Real Analysis Homework 2

import%20marimo%0A%0A__generated_with%20%3D%20%220.17.6%22%0Aapp%20%3D%20marimo.App()%0A%0A%0A%40app.cell%0Adef%20_()%3A%0A%20%20%20%20import%20marimo%20as%20mo%0A%20%20%20%20return%20(mo%2C)%0A%0A%0A%40app.cell%0Adef%20_(mo)%3A%0A%20%20%20%20nav_menu%20%3D%20mo.nav_menu(%0A%20%20%20%20%20%20%20%20%7B%0A%20%20%20%20%20%20%20%20%20%20%20%20%22%2Findex.html%22%3A%20f%22%7Bmo.icon('lucide%3Ahome')%7D%20Home%22%2C%0A%20%20%20%20%20%20%20%20%7D%2C%0A%20%20%20%20%20%20%20%20orientation%3D%22vertical%22%2C%0A%20%20%20%20)%0A%20%20%20%20nav_menu%0A%20%20%20%20return%0A%0A%0A%40app.cell%0Adef%20_(mo)%3A%0A%20%20%20%20mo.md(r%22%22%22%0A%20%20%20%20%23%20Spring%20Semester%20Homework%201%0A%20%20%20%20%22%22%22)%0A%20%20%20%20return%0A%0A%0A%40app.cell%0Adef%20_(mo)%3A%0A%20%20%20%20mo.md(r%22%22%22%0A%20%20%20%20%23%23%20Exercise%201%0A%0A%20%20%20%20**Let%20%24X%24%20consists%20of%20two%20points%20%24a%24%20and%20%24b%24%2C%20put%20%24%5Cmu(%5Clbrace%20a%20%5Crbrace)%20%3D%20%5Cmu(%5Clbrace%20b%20%5Crbrace)%20%3D%201%2F2%24%2C%20and%20let%20%24L%5Ep(%5Cmu)%24%20be%20the%20resulting%20real%20%24L%5Ep%24-space.%20Identify%20each%20real%20function%20%24f%24%20on%20%24X%24%20with%20the%20point%20%24(f(a)%2Cf(b))%24%20in%20the%20plane%2C%20and%20sketch%20the%20unit%20balls%20of%20%24L%5Ep(%5Cmu)%24%20for%20%240%20%3C%20p%20%5Cle%20%5Cinfty%24.%20Note%20that%20they%20are%20convex%20if%20and%20only%20if%20%241%20%5Cle%20p%20%5Cle%20%5Cinfty%24.%20For%20which%20%24p%24%20is%20this%20unit%20ball%20a%20square%3F%20A%20circle%3F%20If%20%24%5Cmu(%5Clbrace%20a%20%5Crbrace)%20%5Cne%20%5Cmu(%5Clbrace%20b%20%5Crbrace)%24%2C%20how%20does%20the%20situation%20differ%20from%20the%20preceding%20one**%3F%0A%0A%20%20%20%20Solution%0A%0A%20%20%20%20We%20want%20to%20find%20the%20unit%20balls%20of%20%24L%5Ep(%5Cmu)%24%2C%20we%20just%20define%20the%20%24L%5Ep%24-norm%20in%20the%20usual%20way%2C%20and%20see%20which%20functions%20are%20of%20%24B_p%20%3D%20%5Clbrace%20f%20%5Cin%20L%5Ep(%5Cmu)%20%7C%20%5C%7C%20f%20%5C%7C_p%20%5Cle%201%20%5Crbrace%24.%20As%20usual%2C%20we%20need%20to%20consider%20two%20cases%3A%0A%0A%20%20%20%20If%20%24p%20%3D%20%5Cinfty%24%2C%20since%20both%20set%20%24%5Clbrace%20a%20%5Crbrace%2C%20%5Clbrace%20b%20%5Crbrace%24%20are%20of%20finite%20non-zero%20measure%2C%20we%20can%20see%20the%20essential%20supremum%20is%20just%20the%20maximum%20of%20%24%7C%20f(a)%20%7C%24%20and%20%24%7C%20f(b)%20%7C%24.%20I.e.%20%24f%20%5Cin%20B_%7B%5Cinfty%7D%24%20if%20and%20only%20if%20%24%5Cmax%7B%5Clbrace%20%7C%20f(a)%20%7C%2C%20%7C%20f(b)%20%7C%20%5Crbrace%7D%20%5Cle%201%24.%20If%20we%20identify%20%24f%24%20on%20the%20real%20plane%2C%20this%20is%20a%20square%20(of%20length%20%242%24%2C%20centered%20at%20origin).%0A%0A%20%20%20%20If%20%240%20%3C%20p%20%3C%20%5Cinfty%24%2C%20%24%5C%7C%20f%20%5C%7C_p%20%3A%3D%20%20(%5Cint_X%20%7C%20f%20%7C%5Epd%5Cmu)%5E%7B1%2Fp%7D%24%2C%20by%20assumption%20this%20equals%20%24(%5Cfrac%7B1%7D%7B2%7D%7C%20f(a)%20%7C%5Ep%20%2B%20%5Cfrac%7B1%7D%7B2%7D%7C%20f(b)%20%7C%5Ep)%5E%7B1%2Fp%7D%24%20%2C%20move%20%24%5Cfrac%7B1%7D%7B2%7D%24%20out%20and%20take%20%24p%24-power%2C%20it%20is%20not%20hard%20to%20see%20%24%5C%7C%20f%20%5C%7C_p%20%5Cle%201%24%20if%20and%20only%20if%20%24%7C%20f(a)%20%7C%5Ep%20%2B%20%7C%20f(b)%20%7C%5Ep%20%5Cle%202%24.%0A%0A%20%20%20%20So%20if%20%24p%20%3D%201%24%2C%20this%20is%20a%20square%20(rotated)%3B%20if%20%24p%20%3D%202%24%20this%20is%20a%20circle%20(I%20think%20we%20should%20really%20call%20it%20a%20disk)%3B%20if%20%241%3Cp%3C2%24%2C%20this%20is%20some%20shape%20in%20between%20a%20square%20and%20a%20circle%3B%20if%20%240%3Cp%20%3C1%24%20the%20shape%20is%20'thinner'%20and%20if%20%242%3Cp%3C%5Cinfty%24%20it%20is%20'fatter'%3A%0A%0A%20%20%20%20%3Cimg%20src%3D%22public%2FPasted%20image%2020220309211216.png%22%20width%3D%22600%22%20%2F%3E%0A%0A%20%20%20%20If%20%24%5Cmu(a)%20%5Cne%20%5Cmu(b)%24%3A%20if%20one%20of%20them%20is%20%24%5Cinfty%24%20then%20%24f%24%20can%20only%20take%20%240%24%20there%2C%20so%20%24(f(a)%2Cf(b))%24%20form%20some%20line%20segment%3B%20if%20one%20of%20them%20is%20%240%24%20then%20it%20is%20a%20band%3B%20if%20one%20of%20them%20is%20%24%5Cinfty%24%20and%20the%20other%20is%20%240%24%20then%20it%20is%20a%20line%3B%20otherwise%20%24%5Cmu(a)%20%5Cne%20%5Cmu(b)%24%20but%20they%20are%20still%20finite%20and%20positive%2C%20then%20the%20%24L%5E%7B%5Cinfty%7D%24%20case%20does%20not%20change%2C%20for%20%24L%5Ep%24%20case%20we%20need%20to%20have%20%24%5Cmu(a)%7C%20f(a)%20%7C%5Ep%20%2B%20%5Cmu(b)%20%7C%20f(b)%20%7C%5Ep%20%5Cle%201%24%2C%20so%20the%20graph%20is%20'stretched'%20so%20it's%20never%20a%20square%20or%20circle%2C%20for%20example%20if%20%24p%20%3D%202%24%20it%20is%20an%20ellipse.%0A%0A%20%20%20%20%23%23%20Exercise%202%0A%0A%20%20%20%20**Prove%20that%20the%20unit%20ball%20(open%20or%20closed)%20is%20convex%20in%20every%20normed%20linear%20space**.%0A%0A%20%20%20%20Proof%0A%0A%20%20%20%20Let%20%24X%24%20be%20a%20normed%20linear%20space%2C%20let%20%24B%24%20be%20its%20open%20unit%20ball%2C%20then%20for%20any%20%24x%2C%20y%20%5Cin%20B%24%20and%20%24t%20%5Cin%20%5B0%2C1%5D%24%20we%20have%20%24%5C%7C%20tx%20%2B%20(1-t)y%20%5C%7C%24%20%24%5Cle%20t%5C%7C%20x%20%5C%7C%20%2B%20(1-t)%5C%7C%20y%20%5C%7C%24%20(by%20the%20definition%20of%20a%20normed%20linear%20space)%20%24%5Cle%20%5Cmax%7B%5Clbrace%20%5C%7C%20x%20%5C%7C%2C%20%5C%7C%20y%20%5C%7C%20%5Crbrace%7D%20%3C%201%24%2C%20so%20%24tx%20%2B(1-t)y%20%5Cin%20B%24%2C%20so%20by%20definition%20%24B%24%20is%20convex.%20For%20the%20closed%20unit%20ball%2C%20just%20change%20%24%3C%24%20to%20%24%5Cle%24%20and%20the%20argument%20is%20the%20same.%0A%0A%20%20%20%20%23%23%20Exercise%203%0A%0A%20%20%20%20**If%20%241%20%3C%20p%20%3C%20%5Cinfty%24%2C%20prove%20that%20the%20unit%20ball%20of%20%24L%5Ep(%5Cmu)%24%20is%20strictly%20convex%3B%20this%20means%20that%20if**%20%24%24%5C%7C%20f%20%5C%7C_p%20%3D%20%5C%7C%20g%20%5C%7C_p%20%3D%201%2C%20f%20%5Cne%20g%2C%20h%20%3D%20%5Cfrac%7B1%7D%7B2%7D(f%2Bg)%2C%24%24%20**then%20%24%5C%7C%20h%20%5C%7C_p%20%3C%201%24.%20Show%20that%20this%20fails%20in%20every%20%24L%5E1(%5Cmu)%24%2C%20%24L%5E%7B%5Cinfty%7D(%5Cmu)%24%20and%20%24C(X)%24%20(ignore%20triviality%2C%20such%20as%20spaces%20consisting%20of%20only%20one%20point)**.%0A%0A%20%20%20%20Proof%0A%0A%20%20%20%20Let%20%24f%2Cg%2Ch%24%20be%20as%20in%20the%20statement%20for%20strictly%20convex.%20By%20Minkowski's%20inequality%20we%20have%20%24%24%5C%7C%20h%20%5C%7C_p%20%3D%20%5C%7C%20%5Cfrac%7B1%7D%7B2%7D(f%2Bg)%20%5C%7C_p%20%5Cle%20(%5C%7C%20f%20%5C%7C_p%20%2B%20%5C%7C%20g%20%5C%7C_p)%2F2%20%3D%201.%24%24%20Now%20remember%20the%20equality%20in%20Minkowski's%20inequality%20holds%20if%20and%20only%20if%3A%0A%0A%20%20%20%201.%20One%20of%20%24f%2C%20g%24%20is%20%240%24%20almost%20everywhere%3A%20we%20cannot%20have%20this%2C%20because%20the%20norms%20are%20assumed%20to%20be%20%241%24%3B%0A%20%20%20%202.%20%24f%20%3D%20ag%24%20almost%20everywhere%20for%20some%20constant%20%24a%24%3A%20we%20cannot%20have%20this%20neither%2C%20because%20again%20the%20norms%20are%20both%20%241%24%2C%20so%20%24a%24%20must%20be%20%241%24%20thus%20%24f%20%3D%20g%24%2C%20but%20we%20assumed%20%24f%20%5Cne%20g%24.%0A%0A%20%20%20%20So%20equality%20cannot%20be%20achieved%20anyway%2C%20so%20we%20have%20%24%5C%7C%20h%20%5C%7C_p%20%3C%201%24%2C%20and%20thus%20we%20have%20strict%20convexity.%0A%0A%20%20%20%20Assume%20the%20space%20is%20not%20trivial%2C%20strictly%20convex%20fails%20in%3A%0A%0A%20%20%20%201.%20Every%20%24L%5E1(%5Cmu)%24%3A%20take%20%24E_1%2C%20E_2%24%20any%20two%20disjoint%20measurable%20sets%20in%20%24X%24%20such%20that%20%240%20%3C%20%5Cmu(E_1)%2C%20%5Cmu(E_2)%20%3C%20%5Cinfty%24%2C%20define%20%24f%20%3D%20%5Cfrac%7B%5Cchi_%7BE_1%7D%7D%7B%5Cmu(E_1)%7D%24%20and%20%24g%20%3D%20%5Cfrac%7B%5Cchi_%7BE_2%7D%7D%7B%5Cmu(E_2)%7D%24%2C%20then%20it%20is%20easy%20to%20see%20%24%5C%7C%20f%20%5C%7C_1%20%3D%20%5C%7C%20g%20%5C%7C_1%20%3D%201%24%2C%20%24f%20%5Cne%20g%24%20and%20%24%5C%7C%20%5Cfrac%7B1%7D%7B2%7D(f%2Bg)%20%5C%7C_1%20%3D%201%24%3B%0A%20%20%20%202.%20Every%20%24L%5E%7B%5Cinfty%7D(%5Cmu)%24%3A%20similarly%2C%20define%20%24f%20%3D%20%5Cchi_%7BE_1%7D%24%20and%20%24g%20%3D%20%5Cchi_%7BE_1%20%5Ccup%20E_2%7D%24%2C%20this%20gives%20a%20counterexample%3B%0A%20%20%20%203.%20Every%20%24C(X)%24%3A%20from%20this%20notation%20in%20the%20book%20we%20should%20assume%20%24X%24%20is%20a%20compact%20normed%20linear%20space%20and%20we%20are%20using%20the%20sup-norm%3B%20we%20also%20know%20%24X%24%20can%20be%20regarded%20as%20a%20metric%20space%2C%20say%20the%20distance%20function%20is%20%24d%24.%20Fix%20a%20point%20%24a%20%5Cin%20X%24%20and%20define%20%24f(x)%20%3D%20%5Cfrac%7B1%7D%7B1%2Bd(x%2Ca)%7D%24%20and%20%24g(x)%20%3D%20%5Cfrac%7B1%7D%7B(1%2Bd(x%2Ca))%5E2%7D%24%2C%20they%20are%20continuous.%20It%20is%20easy%20to%20see%20both%20functions%20achieve%20maximum%20thus%20supremum%20at%20%24x%20%3D%20a%24%20and%20have%20norm%20equals%20%241%24%3B%20%24f%20%5Cne%20g%24%20and%20%24h%20%3D%20%5Cfrac%7B1%7D%7B2%7D(f%2Bg)%24%20also%20has%20norm%20%241%24.%0A%0A%20%20%20%20Those%20counterexamples%20all%20work%20in%20any%20corresponding%20spaces%2C%20so%20we%20have%20the%20desired%20statement.%0A%0A%20%20%20%20%23%23%20Exercise%204%0A%0A%20%20%20%20**Let%20%24C%24%20be%20the%20space%20of%20all%20continuous%20functions%20on%20%24%5B0%2C1%5D%24%20with%20the%20supremum%20norm.%20Let%20%24M%24%20consists%20of%20all%20%24f%20%5Cin%20C%24%20for%20which**%20%24%24%5Cint_0%5E%7B1%2F2%7Df(t)dt%20-%20%5Cint_%7B1%2F2%7D%5E1f(t)dt%20%3D%201.%24%24%20**Prove%20that%20%24M%24%20is%20a%20closed%20convex%20subset%20of%20%24C%24%20which%20contains%20no%20element%20of%20minimal%20norm**.%0A%0A%20%20%20%20Proof%0A%0A%20%20%20%20First%2C%20%24M%24%20is%20not%20empty%3A%20let%20%24f%24%20be%20the%20function%20that%20takes%20%240%24%20on%20%24%5B1%2F2%2C%201%5D%24%20and%20is%20linear%20on%20%24%5B0%2C1%2F2%5D%24%20with%20appropriate%20coefficients%2C%20this%20%24f%20%5Cin%20M%24.%0A%0A%20%20%20%20Suppose%20%24f%2C%20g%20%5Cin%20M%24%20and%20%24s%20%5Cin%20%5B0%2C1%5D%24%2C%20let%20%24h%20%3D%20sf%20%2B%20(1-s)g%24%2C%20we%20want%20to%20show%20%24h%20%5Cin%20M%24%2C%20write%3A%20%24%24%5Cbegin%7Baligned%7D%26~~~~~%5Cint_0%5E%7B0.5%7Dhdt%20-%20%5Cint_%7B0.5%7D%5E1hdt%20%5C%5C%20%26%3D%20%5Cint_0%5E%7B0.5%7Dsf%20%2B%20(1-s)gdt%20-%20%5Cint_%7B0.5%7D%5E1sf%20%2B%20(1-s)gdt%20%5C%5C%20%26%3D%20s(%5Cint_0%5E%7B0.5%7D%20f%20dt%20-%20%5Cint_%7B0.5%7D%5E1fdt)%20%2B%20(1-s)(%5Cint_0%5E%7B0.5%7D%20g%20dt%20-%20%5Cint_%7B0.5%7D%5E1gdt)%20%5C%5C%20%26%3Ds%2B1-s%20%5C%5C%26%3D%201%5Cend%7Baligned%7D%2C%24%24%20also%20%24h%24%20is%20clearly%20continuous%2C%20so%20indeed%20%24h%20%5Cin%20M%24%2C%20%24M%24%20is%20convex.%0A%0A%20%20%20%20Now%20let%20%24f_n%24%20be%20a%20sequence%20of%20functions%20in%20%24M%24%20and%20%24f_n%20%5Cto%20f%24%2C%20we%20want%20to%20show%20%24f%20%5Cin%20M%24.%20We%20have%20%24%5C%7C%20f_n%20-%20f%20%5C%7C%20%3D%20%5Csup%5Climits_x%20%7C%20f_n(x)%20-%20f(x)%20%7C%20%5Cto%200%24%20(remark%3A%20this%20means%20%24f_n%20%5Cto%20f%24%20uniformly).%20In%20particular%20we%20can%20find%20%24N%24%20such%20that%20for%20all%20%24n%3EN%24%2C%20%24%5Csup%5Climits_x%7C%20f_n(x)%20-%20f(x)%20%7C%20%3C%201%24%2C%20so%20%24f_n%24%20is%20bounded%20by%20%24%5Cmax%7B%5Clbrace%20f_1%2C%5Cdots%2Cf_N%2C%20f%2B1%5Crbrace%7D%24%2C%20so%20Dominated%20Convergence%20Theorem%20says%20that%20we%20can%20interchange%20the%20limit%20and%20integrant%20operation%2C%20and%20we%20have%20%24%24%5Cbegin%7Baligned%7D1%20%26%3D%20%5Clim%20%5Cint_0%5E%7B0.5%7Df_ndt%20-%20%5Cint_%7B0.5%7D%5E1f_ndt%20%5C%5C%26%3D%20%5Cint_0%5E%7B0.5%7D%5Clim%20f_ndt%20-%20%5Cint_%7B0.5%7D%5E1%5Clim%20f_ndt%20%5C%5C%26%3D%20%5Cint_0%5E%7B0.5%7Dfdt%20-%20%5Cint_%7B0.5%7D%5E1fdt%5Cend%7Baligned%7D%24%24%20as%20desired.%20So%20%24M%24%20is%20closed.%0A%0A%20%20%20%20Finally%20we%20show%20that%20it%20does%20not%20have%20an%20element%20of%20minimal%20norm.%20First%20notice%20%241%20%3D%20%7C%201%20%7C%24%20%24%3D%20%7C%20%5Cint_0%5E%7B0.5%7Dfdt%20-%20%5Cint_%7B0.5%7D%5E1fdt%20%7C%24%20%24%5Cle%20%5Cint_0%5E%7B0.5%7D%5C%7C%20f%20%5C%7C%20dt%20%2B%20%5Cint_%7B0.5%7D%5E1%5C%7C%20f%20%5C%7C%20dt%24%20%24%3D%5C%7C%20f%20%5C%7C%24.%20Suppose%20%24%5C%7C%20f%20%5C%7C%20%3D%201%24%2C%20then%20the%20equality%20of%20above%20inequality%20holds%2C%20that%20is%20%24%5Cint_0%5E1%7C%20f%20%7C%20dt%20%3D%201%24%2C%20and%20we%20can%20see%20that%20%24%7C%20f%20%7C%20%3D%201%24%20on%20%24%5B0%2C1%5D%24%20(a.e.%2C%20but%20not%20that%20important%2C%20as%20we%20are%20talking%20about%20continuous%20functions%20here)%2C%20but%20in%20this%20case%20to%20have%20%24%5Cint_0%5E%7B0.5%7Dfdt%20-%20%5Cint_%7B0.5%7D%5E1fdt%20%3D%201%24%20we%20must%20have%20%24f%20%3D%201%24%20on%20%24%5B0%2C1%2F2%5D%24%20and%20%24f%20%3D%20-1%24%20on%20%24%5B1%2F2%2C1%5D%24%2C%20which%20makes%20%24f%24%20not%20continuous.%20So%20there%20is%20no%20element%20with%20norm%20%241%24.%20However%2C%20for%20any%20%24%5Cvarepsilon%24%2C%20we%20can%20see%20that%20there%20is%20some%20%24f%20%5Cin%20M%24%20with%20norm%20%241%2B%5Cvarepsilon%24%3A%0A%0A%20%20%20%20%3Cimg%20src%3D%22public%2FPasted%20image%2020220309162929.png%22%20width%3D%22400%22%20%2F%3E%0A%0A%20%20%20%20Some%20work%20is%20needed%20to%20write%20down%20the%20explicit%20function%2C%20but%20it%20is%20definitely%20achievable.%20It%20shows%20that%20%24%5Cinf%5Climits_%7Bf%20%5Cin%20M%7D%5Clbrace%20%5C%7C%20f%20%5C%7C%20%5Crbrace%24%20is%20indeed%20%241%24%20but%20it%20is%20never%20achieved.%0A%0A%20%20%20%20%23%23%20Exercise%205%0A%0A%20%20%20%20**Let%20%24M%24%20be%20the%20set%20of%20all%20%24f%20%5Cin%20L%5E1(%5B0%2C1%5D)%24%20relative%20to%20Lebesgue%20measure%20such%20that**%20%24%24%5Cint_0%5E1f(t)dt%20%3D%201.%24%24%20**Show%20that%20%24M%24%20is%20a%20closed%20convex%20subset%20of%20%24L%5E1(%5B0%2C1%5D)%24%20which%20contains%20infinitely%20many%20elements%20of%20minimal%20norm**.%0A%0A%20%20%20%20Proof%0A%0A%20%20%20%20Let%20%24f%24%20be%20constant%20%241%24%2C%20then%20%24f%20%5Cin%20M%24%2C%20thus%20%24M%24%20is%20not%20empty.%20Suppose%20%24f%2C%20g%20%5Cin%20M%24%20and%20%24s%20%5Cin%20%5B0%2C1%5D%24%2C%20let%20%24h%20%3D%20sf%20%2B%20(1-s)g%24%2C%20then%20%24%24%5Cint_0%5E1%20%7C%20h%20%7C%20dt%20%3D%20%5Cint_0%5E1%7C%20sf%2B(1-s)g%20%7C%20dt%20%5Cle%20t%5Cint_0%5E1%7C%20f%20%7C%20dt%20%2B%20(1-s)%20%5Cint_0%5E1%7C%20g%20%7C%20dt%2C%24%24%20since%20both%20%24f%2Cg%24%20are%20by%20assumption%20%24L%5E1%24%2C%20%24h%24%20is%20%24L%5E1%24.%20Also%20%24%24%5Cint_0%5E1%20h%20dt%20%3D%20%5Cint_0%5E1%20sf%2B(1-s)gdt%20%3D%20s%5Cint_0%5E1fdt%20%2B%20(1-s)%5Cint_0%5E1gdt%20%3D%201%24%24%20thus%20%24h%20%5Cin%20M%24%2C%20thus%20%24M%24%20is%20convex.%0A%0A%20%20%20%20Suppose%20%24f_n%24%20is%20a%20sequence%20in%20%24M%24%20and%20%24f_n%20%5Cto%20f%24%2C%20meaning%20that%20%24%5C%7C%20f_n%20-%20f%20%5C%7C_1%20%5Cto%200%24.%20Since%20convergent%20sequence%20is%20bounded%2C%20%24f_n%20-%20f%24%20must%20be%20%24L%5E1%24%20for%20all%20%24n%24%2C%20thus%20%24%24%7C%20%5Cint_0%5E1%20f_n%20dt%20-%20%5Cint_0%5E1%20f%20dt%20%7C%20%3D%20%7C%20%5Cint_0%5E1%20f_n%20-%20f%20dt%20%7C%20%5Cle%5Cint_0%5E1%20%7C%20f_n%20-%20f%20%7C%20dt%20%3D%3A%20%5C%7C%20f_n%20-%20f%20%5C%7C_1%20%5Cto%200%2C%24%24%20in%20particular%20%241%20%3D%20%5Clim%5Cint_0%5E1%20f_ndt%20%3D%20%5Cint_0%5E1fdt%24%2C%20thus%20%24f%20%5Cin%20M%24%20and%20%24M%24%20is%20closed.%0A%0A%20%20%20%20Since%20%241%20%3D%20%7C%20%5Cint_0%5E1%20f%20dt%20%7C%20%5Cle%20%5Cint_0%5E1%20%7C%20f%20%7C%20dt%20%3D%3A%20%5C%7C%20f%20%5C%7C_1%24%2C%20the%20norm%20of%20any%20%24f%20%5Cin%20M%24%20is%20at%20least%20%241%24.%20For%20each%20%24n%20%3D%201%2C%202%2C%20%5Cdots%24%2C%20define%20%24f_n%24%20as%20%24f_n(x)%20%3D%20nx%5E%7Bn-1%7D%24%2C%20then%20%24f_n%20%5Cne%20f_m%24%20for%20%24n%20%5Cne%20m%24%20and%20each%20%24f_n%24%20is%20non-negative.%20A%20direct%20calculation%20gives%20%24%24%5Cint_0%5E1f_ndt%20%3D%20%5Cint_0%5E1nt%5E%7Bn-1%7Ddt%3D1%5En-0%5En%20%3D%201%24%24%20so%20all%20%24f_n%20%5Cin%20M%24%2C%20and%20that%20%24%24%5C%7C%20f_n%20%5C%7C_1%20%3D%20%5Cint_0%5E1%20%7C%20f_n%20%7C%20dt%20%3D%20%5Cint_0%5E1%20f_n%20dt%20%3D%201%2C%24%24%20thus%20they%20are%20all%20of%20norm%20%241%24%2C%20the%20minimal%20one%2C%20and%20there%20are%20infinitely%20many%20of%20them.%0A%0A%20%20%20%20%23%23%20Exercise%206%0A%0A%20%20%20%20**Let%20%24f%24%20be%20a%20bounded%20linear%20functional%20on%20a%20subspace%20%24M%24%20of%20a%20Hilbert%20space%20%24H%24.%20Prove%20that%20%24f%24%20has%20a%20unique%20norm-preserving%20extension%20to%20a%20bounded%20linear%20functional%20on%20%24H%24%2C%20and%20that%20this%20extension%20vanishes%20on%20%24M%5ET%24**.%0A%0A%20%20%20%20Proof%0A%0A%20%20%20%20Any%20Hilbert%20space%20is%20a%20Banach%20space.%20Let%20%24f%3A%20M%20%5Cto%20%5Cmathbb%7BC%7D%24%20be%20a%20bounded%20linear%20functional.%20By%20Theorem%205.4%20%24f%24%20is%20uniformly%20continuous.%20Since%20%24M%24%20is%20by%20definition%20a%20dense%20subset%20of%20its%20closure%20%24%5Coverline%7BM%7D%24%2C%20there%20is%20a%20unique%20uniformly%20continuous%20extension%20%24%5Coverline%7Bf%7D%3A%20%5Coverline%7BM%7D%20%5Cto%20%5Cmathbb%7BC%7D%24%20of%20%24f%24.%20Thus%20we%20may%20assume%20%24M%24%20is%20closed.%0A%0A%20%20%20%20We%20know%20that%20a%20closed%20subspace%20of%20a%20Hilbert%20space%20is%20again%20a%20Hilbert%20space%2C%20thus%20there%20is%20a%20unique%20%24y%20%5Cin%20M%24%20such%20that%20%24f(x)%20%3D%20%5Clangle%20x%2Cy%20%5Crangle%24%20for%20all%20%24x%20%5Cin%20M%24.%20By%20definition%20%24%5C%7C%20f%20%5C%7C%24%20is%20the%20smallest%20number%20such%20that%20%24%5C%7C%20f(x)%20%5C%7C%20%5Cle%20%5C%7C%20f%20%5C%7C%20%5C%7C%20x%20%5C%7C%24%20for%20all%20%24x%24%2C%20yet%20from%20definition%20of%20the%20inner%20product%20in%20the%20Hilbert%20space%20we%20have%20%24%5C%7C%20f(y)%20%5C%7C%20%3D%20%5Clangle%20y%2Cy%20%5Crangle%20%3D%20%5C%7C%20y%20%5C%7C%5C%7C%20y%20%5C%7C%24%2C%20we%20must%20have%20%24%5C%7C%20f%20%5C%7C%20%3D%20%5C%7C%20y%20%5C%7C%24.%0A%0A%20%20%20%20Define%20%24F%3A%20H%20%5Cto%20%5Cmathbb%7BC%7D%24%20as%20%24F(x)%20%3D%20%5Clangle%20x%2C%20y%20%5Crangle%24%2C%20then%20of%20course%20this%20is%20an%20extension%20of%20%24f%24.%20For%20%24x%20%5Cin%20M%5ET%24%2C%20%24F(x)%20%3D%20%5Clangle%20x%2Cy%20%5Crangle%20%3D%200%24%20because%20%24y%20%5Cin%20M%24.%20By%20a%20very%20same%20argument%2C%20%24%5C%7C%20F%20%5C%7C%20%3D%20%5C%7C%20y%20%5C%7C%24%20as%20well%2C%20so%20the%20extension%20is%20a%20norm-preserving%20extension.%0A%0A%20%20%20%20(The%20above%20part%20can%20also%20start%20by%20using%20Hahn-Banach%20Theorem%2C%20which%20states%20that%20given%20%24M%24%20a%20subspace%20of%20a%20normed%20linear%20space%20(any%20Hilbert%20space%20is%20a%20normed%20linear%20space)%20%24H%24%20and%20%24f%24%20a%20bounded%20linear%20functional%20on%20%24M%24%2C%20then%20%24f%24%20can%20be%20extended%20to%20a%20bounded%20linear%20functional%20%24F%24%20on%20%24H%24%20so%20that%20%24%5C%7C%20F%20%5C%7C%20%3D%20%5C%7C%20f%20%5C%7C%24.%20Here%20we%20don't%20need%20to%20bother%20if%20%24M%24%20is%20closed)%0A%0A%20%20%20%20It%20remains%20to%20prove%20uniqueness.%20Suppose%20%24F'%24%20is%20also%20a%20norm-preserving%20extension%20of%20%24f%24%20such%20that%20%24F'(x)%20%3D%200%24%20for%20%24x%20%5Cin%20M%5ET%24.%20Again%20there%20is%20a%20unique%20%24y'%20%5Cin%20M%24%20such%20that%20%24F'(x)%20%3D%20%5Clangle%20x%2Cy'%20%5Crangle%24.%20Also%20%24%5Clangle%20x%2Cy%20%5Crangle%20%3D%20%5Clangle%20x%2Cy'%20%5Crangle%24%20on%20%24M%24%2C%20so%20%24%5Clangle%20x%2Cy%20%5Crangle%20-%20%5Clangle%20x%2Cy'%20%5Crangle%20%3D%200%24%20and%20thus%20%24%5Clangle%20x%2C%20y-y'%20%5Crangle%20%3D%200%24%2C%20thus%20%24y%20-%20y'%20%5Cin%20M%5ET%24.%20In%20particular%2C%20%24y%24%20can%20be%20uniquely%20written%20as%20%24y%20%3D%20y'%20%2B%20(y%20-%20y')%24%20where%20%24y'%20%5Cin%20M%24%20and%20%24y-y'%20%5Cin%20M%5ET%24%2C%20and%20that%20%24%5C%7C%20y%20%5C%7C%5E2%20%3D%20%5C%7C%20y'%20%5C%7C%5E2%20%2B%20%5C%7C%20y-y'%20%5C%7C%5E2%24.%20Since%20%24%5C%7C%20F'%20%5C%7C%20%3D%20%5C%7C%20y'%20%5C%7C%20%3D%20%5C%7C%20f%20%5C%7C%20%3D%20%5C%7C%20y%20%5C%7C%20%3D%20%5C%7C%20F%20%5C%7C%24%2C%20we%20must%20have%20%24%5C%7C%20y-y'%20%5C%7C%20%3D%200%24%20thus%20%24y%20%3D%20y'%24%2C%20thus%20%24F'%20%3D%20F%24.%0A%0A%20%20%20%20%23%23%20Exercise%207%0A%0A%20%20%20%20**Construct%20a%20bounded%20linear%20functional%20on%20some%20subspace%20of%20some%20%24L%5E1(%5Cmu)%24%20which%20has%20two%20(hence%20infinitely%20many)%20distinct%20norm-preserving%20linear%20extensions%20to%20%24L%5E1(%5Cmu)%24**.%0A%0A%20%20%20%20Solution%0A%0A%20%20%20%20Consider%20the%20Lebesgue%20measure%20on%20%24%5B-1%2C%201%5D%24%2C%20and%20consider%20the%20subspace%20%24%24M%20%3D%20%5Clbrace%20f%20%5Cin%20L%5E1(%5B-1%2C1%5D)%20%7C%20f(x)%20%3D%200%20%5Ctext%7B%20for%20%7Dx%5Cin%20%5B-1%2C0%5D%2C%20f(x)%20%5Cge%200%20%5Ctext%7B%20for%20%7Dx%5Cin%20%5B0%2C1%5D%20%5Crbrace.%24%24%20Define%20%24%5CLambda%3A%20M%20%5Cto%20%5Cmathbb%7BC%7D%24%20by%20%24%5CLambda%20f%20%3D%20%5Cint_0%5E1%20f%20dx%24%2C%20then%20this%20is%20a%20linear%20functional.%20%24%5C%7C%20%5CLambda%20%5C%7C%24%20is%20defined%20to%20be%20%24%5Csup%5Clbrace%20%5C%7C%20%5CLambda%20f%20%5C%7C%20%3A%20%5C%7C%20f%20%5C%7C_1%20%3D%201%20%5Crbrace%24%2C%20which%20can%20be%20written%20as%20%24%3D%5Csup%5Clbrace%20%5Cint_0%5E1%20fdx%20%3A%20%5Cint_%7B-1%7D%5E1%20%7C%20f%20%7C%20dx%20%3D%201%20%5Crbrace%24%2C%20notice%20that%20for%20%24f%20%5Cin%20M%24%20we%20have%20%24%5Cint_%7B-1%7D%5E1%20%7C%20f%20%7C%20dx%20%3D%20%5Cint_0%5E1%20f%20dx%24%2C%20thus%20we%20have%20%24%5C%7C%20%5CLambda%20%5C%7C%20%3D%201%24.%0A%0A%20%20%20%20Now%2C%20define%20%24%5CLambda_0%3A%20L%5E1(%5B-1%2C1%5D)%20%5Cto%20%5Cmathbb%7BC%7D%24%20as%20%24%5CLambda_0%20f%20%3D%20%5Cint_0%5E1%20f%20dx%24%2C%20it%20is%20easy%20to%20be%20seen%20as%20an%20extension%20of%20%24%5CLambda%24%2C%20and%20%24%5C%7C%20%5CLambda_0%20%5C%7C%20%3A%3D%20%5Csup%5Clbrace%20%5C%7C%20%5CLambda_0%20f%20%5C%7C%3A%20%5C%7C%20f%20%5C%7C_1%20%3D%201%20%5Crbrace%24%20is%20still%20%241%24%2C%20so%20it%20is%20norm-preserving.%0A%0A%20%20%20%20Define%20%24%5CLambda_1%20%3A%20L%5E1(%5B-1%2C1%5D)%20%5Cto%20%5Cmathbb%7BC%7D%24%20as%20%24%5CLambda_1%20f%20%3D%20%5Cint_%7B-1%7D%5E1%20f%20dx%24%2C%20by%20our%20assumption%20of%20%24M%24%2C%20we%20should%20see%20that%20%24%5CLambda_1%20f%20%3D%20%5CLambda%20f%24%20for%20%24f%20%5Cin%20M%24.%20Also%20%24%5C%7C%20%5CLambda_1%20%5C%7C%20%3A%3D%20%5Csup%5Clbrace%20%5C%7C%20%5CLambda_1%20f%20%5C%7C%3A%20%5C%7C%20f%20%5C%7C_1%20%3D%201%20%5Crbrace%24%20%24%3D%5Csup%5Clbrace%20%7C%20%5Cint_%7B-1%7D%5E1%20f%20dx%20%7C%20%3A%20%5Cint_%7B-1%7D%5E1%20%7C%20f%20%7C%20dx%20%3D%201%20%5Crbrace%24.%20Since%20we%20know%20that%20%24%7C%20%5Cint_Xfdx%20%7C%20%5Cle%20%5Cint_X%7C%20f%20%7C%20dx%24%20is%20always%20true%2C%20yet%20%24f%20%5Cequiv%201%2F2%24%20is%20an%20example%20that%20%24%7C%20%5Cint_%7B-1%7D%5E1%20f%20dx%20%7C%24%20is%20indeed%20%241%24%2C%20we%20have%20that%20%24%5C%7C%20%5CLambda_1%20%5C%7C%20%3D%201%24.%0A%0A%20%20%20%20Of%20course%20%24%5CLambda_0%20%5Cne%20%5CLambda_1%24%2C%20and%20we%20have%20two%20desired%20extensions.%20In%20fact%2C%20for%20any%20%24a%20%5Cin%20%5B0%2C1%5D%24%20the%20functional%20%24%5CLambda_a%20f%20%3D%20%5Cint_%7B-a%7D%5E1%20f%20dx%24%20is%20a%20different%20extension%2C%20so%20we%20have%20infinite%20many%20of%20them.%0A%0A%20%20%20%20%23%23%20Exercise%208%0A%0A%20%20%20%20**Let%20%24X%24%20be%20a%20normed%20linear%20space%2C%20and%20let%20%24X%5E%7B*%7D%24%20be%20its%20dual%20space%20with%20the%20norm%20%24%5C%7C%20f%20%5C%7C%20%3D%20%5Csup%5Clbrace%20%7C%20f(x)%20%7C%20%3A%20%5C%7C%20x%20%5C%7C%20%5Cle%201%20%5Crbrace%24**.%0A%0A%20%20%20%20%23%23%23%20Part%201%0A%0A%20%20%20%20**Prove%20that%20%24X%5E%7B*%7D%24%20is%20a%20Banach%20space**.%0A%0A%20%20%20%20Proof%0A%0A%20%20%20%20For%20a%20reminder%2C%20%24X%5E%7B*%7D%24%20is%20defined%20as%20the%20collection%20of%20all%20bounded%20linear%20functionals%20on%20%24X%24.%20Follow%20the%20definition%20of%20Banach%20space%3A%0A%0A%20%20%20%201.%20If%20%24%5CLambda%2C%5CTheta%20%5Cin%20X%5E%7B*%7D%24%2C%20then%20%24%7C%20%5CLambda%20x%20%2B%20%5CTheta%20x%20%7C%20%5Cle%20%7C%20%5CLambda%20x%20%7C%20%2B%20%7C%20%5CTheta%20x%20%7C%24%20for%20all%20%24x%24%2C%20thus%20%24%5C%7C%20%5CLambda%20%2B%20%5CTheta%20%5C%7C%20%5Cle%20%5C%7C%20%5CLambda%20%5C%7C%20%2B%20%5C%7C%20%5CTheta%20%5C%7C%24%3B%0A%20%20%20%202.%20If%20%24%5Calpha%24%20is%20a%20scalar%20and%20%24%5CLambda%20%5Cin%20X%5E%7B*%7D%24%2C%20then%20%24%7C%20%5Calpha%20%5CLambda%20x%20%7C%20%3D%20%7C%20%5Calpha%20%5C%7C%20%5CLambda%20x%20%7C%24%20for%20all%20%24x%24%2C%20thus%20%24%5C%7C%20%5Calpha%20%5CLambda%20%5C%7C%20%3D%20%7C%20%5Calpha%20%7C%20%5C%7C%20%5CLambda%20%5C%7C%24%3B%201%20and%202%20also%20shows%20this%20is%20a%20vector%20space%3B%0A%20%20%20%203.