Professional Documents
Culture Documents
Mercan Terenzi Miniprojet 2016
Mercan Terenzi Miniprojet 2016
a stable point (fig. 4) whereas for µ > 0 (fig.2) R∗ becomes (fig 8). Indeed, for further analysis we have chosen with the
unstable (fig. 1). Moreover, we can now find out the critical supercritical model that is more robust because it is lacking
value of µ for which the fixed point at the origin changes the hysteresis phenomenon.
its nature: the critical value µc is equal to zero. Recalling Now that we have understood better our simple model we
that R is the radius of the limit cycle, we would now try to can add an external force to our simplified supercritical Hopf
understand the impact of the bifurcation on the stability of oscillator and obtain equation (2). In order to easily analyze
this equation we need to get rid of the time dependency. We spirals. We can do that with the Jacobian matrix that will
do so by a change of variable that will virtually put ourselves traduce our problem in a 2D linear one. With the trace and
in the oscillating reference: determinant of the Jacobian matrix we will be able to see the
−iΩt
nature of the fixed points. In order to avoid problems derived
Z(t) = z(t)e (8) from the domain of cosine and sine function we write them
Geometrically this applied to our equation is nothing but in function of the other parameters exploiting the fact that
a change in the phase. we have a fixed point in Ṙ = Θ̇ = 0.
Introducing:
−R(µ − R2 )
cos(−Θ) = (15)
∆ = ω − Ω (9) F
−∆R
and applying our change of variables to equation (2) sin(−Θ) = (16)
gives: F
and write the Jacobian matrix only in function of R.
Ż = (µ + i∆)Z(t) − Z 3 + F (10)
(µ − R2 ) − 2R2
− ∆R
That we can express in cartesian coordinates: J= ∆ (17)
R µ − R2
3
ẋ −x + x(µ − y 2 ) − ∆y + F
Plotting the results for the stability analysis gives a char-
= (11) acteristic figure called ”Arnold Tongue” fig. 14. We can see
ẏ −y 3 + y(µ − x2 ) + ∆x that when the difference of phase between the ”Zeitgeber”
We have a fixed point when ẋ and ẏ are both equal to 0 and the endogenous oscillator represented by ∆ is minimal
which is when the two isoclines are null. the force needed by the ”Zeitgeber” to impose synchroniza-
As previously stated we have synchronization when we tion is small and increases as long as the difference of phase
end on a stable fixed point or a stable spiral, equation (11) increase. With Fig. 15 we can see the nature of all the fixed
shows the unequivocal importance that the force of the exter- points and how many there are in every region.
nal oscillator and his frequency will have on the attainment It is remarkable that when we don’t have a stable fixed
of synchronized situation. point or spiral we have an unstable spiral fig. 11, 15. For
To better visualize this result it is helpful to display a the vector field properties and Poincaré - Bendixson theo-
phase portrait fig. 9-13. We will gradually vary F or ∆ while rem we see the formation of a stable limit cycle. This limit
keeping µ constant. Since we have applied the substitution cycle represent an oscillation that in our case is the internal
now |Z(t)| coordinates in a stable fixed point or stable spi- oscillator that will never synchronize with the external one
ral represent the amplitude of the synchronized system com- and for every cycle will return in the same starting condition.
posed by the two oscillators. Applied to a sleep pattern this could suggest that the studied
A more comprehensive way of showing this result is to living organism will let’s say fall asleep the first night at 10
plot an analysis of all the possible combinations of F and ∆ pm and the successive days always some minutes later (or
in a reasonable range to see which kind of fixed point we get sooner depending on initial conditions) until coming back to
for every possible coupling. the first studied day when he felt asleep at 10 pm. It is thus
To do so we transformed our equation in polar coordi- a sort of entrainment, a behaviour that repeats itself during
nates: time.
Another interesting result that could seem trivial is that
when there is no difference in the external and internal os-
R(µ − R2 ) + F cos(−Θ)
Ṙ cillator phase no external force is needed to attain synchro-
= (12)
Θ̇ ∆+ F R sin(−Θ) nization fig. 12-14.
