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
.
W # V . SW C W ) . there is a nonzero subspace W of V . W and V/W respectively. W # V . .I. ml(t). Then V E E [ X ] / Ifor some maximal ideal of B[X]. Let m(t). such that T ( W ) W . Let W be an invariant subspace of V (that is. Make V into an B[X]module by defining for i=O w = T ( v ) for all w E V .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). that is.D.and rnz(t) be the minimal polynomial of S as linear transformation of V. (Indian u) Solution. r n 2 ( t ) . Since R [ X ] is a P.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). Thus a. This is a contradiction. Hence V is not a simple B(X)module. Prove that there is a nonzero subspace W of V .. there exists an irreducible polynomial f(X) of B[X]such that I = (f(X)). (a) Prove that m(t)divides ml(t). Let S be a linear transformation of V into itself. Now suppose V is a simple IR[X]module. This implies that we have an irreducible polynomial f ( A ) with degree 2 3 in B[X].such that T ( W )C W . 1102 Let V be a finite dimensional vector space over a field K .
since {Taw I i 2 0) spans V by (b). Similarly. z = f ( T ) (x). T/ can be viewed as a K[t]module via the linear transformation S . Show that V doesn’t have proper Tinvariant subspace.4 (b) Prove that if ml(t)and mZ(t)are relatively prime. mz(t). So m(t)# ml(t)* m2(t) in this example. Let m(X) be the minimal 3 . then m(t)= m1(t)* m2(t). (c) Give an example of a case in which m(t)# ml(t). Then we have V = lR[X] . As usual.m2(t). (In i m a) d Solution. v c m1(t). ~ ( t m ( t ) E A n n ~ p l V (m(t)). W = 0. m2(t)divides m(t).for any f(A) E lR[X] and z E V. Then it is clear that ( m ( t ) = A n n ~ p l V ’(. (a) Since m1(t). V can be viewed as a module over the polynomial ring l R [ X ] via f ( X ) . ) = Hence m(t)divides m l ( t ) m2(t).m2(t) divides m(t). m2(t). + + 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.Then we have m(t)= ml(t). a cyclic module. ml(t).m l ( t ) ) A n n ~ p l W ) = and (m2(t)) = A%[t] V l W . Since ml(t)and mz(t)are relatively prime. v . Let V = W @ W and S : V t V be the natural extenaion of S to V. (b) Since m(t)E AnnK[t]V G AnnK[t]W = (m1(t>). W is a K[t]submodule of V. rnl(t) divides m(t). Since W is an Sinvariant subspace of V. Then it is clear that m(t)= ml(t) = m2(t)= t 2 1.ml(t)and mz(t)are all monic polynomials. (Indiana) Solution. (c) Let W be a 2dimensional vector space over the field Q of rational numbers and S : W + W be a linear transformation with minimal polynomial t2 1. since m(t).
.. Without loss of generality. Hence we have P'BP = 0. Let B' = diag(B1. . with 0 on the diagonal.. =v R[X]/(m(A)) R[A]* E 21 (we may assume that V # 0). . Then there exists an invertible n x n matrix P such that PAP' = diag{ A 1 .. . A in C. Jt). . Then m(A) E Annwfxl(v). . . PAP' = diag{X1. We have to show that A commutes with some nonzero nilpotent matrix. Then B' . Let diag(J1. is a Jordan block of order ei.. Jz . Sufficiency. 1104 Let A be an n x n matrix with entries in C. ..l B P is a diagonal matrix. . . B'. and A B = B A . . . PAP' = P A P . Show that A has n distinct eigenvalues in C if and only if A commutes with no nonzero nilpotent matrix.l B ' P . B . So V is an irreducible R[A]module. . Since All Az.. we have + V .J t ) be the Jordan canonical form of A and PAP' = diag(J1. But the nilpontency of B implies that P ' B P is nilpotent. Xa. .. Suppose that the characteristic polynomial of A has multiple roots. Suppose that A has n distinct eigenvalues A'. . Jz. we may assume that el > 1 (If all the ei = 1. ) where Bi(i 2 2) = 0 E Me. which is nilpotent. Bz.l . An}. then it is easy to see that A commutes with some nonzero nilpotent matrix). Let B1 be the Jordan block of order e l . where P is an n x n invertible matrix and J. Taking B = P ....5 polynomial of the linear transformation T : V Since m(A) is irreducible. we have t o show B = 0.. Then JIB' = BIJl and Bl(#0 ) is nilpotent. .Xn) commutes with P'BP. . If A commutes with some nilpotent matrix B .(C). V does not have proper Tinvariant subspace. Necessity. So B = 0. . Thus. (Inddana) Solution. A are distinct and . we have B # 0. P .
there exist z. Obviously. f ( A ) 6 F[X]. Show that A is diagonalizable... Here.I>~ o of V over F[X]. . which is the square of an irreducible polynomial in K [ T ] .>o.Let { U ~ . .zl has exactly two nonzero submodules.I.vn}. according to the assumptions. A on V . V = F[X]. T E T }and { T * 8 } . V can be viewed as a module over the polynomial ring F[X] simply by f(X). T : V + V a linear map such that the minimal polynomial of T coincides with the characteristic polynomial.) = where (d*(X)).(X) = det(X1. A = ( u ~ ~ ) the ~ ~ of T relative to the base.A in M.d. so m(X)= d. . . z .. 1106 Let V be a finite dimensional vector space over with basis {wl.} of V over F .dl(X). In general. ~ F [ X ] .A ) .. By the structure theory of finite generated modules over P. be ..> coincide. . . . . l. .D. > are nonzero submodules { . 2 . . Ann(z. So at least two of them coincide. (HU7Wrd) . the three subspaces generated by the sets {T'Z'}. .(X) if i 5 j. U be a base for . . . . U ~ ... the minimal polynomial m(X) of T is det(X1 . Let CT be a permutation on {q. v. z z ~ .{Ta?}.20 and { T Z 3 } .A ) .(X)(d.(z) any z E V .(X)} where the d. Since m(X) is the square of some irreducible polynomial. then at least two of the three subspaces spaned by the sets (T*?it}z>~. . . . + Hence s = 1 and v = F[X] z1 E?! f F[X]/(m(X)) is cyclic. .(F[X]) has the form diag{l.6 1105 Suppose V is a finite dimensional vector space over a field K .T and 3 are any three nonzero vectors in V . z l $ F [ X ] . matrix a normal form for X I .Show that if %'. s ) E V such that V = F [ X ] .} and thus induce a linear transformation . .z = f(T).(X) are monic of positive degree and d. r 0 (Stanford) Solution.(z = 1 .
then all such T's are similar...wn). . X3 + X2 ..1 is the minimal polynomial m(X) of T . Thus X3 X2 . . Then M = i } diag{M1. 0 1 0 0 1 0 0 1 1 0 0 *.1 is irreducible over the field Q of rational numbers.. } ... . 1107 Let V be a finite dimensional vector space over the field of rational numbers. . .... Since T' = T2 T.. . . Prove that 3 divides the dimension of V. ) is . V ~ .. ' we assume that when u is expressed as the product of disjoint cycles (This decomposition may have 1cycles). 1.. V ~ . . V .d. . A be the matrix of T relative t o + + + ns some base of V ... The matrix of A relative to the base { V ~ ... V . ( H U 7 V d ) Solution. Now let n be the dimension of V over Q. Mj be the matrix of Alwj : Let Wi + Wj relative to the base { ~ i ~ . .. . . we may assume that CJ is the ncycle (v1. . Hence.. . . l.. . Then the Wj are invariant subspaces of A and V = W1 @ W2 @ . . . vof~Wj overt.. By reordering the elements 211.1. . . .~ + 1 . Suppose T is a nonsingular linear transformation of V such that T' = T2+T. . . . vij} + for j = 1. This completes the proof that A is diagonalizable. . . . Let Wj be the subspace of V generated by {Kjl 1. and prove that if dimV = 3..s 1 (io = 0. and thus M is diagonaliz able. .. ~ ~ ~ . is+l= n). @ Ws+l.. .* .. ~ ~ ..u. . . diag{l. .+ M = .w2. .7 Solution..Ms+l} is the matrix of A relative to the base ( V ~ . without loss of generality. .dl(X). 'u2.2. 0 0 It is easy to see that the minimal polynomial of M is A" . . . Obviously. So it suffices to prove that every M i is diagonalizable.(X)} be the normal form for X I .A A . ... . T is annihilated by the polynomial X3 X2 . .1. .
. 11: .1 has no multiple root.. . Then m(X) = 1 A" . Thus we have proved that 3 divides the dimension of V .. = &(A) = X3 + X 2 . + X2  1).1.1) and the minimal polyE nomial m(X)of M is A" . If dimV = 3. then X I . .. (Here is a misprint. II(u).A is equivalent to diag(1. . 1 0 a .1. Let 1 : Fq + Fq be the Frobenius automorphism II(z) = z p .l". By the irreduciblity of &(A) = m(X)= X3 + X2 . .11.A" .:. there exists a u E Fpsuch that {u. On the other .M is diag{ 1 . 1..1 are in Fp. M= I. ( X ) . 3 .1. Suppose that 1 is diagonalizable as a linear map over Fp.. . Prove that II 1 considered as a linear map over Fp is diagonalizable if and only if n divides p" . . 1 II(u)..I12(u). ...112. By the Normal Base Theorem. When II is considered as a linear map of Fq over Fp.8 where the &(A) are monk of positive degree and d i ( X ) l d j ( X ) if i 5 j.II"l(u)} is a base for Fq over F p . . and all the roots of A" . It is wellknown that Fp C Fq is a Galois extension with Gal(Fq/Fp) = {1. It follows that all the 1108 Let Fq be a finite field with q = p" elements.. Since det(X1.A ) ) = n. .1 = det(XE .I..M ) .A ) = d l ( X ) .). X3 rational canonical form for A (or 7') is T ' s are similar when dimV = 3.. s = deg(det(XI . II"'(u)} * is 0 1 0 . . It should be "n divides p . .. 1.. The normal form for X .) (Hurvard) Solution. The . d . the matrix of 1 relative to the base {u.IIn1}.. I12(u). where p is a prime. we have &(A) = &(A) = .
So IT is diagonalizable as a linear map in over Fp.ad)(. Show that the derivative of the determinant det A satisfies Then and .a2d).. . .1) .\ .' ) ~ } Mn(Fp). A" . .a("l)d) where d = and a is the generator of the group F. (A . if n divides p . u ( ~ . Let A'(t) denote the matrix formed by the derivatives of the elements. .. Thus n divides Conversely. all the root of An p .Lli)seelements are differentiable functions of real variable t . 1109 Let A(t) be a nonsingular matrix r. Hence M is similar to diag{ 1.a d .1.1.  1 forms a subgroup of FJ = Fp\{O}. .(A .9 hand.1 has no multiple root and A"  1 = (A .
+ det A.) .10 = det A1 + det A2 + . (b) Use the answer to (a) to find the second degree polynomial that solves the minimization problem (Courant Inst. . and 2. 41(z). respectively. Hence we have proved that d (det A) = det A . Define an inner product on V by + bz + cz2 with real (a) Find a n orthonormal basis for V consisting of polynomials 4o(z). and c. and q52(z). trace(A’ . . having degree 0.. b.1. dt 1110 Let V be the vector space of polynomials p ( z ) = a coefficients a. A’).
x. 2 ) .$a(z:)) = ( p o ( z ) . c1 = and no(.))= C l (i = 0 . 9 and 175 cg = 0. we can orthonormalization 1.4i(. 8 3 .x 2 to get an orthonormal basis of V with degree 0.5 z ) 2 d x = . Then for any p(x) E V . By an easy calculation. (a) Since 1.11 Solution.)) =0 (i = 0. Hence P O ( % ) = $ z [:ix3 . Obviously. 3 .x. rl ( . 1 and 2 respectively as usual.2).Zz. we get co = 0. (b) Let and QO(Z) = z3 Po(.) = z3 . 1 .) such that ~ o ( z is orthogonal to V .x 2 is a base of V .1. that is ) ( x 3 .$i(.p o ( z ) .
.**... (b) Show that if det(A) # 0 then a = det(A) and b = det(A) y*Alz. . . ( c ) Is it true that a = 0 if det(A) = O? (Courant Inst. Xn Y = ( . * * aln ann + .. ann * ZlYl .... (a) Show that there exist numbers a and b so that det(A + m y * ) = u + bs. [[ Un1 . 3 det(A + szy*) a11 an1 = det + .).) Solution.. Yn y*=(y1... is minimal..12 Thus.... + Sz1Yn SznY1 SXnYn .. when p ( z ) = gz. and 1111 Suppose that A is a n n x n matrix and that 2 = ( :).. . ”’ . Suppose that all the entries of A. SXIYI + . and y are real. z. (a) Directly. a nn a2n +det anl . . * . X1Yn det a21 . = det +s ... .. yn).
. a = det(A) = 0. ... * . ... ann + .)..... (c) If det(A) = 0.. Aln ... Ann = y*(det(A) ... y*Alx. . ...13 Hence det(A + sxy*) = a bs for some a and b. ... . Ann All ) .... ( If det A # 0.. . .. ... . is the algebraic cofactor of Al Thus 1 det(A) ___ aij. anl *.. a n n .. . A')Z = det(A) ..... for any s.. we have a = det(A) and + + b = det .....In ..+ det n an11 a n ... + det . (b) Since for any s . n X n Yn n where A. .. * * An1 Aln . ... det(A + sxy*) = a bs as in (a).. An1 . . All .. . .
l d i ag ( l ... Define 1 sin(A) = A . and For any v E a".... (el. 5 j 5 n.O) = P'P'l diag(l. v n ).. 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. . v..A3 3! 1 + A5 5! * . (Courant Inst. XI. v 1200 P .Xn aredistinct..14 1112 Let A be an n x n real matrix with distinct (possibly complex) eigenvalues. n+w ncc orthonormal basis of (En. w (Since lim (X. {f2..) = lirn X . .. f z . . For . Xz... . Let P = ( v l .) is a basis o f a n and diag(l. ncc lim A'lw = = lim P d i a g { X . .A2. . f z .. dim(ker(T)) = n . Hence {fl} is a basis of Im(T).O) . V Z . .. diag{l. .Az.. v z . P isaninvertible matrix in illn(@) P . .Y.. . P' . . . Since A1.. Define lirn T : C" + an by T ( v ) = lirn Anv. .... A. .I A P = diag{A1. e2. . and dim(Im(T)) = 1.a. f.O .. P ' . Assume that XI = .l . Then the matrix of T with respect to this basis is P ' ...O. .O)P'P'l is the matrix of T with respect t o the basis ( f l ...1. .Xn} .O.. f n ) . . . 1113 Let A be a matrix. Prove that noo Anv exists. .. e n ) be the standard Let ncr. X ~ ....fn} is a basis of ker(T). . X ~ } . and corresponding eigenvector V ~ . P .O.. Let fn en Then ( f l ..) Solution. 0 } . lim Yn). . . .
15 express sin(A) in closed form. 103 O ( 1 1 ).[. Denote B = ( 3 Then B is similar to ") .) Solution. 1' )'=( 1 ). Hence sin(A)  ( .(4 0 10 ) . 7 and 1 1 0 1 0 1 1 0bviously.. ( 4 3o 0 ( 4 1 1 ) where .L 3!T ) 3 . (Courant Inst.
. all the eigenvalues of A are negative. ..(B) > 1. Let A.1 . a .B)Y >_ 0. Then for any Y E Elo.YT(X. *... Then for any Y E V1. (c) Obviously.I . y2 E R I O such ..I . .) Solution. Taking Y T = ( 1 . (a) Show that (b) Use part (a) to show that A has all negative eigenvalues. z 1 .2. > 1. 0 . this is.(B) be the largest eigenvalue of B . 2 1 1 2 All entries not shown are zero. 1 A= **. . Thus A . which agrees with A except in the first row and column: 1 3 3 2 B= 1 1 2 1 1 *. (a) A direct verification shows that (a) is true. If B has more than two nonnegative eigenvalues.. 0 . Show that A. the symmetric matrix X... (d) Let Vl = { ( O . Y T B Y 5 0 and Y T B Y = 0 if and only if Y = 0. . . (b) Since z T A z 5 0 for all z E lR9 and z T A z = 0 if and only if z = 0.16 1114 Let A be the 9 x 9 tridiagonal matrix '2 1 2 1 1 ..YTY. Y T B Y 5 X. we have y1.B is semidefinite positive. . (Courant Inst. we have Y T B Y = 5 5 X. 0 ) ... (c) Let B be the following 10 x 10 tridiagonal matrix. ~ 9 ) I zi E IR} 5 Elo. Let A1 and A 2 be two of them. (d) Show that B has 9 negative eigenvalues.. ..
2 ) . there exists an n x n orthogonal matrix P such that P'TP = diag(A1. By assumption.y2).. > A. You may wish to express T in a basis of eigenvectors.) Solution. en} be the standard orthonormal basis of Rn.) 1x1 Idiag(Al..a contradiction. Thus B has 9 negative eigenvalues. Show = by showing 5 and 2 separately..) 1x1 . Since dim V1+ dim V = 9 2 = 11 > 10.. .. yTBy = a?A1 u.17 T that y1ly2. L ) z l 1x1 . Hint..A. e 2 .).eZ).. we have min 2EV10) lTxl   min Z€VIO) IP'ldiag(Xl. Then min x €V . A.l V is also twodimensional. PlxI IP1x1 .x1 when V ranges over all two dimensional subspaces of lR". n x n matrix with distinct eigenvalues A 1 > A 2 > .to1 Idiag(Al. . z + + + 1 there exists 0 # y E V n VZ. PlxI 1x1  2EVIO) min If V is a 2dimensional subspace. positivedefinite. .A. Az. .. y E V1) and yTBy 2 0 (y E VZ).. ... Hence max v min Z€V{O} w = m a x 1x1 v min Z€V{O) Idiag(A1. Then for any y = ulyl uzy2 E V2. Then we have yTBy < 0 (y # 0. . (Hyl = Iy1. . V = (e1. (Courant Inst.Xn). 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 .A2. . Since for any orthogonal matrix H and any y E B". 1115 Let T be a real symmetric. and Now let {el. z the subspace spaned by y1 and y2.. P . Let V = (yl. * A 2 2 0. y1 y1 = yTy2 = 1 and Byi = Aiyi (i = 1 .. .. . .
fn} of IR" such that V = (fl.. .fi. ez.. e. . = alf1 Then Q = ( q i j ) . . we can find an orthogonal basis {fl. > A 2 > 0). It follows that I v 1x1 On the other hand.. x n is an orthogonal matrix.fz). . Let (fl. . .. . fz . lT. for any two dimensional subspace V of 1R". and Thus when V ranges over all two dimensional subspace of B". . 1x1 #x + azfz E V .2 Az.18 = A2 (A. max min Z€V{O) ITX . . and if 0 Idiag(A1. A ). .XI so. fn IT = Q(e1. ...
. . + P" Since C k=O 00 IJPk(l co and the norms over vector space are all equivalent. QlpQ= be the Jordan form of P. be the n eigenvalues of P and ' . r). Prove that ll(I . we claim that the norms of all eigenvalues of P are < 1.) Solution. consider the sum R(P)= M c 00 Pk exist and R(P) ( I . Denote ( A1 e f .19 1116 For any n x n matrix P . A. ~ are X . (c) Use part (a) to compute the inverse of 1 A = [ ! i . ) n eigenvalues of m ( the sm= E + P + P 2 + . + z"..P)'ll 5 *. * Then ~ m ( A l ) . then (IP)' < (b) Assume that IlPll < 1 in a matrix norm induced by a vector norm. * ) rjm(z)= 1 + z + * . (a) First. 1 0 0 (Courant Inst. . Let A Az. r j m ( X ~ ) . = k=O (a) Prove that if k=O JIPklJJ00. < 00 k=O . ~ ~ ~ . .P ) .' .
I J . and (I. Q = Q . . Then llull = IlIuII = IIPuII. = We must have u = 0.P)'. m+m lim Sm = lim ( I . 2 . Since llPll < 1. n ) . llull < llull for the corresponding vector norm. Thus all the eigenvalues of I . Since mw lim Q .P)' = (IP)'.u 0 for some vector u.P). .20 is convergent. The matrix I . C m=O W is convergent (i = 1 .~ s .P are nonzero. Hence pal < l ( 1 5 i 5 n).llPull 5 llPll .O)(I. Then we have .P is therefore invertible. mw 1 Q. Now suppose ( I . that is. (b) By definition. and we can write ( I . Thus 00 ( l( I ) PI' R ( P )= k=O Pk = ( I .(IP) rn00 = IP"+'. for any nonzero vector u. l l ~ l l =sup { #lv nonzero vector I ..) exists. . moo lim dm(X.lim ~ + = nrn = (I .P"+')(I  PI' I .P)' = I + P ( I . Since S.1 .~ ( lim S.P ) is invertible.
F p n ) .N .21 Then A = I . since there is a surjective homomorphism G + G' ( 5 EndF(V)).w + 0 * 01g E G}. which are unipotent matrices. It is well known that m j=1 and U= is a Sylow psubgroup of GL(m.. v(a E G) are unipotent matrices. V .Fpn)such that PHP' U . where m = dirnp. . there exists some P E GL.Fan). 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. We fix a base of V over Fan. Using Sylow theory. Then G' is a pgroup. Let G' = {V + V . if we change the base of V via P . ( CoZumbia) Solution. Then G' is isomorphic to a subgroup H of G L ( m . By Sylow's Theorem. Thus. 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.
.. (Stanford) + . S N = C 2s 2 j ( . where r n ( A ) is the minimal polynomial of S . Since the minimal polynomial of S is not equal to its characteristic polynomial. So there is a linear transformation T of V such that T commutes with S but T is not a polynomial in S. . all of whose (complex) eigenvalues are real and positive.j=1 + 1)nj. . Let F [ S ]= {f(S): V t V I f(X) E F [ X ] } . Hint. where S is diagonal and N is strictly triangular.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. N 2 C j=1 8 nj = dimF V.+akSk for some k and ai E F . then by Frobenius theorem. . 1119 Let A be an n x n real matrix.&(A). Show that for any integer rn 2 1. . dimF F [ S ] = degrn(X). dimF F [ S ]< dimF On the other hand. 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+. the dimension of the i vector spcae over F of matrices commutative with the matrix of S is v. ) (Stunfod) Solution. 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). let d l ( X ) . Make use of the Jordan form S N of a conjugate of A . 0bviously. Then as Fvector space.
. . chbml " b I* A = QlB"Q = (QlBQ)" for some invertible n x n real matrix Q. 1. any 1 5 i 5 k. for any integer m 2 1..B2. Hence PlAP = (diag(B1.. Now let A be an n x n real matrix..(Aa). there exists real matrix B. Then there exists an n x n invertible real matrix P such that F I A P= diag(J1.. A and Bm are similar.. '.23 Solution. . J2.lbmn+l cibm1 .. .. C. For any integer m and b 1 0 O b 1 2 1. As proved in the above.* b O 1 b Bm= Obviously. XEA and XEB" have the same invariant divisors (1.Bk))m . such that By = Ji (1 5 i 5 k).... ~..}. all of whose eigenvalues are real and positive.. . . Thus L : b " 0 Chb"l b " Cibm1 .. Suppose first that A= be a Jordan block with a real and positive.. J k ) where Ji ( 1 5 i 5 k) are Jordan blocks with diagonal elements real and positive. .. .2bmn+2 0 . . let b = m & B= 0 0 0 0 It is easy to see that *. .
an are distinct and + By the structure theorem of finitely generated modules over PID. (a) Show that there are 2'invariant subspaces V .~ = ~ i ) ~ ~ j ) . ( Stanford) Solution. for some n 2 1. E K such that V is the direct sum of the V. Viewing V as a module over the polynomial ring K[A] via T. and N ( T ) : V S V is nilpotent. j ) ((A .ail) : V . Denote V. and let T : V 4 V be a linear transformation. we may write where all the w.Then . ( A . (Aa. (A transformation N is nilpotent if N" = 0. 0 where B k 1120 Let V be a finite dimensional vector space over an algebraically closed field K . .a Z ) e z T . Use the Chinese Remainder Theorem.. . V and elements of a.* . . Hint. . .j = K[A]wij. + V. . * * * . is nilpotent. Let (A al)ell. .(Xal)el'l .) (b) show that there are polynomials S ( T ) . ( T ) : V + V is diagonalizable. (Aan)ebe the elementary divisors of VqX]. a2. N ( T )E K [ T ] such that T = S ( T ) + N ( T ) .where a1.24 and B1 B2 0 P'] m = Bm. and (T .. (Xa2)e=.j E V and A n n ( ~ .. .)enl. .
Thus N ( T ) : V + V is nilpotent. ). Thus (b) is proved. by Chinese Remainder Theorem.25 The K j ' s are Tinvariant subspace and ( T . Since ( A . Let N ( X ) = X . . (1 5 i 5 n). Then S ( T ). ( A . T = S ( T ) N ( T ) and X .a I N ( X ) for any i 1 5 i 5 n.a 2 ) e 2 r 2 . waj = Cya . there exists some S(X) E K[X]such that S(X) ai( mod ( A  ai)eir. . Thus are pairwise relatively prime. waj for all wij and it is easy t o see that S ( T ) : V + V is diagonalizable.a i l ) : V. + vj is nilpotent. So the minimal polynomial m(X) I N ( X ) h for some positive integer h. + . Obviously.:j (a) is proved.al)e*'l (A .S(X). .
Let g E G. then N E G and NgN'= 1 0) A l ( . A 1 and A2 be the eigenvalues of g viewed in M2(@). Then A 1 A 2 = 1. g is similar to  in GL2(R). . (Hamud) Solution. i) If A 1 = A 2 . of determinant one. In case a). Describe the set of conjugacy classes of elements of G. let A be an invertible 2 x 2 matrix with real entries such that N=(d)?( if d = det A < 0. g = In case b) g = In case c). .26 SECTION 2 GROUP THEORY 1201 Let G be the group of real 2 x 2 matrices.) or NgN'= (1 1 ). then A 1 = A 2 = 1 or A 1 = A 2 = 1.
AS in case c).') a21 ( E:: :tz ) E G.) in G . There exists M E G L z ( R ) such that We take N = d4 . Thus in case d ) .')I* ( 0' 21 ) ( 0' ) ( i11. Then N E G and Thus g is conjugate to classes ( x". g is similar to similar (conjugate) to is not conjugate to ( i11. we have two conjugacy classes In case d). for if A*( where A = d e t A = a!l i :)A'=( . M i f d = d e t M > 0 or N=(d)i( if det M = d 1 0) M ' < 0.) or in G L ~ ( B ) .2). But (A i ) isnot in G. ) in G . then g is similar to ( 2 x".)]. all = a22 and = 1 which is a contradiction. ii) If X i # and E B (i = 1. then we have and = 0. we have two conjugacy classes [ ( i1 yl')] and [ ( 0' 1.27 Hence g is conjugate to conjugate to ( ( q' ) 1 1 ) or (1 1 ) i n G . g is inGand ( i11' ) GL2(B). ) in A2 Xi [( i!i :)] [( i! . we have the conjugacy . Thus in case c). So in this case.
As in the above. we have the conjugacy classes .isino cos 8 sin8 cos8 ) in GL2(C).7 implies that A and B are similar in M2(nZ).  O i sin 8 ) in G L ~ ( C ) . for if sin8 COSB sin8 cos8 such that cos8 sin8 ) Al = ( cos8 sin0 det A = (a:l + ail) = 1. B E M 2 ( B ) are similar in M2(~. we conclude that g is conjugate to cos$ sin6 or sin8 cos6 sin$ C O S ~ cosB sin8 in G.2).28 iii) If XI # A 2 and X i E C\E (i = 1.i s i n 8 (since XI . which is a contradiction. X2 11 < 8 8 # 0. where T. we denote A1 = cos8 + i s i n 8 and = cos8 . So there exists M E G L 2 ( E )such that cos8 sin0 MgMI= and also similar to ( cos8 ( cos8. So in this case. g is similar to sin8 As it is well known. But cos8 sin$ and cos0 sin0 are not conjugate in G. A2 = 1 and trace ( g ) = A 1 + A 2 E E ) . that two matrices A .
a) When forming a matrix in G L n ( F q ) . the set of conjugate classes of elements of G is U { [ ( $ x". Thus we can conclude that the order of GL.the group of invertible n x n matrices over the finite field Fq with q = p " . c) Determine the number of Gconjugates of U. the third row in qn . d) Show that g E G has ppower order iff g = I N with N n = 0.1 ways (a row of zeros not being allowed).) Xz and XI 3 XZ = 1 sin6 c0s6 cos6 1 0 (61 1202 Let G = G L n ( F q ) . b) Deduce that every psubgroup of G is conjugate to a subgroup of U . ) ] ~ A I .(Fq) is + . Make use of Fqn. Deduce that U is a Sylow psubgroup of G. E El " {[(i Y ) ]I b = * l } { [( )] < < .q ways (no multiple of the first row being allowed). the second row in qn .we may choose the first row in qn . p a prime.q2 ways (no linear combination of the first two rows being allowed). (Columbia) Solution.29 and cos6 sin6 sin6 cos6 To sum up.1 Hint. the subgroup of upper triangular matrices with 1's on the diagonal a) Calculate the orders of G and U . U = Un(Fq). e) Show that G contains elements of order qn . and so on.
j) place.. By this way.l M " h = 0. Since p j [G : U ] = . .1.. d ) Suppose g E G has ppower order. there exists some 1 U12 0 u=[. the second row in q n P 2 ways: the third rows in q n ... 0's elsewhere... .l M h = I + N where N = h .' & U (by matrix multiplication). Conversely.I. Then for any 1 5 i < j 5 n. of matrices ( a i j ) E G with a. . Denote h g h . x F. 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 ) . U is a Sylow psubgroup of G..30 When forming a matrix in U n ( F q ) ..3 ways and so on. 1 . Obviously Mn = 0 and g = I h .. Checking the last row of g ( l + E i j ) and u g . So the order of T is b Hence the number of Gconjugates of U is i=l n (qi n . c) By Sylow's Theorem. i=l b) It follows directly from Sylow's Theorem.l ) .l ways..l .l M h and N" = h . j = 0 if i < n .j = 0 (0 5 j < i 5 n). There exists an h E G such that h g h .we may choose the first row in q n . the number of Gconjugates of U is equal to [G : N G ( U ) ] where N G ( U ) is the normalizer of U in G.. = 0 if i < TI. Suppose g = ( g i j ) n x n E N G ( U ) . j E V such that g ( I E j j ) = u g where Eij is the matrix with a lone 1 in the (i..l ) / ( q l)n. if g E T: It is easy to see that g U g . .1)th row. . Uln ~ 2 3 ::: . + . So gn.l E U . Then checking the ( n . we get g i j = 0 if i > j Hence g E T = T n ( F q ) the set . So g N G ( U ) and we have N G ( U )= T .l = I + M where M is a matrix with zero on and below the diagonal. we get gni = 0.or gUg' = U . matrix multiplication shows that 0 is a group epimorphism whose kernel is precisely U = Un(Fq). we get g n . x . Then ( 9 ) is a psubgroup and ( 9 ) is conjugate to a subgroup of U by b). by mapping a matrix onto its principal diagonal. n + We can define a map 19 : T + F..
(IHI . It follows that G contains element of order q" . We fix a base of Fqn over Fq.1.1) + 1. we have a linear map al : Fqn FqR.31 Conversely. ~ ( u = at. $ G = Gn(Fq). the Jordan canonical form for N is a matrix with zero on and below the diagenal. So the order of g is ppower. 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. (T is injective..1. then we can obtain a group homomorphism (T : FLf. a l ( z ) = a z (over Fq) which is invertible.. f + 1203 [GI = IgEG U gHg' 5 [G : N ( H ) ]. ) Obviously. It is well known that F. has an element with order q" . Thus g is similar (conjugate) to some matrix in U . . For any a(# 0) E Fqn.
.. So (a) is contained in some maximal subgroup of G (since G is finite). = G N is a composition series of G with composition factors H and H i / H i . Then for any a E G . . K = (1) = {*l} be the subgroups of G generated by i and 1 respectively. But G I K = { 1.2.32 (b) First we claim that G is cyclic. (1) = KOa K 1 a . G has only finitely many subgroups of index n. . then (1) u H a H 1 a . UEG 1204 Suppose that H and K are normal subgroups of a finite group G and that GIH = K. (a) < G. (b) Let (1) = KOa K1 4.l Ki/Kil (i = 1. 1205 Prove that if G is a finitely generated infinite group then for each positive integer n. K a G and GIH is cyclic __of order 2. j . (a) Give an example to show that G / K need not be isomorphic to H . aK n P 1a K U G is a normal series of G with factors KiIKi1 (i = 1 . a K. (a) Let G = Q8 = { f l . K is not isomorphic to H . n ) and Hi ~ H i + (i = O . = K be a composition series of . Obviously.l a H . a H. . then G I K N H . l Suppose H is simple. . * * * .. ..1). there exist subgroups Hi such that Hi I> H . n ) and G / K . Suppose that G is not cyclic. Since GIH N K . (b) Show that if H is simple. . .*i}. Then we have G = U gHgl by assumption. 3 t k } be the Hamilton’s Quaternion group and H = (i) = {3tl. . . H i I H N Ki ( i = O . . By (a) G = H .. f j . it is easy to see that G is cyclic of prime power order since all of its maximal subgroup are conjugate. Let H be a maximal subgroup of G. . we have the isomorphism G / K N H . n). . . & i . By the JordanHolder Theorem. K . l .i. Hence G is cyclic. This contradicts the maximality o f k . (Columbia) Solution. Now. n . I c } N K 4 (Klein four group). we have H a G. so it is isomorphic to K . On the other hand. 2 .1 u K .
Since G is finite generated. there are only finite number of homomorphisms from G to S. where GIH denotes the set of all left cosets z H ( z E G ) . Show that if H 5 G (finite index) then there exists a homomorphism of G/HG into S[G:XI. ( Co2umbia) Solution. Thus G has only finitely many subgroups of index n. It follows that the set g€G {HGIH < G . g .. Let H be any subgroup of G such that [G : H ] = n. Let H 5 G and define HG = g€G glHg. H defines an action of G on G I H . Show that G is solvable but not nilpotent. and . [G : H ] = n } is finite. Then G x G I H + G / H .lR).+))is an abelian subgroup of G. (Columbia) Solution. Since G/HG is finite. G/HG has only finitely many subgroups of index n. with HG as its kernel.33 Hint. the symmetric group on n letters. ~ ' ) Matrix multiplication shows that 0 is a homomorphism. where S. This induces a homomorphism G + S. The kernel of this action is HG = glHg. E lR b 1 ( C_ GL(2. 1206 Let G= {( a!l ) la E lR*. Suppose . n Hence the group GIHG is isomorphic to a subgroup of S. Let 0 be the map G + R* x lR* mapping a!l ) to its diagonal ( a . ( z H ) = ( g x ) . and GIN E x IR* is also abelian. Hence G is a solvable group. N = ker0 (21 (lR. denotes the symmetric group on n letters. . ker8= {( ) 16tnJN Obviously.
Since nilm.. Thus T p r n . G = G L F ~ ( V Recall that IGI = (p" ). then f ( X ) = f l ( X ) f 2 ( X ) . Show that H can be simultaneously upper triangularized.fi~(X)I(X~~~l) . Let . f k ( X ) for some distinct irreducible polynomials fi ( 1 5 i 5 k ) . Denote n = degfi(X) i and m = n1 . x!!l So ay+ bx' = xb+yal for any x Thus # 0. Then X 1 f ( X ) .(G) = a * * = G $ where Ci+l(G)/C. y E IR.l = 1. is isomorphic to a subfield of E . ) EG. f(X) is separable (remark: here separability means that f(X) has no multiple roots). Hence f(X) = fl(X). Then for any ( . that is. ( 1 5 i 5 k ) . f i ( A ) ~ ( A P r n .fi(X). Hint. (Indiana) Solution.  .frnl = 1. 1207 Let p be a prime and let V be an ndimensional vector space over Fp.(G) C ( G / C i ( G ) ) = (Co(G)= 1 ) . a = f l .l ) .p ) (p" . So G is not nilpotent. It follows that b = 0. n2 'nk. Hence = C1(G)= C2(G)= * * .p " . Then IEl = p" and Ei is a field of order (1 5 pni. So E contains an element ~i whose minimal polynomial over Fp is f i ( x ) .l)(pn.. (a) Let T E G and f ( X ) be the minimal polynomial of T over Fp.l . that is.l ) .X over Fp and Ei = Fp[X]/(fi(X)) i 5 k ) . Find a Sylow psubgroup of G. there is a basis of V with respect to which all of the elements of H are upper triangular..= C. (b) Let H be a subgroup of G of order a power of p .34 the center of G. Prove that a transformation T E G is semisimple if and only if Tprnl= 1 for some positive integer m. E. A direct discussion as above shows that C ( G / C ( G ) ) ( 1 ) .. (a) Recall that a linear transformation is called semisimple if its minimal polynomial is separable. If T is semisimple. Since ~i # 0 and 7. Let E be a splitting subfield of XPm .
(V). Now let H be a subgroup of G of order a power of p.35 Next suppose TPm' = 1 for some positive integer m.2 . ( c ) First we claim that H is not normal in G. q such that its commutator subgroup K is of order q. (p"' . the subgroup of upper triangular matrices with 1's on the diagonal.we may choose the first row in pn' ways. So H K = K H is a subgroup of G.p ) * * * (p" .l)(pn . U is a Sylow psubgroup of G. Since )GI = (p" . Hence G = H K and H 21 H / ( e ) = H / H n K 21 H K I K E G / K is abelian.. (a) Since K is the commutator subgroup of G.p = pv (when forming a matrix in U . Since (XP"'' . IHJ = p 3 . K is normal in G and G / K is abelian. G = H K is abelian. (b) Show that there are elements h E H and k 6 K such that hk # kh. for any h E H and k E K . Then JUJ p"l . (c) From (b) show that p divides q . (b) Let U = U. the second row in pn2 ways and so on). This contradicts the fact that the commutator subgroup K of G is of order q..  . (p" . Let G be a group of order p 3 . = . Otherwise.1. Thus we have proved that H can be simultaneously upper triangularized. (a) Show that H is abelian and G = H K . Then f(X)I(P"'' m 1). By Sylow Theorem.p"') n n1 = p.l 5 gKg' = U. . 1208 Let p and q be distinct prime numbers.1). (b) Suppose that for any elements h E H and k E K holds hk = kh. By Sylow's Theorem.. H n K = { e } and lHKl = p3 q = IGI. f ( X ) has no multiple roots. Hence there exists an element g in G such that g H g . Since K is abelian and H is abelian. ( p . Thus T is semisimple.1) . hkhlkl E H n K = { e } and so hk = kh. (Indiana) Solution. H is contained in some Sylow psubgroup K of G and K is conjugate t o U .I).p " . This proves (b).2 # 0. Let H be a pSylow subgroup of G. Since IK( = q .1)' = XP .
3e2 6e3. 2 2 . Thus H .36 the number of pSylow subgroups of G is greater than 1 and divides q . q . the alternating group of degree 4. Denote + + + + 9 3 6 . and g.1. it is easy t o see that rp = 1. by Sylow’s Theorem. G may not contain a subgroup of order p . fz = 3el 3ez and f3 = 3el . By Cauchy’s Theorem. Since p . q are primes and p > q > 2 . Show G has a subgroup of order p q . which is normal in G and of order p . plq . is a group of order 3 . + 3f 3e. Let G be a group of order pq’. (b) If q = 2. rPlqz and plrp . p > q > 2 .3f+6g Give a decomposition of A as a direct sum of cyclic groups of prime order or infinite order. Then. So G has only one Sylow psubgroup H . So G has a subgroup K of order q. G also contains an element of order q . 1200 (a) Let p . q. q be primes. (Stanford) Solution. K is a subgroup of order p . (b) What can you say if q = 2 (and p > q ) ? (Indiana) Solution. K be the subgroup of F generated by f1 = 9el 3ez 6e3. A4 does not have a subgroup of order 6 . = 0. q since H is normal in G. For example.1. Aq. Let F be the free abelian g r o u p g e l $Zez $Ze3. Obviously A E F / K . 1210 Let A be the abelian group on generators el f. (a) Let rp be the number of Sylow psubgroups of G. So it is Again by Sylow Theorem. = 0. subject to the relations 9e+3f+6g 3e = 0.
f4)' = Q . fL Then F =Zei $Ze'. e$>' = P .6. Let (e.e.6. e 2 . i.37 It is easy t o get the normal form diag{ 3..l ( e l . e i .0} ( P and Q are invertible matrices in M 3 ( Z ) ) . $ Z e L and K = Zfl + Z f 2 +Zf3 = XCe.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.. a) Show that 2"/MZn finite.. Suppose that M is invertible when viewed as a matrix of rational numbers. that there exists an n x n matrix N with rational entries so that M N equals the n x n identity matrix. e3)' and (fi. 1211 Let M be an n x n matrix of integers. (Stanford) . 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.f 2 . f3)'. View M as an endomorphism of 2". @ Uel. is b) Show that the order of Zn/M57" is equal to the absolute value of the determinant of M . (fi.
} where f.. (ei. f n }. 1 * * f n )'.f )' = Q . .dn} be the normal form of M in M. (Haward) + . } .. ([ + 2 .l).. .Solution. . .(Z) be invertible matrices in M. b) Let D = diag{dl. d z ... . M * is the adjoint of M . 0bviously. . .el.. .ek} is another base ofZ". Obviously det D = dl . . e n } be a base for the free module Z"..... . . . Then express G as a product of cyclic groups. (f. and the fact p n / l M p n l = ( I det MI)". . ..e.. 6 ( r ) = M * X is injective. a) It follows directly from b). e n ) . 2 . e2. . .el. and P . . .(Z) such that D = Q M P .)' Then {ei.. It is also obvious from the facts that the map g : Zn/MZn Z " / l M p " .. = det Q .Denote (ei.fn )I I = Q M P (ei. (el.. . . . el.  Hence 1212 Let D = a [ [ with E = *.. d. . . . . en)' = M P (ei. . . fn)' = M . . (fi f~r : . eh. f. where [MI is + the determinant of M . (fil f.. Let { e l . Then M P is generated by { f l l f 2 . .. eh . Let (fl. d.l).. . ] Calculate the order of the additive group G = D2/K where K is the Dsubmodule of D2 generated by ( 2 t 11[ ..f. M Z n can be generated by {f. det P = det M or . .4) and (21. (fll f~. . Since Q is invertible in M n ( Z ) .det M . .. . 1' 1 = Q . ( . .. . . Q E M. .) / I = diag{dl.21).. e. . e2. 7 * * * .. . I / fz. en)I . .(Z). det M .dz. f. n . where the di # 0 and dildi+l (i = 1 . dz..ez. . ..ek)l = P'(el.
39 Solution. Let { e l l e 2 } be a base of the free Dmodule 0’. 0 . 2el.<ez) + (ez) + b(el+ (el .e2 + 0 .21(el + 21(e2} is a generating +b(O e l subset of K as Zmodule (additive group).e2 . + + (a bt)[(l+ = a ( e l + 2tel .5(ez). . 2 1 e l + 21e2.+ 21 (el + 0 . .2(e2).e2.2/(3) @Z/(21)and the order of D 2 / K is 63. 0 21 .el + (el . {ell(el. For any a bt E D .e2 + + + (e2) + b(2e1 + (4 + Oezl + (1 + t ) e ~ l .e2 . e2 + 21 . e2} is a generating subset of the Dsubmodule K of 0’. Then + 0’= Del @ Dez = Z e l @Z(el @Zez @Z<e2.2(e2. It is routine to get the normal form 1 0 0 0 N = [ 0 .l] 0 . <el + 21 .e2 .9 0 ~ 21 5 ~ . D = Z [ t ] = Z@Z< (as additive group).(el . It follows that D 2 / K .2el + ( e l 4e2 + 5e2. for A in hf6x4(Z) and to find invertible matrices P E &(Z) and Q E 1M4(Z) such that PA& = N .0 .e2 .e2 + (ez. el 21.4ez + (a = a(21 . It is easy t o see that {el + 2te1. e l + b()(2le1+ 21e2) + 0 . 21.1 2 1 1 A= 1 1 2 . e2.  ( a b0[(2 ()el a(2el + < e l . (e2).5(e2. Denote 1 2 . and {(1+2<)e1+(1+<)ez1(2+ ()el + (4 + () .(el .(e2} i s a baseofD2asZmodule. Since ( is a root of the irreducible polynomial z2 z+ 1.
c'. we claim that matrices in S L ( 2 . (Stanford) Solution. If a Hence > c > 1. Similarly. Z )is the group of 2 x 2 matrices with integral coefficients and determinant = 1.there exist some q . Z ) is generated by ( 1 rtl ) ) ( i1 If both a and c are nonzero. a and c are relatively prime.be = 1. 1 5 c1 < a . let c = ha + e l .d' EZ. Since det M = ad .40 1213 where and S L ( 2 . It is easy to know that if a = 1. if c > a > 1. Suppose that M = Prove that S L ( 2 . a1 E Z such that 1 5 al < c and a = qc+al. then for some d' EZ. Z ) with the form ( 1) is a prodct of the for some b'. . and if c = 1. for some b' E Z.
 1215 Let P be a Sylow subgroup of a finite group G. N . Thus we have G = K . c'. there exists h E K such that gPgl = h P h . Hence gPgl is also a Sylow subgroup of K . If N is the normalizer of P in G (i. d' E Z.41 According to the above discussions. X ) with the form for some b'. ( i1 )) where b ' . Let N = { x E G I xPx' = P } . (Indiana) .l . By Sylow Theorem. Z ) = generated by ( 1 3zl ) ( and 0 1 ). then y E N. So we have g = h . H 2 N . it can be derived that M = is a product of the matrices in S L ( 2 . For any g E G. It is easy to check this by the property of the elementary matrices. d ' E Z . K a normal subgroup of G and P a Sylow subgroup of K .l = H . we have 9Pgl C gKg' 2 K since K is normal in G. it suffices to prove that all are in (( 1 fl cc) . Hence h'gP(h'g)' = P and h'g E N . 1214 Let G be a finite group.e. (Indiana) Solution. N . Prove: If y E G such that y H y . This completes the proof. c ' . N = {g E G 1 gPg' = P } then show that G = K . Let H be asubgroup of G. So to prove that S L ( 2 . h'g E K N ..
N cH.(1 1 2 1 . then.b) E G x G I ab = ba}J IG x GI (a) If G is not commutative. P is a Sylow subgroup of H . (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. Obviously. obviously. yPy' is also a Sylow subgroup of H . (b) Give an example such that P ( G ) is exactly 5/8. So 1 y=h. Hence h'yP (hly)' = P and h'y N .42 Solution. If G is not commutative.hlyEH. (Stanfod) Solution. 1216 Let G be a finite group.b)E G x G I ab = bu}l = P(G) = L . prove that P ( G ) 6 5/8. Since J{(a. Since yPy' & yHy' H. + ) 4 =8 5 . So there exists an h E H such that yPy' = hPh'. IG/C(G)I L 4. The probability of G to be commutative is defined by P(G) = I{(a.
3 5 l C ~ ( a )< 8. It is easy to see that P ( G ) = l =/ 518.z2 = y4 = 1 zy = y3z}. obviously. y}.C ( G ) . For any a E G . . Then IGJ = 8 and C(G) = (1. and so ~ C G ( U )4.43 (b) Let G be the dihedral group 0 4 = { z * y j I 0 5 i 5 1 . 0 5 j 5 3. .
Hence Z [ m is Euclidean. Obviously T and 6(T) = 6 ( b ) * ( l p . we define 6 ( a ) = m2 2n2. Z [ G ] is a unique factorization domain. a) It is readily known that Z [ n ] is a subring of a. Suppose that a. (a + t) 2* + 21v  wl2) < 6(b). (1. Then ab' = p v. For any a = m nz/2 E Z[z/2].\/z where p and Y are rational numbers. 1) is the set of units of Z . .v)Q] +v E Z[] Q is in Z [] and r = a . b) Using a).n&? is also a n irreducible divisor of 2 0 . (Hurvard) Solution..G) zg. So we have a map 6 :Z[]* + Z?. find all integer solutions to the equation y2 + 2 = x3. Then w + + + a = b(p+vG) = b [ ( u + p. and if alyo then + + + + n.2112 5 6(b).vI 5 1/2.v)Q.u ) = bq+r where Q = u + (v + v . hence an integral ' domain. [] Suppose (20.u ) + b(v .pl 5 1/2 u and l . 20 is not a prime element in Z[.\/z].yo) be an integer solution to the equation y2 2 = x3.bq = b(p . If a = m nis an irreducible divisor of 2 0 . a H 6(a). ti = m . b # 0 E a[].44 SECTION 3 RING THEORY 1301 a) Prove the ring Z [ n ] is Euclidean. we have (yo Q)(yo . In the ring Z [ G ] . We can find integers u and v such that I . Since = the integral divisor of yo + 2/2 or yo can only be f l . b) By a).
2n3 = (3m2. . 0 < p n 5 9 and A2 = ~  n2 = ( n .(R)*.2n2).22. ? 2 . yo is of the form ( m nJ2)3.a E R*. T .. ..(R) the ring of n x n matrices over R. .S ) n S # 0.S C Z/pZ and Z/pZ has only p elements. (a) If a E R is nilpotent (am = 0 for some m) then 1 . Let S be the set { b 2 I b E a/$}.(R) = M.y 2 = i + j . ( q ) } 2 + 1303 For a ring R. When p = 2 . . 2 (q)2} y . c E (Harvard) Solution. R* denotes its multiplicative group of units.p ) 2 = has exactly elements.j # O ( O < i + j 5 p .1/2.T. Then S = {o. 9 G . 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. : i 0 < i .c3ly0 . M. On the ) are other hand.(q)2} elements in S. n E Z. So without alyo loss of generality.S := {a . Let p be a prime.b2 = c 2 . {o. 2 .b2 I b E Z/@} has elements.. for any A2 E S where 9 < n < p .6mn2 1 = 3m2n. that is a = b2 c 2 . Now for any element a E Z/@. i .. has exactly elements.z.45  from which it follows that a31yo fl. c E a/@ such that a . On one hand for any i # j .T. Assume that p # 2. exactly (a . Hence there exist 6. and GL.Thus 2 S' = { b 2 I b EZ/$} = { a .1. it is trivial. there exist b. Show that for any element a E Z/$.rn..j < 9 if i > j .0 I j 5 e. the set a . Hence + + + yo = m3 . (b) Let J be a nilpotent ideal of R (J" = 0 for some m).0 < j 5 q. n.. . Since a .
(e) We consider the field extension 3 C 3. % ( R / Jis surjective.(J).c E F.s + 3.4 6 Show that GL.} is a pSylow subgroup { A = ( a i j )I = { A E M. 1))  IA =I mod M .3. with kernel ) (d) Show that U = {( l a b 0 1 c 0 0 1 ) a. 2 = diag{i.".s. .. There exsits a E 3. T. 3: H az is an invertible linear . Such that Fi3 = ( a ) .).T. is it easy to see that U is a pSylow subgroup of GL3(3.(R) = I+M. Obviously. ( J ) } (c) By calculating the possibility of the row vectors of A in GL3(Fq). . : F. b. .(R) t G L . .
. bi = 2a. ad. and ” a = 2 4 .47 transformation over Fq.(a) = q3 . Then we have bf2 = So must be a square number. = 3 . we get an infinite set of integral solutions { ( a i . Let A E GL3(Fq)be the matrix of T. 512 31. By (b). bo) is a solution of z2 .Since . = 2aP1 . it is easy to see that if (ao. II = 512. Then J 2 = 0 and % = R / J N Z / U .  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 = 1.1)(52. &om the above consideration. v.2y2 = 1 (i.1 .2y2 = 1. Then a must be b) odd and b even.1. the order of A in GL3(Fq) q3 .. b E Z such that a2 . the kernel of the surjective homomorphism G&3(Z/2w) GL3(Z/5Z) 3 is I and + M 3 ( J ) .e.2y2 = 1 .1 . 3 . Let a = 2a’+ 1 and b = 2b’.bo) is a solution of z2. . Hence either both $ and a f 1 are square ’ numbers or a and ’ are square numbers. a .Then P/ ker cr ‘v U and JP] I kercr] . 2 ) is an integral solution of x 2 . then (al = 2 4 .2b2 = 1). say a’ = 2b. If $ and a ’ 1 are square numbers. Thus P is a 5Sylow subgroup of G L 3 ( Z / 2 U ) . relative to some base for Fq. Suppose that (a.1.2y2 = 1 . = U 1304 Describe an infinite set of integral solutions ( a . + so Let (GL3(Z/2U)I= 2 7 . (ao. ( 3 .lbil (i 2 l). then b = b. (Indiana) Solution. b ) of the Pell’s equation z2 ./Fq. 3 .31.2y2 = 1. So if we take a. is an integral solution of x 2 . b = f2aobo.2y2 = 1 . + + . a’ 1 = a. b = 2aobo) is another solution of z2. s 3 .1 . for some integers ao. bo.bi) 1 i 2 0) of the pell’s equation x 2 . Obviously.1) = 2 7 . bo = 2 and a. is ( f ) Let R = Z/2M and J = U/2U < R .1)(5 . Obviously [ I M 3 ( J ) J= 59 IGL3(Z/U)I = 53(53 .
e. ap' = 1. and for any a E R.p z .Z / p Z .2y2 = 1. the intersection of all prime ideals of R is 0. If R satisfies the a. By (a). Give example (without proof) of such a ring which is an integral domain but not a p.z = 0.1) = 0. RIP . 1305 Let p be a prime. that is. we get a solution (99.2y2 = 1. For example. prove that R is isomorphic (as a ring) to (b) In the general case (i. It follows that R has at most p elements. from the solution ( 7 . (ascending chain condition) for finitely generated ideals then show that R satisfies the a.. then ( a = 2ag + 1 .2y2 = 1 and a = 2 a i + 1.bo) is an solution of ' . for all ideals. (a) If R is an integral domain.i. Char(R) = p and R ZlpZ. RIP is an integral domain and for any a E RIP. if (ao.. (b) Since for any a E R. the nil radical of R is 0.. (a) Suppose that R is an integral domain. a'+ 1 = 2bg. R not necessarily an integral domain).d. 1306 Let R be a commutative ring with 1.c. For any a (f 0) E R. (Indiana) Solution. Hence = all prime . u p = a. a is a root of the polynomial zP . b = 2 f 2 a o b o ) is an solution z2 . each of which is isomorphic to a/$. Conversely. 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.2y2 = 1.c. (Indiana) . Since p is a prime. 5 ) of x 2 . Hence R is a field.a. If a' and are square numbers. b = k2aobo.c. then (ao.c. say a' = ug. Thus we have proved that R is isomorphic to a subring of a direct product of rings. since UP .bo) is an integer solution of z2 .48 Remark.2y2 = 1.a u(aP1. a  P For any prime ideal P of R.70) of z2 . Let R be a commutative ring in which u p = a for all a E R.
(a) Show that S = (1 a I a E A} is a multiplicatively closed subset of R.l R and those prime ideals P of R such that P A # R. For any 1+ a and 1+ b. i.i. It follows that S is a multiplicatively closed subset of R. a .c.c.l A is the unique maximal ideal of S . SlR = R. (d) If A is a maximal ideal of R. This is what we need to prove (b).. for all ideals. 1307 Let R be a commutative ring. what is S . 1 = 1 + 0 E S . b E A.l R by (b).l R ? (d) If A is a maximal ideal of R.l R and those prime ideals P of R such that P n S = 0. then for any ideal I of R.c.the polynomial ring over Z. for finitely generated ideals. + 1308 f :M Let I be a nilpotent ideal in a ring R. and let + N be an Rhomomorphism. (b) Show there is a onetoone correspondence between the prime ideals of S . (b) It is well known that there is an onetoone correspondence between the prime ideas of S .d. then A J ( R ) . Hence Jacobson radical of R. (a) Obviously.49 Solution. we can construct a strictly ascending chain of finitely generated subideals of I .l R is a local ring. 1 is finitely generated. the localization of R at the maximal ideal A. Hence R is a Noetherian ring. but not a p. For otherwise.c.let M and N be Rmodules.satisfies the a. for all ideals. Obviously.c.e. (1 + a ) ( l + b ) = 1 + (u + b+ab) 6 S. R satisfies the a.c. P n S # 0 if and only if P A # R here. So S . (c) If A is contained in every maximal ideal of R. Show that if the induced map . If R satisfies the a. (c) If A is contained in every maximal ideal of R . and it is easy to see that S .l R is isomorphic to RA. what can you say about the structure of SlR? + + (Indiana) Solution. X [ z ] . then S . Let A be an ideal of R. the So every elements in S is invertible in R.
(Columbia) Solution.. I . If is the completion of F ( t ) with respect t o 1 1. . Since I maXMt)l. . we have i) V. 1309 Let F be a finite field containing 5 elements. Since f :M/IM + N/IN is surjective. then f is surjective.50  f :M / I M t NIIN (Harvard) is surjective. Is(t)lPH 1 + m. 1 (called padic absolute value) on F ( t ) .P(t)). * = I n . N/f(M) = I N Hence + f(M)/f(M)= N / f ( M ) . ii) V. It follows that I .(f(t)) + % ( S W and 4 v. we get an absolute value 1 .dt))= V. c ( t ) E F [ t ]and * ( / ( b ( t ) . We define V. : F ( t ) + I?i by %(O) = 00 and Vp(f(t)) = Ic if 0 # f ( t )= ~ ( t b) t )~c ( t ) . Let R = { a ( t )1 a ( t )E F ( t ) .(f(t)) = 00 if and only if f ( t ) = 0. because I is nilpotent. N / f ( M ) = 1 2 .V p ( f ( t ) ) .(f(t) + d t ) ) L fin(Vp(f(t)>. Let p = p ( t ) = t . N / f ( M ) = * . by defining I f ( t ) l .Ia(t)lp5 1) and M = { a ( t )E R I la(t)l < 1). 1 is nonarchimedean (If(t) W l P + Suppose F ( t ) is the completion with respect to 1 Ip. Obviously. Obviously. P ( t ) ) = 1 = (. show that the set of a(t) E F ( t ) satisfying la(t)l 5 1 is a local ring.VP(dt))). Explicitly construct one nonarchimedean absolute value I I on F ( t ) . N / f ( M ) = I . = 0. 4 l IP(t>lP) .(t).where b ( t ) .(f(t). So N = f ( M ) and f is surjective. L m={I4t>Ip. f ( M ) + I N = N .1 E F [ t ] . Then. Solution.N / f ( M ) = . Let t be transcendental over F . = 2 .
. 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 ) . .X z . . bj. Y2. . . a j 2 . . Suppose that { Y l . . . . = UilY1 * * * U l m Y m fa + + + for any X . . 2 . Since X I . . of degree two or greater. b ( t > P ( t ) l P = l+)lPIP(t)lP>  . x . .any For R is a subring of the field F ( t ) and M is an ideal of R. in the polynomial ring a'[X1. where b j l .. and X I .X 2 . ' m aijbjk j=1 = 6ik (i. .l E = F ( t ) is the inverse of a @ ) .X.51 and . . In the same way. . + * * * gj for j = 1 . . ) of the ring C [ X 1 .of degree 2 2) j=1 ( i = 1 . 2 . X .. cannot be generated by fewer than n elements.X ... and f j is a sum of monomials in ' Y 1 .2. 1310 Prove that the ideal generated by X I . We We can write x. . . . a ( t ) .]. . a i m E a . .X 2 . . . Thus R is a local ring with maximal ideal M . k = 1. . X 2 .Y. .So. . of degree two or greater.X .. ] ( StanfonE) Solution. where a i l . n )..rn. E a'.. a ( t )E R\M.n).} is another generating subset of the ideal ( X I . . .X . * * * . .X .Y. . .. .* . we also have y j = bjiXi + bj2X2 + + bjnX. have to show that m 2 n.l E R and a ( t ) is invertible in R. . . . are algebraically independent over a .
1) by f(x)= { 2x. since + 1= (a + b ) 2 n E (a") + (b") = I1 + 12. Prove that A N A1 x A2 for some (nontrivial) rings A1 and A2. By Chinese Remainder Theorem.we have I J N ( A ) (the nil radical of A ) and I + J = A . we have I1 n 12 = 1 1 I2 = (a") . and since ab E I J C N ( A ) . 22  if 0 5 if 1 5 22 22 < 1. b E J such that a + b = 1. There exist a E I . (b") = (a"b") = 0 and thus A N A/I1 x A/I2 where A/I1 and A/I2 nontrivial. there exists an integer n such that (ab)" = an b = 0. 1312 Define f : [0. ( Stanford) Solution. Since every prime ideal of A contains either I or J but no prime ideal contains both I and J . 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 .52 1311 Let A be a commutative ring. < 2. 1. = (Indiana) . Find all z such that f(f(f(f(f(f(z)>>))>2. Let 1 1 = (a") and 12 = (b"). " Then I1 and I2 are proper and I1 I2 = A . 1) + [0.
53 Solution. [a]. Use this result to prove that T is an integral domain which is not a unique factorization domain. By using the orthogonality of the set ( l . l). (2 . sin nz.{ u } } * * . Since for any real number a and positive integer n. sinnz I n E nV}. cos nz + b. 1313 Let T be the ring of all real trigonometric polynomials N f(z) = a0 + n=l a. cos nz + b. g = deg f + deg g. it is easy t o see that N f ( z ) = a0 + n=l a. Define deg f(x) = N where U N or bN # 0. sin n2 0 .[nu .([nu] n [ u ] ) = = nu . ( 2 2 ) .[nu]= { n u } . Then it is clear that f ( z ) = 22 .[ n . (2 * ( 2 . f(f(f(f(f(f(.n . 1). 0 5 k 5 62. [u] denotes the largest integer We denote { u } = a .) z if and only if z = = &. z] if and only if 63z = [64z]. Show that deg f .{ u } ] = n(u .[u]) . Hence f(f(f(f(f(f(4 { n .[u] here. Since E [0. For any real number a .[2z] = ( 2 2 ) 5 a (Gauss function). c o s n z .632 = [64z] = [63z + z] (to be an integer) if and only if 1 2 63 Thus f(f(f(f(f(f(z)). (Columbia) Solution.  0 ) .n[u]] nu . we have = n { u } .[64 .[u])] = nu .) = (2 * (2 . for any z 6 [0.n[u] .[n(u.)))))) z = { 2 6 4 = (642). Notice that 2 = (642) = 642 .
g(z) = 1. d$ = . dM = b N . If f(z). ao. . sinmz.bncm 2 ) sinkx] . . Hence the units of T are {*1}. .g(z) are and a N d M : b N C M respectively. 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). then a N .bN f ( z ) = a0 and g(z) = c g + n= 1 an cos nx + b. 2 and so aNCM = 0.if and only if all the coefficients of f(z)..e. aN and bo. .b N C M . Now f ( z ) . Let N . Since (an cos n x + bn sin n z ) (cm cos m x + dm sin m z ) + mn=k c andm . are zero. CM.g(x)) = N + M = d e g f ( z ) + d e g g ( z ) . So T is an integral domain. Suppose that deg f(z)= N and deg g(z) = M . which is contrary to aN or bN # 0. Since c h d& # 0 .g(x) = 0 can happen only if either f(z)= 0 or g(z) = 0 by the above degree relation. Then we have a N = 0.. sin nz + c M m=l + cm cosmz + d. Thus we have proved that bN(c$ + d&) + = U N d M = 0. then the degree relation implies that d e g f ( z ) = 0 = degg(x). Hence deg(f(z). . If both of them are zero. bN = 0. b l . i. a l .
