How to prove that $sum_k=1^nfrac1sqrt[n]k! sim fracnln n$

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











up vote
12
down vote

favorite
8












Recently I have met this asymptotic estimation:
prove that $sum_k=1^nfrac1sqrt[n]k! sim fracnln n$,$nto infty$.



Here is the way I think:



According to area principle,
beginalign*
Bigg|sum_k=1^nfrac1sqrt[n]k! -int_1^nfrac1sqrt[n]Gamma(x+1)d xBigg|leqslant frac1sqrt[n]Gamma(n+1)
endalign*
Thus we have
beginalign*
fracln nnBigg|sum_k=1^nfrac1sqrt[n]k! -int_1^nfrac1sqrt[n]Gamma(x+1)d xBigg|leqslant fracln nnfrac1sqrt[n]Gamma(n+1)sim fracln nncdot frac1sqrt[n]2pi n fracento 0,~~nto infty
endalign*
To this end we only need to show
beginalign*
int_1^nfrac1sqrt[n]Gamma(x+1)d xsim fracnln n
endalign*
But my train of thought is stuck here.Can you give me any help? Thanks a lot.







share|cite|improve this question






















  • Maybe you can use the stirling approximation again?
    – orion
    Aug 21 at 11:55










  • My computer shows that $fracln nnsum(sth)to0neq1$. Where did you find the result?
    – Aforest
    Aug 21 at 13:46










  • Is your question answered or is something still unclear? ;)
    – user90369
    Aug 27 at 8:26














up vote
12
down vote

favorite
8












Recently I have met this asymptotic estimation:
prove that $sum_k=1^nfrac1sqrt[n]k! sim fracnln n$,$nto infty$.



Here is the way I think:



According to area principle,
beginalign*
Bigg|sum_k=1^nfrac1sqrt[n]k! -int_1^nfrac1sqrt[n]Gamma(x+1)d xBigg|leqslant frac1sqrt[n]Gamma(n+1)
endalign*
Thus we have
beginalign*
fracln nnBigg|sum_k=1^nfrac1sqrt[n]k! -int_1^nfrac1sqrt[n]Gamma(x+1)d xBigg|leqslant fracln nnfrac1sqrt[n]Gamma(n+1)sim fracln nncdot frac1sqrt[n]2pi n fracento 0,~~nto infty
endalign*
To this end we only need to show
beginalign*
int_1^nfrac1sqrt[n]Gamma(x+1)d xsim fracnln n
endalign*
But my train of thought is stuck here.Can you give me any help? Thanks a lot.







share|cite|improve this question






















  • Maybe you can use the stirling approximation again?
    – orion
    Aug 21 at 11:55










  • My computer shows that $fracln nnsum(sth)to0neq1$. Where did you find the result?
    – Aforest
    Aug 21 at 13:46










  • Is your question answered or is something still unclear? ;)
    – user90369
    Aug 27 at 8:26












up vote
12
down vote

favorite
8









up vote
12
down vote

favorite
8






8





Recently I have met this asymptotic estimation:
prove that $sum_k=1^nfrac1sqrt[n]k! sim fracnln n$,$nto infty$.



Here is the way I think:



According to area principle,
beginalign*
Bigg|sum_k=1^nfrac1sqrt[n]k! -int_1^nfrac1sqrt[n]Gamma(x+1)d xBigg|leqslant frac1sqrt[n]Gamma(n+1)
endalign*
Thus we have
beginalign*
fracln nnBigg|sum_k=1^nfrac1sqrt[n]k! -int_1^nfrac1sqrt[n]Gamma(x+1)d xBigg|leqslant fracln nnfrac1sqrt[n]Gamma(n+1)sim fracln nncdot frac1sqrt[n]2pi n fracento 0,~~nto infty
endalign*
To this end we only need to show
beginalign*
int_1^nfrac1sqrt[n]Gamma(x+1)d xsim fracnln n
endalign*
But my train of thought is stuck here.Can you give me any help? Thanks a lot.







share|cite|improve this question














Recently I have met this asymptotic estimation:
prove that $sum_k=1^nfrac1sqrt[n]k! sim fracnln n$,$nto infty$.



Here is the way I think:



According to area principle,
beginalign*
Bigg|sum_k=1^nfrac1sqrt[n]k! -int_1^nfrac1sqrt[n]Gamma(x+1)d xBigg|leqslant frac1sqrt[n]Gamma(n+1)
endalign*
Thus we have
beginalign*
fracln nnBigg|sum_k=1^nfrac1sqrt[n]k! -int_1^nfrac1sqrt[n]Gamma(x+1)d xBigg|leqslant fracln nnfrac1sqrt[n]Gamma(n+1)sim fracln nncdot frac1sqrt[n]2pi n fracento 0,~~nto infty
endalign*
To this end we only need to show
beginalign*
int_1^nfrac1sqrt[n]Gamma(x+1)d xsim fracnln n
endalign*
But my train of thought is stuck here.Can you give me any help? Thanks a lot.









share|cite|improve this question













share|cite|improve this question




share|cite|improve this question








edited Aug 21 at 21:51

























asked Aug 21 at 11:37









mbfkk

324112




324112











  • Maybe you can use the stirling approximation again?
    – orion
    Aug 21 at 11:55










