Scribd Logo
Cargar

¡Te damos la bienvenida a Scribd!

  • Rudin CH 7 PDF

    Cargado por

    Luis Fernando Honorato
    17 vistas21 páginas

    Título original

    rudin ch 7.pdf

    Derechos de autor

    © © All Rights Reserved

    Formatos disponibles

    PDF o lea en línea desde Scribd

    ¿Le pareció útil este documento?

    ¿Este contenido es inapropiado?

    Denunciar este documento

    Copyright:

    © All Rights Reserved
    Descargue como PDF o lea en línea desde Scribd
    Marcar por contenido inapropiado
    17 vistas21 páginas

    Rudin CH 7 PDF

    Cargado por

    Luis Fernando Honorato

    Copyright:

    © All Rights Reserved
    Descargue como PDF o lea en línea desde Scribd
    Marcar por contenido inapropiado
    de 21
    2A {976 supp. WATH Solutions Manual to Walter Rudin’s Principles of Mathematical Analysis Roger Cooke, University of Vermont Chapter 7 Sequences and Series of Functions Exercise 7.1 Prove that every uniformly convergent sequence of bounded func- tions is uniformly bounded. Solution. Let {fn(x)}%2.4 be a uniformly convergent sequence of bounded func- tions, say |fa(z)| < Mn for all z and all n. Since the sequence converges uniformly, it is a uniformly Cauchy sequence. Hence there exists NV such that \fre(2)— fan(a)| <1 for all m,n > N. In particular if m > N, we have |fn(z)| < Viv (2)|+lfm(®) —fnr(2)| S$ My +1, and therefore if M = 1+max(Mi,..., My) we have |fn(x)| 0. There exist Ny and Np such that |fn(z) ~ f(2)| < § for all « ifn > Ny and Ign (2) ~ g(2)| < § for all x ifn > No. Let N = max(N;,.No). Then for n > N we have, for all 2, (fa + 9n)(2) = (F +.9)(2)I < [fn @) - F(2)| + lon (2) - 9(2)| 0, choose Ny and No such that |fn(z)—f(z)| < fg for all x and all n > Ny and |gn(z) — g(z)| < 36 for all z and n > Np. 109 110 CHAPTER 7. SEQUENCES AND SERIES OF FUNCTIONS Again let N = max(1N;, No). We then have, for all z and all n > N, \fu(z)9n(2) — S(z)9(z)| < |fale)gn(z) — fa(z)g(2)| + +1falz)9(e) - f(2)9(x)] S$ Mlgn(#) - 9(z)| + Mifa(x) - F(2)| é € < Mag t+May ze Exercise 7.3 Construct sequences {fn}, {gn} which converge uniformly on some set B, but such that { fngn} does not converge uniformly on E (of course, {Fngn} must converge on E). Solution. Let f(z) = z for all 2 and all n, and let gn(z) = 2 for all x and all n. Then f(z) converges uniformly to z, and gn (2) converges uniformly to 0. Therefore fn()gn (2) converges to 0, but not uniformly. In fact for every n there is an , namely ¢ =n, such that fn(z)ga(z) = 1. Hence, no matter how large n is taken, the inequality |f,(x)gn(z)| <1 will never hold for all Exercise 7.4 Consider For what values of x does the series converge absolutely? On what intervals does it converge uniformly? On what intervals does it fail to converge uniformly? Is f continuous wherever the series converges? Is f bounded? Solution. The series converges for all x except 0 and z = =}, n = 1,2,.... For x = 0 all the terms of the series are defined, but the terms do not tend to 0. For z = 5} the nth term is not defined. For all other values of the series converges. By Theorem 7.10 (the Weierstrass M-test) the series converges uniformly on the interval (6,00) if § > 0, since on that interval Lente ~ 126 Likewise, the series converges uniformly on (-c0, —6] except at the points 2 = —3, since for n > \/? we have 1 2 lee on? ‘The series does not converge uniformly on any interval having 0 as an end- point. This is easy to see in the case when 0 is the left-hand endpoint. For each a of the terms of the series is a bounded function on (0,00). If the series converged uniformly, the limit would be bounded by Problem 1 above, But we have AB)2¥ neil at n 2 Likewise the series cannot be a uniformly Cauchy series (i.e., the sequence of partial sums cannot be a uniformly Cauchy sequence) on any interval (~6,0), since, no matter how large n is taken, there is a point « in this interval, namely 2 = phy, at which the nth term has the value 2. Hence, if Sy denotes the sum of the first n terms, then |S, (z) ~ Sn-x(2)| = ‘The uniform convergence shows that the limiting function f(x) is continu ous wherever it is defined on (—00,6] U [6,+00). Since é is arbitrary, f(z) is continuous wherever it is defined. The argument given above shows that f(r) is not bounded. Exercise 7.5 Let Show that {fn} converges to a continuous function, but not uniformly. Use the series 5> fy to show that absolute convergence, even for all x does not imply uniform convergence. Solution. The limit of fa(z) is zero. If x < 0 or e > 1, then fn(z) = 0 for all n, and so this assertion is obvious, If 0 <2 < 1, then f(z) = 0 for all n > 3, and so once again the assertion is obvious. The convergence is not uniform, since, no matter how large n is taken, there is a point z, namely z= sr for which f(x) = 1. ‘The series fa(z) converges to 0 for z < 0 and x > 1, and to sin? 2 for 0 <2 <1. Since the terms are nonnegative, the series obviously converges absolutely. Since the sum is not continuous at 0, the series does not converge uniformly on any interval containing 0. Exercise 7.6 Prove that the series = att tn yo we converges uniformly in every bounded interval, but does not converge absolutely for any value of x. 12 CHAPTER 7. SEQUENCES AND SERIES OF FUNCTIONS Solution. The series is the sum of two series The first of these converges both uniformly and absolutely on any bounded interval (a, ] by the M-test (with My = 4, where M = max((al,|b/)). The setond is independent of x and converges, hence it converges uniformly in 2. By Exercise 2 above, the sum of the two series converges uniformly. The series does not converge absolutely since the absolute value of each term is at least 2 for any z. Exercise 7.7 For n = 1,2,3,... 1+nz2" eC ————L the equation F'(2) = Jim. f(a) is correct if x #0, but false if x = 0, Solution. The Schwara inequality, which implies that Vial) < ala = abe for z #0, shows that f,(z) tends uniformly to 0. Now fi (2) = a which tends to 0 if «#0, though f%(0) = 1 for all n. Exercise 7.8 If 0 (<0), T(z) 1 (e>0), if {tn} is a sequence of distinct points of (a,b), and if X len| converges, prove that the series fle) = Sento nat (asx 0. Choose N; so large that |fm(x)— f(z) < § for all m > M, ‘Then, since / is continuous at x, choose § > 0 so small that |f(y) — f(2)| < § if |y —2| < 6. Finally, choose No so large that |r, —2| <6 ifn > Nz. Then if n> max(Nj,.N2) we have Jn(@n) — (2) $ |fa(@n) — f(@n)| + |f@n) - £(2)| n+1. Thus fa(2) tends to zero, since f(x) = Oifn > |x, but f_(2) does not converge uniformly, since f,(n-+ 4) = 1 Then for any convergent sequence, say tn + 2, let N > max([x|, [21], [2al,-.