Asymptotical Behavior of Global Solutions of the Navier–Stokes–Korteweg Equations with Respect to
Capillarity Number at Infinity

A classical model to describe the dynamics of an inhomogeneous incompressible fluid endowed with internal capillarity (in the diffuse interface setting) is the following general system of incompressible Navier–Stokes–Korteweg (NSK) equations:

where ${\rho(x,t)}\in\mathbb{R}^{+}$, $v(x,t)\in{\mathbb{R}^{3}}$ and $P(x,t)$ denote the density, velocity and kinetic pressure of the fluid resp. at the spacial position $x\in{\mathbb{R}^{3}}$ for time $t\in\mathbb{R}^{+}_{0}:=[0,{+\infty})$. The differential operator $\mathbb{D}$ is defined by $\mathbb{D}v=\nabla v+\nabla v^{\top}$, where the superscript $\top$ represents the transposition. The shear viscosity function $\mu$, the capillarity function $\tilde{\kappa}$ are known smooth functions $\mathbb{R}^{+}\to\mathbb{R}$, and satisfy $\mu>0$, $\tilde{\kappa}>0$. The general capillary tensor is written as

where $\mathbb{I}$ is the identity matrix. To conveniently investigate the asymptotical behavior of capillarity number, we considers that $\mu$ and $\tilde{\kappa}$ are positive constants [6], thus it holds that

where we have defined that $\kappa:=\sqrt{\tilde{\kappa}}>0$. We call the parameters $\mu$ and $\kappa^{2}$ the viscosity coefficient and the capillarity number, resp..

In classical hydrodynamics, the interface between two immiscible compressible/incompressible fluids is modeled as a free boundary which evolves in time. The equations describing the motion of each fluid are supplemented by boundary conditions at the free surface involving the physical properties of the interface. For instance, in the free-boundary formulation, it is assumed that the interface has an internal surface tension. However, when the interfacial thickness is comparable to the length scale of the phenomena being examined, the free-boundary description breaks down. Diffuse-interface models provide an alternative description where the surface tension is expressed in its simplest form as $\mathrm{div}K$, i.e., the capillary tension which was introduced by Korteweg in 1901 [41]. Later, its modern form was derived by Dunn and Serrin [19].

In the physical view, it can serve as a phase transition model to describe the motion of compressible fluid with capillarity effect. Owing to its importance in mathematics and physics, there has been profuse works for the mathematical theory of the corresponding compressible NSK system, for example, see [2, 7, 24] for the global(-in-time) weak solutions with large initial data, [17, 25, 27, 40, 50] for global strong solutions with small initial data, [12, 26, 42, 43] for local(-in-time) strong/classical solutions with large initial data, [16] for stationary solutions, [20] for highly rotating limit, [15, 28] for the stability of viscous shock wave, [54] for the maximal $L^{p}$–$L^{q}$ regularity theory, [9, 13, 39] for the decay-in-time of global solutions, [5, 14, 21, 38, 30] for the vanishing capillarity limit, [44, 45, 29] for the nonisentropic case, [48, 55, 37] for the low Mach number limit, [3, 4, 8, 57] for the inviscid case, and so on.

The incompressible NSK system (1.1) has been also widely investigated. The local existence of a unique strong solution was obtained by Tan–Wang [58]. Burtea–Charve also established the existence result of strong solutions with small initial perturbations in Lagrangian coordinates [11], and further presented that the lifespan goes to infinity as the capillarity number goes to zero. Yang–Yao–Zhu proved that as both capillarity number and viscosity coefficient vanish, local solutions of the Cauchy problem for the incompressible NSK system converge to the one of the inhomogeneous incompressible Euler equations [60]. Recently a similar result was also established by Wang–Zhang for the fluid domain being a horizontally periodic slab with Navier-slip boundary conditions [59].

