Are you sure?
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
.
. Then V E E [ X ] / Ifor some maximal ideal of B[X]. 1102 Let V be a finite dimensional vector space over a field K .such that T ( W )C W . SW C W ) . W and V/W respectively. Now suppose V is a simple IR[X]module.and rnz(t) be the minimal polynomial of S as linear transformation of V. Let S be a linear transformation of V into itself. Let W be an invariant subspace of V (that is. Since R [ X ] is a P. Make V into an B[X]module by defining for i=O w = T ( v ) for all w E V .I.D. Thus a. This is a contradiction.. Prove that there is a nonzero subspace W of V . This implies that we have an irreducible polynomial f ( A ) with degree 2 3 in B[X]. W # V . (Indian u) Solution. W # 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).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 . So c 3 5 dimR(V) = dimnR[X]/(f(X)) = degf(X). ml(t). that is. Hence V is not a simple B(X)module. there exists an irreducible polynomial f(X) of B[X]such that I = (f(X)). such that T ( W ) W . r n 2 ( t ) . (a) Prove that m(t)divides ml(t). Let m(t). there is a nonzero subspace W of V .
m l ( t ) ) A n n ~ p l W ) = and (m2(t)) = A%[t] V l W . v . Let V = W @ W and S : V t V be the natural extenaion of S to V.Then we have m(t)= ml(t). Since W is an Sinvariant subspace of V. Then it is clear that ( m ( t ) = A n n ~ p l V ’(. ) = Hence m(t)divides m l ( t ) m2(t). v c m1(t). (In i m a) d Solution. W is a K[t]submodule of V. Show that V doesn’t have proper Tinvariant subspace. V can be viewed as a module over the polynomial ring l R [ X ] via f ( X ) . m2(t)divides m(t). (c) Give an example of a case in which m(t)# ml(t). then m(t)= m1(t)* m2(t). W = 0. + + 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. Similarly. So m(t)# ml(t)* m2(t) in this example. Then we have V = lR[X] .m2(t) divides m(t). z = f ( T ) (x).4 (b) Prove that if ml(t)and mZ(t)are relatively prime.ml(t)and mz(t)are all monic polynomials. (Indiana) Solution. Since ml(t)and mz(t)are relatively prime. (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. Let m(X) be the minimal 3 . since {Taw I i 2 0) spans V by (b). ~ ( t m ( t ) E A n n ~ p l V (m(t)). a cyclic module.m2(t). As usual. m2(t). Then it is clear that m(t)= ml(t) = m2(t)= t 2 1. rnl(t) divides m(t). (b) Since m(t)E AnnK[t]V G AnnK[t]W = (m1(t>). since m(t). (a) Since m1(t). T/ can be viewed as a K[t]module via the linear transformation S . mz(t).for any f(A) E lR[X] and z E V. ml(t).
Hence we have P'BP = 0.. . ) where Bi(i 2 2) = 0 E Me.. . we have B # 0. we have t o show B = 0. Thus. .. we may assume that el > 1 (If all the ei = 1. . with 0 on the diagonal. Then B' . . A are distinct and ... . . Suppose that A has n distinct eigenvalues A'. . Jt). Then there exists an invertible n x n matrix P such that PAP' = diag{ A 1 .. Jz. =v R[X]/(m(A)) R[A]* E 21 (we may assume that V # 0). . .5 polynomial of the linear transformation T : V Since m(A) is irreducible. Suppose that the characteristic polynomial of A has multiple roots. .l B ' P . Show that A has n distinct eigenvalues in C if and only if A commutes with no nonzero nilpotent matrix. We have to show that A commutes with some nonzero nilpotent matrix. Bz. . P . PAP' = P A P . So B = 0.J t ) be the Jordan canonical form of A and PAP' = diag(J1. Without loss of generality.Xn) commutes with P'BP. B . where P is an n x n invertible matrix and J. A in C. So V is an irreducible R[A]module. But the nilpontency of B implies that P ' B P is nilpotent. An}. . Let B1 be the Jordan block of order e l . . .l . Let B' = diag(B1. . Taking B = P .. is a Jordan block of order ei. which is nilpotent. Then m(A) E Annwfxl(v). V does not have proper Tinvariant subspace. Since All Az. we have + V .. Let diag(J1. Sufficiency.. If A commutes with some nilpotent matrix B .. PAP' = diag{X1. Necessity. Jz . and A B = B A .(C).. (Inddana) Solution. Then JIB' = BIJl and Bl(#0 ) is nilpotent. then it is easy to see that A commutes with some nonzero nilpotent matrix). Xa. 1104 Let A be an n x n matrix with entries in C. . B'. .l B P is a diagonal matrix.
.I>~ o of V over F[X]. then at least two of the three subspaces spaned by the sets (T*?it}z>~. Here. T : V + V a linear map such that the minimal polynomial of T coincides with the characteristic polynomial. according to the assumptions. . T E T }and { T * 8 } . By the structure theory of finite generated modules over P. .(F[X]) has the form diag{l.{Ta?}.. 1106 Let V be a finite dimensional vector space over with basis {wl.. s ) E V such that V = F [ X ] . z . U ~ . Let CT be a permutation on {q.vn}. + Hence s = 1 and v = F[X] z1 E?! f F[X]/(m(X)) is cyclic. z l $ F [ X ] .. l. so m(X)= d. the minimal polynomial m(X) of T is det(X1 . 2 .) = where (d*(X)). .6 1105 Suppose V is a finite dimensional vector space over a field K .(X) if i 5 j. Show that A is diagonalizable. there exist z. .A ) ..z = f(T). A on V .Let { U ~ . V can be viewed as a module over the polynomial ring F[X] simply by f(X). U be a base for .A ) . .zl has exactly two nonzero submodules. Obviously.I.(X) are monic of positive degree and d.> coincide.(X)(d.(X) = det(X1.D. . r 0 (Stanford) Solution.>o.. .dl(X).(X)} where the d.. V = F[X]. In general. be . . matrix a normal form for X I . (HU7Wrd) . Ann(z.(z) any z E V . .} and thus induce a linear transformation .20 and { T Z 3 } . . A = ( u ~ ~ ) the ~ ~ of T relative to the base. . Since m(X) is the square of some irreducible polynomial. . f ( A ) 6 F[X].Show that if %'. v.d. . . which is the square of an irreducible polynomial in K [ T ] . ~ F [ X ] .(z = 1 . . > are nonzero submodules { . the three subspaces generated by the sets {T'Z'}. So at least two of them coincide.A in M. . .T and 3 are any three nonzero vectors in V .} of V over F . z z ~ . ..
.. then all such T's are similar... .~ + 1 . . is+l= n). .. . . . . .1.(X)} be the normal form for X I . Mj be the matrix of Alwj : Let Wi + Wj relative to the base { ~ i ~ . 0 0 It is easy to see that the minimal polynomial of M is A" .. and prove that if dimV = 3.7 Solution. . ' we assume that when u is expressed as the product of disjoint cycles (This decomposition may have 1cycles).. . .. Then M = i } diag{M1.1 is irreducible over the field Q of rational numbers.. .* . .. The matrix of A relative to the base { V ~ ..+ M = . T is annihilated by the polynomial X3 X2 .. . Now let n be the dimension of V over Q. . and thus M is diagonaliz able.. ..u.. Prove that 3 divides the dimension of V. ..1 is the minimal polynomial m(X) of T .. Since T' = T2 T. . Let Wj be the subspace of V generated by {Kjl 1.d. Obviously.wn). . ) is . .Ms+l} is the matrix of A relative to the base ( V ~ . without loss of generality. A be the matrix of T relative t o + + + ns some base of V . Suppose T is a nonsingular linear transformation of V such that T' = T2+T. . 0 1 0 0 1 0 0 1 1 0 0 *... This completes the proof that A is diagonalizable. X3 + X2 . ( H U 7 V d ) Solution. 1.dl(X). . V . Hence.. So it suffices to prove that every M i is diagonalizable..s 1 (io = 0. vof~Wj overt. V . ~ ~ . . .. vij} + for j = 1.. diag{l.. V ~ .. we may assume that CJ is the ncycle (v1. .A A . Thus X3 X2 .. 'u2. . .1. 1107 Let V be a finite dimensional vector space over the field of rational numbers. @ Ws+l. . .w2. . . . V ~ .. ~ ~ ~ . By reordering the elements 211. . . Then the Wj are invariant subspaces of A and V = W1 @ W2 @ . . . } .2. .. . . l. .
d . then X I ..I.1.M ) . . (Here is a misprint..:.. Thus we have proved that 3 divides the dimension of V . Suppose that 1 is diagonalizable as a linear map over Fp. 11: . 1 II(u). X3 rational canonical form for A (or 7') is T ' s are similar when dimV = 3. II(u).A ) = d l ( X ) . = &(A) = X3 + X 2 . Then m(X) = 1 A" .I12(u). 1. . On the other .II"l(u)} is a base for Fq over F p .. 1. . where p is a prime.) (Hurvard) Solution.... the matrix of 1 relative to the base {u..IIn1}. . By the Normal Base Theorem. we have &(A) = &(A) = .8 where the &(A) are monk of positive degree and d i ( X ) l d j ( X ) if i 5 j. If dimV = 3.1 are in Fp. The . + X2  1). It follows that all the 1108 Let Fq be a finite field with q = p" elements.A ) ) = n.112. 3 .1.1.A" . It should be "n divides p . s = deg(det(XI . . M= I. Let 1 : Fq + Fq be the Frobenius automorphism II(z) = z p .. ( X ) . and all the roots of A" .A is equivalent to diag(1. The normal form for X . Prove that II 1 considered as a linear map over Fp is diagonalizable if and only if n divides p" .11. Since det(X1.1) and the minimal polyE nomial m(X)of M is A" .. By the irreduciblity of &(A) = m(X)= X3 + X2 .). . .. ..M is diag{ 1 .1 has no multiple root. I12(u).. . there exists a u E Fpsuch that {u. It is wellknown that Fp C Fq is a Galois extension with Gal(Fq/Fp) = {1.l". II"'(u)} * is 0 1 0 .. . 1 0 a . .1 = det(XE . . When II is considered as a linear map of Fq over Fp.
.Lli)seelements are differentiable functions of real variable t .\ .a("l)d) where d = and a is the generator of the group F.a2d).9 hand. . Hence M is similar to diag{ 1.(A .. all the root of An p . . Show that the derivative of the determinant det A satisfies Then and . 1109 Let A(t) be a nonsingular matrix r.1 has no multiple root and A"  1 = (A . (A . So IT is diagonalizable as a linear map in over Fp. if n divides p . .ad)(. . Thus n divides Conversely..  1 forms a subgroup of FJ = Fp\{O}.1) . Let A'(t) denote the matrix formed by the derivatives of the elements. u ( ~ . A" .1.' ) ~ } Mn(Fp).a d .1.
41(z). and q52(z). Define an inner product on V by + bz + cz2 with real (a) Find a n orthonormal basis for V consisting of polynomials 4o(z). b.1. respectively. .) . dt 1110 Let V be the vector space of polynomials p ( z ) = a coefficients a. and c. and 2. Hence we have proved that d (det A) = det A . + det A.. having degree 0. trace(A’ . (b) Use the answer to (a) to find the second degree polynomial that solves the minimization problem (Courant Inst. A’). .10 = det A1 + det A2 + .
$i(. 1 and 2 respectively as usual. (a) Since 1. 3 .$a(z:)) = ( p o ( z ) . 9 and 175 cg = 0. c1 = and no(.x.) = z3 . we can orthonormalization 1. 1 .4i(. Obviously.x 2 is a base of V .2).)) =0 (i = 0. Then for any p(x) E V .1. By an easy calculation.5 z ) 2 d x = . that is ) ( x 3 . we get co = 0. rl ( .))= C l (i = 0 . Hence P O ( % ) = $ z [:ix3 .x. 8 3 . (b) Let and QO(Z) = z3 Po(. 2 ) .Zz.x 2 to get an orthonormal basis of V with degree 0.) such that ~ o ( z is orthogonal to V .11 Solution.p o ( z ) .
..) Solution. ann * ZlYl . = det +s . a nn a2n +det anl . * ..... .).. (b) Show that if det(A) # 0 then a = det(A) and b = det(A) y*Alz.. * * aln ann + . when p ( z ) = gz. is minimal. . (a) Directly.. SXIYI + ... 3 det(A + szy*) a11 an1 = det + .. + Sz1Yn SznY1 SXnYn . . (a) Show that there exist numbers a and b so that det(A + m y * ) = u + bs. and 1111 Suppose that A is a n n x n matrix and that 2 = ( :). Yn y*=(y1. ( c ) Is it true that a = 0 if det(A) = O? (Courant Inst.. X1Yn det a21 ..**. . yn). Suppose that all the entries of A. z..12 Thus. .. Xn Y = ( . and y are real. [[ Un1 ... ”’ ..
. * * An1 Aln ... . a n n .... Ann = y*(det(A) . .. * . (c) If det(A) = 0. for any s. .In . Ann All ) .). ..... .. a = det(A) = 0.. Aln .. (b) Since for any s ..... ann + . anl *. . A')Z = det(A) . .. + det .... ...+ det n an11 a n . . is the algebraic cofactor of Al Thus 1 det(A) ___ aij. An1 .13 Hence det(A + sxy*) = a bs for some a and b. n X n Yn n where A.... ( If det A # 0. .. we have a = det(A) and + + b = det . .. y*Alx.. .......... All .. . .. . . det(A + sxy*) = a bs as in (a)....
e2.fn} is a basis of ker(T).Y... and corresponding eigenvector V ~ . ncc lim A'lw = = lim P d i a g { X .) = lirn X .. . (Courant Inst. v z ..O) . ...A3 3! 1 + A5 5! * . .O. f z .Xn aredistinct. . For .) Solution. and For any v E a". A.I A P = diag{A1... Define lirn T : C" + an by T ( v ) = lirn Anv. P . .. 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. diag{l. .O. Assume that XI = . P' . v. {f2. V Z .a.. . Define 1 sin(A) = A .. ...Az. f. X ~ . .. . ..14 1112 Let A be an n x n real matrix with distinct (possibly complex) eigenvalues.. .l d i ag ( l .. . .l . lim Yn). e n ) be the standard Let ncr. Prove that noo Anv exists. n+w ncc orthonormal basis of (En... Hence {fl} is a basis of Im(T). f n ) .O)P'P'l is the matrix of T with respect t o the basis ( f l . . Let fn en Then ( f l .. Xz. Since A1. (el. Then the matrix of T with respect to this basis is P ' ..1. . .. 0 } ..O.. and dim(Im(T)) = 1. dim(ker(T)) = n .O) = P'P'l diag(l. . 5 j 5 n. X ~ } ... w (Since lim (X. . P ' . . XI..O .Xn} . P isaninvertible matrix in illn(@) P . . Let P = ( v l . v n )... .. v 1200 P .) is a basis o f a n and diag(l. f z . ...A2. . 1113 Let A be a matrix.
Denote B = ( 3 Then B is similar to ") . (Courant Inst. 7 and 1 1 0 1 0 1 1 0bviously.[. ( 4 3o 0 ( 4 1 1 ) where .) Solution.15 express sin(A) in closed form. Hence sin(A)  ( .L 3!T ) 3 .(4 0 10 ) . 103 O ( 1 1 ). 1' )'=( 1 )..
. 0 . we have Y T B Y = 5 5 X. > 1. ... (Courant Inst. If B has more than two nonnegative eigenvalues. (c) Let B be the following 10 x 10 tridiagonal matrix. Show that A.1 . Then for any Y E V1. Then for any Y E Elo.B is semidefinite positive.16 1114 Let A be the 9 x 9 tridiagonal matrix '2 1 2 1 1 . Let A1 and A 2 be two of them. . (a) Show that (b) Use part (a) to show that A has all negative eigenvalues. Thus A . (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. 1 A= **. Let A. 0 ...YT(X. z 1 . which agrees with A except in the first row and column: 1 3 3 2 B= 1 1 2 1 1 *. all the eigenvalues of A are negative. *. . (c) Obviously.2. (a) A direct verification shows that (a) is true. . we have y1. (d) Show that B has 9 negative eigenvalues.B)Y >_ 0. the symmetric matrix X. a .(B) be the largest eigenvalue of B . ~ 9 ) I zi E IR} 5 Elo. Y T B Y 5 X. 2 1 1 2 All entries not shown are zero... 0 ) .. Taking Y T = ( 1 .. this is. .I . .. (d) Let Vl = { ( O .YTY.(B) > 1. Y T B Y 5 0 and Y T B Y = 0 if and only if Y = 0.. y2 E R I O such .I .. .) Solution. .
e 2 . . . there exists an n x n orthogonal matrix P such that P'TP = diag(A1.to1 Idiag(Al. .. Then min x €V . and Now let {el. Since for any orthogonal matrix H and any y E B". . .) 1x1 Idiag(Al.. z the subspace spaned by y1 and y2.a contradiction. yTBy = a?A1 u. . ..) 1x1 .. 2 ) . . positivedefinite. A. Show = by showing 5 and 2 separately. y E V1) and yTBy 2 0 (y E VZ)..Xn).. Hence max v min Z€V{O} w = m a x 1x1 v min Z€V{O) Idiag(A1.. By assumption. . en} be the standard orthonormal basis of Rn. Thus B has 9 negative eigenvalues. > A.. 1115 Let T be a real symmetric. * A 2 2 0. Since dim V1+ dim V = 9 2 = 11 > 10. z + + + 1 there exists 0 # y E V n VZ. P . V = (e1.... y1 y1 = yTy2 = 1 and Byi = Aiyi (i = 1 . Hint. . n x n matrix with distinct eigenvalues A 1 > A 2 > .y2). (Hyl = Iy1. L ) z l 1x1 . PlxI 1x1  2EVIO) min If V is a 2dimensional subspace.).A.x1 when V ranges over all two dimensional subspaces of lR".. Az. Let V = (yl. we have min 2EV10) lTxl   min Z€VIO) IP'ldiag(Xl. (Courant Inst. Then we have yTBy < 0 (y # 0. Then for any y = ulyl uzy2 E V2.) Solution. PlxI IP1x1 .A.A2..l V is also twodimensional. 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 . You may wish to express T in a basis of eigenvectors.17 T that y1ly2. ..eZ).
ez.. . for any two dimensional subspace V of 1R". . ..fi. x n is an orthogonal matrix. fz ..18 = A2 (A. we can find an orthogonal basis {fl. A ). . . fn IT = Q(e1. lT. . max min Z€V{O) ITX . Let (fl. . It follows that I v 1x1 On the other hand. 1x1 #x + azfz E V .XI so. = alf1 Then Q = ( q i j ) . . e.2 Az. . . > A 2 > 0). and Thus when V ranges over all two dimensional subspace of B".fz)... ..fn} of IR" such that V = (fl. and if 0 Idiag(A1.
A. 1 0 0 (Courant Inst. Prove that ll(I . * Then ~ m ( A l ) . + P" Since C k=O 00 IJPk(l co and the norms over vector space are all equivalent. Let A Az. (c) Use part (a) to compute the inverse of 1 A = [ ! i .. (a) First. + z".. we claim that the norms of all eigenvalues of P are < 1. QlpQ= be the Jordan form of P.P)'ll 5 *. ) n eigenvalues of m ( the sm= E + P + P 2 + . Denote ( A1 e f .19 1116 For any n x n matrix P . ~ are X . * ) rjm(z)= 1 + z + * . be the n eigenvalues of P and ' .) Solution. < 00 k=O . consider the sum R(P)= M c 00 Pk exist and R(P) ( I . . then (IP)' < (b) Assume that IlPll < 1 in a matrix norm induced by a vector norm. r). ~ ~ ~ . = k=O (a) Prove that if k=O JIPklJJ00. .' .P ) . r j m ( X ~ ) . .
P are nonzero.P)'. moo lim dm(X.I J . Then llull = IlIuII = IIPuII. l l ~ l l =sup { #lv nonzero vector I . llull < llull for the corresponding vector norm.P is therefore invertible. Hence pal < l ( 1 5 i 5 n). m+m lim Sm = lim ( I .P). (b) By definition.~ s . mw 1 Q. The matrix I . . Thus 00 ( l( I ) PI' R ( P )= k=O Pk = ( I . = We must have u = 0.20 is convergent.P ) is invertible. 2 . Then we have .) exists.u 0 for some vector u. Since mw lim Q .. Now suppose ( I .P"+')(I  PI' I .O)(I.P)' = (IP)'. that is. .P)' = I + P ( I .1 . Q = Q .~ ( lim S.(IP) rn00 = IP"+'. Since llPll < 1. for any nonzero vector u. Thus all the eigenvalues of I . n ) . . Since S. C m=O W is convergent (i = 1 . and (I.llPull 5 llPll . and we can write ( I .lim ~ + = nrn = (I .
Thus. It is well known that m j=1 and U= is a Sylow psubgroup of GL(m. Let G' = {V + V . there exists some P E GL. We fix a base of V over Fan.w + 0 * 01g E G}. . where m = dirnp. v(a E G) are unipotent matrices. prove that there exist a basis of V in which all the transformation v + a . the corresponding matrices of the linear transformations in G' are in U .(m. By Sylow's Theorem. since there is a surjective homomorphism G + G' ( 5 EndF(V)). if we change the base of V via P .F p n ) . which are unipotent matrices.Fpn)such that PHP' U .21 Then A = I . Then G' is isomorphic to a subgroup H of G L ( m .Fan). Using Sylow theory.. V . 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. Then G' is a pgroup. ( CoZumbia) Solution.N .
S N = C 2s 2 j ( . where r n ( A ) is the minimal polynomial of S . 0bviously. the dimension of the i vector spcae over F of matrices commutative with the matrix of S is v.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. . Then as Fvector space. then by Frobenius theorem. where S is diagonal and N is strictly triangular.. all of whose (complex) eigenvalues are real and positive. there exists a t least one n x n real matrix B with B" = A . & ( A ) be the invariant factors # 1 of S and let n = degdi(X).j=1 + 1)nj. Make use of the Jordan form S N of a conjugate of A . So there is a linear transformation T of V such that T commutes with S but T is not a polynomial in S. dimF F [ S ]< dimF On the other hand. Let F [ S ]= {f(S): V t V I f(X) E F [ X ] } . 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+. . . . Hint.+akSk for some k and ai E F . (Stanford) + . 1119 Let A be an n x n real matrix.&(A). N 2 C j=1 8 nj = dimF V. Show that for any integer rn 2 1. Since the minimal polynomial of S is not equal to its characteristic polynomial. ) (Stunfod) Solution. let d l ( X ) . . dimF F [ S ] = degrn(X).
2bmn+2 0 .. J k ) where Ji ( 1 5 i 5 k) are Jordan blocks with diagonal elements real and positive. all of whose eigenvalues are real and positive.. such that By = Ji (1 5 i 5 k).23 Solution.. For any integer m and b 1 0 O b 1 2 1. .. Now let A be an n x n real matrix. let b = m & B= 0 0 0 0 It is easy to see that *. Then there exists an n x n invertible real matrix P such that F I A P= diag(J1... . .Bk))m . . C. XEA and XEB" have the same invariant divisors (1.(Aa). Hence PlAP = (diag(B1. . '. J2.. . there exists real matrix B. A and Bm are similar..... any 1 5 i 5 k. Thus L : b " 0 Chb"l b " Cibm1 . ~..B2. ... . As proved in the above. Suppose first that A= be a Jordan block with a real and positive.lbmn+l cibm1 . 1... for any integer m 2 1. chbml " b I* A = QlB"Q = (QlBQ)" for some invertible n x n real matrix Q.* b O 1 b Bm= Obviously.}. .
(a) Show that there are 2'invariant subspaces V . . 0 where B k 1120 Let V be a finite dimensional vector space over an algebraically closed field K . + V.24 and B1 B2 0 P'] m = Bm.. we may write where all the w.~ = ~ i ) ~ ~ j ) .(Xal)el'l .) (b) show that there are polynomials S ( T ) . Hint. . (A transformation N is nilpotent if N" = 0. . ( A . ( Stanford) Solution. is nilpotent. a2. V and elements of a. (Aa.where a1. . j ) ((A .a Z ) e z T . Viewing V as a module over the polynomial ring K[A] via T. N ( T )E K [ T ] such that T = S ( T ) + N ( T ) . * * * . and let T : V 4 V be a linear transformation.. E K such that V is the direct sum of the V. (Aan)ebe the elementary divisors of VqX]. Denote V.. and (T . for some n 2 1. and N ( T ) : V S V is nilpotent. .)enl. Use the Chinese Remainder Theorem.j = K[A]wij. Let (A al)ell.ail) : V . .an are distinct and + By the structure theorem of finitely generated modules over PID. .* . ( T ) : V + V is diagonalizable. . .j E V and A n n ( ~ .Then . (Xa2)e=.
). T = S ( T ) N ( T ) and X .a I N ( X ) for any i 1 5 i 5 n. + vj is nilpotent. (1 5 i 5 n). waj = Cya . Thus (b) is proved. . Since ( A . Then S ( T ).a 2 ) e 2 r 2 . by Chinese Remainder Theorem. + . So the minimal polynomial m(X) I N ( X ) h for some positive integer h.a i l ) : V. waj for all wij and it is easy t o see that S ( T ) : V + V is diagonalizable. .:j (a) is proved.al)e*'l (A . Let N ( X ) = X . ( A .25 The K j ' s are Tinvariant subspace and ( T . Obviously. there exists some S(X) E K[X]such that S(X) ai( mod ( A  ai)eir. .S(X). Thus N ( T ) : V + V is nilpotent. Thus are pairwise relatively prime.
A 1 and A2 be the eigenvalues of g viewed in M2(@). . let A be an invertible 2 x 2 matrix with real entries such that N=(d)?( if d = det A < 0. (Hamud) Solution. g = In case b) g = In case c). Then A 1 A 2 = 1. In case a). Let g E G. then A 1 = A 2 = 1 or A 1 = A 2 = 1. then N E G and NgN'= 1 0) A l ( . of determinant one. .26 SECTION 2 GROUP THEORY 1201 Let G be the group of real 2 x 2 matrices.) or NgN'= (1 1 ). g is similar to  in GL2(R). i) If A 1 = A 2 . Describe the set of conjugacy classes of elements of G.
27 Hence g is conjugate to conjugate to ( ( q' ) 1 1 ) or (1 1 ) i n G . g is similar to similar (conjugate) to is not conjugate to ( i11. then we have and = 0. then g is similar to ( 2 x".2). Then N E G and Thus g is conjugate to classes ( x". for if A*( where A = d e t A = a!l i :)A'=( . we have two conjugacy classes In case d). we have two conjugacy classes [ ( i1 yl')] and [ ( 0' 1. But (A i ) isnot in G.) in G .')I* ( 0' 21 ) ( 0' ) ( i11. ) in G .) or in G L ~ ( B ) . we have the conjugacy . Thus in case c). So in this case. ) in A2 Xi [( i!i :)] [( i! . all = a22 and = 1 which is a contradiction. There exists M E G L z ( R ) such that We take N = d4 . g is inGand ( i11' ) GL2(B). M i f d = d e t M > 0 or N=(d)i( if det M = d 1 0) M ' < 0.') a21 ( E:: :tz ) E G. AS in case c). Thus in case d ) . ii) If X i # and E B (i = 1.)].
we have the conjugacy classes .2). 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.7 implies that A and B are similar in M2(nZ). we denote A1 = cos8 + i s i n 8 and = cos8 .isino cos 8 sin8 cos8 ) in GL2(C).i s i n 8 (since XI . As in the above.  O i sin 8 ) in G L ~ ( C ) . So in this case. B E M 2 ( B ) are similar in M2(~. But cos8 sin$ and cos0 sin0 are not conjugate in G.28 iii) If XI # A 2 and X i E C\E (i = 1. X2 11 < 8 8 # 0. A2 = 1 and trace ( g ) = A 1 + A 2 E E ) . we conclude that g is conjugate to cos$ sin6 or sin8 cos6 sin$ C O S ~ cosB sin8 in G. So there exists M E G L 2 ( E )such that cos8 sin0 MgMI= and also similar to ( cos8 ( cos8. that two matrices A . where T.
E El " {[(i Y ) ]I b = * l } { [( )] < < . the subgroup of upper triangular matrices with 1's on the diagonal a) Calculate the orders of G and U .(Fq) is + . and so on. U = Un(Fq).1 Hint. b) Deduce that every psubgroup of G is conjugate to a subgroup of U . (Columbia) Solution. Thus we can conclude that the order of GL.1 ways (a row of zeros not being allowed). e) Show that G contains elements of order qn . Make use of Fqn.q2 ways (no linear combination of the first two rows being allowed). the third row in qn . ) ] ~ A I .) Xz and XI 3 XZ = 1 sin6 c0s6 cos6 1 0 (61 1202 Let G = G L n ( F q ) . d) Show that g E G has ppower order iff g = I N with N n = 0.29 and cos6 sin6 sin6 cos6 To sum up.q ways (no multiple of the first row being allowed). a) When forming a matrix in G L n ( F q ) .the group of invertible n x n matrices over the finite field Fq with q = p " .we may choose the first row in qn . p a prime. the second row in qn . Deduce that U is a Sylow psubgroup of G. c) Determine the number of Gconjugates of U. the set of conjugate classes of elements of G is U { [ ( $ x".
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 ) . x .I.l E U .or gUg' = U .1. Then ( 9 ) is a psubgroup and ( 9 ) is conjugate to a subgroup of U by b). if g E T: It is easy to see that g U g .30 When forming a matrix in U n ( F q ) . = 0 if i < TI.l M h and N" = h .we may choose the first row in q n . Denote h g h .. the number of Gconjugates of U is equal to [G : N G ( U ) ] where N G ( U ) is the normalizer of U in G.. matrix multiplication shows that 0 is a group epimorphism whose kernel is precisely U = Un(Fq).. There exists an h E G such that h g h . . n + We can define a map 19 : T + F. 0's elsewhere. x F.1)th row.l ) . . j = 0 if i < n . U is a Sylow psubgroup of G.. there exists some 1 U12 0 u=[. by mapping a matrix onto its principal diagonal. Obviously Mn = 0 and g = I h . we get gni = 0. Uln ~ 2 3 ::: . + ..l M " h = 0. . of matrices ( a i j ) E G with a.l ) / ( q l)n. .j) place. j E V such that g ( I E j j ) = u g where Eij is the matrix with a lone 1 in the (i.. d ) Suppose g E G has ppower order.. c) By Sylow's Theorem. So the order of T is b Hence the number of Gconjugates of U is i=l n (qi n . Checking the last row of g ( l + E i j ) and u g . Since p j [G : U ] = .. By this way. the second row in q n P 2 ways: the third rows in q n . we get g n .l M h = I + N where N = h ..j = 0 (0 5 j < i 5 n). So g N G ( U ) and we have N G ( U )= T . So gn. 1 . we get g i j = 0 if i > j Hence g E T = T n ( F q ) the set . Then checking the ( n ...l = I + M where M is a matrix with zero on and below the diagonal. Suppose g = ( g i j ) n x n E N G ( U ) . .. Then for any 1 5 i < j 5 n.l ways.3 ways and so on. Conversely.l .. i=l b) It follows directly from Sylow's Theorem.' & U (by matrix multiplication).
31 Conversely. . we have a linear map al : Fqn FqR. $ G = Gn(Fq). (IHI . For any a(# 0) E Fqn. e) Consider the finite field Fqn as an extension of the field Fq and as a vector space over Fq. if g = I N with N" = 0.1.1... f + 1203 [GI = IgEG U gHg' 5 [G : N ( H ) ]. It follows that G contains element of order q" . ~ ( u = at. So the order of g is ppower. It is well known that F. the Jordan canonical form for N is a matrix with zero on and below the diagenal.1) + 1. has an element with order q" . We fix a base of Fqn over Fq. a l ( z ) = a z (over Fq) which is invertible. (T is injective. then we can obtain a group homomorphism (T : FLf. ) Obviously. Thus g is similar (conjugate) to some matrix in U .
l . then (1) u H a H 1 a . there exist subgroups Hi such that Hi I> H . a K. . . Then we have G = U gHgl by assumption. . = G N is a composition series of G with composition factors H and H i / H i . 3 t k } be the Hamilton’s Quaternion group and H = (i) = {3tl. UEG 1204 Suppose that H and K are normal subgroups of a finite group G and that GIH = K.l Ki/Kil (i = 1. n . we have the isomorphism G / K N H . I c } N K 4 (Klein four group). . (a) < G.32 (b) First we claim that G is cyclic. Let H be a maximal subgroup of G. we have H a G. . j . . it is easy to see that G is cyclic of prime power order since all of its maximal subgroup are conjugate.2. By the JordanHolder Theorem. l Suppose H is simple. . n ) and Hi ~ H i + (i = O .. K is not isomorphic to H .. 2 . This contradicts the maximality o f k . Obviously.. H i I H N Ki ( i = O . . .i. On the other hand. . n). . (b) Let (1) = KOa K1 4. But G I K = { 1. (1) = KOa K 1 a . So (a) is contained in some maximal subgroup of G (since G is finite). . (a) Give an example to show that G / K need not be isomorphic to H . . f j . so it is isomorphic to K . a H. K a G and GIH is cyclic __of order 2.1 u K .*i}. * * * . = K be a composition series of .l a H . G has only finitely many subgroups of index n. Then for any a E G . (Columbia) Solution. 1205 Prove that if G is a finitely generated infinite group then for each positive integer n. then G I K N H . aK n P 1a K U G is a normal series of G with factors KiIKi1 (i = 1 . Hence G is cyclic.1). Since GIH N K . n ) and G / K . K = (1) = {*l} be the subgroups of G generated by i and 1 respectively.. Suppose that G is not cyclic. & i . By (a) G = H . . (a) Let G = Q8 = { f l . K . (b) Show that if H is simple. Now.
Since G/HG is finite. 1206 Let G= {( a!l ) la E lR*. G/HG has only finitely many subgroups of index n. Then G x G I H + G / H . ~ ' ) Matrix multiplication shows that 0 is a homomorphism.+))is an abelian subgroup of G. g . E lR b 1 ( C_ GL(2. and . Let 0 be the map G + R* x lR* mapping a!l ) to its diagonal ( a . ker8= {( ) 16tnJN Obviously. Since G is finite generated. Hence G is a solvable group. . where S.. (Columbia) Solution. and GIN E x IR* is also abelian. there are only finite number of homomorphisms from G to S. Thus G has only finitely many subgroups of index n. N = ker0 (21 (lR. It follows that the set g€G {HGIH < G . Suppose . [G : H ] = n } is finite. Let H be any subgroup of G such that [G : H ] = n. ( Co2umbia) Solution. ( z H ) = ( g x ) . Show that G is solvable but not nilpotent. the symmetric group on n letters. Show that if H 5 G (finite index) then there exists a homomorphism of G/HG into S[G:XI. n Hence the group GIHG is isomorphic to a subgroup of S. Let H 5 G and define HG = g€G glHg. with HG as its kernel.33 Hint.lR). This induces a homomorphism G + S. H defines an action of G on G I H . denotes the symmetric group on n letters. The kernel of this action is HG = glHg. where GIH denotes the set of all left cosets z H ( z E G ) .
Hence = C1(G)= C2(G)= * * .l .34 the center of G. y E IR. Hint.p " . ) EG.= C. a = f l . that is. Then for any ( . 1207 Let p be a prime and let V be an ndimensional vector space over Fp.l ) . n2 'nk.fi(X).p ) (p" .fi~(X)I(X~~~l) . f(X) is separable (remark: here separability means that f(X) has no multiple roots). (a) Let T E G and f ( X ) be the minimal polynomial of T over Fp. Denote n = degfi(X) i and m = n1 . Hence f(X) = fl(X). that is..(G) = a * * = G $ where Ci+l(G)/C. is isomorphic to a subfield of E .frnl = 1. So G is not nilpotent. Since nilm.X over Fp and Ei = Fp[X]/(fi(X)) i 5 k ) .(G) C ( G / C i ( G ) ) = (Co(G)= 1 ) . Show that H can be simultaneously upper triangularized. G = G L F ~ ( V Recall that IGI = (p" ). Prove that a transformation T E G is semisimple if and only if Tprnl= 1 for some positive integer m. ( 1 5 i 5 k ) . (b) Let H be a subgroup of G of order a power of p . Then X 1 f ( X ) . Let . Find a Sylow psubgroup of G.l)(pn.  .l ) . A direct discussion as above shows that C ( G / C ( G ) ) ( 1 ) . f k ( X ) for some distinct irreducible polynomials fi ( 1 5 i 5 k ) . It follows that b = 0. Then IEl = p" and Ei is a field of order (1 5 pni. f i ( A ) ~ ( A P r n . (Indiana) Solution. So E contains an element ~i whose minimal polynomial over Fp is f i ( x ) .. Since ~i # 0 and 7. 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. E. Let E be a splitting subfield of XPm . If T is semisimple. then f ( X ) = f l ( X ) f 2 ( X ) . Thus T p r n . x!!l So ay+ bx' = xb+yal for any x Thus # 0..l = 1.
H is contained in some Sylow psubgroup K of G and K is conjugate t o U .2 .. (p" . (Indiana) Solution.1). (b) Suppose that for any elements h E H and k E K holds hk = kh. Since (XP"'' . Now let H be a subgroup of G of order a power of p. This proves (b). IHJ = p 3 . (a) Since K is the commutator subgroup of G.l 5 gKg' = U. Thus T is semisimple.  .1) . 1208 Let p and q be distinct prime numbers. ( c ) First we claim that H is not normal in G.1)' = XP . Hence G = H K and H 21 H / ( e ) = H / H n K 21 H K I K E G / K is abelian. Let H be a pSylow subgroup of G. ( p . . Hence there exists an element g in G such that g H g . Then f(X)I(P"'' m 1).I)... G = H K is abelian. (b) Show that there are elements h E H and k 6 K such that hk # kh. Since IK( = q .35 Next suppose TPm' = 1 for some positive integer m. By Sylow's Theorem. H n K = { e } and lHKl = p3 q = IGI. Let G be a group of order p 3 .l)(pn .2 # 0. hkhlkl E H n K = { e } and so hk = kh. Then JUJ p"l . K is normal in G and G / K is abelian. (p"' . the second row in pn2 ways and so on).p"') n n1 = p. q such that its commutator subgroup K is of order q. f ( X ) has no multiple roots. Since )GI = (p" . Since K is abelian and H is abelian. U is a Sylow psubgroup of G. (a) Show that H is abelian and G = H K . the subgroup of upper triangular matrices with 1's on the diagonal. = . for any h E H and k E K .p = pv (when forming a matrix in U . Thus we have proved that H can be simultaneously upper triangularized.1. By Sylow Theorem. So H K = K H is a subgroup of G.p " .p ) * * * (p" . (b) Let U = U. This contradicts the fact that the commutator subgroup K of G is of order q.(V). (c) From (b) show that p divides q . Otherwise.we may choose the first row in pn' ways.
rPlqz and plrp . = 0.3e2 6e3. Show G has a subgroup of order p q . fz = 3el 3ez and f3 = 3el . A4 does not have a subgroup of order 6 . (a) Let rp be the number of Sylow psubgroups of G. by Sylow’s Theorem. q since H is normal in G. G also contains an element of order q . + 3f 3e. Let G be a group of order pq’. 2 2 . plq . q are primes and p > q > 2 . So G has only one Sylow psubgroup H .1. (b) What can you say if q = 2 (and p > q ) ? (Indiana) Solution. q be primes. K is a subgroup of order p . q . Aq. p > q > 2 . Since p . q. For example. So G has a subgroup K of order q.36 the number of pSylow subgroups of G is greater than 1 and divides q . 1200 (a) Let p .1. 1210 Let A be the abelian group on generators el f. subject to the relations 9e+3f+6g 3e = 0. Let F be the free abelian g r o u p g e l $Zez $Ze3. the alternating group of degree 4. G may not contain a subgroup of order p . Then. (b) If q = 2. K be the subgroup of F generated by f1 = 9el 3ez 6e3. is a group of order 3 . and g. By Cauchy’s Theorem. = 0. Denote + + + + 9 3 6 . Obviously A E F / K . it is easy t o see that rp = 1. which is normal in G and of order p . (Stanford) Solution. So it is Again by Sylow Theorem.3f+6g Give a decomposition of A as a direct sum of cyclic groups of prime order or infinite order. Thus H .
that there exists an n x n matrix N with rational entries so that M N equals the n x n identity matrix..6. 1211 Let M be an n x n matrix of integers. a) Show that 2"/MZn finite. (fi.. Let (e.37 It is easy t o get the normal form diag{ 3. 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. Suppose that M is invertible when viewed as a matrix of rational numbers. e 2 . f4)' = Q . e3)' and (fi. $ Z e L and K = Zfl + Z f 2 +Zf3 = XCe. (Stanford) . e$>' = P . @ Uel.6. f3)'.l ( e l . i.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. e i .e.f 2 .0} ( P and Q are invertible matrices in M 3 ( Z ) ) . fL Then F =Zei $Ze'. View M as an endomorphism of 2". is b) Show that the order of Zn/M57" is equal to the absolute value of the determinant of M ..
. (fil f. . .Denote (ei.ez.. and P . 2 .4) and (21.. b) Let D = diag{dl. . fn)' = M . . .. . . . dz. It is also obvious from the facts that the map g : Zn/MZn Z " / l M p " . (fll f~.l). . . Then express G as a product of cyclic groups. . f. .. . 0bviously. .. . eh . . where [MI is + the determinant of M ..(Z) such that D = Q M P . Q E M.. f.)' Then {ei.. ( . M Z n can be generated by {f. . .. . . 7 * * * . eh. e2.. (el. e2. .f.} where f... I / fz.ek} is another base ofZ". . Since Q is invertible in M n ( Z ) .ek)l = P'(el. (Haward) + .. . d z . det P = det M or . . 1' 1 = Q .dz.(Z) be invertible matrices in M. ] Calculate the order of the additive group G = D2/K where K is the Dsubmodule of D2 generated by ( 2 t 11[ . where the di # 0 and dildi+l (i = 1 . M * is the adjoint of M .e. en)I . en)' = M P (ei. . . . . Let { e l . (fi f~r : . el. 6 ( r ) = M * X is injective.. . . ... n . = det Q . and the fact p n / l M p n l = ( I det MI)". a) It follows directly from b)..dn} be the normal form of M in M.l). e. .. Let (fl. d.el. } . Then M P is generated by { f l l f 2 . .Solution. .el.fn )I I = Q M P (ei.) / I = diag{dl. . . ([ + 2 . . (ei.. .. . f n }.. e n } be a base for the free module Z"...21).f )' = Q . e n ) .(Z). d. det M .det M . . . ... (f. 1 * * f n )'.  Hence 1212 Let D = a [ [ with E = *. Obviously det D = dl . .
e2 + 0 .e2 + (ez. 0 21 . + + (a bt)[(l+ = a ( e l + 2tel . For any a bt E D .2(e2. 2el. e l + b()(2le1+ 21e2) + 0 .e2 . It is easy t o see that {el + 2te1.(el .+ 21 (el + 0 .l] 0 . {ell(el.4ez + (a = a(21 .5(ez).e2 . Then + 0’= Del @ Dez = Z e l @Z(el @Zez @Z<e2.21(el + 21(e2} is a generating +b(O e l subset of K as Zmodule (additive group). 2 1 e l + 21e2.el + (el .1 2 1 1 A= 1 1 2 .(e2} i s a baseofD2asZmodule.e2 + + + (e2) + b(2e1 + (4 + Oezl + (1 + t ) e ~ l . . for A in hf6x4(Z) and to find invertible matrices P E &(Z) and Q E 1M4(Z) such that PA& = N .5(e2. 0 . el 21. D = Z [ t ] = Z@Z< (as additive group). It follows that D 2 / K .9 0 ~ 21 5 ~ .2el + ( e l 4e2 + 5e2.<ez) + (ez) + b(el+ (el . Since ( is a root of the irreducible polynomial z2 z+ 1. e2 + 21 . It is routine to get the normal form 1 0 0 0 N = [ 0 . .(el . (e2). Denote 1 2 .e2 . e2.2/(3) @Z/(21)and the order of D 2 / K is 63. and {(1+2<)e1+(1+<)ez1(2+ ()el + (4 + () .0 .39 Solution. 21.  ( a b0[(2 ()el a(2el + < e l .e2 .2(e2). e2} is a generating subset of the Dsubmodule K of 0’. <el + 21 .e2. Let { e l l e 2 } be a base of the free Dmodule 0’.
let c = ha + e l . Z ) with the form ( 1) is a prodct of the for some b'. Similarly.40 1213 where and S L ( 2 .there exist some q .c'. Suppose that M = Prove that S L ( 2 . Z )is the group of 2 x 2 matrices with integral coefficients and determinant = 1. a and c are relatively prime. It is easy to know that if a = 1. If a Hence > c > 1. 1 5 c1 < a . Z ) is generated by ( 1 rtl ) ) ( i1 If both a and c are nonzero. a1 E Z such that 1 5 al < c and a = qc+al. and if c = 1.be = 1. for some b' E Z. . we claim that matrices in S L ( 2 . if c > a > 1. then for some d' EZ. (Stanford) Solution. Since det M = ad .d' EZ.
then y E N. K a normal subgroup of G and P a Sylow subgroup of K . (Indiana) . ( i1 )) where b ' . It is easy to check this by the property of the elementary matrices. Hence h'gP(h'g)' = P and h'g E N . By Sylow Theorem. Z ) = generated by ( 1 3zl ) ( and 0 1 ).e. N . d ' E Z . H 2 N .41 According to the above discussions. N .l = H . So we have g = h . 1214 Let G be a finite group. N = {g E G 1 gPg' = P } then show that G = K . (Indiana) Solution. d' E Z.. we have 9Pgl C gKg' 2 K since K is normal in G.l . X ) with the form for some b'. there exists h E K such that gPgl = h P h . For any g E G. c'. This completes the proof. it can be derived that M = is a product of the matrices in S L ( 2 . h'g E K N . If N is the normalizer of P in G (i. it suffices to prove that all are in (( 1 fl cc) . Hence gPgl is also a Sylow subgroup of K . Prove: If y E G such that y H y . So to prove that S L ( 2 .  1215 Let P be a Sylow subgroup of a finite group G. Thus we have G = K . Let H be asubgroup of G. Let N = { x E G I xPx' = P } . c ' .
(b) Give an example such that P ( G ) is exactly 5/8. yPy' is also a Sylow subgroup of H .(1 1 2 1 . Obviously. 1216 Let G be a finite group. Since yPy' & yHy' H.b) E G x G I ab = ba}J IG x GI (a) If G is not commutative. IG/C(G)I L 4. (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. prove that P ( G ) 6 5/8. P is a Sylow subgroup of H . So there exists an h E H such that yPy' = hPh'.N cH. Hence h'yP (hly)' = P and h'y N . + ) 4 =8 5 . So 1 y=h. then. (Stanfod) Solution. Since J{(a.42 Solution.b)E G x G I ab = bu}l = P(G) = L . obviously. The probability of G to be commutative is defined by P(G) = I{(a. If G is not commutative.hlyEH.
obviously. 0 5 j 5 3. y}. and so ~ C G ( U )4. Then IGJ = 8 and C(G) = (1. .C ( G ) . . 3 5 l C ~ ( a )< 8. For any a E G . It is easy to see that P ( G ) = l =/ 518.43 (b) Let G be the dihedral group 0 4 = { z * y j I 0 5 i 5 1 .z2 = y4 = 1 zy = y3z}.
\/z].n&? is also a n irreducible divisor of 2 0 . Obviously T and 6(T) = 6 ( b ) * ( l p . . [] Suppose (20. Since = the integral divisor of yo + 2/2 or yo can only be f l . ti = m . Suppose that a. b # 0 E a[].vI 5 1/2. (Hurvard) Solution. So we have a map 6 :Z[]* + Z?. Z [ G ] is a unique factorization domain. If a = m nis an irreducible divisor of 2 0 .u ) = bq+r where Q = u + (v + v . and if alyo then + + + + n. 1) is the set of units of Z . b) Using a).2112 5 6(b).bq = b(p . For any a = m nz/2 E Z[z/2].v)Q] +v E Z[] Q is in Z [] and r = a . We can find integers u and v such that I .v)Q.yo) be an integer solution to the equation y2 2 = x3. we have (yo Q)(yo . In the ring Z [ G ] . (a + t) 2* + 21v  wl2) < 6(b). b) By a).u ) + b(v . Then ab' = p v. we define 6 ( a ) = m2 2n2.G) zg. Hence Z [ m is Euclidean. (1. a H 6(a). a) It is readily known that Z [ n ] is a subring of a.\/z where p and Y are rational numbers. 20 is not a prime element in Z[. Then w + + + a = b(p+vG) = b [ ( u + p. find all integer solutions to the equation y2 + 2 = x3.pl 5 1/2 u and l .44 SECTION 3 RING THEORY 1301 a) Prove the ring Z [ n ] is Euclidean. hence an integral ' domain..
(R)*. (a) If a E R is nilpotent (am = 0 for some m) then 1 .T. R* denotes its multiplicative group of units. T . Then S = {o. Hence there exist 6. {o. the set a . it is trivial. n E Z. c E (Harvard) Solution. exactly (a .6mn2 1 = 3m2n. 2 .2n3 = (3m2. 9 G .S C Z/pZ and Z/pZ has only p elements.S ) n S # 0..45  from which it follows that a31yo fl...a E R*. Since a . Hence + + + yo = m3 .(R) = M.j < 9 if i > j . When p = 2 . i . c E a/@ such that a . there exist b.y 2 = i + j . On one hand for any i # j . 2 (q)2} y .Thus 2 S' = { b 2 I b EZ/$} = { a . M. has exactly elements.rn. that is a = b2 c 2 .. Now for any element a E Z/@.0 I j 5 e. yo is of the form ( m nJ2)3. . 0 < p n 5 9 and A2 = ~  n2 = ( n . ? 2 . . .22.0 < j 5 q.2n2). 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.T. and GL.z. On the ) are other hand.1/2. Assume that p # 2. n.c3ly0 .S := {a ..b2 = c 2 . for any A2 E S where 9 < n < p . Let S be the set { b 2 I b E a/$}.(q)2} elements in S. .. Let p be a prime.j # O ( O < i + j 5 p . ( q ) } 2 + 1303 For a ring R. (b) Let J be a nilpotent ideal of R (J" = 0 for some m). : i 0 < i . .p ) 2 = has exactly elements.1.(R) the ring of n x n matrices over R.b2 I b E Z/@} has elements. So without alyo loss of generality. Show that for any element a E Z/$.
s. Obviously.". 3: H az is an invertible linear . There exsits a E 3.s + 3. 1))  IA =I mod M .T. .4 6 Show that GL. (e) We consider the field extension 3 C 3. is it easy to see that U is a pSylow subgroup of GL3(3.c E F. . . .} is a pSylow subgroup { A = ( a i j )I = { A E M. with kernel ) (d) Show that U = {( l a b 0 1 c 0 0 1 ) a. % ( R / Jis surjective.(J).). b. : F. ( J ) } (c) By calculating the possibility of the row vectors of A in GL3(Fq).(R) t G L .3. T.(R) = I+M.. 2 = diag{i. Such that Fi3 = ( a ) .
e. 2 ) is an integral solution of x 2 . is an integral solution of x 2 . b = 2aobo) is another solution of z2. bo.1. Then we have bf2 = So must be a square number. say a’ = 2b.lbil (i 2 l). s 3 . Thus P is a 5Sylow subgroup of G L 3 ( Z / 2 U ) .1 . we get an infinite set of integral solutions { ( a i . Hence either both $ and a f 1 are square ’ numbers or a and ’ are square numbers. and ” a = 2 4 .1.47 transformation over Fq. v. Then J 2 = 0 and % = R / J N Z / U . 512 31.. the kernel of the surjective homomorphism G&3(Z/2w) GL3(Z/5Z) 3 is I and + M 3 ( J ) . (Indiana) Solution.31. relative to some base for Fq. Suppose that (a. is ( f ) Let R = Z/2M and J = U/2U < R . 3 . it is easy to see that if (ao.Then P/ ker cr ‘v U and JP] I kercr] . a . = 2aP1 .2y2 = 1 (i.  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 ) .2y2 = 1 . By (b).1 . = U 1304 Describe an infinite set of integral solutions ( a .2y2 = 1. ( 3 . + so Let (GL3(Z/2U)I= 2 7 . bo = 2 and a. II = 512. Let A E GL3(Fq)be the matrix of T. the order of A in GL3(Fq) q3 . bi = 2a.bo) is a solution of z2. + + . for some integers ao. then (al = 2 4 . a’ 1 = a.1) = 2 7 .2y2 = 1.2y2 = 1. b ) of the Pell’s equation z2 ..1)(52.(a) = q3 .1 . bo) is a solution of z2 . then b = b. Obviously [ I M 3 ( J ) J= 59 IGL3(Z/U)I = 53(53 . 3 .Since . .bi) 1 i 2 0) of the pell’s equation x 2 . (ao.2b2 = 1). ad. &om the above consideration. = 3 ./Fq. Let a = 2a’+ 1 and b = 2b’.1)(5 . b E Z such that a2 . b = f2aobo. Obviously. Then a must be b) odd and b even. If $ and a ’ 1 are square numbers. So if we take a.2y2 = 1 .2y2 = 1 .
say a' = ug. (Indiana) . (ascending chain condition) for finitely generated ideals then show that R satisfies the a. b = 2 f 2 a o b o ) is an solution z2 . that is. Conversely. For any a (f 0) E R. Give example (without proof) of such a ring which is an integral domain but not a p. 1305 Let p be a prime. a'+ 1 = 2bg.. Thus we have proved that R is isomorphic to a subring of a direct product of rings. Since p is a prime.. from the solution ( 7 . It follows that R has at most p elements. RIP .2y2 = 1.. prove that R is isomorphic (as a ring) to (b) In the general case (i. (a) Suppose that R is an integral domain. (a) If R is an integral domain.70) of z2 .bo) is an solution of ' .z = 0. each of which is isomorphic to a/$.a. For example. a is a root of the polynomial zP . and for any a E R. (b) Since for any a E R. Char(R) = p and R ZlpZ. Hence R is a field.2y2 = 1.2y2 = 1 and a = 2 a i + 1. Let R be a commutative ring in which u p = a for all a E R.c. If R satisfies the a. if (ao. If a' and are square numbers.48 Remark. 5 ) of x 2 . a  P For any prime ideal P of R.c.p z . then ( a = 2ag + 1 . R not necessarily an integral domain).bo) is an integer solution of z2 . (Indiana) Solution. we get a solution (99.a u(aP1.2y2 = 1. ap' = 1.i.Z / p Z . the nil radical of R is 0. u p = a.2y2 = 1.c.e.d. Hence = all prime . 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.c. the intersection of all prime ideals of R is 0. b = k2aobo. 1306 Let R be a commutative ring with 1.1) = 0. By (a). RIP is an integral domain and for any a E RIP. since UP . for all ideals. then (ao.
c. a . but not a p. for all ideals. (d) If A is a maximal ideal of R.c.49 Solution. 1307 Let R be a commutative ring. then S . For otherwise. (a) Obviously. (c) If A is contained in every maximal ideal of R.c.satisfies the a. R satisfies the a. SlR = R.c. and let + N be an Rhomomorphism.l R by (b). X [ z ] . what is S .the polynomial ring over Z. (b) It is well known that there is an onetoone correspondence between the prime ideas of S .l R and those prime ideals P of R such that P n S = 0. the localization of R at the maximal ideal A. b E A. 1 = 1 + 0 E S . then for any ideal I of R.l R ? (d) If A is a maximal ideal of R. (b) Show there is a onetoone correspondence between the prime ideals of S .l A is the unique maximal ideal of S .l R is isomorphic to RA. then A J ( R ) . + 1308 f :M Let I be a nilpotent ideal in a ring R. the So every elements in S is invertible in R. 1 is finitely generated.l R is a local ring. (1 + a ) ( l + b ) = 1 + (u + b+ab) 6 S. This is what we need to prove (b).d. we can construct a strictly ascending chain of finitely generated subideals of I . (c) If A is contained in every maximal ideal of R .i. Let A be an ideal of R. (a) Show that S = (1 a I a E A} is a multiplicatively closed subset of R.c. for all ideals. P n S # 0 if and only if P A # R here.c..let M and N be Rmodules. what can you say about the structure of SlR? + + (Indiana) Solution. Show that if the induced map . for finitely generated ideals. Hence R is a Noetherian ring. It follows that S is a multiplicatively closed subset of R.e. Hence Jacobson radical of R. If R satisfies the a. So S . i. For any 1+ a and 1+ b. and it is easy to see that S . Obviously.l R and those prime ideals P of R such that P A # R.
So N = f ( M ) and f is surjective.VP(dt))).50  f :M / I M t NIIN (Harvard) is surjective. Obviously. .V p ( f ( t ) ) . Since I maXMt)l. ii) V. = 0. N / f ( M ) = I . 4 l IP(t>lP) .(f(t)) + % ( S W and 4 v.(t). because I is nilpotent. N/f(M) = I N Hence + f(M)/f(M)= N / f ( M ) . show that the set of a(t) E F ( t ) satisfying la(t)l 5 1 is a local ring. We define V.dt))= V. by defining I f ( t ) l . L m={I4t>Ip. = 2 .(f(t).(f(t)) = 00 if and only if f ( t ) = 0. Obviously. : F ( t ) + I?i by %(O) = 00 and Vp(f(t)) = Ic if 0 # f ( t )= ~ ( t b) t )~c ( t ) .N / f ( M ) = . c ( t ) E F [ t ]and * ( / ( b ( t ) . P ( t ) ) = 1 = (. Since f :M/IM + N/IN is surjective. N / f ( M ) = 1 2 .Ia(t)lp5 1) and M = { a ( t )E R I la(t)l < 1). 1309 Let F be a finite field containing 5 elements.. Let p = p ( t ) = t . Explicitly construct one nonarchimedean absolute value I I on F ( t ) . f ( M ) + I N = N . * = I n . we have i) V. It follows that I . If is the completion of F ( t ) with respect t o 1 1.where b ( t ) . 1 (called padic absolute value) on F ( t ) .(f(t) + d t ) ) L fin(Vp(f(t)>. Let t be transcendental over F . Is(t)lPH 1 + m. . 1 is nonarchimedean (If(t) W l P + Suppose F ( t ) is the completion with respect to 1 Ip.P(t)). Then. (Columbia) Solution. Solution. we get an absolute value 1 .1 E F [ t ] . Let R = { a ( t )1 a ( t )E F ( t ) . I . N / f ( M ) = * . then f is surjective.
and f j is a sum of monomials in ' Y 1 . . .X .l E R and a ( t ) is invertible in R.51 and . . cannot be generated by fewer than n elements. .l E = F ( t ) is the inverse of a @ ) . where a i l .2.X 2 .n). . X . . . . . . 2 . and X I . k = 1. In the same way. Y2. . a ( t )E R\M. b ( t > P ( t ) l P = l+)lPIP(t)lP>  .any For R is a subring of the field F ( t ) and M is an ideal of R. . n ). ) of the ring C [ X 1 . in the polynomial ring a'[X1. ' m aijbjk j=1 = 6ik (i. .. . We We can write x. + * * * gj for j = 1 . 1310 Prove that the ideal generated by X I . x . .. . of degree two or greater. X 2 . Since X I ... . we also have y j = bjiXi + bj2X2 + + bjnX. a j 2 . of degree two or greater.Y.]. = UilY1 * * * U l m Y m fa + + + for any X . gj is a sum of monomials in so = 5(5 k=l uijbjk) x k + (terms in X I . . we have la(t)lP 1 and Ia(t)'lP= 1 where a ( t ) . .So. . . have to show that m 2 n.. a ( t ) . .X .. 2 . . .* .rn. . where b j l .of degree 2 2) j=1 ( i = 1 . . are algebraically independent over a . . ] ( StanfonE) Solution.X. Thus R is a local ring with maximal ideal M . . ...X . . . . .X 2 . . . Suppose that { Y l .Y. a i m E a .} is another generating subset of the ideal ( X I .X . . bj. * * * .X z . . E a'.
1. There exist a E I . 1) + [0. 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 . since + 1= (a + b ) 2 n E (a") + (b") = I1 + 12. ( Stanford) Solution. Let 1 1 = (a") and 12 = (b"). 1) by f(x)= { 2x. Since every prime ideal of A contains either I or J but no prime ideal contains both I and J .52 1311 Let A be a commutative ring. = (Indiana) . " Then I1 and I2 are proper and I1 I2 = A .we have I J N ( A ) (the nil radical of A ) and I + J = A . (b") = (a"b") = 0 and thus A N A/I1 x A/I2 where A/I1 and A/I2 nontrivial. and since ab E I J C N ( A ) . 22  if 0 5 if 1 5 22 22 < 1. By Chinese Remainder Theorem. we have I1 n 12 = 1 1 I2 = (a") . Find all z such that f(f(f(f(f(f(z)>>))>2. 1312 Define f : [0. Prove that A N A1 x A2 for some (nontrivial) rings A1 and A2. b E J such that a + b = 1. there exists an integer n such that (ab)" = an b = 0. < 2.
Since E [0. Define deg f(x) = N where U N or bN # 0.53 Solution. ( 2 2 ) .) = (2 * (2 . Since for any real number a and positive integer n. sin n2 0 .)))))) z = { 2 6 4 = (642).n[u]] nu . Use this result to prove that T is an integral domain which is not a unique factorization domain.) z if and only if z = = &.[u])] = nu . For any real number a . Then it is clear that f ( z ) = 22 . f(f(f(f(f(f(.[nu]= { n u } .632 = [64z] = [63z + z] (to be an integer) if and only if 1 2 63 Thus f(f(f(f(f(f(z)). l). it is easy t o see that N f ( z ) = a0 + n=l a. 1313 Let T be the ring of all real trigonometric polynomials N f(z) = a0 + n=l a. cos nz + b.[u] here.n . By using the orthogonality of the set ( l . Hence f(f(f(f(f(f(4 { n .n[u] . Show that deg f . [a]. for any z 6 [0. sinnz I n E nV}. (2 . cos nz + b. (Columbia) Solution. (2 * ( 2 .[64 .{ u } } * * .{ u } ] = n(u .[ n . c o s n z .[nu . z] if and only if 63z = [64z]. [u] denotes the largest integer We denote { u } = a .[n(u.[u]) . we have = n { u } .([nu] n [ u ] ) = = nu .  0 ) . 0 5 k 5 62.[2z] = ( 2 2 ) 5 a (Gauss function). g = deg f + deg g. sin nz. Notice that 2 = (642) = 642 . 1).
If both of them are zero. Then we have a N = 0. bN = 0. a l . sinmz.bN f ( z ) = a0 and g(z) = c g + n= 1 an cos nx + b. which is contrary to aN or bN # 0. CM. sin nz + c M m=l + cm cosmz + d.g(x)) = N + M = d e g f ( z ) + d e g g ( z ) . Hence deg(f(z). b l ..g(z) are and a N d M : b N C M respectively. dM = b N . d$ = . 2 and so aNCM = 0. Let N . So T is an integral domain. then the degree relation implies that d e g f ( z ) = 0 = degg(x).if and only if all the coefficients of f(z). ao.e. . then a N . Hence the units of T are {*1}. Since (an cos n x + bn sin n z ) (cm cos m x + dm sin m z ) + mn=k c andm . . If f(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. Thus we have proved that bN(c$ + d&) + = U N d M = 0. i.bncm 2 ) sinkx] . Now f ( z ) . are zero.. .b N C M . aN and bo. . g(z) = 1. 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). . Since c h d& # 0 .