-» |nls--.). We then have fn (tn) = 0 for alln > N, and 80 f(a) > f(x) This condition does guarantee uniform convergence on any compact set, however. For if {fn(z)} is not a uniformly Cauchy sequence, then for some £o > 0 there is a sequence of integers ny < my < +++ and @ sequence of points 21,22,... such that [Fnon—1(te) ~ Foe (te)| 2 €0 for k = 1,2,.... Since K is compact, some subsequence of {2,} converges, say t, > © as — 00, Now define yn = x for alln # mo,, n # No.1, and let Ynay, 1 = Unox = Tk-» 80 that so that y, + z. Then the sequence {en} = (fn(Un)} is not a Cauchy sequence, since |2nsa, — Znae,—al = Exercise 7-10 Let (x) x 0, and this will show that fq(z) is discontinu- ao Be : Since all the terms of the series for fy are nonnegative, it suffices to show that the limit of the first term is positive, To that end, let § = 34. If “g ql a -6 - from which it follows ous at r= 1 that f4(=) 2 g95- Therefore the lower left-hand limit of f(z) at # = Bis at 4 q 2 Since, by the M-test with M, = an this series converges uniformly and each of its terms is Riemann-integrable, it follows from Theorem 7.16 that the sum of the series is Riemann-integrable. Exercise 7.11 Suppose {fa}; {gn} are defined on E and (a) fp has uniformly bounded partial sums; (8) 9, = 0 uniformly on B; (c) g(x) = g2(x) 2 ga(a) 2 --- for every 2 € E. Prove that > faga converges uniformly on E. Hint: Compare with Theorem 3.42. Solution. Following the hint, we let Sv(z) n 2X, fale)an(z) and F(z) YX, fale) (Fo(2) = 0), so that |Fy(z)| < B for all z. Then if N > M, we have N |Sw(0)-Su(e)| = | 3) (Fale) - Fo-a(e)on(a)| adits = |Fv(e)ow(2) — Facln)anesa(@) + + > Falelon() — gn4a(0)| nants yea < B{lov(2)] +lameale) + Do ln) ~ gnarl} = Bilgw(#)| + lgnesa(@)| + gaeza() - gv(2)], and this last expression can be made uniformly small by choosing M sufficiently large by hypothesis (6). Hypothesis (c) was used in moving the stmmation sign outside the absolute value, Exercise 7.12 Suppose g and fy (n = 1,2,3,...) are defined on (0,00), are Riemann-integrable on [t,T] whenever 0 < t < T < 00, |fnl $9, fn > f uniformly on every compact subset of (0,00), and fF oeyae <0. Prove that. o lim, i Sol) de i fla)de. ene 0 (See Exercises 7 and 8 of Chap. 6 for the relevant definitions.) This is a rather weak form of Lebesgue’s dominated convergence theorem (Theorem 11.32). Even in the context of the Riemann integral, uniform con- vergence can be replaced by pointwise convergence if it is assumed that f € R. (See the articles by F. Cunningham in Math. Mag., vol. 40, 1967, pp. 179-186, and by H. Kestelman in Amer. Math. Monthly, vol. 77, 1970, pp. 182-187.) Solution. We shall prove that f Jn(z) de converges for each n, that the limit 5 lim | fa(x) dx exists, that quantities are equal. J (x) de converges and that these last two 116 CHAPTER 7. SEQUENCES AND SERIES OF FUNCTIONS Since we obviously have |/(2)| < g(z) also, it follows that for any interval {r3] C (0,00) we have | [ tateyan < ff saya [sere s [ote)ae, | sate stedael <2 [" otedac. Now let ¢ > 0. Choose a and b with 0 e > b we have , . [ aayar = [olayac~ f° oteyae ve c3 3 = fF serae— fi oa) b ' 4 [ ol) dx -f o(2) de ° + ee Then for any d > e > b > r and any n we certainly have | [tscae— f° satael =| f° score ‘i Thus by the Cauchy criterion Jim / j,(2)de exists. A similar argument A s [ oear 0 we can choose c and d so that the first and’last terms on the right are less than § (for all n, in the case of the first term). Then, since fa(z) + f(z) uniformly on ¢,d, we can choose no so large that the middle term is less than § if n > no. Exercise 7.13 Assume that {fn} is a sequence of monotonically increasing functions on R! with 0 < f,(2) <1 for all x and all n. (a) Prove that there is a function f and a sequence {ng} such that F(e) = jim fC) for every x € R!. (The existence of such a pointwise convergent subsequence is usually called Helly’s selection theorem.) (0) If, moreover f is continuous, prove that f,, — f uniformly on Rt Hint: (i) Some subsequence {fn,} converges at all