function X R E Cp(lRn) such that XR(<) = 1. Set It is clear that E S ’ ( R n )n Cm(Rn).v2)4dy = 0.463 in the sense of distributions is defined as (b) Let 171 = (u1.u2)dz + (v1 .
G ( t ) must be finite linear combinations of the derivatives of the Dirac delta function S(<) at { 0 } . From the transformed equation. P(D)= C.Transforming the equation. l4lN where a = (a1. Suppose ~(i<) E ~. i. .) and the C. Therefore..D" bllm where a denotes a multiindex Q = (a1.' ( Z ) E S'(nZn).S(")(<).. (Here S' denotes the space of tempered distributions on an. 6109 s.e.(i<>" o for all E # . are constants. . Let G(E) be the Fourier transform of u.) is a multiindex and a.6 E where E = F . are constants. prove that u must be a polynomial.. Therefore.a. .(0).a.(nZ"). Hence is a polynomial..e.a2.~ 2 . . we get  C < P ( g ) G ( [ )= 0. Let P ( D ) be an m t h order linear constant coefficient partial differential operator on En.1 E C. we can see that suppG = { 0).^(€I = + c a.) (Indiana) Solution. . I4<m If u E S' satisfies P ( D ) u = 0. i.464 It is easy to see that p ( i € ) Z ( € ). we get p(D)E.
. exaf4tdx = 1 Vt For any > 0. 4 E D. (a) It is wellknown that 5. (c) Calculate rigorously the derivative of log 1x1 and express the result in terms of your answer to (b). define the Cauchy Principle Value and show that it is in D'). Let supp4. Then we have + 0 .465 6110 in D'.( 1.e. suppose that q5n in D. c ( . a ) Vn.we have Then it is not difficult to verify that Hence t+O+ lim  & ex2/4t 6 .a . Further. as (b) We define the distribution follows: 5 by its Cauchy Principle Value &: The linearity of the functional Vp: is obvious. (d) Calculate the Fourier transform of the distribution (Indiana) Solution. (b) Show how to define as a distribution (i.
' ( c ) For any 4 E D . Therefore.DcIu(c) 2 €mn .we have Here.466 Therefore. 2 (d) By F ( f ) we denote the Fourier transform of a tempered distribution f For any 4 E S. we have applied the integral equality 'J00t 'IF sin(zt)dt = 1 for z > 0. we obtain 6111 a) Let Lu(2) = bI<m c a. is a distribution in D .:.we have V p i Hence &log 1 1 = V.
a) A distribution E E D’(nZn) is called a fundamental solution of the differential operator L if L E ( 2 ) = S(z). 2 is a fundamental solution of the wave operator W ( u )= utt . and a is a multiindex. 00 (f.. Simplify your result to obtain D’Alembert’s solution of this problem. then the operator L has at most one fundamental solution in S‘. b) We need only to show that In fact . u ( z . = #J(z.)1.u. What is a fundamental solution of L? When is it unique? b) If I+denotes the Heaviside step function Show that 1 F ( z . Let L* be the formal adjoint of L defined by L*V(C) = lallm c (l)~aba(aaV(z)). on E 2 . If L*S is dense in S.g. < 2 < 00. #J sufficiently nice) in terms of the above fundamental solution. where S(z) is the Dirac function. O ) = g(z). 0) = f(z).467 where the abs are constants. (Indiana) Solution.t). t > 0 00 < 2 < 00 U t ( 2 . where S is the space of rapidly decreasing functions. c) Express the solution of the Cauchy problem utt . t ) = I+(t)l+(t..u.
468 The solution of the Cauchy problem is given by Therefore. the solution of the original Cauchy problem can be obtained as .
we have ztt+ i t 1 4 6= 0. q t .469 This is just the D’Alembert’s formula when q5 2 0. Assume that there is a constant C such that (Indiana) Solution.0) = w.0) = 0. we have Set 77 = tit in the integral in the right side of the above equality. By solving the above transformed problem. By applying Fourier transform to the problem. 6112 Let u be the solution of the initial problem where summation is understood in the pde. and 4 is a C” function on IR3 with compact support. where G and denote the Fourier transforms of u and q5 with respect to z. respectively. qt. It is reduced to . it yields that 3 %t) = Therefore.
given in the problem. K E E3.470 This implies that In fact. consider the solution u of the SchrGdinger Equation Show that lu(z. if I&(O)l = 2a # 0. (Indiana) Solution. m 4 And when t > 5 . then there exists a positive constant r such that I l2 a. It can be verified that the Poisson's formula for the Cauchy problem of the heat equation applies to the Schr6dinger equation. 6113 Given E S (rapidly decreasing functions) over B1.1 < r.t)I2dz = I&0124 Here 3 denotes the Fourier Transform of 4. So we have And then .we have This contradicts the condition satisfied by the solution u.
we get Here the Fourier transform of 4 is defined by .47 1 Set x = 2&.
zz*(=p Then we have u2(x)dx wnu2(xo)Jlo pn'dp w 2 = u n ( x 0)T n . There exists a point xo E En such that u ( x o ) # 0.pn1U2(x0). Schwarz's inequality gives that is JT u 2 ( x ) d S z >_ w. n .472 SECTION 2 ELLIPTIC EQUATIONS 6201 Let u $ 0 satisfy u E C2(Rn). we have where w . = 0 on Au R". Applying the meanvalue property for the harmonic function u. (Indiana) Solution. Show that does not exist. is the surface area of the unit sphere in En.
as E + 0. an). the relation uAddx = 0 holds for all 4 E C. u. an).. Show that u is then harmonic in R. Set where C is a constant such that For x E R. Therefore u is harmonic in R. is harmonic in K if E < u. 6202 Let u E Co(R) be weakly harmonic in an open set R. we have where This implies that for any compact set K dist(K.473 for any T > 0. Therefore.(R). i. It is wellknown that c R. (Iowa) Solution. +u uniformly in K . The conclusion follows as T + 00.e. set PE(Y) = EP Then it is easy to see that provided E 2Y ( ) 7 P4Y) E C.(f% < dist(z. .
we get we have the same estimation. 1 2 5 M. B4. y) be a locally bounded function. (iii) u is harmonic in z > 0.y)I > 0 be given. y ) is continuous at (zo. .y) E .. These estimations imply that f ( z . (Indiana) . yo) E B4and R such that lf(z. V(z. Then the conclusion that f ( z . as a function on y).e. harmonic in z E lR2 and continuous in y E BIZ. Prove that u E 0 on z 2 0.Show that f is continuous in (2. Then there exists a constant M > 0 lR4. .yo) E B4follows immediately. (Indiana) Solution. 6204 Suppose u is a function defined on nZn such that (i) u is bounded and continuous on the halfspace xC (ii) u = 0 on z = 0. we have By Green's formula. For z. . y ) is continuous at z = z0 uniformly with respect to y near yo.z0l2 + Iy .474 6203 Let f (2. Let (zo. i. Applying the meanvalue theorem to the harmonic function z.y0l2 < 2R2. 2 0 of En.
6205 Let u(x) be C 2 on the halfspace B = {x : x. If u. = 0).. Thanks to the Liouville’s theorem. = {x : x. = . Redefine the function u in the halfspace x. ~ ~I‘ 1 )([1. < 0 as a n odd function of u(x‘.475 Solution. u is bounded and harmonic in B”. n 1.e. and let g(x) be a compactly supported. * .) (Indiana) Solution.bounded. Therefore.xnl). we get u s 0 immediately. > 0) and continuous on y the boundary alR.. is the surface measure of the unit sphere in B”. = u(x’. we can show that the unique bounded solution to the problem (3) is given by e. where X’ = ( ~ 1 .. By the uniqueness of the bounded solution to the Dirichlet problem (3). we can verify that the redefined function u is harmonic in any ball centred a t the origin in lRn. satisfies j = l .1) and w.**. It is not difficult to verify from the above formula that and .2n) 2: . (Warning: you must T justify any assumptions of regularity you make on the solution u. By using the uniqueness theorem of the Dirichlet problem and the Poisson’s formula of the Dirichlet problem in a ball. are bounded in show that its “tangential” derivatives magnitude by maxlVgl on all of B = {x : xn > 0).x~). . C1function on alR: = {x : x. VX. < O where x’ = (xl. = 0 ) .(.* .
