next up previous
Next: The case of three Up: Rescaling of diffusion coefficients Previous: The reduction of the

III. The initial conditions for the reduced equations.

We show in this Section how to obtain the initial conditions for the reduced set of equations, given the initial conditions for the original set. We will work with an expansion such that these initial conditions can be written as

    eqnarray559

Thus, the values tex2html_wrap_inline2234 and tex2html_wrap_inline2238 correspond to the fixed point that the dynamical system (10)-(11) approaches for a particular initial condition tex2html_wrap_inline2308tex2html_wrap_inline2310 . Given the relations (7), the system (10)-(11) has many constants of motion. Namely, it has a total of N. Due to (7) there are N-1 quantities of the form:

  equation590

that are constants of motion for the system (10)-(11). On the other hand, the total mass

  equation595

where tex2html_wrap_inline2316 and tex2html_wrap_inline2318 are the masses of species tex2html_wrap_inline1952 and Q, respectively, is also a constant for the system (10)-(11) (this is a consequence of our assumption that all feeding terms occur on the slow timescale). Therefore, the N+1-dimensional dynamical system (10)-(11) has N constants of motion. Each level set (i.e., the set of points in the space of concentrations for which each of the N constants has a particular given value) is a one-dimensional curve that generically intersects the N-dimensional set of fixed points (20) at isolated points. Notice that, since all the constants are linear in the concentrations, the level sets are straight lines. Given an initial condition that belongs to one particular level set, then the system evolves according to Eqs. (10)-(11) without ever leaving that level set. Since the level sets are one-dimensional objects (straight lines in this case), the only attractors they can contain are fixed points. Assuming that each level set intersects the set of fixed points (20) only once (which is consistent with our assumption that all the points on the set of fixed points are attracting), then the fixed point that each initial condition approaches is uniquely determined: it is the one defined by the same values for the constants of motion as the initial condition. We schematically depict this situation in Fig. 2 (a).

   figure608
Figure: Integration of the fast equations. The straight line is the level set which is uniquely determined by the values of the (in this case two) constants of motion. The arrow indicates how the system evolves in time. Each level set intersects the manifolds depicted in Fig. 1 at isolated points. The first point of intersection provides the initial condition for the reduced (slow) evolution equations. (a) The case of one fast reversible reaction. (b) The case of one fast irreversible reaction.

The system with k'=0 has the same N constants of motion as before. But now, each level set may intersect the set of fixed points more than once. This is shown schematically for the case with three species in Fig 2 (b). In this case, the set of fixed points are the three planes defined by tex2html_wrap_inline2256tex2html_wrap_inline2258 and q=0. A generic straight line in the three-dimensional space tex2html_wrap_inline2342 will intersect all three planes. Therefore, not all the fixed points can be regarded as stable. Of course, the only fixed points (and initial conditions) that are physically relevant are those for which tex2html_wrap_inline2344 , and each level set can contain at most two such fixed points. This is due to the fact that the level sets are straight lines (see Fig.fig:integ (b)). The continuity of the flow guarantees then that one of these two fixed points is stable. Thus, given an initial condition, also in this case there is a unique fixed point of the dynamical system (10)-(11) that the system approaches eventually in time. This fixed point is the initial condition for the reduced set of equations.

As mentioned in the paper, the reduction may also be obtained directly by looking for an algebraic relationship among the variables (see Sec. V of the paper). When this is done, the information on which is the initial condition for the ``slow'' (reduced) equations is lost. However, it is still possible to integrate the fast portion of all the equations simultaneously as done in this Secion. Given that it is assumed that diffusion acts on the slow timescale, this means solving a set of ODE's. The problem when doing this is that the algebraic relationship that is kept in general may not coincide with the manifold of fixed points for the fast equations. Thus, once the fast ODE's are integrated, it is necessary to look for the point on the ``slow'' manifold defined by the algebraic relationship tex2html_wrap_inline2346 that is closer to the fixed point that the real initial condition approaches under the fast dynamics. This point may then be used as the initial condition for the reduced (slow) equations.


next up previous
Next: The case of three Up: Rescaling of diffusion coefficients Previous: The reduction of the
Silvina Ponce Dawson

Fri Aug 27 03:50:25 ART 1999