$f'$ is bounded and $suplimits_cin OVert f'(c) Vert<infty$ if and only if $f$ is globally Lipschitz

The name of the pictureThe name of the pictureThe name of the pictureClash Royale CLAN TAG#URR8PPP











up vote
0
down vote

favorite












Let $f:Osubset BbbR^ntoBbbR^m$ be differentiable and O be convex. Prove that the following are equivalent:



$i$. $f'$ is bounded and $suplimits_cin OVert f'(c) Vert<infty;$



$ii$. $f$ is globally Lipschitz.



My trial:



Proving $iimplies ii$. Let $x,yin O$ such that $[x,y]subset 0.$ Then, by MVT, $exists ;cin [x,y]$ such that beginalignVert f(x)-f(y) Vertleq suplimits_cin [x,y]Vert f'(c) VertVert x-y Vertendalign
beginalignqquadqquadqquadleq suplimits_cin OVert f'(c) VertVert x-y Vertendalign
Since $f'$ is bounded and $suplimits_cin OVert f'(c) Vert<infty;$ then, let $k=suplimits_cin OVert f'(c) Vert.$ So, we have beginalignVert f(x)-f(y) Vertleq kVert x-y Vertendalign This implies that $f$ is globally bounded.



Proving $iiimplies i$. Let $f$ be globally globally Lipschitz, then $exists;k>0$ such that beginalignVert f(x)-f(y) Vertleq kVert x-y Vert,;;forall ;x,yin,Oendalign
beginalignfracVert f(x)-f(y) VertVert x-y Vertleq k,;;forall ;x,yin,Oendalign
Since $f$ is differentiable, we take the directional derivative as $xto y$, we have
beginalignVert f'(y) Vertleq k,;;forall ;yin,Oendalign
Taking $sup,$ we have that
beginalignsuplimits_yin OVert f'(y) Vertleq k.endalign
Please, I'm I correct? If not, alternative proofs will be highly regarded.







share|cite|improve this question


























    up vote
    0
    down vote

    favorite












    Let $f:Osubset BbbR^ntoBbbR^m$ be differentiable and O be convex. Prove that the following are equivalent:



    $i$. $f'$ is bounded and $suplimits_cin OVert f'(c) Vert<infty;$



    $ii$. $f$ is globally Lipschitz.



    My trial:



    Proving $iimplies ii$. Let $x,yin O$ such that $[x,y]subset 0.$ Then, by MVT, $exists ;cin [x,y]$ such that beginalignVert f(x)-f(y) Vertleq suplimits_cin [x,y]Vert f'(c) VertVert x-y Vertendalign
    beginalignqquadqquadqquadleq suplimits_cin OVert f'(c) VertVert x-y Vertendalign
    Since $f'$ is bounded and $suplimits_cin OVert f'(c) Vert<infty;$ then, let $k=suplimits_cin OVert f'(c) Vert.$ So, we have beginalignVert f(x)-f(y) Vertleq kVert x-y Vertendalign This implies that $f$ is globally bounded.



    Proving $iiimplies i$. Let $f$ be globally globally Lipschitz, then $exists;k>0$ such that beginalignVert f(x)-f(y) Vertleq kVert x-y Vert,;;forall ;x,yin,Oendalign
    beginalignfracVert f(x)-f(y) VertVert x-y Vertleq k,;;forall ;x,yin,Oendalign
    Since $f$ is differentiable, we take the directional derivative as $xto y$, we have
    beginalignVert f'(y) Vertleq k,;;forall ;yin,Oendalign
    Taking $sup,$ we have that
    beginalignsuplimits_yin OVert f'(y) Vertleq k.endalign
    Please, I'm I correct? If not, alternative proofs will be highly regarded.







    share|cite|improve this question
























      up vote
      0
      down vote

      favorite









      up vote
      0
      down vote

      favorite











      Let $f:Osubset BbbR^ntoBbbR^m$ be differentiable and O be convex. Prove that the following are equivalent:



      $i$. $f'$ is bounded and $suplimits_cin OVert f'(c) Vert<infty;$



      $ii$. $f$ is globally Lipschitz.



