Various firing activities and finite-

time synchronization of an improved

Hindmarsh–Rose neuron model under
electric field effect

K. Marcel Wouapi, B. Hilaire Fotsin,

F. Patrick Louodop, K. Florent Feudjio,
Z. Tabekoueng Njitacke & T. Hermann
Djeudjo
Cognitive Neurodynamics

ISSN 1871-4080

Cogn Neurodyn
DOI 10.1007/s11571-020-09570-0

1 23
Your article is protected by copyright and
all rights are held exclusively by Springer
Nature B.V.. This e-offprint is for personal
use only and shall not be self-archived
in electronic repositories. If you wish to
self-archive your article, please use the
accepted manuscript version for posting on
your own website. You may further deposit
the accepted manuscript version in any
repository, provided it is only made publicly
available 12 months after official publication
or later and provided acknowledgement is
given to the original source of publication
and a link is inserted to the published article
on Springer's website. The link must be
accompanied by the following text: "The final
publication is available at link.springer.com”.

1 23
 Author's personal copy
Cognitive Neurodynamics
https://doi.org/10.1007/s11571-020-09570-0 (0123456789().,-volV)(0123456789().
,- volV)

RESEARCH ARTICLE

Various firing activities and finite-time synchronization of an improved

Hindmarsh–Rose neuron model under electric field effect
K. Marcel Wouapi1 • B. Hilaire Fotsin1 • F. Patrick Louodop1 • K. Florent Feudjio2 • Z. Tabekoueng Njitacke3 •

T. Hermann Djeudjo4

Received: 3 September 2019 / Revised: 5 January 2020 / Accepted: 10 January 2020

 Springer Nature B.V. 2020

Abstract
Nowadays, it is important to realize systems that can model the electrical activity of neurons taking into account almost all
the properties of the intracellular and extracellular environment in which they are located. It is in this sense that we propose
in this paper, the improved model of Hindmarsh–Rose (HR) which takes into account the fluctuation of the membrane
potential created by the variation of the ion concentration in the cell. Considering the effect of the electric field that is
produced on the dynamic behavior of neurons, the essential properties of the model such as equilibrium point and its
stability, bifurcation diagrams, Lyapunov spectrum, frequency spectra, time series of the membrane potential and phase
portraits are thoroughly investigated. We thus prove that Hopf bifurcation occurs in this system when the parameters are
chosen appropriately. We also observe that by varying specific parameters of the electric field, the model presents a very
rich and striking event, namely hysteresis phenomenon, which justifies the coexistence of multiple attractors. Besides, by
applying a suitable sinusoidal excitation current, we prove that the neuron under electric field effect can present several
important electrical activities including quiescent, spiking, bursting and even chaos. We propose the improved HR model
under electric field effect (mHR) to study the finite-time synchronization between two neurons when performing synapse
coupling across the membrane potential and the electric field coupling. As a result, we find that the synchronization
between the two neurons is weakly influenced by the variation of the intensity of the electric field coupling while it is
strongly impacted when the intensity of the synapse coupling is modified. From these results, it is obvious that the electric
field can be another effective bridge connection to encourage the exchange and coding of the signal. Using the finite-time
synchronization algorithm, we theoretically quantify the synchronization time between these neurons. Finally, Pspice
simulations are presented to show the feasibility of the proposed model as well as that of the developed synchronization
strategy.

Keywords Modified Hindmarsh–Rose (mHR) model  Finite-time synchronization  Synapse coupling  Electric field
coupling  Pspice analog circuit implementation

3
Department of Electrical and Electronic Engineering, College
& K. Marcel Wouapi of Technology (COT), University of Buea,
marcelwk2000@yahoo.fr P.O. Box 63, Buea, Cameroon
4
1
Unité de Recherche de Matière Condensée, d’Electronique et Energy and Environmental Technologies Laboratory,
de Traitement du Signal (URMACETS), Department of Department of Physics, University of Yaounde I,
Physics, University of Dschang, P.O. Box 67, Dschang, PO Box 812, Yaoundé, Cameroon
Cameroon
2
Laboratoire d’Energie et des Systemes Electriques et
Electroniques, Department of Physics, University of Yaounde
I, PO Box 812, Yaoundé, Cameroon

123
 Author's personal copy
Introduction
Like most natural phenomena, the behavior of the neuron
can be described by mathematical equations involving its
main characteristics, notably, the ionic mechanisms that
govern the functioning of a neuron. The knowledge of a
dimensionless dynamic equation system, in this case non-
linear, called the substituting model for the neuron in the
dynamic system, makes it possible, through reference tools,
to have vital information about its functioning and there-
fore, of the nervous system. In this way, Hodgkin–Huxley’s
innovative work aimed at creating a mathematical model
capable of modeling the different behaviors of the elec-
trical activity of neurons.
One of the significant challenges in neuroscience is to be
able to realize mathematical models able to describe the
different behaviors of the electrical activity of the biolog-
ical neurons in the simplest possible way. It is in this sense
that in 1952, physiologists-biophysicists Alan Lloyd
Hodgkin and Andrew Fielding Huxley proposed a neuron
model capable of reproducing the main properties of
electrical activities by considering the effect of ion chan-
nels (Hodgkin and Huxley 1952). This model was taken
over, modified and improved several times; thus, appeared
the model of Fitzhugh–Nagumo (Fitzhugh 1961, 1969).
Subsequently, many other models to describe the neural
activity will be proposed like the Morris-Lecar model
(Morris and Lecar 1981) and the Hindmarsh–Rose model
(HR) (Hindmarsh and Rose 1982) to name these few. The
latter model, due to its algebraic simplicity, has made and
continues to be the subject of several improvements and
studies to analyze and understand the complex dynamics of
membrane potential in the neuron axon (Hindmarsh and
Rose 1984; Ren et al. 2017a, b; Ma et al. 2016). Thus, in
the past decades, several types of HR models such as
classical and modified HR models (Innocenti and Genesio
2009; Innocenti et al. 2007; Gu et al. 2014; Wang and Shi
2020; González-Miranda 2007; Ngouonkadi et al. 2016;
Wu et al. 2016; Mondal et al. 2019; Djeundam et al. 2013),
delayed HR models (Rigatos et al. 2019), fractional HR
models (Dong et al. 2014; Kaslik 2017) and memristor-
based HR models with electromagnetic induction (Ren
et al. 2017a, b; Lv et al. 2016; Ge et al. 2018; Lv and Ma
2016; Lu et al.2017) have been thoroughly examined and
studied by numerous bifurcation analysis methods (Spitzer
2006). Although the three-dimensional HR model has been
proposed since 1984, to the best of our knowledge, a
detailed dynamic study of this model under the effect of the
electric field has not yet been proposed. Indeed, the neuron
is the basic unit of the nervous system, the complex
physical effects such as electromagnetic induction and
polarization must be taken into account when the flow of
123
Cognitive Neurodynamics
charges (calcium, potassium, sodium, for example) spreads
through the membrane. Moreover, ions are pumped from
the endoplasmic reticulum into the cells. Also, the electric
field can cause the polarization of the media, while the
magnetization of the support is important when considering
a time-varying magnetic field because an induced electric
field can be generated in the media. Thus, in this paper,
inspired by the works done by Ma et al. (2019), we propose
an improved HR model that takes into account the effect of
this field.
In order to understand and even master the operating
principle of some neurological processes, it is essential to
study the regulation and transmission of nerve impulses in
the brain. More importantly, complex networks are essen-
tial tools for clarifying the different features of complex
systems (Kivelä et al. 2014; Boccaletti et al. 2014; Estrada
2012; Boccaletti et al. 2006). Synchronization is a uni-
versal phenomenon in complex networks because it is via
this dynamic behavior that the transmission of information
between neurons occurs (Jia et al. 2011; Shi and Wang
2012). In recent years, the study of coupled oscillator
networks and their synchronized activities has interested
many researchers, particularly in biology. Indeed, in neu-
roscience, it is proved that an abnormality in the syn-
chronization capacity of neural networks can be at the
origin of cerebral pathologies such as epilepsy,
schizophrenia, Alzheimer’s disease, Parkinson’s disease,
and autism just to name these few (Uhhaas and Singer
2006). As a consequence, many studies have been carried
out on the phenomenon of synchronization of neurons by
generally considering static couplings (Ma et al. 2017; Perc
2009; Parastesh et al. 2019). Using this as a motivation, we
propose the improved HR model under an electric field
effect to study the finite-time synchronization between two
neurons when performing synapse coupling across the
membrane potential and the electric field coupling. Thanks
to Lyapunov’s theory (used to prove finite-time conver-
gence), one of the main advantages of the synchronization
method we propose (i.e. finite-time synchronization algo-
rithm) is that it allows theoretical determination of the
maximum synchronization time before which the electrical
activity of the coupled neurons behaves identically (Zuppa
et al. 2002; Louodop et al. 2013, 2014a, b).
The rest of this work is organized as follows. We first
describe the improved neuron model by showing how the
1984 HR model is modified so that it can take into account
the effect of the electric field. After that, we present the
analytic expression of the equilibrium point of the system
as well as its different eigenvalues. Then, we present a
study of the stability of the mHR model. The analysis of
the existence of HB in this model for a suitable choice of
parameters is also presented. Thereafter, we present two
essential parts, the first part deals with the study of the
 Author's personal copy
Cognitive Neurodynamics