  • My computer shows that $fracln nnsum(sth)to0neq1$. Where did you find the result?
    – Aforest
    Aug 21 at 13:46










  • Is your question answered or is something still unclear? ;)
    – user90369
    Aug 27 at 8:26
















  • Maybe you can use the stirling approximation again?
    – orion
    Aug 21 at 11:55










  • My computer shows that $fracln nnsum(sth)to0neq1$. Where did you find the result?
    – Aforest
    Aug 21 at 13:46










  • Is your question answered or is something still unclear? ;)
    – user90369
    Aug 27 at 8:26















Maybe you can use the stirling approximation again?
– orion
Aug 21 at 11:55




Maybe you can use the stirling approximation again?
– orion
Aug 21 at 11:55












My computer shows that $fracln nnsum(sth)to0neq1$. Where did you find the result?
– Aforest
Aug 21 at 13:46




My computer shows that $fracln nnsum(sth)to0neq1$. Where did you find the result?
– Aforest
Aug 21 at 13:46












Is your question answered or is something still unclear? ;)
– user90369
Aug 27 at 8:26




Is your question answered or is something still unclear? ;)
– user90369
Aug 27 at 8:26










3 Answers
3






active

oldest

votes

















up vote
9
down vote













Using the stirling formula and partial integration we get:



$displaystylefracln nnsumlimits_k=1^n frac1k!^1/n sim fracln nnintlimits_1^nfracdxGamma(1+x)^1/n = (ln n)intlimits_1/n^1fracdxGamma(1+nx)^1/n
sim (ln n)intlimits_1/n^1fracdx(nx/e)^x$



$displaystyle = left(fracln n(x/e)^xfracn^-x-ln nright)bigg|_1/n^1 + intlimits_1/n^1fracln frac1x(nx/e)^xdx enspace$ with
$enspacedisplaystyle left(fracln n(x/e)^xfracn^-x-ln nright)bigg|_1/n^1 sim 1$



$displaystyle 0< intlimits_1/n^1fracln frac1x(nx/e)^xdx < eintlimits_0^1fracln frac1xn^xdx =fraceln nleft(gamma+lnln n+intlimits_ln n^inftyfracdtte^tright)simfracelnln nln nsim 0$



with the Euler-Mascheroni constant $,gamma,$



It follows $enspacedisplaystylefracln nnsumlimits_k=1^n frac1k!^1/n sim 1enspace$ and therefore the claim.



Note:



$displaystyle intlimits_0^1fracln frac1xe^axdx=fracln a + gamma + Gamma(0,a)asim fracln aaenspace$ with the incomplete Gamma function $,Gamma(.,.)$






share|cite|improve this answer






















  • How can you get 0 after partial integration?I have tried,the part is$nint_frac1n^1fracn^-x-ln ncdot Big(fracexBig)^xln frac1xdx$,and since $1 leqslant Big(fracexBig)^x leqslant e$,so if that part tend to 0,we must have $int_frac1n^1fracnln n n^-xln frac1xdxto 0$,$nto infty$,but it's impossible ,for in fact we can see $int_frac1n^1fracnln n n^-xln frac1xdxto +infty$!
    – mbfkk
    Aug 23 at 13:09











  • @mbfkk : In my first version of my answer I simply forgot $n$ . Hope it's better now. :) I'm sorry about the inconvenience and misunderstandings. I've done the calculation of the last integral of my answer with the help of WolframAlpha. It was necessary to show, that the divergence of the last integral of your comment above is irrelevant regarding $n/ln n$ .
    – user90369
    Aug 25 at 12:54











  • Nice done!Thanks!
    – mbfkk
    Aug 25 at 23:15










  • @mbfkk: You are welcome. :)
    – user90369
    Aug 25 at 23:34

















up vote
7
down vote













For any $x>0$, $$c_1,x^x+frac12e^-xleq Gamma(x+1)leq c_2,x^x+frac12e^-x$$ for some $c_1,c_2>0$ (see for example the Wikipedia page). Therefore, $$frac1sqrt[n]c_2int_1^nx^-x/nx^-frac12ne^x/n,dxleq int_1^nfrac1sqrt[n]Gamma(x+1),dxleqfrac1sqrt[n]c_1int_1^nx^-x/nx^-frac12ne^x/n,dx.$$ Next, note that $$n^-frac12nint_1^nx^-x/ne^x/n,dxleqint_1^nx^-x/nx^-frac12ne^x/n,dxleq int_1^nx^-x/ne^x/n,dx,$$ so it suffices to find the asymptotics of the last integral.






share|cite|improve this answer
















  • 1




    ... which can be done through Laplace method. Nice approach, (+1).
    – Jack D'Aurizio♦
    Aug 21 at 15:57