It is easy to see that lXl uAudxdy = 0.lY I < 1 Integrating the above integral by parts.476 Therefore. _ . ( Cincinnati) Solution. < 1. au au = 0 ax ay for IyI = 1 has at most one solution. We need only prove that the problem Au = 0 for 1x1 < 1. we have uAudxdy . . (Here u = u(x. we have which are just we want to show. 6206 Show that the problem Au = 1 for 1x1 < 1. au . < 1. IyI u = 0 for 1x1 = 1. and is a classical solution).au = 0 for IyI = 1 ax ay has the unique solution u 0.. < 1. IyI u = 0 for 1x1 = 1. y) is a function of two variables x and y.
1)) Then we get u f 0 immediately. 6207 Let R be a bounded normal domain in lR3. Then it is clear that u can be chosen to be realvalued. Show that r (IndianaPurdue) Solution. we obtain (Vu12dw= a2 IuI2dv. . (a) Multiplying the equation in the problem (*) by the complex conjugate of the solution u.u2(1. (b) Suppose a1 # a2 are such that nontrivial real solutions u1 and u2 satisfy (*). which implies that a is a real number.u y 1 . and integrating the resulting equation in R. and suppose nontrivial solution exists to CY is such that a { V2u = a 2 u in R.1) . and that u can be chosen to be realvalued.477 = (u2(1.1) . u=O on a R . 1) 1 2 + u2(1. (a) Show that a is a real number.
one has u 5 h in B. Let u be a continuous function on a. and integrating the resulting equations in R. (i) For every x E R and r > 0 such that the ball B ( x .r ) cc R one has (Here un denotes the surface area of the unit sphere in Enand B ( z . Vu2dv = a: and From the above integral equalities.) (ii) For every ball B cc fl and every harmonic function h E C o ( B )such that u 5 h on d B . . we can show that u can not take its maximum in a unless u is a constant.r ) centered a t x with radius r satisfies B ( z . we get immediately which implies the conclusion of the problem. Let u be subhamonic in the sense of (i). By a standard argument.Prove the equivalance of these two notions of “subharmonic”. Multiplying the V2u1 = a9u1 in R and the equation satisfied by u 2 v2u2 = aiu2 in by u2 and can obtain u1. we Vul . (Indiunu) Solution. respectively.478 (b) Suppose that equation satisfied by u1 u 1 and u2 are chosen to be realvalued.r ) cc R means that the closure of B is contained within R. 6208 Let fl c B“ be an open set.
a) The conclusion is clear. let u be subhamonic in the sense of (ii).)l where d ( z ) = dist(z. a) Show that is harmonic in R for each i. 8 0 ) . We define the function h as follows: Ah = 0 in B ( z . :Y E lVU(. h=u on a B ( z . Conversely. r ) . we have au n (z) =dXi in the wn r n where w.YI = 71. { From the definition of subhamonicity in the sense of (ii). By using the Green's formula. if R is bounded and n E R. the above equality can be rewritten as . SUP~lU(Y)l 1% : 3 : n . /n is the volume of the unit ball. 6209 . b) Applying the meanvalue theorem for the harmonic function ball B . Hence that 5 0 on 8 B implies that w 5 0 in B. L SUP{lU(Y)l 4s) 0) (Indiana) Solution.479 Let h be the function given in (ii). hence w.h is also subharmonic in the sense of (i). b) Let B c a. c) Show that. r ) . and the meanvalue theorem. Then w = u . we have This completes the proof.Let R E B" be an open set. is the surface area of the unit sphere in En. where B is the open ball centered at x of radius r . Show 2 that IV4z)l L. and let u be harmonic in R and continuous in R.
By B ( z o .2'1. ds2) be a fixed positive constsant. Then the spherically symmetric solutions depending only on T satisfy (TU) d2 dr2 + k2TU = 0. Let R < dist(zO. First we look for the fundamental solutions of the equation. Let u E C1(n)n C2(R) solve A u + k 2 u = 0 in R (k > 0).O. it is not difficult to verify that & where T.cos(kr).R) we denote the ball centered at zo with radius R. there are two linearly independent solutions: 1 1 u = . . Applying the above integral formula in . = ( z . (Indiana) Solution. 6210 Let R C lR3 be bounded and open with smooth boundary. T r Let u be a solution to the equation A u + k2u = 0. By the fundamental solution cos k r . we have c) The conclusion can be obtained from the result of the part b) immediately. For this ordinary differential equation. Let T = 1x1.480 And therefore. Derive an appropriate meanvalue property for the solution.sin(kr).
z0) = cot(kR) sin( kr. For this function. E aB(zO.R2) satisfy AU = u + 1.zo). Let u E c2(.). = 0.481 the domain B ( z o .R). 2 g(z. it is easy to verify that Subtracting the equation (2) from the equation (l).) rxox satisfies the above conditions.R) This is a meanvalue theorem for equation A u 6211 + k2u = 0.+ k2)g(z. zo) 1 47Frxo.. Prove that its spherical mean.cot(kR)1 sin(kr. R).rrRsin(kR) J BB(xo.o. which satisfies (A. Txox  4.o. It is easy to see that the function g(z. u(z)dS.R).o.we obtain u(x0) = a . z E B ( z o . defined by .zO)= cos(kT. we get Now we look for a function g(z.
482 and v(0) = u(O). W(T) # 0.y) I x2 + y2 < 1).also satisfies ( l ) . B ( 0 . Therefore + The proof is completed. for z radius T . we get Then it follows that = r(w(r) 1). Prove that the problem Au= in Jm B. 6212 Let B = {(z. Set T = 1 1 Then we have 2. 1 u=O onaB . dSy the element of arc length. Solution. and Applying Green's formula. r )denotes the ball of (Indiana) 1 =21r L B ( 0 . Here. l ) u(rw)dS.
= h[(gradu) (gradv) v)1  + Xuvldz.and g E L2(dR) be given. g and 80 are sufficiently smooth.483 has exactly one weak solution u in HA(B). we get + J. And denote the corresponding norm in W't2((n)by 11 ( ( 1 . Then by applying Cauchy inequality and the trace theorem in Sobolev's spaces. w E L2(B) and f = a v . (b) Will the conclusion of part (a) be valid if X = O? Explain. (Iowa) Solution. and (1) L ( g r a d u ) . This implies that f E H . Hence the problem admits a unique weak solution u E Hip).a w + ax dy in the sense of distributions (see the solution to the problem 6105).  . ( Cincinnati) Solution. Show that the problem (1) has a unique solution u E W1!2(R). write problem (1) in the classical form. Let X v= d 3 w = &5qj5 and f = 1 &Gp' It is easy to see that v . fvdx 1. 6213 (a) Let R be a bounded domain in lRn with smooth boundary f E L 2 ( f i ) . gvda v v E w'q(n) Vu. (c) Soppose that f . (a) Define the inner product in W172(f2) by (u.v E W'y2((n).' ( B ) . (gradv) + Xuvdx = where X > 0. Consider the problem aQ.
it is easy to see that the conclusion of the part (a) is false if X = 0 even for smooth f and g. there exists a function u E W'?'(R) such that F ( v ) = (u. the classical form of the problem (1) is Xu = f in R. Prove the existence of a unique solution of that problem. where IR" 3 2 = (21.= u Inn(z. azn du+ .1(R). Then u is a solution to the problem (1). WX'. Thanks to the Riesz theorem. Denote u+ = u Inn~X. (b) By the classical form of the problem. ( Cincinnati) . an + 6214 Let R be a unit ball in IR" centered at zero..= g on dR. =o} . g and dR are sufficiently smooth.o} and u. and assume that U* are regular enough. Jnn {X .). (c) Suppose that f . Au dU . z n ) = (x'.<o}. Integrating by parts. Therefore.1 on R n {zn = 0).l .. Show that Au* = 0 and that du ax. z n . F(v)= fvdx +lngvdo is a bounded linear functional in W'?'((n). V#€H.44 8 where C is a constant. The uniqueness is clear. v v E W'J(R).Vq5dx = a . V' ). Therefore. Consider the following problem Find u E Ht(R) such that lVu. x.