dynamic behavior of the electrical activity of the neurons. also represents a nonlinear function that defines the trans-
The second part makes it possible to propose an analog membrane current. As is well known, biological neuron
circuit that can be used to make the electronic circuit of the models should take into account the effect of ion channels
model. Next move, we focus on the design of a finite-time that determine ion propagation as well as the membrane
synchronization applied to the modified HR model. For potential. However, the involvement of the field variable E
this, first, we formulate the problem of synchronization; allows us to correctly describe the ion distribution and the
secondly, we present the synchronization process; third, we membrane potential change induced by ion exchange and
present the results of the different numerical simulations; transport in the cell. Therefore, the electric field can be
fourthly, we present the results of Pspice simulations and used as a new variable to estimate the change of ions and
fifthly, we make a discussion on the synchronization fea- the membrane potential of the neuron. Besides, the intrinsic
ture. Finally, a conclusion is made. electric field in the neuron modulated when the media
exposed to the external electric field.
Taking into consideration what is said above, Eq. (2)
Model description becomes:
8
>
> dV
When the concentration of ions (such as calcium, potas- >
> C ¼ f ðV; ic ; pÞ þ Iext ;
>
< ds
sium, sodium) in the cell is changed, this causes the fluc- dic
L ¼ gðV; ic Þ þ k1 E; ð3Þ
tuation of the membrane potential. Thus, when an external >
> ds
>
> dE
electrical excitation beyond the threshold value is applied, >
: ¼ k2 ic þ Eext ;
an action potential may be induced to predict changes in ds
ion distribution density, which may also cause a time- where k2 is a constant that describes the polarization
varying electric field. Considering the electrostatic/elec- property, its value is 2e11 S. Eext represents the external
trodynamics’ hypothesis mentioned above, the electric field
electric field, which can either be a periodic modulation
E across the soma surface (S), charge number q, surface
function or radiation that characterizes a noise. Thus,
charge density r can be estimated as follows:
8 drawing on (Ma et al. 2019), in this work, the neuron model
<E ¼ q ¼ r ; introduced by Hindmarsh and Rose (HR) in 1984 (Hind-
2e1 S 2e1 marsh and Rose 1984) will be improved to take into
: pffiffiffi ð1Þ
DV ¼ k1 E ¼ E S; account the effect of the electric field. The model obtained
after this modification is, therefore, the following:
8
> dx
where e1 represents the dielectric constant associated with >
> ¼ y  ax3 þ bx2  z þ Iext ;
>
> ds
the intrinsic property of the media, DV represents the >
> dy
>
< ¼ c  dx2  y þ k1 E;
potential difference between the plates of the cell, k1 is the
ds ð4Þ
size of the radius when the cell is considered to have a
> dz ¼ r ½sðx þ hÞ  z;
>
>
spherical shape. Briefly recall that one of the most com- >
> ds
>
>
monly used methods of physically modeling the neuron >
: dE ¼ k2 y þ Eext ;
today is to consider the neuron as a neural circuit that can ds
be constructed using a capacitor, inductor, and other nec- where x, y and z respectively describe the membrane
essary electronic components. Therefore, the equation of potential, the slow current associated with the recovery
the generic circuit can be described by: variable and the adaptation current. Generally, the values
8 of the constants set as follows:
> dV
<C ¼ f ðV; ic ; pÞ þ Iext ;
ds ð2Þ a ¼ 1; b ¼ 3; c ¼ 1; d ¼ 5; r ¼ 0:006; s ¼ 4 and
: L dic ¼ gðV; ic Þ;
>
ds h ¼ 1:6: The external stimulus current is a sinusoidal
where C and L represent respectively the capacitance of the function defined by Iext ¼ I1 sinð2pf1 sÞ. For the sake of
membrane and the equivalent inductance of the neuron. V simplicity, the external electric field in this system that we
is the voltage of the membrane, ic represents the current propose is modeled by Eext ¼ I2 sinð2pf2 sÞ.
through the cell, p represents the intrinsic parameter of the
media while Iext indicates the external forcing and synapse
current. The nonlinear function f ðV; ic Þ describes the pro-
cess of the membrane potential that depends on the voltage
of the membrane and the current through the cell, gðV; ic Þ

123
 Author's personal copy
Cognitive Neurodynamics

Equilibrium point and stability m1 ¼ 3ax2e  2bxe þ r þ 1;

m2 ¼ 3aðr þ 1Þx2e þ 2ðd  bðr þ 1ÞÞxe þ r ðs þ 1Þ  k1 k2 ;
We can have an idea about the qualitative behavior of the
system evolution by analyzing what happens around the m3 ¼ 3aðr  k1 k2 Þx2e þ 2ðbðk1 k2  r Þ þ dr Þxe þ r ðs  k1 k2 Þ;
equilibrium point. So, considering the system (4) and m4 ¼ 3ark1 k2 x2e þ 2brk1 k2 xe  rsk1 k2 :
replacing the parameters a; b; c; d; r; s and h by their ð12Þ
values defined previously, this equilibrium point is
obtained by solving the equation system In order to find the different eigenvalues, we will solve
_
x_ ¼ y_ ¼ z_ ¼ E ¼ 0, either: this characteristic equation. Equation (11) can be reduced
8 to the equivalent form:
>
> y  x3 þ 3x2  z þ Iext ¼ 0;
< n4 þ An2 þ Bn þ C ¼ 0; ð13Þ
1  5x2  y þ k1 E ¼ 0;
ð5Þ
>
> 0:024ðx þ 1:6Þ  0:006z ¼ 0; with k ¼ n  m41 , A ¼
3m21
þ m2 ; B ¼
m31
 m12m2 þ m3 and
: 8 8
k2 y þ Eext ¼ 0: 3m41 m2 m21
C ¼ 256  m14m3 þ 16 þ m4 . The terms mi ði ¼ 1; 2; 3 and
Following (Bao et al. 2015, 2018; Xu et al. 2017), we 4Þ are functions of the AC equilibrium point, hence the
solve the system (5) and the AC equilibrium point term B is a function that evolves with time. So, the method
Se ðxe ; ye ; ze ; Ee Þ of the modified HR model is as follows: of solving Eq. (13) depends on the value of B (B 6¼ 0 or
  
Eext 1 2 Eext B ¼ 0).
Se x e ;  ; 4xe þ 6:24; 5xe  1  ; ð6Þ
k2 k1 k2 – When B 6¼ 0, we will use the Ferrari method:
where xe represents the real solutions of the following Equation (13) is equivalent to:
equation: 
  r 2 r2
Eext n2 þ ¼ ðr  AÞn2  Bn þ  C; with r 2 R ð14Þ
Pðxe Þ ¼ x3e  3x2e þ 4xe þ 6:24  Iext þ ¼ 0: ð7Þ 2 4
k2
In Eq. (14), the term on the right is a second-order
Let, polynomial. We are going to look for r 2 R such that the
2
Eext q2 p3 discriminant of ðr  AÞn2  Bn þ r4  C ¼ 0 is equal to
p ¼ 1; q ¼ 8:24  Iext þ and D ¼ þ ; ð8Þ zero, we thus obtain the following equation:
k2 4 27
where D is called Cardan discriminant. From (8), we can r3  Ar2  4Cr þ 4AC  B2 ¼ 0; ð15Þ
easily see that p [ 1, which implies that D [ 0. In this We are brought back to solve a cubic polynomials
case, there exists a single real root which is defined by equation. As PðkÞ is with real coefficients, considering the
(Han et al. 2015; Bao et al. 2018): Cardan discriminant described in (Han et al. 2015; Bao
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi et al. 2018), one of the following roots can be considered as
3 q pffiffiffiffi 3 q pffiffiffiffi
xe ¼  þ D þ   D þ 1: ð9Þ a particular solution of Eq. (15) according to the sign of
2 2
D1 :
The stability of this AC equilibrium point (Se ) can be rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
3 q1 pffiffiffiffiffiffi 3 q1 pffiffiffiffiffiffi A
analyzed by determining the Jacobian matrix associated r1 ¼  þ D1 þ   D1 þ ;
with the system (4): 2 2 3
pffiffiffi rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffi
0 1 1 þ j 3 3 q1 pffiffiffiffiffiffi 1  j 3
3ax2e þ 2bxe 1 1 0 r2 ¼   þ D1 þ
B 2 2 2
2dxe 1 0 k1 C rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Je ¼ B
@
C: ð10Þ p ffiffiffiffiffi
ffi
rs 0 r 0 A 3 q
   D1 þ ;
1 A
ð16Þ
0 k2 0 0 2 3
pffiffiffi rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffi
1  j 3 3 q1 pffiffiffiffiffiffi 1 þ j 3
We can therefore easily deduce the characteristic r3 ¼   þ D1 þ
equation (detðJe  kId Þ ¼ 0, with Id an identity matrix of 2 2 2
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
the same size as Je ) corresponding to the Jacobian matrix 3 q 1
p ffiffiffiffiffi
ffi A
   D1 þ ;
of Eq. (10): 2 3
2
PðkÞ ¼ k4 þ m1 k3 þ m2 k2 þ m3 k þ m4 ¼ 0; ð11Þ with,p1 ¼ 4C  A3 , q1 ¼ 4AC  B2  43 AC  27
2 3
A , D1 ¼
2 3
in which ðq1 =2Þ þðp1 =3Þ and j denotes the unit of the imaginary
number.
Equation (14) can be in the form:

123
 Author's personal copy
Cognitive Neurodynamics
  2
2 r2 B observe that the curve of the xe-component of the equi-
n þ ¼ n ; for r 6¼ A:
2 2ð r  AÞ
The solutions of Eq. (11) belonging to the set of com-
plex numbers are therefore the following:
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 ffi!
1 B m1
k1 ¼ 1 þ 1  2 r   ;
2 ð r  AÞ 4 envisaged (see (Letellier et al. 2013)), the first case is when
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 ffi!
1 B m1
k2 ¼ 1  1  2 r   this case, the equilibrium point is said to be unstable sad-
2 ð r  AÞ 4
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 ffi!
1 B m1
k3 ¼ 1þ 12 rþ  ;
2 ð r  AÞ 4
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 ffi!
1 B m1
k4 ¼ 1 12 rþ  :
2 ð r  AÞ 4 negative real part; in this case, we say that the equilibrium
• When B ¼ 0, this means that Eq. (13) is a simple
quartic equation and it can resolved easily. In this case,
the eigenvalues are as follows:
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi m1
k1 ¼ A þ A2  4C  ;
2 4
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi m1
k2 ¼ A þ A2  4C 
2 4
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi m1
k3 ¼ A  A2  4C  ; remains fixed (I1 ¼ 0:74;I2 ¼ 0:02 and f1 ¼ f2 ¼ 0:01) it is
2 4
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi m1
k4 ¼ A  A2  4C  :
2 4
By varying the normalized time s in an interval from 0
to 200 (or from 0 to 100), the time evolution of the xe-
component of the AC equilibrium point Se and the real
parts of the eigenvalues ki ði ¼ 1; 2; 3; and 4Þ correspond-
ing to this equilibrium point are presented in Fig. 1 for
k1 ¼ 0:000085 and k2 ¼ 0:001. In order to study the stability
of our model, several cases were envisaged. In the first case
(Fig. 1a, b), we assume that the amplitude and frequency of
the excitation current are the same as those of the external
electric field (I1 ¼ I2 ¼ 0:02 and f1 ¼ f2 ¼ 0:01). In the
second case (Fig. 1c, d), we assume that the amplitude of
the external excitation current is different from that of
the external electric field (I1 ¼ 0:74; I2 ¼ 0:02; and
f1 ¼ f2 ¼ 0:01), we can notice in this case that the
amplitude of the excitation current has increased as com-
pared to that of the previous case and that the amplitude of
the external electric field has not changed. In the latter
case (Fig. 1e, f), we consider I1 ¼ 0:74; I2 ¼ 0:02 and
f2 ¼ 0:01, but increase the frequency value of the external
stimulus current from 0.01 to 0.04. From Fig. 1a–d, we
ð17Þ librium point Se evolves over two periods and that the time
evolution of the eigenvalues ki ði ¼ 1; 2; and 3Þ alternates
between negative and positive values while that of the
eigenvalue k4 is almost constant and close to zero. Let us
recall that in order to study the stability of a stationary
point that has four eigenvalues, at least three cases can be
at least one of the four eigenvalues has positive real part, in
ð18Þ dle-point/focus. The second case is when three of the
eigenvalues have a strictly negative real part, and the fourth
value has a zero real part, then the equilibrium point is said
to be marginally stable (quasi-stable) node-point/focus.
The third case is when all eigenvalues have a strictly
point is stable node-point/focus.
It follows that the corresponding stability distributions
evolve in a cycle that goes from a marginally stable inter-
val to an unstable interval, then from this unstable interval
to a marginally stable interval and the cycle starts again in
the next period. Indeed, considering the dynamics of the
system during a cycle (i.e. a period) in the interval [20,
120], the AC equilibrium point Se for I1 ¼ I2 ¼
0:02 and f1 ¼ f2 ¼ 0:01 is unstable in both regions [49.86,
50.20] and [99.79, 100.1] (see Fig. 1b). In this same
interval, when only the amplitude of the external excitation
ð19Þ
current is increased and the value of the other parameters
found that the equilibrium point becomes unstable in the
four regions [44.94, 55.23], [55.8, 58.3], [91.7, 94.2], and
[94.76, 105.1] (see Fig. 1d). As a result, it can be said that
the dynamics of the system becomes more complex as the
magnitude of the forcing current increases. On the other
hand, by observing Fig. 1e, f, it is found that when
increasing the frequency of Iext and the value of the other
parameters is the same as above, this does not have a
significant influence on the stability distributions of the
equilibrium point in a period.
Without loss of generality, consider a particular value of
the normalized time s ¼ 48 set as a representative exam-
ple. In this specific case, the AC equilibrium point becomes
a DC equilibrium point since it no longer evolves
over time (Bao et al. 2018). Using the parameters values of
Fig. 1e, f (i.e.I1 ¼ 0:74; I2 ¼ 0:02;f1 ¼ 0:04,f2 ¼ 0:01,
k1 ¼ 0:000085 and k2 ¼ 0:001) the corresponding equilib-
rium point is unstable (Real (k2) [ 0). Since the coeffi-
cients mi ði ¼ 1; 2; 3 and 4Þ are real numbers and they
depend on the electric field parameters ðk1 ,k2 Þ, it is
deduced that the stability of the modified HR model also
depends on the parameters of the field. Therefore, it is
important to investigate the effects of these parameters on
the equilibrium point stability in order to explain and
123
 Author's personal copy
Cognitive Neurodynamics

(a)

(b)

(c)

(d)

(e)

(f)

123
 Author's personal copy
Cognitive Neurodynamics
b Fig. 1 Time evolution of the xe-component of the AC equilibrium
point Se and the corresponding eigenvalues ki ði ¼ 1; 2; 3; and 4Þ for
I1 ¼ I2 ¼ 0:02 and f1 ¼ f2 ¼ 0:01(a, b); I1 ¼ 0:74; I2 ¼ 0:02; and
f1 ¼ f2 ¼ 0:01 (c, d); I1 ¼ 0:74; I2 ¼ 0:02; f1 ¼ 0:04 and f2 ¼ 0:01
(e, f)
Fig. 2 Basin stability of equilibrium point Se in the ðk1 ; k2 Þ plane
showing the region where this equilibrium point is stable/quasi-
stable (cyan) and the region where it is unstable (magenta) for
0  k1  0:09,109  k2  0:002, while keeping I1 ¼ 0:74; I2 ¼ 0:02;
f1 ¼ 0:04 and f2 ¼ 0:01
highlight possible local bifurcation phenomena. Thus, in
Fig. 2, the parameters k1 and k2 of the electric field vary
simultaneously in the regions 0  k1  0:09 and
109  k2  0:002 respectively, and their influence on the
stability of the equilibrium point is observed (the values of
the other parameters are the same as those of Fig. 1e, f). In
this figure, the pairs of coordinates ðk1 ,k2 Þ for which the
equilibrium point is stable/quasi-stable are represented by
the cyan color zones, while those for which the equilibrium
point is rather unstable are represented by the magenta-
colored zones. Since on this map, one observes a sudden
transition of the equilibrium point stability, which signifies
that there has been a qualitative change of the solutions of
the Eq. (11). Thus, Fig. 3 provides the representation of the
eigenvalues solutions in the plane (Real (k), Imag (k)) for
values of k1 and k2 varying in the interval where this
qualitative change occurs (i.e. 0:001  k1  0:08 and
0:0003  k2  0:001). Note that the coefficients of the
Jacobian matrix Je are all real, which implies that it is the
appearance of complex conjugate eigenvalue pairs that
creates the symmetry observed along the real axis (see
Fig. 3). As a result, the locus intersects the imaginary axis
and thus suggests the possibility of Hopf bifurcation
(Ngouonkadi et al. 2016; Kengne et al. 2017; Wouapi et al.
2019).
0.05
0.04
0.03
0.02
0.01
Imag( λ )
0
-0.01
-0.02
-0.03
-0.04
-0.05
-16 -14 -12 -10 -8 -6 -4 -2 0
Real( λ )
Fig. 3 Representation of the eigenvalues solutions of Eq. (11) in the
complex plane (Real ðkÞ, Imag ðkÞ) for 0:001  k1  0:08,
0:0003  k2  0:001, while keeping I1 ¼ 0:74; I2 ¼ 0:02; f1 ¼ 0:04
and f2 ¼ 0:01. Provided that Je is a real matrix, complex eigenvalues
occur in complex conjugate pairs responsible of the symmetry
observed along the real axis. The locus intersects the imaginary axis
and thus suggests the possibility of Hopf bifurcation
Existence of Hopf bifurcation (HB)
in the model
Theoretical analysis of the existence of HB
It is important to recall that, the transition from stability to
instability in some contests is explicitly linked to the dis-
appearance or birth of a periodic orbit. When a change of
this nature occurs in a system, it is said that there is the
appearance of the Poincare-Andronov-Hopf bifurcation,
better known as the Hopf bifurcation (HB). This bifurca-
tion can provide a good explanation for many physical
phenomena that are usually encountered because it char-
acterizes the simplest mechanism of transition from a sta-
tionary regime to oscillations (or oscillations at a stationary
regime). As an example in neurodynamics, the possible
mechanisms for highlighting how a nervous system reacts
instantly to excitation are described by sudden changes
observed in the dynamic behavior of this system. This
change in the behavior of the nervous system is due to the
appearance of a phenomenon called Hopf bifurcation (HB)
(Ngouonkadi et al. 2016). This is the major reason why the
study and mastery of this type of bifurcation in particular is
very important in the theory of bifurcations. The priority
now is to study the occurrence of a pair of complex-con-
jugate feature of an equilibrium state across the imaginary
axis for the specific normalized time s ¼ 48 (see Fig. 3).
Thus, we perform the analysis of parametric variations with
reference to dynamical bifurcations at the equilibrium
Se ðxe ; ye ; ze ; Ee Þ by employing the classical form theory and
symbolic computations (Wiggins 1990; Kuznetsov 1998).
Note that in Fig. 3, only the electric field parameters k1 and
k2 vary, the other parameters are fixed and it is these fixed
123
 Author's personal copy
Cognitive Neurodynamics

values that will be used in this subsection. The electric field purely imaginary conjugate roots k1;2 ¼  ix0 and a
strength k1 is a significant parameter in the dynamical study strictly negative reals roots k3;4 ¼  m21  12
of this mHR neuron model (see Eq. 3) because as we rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 ffi
mentioned above, it can describe the size of the radius m21  4 m2  m m1 . Our aim now is to deduce a rela-
3

when the cell is considered to have a spherical shape. tionship between system’s parameters corresponding to this
Therefore it is very interesting to find the critical value k1c bifurcation around the equilibrium Se ðxe ; ye ; ze ; Ee Þ. Thus,
for which the HB can appear in the proposed model. We we substitute k ¼ ix0 into the Eq. (11) and we obtain the
briefly recall the following transversality condition to following conditions:
prove that a Hopf bifurcation occurs in a system (Wiggins rffiffiffiffiffiffi
1990; Guckenheimer and Holmes 1983; Wouapi et al. m3
x0 ¼ ; ð22Þ
2019). m1
Suppose that the system x_ ¼ fl ðxÞ ; x 2 Rn ;l 2 R has an and
equilibrium ðx0 ; l0 Þ at which the following properties are
satisfied: m23  m1 m2 m3 þ m21 m4 ¼ 0: ð23Þ

