Abstract
This is the third paper in a three-part sequence in which we prove that steady, incompressible Navier–Stokes flows posed over the moving boundary, \(y = 0\), can be decomposed into Euler and Prandtl flows in the inviscid limit globally in \([1, \infty ) \times [0,\infty )\), assuming a sufficiently small velocity mismatch. In this paper, we prove existence and uniqueness of solutions to the remainder equation.
Similar content being viewed by others
Avoid common mistakes on your manuscript.
1 Introduction
In this paper, we will prove existence and uniqueness in the space Z [the Z norm is recalled below in (1.45)] of solutions to the nonlinear system
with boundary conditions
The terms in Eqs. (1.1)–(1.2) are defined:
For the convenience of the reader, we recall below the main estimates established in [3] which will be in use throughout this paper.
Theorem 1.1
[3] Let\(n \ge 2 \in {\mathbb {N}}\). Let\(\delta , \epsilon \)be sufficiently small relative to universal constants, and\(\epsilon<< \delta \). Let the boundary and in-flow data be prescribed as described in [3]. Then there exist Prandtl profiles\([u^j_p, v^j_p, P^j_p]\)for\(j = 1,\ldots ,n\), Euler profiles\([u^j_e, v^j_e, P^j_e]\)for\(j = 1,\ldots ,n\), and auxiliary pressures\([P^{j,a}_p, P^{j,a}_e]\)for\(j = 1,\ldots ,n\)such that for any\(\gamma \in [0,\frac{1}{4}), n \ge 2\), and for\(\sigma _n = \frac{1}{10000}, \kappa >0\)arbitrarily small, the following remainder estimate holds for any\(k \ge 0\):
The following bounds hold on\([u_R,v_R]\)by construction, for any\([k, j, m] \ge 0\), so long asnis sufficiently large relative tom.
We also recall briefly the estimate, established in [3], on the remainders, \(R^{u,n}, R^{v,n}\) which act as forcing terms:
Lemma 1.2
(Remainder Estimates) For any\(\gamma \in [0,\frac{1}{4}), n \ge 2\), and for\(\sigma _n = \frac{1}{10000}, \kappa >0\)arbitrarily small,
The main result of this paper is:
Theorem 1.3
For\(\epsilon , \delta \)sufficiently small,\(\epsilon<< \delta , \kappa > 0\)small, and\(0 \le \gamma < \frac{1}{4}\), there exists a unique solution\([u,v] \in Z(\Omega )\)to the system (1.1)–(1.3), (1.4), (1.5) satisfying the bound:
The main result of the three-paper sequence, Theorem 1.2 of [3], follows immediately from Theorem 1.3. The proof of this theorem proceeds in several steps, which we now outline:
- (Step 1)
Linear existence of solutions to weighted Stokes system, defined as follows:
$$\begin{aligned}&\varDelta _\epsilon ^2 \psi + \alpha A(\psi ) = F_y - \epsilon G_x \quad {\text { on }} \Omega ^N, \quad F, G \in L^2(\Omega ^N),\end{aligned}$$(1.26)$$\begin{aligned}&\psi |_{y = 0, N} = \psi _y|_{y = 0, N} = 0, \quad {\text { and }} \psi |_{x = 1} = \psi _x|_{x = 1} = 0, \end{aligned}$$(1.27)$$\begin{aligned}&\lim _{x \rightarrow \infty } [\psi _{x}, \psi _y] = 0, \end{aligned}$$(1.28)where \(\alpha > 0\), and
$$\begin{aligned} A(\psi )&= \big [ \psi x^{2m} - \psi _{yy} x^{2m+2} - \partial _x(\psi _x x^{2m+2}) + \psi _{yyyy} x^{2m+4} \nonumber \\&\quad \, + \partial _x ( \psi _{yyx} x^{2m+4} ) + \partial _{xx} ( \psi _{xx} x^{2m+4} ) \big ].\end{aligned}$$(1.29)Here, \(m > 0\) is sufficiently large, and can remain temporarily unspecified. The scaled Bilaplacian is defined as \(\varDelta _\epsilon ^2 := \partial _y^4 + \epsilon \partial _y^2 \partial _x^2 + \epsilon ^2 \partial _x^4\). The right-hand sides, F, G, should be thought of as generic elements satisfying \(F_y - \epsilon G_x \in H^{-1}\). Upon introducing appropriate function spaces, we define the weak formulation of (1.26)–(1.27) in (2.6). Depicting the weak-solution operator to the above system by \(S_{\alpha} ^{-1}\) (see (2.12) for a precise definition), Step 1 amounts to studying the solvability of \(S_{\alpha} \psi = F_y - \epsilon G_x\).
The boundary conditions as \(x \rightarrow \infty \) in (1.28) are selected in order to be consistent with (1.4). However, due to the terms in \(A(\psi )\), the weak solution, \([\psi , u, v]\) exhibits rapid decay as \(x \rightarrow \infty \).