. 1 f f i s i n z are irreducible. If p = 0 or ( p . . we claim that the trace function ~ L I K 0.55 Obviously. y) ) # be the dual basis of u l .sin z) = (1 a s i n z)(1. . for any z E B .* u. If B is the integral closure of A in L . 1. 1314 Let A be a Noetherian integral domain integrally closed in its field of fractions K . First. The bilinear function ( 2 . u 2 .a s i n z). outline a proof that B is Noetherian. Then. a 2 . K Let 211. . Hence B C Avl + Av2 .. (the elements in L satisfying T T L / ~ ( u ~ = ~ . j for all i. = K [ a ]for some a E L. Since ) L _> K is a finite separable extension. Let ~ 1 .cj # 0. . . suppose that p # 0 and p I n. be the ~ +  minimal polynomial of a over K and j be the minimal positive integer such that p ] j and c j # 0. all the factors cosz f sin%. be nL/K(f(a)) = i=l C f(ai). + + Again by the degree relation.  + + + ) + . v. Let L be a finite separable field extension of K . * knvn . Denote a = a 1 . . For any f(a)E L. Hence T is not a unique factorization domain. Let p = Char(K) # and n = [ L : K ] . u2. E B . Since xu. Ici = ~ L / K ( ~ u E ) A for any i.&j) = (l)n+j+l * j . . . n ) = 1. 2 1 2. . a n the set of the conjugates of a. Since A is Noetherian.Let zn . . Thus cos 2% in T does not have an essentially unique factorization into irreducible elements. ( k i E B ) . Now. B is Noetherian. Thus ~ L I # 0. we get TrL. + Av.u. c. + T r L / ~ ( z y is nondegenerate since ' I ~ L I K 0. * * ~ (l).. . and cos z f sin z and 1 f d s i n z are not associates. it is easy to see that nLIK 0 # ( n ~ / K ( b = nb for any b E K ) . 3 has the form lclvl : k2v2 .clznl + c n ~ z . j). be a basis of L over K contained in B.. v 6 . in T we have cos 22 = (cos z sin z)(cos z . ( Stanfod) Solution. Then by using Newton's identities on the elementary symmetric polynomials.
. " Hence f is nilpotent. i. then f is nilpotent. suppose that A is Noetherian and all the elements ai are nilpotent. = fn = 0. bz. that is. Since A is commutative. a0 a12 nilpotent and n=O c anzn is nilpotent. I is finitely generated and nilpotent.and determine the integral closure of R.  1316 Let F be a field. If A is Noetherian.. ak is nilpotent. (Stanford) . + a k Z k Assume that is is nilpotent. then it is easy to see that f = 0. . all the elements ai are nilpotent. (a) Show that R. 03 It is easy to see that f = 0 " + + . Let I be the ideal generated by {ai I 0 5 i < m}. in its field of fractions. that if all the elements ai are nilpotent.b. a i . then all the elements aa E A are 00 n=O nilpotent.b. E I. .56 1315 If A is a commutative ring and A[[%]] the ring of formal power series over is A .. Conversely.. show that if f = c anzn is nilpotent. ao. So. prove the converse. Hence 2 n=k+l is nilpotent and so ak+l is nilpotent. be the subring of the ring of polynomials F[x] consisting of polynomials f0 + f i x + fizz + . If I" = 0. (Stanford) Solution. + f k x k E F [ z ] * such that f1 = fi = . blbz . Suppose that f = implies that a? = 0.e. = 0 for any b l .. Since A is commutative Noetherian. is a Noetherian ring. . is F ( z ) . (b) Show that the field of fractions of R. * . By induction.. and for n 2 1 let R. first we get that a0 is nilpotent. .
Then + 1. R. if a ( x ) / b ( x ) E F ( z ) is integral over R. It is clear that the field of fractions of R.1 + k * 4 E R. On the other hand. S is a multiplicatively closed subset of Z containing 1. we may assume that ( p . is F ( z ) . q ) = 1 and pl q k = 1 for some I . since z is a root of Xn+' . a ( z ) / b ( z )E F [ z ] .. by Let Obviously. z is integral over &. for any q / p E R. Hence F [ x ] is integral over R. a ( x ) / b ( z )is integral over F [ z ] . in its field of fractions. 1317 Describe all subrings of &. Obviously. Let R be a subrings of&. It follows that R = Zs where 1 s={O#PEZ)ER}. Since F [ x ]is an integrally closed domian.pl+qk . For any q / p E as ( q E Z. P P P So p E S and q / p E Zs.) ( Stanford) Solution. E R.57 Solution. (a) Since F [ x ]i s Noetherian and s is finitely generated a &module. Let Z s be the localization of Z at the multiplicatively closed subset S . E S). (b') Observe that for any a ( x ) / b ( x )E F ( x ) .. k E Z. Hence R C Zs.zn+' E R n [ X ] . Hence F [ x ] is the integral closure of R. is a Noetherian ring by ArtinTate Lemma (or Eakin's Theorem). Then R _>a definition. = q . p On the other hand. P . (A subring contains 1 by definition. Hence Z s C R.
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 ) . .
Thus { ~ ( z l ) . (b) If G acts transitively.YY = ( z . x 2 .. Solution. y m } two equivalence classes of zn be determined by z = x1 and y = y1 respectively. So all equivalence classes contain the same number of elements.. Hence m 2 n. z ) .there exists g E G such that g ( x ) = y. ..4)(x2 . If f has exactly p . show that the Galois group of f over & is the symmetric group S. Show the following.Z)(x.2 real roots and 2 complex roots. ~ ~ ~ . (g(zl). x . ( 2 1~ i ) g . ( 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. (a) is a n equivalence relation. Thus m = n.16) .' E . .u(zi))= ~ e .2 ( Cohmbia) is irreducible and determine its Galois group over &. (z2. G. y E X and leaves all other elements y) fixed) is an element of G . Since (x. 1) (a) By the defintion of . (which interchanges z. ~(2. Similarly. then all equivalence classes are distinct and contain the same number of elements. (b) Let { z l . . Hence is an equivalence relation. y 2 .Y # z). it is easy to see that is transitive.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 . Define a relation on X by requiring x y if either x = y or the transposition (2. on p symbols. we have    ( . For any 1 < i 5 n.Y)(Y. # Y. 4z2). 2) Suppose f E &[z] is irreducible and has degree p .  is reflexive and symmetric. a prime number.)) are n distinct elements belonging to the equivalence class determined by y1 = g(x1). and }{ y 1 . . we have n 2 m. Show that the polynomial    (x2 + 4) . Since G acts transitively on X .
Thus the Galois group G of f ( z ) over & is the symmetric group Sp on { T I . r P )is a splitting field of f(z) over & contained in a. Hence the restriction of the conjugation to E is an element of G and it is a transposition.. f(z)= (z2 + 4)2(z2 .1 + .. Since E =&(rl. Then.4)(x2 . and n > 1. there exists an isomorphism of&(r. f2. r p )is a splitting field of f(z) over & ( r i ) .r p } of the (distinct) roots by q + q1x. Consider the conjugation automorphism onG.16) .l. . say n. = rI(z .. 2) Suppose  P f) ( . (z. i. which is generated by all the transpositions.and the graph of y = g(z) has the form Fig. By Eisenstein criterion.f4.y)E G. m = 1 and n = Card(X). . . since f(z) is irreducible and f ( r i ) = 0 = f ( r j ) . @[XI. This maps f(z) to itself. r j E X.. on X .e. all the equivalence classes contain the same number of elements.this isomorphism can be extended to an automorphism q of El&.2 is irreducible over &. For any r i . n. by (b). Identify G = Gal(E/Q) with a permutation group of the set X = { T I . and also over Q ( r j ) . . Let g(2) = (z2 + 4)2(z2. Since Card(X) is a prime. r p } . It follows that G is the whole permutation group of X . The real roots of g(z) are 0. So Card(X) = m . . . Thus for any z. . z y. Thus the conjugation interchanges r1 and r2 = Fl and fixed all other real roots.)/& into&(rj)/& by sending ri to r j .y E X .16).Ti) i=l in So E = & ( T I . Let r1 and r2 be the two nonreal roots of f(z).60 (c) Suppose G acts transitively and contains a t least one transposition.4)(z2. Then q E Gal(E/&) and q(ri) = ~ j which shows G acts transitively . . where m is the number of distinct equivalence classes determined by .
Define f 1 .e. . Now for any i. Since H is normal in G. Hint. Use the generator CT : Cp + ( C p ) g of the Galois group of&(Cp) over &..1 and put f = ( p . a 2 ( ~ p ) . 1402 Let p be an odd prime.1a(base)for&(<p)/&.)g is an automorphism of &(GI over & and G = Gal(&(<. Inv(H)/& is a Galois extension and Gal(Inv(H)/&) 21 G / H .l)/e.qa = C(<p)gej+i j =O Show that. Let Cp be a primitive pth root of unity and g a primitive root ( mod p ) (i. = ~ } is < p Let H =< C T ~>< G. Hence the Galois group of f ( z ) over & is ST.2 has five real roots and two nonreal roots. = @(cP): InvH] = IHI = f and [Inv(H) : &] = e . vi generates a subfield of &(Cp) of degree e over &. . ~ ~ ~ . a divisor of p .Fix e.it is easy to see that f(z)= g(z) . CT : Cp H (&. Obviously.. g is a generator for (Z/pZ)*). Then J H J f. CT’. Since g is a primitive root ( mod p ) .aP’ = 11.61 Since I g ( n ) l > 2 for any odd integer n. for any i. (Columbia) Solution.)/&) =< CT >= {c.up’. { a ( < p ) . C pP .
Thus qi generates a subfield of &(&) of degree e .AB 6 K .yB E K ( y ) [ X ] not 0. (a) Let A . Show in succession that the polynomial A . e. and let H denote the subgroup of G consisting of the substitutions 2 + 2 b with b E K .e. Hence j=O n (z q j ) is the minimal polynomial for qi over Q (for any  i).. is a primitive element for Inv(H)/Q.1 are distinct. that z is is + . B E K [ X ] . Inv(H) =@(qi) for any i.B) = 1... and put y = A(z)/B(z). Put G = Aut(K(z)/K).62 and f I j=O fI j=1 f1 . gcd(A. It follows that 71. 1403 Let K be a field and z be an element of an extension of K such that 2 is transcendental over K . i.
Thus A . Since A(z) . ] Thus A .y B is irreducible in A K [ y ] [ X= K [ X ] [ y ] . then G is infinite and the fixed field of G is K . z is algebraic over K ( y ) . ad .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 ) .that y is transendental over K . B ) = 1 in K [ X ] K ( y ) [ X ] ) . Hence y is transendental over K . deg B}. Find the fixed field of H . Find the fixed field of H . polynomial of z over K ( y ) is the monic polynomial which is a multiple in K ( y ) of A . (a) Since gcd(A.y B ) = max{degA.where K* . If y is algebraic over K . that A . deg B } = 1.63 algebraic over K ( y ) . or if and only if y has the form az+ b/cz +d.y B ( z ) = 0. (c) By (b). [ K ( z ): K ( y ) ]= max{deg A.. z must be algebraic over K .bc # 0. . Again. y = A ( z ) / B ( z )is a generator of K ( z ) / K (i. (b) By (a). K ( y ) = K ( z ) )if and only if max{ deg A.deg B } .z)qZ+l.z)q+l/(z' . E K [ X ]such that A S + B T = 1 T in K [ X ] .y B is irreducible in K ( y ) [ X ] Thus the minimal .Suppose A . which is contrary to A B 6 K . ( CoZumbia) Solution. . I2 = {diag(a.y B = 0 in K ( y ) [ X ] . Conclude that n.12. which is contrary to the assumption.y B is irreducible in K ( y ) [ X ]and that . (b) Show that the elements of G are given by the fractional linear substitutions z +a z + b/cz d with a . (c) Show that when K is infinite. b. B ) = 1.q and the fixed field of G is n. then ordG = q 3 .bc # 0. it is easy to see that G N G L z ( K ) / K *. there exist S. (d) Show that when K = lFq and put + z (zq2 . d E K and ad .h(z).a) 1 a E K*}. since gcd(A. c.yB and (c [ K ( z ): K ( y ) ]= deg(A . Then It follows that degA = 0 and deg B = deg(A) = 0.y B # 0.e.
X)q2g. then by (a). Hence we have [Fq(z): Fq(z)] = q3 .Z)Q+1/(2~2)q2+1) 2. On the other hand. Then c and [F.X)+P Denote and (1 + x . (xq2 X)  1 (XP . we have Inv(H) = K .+ x ( n .4. Thus Inv(G) = Fq(z) and Fq(z) is a Galois extension of a given field L with FP C L C F. + X(Q1)Q)q+l B = (Xq.q . + X(PQ?)q+l n1 (XP . .q . since H is also infinite. Then gcd(A.(z) : Inv(G)] = q3 . we have (XP2 . = Hence F ( z ) c Inv(G). . it is routine to check g(z) = g((zq2 . (d) Suppose K = Fq. Thus we have Inv(G) = K . For any g :2 H az + b/cz + d .X)q2+1 = ((xqx))q+l. . K ( z ) / K ( y )is a finite dimensional simple extension. Obviously. (ad . which is contrary to the facts that G AutK(y)K(z) and G is infinite. .64 If K is infinite. On the other hand. the fixed subfield of G. Hence AutK(y)K(z) is finite. B ) = 1 and max{deg A.x)q+l/(XP .X)n”n A = (1 + XP’ + X(P1)2+ .bc # 0) in G. . K C Inv(G). Similarly. if there exists some y = f(z)/g(z)E Inv(G)\K. so is G.(z) if and only if L _> F ( z ) .1 1 2 + . deg B } = q3 .
(c) Show that EIF is a Galois extension with Galois group G. x + 1 = X2" .C"'y) (X. >rx D. F = C ( 2 ) with Z = Y" + Y". Let A = X .Y")(X" . Then z' E Inv(H) and by (a).. Hence c and T are automorphisms of E/C. G =< u . ord(7) = 2 and Tu = uT in G. Y is a root of X2" .l .2 . and = earriIn..C Y ) . Obviously. 1404 Let E = C(Y) with Y an indeterminate.ZX 1 since + + + x2n  zx + 1 = ( X . (b) Since X2" .z E Fq(z).Z X 1 is irreducible in C[Z][X].for example. Thus = Inv(H) = F.there are unique homomorphisms u and T of E to E over C such that . It is readily verified that X2" ZX+ 1 is irreducible over F (Z is transendental over C and X2" . the Dihedral group of "' ~ order 2n.(Y"+Y") + 1 = (X" . Since u ( Z ) = u(Yn + Y") = Z and T(Z) = 2. and that the subgroup G of Aut(E) which these generate is isomorphic to D. (c) Obviously.z).ZX I.. Since ord(u) = n. using 1403). [Fq(z) : Fq(z')] q.65 Similarly. (a) Show that there are unique automorphisms u and r of El@such that a ( Y ) = CY and T(Y) = Y .(Y) = CY and T ( Y ) Y . E is a splitting field of X2" . .(Y') * (X  y1yl) .Y)(X . (Columbia) Solution.Y") E F[X]. (b) Show that Y is a root of a polynomial of degree 2n with coefficients in < F. we have Z E Inv(G) and F = C(Z) Inv(G).l .y')(x . (a) Since Y is a generator of E over C. we have [Fq(z)Inv(H)] = q since IHI = q.the Dihedral group of order 2n.(z') = F q ( z q . E/Inv(G) is a Galois extension with Galois group G and [E : Inv(G)] = 2n.X : q and B = 1 in F. = (T" = 1 = r 2 .[X] and let z' = zq . ( X .
(b) Obviously U ( Z .e. (a) Since ( ~ ’ ( z ) = z. ?(z) = z and T ( T ( Z ) = = ( T T ( z ) .1) = 1 .66 in E . K C_ F is a Galois extension and [F : K ] = IGI = 4. It follows that (c) By the fundamental theorem of Galois theory. r 2 = 1 and UT = T U . there are exactly 3 subfields lying strictly between K and F which correspond to the three proper . the fixed field of G in F . Denote 7 = (z .. the quotient field of the polynomial ring&[z]). + f ( t )= t4 . Hence&(q) C F is a Galois extension and [ F : &(71)] = 4.x and ~ ( . Consider the elements (T. (c) How many subfields lie strictly between K and F ? How many of these are Galois over K ? Justify your answers. (b) If K is fixed field of G in F .we have (T’ = 1. Hence where K 4 is the Klein 4group. 2. Then&(q) 2 K F .7 in A u b ( F ) (i.qt2 + 1€&(V)[t]. It follows that E / F is Galois and [E : F ] = 2n.z and = 7(z) = A. Then f ( t ) is irreducible in & ( q ) [ t and F is a splitting field of f ( t ) since ] in F [ t ] . find a finite subset S of F such that K = &(S). 1405 Let F = &(z) be the field of rational polynomials in one variable x over & (i..1) = z It is easy to see that ( z .e. (Indiana) Solution. field automorphisms of F ) given by ( ~ ( z ) 2 .1)2 + E K .1 (a) Find the subgroup G (of Aub(F)) generated by (T and 7. Let (x11)2 5.1)’ &. Hence we have F = Inv(G) and E / F is Galois with Galois group G. On the other hand.
:) + & ( t 2 + $) of is irreducible over & (t2 $) and Q ( t ) is a splitting field over separable polynomial f(z). 1406 Consider &(t). u4 = 1 and isomorphic to the Dihedral group (b) Since D4.t ) ( z+ t ) (x z . CT3 ( t ) = and a4(t) = T 3 (E) 3 t. (a) Obviously. It follows that G is the fixed field of H . Thus all these 3 subfields are Galois over K.67 subgroups of G K4. field of quotients of the polynomial ring @[t]. It is easy to see that r2 = 1. (b) Let H be the subgroup of G generated by u2 and T . u 2 ( t ) u = = . u and the Let be elements of Autq(&(t)) given by u ( t )= and ~ ( t=) t. T 2 ( t ) = t . and obviously [&(t) : InvH] = (HI 4.) (.F ) 1 = ( . Since all the subgroups of K4 are normal. all these 3 subfields are normal over K (finite dimensional and separable). So + .f . = On the other hand. Find a E @(t) such that & ( a )is the fixed field of H (Justify your answer). (a) Identify G. Hence. T(T = C3T. (Indiana) Solution.t 2 ) (2 . = (2. Let G be the subgroup of Aub(&(t)) generated by u and T .
then f ( z ) has a factor with degree one in that is. b) Prove that K/& is Galois. Hint. Obviously. b) Let . a) Prove that f(x) is irreducible over &. It follows that that is. crPy = 1 and the discriminant A of f(2) is 34.[& + + + .3x 1.rbe the other two roots of f ( 2 ) = z3 . But fl are not roots of f(z). Since f(2) = x3 . the rational roots must be integral factors of 1. Hence /3 y = a and . Consider a .B. f ( ) = x3 . This is a contradiction. Thus f(x) is irreducible over &.68 is a Galois extension and [&(t) :& (t2 + $)I = 4. f(x) has a root in &. of.I. (Indiana) + Solution.32 1 is reducible over &.2. a) If f(z)= x3 .3a: 1 is monic. let K = & ( a )where a is a root . ' c) Find a generator of the Galois group Gal(K/&). InvH =& (t2+ $) 1407 Let & denote the field of rational numbers.32 + 1 in a splitting field of f(x) over K. cr + p + 7 = 0.
2 is a root of f ( x ) directly by using the hint). = a2 .. c) From a) and b). . 1 a E K } is linearly dependent over K .zn & + anlxnl + + a12 + ' .* a0 E K[z].(Y 2 and . Show that the Gorbit of f spans a finite dimensional Kmodule if and only if f E K[z]. Hence K / Q is Galois since Char(&) = 0. acting on K ( x ) by a E G sending z to z a. (a) Show that the rational functions f a = ( a E K ) are linearly independent over K. it is easy to see that the Galois group Gal(K/&) F A3. ( Y . . Assume that K is infinite.K and some distinct eleE ments a l .69 It follows that (Y. So K is a splitting field of f ( x ) .(Y = a2 . (a) Suppose that {fa = There exist some nonzero elements ( Y ~ . .. . Thus { f a = I a 6 K } is linearly independent. a. (T: + A K=&(a)+K ak+(Y22 is a generator of Gal(K/&). (Columbia) Solution. . . K [ z ]and K ( z ) have what dimensions? (c) Let G denote the additive group of K. (c> Suppose f ( x ) = a. . Let f E K ( z ) .2 are the roots of f(x). (b) As Kmodules. (b) dimK K[x] = dimK K ( z ) = 00. u2. 1408 Let K be a field and 3: an indeterminate. (We may check a2 . E K such that + n Hence and i#i which is not true.
f ( x a l ) .. a 2 .)a =P ~ ( Z + alli + /32(z + a2)a + .+1 .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) .'. Let f ( x + a l ) . a.. = P ~ ( Z + $ I+)P~ ( x + a 2 ) " z and also (Z +.. + P n + l f ( 2 + ~ 1 n + l ) .*f ..h ( z ) E K [ z ]and ( g ( z ) h ( z ) )= 1..*. there exist PI. . .+Pn+l(~+an+l)~ + . 1 ant1 a1 . f ( x an+l)spans the subspace generated by the Gorbit of f . . For any a E G. +an)(ai E K ) span the subspace < {f(x+u) I u E K } >. .70 We claim that for any n 1 distinct elements a l .f ( z a2). . . On the other hand. ./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 . . " . Then for any a E K . . ..01. Hence = P l f ( z + a l )+ P a f ( z + a 2 ) + . . a. .. . and the Gorbit o f f .f ( z + a 2 ) .+a) This shows that { f(x + a i ) 1 1 5 i 5 n + 1) spans the subspace generated by the Gorbit of f . . .f(x . ./32. . . . suppose f ( z ) E K ( z ) ... + Pn+l(z + Qn+1li for 1 5 i < n. . .. a. { f ( z + a ) 1 a E K } spans a finite dimensional Kmodule. a. a n + l in K (note that K is infinite). we take (. . We write f(x) = g ( z ) / h ( x ) where g ( z ) .P 2 . f(. .+1 a: a..
Let a and a’ be roots of e(z) and e‘(z) in some extension field of E .a) for any a E F . Hence we have an isomorphism 7 : F ( a ) + F(a’) which sends a to a (u E F ) and CY to a’. q(P) is a root of q ( q ( z ) ) = g(z). which is irreducible over F . 7 : E ( a ) + E(a’) is an isomorphism and ~ ( a )a’. but :E E is in G.Let e(z) be a monic irreducible factor of f(z) over E. we have h(z +. Hence 7 : F ( a ) + F ( a ’ ) can be extended to an isomorphism 7 of E(a) onto E(a’). + . Note that y ( P ) may not be P. Since E / F is Galois. Obviously E is a splitting field of g(z). where . for any u E G. = we must have (vlE)(e(z)) = e’(z) and dege(z) = dege’(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 ] . Since f ( z ) E F [ z ] . CY and a’ are roots of f(z).h ( z + u)) = 1.71 Since (g(z + a ) . an easy discussion in a splitting field of h ( z ) over F will lead to degh(z) = 0. u(e(z)) is an irreducible factor of f ( z ) over E . we can write E = F ( P ) . 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. is a root of g z . Thus all the irreducible factors of f ( z ) over E are of the same degree. B () a separable irreducible polynomial over F . ( c ozum bia) Solution. Since K is infinite. It follows that Now. Then. For any (T E G = Gal(E/F). Then. a(f(z)) = f(z).)I i=l n n h ( z +(. since a and a’ are roots of the monic irreducible polynomial e(z) and e’(z) over E respectively. We prove in the following that all monic irreducible factors of f ( z ) over E arise in this way. we still denote u to be the isomorphism E[z] t E[z] which extends (T on E and maps x to z.Show that all the irreducible factors of f ( z ) over E are of the same degree. Suppose e’(z) is another monk irreducible factor over E.
the prime field of K .1.1)' = 62t6' = 2t61 # 0. Comparing the coefficients of the term of degree n . Hint. Show that the polynomial tP . in E K . (Indiana) S o htion. (b) Let F be the splitting field of the polynomial t62 .u . Hint.c and let u E E be a root of this polynomial. u E K .t . (b) Let G be all the zeroes of t62 . n w p = 1.1 over Zg. G is a subgroup of F" and G is cyclic of order 6 2 since (t6' It follows that .1 form a cyclic group G of order 62.u . If u is a root of t P . . + + + + in K [ t ] . Then for any m EZ. u)E K . 1 5 n 5 p . So we have n .u . Thus we have tptc= (tum) +. Since p . Let E be a splitting field of tP . Show that [ F :&I = 3.1 in F . g ( t ) in K [ t ]where f ( t ) is a monic polynomial of degree n.il)(t .in).t c = f ( t ) .i2) * * * (t . [ F :as]2 3.1.c in K [ t ]is either irreducible or splits completely into p linear factors over K. Obviously. u il + i2 .c then so is u+ 1. . First prove that the zeroes of t62 . n wp) u = v(n . u = 0 and there exist integers v and w such that w .72 1410 (a) Let K be a field of characteristic p > 0. we obtain n . u = (v . f ( t ) = (t .t . Suppose that t p .t ..
prove that it is a power of p . when decomposed in nZ[a:]. Let GI be its unique subgroup of order 62. that is. and so K' = 1. The set {a:" J a: E K } is a subfield of K that is denoted by KP (no proof required). Show that every finite field extension of L must in fact have degree equal to a power of 2. (b) Let L be the field of real algebraic numbers. If [L : KP] is finite. we must have m = 1. = 1411 (a) Suppose you are given a field L .1.1 = 124. Then E* is a cyclic group of order 53 . & C_ L a'. is a product of irreducible polynomials of degree 2. such that L/& is algebraic and every finite field extension K / L . For this purpose. By (a). If K' is the corresponding subfield of K I L . m where m is odd. Hence. By Sylow's Theorem.1 splits in E and E is a splitting field of t62 . [ K : L] = [ L ( a ): L] = d e g f ( s ) is even. (a) Let K / L ( K C a ' ) be a finite field extension. then [ K : K'] = 2" and [K' : L] = m.1. K 5 a' is of even degree. Since L has no proper odd dimensional extension field. G has a subgroup H of order 2n. Then & 2 L & ( and L/& is algebraic. X Now for any finite field extension K I L . Then all the elements of G' satisfies t6' . it is in fact a power of 2. Thus we have E N F and [F :as] 3. We have to show [ K : L] is a power of 2. there exists some element a: E K such that K = L ( o ) by Primitive Element Theorem. f ( z ) . So t6' . . Let f(a:) be the minimal polynomial of o over L. (Indiana) Solution. and [K : L] = 2n. (b) Show that such a field L actually exists. K a'. (a) Let L be an intermediate field between Kp and K . we may assume that K is Galois over L. So. the subfield of 1R of numbers which are algebraic over &. Let G = GalK/L and JGJ= 2n .73 Let E be a extension field of Z5 such that [E : Z5]= 3. Then f(a:) has no real root since f(a:) is irreducible in L [ z ] . 1412 Let K be a field of characteristic p # 0.
.. . For example.. we can assume that bi $ KP(bl. . (c) Prove that the set B of part (b) satisfies K p ( B ) = K .. : It follows that B is pindependent. b. b 2 . By Zorn’s Lemma... (Indiana) Solution.b. bi1) = KP when i = 1).. * . b2. and bi @ KP(bl. Now suppose that {Bi I i E I} is a chain of pindependent subsets of K . Since Bi is apindependent subset. * * .b. .. b m } C Bi.. tP .l)] ) i=l = p”. Without loss of generality.. Then for any finite set { b l ... . . .. . .. .bn} such that = KP(b1. Hence [KP(bl.bil). [ K P ( b l .. of distinct elements of B ..74 (b) A subset B of K is called pindependent if for any finite set bl.*.b. K contains a maximal pindependent subset.. . . Let B = U Bi. .) : L] n = n [ K P ( b l . . ..+ . (a) If [L : KP] is finite.b2. if 6 E K\Kp. .bp is irreducible in KP(b1. (b) If KP # K . b 2 . B = { b } is a pindependent subset of K . .. b 2 . b m )K P ) =pm.bzl)] = p... . It follows that [L: KP] 1 [KP(bl.bil)[t]..... there exists a finite set of elements {bl. . .bi1) for any ! 1 5 i 5 n (KP(b1. i€I . .bn) L. Prove that if KP # K . b i : KP(bl.} of distinct elements of B . Since .. there exist pindependent subsets. such that { b l . bz) : KP(b1. there exists some i.b2. then K contains a maximal pindependent subset B.
.. . 1414 Let K I F be a finite extension of fields and let L and E be intermediate fields. Then IK(a)l = prm and [ K ( a ): K ]= m. show that f divides t P r n . that is. .) * 1 .b.t ) ..p. with E / F Galois and [K : L] = p . if b @ { b l . . . a prime. we have f ( t ) I ( t P r n . have [KP(bl. then and [KP(bl.) .... .75 (c) Let B be a maximal pindependent subset of K . Since m I n. b 2 .b. Suppose K p ( B ) c K . If m is an integer which divides n and f(t) E K [ t ]is an irreducible polynomial of degree m.}. [KP(bl.. Then (El = prnand [E : K ] = n.l) : KP] Since B C B U { b } .b.. Let b E K\Kp(B).. The reason is that. bml)] = [KP(bl.* . (Indiana) .l} say.b. in B U { b } . bm) : KP] = p m . (Indiana) Solution..* and if b E {bl.t over K .*"l = p"* . . . 1413 Let K be a finite field with pr elements ( p a prime) and n be a positive integer. f ( t ) is the minimal polynomial of p. . Thus we have K P ( B ) = K . Prove that if p does not divide [E : F ] then E L. E contains a subfield L such that K & L E and L N K ( a ) . .. Since pp'" .t . the independency of B U { b } contradicts the maximality of B . . b..p = 0. . . then {bl.b. b 2 .* ... . Then B U { b } is pindependent. b m } . b = b.bz. CB and we still : KP] : KP(b1. Let (Y be a root of the irreducible polynomial f ( t ) in some extension field of K. Let E be a splitting field of t p r n . for any distinct elements b l .. Hence there exists an element p E L C_ E such that f(P) = 0.
It is clear that i=l u Ki is a subfield of a'. L and n E)I [K : L] = p . And further. a p = IGal(E. It follows that C E . Let f(z)be a separable irreducible polynomial over F such that E is its splitting field. Then E . we have N_ G a l ( E . Then E L = L(E) is a splitting field of f ( z ) over L. Hence E L is Galois over L. Let a E E be a root of f(z). Prove that deg(f) 2 5. L = K . 1415 Let Ki be the subfields of a' defined as follows: KO = &. L = L(E) = L(a). If i the smallest subfield of @ containing the set 2 0. contrary to the assumption. Let i=O (1) Prove K is a field. L/L) Gal(E/L n E ) .76 Solution. L. (1) Since KO C_ K1 C_ C_ Ki C_ Ki+l C_ M * * a is a chain of subfields of a'. Then E = F ( a ) and E . L/L)I = IGal(E/L dividers IGal(E/F)I = [E : F ] . For any c E G a l ( E . Let E * L = E(L) L ( E ) be the composite of E and L in K . + it is clear that g (E ~ Aut(E/L n E ) . Thus we have E C L. (Indiana) Solution. Ki+l is (0 €6' 10" E Ki for some n > 0). (2) Let f(z) E K [ z ]be irreducible. L/L) = Gal(L(a)/L). since L Now suppose E prime. .
Let n = [ F p ( a ) Fp]. if aPml I . Hence deg f ( z ) 2 5. 1416 Let K be an extension field of Fp.(u) : F.]([EpP]= m.Then IFp(a)I = p" and upn' = 1. Prove that [Fp(a) Fp] is the smaIlest positive integer : m such that E Fp. 1417 Let F 2 K be a field extension of finite degree m. P. If m and n are coprime then show that f remains irreducible in F [t]. : Thus [Fp(a) F] is the smallest positive integer m such that : E Fp. zPm  so u is a root of m zp . Then [E : Fp] = m and Fp n = [F. the field with p elements.hence in K [ z ] . Let E be a splitting field of F p ( a )C E . Contradicts the irreducibility of f(z). Let E = F(cr). Let f(x) E K [ z ] be an irreducible polynomial with d e g f ( z ) > 1. of . where g ( n ) = p"l.1 (Indiana) Solution.x. 1 5 k < n. Hence z over Fp and a E E . for some n > O}. Let a be an algebraic element in K .(&4)Pl = 1. then On the other hand.h ( t in F[t] and ) where g ( t ) is a irreducible polynomial in F[t] degree k. f(z)splits in Ki+3[z]. E Fp. Suppose d e g f ( z ) 5 4. Suppose that f(t) is reducible in F[t] let f ( t ) = g ( t ) . (Indiana) Solution. Let f E K [ t ]be an irreducible polynomial of degree n. a d n ) E Fp. There exsits some i such that f(x) E Ki[x]. Since ( ~ ( g ( ~ ) ) P = : l apnl = 1.77 (2) (Remark: d e g f ( z ) may be 1). for some positive integer m. cubic and quartic equations and Ki+l 1 (0 E(C 10" E K . By the formulas for the roots of quadratic.
which contradicts ( m . 1419 Find a Galois extension E over & with Gal(E/&) cyclic of order 32. Then [E : F ] = Ic since F [ t ] .Zz @ . It is easy to see that IGal(&(td)/&)I = 4(n) and Gal(&(z. [E : K ] = [E : K(cr)]. [K(a!) K ]= [E : K(cr)]. then & C E is a Galois extension and Gal(E/&) 1 . (Stanford) Solution. . since cr is a root of f ( t ) and f ( t ) is irreducible in K [ t ] . (Stanford) Solution. Aut(G) is cyclic of order n . . .z .Z2. As in 1418.)/&) E Aut( ) where G is the cyclic group of order n. for any positive integer n. E = & ( z 1 7 ) . 4(n) is the Euler $fur Ition.m.1. the cyclotomic field & ( z n ) over & is a Galois extension and [ & ( z n ) : &] = d(n) where z . It is easy to see that IGal(&(zi)/&)I = +(n) and Gal(&(z. Thus f ( t ) is irreducible in F [ t ] . the cyclotomic field &(zn) over Q is a Galois extension and [&(zn): Q]= +(n) where z.. is an nth primitive root of the unit.So [E : K ] = [E : F ] [ F : K ] = Ic . For any positive integer n. then & E is a cyclic extension of order 32. : n. When n = 2" and m 2 3. if we take E = Inv(&).m . On the other hand. Z p . By the Fundamental Theorem of Galois Theory.78 g ( t ) is irreducible in where cr is a root of g ( t ) in some extension field of F . then E is a Galois extension of & with Gal(E/&) cyclic of order 16. : Taking m = 7.n ) = 1 and 1 5 Ic < n. 1418 Find a Galois extension E over & with Gal(E/&) cyclic of order 16. So if we take n = 17. is an nth primitive root of the unit. it is well known that Aut(G) . When n is prime.)/&) N Aut(G) where G is the cyclic group of order n. d(n)is the Euler &function.It follows that n I Ic .
Show that its trace is given by Trp(M. is similar to diag{ul(z). a has the form since { l .z). Let z be a primitive element of E / F . ~ . Z ( z ) .z n ...a.(z) = a z . Since f(z)has distinct roots C T ~ ( Z )g.@) in E. M .l } is a base for E / F . f(z) is also the minimal polynomial of the Flinear map M . . . 3c.. (z) = z . . the matrix of M .. Anyway. Consider the Flinear map M a : E + E..say... we have  n i=l . for any a E E .79 1420 Let E / F be a finite Galois extension.+ . Obviously. g n ( ( z )are distinct. To prove that it suffices to prove that for any 1 5 k 5 n ..znl}. . : E + E . is the minimal polynomial of z over F. Now. relative to the base (1.an(z)} in Mn(E).uz(..) = C g ( a ) where varies over G. M . (Columbia) Solution. z .1. G = Gal(E/F) and a E E . [E : F ] = n and Then . . . and u 1 ( z ) . (E M n ( F ) ) .
1. .and for any 1 5 i 5 n . n This completes the proof.
Part I1 Topology .
.
Since A is connected. and that N is an open neighborhood of AxBcXxY. there exists a neighborhood of zo. It means that A U B is connected. such that A n B # 8.l). If A and B are path connected. 2102 Suppose that A and B are compact subspaces of spaces X and Y respectively. Let zo E AnB.O)) C R2 and 1 B = {(z. we may assume ~ that A c f'(l). f l is also constant. The following example shows that if A and B are path connected then A u B needs not to be path connected. Hence f is not surjective. need A U B be path connected? (Indiana) Solution. Since f is continuous. Without loss of generality. say U . f ( must be constant. but A U B is not path connected. Prove that A U B is connected. But since zo E B. Prove that there are open sets U c X and V c Y such that . Therefore we have B c f'(l). ~ We have f(zo) = 1. 3 : Then A and B are path connected and A n B = A .sin ) I 0 < z 5 1). Let A = ((0. By the same reason.a3 SECTION 1 PO IN T S ET TOPOLOGY 2101 Let A and B be connected subspaces of a topological space X. Let f be any continuous map from A U B to So = {1. such that U c f'(l). there is a point of B which belongs to U .
i=l P Let U ( z ) = fi Uyi(z) V(z) = 0 V. and I) n It is obvious that U ( z ) and V(z) are open i=l sets of X and Y respectively and that {z} x B i=l c U ( z )x V(z) cN. . ' Then by the Urysohn Lemma there exists a continuous function X + [0. . .since (z. c On the other hand.(z) x V. Let X* = X U {m}. We denote by X" the onepoint compactification of X. . y) belongs to A x B and N is an open neighborhood of A x B c X x Y . For any y E B. p } such that B = c U V. with A compact and B closed. (Indiana) Solution. z E A } is an open cover of A . .84 Ax B cU x V cN. . there exists asubcover { U ( z i ) . ( z ) c X and V. Y )E V.il . Let' z be a point of A ..(z).(z). Hence we may assume that X is noncompact. Therefore the family of open sets {Vy(z). q } such that A Let U = sets of j=1 u U(zj). Let A and B be disjoint subsets of X .1] such that f l = 0 and f l = l? ~ ~ + ( Cincinnati) Solution. j = l . 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 .(4 cN.. . Since X is locally compact and Hausdorff. X is compact ' and Hausdorff.. and consequently is also a normal space. { U ( z ) . Since B is compact. Does there exist a continuous function f : X [0. If X is compact. there is asubcover {VYi(z). 2103 j= 1 Let X be a locally compact Hausdorff space.1] ' F: . 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) c Y such that ( Z C . there exist open sets U . U U U ( z j ) and V = j=1 q n V(zj). Since A is also compact.y E B } covers B.
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. 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.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. (a) So * So and S * So are homeomorphic to familiar spaces. and S’ * So is homeomorphic to the unit sphere S2. (c) Show that i f f satisfies the condition of (b) then f is a quotient map. What space ’ are they? (b) Describe X * So for a general space X . (Indiana) Solution.satisfies = 0 and f l = 1. the restriction o f f on X . (a) Let X be a topological space and be an equivalence relation on X. Define by X / the space of equivalence classes under .t’) if and only if { x = x’ and t = t’ = 0 or y = y’ and t = t’ = 1. Y is Hausdorff and f : X + Y is continuous and onto. y. (Indiana) Solution.y. ~ ~ 2104  No proofs. (b) Generally. If X and Y are topological spaces. (a) So * So is homeomorphic to the unit circle S’.  2105 (a) Define quotient map. Therefore. are required for this problem. only the correct answers to the question asked. X * So is homeomorphic to the quotient space X x [0. f.t) is equivalent to (d. (b) Show that if X is compact.
(a) Prove that if X is compact then there is an open neighborhood V of C in Y such that f'(V) c U . there is naturally associated an equivalence relation on X such that z1 5 2 if and only if f(x1) = f(z2). It follows from the continuity of f that f ( A ) is a compact subset of Y . then W is closed in X . More generally. (b) Give a counterexample to show that if X is not compact. Since Y is Hausdorff. (Indiana) Solution. By the result of (b). we see that f f'(v) = x . : X + Y is the continuous function defined f by j ( t ) = earit for t E R.86 class containing x.f ( W ) . W is compact too.U . Under this topology. Since f is a continuous function. A is compact too. Since W c f'(f(W)). (c) Let be the equivalence relation on X associated to the map f. r becomes a continuous map. if f : X + Y is a continuous map. Let V = Y . and let U be an open neighborhood of f'(C) in X.' ( U ) is open in X . the quotient space X / is compact. f ( A ) is closed in Y .  (b) Let X = R and Y = S1. If Y is homeomorphic to X / under the map i : X/ + Y and f = i o T then we call f a quotient map. f ( W ) is a compact set of Y . Since X is compact. The quotient space X / may be topologized by defining a subset U c X/ to be open if and only if r . and consequently is a closed set of Y because Y is a Hausdorff space. Then V is an open neighborhood of C in Y . Let C be a closed subspace of Y .1 map. (b) Let A be a closed subset of X . i is obviously a 1 . Since f is onto. therefore. and. Since X is compact.fl(f(w))c x w = u. (a) Let W = X . So f is a quotient map. Then we define a map i : X/ 4 Y by i([x]) = f ( x ) .       N 2106 Let f : X Y be a continuous function from a space X to a Hausdorff space Y .1 map from the compact space X / to the Hausdorff space Y . Denote by [z] the equivalence class containing 2. Take C = (1) E S1. It is clear that f = i o T . . Hence i is a continuous 1 . Hence f is a closed map. then there need not be such a neighborhood V . is obvious that It fyc)= {n I 72 EX}. is a homeomorphism.
Thus we see that R ( W ~ ) and R ( W ~ ) two disjoint open sets in Y and contain U and V respectively.=(+n. Let X be a connected space. (a) Let R : X Y be the identification map and yo E Y be the point which A collapses to. let f : X +B be a continuous map. 2107 Let X be a normal topological space and A c X a closed subspace. 2108 Let p : E B be a covering map with E locally path connected and simply connected. Then R'(U) and R . are If yo E U (or V). 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.' ( V ) are two nonempty disjoint closed sets in X . Then the quotient space Y obtained by collapsing A to a point yo is not regular. then. n  one can easily prove that there does not exist such a neighborhood V. 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. If yo $ U U V. (a) Show that the quotient space Y obtained by collapsing A t o a point is normal.87 Let U = U Un be the open neighborhood of f .' ( C ) n €I in X . by the normality of X . (b) Does this result hold if normality is replaced with regularity? (Indiana) Solution. we only need to take the sets W1 and Wz without the restrictions that Wi n A = 0. (b) Let X be a regular space which is not a normal space. Let U and V be two nonernpty closed sets in Y such that U n V = 0. where 1 1 U.n+). f . It means that there exist two disjoint closed sets A and B in X such that they cannot be separated by disjoint open sets in X . n n Since nw lim lUnl = 2 = 0.
Let A={ E . g f l ( z 0 ) = f z ( 2 0 ) . we see that A = X. Since p is an open map.X E A} is an open covering of X .' ( W ( z ) ) such that Zi E w.(z). Therefore.' ( W ( z ) ) is mapped topologically onto W ( z )by p . Since p f 1 = p f 2 = f . 2109 Let p : X +X be an nsheeted covering projection. 2 . For any point z E X. n .. Prove that there is a deck transformation such that fi = g f 1 . . Let f2(20) = e l . It is not difficult to prove that A is bothopenandclosed in X. Suppose that U = {UA. i=l Choose another elementary neighborhood V(z) of z such that v ( ~cc ) np(w). W(x) is an pathconnected open neighborhood of x such that each path component of p .' ( z ) = ( 1 . Let c ( z )be the path component of %. we see that e l E pl(b0). Take a point zo E X and let f l ( z 0 ) = eo E E . Choose ix is a U .. Thus there exists a deck transformation g such that g(e0) = e l . Suppose that (In dian a) Solution. let p . By the assumptions p : E + B is the universal covering map. 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 ) } . E Ui.l ( b o ) .(z) n U i containing Zi. n X is compact. which means fi = g f l : X + E .  < 00. f2 : X + g :E +E E be two lifts of f . Prove that 2 is compact. It is obvious that A is not empty. E U such that Z . Let  where w ( )a path component of p . Let W ( z )be an elementary neighborhood. . i. (Indiana) Solution. by the connectedness of X . where bo = p ( e 0 ) .88 and f l .e. Z n } . Thus. Then eo E p . i=l fi p ( c ( z ) )is an open neighborhood of z in X .
. (Recall that 7 c U means that every set in the topology 7 is contained in U. Suppose that C n BdA = 0. nIt follows that p'(V(z)) c .. U which means U = 1.can be covered by a finite number of V(zi). 7 ) and ( X . We use the reductio ad absurdum.A . 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. 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 . U ) to the Hausdorff space (X. ' T ) is the identity map of X. Since 7 c U . Let U be any open neighborhood of 3:o in ( X .U) ( X . . Then it is easy to see is T(z)c E(z) c Ui for i = l .m. . We use the reduction to absurdity. 6 Ui. 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.) (Indiana) Solution. Let ( X . For any point 20 E X .. X i = 1.. i=l c(z) the path component ofpl(V(z)) for any i. It is obvious that h(z0) = zo and h ( U ) = U . and consequently X can be covered by a finite subcover of U. Suppose that 7 c U. then C intersects BdA. ) (Indiana) Solution. Take U = C n x and V = C n ( X . U ) . Since C n A c U and C n ( X . (Recall that BdA = 2n (X. We claim that h is a 7).A ) .89 Let P'(v(zc>) = where that n (JW). Thus h is a homeomorphism from (X. 2111 Let X be a topological space and let A c X.)to (X. 7 ) . I). continuous map. Show that if C is a connected subset of X that intersects both A and X . That is.1 map from + the compact space ( X .A ) . h : (X. This contradicts the assumption.+ . . Since X is compact. Then h is a 1 . Prove that 7 p U.A ) c V . Hence h is continuous at 2 0 .
(b) Let zo be any cluster point of C e ( A ) . Then there exists a sequence {z. a. d ) be a metric space. So C . (a) Let 20 be any point of O . there exists a point a E A such that d(z0. ( A ) . d ( z . I g) . (a) Prove that O . let O. ( A ) . Since A is compact. For any subspace A E cX and real number > 0.e.< ( E . a ) ~ d ( z . ( A ) . Thus we have + which means 20 E C . A which is a contradiction t o the connectedness of C. 6 = E . ( A ) such that z # zo for any n and 2. z o as n + 00. a ) > 0. a ) < E for some a E A } . 2112 Let ( X . Then C a ( A )= [0. that both U and V are open sets of C.} in C . for .from the assumption. But u nv = c n A n ( X  A ) = C n B ~ = 0. If A is not compact.) 5 E. i.. ( A ) . we may assume that a. Take X = [0. A : d(z.d ) < E . for any 2 E 06/4(z0). .. It is clear that C = U U V and both U and V are closed subsets of C . ( A ) is closed in X . + 4 6 which means that 06/4(zo) c O .d(z0. Must C ( A ) be closed . E A such that d(z. + a for some a E A . (b) If A is compact. u ) < E . . 1). z o ) + d ( z o . without loss of generality. and.(A) = C. each z there exists an a. a ). is not closed in X .2] c R and A = [0.(A) = {z E {z E A : d ( z . show that C ( A ) is closed in X . consequently. Then. for a general subspace A of X ? (Indiana) Solution. we give a counterexample as follows. Thus O . By the definition of C E ( A ) . ( A ) is an open subspace of X . we see that both U and V are nonempty subset of C.a)5 E for some a E A } . ( A ) is an open subspace of X . By the definition of O .
l I v ( N ) are also open in X.0).91 2113 Suppose that X is a dense subspace of a topological space Y. then Y is connected. The topology on Y is determined by the family of open sets 7 = { { a . X .U ) would be an open set of X . c ) . would be a nonempty open set of X.' l v ( N ) are open in. If Y were not connected.Prove or give counterexamples to the following assertions: (a) If X is Hausdorff. el. Let A = X n U . f . a) If U and V are open in X . a) Let N be an open set of Y. Let Y = { a . C l . On the other hand.A = X n (Y. therefore. show that f is continuous. f . Thus fF1(N) = flIv(N) uf'Iv(N) . It means A # X . because Y . a) This assertion is not correct. We give a proof to it as follows.' I u ( N ) and f . But U and V are open in X . A X . { b . and f : X + Y be a function so that flu and flv are continuous.U is open in Y . (Indiana) Solution. respectively. 2114 Let X and Y be topological spaces. (b) If X is connected. We give a counterexample as follows. which is bothopenandclosed in X. c}. Since X is dense in Y and Y . But it is easy to see that the subspace { c } is dense in Y and is Hausdorff. U and V . b) Give an example where U and V are not open in X and f is not continuous. { c } .b. X = U U V . This contracts the connectedness of X . Since X is dense in Y and U is open in Y . (Indiana) Solution. b.U is open in Y . Y is not a Hausdorff space.Since flu and fv are continuous. then there would exist a nonempty proper subset U of Y which is bothopenandclosed in Y. Since any neighborhood of the point a always contains the point c.' I u ( N ) and f . Thus A would be a nonempty subset of X . b) This assertion is true. then Y is Hausdorff.A is nonempty. { a .
21. which contradicts the connectedness of q'y. Therefore. and : X + Y be the map defined by q ( x . f : X +R is defined by f(x) = {x 2 XEU.y) E R2 11 5 y 5 2.l ( y ) is connected. Thus. () Since q is a quotient mapping. 1) and V = [l.92 is open in X .q . XEV. But X is not connected. O ) E U and q ( 1 . let y E q ( U ) n q ( V ) . ( Coturnbia) Solution. a) Show that X is connected. b) Let X = [0.which contradicts the connectedness of Y . y ) = 1 for (1. a) Suppose that X is not connected. b) Is X necessarily connected if the map q is not assumed to be a quotient mapping? Justify your assertion. Let X = U U V . q .Y Y ) = (ql(y) n U ) u (ql(y) n V ) . Thus X must be connected. U = [0. and consequently. . = For otherwise.l ( q ( V ) )and U = ql(q(U)). U and V such that X = UUV. b) The following example shows that the assumption that q is a quotient mapping is necessary for X to be connected. q ( U ) and q ( V ) are disjoint nonempty open subsets such that Y = q ( U ) U q ( V ) . V = q . Then q ( U ) n q ( V ) 0. O ) E R2 10 5 and 2 5 1) v = ((1.y) E V . where U = { ( x . 2115 Let q : X t Y be a quotient space projection from a topological space X t o a connected topological space Y. f is continuous. the unit interval of R. but f is not continuous. 0) = x for ( x . there exist two disjoint nonempty open subsets of X . Assume that ql(y) is connected for each y EY. Then f (u and flv are continuous. For each y E Y. ) The topology of X is induced from the topology of R2. Then q is continuous but is not a quotient mapping.1]. Let Y = [0. and it is obvious that ql(y) n U and ql(y) n V are both nonempty open subsets of ql(y).2].
' . ( 2 0 .(ZO) and (Wzo(y)x V.p ( U ) is an open set of X . Let p : X x Y + X be the usual projection onto the first factor. Thus we have 2 c f'(l). for any y E Y . without loss of generality. Since Y is connected. Since Y is compact. But c as wellknown. Let U c X x Y be a closed subset and zo E X . According to the assumptions. ( Cincinnati) Solution. Show that p is a closed map. we see that W(z0) n p ( U ) = 0. Since U is closed... .e.(ZO) of Y such that ( Z O . and let 2 be a set such that Y is a subset of 2 and 2 is a subset of the closure of Y . (Y)z= Y n 2 = 2. Y c f'(l).(y) of X and open set V. there must exist a finite number of V. 2117 Let Y be a connected subset of the topological space. where So = {1. Thus X . Y ) E Wz0(y) x V. (zo).(~~). Prove that 2 is connected.y) 6 U .p ( U ) .p ( U). = ) i=l n n Then W(z0) is an open neighborhood of z o in X . Taking closures of these two sets with respect to 2 and noting that f'(l) is a closed subset of 2. p ( V ) is closed in X . we get (Y)z f'(l).1} is the Odimensional sphere with discrete topology.. W(z0) C X . i. and it follows that f is not surjective. and consequently.i.(zo))nU = 0. we have Y c 2 c F. Let f : 2 + So be any continuous map.e.VYn (zo) such that n y = Let i=l u VYi(XO). Since (W(z0)x Y )n U = 0. It means that there does not exist any continuous surjective map from 2 to So and therefore 2 is connected.93 2116 Let X and Y be topological spaces. which means that p is a closed map. with Y compact. there exist an open set Wz. we may assume that f ( Y ) = {l}. . (Minnesotu) Solution. w ( ~ ~w~.Then.
( Cincinnati) Solution. a) Prove a metric space X has property S if X has a dense subset with property S. Then. we may assume that A C V. a bothopenandclosed subset of X\U. then U U V is connected. Hence W1 is a nonempty bothopenandclosed subset of (U U V )U (X\U) = X . without loss of generality.we see that C U U is a connected subset of X\A containing C . By the above fact. Since C U U c X\A and CnU = 0. we may assume that U c W. We first prove that if X is a connected space. Thus W1 is a bothopenandclosed subset of V . Hence X\C must be connected. there is a cover {Val& E I?} of A by connected sets such that each V. and if V is a bothopenandclosed subset with respect to X\U. consequently. a) Suppose that X has a dense subset A which has property S. is connected. Since U is connected. which contradicts the assumption that X is connected. Now suppose that X\C is not connected. for any E > 0.Since each V. Since A c X\C and A is connected. Must X have property S? ( Cincinnati) Solution. 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. Then let U U V = W1 U W2 where W1 and W2 are two disjoint nonempty bothopenandclosed subsets of U U V.. Suppose that U U V is not connected. show that X\C is connected. U is a connected subset of X . is also connected va .94 2118 Let A be a connected subspace of a connected set X . 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 . C u U is connected because C is connected. Suppose the closure of X has property S. b) Suppose X is a subset of a metric space. has diameter < ~ / 2 . and. which contradicts that C is a component of X\A. If C is a component of X\A.
Show that either f ( X ) = X or else there exists 6  > 0 such that 3 3 < y 5 . c R 2 . s i n r / z ) 10 < z and 5 l} 6 = inf{z 1 (z7sin. Fig.y) 11 5 y u{(z12) 5 2) I 5 5 '1 {(llY)7O 5 y 5 2. Let R denote the euclidean real line. ' Thus X has property S. Since = X and = U Val we see that aEr {ValQ: E I} is a cover of X by connected sets each of which has diameter < E . 22 (Toronto) Solution.rr/z) E A } .} = 0. Then = R has property S. b) We give a counterexample as follows. .2. x 2120 Let X be the topologist's sine curve defined by X = {(z.lr/z) 10 < z 5 1) u ((0. X be the set of all rational numbers. (i) Sketch X . but it is easy to see that X does not have property S.sin.1 (ii) Let A = f ( X ) n { ( z .95 and obviously has diameter < E. (ii) Let f : X + X be continuous. (i) X is shown in the Figure below.
{ u U b } . Hence in this case there exists a 60. we claim that f ( X ) = X .11. we first note that since X is path connected. we see that f ( X ) = X. n a Fig. 3 3 < y 5 } = 0. such that Therefore it is easy to prove that the set {(O. 2 2 2121 Let T = S1x S1denote the torus. 0 < 60 5 1.I and U3 = X .) Then U1. If 6 > 0. using the fact that f ( X ) is path connected. .Y) n 10 < 2 < 6.96 I f 6 = 0. UZ = X . (i) Show that T can be covered by 3 contractible open subsets. (See the Figure above. then it is obvious that f(x) {(z. ( Toronto) Solution. 2.2 Thus let U1 = X . U2 and U3 are contractible open subsets and T = U ~ u U . (ii) Show that T cannot be covered by 2 contractible open subsets. f ( X ) is also path connected. To prove it. U U 3 . (i) It is wellknown that the torus T can be identified to the quotient space of a square X obtained by identifying opposite sides of the square X according to the directions indicated by the arrows as shown in the Figure below.Y) I 1 IY I 1) c f ( X ) * Once again.
. since U is locally finite.particularly. It is clear that V is an open neighborhood of p such that V n ( up) = 0. 2122 Let U = (U. ( Toronto) Solution. U PEK r p is closed. Let U = { f . . a) By definition. Prove that there exists a weakest topology in S amongst all those for which all members of F are continuous. Show further that the resulting topological space satisfies the Hausdorff separation axiom. for any subset K c J . there is an open neighborhood V . n V.}aE~ an open cover of the space X. for each pi. . 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. .97 (ii) Suppose that T could be covered by 2 contractible open subsets. . b) If U is locally finite show that. Then there exists a unique topology 7 on the set S of which U is the topology subbase. if each point p E X has a neighborhood which intersects only a finite number of U..(T) x Z @Z. Suppose that s1 and s2 are . U is said to be locally finite. Up. From the Van Kampen theorem it would follows that the fundamental group ?rl(T) = 0.’ ( ( a . Hence is closed. of p such that V . Then let V = W n V1 n .b) is any open interval of R and f E F } . (Hama4 Solution. be a) Give the definition for “U is locally finite”.. say. there is an open u PEK neighborhood W of p which intersects only a finite number of sets in U . Up. Since p $! r p . It contradicts the known fact that a. b ) ) 1 ( a . only a finite number of Up for /3 E K . . It is easy to see that 7 is the weakest topology on S amongst all those for which all members of F are continuous. b) For any point p $! vp. . = 0..
By the assumption there exists a t least a n f E F such that f(s1) # ~ ( s z ) . bl)) and Uz = f . bz) = 0.bi) for i = 1 . So (S.98 two distinct points of S.bl) and ( a z .b z ) ) are two disjoint open sets of (S. b z ) such that f ( S i ) E (ai.l ( ( a z .for i = 1. Then n Ui = f'((al. is . 2 and (a1.2.7) such that si E U.7) Hausdorff.bl) ( U Z .Hence we may take two open intervals (a1.
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. a> As is wellknown. as shown in Fig. we conclude that ?rl(T. Under the identification the sides a and b each become circles which intersect in the point 20. modulo the smallest normal subgroup of ?rl(U.21) z1) containing the image +*(rl(UnV. TI)). Let V be the image of the interior of the square under the identification. by the Van Kampen theorem. Since V is simply connected. and let U = T .2. where q is the homomorphism induced L . Let y be the center point of the square. I ) T is isomorphic to 7~1(U. we can present the torus T as the space obtained by identifying the opposite sides of a square.{y}. Solution. b) Compute the fundamental group of the figure 8 and draw a piece of its universal covering space.3 (b).
z o ) is a free group on two generators a. respectively. respectively.4 Let X denote the figure 8 space. respectively.20) is a free group on two generators a and p. a2.the equivalence class of a closed path c which circles around the point y once.{ q } . (See Fig. respectively. each level segment is mapped on the circle a .) On the other hand.21).) b) Fig . which are presented by circles a and b. In a similar way.ba.p.100 by the inclusion map q5 : U n V + U . bl. where a l . a 2 .p2].2.5 Under the covering map T. where a and p are presented by circles a and b. we see that rl(T.2. By the Van Kampen theorem we can prove that a l ( X . it is easy to see that x l ( U n V.2.~ f l ‘ .2. 2 1 ) ) is just the commutator subgroup of 7rl(U.~ . Z is )a free group on two generators a’ = F1a6 and p’ = C1p6. It is easily seen that a u b is a deformation retract of U . P are presented by circles a and b.20) ’ is a free abelian group on two generators a and fl.3 (b). Hence a l ( U . and [ailpi] denotes the commutor aipiailpzrl. V = X .p1. (See Fig.aa. we can see that the fundamental group of the oriented surface of genus 2 is a free group on four generators a1. The following picture is a piece of its universal covering space. The smallest normal subgroup of a l ( U . 21) containing & ( a l ( U n V.p2 with the single relation [a1.3 (a). that q5+(y)= c y ’ p I ~ y ’ .4. Let U = X . as shown in Fig. and. Changing to the base point 20. it Fig ./31][a2. ~ where 6 is the equivalence class of a path d from 20 to 21. / 3 2 are presented by circles a l . where a . 1 ) is a free abelian group on 2 two generators a and p’.{ p } . It is also clear that T ~ ( U . Thus ?rl(T. consequently.2.zl) is an infinite cyclic group generated by 7 . p 1 .
B is a nullhomotopic map. {xn}. consequently.B is open. there exist disjoint open subsets of A . Since A is obviously compact and Hausdorff. Thus f N ( A )c W . which is an open subset of A.101 according to the direction indicated by the arrow. where U and V are disjoint nonempty closed subsets of B.W for every n. n times 00 f"(A) is connected. closed and connected. n=l n fn(A) m PYA) = ( f N W n Wl)u (fN(A) W2>.B be a homotopy between y and the constant map. We claim that there exists a positive integer N such that fN(A)c W . is a compact subset. @ W for every n. can see we that xo E B c W . Since F(S1x I) is compact and X . otherwise. For each positive integer n let fn(A) = ~ o ~ o . and that A is compact. Therefore. closed subspace of a compact Hausdorff space X . we see that B = is a nonempty closed subset of A .f"(A) and such that the induced map y : S1+ X . (Indiana) Solution. Noting that .W is closed in A and. A is also compact. Noting that fn(A) is compact for every n and fn+'(A) c fn(A). which is a contradiction. It is obvious that U and V are also closed subsets of A.. For. and therefore. there exists at least a limit point 20 of the sequence { x C n } . n=l ii) Suppose y : S1 X .. Since A .+'(A) c (i) Show that B = n f"(A) c A for any n. is ii) Let F : S1x I 3 X . there would be a squence in A . n which contradicts the fact that fN(A) connected. Suppose that B is not connected. It means that x. A is a normal space. F(S' x I) is also closed in . o ~ ( A ) . i) Since X is compact and A is closed. and suppose f : A + A is a continuous map. Show that there exists a positive integer n such that y(S1) c X . E A . Thus it is easy to see that f"(A) is compact. Thus it follows that B c W1 U Wz = W . 2202 Let A be a connected. such that x. Wl and Wz such that U c W1 and V c W2. Then B = U u V . E f"(A) but 2.fn(A) ' (y'(s) = y(s) for s E S') is also nullhomotopic. . and each vertical segment is mapped on the circle b according t o the direction indicated by the double arrows.
102 X .T(z) tpo. So homotopic and consequently f is also nullhomotopic. such that pf = f .  + fl is null 2204 Let X be the quotient space obtained by collapsing { p t . cX . Define H : RP" x I + R" by  H(z. .t) (1 . there exists open sets U and V such that F(S1x I ) c U and B c V .fN(A). Then H is a homotopy between f and the constant map PO. say f.t). (Indiana) Solution. Let P : " R + T" be the universal covering map. Prove that any continuous map RP" + T m is nullhomotopic. Therefore. Let po be a fixed point of R". . the induced homomorphism f+ : a 1 ( R T ) + nl(T") is trivial.. for any continuous map f : RP" + T". Therefore. y(S') x . } x to a point S1c S1 x S1 .fN(A).The remainder of ii) is obvious. From the proof for i). c F(S1 x I ) c u c Particularly. Since X is normal. Thus there exists a lifting of f . it follows that there exists a positive N such that fN(A) V . It is also wellknown that "  ai(T") M Z e . x S1 ( m factors). (rn times) and that a l ( R P n ) M Z2. 2203 Let RP" be the real projective nspace and T" be the rntorus S1 x . It is wellknown that R is a universal covering space of T".