%20If%20%24%5C%7C%20%5CLambda%20%5C%7C%20%3D%200%24%2C%20then%20by%20definition%20%24%5C%7C%20%5CLambda%20x%20%5C%7C%24%20%24%3D%7C%20%5CLambda%20x%20%7C%20%5Cle%20%5C%7C%20%5CLambda%20%5C%7C%20%5C%7C%20x%20%5C%7C%24%20for%20all%20%24x%24%2C%20so%20%24%7C%20%5CLambda%20x%20%7C%20%5Cequiv%200%24%20and%20%24%5CLambda%20%5Cequiv%200%24%3B%0A%20%20%20%204.%20Suppose%20%24%5Clbrace%20%5CLambda_n%20%5Crbrace%20%5Csubset%20X%5E%7B*%7D%24%20is%20a%20Cauchy%20sequence%20and%20fix%20some%20%24%5Cvarepsilon0%24.%20By%20definition%20there%20is%20some%20%24N%24%20such%20that%20for%20%24n%2Cm%20%3E%20N%24%20we%20have%20%24%5C%7C%20%5CLambda_n%20-%20%5CLambda_m%20%5C%7C%20%3C%20%5Cvarepsilon%24.%20Use%20the%20definition%20of%20norm%20of%20a%20linear%20transformation%2C%20we%20have%20that%20%24%5C%7C%20(%5CLambda_n%20-%5CLambda_m)(x)%20%5C%7C%24%20%24%3D%20%7C%20(%5CLambda_n%20-%20%5CLambda_m)%20x%7C%24%20%24%5Cle%20%5C%7C%20%5CLambda_n%20-%20%5CLambda_m%20%5C%7C%5C%7C%20x%20%5C%7C%24%20%24%3C%5Cvarepsilon%20%5C%7C%20x%20%5C%7C%24%20for%20any%20%24x%24.%20Thus%20if%20we%20fix%20an%20%24x%24%2C%20then%20%24%5Clbrace%20%5CLambda_n%20x%20%5Crbrace%24%20is%20a%20Cauchy%20sequence%20in%20%24%5Cmathbb%7BC%7D%24%2C%20thus%20is%20convergent%2C%20denote%20the%20limit%20as%20%24%5CLambda%20x%24%2C%20and%20do%20this%20for%20all%20%24x%20%5Cin%20X%24.%20We%20want%20to%20show%20that%20%24%5Clbrace%20%5CLambda_n%20%5Crbrace%24%20converges%20to%20%24%5CLambda%24%20and%20%24%5CLambda%20%5Cin%20X%5E%7B*%7D%24%3B%0A%20%20%20%20%091.%20Let%20%24a%2Cb%24%20be%20scalars%20and%20%24x%2Cy%5Cin%20X%24%2C%20then%20%24%5CLambda(ax%2Bby)%20%3A%3D%20%5Clim%20%5CLambda_n(ax%2Bby)%24%20%24%3D%20%5Clim%20a%5CLambda_n%20x%20%2B%20b%5CLambda_n%20y%24%20%24%3D%20a%5CLambda%20x%20%2B%20b%5CLambda%20y%24%2C%20thus%20%24%5CLambda%24%20is%20a%20linear%20functional%3B%0A%20%20%20%20%092.%20Every%20Cauchy%20sequence%20is%20bounded%2C%20thus%20%24%5Clbrace%20%5CLambda_n%20%5Crbrace%24%20is%20bounded.%20Say%20%24%5C%7C%20%5CLambda_n%20%5C%7C%20%5Cle%20M%24%2C%20then%20we%20can%20write%20%24%7C%20%5CLambda%20x%20%7C%20%5Cle%20%7C%20%5CLambda%20x%20-%20%5CLambda_n%20x%20%7C%20%2B%20%7C%20%5CLambda_n%20x%20%7C%24%20(remember%20%24%7C%20(%5CLambda_n%20-%20%5CLambda_m)%20x%7C%20%3C%20%5Cvarepsilon%20%5C%7C%20x%20%5C%7C%24%2C%20taking%20%24m%20%5Cto%20%5Cinfty%24%20now)%20%24%5Cle%20%5Cvarepsilon%20%5C%7C%20x%20%5C%7C%20%2B%20%5C%7C%20%5CLambda_n%20%5C%7C%5C%7C%20x%20%5C%7C%24%20%24%5Cle%20(M%2B%5Cvarepsilon)%5C%7C%20x%20%5C%7C%24%2C%20so%20%24%5C%7C%20%5CLambda%20%5C%7C%20%3D%20%5Csup%20%5Clbrace%20%7C%20%5CLambda%20x%20%7C%20%3A%20%5C%7C%20x%20%5C%7C%20%3D%201%20%5Crbrace%24%20%24%5Cle%20M%2B%5Cvarepsilon%24%20is%20bounded%3B%20so%20%24%5CLambda%24%20is%20a%20bounded%20linear%20functional%20thus%20by%20definition%20is%20in%20%24X%5E%7B*%7D%24%3B%0A%20%20%20%20%093.%20In%20the%20above%20steps%20we%20used%20%24%7C%20%5CLambda%20x%20-%20%5CLambda_n%20x%20%7C%20%5Cle%20%5Cvarepsilon%20%5C%7C%20x%20%5C%7C%24%20for%20%24n%20%3E%20N%24%2C%20so%20%24%5C%7C%20%5CLambda%20-%20%5CLambda_n%20%5C%7C%20%3A%3D%24%20%24%5Csup%20%5Clbrace%20%7C%20%5CLambda%20x%20-%20%5CLambda_n%20x%20%7C%20%3A%20%5C%7C%20x%20%5C%7C%20%3D%201%5Crbrace%24%20%24%5Cle%20%5Cvarepsilon%24%2C%20that%20we%20have%20convergence%20as%20desired.%0A%0A%20%20%20%20This%20shows%20%24X%5E%7B*%7D%24%20is%20complete%20and%20we%20finished%20the%20proof.%0A%0A%20%20%20%20%23%23%23%20Part%202%0A%0A%20%20%20%20**Prove%20that%20for%20each%20%24x%20%5Cin%20X%24%2C%20the%20mapping%20%24f%5Cto%20f(x)%24%20is%20a%20bounded%20linear%20functional%20on%20%24X%5E%7B*%7D%24%20of%20norm%20%24%5C%7C%20x%20%5C%7C%24.%20This%20gives%20a%20natural%20imbedding%20of%20%24X%24%20in%20its%20'second%20dual'%20%24X%5E%7B**%7D%24%2C%20the%20dual%20space%20of%20%24X%5E%7B*%7D%24**.%0A%0A%20%20%20%20Proof%0A%0A%20%20%20%20Let%20us%20call%20this%20map%20%24%5Cvarphi%24.%20As%20per%20defined%2C%20for%20a%20fixed%20%24x%24%2C%20%24%5Cvarphi%3A%20X%5E%7B*%7D%20%5Cto%20%5Cmathbb%7BC%7D%24%20is%20given%20by%20%24%5Cvarphi(f)%20%3D%20f(x)%24.%0A%0A%20%20%20%20Suppose%20%24%5CLambda%2C%20%5CTheta%20%5Cin%20X%5E%7B*%7D%24%20and%20%24a%2C%20b%20%5Cin%20%5Cmathbb%7BC%7D%24%2C%20then%20%24%5Cvarphi(a%5CLambda%20%2B%20b%20%5CTheta)%24%20%24%3D(a%5CLambda%20%2B%20b%5CTheta)x%24%20%24%3Da%5CLambda%20x%20%2B%20b%20%5CTheta%20x%24%20%24%3Da%5Cvarphi(%5CLambda)%20%2B%20b%5Cvarphi(%5CTheta)%24%2C%20thus%20%24%5Cvarphi%24%20is%20a%20linear%20functional.%0A%0A%20%20%20%20By%20definition%20of%20the%20norm%2C%20we%20have%20%24%7C%20%5Cvarphi(f)%20%7C%20%20%3D%20%7C%20f(x)%20%7C%24%20%24%5Cle%20%5C%7C%20f%20%5C%7C%20%5C%7C%20x%20%5C%7C%24%2C%20so%20%24%5C%7C%20%5Cvarphi%20%5C%7C%20%3A%3D%24%20%24%5Csup%20%5Clbrace%20%7C%20%5Cvarphi(f)%20%7C%20%3A%20%5C%7C%20f%20%5C%7C%20%3D%201%20%5Crbrace%24%20%24%5Cle%20%5C%7C%20x%20%5C%7C%24.%20Since%20%24x%24%20is%20fixed%2C%20we%20see%20%24%5Cvarphi%24%20is%20a%20bounded%20linear%20functional%20as%20desired.%0A%0A%20%20%20%20To%20finish%20the%20proof%20we%20need%20the%20other%20way%20of%20the%20inequality%3A%0A%0A%20%20%20%201.%20If%20%24x%20%5Cne%200%24%2C%20Theorem%205.20%20says%20there%20is%20a%20bounded%20linear%20functional%20%24f%20%5Cin%20X%5E%7B*%7D%24%20such%20that%20%24f(x)%20%3D%20%5C%7C%20x%20%5C%7C%24%20and%20%24%5C%7C%20f%20%5C%7C%20%3D%201%24%2C%20thus%20for%20this%20%24f%24%20we%20have%20%24%5C%7C%20%5Cvarphi(f)%20%5C%7C%24%20%24%3D%20%5C%7C%20f(x)%20%5C%7C%24%20%24%3D%20%5C%7C%20x%20%5C%7C%24%20thus%20%24%5C%7C%20%5Cvarphi%20%5C%7C%20%3D%20%5C%7C%20x%20%5C%7C%24%20(because%20it%20is%20a%20achieved%20upper%20bounded)%3B%0A%20%20%20%202.%20If%20%24x%20%3D%200%24%20then%20it%20is%20trivial%2C%20as%20that%20%24%5Cvarphi(f)%20%3D%20f(0)%20%3D%200%24%20for%20any%20(linear)%20%24f%24%20so%20%24%5C%7C%20%5Cvarphi%20%5C%7C%20%3D%20%5C%7C%20x%20%5C%7C%20%3D%200%24%20is%20still%20true.%0A%0A%20%20%20%20%23%23%23%20Part%203%0A%0A%20%20%20%20**Prove%20that%20%24%5Clbrace%20%5C%7C%20x_n%20%5C%7C%20%5Crbrace%24%20is%20bounded%20if%20%24%5Clbrace%20x_n%20%5Crbrace%24%20is%20a%20sequence%20in%20%24X%24%20such%20that%20%24%5Clbrace%20f(x_n)%20%5Crbrace%24%20is%20bounded%20for%20every%20%24f%20%5Cin%20X%5E%7B*%7D%24**.%0A%0A%20%20%20%20Proof%0A%0A%20%20%20%20We%20have%20that%20%24X%5E%7B*%7D%24%20is%20a%20Banach%20space%20and%20%24%5Cmathbb%7BC%7D%24%20is%20a%20normed%20linear%20space.%20Let%20%24%5Clbrace%20%5CLambda_%7Bx_n%7D%20%5Crbrace%24%20be%20the%20collection%20of%20bounded%20linear%20functional%20as%20defined%20in%20part%202%2C%20%24%5CLambda_%7Bx_i%7D(f)%20%3D%20f(x_i)%24%20for%20each%20%24x_i%24.%20We%20apply%20the%20Banach-Steinhaus%20Theorem%3A%0A%0A%20%20%20%20By%20assumption%2C%20for%20each%20%24f%24%20we%20have%20%24%7C%20f(x_n)%20%7C%20%20%3D%20%7C%20%5CLambda_%7Bx_n%7D(f)%20%7C%24%20is%20bounded.%20So%20it%20is%20not%20true%20that%20there%20is%20some%20%24f%24%20so%20that%20%24%5Csup%5Climits_%7Bx_n%7D%20%7C%20%5CLambda_%7Bx_n%7D(f)%20%7C%20%3D%20%5Cinfty%24%2C%20so%20we%20must%20have%20that%20%24%5Clbrace%20%5C%7C%20%5CLambda_%7Bx_n%7D%20%5C%7C%20%5Crbrace%24%20is%20bounded%2C%20but%20as%20showed%20in%20part%202%2C%20this%20is%20the%20same%20sequence%20as%20%24%5Clbrace%20%5C%7C%20x_n%20%5C%7C%20%5Crbrace%24.%0A%0A%20%20%20%20%23%23%20Exercise%209%0A%0A%20%20%20%20**Let%20%24c_0%2C%20%5Cell%5E1%2C%20%5Cell%5E%7B%5Cinfty%7D%24%20be%20the%20Banach%20spaces%20consisting%20of%20all%20complex%20sequences%20%24x%20%3D%20%5Clbrace%20%5Cxi_i%20%5Crbrace%24%20defined%20as%20follows**%3A%0A%0A%20%20%20%201.%20**%24x%20%5Cin%20%5Cell%5E1%24%20if%20and%20only%20if%20%24%5C%7C%20x%20%5C%7C_1%20%3D%20%5Csum%20%7C%20%5Cxi_i%20%7C%20%3C%20%5Cinfty%24**%3B%0A%20%20%20%202.%20**%24x%20%5Cin%20%5Cell%5E%7B%5Cinfty%7D%24%20if%20and%20only%20if%20%24%5C%7C%20x%20%5C%7C_%7B%5Cinfty%7D%20%3D%20%5Csup%20%7C%20%5Cxi_i%20%7C%20%3C%20%5Cinfty%24**%3B%0A%20%20%20%203.%20**%24x%20%5Cin%20c_0%24%20if%20and%20only%20if%20%24%5Cxi_i%20%5Cto%200%24**.%0A%0A%20%20%20%20**Prove%20the%20following%20four%20statements**%3A%0A%0A%20%20%20%201.%20**If%20%24y%20%3D%20%5Clbrace%20%5Ceta_i%20%5Crbrace%20%5Cin%20%5Cell%5E1%24%20and%20%24%5CLambda%20x%20%3D%20%5Csum%20%5Cxi_i%20%5Ceta_i%24%20for%20every%20%24x%20%5Cin%20c_0%24%2C%20then%20%24%5CLambda%24%20is%20a%20bounded%20linear%20functional%20on%20%24c_0%24%20and%20%24%5C%7C%20%5CLambda%20%5C%7C%20%3D%20%5C%7C%20y%20%5C%7C_1%24.%20Moreover%2C%20every%20%24%5CLambda%20%5Cin%20(c_0)%5E%7B*%7D%24%20is%20obtained%20in%20this%20way.%20In%20brief%2C%20%24(c_0)%5E%7B*%7D%20%3D%20%5Cell%5E1%24**%3B%0A%20%20%20%202.%20**In%20the%20same%20sense%2C%20%24(%5Cell%5E1)%5E%7B*%7D%20%3D%20%5Cell%5E%7B%5Cinfty%7D%24**%3B%0A%20%20%20%203.%20**Every%20%24y%20%5Cin%20%5Cell%5E1%24%20induces%20a%20bounded%20linear%20functional%20on%20%24%5Cell%5E%7B%5Cinfty%7D%24%2C%20as%20in%20part%201.%20However%2C%20this%20does%20not%20give%20all%20of%20%24(%5Cell%5E%7B%5Cinfty%7D)%5E%7B*%7D%24%2C%20since%20%24(%5Cell%5E%7B%5Cinfty%7D)%5E%7B*%7D%24%20contains%20non-trivial%20functionals%20that%20vanishes%20on%20all%20of%20%24c_0%24**%3B%0A%20%20%20%204.%20**%24c_0%24%20and%20%24l_1%24%20are%20separable%20but%20%24%5Cell%5E%7B%5Cinfty%7D%24%20is%20not**.%0A%0A%20%20%20%20Proof%0A%0A%20%20%20%20%23%23%23%20Part%201%0A%0A%20%20%20%20Suppose%20%24x%20%3D%20%5Clbrace%20%5Cxi_i%20%5Crbrace%2C%20z%20%3D%20%5Clbrace%20%5Ctheta_i%20%5Crbrace%20%5Cin%20c_0%24%20and%20%24a%2Cb%20%5Cin%20%5Cmathbb%7BC%7D%24%2C%20then%20%24%5CLambda(ax%2Bbz)%20%3D%20%5Csum%5Climits_%7Bi%20%3D%201%7D%5E%7B%5Cinfty%7D(a%5Cxi_i%20%2B%20b%5Ctheta_i)%5Ceta_i%24%20%24%3Da%5Csum%5Climits%5Cxi_i%5Ceta_i%20%2B%20b%20%5Csum%5Climits%5Ctheta_i%5Ceta_i%24%20%24%3Da%5CLambda%20x%20%2B%20b%20%5CLambda%20z%24%2C%20so%20%24%5CLambda%24%20is%20linear.%20For%20any%20%24x%24%2C%20%24%7C%20%5CLambda%20x%20%7C%24%20%24%3D%7C%20%5Csum%5Climits%5Cxi_i%5Ceta_i%20%7C%24%20%24%5Cle%5Csum%5Climits%7C%20%5Cxi_i%20%5C%7C%20%5Ceta_i%20%7C%24%20%24%5Cle%20%5Csup%20%7C%20%5Cxi_i%20%7C%20%5Csum%20%7C%20%5Ceta_i%20%7C%24%20%24%3D%5C%7C%20x%20%5C%7C_%7B%5Cinfty%7D%5C%7C%20y%20%5C%7C_1%24%20(this%20also%20shows%20the%20series%20converges%20absolutely%20so%20%24%5CLambda%24%20is%20well-defined)%2C%20so%20%24%5C%7C%20%5CLambda%20%5C%7C%20%3D%20%5Csup%20%5Clbrace%20%7C%20%5CLambda%20x%20%7C%3A%20x%20%5Cin%20c_0%2C%20%5C%7C%20x%20%5C%7C_%7B%5Cinfty%7D%20%3D1%5Crbrace%24%20%24%5Cle%20%5C%7C%20y%20%5C%7C_1%24.%0A%0A%20%20%20%20For%20the%20other%20direction%2C%20for%20a%20given%20%24y%20%3D%20%5Clbrace%20%5Ceta_i%20%5Crbrace%24%20and%20%24n%20%5Cin%20%5Cmathbb%7BN%7D%24%2C%20take%20%24x_n%20%3D%20%5Clbrace%20%5Cxi_%7Bn%2Ci%7D%20%5Crbrace%24%20where%20%24%5Cxi_%7Bn%2Ci%7D%20%3D%20%5Cfrac%7B%5Coverline%7B%5Ceta_i%7D%7D%7B%7C%20%5Ceta_i%20%7C%7D%24%20if%20%24i%20%5Cle%20n%24%2C%20and%20%240%24%20otherwise.%20Each%20%24x_n%24%20is%20by%20construction%20in%20%24c_0%24%20because%20it%20is%20eventually-zero%2C%20and%20clearly%20%24%5C%7C%20x_n%20%5C%7C_%7B%5Cinfty%7D%20%3D%201%24.%20Now%20for%20this%20particular%20choice%20of%20%24x%24%2C%20%24%5C%7C%20%5CLambda%20%5C%7C%24%20%24%5Cge%20%7C%20%5CLambda%20x_n%20%7C%24%20%24%3D%20%5Csum%5Climits_%7Bi%20%3D%201%7D%5E%7B%5Cinfty%7D%5Cxi_%7Bn%2Ci%7D%5Ceta_i%24%20%24%3D%5Csum%5Climits_%7Bi%3D1%7D%5En%7C%20%5Ceta_i%20%7C%24%20(because%20%24%5Ceta_i%5Coverline%7B%5Ceta_i%7D%20%3D%20%7C%5Ceta_i%7C%5E2%24)%20for%20each%20%24n%24%3B%20take%20%24n%20%5Cto%20%5Cinfty%24%20then%20%24%5C%7C%20%5CLambda%20%5C%7C%20%5Cge%20%5Csum%20%7C%20%5Ceta_i%20%7C%24%20%24%3D%5C%7C%20y%20%5C%7C_1%24.%20So%20we%20have%20equality.%0A%0A%20%20%20%20Define%20a%20function%20%24%5Cvarphi%3A%5Cell%5E1%20%5Cto%20(c_0)%5E%7B*%7D%24%20by%20%24%5Cvarphi(y)%20%3D%20%5CLambda%24%2C%20as%20above%2C%20it%20is%20well-defined.%20We%20want%20to%20show%20it%20is%20a%20bijection%3A%0A%0A%20%20%20%201.%20It%20is%20surjective%3A%20let%20%24%5CLambda%20%5Cin%20(c_0)%5E%7B*%7D%24%2C%20we%20want%20to%20find%20%24y%20%5Cin%20%5Cell%5E1%24%20such%20that%20%24%5Cvarphi(y)%20%3D%20%5CLambda%24.%20Consider%20the%20sequence%20%24%5Clbrace%20e_i%20%5Crbrace%24%20the%20unit%20vector%20basis%20of%20%24c_0%24%2C%20define%20%24y%20%3D%20%5Clbrace%20%5Ceta_i%20%5Crbrace%24%20by%20%24%5Ceta_i%20%3D%20%5CLambda%20e_i%24%20so%20that%20%24%5Cvarphi(y)%24%20indeed%20maps%20each%20%24%5Clbrace%20%5Cxi_i%20%5Crbrace%20%5Cin%20c_0%24%20to%20%24%5Csum%5Climits%5Cxi_i%5Ceta_i%24%20%24%3D%5CLambda%5Csum%5Climits%5Cxi_ie_i%24%20%24%3D%5CLambda%20x%24%2C%20we%20will%20show%20%24y%20%5Cin%20%5Cell%5E1%24.%20For%20each%20%24n%24%2C%20define%20%24x_n%24%20as%20before%20(%24%5Cxi_%7Bn%2Ci%7D%20%3D%20%5Cfrac%7B%5Coverline%7B%5Ceta_i%7D%7D%7B%7C%20%5Ceta_i%20%7C%7D%24%20for%20%24i%20%5Cle%20n%24%2C%20%240%24%20otherwise)%2C%20then%20%24%5CLambda%20x_n%24%20%24%3D%5CLambda%20%5Csum%5Climits%5Cxi_%7Bn%2Ci%7De_i%20%3D%20%5Csum%5Climits%5Cxi_%7Bn%2Ci%7D%5CLambda%20e_i%24%20%24%3D%5Csum%5Climits_%7Bi%3D1%7D%5En%5Cxi_%7Bn%2Ci%7D%5Ceta_i%20%3D%20%5Csum%5Climits_%7Bi%3D1%7D%5En%7C%20%5Ceta_i%20%7C%24.%20On%20the%20other%20hand%20%24%7C%20%5CLambda%20x_n%20%7C%24%20%24%5Cle%20%5C%7C%20%5CLambda%20%5C%7C%5C%7C%20x_n%20%5C%7C_%7B%5Cinfty%7D%20%3D%20%5C%7C%20%5CLambda%20%5C%7C%24%2C%20if%20we%20take%20%24n%20%5Cto%20%5Cinfty%24%20then%20%24%5C%7C%20y%20%5C%7C_1%20%5Cle%20%5C%7C%20%5CLambda%20%5C%7C%24%20which%20is%20by%20assumption%20bounded%2C%20so%20%24y%20%5Cin%20%5Cell%5E1%24%3B%0A%20%20%20%202.%20It%20is%20injective%3A%20Suppose%20%24y%20%3D%20%5Clbrace%20%5Ceta_i%20%5Crbrace%24%20and%20%24z%20%3D%20%5Clbrace%20%5Ctheta_i%20%5Crbrace%24%20are%20both%20%24%5Cell%5E1%24%2C%20and%20%24%5Cvarphi(y)%20%3D%20%5Cvarphi(z)%24.%20To%20avoid%20confusion%2C%20let%20us%20add%20a%20subscript%2C%20denoting%20%24%5CLambda_%7B*%7D%24%20be%20the%20image%20of%20%24*%24%20under%20%24%5Cvarphi%24.%20Then%20%24%5Cvarphi(y%20-%20z)%24%20%24%3D%20%5CLambda_%7By-z%7D%24%20where%20%24%5CLambda_%7By-z%7D(x)%20%3D%24%20%24%5Csum%5Climits%5Cxi_i(%5Ceta_i%20-%20%5Ctheta_i)%24%20%24%3D%20%5Csum%5Climits%5Cxi_i%5Ceta_i%20%20-%20%5Csum%5Climits%5Cxi_i%5Ctheta_i%24%20%24%3D%5CLambda_y%20x%20-%20%5CLambda_z%20x%24%20%24%3D0%24%2C%20thus%20%24%5Cvarphi(y-z)%24%20is%20the%20trivial%20linear%20functional%20sending%20everything%20in%20%24c_0%24%20to%20%240%24%2C%20in%20particular%20we%20must%20have%20%24%5Ceta_i%20-%20%5Ctheta_i%20%3D%200%24%20for%20all%20%24i%24%2C%20thus%20%24y%20%3D%20z%24.%0A%0A%20%20%20%20To%20conclude%20we%20should%20also%20show%20this%20is%20an%20isometry%20(i.e.%20isomorphism)%2C%20but%20this%20is%20just%20what%20we%20proved%2C%20as%20%24%5C%7C%20y%20%5C%7C_1%20%3D%20%5C%7C%20%5CLambda%20%5C%7C%20%3D%20%5C%7C%20%5Cvarphi(y)%20%5C%7C%24.%0A%0A%20%20%20%20%23%23%23%20Part%202%0A%0A%20%20%20%20This%20is%20pretty%20much%20the%20same%20idea%20with%20part%201.%20Fix%20%24y%20%3D%20%5Clbrace%20%5Ceta_i%20%5Crbrace%20%5Cin%20%5Cell%5E%7B%5Cinfty%7D%24%2C%20define%20%24%5CLambda%24%20so%20that%20for%20any%20%24x%20%3D%20%5Clbrace%20%5Cxi_i%20%5Crbrace%20%5Cin%20%5Cell%5E1%24%20we%20have%20%24%5CLambda%20x%20%3D%20%5Csum%5Climits%5Cxi_i%5Ceta_i%24%2C%20with%20a%20very%20same%20argument%2C%20we%20see%20each%20such%20series%20converges%20absolutely%2C%20%24%5CLambda%24%20is%20linear%20and%20it%20is%20bounded%2C%20thus%20%24%5CLambda%20%5Cin%20(%5Cell%5E1)%5E%7B*%7D%24%20by%20definition%20(for%20boundedness%20this%20time%20we%20have%20%24%7C%20%5CLambda%20x%20%7C%24%20%24%5Cle%20%5C%7C%20x%20%5C%7C_1%5C%7C%20y%20%5C%7C_%7B%5Cinfty%7D%24%20so%20that%20%24%5C%7C%20%5CLambda%20%5C%7C%24%20%24%5Cle%20%5C%7C%20y%20%5C%7C_%7B%5Cinfty%7D%24%2C%20basically%20we%20interchange%20%241%24%20with%20%24%5Cinfty%24%20as%20it%20in%20part%201).%0A%0A%20%20%20%20The%20next%20step%20is%20to%20show%20%24%5C%7C%20%5CLambda%20%5C%7C%20%3D%20%5C%7C%20y%20%5C%7C_%7B%5Cinfty%7D%24%3A%20consider%20the%20unit%20vector%20basis%20%24%5Clbrace%20e_i%20%5Crbrace%24%20for%20%24%5Cell%5E1%5Cell%5E1%24%2C%20apparently%20%24%7C%20%5Ceta_i%20%7C%20%3D%20%7C%20%5CLambda%20e_i%20%7C%24%20for%20each%20%24i%24%2C%20we%20may%20continue%20to%20write%20%24%5Cle%20%5C%7C%20%5CLambda%20%5C%7C%20%5C%7C%20e%20%5C%7C_i%20%3D%20%5C%7C%20%5CLambda%20%5C%7C%24%2C%20take%20supremum%20each%20side%20we%20have%20%24%5C%7C%20y%20%5C%7C_%7B%5Cinfty%7D%20%3D%20%5Csup%5C%7C%20%5Ceta_i%20%5C%7C%20%5Cle%20%5C%7C%20%5CLambda%20%5C%7C%24.%20So%20we%20have%20both%20direction%20thus%20we%20have%20equality.%0A%0A%20%20%20%20Finally%2C%20we%20define%20function%20%24%5Cvarphi%3A%20%5Cell%5E%7B%5Cinfty%7D%20%5Cto%20(%5Cell%5E1)%5E%7B*%7D%24%20by%20%24%5Cvarphi(y)%20%3D%20%5CLambda%24%20(or%20%24%3D%5CLambda_y%24%20when%20the%20notation%20is%20needed).%20As%20before%2C%20since%20%24%5C%7C%20y%20%5C%7C_%7B%5Cinfty%7D%20%3D%20%5C%7C%20%5CLambda%20%5C%7C%20%3D%20%5C%7C%20%5Cvarphi(y)%20%5C%7C%24%2C%20this%20is%20an%20isometry.%20We%20just%20need%20to%20show%20it%20is%20a%20bijection%3A%0A%0A%20%20%20%201.%20It%20is%20surjective%3A%20let%20%24%5CLambda%20%5Cin%20(l_1)%5E%7B*%7D%24%2C%20we%20want%20to%20find%20%24y%20%5Cin%20%5Cell%5E%7B%5Cinfty%7D%24%20such%20that%20%24%5Cvarphi(y)%20%3D%20%5CLambda%24.%20Consider%20the%20sequence%20%24%5Clbrace%20e_i%20%5Crbrace%24%20the%20unit%20vector%20basis%20of%20%24l_1%24%2C%20define%20%24y%20%3D%20%5Clbrace%20%5Ceta_i%20%5Crbrace%24%20by%20%24%5Ceta_i%20%3D%20%5CLambda%20e_i%24%20so%20that%20%24%5Cvarphi(y)%24%20indeed%20maps%20each%20%24%5Clbrace%20%5Cxi_i%20%5Crbrace%20%5Cin%20l_1%24%20to%20%24%5Csum%5Climits%5Cxi_i%5Ceta_i%24%20%24%3D%5CLambda%5Csum%5Climits%5Cxi_ie_i%24%20%24%3D%5CLambda%20x%24.%20By%20above%2C%20we%20already%20know%20such%20%24y%24%20lies%20in%20%24%5Cell%5E%7B%5Cinfty%7D%24%3B%0A%20%20%20%202.%20It%20is%20injective%3A%20the%20proof%20is%20the%20same%20as%20above.%0A%0A%20%20%20%20%23%23%23%20Part%203%0A%0A%20%20%20%20For%20each%20%24y%20%3D%20%5Clbrace%20%5Ceta_i%20%5Crbrace%20%5Cin%20%5Cell%5E1%24%2C%20define%20%24%5CLambda%24%20as%20before%20that%20%24%5CLambda%20x%20%3D%20%5Csum%5Climits%5Cxi_i%5Ceta_i%24%20for%20%24x%20%3D%20%5Clbrace%20%5Cxi_i%20%5Crbrace%20%5Cin%20%5Cell%5E%7B%5Cinfty%7D%24%2C%20we%20saw%20this%20series%20converges%20absolutely%2C%20and%20it%20is%20a%20bounded%20linear%20functional%2C%20so%20%24y%24%20induces%20a%20bounded%20linear%20functional%20on%20%24%5Cell%5E%7B%5Cinfty%7D%24.%0A%0A%20%20%20%20Now%20apparently%20%24c_0%20%5Csubsetneq%20%5Cell%5E%7B%5Cinfty%7D%24%2C%20because%20any%20convergent%20series%20must%20be%20bounded%2C%20but%20for%20example%20a%20series%20%24x%20%3D%20%5Clbrace%201%2C%201%2C%20%5Cdots%20%5Crbrace%24%20is%20%24%5Cell%5E%7B%5Cinfty%7D%24%20yet%20not%20%24c_0%24.%20Moreover%2C%20%24c_0%24%20is%20closed%3A%20let%20%24x%20%3D%20%5Clbrace%20%5Cxi_i%20%5Crbrace%20%5Cin%20%5Cell%5E%7B%5Cinfty%7D%20-%20c_0%24%2C%20then%20by%20definition%20%24%5Clim%20%5Cxi_i%20%5Cne%200%24%2C%20meaning%20that%20there%20is%20an%20%24%5Cvarepsilon0%24%20so%20that%20for%20any%20%24N%24%2C%20there%20is%20%24n%3EN%24%20with%20%24%7C%20%5Cxi_n%20%7C%20%5Cge%20%5Cvarepsilon%24%3B%20take%20any%20sequence%20%24y%20%3D%20%5Clbrace%20%5Ceta_i%20%5Crbrace%24%20such%20that%20%24%5C%7C%20x-y%20%5C%7C_%7B%5Cinfty%7D%20%3C%20%5Cvarepsilon%2F2%24%2C%20meaning%20that%20%24%5Csup%20%7C%20%5Cxi_n%20-%20%5Ceta_n%20%7C%20%3C%20%5Cvarepsilon%2F2%24%20(and%20%24y%20%5Cin%20%5Cell%5E%7B%5Cinfty%7D%24)%2C%20then%20by%20triangle%20inequality%20we%20must%20have%20for%20any%20%24N%24%2C%20there%20is%20%24n%3EN%24%20with%20%24%7C%20%5Ceta_i%20%7C%20%5Cge%20%5Cvarepsilon%2F2%24%2C%20meaning%20that%20%24y%24%20is%20also%20not%20%24c_0%24%2C%20so%20%24%5Cell%5E%7B%5Cinfty%7D%20-%20c_0%24%20is%20open%2C%20thus%20%24c_0%24%20is%20closed.%20So%20by%20Theorem%205.19%2C%20apply%20to%20%24c_0%20%3D%20M%20%3D%20%5Coverline%7BM%7D%24%2C%20%24%5Cell%5E%7B%5Cinfty%7D%20%3D%20X%24%20and%20%24x_0%20%5Cin%20%5Cell%5E%7B%5Cinfty%7D-c_0%24%2C%20we%20can%20find%20some%20bounded%20linear%20functional%20%24f%24%20on%20%24%5Cell%5E%7B%5Cinfty%7D%24%20such%20that%20%24f(x)%20%3D%200%24%20for%20all%20%24x%20%5Cin%20c_0%24%20but%20%24f(x_0)%20%5Cne%200%24.%0A%0A%20%20%20%20If%20for%20contrary%20%24y%20%5Cin%20%5Cell%5E1%24%20gives%20all%20of%20%24%5CLambda%20%5Cin%20(%5Cell%5E%7B%5Cinfty%7D)%5E%7B*%7D%24%3A%20by%20part%201%2C%20those%20%24%5CLambda%24%20are%20%24(c_0)%5E%7B*%7D%24%2C%20and%20the%20above%20%24f%24%20is%20such%20a%20functional%2C%20so%20%24f%20%5Cin%20(c_0)%5E%7B*%7D%24%20and%20it%20is%20trivial%20on%20%24c_0%24%3B%20this%20%24f%24%20is%20also%20in%20%24(%5Cell%5E%7B%5Cinfty%7D)%5E%7B*%7D%24%20in%20the%20first%20place%2C%20so%20it%20is%20of%20one-to-one%20correspondence%20with%20the%20trivial%20linear%20functional%20on%20%24c_0%24%2C%20so%20it%20must%20be%20the%20trivial%20functional%20on%20%24%5Cell%5E%7B%5Cinfty%7D%24%2C%20but%20it%20is%20not%2C%20as%20%24f(x_0)%20%5Cne%200%24.%0A%0A%20%20%20%20%23%23%23%20Part%204%0A%0A%20%20%20%201.%20%24c_0%24%3A%20consider%20the%20subset%20%24S_n%20%3D%20%5Clbrace%20%5Clbrace%20%5Cxi_i%20%5Crbrace%3A%20%5Cxi_i%20%5Cin%20%5Cmathbb%7BQ%7D%2C%20%5Cxi_i%20%3D%200%5Ctext%7B%20if%20%7Di%3En%20%5Crbrace%24%20and%20%24S%20%3D%20%5Ccup%20S_n%24%2C%20then%20%24S%24%20is%20countable%2C%20and%20it%20is%20easy%20to%20see%20it%20is%20dense%20in%20%24c_0%24%20(with%20supremum%20norm)%3B%0A%20%20%20%202.%20%24%5Cell%5E1%24%3A%20we%20use%20the%20same%20%24S%24%2C%20and%20remember%20that%20%24%5Csum%20%7C%20%5Cxi_i%20%7C%20%3C%20%5Cinfty%24%20means%20that%20for%20a%20fixed%20%24%5Cvarepsilon%24%20there%20exists%20%24N%24%20so%20that%20%24%5Csum%5Climits_%7Bi%3DN%7D%5E%7B%5Cinfty%7D%7C%20%5Cxi_i%20%7C%20%3C%20%5Cvarepsilon%2F2%24.%20Then%20for%20any%20%24x%20%5Cin%20%5Cell%5E1%24%20and%20any%20%24%5Cvarepsilon%3E0%24%2C%20just%20pick%20%24N%24%20as%20discussed%2C%20then%20choose%20%24y%20%3D%20%5Clbrace%20%5Ceta_i%20%5Crbrace%20%5Cin%20S_N%24%20so%20that%20%24%5Csum%5Climits_%7Bi%3D1%7D%5EN%7C%20%5Cxi_i%20-%20%5Ceta_i%20%7C%20%5Cle%20%5Cvarepsilon%2F2%24%20(it%20is%20certainly%20achievable%20because%20it%20is%20finite)%2C%20and%20thus%20the%20%24%5Cell%5E1%5Cell%5E1%24%20distance%20between%20%24x%24%20and%20%24y%24%20is%20no%20larger%20than%20%24%5Cvarepsilon%24%2C%20which%20proves%20separability%3B%0A%20%20%20%203.%20%24%5Cell%5E%7B%5Cinfty%7D%24%3A%20consider%20the%20set%20%24U%20%3D%20%5Clbrace%20%5Clbrace%20%5Cxi_i%20%5Crbrace%3A%20%5Cxi_i%20%5Cin%20%5Clbrace%200%2C1%20%5Crbrace%5Crbrace%24%2C%20it%20is%20well-known%20that%20%24U%24%20is%20uncountable%3B%20furthermore%2C%20for%20any%20%24x%5Cne%20y%20%5Cin%20U%24%2C%20it%20is%20easy%20to%20see%20%24%5C%7C%20x-y%20%5C%7C_%7B%5Cinfty%7D%20%3D%201%24%20is%20a%20fixed%20positive%20number.%20In%20particular%20it%20means%20%24%5Clbrace%20x%20%5Crbrace%24%20is%20an%20open%20set%20for%20any%20%24x%20%5Cin%20U%24%2C%20so%20any%20dense%20subset%20must%20contain%20all%20%24x%24%20thus%20the%20entire%20%24U%24%2C%20which%20is%20not%20countable%2C%20so%20the%20space%20is%20not%20separable.%0A%0A%20%20%20%20%23%23%20Exercise%2010%0A%0A%20%20%20%20**If%20%24%5Csum%20%5Calpha_i%5Cxi_i%24%20converges%20for%20every%20sequences%20%24%5Clbrace%20%5Cxi_i%20%5Crbrace%24%20such%20that%20%24%5Cxi_i%20%5Cto%200%24%2C%20prove%20that%20%24%5Csum%20%7C%20%5Calpha_i%20%7C%20%3C%20%5Cinfty%24**.%0A%0A%20%20%20%20Proof%0A%0A%20%20%20%20By%20the%20above%20exercise%2C%20if%20we%20denote%20%24x%20%3D%20%5Clbrace%20%5Cxi_i%20%5Crbrace%24%20such%20that%20%24%5Cxi%20%5Cto%200%24%20then%20%24%5Clbrace%20x%20%5Crbrace%20%3D%20c_0%24%20is%20a%20Banach%20space.%20For%20each%20%24n%24%20define%20%24%5CLambda_n%20%3A%20c_0%20%5Cto%20%5Cmathbb%7BC%7D%24%20by%20%24%5CLambda_n(x)%20%3D%20%5Csum%5Climits_%7Bi%3D1%7D%5En%5Calpha_i%5Cxi_i%24.%20Each%20%24%5CLambda_n%24%20is%20obviously%20linear.%0A%0A%20%20%20%20Write%20%24%7C%20%5CLambda_n%20x%20%7C%20%3D%20%5Csum%5Climits_%7Bi%3D1%7D%5En%20%7C%20%5Calpha_i%5Cxi_i%20%7C%24%20%24%5Cle%20%5Csum%5Climits_%7Bi%3D1%7D%5En%20%7C%20%5Calpha_i%20%7C%20%5C%7C%20x%20%5C%7C_%7B%5Cinfty%7D%24%2C%20then%20by%20definition%20of%20the%20norm%20of%20a%20linear%20transformation%2C%20%24%5C%7C%20%5CLambda_n%20%5C%7C%20%5Cle%20%5Csum%5Climits_%7Bi%3D1%7D%5En%20%7C%20%5Calpha_i%20%7C%24.%20On%20the%20other%20hand%2C%20%24%5C%7C%20%5CLambda_n%20%5C%7C%20%3A%3D%24%20%24%5Csup%5Clbrace%20%5C%7C%20%5CLambda_n%20x%20%5C%7C%20%3A%20x%20%5Cin%20c_0%2C%20%5C%7C%20x%20%5C%7C_%7B%5Cinfty%7D%20%5Cle%201%20%5Crbrace%24%2C%20so%20it%20is%20greater%20than%20or%20equal%20to%20any%20particular%20choice%20of%20%24x%24%3A%20take%20%24x%20%3D%20%5Clbrace%20%5Cxi_i%20%5Crbrace%24%20where%20%24%5Cxi_i%20%3D%20%5Cfrac%7B%5Coverline%7B%5Calpha_i%7D%7D%7B%7C%20%5Calpha_i%20%7C%7D%24%20if%20%24i%5Cle%20n%24%20and%20%240%24%20otherwise%2C%20we%20can%20see%20%24%5C%7C%20x%20%5C%7C_%7B%5Cinfty%7D%20%5Cle%201%24%20and%20this%20makes%20%24%5C%7C%20%5CLambda_n%20x%20%5C%7C%20%3D%20%5Csum%5Climits_%7Bi%3D1%7D%5En%20%7C%20%5Calpha_i%20%7C%24.%20So%20in%20conclusion%20we%20have%20%24%5C%7C%20%5CLambda_n%20%5C%7C%20%3D%20%5Csum%5Climits_%7Bi%3D1%7D%5En%20%7C%20%5Calpha_i%20%7C%24.