V 4 d x = where R+ = R n {z. > 0) and R. V4 E H. This proves the existence of the solution to the problem..4dx  l Au. . According to the Riesz theorem. ax. < 0). Write the equation in the following form Vus V 4 d x Vu. + L  J. nn{ x . auax. = 0).a x .485 Solution. Define the scalar product in Hi(R) by By the Cauchy inequality and the trace theory in Sobolev spaces.=o} 4d~'. The uniqueness of the solution to the problem is clear. ""'> / dx' = au+ ax. Integrating the above  l+ 1 Au+ . The proof is completed.n{X. Mx'. we have + 1.= R equation by parts.(Q). we can get I Cll4ll1.. =o} 4dx'. 4 d x + nn{ xn=o} (E. Assume that u* are regular enough. there exists 1 (%4)1 u E Hi(R) such that = / nn{ I=o} . n { x . : where I(. v4 E m Q ) . .11 is the norm in Hi(R). =1 on Q n {x. From the above equation. we get immediately that AU* = o in a&.
in R.’”(R).’”(R) and Now we have Vu V4dx =  J.u. From the above inequality and (I). it holds that IV+(2dx= 0.e. 0). the proposition is true.VU . Let an d(z) = max{l Then it can be verified that .486 6215 J.e. b) Let u E W1?’((n)satisfy (I) above. By the PoincarC inequality with the above estimate. show that 2 inf u an (essinf) ( Cincinnati) Solution. in R. in R .e. Show that u 2 0 a. 4 E W. a. 4 2 0. a) This is a corollary of the conclusion in b) of this problem. an Assume that inf u = I > 00.V 4 d ~2 0 inf u n V 4 E W. b) If inf u = 00. we obtain Hence 4 =0 This implies that u 2 I a. IV412dx 5 0.
. respectively. E W.n11w. that such Awn + u .nw. Au. Vw. and then integrating the resulting equations in R. both equations above can be written as Vu. in R. and w. w strongly in L2(R).'72((n). + o strongly in ~ ~ ( 0 ) . as n ( Cincinnati) Solution. w E W292(R)n W. we get By using Green's formula and noting that u. + = w. + +oo.'12(R). Therefore..487 6216 Consider functions u. Multiplying the first equation and the second equation by u. = 0. Prove that '11.11i2 Then we have  = 0. we have . Multiplying the first equation by v.v. w.} is bounded in L2(R) and w.dx . and integrating it on R.. we obtain {u..
+0 strongly in L’(R). J. Prove that there exist a function u E C(R) satisfying Au u = f and a subsequence {ukj} such that ukj converges uniformly to u on each compact subset of R. uk(x)2dx 5 M for k = 1..Au. Av. Vv.2. .IIVu. = 0. finally (4) By similar consideration. we can get . respectively. (Indiana) + . and Av.~ ~ A .AunAvndx + nllVun11i2 IIVUn11ia Hence + nllVvn11iz = k VW. Vundx.dx  . Multiply the first equation by Av.Suppose uk E cm(n) satisfies Auk +uk = f and has the property that for some positive number M . we can get J.488 By noting (1) and (2). ~ ~ ~ ~ v Vu.dx.dx. 6217 Assume R c Rn is a bounded open set with smooth boundary and let f E cF(R).Vw Vv.. By noting (3) and (4). . This gives us the conclusion of the problem.Il$ = LwAu... it follows from the above equality that Av.12(s2). This implies that {un} is bounded in W. it follows from above equality that Multiplying the first equation and the second equation by Au. and then integrating the resulting equations.dx =  J.
we have + h(T)vk(z) = 1 vk(Y)dSy. aB(X. and therefore in D'(R). r ) we denote the ball with the radius T (< R) centered at z. we find that Integrating the above equality from Here R can be chosen so small that Ivk(z)I 1 h(r)dr # 0. v x E 0 1 . see the solution to the problem 6210). It is clear that v k satisfies A V k + V k = 0 in R and Hereafter. As { u k } i bounded in L2(n). Set vk =w k + z k . Then Au + u = f in the sense of the distribution. Now we show the second conclusion.489 Solution. we have u E Cm(R). it holds that : (1) 5 c.u1. Therefore. by C we denote various constants. " ' . s there exists a subsequence { u k j } such that u k j converges to u weakly in L 2 ( R ) . For any z E 21. by B ( z . Set V k = u k . Let R l be a subdomain of R such that 2 CC R and R = dist(dQR1) 1 fixed. k = 1 . where h(r) is a continuous function of 0 to R with respect to T . By the regularity theorem of the elliptic equations.r) T. 2 . By the meanvalue theorem for the equation Au u = 0 (for the case n = 3. where W k and z k are defined by and .
we prove the second conclusion of the problem immediately.e. VEd Define the norm of w in Hi(R) as .a . 5 c..respectively. (a) Let {un} be the minimizing sequence of I(w) in A. wed (b) Show that L V U V ( W u)dx 2 0 VV E A. .lR. i. k = 1. where hl. we can obtain from the above estimation IVzk(x)) 5 C in any compact subset of R1. (a) Show that if A # 8. . V k = 1. * * . h2 : R + .2. 6218 Let R be a bounded Lipschits domain in B". and A = {v E Hi(R) : h l 5 v 5 h2 a. there exists u E A satisfying I(u) = minI(v). we can get from the estimation (1) lVWk(Z)( <_ It is clear that Igix 1(1 Z.2. And by applying the meanvalue property for the harmonic solutions. are given smooth functions. vx € ill. n+w lim I ( u n )= inf I(w).e. Then by using the AscoliArzela theorem. (Iowa) Solution.. k) a 1 c. By applying the solution formula of the Dirichlet's problem for the Poisson's equation. in R}.
 6219 Let R c IR”. .u)dx 2 0.e. Therefore. we get u E A and that by the weak lower semicontinuity of IIwII1.u)) V(u+ t(w . Therefore we have + F’(0) 2 0. of {u.such that u + u a. be a bounded domain with smooth boundary. .}.denoted still by {u.u))dz.such that u u . for simplicity we still denote it by {u.}. V(u t ( w . This implies that I(u) = min I ( w ) .~(R) u t u in L ~ ( R ) . * Then F ( t ) has the minimum at t = 0. . and in H. Let r + v E A.}. 0 5 t 5 1. w F ( t ) = J. This implies r Vu V(v . (b) It is easy to verify that u t ( v .}. there exists a subsequence Iull.u)E A. in R.491 Then we get the boundedness of { I n l } Therefore. The proof is completed. n > 1. wed The conclusion of (a) is proved. Let u E C1@) be harmonic in R. And there exists a subsequence of {u.
you may cite. it holds that @2(2) vx E s2% .u)= 0.201 I u ( x ). @I(. we have A(Q2 . From the maximum principle. any standard maximum principles you are using. Suppose there exist functions @ I . Prove that ma IVuJ= rnax ( V u ( . x n an (Note: In both parts a and b of this question. () Similarly. without proof.u ) 2 0 and @2 in R . . @ 2 E C 2 ( Qn C1(t) such that ) A@I 5 0. Prove that m3x l v u l s rnax rnax lVaI1. For the function @2 .u. u ( z )5 Therefore. n an i.2 . The conclusion is an immediate corollary of the following inequality: n i. vz E s2.).u(z0) 1% . @I =u = on dQ. %.u) = rnax(a2 . it holds that sup(a.) b..u = 0 on 8Q.492 a. a. we have 5 u(z).e. (Indiana) Solution.2 0 ) This inequality implies the conclusion of the part b. rnax l n an an AQ2 2 0. ( ~ m 2 l ) .@2(z0) ) x .j=l b.
