Hopf bifurcation in delayed nutrient-microorganism model with network structure

In this paper, we introduce and deal with the delayed nutrient-microorganism model with a random network structure. By employing time delay τ as the main critical value of the Hopf bifurcation, we investigate the direction of the Hopf bifurcation of such a random network nutrient-microorganism model. Noticing that the results of the direction of the Hopf bifurcation in a random network model are rare, we thus try to use the method of multiple time scales (MTS) to derive amplitude equation and determine the direction of the Hopf bifurcation. It is showed that the delayed random network nutrient-microorganism model can exhibit a supercritical or subcritical Hopf bifurcation. Numerical experiments are performed to verify the validity of the theoretical analysis.


Introduction
The existence of the Hopf bifurcation in ecological systems is a hot investigation field, since its practical significance in biological control. Therefore, there are many results concerning the Hopf bifurcation in ecological systems that have been reported, see Refs. [5][6][7]12]. One of the most important results induced by the Hopf bifurcation is periodic solutions, including spatially homogeneous and spatially non-homogeneous periodic solutions. As a result, it is necessary to study or determine the stability of the periodic solution. As we know, there are two common techniques to study the direction of the Hopf bifurcation in a delayed differential equation, they are centre manifold reduction (CMR) and MTS, see Refs. [2,8,11].
It is noticed that there are few results about the direction of the Hopf bifurcation in a delayed reaction-diffusion system with network structure have been reported [9,10], where the direction of the Hopf bifurcation is determined by employing the method of CMR. However, different from the technique is adopted in [9,10], we mainly attempt to use MTS to compute amplitude equation, and determine the direction of the Hopf bifurcation in a delayed nutrient-microorganism model with random network structure. More precisely, the system we consider takes the form (1) This model is called nutrient-microorganism system in the sediment, and first proposed by Baurmann and Feudel in [1](continuous form). All parameters d 1 , d 2 , α, β, K in (1) are positive constants; τ ≥ 0 is time delay, it implies that some time delay is required for microorganism to consume nutrient in the sediment.
For system (1), we assume that it is defined on an undirected network with N nodes and there are no self-loops; u i and v i are the densities of the microorganism and the nutrient on node i, respectively; is the N × N discrete Laplacian matrix of network with its elements ij = k i δ ij − L ij , where L is the adjacency matrix encoding the network connection, this indicates it satisfies L ij = 1 if there is a link connecting from patch i to patch j. If not, L ij = 0 when there is no any link connecting from patch i to patch j, and δ ij is the Kronecker's delta [4]. Moreover, the degree of the ith node is defined by k i = N j=1 L ij , and the connection probability between node i and node j for i = j is p(0 ≤ p ≤ 1). This paper is structured as follows. In Section 2, we establish the existence results of the Hopf bifurcation of the model (1). In Section 3, the amplitude equation of the Hopf bifurcation is deduced. As a result, the supercritical or the subcritical Hopf bifurcation can be yield by analysing the amplitude equation. We perform the numerical simulations to verify the theoretical analysis in Section 4, and some discussions are made in Section 5.

Existence of the Hopf bifurcation
In this section, we give some conditions to ensure the existence of the Hopf bifurcation of the networked model (1). To this end, we first consider the existence and stability of positive equilibria for model (1).

Lemma 2.1 ([3]):
The possible positive equilibria of model (1) can be found as follows Thereby, in view of Lemma 2.1 and Lemma 2.2, we only focus on the dynamical behaviours near E * 1 , and we denote it by E * 1 E * = (u * , v * ) for simplicity. Now, we shall perform the linear stability analysis of system (1) near the positive equilibrium E * = (u * , v * ). For this purpose, let u i = u i − u * and v i = v i − v * be the small perturbations, then the linear system of (1) evaluated at E * = (u * , v * ) can be written as follows Let 0 = 1 > 2 > · · · > N be the eigenvalues of the discrete Laplacian matrix , and suppose that L φ = {φ i } N i=1 is the subspace generated by the eigenfunctions associated to the topological eigenvalue i . Then, the general solution of system (2) can be rewritten as It then follows that where Let λ = iω(ω > 0) be the solution of Equation (3), we have Thereby, if one of the conditions in the following is satisfied, one yields We thus obtain the critical value τ of the Hopf bifurcation is where j ∈ N 0 = {0, 1, 2, . . .}. In addition, a straightforward calculation shows that Moreover, note that this implies that the transversality condition of the Hopf bifurcation is satisfied. Combine this with (5) we claim that the delayed network model (1) undergoes the Hopf bifurcation at E * when τ = τ j i for 0 = 1 > 2 > · · · > N and j ∈ N 0 . Especially, the delayed network model (1) admits a Hopf bifurcation at E * when τ = τ 0 0 , and in this case the periodic solution bifurcated from the Hopf bifurcation is spatially homogeneous.