1 f f i s i n z are irreducible. y) ) # be the dual basis of u l . in T we have cos 22 = (cos z sin z)(cos z . First. . If p = 0 or ( p . . * * ~ (l).55 Obviously. Since A is Noetherian. all the factors cosz f sin%.Let zn . . a n the set of the conjugates of a. Now. j). Since ) L _> K is a finite separable extension. u2. j for all i. Hence T is not a unique factorization domain. we get TrL. 2 1 2. E B . c.u. = K [ a ]for some a E L. . ( Stanfod) Solution. The bilinear function ( 2 . it is easy to see that nLIK 0 # ( n ~ / K ( b = nb for any b E K ) . For any f(a)E L. 1314 Let A be a Noetherian integral domain integrally closed in its field of fractions K . Let ~ 1 . be the ~ +  minimal polynomial of a over K and j be the minimal positive integer such that p ] j and c j # 0. u 2 . Since xu. . (the elements in L satisfying T T L / ~ ( u ~ = ~ . . suppose that p # 0 and p I n.. we claim that the trace function ~ L I K 0.cj # 0.. + T r L / ~ ( z y is nondegenerate since ' I ~ L I K 0. . Then by using Newton's identities on the elementary symmetric polynomials. Denote a = a 1 . B is Noetherian. .. v 6 . v. be nL/K(f(a)) = i=l C f(ai)..sin z) = (1 a s i n z)(1. + + Again by the degree relation. n ) = 1. outline a proof that B is Noetherian.a s i n z). Ici = ~ L / K ( ~ u E ) A for any i.  + + + ) + . ( k i E B ) . .* u. K Let 211. Let p = Char(K) # and n = [ L : K ] . Then. a 2 . be a basis of L over K contained in B. . If B is the integral closure of A in L .&j) = (l)n+j+l * j .clznl + c n ~ z . 3 has the form lclvl : k2v2 . for any z E B . Thus cos 2% in T does not have an essentially unique factorization into irreducible elements. Let L be a finite separable field extension of K . . * knvn . Thus ~ L I # 0. 1. and cos z f sin z and 1 f d s i n z are not associates. Hence B C Avl + Av2 . + Av.
Suppose that f = implies that a? = 0. (a) Show that R. a i .e. bz. = 0 for any b l .b. in its field of fractions. ao. ak is nilpotent. i. then all the elements aa E A are 00 n=O nilpotent..  1316 Let F be a field. (Stanford) Solution. that if all the elements ai are nilpotent. + f k x k E F [ z ] * such that f1 = fi = .and determine the integral closure of R. is a Noetherian ring. prove the converse... that is. E I. and for n 2 1 let R. is F ( z ) . Since A is commutative Noetherian. then it is easy to see that f = 0. . = fn = 0. (b) Show that the field of fractions of R. . then f is nilpotent. . So. 03 It is easy to see that f = 0 " + + . be the subring of the ring of polynomials F[x] consisting of polynomials f0 + f i x + fizz + .56 1315 If A is a commutative ring and A[[%]] the ring of formal power series over is A . a0 a12 nilpotent and n=O c anzn is nilpotent. . (Stanford) . first we get that a0 is nilpotent. .. * . show that if f = c anzn is nilpotent. Conversely. + a k Z k Assume that is is nilpotent.b. Since A is commutative.. By induction. If I" = 0. " Hence f is nilpotent. Let I be the ideal generated by {ai I 0 5 i < m}. all the elements ai are nilpotent. If A is Noetherian. suppose that A is Noetherian and all the elements ai are nilpotent. Hence 2 n=k+l is nilpotent and so ak+l is nilpotent. blbz . I is finitely generated and nilpotent.
z is integral over &.1 + k * 4 E R.57 Solution. a ( z ) / b ( z )E F [ z ] .. E S). is F ( z ) . On the other hand. k E Z. For any q / p E as ( q E Z. S is a multiplicatively closed subset of Z containing 1. (a) Since F [ x ]i s Noetherian and s is finitely generated a &module. by Let Obviously. 1317 Describe all subrings of &.pl+qk .) ( Stanford) Solution. Hence F [ x ] is integral over R.. Let R be a subrings of&. = q . Let Z s be the localization of Z at the multiplicatively closed subset S . It is clear that the field of fractions of R. P P P So p E S and q / p E Zs. R. for any q / p E R. Hence R C Zs. a ( x ) / b ( z )is integral over F [ z ] . if a ( x ) / b ( x ) E F ( z ) is integral over R. (A subring contains 1 by definition. Then R _>a definition. (b') Observe that for any a ( x ) / b ( x )E F ( x ) . is a Noetherian ring by ArtinTate Lemma (or Eakin's Theorem).Then + 1. Since F [ x ]is an integrally closed domian. P . we may assume that ( p . since z is a root of Xn+' . p On the other hand. E R.zn+' E R n [ X ] . Hence F [ x ] is the integral closure of R. Obviously. in its field of fractions. It follows that R = Zs where 1 s={O#PEZ)ER}. Hence Z s C R. q ) = 1 and pl q k = 1 for some I .
.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 ) .
we have n 2 m. Show that the polynomial    (x2 + 4) .4)(x2 . (b) If G acts transitively. (g(zl).. y 2 . Since (x. Similarly. we have    ( . Thus m = n. ( 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.Z)(x. Define a relation on X by requiring x y if either x = y or the transposition (2. y m } two equivalence classes of zn be determined by z = x1 and y = y1 respectively. Since G acts transitively on X . . Hence is an equivalence relation. ~ ~ ~ .  is reflexive and symmetric. (which interchanges z.. 4z2). ( 2 1~ i ) g . a prime number.u(zi))= ~ e . then all equivalence classes are distinct and contain the same number of elements. G. Show the following. .' E . For any 1 < i 5 n. (z2. x . Hence m 2 n. ~(2.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 .2 real roots and 2 complex roots. and }{ y 1 .)) are n distinct elements belonging to the equivalence class determined by y1 = g(x1).Y)(Y.16) . 2) Suppose f E &[z] is irreducible and has degree p . show that the Galois group of f over & is the symmetric group S. . # Y. So all equivalence classes contain the same number of elements.Y # z). on p symbols. x 2 . 1) (a) By the defintion of .YY = ( z . (b) Let { z l . Solution.there exists g E G such that g ( x ) = y. y E X and leaves all other elements y) fixed) is an element of G ..2 ( Cohmbia) is irreducible and determine its Galois group over &. Thus { ~ ( z l ) . If f has exactly p . z ) . it is easy to see that is transitive. (a) is a n equivalence relation. .
Consider the conjugation automorphism onG.4)(z2. which is generated by all the transpositions.. r j E X. .and the graph of y = g(z) has the form Fig. Since Card(X) is a prime. By Eisenstein criterion.f4.16). 2) Suppose  P f) ( . Hence the restriction of the conjugation to E is an element of G and it is a transposition. say n. .2 is irreducible over &. So Card(X) = m . This maps f(z) to itself.e. z y. r p } . (z. since f(z) is irreducible and f ( r i ) = 0 = f ( r j ) . Let g(2) = (z2 + 4)2(z2.r p } of the (distinct) roots by q + q1x. i. .4)(x2 .. r p )is a splitting field of f(z) over & ( r i ) . The real roots of g(z) are 0. f2. .. there exists an isomorphism of&(r. f(z)= (z2 + 4)2(z2 . . m = 1 and n = Card(X). @[XI. . all the equivalence classes contain the same number of elements. Thus for any z. . = rI(z . Thus the Galois group G of f ( z ) over & is the symmetric group Sp on { T I . Then. Then q E Gal(E/&) and q(ri) = ~ j which shows G acts transitively . and also over Q ( r j ) . Since E =&(rl. Let r1 and r2 be the two nonreal roots of f(z).y)E G. where m is the number of distinct equivalence classes determined by .)/& into&(rj)/& by sending ri to r j ..1 + .y E X .this isomorphism can be extended to an automorphism q of El&.l. and n > 1. r P )is a splitting field of f(z) over & contained in a. on X . It follows that G is the whole permutation group of X .Ti) i=l in So E = & ( T I . For any r i . n.16) . Identify G = Gal(E/Q) with a permutation group of the set X = { T I . Thus the conjugation interchanges r1 and r2 = Fl and fixed all other real roots. by (b).60 (c) Suppose G acts transitively and contains a t least one transposition. .
up’. a divisor of p . Let Cp be a primitive pth root of unity and g a primitive root ( mod p ) (i..1a(base)for&(<p)/&. 1402 Let p be an odd prime. . = @(cP): InvH] = IHI = f and [Inv(H) : &] = e .61 Since I g ( n ) l > 2 for any odd integer n. Obviously.aP’ = 11.. for any i. (Columbia) Solution. = ~ } is < p Let H =< C T ~>< G. Now for any i. C pP . g is a generator for (Z/pZ)*). CT : Cp H (&. Then J H J f. Define f 1 . Since g is a primitive root ( mod p ) .)g is an automorphism of &(GI over & and G = Gal(&(<. Hence the Galois group of f ( z ) over & is ST. a 2 ( ~ p ) .l)/e. Hint.it is easy to see that f(z)= g(z) .Fix e.2 has five real roots and two nonreal roots. Since H is normal in G. CT’.e. { a ( < p ) . . Inv(H)/& is a Galois extension and Gal(Inv(H)/&) 21 G / H .)/&) =< CT >= {c.1 and put f = ( p . ~ ~ ~ .qa = C(<p)gej+i j =O Show that. Use the generator CT : Cp + ( C p ) g of the Galois group of&(Cp) over &. vi generates a subfield of &(Cp) of degree e over &.
is a primitive element for Inv(H)/Q. gcd(A. Show in succession that the polynomial A .B) = 1.1 are distinct. Put G = Aut(K(z)/K).62 and f I j=O fI j=1 f1 . B E K [ X ] ..AB 6 K . and put y = A(z)/B(z).e. (a) Let A . that z is is + ... Thus qi generates a subfield of &(&) of degree e . 1403 Let K be a field and z be an element of an extension of K such that 2 is transcendental over K . Hence j=O n (z q j ) is the minimal polynomial for qi over Q (for any  i). It follows that 71. Inv(H) =@(qi) for any i. and let H denote the subgroup of G consisting of the substitutions 2 + 2 b with b E K .yB E K ( y ) [ X ] not 0. e. i.
Hence y is transendental over K .bc # 0. B ) = 1.where K* . .y B is irreducible in K ( y ) [ X ] Thus the minimal . (c) By (b).e.63 algebraic over K ( y ) . Thus A . polynomial of z over K ( y ) is the monic polynomial which is a multiple in K ( y ) of A . which is contrary to A B 6 K .y B ( z ) = 0.y B is irreducible in A K [ y ] [ X= K [ X ] [ y ] . (b) By (a). ad .z)qZ+l. deg B}. z must be algebraic over K .12. z is algebraic over K ( y ) . deg B } = 1. then ordG = q 3 .yB and (c [ K ( z ): K ( y ) ]= deg(A . Conclude that n. ( CoZumbia) Solution. (a) Since gcd(A. Since A(z) . (b) Show that the elements of G are given by the fractional linear substitutions z +a z + b/cz d with a . Again.y B = 0 in K ( y ) [ X ] . (c) Show that when K is infinite. d E K and ad .that y is transendental over K . .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 ) . E K [ X ]such that A S + B T = 1 T in K [ X ] . y = A ( z ) / B ( z )is a generator of K ( z ) / K (i. then G is infinite and the fixed field of G is K . since gcd(A. If y is algebraic over K . it is easy to see that G N G L z ( K ) / K *. c. B ) = 1 in K [ X ] K ( y ) [ X ] ) .Suppose A .y B # 0.q and the fixed field of G is n. which is contrary to the assumption.y B is irreducible in K ( y ) [ X ]and that . Find the fixed field of H .h(z). that A .z)q+l/(z' . b.deg B } . K ( y ) = K ( z ) )if and only if max{ deg A. [ K ( z ): K ( y ) ]= max{deg A.a) 1 a E K*}.. ] Thus A . or if and only if y has the form az+ b/cz +d. Find the fixed field of H .bc # 0. Then It follows that degA = 0 and deg B = deg(A) = 0.y B ) = max{degA. there exist S. I2 = {diag(a. (d) Show that when K = lFq and put + z (zq2 .
On the other hand. deg B } = q3 . the fixed subfield of G. .(z) : Inv(G)] = q3 . Thus Inv(G) = Fq(z) and Fq(z) is a Galois extension of a given field L with FP C L C F. K ( z ) / K ( y )is a finite dimensional simple extension.1 1 2 + . Hence we have [Fq(z): Fq(z)] = q3 .x)q+l/(XP . Then c and [F.q .X)q2g. it is routine to check g(z) = g((zq2 . . we have (XP2 . . Similarly.X)n”n A = (1 + XP’ + X(P1)2+ .4. On the other hand. then by (a).(z) if and only if L _> F ( z ) . .Z)Q+1/(2~2)q2+1) 2.q . Obviously. For any g :2 H az + b/cz + d . we have Inv(H) = K .X)q2+1 = ((xqx))q+l.+ x ( n .X)+P Denote and (1 + x . (ad . Thus we have Inv(G) = K . . which is contrary to the facts that G AutK(y)K(z) and G is infinite.bc # 0) in G. if there exists some y = f(z)/g(z)E Inv(G)\K. (d) Suppose K = Fq. since H is also infinite. (xq2 X)  1 (XP . + X(PQ?)q+l n1 (XP . Hence AutK(y)K(z) is finite.64 If K is infinite. + X(Q1)Q)q+l B = (Xq. Then gcd(A. B ) = 1 and max{deg A. K C Inv(G). so is G. = Hence F ( z ) c Inv(G).
. and that the subgroup G of Aut(E) which these generate is isomorphic to D.C Y ) . >rx D. Obviously.Y")(X" .(z') = F q ( z q . Then z' E Inv(H) and by (a).(Y') * (X  y1yl) . E is a splitting field of X2" .Y)(X ..2 . and = earriIn. It is readily verified that X2" ZX+ 1 is irreducible over F (Z is transendental over C and X2" . (c) Obviously.ZX 1 since + + + x2n  zx + 1 = ( X . (Columbia) Solution. = (T" = 1 = r 2 . Hence c and T are automorphisms of E/C. .l .z E Fq(z). we have [Fq(z)Inv(H)] = q since IHI = q. ( X . F = C ( 2 ) with Z = Y" + Y"..there are unique homomorphisms u and T of E to E over C such that . 1404 Let E = C(Y) with Y an indeterminate.C"'y) (X.65 Similarly. G =< u .z). E/Inv(G) is a Galois extension with Galois group G and [E : Inv(G)] = 2n. (b) Show that Y is a root of a polynomial of degree 2n with coefficients in < F. (a) Show that there are unique automorphisms u and r of El@such that a ( Y ) = CY and T(Y) = Y . using 1403).(Y) = CY and T ( Y ) Y . Thus = Inv(H) = F.for example.y')(x . the Dihedral group of "' ~ order 2n. Y is a root of X2" . Since u ( Z ) = u(Yn + Y") = Z and T(Z) = 2. [Fq(z) : Fq(z')] q.ZX I. we have Z E Inv(G) and F = C(Z) Inv(G).Y") E F[X].the Dihedral group of order 2n. ord(7) = 2 and Tu = uT in G.X : q and B = 1 in F.(Y"+Y") + 1 = (X" . (c) Show that EIF is a Galois extension with Galois group G. (b) Since X2" . (a) Since Y is a generator of E over C.l . Since ord(u) = n.[X] and let z' = zq . x + 1 = X2" .Z X 1 is irreducible in C[Z][X]. Let A = X .
. It follows that E / F is Galois and [E : F ] = 2n. the quotient field of the polynomial ring&[z]). K C_ F is a Galois extension and [F : K ] = IGI = 4.1)2 + E K . On the other hand.z and = 7(z) = A. there are exactly 3 subfields lying strictly between K and F which correspond to the three proper . field automorphisms of F ) given by ( ~ ( z ) 2 . (b) If K is fixed field of G in F . (c) How many subfields lie strictly between K and F ? How many of these are Galois over K ? Justify your answers. + f ( t )= t4 . Let (x11)2 5. (a) Since ( ~ ’ ( z ) = z. It follows that (c) By the fundamental theorem of Galois theory.e.e. Denote 7 = (z . 1405 Let F = &(z) be the field of rational polynomials in one variable x over & (i. (Indiana) Solution.1)’ &. (b) Obviously U ( Z . ?(z) = z and T ( T ( Z ) = = ( T T ( z ) . Hence where K 4 is the Klein 4group. find a finite subset S of F such that K = &(S).. Then&(q) 2 K F .66 in E .1 (a) Find the subgroup G (of Aub(F)) generated by (T and 7.we have (T’ = 1.1) = z It is easy to see that ( z . the fixed field of G in F . Consider the elements (T. 2. r 2 = 1 and UT = T U . Then f ( t ) is irreducible in & ( q ) [ t and F is a splitting field of f ( t ) since ] in F [ t ] . Hence&(q) C F is a Galois extension and [ F : &(71)] = 4.qt2 + 1€&(V)[t]. Hence we have F = Inv(G) and E / F is Galois with Galois group G.1) = 1 .7 in A u b ( F ) (i.x and ~ ( .
T(T = C3T. u and the Let be elements of Autq(&(t)) given by u ( t )= and ~ ( t=) t.F ) 1 = ( . field of quotients of the polynomial ring @[t]. So + . :) + & ( t 2 + $) of is irreducible over & (t2 $) and Q ( t ) is a splitting field over separable polynomial f(z). It follows that G is the fixed field of H .t 2 ) (2 .f . 1406 Consider &(t). = (2. Hence. (a) Obviously. (Indiana) Solution.t ) ( z+ t ) (x z . Thus all these 3 subfields are Galois over K. It is easy to see that r2 = 1. Since all the subgroups of K4 are normal. CT3 ( t ) = and a4(t) = T 3 (E) 3 t. Let G be the subgroup of Aub(&(t)) generated by u and T . T 2 ( t ) = t . Find a E @(t) such that & ( a )is the fixed field of H (Justify your answer). and obviously [&(t) : InvH] = (HI 4. (b) Let H be the subgroup of G generated by u2 and T . (a) Identify G. u4 = 1 and isomorphic to the Dihedral group (b) Since D4.) (. u 2 ( t ) u = = . all these 3 subfields are normal over K (finite dimensional and separable).67 subgroups of G K4. = On the other hand.
' c) Find a generator of the Galois group Gal(K/&). This is a contradiction. a) If f(z)= x3 . But fl are not roots of f(z).[& + + + . Hint. Thus f(x) is irreducible over &. cr + p + 7 = 0.I. then f ( z ) has a factor with degree one in that is.32 1 is reducible over &.3a: 1 is monic. Obviously. f(x) has a root in &.32 + 1 in a splitting field of f(x) over K. b) Let . f ( ) = x3 . the rational roots must be integral factors of 1. InvH =& (t2+ $) 1407 Let & denote the field of rational numbers.3x 1.68 is a Galois extension and [&(t) :& (t2 + $)I = 4. It follows that that is. Consider a .2. of. Since f(2) = x3 . Hence /3 y = a and . crPy = 1 and the discriminant A of f(2) is 34.rbe the other two roots of f ( 2 ) = z3 . b) Prove that K/& is Galois. (Indiana) + Solution. a) Prove that f(x) is irreducible over &.B. let K = & ( a )where a is a root .
(b) As Kmodules. So K is a splitting field of f ( x ) . K [ z ]and K ( z ) have what dimensions? (c) Let G denote the additive group of K. 1 a E K } is linearly dependent over K .(Y 2 and . Show that the Gorbit of f spans a finite dimensional Kmodule if and only if f E K[z]. . it is easy to see that the Galois group Gal(K/&) F A3. Thus { f a = I a 6 K } is linearly independent. ( Y . . (We may check a2 . Hence K / Q is Galois since Char(&) = 0. c) From a) and b). (b) dimK K[x] = dimK K ( z ) = 00. Let f E K ( z ) . (c> Suppose f ( x ) = a. .(Y = a2 . a.2 are the roots of f(x).zn & + anlxnl + + a12 + ' . acting on K ( x ) by a E G sending z to z a.. Assume that K is infinite. (a) Suppose that {fa = There exist some nonzero elements ( Y ~ . (Columbia) Solution.K and some distinct eleE ments a l . E K such that + n Hence and i#i which is not true..* a0 E K[z]. u2.2 is a root of f ( x ) directly by using the hint). = a2 . . . (T: + A K=&(a)+K ak+(Y22 is a generator of Gal(K/&). . . (a) Show that the rational functions f a = ( a E K ) are linearly independent over K.69 It follows that (Y. 1408 Let K be a field and 3: an indeterminate.
f ( z + a 2 ) . . . ./32. a. . Let f ( x + a l ) .*.f ( z a2).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) . a. . a n + l in K (note that K is infinite).. .. suppose f ( z ) E K ( z ) .f(x . f ( x a l ) . .*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 . . .+Pn+l(~+an+l)~ + . f ( x an+l)spans the subspace generated by the Gorbit of f . + P n + l f ( 2 + ~ 1 n + l ) .. there exist PI. { f ( z + a ) 1 a E K } spans a finite dimensional Kmodule. . a 2 . +an)(ai E K ) span the subspace < {f(x+u) I u E K } >. and the Gorbit o f f . .. . .P 2 . . . .. a. . .h ( z ) E K [ z ]and ( g ( z ) h ( z ) )= 1. f(. ..+1 .+a) This shows that { f(x + a i ) 1 1 5 i 5 n + 1) spans the subspace generated by the Gorbit of f . we take (. .. Then for any a E K .. For any a E G. 1 ant1 a1 . .01.. " . We write f(x) = g ( z ) / h ( x ) where g ( z ) . = P ~ ( Z + $ I+)P~ ( x + a 2 ) " z and also (Z +. + Pn+l(z + Qn+1li for 1 5 i < n.. a. Hence = P l f ( z + a l )+ P a f ( z + a 2 ) + . On the other hand.70 We claim that for any n 1 distinct elements a l .)a =P ~ ( Z + alli + /32(z + a2)a + .'..+1 a: a. .
q(P) is a root of q ( q ( z ) ) = g(z). Note that y ( P ) may not be P. Thus all the irreducible factors of f ( z ) over E are of the same degree. ( c ozum bia) Solution. Since K is infinite. 7 : E ( a ) + E(a’) is an isomorphism and ~ ( a )a’. We prove in the following that all monic irreducible factors of f ( z ) over E arise in this way. where . an easy discussion in a splitting field of h ( z ) over F will lead to degh(z) = 0.Show that all the irreducible factors of f ( z ) over E are of the same degree. but :E E is in G. Since E / F is Galois.h ( z + u)) = 1.71 Since (g(z + a ) .Let e(z) be a monic irreducible factor of f(z) over E. which is irreducible over F . + . B () a separable irreducible polynomial over F . u(e(z)) is an irreducible factor of f ( z ) over E . we can write E = F ( P ) . Since f ( z ) E F [ z ] . for any u E G. Then.a) for any a E F . Then. Obviously E is a splitting field of g(z).)I i=l n n h ( z +(. CY and a’ are roots of f(z). For any (T E G = Gal(E/F). Hence 7 : F ( a ) + F ( a ’ ) can be extended to an isomorphism 7 of E(a) onto E(a’). we still denote u to be the isomorphism E[z] t E[z] which extends (T on E and maps x to z. Thus we have 1409 Let E be a finite Galois extension of F and let f(z) be an irreducible polynomial in F [ z ] . Suppose e’(z) is another monk irreducible factor over E. since a and a’ are roots of the monic irreducible polynomial e(z) and e’(z) over E respectively. Hence we have an isomorphism 7 : F ( a ) + F(a’) which sends a to a (u E F ) and CY to a’. we have h(z +. It follows that Now. 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. a(f(z)) = f(z). is a root of g z . = we must have (vlE)(e(z)) = e’(z) and dege(z) = dege’(z). Let a and a’ be roots of e(z) and e‘(z) in some extension field of E .
u)E K . Let E be a splitting field of tP .1 over Zg.u . First prove that the zeroes of t62 . Comparing the coefficients of the term of degree n .t .. Show that the polynomial tP .1. 1 5 n 5 p . (Indiana) S o htion. f ( t ) = (t .t .1 in F .t . Show that [ F :&I = 3. Thus we have tptc= (tum) +.in). u = 0 and there exist integers v and w such that w .il)(t . Obviously.t c = f ( t ) . u E K . G is a subgroup of F" and G is cyclic of order 6 2 since (t6' It follows that . u = (v . u il + i2 .1)' = 62t6' = 2t61 # 0.c then so is u+ 1.c and let u E E be a root of this polynomial.1. we obtain n . in E K . + + + + in K [ t ] . n wp) u = v(n .u . Then for any m EZ. Suppose that t p . Hint. .1 form a cyclic group G of order 62. .i2) * * * (t . g ( t ) in K [ t ]where f ( t ) is a monic polynomial of degree n.72 1410 (a) Let K be a field of characteristic p > 0.c in K [ t ]is either irreducible or splits completely into p linear factors over K. If u is a root of t P . Since p . Hint. (b) Let G be all the zeroes of t62 . [ F :as]2 3. (b) Let F be the splitting field of the polynomial t62 . n w p = 1. So we have n .u . the prime field of K .
1412 Let K be a field of characteristic p # 0. then [ K : K'] = 2" and [K' : L] = m. Let f(a:) be the minimal polynomial of o over L. We have to show [ K : L] is a power of 2. Then E* is a cyclic group of order 53 . [ K : L] = [ L ( a ): L] = d e g f ( s ) is even. So t6' . Let GI be its unique subgroup of order 62. K a'. Since L has no proper odd dimensional extension field. Then all the elements of G' satisfies t6' . For this purpose.1 splits in E and E is a splitting field of t62 . Hence. m where m is odd. Then f(a:) has no real root since f(a:) is irreducible in L [ z ] . we must have m = 1. . the subfield of 1R of numbers which are algebraic over &. = 1411 (a) Suppose you are given a field L . If K' is the corresponding subfield of K I L . X Now for any finite field extension K I L . If [L : KP] is finite. Show that every finite field extension of L must in fact have degree equal to a power of 2. such that L/& is algebraic and every finite field extension K / L . f ( z ) . prove that it is a power of p . (Indiana) Solution. (b) Let L be the field of real algebraic numbers. K 5 a' is of even degree. it is in fact a power of 2. (a) Let K / L ( K C a ' ) be a finite field extension. there exists some element a: E K such that K = L ( o ) by Primitive Element Theorem. Then & 2 L & ( and L/& is algebraic. we may assume that K is Galois over L.1 = 124. G has a subgroup H of order 2n. is a product of irreducible polynomials of degree 2.1. Thus we have E N F and [F :as] 3. So. Let G = GalK/L and JGJ= 2n . when decomposed in nZ[a:]. and so K' = 1. (b) Show that such a field L actually exists. & C_ L a'.1. (a) Let L be an intermediate field between Kp and K . that is. By Sylow's Theorem.73 Let E be a extension field of Z5 such that [E : Z5]= 3. By (a). The set {a:" J a: E K } is a subfield of K that is denoted by KP (no proof required). and [K : L] = 2n.
For example. (b) If KP # K . of distinct elements of B . K contains a maximal pindependent subset. . .. .b2. b2.. Now suppose that {Bi I i E I} is a chain of pindependent subsets of K . * .. ...bil).l)] ) i=l = p”. b 2 .bn} such that = KP(b1.bp is irreducible in KP(b1.. Since .. then K contains a maximal pindependent subset B... if 6 E K\Kp.. there exist pindependent subsets. bi1) = KP when i = 1).b2. . b 2 .bn) L. (c) Prove that the set B of part (b) satisfies K p ( B ) = K . Hence [KP(bl.bil)[t]. . b. .. . b m } C Bi.) : L] n = n [ K P ( b l ... Prove that if KP # K .. * * . Then for any finite set { b l .. By Zorn’s Lemma.bi1) for any ! 1 5 i 5 n (KP(b1.b. we can assume that bi $ KP(bl. there exists some i.. and bi @ KP(bl. It follows that [L: KP] 1 [KP(bl. such that { b l . ...b..bzl)] = p.b. B = { b } is a pindependent subset of K .. .. b 2 . . Let B = U Bi.74 (b) A subset B of K is called pindependent if for any finite set bl.*. (a) If [L : KP] is finite. b m )K P ) =pm. (Indiana) Solution. bz) : KP(b1. . . . . .. Since Bi is apindependent subset.. : It follows that B is pindependent. [ K P ( b l .} of distinct elements of B . b i : KP(bl. . there exists a finite set of elements {bl. .... . i€I . tP . . Without loss of generality.+ .
.b.b. . bml)] = [KP(bl. the independency of B U { b } contradicts the maximality of B . If m is an integer which divides n and f(t) E K [ t ]is an irreducible polynomial of degree m.l} say. show that f divides t P r n . Prove that if p does not divide [E : F ] then E L. . with E / F Galois and [K : L] = p . b = b. b. (Indiana) . Let (Y be a root of the irreducible polynomial f ( t ) in some extension field of K. 1413 Let K be a finite field with pr elements ( p a prime) and n be a positive integer.. . we have f ( t ) I ( t P r n . E contains a subfield L such that K & L E and L N K ( a ) . . b 2 . Suppose K p ( B ) c K . then {bl. b 2 .t ) . (Indiana) Solution. Hence there exists an element p E L C_ E such that f(P) = 0.* and if b E {bl. . [KP(bl. . CB and we still : KP] : KP(b1.b.. a prime.p. if b @ { b l .}...t over K . then and [KP(bl. Since m I n. . Then IK(a)l = prm and [ K ( a ): K ]= m.. in B U { b } . Let E be a splitting field of t p r n . b m } . . The reason is that.l) : KP] Since B C B U { b } . for any distinct elements b l . bm) : KP] = p m ... . . Since pp'" ..t . f ( t ) is the minimal polynomial of p. have [KP(bl. Then (El = prnand [E : K ] = n.p = 0.bz... that is..b. Thus we have K P ( B ) = K . Then B U { b } is pindependent. .* . 1414 Let K I F be a finite extension of fields and let L and E be intermediate fields.75 (c) Let B be a maximal pindependent subset of K ... .) ..) * 1 . Let b E K\Kp(B). .b.* .*"l = p"* .
a p = IGal(E. (Indiana) Solution. L/L)I = IGal(E/L dividers IGal(E/F)I = [E : F ] . Let a E E be a root of f(z). Then E = F ( a ) and E . Let f(z)be a separable irreducible polynomial over F such that E is its splitting field. Let i=O (1) Prove K is a field. Then E L = L(E) is a splitting field of f ( z ) over L. It follows that C E . . since L Now suppose E prime. (2) Let f(z) E K [ z ]be irreducible. If i the smallest subfield of @ containing the set 2 0. Prove that deg(f) 2 5.76 Solution. For any c E G a l ( E . L and n E)I [K : L] = p . contrary to the assumption. L/L) Gal(E/L n E ) . And further. Then E . L = K . we have N_ G a l ( E . Ki+l is (0 €6' 10" E Ki for some n > 0). L = L(E) = L(a). 1415 Let Ki be the subfields of a' defined as follows: KO = &. Let E * L = E(L) L ( E ) be the composite of E and L in K . L/L) = Gal(L(a)/L). It is clear that i=l u Ki is a subfield of a'. + it is clear that g (E ~ Aut(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'. Thus we have E C L. L. Hence E L is Galois over L.
hence in K [ z ] . if aPml I . 1416 Let K be an extension field of Fp. where g ( n ) = p"l. of . : Thus [Fp(a) F] is the smallest positive integer m such that : E Fp. zPm  so u is a root of m zp . Let a be an algebraic element in K . 1417 Let F 2 K be a field extension of finite degree m.77 (2) (Remark: d e g f ( z ) may be 1). Contradicts the irreducibility of f(z).(u) : F. Let n = [ F p ( a ) Fp]. Since ( ~ ( g ( ~ ) ) P = : l apnl = 1. a d n ) E Fp. E Fp. Let f E K [ t ]be an irreducible polynomial of degree n. 1 5 k < n. There exsits some i such that f(x) E Ki[x]. By the formulas for the roots of quadratic. If m and n are coprime then show that f remains irreducible in F [t]. (Indiana) Solution. then On the other hand. the field with p elements. Let E be a splitting field of F p ( a )C E .(&4)Pl = 1. Prove that [Fp(a) Fp] is the smaIlest positive integer : m such that E Fp. Then [E : Fp] = m and Fp n = [F. Hence deg f ( z ) 2 5. Let f(x) E K [ z ] be an irreducible polynomial with d e g f ( z ) > 1.x. Let E = F(cr). P.h ( t in F[t] and ) where g ( t ) is a irreducible polynomial in F[t] degree k.1 (Indiana) Solution. Suppose that f(t) is reducible in F[t] let f ( t ) = g ( t ) .Then IFp(a)I = p" and upn' = 1. for some positive integer m. Suppose d e g f ( z ) 5 4. cubic and quartic equations and Ki+l 1 (0 E(C 10" E K . f(z)splits in Ki+3[z]. Hence z over Fp and a E E .]([EpP]= m. for some n > O}.
4(n) is the Euler $fur Ition.z . it is well known that Aut(G) .)/&) N Aut(G) where G is the cyclic group of order n. if we take E = Inv(&). then & C E is a Galois extension and Gal(E/&) 1 .m. It is easy to see that IGal(&(zi)/&)I = +(n) and Gal(&(z. then E is a Galois extension of & with Gal(E/&) cyclic of order 16.n ) = 1 and 1 5 Ic < n. When n = 2" and m 2 3. since cr is a root of f ( t ) and f ( t ) is irreducible in K [ t ] . It is easy to see that IGal(&(td)/&)I = 4(n) and Gal(&(z. : n. As in 1418.It follows that n I Ic .Zz @ . Thus f ( t ) is irreducible in F [ t ] .. On the other hand. d(n)is the Euler &function. By the Fundamental Theorem of Galois Theory. 1418 Find a Galois extension E over & with Gal(E/&) cyclic of order 16.1. E = & ( z 1 7 ) . which contradicts ( m . then & E is a cyclic extension of order 32. (Stanford) Solution. Then [E : F ] = Ic since F [ t ] . So if we take n = 17. the cyclotomic field &(zn) over Q is a Galois extension and [&(zn): Q]= +(n) where z.)/&) E Aut( ) where G is the cyclic group of order n. (Stanford) Solution. : Taking m = 7. is an nth primitive root of the unit. [K(a!) K ]= [E : K(cr)].Z2.m . Aut(G) is cyclic of order n . the cyclotomic field & ( z n ) over & is a Galois extension and [ & ( z n ) : &] = d(n) where z . Z p . is an nth primitive root of the unit. for any positive integer n. . For any positive integer n. [E : K ] = [E : K(cr)]. 1419 Find a Galois extension E over & with Gal(E/&) cyclic of order 32. When n is prime. .So [E : K ] = [E : F ] [ F : K ] = Ic . .78 g ( t ) is irreducible in where cr is a root of g ( t ) in some extension field of F .
Anyway.@) in E.1. we have  n i=l .uz(.an(z)} in Mn(E)..say. M .(z) = a z .. . Show that its trace is given by Trp(M. Obviously. Z ( z ) .l } is a base for E / F . and u 1 ( z ) . for any a E E . ..79 1420 Let E / F be a finite Galois extension.. Let z be a primitive element of E / F . g n ( ( z )are distinct.z n . : E + E . To prove that it suffices to prove that for any 1 5 k 5 n . f(z) is also the minimal polynomial of the Flinear map M . [E : F ] = n and Then .a. (z) = z . 3c. . the matrix of M . G = Gal(E/F) and a E E .+ . .. . M . . is the minimal polynomial of z over F. relative to the base (1. (E M n ( F ) ) .. Consider the Flinear map M a : E + E. z . (Columbia) Solution.z).. is similar to diag{ul(z). ~ .. a has the form since { l ..znl}.) = C g ( a ) where varies over G. Now. Since f(z)has distinct roots C T ~ ( Z )g.
and for any 1 5 i 5 n .1. . n This completes the proof.
Part I1 Topology .
.
we may assume ~ that A c f'(l). ~ We have f(zo) = 1. By the same reason. Hence f is not surjective. If A and B are path connected. It means that A U B is connected. say U . Let f be any continuous map from A U B to So = {1. 2102 Suppose that A and B are compact subspaces of spaces X and Y respectively.l). f ( must be constant.a3 SECTION 1 PO IN T S ET TOPOLOGY 2101 Let A and B be connected subspaces of a topological space X. Without loss of generality. there exists a neighborhood of zo. Prove that A U B is connected. Let zo E AnB. need A U B be path connected? (Indiana) Solution. Let A = ((0. Therefore we have B c f'(l). and that N is an open neighborhood of AxBcXxY.O)) C R2 and 1 B = {(z. Since A is connected. such that A n B # 8. The following example shows that if A and B are path connected then A u B needs not to be path connected.sin ) I 0 < z 5 1). Prove that there are open sets U c X and V c Y such that . there is a point of B which belongs to U . 3 : Then A and B are path connected and A n B = A . Since f is continuous. But since zo E B. f l is also constant. such that U c f'(l). but A U B is not path connected.
. ( z ) c X and V. Since X is locally compact and Hausdorff.(z) c Y such that ( Z C . q } such that A Let U = sets of j=1 u U(zj). . Since B is compact. there exists asubcover { U ( z i ) .1] such that f l = 0 and f l = l? ~ ~ + ( Cincinnati) Solution..since (z. y) belongs to A x B and N is an open neighborhood of A x B c X x Y . Hence we may assume that X is noncompact. with A compact and B closed. Let A and B be disjoint subsets of X . For any y E B. U U U ( z j ) and V = j=1 q n V(zj). . j = l . Does there exist a continuous function f : X [0. ' Then by the Urysohn Lemma there exists a continuous function X + [0. . 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. then X is normal and the existence o f f is obvious. It is easy to see that A and F = B u{m} are two disjoint closed subsets of X .(z).84 Ax B cU x V cN.(4 cN. i=l P Let U ( z ) = fi Uyi(z) V(z) = 0 V. .il ..y E B } covers B. . { U ( z ) . Since A is also compact. p } such that B = c U V. 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 . If X is compact. We denote by X" the onepoint compactification of X. Y )E V. Let' z be a point of A . . (Indiana) Solution.(z) x V. and consequently is also a normal space. c On the other hand. 2103 j= 1 Let X be a locally compact Hausdorff space. Let X* = X U {m}. z E A } is an open cover of A . there is asubcover {VYi(z). X is compact ' and Hausdorff. Therefore the family of open sets {Vy(z).(z). there exist open sets U .1] ' F: . .
f. (b) Show that if X is compact.1]/ obtained from the cylinder X x [0. the join of X and Y is the quotient space * Y = ( X x Y x [ O ) 1])/ x ) where (z. (a) Let X be a topological space and be an equivalence relation on X.t) is equivalent to (d. Therefore. Y is Hausdorff and f : X + Y is continuous and onto. y. (c) Show that i f f satisfies the condition of (b) then f is a quotient map. (a) So * So is homeomorphic to the unit circle S’. (Indiana) Solution. only the correct answers to the question asked. ~ ~ 2104  No proofs. If X and Y are topological spaces.  2105 (a) Define quotient map. the restriction o f f on X .t’) if and only if { x = x’ and t = t’ = 0 or y = y’ and t = t’ = 1. X * So is homeomorphic to the quotient space X x [0. and S’ * So is homeomorphic to the unit sphere S2.satisfies = 0 and f l = 1. are required for this problem. (a) So * So and S * So are homeomorphic to familiar spaces.1] by collapsing X x (0) and X x (1) to two points p and q respectively.85 such that f l = 0 and ~ the requirements that  f l = 1. (b) Generally.y. What space ’ are they? (b) Describe X * So for a general space X . By the quotient map ?r : X + X / we mean the map which assigns to x E X the equivalence   . then f is a closed map. (Indiana) Solution. Define by X / the space of equivalence classes under .
if f : X + Y is a continuous map. therefore. (b) Let A be a closed subset of X .U . It follows from the continuity of f that f ( A ) is a compact subset of Y . (b) Give a counterexample to show that if X is not compact. Since X is compact. there is naturally associated an equivalence relation on X such that z1 5 2 if and only if f(x1) = f(z2). (Indiana) Solution. r becomes a continuous map. By the result of (b). i is obviously a 1 . Since f is a continuous function.1 map from the compact space X / to the Hausdorff space Y .  (b) Let X = R and Y = S1. W is compact too. (a) Let W = X . : X + Y is the continuous function defined f by j ( t ) = earit for t E R. the quotient space X / is compact. Then V is an open neighborhood of C in Y .' ( U ) is open in X . f ( W ) is a compact set of Y . If Y is homeomorphic to X / under the map i : X/ + Y and f = i o T then we call f a quotient map. Since W c f'(f(W)). and let U be an open neighborhood of f'(C) in X. (c) Let be the equivalence relation on X associated to the map f. More generally. f ( A ) is closed in Y .fl(f(w))c x w = u. Denote by [z] the equivalence class containing 2. Since X is compact. Hence i is a continuous 1 .86 class containing x. Hence f is a closed map. Let C be a closed subspace of Y . then W is closed in X . is obvious that It fyc)= {n I 72 EX}. and consequently is a closed set of Y because Y is a Hausdorff space. . (a) Prove that if X is compact then there is an open neighborhood V of C in Y such that f'(V) c U . Since Y is Hausdorff. Take C = (1) E S1. is a homeomorphism.f ( W ) . The quotient space X / may be topologized by defining a subset U c X/ to be open if and only if r .       N 2106 Let f : X Y be a continuous function from a space X to a Hausdorff space Y . Then we define a map i : X/ 4 Y by i([x]) = f ( x ) . Since f is onto. we see that f f'(v) = x .1 map. A is compact too. and. It is clear that f = i o T . Let V = Y . So f is a quotient map. Under this topology. then there need not be such a neighborhood V .
If yo $ U U V. Let X be a connected space. Let U and V be two nonernpty closed sets in Y such that U n V = 0. (a) Let R : X Y be the identification map and yo E Y be the point which A collapses to. n  one can easily prove that there does not exist such a neighborhood V. Then R'(U) and R . 2108 Let p : E B be a covering map with E locally path connected and simply connected. 2107 Let X be a normal topological space and A c X a closed subspace. f . where 1 1 U. by the normality of X . then.' ( C ) n €I in X .' ( V ) are two nonempty disjoint closed sets in X . let f : X +B be a continuous map. (b) Let X be a regular space which is not a normal space. Thus we see that R ( W ~ ) and R ( W ~ ) two disjoint open sets in Y and contain U and V respectively. 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. 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.n+). (b) Does this result hold if normality is replaced with regularity? (Indiana) Solution. Then the quotient space Y obtained by collapsing A to a point yo is not regular. are If yo E U (or V). n n Since nw lim lUnl = 2 = 0. 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 . we only need to take the sets W1 and Wz without the restrictions that Wi n A = 0. (a) Show that the quotient space Y obtained by collapsing A t o a point is normal.=(+n.87 Let U = U Un be the open neighborhood of f .
Take a point zo E X and let f l ( z 0 ) = eo E E . 2109 Let p : X +X be an nsheeted covering projection. For any point z E X.  < 00. . i=l fi p ( c ( z ) )is an open neighborhood of z in X . Suppose that U = {UA. i. we see that e l E pl(b0). Let W ( z )be an elementary neighborhood.(z). Let  where w ( )a path component of p .. By the assumptions p : E + B is the universal covering map.e. Therefore. Suppose that (In dian a) Solution. Prove that 2 is compact.. Choose ix is a U . E Ui. i=l Choose another elementary neighborhood V(z) of z such that v ( ~cc ) np(w).' ( W ( z ) ) such that Zi E w. f2 : X + g :E +E E be two lifts of f . It is not difficult to prove that A is bothopenandclosed in X. Since p is an open map. Then eo E p . we see that A = X. g f l ( z 0 ) = f z ( 2 0 ) . Thus there exists a deck transformation g such that g(e0) = e l . n . It is obvious that each & x is open and R(z) n (x) = 0 for () i # j . x I s f 1 ( z )= f 2 ( z ) } . let p .(z) n U i containing Zi. E U such that Z . n X is compact.l ( b o ) .' ( z ) = ( 1 . Prove that there is a deck transformation such that fi = g f 1 . Let f2(20) = e l . . Let A={ E . where bo = p ( e 0 ) . which means fi = g f l : X + E .X E A} is an open covering of X . by the connectedness of X . Let c ( z )be the path component of %. 2 .88 and f l . Z n } . (Indiana) Solution. W(x) is an pathconnected open neighborhood of x such that each path component of p .' ( W ( z ) ) is mapped topologically onto W ( z )by p . Since p f 1 = p f 2 = f . Thus. It is obvious that A is not empty.
and consequently X can be covered by a finite subcover of U. (Recall that BdA = 2n (X. U which means U = 1.. We use the reductio ad absurdum. Show that if C is a connected subset of X that intersects both A and X .1 map from + the compact space ( X . Since C n A c U and C n ( X . U ) .A ) c V . (Recall that 7 c U means that every set in the topology 7 is contained in U. I).) (Indiana) Solution. Then it is easy to see is T(z)c E(z) c Ui for i = l . Hence h is continuous at 2 0 . Since 7 c U .89 Let P'(v(zc>) = where that n (JW). h : (X.)to (X. U ) to the Hausdorff space (X. . We use the reduction to absurdity.A ) . i=l c(z) the path component ofpl(V(z)) for any i. It is obvious that h(z0) = zo and h ( U ) = U .A ) . continuous map. then C intersects BdA. Since X is compact. Then h is a 1 . For any point 20 E X . 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. 7 ) and ( X . Prove that 7 p U.. .can be covered by a finite number of V(zi). 6 Ui. Suppose that 7 c U. Let U be any open neighborhood of 3:o in ( X . X i = 1. Take U = C n x and V = C n ( X . Thus h is a homeomorphism from (X. 2111 Let X be a topological space and let A c X. ) (Indiana) Solution.U) ( X . Suppose that C n BdA = 0. nIt follows that p'(V(z)) c . We claim that h is a 7). U ) denote the topological spaces of X with respect t o 7 and U respectively.U is also an open neighborhood of z o in ( X .+ .. .A . . This contradicts the assumption. 7 ) . 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. Let ( X . ' T ) is the identity map of X. That is.m.
( A ) . + 4 6 which means that 06/4(zo) c O .a)5 E for some a E A } . d ) be a metric space. Take X = [0. E A such that d(z. i. But u nv = c n A n ( X  A ) = C n B ~ = 0.. d ( z . Thus we have + which means 20 E C . u ) < E .d(z0. . ( A ) such that z # zo for any n and 2. for . For any subspace A E cX and real number > 0. . Must C ( A ) be closed . we see that both U and V are nonempty subset of C.. (a) Let 20 be any point of O . there exists a point a E A such that d(z0. and. is not closed in X .} in C . ( A ) . Thus O . (b) If A is compact. It is clear that C = U U V and both U and V are closed subsets of C .d ) < E . z o ) + d ( z o .e. z o as n + 00. show that C ( A ) is closed in X . A which is a contradiction t o the connectedness of C. Then C a ( A )= [0. each z there exists an a.from the assumption. for a general subspace A of X ? (Indiana) Solution. Then. a ) > 0. 1).(A) = C. + a for some a E A . we give a counterexample as follows. A : d(z. ( A ) . that both U and V are open sets of C. a ) ~ d ( z . 2112 Let ( X . ( A ) is an open subspace of X . 6 = E . consequently. I g) . If A is not compact. Since A is compact.< ( E . (b) Let zo be any cluster point of C e ( A ) .2] c R and A = [0. So C . ( A ) is an open subspace of X . ( A ) . a ). ( A ) is closed in X . we may assume that a. By the definition of C E ( A ) . for any 2 E 06/4(z0). By the definition of O . (a) Prove that O . Then there exists a sequence {z. let O. without loss of generality.) 5 E. a ) < E for some a E A } . a.(A) = {z E {z E A : d ( z .
C l . { c } . X = U U V . then there would exist a nonempty proper subset U of Y which is bothopenandclosed in Y. The topology on Y is determined by the family of open sets 7 = { { a . el. Since X is dense in Y and U is open in Y . Since any neighborhood of the point a always contains the point c. then Y is connected.' l v ( N ) are open in. Since X is dense in Y and Y . f . a) Let N be an open set of Y.U ) would be an open set of X . This contracts the connectedness of X . and f : X + Y be a function so that flu and flv are continuous. therefore.b.' I u ( N ) and f . f . b) Give an example where U and V are not open in X and f is not continuous.Prove or give counterexamples to the following assertions: (a) If X is Hausdorff. c ) .Since flu and fv are continuous.A is nonempty. U and V . We give a counterexample as follows.' I u ( N ) and f . Y is not a Hausdorff space.l I v ( N ) are also open in X.U is open in Y . { b . A X . X . would be a nonempty open set of X. b.0).U is open in Y . a) This assertion is not correct. If Y were not connected. We give a proof to it as follows.91 2113 Suppose that X is a dense subspace of a topological space Y. On the other hand. (b) If X is connected. 2114 Let X and Y be topological spaces.A = X n (Y. b) This assertion is true. { a . But U and V are open in X . respectively. (Indiana) Solution. then Y is Hausdorff. It means A # X . Let A = X n U . Thus fF1(N) = flIv(N) uf'Iv(N) . show that f is continuous. a) If U and V are open in X . Let Y = { a . which is bothopenandclosed in X. Thus A would be a nonempty subset of X . because Y . But it is easy to see that the subspace { c } is dense in Y and is Hausdorff. c}. (Indiana) Solution.
O ) E R2 10 5 and 2 5 1) v = ((1. Thus X must be connected. Let X = U U V . and it is obvious that ql(y) n U and ql(y) n V are both nonempty open subsets of ql(y). b) Let X = [0. the unit interval of R. where U = { ( x .l ( y ) is connected. q ( U ) and q ( V ) are disjoint nonempty open subsets such that Y = q ( U ) U q ( V ) . f is continuous. Let Y = [0. f : X +R is defined by f(x) = {x 2 XEU. b) Is X necessarily connected if the map q is not assumed to be a quotient mapping? Justify your assertion.21. Then f (u and flv are continuous. ) The topology of X is induced from the topology of R2. But X is not connected. q . b) The following example shows that the assumption that q is a quotient mapping is necessary for X to be connected. but f is not continuous.which contradicts the connectedness of Y . ( Coturnbia) Solution.92 is open in X . a) Show that X is connected. XEV. Therefore.q . y ) = 1 for (1. .1]. and consequently. = For otherwise. U = [0. 0) = x for ( x . O ) E U and q ( 1 . Then q is continuous but is not a quotient mapping.y) E V . 1) and V = [l. Assume that ql(y) is connected for each y EY.Y Y ) = (ql(y) n U ) u (ql(y) n V ) . For each y E Y. which contradicts the connectedness of q'y. let y E q ( U ) n q ( V ) . there exist two disjoint nonempty open subsets of X . U and V such that X = UUV. a) Suppose that X is not connected. Then q ( U ) n q ( V ) 0.2]. and : X + Y be the map defined by q ( x . 2115 Let q : X t Y be a quotient space projection from a topological space X t o a connected topological space Y. () Since q is a quotient mapping.l ( q ( V ) )and U = ql(q(U)). Thus. V = q .y) E R2 11 5 y 5 2.
Let f : 2 + So be any continuous map. and let 2 be a set such that Y is a subset of 2 and 2 is a subset of the closure of Y . Let U c X x Y be a closed subset and zo E X . with Y compact. without loss of generality. we may assume that f ( Y ) = {l}. Prove that 2 is connected.(zo))nU = 0. Show that p is a closed map. But c as wellknown. i. Y c f'(l). (Y)z= Y n 2 = 2. Taking closures of these two sets with respect to 2 and noting that f'(l) is a closed subset of 2. Since (W(z0)x Y )n U = 0.(~~). for any y E Y . Since Y is connected.p ( U). Let p : X x Y + X be the usual projection onto the first factor.VYn (zo) such that n y = Let i=l u VYi(XO).(y) of X and open set V. . = ) i=l n n Then W(z0) is an open neighborhood of z o in X . 2117 Let Y be a connected subset of the topological space.e.i. we see that W(z0) n p ( U ) = 0. .Then. Thus we have 2 c f'(l)..p ( U ) . p ( V ) is closed in X .93 2116 Let X and Y be topological spaces. there must exist a finite number of V. W(z0) C X . w ( ~ ~w~. According to the assumptions. we have Y c 2 c F. ( Cincinnati) Solution. there exist an open set Wz. (Minnesotu) Solution.(ZO) and (Wzo(y)x V. (zo). Since U is closed.. Since Y is compact.1} is the Odimensional sphere with discrete topology.p ( U ) is an open set of X .(ZO) of Y such that ( Z O . ( 2 0 . we get (Y)z f'(l). and it follows that f is not surjective. ' . which means that p is a closed map. Thus X ..y) 6 U . where So = {1. Y ) E Wz0(y) x V.e. and consequently. It means that there does not exist any continuous surjective map from 2 to So and therefore 2 is connected.
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. If C is a component of X\A. Must X have property S? ( Cincinnati) Solution. Suppose the closure of X has property S.we see that C U U is a connected subset of X\A containing C . Since U is connected. has diameter < ~ / 2 . and if V is a bothopenandclosed subset with respect to X\U. Then.Since each V. Now suppose that X\C is not connected.94 2118 Let A be a connected subspace of a connected set X . C u U is connected because C is connected. U is a connected subset of X . we may assume that A C V. then U U V is connected.. Hence W1 is a nonempty bothopenandclosed subset of (U U V )U (X\U) = X . which contradicts that C is a component of X\A. 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. Let X\C = U U V where U and V are two disjoint nonempty bothopenandclosed subsets of X\C. So the above statement is proved. a bothopenandclosed subset of X\U. which contradicts the assumption that X is connected. is connected. Since C U U c X\A and CnU = 0. without loss of generality. Then let U U V = W1 U W2 where W1 and W2 are two disjoint nonempty bothopenandclosed subsets of U U V. we may assume that U c W. a) Prove a metric space X has property S if X has a dense subset with property S. By the above fact. ( Cincinnati) Solution. Suppose that U U V is not connected. a) Suppose that X has a dense subset A which has property S. Hence X\C must be connected. b) Suppose X is a subset of a metric space. for any E > 0. consequently. show that X\C is connected. Since A c X\C and A is connected. Thus W1 is a bothopenandclosed subset of V . is also connected va .
2.lr/z) 10 < z 5 1) u ((0. (i) X is shown in the Figure below. x 2120 Let X be the topologist's sine curve defined by X = {(z. Show that either f ( X ) = X or else there exists 6  > 0 such that 3 3 < y 5 . (i) Sketch X .sin.1 (ii) Let A = f ( X ) n { ( z . b) We give a counterexample as follows. but it is easy to see that X does not have property S. 22 (Toronto) Solution. Then = R has property S. Fig. X be the set of all rational numbers. c R 2 . Let R denote the euclidean real line.} = 0. .y) 11 5 y u{(z12) 5 2) I 5 5 '1 {(llY)7O 5 y 5 2.95 and obviously has diameter < E. 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 . ' Thus X has property S. (ii) Let f : X + X be continuous.rr/z) E A } . s i n r / z ) 10 < z and 5 l} 6 = inf{z 1 (z7sin.
Y) I 1 IY I 1) c f ( X ) * Once again.11. If 6 > 0. (i) Show that T can be covered by 3 contractible open subsets. we first note that since X is path connected.96 I f 6 = 0. U2 and U3 are contractible open subsets and T = U ~ u U .{ u U b } .Y) n 10 < 2 < 6. 0 < 60 5 1. such that Therefore it is easy to prove that the set {(O. (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. f ( X ) is also path connected. UZ = X . (ii) Show that T cannot be covered by 2 contractible open subsets. Hence in this case there exists a 60. (See the Figure above. 2. we claim that f ( X ) = X . 3 3 < y 5 } = 0.2 Thus let U1 = X . 2 2 2121 Let T = S1x S1denote the torus. we see that f ( X ) = X. ( Toronto) Solution. . To prove it.I and U3 = X .) Then U1. n a Fig. U U 3 . then it is obvious that f(x) {(z. using the fact that f ( X ) is path connected.
97 (ii) Suppose that T could be covered by 2 contractible open subsets. Up. . Show further that the resulting topological space satisfies the Hausdorff separation axiom. Since p $! r p . b) For any point p $! vp. since U is locally finite.particularly. a) By definition.b) is any open interval of R and f E F } . . . Let U = { f . It contradicts the known fact that a. Then there exists a unique topology 7 on the set S of which U is the topology subbase. of p such that V . (Hama4 Solution.. Up. say. there is an open neighborhood V . if each point p E X has a neighborhood which intersects only a finite number of U.}aE~ an open cover of the space X. 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. It is easy to see that 7 is the weakest topology on S amongst all those for which all members of F are continuous. Suppose that s1 and s2 are . be a) Give the definition for “U is locally finite”. ( Toronto) Solution. Prove that there exists a weakest topology in S amongst all those for which all members of F are continuous. b ) ) 1 ( a . . U is said to be locally finite.. 2122 Let U = (U. . Then let V = W n V1 n . b) If U is locally finite show that. n V.. U PEK r p is closed.(T) x Z @Z. = 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. there is an open u PEK neighborhood W of p which intersects only a finite number of sets in U . Hence is closed. for each pi. for any subset K c J . .’ ( ( a . . It is clear that V is an open neighborhood of p such that V n ( up) = 0.
2.b z ) ) are two disjoint open sets of (S. bz) = 0.bl) and ( a z . By the assumption there exists a t least a n f E F such that f(s1) # ~ ( s z ) . Then n Ui = f'((al.bi) for i = 1 .for i = 1.l ( ( a z .bl) ( U Z . b z ) such that f ( S i ) E (ai. is . bl)) and Uz = f . 2 and (a1.98 two distinct points of S.7) such that si E U.Hence we may take two open intervals (a1. So (S.7) Hausdorff.
b) Compute the fundamental group of the figure 8 and draw a piece of its universal covering space. Let V be the image of the interior of the square under the identification. Solution. as shown in Fig. a> As is wellknown. I ) T is isomorphic to 7~1(U. we conclude that ?rl(T. modulo the smallest normal subgroup of ?rl(U.2.99 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. where q is the homomorphism induced L . by the Van Kampen theorem. Let y be the center point of the square.3 (b). Since V is simply connected. we can present the torus T as the space obtained by identifying the opposite sides of a square. TI)).21) z1) containing the image +*(rl(UnV.{y}. Under the identification the sides a and b each become circles which intersect in the point 20. and let U = T .
and.4.{ q } .3 (a). 2 1 ) ) is just the commutator subgroup of 7rl(U.{ p } .p1. where a and p are presented by circles a and b.~ f l ‘ .p2]. ~ where 6 is the equivalence class of a path d from 20 to 21.5 Under the covering map T. as shown in Fig. respectively. 21) containing & ( a l ( U n V. The following picture is a piece of its universal covering space. that q5+(y)= c y ’ p I ~ y ’ . respectively.ba. a2.2. and [ailpi] denotes the commutor aipiailpzrl. it Fig .2.4 Let X denote the figure 8 space. Let U = X . P are presented by circles a and b. respectively.20) ’ is a free abelian group on two generators a and fl. a 2 .2. where a . It is also clear that T ~ ( U .100 by the inclusion map q5 : U n V + U . Z is )a free group on two generators a’ = F1a6 and p’ = C1p6. The smallest normal subgroup of a l ( U . bl. By the Van Kampen theorem we can prove that a l ( X .the equivalence class of a closed path c which circles around the point y once. which are presented by circles a and b. (See Fig. It is easily seen that a u b is a deformation retract of U . z o ) is a free group on two generators a. we can see that the fundamental group of the oriented surface of genus 2 is a free group on four generators a1. Hence a l ( U ./31][a2. Changing to the base point 20. In a similar way. respectively. 1 ) is a free abelian group on 2 two generators a and p’.2. / 3 2 are presented by circles a l .zl) is an infinite cyclic group generated by 7 . it is easy to see that x l ( U n V. V = X .p. consequently.3 (b). Thus ?rl(T.) On the other hand.aa. we see that rl(T. p 1 .2.21). where a l .20) is a free group on two generators a and p.p2 with the single relation [a1.~ .) b) Fig . each level segment is mapped on the circle a . (See Fig.
we see that B = is a nonempty closed subset of A . Since A is obviously compact and Hausdorff. and suppose f : A + A is a continuous map. @ W for every n. and therefore. otherwise. closed subspace of a compact Hausdorff space X . For each positive integer n let fn(A) = ~ o ~ o .fn(A) ' (y'(s) = y(s) for s E S') is also nullhomotopic. (Indiana) Solution. where U and V are disjoint nonempty closed subsets of B. Noting that . there exist disjoint open subsets of A . consequently. which is a contradiction. A is also compact. F(S' x I) is also closed in . Noting that fn(A) is compact for every n and fn+'(A) c fn(A). Then B = U u V . and each vertical segment is mapped on the circle b according t o the direction indicated by the double arrows. and that A is compact. For. Show that there exists a positive integer n such that y(S1) c X . there exists at least a limit point 20 of the sequence { x C n } . E A . such that x.. o ~ ( A ) . i) Since X is compact and A is closed. A is a normal space. there would be a squence in A .101 according to the direction indicated by the arrow. n which contradicts the fact that fN(A) connected.B is a nullhomotopic map. We claim that there exists a positive integer N such that fN(A)c W .B be a homotopy between y and the constant map. Since A . which is an open subset of A. Thus f N ( A )c W . E f"(A) but 2. Therefore. 2202 Let A be a connected. It means that x. is ii) Let F : S1x I 3 X . Since F(S1x I) is compact and X .B is open. n=l ii) Suppose y : S1 X . n=l n fn(A) m PYA) = ( f N W n Wl)u (fN(A) W2>. Wl and Wz such that U c W1 and V c W2.. n times 00 f"(A) is connected. Thus it follows that B c W1 U Wz = W . closed and connected. . It is obvious that U and V are also closed subsets of A.W for every n.+'(A) c (i) Show that B = n f"(A) c A for any n. Suppose that B is not connected. is a compact subset. {xn}. Thus it is easy to see that f"(A) is compact.W is closed in A and. can see we that xo E B c W .f"(A) and such that the induced map y : S1+ X .
