Blow-up prevention by sub-logistic sources in 2d Keller-Segel system

In this paper, we consider the following chemotaxis model with sub-logistic sources in a smooth bounded domain $\Omega\subset\mathbb{R}^{n}$, where $n\geq 2$:

where $r,\mu$ are positive parameters, and $f$ is a smooth function generalizing the sub-logistic and signal production source respectively,

The system (1.1) is complemented with nonnegative initial conditions in $W^{1,\infty}(\Omega)$ not identically zero:

and homogeneous Neumann boundary condition are imposed as follows:

where $\nu$ denotes the outward normal vector.
The study of chemotaxis, which is the phenomenon where cells or bacteria move towards a chemical signal, has been a topic of intense research since the 1970s. Chemotaxis plays a significant role in various fields, including predicting the formation of aggregations, navigating optimal paths in a complex network, and even in physics, such as particle interaction. Moreover, it presents an intriguing mathematical property known as the critical mass phenomenon. This phenomenon means that if the mass is strictly less than a certain number, solutions exist globally, while if the mass is strictly larger than that number, solutions blow up in finite time. When $f\equiv 0$, it was shown in [8] that the critical mass equals $4\pi$ when $\Omega=B(0,1)$ and $8\pi$ when the initial data are non-negative and radial. However, in higher dimensions, this property no longer holds. Recent research, as reported in [13], has shown that a finite blow-up solution can be constructed in a smooth bounded domain, regardless of how small the mass is.
The logistic sources, $f(u):=ru-\mu u^{2}$, was introduced and studied in [11] that if $\mu>\frac{n-2}{n}$ then solutions exist globally and are bounded at all time in a convex open bounded domain $\Omega\subset\mathbb{R}^{n}$ where $n\geq 2$. In order word, if $\mu$ is sufficiently large, then the quadratic term $-\mu u^{2}$ ensures no occurrence of blow-up solutions in two spacial dimensional domain. This leads to a natural question that whether the term ”$-\mu u^{2}$” is optimal to prevent blow-up solutions. However, it has been discovered in [14] that the answer is negative. To be specific, the ”weaker” term $-\frac{\mu u^{2}}{\ln^{p}(u+e)}$ for $0<p<1$ is sufficient to avoid blow-up solutions for both elliptic-parabolic and fully parabolic minimal Keller-Segel chemotaxis models in a two spacial dimensional domain. Our main work improve the previous finding by showing that $p=1$ can prevent blow-up solutions of the system (1.1).
Our analysis relies on a test function method and Moser iteration technique. It is proved in [4] that if the family of $\left\{\int_{\Omega}u^{\frac{n}{2}}(\cdot,t)\right\}_{t\in(0,T_{\rm max})}$ is equi-integrable, then solutions of (1.1) when $f\equiv 0$ exist globally and remain bounded at all time. In this paper, we give another shorter proof in Proposition 3.1 for that result as well as indicate that the equi-integrability is not optimal to prevent blow-up thank to de la Vallée-Poussin Theorem. Thereafter, we try to a find a suitable functional and establish a differential inequality to obtain a priori estimate for solutions of (1.1) thank to the presence of the sub-logistic quadratic degradation term ”$-\mu\frac{u^{2}}{\ln(u+e)}$”. Notice that the key milestone in this study is the choice of the following functional:

One can also try to examine a functional

to find an appropriate $k$, however, there is no suitable $k$ satisfying the conditions that $\mu$ can be arbitrary small. In order word, this method leads to the choice of $k$, but it does require the largeness assumption for $\mu$. So the functional (1.5) enables us to overcome that obstacle to prove our main theorem as follows:

The local existence and uniqueness of non-negative classical solutions to the system (1.1) can be established by adapting and adjusting the fixed point argument and standard parabolic regularity theory. For further details, we refer the reader to [5, 11, 6]. For convenience, we adopt Lemma 4.1 from [12].

We will use several interpolation inequalities extensively in the following sections. To start, we present an extended version of the Gagliardo-Nirenberg interpolation inequality, which was established in [7].

In [4], an interpolation inequality of Ehrling-type is utilized to show that the equi-integrability of the family $\left\{\int_{\Omega}u^{\frac{n}{2}}(\cdot,t)\right\}_{t\in(0,T_{\rm max})}$ implies the uniform boundedness of solutions. In this paper, we present an interpolation inequality that is similar to [4, Lemma 2.1], and which will be employed to obtain an $L^{q}$ estimate with $q\geq 2$ for the solutions of the system (1.1). To prove this inequality, we adapt the argument used in the proof of inequality (22) in [3], with some modifications. We include a complete proof of this interpolation inequality below for the reader’s convenience.

Let us recall de la Vallée-Poussin Theorem

In this section, $(u,v)$ is a classical solutions as defined in Lemma 2.1 to the system (1.1) with $p=1$. Our aim is to establish a priori estimate for the solutions. While the method in [9] and [14] relies on the $L^{1}$-estimate of $u$ and the absorption of $-\int_{\Omega}|\nabla u^{\frac{1}{2}}|^{2}$ to obtain a $L\ln L$ uniform bound, we take advantage of the term $-\mu\frac{u^{2}}{\ln(u+e)}$ to obtain a weaker $L\ln\ln L$ uniform bound.

Thank to Lemma 2.4, the equi-integrability of $\left\{\int_{\Omega}u^{\frac{n}{2}}(\cdot,t)<\infty\right\}_{t\in(0,T_{\rm max})}$ is equivalent to
$\sup_{t\in(0,T_{\rm max})}\int_{\Omega}G(u^{\frac{n}{2}}(\cdot,t))<\infty$ for some non-negative increasing convex function such that $\lim_{s\to\infty}\frac{G(s)}{s}=\infty$. However, the convexity condition is not necessary, which means that the equi-integrable condition can be relaxed. Indeed, following proposition gives us the $L^{q}$ bounds, where $q>\frac{n}{2}$ for solutions without the convexity assumption.

We are now ready to prove the main result.

The author is indebted to Professor Michael Winkler for his kindly assistance in providing insightful comments, suggestions and valuable references. Additionally, the author extends appreciation to Professor Zhengfang Zhou for thoroughly reviewing the manuscript, engaging in fruitful discussions, and bringing to attention certain errors throughout the course of this project.