You are on page 1of 10

Abstract

Researchers have been studied nonlinear chemical dynamics as an interdisciplinary field, from
chemistry to mathematics, physics, biology and engineering, for many years. This project covers just
e f he d amic m del ed i hi field, called he c bic a ca ala . I he c e
of the project, it was first aimed to analyze analytically the dynamics of this model related to a
hypothetical reaction in detail in the light of corresponding articles and then, the system was
simulated to examine the system behaviour over wide ranges of control parameters. The existent
model was extended by the inclusion of the force term to see if the chaotic solutions are also
possible for this model, as they are found possible for some other nonlinear chemical oscillator
models. At least, it was expected to see some changes on the phase portrait.
Introduction
The main breakthrough in the field of nonlinear chemical dynamics came with Bel d he
bromination of citric acid catalyzed by cerium ions. He noticed that the color of a mixture consisting of
bromate and cerium ions with citric acid in sulfuric acid oscillated back and forth, with a period of a
minute or more, between colorless and pale yellow. After him, Zhabotinsky discovered that when
malonic acid is replaced with citric acid, the reaction shows a more dramatic red-blue color change. This
eac i i k a Bel -Zhab i k (BZ) eac i a d i ha bee he m died eac i
mechanism to understand the chemical oscillators [1,2]. The common thinking of chemical reactions is
that they occur in a monotonic fashion while showing color change, heat generation or absorption as the
reactants are consumed to give products. However, they may give persistent oscillations in the
concentration of reactants and products. In other words, oscillations are possible for far-from-
equilibrium chemical systems where intermediate or catalyst species oscillate around pseudo-steady-
state values. Hence, chemical oscillations are confined to intermediates and catalyst species. Several
mathematical models have been developed for many years to analyze the dynamics of these oscillatory
reactions and to predict the behavior of complex reaction mechanisms. The most common two-variable
m del a e O eg a hich c ide fi e eac ion steps to analyze BZ and CIMA reactions,
Higgi m del hich i ed a al e he ce f gl c l i [2, 3]. The cubic autocatalator is
another useful two-variable model named by Gray and Scott which includes a cubic nonlinearity [1, 4].
The minimal hypothetical mechanism of the cubic autocatalator is given by the following chemical
steps,
P A 𝑅0 = 𝑘 𝑃
2
A+ 2B 3B 𝑅1 = 𝑘1 𝐴 𝐵
B C 𝑅2 = 𝑘2 𝐵
A B 𝑅3 = 𝑘3 𝐴

The two species of our interest are A and B which are autocatalytic species. [P], [A], [B] and [C] are
molar concentration of the species. 𝑅0 − 𝑅3 are the rate expressions for the reactions. k0-k3 are rate
constants. For the dynamical analysis of the system, the reactions are assumed to take place isothermally.
Addi i all , l chemical a ima i i c ide ed f im lici he e he ec P i
assumed to be an excessive amount in the reaction environment that its concentration remains constant
[1,4,5]. In the literature, this reaction system was initially studied without the last step, which a is slow
uncatalysed step, by considering the Landolt clock reaction which is also referred to the iodate-arsenite
reaction [3].
Background analysis of the mechanism:
Mass balances written for two reacting species A and B of interest due to the reaction mechanism are
𝑑𝐴 2
= 𝑘0 𝑃 − 𝑘1 𝐴 𝐵 − 𝑘3 𝐴
𝑑𝑡
𝑑𝐵 2
= 𝑘1 𝐴 𝐵 − 𝑘2 𝐵 + 𝑘3 𝐴
𝑑𝑡

It is convenient to write these equations in terms of non-dimensionalized concentrations X and Y for


