Loading [MathJax]/jax/output/SVG/jax.js
Research article

A proof of Kirchhoff's first law for hyperbolic conservation laws on networks

  • In dynamical systems on networks, Kirchhoff's first law describes the local conservation of a quantity across edges. Predominantly, Kirchhoff's first law has been conceived as a phenomenological law of continuum physics. We establish its algebraic form as a property that is inherited from fundamental axioms of a network's geometry, instead of a law observed in physical nature. To this end, we extend calculus to networks, modeled as abstract metric spaces, and derive Kirchhoff's first law for hyperbolic conservation laws. In particular, our results show that hyperbolic conservation laws on networks can be stated without explicit Kirchhoff-type boundary conditions.

    Citation: Alexandre M. Bayen, Alexander Keimer, Nils Müller. A proof of Kirchhoff's first law for hyperbolic conservation laws on networks[J]. Networks and Heterogeneous Media, 2023, 18(4): 1799-1819. doi: 10.3934/nhm.2023078

    Related Papers:

    [1] Tianyuan Xu, Shanming Ji, Chunhua Jin, Ming Mei, Jingxue Yin . EARLY AND LATE STAGE PROFILES FOR A CHEMOTAXIS MODEL WITH DENSITY-DEPENDENT JUMP PROBABILITY. Mathematical Biosciences and Engineering, 2018, 15(6): 1345-1385. doi: 10.3934/mbe.2018062
    [2] Changwook Yoon, Sewoong Kim, Hyung Ju Hwang . Global well-posedness and pattern formations of the immune system induced by chemotaxis. Mathematical Biosciences and Engineering, 2020, 17(4): 3426-3449. doi: 10.3934/mbe.2020194
    [3] Xu Song, Jingyu Li . Asymptotic stability of spiky steady states for a singular chemotaxis model with signal-suppressed motility. Mathematical Biosciences and Engineering, 2022, 19(12): 13988-14028. doi: 10.3934/mbe.2022652
    [4] Xiaoli Hu, Shengmao Fu . Global boundedness and stability for a chemotaxis model of Boló’s concentric sclerosis. Mathematical Biosciences and Engineering, 2020, 17(5): 5134-5146. doi: 10.3934/mbe.2020277
    [5] Dongxiu Wang, Fugeng Zeng, Lei Huang, Luxu Zhou . Persistence and boundedness in a two-species chemotaxis-competition system with singular sensitivity and indirect signal production. Mathematical Biosciences and Engineering, 2023, 20(12): 21382-21406. doi: 10.3934/mbe.2023946
    [6] Chris Cosner, Nancy Rodríguez . On the Allee effect and directed movement on the whole space. Mathematical Biosciences and Engineering, 2023, 20(5): 8010-8030. doi: 10.3934/mbe.2023347
    [7] Chichia Chiu, Jui-Ling Yu . An optimal adaptive time-stepping scheme for solving reaction-diffusion-chemotaxis systems. Mathematical Biosciences and Engineering, 2007, 4(2): 187-203. doi: 10.3934/mbe.2007.4.187
    [8] Sunwoo Hwang, Seongwon Lee, Hyung Ju Hwang . Neural network approach to data-driven estimation of chemotactic sensitivity in the Keller-Segel model. Mathematical Biosciences and Engineering, 2021, 18(6): 8524-8534. doi: 10.3934/mbe.2021421
    [9] Wenjie Zhang, Lu Xu, Qiao Xin . Global boundedness of a higher-dimensional chemotaxis system on alopecia areata. Mathematical Biosciences and Engineering, 2023, 20(5): 7922-7942. doi: 10.3934/mbe.2023343
    [10] Lin Zhang, Yongbin Ge, Zhi Wang . Positivity-preserving high-order compact difference method for the Keller-Segel chemotaxis model. Mathematical Biosciences and Engineering, 2022, 19(7): 6764-6794. doi: 10.3934/mbe.2022319
  • In dynamical systems on networks, Kirchhoff's first law describes the local conservation of a quantity across edges. Predominantly, Kirchhoff's first law has been conceived as a phenomenological law of continuum physics. We establish its algebraic form as a property that is inherited from fundamental axioms of a network's geometry, instead of a law observed in physical nature. To this end, we extend calculus to networks, modeled as abstract metric spaces, and derive Kirchhoff's first law for hyperbolic conservation laws. In particular, our results show that hyperbolic conservation laws on networks can be stated without explicit Kirchhoff-type boundary conditions.



    In the present work, we shall consider a chemotaxis-haptotaxis model

    {ut=Δuχ(uv)ξ(uw)+μu(1uw),τvt=Δvv+u,wt=vw+ηu(1uw), (1.1)

    where χ and ξ are positive parameters. In the model (1.1), u represents the density of cancer cell, v and w denote the density of matrix degrading enzymes (MDEs) and the extracellular matrix (ECM) with the positive sensitivity χ, ξ, respectively. Such an important extension of chemotaxis to a more complex cell migration mechanism has been proposed by Chaplain and Lolas [3] to describe the cancer cell invasion of tissue. In that process, cancer invasion is associated with the degradation of ECM, which is degraded by MDEs secreted by cancer cells. Besides random motion, the migration of invasive cells is oriented both by a chemotaxis mechanism and by a haptotaxis mechanism.

    In the past ten more years, the global solvability, boundedness and asymptotic behavior for the corresponding no-flux or homogeneous Neumann boundary-initial value problem in bounded domain and its numerous variants have been widely investigated for certain smooth initial data. For the full parabolic system of (1.1), Pang and Wang [4] studied the global boundedness of classical solution in the case τ=1 in 2D domains, and the global solvability also was established for three dimension. When η=0 and τ=1, Tao and Wang [5] proved the existence and uniqueness of global classical solution for any χ>0 in 1D intervals and for small χμ>0 in 2D domains, and Tao [6] improved the results for any μ>0 in two dimension; Cao [7] proved for small χμ>0, the model (1.1) processes a global and bounded classical solution in 3D domains.

    When τ=0, the second equation of (1.1) becomes an elliptic function. In the case of η>0, Tao and Winkler [8] proved the global existence of classical solutions in 2D domains for any μ>0. In the case of η=0, the global existence and boundedness for this simplified model under the condition of μ>(N2)+Nχ in any N-D domains in [9]. Moreover, the stabilization of solutions with on-flux boundary conditions was discussed in [10]. For the explosion phenomenon, Xiang [11] proved that (1.1) possess a striking feature of finite-time blow-up for N3 with μ=η=τ=0; the blow-up results for two dimension was discussed in [2] with wt=vw+ηw(1w) and μ=0.

    When χ=0, the system (1.1) becomes a haptotaxis-only system. The local existence and uniqueness of classical solutions was proved in [12]. In [13,14,15], the authors respectively established the global existence, the uniform-in-time boundedness of classical solutions and the asymptotic behavior. Very recently, Xiang[11] showed that the pure haptotaxis term cannot induce blow-up and pattern for N3 or τ=0 in the case of μ=η=0.

    Without considering the effect of the haptotaxis term in (1.1), we may have the extensively-studied Keller-Segel system, which was proposed in [16] to describe the collective behavior of cells under the influence of chemotaxis

    {tu=Δuχ(uv),τtv=Δuλv+u (1.2)

    with u and v denoting the cell density and chemosignal concentration, respectively. There have been a lot of results in the past years (see [17,18,19,20,21], for instance). Here we only mention some global existence and blow-up results in two dimensional space. For the parabolic-elliptic case of (1.2) with λ=0, 8πχ was proved to be the mass threshold in two dimension in [22,23,24] (see also [25,26] for related results in the bounded domain); namely, the chemotactic collapse (blowup) should occur if and only if u0L1 is greater than 8πχ. If u0L1<8πχ, the existence of free-energy solutions were improved in [22]. Furthermore, the asymptotic behavior was given by a unique self-similar profile of the system (see also [27] for radially symmetric results concerning self-similar behavior). For the results in the threshold 8πχ, we refer readers for [28,29,30] for more details. For the parabolic-elliptic model in higher dimensions (N3) in (1.2), the solvability results were discussed in [31,32,33,34] with small data in critical spaces like LN2(RN),LN2w(RN),MN2(RN), i.e., those which are scale-invariant under the natural scaling. Blowing up solutions to the parabolic-elliptic model of (1.2) in dimension N3 have been studied in [35,36,37,38].