- (Step 2)
Linear existence of compact perturbations to \(S_{\alpha} \). Define the maps:
$$\begin{aligned} T[\psi ]&:= \partial _y \big [ -u_R \psi _{xy} - u_{Rx}\psi _y - (v_R + \epsilon ^{\frac{n}{2}+\gamma } {{\bar{v}}}) \psi _{yy} + u_{Ry} \psi _x \big ] \nonumber \\&\qquad - \epsilon \partial _x \Big [ u_R \psi _{xx} - v_{Ry} \psi _y + v_R \psi _{xy} + v_{Ry} \psi _x \big ], \end{aligned}$$(1.30)$$\begin{aligned} T_0[\psi ]&:= T[\psi ] + \epsilon ^{\frac{n}{2}+\gamma } \partial _y [ {{\bar{v}}} \psi _{yy}], \end{aligned}$$(1.31)$$\begin{aligned} T_a[\psi ]&:= -u_R \psi _{xy} - u_{Rx}\psi _y - v_R \psi _{yy} + u_{Ry} \psi _x, \end{aligned}$$(1.32)$$\begin{aligned} T_b[\psi ]&:= u_R \psi _{xx} - v_{Ry} \psi _y + v_R \psi _{xy} + v_{Ry} \psi _x. \end{aligned}$$(1.33)T has a dependence on \({{\bar{v}}}\), so to be precise we will sometimes write \(T[\psi ; {{\bar{v}}}]\). When there is no danger of confusion, we simply write \(T[\psi ]\). The map \(T_0[\psi ]\) is defined to match the profile terms, \(S_u(u,v), S_v(u,v)\) [see the definition in (1.6)], when they are written in terms of the stream function, \(\psi \). We have defined the notation \(T_a, T_b\) so that we can write \(T_0 = \partial _y T_a - \epsilon \partial _x T_b\). In this step, we are interested in establishing solvability of the system:
$$\begin{aligned}&S_{\alpha} \psi + T[\psi ] = F_y - \epsilon G_x\quad {\text { on }} \Omega ^N, \end{aligned}$$(1.34)$$\begin{aligned}&[\psi = \psi _{x}]|_{x = 1} = [\psi = \psi _{y}]|_{y = 0} = [\psi = \psi _{y}]|_{y = N} = \lim _{x \rightarrow \infty } [\psi _x, \psi _y] = 0. \end{aligned}$$(1.35)The essence of the arguments in this step is that upon applying \(S_{\alpha} ^{-1}\) to both sides above, \(S^{-1}_{\alpha} T\) is seen as a compact perturbation of the identity. Despite \(\Omega ^N\) being unbounded in the x-direction, the required compactness arises from the weights, w, present in \(A(\psi )\) above in (1.29). The solution of (1.34) is known to decay rapidly as \(x \rightarrow \infty \), due to the presence of \(A(\psi )\). This is captured in estimate (3.67).
- (Step 3)
Nonlinear existence of auxiliary system: we first invite the reader to refer back to (1.5) for the definitions of f and g. Given this and the definition of T in (1.30), we define:
$$\begin{aligned} {{\tilde{f}}}({{\bar{u}}}, {{\bar{v}}}):= \epsilon ^{-\frac{n}{2}-\gamma } R^{u,n} + \epsilon ^{\frac{n}{2}+\gamma } {{\bar{u}}} {{\bar{u}}}_x, \quad {\text {so that}}\,\,\, \,\, f(u, {{\bar{u}}}, {{\bar{v}}}) = {{\tilde{f}}}({{\bar{u}}}, {{\bar{v}}}) + \epsilon ^{\frac{n}{2}+\gamma } {{\bar{v}}}u_y. \end{aligned}$$(1.36)The aim of this step is to obtain existence of solutions (which we now index by \(\alpha \) and N for clarity) to the nonlinear system:
$$\begin{aligned} S_{\alpha} \psi ^{\alpha , N} + T[\psi ^{\alpha , N}; v^{\alpha , N}] = {{\tilde{f}}}_y(u^{\alpha , N}, v^{\alpha , N}) + \epsilon g_x(u^{\alpha , N},v^{\alpha , N}) \quad {\text { on }} \Omega ^N. \end{aligned}$$(1.37)This existence is obtained in the unit ball of \(Z(\Omega ^N)\) via Schaefer’s fixed point theorem.
- (Step 4)
Nonlinear existence of solutions to the system (1.1)–(1.3), with f, g as in (1.5): By re-applying the analyses in [4], one obtains the uniform-in-\((\alpha ,N)\) estimate: \(\Vert u^{\alpha , N}, v^{\alpha , N} \Vert _{Z(\Omega ^N)} \lesssim {\mathcal {O}}(\delta ) \epsilon ^{\frac{1}{4}-\gamma - \kappa }\), which then enables the passage to weak limits in the space \(X_1 \cap X_2 \cap X_3\). The weak limit is denoted by [u, v], and is demonstrated to satisfy a weak formulation of system (1.1)–(1.3), see (5.10) for this formulation. Moreover, \([u,v] \in X_1 \cap X_2 \cap X_3\), gives enough regularity to upgrade immediately to a strong solution of (1.1)–(1.3).
Remark 1.4
To establish existence, we rely on compactness methods as opposed to applying a contraction mapping. The essential reason for this is seen by examining calculation (4.5) in [4], in which the structure is not preserved under taking differences.
Remark 1.5
It is important to establish nonlinear existence of the auxiliary system before establishing nonlinear existence of the system (1.1)–(1.3), as opposed to jumping from linear existence of (1.1)–(1.3) to nonlinear existence because the compactness methods we rely on require the weights from \(\alpha A(\psi )\).
-
(Step 5)
Nonlinear uniqueness for solutions to the system (1.1)–(1.3), with f, g as in (1.5): In order to prove uniqueness, we re-apply the estimates from [4] with weights that are weaker by \(x^{-b}\), where \(b < 1\), but is arbitrarily close to 1. This step is necessary (with the weaker weight) due again to the calculation in (4.5) from [4], whose structure is destroyed upon considering differences.