      My trial:



      Proving $iimplies ii$. Let $x,yin O$ such that $[x,y]subset 0.$ Then, by MVT, $exists ;cin [x,y]$ such that beginalignVert f(x)-f(y) Vertleq suplimits_cin [x,y]Vert f'(c) VertVert x-y Vertendalign
      beginalignqquadqquadqquadleq suplimits_cin OVert f'(c) VertVert x-y Vertendalign
      Since $f'$ is bounded and $suplimits_cin OVert f'(c) Vert<infty;$ then, let $k=suplimits_cin OVert f'(c) Vert.$ So, we have beginalignVert f(x)-f(y) Vertleq kVert x-y Vertendalign This implies that $f$ is globally bounded.



      Proving $iiimplies i$. Let $f$ be globally globally Lipschitz, then $exists;k>0$ such that beginalignVert f(x)-f(y) Vertleq kVert x-y Vert,;;forall ;x,yin,Oendalign
      beginalignfracVert f(x)-f(y) VertVert x-y Vertleq k,;;forall ;x,yin,Oendalign
      Since $f$ is differentiable, we take the directional derivative as $xto y$, we have
      beginalignVert f'(y) Vertleq k,;;forall ;yin,Oendalign
      Taking $sup,$ we have that
      beginalignsuplimits_yin OVert f'(y) Vertleq k.endalign
      Please, I'm I correct? If not, alternative proofs will be highly regarded.







      share|cite|improve this question














      Let $f:Osubset BbbR^ntoBbbR^m$ be differentiable and O be convex. Prove that the following are equivalent:



      $i$. $f'$ is bounded and $suplimits_cin OVert f'(c) Vert<infty;$



      $ii$. $f$ is globally Lipschitz.



      My trial:



      Proving $iimplies ii$. Let $x,yin O$ such that $[x,y]subset 0.$ Then, by MVT, $exists ;cin [x,y]$ such that beginalignVert f(x)-f(y) Vertleq suplimits_cin [x,y]Vert f'(c) VertVert x-y Vertendalign
      beginalignqquadqquadqquadleq suplimits_cin OVert f'(c) VertVert x-y Vertendalign
      Since $f'$ is bounded and $suplimits_cin OVert f'(c) Vert<infty;$ then, let $k=suplimits_cin OVert f'(c) Vert.$ So, we have beginalignVert f(x)-f(y) Vertleq kVert x-y Vertendalign This implies that $f$ is globally bounded.



      Proving $iiimplies i$. Let $f$ be globally globally Lipschitz, then $exists;k>0$ such that beginalignVert f(x)-f(y) Vertleq kVert x-y Vert,;;forall ;x,yin,Oendalign
      beginalignfracVert f(x)-f(y) VertVert x-y Vertleq k,;;forall ;x,yin,Oendalign
      Since $f$ is differentiable, we take the directional derivative as $xto y$, we have
      beginalignVert f'(y) Vertleq k,;;forall ;yin,Oendalign
      Taking $sup,$ we have that
      beginalignsuplimits_yin OVert f'(y) Vertleq k.endalign
      Please, I'm I correct? If not, alternative proofs will be highly regarded.









      share|cite|improve this question













      share|cite|improve this question




      share|cite|improve this question








      edited Aug 22 at 14:33

























      asked Aug 22 at 7:26









      Mike

      75615




      75615




















          1 Answer
          1






          active

          oldest

          votes

















          up vote
          1
          down vote



          accepted










          (i) There is something fishy here: You cannot say "$exists c$ such that $Phi$ holds" and then take the sup over all $c$ in $Phi$. Furthermore the statement is wrong if $O$ is not convex. As an example let $f$ be the principal value of the polar angle in the plane slit along the negative $x$-axis. Here $f(-1,epsilon)-f(-1,-epsilon)to2pi$ while $epsilonto0+$.