(b) Show that if X > 0 is sufficiently small. Au(z') 2 0. u = 0 on dR. it is clear that Au(zo) 5 0. Suppose u E C2((n)is a solution of the equation Au = u3 with the property that lVu(z)I 5 1 for each z E dR. (a) Prove that for any solution u of (l). . Hence w 5 1 on 6221 Consider (a nonlinear) equation Au = Xu2(1 . where X > 0 is a parameter.0 5 u 5 1. Set w = IVu12. (Indiana) Solution.493 6220 Let R be an open subset of Rn. the only solution of (1) is u (1) 0.e. then there exists xo E R such that u takes maximum at to and u ( 2 ) > 1. Prove that IVu(z)I 5 1 for all x E R. do. w takes its maximum on (n a t some point on dR implies that w 5 1 in R. (a) If u 2 1 is false. and dR sufficiently smooth. i. ~ ~ 2 U Therefore.u) in R.. It is easy to see that w E C2(R)n C'((n) and n Aw = 2 3. For the maximum point xo.j=l u2izj +~ u ~ I V 0. ( Cincinnati) Solution.
= 0 on dR. we have u E 0 immediately.u3(1 By using the conclusion of (a). 1vul2dz = x J. As Xu2(1 . (IndianaPurdue) Solution.494 This contradicts the assumption that x u y 1 . Assume u E C4((n). 8222 Let R be a bounded normal domain in IRn with smooth boundary dQ.U)(ZO) < 0. Integrating by parts and taking the boundary conditions into account. dn is the trivial one u E 0. Let ?i' be the exterior unit normal on dQ. Show that the only solution of the boundary value problem for the biharmonic equation: A(Au)=O i n Q u = 0 on dR. we have r 0 = JnA(Au)uda . it holds that Applying PoincarC inequality. If X < Cl. Hence u 2 0. u can not take minimum in R unless u f 0. dU . we can get J. Hence u 5 1. (b) Multiplying the equation by u and integrating the resulting equality on R.u) 2 0. we obtain from the above inequality where C is a constant.
af2. we get u 0 . u is harmonic in f2.495 = lA~1~dx. Taking into account u = 0 on immediately. Therefore.
(Indiana) Solution. continuous. 03 lf(”)I2d2 < 0O. Show that there exists a constant C such that for all 00 < 2 < 00. Then we get lu(2. (0O IR1: u(2. J 2 m m By the Cauchy inequality.t)l c I t1/4‘ .496 SECTION 3 PARABOLIC EQUATIONS 6301 Consider the Cauchy problem for the Heat Operator in ut = u. t > 0.O) = f ( 2 ) < 2 < 0O.W d y .t > 0) (0O < 2 < m).t) = . where f is bounded.r nf ( y ) e . The solution to the problem is given by the Poisson’s formula u(2. and satisfies J_. we have Set y = 2 &q immediately + in the second integral of the above inequality..
For any 5 CIIVU(*. m)) is a solution of the equation with u = 0 on a R x [O. we have . Prove that $ ( t ) + 0 as t 4 +m.t)2dz. Multiplying the equation by u and then integrating the resulting equation on R.497 6302 Let R c Bn be an open set with smooth boundary and suppose u E Cm(R x [0. we obtain Solving this differential inequality. we can get The PoincarC inequality gives IIU(*. (Indiana) Solution.t)ll:z(n) where C > 0 is a constant. t)l&(n)? E > 0. it holds that Taking E = 1 / C . 00). Assume that Let $ ( t )= 1 n u(z.
t > 0) u(z. Transforming the following Cauchy problem with the initial data S(z): ut = u .t > O ) . ut where f is bounded and continuous. b) State a maximum principle for this problem. Why is the Tychanov nonuniqueness example possible? c) Suppose in addition that f E L 1 ( B ) .. d) Show how to solve the problem > 0 . < 2 < 00. a) Using the Fourier Transform. ( = 6(2).It is not difficult to verify that if l/f[.t > 0) 4 2 .t) be the Fourier transform of u ( z . construct explicitly a fundamental solution for this problem. (2 by using a) and an appropriate extension of f. a) Let G(<. (00 < 2 < 00. .t)ll&n) . (Indiana) Solution. t ) with respect to z.Show that ta. and write down the solution u. 4 2 . .) 0 as t + +m. This completes the proof.0) = f(z). 6303 Consider the Cauchy problem = u. . lim u(2.O) = f ( z ) ( 2 > 0) U ( 0 .t ) = 0.0) we get . t ) = 0 (t > 0) ut = u .
t) = 4(y)eWcly. < 0. Then the inverse Fourier transform of 6 gives the fundamental solution t o the From this fundamental solution.and satisfies the heat equation in lR x (O. Then it holds that XER J inf f ( z ) 5 u ( z . t ) to the general Cauchy problem with the initial data f ( z ) u(z.T]. 4 2 .) = { f(x>. d) Defined an odd function 4(z) as follows: 4(.t ) = etzt.T]. Then by solving the following Cauchy problem: ut = u x x (cm < 2 < cm. 2n21 b) The maximum principle can be stated i ~ follows: Suppose that u ( z . XER The Techanov nonuniqueness example is possible for some unbounded solutions. c) The conclusion is clear. we can get the explicit representation for the solution u ( z .t > O ) .499 The solution to the transformed problem is given by G(<. we get the solution to the initialboundary problem 6304 Consider the problem . f(z). 2 2 > 0. O )= 4(z).T]with the initial data f ( 2 ) . t ) E lR x (O.t ) s is bounded and continuous on lR x [O. t ) 5 sup f(z) for ( z .
T). we can find that Then by using the Cauchy inequality. Find a bound on ': 1 Ju.t u(2. From the Poisson's formula for the Cauchy problem of the heat equation.(z.. (z E IR.(z.t )12dzdt = I'L: tu. t ) u t ( 2 . . and integrating the resulting equations in x (0.t)I2dz t > 0. . we have 1' 1: tlut(2.O) = uo(z) E > 0) L2(E).Ut = u. in (Indiana) Solution. it holds that We can also get the result by the method of energy integrals.t ) d z d t . Multiplying the heat equation by u and tut respectively. we can obtain and Adding the double of the second equality to the first one. we have And therefore.
we have A/m 2 w oo ~ u ( ~ . 2 ) u. t> . t ) with z E R and t < 2 Let B1 denote the open linesegment ' . can From the above two equalities.ut + a(2. ut E C o ( R U B1) be a solution of ~ %x .. we get an estimation 6305 For the indicated domain R (interior of the parabola) let RT denote the points ( z .t ) u = f(2.501 otherwise an approach to the function U O ( X ) by a sequence of functions in Cr(En) be used. forming the top of ~ R T and let B = BRT\B~.Let u ( z .T)I2dz +T Therefore. t ) E C o ( n ~with . T ) ~ ~Ld_zlu.(z..
(Here a < 0 should be assumed in RT u B1). G O .502 with a.f E Co(D). then every nonconstant solution u assumes its negative minimum (if one exists) on B2.tO) 0. Show that if a < 0... (IndianaPurdue) .6.t0) 0 < 5 and u.t) + (0 5 z 5 A and t > 0).to)u(zo. If a < 0. > + This contradicts f(zo.we can say that every nonconstant solution assumes its positive maximum (if there exists one) on Ba. Use the sequence { f n ( z ) = 2 sin ( y z ) } and an appropriate space of conn+l tinuous functions to decide whether this problem is wellposed. Hence u. ut = 0 u(0.ut(z0. 6306 Consider the boundaryvalue problem B.C. I. u.. UP(. f 2 0 in RT.C.t)= 0. t o ) E RT U B1.(zO. u(z. = 0.O) = f(z).. f 5 0 in RT.tO) 0. Suppose that u assumes its negative minimum at ( .tO) a(zO.t o )5 0.tO) 2 0.1 Solution.(2 0 .. Then we have u(zO. f 2 0 in RT? (Iowa) Fig. What can you say if a < 0.t 0 ) . ut(z0.
