1. Introduction
Let
a function which applies a set
in itself. We iterate
indefinitely. Sometimes, the process converges to some fixed point or to some cycle. But, in many cases, it is quite impossible to know, after a long time, the position of the iteration
.
One of the best deterministic method to study the asymptotic behavior of the iteration is to linearize
. It consists to find a function
such as:
where
is linear.
When this linearization is possible,
and we obtain asymptotic cases, which are the generalization of the unidimensional cases
. But, we meet some well-known difficulties: the function
is solution of a functional equation; the basins of attraction around each fixed point of
may have fuzzy frontiers… If
, we have other difficulties called resonance in
. It is often a good approach near each fixed point. But, in many cases, we don’t know to treat mathematically the problem.
So, it seems very important to understand what happens when we iterate
indefinitely, especially if the set
is bounded when
applies C in itself. In this case, a probabilistic approach with invariant measures gives other information. It is the object of this paper which is the synthesis of a long work: some demonstrations, as the one of the proposition in Section 4.4, can be found in previous articles or book [1].
2. The Perron-Frobenius’s Measure
Let P a measure on a measurable set
and
a measurable function. We note
the transform of P by
. We define the
-invariant measure P of Perron-Frobenius as in [2]:
P is invariant under
if, for all borelian set B, P verifies the Perron-Frobenius’s equation (PF):
This measure P remains invariant when we iterate the measurable function
. Under very general conditions, the solution of this equation is unique.
This measure presents the same difficulties as we have seen with linearization methods for
: we will see that it depends of the fixed points and we meet the resonance’s problems, but it gives us many information about the areas where the iteration belongs more frequently. This information is asymptotic when
but doesn’t give us any result about the transient steps of null measure.
This invariant measure is generally difficult to study. For instance, in the very simple case where
is invertible, its density p verifies the functional equation:
This is very complicated to solve. If
is not invertible, it is more difficult.
2.1. The Fourier-Laplace’s Transform
Here, we seek an analytic approach of the
-invariant measure with the Fourier-Laplace’s transform. We use the known property of this measure: for all positive P-measurable function g, we have the formula [2]:
.
For
, we write the Fourier-Laplace’s transform
with the series
and
. If the measure is invariant:
Here,
or
. If
, we have the characteristic function of the measure P.
Hypothesis
All along the paper, we suppose that the set
is bounded.
applies C in C and
is at least
.
Anymore, the series
is convergent because it is bounded by an exponential series with the diameter of C.
We translate the distribution with a small fixed vector
,
.
So:
by
-transform:
.
If the measure is invariant:
.
Proposition:
The resolving equation Ra of PF is:
.
If
is a
-function, then, for
and
:
.
Let
be the gap of order n:
At
:
.
As
is an identity,
for all p and q.
■ If the random variable
has a measure P with density
, the translated random variable
has the same translated density. Using the convergent series:
, we have for all small translation
of the random vector X, the translated density:
.
So, we can write
as a distribution in the sense of Schwartz:
.
As
.
And
.
By difference, we get:
. ■
Remarks
- We observe that
where
is a Bell-polynomial in y with degree n. We can note the gap:
And:
is a polynomial in y with degree n.
- We obtain
by putting
instead of
in the series of
.
- We study the problem near a fixed point 0 of
:
, then:
, and:
. But, the other
are unknown.
2.2. Consequences
The general solution of the linear equation
has the form
where b is an arbitrary constant real. So, we can write
with the arbitrary constant b. It means that
, for all
. We have a lattice distribution of probability for
.
The solution of the Perron-Frobenius’s equation is a particular case of the equation
.
First, we show the effect of an iteration on
and on
.
Proposition
Iteration
acts as a derivation on
and on
or on
in the sense:
and
induces
or
.
By induction, all the coordinates of
in
are:
.
■ The demonstration in [1] is based on the mean’s formula for
.
For example, we study the impact of
on
.
