A fractional-order locally active memristor is proposed in this paper. When driven by a bipolar periodic signal, the generated hysteresis loop with two intersections is pinched at the origin. The area of the hysteresis loop changes with the fractional order. Based on the fractional-order locally active memristor, a fractional-order memristive system is constructed. The stability analysis is carried out and the stability conditions for three equilibria are listed. The expression of the fractional order related to Hopf bifurcation is given. The complex dynamical behaviors of Hopf bifurcation, period-doubling bifurcation, bistability and chaos are shown numerically. Furthermore, the bistability behaviors of the different fractional order are validated by the attraction basins in the initial value plane. As an alternative to validating our results, the fractional-order memristive system is implemented by utilizing Simulink of MATLAB. The research results clarify that the complex dynamical behaviors are attributed to two facts: one is the fractional order that affects the stability of the equilibria, and the other is the local activeness of the fractional-order memristor.
National Natural Science Foundation of China11602035114021256160106261801054Natural Science Foundation of Jiangsu ProvinceBK201914511. Introduction
Nonlinear electronic circuits provide powerful and analytical platforms for people to realize and understand the complex dynamical behaviors in physics [1]. Chaotic circuits especially have become effective tools for studying chaos theory. The memristor, originally defined as the forth element of the circuit by Chua in 1971 [2], is a nonlinear circuit device besides the nonlinear resistor, capacitor, and inductor. As a result, many novel memristive circuits have been constructed by integrating the memristors with versatile nonlinearities into some existing linear or nonlinear circuits [3–11]. In these memristive circuits, rich dynamical behaviors have been reported and tested by numerical simulations and hardware experiments, such as chaos and hyperchaos [12, 13], hyperchaotic multiwing attractors [14, 15], coexisting multiple attractors [16, 17], hidden attractors [18, 19], and complex transient chaos and hyperchaos [20]. It should be noted that the simplest chaotic circuit has been proposed based on a locally active nonlinear memristive element [4]. Compared to the chaotic circuit shown in [21], the simplest chaotic circuit has following characteristics: (1) the circuit components are connected in a single way, i.e., in series; (2) the number of the circuit components is decreased from four to three; (3) the memristor is locally active.
At a given moment, the resistance of an ideal memristor is represented by the integration of all states before the current moment. This means that the ideal memristor has no memory loss. But the work [22] shows that the width of the doped layer of the HP TiO_{2} linear model cannot be equal to zero or the whole width of the model. The HP TiO_{2} linear memristor has memory loss. From the definition, the fractional-order derivative depends on the previous history of the variable and is not a strictly local operator [23]. The order of the fractional-order derivative is related to the memory loss or the “proximity effect” of some characteristics [12]. Then the nonideal memristor with memory loss mentioned in [22] can be modeled by a fractional-order derivative with the order between 0 and 1 [24]. According to this, there are many memristors modeled with the fractional-order derivative [25–27]. As shown in [24], the fractional-order memristor in the series circuits has capacitive properties or inductive properties by choosing a suitable fractional order; i.e., the fractional order can be regarded as a parameter which is used to control the memory strength and dynamics of the circuit. In [25], the fractional order can be used to control the time period in which the resistance of the memristor increases from the initial value to its maximum. In addition, a noncommensurate fractional-order autonomous memristor-based circuit is proposed in [27], where the chaotic behavior can be suppressed by applying periodic impulses. In addition, the dynamical system with the locally active equipment can exhibit complexity and emergent behaviors [28, 29]. Then, it is significant to model the memristor or locally active memristor with the fractional-order derivative and display the dynamics induced by these fractional-order memristors.
The main purpose of this paper is to study the complex dynamical behaviors of a fractional-order system based on a locally active voltage-controlled memristor. By theoretical analyses, the stability conditions of the fractional-order memristive system are listed. The complex dynamical behaviors, such as Hopf bifurcation, period-doubling bifurcation, bistability, and chaos, are displayed numerically. The rest of the paper is organized as follows. In Section 2, the mathematical model of the fractional-order memristor is presented. The fractional-order memristor’s fingerprints and local activeness are addressed. In Section 3, an integer-order locally active memristive system is generalized into a fractional-order locally active memristive system. The stability conditions are listed. The complex dynamical behaviors are stated, and numerical simulations are displayed. As an alternative to validating the numerical results, the fractional-order memristive system is implemented by utilizing Simulink of MATLAB. In Section 4, the effect of the local activeness on complex dynamical behaviors is stated. Section 5 ends with some concluding remarks.