It is also wellknown that "  ai(T") M Z e . Prove that any continuous map RP" + T m is nullhomotopic. Therefore.102 X . say f. Let po be a fixed point of R". (Indiana) Solution. Therefore. there exists open sets U and V such that F(S1x I ) c U and B c V . it follows that there exists a positive N such that fN(A) V . such that pf = f . for any continuous map f : RP" + T".T(z) tpo. y(S') x .fN(A). Since X is normal.t). Then H is a homotopy between f and the constant map PO. It is wellknown that R is a universal covering space of T". } x to a point S1c S1 x S1 . .t) (1 . x S1 ( m factors). cX .fN(A). Let P : " R + T" be the universal covering map.  + fl is null 2204 Let X be the quotient space obtained by collapsing { p t . c F(S1 x I ) c u c Particularly. From the proof for i). 2203 Let RP" be the real projective nspace and T" be the rntorus S1 x . Thus there exists a lifting of f . the induced homomorphism f+ : a 1 ( R T ) + nl(T") is trivial. (rn times) and that a l ( R P n ) M Z2.The remainder of ii) is obvious.. Define H : RP" x I + R" by  H(z. So homotopic and consequently f is also nullhomotopic. .
by the Van Kampen theorem.2. where S is the equivalence class of a path d connecting 20 and z1. and U n V are path connected. (See Fig.l(X) =z. a F ig. Thus. zl)is an infinite cyclic group on the generator y. we may apply the MayerVietoris sequence to the pair ( U . Does there exist a continuous map f : T + S1 which is not homotopic to a constant? ( Toronto) . where q5* is the homomorphism induced by the inclusion map q5 : U n V + U . 2 1 ) modulo the smallest normal subgroup of R ~ ( U .{y}. Does there exist a continuous map f : S" + S1 which is not homotopic to a constant? (ii) Suppose n 2 2.V). It is easy t o see that u is a deformation ratract of U . al(U. ( X ) .2. V.zl). which is obtained by identifying the sides of a 2gon. zl)). 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. ( X ) . The conclusion is that H i ( X ) is an infinite cyclic group for i equal to 0 . Therefore.6) It is also clear that x l ( U n V. 1 . 7 r l ( X . and V be the interior of the 2gon. To compute H . and V is simply connected. 2205 (i) Suppose n 2 2. U and V are open subsets.6.2. z l ) is isomorphic to the quotient group of a l ( U .6 X may be identified with the space shown in Fig. U = X . Let y be the center of the 2gon.103 Compute x1(X) and H . Therefore we conclude that . (Columbia) Solution.(al(U n V. zl)is an infinite cyclic group on the generator SluS. It is easy to see that &(y) = lrl(u. where y is the equivalence class of a closed path c which goes once round the point y. 2 and is zero otherwise. U . Then. containing the 21) image q5.
eo) + “1(B.that is. F = p . : S” + R such that T? = f. Let p :E B be a fibration. we see that there is a lifting of f. By the same argument as in (i). bo) (Indiana) in which the homomorphisms are i# and p # . which means that f is not homotopic to a constant. b o )where f : I + B is a closed path a t bo . (i) Let cr = [f] E ~ l ( B .l ( b o ) (the “fiber”). ?must be homotopic to a constant map : S” + R. 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 .e.which is also a constant map from S” to S1. 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). e 2 x i t z ) = e27ritl E sl.104 Solution.rrl(B. prove that p# : ?rl(E. Hence f* is not trivial. respectively. (iii) Denote T = S1 x S1 by T = ( e 2 r r i t l .any continuous map f : RPn + S1 induces a trivial homomorphism f* : x 1 ( R r n ) a l ( S 1 ) . 20 5 t l . (ii) Prove that in general the 3term sequence + + + + Tl(F. and eo E F . Since al(S”)= 0 for n 2 2. has a lifting RP” + R such that a o f = f. Solution. Since R is contractible.eo) + Tl(E. which represents a. + . Let G : I x I B be a continuous map defined by f .. hence f is homotopic t o a o = C. Define it f : T + S 1 by f ( p i t l .bo) is surjeceo) tive. e 2 n ).and another generator to zero. Let i : F E denote the inclusion map. there does not exist any continuous map f : Sn + S1which is not homotopic to a constant map. (ii) Since a l ( R P n ) = Zz. bo E B be a base point. t z 5 1. consequently. i. = and f : S” + S1be a continuous map. (i) If F is path connected. 2206 A continuous map of topological spaces: p : E B is called a fibration if it has the homotopy lifting property . (i) Let a : R + S1be the universal covering map defined by ~ ( t )eZrrit. f must be homotopic to a constant.and. is exact.
105 G ( t .Irl(F. Let c1 = H I { o i x r .Then c1. there exists a continuous map H : I x I + E such that P H = G and H ( t . Then pc = f .) Therefore. On the other hand. G = = [c] = i#[c]. It is obvious that p h = GII. and h : I x (0) + E be the constant map such that h(t.  [fi Fig. i#(a)= [pf] = [bo]. cg = H11. h .G(t. Since F is pathconnected. 0) = bo. Therefore there exists a continuous map H : I x I + E such that H / I x { o }= h and p H = G . g2(0) = e l and gz(1) = e l . l which means that p# is surjective. Particularly we have p H ( t . By the homotopy lifting property. kerp# c imi#. ~ ( 1 = c i 1 ( 0 ) and c i 1 ( l ) = eo. ) ) Hence c = c1 x c2 x c i l is a closed path in F with base point eo.{o}. s ) = f ( s t ) for ( s . Hence the sequence mentioned above is exact. (See Fig.2. l I = b = [fl = a . suppose that G = [f] E kerp#.7. cg and c3 are paths in F and cl(0) = eo.. 1) = f ( t ) . Thus f = g2 * c * g1* gF1 is a closed path a t e l . It is clear that e l and e2 belong t o F. Noting that pgl. Let c ( 0 ) = e l and c(1) = e2. 0) = f(t). bg11 . It is easy to see that f is homotopic to C. we have P# rfi = bfi = bg21 . Then there exists a homotopy G : I x I + B between pTand the constant path bo such that G(t. s ) E bo for any s. It is obvious that p# . gl(1) = e l . ~ ( 1 = cz(O). Hence im. and G(0.7 2207 Is the canonical map q : S2 + R P 2 (which identifies antipodal points of S2)nullhomotopic? Why or why not? (Indiana) .{1} and c3 = H l { l } x ~ . b 1 . Let C : I + E be a path in E defined by c ( t ) = H ( t . pg2 and pgz' are all constant path at bo.O) eo.2. s ) G ( l . . where gF1 is the inverse path of g2. c kerp#. we may choose two paths g1 and 92 in F such that gl(0) = e2. 0) = ( p T ) ( t ) .eo). 1). t ) E I x I . (ii) Let a = [f] E . i. where bo is the constant path a t bo.e.
Since I  T" = ( p T ) and . This is a contradiction. it follows that identity map Id : (E.. We use the reduction to absurdity. T : ( E .eo) + ( E . It is wellknown that q is also the universal covering map from S 2 to R P 2 .eo) is obviously a lifting of T n p a t the base point eo. a constant map C : S 2 + S 2 .eo) such that p o ? = T o p. + ( E . we have d e g c = degZ = 0.= l T ( p ~ n .) Show that there is a map ? : E + E so that p o T = T o p and Tn= Id. we see that @ is a lifting of T" o p at the base point eo. ~ " . denoted by r. Let i : X t T 2be the inclusion map. But it is obvious that f is the identity map or the antipodal map from S 2 to S2. = T~ O )  P T " ( e 0 ) = T"'(T(eo)) = T*'(eo) = . Then roi : X + X is . 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 .and therefore q has a lifting : S2 + S2which is also homotopic to the lifting of c . = Id.. < 2208 Let p : E B be an universal cover with E and B pathconnected and locally pathconnected. Suppose that there is a retraction of T 2 to X . Thus we conclude that q is not nullhomotopic. eo) since T" = Id.l = . = T ( e 0 ) = eo.l ( b ) and consider the map T o p : E + B . S1.      2209 Let T 2 = S1x S1. On the other hand. there is a lifting of T o p . We have T o p ( e o ) = T ( b ) = b. Therefore. Choose eo E p . Suppose that q is nullhomotopic. by the uniqueness of lifting. o T .106 Solution. Prove (Indiana) Solution. .. Since ?rl(E)is trivial. Let T : B + B be a map so that T n = Id and so that T(b)= bfor some b E B.. we conclude that r?. On the other hand. f   (Indiana) Solution. Hence degg = f l . (Here T n = T o T o . Then q is homotopic to a constant map c : S 2 + R P 2 .. n times. .
Y. O .2.2. Therefore we conclude that H1(R3 . p l and p2 denote the points (1. p ) is not abelian. Let X = {(x. O . Take U = Y . z) I x2 z 2 = 1) be a circular cylindrical surface.Therefore we have = which is a contradiction. Hence ~. The Euler Characteristic K ( Y ) = 4 .l .0) and ( . + Fig. b) Prove that r l ( R 3. b) Let e be a point on the arc ab different from a and b. Hence the rank of H I ( Y ) is equal to 1 .A ) . a) Compute H 1 ( R 3 . Hence we have ab # ba and i*(a)i*(b) z*(b)i*(a).p2) is a deformations retract of R3 .107 the identity map. 0) respectively.8. It is easy to see that U has the .A .{pl. 1). Then we have Y = U U V .y.8 a) Y has a structure of a graph with 4 vertices and 6 edges. A = {(X.Z) I (Y2 + z2)(x2+ 2)= O}.{ e } . It is also clear that X .{ p l .oi.A . 2210 Let A c R3 be the union of the x and yaxis.A ) M H 1 ( Y ) z Y Z @YZ @Z.: r l ( X ) + r l ( X ) is the identity homomorphism. (Indiana) Solution.6 = 2. and let p = ( O .{ c } and V = Y . Then it is easy to see that X . p z ) is homotopically equivalent to the space Y as shown in Fig.0.(2) = 3. It is wellknown that r1(T2)is abelian and that .1(X) is an nonabelian free group on two generators denoted by a and b.
p ) is not abelian.y).A . and that there is a lifting of f . 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. : ( X .z1) + need not be distinct from  5 TO (?. that (R3 . + Hence 5 is the lifting o f f we want. p ) M ?rl(Y) is a free group on three generators and. Then let fl .y) be a regular covering space.9.g) + (Y.Since Since p is a regular covering space. Therefore we conclude that ?rl(R3 . Since U n V is contractible. (Indiana) . Show that there is a second lifting. there is a deck transformation y such that y(y) = g.) (Indiana) Solution. p*(.rrl(Y. Let fo(zl) = y.A .= r o f o : x Y.g.g) is a normal subgroup of 7rl(Y. ZO) + ( Y . 9. 2211 Let p : (?. consequently. (A 5. : (X. that is. Suppose that f : X + Y is a continuous function from the pathconnected space X to Y with f ( z 0 ) = f(z1) = y. by the Van Kampen theorem we see that ?rl(Y) is a free group generated by ?rl(V) and ?rl(V).108 same homotopy type as S1 and that V has the same homotopy type as the “eight figure space”.
109 Solution. where y is the path class of a path d from z o to z1.2. we may extend f ' .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. it is clear that wl(U. Fig. z o ) is an infinite cyclic group generated by u the closed path class of a. (Indiana) Solution. Then both U and V are path connected subsets of X and X = U U V .9 Take a point y in the open disk.2. Hence the least normal subgroup containing = 4*(rl(U V ) ) is isomorphic to 3 2 .9. Since the circle a is a deformation retract of U . Let denote the complex plane and f (2) be a polynomial of positive degree with complex coeficients.) Let U = X . Therefore r l ( U 121) is an infinite cyclic group on ' generator Z = y'a'y.) It is clear that r l ( U n V. 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 ) ) . by the Van Kampen theorem.where 4 : (U n V ) + rl(U)is n the homomorphism induced by the inclusion 4 : U n V + U . It is also clear that 4+(yc) 3Z. 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. We may consider f t o be a continuous nonconstant map f : a +a. Take a point z1 E U n V as the base point.9. It follows that rl(X) NNZG/U n =as. hence. Note that If(z)l + 00 as JzI + 00. (See Fig. (See Fig.{y} and let V be the open disk.2. Since V is simply connected.
we take U = X 1 V X 2 .{ p } and V = X 1 V X2 .10. (b) Compute the integral homology groups of X . by the Van Kampen.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. the unit disk D2 in the 2y plane.2. where 00 denotes the north pole. (a) Compute the fundamental group of X . (See Fig. call it A. Noting the fact that if a continuous map f : S2 + S2 is not surjective then the degree o f f is zero.2.k > 0. Furthermore. we may prove the Fundamental Theorem of Algebra by means of the reduction to absurdity. and the portion. then the degree of the extension f : S 2 + S 2 is equal t o k. (4 Fig. of the z axis lying within S 2 . 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 . Since U n V is contractible. It is obvious that . Then we may first prove that if f ( z ) = z k . (Indiana) Solution. + 2214 Let X denote the subspace of R3 that is the union of the unit sphere S2.110 to a map of the onepoint compactification of ( X f : s2+ s2 by setting f ( m ) = 00.) To compute ~ l ( X v X 2 ) .(4). Then 1 we have U u V = X1 V X z . 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.
f : X Y a continuous function.2. n ( Y . ? r l ( X )is a free product generated by two generators.) (b) Applying the MayerVietoris sequence to the pair ( V . f*([c'hc]) = [a]. we see that Noting that H i ( X 1 ) M Hi(X2) which are infinite cyclic for i equal to 0. there is a closed path h at 20 such that f*([h]) [Z?'].111 Therefore. We can take as generators the closed path classes which are determined by the closed paths omp and omq respectively. Thus f o (c'hc) and a are homotopic.2 1 Let X be a pathconnected space. consequently. there exists a path C : [0. and.2 and are zero otherwise. (Indiana) Solution. .10.V). Show that is also surjective. and E X . f ( z l ) )where a is a closed path at f ( ~ l ) . It means that f o h and Zu?' = are homotopic. (See Fig. cac is a closed path at f ( z 0 ) .f ( 2 1 ) ) . For any [a] E ? r I ( Y . Hence  f* : m ( X . otherwise.1] + X such that c(0) = 20 and c(1) = 2 1 . 20. By the assumption. Suppose that the induced homomorphism f is surjective. we conclude that i = 0.1. Since X is pathconnected. Then Z= f o c is a path connecting f(z0)and f(z1) I in Y . 2 1 ) is surjective.
{ e o ) is homeomorphic to Y .) Since the automorphism group A ( X . that.11. it is not difficult to see. Since X . where both X and Y are connected. let p .11 are not homeomorphic.2.112 2216 Let B denote the “figure eight space”. Otherwise. Prove that X and Y are homotopy equivalent. But the spaces X and Y shown in Fig. p ) F Z & and X is connected.{ f ( e O ) } is obviously not contractible.l ( z 0 ) = {eo.11 For any 2fold covering map p : X + B . . X has only two different types as shown in Fig. By the formula . 2217 Calculate the fundamental group of the space R P 2 x S2.11. we come to a contradiction. (Indiana) Solution. (See Fig.2.2. suppose that f : X + Y is a homeomorphism. in substance. Then X . a m b Fig. (Indiana) Solution.el}. r ( x Y )= m ( X > m ( Y > .2. by considering the liftings of the circles a and b in X . Then it is easy to see that they are homotopy equivalent to an 3leaved rose G3. rlX @ we see that 7rl(RP1x S2) x T1(RP2)@ “1(S2) a @ {O} M 2 =a. Let p : X + B and q : Y + B be 2fold covering maps. but not necessarily homeomorphic..{ e o } is contractible and Y .{f(eo)}.
h : ~ ' ( 2 . we see that the path component of p .e2.113 2218 TI(.' ( Y ) contains exactly the points {eo.e4} of p'(*) which corresponds to the elements {0. what is the cardinality of the sat A n B?) (b) If G is a finite group.' ( X ) contains exactly the points {eO.+ G a surjection.e4.' ( * ) = {eO. * is generated by 2 and y. ker h is isomor*)/ phic to the group Z/6Z= Zg. where * denotes the vertex Fig. .l ( * ) .' ( Y ) divides the order of G . Then. Y .!?.' ( Y ) = p . Thus ~ l ( 2 . We denote p .' ( X ) with a path component B of p . prove that the number of intersection points of a path component A of p . we conclude that A n B = {eo}. ( p * ( ~ 1 ( 2 .' ( * ) by p ..4}of& respectively.2.e3. Since p'(x) n p . (a) Since h ( z ) = 2 and h(y) = 3 and T ~ ( Z . *) Let 2 denote the figure 8 space. X and Y circles.y E be the elements in 2 defined by X . By the same reason.12 (a) Let h : ~ ' ( 2 .2.+ Z/62be the homomorphism satisfying h ( z ) = 2 and *) h(y) = 3.el.ez.) If A is a path component of p . and let p : 2 + 2 denote the covering space corresponding to the kernel of h. how many intersection points of A and B are there? (i. we see that the path component of p .' ( X ) and *)) B is a path component of p .' ( Y ) . it ) follows that the homomorphism h is surjective.e5}. from h ( z ) = 2. = ker(h). Z = X V Y . e3} of p'(*). Hence the covering space p : 2 + 2 is a 6fold cover.e.   (Indiana) Solution. Let x. and p : 2 + 2 the *) corresponding cover.
) Applying the Van Kampen theorem. we see that the number of intersection points of A and B is equal to the number of elements in H1 n Ha. U and V are open subsets.{y}. Under the identification the edges a become a circle through the point 20. and U n V are arcwise connected. Let V be the image of the interior of the hexagon under the identification.13. (b) Compute.zl) + r1(U. (4 a d @ 2 0 a Fig. Then r generates a subgroup H1 of G and s generates a subgroup of H2 of G.13. V . Let 2 1 be a point in U n V .rrl(X.21) is an epimorphism and its kernel is the smallest normal subgroup containing the image of the homomorphism 41: rl(u n v. the homotopy class of a closed path c which circles around the point y once (see Fig.2. the corresponding cover p : 5 + 2 is a finite fold cover. the fundamental group of X . and V is simply connected.13 Represent X as the space obtained by identifying the edges of a hexagon. U . (a) Compute. . it divides the order of G. as shown in Fig.2. (Indiana) Solution. Then.Irl(U. 21) is a n infinite cyclic group and generated by y. with proof. Let y be the center point of the hexagon.114 (b) Since h is a surjection and G is a finite group. we conclude that $1 : .2. It is clear that U n V has the same homotopy type with S1. By Lagrange’s theorem. In a similar way as in (a). 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. 2 1 ) + . with proof. that wl(U nV. Suppose that h ( z ) = r and h(y) = s. the homology of the universal cover of X . zl). and let U = X .
S1] of homotopy classes of maps from X to S1 inherits a natural group structure. (a) Define this group operation explicitly and indicate the group identity and how inverses are formed. Here ( f a  . then it is known that the set [ 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). But it is easy to see that X ( X ) = 1. i = 0. So we conclude that . S1. 1 where It is obvious that the circle a is a deformation retract of U. H I ( % ) = 0 and H 2 ( 2 ) x H 2 ( X . 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. It is wellknown that the Euler characteristic X ( 2 ) = 6 X ( X ) . g ] . Then the multiplication of [X. hence X ( 2 ) = 6.lrl(U n V . From H o ( 2 ) M I . is defined by [f] [g] = [f . where 2 is the 1skeleton of 2. 7 r l ( U . and T 2= S1x S1. = I Z. S (Indiana) Solution. It is obvious that dl(r) = ar6. z l ) is an infinite cyclic group generated by a' = ?'a?. 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. Hi(%) i 2 3. 2 . ( 0. Hence the smallest normal subgroup containing 41(. (b) Compute this group explicitly for X = point. You do not need to prove your assertions. Thus ?rl(U. zo) is an infinite cyclic group generated by a and.115 and 4 are homomorphism induced by inclusion maps. 2220 i = 1. X ' ) . see ' we that H 2 ( 2 ) is a free Abelian group of rank 5. where the S1] map f .21)) is isomorphic t o U . where 7 is the homotopy class of a path d from zo to 21. consequently. g : X + S1is defined by g)(z) = f ( z ) g(z) for any z E X .
For the case of X = S1.9)2) (deg(f1 * Sl). R is contractible. S l ] x “l(S1)xa. 4 is a homomorphism from [X. We conclude that 3 [X.116 f ( z ) .S 1 ] to Z @Z. Note that f l can be extended to a map from X to S1. The inverse of [ f ] is the homotopy class of a map f. 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 . denoted by f1.S1] is trivial. where the map e : X + S1is defined by e ( z ) E eZni for any 3 E X. and cosequently q is an isomorphism. . 2. M r2(S1)= 0. 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. Thus one can easily prove that 4 is a monomorphism. The identity of this multiplication is [el. then [ X . g ( z ) is defined by the multiplication of complex numbers.sl) = (deg(f * 911. S1defined by t ) = f ( e 2 n i f 3 1e2xit+2xi(lt)+ 1. (b) If X = point. can easily prove that one [S’.degfz).S1] obviously has only one element. For the case of X = S2.S1]+Z@Z is defined by 4([f]) = (degf1. p o )and fZ(6) = f ( p O l e z U i e )for any e2nie E S1. deg(f .92)) = (deg fl + degg1. respectively. deg(f2 . Then by the facts that S2 is simply connected and the universal covering space of S1. Now we discuss the case of X = T 2 = S1x S1. It is clear that 4 is a epimorphism. We have 4“fI . [sl) = = = 4Uf. by still which is homotopic to f under the homotopy map F : X x I + ~ ( ~ 2 r r ie0 n i G .S1] %Z@Z. Then let 4 : [X. which is : defined by f(3:) = l / f ( z ) for any 3: E X . one can easily prove that [S2. So the group [ X . deg f 2 + deg 92) + (deg 91 deg 92) 1 = (deg fll deg f 2 ) 4(V1) + N g 1 )  Therefore.
Since R is simply connected. i. Suppose that ?rl(X) is finite.Then exp(2) is the homotopy between f and c . and f*(?rlT1(X)) for any map f : X + S1. and the same conclusion holds.e. i... locally path connected and the fundamental group of X is finite? (Indiana) Solution. the universal covering space of S1.  Prove that if X is locally path connected and simply connected then every map X 4 S1is homotopic to a constant. Thus it is ) easy to see that x is not a local homeomorphism at Z.~ ( z o must have at least a limit point. Let ?r : 2 + X be the universal cover of X.e. Since X is locally path connected and simply connected.117 2221 If X is a pathconnected space whose universal cover is compact. say Z. X + R such that exp(7) = f . What can you say if we just assume that X is path connected. Since ?rl(S1)% Z and f * ( ~ l ( X ) is a finite ) subgroup of ?rl(S1). ?: 7 . show that ?rl(X. Since 2 is compact. we see that f * ( ~ l ( X )must be trivial. Denote the homotopy between and E by g.20) is finite. then ~ ~ ~ ( would ) a closed set of infinite points in X. 7r1(X) = 0. =0 Hence there exists a lifting of f . f is homotopic to a constant map Z. Let exp : R + 6 denote the exponential covering map. 2 0 be T . 20) were not finite. f is homotopic to a constant map. If n l ( X . So the above ) argument still works in this case.such that n(Z) = 20. (Indiana) Solution. which is a contradiction.
. we see & ( a ) = b. Since P2(62(a) . Let c E C3 and 7 2 ( ~ = 0. . Since b1 is surjective.&(b) E kercl..(b) = 6162(u) = 0.Pz(b) = E z ( a 2 ) . i. Therefore c = S. (Hamud) Solution. Thus there is = P262(a) .b ) a2 E ker 01 = imcyz. Hence there is a n a2 E A2 such that ~ ~ ( a= Pz(b). To prove the exactness at C.Pz(b) = 0 and f ~ 2 is injective.e.h ( b ) = w 2 ( a ) . then the last row is also short exact. By ) 3 the commutativity we see . 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. Hence 7 2 is injective. Then since 2 ) c2(al(a2)) = Pl(EZ(a2)) = Pl(PZ(b)) = 0 and C 2 is injective. an a E A3 such that a z ( a ) = a2.118 SECTION 3 HOMOLOGY THEORY 2301 Prove the 3 x 3 Lemma. there is a b E B such that c = b1(b). we show that 7 2 is injective. we have a l ( a 2 ) = 0.
Therefore. 2 We leave the proof of the exactness a t C1 to the reader. Let Xo = T2x T2. then X ( M ) = 0.E 1 & 2 ( a 2 ) = c. .2. since i and 2m 1 .i have different parity. ( Columbia) Solution.119 Now we prove that ker y1 c imy2.1. 2302 Prove that if M is a compact manifold of odd dimension. Then 72(61(b3))= E l h ( b 3 ) = &l(b). For any c E ker 7 1 . Thus the exactness a t C is proved.1. since ~1 is surjective. we have b . Thus.&). 3 and 4. In a similar way we may prove that imy2 c ker 71. . and X4 = S2x S2. Hence c E imy2. Since a1 is surjective. 2m+ 1 X ( M )= (l)i dimHi(M. Show examples of compact 4manifolds with X = 0. Since M is compact. there is an b E B2 such that c = ~ l ( b ) Thus it is easy to see that Pl(b) E ker C1.2. and consequently that there is an a1 E A1 such that &(u1) = Pl(b).) = i for i equal to 0.4. for any i.3. M is 272orientable. there is an a2 E A2 such that al(u2) = ul. Denote by u the connected sum of h h projective planes.X 1 = R P 2 x R P 2 . Thus by the commutativity. Therefore there is a b3 E B3 such ) that p 2 ( b 3 ) = b . i=O By the PoincarC duality theorem. they appear in the sum with opposite signs. X3 = U3 x U s .~ 2 ( u 2 ) . Then it is wellknown that X ( U h ) = 2h. Therefore X ( M ) = 0. Let X2 = U3 x 174. Suppose that d i m M = 2m + 1.Using the fact that + we see that X(X.~ 2 ( u 2 E kerpl.
s+t. The suspension E X of X is defined t o be the identification space obtained from X x [1. O ) =z3 (T. : C. So Zn(c) = { ( T . ( c ) is isomorphic to Z $Z.l defined by (s . ( c ) = Z for even n. (Illinois) Solution. Set U = E X . Then U and V are open sets of E X . and E X = U UV. s .~) + n even (O.1] by identifying X x (1) to a point and X x { 1) to another point. en}.(C) for all n.s+t) nodd. When n is even. Let pl and pa be the identification point respectively. ~ E Zn(c) if and only s = t . Compute the homology of E X in terms of the homology of X .t.s. Z .s. we have H . Compute H.e. For example the sphere S" is the suspension of S"' with the north and south poles corresponding to the two identification points.t . we have where H q . (Illinois) Solution.~) + (T. O . It is obviously to see that U and V are contractible spaces and X is a deformation retractor of U n V .t)E C . s ) E ) i. By MayerVietoris sequence. = t {C"In>O} C" C..{pl} and V = C X .120 2303 Let X be a topological space. Noting that imdn+l = {(O. ( T . s .l ( X ) denotes the reduced homology of X  2304 Consider the chain complex c d. } for even n. .{ p z } .
S1) Hl(S1) + 0 The square is commutative and the level rows are exact. Therefore. X') T fi* 0 + T fils'+ a : H2(D2. 0 . 0 )E C.121 When n is odd. ( c ) M Z for any n. . Therefore. M z CBZ. H . ~ )E & ( c ) if and only if s t = 0. + 2305 Construct a CW complex which has the following 2homology groups: ( Coz~mbia) Solution. ( T . It is wellknown that H1(X1) = Z$Z. So Zn(c) is isomorphic to Z $Z.HI(X2. one by the map f i : S' + 20 and the other by the map f2 : S' + a such that f2(z) = z2. which has two generators a and b. s . the 2sphere.+l = { ( ~ . Let X2 be the space obtained by attaching two 2cells to X I .} M Z.and noting imd. ~. Denote by X1 the "figure eight space" as shown below. Since H ~ ( xx') = imfl.we have Hn(c)= z . consider the following diagram 0 t HZ(X2) 5 H2(X2.X1) 3 + H'(X1) % H1(X2) . CB imf2. the image of one 2cell under f1 is homeomorphic to S 2 .
Then H p ( S o ( q ) )is a free abelian group of rank q for p equal to 0 and is zero otherwise. Now let X be the space obtained by attaching two 3cells to X2. 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 prove that gp(Sn(q)) is a free abelian group of rank q for p equal t o n and is zero otherwise. M 22 and kera. Thus. = 0 and imf2Isi.X1) = 0.. Denote by S"(q) the space S" V S" V .. q times q + 1 points. we may see that ima. by induction on q and the MayerVietoris sequence of the pair ( U .. V ) . when n > 0. by the map z one and the other by the map z + z6. V S"). ( Columbia) Solution. n 2 0. What is the homology of this space? + z4 (Indiana) .1).{ u } has the homotopy type of Sn(q) and V = Sn(q 1) . + + + Let a E S n ( q ) and b E S". for all p .(S" V S" V .122 imfllsi. It is easy to see that U = S n ( q 1) . + 2306 Compute H.{ b } also has the homotopy type of S"(q).. So(q) has J . Since H1(X2. where fip is the reduced homology group. we may conclude that the space X satisfies the requirements in the problem. V S" When n = 0. It is also easy to see that H z ( X 2 ) =z. Then in a similar way as above. In fact S"(q + 1) = S"(q) v S". = 2!J. z Y . one by the map g1 : S 2 20 and the other by a map g2 : S2 + S 2 such that degg2 = 3. we have H1(X2)M Z $ Z 2 . 2307 Build a CW complex X by adding two 2cells t o S1.
KnA1)d. i = 2 . lskeleton K1 = S1 and Oskeleton KO = { p } . where C n ( K )= Hn(Kn.In the same way we see that the image of (glsi). consider the following diagram: HZ(K1) t HZ(K2) 2 H2(K2.Knl) and j. is isomorphic to the subgroup 22 of H I ( @ ) . : and C n ( K )+ Cnl(K) is defined to be the composition of homomorphisms. Noting f l s i is the map z + z4.we have H 1 ( K ) M 2/22 M &. is isomorphic to Z @&.KO) M 2. It is obvious that H n ( X ) = H. imd2 M Z. K1) 2 HI(K1) 2 H I ( K 1 . I 01 otherwise. Thus we conclude that ( zg3z2. it follows that imd. K 2 is obtained from K 1 by attaching two 2cells via the maps f and g as indicated in Fig.2. M Z @&. we see that the image of ( f l S l ) * is 4Z. It is wellknown that H .(K) = 0 for n >_ 3.14 X is a %dimensional CW complex with %skeleton K 2 = X . is 6Z. we see that H 2 ( K 2 ) imj2 M M ker 8 .2. i = 1. Let K" = K 2 for n 2 3. since j 2 is injective.l is the homomorphism induced by the inclusion map. Therefore.dn}. KO) T f* T (f IS^)* H2(E2. It is clear that j is an isomorphism. is a n isomorphism. where d. KM ) ~ imf. K 1 is obtained from KO by attaching one lcell. Fig.S1) 3 H l ( S 1 ) The square is commutative and it is well known that f. and that ker 3.14. is a monomorphism and 8.123 Solution.. Thus. Noting that 1 ker d l % H 1 ( K 1 . To compute H2(h). Hi(X) = i = 0. Since H ~ ( K ' . Then we have a chain complex K = {Cn(K). ( X ) M H n ( K ) . . CB img. is the boundary operator of the pair (Kn.
y.(f(0))). let Wl = .(0)) g * ( S ) . it follows that f(0) = 0.0)).(0)) M H. From the above results. . if z is in a coordinate axis.(0).((O. b) Prove that any homeomorphism h : X Solution. 1)) = s and W2 = s .124 2308 Let X = {(z.{ 0)). Thus H . a) Let U = D3 n X . Let S = S 2 n X .((11 0. Then S is a deformation retract of U . To compute Z*(S).z) E R3 I zyz = 0).{XI) = Z @Z @Z. a) Compute H . (O.O.we see that b) Let z be any point of X different from the origin.(0)) = H.0). z ) E R3 I z2 + y2 + z 2 < 2). Wz). u .X . x .y.(0)) M H2(X.0. where +X must leave the origin fixed.(U.(1. ( X .I(U. Since the local homology groups are invariant under homeomorphism.(0)) for any q. It is easy to see that U is contractible. otherwise.(U . x . where S 2 is the unit 2sphere. Let h : X + X be any homeomorphism.X . (Indiana) D3 = {(z. X . I). Applying the MayerVietoris sequence of the pair (W1.O. H 2 ( X . Similarly we obtain Hz(X. So H. ( X .
RZ which is split exact.{ p } ) x Hq(S1). (S1 = 1 .+ H q ( D.{ p } have the same homotopy type with S1. A  b) Since S1 x R1 and D . we see that H.l ( D .sphere) b) Use a) to calculate the homology of S1x R1 . the nontrivial part of the MayerVietoris sequence is 0 + g1(D. 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 . 2310 Let the real projective plane R P 2 be embedded in the standard way in the real projective 5space R P 5 . Thus we have %I iYi(S1 x R1 .{ p } . a) The MayerVietoris sequence is .125 2309 a) Write down the MayerVietoris sequence in reduced homology which relates the spaces S1x R1.{ p } ) f gl(S1 x R1)+ 0 .{ p } ) + *. Since D is contractible. . ( R P 5 / R P 2 )x H q ( R P 5R P 2 ) .{ p } ) 3 Hq(S1x R1 .{ p } ) @ @. (Indiana) Solution. Compute Hq ( R P 5 / R P 2 ) .{ p } ) 3 x R1 .. and a disk D in S1 x R1 about the point p.(D . for any q.{ p } ) = otherwise. S1x R1 ..{ p } .(D) f gq(S1X R1) + H g . (Indiana) Solution.. it follows that Hg(S1x R1)x H. where R P 5 / R P 2 is the space obtained from R P 5 by identifying R P 2 to a point.
= 1.it follows that R H.O) E Y x R. from the exact sequence of the pair ( R P 5 .Y x R .(RP5)= { z.5. otherwise.(RP2) = otherwise.2. 0.{0}). (Indiana) Prove that Y x R is not homeomorphic to R 2 .3. U  (0)) M Hil(U . &. we take an open neighborhood of the point 0. Let Y x R = { ( 2 1 .5.{ 0 } ) . (RP5/RP2) = a. and H. otherwise. 2 2 . { z. Therefore Hi(Y x R . P 2 ) . 2 22 0 ) . 0. 0. To compute the local homology groups Hi(Y x R.*. q = 0. 0. 2 3 ) E R 3 1 2 1 2 2 = 0 .1. &.  because U is contractible. 4 = 0. 2311 Let Y = {(z1. where D3 is an open ball centered a t 0. = 0. Solution.z2) E R2 I 2122 = 0 and 22 2 0). Then the point 0 = (0. U = D3 n (Y x R ) . Y x R .126 It is wellknown that H.(0)) M Hi(U. Let X be the space as shown in the following figure . i = 0.*. q = 1. Q q Thus.
Then yo is a strong deformation retract of ?r(U). X U C A is obtained from the subset X x (0) U A x [0. Let V = A x [i.yo). Thus H z ( Y x R.B ) = H * ( Y . Y O ) = H*(Y/B. y o ) + + + H.1] c Y .?r(U)) H.yo) = 0. *. Hence.V ) . for every n.YO) * * * it follows that H. Since B is a strong deformation retract of U . It is wellknown that for any point p E R2 the local homology group H 2 ( R 2 . A ) .v.127 It is easy to see that X is a deformation retract of U .(0) and that H I ( X ) X Z @Z.(Y/B. 1 of 1 X x [0.YO)+ H * ( Y / B .Y x R . By the excision property. H * ( Y . Let U = A x [$.?r(U). the inclusion map of pairs induces an isomorphism H * ( Y / B .(?r(U).l(?r(U). Let ?r : Y + Y / B be the identification map and yo denote the point ?r(B)in Y / B . 1 by identifying A x (1) to a point.(0)) EZCBZ.U ) . 2312 Let A be a nonempty subset of X and X u C A be the union of X with the cone of A.u .?r(U)). ( X U C A ) is isomorphic to H. Let Y = X x (0) U A x [0. that is.1]and B = A x (1). Then X U C A = Y / B . we have an isomorphism 11. U ) M H * ( Y . 1 Prove that the reduced homology group g . H * ( Y . it follows that Y x R is not homeomorphic to R2.(X. (Indiana) Solution.( p } ) GZ.* + Hn(?r(U). Since the local homology groups are isomorphic under homeomorphism.R2 .it follows from the exact sequence that . Thus. in the exact sequence of the triple ( Y / B .
  (ii) Using only the EilenbergSteenrod axioms (Homotopy. it follows that H*(Y/B. ( X .1]. we have X x (0) H. we have H * ( Y . B ) M H . where denotes the equivalence relation generated by requiring that (2.(V)).T ( V ) T ( V ).yo)H * ( Y / B )= G . s) if s = t = 0 or s = t = 1. All of these combine to give an isomorphism H * ( Y / B . On the other hand.n(V)) to give an isomorphism Now the restriction of the map n.128 Thus. let Cf : C X + C Y be the map of suspensions induced by the map ( ~ .1]/ . Dimension) for a homology theory prove that Hi (EX) naturally isois morphic to Eil(X).. ( X UCA).) t I f f : X + Y is a map. (EX is the quotient space (X x [0. . A ) . H.gives a homeomorphism of pairs n. yo) Since Y admits obviously a strong deformation retraction onto which maps B onto A x { 0 } . Exactness.(Y.(X U C A ) is isomorphic to H .t) X x [0.) (Indiana) .yo).u V). M H * ( Y . For any topological space X let EX denote its (unreduced) suspension.v.u  V )+ ( Y / B . t ) (f(z). (Here Hi denotes reduced singular homology. x Therefore. Excision.B ) .: (Y  v. In a similar way the set T ( V )may be excised from the pair ( Y / B . A ) for every 2313 n. + of (i) Prove that if f : X + Y is a homotopy equivalence then so is Cf. B )M H * ( Y . and so an isomorphism of their homology groups. ( X . ) (y. from the exact sequence of the pair ( Y / B .
1).'(Z) means any point (z.s) E EX x I .1) = g(z) for any z E X . if Z = po or p l . It follows from = the homology sequence of the pair (EX.129 Solution. Let c U .O) or (z. By the definition we see that Cf = rIy 0 (f x id) 0 IIXl.O) = f(z) and G ( z .W). 3 It is obvious that U and V are both open sets of EX and that U and V are both contractible. Let f : X + Y be a continuous map. respectively. where. 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.W has the same homotopy type of X . Then v It is easy to see that U . It is also easy to see that H ( Z .W ) we see that Hi(V. (i) Denote by rIx : X x I + C X the quotient map. lI. s) = E(II&'(%).t) 1 t < }.z E X . where id : I + I is the identity map. U .(V) 0 for all i. l). U ) . and by po. By the excision property we have W = {IIx(z.W ) M E&'(U . Thus we conclude that k i ( C X ) k*l(X). (ii) Let 1 u = {rIx(z. 0) = Cf(Z) and H ( E . u . .t)1 t > } 3 and 2 = {rIx(z. from the homolopy sequence of the pair (V. Therefore Cf is homotopic to Cg. for any (2. U ) that fii(CX) M Hi(EX. and consequently it follows that if f : X + Y is a homotopy equivalence then so is Cf. Hence r?. 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 . On the other hand.t ) 1 t 2 2/3}. It is easy to check that H is welldefined and continuous. pl the equivalence classes of (z. 1) = Eg(Z).i(U) = g.
e.a2 = 0. Suppose that y and Po are surjective. and the inclusion U n V + V induces a surjective homomorphism H1(UnV) + H 1 ( V ) . It means that kerag = 0.130 2314 Consider the following commutative diagram of abelian groups in which the row and column are exact sequences. (Here H denotes singular homology.a 2 . there is an a1 E Al such that y ( a 1 ) = /32(a2).. which means that a l ( a 1 ) . Let a3 E A3 such that cyg(a3) = 0.) a) Prove that H 1 ( X ) = 0. 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. H 1 ( U ) = 0.a2 E kerP2. there is a bo E Bo such that Po(b0) = b l . Then from the exactness there is an a2 E A2 such that a z ( u 2 ) = a3. Hence by the commutativity we have a l ( a 1 ) . Since y is surjective. (Indiana) Solution. and there is a bl E B1 such that P l ( b 1 ) = a(a1) . Since PO is surjective. Once again by the exactness we have 0 = PlPo(b0) = P l ( b l ) . a3 is ) injective. Prove that a3 is injective. . i. Hence a3 = ~ 2 ( a 2 = a 2 a l ( a l ) = 0.
So the homomorphism H z ( U n V ) + H z ( V ) induced by the inclusion map is surjective..1)sphere. We use the reduction to absurdity. Suppose that there is a point 20 E D n such that zo @ f(D"). H 1 ( X ) = 0.9 .e. By the hypothesis. a) Consider the MayerVietoris sequence of the pair ( U . z ) E s2I z < }. V and U n V are pathconnected and that H z ( U ) = H2(V)= 0.5 ) and 1 1 v = {(z. By the assumption. i... (Indiana) Solution. and denote the standard (n. 1 (Indiana) Solution. 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 . .1 . Prove that f is surjective. ) n ) 5 ) n Since U n V is pathconnected. we see that 20 E D" . It means H 1 ( X ) = im$*. b) Let X = S 2 . d* is surjective so that Therefore im& = 0.131 b) Give an example t o show that the analogous assertion becomes false in general if H is replaced everywhere by Hz. 2316 Let D" = { z E R" 1121 5 1) denote the standard unit ball in Euclidean space Rn.V ) A . Eo(U n V )= 0.z) E s2I z > . It is obvious that U . + H ~ ( x 3 H ~ ( U v)% H ~ ( u CB H ~ ( v )H ~ ( x + H ~ ( U v).l . y. 2 Then X = U U V . But H 2 ( X ) M Z. the unit 2sphere7 u = {(zclg.
: H2(X2. Therefore = H 2 ( X ) M ker{d. Assume H1(X) x Z @Z/3. X 1 )is free abelian. (Indiana) Solution. Therefore g o i : Sn' + Sn' is a homeomorphism.( 2 0 ) + Sn' be a retraction. which is a homeomorphism. Thus because H 2 ( X 3 . and no higherdimensional cells.1 : 9 . .e. H2(X) Z Z $Z. and H 2 ( X 2 .3 3 = 2. i.Compute the Euler characteristic of X and determine (with proof) all possibilities for H2(X).rankHl(X) + rankH.132 Therefore Sn' is a deformation retract of D" . Therefore rankHz(X) = 2. where d.X O ) } . three 2cells. H 2 ( X ) is also a free abelian group of rank 2. Hence (g 0 i)* : Hnl(S"') + H. Let X k denote the kskeleton of X .'(S"') is an isomorphism. Let i : S"l + Dn be the inclusion map. X 2 ) 0. three 1cells. we see that H 3 ( X 3 .(20).l ( D n ) is t obviously a trivial homomorphism because of Hnl(Dn) = 0. X 1 )= Z $ Z @ Z .XO).X ' ) t H 1 ( X 1 . This is a contradiction.X ' ) 3 H 1 ( X 1 ) 2. Then g = r o f is a continuous map from Dn to S"l and gIs. X ( X ) = 2 .. H1(X1. On the other hand. But ( g o i)* = g* o i. : Hn1(Sn') H n . and i. is the composition of homomorphisms H 2 ( X 2 . + X ( X ) = rankHo(X) . By the assumption. and consequently (g o i)* is trivial. Let T : D" . The Euler characteristic of X . 2317 Let X be a connected CW complex with two 0cells.(X).l + S"' is just the restriction of f to Sn'.
Z) is a nontrivial finite group then Hk+l(X. We claim that &+'(a) Otherwise we would have imd'+l = 0.133 2318 Let X be a CW complex with exactly one k+lcell.x k . Prove that if Hk(X. therefore.l ) is a free abelian group. consequently.Z) = 0.Z) M ker dk+1/imdk+2. Denote by Xk the Icskeleton of the C W complex X . X k . By the hypothesis. 2319 Let and . X k ) zz.1 ) It is wellknown that 3 Hkl(Xkl) 5 H k T l ( X k .X"'). and Hk+l(X. Z )M ker dk c H k ( X k .Z) = 0. We choose a generator a in H k + l ( X k + l . Therefore # 0. Hk+l(X. and by dk the composition of homomorphisms of pairs H k ( X k .l . H ' + l ( X k + l . It follows that kerd"' = 0 and. It means that H k ( X . i Y k ( X . X k ) . (Indiana) Solution. and. kerdk is trivial or free abelian. Z )is not a nontrivial finite group and contradicts the hypothesis. H k ( X . But it is wellknown that H k ( X k . Thus we have d'+'(a) # 0. Z ) M ker dk/imd"' xk2).
x  (O. It is also clear that S1 is an deformation retract of W n Wz. O ) in R4. U . where and D = { ( ~ 1 . O . O .O)) for all i.p4}.O)) = Hi(U . (Indiana) Solution.(0.15) + 2: + 2: = 1). Let 6 be the unit open ball centered at the point ( O . 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.O. applying the MayerVietoris 1 sequence to the pair (W1. we see that Hi+l(U.O)) . 2: (See Fig. W2). Therefore.15.15 Let Wi = C U D .(O. Then U = 6n X is an open neighborhood of ( O . It is obvious that U is contractible and that U . X . other wise. u . 2 3 . 0. 2 2 . O . 0)). 0.O) has a deformation retract C U D. Fig. 2. 0. we see that 2 = 0.{pl.(0.p2) and W2 = C U D . 0. 0.{p3. By the excision property we have &(X.(O.134 Let X = A u B.O)) M H i ( U . O . O .(0.) Then W1 and W2 are open subsets of C U D and W u W2 = C u D. O . 0.O.2. 0.( O . 0) in X . Compute the relative homology groups H i ( X . 2 4 )E R4 I 2 4 = 0.O.2.u . (See Fig. 0. From the homology sequence of the pair ( U .O)) for all a. O .
f(z)) be trivial. Hence we have  7 f* = R.X .z) is finite. Let Y = S 1 x S1x S2. Since ?rl(X. we have It is clear that Rl(S1 x s1x S1) = “l(T2)@ “ l ( S 2 ) = a $a.1 otherwise.x) is finite. Z ) + H i ( Y . But it is obvious that : H i ( X . must . I ) for all i different from 0 and 2.Z) is trivial for i different from 0 and 2. 0)) = i=o.Z) Hi(R2 x S2. It is easy t o see that R2 x S 2 is the universal covering space of Y .(O. 2320 Let X be a path connected and locally path connected space and let x E X . so is . Z ) +  + Hi(Y.O.135 for all i 2 0. then any continuous map f : X + Y induces the trivial map f* : H i ( X . Show that if al(X. By the Kunneth theorem. where R : R2 x S 2 + Y is the covering map. Therefore. (Indiana) Solution. Hence we conclude that Hj(X. 0 f* : H l ( X . the homomorphism f* : R ~ ( Xz) + R ~ ( Y .O. we may lift f t o a map : X + R2 x S 2 such that a o f = f . Z ) for any i.
a’a) 6’6P = . B P1 and C + D 17 B S E 71 1 E D .Therefore. then so is A C ’a l .+ E 1 E PI 6’6 C .6’ya = 0.&)(P.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.&a’a= 6’ya . we have 6P = 7 a and &a’= 6’7. .t F Prove exactness of (Indiana) Solution. we have (6’6.F 6’ are exact. By the assumptions.
( c .e. a’a).. i.+ X be a continuous map.. a’a).a’). The exactness at C @ E is proved. Thus.E ) . By the assumptions.E ) . It is clear that { F n .e. consequently. Then there would exist an 2 0 E A such that z 5 zo for any z E A.y) = ker(c . a’a)(u)E im(0.6) and im(a’. 20 1 = 521. Let ( c . Give an example to show that compactness is essential for this assertion.e ) = (0. Therefore. We give an example to show that compactness is essential for this assertion. It means that ker(6’6 . 6 ( c ) )E ker(E .a’a) c ker(6’S . 00 is a nonempty closed subset of X . i. Then F is a nonempty closed subset of X. S’6(c) E ( e ) = 0. we see that there exists a b E B such that a’(b) = e and y(b) = 6 ( c ) . Hence we have (b. there would exists an z1 E A such that f(z1) = 2 0 . and. . we have i m ( a . we see that the subset A = n=l n F.1] and f : X + X is defined by f(z) = for z E X. Prove that there exists a nonempty closed subset A of X such that f ( A ) = A . e) E ker(6’6 . Therefore zo 2 z1 = 220.S’). there exists an a E A such that a(.) = b and p(a) = c . p ) = ker(y . which is a contradiction. Suppose that a nonempty closed A of X satisfies f ( A ) = A . Since f ( A ) = A iz xEA and zo E A . 2322 Let X be a nonempty compact Hausdorff space and f : X .E ) c im(P. Let F = f ( X ) .137 It follows that im(P. n E Z+} is a sequence of nonempty closed subsets in X and that Since X is compact. (Indiana) Solution. c ) E ker(y . noting that ( e . We claim that f ( A ) = A holds. Let X = (0. We define 1 1 Fn+l = f(Fn) inductively for n E Z+. In fact z o = sup z.6).
Then it is clear that the degree of 7r is equal to 2. 0) is also a homeomorphism.Z) Z . (b) Suppose that there exists a homeomorphism f : X + R2. (c) Prove or disprove: X is homotopy equivalent t o R2.(O. (a) In fact.> I ZY = 01. U and V are both contractible.O. r . U U V = A U B and U n V has four path components which are all contractible. which .o.d e g f . (b) Using part a .(O. 2324 Let x = ((2. Let 7~ : S3 + RP3 denote the universal covering map. (Indiana) Solution. . Let f be a function from S3 to RP3. X = 7 ~ 1 7r2 where 7r1 is the plane y = 0 and 7r2 is the plane U z = 0. show that X is not homeomorphic to R2. ( a ) = 2 . It is wellknown that S3 is the 2fold universal covering space of RP3. O . Denote by a and b the generators of H 3 ( S 3 ) and H3(RP3)respectively.Y.V). Since 7r1(S3)= { 0 } .(0. 0 ) + R2 . Applying the MayerVietoris sequence to the pair ( U . (a) Compute H1(X . 0).) Prove that there is no function f : S3 + RP3 inducing an isomorphism on the third homology. Denote by A and B the unit circle in 7rl and a2 respectively. Therefore.( O .138 2323 Recall that H 3 ( S 3 . O ) ) x H 1 ( A U B ) =Z $Z $Z. 0 ) ) . Z )x H3(RP3. O . we see that H1(X . =  I because deg f E Z. Then flx(o. f : S3 + S3 such that 7 r f = f . (Indiana) Solution. 1). 0. b . Take U = A U B .(0.o) : X .(O. 0 . Then we have f*(a)= 7r*(f*(a)) d e g f .O. l ) and V = A U B . Then it is easy to see that A U B is a deformation retract of X .f ( O . there is a lifting of f. M (RP" is ndimensional real projective space.O.it is obvious that f* is not an isomorphism.
l ( A ) = Z . Then F is a deformation retraction of homotopy equivalent to R 2 . is an isomorphism.f ( O .l ) be the standard ball and sphere pair in R". X onto the plane 7r2. ( X .S"1)+ H " .i(X) t f* a: t fl* i : + t f* Hnl(B") Hn(B". f1. ( X ) 1?. But it is clear that It is a contradiction. (c) Let F : X x I +X be a map defined as follows. (Indiana) Solution.(X. Show that if & ( X ) = 0 then H .S n . = Z. ( X ) = 0. is a monomorphism.l : S"l + A is a homeomorphism.H.A) a. ( X . d. A ) = Z . and therefore It is wellknown that 8: is an isomorphism and H n . = H . O . 0)).S n .l ( S n . O . Hence we have H . Hence X is 2325 Let ( B " .( O . + H.l ) in which the level rows are exact and the squares are commutative.139 wouid induce an isomorphism from H1(X . Consider the following diagram H . Since H . Suppose that f : ( B " .l ( B n ) = 0. . 0)) to H1(R2 .A ) M imd. Thus we have keri. By the assumption. M ker i.i(A) i* t H.l ) + ( X . A ) is a continuous map and f l s n . . n > 1.
21) E S2 x S5. (b) Let X = S 2 x S 5 . (a) Let xo E S 2 . and 7. (Indiana) Solution. and UUV = S 2 x S5. Therefore the CW structure on S2 x S 5 has a 0cell. ez e z ez. What can you say about the possibilities for the relative homology groups H . Hence it is obvious that H i ( S 2 x S 5 ) is an infinite cyclic group for i equal to 0.A)? (Indiana) Solution. and is zero otherwise. Then S2 is obtained by attaching a 2cell t o 2 0 .5. 2. 5 (b) Compute the homology of S2 x S5 with 2 points removed.140 2326 (a) Describe a CW structure on S2x S5 and use it t o compute the homology of s 2 x s .{ p l . State what the homology groups H . x 1 E S 5 . Denote by S2 V S5 the one point union of S2 and S5. 2327 Let A be a subspace of S 2 x S2 homeomorphic to the 2sphere S2. which can be considered as the space obtained by attaching a 2cell and a 5cell to the point (20. Then U . ( S 2 x S2. By the Kiinneth formula we conclude that otherwise.2.{ p 1 } a n d V = S 2 x S 5 . V are both open sets of S2x S5 and UnV = X. p 2 } . ( S 2 x S2)are (no proof is required). a 2cell. . V) and using the fact that we conclude that i = 0. and S5 is obtained by attaching a 5cell to XI. otherwise. It is easy to see that S2VS5 a deformation retract of both U and V . a 5cell and a 7cell.{ p 2 } . 5. U = S 2 x S 5 . It is easy to see that S 2 x S5 is homeomorphic to the space obtained by attaching a 7cell to S2V S5. Applying is the MayerVietoris sequence to the pair (U. i = 1.6.
) Show that there is a natural isomorphism ax : H.(X) H i + l ( C X ) . It is easy to see that X’ has the homotopy type of 5’. 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 . 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. It is clear that Hi(X. where i. Let X’ = X .A ) M keri. a. for all i 2 0. ( S 2 x S 2 .X”) conclude that Hi(X) infinite we is cyclic for i equal to 0. (This is the union of two “cones” on X. (Indiana) Solution. Thus from the homology sequence of the pair ( X .. and H 2 ( S 2 x S 2 .using the above result and the fact that H .A ) + H o ( A ) + Ho(S2 x S 2 ) + H o ( S 2 x S 2 . X”) infinite cyclic for i equal to 1 and is zero is otherwise.{the open interval (1. ( A ) M Hi(S2). Since &(A) = 0. x we have H3(S2 x S 2 . A ) = 0 for i equal to 0 and 1.2 and is zero otherwise. A ) .. is the homomorphism induced by the inclusion map i : A S2 x S 2 . Then the space X may be viewed as a space obtained from X* by attaching a 1cell.we see that H . H. A ) 4 H 2 ( A ) 5 H 2 ( S 2 x S 2 ) 5 H 2 ( S 2 x S 2 .(S2 S 2 . 4 (Indiana) .1. By the exactness.141 From the homology sequence of the pair ( S 2 x S 2 .   2328 Compute the homology groups of the space X obtained as the union of the 2sphere S2 and the xaxis in R3. A )+0 0 + H 1 ( S 2 x S 2 .l) in the xaxis}.A ) M Z @Z/irni. A ) +0.
T 2 ) .) 20 Fig. (See below.. Take U = EX . ( X .16 (a) Calculate H. + a: H 1 ( T 2 )2 H 1 ( X ) + 0 Qt f* t fl* 0 + H2(D2..(21). ( X ) . It is wellknown that H . ) : V .. The inverse of A * . T 2 ) .2. H ... T 2 )= 0 for i # 2 and H 2 ( X . where D 2 is attached to T 2 by identifying dD2 with a meridian curve S1x (20) in the torus. V s l +~ A ++ +  It is clear that U and V are contractible and that U n V has the homotopy type of X .. d D 2 ) ) Z . + ~ ( U ~ 3 )H i + l ( u > ~ H i9 H(~~ > ( E x ) H i ( u n v ) .where z f * is the homomorphism induced by the adjunction map f : D 2 + X . and U u V = EX.(X) for all n 2 0. Consider the MayerVietories sequence of the pair (U. (a) Consider the homology sequence of the pair ( X . (b) Is T 2 a retract of X ? Why or why not? (Indiana) Solution.142 Solution.a D 2 ) 4 w H l ( d D 2 )+ 0 . Then U and V are open sets of E X . The naturality of ax can be derived from the naturality of the MayerVietories sequence. where 20 E S1. 0 4 H 2 ( T 2 )5 H 2 ( X ) 1 H 2 ( X . CTX : & i ( X ) + H i + l ( C X ) is just what we are looking for. 2330 Denote by X the union of the torus S1x S1with the disc D 2 . Let (0)’x X and (1) x X be identified to the points 20 and 2 1 respectively. T 2 ) M f * ( H 2 ( D 2 . 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 & . Thus the nontrivial part of the sequence is as follows.( 2 0 ) and V = C X .
Hence we conclude that (b) We claim that T 2 is not a retract of X. Let i : T2+ X be the inclusion map. consequently. (See Fig. This is a contradiction. we have T* o i. Let C = 2/82 and let g : C + B be the surjection defined by g(1) = 1. We know that H 1 ( T 2 ) = Z @Z is generated by a and b. we have the following commutative diagram  because of H 1 ( T 2 ) = Z @I and H 1 ( X ) =Z. and let h : A + B be the surjection defined by h(1) = 1. = id : H 1 ( T 2 ) H 1 ( T 2 ) . + (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 ) . a )E C x A 1 g ( c ) = h ( a ) } . Prove that the kernel i of g*(A)is isomorphic to the kernel of h.+ B 9 (b) Let A = Z / U . and. In other words. Otherwise. (Indiana) .with explanation. Identify the group g * ( A ) . B = 2/Z. there would exist a retraction map T : X .2. Then r o i = i&z : T 2 + T 2 . 2331 Given homomorphism h : A B and g : C + B .143 It is easy to see that ima: = irnfl. and ker& = ker f 1 + .16) It is clear that f 1 * ( H l ( a D 2 ) )= 2 @ (0).+ T 2 such that ~ ( =pi d p . 9*( A ) c 1 g*(h) A l h . 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 .
(4. g'(O) = {0. (1.X .C ) @ Hi(S3 .{m}. Then g*(h)(c. I>... .0). 2332 Recall that if C is a homeomorphic copy of the circle in S 3 .e. then H i ( S 3C) is infinite cyclic for i equal to 0 or 1 and is zero otherwise. Assuming this fact compute (a) The homology of R3 .2}. u ) E kerg*(h). It is easy to see that F is an isomorphism. C = l. + H.1>... We have the following MayerVietoris sequence. 0. (3.C.0). ( C +  Noting that Ga(S3 .6} and g'(l) = {1. (2.~ . (a) Denote S3 = R3 U { }Let A = S3 . Let F : kerg*(h) + ker h be defined by F(l. (6. a ) = a.3).2). the identity of C. the onepoint union of two circles.C. I). i = 3.(7. (4.. Then A and a . (7. (6..7}. (b) In this case.+'(S3) 2 Hi(R3.3}.B = S3 .1).2). Thus it is clear that the group g*(A) consists of the following 16 elements: (O. and Hi(S3 .3).5. we have h'(O) = {0. (a) Suppose that ( c .R )~+ ..144 Solution.C ) = .3). therefore. B are open subsets in S3. ( L 3 L (3.2. h'(l) = {1. otherwise.2). when C is a homeomorphic copy of the circle in R3. (2. (5.2). . where X c R3 is a homeomorphic copy of the "figureeight space" (i. (5.{ }= G@3) = 0 a ) for any i and that H i p 3 )= { z.a) = l. and A U B = S 3 and A n B = R3 . (b) The homology of Y = R3 . we see that kerg*(h) = {(l. and.C ) % Hi(S3 .O). By the definition of g*(A).. (0.a) I a E kerh}.3.C.0).4.) (Indiana) Solution.{co}) % Hi(s3)A H ~ .
by the definition of ++ we have [c] = + + ( [ c ] .v )= R~ .V ) . we have Y = R~ . It is easy to see that iii(x3 (v n v))M iri(s2)>. 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 . the nontrivial part of the MayerVietoris sequence of the pair (R3 .2. %A7 @Z  MZ + A We claim that the homomorphism is an epimorphism.U n V .V ) + 0.X . otherwise. R3 .17. Fig.0).U ) and gi(R3.145 we can see that  = {: i = 0.(u nv).U ) n (R3 .u) n ( R ~v) ~ and ( R .2.U.V )is infinite cyclic for i equal to 1 or 2 and is zero otherwise. . The claim is proved. Let U = X .( ~ 2 ) .U n V). Due to the above facts. which means imA = 0 too. Hence we can consider the generator of H 2 ( S 2 ( r ) ) .{PI} and V = X .[c]. Then S 2 ( r )is a ) deformation retract of R3 . Then we see that U and V have homotopy type of S 1 and that U n V is contractible. Since U u V = X .V )5 i i 2 ( ~ u n V > 2 U 3 3 %Z$Z H 1 ( Y ) 2 ii1(R3 .u )u ( R .as the generator of E2(R3 .U ) @ ii1(R3 . Thus we get H 1 ( Y ) = Z@Z and a split exact sequence   O + & ( Y ) 5ZCBZ5Z+O.2.1.V ) is o + ii2(y) i i 2 ( ~ 3 ) CB i i 2 ( ~ .17 Represent X as shown in Fig. ~ ~ From the result of (a).x = ( R. we see that Zi(R3.
So we conclude that H l ( Y ) = {:w Z7 I 2333 i 2 3. ~ Z. It is known that if X c S3 is homeomorphic t o S1 then H . this fact to compute the homology of S 3 . Denote Y by Y = A U B.(S3 . .H .~ ( s . q#3.A ) U ( S 3 .+l(~3) . * . H .B ) M H. 3. H.146 It follows that HZ(Y) =Z. 4.(S1). i = 2.B)= S 3 . where A n B = 8 and both A and B are homeomorphic to S1.Y )5 . 4'1. S3 . 4 = 1.Y )=  { 2*Z. A  A  &(s3.Y = ( S 3 . ( S 3 .(S3) = we can easily see that { Z.ii. 4=2.A ) n ( S 3 . . 0. ( s ~ .. . Therefore. ( s ~ ) .A ) @ &(s3 .X ) M H. i=l. .B ) . in the MayerVietoris sequence we have *. ii. Using the fact that Z. Noting S3 .A ) M Hq(S3.B are open sets of S3 and ( S 3 .(S1) = and that 4 # 1.Y ). 4 = 0. (Indiana) Solution.A and S3 . .Y where Y is a subspace Use of S3 homeomorphic to the disjoint union of two copies of S1.(S3 .B) H . i = 0.
the homology of S2 V S1V S 2 .2. Fig. the 2fold covering map: z + z 2 from S1to S1. Thus. (a) .18 The universal covering of the one point union S1VS2 is shown as in Fig.e. is the identity map. Then it is easy to see that S1is a deformation retract of U n V . it is clear that . we take U = S2 V S1V S2 . (Indiana) Solution. restricted on each S2. Fig...{the antipodal point of p 2 } (see the above figure). (b) Compute the homology of the connected 2fold covering space of S1vS2.18. reis stricted on S1.19 The covering map T . and. restricted on each S2. i. is the exponential map: R + S1. and. 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).is the identity map.2. The covering map T . and that S1V S2 is a deformation retract of U and V. b) To compute the homology of the connected 2fold covering space of S1v S 2 . restricted on R..2.{the antipodal point of p l } and V = S 2 V S1V S2.