As
:
So:
When
Then, if we iterate
, that means
, we obtain:
.
If this quantity is null for all a, then
. For similar raisons, if
, then:
. ■
3. Solution of the Resolvent R0
Now, we choose a sufficiently large index
, with
Lemma
For a fixed
, under non-resonance conditions, if a solution of
exists, the zeros of
are zeros of
.
■ The solution of this equation is obtained as the following:
We choose a sufficiently large index
such as:
uniformly.
Then:
verifies uniformly:
.
As
, we search an approximation
, and estimators
such as we have:
.
- For
:
. As
is a polynomial, the condition
implies ether all the coefficients of
are null or the solution is valid only for the y verifying
. But, as the term of highest degree of
is:
, we must have, under non-resonance conditions and for all
,
. (Because all the other gaps
have a lower total degree for all
).
Then, zeros of
are zeros of
. ■
Theorem
Under the non-resonance condition, we can find a unique convergent solution of
, up to an arbitrary constant b:
.
We obtain a lattice distribution defined by the zeros of
.
In the repellent case where
, we have:
.
Then, the distribution of the real zeros of the polynomials
gives the distribution of the Perron-Frobenius’s measure when
.
We obtain a lattice distribution defined by the zeros of
.
■ We note the polynomials
.
and
So:
.
- We search a solution under the condition
.
1) Now, for all
verifying
, can we find a solution of
?
For all
, we note:
.
If
:
Where all the coefficients of
are known because
is defined by the coefficients of the Bell’s polynomials.
So, we study in
all the terms of
with degree
:
for a fixed arbitrarily
.
2) We obtain a finite triangular system of linear equations: it can be solved step by step and we can identify in a unique way all the unknown coefficients
in function of the fixed
and the coefficients
of
with
.
3) This solution is unique for all
arbitrarily fixed, near to the solution of
, as the
converge to the
. So, we have constructed the polynomials
and
and we can write
where
is arbitrary. That means
when
; then, we can choose now
. ■
Different cases can happen according to
or
.
If all the coordinates of
are less than 1, the process converges to the fixed point.
If some of them are less than 1, but others are greater than 1, we have a hyperbolic situation under no resonance conditions.
When
, we can write for large n:
. ■
And now we have to study the zeros of
.
Remark (demonstration in [1])
Under the condition that the set C is rectangular, if
is the density of real zeros of
when
, then the invariant density
of the Perron-Frobenius’s measure is:
.
4. Study of the Zeros of
in the Repellent Case
We suppose
is
, without resonance, and
bounded. The problem is reduced to find the asymptotic distribution of the zeros of
in the repellent case. Here, the distribution of the
-invariant measure P is given in general by the distribution of the real zeros of
when
.
4.1. The Plancherel-Rotach’s Method
We will see soon that all the real zeros of
are distinct when the steepest descent’s method [3] can be applied to
and we get an estimation of the asymptotic distribution of these real zeros.
- First, we use the steepest descent’s method as Plancherel and Rotach do [4]. We recall that the polynomial:
can be represented by the Cauchy’s integral:
where Γ is a closed polydisk around the fixed point 0 of
,
, K is a finite non-null function, without importance in the context [3]. We take
.
We note the integrand
And we call
the Plancherel-Rotach’s function.
- Second, with Plancherel and Rotach, we use the steepest descent’s method. We search the critical point of
. Under the numerous conditions of the general position, the critical point
maximizing
gives the solution. The critical point
is defined by the equation:
.
(A sufficient condition to get this maximum is that the hessian matrix of
, which is Hermitian, is definite negative at
). Let
with
, then:
.
The critical point must be isolated from the other critical points and at a finite distance. Some coordinates of
can be real, the others are complex. Then:
.
(
is the conjugate of
). We notice that
or the real part
of
cannot annul
, but, among the solutions, we have to choose
maximum. Then, only the imaginary part
of
can nullify
.