1. Dx fu0 ðx0 Þ has a simple pair of purely imaginary To obtain the control parameter’s values k1c , we replace

eigenvalues kðlÞ and kðlÞ, and other eigenvalues with in Eq. (23) mi ði ¼ 1; 2; 3 and 4Þ by their expressions
negative real parts; described in (12). Hence, after some algebraic manipula-
 tions, we derive the following second-order polynomial
d 
2. dl ðRðkðuÞÞÞ 6¼ 0.
l¼l0 equation from which solutions give k1c :
Then, the system x_ ¼ fl ðxÞ ; x 2 Rn ;l 2 R has a Hopf b1 k12 þ b2 k1 þ b3 ¼ 0; ð24Þ
bifurcation at the equilibrium ðx0 ; l0 Þ:
where
Since what interests us here is the instability related to
the HB, we consider the characteristic equation described b1 ¼ ð3ak2 x2e þ 2bk2 xe  k2 rÞ2 þ k2 ð3ax2e  2bxe þ r
by Eq. (11) calculate around the point of equilibrium Se . þ 1Þð3ak2 x2e þ 2bk2 xe  k2 rÞ;
Then we derive this equation with respect to the control
b2 ¼ ð2ð3arx2e  2brxe þ 2drxe þ rsÞÞ
parameter k1 and we obtain:
ð3ak2 x2e þ 2bk2 xe  k2 rÞ  ð3ax2e  2bxe þ r þ 1Þ
okðk1 Þ okðk1 Þ
4k3 ðk1 Þ þ g1 ðk1 Þk3 ðk1 Þ þ 3m1 ðk1 Þk2 ðk1 Þ  ð3arx2e þ 3ax2e  2brxe  2bxe þ 2dxe þ rs þ rÞ
ok1 ok1
okðk1 Þ ð3ak2 x2e þ 2bk2 xe  k2 rÞ þ k2 ð3ax2e  2bxe
þ g2 ðk1 Þk2 ðk1 Þ þ 2m2 ðk1 Þkðk1 Þ þ r þ 1Þð3arx2e  2brxe þ 2drxe þ rsÞ
ok1
okðk1 Þ þ ð3ax2e  2bxe þ r þ 1Þ2 ð3ak2 rx2e þ 2bk2 rxe  k2 rsÞ;
þ g3 ðk1 Þkðk1 Þ þ m3 ðk1 Þ þ g4 ðk1 Þ ¼ 0;
ok1
b3 ¼ ð3arx2e  2brxe þ 2drxe þ rsÞ2
ð20Þ
 ð3ax2e  2bxe þ r þ 1Þð3arx2e þ 3ax2e  2brxe  2bxe
omi ðk1 Þ
where gi ðk1 Þ ¼ ¼ 1; 2; 3 and 4:
ok1 ; i þ 2dxe þ rs þ rÞð3arx2e  2brxe þ 2drxe þ rsÞ:
Now we assume that Eq. (11) has a pure imaginary root
kðk1c Þ ¼ jxðk1c Þ; ðx 2 Rþ Þ. By substituting it into Consequently, the following conclusion can be made,
Eq. (11), the real parts gives: when k1 passes through the critical value k1c , system (4)
 ! undergoes a Hopf bifurcation at the equilibrium
okðk1 Þ Se ðxe ; ye ; ze ; Ee Þ.
Re ¼
ok1 k1 ¼k1c
   
g4  x20 g2 m3  3m1 x20 þ x0 g3  x30 g1 2m2 x0  4x30 Numerical verification of the existence of HB
  2  2 :
m3  3m1 x20 þ 2m2 x0  4x30
ð21Þ In order to numerically prove the theoretical developments
presented above, we consider the value of the electric field
   parameter k2 for which there is a sudden change in the sta-

Examining these relations, we see that Re okokðk11 Þ 6¼ bility of the equilibrium point Se ðxe ; ye ; ze ; Ee Þ:k2 ¼ 0:00042
k1 ¼k1c
 (see Fig. 2). By using this value of k2 and considering the
0. Under the restriction that < kj ðk1c Þ\0 for j = 3, 4, the
values of the other parameters used to plot Figs. 2 and 3, it is
second condition for a HB is met and the Poincaré–An-
easily deduced by solving Eq. (24) that the critical values of
dronov–Hopf theorem holds. Then, HB can occur at
the parameter k1 are: k1c1 ¼ 0:07355512804 and k1c2 ¼
ðSe ; k1c Þ of system (3). Obviously, Eq. (11) has a pair of

123
 Author's personal copy
Cognitive Neurodynamics
3699:305315: Since the size of the radius cannot be neg-
ative, we only consider the positive value of k1c .
For this critical value of k1c , we obtain the unique equilib-
rium point Se ð1:305780448; 5:968249191;1:016878208;
21:16866275Þ. For this equilibrium point, we obtain two real
eigenvalues k3 ðk1c Þ ¼ 13:95600730; and k4 ðk1c Þ ¼
0:136883873  103 as well as a pair of purely imaginary
conjugate roots k1;2 ðk1c Þ ¼ j0:04055395391 (i.e.
x0 ¼ 0:04055395391). In addition, we also obtain
  
 some small perturbations of the state variables of the sys-
Re okokðk11 Þ ¼ 0:9456047959  103 6¼ 0; then the
k1 ¼k1c
transversality condition is satisfied. As a result, we conclude
that for this judicious choice of parameters of the mHR
neuron model under electric field effect, there is an occur-
rence of Hopf bifurcation.
Dynamic analysis and electronic circuit
of the model
Dynamic analysis: bifurcation, Lyapunov
exponent and various firing activities
In order to investigate the various important phenomena
that the HR model can present under electric field effect,
we solve system (4) numerically using the fourth-order
Runge–Kutta algorithm. It is important to mention that for
all the results presented in this work, the integration step is
always set to Ds ¼ 0:03715 and the calculations are carried
out using variables and constants parameters in extended
mode. Briefly, recall that two indicators are generally used
to identify chaotic behavior in a system, we have the
bifurcation diagram and the Lyapunov exponent. The first
indicator (i.e. the bifurcation diagram) shows the values
visited or approached asymptotically (fixed points periodic
orbits, or chaotic attractors) by a system as a function of the
system control parameter. In other words, the bifurcation
diagram provides models of transitions and instabilities
when some control parameters are varied (Strogatz et al.
1994). It is obtained using the Runge–Kutta algorithm with
the particularity that one calculates the velocity x_ðsÞ of the
single scalar variable xðsÞ at a moment s and then one
calculates the same velocity x_ðs þ MsÞ at a moment
s þ Ms. A test is then carried out according to whether we
want to represent the local maxima (in this case the con-
dition x_ðsÞ [ 0 and x_ðs þ MsÞ\0 must be verified), the
local minima (in this case the condition x_ðsÞ\0 and
x_ðs þ MsÞ [ 0 must be verified) or the local averages (in
this case the condition x_ðsÞ  x_ðs þ MsÞ\0 must be veri-
fied). For each of these cases, we represent on the ordinate
axis of the diagram the coordinate point xðsÞþx2ðsþMsÞ and on
the abscissa axis the corresponding value of the control
parameter. Note that one of these bifurcation diagrams
(local maxima, local minima or local averages) is sufficient
to have the desired information on the dynamics of the
system because they all have the same characteristics.
Concerning the second indicator (i.e. Lyapunov exponent),
the dynamics of the system is evaluated thanks to the
Lyapunov exponent, which is calculated numerically using
the algorithm of Wolf et al. (1985). In particular, the sign
of the largest Lyapunov exponent determines the rate of
tem and, consequently, the nature of attractor. When
kmax \0, all the disturbances disappear, and the trajectories
start sufficiently close to each other, thus converging
towards the same point of stable equilibrium in the state
space. For kmax ¼ 0, initially closed the orbits remain close
but discrete, corresponding to the oscillatory dynamics on a
limit cycle or a torus; and finally, when kmax [ 0 the small
perturbations grow exponentially, and the system evolves
chaotically, we say in the latter case that the system pre-
sents the phenomenon of chaos. To better understand the
complex dynamics of the new system that is the subject of
our study, we have plotted the time evolution of the state
variables, as well as some phase portraits. When we vary
the parameter I1 which represents the amplitude of the
external excitation current, we obtain the spectrum of the
bifurcation diagram of Fig. 4a which shows the local
maxima of the membrane potential x. Moreover, we pre-
sent in Fig. 4b the equivalent graph of the Lyapunov
exponent of the attractor as a function of the parameter I1
that is varied in the interval 0  I1  1:5 with the initial
condition ð2; 5; 0:8; 1Þ for the following values of
the other parameters: f1 ¼ 0:01; k1 ¼ 0:000085; I2 ¼
0:02; f2 ¼ 0:09; k2 ¼ 0:001.
It can be observed from this diagram that, when some
values of the currents are applied, the model can present
very rich dynamic phenomena such as chaos. It can easily
be seen from Fig. 4a perfect harmony between the bifur-
cation diagram and the Lyapunov exponent.
From the time series of the membrane potential of
Fig. 5a and the two-dimensional phase portrait in the (z–x)
plane of Fig. 5b, the chaotic behavior of the modified HR
model can be seen.
In order to confirm that this new HR model can always
present the essential dynamics of neurological behavior, we
present the results of Fig. 6. In this figure, we show that by
applying a suitable sinusoidal external current, we can have
exciting behaviors such as the quiescent state Fig. 6a,
spiking state Fig. 6b and busting state Fig. 6c. Figure 6d
presents the phase portrait in the (x–z) plane of the time
series of Fig. 6c. As a result, the consideration of electric
field E can modulate the polarization and fast current; thus,
123
 Author's personal copy
Fig. 4 Bifurcation diagram a showing local maxima of the membrane
potential x of the attractor and corresponding graph of largest
Lyapunov exponents ðkmax Þ b versus parameter I1 that is varied in
tiny steps in the range 0  I1  1:5 with the initial condition
ð2; 5; 0:8; 1Þ for f1 ¼ 0:01; k1 ¼ 0:000085; I2 ¼ 0:02; f2 ¼
0:09 and k2 ¼ 0:001. A positive exponent ðkmax [ 0Þ indicates chaos
while regular states are characterized with negative values of
Lyapunov exponent ðkmax  0Þ
chaotic behavior can be suppressed to select spiking or
bursting series in the electric activities.
In order to better understand the dynamics of the
improved HR model, we have produced Fig. 7, where two
bifurcation diagrams are superimposed. These diagrams
showing the local maxima of the membrane potential x as a
function of the electric field parameter k1 which varies in
the very small range 0  k1  0:00001799, are obtained for
the values of the parameters f1 ¼ 0:01; I2 ¼ 0:02; f2 ¼
0:09; k2 ¼ 0:001 and I1 ¼ 0:74, when the initial condition
is fixed at ð0:2;1; 0:8; 0:6Þ: Indeed, the green diagram
(respectively blue) is obtained by increasing (respectively
decreasing) the parameter k1 in this interval starting in each
case with the initial condition ð0:2; 1; 0:8; 0:6Þ. This
technique, where the final state of each iteration is used as
123
Cognitive Neurodynamics
an initial condition for the next iteration is usually known
as a continuation bifurcation diagram (or upward and
backward bifurcation) (Njitacke et al. 2018; Wouapi et al.
2019). These bifurcation diagrams reveal a significant
phenomenon when it is occurring in biological systems,
namely hysteretic dynamics, which has as a first conse-
quence the creation of another phenomenon called the
coexistence of attractors (Njitacke et al. 2017; Njitacke and
Kengne 2018; Negou and Kengne 2018).
Indeed, from Fig. 7, we can further analyze the possible
coexistence of the attractors by choosing a value of k1 (for
example k1 ¼ 1:11  105 ) for which the hysteresis phe-
nomenon is observed (i.e. the regions where the green
diagram and the blue diagram are not overlapped). Con-
sidering the parameters of Fig. 7 and using this value of k1 ,
the phase portraits and the corresponding frequency spec-
trum were obtained to prove the coexistence of two
attractors (bi-stability) created by the hysteresis phe-
nomenon (see Fig. 8). As predicted by the upward bifur-
cation (green bifurcation diagram in Fig. 7), a chaotic
attractor is obtained for the initial conditions ðxð0Þ; yð0Þ;
zð0Þ; Eð0ÞÞ = ð2:7; 7; 2; 4Þ (see Fig. 8a). Similarly, as
predicted by the backward bifurcation (blue bifurcation
diagram), a period-3 limit cycle is obtained for the initial
conditions ðxð0Þ; yð0Þ; zð0Þ; Eð0ÞÞ = ð4; 0; 5; 0Þ (see
Fig. 8c). To better illustrate the difference between these
two attractors, we provide the power spectrum (PS) of
Fig. 8b, d. It should be noted that for periodic motion, all
spikes in the power spectrum are harmonically related to
the fundamental, whereas a broadband power spectrum is
characteristic of a chaotic mode of oscillations. Briefly
recall that the periodicity of the attractor is deduced by
counting the number of spikes located at the left-hand side
of the highest spike of the spectrum (the latter being
included).
Panahi’s and collaborators relevant work have shown
that a healthy brain has neuronal activity that is described
by a chaotic (irregular oscillations) electroencephalogram
(EEG), whereas when the brain presents certain pathology
such as epilepsy, the EEG signal obtained is rather periodic
(regular oscillations) (Panahi et al. 2017). It is therefore
interesting to evaluate the effect of the electric field
parameters on the dynamics of the HR neuron model when
these vary simultaneously and continuously. Thus, in
Fig. 9, the two parameters k1 and k2 of the electric field
vary simultaneously in the regions 0  k1  0:01 and
0  k2  0:008 and their influence on the dynamics of the
mHR is observed (the values of the other parameters are
the same as those of the Fig. 5). For each parameters set-
ting, the system is integrated for a sufficiently long time
and the transient period is canceled. In this figure, the pairs
of ðk1 , k2 Þ coordinates for which the neuron has a chaotic
 Author's personal copy
Cognitive Neurodynamics