4 # 172.148 and &(U) M &(V) x H. we have H. s ~ M Hence. : H1(U n V) + H1(V). the nontrivial part of the MayerVietoris sequence is c Z. It is clear that   H ~ ( v s1v s2) H ~ ( u ) kl(V)/im#. It is easy to see that the inclusionmaps k : unv + U and I : U n V + V induce injective homomorphisms k.e. Hence the homomorphism   4. . (Indiana) Solution. i. ker 4%= 0.. respectively. Fig.(S1) = Therefore.(S2   v s1v S 2 ) = 4 = 1. 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 . MZ. 4 = 1121 0. : G1(u n v)+ H1(u)@ H1(v) is injective. Thus imA = 0. 2335 Compute the homology of S1 x S1point.20 It is easy to see that the space X = A u B is a deformation retract of the space H = S1x S1point.(S2) @ H. where A and B are each homeomorphic to S1and . : H1(U n V) + k 1 ( U ) and I . 2. 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. otherwise.
and consequently.{ b } . 7. b ) = . 07.u E kerj. b } is contractible. Then 7 ’ o j L ( u ) E ker6. is an isomorphism for all n..u. ( X ) = H .{ u } . Construct an exact sequence . b).(b) = j k ( u ) . ( A ) @ H .(a’) = Pn(b). and there exists an a’ E A. Due to the exactness . Thus Pn(b). b) E A. Hn(S1 S1. ( u ) . Let U = X .2.(a’) .lojA).P n ( b ) for any (a’. at C. . o@. + A .. (Indiana) Solution.. @ B. and that U n V = X .there exists a b E B. the rows are exact. ( d )= A(a’. and 7. j n ( b ) = jA(b).. i BL + Anl + * * * Write out the proof of exactness at Bh. +A’. respectively..(b) . Applying MayerVietoris sequence t o U and V.. we have nL2. such that jn(b) = 7 ’ o j . ( B ) = x 2336 Suppose the following diagram is commutative.149 A n B = {zo} as shown in Fig. We give the proof of exactness at Bh as follows. Let u E ker(6. we have j . The following sequence is exact: where A is defined by A(a’. and let V = X . @ B.{ a . .point) = H .i . From the commutativity. It means that u = P. It is clear that A and B are deformation retracts of U and V . .20 Choose points a E A and b E B such that u # z o and b # 2 0 . such that i.
o j . 1) g(z). ) C imA. 0 jL(U) = . H.(b) for some a E Ah and 6 E B. . + A  H .l ( X ) .P. Then we have M = U U V and U n V = X x (1/2.3/4).(u n V ) + H J U ) @ I?p(v)+ H. M is the identification space obtained from the disjoint union of Y . where X x [0.(M) + H q .. 2 In the same way. It is obvious that U n V has the same homotopy type with X .. 0) II f ( z ) . (Indiana) Solution.(Z) = 0. o Pn = 7. Prove that H. 1) with g(z). g : X + 2 are two maps of X into contractible spaces Y and 2.     . we have H. .(b) = 0.3/4] and Y .l ( V ) + ’. we have the following exact sequence: ’  * . u = ih(a’) .1 ( U ) @ H q . is proved. Suppose that u E imA.3/4] = Y/.3/4] = Y denotes the disjoint union of X x [0.(V) E H. In the same way. Noting U . and each (z.~ (n v > u (1) II Since X x (0) = Y / is homeomorphic to Y and is a deformation retract of U . let V = X x (f. where the equivalence     relation is determined by (z. Thus. from the assumption that Y is contractible. o 7 ’ ojk). ’ Since j .. . 6 0 j. Let M be the mapping cylinder o f f and g . . II 1 I Let U = X x [0. it is easy to see that 6 0 7’ .+ H.1] = Z/ ’. Thus the exactness a t B. . the equivalence relation is determined by (z. X x I and Z by identifying each (z. o ih = 0 and j . by MayerVietoris sequence.M Z H. V and U n V are open sets of M . i.k . from (l).1X. .150 So ker(6. Now we prove imA c ker(6. So k .e. ( U n V ) Z & ( X ) . Suppose that X is a space and f : X + Y . ( Y ) = 0.(U) H . ( M ) Z k q . o 7 ’ o j .O) with f (z). that is.
Part 111 Differential Geometry .
.
Define the diameter d. > N . by Blaschke.3. where the oval is counterclockwise orientated and 8 denotes the oriented angle from the zaxis to I. of a ( s ) to be Now assume that the curvature k ( s ) 2 1 for all s. y(e) = p(e) cos e pye) sine. (Indiana) Solution. Prove that d.153 SECTION 1 DIFFERENTIAL GEOMETRY OF CURVES 3101 Let a ( s ) be a closed plane curve.y) of the oval.1 ii) From the hypothesis that k ( s ) 2 1 for all s. i) For any N E Z + sketch an example of such an a(. 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.(e) = p(e)sin 6 p'(0) cos 8.) with d. 5 2 (or some other constant independent of a . i) The following is an example of such an a ( s ) with k ( s ) 2 1 and d. then the oval can be parameterized by 8 as follows . + + . For every oval. we know that the simple closed curve a is an oval. 2 is the best possible such constant). ii) Assume further that a is a simple closed curve. "t Fig. > N .
the relative curvature of the oval is k. Suppose that. the curvatures of C and C1 satisfy the inequality k l ( P l ) 5 k ( P ) . From the above. 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 (. we can say function p ( 8 ) must be the solution to the following initial ODE origin. p ( 8 ) can be uniquely determined by k ( 8 ) = k.3. Fig. a t every pair of points P and PI where C and C1 have the same tangent orientation. Let p ( 8 ) and p l ( 8 ) be the of C and C1.2 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. Fig. by direct computation...3 In fact. Hence. Then the domain encircled by C1 must contain the domain encircled by C. . From this we can conclude that.154 p ( 8 ) is called the support function of the oval. y(8)).(8) = ( p ( 8 ) + (8) ) . take the tangent point 0 of C and C1 as the common tangent line as the zaxis. respectively.1.(a).3.(8) of C.
we also have the similar expression of p l ( 8 ) . Determine all such curves. where 8 and its corresponding original 8 satisfy 8 + 8 = 27~. we know that the curve a must be a circle (or a part of it). Therefore. For convenience. . if 0 5 8 5 x. we always have p l ( 8 ) 2p ( 8 ) . Thus.155 Analogously. firstly. By the hypothesis. for the curve C1. Suppose " the normal vector N ( t ) is proportional to the position vector. then ii) follows immediately from the above assertion. Hence. where c is a smooth function. (Indiana) Solution. 2 0. if x 5 8 5 2x. Secondly. we have.4)  I  + 3102 Let a be a regular C curve in lR3 with nonvanishing curvature. Differentiating both sides of the equality N ( t ) = c(t)cr(t)with respect to t . T ( t ) = 0 and c ( t ) = k. we know that As for the sign of sin(8 . then owing to 0 5 4 5 8. Then . Thus we can immediately obtain k ( t ) = constant. that is. 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. by the fundamental theorem of the theory of curves. The assertion of Blaschke is proved.can get we pl(8) . we assume that the parameter t is the arc length of the curve a. we see that the domain encircled by C must be contained in the domain encircled by C1. and take a as C . Noticing that every oval is the envelope of all its tangent lines.b). Therefore. N ( t ) = c ( t ) a ( t )for all t . we have sin(8 .p ( 8 ) 2 0. p l ( 8 ) .p ( 8 ) 2 0. by the Frenet formula. If we take a circle with radius 1 and centered at a ( s g ) N ( s o ) as C1.
a) Let a circular helix be parameterized as follows y(s) = (Tcosws.v ) = N ( u . Lz.v) = 0 for all ( u . N ( u .w)du2 + 2 M ( u . there are three different asymptotic directions at every point of s.dv) satisfies L(u. Assume there exists a unit vector a E R3 with T . then. v). M ( u . Observe that every asymptotic direction ( d u . v ) are the coefficients of the second fundamental form of S at point p ( u .Tsinws.v)dv2 = 0.156 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.v).v)dudv + N ( u . every point of S is a planar point. v ) . If S is a triply ruled surface. Determine all triply ruled surfaces. L3 lying in S such that L1 n L2 n L3 = { p } . we obtain L(u. a) Show that a circular helix is an example of such a curve. (Indiana) Solution. Noticing that the above equation is of 2nd order with respect to du : dw and it has two roots. N . (Indiana) Solution. by the hypothesis. a = 0. Therefore. B } . b) Show that N . a = constant = cos a. .v). c) Show that k / r = constant = k t a n a . both nonvanishing. S must be a plane (or a part of it).b) 4 R3 be smooth with [y’l = 1 and curvature k and torsion 7.w) = M ( u . In other words. where L(u. Denote the Frenet frame by { T . 3104 Let y : (a. hws).
for all s E ( a . with curvature and torsion k ( s ) and ~ ( s ) . l).b ) .rwcosws. Hence T ( s )= (rwsinws.k ( s ) / ~ ( s ) & t a n a . h.a = 0 for all s E ( a . = 3105 Show that if y is a geodesic on the cone t = then y intersects itself at most a finite number of times. T(s ) ) N (s) . a straight line never intersecting the generator 1. d m .157 where T . If we take a = (O. b ) . then we can conclude that y never intersects itself. we have 0 = (cos a k( s) f sin a . w = ( r 2+ h 2 ) . c) By the property of b). an obtuse angle.hw). Assume k ( s ) # 0. If we cut the cone along a generator 1 and develop it into a plane.O. b) Differentiating T e a = constant with respect t o the arc length parameter s.a = 0. 3106 Let y : (a. Differentiating the above equality with respect to 3 s. we obtain k ( s ) N ( s ). Noticing that the central angle of the sector is f i r . a = hw = constant.i are all constants. two rays which start from the generator I . b ) . B ( s ) . then the cone becomes an infinite sector without the vertex. we may assume that the constant vector a = cosa. and the geodesic y becomes a straight line on the developed infinite sector.b) 4 R3 be a C" curve parameterized by arc length. and the image of y on the sector must be one of the following three cases: a generator.0)) (Indiana) Solution. ~ ( s # 0 for all ) s E ( a . Thus. (z)y) E R2\{0. from which follows N ( s ) . Then it is easy to verify that s is the arc length parameter. and let T and N denote the unit tangent and normal vectors to .T(s) f s i n a . then T ( s ) .
(Indiana) Solution.C2 such that for all s E ( a . there exists a constant C such that r k ( s ) C T ( S = 1 for all s E ( a . In this case. (b) Prove that if y is a Bertrand curve (with r as above).b). namely r = const. (c) Suppose that y1 and 7 2 are the Bertrand mates of y. then there exists a constant C such that r k ( s ) C T ( S= 1 for all s E ( a . . we obtain  k ( s ) N ( s ) .158 y.b). ds If we denote  T ( s )= a ( s ) T ( s ) b ( s ) B ( s ) . (Note that s might not be an arc length parameter for 7. rg. by (b). ds dz . we have that r’(s) = 0. b ) the normal lines of y and 7 a t s are equal. b) + IR3 such that for each s E ( a . (b) From (a). Then.) (a) Prove that r is constant. + then by differentiating with respect to s . then y is a circular + helix.r k ( s ) ) = const. r ( s ) ) N ( s ) b’(s)B(s).r ~ ( s ) = const. namely (1 . The curve y is called a Bertrand curve if there exists a regular curve 7 : (a. now we have  T ( s ) = (1 ds ds rk(s))T(s)  ds TT(S)B(S). there exist constants T I . (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). from N ( s ) = fN(s) we know that a’(s) = b’(s) = 0.= a’(s)T(s) ( a ( s ) k ( s ) b ( s ) . dZ ds + + + Hence. ds dz ) Therefore. + C27(5) = 1.b). C . 1 r 1k( s ) r2k(s) + + ClT(9) = 1. 7 is called the Bertrand mate of y and we can write y(s) = y ( s ) + T N ( s )for some T = ~ ( sE) IR. ) (c) Prove that if 7 has more than one Bertrand mate.
Differentiating (z(s). Differentiating ) ( z ( s ) . Vs. T ( s ) ) 0 with respect to s again.N ( s ) ) = 0. differentiating . still differentiating (z(s). 12 are not equal. N ( s ) ) = & with respect to s. 3107 Let z(s) be a curve in R3 parameterized by arclength. ( s ) . ~ ( s = const. Suppose that z(s) lies on a sphere centered at the origin.k’(s) k 2 ( S ) T ( S) * Conversely. we obtain = + + 1 + k ( s ) ( 4 s ) . (Indiana) k4(s) S o ht ion. we have ( z ( s ) . Then we may assume that Z(S) = a ( s ) T ( s ) b ( s ) N ( s ) c ( s ) B ( s ) .B ( s ) } N is the Fkenet frame field along z(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. Assume that ~ ( s # ) 0 and k’(s) # 0 for all s. Hence y must be ) a circular helix. c ( s ) are suitable functions to be ascertained later. b ( s ) . where { T ( s ) . and a ( s ) . from which it follows that u ( s ) = 0.159 Because the nonzero constants T I .P ( S ) 1 +(s) P ( S ) = constant. t ( 8 ) . T ( s ) = 0. we obtain namely. z(s)) = R2 with respect to s. c(s) = . which means that b ( s ) = 2 At last.
Then. 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 . because neither the top circle is a straight line.160 with respect to s. . Is this curve a geodesic on M ? Explain without long computations. 1) (“top circle”). which is orthogonal to the zaxis. z ) : (y 2)’ z’ = 1) around the zaxis. the curve c ( t ) is not a geodesic on M . namely. Hence (z(s) . y. where k ( t ) is the curvature of the curve. 3108 obtained by rotating the circle ((0. Let M + c R3 be the torus (Indiana) Solution. nor its principal normal. we have that vanishes identically because k’(s) # 0 and Then is a constant vector. denoted by m. and let c ( t ) = (2cost. Observe that the geodesic curvature of a curve on M can be computed as follows k .m .m) = constant. 2 sint. which is parallel to the zaxis. z(s) lies on a sphere centered a t m. = f k ( t ) sin€J(t). is parallel to the normal of M along c ( t ) . z ( s ) .
in the direction of the circle of latitude. f(zp)). that is. which means that y ( s ) B = constant. y lies in a plane. Assume the torsion 7 = 0. f'(zp)= 0. . Namely. Let M be the surface in lR3 obtained by rotating the graph {(z. Thus. f ( z ) ) : z E B}of f in the z z plane about the z axis. (a) Show that y lies in a plane. is a principal curvature. i. SO is a stationary point of y's tangent vector field. Characterize in t e r m o f f the set of z such that && is a principal curvature of M a t (z. the curve {(z. $(y(s)B) = 0.e. (b) The vanishing curvature at point SO means ? ( S O ) = 0. *$J . " .161 3109 Let y : [0. Thus.. 3110 Let f : lR + lR be positive and smooth. 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. Ic. denoted by B .O. At P = ( z p 0. Hint.1] + lR3 be a C curve with 1il = 1 and nonvanishing curvature. that is. (Indiana) Solution. is constant. f ( z ) ) . = 1 implies that the curve is parameterized by arclength s. 0.f(z)) : z E IR} has a tangent parallel t o the z axis at P . (a) 1. (b) What happens if the curvature is allowed to vanish a t a point? (Indiana) Solution. then the corresponding normal curvature Ic. From il the Frenet formula we know that the binormal vector field of y. then it is an inflection point of the curve 7. Since every circle of latitude on M is a line of curvature. If the point is the strict extreme value point of the tangent vector field. Local coordinate computations are not necessary. = implies that cos8(zP) = f l .
i) The hypothesis of torsion ~ ( s ) 0 implies that the curve a is a plane curve. T i) If k ( s ) > 0 and T ( S ) are curvature and torsion of a in E3.) ii) If Icg(s) 1. Integrating the obtained equality. (Indiana) + Solution.)) 0.) E M 2 is parameterized by arclength. 0. show that a ( s ) is a closed curve (kg is the geodesic curvature in M). Its curvature Ic a t P is just the other principal curvature of M .) is always a linear combination of the binormal and normal vectors. a(. whereas the hypothesis of curvature k ( s ) > 0 implies that the plane R . a ( s ) = const. Therefore. 3112 Let M 2 c IR3 be the cylinder z2 y2 = 1. show that a ( s ) is a closed curve. Assume that the position vector a(. we ) will get a ( s ) 5 0. Suppose the curve a(. if a ( s ) passes through the origin. (Here k . we have ( a ( s ) . Thus. the trivial case.O. the second possible case is Therefore. we conclude that the set of L such that f of M a t (z. the meridian passing P is also a line of curvature.162 Besides. (Indiana) Solution. Show that a ( s ) does not pass through 0 E lR3. If the position vector a(. f(z)) is & is a principal curvature 3111 Let a(.) c R3 be a smooth curve parameterized by arclength.) is always a linear combination of the binormal and normal vector B ( s ) . N ( s ) of a ( s ) . then ( T ( s ) .
Therefore. = (z. 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.O. a must be closed. z ( s ) ) where z ( s ) is unknown. i) That z& = and z&. (Indiana) Solution. ii) Given a C curve a ( s ) E B2 = ((z.Y ) . &A) where A = area of R. Thus the curvature of & is = 1 in x. ii) If one develops M into a plane x. we may assume that p(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(.y(s).p ( s 0 ) = (O. then the corresponding plane curve of a .)) = a ( s > . has the same geodesic curvature kg(s) 3 1. & + &} + + ”> for suitable functions a ( s ) and b(s). ii) Let a ( s ) = (z(s).e.B’(s) E Tp(. Y. i..y. the curve a c M n x must be closed. are linearly independent shows that T is not involutive.O) : z. iii) Show that if a ( s ) is a simple closed curve of length L bounding the region R in B2 then @(so + L ) .y E IR} C IR3 and ” a ( s o ) . . y(s). By x ( p ( s ) ) = a ( s ) . So. 0).O). denoted by &. where x(2. Then .163 where the curve a lies does not parallel the generating line of the cylinder M .z(s)(:9 d2 dX [& & + I & { &.f3’(s) E ” q 3 ( $ ) . ds ’ ds ’ ds Hence Thus the problem of finding problem p(s) reduces to solving the following initial value .) implies that d p ’ ( s ) = u(s)b(s) .
respectively. = (sine. T N  = . (ICT... we may assume that (vI G 1. then we easily have + L ) . 3114 Call a normal vector field Y along a space curve y parallel if l is always i tangent t o y. Here s and L denote arclength and the length of y.e. .cose k~ ( B . B ( s ) . + i. i) The hypothesis that l is always tangent to y means that (v.where e(s) is a smooth function globally defined on y.P(s0) = (z(so + L ) .fA). without loss of generality. f ZdY) = (O.TB)+ s i n e . . B ) B +case.Y(SO>.. 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. (Indiana) Solution.l ) 0. N + (4  T )c o s ~B. i.y(so f L ) . Thus we have ) I” r ( s) ds . Therefore. and i i= hence. N + case.z(so). where the sign is determined by the orientation of the curve a.164 When the given plane curve uniquely determined by Q(S) is C”. iii) If P(s0 Q(S) is simple and closed. the normal vector field v has constant norm. N ( s ) sinB(s) .)sine. ii) Show that the total torsion of any closed curve y which lies on a sphere in lR3 must vanish. i. which measures the angle between ~ ( s and N ( s ) .O.e. and v ( s ) = cos O(s) .the unknown function z ( s ) can be and so is the space curve P ( s ) .
the smooth function e ( s ) . t E [0. for every ytrwe have the corresponding map ft such that fo = f . Let k ( s ) be the curvature of the intersection curve C = M n P . s E [O.. Therefore. along the segment where k ( s ) # 0. . we have the relation qS) J.deg f1 = 0. i. the normal of M is parallel to the principal normal of C. = ra T(S)dS +qo). where the integer n is just the degree of the map f. ii) For any closed curve y which lies on a sphere in El3. n = deg f. Furthermore. Let po E y and y contract to po smoothly on S 2 . yt overlaps with y = yo about po = yl. we may assume that for every t E [0.1]. fact that y is The closed means that f +)dS = e ( q . finally we have I” T(S)~S = 27r . Otherwise. B ( L ). Suppose M and P intersect orthogonally.L]can be regarded as a lift of a certain differentiable map f : [O. Because the degree of a map is homotopically invariant.we see that the above equality implies that = ~ ( s )and hence .e(o) = 2n7F. Thus we get a family of curves { y t } . then C is naturally a geodesic on M . Show that the intersection curve (parameterized by arclength) is a geodesic on M .hence the segment is also a geodesic one. l). L] S1into R1.q o ) = J 0 L r(s)ds. Also. fi = the constant map into PO. (Indiana) Solution. the unit normal vector field Y of the sphere along y satisfies the above mentioned hypothesis. On the other hand.e. If k 5 0.165 e(s) Noting that i/ is parallel to T . 3115 Let M be a surface in El3 and let P be a plane.
then a”(s) 0. for each t E lR. we have which implies that f(s)= k2(s). What can you deduce about f(s).. If f ( s ) # 0. a ( s ) is a spherical curve. then ( ( ~ ( s ) )= const.e. where c E lR is an arbitrary fixed constant.e. 0 3117 Suppose y(t) parameterizes a space curve with curvature function K ( t ) . T(S) = 0. i.) be a C2 curve in lR3 parameterized by arclength. Derive the curvature function for this new curve. a ( s ) is a straight line.166 3116 Let a(. differentiating with respect to s. (Indiana) Solution. k’(s) = 0. Hence k ( s ) = const. From 1 f ( s ) a ( s ) a ’ ( s ) d(a’(s))2 = 0 = 2 we have f(s)d(a(s))2 = 0. Define a new curve 7 by setting. i. a”(. . Further~ more. If f ( s ) 0.) = f ( s ) a ( s ) . and a ( s ) is a circle or part of it with radius 1 .. (Y(S)? (Indiana) Solution.r ( t ):= cy(t/c). Suppose that for some function f(s).
) . (O. maybe. then. i) Show that if the torsion of a vanishes identically then there is a plane ?r c lR3 containing a . ( t . ?rl # x2. ii) If the curvature k is allowed to vanish a t a single point c E ( a . Hence. The condition k # 0 means that we can ~) 0 define the F’renet frame field { T .] (Indiana) Solution. . namely.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.f ( t ) . A counterexample is as follows. .0) where the function f is defined by . ( t . c ) must be on a plane T I . $ ( a l l ? )= 0. a ( t )E ?r for all t E ( a . according to the above discussion. b ) .a ( t ( s g ) ) B ) = 0 contains the curve a. swith t ( s 1 ) = a . suppose a = a ( t ( s ) ) . t ( s g ) = b. the connectedness of a and the smoothness of the Frenet frame field imply that T is unique. Then the hypothesis T implies that B is a constant vector field. i) Let a be reparameterized by arclength s. that is.b). but is assumed to be zero a t all other points. s E ( ~ 1 .O. Obviously. the torsion is not defined at c. B Therefore. Of course. f ( t ) ) if t > 0. Let a(t)= { if t < 0. b ) must be on another plane a2. 0) if t = 0.N .b) + lR3 be a smooth curve with nonvanishing curvature.t E ( a . the plane ?r : ( p . from which follows ( a ( t ( s ) )B ) = const = (a(t(so)). and a ( t ) .b)? [Of course.t E ( c .167 Therefore.0. from y ( t )= y ’ ( t / c ) and y”(t) = y ” ( t / c ) / c we have 3118 Let a : ( a . a ( t ) . B } along a.
The curve . (Recall that a is a generalized helix if for some constant vector u # 0. (u.a’(s)) const. Using the Frenet formulas and the hypothesis (T(s).= . Differentiate the obtained expression for . .c ? . Differentiate the equation with respect t o s. that a plane contains it. Deduce that in this case.B(sl) = a ( s ) X ( s ) T ( s ) . 3120 Suppose that the unit normal vector to a surface M c lR3 is constant along c M . u2).) (Indiana) Solution. . a is an asymptotic (Note: A curve in a surface is called asymptotic if its acceleration is everywhere tangent to that surface. Noting that a being a generalized hrelix implies + + we obtain by straightforward calculation i... P ( t ) lies on the line tangent to a a t a ( t )and (a’(t). Let a be parameterized by arclength s .. Firstly.f3’(t)) 0. by the Weingarten formula. (Indiana) a regular curve a Solution.W d X n ds iJ=l 2 ds aut (Z) = W(a’). p is a plane curve. the torsion of p vanishes everywhere.. we have.. and let s . duj o = d= . Then we may assume .e.Tl(sI))= 0.) curve. s1 represent their corresponding points. let a and . be parameterized by their arclength s and B s1 respectively. E where s is arclength for a.s)T(s). For convenience. = Show that every involute of a generalized helix a is a plane curve.f3 is called an involute of a if for all t . Therefore. and that the Gauss curvature of M vanishes a t each point of a.B successively. we can ascertain that P ( s 1 ) = a ( s ) (so .168 3119 Let a and p be two regular curves in lR3. along a . and M be locally parameterized by X(ul.
a ( s g ) . by the Gauss formula. we have d ( a ( s ) .) (Indiana) Fig.a')= 0. is geodesic curvature. = where k. ( ( ~ ( s n) = const = (a(so). along a . namely. Hence a" is everywhere tangent to the surface. and has the following graph.O) andP'(0) = (1. satisfies r" Jo L(s)ds m 3 ~ 1 2 . n )= 0.2HII K I = 0 it follows that along Q + K =.T .0).a ( s o ) . which means that the curve a is contained by a plane ( p . ). Secondly.:2 ) ( + 2H(W(a'). Thirdly. noting that n is constant. ( a ( s ). we immediately deduce that. ifp's curvature function k ( s ) ( s = arclength) is odd. from I11 .4 . 3121 Sketch the closed regular plane curve p : [ . T]+ R2having p(0) = (0.3. n). along a 2 2 duj duk = kgn x a' + (W(a').a')n kgn x a'.n) = 0. n )= 0. (Include an explanation with your sketch.169 Further. Therefore.
which has zaxis as its tangent line a t (0.B’(s) and p’(0).5. too.3. where 6 ( s ) is the oriented angle formed by .170 Solution. And yaxis is almost its “tangent” line at ( O .3. O ) . as shown in Fig.O). This assertion follows from I” k(s)ds = 3a J.” g d s = 6 ( ~) 6(0) x T .5 . The image of the plane curve p is a “figure eight”. Fig.
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.Compute interior of Q J k. d s 2 ) + . For every p E R 2 .(x')*ds2) is an isometry.) be a simple closed geodesic as described above and K ( z ) be the Gaussian curvature of the above metric at 2 E B2. ii) Let a(.By direct calculation. ~ 2 ~ 2 E S2\{N} to (u.v). . 2 2 .1 u2+v2+1' 212 + Then the metric of S2\{N} can be expressed by ds2 = + dv2) (22 + 212 + 1 ) 2 ' 4(du2 Now. we can obtain a 2dimensional Riemannian manifold ( B 2(x')*ds2). using the pull back of ds2 by (x')*. Since all great circles which do not pass through the north pole are closed geodesics on S2\{N}. . which maps (21. Thus. then their images are simple closed geodesics on ( E 2(xl)*ds2). (B2. (Indiana) Solution. Here the integral is with respect to the Riemannian volume induced by the metric. and B2 be the 21 o 2 2 plane. we obtain 3 ) + + 21 = u2 + v2 + 1' 2u 22 = u2 + v2 + 1' 2v 23 = v2 .v) E B2. i) Let S2 = { ( 2 1 .its coordinates with respect to the z1 and 2 2 axes are denoted by (u. 2 3 ) E B3 : 2 : xi 22 = l}. the map x : ( S 2 \ { N } . Suppose that x : S2\{N} + B2 is the stereographic projection from the north pole of S 2 into the plane lR2.
interior of J ( I 3202 Let M 2 c lR3 be a surface containing x = 0. i) By the Meusnier theorem. we know that the Gaussian curvature of M 2 at x = 0 is K ( 0 ) = 25.(O. a is a normal section.e. &)I = 4. using the famous GaussBonnet formula. &) we immediately obtain Ic. the normal curvature of Q at x = 0 is ii) Since a is obtained by intersecting M with 21x3 plane. By the relation among k ( O ) . kn(O. or more directly. we immediately have K ( x ) = 2 T X ( D ) = 2T.(O.(O. K ( 0 ) = det ( y4 ) = 25. &)and the geodesic curvature k. 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. .. 1). O . You may assume that the normal to the surface at x = 0 is ( O . Then. then its curvature at z = 0 is L ( 0 ) = lk. (Indiana) Solution. i. iii) Because the eigenvalues of the Weingarten map L are f 5 . &)= 0.172 ii) Denote by D the simply connected domain encircled by the simple closed geodesic a. 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. Assume that and ax 1 tangent to M a t z = 0.
y. we sketch the surface near x = 0 as follows. Firstly. (Indiana) Solution. z ) E E3 : z = 6 .6 3203 Let M = {(x.0. Therefore. z ) E E3 : x2 y2 _< 4 .(O. we have that + l U p d w = 0. 1) and k. Choose the unit outward pointing normal as the orientation of M . we know that x = 0 is a hyperbolic point of M 2 .(zz y2)} (a paraboloid of revolution). y. and w = yz2dx zzdy x2y2dz.O.173 iv) From iii). s / Fig. B . 1) as its normal vector. we have k [ 6 . z ) E M : z > 21. Orient M and evaluate + + + LzdxAdy+dw. z = 2) and take (0. Noting that the normal to the surface at z = 0 is (O. y.3. Hence. &)= 4 > 0. Then. Let P = {(z.r2)rdr = 16n. by the Stokes’ formula. D = {(x.(z2 y2)]dx A dy = + 12r d 1 2 ( 6 .
noticing that dw = z 2 d y A d x 2yzd. Let ( z 1 . 3205 Let D be the disk { ( x . + + we have ~dw=~4dyAdx+2dxAdy=2 J.0) = r2[1 .2r2 + r4 sin201'. Here we assume the Jordan curve theorem. 3204 Let S be a surface diffeomorphic to the ordinary 2sphere.174 Secondly. Applying the global GaussBonnet formula to the domain D c S and noting that y1 and y2 are geodesics of S. J. (Indiana) Solution.y ) E lR2 : x2 y2 < and let 0 = (0. x 2 )= ( r . then y1 and 7 encircle a domain D such 2 that the Euler characteristic x ( D ) = 0 and d D = 71 uyz. which contradicts the assumption of positive curvature. Suppose there is a C" metric of positive curvature on S for which there exist two simple closed geodesics y1 and 7 2 . d y A d x = . r zdz A dy + dw = 8T. .& . Hence. Show that y1 and yz must intersect. we have K d c = 27rx(D) = 0.O) E D. 0 ) be the usual polar coordinates and define a metric on D\{O} by + i} where h(r.z A d x + z d x A d y +zdz A dy + 2xy2dx A d z 2x2ydy A d z . If y1 and yz do not intersect.
Hint.2r2 + = [cos28 + (1 .2r2 + 4r4 sin219)~ 8)dz2 sin2 on D . we know that both l1 and 12 are geodesics in D\{O}. (b) Show that line segments in D which pass through the origin (suitably parameterized) are geodesics. Hence. .+ 2lii2dxdy + li22. we can express 2\{O} as the union of Il = { ( r . y = r sin 8. the geodesic curvature k .8)de2 + 4r4 sin2 8)2]dxdy +[sin2 8 + cos2 8( 1 .8)d82. y) as the coordinates. Using ds2 = dr2 + h(r. { Then = r.d~~. Z22 = 1 + o ( r ) . Therefore.8 = B0 + T } .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 . on D\{O}. we know that kgIo = 0. = giidx2 When r 0. (b) For any line segment I in D . we see that &I = 1 + o ( r ) . 8 ) : 0 < T < :.2r2 + 4r4 sin2 8)2]dy2 . If we use the extended metric. as a whole line segment.(1 . Because 11 and 12 are geodesics by either metric. So. t 00. e ) : 0 < r < :. of I must be a continuous function of the coordinates.T sin adz). (Indiana) Solution.8 = O 0 } and l2 = { ( r .r sin 8d8 dy = sin 8dr T cos 8d8 + { dr = cos dx sin 8dy d8 = (cos 8dy . we can choose (x. From { follows dx = cos Bdr . Use a smooth coordinate system. $& = o( 1). + Hence the metric of D\{O} may be rewritten as ds2 = dr2 + h(r. and %o( l). 1 is a geodesic. (a) Set x = rCOS8. g12 = o(T).
Ler R be the infimum of their radii. i) The surface M and the sphere S 2 . &. for minimal surfaces. But it contradicts the fact that . l ) be a tangent vector a t the point ( 0 . l ) E H 2 . Then. one concludes that the Gauss curvature of M a t p . the curvature of M n T is not less than that of S2na = S1. a t p . K ( p ) 2 &. Then. Using the result of i). 3207 Let H 2 = {(z. is the usual inner product on B2. the Christoffel symbols).b( t) )be the parallel transport of V(0) along the curve z = t . 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. 6.e. y) (a) Compute the components of the LeviCivita connection (i. (Indiana) Solution. p is an elliptic point of M . Let a be a normal plane of M and S 2 at p . Then there is a family of spheres that contain the surface M inside and have a fixed center. Show that the Gauss curvature of M at p is 2 ii) Show that there is no closed minimal surface in lR3. (b) Let V(0) = ( 0 .y) E lR2. there must be at least one common point of M and the sphere with radius R and centered a t the fixed point. Let V(t) = (a(t). one can easily show that. Suppose M is inside a ball of radius R and suppose a point p E M is on the boundary sphere.. the boundary of the ball. using the local canonical forms of the normal section curves M n 7~ and S 2 n7~ a t p . have the same tangent plane a t p .y > 0) be the upper halfplane in lR2 and let H 2 have the Riemannian metric g such that where (2. .176 3206 i) Let M be a surface (without boundary) in lR3.
by taking the outer unit normal vector field as the orientation of S. k2 are the two principal curvatures of S a t P. Show that V ( t )makes an angle t with the direction of the yaxis in the clockwise direction. Hint. writing a ( t ) = cos@(t). (a) Define the Gauss curvature K of S. the normal section is a circle x 2 + y2 = 4. the corresponding principal curvature Icl = f. (b) Use the definition from part (a) to compute K at ( 0 .177 y = 1. l ) .2. = ( t . 3208 Consider the hyperboloid S of one sheet x 2 + y2 .b ( t ) = sin@(t). ( t ) ) = (wl(t). (a) From the hypothesis we have E = G = I F = 0 and hence ?I2 ' (b) For convenience. Noticing the above expression of is equivalent to rik. in the direction of the circle of latitude. (Indiana) Solution. v 2 ( t ) ) .u 2 ( t ) ) y) and the parallel transport vector field V ( t ) = ( a ( t ) . (a) The Gauss curvature K of S at P is defined by K = k l k 2 . Write a ( t ) = cosO(t).along the curve.vyO)) = ( 0 .1) = ( d ( t ) . 0 ) . (b) At P = (0. we see that the problem = a(t) 9= b(t). Then b V ( t )must be the solution to the following initial value problem (vl(O). denote the curve ( 2 .we have immediately 0 = t.z 2 = 4. (Indiana) Solution. b ( t ) = sinO(t) where O ( t ) is the angle V ( t ) makes with the zaxis. where k l .0). 2 . . Thus. Therefore.
thus the corresponding principal curvature k2 = Therefore.2. Thus.z 2 = 4.& k i ) .(z) dU1 au2 a a = (1 . := P . ( M ) Pc(z) dU' a = (1 . express the principal curvatures of M . By the GaussBonnet formula.. If. then by the Rodriques equation we have on P . In this case. the normal section is a component of the hyperbola y2 . 3210 Let M be a compact surface embedded in R3.178 whereas in the direction of the meridian. in f + (Indiana) Solution.(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 . the Gauss curvature K = k l k 2 = a. and s1 is the spherical cap encircled by C. 1 i. Show that the mapping P. ax aui i = 1.& k 1 ) ( 1  Ek2) ax dul x # 0 au2 ax . (Indiana) Solution. x Pc(z) P. : M R3defined by P. ( M ) at P. locally. we take the lines of curvature as the parametric lines of M .(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. the total geodesic curvature is given by where C is the given circle. with smooth unit normal vector field v.
C . C . .a ) . we can express the metric of the Riemannian surface k2 ds2 = ex(y)(dx2 dy2).)lR2..case + 2 O=n. D are A . in lR2. B ( a .1.a . use the GaussBonnet Theorem to deduce that in fact. immerses M into R3. respectively. (Indiana) Solution. in P . Take a positive number a and consider a square domain D in IR2. by the Liouville formula. whose vertices are A ( . Applying the GaussBonnet . . do 2dE 1 alnE aY coso+ 1 dlnG sin ' .a ) . ( M ) . we can see that the geodesic curvatures of the segments of coordinate curves i g and EE are +  5. w ) := ex(Y)(v. we see that the u1 and u2 parametric lines are the lines of curvature l . w E T(.sin 0 .179 provided that I E ~ is the reciprocal of neither k1 or k2 at x E M . C ( a . which shows that the mapping P.1 and respectively. Furthermore.a . which is obviously by conformal to that of the Euclidean plane B 2 .y) in lR2.a). Using the Descartes coordinates (x. B . and the corresponding principal curvatures are ~ k . and consider R the Riemannian surface obtained by equipping lR2 with a metric g of the form g(v. the corresponding part of 2) is and the corresponding points of A . Then. D ( . D . B . Suppose that. Assuming the Gauss curvature K of this metric is everywhere nonnegative. the surface is flat.6 2 d z ax O=O. a ) respectively.. again by the Rodriques formula.y=a 1 dlnE 2~ aY 1 dlnG a ax whereas both 52are geodesics of k2.y= a k.lco = (do ds and .w ) for v . 3211 Let X : lR + l be a smooth function with compact support.
y.Therefore. dy2  d2X namely ’ dY is a decreasing function. we have < 0 . from the above integral equality we can obtain Because the number a is arbitrarily chosen. i) Show that a(.180 formula to 5 c IR2. Thus. 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 .y . Assume that M 2 n ((2. R is flat.. .e. z ) : z 2 0). by the hypothesis K 2 0. letting a + 0. 2 3212 Let M 2 c R3be a smooth compact surface such that M 2 c ((2.) is an asymptotic curve on M 2 . y=a y=a which combining the above equality shows that % = 0 in B1. ii) Show that a’(s) is always a principal direction. K 0.parameterized by arclength. z ) : z = 0) is a smooth curve a ( s ) . i. it leads to Besides.
a ' ( s ) . Pi draw closed geodesics c. the unit normal vector field n ( s ) of M 2 is constant along a ( s ) . 3213 (a) Construct an example of a noncompact C surface in " sequence of closed geodesics {c. ii) Noticing the above fact. (b) Show that this is not possible if the surface is compact. from which the Rodriques equation says that a ' ( s ) is always a principal direction with the principal curvature 0. along a ( s ) .7 (b) Suppose S is a compact surface. a ( s ) is an asymptotic curve on M 2 . the plane {(z. = .181 iii) Let's now drop the assumption that M 2 n ((2.O .7. Let Pi denote the points where C has horizontal tangents.3.z) : z = 0) is a tangent plane of the surface M 2 . y.z) : z = 0) consists of elliptic or parabolic points of M 2 . which means that the tangent vector of a ( s ) for every s is an asymptotic direction.i. i) The hypotheses imply that along the curve a ( s ) . iii) If we drop the assumption that M 2 n { ( z ly l z ) : t = 0) is a curve. Hence.} such that length (ri)+ 0. rotating around the zaxis which is the asymptote of the curve. z ) : t = 0) is a curve. = 0. If there exist closed geodesics c. and length (ci) 0.illustrated in Fig. + X Fig. Therefore.i such . R3 with a (Indiana) Solution. a t which the plane {(z. y. Then. 3. then the set M 2 n {(z.y. t) : z = 0) be? (Indiana) Solution. (a) Consider the following surface S of revolution generated by a C vi" brating curve C. we see that along a ( s ) .y.y. What kind of a set could M 2 n {(z. z ) : z = 0) is tangent to M 2 . Thus. on S..
K d a = 2 r x ( M Z ) = 4 ~1( 9 ) _< 0.___ a G~ 8 rgl+ 8s 21ae 2~ as 2 ~ a . when i is sufficiently large. The direct computation gives & 6. z ( s ) ) a unit speed curve in B3 with T ( S ) > 0. Because S is compact. Consider be the surface of revolution (s.a . T ( S ) sin 8.. 0 ) + ( ~ ( s COSO. Thus.ae 89 v8.. I = d s 2 + r2(s)d02 Eds2 + Gd02. By the GaussBonnet formula. we may suppose R. 3215 Let M 2 c R3 be an embedded compact surface of genus 2 1. is simply connected. 3214 Let ( r ( s ) .j .a where as = l a a GI G2 r2. denotes the Christoffel symbol ( i .  . z ( s ) ) . setting i + 00 in the above equality. we have J I.E~ a +.2). then by the GaussBonnet formula we have where a. we come t o a contradictory result 0 = 2 ~ .denotes the domain encircled by u. (Indiana) Solution.r ( s ) ae’ e r..182 that length ( u i ) + 0. Ic = 1. On this surface ) compute the covariant derivatives V+& and V A & in terms of and 86 (Indiana) Solution. O . aas 22ae 2 a ~ e a y 2 . Show that the Gauss curvature of M must vanish somewhere on M .8~ 2 ~ a = . = Denoting (s..T ( S ) T ’ ( S ) .we have  a vs.e) = (ul. u 2 ) .%+r2=+.
we would have a contradiction. compute its Gauss curvature function K. G = (u + TCOSU)~. by the connectedness. ( u + TCOSU). we show that there exists another point Q with K(Q) < 0. 0 < u. y.u ) ~ + z 2 = r 2 around the zaxis: T = {(z. Secondly. Z) : (z2 y2 + + z 2 + u2  T ~ . 12T 12' dv cos udu = 0. the continuous function K must vanish somewhere on M 2 . by the continuity. ( a + r c o s u ) s i n w . N = COSU. (Indiana) Solution.F 2 cos u T(U+ TCOSU)' Noting that the area element on T is dA = J E G .w) = ( ( u + r c o s u ) c o s w . M = 0. Therefore. we claim that because of the compactness of M 2 . there must be a point P with K ( P ) > 0. A straightforward computation gives the coefficients of its first and second fundamental forms E = r 2 . .M~ EG .183 Firstly. and verify that KdA = 0 by explicit calculation.F 2 d u d v we have immediately KdA = = T(U + TCOS u)dudv. r s i n u ) . Thus obtained torusofrevolution T can be parameterized by sT X(u. L = T.4u2(z2+ y2) = 0). ) ~ Parameterize this torus. there exists a domain U of P such that KIu > 0. Finally. hence. 3210 Consider the torusofrevolution T obtained by rotating the circle ( z . F = 0. Otherwise.w < 27r. its Gauss curvature function K = L N .
for any Riemannian metric on S 2 . Consider the cone that is tangent t o the unit sphere S 2 along the curve z. the oriented angle formed by the tangent vector with the parallel translation vector X ( t ) of XO is given by 27r . 7 0 5 t < 27r T(L fils2.. By the GaussBonnet formula. Let X o = translation of Xo along z ( t ) . then 47r 5 Is IKldv 5 Area ( S 2 ) .8 = 27r  (i. Compute the parallel 21 a (Indiana) Solution. 5 1. we obtain the result that. 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. where K is the (Indiana) Solution. we have J. for a displacement t of a moving point p along z starting from 0. and characterize (with proof): i) All geodesics on the unit sphere S 2 c IR3. 2 3219 Define “geodesics”. 1 cos(t). Thus. “1 & . if IKI Kdv = 2 x x ( S 2 ) = 47r. . q) 3218 Show that for any Riemannian metric on S2 with II K Gauss curvature. 5 1.184 3217 Let z(t)= (. sin(t). Because the parallel translation in the plane coincides with the normal Euclidean one. the area of S 2 is not less than 47r. 2 2 E be a curve on S 2 c Ill3. (corresponding to the central angle 8 = $t in the domain D).
draw a picture) to show that a connected surface can have two points which are not jointed by any geodesic. every great circle is a geodesic. the principal normal vector N is parallel to n. and kg is the geodesic curvature of a. then if we develop the cylinder to a plane. Then. k. ii) Since geodesics are intrinsic objects. What is the usual topological hypothesis that prevents this problem? (Indiana) . we have d ( S ) = Ic(s)N(s) where u1.k i = i.u2 are local coordinates on M . Therefore. or a circle of latitude. must be a great circle. E 0 or  d2ui ds2 .. n is the unit normal vector field of M along a . because along a . every geodesic on the cylinder must be a helix.185 ii) All geodesics on the surface ((2.j ddsk = o ds j.g. a is a geodesic of M if and only if along a . On the contrary.2. or a straight generating line. i) On the unit sphere. (Indiana) Solution. Let a ( s ) be a curve on a surface M parameterized by arclength. every geodesic must become a straight line. y. du u +Cr*. 3220 Give an example (e. z ) E R3 : z2 + y2 E 1). owing t o the uniqueness of the initial value problem every geodesic on M . Along a . that starts from a given point and is tangent t o a given direction.
i) From x “(sinu.2 F M + G L = 0.(0. and prove that the catenoid.O). w. Let K be a plane and p .8) = (z. 3222 The “Clifford” torus in S3 can be parameterized using charts of the form 1 X ( u . obtained by rotating the graph of y = cosh(z) around the zaxis in R3.sin w). Then the surface K\{T} is a case in point. Calculation is not necessary here. A minimal surface is a surface with mean curvature H= 1E N .186 Solution. iii) Deduce from the result of ii) that the Euler characteristic of a torus is zero. 3221 Define “minimal surface in R3”. i) Compute the metric [gjj] on the Clifford torus for a coordinate chart of the indicated type. cosh(t) cos(8). w) = (cos Jz u .is minimal. A 1 X “a . cos where u and v are constrained to lie within intervals length < 2 ~ .sinv. ii) Figure out the Clifford torus’ Gauss curvature function. sin u. cosh(t) sin(8)) is a minimal surface in R3. “Completeness” is the usual topological hypothesis that prevents this problem. (Indiana) Solution. Let T be an interior point of the line segment p y . Hint. y two points in K.cosu. (Indiana) Solution. 2 EGF2 The straightforward calculation shows that the catenoid X ( Z .O.cosv) 1 .0.
i) Show that the principal curvatures of the graph z = f(Z1. f ( 0 . O ) ) E R3 are the same as the eigenvalues of the Hessian matrix [ a 2 f / a z .2) at ( O . Look a t the “lowest” point on the surface. a 2 jat ( O . O ) ) E B3 are the roots of the following equation . we have Therefore. Hint. the principal curvatures of the graph z = f(Z1. O . a minimum R or maximum) at the origin 2 1 = 2 2 = 0..g. 2 1 g12 = 921 = 0. . by straightforward calculation. iii) Using the GaussBonnet formula. O ) . ii) The Gauss equation implies that the Clifford torus’ Gauss curvature function K 0. Then.22) 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. i) The hypothesis means that $ and vanish at (0.187 we easily have 911 = 922 = . O . (Indiana) Solution.O). ] ii) Show that lR3 contains no compact embedded surface with strictly negative Gauss curvature at all its points. f ( O .
zo) be the “lowest” point on the surface. f ( 2 . Since the inclusion map M + lR3 is an embedding.O. 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.Y>. = z ( u ( 2 .) Since the surface is above the plane T : z = zo and p is the common point of R and the surface. the embedding can be expressed by ( 2 . we have F .f ( z . Y) Y)). ) Noticing that we know from the implicit function theorem that there exists a neighborhood V. v ) .~ ( o = upn M . ~ + ( z ( u .v ) for V and (z. for each coordinate neighborhood V c M of p E M . yo. we can find a neighborhood Up c U of p in lR3.188 ii) Let p ( z 0 . such that V.. y . y ) . such that on V. z ( ~ . { Y= 2 = 2(u. it is easy to conclude that any normal curvature of the surface at p is greater than or equal t o zero. v ) ) . 3224 Let M be a 2dimensional manifold smoothly embedded in lR3 with unit normal n. Y. if we take (O. where f(z. Using the local coordinates ( u . and in terms of the local coordinates (z. c V of p .v) Y(U. setting F : Up t R3 by F ( z . z ) = z . has smooth inverse { 2) =4ZlY) = v(2. By observing the normal section at p . there exists a neighborhood U of p in R3. y ( u . Thus the Gauss curvature of the surface a t p is not less than zero. ~ ) . such that V = U M . (Indiana) Solution. 1) as the unit normal of the surface at p. we can express this embedding as ( u .~ ( 2 . Y)). z ) for U . (The existence of such a point follows from the compactness of the surface.Y) (z. = Up n M . Therefore.y). y. then T is the tangent plane of the surface a t p . Then. )  .
d) Show that J. there exist a coordinate neighborhood C c M of p and a coordinate map h : C + U c B 2 . b) and c).P(. x Xu and K > 0. y ) . we can take Up = R3.+ S2 is a local diffeomorphism. You may use without proof the fact that M lies on one side of each of its tangent planes. 211. c) Assuming that 71 is onetoone. We first prove that q : M .189 b) If M is the graph of a smooth function f : R2 + lR. . define q ( p ) as the end point of the unit exterior normal n(p) after parallel translating it to the origin of E3.>> # 0. y. = K X .> rank = (4% . [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. Assume that n is the exterior normal and that the Gauss curvature K of M is everywhere positive. v) E C C M . prove that is a diffeomorphism onto S2.] a) Define the Gauss (or sphere) map 7) :M $ s2= {(z. a) For each p E M . . Since nu x n. y .v) is the unit normal at X ( u . 3225 Let M be a 2dimensional manifold smoothly embedded in R3 with unit normal n. 7) b) Prove that 7) is onetoone. z ) : x 2 + y2 + z 2 = 1).v). K ( P ) d P = 4?T. where n(u. .) (Indiana) Solution. Assume that M is the boundary of a bounded convex open set.and F : R3+ R defined by F ( z . = n ( u .)..which is a diffeomorphism. such that on U the Gauss map has the following expression q 0 h'(u.then for each p E M .. z ) = z . (You may assume b) and c) if you wish. Xu x Xu Jacobi matrix the rank of the ( ffu ffv 3 /: ) = 2. For each p E M . r(.f ( z ..
: Cp 7 ( C p ) . : CQ + ~ ( C Qare diffeomorphisrns. Suppose that there are two distinct points P. Then the GaussBonnet formula gives 1. da = Hence Kda = 4n. Now we prove that q : M t S2 is globally onetoone by leading to a contradiction. Thus q ( M ) is an open.Q respectively. I . In the end. we see that the globally onetoone.190 Thus.hence 7 o h . Next. we show that q : M S 2 is surjective.l is a local diffeomorphism. fi have local expressions Since Kda > 0.respectively. From the above argument we know that there are neighborhoods C p . on M . q = (7o h') o h is a diffeomorphism. in the neighborhood of h ( p ) . 7 is also a closed map. q l c . such that 7Ic. in M we can choose C ~ . Besides. The connectedness of S 2 implies that q ( M ) = S2. because M is compact and S2 is a Hausdorff space. On the other hand. connected and oriented. Therefore. Now take the inverse images of ~ ( C P ) ( C Q ) under nq q l ~ and q l cQ . closed and nonempty set of S 2 . CQ of P. 5.1 ( 7 ( C P )n 7(CQ)). 7 is a local diffeomorphism. where da. q is an open map. surjective local diffeomorphism is naturally a global diffeomorphism. +  u = ( v I c p ) . we arrive at a contradiction. Q E M such that q ( P ) = q ( Q ) . noticing that differentiability is a local property. which implies q(M\U) = S2. Since 7 : M + S 2 is a local diffeomorphism. Thus. set . u n v = 0 and q ( U ) = q(V). = (vIc*)'(7(cP) v " rl(CQ)) Thus. (by noting K > 0). . 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 ) . Namely. C Q ) so small that C p n CQ = 0. it is easy to show that M is compact. Because P # Q.
last local expression of the covariant derivative of M the involves the tangent vector field and the Christoffel symbols of M . Further. b) The principal curvatures axe not intrinsic. Firstly. discuss why the Gauss curvature of M is intrinsic. as the above.191 Remark. we know that the covering map q must be a homeomorphism and hence a global diffeomorphism. For example. a) Explain what is meant by intrinsic and extrinsic quantities on M . Aw J do so . Using the theory of covering map. they are extrinsic.is intrinsic. they are called “extrinsic quantities”. then the local diffeomorphism q : M + S2 is also a covering map. b) Are the principal curvatures intrinsic? c) Discuss why the covariant derivative on M . because S 2 is simply connected. Let Aw denote the angle variance caused by a tangent vector after parallel translating it around C once. d) For each p E M . In fact. Since M is compact and S 2 is connected. consider a plane and a cylinder. Otherwise. it is intrinsic. Therefore. 3220 Let M be a 2dimensional manifold smoothly embedded in R3with unit normal n. show that q : M t S2 is a local diffeomorphism. we can show that the Gauss curvature of M a t p can be expressed by K ( p ) = lim D+P . and M c R3 is connected and hence path connected. Using the Hopf’s rotation index theorem and the GaussBonnet formula. this problem is the famous Hadamard theorem. c) Although the covariant derivative on M is defined by using the covariant derivative on R3. d) Assuming c). we can simplify its proof as follows. defined using the covariant derivative on R3. (Indiana) Solution. In other words. a) The terminology “intrinsic quantities” means those geometrical quantities that are definable only by the first fundamental form of M and its derivatives. let C be a simple closed curve encircling a simply connected domain D where p lies. intrinsic quantities are those that are invariant under isometric correspondence. but extrinsic ones are not.
The surface M 2 of revolution generated by the curve y is homeomorphic t o a section of a cylindrical surface.8 Solution. by the famous GaussBonnet formula. we have immediately K d A = 0. too.192 Since parallel translation is intrinsic. we get a surface of revolution M 2 c R3. 1)2+ N such that for each fixed wo E (0. b) Show it for arbitrary n. Compute JM K d A . (Indiana) . the boundary of M 2 consists of two circles which are just geodesics of M 2 . Therefore.) (Indiana) z I I Fig.3. wo) is a geodesic in N. l). a) When n = 3. 3227 By revolving the curve y sketched below around the zaxis. then the Gauss curvature is intrinsic. where K is the Gauss curvature and dA the area form. because along these circles the normal vector of M 2 is parallel to the principal normal vector of the two circles respectively. Thus the Euler characteristic number x ( M 2 )= 0. the uparameter curve X ( u . Besides. on M . show that the Gauss curvature of a ruled surface in Rfl is nowhere positive. 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. (Make sure to justify your answer.
C2 d 2X L = (n. we obtain the desired result.(a’(v).Z’(w))2 = la’(w)x I’(w)I2 2 0.Z’(v)) + K = U2II’(V)12)dV2.F2  b) Similar to the above. 1Z(v)1 SE 1. Thus.l(w)) Then. a) Since every geodesic in N = R3 is a straight line. by a routine work. we can easily deduce that the first coefficient of the second fundamental form of i)l(w). we see F 0. that the first fundamental form of C2 is ds2 = d U 2 + (1+ 2u(a’(w). then C2 can be characterized locally by X ( u . which shows that the Gauss curvature of the surface K= LN . ) dU2 f0.iw2 <o. we have Noting that 1l’(v)12 .w) = a(w)+(uwhere Z(v) is the direction of the geodesic corresponding to w E ( 0 .193 Solution. l ) is the arclength of the curve a ( v ) . l ) . Since the Gauss equation says * dz’ it suffices to show that 2 0. and (a’(u). for convenience.F2 EG. EG. suppose C2 can be characterized locally by where. we may assume that v E ( 0 . through computation. . However.
then 2 = aX by. Using this or any other valid method.. ii) For any vector fields X .2. Z . Thus.Y ) Z .( L X .Y. where L is the Weingarten map. fY. iii) Recall that if M" c En+' is a hypersurface.Y]Z and the Riemannian curvature tensor field by R ( X .. W.Y ) Z = ( L Y . E T'(M"). M. Y . Then for any C function f on M " . Y . ? = cX d Y . z.W on M " .y.194 SECTION 3 DIFFERENTIAL GEOMETRY OF MANIFOLD 3301 Let M" be a Riemannian manifold and R a 2plane in T. i) Let { X .Y)Z = V x V y Z . compute all sectional curvatures of the sphere {z E B4: 121 = 3). which proves that K ( R )depends only on R . Y . . where R( ) is the Riemannian curvature tensor field of M " . fW) = f R ( X .V[X.. one can show that V p E M " . Explain why it is a tensor. Y ) . W ) = R ( X . Y } be an orthonormal basis of R .Z ) L X . Use this basis to define the sectional curvature K ( R ) and show that it depends only on R and not the particular basis chosen. W)I.V y V x Z . R ( X .2. Noticing that field. and R( ) is a 4th order covariant tensor  z. Therefore. X . If ( 2 . ii) Define the Riemann curvature tensor in terms of covariant differentiation. 2. we can easily have R ( X . is dependent only on X. Z . W ) W = R ( X . Y ) .R ( X . (Indiana) Solution.W ) . then R ( X . Y . f Z .X . define the Riemannian curvature operator by R ( X . y. we can conclude from straighforward computation that ( 1) + + is an orthogonal matrix. ) = R ( X . in terms of local coordinate frame field. Z. i) The sectional curvature is defined by K ( R ) = .Z ) L Y . R(fX.W ) = ( R ( X .Y . Y ) is another orthonormal basis of R .Y. Y . Y ) = R ( X .respectively. using the properties of covariant differentiation " and of the inner product.W ) . Y .X .
3302 Let C = {z E lR3 : 0 5 3ci 5 1) be the unit cube in B3. one can easily show that V e i z= e i . " iii) For the sphere M n = { z E lR4. if { X . X ) ( L X .V x g = . Y } is any orthonormal basis of an arbitrary 2plane T in T p ( M n ) . we can regard the position vector field 2 as the normal vector field of M n .( R ( X . Suppose F : lR3 + lR4 is a 11 C immersion in some neighborhood of C.1zl= 3 } . Y ). Justify the following formula: . Therefore. Y ) = ( L X . Y )= 1/9. the inner product and covariant differentiation are all C".Then by the original definition of : covariant differentiation. Besides.    K(7r) = .195 R( ) I p is a welldefined 4th order multilinear function. the Weingarten map is L : X F+ L ( X ) = . { e i } be a local orthonormal Let frame field about 3 (as a point) of M n .( L Y . Hence. R( ) thus defined is a C tensor field. where V denotes the covariant differentiation in lR4.X / 3 . Y ) X . X ) ( L Y . The image " F(C) is then a compact Riemannian submanifold of lR4 with boundary and therefore has a volume. Hence.
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
y) : x. O . i) Identify lR6 with {(x. f3) = (11x112. (Indiana) :X. a axl a ax4 i) Is there a 2dimensional submanifold M 2 of lR4 such that for each p E M2. Y ) = 01. Y) = ( f l . Y ) i) Show that M explain how M can ii) Show that M B3. It is easy to verify that the Jacobi matrix of the C map F " .asJ does not satisfy the Frobenius condition. ( X .y E lR3} and define a map F : R6+ IR3 by F ( x .llY1I2.21 1 is a regular submanifold of R4\{(0.X2] = .IIXII = 1. ii) Yes.O. for example. (x.we see that is also a threedimensional submanifold of lR4. llYll = 1. a i) No.(23 2 4 ) . & + 3321 Let M = { ( X .f 2 . X2 on lR4 be defined by x1=  a + 21. we can set f = 21x2 .Y)). there is not.O . O.Y E Solution. a ax2 ax3 x2=+22. 0)) with dimension 3. Noting that E4\{(0.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. 0)) is an open submanifold of B 4 . Xl(P). is orientable. 3320 Let vector fields XI. The reason is that the bracket [Xl. is a smooth compact embedded submanifold of IR6 and be identified with the unit tangent bundle of 5 ' .X2(P) E T.
x I. F ( M ) is also closed in N . Show that if M is complete. .. too. Besides. Therefore. ) au. ( S 2 ) .. 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.. and the image of its end point under F must be y.e. for every (z.dol x 4) form a covering of coherently oriented . ii) Because S 2 is orientable. take S > 0 so small that exp. so is N and F is a covering map. if we regard z E S2 and y E T. .. Thus.) n (Up x I p . 4 x 4. for every (z. Thus. where I.y) = = 1 0. Next. F ( M ) is open in N . Because F is a local isometry.& x $ is . S2 has a covering { ( U .ui) U. (Indiana) Solution. For every x’ E N . : B’(S) + Bi is a diffeomorphism. the Hopffinow theorem means that N is complete. M is orientable. y) E M .212 has constant rank three when fl = 11z1(2 1. i.. a coordinate neighborhood of (2. If (U. #Jp x $ p ) # 0.. then we can identify M with the unit tangent bundle of S 2 . ( U . then there is a point z E M such that there exists a geodesic in N connecting F ( z ) and y. z has the local coordinates (ui. Besides.(S2) with E 1)yI)= 1. Therefore. x I. since II(z. then the transition function . y) to 0. since F maps every geodesic of M into a geodesic of N . where B’(6) = {v E . f2 = ( ( ~ ( =~ 1 and f3 = (z. and y is uniquely determined by the oriented angle 0. we show that F is a covering map. M must be compact. F is surjective. By using identification of M with the unit tangent bundle of S 2 .y) M . = (0.) neighborhoods. Naturally. If y is a limit point of F ( M ) in N . .. 1. the connectedness of N implies that F ( M ) = N .O) = M is a closed embedded submanifold of B6.E . 3322 Let F : M 4 N be a local isometry between connected Riemannian manifolds M and N . y) E ) with E being a suitable positive real number and 4 is the map from (2. has the following J acobian + + which means that { ( V a x I. 0. Hence. # J ~ ) }of coherently oriented coordinate neighborhoods. a t z E from 4 to the unit tangent vector y E T . F’(l.y)II = (llzllz Ilyl12)i = fi which means M is bounded in IR6. E M .
2’. In fact. F o expZm = expzf odF is a diffeomorphism and hence exp. exp. expXm also a diffeomorphism.8. contradicts a # p. L ( y ) = L(y) < S means that z = T ( 0 ) E B r . yp connecting z with .. On the + other hand.) = y = F ( y p ) . we know that d F . Besides. and F : B p 6 B. Secondly. z a ) < a}. Thus the claim is proved. i) Show that each component of X = {z E M : F ( z ) = z } is an embedded totally geodesic submanifold of M .” n B f .“ + B. we have the following commutative diagram and expZe is surjective. are diffeomorphisms.e. Fl(B. Since F is a local isometry. such that y(0) = F ( z ) and y(1) = 2’.l ( B i ) = U B ? . + Let F : M + M be an isometry of a Riemannian manifold M . M is complete. So.)’ is a diffeomorphism. then there exist unique geodesics y. then 0 there is a unique geodesic y : [0. Hence F = exp.213 T . uBr c Fl(B.((M) lvl < 6) and Br = {y E M : d ( y . are isometric. we claim that if Q # .1] B.} c M and set Ba(S) = {v E T. is is an immersion. . i. Because both F : B. In particular. Since . o d F o (exp. we claim that F . f ( N ) : IvI < S} and Bi = {y’ E N : d ( y ’ .) c U B r . if z E F’(B. we have F(y.1] + M such that F ( T ( t ) ) = y ( t ) . z zp respectively. Hence F(y(1)) = z’ and y(1) = z for some a ... ya. . Fl(z’) is discrete. we claim : that Bi is an admissible neighborhood of z’ and F is a covering map. In other words. Thirdly.). which . we say that for any a . Then. Since F is a local isometry. Otherwise. then BF n Bf = 8. z = y a ( l ) = yp(1) = zcp. Therefore.) 0 a is obvious. z ’ ) < 5). Let y be the unique geodesic in Bi connecting F ( z ) and .. Firstly. Denote Fl(z’) = {z. Besides. The uniqueness implies ya = yp.. if there is z E B.yp are the lifts of y through z. V t . F : BO + Bi is a diffeomorphism. there exists a geodesic y : [0.