    In the case τ=1, Calvez and Corrias [1] showed that under hypotheses u0ln(1+|x|2)L1(R2) and u0lnu0L1(R2), any solution exists globally in time if u0L1<8πχ. In [39], the extra assumptions on u0 were removed, while the condition on mass was restricted to u0L1<4πχ. The value 4πχ appeared since a Brezis-Merle type inequality played an essential role there. These results were improved in [40,41] to global existence of all solutions with u0L1<8πχ by two different method. Furthermore the global existence of solutions was also obtained under some condition on u0 in the critical case u0L1=8πχ in[40]. The blow-up results of the parabolic-parabolic case in the whole space were discussed in [42,43] with the second equation was replaced by tv=Δu+u.

    However, the global solvability and explosion phenomenon of chemotaxis-haptotaxis model in the whole space have never been touched. Here we consider the global solvability of a simplified model of (1.1)

    {ut=Δuχ(uv)ξ(uw),xR2, t>0,vt=Δvv+u,xR2, t>0,wt=vw,xR2, t>0,u(x,0)=u0(x),  v(x,0)=v0(x),  w(x,0)=w0(x),xR2. (1.3)

    Main results. We assume that the initial data satisfies the following assumptions:

    (u0,v0,w0)H2(R2)×H3(R2)×H3(R2) and u0,v0,w0 are nonnegative, (1.4)
    u0L1(R2,ln(1+|x|2)dx) and u0lnu0L1(R2) (1.5)

    and

    Δw0L(R2) and w0L(R2). (1.6)

    Theorem 1.1. Let χ>0, ξ>0 and the initial data (u0,v0,w0) satisfy (1.4)–(1.6). If m:=u0L1<8πχ, then the corresponding chemotaxis-haptotaxis system (1.3) possesses a unique global-in-time, nonnegative and strong solution (u,v,w) fulfilling that for any T<

    (u,v,w)C(0,T;H2(R2)×H3(R2)×H3(R2)).

    Remark 1.1. Our theorem extends the previous results in two aspects. First, our result agrees with that in [1] by setting w=0, which proved that if u0L1<8πχ, then the Cauchy problem of the system (1.2) admits a global solution. Secondly, our theorem extends Theorem 1.1 in [2], where the authors proved that 4πχ is the critical mass of the system (1.3) in bounded domains, implying the negligibility of haptotaxis on global existence.

    We obtain the critical mass value using the energy method in [1,22]. The energy functional:

    F(u,v,w)(t)=R2ulnuχR2uvξR2uw+χ2R2(v2+|v|2),t(0,Tmax) (1.7)

    as shown in [2] comes out to be the key ingredient leading to the global existence of solutions under the smallness condition for the mass. Under the assumption

    u0L1<8πχ (1.8)

    and (1.5), we can derive an integral-type Gronwall inequality for F(t). As a result, we can get a priori estimate for the R2ulnu, which is the key step to establish the global existence of solutions to the system (1.3).

    The rest of this paper is organized as follows. In Section 2, we prove local-in-time existence of the solution, and obtain the blow-up criteria for the solution. In Section 3, we give the proof of the Theorem 1.1.

    In the following, (u)+ and (u) will denote the positive and negative part of u as usual, while Lp:=Lp(R2).

    We now establish the local existence and uniqueness of strong solutions to system (1.3). Our strategy is first to construct an iteration scheme for (1.3) to obtain the approximate solutions and then to derive uniform bounds for the approximate solutions to pass the limit.

    Lemma 2.1. Let χ>0, ξ>0 and u00. Then, there exists a maximal existence time Tmax>0, such that, if the initial data (u0,v0,w0) satisfy (1.4), then there exists a unique solution (u,v,w) of (1.3) satisfying for any T<Tmax, and

    (u,v,w)C(0,T;H2(R2)×H3(R2)×H3(R2)). (2.1)

    Furthermore, u, v and w are all nonnegative.

    Proof. To obtain the local solution, we follow similar procedures of an iterative scheme developed in [45,46]. We construct the solution sequence (uj,vj,wj)j0 by iteratively solving the Cauchy problems of the following system

    {tuj+1=Δuj+1χ(uj+1vj)ξ(uj+1wj),xR2, t>0,tvj+1=Δvj+1vj+1+uj,xR2, t>0,twj+1=vj+1wj+1,xR2, t>0,u(x,0)=u0(x),  v(x,0)=v0(x),  w(x,0)=w0(x),xR2. (2.2)

    We first set (u0(x,t),v0(x,t),w0(x,t))=(u0(x),v0(x),w0(x)). We point out that the system is decouple, then by the linear parabolic equations theory in [44,Theorem Ⅲ.5.2], we can obtain the unique solution u1,v1V1,122([0,T]×R2), then we get w1C1([0,T],H1(R2)) by directly solving the ordinary equation. Similarly, we define (uj,vj,wj) iteratively.

    In the following, we shall prove the convergence of the iterative sequences {uj,vj,wj}j1 in C(0,T;X) with X:= H2×H3×H3 for some small T>0. To obtain the uniform estimates, we may use the standard mollifying procedure. However, since the procedure is lengthy, we omit the details, like in the proofs of Theorem 1.1 in [45] and Theorem 2.1 in [46].

    Uniform estimates: We will use the induction argument to show that the iterative sequences {uj,vj,wj}j1 are in C(0,T;X) with X:= H2×H3×H3 for some small T>0, which means that there exists a constant R>0 such that, for any j, the following inequality holds for a small time interval

    sup0tT(ujH2+vjH3+wjH3)R, (2.3)

    where R=2{u0H2+v0H3+w0H3}+8. Due to the definition of R, the case j=0 is obvious. Then, we need to show that (2.3) is also true for j+1. This will be done by establishing the energy estimate for (uj+1,vj+1,wj+1). First, we begin with the estimate of vj+1.

    (ⅰ) Estimates of vj+1. Taking the L2 inner product of the second equation of (2.2) with vj+1, integrating by parts and using Young's inequality, we have

    12ddtvj+1(t)2L2+vj+12L2=R2(vj+1)2+R2vj+1uj12vj+12L2+12uj2L2. (2.4)

    To show the H1 estimate of vj+1, we will multiply the second equation of (2.2) by tvj+1, integrating by parts and then obtain

    12ddtvj+1(t)2L2+tvj+12L2=R2vj+1tvj+1+R2ujtvj+112tvj+12L2+vj+12L2+uj2L2. (2.5)

    For the H2 estimate of vj+1, by Young's inequality, we have

    12ddt2vj+1(t)2L2+Δvj+12L2=R2(Δvj+1)2+R2Δvj+1Δuj12Δvj+12L2+12Δuj2L2. (2.6)

    Similarly, integrating by parts, it is clear that for all t(0,T)

    ddt3vj+1(t)2L2=2R23vj+13(Δvj+1vj+1+uj)=2R2|4vj+1|22R2|3vj+1|22R24vj+12uj4vj+12L223vj+12L2+2uj2L2,

    togethering with (2.3)–(2.6) and adjusting the coefficients carefully, we can find a positive constant α such that

     ddtvj+1(t)2H3+α(vj+12H4+tvj+12L2)c1(vj+12H3+uj2H2) (2.7)

    with c1>0. Here after ci(i=2,3...) denotes the constant independent of R. Integrating on (0,t), we can obtain for all t(0,T)

    vj+1(t)2H3+αt0(vj+1(s)2H4+tvj+1(s)2L2)ec1Tv02H3+ec1TTsupt(0,T)uj(t)2H2ec1Tv02H3+ec1TTR22v02H3+2, (2.8)

    by choosing T>0 small enough to satisfy ec1T<2 and TR2<1.

