Chimera State in the Network of Fractional-Order FitzHugh–Nagumo Neurons

Centre for Artificial Intelligence, Chennai Institute of Technology, Malayambakkam, India Department of Biomedical Engineering, Amirkabir University of Technology, No. 350, Hafez Ave, Valiasr Square, Tehran 159163-4311, Iran Centre for Nonlinear Systems, Chennai Institute of Technology, Malayambakkam, India Health Technology Research Institute, Amirkabir University of Technology, No. 350, Hafez Ave, Valiasr Square, Tehran 159163-4311, Iran Department of Mathematics, Statistics and Physics, Qatar University, Doha 2713, Qatar


Introduction
Fractional-order models have attracted much attention from scientists in different fields such as physics and electronics [1][2][3][4][5]. Considering the fractional derivative, a memory feature is added to the systems. erefore, the fractional-order model can provide a more precise description of the real phenomena than the integer-order [6]. Furthermore, the fractional calculus has found wide applications in controlling the integer-order systems [7]. e fractional derivative also plays important role in demonstrating different firing patterns of the neurons [8]. Consequently, several fractional neuron models have been presented [9][10][11]. e complicated interactions among the neurons cause the neural system to act as a complex network [12]. e emergence of collective behaviors is an important characteristic of complex networks [13]. Some examples of collective behaviors are synchronization [14], chimera state [15], and solitary state [16]. Synchronization is an important phenomenon in many applications [17]. Many studies have focused on the synchronization of chaotic systems [18][19][20][21][22][23]. Furthermore, synchronization manages many neural functions and participates in many brain disorders [24]. In special cases, synchrony and asynchrony are observed simultaneously in a specific region of the brain. For example, the unihemispheric sleep, the neural bump state, and the epileptic seizure disease can be mentioned [25]. is particular condition is called the chimera state [26]. After the foundation of chimera state in 2002 [27], it became the focus of many researchers in a variety of dynamical systems such as the mechanical [28], optical [29], and chemical [30] oscillators and neuronal models [25,31,32]. Furthermore, these studies have represented the chimeras with different spatiotemporal patterns and properties, including the amplitude chimera [33] and traveling chimera [34].
In neuronal studies, the chimeras were under consideration from different perspectives such as the neurons' dynamics, network topology, and coupling scheme. Santos et al. [35] investigated the chimera in 2D networks with regular and fractal topologies and found the spiral chimeras with multiple asynchronous cores. Wang et al. [36] reported the existence of chimera state in the hyperchaotic neurons with hyperchaotic dynamics. Blondeau Soh et al. [37] represented that shifting the neighbors in the coupling leads the network towards the chimera state. Provata and Venetis [38] studied a neuronal network with power-law coupling and showed that the chimera exists in the weak couplings with large exponents. Li et al. [39] considered two unidirectionally coupled layers of neurons and showed different collective behaviors in the master layer and the induced imperfect chimera state in the slave layer.
Among the chimera studies in neuronal networks, a few have considered the fractional models. Vázquez-Guerrero et al. [40] showed that the network of fractional Hindmarsh-Rose neurons is capable of representing chimera state. ey also presented a fractional adaptive controller to obtain the synchronization. In another study, they designed an observer to synchronize the chimera state in coupled fractional neurons [41]. He [42] investigated the magnetic Hindmarsh-Rose model with fractional derivative and showed that analyzing the complexity of the network can help in recognition of its dynamical behavior. In this paper, we study the dynamical behaviors of a network of fractionalorder FitzHugh-Nagumo systems. e effects of the order of derivatives and the coupling strength on the chimera state are under consideration.

The Model
Recently, the scientists have focused on proposing new models for describing the neural behaviors with considering different aspects of neurons [43,44]. Here, we use the FitzHugh-Nagumo (FHN) model with considering the fractional derivative as follows [45]: where u and v are the membrane voltage and the recovery variable and I is the external excitation current fixed at 0.5. e fractional derivative order is denoted by q, and d q /dt q is the Caputo-Fabrizio (CF) fractional operator defined by where Γ is the Gamma function. e dynamics of the model relies on the values of the derivative order. Figure 1(a) represents the bifurcation of the model according to q. To consider the spiking firing for the model, the range of 0.7 < q < 1 is selected in all simulations. e time series and phase spaces of the model for q � 1, 0.9, 0.8 , and 0.7 are shown in Figures 1 . It is observed that by decreasing q, the amplitude of the oscillations decreases and the period increases.
We consider the network of fractional FHN neurons with the following equations: where d is the coupling strength and G is the Laplacian matrix of connections. e network has a ring structure with nonlocal coupling as shown in Figure 2 (each neuron is connected to its 40 nearest neighbors, and N � 100). e coupling between variables is through a rotational matrix as follows [46]: with ϕ � (π/2) − 0.1 being the coupling phase.