1.1 Notations and Norms
We briefly recall the norms introduced in [4] for the convenience of the reader. First define the cut-off functions:
The energy norms are defined as follows:
Definition 1.6
The norms \(Y_2, Y_3\) are strengthenings of \(X_2, X_3\) near the boundary, \(x = 1\), and defined through:
Definition 1.7
The norm Z is defined through:
Here, \(N_i\), are some large universal numbers.
The following embedding result from [4] appears repeatedly in the present paper:
Lemma 1.8
For\(\sigma > 0\)arbitrarily small,
The constant\(C(\sigma ) \uparrow \infty \)as\(\sigma \downarrow 0\). Finally, for\([u,v] \in Z\), we have the following property:
2 Invertibility of Weighted Stokes Operator, \(S_{\alpha} \)
In this step, we study the system (1.26)–(1.27). We remind the reader that still, all integrations and all norms are taken over \(\Omega ^N\) unless otherwise specified. There is an abuse of notation here; \(\psi \) should be indexed by \(\alpha \) and N, but this will not cause any confusion for this step, as we view both \(\alpha \) and N as fixed. Our intention of this section is to exhibit solvability of the system (1.26) in the space \(Z(\Omega ^N)\). Denote by \(\chi _1(x)\) a cut-off function satisfying (refer to (1.38) for the definition of \(\zeta _3\)):
We define higher-order cut-offs similar to (2.1), satisfying the following property: \({\text {support}} \, (\chi _k) \subset \{ \chi _{k-1} = 1 \}\). Define the following auxiliary norms via:
We will also call \(G^k_{w, {\mathrm{loc}}}(\Omega ^N)\) the space such that \(\Vert \psi \Vert _{G^k_{w,B}} \le C(B)\) for all compact subsets B. Define the weak formulation of (1.26) to be:
Above, \(\nabla ^2\) is the Hessian matrix, and the inner product between two matrices is given by \(A:B = {\text {trace}}(AB)\). We will need one more norm:
Relevant spaces are defined here:
Definition 2.1
\(H^2_w(\Omega ^N)\) is defined to be the closure of \(C_0^{\infty }(\Omega ^N)\) under the norm \(\Vert \cdot \Vert _{H^2_w}\). \(H^k_w(\Omega ^N)\) for \(k \ge 3\) consists of the subspace of \(H^2_w(\Omega ^N)\) whose \(H^k_w(\Omega ^N)\) norm is finite. Note that \(H^3_w(\Omega ^N)\) does not contain all of the third derivatives of \(\psi \); it is missing \(\partial _y^3 \psi \), which is the reason for the norm, \(\Vert \cdot \Vert _{G_{w, B}}\).
Remark 2.2
There is a distinction between \(H^2_w(\Omega ^N)\), and \(H^k_w(\Omega ^N)\) in that:
Due to the weights, there is no “\(H = W\)” theorem generically for \(H^k_w(\Omega ^N)\).
Lemma 2.3
For \(\psi \in H^2_w(\Omega ^N)\) , the following boundary conditions are satisfied:
Proof
If \(\psi \in H^2_w(\Omega ^N)\), obtain a sequence \(\phi ^{(n)}\) such that \(\Vert \phi ^{(n)} - \psi \Vert _{H^2_w} \rightarrow 0\). The claim now follows by the standard boundedness properties of the trace operator. \(\square \)
Lemma 2.4
\(H^2_w(\Omega ^N)\)as defined in Definition 2.1is a Banach space.
Proof
Consider the auxiliary space:
Through standard arguments, \(H^2_{0,w}(\Omega ^N)\) is a Banach space. Suppose \(\{\psi ^{(n)}\}\) is a Cauchy sequence in \(H^2_w(\Omega ^N)\). Then \(\{\psi ^{(n)}\}\) is Cauchy in \(H^2_{0,w}(\Omega ^N)\), and so there exists a limit point \(\psi \) such that: \(\Vert \psi - \psi ^{(n)}\Vert _{H^2_w} \xrightarrow {n \rightarrow \infty } 0\). As \(\psi ^{(n)} \in H^2_w(\Omega ^N)\), we may find a sequence \(\{\phi ^{(n)}_m\}_{m \ge 1}\) such that \(\Vert \phi ^{(n)}_m - \psi ^{(n)}\Vert _{H^2_w} \xrightarrow {m \rightarrow \infty } 0\), where \(\phi ^{(n)}_m \in C^\infty _0(\Omega ^N)\). In particular, define, for each n, by selecting m large enough: \(\Vert \phi ^{(n)} - \psi ^{(n)}\Vert _{H^2_w} < 2^{-n}\). Thus, \(\Vert \phi ^{(n)} - \psi \Vert _{H^2_w} \xrightarrow {n \rightarrow \infty } 0\), proving that \(\psi \in \overline{C^\infty _0}^{\Vert \cdot \Vert _{H^2_w}}\). This establishes the desired result. \(\square \)
Lemma 2.5
Endowed with the inner product,
\(H^2_w\)is a Hilbert Space. The inner product in (2.11) induces the norm defined in (2.2).
Proof
One easily verifies the standard axioms of an inner-product for (2.11). Non-degeneracy of (2.11) is obtained via the boundary conditions in (1.27). Completeness is then obtained via Lemma 2.4. \(\square \)
Definition 2.6
The \(\alpha \)-Stokes operator is defined through:
and is equivalent to (2.6) holding.
It is our aim to study the invertibility of \(S_{\alpha} \).