rational points r, say to J (r). (ii) Define f (2) for any x € R! to be sup f(r), the sup being taken over all 1 <2. (ii) Show that fn,(2) > f(z) at every x at which f is continuous. (This where monotonicity is strongly used.) (iv) A subsequence of {f,,} converges at every point of discontinuity of f since there are at most countably many such points. This proves (a). To prove (), modify your proof of (iii) appropriately. Solution. (a) Following the hint, we enumerate the rational numbers (or any countable dense set) as {rn} and use the well-known diagonal procedure to get first a subsequence that converges at r1, then a further subsequence that converges at rz, ete. The sequence formed by taking the nth term of the nth subsequence is itself a subsequence and converges at each rn. (Note that we have not used the fact that 0 < f,(«) < 1 for all x and n, only the much weaker fact that for each x there is an M(x) such that |f,(x)| < M(x) for all z and n.) Let the function f(z) be defined as f(rx) = lima, fa,(re) and (x) = sup{f(re) : re < 7} for all other 2. The second definition could be taken as the general one if we wished, since it is consistent with the definition already given at the points x = T.. Since each of the functions is nondecreasing, it is clear that the function f(z) is nondecreasing. By its definition it is continuous from the left. Suppose f(z) is continuous at ro. Let <> 0 be given. Choose rational numbers r and s with ip, t= or t=. We then have $(0) — § SS) — F< Sault) S fas(B0) S fas(8) < SU) + $$ H(00) + 5- Hence |f(z0) — fn,(t0)| < € if i > tp, which proves the convergence at points of continuity. One more application of the diagonal procedure now allows us to assure that some subsequence converges at every point (since the set of of discontinuities of a nondecreasing function is countable). We can then modify the definition of f(x) at these points. 8 CHAPTER 7. SEQUENCES AND SERIES OF FUNCTIONS The claim that the convergence is uniform if the limit is continuous is not true. Let: Te aay HF each a, vet Inlu) ~ Fly) > y > n, so that the convergence is not uniform. Here the functions f(x) are not continuous, but they could easily be made so without violating the conditions of the problem. To get uniform convergence we must assume in addition that f(x) — 1 as 2 — co and f(z) + 0 as x + oo. Let us grant these relations and assume that f(z) is continuous at all points x. To simplify the notation we shall write fe instead of fay. Given © > 0, choose an interval [2,0] such that f(z) < if It is clear that lim fx(2) = 2 1- 5 if >. Then, since f(z) is uniformly continuous on {a,8], let a = to < ty <+-+ < tq =b be such that f(ti) — f(ti-1) < $. Choose kk so large that |fi(ts) — f(ts)| < § for all ¢=1,...,n and all 1 > k.Then for all y>b=t, we have 13 fy) 2 fits) >1- = and 12s) >1-$>1-4. Hence certainly \fily) - Ff) <1 (1- F) k and all y > b. A similar argument shows that f; converges uniformly to f on (~00,a]. The argument that fi converges uniformly to f on [t;~1, ts] is identical to that given above. Exercise 7.14 Let f be a continuous real function on R? with the following properties: 0 < f(t) < 1, f(t+2) = f(t) for every t, and 0 (O ==et>. ‘This fractional part ties in fe 2,1] if a, = 1, while if a, = 0 it is at least 0 and at most 2. Thus it lies in the interval (0, 4] if a, =0. In either case f(3*to) = ax, as claimed. We therefore have as) = S02 tea =) ya) = oP ome = n=l nat as asserted, Exercise 7.15 Suppose f is a real continuous function on R?, f(t) = f(nt) for n = 1,2,3,..., and {fn} is