We write −Θ in function of R and replace it in equation Waves and defects in a line of nonlinear oscillators
(12), this will give us: Now we will focus on discrete oscillators instead of contin-
r uous like in the previous parts. Namely, let’s have a look
F 2 to a system of 512 non-linear discrete oscillators aligned
Ṙ = R(µ − R2 ) + F 1 + (∆ + ) (13) and coupled to their neighbors. Moreover, we will focus on
R
the phenomenon of synchronization, during which the oscil-
We have a fixed point when Ṙ = 0 so from equation (13) lators have the same behavior. Again we will consider the
we derive: model of a subcritical Hopf system equation (2) where the
2
force, represented this time by the term ∂∂xZ2 , is an internal
R6 − 2µR2 + (µ2 + ∆2 )R − F 2 = 0 (14) force considering the system of all the oscillators, but an ex-
ternal one for every single oscillator. The continuum equa-
Finding the roots of this equations will give us the fixed tion therefore becomes the following:
point of our entrained system.
As previously stated to verify if we have synchronization ∂Z ∂2Z
we need to see if those roots are stable fixed points or stable = (µ + iω)Z − Z|Z|2 + (18)
∂t ∂x2
After discretisation of the oscillators we obtain the fol- when µ is comprised between certain values in the phase
lowing equation: portrait a limit cycle appears. This limit cycle is exactly what
we wanted to find since it represent an oscillatory behavior.
zn+1 + zn−1 − 2zn Any small deviation given by an external input will still end
żn = (µ + iω)zn − zn |zn |2 + (19) on the oscillation after stabilization of the system.
∂x2
This allows us to use this model to represent a endogenous
where we assume periodic boundary conditions, that is biological oscillator.
zN +1 = z1 . If we want to study wave-like discrete oscilla- After having proved the validity of this system we added
tors, the equation (19) should behaves like a wave equation the effect of an external oscillating force.
z(x, t) = R(q)ei(Ω(q)t−qx) (20) The study of this influence has pointed out how which
parameters are important to attain entrainment. Most no-
where q is the wavenumber of the wave (with an analysis tably we can see that the difference of the angular velocity
of dimension is clear that q and x have reciprocal units for strongly changes the behavior of the system, when this dif-
a dimensionless exponent). Equation (20) corresponds to ference is minimal a small force is required to see the two
the solution of the system if we state the conditions in polar oscillators synchronized, whether if this difference increases
coordinates the requested force to observe entrainment increases too fig.
( 9-15.
Ω(q) = ω We have been able to show that when there is no entrain-
p (21) ment a situation of pseudo-entrainment arises, with the ap-
R(q) = µ − q 2
parition in the phase portrait of a limit cycle fig. 11.
In order to numerically demonstrate the convergence of We are thus able to use this system as a model to rep-
the solution, we studied this model (equation 19) using a resent a endogenous biological oscillator influenced by the
numerical software. We set as initial condition for all the 512 environment. The force exerted in our model is the same as
oscillators the equation (20) and we added some noise, pro- the one a ”Zeitgeber” will give to our biological oscillator.
vided with a random function, to it (abs(noise) < 1). We In the last part, after having modeled the interaction of an
demonstrated that the system converges toward the wave- external oscillator with an endogenous one, we have focused
like solution when the time tends to ∞ for small value on a model who could better represent many independent
of the parameter c (fig. 17). However, when the value of non-linear oscillators interacting with each other. This could
c increases, the convergence is less clear and some chaos be an initial model for a system as the SNC, even if it is
appears. Even though a definition of chaos is complicated a oversimplification because we suppose a neuron interacts
and debated, we can say that chaos is ”an aperiodical long- only with the previous and the next one in line.
term behavior in a deterministic system that exhibits sen- We have shown that after a critical value of C is attained a
sitive condition on the initial condition” (Strogatz, 2000). wave pattern spreads between the oscillators that fully syn-
That is exactly what we notice from fig 17-18: the wave- chronize themselves.
like solution seems less stable and some bifurcations appear To have a general picture of the situation we could imag-
in time. More drastically, when c cross from 9 to 10 that ine that an external input given by light influences a neuron
is the wavenumber q 2 = 1/3, the system stop showing a in the SCN that becomes entrained and will pass this impulse
wave-like pattern, i.e. the synchronization in space disap- to the other neurons of the SCN. All the system becomes
pears. However, looking at the real part of the solution we thus synchronous and will be able to pass a consistent signal
notice that all the oscillators have the same value at the same to peripheral clocks in the human body.