Lemma 2.7
Given\(F_y - \epsilon G_x \in H^{-1}(\Omega ^N)\), there exists a unique weak solution\(\psi \in H^2_w(\Omega ^N)\)satisfying (2.6). Such a weak solution satisfies the energy inequality:
Proof
Define:
It is immediate to see that B is bilinear, bounded, and coercive on \(H^2_w(\Omega ^N)\). Next, \(F_y - \epsilon G_x\) act as bounded linear functionals on \(H^2_w(\Omega ^N)\) through the pairing: \(\langle F_y - \epsilon G_x, \phi \rangle _{H^{-2}_w, H^2_w} := \langle F_y - \epsilon G_x, \phi \rangle _{H^{-1}, H^1}\). This follows from: \(| \langle F_y - \epsilon G_x, \phi \rangle _{H^{-1}, H^1} | \le \Vert F_y - \epsilon G_x\Vert _{H^{-1}} \Vert \phi \Vert _{H^2_w}.\) The existence of \(\psi \in H^2_w(\Omega ^N),\) a solution to (2.6) is then a standard application of the Lax–Milgram lemma to the Hilbert space \(H^2_w(\Omega ^N)\). The energy identity above follows from density of \(C_0^{\infty }(\Omega ^N)\) in \(H^2_w(\Omega ^N)\), which enables us to replace \(\phi \) with \(\psi \) in (2.6). \(\square \)
The above lemma then says that \(S_{\alpha} ^{-1}: H^{-1}(\Omega ^N) \rightarrow H^2_w(\Omega ^N)\) is well-defined. Our intention now is to upgrade regularity.
Lemma 2.8
Given\(F_y - \epsilon G_x \in H^{-1}(\Omega )\), the unique weak solution in\(H^2_w(\Omega ^N)\)guaranteed by Lemma 2.7is in \(H^3_w(\Omega ^N)\)and satisfies:
Proof
As our weak solutions are only in \(H^2_w(\Omega ^N)\), we must formally use difference quotients within the weak formulation (2.6) to upgrade to \(H^3_w(\Omega ^N)\). However, we will generate the \(H^3_w\) estimate via differentiating (1.26), with the understanding that everything that is done can be formalized through the use of difference quotients in the standard manner. As such, we take \(\partial _x\) of the system (1.26), which gives:
where
Let \(\chi _1\) be as above in (2.1). Define the quantities:
We now test the above equation, (2.16), against the multiplier \(\rho _M \psi _x\). Doing so first gives from the Bilaplacian:
for constants \(c_0, c_1\). Next, we have the terms coming from A:
Through a direct integration by parts, the commutator contains lower order terms:
For detailed proofs of calculations (2.20) and (2.21), we refer the reader to (3.93)–(3.100). Finally, on the right-hand side of (2.16), we have:
We can send \(M \rightarrow \infty \) so that the weight \(\rho _M \uparrow \chi _1\), resulting in
For the region \(1 \le x \le 20\), and \(0 \le y \le N\), we apply the standard \({{\dot{H}}}^2(\Omega ^N)\) estimate for solutions, \(u^\alpha , v^\alpha \) Stokes’ equation near corners (see [1], Theorems 1 and 2, and Figure 2, P. 562 also in [1] with “C/C” boundary conditions). Formally, fix another cut-off function, \(\chi _2(x,y)\) localized near the corner (1, 0) [the identical argument can be given for the other corner, (1, N)]. First, by calculation, we have:
where the expression for the commutator is given explicitly:
The salient feature of (2.25) will be:
where this is short-hand notation for containing up to three \(\psi \)-derivatives, and localized by \(\chi _2\) (or any derivative of \(\chi _2\) which is also localized). Localizing (1.26) using \(\chi _2\):
Combining (2.27) and (2.23) gives the desired result. \(\square \)
Lemma 2.9
Fix any bounded set\(B \subset \Omega ^N\). Then we have:
where the constantC(B) depends onB.
Proof
This argument proceeds identically to the calculation from the previous lemma which resulted in (2.27) by simply replacing \(\chi _2\) with cut-off functions localized to each interval \(x \in [M, M+1]\). The dependence on B in the constant in (2.28) arises from the weights \(x^{2m},x^{2m+2}, x^{2m+4}\) appearing in Eq. (1.26) through \(A(\psi )\). \(\square \)
The above lemmas roughly show that \(S_{\alpha} ^{-1}\) gains four derivatives. By repeating this procedure for higher-order x-derivatives, we can upgrade to higher-regularity.
Lemma 2.10
Given\((F, G) \in H^{-1}_2\), the unique weak solution guaranteed by Lemma 2.7satisfies:
For \((F, G) \in H^{-1}_2\), we can upgrade weak solutions to strong solutions.
Lemma 2.11
Given\((F,G) \in H^{-1}_2\), the unique weak solution guaranteed by Lemma 2.7is a strong solution of (1.26).
Proof
An integration by parts of the weak formulation (2.6), justified according to the previous lemma, is equivalent to Eq. (1.26) being satisfied pointwise on \(\Omega ^N\). The boundary conditions at \(x = 1, y = 0, y = N\) are satisfied by Lemma 2.3. The boundary condition at \(x \rightarrow \infty \) comes from the norms, (2.2), which when applying with \(k = 5\), imply that up to four derivatives of \(\psi \) vanish rapidly at \(x \rightarrow \infty \). \(\square \)
3 Compact Perturbations, SαΨ+T[Ψ]
For this step, we invite the reader to refer back to the specification of \(T[\psi ]\), given in (1.30), and the system that we will focus on, given in (1.34). Note that \(T[\psi ]\) contains a loss of three derivatives for \(\psi \). Note also the presence of the term \(\epsilon ^{\frac{n}{2}+\gamma } {{\bar{v}}}\). We will now need some compactness lemmas.