(a) (b)

Fig. 5 Numerical simulation of: time series (a) and two dimensional views (b) of the attractor projected, illustrating the complexity of the system
for I1 ¼ 0:745; f1 ¼ 0:01; k1 ¼ 0:000085;I2 ¼ 0:02; f2 ¼ 0:09 and k2 ¼ 0:001 with the initial condition ð2; 5; 0:8; 1Þ

8
firing are represented by the green zones while those for >
>
>
dVCx
¼ 4
1
VC 
1
V3 þ
1
V2 
1
VC þ
1
VIext ;
>
> dt 10 R1 Cx y 104 R2 Cx Cx 104 R3 Cx Cx 104 R4 Cx z 104 R5 Cx
which this neuron has a periodic firing are represented by >
>
>
> dVCy 1 1 1 1
>
< ¼ 4 VC  4 V2  VC þ VC ;
the blue zones. In addition, for more visibility, we define dt 10 R6 Cy 10 R7 Cy Cx 104 R8 Cy y 104 R9 Cy E
>
> dVCz 1 1 1
Yn = sgn(kmax) and represent in 3D the surface of the >
>
> dt
¼ 4 VC þ
10 R10 Cz x 104 R11 Cz
Vrsh  4 VC ;
10 R12 Cz z
>
>
electric field parameters k1 and k2 for which the proposed >
> dV 1 1
>
:
CE
¼ 4 VC þ VE :
dt 10 R13 CE x 104 R14 CE ext
neuron model has an irregular or regular oscillation (see
Fig. 9b). Note that in Fig. 9b, the irregular oscillations ð25Þ
surfaces correspond to Yn = 1 (i.e. kmax [ 0) whereas the From Eq. (25), we can easily establish the expressions
regular oscillations correspond to Yn = - 1 (i.e. kmax- of the parameters of Eq. (4) depending on the value of the
B 0). As a result, it can be seen that certain values of the electronic components of Fig. 10:
electric field parameters can allow the neuron to have a
1 1 1
normal behavior while others can create pathology. a¼ ;b ¼ 4 ;c ¼ 4 VC ; d
104 R2 Cx 10 R3 Cx 10 R6 Cy
1 1
PSpice electronic circuit implementation ¼ 4 ; Iext ¼ 4 VIext ;
10 R7 Cy 10 R5 Cx
of the mHR
1 1 1
k1 ¼ ; rs ¼ 4 ; rsh ¼ 4 Vrsh ; r
In this part of our work, the aim is to be able to set up an 104 R9 Cy 10 R10 Cz 10 R11 Cz
analog circuit that will allow us to make a comparison 1 1
¼ 4 ; k2 ¼ 4 ;
between the theoretical/numerical results obtained previ- 10 R12 Cz 10 R13 CE
ously and the experimental results. The circuit diagram that 1
allows us to perform the various simulations in the PSpice Eext ¼ VEext ; x ¼ VCx ; y ¼ VCy ; z ¼ VCz and E
104 R 14 CE
software presented in Fig. 10. The realization of this circuit ¼ V CE :
is carried out with the help of the operational amplifiers
ð26Þ
TL084 and the associated circuits making it possible to
carry out the basic operations like the addition, the sub- Bearing in mind that the time scaling process offers
traction and the integration, the electronic multipliers analog instruments the ability to work with their band-
(MULT) are the analog component versions AD633JN. widths, the unit of time here is 104 . Indeed, this process
They are used to implement the non-linear term of the offers the opportunity to simulate the behavior of the sys-
system. tem at a given frequency by performing an appropriate time
By applying Kirchhoff’s laws to the electronic circuit of scaling consisting of expressing the MATLAB time vari-
Fig. 10, their circuit equations are deduced in the following able s for the PSpice calculation time variable t: t ¼ RCs ¼
form: 10n s with n 2 N (Kengne et al. 2012). Operational

123
 Author's personal copy
Cognitive Neurodynamics

(a) (b)

(c) (d)

Fig. 6 Numerical simulation of the time series of membrane potential I1 ¼ 0:8; c I1 ¼ 2 with the initial condition ð0; 0; 0; 0Þ. d present
in neuron under different external current at f1 ¼ 0:01; k1 ¼ the two dimensional views of the attractor illustrating the complexity
0:000085; I2 ¼ 0:02; f2 ¼ 0:09; k2 ¼ 0:001, for a I1 ¼ 0:01; b of the system when I1 ¼ 2

amplifiers are polarized at ± 14 VDC using a symmetrical
voltage source.
By choosing Cx ¼ Cy ¼ Cz ¼ CE ¼ C ¼ 10nF, and
considering the following values of the other parameters of
the system (4) ða; b; c;d; h; r; s; k1 ; k2 ; f1 ; I2 ; f2 Þ ¼
ð1; 3; 1; 5; 1:6; 0:006; 4;8:5105 ; 0:001; 0:01; 0:02; 0:09Þ,
the equivalent values of the electronic components of the
circuit of Fig. 10 are:
R1 ¼ R2 ¼ R4 ¼ R5 ¼ R6 ¼ R8 ¼ R11 ¼ R14 ¼ R ¼ 10kX;
R3 ¼ 10 10 10
3 kX; R7 ¼ 2kX; R9 ¼ 85 GX; R10 ¼ 24 MX; R12 ¼
10
6 MX; R13 ¼ 10MX; VC ¼ 1V; Vrsh ¼ 0:0384V; VEext ¼
2
ð2  10 sinð5652tÞÞV; and VIext ¼ ðI1 sin ð628tÞÞV:
It can be seen from the above that the amplitude of
the external stimulus current can be varied in the circuit
of Fig. 10 by changing only I1 , which allows us to
appreciate the effect of the current on the dynamics of
the improved HR model. Figures 11a and 12a–c show
some curves of the time evolution of the membrane
potential, while Figs. 11b and 12d illustrate some
attractors in the (z–x) plane. Indeed, from these figures,
it is clear that by applying a suitable excitation current
on the HR model subjected to the influence of a sinu-
soidal external electric field. It presents very interesting
phenomena and rich in dynamics describing well the
different electrical activities of neurons such as chaos
state (Fig. 11), quiescent state (Fig. 12a), spiking state
(Fig. 12b) and finally bursting state (Fig. 12c, d). Given
all these results, it can be said that the analog simula-
tions are in agreement with the theoretical/numerical
results obtained previously.