(You should be able to find u explicitly.t ) we denote the Fourier transform of u with respect to 2. By setting u = f. 0) az.r].) (Indiana) Solution. 6307 In Probability Theory one encounters the OrnsteinUhlenbeck process. By G(<. t)ll + 00 as n + 00.) = iF(u. leave your result in the form of a onedimensional integral. = uo(2). It is easy to see that F(zu.503 Solution. (x E t .) d at 2 (<G) a at = u I. if not. This implies that the problem is not wellposed in the space of continuous functions with the norm defined above.r. u u(z.x . When n is even integer. we can find the solution u to the problem with the initial data f(z) = fn(z) .t > 0) Assume uo E S and find u. in which the particles of Brownian motion are subjected to an elastic force. . f n ( z ) is the eigenfunction of the corresponding twopoint boundary value problem.(. at .(z)T. is clear that It But for any fixed t > 0 IIu. One is required to solve u = u .(t).  a. for all f E C[O.
. In fact. t ) = et3(e2'. we can find a convenient transformation of variables to the problem. It is not difficult to verify that F'(iio(ett)) = e . we can obtain . and let h be a fixed point IR".504 Then the transformed equation and initial condition are in the following form qt. From the above solution formula. Then the problem is transformed into the following 7 =0 : 21 = Uo(t). 6308 Let L be the usual heat operator L = a let at (: . 2 t = etx. we get the solution to the problem Remark. u(<.+of ax. .1) . The above solution formula can be obtained easily from the Poisson's formula for the Cauchy problem of the heat equation. Solving the above Cauchy problem for the first order linear partial differential equation. I +. a2 1 ' 4 E D(nZ"). set T 1 = (I . Consider the problem .l)caiio(et<).e ..t u o ( e . Therefore.0) = G o ( t ) .t z ) and Fl(eli(e2tl)€2) = 1  d m e 2(e 2 .2 t ) .
t > 0. t ) to the problem This is just the conclusion of the problem. the above problem we obtain { O(<.O)= 4(2). u(2. < E En. a.505 ( L u ) ( z . a.h . t ) . . t 2 E En L which admids the following solution qt. t ) of the above problem in terms of the initial data.O) = m. Transforming (. i& + ([(I2 . for an unknown function u ( z .t)llL~(R")< e(1R2)tl141~L2(Rn). Show then that the solution u satisfies Il"('. Let G( t ) be the Fourier transform of u with respect to 2. (Indiana) Solution.> 0. but you need not prove that it represents a smooth solution.t ) . Derive an explicit representation for a solution u ( z . 2 E B" 2 E B". t )= u(2 . (You should prove that your u is welldefined. ( l€12exP( The inverse Fourier transform of G gives us the explicit representation for the solution u ( z . Assume that 4 is the Fourier transform of a function which vanishes in the ball of radius R 2 0 centered a t the origin.) b. € ) ) t .e"n€)G = 0.t ) = &(()(y i h .
we get from above euualitv E(t) 2 dt I d where E ( t ) = & u 2 d x ..which solves the equation x U+. (Iowa) Solution. oo)). Prove that we can find a constant C such that E(1) = I u 2 ( x .  AU = U 4 (heat conduction with heat loss due to radiation) with the boundary condition u Ian= 0. + 1 IVu12dz =  1 u5dz.506 6309 Let R c IR" be a bounded domain and assume u ( x . we get from (1) I d E(t) 2 dt 5 JRJ+Eqt).e. Multiplying both sides of the equation by u and integrating on f .1 ) d x 5 C regardless of the initial value u ( z . The Hijlder inequality gives i.we have 2 utudx  Au u d x =   I u5dx. t ) 2 0 is a function with u E C2(o [O. By using the Green's formula and noting the boundary condition. .0). Using above estimation.
h < 1.T ) x (O.T) x with g.T]such that u takes the maximum a t ( z o . = < 0. 6310 Let satisfy ut u3 u = h.t o ) 2 1. u3(z0.tO) = (u3 sin(u))(zo. if u( zo. r ) x [O.sin 1 < 0.tO) 1 .t.A+u)(zo. (Indiun u) Solution. Prove that maxu < 1 on B ( 0 .to) This contradicts (1). then there exists (zo.tO) 1. LO.to) E B(0. h continuous and g. TI.u. Therefore. we have Ef(t) It follows that 2 E+(O) + 3(R(. we must have But from the equation we get (ut .to) 0 2 and Az(zo.T ) x (0. B(0. t o ) and u(z0.T]. If the conclusion of the problem is false. to) > 1. TI. It is clear that ut(zo. This completes the proof. + if u(zO.to) 5 0. u= 9. B ( O . + + sin(u) = A.to) sinu(zo. . T ) x (0) W O .507 The above inequality can be written as Integrating the above inequality from 0 t o t .
(n) for all t E (0.to) 2 0 and Au(zo.O) = f ( z ) 2 E a.T]. udx  u4dx.yn) 5 + IlfllLl(n). t > 0: u(z. (c) Prove that the solution t o this boundary value problem is unique. u=o z E d R .to) 0. In fact. where R is a smooth bounded domain in En. TI.and integrating the resulting equation on 0. t > 0. (Indiana) Solution. T)) Co(nx [O. (b) First we show that u can not take the positive maximum in R x (O. t o )E R x (O. the following bound also holds: I I 4 4 ( n .u3 2 E R .to) 5 0. IIUIIL(n)(t) L llfllL(n) for all t E (0. By applying Green's formula. (d) Show that for f 6 L4(R) n H1(R). we get Au . then we must have ut(zo. TI.508 6311 Let u E C2(R x (0. Hence (ut . (a) Prove that IIUllLz(n)(t)5 IlfllL. 2 .T]) a solution to the problem n be { (b) Prove that ~t = AU .Au)(xO. ( t ) + II~llLl(n)(t) Ilfll. if u take the positive maximum a t ( z o . (a) Multiplying the equation by u.T]. we obtain from above integral equality Then we get the conclusion immediately.
J.1 3 ~ ~ I V u (. 1 + u ~ )2 w . IVul'dx) 5 0. we get 1 utdx = J.Tl. By a similar deduction as done in (a). vt E (0.u'dx. w ( z . In a similar way.u'dx) dt 4 5 J.T].This contradicts the inequality u3(xO. t > 0. we have dt ( / . x E R. Au . we can get Il4lL. This implies that u1 = u2.d u'dx <_ 0. and w = u1 . vt E ( 0 7 1 for above problem. lVu12dx) = .and the uniqueness of the solution is proved. we can obtain f and (a Lu'dx) + dt (1 J.(u: + ~ w = 0.utdx .l u i d x 5 0 d = . we have Il. (d) Multiplying the equation by ut and u3 respectively. 0) = 0. t > 0. Adding the both inequalities above.tO) < 0. ~ x 5 (1 J.u3utdx and utu3 = J. Therefore. we can show that u can not take the negative minimum in R x (O. .u2. u3dx . x E $2. By applying Green's formula to the first terms in the right side of the both equalities above.llL(s2)(t> L llfllL(ct).(ct)(t) 50.J.Au . ~ x E a R . and integrating the resulting equations. u 4 d x + J. The w solves the following problem: { Wt = AW. (c) Let u1 and 212 be both solutions t o the boundary value problem.