          Given $x$, $yin O$ you can argue as follows: If $z:=f(y)-f(x)ne0$ consider the auxiliary real-valued function
          $$phi(t)=zcdot fbigl((1-t)x+t ybigr)qquad(0leq tleq 1)$$
          $bigl($note that $(1-t)x+tyin O$ when $0leq tleq 1bigr)$. By the MVT there is a $tauin>]0,1[>$ such that
          $$zcdotbigl(f(y)-f(x)bigr)=phi(1)-phi(0)= phi'(tau)=zcdotbigl( f'bigl((1-tau)x+tau ybigr).(y-x)bigr) ,$$
          and therefore $|z|^2leq |z|sup_cin O|f'(c)|>|y-x|$, or
          $$bigr|f(y)-f(x)bigl|=|z|leq k>|y-x| .$$



          (ii) It's unclear what the directional derivative is for a vector valued function. We need a clear cut proof that the derivative $f'(p)=:Aincal L(mathbb R^n,mathbb R^m)$ at a given point $pin O$ has norm $leq L$, where $L$ is a global Lipschitz constant for $f$. To this end we go back to the definition of $f'(p)$. Let $u$ be an arbitrary unit vector. Then
          $$f(p+t u)- f(p)= A.t u + o(t)qquad(tto0) .$$
          This implies
          $$A.u=f(p+ tu)-f(p)over t +o(1)qquad(tto0) ,$$
          hence
          $$|A.u|leq+o(1)leq L+o(1)qquad(tto0) .tag1$$
          As the LHS of $(1)$ does not depend on $t$ it follows that $|A.u|leq L$, and since $uin S^n-1$ was arbitrary we may conclude that $|A|leq L$.






          share|cite|improve this answer






















          • But $phi$ has to be $C^1$ for that to hold!
            – Mike
            Aug 22 at 9:01










          • O.k. See my edit. (I still have to see a naturally occurring multivariate function that is differentiable, but not continuously differentiable$ldots$)
            – Christian Blatter
            Aug 22 at 10:13










          • Thanks for your feedback Sir!
            – Mike
            Aug 22 at 13:37










          • Thanks for all but I think the first approach is right if I include the theorem I used. Mean Value Theorem for higher dimensions.
            – Mike
            Aug 22 at 14:33










          Your Answer




          StackExchange.ifUsing("editor", function ()
          return StackExchange.using("mathjaxEditing", function ()
          StackExchange.MarkdownEditor.creationCallbacks.add(function (editor, postfix)
          StackExchange.mathjaxEditing.prepareWmdForMathJax(editor, postfix, [["$", "$"], ["\\(","\\)"]]);
          );
          );
          , "mathjax-editing");

          StackExchange.ready(function()
          var channelOptions =
          tags: "".split(" "),
          id: "69"
          ;
          initTagRenderer("".split(" "), "".split(" "), channelOptions);

          StackExchange.using("externalEditor", function()
          // Have to fire editor after snippets, if snippets enabled
          if (StackExchange.settings.snippets.snippetsEnabled)
          StackExchange.using("snippets", function()
          createEditor();
          );

          else
          createEditor();

          );

          function createEditor()
          StackExchange.prepareEditor(
          heartbeatType: 'answer',
          convertImagesToLinks: true,
          noModals: false,
          showLowRepImageUploadWarning: true,
          reputationToPostImages: 10,
          bindNavPrevention: true,
          postfix: "",
          noCode: true, onDemand: true,
          discardSelector: ".discard-answer"
          ,immediatelyShowMarkdownHelp:true
          );



          );








           

          draft saved


          draft discarded


















          StackExchange.ready(
          function ()
          StackExchange.openid.initPostLogin('.new-post-login', 'https%3a%2f%2fmath.stackexchange.com%2fquestions%2f2890725%2ff-is-bounded-and-sup-limits-c-in-o-vert-fc-vert-infty-if-and-only-i%23new-answer', 'question_page');

          );

          Post as a guest






























          1 Answer
          1






          active

          oldest

          votes








          1 Answer
          1






          active

          oldest

          votes









          active

          oldest

          votes






          active

          oldest

          votes








          up vote
          1
          down vote



          accepted










          (i) There is something fishy here: You cannot say "$exists c$ such that $Phi$ holds" and then take the sup over all $c$ in $Phi$. Furthermore the statement is wrong if $O$ is not convex. As an example let $f$ be the principal value of the polar angle in the plane slit along the negative $x$-axis. Here $f(-1,epsilon)-f(-1,-epsilon)to2pi$ while $epsilonto0+$.