Lemma 3.1
Fix two weights, \(w_1 = x^{m_1}\) , and \(w_2 = x^{m_2}\) , where \(m_2 > m_1 \ge 0\) . Then, one has the following compact embedding:
Proof
Consider a family of functions \(\{f^n\}\) defined on \(\Omega ^N\) such that:
and such that \(f_n \in H^1_{{\mathrm {loc}}}(\Omega ^N)\), uniformly in n. By taking Sobolev extensions across \(\partial \Omega ^N\), and subsequently cutting off in the y and negative x directions, we can assume \(\{f_n \}\) are defined on \({\mathbb {R}}^2\), compactly supported in the y direction and negative x direction. Fix any \(\delta ' > 0\). Since \(m_2 > m_1\), there exists a compact set \(K = K(\delta ')\) such that:
On K, by Rellich compactness, there exists a subsequence (depending on \(\delta '\)) such that
Then,
Combining the above two estimates,
Taking successively \(\delta ' = 2^{-n}\) and applying a diagonalization argument give the result. \(\square \)
Lemma 3.2
Let the weight,\(x^{2m}\), in the expression for \(A(\psi )\), Eq. (1.29), be selected for any\(m > 0\). Then the map\(S_{\alpha} ^{-1}T\)is well-defined and compact\(H^2(\Omega ^N) \rightarrow H^2(\Omega ^N)\).
Proof
According to (2.28), this follows from the compactness of \(G^3_{w, {\mathrm{loc}}}(\Omega ^N) \hookrightarrow \hookrightarrow H^2(\Omega ^N)\), which in turn follows from (3.1). The lemma is proven. \(\square \)
We are now ready to study system (1.34). The first task is to obtain an energy estimate to the inhomogeneous problem.
Lemma 3.3
Suppose\(\psi \in H^2(\Omega ^N)\)is a solution to (1.34), where\((F,G) \in H^{-1}_2\), and\(\Vert {{\bar{u}}}, {{\bar{v}}}\Vert _Z \le 1\). Then\(\psi \)obeys the following energy estimate:
Proof
Supposing there exists such a \(\psi \), we would have \(T[\psi ] \in H^{-1}(\Omega ^N)\), and so by (2.15), we know \(\psi \in H^3_w(\Omega ^N)\). By bootstrapping this regularity, we obtain that:
We would like to apply the multiplier \(\psi \) to Eq. (1.34) in order to repeat the energy estimate from Proposition 3.2 in [4]. Select test functions, \(\phi ^{(n)} \in C^\infty _0(\Omega ^N)\), which satisfy:
This is possible according to the density of \(C^\infty _0(\Omega ^N)\) in \(H^2_w\) in Definition 2.1. Multiplying (1.34) by \(\phi ^{(n)}\), then gives on the left-hand side:
First, we shall use (1.31) to write:
According to (3.9), we pass to limits in the following terms:
We have used:
according to (3.8) and the definition in Eq. (1.31). The integration on the right-hand side of (3.12) produces the lower bound:
We may pass to the limit in the final term of (3.11) due to the calculation:
Upon passing to the limit, we integrate by parts:
From here, we estimate:
It remains to treat (3.10), for which we use the compact support of \(\phi ^{(n)}\) to justify the integration by parts:
Passing to the limit, according to (3.9):
On the right-hand side, we have:
Consolidating the previous estimates gives the desired estimate, (3.7). \(\square \)
The task now is to estimate the right-hand side of (3.7) in terms of the left-hand side using the smallness of \({\mathcal {O}}(\delta )\). We refer the reader to Proposition 3.4 in [4], whose proof we follow closely. We will point out the subtle differences.
Lemma 3.4
Suppose\(\psi \in H^2(\Omega ^N)\)is a solution to (1.34), where\((F,G) \in H^{-1}_2\), and\(\Vert {{\bar{u}}}, {{\bar{v}}}\Vert _Z \le 1\). Suppose the weight\(w = x^{2m}\)from Eq. (1.29) is selected such thatmis sufficiently large relative to universal constants. Then\(\psi \)obeys:
Proof
We apply the multiplier \(\psi _x x \chi _{L,\alpha }^2\) to (1.34). Here, \(\chi \) is a normalized cut-off function equal to 1 on [1, 2] and 0 on \([3,\infty )\), and
The necessity of such a cut-off function here is due to the terms arising from \(A(\psi )\). The presence of this cut-off function enables us to justify all integrations by parts in the x-direction. For our fixed \(\alpha > 0\), we will eventually send \(L \rightarrow \infty \). Applying the multiplier \(\psi _x x \chi _{L,\alpha }^2\) to (1.34), gives on the left-hand side:
We will first focus on the first two integrands above. Let us start with the profile terms from \(T_0[\psi ]\). We will transfer all of the terms to velocity formulation.