%20In%20particular%20each%20%24%5CLambda_n%24%20is%20a%20bounded%20linear%20functional.%0A%0A%20%20%20%20We%20assume%20%24%5Csum%5Climits_%7Bi%3D1%7D%5E%7B%5Cinfty%7D%5Calpha_i%5Cxi_i%24%20converges%20for%20any%20%24x%20%5Cin%20c_0%24%2C%20which%20is%20the%20same%20to%20say%20%24%5Csup%5Climits_n%20%7C%20%5CLambda_nx%20%7C%20%3C%20%5Cinfty%24%20for%20any%20%24x%20%5Cin%20c_0%24.%20Thus%20Banach-Steinhaus%20Theorem%20says%20that%20there%20exists%20%24M%20%3C%20%5Cinfty%24%20such%20that%20%24%5Csum%5Climits_%7Bi%3D1%7D%5En%20%7C%20%5Calpha_i%20%7C%24%20%24%3D%20%5C%7C%20%5CLambda_n%20%5C%7C%20%5Cle%20M%24%20for%20all%20%24n%24.%20Take%20limit%20and%20we%20have%20that%20%24%5Csum%5Climits_%7Bi%3D1%7D%5En%20%7C%20%5Calpha_i%20%7C%20%5Cle%20M%20%3C%20%5Cinfty%24%20as%20desired.%0A%0A%20%20%20%20%23%23%20Exercise%2011%0A%0A%20%20%20%20**For%20%240%20%3C%20%5Calpha%20%5Cle%201%24%2C%20let%20%24%5Ctext%7BLip%7D~%5Calpha%24%20denote%20the%20space%20of%20all%20complex%20functions%20%24f%24%20on%20%24%5Ba%2Cb%5D%24%20for%20which**%20%24%24M_f%20%3D%20%5Csup%5Climits_%7Bs%20%5Cne%20t%7D%5Cfrac%7B%7C%20f(s)%20-%20f(t)%20%7C%7D%7B%7C%20s%20-%20t%20%7C%5E%7B%5Calpha%7D%7D%20%3C%20%5Cinfty.%24%24%20**Prove%20that%20%24%5Ctext%7BLip%7D~%5Calpha%24%20is%20a%20Banach%20space%20if%20%24%5C%7C%20f%20%5C%7C%20%3D%20%7C%20f(a)%20%7C%20%2B%20M_f%24%2C%20also%20if%20%24%5C%7C%20f%20%5C%7C%20%3D%20%5Csup%5Climits_x%20%7C%20f(x)%20%7C%20%2B%20M_f%24.%20The%20members%20of%20%24%5Ctext%7BLip%7D~%5Calpha%24%20are%20said%20to%20satisfy%20a%20Lipschitz%20condition%20of%20order%20%24%5Calpha%24**.%0A%0A%20%20%20%20Proof%0A%0A%20%20%20%20First%20a%20property%20of%20%24M_f%24%20that%20will%20be%20used%2C%20for%20%24f%2Cg%20%5Cin%20%5Ctext%7BLip%7D~%5Calpha%24%20and%20%24b%2C%20c%24%20scalars%2C%20%24%24%5Cbegin%7Baligned%7DM_%7Bbf%2Bcg%7D%26%3A%3D%20%5Csup%20%5Cfrac%7B%7C%20bf(s)%20%2B%20cg(s)%20-%20bf(t)%20-%20cg(t)%20%7C%7D%7B%7C%20s%20-%20t%20%7C%5E%7B%5Calpha%7D%7D%20%5C%5C%20%26%5Cle%20%5Csup%20%5Cfrac%7B%7C%20b%5C%7C%20f(s)%20-%20f(t)%20%7C%7D%7B%7C%20s%20-%20t%20%7C%5E%7B%5Calpha%7D%7D%20%2B%20%5Csup%5Cfrac%7B%7C%20c%20%5C%7C%20g(s)%20-%20g(t)%20%7C%7D%7B%7C%20s%20-%20t%20%7C%5E%7B%5Calpha%7D%7D%20%5C%5C%20%26%3D%3A%7C%20b%20%7C%20M_f%20%2B%20%7C%20c%20%7C%20M_g%5Cend%7Baligned%7D.%24%24%20Notice%20if%20(at%20least)%20one%20of%20the%20scalar%20is%20%240%24%20then%20we%20have%20equality%2C%20we%20merely%20pull%20the%20scalar%20out.%20This%20also%20shows%20that%20any%20linear%20combination%20%24bf%2Bcg%24%20is%20also%20a%20member%20of%20%24%5Ctext%7BLip%7D~%5Calpha%24%2C%20thus%20%24%5Ctext%7BLip%7D~%5Calpha%24%20is%20a%20vector%20space.%0A%0A%20%20%20%20Now%20follow%20the%20definition%20of%20a%20Banach%20space%3A%0A%0A%20%20%20%201.%20Let%20%24f%2C%20g%20%5Cin%20%5Ctext%7BLip%7D~%5Calpha%24.%20For%20the%20first%20norm%2C%20we%20have%20%24%5C%7C%20f%20%2B%20g%20%5C%7C%24%20%24%3D%7C%20(f%2Bg)(a)%20%7C%20%2B%20M_%7Bf%2Bg%7D%24%2C%20use%20the%20above%20property%2C%20%24%5Cle%20%7C%20f(a)%20%7C%20%2B%20%7C%20g(a)%20%7C%20%2B%20M_f%20%2B%20M_g%20%3D%20%5C%7C%20f%20%5C%7C%20%2B%20%5C%7C%20g%20%5C%7C%24.%20For%20the%20second%20norm%2C%20we%20have%20%24%5C%7C%20f%2Bg%20%5C%7C%20%3D%20%5Csup%7C%20(f%2Bg)(x)%20%7C%20%2B%20M_%7Bf%2Bg%7D%24%20%24%5Cle%20M_f%20%2B%20M_g%20%2B%20%5Csup%20%7C%20f(x)%20%7C%20%2B%20%5Csup%20%7C%20g(x)%20%7C%24%20%24%3D%20%5C%7C%20f%20%5C%7C%20%2B%20%5C%7C%20g%20%5C%7C%24%3B%0A%20%20%20%202.%20Let%20%24f%20%5Cin%20%5Ctext%7BLip%7D~%5Calpha%24%20and%20%24b%24%20a%20scalar.%20For%20the%20first%20norm%20we%20have%20%24%5C%7C%20bf%20%5C%7C%24%20%24%3D%20%7C%20bf(a)%20%7C%20%2B%20M_%7Bbf%7D%24%20%24%3D%20%7C%20b%20%5C%7C%20f(a)%20%7C%20%2B%20%7C%20b%20%7C%20M_f%24%20%24%3D%20%7C%20b%20%7C%20%5C%7C%20f%20%5C%7C%24.%20For%20the%20second%20norm%20we%20have%20%24%5C%7C%20bf%20%5C%7C%24%20%24%3D%20%5Csup%7C%20bf(x)%20%7C%20%2B%20M_%7Bbf%7D%24%20%24%3D%20%7C%20b%20%7C%20M_f%20%2B%20%7C%20b%20%7C%20%5Csup%20%7C%20f(x)%20%7C%24%20%24%3D%20%7C%20b%20%7C%20%5C%7C%20f%20%5C%7C%24%3B%0A%20%20%20%203.%20If%20%24%5C%7C%20f%20%5C%7C%20%3D%200%24%2C%20then%20for%20the%20first%20norm%20%24%7C%20f(a)%20%7C%24%20and%20%24M_f%24%20are%20%240%24%20because%20they%20are%20non-negative%2C%20but%20then%20%24%7C%20f(s)%20-%20f(a)%20%7C%20%3D%200%24%20for%20all%20possible%20%24s%24%2C%20so%20%24f(x)%20%5Cequiv%200%24%20on%20%24%5Ba%2C%20b%5D%24.%20For%20the%20second%20norm%20we%20have%20%24M_f%24%20and%20%24%5Csup%20%7C%20f(x)%20%7C%24%20are%20both%20%240%24%2C%20again%20it%20is%20easy%20to%20see%20%24f(x)%20%5Cequiv%200%24%3B%0A%20%20%20%204.%20Fix%20%24%5Cvarepsilon%20%3E%200%24%20and%20let%20%24%5Clbrace%20f_n%20%5Crbrace%20%5Csubset%20%5Ctext%7BLip%7D~%5Calpha%24%20be%20a%20Cauchy%20sequence%20with%20respect%20to%20the%20first%20norm%2C%20that%20is%20to%20say%20there%20is%20some%20%24N%24%20such%20that%20for%20%24n%2Cm%3EN%24%20we%20have%20%24%5C%7C%20f_n%20-%20f_m%20%5C%7C%24%20%24%3D%20%7C%20f_n(a)%20-%20f_m(a)%20%7C%20%2B%20M_%7Bf_n%20-%20f_m%7D%20%3C%20%5Cvarepsilon%24.%20So%20%24%7C%20f_n(a)%20-%20f_m(a)%20%7C%20%3C%20%5Cvarepsilon%24%2C%20thus%20%24%5Clbrace%20f_n(a)%5Crbrace%24%20is%20a%20Cauchy%20sequence.%20Since%20it%20lies%20in%20%24%5Cmathbb%7BC%7D%24%20and%20complex%20field%20is%20complete%2C%20%24%5Clbrace%20f_n(a)%20%5Crbrace%24%20is%20convergent.%20Also%20for%20any%20%24x%20%5Cin%20%5Ba%2Cb%5D%24%2C%20we%20have%20%24%5Cvarepsilon%3E%20M_%7Bf_n-f_m%7D%24%20%24%5Cge%20%5Cfrac%7B%7C%20f_n(a)-f_n(x)-f_m(a)%2Bf_m(x)%7C%7D%7B%7C%20a-x%20%7C%5E%7B%5Calpha%7D%7D%24%2C%20this%20shows%20%24f_n(x)%24%20is%20Cauchy%20(i.e.%20%24%7C%20f_n(x)%20-%20f_m(x)%20%7C%20%3C%5Cvarepsilon%24)%20thus%20convergent%20for%20any%20%24x%24.%20Define%20%24f(x)%20%3D%20%5Clim%20f_n(x)%24.%20We%20want%20to%20show%20%24f%20%5Cin%20%5Ctext%7BLip%7D~%5Calpha%24%20and%20%24%5C%7C%20f_n%20-%20f%20%5C%7C%20%5Cto%200%24%3A%0A%20%20%20%20%091.%20By%20convergence%2C%20we%20have%20%24%7C%20f_n(a)%20-%20f(a)%20%7C%20%5Cle%20%5Cvarepsilon%24.%20Also%20%24%5Cvarepsilon%20%3E%20M_%7Bf_n%20-%20f_m%7D%24%20%24%5Cge%20%5Cfrac%7Bf_n(s)%20-%20f_n(t)%20-%20f_m(s)%20%2B%20f_m(t)%7D%7B%7C%20st%20%7C%5E%7B%5Calpha%7D%7D%24%2C%20take%20%24m%20%5Cto%20%5Cinfty%24%20we%20have%20%24M_%7Bf_n%20-%20f%7D%20%5Cle%20%5Cvarepsilon%24.%20So%20%24%5C%7C%20f_n%20-%20f%20%5C%7C%24%20%24%3D%20%7C%20f_n(a)%20-f(a)%20%7C%20%2B%20M_%7Bf_n%20-%20f%7D%24%20%24%5Cle%202%5Cvarepsilon%24%3B%0A%20%20%20%20%092.%20For%20any%20%24s%2Ct%20%5Cin%20%5Ba%2C%20b%5D%24%20and%20%24s%20%5Cne%20t%24%20we%20have%20%24%5Cfrac%7B%7C%20f(s)%20-%20f(t)%20%7C%7D%7B%7C%20s%20-%20t%7C%5E%7B%5Calpha%7D%7D%24%20%24%3D%20%5Cfrac%7B%7C%20f(s)%20-%20f_n(s)%20%2B%20f_n(s)%20-%20f_n(t)%20%2B%20f_n(t)%20-%20f(t)%7C%7D%7B%7C%20s-t%20%7C%5E%7B%5Calpha%7D%7D%24%20%24%5Cle%20%5Cfrac%7B%7C%20f(s)%20-%20f_n(s)%20%2B%20f_n(t)%20-%20f(t)%20%7C%20%2B%20%7C%20f_n(s)%20-%20f_n(t)%20%7C%7D%7B%7C%20s-t%20%7C%5E%7B%5Calpha%7D%7D%24%20%24%5Cle%20M_%7Bf_n%20-%20f%7D%20%2B%20M_%7Bf_n%7D%24%2C%20by%20above%20this%20%24%5Cle%20%5Cvarepsilon%20%2B%20M_%7Bf_n%7D%24.%20Since%20%24f_n%20%5Cin%20%5Ctext%7BLip%7D~%5Calpha%24%2C%20%24M_%7Bf_n%7D%20%3C%20%5Cinfty%24%20and%20thus%20%24M_f%20%3C%20%5Cinfty%24%2C%20thus%20%24f%20%5Cin%20%5Ctext%7BLip%7D~%5Calpha%24.%0A%20%20%20%20%09This%20shows%20%24%5Ctext%7BLip%7D~%5Calpha%24%20is%20complete%20thus%20a%20Banach%20space%20with%20respect%20to%20the%20first%20norm.%20Let%20%24%5Clbrace%20f_n%20%5Crbrace%20%5Csubset%20%5Ctext%7BLip%7D~%5Calpha%24%20be%20a%20Cauchy%20sequence%20with%20respect%20to%20the%20second%20norm%2C%20that%20is%20to%20say%20there%20is%20some%20%24N%24%20such%20that%20for%20%24n%2Cm%3EN%24%20we%20have%20%24%5C%7C%20f_n%20-%20f_m%20%5C%7C%24%20%24%3D%20%5Csup%20%7C%20f_n(x)%20-%20f_m(x)%20%7C%20%2B%20M_%7Bf_n%20-%20f_m%7D%20%3C%20%5Cvarepsilon%24.%20So%20%24%7C%20f_n(x)%20-%20f_m(x)%20%7C%20%3C%20%5Cvarepsilon%24%20for%20all%20%24x%24%20for%20%24n%2C%20m%3E%20N%24%2C%20thus%20%24%5Clbrace%20f_n(x)%5Crbrace%24%20is%20a%20Cauchy%20sequence%2C%20thus%20%24%5Clbrace%20f_n(x)%20%5Crbrace%24%20is%20convergent%2C%20the%20rest%20is%20pretty%20much%20the%20same%20as%20above.%0A%20%20%20%20%09Remark%3A%20another%20way%20is%20to%20first%20show%20that%20%24%5Ctext%7BLip%7D~%5Calpha%24%20is%20complete%20with%20respect%20to%20the%20second%20norm%2C%20denoted%20by%20%24%5C%7C%20%5Ccdot%20%5C%7C_%7B*%7D%24%2C%20then%20show%20%24%5C%7C%20f%20%5C%7C_%7B*%7D%20%5Cle%20c%20%5C%7C%20f%20%5C%7C_a%24%20(%24%5C%7C%20%5Ccdot%20%5C%7C_a%24%20to%20denote%20the%20first%20norm)%20for%20some%20%24c%24%20independent%20from%20the%20choice%20of%20%24%5Clbrace%20f_n%20%5Crbrace%24%3A%0A%20%20%20%20%09For%20any%20%24x%20%5Cin%20(a%2C%20b%5D%24%2C%20we%20have%20%24%24%5Cbegin%7Baligned%7D%20%7C%20f(x)%20%7C%20%26%5Cle%20%7C%20f(a)%20%7C%20%2B%20%7C%20f(x)%20-%20f(a)%20%7C%20%5C%5C%20%26%3D%20%7C%20f(a)%20%7C%20%2B%20%5Cfrac%7B%7C%20f(x)%20-%20f(a)%20%7C%7D%7B%7C%20x%20-%20a%20%7C%5E%7B%5Calpha%7D%7D%7C%20x%20-%20a%20%7C%5E%7B%5Calpha%7D%20%5C%5C%20%26%5Cle%20%7C%20f(a)%20%7C%20%2B%20M_f%20%7C%20b%20-%20a%20%7C%5E%7B%5Calpha%7D%20%5C%5C%20%5Cimplies%20%5C%5C%20%5Csup%7B%7C%20f(x)%20%7C%7D%20%26%5Cle%20%7C%20f(a)%20%7C%20%2B%20M_f%20%7C%20b%20-%20a%20%7C%5E%7B%5Calpha%7D%20%5C%5C%20%5Csup%7B%7C%20f(x)%20%7C%7D%20%2B%20M_f%20%26%5Cle%20%7C%20f(a)%20%7C%20%2B%20M_f%20%7C%20b%20-%20a%20%7C%5E%7B%5Calpha%7D%20%2B%20M_f%20%2B%20%7C%20f(a)%20%5C%7C%20b%20-%20a%20%7C%5E%7B%5Calpha%7D%20%5C%5C%20%5C%7C%20f%20%5C%7C_%7B*%7D%20%26%5Cle%20c%20%5C%7C%20f%20%5C%7C_a%5Cend%7Baligned%7D%24%24%20if%20we%20take%20%24c%20%3D%20%7C%20b%20-%20a%20%7C%5E%7B%5Calpha%7D%20%2B%201%24.%0A%20%20%20%20%09Then%20%24%5C%7C%20f%20%5C%7C_a%20%5Cle%20%5C%7C%20f%20%5C%7C_%7B*%7D%24%20(from%20definition)%20%24%5Cle%20c%5C%7C%20f%20%5C%7C_%7Ba%7D%24%2C%20so%20the%20two%20norms%20are%20equivalent%2C%20and%20thus%20%24%5Ctext%7BLip%7D~%5Calpha%24%20is%20complete%20with%20respect%20to%20the%20first%20norm%20as%20well.%0A%0A%20%20%20%20%23%23%20Exercise%2014%0A%0A%20%20%20%20**Let%20%24C%24%20be%20the%20space%20of%20all%20real%20continuous%20functions%20on%20%24I%24%20with%20the%20supremum%20norm.%20Let%20%24X_n%24%20be%20the%20subset%20of%20%24C%24%20consisting%20of%20those%20%24f%24%20for%20which%20there%20exists%20a%20%24t%20%5Cin%20I%24%20such%20that%20%24%7C%20f(s)%20-%20f(t)%20%7C%20%5Cle%20n%20%7C%20s%20-%20t%20%7C%24%20for%20all%20%24s%20%5Cin%20I%24.%20Fix%20%24n%24%20and%20prove%20that%20each%20open%20set%20in%20%24C%24%20contains%20an%20open%20set%20which%20does%20not%20intersect%20%24X_n%24.%20Show%20that%20this%20implies%20the%20existence%20of%20a%20dense%20%24G_%7B%5Cdelta%7D%24%20in%20%24C%24%20which%20consists%20entirely%20of%20nowhere%20differentiable%20functions**.%0A%0A%20%20%20%20Proof%0A%0A%20%20%20%20Let%20everything%20be%20as%20in%20the%20description%20of%20the%20problem.%20Fix%20%24n%24.%20First%20we%20prove%20%24X_n%24%20is%20closed%20in%20%24C%24.%20Let%20%24%5Clbrace%20f_i%20%5Crbrace%24%20be%20a%20sequence%20of%20functions%20in%20%24X_n%24%20and%20%24f_i%20%5Cto%20f%24%20uniformly.%20By%20the%20assumption%20we%20can%20find%20a%20sequence%20%24%5Clbrace%20t_i%20%5Crbrace%24%20such%20that%20%24%7C%20f_i(s)%20-%20f_i(t_i)%20%7C%20%5Cle%20n%20%7C%20s%20-%20t_i%20%7C%24%20holds%20for%20each%20%24i%24%20and%20all%20%24s%20%5Cin%20I%24.%20Since%20%24%5B0%2C1%5D%24%20is%20compact%2C%20there%20is%20a%20subsequence%20of%20%24t_i%24%20that%20is%20convergent%2C%20for%20simplicity%20of%20notations%2C%20let%20us%20just%20assume%20%24t_i%24%20is%20convergent%20and%20the%20limit%20is%20%24t%24.%20Then%20%24%7C%20f(s)%20-%20f(t)%20%7C%24%20%24%3D%20%5Clim%5Climits_%7Bi%20%5Cto%20%5Cinfty%7D%20%7C%20f_i(s)%20-%20f_i(t_i)%7C%24%20%24%5Cle%20%5Clim%20n%20%7C%20s%20-%20t_i%20%7C%24%20%24%3D%20n%20%7C%20s%20-%20t%20%7C%24.%20Thus%20%24X_n%24%20is%20closed.%0A%0A%20%20%20%20Now%20we%20show%20%24X_n%24%20has%20empty%20interior.%20The%20idea%20is%20to%20uniformly%20approximate%20any%20%24f%20%5Cin%20X_n%24%20using%20some%20%24f'%24%20as%20below%3A%0A%0A%20%20%20%20%3Cimg%20src%3D%22public%2FPasted%20image%2020220309162107.png%22%20width%3D%22600%22%20%2F%3E%0A%0A%20%20%20%20We%20can%20think%20of%20a%20band%20covers%20the%20graph%20of%20%24f%24%20of%20some%20tiny%20width%20%24%5Cvarepsilon%24%2C%20%24f'%24%20moves%20zig-zag%20between%20the%20two%20sides%20of%20the%20band%20by%20some%20line%20segments%20of%20slope%20equals%20%24%5Cpm%202n%24.%20Then%20this%20%24f'%20%5Cnotin%20X_n%24%2C%20as%20that%20by%20construction%20for%20each%20%24t%20%5Cin%20%5B0%2C1%5D%24%20there%20are%20some%20%24s%24%20such%20that%20%24%5Cfrac%7B%7C%20f(s)%20-%20f(t)%7C%7D%7B%7C%20s-t%20%7C%7D%20%3D%202n%20%5Cnot%5Cle%20n%24.%0A%0A%20%20%20%20Thus%20the%20complement%20of%20%24X_n%24%20is%20open%20and%20dense%20in%20%24C%24.%20By%20Baire's%20Theorem%2C%20the%20intersection%20of%20those%20complements%20is%20a%20dense%20%24G_%7B%5Cdelta%7D%24%20in%20%24C%24%2C%20call%20it%20%24G%20%3D%20%5Ccap%20X_n%5Ec%24.%20Now%20notice%20that%20if%20a%20continuous%20function%20is%20differentiable%20at%20some%20point%20in%20%24%5B0%2C1%5D%24%20(which%20is%20compact%2C%20so%20that%20%24f%24%20is%20bounded)%2C%20it%20must%20lie%20inside%20some%20%24X_n%24%2C%20thus%20%24G%24%20consists%20exactly%20those%20nowhere%20differentiable%20functions%2C%20and%20we%20have%20the%20desired%20result.%0A%0A%20%20%20%20%23%23%20Exercise%2016%0A%0A%20%20%20%20**(Closed%20Graph%20Theorem)%20Suppose%20%24X%24%20and%20%24Y%24%20are%20Banach%20spaces%2C%20and%20suppose%20%24%5CLambda%24%20is%20a%20linear%20mapping%20of%20%24X%24%20into%20%24Y%24%2C%20with%20the%20following%20property%3A%20For%20every%20sequence%20%24%5Clbrace%20x_n%20%5Crbrace%24%20in%20%24X%24%20for%20which%20%24x%20%3D%20%5Clim%20x_n%24%20and%20%24y%20%3D%20%5Clim%20%5CLambda%20x_n%24%20exist%2C%20it%20is%20true%20that%20%24y%20%3D%20%5CLambda%20x%24.%20Prove%20that%20%24%5CLambda%24%20is%20continuous**.%0A%0A%20%20%20%20Proof%0A%0A%20%20%20%20Let%20%24X%20%5Coplus%20Y%24%20be%20the%20set%20of%20all%20ordered%20pairs%20%24(x%2Cy)%2C%20x%20%5Cin%20X%24%20and%20%24y%20%5Cin%20Y%24%2C%20with%20addition%20and%20scalar%20multiplication%20defined%20component-wise%2C%20then%20%24X%20%5Coplus%20Y%24%20is%20Banach%20if%20%24%5C%7C%20(x%2Cy)%20%5C%7C%20%3D%20%5C%7C%20x%20%5C%7C%20%2B%20%5C%7C%20y%20%5C%7C%24%3A%0A%0A%20%20%20%201.%20Suppose%20%24(x%2Cy)%2C%20(x'%2Cy')%20%5Cin%20X%20%5Coplus%20Y%24%2C%20then%20%24%5C%7C%20(x%2Cy)%20%2B%20(x'y')%20%5C%7C%24%20%24%3D%20%5C%7C%20(x%2Bx'%2Cy%2By')%20%5C%7C%24%20%24%3D%5C%7C%20x%2Bx'%20%5C%7C%20%2B%20%5C%7C%20y%20%2B%20y'%20%5C%7C%24%20%24%5Cle%20%5C%7C%20x%20%5C%7C%20%2B%20%5C%7C%20x'%20%5C%7C%20%2B%20%5C%7C%20y%20%5C%7C%20%2B%20%5C%7C%20y'%20%5C%7C%24%20%24%3D%5C%7C%20(x%2Cy)%20%5C%7C%20%2B%20%5C%7C%20(x'%2Cy')%20%5C%7C%24%3B%0A%20%20%20%202.%20Suppose%20%24(x%2Cy)%20%5Cin%20X%20%5Coplus%20Y%24%20and%20%24%5Calpha%24%20is%20a%20scalar%2C%20then%20%24%5C%7C%20%5Calpha(x%2Cy)%20%5C%7C%24%20%24%3D%20%5C%7C%20(%5Calpha%20x%2C%20%5Calpha%20y)%20%5C%7C%24%20%24%3D%20%5C%7C%20%5Calpha%20x%20%5C%7C%20%2B%20%5C%7C%20%5Calpha%20y%20%5C%7C%24%20%24%3D%20%7C%20%5Calpha%20%7C%20%5C%7C%20x%20%5C%7C%20%2B%20%7C%20%5Calpha%20%7C%20%5C%7C%20y%20%5C%7C%24%20%24%3D%20%7C%20%5Calpha%20%7C%20%5C%7C%20(x%2Cy)%20%5C%7C%24%3B%0A%20%20%20%203.%20Suppose%20%24%5C%7C%20(x%2Cy)%20%5C%7C%20%3D%200%24%2C%20then%20%24%5C%7C%20x%20%5C%7C%20%2B%20%5C%7C%20y%20%5C%7C%20%3D%200%24%2C%20thus%20both%20%24x%2C%20y%24%20are%20%240%24%20thus%20%24(x%2Cy)%20%3D%200%24%3B%0A%20%20%20%204.%20Suppose%20%24%5Clbrace%20(x_n%2C%20y_n)%20%5Crbrace%24%20is%20a%20Cauchy%20sequence%20and%20fix%20%24%5Cvarepsilon%20%3E%200%24%2C%20then%20there%20is%20some%20%24N%24%20such%20that%20for%20%24n%2C%20m%20%3E%20N%24%20we%20have%20%24%5C%7C%20(x_n%2C%20y_n)%20-%20(x_m%2C%20y_m)%20%5C%7C%24%20%24%3D%5C%7C%20x_n-x_m%20%5C%7C%20%2B%20%5C%7C%20y_n%20-%20y_m%20%5C%7C%24%20%24%3C%20%5Cvarepsilon%24%2C%20so%20%24%5Clbrace%20x_n%20%5Crbrace%24%20and%20%24%5Clbrace%20y_n%20%5Crbrace%24%20are%20both%20Cauchy%2C%20thus%20convergent%20because%20%24X%24%20and%20%24Y%24%20are%20Banach.%20Say%20they%20converge%20to%20%24x%2Cy%24%20respectively%2C%20then%20%24%5C%7C%20(x_n%2Cy_n)%20-%20(x%2Cy)%20%5C%7C%24%20%24%3D%5C%7C%20x_n%20-%20x%20%5C%7C%20%2B%20%5C%7C%20y_n%20-%20y%20%5C%7C%20%5Cto%200%24%2C%20thus%20%24(x_n%2C%20y_n)%20%5Cto%20(x%2Cy)%24%2C%20thus%20this%20sequence%20is%20convergent%2C%20thus%20this%20space%20is%20complete.%0A%0A%20%20%20%20So%20by%20definition%2C%20%24X%20%5Coplus%20Y%24%20is%20Banach.%20Suppose%20%24G%24%20is%20the%20graph%20of%20%24%5CLambda%24%20of%20which%20is%20as%20assumed%20in%20the%20problem%2C%20then%20%24G%24%20is%20closed%20(graph%20of%20continuous%20function%20to%20Hausdorff%20space).%20%24G%24%20is%20a%20subspace%2C%20because%20%24X%24%20is%20a%20vector%20space.%20Thus%20%24G%24%20is%20a%20closed%20subspace%20of%20complete%20metric%20space%20thus%20is%20also%20complete%2C%20thus%20%24G%24%20is%20Banach.%0A%0A%20%20%20%20Notice%20that%20%24%5Cvarphi%3AG%20%5Cto%20X%24%2C%20%24(x%2C%20%5CLambda%20x)%20%5Cmapsto%20x%24%20is%20a%20continuous%20linear%20bijection%2C%20thus%20by%20Theorem%205.4%2C%20%24%5Cvarphi%24%20is%20bounded%2C%20and%20by%20Theorem%205.10%20%24%5Cvarphi%5E%7B-1%7D%24%20is%20also%20bounded%20(and%20continuous).%20Thus%20%24%5C%7C%20x%20%5C%7C%20%2B%20%5C%7C%20%5CLambda%20x%20%5C%7C%24%20%24%3D%5C%7C%20(x%2C%5CLambda%20x)%20%5C%7C%24%20%24%3D%20%5C%7C%20%5Cvarphi%5E%7B-1%7D(x)%20%5C%7C%24%20%24%5Cle%20%5C%7C%20%5Cvarphi%5E%7B-1%7D%20%5C%7C%20%5C%7C%20x%20%5C%7C%24%20thus%20%24%5C%7C%20%5CLambda%20x%20%5C%7C%24%20%24%5Cle%20(%5C%7C%20%5Cvarphi%5E%7B-1%7D%20%5C%7C%20-1)%5C%7C%20x%20%5C%7C%24.%0A%0A%20%20%20%20Now%20from%20definition%2C%20%24%5C%7C%20%5CLambda%20%5C%7C%20%3A%3D%20%5Csup%5Clbrace%20%5C%7C%20%5CLambda%20x%20%5C%7C%20%3A%20x%20%5Cin%20X%2C%20%5C%7C%20x%20%5C%7C%20%3D%201%5Crbrace%24%20%24%5Cle%20%5C%7C%20%5Cvarphi%5E%7B-1%7D%20%5C%7C%20-1%24%2C%20which%20by%20above%20is%20bounded%2C%20apply%20Theorem%205.4%20again%2C%20we%20have%20that%20%24%5CLambda%24%20is%20continuous.%0A%0A%20%20%20%20%23%23%20Exercise%2018%0A%0A%20%20%20%20**Suppose%20%24%5Clbrace%20%5CLambda_n%20%5Crbrace%24%20is%20a%20sequence%20of%20bounded%20linear%20transformations%20from%20a%20normed%20linear%20space%20%24X%24%20to%20a%20Banach%20space%20%24Y%24%2C%20suppose%20%24%5C%7C%20%5CLambda_n%20%5C%7C%20%5Cle%20M%20%3C%20%5Cinfty%24%20for%20all%20%24n%24%2C%20and%20suppose%20there%20is%20a%20dense%20set%20%24E%20%5Csubset%20X%24%20such%20that%20%24%5Clbrace%20%5CLambda_n%20x%20%5Crbrace%24%20converges%20for%20all%20%24x%20%5Cin%20E%24.%20Prove%20that%20%24%5Clbrace%20%5CLambda_n%20x%20%5Crbrace%24%20converges%20for%20each%20%24x%20%5Cin%20X%24**.%0A%0A%20%20%20%20Proof%0A%0A%20%20%20%20Suppose%20%24x%20%5Cin%20X%24%20and%20fix%20%24%5Cvarepsilon%3E0%24.%20Since%20%24E%24%20is%20dense%2C%20we%20can%20find%20%24y%20%5Cin%20E%24%20such%20that%20%24%5C%7C%20x%20-%20y%20%5C%7C%20%3C%20%5Cvarepsilon%2FM%24.%0A%0A%20%20%20%20Since%20%24y%20%5Cin%20E%24%2C%20by%20assumption%20%24%5Clbrace%20%5CLambda_ny%20%5Crbrace%24%20is%20convergent%20thus%20Cauchy%2C%20so%20there%20is%20some%20%24N%24%20such%20that%20for%20%24n%2C%20m%20%3E%20N%24%20we%20have%20%24%5C%7C%20%5CLambda_ny%20-%20%5CLambda_my%20%5C%7C%20%3C%20%5Cvarepsilon%24.%20For%20the%20same%20%24N%24%20and%20for%20%24n%2C%20m%20%3E%20N%24%2C%20we%20can%20write%20%24%24%5Cbegin%7Baligned%7D%5C%7C%20%5CLambda_nx%20-%20%5CLambda_mx%20%5C%7C%20%26%5Cle%20%5C%7C%20%5CLambda_nx%20-%20%5CLambda_ny%20%5C%7C%20%2B%20%5C%7C%20%5CLambda_ny%20-%20%5CLambda_my%20%5C%7C%20%2B%20%5C%7C%20%5CLambda_my%20-%20%5CLambda_mx%20%5C%7C%5Cend%7Baligned%7D%2C%24%24%20plug%20in%20%24%5C%7C%20%5CLambda_ny%20-%20%5CLambda_my%20%5C%7C%20%3C%20%5Cvarepsilon%24%20and%20remember%20%24%5CLambda_i%24's%20are%20linear%2C%20we%20have%20%24%24%5Cbegin%7Baligned%7D%26%3C%20%5C%7C%20%5CLambda_n(x-y)%20%5C%7C%20%2B%20%5C%7C%20%5CLambda_m(x-y)%20%5C%7C%20%2B%20%5Cvarepsilon%20%5C%5C%20%26%5Cle%20%5C%7C%20%5CLambda_n%20%5C%7C%20%5C%7C%20x-y%20%5C%7C%20%2B%20%5C%7C%20%5CLambda_m%20%5C%7C%20%5C%7C%20x-y%20%5C%7C%20%2B%20%5Cvarepsilon%5Cend%7Baligned%7D%2C%24%24%20plug%20in%20everything%20we%20have%2C%20this%20value%20is%20smaller%20than%20%242M%5Ccdot%5Cvarepsilon%2FM%20%2B%20%5Cvarepsilon%20%3D%203%5Cvarepsilon%24.%20Since%20%24%5Cvarepsilon%24%20is%20chosen%20arbitrarily%2C%20this%20shows%20%24%5Clbrace%20%5CLambda_nx%5Crbrace%24%20is%20Cauchy.%20Since%20%24Y%24%20is%20Banach%20thus%20complete%2C%20%24%5Clbrace%20%5CLambda_nx%20%5Crbrace%24%20is%20convergent.%0A%20%20%20%20%22%22%22)%0A%20%20%20%20return%0A%0A%0A%40app.cell%0Adef%20_(mo)%3A%0A%20%20%20%20mo.md(r%22%22%22%0A%20%20%20%20%23%20Spring%20Semester%20Homework%202%0A%20%20%20%20%22%22%22)%0A%20%20%20%20return%0A%0A%0A%40app.cell%0Adef%20_(mo)%3A%0A%20%20%20%20mo.md(r%22%22%22%0A%20%20%20%20%23%23%20Exercise%202%0A%20%20%20%20**Prove%20that%20the%20example%20given%20at%20the%20end%20of%20Sec.%206.10%20has%20the%20stated%20properties**.%0A%0A%20%20%20%20Proof%0A%0A%20%20%20%20So%20we%20want%20to%20show%3A%0A%0A%20%20%20%20%3EIf%20%24%5Cmu%24%20is%20the%20Lebesgue%20measure%20on%20%24(0%2C1)%24%2C%20%24%5Clambda%24%20be%20the%20counting%20measure%20on%20the%20%24%5Csigma%24-algebra%20of%20all%20Lebesgue%20measurable%20sets%20in%20%24(0%2C1)%24%2C%20then%20%24%5Clambda%24%20has%20no%20Lebesgue%20decomposition%20relative%20to%20%24%5Cmu%24%2C%20and%20although%20%24%5Cmu%20%5Cll%20%5Clambda%24%20and%20%24%5Cmu%24%20is%20bounded%2C%20there%20is%20no%20%24h%20%5Cin%20L%5E1(%5Clambda)%24%20such%20that%20%24d%5Cmu%20%3D%20hd%5Clambda%24.%0A%0A%20%20%20%20In%20particular%2C%20%24%5Clambda%24%20is%20not%20%24%5Csigma%24-finite%20(because%20finite%20%24%5Ctimes%24%20countable%20is%20countable%2C%20yet%20%24(0%2C1)%24%20is%20uncountable).%20This%20example%20show%20that%20Lebesgue-Radon-Nikodym%20Theorem%20is%20in%20general%20false%20if%20we%20omit%20the%20%24%5Csigma%24-finite%20condition.%0A%0A%20%20%20%20For%20any%20%24x%20%5Cin%20(0%2C1)%24%2C%20%24%5Clambda(%5Clbrace%20x%20%5Crbrace)%24%20is%20by%20definition%20%241%24%2C%20and%20%24%5Cmu(%5Clbrace%20x%20%5Crbrace)%20%3D%200%24.%20Suppose%20%24%5Clambda%24%20has%20a%20Lebesgue%20decomposition%20%24%5Clambda%20%3D%20%5Clambda_a%20%2B%20%5Clambda_s%24%20where%20%24%5Clambda_a%20%5Cll%20%5Cmu%24%20and%20%24%5Clambda_s%20%5Cperp%20%5Cmu%24%3A%20since%20%24%5Cmu(%5Clbrace%20x%20%5Crbrace)%20%3D%200%24%20we%20must%20have%20%24%5Clambda_a(%5Clbrace%20x%20%5Crbrace)%20%3D%200%24%2C%20so%20we%20must%20have%20%24%5Clambda_s(%5Clbrace%20x%20%5Crbrace)%20%3D%201%24.%20Since%20this%20is%20true%20for%20all%20%24x%20%5Cin%20(0%2C1)%24%2C%20it%20follows%20that%20%24%5Clambda_s%24%20is%20concentrated%20on%20the%20whole%20%24(0%2C1)%24.%20Thus%20we%20cannot%20have%20%24%5Clambda_s%20%5Cperp%20%5Cmu%24%2C%20otherwise%20%24%5Cmu%24%20has%20to%20be%20concentrated%20on%20%24%5Cvarnothing%24%2C%20which%20is%20absurd.%20In%20conclusion%20we%20cannot%20have%20such%20decomposition.%0A%0A%20%20%20%20Now%2C%20if%20%24E%20%5Cin%20M%24%20such%20that%20%24%5Clambda(E)%20%3D%200%24%2C%20then%20%24E%24%20must%20be%20empty%2C%20and%20of%20course%20%24%5Cmu(%5Cvarnothing)%20%3D%200%24%2C%20thus%20%24%5Cmu%20%5Cll%20%5Clambda%24.%20Also%20%24%5Cmu%24%20is%20bounded.%20Suppose%20now%20if%20there%20is%20%24h%20%5Cin%20L%5E1(%5Clambda)%24%20such%20that%20%24d%5Cmu%20%3D%20hd%5Clambda%24%20%24%5Ciff%20%5Cmu(E)%20%3D%20%5Cint_Ed%5Cmu%20%3D%20%5Cint_Ehd%5Clambda%24%20for%20all%20%24E%24.%20For%20any%20%24x%20%5Cin%20(0%2C1)%24%2C%20%24%5Cmu(%5Clbrace%20x%20%5Crbrace)%20%3D%200%24%2C%20but%20%24%5Clambda(%5Clbrace%20x%20%5Crbrace)%20%5Cne%200%24%2C%20so%20we%20must%20have%20%24h(x)%20%3D%200%24%2C%20so%20%24h%20%5Cequiv%200%24%20on%20%24(0%2C1)%24.%20On%20the%20other%20hand%2C%20%24%5Cmu(0%2C1)%20%3D%201%24%2C%20thus%20%24h%24%20cannot%20be%20identically%20%240%24%2C%20so%20such%20%24h%24%20cannot%20exist.%0A%0A%20%20%20%20%23%23%20Exercise%203%0A%0A%20%20%20%20**Prove%20that%20the%20vector%20space%20%24M(X)%24%20of%20all%20complex%20regular%20Boreal%20measures%20on%20a%20locally%20compact%20Hausdorff%20space%20%24X%24%20is%20a%20Banach%20space%20if%20%24%5C%7C%20%5Cmu%20%5C%7C%20%3D%20%7C%20%5Cmu%20%7C(X)%24.%20That%20the%20difference%20of%20any%20two%20members%20of%20%24M(X)%24%20is%20in%20%24M(X)%24%20was%20used%20in%20the%20first%20paragraph%20of%20the%20proof%20of%20Theorem%206.19%3B%20supply%20a%20proof%20of%20this%20fact**.%0A%0A%20%20%20%20Proof%0A%0A%20%20%20%20%23%23%23%20Part%201%0A%0A%20%20%20%20A%20little%20translation%3A%20Riesz%20Representation%20Theorem%20says%20that%20for%20any%20%24%5CPhi%20%5Cin%20(C_0(X))%5E%7B*%7D%24%20we%20have%20a%20unique%20%24%5Cmu_%7B%5CPhi%7D%20%5Cin%20M(X)%24%20such%20that%20%24%5CPhi%20f%20%3D%20%5Cint_X%20fd%5Cmu_%7B%5CPhi%7D%24%20for%20all%20%24f%20%5Cin%20C_0(X)%24%20such%20that%20%24%5C%7C%20%5CPhi%20%5C%7C%20%3D%20%7C%20%5Cmu_%7B%5CPhi%7D%20%7C%20(X)%24.