          Given $x$, $yin O$ you can argue as follows: If $z:=f(y)-f(x)ne0$ consider the auxiliary real-valued function
          $$phi(t)=zcdot fbigl((1-t)x+t ybigr)qquad(0leq tleq 1)$$
          $bigl($note that $(1-t)x+tyin O$ when $0leq tleq 1bigr)$. By the MVT there is a $tauin>]0,1[>$ such that
          $$zcdotbigl(f(y)-f(x)bigr)=phi(1)-phi(0)= phi'(tau)=zcdotbigl( f'bigl((1-tau)x+tau ybigr).(y-x)bigr) ,$$
          and therefore $|z|^2leq |z|sup_cin O|f'(c)|>|y-x|$, or
          $$bigr|f(y)-f(x)bigl|=|z|leq k>|y-x| .$$



          (ii) It's unclear what the directional derivative is for a vector valued function. We need a clear cut proof that the derivative $f'(p)=:Aincal L(mathbb R^n,mathbb R^m)$ at a given point $pin O$ has norm $leq L$, where $L$ is a global Lipschitz constant for $f$. To this end we go back to the definition of $f'(p)$. Let $u$ be an arbitrary unit vector. Then
          $$f(p+t u)- f(p)= A.t u + o(t)qquad(tto0) .$$
          This implies
          $$A.u=f(p+ tu)-f(p)over t +o(1)qquad(tto0) ,$$
          hence
          $$|A.u|leq+o(1)leq L+o(1)qquad(tto0) .tag1$$
          As the LHS of $(1)$ does not depend on $t$ it follows that $|A.u|leq L$, and since $uin S^n-1$ was arbitrary we may conclude that $|A|leq L$.






          share|cite|improve this answer






















          • But $phi$ has to be $C^1$ for that to hold!
            – Mike
            Aug 22 at 9:01










          • O.k. See my edit. (I still have to see a naturally occurring multivariate function that is differentiable, but not continuously differentiable$ldots$)
            – Christian Blatter
            Aug 22 at 10:13










          • Thanks for your feedback Sir!
            – Mike
            Aug 22 at 13:37










          • Thanks for all but I think the first approach is right if I include the theorem I used. Mean Value Theorem for higher dimensions.
            – Mike
            Aug 22 at 14:33














          up vote
          1
          down vote



          accepted










          (i) There is something fishy here: You cannot say "$exists c$ such that $Phi$ holds" and then take the sup over all $c$ in $Phi$. Furthermore the statement is wrong if $O$ is not convex. As an example let $f$ be the principal value of the polar angle in the plane slit along the negative $x$-axis. Here $f(-1,epsilon)-f(-1,-epsilon)to2pi$ while $epsilonto0+$.



          Given $x$, $yin O$ you can argue as follows: If $z:=f(y)-f(x)ne0$ consider the auxiliary real-valued function
          $$phi(t)=zcdot fbigl((1-t)x+t ybigr)qquad(0leq tleq 1)$$
          $bigl($note that $(1-t)x+tyin O$ when $0leq tleq 1bigr)$. By the MVT there is a $tauin>]0,1[>$ such that
          $$zcdotbigl(f(y)-f(x)bigr)=phi(1)-phi(0)= phi'(tau)=zcdotbigl( f'bigl((1-tau)x+tau ybigr).(y-x)bigr) ,$$
          and therefore $|z|^2leq |z|sup_cin O|f'(c)|>|y-x|$, or
          $$bigr|f(y)-f(x)bigl|=|z|leq k>|y-x| .$$



          (ii) It's unclear what the directional derivative is for a vector valued function. We need a clear cut proof that the derivative $f'(p)=:Aincal L(mathbb R^n,mathbb R^m)$ at a given point $pin O$ has norm $leq L$, where $L$ is a global Lipschitz constant for $f$. To this end we go back to the definition of $f'(p)$. Let $u$ be an arbitrary unit vector. Then
          $$f(p+t u)- f(p)= A.t u + o(t)qquad(tto0) .$$
          This implies
          $$A.u=f(p+ tu)-f(p)over t +o(1)qquad(tto0) ,$$
          hence
          $$|A.u|leq+o(1)leq L+o(1)qquad(tto0) .tag1$$
          As the LHS of $(1)$ does not depend on $t$ it follows that $|A.u|leq L$, and since $uin S^n-1$ was arbitrary we may conclude that $|A|leq L$.