<~ o . From d F ( v ) = v follows d F ( T ( 0 ) )= j ( 0 ) . F ( z . d F ( j ( 0 ) ) = j(O). X n B c exp. i. we show that every geodesic y : ( a . b ) . and F be a reflection with respect to the plane z = 0. that F is an isometry implies F(y) = y. lies in X . V is a subspace of T.e. Use exponential coordinates. v E V which means that y E exp.(VnB(G)) is obviously a submanifold of M .) E IR3 : y > 0 .y. Then we claim that XnBa = exp.(VnB(6)). Hence.y. and x = {(2..(V n B(6)). i(so) = ? ( S O ) .z ) .e. because exp. Define V = {v E g T . ~ . F ( < ) = <. y ) < 6}.214 Hint. and F is an isometry.(V n B ( 6 ) )c X n Ba.On the other hand. y .(tv). Thus.(M).1] 4 M be the unique shortest geodesic y(t) = exp. ( M ) : IvI < 6) and B = {y E M : d ( z . For z E X c M . Then. suppose that v E V n B ( 6 ) and g y = exp.y < 0}. ii) Give an example in which the components of X have different dimensions.i. is naturally a geodesic of X. Let y : [0. Then M is a 2dimensional manifold. d F ( v ) = v... z = 0) u {(z. b ) + X parameterized by arclength is also a geodesic of M . : B(6) + B is a diffeomorphism. If this is proved. F ( y ) is also a shortest geodesic connecting F ( z ) = z and F ( y ) = y.1] 4 M be defined by y ( t ) = exp. d F ( ( ( s 0 ) )= ((so) and F is an isometry. Thus the uniqueness implies F ( y ) = y. we show that X has submanifold structures.e. Therefore. X is totally geodesic.z) ~ E 1123 : =o . we first assume that y = X n & and v E B(6) such that exp. v.v = y.e. in a neighborhood of S O . Therefore. exp.y. y is a geodesic of M . For any SO E ( a . then F ( < ) and are two geodesics of M which satisfy the same initial conditions. let <(s) be a geodesic of M such that <(so) = SO). . In order to prove the claim. namely. ( M ) : d F ( v ) = v}.(tv) connecting z and y. z ) = (z. Since z. In particular.z) E ~3 : > o . where a 6 is so small that exp. = 01 u {(z.. ii) Let < < < < M = {(qy. set B(6) = {v E T .y E X . Thus.. Let the geodesic y : [0. i. (Indiana) Solution. which means that y E X n Bg. i) First. In particular. = 0). Since F ( < ( s o ) ) = (‘(SO). F is an isometry.(V n B(6)). we can assert that X has submanifold structures. Next.i. y. Besides.z> E IR3 : 2 = 0. = y. Because so is arbitrarily chosen. F ( y ) = F(y(1)) = y(1) = y.
d)= 0 and S' is connected. w .d) = {df I f E Coo(S1. Le. d )= 0. Let 6 be the polar coordinate characterizing S1.215 3324 Compute the de Rham cohomology groups of the circle S'. (Indiana) Solution. set and E x. A'(S')). E ) := {y E En: Iy .d) = Cm(S1. without citing the de Rham Theorem. observe that Z'(S'. n is globally welldefined on S1. Do so directly. we d) have Zk(S1. and it is not exact. define a function & Because R(0) = R(27r).v E N .d) = Z'(S'. := {z I/ u. because 6 in usual sense is not a globally welldefined function on S'). Besides. Then is a nonvanishing vector field on S1.x + :2 . since there are no nonvanishing kforms on S'.d)/B'(S'. d ) = Zo(S'. IvI < E } . 3325 Let X denote a submanifold of Euclidean space E n . then H * ( S ' . B1(S1.d) = {f E Coo(S1.. since Bo(S1.Let d6 be its dual nonvanishing 1form on S'(Caution: Here d6 is only a formal symbol. For every w = g(O)d6 E C"(S'. B ( X . X .R1)}. I ' For H k ( S 1 . Hence.z1 < E for some z E X } .A'(S1)).d) = (Cd6 I C E Rl}Z R1. = 0 and hence H k ( S 1 . Therefore. denoting we see that w .. k> 1. H1(S1. d ) .R1)df = 0) E B .e. i.Cd6 = do.CdB is exact.
Show that the two are not generally equal.that when 2 is Killing. Y ) g ( X . j ( z ) = g (&.l=l where g . ) Can you give conditions which imply equality? (Indiana) Solution. let the 1parameter group of isornetries generated by a vector field 2 = C zi& i=l m be expressed by 2 = 2(z1. g ) .. If X has no boundary..y > > S(JhX.X c B ( X . i. can show that V E X is a proper subset of B ( X .X = + B ( X . (Indiana) Solution. E )for all E . either compact or complete. t ) such that m i.j=l m k.e. z m . . we have Lzg = 0 . In local coordinates (zl. ii) Show that the expression (**) above is equivalent to where V denotes the LeviCivita connection for ( M . E ) . Z ( g ( X . Setting y = z v immediately implies that V E Xc B ( X . Call a vector field 2 on M a killing vector field if 2 generates a 1parameter group of isometries of M . &). and 2 = 2 ( z . Let X be an open line segment in E 2 .g.9 ) . t ) satisfies . Here L z denotes the Lie derivative along z.216 Show that U.E l . i. e. the end points. then U.. t ):= 2 ( z . (Consider examples of 1dimensional submanifold in E 2 . Then considering the boundary of X . L Z Y ) = + (**) for all vector fields X and Y on M .  a .e. i) Show. z m ) of ( M . 3326 Consider a Riemannian manifold ( M . E ) .g ) .
i = 0. ( X . we easily obtain (w).o X t . we obtain which can be written as k gkjz. = Y.j Xj.i + S i k zkj $9 ( l . i) Show that the integral curves of X are curves of constant geodesic curvature. Conclude that ( X . we may assume that X = X I & and g 1 2 = (&. take a local coordinate neighborhood about p such that the coordinate curves are the integral curves of X and their orthogonal trajectories. From this follows what we desire in ii). i. j = 1 .i V 0. X ) . or equivalently g (V a Z . Y complete vector fields on M .217 Differentiating the obtained equality with respect to t and then setting t = 0. Y ) g ( X . ii) Assume that X and Y are linearly independent at all 2 E M 2and that their flows commute X t o Y. Because the flow X t of X for every t is an isometry. 2 . Y )are constant on M . i) For every p E M . &)= 0. + 3327 Let M 2 be a connected Riemannian manifold and X . Noting that the LeviCivita connection V satisfies Z ( g ( X . Assume that the flows Xtand Yt are isometries of M for all t. Y ) and ( Y . l) j e er3 Z ) =o. (Indiana) Solution. V Z Y ) = and the Lie derivative satisfies L z X = [Z. Namely. + . X I . the vector field X should satisfy the Killing equation Xi. or equivalently.Y)) g ( V z X .
i. i. . using the Liouville formula. all gij's are constants. Then.. ( X . Therefore. by the corresponding Killing equations. 2 . we obtain a922 agll.0 .8x1 a x 2 dx2 dXl = O . we immediately have x ax2 Therefore. Again. then the vector field X = from the corresponding killing equation. if a adopting the original notations. Now. ak = 0. that is equivalent to lay2 d ax1 d [ X . j . Ic. Because their flows are commutative X t o Y. make the Hence If we still adopt the original notations. (Y. about p .g 2 2 ( ~ ~ ) ( d ~ ~ ) ~ .e.e..X ) = gll = const. we can make a suitable coordinate transformation and then. a ( X .Y ) = 922 = const. X1 = X1(xl). i. Noting the expressions of X and Y . +. Y = Y2&. we can obtain = 0. we obtain d=Yol2. taking i = j = 1 and i = j = 2. along every integral curve of X .e. which means that the metric of M 2about p can be written as ds2 = g11(x2)(dx1)2 &. = Y. X = Y = G. we can compute the geodesic curvature of the xlcoordinate curve as follows: where 0 = 0. ax = 0.Y ) = g12 = const. take the integral curves of X and Y as the x1 and x2 coordinate curves. Then. i. o X t .. respectively. dX1 &. we obtain 911% following coordinate transformation = 0.. j = 2.Y 2 . Y ]= . =o 8x1 dxi . = const. we have locally X = X I & . Ic = 1 . ii) Analogously.218 Taking i = 1.
through direct claculation. .e j ) = (e: . Then we Let { e .. } around p and then define (Vf)(p) = C ( e i f ) ( e i ) and (Af>(p> . choose an orthonormal frame field { e l .( v e i e i ) f I = i i Verify first that V f and A f are well defined (i.C [ e i ( e i f ). e i } be another orthonormal frame field around p . j = 1.( v e . . j for suitable functions a!. . they do not depend on the choice of orthonormal frame) and then show that (Indiana) Solution.219 3328 Let M be a compact Riemannian manifold without boundary. . Then Green’s theorem implies . i .. . define O f E X ( M ) and Af E C “ ( M ) as follows: At any p E M ...f A f. . e . . Using the definition of the Laplacian here. we can easily obtain i i C [ e i ( e i f ). For any f E Cm(M).1n  e .e.e3) = S i j . we have div( f V f ) = llVf112 . e i ) f l = C [ e r ( e f f ) ( v e . . we have j Using these equalities and the properties of Riemannian connection. f ] . . e : ) . . i i Hence Of and Af are all well defined. Noting (ei . may suppose that e i = C a ! e r i = l .n.
is Sa. 2. we can easily show that &. 3. Denote a = xf + x i and .2 : = b .o is a 0dimensional manifold. b2 . If a = b = 0 . 1.220 3320 For what a . Assume that F is onto.4a 2 0 is the prerequisite condition. (Indiana) Solution. If a = 0. then when b2 . we can show that they are all 2dimensional submanifolds of R4.4a = 0 Sa. Let X be a smooth vector field on M .f3 = x i + x z .b a manifold? Explain. x : + 2 : = b or 2. Using the theorem of closed regular submanifolds proved by rank theorem. = 2.b = ( 2 E E4I b + 2. If a # 0. Thus S0. Then C Y ~ a and CY + p = b = means that a and . .f3 are two roots of the equation X 2 . then 21 = 2 2 = 2 3 = 2 4 = 0.bX + a = 0. b 2 0.4a > 0 and when b2 .b is a 1dimensional submanifold of R 4 . 3330 " " Let F : M + N be a C map between two C manifolds. then = { Z E B4 1 x1 = 2 2 = 0 . i . x 3 = 2 4 = 0). Therefore. Analogously. (i) Show by an example that d F ( X ) may not be a vector field on N . b # 0. = 2}. + 2. b # 0.
.N = {x : x E E}. i. Let (xl. Show that d F ( [ X l . y(t) has the following expression  21 1 t. X Z l )= [Yl. 3331 Let M" be a Riemannian manifold. .y) : z.221 (ii) Suppose Y = d F ( X ) is a smooth vector field on N . (Indiana) Solution. Hence. (iii) Using local coordinates. as a vector field. by direct computation. if y1 # yz. XJ 1 x 3 0 (.y E IR}. = (xz + y 2 ) & . Show that whenever f : M + lR is a smooth function. along a . Then. Y are related as X and Y in (ii) above. 1 and X z .). we have da ($) = X . we have d F ( X ( x . Show that F takes integral curves of X into integral curves of Y.. Y.yz)).y) = x. x2. If y ( t ) is the ith coordinate curve that passes p . d(F o a) (i) = ( d F o da) (g) = d F ( X ) = Y. d F ( X ) is not well defined in every point of N . z (iii) Suppose XI. (Indiana) Solution.x n ) be the local coordinates about p and set p : (xk.y l ) ) # d F ( X ( z .) be an arbitrary integral curve of X . Then. 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 . (i) Let M = {(x. . one will obtain the desired equality... which means that F takes integral curves of X into integral curves of Y .e. . by . # then. x and F : M + N be defined by F(z. Therefore.YZ].i 4. (ii) Let a(.
is (Indiana) Solution. Suppose that . . Hence .. and r : an RnX" be defined by (21.j=1 c = 1 5.z.. in which the dot product between A = [a.) in S"l to the symmetric matrix ~ ( x= I [ z i z j ] ..B ) := i=l j=1 a.u(x))= z. the pullback via g of the dotproduct metric defined above on RnX" the standard one on S"').zn) E S"l. 3332 The space ELnxn of n x n real matrices forms an n2 dimensional Euclidean space. ) Jz i) Show that u maps S"' into the sphere of radius centered a t the origin in R" n.j b i j . n i) Let x = (zl. It is easy to verify that the above expression of V f (p)is independent of the choice of local coordinates. . 5 i=1 1 " (u(x). ii) Prove that cr is a local isometry (i.e. .222 we have the expression where gij's are the components of the matrix [gijl' = [(&. . Then C z = 1.. Let Sn' be the unit sphere in IR".and define u : S n w 1 E n x n the map + as sending x = (21. 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.. 2. z i. a .) + [yij] = &[zizj]..j]and B = [ b i j ] is given by n n ( A .
a metric of the form [sij] = e 2 f [ S i j ] . .w).Then we have n n i=l i= 1 Therefore Hence we have (dn(v).e. where f : IR" IR is a smooth function. Let M be the Riemannian manifold obtained by equipping B" with a metric conformal t o its usual one. n} = & f ii) Show that when n = 2. 2 .223 belong to Tx(Sn').j=1 (dXj + VjXi)(WiXj +WjXi) = 3333 c n i=l viwi = (v. the sectional curvature of M along an plane a t p is given by e1. i) Show that for arbitrary indices i..dn(w))=  1 . k E { 1 . j . i. c n i. Let ei = denote the standard coordinate basis vector fields. and S i j is the Kronecker delta.e2 (Indiana) .
F = 0. i) Set From gij = e 2 f & it follows that ii) The sectional curvature of M along an e l . 3 4 0 s ii) Now suppose that [ X .Y]= 0. Y be complete vector fields on a manifold M and let X .x. Then . Observe that for a diffeomorphism F : M +M .Y = Y . o yt = yt o X .Y ]= 0. W E R. . i. Yt be the flows induced by them. if and only if F o yt = yt o F . Y ]= 0.t E IR implies [ X . e2 plane at p is just the Gauss curvature of M a t p . the complete vector field Y is Finvariant. is a diffeomorphism for each s. = yt o X. t E R.224 Solution. Thus 1 [ X . ii) Prove the converse to i). (Indiana) Solution. then Y is X.invariant. for all s. from the Gauss equation we obtain 3334 Let X . i) Show that X . Noting that E = G = e 2 f . for all s.Since X . F.. Y ]= L X Y = lim [Y . i) Suppose that X.e.. o r.
we obtain X . . ( X s * Y ) p f ( X o * Y ) p f= Y p f . 4(exp(u)) = exp(+*v). then 4(exptv) is also a 1parameter subgroup of G2 generated by a suitable w E LG2. ox Hence 3335 If 4 : G1 + G2 is a Lie group homomorhism. Y = Y . . Locally. = yt O X .z ) and (y'. Let ( z l . 4 can be expressed by and exp tv and exp tw can be denoted by respectively. .invariant. respectively. t E l is a 1parameter subgroup of G1 generated by R ' v E LG1. Assuming that wel " a =cvi.225 Hence for each p E M and any f E C"(p). yn) be local coordinates of G1 and G2 about " the identities el and e2. . we have for all s E R. = Since f is arbitrary. 4(exp tv) = exp tw. show that for all w E LG1. namely. we have (Indiana) Solution. t E R. Note that exptw. i=l axi we2 = a E w e aye ' n e=1 . . namely. X.for all s . Y is X. . Since 4 is a Lie group homomorphism. Therefore. .
a) Direct calculation gives [r. 2 4 ) E B4 are given by rx : U := B4\0 = = (21. i.23) v. Hence 4(exptw) = expt4.22. 2 2 . ZlYZ .. y2. = 2 (Z) (= I =we.: 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.e. b) Construct the following linear algebraic equations about y 1 .21. 23. 24. which are equivalent to .w. a 1 t=O aY" Furthermore. 3336 Consider the linearly independent vector fields r and v on whose values at x = ( 2 1 .226 we have t=O t=o Therefore. (22.24y3 + 23Y4 = 0. we have d $ ( v ) = w. d*w = w. y 3 and Y4 I Z l Y l + 22y2 22y1+ + 23Y3 + 24Y4 = 0.24). by left translation. Taking t = 1 completes the proof. Thus the Frobenius theorem guarantees that the rank2 distribution defined by r and v in U is completely integrable.v] = 0.. 2 3 .
. and the framefield is leftinvariant. ~ = 2(2: ] zi)v which does not belong to the distribution defined by CY and p. 2224 p = . (Indiana) Solution. 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. ~ 2 .)). By a long but straightforward calculation.n. 1 Veiej = . a) For arbitrary smooth vector fields X and Y on G.we have  n n .2223 Setting and respectively. b) Show that when G is a Lie group with a biinvariant metric ( . + 3337 Let G be a Riemannian manifold with a global framefield {ei}y=l.:). ( : . we have [ c Y . Thus the F’robenius theorem tells us that this distribution is not completely integrable.n. k 2.j = 1.without loss of generality. 2123.2223. (2122 222322124. 0. ).e. by the vector fields Veiej. 2 32...s+.2124 2124 2224 + 2123 . . ( 7 )) by setting. we may assume 2 1 # 0. Therefore. 4 ) E B 4 \ 0 = U. o). + (.i. a) Show that any connection on G is competely determined by its effect on the frame field.[ e i . . we obtain 2123 2223 +2224 .227 Because ( 2 1 . e j ] 2 for all i. i. we characterize the LeviCivita connection on (G. j = 1.
Thus.V e j e i= [ e i . .ej . where e is the unit element of G. the LeviCivita connection V satisfies V. motivated by the properties of connection. we obtain V e i e j= . 2 1 .[ e i . we can well define x'e. because the metric of G is biinvariant. e j ] . then V e i + e j ( e i e j ) = 0 implies that Veiej Veje.(g)ej + C zigveiej. e j ] .. i. = X . b) For every left invariant vector field X on G. Especially. + + + On the other hand. we know that V x X = 0 is valid everywhere. if X = ei e j . Noting that g ( t ) is a geodesic of G and Xg(tl = we have obviously 9 y.228 Then. = 0. let g ( t ) denote the unique 1parameter subgroup of G such that g(0) = e .j=l n Thus defined operator V certainly satisfies all properties of a linear connection on G. vx.x =  t=O Besides.
Part IV Real Analysis .
.
E LO. 1) I h P n ( z ) # 2 and P() n . * p s Let = (1) U{z u{ .1 kn1 lo*1 + 4 1uR) and B = C u [$+. p i ( z ) # 2 and pi(z) # 3). 3rn 2 n : p . p n ( z ) = 2 .+ kn1 n = l kj#2.+ . For any z E [0.1] 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. 1) I V n .3 101 + . p n ( z ) # 2 and p n ( z ) # 3) and B = {z E [ O .. kn1 n=lki#2.+ . V i < n . # 3) E [0.2 .pn(z)= 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 9.231 SECTION 1 MEASURABLITY AND MEASURE 4101 Let S c [0. 10" l o + 3 It follows that both A and B are measurable and therefore is S. 10n 10 +..3 10n1 kl + . 2 kl . (2) Vn.. V i < n .1) I 3 n . Since c y ) n . (Stanford) Solution. Prove that S is Lebesgue measurable and find its measure. p i ( z ) # 2 and pi(.k . A = {z E [O. ( z ) 58 and (3) z = n=l . . l ) Then 00 1 3n..) # 3).
each ( E S1/ choose a representative f E S1. + 2 4102 Define S1= IR/Z endowed with the natural Lebesgue measure. ( StanfoTd) Solution. a w €P. It follows that { E + [TI 1 T E [0. [y]) H [z + y ] and the inverse operation. 1) m(E+[r]) = C T m(E).where T E [0. too. Thus [r] = [ s ] . which then implies that T = s since 1 < r .e. If m ( E ) = 0 then m(S1) = 0. and let E c S1be the set of these points. 1) n&} is a partition of S1. [y] E E y mod& and therefore such that [z] [T] = [y] [ s ] . If E is measurable then + + + T €[O.1) n& is such that r G z .1) n& are such that ( E [TI) n ( E [ s ] ) # 0 there are [z].232 and we have 1 m ( S )= m(A) m ( B ) = . Consider on S1the equivalence relation: [z] [y] # z .  E = {F Show that E is not measurable..y E IR and [ ] denoting the class in S1. For any [z] E S1there is by the construction of E a unique element [y] E E such that z . s E [0. for z.l)W + m(S1) = T€ C [O.y E &. a contradiction. Let m denote the natural Lebesgue measure on S1such that m(S1) = 1.V[ E S'/ }. Otherwise m ( S 1 ) = contradiction.y m o d Z . i.y E &. + If T .s < 1. . I E <. It follows that z [z] = [y] by the construction of E . For . Then [z] = [y] [r]. [z] ++[.I. S1is an abelian group under the binary operation ([z]. So we conclude that S1= U ( E + [TI).1)W 00.
We have since the former is a ring containing D. D c a c P ( X ) such that R is closed under the following operations: (i) finite disjoint unions. let a be the smallest system. Denote ' by R the ring generated by D. (Iowa) Solution. A )< T) 2 .233 4103 Let X be a set and D c ' P ( X ) . y E B } > 0. Let U = {Z E I d ( z . We will show the equality. B c A. Also we have since by the equality (1) the former is a ring containing 2). 4104 Let p* be the Lebesgue outer measure on IR and A . R = R since for any A E R. Let T = d ( A . (ii) differences A\B.'D closed under finite intersection. x C R. Prove that a = R. B subsets of I such R that inf{ 1 . F'uthermore. Obviously.A = A n A.yI I z E A .B ) > 0. (Iowa) Solution. By equality (2). 2 Prove or disprove that p * ( AU B ) = p * ( A ) +p*(B).
Finite additivity means that whenever { B i } is a finite collection of mutually disjoint measurable sets. B C W2 C V} P * ( A )+ P * ( B ) . positive. finitely additive measure on B. X. Prove that p is countably additive if and only if it satisfies the following condition: If A.234 Then U and V axe disjoint open sets containing A and B respectively. 4105 a) Consider a measurable space (X. that is for each B E B. T00 b) Now that suppose X is a locally compact Hausdorff space. We have \ o o m / n . A U B C W C U U V ) E = + p ( W 2 ) I W open. p ) with a finite. Let {B. A 5 cU } ~1 WI +inf{p(W2) I W2 open. is a decreasing sequence of sets with empty intersection then lim p(An) = 0. Then n=l n An = 0. A C WI C U . then p(UBi) = C p ( B i ) . B C W2 2 V } i inf(p(W1) 1 open. and that p is a finite. finitely additive measure p. Let A. Suppose moreover that p is regular. = i=n+l U Bi. that B is the Bore1 aalgebra. (Iowa) Solution.} be countably many measurable sets which are 00 mutually disjoint.Y where the equality (1) follows from the following equality GR. a) The sufficiency. positive. inf(X+Y)=infX+infY.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. p ( B ) = sup{p(K) I K 5 B and K is compact}. Prove that p is countably additive.
n For each n there is a compact Kn contained in A . . Denote F ={ .21 < 1. It suffices to prove m ( f ( F ) )= 0.too lim p(A. lf'(4I < MI.235 Therefore Y is a measure. So n K. b) If /I is not countably additive. by a) there is a decreasing sequence {A. 03 # 0.} of measurable sets with empty intersection such that . 12 E E .) = inf p ( A n ) > 0. n . The necessity is obvious. Set Fn = I z E F .f(z>II MIY. 1 where M is any positive number. If(y) .21 if I Y . which contradicts the fact that n=l 4106 For f : [0. 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 K1.1] + B. If m ( E ) = 0. let E C { z I f'(z) exists}. ( IndianaPurdue) Solution. show that m(f(E)) = 0.
k ) ) m ( I n . = 4107 Suppose A c IR is Lebesgue measurable and assume that for any a. lim For any E > 0. which implies that m ( f ( F ) )= 0. .n + 1) such that + 1)) # 0. take a sequence {In$} of open intervals such that For 2 . .m ( F n ) = 0. (Iowa) Solution. + 1) n . and f(F)c uf(Fn) = ncu f(Fn).a < b. k ) < ME 5 M Thus m*(f(Fn))0. b E IR. There is an A n (n.236 Then F1 C F2 C * . If m ( A ) # 0 there is an n such that m ( A n (n. Prove that m ( A )= 0.n + 1) C U C (n.y E Fn f l I n $ . we have Therefore 5 k m * ( f ( F nf' I n .n open subset U in ( n .
and keep doing this ad infinitum.237 where E < m(An (n. 1 Remove from each of them their middle parts of lengths X l l i j l . We are left with the set K x . Claim. j We have which deduces that m ( A n (n. There are at most countably many disjoint intervals ( u j .n 1)). 4108 Choose 0 < X < 1 and construct the Cantor set Kx as follows: Remove from [0. (Stanford) Solution.1] its middle part of length A. Then the total length of intervals removed in the k 1th step is + k i=l . Assume that it holds for any n 5 k . For any n E W . n + 1) = U A n ( a j . b j ) . Prove that the set Kx has Lebesgue measure zero.n a contradiction. the total length of intervals removed in the nth step is X ( l The claim holds for n = 1. etc. we are left with two intervals 1 and 12. + 1))< E. bj)’s such that + Then A n ( n .
From the rightcontinuity of g. y ] = g(y) . from either A 7 . (Iowa) Solution.11) = P([O. Moreover. Y1. p ( z .O]) n+oo n 1 = 1 . 4109 Suppose p is a positive Borel measure on lR such that (i) p([O.g()) = 1. However.lim (g(O) .1 = 0. y E B. g(1) = p((0. It follows that the measure p is induced by g. I. Does this imply that p is the Lebesgue measure? Justify your answer. Y L0 <0 + Y. n+oo n and therefore p is the Lebesgue measure. For any z . for any R z .lim p((. Yes.l]) = 1. 11) .P ( { W 1 = 1 . or p(z z + Y1 = 4 0 . p is the Lebesgue measure. y E IR with z < y. It follows that g ( z ) 2 z .A y . = P(Y. 01. For any z E IR define + Then g : lR + l is nondecreasing and rightcontinuous. Y we have g(z + Y) = g ( z ) + d Y ) .g(z). It follows that the Lebesgue measure of Kx is 1 c n=l X ( 1 . and (ii) p ( E ) = p ( z E) for any Borel set E of 2R and every z E B.238 By induction the claim holds for any n E JV. we conclude that g ( z ) = zg(1).
ndoo is a signed measure. is a finite measure and pn’s are absolutely continuous with respect to Z .Show that for any Lebesgue measurable set E c 1R with X(E) = 1. for each E E U . there is a Lebesgue measurable set A c E with X(A) = !. by the VitaliHahnSaks Then k=l which shows that p is a generalized measure and therefore a signed measure. j (Iowa) . (Iowa) Solution. 4111 Let X be Lebesgue measure on R.G. Prove that u . Then . Let.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. Given any mutually disjoint measurable sets En’s.
then . u (Iowa) Solution.) = m(A) > 0. . Define the function f : IR f(2) = +[0. 1 . 2 E E. . 4112 Let p be a complex Bore1 measure on [0. 21. It follows from the StoneWeierstrass Theorem. ( IndianaPurdue) Solution.YI. c [I. 1 be a measurable set of positive measure. Z. co). 20 lim f(z) = 0 and lim f(z) = 1. + T.. A .f(Y)I x+ w 5 1 2 . = 0. 12 E A}.240 Solution.zo].1] by rl. 1 by 1 X(E n (m.) Since f(z0) x+w . Put A = E n (co. .T. S is dense in C([O. Show that there 1 exist two points z/ # z” in A with 2’ . . . which then implies that p = 0. . 4113 Let A c [O. Show that if r w for all n = 0 . . 2 . which is required. .Y E E . Denote A.. co]..z]). It is continuous by the following inequality If(.oo]). = {z Then m(A. . there is a point E E such that = f.m]) under the supremum norm topology. Let S = span{ePa5 I n E lN u (0)) then S is a selfadjoint subalgebra of C([0. Therefore for any f E C([O.7 3 . Denote all the rational numbers in [1. Thus . . co])separating the points of [0.z” rational.
Let n = min{k I 2 .+. Prove that E is Lebesgue measurable with Lebesgue measure 0. .. 2 E A such that ' ' z =2 ' + r. n=l 00 Suppose that A. Therefore there must be some m. .= 2) then 0 1 a:€ I l). By equality ( l ) . We will show next that * *  [O.T. Take 2 E An n Am.. I 2 and for any n there is an m 2 n such that 2 . 5 zl. Obviously.. which contradicts m ( A ) > 0. E c [0.1 . 4114 2.+ . 2 = n=l ? $ (where 0 I z. E is measurable and since + ["3 + 2.+ + * a +  3"1 The converse inclusion is obvious. Also show that E + E = [0.21) n=l 00 = 3.241 U An c [1721.. Then C m(An) I m([1. n Am = 0 if n # m. Then we c a n find z'.. there is at least one n such that 2.. 1)\E = u { 00 Indeed. for any 2 E [O. 3 31  3n1 xnl 7 = 2.1 2 2 + + 3" . Thus  = 'I'm .1).1]. = 0 or 1 .+1.xnlI 1 and therefore 2 21 + ..l)\E. n such that A n n Am # 0... (Iowa) Solution. =2 I1 + T.. 3"' 3 [? 2.
Obvior. n \ arcctgrc = T . [If I{. CF n= 1 03 5 z . I g ( 4 = . For any z E [0. I 2.242 m ( E )= 0.T } \ = 0. < T.1) say z = 2 .gl)loo 5 E .sly. 1 gl(z). 1z . T. . L E.E + E [0. 13. 2 ' ' = C . and z . and let E > 0. 4115 Let f : It2 + IR+ be measurable.+~ r .  Set g(z) = gl(z) . () . (IndianaPurdue) Solution. (where 0 Define i]. = +m.} such that 0 < r1 < r2 < . limr..gllm 5 E (ii) for every T E B. . Since E C [0. Take {r.. and for any n. \ { z g z = T } \ = 0. there is an m 2 n such that # 2). {zIg(z)= r } = {z [If gllco = .=l 00 X" $ belong to E so that = z.arcctgz. n E N. . We have g 2 0 and For every r E B.arcctgz = T } TI. for all n. Denote () En = { z Set g1 = TnXE. ~'$2'' Then z' = C 0 0 3 and . = 2. < E . I r n .1 < fi(z) I ~n}. . Show that there exists g : I22 + El+ measurable such that (i) .glllcx. Set fl(z)= f z + arcctgz. n= 1 This completes the proof. = Since \ ( r c U{zI z E Enlarcctgz = r. .}I = 0. llfl Then g ( )2a c c t g z . < .
h0 lim m ( ( E+ h ) n E ) = m ( E ) . .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 . and 4117 Let E be a Lebesgue measurable subset of lR with m ( E ) < 03 and let f ( ~= m ( ( ~2 ) n E ) .1 2 ) . a) Prove that f is measurable. . We will show first that for any Lebesgue measurable set E . we have which then is measurable. . O .U p . b) Evaluate 1 1 f(z)dz. ) Show that (a) f ( z ) is a continuous function on R. 1]\K = p 2 1 ai=0. By the definition. a 1 . (b) lim f ( z ) = 0. (Stanford) Solution.1 1 .P . * a p . z++w + (Illinois) Solution. Let K denote the Cantor ternary set.243 4116 Let f be the function on [0. a 1 a 2 .2 uu (0 . . Then [O.
there is an increasing sequence { E n } of compact sets such that En C E and lirn m(E. 3 ~ . h0 In general. then n 5 hm(En(E+h)) h+O <_ h+O lim m ( E n ( E + h ) ) 5 m ( E ) .G( a . an open set E’ of the above form such that E c E’ and m(E’\E) < E . If E is a compact set.@i)’s are finitely many mutually disjoint open intervals of finite length.244 If E is of the form . for any E > 0. then there is. n+m Then m(E) = 5 + h) n En) lim lim m ( ( E + h) n E ) = lim m ( ( E + h) n E ) . 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. so hO lirn m ( ( E + h)nE ) = m(E). n+mh+O n+m lim m ( E n )= lim lirn m ( ( E n n+m h10 h+O .) = m ( E ) .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 ) . ) o i )where ) r=l n (q.
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 . z b i ) (ai.m ( E (Y .bi) n i=l <m and any z >0 we c ZB nB.245 so h0 lim m ( ( E + h)nE ) = m(E).z)) n E ) m ( ~ ) + + + + M ( ( E + z)\(E + Y)) + + . (b) If E is compact.y ) \ E ) = m ( E ).m ( ( E + (z .b i ) .z)\E) m ( E (z . the line y = mz has a nonvoid m 1 intersection with A x A .m ( ( E + z)\(E (a) For any y. (Iowa) Solution.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 ) . then for any n have n U ( z a i . Prove that cp(z) = J X & ) X A ( t ) d t is continuous a t z = 1.f(.1 < E . If B is a bounded open set.y)) n E ) + 0. n E ) .)\(En + + z)) + m(E\En)) = 0. then there is an T > 0 such that for any z > T . z E 2R If(Y) . 4118 Let A c l be a set of positive R Lebesgue measure. Use this result to prove that there exists an E > 0 such that for any m E 2R with I . The claim follows. . Then + E ) n E = 0. say (a*. where 2 5 m 5 03 and ( a j . In general there is an increasing sequence { E n } of compact sets such that En E and lim m ( E n )= m ( E ) .)I = Im(((E+ Y)\(E + + Y) ) l 5 m ( ( E+ Y)\(E + = m ( E (Y . (z as y + z.
C A and p(A\ n=l 00 n= 1 u K.} of compact sets such that U K . n n .) aJ = 0. 00 Since p ( z B .p ( z K n K ) I P(Z(B. n Bn) .a) 5 ~ P ( z B ~ B ) 21 5 X’lim ~ l ( Z nB ) B 5~L(B).246 We have p(B) = sup n<m a = 1 C(bi . there is an increasing sequence {K. there is a decreasing sequence {B.} of bounded open sets such that K = n=l n B. By (2) we have Since . If K is a compact set. (1) where p is the Lebesgue measure.\K)) we have by (1) + P(Bn\K) In general..
then 03 = .e. ) } increasing to p. n03 a contradiction. if A is a measurable set with p ( A ) > 0.otherwise.247 If there is no E > 0 with the property mentioned in the question. Given a totally ordered set Q of P . p(S)]. then [B]5 [ A ] .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 . 4119 Let p be a countably additive measure on a set S with p(S) < $00 that is without atoms. i. If there is a t o E (O. Prove that the range of p is the closed interval [0. let . However. For any [ B ] E Q. Let where is an equivalence relation: A B if p(A\B) = p(B\A) = 0. then there is a measurable set B c A such that 0 < p ( B ) < p ( A ) . Let A = n=l  P = { A I A measurable and p ( A ) < t o } / . [B]= [A] since P(A\B) 5 C p(An\B) n=l 03 =0 . SO p ( A ) = /3 < t o and therefore [A] E 7'. then for 2 1 any n E mT there is an m. this implies that m . if [ B ] 5 [An] for some n. and therefore p ( A ) = lim p ( m n ( An A ) ) = 0.p(S)) not in the range of p. ( A n A ) = 0. ] } of Q with { p ( A .  u A n .) Solution.. Then P is a partially ordered set: [41 5 [B]if p(A\B) = 0. (Courant Inst.B = sup{p(A) I [A] E Q } . ) = P 5 to. E B such that Imn .403 lim p ( A .
Let  R ={BI B c S\Ao. then [B]5 [C]. (Iowa) . There is by Zorn’s Lemma a minimal element [Bo] R. by the assumption that there is a subset of COof Bo with 0 < ~ ( C O ) ( B o[CO] [Bo] [CO] [Bo]. (e. p(S\A) > t o . Let n=l then So p ( B ) = a > 0. Given a = inf{p(B) I [B]E S}. For any [Cl E S.248 It follows that [A] is an upper bound of Q in P. It follows that [ B ]is a lower bound of S in R. then cr 2 t o . 5 and # a 4120 Let X be a compact Hausdorff space.p ( B n C ) = a = 0. <~ ).p ( A 0 ) . Therefore p ( B ) > t o . 5 as above. [B]E R. if [Bn]I [C]for some n. ) } decreasing t o a.p ( A 0 ) whenever B C S\Ao and p ( B ) > 0.g.otherwise [B]= [C]since n=l and ~ ( B \ C )= p ( B ) . contradiction.p ( A 0 ) ) . Consider the calgebra 2 generated 3 by the compact G sets. It follows that p ( B ) > t o .P(B) > O}/ . Show that any positive measure p on B which is finite a on compact sets is automatically regular. There is by Zorn’s Lemma a maximal element [Ao]of P. However. Equip 72 with the partial order a totally ordered set S of R. Let where is defined as above.p ( A ) and there is a decreasing sequence { [Eln]}of S with { p ( B .
249 Solution. Recall that sets in 23 are called Baire sets and each compact Baire set K is a Ga set. there is a V in 0 such that X\U C V and p ( V ) < p(X\V) E . Let us show each set U in 0 is inner regular. respec3 tively. L in K . . 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 .) < . E i = 1. there are B1.\E. If { E .U (Ki\Li) I K i .. K\L is regular. i= 1 n} is a finite class of mutually disjoint. Then 7E is a ring such that 23 = By definition. 7 ~ . where the symbol 0 means disjoint union. I 1 5 i ? sets. Thus K\L is regular. each set in K is outer regular. Let us preceed in five steps to show the regularity of p. M) + E. E 0 such that E C Bi with i p(B. . 6 i=l Ei C ( Bi J i=l . Let K and 0 denote the classes of compact sets and open sets in 2 . Consequently. Then X\V 2 U and p ( U ) < p(X\V) E . . then U Ei is regular. . . .V open and V E 23) p ( E ) = sup{p(K) 1 K C E. B. 5 P(B\L) < P(K\L) p(K\L) < p ( K +E. Also recall that E is outer regular if p ( E ) = inf{p(V) 1 E and E is inner regular if C V. For any pair K . Let a = { ¶=1 . For any c > 0.K ? compact and K E B } . Li E K}. Then we have V). .. + + + K\L K nM c B\L. Step 2. regular Obviously For any E > 0. C K\L. Step 1.
. n=l n En is . n=l Let E = u En. If { E n 03 In E IN} is an increasing sequence of regular sets then (J En is regular. while the inner regularity follows from the following inequalities co Step 4. Let V = U1 Vn E 0. . .. Obviously 00 n=l p ( E ) I inf{p(V) I E For any E G V E O}. E 0 such that C Vn and p(Vn) < p(En) + &. . then n= which shows the outer regularity of E . . If { E n } is a decreasing sequence of regular sets.250 and which shows the outer regularity of the following inequalities u Ei. En > 0 there are 03 V1. V. then regular.. The inner regularity follows from n i= 1 Step 3.
t monotone class..there is a K . Let S = { E E 23 1 E is regular }. > n> . E ?E such that n= 1 and which shows the inner regularity of E . Let Assume f ( ~is finite for all ) T > 0 and assume Prove that p is identically 0.S is follows that S = B.. It 4121 Suppose that p is a nonnegative Bore1 measure on En. By steps 3 and 4.251 Let E = n En.u(En) = p ( E n 5 V E 0) Step 5. (Indianu) . inf{p(V) I E V E 0} 5 inf inf{p(V) I En = inf. The outer regularity follows from the following inequalities. S contains a. 00 For any E > 0. n E lN. By steps 1 and 2.
. i= 1 Letting E + 0. 4122 Let 1. (Indiana) Solution. zl.. 1) denotes the open ball centered at z of radius 1. T ) c B ( z . . It is easy to show that Assume without loss of generality that B(z.2s) and C(z*. zk E IRn such that . in K . and define a function f on Enby f ( z ) = p ( B ( z . There exists an sequence { z k } of K such that k+oo lim f(zk) =a. Prove that f attains its minimum on each compact set of IR". K c u k i=l C(z"7) c C(0. let s > 0 be such that K c C (0 . a s ) ) & .l) %k + z B(Xk31)k+m Hence .&?Fir).s). E Given a compact set K of IR". we get that p ( K ) = 0 and therefore p is identically 0.1 be a finite Bore1 measure defined on lRn. Let K be any nonempty compact set of 1R" and let a = i n f f ( K ) . ) ' S are mutually disjoint.1))where B ( z . Then T k i= 1 k i=l k = Z A ( C ( z $ T ) ) & 5 X(C(0.252 Since C(z. For any > 0 there exists an T > 0 such that g ( T ) < ( ~ T ) ~ EThere exist finitely many .
E a contradiction.  1 27 28 (Indiana) Solution. It follows that 0.= 0. So E is Borel measurable and W m ( E )= 1 n=l 9"l x = 1  1 10" 1 = 0.anl 7: n = 1. . 1 0 .. 2 (*> Show that if E c IR is a measurable set with m l ( E ) = 0 then m . If ak = 0 for all k 3 : > n then = 0.f(y)ln 5 e'zl+lyI 1 .an17.}..~~lS..alaz. ... .2800. ) Hence a is finite and f attains its minimum on K .= 0..) . . E. 4124 Let f : IR + R" be a function such that for all z.ala2. . .Ulaz. For any z E [0. 4123 Let E be the set of all numbers in [0. y E R If(. (ii) Determine whether E is a Borel set. 3 ' 100 ' 100 (i) Compute the Lebesgue measure of E .253 which shows that P ( B ( X . The reverse inclusion is obvious.1] which can be written in a decimal expansion with no sevens appearing..yI.... .UlU2. and therefore I # [O.a1a2 .0. 3 3 3 . l]\E. ..2'. ( f ( E ) ) = 0. .611a2 * * * Un17 = O.a. Let n = min{k 2 1 I ak = 7).aia2 * * * ~~1699.2699.. . (Indiana) .. 1) = Q.Un17< x < O.. . . Thus.~~18)~l. E E .l]\EU{(O.. . write z = 0..
bi)) i c ~(f(ai). Prove that f is measurable. m * ( { zI lfk f )(. where (ai. (Indiana) Solution. Prove that either A or R\A has Lebesgue measure zero. Let fk = fXEn\u..bi)'s are mutually disjoint. 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 ( . then f k is measurable.a i l ) ' ) . For any E > 0. Assume without loss of generality that E is bounded. ( f ( ~ ) _< ) C C. Let u k be an open set such that x ( u k ) < and f is continuous on R n \ u k . Assume that r A = A for every nonzero rational number r.254 Solution. Let E 2 (r. (eZrlbi.I 2 E l ) = m * ( { zE uk I If(x)I 2 E } ) 5 C. there is an open set U with X(U) < E such that f is continuous on E"\U (in the relative topology). a which implies that rn:(f(E)) = 0 and therefore f ( E ) is measurable.T . (Indiana) . Write U = U(uilbi). 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. 4125 Let f : R" + lR be an arbitrary function having the property that for each E > 0. Then condition (*) implies that f((ai.ail < CneZr&. Hence mn(f(E)) = 0. Since the Lebesgue measure is complete f is measurable. It follows that r n .r). T ) and rnl(U\E) < E .
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. For each measurable set E QJ then Y is the desired probability measure.m). Let lR* = R\{O} and 3 = A\{O}. m ) ( ~ ( z = 1) such that p is absolutely ) continuous with respect to v . c )by m f(2) = 1. Then f is continuous and for any nonzero rational number Hence f( z) = 0 for any z E R* Since . dY XK(Y)XLl(Ylz). . Y 2 E a*. and v is absolutely continuous with respect to p.255 Solution. Prove that there exists a probability measure Y on ( X . define function f : lR* + [0. Then r B = B and T(R*\B) = R*\B for every nonzero rational number T .. T. If m(lR*\B) = m(R\A) > 0 there exists a compact subset K of E*\B with positive Lebesgue measure. 4127 Let p be a cfinite measure on the measure space ( X . we conclude that m ( L . It follows that m ( B ) = 0 and therefore m ( A ) = 0.l ) = 0 and therefore m ( L ) = 0. (Indiana) Solution. For any compact subset L of B .
i. g : IR + IR is still integrable. However. 1). 1)\E i s a Lebesgue point o f f .) < E . 2 ( 0 . since g o f (z) 5 1. and fdm = 0. For example.e.l}(x). . l ) such that E > 0. there is x EJ. 4202 Let f E &(O. Prove that for every open interval I E ( 0 . It is not true. the function g o f is not integrable. There is a measurable set E of measure zero such that any x E (0.. l ) and every an open interval J. 1 = x{o. (Stanford) Solution. Then f and g are integrable functions with compact support. l ) fdm = 0 . (Illinois) Solution. let f. (> = X { 0 } ( 4 and d ..256 SECTION 2 INTEGRAL 4201 Prove or disprove that the composition of any two Lebesgue integrable functions with compact support f . m(J. Assume that for any x E ( 0 .
one has Jx f d p < CO) if and only if the series n=O c 00 I f (z) 2 n}) (Iowa) converges. .257 For any 2 E (0.) < and It follows by equality (1) that f ( z ) = 0.. p ) be a positive measure space with p ( X ) < 03.. 03) is integrable (i.. Solution. Therefore 4203 Let (X.e. l)\E and for any n there is an open interval Jn & ( 0 . Show that a M measurable function f : X + [0. f ( z ) = 0 a. Then W . i. Suppose that f is integrable.e.e. m(J.. l ) such that z E J.
1]. (X. (a) Yes. ) (b) No. which shows that f is integrable. + E . fdp = f (0) for all continuous f : B + (c of compact support? Justify.258 Conversely.?r. 4204 (a) Is there a Borel measure p (positive or complex) on IR with the property that J . / p(t)dt > 0 4205 Let E be a Banach space. Let p ( E ) = x ~ ( 0for any Borel set E . taking value one on [1. let ‘p 2 0 be a continuously differentiable function of compact support. If there were such a Borel measure. Then a contradiction occurs from the following limits JL% and Irn ( p ( t ) e t d t= .p ) a probability space. f d p = f’(0) + (c for all continuously differentiable f : R of compact support? Justify. and f : X such that g o f is pintegrable for every g E E‘. (b) Is there a Borel measure p (positive or complex) on B with the property that J. (Iowa) Solution.
Since .. If (b) does not hold. Assume that pn + ‘p in E’ and T(pn) + h in L1(X).259 Define L : E‘ + IR. p ) be a positive measure space with p(X) g be realvalued measurable functions with < co. (Iowa) Solution. L ( g ) = I g 0 fdp. By the closed graph theorem. It is ture that L E E”. We have So L is bounded.and let f and Show that either (a) f = g a. M . or (b) there exists an E E M such that (Iowa) Solution.e. Does L E E”? Justify your answer. p H p of. 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 ) . then for any E E M . 4206 Let ( X . Define the linear operator T : E’ + L1(X). T is bounded.
4207 Let ( X . Since a' is second countable. J. I f(z)@ S } has measure (Iowa) Solution. From the equalities and we conclude that p ( E + ) = p ( E . and S a closed set ina'. . > 0 the sets are measurable. d P .If .) = 0. For any E fdP = J. Show that { z E X 0. one finds that a'\S is the union of countably many closed balls { z E C I Iz .M .260 for any E E M .A 1 5 E ~ } ' s . p ) be a positive measure space. It follows that Therefore (a) holds. Suppose that 1 JE ) P( E f@€S whenever E E M and p ( E ) > 0.
) . Denote El = E (f Then m(E2) < 5 .1] with m ( E ) 2 a .not to S.XnI I If(.1] + (0. Show that where inf is extended over all measurable E c [0. m) and let 0 < a 5 1. 4208 Let f : [0.X. where 5 E ~ } ) .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. Therefore 2 +). m ( E ) > a.261 there is a t least an n such that P ( { ~ E Xf I ~ ) . so m(E1) > $.1]. I 5 E ~ } . Take an N such that Suppose that E C [0. 0bviously. (IndianaPurdue) Solution. EZ = E\E1. Thus .
a measurable set E of measure zero such that for any t E B\E.l ) ( t ) k=2 converges.f k . with respect to the Lebesgue measure on lR. It follows that fn + f almost everywhere.e. by Levi’s Lemma. Since fn + f almost everywhere with respect to Lebesgue measure. n Therefore for any t E lR\E. 4210 Let /I be a finite measure on lR. Let . Prove that Solution. (I Zin o is) l there is.262 4209 Let {fn} be a sequence of realvalued functions in L 1 ( B )and suppose that for some f E L1(IR). (Indiana) Solution. n fn(t) = fl(t) k C ( f k .and define Show that f ( z ) is finite a.
[p]. .e. p ) be a positive measure space. m] an integrable function. i. d v ) .M . The conclusion follows from that the measure v and the Lebesgue measure are equivalent. Assume. the function is finite a.e.263 then g E L 1 ( R .where dv(z) = 1 dx +22’ by h b i n i ’ s Theorem. without loss of generality. f n : X + [0. CQ] a sequence of integrable functions. that p is totally afinite.and that Prove that (Iowa) Solution.. Suppose that fn +f a. and f : X t [0. with respect to the measure v.e. 4211 Let ( X .
For any A E M . There is by Egroff’s Theorem an E with v ( E ) < 6 such that converges t o f uniformly on X\E. Then 9 { $} . 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. Let d X n ( z ) 1f n ( z ) d p ( z ) . Then v is a finite measure on X . 71 E m* Then An << v. equivalent t o p. Let Then g is integrable. then by Fatou’s Lemma one has {s. Let d v ( 2 ) = g ( z ) d p ( z ) . by the VitaliHahnSaks Theorem.264 n=l where Xn’s are mutually disjoint measurable sets of finite and positive measure. f n d p } is bounded and therefore admits a convergent subsequence. {s. If f n . d p } is any such subsequence.
then (Iowa) Solution. as desired. Let W=C1 O0 n=l 2 1 . finite measure on M . The set function is of course a positive. and n=l 03 p ( X . Conversely. P . finite measure on M . ) > 0. and that P ( E ) = 0 ep ( E ) = 0 for E E M . (a) Let {X. l ) . (b) Show that if f is a complex function on X which is measurable with respect to M.}be a disjoint sequence of M such that X = U X .265 4212 Let p be a afinite. If p ( E ) = 0 then P(E) = 0. (a) Show that there exists W E L1(p) which takes its values in the open interval ( 0 . positive measure on a ualgebra M in a set X. + p(Xn) xx. Show also that r is a positive.
Since and lim fn n+cc = f. if f is integrable with respect to the measure p. there is a sequence { fn} of simple functions integrable with respect to p such that Then fnW’s integrable with respect to p . then both sxf d j i and sxfWdp equal to 2 aiii(Ei). i. . . . Then fnW is integrable with respect noo to p and hence fn is integrable with respect to p. and since are lim n+w Jx\jnwfmWldp= lim n+w m+cc ~f~fmldji=~.. 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 . n In general.f is integrable with respect to p. m+oo ncc lim fnW = f W . We have and therefore J. we see that fW is integrable with respect to p. .266 (b) If f is a simple function. n f =CaixEi! i=l where p ( E i ) i= 1 < +a.e. Accordingly.= i l .l n .fdF=J X fWdP Conversely.
IJ.1] t R such that (3) 0 5 g 5 1. there is a 6 > 0 such that IE . ( 4 ) g = 1 on K and (5) g = 0 outside U . fndm= 0. if : ( IndianaPvrdm) . 4214 Let { fn} be a sequence of nonnegative measurable functions in LP( R) for some 1 < p < 00.Suppose that (i) fngdm + 0 for each g E C([O.267 4213 Let m denote the Lebesgue measure on [O.1]. For any E > 0. Then we have = n+m lilfnXAdml It follows that Jiir J. Show that J. 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. fndm + 0 for each Bore1 subset A [0. There is a continuous function g : [0. (Iowa) Solution.fndml % hdm < E whenever m(E) < 6. 1 and let (fn) be a sequence in 3 L 1 ( m )and h a nonnegative element of L1(m). Show that fn 4 f(P) and only if f 4 fP(L1).11) and (ii) Ifn I h for all n.
”+fP. Just as above.fP)/ I J tfi Jfp. take . 2 f. +  Since IJu~ > 0. we have Since T:+X SO =f.f I L Ifn . it follows that +  Jfi For any c N1 such that .” Then fn f p l .  PI 0.268 Then f n 5 f and IL . Suppose that J If.fl which implies that IIE .f (Ip + 0.+0. Therefore Conversely. f > 0.
Vn . Denote 7 = &/m(A)f.f l > q ) < S i f n > N . < EP e if m ( e ) < S and n > N1... Take N such that N 2 a n d m ( l f n . < f P &PP Then J.. All parts refer to Lebesgue measure on R. (a) Give an example where { f n } converges to f pointwise. llfnllp and llfn . 5 M . a measurable set. 4215 Let 1 5 p < 00. f: > N1.fllp f . Thus Jf. UP Ifn flp) for any n >N. Wehave N1 (J. Take S > 0 such that < EP for n if m ( e ) < 6. Take A .269 for n > N1 and any measurable set E . 0. such that m ( A ) < +m and J.
(c) Show that if { f n } converges to f pointwise and llfnllp + IlfllP. Using the Fatou lemma. on [0. we have Therefore no3 I f n .f l P . (a) Consider L[O.270 (b) If { f n } converges to f pointwise and l ] f n l l p + M < $00. Then zE lim f n ( z )= f ( z ) = 0.1]. (c) Set gn = 2"(lfn Ip + Ifl") .fll = llfnll = 1. then +  llfn .e.l f n . By the Fatou's lemma. i ) .fJPdz = 0. Then gn 2 0. [O. 1 1 and (b) We conclude that llfn . 4216 Suppose f n is a sequence of measurable functions on [0. Prove 1' lfnldz t 0 . and n+m lim g n ( z ) = 2 p + ' l f l p pointwise.fllp 0 (IndianaPurdue) Solution.13. what can you conclude about Ilfllp? Justify this conclusion. Set ncc fn =n ~ ( ~ .1] with 1' and f n + lfn(z)12dz 5 10 0 a. llfllp 5 M .
00).M . 1 We have by the CauchySchwartz Inequality and therefore Let E + 0.1 I E ) < E and (2) f n converges to 0 uniformly on [O. (Stanford) Solution. and we have rl 4217 Let (X. then lim P10 (J.. 1]\E. It suffices to show that If f = 1 a. equality (1) holds.27 1 Hint. Show that i f f is an integrable function with values in [1. For any E > 0 there is by Egorov’s Theorem a measurable set E c [0. (Iowa) Solution. Use Egorov and CauchySchwartz.e.log f d p ) .f p d p ) ’ = exp ( J . Otherwise.1] such that (1) m([O. for any 0 < p < 1 and any 2 EX . p ) be a probability space (a positive measure space with p ( X ) = 1).
.272 By the Dominated Convergence Theorem we have nco lim log(Jx f P n d P ) Pn = = n lim 00 log(J. l ) decreasing to 0. Consider 00 n=2 a) Prove that this series converges in L1[2. k=2 x 2 2. (Stanford) Solution. 00) b) Prove that where the sum on the right is the pointwise limit (you need not prove that this pointwise limit exists). where { p n } is any sequence of ( 0 . n fn(x) = j+k5.m).(fP"   1)dP) J#"  1)dP Pn = Llogfdp. 4218 Let { a n } n > 2 be a sequence of real numbers with lan[ < logn. a) Since The series b) Let C n=2 00 ann2 converges in L1[2. 1+ ( f P n Pn  lim n00 ( log(1 + JJfP" J.
(Iowa) Solution. 2 Theorem. 1 2 . H IakIk" k=2 is integrable. and by the Dominated Convergence 4219 Let S be a bounded Lebesgue measurable set in IR. Show that where X is Lebesgue measure. By the RiemannLebesgue Lemma. we have = X(S). 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).
(a) The function f is not integrable on the unit square.. for ) b . Y)ldXdY rdrde drde (b) We have . (c) Does the integral exist? (If you use a theorem. 0 5 y 5 l? (b) Compute the repeated integrals in the two orders.274 4220 (a) Is the function Lebesgue integrable on the unit square 0 5 x 5 1. Y ?M2+Y211) If(. be explicit!) (Iowa) Solution.
p ) be a measure space and f E L ( p ) . 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. (Iowa) Solution.275 and 4221 Let ( X . Evaluate ncc lim L n l n ( l + (!)’) dp. It is easy to show that for any x 20 2. Define the function H .l). In( 1 + x’) 5 It follows that for any n E lN the function n In( 1 !$) is dominated by and therefore integrable.M .
Prove that 'p is differentiable on (. It kkz By the same method as above.1 . and we conclude that t H e " t f ( t ) is also in L1(E). t t Iy . +m etztf(t)dt. For any fixed z E (1. for any y ) > z. (Iowa) Solution. I).1.276 for all x E (. et"t2f(t) is integrable. 2 0 and x < y < H < 0 and x < y < A' lzz.x ) ~ (x < z' < z ) = I(ett  e")tf(t)l. Next we show that dx E (1. Since for any x E (1. we see that t follows that 4223 Evaluate . l.zlet"t2f(t). e"f(t)dt = m J_.l).l). < z < y) = \ e t z " t 2 f ( t ) ( z.z l e ' z X t f(t). (x 4 and t A 2 Iy . l ) .
. Hint.)n =ex. we have (I+ x ) n s e x . (Stanfod) Solution.(lS Z! i=l + and ncc lim ( 1 + . .277 justifying any interchange of limits you use.. Since for any n E IV and z 2 0. i=l )(l&). First show that (1 E)n 5 e" for x + 2 0. By the Dominated Convergence Theorem = brn e"dx = 1.
co) a measurable (Iowa) Solution. 4225 For i = 1.2. Let Mi = 2N (the galgebras of all subsets of nV) and let pi be the counting measure. let Xi = nV (the natural numbers). For the function f : X I x X z + B defined by . since ' '{ f ( x ) 5 el "f'ln f . otherwise and 0 5 In f 5 f In f we see that both f and In f are integrable. and Indeed. Prove or disprove: + [I. p ) be a probability space and f : X function. Under the condition that f In f is integrable we conclude that both f and In f are integrable. Since l n t 5 t l n t for any t > 0.278 4224 Let ( X .A.
f ( i . j = is 1..j)is integrable and J.2i. j ) is integrable and Since f is integrable ( C I f l ( i .p ) be any measure space. Compute the iterated integrals and How do you reconcile your answers with Fubini's Theorem? (Iowa) Solution. For f E L1(Q. f ( i . 4226 Let (0. For any i E X I . otherwise. j ) d p z ( j )= 2i + 2i Therefore For any j E X 2 . j H f(i. 2i. and for X > 0 define Cp(4 = P({Z E Q I f ( Z ) > XI) . the two iterated integrals exist Li€N and coincide by the Fubini'i Theorem. i H = 0. p ) . j = i. f ( i . j ) = 2). j )= { 279 .
280 and 4(A)= P ( { Z Show that the functions 'p E c I f(z) > A>). ~ ) ( l f l ( zA) is measurable. A) H x ~ O . The function (z. are Bore1 measurable and that llflll = J m ( ' p ( A ) + +(A))dA. 4227 Let f E L 2 ( 0 . 2 and II. (Iowa) Solution. . By Fubini's Theorem we have provided that either side exists.) + but on the other.I). it may be helpful to consider f positive first. As usual. Set . On one hand. The measurablity of 'p and follows from that they are monotone. X . 0 Hint.
F ( z ) h l2 dz) 4228 Let f. Imine) Solution. y E [ O . We have ( I J1h 0 F(z + h) . (UC.g E L1(O. . l ] . 1) and assume that f(z)g(y) = f ( y ) g ( z ) for all z. 0 < h < 1 where c is a positive number independent of f.281 Show that for each h.
.282 where A = {(Z.Y) I 0 5 2 < Y 511 (Iowa) and dA denotes the planar Lebesgue measure.
n}. .n. let M such that p.=. must it be the case that fifi . Show that for any [0. . .p > 1 and be positive numbers c. By Holder’s inequality. . n 1 1 i=l If. The function I is strictly increasing if and only if f is not almost everywhere equal to the constant function.. we have + which shows that I ( p ) is a nondecreasing function of p E (1.w)valued measurable function f on X . fi E CPi ( p ) . For any p . i = 1 . .m). for i E { 1. .. is a nondecreasing function of p E (1. p ) be a fixed measure space. let CY = and = then $ = 1. m). . * f n E Cp(/l)? (Iowa) Justify. q E (1.. w) with p < q . > 1. 4302 Let (X..283 SECTION 3 SPACE OF INTEGRABLE FUNCTIONS 4301 Let X be a positive measure space of total measure 1. . Under what circumstances is I ( p ) strictly increasing as a function of p? (Iowa) Solution.
P({b)) = 1 1 and P({. Then for n=k+l.fnllp 5 llflllpl .and L"O(p)? Justify your answers. . L z ( p ) . (1) In case n = 2.)> = 00.. f k ) f k + l E LP and which completes the proof. 1 o < +. (Iowa) . and let p be the measure determined by P ( { a ) ) = 0.1.1  which shows that ppk c+k12p > Pl P k ? ' Pk+l PPk 1 pk+l?p < ' 9 1. b. and by H6lder's inequality.*. By inductive assumption f i e . c}. let M be the cralgebra of all subsets of X .. . ..284 Solution. IIfnIlp. lfilp E L"(p) and PI If2Ip E L F ( p ) . Assume that the conclusion and the inequality (1) hold for n = k . f i * . Yes.+ 1= . fk belongs to 1 L Pk+:P ( P ) and Then by the case n = 2 we conclude that (fl . 4303 Let X = { a .. Moreover llfl.. fn E LP(p).1 . What are the dimensions of the vector spaces L 1 ( p ) . .
we conclude that dimL'(p) = dimL2(p) = 1. We have dimL1(p) = dimL2(p) = 1 and 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 ) . + < oc) and let . c } we have dimLW(p) = 2. Since two functions are equal in L"(p) if and only if they are identical on { b . Also let f h ( z ) = f ( z h).285 Solution. s > 0. ( IndianaPurdue) Solution. Denoting we have 4305 T +s =p. Let f be a nonnegative function in P ( E )for some 1 5 p T > 0. 4304 lo Show that f d m = 0.
(ii) We claim that lim lhl+m 11 f i f s I I 1 = 0. (ii) Investigate what happens to 11 fifslll as fi E Ihl + co. take N such that JEc f P < E where E = [ .286 (i) Show that fif s E L1(R). then z h 4 E whenever z E E. f" E L f . Now if Ihl > 2 N . Show that (IndianaPPurdue) Solution. Then p > 1 and f + & = 1. and let 0 < T < 00. (IndianaPurdue) Solution. it E L1(R). (i) Obviously follows that fh E Lp(R). Therefore + 4306 Let f : R +R+ be measurable. From f[fS L f . N ] . For any E > 0. Since we have . Set p = 1 + f . and P +P = 1.N .
: Therefore .287 which implies 4307 Let f be a bounded measurable function on (0. llfllm = I I Ilf llm.l). Prove that (IZZinois) Solution. For any Then E > 0.
1.1] satisfying Determine those values of p . m M it follows that f P is integrable. 3 (Iowa) Solution. Then f 2 1 2 is not integrable.288 4308 Let f be a nonnegative measurable function on [0. The minimum value of p for which f may fail to be in LP is 2. Indeed. 2). Let f(z)= However for any t > 0 & . (Illinois) Solution. 1 l+t 4309 With Lebesgue measure on [0. 11).13. 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. then f E LP. 1 .1 I llfllcc I 1) 1 is not compact in L1[O.  1 (1+t)2  < 2. for any p E [l.let . It suffices to show that S is not closed in L1([O. If 1 5 p < 2. prove that s = {f E C[O. For each n E IN.
1]. i. with inner product defined by 7) ( u . (Stanford) Solution. in the strong (i.2 1 . but not compact.too lim (un.271).~ is bounded.e. the set {u. n = 1 . w E 'H.}.i. n 2 1 with m # n which shows that {urn I m E IN} is closed.. Since it admits no convergent subsequence..w E L1(0. the Hilbert space consisting of functions f E L2[0. By RiemannLebesgue Lemma. . (a) Obviously. For any m.(z) = sin(nz) for z E (0. for every w E 'H.  {u. (b) For any w E X . 4311 Let H1 be the Sobolev's space on the unit interval [0.}rZl . ..e.w) = 0.". u . 2 . E 'H.2r) by Holder's inequality. norm) topology of 2.1 such that 1 . defined by u. Show that (a) the set is closed and bounded. + 0 as n + 00 in the weak topology of X . it is not compact. (b) u.e.w) = 1 2= u(z)w(z)dz.  Consider the elements u.289 4310 Let 'H be Hilbert space L 2 ( 0 ..
Suppose For each h E lR define the function fh by f h ( z ) = f ( z . (i) Give the Fourier expansion of fh.1] too the series n=w C f^(n)e2ninzconverges to f(z) both in H1 and in L"O.290 n=cc where are Fourier coeffents o f f . Solution. (Stanford) . (iii) Prove that unless f is constant almost everywhere. 2 T ~ belongs to ~ ~ ( O .fllz in terms of the a. Show that there exists constant C > 0 such that (Stanford) I l f I l P I Cllflll. 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. (ii) Find the &norm llfh . 2 a ) . Since for any z E [O. and h.h).