%20Thus%20if%20we%20define%20%24g%3A%20(C_0(X))%5E%7B*%7D%20%5Cto%20M(X)%24%20by%20%24g(%5CPhi)%20%3D%20%5Cmu_%7B%5CPhi%7D%24%2C%20then%3A%0A%0A%20%20%20%201.%20%24g%24%20is%20well-defined%20by%20Riesz%20Representation%20Theorem%3B%0A%20%20%20%202.%20%24g%24%20is%20surjective%20by%20the%20'inverse'%20of%20Riesz%20Representation%20Theorem%3B%0A%20%20%20%203.%20%24g%24%20is%20injective%20because%20%24%5Cmu_%7B%5CPhi%7D%24%20is%20the%20trivial%20measure%20if%20and%20only%20if%20%24%5CPhi%20f%20%5Cequiv%200%24%2C%20thus%20%24%5CPhi%20%5Cequiv%200%24%20and%20%24g%24%20has%20trivial%20kernel%3B%0A%20%20%20%204.%20%24g%24%20is%20linear%20because%20%24%5Cint_Xfd(%5Cmu_%7Ba%5CPhi%2Bb%5CPsi%7D)%24%20%24%3D(a%5CPhi%2Bb%5CPsi)f%24%20%24%3Da%5CPhi%20f%20%2B%20b%5CPsi%20f%24%20%24%3D%20a%5Cint_X%20f%20d%5Cmu_%7B%5CPhi%7D%20%2B%20b%5Cint_X%20f%20%5Cmu_%7B%5CPsi%7D%24%20%24%3D%20%5Cint_X%20f%20da%5Cmu_%7B%5CPhi%7D%20%2B%20%5Cint_X%20f%20db%5Cmu_%7B%5CPsi%7D%24%20%24%3D%20%5Cint_X%20f%20d(a%5Cmu_%7B%5CPhi%7D%20%2B%20b%5Cmu_%7B%5CPsi%7D)%24%2C%20meaning%20that%20%24%5Cmu_%7Ba%5CPhi%20%2B%20b%5CPsi%7D%20%3D%20a%5Cmu_%7B%5CPhi%7D%20%2B%20b%5Cmu_%7B%5CPsi%7D%24%2C%20which%20just%20means%20%24g(a%5CPhi%20%2B%20b%5CPsi)%20%3D%20ag(%5CPhi)%20%2B%20bg(%5CPsi)%24%3B%0A%20%20%20%205.%20%24g%24%20is%20an%20isometry%20as%20%24%5C%7C%20%5CPhi%20%5C%7C%24%20%24%3D%20%7C%20%5Cmu_%7B%5CPhi%7D%20%7C%20(X)%24%20%24%3D%20%5C%7C%20%5Cmu_%7B%5CPhi%7D%20%5C%7C%24%20by%20Riesz%20Representation%20Theorem%20and%20the%20assumption.%0A%0A%20%20%20%20So%20%24g%24%20is%20an%20isometry%20isomorphism%20of%20vector%20spaces.%20We%20know%20%24C_0(X)%24%20is%20a%20Banach%20space%2C%20thus%20the%20dual%20space%20%24(C_0(X))%5E%7B*%7D%24%20is%20a%20Banach%20space%2C%20thus%20the%20isometry%20isomorphism%20image%20%24M(X)%24%20is%20also%20a%20Banach%20space.%0A%0A%20%20%20%20%23%23%23%20Part%202%0A%0A%20%20%20%20This%20part%20essentially%20want%20us%20to%20verify%20%24M(X)%24%20is%20a%20vector%20space%2C%20i.e.%20if%20%24%5Cmu%2C%20%5Clambda%20%5Cin%20M(X)%24%20and%20%24a%24%20is%20a%20scalar%2C%20then%20%24a%5Cmu%24%2C%20%24%5Cmu%20%2B%20%5Clambda%24%20are%20also%20in%20%24M(X)%24.%0A%0A%20%20%20%20So%20suppose%20%24%5Cmu%2C%20%5Clambda%20%5Cin%20M(X)%24%20and%20%24a%24%20is%20a%20scalar%3B%20fix%20%24%5Cvarepsilon0%24.%20By%20definition%20it%20means%20%24%7C%20%5Cmu%20%7C%2C%20%7C%20%5Clambda%20%7C%24%20are%20(positive)%20regular%20Borel%20measure.%20Let%20%24E%24%20be%20any%20Borel%20set%2C%20there%20are%20%24V_1%2CV_2%24%20open%20sets%20containing%20%24E%24%20such%20that%20%24%7C%20%5Cmu%20%7C(V_1)%20%3C%20%7C%20%5Cmu%20%7C%20(E)%20%2B%20%5Cvarepsilon%20%5Cimplies%20%7C%20%5Cmu%20%7C(V_1%20-%20E)%20%3C%20%5Cvarepsilon%24%20and%20%24%7C%20%5Clambda%20%7C(V_2%20-%20E)%20%3C%20%5Cvarepsilon%24%3B%20also%20there%20are%20%24K_1%2CK_2%24%20compact%20sets%20contained%20in%20%24E%24%20such%20that%20%24%7C%20%5Cmu%20%7C%20(K_1)%20%3E%20%7C%20%5Cmu%20%7C%20(E)%20-%20%5Cvarepsilon%20%5Cimplies%20%7C%20%5Cmu%20%7C(E%20-%20K_1)%20%3C%20%5Cvarepsilon%24%20and%20%24%7C%20%5Clambda%20%7C(E%20-%20K_2)%20%3C%20%5Cvarepsilon%24.%0A%0A%20%20%20%20Consider%20%24%7C%20%5Cmu%20%2B%20%5Clambda%20%7C%20(E)%24%2C%20it%20by%20definition%20equals%20%24%5Csup%5Climits_%7B%5Ctext%7Bpartition%20%7D%5Clbrace%20E_i%20%5Crbrace%7D%20%5Csum%5Climits_i%20%7C%20(%5Cmu%2B%5Clambda)(E_i)%20%7C%24%20%24%5Cle%20%5Csup%5Csum%5Climits(%7C%5Cmu%7C(E)%20%2B%20%7C%5Clambda%7C(E))%24%20%24%3D%7C%20%5Cmu%20%7C(E)%20%2B%20%7C%20%5Clambda%20%7C%20(E)%24.%0A%0A%20%20%20%20Let%20%24V%20%3D%20V_1%20%5Ccap%20V_2%24%20and%20%24K%20%3D%20K_1%20%5Ccup%20K_2%24%2C%20then%20%24K%20%5Csubset%20E%20%5Csubset%20V%24%2C%20%24K%24%20is%20compact%20and%20%24V%24%20is%20open%2C%20and%20we%20have%3A%20%24%24%5Cbegin%7Baligned%7D%20%7C%20%5Cmu%2B%5Clambda%20%7C(V)%20%26%3D%20%7C%20%5Cmu%2B%20%5Clambda%20%7C(E)%20%2B%20%7C%20%5Cmu%20%2B%20%5Clambda%20%7C%20(V-E)%5C%5C%26%20%5Cle%7C%20%5Cmu%2B%20%5Clambda%20%7C(E)%20%2B%7C%5Cmu%7C(V-E)%2B%7C%20%5Clambda%20%7C(V-E)%2C%20%5Ctext%7B%20by%20above%20inequality%3B%7D%20%5C%5C%20%26%5Cle%7C%20%5Cmu%2B%20%5Clambda%20%7C(E)%2B%7C%20%5Cmu%20%7C(V_1%20-%20E)%2B%7C%20%5Clambda%20%7C(V_2-E)%2C%5Ctext%7B%20because%20those%20are%20larger%20sets%3B%7D%20%5C%5C%20%26%3C%7C%20%5Cmu%2B%20%5Clambda%20%7C(E)%2B2%5Cvarepsilon%2C%5Ctext%7B%20by%20regularity%20discussed%20above.%7D%5Cend%7Baligned%7D%24%24%20This%20shows%20%24%7C%20%5Cmu%20%2B%20%5Clambda%20%7C%24%20is%20outer%20regular.%20Similarly%20%24%24%5Cbegin%7Baligned%7D%20%7C%20%5Cmu%20%2B%5Clambda%20%7C(K)%20%26%3D%20%7C%20%5Cmu%2B%5Clambda%20%7C(E)%20-%20%7C%20%5Cmu%2B%5Clambda%20%7C(E-K)%20%5C%5C%20%26%5Cge%20%7C%20%5Cmu%2B%5Clambda%20%7C(E)%20-%20%7C%20%5Cmu%20%7C(E-K_1)%20-%20%7C%20%5Clambda%20%7C%20(E-K_2)%20%5C%5C%20%26%3E%7C%20%5Cmu%2B%5Clambda%20%7C(E)%20-%202%20%5Cvarepsilon%5Cend%7Baligned%7D.%24%24%20Thus%20%24%7C%20%5Cmu%20%2B%20%5Clambda%20%7C%24%20is%20inner%20regular%2C%20thus%20it%20is%20regular%2C%20and%20by%20definition%20%24%5Cmu%20%2B%20%5Clambda%24%20is%20regular%2C%20thus%20in%20%24M(X)%24.%20Regularity%20for%20%24%7C%5Calpha%5Cmu%7C%24%20is%20trivial%20-%20all%20we%20need%20to%20do%20is%20multiply%20everything%20by%20a%20scalar.%20In%20conclusion%2C%20%24M(X)%24%20is%20a%20vector%20space%2C%20thus%20in%20particular%20the%20difference%20of%20any%20two%20members%20of%20%24M(X)%24%20is%20still%20in%20%24M(X)%24.%0A%0A%20%20%20%20%23%23%20Exercise%206%0A%0A%20%20%20%20**Suppose%20%241%20%3C%20p%20%3C%20%5Cinfty%24%2C%20prove%20that%20%24L%5Eq(%5Cmu)%24%20is%20the%20dual%20space%20of%20%24L%5Ep(%5Cmu)%24%20even%20if%20%24%5Cmu%24%20is%20not%20%24%5Csigma%24-finite**.%0A%0A%20%20%20%20Proof%0A%0A%20%20%20%20The%20question%20essentially%20want%20us%20to%20show%20a%20similar%20statement%20as%20Theorem%206.16%3A%0A%0A%20%20%20%20%3ESuppose%20%241%20%3C%20p%20%3C%20%5Cinfty%24%2C%20%24%5Cmu%24%20is%20a%20positive%20measure%20on%20%24X%24%2C%20and%20%24%5CPhi%24%20is%20a%20bounded%20linear%20functional%20on%20%24L%5Ep(%5Cmu)%24%2C%20then%20there%20is%20a%20unique%20%24g%20%5Cin%20L%5Eq(%5Cmu)%24%20where%20%24q%24%20is%20the%20exponent%20conjugate%20to%20%24p%24%20such%20that%20%24%5CPhi(f)%20%3D%20%5Cint_Xfgd%5Cmu%24%20for%20all%20%24f%20%5Cin%20L%5Ep(%5Cmu)%24%20and%20%24%5C%7C%20%5CPhi%20%5C%7C%20%3D%20%5C%7C%20g%20%5C%7C_q%24.%0A%0A%20%20%20%20Suppose%20we%20have%20a%20pair%20of%20conjugate%20exponents%20%241%20%3C%20p%2Cq%20%3C%20%5Cinfty%24%3B%20we%20are%20given%20an%20arbitrary%20positive%20measure%20%24%5Cmu%24%20and%20a%20bounded%20linear%20functional%20%24%5CPhi%24%20on%20%24L%5Ep(%5Cmu)%24.%0A%0A%20%20%20%20By%20Theorem%206.16%2C%20on%20a%20%24%5Csigma%24-finite%20subset%20%24E%20%5Csubset%20X%24%20we%20know%20we%20have%20the%20result%2C%20i.e.%20there%20is%20some%20%24g_E%20%5Cin%20L%5Eq(E)%24%20(denote%20for%20those%20%24L%5Eq(%5Cmu)%24%20functions%20such%20that%20%24g(x)%20%3D%200%24%20for%20%24x%20%5Cnotin%20E%24)%20such%20that%20%24%5CPhi(f)%20%3D%20%5Cint_Xfg_Ed%5Cmu%24%2C%20and%20%24f%20%5Cin%20L%5Ep(E)%24.%20We%20should%20see%20that%20under%20those%20restrictions%20there%20is%20no%20need%20to%20restrict%20%24%5Cmu%24%20as%20well.%20We%20also%20have%20that%20%24%5C%7C%20g_E%20%5C%7C_q%20%3D%20%5C%7C%20%5CPhi%7C_%7BL%5Ep(E)%7D%20%5C%7C%24%20(i.e.%20%24%5CPhi%24%20restricted%20to%20those%20%24f%20%5Cin%20L%5Ep(E)%24)%20which%20%24%5Cle%20%5C%7C%20%5CPhi%20%5C%7C%24.%20In%20particular%20if%20%24E'%24%20contains%20%24E%24%2C%20then%20%24%5C%7C%20g_%7BE'%7D%20%5C%7C_q%20%5Cge%20%5C%7C%20g_E%20%5C%7C_q%24.%0A%0A%20%20%20%20Let%20%24M%20%3D%20%5Csup%5Climits_E%5C%7C%20g_E%20%5C%7C_q%24%20where%20%24E%24%20is%20%24%5Csigma%24-finite%2C%20then%20%24M%20%5Cle%20%5C%7C%20%5CPhi%20%5C%7C%24.%20Choose%20a%20sequence%20%24%5Clbrace%20E_n%20%5Crbrace%24%20so%20that%20%24%5Clim%5Climits_%7Bn%20%5Cto%20%5Cinfty%7D%5C%7C%20g_%7BE_n%7D%20%5C%7C_q%20%3D%20M%24%2C%20and%20let%20%24E%20%3D%20%5Cbigcup%5Climits_%7Bi%3D1%7D%5E%7B%5Cinfty%7DE_i%24%2C%20so%20%24E%24%20is%20%24%5Csigma%24-finite.%20By%20the%20above%20inequality%20we%20have%20%24%5C%7C%20g_E%20%5C%7C_q%20%5Cge%20%5C%7C%20g_%7BE_n%7D%20%5C%7C_q%24%20for%20any%20%24n%24%2C%20thus%20by%20definition%20of%20supremum%20and%20limit%20we%20must%20have%20%24%5C%7C%20g_E%20%5C%7C_q%20%3D%20M%24.%20Now%20if%20%24E'%24%20is%20a%20%24%5Csigma%24-finite%20subset%20containing%20%24E%24%2C%20then%20%24E'%20-%20E%24%20is%20also%20%24%5Csigma%24-finite%2C%20so%20we%20can%20write%20%24%24(%5Cint_X%7C%20g_E%20%7C%5Eq%20d%5Cmu%5Cle)%5Cint_X%7C%20g_E%20%7C%5Eq%20d%5Cmu%20%2B%20%5Cint_X%7C%20g_%7BE'-E%7D%20%7C%20%5Eqd%5Cmu%20%3D%20%5Cint_X%20%7C%20g_%7BE'%7D%20%7C%5Eqd%5Cmu%24%24%20because%20they%20only%20'concentrate'%20on%20disjoint%20sets.%20Since%20%24E'%24%20is%20%24%5Csigma%24-finite%2C%20we%20must%20have%20%24%5C%7C%20g_%7BE'%7D%20%5C%7C_q%20%5Cle%20M%20%3D%20%5C%7C%20g_%7BE%7D%20%5C%7C_q%24%2C%20in%20particular%20%24%5Cint_X%20%7C%20g_%7BE'%7D%20%7C%5Eqd%5Cmu%20%5Cle%20%5Cint_X%20%7C%20g_%7BE%7D%20%7C%5Eqd%5Cmu%24%2C%20so%20by%20a%20squeeze%20argument%20we%20see%20%24%24%5Cint_X%7C%20g_%7BE'-E%7D%20%7C%5Eqd%5Cmu%3D%200.%24%24%20And%20thus%20%24%24%5Cint_X%7C%20g_E%7C%5Eqd%5Cmu%20%3D%20%5Cint_X%20%7C%20g_%7BE'%7D%7C%5Eqd%5Cmu%2C%24%24%20which%20means%20%24g_%7BE'%7D%20%3D%20g_E%24%20a.e.%20on%20%24X%24%2C%20in%20particular%20they%20are%20identified%20in%20%24L%5Eq(%5Cmu)%24.%0A%0A%20%20%20%20The%20above%20argument%20doesn't%20really%20work%20if%20%24p%20%3D%201%24%20so%20that%20%24q%20%3D%20%5Cinfty%24%2C%20as%20the%20%24L%5E%5Cinfty%24-norm%20is%20defined%20in%20another%20way.%0A%0A%20%20%20%20Now%20for%20any%20%24f%20%5Cin%20L%5Ep(%5Cmu)%24%2C%20%24E'%20%3D%20E%20%5Ccup%20%5Clbrace%20x%20%7C%20f(x)%20%5Cne%200%20%5Crbrace%24%20%24%3DE%20%5Ccup%20%5Clbrace%20x%20%3A%20%7C%20f(x)%20%7C%20%3E%201%20%5Crbrace%20%5Ccup%20%5Clbrace%20x%20%3A%20%7C%20f(x)%20%7C%20%3E%201%2F2%20%5Crbrace%20%5Ccup%20%5Cdots%24%20is%20%24%5Csigma%24-finite%20(if%20any%20of%20the%20set%20is%20of%20infinite%20measure%2C%20%24f%24%20won't%20lie%20in%20%24L%5Ep(%5Cmu)%24).%20Since%20%24f%3D0%24%20outside%20%24E'%24%2C%20we%20can%20safely%20write%3A%20%24%24%5CPhi(f)%20%3D%20%5Cint_Xfg_%7BE'%7Dd%5Cmu%20%3D%20%5Cint_Xfg_Ed%5Cmu.%24%24%20(The%20second%20equality%20is%20due%20to%20the%20above%20discussion%20and%20%24E'%24%20contains%20%24E%24)%20So%20%24g%20%3D%20g_E%24%20will%20satisfy%20%24%5CPhi(f)%20%3D%20%5Cint_Xfgd%5Cmu%24%20for%20all%20%24f%24.%20The%20other%20properties%20are%20built-in%2C%20so%20we%20are%20finished.%0A%0A%20%20%20%20%23%23%20Exercise%207%0A%0A%20%20%20%20**Suppose%20%24%5Cmu%24%20is%20a%20complex%20Borel%20measure%20on%20%24%5B0%2C%202%5Cpi)%24%20(or%20on%20the%20unit%20circle%20%24T%24)%2C%20and%20define%20the%20Fourier%20coefficients%20of%20%24%5Cmu%24%20by**%20%24%24%5Chat%7B%5Cmu%7D(n)%20%3D%20%5Cint%20e%5E%7B-int%7Dd%5Cmu(t)%2C%20n%20%5Cin%20%5Cmathbb%7BZ%7D.%24%24%20**Assume%20that%20%24%5Chat%7B%5Cmu%7D(n)%20%5Cto%200%24%20as%20%24n%20%5Cto%20%2B%5Cinfty%24%2C%20prove%20that%20%24%5Chat%7B%5Cmu%7D(n)%20%5Cto%200%24%20as%20%24n%20%5Cto%20-%5Cinfty%24**.%0A%0A%20%20%20%20Proof%0A%0A%20%20%20%20By%20the%20hint%2C%20the%20assumption%20holds%20with%20%24fd%5Cmu%24%20in%20place%20of%20%24d%5Cmu%24%20if%20%24f%24%20is%20a%20trigonometric%20polynomial%2C%20hence%20if%20%24f%24%20is%20continuous%2C%20hence%20if%20%24f%24%20is%20bounded%20Borel%20measurable.%20Now%20by%20Theorem%206.12%2C%20there%20is%20a%20(Borel)%20measurable%20%24h%24%20such%20that%20%24%7C%20h(x)%20%7C%20%3D%201%24%20(so%20it%20is%20bounded)%20and%20%24%24d%5Cmu%20%3D%20hd%7C%20%5Cmu%20%7C%2C%24%24%20which%20can%20be%20rewritten%20as%20%24%24d%7C%20%5Cmu%20%7C%20%3D%20%5Coverline%7Bh%7Dd%5Cmu%24%24%0A%20%20%20%20because%20%24h%5Coverline%7Bh%7D%20%5Cequiv%201%24.%20%24%5Coverline%7Bh%7D%24%20is%20bounded%20Borel%20measurable%20because%20%24h%24%20is%2C%20thus%20we%20can%20replace%20%24d%5Cmu%24%20by%20%24%5Coverline%7Bh%7Dd%5Cmu%24%20which%20equals%20%24d%7C%20%5Cmu%20%7C%24.%0A%0A%20%20%20%20Now%20%24%7C%20%5Cmu%20%7C%24%20is%20a%20real%20and%20positive%20Borel%20measure.%20For%20simplicity%20of%20notation%20we%20just%20write%20%24%5Cmu%24%20and%20assume%20it%20is%20now%20a%20real%20Boreal%20measure%2C%20and%20we%20have%20%24%24%5Chat%7B%5Cmu%7D(-n)%20%3D%20%5Cint%20e%5E%7Bint%7Dd%5Cmu(t)%20%3D%20%5Cint%20%5Coverline%7Be%5E%7B-int%7D%7Dd%5Cmu(t)%20%3D%20%5Coverline%7B%5Cint%20e%5E%7B-int%7Dd%5Cmu(t)%7D%20%3D%20%5Coverline%7B%5Chat%7B%5Cmu%7D(n)%7D%24%24%20(the%20third%20equality%20(move%20the%20conjugate%20out)%20fails%20if%20%24%5Cmu%24%20is%20complex).%20Thus%20%24%24%5Clim%5Climits_%7Bn%5Cto%20-%5Cinfty%7D%5Chat%7B%5Cmu%7D(n)%20%3D%20%5Clim%5Climits_%7Bn%5Cto%20%5Cinfty%7D%5Chat%7B%5Cmu%7D(-n)%20%3D%20%5Clim%5Climits_%7Bn%5Cto%20%5Cinfty%7D%5Coverline%7B%5Chat%7B%5Cmu%7D(n)%7D%20%3D%200%24%24%20by%20assumption.%0A%0A%20%20%20%20%23%23%20Exercise%2010%0A%0A%20%20%20%20**Let%20%24(X%2C%20M%2C%20%5Cmu)%24%20be%20a%20positive%20measure%20space.%20Call%20a%20set%20%24%5CPhi%20%5Csubset%20L%5E1(%5Cmu)%24%20uniformly%20integrable%20if%20to%20each%20%24%5Cvarepsilon%20%3E%200%24%20corresponds%20a%20%24%5Cdelta%3E0%24%20such%20that**%20%24%24%7C%20%5Cint_Efd%5Cmu%20%7C%20%3C%20%5Cvarepsilon%24%24%20**whenever%20%24f%20%5Cin%20%5CPhi%24%20and%20%24%5Cmu(E)%20%3C%20%5Cdelta%24**.%0A%0A%20%20%20%201.%20**Prove%20that%20every%20finite%20subset%20of%20%24L%5E1(%5Cmu)%24%20is%20uniformly%20integrable**%3B%0A%20%20%20%202.%20**Prove%20the%20following%20convergence%20theorem%20of%20Vitali%3A%20If**%3A%0A%20%20%20%20%20%20%20%201.%20**%24%5Cmu(X)%20%3C%20%5Cinfty%24**%3B%0A%20%20%20%20%20%20%20%202.%20**%24%5Clbrace%20f_n%20%5Crbrace%24%20is%20uniformly%20integrable**%3B%0A%20%20%20%20%20%20%20%203.%20**%24f_n(x)%20%5Cto%20f(x)%24%20a.e.%20as%20%24n%20%5Cto%20%5Cinfty%24**%3B%0A%20%20%20%20%20%20%20%204.%20**%24%7C%20f(x)%20%7C%20%3C%20%5Cinfty%24%20a.e**.%2C%0A%0A%20%20%20%20%20%20%20%20**Then%20%24f%20%5Cin%20L%5E1(%5Cmu)%24%20and**%20%24%24%5Clim%5Climits_%7Bn%20%5Cto%20%5Cinfty%7D%5Cint_X%7C%20f_n%20-%20f%20%7C%20d%5Cmu%20%3D%200%3B%24%24%0A%20%20%20%203.%20**Show%20that%20part%202%20fails%20if%20%24%5Cmu%24%20is%20Lebesgue%20measure%20on%20%24(-%5Cinfty%2C%20%5Cinfty)%24%20even%20if%20%24%5Clbrace%20%5C%7C%20f_n%20%5C%7C_1%20%5Crbrace%24%20is%20assumed%20to%20be%20bounded.%20Hypothesis%201%20can%20therefore%20not%20be%20omitted%20in%20part%202**%3B%0A%20%20%20%204.%20**Show%20that%20hypothesis%204%20is%20redundant%20in%20part%202%20for%20some%20%24%5Cmu%24%20(for%20instance%2C%20for%20Lebesgue%20measure%20on%20a%20bounded%20interval)%2C%20but%20that%20there%20are%20finite%20measures%20for%20which%20the%20omission%20of%204%20would%20make%20part%202%20false**%3B%0A%20%20%20%205.%20**Show%20that%20Vitali's%20Theorem%20implies%20Lebesgue's%20Dominated%20Convergence%20Theorem%2C%20for%20finite%20measurable%20spaces.%20Construct%20an%20example%20in%20which%20Vitali's%20Theorem%20applies%20although%20the%20hypotheses%20of%20Lebesgue's%20Theorem%20do%20not%20hold**.%0A%0A%20%20%20%20Proof%0A%0A%20%20%20%20%23%23%23%20Part%201%0A%0A%20%20%20%20Let%20%24%5CPhi%20%3D%20%5Clbrace%20f_1%2C%5Cdots%2C%20f_k%20%5Crbrace%20%5Csubset%20L%5E1(%5Cmu)%24%2C%20fix%20%24%5Cvarepsilon%3E0%24.%20Since%20%24f_i%20%5Cin%20L%5E1(%5Cmu)%24%2C%20for%20each%20%24i%24%20we%20can%20find%20%24%5Cdelta_i%24%20such%20that%20%24%24%7C%20%5Cint_%7BE%7Df_i%20d%5Cmu%20%7C%20%3C%20%5Cvarepsilon%24%24%20whenever%20%24%5Cmu(E)%20%3C%20%5Cdelta_i%24%20(this%20is%20an%20exercise%20we%20did%20in%20Chapter%201).%20Take%20%24%5Cdelta%20%3D%20%5Cmin%5Climits_i%7B%5Cdelta_i%7D%24%2C%20then%20we%20get%20the%20desired%20condition.%0A%0A%20%20%20%20In%20particular%2C%20any%20%24L%5E1(%5Cmu)%24%20function%20by%20itself%20is%20uniformly%20integrable.%0A%0A%20%20%20%20%23%23%23%20Part%202%0A%0A%20%20%20%20First%20we%20need%20to%20improve%20the%20uniform%20integrability%20condition%20a%20little%20bit%3A%0A%20%20%20%20Let%20us%20fix%20%24%5Cvarepsilon%20%3E%200%24.%20Since%20%24%5Clbrace%20f_n%20%5Crbrace%24%20is%20uniformly%20integrable%2C%20there%20is%20%24%5Cdelta%20%3E%200%24%20such%20that%20%24%7C%20%5Cint_E%20f_i%20d%5Cmu%20%7C%20%3C%20%5Cvarepsilon%24%20whenever%20%24%5Cmu(E)%20%3C%20%5Cdelta%24.%20Suppose%20%24f_i%24's%20are%20real-valued%2C%20then%20%24%24%5Cint_E%7C%20f_i%20%7C%20d%5Cmu%20%3D%20%7C%20%5Cint_E%7C%20f_i%7C%20d%5Cmu%7C%20%3D%20%7C%20%5Cint_%7BE_%2B%7Df_i%20d%5Cmu-%5Cint_%7BE_-%7Df_i%20d%5Cmu%7C%2C%24%24%20where%20%24E_%2B%20%3D%20%5Clbrace%20x%20%5Cin%20E%20%7C%20f(x)%20%5Cge%200%20%5Crbrace%24%2Csimilar%20definition%20for%20%24E_-%24.%20In%20particular%20%24%5Cmu(E_%2B)%2C%5Cmu(E_-)%5Cle%20%5Cmu(E)%24%20.%20Essentially%20%24%5Cint_%7BE_%2B%7Df_i%20d%5Cmu%24%20is%20the%20same%20as%20%24%5Cint_Ef_i%5E%2Bd%5Cmu%24%20(we%20do%20this%20because%20%24f_i%5E%2B%20%5Cnotin%20%5CPhi%24%20in%20general).%20And%20we%20can%20continue%20to%20write%20%24%24%5Cle%20%7C%20%5Cint_%7BE_%2B%7Df_i%20d%5Cmu%7C%20%2B%20%7C%20%5Cint_%7BE_-%7Df_i%20d%5Cmu%7C%20%3C%202%5Cvarepsilon.%24%24%20Since%20this%20works%20for%20arbitrary%20%24%5Cvarepsilon%24%2C%20it%20follows%20that%20(for%20real%20valued%20set%20of%20%24f%24)%20we%20have%20%24%5Cint_E%20%7C%20f%20%7C%20d%5Cmu%20%3C%20%5Cvarepsilon%24%20(move%20the%20absolute%20value%20sign%20inside).%20Use%20usual%20procedure%2C%20the%20statement%20is%20also%20true%20for%20complex%20valued%20functions.%0A%0A%20%20%20%20In%20conclusion%2C%20we%20have%20a%20%24%5Cdelta%20%3E%200%24%20such%20that%20%24%24%5Cint_E%7C%20f_n%20%7C%20d%5Cmu%20%3C%20%5Cvarepsilon%24%24%20whenever%20%24%5Cmu(E)%20%3C%20%5Cdelta%24.%0A%0A%20%20%20%20Next%20we%20follow%20the%20hint%20and%20prove%20the%20theorem.%20We%20use%20Egoroff's%20Theorem%2C%20it%20appears%20in%20Chapter%203's%20exercises.%20We%20didn't%20prove%20it%2C%20but%20let's%20believe%20it%20for%20now%3A%0A%0A%20%20%20%20%3EEgoroff's%20Theorem%0A%20%20%20%20%3EIf%20%24%5Cmu(X)%20%3C%20%5Cinfty%24%2C%20%24%5Clbrace%20f_n%20%5Crbrace%24%20is%20a%20sequence%20of%20complex%20measurable%20functions%20which%20converges%20point-wise%20at%20every%20point%20of%20%24X%24%2C%20then%20for%20any%20%24%5Cvarepsilon%20%3E%200%24%2C%20there%20is%20a%20measurable%20set%20%24E%20%5Csubset%20X%24%20such%20that%20%24%5Cmu(E%5Ec)%20%3C%20%5Cvarepsilon%24%20and%20%24%5Clbrace%20f_n%20%5Crbrace%24%20converges%20uniformly%20on%20%24E%24.%0A%0A%20%20%20%20It%20follows%20that%20(we%20are%20using%20our%20%24%5Cdelta%24%20in%20place%20of%20%24%5Cvarepsilon%24%20in%20the%20statement%20of%20Egoroff's%20Theorem%2C%20and%20using%20%24E%24%20in%20place%20of%20%24E%5Ec%24)%20there%20is%20some%20%24E%24%20such%20that%20%24%5Cmu(E)%20%3C%20%5Cdelta%24%20and%20%24%5Clbrace%20f_n%20%5Crbrace%24%20converges%20uniformly%20(to%20%24f%24%2C%20by%20assumption)%20on%20%24E%5Ec%24%2C%20which%20means%20that%20there%20is%20%24N%24%20such%20that%20for%20%24n%20%3E%20N%24%20we%20have%20%24%5Cint_%7BE%5Ec%7D%7C%20f_n%20-%20f%20%7C%20d%5Cmu%20%3C%20%5Cvarepsilon%24.%20Since%20%24%7C%20f(x)%20%7C%24%20is%20by%20assumption%20bounded%20a.e.%2C%20we%20may%20assume%20that%20%24%5Cint_E%20%7C%20f%20%7C%20d%5Cmu%20%3C%20%5Cvarepsilon%24%20whenever%20%24%5Cmu(E)%20%3C%20%5Cdelta%24%20as%20well%20(or%20just%20take%20the%20minimal%20of%20the%20%24%5Cdelta%24's).%20Now%20we%20have%20everything%20we%20need%20and%3A%20%24%24%5Cbegin%7Baligned%7D%5Cint_X%7C%20f_n%20-%20f%20%7C%20d%5Cmu%20%26%3D%5Cint_%7BE%5Ec%7D%7C%20f_n%20-%20f%20%7C%20d%5Cmu%20%2B%20%5Cint_%7BE%7D%7C%20f_n%20-%20f%20%7C%20d%5Cmu%20%5C%5C%26%5Cle%20%5Cint_%7BE%5Ec%7D%7C%20f_n%20-%20f%20%7C%20d%5Cmu%20%2B%20%5Cint_E%7C%20f_n%20%7C%20d%5Cmu%20%2B%20%5Cint_%7BE%7D%7C%20f%20%7C%20d%5Cmu%20%5C%5C%26%3C%5Cvarepsilon%20%2B%20%5Cvarepsilon%20%2B%20%5Cvarepsilon%20%3D%203%5Cvarepsilon%20%5Cto%200%5Cend%7Baligned%7D.%24%24%20This%20in%20particular%20shows%20%24%7C%20f_n%20-%20f%20%7C%20%3D%20%7C%20f%20-%20f_n%20%7C%24%20is%20in%20%24L%5E1(%5Cmu)%24%3B%20since%20%24%7C%20f%20%7C%20%5Cle%20%7C%20f_n%20%7C%20%2B%20%7C%20f%20-%20f_n%20%7C%24%2C%20we%20have%20that%20%24f%24%20is%20in%20%24L%5E1(%5Cmu)%24%20because%20both%20of%20the%20summand%20terms%20are.%0A%0A%20%20%20%20%23%23%23%20Part%203%0A%0A%20%20%20%20Let%20%24%5Clbrace%20f_n%20%5Crbrace%24%20be%20indexed%20on%20the%20integers%2C%20and%20define%20each%20%24f_n%24%20as%20%24%24f_n%20%3D%20%5Cbegin%7Bcases%7D%201%2C%20n%5Cle%20x%20%5Cle%20n%2B1%5C%5C0%2C%20%5Ctext%7Botherwise%7D%20%5Cend%7Bcases%7D.%24%24%20%24%5CPhi%20%3D%20%5Clbrace%20f_n%20%5Crbrace%24%20is%20obviously%20uniformly%20integrable%20(just%20take%20%24%5Cdelta%20%3D%20%5Cvarepsilon%24).%20Also%20%24f_n%24%20clearly%20converges%20to%20%24f%20%5Cequiv%200%24%20point-wisely%2C%20in%20particular%20%24%7C%20f(x)%20%7C%20%3C%20%5Cinfty%24.%20However%20%24%5Cint_X%7C%20f_n%20-%20f%20%7C%20d%5Cmu%20%3D%201%24%20for%20any%20%24f_n%24%2C%20so%20the%20limit%20must%20be%20%241%24%2C%20not%20%240%24.%0A%0A%20%20%20%20%23%23%23%20Part%204%0A%0A%20%20%20%20Consider%20the%20Lebesgue%20measure%20%24m%24%20on%20%24%5B0%2C1%5D%24%2C%20if%20we%20can%20show%20hypothesis%201%2C2%2C3%20implies%20hypothesis%204%2C%20we%20are%20finished.%20So%20suppose%20%24%7C%20f(x)%20%7C%24%20is%20not%20bounded%20a.e.%2C%20i.e.%20%24m(F)%20%3E%200%24%20for%20%24F%20%3D%20%5Clbrace%20x%20%3A%20%7C%20f(x)%7C%20%3D%20%5Cinfty%20%5Crbrace%24.%20Fix%20%24%5Cvarepsilon%20%3E%200%24%2C%20uniform%20integrability%20and%20part%202%20says%20that%20there%20is%20%24%5Cdelta%20%3E%200%24%20such%20that%20%24%5Cint_E%20%7C%20f_n%20%7C%20dm%20%3C%20%5Cvarepsilon%24%20for%20all%20%24E%24%20with%20%24m(E)%3C%20%5Cdelta%24.%0A%0A%20%20%20%20Choose%20%24E%20%5Csubset%20F%24%20such%20that%20%240%3Cm(E)%3C%5Cdelta%24%2C%20this%20is%20always%20doable%20in%20Lebesgue%20measure%2C%20now%3A%0A%0A%20%20%20%201.%20%24%5Cint_E%20%7C%20f%20%7C%20dm%20%3D%20%5Cint_E%20%5Clim%20%7C%20f_n%20%7C%20dm%24%20%24%5Cle%20%5Climinf%20%5Cint_E%7C%20f_n%20%7C%20dm%20%3C%20%5Cvarepsilon%24%3B%0A%20%20%20%202.%20Since%20%24E%20%5Csubset%20F%24%2C%20%24%7C%20f(x)%7C%20%3D%20%5Cinfty%24%20on%20%24E%24%2C%20since%20%24m(E)%3E0%24%2C%20we%20must%20have%20%24%5Cint_E%20%7C%20f%20%7C%20dm%20%3D%20%5Cinfty%24.%0A%0A%20%20%20%20So%20there%20is%20a%20contradiction%2C%20thus%20we%20must%20have%20%24%7C%20f(x)%20%7C%20%3C%20%5Cinfty%24%20a.e.%2C%20result%20follows.%0A%0A%20%20%20%20The%20key%20of%20above%20example%20is%20that%20for%20a%20given%20%24%5Cdelta%20%3E%200%24%20we%20can%20always%20find%20a%20measurable%20set%20%24E%24%20with%20%240%20%3C%20m(E)%20%3C%20%5Cdelta%24.%20Otherwise%2C%20say%20%24%5Cmu%24%20is%20the%20measure%20on%20%24%5Cmathbb%7BR%7D%24%2C%20such%20that%20if%20%240%20%5Cin%20E%24%20then%20%24%5Cmu(E)%20%3D%201%24%2C%20otherwise%20%24%5Cmu(E)%20%3D%200%24.%20This%20is%20a%20finite%20measure%20such%20that%20the%20above%20property%20no%20longer%20exists.%20Now%20just%20take%20%24%5Clbrace%20f_n%20%5Crbrace%24%20to%20be%20the%20functions%20taking%20%24n%24%20at%20%240%24%20(and%20arbitrary%20elsewhere)%2C%20then%20they%20are%20uniformly%20integrable%3A%20for%20any%20%24%5Cvarepsilon%3E0%24%20we%20can%20take%20%24%5Cdelta%20%3D%201%24%2C%20then%20%24%5Cmu(E)%3C%5Cdelta%24%20means%20%24%5Cmu(E)%20%3D%200%24%2C%20uniform%20integrability%20follows%20trivially.%20If%20we%20let%20%24f%24%20be%20the%20function%20taking%20%24%5Cinfty%24%20at%20%240%24%2C%20then%20hypothesis%201%2C2%2C3%20are%20all%20satisfied%2C%20yet%20hypothesis%204%20is%20not.%0A%0A%20%20%20%20If%20hypothesis%204%20is%20not%20satisfied%2C%20then%20of%20course%20%24f%24%20is%20not%20in%20%24L%5E1(%5Cmu)%24%2C%20so%20we%20don't%20have%20the%20result%20of%20Vitali's%20Theorem.%0A%0A%20%20%20%20%23%23%23%20Part%205%0A%0A%20%20%20%20Suppose%20%24%5Clbrace%20f_n%20%5Crbrace%24%20is%20a%20sequence%20of%20complex%20measurable%20functions%20on%20%24X%24%20such%20that%20%24%5Clim%20f_n(x)%20%3D%20f(x)%24%20exists%20for%20every%20%24x%20%5Cin%20X%24%2C%20and%20there%20is%20a%20function%20%24g%20%5Cin%20L%5E1(%5Cmu)%24%20such%20that%20%24%7C%20f_n(x)%20%7C%20%5Cle%20g(x)%24%20for%20all%20%24n%24%20and%20%24x%24%20(these%20are%20hypothesis%20of%20LDCT).%20We%20also%20assumed%20that%20%24%5Cmu(X)%20%3C%20%5Cinfty%24%20as%20stated%20in%20the%20problem.%0A%0A%20%20%20%20So%20hypothesis%201%20and%203%20of%20Vitali's%20Theorem%20are%20built-in.%20As%20we%20proved%20in%20part%201%2C%20a%20single%20%24L%5E1(%5Cmu)%24%20function%2C%20in%20particular%20%24g%24%2C%20is%20always%20uniformly%20integrable.%20Let%20us%20fix%20%24%5Cvarepsilon%24%2C%20it%20means%20that%20there%20is%20%24%5Cdelta%20%3E%200%24%20such%20that%20%24%5Cint_E%20%7C%20g%20%7C%20d%5Cmu%20%3C%20%5Cvarepsilon%24%20whenever%20%24%5Cmu(E)%20%3C%20%5Cdelta%24%20(again%20by%20part%202).