          share|cite|improve this answer






















          • But $phi$ has to be $C^1$ for that to hold!
            – Mike
            Aug 22 at 9:01










          • O.k. See my edit. (I still have to see a naturally occurring multivariate function that is differentiable, but not continuously differentiable$ldots$)
            – Christian Blatter
            Aug 22 at 10:13










          • Thanks for your feedback Sir!
            – Mike
            Aug 22 at 13:37










          • Thanks for all but I think the first approach is right if I include the theorem I used. Mean Value Theorem for higher dimensions.
            – Mike
            Aug 22 at 14:33












          up vote
          1
          down vote



          accepted







          up vote
          1
          down vote



          accepted






          (i) There is something fishy here: You cannot say "$exists c$ such that $Phi$ holds" and then take the sup over all $c$ in $Phi$. Furthermore the statement is wrong if $O$ is not convex. As an example let $f$ be the principal value of the polar angle in the plane slit along the negative $x$-axis. Here $f(-1,epsilon)-f(-1,-epsilon)to2pi$ while $epsilonto0+$.



          Given $x$, $yin O$ you can argue as follows: If $z:=f(y)-f(x)ne0$ consider the auxiliary real-valued function
          $$phi(t)=zcdot fbigl((1-t)x+t ybigr)qquad(0leq tleq 1)$$
          $bigl($note that $(1-t)x+tyin O$ when $0leq tleq 1bigr)$. By the MVT there is a $tauin>]0,1[>$ such that
          $$zcdotbigl(f(y)-f(x)bigr)=phi(1)-phi(0)= phi'(tau)=zcdotbigl( f'bigl((1-tau)x+tau ybigr).(y-x)bigr) ,$$
          and therefore $|z|^2leq |z|sup_cin O|f'(c)|>|y-x|$, or
          $$bigr|f(y)-f(x)bigl|=|z|leq k>|y-x| .$$



          (ii) It's unclear what the directional derivative is for a vector valued function. We need a clear cut proof that the derivative $f'(p)=:Aincal L(mathbb R^n,mathbb R^m)$ at a given point $pin O$ has norm $leq L$, where $L$ is a global Lipschitz constant for $f$. To this end we go back to the definition of $f'(p)$. Let $u$ be an arbitrary unit vector. Then
          $$f(p+t u)- f(p)= A.t u + o(t)qquad(tto0) .$$
          This implies
          $$A.u=f(p+ tu)-f(p)over t +o(1)qquad(tto0) ,$$
          hence
          $$|A.u|leq+o(1)leq L+o(1)qquad(tto0) .tag1$$
          As the LHS of $(1)$ does not depend on $t$ it follows that $|A.u|leq L$, and since $uin S^n-1$ was arbitrary we may conclude that $|A|leq L$.






          share|cite|improve this answer














          (i) There is something fishy here: You cannot say "$exists c$ such that $Phi$ holds" and then take the sup over all $c$ in $Phi$. Furthermore the statement is wrong if $O$ is not convex. As an example let $f$ be the principal value of the polar angle in the plane slit along the negative $x$-axis. Here $f(-1,epsilon)-f(-1,-epsilon)to2pi$ while $epsilonto0+$.