= 0.1]. Prove that for all 2 E [0. there is an n # 0 such that an # 0. (i) We have n=w +w n=w (ii) We have n=w (iii) I f f is constant almost everywhere.. I f f is not constant almost everywhere. (Iowa) Solution.1).f l l 2 = 0. By Bessel’s inequality . Then 4313 Let f n ( x ) be an orthonormal family of functions in the Hilbert space L2(0. n # 0. and that this inequality is sharp (equality) if and only if {f.291 Solution. then a.1). 1 n = 1..}span a dense subspace of L 2 ( 0 . It follows that fh = f and 11th ..
then t+O lim llft .f n t l l 2 + llfnt . Then L t+O lim llft .1 b l Ia Ib 5 I} O is a dense subspace of L2(0. since s ~ a n { x ( o .1). Suppose that f : lR + a' be a continuous function with compact support. (Stanfonl) Solution. l]} is compact. {fn} span a dense subspace of L2(0. it follows and therefore. I~0] 5 z I 1) = s p a n { ~ ( ~ . 4314 I f f is a function on IR.[1.fnll2  + llfn .fllL2(nz) = 0.1) while is a closed subspace of L2(0. Then ft converges to f uniformly.l). there is a sequence {fn} of continuous functions with compact support converging to f in L2(IR).fllz 5 S(1lft .1] = { z y 12 E K .l). E nV} span a dense subspace of L 2 ( 0 .fllz) i t0 2 l l f n W . In general. let ft be the translate ft(z) = f ( t + z). y E [1.292 If {fn I R.then {fn} is a n orthonomal basis of L2(0. Prove that i f f is square integrable with respect to Lebesgue measure.1) and therefore (1) is sharp. We have where K = supp(f) and .fllz + llfnt fnllz) = 2llfn . Conversely. .fllz.z] 1 0 5 z 5 l}. containing span{x(o.
Hint.00. Then . 4315 A trigonometric polynomial is a function p on lR of the form N n=N for some Cn €a'. You may wish t o think about harmonic functions on the unit disk.293 Letting n . Regard f as a continuous function on the unit circle S1 of a'. (Stanford) Solution. there exists a trigonometric polynomial N n=N such that Ilf and PI103 <E ICnl 5 lan] for all n. Let u be the harmonic extension o f f to the unit disc D = { z 1 IzI 5 1). Suppose f is continuous and 21rperiodic on El. t0 lim lift .fll2 = 0. we have . N 2 0.and let Show that for every E > 0. Then where a ( d z ) is the Harr measure on S1such that a(S') = 27r.
where {fn}r=l + and lim gn = g in L Q . 4316 Prove or disprove the following statements: (i) the set of continuous functions on the interval [0. (Iowa) Solution. l]). Yes. 3 n+1 n Then as a subset of L”([O.1 ) is not 3 dense in Loo([O.11) is separable while Lw([O. Indeed. If lim f n = f in l.1] is dense in Loo([O. n E m.P {gn}T=l a sequence in Lq([O. However.11) is not. (ii) False. since the distance between any two elements in S is 1. fngn + f g in measure. 13). 11).S is of cardinal N. since .294 There is an N such that where r is fixed so that Put N n=N which is required. Let 1 1 S = {f E L”([0.f(0) = 0). S is not separable. < 2 5 .l]) (with Lebesgue measure) and = 1. (ii) Lw([O. C([O. 1 ) I f ( z ) = 0 or 1.l]). 4317 Let be a sequence in LP([O. Since C([O. (i) False.is it true that f n g n nw nw + f g in measure? Justify. is a separable metric space. 11) ( Stanfo rd) Solution.
g ) = p(f . For f .g) is a metric on S.g) = 0 iff g) f = g in S. define Show that d ( f . g . as usual. Obviously.. S = {(equivalence class) of measurable complex functions on X}. and that if and only if fn + f in measure. (Here.295 we have for any E >0 where A is the Lebesgue measure. fn + f in this metric (Stanford) Solution. f) 2 0 and d(f.e.e. 4318 Let X be a measure space with measure p and suppose that p ( X ) < Let 00. = d(g. two measurable complex functions are equivalent if they agree a. we have &  0 a. h E S. iff . d(f.1 For f E S.
By the Dominated Convergence Theorem d(fn.fl Ifn  fl > &} =  { 2 E XI Ifn  f l 2 lE '} (O<&<l). If fn + f in measure. f ++ f(z)g(z)dz is continuous on P ( R ) and therefore there is an h E L q ( R )such that . h E L q ( R ) . By the BanachSteinhaus Theorem.s which then implies that g "2.296 Therefore d is a metric on S. 5 n and 11 2 5 n.f) & 1+E + 0 as n t m. Prove that there is a constant M such that . then  d(fn.5 Isfl. 191fl . 4319 Let g be a measurable function such that lfgl < 00 for every f E P ( B ) (fixed p >_ 1).(z)dz on P ( I Rconverges to f H . ) J pointwise. * . ( st u n f o d ) Solution. Then gn E Lq(IR) and for any f E L p ( l R ) .* L lsnfl 5 . Moreover.f ) + 0. .f) +0.. Define for each n E B a measurable function gn V gn(2) = { g!x" 5 hfl 5 if Ig(z)I else. It follows that the sequence of bounded linear ntm functionah f H J f(z)g. . f(z)g(z)dz. lim gn f = gf.s all f E Lp(IR). then %0 in measure by the equality f 1+f + { 2 EX I 1 + Ifn . If d(fn.
Suppose that f is an element of L2(S1) with the property that its Fourier coefficient is nonzero for all n E Z. or the real numbers modulo 2a). Let for any n E Z.297 4320 Let S1 denote the unit circle (the set of complex numbers with modulus one. where the integral is with respect to Lebesgue measure. g E L 2 ( R ) .x. We have ~ ~is an orthonormal basis of } ~ ~ a Therefore {f * k I k E L 2 ( S 1 ) }contains L2( S1). too f(y)g(z . .define the convolution f * g by f * 9 = J_. The convolution of two functions on S1 is f *g(a) =  2'. {xn I n €27) and therefore is dense in 4321 For functions f .Do not neglect to check the measurablity of f *!l.(O) = elne then { L2(S1).Show that the linear space { f dense in L2(S1).Y P Y . *k 1 k E L2(S1)} is (Iowa) Solution. I'" f(%7(ct  wo. a) Show that f * g E L o o ( R ) .
a) We will show that f * g is uniformly continuous. Since l(f * 9 n . 4322 I f f E L 1 ( R ) . Hint. Show that Tf is a bounded linear operator. b) Let { g n } be a sequence of L2(IR) such that gn t g and Tfgn+ h in L 2 ( R ) . 2 E lR * g = T f g . oo  YVdY) . Tfis bounded.define + (1) + i ([r If(y)l2dY) (Jt" Id. (Iowa) Solution. By (l). closed graph theorem.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). = It is evident that Tf is a linear operator. By the Closed Graph Theorem.f * g)(z)l I llfllzllgn . I r+m I = 0.Y)ldY II l f l l 2 l l s l l 2 . f * g(z) is well defined.f * g is bounded and therefore belongs to L m ( B ) . Define Tf : L 2 ( R )+ L2(BZ)by T f ( g ) f * g .9 / 1 2 we see that h = f + 0. Since by Holder's inequality If(y)g(z . Indeed. you need not check this.
have we 4323 Let f : [0. For any a with 0 < < Ilflloo. ?(() Solution. (Y let . ( Stanford) (1) In general. Show (Indiana) Solution. If f is simple.299 Prove that. there is a sequence {fn} of functions of the form (1) such that Then Letting n + +a. say +0 as + 00. llflloo > 0.1] + [0. co) be an essentially bounded function. for any f E L1(IR).
2. For any k E N . Since { l f k l } is bounded in P ( X ) {(fk. then X(E. where X is the Lebesgue measure.p(X). Prove that (Indiana) Solution. we get that 4324 0 . = [O. such that lim km > 0 there exists asubsequence of {fk}.) > 0.300 and F. If on the contrary k+m lim Ilfkflll llfk denoted {fk}. by = 0. the Dominated Convergence Theorem. Hence Letting a 1 llflloo. I} of {lfkl Ik E m}. p ) be a measure space for which p ( X ) < 00. M . 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 . .l]\E. also fill > 0. 0 Let ( X . and Lp(x) reflexive there exists a subsequence is converging to If1 weakly in I.
Then Letting E + 0. . .2. we get that a contradiction. 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.301 By the VitaliHahnSaks Theorem. for any that for k = 1.. E > 0 there exists a 6 > 0 such r r whenever p ( E ) < 6. a.
. Again by (1) f is Lipschitz and therefore absolutely continuous. y).e. 1 + lR satisfy the inequality 3 f(Y) . However. So lim g(z.1] x [O. i.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. z) then 4(z) 4 is nondecreasing and L !J(Y. y) Y+X # 4(z) exists except in a countable set.f(. Observe that (*) is equivalent to (Iowa) Solution. . u 5 z.13 + lR and g : [0. Show that lim g(z. y) = +(z) exists except in a countable Y+X + set and that for 0 5 z 5 y 5 1 we have Hint.f’(z) = 4(z) whenever # is continuous at z.302 SECTION 4 DIFFERENTIAL 440 1 Let f : [0. since by (l). Y E (071)). Let #(z) = g(z.) L g(Y1 Z)(Y  ( for all 2. v) 5 g(z. v 5 y g(u. (*) Assume also that g is nondecreasing in each variable.
1 1 f) ( . with for all 2. (11) fn(x) + f ( x ) for all 1 E [0. 1 1 For any n E A .1]. There are several different ways to do this problem.x E [0. Since I ‘ z I f() such that and = f(0) + i= 2 f’(x)dz. E [O. : Hint. then f is differentiable almost everywhere and moreover for any x E [O.303 4402 Prove that a function f : [0. If f is Lipschitz. (Stanford) Solution.1] there is a sequence { g n } of continuous functions Ign(z)l rl IM.1]. by the Mean Value Theorem This completes the proof. one is to use the Fundamental Theorem of Calculus. 11. define T which is required. 5 M. 1 + R such that 1 (I) [fA(x)l I M for all x E [O.1] + R is Lipschitz. Conversely.y E [0. if and only if there is a sequence of continuously differentiable functions f n : [O. .
1]. 3 Show that f is absolutely continuous on [0.1] such that (a) f(z)= n=l C 00 fn(z) converges for every 2 E [O. (nlinois) Solution. 1 . 4404 Assume that f E A C ( I ) for every I c R. (ii) f ( z ) + 0 as 121 + 00. SR . Then 1 It follows that f is absolutely continuous.for every I c R. 1 . Since f E A C ( I ) .304 4403 Let {fn} be a sequence of absolutely continuous realvalued functions on [0. If both f and f’ are in L1(R) show that (i) f’ = 0. (IndianaPurdue) Solution. Let n=l in L1\O.
1 ) such that 1 f n + f uniformly on [0. .f(O) = f(o). must have lim f ( z ) = 0. 0 f’(z)dz + I’” f ’ ( z ) d z = 0. Therefore we f’(z)dz = Iim f(z) . 11).l]). If {f. Thus So there exists an f E C([O. L’ which implies f E AC( [0. there exists g E L1 such that f : + g.305 Since f’ E L1. Then Moreover. Since x++w f E L1.fn(0) = 0 for every n. Since fn E AC( [0.13. 4405 Let { f n } C AC([O.1].1 ) such that fn + f uniformly on [0. we have lim f(z)= 0 and x+w Thus L f‘(z)dz = J_. l]).show that there is f E AC([O. which means f x+w x++w lim f(z)exists. lim x+w sox’ ( t ) d t exists. In the same way. fn(z) = 1’ fXW.}is Cauchy ( L l ) . 1 ( IndianaPurdue) Solution.
Take 6 > 0 such that If(bj)f(aj)l 5 E for any Ij = [ u j . Suppose that 22 . 4407 Let Gn. c.. ..b j ] } c I with (IndianaPurdue) Solution. Take N such that f 5 < 6.2 1 < 5. So we have 4< N(z2 . 2 . 2 1 < 2 2 . such that Then N l f ( z 1 ) . < C N = 2 2 . I f ( 4 . lGnl I .2 E I .1)1 I Therefore we find a Lipschitz constant 1 $.21) < 5.1 < 6. (2) 3 (1).1] such that GI 3 G 2 3 * * * .c. Show that f is Lipschitz. (1) f : I + R is Lipschitz (2) E > 0 3 35 = S ( E ) > 0 such that { I j = [ a j . (b) Show that the following two statements are equivalent. be open subsets of [0. 2 (b) (1) =+ (2) obviously. (a) For any zl. I. bj] c IIj I 5 6. Then Suppose that 22 . 2 2 .2 1 2 6. :2n 1 Let . Take N 2 0. n = 1 . Take { c j } such that x1 = cg < c1 < .f(zl)l < 1. . For any 2 1 .f(.306 4406 (a) Assume that f E A C ( I ) and f’ E L”(I). E I . we give the following fact.
then < & + N C(bi U i ) < & + . n 2 no.11).11).f(x’)l = C IG. a < b. (ii) If(x’) .307 Show that AC([O. Set 6 = &. b E (x . b. Gkt # 0. If(x”) . + a .a i ) < 6. suppose that for any natural number K . Solution. n [x’.al. x’. there is k’ > K . . i=l n=l K n=l K 1 = Klb. (i) For any (9 f 5 MIX’. x Then + 6) c GK = K Gi. Then .1] iff there exists no such that (IndianaPurdue) E > 0.6. &  2 which means f E AC([O.x 6) c GK.x‘’]~5 no1x” 1 no Conversely. take N such that C N+1 M & < 5. Take a. If {(ailbi)} is a sequence of disjoint open intervals in [0.S. Then GK # 0. for > x’.1] and C ( b i . 6 > 0 such that (x .f(x”)I Gn = 0.2’1.= &. (ii) Suppose that there exists no such that Gn = 2‘‘ 2 2 8 for n 2 n o . x” E [0.x”1. Take x E GK.
308 4408 For what values ‘a’ and ‘b’ is the function (i) of bounded variation in (1. (ii) defferentiable a t ‘0’. for any of bounded variation on [ E . The function f is of bounded variation if and only if To show (l). since f is continuously differentiable on (0.let us first establish the following equality provided that either side is finite.l). 1 and 1 E > 0. f is Then 1 E 1 1 . (Iowa) Solution.1]. Indeed.
If a + b 5 0. Case 111. there is a 6 that 1bl2 Let N = > 0 such that 0 < x 5 6 implies d3 . b = 0. b < 0. 0<€<1 00 E Case 11. while a 2 0 implies that sup O<e<l J' lf'(x)ldx = sup J' . Then sup 0<€<1 If (&)I < 00 iff a 2 0.309 Case I. since and .:l!al € dx < oc).> Ibl xb . Then sup O<E<' If(&)I < iff a + b 2 0. 1& + 1. for Since a + 6 = 0 and a + b > 0 imply respectively that and ' the function f is integrable on (0. Then b1 If a + b < 0.2 I4 .1]. b > 0.
] Then f must be constant.27r]. Then f must be constant. (ii) f is differentiable a t 0 if and only if { %>l or { b<o a > 1. a) The conclusion is ture.) f is absolutely continuous. = +m.f(y)l <: 1 . (Stanford) Solution.y E [0.yI.e.27r] satisfying .310 I’ and If’(. b) Let f be a real valued function of bounded variation on [ 0 .)ld. 4409 Prove or disprove a) Let f be a real function on [0. we have rl If a + b > 0. Since by the condition that If(. all 2 2 . . 2 ~ with f’(z) = 0 a.27r] and f’(z) = 0 a. then and and therefore f’ is integrable.e.. on [0. f is constant.
4410 Let f . that f and g are realvalued. 27~1.e. c such that f ( u + t )= J. it is differentiable almost everywhere by Lebesgue’s Theorem. g E L1(IR). (S t a n f o d ) Solution. without loss of generality. 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. Since the function is of bounded variation. T ) and f ( x ) = 1 on [T.e. Show a1 b) if a+h h+O f(z)dz = c. Can you deduce that f’(x) = g(z) a. Then f is of bounded variation with f’(z) = 0 a.e.31 1 b) is false. Assume. Let f(x) = 0 on [0. c ) if then there are constants a .( t ) d t cc oo .? State explicitly the theorems you use. a+t g(z)dz +c a.. but f is not constant.
f ( x ) d Z h = f ( a + t ) .c a. there is an a E BZ and a c such that I ra+h J a k. c) By a). . we see that and therefore b) We have by a) h0 lim f(' a + h. f ( z ) d x = c.e. t E IR.312 We have Letting c t 0. .
(2) Show that the smallest constant M in (b) is equal to Ilfll.b) + Dl be measurable. < 2 of ( a . a.e. t E IR.. 1.313 By b) and by the assumption.b).. (b) There is an M > 0 such that for any finite partition zo < z 1 < .e.’. 4411 Let F : ( a . (1) Prove that the following two statements are equivalent: (a) There is an f E L2(a. It follows that f’(z) = g(z) a. we have = 0.e..b) such that F(z)= lz fdm ( m the Lebesgue measure). . (Iowa) .
& ) (A. 2 E (a. We will show that if E l .b).b ) . .b ) E such that f F(x) = LZ f(t)dt + c.b). and therefore there is an integrable function f : ( a . say Ei = 3EAi . .. o ( a i j .En are any mutually disjoint measurable sets. Since a F is absolutely continuous. .we have on)of ( a . then for any E > 0 If Ei's are open sets of ( a . n E lN. a t most countable). (1) (a) 3 (b).314 Solution.. for any mutually disjoint intervals (a. By the CauchySchwarz inequality (b) 3 (a) Obviously. .
In general. ) . there is for each i an increasing sequence { E . I 65 f ( z ) < 1f ( z )5 and E.a i j 5 M. . and lim p ( E . . . j } E 1of compact sets such that j=1 u Eij 00 E. .315 then by inequality (2) I 2 00 It:.Enj and passing to the limits we see that (3) holds in general. . Applying (3) to the 300 mutually disjoint compact sets Elj.=I j € A i P i j . Applying (3) to the mutually disjoint open sets Elj. En. . Let E = {z ’ . it suffices to show that 00 where En = {z In 5 If(z)12 < n + 1). . If Ei’s are compact sets there is for each i a decreasing sequence { E i j } E 1 of open sets of ( a . b) such that Ei = j=1 Eij and Eil n E k l = 0 (i # k ) . . . . = {x Then M I < 6}. . f I 2 . . Enj and passing to the limits we see that (3) holds for E l . j ) = m ( E . To show that f is square integrable.
For any E (zip'.b]. It follows that sequence {fn}rz1 b] such that 11 of C[a. then 11. z i ) (xn = a k ( b .b ) there is a fll2 f 0.a ) ) such that + Then Illf1I2. (2) Define. b )by (3). there is is a norm on L 2 ( a . IlfII = nlimO llfnll = 40 llfll = Ilfll2 for any f E C[a.11 f E C[a. so We have Ilfnll2 = Ilf112.b ) . For any fn  f E L 2 ( a . for each f E L 2 ( a .316 5 M. Moreover 5 l l f l l n by (1). . This completes the proof.b].
(Iowa) Solution. For any Borel subset E of X . M the calgebra of Borel subsets of X. A. . Justify.Y and q5 on M by 1 E Ida and $ ( E )= IdP.e. 2 p ( E )= and ] 2tdt r E + ( E )= We have 1 E 3t2dt. t E E e X(E) = 0 e q5(E) = o . p(E)= 0 It follows that p 4  e 2t = 0 a.1].317 4412 p(t) = t 3 and define measures p ( E )= Let X = [0. We have and 4413 Let X be the Lebesgue measure on R2 and R defined by: 2 /I and v the Borel measures on .e. t E E @ X(E) = 0 e 3 t 2 = 0 a. Let a ( t ) = t 2 . Does 4~ exist? Does exist? Compute the value of the RadonNikodym d? derivatives that exist. where X the Lebesgue measure.
i. Because p ( A ) = 0 implies for all n E N .e. << p2 or/and p2 << pl? (Iowa) . let A = 21 and then v ( A )= 0 but p ( A ) = We show next that (g. X((n. m]. Since and we have " 1 du n=l and dv d X Therefore (1) holds. (Iowa) Solution.318 Is p << v or u << p? Find the corresponding derivaties if they exist. i. It is true that v << p and it is false that p << v.pn) n A ) = 0. 4414 Let p be the Lebesgue measure on [0. however.. Is p If so. find the corresponding derivatives. = n=l c " 1 $~(n. v ( A ) = 0.2n)n A ) = 0 and therefore p ( ( n .in]. m]. Define for any Lebesgue measurable subset E of [0.
Moreover From p2([0.)) E L(lR) for all t E R). that q5(tg(.. and assume g E L(lR). t E IR. .1)) = 0 and p([O.e. continuously differentiable function with a R R bounded derivative. l R a) What additional assumption on 4 will insure that f is well defined? (i.1)) = 1. Define f(t) = J dJ(tg(z))dz. 4415 Let C#J : l + l be a bounded. p << p2 does not hold.39 1 Solution. Since r 1 we have It follows that both p1 << p2 and p 2 << p1.
) = 1 . I). There are sequences {a. nco < 1. 4416 Let ( f k } denote a sequence of nondecreasing functions defined on ( 0 .} such that 0 < a.} and {b. < b.320 b) Under the additional assumption in a) above. and The conclusion follows from the Dominated Convergence Theorem. for any s. i. f’(t) exists for all t E IR. Prove that . Assume 4(0) = 0 then f is well defined. (Indiana) Solution.l). Indeed. lim (bn .. t E R (s # t). z there exists an s with 0 < s < 1 such that Since b.e. for any t . a. show that f is differentiable. l ) with the property that lim f k ( Z ) = 1 for almost every z E (0. We will show that Indeed. (Indiana) Solution.a.fi(z) = O for almost every lim krn k+oo z E (0.
l). lim fL(z) = 0 for almost every z E (0. is integrable and I’ ks Hence k+co f i ( z ) d z = 0.321 and k+m lim fk(a. .) = lim kr 00 fk(bn) =1 It follows that lim krm f.
Show that is countable.). For any r UIa\UI. 4502 Let { I a } . 51 (IndianaPardae) Solution.Q E r.?.I for every finite subset {!I. We have u ma 00 : = 1 z € UI%P. Obviously u1 : aEr is an open set. Take n such that but : . . Show that S is countable.322 SECTION 5 MISCELLANEOUS PROBLEMS 4501 Let S c R be a set of real numbers with the property that IS1+. be a collection of closed intervals in R. n=l where cyn.+S. there must be a (Y such that m x E la. Sn}c S. . S n (n. Since S n (i.pnTU I:. . . i) be finite and n ) and must S is countable. x can not be in 1. ( IndianaPurdue) Solution.". . .
since {Ic: . mutually disjoint.which is a contradiction. Thus we must have x = on or P. Let 2 A = {x E (0. Let {Ui I i E I } be such an infinite set. lim sin(nx) = 1.323 If I . ( st anford) Solution.is irrational}. 7F Then A is a measurable set of measure 27r. i H ai of I to D is injective and therefore I is countable. Therefore a a n which implies uIa\ u1 : a a is countable. 4504 Prove that for almost every x E [0. For each i E I there is an ai E D such that a E Ui.27r] n+co lim sin(nz) = 1.. for any x E A . It follows that the map. 6 Icy.} o f Z such that nm 2 1 lirn (kn.27r) nm I . 4503 Show that any infinite set of nonempty.21.) = . Indeed for any x E A . open sets in a separable metric space X is countable. 7~ 2 .where D is a countable dense subset of X . c ( a n . Moreover. P n )then x .21 I Ic. there are sequences {kn} and {I. (Stanford) Solution. Then ai # aj i whenever i # j .. I E a } is a dense subgroup of lR.
For any n. {k. .2 1 I k. Show that there is a sequence of odd polynomials p n ( z ) with pn + f uniformly on [O.} admits a subsequence {k.1] iff f(0) = 0. Suppose that there is a sequence of odd polynomials pn with p .} either increasing to t 0 0 or decreasing to m. ~ )= 1. ( IndianaPurdue) Solution. + f .324 Since f 4 { k : . + 1)w) = 1.2 1 .I EZ}. ( IndianaPurdzle) Solution. + 1))= 2 and therefore n+m lirn sin(3kAz) = lirn sin(3khx ncc + 2(31. Then f ( O ) = lim pn(O) = 0. nw .'for every z E I . 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 . n+cc Otherwise lirn (3kA) = +co and ncc lirn ((3kh)n+cc 2 T 1 + 2(31. = $00 then nco ncc lim sin(khz) = lim sin(khz . 4505 Let f E C ( I ) .11). If lirn Ic. take a polynomial pn such that 4506 Let f E C([O.
(1) . a) For any f and g in C([O.325 Conversely. ex + zf(x2). b) By the Banach’s Theorem. (Stanford) 1. We have uniformly on [0. Then f is a continuous function on [1.1 + C[O.1]. + f uniformly on [1. Then Pn is an odd polynomial for any n. 12 f(t)dt. b) Show that there exists one and only one smooth function f on [0.1 1 1 1 r+’ is a contracting mapping on C[O.1 such 1 that I(f) = f. there is only one function f E C[O.1] satisfying the conditions: { f(0)( z=) = &f Solution. 11. f(z) = es + f J. suppose that f(0) = 0. 2 E [1.0]. Set f ( z ) = f(z).l]). Take a sequence of polynomials p .e. ( z ) with p .1]. Set 1 pn(x) = [pn(x) 2 ~n(x)]. i.. 4507 a) Show that the mapping I : C[O. with the supremum norm. I].
2 . lKnl Show K . Thus Take 6 > 0 such that lf(z . . If g is another smooth function satisfying (2). K .e.g = f. For any E > 0. }. . take Nz such that . (ii) ~ K . ..3.61 + O as n + 00 for all 6 > 0. 3 .+ o o . (UC.)l < 4M & E R any IyI < 6. .326 By (1) f is smooth and satisfies the conditions: f(0) = 1. (iii) J12:121. by the uniqueness of f. then i. by (ii) there is an N1 such that If n > N1. .. 4508 Given f : R + R bounded and uniformly continuous and {K .2. For the above 6. n = 1 . * f + f uniformly.+ l a s n . E L1 such that (i) llKnlll 5 M < 00.. I ( g ) = g . n = 1.. . Take M1 > 0 such that If(z)l < M I for all z E R. .Y) holds for all 3:  f(. Irvine) Solution.
then m ( z ) = aon I.327 & < . (c) Show that S is nowhere dense. (a) For any zo E S". Let S = { z I f(z) .+ M . (b) Show: If I is an open interval contained in S.N2). (UC.m ( z ) 2 1)... It follows from (l). there exists E > 0 such that f(z0)  m(z0) < 1 . . co) be upper semicontinuous and define m : R [oo. .oo) m ( z )= liminf f(y). Irvine) Solution. by YX + (a) Show S is closed.E < 1.&z . & 4 4M 2 Set N = max(N1.(2) that < holds for all z E R if n & & +=& 2 2 >N 4509 Let f : R + (03.
Take an open interval I1 such that 7 c A l . i.d ) is a metric space.o) + . Al = {Z E I.) < 1). () (c) Suppose that S is not nowhere dense. (Stanford) Solution. In the 1 same way. there is an open interval I c S.l . c I n ..e. .6). i. . . By the definition of rn(z). Thus f) ( . we can take an open interval I2 satisfying 7 2 c I1 and f(z) < 2 for z E 7 2 . f(z)< n for z E 7 Obviously f ( z ) = oo for 2 E n . Define another metric d o n X by It is easy to show that the identity mapping of X is a homeomorphism of ( X . As in (a).& & holds for z E O(z0. .6). we obtain a sequence of open intervals { I n } such that 7.6) such that f(5)< rn(z0) Therefore f: . 6) c I . Z ) ..328 Take 6 > 0 such that for 2 E O(z0. 2 E S". (b) Suppose that there is zo E I such that rn(z0) > m.m(2)< f(20) + .for 2 E O(z0. ? which is a contradiction.. we can find 6 > 0 such that m ( z ) > rn(z0).e. Denote i + i.f(. From (b).m(.m(5)< 1. m ( z )= m for z E I .d ) to ( X . So S" is open.S).there must be F E O(z0. By induction.< 1 2 2 & & +& = 1 if z E O(zo. Since S is closed. Therefore m ( z )= liminf f ( y ) Y+Z 2 rn(zo). 4510 Prove that any topological metric space is homeomorphic to a bounded metric space. 5 6 S which contradicts % E I c S. A1 is a nonempty open set. Suppose that ( X .
~ ( z . periodic of order 3. l ) and (0. too. n there is a subsequence {Z~. is a periodic map of order 3. z) E M . + t B be a continuous convex 00. (ii) Let M be the 2dimensional sphere M = { ( z .z) = (e. y). a distance nonincreasing periodic map of order 3. z) E C x 2R The isometry A : M +M I 1 . i. z).. which has exactly two fixed points.e. suppose A : M + M is . Therefore A is an isometry and therefore continuous. ~ 1+ z2 = 13. ) = +00.}F=~ that such krco l/znk (1 f +m. If SUP llznll = $00. (Stanford) Solution. having exactly two fixed points ( 0 . (ii) Give a n example of a complete metric space M and an isometry A on M . 1). Prove that f attains a (SVNY. A' is distance nonincreasing.Ay) 5 d ( z . then a = lim f ( z n . (i) Show that A is a continuous bijective isometry.329 4511 Let M be a metric space with distance function d. 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. (i) Since A' = A 2 . A o A o A = Id and d ( A z .*'z. defined by A ( z . Stony Brook) Solution. . Let (Y = inf f ( H ) .
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.) (Stanford) Solution. y ) = (z.Riesz Theorem there is a subsequence {fnk} converging to f almost everywhere.. f ( y ) < .. on aZ. ball of H .y)I llYll51 114151 5 IlAll . ( StunfoTd) Solution. 1 k >N} W = coV{znk I k > N } 11. there is a subsequence { z n k } For any p > a there is an N E lT such that for any k > N . (b) Show that AA* . Prove that A is closed in the norm topology of L1(lR). A*y) for all z. exists in L1(BZ). Then n+m {fn} converges to f in measure and therefore by F.e. where IlAll is the norm of A. It follows that I ( ) 5 fzI 1 a. (a) Show that IIA*ll = IIAll. on lR.330 a contradiction.B.11 Therefore and for any y E cov{z. one has f ( z ) 5 p. So A is closed. f ( x n k ) < 0. 4513 fzI A is the subset of L1(BZ)consisting of all functions f satistying I ( ) 5 1 a. It follows that a is finite and f attains its mininum at z. (You may wish to use (a) prove this. I k > N } .A*A cannot be the identity on H. Recall that the adjoint A* is the unique bounded linear operater on H such that ( A z .e.. By the weak compactness of the closed converging weakly to a point z E H . h Since X E n * cov{z. If {fn} is a sequence of A such that f = lim f.y E H . 4514 Let A be a bounded linear operator on Hilbert space H . So sup llzn[l < +m. (a) Indeed.
Since = m(A)m(B) > 0. B c 0. m ( B ) > contains an interval of positive Lebesgue measure. Show that IR are Lebesgue measurable. (b) First. . AA* .A*A = I then = + It follows that I)A))2 JJAJ12 1. with m(A) > 0. 4515 Suppose that A . (Indiana) Solution.33 1 and a fortiori IlAll = 11A*\\. a contradiction. Assume without less of generality that A and B are compact sets. we will show that Indeed equalities (1) follow from the following inequalities If for some bounded linear operator A on H .
. Since z H m((zB)nA) is continuous. [p]? Justify. For any n E l there is a compact subset F. . Then (1) i=l is for f = f on X.. = 00 u Fi. .332 there exists an zo E IRsuch that m((zoB)nA) > 0.) f is continuous on X. Let X . ..e. 4516 Let X be a compact Hausdorff topological space and let p be a finite regular Bore1 measure on X .) < and f is continuous on F. + l R such that b].e. of X such that p(X\F. Then the conclusion follows from the following inclusion { Z I m((z . 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. nco < i.B ) n A > 0) is nonempty and open. the set { z I m((z . Then lim f = f a..) p(X\X. n0 (Iowa) Solution. and (3) p (2 each n E lN a realvalued continuous function f : X . N Yes.B ) n A ) > 0) CA + B.
Part V Complex Analysis .
.
. vy exist at z = 0 with u. If the assertion is true. If neither of f and g is identically zero in R . and continuous on with If(iy)I E 1. (e) True. f satisfies Cauchy(a) False. which is analytic in R. then f E 0. a t z = 0. (b) if f(z) is analytic in R and has infinitely many zeros in R . continuous on with If(iy)I 5 1 (co < y < +co).. y ) = Riemann equations a t z = 0. give a counterexample or other justification. z n has radius of convergence exactly R. (e) if C a n z n has radius of convergence exactly R. (d) False. Rez > 0 } . (IndianaPurdue) Solution. and the zeros have no limit point in R. then f'(0) exists. if false. then C n 3 a . (c) True. . then both f and g have a t most countably many zeros in R. quote a relevant theorem or reason.335 SECTION 1 ANALYTIC AND HARMONIC FUNCTIONS 5101 TrueFalse. 6. (b) False. and the partials u. but f'(0) doesn't exist. (a) if f(z) = u iv is continuous at z = 0. then either f E 0 or g 0. a n+m . Because lim G = it follows from 1. (c) if f and g are analytic in R and f(z) . A counterexample is f(z) = sin&.and has zeros z = 1 n = 1.v. A counterexample is f ( z . g ( z ) E 0 in R.2. uy. But f is not identically zero in R. But f ( z ) is not bounded in R. (f) sin f i is an entire function. then If(z)I 2 1 ( z E R).g ( z ) is not identically zero in R. A counterexample is f(z) = e". Then f(z) . = vy and 1ly = 21. + m.. f is analytic in R = { z : IzI < l}. (d) if f(z) is analytic in R = { z .
(a) It is well known that the Laplacian can be written as [). When 17iz IzI = 1. What can you conclude if this holds with equality.( < 1). (c) Determine all entire function f that < lf(z)I. sin f i is not analytic at z = 0. Actually. 5102 (a) Let f(z) be a complexvalued function of a complex variable. (Stunf07d) Solution. If both f ( z ) and z f ( z ) are harmonic in a domain Q.(’fI and that which implies that f ( z ) is analytic in R. prove that f is analytic there. (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+az hence . z = 0 is a branch point of sin &. Show that If(0)l 5 u2.336 (f) False. then F ( z ) is analytic in 1.
i% is an analytic function on G. then for any closed curve c c G. Hence f (z) = c c’e‘”.=aQ  ay ax axay azay = 0. I $$I + % 5103 Let G be a region in a‘ and suppose u : G t lR is a harmonic function. we obtain If(0)l I bI2. Integrating on both sides. which implies that fIIl is also an entire function. (a) Let 2 i e Because u is a harmonic function. the integral 2 i g . we have F ( z ) f ( z ) = a. we obtain logf(z) = cz d..’ zu 1Tiz z + a .+ . Take z = 0. elel which is equivalent to When it holds with equality. we know that f has no zero in a‘. au ax au ay aZU azu So P ( x . 1+zz (c) From lf’(z)l < lf(z)I. It follows from < 1 that = c.337 which implies that IF(.y ) satisfy the CauchyRiemann equations. (a) Show that . (Indiana) Solution. (b) Show that u has a harmonic conjugate on G if and only if has a primitive (antiderivative) on G. ay2 We also have ap . hence P+iQ=i is analytic on G. y) and Q ( x .)[ 5 1 for IzI < 1. (b) If has a primtive. we have ap ax aQ ay  aZu + ax2 ~ = 0. f(2) I I < 1. where c and c’ are constants and JcI< 1.
 av au ay ax’ we know that v(x. On the contrary.__ ax dy’ . Hence we can define a singlevalued function where zo. Because av . y) on G. we have = dx dU dU + dy. if u has a harmonic conjugate v on G. Prove that u iv must be analytic on 0.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. . + (IndianaPurdue) Solution. fi = .ik and f2 = aY aY are analytic functions on a. 2 .iau Because u and v are harmonic functions. The reason lies on the fact that the real and imaginary parts of fi and f2 satisfy the CauchyRiemann equations respectively (see 5103 (a)).338 It follows that Kdx dU du +z d y =0 holds for any closed curve c c G.au _ . ay ax For any closed curve c C = ld(u Hence + iv) = 0. % . y) is a harmonic conjugate of u(x. and the integral is taken along any curve connecting zo and z in G . then dv = dx av + dy aV ax aY G . z E G.
Hence fi = & .1] x [0. Show that f is constant. Suppose that f(z f(z + 1). 5105 Let Q = [0. It follows from fi = i f 2 for z E R that for z E Q.f ( z ) is real and + i) .i a u = if 2 = i ( g . we know that f1 = if2 holds for all z E Q.i $ ) bu & when z E S. (Indiana) Solution. by Cauchy integral theorem. ax dy’ du dy av __ ax when z E S c Q.339 By the assumption of the problem.1] c a‘ be the unit square. Because the subset S has a limit point in Q. by the uniqueness theorem of analytic functions. au av _.i] 2 0 for z E [0. which implies that u + iw is analytic on Q. and let f be holomorphic in a neighborhood of Q . Because f is holomorphic on the closed unit square Q. we have As .1].f ( z ) is real and 2 0 for z E [O.
+ f(z)dz = (udz . by comparing the real and imaginary parts in the above identity. Hence f ( z ) can be analytically extended to a doubleperiodic function by f(z) = f ( z 5 1 and + 1) = f ( z + i ) . Because G ( z )has no zeros in S. then si as. On 3 n ({y = 0) U {y = 1)) we have u = 0 and dy = 0.Then G(z) is analytic in S and IG(z)I = 1 when z E 85’.prove that f = 0 everywhere on S. Let G ( z ) = e F ( * ) . If Rf = 0 on sn ({y = O } U {y = l)). Then F ( z ) is analytic in S and continuous on S. which is holomorphic in a and satisfies ‘ lf(z)I I %${If (. (Indiana) Solution.340 for 0 5 z f(1+ Yi) . For z E we choose the integral path on dS and consider the differential form f(z)dz in the integral. . 5106 Let f be continuous on the closure 3 of the unit square s = { z = z + i y E @ ’ : 0 < 2 < l . Apply the maximum modulus principle to both G ( z ) and l/G(z). 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. and let f be analytic on S. o < y < l}. Define F ( z ) = f(z)dz. where the integral is taken along any curve in which has endpoints 0 and z. Hence f ( z ) = F’(z) G 0. we obtain IG(z)I 1 for z E S. It follows from G(z) = eF(’) that F ( z ) is also a constant. Let f = u i w . which implies that G ( z ) is a constant of modulus 1.)I1 < This shows that f ( z ) must be a constant.and if If = 0 on S n ({z = 0) U ( 2 = l}). so l/G(z) is also analytic in S and 11/G(z)l = 1 when z E dS. Hence we obtain Re(f(z)dz) = 0 on dS which implies ReF(z) = 0 when z E dS. and on sn ({z = 0) u { z = 1)) we have v = 0 and dz = 0.vdy) + i(wdz + u d y ) .f ( Y 4 2 0 for 0 I y 5 1.
(a) As 241 lim(z . we obtain that f ( z ) has no singular point in { z : IzI 5 l}.4 22 1 + + 22 + z .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. which shows that z = 1 is the only zero of z 4 + z 3 z2 + z . . By (a). (Iowa) Solution.+ +1 + z . .1) = lim =  2 4 1 (z4 .4 with residue equal t o z Hence when c = f ( z ) is holomorphic in a neighborhood of z = 1. lz4 + z3 + + z2 . .41 2 4 . we have A. &. A “partial fractions” expression of $ has the form: . zd.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 . .4 in { z : I z J5 l}. hence f ( z ) is holomorphic on an open set containing { z : IzI 5 I}.lz4 + 2 + + z~ 2 4 z2 144  lz1 3 . + 5108 Let P ( z ) be a polynomial of degree d with simple roots z1. z 2 . (b) Show that the function f is holomorphic on an open set containing the closed disk { z : IzI 5 1). (b) When IzI 5 1. and the equalities hold if and only if z = 1.lz12  lz1 1 0.
a t z = z.) . (b) Show that 2 really has a representation of the form (*). P(.zn  1 P(z) P’(zn)’ which is the residue of 2. () (c) Give a formula similar to (*) that works when z1 = z2 but all other roots are simple.zm z . f(z) is I 4 0 0 identically equal to zero.) Solution. p. By Liouville’s theorem. (4 c n = lim z+. hence (c) Denote the Taylor expansion of P ( z ) at z = z1(= z2) by P(2)= c 00 a n ( z .Z1)n. Then the Laurent expansion of 2 a t z = z1 is P ( Z ) Hence 1 the form: has P(.) (b) Let Then f ( z ) is analytic o n c and lim f ( z ) = 0.342 (a) Give a direct formula for cn in terms of P.. (Courant Inst.
Suppose g ( z ) B1. 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.a ) ( l . B = (ZAl + a z l ) + B1(1+ (at2)E lR. TI < )Z . ... Choose c1 > 0 sufficiently small.z l ( 5 rl}. 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. such that C1 = { Z. j=1 n (Indiana) Solution.Aizz 1z z.a ) ( l .Tiz) * Solution. g ( z ) can be extended to an analytic function on the Riemann sphere hence g ( z ) must be a constant..ZB1. K2.) = ‘(z . Assume that the residue o f f at z = a is Al. Kn be pairwise disjoint disks in C. 5110 Let K1. . then B1 is real and  Az2 + Bz + z ( z . . n n f1.Tia)’ where A = 2 1 . and j=1 U Kj. Define g(z) = f ( z )   Ai za . Assume K1 = { z .. If f ( a D ) C IR.~1 < + ~ 1 C ) C\ UKj j= 1 n * . It is obvious that g ( z ) is analytic on D and g ( a D ) C lR. By the reflection principle. ( a . fz. show that there is a complex constant A and a real constant B such that Az2 + Bz + z (Indiana) f(. .
f ( z ) .f ( z ) has the Laurent expansion f(z) = c +m up(z .fl(z)fZ(Z) . consisting . @ u {m}}. Repeat the above procedure n . and the above identity holds for z E @\ U Kj.f n . such Laurent expansion 0 Set f2(z) = k=m uk ( z .fl(z) is analytic in @\ Assume K 2 u Kj. fl(z)f2(z) has an analytic continuation to in @\ f(z)fl(z)f2(z) is analytic ij j=3 ~ j . Because f ( z ) .1 times. Choose ~2 n = { z . Set Then we have n j=1 where fj(z) is analytic in@\Kj. we get a function f ( z ) .344 In C1.221 5 72).z1)5 fl(z) is analytic in @\K1. 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). Because f(z) K2.~ (2) 2 ) f2(z) ~ .fi(z) has an analytic continuation j=2 to K1.* * * . is analytic in@\KZ. which is analytic in @\Kn.I)? k=m Set fl(Z) = c 0 k=oo u p ( z . Iz . > 0 sufficiently small.l ( z ) .
n) be all the zeros and poles of f(z) with multiplicities npirespectively.p i ) n p i and m Then n n n m and m m m n . C npi= 0. is a zero of f when npi> 0 and a pole of f when npi < 0. let q j ( j = 1 . . Stony Brook) Solution. Let pi (i = 1.m) be all the zeros and poles of g ( z ) m with multiplicities mqjrespectively. . i=l n Similarly. Let f and g be two rational functions with disjoint divisors. we have  1 . By the property of rational functions.2. . Prove that (SUNY. 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. then we have j=1 mqj= 0. then f n f ( z ) = A JJ(z .345 counted with their multiplicities np E 2. It should be noted that p. 2 . . .
346 In case the point 00 is a zero or a pole of f or g. 3i. Hence n. simplifying the (a) Find the Taylor polynomial of f of degree 2 a t 4 coefficients. we may assume p . + + .) j=1 = g(m) = B . j=1 Since C mpj = 0. (b) Find the radius of convergence R of the Taylor series T ( z ) of f ( z ) at 4 3i. we may assume that g(p. = without loss of generality. 5112 Let f ( z ) be the "branch" of logz defined off the negative real axis so that f( 1) = 0.1 i=l n1 m and i = l j=1 which completes the proof of the problem. Then n1 00 f(z) = A n ( z i=l and g(z) m =B n(~ m  4j)"'j.
5. Hence + + + + + + .Imz < 0 } . is this something special to this example. T ( z ) # f(z).3i) + cg(z + 4 . It is because T ( z ) in R is the continuation of logz at 4 3i in the disk { z : Iz 4 .3i( < 5. Because log z has only z = 0 and z = 00 as its branch points. the radius of convergence R of T ( z ) is equal to the distance between z = 4 3i and z = 0.347 (c) Identify on a picture any points z where T ( z ) converges but T ( z ) # f(z).arcsin ).1 by a...01. (c) Denote the shaded domain shown in Fig. When z E Cl = { z : Iz 4 . 5 4 =  + 3i 25 ' 25 24i 1250 ' and c2 = + (b) Denote the Taylor series of f(z) at 4 3i by T ( z ) .4 + 3 i I ( ) 2 +. or a general impossibility? Explain and/or give examples.* Hence the Taylor polynomial of co f of degree 2 at 4 + 3i is + c1(z + 4 .3il < 5}. (a) When z is in the neighborhood of zo = 4 f ( z ) = logz + 3i. Hence R = 5.3i)] z+43i log( 4 + 3i) + log 1 + [3 z4+ 4+3ii 3 log 5 + i(7r .arcsin ) + 5 4+3i 1 z+43i 5 .3 4 2 . and has no other singular point. If there are no such points. (Minnesota) Solution. and describe the relationship between f and T at such points. co = log 5 c1 where 3 + i(7r . we have = l0g[(4 + 3i) + (Z + 4 . while f(z) in R is the continuation of log z a t 4 3i in the slit plane @\(m.
2). We have Arargf(z) 1 1 1 + Ararg(z + 1) = (27~)+ 2( . Express g in terms of f .41 < 4) so that f ( z ) = z i ( z + 1)i in A and f ( z ) is positive for 0 < 2 < 8. An analytic function g in A is obtained from f by analytic continuation along the path starting and ending at z = 4 (see Fig. and denote the change of $ ( z ) when z goes along r from the start point to the end point by A.. i.1 5113 Let f be the analytic function defined in the disk A = { z : Iz . f. T(z) = f ( z ) + 27ri.$(. Then g(z) = Ig(z)leia%dz) 1 = Arargz 3 = ~() I.2 ~ ) 3 2 Hence .2 by I?.e.).348 the difference is 2ai.5.5.5. Denote the closed path in Fig.5. Fig.l(argf(z)+Arargf(zr)).2 Solution. (Indiana) Fig.
.=25 (Indiana) Solution.l ) ~ ( e = ~ enn). f i is changed to fi. It is obvious that z = n2R2 (n = 1 . . : (b) We have which shows that z = n2r2are simple poles of f ( z ) with residues 2 n ~ ( . 4x. When z is in the small neighborhood of z = 0. Hence f ( z ) is singlevalued and analytic inC\{z = n2r2 n = 1 .6 .349 5114 Define eJ. Let r = { z : IzI = T } . Hence z = 0 and z = 00 are no longer the branch points of f(z). 2 .e& sin(&)  e f i . and when z goes along I’ once in the counterclockwise sense. . 2 . (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 ) . (a) It is known that z = 0 and z = 00 are the branch points of function &. sin f i which is still f(z). efi f(z) = (a) Where is sin& * f singlevalued and analytic? (b) Classify the singularities o f f . .) are poles of f ( z ) . . . while f(z) is changed to e . .
and hence is a nonisolated singular point of f(z). Which of these branches are (is) the derivative of a singlevalued holomorphic function in 0. y = 0) not inside r 5 z 5 1 .{+log(l+z)+log(ll)} Because and 1 Ar[argz .r#J(z)the change of r#J(z)when z goes continuously along I counterclockwise once. Why? (IndianaPurdue) Solution.z ) ] = 2 2 5 z 5 1 . Hence f(z)dz = 2 ~ i R e s ( fr 2 ) .350 As to z = 00. y = 0) inside I? . It is ' known that f and g can be represented by and  g(z)  4 1  . it is the limit point of the poles of f(z). y = 0) deleted. Let I? be an arbitrary simple closed curve in a.arg(l . For (b) g ( z ) = can we choose singlevalued branches which are holomorphic in fl. 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 . (c) f ( z ) has only one pole z = 7r2 in the disk { z : IzI < 25).arg(l 2 +z) 0 27r  1 arg(l 2 (1 (1  z)] =0 1 1 Ar[arg(l + z ) . and denote by A.
el5 E (e4A. Because . By lifting the integration to a n appropriate covering space. If the segment (1 5 z 5 1. The singlevalued branch of logz is defined by loge = 1. e2=). Find log el5. If the segment (1 5 z 5 . and the primitives are f(z)dz c. 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 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 both f ( z ) and g ( z ) have singlevalued branches which are holomorphic in Q. d z (HUT7NZrd) Solution. Justify your answer. 1. Hence loge15 = loge A r l o g z . Let log be a branch of the logarithm on D such that loge = 1. Hence log el5 = 1+ (14 4ai) = 15 4ai. f(z) g(2) = *i(l z  )5 22 1 1 = f i (1 + a2 + 7 + * * 7 a4 . z 23 zi It follows that f(z)dz = 0 and lrg(z)dz = f 2 a . . and each of f and g has two singlevalued branches.y = 0) is not inside I?.*2.1 = 14.. we consider the integrals 1 f(z)dz and 1 g(z)dz.351 we have A r f ( z ) = 0 and Apg(z) = 0. e6=). oriented counterclockwise. But neither of the branches of g is the derivative of a singlevalued holomorphic function in a. and A r l o g IzI = 15 . i = *(1 )T 22 1 1 1 b3 b5 =fi . (b) Let y denote the unit circle. Because e E (e'. we know that when I connects e and el5 in D.+ T +.+ . ' Arargz must be 4a. 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.) . 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. where the integral is taken along any curve connecting zo and z in 0. It follows that A r l o g z = Aplog IzI + iArargz.*1. we consider the Laurent expansion of f and g about z = 03: s . Hence we obtain that both of the singlevalued branches of f are the derivatives of singlevalued holomorphic functions in Q. In order t o know which of f and g has a singlevalued primitive in R.y = 0) is inside I?.. (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.
+ . we have reason to assume that the function f (with real variable t ) has continuous derivatives f’(t) and f”(t). has a singlevalued harmonic conjugate in C\O. Hence the most general harmonic function of the form f(1). + 5117 Find the most general harmonic function of the form f(lzI). and the solution is f ( t ) = Q log t + 0. 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}. 1 ) is harmonic. the segment on wplane can be denoted by [it. it must be a constant.47r2) . 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 . Since log IzI has no singlevalued harmonic conjugate inC\O. Because f ( I . z E C\O. Which of these f(I. in C\O is Q log IzI 0. where (Y.352 both the starting point of y and the singlevalued branch of log z on y can be arbitrarily chosen.(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.Pare two real constants. where t can be any real number.1) have a single valued harmonic conjugate? (Indiana) Solution. we know that when f(1).i(t+ 27r)). Note that the Laplacian A=+=4ax2 ay2 a2 a2 a2 azaz7 and we obtian where t = IzI.
By the symmetry of 0 on the rest of the boundary by domain D. Fig. 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. .. 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 ) . .3 by D . 10. . What is the value of u at the origin? Justify your claim. Denote the interior domain of the pentagram shown in Fig. . put in order of counterclockwise. k = 1 . . .5. 110. (Stanford) Solution.353 5118 Consider the regular pentagram centered at the origin in the complex plane. 2 .5. . and the ten segments of the boundary by I l l I z .
' r d . and h ( z ) is continuous o n C . is d m . Since d m is analytic in C\[O. Y<O Hence h ( z ) = 0 when z = 2. because z = 0 and z = 1 are branch points of Remark. and h : G + lR is continuous. But h ( z ) is not harmonic on&. (Indiana) Solution. Suppose that u and v are harmonic in D = { z : IzI < 1) and continuously differentiable on D u y. The answer is No. A counterexample is h ( z ) = Red. 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 .l] holomorphic. When 0 5 2 5 1. If the problem is changed to h : G + C is continuous and hl~\[o. [0. 0 5 2 5 1. h l . 5120 Let 7 be an arc of the unit circle. 11. Y>O lim z=x+yi+x d m = 4i. where the singlevalued branch of d m is chosen by = 4. Let u* be a conjugate harmonic function of u in D and v* be a conjugate harmonic function of v in D. does this implies that h is harmonic on G? (Iowa) Solution. q p l ] is harmonic.354 5119 Suppose G is a region in C. dmlz=z lim z=x+yi+z &zij=&cT)i. then h must be holomorphic on G.1] c G. h ( z ) is harmonic there. If the boundary values satisfy u = v on y and the radial derivatives satisfy = on y. prove that u = v in D.
. Hence s i U ( z ) = u(i) z1 zS1 = arg() T 2 z+l z1 .and for z E y denote the subarc of y from zo to z by y z . such that f = g on y. 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). Let zo be a fixed point on y. = lz :dB = v*(z). Then by the reflection principle.2.g. Without loss of generality.1 is a harmonic function in H and assume $1 on the negative real axis and 1 on the positive real axis.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. + + 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. Hence we obtain two functions f = u iu* and g = v iv* which are analytic in D and continuous on D Uy. Let F = f . F can be analytically extended to an analytic function on D u y U D*. 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 ~ } . Give the correct determination of any multiplevalued functions appearing in your answer. which implies u = v in D. where D* = {z : IzI > 1). It is well known that u(w) = targw . we obtain F 0 on D U y U D*. Since F = 0 on y. Then for z E y. (Courant Insf. we may assume that u * ( z o ) = v*(zo) = 0.) Solution.
argc) . Define + f ( z ) = (u(2) + i.{2arg(C .dz u.dy and A = h z l = r * d u . Let * d u = u. but may be not singlevalued. 2R A . 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). (Minnesota) Solution.z ) .1 A  2 z+l arg2. 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). This problem can be solved directly from the Poisson formula as follows: 2) 1 = A.. Denote w(z) = Jz * d u .(Z)) . where A is a real number not necessarily zero.log 2. Remark. then w(z) is the conjugate harmonic function of u ( z ) . is a harmonic function in D = { z : I1 < 1) with the boundary values $1 on t and 1 on I'2..where the singlevalued branch of arg( is defined by arg( %)lr=o = a.
. . show that . From the above discussion. let n(t)be the number of (j= 1 . Jensen's formula states that Prove this.357 then f ( z ) is a singlevalued analytic function on (0 identically zero on { IzI = 1). . for n = 1 . counted with multiplicity. Then f ( z )  g(z) = n=O c.. we finally obtain + +cc u ( z ) = Re n=m b. n. Suppose further that luj I < T for all j = I. Define g(z) = n=m C 00 bnzn.1 = l}. + . . = bn for n = 0. 2. it follows from Rebo = 0 and 1 n=cc 03 03 n=l n=l 00 n=l that Reg(z) = 0. 5 1.1.zn be the Laurent expansion of f ( z ) on (0 ncc < IzI < lim = 0 and lim 0 . hence f ( z ) = g ( z ) ia.n ) such that ( u j ( 5 t . and If(0)l = 1. = 1). (b) With the hypotheses and notations of (a).2. = bn and  n+cc lim = 0. When IzI = 1. then . satisfying 6.( < 1) and Re(f(z) .zn is an analytic function in 1. where b. we have F ( z ) i~ e i a . . 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. .zn A + log 21r Iz]. and IF(z)I = 1 on (1. 2 . . . and b.g ( z ) ) is identically zero on (1. Then g ( z ) is an analytic function on (0 < IzI < +m}. . Consider P ( z ) = e f ( ' ) . Using Jensen's formula. by the maximum and minimum modulus principles. 2 . a .too < Iz1 < 1) and Ref(z) is n=m 5  u. Let f ( z ) = l}.g ( ' ) which is analytic and does not assume zero in { IzI < l}. = a _ .. uj .
(c) Apply the identity in (b). which implies that log IF(. we have & Lzx 0<8<2R log Jf(Re2')ldB= R Denote max log If(Re")J by M ( R ) . By the definition of the function which shows that the identity holds.)/ is harmonic in 1.358 (c) For r < R deduce an estimate on n ( r ) in terms of max log If(Reie)l.( 5 r } . 0<852x (d) What can be said about the zeros of bounded holomorphic functions in the unit disc? (Hamard) Solution. . n(t)we have (b) It is obvious that log 2 = la3 I h:j. (a) Let then F ( z ) is holomorphic and has no zero in the disc ( 1 . ~ 1 5 r } . we obtain n ( r )5 M(R)/log R . $. By the mean value theorem of harmonic functions.
Apply Jensen's formula in (a) to F(z) = with 0 < T < 1 such that there is no zero of f on ( 1 . we Since f(z) is bounded..359 (d) Let f ( z ) be a bounded holomorphic function in { z : IzI < 1). We know that f(z) can have countably many zeros.. Obviously Ian1 + 1. Suppose z = 0 is a zero of f ( z ) of multiplicity m 2 0 with = a.and hence Let T + I. ~ 1 = T } . we assume For any n. and let the other zeros be ordered by 0 < lull 5 la21 I . j=1 00 . we obtain n which implies that the series C (1 .luj I) is convergent. we can choose T such that T > lun[.
5202 (a) Find a function f that conformally maps the region { z : largzl < 1) onetoone onto the region {w : 1wI < l}. z + 1 : { z : IzI < 1) z1 f (21 : Imzl < O}. 43(22) = 22” : {z2 : 1221 < 1 and Imz2 > 0} + {w : Iw( < l}\{w : Imw = 0. give an explicit map. {zi : Imzl w 1 < 0) t {z2 : 1.O 5 Rew < 1).b o n e holomorphic map from the unit disk {IzI < l} onto the slit disk {IwI < 1}\{[0. if No. explain why not.~21< 1 and Imzz > O}. Then w = 43 0 4 2 o 41(z) = f ( z ) is a o n e . (Illinois) Solution.{[0. We construct the map by the following steps: 2 1 = 41(z) = 2. Stony B Solution. 5 l o g r (log 1 = 0) is a conformal map of { z : largzl < l} onto {< : Re( > 0}. < l} ont ( S U N Y . 1(. Hence w < 5 . (b) Is it possible to require that f(1) = 0 and f(2) = $? If yes.1)}. (a) = f l ( z ) = z: = . and w = f 2 ( ( ) = is a conformal map of {< : Re< > 0) onto { w : II < l}. Show that the function you have found satisfies the required conditions.SECTION 2 GEOMETRY OF ANALYTIC FUNCTIONS 5201 Find a onetoone holomorphic map from the unit disk slit disk {IwI < 1) .1)).
0 < b < 1) conformally on the full disk {IwI < 1) by means of a function . . we have both [ F ( G ) 5 1 and I F ( w ) l 5 Iw(.y) : x 2 y2 < 1 and y < 0). 0 & ( z ) . (2) I f f is an arbitrary conformal map satisfying the condition of (l). s. d where 1 = .)( = lY(z)I for every z E { z : (argzl < 1). b.i a a is z=e 4 2 + iy : 2 > required conformal map. Then w = 4 2 o &(z) = d n + z 2 .11. Then w = F(6)= foY'(G) is a conformal map 6 of { G : 1 < 1) onto {w : II < 1) with F ( 0 ) = 0. c ._ (b) Suppose W = f ( z ) is an arbitrary conformal map of {z : largzl < 1) onto {W : 11< 1) with f(1) = 0. By w 6 Schwarz's lemma. 5204 a Map the disk { Izl < 1) with slits along the segments [a. Since    we cannot require that f(2) =  i. 2r+1 .' ( w ) 6 1 w is a conformal map of {w : II < 1) onto {W : 11 < l} with F ( 0 ) = 0. (1) Let C = 41(z) = z 2 . = . It map of {C = < + iq : 7 > 0) onto {w = u + iw : u2 + v2 < 1 and w < 0). ad . [1. where d 1 =.bc > 0. 5203 ( ( 2 . ( Toronto) + Solution. and W = F(w)= f o f . b] (0 < < 1.then 4T1 o f o 4F1(C) is a conformal map of the upper half plane onto itself.361 is a conformal map of { z : largzl < 1) onto {w : IwI < 1) with f(1) = 0 and f(2) = 2q1. ~ ): 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. It is a conformal map of { z = 0 and y > 0) onto {C = iv : 77 > 0). (2) Find all such f . ( 6 1 which implies that If(.i is a conformal a. which can be represented by $([) = where a. d E IR. Let w = 4 2 ( ( ) = (=a <+ dm+(. Hence f can be written as 4 2 o II.
42(. ( i ) z 1 = 4 1 ( z ) = z + i : { I z I < 1}\{[a. 41(b) = b . let A= L4 &+&' i.. B= 24' 6 t &+4 We also know that +i(l)= (iv) w = 44(z3) = ~3: { z 3 : Imz3 . We construct the conformal mapping by the following steps.i. > 0).a + i]} C\[O.) = 1. Compute f'(0) and the lengths of the arcs (Haruard) Solution.362 w = f(z) with f(0) = 0.a+~]}. > 0. It has the point (iii) z3 = 43(~2) = correspondences fi : C\[O. + ) It has a .b]}+~\{[b~.. (ii) z2 = 4 2 ( 4 = the point correspondences z i + ( b + ") Zl+(aP+): @\{[6 . > 0) {w : ll < 1).. It has the point correspondences #q(O) = 00. 1 + + #1(1)= 2 and &(1) = 2. f'(0) corresponding to the slits. 4 l ( a ) = a +. +m) + (z3 : Imz3 and For the convenience of computation.. and 4 2 (2) and it is easy to know that ( 4 2 0 41)'(0) = (a 1 + 1 + b + ) b < 0. It is obvious w .1]U[1.rs+i that 44(i) = 0 and d k ( i ) = f.
= 4arctg B+z B A+& 5205 denote the arc {eit : E 5 t 5 27r . > I}.363 Now we define w = f(z) = 4 4 o 4 3 o 4 2 o & ( z ) .E } and let fir be Let 0 < E < a. describe the part of the unit disc that f maps onto 1(. (Stanford) Solution.arg= 4arctgA = 4arctg A+z Ai T+& B+i B2 5 4 and I2 = xg . I f f is the conformal map of the unit disk onto a.. f'(0) > 0. We are going to find the map f by the following steps: . From the above discussion.arg Bi 1 $4 = 4arctg. f(0) = 0. we know that f maps the unit disk with slits [1. let the complement of yE in the Riemann sphere.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. b] and [a. The lengths of the two arcs are and containing 2 I1 Ai A+i = arg.
and the symmetric domain of D1 with respect t o I. @ ( z ) can be extended t o a conformal map by of fie onto {C : ICI < 1). + Fig.4).. (1)Suppose = g ( z ) is another conformal map of G onto a circular disk D. and the domain f maps onto 1. which is C } c.( < 1) with respect to arc { z = eit : It( < E } is 1(. is uniquely determined by G and a. with center C = 0 and radius T I . such that f ( a ) = 0 and If'(a)l = 1 for some point a E G..364 (shown in Fig. is D 2= (1 < 11 l}\n1. W(0) > 0. radius r . (2) Determine r ( G .which is bounded by circular arcs I. After considering the boundary correspondence. and if a = 1 i.T a in {/<I 5 1). C . 5206 Suppose that w = f ( z ) maps a simply conncted region G onetoone and conformally onto a circular disk DT with center w = 0. Hence the conformal map f in the problem is nothing but the inverse of @. such that g ( a ) = 0 and Ig'(a)l = 1. then @ maps { z : IzJ < l} conformally onto D 1 with @(O) = 0. with q53(e+a) = 0.4 Let + ( z ) = $3 o $2 o # q ( z ) . orthogonal t o = 1) and connects points e f * and e . then w = F ( C )= f o g'(C) is a conformal map of {C : 1" < r 1 ) onto {w : Iwl < r } + c . where D1 is a domain bounded by { = eie. > l}. (1) Prove that the radius r = r ( G . and {C = eie : (01 5 i}.5. a ) of D. the reflection principle. (Indianu ) Solution.5. we know that 1. a ) if G is the region between the hyperbola zy = 1 ( z > 0. arg$$(eT*) = f 5 . corresponds t o the arc { z = eit : It(< E } under the map @.( > 1) is D2. and an circular arc I. Since the symmetric domain of 1. 5 B 5 27r . y > 0) and the positive axes.
in the following steps: 2.. which implies T I = T .2. then w = f ( z ) : G f ( a ) = 0 . with 5207 Let T ( z ) = be a Mobius transformation.($ + i)l with 43(ie%") 0.e. 11#:(1+ f).z2 (b) Use (a) to find an expression for T " ( z ) . a ) = * f {w : JwI < &}.i = 1.32 (Iowa) Solution. (2) We construct a conformal map of G onto a circular disk D. Because the cross ratio is invariant under Mobius transformations.365 with F ( 0 ) = 0 and IF'(0)l = = 1. Show that there exists a constant c such that 3 T ( z ). Apply Schwarz's lemma to F(C) and we have IF'(0)l 5 hence TI 5 T . > 01.' ( w ) and we have T 5 T I . Irj5i(iegx)I = 2 = d%lret=* Define f ( z ) = 4 3 . 3 . we have . . n = 1 . l#l z1 = & ( z ) = z2 : G + { q : 0 < Imzl < 2}. For the same reason. 4 2 o 41(z). z T ( z ) . = 4 2 ( z 1 ) = eHZ1: ( 2 1 : o < Imzl < 2 ) =fey. .z 1 z = c. + (22 : Imzz with 42(: + i) = icy. Id..z2 21 . In other words. (a) Let a E a' be a point different from z1. with 41(1+ 22 f) = $ + i. Hence r ( G . T ( z i ) = zi. z2. If'(a)I = 1. i. 2 . T is uniquely determined by G and a. z2 E a' are two distinct fixed points for T ./ = A. 23 1 .if T ( z ) = . apply Schwarz's lemma to F . (a) Assume that z1.