these two chemical species A and B according to the relations X=[A] 𝑘1 /𝑘2 , Y=[B] 𝑘1 /𝑘2 and the
dimensionless time 𝜏 = 𝑘2 𝑡. The problem is then seen to be dependent upon two dimensionless
𝑘1 𝑘 0 𝑃
parameters, 𝜇 = 𝑘2 𝑘2
and 𝑟 = 𝑘3 /𝑘2 , (𝜇 ≥ 0) and 𝑟 ≥ 0 . Then, two differential equation
take the forms of,
𝑑𝑋 𝑑𝑌
= 𝜇 − 𝑋𝑌 2 − 𝑟𝑋 = 𝑋𝑌 2 − 𝑌 + 𝑟𝑋
𝑑𝜏 𝑑𝜏
Here, the first parameter is directly related to the amount of precursor [P] present in the reaction
environment. Since we assumed it is kept constant through the reaction, it can be interpreted as 𝑃 =
𝑃 0 . In the first case, r was set to 0 to investigate only the effect of the precursor concentration and
included with an extensive discussion of the pool-chemical approximation [1,5]. However, it has been
seen that the last step has a dramatic effect on the overall reaction since this additional production of B
triggers the autocatalytic step (second reaction). Moreover, some studies have shown that without this
step, the pooled chemical approximation became invalid for some parameters.
Case 1: The effect of precursor concentration with parameter µ where r=0 [3, 4]
𝜇 − 𝑋𝑌 2 = 0

𝑋𝑌 2 − 𝑌 = 0
1
where the equilibrium points for these two equations occur at 𝑋 ∗ , 𝑌 ∗ = ,𝜇 . The Jacobian of the
𝜇
2 −µ2 −2
system is J= −𝑌2 −2𝑋𝑌 and is evaluated at fixed points as, 𝐽𝑋 ∗ ,𝑌 ∗ =[ ].
𝑌 2𝑋𝑌 − 1 𝜇2 1
Then, the trace and determinant were found as, 𝜏 = 1 − 𝜇2 and ∆= 𝜇2 , respectively. Since ∆≥ 0,
solutions are stable/unstable node or spiral. Where >1, solutions are stable (stable spirals for 1< <√2 +
1, stable nodes for > √2 + 1, as seen in Fig.1a and 1b, respectively) and <1 are unstable solutions
(unstable nodes for 0< <√2 − 1, unstable spirals start for √2 − 1< <1 as seen in Fig.1c). In the
corresponding article, the authors have shown that periodic solutions are possible values of the
a ame e i he i e al f 𝜇0 < 𝜇 < 1, with the value 𝜇0 determined numerically to be 𝜇0 ≈
0.9003. This result can be also seen clearly in Fig.2. The periodic solutions arise at =1 through a
supercritical Hopf bifurcation since the stable spiral solutions turn to stable limit cycle solutions. Stable
limit cycles are seen nearly between 0.9 < <1. The amplitude of these oscillations increases with
dec ea i g coming near to 𝜇0 .

(a) (b) (c)

Figure 1: (a) 𝜇 > 1 Stable solutions, (b) 𝜇 > 1 Stable spirals better seen near 1, (c) 𝜇 < 1, unstable solutions

The initial concentration of X and Y were chosen to be 1 for


solving the equations. Regarding the comparison seen in
Fig.3, we can observe the parameter dependence of the system
much better when the initial conditions are set to 1 or above.
It is also clear that the system does not show any sensitivity to
initial conditions. This is expected since this Autocatalator
model is a two-variable model that includes two-oscillating
intermediate X and Y. The sensitivity to the initial conditions
is actually one of the essential properties of the chaotic
dynamics and its possibility starts with the three-variable
Figure 2: 0.89 < 𝜇 < 0.99 Limit cycles autocatalators [7].
Figure 3: Solutions with different initial conditions at different parameter regions (from left to right: =0.6, =0.92, =1.5)
A further study including an extra parameter (r 0) showed that the amplitude of the oscillations does
not become infinite near = 𝜇0 , instead, it rises very sharply to some large finite value [5]. Moreover,
they realized, in the vicinity of this rapid change of the amplitude (Fig.6(b)), there may be a chaotic
behaviour because the numerical round-off error causes the solution scheme to jump almost randomly
from the large to the small amplitude solutions . Thus, they suggested the possible existence of unstable
limit cycles in that vicinity since the solutions in this parameter region are already known as unstable.
Case 2: Two parameters system (µ, r) [1, 5,6]
The equilibrium solutions for the equations (r 0) are obtained again easily by 𝑥 = 𝑦 = 0, where the
2
, 𝜇 . J=[−𝑌2 − 𝑟 −2𝑋𝑌 ]
𝜇
equilibrium points for these two equations occur at 𝑋 ∗ , 𝑌 ∗ = 𝜇 2 +𝑟
and is
𝑌 +𝑟 2𝑋𝑌 − 1
𝜇
−µ2 − 𝑟 −2 𝜇
𝜇2 +𝑟
evaluated at fixed points as, 𝐽𝑋 ∗,𝑌 ∗ = 2 𝜇 . Two Hopf bifurcation points exists at the
𝜇 +𝑟 −2 𝜇−1
𝜇2 +𝑟
values,