V. U and V are open subsets.103 Compute x1(X) and H . 7 r l ( X . zl)). U = X . ( X ) .l(X) =z. Does there exist a continuous map f : T + S1 which is not homotopic to a constant? ( Toronto) .2.(al(U n V. Thus. al(U. and U n V are path connected. and V is simply connected. we may apply the MayerVietoris sequence to the pair ( U . U .zl). (Columbia) Solution. z l ) is isomorphic to the quotient group of a l ( U . where y is the equivalence class of a closed path c which goes once round the point y. containing the 21) image q5.V). 2205 (i) Suppose n 2 2. Let y be the center of the 2gon. (See Fig. 2 and is zero otherwise. The conclusion is that H i ( X ) is an infinite cyclic group for i equal to 0 . 2 1 ) modulo the smallest normal subgroup of R ~ ( U . a F ig.2. zl)is an infinite cyclic group on the generator y. where q5* is the homomorphism induced by the inclusion map q5 : U n V + U . by the Van Kampen theorem. Therefore. Does there exist a continuous map f : S" + S1 which is not homotopic to a constant? (ii) Suppose n 2 2.2. where S is the equivalence class of a path d connecting 20 and z1. 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.6 X may be identified with the space shown in Fig. ( X ) .6) It is also clear that x l ( U n V. 1 .{y}. Therefore we conclude that . zl)is an infinite cyclic group on the generator SluS. and V be the interior of the 2gon. which is obtained by identifying the sides of a 2gon. Then. It is easy t o see that u is a deformation ratract of U . It is easy to see that &(y) = lrl(u.6. To compute H .
t z 5 1. there does not exist any continuous map f : Sn + S1which is not homotopic to a constant map. hence f is homotopic t o a o = C.and. Define it f : T + S 1 by f ( p i t l . 2206 A continuous map of topological spaces: p : E B is called a fibration if it has the homotopy lifting property . which represents a. Let i : F E denote the inclusion map.l ( b o ) (the “fiber”).e. By the same argument as in (i). Solution.any continuous map f : RPn + S1 induces a trivial homomorphism f* : x 1 ( R r n ) a l ( S 1 ) . (iii) Denote T = S1 x S1 by T = ( e 2 r r i t l . 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 . f must be homotopic to a constant. Since R is contractible. respectively. consequently. we see that there is a lifting of f. F = p .eo) + Tl(E. Hence f* is not trivial.which is also a constant map from S” to S1. + . which means that f is not homotopic to a constant. : S” + R such that T? = f. and eo E F . Let p :E B be a fibration. 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. has a lifting RP” + R such that a o f = f. i.rrl(B. (i) If F is path connected. prove that p# : ?rl(E. = and f : S” + S1be a continuous map. ?must be homotopic to a constant map : S” + R. Let G : I x I B be a continuous map defined by f . (ii) Prove that in general the 3term sequence + + + + Tl(F.bo) is surjeceo) tive. b o )where f : I + B is a closed path a t bo . 20 5 t l . e 2 n ).that is.. bo E B be a base point. (i) Let cr = [f] E ~ l ( B . bo) (Indiana) in which the homomorphisms are i# and p # . is exact.104 Solution. (i) Let a : R + S1be the universal covering map defined by ~ ( t )eZrrit. Since al(S”)= 0 for n 2 2.eo) + “1(B.and another generator to zero. (ii) Since a l ( R P n ) = Zz.
b 1 . t ) E I x I . 0) = f(t). l I = b = [fl = a . ~ ( 1 = c i 1 ( 0 ) and c i 1 ( l ) = eo. suppose that G = [f] E kerp#. 1) = f ( t ) . cg and c3 are paths in F and cl(0) = eo. By the homotopy lifting property. Since F is pathconnected.7 2207 Is the canonical map q : S2 + R P 2 (which identifies antipodal points of S2)nullhomotopic? Why or why not? (Indiana) .Then c1. Let c1 = H I { o i x r . h . s ) G ( l . Noting that pgl. Then there exists a homotopy G : I x I + B between pTand the constant path bo such that G(t. (See Fig. It is clear that e l and e2 belong t o F. we have P# rfi = bfi = bg21 . and G(0. s ) = f ( s t ) for ( s . and h : I x (0) + E be the constant map such that h(t. It is easy to see that f is homotopic to C. Then pc = f ..{o}. bg11 . l which means that p# is surjective. kerp# c imi#.eo). 0) = bo. ) ) Hence c = c1 x c2 x c i l is a closed path in F with base point eo.2.) Therefore. pg2 and pgz' are all constant path at bo. Thus f = g2 * c * g1* gF1 is a closed path a t e l .O) eo. where bo is the constant path a t bo. s ) E bo for any s. G = = [c] = i#[c].2. cg = H11. Particularly we have p H ( t . where gF1 is the inverse path of g2. Let C : I + E be a path in E defined by c ( t ) = H ( t . Therefore there exists a continuous map H : I x I + E such that H / I x { o }= h and p H = G . gl(1) = e l . Hence im.  [fi Fig. ~ ( 1 = cz(O).7. 0) = ( p T ) ( t ) . It is obvious that p# . i. there exists a continuous map H : I x I + E such that P H = G and H ( t .{1} and c3 = H l { l } x ~ . g2(0) = e l and gz(1) = e l .105 G ( t . On the other hand. i#(a)= [pf] = [bo]. It is obvious that p h = GII. 1).e. Let c ( 0 ) = e l and c(1) = e2. .Irl(F. (ii) Let a = [f] E . Hence the sequence mentioned above is exact. we may choose two paths g1 and 92 in F such that gl(0) = e2.G(t. c kerp#.
T : ( E . (Here T n = T o T o . denoted by r. 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 . Hence degg = f l . n times. Let i : X t T 2be the inclusion map. Therefore. < 2208 Let p : E B be an universal cover with E and B pathconnected and locally pathconnected. we see that @ is a lifting of T" o p at the base point eo. Let T : B + B be a map so that T n = Id and so that T(b)= bfor some b E B. Prove (Indiana) Solution.= l T ( p ~ n . On the other hand. It is wellknown that q is also the universal covering map from S 2 to R P 2 .) Show that there is a map ? : E + E so that p o T = T o p and Tn= Id. there is a lifting of T o p . = Id. = T ( e 0 ) = eo. = T~ O )  P T " ( e 0 ) = T"'(T(eo)) = T*'(eo) = . f   (Indiana) Solution. This is a contradiction. Since I  T" = ( p T ) and . On the other hand. Then roi : X + X is . Suppose that there is a retraction of T 2 to X . Thus we conclude that q is not nullhomotopic. Suppose that q is nullhomotopic.l = .and therefore q has a lifting : S2 + S2which is also homotopic to the lifting of c .eo) + ( E . . But it is obvious that f is the identity map or the antipodal map from S 2 to S2. eo) since T" = Id..      2209 Let T 2 = S1x S1. We have T o p ( e o ) = T ( b ) = b...l ( b ) and consider the map T o p : E + B . S1.106 Solution. Then q is homotopic to a constant map c : S 2 + R P 2 . o T . by the uniqueness of lifting.eo) such that p o ? = T o p. a constant map C : S 2 + S 2 . Since ?rl(E)is trivial... ~ " .eo) is obviously a lifting of T n p a t the base point eo. . + ( E . it follows that identity map Id : (E. We use the reduction to absurdity. Choose eo E p . we conclude that r?. we have d e g c = degZ = 0.
A ) M H 1 ( Y ) z Y Z @YZ @Z. (Indiana) Solution. 0) respectively. The Euler Characteristic K ( Y ) = 4 .0. + Fig. 2210 Let A c R3 be the union of the x and yaxis.{ e } . Let X = {(x.1(X) is an nonabelian free group on two generators denoted by a and b. Hence we have ab # ba and i*(a)i*(b) z*(b)i*(a).y. z) I x2 z 2 = 1) be a circular cylindrical surface. p ) is not abelian.8.{ p l .Z) I (Y2 + z2)(x2+ 2)= O}. Then it is easy to see that X . 1). a) Compute H 1 ( R 3 . p l and p2 denote the points (1. A = {(X.Y.A .l . It is also clear that X .8 a) Y has a structure of a graph with 4 vertices and 6 edges. Then we have Y = U U V . b) Let e be a point on the arc ab different from a and b. Therefore we conclude that H1(R3 .: r l ( X ) + r l ( X ) is the identity homomorphism.0) and ( . Take U = Y .oi.(2) = 3. and let p = ( O .6 = 2. O .A .2. It is easy to see that U has the . Hence the rank of H I ( Y ) is equal to 1 .{pl.{ c } and V = Y . It is wellknown that r1(T2)is abelian and that .A ) . p z ) is homotopically equivalent to the space Y as shown in Fig.107 the identity map. b) Prove that r l ( R 3. Hence ~.2.Therefore we have = which is a contradiction. O .p2) is a deformations retract of R3 .
Since U n V is contractible. (Indiana) .A . : ( X .9.108 same homotopy type as S1 and that V has the same homotopy type as the “eight figure space”. + Hence 5 is the lifting o f f we want. (A 5. 2211 Let p : (?.y) be a regular covering space. ZO) + ( Y . p*(.g. 9. that (R3 . Therefore we conclude that ?rl(R3 . 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). consequently. Then let fl .A . that is. Show that there is a second lifting. 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.g) + (Y. Let fo(zl) = y.= r o f o : x Y.g) is a normal subgroup of 7rl(Y. there is a deck transformation y such that y(y) = g.rrl(Y. p ) M ?rl(Y) is a free group on three generators and. : (X.Since Since p is a regular covering space.y). p ) is not abelian.) (Indiana) Solution.z1) + need not be distinct from  5 TO (?. and that there is a lifting of f .
z o ) is an infinite cyclic group generated by u the closed path class of a. We may consider f t o be a continuous nonconstant map f : a +a.{y} and let V be the open disk.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. 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 ) ) .2. Note that If(z)l + 00 as JzI + 00. 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.109 Solution.2.) Let U = X . Take a point z1 E U n V as the base point. Then both U and V are path connected subsets of X and X = U U V . Since the circle a is a deformation retract of U . It is also clear that 4+(yc) 3Z. (See Fig. It follows that rl(X) NNZG/U n =as. Since V is simply connected.where 4 : (U n V ) + rl(U)is n the homomorphism induced by the inclusion 4 : U n V + U . (See Fig. Let denote the complex plane and f (2) be a polynomial of positive degree with complex coeficients. it is clear that wl(U.2. hence. Hence the least normal subgroup containing = 4*(rl(U V ) ) is isomorphic to 3 2 .9. we may extend f ' .9 Take a point y in the open disk. by the Van Kampen theorem.9. Therefore r l ( U 121) is an infinite cyclic group on ' generator Z = y'a'y. Fig. where y is the path class of a path d from z o to z1. (Indiana) Solution.) It is clear that r l ( U n V.
+ 2214 Let X denote the subspace of R3 that is the union of the unit sphere S2. 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. and the portion. (Indiana) Solution.(4). (See Fig. Noting the fact that if a continuous map f : S2 + S2 is not surjective then the degree o f f is zero. Then we may first prove that if f ( z ) = z k . of the z axis lying within S 2 . where 00 denotes the north pole. 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 . we take U = X 1 V X 2 .) To compute ~ l ( X v X 2 ) .2. (4 Fig.110 to a map of the onepoint compactification of ( X f : s2+ s2 by setting f ( m ) = 00. we may prove the Fundamental Theorem of Algebra by means of the reduction to absurdity.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.{ p } and V = X 1 V X2 .10. (a) Compute the fundamental group of X . call it A. by the Van Kampen. It is obvious that . Furthermore. the unit disk D2 in the 2y plane. Then 1 we have U u V = X1 V X z . Since U n V is contractible. (b) Compute the integral homology groups of X .k > 0.2. then the degree of the extension f : S 2 + S 2 is equal t o k.
V). . 2 1 ) is surjective.2 and are zero otherwise. there is a closed path h at 20 such that f*([h]) [Z?']. ? r l ( X )is a free product generated by two generators.10. cac is a closed path at f ( z 0 ) . (See Fig. Suppose that the induced homomorphism f is surjective. 20. otherwise. and E X . consequently.1] + X such that c(0) = 20 and c(1) = 2 1 .2 1 Let X be a pathconnected space. Hence  f* : m ( X . we conclude that i = 0. We can take as generators the closed path classes which are determined by the closed paths omp and omq respectively. It means that f o h and Zu?' = are homotopic. f ( z l ) )where a is a closed path at f ( ~ l ) . By the assumption.) (b) Applying the MayerVietoris sequence to the pair ( V .111 Therefore. we see that Noting that H i ( X 1 ) M Hi(X2) which are infinite cyclic for i equal to 0. n ( Y . Since X is pathconnected. For any [a] E ? r I ( Y .1. and.f ( 2 1 ) ) . Thus f o (c'hc) and a are homotopic. Then Z= f o c is a path connecting f(z0)and f(z1) I in Y . f : X Y a continuous function. Show that is also surjective. (Indiana) Solution.2. f*([c'hc]) = [a]. there exists a path C : [0.
Prove that X and Y are homotopy equivalent..2.{ f ( e O ) } is obviously not contractible.11 For any 2fold covering map p : X + B . X has only two different types as shown in Fig.112 2216 Let B denote the “figure eight space”.11.2.2. it is not difficult to see.11 are not homeomorphic. p ) F Z & and X is connected. we come to a contradiction.el}.) Since the automorphism group A ( X . that.{ e o } is contractible and Y . Since X . rlX @ we see that 7rl(RP1x S2) x T1(RP2)@ “1(S2) a @ {O} M 2 =a. Then X . (Indiana) Solution.{f(eo)}. where both X and Y are connected. but not necessarily homeomorphic. let p .{ e o ) is homeomorphic to Y . suppose that f : X + Y is a homeomorphism. Otherwise.11. in substance. 2217 Calculate the fundamental group of the space R P 2 x S2. Then it is easy to see that they are homotopy equivalent to an 3leaved rose G3. Let p : X + B and q : Y + B be 2fold covering maps. by considering the liftings of the circles a and b in X . r ( x Y )= m ( X > m ( Y > . (Indiana) Solution. By the formula . . (See Fig. a m b Fig. But the spaces X and Y shown in Fig.2.l ( z 0 ) = {eo.
' ( * ) = {eO. Since p'(x) n p . and let p : 2 + 2 denote the covering space corresponding to the kernel of h. where * denotes the vertex Fig.) If A is a path component of p . * is generated by 2 and y. Y . . Z = X V Y . we see that the path component of p .l ( * ) . we conclude that A n B = {eo}.. Then.2.e3.y E be the elements in 2 defined by X .12 (a) Let h : ~ ' ( 2 . prove that the number of intersection points of a path component A of p .+ Z/62be the homomorphism satisfying h ( z ) = 2 and *) h(y) = 3.' ( X ) with a path component B of p .e2.113 2218 TI(. (a) Since h ( z ) = 2 and h(y) = 3 and T ~ ( Z . We denote p .+ G a surjection. By the same reason. = ker(h).' ( Y ) divides the order of G .e4. ker h is isomor*)/ phic to the group Z/6Z= Zg. how many intersection points of A and B are there? (i. X and Y circles. and p : 2 + 2 the *) corresponding cover.' ( Y ) .e4} of p'(*) which corresponds to the elements {0. Let x.ez.' ( Y ) = p .' ( * ) by p . what is the cardinality of the sat A n B?) (b) If G is a finite group.' ( X ) contains exactly the points {eO.2.!?.   (Indiana) Solution. Thus ~ l ( 2 . *) Let 2 denote the figure 8 space. h : ~ ' ( 2 . we see that the path component of p .4}of& respectively. Hence the covering space p : 2 + 2 is a 6fold cover.el.' ( X ) and *)) B is a path component of p .e. e3} of p'(*). it ) follows that the homomorphism h is surjective.e5}.' ( Y ) contains exactly the points {eo. ( p * ( ~ 1 ( 2 . from h ( z ) = 2.
2 1 ) + . U and V are open subsets.2.13. zl).rrl(X. Suppose that h ( z ) = r and h(y) = s. It is clear that U n V has the same homotopy type with S1.2. the fundamental group of X . 21) is a n infinite cyclic group and generated by y. with proof. (Indiana) Solution.{y}. Under the identification the edges a become a circle through the point 20. we conclude that $1 : .) Applying the Van Kampen theorem. By Lagrange’s theorem. Let y be the center point of the hexagon. Let 2 1 be a point in U n V . (4 a d @ 2 0 a Fig.13 Represent X as the space obtained by identifying the edges of a hexagon. we see that the number of intersection points of A and B is equal to the number of elements in H1 n Ha. V .zl) + r1(U. . with proof. and let U = X .13. it divides the order of G.114 (b) Since h is a surjection and G is a finite group.Irl(U.2. Let V be the image of the interior of the hexagon under the identification. 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. and V is simply connected. (a) Compute. and U n V are arcwise connected. the corresponding cover p : 5 + 2 is a finite fold cover. Then. the homotopy class of a closed path c which circles around the point y once (see Fig. as shown in Fig. (b) Compute. U . the homology of the universal cover of X .21) is an epimorphism and its kernel is the smallest normal subgroup containing the image of the homomorphism 41: rl(u n v. that wl(U nV. Then r generates a subgroup H1 of G and s generates a subgroup of H2 of G.
1 where It is obvious that the circle a is a deformation retract of U. g ] . Here ( f a  . It is obvious that dl(r) = ar6. X ' ) . From H o ( 2 ) M I . Thus ?rl(U.lrl(U n V . Hi(%) i 2 3. and T 2= S1x S1. then it is known that the set [ X . S (Indiana) Solution. hence X ( 2 ) = 6.21)) is isomorphic t o U . z l ) is an infinite cyclic group generated by a' = ?'a?. Hence the smallest normal subgroup containing 41(. So we conclude that . where 7 is the homotopy class of a path d from zo to 21. see ' we that H 2 ( 2 ) is a free Abelian group of rank 5. where 2 is the 1skeleton of 2.115 and 4 are homomorphism induced by inclusion maps. (a) Define this group operation explicitly and indicate the group identity and how inverses are formed. consequently. = I Z. 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. 2220 i = 1. i = 0. Then the multiplication of [X. 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. 2 . H I ( % ) = 0 and H 2 ( 2 ) x H 2 ( X . g : X + S1is defined by g)(z) = f ( z ) g(z) for any z E X . (b) Compute this group explicitly for X = point. is defined by [f] [g] = [f . (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. ( 0. You do not need to prove your assertions. 7 r l ( U . S1. zo) is an infinite cyclic group generated by a and.S1] of homotopy classes of maps from X to S1 inherits a natural group structure. It is wellknown that the Euler characteristic X ( 2 ) = 6 X ( X ) . where the S1] map f .
g ( z ) is defined by the multiplication of complex numbers. S1defined by t ) = f ( e 2 n i f 3 1e2xit+2xi(lt)+ 1. by still which is homotopic to f under the homotopy map F : X x I + ~ ( ~ 2 r r ie0 n i G . R is contractible. 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 . 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. It is clear that 4 is a epimorphism. where the map e : X + S1is defined by e ( z ) E eZni for any 3 E X. Then let 4 : [X. and cosequently q is an isomorphism. 2. deg(f . which is : defined by f(3:) = l / f ( z ) for any 3: E X . M r2(S1)= 0. [sl) = = = 4Uf. We conclude that 3 [X. can easily prove that one [S’.S1] obviously has only one element. Now we discuss the case of X = T 2 = S1x S1. respectively. denoted by f1. Then by the facts that S2 is simply connected and the universal covering space of S1.S1] %Z@Z. 4 is a homomorphism from [X. deg(f2 . Note that f l can be extended to a map from X to S1. S l ] x “l(S1)xa.9)2) (deg(f1 * Sl). The inverse of [ f ] is the homotopy class of a map f. p o )and fZ(6) = f ( p O l e z U i e )for any e2nie E S1.S 1 ] to Z @Z. one can easily prove that [S2. Thus one can easily prove that 4 is a monomorphism. then [ X . For the case of X = S1.sl) = (deg(f * 911. deg f 2 + deg 92) + (deg 91 deg 92) 1 = (deg fll deg f 2 ) 4(V1) + N g 1 )  Therefore.116 f ( z ) . .degfz). For the case of X = S2. So the group [ X . The identity of this multiplication is [el. (b) If X = point.S1] is trivial.92)) = (deg fl + degg1. We have 4“fI .S1]+Z@Z is defined by 4([f]) = (degf1.
Then exp(2) is the homotopy between f and c . 7r1(X) = 0. 2 0 be T . So the above ) argument still works in this case.e. Since X is locally path connected and simply connected.  Prove that if X is locally path connected and simply connected then every map X 4 S1is homotopic to a constant. i. ?: 7 .117 2221 If X is a pathconnected space whose universal cover is compact. we see that f * ( ~ l ( X )must be trivial.. show that ?rl(X. If n l ( X . locally path connected and the fundamental group of X is finite? (Indiana) Solution. (Indiana) Solution.20) is finite. Thus it is ) easy to see that x is not a local homeomorphism at Z. Let exp : R + 6 denote the exponential covering map. Since R is simply connected. Let ?r : 2 + X be the universal cover of X.~ ( z o must have at least a limit point.such that n(Z) = 20. and f*(?rlT1(X)) for any map f : X + S1. =0 Hence there exists a lifting of f . What can you say if we just assume that X is path connected. and the same conclusion holds. say Z. X + R such that exp(7) = f . then ~ ~ ~ ( would ) a closed set of infinite points in X. i. which is a contradiction. Denote the homotopy between and E by g. Since 2 is compact. f is homotopic to a constant map. the universal covering space of S1. Suppose that ?rl(X) is finite. 20) were not finite. f is homotopic to a constant map Z.. Since ?rl(S1)% Z and f * ( ~ l ( X ) is a finite ) subgroup of ?rl(S1).e.
Since P2(62(a) .. Thus there is = P262(a) . an a E A3 such that a z ( a ) = a2. Hence 7 2 is injective. there is a b E B such that c = b1(b).118 SECTION 3 HOMOLOGY THEORY 2301 Prove the 3 x 3 Lemma. .Pz(b) = E z ( a 2 ) ..&(b) E kercl.(b) = 6162(u) = 0.h ( b ) = w 2 ( a ) . Hence there is a n a2 E A2 such that ~ ~ ( a= Pz(b). Since b1 is surjective. By ) 3 the commutativity we see . then the last row is also short exact. Therefore c = S.e. Consider the following commutative diagram of abelian groups 0 1 0 t A3 62 1 B3 0 0 + 61 + 1 c 3 1 0 If all 3 columns and the first two rows are short exact.Pz(b) = 0 and f ~ 2 is injective.b ) a2 E ker 01 = imcyz. Then since 2 ) c2(al(a2)) = Pl(EZ(a2)) = Pl(PZ(b)) = 0 and C 2 is injective. we show that 7 2 is injective. we have a l ( a 2 ) = 0. i. Let c E C3 and 7 2 ( ~ = 0. (Hamud) Solution. we see & ( a ) = b. To prove the exactness at C.
X3 = U3 x U s . Thus. there is an a2 E A2 such that al(u2) = ul. Therefore. for any i. Let X2 = U3 x 174. Therefore X ( M ) = 0.1.2.i have different parity. Let Xo = T2x T2. M is 272orientable.E 1 & 2 ( a 2 ) = c. Thus by the commutativity. . Denote by u the connected sum of h h projective planes. Thus the exactness a t C is proved. In a similar way we may prove that imy2 c ker 71. Since M is compact. . we have b . and consequently that there is an a1 E A1 such that &(u1) = Pl(b).1.X 1 = R P 2 x R P 2 .~ 2 ( u 2 ) . Then 72(61(b3))= E l h ( b 3 ) = &l(b).Using the fact that + we see that X(X.) = i for i equal to 0. then X ( M ) = 0. i=O By the PoincarC duality theorem. there is an b E B2 such that c = ~ l ( b ) Thus it is easy to see that Pl(b) E ker C1. ( Columbia) Solution. since ~1 is surjective. 2 We leave the proof of the exactness a t C1 to the reader. Therefore there is a b3 E B3 such ) that p 2 ( b 3 ) = b . since i and 2m 1 . Hence c E imy2. For any c E ker 7 1 . they appear in the sum with opposite signs. 2302 Prove that if M is a compact manifold of odd dimension. 2m+ 1 X ( M )= (l)i dimHi(M.&).4. Since a1 is surjective. and X4 = S2x S2.~ 2 ( u 2 E kerpl. Then it is wellknown that X ( U h ) = 2h.119 Now we prove that ker y1 c imy2. 3 and 4.3. Show examples of compact 4manifolds with X = 0. Suppose that d i m M = 2m + 1.2.
s+t) nodd.l defined by (s .t)E C . Let pl and pa be the identification point respectively. By MayerVietoris sequence. ( c ) = Z for even n. we have H . ( T . O ) =z3 (T.1] by identifying X x (1) to a point and X x { 1) to another point. and E X = U UV. It is obviously to see that U and V are contractible spaces and X is a deformation retractor of U n V .e.s.{ p z } .~) + (T. Noting that imdn+l = {(O.s.(C) for all n. Z . s ) E ) i. = t {C"In>O} C" C. So Zn(c) = { ( T .~) + n even (O. For example the sphere S" is the suspension of S"' with the north and south poles corresponding to the two identification points. s . ~ E Zn(c) if and only s = t . s . (Illinois) Solution. ( c ) is isomorphic to Z $Z.t . . The suspension E X of X is defined t o be the identification space obtained from X x [1. Compute the homology of E X in terms of the homology of X . Set U = E X . Compute H. When n is even. en}. } for even n.t. O .120 2303 Let X be a topological space.s+t. : C. we have where H q .l ( X ) denotes the reduced homology of X  2304 Consider the chain complex c d. Then U and V are open sets of E X .{pl} and V = C X .. (Illinois) Solution.
( c ) M Z for any n. ( T . Denote by X1 the "figure eight space" as shown below. So Zn(c) is isomorphic to Z $Z. It is wellknown that H1(X1) = Z$Z. consider the following diagram 0 t HZ(X2) 5 H2(X2.and noting imd. CB imf2. one by the map f i : S' + 20 and the other by the map f2 : S' + a such that f2(z) = z2. H . the image of one 2cell under f1 is homeomorphic to S 2 . Therefore. Therefore.we have Hn(c)= z . X') T fi* 0 + T fils'+ a : H2(D2. the 2sphere. Since H ~ ( xx') = imfl. + 2305 Construct a CW complex which has the following 2homology groups: ( Coz~mbia) Solution. ~. . 0 )E C. which has two generators a and b. s . 0 . Let X2 be the space obtained by attaching two 2cells to X I .121 When n is odd.X1) 3 + H'(X1) % H1(X2) .S1) Hl(S1) + 0 The square is commutative and the level rows are exact.} M Z.+l = { ( ~ .HI(X2. M z CBZ. ~ )E & ( c ) if and only if s t = 0.
1). Denote by S"(q) the space S" V S" V . 2307 Build a CW complex X by adding two 2cells t o S1. + 2306 Compute H.X1) = 0. V ) . n 2 0. Since H1(X2..{ b } also has the homotopy type of S"(q).(S" V S" V . V S"). when n > 0. by induction on q and the MayerVietoris sequence of the pair ( U .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. one by the map g1 : S 2 20 and the other by a map g2 : S2 + S 2 such that degg2 = 3. Then H p ( S o ( q ) )is a free abelian group of rank q for p equal to 0 and is zero otherwise. M 22 and kera. = 0 and imf2Isi. What is the homology of this space? + z4 (Indiana) . Then in a similar way as above.{ u } has the homotopy type of Sn(q) and V = Sn(q 1) . we may see that ima. So(q) has J . It is also clear that Sn(q 1) = U U V and U n V has the homotopy type of S n ( q . where fip is the reduced homology group. we may conclude that the space X satisfies the requirements in the problem. In fact S"(q + 1) = S"(q) v S".. V S" When n = 0. Now let X be the space obtained by attaching two 3cells to X2.122 imfllsi. q times q + 1 points. for all p . Thus. ( Columbia) Solution. = 2!J. z Y . we have H1(X2)M Z $ Z 2 ... It is also easy to see that H z ( X 2 ) =z. It is easy to see that U = S n ( q 1) . by the map z one and the other by the map z + z6. + + + Let a E S n ( q ) and b E S".
since j 2 is injective. KO) T f* T (f IS^)* H2(E2. Since H ~ ( K ' . : and C n ( K )+ Cnl(K) is defined to be the composition of homomorphisms.Knl) and j. To compute H2(h). I 01 otherwise. It is wellknown that H .14. it follows that imd. CB img. i = 2 . is 6Z. Hi(X) = i = 0. is the boundary operator of the pair (Kn. Fig. is isomorphic to Z @&. KM ) ~ imf.KO) M 2.KnA1)d.we have H 1 ( K ) M 2/22 M &.dn}.(K) = 0 for n >_ 3. Let K" = K 2 for n 2 3. consider the following diagram: HZ(K1) t HZ(K2) 2 H2(K2. Thus. lskeleton K1 = S1 and Oskeleton KO = { p } .l is the homomorphism induced by the inclusion map.14 X is a %dimensional CW complex with %skeleton K 2 = X . ( X ) M H n ( K ) . K 2 is obtained from K 1 by attaching two 2cells via the maps f and g as indicated in Fig..123 Solution. Then we have a chain complex K = {Cn(K). is isomorphic to the subgroup 22 of H I ( @ ) .2. where d. Therefore. imd2 M Z.In the same way we see that the image of (glsi). i = 1. K1) 2 HI(K1) 2 H I ( K 1 . where C n ( K )= Hn(Kn. Noting that 1 ker d l % H 1 ( K 1 . It is clear that j is an isomorphism.S1) 3 H l ( S 1 ) The square is commutative and it is well known that f. M Z @&. It is obvious that H n ( X ) = H. Thus we conclude that ( zg3z2. K 1 is obtained from KO by attaching one lcell. we see that H 2 ( K 2 ) imj2 M M ker 8 . Noting f l s i is the map z + z4. . is a n isomorphism.2. is a monomorphism and 8. and that ker 3. we see that the image of ( f l S l ) * is 4Z.
0).0. a) Let U = D3 n X .(0).X . ( X .((O. Since the local homology groups are invariant under homeomorphism. Wz). let Wl = . X . Let h : X + X be any homeomorphism.124 2308 Let X = {(z. From the above results. u .O.(0)) M H2(X. Then S is a deformation retract of U . 1)) = s and W2 = s . It is easy to see that U is contractible.(f(0))).z) E R3 I zyz = 0). Let S = S 2 n X . where +X must leave the origin fixed. So H.X . otherwise. Thus H .(0)) M H.we see that b) Let z be any point of X different from the origin. Similarly we obtain Hz(X. (O. b) Prove that any homeomorphism h : X Solution.I(U.{ 0)). where S 2 is the unit 2sphere. I). (Indiana) D3 = {(z.y. if z is in a coordinate axis.(U . Applying the MayerVietoris sequence of the pair (W1. x . ( X .0)). H 2 ( X .O.(0)) for any q. a) Compute H . To compute Z*(S). y. x . z ) E R3 I z2 + y2 + z 2 < 2). .{XI) = Z @Z @Z.((11 0.(0)) = H.(1.(U.(0)) g * ( S ) . it follows that f(0) = 0.
Thus we have %I iYi(S1 x R1 .{ p } ) = otherwise.{ p } ) x Hq(S1).+ H q ( D.(D . .sphere) b) Use a) to calculate the homology of S1x R1 . it follows that Hg(S1x R1)x H. Compute Hq ( R P 5 / R P 2 ) . (S1 = 1 . S1x R1 .{ p } .{ p } ) f gl(S1 x R1)+ 0 . (Indiana) Solution. Since R P 5 is compact Hausdorff and R P 2 is a strong deformation retract of a compact neighborhood of R P 2 in R P 5 . the nontrivial part of the MayerVietoris sequence is 0 + g1(D..(D) f gq(S1X R1) + H g .{ p } ) @ @. RZ which is split exact.. and a disk D in S1 x R1 about the point p. Since D is contractible.{ p } ) 3 Hq(S1x R1 . (Indiana) Solution.125 2309 a) Write down the MayerVietoris sequence in reduced homology which relates the spaces S1x R1. 2310 Let the real projective plane R P 2 be embedded in the standard way in the real projective 5space R P 5 ..{ p } ) 3 x R1 . where R P 5 / R P 2 is the space obtained from R P 5 by identifying R P 2 to a point.{ p } ) + *. A  b) Since S1 x R1 and D . ( R P 5 / R P 2 )x H q ( R P 5R P 2 ) .{ p } . a) The MayerVietoris sequence is .{ p } have the same homotopy type with S1. for any q. we see that H.l ( D .
2 2 . q = 0.Y x R . otherwise.5. Solution. P 2 ) .5.*. { z. i = 0.2. 0. 2 3 ) E R 3 1 2 1 2 2 = 0 .3. where D3 is an open ball centered a t 0. 0. (RP5/RP2) = a. 4 = 0. from the exact sequence of the pair ( R P 5 . U = D3 n (Y x R ) . Q q Thus. 0.{ 0 } ) . otherwise.{0}). (Indiana) Prove that Y x R is not homeomorphic to R 2 . = 1.z2) E R2 I 2122 = 0 and 22 2 0).  because U is contractible. &.126 It is wellknown that H. Therefore Hi(Y x R .O) E Y x R. Then the point 0 = (0. we take an open neighborhood of the point 0. and H. 2311 Let Y = {(z1.(0)) M Hi(U.(RP2) = otherwise. 0.it follows that R H. q = 1. Let X be the space as shown in the following figure .*. &. To compute the local homology groups Hi(Y x R.1. = 0. Y x R .(RP5)= { z. Let Y x R = { ( 2 1 . 2 22 0 ) . U  (0)) M Hil(U .
(X.it follows from the exact sequence that .YO)+ H * ( Y / B . Hence.yo) = 0.u .B ) = H * ( Y . Let U = A x [$. *. Thus.(?r(U). Thus H z ( Y x R. Let ?r : Y + Y / B be the identification map and yo denote the point ?r(B)in Y / B . Then yo is a strong deformation retract of ?r(U). Let Y = X x (0) U A x [0. A ) .v.YO) * * * it follows that H. we have an isomorphism 11.1]and B = A x (1). ( X U C A ) is isomorphic to H. Since B is a strong deformation retract of U . that is. for every n. 2312 Let A be a nonempty subset of X and X u C A be the union of X with the cone of A. It is wellknown that for any point p E R2 the local homology group H 2 ( R 2 .Y x R .?r(U).R2 .( p } ) GZ.?r(U)) H. in the exact sequence of the triple ( Y / B . 1 Prove that the reduced homology group g .(0)) EZCBZ.(Y/B.V ) .127 It is easy to see that X is a deformation retract of U . (Indiana) Solution. By the excision property. Let V = A x [i. 1 by identifying A x (1) to a point. H * ( Y .U ) . 1 of 1 X x [0. Then X U C A = Y / B .1] c Y . the inclusion map of pairs induces an isomorphism H * ( Y / B .* + Hn(?r(U).l(?r(U).yo).(0) and that H I ( X ) X Z @Z.?r(U)). X U C A is obtained from the subset X x (0) U A x [0. y o ) + + + H. H * ( Y . Since the local homology groups are isomorphic under homeomorphism. U ) M H * ( Y . it follows that Y x R is not homeomorphic to R2. Y O ) = H*(Y/B.
we have X x (0) H.v. ( X UCA). yo) Since Y admits obviously a strong deformation retraction onto which maps B onto A x { 0 } .: (Y  v. A ) . A ) for every 2313 n. x Therefore. Excision. where denotes the equivalence relation generated by requiring that (2.) (Indiana) . let Cf : C X + C Y be the map of suspensions induced by the map ( ~ . . ) (y.u  V )+ ( Y / B . On the other hand.u V). For any topological space X let EX denote its (unreduced) suspension. ( X .128 Thus. Dimension) for a homology theory prove that Hi (EX) naturally isois morphic to Eil(X). ( X . B )M H * ( Y . s) if s = t = 0 or s = t = 1.T ( V ) T ( V ). we have H * ( Y . B ) M H .   (ii) Using only the EilenbergSteenrod axioms (Homotopy.. and so an isomorphism of their homology groups. Exactness. All of these combine to give an isomorphism H * ( Y / B . (Here Hi denotes reduced singular homology.B ) . from the exact sequence of the pair ( Y / B . H. it follows that H*(Y/B.(V)).(X U C A ) is isomorphic to H . In a similar way the set T ( V )may be excised from the pair ( Y / B .(Y.yo). (EX is the quotient space (X x [0. M H * ( Y .1]/ .t) X x [0.1]. t ) (f(z).n(V)) to give an isomorphism Now the restriction of the map n. + of (i) Prove that if f : X + Y is a homotopy equivalence then so is Cf.) t I f f : X + Y is a map.gives a homeomorphism of pairs n.yo)H * ( Y / B )= G .
3 It is obvious that U and V are both open sets of EX and that U and V are both contractible. pl the equivalence classes of (z.1) = g(z) for any z E X . 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 . and consequently it follows that if f : X + Y is a homotopy equivalence then so is Cf. On the other hand. s) = E(II&'(%). By the definition we see that Cf = rIy 0 (f x id) 0 IIXl. (ii) Let 1 u = {rIx(z.i(U) = g. U ) . for any (2. lI.W). U . from the homolopy sequence of the pair (V. It follows from = the homology sequence of the pair (EX. (i) Denote by rIx : X x I + C X the quotient map.W has the same homotopy type of X .t) 1 t < }. 1) = Eg(Z). where id : I + I is the identity map. By the excision property we have W = {IIx(z.t)1 t > } 3 and 2 = {rIx(z. It is easy to check that H is welldefined and continuous.W ) M E&'(U . Then v It is easy to see that U .O) = f(z) and G ( z . 0) = Cf(Z) and H ( E . respectively. where.'(Z) means any point (z. l).s) E EX x I .t ) 1 t 2 2/3}.(V) 0 for all i. Therefore Cf is homotopic to Cg. U ) that fii(CX) M Hi(EX. and by po. 0) and (z. . 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. 1).z E X . Hence r?.129 Solution.O) or (z. u . Let f : X + Y be a continuous map. It is also easy to see that H ( Z . if Z = po or p l . Thus we conclude that k i ( C X ) k*l(X). Let c U .W ) we see that Hi(V.
a2 E kerP2. Once again by the exactness we have 0 = PlPo(b0) = P l ( b l ) . and the inclusion U n V + V induces a surjective homomorphism H1(UnV) + H 1 ( V ) . Then from the exactness there is an a2 E A2 such that a z ( u 2 ) = a3.) a) Prove that H 1 ( X ) = 0. there is a bo E Bo such that Po(b0) = b l . Suppose that y and Po are surjective. a3 is ) injective.a 2 . . Prove that a3 is injective. It means that kerag = 0. H 1 ( U ) = 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.130 2314 Consider the following commutative diagram of abelian groups in which the row and column are exact sequences.a2 = 0. there is an a1 E Al such that y ( a 1 ) = /32(a2). Since PO is surjective. (Indiana) Solution. Hence a3 = ~ 2 ( a 2 = a 2 a l ( a l ) = 0. Since y is surjective. i. and there is a bl E B1 such that P l ( b 1 ) = a(a1) . Let a3 E A3 such that cyg(a3) = 0. (Here H denotes singular homology.. Hence by the commutativity we have a l ( a 1 ) .e. which means that a l ( a 1 ) .
V ) A . d* is surjective so that Therefore im& = 0. y. It is obvious that U . we see that 20 E D" .e. (Indiana) Solution..1 . 1 (Indiana) Solution.l . + H ~ ( x 3 H ~ ( U v)% H ~ ( u CB H ~ ( v )H ~ ( x + H ~ ( U v).. a) Consider the MayerVietoris sequence of the pair ( U .131 b) Give an example t o show that the analogous assertion becomes false in general if H is replaced everywhere by Hz.5 ) and 1 1 v = {(z. ) n ) 5 ) n Since U n V is pathconnected. But H 2 ( X ) M Z.1)sphere. 2316 Let D" = { z E R" 1121 5 1) denote the standard unit ball in Euclidean space Rn.. . z ) E s2I z < }. So the homomorphism H z ( U n V ) + H z ( V ) induced by the inclusion map is surjective. Suppose that there is a point 20 E D n such that zo @ f(D"). H 1 ( X ) = 0. Suppose that f : D n t D" is a continuous map such that the restriction of f to S"l is a homeomorphism from S"l to s n . V and U n V are pathconnected and that H z ( U ) = H2(V)= 0. 2 Then X = U U V . and denote the standard (n. Prove that f is surjective. b) Let X = S 2 .9 . By the hypothesis. We use the reduction to absurdity. It means H 1 ( X ) = im$*. Eo(U n V )= 0.z) E s2I z > . the unit 2sphere7 u = {(zclg. i. By the assumption.
which is a homeomorphism.X ' ) 3 H 1 ( X 1 ) 2. Assume H1(X) x Z @Z/3. 2317 Let X be a connected CW complex with two 0cells. and i.l ( D n ) is t obviously a trivial homomorphism because of Hnl(Dn) = 0.X O ) } . Let X k denote the kskeleton of X .e.1 : 9 . By the assumption. Then g = r o f is a continuous map from Dn to S"l and gIs.rankHl(X) + rankH. X 1 )is free abelian. and H 2 ( X 2 . The Euler characteristic of X .132 Therefore Sn' is a deformation retract of D" . H2(X) Z Z $Z. Hence (g 0 i)* : Hnl(S"') + H. where d.XO)..X ' ) t H 1 ( X 1 . (Indiana) Solution. Let T : D" . This is a contradiction.(20). H 2 ( X ) is also a free abelian group of rank 2. H1(X1.( 2 0 ) + Sn' be a retraction. Thus because H 2 ( X 3 .Compute the Euler characteristic of X and determine (with proof) all possibilities for H2(X). is the composition of homomorphisms H 2 ( X 2 . : H2(X2. + X ( X ) = rankHo(X) . Therefore rankHz(X) = 2. : Hn1(Sn') H n . and no higherdimensional cells. X ( X ) = 2 . we see that H 3 ( X 3 .(X).3 3 = 2. X 1 )= Z $ Z @ Z . and consequently (g o i)* is trivial. X 2 ) 0. On the other hand. . Therefore = H 2 ( X ) M ker{d. Therefore g o i : Sn' + Sn' is a homeomorphism. i. But ( g o i)* = g* o i. Let i : S"l + Dn be the inclusion map. three 2cells. three 1cells.l + S"' is just the restriction of f to Sn'.'(S"') is an isomorphism.
l . i Y k ( X . (Indiana) Solution. Z )is not a nontrivial finite group and contradicts the hypothesis. and Hk+l(X. H ' + l ( X k + l . But it is wellknown that H k ( X k . X k ) . kerdk is trivial or free abelian. We claim that &+'(a) Otherwise we would have imd'+l = 0. consequently. Thus we have d'+'(a) # 0. therefore. Denote by Xk the Icskeleton of the C W complex X .1 ) It is wellknown that 3 Hkl(Xkl) 5 H k T l ( X k . and. By the hypothesis.Z) M ker dk+1/imdk+2. Hk+l(X. Z ) M ker dk/imd"' xk2).133 2318 Let X be a CW complex with exactly one k+lcell.X"'). X k ) zz.Z) is a nontrivial finite group then Hk+l(X. 2319 Let and . We choose a generator a in H k + l ( X k + l .x k . X k . Prove that if Hk(X.Z) = 0. H k ( X . Therefore # 0. It follows that kerd"' = 0 and. It means that H k ( X . and by dk the composition of homomorphisms of pairs H k ( X k . Z )M ker dk c H k ( X k .Z) = 0.l ) is a free abelian group.
15.15) + 2: + 2: = 1).134 Let X = A u B. X . It is obvious that U is contractible and that U . u . (Indiana) Solution.O)) = Hi(U . 0.{p3. O . O . 0) in X . 0.(0. O . we see that 2 = 0.O)) . other wise. 0. O .p4}. Therefore.O) has a deformation retract C U D. 0. O .15 Let Wi = C U D . W2).(O.(O. we see that Hi+l(U. 2 4 )E R4 I 2 4 = 0. 0)). Compute the relative homology groups H i ( X .O.u .x  (O. 0.{pl.O. It is also clear that S1 is an deformation retract of W n Wz.p2) and W2 = C U D . Then U = 6n X is an open neighborhood of ( O .O)) for all a. 2: (See Fig. Fig. applying the MayerVietoris 1 sequence to the pair (W1. (See Fig. 0. Let 6 be the unit open ball centered at the point ( O .(0. 0.) Then W1 and W2 are open subsets of C U D and W u W2 = C u D.(0. where and D = { ( ~ 1 . 2 3 . It is clear 1 that both C and D are homeomorphic to the unit sphere S2 and that D and C are deformation retracts of W1 and Wz respectively. 0.2.( O . By the excision property we have &(X.O)) M H i ( U . 2.2.O. O . 2 2 .O)) for all i. O ) in R4. O . From the homology sequence of the pair ( U . U .
(Indiana) Solution. It is easy t o see that R2 x S 2 is the universal covering space of Y .O. 0 f* : H l ( X .Z) is trivial for i different from 0 and 2. then any continuous map f : X + Y induces the trivial map f* : H i ( X .135 for all i 2 0. I ) for all i different from 0 and 2. we have It is clear that Rl(S1 x s1x S1) = “l(T2)@ “ l ( S 2 ) = a $a. Hence we conclude that Hj(X. Show that if al(X. where R : R2 x S 2 + Y is the covering map. the homomorphism f* : R ~ ( Xz) + R ~ ( Y .(O. must . Z ) + H i ( Y . Since ?rl(X.O. f(z)) be trivial. Hence we have  7 f* = R.Z) Hi(R2 x S2.X . 2320 Let X be a path connected and locally path connected space and let x E X . Therefore. Let Y = S 1 x S1x S2. By the Kunneth theorem.x) is finite. Z ) +  + Hi(Y.1 otherwise. Z ) for any i. so is . 0)) = i=o. we may lift f t o a map : X + R2 x S 2 such that a o f = f . But it is obvious that : H i ( X .z) is finite.
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. By the assumptions.&a’a= 6’ya . we have (6’6.F 6’ are exact.6’ya = 0. B P1 and C + D 17 B S E 71 1 E D .+ E 1 E PI 6’6 C . then so is A C ’a l .Therefore.t F Prove exactness of (Indiana) Solution.&)(P. we have 6P = 7 a and &a’= 6’7. .a’a) 6’6P = .
+ X be a continuous map.e.y) = ker(c .S’). a’a). Then there would exist an 2 0 E A such that z 5 zo for any z E A. We give an example to show that compactness is essential for this assertion.. there would exists an z1 E A such that f(z1) = 2 0 .) = b and p(a) = c .. We claim that f ( A ) = A holds. It means that ker(6’6 . Let ( c . 6 ( c ) )E ker(E .137 It follows that im(P. In fact z o = sup z. (Indiana) Solution. Then F is a nonempty closed subset of X. 20 1 = 521. p ) = ker(y . Give an example to show that compactness is essential for this assertion. c ) E ker(y . which is a contradiction.6). i. noting that ( e . The exactness at C @ E is proved. Prove that there exists a nonempty closed subset A of X such that f ( A ) = A .a’).1] and f : X + X is defined by f(z) = for z E X. we see that the subset A = n=l n F. and. Since f ( A ) = A iz xEA and zo E A . a’a). e) E ker(6’6 . consequently.E ) . It is clear that { F n . a’a)(u)E im(0. 00 is a nonempty closed subset of X . 2322 Let X be a nonempty compact Hausdorff space and f : X . we see that there exists a b E B such that a’(b) = e and y(b) = 6 ( c ) .6) and im(a’. Hence we have (b. n E Z+} is a sequence of nonempty closed subsets in X and that Since X is compact.E ) c im(P. Therefore zo 2 z1 = 220. Thus.E ) . Let X = (0. Let F = f ( X ) . we have i m ( a . Suppose that a nonempty closed A of X satisfies f ( A ) = A . S’6(c) E ( e ) = 0.a’a) c ker(6’S . ( c . We define 1 1 Fn+l = f(Fn) inductively for n E Z+.e ) = (0. i. there exists an a E A such that a(. By the assumptions. .e. Therefore.
Z) Z .> I ZY = 01. show that X is not homeomorphic to R2. Therefore. Then it is easy to see that A U B is a deformation retract of X .d e g f . Applying the MayerVietoris sequence to the pair ( U . (Indiana) Solution. Then it is clear that the degree of 7r is equal to 2.O. Then flx(o. M (RP" is ndimensional real projective space. Let f be a function from S3 to RP3. (b) Using part a . 0 . Take U = A U B . 1).(O. It is wellknown that S3 is the 2fold universal covering space of RP3.) Prove that there is no function f : S3 + RP3 inducing an isomorphism on the third homology.( O . O . ( a ) = 2 . (b) Suppose that there exists a homeomorphism f : X + R2.(O. O .O. X = 7 ~ 1 7r2 where 7r1 is the plane y = 0 and 7r2 is the plane U z = 0. Denote by A and B the unit circle in 7rl and a2 respectively. U U V = A U B and U n V has four path components which are all contractible. which . 0 ) ) . (a) Compute H1(X . (c) Prove or disprove: X is homotopy equivalent t o R2.Y. Denote by a and b the generators of H 3 ( S 3 ) and H3(RP3)respectively. l ) and V = A U B . (a) In fact. .(O.138 2323 Recall that H 3 ( S 3 .(0. 0) is also a homeomorphism. Then we have f*(a)= 7r*(f*(a)) d e g f .(0. b .V). there is a lifting of f.f ( O . we see that H1(X . 0 ) + R2 .o. (Indiana) Solution.it is obvious that f* is not an isomorphism.O. 0). Z )x H3(RP3. Since 7r1(S3)= { 0 } . 2324 Let x = ((2. 0. =  I because deg f E Z. Let 7~ : S3 + RP3 denote the universal covering map. r .o) : X . O ) ) x H 1 ( A U B ) =Z $Z $Z. f : S3 + S3 such that 7 r f = f . U and V are both contractible.
d. Consider the following diagram H . X onto the plane 7r2. ( X . Then F is a deformation retraction of homotopy equivalent to R 2 . . = H . Hence X is 2325 Let ( B " .l : S"l + A is a homeomorphism.H.l ) be the standard ball and sphere pair in R". and therefore It is wellknown that 8: is an isomorphism and H n . By the assumption. Since H . . ( X ) = 0. f1.( O .(X.i(X) t f* a: t fl* i : + t f* Hnl(B") Hn(B". ( X . is an isomorphism.f ( O . 0)).l ) in which the level rows are exact and the squares are commutative.i(A) i* t H. Show that if & ( X ) = 0 then H . Thus we have keri. ( X ) 1?. is a monomorphism.S n . Suppose that f : ( B " . = Z.A ) M imd.l ( S n .l ( B n ) = 0. n > 1. 0)) to H1(R2 . A ) is a continuous map and f l s n .139 wouid induce an isomorphism from H1(X . (c) Let F : X x I +X be a map defined as follows. Hence we have H . O . + H. M ker i.S"1)+ H " . O .S n . (Indiana) Solution. A ) = Z .A) a.l ( A ) = Z . But it is clear that It is a contradiction.l ) + ( X .
p 2 } . By the Kiinneth formula we conclude that otherwise.2. ez e z ez. and S5 is obtained by attaching a 5cell to XI. otherwise. x 1 E S 5 . 2. (b) Let X = S 2 x S 5 . V are both open sets of S2x S5 and UnV = X. ( S 2 x S2)are (no proof is required). Then S2 is obtained by attaching a 2cell t o 2 0 . and is zero otherwise. and UUV = S 2 x S5. ( S 2 x S2. which can be considered as the space obtained by attaching a 2cell and a 5cell to the point (20. U = S 2 x S 5 . Hence it is obvious that H i ( S 2 x S 5 ) is an infinite cyclic group for i equal to 0.{ p 2 } . Therefore the CW structure on S2 x S 5 has a 0cell.6.140 2326 (a) Describe a CW structure on S2x S5 and use it t o compute the homology of s 2 x s . 5 (b) Compute the homology of S2 x S5 with 2 points removed. State what the homology groups H . a 2cell.A)? (Indiana) Solution. It is easy to see that S 2 x S5 is homeomorphic to the space obtained by attaching a 7cell to S2V S5. It is easy to see that S2VS5 a deformation retract of both U and V . (Indiana) Solution. Then U . and 7. i = 1.{ p 1 } a n d V = S 2 x S 5 . 21) E S2 x S5. Denote by S2 V S5 the one point union of S2 and S5. Applying is the MayerVietoris sequence to the pair (U. V) and using the fact that we conclude that i = 0.{ p l . a 5cell and a 7cell. . (a) Let xo E S 2 . 5.5. 2327 Let A be a subspace of S 2 x S2 homeomorphic to the 2sphere S2. What can you say about the possibilities for the relative homology groups H .
  2328 Compute the homology groups of the space X obtained as the union of the 2sphere S2 and the xaxis in R3.A ) M keri. X”) infinite cyclic for i equal to 1 and is zero is otherwise. x we have H3(S2 x S 2 . ( A ) M Hi(S2). 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.1.A ) M Z @Z/irni. is the homomorphism induced by the inclusion map i : A S2 x S 2 .) Show that there is a natural isomorphism ax : H. and H 2 ( S 2 x S 2 . Since &(A) = 0.. (Indiana) Solution. A )+0 0 + H 1 ( S 2 x S 2 .using the above result and the fact that H .we see that H . It is easy to see that X’ has the homotopy type of 5’. A ) . (This is the union of two “cones” on X. It is clear that Hi(X. H. ( S 2 x S 2 . A ) = 0 for i equal to 0 and 1. By the exactness.X”) conclude that Hi(X) infinite we is cyclic for i equal to 0. A ) 4 H 2 ( A ) 5 H 2 ( S 2 x S 2 ) 5 H 2 ( S 2 x S 2 .141 From the homology sequence of the pair ( S 2 x S 2 .l) in the xaxis}.. 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 . 4 (Indiana) .(S2 S 2 .(X) H i + l ( C X ) . Then the space X may be viewed as a space obtained from X* by attaching a 1cell.2 and is zero otherwise.{the open interval (1. where i. Let X’ = X . A ) +0. Thus from the homology sequence of the pair ( X .A ) + H o ( A ) + Ho(S2 x S 2 ) + H o ( S 2 x S 2 . for all i 2 0. a.
a D 2 ) 4 w H l ( d D 2 )+ 0 . 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 & . CTX : & i ( X ) + H i + l ( C X ) is just what we are looking for. Thus the nontrivial part of the sequence is as follows. where 20 E S1. ) : V .( 2 0 ) and V = C X ..142 Solution. It is wellknown that H . 2330 Denote by X the union of the torus S1x S1with the disc D 2 .16 (a) Calculate H. (a) Consider the homology sequence of the pair ( X ...(21).) 20 Fig. Let (0)’x X and (1) x X be identified to the points 20 and 2 1 respectively. Consider the MayerVietories sequence of the pair (U. Take U = EX .. (See below. T 2 ) .. and U u V = EX. The naturality of ax can be derived from the naturality of the MayerVietories sequence. 0 4 H 2 ( T 2 )5 H 2 ( X ) 1 H 2 ( X .where z f * is the homomorphism induced by the adjunction map f : D 2 + X . where D 2 is attached to T 2 by identifying dD2 with a meridian curve S1x (20) in the torus.(X) for all n 2 0. ( X ) . + ~ ( U ~ 3 )H i + l ( u > ~ H i9 H(~~ > ( E x ) H i ( u n v ) . V s l +~ A ++ +  It is clear that U and V are contractible and that U n V has the homotopy type of X . T 2 )= 0 for i # 2 and H 2 ( X .2. ( X . T 2 ) . (b) Is T 2 a retract of X ? Why or why not? (Indiana) Solution. d D 2 ) ) Z . T 2 ) M f * ( H 2 ( D 2 . The inverse of A * . Then U and V are open sets of E X . H .. + a: H 1 ( T 2 )2 H 1 ( X ) + 0 Qt f* t fl* 0 + H2(D2.
a )E C x A 1 g ( c ) = h ( a ) } .16) It is clear that f 1 * ( H l ( a D 2 ) )= 2 @ (0). and. we have the following commutative diagram  because of H 1 ( T 2 ) = Z @I and H 1 ( X ) =Z. the pull back of h via g is the group g * ( A )= { ( c . Thus we see that H 1 ( X ) x Z @Z/Z M Z and H z ( X ) M H 2 ( T 2 )x Z . Then r o i = i&z : T 2 + T 2 . 2331 Given homomorphism h : A B and g : C + B . This is a contradiction. Let C = 2/82 and let g : C + B be the surjection defined by g(1) = 1. Hence we conclude that (b) We claim that T 2 is not a retract of X. (Indiana) .143 It is easy to see that ima: = irnfl. Let i : T2+ X be the inclusion map. and let h : A + B be the surjection defined by h(1) = 1. there would exist a retraction map T : X . consequently.+ T 2 such that ~ ( =pi d p . 9*( A ) c 1 g*(h) A l h . we have T* o i. = id : H 1 ( T 2 ) H 1 ( T 2 ) . (See Fig. B = 2/Z. and ker& = ker f 1 + . + (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 ) . Otherwise.+ B 9 (b) Let A = Z / U . Prove that the kernel i of g*(A)is isomorphic to the kernel of h. We know that H 1 ( T 2 ) = Z @Z is generated by a and b.with explanation.2. Identify the group g * ( A ) . In other words.
C.3. .~ .a) = l. (7.3).. 2332 Recall that if C is a homeomorphic copy of the circle in S 3 .a) I a E kerh}.{m}. therefore. the identity of C. we have h'(O) = {0. (0. then H i ( S 3C) is infinite cyclic for i equal to 0 or 1 and is zero otherwise.C ) @ Hi(S3 .. Assuming this fact compute (a) The homology of R3 . otherwise.4.e.R )~+ . (5... Thus it is clear that the group g*(A) consists of the following 16 elements: (O.C ) % Hi(S3 . and.{ }= G@3) = 0 a ) for any i and that H i p 3 )= { z.. a ) = a.6} and g'(l) = {1.5.B = S3 .. where X c R3 is a homeomorphic copy of the "figureeight space" (i. h'(l) = {1.0). i = 3.+'(S3) 2 Hi(R3. (a) Denote S3 = R3 U { }Let A = S3 .2.7}.144 Solution. when C is a homeomorphic copy of the circle in R3.) (Indiana) Solution. g'(O) = {0. Then g*(h)(c. (5. (4. (6. ( L 3 L (3.2). the onepoint union of two circles.C ) = . and Hi(S3 .(7.3). (2. (a) Suppose that ( c .3).O).. and A U B = S 3 and A n B = R3 .1). (1.2). we see that kerg*(h) = {(l. .X .{co}) % Hi(s3)A H ~ . (3. By the definition of g*(A). (b) The homology of Y = R3 .0). C = l. I>. + H.C. I). ( C +  Noting that Ga(S3 .0). (4. (2. (b) In this case. Then A and a .2). 0. Let F : kerg*(h) + ker h be defined by F(l. It is easy to see that F is an isomorphism.2)..2}. u ) E kerg*(h).3}. (6.C. We have the following MayerVietoris sequence.1>. B are open subsets in S3.
{PI} and V = X . Thus we get H 1 ( Y ) = Z@Z and a split exact sequence   O + & ( Y ) 5ZCBZ5Z+O.x = ( R.17 Represent X as shown in Fig.V ) + 0.U n V .2.17.[c]. Since the representative chain c of [c] can also be considered as a chain of (R3 .145 we can see that  = {: i = 0. Since U u V = X .U ) and gi(R3.V )5 i i 2 ( ~ u n V > 2 U 3 3 %Z$Z H 1 ( Y ) 2 ii1(R3 . we have Y = R~ . ~ ~ From the result of (a).X .U.( ~ 2 ) . Then we see that U and V have homotopy type of S 1 and that U n V is contractible.U n V).u) n ( R ~v) ~ and ( R .1.(u nv).V ) is o + ii2(y) i i 2 ( ~ 3 ) CB i i 2 ( ~ . which means imA = 0 too. we see that Zi(R3.u )u ( R .as the generator of E2(R3 . Due to the above facts. Let U = X . It is easy to see that iii(x3 (v n v))M iri(s2)>.U ) n (R3 .2. %A7 @Z  MZ + A We claim that the homomorphism is an epimorphism.V )is infinite cyclic for i equal to 1 or 2 and is zero otherwise. the nontrivial part of the MayerVietoris sequence of the pair (R3 .V ) . . Take a sufficiently large T > 0 such that the sphere S 2 ( ~belongs to R3 .v )= R~ . Then S 2 ( r )is a ) deformation retract of R3 . The claim is proved. Fig. otherwise. R3 .2. Hence we can consider the generator of H 2 ( S 2 ( r ) ) .by the definition of ++ we have [c] = + + ( [ c ] .0).U ) @ ii1(R3 .
(S3 . this fact to compute the homology of S 3 .B are open sets of S3 and ( S 3 . . q#3. 4'1. ii. 4=2. .H . in the MayerVietoris sequence we have *. where A n B = 8 and both A and B are homeomorphic to S1. . S3 ..A ) @ &(s3 .(S1) = and that 4 # 1. 4 = 0. i = 0.Y )5 .B ) . (Indiana) Solution. ( S 3 . Noting S3 . 4.(S3) = we can easily see that { Z.A and S3 . Denote Y by Y = A U B.X ) M H. 3. ~ Z. i=l.B)= S 3 .ii.A ) n ( S 3 . .+l(~3) . 0. A  A  &(s3.Y )=  { 2*Z. H.Y where Y is a subspace Use of S3 homeomorphic to the disjoint union of two copies of S1.B) H .A ) U ( S 3 . .B ) M H.(S3 .146 It follows that HZ(Y) =Z. i = 2. 4 = 1.Y ). ( s ~ .A ) M Hq(S3.Y = ( S 3 . It is known that if X c S3 is homeomorphic t o S1 then H . Using the fact that Z. H .(S1).~ ( s .So we conclude that H l ( Y ) = {:w Z7 I 2333 i 2 3. ( s ~ ) . Therefore. * .
it is clear that . The covering map T . The connected 2fold covering of S1v S2 is shown as follows. (b) Compute the homology of the connected 2fold covering space of S1vS2. we take U = S2 V S1V S2 ..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). Thus. (a) . and. is the identity map.2. and. is the exponential map: R + S1. reis stricted on S1. i. and that S1V S2 is a deformation retract of U and V.2. the homology of S2 V S1V S 2 . (Indiana) Solution.{the antipodal point of p l } and V = S 2 V S1V S2..2. Fig. b) To compute the homology of the connected 2fold covering space of S1v S 2 .18 The universal covering of the one point union S1VS2 is shown as in Fig. restricted on R.18. restricted on each S2. Fig.. restricted on each S2.is the identity map. Then it is easy to see that S1is a deformation retract of U n V .19 The covering map T .e.{the antipodal point of p 2 } (see the above figure). the 2fold covering map: z + z 2 from S1to S1.
otherwise. Hence the homomorphism   4. where A and B are each homeomorphic to S1and .(S2) @ H. 4 = 1121 0. Thus imA = 0. It is clear that   H ~ ( v s1v s2) H ~ ( u ) kl(V)/im#. 4 # 172. : G1(u n v)+ H1(u)@ H1(v) is injective. : H1(U n V) + H1(V). we have H. 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. : H1(U n V) + k 1 ( U ) and I . . ker 4%= 0. MZ.(S1) = Therefore. 2335 Compute the homology of S1 x S1point. It is easy to see that the inclusionmaps k : unv + U and I : U n V + V induce injective homomorphisms k.20 It is easy to see that the space X = A u B is a deformation retract of the space H = S1x S1point.. the nontrivial part of the MayerVietoris sequence is c Z.148 and &(U) M &(V) x H. Fig. s ~ M Hence. 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 . 2.e. i.(S2   v s1v S 2 ) = 4 = 1. respectively. (Indiana) Solution.