equicontinuous on [0,1]. What conclusion can you draw about /? Solution. The function f(t) must be constant on 0,00). For if f(x) # f(y) and O0<2 0, it follows that |fn(2)—fn(2)| =e for all n, Since #=¥ — 0, it follows that the family {fn} cannot be equicontinuous on (0,1), or, indeed, on any neighborhood of 0. Exercise 7-16 Suppose {/,} is an equicontinuous sequence of functions on a compact set K’, and { fa} converges pointwise on K. Prove that {fn} converges uniformly on K. Solution. Let ¢ > 0. Choose > 0 such that |fa(z) — fa(y)| < § for all nm if 2,y € K and |x — yl <4, Choose a finite number of points 21,...,2,7 such that for every 2 © K there exists j with |x ~ 25| <6. (Such a finite set exists; otherwise we could inductively select a sequence {zn} such that [tm —tn| > 6 120 CHAPTER 7. SEQUENCES AND SERIES OF FUNCTIONS for all n, and this sequence would have no Cauchy subsequence, contradicting the compactness of K.) Then choose no so large that |fm(j) — falz3)| < § for all m,n > no and all j = 1,2,...,.V. Then for any point ¢ € K, fix j so that lz ~2j| <6. If m,n > mo we have fons) — fn(®)| < |For a) — fn ay) + [fn (@s) — fn(@3)| + [fa(23) ~ fa(e)| ‘The first and last terms are smaller than § because |r—z,| < 6; the middle term is smaller than £ since m,n > mo. Thus the sequence is a uniformly Cauchy sequence. Exercise 7.17 Define the notions of uniform convergence and equicontinuity for mappings into any metric space. Show that Theorems 7.9 and 7.12 are valid for mappings into any metric space, that Theorems 7.8 and 7.11 are valid for mappings into any complete metric space, and that Theorems 7.10, 7.16, 7.17, 7.24, and 7.25 hold for vector-valued functions, that is, for mappings into any B Solution. Let X and Y be any metric spaces. The sequence {fn}, where fa : X = Y, converges uniformly to f : X — Y if for every ¢ > 0 there exists N such that dy(f,(2), f(2)) < ¢ for all x € X and alln > N. A family of functions is equicontinuous if for every ¢ > 0 there exists § > 0 such that dy(f(21), f(z2)) 0 there exists N such that dy(fm(:2), fn(t)) < ¢ for all myn > N and all x. Then in particular, for each x € X, the sequence {f,(z)} is a Cauchy sequence in Y. Since Y is complete, this sequence converges to a value that we shall call f(z). We now claim that {f,} converges uniformly to f. Indeed, given e > 0 choose N so that dy(fm(),fa(t)) < § if m,n > N. Since a metric is a continuous function, it follows that dy(f(x), fa(z)) < § <¢ ifm > N, that is {fn} converges uniformly to f. This is Theorem 7.8. Suppose now {f,} converges uniformly to f, Y is complete, ro € X, and dim fale) = An for n =.1,2,.... Then {An} converges, and lim f(z) = jim An. (This is Theorem 7.11.) The proof is as follows. Given € > 0 choose N so that dy(f(z); fa(z)) < § for all x ifn > Let n > N be fixed. Choose 6 > 0 (depending on n and ¢ in general) such that dy (fn(z),An) < § if 0 N, dy (Am, An) S dy (Am; fm(2)) + dy (f(z); f(z) + dy (Fn(2); f(2)) | | 121 ‘The middle term is less than § for all m,n > N and all 2 € X. If m and nate then fixed integers larger than NV, the first and last terms can be made smaller than £ by choosing sufficiently close to zo. Hence we have dy-(Am; An) < € if m,n > N. Since Y is complete, the sequence {An} converges, say to A. Now observe that dy (f(z), A) $ dy(f(2), fal2)) + dy (fa), An) + dy (An; A). If N is chosen sufficiently large, the first and last terms on the right-hand side will be less than § (for all z, in the case of the first term). For a fixed n satisfying these conditions, if 6 > 0 is