1 − 2𝑟 ∓ 1 − 8𝑟
𝜇∓ = ( )
2
These two values of approach 0 and 1 as r 0. Periodic
solutions are generated at these points via supercritical Hopf
bifurcations. Moreover, it was seen that towards very small
r, the limit cycle solution of these two 𝜇∓ limit values meets
abruptly around 𝜇∓ =𝜇0 =0.9003. Thus, they concluded that
limit cycles are expected in the interval of 𝜇− < 𝜇 < 𝜇+
and the length of this interval decreases as r is increased.
This interval becomes zero at r=1/8, so any limit cycle is not
possible for r >1/8. In the paper of interest, this condition
was also proved via a theorem via transforming the system
to a Lienard equation [1].
Figure 4: Similar to case 1 at very low r values
The main aim of this paper is to investigate if the system
has any periodic solutions and shows limit cycles at the region where 𝜇>1 with different values of r <1/8
and examine the solutions in detail at the vicinity of the sharp increase to find if there is any unstable
limi c cle l i . H e e , he did fi d a e i dic l i he e ab e 1. Whe e l e
the system at very low r values, the results are found very similar to the results of r=0, and limit cycle
solution starts to occur around 𝜇0 = ≅0.9 as seen in Fig.4. However, if r is increased, periodic solutions
a cc al a l e al e i h a e high am li de (Fig.5(a)). These very high amplitudes
decrease with further increase in r value until the limiting value 1/8 (Fig.5(b), (c)). Then, after r=1/8,
periodic solutions are lost (Fig.5(d)).
(a) (b)

(c) (d)

Figure 5: Dynamic behaviour of the system with different values of r parameter


In addition, the amplitude curves in Fig. 6 showed that we have still a sharper fall in the amplitude at
a d =0.9 i h decreasing r. The paper results show the same non-continuous behaviour around this
fall region when r goes up to 0.002. Thi ha cha ge a f d a g i g i fi i a 0 i he
analysis of the first case, also compared in Fig.6 (b). As I mentioned before, Gray and Thuraisingham
have previously suggested that the two diffe e i f he l i c e f =0.001 a d =0.9
might be connected by a family of unstable periodic solutions. However, when this region is analysed
m e i de ail a i g i h =0.89 =0.91, as the authors of the paper also found, there is no any
unstable limit cycle. Then, by looking at Fig.6(a), it is also proved that there is no possible periodic
solution above r=1/8 where the amplitude gets lost after this point. Thus, solving the system after this
value of r gives only stable spirals and stable nodes as seen in Fig.5(d).

(a) (b)

Figure 6: Amplitude of the oscillations at different values of r parameter


Contribution
Up to this part of the project, the dynamic system related to the reaction mechanism was reviewed in
detail and the results are extended. In addition, as mentioned shortly at the beginning, two-variable
systems do not show any chaotic dynamics (Poincare-Bendixson theorem) which was also proved by
the analysis in the previous section examining wide ranges of the two parameters. Actually, chaos in the
reactions can be regarded as a dangerous or a desired behaviour". For example, chaotic behavior is
desirable for combustion processes to enhance the agitation of the air-fuel mix and, consequently, to
accelerate the process [7]. Regarding this, chaos control plays an important role. Therefore, there are
several studies going on for chaotic dynamics of the chemical reactions and the control of the chaos.
These researches showed that chaotic behaviour is possible also for the two-variable autocatalators when
a certain force acts on the system. Moreover, controlling this chaotic behaviour becomes possible via
manipulating the force applied. It is most common to see such kind of forced systems applied to the
Brusselator model [7,8]. Therefore, this project aims to apply a periodic force to this previously
examined reaction mechanism and to see if this chaotic behaviour is also possible for this cubic-
autocatalator model. The forced cubic autocatalator has a form of,
𝑑𝑋
= 𝜇 − 𝑋𝑌 2 − 𝑟𝑋 + 𝒇𝐜𝐨𝐬 𝒘 ∗ 𝒕
𝑑𝜏
𝑑𝑌
= 𝑋𝑌 2 − 𝑌 + 𝑟𝑋
𝑑𝜏

where f is the intensity of the sinusoidal force and w is the angular velocity. Then, the system can be
controlled by parameter r, so-called chaos control parameter.

