Dynamics and bifurcation analysis of a state-dependent impulsive SIS model

Recently, considering the susceptible population size-guided implementations of control measures, several modelling studies investigated the global dynamics and bifurcation phenomena of the state-dependent impulsive SIR models. In this study, we propose a state-dependent impulsive model based on the SIS model. We firstly recall the complicated dynamics of the ODE system with saturated treatment. Based on the dynamics of the ODE system, we firstly discuss the existence and the stability of the semi-trivial periodic solution. Then, based on the definition of the Poincaré map and its properties, we systematically investigate the bifurcations near the semi-trivial periodic solution with all the key control parameters; consequently, we prove the existence and stability of the positive periodic solutions.

impulsive models have been proposed [15][16][17][18][19]. Fractional calculus is a generalization of the standard integer calculus. It has become a significant field of investigation due to its immense opportunity and wide applications. A considerable number of articles have been concerned with reviewing the presence of solutions to fractional systems. Several researchers pioneered their attempts to extend the impulsive modelling methods to the fractional models [20][21][22][23] or the fractional integro-differential equations [24,25], with various new analytic techniques developed and lots of interesting results obtained [26]. Gupta et al. in [20] investigated the existence and uniqueness of impulsive dynamical fractional systems with quadratic perturbation of second type subject to nonlocal boundary conditions. Kumar et al. in [22] explored the existence of solution of non-autonomous fractional differential equations with integral impulse condition by the measure of noncompactness (MNC), fixed point theorems, and k-set contraction. Ravichandran in [24] discussed the existence and uniqueness of solution to the integro-differential equations involving Atangana-Baleanu fractional derivatives.
Recently, several studies have introduced the susceptible population guided impulsive control into the SIR-type models and systematically investigated the dynamic behaviors [27][28][29][30]. In this study, we extend this modelling idea to the SIS-type model and investigate the rich dynamical behaviors and the bifurcations. The classical SIS-type model is given where S and I are the populations of susceptible and infectious, respectively. A is the recruitment rate of the susceptible populations, and d is the natural death rate. β is the property of transmission per contact, and the incidence rate is denoted by βSI. The recovery from the infected compartment is presented by vI. Furthermore, considering the continuous interventions, including the treatment and vaccination for controlling infectious diseases, system (1.1) becomes the following three-dimensional SIVS model: (1.2) Here, the added compartment V denotes the vaccinated population. The recovery from the infected compartment with hospital treatment is represented by cI b+I . Moreover, the susceptible population is vaccinated with rate q, and the vaccine protection wanes with rate θ . Then, we can extend model (1.2) by including the state-dependent control strategies and replacing the continuous vaccination with impulsive vaccination, and we obtain the following state-dependent impulsive model: Here, q ∈ (0, 1) denotes the vaccination rate of the susceptible population.

Preliminaries
Subject to the restriction which means that N(t) will tend to A d as t approaches infinity. Without loss of generality, the ODE system in model (1.3) can be reduced to the following system: (2.1) Consequently, the proposed state-dependent impulsive model (i.e. system (1.3)) is reduced to the following system: We start with concluding the main dynamics of ODE system (2.1), while the detailed proof is similar to the existing study [31]. From biological considerations, we study (2.1) in the closed set D = (S, I)|S, I ≥ 0, S + I ≤ A/d , which is invariant set under nonnegative initial conditions. Let the right-hand side of (2.1) be zero, we obtain that (2.1) has a disease-free equilibrium E 0 = ( A d , 0). Using the notation in van den Driessche and Watmough [32], the reproduction number is given by .
Note that the characteristic equation of (2.1) at E 0 reads In order to obtain the existence of endemic equilibria, we give the following quadratic equation from model (2.1) depending on the solutions of For convenience, we denote with = a 2 1 -4a 2 . We further denote Then we get the result regarding the number of endemic equilibrium. Theorem 2 For model (2.1), with A 1 and R 0 defined as above, we have: (1) When R 0 > 1 or R 0 = 1 and A > A 1 , there is a unique endemic equilibrium E * .
(2) When R 0 < 1 and A ≤ A 1 , there is no endemic equilibrium.
(3) When 1 > R 0 > R 0 and A > A 1 , there are two endemic equilibria E * and E * . (4) When R 0 = R 0 and A > A 1 , E * and E * coincide into a unique endemic equilibrium of multiplicity 2.
(5) When R 0 < R 0 and A > A 1 , there is no endemic equilibrium.
Remark 1 When A > A 1 , this theorem shows that there exists R 0 (0 < R 0 < 1) such that the model has a unique endemic equilibrium for R 0 = R 0 , then the model has two endemic equilibria for 1 > R 0 > R 0 and a unique endemic equilibrium for R 0 = 1. This situation corresponds to a backward bifurcation which occurs at R 0 = 1. There is also a forward transcritical bifurcation at E 0 (where R 0 = 1) if A ≤ A 1 .
When model (2.1) has no endemic equilibria, we get the global stability of the diseasefree equilibrium E 0 . Theorem 3 If R 0 < 1 and A ≤ A 1 or R 0 < R 0 and A > A 1 hold, then the disease-free equilibrium E 0 of (2.1) is globally asymptotically stable.
Next, we study the stability of the endemic equilibria. Denote H(I) = θ +d +q +βI -cI (b+I) 2 , we get the following results.
Theorem 4 Suppose R 0 > 1 or 1 > R 0 > R 0 and A > A 1 , the endemic equilibrium E * of model (2.1) is a stable node or focus when H(I * ) > 0; E * is an unstable node or focus when H(I * ) < 0; and E * is a center of the linear system of (2.1) when H(I * ) = 0.
To investigate the dynamical behaviors of system (2.2), we first briefly summarize the basic definitions and properties of the impulsive semi-dynamical systems. We consider the following generalized planer impulsive semi-dynamical system with state-dependent feedback control: with z = (x + , y + ) called an impulsive point of z = (x, y). Then, based on the definitions in [12], we can define the impulsive semi-dynamical system and the order-k period solution. Particularly, the following analogue of Poincaré criterion in [33] determines the local stability of an order-k periodic solution.