2. Fractional-Order Locally Active Nonlinear Memristor2.1. The Model of the Fractional-Order Memristor
Generally, the memristor can be seen as a sliding resistor whose resistance changes with the charge crossing it. Driven by a bipolar periodic signal, the memristor exhibits a hysteresis loop pinched at the origin in the current-voltage plane. An integer-order nonlinear voltage-controlled memristor is stated as follows [30]:(1)i=Wxmv=xm2−xm−1v,x˙m=p1xm−p2xm3+p4v−p4xmv,where v and i are the voltage and current of the memristor, respectively, xm is the internal state of the memristor and Wxm=xm2−xm−1 is the memductance, and p_{1}, p_{2}, p_{3}, and p_{4} are the system parameters. By using the trial and error method [30], the parameters are decided as p_{1} = 1.8, p_{2} = 3.9, p_{3} = 1.4, and p_{4} = 1.5. Considering the memory effect from the memristor, a fractional-order voltage-controlled memristor M_{α} corresponding to (1) is modeled as follows:(2)i=Wαxmv=xm2−xm−1v,D0Ctαxm=1.8xm−3.9xm3+1.4v−1.5xmv,where(3)D0Ctαxmt=1Γ1−α∫0tx˙mτt−ταdτ,0<α≤1,is α-order derivative of xmt in the sense of Caputo’s definition given in [23], x˙mτ denotes the first-order derivative of xmτ with respect to τ, and Wαxm=xm2−xm−1 is the memductance of α-order memristor M_{α}. The integral process in (3) is the memory process of the memristor. In the circuits, the proposed fractional-order memristor is marked as Figure 1(a).
(a) The fractional-order memristor M_{α}; (b) the fractional-order memristive circuit.
2.2. The Characteristics of the Fractional-Order Memristor
Driven by a sinusoidal voltage source vt=sinωt, hysteresis loops generated in the current-voltage plane are plotted numerically in Figure 2, where ω is the stimulus frequency. One has the following:
Under different stimulus frequencies or different fractional orders, the hysteresis loops of the fractional-order memristor are pinched at the origin.
The larger the area of the hysteresis loop, the stronger the memory [31]. Let the order α = 0.98. Figure 2(a) shows that the smaller the stimulus frequency ω, the stronger the memory. As ω = 1 rad/s, there is another intersection in the hysteresis loop besides the origin and another area S∗ is displayed. Currently, there are few reports on the new intersection which reflects the nonlinearity of the memristor.
As fixing ω = 1 rad/s and decreasing α from 1 to 0, the area of the hysteresis loop increases; i.e., the strength of the memory increases, referring to Figure 2(b). Simultaneously, the quadrants which the hysteresis loops lie in change from II and IV to II, III, and IV.
vt versus it of the fractional-order memristor. (a) The order α = 0.98 with two different frequencies; (b) the frequency ω = 1 rad/s with three different fractional orders.
2.3. Local Activeness of the Fractional-Order Memristor
A component being capable of providing a power gain is called an active component. If the component provides the power gain within the local range of its variables, the component is locally active.
Based on (2), as 1−5/2<xm<1+5/2, one has Wαxm=xm2−xm−1<0, the power p=v⋅i=Wαxm⋅v2<0, and the memristor can provide the power gain; as xm<1−5/2 or xm>1−5/2, one has Wαxm=xm2−xm−1>0, the power p=v⋅i=Wαxm⋅v2>0, and the memristor cannot provide the power gain. So the memristor is locally active.
The above statement implies that the local activeness of the memristor can be decided by the sign of the memductance Wαxm. Referring to Figure 2, the slope of the hysteresis loop is the memductance. Obviously, the changing of the quadrants of the hysteresis loops changes the sign of the slope or the sign of the memductance Wαxm. Remembering the characteristics shown in Figure 2(b), it is easy to know that the fractional order has influences on the activeness of the memristor.
3. Fractional-Order Memristor-Based System
Besides the memristor can be modeled by the fractional-order derivative, the capacitor and the inductor can also be modeled by the fractional-order derivative due to the memory effect [12, 32]. With the fractional-order locally active memristor, a fractional-order memristive circuit is generalized from an integer-order memristive circuit [30], as shown in Figure 1(b), which is modeled by Caputo’s fractional-order derivative(4)D0CtαvC=−axm2−xm−1vC+iL,D0CtαiL=bvC,D0Ctαxm=1.8xm−3.9xm3+1.4vC−1.5xmvC,where 0 < α ≤ 1, a=1/C>0, b=1/L>0, and C, L, vC, and i_{L} are the capacitance, inductance, capacitor voltage, and inductor current, respectively. Letting α = 1, model (4) is changed into the integer-order model stated in [30].
