Dismiss Notice
Join Physics Forums Today!
The friendliest, high quality science and math community on the planet! Everyone who loves science is here!

Homework Help: Uniform conevergence

  1. Jul 11, 2006 #1
    Uniform convergence

    I've got a short question to an example.
    I should check the following sequence for uniform convergence on the whole of [itex]\mathbb{R}[/itex]:
    It says that the conevergence is nonuniform, because:
    Obviously they put [itex]x=1/n[/itex]. I cannot see why this is true.
    I tried to differentiate the given function sequence and see, wether the derivative can become zero or not, i.e. I look for maximum values, but the derivative term gets difficult to manage in the end.
    How can I make it clear to myself that the above inequality holds?
    Last edited: Jul 11, 2006
  2. jcsd
  3. Jul 11, 2006 #2


    User Avatar
    Gold Member

    Hello little Hilbert,

    Are you sure that the inequality shouldn't read something like:

    [itex]sup_{\mathbb{R}}|\frac{nx(7+sin(nx))}{4+n^2x^2}|\leq{\frac{7+sin1}{5}[/itex] ?

    You need to find a supremum to show that the series of functions converges uniformly on R.

    Obviously the sinus term can be 1 at max. And after canceling nx you will have the given numerator already. Can you go from there and find the denominator?


    Last edited: Jul 11, 2006
  4. Jul 11, 2006 #3
    No, the greater-equals-sign is correct.
    Estimatation from above is:
    [itex]|\frac{nx(7+sin(nx))}{4+n^2x^2}|\le{\frac{8n|x|}{4+n^2x^2}[/itex], because [itex]|sin(nx)|\le{1}[/itex]...but it leads me to nowhere because I need a majorant independent of x.
    Last edited: Jul 11, 2006
  5. Jul 11, 2006 #4


    User Avatar
    Gold Member

    My bad, I thought you'd have to show that it does converge uniformly, sorry. :rolleyes:
  6. Jul 11, 2006 #5
    [itex]|\frac{nx(7+sin(nx))}{4+n^2x^2}|\le{\frac{8n|x|}{4 +n^2x^2}\le{\frac{8n|x|}{n^2x^2}\le{\frac{8}{nx} [/itex]

    and therefore f(x) is maximal for x = 1/n . You must now proove that

    lim(n->infinity) [ sup (|fn(x)-f(x)|)] > 0

    So : lim(n->infinity) [|8/(nx) - 8|] = 8 > 0

    Now you need to write this properly but the main ideas are there. My experiences with analysis have teached me that you cannot be rigourous enough so if anyone sees a flaw in my reasoning, don't hesitate to rectify my monstrosities.
    Last edited: Jul 11, 2006
  7. Jul 12, 2006 #6


    User Avatar
    Science Advisor
    Homework Helper

    The maximum of f(x) is not at x=1/n. It's probably at someplace gross. Fortunately we don't need to know where it is.

    Every place you see an "n" in your function f_n(x), there's an "x" living next to it, you can think of it as a function of "nx". So if you set x=1/n the result will be entirely devoid of x and n, it's just a convenient place to evaluate f at that gives a constant. You could have taken x=2/n, or pi/n, etc.

    So we know for any n,


    Hence for any n we have


    since f_n(x) takes on a value larger than 1.
  8. Jul 12, 2006 #7
    Taking x = 2/n or pi/n does not bound 8/(nx) maximaly but work just fine too given the fact that you must disprove the unif.conv. The use of x = 1/n seems to be forced by the " sup " part of the equation. Nevertheless, I seem to be using a slightly different theorem than you ( even if they are equiavlent) and I thank you for bringing to my attention an easier way to investigate such interesting matter.
    Last edited: Jul 12, 2006
  9. Jul 12, 2006 #8


    User Avatar
    Science Advisor
    Homework Helper

    Nor would x=1/n. But making 8/(nx) large won't help an upper bound.

    All you've shown is there that f_n(1/n)<8. Bounding one point of the function from above does not tell you anything at all about the sup over the reals. (again I'll points out that the maximum of f_n(x) does not occur at x=1/n, and again that we don't even care)

    You want to bound the sup from below in any case. For that it's enough to have a sequence of points where [tex]f_n(x_n)[/tex] is bounded away from zero, i.e. they are all >=k for some fixed k>0.
Share this great discussion with others via Reddit, Google+, Twitter, or Facebook