(b) .. b } is contractible.point) = H . ( B ) = x 2336 Suppose the following diagram is commutative. o@..u E kerj. i BL + Anl + * * * Write out the proof of exactness at Bh. j n ( b ) = jA(b). Construct an exact sequence .lojA).. Thus Pn(b).(a’) . We give the proof of exactness at Bh as follows. Due to the exactness . Let U = X . ( A ) @ H . .(a’) = Pn(b).i . and consequently. .{ a .149 A n B = {zo} as shown in Fig. 7. and let V = X . at C.. and that U n V = X . ( d )= A(a’. and 7.. Then 7 ’ o j L ( u ) E ker6. It means that u = P. +A’.P n ( b ) for any (a’. 07. From the commutativity. + A . b).(b) = j k ( u ) . the rows are exact. Let u E ker(6. we have nL2. b ) = . such that jn(b) = 7 ’ o j .u. Applying MayerVietoris sequence t o U and V.{ u } .. It is clear that A and B are deformation retracts of U and V .{ b } .there exists a b E B. and there exists an a’ E A. @ B. (Indiana) Solution. .2. The following sequence is exact: where A is defined by A(a’. ( u ) . respectively. ( X ) = H . @ B. Hn(S1 S1.20 Choose points a E A and b E B such that u # z o and b # 2 0 . we have j . b) E A. such that i. is an isomorphism for all n.
from (l). it is easy to see that 6 0 7’ .M Z H.(b) = 0. i. . Suppose that X is a space and f : X + Y . Let M be the mapping cylinder o f f and g .(U) H . Noting U . II 1 I Let U = X x [0. u = ih(a’) . (Indiana) Solution.e. where X x [0. ( U n V ) Z & ( X ) . o j .P.     . In the same way.+ H. 0 jL(U) = . .3/4] = Y/. Thus the exactness a t B. It is obvious that U n V has the same homotopy type with X . from the assumption that Y is contractible.l ( X ) . Now we prove imA c ker(6. Thus. H.1 ( U ) @ H q .. o 7 ’ o j .l ( V ) + ’.150 So ker(6. V and U n V are open sets of M . the equivalence relation is determined by (z.. 1) with g(z). is proved. where the equivalence     relation is determined by (z. . X x I and Z by identifying each (z. . Suppose that u E imA. 2 In the same way. + A  H .k .1X. o Pn = 7.O) with f (z). o 7 ’ ojk). ( M ) Z k q . ’ Since j . 1) g(z). 0) II f ( z ) .1] = Z/ ’. 6 0 j.3/4] and Y . . Prove that H. we have the following exact sequence: ’  * .(M) + H q . ( Y ) = 0. .3/4] = Y denotes the disjoint union of X x [0. o ih = 0 and j . g : X + 2 are two maps of X into contractible spaces Y and 2.(u n V ) + H J U ) @ I?p(v)+ H. let V = X x (f. So k . ) C imA. M is the identification space obtained from the disjoint union of Y .(b) for some a E Ah and 6 E B. by MayerVietoris sequence.. Then we have M = U U V and U n V = X x (1/2. and each (z. that is. we have H.(V) E H.(Z) = 0.3/4).~ (n v > u (1) II Since X x (0) = Y / is homeomorphic to Y and is a deformation retract of U .
Part 111 Differential Geometry .
.
i) For any N E Z + sketch an example of such an a(. by Blaschke.(e) = p(e)sin 6 p'(0) cos 8. 5 2 (or some other constant independent of a .) with d. of a ( s ) to be Now assume that the curvature k ( s ) 2 1 for all s.y) of the oval.153 SECTION 1 DIFFERENTIAL GEOMETRY OF CURVES 3101 Let a ( s ) be a closed plane curve. Define the diameter d. where the oval is counterclockwise orientated and 8 denotes the oriented angle from the zaxis to I.1 ii) From the hypothesis that k ( s ) 2 1 for all s. + + . 2 is the best possible such constant). ii) Assume further that a is a simple closed curve. > N . i) The following is an example of such an a ( s ) with k ( s ) 2 1 and d.3. "t Fig. Prove that d. > N . 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. (Indiana) Solution. we know that the simple closed curve a is an oval. then the oval can be parameterized by 8 as follows . y(e) = p(e) cos e pye) sine. For every oval.
Suppose that. respectively. From this we can conclude that. . a t every pair of points P and PI where C and C1 have the same tangent orientation. 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 (. p ( 8 ) can be uniquely determined by k ( 8 ) = k. From the above.. by direct computation. Then the domain encircled by C1 must contain the domain encircled by C. y(8)).(a).(8) = ( p ( 8 ) + (8) ) . Hence. we can say function p ( 8 ) must be the solution to the following initial ODE origin.1. take the tangent point 0 of C and C1 as the common tangent line as the zaxis.(8) of C. the relative curvature of the oval is k. the curvatures of C and C1 satisfy the inequality k l ( P l ) 5 k ( P ) . Fig. Fig.3.2 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 0.3 In fact..154 p ( 8 ) is called the support function of the oval.3. Let p ( 8 ) and p l ( 8 ) be the of C and C1.
Secondly. Therefore. we also have the similar expression of p l ( 8 ) . Then . For convenience. we always have p l ( 8 ) 2p ( 8 ) . we have.b). Thus. 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. firstly. T ( t ) = 0 and c ( t ) = k. we know that As for the sign of sin(8 . for the curve C1. N ( t ) = c ( t ) a ( t )for all t .can get we pl(8) . we see that the domain encircled by C must be contained in the domain encircled by C1. we have sin(8 . Thus we can immediately obtain k ( t ) = constant. (Indiana) Solution. 2 0. we assume that the parameter t is the arc length of the curve a. The assertion of Blaschke is proved. By the hypothesis. by the fundamental theorem of the theory of curves. then owing to 0 5 4 5 8. that is. Therefore. Suppose " the normal vector N ( t ) is proportional to the position vector. .4)  I  + 3102 Let a be a regular C curve in lR3 with nonvanishing curvature. Hence. by the Frenet formula. then ii) follows immediately from the above assertion. if x 5 8 5 2x. where c is a smooth function. Noticing that every oval is the envelope of all its tangent lines.p ( 8 ) 2 0. if 0 5 8 5 x. Differentiating both sides of the equality N ( t ) = c(t)cr(t)with respect to t . and take a as C .155 Analogously. we know that the curve a must be a circle (or a part of it). If we take a circle with radius 1 and centered at a ( s g ) N ( s o ) as C1.p ( 8 ) 2 0. p l ( 8 ) . Determine all such curves. where 8 and its corresponding original 8 satisfy 8 + 8 = 27~.
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. a) Show that a circular helix is an example of such a curve. a) Let a circular helix be parameterized as follows y(s) = (Tcosws.dv) satisfies L(u.v)dv2 = 0. every point of S is a planar point. v ) are the coefficients of the second fundamental form of S at point p ( u . Determine all triply ruled surfaces. both nonvanishing. c) Show that k / r = constant = k t a n a . In other words. Observe that every asymptotic direction ( d u . then.v). (Indiana) Solution. N ( u . Therefore.v). .v)dudv + N ( u . L3 lying in S such that L1 n L2 n L3 = { p } .w) = M ( u . we obtain L(u. S must be a plane (or a part of it). 3104 Let y : (a. Noticing that the above equation is of 2nd order with respect to du : dw and it has two roots. hws).Tsinws.b) 4 R3 be smooth with [y’l = 1 and curvature k and torsion 7. b) Show that N . v). (Indiana) Solution.v ) = N ( u . Lz. a = constant = cos a. If S is a triply ruled surface. Denote the Frenet frame by { T .w)du2 + 2 M ( u . a = 0. v ) . N . by the hypothesis. Assume there exists a unit vector a E R3 with T . there are three different asymptotic directions at every point of s. M ( u . where L(u. B } .v) = 0 for all ( u .
and the geodesic y becomes a straight line on the developed infinite sector. and let T and N denote the unit tangent and normal vectors to . c) By the property of b). a straight line never intersecting the generator 1. Then it is easy to verify that s is the arc length parameter.k ( s ) / ~ ( s ) & t a n a . then T ( s ) . (z)y) E R2\{0.a = 0. If we take a = (O. T(s ) ) N (s) . = 3105 Show that if y is a geodesic on the cone t = then y intersects itself at most a finite number of times. then we can conclude that y never intersects itself. Noticing that the central angle of the sector is f i r . then the cone becomes an infinite sector without the vertex. Differentiating the above equality with respect to 3 s. l). b ) .T(s) f s i n a . h. from which follows N ( s ) . B ( s ) .hw).rwcosws.157 where T . we may assume that the constant vector a = cosa. If we cut the cone along a generator 1 and develop it into a plane.b) 4 R3 be a C" curve parameterized by arc length. Assume k ( s ) # 0. for all s E ( a .0)) (Indiana) Solution. we obtain k ( s ) N ( s ). b ) . with curvature and torsion k ( s ) and ~ ( s ) .b ) . 3106 Let y : (a. ~ ( s # 0 for all ) s E ( a . Hence T ( s )= (rwsinws. Thus.a = 0 for all s E ( a . a = hw = constant.O. two rays which start from the generator I .i are all constants. w = ( r 2+ h 2 ) . and the image of y on the sector must be one of the following three cases: a generator. d m . we have 0 = (cos a k( s) f sin a . an obtuse angle. b) Differentiating T e a = constant with respect t o the arc length parameter s.
7 is called the Bertrand mate of y and we can write y(s) = y ( s ) + T N ( s )for some T = ~ ( sE) IR. (Indiana) Solution. namely r = const. r ( s ) ) N ( s ) b’(s)B(s). (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). . (c) Suppose that y1 and 7 2 are the Bertrand mates of y.C2 such that for all s E ( a . then y is a circular + helix. (Note that s might not be an arc length parameter for 7. Then. + C27(5) = 1. (b) From (a).b).b). namely (1 .= a’(s)T(s) ( a ( s ) k ( s ) b ( s ) . then there exists a constant C such that r k ( s ) C T ( S= 1 for all s E ( a . we have that r’(s) = 0. there exists a constant C such that r k ( s ) C T ( S = 1 for all s E ( a .) (a) Prove that r is constant. now we have  T ( s ) = (1 ds ds rk(s))T(s)  ds TT(S)B(S). b) + IR3 such that for each s E ( a . ds dz ) Therefore.r ~ ( s ) = const.r k ( s ) ) = const.158 y. dZ ds + + + Hence. 1 r 1k( s ) r2k(s) + + ClT(9) = 1. we obtain  k ( s ) N ( s ) . ds If we denote  T ( s )= a ( s ) T ( s ) b ( s ) B ( s ) . b ) the normal lines of y and 7 a t s are equal. there exist constants T I . (b) Prove that if y is a Bertrand curve (with r as above). C . The curve y is called a Bertrand curve if there exists a regular curve 7 : (a. from N ( s ) = fN(s) we know that a’(s) = b’(s) = 0. ) (c) Prove that if 7 has more than one Bertrand mate. + then by differentiating with respect to s .b). by (b). rg. ds dz . In this case.
( s ) .N ( s ) ) = 0. Vs. 3107 Let z(s) be a curve in R3 parameterized by arclength. Differentiating ) ( z ( s ) .B ( s ) } N is the Fkenet frame field along z(s).k’(s) k 2 ( S ) T ( S) * Conversely. t ( 8 ) . and a ( s ) .159 Because the nonzero constants T I . still differentiating (z(s). b ( s ) . Hence y must be ) a circular helix. Suppose that z(s) lies on a sphere centered at the origin. which means that b ( s ) = 2 At last.P ( S ) 1 +(s) P ( S ) = constant. ~ ( s = const. we have ( z ( s ) . where { T ( s ) . from which it follows that u ( s ) = 0. (Indiana) k4(s) S o ht ion. z(s)) = R2 with respect to s. we obtain = + + 1 + k ( s ) ( 4 s ) . the above system of linear algebraic equations has solution k ( s ) = const. N ( s ) ) = & with respect to s. differentiating . T ( s ) ) 0 with respect to s again. c(s) = . we obtain namely. Differentiating (z(s). c ( s ) are suitable functions to be ascertained later. T ( s ) = 0. Assume that ~ ( s # ) 0 and k’(s) # 0 for all s. 12 are not equal. Then we may assume that Z(S) = a ( s ) T ( s ) b ( s ) N ( s ) c ( s ) B ( s ) . Show that a necessary and sufficient condition for z(s) to lie on a sphere is that 1 +.
160 with respect to s. which is orthogonal to the zaxis. where k ( t ) is the curvature of the curve. Let M + c R3 be the torus (Indiana) Solution. Is this curve a geodesic on M ? Explain without long computations. is parallel to the normal of M along c ( t ) . nor its principal normal. 1) (“top circle”). 2 sint. . 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 . namely. denoted by m. y. the curve c ( t ) is not a geodesic on M .m) = constant. Observe that the geodesic curvature of a curve on M can be computed as follows k . we have that vanishes identically because k’(s) # 0 and Then is a constant vector. Hence (z(s) . Then. which is parallel to the zaxis. = f k ( t ) sin€J(t). 3108 obtained by rotating the circle ((0. and let c ( t ) = (2cost. because neither the top circle is a straight line.m . z ( s ) . z(s) lies on a sphere centered a t m. z ) : (y 2)’ z’ = 1) around the zaxis.
denoted by B . Let M be the surface in lR3 obtained by rotating the graph {(z. (Indiana) Solution.1] + lR3 be a C curve with 1il = 1 and nonvanishing curvature. y lies in a plane. which means that y ( s ) B = constant. then the corresponding normal curvature Ic. = implies that cos8(zP) = f l .e. (a) Show that y lies in a plane. 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. i.161 3109 Let y : [0. the corresponding normal curvature where O(zp)is the angle between the normal of M and the principal normal of the circle of latitude at P. .f(z)) : z E IR} has a tangent parallel t o the z axis at P . (b) What happens if the curvature is allowed to vanish a t a point? (Indiana) Solution. the curve {(z. (a) 1. f(zp)). = 1 implies that the curve is parameterized by arclength s. Thus. " . Namely. that is. that is.in the direction of the circle of latitude. 0. From il the Frenet formula we know that the binormal vector field of y. Since every circle of latitude on M is a line of curvature. $(y(s)B) = 0. f'(zp)= 0. (b) The vanishing curvature at point SO means ? ( S O ) = 0. is a principal curvature. Assume the torsion 7 = 0. Ic. Hint. then it is an inflection point of the curve 7. f ( z ) ) . is constant. At P = ( z p 0. f ( z ) ) : z E B}of f in the z z plane about the z axis. Thus. Local coordinate computations are not necessary.O. *$J . If the point is the strict extreme value point of the tangent vector field.. 3110 Let f : lR + lR be positive and smooth.
Assume that the position vector a(. (Indiana) + Solution. we conclude that the set of L such that f of M a t (z. we have ( a ( s ) . a(. T i) If k ( s ) > 0 and T ( S ) are curvature and torsion of a in E3.O. Its curvature Ic a t P is just the other principal curvature of M . Show that a ( s ) does not pass through 0 E lR3. the second possible case is Therefore.) is always a linear combination of the binormal and normal vectors. Integrating the obtained equality.) E M 2 is parameterized by arclength. show that a ( s ) is a closed curve. 0. show that a ( s ) is a closed curve (kg is the geodesic curvature in M). then ( T ( s ) . the meridian passing P is also a line of curvature. If the position vector a(. N ( s ) of a ( s ) . a ( s ) = const. we ) will get a ( s ) 5 0. (Indiana) Solution.) is always a linear combination of the binormal and normal vector B ( s ) . Thus. if a ( s ) passes through the origin. the trivial case.) c R3 be a smooth curve parameterized by arclength. f(z)) is & is a principal curvature 3111 Let a(. whereas the hypothesis of curvature k ( s ) > 0 implies that the plane R . i) The hypothesis of torsion ~ ( s ) 0 implies that the curve a is a plane curve.162 Besides.) ii) If Icg(s) 1. 3112 Let M 2 c IR3 be the cylinder z2 y2 = 1. (Here k . Suppose the curve a(. Therefore.)) 0.
Y ) . & + &} + + ”> for suitable functions a ( s ) and b(s). i. Thus the curvature of & is = 1 in x..y(s). then the corresponding plane curve of a . a must be closed. y(s). .f3’(s) E ” q 3 ( $ ) . = (z. 0). By x ( p ( s ) ) = a ( s ) . ii) If one develops M into a plane x. has the same geodesic curvature kg(s) 3 1. Then . denoted by &. Therefore.O.e.) implies that d p ’ ( s ) = u(s)b(s) .163 where the curve a lies does not parallel the generating line of the cylinder M . we may assume that p(s) = (z(s).O). 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.y. where x(2. i) That z& = and z&. the curve a c M n x must be closed.show that there exists a unique C curve p(s) E B3 such that . P ( S O ) = a ( s 0 ) and x(P(. (Indiana) Solution. Y.y E IR} C IR3 and ” a ( s o ) .B’(s) E Tp(. are linearly independent shows that T is not involutive. ii) Let a ( s ) = (z(s).p ( s 0 ) = (O. So.)) = a ( s > .O) : z. z ( s ) ) where z ( s ) is unknown.z(s)(:9 d2 dX [& & + I & { &. iii) Show that if a ( s ) is a simple closed curve of length L bounding the region R in B2 then @(so + L ) . ii) Given a C curve a ( s ) E B2 = ((z. &A) where A = area of R. ds ’ ds ’ ds Hence Thus the problem of finding problem p(s) reduces to solving the following initial value .
. N + case. where the sign is determined by the orientation of the curve a.l ) 0.P(s0) = (z(so + L ) . i. then we easily have + L ) .)sine.. B ( s ) . 3114 Call a normal vector field Y along a space curve y parallel if l is always i tangent t o y. i.cose k~ ( B .e.where e(s) is a smooth function globally defined on y. ii) Show that the total torsion of any closed curve y which lies on a sphere in lR3 must vanish. T N  = .y(so f L ) . B ) B +case. iii) If P(s0 Q(S) is simple and closed. N + (4  T )c o s ~B. + i.O. i) The hypothesis that l is always tangent to y means that (v.TB)+ s i n e .fA). f ZdY) = (O. (Indiana) Solution. 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. the normal vector field v has constant norm.e. and v ( s ) = cos O(s) . Therefore. respectively.the unknown function z ( s ) can be and so is the space curve P ( s ) . N ( s ) sinB(s) . without loss of generality. Here s and L denote arclength and the length of y. we may assume that (vI G 1. (ICT.z(so). and i i= hence..Y(SO>.. . Thus we have ) I” r ( s) ds . = (sine.164 When the given plane curve uniquely determined by Q(S) is C”. which measures the angle between ~ ( s and N ( s ) .
Let k ( s ) be the curvature of the intersection curve C = M n P . Let po E y and y contract to po smoothly on S 2 . B ( L ).e(o) = 2n7F.hence the segment is also a geodesic one.1]. = ra T(S)dS +qo). Also. we have the relation qS) J.L]can be regarded as a lift of a certain differentiable map f : [O. for every ytrwe have the corresponding map ft such that fo = f . the smooth function e ( s ) . finally we have I” T(S)~S = 27r . Thus we get a family of curves { y t } . Furthermore. Otherwise. where the integer n is just the degree of the map f. Therefore. the normal of M is parallel to the principal normal of C.165 e(s) Noting that i/ is parallel to T . On the other hand. fact that y is The closed means that f +)dS = e ( q . l). n = deg f. Show that the intersection curve (parameterized by arclength) is a geodesic on M . . then C is naturally a geodesic on M . the unit normal vector field Y of the sphere along y satisfies the above mentioned hypothesis. If k 5 0.we see that the above equality implies that = ~ ( s )and hence .. ii) For any closed curve y which lies on a sphere in El3. along the segment where k ( s ) # 0. (Indiana) Solution. t E [0. L] S1into R1. fi = the constant map into PO. yt overlaps with y = yo about po = yl. we may assume that for every t E [0. s E [O. Because the degree of a map is homotopically invariant. i. Suppose M and P intersect orthogonally.q o ) = J 0 L r(s)ds.deg f1 = 0.e. 3115 Let M be a surface in El3 and let P be a plane.
for each t E lR.e.) be a C2 curve in lR3 parameterized by arclength. . we have which implies that f(s)= k2(s). If f ( s ) 0. a”(. What can you deduce about f(s). T(S) = 0.. Suppose that for some function f(s). Derive the curvature function for this new curve. i. Hence k ( s ) = const.) = f ( s ) a ( s ) .. Define a new curve 7 by setting. then ( ( ~ ( s ) )= const. k’(s) = 0. a ( s ) is a spherical curve. a ( s ) is a straight line. i.r ( t ):= cy(t/c). and a ( s ) is a circle or part of it with radius 1 .166 3116 Let a(. From 1 f ( s ) a ( s ) a ’ ( s ) d(a’(s))2 = 0 = 2 we have f(s)d(a(s))2 = 0. If f ( s ) # 0. 0 3117 Suppose y(t) parameterizes a space curve with curvature function K ( t ) . then a”(s) 0. where c E lR is an arbitrary fixed constant. differentiating with respect to s. (Indiana) Solution. (Y(S)? (Indiana) Solution.e. Further~ more.
Is ?r unique? ii) Is the conclusion of i) still true if the curvature is allowed to vanish at a single point c E (a.t E ( a .0) where the function f is defined by . from which follows ( a ( t ( s ) )B ) = const = (a(t(so)). the torsion is not defined at c. but is assumed to be zero a t all other points. B Therefore. namely. ( t . Then the hypothesis T implies that B is a constant vector field. that is. ) . ?rl # x2.O. s E ( ~ 1 .b) + lR3 be a smooth curve with nonvanishing curvature. the plane ?r : ( p . $ ( a l l ? )= 0.a ( t ( s g ) ) B ) = 0 contains the curve a. suppose a = a ( t ( s ) ) .167 Therefore.b). according to the above discussion. Obviously. 0) if t = 0.] (Indiana) Solution. a ( t )E ?r for all t E ( a .b)? [Of course. The condition k # 0 means that we can ~) 0 define the F’renet frame field { T . the connectedness of a and the smoothness of the Frenet frame field imply that T is unique. ( t . and a ( t ) . a ( t ) . c ) must be on a plane T I . then. b ) must be on another plane a2. A counterexample is as follows. Let a(t)= { if t < 0.0.f ( t ) . maybe. i) Let a be reparameterized by arclength s. . b ) . B } along a. i) Show that if the torsion of a vanishes identically then there is a plane ?r c lR3 containing a . from y ( t )= y ’ ( t / c ) and y”(t) = y ” ( t / c ) / c we have 3118 Let a : ( a . f ( t ) ) if t > 0. Of course. Hence. t ( s g ) = b. (O.N .t E ( c . swith t ( s 1 ) = a . ii) If the curvature k is allowed to vanish a t a single point c E ( a . .
f3’(t)) 0. Differentiate the equation with respect t o s. For convenience. Then we may assume ...B successively.Tl(sI))= 0. (Recall that a is a generalized helix if for some constant vector u # 0. let a and .s)T(s). and M be locally parameterized by X(ul. (u.c ? .B(sl) = a ( s ) X ( s ) T ( s ) . E where s is arclength for a. P ( t ) lies on the line tangent to a a t a ( t )and (a’(t). Using the Frenet formulas and the hypothesis (T(s).. Therefore. Let a be parameterized by arclength s . that a plane contains it.a’(s)) const. Noting that a being a generalized hrelix implies + + we obtain by straightforward calculation i. and that the Gauss curvature of M vanishes a t each point of a.= . along a .e. Deduce that in this case. = Show that every involute of a generalized helix a is a plane curve. Firstly. p is a plane curve. Differentiate the obtained expression for . s1 represent their corresponding points. (Indiana) a regular curve a Solution. we have.. The curve .f3 is called an involute of a if for all t . we can ascertain that P ( s 1 ) = a ( s ) (so .W d X n ds iJ=l 2 ds aut (Z) = W(a’). a is an asymptotic (Note: A curve in a surface is called asymptotic if its acceleration is everywhere tangent to that surface. be parameterized by their arclength s and B s1 respectively. the torsion of p vanishes everywhere. 3120 Suppose that the unit normal vector to a surface M c lR3 is constant along c M .168 3119 Let a and p be two regular curves in lR3. .) curve. and let s .. .) (Indiana) Solution. u2). by the Weingarten formula. duj o = d= .
Thirdly.2HII K I = 0 it follows that along Q + K =. n )= 0.a')= 0.:2 ) ( + 2H(W(a'). we immediately deduce that. Hence a" is everywhere tangent to the surface. and has the following graph. n). by the Gauss formula. = where k. from I11 .) (Indiana) Fig. we have d ( a ( s ) .4 . which means that the curve a is contained by a plane ( p . ). satisfies r" Jo L(s)ds m 3 ~ 1 2 .a ( s o ) .3.a ( s g ) . ifp's curvature function k ( s ) ( s = arclength) is odd. Secondly. n )= 0. ( a ( s ). ( ( ~ ( s n) = const = (a(so). T]+ R2having p(0) = (0. along a 2 2 duj duk = kgn x a' + (W(a'). namely.169 Further.T .0). Therefore.n) = 0. is geodesic curvature. along a . noting that n is constant.O) andP'(0) = (1. 3121 Sketch the closed regular plane curve p : [ .a')n kgn x a'. (Include an explanation with your sketch.
which has zaxis as its tangent line a t (0.” g d s = 6 ( ~) 6(0) x T .5. The image of the plane curve p is a “figure eight”. O ) . as shown in Fig.5 . where 6 ( s ) is the oriented angle formed by . And yaxis is almost its “tangent” line at ( O .3. Fig. This assertion follows from I” k(s)ds = 3a J.O).B’(s) and p’(0).3.170 Solution. too.
(B2. 2 2 . Thus.v). then their images are simple closed geodesics on ( E 2(xl)*ds2).Compute interior of Q J k. ~ 2 ~ 2 E S2\{N} to (u. d s 2 ) + . using the pull back of ds2 by (x')*. 2 3 ) E B3 : 2 : xi 22 = l}. (Indiana) Solution. ii) Let a(. the map x : ( S 2 \ { N } . i) Let S2 = { ( 2 1 . Since all great circles which do not pass through the north pole are closed geodesics on S2\{N}.its coordinates with respect to the z1 and 2 2 axes are denoted by (u. which maps (21. Here the integral is with respect to the Riemannian volume induced by the metric. Suppose that x : S2\{N} + B2 is the stereographic projection from the north pole of S 2 into the plane lR2.) be a simple closed geodesic as described above and K ( z ) be the Gaussian curvature of the above metric at 2 E B2.171 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.1 u2+v2+1' 212 + Then the metric of S2\{N} can be expressed by ds2 = + dv2) (22 + 212 + 1 ) 2 ' 4(du2 Now. For every p E R 2 . .v) E B2. . and B2 be the 21 o 2 2 plane. we obtain 3 ) + + 21 = u2 + v2 + 1' 2u 22 = u2 + v2 + 1' 2v 23 = v2 .(x')*ds2) is an isometry. we can obtain a 2dimensional Riemannian manifold ( B 2(x')*ds2).By direct calculation.
Assume that and ax 1 tangent to M a t z = 0. &) we immediately obtain Ic. . &)and the geodesic curvature k. i. we immediately have K ( x ) = 2 T X ( D ) = 2T. 1). or more directly. (Indiana) Solution. You may assume that the normal to the surface at x = 0 is ( O . 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. Then..(O. i) By the Meusnier theorem. 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.(O. By the relation among k ( O ) .172 ii) Denote by D the simply connected domain encircled by the simple closed geodesic a.e. interior of J ( I 3202 Let M 2 c lR3 be a surface containing x = 0. the normal curvature of Q at x = 0 is ii) Since a is obtained by intersecting M with 21x3 plane. using the famous GaussBonnet formula. a is a normal section. then its curvature at z = 0 is L ( 0 ) = lk. &)I = 4. K ( 0 ) = det ( y4 ) = 25. kn(O. &)= 0. iii) Because the eigenvalues of the Weingarten map L are f 5 . O . we know that the Gaussian curvature of M 2 at x = 0 is K ( 0 ) = 25.(O.
6 3203 Let M = {(x.(z2 y2)]dx A dy = + 12r d 1 2 ( 6 . (Indiana) Solution. Choose the unit outward pointing normal as the orientation of M . we have k [ 6 .173 iv) From iii). B . z ) E M : z > 21. &)= 4 > 0. z ) E E3 : x2 y2 _< 4 . and w = yz2dx zzdy x2y2dz.3. we know that x = 0 is a hyperbolic point of M 2 . z = 2) and take (0. y. Then. z ) E E3 : z = 6 . Therefore. s / Fig. y.r2)rdr = 16n. we sketch the surface near x = 0 as follows. by the Stokes’ formula. Firstly.(O. 1) as its normal vector. D = {(x.0. we have that + l U p d w = 0.(zz y2)} (a paraboloid of revolution). Noting that the normal to the surface at z = 0 is (O. Hence. Orient M and evaluate + + + LzdxAdy+dw. y. Let P = {(z. 1) and k.O.
3204 Let S be a surface diffeomorphic to the ordinary 2sphere. d y A d x = . Hence. r zdz A dy + dw = 8T. . Applying the global GaussBonnet formula to the domain D c S and noting that y1 and y2 are geodesics of S.2r2 + r4 sin201'. Show that y1 and yz must intersect. J. Here we assume the Jordan curve theorem. Suppose there is a C" metric of positive curvature on S for which there exist two simple closed geodesics y1 and 7 2 .0) = r2[1 . Let ( z 1 .y ) E lR2 : x2 y2 < and let 0 = (0.O) E D. we have K d c = 27rx(D) = 0. (Indiana) Solution. 0 ) be the usual polar coordinates and define a metric on D\{O} by + i} where h(r. If y1 and yz do not intersect. 3205 Let D be the disk { ( x . x 2 )= ( r . noticing that dw = z 2 d y A d x 2yzd. then y1 and 7 encircle a domain D such 2 that the Euler characteristic x ( D ) = 0 and d D = 71 uyz.& . + + we have ~dw=~4dyAdx+2dxAdy=2 J.z A d x + z d x A d y +zdz A dy + 2xy2dx A d z 2x2ydy A d z . which contradicts the assumption of positive curvature.174 Secondly.
as a whole line segment. y) as the coordinates.+ 2lii2dxdy + li22. { Then = r. t 00. Use a smooth coordinate system.(1 . this metric can extend to a smooth metric az = % i +2 cos 8 sin 8[l . Hint. we can choose (x.8)d82. we know that kgIo = 0. (b) For any line segment I in D .2r2 + 4r4 sin2 8)2]dy2 . From { follows dx = cos Bdr . 1 is a geodesic. we can express 2\{O} as the union of Il = { ( r . + Hence the metric of D\{O} may be rewritten as ds2 = dr2 + h(r. = giidx2 When r 0. 8 ) : 0 < T < :. Therefore. . of I must be a continuous function of the coordinates. Using ds2 = dr2 + h(r.2r2 + = [cos28 + (1 .8)de2 + 4r4 sin2 8)2]dxdy +[sin2 8 + cos2 8( 1 . the geodesic curvature k . So. (b) Show that line segments in D which pass through the origin (suitably parameterized) are geodesics. and %o( l). (Indiana) Solution. If we use the extended metric. we know that both l1 and 12 are geodesics in D\{O}.d~~. (a) Set x = rCOS8.r sin 8d8 dy = sin 8dr T cos 8d8 + { dr = cos dx sin 8dy d8 = (cos 8dy . g12 = o(T). Z22 = 1 + o ( r ) .T sin adz). y = r sin 8.8 = B0 + T } . Hence. e ) : 0 < r < :.175 (a) Prove that this metric extends to a smooth metric on D.2r2 + 4r4 sin219)~ 8)dz2 sin2 on D . we see that &I = 1 + o ( r ) . on D\{O}. $& = o( 1). Because 11 and 12 are geodesics by either metric.8 = O 0 } and l2 = { ( r .
there must be at least one common point of M and the sphere with radius R and centered a t the fixed point.. . for minimal surfaces. using the local canonical forms of the normal section curves M n 7~ and S 2 n7~ a t p . (Indiana) Solution. a t p . 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. Suppose M is inside a ball of radius R and suppose a point p E M is on the boundary sphere. (b) Let V(0) = ( 0 . y) (a) Compute the components of the LeviCivita connection (i. is the usual inner product on B2. 3207 Let H 2 = {(z. 6. Show that the Gauss curvature of M at p is 2 ii) Show that there is no closed minimal surface in lR3.176 3206 i) Let M be a surface (without boundary) in lR3. &. Let a be a normal plane of M and S 2 at p . the Christoffel symbols). Ler R be the infimum of their radii. l ) be a tangent vector a t the point ( 0 . Using the result of i).b( t) )be the parallel transport of V(0) along the curve z = t . the curvature of M n T is not less than that of S2na = S1. But it contradicts the fact that . Then. Then there is a family of spheres that contain the surface M inside and have a fixed center. p is an elliptic point of M . i) The surface M and the sphere S 2 .y) E lR2.e. one can easily show that. K ( p ) 2 &. have the same tangent plane a t p . Let V(t) = (a(t). Then. one concludes that the Gauss curvature of M a t p . the boundary of the ball. l ) E H 2 .y > 0) be the upper halfplane in lR2 and let H 2 have the Riemannian metric g such that where (2.
we see that the problem = a(t) 9= b(t). (b) Use the definition from part (a) to compute K at ( 0 .u 2 ( t ) ) y) and the parallel transport vector field V ( t ) = ( a ( t ) . (Indiana) Solution. Write a ( t ) = cosO(t). b ( t ) = sinO(t) where O ( t ) is the angle V ( t ) makes with the zaxis. the normal section is a circle x 2 + y2 = 4. = ( t . (a) From the hypothesis we have E = G = I F = 0 and hence ?I2 ' (b) For convenience. 3208 Consider the hyperboloid S of one sheet x 2 + y2 . k2 are the two principal curvatures of S a t P. Hint. 0 ) . (Indiana) Solution. Noticing the above expression of is equivalent to rik. (a) The Gauss curvature K of S at P is defined by K = k l k 2 . the corresponding principal curvature Icl = f. l ) .along the curve. 2 . Thus. by taking the outer unit normal vector field as the orientation of S. (a) Define the Gauss curvature K of S. Therefore. .vyO)) = ( 0 .177 y = 1. Then b V ( t )must be the solution to the following initial value problem (vl(O). (b) At P = (0.z 2 = 4.we have immediately 0 = t.1) = ( d ( t ) . denote the curve ( 2 . Show that V ( t )makes an angle t with the direction of the yaxis in the clockwise direction. v 2 ( t ) ) .0). where k l . writing a ( t ) = cos@(t).2. ( t ) ) = (wl(t). in the direction of the circle of latitude.b ( t ) = sin@(t).
x Pc(z) P. and s1 is the spherical cap encircled by C. In this case. with smooth unit normal vector field v. we take the lines of curvature as the parametric lines of M .178 whereas in the direction of the meridian.. 3209 Calculate the total geodesic curvature of a circle of radius r on a sphere of radius R > r.z 2 = 4.(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 .(x) terms of k l and k 2 .(z) dU1 au2 a a = (1 .& k 1 ) ( 1  Ek2) ax dul x # 0 au2 ax . := P . locally. Thus. : M R3defined by P. Show that the mapping P. If.& k i ) . thus the corresponding principal curvature k2 = Therefore. the Gauss curvature K = k l k 2 = a. ( M ) Pc(z) dU' a = (1 .2. ax aui i = 1. (Indiana) Solution. then by the Rodriques equation we have on P . express the principal curvatures of M . the total geodesic curvature is given by where C is the given circle. 1 i. the normal section is a component of the hyperbola y2 . ( M ) at P. By the GaussBonnet formula. 3210 Let M be a compact surface embedded in R3. in f + (Indiana) Solution.
B ( a . ( M ) .y) in lR2. which is obviously by conformal to that of the Euclidean plane B 2 .w ) for v .a ) . Take a positive number a and consider a square domain D in IR2.lco = (do ds and .. respectively. in lR2. D are A . we can express the metric of the Riemannian surface k2 ds2 = ex(y)(dx2 dy2). D ( . 3211 Let X : lR + l be a smooth function with compact support.a ) . Applying the GaussBonnet . the corresponding part of 2) is and the corresponding points of A . (Indiana) Solution. the surface is flat. w ) := ex(Y)(v. use the GaussBonnet Theorem to deduce that in fact. and consider R the Riemannian surface obtained by equipping lR2 with a metric g of the form g(v. B . Furthermore.6 2 d z ax O=O.case + 2 O=n. again by the Rodriques formula. . Then. we can see that the geodesic curvatures of the segments of coordinate curves i g and EE are +  5. C .y= a k.a . B . we see that the u1 and u2 parametric lines are the lines of curvature l . Using the Descartes coordinates (x.y=a 1 dlnE 2~ aY 1 dlnG a ax whereas both 52are geodesics of k2. C ( a . .)lR2. in P . C . immerses M into R3. and the corresponding principal curvatures are ~ k . by the Liouville formula.. Suppose that. D .sin 0 .1 and respectively. do 2dE 1 alnE aY coso+ 1 dlnG sin ' .179 provided that I E ~ is the reciprocal of neither k1 or k2 at x E M . which shows that the mapping P.a). Assuming the Gauss curvature K of this metric is everywhere nonnegative. w E T(. a ) respectively.a . whose vertices are A ( .1.
) is an asymptotic curve on M 2 . y=a y=a which combining the above equality shows that % = 0 in B1. z ) : z 2 0).parameterized by arclength. 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 . i. from the above integral equality we can obtain Because the number a is arbitrarily chosen. .e. R is flat. we have < 0 . letting a + 0. 2 3212 Let M 2 c R3be a smooth compact surface such that M 2 c ((2.180 formula to 5 c IR2. ii) Show that a’(s) is always a principal direction.y .Therefore. K 0. it leads to Besides. z ) : z = 0) is a smooth curve a ( s ) . Assume that M 2 n ((2. Thus. i) Show that a(. dy2  d2X namely ’ dY is a decreasing function..y. by the hypothesis K 2 0.
181 iii) Let's now drop the assumption that M 2 n ((2. R3 with a (Indiana) Solution.y. y. 3213 (a) Construct an example of a noncompact C surface in " sequence of closed geodesics {c. = .a ' ( s ) . 3. y. i) The hypotheses imply that along the curve a ( s ) .y. which means that the tangent vector of a ( s ) for every s is an asymptotic direction. along a ( s ) . (b) Show that this is not possible if the surface is compact. on S. Hence.7. Therefore.i.7 (b) Suppose S is a compact surface. + X Fig. from which the Rodriques equation says that a ' ( s ) is always a principal direction with the principal curvature 0. the unit normal vector field n ( s ) of M 2 is constant along a ( s ) . the plane {(z. iii) If we drop the assumption that M 2 n { ( z ly l z ) : t = 0) is a curve. z ) : z = 0) is tangent to M 2 .. z ) : t = 0) is a curve. a ( s ) is an asymptotic curve on M 2 . ii) Noticing the above fact.illustrated in Fig. we see that along a ( s ) .O . (a) Consider the following surface S of revolution generated by a C vi" brating curve C. What kind of a set could M 2 n {(z.3. = 0. then the set M 2 n {(z. If there exist closed geodesics c. t) : z = 0) be? (Indiana) Solution. a t which the plane {(z. rotating around the zaxis which is the asymptote of the curve.y.i such . Pi draw closed geodesics c. Let Pi denote the points where C has horizontal tangents.z) : z = 0) is a tangent plane of the surface M 2 . and length (ci) 0. Thus.z) : z = 0) consists of elliptic or parabolic points of M 2 . Then.} such that length (ri)+ 0.
T ( S ) sin 8. denotes the Christoffel symbol ( i . 0 ) + ( ~ ( s COSO. I = d s 2 + r2(s)d02 Eds2 + Gd02. Thus. 3214 Let ( r ( s ) . K d a = 2 r x ( M Z ) = 4 ~1( 9 ) _< 0. Show that the Gauss curvature of M must vanish somewhere on M . u 2 ) ..we have  a vs. = Denoting (s. Because S is compact.r ( s ) ae’ e r. 3215 Let M 2 c R3 be an embedded compact surface of genus 2 1. On this surface ) compute the covariant derivatives V+& and V A & in terms of and 86 (Indiana) Solution.e) = (ul. then by the GaussBonnet formula we have where a. (Indiana) Solution.a where as = l a a GI G2 r2.a . z ( s ) ) . z ( s ) ) a unit speed curve in B3 with T ( S ) > 0.denotes the domain encircled by u. O ..ae 89 v8.___ a G~ 8 rgl+ 8s 21ae 2~ as 2 ~ a .8~ 2 ~ a = .T ( S ) T ’ ( S ) . Consider be the surface of revolution (s. we have J I. we come t o a contradictory result 0 = 2 ~ . Ic = 1. By the GaussBonnet formula.. we may suppose R.j . when i is sufficiently large..2).E~ a +. setting i + 00 in the above equality.182 that length ( u i ) + 0.%+r2=+.  . is simply connected. The direct computation gives & 6. aas 22ae 2 a ~ e a y 2 .
F 2 d u d v we have immediately KdA = = T(U + TCOS u)dudv. there must be a point P with K ( P ) > 0. N = COSU. we claim that because of the compactness of M 2 . Otherwise.183 Firstly.u ) ~ + z 2 = r 2 around the zaxis: T = {(z. F = 0.w < 27r. ) ~ Parameterize this torus.w) = ( ( u + r c o s u ) c o s w . hence. . A straightforward computation gives the coefficients of its first and second fundamental forms E = r 2 . Thus obtained torusofrevolution T can be parameterized by sT X(u. Secondly. G = (u + TCOSU)~. compute its Gauss curvature function K. Finally. 0 < u.4u2(z2+ y2) = 0). we would have a contradiction. r s i n u ) . M = 0.F 2 cos u T(U+ TCOSU)' Noting that the area element on T is dA = J E G .M~ EG . and verify that KdA = 0 by explicit calculation. y. 3210 Consider the torusofrevolution T obtained by rotating the circle ( z . Therefore. (Indiana) Solution. the continuous function K must vanish somewhere on M 2 . L = T. we show that there exists another point Q with K(Q) < 0. there exists a domain U of P such that KIu > 0. by the continuity. 12T 12' dv cos udu = 0. Z) : (z2 y2 + + z 2 + u2  T ~ . by the connectedness. its Gauss curvature function K = L N . ( u + TCOSU). ( a + r c o s u ) s i n w .
we obtain the result that. (corresponding to the central angle 8 = $t in the domain D).. Let X o = translation of Xo along z ( t ) . we have J. then 47r 5 Is IKldv 5 Area ( S 2 ) . Consider the cone that is tangent t o the unit sphere S 2 along the curve z. q) 3218 Show that for any Riemannian metric on S2 with II K Gauss curvature. and characterize (with proof): i) All geodesics on the unit sphere S 2 c IR3. sin(t).8 = 27r  (i. 7 0 5 t < 27r T(L fils2. 2 2 E be a curve on S 2 c Ill3. 0 < 8 < AT. 1 cos(t). the oriented angle formed by the tangent vector with the parallel translation vector X ( t ) of XO is given by 27r . 2 3219 Define “geodesics”. for a displacement t of a moving point p along z starting from 0. if IKI Kdv = 2 x x ( S 2 ) = 47r. 5 1. “1 & . for any Riemannian metric on S 2 . . Thus. where K is the (Indiana) Solution. Because the parallel translation in the plane coincides with the normal Euclidean one. This cone minus one generator is isometric to an open set D c lR2 given in polar coordinates by 0 < p < +m. 5 1. Compute the parallel 21 a (Indiana) Solution. the area of S 2 is not less than 47r. By the GaussBonnet formula.184 3217 Let z(t)= (.
or a circle of latitude. ii) Since geodesics are intrinsic objects.k i = i. 3220 Give an example (e.2. On the contrary. Let a ( s ) be a curve on a surface M parameterized by arclength. (Indiana) Solution. z ) E R3 : z2 + y2 E 1). What is the usual topological hypothesis that prevents this problem? (Indiana) .. because along a . we have d ( S ) = Ic(s)N(s) where u1. every geodesic on the cylinder must be a helix. du u +Cr*. must be a great circle. E 0 or  d2ui ds2 . a is a geodesic of M if and only if along a .185 ii) All geodesics on the surface ((2. k.g. Along a . and kg is the geodesic curvature of a. then if we develop the cylinder to a plane. or a straight generating line. the principal normal vector N is parallel to n. Then. n is the unit normal vector field of M along a . draw a picture) to show that a connected surface can have two points which are not jointed by any geodesic. owing t o the uniqueness of the initial value problem every geodesic on M . i) On the unit sphere. Therefore.u2 are local coordinates on M . y. every geodesic must become a straight line. every great circle is a geodesic. that starts from a given point and is tangent t o a given direction.j ddsk = o ds j.
Hint. w) = (cos Jz u .cosv) 1 . cosh(t) cos(8).O. Let T be an interior point of the line segment p y .0. (Indiana) Solution.8) = (z. and prove that the catenoid. Then the surface K\{T} is a case in point. w. ii) Figure out the Clifford torus’ Gauss curvature function. i) From x “(sinu. “Completeness” is the usual topological hypothesis that prevents this problem. Let K be a plane and p .is minimal. A minimal surface is a surface with mean curvature H= 1E N . cosh(t) sin(8)) is a minimal surface in R3. obtained by rotating the graph of y = cosh(z) around the zaxis in R3.186 Solution. y two points in K.2 F M + G L = 0.sinv. sin u.(0. A 1 X “a . i) Compute the metric [gjj] on the Clifford torus for a coordinate chart of the indicated type. 3221 Define “minimal surface in R3”. cos where u and v are constrained to lie within intervals length < 2 ~ .O). iii) Deduce from the result of ii) that the Euler characteristic of a torus is zero.cosu. 2 EGF2 The straightforward calculation shows that the catenoid X ( Z . (Indiana) Solution. 3222 The “Clifford” torus in S3 can be parameterized using charts of the form 1 X ( u . Calculation is not necessary here.sin w).
22) at ( O .187 we easily have 911 = 922 = . O . a 2 jat ( O . O ) ) E B3 are the roots of the following equation . ] ii) Show that lR3 contains no compact embedded surface with strictly negative Gauss curvature at all its points..g. f ( O . ii) The Gauss equation implies that the Clifford torus’ Gauss curvature function K 0. O . iii) Using the GaussBonnet formula. we see that the Euler characteristic of a torus T 3223 Let f : IR2 + l be a smooth function with a critical point (e. O ) ) E R3 are the same as the eigenvalues of the Hessian matrix [ a 2 f / a z . i) The hypothesis means that $ and vanish at (0. the principal curvatures of the graph z = f(Z1. 2 1 g12 = 921 = 0. Then. a minimum R or maximum) at the origin 2 1 = 2 2 = 0. Hint. i) Show that the principal curvatures of the graph z = f(Z1. by straightforward calculation. . f ( 0 . we have Therefore. Look a t the “lowest” point on the surface.O).2) at ( O . O ) . (Indiana) Solution.
y. we have F .O.) Since the surface is above the plane T : z = zo and p is the common point of R and the surface. z ) for U . such that V. Thus the Gauss curvature of the surface a t p is not less than zero. we can find a neighborhood Up c U of p in lR3. y ) . setting F : Up t R3 by F ( z . y . if we take (O. Using the local coordinates ( u . we can express this embedding as ( u . ~ ) . Then.~ ( o = upn M . Y. c V of p .v) Y(U.~ ( 2 . ) Noticing that we know from the implicit function theorem that there exists a neighborhood V. By observing the normal section at p . (Indiana) Solution. yo. Since the inclusion map M + lR3 is an embedding. has smooth inverse { 2) =4ZlY) = v(2. such that on V.Y>. z ) = z . the embedding can be expressed by ( 2 . Y)). 1) as the unit normal of the surface at p.y). 3224 Let M be a 2dimensional manifold smoothly embedded in lR3 with unit normal n. )  . z ( ~ . where f(z. Therefore. b) Find F if M is the graph of a smooth function f : lR2 + lR.188 ii) Let p ( z 0 . y ( u . { Y= 2 = 2(u.Y) (z. zo) be the “lowest” point on the surface. v ) .y). such that V = U M .f ( z . it is easy to conclude that any normal curvature of the surface at p is greater than or equal t o zero.v ) for V and (z. (The existence of such a point follows from the compactness of the surface.. for each coordinate neighborhood V c M of p E M . = z ( u ( 2 . f ( 2 . = Up n M . there exists a neighborhood U of p in R3. and in terms of the local coordinates (z. Y) Y)). v ) ) . 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 . then T is the tangent plane of the surface a t p . ~ + ( z ( u .
189 b) If M is the graph of a smooth function f : R2 + lR. You may use without proof the fact that M lies on one side of each of its tangent planes. we can take Up = R3. (You may assume b) and c) if you wish. prove that is a diffeomorphism onto S2. Since nu x n. Xu x Xu Jacobi matrix the rank of the ( ffu ffv 3 /: ) = 2. = n ( u .. y ) . = K X . define q ( p ) as the end point of the unit exterior normal n(p) after parallel translating it to the origin of E3.. 211.+ S2 is a local diffeomorphism.then for each p E M .).P(. v) E C C M . K ( P ) d P = 4?T. where n(u.. x Xu and K > 0. . c) Assuming that 71 is onetoone. Assume that n is the exterior normal and that the Gauss curvature K of M is everywhere positive. For each p E M .>> # 0. . b) and c). 7) b) Prove that 7) is onetoone.v) is the unit normal at X ( u .f ( z . there exist a coordinate neighborhood C c M of p and a coordinate map h : C + U c B 2 . z ) : x 2 + y2 + z 2 = 1). d) Show that J. such that on U the Gauss map has the following expression q 0 h'(u.and F : R3+ R defined by F ( z . Assume that M is the boundary of a bounded convex open set. r(. z ) = z . .] a) Define the Gauss (or sphere) map 7) :M $ s2= {(z. We first prove that q : M . [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. y. y .) (Indiana) Solution.v).which is a diffeomorphism. 3225 Let M be a 2dimensional manifold smoothly embedded in R3 with unit normal n. a) For each p E M .> rank = (4% .
. q = (7o h') o h is a diffeomorphism. such that 7Ic. C Q ) so small that C p n CQ = 0. 7 is a local diffeomorphism. Now we prove that q : M t S2 is globally onetoone by leading to a contradiction. +  u = ( v I c p ) . q l c . it is easy to show that M is compact. we show that q : M S 2 is surjective. 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 ) . we see that the globally onetoone. fi have local expressions Since Kda > 0. noticing that differentiability is a local property. set .190 Thus. surjective local diffeomorphism is naturally a global diffeomorphism. Now take the inverse images of ~ ( C P ) ( C Q ) under nq q l ~ and q l cQ . CQ of P. closed and nonempty set of S 2 . in the neighborhood of h ( p ) . connected and oriented. because M is compact and S2 is a Hausdorff space. where da. Next.1 ( 7 ( C P )n 7(CQ)). Thus q ( M ) is an open. Because P # Q.hence 7 o h . on M . I . we arrive at a contradiction. = (vIc*)'(7(cP) v " rl(CQ)) Thus. Since 7 : M + S 2 is a local diffeomorphism. In the end. Q E M such that q ( P ) = q ( Q ) .Q respectively. Therefore. which implies q(M\U) = S2. The connectedness of S 2 implies that q ( M ) = S2. Thus.respectively. Then the GaussBonnet formula gives 1. Namely. 5. From the above argument we know that there are neighborhoods C p .l is a local diffeomorphism. (by noting K > 0). : CQ + ~ ( C Qare diffeomorphisrns. 7 is also a closed map. : Cp 7 ( C p ) . q is an open map. in M we can choose C ~ . Suppose that there are two distinct points P. u n v = 0 and q ( U ) = q(V). Besides. On the other hand. da = Hence Kda = 4n.
defined using the covariant derivative on R3. we can simplify its proof as follows. show that q : M t S2 is a local diffeomorphism. they are called “extrinsic quantities”. last local expression of the covariant derivative of M the involves the tangent vector field and the Christoffel symbols of M . intrinsic quantities are those that are invariant under isometric correspondence. discuss why the Gauss curvature of M is intrinsic. but extrinsic ones are not. b) Are the principal curvatures intrinsic? c) Discuss why the covariant derivative on M . Therefore. we know that the covering map q must be a homeomorphism and hence a global diffeomorphism. as the above. d) Assuming c).is intrinsic. c) Although the covariant derivative on M is defined by using the covariant derivative on R3. Firstly. a) The terminology “intrinsic quantities” means those geometrical quantities that are definable only by the first fundamental form of M and its derivatives. Since M is compact and S 2 is connected. Aw J do so . (Indiana) Solution.191 Remark. we can show that the Gauss curvature of M a t p can be expressed by K ( p ) = lim D+P . For example. b) The principal curvatures axe not intrinsic. this problem is the famous Hadamard theorem. and M c R3 is connected and hence path connected. Using the Hopf’s rotation index theorem and the GaussBonnet formula. 3220 Let M be a 2dimensional manifold smoothly embedded in R3with unit normal n. they are extrinsic. then the local diffeomorphism q : M + S2 is also a covering map. In fact. a) Explain what is meant by intrinsic and extrinsic quantities on M . d) For each p E M . Using the theory of covering map. let C be a simple closed curve encircling a simply connected domain D where p lies. Otherwise. In other words. Further. consider a plane and a cylinder. Let Aw denote the angle variance caused by a tangent vector after parallel translating it around C once. it is intrinsic. because S 2 is simply connected.
the uparameter curve X ( u . a) When n = 3. (Indiana) .) (Indiana) z I I Fig. 1)2+ N such that for each fixed wo E (0. Thus the Euler characteristic number x ( M 2 )= 0. we get a surface of revolution M 2 c R3. wo) is a geodesic in N. Therefore. l). b) Show it for arbitrary n.8 Solution. too. on M . The surface M 2 of revolution generated by the curve y is homeomorphic t o a section of a cylindrical surface. by the famous GaussBonnet formula. 3227 By revolving the curve y sketched below around the zaxis.3. 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. show that the Gauss curvature of a ruled surface in Rfl is nowhere positive. Besides. Compute JM K d A . the boundary of M 2 consists of two circles which are just geodesics of M 2 . where K is the Gauss curvature and dA the area form. (Make sure to justify your answer. we have immediately K d A = 0. then the Gauss curvature is intrinsic.192 Since parallel translation is intrinsic. because along these circles the normal vector of M 2 is parallel to the principal normal vector of the two circles respectively.
by a routine work. we obtain the desired result. Thus. EG.l(w)) Then. ) dU2 f0.Z’(w))2 = la’(w)x I’(w)I2 2 0. However. .(a’(v). a) Since every geodesic in N = R3 is a straight line.F2  b) Similar to the above. we see F 0. which shows that the Gauss curvature of the surface K= LN . we have Noting that 1l’(v)12 . l ) is the arclength of the curve a ( v ) .w) = a(w)+(uwhere Z(v) is the direction of the geodesic corresponding to w E ( 0 . that the first fundamental form of C2 is ds2 = d U 2 + (1+ 2u(a’(w). 1Z(v)1 SE 1.193 Solution. we may assume that v E ( 0 . for convenience. we can easily deduce that the first coefficient of the second fundamental form of i)l(w). suppose C2 can be characterized locally by where. through computation. C2 d 2X L = (n. Since the Gauss equation says * dz’ it suffices to show that 2 0. and (a’(u). l ) . then C2 can be characterized locally by X ( u .Z’(v)) + K = U2II’(V)12)dV2.F2 EG.iw2 <o.
Y . ) = R ( X .W ) .( L X . R(fX. one can show that V p E M " .W on M " . 2. Y ) . E T'(M"). Y . Then for any C function f on M " . W.Y . Thus.Z ) L X . If ( 2 . Y . i) Let { X . Y . Y .V[X.X .X . R ( X . Therefore. (Indiana) Solution. using the properties of covariant differentiation " and of the inner product.Y. Y ) . X .Y ) Z = ( L Y .Z ) L Y . which proves that K ( R )depends only on R . then R ( X .194 SECTION 3 DIFFERENTIAL GEOMETRY OF MANIFOLD 3301 Let M" be a Riemannian manifold and R a 2plane in T. define the Riemannian curvature operator by R ( X . Using this or any other valid method.Y]Z and the Riemannian curvature tensor field by R ( X . f Z . z.. is dependent only on X. M. Z .W ) .2.V y V x Z . Y ) = R ( X . we can conclude from straighforward computation that ( 1) + + is an orthogonal matrix. we can easily have R ( X .Y. y.y.Y ) Z . fW) = f R ( X .R ( X . W ) W = R ( X . where R( ) is the Riemannian curvature tensor field of M " . ii) For any vector fields X . ? = cX d Y . compute all sectional curvatures of the sphere {z E B4: 121 = 3). fY. in terms of local coordinate frame field. where L is the Weingarten map. .2.W ) = ( R ( X .respectively. iii) Recall that if M" c En+' is a hypersurface. and R( ) is a 4th order covariant tensor  z.. Y ) is another orthonormal basis of R . Z. then 2 = aX by. Y . Z . W)I. Use this basis to define the sectional curvature K ( R ) and show that it depends only on R and not the particular basis chosen. i) The sectional curvature is defined by K ( R ) = . W ) = R ( X . Y } be an orthonormal basis of R . Noticing that field. ii) Define the Riemann curvature tensor in terms of covariant differentiation..Y)Z = V x V y Z . Explain why it is a tensor.
X / 3 . Justify the following formula: . Therefore. " iii) For the sphere M n = { z E lR4. we can regard the position vector field 2 as the normal vector field of M n .1zl= 3 } . Y ) X . Hence. one can easily show that V e i z= e i . R( ) thus defined is a C tensor field. { e i } be a local orthonormal Let frame field about 3 (as a point) of M n . the inner product and covariant differentiation are all C". X ) ( L Y .Then by the original definition of : covariant differentiation.( L Y . where V denotes the covariant differentiation in lR4.195 R( ) I p is a welldefined 4th order multilinear function. Hence. Y ) = ( L X . Y ).V x g = . X ) ( L X . Besides. 3302 Let C = {z E lR3 : 0 5 3ci 5 1) be the unit cube in B3. if { X . The image " F(C) is then a compact Riemannian submanifold of lR4 with boundary and therefore has a volume. the Weingarten map is L : X F+ L ( X ) = . Suppose F : lR3 + lR4 is a 11 C immersion in some neighborhood of C. Y )= 1/9.    K(7r) = . Y } is any orthonormal basis of an arbitrary 2plane T in T p ( M n ) .( R ( X .
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
a ax2 ax3 x2=+22.IIXII = 1.O .(23 2 4 ) . is a smooth compact embedded submanifold of IR6 and be identified with the unit tangent bundle of 5 ' . ( X .we see that is also a threedimensional submanifold of lR4. 0)) is an open submanifold of B 4 . X2 on lR4 be defined by x1=  a + 21.Y)).21 1 is a regular submanifold of R4\{(0. Xl(P).llY1I2.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. O .O.y E lR3} and define a map F : R6+ IR3 by F ( x .X2(P) E T. Noting that E4\{(0.f 2 . & + 3321 Let M = { ( X . 0)) with dimension 3. Y ) i) Show that M explain how M can ii) Show that M B3. is orientable. there is not.Y E Solution.y) : x. llYll = 1. we can set f = 21x2 . for example. (Indiana) :X. It is easy to verify that the Jacobi matrix of the C map F " . The reason is that the bracket [Xl. ii) Yes. a i) No. i) Identify lR6 with {(x. Y) = ( f l . O.X2] = . Y ) = 01. 3320 Let vector fields XI. a axl a ax4 i) Is there a 2dimensional submanifold M 2 of lR4 such that for each p E M2.asJ does not satisfy the Frobenius condition. f3) = (11x112. (x.
Next.e.dol x 4) form a covering of coherently oriented . if we regard z E S2 and y E T. 3322 Let F : M 4 N be a local isometry between connected Riemannian manifolds M and N . Naturally. F ( M ) is also closed in N . F ( M ) is open in N .. x I.. z has the local coordinates (ui. By using identification of M with the unit tangent bundle of S 2 . 1. since F maps every geodesic of M into a geodesic of N . F’(l.ui) U. y) to 0. y) E M . For every x’ E N . Therefore. x I. 0. Hence.212 has constant rank three when fl = 11z1(2 1. Therefore. where I. = (0. If y is a limit point of F ( M ) in N . F is surjective.y) M . and y is uniquely determined by the oriented angle 0.(S2) with E 1)yI)= 1. # J ~ ) }of coherently oriented coordinate neighborhoods. : B’(S) + Bi is a diffeomorphism.E . E M . Besides. 4 x 4.& x $ is . i. the connectedness of N implies that F ( M ) = N . and the image of its end point under F must be y.. then the transition function . then there is a point z E M such that there exists a geodesic in N connecting F ( z ) and y. we show that F is a covering map. the Hopffinow theorem means that N is complete. Besides. . ( 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. y) E ) with E being a suitable positive real number and 4 is the map from (2. M is orientable.y)II = (llzllz Ilyl12)i = fi which means M is bounded in IR6. then we can identify M with the unit tangent bundle of S 2 . for every (z.O) = M is a closed embedded submanifold of B6. S2 has a covering { ( U .) neighborhoods. f2 = ( ( ~ ( =~ 1 and f3 = (z. If (U.) n (Up x I p . so is N and F is a covering map. ii) Because S 2 is orientable. Thus.. . too.y) = = 1 0. (Indiana) Solution. for every (z. a t z E from 4 to the unit tangent vector y E T . #Jp x $ p ) # 0. since II(z.. Thus. .. where B’(6) = {v E .. take S > 0 so small that exp. M must be compact. has the following J acobian + + which means that { ( V a x I. ) au. Show that if M is complete. a coordinate neighborhood of (2. Because F is a local isometry. ( U .
Besides. Since F is a local isometry..yp are the lifts of y through z.} c M and set Ba(S) = {v E T.1] + M such that F ( T ( t ) ) = y ( t ) . Hence F = exp. On the + other hand. f ( N ) : IvI < S} and Bi = {y’ E N : d ( y ’ .) = y = F ( y p ) . L ( y ) = L(y) < S means that z = T ( 0 ) E B r . then BF n Bf = 8. Firstly. z ’ ) < 5). In fact.. uBr c Fl(B. i. expXm also a diffeomorphism.) c U B r . V t . F o expZm = expzf odF is a diffeomorphism and hence exp.“ + B. o d F o (exp.). we have the following commutative diagram and expZe is surjective. Besides. Fl(B.l ( B i ) = U B ? . . exp. So. z = y a ( l ) = yp(1) = zcp. Hence F(y(1)) = z’ and y(1) = z for some a . Therefore. Since .” n B f . Because both F : B. are isometric. Thus the claim is proved. Otherwise. z zp respectively. we have F(y. Fl(z’) is discrete. we claim : that Bi is an admissible neighborhood of z’ and F is a covering map.. Denote Fl(z’) = {z. Then. In particular. there exists a geodesic y : [0. Thirdly. F : BO + Bi is a diffeomorphism.((M) lvl < 6) and Br = {y E M : d ( y . contradicts a # p. + Let F : M + M be an isometry of a Riemannian manifold M . and F : B p 6 B. we know that d F . then there exist unique geodesics y. then 0 there is a unique geodesic y : [0. .) 0 a is obvious. we say that for any a .. 2’.213 T . Secondly. The uniqueness implies ya = yp. ya. such that y(0) = F ( z ) and y(1) = 2’. which . yp connecting z with .)’ is a diffeomorphism. Let y be the unique geodesic in Bi connecting F ( z ) and .8. are diffeomorphisms. if z E F’(B. In other words.e. M is complete..1] B. is is an immersion. we claim that F . if there is z E B. z a ) < a}. i) Show that each component of X = {z E M : F ( z ) = z } is an embedded totally geodesic submanifold of M . we claim that if Q # . Since F is a local isometry.
y is a geodesic of M .. y .. we show that X has submanifold structures.(tv) connecting z and y.1] 4 M be the unique shortest geodesic y(t) = exp.. Since z. that F is an isometry implies F(y) = y. Define V = {v E g T . Thus.1] 4 M be defined by y ( t ) = exp.(V n B(6)).e. Hence. F ( z . ~ . z ) = (z. v.214 Hint. For z E X c M .On the other hand. Thus the uniqueness implies F ( y ) = y. and F is an isometry. d F ( v ) = v. d F ( j ( 0 ) ) = j(O). X is totally geodesic. X n B c exp. suppose that v E V n B ( 6 ) and g y = exp. Next.z> E IR3 : 2 = 0.y E X .z ) . In particular.) E IR3 : y > 0 . we can assert that X has submanifold structures. Besides. F ( y ) = F(y(1)) = y(1) = y. = 01 u {(z..z) E ~3 : > o .e.(V n B ( 6 ) )c X n Ba. in a neighborhood of S O . Let the geodesic y : [0.i. and x = {(2.z) ~ E 1123 : =o .(tv). <~ o . b ) . i. F ( < ) = <. and F be a reflection with respect to the plane z = 0.(VnB(G)) is obviously a submanifold of M . Because so is arbitrarily chosen.y < 0}. (Indiana) Solution. b ) + X parameterized by arclength is also a geodesic of M . let <(s) be a geodesic of M such that <(so) = SO). ii) Give an example in which the components of X have different dimensions. namely.i. lies in X . ( M ) : IvI < 6) and B = {y E M : d ( z . Then M is a 2dimensional manifold. . Therefore.y. Use exponential coordinates. Thus. i. = y.e. From d F ( v ) = v follows d F ( T ( 0 ) )= j ( 0 ) . For any SO E ( a . we first assume that y = X n & and v E B(6) such that exp. : B(6) + B is a diffeomorphism. set B(6) = {v E T . = 0). Then. we show that every geodesic y : ( a . Let y : [0.(V n B(6)). ii) Let < < < < M = {(qy. v E V which means that y E exp. In order to prove the claim. d F ( ( ( s 0 ) )= ((so) and F is an isometry. because exp. i(so) = ? ( S O ) . Then we claim that XnBa = exp. then F ( < ) and are two geodesics of M which satisfy the same initial conditions. where a 6 is so small that exp. y..y. Since F ( < ( s o ) ) = (‘(SO). exp. If this is proved. V is a subspace of T.y.e. F is an isometry. In particular. is naturally a geodesic of X.v = y. i) First. F ( y ) is also a shortest geodesic connecting F ( z ) = z and F ( y ) = y.(M). z = 0) u {(z. Therefore. which means that y E X n Bg. y ) < 6}. ( M ) : d F ( v ) = v}.(VnB(6)).