  • Thanks for your reply! I'm still puzzled at how to get the asymptotics of $int_1^nx^-x/ne^x/n,dx$.Can you give any clearer hint?
    – mbfkk
    Aug 23 at 13:16

















up vote
3
down vote













The terms $frac1sqrt[n]k!$, for $kin[1,n]$, are roughly of the same size, hence the upper bound provided by Holder's inequality
$$ sum_k=1^nfrac1sqrt[n]k!leq sqrt[n]n^n-1sum_k=1^nfrac1k!leq fracnsqrt[n]n/e$$
is expected to be close to the asymptotic behaviour of the RHS.
It can be improved via (I am going to outline the case $n=4$ for simplicity)
$$sum_k=1^4frac1sqrt[4]k!leq sqrt[4]left(tfrac11+tfrac12+tfrac13+tfrac14right)left(tfrac11+tfrac11+tfrac12+tfrac13right)left(tfrac11+tfrac11+tfrac11+tfrac12right)left(tfrac11+tfrac11+tfrac11+tfrac11right)$$
leading to
$$ sum_k=1^nfrac1sqrt[n]k!leq sqrt[n]prod_h=0^n-1left(h+H_n-hright).tagU$$
On the other hand Holder's inequality can be used also for producing a lower bound:



$$ left(sum_k=1^nfrac11right)left(sum_k=1^nfrac1sqrt[n]k!right)^n geq left(sum_k=1^nfrac1sqrt[n+1]k!right)^n+1$$
$$ left(sum_k=1^nfrac11right)left(sum_k=1^nfrac1sqrt[n+1]k!right)^n+1 geq left(sum_k=1^nfrac1sqrt[n+2]k!right)^n+2$$



lead to
$$ sum_k=1^nfrac1sqrt[n]k!geq frac1n^1/nleft(sum_k=1^nfrac1sqrt[n+1]k!right)^fracn+1ngeq frac1n^2/nleft(sum_k=1^nfrac1sqrt[n+2]k!right)^fracn+2ngeqldotsgeqfrac1nleft(sum_k=1^nfrac1sqrt[2n]k!right)^2,$$



$$ sum_k=1^nfrac1sqrt[n]k!geqfrac1n^nleft(sum_k=1^nfrac1sqrt[n^2]k!right)^n+1tagL$$
and for any $kin[1,n]$ the distance between $sqrt[2n]k!$ and $1$ is $Oleft(fraclog nnright)$ by Stirling's approximation.
The asymptotic behaviour of your sum should now follow by comparing the accurate upper bound $(U)$ and the accurate lower bound $(L)$.






share|cite|improve this answer




















    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%2f2889768%2fhow-to-prove-that-sum-k-1n-frac1-sqrtnk-sim-fracn-ln-n%23new-answer', 'question_page');

    );

    Post as a guest






























    3 Answers
    3






    active

    oldest

    votes








    3 Answers
    3






    active

    oldest

    votes









    active

    oldest

    votes






    active

    oldest

    votes








    up vote
    9
    down vote













    Using the stirling formula and partial integration we get:



    $displaystylefracln nnsumlimits_k=1^n frac1k!^1/n sim fracln nnintlimits_1^nfracdxGamma(1+x)^1/n = (ln n)intlimits_1/n^1fracdxGamma(1+nx)^1/n
    sim (ln n)intlimits_1/n^1fracdx(nx/e)^x$



    $displaystyle = left(fracln n(x/e)^xfracn^-x-ln nright)bigg|_1/n^1 + intlimits_1/n^1fracln frac1x(nx/e)^xdx enspace$ with
    $enspacedisplaystyle left(fracln n(x/e)^xfracn^-x-ln nright)bigg|_1/n^1 sim 1$



    $displaystyle 0< intlimits_1/n^1fracln frac1x(nx/e)^xdx < eintlimits_0^1fracln frac1xn^xdx =fraceln nleft(gamma+lnln n+intlimits_ln n^inftyfracdtte^tright)simfracelnln nln nsim 0$



    with the Euler-Mascheroni constant $,gamma,$



    It follows $enspacedisplaystylefracln nnsumlimits_k=1^n frac1k!^1/n sim 1enspace$ and therefore the claim.



    Note:



    $displaystyle intlimits_0^1fracln frac1xe^axdx=fracln a + gamma + Gamma(0,a)asim fracln aaenspace$ with the incomplete Gamma function $,Gamma(.,.)$






    share|cite|improve this answer






















    • How can you get 0 after partial integration?I have tried,the part is$nint_frac1n^1fracn^-x-ln ncdot Big(fracexBig)^xln frac1xdx$,and since $1 leqslant Big(fracexBig)^x leqslant e$,so if that part tend to 0,we must have $int_frac1n^1fracnln n n^-xln frac1xdxto 0$,$nto infty$,but it's impossible ,for in fact we can see $int_frac1n^1fracnln n n^-xln frac1xdxto +infty$!
      – mbfkk
      Aug 23 at 13:09











    • @mbfkk : In my first version of my answer I simply forgot $n$ . Hope it's better now. :) I'm sorry about the inconvenience and misunderstandings. I've done the calculation of the last integral of my answer with the help of WolframAlpha. It was necessary to show, that the divergence of the last integral of your comment above is irrelevant regarding $n/ln n$ .
      – user90369
      Aug 25 at 12:54











    • Nice done!Thanks!
      – mbfkk
      Aug 25 at 23:15










    • @mbfkk: You are welcome. :)
      – user90369
      Aug 25 at 23:34














    up vote
    9
    down vote













    Using the stirling formula and partial integration we get:



