[Search for users] [Overall Top Noters] [List of all Conferences] [Download this site]

Conference rusure::math

Title:Mathematics at DEC
Moderator:RUSURE::EDP
Created:Mon Feb 03 1986
Last Modified:Fri Jun 06 1997
Last Successful Update:Fri Jun 06 1997
Number of topics:2083
Total number of notes:14613

1393.0. "question on series expansion" by SMAUG::ABBASI () Wed Feb 27 1991 23:59

    I gave up on this folks,
    any one could say why this series has this closed form?
    Thanks!
    -------------------------------------------------------------------
    oo
    ----
     \       j(2*Pi/6)   -1    n
     /   {  e           Z     }     
    ---
    n=0
    
             1
    =   -----------------------
             j(2*Pi/6)   -1
        1- e            Z
    
    
    where j= sqrt(-1)
          Z is gerenal complex number (a+jb)
    
    this must have been too obviouse to the author, as they dont mention
    slightest clue how it is so. which makes me feel even more depressed,
    never mind other worst feelings..
    /naser
    
T.RTitleUserPersonal
Name
DateLines
1393.1hopefully helpfulHERON::BUCHANANHoldfast is the only dog, my duck.Thu Feb 28 1991 06:4670
(1) Morale

>    this must have been too obvious to the author, as they dont mention
>    slightest clue how it is so. which makes me feel even more depressed,
>    never mind other worst feelings..

	Hey, don't beat yourself up!   I know how easy it is to take it
all personally, but really You're OK.   We've all been there!


(2) As simple as A,B,C=(B�/A)...:-)

>    any one could say why this series has this closed form?

>    oo
>    ----
>     \       j(2*Pi/6)   -1    n
>     /   {  e           Z     }     
>    ---
>    n=0
    
>                   1
>    =   -----------------------
>             j(2*Pi/6)   -1
>        1- e            Z
>    
>    
>    where j= sqrt(-1)
>          Z is general complex number (a+jb)
    
	OK, let's write u = e^(2*j*PI/6)/z.   Then you want

  sum(n=0...oo) u^n.

= lim N->oo sum(n=0...N) u^n	[*]

= lim N->oo (1-u^(N+1))/(1-u)	[multiply * by 1-u to see this]
				[assuming u <> 1 for the moment]

= 1/1-u - lim N->oo (u^(N+1))/(1-u).

	Now the question is, what is the limit of this second term?   It
depends on the modulus of u.   If |u| < 1, I hope you can see that this
term will go to zero.   If |u| > 1, then the term will go to infinity and
so your series doesn't converge.

	Lastly, let's look at what happens when |u| = 1.   If u = 1, then
the series diverges and we could never get the expression * above.   For
other u, we can express u = exp^(2*j*PI*�), where � lies in (0,1).   u^N
does not converge to 0, it just whizzes round with modulus always 1, so there 
is no way that your series could converge to anything finite.

	This is all actually a very common situation.   We say that your series
has a *radius of convergence*, which equals 1 in this case, and is here
centred on the origin.   While u is inside the circle, the series converges to 
1/1-u.   While u is outside, the series diverges to oo.   What happens *on*
the circumference can vary, and in this case it's always divergent.   

	If you substitute back for u:
	|u| = |e^(2*j*PI/6)/z|
	    = |e^(2*j*PI/6)|/|z|
	    = 1/|z|

	So, your series converges not for all z, but for those z such that
|z| > 1, ie *outside* the unit circle.   "general z" don't work.

	I hope this is helpful...

Regards,
Andrew.
1393.2ALLVAX::JROTHThe ships of state sail on mirage...Thu Feb 28 1991 08:5828
    This is a geometric series that collapses, much like compound
    interest (where I first saw this as a kid).

    A heuristic way to think about it is to formally perform long
    division:

	  1 + z + z^2 + ...
        ____________________________
    1-z ) 1
	  1-z
	  ---
	    z
	    z-z^2
	    -----
	      z^2
	      .......

    Breaking this off after n terms we have a remainder of z^n/(1-z)
    so loosely speaking we might hope that if z is less than 1 this
    remainder will approach zero.

    The full theory of convergence Andrew mentions actually appeared centuries
    after mathematicians had started playing with series in a formal way.

    It was suspicion about the validity of these manipulations that brought
    on the seemingly involved theorems we take for granted today.

    - Jim
1393.3formal rules.CADSYS::COOPERTopher CooperThu Feb 28 1991 10:4741
RE: .1 (Andrew)

    Without disagreeing with your conclusions, I have a minor nit to pick
    with your argument.

>= lim N->oo (1-u^(N+1))/(1-u)	[...]
>
>= 1/1-u - lim N->oo (u^(N+1))/(1-u).
>
>	Now the question is, what is the limit of this second term?   It
>depends on the modulus of u.   ...  If |u| > 1, then the term will go to
>infinity and so your series doesn't converge.

    You got from the first step above to the second by applying to
    algebraic rulse for limits:

      (1) lim x->a (f(x) - g(x)) = (lim x->a (f(x))) - (lim x->a (g(x)))
      (2) lim x->a (f) = f when f is independent of x.

    These are quite correct but there is a restriction on (1) -- it is
    only applicable when both the limits on the right converge.  It is
    possible for them to fail to converge and yet the expression on the
    right to converge.  For example:

	lim N->oo (0) = lim N->oo (N-N) = (lim N->oo (N)) - (lim N->oo (N))

    obviously (by the second rule above) the limit on the left converges
    while neither of the two on the right do.  You cannot, therefore,
    simply conclude from the non-convergence of the expression on the
    right that the original expression will not converge.

    You might be able to show that for the particular case of

	    h = lim x->a (f - g(x)) {x not in f}
	    f - g(x) converges iff g(x) converges,

    but I haven't seen that anywhere.

    As I said, just a nit.

				    Topher
1393.4HERON::BUCHANANHoldfast is the only dog, my duck.Fri Mar 01 1991 08:574
Sheesh, if I can't even take a constant out of a limit without being jumped on!

Regards,
Andrew.
1393.5ApologyCADSYS::COOPERTopher CooperFri Mar 01 1991 12:4014
RE: .4 (Andrew)

    Sorry.  My note came across as much more critical than I intendend --
    written too quickly without having read it over.  Of course you knew
    what you were doing.  What I was trying to do is to warn people that
    what you did is not, *in general*, correct.  The reason I thought
    it was worth "nitting" you about is that it is a common error for
    beginners -- they partition some perfectly reasonable, convergent
    expression into non-convergent expressions and then give up on the
    assumption that the original was not convergent.  I wanted to put
    in a "watch it" for the less experienced.

					Topher

1393.6Other collapsing operations :-)CHOVAX::YOUNGBulldozer &amp; Buzzsaw MethodologySun Mar 03 1991 16:488
    Re .2:
    
>    This is a geometric series that collapses, much like compound
>    interest ...
    
    and much like many of those institutions that deal in compund interest.
    
    --  Barry