We will treat the three terms on the right-hand side above, using (1.14) and (1.20) starting with:
Above, we have used the Hardy inequality:
Next, by (1.12), (1.22), we have:
Next, by (1.16), (1.18), (1.20) we have:
Summarizing the previous four terms:
We will now move to the profile terms from \(S_v\). First,
Referring to definition (1.5), consider the term \(u_R v_x\) in \(S_v\), which is the most delicate profile term:
The first term above in (3.32) gives positivity:
We will treat the second term on the right-hand side of (3.32):
The first term on the right-hand side of (3.34) yields:
We now estimate:
This same estimate can be performed for all the terms in (3.35). Consolidating these bounds:
It remains now to treat the remaining three terms in \(S_v\). The second, third, and fourth terms from \(S_v\) are estimated immediately:
We now turn back to (3.31), addressing the second term in the bracket for the final three profile terms from \(S_v\):
First, through the Hardy inequality and (1.10), (1.19):
so long as \(w = x^m\) is selected larger than \(x^{\frac{1}{4}}\), which is true by the assumption of this lemma. Next, through an integration by parts and (1.12), (1.22):
The final task for the \(S_v\) profile contributions is the third term from (3.31):
so long as \(w = x^m\) is selected larger than x, which is true by assumption of this lemma. Let us consolidate all of the calculations from \(S_v\):
It now remains to come to those terms contributed by \(\varDelta _\epsilon ^2 \psi \) into (3.24). Again, we will write these terms in the velocity form. First,
Next,
We now move to the terms from \(\varDelta _\epsilon v\), starting with:
Finally, we have:
By combining calculations (3.30), (3.46), (3.47)–(3.50), and absorbing relevant terms to the left-hand side below, we have:
Via direct integration by parts, which is justified due to the presence of the cut-off function in x, we compute:
Let us compute each term in \(A(\psi )\) to verify (3.52), referring to the definition in (1.29), starting with:
For the second term in (3.53), we have used: \(|\frac{\alpha }{L} x \chi _{L,\alpha }| \lesssim 1\). Next, let us turn to:
Next,
Next,
The final term in \(A(\psi )\) is:
This concludes all the terms in \(A(\psi )\), according to (1.29). Estimating the next term in (3.51) yields:
Finally, we come to the right-hand side:
Inserting the previous few calculations into estimate (3.51) gives:
We now send \(L \rightarrow \infty \), and appeal to Monotone Convergence Theorem, as \(\chi _{L,\alpha } \uparrow 1\) to establish the desired result. \(\square \)
Having understood the inhomogeneous problem:
Lemma 3.5
For\((F,G) \in H^{-1}_2\), and\(\Vert {{\bar{u}}}, {{\bar{v}}}\Vert _Z \le 1\), there exists a unique weak solution\(\psi \in H^2_w(\Omega ^N)\)to the system (1.34).
Proof
We apply \(S_{\alpha} ^{-1}\) to both sides of (1.34), which is valid as the right-hand side and therefore the left-hand side is assumed to be in at least \(H^{-1}(\Omega ^N)\), thereby yielding:
We will study Eq. (3.62) as an equality in the space \(H^2(\Omega ^N)\). According to the Fredholm alternative, which is available according to Lemma 3.2, there either exists a unique solution \(\psi \in H^2(\Omega ^N)\) to the system (3.62), or a non-trivial solution \(\psi \in H^2(\Omega ^N)\) to:
Therefore, coupling (3.22) with (3.7), taking \(F = G = 0\), we have:
implying \(\psi , u, v = 0\). Thus, by the Fredholm alternative, there exists a unique solution \(\psi \in H^2(\Omega ^N)\) to (3.62). Rearranging (3.62):
where \(F_y - \epsilon G_x - T\psi \in H^{-1}(\Omega ^N)\), and so an application of (2.13) shows that \(\psi \in H^2_w(\Omega ^N)\). This concludes the proof. \(\square \)
Lemma 3.6
Let\(\psi \)be the unique\(H^2_w(\Omega ^N)\)weak solution from Lemma 3.5. Then for\((F,G) \in H^{-1}_2\), \(\psi \in H^5_w(\Omega ^N)\).
Proof
\(T \psi \in H^{-1}(\Omega ^N)\), so \(S_{\alpha} \psi = - T\psi + F_y - \epsilon G_x \in H^{-1}(\Omega ^N)\), which implies that \(\psi \in H^3_w(\Omega ^N)\) according to (2.15). Iterating this regularity then gives \(\psi \in H^5_w(\Omega ^N)\). \(\square \)
We now introduce more notations, which are more suitable for the velocities:
Summarizing the established results, we have:
Corollary 3.7
For\({{\tilde{f}}}, g \in H^{-1}_2(\Omega ^N), \alpha > 0\), and\(\Vert {{\bar{u}}}, {{\bar{v}}}\Vert _{Z(\Omega ^N)} \le 1\), the map \(L_{\alpha , {{\bar{v}}}}[u,v]\)is invertible, where
Moreover, the boundary conditions (1.27) are satisfied by\([u,v] = L_{\alpha , {{\bar{v}}}}^{-1}[{{\tilde{f}}},g]\).
It is now our intention to obtain second- and third-order energy and positivity estimates for our new system (1.34). For this, we will need to understand several calculations. First, we introduce some norms:
The essential difference between these \(J^k\)-norms and the \(H^k_w\) norms introduced in (2.2) are the growing weights of x which each application of \(\partial _x\), which mimics the structure of the energy norms, \(X_k\).