    (ⅱ) The estimate of wj+1. In fact, the third component of the above solution of (2.2) can be expressed explicitly in terms of vj+1. This leads to the representation formulae

    wj+1(x,t)=w0(x)et0vj+1(x,s)ds, (2.9)
    wj+1(x,t)=w0(x)et0vj+1(x,s)dsw0(x)et0vj+1(x,s)dst0vj+1(x,s)ds (2.10)

    as well as

    Δwj+1(x,t)=Δw0(x)et0vj+1(x,s)ds2et0vj+1(x,s)dsw0(x)t0vj+1(x,s)ds+w0(x)et0vj+1(x,s)ds|t0vj+1(x,s)ds|2w0(x)et0vj+1(x,s)dst0Δvj+1(x,s)ds. (2.11)

    From (2.9), we can easily get for t(0,T)

    wj+1Lpw0Lp,p(1,]. (2.12)

    From (2.10), by (2.8), the definition of R and the following inequality

     t0f(x,s)dsLp={R2|t0f(x,s)ds|pdx}1p{tp1R2t0|f|pdsdx}1ptsups(0,t)f(s)Lp,for all p(1,), (2.13)

    we can obtain

    wj+1L2w0L2+w0et0vj+1dst0vj+1dsL2w0L2+w0Lt0vj+1dsL2w0L2+w0LTsupt(0,T)vj+1L2w0L2+c2R2Tw0L2+1 (2.14)

    by setting T small enough to satisfy c2R2T<1.

    Similarly, by the embedding H2W1,4 and (2.13), we can obtain from (2.11) for all t(0,T)

    Δwj+1L2Δw0L2+2w0t0vj+1dsL2+w0|t0vj+1ds|2L2 +w0t0Δvj+1dsL2Δw0L2+2w0L4t0vj+1dsL4+w0L|t0vj+1ds|2L2 +w0Lt0Δvj+1dsL2Δw0L2+2w0L4Tsupt(0,T)vj+1L4+w0LT2supt(0,T)vj+12L2 +w0LTsupt(0,T)Δvj+1L2Δw0L2+c3R2T+c3R3T2Δw0L2+2 (2.15)

    by setting T small enough to satisfy c3R2T<1 and c3R3T2<1.

    Now we deduce the L2 norm of 3wj+1. According to the equation of w and Hölder inequality, we can easily get for all t(0,T)

    ddt3wj+1(t)2L2=2R23wj+13wj+1t=2R23wj+13(vj+1wj+1)23wj+1L23(vj+1wj+1)L2c43wj+1L23wj+1vj+1+32wj+1vj+1+32vj+1wj+1+3vj+1wj+1L2c4{3wj+12L2vj+1L+3wj+1L22wj+1L4vj+1L4+3wj+1L22vj+1L2wj+1L+3wj+1L23vj+1L2wj+1L}. (2.16)

    By Galiardo-Nirenberg inequality, we have

    2wj+1L4c53wj+134L2wj+114L,wj+1Lc53wj+112L2wj+112L.

    Together with Young's inequality, (2.8), (2.12) and (2.16), we can get

    ddt3wj+1(t)2L2c6(vj+12H3+1)3wj+12L2+c6wj+12Lc7R23wj+12L2+c6w02Lc7R23wj+12L2+c7R2. (2.17)

    Then, we can deduce from Gronwall's inequality that for all t(0,T)

    3wj+12L2ec7R2T33w02L2+ec7R2T3c7R2T23w02L2+2 (2.18)

    by setting T small enough to satisfy ec7R2T3<2 and c7R2T<1.

    Combining with (2.12)–(2.15) and (2.18), we can see that

    wj+1H32w0H3+5. (2.19)

    (ⅲ) Estimates of uj+1. Taking the L2 inner product of the equation of uj+1 in (2.2), integrating by part we obtain

    12ddtuj+1(t)2L2+uj+12L2=χR2uj+1vjuj+1+ξR2uj+1wjuj+1uj+1L2vjLuj+1L2+uj+1L2wjLuj+1L2. (2.20)

    By (2.8), (2.19) and (2.20) and the embedding H3W1,, we can see for all t(0,T)

    12ddtuj+1(t)2L2+uj+12L2c8R2uj+12L2+12uj+12L2. (2.21)

    Now we turn to show the L2-estimate of uj+1. Multiplying Δuj+1 to both sides of the first equation of (2.3) and integrating by parts, we obtain for all t(0,T)

    12ddtuj+1(t)2L2+Δuj+12L2=χR2Δuj+1(uj+1vj)+ξR2Δui+1(ui+1wj)=I1+I2. (2.22)

    By Hölder inequality and Young's inequality, it yields that

    I1c9Δuj+1L2(uj+1vj)L2c10Δuj+1L2{uj+1H1vjH2}14Δuj+12L2+c210vj2H2uj+12H1.

    Applying the similar procedure to I2, we can obtain

    I214Δuj+12L2+c210wj2H2uj+12H1,

    which entails that for all t(0,T)

    12ddtuj+1(t)2L2+12Δuj+12L2c11R2uj+12H1. (2.23)

    Similar as (2.16), we can get

    ddt2uj+1(t)2L2=2R22uj+12uj+1t=2R22uj+1{2(Δuj+1χ(uj+1vj)ξ(uj+1wj))}=2R22uj+12Δuj+12χR22uj+13(uj+1vj)2ξR22ui+13(uj+1wj)=23uj+12L2+2χR23uj+12(uj+1vj)+2ξR23ui+12(ui+1wj)3uj+12L2+c12{2(uj+1vj)2L2+2(ui+1wj)2L2}3uj+12L2+c13R2uj+12H2. (2.24)

    Together with (2.21), (2.23) and (2.24), and adjusting the coefficients carefully, we can find a positive constant β such that

    ddtuj+1(t)2H2+βuj+12H3c14R2uj+12H2, (2.25)

    which implies from Gronwall's inequality that

    uj+12H2+βt0uj+12H3ec14R2Tu02H22u02H2,for all t(0,T), (2.26)

    by choosing T small enough to satisfy ec14R2T<2.

    Combining (2.8), (2.19) and (2.26), we can get for all t(0,T)

    uj+1H2+vj+1H2+wj+1H32{u0H1+v0H2+w0H3}+7R, (2.27)

    by the definition of R.

    Convergence: The derivation of the relevant estimates of uj+1uj, vj+1vj and wj+1wj are similar to the ones of uj+1, vj+1 and wj+1, so we omit the details. For simplicity, we denote δfj+1:=fj+1fj. Subtracting the j-th equations from the (j+1)-th equations, we have the following equations for δuj+1,δvj+1 and δwj+1 :

    {tδuj+1=Δδuj+1χ(δuj+1vj)χ(ujδvj)ξ(δuj+1wj)ξ(ujδwj),tδvj+1=Δδvj+1δvj+1+δuj,[1mm]tδwj+1=vj+1δwj+1δvj+1wj+1. (2.28)

    (ⅰ) Estimates of δvj+1. Using the same procedure as proving (2.8), we can obtain that for all t(0,T)

    δvj+1(t)2H3+αt0(δvj+1(s)2H4+tδvj+1(s)2L2)dsec15Tc15Tsup0tTδuj(t)2H2. (2.29)

    (ⅱ) Estimates of δwj+1. According to the third equation of (2.28), we have for all t(0,T)

    δwj+1(t)=t0etsvj+1(τ)dτwj+1(s)δvj+1(s)ds. (2.30)

    Using the same procedure as proving (2.19) entails that for all t(0,T)

    δwj+1L2supt(0,T)wj+1Lsupt(0,T)δvj+1L2Tc16RTsupt(0,T)δvj+1L2,
    δwj+1L2t0wj+1(s)δvj+1(s)dsL2+t0wj+1(s)δvj+1(s)dsL2+t0wj+1δvj+1tsvj+1dτdsL2c17RTsupt(0,T)δvj+1L2+c17RTsupt(0,T)δvj+1L2+c17R2T2supt(0,T)δvj+1L2c18(R2+R)Tsupt(0,T)δvj+1H1,Δδwj+1L2=t0Δ{etsvj+1dτwj+1δvj+1}dsL2Tsupt(0,T)Δ{etsvj+1dτwj+1δvj+1}L2Tsupt(0,T){(vj+12H2T2+vj+1H2T)wj+1H2δvj+1H2}c19(R3+R2)Tsupt(0,T)δvj+1H2

    and

    3δwj+12L2ec20R2TTsupt(0,T)δvj+1H3,

    which imply that for all t(0,T)

    δwj+1H3c21(R3+R2+R)Tsup0tTδuj(t)H2. (2.31)

    by setting T<1.