Poincaré map and its properties
In this section, we first define the Poincaré map of system (2.2), and then we discuss its main properties. Before moving to the details, we firstly provide some intuitive basis on the Poincaré map. Poincaré used a cross section (called the Poincaré section) to transverse the trajectory of continuous motion. Then, according to the discrete motion of the intersection points of the trajectory to the cross section, we can simply judge the trend of the continuous motion. In the diagram on the Poincaré section, the succeed point where the trajectory crosses the Poincaré section can be regarded as a mapping of the point where the trajectory crosses the Poincaré section last time, which can be denoted by x n+1 = f (x n ), n = 1, 2, 3, . . . . This is actually the Poincaré map, and its function is to study the continuous motion through a simple discrete mapping. Based on the above interpretation, we can easily know that the fixed points of the Poincaré map actually reflect the periodic motion of the phase space.
We assume that S T < A d holds true. Denote the two isolines of the system as follows: and define two sections as follows: According to the definitions in the last section, we have that the impulsive function H(S, I) can be defined as Set the section S qS T as a Poincaré section. Choosing an initial point P + k = ((1q)S T , I + k ) on the Poincaré section, and the orbit starting from P + k reaches S S T . We denote the intersection point as P k+1 = (S T , I k+1 ), then the trajectory will jump to P + k+1 = ((1q)S T , I k+1 ) on the section S qS T . It follows from the existence and uniqueness of solutions that I k+1 is identically determined by I + k . Therefore, we can define a function g with g(I + k ) = I k+1 , and then the Poincaré map P 1 can be defined as follows: Since the domain and range of the Poincaré map depends on the dynamics of system (3.1), we conclude the dynamics of system (3.1) by considering the following cases: (C 1 )R 0 < 1 and A ≤ A 1 or R 0 < R 0 and A > A 1 ; (C 2 )R 0 > 1 and H(I * ) > 0; (C 3 )R 0 > 1 and H(I * ) < 0; Note that the phase trajectories of system (3.1) for case (C 3 ) or (C 4 ) show complexity, which leads to the complicated impulsive and phase sets of impulsive system (2.1). In what follows we discuss the definitions of the impulsive set and the phase set of system (2.2) for case (C 1 ) and case (C 2 ). Under case (C 1 ), the DFE E 0 ( A d , 0) is globally asymptotically stable. It follows from the properties of the vector fields of system (3.1) that there exists an orbit The intersection point of 1 to S S T is denoted by Then the impulsive set is and the phase set N can be defined as follows: For case (C 2 ), the endemic equilibrium E * (S * , I * ) is a stable focus or node. If S T < S * holds, the impulsive set and the phase set of system (2.2) are similar to case (C 1 ). If S T > S * , we let We assume B 1 < 0, then there exists an orbit 2 tangent to the set S S T at the point

Semi-trivial periodic solution 4.1 Existence and stability of semi-trivial periodic solution
Let I(t) = 0 for all t ∈ (0, +∞), system (2.2) becomes the following subsystem: Integrating the first equation of (4.1) with the initial conditions S(0) = (1q)S T , we have and solving it with respect to T, we get the period Therefore, system (2.2) has a semi-trivial periodic solution with period T, which is given as follows: Denote ( S(t), I(t)) = (ξ (t), 0), then we discuss the stability of the semi-trivial periodic solution (ξ (t), 0). There are It is easy to calculate that Moreover, there is Thus, the Floquet multiplier μ 2 can be calculated as It follows from the definition of Poincaré map and the property of system (3.1) that the Poincaré map P 1 is monotonically decreasing under case (C 1 ) on the whole domain of definition. Therefore, the semi-trivial periodic solution (ξ (t), 0) is globally attractive. Based on the above discussion, we have the following conclusion.
Theorem 8 If R 0 < 1, then the semi-trivial periodic solution is orbitally asymptotically stable. Particularly, for case (C 1 ), the semi-trivial periodic solution of system (2.2) is globally stable. Therefore, we have I(s, 0))
In what follows, we consider the bifurcations with respect to q under the conditions of the existence of q * . It is easy to see that there is P 1 (0, q) = 0 for all q ∈ (0, 1), and Moreover, we have As a result of the indeterminacy of the sign of m(s) in the interval s ∈ ((1q * 1 )S T , S T ), the sign of g (0; q * 1 ) is undetermined. If we assume g (0; q * 1 ) > 0, then we have Based on the above discussion, we have the following conclusion.