Hence.. observe that Z'(S'.e. B ( X . := {z I/ u. 3325 Let X denote a submanifold of Euclidean space E n .d) = {df I f E Coo(S1. because 6 in usual sense is not a globally welldefined function on S').d) = Z'(S'. Do so directly. Then is a nonvanishing vector field on S1. since there are no nonvanishing kforms on S'. Therefore.d)/B'(S'.CdB is exact. set and E x. since Bo(S1. A'(S')).d) = (Cd6 I C E Rl}Z R1. n is globally welldefined on S1.A'(S1)).d)= 0 and S' is connected.. w . and it is not exact.R1)}. X . I ' For H k ( S 1 . k> 1. denoting we see that w . i.R1)df = 0) E B . For every w = g(O)d6 E C"(S'.Let d6 be its dual nonvanishing 1form on S'(Caution: Here d6 is only a formal symbol. = 0 and hence H k ( S 1 . d ) .Cd6 = do. E ) := {y E En: Iy . Let 6 be the polar coordinate characterizing S1. Le.v E N . without citing the de Rham Theorem.d) = {f E Coo(S1. IvI < E } . Besides. define a function & Because R(0) = R(27r). we d) have Zk(S1.215 3324 Compute the de Rham cohomology groups of the circle S'. d ) = Zo(S'. B1(S1. d )= 0.x + :2 .d) = Cm(S1.z1 < E for some z E X } . (Indiana) Solution. then H * ( S ' . H1(S1.
E ) .l=l where g . z m .j=l m k.X = + B ( X . . Y ) g ( X . e. then U. E ) . ii) Show that the expression (**) above is equivalent to where V denotes the LeviCivita connection for ( M . Z ( g ( X . (Indiana) Solution. and 2 = 2 ( z . Let X be an open line segment in E 2 . &). j ( z ) = g (&. ) Can you give conditions which imply equality? (Indiana) Solution. i. can show that V E X is a proper subset of B ( X .e. Then considering the boundary of X .  a . If X has no boundary. E )for all E .g ) . 3326 Consider a Riemannian manifold ( M . i. L Z Y ) = + (**) for all vector fields X and Y on M . we have Lzg = 0 . 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 . t ):= 2 ( z . In local coordinates (zl. Show that the two are not generally equal.g.X c B ( X . i) Show. the end points. either compact or complete.9 ) . Here L z denotes the Lie derivative along z. let the 1parameter group of isornetries generated by a vector field 2 = C zi& i=l m be expressed by 2 = 2(z1...216 Show that U.. (Consider examples of 1dimensional submanifold in E 2 .that when 2 is Killing. t ) such that m i. g ) .y > > S(JhX.E l .e. z m ) of ( M . t ) satisfies .
Conclude that ( X . Because the flow X t of X for every t is an isometry.i = 0.Y)) g ( V z X . + .j Xj. = Y. Y ) and ( Y . j = 1 . i. X ) . Y )are constant on M . X I . i) Show that the integral curves of X are curves of constant geodesic curvature. or equivalently g (V a Z . Noting that the LeviCivita connection V satisfies Z ( g ( X . (Indiana) Solution. i) For every p E M .i + S i k zkj $9 ( l .o X t . we obtain which can be written as k gkjz. Assume that the flows Xtand Yt are isometries of M for all t. ii) Assume that X and Y are linearly independent at all 2 E M 2and that their flows commute X t o Y. take a local coordinate neighborhood about p such that the coordinate curves are the integral curves of X and their orthogonal trajectories. or equivalently. + 3327 Let M 2 be a connected Riemannian manifold and X . we easily obtain (w). we may assume that X = X I & and g 1 2 = (&. Y ) g ( X .i V 0. Y complete vector fields on M . &)= 0. From this follows what we desire in ii). V Z Y ) = and the Lie derivative satisfies L z X = [Z. ( X . 2 .217 Differentiating the obtained equality with respect to t and then setting t = 0. l) j e er3 Z ) =o. the vector field X should satisfy the Killing equation Xi. Namely.
that is equivalent to lay2 d ax1 d [ X .Y ) = 922 = const. = const. we can compute the geodesic curvature of the xlcoordinate curve as follows: where 0 = 0. make the Hence If we still adopt the original notations. dX1 &. Ic. i. if a adopting the original notations. along every integral curve of X . i.. Now.218 Taking i = 1.X ) = gll = const.Y ) = g12 = const. about p . =o 8x1 dxi . by the corresponding Killing equations. Noting the expressions of X and Y . we immediately have x ax2 Therefore.8x1 a x 2 dx2 dXl = O . using the Liouville formula. we can make a suitable coordinate transformation and then.. X = Y = G.0 . +. Then. = Y. 2 . (Y. ax = 0. Again. Then. which means that the metric of M 2about p can be written as ds2 = g11(x2)(dx1)2 &. take the integral curves of X and Y as the x1 and x2 coordinate curves. Y ]= . we obtain a922 agll. taking i = j = 1 and i = j = 2. ii) Analogously. X1 = X1(xl).Y 2 . then the vector field X = from the corresponding killing equation. Therefore. j .g 2 2 ( ~ ~ ) ( d ~ ~ ) ~ . we obtain 911% following coordinate transformation = 0. ak = 0. we can obtain = 0. ( X . i. we have locally X = X I & . a ( X . . all gij's are constants.. o X t . Ic = 1 . we obtain d=Yol2. j = 2. respectively. Y = Y2&.e. i..e.e. Because their flows are commutative X t o Y.
through direct claculation.... . Using the definition of the Laplacian here.219 3328 Let M be a compact Riemannian manifold without boundary.e3) = S i j . f ] .n. . . define O f E X ( M ) and Af E C “ ( M ) as follows: At any p E M . . j = 1. i .e j ) = (e: . e i ) f l = C [ e r ( e f f ) ( v e .( v e i e i ) f I = i i Verify first that V f and A f are well defined (i. we can easily obtain i i C [ e i ( e i f ). .C [ e i ( e i f ). Noting (ei . j for suitable functions a!. we have j Using these equalities and the properties of Riemannian connection. . they do not depend on the choice of orthonormal frame) and then show that (Indiana) Solution.. e i } be another orthonormal frame field around p . } around p and then define (Vf)(p) = C ( e i f ) ( e i ) and (Af>(p> . . e : ) . .e. . Then we Let { e .( v e . may suppose that e i = C a ! e r i = l . For any f E Cm(M).f A f. Then Green’s theorem implies . we have div( f V f ) = llVf112 . i i Hence Of and Af are all well defined. e .1n  e . . choose an orthonormal frame field { e l .
Using the theorem of closed regular submanifolds proved by rank theorem. we can easily show that &. Let X be a smooth vector field on M . b 2 0.2 : = b .f3 are two roots of the equation X 2 . 3. Analogously. 2.o is a 0dimensional manifold. b # 0. Therefore. = 2}. is Sa. x : + 2 : = b or 2. x 3 = 2 4 = 0). If a = b = 0 . Denote a = xf + x i and . then when b2 .b is a 1dimensional submanifold of R 4 .220 3320 For what a . 1. i .4a > 0 and when b2 . (i) Show by an example that d F ( X ) may not be a vector field on N . b # 0. we can show that they are all 2dimensional submanifolds of R4. Assume that F is onto. then = { Z E B4 1 x1 = 2 2 = 0 . b2 .b a manifold? Explain. Then C Y ~ a and CY + p = b = means that a and .f3 = x i + x z .bX + a = 0. .4a = 0 Sa. then 21 = 2 2 = 2 3 = 2 4 = 0.4a 2 0 is the prerequisite condition. Thus S0. If a # 0. + 2.b = ( 2 E E4I b + 2. 3330 " " Let F : M + N be a C map between two C manifolds. If a = 0. = 2. (Indiana) Solution.
along a . which means that F takes integral curves of X into integral curves of Y . . (Indiana) Solution. 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 . we have da ($) = X .N = {x : x E E}.x n ) be the local coordinates about p and set p : (xk. XJ 1 x 3 0 (. by . d F ( X ) is not well defined in every point of N . as a vector field. (ii) Let a(. Y are related as X and Y in (ii) above.. Hence. 1 and X z . # then.y) : z. d(F o a) (i) = ( d F o da) (g) = d F ( X ) = Y. if y1 # yz.i 4.). z (iii) Suppose XI. = (xz + y 2 ) & . Y.y) = x. (Indiana) Solution. Then. Then.yz))... 3331 Let M" be a Riemannian manifold.y E IR}. (iii) Using local coordinates. . . Show that d F ( [ X l . by direct computation. x2. Therefore. x and F : M + N be defined by F(z.y l ) ) # d F ( X ( z . Let (xl.. one will obtain the desired equality. X Z l )= [Yl.221 (ii) Suppose Y = d F ( X ) is a smooth vector field on N . we have d F ( X ( x . i. Show that whenever f : M + lR is a smooth function.) be an arbitrary integral curve of X . y(t) has the following expression  21 1 t.YZ]. Show that F takes integral curves of X into integral curves of Y.e. If y ( t ) is the ith coordinate curve that passes p . (i) Let M = {(x.
and define u : S n w 1 E n x n the map + as sending x = (21.222 we have the expression where gij's are the components of the matrix [gijl' = [(&.j b i j .e. Suppose that . ) Jz i) Show that u maps S"' into the sphere of radius centered a t the origin in R" n. is (Indiana) Solution. It is easy to verify that the above expression of V f (p)is independent of the choice of local coordinates. Then C z = 1. the pullback via g of the dotproduct metric defined above on RnX" the standard one on S"').) in S"l to the symmetric matrix ~ ( x= I [ z i z j ] . and r : an RnX" be defined by (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. Let Sn' be the unit sphere in IR". .j]and B = [ b i j ] is given by n n ( A . 3332 The space ELnxn of n x n real matrices forms an n2 dimensional Euclidean space. n i) Let x = (zl. 2. 5 i=1 1 " (u(x). ii) Prove that cr is a local isometry (i.u(x))= z. Hence .) + [yij] = &[zizj].B ) := i=l j=1 a.j=1 c = 1 5. in which the dot product between A = [a..z. . ... z i. a .. ..zn) E S"l.
Then we have n n i=l i= 1 Therefore Hence we have (dn(v).e. Let M be the Riemannian manifold obtained by equipping B" with a metric conformal t o its usual one.j=1 (dXj + VjXi)(WiXj +WjXi) = 3333 c n i=l viwi = (v.223 belong to Tx(Sn'). 2 . c n i. and S i j is the Kronecker delta.dn(w))=  1 . j . k E { 1 .. i) Show that for arbitrary indices i. where f : IR" IR is a smooth function. n} = & f ii) Show that when n = 2.w).e2 (Indiana) . Let ei = denote the standard coordinate basis vector fields. i. a metric of the form [sij] = e 2 f [ S i j ] . the sectional curvature of M along an plane a t p is given by e1. .
(Indiana) Solution.x. for all s. for all s. Y ]= 0. e2 plane at p is just the Gauss curvature of M a t p . i) Show that X .. Y ]= L X Y = lim [Y . Observe that for a diffeomorphism F : M +M . if and only if F o yt = yt o F . i) Set From gij = e 2 f & it follows that ii) The sectional curvature of M along an e l . W E R. then Y is X. t E R. F. = yt o X.. Thus 1 [ X . F = 0.Since X .t E IR implies [ X .e.224 Solution. o yt = yt o X . . ii) Prove the converse to i). i. Yt be the flows induced by them. Noting that E = G = e 2 f . Then . is a diffeomorphism for each s.Y ]= 0.Y]= 0.invariant.Y = Y . Y be complete vector fields on a manifold M and let X . o r. i) Suppose that X. from the Gauss equation we obtain 3334 Let X . 3 4 0 s ii) Now suppose that [ X . the complete vector field Y is Finvariant.
we have (Indiana) Solution. = yt O X . namely. . . show that for all w E LG1. Let ( z l . respectively. i=l axi we2 = a E w e aye ' n e=1 . .for all s . we have for all s E R. . = Since f is arbitrary. X. Therefore. Y = Y .invariant. Locally. Y is X. namely. . Note that exptw. . 4(exp tv) = exp tw.z ) and (y'. Since 4 is a Lie group homomorphism. then 4(exptv) is also a 1parameter subgroup of G2 generated by a suitable w E LG2. t E l is a 1parameter subgroup of G1 generated by R ' v E LG1. we obtain X . 4(exp(u)) = exp(+*v). Assuming that wel " a =cvi.225 Hence for each p E M and any f E C"(p). . ox Hence 3335 If 4 : G1 + G2 is a Lie group homomorhism. ( X s * Y ) p f ( X o * Y ) p f= Y p f . 4 can be expressed by and exp tv and exp tw can be denoted by respectively. t E R. yn) be local coordinates of G1 and G2 about " the identities el and e2.
.v] = 0. b) Construct the following linear algebraic equations about y 1 . y2. i. a) Direct calculation gives [r. Taking t = 1 completes the proof.e. 2 2 .22.21. 2 3 . a 1 t=O aY" Furthermore.24y3 + 23Y4 = 0.24). 3336 Consider the linearly independent vector fields r and v on whose values at x = ( 2 1 . 2 4 ) E B4 are given by rx : U := B4\0 = = (21. Hence 4(exptw) = expt4. Thus the Frobenius theorem guarantees that the rank2 distribution defined by r and v in U is completely integrable. we have d $ ( v ) = w. 24. d*w = w. which are equivalent to .: 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.23) v.226 we have t=O t=o Therefore. y 3 and Y4 I Z l Y l + 22y2 22y1+ + 23Y3 + 24Y4 = 0. = 2 (Z) (= I =we. ZlYZ . by left translation. (22.w. 23..
e. a) Show that any connection on G is competely determined by its effect on the frame field.n.. 2224 p = .)).i. we may assume 2 1 # 0. 0.2223. Thus the F’robenius theorem tells us that this distribution is not completely integrable. By a long but straightforward calculation. and the framefield is leftinvariant. .:).[ e i . 4 ) E B 4 \ 0 = U. 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+2224.227 Because ( 2 1 . ~ = 2(2: ] zi)v which does not belong to the distribution defined by CY and p.s+.2124 2124 2224 + 2123 . i. j = 1.n. k 2.j = 1.. ~ 2 . . ( : . by the vector fields Veiej. we obtain 2123 2223 +2224 . + 3337 Let G be a Riemannian manifold with a global framefield {ei}y=l. . + (. ). 2123. (2122 222322124. ( 7 )) by setting. b) Show that when G is a Lie group with a biinvariant metric ( . o). we characterize the LeviCivita connection on (G.2223 Setting and respectively. we have [ c Y . Therefore. e j ] 2 for all i.we have  n n . 2 32. a) For arbitrary smooth vector fields X and Y on G. 1 Veiej = .without loss of generality. (Indiana) Solution.
ej . i. = 0.V e j e i= [ e i . 2 1 . we obtain V e i e j= . e j ] .x =  t=O Besides.. b) For every left invariant vector field X on G. where e is the unit element of G. = X .228 Then. Especially. Thus. 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. we know that V x X = 0 is valid everywhere. + + + On the other hand. e j ] . we can well define x'e. Noting that g ( t ) is a geodesic of G and Xg(tl = we have obviously 9 y. if X = ei e j . motivated by the properties of connection. because the metric of G is biinvariant. . vx.(g)ej + C zigveiej. the LeviCivita connection V satisfies V. then V e i + e j ( e i e j ) = 0 implies that Veiej Veje.[ e i .
Part IV Real Analysis .
.
Since c y ) n . ( z ) 58 and (3) z = n=l .. 3rn 2 n : p .+ kn1 n = l kj#2..+ . p n ( z ) # 2 and p n ( z ) # 3) and B = {z E [ O . A = {z E [O.1 kn1 lo*1 + 4 1uR) and B = C u [$+. E LO. 10n 10 +. p i ( z ) # 2 and pi(. l ) Then 00 1 3n. V i < n .1) I 3 n . For any z E [0. 2 kl .pn(z)= 2 .) # 3). * p s Let = (1) U{z u{ ..+ . (Stanford) Solution. (2) Vn. V i < n . p n ( z ) = 2 .2 .231 SECTION 1 MEASURABLITY AND MEASURE 4101 Let S c [0. 1) I h P n ( z ) # 2 and P() n .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.. Prove that S is Lebesgue measurable and find its measure.3 10n1 kl + . 10" l o + 3 It follows that both A and B are measurable and therefore is S.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. # 3) E [0. p i ( z ) # 2 and pi(z) # 3). 1) I V n . kn1 n=lki#2. .3 101 + .k .
a w €P. i. If m ( E ) = 0 then m(S1) = 0. which then implies that T = s since 1 < r .y m o d Z . .y E IR and [ ] denoting the class in S1. s E [0.l)W + m(S1) = T€ C [O.y E &.e.232 and we have 1 m ( S )= m(A) m ( B ) = .where T E [0.  E = {F Show that E is not measurable. ( StanfoTd) Solution.. a contradiction. Consider on S1the equivalence relation: [z] [y] # z . Otherwise m ( S 1 ) = contradiction.V[ E S'/ }. I E <. too. + If T . + 2 4102 Define S1= IR/Z endowed with the natural Lebesgue measure. Let m denote the natural Lebesgue measure on S1such that m(S1) = 1. 1) n&} is a partition of S1. [z] ++[. For any [z] E S1there is by the construction of E a unique element [y] E E such that z .y E &. for z.s < 1. S1is an abelian group under the binary operation ([z].1) n& is such that r G z . It follows that z [z] = [y] by the construction of E . [y] E E y mod& and therefore such that [z] [T] = [y] [ s ] . 1) m(E+[r]) = C T m(E). It follows that { E + [TI 1 T E [0.I. For . Thus [r] = [ s ] .1) n& are such that ( E [TI) n ( E [ s ] ) # 0 there are [z]. Then [z] = [y] [r]. So we conclude that S1= U ( E + [TI).1)W 00. [y]) H [z + y ] and the inverse operation. If E is measurable then + + + T €[O. each ( E S1/ choose a representative f E S1. and let E c S1be the set of these points.
233 4103 Let X be a set and D c ' P ( X ) . We will show the equality.A = A n A. Let T = d ( A . Let U = {Z E I d ( z . 2 Prove or disprove that p * ( AU B ) = p * ( A ) +p*(B). x C R. By equality (2). Prove that a = R.yI I z E A . A )< T) 2 . Also we have since by the equality (1) the former is a ring containing 2).B ) > 0. D c a c P ( X ) such that R is closed under the following operations: (i) finite disjoint unions. y E B } > 0. B subsets of I such R that inf{ 1 . We have since the former is a ring containing D. R = R since for any A E R. Obviously. F'uthermore. (Iowa) Solution. B c A. (ii) differences A\B. let a be the smallest system.'D closed under finite intersection. Denote ' by R the ring generated by D. (Iowa) Solution. 4104 Let p* be the Lebesgue outer measure on IR and A .
that B is the Bore1 aalgebra. (Iowa) Solution. finitely additive measure on B. B C W2 2 V } i inf(p(W1) 1 open. A 5 cU } ~1 WI +inf{p(W2) I W2 open. then p(UBi) = C p ( B i ) . Let A. is a decreasing sequence of sets with empty intersection then lim p(An) = 0. p ( B ) = sup{p(K) I K 5 B and K is compact}. p ) with a finite. a) The sufficiency.A U B inf(p(W1) C W} : inf{p(W) I W o p e n . 4105 a) Consider a measurable space (X. that is for each B E B.Y where the equality (1) follows from the following equality GR. Suppose moreover that p is regular. Then n=l n An = 0. positive. Finite additivity means that whenever { B i } is a finite collection of mutually disjoint measurable sets. A U B C W C U U V ) E = + p ( W 2 ) I W open. inf(X+Y)=infX+infY.234 Then U and V axe disjoint open sets containing A and B respectively. X. We have p * ( Au B ) = = = inf{p(W) I W open. T00 b) Now that suppose X is a locally compact Hausdorff space. and that p is a finite. Let {B. B C W2 C V} P * ( A )+ P * ( B ) . Prove that p is countably additive. Prove that p is countably additive if and only if it satisfies the following condition: If A. positive.} be countably many measurable sets which are 00 mutually disjoint. finitely additive measure p. We have \ o o m / n . = i=n+l U Bi. A C WI C U .
b) If /I is not countably additive.} of measurable sets with empty intersection such that .1] + B. n For each n there is a compact Kn contained in A .) = inf p ( A n ) > 0. show that m(f(E)) = 0. let E C { z I f'(z) exists}. If m ( E ) = 0.too lim p(A. Set Fn = I z E F . 1 where M is any positive number. Denote F ={ .f(z>II MIY. So n K. lf'(4I < MI. by a) there is a decreasing sequence {A. 03 # 0. which contradicts the fact that n=l 4106 For f : [0.21 if I Y . If(y) . The necessity is obvious. 12 E E . . 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 K1.235 Therefore Y is a measure.21 < 1. It suffices to prove m ( f ( F ) )= 0. ( IndianaPurdue) Solution. n .
lim For any E > 0. .n + 1) such that + 1)) # 0.n + 1) C U C (n. . take a sequence {In$} of open intervals such that For 2 . + 1) n . b E IR.y E Fn f l I n $ . k ) < ME 5 M Thus m*(f(Fn))0.a < b.m ( F n ) = 0.n open subset U in ( n .236 Then F1 C F2 C * . and f(F)c uf(Fn) = ncu f(Fn). = 4107 Suppose A c IR is Lebesgue measurable and assume that for any a. we have Therefore 5 k m * ( f ( F nf' I n . (Iowa) Solution. Prove that m ( A )= 0. There is an A n (n. k ) ) m ( I n . which implies that m ( f ( F ) )= 0. If m ( A ) # 0 there is an n such that m ( A n (n.
we are left with two intervals 1 and 12. and keep doing this ad infinitum. etc. the total length of intervals removed in the nth step is X ( l The claim holds for n = 1. Claim. Then the total length of intervals removed in the k 1th step is + k i=l . b j ) . 1 Remove from each of them their middle parts of lengths X l l i j l . For any n E W . Assume that it holds for any n 5 k . j We have which deduces that m ( A n (n. + 1))< E. There are at most countably many disjoint intervals ( u j . We are left with the set K x .n a contradiction. bj)’s such that + Then A n ( n . (Stanford) Solution.237 where E < m(An (n.n 1)). Prove that the set Kx has Lebesgue measure zero. n + 1) = U A n ( a j . 4108 Choose 0 < X < 1 and construct the Cantor set Kx as follows: Remove from [0.1] its middle part of length A.
Y L0 <0 + Y.P ( { W 1 = 1 . or p(z z + Y1 = 4 0 . From the rightcontinuity of g.238 By induction the claim holds for any n E JV. It follows that g ( z ) 2 z .O]) n+oo n 1 = 1 . p ( z .11) = P([O. from either A 7 . 4109 Suppose p is a positive Borel measure on lR such that (i) p([O. I. It follows that the measure p is induced by g. Yes. However. for any R z . 01.l]) = 1. y E B. g(1) = p((0.lim (g(O) . 11) . Moreover. It follows that the Lebesgue measure of Kx is 1 c n=l X ( 1 .g(z).1 = 0.g()) = 1. p is the Lebesgue measure. we conclude that g ( z ) = zg(1). and (ii) p ( E ) = p ( z E) for any Borel set E of 2R and every z E B. Y1. y ] = g(y) .A y . For any z E IR define + Then g : lR + l is nondecreasing and rightcontinuous. = P(Y. (Iowa) Solution. For any z . n+oo n and therefore p is the Lebesgue measure. y E IR with z < y. Y we have g(z + Y) = g ( z ) + d Y ) .lim p((. Does this imply that p is the Lebesgue measure? Justify your answer.
Show that for any Lebesgue measurable set E c 1R with X(E) = 1. (Iowa) Solution. is a finite measure and pn’s are absolutely continuous with respect to Z .G. Given any mutually disjoint measurable sets En’s. Prove that u . 4111 Let X be Lebesgue measure on R. j (Iowa) . Then . Let. there is a Lebesgue measurable set A c E with X(A) = !. ndoo is a signed measure. for each E E U .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. by the VitaliHahnSaks Then k=l which shows that p is a generalized measure and therefore a signed measure.
f(Y)I x+ w 5 1 2 . .) Since f(z0) x+w . .oo]).240 Solution. Denote all the rational numbers in [1. It follows from the StoneWeierstrass Theorem. co])separating the points of [0. + T. which is required.. Thus . 2 . A .zo]. Let S = span{ePa5 I n E lN u (0)) then S is a selfadjoint subalgebra of C([0.z]).z” rational. Denote A. u (Iowa) Solution. 12 E A}. = 0. Define the function f : IR f(2) = +[0. 1 be a measurable set of positive measure. there is a point E E such that = f. Put A = E n (co.Y E E . co). S is dense in C([O.YI. . . Therefore for any f E C([O. . .7 3 . 1 . which then implies that p = 0. . . c [I.. co]. .) = m(A) > 0. It is continuous by the following inequality If(. 4112 Let p be a complex Bore1 measure on [0. 21. Show that there 1 exist two points z/ # z” in A with 2’ . = {z Then m(A. . 2 E E. Show that if r w for all n = 0 .T. 20 lim f(z) = 0 and lim f(z) = 1. Z. then .m]) under the supremum norm topology. 4113 Let A c [O.. 1 by 1 X(E n (m. ( IndianaPurdue) Solution.1] by rl.
= 2) then 0 1 a:€ I l).. We will show next that * *  [O.. n such that A n n Am # 0.+ + * a +  3"1 The converse inclusion is obvious.1). By equality ( l ) . Let n = min{k I 2 .1 2 2 + + 3" . n=l 00 Suppose that A.. Then we c a n find z'. 3"' 3 [? 2.. n Am = 0 if n # m. for any 2 E [O. =2 I1 + T.xnlI 1 and therefore 2 21 + . 1)\E = u { 00 Indeed.1].241 U An c [1721. Then C m(An) I m([1.21) n=l 00 = 3.l)\E. 5 zl.+ . = 0 or 1 . ...+1. E is measurable and since + ["3 + 2. Also show that E + E = [0. Thus  = 'I'm .1 . 2 = n=l ? $ (where 0 I z. Therefore there must be some m. Take 2 E An n Am. .+. (Iowa) Solution.. 3 31  3n1 xnl 7 = 2. which contradicts m ( A ) > 0.T. E c [0.. Obviously. I 2 and for any n there is an m 2 n such that 2 . Prove that E is Lebesgue measurable with Lebesgue measure 0. 2 E A such that ' ' z =2 ' + r. there is at least one n such that 2.. 4114 2..
1z . 4115 Let f : It2 + IR+ be measurable. CF n= 1 03 5 z . = 2. () ..T } \ = 0. n= 1 This completes the proof. T. = +m.}I = 0. Obvior. limr. We have g 2 0 and For every r E B. (where 0 Define i]. I r n . and for any n. L E. I g ( 4 = . < E . Take {r. 1 gl(z). n E N.} such that 0 < r1 < r2 < . Show that there exists g : I22 + El+ measurable such that (i) .1) say z = 2 . 2 ' ' = C .gllm 5 E (ii) for every T E B. and z .gl)loo 5 E . < T. Denote () En = { z Set g1 = TnXE. . . for all n.glllcx. n \ arcctgrc = T . .sly. Set fl(z)= f z + arcctgz. {zIg(z)= r } = {z [If gllco = .242 m ( E )= 0. For any z E [0.arcctgz.arcctgz = T } TI.=l 00 X" $ belong to E so that = z. .E + E [0.. . Since E C [0. I 2.+~ r . (IndianaPurdue) Solution. < .  Set g(z) = gl(z) . and let E > 0. [If I{. = Since \ ( r c U{zI z E Enlarcctgz = r. llfl Then g ( )2a c c t g z . there is an m 2 n such that # 2). ~'$2'' Then z' = C 0 0 3 and . 13.1 < fi(z) I ~n}. \ { z g z = T } \ = 0.
z++w + (Illinois) Solution. (b) lim f ( z ) = 0.1 2 ) . a 1 . Then [O. a) Prove that f is measurable.1 1 . . .243 4116 Let f be the function on [0. and 4117 Let E be a Lebesgue measurable subset of lR with m ( E ) < 03 and let f ( ~= m ( ( ~2 ) n E ) .P . * a p . Let K denote the Cantor ternary set. By the definition. . h0 lim m ( ( E+ h ) n E ) = m ( E ) . . (Stanford) Solution. b) Evaluate 1 1 f(z)dz.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 .U p . We will show first that for any Lebesgue measurable set E .2 uu (0 . 1]\K = p 2 1 ai=0. ) Show that (a) f ( z ) is a continuous function on R. we have which then is measurable. O . a 1 a 2 .
there is an increasing sequence { E n } of compact sets such that En C E and lirn m(E. h+O It follows that m ( E ) = lim m ( ( E+ h ) n E ) .5 5 h+Om ( ( E +h ) n E ) + 3~ 5 m ( E )3 . then n 5 hm(En(E+h)) h+O <_ h+O lim m ( E n ( E + h ) ) 5 m ( E ) . ) o i )where ) r=l n (q. If E is a compact set.244 If E is of the form . then there is.) = m ( E ) . h0 In general.G( a . for any E > 0. an open set E’ of the above form such that E c E’ and m(E’\E) < E . n+mh+O n+m lim m ( E n )= lim lirn m ( ( E n n+m h10 h+O . 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. 3 ~ .@i)’s are finitely many mutually disjoint open intervals of finite length. so hO lirn m ( ( E + h)nE ) = m(E).
1 < E . The claim follows.m ( E (Y . ncc lim m ( ( E z) X++OO = = zr+cc lim ncc lim x++m + nE ) m ( ( E+ z) n E ) .f(.z)\E) m ( E (z . say (a*. (z as y + z.m ( ( E + (z . Use this result to prove that there exists an E > 0 such that for any m E 2R with I . (b) If E is compact. z E 2R If(Y) . Prove that cp(z) = J X & ) X A ( t ) d t is continuous a t z = 1. If B is a bounded open set. the line y = mz has a nonvoid m 1 intersection with A x A .)I = Im(((E+ Y)\(E + + Y) ) l 5 m ( ( E+ Y)\(E + = m ( E (Y .m((En+ n E n ) + m((En + n E n ) lim m ( ( ( E+ z) n E)\((En z) n E n ) ) + m( (En + z) n En) + Z) Z) 5 nrn lim ( m ( ( E . where 2 5 m 5 03 and ( a j . n E ) . Then + E ) n E = 0.y ) \ E ) = m ( E ). z b i ) (ai. In general there is an increasing sequence { E n } of compact sets such that En E and lim m ( E n )= m ( E ) .bi) n i=l <m and any z >0 we c ZB nB. (Iowa) Solution.bj)’s i<m u are mutually disjoint open intervals. then for any n have n U ( z a i .m ( ( E + z)\(E (a) For any y. then there is an T > 0 such that for any z > T .z)) n E ) m ( ~ ) + + + + M ( ( E + z)\(E + Y)) + + . .b i ) .245 so h0 lim m ( ( E + h)nE ) = m(E). 4118 Let A c l be a set of positive R Lebesgue measure.)\(En + + z)) + m(E\En)) = 0.y)) n E ) + 0.
a) 5 ~ P ( z B ~ B ) 21 5 X’lim ~ l ( Z nB ) B 5~L(B). n n . there is a decreasing sequence {B. n Bn) . If K is a compact set. 00 Since p ( z B . By (2) we have Since .} of compact sets such that U K .} of bounded open sets such that K = n=l n B. C A and p(A\ n=l 00 n= 1 u K. there is an increasing sequence {K.\K)) we have by (1) + P(Bn\K) In general.. (1) where p is the Lebesgue measure.246 We have p(B) = sup n<m a = 1 C(bi .) aJ = 0.p ( z K n K ) I P(Z(B.
n03 a contradiction. then [B]5 [ A ] . if [ B ] 5 [An] for some n. Prove that the range of p is the closed interval [0. For any [ B ] E Q. and therefore p ( A ) = lim p ( m n ( An A ) ) = 0. [B]= [A] since P(A\B) 5 C p(An\B) n=l 03 =0 .otherwise. Then P is a partially ordered set: [41 5 [B]if p(A\B) = 0.1 < and the line y = mnx has a void intersection with A x A . then 03 = . If there is a t o E (O. (Courant Inst. then for 2 1 any n E mT there is an m. SO p ( A ) = /3 < t o and therefore [A] E 7'.p(S)) not in the range of p.247 If there is no E > 0 with the property mentioned in the question. i. ) = P 5 to. if A is a measurable set with p ( A ) > 0. p(S)].403 lim p ( A .e.  u A n . Let where is an equivalence relation: A B if p(A\B) = p(B\A) = 0. However.B = sup{p(A) I [A] E Q } . this implies that m .. ) } increasing to p. 4119 Let p be a countably additive measure on a set S with p(S) < $00 that is without atoms.) Solution. ] } of Q with { p ( A . then there is a measurable set B c A such that 0 < p ( B ) < p ( A ) . let . E B such that Imn . Given a totally ordered set Q of P . Then P 5 t o and there is an increasing sequence { [ A . Let A = n=l  P = { A I A measurable and p ( A ) < t o } / . ( A n A ) = 0.
p ( A ) and there is a decreasing sequence { [Eln]}of S with { p ( B . It follows that [ B ]is a lower bound of S in R. For any [Cl E S.g. There is by Zorn’s Lemma a minimal element [Bo] R.otherwise [B]= [C]since n=l and ~ ( B \ C )= p ( B ) . if [Bn]I [C]for some n. Let n=l then So p ( B ) = a > 0. by the assumption that there is a subset of COof Bo with 0 < ~ ( C O ) ( B o[CO] [Bo] [CO] [Bo].p ( A 0 ) .p ( A 0 ) ) . (e. Consider the calgebra 2 generated 3 by the compact G sets. Let  R ={BI B c S\Ao. [B]E R. then [B]5 [C].P(B) > O}/ . ) } decreasing t o a. There is by Zorn’s Lemma a maximal element [Ao]of P. 5 and # a 4120 Let X be a compact Hausdorff space. 5 as above. (Iowa) . Given a = inf{p(B) I [B]E S}. Equip 72 with the partial order a totally ordered set S of R. <~ ).p ( A 0 ) whenever B C S\Ao and p ( B ) > 0.248 It follows that [A] is an upper bound of Q in P. p(S\A) > t o . Show that any positive measure p on B which is finite a on compact sets is automatically regular. contradiction. However. then cr 2 t o .p ( B n C ) = a = 0. Therefore p ( B ) > t o . Let where is defined as above. It follows that p ( B ) > t o .
+ + + K\L K nM c B\L. Thus K\L is regular. respec3 tively. C K\L.V open and V E 23) p ( E ) = sup{p(K) 1 K C E. Let us preceed in five steps to show the regularity of p. E i = 1. Step 2. there is a V in 0 such that X\U C V and p ( V ) < p(X\V) E . then U Ei is regular.\E. Consequently. B. Also recall that E is outer regular if p ( E ) = inf{p(V) 1 E and E is inner regular if C V. I 1 5 i ? sets. Recall that sets in 23 are called Baire sets and each compact Baire set K is a Ga set.) < . Let K and 0 denote the classes of compact sets and open sets in 2 . L in K . For any c > 0.249 Solution. M) + E. where the symbol 0 means disjoint union. . regular Obviously For any E > 0. i= 1 n} is a finite class of mutually disjoint. For any pair K . K\L is regular. . 5 P(B\L) < P(K\L) p(K\L) < p ( K +E. . 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 . Then 7E is a ring such that 23 = By definition..U (Ki\Li) I K i . Let a = { ¶=1 . Then we have V). E 0 such that E C Bi with i p(B. Then X\V 2 U and p ( U ) < p(X\V) E . Step 1. 7 ~ . . If { E .. . each set in K is outer regular. . 6 i=l Ei C ( Bi J i=l . there are B1. Let us show each set U in 0 is inner regular. .K ? compact and K E B } . Li E K}.
. En > 0 there are 03 V1. E 0 such that C Vn and p(Vn) < p(En) + &.250 and which shows the outer regularity of the following inequalities u Ei.. . then n= which shows the outer regularity of E . Obviously 00 n=l p ( E ) I inf{p(V) I E For any E G V E O}. n=l Let E = u En. . V. The inner regularity follows from n i= 1 Step 3. n=l n En is . while the inner regularity follows from the following inequalities co Step 4. If { E n } is a decreasing sequence of regular sets. If { E n 03 In E IN} is an increasing sequence of regular sets then (J En is regular. then regular. . Let V = U1 Vn E 0. . .
n E lN.u(En) = p ( E n 5 V E 0) Step 5.there is a K . It 4121 Suppose that p is a nonnegative Bore1 measure on En. inf{p(V) I E V E 0} 5 inf inf{p(V) I En = inf.S is follows that S = B.. (Indianu) . By steps 1 and 2. The outer regularity follows from the following inequalities. Let S = { E E 23 1 E is regular }. E ?E such that n= 1 and which shows the inner regularity of E . By steps 3 and 4. S contains a. 00 For any E > 0. Let Assume f ( ~is finite for all ) T > 0 and assume Prove that p is identically 0. > n> ..251 Let E = n En. t monotone class.
It is easy to show that Assume without loss of generality that B(z. . zk E IRn such that . (Indiana) Solution.2s) and C(z*.l) %k + z B(Xk31)k+m Hence . For any > 0 there exists an T > 0 such that g ( T ) < ( ~ T ) ~ EThere exist finitely many . and define a function f on Enby f ( z ) = p ( B ( z . Then T k i= 1 k i=l k = Z A ( C ( z $ T ) ) & 5 X(C(0..1))where B ( z . There exists an sequence { z k } of K such that k+oo lim f(zk) =a..s). we get that p ( K ) = 0 and therefore p is identically 0. in K . i= 1 Letting E + 0. Let K be any nonempty compact set of 1R" and let a = i n f f ( K ) .1 be a finite Bore1 measure defined on lRn.252 Since C(z. Prove that f attains its minimum on each compact set of IR". E Given a compact set K of IR". 4122 Let 1.&?Fir). 1) denotes the open ball centered at z of radius 1. K c u k i=l C(z"7) c C(0. zl. ) ' S are mutually disjoint. let s > 0 be such that K c C (0 . T ) c B ( z . a s ) ) & .
1] which can be written in a decimal expansion with no sevens appearing. .. .. 3 ' 100 ' 100 (i) Compute the Lebesgue measure of E . Thus. ( f ( E ) ) = 0. 3 3 3 .l]\EU{(O. 4123 Let E be the set of all numbers in [0. .2699.. . . 2 (*> Show that if E c IR is a measurable set with m l ( E ) = 0 then m . For any z E [0. write z = 0. .an17.Ulaz.0.a. 4124 Let f : IR + R" be a function such that for all z.611a2 * * * Un17 = O. y E R If(.~~18)~l. E.~~lS.= 0. ..anl 7: n = 1.ala2..alaz.253 which shows that P ( B ( X . So E is Borel measurable and W m ( E )= 1 n=l 9"l x = 1  1 10" 1 = 0.yI.  1 27 28 (Indiana) Solution. The reverse inclusion is obvious.....2'.) .a1a2 .f(y)ln 5 e'zl+lyI 1 .}. It follows that 0.Un17< x < O. ) Hence a is finite and f attains its minimum on K . . l]\E. E a contradiction. and therefore I # [O.. ... 1) = Q. . 1 0 . If ak = 0 for all k 3 : > n then = 0.2800.aia2 * * * ~~1699.= 0. . Let n = min{k 2 1 I ak = 7). .UlU2.... E E . . (Indiana) .. (ii) Determine whether E is a Borel set.
1 It follows that {fk} converges t o f in measure. (Indiana) .ail < CneZr&. Let u k be an open set such that x ( u k ) < and f is continuous on R n \ u k . For any E > 0 there exists an open set U such that E U C ( . Write U = U(uilbi). Prove that f is measurable. For any E > 0. Prove that either A or R\A has Lebesgue measure zero. where (ai.bi)) i c ~(f(ai). Let E 2 (r. there is an open set U with X(U) < E such that f is continuous on E"\U (in the relative topology). (eZrlbi. 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. Hence mn(f(E)) = 0. Assume without loss of generality that E is bounded.I 2 E l ) = m * ( { zE uk I If(x)I 2 E } ) 5 C. It follows that r n . Assume that r A = A for every nonzero rational number r.254 Solution.. 4125 Let f : R" + lR be an arbitrary function having the property that for each E > 0. Then condition (*) implies that f((ai.a i l ) ' ) . Since the Lebesgue measure is complete f is measurable. ( f ( ~ ) _< ) C C. m * ( { zI lfk f )(.bi)'s are mutually disjoint. (Indiana) Solution.T . T ) and rnl(U\E) < E .r). a which implies that rn:(f(E)) = 0 and therefore f ( E ) is measurable. then f k is measurable.e2'Ibi . Let fk = fXEn\u.
Y 2 E a*. For each measurable set E QJ then Y is the desired probability measure. m ) ( ~ ( z = 1) such that p is absolutely ) continuous with respect to v . Then f is continuous and for any nonzero rational number Hence f( z) = 0 for any z E R* Since . c )by m f(2) = 1. we conclude that m ( L . Let lR* = R\{O} and 3 = A\{O}.m). 4127 Let p be a cfinite measure on the measure space ( X .. It follows that m ( B ) = 0 and therefore m ( A ) = 0. (Indiana) Solution. dY XK(Y)XLl(Ylz). define function f : lR* + [0. 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. Then r B = B and T(R*\B) = R*\B for every nonzero rational number T . Prove that there exists a probability measure Y on ( X . and v is absolutely continuous with respect to p.255 Solution. If m(lR*\B) = m(R\A) > 0 there exists a compact subset K of E*\B with positive Lebesgue measure. T.l ) = 0 and therefore m ( L ) = 0. For any compact subset L of B . .
Prove that for every open interval I E ( 0 . (> = X { 0 } ( 4 and d . l ) and every an open interval J. l ) fdm = 0 . let f. and fdm = 0.) < E . i.. l ) such that E > 0. (Stanford) Solution.e..l}(x). the function g o f is not integrable. Then f and g are integrable functions with compact support. there is x EJ. 1). For example. (Illinois) Solution. However. 4202 Let f E &(O. . since g o f (z) 5 1. There is a measurable set E of measure zero such that any x E (0. Assume that for any x E ( 0 . m(J. 2 ( 0 . 1 = x{o.256 SECTION 2 INTEGRAL 4201 Prove or disprove that the composition of any two Lebesgue integrable functions with compact support f . 1)\E i s a Lebesgue point o f f . It is not true. g : IR + IR is still integrable.
e. m(J. l)\E and for any n there is an open interval Jn & ( 0 .. . p ) be a positive measure space with p ( X ) < 03.e. i. l ) such that z E J. Therefore 4203 Let (X. one has Jx f d p < CO) if and only if the series n=O c 00 I f (z) 2 n}) (Iowa) converges. Then W . Suppose that f is integrable.) < and It follows by equality (1) that f ( z ) = 0. 03) is integrable (i.e.. Show that a M measurable function f : X + [0... f ( z ) = 0 a. Solution.257 For any 2 E (0.
(b) Is there a Borel measure p (positive or complex) on B with the property that J. (a) Yes. + E . which shows that f is integrable.1]. ) (b) No. Then a contradiction occurs from the following limits JL% and Irn ( p ( t ) e t d t= . / p(t)dt > 0 4205 Let E be a Banach space. let ‘p 2 0 be a continuously differentiable function of compact support.?r. If there were such a Borel measure.258 Conversely. (Iowa) Solution.p ) a probability space. fdp = f (0) for all continuous f : B + (c of compact support? Justify. Let p ( E ) = x ~ ( 0for any Borel set E . f d p = f’(0) + (c for all continuously differentiable f : R of compact support? Justify. (X. 4204 (a) Is there a Borel measure p (positive or complex) on IR with the property that J . taking value one on [1. and f : X such that g o f is pintegrable for every g E E‘.
L ( g ) = I g 0 fdp. Define the linear operator T : E’ + L1(X). By the closed graph theorem. Assume that pn + ‘p in E’ and T(pn) + h in L1(X). p ) be a positive measure space with p(X) g be realvalued measurable functions with < co.. If (b) does not hold. It is ture that L E E”. Does L E E”? Justify your answer. p H p of. then for any E E M . Since .259 Define L : E‘ + IR.and let f and Show that either (a) f = g a. 4206 Let ( X .e. We have So L is bounded. M . It follows that pn o f converges t o p o f everywhere and to h in measure and therefore h = p o f in L 1 ( X ) . T is bounded. or (b) there exists an E E M such that (Iowa) Solution. (Iowa) Solution.
Since a' is second countable. For any E fdP = J. Show that { z E X 0. J. one finds that a'\S is the union of countably many closed balls { z E C I Iz . d P . I f(z)@ S } has measure (Iowa) Solution.If .A 1 5 E ~ } ' s . and S a closed set ina'.260 for any E E M . > 0 the sets are measurable. Suppose that 1 JE ) P( E f@€S whenever E E M and p ( E ) > 0. From the equalities and we conclude that p ( E + ) = p ( E . p ) be a positive measure space. .) = 0. It follows that Therefore (a) holds. 4207 Let ( X .M .
1] + (0.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 . Take an N such that Suppose that E C [0. Denote El = E (f Then m(E2) < 5 . 4208 Let f : [0. EZ = E\E1.1] with m ( E ) 2 a .X.not to S. Therefore 2 +). m) and let 0 < a 5 1. (IndianaPurdue) Solution. m ( E ) > a.) . Thus .1]. I 5 E ~ } .261 there is a t least an n such that P ( { ~ E Xf I ~ ) . 0bviously. Show that where inf is extended over all measurable E c [0.a contradiction. so m(E1) > $.XnI I If(. where 5 E ~ } ) .
Let . (Indiana) Solution.f k . 4210 Let /I be a finite measure on lR. (I Zin o is) l there is. n Therefore for any t E lR\E. n fn(t) = fl(t) k C ( f k . by Levi’s Lemma.and define Show that f ( z ) is finite a. a measurable set E of measure zero such that for any t E B\E. It follows that fn + f almost everywhere.l ) ( t ) k=2 converges. Prove that Solution.e. with respect to the Lebesgue measure on lR. Since fn + f almost everywhere with respect to Lebesgue measure.262 4209 Let {fn} be a sequence of realvalued functions in L 1 ( B )and suppose that for some f E L1(IR).
f n : X + [0. and f : X t [0. p ) be a positive measure space. Suppose that fn +f a. without loss of generality.e. that p is totally afinite. 4211 Let ( X .M .263 then g E L 1 ( R . Assume. with respect to the measure v. CQ] a sequence of integrable functions. m] an integrable function. [p]. The conclusion follows from that the measure v and the Lebesgue measure are equivalent. .where dv(z) = 1 dx +22’ by h b i n i ’ s Theorem. the function is finite a.e.and that Prove that (Iowa) Solution.e. d v ) . i..
then by Fatou’s Lemma one has {s. For any A E M . Let d v ( 2 ) = g ( z ) d p ( z ) . by the VitaliHahnSaks Theorem. f n d p } is bounded and therefore admits a convergent subsequence. If f n . There is by Egroff’s Theorem an E with v ( E ) < 6 such that converges t o f uniformly on X\E. d p } is any such subsequence. equivalent t o p. 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 ) .264 n=l where Xn’s are mutually disjoint measurable sets of finite and positive measure. 71 E m* Then An << v. Let Then g is integrable. Then v is a finite measure on X . {s. Then 9 { $} .
}be a disjoint sequence of M such that X = U X . If p ( E ) = 0 then P(E) = 0. as desired. then (Iowa) Solution. (a) Show that there exists W E L1(p) which takes its values in the open interval ( 0 . finite measure on M . (b) Show that if f is a complex function on X which is measurable with respect to M. and n=l 03 p ( X . The set function is of course a positive. Conversely. Let W=C1 O0 n=l 2 1 . and that P ( E ) = 0 ep ( E ) = 0 for E E M . ) > 0. l ) . positive measure on a ualgebra M in a set X. Show also that r is a positive. + p(Xn) xx. P . (a) Let {X. finite measure on M .265 4212 Let p be a afinite.
there is a sequence of simple functions { f n } such that lfnl 5 f and lim fn = f . i. . we see that fW is integrable with respect to p..fdF=J X fWdP Conversely. Then fnW is integrable with respect noo to p and hence fn is integrable with respect to p. We have and therefore J. m+oo ncc lim fnW = f W . n In general. if fW is integrable with respect to pl assuming without loss of generality that f is positivevalued. if f is integrable with respect to the measure p. . . and since are lim n+w Jx\jnwfmWldp= lim n+w m+cc ~f~fmldji=~. Since and lim fn n+cc = f.e. .266 (b) If f is a simple function.f is integrable with respect to p. Accordingly.l n . 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). n f =CaixEi! i=l where p ( E i ) i= 1 < +a.= i l .
Show that J. 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).1] t R such that (3) 0 5 g 5 1. 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. (Iowa) Solution. if : ( IndianaPvrdm) . there is a 6 > 0 such that IE . fndm= 0. For any E > 0.Suppose that (i) fngdm + 0 for each g E C([O. fndm + 0 for each Bore1 subset A [0. There is a continuous function g : [0.11) and (ii) Ifn I h for all n.fndml % hdm < E whenever m(E) < 6. Then we have = n+m lilfnXAdml It follows that Jiir J.1]. IJ.267 4213 Let m denote the Lebesgue measure on [O. 4214 Let { fn} be a sequence of nonnegative measurable functions in LP( R) for some 1 < p < 00. ( 4 ) g = 1 on K and (5) g = 0 outside U .
”+fP.  PI 0. take . Just as above. we have Since T:+X SO =f.268 Then f n 5 f and IL .+0.fl which implies that IIE . it follows that +  Jfi For any c N1 such that . Therefore Conversely.” Then fn f p l .fP)/ I J tfi Jfp.f I L Ifn . Suppose that J If. +  Since IJu~ > 0.f (Ip + 0. 2 f. f > 0.
All parts refer to Lebesgue measure on R. (a) Give an example where { f n } converges to f pointwise. 0.269 for n > N1 and any measurable set E . Denote 7 = &/m(A)f. f: > N1. a measurable set. Take S > 0 such that < EP for n if m ( e ) < 6. 4215 Let 1 5 p < 00. Take A . 5 M . llfnllp and llfn .. Wehave N1 (J. UP Ifn flp) for any n >N. Take N such that N 2 a n d m ( l f n ..fllp f . Vn .. < EP e if m ( e ) < S and n > N1. such that m ( A ) < +m and J. Thus Jf.f l > q ) < S i f n > N . < f P &PP Then J.
Then gn 2 0.e.fJPdz = 0. on [0. Prove 1' lfnldz t 0 .fllp 0 (IndianaPurdue) Solution. 4216 Suppose f n is a sequence of measurable functions on [0. By the Fatou's lemma. Then zE lim f n ( z )= f ( z ) = 0.l f n . Using the Fatou lemma. i ) . (c) Show that if { f n } converges to f pointwise and llfnllp + IlfllP. what can you conclude about Ilfllp? Justify this conclusion.fll = llfnll = 1.270 (b) If { f n } converges to f pointwise and l ] f n l l p + M < $00. and n+m lim g n ( z ) = 2 p + ' l f l p pointwise. llfllp 5 M . [O. then +  llfn . (c) Set gn = 2"(lfn Ip + Ifl") . we have Therefore no3 I f n .1]. (a) Consider L[O.13.1] with 1' and f n + lfn(z)12dz 5 10 0 a.f l P . Set ncc fn =n ~ ( ~ . 1 1 and (b) We conclude that llfn .
Use Egorov and CauchySchwartz.1 I E ) < E and (2) f n converges to 0 uniformly on [O.log f d p ) . and we have rl 4217 Let (X. 1 We have by the CauchySchwartz Inequality and therefore Let E + 0.1] such that (1) m([O. For any E > 0 there is by Egorov’s Theorem a measurable set E c [0. It suffices to show that If f = 1 a.. then lim P10 (J. 1]\E. (Stanford) Solution.00). p ) be a probability space (a positive measure space with p ( X ) = 1).M . Show that i f f is an integrable function with values in [1. equality (1) holds.e.f p d p ) ’ = exp ( J . for any 0 < p < 1 and any 2 EX . (Iowa) Solution.27 1 Hint. Otherwise.
n fn(x) = j+k5. Consider 00 n=2 a) Prove that this series converges in L1[2. where { p n } is any sequence of ( 0 .272 By the Dominated Convergence Theorem we have nco lim log(Jx f P n d P ) Pn = = n lim 00 log(J. 00) b) Prove that where the sum on the right is the pointwise limit (you need not prove that this pointwise limit exists). k=2 x 2 2. .m). 1+ ( f P n Pn  lim n00 ( log(1 + JJfP" J. l ) decreasing to 0.(fP"   1)dP) J#"  1)dP Pn = Llogfdp. (Stanford) Solution. a) Since The series b) Let C n=2 00 ann2 converges in L1[2. 4218 Let { a n } n > 2 be a sequence of real numbers with lan[ < logn.
k=2 00 By a). (Iowa) Solution. and by the Dominated Convergence 4219 Let S be a bounded Lebesgue measurable set in IR. H IakIk" k=2 is integrable. and let { c n } be a sequence in lR. By the RiemannLebesgue Lemma. Show that where X is Lebesgue measure. we have = X(S). 2 Theorem.273 Then (i) nw fn(z) = lim 00 cc C aklc" k=2 and (ii) Ifn(z)I 5 C lanln". 1 2 .
Y ?M2+Y211) If(. (a) The function f is not integrable on the unit square.274 4220 (a) Is the function Lebesgue integrable on the unit square 0 5 x 5 1. 0 5 y 5 l? (b) Compute the repeated integrals in the two orders. (c) Does the integral exist? (If you use a theorem. Y)ldXdY rdrde drde (b) We have .. for ) b . be explicit!) (Iowa) Solution.
275 and 4221 Let ( X . Define the function H . By the Dominated Convergence Theorem + If I 4222 e t z f ( t ) is in Suppose that f is a measurable real valued function on l such that t R L 1 ( E )for all x E (1. Evaluate ncc lim L n l n ( l + (!)’) dp. In( 1 + x’) 5 It follows that for any n E lN the function n In( 1 !$) is dominated by and therefore integrable. It is easy to show that for any x 20 2. (Iowa) Solution.l).M . p ) be a measure space and f E L ( p ) .
2 0 and x < y < H < 0 and x < y < A' lzz.l). l ) .x ) ~ (x < z' < z ) = I(ett  e")tf(t)l.z l e ' z X t f(t). +m etztf(t)dt. (x 4 and t A 2 Iy . t t Iy . for any y ) > z.Prove that 'p is differentiable on (.zlet"t2f(t).1 . we see that t follows that 4223 Evaluate .1. For any fixed z E (1. It kkz By the same method as above. (Iowa) Solution. Next we show that dx E (1. l. et"t2f(t) is integrable. e"f(t)dt = m J_. I). < z < y) = \ e t z " t 2 f ( t ) ( z. and we conclude that t H e " t f ( t ) is also in L1(E).l).276 for all x E (. Since for any x E (1.
Hint.)n =ex. i=l )(l&).277 justifying any interchange of limits you use. First show that (1 E)n 5 e" for x + 2 0... By the Dominated Convergence Theorem = brn e"dx = 1.(lS Z! i=l + and ncc lim ( 1 + . we have (I+ x ) n s e x . Since for any n E IV and z 2 0. . (Stanfod) Solution.
For the function f : X I x X z + B defined by .co) a measurable (Iowa) Solution. Since l n t 5 t l n t for any t > 0.278 4224 Let ( X . 4225 For i = 1. Under the condition that f In f is integrable we conclude that both f and In f are integrable. 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. Let Mi = 2N (the galgebras of all subsets of nV) and let pi be the counting measure. let Xi = nV (the natural numbers).2. Prove or disprove: + [I.A. p ) be a probability space and f : X function. and Indeed.
j = is 1. Compute the iterated integrals and How do you reconcile your answers with Fubini's Theorem? (Iowa) Solution. j )= { 279 . j ) d p z ( j )= 2i + 2i Therefore For any j E X 2 . and for X > 0 define Cp(4 = P({Z E Q I f ( Z ) > XI) . i H = 0. j ) is integrable and Since f is integrable ( C I f l ( i . p ) .p ) be any measure space. otherwise.2i.f ( i . 4226 Let (0. f ( i . For f E L1(Q.j)is integrable and J. the two iterated integrals exist Li€N and coincide by the Fubini'i Theorem. j = i. j H f(i. f ( i . j ) = 2).. For any i E X I . 2i.
As usual. 0 Hint. ~ ) ( l f l ( zA) is measurable. The function (z. On one hand. it may be helpful to consider f positive first. 4227 Let f E L 2 ( 0 . X . are Bore1 measurable and that llflll = J m ( ' p ( A ) + +(A))dA.I). Set . By Fubini's Theorem we have provided that either side exists.) + but on the other. A) H x ~ O . The measurablity of 'p and follows from that they are monotone. (Iowa) Solution. .280 and 4(A)= P ( { Z Show that the functions 'p E c I f(z) > A>). 2 and II.
. y E [ O . 0 < h < 1 where c is a positive number independent of f. 1) and assume that f(z)g(y) = f ( y ) g ( z ) for all z. (UC.F ( z ) h l2 dz) 4228 Let f.g E L1(O. We have ( I J1h 0 F(z + h) .281 Show that for each h. Imine) Solution. l ] .
Y) I 0 5 2 < Y 511 (Iowa) and dA denotes the planar Lebesgue measure. .282 where A = {(Z.
. q E (1.m). . .p > 1 and be positive numbers c.w)valued measurable function f on X . w) with p < q . The function I is strictly increasing if and only if f is not almost everywhere equal to the constant function.. > 1. Show that for any [0. m). is a nondecreasing function of p E (1. let CY = and = then $ = 1. For any p . .. n 1 1 i=l If.283 SECTION 3 SPACE OF INTEGRABLE FUNCTIONS 4301 Let X be a positive measure space of total measure 1. By Holder’s inequality. fi E CPi ( p ) . 4302 Let (X. .n}.n. . * f n E Cp(/l)? (Iowa) Justify. Under what circumstances is I ( p ) strictly increasing as a function of p? (Iowa) Solution. . let M such that p. for i E { 1.=.. p ) be a fixed measure space. . we have + which shows that I ( p ) is a nondecreasing function of p E (1. . must it be the case that fifi . i = 1 .
and L"O(p)? Justify your answers.fnllp 5 llflllpl . lfilp E L"(p) and PI If2Ip E L F ( p ) . L z ( p ) . Yes. and let p be the measure determined by P ( { a ) ) = 0. and by H6lder's inequality.1 .. .*. 1 o < +. .1  which shows that ppk c+k12p > Pl P k ? ' Pk+l PPk 1 pk+l?p < ' 9 1. fn E LP(p).... Moreover llfl. . . c}. Then for n=k+l. By inductive assumption f i e .. f i * . IIfnIlp. P({b)) = 1 1 and P({. (1) In case n = 2. 4303 Let X = { a . Assume that the conclusion and the inequality (1) hold for n = k .1. What are the dimensions of the vector spaces L 1 ( p ) .284 Solution. f k ) f k + l E LP and which completes the proof.)> = 00. b. let M be the cralgebra of all subsets of X . fk belongs to 1 L Pk+:P ( P ) and Then by the case n = 2 we conclude that (fl .+ 1= . (Iowa) .
We have dimL1(p) = dimL2(p) = 1 and dimLW(p) = 2.285 Solution. + < 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 . Let f be a nonnegative function in P ( E )for some 1 5 p T > 0. 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 ) . ( IndianaPurdue) Solution. s > 0. we conclude that dimL'(p) = dimL2(p) = 1. 4304 lo Show that f d m = 0. Also let f h ( z ) = f ( z h). Denoting we have 4305 T +s =p.
From f[fS L f . (ii) We claim that lim lhl+m 11 f i f s I I 1 = 0. Set p = 1 + f . Now if Ihl > 2 N . then z h 4 E whenever z E E. Since we have . Therefore + 4306 Let f : R +R+ be measurable. For any E > 0. take N such that JEc f P < E where E = [ .286 (i) Show that fif s E L1(R). (IndianaPurdue) Solution. (ii) Investigate what happens to 11 fifslll as fi E Ihl + co. and P +P = 1. and let 0 < T < 00.N . Show that (IndianaPPurdue) Solution. f" E L f . Then p > 1 and f + & = 1. N ] . (i) Obviously follows that fh E Lp(R). it E L1(R).
For any Then E > 0. llfllm = I I Ilf llm. : Therefore .287 which implies 4307 Let f be a bounded measurable function on (0.l). Prove that (IZZinois) Solution.
Let f(z)= However for any t > 0 & . (Illinois) Solution. 11).1] satisfying Determine those values of p . Then f 2 1 2 is not integrable. Indeed.1 I llfllcc I 1) 1 is not compact in L1[O. for any p E [l.let .13. 2). prove that s = {f E C[O. It suffices to show that S is not closed in L1([O. 1 l+t 4309 With Lebesgue measure on [0. The minimum value of p for which f may fail to be in LP is 2. 3 (Iowa) Solution. 1 . If 1 5 p < 2. 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.  1 (1+t)2  < 2. m M it follows that f P is integrable. For each n E IN. then f E LP.1.288 4308 Let f be a nonnegative measurable function on [0.
(b) For any w E X .}rZl .~ is bounded. .w) = 0.i. in the strong (i.e. E 'H..w) = 1 2= u(z)w(z)dz. norm) topology of 2. For any m. By RiemannLebesgue Lemma. w E 'H. (Stanford) Solution.2 1 . .271).w E L1(0. n = 1 .too lim (un. the Hilbert space consisting of functions f E L2[0. (b) u. 4311 Let H1 be the Sobolev's space on the unit interval [0. the set {u...  {u. but not compact.}. + 0 as n + 00 in the weak topology of X .  Consider the elements u.1 such that 1 . for every w E 'H. n 2 1 with m # n which shows that {urn I m E IN} is closed.e.289 4310 Let 'H be Hilbert space L 2 ( 0 . (a) Obviously.1]. i.". defined by u.(z) = sin(nz) for z E (0. it is not compact. Show that (a) the set is closed and bounded. Since it admits no convergent subsequence. 2 . with inner product defined by 7) ( u . u ..e.2r) by Holder's inequality.
1] too the series n=w C f^(n)e2ninzconverges to f(z) both in H1 and in L"O. Solution. It follows from (1) that IlfllP I Cllflll. (Stanford) . (ii) Find the &norm llfh . Suppose For each h E lR define the function fh by f h ( z ) = f ( z . 2 T ~ belongs to ~ ~ ( O . (i) Give the Fourier expansion of fh.h). (iii) Prove that unless f is constant almost everywhere. 4312 Let f be a periodic function on lR with period 2 a such that f 1 [ 0 . and h. Show that there exists constant C > 0 such that (Stanford) I l f I l P I Cllflll.290 n=cc where are Fourier coeffents o f f . Since for any z E [O.fllz in terms of the a. 2 a ) .
and that this inequality is sharp (equality) if and only if {f.}span a dense subspace of L 2 ( 0 .1). Then 4313 Let f n ( x ) be an orthonormal family of functions in the Hilbert space L2(0. Prove that for all 2 E [0.. By Bessel’s inequality . (Iowa) Solution. there is an n # 0 such that an # 0. It follows that fh = f and 11th . 1 n = 1.. I f f is not constant almost everywhere.1]. then a. n # 0.291 Solution.1). = 0. (i) We have n=w +w n=w (ii) We have n=w (iii) I f f is constant almost everywhere..f l l 2 = 0.