%20Since%20%24%7C%20f_n(x)%20%7C%20%3C%20g(x)%24%20by%20assumption%2C%20all%20%24f_n%24's%20are%20apparently%20%24L%5E1(%5Cmu)%24%2C%20and%20%24%24%7C%20%5Cint_E%20f_n%20d%5Cmu%20%7C%20%5Cle%20%5Cint_E%20%7C%20f_n%20%7C%20d%5Cmu%20%5Cle%20%5Cint_E%20%7C%20g%20%7C%20d%5Cmu%20%3C%20%5Cvarepsilon%2C%24%24%20so%20%24%5Clbrace%20f_n%20%5Crbrace%24%20is%20uniformly%20integrable.%20Also%20we%20have%20%24%5Cint_X%20%7C%20f%20%7C%20d%5Cmu%20%3D%20%5Cint_X%20%5Clim%20%7C%20f_n%20%7C%20d%5Cmu%24%20%24%5Cle%20%5Climinf%20%5Cint_X%7C%20f_n%20%7C%20d%5Cmu%24%20%24%3C%20%5Cint_X%20%7C%20g%20%7C%20d%5Cmu%3C%5Cinfty%24%2C%20so%20%24f%24%20is%20%24L%5E1%24%20and%20in%20particular%20%24%7C%20f(x)%20%7C%20%3C%20%5Cinfty%24%20a.e..%0A%0A%20%20%20%20So%20all%20the%20hypothesis%20of%20Vitali's%20Theorem%20holds%2C%20thus%20we%20have%20%24%24%5Clim%5Climits_%7Bn%20%5Cto%20%5Cinfty%7D%20%5Cint_X%7C%20f_n%20-%20f%20%7C%20d%5Cmu%20%3D%200%2C%24%24%20and%20thus%20%24%24%7C%20%5Cint_X%20f_n%20-%20f%20d%5Cmu%20%7C%20%5Cle%20%5Cint_X%20%7C%20f_n%20-%20f%20%7C%20d%5Cmu%20%5Cimplies%20%5Clim%7C%20%5Cint_X%20f_n%20-%20f%20d%5Cmu%20%7C%3D0%2C%24%24%20which%20then%20implies%20%24%24%5Clim%5Cint_Xf_nd%5Cmu%20%3D%20%5Cint_Xfd%5Cmu%2C%24%24%0A%20%20%20%20which%20are%20exactly%20the%20results%20of%20LDCT.%0A%0A%20%20%20%20Examine%20the%20hypothesis%20of%20the%20two%20theorems%2C%20we%20shall%20see%20Vitali's%20Theorem%20never%20requires%20a%20%24g%24%20that%20is%20%24L%5E1(%5Cmu)%24.%20From%20here%20we%20can%20construct%20something%20that%20satisfies%20Vitali%20but%20not%20Lebesgue%3A%0A%0A%20%20%20%20Let's%20take%20%24%5Cmu%24%20be%20the%20Lebesgue%20measure%20on%20%24(0%2C1)%24.%20Let%20%24%5Clbrace%20f_n%20%5Crbrace%24%20be%20the%20set%20of%20functions%20indexed%20on%20%24%5Cmathbb%7BN%7D%24%20and%20%24f_n(x)%20%3D%20%5Cfrac%7B1%7D%7Bx%7D%5Cchi_%7B%5B%5Cfrac%7B1%7D%7B1%2Bn%7D%2C%5Cfrac%7B1%7D%7Bn%7D)%7D%24.%20Each%20%24f_n%24%20is%20%24L%5E1%24%20and%20they%20have%20point-wise%20limit%20%24f%20%5Cequiv%200%24.%20It%20is%20not%20hard%20to%20see%20the%20lowest%20upper%20bound%20is%20the%20function%20%24%5Cfrac%7B1%7D%7Bx%7D%24%2C%20so%20%24g(x)%20%5Cge%20%5Cfrac%7B1%7D%7Bx%7D%24%20for%20all%20%24x%24%2C%20but%20this%20is%20a%20well-known%20function%20that%20is%20not%20in%20%24L%5E1(m)%24.%20So%20the%20hypothesis%20of%20LDCT%20are%20not%20satisfied.%20It%20remains%20to%20check%20if%20%24%5Clbrace%20f_n%20%5Crbrace%24%20is%20uniformly%20integrable%2C%20then%20the%20hypothesis%20of%20Vitali's%20Theorem%20are%20satisfied%20and%20we%20are%20done%3A%0A%0A%20%20%20%20Fix%20%24%5Cvarepsilon%3E0%24.%20For%20each%20%24n%24%2C%20%24%7C%20%5Cint_X%20f_n%20dm%20%7C%24%20%24%3D%20%5Cint%5Climits_%7B%5Cfrac%7B1%7D%7Bn%2B1%7D%7D%5E%7B%5Cfrac%7B1%7D%7Bn%7D%7D%5Cfrac%7B1%7D%7Bx%7Ddx%24%20%24%3D%20%5Cln%7B%5Cfrac%7Bn%2B1%7D%7Bn%7D%7D%24%2C%20which%20goes%20to%20%240%24%20as%20%24n%20%5Cto%20%5Cinfty%24.%20So%20we%20can%20find%20%24N%24%20such%20that%20for%20all%20%24n%20%3E%20N%24%2C%20%24%7C%20%5Cint_X%20f_n%20dm%20%7C%20%3C%20%5Cvarepsilon%24%2C%20that%20in%20particular%20for%20all%20%24n%20%3E%20N%24%20we%20can%20choose%20%24%5Cdelta_1%3D1%24%20(or%20any%20positive%20number)%2C%20so%20that%20%24%7C%20%5Cint_E%20f_n%20dm%20%7C%20%3C%20%5Cvarepsilon%24%20whenever%20%24m(E)%20%3C%20%5Cdelta_1%24.%20Now%20%24%5Clbrace%20f_1%2C%5Cdots%2C%20f_N%20%5Crbrace%24%20is%20a%20finite%20collection%20of%20%24L%5E1%24%20functions%20thus%20is%20uniformly%20integrable%2C%20i.e.%20we%20may%20find%20%24%5Cdelta_2%24%20so%20that%20%24%7C%20%5Cint_E%20f_%7B1%2C%5Cdots%2CN%7D%20dm%20%7C%20%3C%20%5Cvarepsilon%24%20whenever%20%24m(E)%20%3C%20%5Cdelta_2%24.%20Take%20%24%5Cdelta%20%3D%20%5Cmin%7B%5Cdelta_1%2C%20%5Cdelta_2%7D%24%20then%20this%20delta%20proves%20uniform%20integrability%20for%20the%20whole%20set%20%24%5Clbrace%20f_n%20%5Crbrace%24.%0A%20%20%20%20%22%22%22)%0A%20%20%20%20return%0A%0A%0A%40app.cell%0Adef%20_(mo)%3A%0A%20%20%20%20mo.md(r%22%22%22%0A%20%20%20%20%23%20Spring%20Semester%20Homework%203%0A%20%20%20%20%22%22%22)%0A%20%20%20%20return%0A%0A%0A%40app.cell%0Adef%20_(mo)%3A%0A%20%20%20%20mo.md(r%22%22%22%0A%20%20%20%20%23%23%20Exercise%201%0A%0A%20%20%20%20**Show%20that%20%24%7C%20f(x)%20%7C%20%5Cle%20(Mf)(x)%24%20at%20every%20Lebesgue%20point%20of%20%24f%24%20if%20%24f%20%5Cin%20L%5E1(%5Cmathbb%7BR%7D%5Ek)%24**.%0A%0A%20%20%20%20Proof%0A%0A%20%20%20%20Suppose%20%24f%20%5Cin%20L%5E1(%5Cmathbb%7BR%7D%5Ek)%24%20and%20%24x%24%20is%20a%20Lebesgue%20point%20of%20%24f%24.%20By%20definition%20of%20Lebesgue%20point%2C%20we%20know%20that%3A%20%24%24%5Clim_%7Br%20%5Cto%200%7D%5Cfrac%7B1%7D%7Bm(B_r)%7D%5Cint_%7BB(x%2Cr)%7D%7C%20f(y)%20-%20f(x)%20%7C%20dm(y)%20%3D%200.%24%24%20As%20usual%2C%20since%20%24x%24%20in%20the%20above%20expression%20is%20constant%2C%20we%20can%20pull%20it%20out%2C%20fix%20any%20%24r%20%3E%200%24%2C%20and%20we%20have%3A%20%24%24%5Cbegin%7Baligned%7D%26~~~~%5Cfrac%7B1%7D%7Bm(B_r)%7D%5Cint_%7BB(x%2Cr)%7D%7C%20f(y)%20-%20f(x)%20%7C%20dm(y)%5C%5C%26%5Cge%20%7C%5Cfrac%7B1%7D%7Bm(B_r)%7D%5Cint_%7BB(x%2Cr)%7D%20f(y)%20-%20f(x)%20dm(y)%20%7C%20%5C%5C%20%26%3D%20%7C%20f(x)%20-%5Cfrac%7B1%7D%7Bm(B_r)%7D%5Cint_%7BB(x%2Cr)%7Df(y)%20dm(y)%7C.%20%5Cend%7Baligned%7D%24%24%20Then%20by%20triangle%20inequality%2C%20the%20above%20inequality%2C%20and%20the%20definition%20of%20%24Mf%24%2C%20we%20have%3A%20%24%24%5Cbegin%7Baligned%7D%7C%20f(x)%20%7C%20%26%5Cle%20%7C%20f(x)%20-%5Cfrac%7B1%7D%7Bm(B_r)%7D%5Cint_%7BB(x%2Cr)%7Df(y)%20dm(y)%7C%20%2B%20%7C%20%5Cfrac%7B1%7D%7Bm(B_r)%7D%5Cint_%7BB(x%2Cr)%7Df(y)%20dm(y)%7C%20%5C%5C%20%26%5Cle%20%5Cfrac%7B1%7D%7Bm(B_r)%7D%5Cint_%7BB(x%2Cr)%7D%7C%20f(y)%20-%20f(x)%20%7C%20dm(y)%20%2B%20(Mf)(x).%5Cend%7Baligned%7D%24%24%20Pass%20%24r%20%5Cto%200%24%2C%20by%20definition%20of%20Lebesgue%20point%20the%20first%20term%20of%20the%20expression%20goes%20to%20%240%24%2C%20and%20we%20have%20%24%24%7C%20f(x)%20%7C%20%5Cle%20(Mf)(x).%24%24%0A%0A%20%20%20%20%23%23%20Exercise%2010%0A%0A%20%20%20%20**If%20%24f%20%5Cin%20%5Ctext%7BLip%7D~%7B1%7D%24%20on%20%24%5Ba%20%2C%20b%5D%24%2C%20prove%20that%20%24f%24%20is%20absolutely%20continuous%20and%20that%20%24f'%20%5Cin%20L%5E%7B%5Cinfty%7D%24**.%0A%0A%20%20%20%20Proof%0A%0A%20%20%20%20By%20definition%20of%20Lipschitz%20condition%2C%20we%20have%20that%20%24f%24%20satisfies%3A%20%24%24M_f%20%3D%20%5Csup%5Climits_%7Bs%20%5Cne%20t%7D%5Cfrac%7B%7C%20f(s)%20-%20f(t)%20%7C%7D%7B%7C%20s%20-%20t%20%7C%7D%20%3C%20%5Cinfty.%24%24%20Which%20in%20particular%20means%20%24%24%5Cinfty%3EM_f%7C%20s%20-%20t%20%7C%20%5Cge%20%7C%20f(s)%20-%20f(t)%20%7C%2C%20%5Cforall%20s%5Cne%20t%20%5Cin%20%5Ba%2Cb%5D.%24%24%20Suppose%20%24%5Cvarepsilon%20is%20given%2C%20take%20%24%5Cdelta%20%3D%20%5Cvarepsilon%24%2C%20then%20if%20%24(a_1%2Cb_1)%2C%5Cdots%2C(a_n%2Cb_n)%24%20are%20disjoint%20segments%20in%20%24%5Ba%2Cb%5D%24%20such%20that%20the%20sum%20of%20length%20is%20less%20than%20%24%5Cdelta%20%3D%20%5Cvarepsilon%24%20(and%20of%20course%2C%20we%20are%20assuming%20%24b_i%20%3E%20a_i%24%20for%20all%20%24i%24)%2C%20we%20get%3A%20%24%24%5Cbegin%7Baligned%7D%5Csum_%7Bi%20%3D%201%7D%5En%20%7C%20f(b_i)%20-%20f(a_i)%20%7C%20%26%5Cle%20%5Csum_%7Bi%3D1%7D%5EnM_f%7C%20b_i%20-%20a_i%7C%5C%5C%26%3DM_f%5Csum_%7Bi%3D1%7D%5En(b_i%20-%20a_i)%5C%5C%26%3CM_f%5Cvarepsilon%2C%5Cend%7Baligned%7D%24%24%20where%20%24M_f%24%20is%20a%20fixed%20number%20thus%20the%20last%20quantity%20goes%20to%20%240%24%2C%20then%20by%20definition%20we%20have%20that%20%24f%24%20is%20AC.%0A%0A%20%20%20%20It%20follows%20that%20%24f'%24%20exists%20a.e.%20on%20%24%5Ba%2C%20b%5D%24.%20For%20any%20point%20%24s%20%5Cin%20%5Ba%2Cb%5D%24%20such%20that%20the%20derivative%20exists%2C%20use%20the%20definition%20of%20derivative%3A%20%24%24%5Cbegin%7Baligned%7D%7C%20f'(s)%20%7C%20%26%3D%20%7C%20%5Clim_%7Bt%20%5Cto%20s%7D%5Cfrac%7Bf(t)-f(s)%7D%7Bt%20-%20s%7D%20%7C%20%5C%5C%26%3D%5Clim_%7Bt%5Cto%20s%7D%5Cfrac%7B%7C%20f(t)%20-%20f(s)%20%7C%7D%7B%7C%20t%20-%20s%20%7C%7D%20%5C%5C%26%5Cle%20%5Clim_%7Bt%20%5Cto%20s%7D%5Cfrac%7BM_f%20%7C%20t%20-%20s%20%7C%7D%7B%7C%20t%20-%20s%20%7C%7D%20%5C%5C%26%3DM_f%20%3C%20%5Cinfty.%5Cend%7Baligned%7D%24%24%20Thus%20%24f'%24%20is%20%24L%5E%7B%5Cinfty%7D%24%20by%20definition.%0A%0A%20%20%20%20%23%23%20Exercise%2011%0A%0A%20%20%20%20**Assume%20that%20%241%20%3C%20p%20%3C%20%5Cinfty%24%2C%20%24f%24%20is%20absolutely%20continuous%20on%20%24%5Ba%2C%20b%5D%24%2C%20%24f'%20%5Cin%20L%5Ep%24%2C%20and%20%24%5Calpha%20%3D%201%2Fq%24%2C%20where%20%24q%24%20is%20the%20exponent%20conjugate%20to%20%24p%24%2C%20prove%20that%20%24f%20%5Cin%20%5Ctext%7BLip%7D~%7B%5Calpha%7D%24**.%0A%0A%20%20%20%20Proof%0A%0A%20%20%20%20Since%20%24f%24%20is%20AC%20on%20%24%5Ba%2C%20b%5D%24%2C%20we%20know%20from%20Theorem%207.20%20that%20%24%24f(x)%20-%20f(a)%20%3D%20%5Cint_a%5Ex%20f'(t)dt%24%24%20for%20all%20%24x%20%5Cin%20%5Ba%2C%20b%5D%24%2C%20and%20%24f'%24%20is%20in%20%24L%5E1%24.%20For%20any%20%24x%20%5Cne%20y%20%5Cin%20%5Ba%2C%20b%5D%24%20we%20can%20take%20difference%20and%20get%20%24%24%5Cbegin%7Baligned%7Df(x)%20-%20f(y)%20%3D%20f(x)%20-%20f(a)%20-%20f(y)%20%2B%20f(a)%20%5C%5C%3D%20%5Cint_a%5Ex%20f'(t)dt%20-%20%5Cint_a%5Ey%20f'(t)dt%20%3D%20%5Cint_y%5Ex%20f'(t)dt.%5Cend%7Baligned%7D%24%24%20Since%20%24f'%24%20is%20in%20%24L%5E1%24%20we%20have%3A%20%24%24%7C%20f(x)%20-%20f(y)%20%7C%20%3D%20%7C%20%5Cint_y%5Exf'(t)dt%20%7C%20%5Cle%20%5Cint_y%5Ex%7C%20f'(t)%20%7C%20dt.%24%24%20By%20H%C3%B6lder's%20inequality%20we%20get%3A%20%24%24%3D%5Cint_y%5Ex%7C%20f'(t)%7C%20%5Ccdot%201%20dt%20%5Cle%20(%5Cint_y%5Ex%20%7C%20f'(t)%20%7C%5Ep%20dt)%5E%7B%5Cfrac%7B1%7D%7Bp%7D%7D(%5Cint_y%5Ex1dt)%5E%7B%5Cfrac%7B1%7D%7Bq%7D%7D.%24%24%20Since%20we%20are%20using%20Lebesgue%20measure%2C%20the%20last%20term%20is%20just%20%24%7C%20x%20-%20y%20%7C%5E%7B%5Cfrac%7B1%7D%7Bq%7D%7D%24%2C%20or%20%24%7C%20x%20-%20y%20%7C%5E%7B%5Calpha%7D%24.%20It%20is%20of%20course%20positive%20and%20finite%20so%20we%20can%20move%20it%20to%20the%20left%20and%20we%20get%3A%20%24%24%5Cfrac%7B%7C%20f(x)%20-%20f(y)%20%7C%7D%7B%7C%20x%20-%20y%20%7C%5E%7B%5Calpha%7D%7D%20%5Cle%20(%5Cint_y%5Ex%20%7C%20f'(t)%20%7C%5Ep%20dt)%5E%7B%5Cfrac%7B1%7D%7Bp%7D%7D%20%5Cle%20(%5Cint_a%5Eb%20%7C%20f'(t)%20%7C%5Ep%20dt)%5E%7B%5Cfrac%7B1%7D%7Bp%7D%7D%2C%24%24%20the%20RHS%20is%20exactly%20the%20%24L%5Ep%24%20norm%20of%20%24f'%24%2C%20which%20is%20finite%20by%20assumption%2C%20and%20which%20is%20a%20fixed%20number%20that%20does%20not%20depend%20on%20%24x%24%20or%20%24y%24.%20Since%20the%20inequality%20holds%20for%20any%20%24x%5Cne%20y%20%5Cin%20%5Ba%2C%20b%5D%24%2C%20we%20get%20%24%24%5Csup_%7Bx%5Cne%20y%7D%20%5Cfrac%7B%7C%20f(x)%20-%20f(y)%20%7C%7D%7B%7C%20x%20-%20y%20%7C%5E%7B%5Calpha%7D%7D%20%5Cle%20%5C%7C%20f'%20%5C%7C_p%20%3C%20%5Cinfty%24%24%20which%20means%20%24f%20%5Cin%20%5Ctext%7BLip%7D~%5Calpha%24%20by%20definition.%0A%0A%20%20%20%20%23%23%20Exercise%2016%0A%0A%20%20%20%20**Suppose%20%24E%20%5Csubset%20%5Ba%2C%20b%5D%24%2C%20%24m(E)%20%3D%200%24.%20Construct%20an%20absolutely%20continuous%20monotonic%20function%20%24f%24%20on%20%24%5Ba%2C%20b%5D%24%20so%20that%20%24f'(x)%20%3D%20%5Cinfty%24%20at%20every%20%24x%20%5Cin%20E%24**.%0A%0A%20%20%20%20Solution%0A%0A%20%20%20%20Let's%20follow%20the%20hint.%20Since%20the%20Lebesgue%20measure%20is%20(outer)%20regular%2C%20we%20can%20find%20a%20sequence%20of%20open%20sets%20%24V_1%20%5Csupset%20V_2%20%5Csupset%20%5Cdots%20%5Csupset%20E%24%20(so%20%24E%20%5Csubseteq%20%5Ccap%20V_n%24)such%20that%20%24m(V_n)%20%3C%202%5E%7B-n%7D%24.%20WLOG%20let%20us%20assume%20%24V_1%24%20lies%20entirely%20inside%20%24%5Ba%2C%20b%5D%24%2C%20otherwise%20just%20take%20intersection.%20Consider%20the%20function%20%24F%3A%5Ba%2C%20b%5D%20%5Cto%20%5Cmathbb%7BR%7D%24%20given%20by%20%24F(x)%20%3D%20%5Csum%5Climits_%7Bn%20%3D%201%7D%5E%7B%5Cinfty%7D%5Cchi_%7BV_n%7D(x)%24.%20Since%20%24%24%5Cint_a%5EbFdm%20%3D%20%5Cint_a%5Eb%5Csum%5Cchi_%7BV_n%7Ddm%20%3D%20%5Csum%5Cint_a%5Eb%5Cchi_%7BV_n%7Ddm%20%3C%20%5Csum%202%5E%7B-n%7D%20%5Cto%201%20%3C%20%5Cinfty.%24%24%20Thus%20%24F%20%5Cin%20L%5E1(m)%24.%20By%20Theorem%207.11%2C%20if%20we%20define%20%24f%3A%5Ba%2Cb%5D%20%5Cto%20%5Cmathbb%7BR%7D%24%20by%20%24%24f(x)%20%3D%20%5Cint_a%5ExFdm%24%24%20then%20%24f'(x)%20%3D%20F(x)%24%20almost%20everywhere.%0A%0A%20%20%20%20Since%20%24F%24%20is%20summation%20of%20characteristic%20functions%2C%20%24F%24%20is%20non-negative%2C%20thus%20%24f%24%20is%20non-decreasing.%20Use%20Theorem%207.18%2C%20we%20have%20that%20%24f%24%20satisfies%20condition%20c%2C%20thus%20it%20satisfies%20condition%20a%2C%20i.e.%20%24f%24%20is%20AC%20on%20%24%5Ba%2C%20b%5D%24.%20It%20remains%20to%20show%20%24f'(x)%20%3D%20%5Cinfty%24%20on%20%24E%24.%0A%0A%20%20%20%20So%20suppose%20%24x%20%5Cin%20E%24%2C%20we%20want%20to%20show%20%24f'(x)%20%3D%20%5Clim%5Climits_%7By%20%5Cto%20x%7D%5Cfrac%7Bf(y)%20-%20f(x)%7D%7By%20-%20x%7D%20%3D%20%5Cinfty%24.%20By%20construction%20%24x%20%5Cin%20V_n%24%20for%20any%20%24n%24%2C%20pick%20an%20arbitrary%20%24n%24.%20Since%20%24V_n%24%20is%20open%2C%20we%20can%20find%20an%20open%20interval%20%24U%24%20contains%20%24x%24%20lies%20inside%20%24V_n%24.%20Since%20we%20are%20taking%20%24y%20%5Cto%20x%24%2C%20we%20can%20choose%20to%20only%20care%20about%20%24y%20%5Cin%20U%24.%20So%20suppose%20%24y%20%5Cin%20U%24%2C%20from%20the%20definition%20of%20%24f%24%20we%20have%20%24f(y)%20-%20f(x)%20%3D%20%5Cint_x%5EyFdm%24.%20Since%20%24y%20%5Cin%20V_n%24%20but%20not%20necessarily%20in%20%24V_m%24%20for%20%24m%20%3E%20n%24%2C%20we%20have%20that%20%24%24%5Cint_x%5EyFdm%20%5Cge%20%5Cint_x%5Ey%20%5Csum_%7Bi%3D1%7D%5En%5Cchi_%7BV_i%7Ddm%5Cge%20n(y-x).%24%24%20The%20following%20picture%20illustrates%20why%20we%20have%20the%20inequality%3A%0A%0A%20%20%20%20%3Cimg%20src%3D%22public%2FPasted%20image%2020220327095614.png%22%20width%3D%22600%22%20%2F%3E%0A%0A%20%20%20%20The%20point%20is%20that%20by%20the%20construction%20the%20entire%20%24U%24%2C%20thus%20in%20particular%20%24%5By%2Cx%5D%24%2C%20lies%20inside%20%24V_n%24%20thus%20%24V_m%24%20for%20%24m%3Cn%24%2C%20thus%20%24%5Cint_x%5EyFdm%24%20(area%20by%20blue)%20is%20no%20less%20than%20%24%5Cint_x%5Ey%20%5Csum_%7Bi%3D1%7D%5En%5Cchi_%7BV_i%7Ddm%24%20(area%20by%20red).%0A%0A%20%20%20%20It%20follows%20that%20for%20%24y%24%20close%20to%20%24x%24%2C%20we%20have%20that%20%24f(y)%20-%20f(x)%20%5Cge%20n(y%20-%20x)%24%20for%20arbitrary%20%24n%24%2C%20which%20implies%20%24f'(x)%20%3D%20%5Clim%5Climits_%7By%20%5Cto%20x%7D%5Cfrac%7Bf(y)%20-%20f(x)%7D%7By%20-%20x%7D%20%3D%20%5Cinfty%24%20as%20desired.%0A%0A%20%20%20%20%23%23%20Exercise%2017%0A%0A%20%20%20%20**Suppose%20%24%5Clbrace%20%5Cmu_n%20%5Crbrace%24%20is%20a%20sequence%20of%20positive%20Borel%20measures%20on%20%24%5Cmathbb%7BR%7D%5Ek%24%20and**%20%24%24%5Cmu(E)%20%3D%20%5Csum_%7Bn%20%3D1%7D%5E%7B%5Cinfty%7D%5Cmu_n(E).%24%24%20**Assume%20%24%5Cmu(%5Cmathbb%7BR%7D%5Ek)%20%3C%20%5Cinfty%24.%20Show%20that%20%24%5Cmu%24%20is%20a%20Borel%20measure.%20What%20is%20the%20relation%20between%20the%20Lebesgue%20decompositions%20of%20the%20%24%5Cmu_n%24%20and%20that%20of%20%24%5Cmu%24%3F%20Prove%20that**%20%24%24(D_%7B%5Cmu%7D)(x)%20%3D%20%5Csum_%7Bn%20%3D%201%7D%5E%7B%5Cinfty%7D(D_%7B%5Cmu_n%7D)(x)%2C%20a.e.%20%5Bm%5D%2C%24%24%20**derive%20corresponding%20theorems%20for%20sequences%20%24%5Clbrace%20f_n%20%5Crbrace%24%20of%20positive%20non-decreasing%20functions%20on%20%24%5Cmathbb%7BR%7D%24%20and%20their%20sums%20%24f%20%3D%20%5Csum%20f_n%24**.%0A%0A%20%20%20%20Proof%0A%0A%20%20%20%20%23%23%23%20Part%201%0A%0A%20%20%20%20By%20the%20construction%2C%20%24%5Cmu%24%20is%20of%20course%20defined%20on%20any%20Borel%20set%20%24E%24%2C%20so%20we%20just%20need%20to%20prove%20it%20is%20a%20measure%2C%20i.e.%20has%20countable%20additivity.%20Suppose%20%24%5Clbrace%20E_i%20%5Crbrace%24%20is%20a%20collection%20of%20mutually%20disjoint%20Borel%20set%2C%20then%20%24%24%5Cbegin%7Baligned%7D%5Cmu(%5Cbigcup_%7Bi%7D%20E_i)%20%26%3D%20%5Csum_n%5Cmu_n(%5Cbigcup_i%20E_i)%26%5Ctext%7B(by%20definition)%7D%5C%5C%26%3D%5Csum_n%5Csum_i%5Cmu_n(E_i)%26%5Ctext%7B(because%20each%20%7D%5Cmu_n%5Ctext%7B%20is%20a%20measure)%7D%5C%5C%26%3D%5Csum_i%5Csum_n%5Cmu_n(E_i)%26%5Ctext%7B(because%20each%20%7D%5Cmu_n%5Ctext%7B%20is%20positive)%7D%5C%5C%26%3D%5Csum_i%5Cmu(E_i).%5Cend%7Baligned%7D%24%24%20And%20we%20are%20done.%0A%0A%20%20%20%20%23%23%23%20Part%202%0A%0A%20%20%20%20Suppose%20%24%5Clambda%24%20is%20a%20positive%20%24%5Csigma%24-finite%20measure%20on%20the%20Borel%20measure%20on%20%24%5Cmathbb%7BR%7D%5Ek%24%20(such%20measure%20exists%2C%20for%20example%2C%20the%20Lebesgue%20measure)%2C%20then%20by%20Lebesgue-Radon-Nikodym%20we%20have%20the%20Lebesgue%20decomposition%20%24%5Cmu%20%3D%20%5Cmu_a%20%2B%20%5Cmu_s%24%2C%20and%20%24%5Cmu_n%20%3D%20%5Cmu_%7Bn%2Ca%7D%20%2B%20%5Cmu_%7Bn%2Cs%7D%24%20for%20each%20%24n%24%2C%20such%20that%20%24%5Cmu_a%2C%20%5Cmu_%7Bn%2C%20a%7D%20%5Cll%20%5Clambda%24%20and%20%24%5Cmu_s%2C%20%5Cmu_%7Bn%2Cs%7D%20%5Cperp%20%5Clambda%24.%20We%20claim%20that%20%24%5Cmu_a%20%3D%20%5Csum%5Climits_n%20%5Cmu_%7Bn%2Ca%7D%24%20and%20%24%5Cmu_s%20%3D%20%5Csum%5Climits_n%5Cmu_%7Bn%2Cs%7D%24.%0A%0A%20%20%20%20First%2C%20%24%5Cmu%20%3D%20%5Csum%5Climits_n%20%5Cmu_n%24%20%24%3D%20%5Csum%5Climits_n(%5Cmu_%7Bn%2Ca%7D%20%2B%20%5Cmu_%7Bn%2Cs%7D)%24%20%24%3D%20%5Csum%5Climits_n%5Cmu_%7Bn%2Ca%7D%20%2B%20%5Csum%5Climits_n%5Cmu_%7Bn%2Cs%7D%24.%20Also%20since%20all%20%24%5Cmu_%7Bn%2C%20a%7D%24's%20are%20absolutely%20continuous%20with%20respect%20to%20%24%5Clambda%24%2C%20the%20summation%20%24%5Csum%5Climits_n%5Cmu_%7Bn%2Ca%7D%20%5Cll%20%5Clambda%24%20as%20well%2C%20and%20similarly%20we%20get%20%24%5Csum_%7Bn%7D%5Cmu_%7Bn%2Cs%7D%20%5Cperp%20%5Clambda%24.%20Thus%20%24%5Cmu%20%3D%20%5Csum%5Climits_n%5Cmu_%7Bn%2Ca%7D%20%2B%20%5Csum%5Climits_n%5Cmu_%7Bn%2Cs%7D%24%20is%20also%20a%20Lebesgue%20decomposition%2C%20but%20we%20know%20the%20decomposition%20is%20unique%2C%20thus%20we%20have%20our%20claim%20proved.%0A%0A%20%20%20%20%23%23%23%20Part%203%0A%0A%20%20%20%20By%20LRN%20and%20Theorem%207.8%2C%20and%20use%20the%20above%20notations%2C%20we%20get%20that%20for%20any%20Borel%20measure%20%24E%24%20we%20have%20%24%24%5Cmu_a(E)%20%3D%20%5Cint_ED%5Cmu_adm%2C%20%5Cforall%20E%2C%24%24%20and%20%24%24%5Cmu_%7Bn%2Ca%7D(E)%20%3D%20%5Cint_ED%5Cmu_%7Bn%2Ca%7Ddm%2C%20%5Cforall%20E.%24%24%20By%20part%202%20we%20get%20%24%24%5Cbegin%7Baligned%7D%5Cint_ED%5Cmu_adm%20%26%3D%20%5Csum_n%5Cint_ED%5Cmu_%7Bn%2Ca%7Ddm%2C%20%5Cforall%20E%5C%5C%26%3D%5Cint_E%5Csum_nD%5Cmu_%7Bn%2Ca%7Ddm%2C%20%5Cforall%20E.%5Cend%7Baligned%7D%24%24%20So%20%24%24%5Cint_E(D%5Cmu_a%20-%20%5Csum_nD%5Cmu_%7Bn%2Ca%7D)dm%20%3D%200%2C%20%5Cforall%20E%2C%24%24%20which%20implies%20%24%24(D%7B%5Cmu_a%7D)(x)%20%3D%20%5Csum_%7Bn%20%3D%201%7D%5E%7B%5Cinfty%7D(D%7B%5Cmu_%7Bn%2Ca%7D%7D)(x)%2C%20a.e.%20%5Bm%5D.%24%24%20Also%20we%20know%20from%20Theorem%207.14%20that%20%24D%5Cmu_s%24%20and%20%24D%5Cmu_%7Bn%2Cs%7D%20%3D%200%24%20a.e.%20%24%5Bm%5D%24%2C%20and%20from%20the%20definition%20of%20symmetric%20derivative%20we%20see%20%24D%5Cmu%20%3D%20D%5Cmu_a%20%2B%20D%5Cmu_s%24%20and%20%24D%5Cmu_%7Bn%7D%20%3D%20D%5Cmu_%7Bn%2Ca%7D%20%2B%20D%5Cmu_%7Bn%2Cs%7D%24%2C%20and%20we%20have%20that%20%24%24(D%7B%5Cmu%7D)(x)%20%3D%20%5Csum_%7Bn%20%3D%201%7D%5E%7B%5Cinfty%7D(D%7B%5Cmu_n%7D)(x)%2C%20a.e.%20%5Bm%5D%24%24%20as%20desired.%0A%0A%20%20%20%20%23%23%23%20Part%204%0A%0A%20%20%20%20The%20theorem%20we%20would%20like%20to%20prove%20is%20that%3A%20If%20%24%5Clbrace%20f_n%20%5Crbrace%24%20is%20a%20sequence%20of%20positive%20non-decreasing%20functions%20on%20%24%5Cmathbb%7BR%7D%24%20such%20that%20%24f%20%3D%20%5Csum%5Climits_n%20f_n%24%2C%20then%20%24f'%20%3D%20%5Csum%5Climits_n%20f'_n%24%20almost%20everywhere%20with%20respect%20to%20%24m%24.%0A%0A%20%20%20%20From%20the%20assumption%2C%20%24f%24%20is%20also%20positive%20non-decreasing%2C%20so%20%24f'%24%20and%20%24f'_n%24%20all%20exist%20a.e.%20%24%5Bm%5D%24%20(a%20result%20of%20problem%207.12).%0A%0A%20%20%20%20%24f_n%24%20(and%20thus%20%24f%24)%20are%20non-decreasing%20functions%2C%20thus%20for%20any%20partial%20sum%20we%20have%20%24%24%5Cfrac%7Bf(y)%20-%20f(x)%7D%7By-x%7D%5Cge%20%5Csum_%7Bn%3D1%7D%5EN%5Cfrac%7Bf_n(y)%20-%20f_n(x)%7D%7By-x%7D%24%24%20because%20each%20fraction%20is%20positive%20(as%20long%20as%20%24y%5Cne%20x%24%2C%20of%20course)%2C%20and%20%24f%24%20is%20defined%20to%20be%20the%20infinite%20sum.%20In%20particular%2C%20taking%20%24y%20%5Cto%20x%24%2C%20this%20means%20%24f'(x)%20%5Cge%20%5Csum%5Climits_%7Bn%3D1%7D%5ENf'_n(x)%24%20thus%20%24f'(x)%20%5Cge%20%5Csum%5Climits_%7Bn%3D1%7D%5E%7B%5Cinfty%7Df'_n(x)%24%2C%20a.e.%20%24%5Bm%5D%24%2C%20which%20implies%20%24%5Clim%5Climits_%7Bn%5Cto%20%5Cinfty%7Df'_n(x)%20%3D%200%24.%0A%0A%20%20%20%20Since%20%24%5Csum%5Climits%5ENf_n(x)%20%5Cto%20f(x)%24%2C%20we%20can%20take%20a%20subsequence%20%24N_k%24%20so%20that%20%24g_k(x)%20%3D%20f(x)%20-%20%5Csum%5Climits%5E%7BN_k%7Df_n(x)%20%3C%202%5E%7B-k%7D%24.%20Since%20%24g_k(x)%24%20can%20be%20rewritten%20as%20%24%5Csum%5Climits_%7Bn%20%3D%20N_k%2B1%7D%5E%7B%5Cinfty%7Df_n(x)%24%2C%20%24g_k%24%20is%20non-decreasing.%20By%20Weierstrass%20M%20test%2C%20%24%5Csum%20g_k%24%20converges%20(uniformly).%20Thus%20by%20the%20same%20argument%20as%20preceding%20paragraph%2C%20%24%5Clim%5Climits_%7Bk%20%5Cto%20%5Cinfty%7Dg'_k(x)%20%3D%20%5Clim%5Climits_%7Bk%20%5Cto%20%5Cinfty%7D%20(f'(x)%20-%20%5Csum%5Climits%5E%7BN_k%7Df_n'(x))%20%3D%200%24.%20This%20combine%20with%20the%20preceding%20paragraph%20give%20us%20that%20%24%24f'(x)%20%3D%20%5Csum%20f'_n(x)%24%24%20as%20desired.%0A%0A%20%20%20%20%23%23%20Exercise%2021%0A%0A%20%20%20%20**If%20%24f%24%20is%20a%20real%20function%20on%20%24%5B0%2C1%5D%24%20and**%20%24%24%5Cgamma(t)%20%3D%20t%20%2B%20if(t)%2C%24%24%20**the%20length%20of%20the%20graph%20of%20%24f%24%20is%2C%20by%20definition%2C%20the%20total%20variation%20of%20%24%5Cgamma%24%20on%20%24%5B0%2C%201%5D%24.%20Show%20that%20this%20length%20is%20finite%20if%20and%20only%20if%20%24f%20%5Cin%20BV%24%20(the%20class%20of%20all%20%24f%24%20on%20%24%5B0%2C1%5D%24%20that%20have%20bounded%20variation%20on%20%24%5B0%2C1%5D%24).%20Show%20that%20it%20is%20equal%20to**%24%24%5Cint_0%5E1%20%5Csqrt%7B1%2B%5Bf'(t)%5D%5E2%7Ddt%24%24%20**if%20%24f%24%20is%20absolutely%20continuous**.%0A%0A%20%20%20%20Proof%0A%0A%20%20%20%20If%20%24f%2C%20g%20%5Cin%20BV%24%20then%20%24%5Csup%20%5Csum%20%7C%20f(t_i)%20-%20f(t_%7Bi-1%7D)%20%7C%24%20and%20%24%5Csup%20%5Csum%20%7C%20g(t_j)%20-%20g(t_%7Bj-1%7D)%20%7C%24%20are%20both%20finite%2C%20then%20for%20any%20partition%20of%20%24%5B0%2Cx%5D%24%20we%20have%20%24%24%5Cbegin%7Baligned%7D%26~~~~~%5Csum%20%7C%20(f%2Bg)(t_k)%20-%20(f%2Bg)(t_%7Bk-1%7D)%7C%20%5C%5C%20%26%3D%20%5Csum%20%7C%20f(t_k)%20-%20f(t_%7Bk-1%7D)%20%2B%20g(t_k)%20-%20g(t_%7Bk-1%7D)%7C%20%5C%5C%20%26%5Cle%20%5Csum%20%7C%20f(t_k)%20-%20f(t_%7Bk-1%7D)%20%7C%20%2B%20%7C%20g(t_k)%20-%20g(t_%7Bk-1%7D)%7C%20%5C%5C%20%26%5Cle%20%5Csup%20%5Csum%20%7C%20f(t_i)%20-%20f(t_%7Bi-1%7D)%20%7C%20%2B%20%5Csup%20%5Csum%20%7C%20g(t_j)%20-%20g(t_%7Bj-1%7D)%20%7C%2C%5Cend%7Baligned%7D%24%24%20which%20is%20by%20assumption%20finite%2C%20and%20it%20follows%20that%20%24f%2Bg%20%5Cin%20BV%24.%20Also%20it%20is%20clear%20that%20if%20%24f%20%5Cin%20BV%24%20then%20%24cf%20%5Cin%20BV%24%20for%20any%20scaler%20%24c%24%20(just%20multiply%20everything%20by%20%24c%24).%20In%20particular%2C%20since%20the%20identity%20function%20is%20in%20%24BV%24%20(the%20total%20variation%20of%20%24%5Ctext%7Bid%7D%24%20on%20%24%5Ba%2C%20x%5D%24%20is%20just%20%24x-a%24)%2C%20we%20get%20%24%5Cgamma%20%5Cin%20BV%24%20if%20and%20only%20if%20%24f%20%5Cin%20BV%24.%0A%0A%20%20%20%20For%20the%20second%20part%2C%20let's%20first%20add%20a%20notation%20to%20make%20things%20clearer%3A%20if%20we%20denote%20the%20total%20variation%20of%20%24%5Cgamma%24%20to%20be%20%24%5CGamma%24%2C%20the%20question%20wants%20us%20to%20show%20that%20%24%24%5CGamma(1)%20%3D%20%5Cint_0%5E1%5Csqrt%7B1%2B%5Bf'(t)%5D%5E2%7Ddt.%24%24%20Another%20thing%20is%20that%20if%20%24%5Cgamma(t)%20%3D%20t%20%2B%20if(t)%24%20then%20%24%5Cgamma'(t)%20%3D%201%20%2B%20if'(t)%24%20then%20%24%7C%20%5Cgamma'(t)%20%7C%24%20%24%3D%20%5Csqrt%7B1%20%2B%20%5Bf'(t)%5D%5E2%7D%24%2C%20that%20is%20where%20this%20expression%20comes%20from.%0A%0A%20%20%20%20Now%2C%20with%20a%20very%20similar%20argument%20as%20the%20first%20part%2C%20we%20see%20the%20sum%20of%20two%20AC%20functions%20is%20still%20AC%2C%20so%20does%20multiply%20by%20a%20scaler%2C%20thus%20if%20%24f%24%20is%20AC%20then%20%24%5Cgamma%24%20is%20AC%2C%20on%20%24I%24.%20It%20follows%20that%20%24%5Cgamma%24%20is%20differentiable%20a.e.%20on%20%24I%24%2C%20%24%5Cgamma'%20%5Cin%20L%5E1%24%2C%20and%20%24%5Cgamma(x)%20-%20%5Cgamma(0)%20%3D%20%5Cint_0%5Ex%20%5Cgamma'(t)dt%24%20by%20Theorem%207.18%20(remember%20this%20direction%20does%20not%20need%20%24%5Cgamma%24%20to%20be%20non-decreasing.%20By%20the%20preceding%20paragraph%2C%20for%20any%20%24x%20%3C%20y%20%5Cin%20%5B0%2C1%5D%24%20we%20have%20%24%24%7C%20%5Cgamma(y)%20-%20%5Cgamma(x)%20%7C%20%3D%20%7C%20%5Cint_x%5Ey%5Cgamma'(t)dt%20%7C%20%5Cle%20%5Cint_x%5Ey%20%7C%20%5Cgamma'(t)%20%7C%20dt%20%3D%20%5Cint_x%5Ey%5Csqrt%7B1%2B%5Bf'(t)%5D%5E2%7Ddt.