Direction of the Hopf bifurcation
In this section, we shall employ MTS to derive amplitude equations and determine the direction of the Hopf bifurcation. We first define τ 0 0 τ * , ω 0 ω * and τ → t/τ , where ω 0 and τ 0 0 can be found in (4) and (5), respectively. Now, a rewritten form of the networked system (1) can be read asU where we set In addition, the nonlinear term F( To employ the technique of MTS, let ε be a small perturbation parameter. Then introducing the time scales T 0 = t, T 2 = ε 2 t, this induces that Solution U i can be decribed as We thus obtain a fact that By a similar manner, we can write the nonlinear term F as follows where Next, we introduce the small perturbation of the Hopf bifurcation parameter τ = τ * + ε 2 δ with δ > 0. Keep this in mind, denote L t | τ =τ * = L * t , L τ | τ =τ * = L * τ and put (7)-(10) into (6), one has O(ε): O(ε 2 ): O(ε 3 ): Considering the solution of perturbation Equation (11) near the Hopf mode, we have where we assume that H(T 2 ) is the complex amplitude and We thus obtain This means that the perturbation Equation (12) has a particular solution with the form where the vectors (q 11 , q 12 ) T and (q 21 , q 22 ) T are unknown. Thence, to determine them we should insert (15) into (12). Then equating like terms of e 2iω * τ * T 0 H 2 (T 2 ) and |H(T 2 )| 2 , we have As a result, it follows from (14) and (15) that Now inserting all of them and (14)-(15) into the perturbation (13), one obtains where c.c. is the conjugation term, NST represents the non-secular term and w = (w (1) , w (2) ) T with Then perturbation Equation (16) possesses its solution when the Fredholm alternative condition is satisfied. It implies that the following orthogonal condition should be satisfied where we take the inner product satisfies < a, b >=ā T b. Then using (17) and noticingp * 11 + p 12p * where we denote η 1 = (a 11 + a 12 p 12 )p * 11 + (a 21 + a 22 p 12 − p 12 e −iω * τ * )p * In what follows, for the purpose of investigating the Hopf bifurcation near τ = τ * , we shall find the amplitude equation on the centre manifold, it is useful for investigating the Hopf bifurcation. To this end, let H(T 2 ) = ze −iω * τ * T 2 with z = x + iy, then it follows from (18) that ẋ = −ω * τ * y + δ(Re{η 1 }x − Im{η 1 }y) + (Re{η 2 }x − Im{η 2 }y)(x 2 + y 2 ), y = ω * τ * y + δ(Re{η 1 }y − Im{η 1 }x) + (Re{η 2 }y − Im{η 2 }x)(x 2 + y 2 ), where Re{•} and Im{•} represent the real part and imaginary part of •. Now, we let x = ρ cos θ and y = ρ sin θ, then (19) becomes where K 1 = Re{η 1 } and K 2 = Re{η 2 }. By virtue of (20), we have the following results about the direction of the Hopf bifurcation. (ii-b) it is subcritical when K 1 < 0, and periodic solution bifurcated from the Hopf bifurcation is unstable; Proof: It is easy to verify that (20) has a unique positive equilibrium (corresponding to the Hopf bifurcation) ρ = −δK 1 K 2 , and thus its existence condition is K 1 K 2 < 0. It indicates (i) is true. Moreover, Equation (20) has a unique eigenvalue, say λ, and λ = −2δK 1 . By employing linear stability analysis theory, we know that (ii) is valid. The proof is completed.

Numerical simulations
In this section, we main verify the effectiveness of Theorem 3.1 by numerical simulations. We assume that the numbers of the node are 100, namely we take N = 100 in the random network model (1). Moreover, we fix the connecting probability p = 0.35 between different nodes u i and u j for i = j. We now choose the parameters in system (1) are α = 1.5, β = 2.5, K = 0.05, d 1 = 2 and d 2 = 0.5, then one obtains the positive equilibrium E * = (0.3618, 1.1585), ω * = 1.5364, τ * = 1.1585, η 1 = 0.5465 + 1.2577i and η 2 = −0.3240 − 16.3341i. This means that K 1 = Re{η 1 } = 0.5465 > 0 and K 1 K 2 = −0.1771 < 0 are satisfied in Theorem 3.1. Figure 1 shows that the distribution of solutions u i (1 ≤ i ≤ 100) with the development of moments. When taking time delay 0.95 = τ < τ * , we find that the positive equilibrium E * is locally asymptotically stable, see Figure 2. Figure 3 suggests that there is a supercritical Hopf bifurcation, and the periodic solution with the spatial homogeneity bifurcated from the Hopf bifurcation is stable in the delayed network model (1), where we choose 1.1587 = τ > τ * . As such, the results in Theorem 3.1 are valid.

Conclusions
We deal with a nutrient-microorganism model with time delay and random network structure in this paper. By employing time delay τ as the critical parameter of the Hopf bifurcation, we explore its occurrence conditions. For the direction of the Hopf bifurcation, we try to adopt MTS to derive amplitude equation near τ = τ * , it is found that the sign of K 1 K 2 determines the existence of the Hopf bifurcation. Namely the Hopf bifurcation exists when K 1 K 2 < 0, and there is no Hopf bifurcation when K 1 K 2 > 0. Moreover, the sign of K 1 determines the direction of the Hopf bifurcation with hypothesis K 1 K 2 < 0. More precisely, the Hopf bifurcation is supercritical (resp. subcritical) and the periodic solution is stable (resp. unstable) when K 1 > 0 (resp. K 1 < 0). Numerical simulations indicate that our theoretical analysis is valid, and compared with the works done in [9,10] we claim that MTS is a easier technique to determine the direction of the Hopf bifurcation in a delayed network model than CMR. For more interesting results about this networked model, for example, resonant/nonresonant Hopf bifurcation, will be further considered.

Disclosure statement
No potential conflict of interest was reported by the author(s).

Funding
This work was supported by the National Natural Science Foundation of China [grant number 11971032], and Young Talent Support Project of Henan [grant number 2020HYTP012].