Then ft converges to f uniformly.fllz 5 S(1lft . Suppose that f : lR + a' be a continuous function with compact support.fnll2  + llfn . Conversely.1] = { z y 12 E K . 4314 I f f is a function on IR. E nV} span a dense subspace of L 2 ( 0 . let ft be the translate ft(z) = f ( t + z).fllL2(nz) = 0. Then L t+O lim llft . . there is a sequence {fn} of continuous functions with compact support converging to f in L2(IR). since s ~ a n { x ( o . y E [1. then t+O lim llft .fllz) i t0 2 l l f n W .fllz.then {fn} is a n orthonomal basis of L2(0.z] 1 0 5 z 5 l}.1) while is a closed subspace of L2(0.[1. Prove that i f f is square integrable with respect to Lebesgue measure. We have where K = supp(f) and .1 b l Ia Ib 5 I} O is a dense subspace of L2(0. l]} is compact. {fn} span a dense subspace of L2(0.l).1) and therefore (1) is sharp. I~0] 5 z I 1) = s p a n { ~ ( ~ .fllz + llfnt fnllz) = 2llfn .292 If {fn I R.f n t l l 2 + llfnt .1). containing span{x(o. In general.l). it follows and therefore. (Stanfonl) Solution.
4315 A trigonometric polynomial is a function p on lR of the form N n=N for some Cn €a'. Then where a ( d z ) is the Harr measure on S1such that a(S') = 27r. Suppose f is continuous and 21rperiodic on El. Let u be the harmonic extension o f f to the unit disc D = { z 1 IzI 5 1). there exists a trigonometric polynomial N n=N such that Ilf and PI103 <E ICnl 5 lan] for all n.293 Letting n . we have . Hint. Then . t0 lim lift . N 2 0. (Stanford) Solution.fll2 = 0.and let Show that for every E > 0.00. You may wish t o think about harmonic functions on the unit disk. Regard f as a continuous function on the unit circle S1 of a'.
P {gn}T=l a sequence in Lq([O.l]). where {fn}r=l + and lim gn = g in L Q . Yes.11) is not. S is not separable. l]).1 ) is not 3 dense in Loo([O.l]) (with Lebesgue measure) and = 1. Since C([O.S is of cardinal N.1] is dense in Loo([O. 3 n+1 n Then as a subset of L”([O. (i) False. < 2 5 .11) is separable while Lw([O. (ii) False. 1 ) I f ( z ) = 0 or 1.is it true that f n g n nw nw + f g in measure? Justify. (Iowa) Solution. is a separable metric space. 11) ( Stanfo rd) Solution. fngn + f g in measure. since .294 There is an N such that where r is fixed so that Put N n=N which is required. 4316 Prove or disprove the following statements: (i) the set of continuous functions on the interval [0. Let 1 1 S = {f E L”([0. since the distance between any two elements in S is 1. If lim f n = f in l.f(0) = 0). n E m. C([O. However. (ii) Lw([O. Indeed. 11). 4317 Let be a sequence in LP([O. 13).
S = {(equivalence class) of measurable complex functions on X}.e. f) 2 0 and d(f. define Show that d ( f .295 we have for any E >0 where A is the Lebesgue measure.g) is a metric on S. as usual. For f .g) = 0 iff g) f = g in S. fn + f in this metric (Stanford) Solution. = d(g. iff . g ) = p(f . Obviously. 4318 Let X be a measure space with measure p and suppose that p ( X ) < Let 00.. two measurable complex functions are equivalent if they agree a. we have &  0 a.e. and that if and only if fn + f in measure. g . h E S.1 For f E S. d(f. (Here.
then %0 in measure by the equality f 1+f + { 2 EX I 1 + Ifn . .(z)dz on P ( I Rconverges to f H . It follows that the sequence of bounded linear ntm functionah f H J f(z)g. * .s all f E Lp(IR). then  d(fn.5 Isfl.f) & 1+E + 0 as n t m. Moreover. By the BanachSteinhaus Theorem. ( st u n f o d ) Solution. Prove that there is a constant M such that . Then gn E Lq(IR) and for any f E L p ( l R ) . 5 n and 11 2 5 n. If fn + f in measure. h E L q ( R ) .f) +0. 4319 Let g be a measurable function such that lfgl < 00 for every f E P ( B ) (fixed p >_ 1).. By the Dominated Convergence Theorem d(fn.296 Therefore d is a metric on S.fl Ifn  fl > &} =  { 2 E XI Ifn  f l 2 lE '} (O<&<l).f ) + 0.* L lsnfl 5 . f ++ f(z)g(z)dz is continuous on P ( R ) and therefore there is an h E L q ( R )such that . . Define for each n E B a measurable function gn V gn(2) = { g!x" 5 hfl 5 if Ig(z)I else. 191fl . ) J pointwise.s which then implies that g "2. f(z)g(z)dz. lim gn f = gf. If d(fn.
a) Show that f * g E L o o ( R ) . We have ~ ~is an orthonormal basis of } ~ ~ a Therefore {f * k I k E L 2 ( S 1 ) }contains L2( S1).x.297 4320 Let S1 denote the unit circle (the set of complex numbers with modulus one. I'" f(%7(ct  wo. g E L 2 ( R ) .(O) = elne then { L2(S1). *k 1 k E L2(S1)} is (Iowa) Solution.define the convolution f * g by f * 9 = J_. or the real numbers modulo 2a). too f(y)g(z . .Show that the linear space { f dense in L2(S1).Do not neglect to check the measurablity of f *!l. {xn I n €27) and therefore is dense in 4321 For functions f . Let for any n E Z.Y P Y . The convolution of two functions on S1 is f *g(a) =  2'. Suppose that f is an element of L2(S1) with the property that its Fourier coefficient is nonzero for all n E Z. where the integral is with respect to Lebesgue measure.
Y)ldY II l f l l 2 l l s l l 2 . By (l). By the 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). I r+m I = 0. 2 E lR * g = T f g . Since by Holder's inequality If(y)g(z . Hint.define + (1) + i ([r If(y)l2dY) (Jt" Id.f * g)(z)l I llfllzllgn . = It is evident that Tf is a linear operator. closed graph theorem. (Iowa) Solution. Indeed. Tfis bounded. Define Tf : L 2 ( R )+ L2(BZ)by T f ( g ) f * g . f * g(z) is well defined. oo  YVdY) . b) Let { g n } be a sequence of L2(IR) such that gn t g and Tfgn+ h in L 2 ( R ) .f * g is bounded and therefore belongs to L m ( B ) . Since l(f * 9 n . you need not check this.9 / 1 2 we see that h = f + 0. 4322 I f f E L 1 ( R ) . Show that Tf is a bounded linear operator. a) We will show that f * g is uniformly continuous.
If f is simple.1] + [0. ?(() Solution. for any f E L1(IR). (Y let . co) be an essentially bounded function. there is a sequence {fn} of functions of the form (1) such that Then Letting n + +a. llflloo > 0.299 Prove that. Show (Indiana) Solution. ( Stanford) (1) In general. have we 4323 Let f : [0. For any a with 0 < < Ilflloo. say +0 as + 00.
. such that lim km > 0 there exists asubsequence of {fk}. the Dominated Convergence Theorem. 2.300 and F. M . If on the contrary k+m lim Ilfkflll llfk denoted {fk}. where X is the Lebesgue measure. we get that 4324 0 . Let 1 < p < Suppose that { f k } is a sequence in p ( x )such that ~upIIfkI(~ 00 and < k fk(Z) n+m lim = f ( z ) exists for pa .p(X). 0 Let ( X .l]\E. Since { l f k l } is bounded in P ( X ) {(fk. Hence Letting a 1 llflloo. I} of {lfkl Ik E m}. = [O. by = 0. Prove that (Indiana) Solution. and Lp(x) reflexive there exists a subsequence is converging to If1 weakly in I.e . For any k E N . p ) be a measure space for which p ( X ) < 00.) > 0. also fill > 0. then X(E.
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. a. .301 By the VitaliHahnSaks Theorem.. .2. for any that for k = 1. Then Letting E + 0. E > 0 there exists a 6 > 0 such r r whenever p ( E ) < 6.
u 5 z. v 5 y g(u. (*) Assume also that g is nondecreasing in each variable.f(. So lim g(z.e. i. . Let #(z) = g(z. Again by (1) f is Lipschitz and therefore absolutely continuous. since by (l). z) then 4(z) 4 is nondecreasing and L !J(Y. Show that lim g(z.1] x [O.302 SECTION 4 DIFFERENTIAL 440 1 Let f : [0. y) Y+X # 4(z) exists except in a countable set.13 + lR and g : [0. 1 + lR satisfy the inequality 3 f(Y) . Observe that (*) is equivalent to (Iowa) Solution.) L g(Y1 Z)(Y  ( for all 2.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. However. y) = +(z) exists except in a countable Y+X + set and that for 0 5 z 5 y 5 1 we have Hint. v) 5 g(z. y).f’(z) = 4(z) whenever # is continuous at z.. Y E (071)).
with for all 2. (Stanford) Solution. then f is differentiable almost everywhere and moreover for any x E [O. 1 1 f) ( .303 4402 Prove that a function f : [0. (11) fn(x) + f ( x ) for all 1 E [0. if and only if there is a sequence of continuously differentiable functions f n : [O. define T which is required.1]. by the Mean Value Theorem This completes the proof.1] there is a sequence { g n } of continuous functions Ign(z)l rl IM. 11.y E [0. . 1 + R such that 1 (I) [fA(x)l I M for all x E [O.x E [0. If f is Lipschitz.1]. : Hint. Conversely. There are several different ways to do this problem. 5 M. E [O. 1 1 For any n E A . one is to use the Fundamental Theorem of Calculus.1] + R is Lipschitz. Since I ‘ z I f() such that and = f(0) + i= 2 f’(x)dz.
3 Show that f is absolutely continuous on [0.304 4403 Let {fn} be a sequence of absolutely continuous realvalued functions on [0. Since f E A C ( I ) . 4404 Assume that f E A C ( I ) for every I c R. 1 . (ii) f ( z ) + 0 as 121 + 00. Then 1 It follows that f is absolutely continuous. If both f and f’ are in L1(R) show that (i) f’ = 0. (IndianaPurdue) Solution. Let n=l in L1\O.for every I c R. SR .1] such that (a) f(z)= n=l C 00 fn(z) converges for every 2 E [O. 1 .1]. (nlinois) Solution.
l]).13. In the same way. there exists g E L1 such that f : + g. If {f. Therefore we f’(z)dz = Iim f(z) .1 ) such that 1 f n + f uniformly on [0. 1 ( IndianaPurdue) Solution.1 ) such that fn + f uniformly on [0. Since fn E AC( [0. Since x++w f E L1.}is Cauchy ( L l ) . 0 f’(z)dz + I’” f ’ ( z ) d z = 0. 11). fn(z) = 1’ fXW. which means f x+w x++w lim f(z)exists. .show that there is f E AC([O.f(O) = f(o). must have lim f ( z ) = 0. Thus So there exists an f E C([O.fn(0) = 0 for every n. we have lim f(z)= 0 and x+w Thus L f‘(z)dz = J_.305 Since f’ E L1. lim x+w sox’ ( t ) d t exists.l]). L’ which implies f E AC( [0. 4405 Let { f n } C AC([O. Then Moreover.1].
Take N such that f 5 < 6.2 1 < 5. 4407 Let Gn. (1) f : I + R is Lipschitz (2) E > 0 3 35 = S ( E ) > 0 such that { I j = [ a j ..2 1 2 6. we give the following fact.f(zl)l < 1. . Suppose that 22 .. Take N 2 0.1 < 6. c. (2) 3 (1).c. 2 . I f ( 4 . Show that f is Lipschitz. I. E I . 2 2 . bj] c IIj I 5 6. such that Then N l f ( z 1 ) .1] such that GI 3 G 2 3 * * * . Then Suppose that 22 . n = 1 . 2 (b) (1) =+ (2) obviously.1)1 I Therefore we find a Lipschitz constant 1 $.f(. Take 6 > 0 such that If(bj)f(aj)l 5 E for any Ij = [ u j . So we have 4< N(z2 . (a) For any zl. . :2n 1 Let . For any 2 1 .2 E I . be open subsets of [0.21) < 5. Take { c j } such that x1 = cg < c1 < .306 4406 (a) Assume that f E A C ( I ) and f’ E L”(I). < C N = 2 2 . 2 1 < 2 2 .b j ] } c I with (IndianaPurdue) Solution. (b) Show that the following two statements are equivalent. lGnl I .
= &. If(x”) . n [x’. x’. Gkt # 0. x Then + 6) c GK = K Gi. Set 6 = &. b E (x . i=l n=l K n=l K 1 = Klb. Take a.1] and C ( b i .6. Take x E GK.307 Show that AC([O.al. 6 > 0 such that (x . for > x’.x‘’]~5 no1x” 1 no Conversely.f(x”)I Gn = 0. Then GK # 0.2’1.x 6) c GK. b. (i) For any (9 f 5 MIX’. a < b. (ii) If(x’) . take N such that C N+1 M & < 5.S.f(x’)l = C IG. there is k’ > K . If {(ailbi)} is a sequence of disjoint open intervals in [0. (ii) Suppose that there exists no such that Gn = 2‘‘ 2 2 8 for n 2 n o . then < & + N C(bi U i ) < & + . Solution. suppose that for any natural number K .1] iff there exists no such that (IndianaPurdue) E > 0. n 2 no. x” E [0.11). .11).x”1. Then .a i ) < 6. + a . &  2 which means f E AC([O.
let us first establish the following equality provided that either side is finite.308 4408 For what values ‘a’ and ‘b’ is the function (i) of bounded variation in (1. (Iowa) Solution.l). 1 and 1 E > 0. since f is continuously differentiable on (0. for any of bounded variation on [ E . Indeed. The function f is of bounded variation if and only if To show (l).1]. f is Then 1 E 1 1 . (ii) defferentiable a t ‘0’.
b < 0.309 Case I. 1& + 1. 0<€<1 00 E Case 11. while a 2 0 implies that sup O<e<l J' lf'(x)ldx = sup J' . Case 111. b = 0.:l!al € dx < oc). for Since a + 6 = 0 and a + b > 0 imply respectively that and ' the function f is integrable on (0. Then sup O<E<' If(&)I < iff a + b 2 0. there is a 6 that 1bl2 Let N = > 0 such that 0 < x 5 6 implies d3 . since and . b > 0.2 I4 . Then sup 0<€<1 If (&)I < 00 iff a 2 0. If a + b 5 0.> Ibl xb .1]. Then b1 If a + b < 0.
2 ~ with f’(z) = 0 a. Since by the condition that If(.27r] and f’(z) = 0 a. = +m. 4409 Prove or disprove a) Let f be a real function on [0.. b) Let f be a real valued function of bounded variation on [ 0 . a) The conclusion is ture. Then f must be constant.e.310 I’ and If’(.e.27r] satisfying .yI.27r]. then and and therefore f’ is integrable. (Stanford) Solution. on [0.f(y)l <: 1 . we have rl If a + b > 0.y E [0. .)ld. (ii) f is differentiable a t 0 if and only if { %>l or { b<o a > 1. f is constant.) f is absolutely continuous. all 2 2 . ] Then f must be constant.
c such that f ( u + t )= J. but f is not constant. g E L1(IR).( t ) d t cc oo . without loss of generality.e. (S t a n f o d ) Solution. c ) if then there are constants a . Show a1 b) if a+h h+O f(z)dz = c.? State explicitly the theorems you use.31 1 b) is false. it is differentiable almost everywhere by Lebesgue’s Theorem. Then f is of bounded variation with f’(z) = 0 a. Can you deduce that f’(x) = g(z) a. 27~1.. Let f(x) = 0 on [0. a+t g(z)dz +c a. T ) and f ( x ) = 1 on [T. 4410 Let f . Since the function is of bounded variation.e. 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.e. Assume. that f and g are realvalued.
e. .f ( x ) d Z h = f ( a + t ) . there is an a E BZ and a c such that I ra+h J a k. c) By a).c a. f ( z ) d x = c.312 We have Letting c t 0. t E IR. . we see that and therefore b) We have by a) h0 lim f(' a + h.
e. (Iowa) . It follows that f’(z) = g(z) a. we have = 0.. < 2 of ( a . 1.b) such that F(z)= lz fdm ( m the Lebesgue measure).’.b). (1) Prove that the following two statements are equivalent: (a) There is an f E L2(a. 4411 Let F : ( a . (b) There is an M > 0 such that for any finite partition zo < z 1 < .e.e. t E IR.b) + Dl be measurable..313 By b) and by the assumption. a.. (2) Show that the smallest constant M in (b) is equal to Ilfll. .
En are any mutually disjoint measurable sets. o ( a i j .b). a t most countable). n E lN. then for any E > 0 If Ei's are open sets of ( a . .b ) . (1) (a) 3 (b).b ) E such that f F(x) = LZ f(t)dt + c. and therefore there is an integrable function f : ( a .. for any mutually disjoint intervals (a. Since a F is absolutely continuous. .314 Solution. 2 E (a.b).. say Ei = 3EAi . . & ) (A.we have on)of ( a . . We will show that if E l . By the CauchySchwarz inequality (b) 3 (a) Obviously.
and lim p ( E . .Enj and passing to the limits we see that (3) holds in general. = {x Then M I < 6}. . If Ei’s are compact sets there is for each i a decreasing sequence { E i j } E 1 of open sets of ( a . . . Applying (3) to the 300 mutually disjoint compact sets Elj. . To show that f is square integrable.=I j € A i P i j . . there is for each i an increasing sequence { E . it suffices to show that 00 where En = {z In 5 If(z)12 < n + 1). j } E 1of compact sets such that j=1 u Eij 00 E. . j ) = m ( E . ) . f I 2 . b) such that Ei = j=1 Eij and Eil n E k l = 0 (i # k ) .315 then by inequality (2) I 2 00 It:. In general. . I 65 f ( z ) < 1f ( z )5 and E. . .a i j 5 M. En. . Enj and passing to the limits we see that (3) holds for E l . Let E = {z ’ . Applying (3) to the mutually disjoint open sets Elj. .
316 5 M. (2) Define. . z i ) (xn = a k ( b . b )by (3). Moreover 5 l l f l l n by (1).b]. For any fn  f E L 2 ( a .b]. IlfII = nlimO llfnll = 40 llfll = Ilfll2 for any f E C[a.b ) there is a fll2 f 0.a ) ) such that + Then Illf1I2.11 f E C[a. there is is a norm on L 2 ( a . It follows that sequence {fn}rz1 b] such that 11 of C[a. For any E (zip'. This completes the proof. then 11. for each f E L 2 ( a . so We have Ilfnll2 = Ilf112.b ) .
Y and q5 on M by 1 E Ida and $ ( E )= IdP. 2 p ( E )= and ] 2tdt r E + ( E )= We have 1 E 3t2dt. A. For any Borel subset E of X . .e. t E E @ X(E) = 0 e 3 t 2 = 0 a. Justify. M the calgebra of Borel subsets of X. Let a ( t ) = t 2 . where X the Lebesgue measure. p(E)= 0 It follows that p 4  e 2t = 0 a.317 4412 p(t) = t 3 and define measures p ( E )= Let X = [0. (Iowa) Solution.e. Does 4~ exist? Does exist? Compute the value of the RadonNikodym d? derivatives that exist.1]. t E E e X(E) = 0 e q5(E) = o . We have and 4413 Let X be the Lebesgue measure on R2 and R defined by: 2 /I and v the Borel measures on .
i. (Iowa) Solution. Is p If so.e.2n)n A ) = 0 and therefore p ( ( n . i.. << p2 or/and p2 << pl? (Iowa) . Define for any Lebesgue measurable subset E of [0. Because p ( A ) = 0 implies for all n E N .in].pn) n A ) = 0. m]. find the corresponding derivatives. 4414 Let p be the Lebesgue measure on [0. m].318 Is p << v or u << p? Find the corresponding derivaties if they exist. X((n. Since and we have " 1 du n=l and dv d X Therefore (1) holds. however. let A = 21 and then v ( A )= 0 but p ( A ) = We show next that (g. It is true that v << p and it is false that p << v. v ( A ) = 0. = n=l c " 1 $~(n.
4415 Let C#J : l + l be a bounded.39 1 Solution.e. and assume g E L(lR).)) E L(lR) for all t E R).. Moreover From p2([0. . Define f(t) = J dJ(tg(z))dz. that q5(tg(. t E IR.1)) = 1.1)) = 0 and p([O. l R a) What additional assumption on 4 will insure that f is well defined? (i. p << p2 does not hold. Since r 1 we have It follows that both p1 << p2 and p 2 << p1. continuously differentiable function with a R R bounded derivative.
and The conclusion follows from the Dominated Convergence Theorem.e. Assume 4(0) = 0 then f is well defined. for any t . (Indiana) Solution. for any s.a. Prove that . Indeed.l). 4416 Let ( f k } denote a sequence of nondecreasing functions defined on ( 0 . a. l ) with the property that lim f k ( Z ) = 1 for almost every z E (0. lim (bn . i.) = 1 .} and {b. (Indiana) Solution.} such that 0 < a. t E R (s # t). z there exists an s with 0 < s < 1 such that Since b.. We will show that Indeed. < b. show that f is differentiable. nco < 1. I). There are sequences {a. f’(t) exists for all t E IR.320 b) Under the additional assumption in a) above.fi(z) = O for almost every lim krn k+oo z E (0.
lim fL(z) = 0 for almost every z E (0. .l).321 and k+m lim fk(a. is integrable and I’ ks Hence k+co f i ( z ) d z = 0.) = lim kr 00 fk(bn) =1 It follows that lim krm f.
?. 4502 Let { I a } . Show that is countable.I for every finite subset {!I. x can not be in 1. .). Take n such that but : . S n (n.+S. . be a collection of closed intervals in R.322 SECTION 5 MISCELLANEOUS PROBLEMS 4501 Let S c R be a set of real numbers with the property that IS1+. i) be finite and n ) and must S is countable. there must be a (Y such that m x E la. Since S n (i. . ( IndianaPurdue) Solution. Sn}c S. Show that S is countable.". . Obviously u1 : aEr is an open set. We have u ma 00 : = 1 z € UI%P. n=l where cyn. For any r UIa\UI.pnTU I:. 51 (IndianaPardae) Solution. .Q E r. .
It follows that the map. Let {Ui I i E I } be such an infinite set. Moreover. i H ai of I to D is injective and therefore I is countable.323 If I .where D is a countable dense subset of X . open sets in a separable metric space X is countable.} o f Z such that nm 2 1 lirn (kn. 7~ 2 .. there are sequences {kn} and {I. since {Ic: . (Stanford) Solution.27r) nm I . 6 Icy. Thus we must have x = on or P.21.21 I Ic.which is a contradiction. Let 2 A = {x E (0. For each i E I there is an ai E D such that a E Ui.) = . c ( a n . lim sin(nx) = 1. 4504 Prove that for almost every x E [0.is irrational}. 4503 Show that any infinite set of nonempty. for any x E A . Indeed for any x E A . ( st anford) Solution.27r] n+co lim sin(nz) = 1. Then ai # aj i whenever i # j .. P n )then x . 7F Then A is a measurable set of measure 27r. I E a } is a dense subgroup of lR. mutually disjoint. Therefore a a n which implies uIa\ u1 : a a is countable.
'for every z E I . {k. + 1)w) = 1. + f . ( IndianaPurdue) Solution. For any n.11). Suppose that there is a sequence of odd polynomials pn with p . + 1))= 2 and therefore n+m lirn sin(3kAz) = lirn sin(3khx ncc + 2(31. n+cc Otherwise lirn (3kA) = +co and ncc lirn ((3kh)n+cc 2 T 1 + 2(31. Show that there is a sequence of odd polynomials p n ( z ) with pn + f uniformly on [O.2 1 . = $00 then nco ncc lim sin(khz) = lim sin(khz .324 Since f 4 { k : . . take a polynomial pn such that 4506 Let f E C([O.1] iff f(0) = 0.} admits a subsequence {k. ~ )= 1.} either increasing to t 0 0 or decreasing to m. ( IndianaPurdzle) Solution. nw . 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 .I EZ}. If lirn Ic.2 1 I k. Then f ( O ) = lim pn(O) = 0. 4505 Let f E C ( I ) .
(1) .325 Conversely. I]. with the supremum norm.1 1 1 1 r+’ is a contracting mapping on C[O. Then f is a continuous function on [1. b) Show that there exists one and only one smooth function f on [0. Set f ( z ) = f(z). (Stanford) 1.. 2 E [1. there is only one function f E C[O. ( z ) with p . 11. + f uniformly on [1. 12 f(t)dt. Set 1 pn(x) = [pn(x) 2 ~n(x)]. b) By the Banach’s Theorem. f(z) = es + f J. Take a sequence of polynomials p . We have uniformly on [0. 4507 a) Show that the mapping I : C[O.1].1 + C[O.1]. suppose that f(0) = 0.0]. i. ex + zf(x2).1] satisfying the conditions: { f(0)( z=) = &f Solution. a) For any f and g in C([O. Then Pn is an odd polynomial for any n.e.l]).1 such 1 that I(f) = f.
. by (ii) there is an N1 such that If n > N1. n = 1. 4508 Given f : R + R bounded and uniformly continuous and {K . For any E > 0. E L1 such that (i) llKnlll 5 M < 00. For the above 6. K . 3 . . then i. take Nz such that . by the uniqueness of f. (UC.Y) holds for all 3:  f(. n = 1 . Take M1 > 0 such that If(z)l < M I for all z E R. Thus Take 6 > 0 such that lf(z . lKnl Show K .. .e. I ( g ) = g .)l < 4M & E R any IyI < 6..2. If g is another smooth function satisfying (2). (iii) J12:121. . . * f + f uniformly.3. 2 ..61 + O as n + 00 for all 6 > 0. }..+ l a s n . . Irvine) Solution. (ii) ~ K .g = f. .326 By (1) f is smooth and satisfies the conditions: f(0) = 1.+ o o .
E < 1..N2). co) be upper semicontinuous and define m : R [oo. (UC.oo) m ( z )= liminf f(y). (a) For any zo E S". It follows from (l).327 & < . (b) Show: If I is an open interval contained in S. then m ( z ) = aon I.&z . .m ( z ) 2 1). there exists E > 0 such that f(z0)  m(z0) < 1 . Irvine) Solution.(2) that < holds for all z E R if n & & +=& 2 2 >N 4509 Let f : R + (03. . Let S = { z I f(z) . (c) Show that S is nowhere dense.+ M . & 4 4M 2 Set N = max(N1. by YX + (a) Show S is closed..
. (b) Suppose that there is zo E I such that rn(z0) > m. As in (a)..f(. we obtain a sequence of open intervals { I n } such that 7. i.) < 1). 2 E S". Denote i + i. . Therefore m ( z )= liminf f ( y ) Y+Z 2 rn(zo).there must be F E O(z0.m(2)< f(20) + .m(5)< 1.d ) is a metric space.< 1 2 2 & & +& = 1 if z E O(zo. From (b).o) + .. () (c) Suppose that S is not nowhere dense. .. 4510 Prove that any topological metric space is homeomorphic to a bounded metric space.6). we can find 6 > 0 such that m ( z ) > rn(z0). ? which is a contradiction. Take an open interval I1 such that 7 c A l .6) such that f(5)< rn(z0) Therefore f: .328 Take 6 > 0 such that for 2 E O(z0.e. . m ( z )= m for z E I . Suppose that ( X .l . Z ) . i. So S" is open. we can take an open interval I2 satisfying 7 2 c I1 and f(z) < 2 for z E 7 2 . (Stanford) Solution. c I n .e. f(z)< n for z E 7 Obviously f ( z ) = oo for 2 E n .6). Define another metric d o n X by It is easy to show that the identity mapping of X is a homeomorphism of ( X . A1 is a nonempty open set.S). By the definition of rn(z). In the 1 same way. Al = {Z E I. 5 6 S which contradicts % E I c S.m(. 6) c I .d ) to ( X . there is an open interval I c S. Since S is closed.& & holds for z E O(z0. Thus f) ( .for 2 E O(z0. By induction.
Stony Brook) Solution. (i) Show that A is a continuous bijective isometry. z) E C x 2R The isometry A : M +M I 1 . (Stanford) Solution. ) = +00. Prove that f attains a (SVNY.z) = (e. A' is distance nonincreasing. ~ ( z . A o A o A = Id and d ( A z . is a periodic map of order 3. Therefore A is an isometry and therefore continuous. 1). which has exactly two fixed points. Let (Y = inf f ( H ) . l ) and (0. having exactly two fixed points ( 0 . There is a sequence {zn} of H such that nco lim f(zn)= a .*'z. defined by A ( z . ~ 1+ z2 = 13. suppose A : M + M is .329 4511 Let M be a metric space with distance function d. + t B be a continuous convex 00. If SUP llznll = $00. a distance nonincreasing periodic map of order 3. y).Ay) 5 d ( z . then a = lim f ( z n .. z). 4512 Let H be a Hilbert space and let f : H function such that f ( z n ) + oa whenever llznll minimum. too. .e.}F=~ that such krco l/znk (1 f +m. (i) Since A' = A 2 . n there is a subsequence {Z~. (ii) Let M be the 2dimensional sphere M = { ( z . i. z) E M . periodic of order 3. (ii) Give a n example of a complete metric space M and an isometry A on M .
A*A cannot be the identity on H.) (Stanford) Solution.y E H . (You may wish to use (a) prove this. Then n+m {fn} converges to f in measure and therefore by F. 4514 Let A be a bounded linear operator on Hilbert space H .e..330 a contradiction. It follows that I ( ) 5 fzI 1 a. (a) Indeed. 4513 fzI A is the subset of L1(BZ)consisting of all functions f satistying I ( ) 5 1 a.e. on lR. So sup llzn[l < +m. 1 k >N} W = coV{znk I k > N } 11. Recall that the adjoint A* is the unique bounded linear operater on H such that ( A z . ball of H . exists in L1(BZ). f ( y ) < .y)I llYll51 114151 5 IlAll . By the weak compactness of the closed converging weakly to a point z E H .11 Therefore and for any y E cov{z.B. where IlAll is the norm of A.. It follows that a is finite and f attains its mininum at z. on aZ.. there is a subsequence { z n k } For any p > a there is an N E lT such that for any k > N . ( StunfoTd) Solution. h Since X E n * cov{z.Riesz Theorem there is a subsequence {fnk} converging to f almost everywhere. Prove that A is closed in the norm topology of L1(lR). I k > N } . f ( x n k ) < 0. (b) Show that AA* . one has f ( z ) 5 p. If {fn} is a sequence of A such that f = lim f. A*y) for all z. (a) Show that IIA*ll = IIAll. 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. y ) = (z. So A is closed.
A*A = I then = + It follows that I)A))2 JJAJ12 1. we will show that Indeed equalities (1) follow from the following inequalities If for some bounded linear operator A on H . (b) First. . Show that IR are Lebesgue measurable. 4515 Suppose that A . AA* . with m(A) > 0. Assume without less of generality that A and B are compact sets. Since = m(A)m(B) > 0.33 1 and a fortiori IlAll = 11A*\\. a contradiction. (Indiana) Solution. B c 0. m ( B ) > contains an interval of positive Lebesgue measure.
N Yes.332 there exists an zo E IRsuch that m((zoB)nA) > 0. of X such that p(X\F. . [p]? Justify. . + l R such that b]. Then (1) i=l is for f = f on X.B ) n A ) > 0) CA + B..e. For any n E l there is a compact subset F.B ) n A > 0) is nonempty and open. 4516 Let X be a compact Hausdorff topological space and let p be a finite regular Bore1 measure on X .. Then lim f = f a. the set { z I m((z .) < and f is continuous on F. 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. Let X . .e.) p(X\X. Since z H m((zB)nA) is continuous. = 00 u Fi. nco < i. n0 (Iowa) Solution.. and (3) p (2 each n E lN a realvalued continuous function f : X .) f is continuous on X. . Then the conclusion follows from the following inclusion { Z I m((z .
Part V Complex Analysis .
.
a n+m . which is analytic in R. but f'(0) doesn't exist. then both f and g have a t most countably many zeros in R.335 SECTION 1 ANALYTIC AND HARMONIC FUNCTIONS 5101 TrueFalse. A counterexample is f(z) = sin&. Then f(z) . uy. 6. and the partials u. (c) True. + m. Rez > 0 } . But f ( z ) is not bounded in R. z n has radius of convergence exactly R. A counterexample is f(z) = e". If neither of f and g is identically zero in R . Because lim G = it follows from 1. then f E 0. a t z = 0. (IndianaPurdue) Solution. then C n 3 a . then If(z)I 2 1 ( z E R). = vy and 1ly = 21.v. (c) if f and g are analytic in R and f(z) . and continuous on with If(iy)I E 1. quote a relevant theorem or reason. f satisfies Cauchy(a) False.g ( z ) is not identically zero in R. (b) if f(z) is analytic in R and has infinitely many zeros in R ... . if false.2. A counterexample is f ( z . g ( z ) E 0 in R. then either f E 0 or g 0. (a) if f(z) = u iv is continuous at z = 0. But f is not identically zero in R.and has zeros z = 1 n = 1. If the assertion is true. f is analytic in R = { z : IzI < l}. then f'(0) exists. (f) sin f i is an entire function. vy exist at z = 0 with u. (b) False.. (d) if f(z) is analytic in R = { z . give a counterexample or other justification. (d) False. (e) True. (e) if C a n z n has radius of convergence exactly R. continuous on with If(iy)I 5 1 (co < y < +co). and the zeros have no limit point in R. y ) = Riemann equations a t z = 0.
336 (f) False. 5102 (a) Let f(z) be a complexvalued function of a complex variable. then F ( z ) is analytic in 1. (Stunf07d) Solution. Actually. sin f i is not analytic at z = 0. z = 0 is a branch point of sin &. What can you conclude if this holds with equality. prove that f is analytic there.( < 1).(’fI and that which implies that f ( z ) is analytic in R. (c) Determine all entire function f that < lf(z)I. If both f ( z ) and z f ( z ) are harmonic in a domain Q. Show that If(0)l 5 u2. (a) It is well known that the Laplacian can be written as [). When 17iz IzI = 1. (b) Suppose that f is analytic with If(z)I < 1 in IzI < 1 and that f(iu) = 0 where u is a complex number with 0 < la1 < 1. 1+az hence .
(a) Let 2 i e Because u is a harmonic function. hence P+iQ=i is analytic on G. we have F ( z ) f ( z ) = a.i% is an analytic function on G. Integrating on both sides.)[ 5 1 for IzI < 1.’ zu 1Tiz z + a . where c and c’ are constants and JcI< 1. ay2 We also have ap .=aQ  ay ax axay azay = 0. we obtain logf(z) = cz d..y ) satisfy the CauchyRiemann equations. y) and Q ( x . (b) If has a primtive. (b) Show that u has a harmonic conjugate on G if and only if has a primitive (antiderivative) on G. the integral 2 i g .+ . I $$I + % 5103 Let G be a region in a‘ and suppose u : G t lR is a harmonic function. we know that f has no zero in a‘. 1+zz (c) From lf’(z)l < lf(z)I. au ax au ay aZU azu So P ( x . Take z = 0. f(2) I I < 1. elel which is equivalent to When it holds with equality. we have ap ax aQ ay  aZu + ax2 ~ = 0. which implies that fIIl is also an entire function. (a) Show that . we obtain If(0)l I bI2. Hence f (z) = c c’e‘”.337 which implies that IF(. then for any closed curve c c G. (Indiana) Solution. It follows from < 1 that = c.
Prove that u iv must be analytic on 0.ik and f2 = aY aY are analytic functions on a. ay ax For any closed curve c C = ld(u Hence + iv) = 0. y) is a harmonic conjugate of u(x. The reason lies on the fact that the real and imaginary parts of fi and f2 satisfy the CauchyRiemann equations respectively (see 5103 (a)).au _ . % . y) on G. + (IndianaPurdue) Solution.iau Because u and v are harmonic functions.__ ax dy’ . z E G. Hence we can define a singlevalued function where zo. then dv = dx av + dy aV ax aY G . fi = . On the contrary. we have = dx dU dU + dy. 2 . and the integral is taken along any curve connecting zo and z in G . . av au ay ax’ we know that v(x.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. if u has a harmonic conjugate v on G.338 It follows that Kdx dU du +z d y =0 holds for any closed curve c c G. Because av .
au av _. and let f be holomorphic in a neighborhood of Q . which implies that u + iw is analytic on Q.1] x [0. by the uniqueness theorem of analytic functions. It follows from fi = i f 2 for z E R that for z E Q.f ( z ) is real and + i) .1] c a‘ be the unit square. (Indiana) Solution. Hence fi = & . we have As . Suppose that f(z f(z + 1).i] 2 0 for z E [0.i a u = if 2 = i ( g . Because the subset S has a limit point in Q. 5105 Let Q = [0. ax dy’ du dy av __ ax when z E S c Q.1]. we know that f1 = if2 holds for all z E Q. Show that f is constant.f ( z ) is real and 2 0 for z E [O.339 By the assumption of the problem.i $ ) bu & when z E S. by Cauchy integral theorem. Because f is holomorphic on the closed unit square Q.
Define F ( z ) = f(z)dz. which implies that G ( z ) is a constant of modulus 1. which is holomorphic in a and satisfies ‘ lf(z)I I %${If (. For z E we choose the integral path on dS and consider the differential form f(z)dz in the integral. Hence f ( z ) can be analytically extended to a doubleperiodic function by f(z) = f ( z 5 1 and + 1) = f ( z + i ) . + f(z)dz = (udz . then si as.340 for 0 5 z f(1+ Yi) . . 5106 Let f be continuous on the closure 3 of the unit square s = { z = z + i y E @ ’ : 0 < 2 < l . Then F ( z ) is analytic in S and continuous on S. Hence f ( z ) = F’(z) G 0.f ( Y 4 2 0 for 0 I y 5 1. It follows from G(z) = eF(’) that F ( z ) is also a constant.Then G(z) is analytic in S and IG(z)I = 1 when z E 85’. by comparing the real and imaginary parts in the above identity.prove that f = 0 everywhere on S. Because G ( z )has no zeros in S. (Indiana) Solution. where the integral is taken along any curve in which has endpoints 0 and z. we obtain IG(z)I 1 for z E S.vdy) + i(wdz + u d y ) . Hence we obtain Re(f(z)dz) = 0 on dS which implies ReF(z) = 0 when z E dS. o < y < l}. Let f = u i w . On 3 n ({y = 0) U {y = 1)) we have u = 0 and dy = 0. Apply the maximum modulus principle to both G ( z ) and l/G(z). and let f be analytic on S. so l/G(z) is also analytic in S and 11/G(z)l = 1 when z E dS. 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 on sn ({z = 0) u { z = 1)) we have v = 0 and dz = 0.)I1 < This shows that f ( z ) must be a constant. Let G ( z ) = e F ( * ) .and if If = 0 on S n ({z = 0) U ( 2 = l}). If Rf = 0 on sn ({y = O } U {y = l)).
zd.4 with residue equal t o z Hence when c = f ( z ) is holomorphic in a neighborhood of z = 1.4 in { z : I z J5 l}. + 5108 Let P ( z ) be a polynomial of degree d with simple roots z1.lz12  lz1 1 0. A “partial fractions” expression of $ has the form: . &. lz4 + z3 + + z2 . . (Iowa) Solution.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 . (b) Show that the function f is holomorphic on an open set containing the closed disk { z : IzI 5 1). we obtain that f ( z ) has no singular point in { z : IzI 5 l}. .341 5107 (a) Find the constant c such that the function 1 f(z)= z4+z3+22+24  C z1 is holomorphic in a neighborhood of z = 1.41 2 4 . . (b) When IzI 5 1.4 22 1 + + 22 + z .lz4 + 2 + + z~ 2 4 z2 144  lz1 3 . hence f ( z ) is holomorphic on an open set containing { z : IzI 5 I}. z 2 . we have A.+ +1 + z . which shows that z = 1 is the only zero of z 4 + z 3 z2 + z . By (a). (a) As 241 lim(z . and the equalities hold if and only if z = 1. .1) = lim =  2 4 1 (z4 .
zn  1 P(z) P’(zn)’ which is the residue of 2. (4 c n = lim z+. P(.) (b) Let Then f ( z ) is analytic o n c and lim f ( z ) = 0.zm z . f(z) is I 4 0 0 identically equal to zero. (Courant Inst.342 (a) Give a direct formula for cn in terms of P.. Then the Laurent expansion of 2 a t z = z1 is P ( Z ) Hence 1 the form: has P(. () (c) Give a formula similar to (*) that works when z1 = z2 but all other roots are simple.) .Z1)n.a t z = z. (b) Show that 2 really has a representation of the form (*). hence (c) Denote the Taylor expansion of P ( z ) at z = z1(= z2) by P(2)= c 00 a n ( z . By Liouville’s theorem.) Solution. p.
~1 < + ~ 1 C ) C\ UKj j= 1 n * .. 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.ZB1.z l ( 5 rl}.a ) ( l . show that there is a complex constant A and a real constant B such that Az2 + Bz + z (Indiana) f(.. .. K2. .Tia)’ where A = 2 1 . Kn be pairwise disjoint disks in C. such that C1 = { Z. then B1 is real and  Az2 + Bz + z ( z . Assume that the residue o f f at z = a is Al. n n f1.Aizz 1z z.) = ‘(z . Assume K1 = { z . Choose c1 > 0 sufficiently small. 5110 Let K1. It is obvious that g ( z ) is analytic on D and g ( a D ) C lR. Define g(z) = f ( z )   Ai za .a ) ( l . B = (ZAl + a z l ) + B1(1+ (at2)E lR. and j=1 U Kj. fn such that (b) f ( z ) = C fj(z) for z €C\ U Kj. j=1 n (Indiana) Solution. fz. By the reflection principle.Tiz) * Solution. g ( z ) can be extended to an analytic function on the Riemann sphere hence g ( z ) must be a constant. If f ( a D ) C IR. . Suppose g ( z ) B1. TI < )Z . .. ( a .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.
we get a function f ( z ) . Choose ~2 n = { z .221 5 72). and the above identity holds for z E @\ U Kj. Because f ( z ) .f ( z ) has the Laurent expansion f(z) = c +m up(z .fl(z)fZ(Z) . consisting .~ (2) 2 ) f2(z) ~ . > 0 sufficiently small.f n . Iz . Because f(z) K2.* * * . 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).I)? k=m Set fl(Z) = c 0 k=oo u p ( z . which is analytic in @\Kn. is analytic in@\KZ. @ u {m}}.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 . f ( z ) . such Laurent expansion 0 Set f2(z) = k=m uk ( z .344 In C1. Set Then we have n j=1 where fj(z) is analytic in@\Kj.l ( z ) . Repeat the above procedure n .z1)5 fl(z) is analytic in @\K1.fi(z) has an analytic continuation j=2 to K1.1 times.
p i ) n p i and m Then n n n m and m m m n . Stony Brook) Solution. i=l n Similarly. is a zero of f when npi> 0 and a pole of f when npi < 0.m) be all the zeros and poles of g ( z ) m with multiplicities mqjrespectively. 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. . n) be all the zeros and poles of f(z) with multiplicities npirespectively. . 2 . Let f and g be two rational functions with disjoint divisors. Prove that (SUNY. we have  1 . . It should be noted that p. . By the property of rational functions. then f n f ( z ) = A JJ(z .345 counted with their multiplicities np E 2.2. C npi= 0. let q j ( j = 1 . Let pi (i = 1. . then we have j=1 mqj= 0.
Then n1 00 f(z) = A n ( z i=l and g(z) m =B n(~ m  4j)"'j.346 In case the point 00 is a zero or a pole of f or g. + + . 5112 Let f ( z ) be the "branch" of logz defined off the negative real axis so that f( 1) = 0. we may assume that g(p. simplifying the (a) Find the Taylor polynomial of f of degree 2 a t 4 coefficients. Hence n. j=1 Since C mpj = 0.1 i=l n1 m and i = l j=1 which completes the proof of the problem.) j=1 = g(m) = B . 3i. = without loss of generality. (b) Find the radius of convergence R of the Taylor series T ( z ) of f ( z ) at 4 3i. we may assume p .
the radius of convergence R of T ( z ) is equal to the distance between z = 4 3i and z = 0.3i( < 5. T ( z ) # f(z).3i) + cg(z + 4 .347 (c) Identify on a picture any points z where T ( z ) converges but T ( z ) # f(z). we have = l0g[(4 + 3i) + (Z + 4 . Hence R = 5. (Minnesota) Solution. 5 4 =  + 3i 25 ' 25 24i 1250 ' and c2 = + (b) Denote the Taylor series of f(z) at 4 3i by T ( z ) . or a general impossibility? Explain and/or give examples.3il < 5}. Hence + + + + + + .arcsin ). It is because T ( z ) in R is the continuation of logz at 4 3i in the disk { z : Iz 4 . and has no other singular point.5. (a) When z is in the neighborhood of zo = 4 f ( z ) = logz + 3i. When z E Cl = { z : Iz 4 . If there are no such points.1 by a.01. Because log z has only z = 0 and z = 00 as its branch points..arcsin ) + 5 4+3i 1 z+43i 5 .3i)] z+43i log( 4 + 3i) + log 1 + [3 z4+ 4+3ii 3 log 5 + i(7r . and describe the relationship between f and T at such points.Imz < 0 } .* Hence the Taylor polynomial of co f of degree 2 at 4 + 3i is + c1(z + 4 .. co = log 5 c1 where 3 + i(7r . while f(z) in R is the continuation of log z a t 4 3i in the slit plane @\(m. (c) Denote the shaded domain shown in Fig.3 4 2 .4 + 3 i I ( ) 2 +. is this something special to this example.
We have Arargf(z) 1 1 1 + Ararg(z + 1) = (27~)+ 2( .2 by I?. Express g in terms of f .). Fig. and denote the change of $ ( z ) when z goes along r from the start point to the end point by A.$(. (Indiana) Fig.e. f.1 5113 Let f be the analytic function defined in the disk A = { z : Iz .41 < 4) so that f ( z ) = z i ( z + 1)i in A and f ( z ) is positive for 0 < 2 < 8. T(z) = f ( z ) + 27ri.l(argf(z)+Arargf(zr)).5.2 ~ ) 3 2 Hence . An analytic function g in A is obtained from f by analytic continuation along the path starting and ending at z = 4 (see Fig.5.348 the difference is 2ai.5..5.2 Solution. Then g(z) = Ig(z)leia%dz) 1 = Arargz 3 = ~() I. Denote the closed path in Fig.2). i.
349 5114 Define eJ. Let r = { z : IzI = T } . When z is in the small neighborhood of z = 0. f i is changed to fi. (a) It is known that z = 0 and z = 00 are the branch points of function &. . .=25 (Indiana) Solution. .6 . . 2 . . while f(z) is changed to e . : (b) We have which shows that z = n2r2are simple poles of f ( z ) with residues 2 n ~ ( . efi f(z) = (a) Where is sin& * f singlevalued and analytic? (b) Classify the singularities o f f . (c) Evaluate f(z)dz. } . It is obvious that z = n2R2 (n = 1 . Hence f ( z ) is singlevalued and analytic inC\{z = n2r2 n = 1 . 2 . 4x.eJ. Hence z = 0 and z = 00 are no longer the branch points of f(z).) are poles of f ( z ) .e& sin(&)  e f i . . and when z goes along I’ once in the counterclockwise sense.l ) ~ ( e = ~ enn). . sin f i which is still f(z). 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 ) .
y = 0) deleted. Why? (IndianaPurdue) Solution.arg(l . y = 0) inside I? .r#J(z)the change of r#J(z)when z goes continuously along I counterclockwise once.z ) ] = 2 2 5 z 5 1 . Which of these branches are (is) the derivative of a singlevalued holomorphic function in 0. Hence f(z)dz = 2 ~ i R e s ( fr 2 ) . For (b) g ( z ) = can we choose singlevalued branches which are holomorphic in fl. and hence is a nonisolated singular point of f(z).350 As to z = 00. 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 . y = 0) not inside r 5 z 5 1 . it is the limit point of the poles of f(z). It is ' known that f and g can be represented by and  g(z)  4 1  . and denote by A. (c) f ( z ) has only one pole z = 7r2 in the disk { z : IzI < 25).{+log(l+z)+log(ll)} Because and 1 Ar[argz .arg(l 2 +z) 0 27r  1 arg(l 2 (1 (1  z)] =0 1 1 Ar[arg(l + z ) . Let I? be an arbitrary simple closed curve in a.
By lifting the integration to a n appropriate covering space. i = *(1 )T 22 1 1 1 b3 b5 =fi . it is obvious that f (z)dz = 0 and Sr g(z)dz = 0. el5 E (e4A.. d z (HUT7NZrd) Solution. (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. Hence log el5 = 1+ (14 4ai) = 15 4ai.*1.1 = 14. (a) The set {ee+ie I 6 E IR} U (0) is a spiral which intersects the positive real axis at {e2nn : n = 0. we consider the integrals 1 f(z)dz and 1 g(z)dz. oriented counterclockwise.*2.. Hence we obtain that both of the singlevalued branches of f are the derivatives of singlevalued holomorphic functions in Q.y = 0) is inside I?. we consider the Laurent expansion of f and g about z = 03: s .) . and A r l o g IzI = 15 . If the segment (1 5 z 5 1. Hence loge15 = loge A r l o g z . f(z) g(2) = *i(l z  )5 22 1 1 = f i (1 + a2 + 7 + * * 7 a4 . 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). If the segment (1 5 z 5 . e2=). But neither of the branches of g is the derivative of a singlevalued holomorphic function in a. Because e E (e'. . e6=). where the integral is taken along any curve connecting zo and z in 0. 1. give a precise meaning to the integral J'(1og ~ ) ~ and find all possible values which can be assigned to it.. (b) Let y denote the unit circle. we know that when I connects e and el5 in D. + + + . Find log el5. Because . and each of f and g has two singlevalued branches.y = 0) is not inside I?.+ T +. Let log be a branch of the logarithm on D such that loge = 1. a ) .351 we have A r f ( z ) = 0 and Apg(z) = 0. In order t o know which of f and g has a singlevalued primitive in R. z 23 zi It follows that f(z)dz = 0 and lrg(z)dz = f 2 a . The singlevalued branch of logz is defined by loge = 1. Hence both f ( z ) and g ( z ) have singlevalued branches which are holomorphic in Q. It follows that A r l o g z = Aplog IzI + iArargz.+ . 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. Justify your answer. ' Arargz must be 4a.
we know that when f(1). Hence the most general harmonic function of the form f(1).352 both the starting point of y and the singlevalued branch of log z on y can be arbitrarily chosen. Because f ( I . and the solution is f ( t ) = Q log t + 0.Pare two real constants. it must be a constant. in C\O is Q log IzI 0. z E C\O. + . 1 ) is harmonic. Which of these f(I. 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 . we have reason to assume that the function f (with real variable t ) has continuous derivatives f’(t) and f”(t). the segment on wplane can be denoted by [it. where t can be any real number. where (Y.47r2) . Note that the Laplacian A=+=4ax2 ay2 a2 a2 a2 azaz7 and we obtian where t = IzI. This differential equation is easy to solve.i(t+ 27r)). Since log IzI has no singlevalued harmonic conjugate inC\O. 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}. has a singlevalued harmonic conjugate in C\O.1) have a single valued harmonic conjugate? (Indiana) Solution. + 5117 Find the most general harmonic function of the form f(lzI).(2wew)lit+ Z U ) + 2 1 t i(t+la) = 47~(t+ ?r + i)eit = 47r(t + 7~ + i ) e i ( t + 7 r ) .
. .5. What is the value of u at the origin? Justify your claim.5. and the ten segments of the boundary by I l l I z . 10. . .3 Then denote the harmonic function on D with boundary values 1 on l k and u k ( z ) . 110. By the symmetry of 0 on the rest of the boundary by domain D. (Stanford) Solution. Denote the interior domain of the pentagram shown in Fig.. . we have It follows from in . Fig. put in order of counterclockwise. . . 2 . k = 1 .3 by D . 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.353 5118 Consider the regular pentagram centered at the origin in the complex plane.
h ( z ) is harmonic there. Since d m is analytic in C\[O. and h : G + lR is continuous. If the boundary values satisfy u = v on y and the radial derivatives satisfy = on y. prove that u = v in D. [0. h l . then h must be holomorphic on G. 11. The answer is No. 0 5 2 5 1. dmlz=z lim z=x+yi+z &zij=&cT)i. But h ( z ) is not harmonic on&. Suppose that u and v are harmonic in D = { z : IzI < 1) and continuously differentiable on D u y. is d m . (Indiana) Solution. If the problem is changed to h : G + C is continuous and hl~\[o. where the singlevalued branch of d m is chosen by = 4. When 0 5 2 5 1. Y>O lim z=x+yi+x d m = 4i. because z = 0 and z = 1 are branch points of Remark. and h ( z ) is continuous o n C . q p l ] is harmonic. does this implies that h is harmonic on G? (Iowa) Solution.1] c G. Let u* be a conjugate harmonic function of u in D and v* be a conjugate harmonic function of v in D.' r d . 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 .354 5119 Suppose G is a region in C. Y<O Hence h ( z ) = 0 when z = 2.l] holomorphic. A counterexample is h ( z ) = Red. 5120 Let 7 be an arc of the unit circle.
we obtain F 0 on D U y U D*. which implies u = v in D. + + 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. Then for z E y. Without loss of generality. Since F = 0 on y. = lz :dB = v*(z). Let zo be a fixed point on y. where D* = {z : IzI > 1). 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 ~ } . . It is well known that u(w) = targw .g. we may assume that u * ( z o ) = v*(zo) = 0. Then by the reflection principle.and for z E y denote the subarc of y from zo to z by y z . Hence s i U ( z ) = u(i) z1 zS1 = arg() T 2 z+l z1 .2. 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). Hence we obtain two functions f = u iu* and g = v iv* which are analytic in D and continuous on D Uy. F can be analytically extended to an analytic function on D u y U D*. Give the correct determination of any multiplevalued functions appearing in your answer.1 is a harmonic function in H and assume $1 on the negative real axis and 1 on the positive real axis.) Solution. (Courant Insf.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. such that f = g on y. Let F = f .
where the singlevalued branch of arg( is defined by arg( %)lr=o = a. (Minnesota) Solution.1 A  2 z+l arg2.(Z)) .log 2. but may be not singlevalued.dz u. Let * d u = u. This problem can be solved directly from the Poisson formula as follows: 2) 1 = 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). Denote w(z) = Jz * d u . where A is a real number not necessarily zero.{2arg(C .dy and A = h z l = r * d u .argc) . then w(z) is the conjugate harmonic function of u ( z ) .. Remark.. 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). is a harmonic function in D = { z : I1 < 1) with the boundary values $1 on t and 1 on I'2. Define + f ( z ) = (u(2) + i. 2R A .z ) .
5 1. . show that . = a _ . where b.( < 1) and Re(f(z) . Then g ( z ) is an analytic function on (0 < IzI < +m}. . . it follows from Rebo = 0 and 1 n=cc 03 03 n=l n=l 00 n=l that Reg(z) = 0. 2 . . = bn for n = 0.357 then f ( z ) is a singlevalued analytic function on (0 identically zero on { IzI = 1). then . hence f ( z ) = g ( z ) ia. and If(0)l = 1.. Then f ( z )  g(z) = n=O c. Let f ( z ) = l}. . Using Jensen's formula.n ) such that ( u j ( 5 t . and b.g ( z ) ) is identically zero on (1. satisfying 6. n. a .1. 2. and IF(z)I = 1 on (1.2. When IzI = 1. 2 . Define g(z) = n=m C 00 bnzn. Consider P ( z ) = e f ( ' ) .zn is an analytic function in 1.zn be the Laurent expansion of f ( z ) on (0 ncc < IzI < lim = 0 and lim 0 . . .too < Iz1 < 1) and Ref(z) is n=m 5  u. let n(t)be the number of (j= 1 . = 1). for n = 1 . Suppose further that luj I < T for all j = I. . + . we finally obtain + +cc u ( z ) = Re n=m b. by the maximum and minimum modulus principles.. (b) With the hypotheses and notations of (a).1 = l}. = bn and  n+cc lim = 0. we have F ( z ) i~ e i a .g ( ' ) which is analytic and does not assume zero in { IzI < l}. counted with multiplicity. From the above discussion. 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. uj .zn A + log 21r Iz]. . Jensen's formula states that Prove this.
358 (c) For r < R deduce an estimate on n ( r ) in terms of max log If(Reie)l. n(t)we have (b) It is obvious that log 2 = la3 I h:j. which implies that log IF(.)/ is harmonic in 1. By the definition of the function which shows that the identity holds. (a) Let then F ( z ) is holomorphic and has no zero in the disc ( 1 . 0<852x (d) What can be said about the zeros of bounded holomorphic functions in the unit disc? (Hamard) Solution. $. By the mean value theorem of harmonic functions. ~ 1 5 r } . we have & Lzx 0<8<2R log Jf(Re2')ldB= R Denote max log If(Re")J by M ( R ) .( 5 r } . (c) Apply the identity in (b). we obtain n ( r )5 M(R)/log R . .
We know that f(z) can have countably many zeros. we Since f(z) is bounded. Apply Jensen's formula in (a) to F(z) = with 0 < T < 1 such that there is no zero of f on ( 1 . we assume For any n.. and let the other zeros be ordered by 0 < lull 5 la21 I .359 (d) Let f ( z ) be a bounded holomorphic function in { z : IzI < 1).and hence Let T + I. Obviously Ian1 + 1.luj I) is convergent.. we can choose T such that T > lun[. ~ 1 = T } . we obtain n which implies that the series C (1 . j=1 00 . Suppose z = 0 is a zero of f ( z ) of multiplicity m 2 0 with = a.
give an explicit map. Then w = 43 0 4 2 o 41(z) = f ( z ) is a o n e .1)). We construct the map by the following steps: 2 1 = 41(z) = 2.~21< 1 and Imzz > O}.1)}.b o n e holomorphic map from the unit disk {IzI < l} onto the slit disk {IwI < 1}\{[0. {zi : Imzl w 1 < 0) t {z2 : 1. Stony B Solution. 5 l o g r (log 1 = 0) is a conformal map of { z : largzl < l} onto {< : Re( > 0}. 43(22) = 22” : {z2 : 1221 < 1 and Imz2 > 0} + {w : Iw( < l}\{w : Imw = 0.{[0. (Illinois) Solution. < l} ont ( S U N Y .O 5 Rew < 1).SECTION 2 GEOMETRY OF ANALYTIC FUNCTIONS 5201 Find a onetoone holomorphic map from the unit disk slit disk {IwI < 1) . 5202 (a) Find a function f that conformally maps the region { z : largzl < 1) onetoone onto the region {w : 1wI < l}. (a) = f l ( z ) = z: = . z + 1 : { z : IzI < 1) z1 f (21 : Imzl < O}. Hence w < 5 . (b) Is it possible to require that f(1) = 0 and f(2) = $? If yes. Show that the function you have found satisfies the required conditions. and w = f 2 ( ( ) = is a conformal map of {< : Re< > 0) onto { w : II < l}. 1(. explain why not. if No.
5204 a Map the disk { Izl < 1) with slits along the segments [a. which can be represented by $([) = where a.i is a conformal a. It is a conformal map of { z = 0 and y > 0) onto {C = iv : 77 > 0). Hence f can be written as 4 2 o II. s. (2) I f f is an arbitrary conformal map satisfying the condition of (l). ~ ): 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 map of {C = < + iq : 7 > 0) onto {w = u + iw : u2 + v2 < 1 and w < 0). 0 & ( z ) . Let w = 4 2 ( ( ) = (=a <+ dm+(. c .then 4T1 o f o 4F1(C) is a conformal map of the upper half plane onto itself. 2r+1 ._ (b) Suppose W = f ( z ) is an arbitrary conformal map of {z : largzl < 1) onto {W : 11< 1) with f(1) = 0.)( = lY(z)I for every z E { z : (argzl < 1). Then w = 4 2 o &(z) = d n + z 2 . (2) Find all such f . b. d E IR. [1. where d 1 =. b] (0 < < 1.y) : x 2 y2 < 1 and y < 0).361 is a conformal map of { z : largzl < 1) onto {w : IwI < 1) with f(1) = 0 and f(2) = 2q1. and W = F(w)= f o f . we have both [ F ( G ) 5 1 and I F ( w ) l 5 Iw(. By w 6 Schwarz's lemma.bc > 0. Then w = F(6)= foY'(G) is a conformal map 6 of { G : 1 < 1) onto {w : II < 1) with F ( 0 ) = 0. . ( 6 1 which implies that If(.' ( w ) 6 1 w is a conformal map of {w : II < 1) onto {W : 11 < l} with F ( 0 ) = 0. 5203 ( ( 2 . Since    we cannot require that f(2) =  i. ( Toronto) + Solution. (1) Let C = 41(z) = z 2 .i a a is z=e 4 2 + iy : 2 > required conformal map. = .0 < b < 1) conformally on the full disk {IwI < 1) by means of a function . d where 1 = . ad .11.
i..362 w = f(z) with f(0) = 0. 41(b) = b . > 0) {w : ll < 1). +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. > 0). Compute f'(0) and the lengths of the arcs (Haruard) Solution..b]}+~\{[b~. let A= L4 &+&' i.1]U[1. > 0. 4 l ( a ) = a +.) = 1. ( i ) z 1 = 4 1 ( z ) = z + i : { I z I < 1}\{[a. It has the point (iii) z3 = 43(~2) = correspondences fi : C\[O. + ) It has a ...a+~]}. (ii) z2 = 4 2 ( 4 = the point correspondences z i + ( b + ") Zl+(aP+): @\{[6 . B= 24' 6 t &+4 We also know that +i(l)= (iv) w = 44(z3) = ~3: { z 3 : Imz3 . f'(0) corresponding to the slits. It is obvious w . 1 + + #1(1)= 2 and &(1) = 2.a + i]} C\[O. We construct the conformal mapping by the following steps. 42(. It has the point correspondences #q(O) = 00.rs+i that 44(i) = 0 and d k ( i ) = f.
arg Bi 1 $4 = 4arctg.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. The lengths of the two arcs are and containing 2 I1 Ai A+i = arg.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. we know that f maps the unit disk with slits [1.E } and let fir be Let 0 < E < a. > I}.. (Stanford) Solution. From the above discussion. We are going to find the map f by the following steps: .= 4arctg B+z B A+& 5205 denote the arc {eit : E 5 t 5 27r .363 Now we define w = f(z) = 4 4 o 4 3 o 4 2 o & ( z ) . let the complement of yE in the Riemann sphere. b] and [a. f'(0) > 0. describe the part of the unit disc that f maps onto 1(. f(0) = 0.