    (ⅲ) Estimates of δuj+1. Using the same procedure as proving (2.26) entails that for all t(0,T)

    sup0tTδuj+12H2+βt0δuj+12H3ec22RTc22RTsup0tT(δvj2H3+δwj2H2). (2.32)

    Combining with (2.29), (2.31) and (2.32), we can obtain for all t(0,T)

    sup0tT(δuj+1H2+δvj+1H3+δwj+1H3)ec23(R3+R2+R)Tc23(R3+R2+R)Tsup0tT(δujH2+δvjH3+δwjH3). (2.33)

    Taking T>0 small enough, we can find a constant r(0,1) such that

    sup0tT(δuj+1H2+δvj+1H3+δwj+1H3)rsup0tT(δujH2+δvjH3+δwjH3) (2.34)

    for any j1 and t(0,T). From the above inequality, we find that (uj,vj,wj) is a Cauchy sequence in the Banach space C(0,T;X) for some small T>0, and thus its corresponding limit denoted by (u,v,w) definitely exists in the same space.

    Uniqueness: If (ˉu,ˉv,ˉw) is another local-in-time solution of system (1.3), (˜u,˜v,˜w):=(uˉu,vˉv,wˉw) solves

    {t˜u=Δ˜uχ(˜uˉv)χ(u˜v)ξ(˜uˉw)ξ(u˜w),xR2, 0<tT,t˜v=Δ˜v˜v+˜u,xR2, 0<tT,t˜w=v˜w˜vˉw,xR2, 0<tT,˜u(x,0)=˜v(x,0)=˜w(x,0)=0,xR2,

    where T is any time before the maximal time of existence. Following a same procedure as (2.34), we can deduce that ˜u=˜v=˜w=0, which implies the uniqueness of the local solution.

    Nonnegativity: The nonnegativity of wj can be easily obtained by (2.9) and the nonnegativity of w0. We will use the induction argument to show that uj and vj are nonnegative for all j>0. We assume that uj and vj are nonnegative. If we apply the maximum principle to the second equation of (2.2), we find vj+1 is nonnegative (uj is nonnegative). Then we turn to deal with uj+1. Let us decompose uj+1=uj+1+uj+1, where uj+1+={uj+1uj+100uj+1<0anduj+1={uj+1uj+100uj+1>0. Now multiplying the negative part uj+1 on both sides of the first equation of (2.2) and integrating over [0,t]×R2, we can get

    t0R2tuj+1uj+1dxds=t0(uj+1)2L2ds+χt0R2uj+1vjuj+1+ξR2uj+1wjuj+1Ct0(uj+1)2L2(vj2L+wj2L)+12(uj+1)2L2ds

    by Young's inequality and the fact the weak derivative of uj+1 is uj+1 if uj+1<0, otherwise zero. Since uj+1, tuj+1L2(0,T;L2(R2)), we can have

    t0R2tuj+1(uj+1)dxds=12((uj+1)(0)L2(uj+1)(0)L2),

    together with the above inequality, it holds that

    (uj+1)(t)2L2(uj+1)(0)2L2exp(Ct0(vj2L+wj2L)ds).

    Due to the fact uj+1(0) is nonnegative, we can deduce that uj+1 is nonnegative. This completes the proof of Lemma 2.1.

    Remark 2.1. Since the above choice of T depends only on u0H2(R2), v0H3(R2) and w0H3(R2), it is clear by a standard argument that (u,v,w) can be extended up to some Tmax. If Tmax< in Lemma 2.1, then

    lim suptTmax(u(t)H2(R2)+v(t)H3(R2)+w(t)H3(R2))=. (2.35)

    In order to show the H2×H3×H3-boundedness of (u,v,w), it suffices to estimate a suitable Lp-norm of u, with some large, but finite p.

    Lemma 2.2. Suppose that χ,ξ>0 and the initial data (u0,v0,w0) satisfy all the assumptions presented in Lemma 2.1. Then for every K>0 there is C>0 such that whenever (u,v,w)C(0,T;H2(R2)×H3(R2)×H3(R2)) solves (1.3) for some T>0 and q0>2 satisfies

    u(,t)Lq0K,for allt(0,T), (2.36)

    then

    u(t)H2(R2)+v(t)H3(R2)+w(t)H3(R2)Cfor allt(0,T). (2.37)

    Proof. Firstly, we suppose that for some q0>2 and K>0

    u(t)Lq0(R2)K,for all t(0,T). (2.38)

    By the Duhamel principle, we represent u and v of the following integral equations

    u(t)=etΔu0χt0e(tτ)Δ(uv)(τ)dτξt0e(tτ)Δ(uw)(τ)dτ (2.39)

    and

    v(t)=et(Δ+1)v0+t0e(tτ)(Δ+1)u(τ)dτ. (2.40)

    where etΔf(x)=R2G(xy,t)f(y)dy and

    G(x,t)=Gt(x):=1(4πt)exp(|x|24t)

    is the Gaussian heat kernel. The following well-known LpLq estimates of the heat semigroup play an important role in the proofs [47,48]. For 1pq and fLq(R2), we have

    etΔfLp(4πt)(1q1p)fLq,etΔfLpC3t12(1q1p)fLq,

    where C3 is a constant depending on p and q. Then, according to (2.39), we can see that for q0>2 and all t(0,T)

    u(t)Lu0L+χt0(tτ)121q0uvLq0dτ+ξt0(tτ)121q0uwLq0dτu0L+Kχt0(tτ)121q0vLdτ+Kξt0(tτ)121q0wLdτ. (2.41)

    From (2.40) and the above LpLq estimates of the heat semigroup, we have

    v(t)Lqv0Lq+t0e(tτ)(tτ)(1q01q)u(t)Lq0dτC4,q(1,] (2.42)

    and by the embedding H3(R2)W1,(R2)

    v(t)Lv0L+t0e(tτ)(tτ)121q0u(t)Lq0dτC5, (2.43)

    where C4 and C5 depend on v0H3 and K in (2.38).

    According to the equation of w, we can see that for some C6=C6(w0H3,v0H2,K,T)>0 and all t(0,T)

    w(t)Lw0et0v(s)dsL+w0et0v(s)dst0v(s)dsLw0H3+w0Lsupt(0,T)v(t)LTC6 (2.44)

    by the embedding H3(R2)W1,(R2). Inserting (2.43) and (2.44) into (2.41), this yields for all t(0,T)

    u(t)Lu0L+KC5χt0(tτ)121q0dτ+KC6ξt0(tτ)121q0dτC7, (2.45)

    where C7 depends on u0H2, v0H2, w0H3, K and T.

    Integrating by parts and by Young's inequality, we can obtain from the second equation of (1.3) that for all t(0,T)

    12ddttv(t)2L2+tv2L2=R2(tv)2+R2tvtu12tv2L2+12tu2L2 (2.46)

    and

    12ddtv(t)2L2+Δv2L2=R2vΔv+R2uΔvv2L2+12u2L2+12Δv2L2. (2.47)

    Similarly, according to the first equation of (1.3), (2.43) and (2.44), we have for all t(0,T)

    12ddtu(t)2L2+u2L2=χR2uvu+ξR2uwuχ22uv2L2+ξ22uw2L2+12u2L2χ2C252u2L2+ξ2C262u2L2+12u2L2 (2.48)

    and by (2.27) for some θ>0

    12ddttu(t)2L2+tu2L2=χR2tuvtu+χR2utvtu+ξR2tuwtu+ξR2utwtuC8(vL+wL)(θtu2L2+1θtu2L2)+C8uL(θtv2L2+θtw2L2+2θtu2L2)C8(C5+C6)θtu2L2+C8(C5+C6+2C7)θtu2L2+θC8C7(tv2L2+tw2L2). (2.49)

    Now we turn to estimate the last term of the right side of (2.49). According to the third equation of (1.3), (2.42) and (2.43), we obtain for some C9>0

    tw2L2v2L2w2L+w2L2v2Lw02Lv2L2+C24(w02L2+T2supt(0,T)v2Lw02L2)C9v2L2+C9. (2.50)