Bifurcation with respect to S T
Similar to the case for q, we first analyze the existence of S * T such that μ 2 | S T =S * T = 1. Taking the derivative of μ 2 with respect to S T yields The roots of the equation f (S T ) = 0, denoted by S T , satisfy the following equation: Thus, > 0 is equivalent to R 0 > 4(1-q) (q-2) 2 . Denote K(q) = 4(1-q) (q-2) 2 . There are K(0) = 1 and K (q) = 4q (q-2) 3 < 0. Therefore, we have that K(q) < 1 holds for all q ∈ (0, 1). This means that > 0 when R 0 > 1, hence equation (4.4) has two roots, denoted by We can verify that 0 < S T 1 < A d < S T 2 . Thus, μ 2 is decreasing when S T ∈ (0, S T 1 ) and increasing when S T ∈ (S T 1 , A d ). Furthermore, when R 0 > 1, there are Therefore, there is a unique S * Based on the above discussion, we can conclude the following.

Proposition 12
Then the semi-trivial periodic solution (ξ (t), 0) is orbitally asymptotically stable for S T ∈ (0, S * T ) and unstable for S T ∈ (S * T , A d ).
In what follows, we investigate the bifurcations of the semi-trivial periodic solution at S * T . We can easily verify that P 1 (0, S T ) = 0 holds true for all S T ∈ (0, A d ), and Moreover, we have As a result of the indeterminacy of the sign of m(s) in the interval s ∈ ((1q)S * T , S * T ), the sign of g (0; S * T ) is undetermined. If we assume g (0; S * T ) > 0, then we have Based on the above discussion, we have the following conclusion.
Based on the above discussion, we have the following conclusion. By choosing S T , q, A as the bifurcation parameters and fixing all the other parameters, we numerically verified the existence of the critical bifurcation point with respect to these three parameters in Fig. 1. In Fig. 2(A), we showed that the semi-trivial periodic solution is globally stable when we choose S T = 6.3. As S T increases to 8, then an unstable periodic solution appears, the semi-trivial periodic solution and the positive equilibrium E * are bistable, as shown in Fig. 2(B).

Conclusion and discussion
Recently, several studies pioneered the attempt to include the susceptible populationguided interventions for controlling infectious diseases into the SIR systems [27][28][29][30]. As highlighted in these studies, with this kind of control strategies, it is possible to define the control reproduction number for the impulsive system compared with the infectedpopulation induced control interventions. In this study, we extended the SIS model by including the control strategy i.e. susceptible population-guided impulsive control, and systematically studied its dynamics and bifurcations.
We started with recalling the dynamic behavior of the ODE system. By defining the Poincaré map of the proposed model and presenting the proof of its properties, we explored the existence and stability of the semi-trivial periodic solution. We found a threshold parameter, which can be defined as the control reproduction number, determining the stability of the semi-trivial periodic solution. In detail, it is locally stable when the control reproduction number is less than 1 and unstable when the reproduction number exceeds the threshold value 1.
Furthermore, we investigated the bifurcations near the semi-trivial periodic solution considering the key parameters, including the constant recruitment rate A, the threshold level of the susceptible population S T , and the pulse vaccination rate q. We proved that as the bifurcation parameters vary, the system can undergo the transcritical or pitchfork bifurcation near the semi-trivial periodic solution; consequently, the semi-trivial periodic solution loses its stability, while an unstable positive periodic solution appears. It is also interesting to summarize the biological implications by this impulsive SIS model and its dynamic behaviors and bifurcations. Through the bifurcation analysis with respect to S T , we obtained a critical value of the threshold to guarantee the disease-free periodic solution to be stable. This means that by choosing a proper threshold of the susceptible population size, this kind of control strategy can indeed help to eliminate the disease. The bifurcations with respect to other parameters can actually reflect similar implications, like choosing a proper vaccination rate. On the other hand, the bistability of the positive equilibrium and the semi-trivial periodic solution indicate that the outcomes under this state-dependent control strategy depend on the initial conditions of the susceptible population and the infected population, hence a personalized strategy is recommended. Our model does not cover spatio-temporal delay, thus considering the effect of spatio-temporal delay [7] could be a valuable issue for future research.