> l}. is uniquely determined by G and a. then w = F ( C )= f o g'(C) is a conformal map of {C : 1" < r 1 ) onto {w : Iwl < r } + c . arg$$(eT*) = f 5 . with q53(e+a) = 0. orthogonal t o = 1) and connects points e f * and e . and the domain f maps onto 1.364 (shown in Fig. y > 0) and the positive axes. 5 B 5 27r .5.and the symmetric domain of D1 with respect t o I. Since the symmetric domain of 1. is D 2= (1 < 11 l}\n1. and an circular arc I. C .. a ) if G is the region between the hyperbola zy = 1 ( z > 0. (2) Determine r ( G . then @ maps { z : IzJ < l} conformally onto D 1 with @(O) = 0. + Fig. with center C = 0 and radius T I . Hence the conformal map f in the problem is nothing but the inverse of @. After considering the boundary correspondence. a ) of D. (1)Suppose = g ( z ) is another conformal map of G onto a circular disk D. where D1 is a domain bounded by { = eie.T a in {/<I 5 1). @ ( z ) can be extended t o a conformal map by of fie onto {C : ICI < 1).( > 1) is D2.4 Let + ( z ) = $3 o $2 o # q ( z ) .. radius r .which is bounded by circular arcs I. (Indianu ) Solution. which is C } c. we know that 1. (1) Prove that the radius r = r ( G . W(0) > 0.5. the reflection principle. such that g ( a ) = 0 and Ig'(a)l = 1. and if a = 1 i. corresponds t o the arc { z = eit : It(< E } under the map @. and {C = eie : (01 5 i}. such that f ( a ) = 0 and If'(a)l = 1 for some point a E G.( < 1) with respect to arc { z = eit : It( < E } is 1(.4). 5206 Suppose that w = f ( z ) maps a simply conncted region G onetoone and conformally onto a circular disk DT with center w = 0.
i./ = A.2. . z2.z2 21 .32 (Iowa) Solution. z T ( z ) . Apply Schwarz's lemma to F(C) and we have IF'(0)l 5 hence TI 5 T . (a) Let a E a' be a point different from z1. + (22 : Imzz with 42(: + i) = icy. Because the cross ratio is invariant under Mobius transformations. 4 2 o 41(z).if T ( z ) = . Id.' ( w ) and we have T 5 T I .e. (2) We construct a conformal map of G onto a circular disk D. (a) Assume that z1. If'(a)I = 1. Hence r ( G . Show that there exists a constant c such that 3 T ( z ).z2 (b) Use (a) to find an expression for T " ( z ) .i = 1. T is uniquely determined by G and a. 11#:(1+ f). which implies T I = T .($ + i)l with 43(ie%") 0. 2 . we have . Irj5i(iegx)I = 2 = d%lret=* Define f ( z ) = 4 3 .. a ) = * f {w : JwI < &}. l#l z1 = & ( z ) = z2 : G + { q : 0 < Imzl < 2}. = 4 2 ( z 1 ) = eHZ1: ( 2 1 : o < Imzl < 2 ) =fey. T ( z i ) = zi. apply Schwarz's lemma to F . 23 1 . then w = f ( z ) : G f ( a ) = 0 .z 1 z = c.. > 01. For the same reason. n = 1 . with 41(1+ 22 f) = $ + i.365 with F ( 0 ) = 0 and IF'(0)l = = 1. In other words. in the following steps: 2. with 5207 Let T ( z ) = be a Mobius transformation. 3 . z2 E a' are two distinct fixed points for T . .
a . It follows from T"(z)l 21 = 2"Tn(z)+l z+l that T"(z) = (2" (2" + 1). CT"l(z) .3 2 we obtain T ( z ).2 2  T " .z1 = c.z1 Cn .  2 z z1 z zZ (b) Since T " ( z ) = T(T"l(z)).(2" .22 z .Y2 b2+X 1 form a family of confocal conics".z 1 T ( z ). (Humurd) Solution.l ( z ) . .(2n .31 . When a2 < X < b2. then T ( a ) = 5. Choose a = 2.z2 When T ( z )= by solving the equation = z.3 2 . the curves form a family of hyperbolas. (c) Show that the transformation 20 = $(z+:) carries straight lines through the origin and circles centered at the origin into a family of confocal conics. while when X > b2. hence c = 51 : 21 5+1 2+1 s. (b) Prove that such confocal conics intersect orthogonally. (a) Without loss of generality.~  c. T ( a ).. easy to have it is T"(3) 31 T " ( z ).z1 = czT"2(z) .z1 .a .366 which is Denoting T ( a ). we assume a > b > 0. 2 .22 T"2(z) .1) + 1) . the curves form . if a t all. = z .1)z' 5208 (a) Justify the statement that "the curves 22 a2+X +. we obtain that z = f l are two fixed points of T...
Noting that the tangent vector of L1 at (xo. (b) Suppose (xO. we have (u2 Y. Suppose the focuses of the conics are ( ~ c ( X ) .A () When A = J(a2 + A) + [(b2 + A)] = m. and w =u + iv = 1z + 1 ( ) 2 Z = 1 (T 2 1 i + ) cos 8 + (r 2 T .yo) is the intersection point of L 1 : . > b2.yo) is t i = (&.yo) is 7 2 = ($&.+X2 a2+ A2 b2+ Y2 A2 . where A1 # X 2 . Hence the curves 22 a 2 + A Y2 +=1 b2+A form a family of confocal conics. It followfrom and that = O. the tangent vector of L2 at (xo. O ) . *).) + ( b 2 + A l ) ( b 2 + A.1. 4 + Al)(U2 + A. &). (c) Let z = Tez'. When a2 <A <4 2 .+X2  u2+Ai Y2 =1 b2+Xi and L 2 : .) and which implies that the confocal conics intersect orthogonally.T sin 8. if at all. . ) 1 .367 a family of ellipses.
prove that (Indiana) Solution. the focuses of the hyperbolas are ( f l . and if IRef(z)l < 1 for all z E D(0. The image of circles centered at the origin is which are ellipses in wplane. Because cos2 e + sin2 e = I.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. the focuses 1 w = z(z + ). It is easy to know that eYc +1 is a conformal mapping of the domain {C : lReCl < 1) onto the unit disk {w : lwl < 1) with g(0) = 0.$ ( r . 5209 If f : D(0. l).1 carries straight lines through the origin and circles centered at the origin into a family of confocal conics. Hence w = F ( z ) = g o f ( z ) is analytic in . Hence the transformation 0). Because a(r :)2 . + = 1.0).:)' of the ellipses are ( f l .
7r 4 5210 Let R = { z E a'.l) and satisfies F ( 0 ) = 0 and IF(.)[ IF'(0)l 5 1. Find (Indiana) Solution.369 D(0. For any analytic function w = f ( z ) : fl +a' such that If1 < 1 and f ( 0 ) = 0. By Schwarz's lemma. The equality holds if and only if F(C) = e"C. IF(C>l ICI. which implies and the supremum is attained by f ( z ) = e"fo(z). and let F be the family of all analytic < 1 on R and f(0) = 0.we have eT+1 E . . I Choose (0 = w. Because < 1. F(C) is analytic in the unit disk such that IF(C)I < 1 and F ( 0 ) = 0. we have it follows from f(0) = 0 that If'(0)l I . By Schwarz's lemma. where 8 is a real number. 1 < Imz functions f : R +a' such that < If1 l}. is a conformal mapping of fl onto the unit disk with the origin fixed. we consider the composite function w = F(C) = f o f{'(C).
(Minnesota) Solution. which implies I F ( $ ) ]5 It should be noted that when f changes in P. s) . (ii) f ( 0 ) = 1 + i. Then g o f ( z ) is an analytic function + + + E. on D with g o f ( 0 ) = 0 and 1g o f ( z ) I < 1.370 5211 Let f be an analytic function on D = { z . Because w = 23 1c (that is the inverse of C = disc {C : i. 1 < (Indiana) Solution. we have IF(z)I 5 121. It follows from 1 f ( z ) I < 1 1 If(z)I that f ( 0 C R. By Schwarz’s lemma. Let E = : f E P}. F ( i ) can take any value in the 1+c is a 5 f}. 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. Describe E explicitly. we obtain If’(O>l L 4. Prove that If’(0)l 5 4. IzI < 1) such that f ( 0 ) = 1. and suppose that 1 f ( z ) I < 1+ I f ( z ) I whenever J z J 1. w=1 I(s O f)’(O)lL 1. ) Set g(w) = (i f t = 2). Let f E P and define {f(i) < = F ( z ) = f2(z).2i + 2i’ f2(z) Then F is a holomorphic function on the unit disc with F ( 0 ) = 0 and IF(z)I < 1. By Schwarz’s lemma. Let R = C\{W = u + iv : u 2 0 and w = 0).
2 5 3 1 l r 1 4 20 E = {f() : f E P } = {reio : 1 . 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. fE3 sup If(8)l = 4  a. find sup If(8)l. .)I 5 1. f€3 and you should prove your assertion.371 conformal mapping of { C : 5 :} onto {w : I ..r2sin28 8 2 4 .y p s i n 4 4 = 0 by PI(+). Hence If(8)l = IF(4 .$i w ' l the set : f E P} is equal to {f2(i) 10. p 2 ( 4 ) where pl(q5) p z ( 4 ) and Iq5 .$1 5 arcsin $. (Indiana) Solution.&)I 54 a. 3 5 $}. 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. + 4)) a) IF(. 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. Define w = + ( z ) = 2(5+ :). we obtain that {w:Iwz~< Hence }={w=pe*+: 8 3 lr 4 20 Id1<arcsin. Your answer should be an explicit number. 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. W+CC R = {w = u + i w : u 2 + 52 w2 > l}. By Schwarz's lemma. If we denote the two roots of p2 . r 4 .2 5 3 + 4 5 0).1 < arcsin .p~psind+4<0).4.
t E [0. Thus we have a + d = 0 and the inequality a d . L c il. Let z1 # z2 be two arbitrary points in R.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. Then (a2 s. 1 3 is the line segment connecting z1 and 22. which contradicts the condition f # id. Since R is convex. Hence adbc > 0 and a2+bc = d2+bcimpies f = id.b = 0.bc > 0. one in the upperhalf plane and the other in the lowerhalf plane. open subset of a‘ and let f : R a‘ be an analytic function satisfying Ref’(z) > 0. which is equivalent to cz2 + (d .z l ) . we have f Hence > 0 for z E Q. c. it can be written as f ( z ) = where a.. d E IR and ad . + bc # 0. Now we consider the equation f ( z ) = = z .bc > 0 can be written as bc a2 < 0. ( S U N Y . Prove that f is onetoone in 51 (i. (Indiana) Solution. we know that f ( z ) = z has two conjugate roots. Since A = (d . then b = c = 0. z E 51. It follows from f o f = id that b(a + d) = c ( a + d) = 0 and a2 + bc = + 5215 Let fl be a convex. b. which implies . So f has a unique fixed point inside D. Since Ref‘(z) Jif‘(z(t))dt # 0.a ) z . Since f is a conformal map of D onto itself. L : z ( t ) = zl +t(zz . f is injective). + bc)z + b(a + d) + d)z + d2 + bc * f(z)= C(U d2 If a + d # 0. we know that f ( z 1 ) # f ( z 2 ) whenever z1 # 2 2 .e. Prove that f has a unique fixed point inside D. S t o n y Brook) Solution.a ) 2 + 4bc is equal to 4bc + 4a2 < 0.
[ 1. . Let z1.l ) ~ . then Izj 2 1 for j = 1 . and P ( z ) = u ( z .~ + . + P ( z ) z . (a) If all roots of P ( z ) lie in H . IzI < 1 and a1 = 1. not identically zero. .. z 2 . 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 . z z . .. ( z .22 z . . . It follows from the univalence of P ( z ) in { IzI < 1) that P'(z) = nu. Because P ' ( z ) can also be written as nu. by comparing the constant terms. Rezj > 0 ( j = 1. . By assumption. z. It follows that 1 1 P'k) + 1 (logP(z))' = .z. be the zeros of P ( z ) . 2 .2. we have n + > 1.zn un1zn' alz ao.1. 5217 Let P ( z ) be a polynomial on the complex plane. we obtain Iul = n.21 z . let H = { z : Rez > 0).. the roots of P ' ( z ) are all situated outside the open unit disk.z.z 2 ) .z.l). + 2 u ~ z + u l# 0 for all z E 1. (a) Let z1. . Since a1 = 1. .1 be the roots of P ' ( z ) .). . 5 ( S U N Y . ~ n1 j=1 n1 4 lal' 5 1. z..2 2 ) ( z . Stony Brook) Solution. .l z " .+ . .373 5216 Show that if the polynomial P ( z ) = a. (b) For any nonvanishing polynomial P ( z ) ... then [nu. ( z .. . j=1 zj = al.( < 1).n .) Solution.zl)(z . . . In other words. .^^^ + (n. is onetoone in the unit disk + + + n1 (l)nlna.z1)(z . show that the same is true for the roots of dP/dz. .n). (Courant Inst.
Re& < C j=1 n & < 0. convergent inC\{O}. 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. (a) Let f ( z ) = too n=cu C bnzn. with residue b at z = 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 ) . Hence Re < arg(zzj) < $. we obtain that the zeros of P’(z) lie on the convex hull of the zeros of P ( z ) .z j ) ) < 0. then Hence . which shows can not be zero on { z : Rez 5 0).374 When z E { z : Rez 5 0}. . . . we have Re{eae(z . z 2 . z. (a) Show that there exists on { z : IzI = 1) with < (b) Characterize those functions with (Minnesota) Solution. then 0.) or equivalently. Hence which shows that the zeros of P ’ ( z ) do not lie on the left side of I . 5218 Let f ( z ) be a Laurent series centered a t 0. . (b) Let z1. When z is on the left side of I. be the zeros of P ( z ) . Denote the intersectional angle from the positive direction of the imaginary axis to I by 8. p(.
f’(z) # 0.375 If If() .1 1 holds for all C with I(I = 1. where D = { z : IzI < 1). and hence f ( 0 .l)ei+(e). where ( = eie and 4(8) is a continuous realvalued function. and f maps d D into a l l . f maps D into D . Let f(() . .<l = ( b . ( a ) Show that Vz E aD. 5219 Assume f is analytic in a neighborhood of 0. (b) Show that &[argf(e”)] > 0 for 8 in R. Apply the discreteness of zeros for analytic functions to f ( z ) . It follows fiom that which implies that (p(8)= 8. b1 p.P I = Ib .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) . p. then which is a contradiction.cl = 7 holds on {C : 1[1 = 1).1)holds for all C with I(I = 1.<l1 < J b .11.
z o ) + b z ( z . ( IndianaPurdue) Solution. If g ( e i e ) < 0 . But from (a) it is impossible. it follows from = aR that %(e i6 ) < 0.zoI = r ) . which implies that f ( z ) has two zeros (counted by multiplicity) in D. that Ararg(f(z) . Then where n 2 2. we know that % ( e i 6 ) = 0. Show that f(z) # 0 whenever 0 < IzI < 1.376 (c) Assume that f(0) = f'(0) = 0 and f l a ~ a twotoone map from a D is onto dD.z o ) " g ( z ) ( n >_ 2).. If g ( e i e )1 0 . Let I be an arc in ' defined by I = { z E % : ) z .~ 0 ) 2 + . and w = f(z) = Rei@. (b) Let z = rei6. It follows from that Ararg(z .. &Alzl=largf(z) = is 2. . ' and denote by Ar#(Z) the change of # ( z ) when z goes along the arc I in ' the counterclockwise sense. Since f(0) = f'(0) = 0. with bo # 0. where zo E d o . Hence we obtain 2 2 7s Rg (c) Because f l a ~ a twetoone map from 8 0 onto a D . 2 3 which implies that f(z) assumes values outside the disk when z E I. g(z) = b o + b l ( z .A variation of the CauchyRiemann equations for analytic function w = f(z) is Since f maps a D into a l l . z = 0 is a zero of f of multiplicity m = 2. Hence f(z) has no zero in (0 < I1 < I}. t . It is demanded that r is sufficiently small such when z E I?.> K . It is ' a contradiction to the fact that f maps D into D. This is also impossible. Since R = 1 when r = 1.W O ) = nArarg(z .wo = ( z .zo) > and Ig(z) . = = = 0. . which implies that g ( e i 6 ) = f'(ei0) = 0. .bol < f ( z ) . then = at point ei6. Let f(z0) = wo E all. (a) Assume f'(z0) = 0. Hence f'(z) # 0 for all 2 € aD. g ( e i 6 ) < 0 implies that R > 1 when r < 1.ZO) 3T n+ Arargg(z) > .
. Function e e z is analytic in { z : 0 around z = 0 is: < IzI < +co}. . + By the residue theorem. and its Laurent expansion The coefficient of the term $ in the above development ~ is 1 1 + 1 + 2! + .l)! 5302 Evaluate where y is the positively oriented circle { Jz1= 1).= e.377 SECTION 3 COMPLEX INTEGRATION 5301 Evaluate the integral (Indiana) Solution.. (Indiana) . ( n. we obtain 1 +.
Hence we obtain a = 2.)Z+.. and with z = 0 as a pole.) (: coszdz (Indiana) Solution. i.. the coefficient of the term in the above development is zero... its and F(z) = (.. 1 1 1 2 t4!z4. + 5) +. = 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 . is $.378 Solution. . It is obvious that is analytic in { z : 0 < IzI 5 l}. 1 U a ( . ... The Laurent expansion of around z = 0 can be obtained as follows: &  1 sin3 z  1 (z  k. Function F ( z ) = ( $ 5 ) c o s z is analytic in { z : 0 < IzI Laurent expansion around z = 0 is: + < +a}.e.+. sin z =~ i ..2aiRes 5303 For what value of a is the function f ( z )= singlevalued? LZ(. I ~ Hence the coefficient of the term residue theorem.23 + $25 1  .)cosz= 2+('2)P+(. we have p in the above development ( 3 1O ) .
IzI < 2). dt It follows from let + ztl 2 et . we know that h ( z ) is analytic in { z : IzI < e } .et dt. + What is the largest possible P so that h ( z ) is analytic for IzI < P? ( IndianaPurdue) Solution. It is easy to see that when t + 1. et . which implies that the integral is divergent. Consider the integral et + z t dt. 5305 Let f ( z ) be analytic in S = { z €67.379 5304 Define h ( z ) = L m ( l ztet)let cos(t2)dt. Hence the largest possible P is equal t o e. cos(t2) et . Show that .et  A (t . For any r < e . By Weierstrass theorem.rt and the convergence of the integral that is uniformly convergent in any compact subset of { z : IzI < e}. Hence P can not be larger than e . When z = e. let IzI 5 T .1)2’ where A = $ cos 1.
Let f ( z ) = u ( z ) iw(z). Then f ( z ) is analytic in { z : IzI < 2}.t E [0. dz dz z where y(t) = eZnit.380 (Iowa) Solution. 1 . Show that < 2}. 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. Stony Brook) Solution. 1 ( S U N Y . and we have + .
381 It follows from u ( z )= and f(z) +m 2 f f (> . z E D ( 0 . and if Ref(z) /2T 0 enie[Ref (eie)]dO.l) = { z . l). (a) Prove that d”f(0) = dzn (b) If f(0) = 1. prove that > 0 for all points (Indianu ) . w(z) = that  22 5307 Let f be an analytic fqnction on an open set containing D(0.IzI 5 11.
by the meanvalue formula of harmonic functions. (a) Assume that 00 k=O we have By Cauchy Integral Formula. Hence (b) Because Ref(z) is harmonic on D(0. l).382 Solution. 2 0. kI'" Noting that Ref(eis) Ref(e")de = Ref(0) = 1. . we have = 2(n!).
5309 Let f = u + iw be an entire function. show that the coefficients in the expansion n =O satisfy lunl < e for n 2 1. (Stanford) Solution.. (0 < T < 1). n=l where nu.383 5308 I f f is analytic in the unit disk and its derivative satisfies IfWI I (1. = 1 f)odz. . : (1 + )I n1 1 < e. we choose T = 1 . It follows from f ( z >= 00 C anzn n=O that f’(2) = c 00 nunznl. It is obvious that lull = I’0l I 1 < e .14)1. f() For n > 1. where e is the base of natural logarithms.
(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 .f 2 ( z ) = .'(. = 0. (b) Let For any integer n > h. ( Stanford) Solution. (Indiana) Solution. Then it follows from Schwarz's theorem that when IzI < R.(ua(z)v2(z))Ziu(L)v(r) Then F ( z ) is an entire function with IF(z>I = e (. Letting R + +m. Show that f is a polynomial of degree at most n. Let R be an arbitrary positive number. h. B .). where n is a positive integer. then f must be a constant. F ( z ) must be a constant. which implies that f ( z ) is a polynomial of degree bounded by h. which implies that f ( z ) is a constant.384 (a) Show that if u 2 ( z ) 2 wz(z) for all z EC. By Liouville's theorem. 5 A+BR~ R" . then f ( z ) is a polynomial of degree bounded by h.)) 5 1. we obtain that a."(. (a) Let F ( z ) = e .
B such that If(z>I 5 AlzY holds for all z EC. Let +B where Hence when k > n. cos z d x d y where D is the disk given by { z = x + i y E a' : x 2 + y 2 < 1).385 Especially when IzI = +. 5311 Compute the double integral 1J . First we have the following complex forms of Green's formula: . (Iowa) Solution. 3R" * 2n + IIm{f(O))l = 3R" If(. which shows that f(z) is a polynomial of degree at most n.)l L 1 + IIm{f(O))lr which implies that there exist constants A .
(a) B Y n=l we have . (a) Express A in terms of the coefficients an. 1 sin zd( ) Z 5312 n=O be analytic in D = (1. (b) Prove that for z E D.sin z dz = R. (Indiana) Solution. < 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? = .
The answer to the problem is Yes. we have 5313 Let f be analytic in (0 < IzI < 1) and in L2 with respect to planar Lebesque measure.387 Noting that we obtain that . Is 0 a removable singularity? Proof or counterexample. Let the Laurent expansion of f in {z : 0 < IzI < l} be 00 From . ( Stanford) Solution. 1 00 00 (b) By Cauchy’s inequality.
Hence n=O which shows that z = 0 is a removable singularity of f . let E + +0. Otherwise.21r 1' la. and then . a = rei+.388 we have Let E < 1 be a small positive number. 5314 Evaluate the integral (Indiana) Solution. which leads to a contradiction. must be zero when n 5 1. . 12r2n+1dr< Then a. Let z = peie. while the right side of the inequality is finite. the left side of the above inequality will tend to infinity.
Denote It is obvious that I ( u ) is an analytic function in { u : la1 lr > 1). Then we have I(a) = dx 2dx dx . (Columbia) Solution. When r > p.When r < p. 5315 Evaluate by the method of residues.
Since z1 z2 = 1.z2 7T 27r d(Zl t 2 2%& * ~ ~ ) ~ 42122 .2(2u 1).390 Let z = eisl then dx cosx and = = z dz iz ' + 21 2 ' = Denote the two roots of z 2 .  2dm' 0.6)de when ( z ( < 1. we may assume that ll > 1. 1 = 0 by z1 and z2.=l (z  idz  z1)(z . It should be noted that branch of is also analytic in { u : la1 > l}. 1221 < 1.27T 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. By the residue theorem we have zl iz.=l + + 22  idz 2(2u + l ) z + 1' =  iz. (b) Integrate the Taylor expansion n=O term by term to find the coefficients in the Taylor expansion .
m in 1. Because IC1 . z k .l ) " C I f . C21 = 1. k=O (1.eiO sin8 = 22  C2 1 2ic ' and where of d ci = $(d=l).1 ) ' s i n k 8 . and f(z)=$' (g(l)ksink8. Z 2 " n=O .391 (c) Verify directly that (a) and (b) agree when IzI < 1.27r]. Since the series converges uniformly for all 8 E [0. and the singlevalued branch by 4 I r = ~ = 1.zk k=O 00 where Uk 2r = (1)' sink ode.i0 . (a) Let c = ele. the integration with respect to 8 can be taken term by term.) Solution.( (2 < 1) is defined =i(dnl). < 1). Then . we develop the f(z) = 2K ~ 00  dC7 27r(l z"i = 2 7 r . 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. (b) It follows from I sin81 5 1 and 00 IzI < 1 that g(z.y . (Courant Inst.e) = C ( .
by Cauchy integral theorem.5 As f ( z ) is analytic inside I. show that exists and is independent of a.5. rl U I'2 U Fig.5.5. ' .392 we know that the results in (a) and (b) agree when J z J< 1. Irvine) Solution. (UC. ' Define f ( z ) = e" and choose the contour of integration I = r3 u r4 as shown in Fig. 5317 If a is real. First we have It follows from the existence of hllRex2dz R that exists.
6 Let f ( z ) = &. we have f(z)dz = 2~iRes(f. . By the residue theorem.5. 0 R Fig . dY it follows from the facts that eR2 + 0 ( R + CQ) and 5318 Let n 2 2 be an integer. Letting R + 00. Compute (Iowa) Solution.393 ieR2 J.e:*). and select the integral contour r as shown in Fig.6.' ey2+2Ryi = 0. f ( z ) has one simple pole z = eEi inside I'.5.
z 2 .394 and 5319 Evaluate JD cos (x') dx with full justification. B caus is analytic on I' and inside I by Cauchy integral theorem. we ' .7.7 Define f(z)= e . (Minnesota) Solution.5. Fig. 3 j=1 . and choose the contour of integration f(z) = have eIZ r= 3 rj j=1 as shown in Fig.5.
2 7F By Jordan's lemma.sin z2)dz = 2 ' which implies 5320 Evaluate (Iowa) . we have For the integral of f ( z ) on r3. then where 72 = { w :lwl= R2. L m ( c o s z2 f(z)dz = I" / 0 ex2dz + 2 f fi when R +00.395 For the integral of f ( z ) on r2.O 5 argw 5 }. we make a change of variable by w = z2. Hence we obtain by letting R co that + sin x2)dz + i 00 & (cos z2 . we have It is well known that l.
by Cauchy integral theorem.5.8.8 Define and select the integral contour as shown in Fig.396 Solution. Fig. where It is easy to see that lim Rr 00 I' iReief (Reie)d6= 0 and .5. Because f(z) is analytic inside r.
Stony Brook) Solution. we know that Res(f. 2 7r 5321 Let f ( z ) be holomorphic in the unit disk IzI 5 1. The branch of log is chosen to be real for positive z.5. Letting E + 0 and R + ca.9 ' Let the contour of integration I be shown as in 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 = 1.5. . J. 0) = 22.f ( z )logzdz = 0.397 Since the Laurent expansion of f about z = 0 is 00 where a1 = 2i. and the singlevalued branch of logz be chosen by argzl.we obtain dx = . Fig. Since f(z)logz is holomorphic inside the contour I?.=I = T . ( S U N Y .9. by Cauchy integral theorem.
we obtain where the integration contour IzI = 1 has starting point and end point z = 1. 5322 where a and b are complex constants. Take ei@ = z . . not both equal to zero. and the value of logz at the starting point z = 1 is defined as 0. (Hamud) Solution. then d4 = and + g. and then the multivalued analytic function log(a bz) has singlevalued branch on { z : IzI 5 1). I 2r log la + be'@]& = Re = R e ( 2 ~ l o g a )= 2rlog lal. First we assume la1 > lbl.398 where It is easy t o see that Letting E + 0.
we have In the case la1 = lbl. Choose a contour of integration = r. let b = ae".399 When la1 < lbl. as shown in Fig. Then Fig. Uy. f(z)dz = 0. Because + s . by Cauchy integral theorem. Since f ( z ) is analytic on I' and inside r.10 In order t o evaluate the integral we define f.5.5. we have JR I*log 11+ ei@1d4 = Re JR IR log(1 + e*+)d4 .10. () = log( 1 z +z) 7 where the singlevalued branch is defined by log(1 z ) lz=o= 0.
11. The singlevalued branch of logz is chosen by argzI. By the residue theorem.'" Evaluate log la + beiGldq5 = 2 7 max(l0g la[.5. l).400 Hence we obtain J. ~ 5323 (Iowa) Solution.1 = ?r. we have J..11 Let and select the integral path I? as shown in Fig.log lbl}. f(z)dz = 2?riRes(f. .5. Fig.
AS E + 0 and R + 00. ice"f (&eie)d6. we consider the Laurent expansion of f about z = 1: = where a1 n=3 c a n ( z + lyl = 1 .ai. we obtain . Comparing the imaginary parts on the two sides of the above identity.l). it turns out that 4ai log x + 4a2 dx = 2ai + 2 2 .'" iReief(Reie)d6 = 0 and In order to find Res(f. Rrm lim J.401 where 2X iRegef (Reie)d6 + (logx + 2ai)' dx rO + It is obvious that 1. Hence 2aiRes( f . 1) = 27ri + 2 2 .
and the contour I of integration is shown in Fig.5.12.) Solution.5.402 5324 Evaluate the following integrals: +im at (a) Sim ( t 2 .12 Define 1 The singlevalued branch for log(z + 1) is chosen by log(z + l ) l z = o = 0.4 ) l o g ( r + l ) (the integration is over the imaginary axis). we have f ( z ) d z = 2 ~ i R e sf(. (a) Fig. (b) (Courant Inst. 2). As f ( z ) is analytic on and ' inside J? except a simple pole at z = 2. where and Because . &dz for CY in the range 1 < QI < 2.
we obtain .efa) = lim ( z . I F where and Res(f.4)log(z + 1) = (14 Ti ). +ice dz ( z 2 ..1  3e f ( 2 . a simple pole at z = e:'. . e f i 3e +a . and the contour 'I of integration is shown in Fig.5.2 log3 Fig.13 Define The singlevalued branch for zm is chosen by argzIr=z./.403 and by letting E + 0 and R + 00.efa) f ( z ) z+e:' .. As f ( z ) is analytic on and inside r except .m ) i ' . we have f ( z ) d z = 2 ~ i R e s ( f e:')).5.13.O = 0.
5.14. 5. What happens if a = l? Solution. we have f ( z ) d z = 2aiRes(f(z). Fig. and select the integral path I as shown in Fig. f(Reie)iReied8 = 0 2a  when a < 2 and lilil ' + f(&eie)iseied8= 0 + co. i ) . By the residue theorem. by letting E 0 and R obtain dx = 5325 Show that J.we 7r when a > 1. a # 1.14 where . Let ' where (argza)z=z>O = 0.a ) ( 1 + x 2 ) 2 d x = 4 cos( y )' < a < 3.404 Because lirn R U C 2z . for 1 00 Xu 7r( 1 .
a(1 . I. When a # 1. and from a > 1 that Letting E + 0 and R $ 00.a ) x y d X= 4 cos( y) ' X T(1a)  1 5326 (a) Prove that converges if 0 <a < 1 .405 + J.i) = lim [ ( ~ 2'2 2 ta i)2 + 1'9  e"i 2 ." iEeie f (&eie)d8 and Res(f(z). we obtain XQ a(1a) 2 e 1" 2 . 2 " al a) ( .+ (1 when a = 1. It follows from a < 3 that lim RC .O iReief (Re")dB = 0 .
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
+
z ) = 2eioz2. By the maximum modulus principle. limit of the Hence can not be larger than 2. = 0. a4 = a6 = .z ) = 2eiaz2 = eio. (a) Let F ( z ) = f(z) + f(z). If equality occurs for some nonzero z in D . then F(O) = 0. we have + I I f(z) + f ( . which implies f ( z ) = e i a z 2 . Because If(z)I < 1 for z E D .. we have Since a2 = e i a . If(. the other coefficients must be zero. where a. is analytic in D . is a real constant.. and when z tends to dD.419 (b) Prove that (Indiana) Solution.) f(z)I 5 21zI2 holds for all z E D . (b) . Let M n=l it follows from that a2 f ( z >+ f ( .
Since G ( z )has exactly one zero inside I.< argz 5 }. where > 2A.Z.z and G ( z )= z. IF(z) . Since G ( z )has only one simple zero in { z : IzI < 1))we conclude that f ( z ) . Hence f(z) takes value X exactly once in the right half plane. 2 2 Define F ( z ) = X . Prove or disprove: f ( z ) takes the value X exactly once in the right halfplane. (Iowa) Solution. By RouchC’s theorem) F ( z ) and G ( z ) have the same number of zeros in { z : IzI < 1). prove that f ( z ) has a fixed (Rutgers) Solution. ?r Take a closed ?r I’ = { z = t + i y : t = 0.G(z)l = le”l 5 1 < IG(z)I.’ . When IzI = 1. 5411 Let f ( z ) = z + e . If the answer is yes. which implies that f ( z ) has a fixed point in { z : IzI < 1). IF(z) .e” and G ( z )= X When z E I?. Let F ( z ) = f(z) . < 1 on { z : IzI 5 l}. F ( z ) has exactly one zero in the right halfplane. it follows from RouchC’s theorem that ’ F ( z ) has exactly one zero inside I?. .z has one zero in { z : IzI < l}. Let R be a sufficiently large real number such that R curve I’ on the right halfplane. . X > 1. Because R can be arbitrarily large.z .G ( z ) /= lf(z)I < 1 = IG(z)l. is the point necessarily real? Justify. R 5 y 5 R} u { z : Izl = R.420 5410 I f f is analytic and lf(z)I point. .
are the zeros of f ( z ) in {z : II . there must exist 20. +.42 1 Take z = x 2 0.Zj). Assume z1. (Iowa) Solution. 5412 Suppose f is analytic in a region which contains the closed unit disc {z : ( z ( 5 1).x . Suppose that and Find the location of the zeros o f f in the open unit disc {z : II z < 1). the point z in the right halfplane such that f ( z ) = X is necessarily real. Since F ( x ) is continuous and F(O) > 0. Let C denote z the unit circle traversed in the counterclockwise sense. such that F ( z 0 ) = 0. Assume f is nonzero on the unit circle {z : II = 1). In other words. x++w lim F ( z ) = m. 0 < xo < $00. n . where g(z) is analytic and has no zero in {z : Il z 5 1). multiple zeros f(z) = !dz) j=1 nc. z being repeated. .e". We have F ( x ) = X .z2. which is a realvalued function of real variable x. Then < l). We have .z.
.e. f ( z ) has two zeros in the unit disk.O 5 argz 5 } z 2 R}.y = 0) U { z : II = R. (a) Let R be sufficiently large such that when IzI = R.z1)(z .O 5 z 5 R. z1. 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 $? (IndianaPurdue) Solution. and which show that unit disc.O<yL . i. U{z=z+iy:z=O. lr Choose a closed curve ' = I { z = z +iy. Then for f(z) = ( z .422 It follows from (1) that n = 2.z z ) g ( z ) . Set f ( z ) = z4 +z +5 and g(z) = z4 + 5.2 = ki.
arg (1 + F) . By RouchC's theorem. Ap.argf(z) = Ar. we know that ' . We also know that Ar. 2 A It is easy t o see that Imf(z) > 0 when f(0) = 5.) . while .)l holds when z E I?.< argz 4  A 5 }. where Ar.argz4 = A. Set f(z)=z4+z+5 and 9is approximately and r3 = { z : IzI = R. i) f (a$*) E {w : ?r E < argw < where E > 0 is very small. the numbers of the zeros of f and g inside I' are equal. . and f ( ~ E {w : o < argw < E } . (b) Let R be sufficiently large such that when la1 = R. f has also one zero inside I Noting that R can be arbitrarily large.mgf(z) denotes the change of argf(z) when z goes continuously from Re<' to R along I3 It is obvious that i '. z4 z + +5 = 0 has one root in the first quadrant.g(z)I < Is(z. Since g has only one zero inside I?. zero.argz4 + Ar.423 It is obvious that If(.
z and z = z + i y . 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 < }. choose a positive number t >> logn and a closed contour I' = rl U I'2 U r3 U r4 in the counterclockwise sense. (Indiana) Solution. where . 2 x 5414 Prove that the equation sin z = z has infinitely many solutions in C. Let I' = I'l U r2 U r is taken once counterclockwise.424 is very small. Let f(z) = sinz . then f ( z ) can be written as For any fixed natural number n.
f(rz) {w = u + i v = with the direction upwards.2 7 ~ v ) 0) ~= with the direction from the right to the left. with the direction downwards. Assume that f ( z ) = sinz . * . we conclude that sinz = z has infinitely many solutions in c. o 5w 5 1 .et ) .z has only finite zeros in (X.425 Then we consider the image of r under w = f (2): f(rl) {w = u + i~ = : 2(n +1 ) 5~u 5 .z n . This problem can also be proved by Hadamard’s theorem. and denote all the zeros by z1. multiple zeros being repeated. Since n is arbitrarily chosen. . z z .t ) lies in the annulus starting from w = 2(n + 1)n + i in the counterclockwise sense.z has one zero inside the contour I?. By Hadamard’s theorem. Hence the winding number of f(r)around w = 0 is 1. f ( z ) = sinz . : t = 2(n+ i f(I’3) ~ ) T .(et 2 .zg) and g( z) is a polynomial. By the argument principle. Remark. . f ( z ) can be written as f(z) = e g ( z ) k ) ) P where P(Z> n = k=l .
Hence we have sin z . and comparing the increasing order on both sides. Show that f has at least two zeros.argf(z) 2 1 and &Ar. Because f is nonconstant analytic in D and If1 = 5 when z E dD.we know that both f(I'1) and f(rz) must be {w : lwl = 5) in the counterclockwise sense. Let g(C) be a conformal map of {(=<+irl:$+$<l. (a) Let D = { z : 1 < I1 < 2) and d D = rlU r2. In other words. f has no zero in D.where rl = { z : IzI = 2) t is in the counterclockwise sense. By letting y + +oo and y + oo respectively. and I'2 = { z : IzI = 1) is in the clockwise sense. which shows by the argument principle that f has at least two zeros in D. In the latter case. This contradiction implies that sin z = z has infinite many solutions in C.426 It is obvious that f ( z ) is an entire function of order X = 1. The following is a counterexample. Let z = z+iy and z be fixed. the statement in (a) is not true. . (b) I f f is meromorphic in the annulus. where T03 which implies that g ( z ) must be a polynomial of degree 1. Hence &Ap. (b) If f is meromorphic in D. It might occur that f(rl) and f(r2)are two subarcs of {w : IwI = 5).z = e a r + b p ( z ) .argf(z) 2 1. we obtain that Imu < 0 in the former case and that Imu > 0 in the latter case. or both f(r1) and f ( r 2 ) are {w : IwI = 5) in the clockwise sense. is the statement in part (a) still true? (Stanfod) Solution. 5415 (a) Let f be a nonconstant analytic function in the annulus (1 < IzI < 2) and suppose that If I = 5 on the boundary.
But Let n be a positive integer. g'(0) > 0.y = 01. > Take a closed contour I? = I'l U I'2 in the counterclockwise sense. 0 where each aj €67. 5416 I f I = 5 when t E LID. and suppose that Prove that P has exactly TI roots (counted with multiplicity) in the open upper half plane { z E 67 : Imz > 0). where rl = { Z = TP: o 5 8 L ) .427 onto {w : IwI > 5) with the normalization g(0) = M. Let T > 0 be sufficiently large such that when IzI = T .. Then is a nonconstant meromorphic function in D with f has no zero in D.+azn_1z2"11. (Indiana) Solution. and let P be a polynomial of exact degree 2n: P ( Z )= a0 + alz + a2z2 + + aZntzn. and aZn # 0. Suppose that there is no real number x such that P ( z ) = 0. e .. ( a 2 n ~ 2 n ( (ao+alz+. and r2= { z = X + i y : T 5 T. Then the number of zeros of P ( z ) inside I is equal to ' It is already known that .
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. argP(z) 1 Note that and 1 Ap.e3+ * .+  z 2!z2 1 1 + .arg 2a (u2nz2n) =n we obtain that P has exactly n roots (counted with multiplicity) in the open upper half plane.) Solution. * . From .+gcn. (a) Let 1 F(C) = f() = 1 + c c 1 1 1 + 3c2+ . 5417 Consider the function 1 1 1 f(z) = 1 + .428 We also have d log P(z) = G A r . . +.
Show that this solution is given by the formula W (b) Show that. > T}. there is no zero of f ( z ) in { IzI > T } . if the integer n is sufficiently large. we obtain {/(I < :} for fixed T and large n. In other words. which is just the number of zeros of f ( z ) in 1(. we conclude that for any fixed T > 0.ec I < m I 1ecI for ICI = : which implies the numbers of zeros for F(C) and e in . all the n zeros of f ( z ) are in { IzI 5 T } . and that there is a unique. 5418 (a) Suppose that f ( z ) is analytic in the closed disk IzI 5 R. lF(C) . (c) From the above discussion. when n is sufficiently large. ICl=? then m > 0.429 we know that the negative of 11 represents the number of zeros of F(C) in (1 < :}. < R}. the equation has exactly one solution with ( z ( < 2. C are equal. Let min leCl = m. When n is sufficiently large. Since ec has no zero in a'. (b) When n + 00. . simple solution z1 of the equation f(z) = w in 1(. F(C) converges to ec uniformly in any compact subset of@.
(a) Let f ( z ) .zi)&(z). show that lim (zl.z ~ ) l o g & ( ~ ) ]= .) 1 1 Solution.E L Ig(z>I. Then f 1 f’(z) = [log(f(z) . When n is sufficiently large. ncc 2 (Courant Inst.w)]’ [log(z . we have fn (1) < 0 and fn > 0. $). where Q ( z ) is analytic and has no zero in { IzI < R}. (c) f. E Hence fn(z) and g ( z ) have the same number of zeros in { IzI < 2 . Hence we have z?’ E (1.1 and rE = For fixed large n.& ’ ( z ) = ’ w zzi Q(z)* + + Hence (b) Let fn(z) = z .€1.w = ( z . Ifn(z) . we choose E = 2 .430 (c) If z1 is the solution in (b). and the number is 1. < 2). Since E can be arbitrarily small.1).(z) is a continuous realvalued function for 1 5 2 5 $.g ( z ) l = 11 5 z n = (1  5)” < 1 . the equation z = 1 + (5)” has exactly one solution (denoted by z p ’ ) in 1(.E } .= . It follows from (z) that . z > 0 sufficiently small such that when E re.
Now we claim that the case that = f(. which is a contradiction. If. f * 7 or f * y2 1 will be homologous to zero (mod R). and the directions of 7 and 7 2 are both in the counter1 clockwise sense. f * yl and 1 f *y2 are also not homologous to zero (mod R).l)\{. As E tends to zero. Since 7 . 2 2 Find all analytic functions f : S2 + R with the following property: if y is any cycle in R which is not homologous to zero (mod R). Hence we obtain f(*) = f2.f of order n and z = f is a zero of f(z) . }. f(rl) and 1 f(y2) will tend to either 20 = 5 or 2u = . for example. (Indiana) Solution. then f * y is not homologous to zero (mod R).7 2 are not homologous to zero (mod R). because otherwise. Since f is analytic in R and bounded by lf(z)I < 1.$ of order m.43 1 which implies 5419 Let 1 1 R = D(O. 1 1 1 f(1 = f() = 5. Thus we obtain either f() 1 1 =2 2' f() =  1 2 1 2 . while my1 . Let where E > 0 is small.nr2) 1 2 1 f(i) is homologous to zero (mod Q). then f * (my1 . the points z = =kf must be the removable singularities o f f .ny2 is not homologous to zero (mod R). 2 2 we assume that z = f is a zero of f(z) . .) will not happen.
i ) = 1 that G(z) 1.zz which is also analytic in D(0. It follows from F ( that F ( z ) 1.l) and satisfies IG(z)I 5 1.t .)   z . 1. which implies that f ( z ) = z.l) and satisfies IF(z)I 5 1. Thus we conclude that the functions which satisfy the requirements of the problem are f(z) = z and f(z) = 2. 2 2 1 1 we consider the function F(z)= fk). = .t z 1 which is analytic in D(0.i f ( z ) 2 1. 1 2 f() = . It follows from G ( . which implies that f(z) = z. = i)= 1 we consider the function f(z)+i G(z) = 1 + if(.432 or f() 1 2 = .51 1 1.
which implies It follows that . .z . In other words.zo 00 is convergent a t z = zo.zo z Then F ( z ) is analytic on { z : 1zJ5 T } . on I1 = T . n+w (Indiana) Solution. and define F(z)= f ( z ) A . which is a simple pole.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 . Hence the power series F(z) = n=O A Cantn. the Taylor expansion of F ( z ) a t z = 0 has a radius of convergence larger than T . Prove that lim an/an+l = zo.
( < 1). which shows that .C $ is convergent and C n>l n>l E C with la1 = 1. 2 ~ )then .a ) for such a. c. (Minnesota) Solution. we have 1 1Z . (2) Show that this series converges to log(1 .434 and and we obtain lim = zo.=c an n>l n>l cos nt + i sin nt n For t E ( 0 . (1) Let a = eit. (2) Let f(z) =  n>l c n> 1  zn n 1. by Dirichlet’s criterion we know that converge. Differentiating term by term. n+co % + I an 5502 (1) Show that the series n> 1 is convergent for 1 # a E C with la/ = 1. t E ( 0 . 2 ~ we have ) and Because both C for 1 # a n>l tends to zero monotonically.
. .a). i. I n= 1 Show that the series converges to a holomorphic function on the open unit disk centered at origin.z ) . then for fixed 2 E ( 0 .e. ( Columbia) Solution. 5503 Consider a power series O01 C nz n . 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.e i t ) = log(1 . z = Teat where 0 < T < 1. we prove the following proposition: If the radius of convergence of n =O is equal to 1 and an 2 0 for all n. we obtain f(z) = log(1. Suppose the series : is + . z = 1 is a regular point of f. then a = 1 is a singular point of f(z). 0 < t < 2a. First of all. It follows from Abel's limit theorem that . Let Q = eit.435 Integrating both sides on the above identity. for (a1 < 1.= n> 1 r+l eint limCr+1n>l (Teit)n 1 n lim log(1 T e i t ) = log(1. Assume the proposition is false. Prove that the boundary of the disk is the natural boundary of the function.C n an n> = = C.
&I n.. + + n=O W k=O . z = 1 is a singular point of F ( z ) . . For any natural numbers p and q .k + k=On=k is convergent at z = 1 6 . we assert that when z = 1 6 .Ic + 1 )an(z . . By the above proposition. n=O which contradicts the statement that the radius of convergence of n=O is equal to 1.1) . and hence the order of summation can be changed. a ( . c .436 Thus 00 w w x b k ( z z)k = k=O n(n .z ) k z n . Noting that when z = 1 6 the right side in the above identity is a convergent double series with positive terms.". 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.
< 1) with f(0) = 0 and If(z)I U . prove that either g ( z ) = 0 or g ( z ) = z . By the above discussion. In other words. 5504 z E Suppose f is analytic in U = 1(. which implies that f ( z ) = z and g ( z ) = t. Remark.and if If(z)I = 1.2. Since the set { e .1 for some z # 0.(fI for all z E U .i is a singular aq point of F ( z ) . then f(z) = einz where a is a real number. By Schwarz's lemma. f n ( z ) = einaz. we conclude that the unit circle {IzI = 1) is the natural boundary of F ( z ) ..p r i : p . q = 1. 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. Since f n ( z ) is convergent. z = e?4. it follows from f(0) = 0 and If(z)l < 1 that 5 IzI for all z E U . (IndianaPurdue) Solution.. we have [).. n n=p it is also a singular point of F ( z e 9 " ' ) . In the case when f ( z ) = eiaz. In other cases.437 Since z = 1 is a singular point of " 1 I zn. we obtain a = 0. Let 0 < r < 1.} is dense on {It1= l}. Then .. 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 know that for sufficiently large n. . T}. there exists T > 0. fn must have a zero in { z : Iz . such that f(z) # 0 when z E {Z : 0 < IZ 5 T } C D. we have . (b) If each fn is onetoone on D . 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. Since the zeros of a nonzero analytic function are isolated.z0I < T } . As {fn} converges to f(z) uniformly on compact subsets of D . (a) First of all. we know from Weierstrass’ theorem that f is analytic on D .438 For all z E { IzI 5 T } . which is a contradiction to the assumption that fn(z) # 0 for all z E D. Suppose f is pot identically zero. (UC. (a) Prove that if fn(z) # 0 for all n 2 1 and z E D .z01 = 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 . Ifn(z>..z0I < T } . Since zo is a zero of f . show that f is either constant or onetoone on D . Irvine) Solution. Then m > 0. {fn}r=l ZOI Let m be the minimum value of /).z0I = T } . Hence fn(z) converges to zero uniformly in{lzl 5 arbitrarily chosen. but has a zero point zo E D .(fI on { z : Iz .f(z>I< m 5 lf(z>I holds on {z : Iz .. It follows from RouchC’s theorem that fn and f have the same number of zeros in { z : Iz .
In other words. 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 and f(z) .211 <T} and z: E { z : (z .a # 0 in {z : 0 < ( z .al { z : Iz .z 2 ( = T } .u = 0. This is a contradiction to the assumption that f n is onetoone on 2 D.z2( < T } .439 (b) Suppose f is not a constant. It follows from RouchC’s theorem E { z : Iz .zll = T or Iz .a have the same number of zeros in {z : ) z. Prove that n+m lim f n ( z ) = b for every z E D. such that f(z1) = f(z2) (denote it by u ) .221 5 T } . which implies fn(z{) = fn(zi) ( : # 24). I(fn(z) .a1 on { z : IZ . such that fn(Z{) .)I = Ifn(z) .zll = T or (z . when n is sufficiently large. there exists 2 : . With the same reason as in (a). Then there exist zl.(f(z). 5506 Let D c a! be a bounded domain.) .a = 0 and fn(zi) .a ) holds on that f n (z) .f(z)I < m L If(. (Indiana) .z 2 ) = T } .221 < T } respectively.z11 < T } and { z : (z ..z l ( 5 T } U {z : 0 < Iz .) . Let rn be the minimum value of If(.z2 E D (z1 # zz). and is not onetoone on D. Then m > 0. Choose T > 0 sufficiently small.
(a) 3 = {f : f is analytic in D . Here the diameter of a set S is diam S = sup{lz .440 Solution. Since { f n ( z ) }is a normal family. Because f(u) # f ( a o ) . (Indiana) S o htion. then m = min{lf(z) .m n+ca lim fn(ao) = bo # b.a = T } .Re{g} > 0. which obviously satisfies the conditions that f(z) is analytic in D and f(0) = 0. 5507 Which of the following families are normal. f(0) = 0.a1 = T } > 0. which shows that 7 is normal. there is a subsequence { f n k ( z ) ) converging uniformly on compact subsets of D to f(z). By Rouche’s theorem. f(z) is a nonconstant analytic function of D . Without loss of generality. which is a contradiction to the fact that fnk does not assume the value b E d D in D because f n k is an automorphism of D .a1 < T } . Let T be sufficiently small such that f ( z ) .b has zero(s) in { z : Iz . ao # a. and which is compact? Justify your answers. we assume lim fn(u) = b 6 d D . 1 1 on { z : Iz .b has no zero in { z : 0 < I z .C E D . diam f ( D ) 5 2) (b) B = { g : g is analytic in D. Since { f n k } converges uniformly to f on { z : ( z . diam g ( D ) 2 1). Take a0 E D . there exists a subsequence of {fn(ao)} converging to bo # b. If {fn(ao)} does not converge t o b. (a) For any f E F. Then there exists a subsequence { f n k } converging uniformly in compact subsets of D to f(z). fnk(z) .bl : )z .a = T } .it follows from f(0) = 0 and diam f ( D ) 5 2 that lf(z)I 5 2. when Ic is sufficiently large. Let {fn} be a sequence of functions in F.CI : z .g(O) = l. we have . C E S}.u [ 5 T } c D . nP . For any two fixed points z.
C E K . we have IGn(z)l < 1. we obtain diam f ( D ) 5 2. which shows that family Q is normal.(z) # 0 for all z E D .1 < 1 which is called the kernel of {R. (b) Let { g . = {w : I . Hence we can define an analytic function g ( z ) = log G(z). which shows that is not compact. ( D ) 5 2.} converges to R = {w : ( w .f(0l 52. 5508 Suppose that 1 5 p < 00 and c 2 0 is a real number. It follows from the uniform convergence of {f..) = logGn. where the singlevalued branch is chosen by g(0) = . Because the domain sequence {a. and it is obvious that g n satisfies all the conditions required by the family Q. Then for Gn(z) = egn('). such a mapping gn exists and is unique. We choose a compact subset K c D such that z. by Hurwitz's theorem. since G. } be any sequence of functions in Q. then the constant is el because ~ ( 0 = lim ) egn(') nvw = e1.) .log G(0) = 1. we have G(z) # 0 for all z E D . If G(z) is a constant. and we conclude that g n. Since z . c E D can be arbitrarily chosen. hence f(z) E 7 . gL(0) > 0.441 because diam f n .} on K that If(. 1 < Rew < 3) w 1 w 4 1 1 n satisfying gn(0) = 1. Let 7 be the set of all analytic functions f on { IzI < 1) such that . By the Riemann mapping theorem. if G(z) is nonconstant analytic.31 < 1) U {w : IImwl < . Since diam g(D) = g ( z ) does not belong to the family 0.(z) converges uniformly in compact subsets of D to g ( z ) .} with respect to w = 1. But family Q is not compact.1 < } U {w : I . 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 . by Caratheodory's theorem. 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. a} i.which shows that F is also compact.( . { g n ( z ) } converges uniformly in compact subsets of D to g ( z ) which is a conformal mapping of D onto 52.
(Illinois) Solution. while the right side of the inequality is a constant. It suffices to prove that the functions in 3 are uniformly bounded on every compact set of {zI < 1). Then when n is sufficiently large. B y Hence where +=l. the left side of the above inequality tends to infinity. Let 1 . If it is not the case. Izn .442 sup O<r<l /'^ o If(reioe)lPde 5 c. . < r. then there exist zn E D .f n E F such that zn + zo E D and fn(Zn) + 00. Show that 3 is a normal family.201 Cauchy integral formula. P q 1 1 Then As n + co. We prove the assertion by contradiction.lzol = 3r. The contradiction implies that 3 is a normal family.
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. and that z = g(w)is the unique solution of that tends to zero with w. Hence tf'(t)dt g(w)= GJ. (a) It follows from IwI <m= rp lf(pe")l n that when t E C. ( t ). f(z) # 0 for 0 < IzI < r 5 R. Solution. and apply this to find the explicit (b) Find the Taylor's expansion of g(w). Let C be the circle IzI = p where p < T .w = 0 that tends t o zero with w.w f 1 =5(&J. which implies that g(w)is holomorphic in {w : II w < m}. . f'(0) # 0.p n=O t f ' ( t ) dt) w n . Show that 1 tf'(t)dt g(w)= 2?ri f(t)w J. series expansion of the root of the equation z3+ 32 .
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. Since z = 0 is the only simple zero of f in { z : IzI < p } .444 Let r be the image of C under f where C is the circle { z : IzI = p } taken once counterclockwise.z1) f(t) . we assert that the unique solution g(w) tends to zero together with w .w = (t .zi)Q(t) where Q(t) is analytic and has no zero in {t : I1 t < p } . we know that f(z) = w has a unique solution in { z : 1x1 < p } . + then where . (b) Let f(z) = z3 32.w)]’ = [log(t . Denote the unique solution by z l . and * f’(t) = pog(f(t) . 9 @.w have the same number of zeros in { z : IzI < p } . then f(t).O) = n(r. As the constant term in the Taylor expansion of g(w) is which is obviously zero.w). which shows that f(z) and f ( z ) . Because lwl the winding number < m = min If(pe”)l.
N 1 with residue n at z = n. . we construct For any natural number N .2 . . ..with residue n a t z = n.1 (5)k'] dt (3ck2c". Prove in detail that your function has all the required properties.:) . 5 N } to a function 2N1 n=l which is analytic in { IzI <N}. So f ( z ) is analytic in 1. z2 is a meromorphic function whose only singularities in { IzI < N } are simple poles at z = 1. n 2 2N. * . (Illinois) Solution.( < N } \ { . . 2k t 5510 Find an explicit formula for a meromorphic function f whose only singularities are simple poles at 1. . 2N2 Hence converges uniformly in { IzI In addition. By MittagLeffler's theorem..N + 1 } . .N+lareitssimple poles with residue n at z = n. . .. . it is obvious that f ( z ) has all the required properties of the problem..445 After some computation. 2.l . + .2.2 (:)k2 k + '2kk . 2.3 . Because N can be chosen arbitrarily large..... we obtain a2k1 a2k = 0 and =  zl 1 33k2 32klt2k1 [ 2 '2k . when IzI 5 N . a n d z = 1.
2. (b) Yes. let (z( R and choose N 5 + > R2. The contradiction follows by taking z = 0 in the inequality.1 is an analytic function in maximum modulus principle that 1. It is easy to see that . (b) Does there exist an entire function g whose zeroset is {+( 1+ i) : n = 0.446 5511 (a) Does there exist a sequence of polynomials {P. IPn(z) < E holds for all z E (1 < Izl < 2). .3. if No.. Multiply both sides by )zI2. then for any E E (0. )? If Yes. If there exists asequence of polynomials {P.1 < 2 } . The function g can be chosen as where an = f i ( 1 i).1. give an explicit formula for the Pn . if No.} such that P n ( z ) i uniformly on the annulus 1 < IzI < 2? If Yes. (Illinois) Solution. explain why not.} such that P n ( z ) + uniformly on (1 < 1.we have 5 $1 a). For any R > 0. it follows from the holds for all z E { Izl < 2 ) . (a) No. there exists N > 0 such that when n > N .( < 2}. 5 Because z 2 P n ( z ). explain why not. Then when n 2N . give an explicit formula for g.
al. False... we know that is analytic in {z : IzI < R}.a2. we have 1 5 Refn(z) 5 1 + l o g 2 . and prove your assertion. and its zeros in { z : 1z1< R} are 0. We prove the assertion by contradiction.. (q 5512 State whether the following statement is True or False. we see that g(z) is a n entire function with the required zereset.447 Because C n=N 00 5 converges. Then Fn(z) are entire functions with no zeros. For each positive integer n there exists an entire function fn such that (Indiana) Solution.ak . which implies m is analytic in { z : IzI < R } . Hence g(z) is analytic in { z : IzI < R}. and 5 e 1 ~ ~ ~ ( zeRefm(z) < 2e > l =  . Since R can be arbitrarily large. If for each positive number n there exists an entire function fn such that then for 15 IzI 5 2..1 5 IC < $). Define Fn(z) = efn(').
(a) zi is defined by eil0gz. which implies that F ( z ) = cyz with = 1 in { z : I1 < 2). and there exists a subsequence {Fnk(z)} converging locally uniformly to an analytic function F ( z ) in { z : IzI < 2).48 4 for 15 IzI 5 2. 5 (Indiana) Solution. (b) Why should there exist a sequence of polynomials P. (a) Give a singlevalued definition for zi in G.(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. where D = { z : IzI < 1).(z)I 5 2e for IzI 5 2. But we have for 1 5 IzI < 2. = G . Hence {F. is connected and contains z = 00. For example. C Kn+lr and nrco lim K. Since IFn(z)l 2 for 1 5 IzI 5 2.0]. (b) Choose where n > 2. Then K. and by Hurwitz's theorem F ( z ) has no zero in { z : 1x1 < 2). t contradiction to the fact that F ( z ) has no zero in { z : IzI < 2). In domain G. This is a 5513 Let G = D\(1. we know by Runge's theorem that there exists a sequence of polynomials which converges . such that nrm lim P. Because the complement of K. singlevalued branch of logz can be chosen.(z)} is a normal family in { z : ( z ( < 2). F ( z ) cannot be identically zero. a singlevalued branch of zi in G can be defined by argzlo<2<l= 0. By the maximum modulus principle. IF.
. * .2.) is . a2 1 The singular part of f at z = n (n = 0. (a) Let f(z) = G. Since Pn(z) converges t o zi in G.} converges to zi uniformly on n compact subsets of G. ( z ) which converges uniformly on compact subsets of D to an analytic function f ( z ) in D ... If there exists M with IPn(z)l5 M for all z E G and all n.449 uniformly on K . which implies that z' can be extended to a singlevalued analytic function in D. . the series  . IPn(z)l 5 M for all z E D and all n. and there exists a subsequence P. . (c) Develop a cot az in a Laurent series about the origin directly and by use of (b). f2. with enough terms to find the values of n=l C 5 and C 00 00 1 n=l (Hamad) Solution. fl. It follows that {Pn(z)} is a normal family in D . (c) No. so the contradiction is obtained.n2 ' n=l Justify your steps. It is obvious impossible. then because Pn(z) are continuous on D. In other words. hence zi = f ( z ) for z E G.Now we consider (2n)2. 5514 (a) Prove that (b) Use this to show that acotaz = Z l o o + c2z 22 . to z i . Hence {Pn(Z). we can find a polynomial Pn(z) such that for all z E K. = 1.
. the limit of the series for Imy + fco can be obtained by taking the limit in each term and the limit is also zero. Hence g ( z ) is a bounded entire function. It is obvious that lim f ( z ) = 0. As f(z) and are both periodic functions with period equal to 1. 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. we restrict z in the strip { z :0 < Rez 5 1). Imt+f00 As the convergence of is uniform for lIm4 2 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.450 For any natural number N and la1 5 N. holds for n 2 2N and n n=2N 5 2N.
The singular part of F at z = n (n = 0.n2 + c. we know that is a meromorphic function which has the same singularities as F ( z ) . we consider Now 22 With similar discussion to that in (a).n2 + Then G ( z ) is an entire function. it follows from the fact that F ( z ) and 1 . .G'(z) Comparing this identity with (l). For z2 n=l . Thus we obtain the identity (b) Let F ( z ) = 7rctg7rz. Hence we obtain 7rcot?fz= +c1 " 22 z2 n=l .n2' . . + n=l c 22 . we obtain T2 1 " n=l 1 1 .45 1 which implies that g ( z ) is a constant. Differentiating both sides of the above identity.. have G'(z) = 0 which implies that G = c we ( c is a constant). f l .)is the series 1 " n=l  A.&2. It is obvious that the constant must be zero."+ E n 22 n=l are both odd functions that c = 0. Let 22 1 " F(z)= .G(z).
. = ( z . we obtain After differentiating (4) on both sides. ..2 and g(z) Let f and g be polynomials.z2.. .zn).2 with f (zj) = uj if and only if n (c) Given a sequence of complex numbers z1. f of degree .23 . Take z = 0. 5 n . (a) Show that (b) Show that there exists a polynomial of degree 5 n ..z 1 z 7r2 3 7r4 . rz x .tl) 9 * . such that lzn 1 + 00.. . 45 (3) It follows from (2) and (3) that around the origin.( z . does there exist an entire function f with f(zj) = aj? Can you write this function down?  (Hamad) .452 (c) The Laurent expansion of 7rcot 7rz about the origin is j cot l .. we can also obtain 5515 Let z1. Zn be distinct complex numbers.
453 Solution.2.q ) ( z .z. < RS. . and the coefficient of zn' is 1.z 2 ) * .. then by (a). . we obtain ?#=O.Z n ) is of degree n. (a) Take R sufficiently large such that Zl.Z j + l ) .z j . .(2 .z 1 ) . j=1 (b) If f ( z ) is a polynomial of degree 1.. dz dz = lim = 0.Z. while f ( z ) is of degree 5 n .2. z ) g( z .l ) ( z . Since . E 1(..2 with f(zj) = uj (j = If u l .an are n complex numbers such that n we construct the function f ( z ) by For each j .( z . Because g ( z ) = ( z .zj ."'.( z . . we have 5 n .. u 2 .1. . . .Z2. ( z .. n).) is a polynomial of degree n .
aj. so that f(z) is analytic in 1.( holds for all IzI arbitrarily large.I > 2R 2 00 1 u n ( a ) converges uni5 R when n 2 N ..I + 00. It is easy to see that while for k # n.Zj) . . f(z) is a polynomial of degree 5 n .2. there exists N > 0 such that Iun(z>l 5 Iz. Because lim z+rj . such that .454 the coefficient of znl of f(z) is zero. Then we define  a. uk(zn) = 0. f ( z ) is an entire function.. z2. is chosen such that when IzI 5 k$. Because Iz. . . Since R can be formly for all IzI 5 R. In other words. for any R when n 2 N . by the Weierstrass theorem about the canonical product of entire functions. n=l < R}. while for k #j. which implies that f ( z ) is an entire function satisfying the required condition. (c) For the given sequence z1.  where 7.’ajz j ) g ( dz) ~ ( z . we can construct an entire function g ( z ) with simple zeros z l .2. Hence > 0. z2.1 I +00.n). f ( z ) satisfies the condition f(zj) = aj ( j = 1. In other words.
Part VI Partial Differential Equations .
.
. Here (. . a) It is not difficult to verify that aijxjxjdx = 0. Show that ( 2 .trace A Wn 1 n n(n + 2) trace A.j)&=l be a real matrix.457 SECTION 1 GENERAL THEORY 6101 a) Let A = (a. b) Show that u E C:(R") implies (Iowa) Solution. Ax)dx = Wn L<l n(n + 2) trace A. lxl<l Vi # j and Therefore. =  . ) is the dot product of vectors in R" and wn is the area of the unit sphere in En.