In addition, it also has been investigated that the capillarity under the large capillarity number stabilizes the motion of the incompressible fluid endowed with internal capillarity. Bresch–Desjardins–Gisclon–Sart derived that the capillarity slows down the growth rate of linear Rayleigh–Taylor (RT) instability, which occurs when a denser fluid lies on top of a lighter fluid under gravity [6]. Li–Zhang proved that the capillarity can inhibit RT instability for properly large capillarity number by the linearized motion equations [46] (Interesting readers further refer to [51, 46, 62, 61] for the existence of RT instability solutions in the incompressible NSK system with small capillarity number). Later, motivated by the result of magnetic tension inhibiting the RT instability in the two-dimensional (2D) non-resistive magnetohydrodynamic (MHD) fluid in [33], Jiang–Li–Zhang mathematically proved that the capillarity inhibits RT instability in the 2D NSK equations in the Lagrangian coordinates [34]. Recently Jiang–Zhang–Zhang further verified such inhibition phenomenon for the 3D case in Eulerian coordinates [35]. Jiang–Zhang–Zhang’s result rigorously presents that the capillarity has the stabilizing effect as well as the magnetic tension in the MHD fluid.

It is well-known that the larger the magnetic tension, the stronger the field intensity in the MHD fluid. Moreover the local solutions of the system of the idea MHD fluid with well-prepared initial data tend to a solution of a 2D Euler flow coupled with a linear transport equation as the field intensity goes to infinity, see [36, 22, 10] for relevant results. This naturally motivates us to except a similar result in the incompressible NSK system. In this paper, under the well-prepared initial data, we also construct a family of the global strong solutions of the 3D incompressible NSK equations with respect to $\kappa$, where the solutions converge to a unique solution of a 2D incompressible Navier–Stokes-like equations as $\kappa$ goes to infinity, see Theorem 2.2. As Lin pointed out in [49], the stratified fluids with the internal surface tension, the viscoelastic fluids, the non-resistive MHD fluids and the diffuse-interface model (or NSK model) can also be regarded as complex fluids with elasticity. Our mathematical result in Theorem 2.2 supports that the capillarity has the role of elasticity again.

This section is devoted to introducing our main results in details. Let us consider a rest stat of the system (1.1). We choose the equilibrium density $\bar{\rho}$ in the rest state to satisfy

Without loss of generality, we consider

Denoting the perturbations around the rest state $(\bar{\rho},0)$ by

and then recalling the relation

we obtain the following perturbation system from (1.1) with (1.2):

where $\mathbf{e}^{3}$ represents the unit vector with the third component being $1$ and we have defined that $\beta:=P-\kappa^{2}\rho\Delta\rho$. The initial condition reads as follows

Here and in what follows, we always use the right superscript $0$ to emphasize the initial data.

We consider that the fluid domain denoted by $\Omega$ is a slab, i.e.

The 2D domain $\mathbb{R}^{2}\times\{0,h\}$, denoted by $\partial\Omega$, is the boundary of $\Omega$. We focus on the following Navier(-slip) boundary conditions on the velocity on $\partial\Omega$:

where $\mathbf{n}$ denotes the unit outward normal to $\Omega$ and the subscript “tan” means the tangential component of a vector (for example, $u_{\mathrm{tan}}=u-(u\cdot\mathbf{n})\mathbf{n}$) [18, 47]. The Navier slip boundary conditions describe an interaction between a viscous fluid and a solid wall. Since $\Omega$ is a slab domain, the Navier boundary conditions are equivalent to the boundary condition

We mention that it is not clear to us whether our main results in Theorems 2.1 and 2.2 can be extended to the non-slip boundary condition of velocity.

However, to investigate the asymptotical behavior of solutions with respect to capillarity number at infinity and to except the vanishing of $\varrho$, we set

and then transform (2.3) into

The corresponding initial condition reads as follows

In addition, to avoid the vacuum of density, we assume that

Before stating our main results, we shall introduce simplified notations, which will be used throughout this paper.