Proposition
Under the conditions of the general position, the critical point
of the PR-function
gives the real zeros of
. For all complex coordinates
of
:
.
As each iteration
acts as a derivation on
, we see:
.
We obtain asymptotically
when
.
The
have an identic independent uniform distribution on
.
In the unidimensional case, the repartition of the zeros is:
.
■ We can tie these distributions of the PF-equation to each fixed point
. Then, we have local solutions. All these distributions can be masked in various situations. The principle of the maximum of the real part
of
provides a method to define the fuzzy frontiers of the different domains of attraction.
As everybody knows, the steepest descent’s method is difficult to use, but it indicates a very large variety of behaviors [3].
In the case of unidimensional function, the repartition of the zeros verifies:
with
So:
Because:
. ■
Remark
If the
are not equal, we take
, and we fix:
. Then, if
and
, the Plancherel-Rotach’s function is:
If
, we have:
.
4.2. Real or Imaginary Coordinates of the Solutions
The reality or the imaginary of the coordinates of
may vary with the orientation of y. The right framework to analyze this question seems to be the Morse’s theory.
4.3. Examples
- Let the logistic map [4]:
; and
;
we put
, we have:
with roots:
and:
.
So:
.
If we put
Then t follows:
.
We recover directly a well-known result: Let
where
is (with easy transformations) like the generatrix function
of the Hermite polynomials
. The law of the zeros of
is known as the semi-circular Wigner’s law:
.
- Then, the density of the logistic corresponding to q(s) is:
.
We deduce that the density of the logistic map follows a Beta (1/2, 1/2) low in a more general situation than in the Ulam-Von Neumann’s case [5].
- The map
, for
. We have neglected this important case in
our previous papers. In general, this iteration tends to infinite. The corresponding Hermite polynomials
are always positive except if
for the odd index n. It seems that this iteration can serve as a parameter in multidimensional case. So, we will say that “
half the time” and arbitrary for even indexes n.
- m-Hermitian case:
.
The Plancherel-Rotach’s function is:
.
With the critical point
defined by the trinomial equation:
studied by H. Fell.
Consequence
We take now a quadratic function
in
with
. We write the PR function
for every fixed point 0 of
:
the hessian of sQ is symmetric. For all s such as sQ is non-degenerate, it exists an orthogonal transformation T:
, with:
, the diagonal matrix of eigenvalues of sQ and:
Because the volume
is invariant under an orthogonal transformation.
We note
if
and
if
.
Then, the P.R. function
becomes:
where
.
If we note:
and
So:
.
And, applying the logistic calculus to each
and
, we obtain p conditions
half the time and d-p random independent variables following a Beta (1/2, 1/2) low. But, we may have other fixed points:
.
Remark
- We can extend these results to a
function
with the Morse-Palais Lemma as in [6], (p.174 et seq.), if the hessian is definite.
5. A differential Equation as a Repellent Iteration
We consider ordinary differential equation [1]:
where
or
,
,
is a
-application of
in C. The domain C is supposed bounded. The problem is to find a function
verifying this equation with an initial condition:
.
We use the theorical solution of Caratheodory
for
:
.
The differential iteration
We associate the differential iteration
belonging in the bounded domain C:
where
is the path. When we iterate n times, we have:
.
The method gives the solution
by iterating n times
from a starting point
with the path
and this solution
when
:
For
:
.
Then, when
:
with
.
The fixed points of a differential iteration are the zeros
of F:
.
5.1. The Invariant Measure of a Differential Iteration
Now, we submit a probabilistic version of the Poincaré-Bendixon’s problem in
.
Proposition
Under the previous hypothesis, all the non-null measures verify:
.
Then, we have asymptotic random cycles around each fixed point. For all these cycles, the times of return in each very small borelian set around a point of a cycle are constant in probability. Along each cycle, the conditional probability has a constant density.