    Combining with (2.46)–(2.50) and setting θ>0 to satisfy C8(C5+C6+2C7)θ<12, we can obtain such Gronwall-type inequality

    ddt{tv(t)2L2+v(t)2L2+u(t)2L2+tu(t)2L2}+C10(tv2L2+u2L2+v2L2+Δv2L2+tu2L2)C10(tv2L2+v2L2+u2L2+tu2L2+1),for all t(0,T), (2.51)

    then by direct integration, we can have for some C11=C11(w0H3,v0H2,u0H2,K,T)>0

     t0(tv2L2+u2L2+v2L2+Δv2L2+tu2L2) +{tv2L2+v2L2+u2L2+tu2L2}C11,for all t(0,T). (2.52)

    By (2.42) and (2.52) and the second equation of v in (1.3), we have

    ΔvL2tvL2+vL2+uL2C12,for all t(0,T), (2.53)

    where C12=C12(w0H3,v0H2,u0H2,K,T). Hence by the equation of w and Young's inequality, we obtain for some C13=C13(w0H3,v0H2,u0H2,K,T)>0

    ΔwL2Δw0L2+2w0L4t0vL4+w0L|t0v|2L2+w0Lt0ΔvL2Δw0L2+2w0L4supt(0,T)vL4T+w0Lsupt(0,T)v2L4T+w0Lsupt(0,T)ΔvL2TC13,for all t(0,T). (2.54)

    By (2.42) and (2.43) and the embedding W2,2(R2)W1,4(R2), we can see

    ddt3w(t)2L2C{3w2L2vL+3wL22wL4vL4+3wL22vL2wL+3wL2wL3vL2}C143w2L2+C143v2L2+C14,for all t(0,T). (2.55)

    Integrating on (0,t), we have

    3w2L2C153w02L2+C15t03v2L2+C15,for all t(0,T). (2.56)

    Now we turn to estimate the second integral of the right side of (2.56). Applying to the second equation of (1.3) and rewriting the equation as Δv=vt+vu, then by (2.52) we have that

    t03v2L2t0vt2L2+t0v2L2+t0u2L2C11,for all t(0,T). (2.57)

    Inserting (2.57) into (2.56), we can obtain that for some C16>0

    3wL2C16,for all t(0,T). (2.58)

    Now we deduce the L2-norm of u and 2u. We multiply the first equation of (1.3) by Δu, integrate by parts and then obtain

    12ddtu(t)2L2+Δu2L2=χR2Δu(uv)+ξR2Δu(uw)=χR2Δuuv+χR2uΔuΔv+ξR2Δuuw+ξR2uΔuΔw=I1+I2+I3+I4. (2.59)

    Then by (2.42), (2.45) and (2.52), Hölder's inequality and Young's inequality, we have

    I1+I2χvLΔuL2uL2+χΔvL2ΔuL2uL14Δu2L2+C17u2L2+C17.

    Similarly, according to (2.43), (2.45) and (2.56), we can obtain

    I3+I4χvLΔuL2uL2+χΔvL2ΔuL2uL14Δu2L2+C17u2L2+C18.

    Then we have

    12ddtu(t)2L2+12Δu2L2C19u2L2+C19.

    Integrating on (0,t), we have for some C20>0

    uL2C20,for all t(0,T). (2.60)

    Rewriting the first equation of (1.3) as Δu=ut+χ(uv)+ξ(uw), and by (2.42)–(2.44), (2.53) and (2.54), we have for some C21>0

    ΔuL2tuL2+χuvL2+χuΔvL2+ξuwL2+ξuΔwL2tuL2+χuL2vL+χΔvL2uL+ξuL2wL+ξΔwL2uLC21. (2.61)

    For the L2-norm of 3v, integrating by parts, we deduce that

    ddt3v(t)2L2=2R23v3(Δvv+u)=2R2|4v|22R2|3v|2+2R23v3u4v2L223v2L2+2u2L2,

    then by (2.61) and Gronwall's inequality, we can see that for all t(0,T)

    3v(t)L2C22. (2.62)

    Putting (2.52)–(2.54), (2.58) and (2.60)–(2.62) together, we conclude that for some C>0

    u(t)H2(R2)+v(t)H3(R2)+w(t)H3(R2)C,t(0,T), (2.63)

    which completes the proof.

    As a preparation, we first state some results concerning the system which will be used in the proof of Theorem 1.1.

    Lemma 3.1. The local-in-time classical solution (u,v,w) of system (1.3) satisfies

    u(t)L1=u0L1:=M,t(0,Tmax) (3.1)

    and

    v(t)L1=u0L1+(v0L1u0L1)et,t(0,Tmax). (3.2)

    Proof. Integrating the first and second equation of (1.3) on R2, we can obtain

    ddtR2u=R2ΔuχR2(uv)ξR2(uw)=0

    and

    ddtR2v=R2ΔvR2v+R2u=R2v+R2u,

    which can easily yield (3.1) and (3.2).

    The following energy

    F(t)=R2ulnuχR2uvξR2uw+χ2R2(v2+|v|2)

    plays a key role in the proof. The main idea of the proof is similar to the strategy introduced in [2].

    Lemma 3.2. Assume that (1.4) and (1.5) holds. Let (u,v,w) be the local-in-time classical solution of system (1.3). Then F(t) satisfies

    F(t)+χt0R2v2t+t0R2u|(lnuχvξw)|2=F(0)+ξt0R2uvw,t(0,Tmax). (3.3)

    Proof. We use the same ideas as in the proofs of [45,Theorem 1.3], [46,Lemma 3.1] and [1,Theorem 3.2]. The equation of u can be written as ut=(u(lnuχvξw)). Multiplying by lnuχvξw and integrating by parts, we obtain

    R2u|(lnuχvξw)|2=R2ut(lnuχvξw)=ddtR2(ulnuχuvξuw)+χR2uvt+ξR2uwt. (3.4)

    Substituting the second and third equation of (1.3) into (3.4) and integrating by parts, we have

    R2u|(lnuχvξw)|2=ddtR2(ulnuχuvξuw)+χR2(vtΔv+v)vt+ξR2uvw=ddtR2(ulnuχuvξuw)χR2v2t+χ2ddtR2(v2+|v|2)+ξR2uvw, (3.5)

    which, upon being integrated from 0 to t, yields simply that (3.3). We give some lemmas to deal with the term R2ulnu in (1.7).

    Lemma 3.3. ([1,Lemma 2.1]) Let ψ be any function such that eψL1(R2) and denote ˉu=Meψ(R2eψdx)1 with M a positive arbitrary constant. Let E:L1+(R2)R{} be the entropy functional

    E(u;ψ)=R2(ulnuuψ)dx

    and let RE:L1+(R2)R{} defined by

    RE(uˉu)=R2uln(uˉu)dx

    be the relative (to ˉu) entropy.

    Then E(u;ψ) and RE(uˉu) are finite or infinite in the same time and for all u in the set U={uL1+(R2),R2u(x)dx=M} and it holds true that

    E(u;ψ)E(ˉu;ψ)=RE(uˉu)0.

    Next, we give a Moser-Trudinger-Onofri inequality.

    Lemma 3.4. ([1,Lemma 3.1]) Let H be defined as H(x)=1π1(1+|x|2)2. Then

    R2eφ(x)H(x)dxexp{R2φ(x)H(x)dx+116πR2|φ(x)|2dx}, (3.6)

    for all functions φL1(R2,H(x)dx) such that |φ(x)|L2(R2,dx).

    Lemma 3.5. ([1,Lemma 2.4]) Let ψ be any function such that eψL1(R2), and let f be a non-negative function such that (f1{f1})L1(R2)L1(R2,|ψ(x)|dx). Then there exists a constant C such that

    R2f(x)(lnf(x))dxC{f1}f(x)ψ(x)dx.

    With the help of Lemma 3.2–3.5, we now use the subcritical mass condition (1.8) to derive a Gronwall-type inequality and to get a time-dependent bound for (ulnu)(t)L1.

    Lemma 3.6. Under the subcritical mass condition (1.8) and (1.5), there exists C=C(u0,v0,w0)>0 such that

    (ulnu)(t)L1+v(t)2H1CeξKγt,t(0,Tmax), (3.7)

    where K>0 and γ are defined by (3.8) and (3.10) below, respectively.

