Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]


Groups > comp.soft-sys.math.mathematica > #12135 > unrolled thread

Separating functions

Started bycarlos@colorado.edu
First post2012-10-21 06:10 +0000
Last post2012-10-23 05:03 +0000
Articles 5 — 2 participants

Back to article view | Back to comp.soft-sys.math.mathematica


Contents

  Separating functions carlos@colorado.edu - 2012-10-21 06:10 +0000
    Re: Separating functions Roland Franzius <roland.franzius@uos.de> - 2012-10-22 06:11 +0000
    Re: Separating functions carlos@colorado.edu - 2012-10-23 05:02 +0000
    Re: Separating functions carlos@colorado.edu - 2012-10-23 05:02 +0000
    Re: Separating functions carlos@colorado.edu - 2012-10-23 05:03 +0000

#12135 — Separating functions

Fromcarlos@colorado.edu
Date2012-10-21 06:10 +0000
SubjectSeparating functions
Message-ID<k603ld$7lp$1@smc.vnet.net>
Hi,

I have a problem in identification.  In the middle of the solution of a wave propagation problem on a regular lattice using a separation-of-variables method, I get a function of the form

               F(x,y,t)= f(x,y) g(t)

in which x,y are space coordinates and t is time. Both can be very complicated and vary from problem to problem, but Mathematica' Simplify is able to factor them. Here is a fairly trivial 1D example with an F output by //InputForm:


      Bx*(-5*Em - 3*Em*=CE=BD + 2*a^2*=CF=81*=CF=89^2 - 2*a^2*=CE=BD^2*=CF=81*=CF=89^2 + 4*Em*(1 + =CE=BD)*Cos[(a*kx)/
          Sqrt[2]] - Em*(-1 + =CE=BD)*Cos[Sqrt[2]*a*kx])*(Cos[t*=CF=89] - I*Sin[t*=CF=89]))/(2*c0^2*(-1 + =CE=BD^2)*=CF=81)

Here g(t)=(Cos[t*=CF=89] - I*Sin[t*=CF=89])). For more complicated 2D cases, F(x,y,t) becomes a array of product functions, each taking 20-30 lines.

Question: is there a way to separate  f(x,y) and g(t) that does not require external intervention?

Thanks.

[toc] | [next] | [standalone]


#12141

FromRoland Franzius <roland.franzius@uos.de>
Date2012-10-22 06:11 +0000
Message-ID<k62o1v$chk$1@smc.vnet.net>
In reply to#12135
Am 21.10.2012 08:10, schrieb carlos@colorado.edu:
> Hi,
>
> I have a problem in identification.  In the middle of the solution of a wave propagation problem on a regular lattice using a separation-of-variables method, I get a function of the form
>
>                 F(x,y,t)= f(x,y) g(t)
>
> in which x,y are space coordinates and t is time.

> Both can be very complicated and vary from problem to problem, but Mathematica' Simplify is able to factor them. Here is a fairly trivial 1D example with an F output by //InputForm:
>
>
>        Bx*(-5*Em - 3*Em*=CE=BD + 2*a^2*=CF=81*=CF=89^2 - 2*a^2*=CE=BD^2*=CF=81*=CF=89^2 + 4*Em*(1 + =CE=BD)*Cos[(a*kx)/
>            Sqrt[2]] - Em*(-1 + =CE=BD)*Cos[Sqrt[2]*a*kx])*(Cos[t*=CF=89] - I*Sin[t*=CF=89]))/(2*c0^2*(-1 + =CE=BD^2)*=CF=81)
>
> Here g(t)=(Cos[t*=CF=89] - I*Sin[t*=CF=89])). For more complicated 2D cases, F(x,y,t) becomes a array of product functions, each taking 20-30 lines.
>
> Question: is there a way to separate  f(x,y) and g(t) that does not require external intervention?


Trivially, a separating factor can be fixed by setting the variable t to 
a fixed value, most commonly the starting value

Assuming[0 < {t,x,y} < 1,
    f[x_,y_]=FullSimplify[F[0,x,y]];
    g[t_]=FullSimplify[F[t,x,y]/f[x,y]  ]

With logs and trig functions sometimes more sophisticated methods using 
FunctionExpand, TrigToExp, ComplexExpand will be necessary to separate 
products like Exp(i omega t ) Exp(x/lambda) or Sin[omega t] Cos[phi] to 
their most simple forms.