123
 Author's personal copy
Cognitive Neurodynamics

Finite-time synchronization

Problem statement

It is known today that the extracellular electric field is

Fig. 7 Bifurcation diagram showing the presence of large windows of
hysteretic dynamics. This region corresponds to values of k1 in
the range:0  k1  0:00001799 with I2 ¼ 0:02;f2 ¼ 0:09; k2 ¼
0:001 f1 ¼ 0:01 and I1 ¼ 0:74. The direction of the green arrow
indicates that green bifurcation diagram is obtained by incrementing
the control parameter k1 from 0 to 0.000017999 with an integration
step of 10-8 (upward bifurcation) while the direction of the blue
arrow indicates that blue bifurcation diagram is obtained by
decrementing the same control parameter k1 from 0.000017999 to 0
with an integration step of - 10-8 (backward bifurcation). Note that
this diagram illustrating the zone in which the system exhibits
possible coexistence of different chaotic attractors
continuously present in the living nervous system and may
have an impact on the synaptic force, which ensures the
connection between the neurons. Indeed, the synaptic link
plays an important role in the exchange of energy and the
propagation of signal between neurons. Also, the chemical
synapse, in particular, plays a significant role in the coding
and synchronization of neurons through the release of the
neurotransmitter. As a first consequence of electrical field
involvement, the synaptic function can be modified
because this field can influence the release of the neuro-
transmitter. Thus, studying synchronization and synchro-
nization patterns in neural networks is very crucial in
understanding neurological activities and is an exciting
topic in neuroscience. Using this as a motivation, in this
part of our work, the goal is to propose the improved HR
model under electric field effect (see Eq. 4) to study the
finite-time synchronization between two neurons when

Fig. 8 Phase orbits (left) and 2

(a) (b)
corresponding frequency spectra 2
1.5 10
(right) showing coexistence of
two different attractors (a
1
chaotic attractors and a period-3
0
limit cycle) for 0.5
10
k1 ¼ 1:11  105 . Initial
PS(y)

conditions 0
x

ðxð0Þ; yð0Þ; zð0Þ; Eð0ÞÞ are -2

10
ð2:7; 7; 2; 4Þ and -0.5
ð4; 0; 5; 0Þ, respectively
-1
-4
10
-1.5

-2 -6
10
0 0.2 0.4 0.6 0.8 0 2 4 6 8 10
z Freq.

2
2 (d) 10
(c)
1.5

1 0
10
0.5
PS(y)

0
X

-2
10
-0.5

-1
-4
-1.5 10

-2
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0 2 4 6 8 10
z Freq.

123
 Author's personal copy
Cognitive Neurodynamics

Fig. 9 Two-parameter phase diagram in the plane ðk2 ; k1 Þ in (a) and irregular oscillation of the neuron corresponds to Yn = 1 while the
in space ðk2 ; k1 ; Yn Þ in (b) showing respectively the region of regular regular oscillation corresponds to Yn = - 1. (Color figure online)
oscillations (blue) and the region of irregular oscillations (red). The

performing the synapse coupling across the membrane addition, finite-time synchronization implies that for all
potential and the electric field coupling, this in bidirec- initial conditions ðe1 ð0Þ; e2 ð0Þ; e3 ð0Þ; e4 ð0ÞÞ, the solution of
tionally way (see the schematic diagram in Fig. 13). the system (29) reaches the origin (e1 ðsÞ ¼ e2 ðsÞ ¼
Considering that the first neuron N1 has a state variable e3 ðsÞ ¼ e4 ðsÞ ¼ 0) after a finite-time ss , that is:
N1 ðx1 ; y1 ; z1 ; E1 ÞT and the second neuron N2 has a state lim keðsÞk ¼ 0 and eðsÞ ¼ 0; 8s  ss : ð30Þ
s!ss
variable N2 ðx2 ; y2 ; z2 ; E2 ÞT , the dynamic equations of these
two neurons under synapse and field coupling are described
by: Synchronization process
8
>
> x_1 ¼ y1  ax31 þ bx21  z1 þ Iext þ g1 ðx2  x1 Þ;
>
> To attain the synchronization objective previously stated,
>
> y_1 ¼ c  dx21  y1 þ k1 E1 ;
>
> z_1 ¼ r ½sðx1 þ hÞ  z1 ; we use the Lyapunov stability theory. The aim here is to
>
>
< _ prove that, the states of the synchronization system con-
E1 ¼ k2 y1 þ Eext þ g2 ðE2  E1 Þ;
ð27Þ verge to zero at a finite horizon. In other words, for the
>
> x_2 ¼ y2  ax32 þ bx22  z2 þ Iext þ g1 ðx1  x2 Þ;
>
> 2 conditions of Eq. (30) to be satisfied, it is essential to find a
>
> y_2 ¼ c  dx2  y2 þ k1 E2 ;
>
>
>
> z_ ¼ r ½sðx2 þ hÞ  z2 ; Lyapunov function necessary to achieve the finite-time
: _2
E2 ¼ k2 y2 þ Eext þ g2 ðE1  E2 Þ; synchronization of the two neurons. This can only be
possible if the Lyapunov function is minimal when all its
where the constants g1 and g2 represent respectively the
variables close to zero. Therefore, we propose the follow-
intensity for synapse coupling and electric field coupling.
ing specific Lyapunov function that satisfies the conditions
We define the synchronization error eðe1 ; e2 ; e3 ; e4 ÞT in the specified in (Zuppa et al. 2002; Louodop et al.
following way: 2013, 2014a, b; Paden Brad and Shankar 1987):
e ¼ N2  N1 ð28Þ   Z s
1 2 2 e23 2
V¼ e þ e2 þ þ e4 þ e24 ds: ð31Þ
Thus, we can easily deduce the system below, which 2 1 r 0
makes it possible to describe the dynamics of the syn- Deriving Eq. (31) with respect to time and using
chronization error: Eq. (29) we obtain:
8  3 
>
> e_ ¼ e2   a2 x2  x31 þ b x22  x21  e3  2g1 e1 ;
< 1   
e_2 ¼ d x2  x21  e2 þ k1 e4 ;
ð29Þ V_ ¼ e2  a x32  x31 þ b x22  x21  e3  2g1 e1 e1
>
> e_ ¼ r ðse1  e3 Þ;  
: 3 þ d x22  x21  e2 þ k1 e4 e2
e_4 ¼ k2 e2  2g2 e4 :
The problem of synchronization between the two neu- þ ðse1  e3 Þe3 þ ðk2 e2  2g2 e4 Þe4 þ e24  e24 ð0Þ:
rons now amounts to ensuring that the error dynamics Taking into account the properties of the Lipschitz func-
described by Eq. (29) are asymptotically stable. In tions set out in (Khalil 2007; Ngouonkadi et al. 2014; Cho

123
 Author's personal copy
Cognitive Neurodynamics

Fig. 10 Analog circuit of the R

improved Hindmarsh–Rose R2
neuron model under electric
field effect
MULT
R3
R
Cx
R
Iext R5
MULT R

-x
R4 R10
R12
R R7
Cz
R
R1 R11

R8
R
Cy
R
R6

R13
R9
CE

-Eext R14

!
and Rajamani 1997), we suppose that there are two positive ð s  1 Þ 2
constants L1 and L2 such that: V_  aL1 þ bL2  g1 þ e21
 3    4
x  x3   L1 je1 j and x2  x2   L2 je1 j: !
2 1 2 1
It follows that ð1 þ dL2 Þ2 ðk1 þ k2 Þ2
þ þ  1 e22 þ ð1  g2 Þe24  e24 ð0Þ
4g1 4g2
   
ð1 þ dL2 Þje2 j 2 ðs  1Þje1 j 2
 g1 e 1   e3 
2g1 2
 2
ðk 1 þ k 2 Þj e 2 j
 g2 e 4  :
2g2

At this level, we assume that the intensities of synapse

coupling and electric field coupling obey the following
conditions:

123
 Author's personal copy
Cognitive Neurodynamics

Fig. 11 PSpice simulation of: time series (a) and two dimensional views (b) of the attractor projected, illustrating the complexity of the system
for I1 ¼ 0:745; f1 ¼ 0:01; k1 ¼ 0:000085; I2 ¼ 0:02; f2 ¼ 0:09 and k2 ¼ 0:001

8
>
>
> ð s  1Þ 2 Equation (34) allows us to know the maximum time
>
> g 1  aL 1 þ bL 2 þ ;
< 4 ! before which there is synchronization between the two
2
1 ð1 þ dL2 Þ ðk1 þ k2 Þ2 modified HR systems. Finite-time synchronization is sat-
>
> 1 þ
>
> 4 g1 g2 isfied when the numerical synchronization time ssNU is less
>
:
g2  1; than or equal to the theoretical synchronization time ssTH
ð s  1Þ 2 (Louodop et al. 2013, 2014a).
g1  aL1 þ bL2 þ ;
4 !
) ð32Þ Numerical results
1 ð1 þ dL2 Þ2 ðk1 þ k2 Þ2
g2  þ ;
4 g1 g2
In order to demonstrate the feasibility of the proposed
when these conditions are satisfied, we have synchronization method, numerical simulations are pre-
sented in this part of the work. These simulations are
V_   e24 ð0Þ: ð33Þ
performed considering the values of the system parameters
For Eq. 33, we can say that the finite-time convergence for which the neurons have a chaotic behavior (see Fig. 5)
of Eq. (29) is satisfied, according to Theorem 2 stated in and thus, the initial conditions N1 ðx1 ð0Þ; y1 ð0Þ; z1 ð0Þ;E1 ð0ÞÞ
(Paden Brad and Shankar 1987). and N2 ðx2 ð0Þ; y2 ð0Þ; z2 ð0Þ; E2 ð0ÞÞ are the following
However, to estimate the theoretical finite-time syn- N1 ð2; 5; 0:8; 1Þ and N2 ð2; 5; 2:26; 1Þ. Using these
chronization between the two neurons, we integrate parameter values and these initial conditions, the theoreti-
Eq. (33) from 0 to ss , we obtain: cal time for finite-time synchronization is ssTH ¼ 210:075.
V ðss Þ  V ð0Þ   e24 ð0Þss : Figure 14a presents the time evolution of the synchro-
nization error between the membrane potentials of the
Knowing that V ðss Þ ! 0 when s ! ss , we have neurons N1 and N2 ; it can be noticed that after a numerical
V ð0Þ  V ðss Þ  V ð0Þ   e24 ð0Þss ; synchronization time ssNU ¼ 150, the two neurons evolve
identically. The same observation can be made for the
which implies that other variables (see Fig. 14b–d) where we see that each
V ð 0Þ time, the condition ssNU  ssTH is verified.
ss  ; Furthermore, one of our major concerns in this work is
e24 ð0Þ
  ð34Þ to be able to quantitatively show the effect of some
1 2 2 e23 ð0Þ 2 important parameters on the coupling dynamics of the
¼ e ð 0 Þ þ e ð 0 Þ þ þ e ð 0 Þ :
2e24 ð0Þ 1 2
r 4