Denoting x=vC, y = i_{L}, and z = x_{m}, (4) is converted into the dimensionless form(5)D0Ctαx=−az2−z−1x+y,D0Ctαy=bx,D0Ctαz=1.8z−3.9z3+1.4x−1.5zx.
For any values of the parameters a and b, system (5) has three equilibria: E_{1} = (0, 0, −0.6794), E_{2} = (0, 0, 0), and E_{3} = (0, 0, 0.6794). For simplicity, the three equilibria are denoted uniformly by Ez0, where Ez0 = (0, 0, z_{0}). Thus, Ez0 = E_{1} as z_{0} = −0.6794, Ez0 = E_{2} as z_{0} = 0, and Ez0 = E_{3} as z_{0} = 0.6794.
3.1. The Stability of the Equilibria
The Jacobian matrix of system (5) at Ez0 is(6)JE=−az02−z0−1−a0b001.4−1.5z001.8−11.7z02.
The characteristic polynomial equation of system (5) at Ez0 is yielded as(7)λ−1.8+11.7z02λ2+az02−z0−1λ+ab=0,which indicates that roots of (7) depend on the three equilibria. The roots of (7) are called the eigenvalues of Jacobian matrix JE. The following lemma is needed.
The fractional-order nonlinear system(8)D0CtαX=fX,0<α≤1,is asymptotically stable at the equilibrium E = (x_{0}, y_{0}, z_{0}) if all eigenvalues λ of Jacobian matrix JE satisfy the condition(9)argλ>απ2,where X=x,y,zT, fX=f1X,f2X,f3XT, fiX=fix,y,zi=1,2,3, and argλ is the principal argument of the eigenvalue λ.
Obviously, equation (7) has a real root λ1=1.8−11.7z02. Considering the sign of λ_{1}, two cases are discussed hereinafter.
Case 1.
λ_{1} is positive
If λ_{1} > 0, one has z0<0.3922 and the equilibrium Ez0 is unstable. Besides the positive root λ_{1}, equation (7) has another two roots:(10)λ2,3=−az02−z0−1±a2z02−z0−12−4ab2.
It is easy to know that −0.454<z02−z0−1<−1.2384 as z0<0.3922. Two cases are listed:
λ_{2} and λ_{3} are the real roots and z02−z0−12≥4b/a holds. Based on (10), one has λ_{1} > 0, λ_{2} > 0, and λ_{3} > 0.
λ_{2} and λ_{3} are the complex roots; i.e., z02−z0−12<4b/a holds. The real parts of the conjugate complex roots are Reλ2,3=−az02−z0−1/2>0.
Case 2.
λ_{1} is negative
As λ1=1.8−11.7z02<0, one has z0>0.3922. Similar to Case 1, two cases are stated as follows:
λ_{2} and λ_{3} are the real roots. Then z02−z0−12≥4b/a holds and
(11)sgnλ2=sgnλ3=sgn−az02−z0−1,
where sgn⋅ is the symbolic function. Due to
(12)gz0=z02−z0−1=z0−1−52z0−1+52,
one has
(13)−az02−z0−1<0,z0<−0.618 or z0>1.618,>0,−0.618<z0<−0.3299 or 0.3299<z0<1.618.
where 1−5/2≈−0.618 and 1+5/2≈1.618. The inequalities −0.618 < z_{0} < −0.3299 and z_{0} > 1.618 can be neglected because z_{0} of the three equilibria is not in these regions. As z_{0} < −0.618, one has λ_{2} < 0 and λ_{3} < 0. As 0.3229 < z_{0} < 1.618, one has λ_{2} > 0 and λ_{3} > 0.
λ_{2} and λ_{3} are the complex roots. The inequality z02−z0−12<4b/a holds. As z_{0} < −0.618, λ_{2} and λ_{3} are the conjugate complex roots with negative real parts Reλ2,3=−az02−z0−1/2<0. As 0.3229 < z_{0} < 1.618, λ_{2} and λ_{3} are the conjugate complex roots with positive real parts Reλ2,3=−az02−z0−1/2>0.
The above discussion can be concluded in Tables 1 and 2.