(a) (b)

Figure 7: Oscillations at different values of chaos control parameter f


First, it was seen that the sensitivity of the system to the initial conditions is highly affected by forcing.
Moreover, numerical simulations were shown that forcing this autocatalator is not easy as in the case of
Brusselator where even very small f values cause chaotic dynamics. Hence, chaos control is possible for
those reactions. However, cubic autocatalator system requires quite a dominant forcing term to have a
chaos behaviour. As seen in Fig.7(a), we have periodic solutions at small values of parameter f. When
it is increased up to 5, where r=0.01 and =0.8, period-doubling bifurcation of cycles occurs (Fig.7(b)).
This behaviour is actually common for forced oscillators.

Figure 8: Oscillation and the phase plane at f=0.5


Moreover, it was observed that the system has a really different periodic behaviour at f=0.5 and that
causes a change in the phase plane as seen in Fig.8. As the uniqueness theorem tells us trajectories
cannot cross and merge however the common illusion happens here. It shows that the system behaviour
is chaotic for these parameters (f=0.5, r=0.01, =0.8) because neighbouring trajectories separate by
spiraling out and then, cross without intersecting by going into the third dimension [9]. Here, by looking
at this behaviour, we can also see why three dimensions are needed for chaotic behaviour.
Conclusion
In this project, the cubic autocatalator model for oscillating chemical reactions was studied in detail to
sum up the complete idea of the dynamic behaviour of this system. It was seen that the inclusion of the
uncatalysed step affects the dynamic behaviour of the system, especially by affecting the parameter
range where limit cycles appear and the amplitude of these oscillatory solutions. Moreover, the forced
cubic autocatalator system was built to obtain a chaotic reaction behaviour that was inspired by the
studies on the forced brusselators. However, the cubic autocatalator system has been found less sensitive
to the force acting on it. In other words, the force should be made very dominant via a high value of
control parameter to see any chaotic change in the phase plane. At some particular values of control
parameter f, it was seen a different dynamic behaviour, may be called as a strange attractor, for this
system.

References
1. Forbes, L. K., & Holmes, C. A. (1990). Limit-cycle behaviour in a model chemical reaction: the cubic autocatalator. Journal of engineering
mathematics, 24(2), 179-189.
2. Epstein, I. R., & Showalter, K. (1996). Nonlinear chemical dynamics: oscillations, patterns, and chaos. The Journal of Physical Chemistry,
100(31), 13132-13147.
3. Beutel, K. M., & Peacock-L e , E. (2007). Che ca c a : -variable chemical models. Parameters, 2, 10-3M.
4. Merkin, J. H., Needham, D. J., & Scott, S. K. (1987). On the creation, growth and extinction of oscillatory solutions for a simple pooled
chemical reaction scheme. SIAM Journal on Applied Mathematics, 47(5), 1040-1060.
5. Gray, B. F., & Thuraisingham, R. A. (1989). The cubic autocatalator: the influence of degenerate singularities in a closed system. Journal
of engineering mathematics, 23(3), 283-293.
6. Merkin, J. H., Needham, D. J., & Scott, S. K. (1987). On the structural stability of a simple pooled chemical system. Journal of engineering
mathematics, 21(2), 115-127.
7. Sanayei, A. (2010, June). Controlling chaotic forced Brusselator chemical reaction. In Proceedings of the World Congress on Engineering
(Vol. 3).
8. Bashkirtseva, I. A., & Ryashko, L. B. (2000). Sensitivity analysis of the stochastically and periodically forced Brusselator. Physica A:
Statistical Mechanics and its Applications, 278(1-2), 126-139.
9. Strogatz, Steven H. (Steven Henry) author. (2015). Nonlinear dynamics and chaos : with applications to physics, biology, chemistry, and
engineering. Boulder, CO :Westview Press, a member of the Perseus Books Group,
APPENDIX
A1. MATLAB code for the first case:
close all
clear all
global mu;

N=5; %number of different mu values


u=zeros(1,N);
u(1)=0.88;

for i=1:N
u(i+1)=u(i)+0.02;
mu=u(i+1);
time=500;
%ode solver
F0=[1 1]'; %initial conditions
[t,F]=ode45(@odesolver, [0 time], F0);

X=F(:,1);
Y=F(:,2);