■ With
for every measurable function F. Then, for
with
, we must have the resolving equation in the neighborhood each fixed point for one or n iterations:
.
That means, especially for
:
As:
.
By continuity:
But
.
In consequence, if
, we have non-null measures verifying
. In other words,
for the invariant non-null measure and some
. Under this condition, the asymptotic behavior is random periodic cycles with an unknown almost period
.
But, when we have many fixed points, the complete solution is more difficult because we meet some problems with domains of domination (see Section 4) and transitions from a domain of a fixed point to an another. ■
Remarks
- Theoretically, if we know the probability’s measure, we can define some statistics (mean, standard deviation…).
- We can try to extend these results to PDE equations and obtain other new results, as in the following:
Let the PDE:
Where
or
,
with
.
After transformation of the PDE into iterations, suppose that one can use the Caratheodory’s solution for the PDE:
.
And we see that the only asymptotic solutions for a non-null measure are periodic cycles with the unknown almost period
.
5.2. Examples
1) Suppose that F has a hessian definite negative, then, when
, it is easy to verify that the critical point verifies:
, with an approximation of
for
.
The critical point
is real and we don’t have a probabilistic solution.
2) Suppose we have a linearity in b: Let
with
and
.
We write:
as: with
;
where:
in order to have
.
We write the Plancherel-Rotach’s function with then
,
and
Putting
and
such as
for
, we obtain when
:
Where
(The change
doesn’t modify the equation
). And the critical point is defined by:
So:
.
The imaginary critical points give the distribution of the cycles. Under general conditions, this distribution doesn’t depend on
but only on
.
6. Critical frequencies
Asymptotically, we have random cycles. Let
be a point on a such asymptotic cycle and a very small invariant borelian around this point. So, we have many large times to return in this borelian. In the differential iteration, we have many and large
which give the same
where T is a random quasi-period [6].
Proposition
When the number of iterations
and if the la hessian of yF is definite negative, the approximation with defines
in function of the critical point
:
where
.
If
is a particular solution and if
is an eigenvector of
for the eigenvalue 1/t, the general solution is disjunctive:
if
or
if
.
The eigenvalue 1/t can be interpreted as a critical asymptotic frequency.
■ Contrary to the previous Section 5, we don’t write the critical point
as a function of s, but s as a function of
. For fixed a on an asymptotic cycle, we recognize the linear affine equation of s depending on the parameter t. We have to find a particular solution
:
.
Formally:
.
This solution
is valid for all
where
is eigenvalue of
at the critical point a. If
,
is a particular solution of the equation and the general solution will be
, then:
And:
.
As
,
. The general solution is disjunctive and shows a discontinuity at the eigenvalues
. ■
Remark: calculation of
is obtained with
for all
which doesn’t belong to the spectrum of
with the series development of
.
Remark: the Fredholm alternative
Here, we have the Fredholm alternative: either we have
for all
or
for
. Suppose we start with
, but
is increasing: what happens when
? What is the physical interpretation? Can we connect this phenomenon to some physical constants or boundaries?
7. Case Where the Hessian Is Degenerated: The Lorenz’s Equation
Generally, the hessian is not definite negative. The Lorenz’s equation [7] is a particularly important example because the differential iteration can be broken down into three independent iterations which have a remarkable feature: a partial linearity; an iteration with a negative hessian which induces a probabilistic solution and another with a positive hessian. It is an ideal example to clarify the previous results.
However, as there is an interpenetration of the distributions related to each fixed point, the connection between the various results remains delicate. The probabilistic presentation seems to be the least bad: it gives the probability of presence except at the places where the domination changes; in this case, we go from a basin to an another.
7.1. The Iteration at Its Repellent Fixed Points
The vectors of this equation are written in bold notations:
where
:
.
The differential equation applies a bounded set C in itself for
(the phenomenon is occurring between a cold sphere at −50˚ and hot sphere, the earth, at +15˚ as the terrestrial atmosphere is modelled by Lorenz).