    Proof. According to the third equation of (1.3), we have for all t(0,T)

    wLw0L:=K, (3.8)

    then we apply the estimate of (3.3) to find that

    F(t)+t0R2u|(lnuχvξw)|2F(0)+ξKt0R2uv,t(0,Tmax). (3.9)

    For our later purpose, since M<8πχ, we first choose some positive constant γ>0 small enough to satisfy

    χM(χ+γ)28π>0, (3.10)

    then by the definition of F(t) in (1.7), we use (3.1) and (3.8) to deduce that

    F(t)=R2ulnuχR2uvξR2uw+χ2R2(v2+|v|2)R2ulnu(χ+γ)R2uvξKM+χ2R2(v2+|v|2)+γR2uv. (3.11)

    Similar as the calculation shown in [1], we set ˉu(x,t)=Me(χ+γ)v(x,t)H(x)(R2e(χ+γ)v(x,t)H(x)dx)1, where H(x) is defined in Lemma 3.4. Then, we can apply the Entropy Lemma 3.3 with ψ=(χ+γ)v+lnH to obtain

    E(u;(χ+γ)v+lnH)E(ˉu;(χ+γ)v+lnH)=MlnMMln(R2e(χ+γ)v(x,t)H(x)dx). (3.12)

    Furthermore, applying Lemma 3.4 with φ=(χ+γ)v to the last term in the right hand side of (3.12), we have that

    E(u;(χ+γ)v+lnH)=R2ulnu(χ+γ)R2uvR2ulnHMlnMMln(R2e(χ+γ)v(x,t)H(x)dx)MlnMM(χ+γ)R2vHM(χ+γ)216πR2|v|2. (3.13)

    Then by Young's inequality, we have M(χ+γ)R2vHM(χ+γ)216πR2v2+4MπR2H2. Together with (3.13) and the fact R2H2(x)dx=13π, we can easily obtain

    R2ulnu(χ+γ)R2uvR2ulnHMlnMM(χ+γ)R2vHM(χ+γ)216πR2|v|2MlnMM(χ+γ)216πR2v2M(χ+γ)216πR2|v|243M. (3.14)

    Substituting (3.14) into (3.11), we have

    F(t)R2ulnu(χ+γ)R2uv+χ2R2(v2+|v|2)+γR2uvξKMMlnM+R2ulnH+(χ2M(χ+γ)216π)R2(v2+|v|2)+γR2uv(ξK+43)MMlnM+R2ulnH+γR2uv(ξK+43)M (3.15)

    by (3.10). Now we turn to estimate the second term on the right side of (3.15). We set ϕ(x)=ln(1+|x|2), then we can obtain by Young's inequality

     ddtR2uϕ=R2utϕ=R2uϕ(lnuχvξw)R2u|ϕ|2+14R2u|(lnuχvξw)|2,for all t(0,Tmax).

    By the fact |ϕ(x)|=|2x1+|x|2|1, we have

    ddtR2uϕR2u+14R2u|(lnuχvξw)|2M+14R2u|(lnuχvξw)|2,

    upon being integrated from 0 to t, which yields simply that for all t(0,Tmax)

    R2uln(1+|x|2)R2u0ln(1+|x|2)+Mt+14t0R2u|(lnuχvξw)|2. (3.16)

    By the definition of H(x), we have for all t(0,Tmax)

    R2ulnH=2R2uln(1+|x|2)Mlnπ2R2u0ln(1+|x|2)2Mt12t0R2u|(lnuχvξw)|2Mlnπ. (3.17)

    Substituting (3.15) and (3.17) into (3.9), we have for all t(0,Tmax)

    γR2uvξKt0R2uv+2Mt+F(0)+2R2u0ln(1+|x|2)+(lnπ+ξK+43lnM)M. (3.18)

    From (1.4), we have assumed for convenience that u0lnu0 and u0ln(1+|x|2) belongs to L1(R2) for convenience. Then we conclude an integral-type Gronwall inequality as follows

    γR2uvξKt0R2uv+2Mt+C1,t(0,Tmax), (3.19)

    where C1=F(0)+2R2u0ln(1+|x|2)+(lnπ+ξK+43lnM)M is a finite number. Solving the integral-type Gronwall inequality (3.19) via integrating factor method, we infer that for some C2>0

    R2uv+t0R2uvC2eξKγt,t(0,Tmax).

    Then by (3.9), one can simply deduce that F(t) grows no great than exponentially as well:

    F(t)C3eξKγt,t(0,Tmax). (3.20)

    Similarly, this along with (1.7) shows that for some C4>0

    R2ulnu+R2v2+R2|v|2C4eξKγt,t(0,Tmax). (3.21)

    According to Lemma 3.5 with ψ=(1+δ)ln(1+|x|2), for arbitrary δ>0 in order to have e(1+δ)ln(1+|x|2)L1(R2), we have for all t(0,Tmax)

    R2u(lnu)dx(1+δ)R2uln(1+|x|2)dx+C5(1+δ){R2u0ln(1+|x|2)+Mt+14t0R2u|(lnuχvξw)|2}+C51+δ4{F(0)F(t)+ξKt0R2uv}+M(1+δ)t+C6C7eξKγt (3.22)

    for some Ci>0 (i=5,6,7). Finally, the identity

    R2|ulnu|dx=R2ulnudx+2R2u(lnu)dx (3.23)

    gives that (ulnu)(t)L1C8eξKγt for some C8>0. Together with (3.21), this easily yield (3.7).

    Next, we wish to raise the regularity of u based on the local L1-boundedness of ulnu. In particular, for subcritical mass M, we have R2(u(x,t)k)+dxM for any k>0, while for k>1 we have for all t(0,Tmax)

    R2(u(x,t)k)+dx1lnkR2(u(x,t)k)+lnu(x,t)dx1lnkR2u(x,t)(lnu(x,t))+dxCeξKγtlnk. (3.24)

    Lemma 3.7. Under the condition (1.5) and (1.8), for any T(0,Tmax), there exists C(T)>0 such that the local solution (u,v,w) of (1.1) verifies that for any p2

    R2up(x,t)dxC(T), t(0,T], (3.25)

    where C(T)=2pˉC(T)+(2k)p1M with k and ˉC(T) respectively given by (3.37) and (3.40) below, which are finite for any T>0.

    Proof. Let k>0, to be chosen later. We derive a non-linear differential inequality for the quantity Yp(t):=R2(u(x,t)k)p+dx, which guarantees that the Lp-norm of u remains finite.

    Multiplying the equation of u in (1.3) by p(uk)p1+ yields, using integration by parts,

    ddtR2(uk)p+dx=4(p1)pR2|(uk)p2+|2dx(p1)χR2(uk)p+ΔvdxpkχR2(uk)p1+Δvdx(p1)ξR2(uk)p+ΔwdxpkξR2(uk)p1+Δwdx=I1+I2+I3+I4+I5. (3.26)

    Now using the equation of v in (1.3) and the nonnegativity of v, one obtains

    I2=(p1)χR2(uk)p+Δvdx=(p1)χR2(uk)p+(vtv+u)dx(p1)χR2(uk)p+vt+(p1)χR2(uk)p+1+dx+(p1)kχR2(uk)p+dx (3.27)

    and

    I3=pkχR2(uk)p1+Δvdx=pkχR2(uk)p1+(vtv+u)dxpkχR2(uk)p1+vtdx+pkχR2(uk)p+dx+pk2χR2(uk)p1+dx. (3.28)

    Using Gagliardo-Nirenberg inequality R2f4(x)dxCR2f2(x)dxR2|f(x)|2dx with f=(uk)p2+ and Hölder inequality, we obtain for ε>0

    |R2(uk)p+vtdx|(R2(uk)2p+dx)12vtL2C(R2(uk)p+dx)12(R2|(uk)p2+|2dx)1/2vtL2C(p)εvt2L2R2(uk)p+dx+2εpR2|(uk)p2+|2dx. (3.29)

    Similarly, we have, for p32

    |R2(uk)p1+vtdx|(R2(uk)2(p1)+dx)12vtL2(C(M,p)+C(p)R2(uk)2p+dx)12vtL2C(M,p)vtL2+C(p)εvt2L2R2(uk)p+dx+p1εp2kR2|(uk)p2+|2dx. (3.30)

