Spectral analysis for stability of bubble steady states of a Keller-Segel's minimal chemotaxis model (Q321865)

This paper is a continuation of the work [\textit{X. Lai} et al., Discrete Contin. Dyn. Syst. 36, No. 2, 805--832 (2016; Zbl 1326.92010)]. Eigenvalue problem arising from the stability analysis of bubble steady states of a Keller-Segel's minimal chemotaxis model \[ \begin{cases} u_x=kf(u)v_x>0,&\text{ in }\Omega\\ v_{xx}-v+g(u)=0,&\text{ in }\Omega\\ u_x=v_x=0,&\text{ on }\partial\Omega\\ \int_\Omega u(x)\,dx=m. \end{cases}\eqno(1) \] where \(\Omega=(0,l)\), \(k\gg1\), \(0<m<l\), \(f\) and \(g\) are given functions satisfying conditions: (f) \(f\in C^3([0,1])\), \(f(0)=f(1)=0\), \(f>0\) in \((0,1)\), \(f'(0)>0>f'(1)\); (g) \(g\in C^2([0,1])\), \(g(0)=0\), \(g(1)=1\), \(g'>0\) in \((0,1)\) is studied. Main results are formulated is tree theorems. Theorem 1. For each \(k\gg1\), let \((u(k, x), v(k, x))\) be the unique solution of (1) and \(\{\mu_i\}_{i=1}^\infty\) (in increasing order) be all eigenvalues of the problem: \(-(pw_x)_x=\mu pw \text{ in } \Omega;\, w_x=0 \text{ on } \partial\Omega\) with \(p(k,x)=kf(u(k, x))\). Then \[ \mu_1=0,\quad \mu_2=2k\min\{b1, b2\} + \frac12+ \frac{O(1)}k, \quad \mu_4\geq6k\min\{b1, b2\} + \frac92+ \frac{O(1)}k, \] \[ \mu_3=\min\big\{2k\max\{b1, b2\}+\frac12, 6k\min\{b1, b2\}+ \frac92\big\}+ \frac{O(1)}k, \] where \(O(1)\) stands for a quantity bounded by a constant that does not depend on \(k\), \(b1=f'(0)v(k, 0)/2\), \(b2=|f'(1)|[1-v(k,l)]/2\), and \(v(k,0)\) and \(1-v(k,l)\) are given. Theorem 2. Assume the same assumptions as in Theorem 1 and let \(\lambda_1\) be the eigenvalue of the problem \[ \begin{cases} \tau\lambda\phi=(\phi_x-kf(u)\psi_x- kf'(u)v_x\phi)_x,&\text{ in }\Omega\\ \lambda\psi=\psi_{xx}-\psi+g'(u)\phi,&\text{ in }\Omega\\ \phi_x=\psi_x=0,&\text{ on }\partial\Omega\\ \int_\Omega \phi\,dx=0. \end{cases} \] with largest real part. There exist positive constants \(\sigma\), \(K\), and \(M\) such that for each fixed \(k\gg K\), the following holds: (i) if \(0\leq\tau<\mu_2-Me^{-\sigma k}\), then \(\lambda_1=-1\); (ii) if \(\tau>\mu_2+Me^{-\sigma k}\), then \(|\lambda_1+\mu_2/\tau|\leq M e^{-\sigma k}/\tau\); and (iii) if \(\tau\in[\mu_2-Me^{-\sigma k}, \mu_2+Me^{-\sigma k}]\), then \(|\lambda_1+ 1|\leq2Me^{-\sigma k}/\tau\). Theorem 3. For each positive integer \(i\), let \(U_i\) be an eigenfunction of the problem: \(LU=\mu U\) in \(\Omega\); \(U_x=0\) on \(\partial\Omega\), where \(LU=-U_{xx}+\frac{(\sqrt{p})_{xx}} {\sqrt{p}}U\) with respect to the \(i\)th eigenvalue \(\mu_i\), normalized by \(\|U_i\|_{L^2(\Omega)}=1\). There exists a positive constant \(\sigma\) that does not depend on \(k\) such that \[ \int_{\Omega}\sqrt{p}[|U_2|+|U_3|]\,dx=O(e^{-\sigma k}). \]