Another way working independently of the art of algebraic factoring 
processes is to take the linear coefficient (or any of the higher order 
ones) of the Taylor expansion of F in the separating variable t

Assuming[0<{t,x,y}<1},
(f[x_,y_] =
   FullSimplify[ Coefficient[Normal[Series[F[t,x,y],{t,0,1}]], 1]];
g[t_] = FullSimplify[F[t,x,y]/f[x,y] ]  )]

Probably it pays to search for a most trival algebraic expression of the 
expansion point t-> t0, where a maximum of subexpressions disappear or 
cancel from the constant expression g[t0].


-- 

Roland Franzius


[toc] | [prev] | [next] | [standalone]


#12145

Fromcarlos@colorado.edu
Date2012-10-23 05:02 +0000
Message-ID<k658d6$h6a$1@smc.vnet.net>
In reply to#12135
On Sunday, October 21, 2012 12:12:42 AM UTC-6, car...@colorado.edu wrote:
> Hi,
> 
> 
> 
> I have a problem in identification.  In the middle of the solution of a wave propagation problem on a regular lattice using a separation-of-variables method, I get a function of the form
> 
> 
> 
>                F(x,y,t)= f(x,y) g(t)
> 
> 
> 
> in which x,y are space coordinates and t is time. Both can be very complicated and vary from problem to problem, but Mathematica' Simplify is able to factor them. Here is a fairly trivial 1D example with an F output by //InputForm:
> 
> 
> 
> 
> 
>       Bx*(-5*Em - 3*Em*=CE=BD + 2*a^2*=CF=81*=CF=89^2 - 2*a^2*=CE=BD^2*=CF=81*=CF=89^2 + 4*Em*(1 + =CE=BD)*Cos[(a*kx)/
> 
>           Sqrt[2]] - Em*(-1 + =CE=BD)*Cos[Sqrt[2]*a*kx])*(Cos[t*=CF=89] - I*Sin[t*=CF=89]))/(2*c0^2*(-1 + =CE=BD^2)*=CF=81)
> 
> 
> 
> Here g(t)=(Cos[t*=CF=89] - I*Sin[t*=CF=89])). For more complicated 2D cases, F(x,y,t) becomes a array of product functions, each taking 20-30 lines.
> 
> 
> 
> Question: is there a way to separate  f(x,y) and g(t) that does not require external intervention?
> 
> 
> 
> Thanks.

Actually this has nothing to do with the topic, but is about font mauling (see above)   Why does the new Google Groups 
convert Mathematics  greek macros like \[Nu] to meaningless character combinations?  In the old Groups 
they appeared like \[Nu] so the expression can be easily cut and pasted into a notebook.  

[toc] | [prev] | [next] | [standalone]


#12147

Fromcarlos@colorado.edu
Date2012-10-23 05:02 +0000
Message-ID<k658dq$h6f$1@smc.vnet.net>
In reply to#12135
Can the following idea be exploited in some way?

If F(x,y,t)=f(x,y) g(t), then F'(x,y,t)=F(x,y) g'(t), where prime is an abbreviation for d/dt (actually
the partial derivative wrt t). Then F'/F = g'(t)/g(t) = (log(g(t)))'.  Integrating g'(t)/g(t) wrt t would
provide G(t)=log(g(t))+C, a generally complex function.  Then g(t)=exp(G) except for a 
factor exp(C).  How do I get rid of the factor?


[toc] | [prev] | [next] | [standalone]


#12148

Fromcarlos@colorado.edu
Date2012-10-23 05:03 +0000
Message-ID<k658e5$h6h$1@smc.vnet.net>
In reply to#12135
Here is an example script:


F=(2-4*x+5*y+x^2+y^2); g=(1+t+t^2+Cos[20*t]); F=f*g;
G=Simplify[D[F,t]/F]; Print["F=",F," G=",G];
H=Simplify[Integrate[G,t]]; H=Simplify[Exp[H]]; Print["H=",H];
Plot[H,{t,0,1}]; Plot[g,{t,0,1}];

[toc] | [prev] | [standalone]


Back to top | Article view | comp.soft-sys.math.mathematica


csiph-web