%24%24%20So%20by%20definition%20we%20have%20(%24%5Csup%24%20is%20taken%20over%20partitions%20of%20%24%5B0%2C1%5D%24)%20%24%24%5CGamma(1)%20%3D%20%5Csup%5Csum%20%7C%20%5Cgamma(t_i)%20-%20%5Cgamma(t_%7Bi-1%7D)%7C%20%5Cle%20%5Csup%20%5Cint_%7Bt_i%7D%5E%7Bt_%7Bi-1%7D%7D%5Csqrt%7B1%2B%5Bf'(t)%5D%5E2%7Ddt%2C%24%24%20where%20the%20last%20expression%20is%20by%20definition%20%24%5Cint_0%5E1%5Csqrt%7B1%2B%5Bf'(t)%5D%5E2%7Ddt%24%2C%20so%20we%20get%20%24%24%5CGamma(1)%20%5Cle%20%5Cint_0%5E1%5Csqrt%7B1%2B%5Bf'(t)%5D%5E2%7Ddt.%24%24%20Notice%20that%20for%20any%20%24x%20%3C%20y%20%5Cin%20%5B0%2C%201%5D%24%2C%20we%20have%20%24%5CGamma(y)%20-%20%5CGamma(x)%20%3D%20%5Csup%5Csum%7C%5Cgamma(t_i)%20-%20%5Cgamma(t_%7Bi-1%7D)%7C%24%20where%20the%20%24%5Csup%24%20is%20taken%20over%20partitions%20of%20%24%5Bx%2C%20y%5D%24%2C%20which%20in%20particular%20no%20less%20than%20%24%7C%20%5Cgamma(y)%20-%20%5Cgamma(x)%20%7C%24%2C%20which%20equals%20%24%5Csqrt%7B(y-x)%5E2%20%2B%20(f(y)%20-%20f(x))%5E2%7D%24%20%2C%20similar%20with%20what%20we%20did%20to%20%24%7C%20%5Cgamma'%20%7C%24.%20So%20we%20have%20%24%24%5Cbegin%7Balign*%7D%20%5Csqrt%7B(y-x)%5E2%20%2B%20(f(y)%20-%20f(x))%5E2%7D%20%26%5Cle%20%5CGamma(y)%20-%20%5CGamma(x)%2C%20%5C%5C%20%5Csqrt%7B1%2B(%5Cfrac%7Bf(y)-f(x)%7D%7By-x%7D)%5E2%7D%26%5Cle%20%5Cfrac%7B%5CGamma(y)%20-%20%5CGamma(x)%7D%7By-x%7D.%5Ctag%7B*%7D%5Cend%7Balign*%7D%24%24%20By%20Theorem%207.19%2C%20since%20we%20know%20%24%5Cgamma%24%20is%20AC%2C%20%24%5CGamma%24%20is%20non-decreasing%20and%20AC%20on%20%24I%24%2C%20so%20by%20Theorem%207.18%20we%20get%20%24%24%5CGamma(1)%20%3D%20%5Cint_0%5E1%5CGamma'(t)dt%24%24%20(%24%5CGamma(0)%24%20is%20of%20course%20%240%24).%20Since%20both%20%24f'%24%20and%20%24%5CGamma'%24%20exist%20a.e.%20on%20%24I%24%2C%20%24(*)%24%20implies%20that%20%24%5Csqrt%7B1%2B%5Bf'(x)%5D%5E2%7D%20%5Cle%20%5CGamma'(x)%24%20a.e.%20on%20%24I%24%2C%20thus%20%24%24%5CGamma(1)%20%3D%20%5Cint_0%5E1%5CGamma'(t)dt%20%5Cge%20%5Cint_0%5E1%20%5Csqrt%7B1%2B%5Bf'(x)%5D%5E2%7Ddt.%24%24%20And%20we%20are%20finished.%0A%20%20%20%20%22%22%22)%0A%20%20%20%20return%0A%0A%0A%40app.cell%0Adef%20_(mo)%3A%0A%20%20%20%20mo.md(r%22%22%22%0A%20%20%20%20%23%20Spring%20Semester%20Homework%204%0A%20%20%20%20%22%22%22)%0A%20%20%20%20return%0A%0A%0A%40app.cell%0Adef%20_(mo)%3A%0A%20%20%20%20mo.md(r%22%22%22%0A%20%20%20%20%23%23%20Exercise%201%0A%0A%20%20%20%20**Find%20a%20monotone%20class%20%24M%24%20in%20%24%5Cmathbb%7BR%7D%24%20which%20is%20not%20a%20%24%5Csigma%24-algebra%2C%20even%20though%20%24%5Cmathbb%7BR%7D%20%5Cin%20M%24%20and%20%24%5Cmathbb%7BR%7D-A%20%5Cin%20M%24%20for%20every%20%24A%20%5Cin%20M%24**.%0A%0A%20%20%20%20Solution%0A%0A%20%20%20%20Define%20%24M%24%20to%20be%20the%20collection%20%24%5Clbrace%20%5Cvarnothing%2C%20%5Cmathbb%7BR%7D%2C(-%5Cinfty%2Ca)%2C%20(-%5Cinfty%2Ca%5D%2C%20(a%2C%5Cinfty)%2C%5Ba%2C%5Cinfty)%20%7C%20a%20%5Cin%20%5Cmathbb%7BR%7D%5Crbrace%24.%20It%20is%20clearly%20not%20a%20%24%5Csigma%24-algebra%3A%20the%20difference%20of%20two%20elements%2C%20say%20%24(0%2C%5Cinfty)%20-%20(1%2C%5Cinfty)%24%2C%20is%20not%20(necessarily)%20in%20%24M%24.%20From%20the%20definition%20it%20is%20also%20easy%20to%20see%20that%20%24%5Cmathbb%7BR%7D%20%5Cin%20M%24%20and%20%24%5Cmathbb%7BR%7D%20-%20A%20%5Cin%20M%24%20whenever%20%24A%20%5Cin%20M%24%2C%20it%20remains%20to%20show%20%24M%24%20is%20a%20monotone%20class.%0A%0A%20%20%20%20Let's%2C%20for%20simplicity%20of%20notations%2C%20call%20element%20of%20%24M%24%20of%20the%20form%20%24(-%5Cinfty%2Ca)%24%20and%20%24(-%5Cinfty%2Ca%5D%24%20type%201%2C%20%24(a%2C%5Cinfty)%24%20and%20%24%5Ba%2C%5Cinfty)%24%20type%202%2C%20and%20%24%5Cvarnothing%2C%20%5Cmathbb%7BR%7D%24%20type%203.%0A%0A%20%20%20%20Suppose%20we%20have%20an%20ascending%20chain%20%24A_i%20%5Csubseteq%20A_%7Bi%2B1%7D%24%20such%20that%20each%20%24A_i%24%20is%20element%20of%20%24M%24%2C%20then%20either%20all%20%24A_i%24's%20are%20of%20type%201%20and%20type%203%2C%20or%20they%20are%20all%20of%20type%202%20and%20type%203%3A%20because%2C%20for%20example%2C%20%24(-%5Cinfty%2Ca)%24%20cannot%20contain%20%24%5Bb%2C%5Cinfty)%24%20for%20any%20%24a%2Cb%24.%20WLOG%20say%20all%20%24A_i%24's%20are%20of%20type%201%20and%20type%203%2C%20then%20if%20any%20of%20%24A_i%24%20is%20%24%5Cmathbb%7BR%7D%24%20then%20their%20union%20is%20%24%5Cmathbb%7BR%7D%24%2C%20and%20if%20all%20of%20%24A_i%24%20is%20%24%5Cvarnothing%24%20then%20their%20union%20is%20%24%5Cvarnothing%24%2C%20otherwise%20%24A_i%24's%20are%20of%20type%201%20except%20for%20finitely%20many%20of%20them%20being%20%24%5Cvarnothing%24%2C%20and%20denote%20them%20%24A_i%20%3D%20(-%5Cinfty%2Ca_i)%24%20or%20%24(-%5Cinfty%2Ca_i%5D%24.%20Then%20%24%5Ccup%20A_i%20%3D%20(-%5Cinfty%2Ca)%24%20or%20%24(-%5Cinfty%2Ca%5D%24%20where%20%24a%20%3D%20%5Clim%20a_i%24%20(possibly%20infinity%2C%20in%20this%20case%20the%20union%20is%20just%20%24%5Cmathbb%7BR%7D%24)%20depending%20on%20the%20openness%20or%20closedness%20of%20%24A_i%24's.%20In%20any%20of%20the%20cases%2C%20the%20union%20is%20in%20%24M%24.%0A%0A%20%20%20%20Suppose%20we%20have%20a%20decending%20chain%20%24B_i%20%5Csupseteq%20B_%7Bi%2B1%7D%24%20such%20that%20each%20%24B_i%24%20is%20element%20of%20%24M%24%2C%20then%20with%20the%20samiliar%20reason%20we%20can%20see%20either%20all%20%24B_i%24's%20are%20of%20type%201%20and%20type%203%2C%20or%20they%20are%20all%20of%20type%202%20and%20type%203.%20Their%20intersection%20could%20be%20%24%5Cvarnothing%24%20(if%20one%20of%20%24B_i%24%20is%20empty%2C%20or%20if%20the%20lower%20bound%20of%20%24B_i%24's%20goes%20to%20infinity%2C%20or%20if%20the%20upper%20bound%20of%20%24B_i%24's%20goes%20to%20negative%20infinity)%2C%20%24%5Cmathbb%7BR%7D%24%20(if%20all%20of%20them%20are%20%24%5Cmathbb%7BR%7D%24)%2C%20or%20some%20set%20of%20the%20type%201%20or%202.%20In%20any%20of%20the%20cases%2C%20the%20intersection%20is%20in%20%24M%24.%0A%0A%20%20%20%20%23%23%20Exercise%202%0A%0A%20%20%20%20**Suppose%20%24f%24%20is%20a%20Lebesgue%20measurable%20non-negative%20real%20function%20on%20%24%5Cmathbb%7BR%7D%24%20and%20%24A(f)%24%20is%20the%20ordinate%20set%20of%20%24f%24.%20This%20is%20the%20set%20of%20all%20points%20%24(x%2Cy)%20%5Cin%20%5Cmathbb%7BR%7D%5E2%24%20for%20which%20%240%20%3C%20y%20%3Cf(x)%24**.%0A%0A%20%20%20%201.%20**Is%20it%20true%20that%20%24A(f)%24%20is%20Lebesgue%20measurable%2C%20in%20the%20two-dimensional%20sense**%3F%0A%20%20%20%202.%20**If%20the%20answer%20to%20part%201%20is%20affirmative%2C%20is%20the%20integral%20of%20%24f%24%20over%20%24%5Cmathbb%7BR%7D%24%20equal%20to%20the%20measure%20of%20%24A(f)%24**%3F%0A%20%20%20%203.%20**Is%20the%20graph%20of%20%24f%24%20a%20measurable%20subset%20of%20%24%5Cmathbb%7BR%7D%5E2%24**%3F%0A%20%20%20%204.%20**If%20the%20answer%20to%20part%203%20is%20affirmative%2C%20is%20the%20measure%20of%20the%20graph%20equal%20to%20zero**%3F%0A%0A%20%20%20%20Solution%0A%0A%20%20%20%20%23%23%23%20Part%201%0A%0A%20%20%20%20Suppose%20first%20%24f%24%20is%20simple%20(and%20Lebesgue%20measurable%20non-negative)%2C%20then%20%24f%20%3D%20%5Csum%5Climits_i%5Ena_i%5Cchi_%7BA_i%7D%24%20where%20%24a_i%20%5Cge%200%24%20are%20the%20finitely%20many%20images%2C%20and%20%24%5Cchi_%7BA_i%7D%24%20are%20the%20preimage%20of%20%24a_i%24%20(which%20is%20by%20assumption%20measurable).%20Then%20%24A(f)%24%20is%20easily%20seen%20to%20be%20%24%5Cbigcup%5Climits_%7Bi%7D%5En(A_i%20%5Ctimes%20(0%2Ca_i))%24%2C%20which%20is%20%24%5Cmathbb%7BR%7D%5E2%24%20Lebesgue%20measurable%20by%20definition%20of%20%24%5Csigma%24-algebra%2C%20definition%20of%20measurable%20rectangle%2C%20and%20Theorem%208.11.%0A%0A%20%20%20%20Now%20we%20lift%20the%20assumption%20that%20%24f%24%20is%20simple%2C%20but%20since%20it%20is%20measurable%2C%20we%20know%20that%20%24f%24%20can%20be%20approximated%20by%20a%20non-decreasing%20sequence%20of%20simple%20functions%20%24%5Clbrace%20s_n%20%5Crbrace%24.%20Claim%20that%20%24A(f)%20%3D%20%5Ccup%20A(s_n)%24.%20If%20the%20claim%20is%20true%20then%20%24A(f)%24%20be%20a%20countable%20union%20of%20measurable%20sets%20is%20measurable.%20Proof%20of%20the%20claim%3A%0A%0A%20%20%20%201.%20From%20the%20definition%20we%20can%20see%20%24A(s_n)%20%5Csubseteq%20A(f)%24%20for%20each%20%24n%24%20because%20%24s_n%20%5Cle%20f%24%20for%20each%20%24n%24%2C%20thus%20%24A(f)%20%5Csupseteq%20%5Ccup%20A(s_n)%24%3B%0A%20%20%20%202.%20Suppose%20%24(x%2Cy)%20%5Cin%20A(f)%24%2C%20then%20%240%20%3C%20y%20%3C%20f(x)%24.%20By%20construction%20and%20definition%20of%20limit%2C%20we%20can%20pick%20%24n%24%20so%20that%20%24y%20%3C%20s_n(x)%20%5Cle%20x%24%2C%20so%20by%20definition%20%24(x%2Cy)%20%5Cin%20A(s_n)%24%2C%20thus%20%24(x%2Cy)%20%5Cin%20%5Ccup%20A(s_n)%24%2C%20thus%20%24A(f)%20%5Csubseteq%20%5Ccup%20A(s_n)%24.%0A%0A%20%20%20%20Since%20we%20have%20both%20inclusion%2C%20the%20claim%20%24A(f)%20%3D%20%5Ccup%20A(s_n)%24%20is%20proved%2C%20and%20we%20have%20that%20%24A(f)%24%20is%20(two-dimensional%20Lebesgue)%20measurable.%0A%0A%20%20%20%20%23%23%23%20Part%202%0A%0A%20%20%20%20We%20know%20%24m_2%24%20is%20completion%20of%20%24m_1%20%5Ctimes%20m_1%24%2C%20so%20%24%24m_2(A(f))%20%3D%20(m_1%5Ctimes%20m_1)(A(f))%20%3D%20%5Cint_%7B%5Cmathbb%7BR%7D%7Dm(A(f)_x)dm%24%24%20by%20definition%208.7.%20%24A(f)_x%24%20(%24x%24%20section)%20is%20defined%20to%20be%20%24%5Clbrace%20y%7C%20(x%2Cy)%20%5Cin%20A(f)%5Crbrace%24%20thus%20is%20the%20open%20segment%20%24(0%2Cf(x))%24%2C%20thus%20%24m(A(f)_x)%20%3D%20f(x)%24%2C%20plug%20this%20in%20we%20get%20%24m_2(A(f))%20%3D%20%5Cint_%7B%5Cmathbb%7BR%7D%7Df(x)dm%24%2C%20so%20the%20statement%20is%20affirmative.%0A%0A%20%20%20%20%23%23%23%20Part%203%0A%0A%20%20%20%20The%20graph%20of%20%24f%24%20is%20defined%20to%20be%20%24G%20%3D%20%5Clbrace%20(x%2Cy)%20%7C%20y%20%3D%20f(x)%20%5Crbrace%24.%20Consider%20the%20sequence%20%24%5Clbrace%20%5Cfrac%7B1%7D%7Bn%7D%20%5Crbrace%24%20and%20%24%5Ccap%20A(f%2B%5Cfrac%7B1%7D%7Bn%7D)%24%2C%20it%20is%20not%20hard%20to%20see%20%24%5Ccap%20A(f%2B%5Cfrac%7B1%7D%7Bn%7D)%20%3D%20%5Clbrace%20(x%2Cy)%20%7C%200%3Cy%20%5Cle%20f(x)%20%5Crbrace%24%20thus%20%24%5Ccap%20A(f%2B%5Cfrac%7B1%7D%7Bn%7D)%20-%20A(f)%20%3D%20%5Clbrace%20(x%2Cy)%20%7C%20y%20%3D%20f(x)%20%5Ctext%7B%20and%20%7Dy%20%5Cne%200%5Crbrace%24%2C%20so%20we%20have%20%24%24G%20%3D%20(%5Ccap%20A(f%2B%5Cfrac%7B1%7D%7Bn%7D)%20-%20A(f))%20%5Csqcup%20%5Clbrace%20(x%2C0)%20%7C%20f(x)%20%3D%200%20%5Crbrace%2C%20%24%24%20or%20equivalently%20%24%24G%20%3D%20(%5Ccap%20A(f%2B%5Cfrac%7B1%7D%7Bn%7D)%20%20-%20A(f))%20%5Csqcup%20(f%5E%7B-1%7D(0)%5Ctimes%20%5Clbrace%200%20%5Crbrace).%24%24%20Now%20it%20is%20not%20hard%20to%20see%20each%20term%20is%20Lebesgue%20measurable%2C%20thus%20%24G%24%20must%20be%20Lebesgue%20measurable.%0A%0A%20%20%20%20%23%23%23%20Part%204%0A%0A%20%20%20%20Yes%2C%20just%20use%20definition%20like%20part%202%20again%3A%20%24%24m_2(G)%20%3D%20(m_1%5Ctimes%20m_1)(G)%20%3D%20%5Cint_%7B%5Cmathbb%7BR%7D%7Dm(G_x)dm%2C%24%24%20now%20%24G_x%24%20is%20by%20definition%20%24%5Clbrace%20y%20%7C%20(x%2Cy)%20%5Cin%20G%20%5Crbrace%20%3D%20%5Clbrace%20f(x)%20%5Crbrace%24%20is%20a%20single%20point%2C%20thus%20%24m(G_x)%20%5Cequiv%200%24%20thus%20the%20integral%20equals%20%240%24%2C%20result%20follows.%0A%0A%20%20%20%20%23%23%20Exercise%203%0A%0A%20%20%20%20**Find%20an%20example%20of%20a%20positive%20continuous%20function%20%24f%24%20in%20the%20open%20unit%20square%20in%20%24%5Cmathbb%7BR%7D%5E2%24%2C%20whose%20integral%20(relative%20to%20Lebesgue%20measure)%20is%20finite%20but%20such%20that%20%24%5Cvarphi(x)%24%20(in%20the%20notation%20of%20Theorem%208.8)%20is%20infinite%20for%20some%20%24x%20%5Cin%20(0%2C1)%24**.%0A%0A%20%20%20%20Solution%0A%0A%20%20%20%20Let%20%24f(x%2Cy)%24%20be%20defined%20on%20%24(0%2C1)%20%5Ctimes%20(0%2C1)%24%20be%20positive%20continuous%2C%20then%20by%20Theorem%208.8%2C%20define%20%24%24%5Cvarphi(x)%20%3D%20%5Cint_0%5E1f_x(y)dy%2C%24%24%20then%20%24%24%5Cint_0%5E1%5Cvarphi(x)dx%20%3D%20%5Cint_0%5E1%5Cint_0%5E1f(x%2Cy)dydx.%24%24%20So%20we%20want%20an%20example%20with%20finite%20%24%5Cint_0%5E1%5Cvarphi(x)dx%24%20but%20%24%5Cint_0%5E1f_x(y)dy%20%3D%20%5Cinfty%24.%0A%0A%20%20%20%20Consider%20%24%5Cvarphi(x)%20%3D%20%5Cfrac%7B1%7D%7B%5Csqrt%7B%7C%20x-%5Cfrac%7B1%7D%7B2%7D%20%7C%7D%7D%24.%20Clearly%20%24%5Cvarphi(%5Cfrac%7B1%7D%7B2%7D)%20%3D%20%5Cinfty%24%2C%20and%20%24%5Cint_0%5E1%20%5Cfrac%7B1%7D%7B%5Csqrt%7B%7C%20x%20-%20%5Cfrac%7B1%7D%7B2%7D%7C%7D%7Ddx%24%20%24%3D%20%5Cint_0%5E%7B1%2F2%7D%20%5Cfrac%7B1%7D%7B%5Csqrt%7B%7C%20x%20-%20%5Cfrac%7B1%7D%7B2%7D%7C%7D%7Ddx%20%2B%20%5Cint_%7B1%2F2%7D%5E1%20%5Cfrac%7B1%7D%7B%5Csqrt%7B%7C%20x%20-%20%5Cfrac%7B1%7D%7B2%7D%7C%7D%7Ddx%24%20%24%3D%202%5Csqrt%7B2%7D%24%20(the%20point%20here%2C%20is%20that%20integrate%20%24%5Cfrac%7B1%7D%7Bx%5Ea%7D%24%2C%20%24a%20%3C%201%24%2C%20is%20finite%20from%20%240%24%20to%20%241%24%2C%20is%20infinite%20from%20%241%24%20to%20%24%5Cinfty%24%3B%20and%20if%20%24a%3E1%24%20then%20the%20integrant%20is%20finite%20if%20it%20is%20taken%20from%20%241%24%20to%20%24%5Cinfty%24%2C%20is%20infinite%20from%20%240%24%20to%20%241%24).%20Then%20the%20natural%20choice%20for%20is%20%24f(x%2Cy)%20%3D%20y%5E%7B%5Csqrt%7B%7C%20x%20-%20%5Cfrac%7B1%7D%7B2%7D%7C%7D%7D%20-1%24%2C%20and%20we%20have%20a%20desired%20example.%0A%0A%20%20%20%20%23%23%20Exercise%204%0A%0A%20%20%20%20**Suppose%20%241%20%5Cle%20p%20%5Cle%20%5Cinfty%24%2C%20%24f%20%5Cin%20L%5E1(%5Cmathbb%7BR%7D)%24%20and%20%24g%20%5Cin%20L%5Ep(%5Cmathbb%7BR%7D)%24**.%0A%0A%20%20%20%201.%20**Imitate%20the%20proof%20of%20Theorem%208.14%20to%20show%20that%20the%20integral%20defining%20%24(f*%20g)(x)%24%20exists%20for%20almost%20all%20%24x%24%2C%20that%20%24f*%20g%20%5Cin%20L%5Ep(%5Cmathbb%7BR%7D)%24%2C%20and%20that%20%24%5C%7C%20f*%20g%20%5C%7C_p%20%5Cle%20%5C%7C%20f%20%5C%7C_1%5C%7C%20g%5C%7C_p%24**%3B%0A%20%20%20%202.%20**Show%20that%20equality%20can%20hold%20in%20part%201%20if%20%24p%3D%201%24%20and%20if%20%24p%20%3D%20%5Cinfty%24%2C%20and%20find%20the%20conditions%20under%20which%20this%20happens**%3B%0A%20%20%20%203.%20**Assume%20%241%20%3C%20p%20%3C%20%5Cinfty%24%2C%20and%20equality%20holds%20in%20part%201%2C%20show%20that%20then%20either%20%24f%20%3D%200%24%20a.e.%20or%20%24g%20%3D%200%24%20a.e.**%3B%0A%20%20%20%204.%20**Assume%20%241%20%5Cle%20p%20%5Cle%20%5Cinfty%24%2C%20%24%5Cvarepsilon%200%24%2C%20and%20show%20that%20there%20exist%20%24f%20%5Cin%20L%5E1(%5Cmathbb%7BR%7D)%24%20and%20%24g%20%5Cin%20L%5Ep(%5Cmathbb%7BR%7D)%24%20such%20that%20%24%5C%7C%20f%20*%20g%20%5C%7C_p%20%3E%20(1-%5Cvarepsilon)%5C%7C%20f%20%5C%7C_1%5C%7C%20g%20%5C%7C_p%24**.%0A%0A%20%20%20%20Proof%0A%0A%20%20%20%20%23%23%23%20Part%201%0A%0A%20%20%20%20As%20discussed%20in%20the%20proof%20of%20Theorem%208.14%2C%20we%20can%20assume%20%24f%2Cg%24%20are%20Borel.%20We%20separate%20the%20cases%20when%20%24p%20%3D%20%5Cinfty%24%20and%20%241%20%5Cle%20p%20%3C%20%5Cinfty%24%3A%0A%0A%20%20%20%201.%20If%20%24p%20%3D%20%5Cinfty%24%2C%20then%20%24%24%5Cbegin%7Baligned%7D%7C%20f*%20g%20(x)%7C%20%26%3A%3D%20%7C%5Cint_%7B%5Cmathbb%7BR%7D%7Df(x-y)g(y)dy%7C%20%5C%5C%20%26%5Cle%20%5Cint_%7B%5Cmathbb%7BR%7D%7D%7C%20f(x-y)%20%5C%7C%20g(y)%7C%20dy%20%20%26%20%5Ctext%7B(product%20is%20in%20%7D%20L%5E1%20%5Ctext%7B)%7D%5C%5C%20%26%5Cle%20%5C%7C%20g%20%5C%7C_%7B%5Cinfty%7D%5Cint_%7B%5Cmathbb%7BR%7D%7D%7C%20f(x-y)%20%7C%20dy%20%5C%5C%26%3D%20%5C%7C%20g%20%5C%7C_%7B%5Cinfty%7D%5C%7C%20f%20%5C%7C_1%26%5Ctext%7B(translate%20invariant)%7D%5Cend%7Baligned%7D%24%24%20which%20proves%20all%20three%20statements%3B%0A%20%20%20%202.%20If%20%241%20%5Cle%20p%20%3C%20%5Cinfty%24%2C%20write%20%24g%24%20as%20a%20summation%20%24g_1%20%2B%20g_2%24%20where%20%24g_1%20%5Cin%20L%5E1(%5Cmathbb%7BR%7D)%24%20and%20%24g_2%20%5Cin%20L%5E%7B%5Cinfty%7D(%5Cmathbb%7BR%7D)%24%20except%20for%20possibly%20on%20a%20set%20of%20measure%20zero%20(if%20%24g%24%20cannot%20be%20written%20in%20this%20way%20then%20it%20means%20it%20attains%20infinity%20on%20a%20set%20of%20positive%20measure%2C%20meaning%20%24g%24%20is%20not%20%24L%5Ep%24)%2C%20then%20%24f*%20g(x)%24%20%24%3D%20%5Cint%20f(x-y)g(y)%20dy%24%20%24%3D%20%5Cint%20f(x-y)(g_1(y)%20%2B%20g_2(y))dy%24.%20Also%20it%20is%20Borel%20by%20the%20proof%20of%20Theorem%208.14%2C%20thus%20the%20integral%20defining%20this%20convolution%20exists%20almost%20everywhere%20(where%20%24g%20%3D%20g_1%2Bg_2%24)%20by%20Theorem%208.14%20(%24L_1%24%20part%20is%20well-defined)%20and%20above%20(%24L%5E%7B%5Cinfty%7D%24%20part%20is%20well-defined).%0A%20%20%20%20%09Let%20%24q%24%20be%20%24p%24's%20conjugate%20exponent%2C%20write%20%24%24%5Cbegin%7Baligned%7D%20%26%5Cint_%7B%5Cmathbb%7BR%7D%7D%7C%20f(x-y)g(y)%7C%20dy%20%3D%20%5Cint_%7B%5Cmathbb%7BR%7D%7D%7C%20f(x-y)%20%7C%5E%7B1%2Fq%7D%20%7C%20f(x-y)%20%7C%5E%7B1%2Fp%7D%20%7C%20g(y)%20%7C%20dy%20%5C%5C%26%5Cle%20(%5Cint_%7B%5Cmathbb%7BR%7D%7D%7C%20f(x-y)%20%7C%20dy)%5E%7B1%2Fq%7D%20(%5Cint_%7B%5Cmathbb%7BR%7D%7D%7C%20f(x-y)%20%5C%7C%20g(y)%7C%5Ep%20dy)%5E%7B1%2Fp%7D%20%26%5Ctext%7B(H%C3%B6lder)%7D%20%5C%5C%26%3D%5C%7C%20f%5C%7C_1%5E%7B1%2Fq%7D(%5Cint_%7B%5Cmathbb%7BR%7D%7D%7C%20f(x-y)%20%5C%7C%20g(y)%7C%5Ep%20dy)%5E%7B1%2Fp%7D%5Cend%7Baligned%7D%24%24%20so%20now%20%24%24%5Cbegin%7Baligned%7D%5Cint_%7B%5Cmathbb%7BR%7D%7D%7C%20(f*%20g)(x)%20%7C%5Ep%20dx%20%26%3D%5Cint_%7B%5Cmathbb%7BR%7D%7D(%7C%20%5Cint_%7B%5Cmathbb%7BR%7D%7D%20f(x-y)g(y)dy%7C%5Ep)%20dx%20%5C%5C%20%26%5Cle%20%5Cint_%7B%5Cmathbb%7BR%7D%7D(%5Cint_%7B%5Cmathbb%7BR%7D%7D%7C%20f(x-y)g(y)%7C%20dy)%5Ep%20dx%20%5C%5C%20%26%5Cle%20%5Cint_%7B%5Cmathbb%7BR%7D%7D%20(%5C%7C%20f%20%5C%7C_1%5E%7Bp%2Fq%7D%5Cint_%7B%5Cmathbb%7BR%7D%7D%7C%20f(x-y)%20%5C%7C%20g(y)%20%7C%5Epdy)%20dx%20%26%5Ctext%7B(by%20above)%7D%20%5C%5C%20%26%3D%5C%7C%20f%20%5C%7C_1%5E%7Bp%2Fq%7D%20%5Cint_%7B%5Cmathbb%7BR%7D%7D%5Cint_%7B%5Cmathbb%7BR%7D%7D%20%7C%20f(x-y)%20%5C%7C%20g(y)%7C%5Ep%20dxdy%20%26%5Ctext%7B(Fubini)%7D%20%5C%5C%26%3D%5C%7C%20f%20%5C%7C_1%5E%7Bp%2Fq%7D%5Ccdot%5C%7C%20f%5C%7C_1%5Ccdot%5C%7C%20g%20%5C%7C_p%5E%7Bp%7D%20%5C%5C%26%3D%5C%7C%20f%20%5C%7C_1%5E%7Bp%7D%5Ccdot%20%5C%7C%20g%20%5C%7C_p%5E%7Bp%7D%20%5Cend%7Baligned%7D%24%24%20which%20proves%20the%20other%20two%20statements.%20Last%20step%20is%20due%20to%20%24p%2Cq%24%20conjugate%20exponents%20%24%5Cimplies%20%5Cfrac%7B1%7D%7Bp%7D%20%2B%20%5Cfrac%7B1%7D%7Bq%7D%20%3D%201%24%20%24%5Cimplies%20q%20%3D%20%5Cfrac%7Bp%7D%7Bp-1%7D%24%20%24%5Cimplies%20%5Cfrac%7Bp%7D%7Bq%7D%2B1%20%3D%20p%24.%0A%0A%20%20%20%20%23%23%23%20Part%202%0A%0A%20%20%20%20Examine%20the%20proof%20of%20Theorem%208.14%20and%20part%201.%20If%20%24p%20%3D%201%24%2C%20the%20only%20inequality%20appears%20is%20that%20%24%24%5Cint_%7B%5Cmathbb%7BR%7D%7D%7C%20h(x)%20%7C%20dx%20%5Cle%20%5Cint_%7B%5Cmathbb%7BR%7D%7Ddx%20%5Cint_%7B%5Cmathbb%7BR%7D%7D%7C%20F(x%2Cy)%20%7C%20dy%2C%24%24%20and%20equality%20is%20attained%20as%20long%20as%20both%20%24f%2C%20g%24%20are%20non-negative.%0A%0A%20%20%20%20If%20%24p%20%3D%20%5Cinfty%24%2C%20the%20inequalities%20appear%20are%3A%0A%0A%20%20%20%201.%20%24%7C%5Cint_%7B%5Cmathbb%7BR%7D%7Df(x-y)g(y)dy%7C%20%5Cle%20%5Cint_%7B%5Cmathbb%7BR%7D%7D%7C%20f(x-y)%20%5C%7C%20g(y)%7C%20dy%24%2C%20which%20attains%20equality%20again%20if%20%24f%2Cg%24%20are%20non-negative%3B%0A%20%20%20%202.%20%24%7C%20g(y)%20%7C%20%5Cle%20%5C%7C%20g%20%5C%7C_%7B%5Cinfty%7D%24%2C%20which%20attains%20equality%20if%20%24g%24%20is%20a%20constant.%0A%0A%20%20%20%20Put%20them%20together%2C%20in%20the%20%24p%20%3D%20%5Cinfty%24%20case%2C%20equality%20is%20attained%20if%20%24f%24%20is%20non-negative%2C%20and%20%24g%24%20is%20a%20non-negative%20constant%20function.%0A%0A%20%20%20%20%23%23%23%20Part%203%0A%0A%20%20%20%20Examine%20the%20proof%20of%20part%202%20of%20part%201%2C%20we%20can%20see%20the%20equality%20holds%20only%20if%20we%20have%20%24%24%5Cbegin%7Baligned%7D%26%5Cint_%7B%5Cmathbb%7BR%7D%7D%7C%20f(x-y)%20%7C%5E%7B1%2Fq%7D%20%7C%20f(x-y)%20%7C%5E%7B1%2Fp%7D%20%7C%20g(y)%20%7C%20dy%20%5C%5C%26%3D%20(%5Cint_%7B%5Cmathbb%7BR%7D%7D%7C%20f(x-y)%20%7C%20dy)%5E%7B1%2Fq%7D%20(%5Cint_%7B%5Cmathbb%7BR%7D%7D%7C%20f(x-y)%20%5C%7C%20g(y)%7C%5Ep%20dy)%5E%7B1%2Fp%7D%2C%5Cend%7Baligned%7D%24%24%20i.e.%20equality%20in%20H%C3%B6lder%20holds%2C%20so%20we%20need%20to%20have%2C%20for%20almost%20all%20%24x%24%20and%20almost%20all%20%24y%24%3A%20%24%24a(x)f(x-y)%20%3D%20b(x)f(x-y)g(y)%5Ep%24%24%20where%20%24a(x)%2C%20b(x)%24%20are%20numbers%20depending%20on%20%24x%24%20that%20are%20not%20both%20zero.%20Fix%20an%20%24x_0%24%20such%20that%20the%20equality%20holds%20for%20%24x_0%24%20and%20almost%20all%20%24y%24%3A%0A%0A%20%20%20%201.%20Suppose%20%24a(x_0)%24%20and%20%24b(x_0)%24%20are%20both%20non-zero%2C%20and%20suppose%20both%20%24f%2Cg%24%20are%20not%20%240%24%20a.e.%2C%20then%20we%20can%20cancel%20things%20out%20and%20get%20%24c%20%3D%20g(y)%5Ep%24%20for%20some%20non-zero%20constant%20%24c%24%2C%20but%20then%20%24g%24%20is%20not%20%24L%5Ep%24%3B%0A%20%20%20%202.%20Suppose%20%24b(x_0)%20%3D%200%24%2C%20then%20%24a(x_0)%24%20is%20not%20%240%24%2C%20thus%20we%20need%20to%20have%20%24f(x_0-y)%20%3D%200%24%20a.e.%2C%20which%20implies%20%24f%20%3D%200%24%20a.e.%3B%0A%20%20%20%203.%20Suppose%20%24a(x_0)%20%3D%200%24%2C%20then%20%24b(x_0)%24%20is%20not%20%240%24%2C%20then%20%24f(x_0-y)g(y)%5Ep%20%3D%200%24%20a.e.%2C%20the%20only%20problematic%20case%20is%20that%20%24f(x_0-y)%2Cg(y)%24%20both%20not%20%240%24%20a.e.%2C%20but%20the%20intersection%20of%20the%20set%20where%20%24f(x_0-y)%2Cg(y)%24%20attains%20non-zero%20values%20is%20of%20measure%20zero%20(for%20example%2C%20%24f(x_0-y)%20%3D%200%24%20if%20%24y%20%5Cin%20(-%5Cinfty%2C0)%24%20and%20%24g(y)%20%3D%200%24%20if%20%24y%20%5Cin(0%2C%5Cinfty)%24).%0A%20%20%20%20%09Suppose%20this%20is%20the%20case%2C%20i.e.%20%24f(x_0-y)%20%5Cbegin%7Bcases%7D%3D0%2C%20y%20%5Cin%20E%5C%5C%5Cne%200%2Cy%20%5Cin%20E'%5Cend%7Bcases%7D%24%20and%20%24g(y)%20%5Cbegin%7Bcases%7D%3D0%2C%20y%20%5Cin%20E'%5C%5C%5Cne%200%2Cy%20%5Cin%20E%5Cend%7Bcases%7D%24%20for%20some%20%24E%2C%20E'%24%20both%20of%20positive%20measure%20and%20%24m(%5Cmathbb%7BR%7D-E-E')%20%3D%200%24.%20Pick%20any%20%24x_1%20%5Cne%20x_0%24%20(such%20that%20the%20equality%20in%20H%C3%B6lder%20should%20hold)%2C%20then%20%24f(x_1%20-y)%20%5Cbegin%7Bcases%7D%3D%200%2C%20y%20%5Cin%20E%2Bx_1-x_0%5C%5C%20%5Cne%200%2C%20y%20%5Cin%20E'%2Bx_1-x_0%5Cend%7Bcases%7D%24.%20Apparently%20%24(E'%2Bx_1-x_0)%5Ccap%20E%20%3D%20E_%7Bx_1%7D%24%20has%20positive%20measure.%20Pick%20any%20%24y%20%5Cin%20E_%7Bx_1%7D%24%2C%20then%20by%20construction%20%24f(x_1%20-%20y)%5Cne%200%24%20and%20%24g(y)%20%5Cne%200%24.%20In%20particular%2C%20if%20we%20consider%20%24a(x_1)f(x_1-y)%20%3D%20b(x_1)f(x_1-y)g(y)%5Ep%24%2C%20then%20%24a(x_1)%20%3D%20b(x_1)%20g(y)%5Ep%24.%20Since%20%24g(y)%20%5Cne%200%24%2C%20%24a(x_1)%20%3D%200%24%20if%20and%20only%20if%20%24b(x_1)%20%3D%200%24%20(which%20shouldn't%20be%20the%20case)%2C%20thus%20neither%20of%20them%20is%20zero%2C%20thus%20%24g(y)%5Ep%20%5Cequiv%20c%5Ep%20%5Cimplies%20g(y)%20%5Cequiv%20c%24%20on%20%24E_%7Bx_1%7D%24%20for%20some%20non-zero%20constant%20%24c%24.%0A%20%20%20%20%09Now%2C%20remember%20this%20works%20for%20almost%20all%20%24x_1%20%5Cin%20%5Cmathbb%7BR%7D%24.%20By%20choosing%20all%20the%20possible%20%24x_1%24's%2C%20we%20should%20see%20that%20%24g(y)%20%3D%20c%24%20for%20all%20%24y%24%20in%20%24E%24%20(because%20%24%5Clbrace%20E_%7Bx_1%7D%20%7C%20x_1%20%5Cin%20%5Cmathbb%7BR%7D%20%5Crbrace%24%20covers%20%24E%24).%20If%20%24m(E)%20%3D%20%5Cinfty%24%20then%20%24g%20%5Cnotin%20L%5Ep%24%20gives%20a%20contradiction.%20If%20not%2C%20do%20the%20same%20but%20change%20place%20of%20%24f%2Cg%24%2C%20we%20get%20that%20%24m(E')%20%3C%20%5Cinfty%24%20otherwise%20%24f%20%5Cnotin%20L%5E1%24.%20But%20then%20both%20%24m(E)%2Cm(E')%3C%5Cinfty%24%2C%20which%20is%20also%20impossible.%0A%0A%20%20%20%20%23%23%23%20Part%204%0A%0A%20%20%20%201.%20Suppose%20%24p%20%3D%20%5Cinfty%24%2C%20we%20can%20take%20%24f%20%3D%20%5Cchi_%7B%5B0%2C1%5D%7D%24%20and%20%24g%20%3D%201%24%2C%20then%20%24(f*%20g)(x)%20%3D%20%5Cint_%7B%5Cmathbb%7BR%7D%7Df(x-y)g(y)dy%24%20%24%3D%5Cint_%7B%5Cmathbb%7BR%7D%7Df(x-y)dy%24%20%24%3D%5Cint_0%5E11dy%24%20%24%3D1%24%20and%20%24%5C%7C%20f%20%5C%7C_1%20%5C%7C%20g%20%5C%7C_p%20%3D%201%24%20as%20well%2C%20thus%20%24%5C%7C%20f%20*%20g%20%5C%7C_p%20%3E%20(1-%5Cvarepsilon)%5C%7C%20f%20%5C%7C_1%5C%7C%20g%20%5C%7C_p%24%20holds%20for%20any%20positive%20%24%5Cvarepsilon%24%3B%0A%20%20%20%202.%20Suppose%20%241%20%5Cle%20p%20%3C%20%5Cinfty%24.%20Let%20%24a%24%20be%20a%20positive%20real%20number%20to%20be%20defined%2C%20define%20%24f%20%3D%20%5Cfrac%7B1%7D%7B2a%7D%5Cchi_%7B%5B-a%2Ca%5D%7D%24%2C%20define%20%24g%20%3D%20%5Csqrt%5Bp%5D%7B%5Cfrac%7B1%7D%7B2%7D%7D%5Cchi_%7B%5B-1%2C1%5D%7D%24.%20Clearly%20%24f%24%20is%20%24L%5E1%24%20and%20%24g%24%20is%20%24L%5Ep%24.%20Also%20%24%5C%7C%20f%20%5C%7C_1%20%3D%201%24%20no%20matter%20which%20%24a%24%20we%20choose%2C%20and%20%24%5C%7C%20g%20%5C%7C_p%20%3D%20(%5Cint_%7B-1%7D%5E1%20%5Cfrac%7B1%7D%7B2%7D%20dx)%5Ep%20%3D%201%24.