Lemma 3.8
Proof
Referring to (1.29), the first term is:
The next terms, via an integration by parts in y:
Next,
Above, we have used the calculation:
For the second term on the right-hand side of (3.75), we estimate: \(\frac{2\alpha }{L} x \chi _{L,\alpha } \lesssim 1\). For the third term on the right-hand side, we use that the support of \(\rho _{k+1}'\) is localized in x. We also use that: \({\text {support}} (\rho _k ) \subset \{\rho _{k-1} = 1\}\). Next, we integrate by parts twice in x to obtain:
The final two terms on the right-hand side of (3.77) are estimated through further integrations by parts:
Finally,
Piecing all of the above estimates together yields the desired bound. \(\square \)
Lemma 3.9
Proof
Again, referring to definition (1.29), we will proceed term by term, starting with the following, for which we integrate by parts once:
Let us expand the product rule above:
Thus, the term (3.80) can be controlled via:
The second term in (1.29) is treated via:
We have expanded the product in the first term on the right-hand side of (3.84):
Next, we have:
We will expand the product rule above:
Inserting this above yields: \(| (3.87)| \lesssim \sum _{i=0}^k \Vert \psi \Vert _{J^{i+2}}^2\). Next, after two integrations by parts in y, and one in x:
Expanding the product rule above yields:
Inserting above yields: \(|(3.89)| \lesssim \sum _{i=0}^k \Vert \psi \Vert _{J^{i+2}}^2\). The next term from \(A(\psi )\) in definition (1.29) is:
The final term from \(A(\psi )\) in definition (1.29) is:
This concludes the proof of the desired estimate, (3.79). \(\square \)
Lemma 3.10
Proof
To keep notations simple, we will prove the \(k = 1\) case, with the \(k \ge 2\) cases following identically. We will proceed term by term from the commutator expression in (2.17). First,
Next,
Next,
We will now move to the higher-order terms, starting with:
Next, again integrating by parts several times:
The final term from \(A(\psi )\), which after integrating by parts several times in the same way as above,
This concludes the proof of (3.93). \(\square \)
Lemma 3.11
Proof
This estimate proceeds in the same manner as those from (3.93), with the adjustment that the extra derivative in the multiplier from (3.100) is accounted for by the increment in order on the right-hand sides of (3.100) versus (3.93). Indeed, let us take the highest order term from the commutator, \([\partial _x, A]\psi \):
The first term on the right-hand side above can be controlled by \(\Vert \psi \Vert _{J^3}^2\), as can be seen from a comparison to (3.69) with \(k = 1\). The remaining terms work identically. \(\square \)
Using the above calculations, we may repeat the energy and positivity estimates, for \(k \ge 1\).
Lemma 3.12
((k + 1)'th order Auxiliary Energy Estimate) Let\(k = 1,2\). Then,
Proof
We apply the operator \(\partial _x^k\) to the system (1.34):
We subsequently apply the multiplier \(\partial _x^k \psi x^{2k} \rho _{k+1}^{2k} \chi _{L,\alpha }^2\):
The desired estimate now follows using similar calculations as in Lemma 3.3. \(\square \)
Lemma 3.13
((k + 1)'th order Auxiliary Positivity Estimate)
Proof
We apply the multiplier \(\partial _x^{k+1} \psi x^{2k+1} \rho _{k+1}^{2k+1} \chi _{L,\alpha }^2 \) to the system (3.103):
The desired estimate now follows using similar calculations as in Lemma 3.4. \(\square \)
4 Nonlinear Existence of Auxiliary Systems
For this section, it is necessary to be more precise with notation; we will index solutions by \((\alpha , N)\) and also specify domains over which norms are being taken. We shall also transition our right-hand sides from being generic (F, G) to being the particular right-hand sides of interest, \(({{\tilde{f}}}, g)\) as defined in (1.36). Our intention now is to study the map, \(M^\alpha \):
which corresponds to the system written in vorticity form:
A fixed point of (4.2) corresponds to the desired solution of (1.37).
Lemma 4.1
Suppose\(\Vert {{\bar{u}}}^{\alpha , N}, {{\bar{v}}}^{\alpha , N}\Vert _{Z(\Omega ^N)} \le 1\). Fix any open set\(B \subset \Omega ^N\). Let\(\alpha > 0\)and\(N>> 1\). Solutions\(\psi ^{\alpha , N}\), or equivalently\([u^{\alpha , N},v^{\alpha , N}]\), to the system (4.2) satisfy the following estimates, independent ofN, where\(\omega (N_i)\)is based on universal constants:
The following energy and positivity estimates hold:
Finally, one has:
All constants appearing in the above estimates are independent of\((\alpha , N)\).
Proof of Estimate (4.3)
This follows by repeating the proofs of elliptic regularity in Subsection 2.1, namely Lemmas 2.11 and 2.13 in [4], to the new system, (4.2). At this point, one repeats the estimates in Subsection 2.2 of [4], which hold independent of any equation. \(\square \)
Proof of Estimate (4.4)
This follows from Lemmas 3.12–3.13, and subsequently comparing \(\Vert \cdot \Vert _{J^k}\) with \(\Vert \cdot \Vert _{H^k_w}\). \(\square \)
Proof of Estimate (4.5)
This follows by repeating the proof of Lemma 4.1 of [4]. \(\square \)
Motivated by (4.5), we define the notation:
Lemma 4.2
(Properties of \(M^\alpha \)) Fix any\(\alpha > 0\)and any\(N > 0\), and\(\gamma , \kappa > 0\)arbitrarily small.
Proof
The outline of this proof is as follows. The map \(M^\alpha \) is shown to be well-defined in the appropriate domains and codomains, according to (1) above. Continuity of \(M^\alpha \) is investigated by considering differences, and compactness of \(M^\alpha \) is obtained using our compactness lemmas above. One then applies a fixed point argument to prove (3) and (4).