2 .31 .3 2 we obtain T ( z ).z1 .z1 Cn . (c) Show that the transformation 20 = $(z+:) carries straight lines through the origin and circles centered at the origin into a family of confocal conics.366 which is Denoting T ( a ). while when X > b2.  2 z z1 z zZ (b) Since T " ( z ) = T(T"l(z)). (b) Prove that such confocal conics intersect orthogonally.. easy to have it is T"(3) 31 T " ( z ).(2n .~  c.22 z . the curves form . . a .Y2 b2+X 1 form a family of confocal conics".3 2 .z2 When T ( z )= by solving the equation = z. When a2 < X < b2. we assume a > b > 0.z 1 T ( z ). then T ( a ) = 5.22 T"2(z) . hence c = 51 : 21 5+1 2+1 s. (Humurd) Solution. if a t all..l ( z ) . = z . (a) Without loss of generality. T ( a ).1)z' 5208 (a) Justify the statement that "the curves 22 a2+X +.a . we obtain that z = f l are two fixed points of T..z1 = c. It follows from T"(z)l 21 = 2"Tn(z)+l z+l that T"(z) = (2" (2" + 1).z1 = czT"2(z) .2 2  T " .1) + 1) .(2" . CT"l(z) . the curves form a family of hyperbolas. Choose a = 2.
When a2 <A <4 2 . *). (b) Suppose (xO.A () When A = J(a2 + A) + [(b2 + A)] = m.1.) and which implies that the confocal conics intersect orthogonally. &).T sin 8. ) 1 . and w =u + iv = 1z + 1 ( ) 2 Z = 1 (T 2 1 i + ) cos 8 + (r 2 T .+X2  u2+Ai Y2 =1 b2+Xi and L 2 : . O ) .+X2 a2+ A2 b2+ Y2 A2 . if at all. where A1 # X 2 .yo) is t i = (&. .) + ( b 2 + A l ) ( b 2 + A. > b2.367 a family of ellipses. Noting that the tangent vector of L1 at (xo. Hence the curves 22 a 2 + A Y2 +=1 b2+A form a family of confocal conics. 4 + Al)(U2 + A. Suppose the focuses of the conics are ( ~ c ( X ) . (c) Let z = Tez'. we have (u2 Y.yo) is 7 2 = ($&.yo) is the intersection point of L 1 : . It followfrom and that = O. the tangent vector of L2 at (xo.
Hence the transformation 0). 5209 If f : D(0. The image of circles centered at the origin is which are ellipses in wplane.:)' of the ellipses are ( f l .0). Because a(r :)2 . 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. Hence w = F ( z ) = g o f ( z ) is analytic in . the focuses 1 w = z(z + ).$ ( r . Because cos2 e + sin2 e = I. the focuses of the hyperbolas are ( f l . and if IRef(z)l < 1 for all z E D(0. prove that (Indiana) Solution.The image of straight lines through the origin is which are hyperbolas in wplane.l) = { z : IzJ < 1) fa' is an analytic function which satisfies f(0) = 0. + = 1.1 carries straight lines through the origin and circles centered at the origin into a family of confocal conics. l).
we have it follows from f(0) = 0 that If'(0)l I . where 8 is a real number. is a conformal mapping of fl onto the unit disk with the origin fixed. F(C) is analytic in the unit disk such that IF(C)I < 1 and F ( 0 ) = 0. By Schwarz's lemma. 1 < Imz functions f : R +a' such that < If1 l}. which implies and the supremum is attained by f ( z ) = e"fo(z).369 D(0. I Choose (0 = w. For any analytic function w = f ( z ) : fl +a' such that If1 < 1 and f ( 0 ) = 0.)[ IF'(0)l 5 1. . By Schwarz's lemma. IF(C>l ICI. 7r 4 5210 Let R = { z E a'. Find (Indiana) Solution.l) and satisfies F ( 0 ) = 0 and IF(.we have eT+1 E . and let F be the family of all analytic < 1 on R and f(0) = 0. Because < 1. The equality holds if and only if F(C) = e"C. we consider the composite function w = F(C) = f o f{'(C).
Let R = C\{W = u + iv : u 2 0 and w = 0). By Schwarz’s lemma.2i + 2i’ f2(z) Then F is a holomorphic function on the unit disc with F ( 0 ) = 0 and IF(z)I < 1. It follows from 1 f ( z ) I < 1 1 If(z)I that f ( 0 C R. ) Set g(w) = (i f t = 2). Because w = 23 1c (that is the inverse of C = disc {C : i. 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 < 1.370 5211 Let f be an analytic function on D = { z . Then g o f ( z ) is an analytic function + + + E. F ( i ) can take any value in the 1+c is a 5 f}. w=1 I(s O f)’(O)lL 1. Let f E P and define {f(i) < = F ( z ) = f2(z). (ii) f ( 0 ) = 1 + i. s) . IzI < 1) such that f ( 0 ) = 1. which implies I F ( $ ) ]5 It should be noted that when f changes in P. Let E = : f E P}. Describe E explicitly. Prove that If’(0)l 5 4. on D with g o f ( 0 ) = 0 and 1g o f ( z ) I < 1. we obtain If’(O>l L 4. and suppose that 1 f ( z ) I < 1+ I f ( z ) I whenever J z J 1. 1 < (Indiana) Solution. By Schwarz’s lemma. we have IF(z)I 5 121. (Minnesota) Solution.
1 < arcsin .4.&)I 54 a. 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 ) = 0.2 5 3 + 4 5 0). Hence If(8)l = IF(4 . 2 5 3 1 l r 1 4 20 E = {f() : f E P } = {reio : 1 .r2sin28 8 2 4 . 3 5 $}.371 conformal mapping of { C : 5 :} onto {w : I . Then F ( z ) = f o 4 ( z ) = f ( 2 ( 5 is analytic in D and satisfies F ( 0 ) = 0 and IF(z)I 5 1. . fE3 sup If(8)l = 4  a.. Your answer should be an explicit number. If we denote the two roots of p2 . p 2 ( 4 ) where pl(q5) p z ( 4 ) and Iq5 . + 4)) a) IF(.$i w ' l the set : f E P} is equal to {f2(i) 10. f€3 and you should prove your assertion. Define w = + ( z ) = 2(5+ :). 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.)I 5 1. r 4 . By Schwarz's lemma. (Indiana) Solution.p~psind+4<0). we obtain that {w:Iwz~< Hence }={w=pe*+: 8 3 lr 4 20 Id1<arcsin. W+CC R = {w = u + i w : u 2 + 52 w2 > l}.$1 5 arcsin $.y p s i n 4 4 = 0 by PI(+). 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.
b. 1 3 is the line segment connecting z1 and 22. Since R is convex. we know that f ( z 1 ) # f ( z 2 ) whenever z1 # 2 2 . c. + bc)z + b(a + d) + d)z + d2 + bc * f(z)= C(U d2 If a + d # 0. L : z ( t ) = zl +t(zz . Now we consider the equation f ( z ) = = z . Prove that f has a unique fixed point inside D.e. So f has a unique fixed point inside D.. Hence adbc > 0 and a2+bc = d2+bcimpies f = id. (Indiana) Solution. which implies .z l ) . + bc # 0. one in the upperhalf plane and the other in the lowerhalf plane. which is equivalent to cz2 + (d . Then (a2 s.a ) z . ( S U N Y . Since A = (d . Prove that f is onetoone in 51 (i.b = 0.bc > 0. It follows from f o f = id that b(a + d) = c ( a + d) = 0 and a2 + bc = + 5215 Let fl be a convex. t E [0. we know that f ( z ) = z has two conjugate roots. L c il. open subset of a‘ and let f : R a‘ be an analytic function satisfying Ref’(z) > 0. which contradicts the condition f # id. it can be written as f ( z ) = where a. Since Ref‘(z) Jif‘(z(t))dt # 0. then b = c = 0.a ) 2 + 4bc is equal to 4bc + 4a2 < 0. we have f Hence > 0 for z E Q. Let z1 # z2 be two arbitrary points in R. Thus we have a + d = 0 and the inequality a d . z E 51. f is injective). d E IR and ad .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. S t o n y Brook) Solution.bc > 0 can be written as bc a2 < 0. Since f is a conformal map of D onto itself.
Rezj > 0 ( j = 1.l ) ~ .[ 1.z. .z 2 ) . .373 5216 Show that if the polynomial P ( z ) = a. . .z. ..22 z . It follows that 1 1 P'k) + 1 (logP(z))' = . (b) For any nonvanishing polynomial P ( z ) . show that the same is true for the roots of dP/dz. ( z . let H = { z : Rez > 0).) Solution.. It follows from the univalence of P ( z ) in { IzI < 1) that P'(z) = nu. .. is onetoone in the unit disk + + + n1 (l)nlna. ~ n1 j=1 n1 4 lal' 5 1. Since a1 = 1. . z. (a) If all roots of P ( z ) lie in H . + 2 u ~ z + u l# 0 for all z E 1. . not identically zero. . .1 be the roots of P ' ( z ) . z z . (a) Let z1. z 2 .2. Stony Brook) Solution. be the zeros of P ( z ) .n . . . IzI < 1 and a1 = 1.zl)(z .1.l).. (Courant Inst. then Izj 2 1 for j = 1 .~ + .^^^ + (n. . Because P ' ( z ) can also be written as nu.z. By assumption. 2 .). + P ( z ) z . 5217 Let P ( z ) be a polynomial on the complex plane. 5 ( S U N Y . we obtain Iul = n. we have n + > 1.+ . z. In other words. and P ( z ) = u ( z .z1)(z .. . the roots of P ' ( z ) are all situated outside the open unit disk.( < 1). j=1 zj = al. .2 2 ) ( z .zn un1zn' alz ao. . then [nu. by comparing the constant terms. 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 .l z " .. Let z1.n).21 z . ( z ..
we obtain that the zeros of P’(z) lie on the convex hull of the zeros of P ( z ) .) or equivalently. (a) Show that there exists on { z : IzI = 1) with < (b) Characterize those functions with (Minnesota) Solution. . . p(. 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.374 When z E { z : Rez 5 0}. then 0. Hence Re < arg(zzj) < $. convergent inC\{O}. . (a) Let f ( z ) = too n=cu C bnzn. with residue b at z = 0. be the zeros of P ( z ) .z j ) ) < 0. 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. (b) Let z1. then Hence . Denote the intersectional angle from the positive direction of the imaginary axis to I by 8. z 2 . Re& < C j=1 n & < 0. we have Re{eae(z . Hence which shows that the zeros of P ’ ( z ) do not lie on the left side of I . which shows can not be zero on { z : Rez 5 0). When z is on the left side of I. 5218 Let f ( z ) be a Laurent series centered a t 0.
1)holds for all C with I(I = 1.11. f maps D into D . It follows fiom that which implies that (p(8)= 8.cl = 7 holds on {C : 1[1 = 1). Apply the discreteness of zeros for analytic functions to f ( z ) .we obtain f(z) = z EC\{O}. Hence there exists ( with = 1 such that (b) If max ICI=1 If(<)  = Ib . it follows from that If(C) . (b) Show that &[argf(e”)] > 0 for 8 in R. . where D = { z : IzI < 1). then which is a contradiction.375 If If() .f’(z) # 0. and f maps d D into a l l . p. ( a ) Show that Vz E aD. 5219 Assume f is analytic in a neighborhood of 0.P I = Ib .l)ei+(e). and hence f ( 0 .<l1 < J b .<l = ( b . b1 p.1 1 holds for all C with I(I = 1. Let f(() . where ( = eie and 4(8) is a continuous realvalued function.
which implies that f ( z ) has two zeros (counted by multiplicity) in D. it follows from = aR that %(e i6 ) < 0. t . &Alzl=largf(z) = is 2. g(z) = b o + b l ( z . and w = f(z) = Rei@.> K .376 (c) Assume that f(0) = f'(0) = 0 and f l a ~ a twotoone map from a D is onto dD. with bo # 0. 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. z = 0 is a zero of f of multiplicity m = 2. Hence f'(z) # 0 for all 2 € aD.wo = ( z . . If g ( e i e ) < 0 . Let I be an arc in ' defined by I = { z E % : ) z . It is ' a contradiction to the fact that f maps D into D. If g ( e i e )1 0 . . It follows from that Ararg(z . Since R = 1 when r = 1.A variation of the CauchyRiemann equations for analytic function w = f(z) is Since f maps a D into a l l . 2 3 which implies that f(z) assumes values outside the disk when z E I. (a) Assume f'(z0) = 0. It is demanded that r is sufficiently small such when z E I?. (b) Let z = rei6.ZO) 3T n+ Arargg(z) > .. Hence f(z) has no zero in (0 < I1 < I}. This is also impossible. = = = 0.~ 0 ) 2 + . But from (a) it is impossible.zoI = r ) .. where zo E d o .W O ) = nArarg(z . ( IndianaPurdue) Solution.z o ) + b z ( z . Then where n 2 2. which implies that g ( e i 6 ) = f'(ei0) = 0. g ( e i 6 ) < 0 implies that R > 1 when r < 1. that Ararg(f(z) .zo) > and Ig(z) .z o ) " g ( z ) ( n >_ 2). Since f(0) = f'(0) = 0. Show that f(z) # 0 whenever 0 < IzI < 1. we know that % ( e i 6 ) = 0. . Hence we obtain 2 2 7s Rg (c) Because f l a ~ a twetoone map from 8 0 onto a D .bol < f ( z ) . then = at point ei6.
377 SECTION 3 COMPLEX INTEGRATION 5301 Evaluate the integral (Indiana) Solution. Function e e z is analytic in { z : 0 around z = 0 is: < IzI < +co}.. + By the residue theorem.= e. (Indiana) .l)! 5302 Evaluate where y is the positively oriented circle { Jz1= 1). we obtain 1 +.. . and its Laurent expansion The coefficient of the term $ in the above development ~ is 1 1 + 1 + 2! + . ( n.
its and F(z) = (. and with z = 0 as a pole.. I ~ Hence the coefficient of the term residue theorem.. i.378 Solution.e. we have p in the above development ( 3 1O ) . sin z =~ i ..... . Function F ( z ) = ( $ 5 ) c o s z is analytic in { z : 0 < IzI Laurent expansion around z = 0 is: + < +a}. = 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.. By the .)cosz= 2+('2)P+(.)Z+. . + 5) +. It is obvious that is analytic in { z : 0 < IzI 5 l}. the coefficient of the term in the above development is zero. The Laurent expansion of around z = 0 can be obtained as follows: &  1 sin3 z  1 (z  k.2aiRes 5303 For what value of a is the function f ( z )= singlevalued? LZ(.. Hence we obtain a = 2. is $.) (: coszdz (Indiana) Solution.+. 1 U a ( .23 + $25 1  . 1 1 1 2 t4!z4.
Consider the integral et + z t dt. By Weierstrass theorem. Show that . It is easy to see that when t + 1. Hence the largest possible P is equal t o e.IzI < 2).et dt. Hence P can not be larger than e . let IzI 5 T . 5305 Let f ( z ) be analytic in S = { z €67. we know that h ( z ) is analytic in { z : IzI < e } .1)2’ where A = $ cos 1.379 5304 Define h ( z ) = L m ( l ztet)let cos(t2)dt. When z = e. dt It follows from let + ztl 2 et . + What is the largest possible P so that h ( z ) is analytic for IzI < P? ( IndianaPurdue) Solution. cos(t2) et .et  A (t .rt and the convergence of the integral that is uniformly convergent in any compact subset of { z : IzI < e}. et . which implies that the integral is divergent. For any r < e .
380 (Iowa) Solution. 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.t E [0. dz dz z where y(t) = eZnit. 1 ( S U N Y . Then f ( z ) is analytic in { z : IzI < 2}. and we have + . Stony Brook) Solution. Show that < 2}. Let f ( z ) = u ( z ) iw(z).
381 It follows from u ( z )= and f(z) +m 2 f f (> .IzI 5 11. and if Ref(z) /2T 0 enie[Ref (eie)]dO. l). w(z) = that  22 5307 Let f be an analytic fqnction on an open set containing D(0. prove that > 0 for all points (Indianu ) . z E D ( 0 .l) = { z . (a) Prove that d”f(0) = dzn (b) If f(0) = 1.
(a) Assume that 00 k=O we have By Cauchy Integral Formula.382 Solution. we have = 2(n!). Hence (b) Because Ref(z) is harmonic on D(0. l). by the meanvalue formula of harmonic functions. . kI'" Noting that Ref(eis) Ref(e")de = Ref(0) = 1. 2 0.
(0 < T < 1).14)1. It is obvious that lull = I’0l I 1 < e . f() For n > 1. we choose T = 1 . : (1 + )I n1 1 < e. where e is the base of natural logarithms. (Stanford) Solution. n=l where nu. . 5309 Let f = u + iw be an entire function. = 1 f)odz. It follows from f ( z >= 00 C anzn n=O that f’(2) = c 00 nunznl. show that the coefficients in the expansion n =O satisfy lunl < e for n 2 1..383 5308 I f f is analytic in the unit disk and its derivative satisfies IfWI I (1.
Letting R + +m. By Liouville's theorem. which implies that f ( z ) is a constant. (Indiana) Solution. (b) Show that if If(z)I 5 A+BIzlh for all z € @ w i t h some positive numbers A . 5310 Let f be an entire function that satisfies IRe{f(z)}I 5 lzln for all z .). . where n is a positive integer.f 2 ( z ) = .)) 5 1. F ( z ) must be a constant. then f must be a constant.(ua(z)v2(z))Ziu(L)v(r) Then F ( z ) is an entire function with IF(z>I = e (. Show that f is a polynomial of degree at most n. 5 A+BR~ R" . h."(. then f ( z ) is a polynomial of degree bounded by h. (b) Let For any integer n > h.'(. ( Stanford) Solution. Let R be an arbitrary positive number. (a) Let F ( z ) = e . which implies that f ( z ) is a polynomial of degree bounded by h. B . Then it follows from Schwarz's theorem that when IzI < R. we obtain that a. = 0.384 (a) Show that if u 2 ( z ) 2 wz(z) for all z EC.
385 Especially when IzI = +.)l L 1 + IIm{f(O))lr which implies that there exist constants A . 5311 Compute the double integral 1J . cos z d x d y where D is the disk given by { z = x + i y E a' : x 2 + y 2 < 1). which shows that f(z) is a polynomial of degree at most n. First we have the following complex forms of Green's formula: . (Iowa) Solution. B such that If(z>I 5 AlzY holds for all z EC. 3R" * 2n + IIm{f(O))l = 3R" If(. Let +B where Hence when k > n.
(a) Express A in terms of the coefficients an. < 1) and assume that the integral is finite.386 The problem can be solved directly by either one of the above two forms: o r JJDcoszdxdy = sin zd? = . 1 sin zd( ) Z 5312 n=O be analytic in D = (1.sin z dz = R. (Indiana) Solution. (a) B Y n=l we have . (b) Prove that for z E D.
we have 5313 Let f be analytic in (0 < IzI < 1) and in L2 with respect to planar Lebesque measure. Is 0 a removable singularity? Proof or counterexample.387 Noting that we obtain that . ( Stanford) Solution. Let the Laurent expansion of f in {z : 0 < IzI < l} be 00 From . 1 00 00 (b) By Cauchy’s inequality. The answer to the problem is Yes.
a = rei+. let E + +0. which leads to a contradiction.21r 1' la. while the right side of the inequality is finite. 5314 Evaluate the integral (Indiana) Solution. Let z = peie. the left side of the above inequality will tend to infinity. . Hence n=O which shows that z = 0 is a removable singularity of f .388 we have Let E < 1 be a small positive number. 12r2n+1dr< Then a. and then . Otherwise. must be zero when n 5 1.
5315 Evaluate by the method of residues. When r > p. (Columbia) Solution. Then we have I(a) = dx 2dx dx . Denote It is obvious that I ( u ) is an analytic function in { u : la1 lr > 1).When r < p.
27T z2) z1 . 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.6)de when ( z ( < 1. (b) Integrate the Taylor expansion n=O term by term to find the coefficients in the Taylor expansion . By the residue theorem we have zl iz. 1 = 0 by z1 and z2.390 Let z = eisl then dx cosx and = = z dz iz ' + 21 2 ' = Denote the two roots of z 2 .  2dm' 0.=l (z  idz  z1)(z . Since z1 z2 = 1. It should be noted that branch of is also analytic in { u : la1 > l}. 1221 < 1.=l + + 22  idz 2(2u + l ) z + 1' =  iz.2(2u 1). we may assume that ll > 1.z2 7T 27r d(Zl t 2 2%& * ~ ~ ) ~ 42122 .
27r]. Because IC1 . k=O (1. (Courant Inst. 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.( (2 < 1) is defined =i(dnl). C21 = 1.e) = C ( . and the singlevalued branch by 4 I r = ~ = 1.zk k=O 00 where Uk 2r = (1)' sink ode.i0 .) Solution.y . the integration with respect to 8 can be taken term by term. Since the series converges uniformly for all 8 E [0. z k . we develop the f(z) = 2K ~ 00  dC7 27r(l z"i = 2 7 r . Then .l ) " C I f .391 (c) Verify directly that (a) and (b) agree when IzI < 1.1 ) ' s i n k 8 . (a) Let c = ele.eiO sin8 = 22  C2 1 2ic ' and where of d ci = $(d=l). (b) It follows from I sin81 5 1 and 00 IzI < 1 that g(z. and f(z)=$' (g(l)ksink8. Z 2 " n=O . m in 1. < 1).
Irvine) Solution.5. show that exists and is independent of a.5 As f ( z ) is analytic inside I.392 we know that the results in (a) and (b) agree when J z J< 1. rl U I'2 U Fig. ' . by Cauchy integral theorem. 5317 If a is real.5. First we have It follows from the existence of hllRex2dz R that exists.5. (UC. ' Define f ( z ) = e" and choose the contour of integration I = r3 u r4 as shown in Fig.
5.393 ieR2 J. 0 R Fig . f ( z ) has one simple pole z = eEi inside I'. Compute (Iowa) Solution.e:*). we have f(z)dz = 2~iRes(f.' ey2+2Ryi = 0. By the residue theorem. dY it follows from the facts that eR2 + 0 ( R + CQ) and 5318 Let n 2 2 be an integer. Letting R + 00. . and select the integral contour r as shown in Fig.5.6 Let f ( z ) = &.6.
Fig. (Minnesota) Solution. B caus is analytic on I' and inside I by Cauchy integral theorem.5. and choose the contour of integration f(z) = have eIZ r= 3 rj j=1 as shown in Fig. we ' .7 Define f(z)= e .7.394 and 5319 Evaluate JD cos (x') dx with full justification.5. 3 j=1 .z 2 .
sin z2)dz = 2 ' which implies 5320 Evaluate (Iowa) .O 5 argw 5 }. 2 7F By Jordan's lemma. Hence we obtain by letting R co that + sin x2)dz + i 00 & (cos z2 .395 For the integral of f ( z ) on r2. we make a change of variable by w = z2. L m ( c o s z2 f(z)dz = I" / 0 ex2dz + 2 f fi when R +00. we have For the integral of f ( z ) on r3. we have It is well known that l. then where 72 = { w :lwl= R2.
5. by Cauchy integral theorem. where It is easy to see that lim Rr 00 I' iReief (Reie)d6= 0 and . Fig.8. Because f(z) is analytic inside r.396 Solution.5.8 Define and select the integral contour as shown in Fig.
9. we know that Res(f. Fig. 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. and the singlevalued branch of logz be chosen by argzl.397 Since the Laurent expansion of f about z = 0 is 00 where a1 = 2i. J. Letting E + 0 and R + ca. The branch of log is chosen to be real for positive z.f ( z )logzdz = 0. .5.we obtain dx = .5. by Cauchy integral theorem.=I = T . Stony Brook) Solution. 0) = 22. Since f(z)logz is holomorphic inside the contour I?. 2 7r 5321 Let f ( z ) be holomorphic in the unit disk IzI 5 1.9 ' Let the contour of integration I be shown as in Fig. ( S U N Y .
and the value of logz at the starting point z = 1 is defined as 0. First we assume la1 > lbl. (Hamud) Solution. and then the multivalued analytic function log(a bz) has singlevalued branch on { z : IzI 5 1). . 5322 where a and b are complex constants. I 2r log la + be'@]& = Re = R e ( 2 ~ l o g a )= 2rlog lal. Take ei@ = z .398 where It is easy t o see that Letting E + 0. not both equal to zero. then d4 = and + g. we obtain where the integration contour IzI = 1 has starting point and end point z = 1.
399 When la1 < lbl. as shown in Fig. () = log( 1 z +z) 7 where the singlevalued branch is defined by log(1 z ) lz=o= 0. Uy.5.10. we have JR I*log 11+ ei@1d4 = Re JR IR log(1 + e*+)d4 . Then Fig.10 In order t o evaluate the integral we define f. let b = ae". we have In the case la1 = lbl. Since f ( z ) is analytic on I' and inside r. Because + s .5. Choose a contour of integration = r. by Cauchy integral theorem. f(z)dz = 0.
.'" Evaluate log la + beiGldq5 = 2 7 max(l0g la[.. l). The singlevalued branch of logz is chosen by argzI. Fig.log lbl}.5. By the residue theorem. f(z)dz = 2?riRes(f. ~ 5323 (Iowa) Solution.1 = ?r.400 Hence we obtain J. we have J.11.5.11 Let and select the integral path I? as shown in Fig.
we consider the Laurent expansion of f about z = 1: = where a1 n=3 c a n ( z + lyl = 1 .401 where 2X iRegef (Reie)d6 + (logx + 2ai)' dx rO + It is obvious that 1. Comparing the imaginary parts on the two sides of the above identity.ai. we obtain . Rrm lim J. it turns out that 4ai log x + 4a2 dx = 2ai + 2 2 .'" iReief(Reie)d6 = 0 and In order to find Res(f. AS E + 0 and R + 00. Hence 2aiRes( f . 1) = 27ri + 2 2 .l). ice"f (&eie)d6.
we have f ( z ) d z = 2 ~ i R e sf(. (a) Fig.12. 2). As f ( z ) is analytic on and ' inside J? except a simple pole at z = 2.) Solution.402 5324 Evaluate the following integrals: +im at (a) Sim ( t 2 . &dz for CY in the range 1 < QI < 2.5.5. where and Because . and the contour I of integration is shown in Fig.12 Define 1 The singlevalued branch for log(z + 1) is chosen by log(z + l ) l z = o = 0. (b) (Courant Inst.4 ) l o g ( r + l ) (the integration is over the imaginary axis).
2 log3 Fig. a simple pole at z = e:'. I F where and Res(f. and the contour 'I of integration is shown in Fig. +ice dz ( z 2 .efa) = lim ( z .4)log(z + 1) = (14 Ti )./.13.5.1  3e f ( 2 .O = 0... e f i 3e +a . we have f ( z ) d z = 2 ~ i R e s ( f e:')). As f ( z ) is analytic on and inside r except .efa) f ( z ) z+e:' .13 Define The singlevalued branch for zm is chosen by argzIr=z. .403 and by letting E + 0 and R + 00.m ) i ' .5. we obtain .
Fig. by letting E 0 and R obtain dx = 5325 Show that J.14. Let ' where (argza)z=z>O = 0.404 Because lirn R U C 2z . What happens if a = l? Solution.a ) ( 1 + x 2 ) 2 d x = 4 cos( y )' < a < 3. and select the integral path I as shown in Fig. 5. we have f ( z ) d z = 2aiRes(f(z). By the residue theorem.5.14 where .we 7r when a > 1. a # 1. i ) . f(Reie)iReied8 = 0 2a  when a < 2 and lilil ' + f(&eie)iseied8= 0 + co. for 1 00 Xu 7r( 1 .
a(1 .i) = lim [ ( ~ 2'2 2 ta i)2 + 1'9  e"i 2 . I.a ) x y d X= 4 cos( y) ' X T(1a)  1 5326 (a) Prove that converges if 0 <a < 1 .+ (1 when a = 1. and from a > 1 that Letting E + 0 and R $ 00. we obtain XQ a(1a) 2 e 1" 2 . When a # 1.O iReief (Re")dB = 0 . It follows from a < 3 that lim RC ." iEeie f (&eie)d8 and Res(f(z).405 + J. 2 " al a) ( .
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 analytic in D . is a real constant. and when z tends to dD. a4 = a6 = .) 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. which implies f ( z ) = e i a z 2 . we have Since a2 = e i a . If equality occurs for some nonzero z in D . the other coefficients must be zero. Because If(z)I < 1 for z E D . If(.. we have + I I f(z) + f ( .z ) = 2eioz2. (b) . limit of the Hence can not be larger than 2. (a) Let F ( z ) = f(z) + f(z). where a.z ) = 2eiaz2 = eio. By the maximum modulus principle.. = 0.419 (b) Prove that (Indiana) Solution.
z has one zero in { z : IzI < l}. is the point necessarily real? Justify.Z. ?r Take a closed ?r I’ = { z = t + i y : t = 0. .G(z)l = le”l 5 1 < IG(z)I. prove that f ( z ) has a fixed (Rutgers) Solution.G ( z ) /= lf(z)I < 1 = IG(z)l. Because R can be arbitrarily large. Let R be a sufficiently large real number such that R curve I’ on the right halfplane. it follows from RouchC’s theorem that ’ F ( z ) has exactly one zero inside I?. Hence f(z) takes value X exactly once in the right half plane. Prove or disprove: f ( z ) takes the value X exactly once in the right halfplane. IF(z) . . 2 2 Define F ( z ) = X . Let F ( z ) = f(z) . By RouchC’s theorem) F ( z ) and G ( z ) have the same number of zeros in { z : IzI < 1). F ( z ) has exactly one zero in the right halfplane. R 5 y 5 R} u { z : Izl = R. where > 2A.z and G ( z )= z. (Iowa) Solution.< argz 5 }. Since G ( z )has exactly one zero inside I.420 5410 I f f is analytic and lf(z)I point.e” and G ( z )= X When z E I?. If the answer is yes.’ . < 1 on { z : IzI 5 l}. . IF(z) . X > 1. 5411 Let f ( z ) = z + e . Since G ( z )has only one simple zero in { z : IzI < 1))we conclude that f ( z ) . which implies that f ( z ) has a fixed point in { z : IzI < 1). When IzI = 1.z .
We have F ( x ) = X . (Iowa) Solution.z. Assume f is nonzero on the unit circle {z : II = 1). Let C denote z the unit circle traversed in the counterclockwise sense. z being repeated.z2. We have .e". are the zeros of f ( z ) in {z : II . multiple zeros f(z) = !dz) j=1 nc. Assume z1. n . which is a realvalued function of real variable x. . x++w lim F ( z ) = m. In other words. Suppose that and Find the location of the zeros o f f in the open unit disc {z : II z < 1).x . Then < l). 0 < xo < $00. such that F ( z 0 ) = 0. Since F ( x ) is continuous and F(O) > 0.Zj). where g(z) is analytic and has no zero in {z : Il z 5 1). 5412 Suppose f is analytic in a region which contains the closed unit disc {z : ( z ( 5 1).42 1 Take z = x 2 0. the point z in the right halfplane such that f ( z ) = X is necessarily real. there must exist 20. +.
e. (a) Let R be sufficiently large such that when IzI = R..z z ) g ( z ) . f ( z ) has two zeros in the unit disk. lr Choose a closed curve ' = I { z = z +iy. 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 $? (IndianaPurdue) Solution. Then for f(z) = ( z .O 5 z 5 R.z1)(z . z1.422 It follows from (1) that n = 2. U{z=z+iy:z=O.y = 0) U { z : II = R. i.O 5 argz 5 } z 2 R}.2 = ki. and which show that unit disc.O<yL . Set f ( z ) = z4 +z +5 and g(z) = z4 + 5.
.argz4 + Ar. the numbers of the zeros of f and g inside I' are equal. Since g has only one zero inside I?. 2 A It is easy t o see that Imf(z) > 0 when f(0) = 5. Set f(z)=z4+z+5 and 9is approximately and r3 = { z : IzI = R. while .)l holds when z E I?.argf(z) = Ar.423 It is obvious that If(. z4 z + +5 = 0 has one root in the first quadrant. Ap. zero.g(z)I < Is(z.< argz 4  A 5 }. i) f (a$*) E {w : ?r E < argw < where E > 0 is very small.mgf(z) denotes the change of argf(z) when z goes continuously from Re<' to R along I3 It is obvious that i '.) .arg (1 + F) . and f ( ~ E {w : o < argw < E } . We also know that Ar. we know that ' . where Ar.argz4 = A. By RouchC's theorem. f has also one zero inside I Noting that R can be arbitrarily large. (b) Let R be sufficiently large such that when la1 = R.
it follows 3 from the above discussion that By the argument principle. 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 < }. Let f(z) = sinz . Let I' = I'l U r2 U r is taken once counterclockwise. 2 x 5414 Prove that the equation sin z = z has infinitely many solutions in C. where .z and z = z + i y . then f ( z ) can be written as For any fixed natural number n. (Indiana) Solution.424 is very small. choose a positive number t >> logn and a closed contour I' = rl U I'2 U r3 U r4 in the counterclockwise sense.
with the direction downwards. Remark. f ( z ) can be written as f(z) = e g ( z ) k ) ) P where P(Z> n = k=l .(et 2 .z has only finite zeros in (X. * . By the argument principle.425 Then we consider the image of r under w = f (2): f(rl) {w = u + i~ = : 2(n +1 ) 5~u 5 .t ) lies in the annulus starting from w = 2(n + 1)n + i in the counterclockwise sense. o 5w 5 1 . multiple zeros being repeated.2 7 ~ v ) 0) ~= with the direction from the right to the left. f ( z ) = sinz .z n .z has one zero inside the contour I?. f(rz) {w = u + i v = with the direction upwards. Since n is arbitrarily chosen.zg) and g( z) is a polynomial. This problem can also be proved by Hadamard’s theorem. z z . Hence the winding number of f(r)around w = 0 is 1. Assume that f ( z ) = sinz .et ) . and denote all the zeros by z1. . we conclude that sinz = z has infinitely many solutions in c. By Hadamard’s theorem. : t = 2(n+ i f(I’3) ~ ) T . .
where rl = { z : IzI = 2) t is in the counterclockwise sense. and comparing the increasing order on both sides. or both f(r1) and f ( r 2 ) are {w : IwI = 5) in the clockwise sense. By letting y + +oo and y + oo respectively. (b) I f f is meromorphic in the annulus. (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. is the statement in part (a) still true? (Stanfod) Solution. and I'2 = { z : IzI = 1) is in the clockwise sense.426 It is obvious that f ( z ) is an entire function of order X = 1. the statement in (a) is not true.argf(z) 2 1 and &Ar.argf(z) 2 1.z = e a r + b p ( z ) . we obtain that Imu < 0 in the former case and that Imu > 0 in the latter case. It might occur that f(rl) and f(r2)are two subarcs of {w : IwI = 5). Let z = z+iy and z be fixed. which shows by the argument principle that f has at least two zeros in D. In the latter case. The following is a counterexample. Hence &Ap. Let g(C) be a conformal map of {(=<+irl:$+$<l. In other words. . Show that f has at least two zeros.we know that both f(I'1) and f(rz) must be {w : lwl = 5) in the counterclockwise sense. (b) If f is meromorphic in D. where T03 which implies that g ( z ) must be a polynomial of degree 1. This contradiction implies that sin z = z has infinite many solutions in C. Hence we have sin z . f has no zero in D. 5415 (a) Let f be a nonconstant analytic function in the annulus (1 < IzI < 2) and suppose that If I = 5 on the boundary.
y = 01. where rl = { Z = TP: o 5 8 L ) . (Indiana) Solution. But Let n be a positive integer. Let T > 0 be sufficiently large such that when IzI = T .427 onto {w : IwI > 5) with the normalization g(0) = M. Then the number of zeros of P ( z ) inside I is equal to ' It is already known that . ( a 2 n ~ 2 n ( (ao+alz+. g'(0) > 0. e . 0 where each aj €67.+azn_1z2"11. > Take a closed contour I? = I'l U I'2 in the counterclockwise sense. 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). Suppose that there is no real number x such that P ( z ) = 0.. 5416 I f I = 5 when t E LID. and aZn # 0. and let P be a polynomial of exact degree 2n: P ( Z )= a0 + alz + a2z2 + + aZntzn.. and r2= { z = X + i y : T 5 T.
.+gcn. argP(z) 1 Note that and 1 Ap. From .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. (a) Let 1 F(C) = f() = 1 + c c 1 1 1 + 3c2+ . * .428 We also have d log P(z) = G A r .arg 2a (u2nz2n) =n we obtain that P has exactly n roots (counted with multiplicity) in the open upper half plane. 5417 Consider the function 1 1 1 f(z) = 1 + .) Solution.e3+ * . +.+  z 2!z2 1 1 + .
< R}. simple solution z1 of the equation f(z) = w in 1(. When n is sufficiently large. lF(C) . there is no zero of f ( z ) in { IzI > T } . Let min leCl = m. ICl=? then m > 0.429 we know that the negative of 11 represents the number of zeros of F(C) in (1 < :}. 5418 (a) Suppose that f ( z ) is analytic in the closed disk IzI 5 R. if the integer n is sufficiently large.ec I < m I 1ecI for ICI = : which implies the numbers of zeros for F(C) and e in . which is just the number of zeros of f ( z ) in 1(. > T}. we conclude that for any fixed T > 0. (c) From the above discussion. C are equal. Since ec has no zero in a'. the equation has exactly one solution with ( z ( < 2. F(C) converges to ec uniformly in any compact subset of@. when n is sufficiently large. (b) When n + 00. . and that there is a unique. In other words. we obtain {/(I < :} for fixed T and large n. Show that this solution is given by the formula W (b) Show that. all the n zeros of f ( z ) are in { IzI 5 T } .
= .& ’ ( z ) = ’ w zzi Q(z)* + + Hence (b) Let fn(z) = z . E Hence fn(z) and g ( z ) have the same number of zeros in { IzI < 2 . we have fn (1) < 0 and fn > 0.1 and rE = For fixed large n.zi)&(z). we choose E = 2 . Since E can be arbitrarily small. It follows from (z) that .w)]’ [log(z .g ( z ) l = 11 5 z n = (1  5)” < 1 .w = ( z .) 1 1 Solution.(z) is a continuous realvalued function for 1 5 2 5 $.E } . (a) Let f ( z ) .€1. (c) f. the equation z = 1 + (5)” has exactly one solution (denoted by z p ’ ) in 1(. z > 0 sufficiently small such that when E re. $). ncc 2 (Courant Inst. where Q ( z ) is analytic and has no zero in { IzI < R}. Ifn(z) .1). and the number is 1. Hence we have z?’ E (1.430 (c) If z1 is the solution in (b). show that lim (zl.E L Ig(z>I. When n is sufficiently large. < 2).z ~ ) l o g & ( ~ ) ]= . Then f 1 f’(z) = [log(f(z) .
1 1 1 f(1 = f() = 5. while my1 . for example. Let where E > 0 is small. Hence we obtain f(*) = f2. .f of order n and z = f is a zero of f(z) . If. then f * (my1 .nr2) 1 2 1 f(i) is homologous to zero (mod Q).l)\{. 2 2 we assume that z = f is a zero of f(z) . }.43 1 which implies 5419 Let 1 1 R = D(O. f(rl) and 1 f(y2) will tend to either 20 = 5 or 2u = .7 2 are not homologous to zero (mod R).) will not happen. the points z = =kf must be the removable singularities o f f . and the directions of 7 and 7 2 are both in the counter1 clockwise sense. As E tends to zero. then f * y is not homologous to zero (mod R). Since 7 .ny2 is not homologous to zero (mod R). Now we claim that the case that = f(. (Indiana) Solution. Thus we obtain either f() 1 1 =2 2' f() =  1 2 1 2 . 2 2 Find all analytic functions f : S2 + R with the following property: if y is any cycle in R which is not homologous to zero (mod R). f * 7 or f * y2 1 will be homologous to zero (mod R). f * yl and 1 f *y2 are also not homologous to zero (mod R).$ of order m. which is a contradiction. Since f is analytic in R and bounded by lf(z)I < 1. because otherwise.
It follows from F ( that F ( z ) 1.l) and satisfies IG(z)I 5 1.432 or f() 1 2 = . It follows from G ( . 1. = .51 1 1. 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.t z 1 which is analytic in D(0.)   z .l) and satisfies IF(z)I 5 1. = i)= 1 we consider the function f(z)+i G(z) = 1 + if(. 2 2 1 1 we consider the function F(z)= fk).i f ( z ) 2 1. which implies that f(z) = z. which implies that f ( z ) = z.i ) = 1 that G(z) 1. 1 2 f() = .t .
the Taylor expansion of F ( z ) a t z = 0 has a radius of convergence larger than T . In other words. Prove that lim an/an+l = zo. which is a simple pole.zo z Then F ( z ) is analytic on { z : 1zJ5 T } . and define F(z)= f ( z ) A . n+w (Indiana) Solution. on I1 = T .zo 00 is convergent a t z = zo.z . Hence the power series F(z) = n=O A Cantn. . which implies It follows that .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. Assume that the residue of f ( z ) at zo is A .
a ) for such a.( < 1). t E ( 0 . (Minnesota) Solution. (1) Let a = eit. which shows that . 2 ~ we have ) and Because both C for 1 # a n>l tends to zero monotonically. we have 1 1Z . (2) Show that this series converges to log(1 . Differentiating term by term. c. 2 ~ )then . n+co % + I an 5502 (1) Show that the series n> 1 is convergent for 1 # a E C with la/ = 1.C $ is convergent and C n>l n>l E C with la1 = 1.=c an n>l n>l cos nt + i sin nt n For t E ( 0 .434 and and we obtain lim = zo. by Dirichlet’s criterion we know that converge. (2) Let f(z) =  n>l c n> 1  zn n 1.
Assume the proposition is false. then for fixed 2 E ( 0 .e. z = 1 is a regular point of f.C n an n> = = C.e i t ) = log(1 . Prove that the boundary of the disk is the natural boundary of the function.= n> 1 r+l eint limCr+1n>l (Teit)n 1 n lim log(1 T e i t ) = log(1. ( Columbia) Solution. I n= 1 Show that the series converges to a holomorphic function on the open unit disk centered at origin. Suppose the series : is + .a).435 Integrating both sides on the above identity. i. for (a1 < 1. . It follows from Abel's limit theorem that . we obtain f(z) = log(1.z ) . First of all. z = Teat where 0 < T < 1. 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. we prove the following proposition: If the radius of convergence of n =O is equal to 1 and an 2 0 for all n. then a = 1 is a singular point of f(z). 5503 Consider a power series O01 C nz n . 0 < t < 2a. Let Q = eit..
. By the above proposition. and hence the order of summation can be changed. &I n. z = 1 is a singular point of F ( z ) .436 Thus 00 w w x b k ( z z)k = k=O n(n . Noting that when z = 1 6 the right side in the above identity is a convergent double series with positive terms.". we assert that when z = 1 6 .k + k=On=k is convergent at z = 1 6 . For any natural numbers p and q . .1) . c . a ( .z ) k z n .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. . + + n=O W k=O . n=O which contradicts the statement that the radius of convergence of n=O is equal to 1.
Since the set { e . In other words. q = 1. prove that either g ( z ) = 0 or g ( z ) = z . If the sequence { f n } is defined by composition < 1 for all and f n ( z >4 g(z) for all z E U . we obtain a = 0. 5504 z E Suppose f is analytic in U = 1(.1 for some z # 0. Then . we conclude that the unit circle {IzI = 1) is the natural boundary of F ( z ) . z = e?4.(fI for all z E U . (IndianaPurdue) Solution. Remark. Let 0 < r < 1.2. In the case when f ( z ) = eiaz.and if If(z)I = 1. it follows from f(0) = 0 and If(z)l < 1 that 5 IzI for all z E U .} is dense on {It1= l}. Since f n ( z ) is convergent.i is a singular aq point of F ( z ) .. which implies that f ( z ) = z and g ( z ) = t. In other cases..p r i : p .. we have [). By the above discussion. n n=p it is also a singular point of F ( z e 9 " ' ) ..437 Since z = 1 is a singular point of " 1 I zn. By Schwarz's lemma. then f(z) = einz where a is a real number. 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. f n ( z ) = einaz. < 1) with f(0) = 0 and If(z)I U .
{fn}r=l ZOI Let m be the minimum value of /). (a) First of all. .. we know that for sufficiently large n.. T}. we have . such that f(z) # 0 when z E {Z : 0 < IZ 5 T } C D.z0I = T } . Since zo is a zero of f .. (b) If each fn is onetoone on D . Ifn(z>. As {fn} converges to f(z) uniformly on compact subsets of D . Irvine) Solution. Hence fn(z) converges to zero uniformly in{lzl 5 arbitrarily chosen.(fI on { z : Iz . we know from Weierstrass’ theorem that f is analytic on D . show that f is either constant or onetoone on D . fn must have a zero in { z : Iz . but has a zero point zo E D .z0I < T } . there exists T > 0.z01 = T}. (UC. . Since the zeros of a nonzero analytic function are isolated. (a) Prove that if fn(z) # 0 for all n 2 1 and z E D .f(z>I< m 5 lf(z>I holds on {z : Iz . Then m > 0.z0I < T } . we obtain g ( z ) = 0 for all z E U . then either f is identically zero in D or f(z) # 0 for all z E D. which is a contradiction to the assumption that fn(z) # 0 for all z E D. Suppose f is pot identically zero. It follows from RouchC’s theorem that fn and f have the same number of zeros in { z : Iz .438 For all z E { IzI 5 T } . 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 .
With the same reason as in (a).221 5 T } .z l ( 5 T } U {z : 0 < Iz . Choose T > 0 sufficiently small.a = 0 and fn(zi) . 5506 Let D c a! be a bounded domain. such that and f ( 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 have the same number of zeros in {z : ) z. there exists 2 : .z2( < T } .z2 E D (z1 # zz).) . Then m > 0.al { z : Iz .221 < T } respectively. It follows from RouchC’s theorem E { z : Iz .z 2 ( = T } . I(fn(z) . such that f(z1) = f(z2) (denote it by u ) .zll = T or Iz .a # 0 in {z : 0 < ( z . and is not onetoone on D. Prove that n+m lim f n ( z ) = b for every z E D. This is a contradiction to the assumption that f n is onetoone on 2 D. Let rn be the minimum value of If(. (Indiana) .u = 0.a1 on { z : IZ .z11 < T } and { z : (z .a ) holds on that f n (z) . when n is sufficiently large. Then there exist zl.(f(z).zll = T or (z .)I = Ifn(z) .z 2 ) = T } . such that fn(Z{) .439 (b) Suppose f is not a constant.211 <T} and z: E { z : (z .f(z)I < m L If(. which implies fn(z{) = fn(zi) ( : # 24)..a and f(z) . In other words.) .
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 .bl : )z . Let T be sufficiently small such that f ( z ) .Re{g} > 0. Let {fn} be a sequence of functions in F.g(O) = l. we assume lim fn(u) = b 6 d D .C E D . diam g ( D ) 2 1).a1 = T } > 0.a1 < T } . Since { f n k } converges uniformly to f on { z : ( z . For any two fixed points z. (Indiana) S o htion. By Rouche’s theorem. nP . there exists a subsequence of {fn(ao)} converging to bo # b. f(z) is a nonconstant analytic function of D . there is a subsequence { f n k ( z ) ) converging uniformly on compact subsets of D to f(z). ao # a.u [ 5 T } c D .b has zero(s) in { z : Iz . we have . 1 1 on { z : Iz .it follows from f(0) = 0 and diam f ( D ) 5 2 that lf(z)I 5 2. which shows that 7 is normal. f(0) = 0.b has no zero in { z : 0 < I z . then m = min{lf(z) . diam f ( D ) 5 2) (b) B = { g : g is analytic in D. (a) For any f E F.m n+ca lim fn(ao) = bo # b. fnk(z) . Because f(u) # f ( a o ) .a = T } . when Ic is sufficiently large.a = T } . Here the diameter of a set S is diam S = sup{lz . Then there exists a subsequence { f n k } converging uniformly in compact subsets of D to f(z). (a) 3 = {f : f is analytic in D . If {fn(ao)} does not converge t o b. Take a0 E D .440 Solution. Without loss of generality.CI : z . 5507 Which of the following families are normal. Since { f n ( z ) }is a normal family. C E S}. which obviously satisfies the conditions that f(z) is analytic in D and f(0) = 0. and which is compact? Justify your answers.
we have IGn(z)l < 1. by Hurwitz's theorem. By the Riemann mapping theorem. and we conclude that g n. we have G(z) # 0 for all z E D . { g n ( z ) } converges uniformly in compact subsets of D to g ( z ) which is a conformal mapping of D onto 52.(z) # 0 for all z E D .} converges to R = {w : ( w .C E K .} with respect to w = 1. } be any sequence of functions in Q. since G.which shows that F is also compact.) . if G(z) is nonconstant analytic. such a mapping gn exists and is unique. which shows that is not compact. But family Q is not compact. (b) Let { g . It follows from the uniform convergence of {f.} on K that If(. = {w : I .1 < 1 which is called the kernel of {R.1 < } U {w : I .) = logGn. Because the domain sequence {a. hence f(z) E 7 . 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 . and it is obvious that g n satisfies all the conditions required by the family Q.f(0l 52. gL(0) > 0. 1 < Rew < 3) w 1 w 4 1 1 n satisfying gn(0) = 1.31 < 1) U {w : IImwl < . If G(z) is a constant.log G(0) = 1. Since z . Hence we can define an analytic function g ( z ) = log G(z).. ( D ) 5 2. Then for Gn(z) = egn('). Let 7 be the set of all analytic functions f on { IzI < 1) such that . 5508 Suppose that 1 5 p < 00 and c 2 0 is a real number. a} i.(z) converges uniformly in compact subsets of D to g ( z ) . which shows that family Q is normal. c E D can be arbitrarily chosen. where the singlevalued branch is chosen by g(0) = . then the constant is el because ~ ( 0 = lim ) egn(') nvw = e1. we obtain diam f ( D ) 5 2. 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.441 because diam f n .( . Since diam g(D) = g ( z ) does not belong to the family 0. by Caratheodory's theorem. We choose a compact subset K c D such that z.
Let 1 . (Illinois) Solution. then there exist zn E D .442 sup O<r<l /'^ o If(reioe)lPde 5 c. Show that 3 is a normal family. The contradiction implies that 3 is a normal family. Izn . P q 1 1 Then As n + co.201 Cauchy integral formula. .lzol = 3r. < r. Then when n is sufficiently large. If it is not the case. while the right side of the inequality is a constant. B y Hence where +=l. We prove the assertion by contradiction.f n E F such that zn + zo E D and fn(Zn) + 00. the left side of the above inequality tends to infinity. It suffices to prove that the functions in 3 are uniformly bounded on every compact set of {zI < 1).
Show that 1 tf'(t)dt g(w)= 2?ri f(t)w J. Let C be the circle IzI = p where p < T . series expansion of the root of the equation z3+ 32 . define a holomorphic function of w for 1I w < m = m p If(pe")l.443 5500 (a) Let f be holomorphic for 121 < R and satisfy f(0) = 0. ( t ).p n=O t f ' ( t ) dt) w n .w f 1 =5(&J. Hence tf'(t)dt g(w)= GJ. . which implies that g(w)is holomorphic in {w : II w < m}. f(z) # 0 for 0 < IzI < r 5 R. and that z = g(w)is the unique solution of that tends to zero with w. (a) It follows from IwI <m= rp lf(pe")l n that when t E C. f'(0) # 0. Solution.w = 0 that tends t o zero with w. and apply this to find the explicit (b) Find the Taylor's expansion of g(w).
(b) Let f(z) = z3 32.w). we know that f(z) = w has a unique solution in { z : 1x1 < p } . Since z = 0 is the only simple zero of f in { z : IzI < p } .O) = n(r. then f(t). which shows that f(z) and f ( z ) . and * f’(t) = pog(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)]’ = [log(t .w = (t .zi)Q(t) where Q(t) is analytic and has no zero in {t : I1 t < p } .444 Let r be the image of C under f where C is the circle { z : IzI = p } taken once counterclockwise. we assert that the unique solution g(w) tends to zero together with w . + then where . Because lwl the winding number < m = min If(pe”)l.w have the same number of zeros in { z : IzI < p } . As the constant term in the Taylor expansion of g(w) is which is obviously zero. 9 @.z1) f(t) . Denote the unique solution by z l .
( < N } \ { . Because N can be chosen arbitrarily large.3 . Prove in detail that your function has all the required properties.. So f ( z ) is analytic in 1. 2.. 2. we construct For any natural number N . n 2 2N.. a n d z = 1. .l .. .N+lareitssimple poles with residue n at z = n. ..N + 1 } . 5 N } to a function 2N1 n=l which is analytic in { IzI <N}. 2N2 Hence converges uniformly in { IzI In addition. + ..2. . .2 (:)k2 k + '2kk .2 . .with residue n a t z = n.. By MittagLeffler's theorem.:) . we obtain a2k1 a2k = 0 and =  zl 1 33k2 32klt2k1 [ 2 '2k .N 1 with residue n at z = n. 2k t 5510 Find an explicit formula for a meromorphic function f whose only singularities are simple poles at 1.. (Illinois) Solution. . * . z2 is a meromorphic function whose only singularities in { IzI < N } are simple poles at z = 1.445 After some computation. . it is obvious that f ( z ) has all the required properties of the problem. when IzI 5 N .1 (5)k'] dt (3ck2c". .
3. give an explicit formula for g.1 is an analytic function in maximum modulus principle that 1. (Illinois) Solution. )? If Yes. IPn(z) < E holds for all z E (1 < Izl < 2).1. It is easy to see that . Multiply both sides by )zI2. if No. let (z( R and choose N 5 + > R2. (b) Yes. if No. Then when n 2N .1 < 2 } .} such that P n ( z ) + uniformly on (1 < 1.. 5 Because z 2 P n ( z ). The contradiction follows by taking z = 0 in the inequality.} such that P n ( z ) i uniformly on the annulus 1 < IzI < 2? If Yes. explain why not.we have 5 $1 a). If there exists asequence of polynomials {P.446 5511 (a) Does there exist a sequence of polynomials {P. it follows from the holds for all z E { Izl < 2 ) . explain why not.( < 2}. give an explicit formula for the Pn . there exists N > 0 such that when n > N . . For any R > 0. The function g can be chosen as where an = f i ( 1 i). (a) No. then for any E E (0.2. (b) Does there exist an entire function g whose zeroset is {+( 1+ i) : n = 0.
and prove your assertion. If for each positive number n there exists an entire function fn such that then for 15 IzI 5 2..447 Because C n=N 00 5 converges. Hence g(z) is analytic in { z : IzI < R}. False.a2. (q 5512 State whether the following statement is True or False... we know that is analytic in {z : IzI < R}. which implies m is analytic in { z : IzI < R } . and 5 e 1 ~ ~ ~ ( zeRefm(z) < 2e > l =  . and its zeros in { z : 1z1< R} are 0. Define Fn(z) = efn(').al.ak . we have 1 5 Refn(z) 5 1 + l o g 2 . We prove the assertion by contradiction.. For each positive integer n there exists an entire function fn such that (Indiana) Solution. Since R can be arbitrarily large.1 5 IC < $). we see that g(z) is a n entire function with the required zereset. Then Fn(z) are entire functions with no zeros.
But we have for 1 5 IzI < 2. and by Hurwitz's theorem F ( z ) has no zero in { z : 1x1 < 2). Since IFn(z)l 2 for 1 5 IzI 5 2. F ( z ) cannot be identically zero. Because the complement of K.(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. = G . C Kn+lr and nrco lim K. (a) Give a singlevalued definition for zi in G. (a) zi is defined by eil0gz. In domain G.0].48 4 for 15 IzI 5 2. which implies that F ( z ) = cyz with = 1 in { z : I1 < 2). This is a 5513 Let G = D\(1. IF.(z)I 5 2e for IzI 5 2. t contradiction to the fact that F ( z ) has no zero in { z : IzI < 2). For example. and there exists a subsequence {Fnk(z)} converging locally uniformly to an analytic function F ( z ) in { z : IzI < 2). 5 (Indiana) Solution. Then K. such that nrm lim P. is connected and contains z = 00. singlevalued branch of logz can be chosen. where D = { z : IzI < 1). we know by Runge's theorem that there exists a sequence of polynomials which converges . (b) Why should there exist a sequence of polynomials P.(z)} is a normal family in { z : ( z ( < 2). Hence {F. a singlevalued branch of zi in G can be defined by argzlo<2<l= 0. (b) Choose where n > 2. By the maximum modulus principle.
.n2 ' n=l Justify your steps. IPn(z)l 5 M for all z E D and all n. which implies that z' can be extended to a singlevalued analytic function in D. to z i . If there exists M with IPn(z)l5 M for all z E G and all n. (c) Develop a cot az in a Laurent series about the origin directly and by use of (b).} converges to zi uniformly on n compact subsets of G. In other words. = 1.Now we consider (2n)2. ( z ) which converges uniformly on compact subsets of D to an analytic function f ( z ) in D . with enough terms to find the values of n=l C 5 and C 00 00 1 n=l (Hamad) Solution.) is . 5514 (a) Prove that (b) Use this to show that acotaz = Z l o o + c2z 22 . and there exists a subsequence P. then because Pn(z) are continuous on D. Hence {Pn(Z). we can find a polynomial Pn(z) such that for all z E K. .2. so the contradiction is obtained. * .. a2 1 The singular part of f at z = n (n = 0. fl. It follows that {Pn(z)} is a normal family in D .449 uniformly on K . f2. It is obvious impossible. (a) Let f(z) = G. (c) No. the series  .. hence zi = f ( z ) for z E G.. Since Pn(z) converges t o zi in G.
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. Imt+f00 As the convergence of is uniform for lIm4 2 1.450 For any natural number N and la1 5 N. . 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. we obtain the result that 00 1 n=cc is a meromorphic function which has the same singulaxities as f(z). Let Then g ( z ) is an entire function. holds for n 2 2N and n n=2N 5 2N. It is obvious that lim f ( z ) = 0. As f(z) and are both periodic functions with period equal to 1.
we know that is a meromorphic function which has the same singularities as F ( z ) .. Hence we obtain 7rcot?fz= +c1 " 22 z2 n=l . Differentiating both sides of the above identity.n2 + c.n2' . The singular part of F at z = n (n = 0.&2. f l .G(z). Thus we obtain the identity (b) Let F ( z ) = 7rctg7rz.G'(z) Comparing this identity with (l). have G'(z) = 0 which implies that G = c we ( c is a constant). It is obvious that the constant must be zero. we obtain T2 1 " n=l 1 1 .45 1 which implies that g ( z ) is a constant. . it follows from the fact that F ( z ) and 1 .)is the series 1 " n=l  A.n2 + Then G ( z ) is an entire function. Let 22 1 " F(z)= ."+ E n 22 n=l are both odd functions that c = 0. For z2 n=l . we consider Now 22 With similar discussion to that in (a). + n=l c 22 . .
. f of degree .z 1 z 7r2 3 7r4 . rz x . .. = ( z .23 . such that lzn 1 + 00. does there exist an entire function f with f(zj) = aj? Can you write this function down?  (Hamad) ..2 with f (zj) = uj if and only if n (c) Given a sequence of complex numbers z1.. . we can also obtain 5515 Let z1..zn).( z . (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 ..tl) 9 * . 45 (3) It follows from (2) and (3) that around the origin. 5 n . Take z = 0.2 and g(z) Let f and g be polynomials. we obtain After differentiating (4) on both sides.z2. Zn be distinct complex numbers.
. .q ) ( z .. then by (a). ( z . .z 2 ) * .zj . dz dz = lim = 0.( z . n)..z 1 ) . and the coefficient of zn' is 1. z ) g( z .2 with f(zj) = uj (j = If u l .Z n ) is of degree n. Since .. we obtain ?#=O.z j ..Z.z. while f ( z ) is of degree 5 n . (a) Take R sufficiently large such that Zl.l ) ( z .Z j + l ) . < RS.2.(2 .( z .. E 1(.453 Solution.) is a polynomial of degree n .Z2. . j=1 (b) If f ( z ) is a polynomial of degree 1. we have 5 n . . . .2.1. u 2 . Because g ( z ) = ( z .an are n complex numbers such that n we construct the function f ( z ) by For each j ."'.
Zj) . In other words. so that f(z) is analytic in 1. Because lim z+rj .1 I +00. f(z) is a polynomial of degree 5 n .( holds for all IzI arbitrarily large. Since R can be formly for all IzI 5 R. . Because Iz.2. there exists N > 0 such that Iun(z>l 5 Iz. uk(zn) = 0. for any R when n 2 N .I > 2R 2 00 1 u n ( a ) converges uni5 R when n 2 N .n). (c) For the given sequence z1. while for k #j.454 the coefficient of znl of f(z) is zero.’ajz j ) g ( dz) ~ ( z .2..  where 7. . f ( z ) is an entire function. Hence > 0. we can construct an entire function g ( z ) with simple zeros z l . In other words. is chosen such that when IzI 5 k$. z2. by the Weierstrass theorem about the canonical product of entire functions. .. such that . It is easy to see that while for k # n. which implies that f ( z ) is an entire function satisfying the required condition.I + 00. f ( z ) satisfies the condition f(zj) = aj ( j = 1. n=l < R}.aj. z2. Then we define  a.
Part VI Partial Differential Equations .
.
457 SECTION 1 GENERAL THEORY 6101 a) Let A = (a. =  .trace A Wn 1 n n(n + 2) trace A. Here (. b) Show that u E C:(R") implies (Iowa) Solution. lxl<l Vi # j and Therefore. a) It is not difficult to verify that aijxjxjdx = 0. Show that ( 2 .j)&=l be a real matrix.. Ax)dx = Wn L<l n(n + 2) trace A. . ) is the dot product of vectors in R" and wn is the area of the unit sphere in En.