    Then we can see that

    I2+I3(p1)R2(uk)p+1+dx+(p1)pR2|(uk)p2+|2dx +C(p,χ)(k+1)vt2L2R2(uk)p+dx+C(M,p,χ)kvtL2 +(2p1)kχR2(uk)p+dx+pk2χR2(uk)p1+dx (3.31)

    by setting ε=4χ. According to the equation of w and v and (3.8), one obtains for all t(0,T)

    Δw(x,t)=Δw0(x)et0v(x,s)ds+2et0v(x,s)dsw0(x)t0v(x,s)dsw0(x)et0v(x,s)ds|t0v(x,s)ds|2+w0(x)et0v(x,s)dst0Δv(x,s)dsΔw0Let0v(x,s)ds(w0t0v(x,s)dsw0w0)2+et0v(x,s)ds|w0|2w0+w0(x)et0v(x,s)dst0(vs(x,s)+vu)ds. (3.32)

    Here to estimate the last integral of the right side of (3.32) we first note (1.7) guarantees that

    w0(x)et0v(x,s)dst0(vs(x,s)+vu)dsw0Let0v(x,s)ds[v(x,t)v0+t0v(x,s)ds]w0Lv+w0Le,t(0,T)

    by the nonnegativity of w0 and v0 and the fact exx1e for all x>0. Substituting (3.8) and (3.32) into (3.26), we have

    I4+I5=(p1)ξR2(uk)p+ΔwdxpkξR2(uk)p1+Δwdx(p1)KξR2(uk)p+vdx+(p1)K1ξR2(uk)p+dx+pkKξR2(uk)p1+vdx+pkK1ξR2(uk)p1+dx,

    where K1=Δw0L+4w02L+Ke. Applying similar procedure as (3.29) and (3.30) to R2(uk)p+vdx and R2(uk)p1+vdx, this yields

    I4+I5(p1)pR2|(uk)p2+|2dx+C(p,K,ξ)kv2L2R2(uk)p+dx+C(M,p,K,ξ)kvL2+(p1)K1ξR2(uk)p+dx+pkK1ξR2(uk)p1+dx (3.33)

    by setting ε=2Kξ. Combining (3.26), (3.31) and (3.33), we have for all t(0,T)

    ddtR2(uk)p+dx2(p1)pR2|(uk)p2+|2dx+(p1)R2(uk)p+1+dx+[(2p1)kχ+(p1)K1ξ]R2(uk)p+dx+(pk2χ+pkK1ξ)R2(uk)p1+dx+C(p,K,χ,ξ)(k+1)(tv2L2+v2L2)R2(uk)p+dx+C(M,p,K,χ,ξ)k(tvL2+vL2). (3.34)

    Next, we estimate the nonlinear and negative contribution 2(p1)pR2|(uk)p2+|2dx in terms of R2(uk)p+1+dx, with the help of the Sobolev's inequality f2L2c1f2L1. Indeed, by (3.24),

    R2(uk)p+1+dx=R2((uk)(p+1)2+)2dxc1(R2|(uk)(p+1)2+|dx)2=C(p)(R2(uk)12+|(uk)p2+|dx)2C(p)R2(uk)+dxR2|(uk)p/2+|2dxC(p)CeξKγTlnkR2|(uk)p/2+|2dx,0<tT. (3.35)

    Moreover, since for p2 it holds true that

    R2(uk)p1+dxR2(uk)+dx+R2(uk)p+dx. (3.36)

    Inserting (3.35) and (3.36) into (3.34) gives for p2 and 0<tT that

    ddtR2(uk)p+dx(p1)(12lnkpC(p)CeξKγT)R2(uk)p+1+dx+C(p,K,χ,ξ)k(1+tv2L2+v2L2)R2(uk)p+dx+C(M,p,K,χ,ξ)k(tvL2+vL2+1).

    For any fixed p we can choose k=k(p,T) sufficiently large such that

    δ:=2lnkpC(p)CeξKγT1>0, (3.37)

    namely, k=exp((1+δ)pC(p)CeξKγT2). For such a k, using the interpolation

    R2(uk)p+dx(R2(uk)+dx)1p(R2(uk)p+1+dx)(11p)M1p(R2(uk)p+1+dx)(11p),

    we end up with the following differential inequality for Yp(t), p2 fixed and 0<tT

    ddtYp(t)(p1)M1p1δYβp(t)+c2(p,K,χ,ξ)k(1+tv2L2+v2L2)Yp(t)+c3(M,p,K,χ,ξ)k(1+tv2L2+v2L2), (3.38)

    where β=pp1>1. Let us write the differential inequality (3.38) as follows for simplicity:

    ddtYp(t)˜CYβp(t)+g(t)Yp(t)+g(t),0<tT, (3.39)

    where g(t)=ˉC(M,p,K,χ,ξ)k(1+tv2L2+v2L2) and ˜C=(p1)M1p1δ>0. According to (3.7), (3.9) and (3.20), we can see that g(t)ˉC(M,p,K,χ,ξ)keξKtγ. Then by comparison inequality, we show that there exists a constant ˉC(T) such that for all t(0,T)

    Yp(t)Yp(0)exp(t0g(s)ds)+t0g(τ)exp(tτg(s)ds)dτYp(0)ˉC(M,p,K,χ,ξ)keξKTγT+ˉC(M,p,K,χ,ξ)keξKTγeˉC(M,p,K,χ,ξ)keξKTγT:=ˉC(T). (3.40)

    It is sufficient to observe that for any k>0

    R2up(x,t)dx={u2k}up(x,t)dx+{u>2k}up(x,t)dx(2k)p1M+2p{u>2k}(u(x,t)k)pdx(2k)p1M+2pR2(u(x,t)k)p+dx, (3.41)

    where the inequality xp2p(xk)p, for x2k, has been used. Therefore, (3.25) follows for any p2 by (3.40) and (3.41) choosing k=k(p,T) sufficiently large such that (3.37) holds true.

    Proof of Theorem 1.1. According to the local Lpboundedness of Lemma 3.7 and Lemma 2.2 we must have the local H2×H3×H3-boundedness of (u,v,w), which contracts the extensibility criteria in (2.35). Then we must obtain that Tmax=, that is, the strong solution (u,v,w) of (1.3) exists globally in time and is locally bounded as in (2.2).

    The authors convey sincere gratitude to the anonymous referees for their careful reading of this manuscript and valuable comments which greatly improved the exposition of the paper. The authors are supported in part by National Natural Science Foundation of China (No. 12271092, No. 11671079).

    The authors declare there is no conflict of interest.