510 This differential inequality.v and Then from the given differential inequalities. combining with (a) if necessary. . t )R x ( 0 . respectively. 6312 For each p E En. Assume A E C ' ( B n ) . ~ ) E and wt 2 aij(Vw)vzizj for ( z . implies the conclusion of (d). This is the result we need. co). we find w 5 o on I x [0. we can obtain a differential inequality for w as follow: n n ij=l k=l By applying the maximum principle t o the above differential inequality. that u and w are continuous on and satisfy the condition u 5 w on the parabolic boundary n x [O. co). co) (Indiana) Solution. t ) E R x (0. Let R C B" be a bounded open set. furthermore. A(p) be a real positive definite n x n matrix with let entries aij(p). Set w = u . Prove (from first principles) the following comparison principle: Let u and v be C2 solutions t o the inequalities ~t 5 aij(Vu)uzizj for ( ~ . Suppose.
j=l n a i j ( z . it must hold that Wt 5 0. t ) = min u(z.j=l n for every t E R~.t o )< min u.{ o ) . T I .T]. T and E I u(zo. u ) w Z i > 0. i. Q . t . ~ u ) w . t ) Q 0x P. First we prove that if w E C2(ax [O. z ) are continuous functions of their arguments and assume that C ~ i j ( z . m)).m)) satisfies wt n n C a i j ( z . ~C~b i ( t .j=l i=l at ( q t )E 0 x ( O . z . t )can not be the minimum point of w in R x (O.T]). vu)uZixj  C b i ( z . p ) and b i ( z . (Indiana) Solution. u.TI where Q = (R x (0)) U (dR x [O. Now suppose that u takes the minimum a t (%'. t . wxi = 0 and which contradicts (1). Suppose u satisfies the inequality ut  C i.t .t . Prove that for any T >0 min u(z. u . a t the minimum point (z. let u E C2(nx [0. z .511 6313 For R c Rnbounded and open. where X is a positive function. t ) . t > 0. t . . t . (1) then ( z .to) R x (0.u)uxi o >_ i=l n for z E R. p ) t i tLj x ( z . z . z . p ) 1 t 1 2> o i.In fact. t . Assume u i j ( z .
Then takes its minimum at some point in R x (O. for the function w we have E >O is sufficiently small. t .if However. w = u . .u.u. t .512 Take a > 0 large enough so that a u l l ( z .&eaxl for ( z . This is in contradiction with w having minimum point in R x (0.V u ) > 0. T ] .T]. TI. Vu) + b l ( z . t )E s2 x [ o .
O) Show that the LP norm of .t) = sin(cl<lt).. we have I and Then g E L1(IR") implies that 2tt K O ) = 0. ?([) is bounded in E n .c2AU = 0. > n is bounded by a constant (Indiana) Solution. it holds with a positive constant M that Set 7 = c<t in the integrand.^(. +. Let G ( < . Therefore. %(t. = $(tL 0) 1 ?(<I I cK + c21<)22 = 0.513 SECTION 4 HYPERBOLIC EQUATIONS 6401 Assume that g is smooth and in L 1 ( R f l ) . Then we get . Transforming the equation and the initial conditions. G(<.t)at time t for p times t lnlp. u(2.O)= 0.n arbitrary. = 9(z). t ) be the Fourier transform (with respect to z) of the solution to the ndimensional wave equation: Utt .
= Gl(€). . utt + l€12G(t..t>= Go(€) cos(ltlt) + sin( I€lt) It I . q€.t) = 0. 6402 Consider the Cauchy problem for the wave equation Assume U O . 0) G(€. = Go(€).. i.we have . (Iowa) Solution. .. E (a) Use the Fouriertransform t o show that the total energy at time t is constant in time. we get < %(€) qt.514 p > n ensures the convergence of the integral.e. U ~ S. Show that . (b) Let Eo denote the constant energy in (a). E ( t ) Eo. The above estimation gives the conclusion of the problem. Solving the above initial problem of the ordinary differential equation. 0) where = (€l.zn).[n). (a) Taking the Fouriertransforms of the wave equation and of the initial conditions with respect to the variable 2 = (zl.
Ut(2. 6403 Consider the initial problem for the wave equation in three space variables: u. 2 E B3.O)= qqz). .. Utt = 0.55 1 Integrating the above equation on En and noting that we obtain Then the Plancherel theorem gives us Combining this with (a)..+ u. . z E R3. + %.. = $(z). we finish the proof of (b)..t> 0.. u(z...O) z E B3.
(IndianaPurdue) Solution. for 15 1x1 5 2.2 3 > 0).(Y) = 0.23) E lR3. find u ( 0 . where k is a constant. (Your answer should be explicit. (4 (b) By applying the domain of dependence for the solutions t o the wave equation. . we can verify that u vanishes when t = 10 for 1 1 5 7 or 1 1 2 13.22. 0. O<t<l. we can show that the above equality holds when (c) It is easy to verify that 2 13.and $(z) = { 0. when 1 1 5 7. it holds that 2 IyI 2 3 for all y E {y : I . find the set of points in lR3 where you are sure that u vanishes when t = 10.21 = 10). y Therefore J q5(Y)dS = IyzI= 10 J II. vanish outside a ball of radius 3 centered a t the origin. 0.516 (a) Write down the formula for the solution of the problem. 11 2 lyzl=lO Similarly. (b) If q5 and II. 6404 Let R be the upperhalf space in lRn R = {(21. 0. 2 2 In fact. for 2 < 121. 2 k . no integrals). t ) for all t 2 0. (c) If 4 vanishes everywhere in R3. for 1 1 < 1. 2<t.
the domain of dependence for the value u ( x . ) is given by the formula for the Cauchy t problem of the equation: (b) Define *(x) and @(x) as both odd functions with respect t o as follows: 4(21. Ys=~/ta(Ylzl)a(Yzza)a Y3)dS = 0. satisfy certain compatibility conditions. (b) Find a formula for the value u ( z . < 0. t ) is in R. ~Y2. z2. @(z)= x3). E R. x E R. t ) . it is easy t o see that u(x. X E R .23)7 23 23 23 23 23 in E3 2 0.Y2.m. ax.~Y 3 W x ~ / +(Yl.. = ~ t ~ . t ) = 0. The value u(x. (a) When x E R and 0 < t < 2 3 .~ y l . . 2 0.( y+ l( Y l . 2 3 ) . ) . When 2 3 = 0. 2 2 . Then the value u(x. ) given by the above formula satisfies the wave equation and the t initial conditions. at2 2 U(X. 231. ‘$(xl. 4(x1. E aa. t ) of the solution at ( x . ax: ax. J y . (a) Write down a formula for the value u ( x .) is given by t If both 4 and II.517 Consider the initialboundary value problem azu azu azu a z U + + . +(% 2 2 . respectively. W ( X .. t > O . < 0.t ) at (2.t 2 0. x t E R for (IndianaPurdue) Solution. 0) = +(x). when t < 23.t ) of the solution at all t > 0.x * ~ ~ . (2.= 0.O) = d(X).
(iv) Let 7~ = 2. 0 for 11 2 2 k. t+O" (iii) Do (ii) above when n = 1. # 0 and suppose that $ ( ~ 1 ~ 2 2 Show that u satisfies ) ." IVuj2)dSdp. 6405 Let u(2. evaluate lim E ~ ( u ( t ) ) .) (Indiana) Solution. we have J Jyxl=t @(y)ds = 0.518 In a similar way. have we ERT+t(u(t)) = (u.T]. Define the local energy by (i) Use the energy indentity to show that Conclude that the total energy E W ( u ( t ) is conserved. u(z. (i) For t E [O. on the set T < t .4k. 23 Therefore.t) satisfies the boundary condition on = 0. ) (ii) When n = 3. + . when 23 = 0. where 2: = for some constant c.t) be a solution of class C2 to the Cauchy problem with compact support. (Here T = 1.
we can calculate that Hence ERT(U(0)) 5 ERT+t(U(t)) 5 ER(u(t))Similarly.(u(t)) equals to E.519 By the above expression. we can show that for t E [O.T] It is clear that the total energy E. (ii) From the formula of the solution to the problem it is easy to verify E ~ ( u ( t =)0 if t > R ) + a. .(u(O)).