time. In other worlds, all the oscillators behave in the same It is interesting to see how we can apply different mod-
manner for a given time; thus, we are able to identify a pat- els to represent several aspects of the same global system
tern of synchronization in time but not in space. To sum up, and it is evident how many insights we can discover about
we face three different pattern depending on the value of the biological system using mathematical models to describe
wavenumber c. Firstly, when c is low the wave like solu- them. We should clarify that there are other models that
tion is stable and the oscillators are synchronized in time. are more robust as the biophysical circadian clock Becker-
Secondly, when c is intermediate, i.e. less than 9 and more WeimannBernard model that is closer to reality that could
than 4, a chaotic pattern appears. Lastly, when c crosses from have been used instead of the Hopf oscillator. Since the cir-
9 to 10, a synchronization in time appears even though the cadian clock is the master clock of human body and influ-
wave like pattern disappears. Therefore, we have numeri- ences all the peripheral ones having a deeper knowledge
cally demonstrated that for a higher number of wavenumber of this master switch could allow us to better control and
the synchronization in space and the wave-like pattern is less cure many deficits or illness derived from errors in periph-
probable. eral clocks.
Lastly, we would like to conclude recalling that numeri-
cal demonstrations are usually very robust but do not con-
Conclusions stitute any formal proof. This work, especially on the dis-
With the analysis of the first modified Hopf oscillator, which crete oscillators, should be considered as a first intuition and
should represent a biological oscillator when no external would therefore require a more formal mathematical demon-
force is exerted on it, we have been able to point out that stration, that could be maybe not so straightforward.
References Figures
1. Strogatz, Steven H. Nonlinear Dynamics and Chaos: With
Applications to Physics, Biology, Chemistry, and Engi-
neering. Cambridge, MA: Westview, 2000. Print.
2. Friesen WO, and Block GD. ”What Is a Biological Os-
cillator?” The American Journal of Physiology. U.S. Na-
tional Library of Medicine, n.d.
3. Honma K, Honma S, and Wada T. ”Phase-dependent Shift
of Free-running Human Circadian Rhythms in Response
to a Single Bright Light Pulse.” Experientia. U.S. National
Library of Medicine, 1 Dec. 1987. Web.
4. Roenneberg T, Kumar CJ, and Merrow M. ”The Human
Circadian Clock Entrains to Sun Time.” Current Biology
: CB. U.S. National Library of Medicine, 23 Jan. 2007.
Web.
5. Zee PC, and Manthena P. ”The Brain’s Master Circa-
dian Clock: Implications and Opportunities for Therapy
of Sleep Disorders.” Sleep Medicine Reviews. U.S. Na-
tional Library of Medicine, 11 Feb. 2007. Figure 1: Changing the coordinates of the system to polar
simplify the analysis of the fixed point: Ṙ in function of R
6. Honma K, and Honma S. ”Circabidian Rhythm: Its
Appearance and Disappearance in Association with a
Bright Light Pulse.” Experientia. U.S. National Library After changing the coordinates of the system to polar, the
of Medicine, 1 Dec. 1988. analysis of the fixed point is brought to a system of 1 di-
mension. The intersection of the function Ṙ with the x-axis
7. Ute Abraham, Adrin E Granada, Pl O Westermark, represents the fixed point of the system i.e. the radii of √
the
Markus Heine, Achim Kramer, and Hanspeter Herzel. limit cycle. Therefore, we can observe a radius of R∗ = µ
”Coupling Governs Entrainment Range of Circadian with g = 1 and the unstable fixed point at the origin. The
Clocks.” Molecular Systems Biology. U.S. National Li- negative radius does not make sense because R > 0.
brary of Medicine, 2010.
Appendix
Poincaré - Bendixson Theorem
Hypotesis The Poincaré-Bendixson theorem suppose
that:
1. R is closed, bounded subset of plane;
2. ẋ = f (x) is a continuously differential vector field on an
open set containing R;
3. R does not contain any fixed points; and
4. There exists a trajectory C that is ”confined” in R, in the
sense that it starts in R and stays in R for all future time.
Figure 2: The radius R in function of the parameter µ.
Figure 20: C = 11
Figure 18: C = 9