(1) Simplified basic notations: $I_{a}:=(0,a)$ denotes an interval, in particular, $I_{\infty}=\mathbb{R}^{+}=(0,\infty)$. $\overline{S}$ denotes the closure of a set $S\subset\mathbb{R}^{n}$ with $n\geqslant 1$, in particular, $\overline{I_{T}}=[0,T]$ and $\overline{I_{\infty}}=\mathbb{R}^{+}_{0}$. $a\lesssim b$ means that $a\leqslant cb$ for some constant $c>0$, where $c>0$ at most depends on the domain $\Omega$, $\bar{\rho}$, $\mu$, $d$ in (2.8), and may vary from line to line. Sometimes we also denote $c$ by $c_{i}$ for emphasizing that $c_{i}$ is fixed, where $i=1$, $2$. $\partial_{i}:=\partial_{x_{i}}$, where $1\leqslant i\leqslant 3$. Let $f:=(f_{1},f_{2},f_{3})^{\top}$ be a vector function defined in a 3D domain, we define that $f_{\mathrm{h}}:=(f_{1},f_{2})^{\top}$ and $\mathrm{curl}{f}:=(\partial_{2}f_{3}-\partial_{3}f_{2},\partial_{3}f_{1}-\partial_{1}f_{3},\partial_{1}f_{2}-\partial_{2}f_{1})^{\top}$. $\nabla^{\bot}=(-\partial_{2},\partial_{1},0)$, $\nabla_{\mathrm{h}}:=(\partial_{1},\partial_{2})^{\top}$, $\nabla^{\perp}_{\mathrm{h}}:=(-\partial_{2},\partial_{1})^{\top}$, $\mathrm{div}_{\mathrm{h}}\cdot:=\partial_{1}\cdot+\partial_{2}\cdot$ and $\Delta_{\mathrm{h}}:=\partial^{2}_{1}+\partial^{2}_{2}$. For the simplicity, we denote $\sqrt{\sum_{i=1}^{n}\|w_{i}\|_{X}^{2}}$ by $\|(w_{1},\cdots,w_{n})\|_{X}$, where $\|\cdot\|_{X}$ represents a norm, and $w_{i}$ is scalar function or vector function for $1\leqslant i\leqslant n$.

(2) Simplified Banach spaces, norms and semi-norms:

where $1\leqslant p\leqslant\infty$ is a real number, and $i$, $l\geqslant 0$, $j\geqslant 1$, $k\geqslant 2$ are integers.

(3) Simplified spaces of functions with values in a Banach space:

(4) Energy/dissipation functionals (generalized):

Now we state the first result, which presents that the initial-boundary value problem (2.5)–(2.7) with small initial data admits a unique global strong solution with uniform estimates with respect to $\kappa$.

The existence of strong solutions with small initial perturbations to the initial-boundary value problem (2.3)–(2.5) in a horizontally periodic domain $\Omega_{\mathrm{p}}:=\mathbb{T}^{2}\times(0,h)$ has been established by Jiang–Zhang–Zhang, where $\mathbb{T}:=\mathbb{R}/\mathbb{Z}$. Let us first roughly recall some key ideas in Jiang–Zhang–Zhang’s proof in [35].

The standing point of Jiang–Zhang–Zhang’s proof is the basic energy identity in differential version for the problem (2.3)–(2.5):