sufficiently small, the second term will be less than § whenever 0 < dx(z,20) <6, and hence dy (f(z), A) 0 choose 6 > 0 such that |[fn(z) — f,(y)|| < ¢ whenever d(z,y) <6. Then for each component fi of f, we have |fi(z)] < |lfnl] $M and filz)—Fily)| $ fn(z)—fn(u) || < € whenever d(x,y) <6. Hence each sequence of components {fi}n, i= 1,...,h, is bounded and equicontinuous. Therefore for each i there is a subsequence {n-} such that fi, converges uniformly. By refining to subsubsequences, we can obtain a single subsequence {n-} such that {f1,} converges uniformly for all i, say to f*(z). Then, given ¢ > 0, choose ro so large that |fi (x) ~ fi(a)| < £ for i= 1,2,...,k and r > ro. It then follows that |lfp, («) — £(2)|| <€ ifr > ro. The proofs of the other results all follow this model argument. Exercise 7.18 Let {fn} be a uniformly bounded sequence of functions which are Riemann integrable on (a, 0], and put Ae= [ma (axes) Prove that there exists a subsequence {Fn,} which converges uniformly on (a, 6]. Solution. Let M be such that |fn(z)| < M for all n and x. Then clearly |F,(z)| $ M(b =a) for all'n, so that {F,} is uniformly bounded. Also, given £>0, let 6= %. Then if x 0 such that for all § > 0 there exist z,y and g € $ such that d(x,y) <6 and |g(x) — g(y)| > €. Let 2,,4m € K and gn € S be such that d(tn,Yn) < 4 and |gn(tn) — gn(¥n)| > €. Then no subsequence of {gn} can be equicontinuous, since |9n, (2m, — Guy (Ym) > €. Hence by Theorem 7.24 no subsequence of {gn} can converge in C(K), and so S cannot be compact. We conclude, then, that if S is compact, then S is closed, bounded, and equicontinuous. Conversely, if $ is closed, pointwise bounded, and equicontinuous, then every sequence {gn} contains a subsequence that converges uniformly, hence converges in the metric of C(K’) (by Ascoli’s theorem). Since $ is closed, the limit belongs to S, and so $ is compact by Exercise 26 of Chapter 2. Exercise 7.20 If f is continuous on [0,1] and if [ Sera=o (n=0,1,2...), 0 prove that f(z) = 0 on [0,1]. Hint: The integral of the product of f with any 2 polynomial is zero. Use the Weierstrass theorem to show that [ f(z) dz =0. o Solution. There exists a sequence of polynomials pa(z) such that pa(z) con- verges uniformly to f(x). Since f is bounded, {pn} is uniformly bounded, and hence pn f converges uniformly to f?. Then by Theorem 7.16 : : [ Pee km [ pa(2) f(a) dz = 0. f anee Jo But we know already (Exercise 2 of Chapter 6) that this implies f?(cr) = 0. Exercise 7.21 Let K be the unit circle in the complex plane (i.e,, the set of all 2 with |z| = 1), and let A be the algebra of all functions of the form iv Fe) = Srene™® (6 reall). a0 | | | 123 ‘The A separates points on K,, and A vanishes at no point of K, but nevertheless there are continuous functions on K which are not in the uniform closure of A. Hint: For every f € A 2 Flee! d0 =0, and this is also true for every f in the closure of A Solution. The function f(z) = 2 € A separates points on K and never vanishes. The equality given in the hint is a straighforward computation. It implies that the continuous function +, which is e~¥, is not in the uniform closure of A, since 2 [ ene! db = On. lo Exercise 7.22 Assume f € R(a) on [a,2], and prove that there are polynomials Pr such that, . lim [ If - Pal?da =0. (Compare with Exercise 12, Chap. 6.) Solution. The parenthetical remark refers to the proof that there is @ sequence of continuous functions {f,} such that sim, [Ur —talPda =o All that is now needed is to note that one can find polynomials P, such that Jn(t) — Pa(2)| < 