1. Not finding help here? Sign up for a free 30min tutor trial with Chegg Tutors
    Dismiss Notice
Dismiss Notice
Join Physics Forums Today!
The friendliest, high quality science and math community on the planet! Everyone who loves science is here!

Closure of set

  1. Sep 25, 2008 #1
    1. The problem statement, all variables and given/known data

    Let A = {m + nx | m, n in Z} and x irrational. Show cl(A) = R.
     
  2. jcsd
  3. Sep 25, 2008 #2
    (1) what have you tried so far? I suggest writing down the definition of closure and what you know about A.
    (2) Are you referring to the extended reals? The real line is usually not a closed set, but [tex](-\infty,\infty)[/tex].
     
    Last edited: Sep 25, 2008
  4. Sep 25, 2008 #3
    cellotim

    i think i am coming close to the solution

    basically we need to show there exists (y - (m + nx), y + (m + nx)) C (y - e/2, y + e/2) for any e > 0. So take y - e/2 which is in R. since x is in R as well, we can apply the standard result from R is an archimedean ordered field to see that there is a unique nx, with n in Z, such that nx < y - e/2 < (n + 1)x. let m be the least integer greater than x, which exists by the AP. then, y - e/2 < (n + 1)x < nx + m since x is in R - Q.

    etc. we can do this on other side and then show y - (m + nx) < y + (m + nx) which i think maybe completes solution by showing y in closure of A, and closure of A is R.

    is this okay

    edit: no maybe i think i need to show first on e/2, then show since e/2 < nx + m, y - e/2 < y - (nx + m) so this part is good. then on other side we see there is nx < e/2 < (n + 1)x. how we can fit this right side into y + e/2?
     
    Last edited: Sep 25, 2008
  5. Sep 25, 2008 #4
    To prove the closure of a set we must show that, for every [tex] y\in R[/tex] and [tex]\varepsilon > 0[/tex], the open set [tex](y-\varepsilon,y+\varepsilon)[/tex] contains a point in A. So let us find it. We are given [tex]y,\varepsilon[/tex]. Can we construct a point that is in A and closer to [tex]y[/tex] than [tex]\varepsilon[/tex]? We can subtract [tex]m[/tex] and divide by [tex]n[/tex], so we only need to show that [tex]x[/tex] is within [tex](y' - \varepsilon',y' + \varepsilon')[/tex], where primes denote the open set after subtracting and dividing. We can assume [tex]y'[/tex] is rational, since if it were irrational we would be done. If we take a [tex]\delta<\varepsilon'[/tex], irrational, can we show that [tex]y'+\delta[/tex] is irrational?
     
  6. Sep 25, 2008 #5
    And if you want to prove [tex]\delta[/tex] exists, just let it be [tex]\sqrt{2}/N[/tex] for sufficiently large N.
     
  7. Sep 25, 2008 #6
    okay i am understanding your argument. in previous problem i proved closure R - Q is R. so, we can choose 0 < delta < e', and if y' + delta is rational then we obtain simple contradiction.

    is it okay?
     
  8. Sep 25, 2008 #7
    Sure, let [tex]y'=\frac{a}{b}[/tex]. Take [tex]\frac{a}{b} + \frac{\sqrt{2}}{N}[/tex], for example. You can show that if this is rational, then it proves that [tex]\sqrt{2}[/tex] is rational, which is a contradiction.
     
  9. Sep 25, 2008 #8
    yes

    also one more thing, iam feeling little bit uneasy when you give "subtract then divide everything in interval"; it is valid to say we have x in (changed interval), then x(operation) in (interval * operation)? because maybe i am thinking "there exists some pair (m, n)"
     
  10. Sep 25, 2008 #9
    If x is in the interval [tex](y' - \varepsilon',y' + \varepsilon')[/tex], then it must be the case that m + nx is in the interval [tex](y - \varepsilon,y+\varepsilon)[/tex]. You can do it explicitly with some algebra on the inequalities.
     
  11. Sep 25, 2008 #10

    Dick

    User Avatar
    Science Advisor
    Homework Helper

    Do feel uneasy with this 'proof'. Feel VERY uneasy. Dividing by an integer changes numbers of the form m+nx (m,n integers) to p+qx (p,q rationals). The simplest proof of the denseness I know starts by saying for every integer n, there is an integer m(n) such that m(n)+nx is in [0,1]. (Basically you just add or subtract an integer to get rid of the whole number part of nx). Now if n and n' are distinct integers then m(n)+nx and m(n')+n'x are also not equal, right? (Why?). Now you have an infinite set of distinct numbers of the form m+nx in [0,1]. Can you see where to go from there? (Hint: think pigeonhole principle and try to show the numbers are dense on [0,1]).