Tables 1 and 2 show that E_{1} = (0, 0, −0.6794) is stable and E_{2} = (0, 0, 0) is unstable for any a > 0, b > 0, and any order α∈0,1. But for E_{3} = (0, 0, 0.6794), two cases are stated:
if 1.4831≥4b/a, E_{3} is unstable for any order α because λ_{2} > 0 and λ_{3} > 0;
if 1.4831<4b/a, there are two conjugated complex roots at E_{3} as
(14)λ2,3=Reλ2,3±jImλ2,3=−az02−z0−1±j4ab−a2z02−z0−122,where j2=−1. If argλ2,3>απ/2 holds, one has(15)4ab−1.4831a21.2178a>tanαπ2.
By Lemma 1, E_{3} = (0, 0, 0.6794) is stable as 4ab−1.4831a2/1.2178a>tanαπ/2 holds; E_{3} is unstable as 4ab−1.4831a2/1.2178a>tanαπ/2 does not hold.
Based on the above discussion, the following theorem is established.
The eigenvalues as z02−z0−12≥4b/a.
z_{0}
(−∞, −0.618)
(−0.3299, 0.3299)
(0.3299, 1.618)
λ_{1}
λ_{1} < 0
λ_{1} > 0
λ_{1} < 0
λ_{2}
λ_{2} < 0
λ_{2} > 0
λ_{2} > 0
λ_{3}
λ_{3} < 0
λ_{3} > 0
λ_{3} > 0
Equilibrium
E_{1}
E_{2}
E_{3}
The eigenvalues as z02−z0−12<4b/a.
z_{0}
(−∞, −0.618)
(−0.3299, 0.3299)
(0.3299, 1.618)
λ_{1}
λ_{1} < 0
λ_{1} > 0
λ_{1} < 0
λ_{2}
Reλ2<0
Reλ2>0
Reλ2>0
λ_{3}
Reλ3<0
Reλ3>0
Reλ3>0
Equilibrium
E_{1}
E_{2}
E_{3}
Theorem 1.
For system (5) (α∈0,1), the stabilities of three equilibria E_{1}, E_{2}, and E_{3} are as follows:
Equilibrium E_{1} = (0, 0, −0.6794) is stable for any a > 0, b > 0 and any order α∈0,1;
Equilibrium E_{2} = (0, 0, 0) is unstable for any a > 0, b > 0 and any order α∈0,1;
As 1.4831≥4b/a, E_{3} is unstable for any order α∈0,1; as 1.4831<4b/a, E_{3} is stable if 4ab−1.4831a2/1.2178a>tanαπ/2, and E_{3} is unstable if 4ab−1.4831a2/1.2178a<tanαπ/2.
Remark 1.
For a fractional-order system, at a parameter ε = ε_{0}, a pair of conjugated complex eigenvalues λ_{1,2} satisfy argλ1,2=απ/2 and other eigenvalues are in stable zones. While the parameter ε > ε_{0}, argλ1,2<απ/2, Hopf bifurcation is generated at ε = ε_{0} [34].
For system (5), based on Table 2, at E_{3} = (0, 0, 0.6794), as 1.4831<4b/a holds, one has
λ_{1} < 0;
argλ2,3>απ/2 as α<2/πarctan4ab−1.4831a2/1.2178a;
argλ2,3<απ/2 as α>2/πarctan4ab−1.4831a2/1.2178a. This means that Hopf bifurcation is generated as the order α=2/πarctan4ab−1.4831a2/1.2178a at the equilibrium E_{3}.
Remark 2.
For simplicity, the eigenvalues (λ_{1}, λ_{2}, λ_{3}) are denoted as (γ, σ + jω, σ − jω), where γ, σ, and ω are all real numbers. A saddle-focus point is called a saddle-focus point of index 1 if γ > 0 and σ < 0, and a saddle-focus point is called a saddle-focus point of index 2 if γ < 0 and σ > 0 [35]. As pointed out in [35], the saddle-focus points of index 2 are crucial to the generation of chaotic attractors. Usually, in chaotic systems, scrolls are generated around the saddle-focus points of index 2, and the saddle-focus points of index 1 are responsible only for connecting the scrolls.
3.2. Numerical Illustrations
For better comparisons with the integer-order memristive circuit systems, in this section, the parameter is chosen as a=10/3 [30].
Case 1.
a = 10/3, and b = 10.
In this case, a = 10/3 (C = 300 mF) and b = 10 (L = 100 mH) satisfy 4b/a=12>1.4831. To make E_{3} stable, by Theorem 1, the order α is satisfied as(16)α<2πarctan4ab−1.4831a21.2178a=2πarctan4⋅10/3⋅10−1.4831⋅10/321.2178⋅10/3=0.7713.