2 for all « € (a, ] and all n. Exercise 7.23 Put P, =0, and define, for n = 0,1,2,..., 2 Ps) Posi(a) = Pa(e) + 5 Prove that lim Pa(z) = uniformly on (1,1) (This makes it possible to prove the Stone-Weierstrass theorem withont first proving Theorem 7.26. Hint: Use the identity — Pri) 124 CHAPTER 7. SEQUENCES AND SERIES OF FUNCTIONS to prove that 0 < Py(z) $ Pnii(z) $ |e| if |r| <1, and that ~#) eee nel if |x| <1. poe The identity given in the hint is a trivial consequence of the identity 2? — P2(2) — P,(2)]{lz| + Pa(z)]. Then, granting that 0 < P,(2) < ‘el, pe corinne ener nee ure iz] - Pa4a(z) < |x| — Pa(x), which gives all of the desired inequalities. An immediate corollary of the same identity (obtained by replacing P,(z) by 0 in the second factor on the right-hand side) is lal — Pasa(2) < [lel - Pa(e)|(1- ), and this inequality makes it possible to obtain the inequality ial Pa) < ti(a- EL)” by induction on n. Finally, by symmetry, the maximum of |z|(1 — #1)” on {-1,1] is its maximum on {0,1}, and this can be found by simple calculus to occur at . Since this function is always less than |z|, the final inequality now follows. a Exercise 7.24 Let X be a metric space, with metric d. Fix a point @ € X. Assign to each p € X the function fy defined by fy(a) = d(z,p)—d(z,a) (x €X). Prove that |fp(z)| < d(a,p) for all x € X, and therefore f, € C(X) Prove that llf — fall = dp, 9) for all p,q € X. H®(p) = fy,it ae that & is an isometry (a distance-preserving mapping) of X onto #(X) c C(X’ Let Y be the cae ‘of ®(X) in C(X). Show that ¥ is complete. Conclusion: X is isometric to a dense subset of a complete metric space Y- (Exercise 24, Chap. 3 contains a different proof of this.) Solution. The inequality | f;(2)| < d(a,p) is well-known, i.e., the fact that ld(z, p) — d(x,a)| < d(a,p) 125 and follows from the triangle inequality by merely transposing a term. (The left-hand side is either d(z,p) — d(z,a) or d(z,a) — d(z,p). Whichever is the case, if the subtracted term is moved to the other side, we have the ordinary triangle inequality.) ‘As for the isometry, we certainly have, for all 2, Ifa and equality holds here if x = q or z = p. Hence the supremum over all x is exactly d(p,q): ‘As for the closure ¥ of @(X) being complete, it is a closed subset of a complete metric space, hence necessarily complete. By definition of closure, 5(X) is dense in Y. — fo(z)| = ld(x, 9) — d(x, p)| < d(p,9) Exercise 7.25 Suppose ¢ is a continuous bounded real function in the strip defined by 0 < 2 <1, co < y < oo. Prove that the initial-value problem y¥ =9(z,y), y(0) =e has a solution. (Note that the hypotheses of this existence theorem are less strigent than those of the corresponding uniqueness theorem; see Exercise 27, Chap. 5.) Hint: Fix n. For i =0, on [0,1] such that fn (0) = ¢ nj put = i/n. Let fa be a continuous function Fat) = O(ts, fn(@s)) if ect OF) uniformly on (0, 1. 126 CHAPTER 7. SEQUENCES AND SERIES OF FUNCTIONS (e) Aa(t) +0 uniformly on (0, 1], since Malt) = (as falas) ~ (ts fall) in (21,2443). (f) Hence : =c+ L(t) ae fe)=ex [ote.s)at This f is a solution of the given problem. Solution, It will save trouble if we assume that ¢ is a bounded continuous mapping from (0, 1] x R* into R* and that c is a vector in R*. That way we can do Exercise 26 simultaneously with this one. Since we are defining the functions f(t) to be piecewise-linear, there is no difficulty in doing this with vector-valued functions. We simply define f,(t) = c+ #6(0,c) for 0 0 let & > 0 be such that |¢(t,y) — O(t,z)]

    También podría gustarte