MathGroup Archive 1997

[Date Index] [Thread Index] [Author Index]

Search the Archive

Q: complex Log[z] and linear dynamics


anyone know why the following compuation fails:

the linear, planar dynamical system dx/dt = A.x where 
A = {{a,-b},{b,a}} with complex eigenvalues a+ib, a-ib is isomorphic to
the complex dynamical system dz/dt = (a+ib).z where z is a complex
number.

the solution to this system is z(t) = Exp[(a+ib)*t]*z(0) where z(0) is
the initial condition....so if z(t) and z(0) are connected by a
solution and the real part of a+ib is nonzero (the origin is a spiral
attractor/repellor depending on the sign of 'a'), then there is a
unique time 't' that should be obtainable by 

                 t = Log[(z(t)/z(0))]/(a+ib)

yet (at least in Mathematica 3.0) with the choice a+ib = -0.5 + i 2.0
and the initial conditions z(0) = 1 i get the correct (real) time only
when z(t) is about t = +/- 1.5 'away from' z(0) and i get complex valued
't' when z(t) is 'farther away' than that....for example, z(4) =
-0.0197 + i 0.1339 and yet i obtain a complex 't' when using the
formula in the previous paragraph. i've checked the modulus of the
complex time but that's also wrong.

is there a singularity in Log[z], is this a numerical problem, or did
i make an error somewhere?...thanks for the info.

 +---------------------------------+
 |          Alan Calvitti          |
 |       Control Engineering       |
 | Case Western Reserve University |
 +---------------------------------+





  • Prev by Date: Re: How to use inequalities with Solve?
  • Next by Date: [Q] Changing TICK MARKS
  • Previous by thread: InputForm, etc...
  • Next by thread: [Q] Changing TICK MARKS