Then equilibrium E_{3} = (0, 0, 0.6794) is stable as the order α < 0.7713, and E_{3} is unstable as the order α > 0.7713. By Remark 1, at equilibrium E_{3}, Hopf bifurcation is generated as α = 0.7713.
At the equilibrium E_{2}, the two complex eigenvalues are(17)λ2,3=−az02−z0−1±j4ab−a2z02−z0−122=5±511j2,j2=−1.
If tanargλ2,3=11>tanαπ/2, one has α < 0.8136. This means that the two complex eigenvalues λ_{2,3} of the equilibrium E_{2} lie in the stable zone. Due to λ_{1} > 0 and Remark 2, E_{2} is an unstable saddle-focus with index 1 and E_{3} is an unstable saddle-focus with index 2. While α > 0.8136, the two complex eigenvalues λ_{2,3} of the equilibrium E_{2} lie in the unstable zone, E_{2} is an unstable node-focus, and E_{3} is an unstable saddle-focus with index 2. The types of three equilibria are listed in Table 3. One has the following:
As 0 < α < 0.7713, there are two steady states of E_{1} and E_{3}.
As 0.7713 < α < 0.8136, there are two steady states of E_{1} and the limit cycle bifurcated from unstable saddle-focus E_{3}.
As 0.8136 < α < 1, the stability of E_{1} is unchanged. E_{2} is changed from unstable saddle-focus into unstable node-focus because the two complex roots λ_{2,3} of E_{2} enter into the unstable zone.
The types of the equilibria.
α
E_{1}
E_{2}
E_{3}
(0, 0.7713)
Stable
Unstable saddle-focus
Stable
0.7713
Stable
Unstable saddle-focus
Hopf bifurcation point
(0.7713, 0.8136)
Stable
Unstable saddle-focus
Unstable saddle-focus
0.8136
Stable
Unstable point
Unstable saddle-focus
(0.8136, 1)
Stable
Unstable node-focus
Unstable saddle-focus
Remark 3.
It is found that E_{3} is an unstable saddle-focus with index 2 as the order α > 0.7713. Due to Remark 2, for system (5), chaotic attractors may be generated as the order α > 0.7713.
Fix the initial values (0.1, 0.1, 0.2). Figure 3(a) is the bifurcation diagram of the local maxima of the variable z about the order α, which shows that Hopf bifurcation is generated at α = 0.7713. As 0.7713 < α < 0.8136, system (5) displays a limit cycle bifurcated from the equilibrium E_{3}. As 0.8136 < α < 0.84, the limit cycle induced by Hopf bifurcation disappears and the phase portrait limits to the stable point E_{1}. Increasing α from 0.84 to 1, the period-doubling bifurcation occurs in system (5). Figure 3(b) is the first two Lyapunov exponents of system (5) according to the MATLAB code of [36]. As α > 0.97, the first Lyapunov exponent LE_{1} > 0 and system (5) enters into the chaos.
Numerical simulations of system (5), a = 10/3, b = 10, and the initial values (0.1, 0.1, 0.2). (a) Bifurcation diagram of the local maxima of the variable z about the order α; (b) first two Lyapunov exponents of Case 1; (c) bifurcation diagram of the local maxima of the variable z about the parameter L; (d) first two Lyapunov exponents of Case 2.
Case 2.
α = 0.99, and a = 10/3.
In this case, if the inequality 4b/a>1.4831 holds, one has b > 1.2359. Same as the integer-order case [30], the parameter L is set in [70 mH; 100 mH]; i.e., the parameter b is in [10, 100/7]. If E_{3} is stable, by Theorem 1, the parameter b should be satisfied as(18)b>1.21782⋅a⋅tan2απ/2+1.4831a4=1.21782⋅10/3⋅tan20.99π/2+1.483110/34=5009.2.
This means that E_{3} is a saddle-focus with index 2 for any b∈10,100/7 or any L∈70 mH,100 mH. Thus, chaotic attractors may be generated at E_{3} as L increases from 70 mH to 100 mH.
At the equilibrium E_{2}, two complex eigenvalues are(19)λ2,3=−az02−z0−1±j4ab−a2z02−z0−122=10/3±j410/3b−10/322.
If tanargλ2,3=120b−100/10>tan0.99π/2, one has b > 3377.65 or L < 0.2961 mH. This means that two complex eigenvalues λ_{2,3} of the equilibrium E_{2} lie in the unstable zone as L∈70 mH,100 mH. Due to λ_{1} > 0, E_{2} is an unstable node-focus which is that same as the former case of 0.8136 < α < 1.