Hence (iii) By the formula it is easy to evalute too lim ER(u(t))= 0.1x1 > 4k. we have IY  4 I 1 + 2k.IzI . it is easy to see that Hence the formula of the solution can be rewritten as When 1x1 < t . t . 4 1 and hence 1 J W J(t + IzI + 2 k ) ( t .4k. we obtain t .520 where a is the radius of a ball in which supp4 U supp.4k and IyI I k . is contained.Izl2k  .J Therefore. (iv) When n = 2. the solution is given by If 1x1 < t .ll.1x1 + 2 k < d3 V t .2 k ) ' I It is easy to verify that .
Then E ( t ) I exp(Ct)E(O) is an immediate consequence of the above differential inequality. oo)). With prove that there is a C < 00 such that for all t 2 0. t ) is a function with u E C2(ax [0.521 6406 domain and assume u ( z . we can conclude that Thus we obtain a differential inequality for E ( t ) d E(t) dt I CE(t). have we Here we have applied the Green's formula: By the Poincar6 inequality. . Multiplying the equation by ut and integrating the resulting equation with respect to x in 0. (Iowa) Solution.which solves the equation Let R c B" be a bounded with the boundary condition = 0.
E ( t ) = E(O).O) = 2 E B3. z E r B3. the equation can be written as + .t > 0. u(z. Find the spherically symmetric solution to the problem. . r = 0. = wtt. and then we can get (b) The energy of the problem is given by E ( t )= 2 It is easy to verify that 1 L.(z. + (ut” IVul2)dz. T > 0. w = 0. In the symmetric case. vt > 0. fo.522 6407 (a) Find the explicit form of the solution of the initial value problem v2u= U t t .t > 0. m. ar2 Set w = T U . (b) What is its energy at time t = 10. (IndianaPurdue) Solution.= U t t . and f(q) is a C3function on the real line with compact support such that where f’(q) is the derivative of f with respect to q. where T’ = z:+zg+zi.O) = u. r 2 E B3. t = 0. Solving this problem. w = f ( r ) . wt = f’(r).. Then w solves the following problem: aZu 2au r ar { w.
Au = 0 for x E B 3 . v x E R3. where g E Cr(R3). Prove that there exists a constant the given data such that C depending only on (Indiana) Solution.523 And therefore = 2F 7 6408 Consider the wave equation in R3: utt . t > 0. u(z.O)= 0 .X I = t } n BR Iy 5 The area of aBR.O) = g(z). . . From the Poisson’s formula. we have It is easy to verify that The area of the intersection {y E R3. Let R and M be positive constants such that and Ig(z)l 5 M . Ut(2.
and integrating the resulting equation with respect to x in R.5 t o .)be a solution of the equation utt Au = 0.524 Therefore.T ) ~ z . 4 r R 2 = . x (Indiana) Solution.Rn+l(x. M .”+ IVuI2)(x. It is not difficult to verify that Therefore..t > 0. t = T } 1 2 1 for 0 5 r 5 to. we can obtain Set 1 E(7)= 5 lr(u. Multiplying the wave equation by ut.O) : 1 1 5 t o } in the plane t = 0.t)l 5 . we have . MR2 47rt t ’ 6409 Let u E C2(. x Prove that u vanishes in the conical region D = {(x. Suppose t) also that u = ut = 0 on the ball B = {(z..t } .t) : 0 5 t 5 t o and 11 5 t o . we obtain 1 lu(z.t) E En+1. Let = {(Z.t .
we have uttdx = J. 0) = f(z). Hence E ( 7 ) = 0. we get the following equality from the equation J.. > 0 ) . t ) be a solution of the Cauchy problem utt = AU (Z E B3. gdx = 0. Prove that if = dx). This implies that vo 5 7. (Indiana) Solution.O) u(z. t Ut(Z.525 It is clear that E ( 0 ) = 0. Therefore. g E Cr(nZ3)..5 to.. then I. u=O 6410 Let U(Z. in D. . /ltnL8u 2 dx = 0. Assume that f . Audx = 0. By using Green's formula. Let R > 0 such that It is not difficult to verify that for any t > R.
by Cauchy inequality it follows that 21ax J R S 5 (J.By taking the Fourier transforms of the wave equation and the initial conditions. .. = 0. This contradicts (1) provided that J.526 and by the initial conditions.m a x l u l ? CI for all sufficiently large t. we can get .t > 0) with data u(x. .. = +(x) E Cp. JlzIRs X b) Show that there is a positive constant C such that 1 t . 6411 Consider the solution of the wave equation Utt = nu (x E R3. zn). a) Let 6 denote the Fourier transform of u with respect to the variable x = (XI. (Indiana) Solution. 0) 0) a) Use Fourier transform to show that there is a positive constant c such that u2 dx = c. ..gdx # 0. In the other side.iax)t (/ t R< ( C<t+R 3l ax) l I 2 for any t > R. ut(x.
’. .z3) for t > 0. ~ 2 This gives Iu(2.527 Therefore we have Then we get where if $0. then there exists a sequence { i n }such that t. If the conclusion is false.m a x ~ u ( z . Then we have This contradicts (a).+ c c t+b and t. t . The proof of (b) is completed.)l+t. )< la(t.tn)l I IQ(t. a s n .>l+. t ) with respect to the variable 2 = (21. 400.21. . b) Let Q ( t )be the support of u ( z .
a) Assume that R > 0 is so large that suppq5 u supp11. (Indiana) Solution. J u4(2. 11. l4<$t 4 3 0. t )=  .Show that for any 1 < 1. c ( 2 f IR3.11. we conclude that if t > R and 11 2 < t . the solution to the 2dimensional problem is given by u(21. which implies that t ) = 0.528 6412 Consider the Cauchy problem for the wave equation a) Let n = 3. then u ( 2 .t)l = O(t') Ixl9t as t + o) c. Let 8 behavior of the integral < 1. t ) = 0. Therefore. From the Huygens's principle for the %dimensional wave eauation. it holds that +. we have 9 sup Iu(z. E C F ( R 2 ) . lxl<@t when t > __ 18' R b) From the Poisson's formula. when t > A.R. 2 2 . u4(2. q5.t ) d z = 0.t)dz. E CF(R3). vlzl< et. (21< R}. Determine the large time J b) Let n = 2.
. if f > 0 for all z. t (z E B3. i.'u.v) 5 h(z). Consider the initialvalue problem for this. The proof is completed.where h E L1(B3). we have dt2..t) = J. j~~ 1 . 2 When t > 5. v. where C is a constant. a) Show that. 0 j given.zI2 2 t Therefore.(y. clear that if it is IzJ5 B t and Iyl 5 R then JY . U ) =j v).529 Let R > 0 be so large that suppllr C { z E IR2. then the same is true for f at all later times . Show that supp = o(t3) as t X (Indiana) . E IR3. (c.X I I et +R and d2 . v.e. let f ( c . V x f = 0.. b) Define the local density p by p(. Suppose that 0 5f 0 (z.t)dv+ oo.( B t + R)2. 6413 The linear transport equation is following PDE for a scalar function f of seven variables z. f(Z.): v ft + v . t E R. v. 1 1 < R } .
(2) From (1) and (2). 6414 Solve the Cauchy problem xu. v.v)dv O This completes the proof.= 0. a) The characteristic curves of the equation are defined by dx = v dt and along which it holds that dt Therefore. b) From (3). The system of the characteristic equations for the problem is given by . the forward characteristic curve departing from t = 0. = xy ( x > 0) u=5 o n x y = 1.tv. we get the solution to the initialvalue problem 0 f (2. + tv. if a solution exists.Solution. x = a is x = (Y df .tv.t) It is clear that f 2 0 i f f ? 0. (h diana) Solution.v). =f ( x . we have 0 = JR3 f ( x . + yuu.