123
 Author's personal copy
Cognitive Neurodynamics

Fig. 12 PSpice simulation of the time series of membrane potential in neuron under different external current at f1 ¼ 0:01; k1 ¼
0:000085; I2 ¼ 0:02; f2 ¼ 0:09; k2 ¼ 0:001, for a I1 ¼ 0:01; b I1 ¼ 0:8; c I1 ¼ 2 with the initial condition ð0; 0; 0; 0Þ. d Present the two
dimensional views of the attractor illustrating the complexity of the system when I1 ¼ 2

Fig. 13 Schematic diagram for

two bidirectionally coupled
neurons under electric field
effect

N1 N2

Electric field coupling

123
 Author's personal copy
Fig. 14 Time histories of error (a)
signals
(c)
mHR model of Eq. (27). For this, we define the
following synchronization error norms e n ð sÞ ¼
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
e21 ðsÞ þ e22 ðsÞ þ e23 ðsÞ þ e24 ðsÞ. In Fig. 15a, it can be seen
that for low values of synapse coupling intensity
(0  g1  0:4), the behavior of the two coupled neurons
cannot be identical (desynchronization). It can be seen
from Fig. 15b that the influence of the variation of the
intensity of the electric field coupling on the synchroniza-
tion of these two neurons is very small. In addition, the
change in the amplitude of the external excitation current I1
(see Fig. 15c) or that of the electric field parameter k1 (see
Fig. 15d) do not have a significant impact after a time
(which can be considered as transient period) of s ¼ 150:
By adjusting the amplitude (Fig. 15e) and the frequency
(Fig. 15f) of the external electric field Eext to appreciate
their impact on the synchronization of membrane potentials
x1 and x2 ; it is noted that, just as the effect of electric field
coupling, these impacts are very small.
Pspice simulation
One of the significant challenges in the study of non-linear
biological systems is to be able to physically implement the
different results obtained through theoretical and numerical
studies (Wu et al. 2019). When these results are confirmed
experimentally, they can be used in medical engineering to
solve specific problems that are encountered in everyday
123
Cognitive Neurodynamics
(b)
(d)
life. This is the case, for example, of cardiac simulators
(pacemakers) which are manufactured by the electronic
circuit and which allow the patient’s heart to have normal
electrical activity (ECG: electrocardiogram) via the syn-
chronization phenomenon between the pacemaker and the
sick heart (Lopez et al. 2015). Then, the objective here is to
implement an electronic circuit that can provide the syn-
chronization strategy proposed above to verify the effi-
ciency and practical feasibility of this synchronization. In
order to electronically implement the schematic diagram of
Fig. 13, we consider two identical neurons whose circuit of
each corresponds to that of Fig. 10. Thus, we use the values
of the parameters defined in the subsection above to
determine using the Eq. (26), the values of the equivalent
electronic components. These values of the components are
the same for their respective circuits. To realize the elec-
tronic circuit of each coupling (synapse coupling and
electrical coupling) between these neurons, we use as
analog tools an inverter summing assembly whose output is
connected to an inverting operational amplifier (see
Fig. 16).
From this figure, we observe two blocks surrounded by
red dashed lines; these blocks represent the synapse cou-
pling and the electric field coupling between the two neu-
rons. Moreover, from Fig. 16, we deduce the values of the
intensities of synapse coupling (g1 ) and the electric field
coupling (g2 ) as follows:
 Author's personal copy
Cognitive Neurodynamics

Fig. 15 Time evolution of error norm when varied: a the intensity of k1 , e the amplitude of the external electric field I2 , f the frequency of
synapse coupling g1 , b the intensity of the electric field coupling g2 , c the external electric field f2
the amplitude of the external forcing current I1 , d the feedback gain

123
 Author's personal copy
Cognitive Neurodynamics

Fig. 16 Analog circuit for two coupled mHR neurons with synapse coupling and electric field coupling. (Color figure online)

g1 ¼
Rg1
and g2 ¼
Rg2
: ð35Þ ts ¼ RCssNU ¼ 104 ssNU with R ¼ 10 kX and C ¼ 10 nF:
R R ð36Þ
It is therefore evident that in this circuit (see Fig. 16) it
PSpice simulations results illustrating the finite-time
is easy to adjust the intensity of the couplings g1 and g2 by
synchronization of the two coupled neurons are presented
merely modifying the values of the resistances Rg1 and Rg2
in Fig. 17. From these results (see Fig. 17a–d), it can be
respectively (since R ¼ 10 kX). Recall that, the relation-
seen that the synchronization time obtained after the
ship between the synchronization time ssNU obtained pre-
PSpice simulation is always less than or equal to the the-
viously thanks to the numerical simulation and that of the
oretical synchronization time obtained previously
synchronization time ts obtained thanks to the PSpice
(ts  ssTH ). Therefore, we can say through these PSpice
simulation is given by:
simulations that the synchronization method/strategy