The bifurcation diagram of the variable z about the parameter L is plotted numerically in Figure 3(c). The first two Lyapunov exponents are shown in Figure 3(d). It is found that system (5) goes into chaos by the period-doubling bifurcation. After system (5) enters into chaos, there suddenly appears several periodic windows (PWs) as L > 86 mH. Compared to the integer-order model in [30], the fractional-order memristive system described by system (5) has more periodic windows. The minimum of the parameter L for system (5) entering into chaos is larger than the minimum of the parameter L shown in [30].
Therefore, the generated complicated dynamical behaviors of system (5) are related to the coexistence of the stable point E_{1}, the unstable node-focus E_{2}, and the saddle-focus E_{3}.
3.3. Bistability Behaviors
Fixing the parameter values and choosing different initial values, a nonlinear system shows two steady states. This behavior is called the bistability behavior. The bistability behavior reflects the sensitivity of the system to its initial values. For different order α or different inductance L, different bistability behaviors appear in system (5), which is listed in Tables 4 and 5.
Bistability in the case of a = 10/3 and b = 10.
α
Two steady states
(0, 0.7713)
Two stable points of E_{1} and E_{2}
(0.7713, 0.84)
Stable point E_{1} and limit cycle
(0.84, 0.97)
Stable point E_{1} and limit cycle
(0.97, 1)
Stable point E_{1} and chaotic attractor
Bistability in the case of α = 0.99 and a = 10/3.
L mH
Two steady states
(70, 86)
Stable point E_{1} and limit cycle
(86, 100)
Stable point E_{1} and chaotic attractor
Bistability behaviors for different fractional orders are plotted numerically in Figure 4. Figure 4(a) is two steady states of the stable points E_{1} and E_{3}, Figure 4(b) is two steady states of the stable point E_{1} and period-2 cycle, Figure 4(c) is two steady states of the stable point E_{1} and period-4 cycle, and Figure 4(d) is two steady states of the stable point E_{1} and chaotic attractor. Compared to the bistability of stable point and chaotic attractor in the integer-order model, a conclusion that the fractional-order derivative can enrich the bistability behaviors is drawn.
Bistability behaviors with different fractional orders. (a) α = 0.75, two sets of the initial values (0.1, 0.1, 0.2) and (0.1, 0.1, −0.2); (b) α = 0.96 and two sets of the initial values (4, 0.1, 4) and (−2, 0.1, 1); (c) α = 0.97 and two sets of the initial values (4, 0.1, 4) and (−2, 0.1, 1); (d) α = 0.99 and two sets of the initial values (−4, 0.1, 4) and (0, 0.1, −1).
Furthermore, the attraction basins in the x0−z0 plane are used to validate the bistability behaviors of system (5) for four different orders α, as shown in Figure 5, where y0=0.1. The light blue, green, yellow, and magenta regions represent the initial value regions for generating period-2, period-4, chaotic, and stable point behaviors, respectively.
Bistability behaviors demonstrated by the attraction basins in the x0−y0 plane for four different fractional orders with y0=0.1. (a) Attraction basin for α = 0.96; (b) attraction basin for α = 0.97; (c) attraction basin for α = 0.98; (d) attraction basin for α = 0.99.
3.4. Block Designs of System (<xref ref-type="disp-formula" rid="EEq5">5</xref>) in Simulink of MATLAB
By utilizing Simulink of MATLAB, the fractional-order system (5) can be implemented to confirm the above numerical plots.
Figure 6 is the block diagrams in Simulink of MATLAB. Figure 6(a) is the α-order differentiator block design. The top in Figure 6(a) is the masked block of the α-order derivative and the bottom in Figure 6(a) is the filter in fractional-order differentiator (here the Oustaloup recursive filter is used). The masking technique of fo_diff.mdl is provided in [37]. Double clicking the block of Fractional Der s^{α}, the order α can be changed by the parameter dialog box. Furthermore, if(20)D0Ctαxt=ft,0<α<1,one has [38](21)D0Ct1−αD0Ctαxt=dxtdt,or(22)dxtdt=D0Ct1−αft.
Block diagrams in Simulink of MATLAB. (a) Differentiator block design4 of the order α; (b) block diagram of the fractional-order memristor.
Then in the block diagram of the fractional-order memristor or in Figure 6(b), the α-order fractional-order derivative of the state is obtained by an integrator.mdl and the fo_diff.mdl of Fractional Der s^{(1−α)}.