    BTW, cellotim, the real line IS a closed set. Ask around.
     
  12. Sep 25, 2008 #11
    My proof is correct. I was not changing the number m + nx, m and n integers into p + qx, p,q rationals, but subtracting then dividing to get x. The proof of denseness only requires that every open set around a member of the reals contains a member of A, one member, which I constructed in my proof.

    Furthermore, the reals are only a closed set if they contain -infinity and +infinity, which, in my graduate courses was referred to as the extended reals. A set is closed if and only if it contains all its limit points. Therefore, whether the reals are a closed set depends on how you define them.
     
  13. Sep 25, 2008 #12

    Dick

    User Avatar
    Science Advisor
    Homework Helper

    I define the 'reals' as the 'reals'. The usual one. Standard topology. The real line contains all of it's limit points without being extended. I reread the proof. As near as I can tell it looks like you are taking the set A to be {m+nx : m and n integers, x ANY irrational}. That would make A the set of ALL irrationals. We already know that is dense. The set A they are referring to is for x a fixed irrational. Like all numbers of the form m+n*sqrt(2).
     
  14. Sep 25, 2008 #13
    "Can you see where to go from there? (Hint: think pigeonhole principle and try to show the numbers are dense on [0,1])."

    can you illustrate for me how ot use pigeonhole principle here
     
  15. Sep 25, 2008 #14
    Yes, I see my mistake. In which case we are looking for an m such that mx - floor(mx) is in the open set [tex](y-\varepsilon,y+\varepsilon), y\in[0,1][/tex]. On that point you are correct.

    On the second, I think not. This is largely semantics, but perhaps important. Let me give you a quote from Dudley's Real Analysis. "[T]he two points [tex]-\infty[/tex] and [tex]+\infty[/tex] are adjoined to R . . . The result is the so-called set of extended real numbers, [tex][-\infty,\infty][/tex]." There is a wikipedia entry on it: en.wikipedia.org/wiki/Extended_real_number_line.
     
  16. Sep 25, 2008 #15
    Take an integer n and the numbers 0*x, 1*x, ..., n*x of which there are n+1. Now partition [0,1] into [0, 1/n), [1/n,2/n), ... ,[n-1/n,1] intervals of which there are n. By the pigeonhole principle, two of the fractional parts of the n+1 numbers must fall into one of the n partitions.
     
  17. Sep 25, 2008 #16

    Dick

    User Avatar
    Science Advisor
    Homework Helper

    I know you can extend the real line. But I'm talking about the un-extended real line. It's closed.
     
  18. Sep 25, 2008 #17

    Dick

    User Avatar
    Science Advisor
    Homework Helper

    That works. I was just thinking of dividing [0,1] into N equal intervals. Now since there are an infinite number of numbers of the form m+nx in [0,1], one interval contains at least two such numbers. j0k3R_, what you are trying to show with this is that for every epsilon>0, there is an m and n such that 0<m+nx<epsilon.
     
  19. Sep 25, 2008 #18
    For any [tex]x\in R[/tex], one can construct an open ball that is in R. This shows that R is open.
     
  20. Sep 25, 2008 #19

    Dick

    User Avatar
    Science Advisor
    Homework Helper

    Sure it's open. It's open AND closed.
     
  21. Sep 25, 2008 #20
    hi again

    "
    Take an integer n and the numbers 0*x, 1*x, ..., n*x of which there are n+1. Now partition [0,1] into [0, 1/n), [1/n,2/n), ... ,[n-1/n,1] intervals of which there are n. By the pigeonhole principle, two of the fractional parts of the n+1 numbers must fall into on"

    thank you -- eventually i did realize precisely what Dick meant with pigeonhole argument so i wrote precisely this :)

    "
    That works. I was just thinking of dividing [0,1] into N equal intervals. Now since there are an infinite number of numbers of the form m+nx in [0,1], one interval contains at least two such numbers. j0k3R_, what you are trying to show with this is that for every epsilon>0, there is an m and n such that 0<m+nx<epsilon."

    yes :)
     
Know someone interested in this topic? Share this thread via Reddit, Google+, Twitter, or Facebook

Have something to add?