    $displaystylefracln nnsumlimits_k=1^n frac1k!^1/n sim fracln nnintlimits_1^nfracdxGamma(1+x)^1/n = (ln n)intlimits_1/n^1fracdxGamma(1+nx)^1/n
    sim (ln n)intlimits_1/n^1fracdx(nx/e)^x$



    $displaystyle = left(fracln n(x/e)^xfracn^-x-ln nright)bigg|_1/n^1 + intlimits_1/n^1fracln frac1x(nx/e)^xdx enspace$ with
    $enspacedisplaystyle left(fracln n(x/e)^xfracn^-x-ln nright)bigg|_1/n^1 sim 1$



    $displaystyle 0< intlimits_1/n^1fracln frac1x(nx/e)^xdx < eintlimits_0^1fracln frac1xn^xdx =fraceln nleft(gamma+lnln n+intlimits_ln n^inftyfracdtte^tright)simfracelnln nln nsim 0$



    with the Euler-Mascheroni constant $,gamma,$



    It follows $enspacedisplaystylefracln nnsumlimits_k=1^n frac1k!^1/n sim 1enspace$ and therefore the claim.



    Note:



    $displaystyle intlimits_0^1fracln frac1xe^axdx=fracln a + gamma + Gamma(0,a)asim fracln aaenspace$ with the incomplete Gamma function $,Gamma(.,.)$






    share|cite|improve this answer






















    • How can you get 0 after partial integration?I have tried,the part is$nint_frac1n^1fracn^-x-ln ncdot Big(fracexBig)^xln frac1xdx$,and since $1 leqslant Big(fracexBig)^x leqslant e$,so if that part tend to 0,we must have $int_frac1n^1fracnln n n^-xln frac1xdxto 0$,$nto infty$,but it's impossible ,for in fact we can see $int_frac1n^1fracnln n n^-xln frac1xdxto +infty$!
      – mbfkk
      Aug 23 at 13:09











    • @mbfkk : In my first version of my answer I simply forgot $n$ . Hope it's better now. :) I'm sorry about the inconvenience and misunderstandings. I've done the calculation of the last integral of my answer with the help of WolframAlpha. It was necessary to show, that the divergence of the last integral of your comment above is irrelevant regarding $n/ln n$ .
      – user90369
      Aug 25 at 12:54











    • Nice done!Thanks!
      – mbfkk
      Aug 25 at 23:15










    • @mbfkk: You are welcome. :)
      – user90369
      Aug 25 at 23:34












    up vote
    9
    down vote










    up vote
    9
    down vote









    Using the stirling formula and partial integration we get:



    $displaystylefracln nnsumlimits_k=1^n frac1k!^1/n sim fracln nnintlimits_1^nfracdxGamma(1+x)^1/n = (ln n)intlimits_1/n^1fracdxGamma(1+nx)^1/n
    sim (ln n)intlimits_1/n^1fracdx(nx/e)^x$



    $displaystyle = left(fracln n(x/e)^xfracn^-x-ln nright)bigg|_1/n^1 + intlimits_1/n^1fracln frac1x(nx/e)^xdx enspace$ with
    $enspacedisplaystyle left(fracln n(x/e)^xfracn^-x-ln nright)bigg|_1/n^1 sim 1$



    $displaystyle 0< intlimits_1/n^1fracln frac1x(nx/e)^xdx < eintlimits_0^1fracln frac1xn^xdx =fraceln nleft(gamma+lnln n+intlimits_ln n^inftyfracdtte^tright)simfracelnln nln nsim 0$



    with the Euler-Mascheroni constant $,gamma,$



    It follows $enspacedisplaystylefracln nnsumlimits_k=1^n frac1k!^1/n sim 1enspace$ and therefore the claim.



    Note:



    $displaystyle intlimits_0^1fracln frac1xe^axdx=fracln a + gamma + Gamma(0,a)asim fracln aaenspace$ with the incomplete Gamma function $,Gamma(.,.)$






    share|cite|improve this answer














    Using the stirling formula and partial integration we get:



    $displaystylefracln nnsumlimits_k=1^n frac1k!^1/n sim fracln nnintlimits_1^nfracdxGamma(1+x)^1/n = (ln n)intlimits_1/n^1fracdxGamma(1+nx)^1/n
    sim (ln n)intlimits_1/n^1fracdx(nx/e)^x$



    $displaystyle = left(fracln n(x/e)^xfracn^-x-ln nright)bigg|_1/n^1 + intlimits_1/n^1fracln frac1x(nx/e)^xdx enspace$ with
    $enspacedisplaystyle left(fracln n(x/e)^xfracn^-x-ln nright)bigg|_1/n^1 sim 1$



    $displaystyle 0< intlimits_1/n^1fracln frac1x(nx/e)^xdx < eintlimits_0^1fracln frac1xn^xdx =fraceln nleft(gamma+lnln n+intlimits_ln n^inftyfracdtte^tright)simfracelnln nln nsim 0$



    with the Euler-Mascheroni constant $,gamma,$



    It follows $enspacedisplaystylefracln nnsumlimits_k=1^n frac1k!^1/n sim 1enspace$ and therefore the claim.