By applying Green's formula. IR Take a function p E C r ( l R )such that J. 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. Show that T = const. R (Indiana) Solution. the conclusion is true for u E Ci(IRn). 6102 Let T be a distribution on IR and suppose that T' = 0 on IR. T' = 0 if and only if T(q5')= 0. g d x = 0.. we get immediately As Cr(lRn)is dense in Hi(IRn). show that there is a number a such that T(q5)= / aq5dx for all q5 E C r ( l R ) . i.458 b) First let u E Cr(IRn). Vq5 E C r ( B ) .e. . Then it is easy to verify that for any q E C r ( I R ) 5 satisfies the condition ? J.pdx = 1.
l)) by Determine whether u is a distribution on (0. as Ic + 00. Then we get immediately + + 6104 Let u be defined for f E D((0. Then we have a compact interval [a. It is clear that u. (Indiana) Solution. is a linear functional in D((0. l). we have 6 E L 1 ( R n ) . 14+w ( cincinnat2) Solution. u E H s ( R n ) if and only if (1 lt12)'/2G E L 2 ( R n ) . Therefore.a / 2 E L2(Rn) if s > n/2.l)). In fact.b] c ( 0 . Let {fk} be a sequence in D((0.l)) such that fk + 0 in D ( ( 0 . We show first that 6 E L 1 ( B n )if u E H s ( E n ) with s > n/2. Show that lim u ( z ) = 0. l ) such that SUPPfk c [a.bl. vk: . l)). defined above. And it is clear that (1 1[ 1 2) . and support your answer.459 Hence T($) 0 and = 6103 Let u E Hs(En) with s > n/2.
dxdy = J. XY 2 ( . ) 5 belong to WIJ’(B)? (Iowa) Solution. u is continuous in D ( ( 0 . u. b] as for any nonnegative integer n. in the sense of distributions.l).460 and jp). Let uy(x7 y. 1 N Then we have Therefore.and is a distribution on (0. that is J.< a.#dxdy .uy E L1(B). + oo.l)). 6105 Let B = {(x. y) I x2 For which p 2 1 does the function + Y2 < 1. o uniformly on [u. + y2)3/2 * It is clear that us. u#. Let N be a positive integer such that .
Derive a necessary and sufficient condition in terms of v +. Denote by v = v(x) the unit normal to C pointing into R 2 . x)ds for 1 5 p < 2.C ( *I in the classical (C')sense where b is a continuous function. Then we have < 1... (Indiana) Solution.461 and r JB u+ydxdy = . we get &u = u.dzdy =  lc ux4dxdy + 1. v.and v for ZI to be a distribution solution of (*) on all of R".. Similarly L u = 2ly. And we can verify that for which means that u @ W1yP(B) p = 2. Letting E . 0 in the above inequality. Assume furthermore that v lies in both Co(?&) and C0((n2)(but not necessarily Co((IRn)!!) with v+(xo) lim v(x). Therefore. u E W1ip(B)for 1 5 p < 2. u~cos(?~'. B.u4. v is a distribution solution of (*) in En and only if if . we denote the ball centred at the origin with the By and S. Suppose v : R" + R" solves the equation divv(x) = b(x) for x E R" .JB uy#dxdy r for all radius E E Cr(B). 6106 Q . = aB. Let RE = B\BE for any E C r ( B ) . and Let C be a smooth hypersurface dividing R" into two disjoint open regions R2. ' BY It is not difficult to verify that + J. vu(xo) lim v(x) xxo xDO +En1 XEn* for each xo E C .
vds = 0. V#dx  J.v E Lloc(B2)... we have Ln 2. Therefore v is a distribution solution t o (*) on all of B* if and only if (w+ . 2.(IRn). V#dx = IRn V# E CF(lRn)..V) .. V+dx = . 2). 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.v . (a) Let u. (Indiana) Solution. describe as completely as possible the curve r.y)): vxuu~2Lyvu.v) uds + lRn divv . Then curl U = 0 . Recall the definition of the curl operator in two dimensions acting on a vector field U(z.v(x. J.462  b#dx. and U1 and U2 are (different) constant vectors.y) = (u(z. If U has curl zero in lR2 in the sense of distributions. (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# E C.y). By Green’s formula..J. V+dx =  +(v+ . #dx.v) .Y = o 6107 on C.I’ z d a l n a 5 2 2 is a smooth curve.
4  Construct a function X R E Cp(lRn) such that XR(<) = 1. Prove that there exists a tempered distribution E on lRn such that p ( D ) E . Therefore. (Indiana) Solution. WI). and assume that the set {z : p ( i z ) = 0) is bounded. w1. As the set {< : p ( i < ) = 0) is bounded.(lR”.6 E S.u2)dz + (v1 .463 in the sense of distributions is defined as (b) Let 171 = (u1.vz).v2)dy = 0. we have J. I’ is a straight line defined by (u1 . U2 = (uz. u2 and 712 are constants. where S’(W) denotes the space of tempered distributions on B”.v2)4dy = 0. where Then curl U = 0 if and only if u1.(ul . and 6 is the Dirac distribution defined by 6 4 = 4(0).%)4dz + (v1 . that and E . v< E {< : < R}. there exists a positive constant R such that {< : P(i<) = 0) c {< : 1 < R). Integrating by parts. 6108 Let p be a polynomial in n variables. v4 E C. Here S is the Schwartz class of rapidly decreasing functions. Set It is clear that E S ’ ( R n )n Cm(Rn).
) and the C. From the transformed equation. Let G(E) be the Fourier transform of u. Therefore.D" bllm where a denotes a multiindex Q = (a1.a2. .464 It is easy to see that p ( i € ) Z ( € ). P(D)= C.Transforming the equation. .a. prove that u must be a polynomial.. we get  C < P ( g ) G ( [ )= 0. (Here S' denotes the space of tempered distributions on an. Let P ( D ) be an m t h order linear constant coefficient partial differential operator on En.. . we can see that suppG = { 0). Therefore.(i<>" o for all E # .) (Indiana) Solution. 6109 s. we get p(D)E.(nZ").S(")(<).e. . are constants.' ( Z ) E S'(nZn). are constants. Suppose ~(i<) E ~. G ( t ) must be finite linear combinations of the derivatives of the Dirac delta function S(<) at { 0 } .^(€I = + c a.a.(0). l4lN where a = (a1. i..e. .. Hence is a polynomial. i. I4<m If u E S' satisfies P ( D ) u = 0.6 E where E = F .1 E C.~ 2 .) is a multiindex and a.
Further. c ( . a ) Vn. (d) Calculate the Fourier transform of the distribution (Indiana) Solution.. define the Cauchy Principle Value and show that it is in D'). as (b) We define the distribution follows: 5 by its Cauchy Principle Value &: The linearity of the functional Vp: is obvious.( 1.e. 4 E D. exaf4tdx = 1 Vt For any > 0.465 6110 in D'. Let supp4. (b) Show how to define as a distribution (i. (c) Calculate rigorously the derivative of log 1x1 and express the result in terms of your answer to (b). Then we have + 0 . (a) It is wellknown that 5.a . suppose that q5n in D.we have Then it is not difficult to verify that Hence t+O+ lim  & ex2/4t 6 .
' ( c ) For any 4 E D . is a distribution in D .466 Therefore. we obtain 6111 a) Let Lu(2) = bI<m c a. we have applied the integral equality 'J00t 'IF sin(zt)dt = 1 for z > 0. Therefore.we have V p i Hence &log 1 1 = V.DcIu(c) 2 €mn .:.we have Here. 2 (d) By F ( f ) we denote the Fourier transform of a tempered distribution f For any 4 E S.
and a is a multiindex.t).)1. (Indiana) Solution. then the operator L has at most one fundamental solution in S‘. If L*S is dense in S... 0) = f(z). 2 is a fundamental solution of the wave operator W ( u )= utt .u. c) Express the solution of the Cauchy problem utt . Let L* be the formal adjoint of L defined by L*V(C) = lallm c (l)~aba(aaV(z)). Simplify your result to obtain D’Alembert’s solution of this problem.g. 00 (f. #J sufficiently nice) in terms of the above fundamental solution.u.467 where the abs are constants. where S(z) is the Dirac function. = #J(z. where S is the space of rapidly decreasing functions. O ) = g(z). t ) = I+(t)l+(t. u ( z . on E 2 . t > 0 00 < 2 < 00 U t ( 2 . What is a fundamental solution of L? When is it unique? b) If I+denotes the Heaviside step function Show that 1 F ( z . a) A distribution E E D’(nZn) is called a fundamental solution of the differential operator L if L E ( 2 ) = S(z). < 2 < 00. b) We need only to show that In fact .
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. q t . it yields that 3 %t) = Therefore. By applying Fourier transform to the problem. respectively.469 This is just the D’Alembert’s formula when q5 2 0. 6112 Let u be the solution of the initial problem where summation is understood in the pde. we have Set 77 = tit in the integral in the right side of the above equality.0) = w. qt. we have ztt+ i t 1 4 6= 0. Assume that there is a constant C such that (Indiana) Solution.0) = 0. and 4 is a C” function on IR3 with compact support. It is reduced to . By solving the above transformed problem.
470 This implies that In fact. given in the problem.t)I2dz = I&0124 Here 3 denotes the Fourier Transform of 4. K E E3. then there exists a positive constant r such that I l2 a.1 < r. m 4 And when t > 5 . (Indiana) Solution. consider the solution u of the SchrGdinger Equation Show that lu(z. It can be verified that the Poisson's formula for the Cauchy problem of the heat equation applies to the Schr6dinger equation.we have This contradicts the condition satisfied by the solution u. So we have And then . 6113 Given E S (rapidly decreasing functions) over B1. if I&(O)l = 2a # 0.
we get Here the Fourier transform of 4 is defined by .47 1 Set x = 2&.
472 SECTION 2 ELLIPTIC EQUATIONS 6201 Let u $ 0 satisfy u E C2(Rn). Applying the meanvalue property for the harmonic function u. is the surface area of the unit sphere in En. (Indiana) Solution. Schwarz's inequality gives that is JT u 2 ( x ) d S z >_ w. There exists a point xo E En such that u ( x o ) # 0. Show that does not exist. n . we have where w . zz*(=p Then we have u2(x)dx wnu2(xo)Jlo pn'dp w 2 = u n ( x 0)T n .pn1U2(x0). = 0 on Au R".
(Iowa) Solution. Set where C is a constant such that For x E R. It is wellknown that c R. an). we have where This implies that for any compact set K dist(K. the relation uAddx = 0 holds for all 4 E C. set PE(Y) = EP Then it is easy to see that provided E 2Y ( ) 7 P4Y) E C. Therefore. . 6202 Let u E Co(R) be weakly harmonic in an open set R. +u uniformly in K . as E + 0. an).e. is harmonic in K if E < u. Therefore u is harmonic in R.(f% < dist(z.473 for any T > 0. The conclusion follows as T + 00.. i. u.(R). Show that u is then harmonic in R.
. Then the conclusion that f ( z . . yo) E B4and R such that lf(z. i.474 6203 Let f (2. y) be a locally bounded function.y) E . we have By Green's formula. Let (zo.. Then there exists a constant M > 0 lR4.y)I > 0 be given. y ) is continuous at (zo. as a function on y). (Indiana) Solution. Prove that u E 0 on z 2 0. we get we have the same estimation. . For z.e. (iii) u is harmonic in z > 0. V(z. 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. (Indiana) . harmonic in z E lR2 and continuous in y E BIZ. Applying the meanvalue theorem to the harmonic function z. These estimations imply that f ( z . B4.Show that f is continuous in (2.yo) E B4follows immediately.z0l2 + Iy . y ) is continuous at z = z0 uniformly with respect to y near yo. 2 0 of En. 1 2 5 M.y0l2 < 2R2.
is the surface measure of the unit sphere in B”.475 Solution. Redefine the function u in the halfspace x. > 0) and continuous on y the boundary alR. are bounded in show that its “tangential” derivatives magnitude by maxlVgl on all of B = {x : xn > 0). u is bounded and harmonic in B”.e. = u(x’. By the uniqueness of the bounded solution to the Dirichlet problem (3).1) and w. we can verify that the redefined function u is harmonic in any ball centred a t the origin in lRn. Thanks to the Liouville’s theorem.. < 0 as a n odd function of u(x‘.x~). where X’ = ( ~ 1 .* . n 1. It is not difficult to verify from the above formula that and . = 0)..**. = . < O where x’ = (xl. If u. . (Warning: you must T justify any assumptions of regularity you make on the solution u. C1function on alR: = {x : x. Therefore.bounded. we get u s 0 immediately. By using the uniqueness theorem of the Dirichlet problem and the Poisson’s formula of the Dirichlet problem in a ball.xnl). VX.(.. ~ ~I‘ 1 )([1. 6205 Let u(x) be C 2 on the halfspace B = {x : x. we can show that the unique bounded solution to the problem (3) is given by e. satisfies j = l . = 0 ) . and let g(x) be a compactly supported.2n) 2: . = {x : x.) (Indiana) Solution. * .
6206 Show that the problem Au = 1 for 1x1 < 1. ( Cincinnati) Solution. < 1. _ . we have uAudxdy . < 1. < 1. (Here u = u(x. It is easy to see that lXl uAudxdy = 0.au = 0 for IyI = 1 ax ay has the unique solution u 0.lY I < 1 Integrating the above integral by parts. au .. IyI u = 0 for 1x1 = 1. y) is a function of two variables x and y.476 Therefore. We need only prove that the problem Au = 0 for 1x1 < 1. we have which are just we want to show. au au = 0 ax ay for IyI = 1 has at most one solution. IyI u = 0 for 1x1 = 1. . and is a classical solution).
u2(1. .u y 1 . and integrating the resulting equation in R. (a) Multiplying the equation in the problem (*) by the complex conjugate of the solution u.1) . 1)) Then we get u f 0 immediately. 1) 1 2 + u2(1. and suppose nontrivial solution exists to CY is such that a { V2u = a 2 u in R. Show that r (IndianaPurdue) Solution. u=O on a R . Then it is clear that u can be chosen to be realvalued. and that u can be chosen to be realvalued. which implies that a is a real number.1) . (a) Show that a is a real number. (b) Suppose a1 # a2 are such that nontrivial real solutions u1 and u2 satisfy (*).477 = (u2(1. 6207 Let R be a bounded normal domain in lR3. we obtain (Vu12dw= a2 IuI2dv.
r ) cc R means that the closure of B is contained within R. By a standard argument. we can show that u can not take its maximum in a unless u is a constant. Let u be a continuous function on a. one has u 5 h in B. . we get immediately which implies the conclusion of the problem. Let u be subhamonic in the sense of (i).Prove the equivalance of these two notions of “subharmonic”.) (ii) For every ball B cc fl and every harmonic function h E C o ( B )such that u 5 h on d B . we Vul . Vu2dv = a: and From the above integral equalities.478 (b) Suppose that equation satisfied by u1 u 1 and u2 are chosen to be realvalued. respectively. Multiplying the V2u1 = a9u1 in R and the equation satisfied by u 2 v2u2 = aiu2 in by u2 and can obtain u1. 6208 Let fl c B“ be an open set. and integrating the resulting equations in R.r ) cc R one has (Here un denotes the surface area of the unit sphere in Enand B ( z . (Indiunu) Solution.r ) centered a t x with radius r satisfies B ( z . (i) For every x E R and r > 0 such that the ball B ( x .
By using the Green's formula.YI = 71.)l where d ( z ) = dist(z. /n is the volume of the unit ball. r ) . a) Show that is harmonic in R for each i.h is also subharmonic in the sense of (i).Let R E B" be an open set. if R is bounded and n E R. where B is the open ball centered at x of radius r . b) Applying the meanvalue theorem for the harmonic function ball B . :Y E lVU(. c) Show that. and let u be harmonic in R and continuous in R. h=u on a B ( z . we have au n (z) =dXi in the wn r n where w. let u be subhamonic in the sense of (ii). Conversely. and the meanvalue theorem. the above equality can be rewritten as . a) The conclusion is clear. r ) . is the surface area of the unit sphere in En. b) Let B c a.479 Let h be the function given in (ii). L SUP{lU(Y)l 4s) 0) (Indiana) Solution. { From the definition of subhamonicity in the sense of (ii). 8 0 ) . We define the function h as follows: Ah = 0 in B ( z . Show 2 that IV4z)l L. we have This completes the proof. Then w = u . hence w. SUP~lU(Y)l 1% : 3 : n . Hence that 5 0 on 8 B implies that w 5 0 in B. 6209 .
Let u E C1(n)n C2(R) solve A u + k 2 u = 0 in R (k > 0). T r Let u be a solution to the equation A u + k2u = 0. we have c) The conclusion can be obtained from the result of the part b) immediately. Let R < dist(zO.cos(kr). By the fundamental solution cos k r .480 And therefore. . there are two linearly independent solutions: 1 1 u = .sin(kr). 6210 Let R C lR3 be bounded and open with smooth boundary. ds2) be a fixed positive constsant. For this ordinary differential equation. Then the spherically symmetric solutions depending only on T satisfy (TU) d2 dr2 + k2TU = 0. Applying the above integral formula in . (Indiana) Solution. it is not difficult to verify that & where T. = ( z .O.R) we denote the ball centered at zo with radius R. Let T = 1x1. By B ( z o . First we look for the fundamental solutions of the equation.2'1. Derive an appropriate meanvalue property for the solution.
E aB(zO. It is easy to see that the function g(z.we obtain u(x0) = a ..) rxox satisfies the above conditions. Txox  4. we get Now we look for a function g(z. Prove that its spherical mean. it is easy to verify that Subtracting the equation (2) from the equation (l).zO)= cos(kT.R).z0) = cot(kR) sin( kr. which satisfies (A.+ k2)g(z. defined by . 2 g(z. For this function.R2) satisfy AU = u + 1.o.o. R).zo).cot(kR)1 sin(kr.rrRsin(kR) J BB(xo.R). Let u E c2(. z E B ( z o .).R) This is a meanvalue theorem for equation A u 6211 + k2u = 0. u(z)dS. zo) 1 47Frxo.o. = 0.481 the domain B ( z o .
r )denotes the ball of (Indiana) 1 =21r L B ( 0 .482 and v(0) = u(O). 6212 Let B = {(z.y) I x2 + y2 < 1). Prove that the problem Au= in Jm B. Here. l ) u(rw)dS. dSy the element of arc length. 1 u=O onaB . Therefore + The proof is completed. B ( 0 . W(T) # 0. for z radius T . and Applying Green's formula.also satisfies ( l ) . Solution. Set T = 1 1 Then we have 2. we get Then it follows that = r(w(r) 1).
And denote the corresponding norm in W't2((n)by 11 ( ( 1 . fvdx 1. This implies that f E H .' ( B ) . (c) Soppose that f . we get + J. and (1) L ( g r a d u ) .a w + ax dy in the sense of distributions (see the solution to the problem 6105). g and 80 are sufficiently smooth. (a) Define the inner product in W172(f2) by (u.  . = h[(gradu) (gradv) v)1  + Xuvldz. (Iowa) Solution.v E W'y2((n). (b) Will the conclusion of part (a) be valid if X = O? Explain. w E L2(B) and f = a v . Show that the problem (1) has a unique solution u E W1!2(R). Consider the problem aQ. gvda v v E w'q(n) Vu. Then by applying Cauchy inequality and the trace theorem in Sobolev's spaces. Hence the problem admits a unique weak solution u E Hip).and g E L2(dR) be given. Let X v= d 3 w = &5qj5 and f = 1 &Gp' It is easy to see that v . ( Cincinnati) Solution. (gradv) + Xuvdx = where X > 0.483 has exactly one weak solution u in HA(B). write problem (1) in the classical form. 6213 (a) Let R be a bounded domain in lRn with smooth boundary f E L 2 ( f i ) .
WX'. z n . it is easy to see that the conclusion of the part (a) is false if X = 0 even for smooth f and g.1 on R n {zn = 0). V' ).= u Inn(z. Consider the following problem Find u E Ht(R) such that lVu. ( Cincinnati) . (c) Suppose that f . Denote u+ = u Inn~X.<o}. an + 6214 Let R be a unit ball in IR" centered at zero. (b) By the classical form of the problem. The uniqueness is clear.= g on dR.44 8 where C is a constant. Therefore. where IR" 3 2 = (21. z n ) = (x'. the classical form of the problem (1) is Xu = f in R.Vq5dx = a . Show that Au* = 0 and that du ax. Jnn {X . Then u is a solution to the problem (1).l ... =o} . Integrating by parts. x. V#€H. Prove the existence of a unique solution of that problem. Thanks to the Riesz theorem. F(v)= fvdx +lngvdo is a bounded linear functional in W'?'((n). Au dU . Therefore. g and dR are sufficiently smooth. there exists a function u E W'?'(R) such that F ( v ) = (u. azn du+ . v v E W'J(R). and assume that U* are regular enough.).o} and u.1(R).
n { x .= R equation by parts. The proof is completed. there exists 1 (%4)1 u E Hi(R) such that = / nn{ I=o} .=o} 4d~'.(Q).n{X. Write the equation in the following form Vus V 4 d x Vu. nn{ x . Define the scalar product in Hi(R) by By the Cauchy inequality and the trace theory in Sobolev spaces. < 0).a x . ""'> / dx' = au+ ax.. The uniqueness of the solution to the problem is clear. =1 on Q n {x.11 is the norm in Hi(R). = 0). =o} 4dx'.4dx  l Au. auax. v4 E m Q ) . Integrating the above  l+ 1 Au+ . we get immediately that AU* = o in a&. : where I(..V 4 d x = where R+ = R n {z. we have + 1. Mx'. From the above equation. ax. According to the Riesz theorem. This proves the existence of the solution to the problem. + L  J. . V4 E H. Assume that u* are regular enough.485 Solution. > 0) and R. we can get I Cll4ll1. . 4 d x + nn{ xn=o} (E.
in R.V 4 d ~2 0 inf u n V 4 E W.u. IV412dx 5 0.486 6215 J. 4 2 0.e.’”(R). 0). Show that u 2 0 a. From the above inequality and (I). 4 E W. an Assume that inf u = I > 00. the proposition is true. we obtain Hence 4 =0 This implies that u 2 I a.e. b) Let u E W1?’((n)satisfy (I) above. show that 2 inf u an (essinf) ( Cincinnati) Solution. in R. a. a) This is a corollary of the conclusion in b) of this problem. b) If inf u = 00.’”(R) and Now we have Vu V4dx =  J.e. in R .VU . it holds that IV+(2dx= 0. By the PoincarC inequality with the above estimate. Let an d(z) = max{l Then it can be verified that .
= 0.'72((n). we get By using Green's formula and noting that u.nw.11i2 Then we have  = 0. + = w..n11w. and integrating it on R. Multiplying the first equation by v. Au.dx . we obtain {u.} is bounded in L2(R) and w. both equations above can be written as Vu. + o strongly in ~ ~ ( 0 ) . and then integrating the resulting equations in R. + +oo. w E W292(R)n W. respectively.. Vw. that such Awn + u . and w. w strongly in L2(R).v.'12(R).. Multiplying the first equation and the second equation by u. w. as n ( Cincinnati) Solution. Therefore. in R. . we have .487 6216 Consider functions u. Prove that '11. E W.
This gives us the conclusion of the problem. 6217 Assume R c Rn is a bounded open set with smooth boundary and let f E cF(R). J. and Av.AunAvndx + nllVun11i2 IIVUn11ia Hence + nllVvn11iz = k VW. it follows from above equality that Multiplying the first equation and the second equation by Au. This implies that {un} is bounded in W.2. Vv. Multiply the first equation by Av.12(s2).dx =  J. = 0. finally (4) By similar consideration. it follows from the above equality that Av. Vundx. .. respectively..dx  .. Av. we can get J.. uk(x)2dx 5 M for k = 1.488 By noting (1) and (2).Suppose uk E cm(n) satisfies Auk +uk = f and has the property that for some positive number M .dx.~ ~ A . we can get . and then integrating the resulting equations. (Indiana) + . ~ ~ ~ ~ v Vu.Vw Vv. +0 strongly in L’(R). .dx.IIVu. 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. By noting (3) and (4).Au.Il$ = LwAu.
r) T. v x E 0 1 . By the regularity theorem of the elliptic equations. Now we show the second conclusion. we have u E Cm(R). Let R l be a subdomain of R such that 2 CC R and R = dist(dQR1) 1 fixed. where W k and z k are defined by and . Set V k = u k . by B ( z . s there exists a subsequence { u k j } such that u k j converges to u weakly in L 2 ( R ) . we have + h(T)vk(z) = 1 vk(Y)dSy. " ' . As { u k } i bounded in L2(n). by C we denote various constants. aB(X. we find that Integrating the above equality from Here R can be chosen so small that Ivk(z)I 1 h(r)dr # 0. Then Au + u = f in the sense of the distribution. For any z E 21. Therefore. it holds that : (1) 5 c. r ) we denote the ball with the radius T (< R) centered at z. It is clear that v k satisfies A V k + V k = 0 in R and Hereafter. Set vk =w k + z k . see the solution to the problem 6210). and therefore in D'(R).u1. By the meanvalue theorem for the equation Au u = 0 (for the case n = 3. 2 .489 Solution. k = 1 . where h(r) is a continuous function of 0 to R with respect to T .
V k = 1. k = 1. * * .2. are given smooth functions.a . k) a 1 c.e. 6218 Let R be a bounded Lipschits domain in B". (a) Let {un} be the minimizing sequence of I(w) in A.respectively. there exists u E A satisfying I(u) = minI(v). (a) Show that if A # 8. vx € ill. and A = {v E Hi(R) : h l 5 v 5 h2 a. VEd Define the norm of w in Hi(R) as . we prove the second conclusion of the problem immediately. Then by using the AscoliArzela theorem. 5 c. By applying the solution formula of the Dirichlet's problem for the Poisson's equation. we can get from the estimation (1) lVWk(Z)( <_ It is clear that Igix 1(1 Z.2. .e. n+w lim I ( u n )= inf I(w). i. (Iowa) Solution. wed (b) Show that L V U V ( W u)dx 2 0 VV E A. h2 : R + . And by applying the meanvalue property for the harmonic solutions..lR.. where hl. . in R}. we can obtain from the above estimation IVzk(x)) 5 C in any compact subset of R1.
The proof is completed.491 Then we get the boundedness of { I n l } Therefore.  6219 Let R c IR”. . in R.such that u u . This implies that I(u) = min I ( w ) . 0 5 t 5 1. and in H.denoted still by {u. there exists a subsequence Iull.~(R) u t u in L ~ ( R ) . Therefore we have + F’(0) 2 0.u)E A.}.}.}. V(u t ( w . wed The conclusion of (a) is proved. we get u E A and that by the weak lower semicontinuity of IIwII1. be a bounded domain with smooth boundary. w F ( t ) = J. of {u. n > 1.u))dz.e. * Then F ( t ) has the minimum at t = 0.u)) V(u+ t(w .u)dx 2 0. Let r + v E A. This implies r Vu V(v . Therefore.such that u + u a. Let u E C1@) be harmonic in R. . And there exists a subsequence of {u.}. for simplicity we still denote it by {u. . (b) It is easy to verify that u t ( v .
201 I u ( x ).u)= 0. you may cite.@2(z0) ) x . u ( z )5 Therefore.2 . Prove that m3x l v u l s rnax rnax lVaI1..492 a. rnax l n an an AQ2 2 0. @ 2 E C 2 ( Qn C1(t) such that ) A@I 5 0. we have A(Q2 . any standard maximum principles you are using.u = 0 on 8Q. vz E s2.) b.u ) 2 0 and @2 in R . without proof.u. The conclusion is an immediate corollary of the following inequality: n i. () Similarly. From the maximum principle. Prove that ma IVuJ= rnax ( V u ( . it holds that @2(2) vx E s2% . (Indiana) Solution. ( ~ m 2 l ) . @I(. n an i.j=l b.u(z0) 1% .). a.e. we have 5 u(z). it holds that sup(a. For the function @2 . . @I =u = on dQ.u) = rnax(a2 .2 0 ) This inequality implies the conclusion of the part b. %. x n an (Note: In both parts a and b of this question. Suppose there exist functions @ I .
do. i. ~ ~ 2 U Therefore. (b) Show that if X > 0 is sufficiently small. . the only solution of (1) is u (1) 0.0 5 u 5 1. (a) If u 2 1 is false. (a) Prove that for any solution u of (l). 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.j=l u2izj +~ u ~ I V 0. u = 0 on dR. For the maximum point xo.493 6220 Let R be an open subset of Rn. It is easy to see that w E C2(R)n C'((n) and n Aw = 2 3. Prove that IVu(z)I 5 1 for all x E R. (Indiana) Solution.. Set w = IVu12. Au(z') 2 0.e. then there exists xo E R such that u takes maximum at to and u ( 2 ) > 1.u) in R. and dR sufficiently smooth. ( Cincinnati) Solution. 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. Hence w 5 1 on 6221 Consider (a nonlinear) equation Au = Xu2(1 . where X > 0 is a parameter.
= 0 on dR.U)(ZO) < 0. Show that the only solution of the boundary value problem for the biharmonic equation: A(Au)=O i n Q u = 0 on dR. Let ?i' be the exterior unit normal on dQ. it holds that Applying PoincarC inequality. we can get J. we obtain from the above inequality where C is a constant. If X < Cl. Assume u E C4((n). Hence u 5 1. As Xu2(1 . we have r 0 = JnA(Au)uda . (IndianaPurdue) Solution.u3(1 By using the conclusion of (a).494 This contradicts the assumption that x u y 1 . 1vul2dz = x J. dU . Integrating by parts and taking the boundary conditions into account. u can not take minimum in R unless u f 0. Hence u 2 0. (b) Multiplying the equation by u and integrating the resulting equality on R. dn is the trivial one u E 0. 8222 Let R be a bounded normal domain in IRn with smooth boundary dQ.u) 2 0. we have u E 0 immediately.
Taking into account u = 0 on immediately. af2. u is harmonic in f2.495 = lA~1~dx. we get u 0 . Therefore.
. Then we get lu(2. J 2 m m By the Cauchy inequality. we have Set y = 2 &q immediately + in the second integral of the above inequality. t > 0. and satisfies J_. (0O IR1: u(2.t > 0) (0O < 2 < m).r nf ( y ) e . Show that there exists a constant C such that for all 00 < 2 < 00.O) = f ( 2 ) < 2 < 0O.496 SECTION 3 PARABOLIC EQUATIONS 6301 Consider the Cauchy problem for the Heat Operator in ut = u. (Indiana) Solution. continuous.W d y . where f is bounded.t) = . The solution to the problem is given by the Poisson’s formula u(2. 03 lf(”)I2d2 < 0O.t)l c I t1/4‘ .
t)l&(n)? E > 0.497 6302 Let R c Bn be an open set with smooth boundary and suppose u E Cm(R x [0.t)2dz. 00). we obtain Solving this differential inequality. we have . Multiplying the equation by u and then integrating the resulting equation on R.t)ll:z(n) where C > 0 is a constant. (Indiana) Solution. Assume that Let $ ( t )= 1 n u(z. it holds that Taking E = 1 / C . Prove that $ ( t ) + 0 as t 4 +m. we can get The PoincarC inequality gives IIU(*. For any 5 CIIVU(*. m)) is a solution of the equation with u = 0 on a R x [O.
. t ) = 0 (t > 0) ut = u .Show that ta. d) Show how to solve the problem > 0 .O) = f ( z ) ( 2 > 0) U ( 0 .0) = f(z).t)ll&n) . 6303 Consider the Cauchy problem = u. a) Using the Fourier Transform. t > 0) u(z.0) we get . ut where f is bounded and continuous. . . and write down the solution u. Transforming the following Cauchy problem with the initial data S(z): ut = u . construct explicitly a fundamental solution for this problem. < 2 < 00.t ) = 0. Why is the Tychanov nonuniqueness example possible? c) Suppose in addition that f E L 1 ( B ) . (00 < 2 < 00.t > O ) . t ) with respect to z. (Indiana) Solution.t) be the Fourier transform of u ( z . ( = 6(2). 4 2 . a) Let G(<. . (2 by using a) and an appropriate extension of f.It is not difficult to verify that if l/f[. This completes the proof. lim u(2.t > 0) 4 2 . b) State a maximum principle for this problem.) 0 as t + +m.
t ) s is bounded and continuous on lR x [O.T].T]. t ) E lR x (O. c) The conclusion is clear. t ) 5 sup f(z) for ( z . < 0.499 The solution to the transformed problem is given by G(<.t ) = etzt.) = { f(x>.t > O ) . Then it holds that XER J inf f ( z ) 5 u ( z . f(z). we get the solution to the initialboundary problem 6304 Consider the problem . t ) to the general Cauchy problem with the initial data f ( z ) u(z. 4 2 .t) = 4(y)eWcly.and satisfies the heat equation in lR x (O. we can get the explicit representation for the solution u ( z . Then the inverse Fourier transform of 6 gives the fundamental solution t o the From this fundamental solution. Then by solving the following Cauchy problem: ut = u x x (cm < 2 < cm.T]with the initial data f ( 2 ) . O )= 4(z). XER The Techanov nonuniqueness example is possible for some unbounded solutions. 2n21 b) The maximum principle can be stated i ~ follows: Suppose that u ( z . d) Defined an odd function 4(z) as follows: 4(. 2 2 > 0.
(z E IR.O) = uo(z) E > 0) L2(E). . we have 1' 1: tlut(2.Ut = u.(z.t)I2dz t > 0.t ) d z d t . it holds that We can also get the result by the method of energy integrals. From the Poisson's formula for the Cauchy problem of the heat equation. in (Indiana) Solution. t ) u t ( 2 . Find a bound on ': 1 Ju.(z. we can find that Then by using the Cauchy inequality..t u(2. we have And therefore. Multiplying the heat equation by u and tut respectively.t )12dzdt = I'L: tu. . we can obtain and Adding the double of the second equality to the first one. and integrating the resulting equations in x (0. T).
forming the top of ~ R T and let B = BRT\B~. t ) with z E R and t < 2 Let B1 denote the open linesegment ' .Let u ( z .t ) u = f(2.. we get an estimation 6305 For the indicated domain R (interior of the parabola) let RT denote the points ( z .(z.T)I2dz +T Therefore. 2 ) u.ut + a(2. t> .501 otherwise an approach to the function U O ( X ) by a sequence of functions in Cr(En) be used. we have A/m 2 w oo ~ u ( ~ . ut E C o ( R U B1) be a solution of ~ %x . can From the above two equalities. T ) ~ ~Ld_zlu.. t ) E C o ( n ~with .
.(2 0 . Suppose that u assumes its negative minimum at ( .tO) 0.C. (Here a < 0 should be assumed in RT u B1). What can you say if a < 0..502 with a.t o )5 0.f E Co(D).. Show that if a < 0. f 5 0 in RT.tO) 0..t 0 ) . f 2 0 in RT? (Iowa) Fig. = 0.tO) a(zO.(zO.t)= 0. then every nonconstant solution u assumes its negative minimum (if one exists) on B2. G O .1 Solution. u(z. If a < 0. f 2 0 in RT. Hence u..6.ut(z0. 6306 Consider the boundaryvalue problem B. I.tO) 2 0.to)u(zo.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. ut(z0. u. UP(. ut = 0 u(0. Then we have u(zO.we can say that every nonconstant solution assumes its positive maximum (if there exists one) on Ba.C. > + This contradicts f(zo. (IndianaPurdue) .t) + (0 5 z 5 A and t > 0). t o ) E RT U B1.O) = f(z).
(x E t . we can find the solution u to the problem with the initial data f(z) = fn(z) .x .  a.) d at 2 (<G) a at = u I.503 Solution. It is easy to see that F(zu. By G(<. t)ll + 00 as n + 00.t > 0) Assume uo E S and find u.) (Indiana) Solution. if not. By setting u = f.r.r]. in which the particles of Brownian motion are subjected to an elastic force. 0) az. leave your result in the form of a onedimensional integral. f n ( z ) is the eigenfunction of the corresponding twopoint boundary value problem.) = iF(u. When n is even integer. One is required to solve u = u . . (You should be able to find u explicitly. is clear that It But for any fixed t > 0 IIu. = uo(2). u u(z.(t).(. This implies that the problem is not wellposed in the space of continuous functions with the norm defined above. 6307 In Probability Theory one encounters the OrnsteinUhlenbeck process. at .(z)T.t ) we denote the Fourier transform of u with respect to 2. for all f E C[O.
2 t ) . a2 1 ' 4 E D(nZ"). t ) = et3(e2'.1) . set T 1 = (I . . we get the solution to the problem Remark. we can find a convenient transformation of variables to the problem. and let h be a fixed point IR".t u o ( e . Therefore.t z ) and Fl(eli(e2tl)€2) = 1  d m e 2(e 2 .e . Solving the above Cauchy problem for the first order linear partial differential equation.+of ax. I +.. Consider the problem . u(<. Then the problem is transformed into the following 7 =0 : 21 = Uo(t). we can obtain . From the above solution formula. The above solution formula can be obtained easily from the Poisson's formula for the Cauchy problem of the heat equation. In fact. 2 t = etx.. 6308 Let L be the usual heat operator L = a let at (: .0) = G o ( t ) .504 Then the transformed equation and initial condition are in the following form qt.l)caiio(et<). It is not difficult to verify that F'(iio(ett)) = e .
e"n€)G = 0. i& + ([(I2 . the above problem we obtain { O(<. t 2 E En L which admids the following solution qt. Show then that the solution u satisfies Il"('. t ) to the problem This is just the conclusion of the problem.505 ( L u ) ( z .t > 0. (Indiana) Solution.t ) = &(()(y i h .h .O) = m. u(2.) b. Let G( t ) be the Fourier transform of u with respect to 2. < E En. Transforming (. for an unknown function u ( z . a. ( l€12exP( The inverse Fourier transform of G gives us the explicit representation for the solution u ( z .t)llL~(R")< e(1R2)tl141~L2(Rn). (You should prove that your u is welldefined. € ) ) t . t ) of the above problem in terms of the initial data.O)= 4(2). a. 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. but you need not prove that it represents a smooth solution. 2 E B" 2 E B". t )= u(2 .t ) .> 0. Derive an explicit representation for a solution u ( z . t ) . .
we get from above euualitv E(t) 2 dt I d where E ( t ) = & u 2 d x . (Iowa) Solution. + 1 IVu12dz =  1 u5dz. we get from (1) I d E(t) 2 dt 5 JRJ+Eqt). Using above estimation.we have 2 utudx  Au u d x =   I u5dx.0). Prove that we can find a constant C such that E(1) = I u 2 ( x . The Hijlder inequality gives i.506 6309 Let R c IR" be a bounded domain and assume u ( x .  AU = U 4 (heat conduction with heat loss due to radiation) with the boundary condition u Ian= 0. t ) 2 0 is a function with u E C2(o [O. By using the Green's formula and noting the boundary condition.e. . oo)).which solves the equation x U+..1 ) d x 5 C regardless of the initial value u ( z . Multiplying both sides of the equation by u and integrating on f .
tO) = (u3 sin(u))(zo.to) 5 0. if u( zo. (Indiun u) Solution.tO) 1 .to) sinu(zo.T) x with g. u= 9. Therefore. then there exists (zo. to) > 1. T ) x (0) W O . LO.to) E B(0. h continuous and g.507 The above inequality can be written as Integrating the above inequality from 0 t o t .T]. TI. r ) x [O.t o ) 2 1. If the conclusion of the problem is false.tO) 1. B ( O .T ) x (O. .u. t o ) and u(z0. TI.T]such that u takes the maximum a t ( z o . + + sin(u) = A. This completes the proof. h < 1. Prove that maxu < 1 on B ( 0 . + if u(zO. we must have But from the equation we get (ut . It is clear that ut(zo. B(0. 6310 Let satisfy ut u3 u = h.to) This contradicts (1).t. = < 0. u3(z0.A+u)(zo.sin 1 < 0.to) 0 2 and Az(zo. we have Ef(t) It follows that 2 E+(O) + 3(R(.T ) x (0.
(b) First we show that u can not take the positive maximum in R x (O. ( t ) + II~llLl(n)(t) Ilfll. t > 0: u(z. In fact. TI. we obtain from above integral equality Then we get the conclusion immediately. udx  u4dx.to) 5 0.T]. the following bound also holds: I I 4 4 ( n . Hence (ut . By applying Green's formula.T]) a solution to the problem n be { (b) Prove that ~t = AU .508 6311 Let u E C2(R x (0. (a) Multiplying the equation by u. (c) Prove that the solution t o this boundary value problem is unique. t o )E R x (O.yn) 5 + IlfllLl(n).Au)(xO. (d) Show that for f 6 L4(R) n H1(R).T]. u=o z E d R . IIUIIL(n)(t) L llfllL(n) for all t E (0. TI. (Indiana) Solution. then we must have ut(zo. where R is a smooth bounded domain in En. if u take the positive maximum a t ( z o .O) = f ( z ) 2 E a. (a) Prove that IIUllLz(n)(t)5 IlfllL.(n) for all t E (0. we get Au . 2 .u3 2 E R . T)) Co(nx [O.to) 0.and integrating the resulting equation on 0. t > 0.to) 2 0 and Au(zo.
(c) Let u1 and 212 be both solutions t o the boundary value problem. lVu12dx) = . vt E (0.u'dx. vt E ( 0 7 1 for above problem. we can show that u can not take the negative minimum in R x (O. Au . we get 1 utdx = J. we can obtain f and (a Lu'dx) + dt (1 J.J. By applying Green's formula to the first terms in the right side of the both equalities above.d u'dx <_ 0.tO) < 0. By a similar deduction as done in (a).utdx . 0) = 0. u3dx . w ( z . Therefore. 1 + u ~ )2 w . ~ x 5 (1 J. and integrating the resulting equations. IVul'dx) 5 0. and w = u1 .This contradicts the inequality u3(xO. we have dt ( / .u2.Au .Tl.T]. we have Il. The w solves the following problem: { Wt = AW.(u: + ~ w = 0. u 4 d x + J. . This implies that u1 = u2. x E R.u3utdx and utu3 = J.J. t > 0. x E $2. we can get Il4lL. Adding the both inequalities above.and the uniqueness of the solution is proved.u'dx) dt 4 5 J.l u i d x 5 0 d = . t > 0.llL(s2)(t> L llfllL(ct).1 3 ~ ~ I V u (.(ct)(t) 50. In a similar way. (d) Multiplying the equation by ut and u3 respectively. ~ x E a R .
Let R C B" be a bounded open set. furthermore. This is the result we need. Set w = u . A(p) be a real positive definite n x n matrix with let entries aij(p). combining with (a) if necessary. ~ ) E and wt 2 aij(Vw)vzizj for ( z . respectively. co). t ) E R x (0. co) (Indiana) Solution. 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. co). Suppose. implies the conclusion of (d). t )R x ( 0 .510 This differential inequality. . 6312 For each p E En. 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 ( ~ . we find w 5 o on I x [0.v and Then from the given differential inequalities. Assume A E C ' ( B n ) . that u and w are continuous on and satisfy the condition u 5 w on the parabolic boundary n x [O.
t .u)uxi o >_ i=l n for z E R. ~C~b i ( t . t > 0. . t ) Q 0x P. t ) = min u(z. Assume u i j ( z .{ o ) . let u E C2(nx [0.j=l n for every t E R~.In fact. Suppose u satisfies the inequality ut  C i. t )can not be the minimum point of w in R x (O. wxi = 0 and which contradicts (1).T]). p ) t i tLj x ( z . (1) then ( z .j=l i=l at ( q t )E 0 x ( O .to) R x (0. T and E I u(zo.t . (Indiana) Solution. vu)uZixj  C b i ( z . z .t . p ) and b i ( z . Now suppose that u takes the minimum a t (%'. t . p ) 1 t 1 2> o i. z . z ) are continuous functions of their arguments and assume that C ~ i j ( z . it must hold that Wt 5 0. z . where X is a positive function. t . i. u. m)). t ) . u .j=l n a i j ( z .TI where Q = (R x (0)) U (dR x [O. T I .T]. Q . t . z . First we prove that if w E C2(ax [O. ~ u ) w . Prove that for any T >0 min u(z. t . u ) w Z i > 0.t o )< min u. a t the minimum point (z.m)) satisfies wt n n C a i j ( z .511 6313 For R c Rnbounded and open.
if However. Then takes its minimum at some point in R x (O.512 Take a > 0 large enough so that a u l l ( z .u. Vu) + b l ( z . T ] . .&eaxl for ( z . w = u . This is in contradiction with w having minimum point in R x (0. t .u. for the function w we have E >O is sufficiently small. t . t )E s2 x [ o .T].V u ) > 0. TI.
= 9(z). Therefore. Then we get . Transforming the equation and the initial conditions.O) Show that the LP norm of . > n is bounded by a constant (Indiana) Solution. ?([) is bounded in E n . +.t)at time t for p times t lnlp..n arbitrary. Let G ( < . = $(tL 0) 1 ?(<I I cK + c21<)22 = 0. we have I and Then g E L1(IR") implies that 2tt K O ) = 0.513 SECTION 4 HYPERBOLIC EQUATIONS 6401 Assume that g is smooth and in L 1 ( R f l ) . u(2.c2AU = 0.t) = sin(cl<lt). it holds with a positive constant M that Set 7 = c<t in the integrand. t ) be the Fourier transform (with respect to z) of the solution to the ndimensional wave equation: Utt . %(t. G(<.^(.O)= 0.
we have . we get < %(€) qt. E ( t ) Eo. (Iowa) Solution. The above estimation gives the conclusion of the problem. (b) Let Eo denote the constant energy in (a).zn).e. Solving the above initial problem of the ordinary differential equation. . i. E (a) Use the Fouriertransform t o show that the total energy at time t is constant in time.[n). . (a) Taking the Fouriertransforms of the wave equation and of the initial conditions with respect to the variable 2 = (zl. q€. utt + l€12G(t.. = Gl(€). 6402 Consider the Cauchy problem for the wave equation Assume U O . Show that . = Go(€).t>= Go(€) cos(ltlt) + sin( I€lt) It I . U ~ S. 0) where = (€l.514 p > n ensures the convergence of the integral...t) = 0. 0) G(€.
+ %. Utt = 0..O)= qqz). = $(z). ... z E R3.. 6403 Consider the initial problem for the wave equation in three space variables: u.55 1 Integrating the above equation on En and noting that we obtain Then the Plancherel theorem gives us Combining this with (a)..+ u.t> 0.. 2 E B3.O) z E B3. we finish the proof of (b). Ut(2.. u(z. .
0.22. we can show that the above equality holds when (c) It is easy to verify that 2 13. it holds that 2 IyI 2 3 for all y E {y : I . 2 2 In fact. 2 k .and $(z) = { 0. (4 (b) By applying the domain of dependence for the solutions t o the wave equation. . (b) If q5 and II. 0. when 1 1 5 7.(Y) = 0. t ) for all t 2 0. y Therefore J q5(Y)dS = IyzI= 10 J II. find the set of points in lR3 where you are sure that u vanishes when t = 10. for 1 1 < 1. 2<t. vanish outside a ball of radius 3 centered a t the origin. no integrals). (IndianaPurdue) Solution.2 3 > 0). for 2 < 121. we can verify that u vanishes when t = 10 for 1 1 5 7 or 1 1 2 13.23) E lR3. (Your answer should be explicit. 11 2 lyzl=lO Similarly. 6404 Let R be the upperhalf space in lRn R = {(21. for 15 1x1 5 2. where k is a constant.21 = 10). 0. (c) If 4 vanishes everywhere in R3.516 (a) Write down the formula for the solution of the problem. find u ( 0 . O<t<l.
. x t E R for (IndianaPurdue) Solution. W ( X . when t < 23. When 2 3 = 0. (a) Write down a formula for the value u ( x . it is easy t o see that u(x. ) . ‘$(xl. ax. (2. x E R.~Y 3 W x ~ / +(Yl. ) 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.t 2 0. 0) = +(x). The value u(x. @(z)= x3). t ) . (b) Find a formula for the value u ( z . . ax: ax. at2 2 U(X.O) = d(X).t ) at (2.~ y l . satisfy certain compatibility conditions. X E R . E R. z2. J y . (a) When x E R and 0 < t < 2 3 .t ) of the solution at all t > 0. +(% 2 2 . = ~ t ~ . t ) of the solution at ( x . the domain of dependence for the value u ( x . Ys=~/ta(Ylzl)a(Yzza)a Y3)dS = 0. E aa.) is given by t If both 4 and II. ~Y2. 2 0. 2 2 .23)7 23 23 23 23 23 in E3 2 0. 4(x1. respectively.( y+ l( Y l . ) given by the above formula satisfies the wave equation and the t initial conditions. t > O . 231.Y2. t ) is in R. t ) = 0.x * ~ ~ .m. 2 3 ) . < 0.= 0.517 Consider the initialboundary value problem azu azu azu a z U + + . Then the value u(x.. < 0.
u(z." IVuj2)dSdp. (iv) Let 7~ = 2. evaluate lim E ~ ( u ( t ) ) . Define the local energy by (i) Use the energy indentity to show that Conclude that the total energy E W ( u ( t ) is conserved. # 0 and suppose that $ ( ~ 1 ~ 2 2 Show that u satisfies ) . ) (ii) When n = 3. when 23 = 0.t) satisfies the boundary condition on = 0.4k. (i) For t E [O.) (Indiana) Solution.T].t) be a solution of class C2 to the Cauchy problem with compact support. + . 0 for 11 2 2 k. have we ERT+t(u(t)) = (u. (Here T = 1. on the set T < t . where 2: = for some constant c. 23 Therefore.518 In a similar way. t+O" (iii) Do (ii) above when n = 1. we have J Jyxl=t @(y)ds = 0. 6405 Let u(2.
(ii) From the formula of the solution to the problem it is easy to verify E ~ ( u ( t =)0 if t > R ) + a.(u(O)). .519 By the above expression.(u(t)) equals to E.T] It is clear that the total energy E. 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.
is contained.ll. (iv) When n = 2.J Therefore. t .IzI .4k and IyI I k . it is easy to see that Hence the formula of the solution can be rewritten as When 1x1 < t .1x1 + 2 k < d3 V t .4k.1x1 > 4k.Izl2k  . we obtain t .2 k ) ' I It is easy to verify that . Hence (iii) By the formula it is easy to evalute too lim ER(u(t))= 0.520 where a is the radius of a ball in which supp4 U supp. we have IY  4 I 1 + 2k. 4 1 and hence 1 J W J(t + IzI + 2 k ) ( t . the solution is given by If 1x1 < t .
oo)). . we can conclude that Thus we obtain a differential inequality for E ( t ) d E(t) dt I CE(t).521 6406 domain and assume u ( z .which solves the equation Let R c B" be a bounded with the boundary condition = 0. Then E ( t ) I exp(Ct)E(O) is an immediate consequence of the above differential inequality. Multiplying the equation by ut and integrating the resulting equation with respect to x in 0. have we Here we have applied the Green's formula: By the Poincar6 inequality. (Iowa) Solution. With prove that there is a C < 00 such that for all t 2 0. t ) is a function with u E C2(ax [0.
E ( t ) = E(O).t > 0. t = 0. Then w solves the following problem: aZu 2au r ar { w. r 2 E B3. fo. + (ut” IVul2)dz. the equation can be written as + . In the symmetric case. (IndianaPurdue) Solution..O) = u. r = 0. m. w = f ( r ) . 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.O) = 2 E B3. vt > 0. Find the spherically symmetric solution to the problem. = wtt.t > 0.(z. (b) What is its energy at time t = 10. . Solving this problem. ar2 Set w = T U .522 6407 (a) Find the explicit form of the solution of the initial value problem v2u= U t t . 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. u(z. w = 0. T > 0.= U t t . wt = f’(r). z E r B3. where T’ = z:+zg+zi.
O) = g(z). From the Poisson’s formula. we have It is easy to verify that The area of the intersection {y E R3.Au = 0 for x E B 3 . Prove that there exists a constant the given data such that C depending only on (Indiana) Solution. . Let R and M be positive constants such that and Ig(z)l 5 M . u(z. where g E Cr(R3). Ut(2. .X I = t } n BR Iy 5 The area of aBR.O)= 0 .523 And therefore = 2F 7 6408 Consider the wave equation in R3: utt . v x E R3. t > 0.
t = T } 1 2 1 for 0 5 r 5 to.t > 0.T ) ~ z . Multiplying the wave equation by ut. It is not difficult to verify that Therefore. we have .O) : 1 1 5 t o } in the plane t = 0.5 t o ..524 Therefore.”+ IVuI2)(x. MR2 47rt t ’ 6409 Let u E C2(.t) E En+1. we can obtain Set 1 E(7)= 5 lr(u.)be a solution of the equation utt Au = 0.t } .t) : 0 5 t 5 t o and 11 5 t o . 4 r R 2 = . Let = {(Z.t . and integrating the resulting equation with respect to x in R. M . x Prove that u vanishes in the conical region D = {(x. we obtain 1 lu(z. x (Indiana) Solution..t)l 5 .Rn+l(x. Suppose t) also that u = ut = 0 on the ball B = {(z.
u=O 6410 Let U(Z.525 It is clear that E ( 0 ) = 0. Let R > 0 such that It is not difficult to verify that for any t > R.5 to. gdx = 0. 0) = f(z). we have uttdx = J. > 0 ) . Therefore. g E Cr(nZ3). By using Green's formula. t ) be a solution of the Cauchy problem utt = AU (Z E B3. Hence E ( 7 ) = 0... /ltnL8u 2 dx = 0. Assume that f . then I.O) u(z. . t Ut(Z. we get the following equality from the equation J. in D. Prove that if = dx). (Indiana) Solution.. Audx = 0. This implies that vo 5 7.
. by Cauchy inequality it follows that 21ax J R S 5 (J. ut(x.m a x l u l ? CI for all sufficiently large t. we can get . = +(x) E Cp.. In the other side. (Indiana) Solution. This contradicts (1) provided that J.iax)t (/ t R< ( C<t+R 3l ax) l I 2 for any t > R. = 0.t > 0) with data u(x. 0) 0) a) Use Fourier transform to show that there is a positive constant c such that u2 dx = c. . a) Let 6 denote the Fourier transform of u with respect to the variable x = (XI. . 6411 Consider the solution of the wave equation Utt = nu (x E R3... zn).526 and by the initial conditions.gdx # 0. JlzIRs X b) Show that there is a positive constant C such that 1 t .By taking the Fourier transforms of the wave equation and the initial conditions.
. The proof of (b) is completed. ~ 2 This gives Iu(2. t ) with respect to the variable 2 = (21. 400. If the conclusion is false.21. b) Let Q ( t )be the support of u ( z . t . Then we have This contradicts (a). .+ c c t+b and t.’. a s n .>l+.527 Therefore we have Then we get where if $0. then there exists a sequence { i n }such that t. )< la(t.tn)l I IQ(t.z3) for t > 0.)l+t.m a x ~ u ( z .
when t > A. Determine the large time J b) Let n = 2. 2 2 .t)dz.528 6412 Consider the Cauchy problem for the wave equation a) Let n = 3. then u ( 2 . t )=  . vlzl< et. u4(2.t ) d z = 0. which implies that t ) = 0. q5.t)l = O(t') Ixl9t as t + o) c. l4<$t 4 3 0. a) Assume that R > 0 is so large that suppq5 u supp11. Let 8 behavior of the integral < 1.Show that for any 1 < 1. lxl<@t when t > __ 18' R b) From the Poisson's formula. J u4(2. Therefore. we conclude that if t > R and 11 2 < t . it holds that +. 11. E CF(R3). (Indiana) Solution.11. the solution to the 2dimensional problem is given by u(21. we have 9 sup Iu(z. t ) = 0. (21< R}. From the Huygens's principle for the %dimensional wave eauation.R. E C F ( R 2 ) . c ( 2 f IR3.
if f > 0 for all z.e. then the same is true for f at all later times . Consider the initialvalue problem for this. b) Define the local density p by p(. clear that if it is IzJ5 B t and Iyl 5 R then JY . V x f = 0.zI2 2 t Therefore. Show that supp = o(t3) as t X (Indiana) . E IR3. let f ( c . j~~ 1 .v) 5 h(z).t)dv+ oo. f(Z.( B t + R)2.. 6413 The linear transport equation is following PDE for a scalar function f of seven variables z. 2 When t > 5. t E R. 1 1 < R } . 0 j given. v. v.where h E L1(B3).t) = J. we have dt2. U ) =j v).(y.. (c. Suppose that 0 5f 0 (z. The proof is completed. where C is a constant.529 Let R > 0 be so large that suppllr C { z E IR2. a) Show that. i.'u.): v ft + v . v.X I I et +R and d2 . t (z E B3..
+ tv. v.tv. a) The characteristic curves of the equation are defined by dx = v dt and along which it holds that dt Therefore. we get the solution to the initialvalue problem 0 f (2. + yuu. (h diana) Solution. (2) From (1) and (2). 6414 Solve the Cauchy problem xu. the forward characteristic curve departing from t = 0. we have 0 = JR3 f ( x . = xy ( x > 0) u=5 o n x y = 1.v)dv O This completes the proof. The system of the characteristic equations for the problem is given by .tv.= 0. if a solution exists.t) It is clear that f 2 0 i f f ? 0. x = a is x = (Y df . b) From (3).Solution.v). =f ( x .
= xy ds du s=O:x=a. The initial data gives C = diately u = 1 43cxy. Set v = u . we get vt + 22121. with the initial conditions _dx dY z= y u .531 ds . y=. W(2.~ e 'ds = _ ds (ae'y ds 1 u2 2 + u). = 0. + 6415 Compute an explicit solution u(x.O) = z2. F. t ) to the initialvalue problem Ut + = 0. . The characteristic curve of the quasilinear equation (1) is given by dx ..O) = 22. we have immediately and 21 du du d = . (Indiana) Solution. 1 u=5. respect to 2. U(X. where C is a constant. Differentiating the equation and the initial condition with . (Y Fkom the above Cauchy problem.= 2v. dt . Then we obtain imme This implies that + u + x y = c.
. 1p. i. t = o : x = p .= 0. 1 . u(z. 05z< 1 .0 . dt where u o ( z )= u(x. giving explicitly the discontinuity curve and the RankineHugoniot jump condition along it. The characteristic curve departing from t = 0. P < 0.t 2 > 0.532 along which it holds that dv . < 0. and find the discontinuous entropy solution.O). 12. + u = uO(P). 2>1. ( IndianaPurdue) Solution. we can obtain immediately v=Hence u= 4t 1' 22 22 4t 1' + + 6416 Solve the following initial value problem.e. dt From the above characteristic equations. u t + uus = 0. du _ . 21 = { P> 1. 0. we have 2 = uO(P)t p. t = 0 : u = u"(p). and P < 0. From (1) and (2). o<p<1. p > 1.O)= Show that the continuous solution exists only for a finite time. . z = P is given by dx =u. dt along which u is a constant. __ {:: 2 E IR.
Fig. u.6.2 . For t 2 1. dx 1 .= (u+ dt 2 Noting u+ = 0. we have p=. the RankineHugoniot condition along the discontinuous curve is given by dt i. 1. we have + u)..2).e.2 When t = 1. 1  dx _Therefore.6. 0.= 1.t 1t 2t . 0 5 5 1.For 0 < t < 1.1x 1t 1t' Therefore we get the continuous solution to the problem u = 1..2' 1 x=(t+1). and hence x < t. 2 > 1. ' t<X<l. the continuous solution to the problem blows up. 2 t l l . the discontinuous curve is dt . 11 for 0 5 t < 1 (see Fig.
e bt v) dX = 0. for 0 < t < T.u)bu=O dX at at4 aU forO<t<T. t ) 5 pebt for all x and all t E (0.534 and the entropy solution to the problem is given by 2 0. aebt 5 u ( z . 3 > z ( t + 1).t)be smooth solution of dU at 8 + a(u)U = 0 dX = UO(X). Then v satisfies the following equation dV at + U(X.Y). .O) where uo and a are smooth. t ) be a smooth solution of +a(z. ii) Show that. 2 < ' ( t + l). Show that. T ) . Show that if a then 5 u ( z .0) 5 p for all X. t ) = U o ( Y . 6417 a) Let u ( z . where b is a constant and a is smooth. then u z ( 4 t . y ) is the characteristic through (Y. if 0 5 t l < t2 < T . 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. a) Let u = ebtv.O). b) Let u(X. if z ( t . U(X. then 00 00 (Indiana) Solution.
we have This is just that we want to show. we can get the Y following conclusions easily. ~ ) .535 The solution v maintains a constant along a characteristic of the above equation. a Then it follows from above problem that Therefore. FZ(0. i) Let m = inf a‘(uo(y))ub(y). If m >_ 0. defined by the equation . we get g (g+. From the above expression. y). Y ) = Y.Y)) { = a’(uo(y))ub(y). . I?”). Hence Differertiating above equality with respect t o y. { 4 0 . a Differentiating this problem with respect to y.O) 5 v(z. we have The characteristic x(t.t)5 supv(z. X 2 dx dt This gives the estimation in the problem immediately. 0) is given by the following Cauchy problem: &. it holds that infw(x. Y) = a(uo(y)).O). Therefore. y) through (y.Y) = 1. t ) is a constant along the characteristic z = x(t.= a(x. the Cauchy problem admits a global smooth solution in ( 0 . b) u(x.
'1 = 0. If uo(z) E 0. the largest T for which the problem admits a smooth solution in (0. we obtain 00 00 This completes the proof of the problem.. (Indiana) Solution. = f ( t .(z(t.T) = 0. . prove that the only LP solution of (2') is u G 0. t ) and &z(t.536 If co < m < 0. u(z. on x [0. we have 00 Taking the expressions of u. by the change of variables z = z(t. . IR x {T}. on on IR x [O. t )E P ( I R x [O. A function u E P ( I R x [O. Consider the following Cauchy problem: 4t + 4.7'].  44YI s))dyds (2') + J_.z).y) in the integral. T).uO(Y)4(Y? O P Y = 0 for all test functions 4 ( z . ii) For any fixed t E (0.. on 4(. y) into account. 0418 (Uniqueness of weak solutions) Consider the Cauchy problem 'Zlt + 1. T) is given by mT+ 1 = 0.O) = uo(z).)(4t(Y.T]) defined to be a weak solution of (2) iff is lT : / 00 4 Y .T].T]) which are compactly supported and vanish on {t = 7) '. y). IR x (0).
. .T].TI).)) each t > 0 (where T V = “total variation”). dt a) Suppose that b 0. meets the line t = 0 at ( < i . where x E t B and t > 0.a * .. n .z . . .T]). .. 6419 Consider the Cauchy problem for a function u = u(x.~ .l).T) .  . 5 TV(u0) t)) 5))ds. Hence we have u 0.’ TV(b(*.T]) with a compact support. l .. ) )5 TV(uo(.. Then u is a constant along a characteristic curve. ) Then we have xi. We divide the line t = T into subintervals by the points * * . . we have 6’J_+_m U(Y.. .t).T]) which is compactly supported. + a(z.. O .t). Therefore. (Indiana) Solution.is dense in L * ( B x [O.537 where f E C m ( Bx [0. the equality (*) is valid for all functions f E L q ( B x [0. 0) = uo(x) E C(B2). one still has the bound T V ( u ( * .5)dYd5 =0 (*I for all function f E C m ( Bx [O.T]) with a compact support. O ) (i = n.. 21..^) = uo(ti). b) Show that when b $ 0. The characterstic curves of the equation are given by the equation dx .. consisting of all Cmfunctions with compact support in E x [O..* . dt along which the solutions to the equation satisfy du . where q = p / ( p . t ) . ut u ( z . zl. l . = b(x. + J. It is not difficult to show that the above Cauchy problem admits a solution ~5 E Cm(Ex [O. ) . Suppose that the backward characteristic curve through (xi. then TV(u(.t)u. .. provided uo(x) 0.. t for a) Show that if b 0.TI). . zo.= u(2. * ..x . As the space Cr(B2 x [0.)f(Y.= b(x.
where by z = Z i ( t ) we denote the characteristic curve through ( z i . we have u ( z i .538 Therefore.. we get . T ) = uo(ti) + Ir 0 b(zi(s). VT > 0. b) Let { z i } and {ti} be given as in a). we get TV(U(.T)) 5 TV(UO(')). the above inequalities. When b $ 0. ~ ) From . s)ds.
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. New York University Harvard Illinois Indiana IndianaPurdue Iowa Minnesota Rutgers Stanford SUNY. Irvine . University of Cincinnati Columbia University Courant Institute of Mathematical Sciences.539 Abbreviations of Universities in This Book Cincinnati Columbia Courant Inst.