Set v = u . t ) to the initialvalue problem Ut + = 0.O) = z2. where C is a constant. we get vt + 22121.O) = 22. y=. Then we obtain imme This implies that + u + x y = c. .~ e 'ds = _ ds (ae'y ds 1 u2 2 + u).531 ds .= xy ds du s=O:x=a. with the initial conditions _dx dY z= y u . (Indiana) Solution. dt .= 2v. + 6415 Compute an explicit solution u(x. (Y Fkom the above Cauchy problem. 1 u=5. = 0. we have immediately and 21 du du d = . The characteristic curve of the quasilinear equation (1) is given by dx . W(2.. respect to 2. The initial data gives C = diately u = 1 43cxy. F. U(X. Differentiating the equation and the initial condition with .
u t + uus = 0. p > 1. u(z. The characteristic curve departing from t = 0. and P < 0. 0. and find the discontinuous entropy solution. i. .532 along which it holds that dv . 2>1. dt along which u is a constant..= 0. z = P is given by dx =u. From (1) and (2). __ {:: 2 E IR. 12.O). o<p<1.O)= Show that the continuous solution exists only for a finite time. 1p. P < 0. dt From the above characteristic equations. dt where u o ( z )= u(x. we have 2 = uO(P)t p.e. we can obtain immediately v=Hence u= 4t 1' 22 22 4t 1' + + 6416 Solve the following initial value problem. 21 = { P> 1. < 0.0 . 1 . giving explicitly the discontinuity curve and the RankineHugoniot jump condition along it. t = o : x = p .t 2 > 0. du _ . + u = uO(P). 05z< 1 . ( IndianaPurdue) Solution. t = 0 : u = u"(p).
' t<X<l. the discontinuous curve is dt . 1  dx _Therefore. the continuous solution to the problem blows up. dx 1 .= 1. and hence x < t... we have + u). 1.2 When t = 1. the RankineHugoniot condition along the discontinuous curve is given by dt i. 0.6.t 1t 2t .2' 1 x=(t+1).2 .2). 2 t l l .= (u+ dt 2 Noting u+ = 0. 2 > 1. Fig. For t 2 1. 0 5 5 1. u. we have p=.e. 11 for 0 5 t < 1 (see Fig.For 0 < t < 1.1x 1t 1t' Therefore we get the continuous solution to the problem u = 1.6.
e bt v) dX = 0. 3 > z ( t + 1). y ) is the characteristic through (Y. .u)bu=O dX at at4 aU forO<t<T. Then v satisfies the following equation dV at + U(X. t ) 5 pebt for all x and all t E (0. 6417 a) Let u ( z . 0) ( & + h 3 ( z ) ) l z = y ) 1' t + i) Use this to compute the largest T for which a smooth solution can exist up to time T. b) Let u(X.0) 5 p for all X. if 0 5 t l < t2 < T . if z ( t . Show that if a then 5 u ( z . ii) Show that. U(X. then 00 00 (Indiana) Solution. 2 < ' ( t + l). T ) .Y).t)be smooth solution of dU at 8 + a(u)U = 0 dX = UO(X). then u z ( 4 t . Show that. where b is a constant and a is smooth. for 0 < t < T. t ) = U o ( Y .O). t ) be a smooth solution of +a(z. aebt 5 u ( z .O) where uo and a are smooth.534 and the entropy solution to the problem is given by 2 0. a) Let u = ebtv.
y) through (y. From the above expression. Hence Differertiating above equality with respect t o y.O) 5 v(z. { 4 0 . the Cauchy problem admits a global smooth solution in ( 0 . we have The characteristic x(t. FZ(0.Y)) { = a’(uo(y))ub(y). we have This is just that we want to show.O). i) Let m = inf a‘(uo(y))ub(y). t ) is a constant along the characteristic z = x(t.t)5 supv(z. I?”).= a(x. Y) = a(uo(y)). y). a Then it follows from above problem that Therefore. b) u(x. Therefore. . we get g (g+. a Differentiating this problem with respect to y. Y ) = Y. it holds that infw(x. defined by the equation . ~ ) . If m >_ 0.Y) = 1.535 The solution v maintains a constant along a characteristic of the above equation. 0) is given by the following Cauchy problem: &. we can get the Y following conclusions easily. X 2 dx dt This gives the estimation in the problem immediately.
If uo(z) E 0. on on IR x [O. u(z. prove that the only LP solution of (2') is u G 0. .  44YI s))dyds (2') + J_. A function u E P ( I R x [O. IR x (0).uO(Y)4(Y? O P Y = 0 for all test functions 4 ( z . on 4(.7'].536 If co < m < 0. t ) and &z(t.)(4t(Y. T). = f ( t .y) in the integral. y).T]) defined to be a weak solution of (2) iff is lT : / 00 4 Y .z). y) into account. ii) For any fixed t E (0. we have 00 Taking the expressions of u. . Consider the following Cauchy problem: 4t + 4.O) = uo(z).(z(t. by the change of variables z = z(t. 0418 (Uniqueness of weak solutions) Consider the Cauchy problem 'Zlt + 1.T]. '1 = 0.. we obtain 00 00 This completes the proof of the problem.T]) which are compactly supported and vanish on {t = 7) '. IR x {T}. (Indiana) Solution.. T) is given by mT+ 1 = 0.T) = 0. on x [0. the largest T for which the problem admits a smooth solution in (0. t )E P ( I R x [O.
. provided uo(x) 0. t for a) Show that if b 0. Hence we have u 0.T].)) each t > 0 (where T V = “total variation”). ) )5 TV(uo(. where q = p / ( p .T]) which is compactly supported. . Then u is a constant along a characteristic curve. . consisting of all Cmfunctions with compact support in E x [O. As the space Cr(B2 x [0. zo.T) .z . l .. + a(z. O ) (i = n..T]) with a compact support.. Therefore.T]) with a compact support. ) Then we have xi. The characterstic curves of the equation are given by the equation dx . . one still has the bound T V ( u ( * .’ TV(b(*.TI). .^) = uo(ti). meets the line t = 0 at ( < i . ut u ( z ... . where x E t B and t > 0. * . 5 TV(u0) t)) 5))ds.T]). . . We divide the line t = T into subintervals by the points * * . t ) . Suppose that the backward characteristic curve through (xi. then TV(u(. the equality (*) is valid for all functions f E L q ( B x [0.x . dt along which the solutions to the equation satisfy du .. dt a) Suppose that b 0. 21. 6419 Consider the Cauchy problem for a function u = u(x. (Indiana) Solution. ) ..537 where f E C m ( Bx [0.TI).* .t).)f(Y...= b(x.a * . .~ ..l). = b(x..t)u. we have 6’J_+_m U(Y.. It is not difficult to show that the above Cauchy problem admits a solution ~5 E Cm(Ex [O.= u(2. zl. O .5)dYd5 =0 (*I for all function f E C m ( Bx [O.is dense in L * ( B x [O. b) Show that when b $ 0..  . n .t). + J. 0) = uo(x) E C(B2). l .
538 Therefore. we get . T ) = uo(ti) + Ir 0 b(zi(s). s)ds.T)) 5 TV(UO(')). we get TV(U(. b) Let { z i } and {ti} be given as in a). ~ ) From . VT > 0. we have u ( z i . the above inequalities.. When b $ 0. where by z = Z i ( t ) we denote the characteristic curve through ( z i .
New York University Harvard Illinois Indiana IndianaPurdue Iowa Minnesota Rutgers Stanford SUNY. Irvine . Stony Brook Toronto Harvard University University of Illinois (Urbana) Indiana University (Bloomington) Indiana UniversityPurdue University (Indianapolis) University of Iowa (Iowa City) University of Minnesota (Minneapolis) Rutgers University Stanford University State University of New York (Stony Brook) University of Toronto University of California (Irvine) UC.539 Abbreviations of Universities in This Book Cincinnati Columbia Courant Inst. University of Cincinnati Columbia University Courant Institute of Mathematical Sciences.
This action might not be possible to undo. Are you sure you want to continue?
We've moved you to where you read on your other device.
Get the full title to continue reading from where you left off, or restart the preview.