We call (2.14) the zero-order energy estimate of $(\varrho,u)$. Then they further derived energy estimates for the high-order spacial derivatives and the temporal derivatives of $(\varrho,u)$. However, the integrals related to the nonlinear terms in the system (2.3) appear in the high-order energy estimates. We call such integrals the nonlinear integrals for simplicity. In particular, there exist a troublesome nonlinear integral $\int(\partial_{3}^{3}\varrho)^{2}\partial_{3}u_{3}\mathrm{d}x$, which is equal to $-\int(\partial_{3}^{3}\varrho)^{2}\mathrm{div}_{\mathrm{h}}u_{\mathrm{h}}\mathrm{d}x$ due to the incompressible condition (2.3)₃. Jiang–Zhang–Zhang naturally expect that $\nabla_{\mathrm{h}}u$ enjoy a fine decay-in-time, which contributes to close the troublesome nonlinear integral. However one can not directly derive the decay-in-time of $\nabla_{\mathrm{h}}u$ from high-order energy estimates due to the absence of the dissipation of $\nabla_{\mathrm{h}}\varrho$. Fortunately, they can capture the dissipation of $\nabla_{\mathrm{h}}\varrho$ from the vortex equations, which can be obtained by applying $\mathrm{curl}$ to (2.3)₂. The dissipation of $(\nabla_{\mathrm{h}}\varrho,\nabla_{\mathrm{h}}u)$, together with the horizontal periodicity of $\Omega_{\mathrm{p}}$, results in the decay-in-time of $\nabla_{\mathrm{h}}\varrho$ and $\nabla_{\mathrm{h}}u$ by energy method with extremely fine estimates.

Following Jiang–Zhang–Zhang’s ideas in [35] for the initial-boundary vaule problem (2.5)–(2.7), however it seems to be difficult to establish the desired decay-in-time of $\nabla_{\mathrm{h}}u$ due to the unboundedness of the fluid domain $\Omega$. This results in that we shall develop a new alternative method to estimate for the troublesome nonlinear integral $\int(\partial_{3}^{3}\sigma)^{2}\partial_{3}u_{3}\mathrm{d}x$ (recalling the linear relation $\sigma=\kappa\varrho$). More precisely, we use the transport equation (2.6)₁ twice and the anisotropic Gagliardo–Nirenberg–Sobolev type estimate in A.6 to estimate $\int(\partial_{3}^{3}\sigma)^{2}\partial_{3}u_{3}\mathrm{d}x$, see (3.20) for the detailed derivation. Based on this new idea, we can refine the energy method in [35] to establish Theorem 2.1.

It is easy see from (2.9) that $\chi$ for $\kappa\geqslant 1$ reduces to $\chi=c_{2}$. Consequently, we can make use of the uniform-in-$\kappa$ estimates in (2.11)–(2.13) with $\kappa\geqslant 1$ to establish the asymptotical behavior with respect to capillarity number at infinity. More precisely, for any given $\kappa>0$, we choose initial data $\sigma^{0}_{\kappa}$ and $u^{0}_{\kappa}$, which satisfy (2.8) and (2.10) with $(\sigma^{0}_{\kappa},u^{0}_{\kappa})$ in place of $(\sigma^{0},u^{0})$. In view of Theorem 2.1, the initial-boundary value problem (2.3)–(2.5) with $(\varrho^{0}_{\kappa},u^{0}_{\kappa})$ in place of $(\varrho^{0},u^{0})$ admits a global strong solution denoted by $(\varrho^{\kappa},u^{\kappa})$, where we have defined that $\varrho^{0}_{\kappa}:=\kappa^{-1}\sigma^{0}_{\kappa}$. Moreover the solutions have the following asymptotical behavior with respect to $\kappa$ at infinity

We will prove the asymptotical behavior in Theorem 2.2 by exploiting a compactness argument, the details of which will be presented in Section 5.

This section is devoted to the proof of the global(-in-time) solvability with uniform-in-$\kappa$ estimates for the initial-boundary value problem (2.5)–(2.7). The key point is to a priori derive the uniform-in-$\kappa$ estimate in (2.11). For this purpose, let $T>0$ be a fixed time and $(\sigma,u)$ a solution to (2.5)–(2.7) on $\Omega\times I_{T}$ with initial data $(\sigma^{0},u^{0})\in{{H}}^{3}_{\bar{\rho}}\times{\mathcal{H}^{2}_{\mathrm{s}}}$. Moreover, we assume that $\sigma^{0}$ satisfies (2.8) and the solution $(\sigma,u)$ is sufficiently regular so that the procedure of formal deduction makes sense.