The differential iteration
associated with a given path
is:
.
This iteration is quadratic, but has a linearity in
.
We recall the known results concerning the fixed points:
The fixed points are zeros of
. If
and
, it exists three fixed points: the point
, and two others symmetric with respect to the axis of c:
and
.
At 0, the eigenvalue’s equation
of the linear part is:
,
But, at
or at
:
Coefficients
are such as these three fixed points are repellent; that means we have to study the distributions around each fixed point. We don’t speak here about attractive cycles, resonances, and some particular values of the parameters, etc. It remains many points to clarify.
7.2. Analysis of the Hessian
Projecting
onto an axis
, we write:
where
is linear for
:
with:
and
is quadratic:
.
The hessian
is degenerated and not definite negative. But, Q doesn’t change when we translate the origin from a fixed point to an another.
First, we examine the matrix of
:
Let
the positive eigenvalue of the characteristic equation of Q:
The matrix of the eigenvectors T is orthogonal and constant for all
.
Corresponding to the diagonal matrix of the eigenvectors:
.
7.3. Change of Basis Near 0
- We calculate in the basis of eigenvectors directly with the Hermite’s polynomials. As T is orthogonal, the transposed T’ is also its inverse:
.
Then, the application
with
transforms:
.
Now, in the basis
, the function
is factorized into three independent functions:
with:
;
;
.
Where:
- To calculate
,
et
, we form:
with:
;
;
Then:
.
- We get 3 independent iterations:
. the first iteration
is linear;
. the second
is a random iteration;
. the third
remains positive, except if
half the time.
- Let the resolving gap
For
, putting
, we have:
.
This gives:
And:
.
Proposition
The solution around the fixed point 0 consists of the intersection of the family of random surfaces defined by:
with the surfaces
and
.
■ With the same calculations of encodings and interchanging the derivations, we have:
;
;
We study separately the three expressions:
- First:
.
Either
, or:
- Second: the polynomial
when
is a Hermite’s polynomial
where x is
. This polynomial
is always positive half the time. In a general way:
. So:
,
And
half the time.
- Third: in the case of
, in addition to the solution
, we have to find the possible invariant distribution of
.
Let the integrand of
.
When
,
with
.
By normalization of the coordinates
, we obtain:
Putting
, we have:
.
We search the critical point:
The imaginary roots are:
.
Under the condition:
:
half the time,
Implies:
The condition becomes:
implies
, then:
.
In any case, we observe that the conditions
allow us to express r et t depending on s and we can write that the density of zeros of s is now:
.
Then,
follows a uniform low on (0, 1) with:
(or:
) and:
.
We also remark that the normalization doesn’t affect the coefficients of the orthogonal matrix:
. ■
7.4. Analysis near
and
We now verify similar results the two other fixed points
and
.
We search the distributions around the two other fixed points. To pass from the fixed point 0 to the fixed point
or
, we have just to put in the iteration instead of
:
or
.
- Calculation for
So, for
;
and
where
becomes
;
then:
As:
becomes for
:
The projection of
on an axis
can be written:
and
is invariant:
is linear for
:
with:
;
;
.
Then T and Λ remain invariant. The following is only a calculus.
We calculate
,
et
, with
:
where
;
;
And:
.
The results are modified; if
is related to 0 and
to
The following calculations remain the same with these modifications.
- Calculation for
When
becomes
the calculation is the same with the coordinates:
.
It remains the problems of domination and frontiers between the various distributions attached at each fixed point.
Remark
We have to go back to the original coordinates. And the solution gives only the probabilities of presence...
8. Conclusions
After this study, we can say, under good conditions, that an EDO is deterministic near the origin of the process, but have random or fixed cycles after a very long time.
With this probabilistic method, we obtain some new results, but we meet also many new difficulties due to the particular steepest descent’s method used to study the Plancherel-Rotach’s function. Many things have to be lightened