A(i)=max(X); %amplitude of oscillations


% hold on
% plot(t,X)
legends{i}=[ 'u=', num2str((mu))];
hold all;
plot(X,Y, 'LineWidth',2) %plotting phase portrait
colororder((parula(i)))
axis([0 1.3 0 1.7])
legends{i}=[ 'u=', num2str((mu))];
title('\fontsize{14}\mu >1')
%title('\fontsize{14}0.9003 (\mu_o) < \mu <0.99')
xlabel('X')
ylabel('Y')
end
legend(legends)
% plot(u(2:N+1),A,'-*','LineWidth',2) %plotting amplitude
% xlabel('\mu')
% ylabel('A')
% title('Amplitude of the oscillations with decreasing \mu')

function dF= odesolver(t,F)


global mu;
X=F(1);
Y=F(2);
dXdt=mu-X*Y^2;
dYdt=X*Y^2-Y;
dF= [dXdt dYdt]';
end

A2. MATLAB code for the second case:


%very similar, just has another for loop for the second parameter r
rvalues=[0 0.0001 0.001 0.003 0.01];

global mu; global r;


for j=1:size(rvalues,2)
r=rvalues(j);
u(1)=0.889;
Nu=11;
for i=1:Nu
u(i+1)=u(i)+0.01;
mu=u(i+1);

time=500;
%ode solver
F0=[1 1]';
[t,F]=ode45(@odesolver, [0 time], F0);
X=F(:,1);
Y=F(:,2);

A(i)=max(X); %amplitude of oscillations


% hold on
% plot(t,X, 'LineWidth',2) %plotting time evolution of X
% xlabel('t')
% ylabel('X')
% title('\fontsize{14}r=0.001')
hold on;
plot(X,Y, 'LineWidth',2) %phase portrait
colororder(flipud(parula(i)))
axis([0 13 0 13])
legends{i}=[ 'u=', num2str((mu))];
title('\fontsize{14}0 < \mu <1 where r=0.003')
xlabel('X')
ylabel('Y')
end
% hold on
% plot(u(2:Nu+1),A,'.-','LineWidth',1,'MarkerSize',12) %plotting amplitude
% xlabel('\mu')
% ylabel('A_x')
% axis([0.88 1 0 80])
% title('Amplitude of the oscillations')
% legendsA{j}=[ 'r=', num2str((r))];
% colororder(flipud(winter(j)))
end
% legend(legends)
% legend(legendsA)

function dF= odesolver(t,F)


global mu; global r;
X=F(1);
Y=F(2);
dXdt=mu-X*Y^2-r*X; %now, equations contains r
dYdt=X*Y^2-Y+r*X;
dF= [dXdt dYdt]';
end

A3. MATLAB code for the forced case:


rvalues=[0.01]; %can be changed as we want to try
f=[5]; %related to the force term
w=[0.9]; %related to the force term
global mu; global r; global f; global w; %another parameter related to the
force

for j=1:size(rvalues,2)
r=rvalues(j);
u(1)=0.8;
Nu=1;

for i=1:Nu
u(i+1)=u(i);
mu=u(i+1);
time=500;
%ode solver
F0=[1 1]';
[t,F]=ode45(@odesolver, [0 time], F0);
X=F(:,1);
Y=F(:,2);

%A(i)=max(X); %amplitude of oscillations


figure(1)
%hold on
plot(t,X, 'LineWidth',2)
xlabel('t')
ylabel('X')
title('\fontsize{14}f=0.00005' )

figure(2)
plot(X,Y, 'LineWidth',1)
colororder(flipud(copper(i)))
% axis([0 5 0 5])
legends{i}=[ 'u=', num2str((mu))];
title('\fontsize{14}\mu=0.8 r=0.05 f=5')
xlabel('X')
ylabel('Y')
end
% hold on
% plot(u(2:Nu+1),A,'.-','LineWidth',1,'MarkerSize',12)
% xlabel('\mu')
% ylabel('A_x')
% axis([0.89 1 0 80])
% title('Amplitude of the oscillations')
% legendsA{j}=[ 'r=', num2str((r))];
% colororder(flipud(winter(j)))
end
%legend(legends)
% legend(legendsA)

function dF= odesolver(t,F)


global mu; global r; global f; global w;

X=F(1);
Y=F(2);
dXdt=mu-X*Y^2-r*X+f*cos(w*t); %forced version
dYdt=X*Y^2-Y+r*X;
dF= [dXdt dYdt]';
end

You might also like