%0A%20%20%20%20%09By%20definition%2C%20%24%24%5Cbegin%7Baligned%7D(f%20*%20g%20)(x)%20%26%3D%20%5Cint_%7B-%5Cinfty%7D%5E%7B%5Cinfty%7D%5Cfrac%7B1%7D%7B2a%7D%5Cchi_%7B%5B-a%2Ca%5D%7D(x-y)%5Csqrt%5Bp%5D%7B%5Cfrac%7B1%7D%7B2%7D%7D%5Cchi_%7B%5B-1%2C1%5D%7D(y)dy%20%5C%5C%20%26%3D%5Cfrac%7B1%7D%7B2%5Csqrt%5Bp%5D%7B2%7Da%7D%5Cint_%7B-1%7D%5E1%20%5Cchi_%7B%5B-a%2Ca%5D%7D(x-y)dy%5Cend%7Baligned%7D%24%24%20%24%5Cchi_%7B%5B-a%2Ca%5D%7D(x-y)%20%3D%201%24%20if%20and%20only%20if%20%24-a%20%5Cle%20x%20-%20y%20%5Cle%20a%24%20%24%5Cimplies%20-a%2Bx%20%5Cle%20y%20%5Cle%20a%2Bx%24%2C%20so%20it%20can%20be%20rewrite%20as%20%24%5Cchi_%7B%5B-a%2Bx%2Ca%2Bx%5D%7D(y)%24.%20The%20valuation%20of%20%24(f*%20g)%24%20depends%20on%20the%20intersection%20of%20%24%5B-a%2Bx%2Ca%2Bx%5D%24%20and%20%24%5B-1%2C1%5D%24%3A%20%24%24(f*%20g)(x)%20%3D%20%5Cbegin%7Bcases%7D0%2C%26x%5Cle%20-1-a%5C%5C%20%5Cfrac%7Ba%2Bx%2B1%7D%7B2%5Csqrt%5Bp%5D%7B2%7Da%7D%2C%26-1-a%3Cx%5Cle%20-1%2Ba%20%5C%5C%20%5Cfrac%7B2a%7D%7B2%5Csqrt%5Bp%5D%7B2%7Da%7D%20%3D%20%5Cfrac%7B1%7D%7B%5Csqrt%5Bp%5D%7B2%7D%7D%2C%26-1%2Ba%3Cx%5Cle%201-a%20%5C%5C%20%5Cfrac%7B1%2Ba-x%7D%7B2%5Csqrt%5Bp%5D%7B2%7Da%7D%2C%261-a%3Cx%5Cle%201%2Ba%5C%5C0%2C%26x%3E1%2Ba%5Cend%7Bcases%7D%24%24%20We%20can%20calculate%20%24%24%5Cbegin%7Baligned%7D%5C%7C%20f*%20g%20%5C%7C_p%5Ep%20%26%3D%20%5Cint_%7B%5Cmathbb%7BR%7D%7D%7C%20f*%20g%20%7C%5Epdx%20%5C%5C%20%26%3D%5Cint_%7B-1-a%7D%5E%7B-1%2Ba%7D%5Cfrac%7B(a%2Bx%2B1)%5Ep%7D%7B2%5E%7Bp%2B1%7Da%5Ep%7Ddx%20%2B%20%5Cint_%7B-1%2Ba%7D%5E%7B1-a%7D%5Cfrac%7B1%7D%7B2%7Ddx%20%2B%20%5Cint_%7B1-a%7D%5E%7B1%2Ba%7D%5Cfrac%7B(1%2Ba-x)%5Ep%7D%7B2%5E%7Bp%2B1%7Da%5Ep%7Ddx%20%5C%5C%20%26%3D%20%5Cint_%7B-1-a%7D%5E%7B-1%2Ba%7D%5Cfrac%7B(a%2Bx%2B1)%5Ep%7D%7B2%5Epa%5Ep%7Ddx%20%2B%201%20-%20a%5Cend%7Baligned%7D%24%24%20Now%20take%20%24a%24%20so%20that%20%24a%20%3C%201-(1-%5Cvarepsilon)%5Ep%24%2C%20also%20notice%20%24a%2Bx%2B1%3E0%24%20for%20%24-1-a%20%3C%20x%20%5Cle%20-1%2Ba%24%2C%20then%20%24%5C%7C%20f%20*%20g%20%5C%7C_p%5Ep%20%5Cge%201-a%20%3E%20(1-%5Cvarepsilon)%5Ep%24%2C%20which%20gives%20%24%24%5C%7C%20f%20*%20g%20%5C%7C_p%20%3E(1-%5Cvarepsilon)%20%3D%20(1-%5Cvarepsilon)%5C%7C%20f%20%5C%7C_1%5C%7C%20g%20%5C%7C_p.%24%24%0A%0A%20%20%20%20%23%23%20Exercise%205%0A%0A%20%20%20%20**Let%20%24M%24%20be%20the%20Banach%20space%20of%20all%20complex%20Borel%20measures%20on%20%24%5Cmathbb%7BR%7D%24.%20The%20norm%20in%20%24M%24%20is%20%24%5C%7C%20%5Cmu%20%5C%7C%20%3D%20%7C%20%5Cmu%20%7C%20(%5Cmathbb%7BR%7D)%24.%20Associate%20to%20each%20Borel%20set%20%24E%20%5Csubset%20%5Cmathbb%7BR%7D%24%20the%20set%20%24E_2%20%3D%20%5Clbrace(x%2Cy)%20%7C%20x%2By%20%5Cin%20E%20%5Crbrace%20%5Csubset%20%5Cmathbb%7BR%7D%5E2%24.%20If%20%24%5Cmu%24%20and%20%24%5Clambda%20%5Cin%20M%24%2C%20define%20their%20convolution%20%24%5Cmu%20*%20%5Clambda%24%20to%20be%20the%20set%20function%20given%20by**%20%24%24(%5Cmu%20*%20%5Clambda)(E)%20%3D%20(%5Cmu%20%5Ctimes%20%5Clambda)(E_2)%24%24%20**for%20every%20Borel%20set%20%24E%20%5Csubset%20%5Cmathbb%7BR%7D%24%3B%20%24%5Cmu%20%5Ctimes%20%5Clambda%24%20is%20as%20in%20Definition%208.7**.%0A%0A%20%20%20%201.%20**Prove%20that%20%24%5Cmu%20*%20%5Clambda%20%5Cin%20M%24%20and%20that%20%24%5C%7C%20%5Cmu%20*%20%5Clambda%20%5C%7C%20%5Cle%20%5C%7C%20%5Cmu%20%5C%7C%20%5C%7C%20%5Clambda%20%5C%7C%24**%3B%0A%20%20%20%202.%20**Prove%20that%20%24%5Cmu%20*%20%5Clambda%24%20is%20the%20unique%20%24v%20%5Cin%20M%24%20such%20that**%20%24%24%5Cint%20f%20dv%20%3D%20%5Cint%20%5Cint%20f(x%2By)d%5Cmu(x)d%5Clambda(y)%24%24%20**for%20every%20%24f%20%5Cin%20C_0(%5Cmathbb%7BR%7D)%24%20(all%20integrals%20extend%20over%20%24%5Cmathbb%7BR%7D%24)**%3B%0A%20%20%20%203.%20**Prove%20that%20convolution%20in%20%24M%24%20is%20commutative%2C%20associative%2C%20and%20distributive%20with%20respect%20to%20addition**%3B%0A%20%20%20%204.%20**Prove%20the%20formula**%20%24%24(%5Cmu%20*%20%5Clambda)(E)%20%3D%20%5Cint%20%5Cmu(E%20-t)d%5Clambda(t)%24%24%20**for%20every%20%24%5Cmu%24%20and%20%24%5Clambda%20%5Cin%20M%24%20and%20every%20Borel%20set%20%24E%24.%20Here%20%24E%20-%20t%20%3D%20%5Clbrace%20x%20-%20t%20%7C%20x%20%5Cin%20E%20%5Crbrace%24**%3B%0A%20%20%20%205.%20**Define%20%24%5Cmu%24%20to%20be%20discrete%20if%20%24%5Cmu%24%20is%20concentrated%20on%20a%20countable%20set%3B%20define%20%24%5Cmu%24%20to%20be%20continuous%20if%20%24%5Cmu(%5Clbrace%20x%20%5Crbrace)%20%3D%200%24%20for%20every%20point%20%24x%20%5Cin%20%5Cmathbb%7BR%7D%24%3B%20let%20%24m%24%20be%20Lebesgue%20measure%20on%20%24%5Cmathbb%7BR%7D%24%20(note%20that%20%24m%20%5Cnotin%20M%24).%20Prove%20that%20%24%5Cmu%20*%20%5Clambda%24%20is%20discrete%20if%20both%20%24%5Cmu%24%20and%20%24%5Clambda%24%20are%20discrete%2C%20that%20%24%5Cmu%20*%20%5Clambda%24%20is%20continuous%20if%20%24%5Cmu%24%20is%20continuous%20and%20%24%5Clambda%20%5Cin%20M%24%2C%20and%20that%20%24%5Cmu%20*%20%5Clambda%20%5Cll%20m%24%20if%20%24%5Cmu%20%5Cll%20m%24**%3B%0A%20%20%20%206.%20**Assume%20%24d%5Cmu%20%3D%20f%20dm%24%2C%20%24d%5Clambda%20%3D%20gdm%24%2C%20%24f%20%5Cin%20L%5E1(%5Cmathbb%7BR%7D)%24%2C%20and%20%24g%20%5Cin%20L%5E1(%5Cmathbb%7BR%7D)%24%2C%20and%20prove%20that%20%24d(%5Cmu%20*%20%5Clambda)%20%3D%20(f*%20g)dm%24**%3B%0A%20%20%20%207.%20**Properties%201%20and%203%20show%20that%20the%20Banach%20space%20%24M%24%20is%20what%20one%20calls%20a%20commutative%20Banach%20algebra.%20Show%20that%20part%205%20and%206%20imply%20that%20the%20set%20of%20all%20discrete%20measures%20in%20%24M%24%20is%20a%20sub-algebra%20of%20%24M%24%2C%20that%20the%20continuous%20measures%20form%20an%20ideal%20in%20%24M%24%2C%20and%20that%20the%20absolutely%20continuous%20measures%20(relative%20to%20%24m%24)%20form%20an%20ideal%20in%20%24M%24%20which%20is%20isomorphic%20(as%20an%20algebra)%20to%20%24L%5E1(%5Cmathbb%7BR%7D)%24**%3B%0A%20%20%20%208.%20**Show%20that%20%24M%24%20has%20a%20unit%2C%20i.e.%20show%20that%20there%20exists%20a%20%24%5Cdelta%20%5Cin%20M%24%20such%20that%20%24%5Cdelta%20*%20%5Cmu%20%3D%20%5Cmu%24%20for%20all%20%24%5Cmu%20%5Cin%20M%24**%3B%0A%20%20%20%209.%20**Only%20two%20properties%20of%20%24%5Cmathbb%7BR%7D%24%20have%20been%20used%20in%20this%20discussion%3A%20%24%5Cmathbb%7BR%7D%24%20is%20a%20commutative%20group%20under%20addition%2C%20and%20there%20exists%20a%20translation%20invariant%20Borel%20measure%20%24m%24%20on%20%24%5Cmathbb%7BR%7D%24%20which%20is%20not%20identically%20%240%24%20and%20which%20is%20finite%20on%20all%20compact%20subsets%20of%20%24%5Cmathbb%7BR%7D%24.%20Show%20that%20the%20same%20result%20hold%20if%20%24%5Cmathbb%7BR%7D%24%20is%20replaced%20by%20%24%5Cmathbb%7BR%7D%5Ek%24%20or%20by%20%24T%24%20(the%20unit%20circle)%20or%20by%20%24T%5Ek%24%20(the%20%24k%24-dimensional%20torus%2C%20the%20cartesian%20product%20of%20%24k%24%20copies%20of%20%24T%24)%2C%20as%20soon%20as%20the%20definitions%20are%20properly%20formulated**.%0A%0A%20%20%20%20Proof%0A%0A%20%20%20%20%23%23%23%20Part%201%0A%0A%20%20%20%20From%20the%20definition%2C%20%24%5Cmu%20*%20%5Clambda%24%20is%20defined%20on%20each%20Borel%20set%2C%20so%20we%20just%20need%20to%20check%20it%20is%20a%20measure%20(countable%20additivity).%20Suppose%20%24%5Clbrace%20E%5Ei%20%5Crbrace%24%20is%20a%20sequence%20of%20disjoint%20measurable%20set.%20Then%20%24(%5Ccup%20E%5Ei)_2%20%3D%20%5Clbrace%20(x%2Cy)%20%7C%20x%2By%20%5Cin%20%5Ccup%20E%5Ei%20%5Crbrace%24%20%24%3D%5Clbrace%20(x%2Cy)%20%7C%20x%2By%20%5Cin%20E%5Ei%5Ctext%7B%20for%20some%20%7Di%20%5Crbrace%24%20%24%3D%20%5Ccup%20E_2%5Ei%24.%20Since%20%24%5Cmu%20%5Ctimes%20%5Clambda%24%20is%20a%20measure%20(has%20countable%20additivity)%2C%20%24%24%5Cbegin%7Baligned%7D(%5Cmu%20*%20%5Clambda)(%5Ccup%20E%5Ei)%20%26%3D%20(%5Cmu%20%5Ctimes%20%5Clambda)((%5Ccup%20E%5Ei)_2)%20%5C%5C%26%3D%20(%5Cmu%20%5Ctimes%20%5Clambda)(%5Ccup%20E%5Ei_2)%5C%5C%20%26%3D%5Csum(%5Cmu%20%5Ctimes%20%5Clambda)(E_2%5Ei)%20%5C%5C%26%3D%5Csum(%5Cmu%20*%20%5Clambda)(E%5Ei)%5Cend%7Baligned%7D%24%24%20so%20does%20%24%5Cmu%20*%20%5Clambda%24.%0A%0A%20%20%20%20Let%20%24%5Clbrace%20E%5Ei%20%5Crbrace%24%20be%20an%20arbitrary%20partition%20of%20%24%5Cmathbb%7BR%7D%24%20(notice%20then%20%24E_2%5Ei%24's%20are%20then%20disjoint)%2C%20then%20%24%24%5Cbegin%7Baligned%7D%20%7C%5Cmu*%5Clambda%7C(%5Cmathbb%7BR%7D)%20%26%5Cle%20%5Csum%7C%20(%5Cmu%20*%20%5Clambda)(E%5Ei)%7C%20%5C%5C%26%3D%5Csum%20%7C%20(%5Cmu%5Ctimes%20%5Clambda)(E%5Ei_2)%7C%20%5C%5C%26%20%5Cle%20%7C%20%5Cmu%20%5Ctimes%20%5Clambda%20%7C%20(%5Cmathbb%7BR%7D%5E2)%20%26(*)%5Cend%7Baligned%7D%24%24%20Remember%20there%20exists%20%24h_1%2C%20h_2%24%20(real)%20functions%20with%20%24%7C%20h_1%20%7C%20%3D%20%7C%20h_2%20%7C%20%3D1%24%20(Radon-Nikodym)%20such%20that%20%24d%5Cmu%20%3D%20h_1d%7C%20%5Cmu%20%7C%24%20and%20%24d%5Clambda%20%3D%20h_2%20d%7C%20%5Clambda%20%7C%24.%20By%20definition%20of%20product%20of%20measure%2C%20we%20have%3A%20%24%24%5Cbegin%7Baligned%7D(%5Cmu%20%5Ctimes%20%5Clambda)(E_2)%20%26%3D%20%5Cint_%7B%5Cmathbb%7BR%7D%7D%5Clambda(E_%7B2y%7D)h_1(x)d%7C%5Cmu%7C%20%5C%5C%20%26%3D%5Cint_%7B%5Cmathbb%7BR%7D%7D(%5Cint_%7B%5Cmathbb%7BR%7D%7D%5Cchi_%7BE_2%7Dh_2d%7C%20%5Clambda%7C)%20h_1d%7C%20%5Cmu%7C%20%5C%5C%20%26%3D%20%5Cint_%7BE_2%7Dh_2h_1d(%7C%20%5Cmu%20%7C%20%5Ctimes%20%7C%20%5Clambda%20%7C)%5Cend%7Baligned%7D%24%24%20(last%20step%20is%20due%20to%20Fubini%2C%20it%20is%20applicable%20because%20we%20know%20%24%7C%20%5Cmu%20%7C%24%20and%20%24%7C%20%5Clambda%20%7C%24%20are%20finite).%20Thus%20%24%24%5Cbegin%7Baligned%7D%7C%20%5Cmu%20%5Ctimes%20%5Clambda%20%7C%20(E_2)%20%26%3D%20%5Cint_%7BE_2%7D%7C%20h_2%20h_1%7C%20d(%7C%20%5Cmu%20%7C%20%5Ctimes%20%7C%20%5Clambda%20%7C)%20%5C%5C%26%3D%5Cint_%7BE_2%7D1d(%7C%20%5Cmu%20%7C%20%5Ctimes%20%7C%20%5Clambda%20%7C)%20%5C%5C%20%26%3D%20(%7C%20%5Cmu%20%7C%20%5Ctimes%20%7C%20%5Clambda%20%7C)(E_2)%20%5Cend%7Baligned%7D%24%24%20thus%20the%20total%20variation%20of%20the%20product%20is%20the%20product%20of%20total%20variation.%20Go%20back%20to%20(%24*%24)%20we%20have%20%24%7C%20%5Cmu%20*%20%5Clambda%20%7C(%5Cmathbb%7BR%7D)%20%5Cle%20(%7C%20%5Cmu%7C%20%5Ctimes%20%7C%20%5Clambda%20%7C)%20(%5Cmathbb%7BR%7D%5E2)%24%20and%20result%20follows.%0A%0A%20%20%20%20%23%23%23%20Part%202%0A%0A%20%20%20%20First%20we%20proof%20%24%5Cmu%20*%20%5Clambda%24%20is%20such%20a%20measure.%20Let%20%24E%24%20be%20a%20(Borel)%20measurable%20set%2C%20then%20we%20have%3A%20%24%24%5Cbegin%7Baligned%7D%5Cint_%7B%5Cmathbb%7BR%7D%7D%5Cchi_E(x)d(%5Cmu%20*%20%5Clambda)%20%26%3D%20(%5Cmu%20*%20%5Clambda)(E)%20%5C%5C%20%26%3D%20(%5Cmu%20%5Ctimes%20%5Clambda)(E_2)%20%5C%5C%20%26%3D%20%5Cint_%7B%5Cmathbb%7BR%7D%7D%5Clambda(E_%7B2x%7D)d%5Cmu(x)%20%5C%5C%20%26%3D%20%5Cint_%7B%5Cmathbb%7BR%7D%7D%20%5Cint_%7B%5Cmathbb%7BR%7D%7D%5Cchi_E(x%2By)d%5Clambda(y)%20d%5Cmu(x)%20%5C%5C%20%26%3D%20%5Cint_%7B%5Cmathbb%7BR%7D%7D%20%5Cint_%7B%5Cmathbb%7BR%7D%7D%5Cchi_E(x%2By)d%5Cmu(x)%20d%5Clambda(y)%5Cend%7Baligned%7D%24%24%20so%20%24%5Cmu%20*%20%5Clambda%24%20satisfies%20the%20condition%20with%20characteristic%20functions%20on%20Borel%20set%2C%20thus%20with%20simple%20functions%2C%20thus%20with%20Borel%20measurable%20functions%2C%20and%20thus%20with%20%24f%20%5Cin%20C_0(%5Cmathbb%7BR%7D)%24.%0A%0A%20%20%20%20By%20Riesz%20Representation%20Theorem%2C%20this%20measure%20is%20unique.%0A%0A%20%20%20%20%23%23%23%20Part%203%0A%0A%20%20%20%201.%20Commutative%3A%20%24%24%5Cbegin%7Baligned%7D%5Cint_%7B%5Cmathbb%7BR%7D%7Dfd(%5Cmu%20*%20%5Clambda)%20%26%3D%20%5Cint_%7B%5Cmathbb%7BR%7D%7D%5Cint_%7B%5Cmathbb%7BR%7D%7Df(x%2By)d%5Cmu%20d%5Clambda%20%5C%5C%26%3D%20%5Cint_%7B%5Cmathbb%7BR%7D%7D%20%5Cint_%7B%5Cmathbb%7BR%7D%7Df(y%2Bx)d%5Clambda%20d%5Cmu%20%5C%5C%26%3D%20%5Cint_%7B%5Cmathbb%7BR%7D%7Dfd(%5Clambda%20*%20%5Cmu)%5Cend%7Baligned%7D%24%24%0A%20%20%20%202.%20Associativity%3A%20%24%24%5Cbegin%7Baligned%7D%5Cint_%7B%5Cmathbb%7BR%7D%7Dfd((%5Cmu*%20%5Clambda)*%20%5Cdelta)%20%26%3D%20%5Cint_%7B%5Cmathbb%7BR%7D%7D%20%5Cint_%7B%5Cmathbb%7BR%7D%7D%5Cint_%7B%5Cmathbb%7BR%7D%7Df(x%2By%2Bz)d%5Cmu(x)d%5Clambda(y)d%5Cdelta(z)%20%5C%5C%20%5Cint_%7B%5Cmathbb%7BR%7D%7Dfd(%5Cmu*%20(%5Clambda*%20%5Cdelta))%20%26%3D%20%5Cint_%7B%5Cmathbb%7BR%7D%7D(%5Cint_%7B%5Cmathbb%7BR%7D%7Df(x%2Bw)d%5Cmu(x))d(%5Clambda*%5Cdelta)(w)%20%5C%5C%26%3D%5Cint_%7B%5Cmathbb%7BR%7D%7D%5Cint_%7B%5Cmathbb%7BR%7D%7D%5Cint_%7B%5Cmathbb%7BR%7D%7Df(x%2By%2Bz)d%5Cmu(x)d%5Clambda(y)d%5Cdelta(z)%5Cend%7Baligned%7D%24%24%20so%20they%20are%20equal%3B%0A%20%20%20%203.%20Distributivity%3A%20%24%24%5Cbegin%7Baligned%7D%5Cint_%7B%5Cmathbb%7BR%7D%7Dfd(%5Cmu%20*(%5Clambda%20%2B%20%5Cdelta))%20%26%3D%20%5Cint_%7B%5Cmathbb%7BR%7D%7D%5Cint_%7B%5Cmathbb%7BR%7D%7Df(x%2By)d%5Cmu(x)d(%5Clambda(y)%2B%5Cdelta(y))%20%5C%5C%20%26%3D%20%5Cint_%7B%5Cmathbb%7BR%7D%7D%5Cint_%7B%5Cmathbb%7BR%7D%7Df(x%2By)d%5Cmu%20d%5Clambda%20%2B%20%5Cint_%7B%5Cmathbb%7BR%7D%7D%5Cint_%7B%5Cmathbb%7BR%7D%7Df(x%2By)d%5Cmu%20d%5Cdelta%5C%5C%26%3D%5Cint_%7B%5Cmathbb%7BR%7D%7Dfd(%5Cmu*%20%5Clambda)%2B%5Cint_%7B%5Cmathbb%7BR%7D%7Dfd(%5Cmu*%5Cdelta)%20%5C%5C%26%3D%5Cint_%7B%5Cmathbb%7BR%7D%7Dfd(%5Cmu*%20%5Clambda%20%2B%20%5Cmu%20*%20%5Cdelta)%5Cend%7Baligned%7D%24%24%0A%0A%20%20%20%20%23%23%23%20Part%204%0A%0A%20%20%20%20By%20definition%20we%20have%3A%20%24%24(%5Cmu%20*%20%5Clambda)(E)%20%3D%20(%5Cmu%5Ctimes%20%5Clambda)(E_2)%20%3D%20%5Cint_%7B%5Cmathbb%7BR%7D%7D%5Cmu(E_2%5Ey)d%5Clambda(y)%20%3D%20%5Cint_%7B%5Cmathbb%7BR%7D%7D%5Cint_%7BE_2%5Ey%7Dd%5Cmu(x)d%5Clambda(y).%24%24%20Notice%20now%20%24%24E_2%5Ey%3A%3D%20%5Clbrace%20x%20%7C%20(x%2Cy)%20%5Cin%20E_2%5Crbrace%20%3D%20%5Clbrace%20x%20%7C%20x%2By%20%5Cin%20E%20%5Crbrace%20%3D%20%5Clbrace%20x-y%20%7C%20x%20%5Cin%20E%20%5Crbrace%20%3D%20E-y.%24%24%20So%20%24%24(%5Cmu%20*%20%5Clambda)(E)%20%3D%20%5Cint_%7B%5Cmathbb%7BR%7D%7D%5Cint_%7BE_2%5Ey%7Dd%5Cmu(x)d%5Clambda(y)%20%3D%20%5Cint_%7B%5Cmathbb%7BR%7D%7D%5Cint_%7BE-y%7Dd%5Cmu%20d%5Clambda%20%3D%20%5Cint_%7B%5Cmathbb%7BR%7D%7D%5Cmu(E-y)d%5Clambda(y)%24%24%20as%20desired.%0A%0A%20%20%20%20%23%23%23%20Part%205%0A%0A%20%20%20%201.%20Suppose%20%24%5Cmu%2C%20%5Clambda%24%20concentrate%20on%20countable%20sets%20%24A%2C%20B%24%2C%20then%20%24A%20%5Ccup%20B%20%3D%20C%24%20is%20also%20countable%2C%20I%20claim%20that%20%24%5Cmu%20*%20%5Clambda%24%20concentrate%20on%20%24C%24%3A%20Let%20%24E%24%20be%20any%20Borel%20set%20disjoint%20from%20%24C%24%2C%20then%20%24%24%5Cmu%20*%20%5Clambda(E)%20%3D%20%5Cint_%7B%5Cmathbb%7BR%7D%7D%5Cmu(E-y)d%5Clambda(y).%24%24%20Now%20if%20%24y%20%5Cin%20C%24%20then%20%24E-y%24%20is%20disjoint%20from%20%24C%24%20thus%20disjoint%20from%20%24A%24%2C%20thus%20%24%5Cmu(E%20-%20y)%20%3D%200%24%2C%20and%20if%20%24y%20%5Cnotin%20C%24%20then%20%24y%20%5Cnotin%20B%24%20thus%20%24%5Clambda(y)%20%3D%200%24%2C%20so%20the%20integral%20equals%20%240%24%2C%20result%20follows%3B%0A%20%20%20%202.%20Suppose%20%24%5Cmu%24%20is%20continuous%20and%20%24%5Clambda%20%5Cin%20M%24%2C%20then%20for%20any%20%24x%20%5Cin%20%5Cmathbb%7BR%7D%24%20we%20have%20%24%24%5Cbegin%7Baligned%7D(%5Cmu%20*%20%5Clambda)(%5Clbrace%20x%20%5Crbrace)%20%26%3D%20%5Cint_%7B%5Cmathbb%7BR%7D%7D%5Cmu(%5Clbrace%20x%20%5Crbrace%20-%20t)d%5Clambda(t)%20%5C%5C%20%26%3D%20%5Cint_%7B%5Cmathbb%7BR%7D%7D%5Cmu(%5Clbrace%20x%20-%20t%20%5Crbrace)d%5Clambda(t)%20%5C%5C%20%26%3D%20%5Cint_%7B%5Cmathbb%7BR%7D%7D%200%20d%5Clambda(t)%20%3D%200%20%26%5Ctext%7B(because%20%7D%5Cmu%20%5Ctext%7B%20is%20continuous)%7D%5Cend%7Baligned%7D%24%24%20thus%20%24%5Cmu%20*%20%5Clambda%24%20is%20continuous%3B%0A%20%20%20%203.%20Suppose%20%24%5Cmu%5Cll%20m%24%2C%20let%20%24E%24%20be%20an%20arbitrary%20Borel%20set%20that%20has%20%24m(E)%20%3D%200%24.%20Then%20%24m(E-t)%20%3D%200%24%20for%20all%20%24t%24%20thus%20%24%5Cmu(E-t)%20%3D%200%24%20for%20all%20%24t%24.%20Thus%20%24%24(%5Cmu%20*%20%5Clambda)(E)%20%3D%20%5Cint_%7B%5Cmathbb%7BR%7D%7D%5Cmu(E-t)d%5Clambda(t)%20%3D%20%5Cint_%7B%5Cmathbb%7BR%7D%7D0d%5Clambda%20%3D%200%24%24%20thus%20%24%5Cmu*%5Clambda%20%5Cll%20m%24.%0A%0A%20%20%20%20%23%23%23%20Part%206%0A%0A%20%20%20%20We%20have%20%24%24%5Cint_%7B%5Cmathbb%7BR%7D%7Dd(%5Cmu*%20%5Clambda)%20%3D%20%5Cint_%7B%5Cmathbb%7BR%7D%7D%5Cint_%7B%5Cmathbb%7BR%7D%7Dd%5Cmu%20d%5Clambda%20%3D%20%5Cint_%7B%5Cmathbb%7BR%7D%7D%5Cint_%7B%5Cmathbb%7BR%7D%7Df(x)dm(x)%20g(y)dm(y)%24%24%20by%20translate%20invariance%2C%20we%20get%20%24%24%5Cbegin%7Baligned%7D%E2%80%A6%20%26%3D%20%5Cint_%7B%5Cmathbb%7BR%7D%7D%5Cint_%7B%5Cmathbb%7BR%7D%7Df(x-y)g(y)dm(x)dm(y)%20%5C%5C%20%26%3D%20%5Cint_%7B%5Cmathbb%7BR%7D%7D(f*%20g)dm(y)%5Cend%7Baligned%7D%24%24%20which%20exactly%20means%20%24d(%5Cmu%20*%20%5Clambda)%20%3D%20(f*%20g)dm%24.%0A%0A%20%20%20%20%23%23%23%20Part%208%0A%0A%20%20%20%20Consider%20the%20measure%20%24%5Cdelta%24%20defined%20as%20%24%5Cdelta(E)%20%3D%20%5Cbegin%7Bcases%7D1%2C0%5Cin%20E%5C%5C0%2C0%5Cnotin%20E%5Cend%7Bcases%7D%24.%20Apparently%20this%20is%20a%20complex%20Borel%20measure.%20Let%20%24E%24%20be%20an%20arbitrary%20Borel%20set%2C%20and%20%24%5Cmu%20%5Cin%20M%24%20be%20an%20arbitrary%20Borel%20measure%2C%20then%20%24%24(%5Cdelta*%5Cmu)(E)%20%3D%20%5Cint_%7B%5Cmathbb%7BR%7D%7D%5Cdelta(E-t)d%5Cmu(t)%20%3D%20%5Cint_E%201%20d%5Cmu(t)%20%3D%20%5Cmu(E)%24%24%20because%20%24%5Cdelta(E-t)%20%3D%201%24%20if%20and%20only%20if%20%24t%20%5Cin%20E%24%20so%20that%20%240%20%5Cin%20E-t%24%2C%20which%20proves%20%24%5Cdelta%20*%20%5Cmu%20%3D%20%5Cmu%24%20for%20any%20%24%5Cmu%24%20as%20desired.%0A%0A%20%20%20%20%23%23%20Exercise%2011%0A%0A%20%20%20%20**Let%20%24%5Cmathscr%7BB%7D_k%24%20be%20the%20%24%5Csigma%24-algebra%20of%20all%20Borel%20sets%20in%20%24%5Cmathbb%7BR%7D_k%24.%20Prove%20that%20%24%5Cmathscr%7BB%7D_%7Bm%2Bn%7D%20%3D%20%5Cmathscr%7BB%7D_m%20%5Ctimes%20%5Cmathscr%7BB%7D_n%24.%20This%20is%20relevant%20in%20Theorem%208.14**.%0A%0A%20%20%20%20Proof%0A%0A%20%20%20%20We%20know%20%24%5Cmathbb%7BR%7D_k%24%20is%20second-countable%2C%20i.e.%20the%20open%20sets%20are%20generated%20by%20a%20countable%20collection%20%24%5Cmathcal%7BB%7D_k%24.%20Then%20%24%5Cmathbb%7BR_%7Bm%2Bn%7D%7D%24%20has%20a%20basis%20%24%5Cmathcal%7BB%7D_%7Bm%2Bn%7D%20%3D%20%5Clbrace%20B_m%20%5Ctimes%20B_n%20%7C%20B_m%20%5Cin%20%5Cmathcal%7BB%7D_m%2C%20B_n%20%5Cin%20%5Cmathcal%7BB%7D_n%20%5Crbrace%24%20which%20is%20countable.%0A%0A%20%20%20%20Suppose%20we%20have%20an%20open%20set%20in%20%24%5Cmathbb%7BR_%7Bm%2Bn%7D%7D%24%2C%20then%20it%20is%20a%20countable%20union%20of%20elements%20of%20%24%5Cmathcal%7BB%7D_%7Bm%2Bn%7D%24%2C%20thus%20it%20is%20an%20open%20set%20in%20%24%5Cmathbb%7BR%7D_m%20%5Ctimes%20%5Cmathbb%7BR%7D_n%24%20thus%20it%20is%20in%20%24%5Cmathscr%7BB%7D_m%20%5Ctimes%20%5Cmathscr%7BB%7D_n%24.%20Since%20%24%5Cmathscr%7BB%7D_m%20%5Ctimes%20%5Cmathscr%7BB%7D_n%24%20is%20a%20%24%5Csigma%24-algebra%2C%20we%20have%20%24%5Cmathscr%7BB%7D_%7Bm%2Bn%7D%20%5Csubseteq%20%5Cmathscr%7BB%7D_m%20%5Ctimes%20%5Cmathscr%7BB%7D_n%24.%0A%0A%20%20%20%20Suppose%20now%20%24U%24%20is%20an%20open%20set%20in%20%24%5Cmathbb%7BR%7D_m%24%20and%20%24V%24%20is%20an%20open%20set%20in%20%24%5Cmathbb%7BR%7D_n%24%2C%20then%20%24U%20%5Ctimes%20%5Cmathbb%7BR%7D_n%24%20and%20%24%5Cmathbb%7BR%7D_m%20%5Ctimes%20V%24%20are%20open%20sets%20in%20%24%5Cmathbb%7BR%7D_%7Bm%2Bn%7D%24%2C%20thus%20%24U%20%5Ctimes%20V%20%3D%20(U%20%5Ctimes%20%5Cmathbb%7BR%7D_n)%20%5Ccap%20(%5Cmathbb%7BR%7D_m%20%5Ctimes%20V)%24%20is%20open%20in%20%24%5Cmathbb%7BR%7D_%7Bm%2Bn%7D%24%20thus%20in%20%24%5Cmathscr%7BB%7D_%7Bm%2Bn%7D%24.%20Since%20%24%5Cmathscr%7BB%7D_%7Bm%2Bn%7D%24%20is%20a%20%24%5Csigma%24-algebra%2C%20we%20have%20the%20other%20inclusion.%0A%0A%20%20%20%20%23%23%20Exercise%2012%0A%0A%20%20%20%20**Use%20Fubini's%20Theorem%20and%20the%20relation**%20%24%24%5Cfrac%7B1%7D%7Bx%7D%20%3D%20%5Cint_0%5E%7B%5Cinfty%7De%5E%7B-xt%7Ddt%2C%20x%20%3E%200%24%24%20**to%20prove%20that**%20%24%24%5Clim_%7BA%20%5Cto%20%5Cinfty%7D%5Cint_0%5EA%5Cfrac%7B%5Csin(x)%7D%7Bx%7Ddx%20%3D%20%5Cfrac%7B%5Cpi%7D%7B2%7D.%24%24%0A%0A%20%20%20%20Proof%0A%0A%20%20%20%20Let%20%24f%20%3D%20%5Csin(x)%5Ccdot%20e%5E%7B-xt%7D%24%2C%20it%20is%20continuous%20thus%20Borel.%20By%20the%20given%20relation%20we%20have%20%24%24%5Cbegin%7Baligned%7D%5Cint_0%5EA%5Cfrac%7B%5Csin(x)%7D%7Bx%7Ddx%20%26%3D%20%5Cint_0%5EA%5Csin(x)(%5Cint_0%5E%7B%5Cinfty%7De%5E%7B-xt%7Ddt)dx%20%5C%5C%20%26%3D%20%5Cint_0%5EA%5Cint_0%5E%7B%5Cinfty%7D%5Csin(x)~e%5E%7B-xt%7Ddtdx%5C%5C%20%26%3D%20%5Cint_0%5E%7B%5Cinfty%7D%5Cint_0%5E%7BA%7D%5Csin(x)~e%5E%7B-xt%7Ddxdt%20%26%5Ctext%7B(Fubini)%7D%5C%5C%20%26%3D%20%5Cint_0%5E%7B%5Cinfty%7D%20%5Cfrac%7B1%7D%7Bt%5E2%2B1%7D%20-%20%5Cfrac%7Be%5E%7B-At%7D(t%5Csin(A)%2B%5Ccos(A))%7D%7Bt%5E2%2B1%7Ddt%20%5C%5C%20%26%3D%20%5Cfrac%7B%5Cpi%7D%7B2%7D%20-%20%5Cint_0%5E%7B%5Cinfty%7D%20%5Cfrac%7Be%5E%7B-At%7D(t%5Csin(A)%2B%5Ccos(A))%7D%7Bt%5E2%2B1%7Ddt%5Cend%7Baligned%7D%24%24%20by%20evaluating%20the%20usual%20Riemann%20integral.%20Now%20for%20any%20%24A%24%2C%20%24%24%5Cbegin%7Baligned%7D%7C%20%5Cfrac%7Be%5E%7B-At%7D(t%5Csin(A)%2B%5Ccos(A))%7D%7Bt%5E2%2B1%7D%20%7C%20%26%5Cle%20%7C%5Cfrac%7Be%5E%7B-At%7D(1%2Bt)%7D%7Bt%5E2%2B1%7D%20%7C%20%5C%5C%20%26%5Cle%20%7C%5Cfrac%7Be%5E%7B-At%7D%7D%7Bt%5E2%2B1%7D%20%7C%20%2B%20%7C%5Cfrac%7Be%5E%7B-At%7Dt%7D%7Bt%5E2%2B1%7D%20%7C%20%5C%5C%20%26%5Cle%202e%5E%7B-At%7D%5Cend%7Baligned%7D%24%24%20%24%24%5Clim_%7BA%20%5Cto%20%5Cinfty%7D%5Cint_0%5E%7B%5Cinfty%7D%5Cfrac%7Be%5E%7BAt%7D(t%5Csin(A)%2B%5Ccos(A))%7D%7Bt%5E2%2B1%7Ddt%20%3D%20%5Cint_0%5E%7B%5Cinfty%7D0dt%20%3D%200.%24%24%0A%20%20%20%20%22%22%22)%0A%20%20%20%20return%0A%0A%0A%40app.cell%0Adef%20_(mo)%3A%0A%20%20%20%20mo.md(r%22%22%22%0A%20%20%20%20%23%20Spring%20Semester%20Homework%205%0A%20%20%20%20%22%22%22)%0A%20%20%20%20return%0A%0A%0A%40app.cell%0Adef%20_(mo)%3A%0A%20%20%20%20mo.md(r%22%22%22%0A%20%20%20%20%23%23%20Exercise%201%0A%0A%20%20%20%20**Suppose%20%24f%20%5Cin%20L%5E1%24%2C%20%24f%3E0%24.%20Prove%20that%20%24%7C%20%5Chat%7Bf%7D(y)%20%7C%20%3C%20%5Chat%7Bf%7D(0)%24%20for%20all%20%24y%20%5Cne%200%24**.%0A%0A%20%20%20%20Proof%0A%0A%20%20%20%20Use%20the%20fact%20that%20%24%7C%20e%5E%7Bix%7D%20%7C%20%3D%201%24%20for%20any%20%24x%24%2C%20we%20get%20%24%7C%20f(x)%20e%5E%7B-ixy%7D%20%7C%20%3D%20%7C%20f(x)%20%7C%20%3D%20f(x)%24%20(because%20%24f%24%20is%20positive)%2C%20and%3A%20%24%24%7C%20%5Chat%7Bf%7D(y)%20%7C%20%3D%20%7C%20%5Cint_%7B-%5Cinfty%7D%5E%7B%5Cinfty%7Df(x)e%5E%7B-ixy%7Ddx%20%7C%20%5Cle%20%5Cint_%7B-%5Cinfty%7D%5E%7B%5Cinfty%7D%7C%20f(x)e%5E%7B-ixy%7D%20%7C%20dx%20%3D%20%5Cint_%7B-%5Cinfty%7D%5E%7B%5Cinfty%7Df(x)%20dx%20%3D%20%5Chat%7Bf%7D(0).%24%24%20If%20there%20is%20some%20%24y%20%5Cne%200%24%20with%24%7C%20%5Chat%7Bf%7D(y)%20%7C%20%3D%20%5Chat%7Bf%7D(0)%24%2C%20then%20the%20equality%20in%20above%20inequality%20holds.%20From%20chapter%201%20we%20know%20this%20means%20there%20is%20a%20constant%20%24a%24%20such%20that%20%24af(x)e%5E%7B-ixy%7D%20%3D%20%7C%20f(x)e%5E%7B-ixy%7D%7C%20%3D%20f(x)%24%20a.e..%20Since%20%24f(x)%20%5Cne%200%24%20we%20can%20cancel%20it%20and%20get%20%24a%20e%5E%7B-ixy%7D%20%3D%201%24%20a.e.%2C%20but%20that%20is%20only%20possible%20when%20%24e%5E%7B-ixy%7D%24%20is%20a%20constant%20a.e.%20(w.r.t.%20%24x%24)%2C%20which%20only%20happens%20if%20%24y%20%3D%200%24.%20So%20if%20%24y%20%5Cne%200%24%20we%20cannot%20have%20equality%2C%20and%20we%20have%20desired%20statement.%0A%0A%20%20%20%20%23%23%20Exercise%202%0A%0A%20%20%20%20**Compute%20the%20Fourier%20transform%20of%20the%20characteristic%20function%20of%20an%20interval.%20For%20%24n%20%3D%201%2C2%2C3%2C%5Cdots%24%2C%20let%20%24g_n%24%20be%20the%20characteristic%20function%20of%20%24%5B-n%2Cn%5D%24%2C%20let%20%24h%24%20be%20the%20characteristic%20function%20of%20%24%5B-1%2C1%5D%24%2C%20and%20compute%20%24g_n%20*%20h%24%20explicitly%20(the%20graph%20is%20piece-wise%20linear).%20Show%20that%20%24g_n%20*%20h%24%20is%20the%20Fourier%20transform%20of%20a%20function%20%24f_n%20%5Cin%20L%5E1%24%3B%20except%20for%20a%20multiplicative%20constant%2C**%20%24%24f_n(x)%20%3D%20%5Cfrac%7B%5Csin(x)%5Csin(nx)%7D%7Bx%5E2%7D.%24%24%20**Show%20that%20%24%5C%7C%20f_n%20%5C%7C_1%20%5Cto%20%5Cinfty%24%20and%20conclude%20that%20the%20mapping%20%24f%20%5Cto%20%5Chat%7Bf%7D%24%20maps%20%24L%5E1%24%20into%20a%20proper%20subset%20of%20%24C_0%24**.%0A%0A%20%20%20%20Solution%0A%0A%20%20%20%20Consider%20an%20interval%20%24%5Ba%2Cb%5D%24%20and%20the%20characteristic%20function%20%24%5Cchi%20%3D%20%5Cchi_%7B%5Ba%2Cb%5D%7D%24%2C%20then%20follow%20definition%2C%20%24%24%5Cbegin%7Baligned%7D%5Chat%7B%5Cchi%7D(t)%20%26%3D%20%5Cint_%7B-%5Cinfty%7D%5E%7B%5Cinfty%7D%20%5Cchi(x)%20e%5E%7B-ixt%7D%20dx%20%5C%5C%20%26%3D%20%5Cint%5E%7Bb%7D_%7Ba%7De%5E%7B-ixt%7Ddx%20%5C%5C%20%26%3D%20%5Cbegin%7Bcases%7D%20-%5Cfrac%7B1%7D%7Bit%7D(e%5E%7B-i%20b%20t%7D%20-%20e%5E%7B-iat%7D)%2C%20%26%20t%20%5Cne%200%2C%20%5C%5C%20b%20-%20a%2C%20%26%20t%20%3D%200.%20%5Cend%7Bcases%7D%5Cend%7Baligned%7D%24%24%20For%20obvious%20reason%2C%20characteristic%20function%20of%20half%20open%20interval%20or%20open%20interval%20has%20the%20same%20transformation.%20In%20particular%20if%20%24a%20%3D%20-b%24%2C%20or%20say%20the%20interval%20is%20of%20the%20form%20%24%5B-n%2Cn%5D%24%2C%20then%20by%20property%20of%20complex%20number%2C%20we%20have%20%24%24%5Cwidehat%7B%5Cchi_%7B%5B-n%2Cn%5D%7D%7D(t)%20%3D%20%5Cbegin%7Bcases%7D%20%5Cfrac%7B2%20%5Csin(nt)%7D%7Bt%7D%2C%20%26t%20%3D%200%2C%5C%5C2n%2C%26t%20%5Cne%200.