    [1] L. Alon, R. Band, G. Berkolaiko, Universality of nodal count distribution in large metric graphs, Exp Math, (2022), 1–35. https://doi.org/10.1080/10586458.2022.2092565
    [2] G. Bastin, J. M. Coron, Exponential stability of networks of density-flow conservation laws under PI boundary control, IFAC Proceedings Volumes, 46 (2013), 221–226. https://doi.org/10.3182/20130925-3-FR-4043.00029
    [3] G. Bastin, B. Haut, J. M. Coron, B. d'Andrea-Novel, Lyapunov stability analysis of networks of scalar conservation laws, Netw. Heterog. Media, 2 (2007), 751–759. https://doi.org/10.3934/nhm.2007.2.751 doi: 10.3934/nhm.2007.2.751
    [4] A. Bayen, J. Friedrich, A. Keimer, L. Pflug, T. Veeravalli, Modeling multilane traffic with moving obstacles by nonlocal balance laws, SIAM J. Appl. Dyn. Syst., 21 (2022), 1495–1538. https://doi.org/10.1137/20M1366654 doi: 10.1137/20M1366654
    [5] A. Bayen, A. Keimer, E. Porter, M. Spinola, Time-continuous instantaneous and past memory routing on traffic networks: A mathematical analysis on the basis of the link-delay model, SIAM J. Appl. Dyn. Syst., 18 (2019), 2143–2180. https://doi.org/10.1137/19M1258980 doi: 10.1137/19M1258980
    [6] G. Boeing, Osmnx: New methods for acquiring, constructing, analyzing, and visualizing complex street networks, Comput Environ Urban Syst, 65 (2017), 126–139. https://doi.org/10.1016/j.compenvurbsys.2017.05.004 doi: 10.1016/j.compenvurbsys.2017.05.004
    [7] D. Braess, Über ein Paradoxon aus der Verkehrsplanung, Unternehmensforschung, 12 (1968), 258–268. https://doi.org/10.1007/BF01918335 doi: 10.1007/BF01918335
    [8] A. Bressan, Hyperbolic Systems of Conservation Laws: The One-dimensional Cauchy Problem, Oxford: Oxford University Press, 2000.
    [9] A. Bressan, S. Čanić, M. Garavello, M. Herty, B. Piccoli, Flows on networks: recent results and perspectives, EMS Surv Math SCI, 1 (2014), 47–111. https://doi.org/10.4171/emss/2 doi: 10.4171/emss/2
    [10] G. M. Coclite, M. Garavello, Vanishing viscosity for traffic on networks, SIAM J. Math. Anal., 42 (2010), 1761–1783. https://doi.org/10.1137/090771417 doi: 10.1137/090771417
    [11] G. M. Coclite, M. Garavello, B. Piccoli, Traffic flow on a road network, SIAM J. Math. Anal., 36 (2005), 1862–1886. https://doi.org/10.1137/S0036141004402683 doi: 10.1137/S0036141004402683
    [12] M. Düfel, J. B. Kennedy, D. Mugnolo, M. Plümer, M. Täufer, Boundary conditions matter: On the spectrum of infinite quantum graphs, arXiv: 2207.04024, [Preprint], (2022), [cited 2023 Nov 13]. Available from: https://doi.org/10.48550/arXiv.2207.04024
    [13] M. Garavello, K. Han, B. Piccoli, Models for Vehicular Traffic on Networks, Springfield Missouri: American Institute of Mathematical Sciences, 2016.
    [14] M. Garavello, B. Piccoli, Conservation laws on complex networks, Ann. Inst. Henri Poincare (C) Anal. Non Lineaire, 26 (2009), 1925–1951. https://doi.org/10.1016/j.anihpc.2009.04.001 doi: 10.1016/j.anihpc.2009.04.001
    [15] F. R. Guarguaglini, R. Natalini, Global smooth solutions for a hyperbolic chemotaxis model on a network, SIAM J. Math. Anal., 47 (2015), 4652–4671. https://doi.org/10.1137/140997099 doi: 10.1137/140997099
    [16] M. Gugat, M. Herty, A. Klar, G. Leugering, Optimal control for traffic flow networks, J Optim Theory Appl, 126 (2005), 589–616. https://doi.org/10.1007/s10957-005-5499-z doi: 10.1007/s10957-005-5499-z
    [17] M. Gugat, Nodal control of conservation laws on networks, In: J. Cagnol, J. P. Zolesio (Eds.) Control and boundary analysis, Boca Raton: CRC Press, 2005, 221–236.
    [18] M. Gugat, A. Keimer, G. Leugering, Z. Wang, Analysis of a system of nonlocal conservation laws for multi-commodity flow on networks, Netw. Heterog. Media, 10 (2015), 749–758. https://doi.org/10.3934/nhm.2015.10.749 doi: 10.3934/nhm.2015.10.749
    [19] H. Holden, N. H. Risebro, A mathematical model of traffic flow on a network of roads, SIAM J. Math. Anal., 26 (1995), 999–1017. https://doi.org/10.1137/S0036141093243289 doi: 10.1137/S0036141093243289
    [20] J. Jost, X. Li-Jost, Calculus of Variations, Cambridge: Cambridge University Press, 1998.
    [21] G. Kirchhoff, Über die Auflösung der Gleichungen, auf welche man bei der Untersuchung der linearen Vertheilung galvanischer Ströme geführt wird, Annalen der Physik, 148 (1847), 497–508.
    [22] T. Kottos, U. Smilansky, Periodic orbit theory and spectral statistics for quantum graphs, Ann Phys (N Y), 274 (1999), 76–124. https://doi.org/10.1006/aphy.1999.5904 doi: 10.1006/aphy.1999.5904
    [23] P. Kuchment, Quantum graphs: I. Some basic structures, Waves in Random Media, 14 (2003), S107. https://doi.org/10.1088/0959-7174/14/1/014 doi: 10.1088/0959-7174/14/1/014
    [24] N. Laurent-Brouty, A. Keimer, P. Goatin, A. M. Bayen, A macroscopic traffic flow model with finite buffers on networks: well-posedness by means of hamilton–jacobi equations, Commun Math Sci, 18 (2020), 1569–1604. https://dx.doi.org/10.4310/CMS.2020.v18.n6.a4 doi: 10.4310/CMS.2020.v18.n6.a4
    [25] D. Mugnolo, Semigroup Methods for Evolution Equations on Networks, Cham: Springer, 2014.
    [26] D. Mugnolo, What is actually a metric graph? arXiv: 1912.07549 [Preprint], (2022), [cited 2023 Nov 13]. Available from: https://doi.org/10.48550/arXiv.1912.07549
    [27] D. Mugnolo, S. Romanelli, Dynamic and generalized wentzell node conditions for network equations, Math. Methods Appl. Sci, 30 (2007), 681–706. https://doi.org/10.1002/mma.805 doi: 10.1002/mma.805
    [28] L. O. Müller, P. J. Blanco, A high order approximation of hyperbolic conservation laws in networks: Application to one-dimensional blood flow, J. Comput. Phys., 300 (2015), 423–437. https://doi.org/10.1016/j.jcp.2015.07.056 doi: 10.1016/j.jcp.2015.07.056
    [29] M. Musch, U. S. Fjordholm, N. H. Risebro, Well-posedness theory for nonlinear scalar conservation laws on networks, Netw. Heterog. Media, 17 (2022), 101–128. https://doi.org/10.3934/nhm.2021025 doi: 10.3934/nhm.2021025
    [30] OpenStreetMap Foundation board, OpenStreetMap contributors, Available from: https://planet.osm.org.
    [31] L. Pauling, The diamagnetic anisotropy of aromatic molecules, J. Chem. Phys., 4 (1936), 673–677. https://doi.org/10.1063/1.1749766 doi: 10.1063/1.1749766
    [32] J. R. Platt, Classification of spectra of cata‐condensed hydrocarbons, J. Chem. Phys., 17 (1949), 484–495. https://doi.org/10.1063/1.1747293 doi: 10.1063/1.1747293
    [33] M. Richardson, N. Balazs, On the network model of molecules and solids, Ann Phys (N Y), 73 (1972), 308–325. https://doi.org/10.1016/0003-4916(72)90285-0 doi: 10.1016/0003-4916(72)90285-0
    [34] C. Wheatstone, XIII. The Bakerian lecture.–An account of several new instruments and processes for determining the constants of a voltaic circuit, Philos. Trans. R. Soc. London, 133 (1843), 303–327. https://doi.org/10.1098/rstl.1843.0014 doi: 10.1098/rstl.1843.0014
  • This article has been cited by:

    1. Yuanlin Chen, Tian Xiang, Negligibility of haptotaxis on global dynamics in a chemotaxis-haptotaxis system with indirect signal production, 2024, 409, 00220396, 1, 10.1016/j.jde.2024.06.034
    2. Poonam Rani, Jagmohan Tyagi, A quasilinear chemotaxis-haptotaxis system: Existence and blow-up results, 2024, 402, 00220396, 180, 10.1016/j.jde.2024.04.034
    3. Poonam Rani, Jagmohan Tyagi, Finite-time blow-up in the higher dimensional parabolic-parabolic-ODE minimal chemotaxis-haptotaxis system, 2025, 423, 00220396, 133, 10.1016/j.jde.2024.12.030
  • Reader Comments
  • © 2023 the Author(s), licensee AIMS Press. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0)
通讯作者: 陈斌, bchen63@163.com
  • 1. 

    沈阳化工大学材料科学与工程学院 沈阳 110142

  1. 本站搜索
  2. 百度学术搜索
  3. 万方数据库搜索
  4. CNKI搜索

Metrics

Article views(1311) PDF downloads(90) Cited by(0)

Figures and Tables

Figures(6)

Other Articles By Authors

/

DownLoad:  Full-Size Img  PowerPoint
Return
Return

Catalog