The input voltage in Figure 6(b) is vt=sinωt. Fixing the order α = 0.98, the hysteresis loops of different input frequency ω obtained in the scope (XY graph) are shown in Figure 7(a). Fixing the input frequency ω = 1 rad/s, the hysteresis loops of different order α obtained in the scope (XY graph) are shown in Figure 7(b). Figure 7 plotted in Simulink of MATALB is consistent with Figure 2.
vt versus it of the fractional-order memristor from the scope (XY Graph) in Figure 6(b). (a) The order α = 0.98 with two different frequencies; (b) the frequency ω = 1 rad/s with three different fractional orders.
The block diagram of system (5) in Simulink of MATLAB is designed in Figure 8. By using (22), the α-order fractional-order derivative of the state is obtained by an integrator.mdl and the fo_diff.mdl of Fractional Der s^{(1−α)}. The initial values of the states are set in three integrators. Setting 1 − α = 0.25 in Figure 8, bistability behaviors of α = 0.75 obtained in the scope (XZ graph) are shown in Figure 9(a). Setting 1 − α = 0.01 in Figure 8, bistability behaviors of α = 0.99 obtained in the scope (XZ graph) are shown in Figure 9(b).
The block diagram of system (5) in Simulink of MATLAB.
Bistability behaviors obtained in Simulink of MATLAB, the running time t = 100 s. (a) α = 0.75, two sets of the initial values (0.1, 0.1, 0.2) and (0.1, 0.1, −0.2); (b) α = 0.99, two sets of the initial values (−4, 0.1, 4) and (0, 0.1, −1).
To obtain the bistability behaviors of α = 0.96 and α = 0.97 in Simulink of MATLAB, the values of (1 − α) in Figure 8 are set at 1 − α = 0.04 and 1 − α = 0.03, which are omitted here.
4. Local Activeness and Stability
A locally active kinetic equation can exhibit complex dynamics such as limit cycles or chaos. The passive (not locally active) kinetic equation must converge to a unique steady state [39]. Furthermore, the time t can be set to be large when the steady states of the system are concerned. As parameters a = 10/3 and b = 10, for the fractional-order memristor in system (5) with the large time t, one has the following.
Fix the initial values (0.1, 0.1, 0.2). As α = 0.75, the memristor is active because the power p=W0.75xm⋅v2 keeps negative, which is shown in Figure 10(a); as α = 0.82, the memristor is passive because the power p=W0.82xm⋅v2 keeps positive for time t > t_{0} (such as t_{0} = 50), which is shown in Figure 10(b). As mentioned before, system (5) with α = 0.75 or α = 0.82 converges to stable points. As α = 0.9 or α = 0.99, the memristor is locally active because the power p=Wαxm⋅v2 changes between the positive and the negative, as shown in Figures 10(c) and 10(d). When the locally active memristor is included in system (5), the limit circle is generated as α = 0.9 and chaos is generated as α = 0.99.
It is found that the complex dynamics of the limit cycle and chaos are not displayed in system (5) when the memristor is active (α = 0.75) or the memristor is passive (α = 0.82), while the complex dynamics of the limit cycle and the chaos are generated when the memristor is locally active (α = 0.9 and α = 0.99). As stated at the beginning of this section, complex dynamics of limit cycle and chaos in system (5) are related to the local activeness of the memristor.
5. Conclusions
In this paper, a chaotic system with a fractional-order locally active memristor is discussed. The fractional order in the memristive system makes the equilibrium vary from unstable to stable, leading to the occurrence of Hopf bifurcation. Moreover, the fractional-order memristive system enters into chaos via period-doubling bifurcation route and triggers more periodic windows than the corresponding integer-order system. Given the suitable parameters, say, a and b, the fractional-order memristive system shows bistability behaviors. For different fractional order and different inductance, the fractional-order memristive system displays different bistability behaviors. The fractional order of the system and the local activeness of the memristor are the main reasons for the complicated dynamical behaviors. Besides, the fractional-order memristive system is implemented using the block diagram of Simulink of MATLAB and its hardware implementation and corresponding experiments will be our future works.
Data Availability
The data used to support the findings of this study are available from the corresponding author upon request.
Conflicts of Interest
The authors declare that they have no conflicts of interest.
Acknowledgments
This work was supported by the National Natural Science Foundations of China under Grant nos. 11602035, 11402125, 61601062, and 61801054 and the Natural Science Foundation of Jiangsu Province under Grant no. BK20191451.