%5Cend%7Bcases%7D%24%24%20Or%20simply%20%24%24%5Cwidehat%7B%5Cchi_%7B%5B-n%2Cn%5D%7D%7D(t)%20%3D%20%5Cfrac%7B2%20%5Csin(nt)%7D%7Bt%7D%24%24%20because%20the%20limit%20agrees.%0A%0A%20%20%20%20Let%20%24g_n%24%20and%20%24h%24%20be%20as%20given%2C%20we%20have%3A%20%24%24%5Cbegin%7Baligned%7D(g_n%20*%20h)(x)%20%26%3D%20%5Cint_%7B-%5Cinfty%7D%5E%7B%5Cinfty%7D%5Cchi_%7B%5B-n%2Cn%5D%7D(x-y)%5Cchi_%7B%5B-1%2C1%5D%7D(y)dy%20%5C%5C%20%26%3D%5Cint_%7B-1%7D%5E1%5Cchi_%7B%5B-n%2Cn%5D%7D(x-y)dy%20%5C%5C%26%3Dm(%5B-1%2Bx%2C%201%2Bx%5D%20%5Ccap%20%5B-n%2Cn%5D)%5Cend%7Baligned%7D%24%24%20Notice%20it%20is%20bounded%20above%20by%20%242%24.%0A%0A%20%20%20%20%24f_n%20%5Cin%20L%5E1%24%20because%20on%20%24%5B-1%2C1%5D%24%20it%20is%20bounded%20by%20%24n%24%20and%20otherwise%20its%20absolute%20value%20is%20bounded%20by%20%24%5Cfrac%7B1%7D%7Bx%5E2%7D%24%20which%20is%20a%20well-known%20%24L%5E1%24%20function.%20Also%20%24g_n%24%20and%20%24h%24%20are%20both%20%24L%5E1%24%2C%20and%20we%20have%20%24%5Cwidehat%7Bg%20*%20h%7D%20%3D%20%5Chat%7Bg_n%7D%5Ccdot%20%5Chat%7Bh%7D%24.%20Use%20the%20formula%20we%20had%20above%2C%20we%20get%3A%20%24%24(%5Cwidehat%7Bg_n%20*%20h%7D)(t)%20%3D%20%5Cfrac%7B4%5Csin(nt)%5Csin(t)%7D%7Bt%5E2%7D%24%24%20which%20is%20%244f_n(t)%24.%20Notice%20%24f_n%24%20is%20even%20function%2C%20so%20%24f_n(t)%20%3D%20f_n(-t)%24%2C%20we%20have%3A%20%24%24%5Cbegin%7Baligned%7D%5Chat%7Bf_n%7D(t)%3D%20%5Chat%7Bf_n%7D(-t)%20%26%3D%20%5Cint_%7B-%5Cinfty%7D%5E%7B%5Cinfty%7Df_n(-x)e%5E%7Bixt%7Ddx%20%5C%5C%26%3D%20%5Cint%20f_n(x)e%5E%7Bixt%7D%20dx%5C%5C%26%3D%5Cfrac%7B1%7D%7B4%7D%5Cint%20%5Cwidehat%7Bg_n%20*%20h%7D(x)e%5E%7Bixt%7Ddx%20%20%5Cend%7Baligned%7D%24%24%20The%20condition%20of%20the%20Inversion%20Theorem%20is%20satisfied%2C%20and%20we%20have%20%244%5Chat%7Bf_n%7D(t)%20%3D%20(g_n*%20h)(t)%24%20a.e.%2C%20i.e.%20%24g_n*%20h%24%20is%20the%20Fourier%20transformation%20of%20%24f_n%24%20up%20to%20a%20constant.%0A%0A%20%20%20%20We%20can%20write%20%24%24%5C%7C%20f_n%20%5C%7C_1%20%3D%20%5Cint_%7B-%5Cinfty%7D%5E%7B%5Cinfty%7D%7C%5Cfrac%7B%5Csin(x)%7D%7Bx%7D%7C%20%5Ccdot%7C%5Cfrac%7B%5Csin(nx)%7D%7Bx%7D%7C%20dx.%24%24%20Pick%20arbitrary%20%24c%20%5Cin%20(0%2C1)%24%2C%20it%20is%20easy%20to%20see%20we%20can%20find%20an%20interval%20%24%5B0%2Ca%5D%24%20for%20certain%20%24a%24%20such%20that%20%24%5Cfrac%7B%5Csin(x)%7D%7Bx%7D%20%3E%20c%24%20whenever%20%24x%20%5Cin%20%5B0%2Ca%5D%24.%20It%20follows%20that%20if%20we%20can%20show%20%24%24%5Clim_%7Bn%20%5Cto%20%5Cinfty%7D%5Cint_0%5Ea%20%7C%20%5Cfrac%7B%5Csin(nx)%7D%7Bx%7D%20%7C%20dx%20%3D%20%5Cinfty%24%24%20then%20%24%5C%7C%20f_n%20%5C%7C_1%24%20also%20goes%20to%20infinity%20because%20it%20is%20greater%20than%20the%20above%20expression%20up%20to%20a%20multiplication%20of%20constant.%20But%20%24%24%5Clim_%7Bn%20%5Cto%20%5Cinfty%7D%5Cint_0%5Ea%20%7C%20%5Cfrac%7B%5Csin(nx)%7D%7Bx%7D%20%7C%20dx%20%3D%5E%7Bu%20%3D%20nx%7D%20%5Clim_%7Bn%7D%5Cint_0%5E%7Ban%7D%7C%5Cfrac%7B%5Csin(u)%7D%7Bu%7D%20%7C%20du%20%3D%20%5Cint_%7B0%7D%5E%7B%5Cinfty%7D%7C%20%5Cfrac%7B%5Csin(u)%7D%7Bu%7D%20%7C%20du%24%24%20which%20is%20well-known%20to%20be%20infinite%2C%20so%20we%20are%20finished.%0A%0A%20%20%20%20The%20final%20assertion%20was%20proved%20in%20the%20book%20(Theorem%205.15)%20for%20%24T%24%2C%20follow%20that%20proof%2C%20the%20connection%20here%20is%20that%20if%20this%20function%20is%20(injective%20and)%20surjective%20then%20there%20is%20a%20%24%5Cdelta%20%3E%200%24%20such%20that%20%24%5C%7C%20%5Chat%7Bf%7D%20%5C%7C_%7B%5Cinfty%7D%20%5Cge%20%5Cdelta%20%5C%7C%20f%20%5C%7C_1%24%20for%20every%20%24f%20%5Cin%20L%5E1%24%2C%20but%20with%20the%20above%20example%20we%20have%20%24%24%5C%7C%20%5Chat%7Bf_n%7D%20%5C%7C_%7B%5Cinfty%7D%20%3D%20%5C%7C%20g_n%20*%20h%20%5C%7C_%7B%5Cinfty%7D%20%3D%202%24%24%20yet%20%24%5C%7C%20f%20%5C%7C_1%20%5Cto%20%5Cinfty%24%2C%20so%20such%20%24%5Cdelta%24%20cannot%20exist.%0A%0A%20%20%20%20%23%23%20Exercise%203%0A%0A%20%20%20%20**Find**%20%24%24%5Clim_%7BA%5Cto%20%5Cinfty%7D%5Cint_%7B-A%7D%5EA%5Cfrac%7B%5Csin(%5Clambda%20t)%7D%7Bt%7De%5E%7Bitx%7Ddt%2C%20(-%5Cinfty%20%3C%20x%20%3C%20%5Cinfty)%24%24%20**where%20%24%5Clambda%24%20is%20a%20positive%20constant**.%0A%0A%20%20%20%20Solution%0A%0A%20%20%20%20Remember%20we%20have%20that%20if%20%24g%20%3D%20%5Cchi_%7B%5B-%5Clambda%2C%20%5Clambda%5D%7D%24%20then%20%24%5Chat%7Bg%7D%20%3D%202%5Cfrac%7B%5Csin(%5Clambda%20t)%7D%7Bt%7D%24.%20However%20%24%5Chat%7Bg%7D%24%20is%20%24L%5E2%24%20yet%20not%20%24L%5E1%24%20so%20the%20Inversion%20Theorem%20is%20not%20applicable.%20We%20can%20only%20use%20Inversion%20Theorem%20for%20%24L%5E2%24%20function%20and%20get%20that%20%24%24%5Cpsi_A(x)%20%3D%202%5Cint_%7B-A%7D%5EA%5Cfrac%7B%5Csin(%5Clambda%20t)%7D%7Bt%7De%5E%7Bixt%7Ddt%24%24%20with%20%24%5C%7C%20%5Cpsi_A%20-%20g%20%5C%7C_2%20%5Cto%200%24.%20Write%20explicitly%3A%20%24%24%5C%7C%20%5Cint_%7B-A%7D%5EA%5Cfrac%7B%5Csin(%5Clambda%20t)%7D%7Bt%7De%5E%7Bixt%7Ddt%20-%20%5Cfrac%7B1%7D%7B2%7D%5Cchi_%7B%5B-%5Clambda%2C%20%5Clambda%5D%7D%20%5C%7C_2%20%5Cto%200%2C%24%24%20if%20we%20only%20consider%20%24A%20%5Cin%20%5Cmathbb%7BZ%7D%24%20then%20this%20means%20that%20%24%5Clbrace%5Cint_%7B-A%7D%5EA%5Cfrac%7B%5Csin(%5Clambda%20t)%7D%7Bt%7De%5E%7Bixt%7Ddt%5Crbrace%24%20has%20subsequence%20converges%20point-wise%20to%20%24%5Cfrac%7B1%7D%7B2%7D%5Cchi_%7B%5B-%5Clambda%2C%20%5Clambda%5D%7D%24%20a.e..%0A%0A%20%20%20%20%23%23%20Exercise%204%0A%0A%20%20%20%20**Give%20examples%20of%20%24f%20%5Cin%20L%5E2%24%20such%20that%20%24f%20%5Cnotin%20L%5E1%24%20but%20%24%5Chat%7Bf%7D%20%5Cin%20L%5E1%24.%20Under%20what%20circumstances%20can%20this%20happen**%3F%0A%0A%20%20%20%20Solution%0A%0A%20%20%20%20First%2C%20the%20underlying%20space%20must%20not%20be%20compact%2C%20otherwise%20%24L%5E2%20%5Csubset%20L%5E1%24.%20One%20possible%20circumstance%20is%20that%20suppose%20we%20have%20some%20%24g%20%5Cin%20L%5E2%20%5Ccap%20L%5E1%24%20and%20%24%5Chat%7Bg%7D%24%20not%20in%20%24L%5E1%24%2C%20then%20%24%5Chat%7Bg%7D%20%5Cin%20L%5E2%24%20(Theorem%209.13).%20Then%20if%20we%20define%20%24f(x)%24%20to%20be%20%24%5Chat%7Bg%7D(-x)%24%20then%20%24f%24%20satisfies%20the%20condition.%20Proof%3A%0A%0A%20%20%20%201.%20%24f%20%5Cin%20L%5E2%24%20and%20%24f%20%5Cnotin%20L%5E1%24%20are%20immediate%20from%20construction%3B%0A%20%20%20%202.%20Notice%20we%20have%20%24%5Chat%7B%5Chat%7Bg%7D%7D(x)%20%3D%20g(-x)%24%20a.e..%20Since%20%24f(x)%20%3D%20%5Chat%7Bg%7D(-x)%24%20we%20get%20%24%5Chat%7Bf%7D(x)%20%3D%20%5Chat%7B%5Chat%7Bg%7D%7D(-x)%20%3D%20g(x)%24%2C%20so%20%24%5Chat%7Bf%7D%24%20is%20%24L%5E1%24.%0A%0A%20%20%20%20Now%20consider%20%24g%20%3D%20%5Cchi_%7B%5B-1%2C1%5D%7D%24%2C%20it%20is%20in%20%24L%5E2%20%5Ccap%20L%5E1%24.%20By%20above%20exercise%20we%20know%20%24%5Chat%7Bg%7D(x)%20%3D%20%5Cfrac%7B2%5Csin(x)%7D%7Bx%7D%24%2C%20which%20is%20not%20an%20%24L%5E1%24%20function%2C%20so%20if%20we%20define%20%24f(x)%20%3D%20%5Chat%7Bg%7D(-x)%20%3D%20%5Cfrac%7B2%5Csin(x)%7D%7Bx%7D%24%20(sine%20is%20odd%20function)%20then%20it%20is%20an%20example%20we%20want.%0A%0A%20%20%20%20%23%23%20Exercise%208%0A%0A%20%20%20%20**If%20%24p%24%20and%20%24q%24%20are%20conjugate%20exponents%2C%20%24f%20%5Cin%20L%5Ep%24%2C%20%24g%20%5Cin%20L%5Eq%24%2C%20and%20%24h%20%3D%20f%20*%20g%24%2C%20prove%20that%20%24h%24%20is%20uniformly%20continuous.%20If%20also%20%241%20%3C%20p%20%3C%20%5Cinfty%24%2C%20then%20%24h%20%5Cin%20C_0%24%3B%20show%20that%20this%20fails%20for%20some%20%24f%20%5Cin%20L%5E1%2C%20g%20%5Cin%20L%5E%7B%5Cinfty%7D%24**.%0A%0A%20%20%20%20Proof%0A%0A%20%20%20%20Let%20%24p%2Cq%24%20be%20conjugate%20exponents%2C%20fix%20arbitrary%20%24x%20%5Cin%20%5Cmathbb%7BR%7D%24%20and%20let%20%24a%20%5Cin%20%5Cmathbb%7BR%7D%24%2C%20then%20%24%24%5Cbegin%7Baligned%7D%20%7C%20h(x-a)%20-%20h(x)%20%7C%20%26%20%3D%20%7C%20%5Cint_%7B-%5Cinfty%7D%5E%7B%5Cinfty%7D%20f(x-a-y)g(y)%20dy%20-%20%5Cint_%7B-%5Cinfty%7D%5E%7B%5Cinfty%7D%20f(x-y)g(y)dy%7C%20%5C%5C%26%3D%20%7C%20%5Cint_%7B-%5Cinfty%7D%5E%7B%5Cinfty%7D%20f(x-y)g(y-a)%20dy%20-%20%5Cint_%7B-%5Cinfty%7D%5E%7B%5Cinfty%7D%20f(x-y)g(y)dy%7C%20%5C%5C%20%26%3D%20%7C%20%5Cint_%7B-%5Cinfty%7D%5E%7B%5Cinfty%7Df(x-y)(g(y-a)%20-%20g(y))dy%20%7C%20%5C%5C%20%26%5Cle%20%5Cint_%7B-%5Cinfty%7D%5E%7B%5Cinfty%7D%7C%20f(x-y)%20%7C%20%5Ccdot%7C%20(g(y-a)%20-%20g(y))%7C%20dy%20%5C%5C%26%5Cle%20(%5Cint_%7B-%5Cinfty%7D%5E%7B%5Cinfty%7D%7C%20f(x-y)%20%7C%5Epdy)%5E%7B1%2Fp%7D(%5Cint_%7B-%5Cinfty%7D%5E%7B%5Cinfty%7D%7C%20g(y-a)%20-%20g(y)%20%7C%5Eqdy)%5E%7B1%2Fq%7D%5Cend%7Baligned%7D%24%24%20The%20last%20step%20is%20due%20to%20H%C3%B6lder.%20Since%20%24g%24%20is%20%24L%5Eq%24%2C%20%24g'(y)%20%3D%20g(y-a)%20-%20g(y)%24%20is%20also%20%24L%5Eq%24%2C%20in%20conclusion%20we%20have%20%24%24%7C%20h(x-a)%20-%20h(x)%20%7C%20%5Cle%20%5C%7C%20f%20%5C%7C_p%20%5C%7C%20g'%20%5C%7C_q%20%3C%20%5Cinfty.%24%24%20In%20particular%2C%20for%20any%20given%20%24%5Cvarepsilon%20choose%20%24%5Cdelta%20%3E%200%24%20so%20that%20%24%5C%7C%20g'%20%5C%7C_q%20%3C%20%5Cvarepsilon%2F%5C%7C%20f%20%5C%7C_p%24%20whenever%20%24%7C%20a%20%7C%20%3C%20%5Cdelta%24%20(this%20is%20apparently%20doable%2C%20because%20%24g(y-a)%20%5Cto%20g(y)%24%20with%20respect%20to%20%24L%5Eq%24%20norm%20when%20%24a%20%5Cto%200%24)%2C%20then%20%24%7C%20h(x-a)%20-%20h(x)%20%7C%20%3C%20%5Cvarepsilon%24.%20Since%20%24a%24%20does%20not%20depend%20on%20%24x%24%2C%20we%20have%20uniform%20continuity.%0A%0A%20%20%20%20Suppose%20now%20%241%20%3C%20p%2Cq%20%3C%20%5Cinfty%24.%20We%20know%20that%20%24C_C(%5Cmathbb%7BR%7D)%24%20is%20dense%20in%20%24L%5Ep%24%20and%20%24L%5Eq%24%20(and%20it%20fails%20to%20the%20%24L%5E%7B%5Cinfty%7D%24%20case%2C%20so%20this%20argument%20does%20not%20apply%20to%20%24p%20%3D%201%2C%20q%20%3D%5Cinfty%24%20at%20this%20very%20beginning).%20So%20choose%20sequences%20%24f_n%24%20and%20%24g_n%24%20so%20that%20%24f_n%20%5Cto%20f%24%20and%20%24g_n%20%5Cto%20g%24%20with%20respect%20to%20%24L%5Ep%24%20and%20%24L%5Eq%24%20norm%20respectively.%20For%20each%20%24n%24%2C%20denote%20%24h_n%20%3D%20f_n%20*%20g_n%24.%20First%20we%20will%20show%20that%20%24h_n%24%20is%20continuous%20and%20compact%20supported%2C%20and%20then%20we%20show%20%24h_n%20%5Cto%20h%24%20with%20respect%20to%20the%20%24%5Csup%24%20norm.%20Since%20we%20know%20%24C_0%24%20is%20completion%20of%20%24C_C%24%2C%20these%20would%20imply%20that%20%24h%24%20is%20in%20%24C_0%24%3A%0A%0A%20%20%20%201.%20First%20we%20fix%20%24n%24%20and%20show%20%24h_n%24%20is%20continuous%20and%20compact%20supported%3A%0A%0A%20%20%20%20%09Since%20%24f_n%24%20is%20continuous%20and%20compact%20supported%2C%20it%20is%20uniformly%20continuous.%20Fix%20%24%5Cvarepsilon%20%3E%200%24%2C%20choose%20%24%5Cdelta%20%3E%200%24%20so%20that%20%24%7C%20f_n(x)%20-%20f_n(y)%20%7C%20%3C%20%5Cvarepsilon%24%20whenever%20%24%7C%20x%20-%20y%20%7C%20%3C%20%5Cdelta%24%2C%20then%20%24%24%7C%20h_n(x)%20-%20h_n(y)%20%7C%20%5Cle%20%5Cint_%7B-%5Cinfty%7D%5E%7B%5Cinfty%7D%7C%20f_n(x-z)-%20f_n(y-z)%20%7C%20%5Ccdot%20%7C%20g_n(z)%20%7C%20dz%20%5Cle%20%5Cvarepsilon%20%5Cint_%7B-%5Cinfty%7D%5E%7B%5Cinfty%7D%7C%20g_n(z)%20%7C%20dz%2C%24%24%20now%20use%20the%20fact%20that%20%24g_n%24%20is%20a%20compact%20supported%20function%2C%20we%20can%20conclude%20that%20%24%5Cint_%7B-%5Cinfty%7D%5E%7B%5Cinfty%7D%7C%20g_n(z)%20%7C%20dz%24%20is%20finite%20thus%20the%20above%20quantity%20goes%20to%20%240%24%20as%20%24%5Cvarepsilon%20%5Cto%200%24%2C%20which%20proves%20(uniform)%20continuity%20of%20%24h_n%24.%0A%0A%20%20%20%20%09Choose%20%24M%24%20large%20enough%20so%20that%20support%20of%20%24f_n%24%20and%20%24g_n%24%20both%20lie%20in%20%24%5B-M%2C%20M%5D%24%2C%20then%20for%20any%20%24x%24%20so%20that%20%24%7C%20x%20%7C%20%3E%202M%24%2C%20if%20%24y%20%5Cin%20%5B-M%2C%20M%5D%24%20then%20%24x%20-%20y%20%5Cnotin%20%5B-M%2C%20M%5D%24.%20This%20implies%20that%20for%20%24%7C%20x%20%7C%20%3E%202M%24%2C%20we%20have%20%24f_n(x-y)%20g_n(y)%20%3D%200%24%20for%20any%20%24y%24%2C%20which%20means%20%24h_n(x)%20%3D%200%24%2C%20meaning%20that%20the%20support%20of%20%24h_n%24%20lie%20inside%20%24%5B-2M%2C%202M%5D%24%20thus%20is%20bounded%20thus%20compact%2C%20thus%20%24h_n%24%20is%20compact%20supported%3B%0A%20%20%20%202.%20Now%20we%20show%20%24h_n%20%5Cto%20h%24%20with%20respect%20to%20the%20%24%5Csup%24%20norm.%20We%20have%3A%20%24%24%5Cbegin%7Baligned%7D%20%7C%20h_n(x)%20-%20h(x)%20%7C%20%26%5Cle%20%7C%20h_n(x)%20-%20(f_n%20*%20g)(x)%20%7C%20%2B%20%7C%20(f_n*%20g)(x)%20-%20h(x)%20%7C%20%5C%5C%26%3D%7C%20(f_n%20*%20(g_n%20-%20g))(x)%20%7C%20%2B%20%7C%20((f_n-f)*%20g)(x)%20%7C%20%5C%5C%20%26%20%5Cle%20%5C%7C%20f_n%20%5C%7C_p%5C%7C%20g_n%20-%20g%20%5C%7C_q%20%2B%20%5C%7C%20f_n%20-%20f%20%5C%7C_p%20%5C%7C%20g%20%5C%7C_q%20%26(*)%20%5C%5C%20%26%5Cto%200%20%5Cend%7Baligned%7D%24%24%20where%20(%24*%24)%20comes%20from%20Holder%20(say%20%24f%20%5Cin%20L%5Ep%24%20and%20%24g%20%5Cin%20L%5Eq%24)%3A%20%24%24%5Cbegin%7Baligned%7D%20%7C%20(f%20*%20g%20)%20(x)%7C%20%26%3D%20%7C%20%5Cint%20f(x-y)g(y)dy%7C%20%5C%5C%20%26%5Cle%20%5Cint%20%7C%20f(x-y)%20%5C%7C%20g(y)%20%7C%20dy%20%5C%5C%20%26%5Cle%20(%5Cint%20%7C%20f(x-y)%20%7C%5Ep%20dy)%5E%7B1%2Fp%7D(%5Cint%20%7C%20g(y)%20%7C%5Eq%20dy)%5E%7B1%2Fq%7D%20%5C%5C%20%26%3D%20%5C%7C%20f%20%5C%7C_p%20%5C%7C%20g%20%5C%7C_q%5Cend%7Baligned%7D%24%24%20This%20proves%20convergence%2C%20and%20by%20the%20argument%20above%2C%20we%20finish%20the%20proof.%0A%0A%20%20%20%20For%20%24p%20%3D%201%24%20and%20%24q%20%3D%20%5Cinfty%24%3A%20consider%20%24f%20%3D%20%5Cchi_%7B%5B-1%2C1%5D%7D%24%20and%20%24g%20%5Cequiv%201%24%20(which%20is%20%24L%5E%7B%5Cinfty%7D%24%20but%20is%20not%20%24L%5Eq%24%20for%20any%20finite%20%241%20%5Cle%20q%20%3C%20%5Cinfty%24)%2C%20then%20%24h(x)%20%3A%3D%20%5Cint%20%5Cchi_%7B%5B-1%2C1%5D%7D(x-y)%20dy%20%5Cequiv%202%24%20so%20%24h%20%5Cnotin%20C_0%24.%0A%0A%20%20%20%20%23%23%20Exercise%209%0A%0A%20%20%20%20**Suppose%20%241%20%5Cle%20p%20%3C%20%5Cinfty%24%2C%20%24f%20%5Cin%20L%5Ep%24%2C%20and**%20%24%24g(x)%20%3D%20%5Cint_x%5E%7Bx%2B1%7Df(t)dt.%24%24%20**Prove%20that%20%24g%20%5Cin%20C_0%24.%20What%20can%20you%20say%20about%20%24g%24%20if%20%24f%20%5Cin%20L%5E%7B%5Cinfty%7D%24**%3F%0A%0A%20%20%20%20Proof%0A%0A%20%20%20%20First%20we%20need%20to%20show%20%24g%24%20is%20continuous.%20Write%20%24g(x)%20%3D%20%5Cint_0%5E1f(x%2Bt)dt%24%20and%20denote%20%24f(x%2Bt)%24%20as%20%24f_x(t)%24.%20Then%20%24%7C%20g(x)%20-%20g(y)%20%7C%20%3D%20%7C%20%5Cint_0%5E1f_xdt%20-%20%5Cint_0%5E1%20f_ydt%20%7C%24%20%24%5Cle%20%5Cint_0%5E1%7C%20f_x%20-%20f_y%20%7C%20dt%24%20%24%5Cle%20(%5Cint_0%5E1%201%5Eqdt)%5E%7B1%2Fq%7D%20%5Ccdot%20(%5Cint_0%5E1%7C%20f_x%20-%20f_y%20%7C%5Ep)%5E%7B1%2Fp%7D%24%20%24%3D%20%5C%7C%20f_x%20-%20f_y%20%5C%7C_p%24.%20By%20Theorem%209.5%2C%20%24x%20%5Cmapsto%20f_x%24%20is%20uniformly%20continuous%2C%20thus%20%24g%24%20is%20also%20(uniformly)%20continuous.%0A%0A%20%20%20%20If%20%24g%20%5Cnotin%20C_0%24%2C%20then%20by%20definition%20there%20is%20some%20%24%5Cvarepsilon%20%3E%200%24%20so%20that%20there%20is%20no%20compact%20set%20%24K%20%5Csubset%20%5Cmathbb%7BR%7D%24%20with%20%24%7C%20g(x)%20%7C%20%3C%20%5Cvarepsilon%24%20for%20all%20%24x%24%20outside%20%24K%24.%20In%20particular%2C%20there%20is%20a%20sequence%20%24%5Clbrace%20x_n%20%5Crbrace%24%2C%20which%20we%20may%20choose%20to%20have%20%24x_1%20%3E%200%24%20and%20%24x_n%20-%20x_%7Bn-1%7D%20%3E%201%24%20for%20all%20%24n%24%2C%20and%20%24%7C%20g(x_n)%20%7C%20%5Cge%20%5Cvarepsilon%24%20for%20any%20%24n%24.%20But%20then%20%24%24%5Cint_%7B-%5Cinfty%7D%5E%7B%5Cinfty%7D%7C%20f(t)%20%7C%20%5Epdt%20%5Cge%20%5Csum_n%20%5Cint_%7Bx_n%7D%5E%7Bx_n%2B1%7D%7C%20f(t)%20%7C%5Epdt%20%5Cge%20%5Csum_n%20%5Cvarepsilon%5Ep%20%3D%20%5Cinfty%24%24%20so%20%24f%24%20is%20not%20%24L%5Ep%24%20for%20any%20%241%20%5Cle%20p%20%3C%20%5Cinfty%24.%20This%20proves%20the%20negation%20thus%20the%20original%20statement.%0A%0A%20%20%20%20If%20%24f%20%5Cin%20L%5E%7B%5Cinfty%7D%24%20then%20%24g%24%20may%20not%20be%20%24C_0%24%20(just%20take%20%24f%20%5Cequiv%201%24%20then%20%24g%20%5Cequiv%201%24%20as%20well).%20We%20could%20have%20that%20%24%24%5Cbegin%7Baligned%7D%7C%20g(x)%20%7C%20%26%3D%20%7C%20%5Cint_x%5E%7Bx%2B1%7Df(t)dt%20%7C%20%5Cle%20%5Cint_x%5E%7Bx%2B1%7D%20%7C%20f(t)%20%7C%20dt%20%20%5C%5C%20%26%5Cle%5Cint_x%5E%7Bx%2B1%7D%5C%7C%20f%20%5C%7C_%7B%5Cinfty%7Ddt%20%5C%5C%20%26%3D%20%5C%7C%20f%20%5C%7C_%7B%5Cinfty%7D%5Cend%7Baligned%7D%24%24%20(because%20%24%5C%7C%20f%20%5C%7C_%7B%5Cinfty%7D%24%20is%20independent%20of%20%24x%24)%2C%20so%20%24g%24%20is%20%24L%5E%7B%5Cinfty%7D%24%20and%20%24%5C%7C%20g%20%5C%7C_%7B%5Cinfty%7D%24%20is%20bounded%20by%20%24%5C%7C%20f%20%5C%7C_%7B%5Cinfty%7D%24.%0A%0A%20%20%20%20Also%20%24g%20%5Cin%20%5Ctext%7BLip%7D~1%24%3A%20%24%24%5Cbegin%7Baligned%7D%7C%20g(x)%20-%20g(y)%20%7C%20%26%3D%20%7C%20%5Cint_x%5E%7Bx%2B1%7Dfdt%20-%20%5Cint_%7By%7D%5E%7By%2B1%7Dfdt%20%7C%20%5C%5C%20%26%20%5Cle%20%7C%20%5Cint_%7By%2B1%7D%5E%7Bx%2B1%7Dfdt%20%7C%20%2B%20%7C%20%5Cint_x%5Eyfdt%7C%20%26(1)%5C%5C%20%26%20%5Cle%202%5C%7C%20f%20%5C%7C_%7B%5Cinfty%7D%7C%20xy%20%7C.%5Cend%7Baligned%7D%24%24%20(in%20fact%20if%20%24f%20%5Cin%20L%5Ep%24%20then%20%24g%20%5Cin%20%5Ctext%7BLip%7D~%5Calpha%24%20for%20%24%5Calpha%20%3D%201%2Fq%24%20for%20%24p%24%20other%20than%20%24%5Cinfty%24%20as%20well%3A%20continue%20from%20%24(1)%24%20above%2C%20apply%20Holder%20and%20we%20get%20%24%7C%20g(x)%20-%20g(y)%20%7C%20%5Cle%202%5C%7C%20f%20%5C%7C_p%20%7C%20x-y%7C%5E%7B1%2Fq%7D%24)%0A%0A%20%20%20%20%23%23%20Exercise%2010%0A%0A%20%20%20%201.%20**Let%20%24C%5E%7B%5Cinfty%7D%24%20be%20the%20class%20of%20all%20infinitely%20differentiable%20complex%20functions%20on%20%24%5Cmathbb%7BR%7D%24%2C%20and%20let%20%24C_C%5E%7B%5Cinfty%7D%24%20consist%20of%20all%20%24g%20%5Cin%20C%5E%7B%5Cinfty%7D%24%20whose%20supports%20are%20compact.%20Show%20that%20%24C_C%5E%7B%5Cinfty%7D%24%20does%20not%20consists%20of%20%240%24%20alone**%3B%0A%20%20%20%202.%20**Let%20%24L%5E1_%7Bloc%7D%24%20be%20the%20class%20of%20all%20%24f%24%20belong%20to%20%24L%5E1%24%20locally%3B%20that%20is%2C%20%24f%20%5Cin%20L%5E1_%7Bloc%7D%24%20provided%20that%20%24f%24%20is%20measurable%20and%20%24%5Cint_I%7C%20f%20%7C%20%3C%20%5Cinfty%24%20for%20every%20bounded%20interval%20%24I%24.%20If%20%24f%20%5Cin%20L%5E1_%7Bloc%7D%24%20and%20%24g%20%5Cin%20C_C%5E%7B%5Cinfty%7D%24%2C%20prove%20that%20%24f%20*%20g%20%5Cin%20C%5E%7B%5Cinfty%7D%24**%3B%0A%20%20%20%203.%20**Prove%20that%20there%20are%20sequences%20%24%5Clbrace%20g_n%20%5Crbrace%24%20in%20%24C%5E%7B%5Cinfty%7D_C%24%20such%20that**%20%24%24%5C%7C%20f*%20g_n%20-%20f%20%5C%7C_1%20%5Cto%200%24%24%20**as%20%24n%20%5Cto%20%5Cinfty%24%2C%20for%20every%20%24f%20%5Cin%20L%5E1%24%20(compare%20Theorem%209.10).%20Prove%20that%20%24%5Clbrace%20g_n%20%5Crbrace%24%20can%20also%20be%20so%20chosen%20that%20%24(f*%20g_n)(x)%20%5Cto%20f(x)%24%20a.e.%2C%20for%20every%20%24f%20%5Cin%20L_%7Bloc%7D%5E1%24%3B%20in%20fact%2C%20for%20suitable%20%24%5Clbrace%20g_n%20%5Crbrace%24%20the%20convergence%20occurs%20at%20every%20point%20%24x%24%20at%20which%20%24f%24%20is%20the%20derivative%20of%20its%20indefinite%20integral**.%0A%0A%20%20%20%20Proof%0A%0A%20%20%20%20%23%23%23%20Part%201%0A%0A%20%20%20%20A%20famous%20example%20is%20%24%5CPsi(x)%20%3D%20%5Cbegin%7Bcases%7D%20e%5E%7B-%5Cfrac%7B1%7D%7B1-x%5E2%7D%7D%2C%26x%20%5Cin%20(-1%2C1)%5C%5C%200%2C%26%5Ctext%7Botherwise%7D%5Cend%7Bcases%7D%24.%20It%20is%20smooth%20and%20clearly%20compact%20supported%2C%20i.e.%20it%20is%20a%20non-trivial%20member%20of%20%24C_C%5E%7B%5Cinfty%7D%24.%20It%20is%20so%20called%20the%20'bump%20function'.%0A%0A%20%20%20%20%23%23%23%20Part%202%0A%0A%20%20%20%20Since%20%24g%20%5Cin%20C_C%5E%7B%5Cinfty%7D%24%2C%20let%20us%20assume%20support%20of%20%24g%24%20is%20contained%20in%20some%20interval%20%24%5B-M%2C%20M%5D%24%20(which%20is%20apparently%20bounded).%20First%2C%20%24%24(f%20*%20g)(x)%20%3D%20%5Cint_%7B-%5Cinfty%7D%5E%7B%5Cinfty%7Df(x-y)g(y)dy%20%3D%20%5Cint_%7B-M%7D%5EM%20f(x-y)g(y)dy%24%24%20where%20by%20assumption%20of%20%24f%24%2C%20on%20this%20interval%20%24f%24%20is%20bounded%2C%20and%20%24g%24%20is%20continuous%20on%20this%20compact%20interval%20thus%20also%20bounded%2C%20so%20the%20above%20quantity%20is%20bounded%2C%20which%20means%20%24f*%20g%24%20is%20well-defined.%0A%0A%20%20%20%20Now%20we%20claim%20that%20%24(f%20*%20g)'%20%3D%20f%20*%20g'%24%2C%20if%20we%20prove%20this%20claim%20then%20it%20is%20not%20hard%20to%20see%20%24g'%24%20is%20also%20in%20%24C_C%5E%7B%5Cinfty%7D%24%20(because%20whenever%20%24g%20%3D%200%24%2C%20%24g'%24%20has%20to%20be%20%240%24)%2C%20thus%20%24(f*%20g)%5E%7B(n)%7D%20%3D%20f%20*%20g%5E%7B(n)%7D%24%20which%20proves%20%24f%20*%20g%24%20is%20infinitely%20differentiable%20thus%20in%20%24C%5E%7B%5Cinfty%7D%24.%20Proof%20is%20as%20following%3A%0A%0A%20%20%20%20First%20we%20have%20(let%20%24h%24%20be%20a%20number)%3A%20%24%24%5Cbegin%7Baligned%7D%20%5Cfrac%7B(f*%20g)(x%2Bh)%20-%20(f*%20g)(x)%7D%7Bh%7D%20%26%3D%20%5Cfrac%7B1%7D%7Bh%7D%20(%5Cint_%7B-M%7D%5EMf(x%2Bh-y)g(y)dy%20-%20%5Cint_%7B-M%7D%5EMf(x-y)g(y)dy)%20%5C%5C%20%26%3D%20%5Cfrac%7B1%7D%7Bh%7D%20(%5Cint_%7B-M%2Bx%2Bh%7D%5E%7BM%2Bx%2Bh%7Df(y)g(x%2Bh-y)dy%20-%20%5Cint_%7B-M%2Bx%7D%5E%7BM%2Bx%7Df(y)g(x-y)dy)%20%5C%5C%20%26%3D%20%5Cfrac%7B1%7D%7Bh%7D%20(%5Cint_%7B-M%2Bx%7D%5E%7BM%2Bx%7Df(y)g(x%2Bh-y)dy%20-%20%5Cint_%7B-M%2Bx%7D%5E%7BM%2Bx%7Df(y)g(x-y)dy)%20%5C%5C%20%26%3D%20%5Cint_%7B-M%2Bx%7D%5E%7BM%2Bx%7Df(y)%5Ccdot%5Cfrac%7Bg(x%2Bh-y)-g(x-y)%7D%7Bh%7Ddy%5Cend%7Baligned%7D%24%24%20(using%20the%20fact%20that%20%24g%24%20is%20%240%24%20outside%20of%20%24%5B-M%2CM%5D%24%2C%20change%20of%20variable%2C%20and%20translate%20invariant).%20Now%20we%20send%20%24h%20%5Cto%200%24%2C%20then%20the%20LHS%20is%20by%20definition%20%24(f%20*%20g)'(x)%24%2C%20and%20RHS%20is%20%24%24%5Clim_%7Bh%20%5Cto%200%7D%5Cint_%7B-M%2Bx%7D%5E%7BM%2Bx%7Df(y)%5Ccdot%5Cfrac%7Bg(x%2Bh-y)-g(x-y)%7D%7Bh%7Ddy.%24%24%20If%20we%20view%20the%20integrant%20as%20a%20sequence%20of%20function%20indexed%20on%20%24h%24%2C%20it%20is%20dominated%20by%20some%20%24L%5E1%24%20function%3A%20%24f%24%20is%20in%20%24L%5E%7Bloc%7D_1%24%2C%20and%20the%20expression%20with%20%24g%24%20is%20dominated%20by%20some%20constant%20(on%20a%20bounded%20interval%20thus%20%24L%5E1%24)%20because%20%24g%24%20is%20uniformly%20continuous%20and%20differentiable%20thus%20the%20derivative%20is%20bounded.%20Thus%20by%20Lebesgue%20Dominated%20Convergence%20Theorem%20we%20can%20interchange%20the%20%24%5Clim%24%20and%20%24%5Cint%24%20operators%20and%20get%3A%20%24%24%5Cbegin%7Baligned%7D(f*%20g)'(x)%20%26%3D%20%5Cint_%7B-M%2Bx%7D%5E%7BM%2Bx%7D%5Clim_%7Bh%20%5Cto%200%7D%5Cfrac%7Bg(x-y%2Bh)%20-%20g(x-y)%7D%7Bh%7Df(y)dy%20%5C%5C%20%26%3D%20%5Cint_%7B-M%2Bx%7D%5E%7BM%2Bx%7Dg'(x-y)%20f(y)%20dy%20%5C%5C%26%20%3D%20(g'*%20f)(x)%20%3D%20(f*%20g')(x)%5Cend%7Baligned%7D%24%24%20as%20claimed.%0A%0A%20%20%20%20%23%23%23%20Part%203%0A%0A%20%20%20%20Consider%20the%20about%20bump%20function%2C%20rescale%20so%20that%20we%20assume%20%24%5Cint_%7B-1%7D%5E1%5CPsi(x)dx%20%3D%201%24.%20For%20each%20%24%5Cvarepsilon0%24%2C%20define%20%24%5CPsi_%7B%5Cvarepsilon%7D(x)%20%3D%20%5Cfrac%7B%5CPsi(x%2F%5Cvarepsilon)%7D%7B%5Cvarepsilon%7D%24%2C%20it%20is%20not%20hard%20to%20see%20%24%5CPsi_%7B%5Cvarepsilon%7D%24%20is%20in%20%24C_C%5E%7B%5Cinfty%7D%24%2C%20supported%20in%20%24%5B-%5Cvarepsilon%2C%5Cvarepsilon%5D%24%2C%20and%20have%20%24L%5E1%24%20norm%20%24%3D1%24.%20We%20have%3A%20%24%24%5Cbegin%7Baligned%7D%20%5C%7C%20f%20*%20%5CPsi_%7B%5Cvarepsilon%7D%20-%20f%20%5C%7C_1%20%26%3D%20%5Cint_%7B-%5Cinfty%7D%5E%7B%5Cinfty%7D%7C%20(f%20*%20%5CPsi_%7B%5Cvarepsilon%7D)(x)%20-%20f(x)%7C%20dx%20%5C%5C%20%26%3D%20%5Cint_%7B-%5Cinfty%7D%5E%7B%5Cinfty%7D%7C%20%5Cint_%7B-%5Cinfty%7D%5E%7B%5Cinfty%7D%20(f(x-y)%20-%20f(x))%5CPsi_%7B%5Cvarepsilon%7D(y)%20dy%20%7C%20dx%20%20%5C%5C%20%26%3D%20%5Cint_%7B-%5Cinfty%7D%5E%7B%5Cinfty%7D%20%7C%20%5Cint_%7B-%5Cinfty%7D%5E%7B%5Cinfty%7D%20(f(x-y)%20-%20f(x))%20%5Cfrac%7B%5CPsi(y%2F%5Cvarepsilon)%7D%7B%5Cvarepsilon%7Ddy%20%7C%20dx%20%5C%5C%20%26%3D%5E%7Bz%20%3D%20y%2F%5Cvarepsilon%7D%20%5Cint_%7B-%5Cinfty%7D%5E%7B%5Cinfty%7D%20%7C%20%5Cint_%7B-%5Cinfty%7D%5E%7B%5Cinfty%7D(f(x-%5Cvarepsilon%20z)-%20f(x))%5CPsi(z)dz%20%7C%20dx%20%5C%5C%20%26%5Cle%20%20%5Cint_%7B-%5Cinfty%7D%5E%7B%5Cinfty%7D%20%5Cint_%7B-%5Cinfty%7D%5E%7B%5Cinfty%7D%20%7C%20f(x-%5Cvarepsilon%20z)%20-%20f(x)%20%5C%7C%20%5CPsi(z)%20%7C%20dzdx%20%5C%5C%20%26%3D%20%5Cint_%7B-%5Cinfty%7D%5E%7B%5Cinfty%7D%5Cint_%7B-%5Cinfty%7D%5E%7B%5Cinfty%7D%20%7C%20f(x%20-%20%5Cvarepsilon%20z)%20-%20f(x)%20%5C%7C%20%5CPsi(z)%20%7C%20dxdz%20%5C%5C%20%26%3D%20%5Cint_%7B-%5Cinfty%7D%5E%7B%5Cinfty%7D%20%5C%7C%20f_%7B%5Cvarepsilon%20z%7D%20-%20f%20%5C%7C_1%20%7C%20%5CPsi(z)%20%7C%20dz%5Cend%7Baligned%7D%24%24%20Prove%20that%20%24(f*%20h_%7B%5Clambda%7D)(x)%20%5Cto%20f(x)%24%20a.e.%20if%20%24f%20%5Cin%20L%5E1%24%2C%20as%20%24%5Clambda%20%5Cto%200%24%2C%20and%20that%20%24f%20*%20h_%7B%5Clambda%7D%20%5Cin%20C%5E%7B%5Cinfty%7D%24%2C%20although%20%24h_%7B%5Clambda%7D%24%20does%20not%20have%20compact%20support%20(%24h_%7B%5Clambda%7D%24%20as%20defined%20in%20Section%209.7).%0A%20%20%20%20%22%22%22)%0A%20%20%20%20return%0A%0A%0Aif%20__name__%20%3D%3D%20%22__main__%22%3A%0A%20%20%20%20app.run()%0A