    Note:



    $displaystyle intlimits_0^1fracln frac1xe^axdx=fracln a + gamma + Gamma(0,a)asim fracln aaenspace$ with the incomplete Gamma function $,Gamma(.,.)$







    share|cite|improve this answer














    share|cite|improve this answer



    share|cite|improve this answer








    edited Aug 25 at 19:52

























    answered Aug 21 at 16:11









    user90369

    7,891925




    7,891925











    • How can you get 0 after partial integration?I have tried,the part is$nint_frac1n^1fracn^-x-ln ncdot Big(fracexBig)^xln frac1xdx$,and since $1 leqslant Big(fracexBig)^x leqslant e$,so if that part tend to 0,we must have $int_frac1n^1fracnln n n^-xln frac1xdxto 0$,$nto infty$,but it's impossible ,for in fact we can see $int_frac1n^1fracnln n n^-xln frac1xdxto +infty$!
      – mbfkk
      Aug 23 at 13:09











    • @mbfkk : In my first version of my answer I simply forgot $n$ . Hope it's better now. :) I'm sorry about the inconvenience and misunderstandings. I've done the calculation of the last integral of my answer with the help of WolframAlpha. It was necessary to show, that the divergence of the last integral of your comment above is irrelevant regarding $n/ln n$ .
      – user90369
      Aug 25 at 12:54











    • Nice done!Thanks!
      – mbfkk
      Aug 25 at 23:15










    • @mbfkk: You are welcome. :)
      – user90369
      Aug 25 at 23:34
















    • How can you get 0 after partial integration?I have tried,the part is$nint_frac1n^1fracn^-x-ln ncdot Big(fracexBig)^xln frac1xdx$,and since $1 leqslant Big(fracexBig)^x leqslant e$,so if that part tend to 0,we must have $int_frac1n^1fracnln n n^-xln frac1xdxto 0$,$nto infty$,but it's impossible ,for in fact we can see $int_frac1n^1fracnln n n^-xln frac1xdxto +infty$!
      – mbfkk
      Aug 23 at 13:09











    • @mbfkk : In my first version of my answer I simply forgot $n$ . Hope it's better now. :) I'm sorry about the inconvenience and misunderstandings. I've done the calculation of the last integral of my answer with the help of WolframAlpha. It was necessary to show, that the divergence of the last integral of your comment above is irrelevant regarding $n/ln n$ .
      – user90369
      Aug 25 at 12:54











    • Nice done!Thanks!
      – mbfkk
      Aug 25 at 23:15










    • @mbfkk: You are welcome. :)
      – user90369
      Aug 25 at 23:34















    How can you get 0 after partial integration?I have tried,the part is$nint_frac1n^1fracn^-x-ln ncdot Big(fracexBig)^xln frac1xdx$,and since $1 leqslant Big(fracexBig)^x leqslant e$,so if that part tend to 0,we must have $int_frac1n^1fracnln n n^-xln frac1xdxto 0$,$nto infty$,but it's impossible ,for in fact we can see $int_frac1n^1fracnln n n^-xln frac1xdxto +infty$!
    – mbfkk
    Aug 23 at 13:09





    How can you get 0 after partial integration?I have tried,the part is$nint_frac1n^1fracn^-x-ln ncdot Big(fracexBig)^xln frac1xdx$,and since $1 leqslant Big(fracexBig)^x leqslant e$,so if that part tend to 0,we must have $int_frac1n^1fracnln n n^-xln frac1xdxto 0$,$nto infty$,but it's impossible ,for in fact we can see $int_frac1n^1fracnln n n^-xln frac1xdxto +infty$!
    – mbfkk
    Aug 23 at 13:09













    @mbfkk : In my first version of my answer I simply forgot $n$ . Hope it's better now. :) I'm sorry about the inconvenience and misunderstandings. I've done the calculation of the last integral of my answer with the help of WolframAlpha. It was necessary to show, that the divergence of the last integral of your comment above is irrelevant regarding $n/ln n$ .
    – user90369
    Aug 25 at 12:54





    @mbfkk : In my first version of my answer I simply forgot $n$ . Hope it's better now. :) I'm sorry about the inconvenience and misunderstandings. I've done the calculation of the last integral of my answer with the help of WolframAlpha. It was necessary to show, that the divergence of the last integral of your comment above is irrelevant regarding $n/ln n$ .
    – user90369
    Aug 25 at 12:54













    Nice done!Thanks!
    – mbfkk
    Aug 25 at 23:15




    Nice done!Thanks!
    – mbfkk
    Aug 25 at 23:15












    @mbfkk: You are welcome. :)
    – user90369
    Aug 25 at 23:34




    @mbfkk: You are welcome. :)
    – user90369
    Aug 25 at 23:34










    up vote
    7
    down vote













    For any $x>0$, $$c_1,x^x+frac12e^-xleq Gamma(x+1)leq c_2,x^x+frac12e^-x$$ for some $c_1,c_2>0$ (see for example the Wikipedia page). Therefore, $$frac1sqrt[n]c_2int_1^nx^-x/nx^-frac12ne^x/n,dxleq int_1^nfrac1sqrt[n]Gamma(x+1),dxleqfrac1sqrt[n]c_1int_1^nx^-x/nx^-frac12ne^x/n,dx.$$ Next, note that $$n^-frac12nint_1^nx^-x/ne^x/n,dxleqint_1^nx^-x/nx^-frac12ne^x/n,dxleq int_1^nx^-x/ne^x/n,dx,$$ so it suffices to find the asymptotics of the last integral.