By applying Green's formula. Show that T = const. i. the conclusion is true for u E Ci(IRn). Then it is easy to verify that for any q E C r ( I R ) 5 satisfies the condition ? J. g d x = 0. we get immediately As Cr(lRn)is dense in Hi(IRn). 6102 Let T be a distribution on IR and suppose that T' = 0 on IR.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. IR Take a function p E C r ( l R )such that J. show that there is a number a such that T(q5)= / aq5dx for all q5 E C r ( l R ) .458 b) First let u E Cr(IRn). . T' = 0 if and only if T(q5')= 0. R (Indiana) Solution. Vq5 E C r ( B ) .pdx = 1..
l)).b] c ( 0 . And it is clear that (1 1[ 1 2) . Let {fk} be a sequence in D((0. l)). we have 6 E L 1 ( R n ) . defined above. Then we get immediately + + 6104 Let u be defined for f E D((0.a / 2 E L2(Rn) if s > n/2. u E H s ( R n ) if and only if (1 lt12)'/2G E L 2 ( R n ) . as Ic + 00.l)) by Determine whether u is a distribution on (0. and support your answer.bl. is a linear functional in D((0. In fact. l). It is clear that u.l)) such that fk + 0 in D ( ( 0 . Therefore. (Indiana) Solution. Show that lim u ( z ) = 0. Then we have a compact interval [a. 14+w ( cincinnat2) Solution. vk: . We show first that 6 E L 1 ( B n )if u E H s ( E n ) with s > n/2. l ) such that SUPPfk c [a.459 Hence T($) 0 and = 6103 Let u E Hs(En) with s > n/2.
XY 2 ( .l). in the sense of distributions. u is continuous in D ( ( 0 .l)).< a. + oo. y) I x2 For which p 2 1 does the function + Y2 < 1. u.uy E L1(B). u#.dxdy = J. 1 N Then we have Therefore. ) 5 belong to WIJ’(B)? (Iowa) Solution. + y2)3/2 * It is clear that us.460 and jp). Let N be a positive integer such that . o uniformly on [u. 6105 Let B = {(x.#dxdy .and is a distribution on (0. that is J. Let uy(x7 y. b] as for any nonnegative integer n.
461 and r JB u+ydxdy = ..dzdy =  lc ux4dxdy + 1. v is a distribution solution of (*) in En and only if if .JB uy#dxdy r for all radius E E Cr(B). we get &u = u. Let RE = B\BE for any E C r ( B ) . And we can verify that for which means that u @ W1yP(B) p = 2.. we denote the ball centred at the origin with the By and S. 6106 Q . Therefore. Then we have < 1. Derive a necessary and sufficient condition in terms of v +. x)ds for 1 5 p < 2. 0 in the above inequality. u~cos(?~'. = aB. Letting E . 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 . B.. and Let C be a smooth hypersurface dividing R" into two disjoint open regions R2.and v for ZI to be a distribution solution of (*) on all of R". u E W1ip(B)for 1 5 p < 2.C ( *I in the classical (C')sense where b is a continuous function. v. Assume furthermore that v lies in both Co(?&) and C0((n2)(but not necessarily Co((IRn)!!) with v+(xo) lim v(x). ' BY It is not difficult to verify that + J. Suppose v : R" + R" solves the equation divv(x) = b(x) for x E R" . Similarly L u = 2ly.u4. (Indiana) Solution.
Therefore v is a distribution solution t o (*) on all of B* if and only if (w+ .(IRn). and U1 and U2 are (different) constant vectors.I’ z d a l n a 5 2 2 is a smooth curve.v) uds + lRn divv . Recall the definition of the curl operator in two dimensions acting on a vector field U(z..y) = (u(z. Hence the equation (1) is equivalent to #(v+ . (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. 2).Y = o 6107 on C. 2.J.v) . V#dx  J. vds = 0. V#dx = IRn V# E CF(lRn).v .V) .. we have Ln 2. If U has curl zero in lR2 in the sense of distributions.. V+dx = ..y)): vxuu~2Lyvu. V# E C. (b) Suppose U takes the form 2 2 1 2 where 5 1 and a are open sets such that a U n = B2. V+dx =  +(v+ .y).. By Green’s formula. (a) Let u. describe as completely as possible the curve r.462  b#dx.v E Lloc(B2).. J. Then curl U = 0 . #dx. (Indiana) Solution.v(x.
As the set {< : p ( i < ) = 0) is bounded. where Then curl U = 0 if and only if u1. 4  Construct a function X R E Cp(lRn) such that XR(<) = 1.vz). w1. Therefore.6 E S.v2)dy = 0. v4 E C. v< E {< : < R}.v2)4dy = 0. that and E . WI).(lR”. 6108 Let p be a polynomial in n variables. there exists a positive constant R such that {< : P(i<) = 0) c {< : 1 < R). and assume that the set {z : p ( i z ) = 0) is bounded. I’ is a straight line defined by (u1 . Prove that there exists a tempered distribution E on lRn such that p ( D ) E . Integrating by parts. Set It is clear that E S ’ ( R n )n Cm(Rn).463 in the sense of distributions is defined as (b) Let 171 = (u1.u2)dz + (v1 . u2 and 712 are constants. we have J.%)4dz + (v1 . where S’(W) denotes the space of tempered distributions on B”. and 6 is the Dirac distribution defined by 6 4 = 4(0). (Indiana) Solution. U2 = (uz.(ul . Here S is the Schwartz class of rapidly decreasing functions.
(nZ"). Suppose ~(i<) E ~.(i<>" o for all E # .~ 2 . Therefore..e. . we get  C < P ( g ) G ( [ )= 0.1 E C.a. Let G(E) be the Fourier transform of u.e. From the transformed equation.. i. i. .) and the C..a.6 E where E = F .Transforming the equation.D" bllm where a denotes a multiindex Q = (a1. I4<m If u E S' satisfies P ( D ) u = 0.a2. Let P ( D ) be an m t h order linear constant coefficient partial differential operator on En. are constants. we get p(D)E. .) (Indiana) Solution. are constants.^(€I = + c a. . we can see that suppG = { 0). Therefore. . G ( t ) must be finite linear combinations of the derivatives of the Dirac delta function S(<) at { 0 } ..' ( Z ) E S'(nZn).) is a multiindex and a. l4lN where a = (a1. Hence is a polynomial. prove that u must be a polynomial. (Here S' denotes the space of tempered distributions on an.S(")(<).464 It is easy to see that p ( i € ) Z ( € ). P(D)= C.(0). 6109 s.
Further. Then we have + 0 . c ( . (d) Calculate the Fourier transform of the distribution (Indiana) Solution.we have Then it is not difficult to verify that Hence t+O+ lim  & ex2/4t 6 . (c) Calculate rigorously the derivative of log 1x1 and express the result in terms of your answer to (b).a . Let supp4.e..( 1. (b) Show how to define as a distribution (i. (a) It is wellknown that 5. 4 E D. suppose that q5n in D. as (b) We define the distribution follows: 5 by its Cauchy Principle Value &: The linearity of the functional Vp: is obvious.465 6110 in D'. a ) Vn. exaf4tdx = 1 Vt For any > 0. define the Cauchy Principle Value and show that it is in D').
is a distribution in D . 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.466 Therefore.we have V p i Hence &log 1 1 = V. Therefore.DcIu(c) 2 €mn . ' ( c ) For any 4 E D . 2 (d) By F ( f ) we denote the Fourier transform of a tempered distribution f For any 4 E S.we have Here.:.
and a is a multiindex. t > 0 00 < 2 < 00 U t ( 2 . 2 is a fundamental solution of the wave operator W ( u )= utt . 00 (f. If L*S is dense in S. < 2 < 00.g. = #J(z.. b) We need only to show that In fact .u.)1. where S is the space of rapidly decreasing functions. O ) = g(z).467 where the abs are constants. t ) = I+(t)l+(t.u. Let L* be the formal adjoint of L defined by L*V(C) = lallm c (l)~aba(aaV(z)). on E 2 . where S(z) is the Dirac function. #J sufficiently nice) in terms of the above fundamental solution. (Indiana) Solution. Simplify your result to obtain D’Alembert’s solution of this problem. u ( z . then the operator L has at most one fundamental solution in S‘. What is a fundamental solution of L? When is it unique? b) If I+denotes the Heaviside step function Show that 1 F ( z . 0) = f(z).t).. c) Express the solution of the Cauchy problem utt . a) A distribution E E D’(nZn) is called a fundamental solution of the differential operator L if L E ( 2 ) = S(z).
the solution of the original Cauchy problem can be obtained as .468 The solution of the Cauchy problem is given by Therefore.
where G and denote the Fourier transforms of u and q5 with respect to z. we have ztt+ i t 1 4 6= 0. By applying Fourier transform to the problem. we have Set 77 = tit in the integral in the right side of the above equality.469 This is just the D’Alembert’s formula when q5 2 0. it yields that 3 %t) = Therefore. It is reduced to .0) = 0. and 4 is a C” function on IR3 with compact support. Assume that there is a constant C such that (Indiana) Solution. q t . respectively. qt. 6112 Let u be the solution of the initial problem where summation is understood in the pde. By solving the above transformed problem.0) = w.
(Indiana) Solution. given in the problem. then there exists a positive constant r such that I l2 a. m 4 And when t > 5 . if I&(O)l = 2a # 0.we have This contradicts the condition satisfied by the solution u. 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. K E E3.470 This implies that In fact. So we have And then . consider the solution u of the SchrGdinger Equation Show that lu(z.t)I2dz = I&0124 Here 3 denotes the Fourier Transform of 4.
we get Here the Fourier transform of 4 is defined by .47 1 Set x = 2&.
Schwarz's inequality gives that is JT u 2 ( x ) d S z >_ w. (Indiana) Solution. Applying the meanvalue property for the harmonic function u. is the surface area of the unit sphere in En. we have where w . = 0 on Au R". There exists a point xo E En such that u ( x o ) # 0. Show that does not exist.472 SECTION 2 ELLIPTIC EQUATIONS 6201 Let u $ 0 satisfy u E C2(Rn). zz*(=p Then we have u2(x)dx wnu2(xo)Jlo pn'dp w 2 = u n ( x 0)T n . n .pn1U2(x0).
.(f% < dist(z. 6202 Let u E Co(R) be weakly harmonic in an open set R. (Iowa) Solution.473 for any T > 0. the relation uAddx = 0 holds for all 4 E C. The conclusion follows as T + 00. is harmonic in K if E < u. an).(R). Therefore. Show that u is then harmonic in R. Therefore u is harmonic in R. u. +u uniformly in K .. set PE(Y) = EP Then it is easy to see that provided E 2Y ( ) 7 P4Y) E C. an). Set where C is a constant such that For x E R. we have where This implies that for any compact set K dist(K. i. as E + 0. It is wellknown that c R.e.
harmonic in z E lR2 and continuous in y E BIZ.Show that f is continuous in (2. Then the conclusion that f ( z . . For z. (Indiana) Solution. as a function on y). Applying the meanvalue theorem to the harmonic function z. Prove that u E 0 on z 2 0.z0l2 + Iy .e. y) be a locally bounded function.474 6203 Let f (2. These estimations imply that f ( z . . 2 0 of En. yo) E B4and R such that lf(z. (iii) u is harmonic in z > 0.y)I > 0 be given. y ) is continuous at z = z0 uniformly with respect to y near yo. i. we have By Green's formula.yo) E B4follows immediately..y) E . V(z. Let (zo. we get we have the same estimation. 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. y ) is continuous at (zo. B4. 1 2 5 M. (Indiana) . Then there exists a constant M > 0 lR4. .y0l2 < 2R2.
. > 0) and continuous on y the boundary alR. By using the uniqueness theorem of the Dirichlet problem and the Poisson’s formula of the Dirichlet problem in a ball. n 1. we can verify that the redefined function u is harmonic in any ball centred a t the origin in lRn. are bounded in show that its “tangential” derivatives magnitude by maxlVgl on all of B = {x : xn > 0). If u. is the surface measure of the unit sphere in B”. By the uniqueness of the bounded solution to the Dirichlet problem (3). < O where x’ = (xl. we get u s 0 immediately. (Warning: you must T justify any assumptions of regularity you make on the solution u.e. Thanks to the Liouville’s theorem. we can show that the unique bounded solution to the problem (3) is given by e. ~ ~I‘ 1 )([1. = 0 ) . VX.bounded. satisfies j = l . where X’ = ( ~ 1 . < 0 as a n odd function of u(x‘. and let g(x) be a compactly supported.(.475 Solution. C1function on alR: = {x : x.xnl).1) and w. . = u(x’. u is bounded and harmonic in B”. = {x : x.2n) 2: . Redefine the function u in the halfspace x. = . 6205 Let u(x) be C 2 on the halfspace B = {x : x. * .x~).* .) (Indiana) Solution. It is not difficult to verify from the above formula that and ..**. Therefore. = 0)..
(Here u = u(x. ( Cincinnati) Solution.lY I < 1 Integrating the above integral by parts. au au = 0 ax ay for IyI = 1 has at most one solution. y) is a function of two variables x and y. < 1.476 Therefore. and is a classical solution). It is easy to see that lXl uAudxdy = 0.au = 0 for IyI = 1 ax ay has the unique solution u 0. IyI u = 0 for 1x1 = 1. . we have which are just we want to show. au . < 1. we have uAudxdy . 6206 Show that the problem Au = 1 for 1x1 < 1. < 1. IyI u = 0 for 1x1 = 1. _ . We need only prove that the problem Au = 0 for 1x1 < 1..
u2(1. (b) Suppose a1 # a2 are such that nontrivial real solutions u1 and u2 satisfy (*).477 = (u2(1. and suppose nontrivial solution exists to CY is such that a { V2u = a 2 u in R. . 1) 1 2 + u2(1. (a) Multiplying the equation in the problem (*) by the complex conjugate of the solution u. which implies that a is a real number.1) . and that u can be chosen to be realvalued. Show that r (IndianaPurdue) Solution. 6207 Let R be a bounded normal domain in lR3. and integrating the resulting equation in R. we obtain (Vu12dw= a2 IuI2dv. (a) Show that a is a real number.u y 1 .1) . u=O on a R . 1)) Then we get u f 0 immediately. Then it is clear that u can be chosen to be realvalued.
6208 Let fl c B“ be an open set. respectively. (Indiunu) Solution. we get immediately which implies the conclusion of the problem. . one has u 5 h in B. By a standard argument.478 (b) Suppose that equation satisfied by u1 u 1 and u2 are chosen to be realvalued. Multiplying the V2u1 = a9u1 in R and the equation satisfied by u 2 v2u2 = aiu2 in by u2 and can obtain u1. and integrating the resulting equations in R. Vu2dv = a: and From the above integral equalities. (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 .r ) cc R means that the closure of B is contained within R.) (ii) For every ball B cc fl and every harmonic function h E C o ( B )such that u 5 h on d B . Let u be a continuous function on a.Prove the equivalance of these two notions of “subharmonic”. 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 Vul . Let u be subhamonic in the sense of (i).
the above equality can be rewritten as . :Y E lVU(. a) The conclusion is clear. where B is the open ball centered at x of radius r . b) Let B c a. r ) .)l where d ( z ) = dist(z. we have au n (z) =dXi in the wn r n where w. Conversely. hence w. we have This completes the proof. SUP~lU(Y)l 1% : 3 : n . L SUP{lU(Y)l 4s) 0) (Indiana) Solution. b) Applying the meanvalue theorem for the harmonic function ball B . Then w = u . is the surface area of the unit sphere in En.479 Let h be the function given in (ii). Hence that 5 0 on 8 B implies that w 5 0 in B. Show 2 that IV4z)l L. and let u be harmonic in R and continuous in R. a) Show that is harmonic in R for each i. if R is bounded and n E R.YI = 71. /n is the volume of the unit ball. c) Show that. By using the Green's formula. { From the definition of subhamonicity in the sense of (ii). We define the function h as follows: Ah = 0 in B ( z . 8 0 ) . h=u on a B ( z . let u be subhamonic in the sense of (ii). and the meanvalue theorem. r ) .h is also subharmonic in the sense of (i).Let R E B" be an open set. 6209 .
(Indiana) Solution.sin(kr). Let T = 1x1. For this ordinary differential equation. By the fundamental solution cos k r . = ( z . we have c) The conclusion can be obtained from the result of the part b) immediately. Derive an appropriate meanvalue property for the solution. ds2) be a fixed positive constsant. T r Let u be a solution to the equation A u + k2u = 0.R) we denote the ball centered at zo with radius R. it is not difficult to verify that & where T. Then the spherically symmetric solutions depending only on T satisfy (TU) d2 dr2 + k2TU = 0.2'1. First we look for the fundamental solutions of the equation. By B ( z o . Applying the above integral formula in . 6210 Let R C lR3 be bounded and open with smooth boundary.cos(kr). Let R < dist(zO. there are two linearly independent solutions: 1 1 u = . . Let u E C1(n)n C2(R) solve A u + k 2 u = 0 in R (k > 0).O.480 And therefore.
= 0. Prove that its spherical mean.we obtain u(x0) = a . It is easy to see that the function g(z.R2) satisfy AU = u + 1.R).+ k2)g(z. R). For this function.zO)= cos(kT.rrRsin(kR) J BB(xo.o.z0) = cot(kR) sin( kr. zo) 1 47Frxo. u(z)dS. it is easy to verify that Subtracting the equation (2) from the equation (l).cot(kR)1 sin(kr.zo). z E B ( z o . 2 g(z. E aB(zO.o..481 the domain B ( z o .R). defined by . we get Now we look for a function g(z. Txox  4. which satisfies (A.).R) This is a meanvalue theorem for equation A u 6211 + k2u = 0.) rxox satisfies the above conditions.o. Let u E c2(.
W(T) # 0. Therefore + The proof is completed. and Applying Green's formula. Here.y) I x2 + y2 < 1). B ( 0 . Solution. 1 u=O onaB .also satisfies ( l ) . Set T = 1 1 Then we have 2. r )denotes the ball of (Indiana) 1 =21r L B ( 0 .482 and v(0) = u(O). 6212 Let B = {(z. for z radius T . we get Then it follows that = r(w(r) 1). dSy the element of arc length. l ) u(rw)dS. Prove that the problem Au= in Jm B.
g and 80 are sufficiently smooth. Show that the problem (1) has a unique solution u E W1!2(R). gvda v v E w'q(n) Vu. This implies that f E H . (b) Will the conclusion of part (a) be valid if X = O? Explain. (c) Soppose that f . (Iowa) Solution. 6213 (a) Let R be a bounded domain in lRn with smooth boundary f E L 2 ( f i ) . Then by applying Cauchy inequality and the trace theorem in Sobolev's spaces. Hence the problem admits a unique weak solution u E Hip). = h[(gradu) (gradv) v)1  + Xuvldz.  . and (1) L ( g r a d u ) . And denote the corresponding norm in W't2((n)by 11 ( ( 1 .' ( B ) .v E W'y2((n). Let X v= d 3 w = &5qj5 and f = 1 &Gp' It is easy to see that v . (a) Define the inner product in W172(f2) by (u. (gradv) + Xuvdx = where X > 0. w E L2(B) and f = a v . Consider the problem aQ. ( Cincinnati) Solution.483 has exactly one weak solution u in HA(B). fvdx 1.and g E L2(dR) be given. write problem (1) in the classical form.a w + ax dy in the sense of distributions (see the solution to the problem 6105). we get + J.
the classical form of the problem (1) is Xu = f in R. there exists a function u E W'?'(R) such that F ( v ) = (u. F(v)= fvdx +lngvdo is a bounded linear functional in W'?'((n). ( Cincinnati) . (b) By the classical form of the problem. and assume that U* are regular enough. z n . Denote u+ = u Inn~X. v v E W'J(R). Thanks to the Riesz theorem. =o} . V#€H. Integrating by parts.. Au dU ..l . z n ) = (x'. (c) Suppose that f . g and dR are sufficiently smooth. azn du+ .1 on R n {zn = 0). Therefore.= g on dR. Show that Au* = 0 and that du ax. an + 6214 Let R be a unit ball in IR" centered at zero. x.1(R). V' ). it is easy to see that the conclusion of the part (a) is false if X = 0 even for smooth f and g.44 8 where C is a constant.).<o}. The uniqueness is clear. Prove the existence of a unique solution of that problem. Consider the following problem Find u E Ht(R) such that lVu. Therefore. where IR" 3 2 = (21. Jnn {X .o} and u.= u Inn(z.Vq5dx = a . Then u is a solution to the problem (1). WX'.
. : where I(. we can get I Cll4ll1.11 is the norm in Hi(R).=o} 4d~'. . The proof is completed. < 0). ax. From the above equation. n { x . > 0) and R. there exists 1 (%4)1 u E Hi(R) such that = / nn{ I=o} .= R equation by parts.V 4 d x = where R+ = R n {z.(Q). nn{ x . we get immediately that AU* = o in a&. . 4 d x + nn{ xn=o} (E. Define the scalar product in Hi(R) by By the Cauchy inequality and the trace theory in Sobolev spaces. Write the equation in the following form Vus V 4 d x Vu. =1 on Q n {x. Mx'. This proves the existence of the solution to the problem. = 0). we have + 1. auax. v4 E m Q ) . ""'> / dx' = au+ ax.485 Solution.n{X. =o} 4dx'. Assume that u* are regular enough.a x . According to the Riesz theorem.4dx  l Au. The uniqueness of the solution to the problem is clear. Integrating the above  l+ 1 Au+ . V4 E H.. + L  J.
’”(R).’”(R) and Now we have Vu V4dx =  J. the proposition is true. 4 E W. By the PoincarC inequality with the above estimate. we obtain Hence 4 =0 This implies that u 2 I a. Let an d(z) = max{l Then it can be verified that . From the above inequality and (I). in R . 0).VU .u. show that 2 inf u an (essinf) ( Cincinnati) Solution. in R. a. 4 2 0. b) Let u E W1?’((n)satisfy (I) above. IV412dx 5 0. in R. b) If inf u = 00. it holds that IV+(2dx= 0. Show that u 2 0 a.486 6215 J. an Assume that inf u = I > 00. a) This is a corollary of the conclusion in b) of this problem.e.V 4 d ~2 0 inf u n V 4 E W.e.e.
} is bounded in L2(R) and w. and integrating it on R. as n ( Cincinnati) Solution.. w E W292(R)n W. w. that such Awn + u .. Vw.'12(R). Prove that '11. we have . respectively.dx . both equations above can be written as Vu. we get By using Green's formula and noting that u.v. and then integrating the resulting equations in R. Multiplying the first equation by v. Therefore. = 0.11i2 Then we have  = 0..nw. . + = w.'72((n). in R.n11w. and w. E W. + o strongly in ~ ~ ( 0 ) . Multiplying the first equation and the second equation by u. we obtain {u. w strongly in L2(R). + +oo. Au.487 6216 Consider functions u.
IIVu. = 0. 6217 Assume R c Rn is a bounded open set with smooth boundary and let f E cF(R). we can get . .dx. By noting (3) and (4). Vundx. This gives us the conclusion of the problem.. This implies that {un} is bounded in W. 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.AunAvndx + nllVun11i2 IIVUn11ia Hence + nllVvn11iz = k VW.dx. Av. Multiply the first equation by Av.Vw Vv. it follows from the above equality that Av. it follows from above equality that Multiplying the first equation and the second equation by Au.. .12(s2). and Av. (Indiana) + . +0 strongly in L’(R). uk(x)2dx 5 M for k = 1. finally (4) By similar consideration... respectively.~ ~ A . we can get J.488 By noting (1) and (2). Vv. and then integrating the resulting equations. ~ ~ ~ ~ v Vu.dx  . J.2.Il$ = LwAu.Au.dx =  J.Suppose uk E cm(n) satisfies Auk +uk = f and has the property that for some positive number M .
" ' . see the solution to the problem 6210). 2 . For any z E 21. we have + h(T)vk(z) = 1 vk(Y)dSy. we have u E Cm(R). v x E 0 1 . k = 1 .u1. Let R l be a subdomain of R such that 2 CC R and R = dist(dQR1) 1 fixed. Set vk =w k + z k . r ) we denote the ball with the radius T (< R) centered at z.r) T. By the regularity theorem of the elliptic equations. By the meanvalue theorem for the equation Au u = 0 (for the case n = 3. 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 ) . It is clear that v k satisfies A V k + V k = 0 in R and Hereafter. by B ( z . As { u k } i bounded in L2(n). where W k and z k are defined by and . Therefore. where h(r) is a continuous function of 0 to R with respect to T . aB(X.489 Solution. we find that Integrating the above equality from Here R can be chosen so small that Ivk(z)I 1 h(r)dr # 0. Set V k = u k . it holds that : (1) 5 c. and therefore in D'(R). Then Au + u = f in the sense of the distribution. Now we show the second conclusion.
n+w lim I ( u n )= inf I(w). wed (b) Show that L V U V ( W u)dx 2 0 VV E A.. 6218 Let R be a bounded Lipschits domain in B". and A = {v E Hi(R) : h l 5 v 5 h2 a. . (a) Let {un} be the minimizing sequence of I(w) in A.lR. Then by using the AscoliArzela theorem. V k = 1. there exists u E A satisfying I(u) = minI(v). * * . VEd Define the norm of w in Hi(R) as .a . (Iowa) Solution. we can get from the estimation (1) lVWk(Z)( <_ It is clear that Igix 1(1 Z. k) a 1 c. i.e. we can obtain from the above estimation IVzk(x)) 5 C in any compact subset of R1. And by applying the meanvalue property for the harmonic solutions. (a) Show that if A # 8.respectively. are given smooth functions. vx € ill. where hl. . in R}. we prove the second conclusion of the problem immediately. 5 c. k = 1. By applying the solution formula of the Dirichlet's problem for the Poisson's equation.2.2. h2 : R + ..e.
 6219 Let R c IR”. of {u.denoted still by {u. (b) It is easy to verify that u t ( v . Let u E C1@) be harmonic in R. there exists a subsequence Iull. * Then F ( t ) has the minimum at t = 0. This implies that I(u) = min I ( w ) . in R.u)dx 2 0. n > 1.u)) V(u+ t(w .}. 0 5 t 5 1. . The proof is completed.}. . we get u E A and that by the weak lower semicontinuity of IIwII1.e. Therefore. for simplicity we still denote it by {u. and in H.}.u))dz. Let r + v E A. V(u t ( w . be a bounded domain with smooth boundary.491 Then we get the boundedness of { I n l } Therefore.}.such that u + u a.u)E A.~(R) u t u in L ~ ( R ) .such that u u . w F ( t ) = J. And there exists a subsequence of {u. This implies r Vu V(v . wed The conclusion of (a) is proved. . Therefore we have + F’(0) 2 0.
@I =u = on dQ. @I(. a. For the function @2 . n an i.u(z0) 1% . x n an (Note: In both parts a and b of this question. %. (Indiana) Solution. you may cite.492 a. Prove that ma IVuJ= rnax ( V u ( .).u. . @ 2 E C 2 ( Qn C1(t) such that ) A@I 5 0. it holds that sup(a.u ) 2 0 and @2 in R . The conclusion is an immediate corollary of the following inequality: n i.201 I u ( x ).. u ( z )5 Therefore. Suppose there exist functions @ I . Prove that m3x l v u l s rnax rnax lVaI1.j=l b.2 . rnax l n an an AQ2 2 0. any standard maximum principles you are using. vz E s2. ( ~ m 2 l ) .u = 0 on 8Q.u)= 0.) b.e.u) = rnax(a2 .@2(z0) ) x . From the maximum principle. without proof. we have A(Q2 . it holds that @2(2) vx E s2% . () Similarly. we have 5 u(z).2 0 ) This inequality implies the conclusion of the part b.
then there exists xo E R such that u takes maximum at to and u ( 2 ) > 1. (a) If u 2 1 is false. (Indiana) Solution. ~ ~ 2 U Therefore. For the maximum point xo.. i.0 5 u 5 1. Set w = IVu12.u) in R. (a) Prove that for any solution u of (l). the only solution of (1) is u (1) 0.e.493 6220 Let R be an open subset of Rn.j=l u2izj +~ u ~ I V 0. Au(z') 2 0. Prove that IVu(z)I 5 1 for all x E R. It is easy to see that w E C2(R)n C'((n) and n Aw = 2 3. do. it is clear that Au(zo) 5 0. . w takes its maximum on (n a t some point on dR implies that w 5 1 in R. and dR sufficiently smooth. ( Cincinnati) Solution. Hence w 5 1 on 6221 Consider (a nonlinear) equation Au = Xu2(1 . u = 0 on dR. (b) Show that if X > 0 is sufficiently small. 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. where X > 0 is a parameter.
494 This contradicts the assumption that x u y 1 . it holds that Applying PoincarC inequality. If X < Cl. (b) Multiplying the equation by u and integrating the resulting equality on R. Hence u 2 0. (IndianaPurdue) Solution. we have r 0 = JnA(Au)uda . u can not take minimum in R unless u f 0. dn is the trivial one u E 0. As Xu2(1 .u3(1 By using the conclusion of (a). 8222 Let R be a bounded normal domain in IRn with smooth boundary dQ. we obtain from the above inequality where C is a constant. Let ?i' be the exterior unit normal on dQ.U)(ZO) < 0. 1vul2dz = x J. dU .= 0 on dR. we can get J.u) 2 0. Assume u E C4((n). Show that the only solution of the boundary value problem for the biharmonic equation: A(Au)=O i n Q u = 0 on dR. Hence u 5 1. we have u E 0 immediately. Integrating by parts and taking the boundary conditions into account.
Taking into account u = 0 on immediately. af2.495 = lA~1~dx. we get u 0 . Therefore. u is harmonic in f2.
W d y . 03 lf(”)I2d2 < 0O. Show that there exists a constant C such that for all 00 < 2 < 00. we have Set y = 2 &q immediately + in the second integral of the above inequality.t) = . Then we get lu(2.t > 0) (0O < 2 < m).r nf ( y ) e . J 2 m m By the Cauchy inequality.496 SECTION 3 PARABOLIC EQUATIONS 6301 Consider the Cauchy problem for the Heat Operator in ut = u.. (Indiana) Solution.t)l c I t1/4‘ .O) = f ( 2 ) < 2 < 0O. where f is bounded. (0O IR1: u(2. t > 0. and satisfies J_. continuous. The solution to the problem is given by the Poisson’s formula u(2.
00). we can get The PoincarC inequality gives IIU(*. m)) is a solution of the equation with u = 0 on a R x [O. we have . For any 5 CIIVU(*.t)ll:z(n) where C > 0 is a constant.t)2dz. Multiplying the equation by u and then integrating the resulting equation on R. (Indiana) Solution. t)l&(n)? E > 0. it holds that Taking E = 1 / C . Prove that $ ( t ) + 0 as t 4 +m. Assume that Let $ ( t )= 1 n u(z.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.
a) Using the Fourier Transform.) 0 as t + +m. 4 2 . ut where f is bounded and continuous. (00 < 2 < 00.t)ll&n) .It is not difficult to verify that if l/f[.t > O ) .Show that ta. Transforming the following Cauchy problem with the initial data S(z): ut = u .t > 0) 4 2 . d) Show how to solve the problem > 0 .t ) = 0. Why is the Tychanov nonuniqueness example possible? c) Suppose in addition that f E L 1 ( B ) . This completes the proof. and write down the solution u. (Indiana) Solution.0) we get . . t ) = 0 (t > 0) ut = u . t > 0) u(z.0) = f(z). lim u(2. .. construct explicitly a fundamental solution for this problem. < 2 < 00. 6303 Consider the Cauchy problem = u.t) be the Fourier transform of u ( z . (2 by using a) and an appropriate extension of f. ( = 6(2). a) Let G(<. b) State a maximum principle for this problem.O) = f ( z ) ( 2 > 0) U ( 0 . . t ) with respect to z.
t ) s is bounded and continuous on lR x [O. t ) to the general Cauchy problem with the initial data f ( z ) u(z. d) Defined an odd function 4(z) as follows: 4(.T]with the initial data f ( 2 ) .and satisfies the heat equation in lR x (O. Then the inverse Fourier transform of 6 gives the fundamental solution t o the From this fundamental solution.t) = 4(y)eWcly. XER The Techanov nonuniqueness example is possible for some unbounded solutions.t > O ) .) = { f(x>. Then by solving the following Cauchy problem: ut = u x x (cm < 2 < cm.t ) = etzt. t ) E lR x (O. t ) 5 sup f(z) for ( z . 4 2 .T]. Then it holds that XER J inf f ( z ) 5 u ( z . 2 2 > 0. O )= 4(z). we can get the explicit representation for the solution u ( z .499 The solution to the transformed problem is given by G(<. c) The conclusion is clear. < 0. we get the solution to the initialboundary problem 6304 Consider the problem . f(z).T]. 2n21 b) The maximum principle can be stated i ~ follows: Suppose that u ( z .
From the Poisson's formula for the Cauchy problem of the heat equation. t ) u t ( 2 . T).(z. Find a bound on ': 1 Ju..t ) d z d t .Ut = u. . we have And therefore. we have 1' 1: tlut(2. .t u(2.(z.O) = uo(z) E > 0) L2(E). (z E IR. in (Indiana) Solution. and integrating the resulting equations in x (0.t)I2dz t > 0. we can find that Then by using the Cauchy inequality. Multiplying the heat equation by u and tut respectively. we can obtain and Adding the double of the second equality to the first one. it holds that We can also get the result by the method of energy integrals.t )12dzdt = I'L: tu.
t> .ut + a(2..t ) u = f(2. forming the top of ~ R T and let B = BRT\B~. we get an estimation 6305 For the indicated domain R (interior of the parabola) let RT denote the points ( z . ut E C o ( R U B1) be a solution of ~ %x . can From the above two equalities.T)I2dz +T Therefore.(z. t ) E C o ( n ~with . t ) with z E R and t < 2 Let B1 denote the open linesegment ' . 2 ) u.501 otherwise an approach to the function U O ( X ) by a sequence of functions in Cr(En) be used.. T ) ~ ~Ld_zlu.Let u ( z . we have A/m 2 w oo ~ u ( ~ .
.tO) a(zO.tO) 0.tO) 0. f 2 0 in RT? (Iowa) Fig.t0) 0 < 5 and u. 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. Suppose that u assumes its negative minimum at ( ..f E Co(D)..C. u(z.6. then every nonconstant solution u assumes its negative minimum (if one exists) on B2.(zO.to)u(zo.t)= 0.t) + (0 5 z 5 A and t > 0).C.we can say that every nonconstant solution assumes its positive maximum (if there exists one) on Ba. Then we have u(zO. ut(z0.t o )5 0. > + This contradicts f(zo.t 0 ) . What can you say if a < 0. If a < 0. f 5 0 in RT.(2 0 .ut(z0. 6306 Consider the boundaryvalue problem B. I.tO) 2 0. (Here a < 0 should be assumed in RT u B1). ut = 0 u(0. (IndianaPurdue) . UP(. f 2 0 in RT. u. Show that if a < 0. Hence u. = 0..O) = f(z). t o ) E RT U B1.1 Solution.. G O .502 with a.
This implies that the problem is not wellposed in the space of continuous functions with the norm defined above. if not.x .(z)T. is clear that It But for any fixed t > 0 IIu.(t).t > 0) Assume uo E S and find u. 0) az.) = iF(u.r].) (Indiana) Solution.(. leave your result in the form of a onedimensional integral.r.t ) we denote the Fourier transform of u with respect to 2.) d at 2 (<G) a at = u I. One is required to solve u = u . When n is even integer.  a. f n ( z ) is the eigenfunction of the corresponding twopoint boundary value problem. in which the particles of Brownian motion are subjected to an elastic force. t)ll + 00 as n + 00. 6307 In Probability Theory one encounters the OrnsteinUhlenbeck process. u u(z.503 Solution. By setting u = f. (You should be able to find u explicitly. . at . By G(<. = uo(2). (x E t . we can find the solution u to the problem with the initial data f(z) = fn(z) . for all f E C[O. It is easy to see that F(zu.
Solving the above Cauchy problem for the first order linear partial differential equation.+of ax. . we can obtain . In fact. Therefore. we get the solution to the problem Remark.t u o ( e . 6308 Let L be the usual heat operator L = a let at (: .504 Then the transformed equation and initial condition are in the following form qt. From the above solution formula. set T 1 = (I .2 t ) . Then the problem is transformed into the following 7 =0 : 21 = Uo(t). u(<. we can find a convenient transformation of variables to the problem. It is not difficult to verify that F'(iio(ett)) = e .t z ) and Fl(eli(e2tl)€2) = 1  d m e 2(e 2 . Consider the problem . t ) = et3(e2'. a2 1 ' 4 E D(nZ"). and let h be a fixed point IR".0) = G o ( t ) . 2 t = etx. The above solution formula can be obtained easily from the Poisson's formula for the Cauchy problem of the heat equation.1) . I +...e .l)caiio(et<).
t )= u(2 .t)llL~(R")< e(1R2)tl141~L2(Rn). € ) ) t . 2 E B" 2 E B". i& + ([(I2 . but you need not prove that it represents a smooth solution. t 2 E En L which admids the following solution qt.e"n€)G = 0.> 0. (Indiana) Solution.t > 0.O) = m.h .O)= 4(2). < E En. u(2. t ) of the above problem in terms of the initial data. Derive an explicit representation for a solution u ( z . (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 . a.) b.t ) . the above problem we obtain { O(<. .t ) = &(()(y i h .505 ( L 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. Let G( t ) be the Fourier transform of u with respect to 2. Transforming (. t ) . for an unknown function u ( z . t ) to the problem This is just the conclusion of the problem. a. Show then that the solution u satisfies Il"('.
t ) 2 0 is a function with u E C2(o [O.506 6309 Let R c IR" be a bounded domain and assume u ( x . Prove that we can find a constant C such that E(1) = I u 2 ( x . Using above estimation. oo)).. (Iowa) Solution. By using the Green's formula and noting the boundary condition.0).  AU = U 4 (heat conduction with heat loss due to radiation) with the boundary condition u Ian= 0.1 ) d x 5 C regardless of the initial value u ( z .e. . we get from above euualitv E(t) 2 dt I d where E ( t ) = & u 2 d x .which solves the equation x U+. The Hijlder inequality gives i. we get from (1) I d E(t) 2 dt 5 JRJ+Eqt). Multiplying both sides of the equation by u and integrating on f . + 1 IVu12dz =  1 u5dz.we have 2 utudx  Au u d x =   I u5dx.
T]. + if u(zO. Therefore. 6310 Let satisfy ut u3 u = h.u. If the conclusion of the problem is false.A+u)(zo.tO) 1 .sin 1 < 0. t o ) and u(z0. then there exists (zo. we must have But from the equation we get (ut .t. r ) x [O. we have Ef(t) It follows that 2 E+(O) + 3(R(.T) x with g. to) > 1. TI.t o ) 2 1. u= 9. B ( O .to) 0 2 and Az(zo. (Indiun u) Solution. = < 0.507 The above inequality can be written as Integrating the above inequality from 0 t o t . T ) x (0) W O . Prove that maxu < 1 on B ( 0 .T]such that u takes the maximum a t ( z o .tO) 1. It is clear that ut(zo.T ) x (O.to) 5 0. B(0.to) sinu(zo. if u( zo. + + sin(u) = A.tO) = (u3 sin(u))(zo. This completes the proof.T ) x (0.to) E B(0. h continuous and g. LO. TI. u3(z0. .to) This contradicts (1). h < 1.
(d) Show that for f 6 L4(R) n H1(R). we obtain from above integral equality Then we get the conclusion immediately. By applying Green's formula. (a) Multiplying the equation by u. the following bound also holds: I I 4 4 ( n . (Indiana) Solution. In fact. udx  u4dx. IIUIIL(n)(t) L llfllL(n) for all t E (0. Hence (ut . if u take the positive maximum a t ( z o . u=o z E d R . (a) Prove that IIUllLz(n)(t)5 IlfllL.yn) 5 + IlfllLl(n).and integrating the resulting equation on 0.(n) for all t E (0.T].Au)(xO.to) 2 0 and Au(zo. t o )E R x (O.T]) a solution to the problem n be { (b) Prove that ~t = AU .508 6311 Let u E C2(R x (0. we get Au . 2 . (b) First we show that u can not take the positive maximum in R x (O.to) 5 0.u3 2 E R . t > 0: u(z.O) = f ( z ) 2 E a. (c) Prove that the solution t o this boundary value problem is unique.T]. where R is a smooth bounded domain in En. t > 0.to) 0. ( t ) + II~llLl(n)(t) Ilfll. TI. T)) Co(nx [O. TI. then we must have ut(zo.
Tl. we have Il. (c) Let u1 and 212 be both solutions t o the boundary value problem. (d) Multiplying the equation by ut and u3 respectively. u3dx . u 4 d x + J. and integrating the resulting equations.1 3 ~ ~ I V u (. we can get Il4lL. we can show that u can not take the negative minimum in R x (O. Therefore. Adding the both inequalities above. In a similar way.llL(s2)(t> L llfllL(ct).This contradicts the inequality u3(xO.J. 1 + u ~ )2 w . w ( z . we can obtain f and (a Lu'dx) + dt (1 J. x E R.and the uniqueness of the solution is proved.(ct)(t) 50. By applying Green's formula to the first terms in the right side of the both equalities above.Au . vt E (0.u3utdx and utu3 = J.d u'dx <_ 0. The w solves the following problem: { Wt = AW. lVu12dx) = .tO) < 0.u'dx) dt 4 5 J. By a similar deduction as done in (a).utdx .(u: + ~ w = 0.l u i d x 5 0 d = . t > 0.J. we get 1 utdx = J. ~ x E a R . vt E ( 0 7 1 for above problem.u'dx. This implies that u1 = u2. Au . and w = u1 . we have dt ( / . 0) = 0.T]. ~ x 5 (1 J. t > 0.u2. . IVul'dx) 5 0. x E $2.
. respectively. A(p) be a real positive definite n x n matrix with let entries aij(p). 6312 For each p E En. furthermore. we find w 5 o on I x [0. Set w = u . co). Assume A E C ' ( B n ) . co). This is the result we need. implies the conclusion of (d). Suppose. that u and w are continuous on and satisfy the condition u 5 w on the parabolic boundary n x [O. 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 ( ~ .510 This differential inequality. t )R x ( 0 .v and Then from the given differential inequalities. ~ ) E and wt 2 aij(Vw)vzizj for ( z . combining with (a) if necessary. 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. Let R C B" be a bounded open set. t ) E R x (0. co) (Indiana) Solution.
z . T and E I u(zo. t . z ) are continuous functions of their arguments and assume that C ~ i j ( z . vu)uZixj  C b i ( z . z . Prove that for any T >0 min u(z. t )can not be the minimum point of w in R x (O. where X is a positive function.TI where Q = (R x (0)) U (dR x [O. (1) then ( z . t . ~C~b i ( t . wxi = 0 and which contradicts (1).j=l n a i j ( z . p ) t i tLj x ( z . a t the minimum point (z. z .m)) satisfies wt n n C a i j ( z .{ o ) . z . p ) and b i ( z .T]).t o )< min u. First we prove that if w E C2(ax [O. u .j=l i=l at ( q t )E 0 x ( O . ~ u ) w . t ) = min u(z. t . Assume u i j ( z . . (Indiana) Solution.t . t . it must hold that Wt 5 0.In fact. p ) 1 t 1 2> o i.to) R x (0.511 6313 For R c Rnbounded and open. u.t . i. m)).u)uxi o >_ i=l n for z E R.T]. let u E C2(nx [0. t . Now suppose that u takes the minimum a t (%'. t > 0. t ) .j=l n for every t E R~. u ) w Z i > 0. Suppose u satisfies the inequality ut  C i. t ) Q 0x P. T I . Q .
t . Vu) + b l ( z . TI. This is in contradiction with w having minimum point in R x (0. w = u . t )E s2 x [ o . Then takes its minimum at some point in R x (O. for the function w we have E >O is sufficiently small.u.V u ) > 0.T]. t . .&eaxl for ( z .512 Take a > 0 large enough so that a u l l ( z . T ] .u.if However.
^(.513 SECTION 4 HYPERBOLIC EQUATIONS 6401 Assume that g is smooth and in L 1 ( R f l ) .t) = sin(cl<lt). = 9(z).c2AU = 0.O) Show that the LP norm of . %(t.t)at time t for p times t lnlp.O)= 0. Let G ( < . Therefore. > n is bounded by a constant (Indiana) Solution.. t ) be the Fourier transform (with respect to z) of the solution to the ndimensional wave equation: Utt . u(2. it holds with a positive constant M that Set 7 = c<t in the integrand. +. Then we get . = $(tL 0) 1 ?(<I I cK + c21<)22 = 0.n arbitrary. ?([) is bounded in E n . G(<. Transforming the equation and the initial conditions. we have I and Then g E L1(IR") implies that 2tt K O ) = 0.
we get < %(€) qt. The above estimation gives the conclusion of the problem. i.zn). E (a) Use the Fouriertransform t o show that the total energy at time t is constant in time.t>= Go(€) cos(ltlt) + sin( I€lt) It I . 0) G(€.514 p > n ensures the convergence of the integral. = Gl(€). . = Go(€)... q€.e. (a) Taking the Fouriertransforms of the wave equation and of the initial conditions with respect to the variable 2 = (zl.. E ( t ) Eo.t) = 0. (b) Let Eo denote the constant energy in (a).we have . Show that .[n). . 6402 Consider the Cauchy problem for the wave equation Assume U O . U ~ S. 0) where = (€l. utt + l€12G(t. Solving the above initial problem of the ordinary differential equation. (Iowa) Solution.
. + %.+ u.O) z E B3... ...55 1 Integrating the above equation on En and noting that we obtain Then the Plancherel theorem gives us Combining this with (a). z E R3.O)= qqz). u(z. 2 E B3.t> 0. Ut(2. we finish the proof of (b). .. Utt = 0.. = $(z). 6403 Consider the initial problem for the wave equation in three space variables: u.
2 k .22. 11 2 lyzl=lO Similarly. 6404 Let R be the upperhalf space in lRn R = {(21. . (Your answer should be explicit.516 (a) Write down the formula for the solution of the problem. for 1 1 < 1.and $(z) = { 0. no integrals).23) E lR3. it holds that 2 IyI 2 3 for all y E {y : I . when 1 1 5 7. 0.21 = 10). y Therefore J q5(Y)dS = IyzI= 10 J II. find the set of points in lR3 where you are sure that u vanishes when t = 10. find u ( 0 .2 3 > 0). for 15 1x1 5 2. (c) If 4 vanishes everywhere in R3. for 2 < 121.(Y) = 0. we can show that the above equality holds when (c) It is easy to verify that 2 13. (4 (b) By applying the domain of dependence for the solutions t o the wave equation. 2<t. 0. O<t<l. (IndianaPurdue) Solution. where k is a constant. 0. vanish outside a ball of radius 3 centered a t the origin. t ) for all t 2 0. (b) If q5 and II. 2 2 In fact. we can verify that u vanishes when t = 10 for 1 1 5 7 or 1 1 2 13.
Y2. respectively. .t ) of the solution at all t > 0. E R.= 0. 2 2 .O) = d(X). @(z)= x3). (a) Write down a formula for the value u ( x . 0) = +(x). z2. < 0. = ~ t ~ .) is given by t If both 4 and II.( y+ l( Y l . J y . The value u(x. t > O . t ) of the solution at ( x . 4(x1. ax.23)7 23 23 23 23 23 in E3 2 0. ~Y2. ‘$(xl. (b) Find a formula for the value u ( z . (2. at2 2 U(X. t ) = 0. ) 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. when t < 23..t ) at (2. ax: ax. t ) is in R. < 0. it is easy t o see that u(x. 2 3 ) . t ) . the domain of dependence for the value u ( x . x E R. ) given by the above formula satisfies the wave equation and the t initial conditions.. satisfy certain compatibility conditions.~Y 3 W x ~ / +(Yl.~ y l . Ys=~/ta(Ylzl)a(Yzza)a Y3)dS = 0. W ( X . ) . 2 0.x * ~ ~ . (a) When x E R and 0 < t < 2 3 . x t E R for (IndianaPurdue) Solution. Then the value u(x.517 Consider the initialboundary value problem azu azu azu a z U + + . 231. X E R . +(% 2 2 . E aa.m. When 2 3 = 0.t 2 0.
23 Therefore. (Here T = 1.518 In a similar way. on the set T < t .t) be a solution of class C2 to the Cauchy problem with compact support. (iv) Let 7~ = 2.) (Indiana) Solution.4k. have we ERT+t(u(t)) = (u. evaluate lim E ~ ( u ( t ) ) . # 0 and suppose that $ ( ~ 1 ~ 2 2 Show that u satisfies ) . Define the local energy by (i) Use the energy indentity to show that Conclude that the total energy E W ( u ( t ) is conserved. when 23 = 0. t+O" (iii) Do (ii) above when n = 1.T]. 0 for 11 2 2 k. ) (ii) When n = 3. 6405 Let u(2." IVuj2)dSdp. where 2: = for some constant c. we have J Jyxl=t @(y)ds = 0.t) satisfies the boundary condition on = 0. u(z. (i) For t E [O. + .
519 By the above expression.(u(t)) equals to E.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. we can show that for t E [O. . we can calculate that Hence ERT(U(0)) 5 ERT+t(U(t)) 5 ER(u(t))Similarly.(u(O)).
is contained. we obtain t . we have IY  4 I 1 + 2k.4k.520 where a is the radius of a ball in which supp4 U supp.Izl2k  . 4 1 and hence 1 J W J(t + IzI + 2 k ) ( t .ll. (iv) When n = 2.2 k ) ' I It is easy to verify that . the solution is given by If 1x1 < t .IzI .J Therefore.1x1 + 2 k < d3 V t . Hence (iii) By the formula it is easy to evalute too lim ER(u(t))= 0.4k and IyI I k . t .1x1 > 4k. it is easy to see that Hence the formula of the solution can be rewritten as When 1x1 < t .
oo)).521 6406 domain and assume u ( z . (Iowa) Solution. Multiplying the equation by ut and integrating the resulting equation with respect to x in 0. Then E ( t ) I exp(Ct)E(O) is an immediate consequence of the above differential inequality.which solves the equation Let R c B" be a bounded with the boundary condition = 0. t ) is a function with u E C2(ax [0. have we Here we have applied the Green's formula: By the Poincar6 inequality. we can conclude that Thus we obtain a differential inequality for E ( t ) d E(t) dt I CE(t). With prove that there is a C < 00 such that for all t 2 0. .
(z. z E r B3. Then w solves the following problem: aZu 2au r ar { w.= U t t . t = 0. vt > 0. In the symmetric case. 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.t > 0. Solving this problem.O) = u. where T’ = z:+zg+zi.O) = 2 E B3. (b) What is its energy at time t = 10. r 2 E B3. 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. m. Find the spherically symmetric solution to the problem. + (ut” IVul2)dz. T > 0. ar2 Set w = T U . w = 0. (IndianaPurdue) Solution. the equation can be written as + . r = 0. = wtt. wt = f’(r). . u(z. w = f ( r ) .522 6407 (a) Find the explicit form of the solution of the initial value problem v2u= U t t .. E ( t ) = E(O).t > 0. fo.
t > 0.X I = t } n BR Iy 5 The area of aBR. . From the Poisson’s formula.523 And therefore = 2F 7 6408 Consider the wave equation in R3: utt . u(z. Ut(2.O)= 0 .O) = g(z). we have It is easy to verify that The area of the intersection {y E R3. . v x E R3.Au = 0 for x E B 3 . Let R and M be positive constants such that and Ig(z)l 5 M . where g E Cr(R3). Prove that there exists a constant the given data such that C depending only on (Indiana) Solution.
x (Indiana) Solution. Suppose t) also that u = ut = 0 on the ball B = {(z.T ) ~ z . we have .t } .Rn+l(x.)be a solution of the equation utt Au = 0.t .t)l 5 .O) : 1 1 5 t o } in the plane t = 0.5 t o . Multiplying the wave equation by ut. we obtain 1 lu(z. t = T } 1 2 1 for 0 5 r 5 to. M .524 Therefore.t) : 0 5 t 5 t o and 11 5 t o .”+ IVuI2)(x. and integrating the resulting equation with respect to x in R. It is not difficult to verify that Therefore.t > 0. Let = {(Z.. we can obtain Set 1 E(7)= 5 lr(u. 4 r R 2 = . MR2 47rt t ’ 6409 Let u E C2(.t) E En+1.. x Prove that u vanishes in the conical region D = {(x.
(Indiana) Solution. we have uttdx = J. gdx = 0. in D. Audx = 0. Assume that f . Let R > 0 such that It is not difficult to verify that for any t > R. This implies that vo 5 7.5 to.O) u(z. . By using Green's formula. Prove that if = dx). Hence E ( 7 ) = 0. t ) be a solution of the Cauchy problem utt = AU (Z E B3. u=O 6410 Let U(Z. t Ut(Z. 0) = f(z). /ltnL8u 2 dx = 0. then I..525 It is clear that E ( 0 ) = 0. we get the following equality from the equation J. > 0 ) . Therefore... g E Cr(nZ3).
JlzIRs X b) Show that there is a positive constant C such that 1 t . In the other side. by Cauchy inequality it follows that 21ax J R S 5 (J. we can get . . . 6411 Consider the solution of the wave equation Utt = nu (x E R3.t > 0) with data u(x. ut(x.gdx # 0. . 0) 0) a) Use Fourier transform to show that there is a positive constant c such that u2 dx = c.. (Indiana) Solution. This contradicts (1) provided that J.526 and by the initial conditions. a) Let 6 denote the Fourier transform of u with respect to the variable x = (XI. = 0..m a x l u l ? CI for all sufficiently large t. zn).. = +(x) E Cp.By taking the Fourier transforms of the wave equation and the initial conditions.iax)t (/ t R< ( C<t+R 3l ax) l I 2 for any t > R.
t . Then we have This contradicts (a).z3) for t > 0. 400. )< la(t. .m a x ~ u ( z .>l+. t ) with respect to the variable 2 = (21. b) Let Q ( t )be the support of u ( z .)l+t. a s n . then there exists a sequence { i n }such that t. ~ 2 This gives Iu(2. .’.tn)l I IQ(t. The proof of (b) is completed.21.527 Therefore we have Then we get where if $0. If the conclusion is false.+ c c t+b and t.
t )=  .R. E C F ( R 2 ) . we conclude that if t > R and 11 2 < t . 2 2 .t ) d z = 0. which implies that t ) = 0. c ( 2 f IR3.t)dz. then u ( 2 . l4<$t 4 3 0. we have 9 sup Iu(z. when t > A. From the Huygens's principle for the %dimensional wave eauation.528 6412 Consider the Cauchy problem for the wave equation a) Let n = 3.11. Let 8 behavior of the integral < 1. Therefore. (21< R}. a) Assume that R > 0 is so large that suppq5 u supp11. q5. lxl<@t when t > __ 18' R b) From the Poisson's formula. (Indiana) Solution. J u4(2. vlzl< et. t ) = 0. the solution to the 2dimensional problem is given by u(21. Determine the large time J b) Let n = 2.t)l = O(t') Ixl9t as t + o) c.Show that for any 1 < 1. it holds that +. E CF(R3). 11. u4(2.
V x f = 0.X I I et +R and d2 . i..( B t + R)2. if f > 0 for all z.zI2 2 t Therefore. Suppose that 0 5f 0 (z. 0 j given. Show that supp = o(t3) as t X (Indiana) . let f ( c .. U ) =j v). then the same is true for f at all later times . f(Z.t) = J. b) Define the local density p by p(. j~~ 1 . Consider the initialvalue problem for this. we have dt2.'u. (c.v) 5 h(z). t (z E B3. where C is a constant. The proof is completed.where h E L1(B3). 6413 The linear transport equation is following PDE for a scalar function f of seven variables z.): v ft + v . v. clear that if it is IzJ5 B t and Iyl 5 R then JY .t)dv+ oo.e.529 Let R > 0 be so large that suppllr C { z E IR2. 1 1 < R } . v. v.. a) Show that. t E R.(y. 2 When t > 5. E IR3.
t) It is clear that f 2 0 i f f ? 0. + tv.Solution. b) From (3). we have 0 = JR3 f ( x .= 0. a) The characteristic curves of the equation are defined by dx = v dt and along which it holds that dt Therefore. the forward characteristic curve departing from t = 0. we get the solution to the initialvalue problem 0 f (2. (2) From (1) and (2). + yuu. v.tv. 6414 Solve the Cauchy problem xu. The system of the characteristic equations for the problem is given by .v). = xy ( x > 0) u=5 o n x y = 1. (h diana) Solution. x = a is x = (Y df . if a solution exists. =f ( x .tv.v)dv O This completes the proof.
~ e 'ds = _ ds (ae'y ds 1 u2 2 + u). F. we get vt + 22121. + 6415 Compute an explicit solution u(x. Set v = u . dt . W(2. The characteristic curve of the quasilinear equation (1) is given by dx . where C is a constant. Then we obtain imme This implies that + u + x y = c. (Y Fkom the above Cauchy problem. y=. 1 u=5.= 2v.. (Indiana) Solution. Differentiating the equation and the initial condition with .= xy ds du s=O:x=a. with the initial conditions _dx dY z= y u . = 0.531 ds . t ) to the initialvalue problem Ut + = 0.O) = 22. U(X.O) = z2. . respect to 2. we have immediately and 21 du du d = . The initial data gives C = diately u = 1 43cxy.
__ {:: 2 E IR. we can obtain immediately v=Hence u= 4t 1' 22 22 4t 1' + + 6416 Solve the following initial value problem. p > 1. 05z< 1 .e. The characteristic curve departing from t = 0. we have 2 = uO(P)t p. du _ .0 . u t + uus = 0.532 along which it holds that dv . 0. giving explicitly the discontinuity curve and the RankineHugoniot jump condition along it. 1 . ( IndianaPurdue) Solution. 12. From (1) and (2). 1p. z = P is given by dx =u. and find the discontinuous entropy solution. i.t 2 > 0. and P < 0. dt along which u is a constant.O)= Show that the continuous solution exists only for a finite time. dt where u o ( z )= u(x.O). t = 0 : u = u"(p). u(z. o<p<1. t = o : x = p . P < 0. < 0. 21 = { P> 1. . dt From the above characteristic equations. 2>1.= 0. + u = uO(P)..
 and hence x < t.e.. the continuous solution to the problem blows up. 0 5 5 1. 2 > 1.For 0 < t < 1.2 When t = 1. 11 for 0 5 t < 1 (see Fig. we have p=. dx 1 .6. we have + u). the RankineHugoniot condition along the discontinuous curve is given by dt i. ' t<X<l. Fig. For t 2 1. 2 t l l .2).= (u+ dt 2 Noting u+ = 0.2 .2' 1 x=(t+1). u.= 1. 1  dx _Therefore.1x 1t 1t' Therefore we get the continuous solution to the problem u = 1.6. 0.t 1t 2t . 1.. the discontinuous curve is dt .
t ) 5 pebt for all x and all t E (0. where b is a constant and a is smooth. then 00 00 (Indiana) Solution. U(X. t ) = U o ( Y . if z ( t . Show that if a then 5 u ( z . ii) Show that. T ) . Then v satisfies the following equation dV at + U(X.u)bu=O dX at at4 aU forO<t<T. . 6417 a) Let u ( z .e bt v) dX = 0. t ) be a smooth solution of +a(z. 3 > z ( t + 1). aebt 5 u ( z . y ) is the characteristic through (Y.t)be smooth solution of dU at 8 + a(u)U = 0 dX = UO(X). a) Let u = ebtv. then u z ( 4 t . for 0 < t < T. if 0 5 t l < t2 < T . 2 < ' ( t + l).0) 5 p for all X.O). b) Let u(X.Y). Show that.534 and the entropy solution to the problem is given by 2 0.O) where uo and a are smooth. 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.
{ 4 0 . we have The characteristic x(t. i) Let m = inf a‘(uo(y))ub(y). a Then it follows from above problem that Therefore. it holds that infw(x. 0) is given by the following Cauchy problem: &. we get g (g+. Y) = a(uo(y)). X 2 dx dt This gives the estimation in the problem immediately. we can get the Y following conclusions easily. t ) is a constant along the characteristic z = x(t.O).535 The solution v maintains a constant along a characteristic of the above equation. ~ ) . y). a Differentiating this problem with respect to y.O) 5 v(z. From the above expression.Y)) { = a’(uo(y))ub(y).t)5 supv(z.= a(x. I?”). b) u(x. If m >_ 0. we have This is just that we want to show. . Hence Differertiating above equality with respect t o y. defined by the equation . Y ) = Y. y) through (y.Y) = 1. Therefore. the Cauchy problem admits a global smooth solution in ( 0 . FZ(0.
. 0418 (Uniqueness of weak solutions) Consider the Cauchy problem 'Zlt + 1.y) in the integral. If uo(z) E 0.z). u(z. the largest T for which the problem admits a smooth solution in (0.. on 4(. on x [0. prove that the only LP solution of (2') is u G 0. t ) and &z(t. (Indiana) Solution. IR x (0).7']. we have 00 Taking the expressions of u. on on IR x [O.  44YI s))dyds (2') + J_. we obtain 00 00 This completes the proof of the problem. Consider the following Cauchy problem: 4t + 4. y). T) is given by mT+ 1 = 0. t )E P ( I R x [O. A function u E P ( I R x [O.T]) defined to be a weak solution of (2) iff is lT : / 00 4 Y . y) into account. IR x {T}. ii) For any fixed t E (0.T) = 0. = f ( t .(z(t.)(4t(Y.536 If co < m < 0..uO(Y)4(Y? O P Y = 0 for all test functions 4 ( z .O) = uo(z). T).T]) which are compactly supported and vanish on {t = 7) '. by the change of variables z = z(t.T]. '1 = 0. .
* .a * . ) )5 TV(uo(. ) Then we have xi. O ) (i = n... dt along which the solutions to the equation satisfy du . provided uo(x) 0. t ) . . Suppose that the backward characteristic curve through (xi. l .z .is dense in L * ( B x [O. ..TI). where x E t B and t > 0.T]. As the space Cr(B2 x [0.T]) with a compact support. then TV(u(. Therefore.= u(2.= b(x.  . It is not difficult to show that the above Cauchy problem admits a solution ~5 E Cm(Ex [O.TI). t for a) Show that if b 0..^) = uo(ti). Then u is a constant along a characteristic curve. * . . l . dt a) Suppose that b 0. 6419 Consider the Cauchy problem for a function u = u(x.. consisting of all Cmfunctions with compact support in E x [O. where q = p / ( p . ..5)dYd5 =0 (*I for all function f E C m ( Bx [O.T]) with a compact support. + a(z. ) . 21. .x .T]) which is compactly supported. we have 6’J_+_m U(Y. . + J.. n ..t)...T) . ut u ( z . one still has the bound T V ( u ( * . The characterstic curves of the equation are given by the equation dx . the equality (*) is valid for all functions f E L q ( B x [0. Hence we have u 0. .T])..)f(Y.. b) Show that when b $ 0. 0) = uo(x) E C(B2). meets the line t = 0 at ( < i . .t)u. zl. zo.537 where f E C m ( Bx [0.l). (Indiana) Solution.~ .. = b(x. We divide the line t = T into subintervals by the points * * .’ TV(b(*. O .)) each t > 0 (where T V = “total variation”)..t). 5 TV(u0) t)) 5))ds.
T)) 5 TV(UO(')).538 Therefore. When b $ 0. VT > 0. we have u ( z i . the above inequalities. s)ds. we get TV(U(. ~ ) From . we get . T ) = uo(ti) + Ir 0 b(zi(s).. b) Let { z i } and {ti} be given as in a). where by z = Z i ( t ) we denote the characteristic curve through ( z i .
Irvine . University of Cincinnati Columbia University Courant Institute of Mathematical Sciences.539 Abbreviations of Universities in This Book Cincinnati Columbia Courant Inst. New York University Harvard Illinois Indiana IndianaPurdue Iowa Minnesota Rutgers Stanford SUNY. 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.
This action might not be possible to undo. Are you sure you want to continue?
Use one of your book credits to continue reading from where you left off, or restart the preview.