123
 Author's personal copy
Cognitive Neurodynamics
Fig. 17 Time histories of error signals in Pspice
presented throughout this part of our work is physically
feasible.
Discussion on the synchronization feature
Recent research has confirmed that electric/magnetic field
coupling is very effective in achieving signal propagation
between neurons (Ren et al. 2019). However, Ma et al.
relevant work has shown that, under certain conditions,
field coupling can create synchronization or desynchro-
nization between coupled chaotic oscillators (Ma et al.
2018). Knowing this, one of our objectives in this contri-
bution was to investigate the effect of the variation of the
electrical coupling intensity on the signal propagation
between the neurons. Indeed, in (Antonopoulos et al.
2019), it is clearly explained that the brain is a network of
billions of neurons, and that these neurons are intercon-
nected through synaptic and electrical couplings that allow
them to exchange and code the information. Hence, in this
contribution, we focus on the influence of the variation of
synapse coupling (g1) on the membrane potential and the
electrical coupling (g2) on the electric field. In regards to
the electrical coupling g2, we quantitatively evaluate
through Fig. 15b the impact that this coupling can have on
the synchronization of two neurons when its intensity is
modified. As mentioned previously, pathologies such as
epilepsy, schizophrenia, Alzheimer’s disease, Parkinson’s
disease, autism and many others, occur when coding and
transmission of signal/information between coupled neu-
rons does not occur normally, which leads to a desyn-
chronization between the neurons (Uhhaas and Singer
2006). Knowing this, we show thanks to Fig. 15b that, for
the proposed model, electrical coupling favors the trans-
mission and coding of information because, although it is
another link between neurons, it does not prevent them
from having the same behavior (i.e. they synchronize)
when its intensity is changed. Conversely, we show that it
is rather the synapse coupling that can, for some values of
its intensity create a desynchronization which could prob-
ably have caused pathologies (see Fig. 15a). Since the
electric field and the magnetic field are both plane waves,
we deduce that these results are in accordance with those
123
 Author's personal copy
obtained in (Parastesh et al. 2019) where the authors
studied the synchronization phenomenon between the
neurons by realizing synaptic coupling and magnetic cou-
pling (ephaptic coupling). As a result, the electric field may
be another effective bridge connection to promote the
exchange and coding of the signal.
Conclusion
One of the main challenges in the modeling of neurological
systems is to be able to consider all the parameters that can
help to establish a perfect match between the real model
and the mathematical model. Using this as a motivation, we
considered in this work the Hindmarsh–Rose neuron model
proposed in 1984 where we introduced a new variable E
that describes the effect of the electric field created by the
fluctuation of the membrane potential. A basic dynamic
analysis of the model obtained allowed us to highlight the
rich and interesting phenomena of electrical activity. We
prove that Hopf bifurcation occurs in this system when the
parameters are chosen appropriately. A hysteretic dynamic
has also been observed, which justifies the coexistence of
two attractors in the proposed model. In reality, neurons do
not evolve solitarily, so we performed a synapse coupling
through the membrane potential and an electric field cou-
pling between two neurons. It appears from the dynamic
study of this coupling that the effect of the variation of the
intensity of the electric field coupling on the synchroniza-
tion of the two neurons is almost negligible whereas that of
the intensity of the synapse coupling is very significant. As
a first consequence, it can be deduced that the electric field
may be another effective bridge connection to encourage
the exchange and coding of the signal. In order to validate
experimentally the theoretical/numerical results obtained,
we presented PSpice simulations which made it possible to
show the feasibility of our working hypotheses.
References
Antonopoulos CG, Martinez EB, Baptista MS (2019) Evaluating
performance of neural codes in model neural communication
networks. Neural Netw 109:90–102
Bao BC, Jiang P, Wu HG, Hu FW (2015) Complex transient
dynamics in periodically forced memristive chua’s circuit.
Nonlinear Dyn 79:2333–2343
Bao BC, Hu A, Xu Q, Bao H, Hu W, Chen M (2018) AC-induced
coexisting asymetric bursters in the improved Hindmarsh–Rose
model. Nonlinear Dyn 92:1695
Boccaletti S, Latora V, Moreno Y, Chavez M et al (2006) Complex
networks: structure and dynamics. Phys Rep 424:175–308
Boccaletti S, Bianconi G, Criado R, Wang Z, Zanin M et al (2014)
The structure and dynamics of multilayer networks. Phys Rep
544:1–122
123
Cognitive Neurodynamics
Cho YM, Rajamani R (1997) A systematic approach to adaptive
observer synthesis for nonlinear systems. IEEE Trans Autom
Control 42:534–537
Djeundam SRD, Yamapi R, Kofane TC, Azizalaoui MA (2013)
Deterministic and stochastic bifurcations in the Hindmarsh–Rose
neuronal model. Chaos 23:033125
Dong J, Zhang GJ, Xie Y, Yao H, Wang J (2014) Dynamic behavior
analysis of fractional-order Hindmarsh–Rose neuronal model.
Cogn Neurodyn 8:167–175
Estrada E (2012) The structure of complex networks: theory and
applications. Oxford University Press, Oxford
Fitzhugh R (1961) Impulses and physiological states in theoretical
models of nerve membrane. Biophys J 1:445
Fitzhugh R (1969) Mathematical models of excitation and propaga-
tion in nerve. In: Schwan HP (ed) Biological engineering. Mc
Graw-Hill, New-York
Ge M, Jia Y, Xu Y, Yang L (2018) Mode transition in electrical
activities of neuron driven by high and low frequency stimulus in
the presence of electromagnetic induction and radiation. Non-
linear Dyn 91:515–523
González-Miranda JM (2007) Complex bifurcation structures in the
Hindmarsh–Rose neuron model. Int J Bifurc Chaos
17:3071–3083
Gu HG, Pan BB, Chen GR, Duan LX (2014) Biological experimental
demonstration of bifurcations from bursting to spiking predicted
by theoretical models. Nonlinear Dyn 78:391–407
Guckenheimer J, Holmes P (1983) Nonlinear oscillations, dynamical
systems and bifurcation of vector field. Springer, New York
Han C, Yu S, Wang GA (2015) Sinusoidally driven Lorenz system
and circuit implementation. Math Prob Eng 2015:706902
Hindmarsh JL, Rose RM (1982) A model of the nerve impulse using
two first-order differential equations. Nature 296:162–164
Hindmarsh JL, Rose RM (1984) A model of neuronal bursting using
three coupled first order differential equations. Proc R Soc Lond
B Biol Sci 221:87–102
Hodgkin AL, Huxley AF (1952) A quantitative description of
membrane current and its application to conduction and exci-
tation in nerve. J Physiol 117:500–544
Innocenti G, Genesio R (2009) On the dynamics of chaotic spiking–
bursting transition in the Hindmarsh–Rose neuron. Chaos
19:023124
Innocenti G, Morelli A, Genesio R, Torcini A (2007) Dynamical
phases of the Hindmarsh–Rose neuronal model: studies of the
transition from bursting to spiking chaos. Chaos 17:043128
Jia C, Wang J, Deng B, Wei X, Che Y (2011) Estimating and
adjusting abnormal networks with unknown parameters and
topology. Chaos 21:013109
Kaslik E (2017) Analysis of two- and three-dimensional fractional-
order Hindmarsh–Rose type neuronal models. Frac Calc Appl
Anal 20:623–645
Kengne J, Chedjou JC, Kenne G, Kyamakya K, Kom GH (2012)
Analog circuit implementation and synchronization of a system
consisting of a van der Pol oscillator linearly coupled to a
Duffing oscillator. Nonlinear Dyn 70:2163–2173
Kengne J, Jafari S, Njitacke ZT, Yousefi AK, Cheukem A (2017)
Dynamic analysis and electronic circuit implementation of a
novel 3D autonomous system without linear terms. Commun
Nonlinear Sci Numer Simul. https://doi.org/10.1016/j.cnsns.
2017.04.017
Khalil HK (2007) Nonlinear systems, 3rd edn. Prentice Hall, Upper
Saddle River
Kivelä M, Arenas A, Barthelemy M, Gleeson JP, Moreno Y, Porter
MA (2014) Multilayer networks. J Complex Netw 2:203–271
Kuznetsov YA (1998) Elements of applied bifurcation theory.
Springer, New York
 Author's personal copy
Cognitive Neurodynamics
Letellier C, Denis F, Aguirre LA (2013) What can be learned from a
chaotic cancer model ? J Theor Biol 322:7–16
Lopez MJ, Consegliere A, Garcia L, Lorenzo J (2015) Simulation and
control of heart rhythm dynamics. Adv Biomed Res 1:509–516
Louodop P, Fotsin H, Kountchou M, Bowong S (2013) Finite-time
synchronization of Lorenz chaotic systems: theory and circuits.
IOP Sci 88:045002
Louodop P, Fotsin H, Kountchou M, Ngouonkadi LBM, Cerdeira HA,
Bowong S (2014a) Finite-time synchronization of tunnel-diode-
based chaotic oscillators. Phys Rev E 89:032921
Louodop P, Kountchou M, Fotsin H, Bowong S (2014b) Practical
finite-time synchronization of Jerk systems: theory and exper-
iment. Nonlinear Dyn 78:597
Lu L, Jia Y, Liu W, Yang L (2017) Mixed stimulus-induced mode
selection in neural activity driven by high and low frequency
current under electromagnetic radiation. Complexity
7628537:1–11
Lv M, Ma J (2016) Multiple modes of electrical activities in a new
neuron model under electromagnetic radiation. Neurocomputing
205:375–381
Lv M, Wang CN, Ren GD, Ma J (2016) Model of electrical activity in
a neuron under magnetic flow effect. Nonlinear Dyn
85:1479–1490
Ma J, Xu Y, Wang CN, Jin WY (2016) Pattern selection and self-
organization induced by random boundary initial values in a
neuronal network. Phys A 461:586–594
Ma J, Wu F, Wang C (2017) Synchronization behaviors of coupled
neurons under electromagnetic radiation. Int J Mod Phys B
31:1650251
Ma J, Fuqiang W, Ahmed A, Jun T (2018) Crack synchronization of
chaotic circuits under field coupling. Nonlinear Dyn. https://doi.
org/10.1007/s11071-018-4307-x
Ma J, Zhang G, Hayat T, Ren GD (2019) Model electrical activity of
neuron under electric field. Nonlinear Dyn 95:1585
Mondal A, Upadhyay RK, Ma J et al (2019) Bifurcation analysis and
diverse firing activities of a modified excitable neuron model.
Cogn Neurodyn 13:393–407
Morris C, Lecar H (1981) Voltage oscillations in the barnacle giant
muscle fiber. Biophys J 35:193–213
Negou AN, Kengne J (2018) Dynamic analysis of a unique jerk
system with a smoothly adjustable symmetry and nonlinearity:
reversals of period doubling, offset boosting and coexisting
bifurcations. Int J Electron Commun (AEÜ) 90:1–19
Ngouonkadi EB, Fotsin HB, Louodop F (2014) Implementing a
memristive Van der Pol oscillator coupled to a linear oscillator:
synchronization and application to secure communication. IOP
Sci 89:035201
Ngouonkadi EBM, Fotsin HB, Fotso PL, Tamba VK, Cerdeira HA
(2016) Bifurcations and multistability in the extended Hind-
marsh–Rose neuronal oscillator. Chaos, Solitons Fractals
85:151–163
Njitacke ZT, Kengne J (2018) Complex dynamics of a 4D Hopfield
neural networks (HNNs) with a nonlinear synaptic weight:
coexistence of multiple attractors and remerging Feigenbaum
trees. Int J Electron Commun (AEÜ) 93:242–252
Njitacke ZT, Kengne J, Negou AN (2017) Dynamical analysis and
electronic circuit realization of an equilibrium free 3D chaotic
system with a large number of coexisting attractors. Optik
130:356–364
Njitacke ZT, Kengne J, Fonzin FT, Leutcha PB, Fotsin HB (2018)
Dynamical analysis of a novel 4-neurons based Hopfield neural
network: emergences of antimonotonicity and coexistence of
multiple stable states. Int J Dyn Control. https://doi.org/10.1007/
s40435-019-00509-w
Paden Brad E, Shankar Sastry (1987) A calculus for computing
filippov’s differential inclusion with application to the variable
structure control of robot. IEEE Trans Circuit Systems 35:73–82
Panahi S, Aram Z, Jafari S, Ma M, Sprott JC (2017) Modeling of
epilepsy based on chaotic artificial neural network. Chaos
Solitons Fractals 105:150–156
Parastesh F, Azarnoush H, Jafari S et al (2019) Synchronizability of
two neurons with switching in the coupling. Appl Math Comput
350:217–223
Perc M (2009) Optimal spatial synchronization on scale-free networks
via noisy chemical synapses. Biophys Chem 141:175–179
Ren G, Xu Y, Wang C (2017a) Synchronization behavior of coupled
neuron circuits composed of memristors. Nonlinear Dyn
88:893–901
Ren GD, Zhou P, Ma J, Cai N, Alsaedi A, Ahmad B (2017b)
Dynamical response of electrical activities in digital neuron
circuit driven by autapse. Int J Bifurc Chaos 27:1750187
Ren G, Xue Y, Li Y, Ma J (2019) Field coupling benefits signal
exchange between Colpitts systems. Appl Math Comput
342:45–54
Rigatos G, Wira P, Melkikh A (2019) Nonlinear optimal control for
the synchronization of biological neurons under time-delays.
Cogn Neurodyn 13:89–103
Shi X, Wang Z (2012) Adaptive synchronization of time delay
Hindmarsh–Rose neuron system via self-feedback. Nonlinear
Dyn 69:21472153
Spitzer NC (2006) Electrical activity in early neuronal development.
Nature 444:707–712
Strogatz SH, Friedman M, Mallinckrodt AJ, Mckay S (1994)
Nonlinear dynamics and chaos: with applications to physics,
biology, chimestry and engineering. Comput Phys 8(5):532
Uhhaas PJ, Singer W (2006) Neural synchrony in brain disorders:
relevance for cognitive dysfunctions and pathophysiology.
Neuron 52:155–168
Wang Z, Shi X (2020) Electric activities of time delay memristive
neuron disturbed by Gaussian white noise. Cogn Neurodyn
14:115–124
Wiggins S (1990) Introduction to applied nonlinear dynamical
systems and chaos. Springer, New York
Wolf A, Swift JB, Swinney HL, Wastano JA (1985) Determining
Lyapunov exponents from time series. Phys D 16:285–317
Wouapi KM, Fotsin HB, Feudjio KF, Njitacke ZT (2019) Hopf
bifurcation, offset boosting and remerging Feigenbaum trees in
an autonomous chaotic system with exponential nonlinearity. SN
Appl Sci 1:1715
Wu KJ, Luo TQ, Lu HW, Wang Y (2016) Bifurcation study of neuron
firing activity of the modified Hindmarsh–Rose model. Neural
Comput Appl 27:739–747
Wu F, Ma J, Zhang G (2019) A new neuron model under
electromagnetic field. Appl Math Comput 347:590–599
Xu Q, Zhang QL, Bao BC, Hu YH (2017) Non-autonomous second-
order memristive chaotic circuit. IEEE Access 5:21039–21045
Zuppa LA, Hernandez CC, Bustos AYA (2002) Finite synchroniza-
tion of Lorenz-based chaotic systems. www.wseas.us/e-library/
conferences/mexico2002/papers/249.pdf. Accessed 4 April 2017
Publisher’s Note Springer Nature remains neutral with regard to
jurisdictional claims in published maps and institutional affiliations.
123