KuznetsovA. P.KuznetsovS. P.MosekildeE.StankevichN. V.Co-existing hidden attractors in a radio-physical systemChuaL.Memristor-the missing circuit elementBaoB.JiangT.XuQ.ChenM.WuH.HuY.Coexisting infinitely many attractors in active band-pass filter-based memristive circuitMuthuswamyB.ChuaL. O.Simplest chaotic circuitWuH.BaoB.LiuZ.XuQ.JiangP.Chaotic and periodic bursting phenomena in a memristive Wien-bridge oscillatorBaoB.-C.ChenM.BaoH.XuQ.Extreme multistability in a memristive circuitLiQ.ZengH.LiJ.Hyperchaos in a 4D memristive circuit with infinitely many stable equilibriaAlombahN. H.FotsinH.NgouonkadiE. B. M.NguazonT.Dynamics, analysis and implementation of a multi-scroll memristor-based chaotic circuitKountchouM.LouodopP.BowongS.FotsinH.KurthsJ.Optimal synchronization of a memristive chaotic circuitXuQ.LinY.BaoB.ChenM.Multiple attractors in a non-ideal active voltage-controlled memristor based Chua’s circuitBaoB. C.WuP. Y.BaoH.XuQ.ChenM.Numerical and experimental confirmations of quasi-periodic behavior and chaotic bursting in third-order autonomous memristive oscillatorPetrasI.Fractional-order memristor-based Chua’s circuitKengneJ.LeutchoG. D.Kengnou TelemA. N.Reversals of period doubling, coexisting multiple attractors, and offset boosting in a novel memristive diode bridge-based hyperjerk circuitMaJ.ChenZ.WangZ.ZhangQ.A four-wing hyper-chaotic attractor generated from a 4-D memristive system with a line equilibriumZhouL.WangC.ZhouL.Generating hyperchaotic multi-wing attractor in a 4D memristive circuitKengneJ.Njitacke TabekouengZ.Nguomkam NegouV.NegouA.Periodicity, chaos, and multiple attractors in a memristor-based Shinriki’s circuitNjitackeZ. T.KengneJ.FotsinH. B.NegouA. N.TchiotsopD.Coexistence of multiple attractors and crisis route to chaos in a novel memristive diode bridge-based Jerk circuitSahaP.SahaD. C.RayA.ChowdhuryA. R.Memristive non-linear system and hidden attractorBaoB. C.HuA. H.BaoH.XuQ.ChenM.WuH. G.Three-dimensional memristive hindmarsh-rose neuron model with hidden coexisting asymmetric behaviorsAhamedA. I.LakshmananM.Non-smooth bifurcations, transient hyper-chaos and hyperchaotic beats in a memristive Murali-Lakshmanan-Chua circuitBarbozaR.ChuaL. O.The four-element Chua’s circuitStrukovD. B.SniderG. S.StewartD. R.WilliamsR. S.The missing memristor foundPodlubnyI.YuY. J.WangZ. H.A fractional-order memristor model and the fingerprint of the simple series circuits including a fractional-order memristorFoudaM. E.RadwanonA. G.The fractional-order memristor modelCafagnaD.GrassiG.On the simplest fractional-order memristor-based chaotic systemDancaM.-F.TangW. K. S.ChenG.Suppressing chaos in a simplest autonomous memristor-based circuit of fractional order by periodic impulsesDogaruR.ChuaL. O.Edge of chaos and local activity domain of FitzHugh-Nagumo equationDogaruR.ChuaL. O.Edge of chaos and local activity domain of the Brusselator CNNJinP.WangG.IuH. H.-C.FernandoT.A locally active memristor and its application in a chaotic circuitBiolekD.BiolkováV.BiolekZ.Interpreting area of pinched memristor hysteresis loopWesterlundS.EkstamL.Capacitor theoryAhmedE.El-SayedA. M. A.El-SakaH. A. A.Equilibrium points, stability and numerical solutions of fractional-order predator-prey and rabies modelsEl-SakaH. A.AhmedE.ShehataM. I.El-SayedA. M. A.On stability, persistence, and Hopf bifurcation in fractional order dynamical systemsChuaL.KomuroM.MatsumotoT.The double scroll familyDancaM.-F.KuznetsovN.Matlab code for Lyapunov exponents of fractional-order systemsMonjeC. A.ChenY. Q.VinagreB. M.XueD. Y.FeliuV.LiC.DengW.Remarks on fractional derivativesChuaL. O.Local activity is the origin of complexity