          Given $x$, $yin O$ you can argue as follows: If $z:=f(y)-f(x)ne0$ consider the auxiliary real-valued function
          $$phi(t)=zcdot fbigl((1-t)x+t ybigr)qquad(0leq tleq 1)$$
          $bigl($note that $(1-t)x+tyin O$ when $0leq tleq 1bigr)$. By the MVT there is a $tauin>]0,1[>$ such that
          $$zcdotbigl(f(y)-f(x)bigr)=phi(1)-phi(0)= phi'(tau)=zcdotbigl( f'bigl((1-tau)x+tau ybigr).(y-x)bigr) ,$$
          and therefore $|z|^2leq |z|sup_cin O|f'(c)|>|y-x|$, or
          $$bigr|f(y)-f(x)bigl|=|z|leq k>|y-x| .$$



          (ii) It's unclear what the directional derivative is for a vector valued function. We need a clear cut proof that the derivative $f'(p)=:Aincal L(mathbb R^n,mathbb R^m)$ at a given point $pin O$ has norm $leq L$, where $L$ is a global Lipschitz constant for $f$. To this end we go back to the definition of $f'(p)$. Let $u$ be an arbitrary unit vector. Then
          $$f(p+t u)- f(p)= A.t u + o(t)qquad(tto0) .$$
          This implies
          $$A.u=f(p+ tu)-f(p)over t +o(1)qquad(tto0) ,$$
          hence
          $$|A.u|leq+o(1)leq L+o(1)qquad(tto0) .tag1$$
          As the LHS of $(1)$ does not depend on $t$ it follows that $|A.u|leq L$, and since $uin S^n-1$ was arbitrary we may conclude that $|A|leq L$.







          share|cite|improve this answer














          share|cite|improve this answer



          share|cite|improve this answer








          edited Aug 22 at 13:53

























          answered Aug 22 at 8:41









          Christian Blatter

          165k7109310




          165k7109310











          • But $phi$ has to be $C^1$ for that to hold!
            – Mike
            Aug 22 at 9:01










          • O.k. See my edit. (I still have to see a naturally occurring multivariate function that is differentiable, but not continuously differentiable$ldots$)
            – Christian Blatter
            Aug 22 at 10:13










          • Thanks for your feedback Sir!
            – Mike
            Aug 22 at 13:37










          • Thanks for all but I think the first approach is right if I include the theorem I used. Mean Value Theorem for higher dimensions.
            – Mike
            Aug 22 at 14:33
















          • But $phi$ has to be $C^1$ for that to hold!
            – Mike
            Aug 22 at 9:01










          • O.k. See my edit. (I still have to see a naturally occurring multivariate function that is differentiable, but not continuously differentiable$ldots$)
            – Christian Blatter
            Aug 22 at 10:13










          • Thanks for your feedback Sir!
            – Mike
            Aug 22 at 13:37










          • Thanks for all but I think the first approach is right if I include the theorem I used. Mean Value Theorem for higher dimensions.
            – Mike
            Aug 22 at 14:33















          But $phi$ has to be $C^1$ for that to hold!
          – Mike
          Aug 22 at 9:01




          But $phi$ has to be $C^1$ for that to hold!
          – Mike
          Aug 22 at 9:01












          O.k. See my edit. (I still have to see a naturally occurring multivariate function that is differentiable, but not continuously differentiable$ldots$)
          – Christian Blatter
          Aug 22 at 10:13




          O.k. See my edit. (I still have to see a naturally occurring multivariate function that is differentiable, but not continuously differentiable$ldots$)
          – Christian Blatter
          Aug 22 at 10:13












          Thanks for your feedback Sir!
          – Mike
          Aug 22 at 13:37




          Thanks for your feedback Sir!
          – Mike
          Aug 22 at 13:37












          Thanks for all but I think the first approach is right if I include the theorem I used. Mean Value Theorem for higher dimensions.
          – Mike
          Aug 22 at 14:33




          Thanks for all but I think the first approach is right if I include the theorem I used. Mean Value Theorem for higher dimensions.
          – Mike
          Aug 22 at 14:33












           

          draft saved


          draft discarded


























           


          draft saved


          draft discarded














          StackExchange.ready(
          function ()
          StackExchange.openid.initPostLogin('.new-post-login', 'https%3a%2f%2fmath.stackexchange.com%2fquestions%2f2890725%2ff-is-bounded-and-sup-limits-c-in-o-vert-fc-vert-infty-if-and-only-i%23new-answer', 'question_page');

          );

          Post as a guest













































































          這個網誌中的熱門文章

          How to combine Bézier curves to a surface?

          Mutual Information Always Non-negative

          Why am i infinitely getting the same tweet with the Twitter Search API?