    share|cite|improve this answer
















    • 1




      ... which can be done through Laplace method. Nice approach, (+1).
      – Jack D'Aurizio♦
      Aug 21 at 15:57










    • Thanks for your reply! I'm still puzzled at how to get the asymptotics of $int_1^nx^-x/ne^x/n,dx$.Can you give any clearer hint?
      – mbfkk
      Aug 23 at 13:16














    up vote
    7
    down vote













    For any $x>0$, $$c_1,x^x+frac12e^-xleq Gamma(x+1)leq c_2,x^x+frac12e^-x$$ for some $c_1,c_2>0$ (see for example the Wikipedia page). Therefore, $$frac1sqrt[n]c_2int_1^nx^-x/nx^-frac12ne^x/n,dxleq int_1^nfrac1sqrt[n]Gamma(x+1),dxleqfrac1sqrt[n]c_1int_1^nx^-x/nx^-frac12ne^x/n,dx.$$ Next, note that $$n^-frac12nint_1^nx^-x/ne^x/n,dxleqint_1^nx^-x/nx^-frac12ne^x/n,dxleq int_1^nx^-x/ne^x/n,dx,$$ so it suffices to find the asymptotics of the last integral.






    share|cite|improve this answer
















    • 1




      ... which can be done through Laplace method. Nice approach, (+1).
      – Jack D'Aurizio♦
      Aug 21 at 15:57










    • Thanks for your reply! I'm still puzzled at how to get the asymptotics of $int_1^nx^-x/ne^x/n,dx$.Can you give any clearer hint?
      – mbfkk
      Aug 23 at 13:16












    up vote
    7
    down vote










    up vote
    7
    down vote









    For any $x>0$, $$c_1,x^x+frac12e^-xleq Gamma(x+1)leq c_2,x^x+frac12e^-x$$ for some $c_1,c_2>0$ (see for example the Wikipedia page). Therefore, $$frac1sqrt[n]c_2int_1^nx^-x/nx^-frac12ne^x/n,dxleq int_1^nfrac1sqrt[n]Gamma(x+1),dxleqfrac1sqrt[n]c_1int_1^nx^-x/nx^-frac12ne^x/n,dx.$$ Next, note that $$n^-frac12nint_1^nx^-x/ne^x/n,dxleqint_1^nx^-x/nx^-frac12ne^x/n,dxleq int_1^nx^-x/ne^x/n,dx,$$ so it suffices to find the asymptotics of the last integral.






    share|cite|improve this answer












    For any $x>0$, $$c_1,x^x+frac12e^-xleq Gamma(x+1)leq c_2,x^x+frac12e^-x$$ for some $c_1,c_2>0$ (see for example the Wikipedia page). Therefore, $$frac1sqrt[n]c_2int_1^nx^-x/nx^-frac12ne^x/n,dxleq int_1^nfrac1sqrt[n]Gamma(x+1),dxleqfrac1sqrt[n]c_1int_1^nx^-x/nx^-frac12ne^x/n,dx.$$ Next, note that $$n^-frac12nint_1^nx^-x/ne^x/n,dxleqint_1^nx^-x/nx^-frac12ne^x/n,dxleq int_1^nx^-x/ne^x/n,dx,$$ so it suffices to find the asymptotics of the last integral.







    share|cite|improve this answer












    share|cite|improve this answer



    share|cite|improve this answer










    answered Aug 21 at 14:54









    detnvvp

    6,7441018




    6,7441018







    • 1




      ... which can be done through Laplace method. Nice approach, (+1).
      – Jack D'Aurizio♦
      Aug 21 at 15:57










    • Thanks for your reply! I'm still puzzled at how to get the asymptotics of $int_1^nx^-x/ne^x/n,dx$.Can you give any clearer hint?
      – mbfkk
      Aug 23 at 13:16












    • 1




      ... which can be done through Laplace method. Nice approach, (+1).
      – Jack D'Aurizio♦
      Aug 21 at 15:57










    • Thanks for your reply! I'm still puzzled at how to get the asymptotics of $int_1^nx^-x/ne^x/n,dx$.Can you give any clearer hint?
      – mbfkk
      Aug 23 at 13:16







    1




    1




    ... which can be done through Laplace method. Nice approach, (+1).
    – Jack D'Aurizio♦
    Aug 21 at 15:57




    ... which can be done through Laplace method. Nice approach, (+1).
    – Jack D'Aurizio♦
    Aug 21 at 15:57












    Thanks for your reply! I'm still puzzled at how to get the asymptotics of $int_1^nx^-x/ne^x/n,dx$.Can you give any clearer hint?
    – mbfkk
    Aug 23 at 13:16