Results
e network is solved numerically by using the Adam-Bashforth method based on the algorithm proposed in [47] with random initial conditions. To identify different behaviors, the strength of incoherence (SI) is used [48]. To find this measure, at first, the variables are transformed into new ones as x i � u i − u i+1 , i � 1, . . . , N. en, the network is divided into M � N/n groups of n neurons, and the local standard deviation is computed as follows: where θ is the Heaviside function and δ is a threshold set at 0.23, here, and n � 4. e value of SI determines the behavior of the network by SI � 0, 0 < SI < 1, and SI � 1 for synchronization, chimera, and asynchronization, respectively. e network of equation (3) with integer-order (q � 1) represents different dynamical behaviors with varying the coupling strength (d). However, the dynamical changes occur in very small coupling strengths. e patterns of the neurons by varying d values are demonstrated in Figure 3. e left panel represents the space-time plots, and the right 2 Complexity panel represents the time snapshots of the neurons. By increasing the coupling strength, the initial asynchronous pattern of the neurons changes to the chimera state. For d � 10 − 7 , a chimera state is created (shown in Figure 3(a)), and it can be seen that there are synchronous and asynchronous neurons in the network. With an increment in the coupling strength, more neurons locate in the synchronous group. e behavior of the network for d � 1.7 × 10 − 5 is illustrated in Figure 3(b). In this case, most neurons are synchronous, while a few oscillate differently. is behavior is called the solitary state [49,50]. As the coupling becomes stronger, different neurons are attracted to the synchronous group and a complete synchronization is observed (Figure 3(c)). e behavior of the neurons is considerably influenced when the derivative of the network's equations changes to the fractional. To investigate this, the coupling strength is considered to be fixed at d � 2 × 10 − 5 , where the integerorder network shows the solitary state (similar to Figure 3(b)) and the fractional order is changed. With decreasing the fractional order (q), firstly, the network synchronization is enhanced. is is shown in Figure 4(a) for q � 0.9. However, more decrement of q disturbs the synchronous behavior of the network. Figure 4(b) represents the chimera state for q � 0.8 with a similar coupling strength value. When q decreases to q � 0.7, the network becomes completely asynchronous (Figure 4(c)). e waveforms of the neurons in each case are depicted in the right panel of Figure 4.
For different fractional orders, the range of the coupling strength for the appearance of different dynamical behaviors is different. As the fractional order decreases, the chimera Complexity 3 state is formed in higher coupling strengths. Figure 5(a) shows the chimera state in q � 0.9 for d � 4.6 × 10 − 6 . In this case, there are several groups containing a few synchronous neurons. Forq � 0.8, the chimera is observed for d � 3 × 10 − 5 and more neurons are involved in the synchronous cluster ( Figure 5(b)). When q decreases to q � 0.7, in some time intervals, some neurons become synchronous. erefore, the chimera for this derivative is nonstationary. e chimera for q � 0.7 is shown in Figure 5(c). To illustrate the coherent and incoherent clusters better, the local order parameter is computed and shown in the right column of Figure 5. is parameter can be obtained as L k � |1/2p |k− l|≤p exp (jϕ l )|, k � 1, . . . , N, where ϕ l is the geometric phase of lth oscillator calculated by ϕ l � a tan(y l /x l ). e size of the spatial window is denoted by p. When L k � 1, the kth oscillator belongs to a coherent group. Figure 6 represents the strength of incoherence of the network for different fractional orders. For q � 1, the network is in chimera state until d � 5 × 10 − 5 and becomes synchronous for larger coupling strengths (Figure 6(a)). A similar pattern is observed for q � 0.95 (Figure 6(b)). When q decreases to q � 0.9, the synchronization occurs for very smaller coupling strengths (d � 2.2 × 10 − 5 ). For q � 0.85, the network's dynamical behavior returns to the q � 1 manner. With more decrement of q, a stronger coupling is needed for the synchronization. Figure 6(e) shows that q � 0.8 has the larger chimera region. For q � 0.7 and q � 0.75, a large asynchronization region is

Complexity
observed. e variation of SI according to q is illustrated in Figure 7. It is observed that for low coupling strength, the integer network is in a chimera state. e chimera state is preserved until q � 0.835 , and then all neurons become asynchronous. For strong couplings, the integer-order network is synchronous. With decreasing q, the synchronization remains until q � 0.84. For q < 0.84, the chimera state is formed. However, in the range 0.8 < q < 0.823, the synchronization may appear in the network determined by the initial conditions.

Conclusion
In this paper, a network of coupled fractional-order Fitz-Hugh-Nagumo neurons was studied. e dynamical behavior of the network was investigated under the variation of the coupling strength and the derivative order. e bifurcation diagram of the fractional system with respect to the derivative order revealed that the dynamics of the model is dependent on the fractional order. Consequently, by changing the value of the derivative order, various collective behaviors of the neurons can be found. e integer-order neurons experience asynchronization, chimera, and synchronization with increasing the coupling strength, respectively. In the fractional-order network, decreasing the derivative order for the constant coupling strength, resulted in the lower synchrony level in the network. erefore, the fractional-order network has the same state transition; however, it occurs in higher coupling strengths. us, the chimera or synchronous states appear for stronger couplings. Furthermore, the pattern of the chimera was changed with varying the derivative order. For q � 0.9, some small synchronous clusters were formed, while in q � 0.8, a large  cluster of synchronous neurons was observed. For lower q values (q � 0.7), the position of the synchronous cluster was time-dependent and the nonstationary chimera state appeared.

Data Availability
e data used to support the findings of the study are included within the article.

Conflicts of Interest
e authors declare that there are no conflicts of interest.