    Thanks for your reply! I'm still puzzled at how to get the asymptotics of $int_1^nx^-x/ne^x/n,dx$.Can you give any clearer hint?
    – mbfkk
    Aug 23 at 13:16










    up vote
    3
    down vote













    The terms $frac1sqrt[n]k!$, for $kin[1,n]$, are roughly of the same size, hence the upper bound provided by Holder's inequality
    $$ sum_k=1^nfrac1sqrt[n]k!leq sqrt[n]n^n-1sum_k=1^nfrac1k!leq fracnsqrt[n]n/e$$
    is expected to be close to the asymptotic behaviour of the RHS.
    It can be improved via (I am going to outline the case $n=4$ for simplicity)
    $$sum_k=1^4frac1sqrt[4]k!leq sqrt[4]left(tfrac11+tfrac12+tfrac13+tfrac14right)left(tfrac11+tfrac11+tfrac12+tfrac13right)left(tfrac11+tfrac11+tfrac11+tfrac12right)left(tfrac11+tfrac11+tfrac11+tfrac11right)$$
    leading to
    $$ sum_k=1^nfrac1sqrt[n]k!leq sqrt[n]prod_h=0^n-1left(h+H_n-hright).tagU$$
    On the other hand Holder's inequality can be used also for producing a lower bound:



    $$ left(sum_k=1^nfrac11right)left(sum_k=1^nfrac1sqrt[n]k!right)^n geq left(sum_k=1^nfrac1sqrt[n+1]k!right)^n+1$$
    $$ left(sum_k=1^nfrac11right)left(sum_k=1^nfrac1sqrt[n+1]k!right)^n+1 geq left(sum_k=1^nfrac1sqrt[n+2]k!right)^n+2$$



    lead to
    $$ sum_k=1^nfrac1sqrt[n]k!geq frac1n^1/nleft(sum_k=1^nfrac1sqrt[n+1]k!right)^fracn+1ngeq frac1n^2/nleft(sum_k=1^nfrac1sqrt[n+2]k!right)^fracn+2ngeqldotsgeqfrac1nleft(sum_k=1^nfrac1sqrt[2n]k!right)^2,$$



    $$ sum_k=1^nfrac1sqrt[n]k!geqfrac1n^nleft(sum_k=1^nfrac1sqrt[n^2]k!right)^n+1tagL$$
    and for any $kin[1,n]$ the distance between $sqrt[2n]k!$ and $1$ is $Oleft(fraclog nnright)$ by Stirling's approximation.
    The asymptotic behaviour of your sum should now follow by comparing the accurate upper bound $(U)$ and the accurate lower bound $(L)$.






    share|cite|improve this answer
























      up vote
      3
      down vote













      The terms $frac1sqrt[n]k!$, for $kin[1,n]$, are roughly of the same size, hence the upper bound provided by Holder's inequality
      $$ sum_k=1^nfrac1sqrt[n]k!leq sqrt[n]n^n-1sum_k=1^nfrac1k!leq fracnsqrt[n]n/e$$
      is expected to be close to the asymptotic behaviour of the RHS.
      It can be improved via (I am going to outline the case $n=4$ for simplicity)
      $$sum_k=1^4frac1sqrt[4]k!leq sqrt[4]left(tfrac11+tfrac12+tfrac13+tfrac14right)left(tfrac11+tfrac11+tfrac12+tfrac13right)left(tfrac11+tfrac11+tfrac11+tfrac12right)left(tfrac11+tfrac11+tfrac11+tfrac11right)$$
      leading to
      $$ sum_k=1^nfrac1sqrt[n]k!leq sqrt[n]prod_h=0^n-1left(h+H_n-hright).tagU$$
      On the other hand Holder's inequality can be used also for producing a lower bound:



      $$ left(sum_k=1^nfrac11right)left(sum_k=1^nfrac1sqrt[n]k!right)^n geq left(sum_k=1^nfrac1sqrt[n+1]k!right)^n+1$$
      $$ left(sum_k=1^nfrac11right)left(sum_k=1^nfrac1sqrt[n+1]k!right)^n+1 geq left(sum_k=1^nfrac1sqrt[n+2]k!right)^n+2$$



      lead to
      $$ sum_k=1^nfrac1sqrt[n]k!geq frac1n^1/nleft(sum_k=1^nfrac1sqrt[n+1]k!right)^fracn+1ngeq frac1n^2/nleft(sum_k=1^nfrac1sqrt[n+2]k!right)^fracn+2ngeqldotsgeqfrac1nleft(sum_k=1^nfrac1sqrt[2n]k!right)^2,$$



      $$ sum_k=1^nfrac1sqrt[n]k!geqfrac1n^nleft(sum_k=1^nfrac1sqrt[n^2]k!right)^n+1tagL$$
      and for any $kin[1,n]$ the distance between $sqrt[2n]k!$ and $1$ is $Oleft(fraclog nnright)$ by Stirling's approximation.
      The asymptotic behaviour of your sum should now follow by comparing the accurate upper bound $(U)$ and the accurate lower bound $(L)$.






      share|cite|improve this answer






















        up vote
        3
        down vote










        up vote
        3
        down vote









        The terms $frac1sqrt[n]k!$, for $kin[1,n]$, are roughly of the same size, hence the upper bound provided by Holder's inequality
        $$ sum_k=1^nfrac1sqrt[n]k!leq sqrt[n]n^n-1sum_k=1^nfrac1k!leq fracnsqrt[n]n/e$$
        is expected to be close to the asymptotic behaviour of the RHS.
        It can be improved via (I am going to outline the case $n=4$ for simplicity)
        $$sum_k=1^4frac1sqrt[4]k!leq sqrt[4]left(tfrac11+tfrac12+tfrac13+tfrac14right)left(tfrac11+tfrac11+tfrac12+tfrac13right)left(tfrac11+tfrac11+tfrac11+tfrac12right)left(tfrac11+tfrac11+tfrac11+tfrac11right)$$
        leading to
        $$ sum_k=1^nfrac1sqrt[n]k!leq sqrt[n]prod_h=0^n-1left(h+H_n-hright).tagU$$
        On the other hand Holder's inequality can be used also for producing a lower bound:



        $$ left(sum_k=1^nfrac11right)left(sum_k=1^nfrac1sqrt[n]k!right)^n geq left(sum_k=1^nfrac1sqrt[n+1]k!right)^n+1$$
        $$ left(sum_k=1^nfrac11right)left(sum_k=1^nfrac1sqrt[n+1]k!right)^n+1 geq left(sum_k=1^nfrac1sqrt[n+2]k!right)^n+2$$



        lead to
        $$ sum_k=1^nfrac1sqrt[n]k!geq frac1n^1/nleft(sum_k=1^nfrac1sqrt[n+1]k!right)^fracn+1ngeq frac1n^2/nleft(sum_k=1^nfrac1sqrt[n+2]k!right)^fracn+2ngeqldotsgeqfrac1nleft(sum_k=1^nfrac1sqrt[2n]k!right)^2,$$



        $$ sum_k=1^nfrac1sqrt[n]k!geqfrac1n^nleft(sum_k=1^nfrac1sqrt[n^2]k!right)^n+1tagL$$
        and for any $kin[1,n]$ the distance between $sqrt[2n]k!$ and $1$ is $Oleft(fraclog nnright)$ by Stirling's approximation.
        The asymptotic behaviour of your sum should now follow by comparing the accurate upper bound $(U)$ and the accurate lower bound $(L)$.






        share|cite|improve this answer












        The terms $frac1sqrt[n]k!$, for $kin[1,n]$, are roughly of the same size, hence the upper bound provided by Holder's inequality
        $$ sum_k=1^nfrac1sqrt[n]k!leq sqrt[n]n^n-1sum_k=1^nfrac1k!leq fracnsqrt[n]n/e$$
        is expected to be close to the asymptotic behaviour of the RHS.
        It can be improved via (I am going to outline the case $n=4$ for simplicity)
        $$sum_k=1^4frac1sqrt[4]k!leq sqrt[4]left(tfrac11+tfrac12+tfrac13+tfrac14right)left(tfrac11+tfrac11+tfrac12+tfrac13right)left(tfrac11+tfrac11+tfrac11+tfrac12right)left(tfrac11+tfrac11+tfrac11+tfrac11right)$$
        leading to
        $$ sum_k=1^nfrac1sqrt[n]k!leq sqrt[n]prod_h=0^n-1left(h+H_n-hright).tagU$$
        On the other hand Holder's inequality can be used also for producing a lower bound:



        $$ left(sum_k=1^nfrac11right)left(sum_k=1^nfrac1sqrt[n]k!right)^n geq left(sum_k=1^nfrac1sqrt[n+1]k!right)^n+1$$
        $$ left(sum_k=1^nfrac11right)left(sum_k=1^nfrac1sqrt[n+1]k!right)^n+1 geq left(sum_k=1^nfrac1sqrt[n+2]k!right)^n+2$$



        lead to
        $$ sum_k=1^nfrac1sqrt[n]k!geq frac1n^1/nleft(sum_k=1^nfrac1sqrt[n+1]k!right)^fracn+1ngeq frac1n^2/nleft(sum_k=1^nfrac1sqrt[n+2]k!right)^fracn+2ngeqldotsgeqfrac1nleft(sum_k=1^nfrac1sqrt[2n]k!right)^2,$$



        $$ sum_k=1^nfrac1sqrt[n]k!geqfrac1n^nleft(sum_k=1^nfrac1sqrt[n^2]k!right)^n+1tagL$$
        and for any $kin[1,n]$ the distance between $sqrt[2n]k!$ and $1$ is $Oleft(fraclog nnright)$ by Stirling's approximation.
        The asymptotic behaviour of your sum should now follow by comparing the accurate upper bound $(U)$ and the accurate lower bound $(L)$.







        share|cite|improve this answer












        share|cite|improve this answer



        share|cite|improve this answer










        answered Aug 21 at 14:38









        Jack D'Aurizio♦

        272k32267634




        272k32267634






















             

            draft saved


            draft discarded


























             


            draft saved


            draft discarded














            StackExchange.ready(
            function ()
            StackExchange.openid.initPostLogin('.new-post-login', 'https%3a%2f%2fmath.stackexchange.com%2fquestions%2f2889768%2fhow-to-prove-that-sum-k-1n-frac1-sqrtnk-sim-fracn-ln-n%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?