(1) Suppose \([{{\bar{u}}}, {{\bar{v}}}] \in Z(\Omega ^N)\). This implies that \(({{\tilde{f}}},g) \in H^{-1}_2\), so by (3.67), the map \(M^\alpha \) is well-defined on \(Z(\Omega ^N)\). Lemma 2.7 and the definition of \(H^2_w(\Omega ^N)\). Definition 2.1 ensures that \([u^\alpha , v^\alpha ]\) are contained in \(\overline{C^\infty _{0,D}}^{\Vert \cdot \Vert _{X_1}}\). Supposing the pre-images are contained in the unit ball of \(Z(\Omega ^N)\), \(\Vert {{\bar{u}}}^{\alpha , N}, {{\bar{v}}}^{\alpha , N}\Vert _{Z(\Omega ^N)} \le 1\), one has estimate (4.5), which implies that \(M^\alpha ({{\bar{u}}}, {{\bar{v}}}) \in B_Z(1)\).
(2) To check continuity of the map \(M^\alpha \) on \(B_Z(1)\), suppose:
where
Define the notation for the differences,
By consulting (4.2), one then obtains the following system satisfied by the differences:
We may then repeat the estimates which resulted in (4.3)–(4.5) to obtain:
The only non-trivial calculation when repeating the estimates which resulted in (4.3)–(4.5) is the following:
A straightforward calculation gives:
For the first term in the right-hand side above, we estimate
This then gets absorbed into the left-hand side of (4.13). For the second term on the right-hand side above, we estimate:
where we have used (4.9) coupled with (4.5) to conclude that: \(\Vert u^{\alpha , N}_1\Vert _{{\mathcal {F}}(\Omega ^N)} \lesssim \epsilon ^{\frac{1}{4}-\gamma - \kappa }\). The weight, \(x^m\), arises from the definition (1.29), and consequently in (2.2). The first term on the right-hand side of (4.16) is absorbed into the left-hand side of (4.13), whereas the second term contributes to the right-hand side of (4.13). All of the remaining calculations which produced (4.5) can be repeated in a similar fashion. Estimate (4.13) then implies the continuity of \(M^\alpha \) on \(B_Z(1)\). The modulus of continuity of \(M^\alpha \) is \(\frac{1}{\alpha ^2}\), which prevents \(M^\alpha \) from being a contraction map. Nevertheless, continuity is retained for all \(\alpha > 0\).
We now turn to compactness. According to Lemma 3.2, (4.5) shows that \(M^\alpha (B_Z(1))\) is compactly embedded in \(B_Z(1)\) so long as m is sufficiently large.
(3) and (4) Consider the family of solutions:
By (4.1) and linearity of \(L_\alpha ^{-1}\), this occurs if and only if
By repeating the estimates which culminated in (4.5), one sees the uniform in \(\lambda \) bound:
Thus, Schaefer’s fixed point theorem applied to the convex subset \(B_Z(1) \subset Z(\Omega ^N)\) produces a fixed point, \([u^{\alpha , N}, v^{\alpha , N}] \in B_Z(1)\). The estimate it obeys follows from (4.5). \(\square \)
5 Nonlinear Existence
We now need to pass to the limit as \(\alpha \rightarrow 0\) and as \( N \rightarrow \infty \). The fixed point of the system (4.2), from Lemma 4.2 satisfies the following integral identity for any \(\phi \in C^\infty _0(\Omega ^N)\):
First, we shall pass to the limit as \(\alpha \rightarrow 0\), fixing an N. To do so, we first use (4.7) to obtain a weak subsequential limit point:
It is now our task to pass to the limit in the equation, (5.1), along the subsequence \(\alpha \rightarrow 0\). Given a test-function, denote by \(U_\phi \) to be the support of \(\phi \). As \(U_\phi \) is bounded, we have Poincare inequalities available:
For all of the linear terms, we use the weak convergence in \((X_1 \cap X_2 \cap X_3)(\Omega ^N)\):
Finally, we turn to the nonlinear terms for which we integrate by parts:
Fixing a compactly supported \(\phi \), we can localize the integrations above to \(U_\phi \). On this set, the weak convergence of \(u^{N,\alpha } {\mathop {\rightharpoonup }\limits ^{X_1 \cap X_2 \cap X_3}} u^N\) implies strong convergence in \(L^2\). Thus,
The right-hand side converges to zero. The same bound works for all of the other nonlinear terms. Thus, the weak limit \([u^N,v^N]\) or equivalently \(\psi ^N\) satisfies the weak formulation:
The weak limit \([u^N,v^N]\) must satisfy the bound:
independent of N. We may now repeat this exact procedure with the subsequential N limit: denote by [u, v] and \(\psi \) the subsequential \((X_1 \cap X_2 \cap X_3)(\Omega )\)-weak limit as \(N \rightarrow \infty \), guaranteed by (5.9). One then passes to the limit in Eq. (5.8) to obtain:
with the limit satisfying:
We now state the main existence result.
Theorem 5.1
For\(\epsilon , \delta \)sufficiently small,\(\kappa > 0\)small, and\(0 \le \gamma < \frac{1}{4}\), there exists a solution to the system (1.1)–(1.3), (1.4), (1.5) satisfying:
Proof
Estimate (5.11) implies enough regularity to integrate by parts identity (5.1) to:
which then implies that the PDE is satisfied pointwise in \(\Omega \). The boundary conditions (1.4) are satisfied by elements in \((X_1 \cap X_2 \cap X_3)(\Omega )\). From here, one applies the available embedding theorems for the norm Z which gives the estimate (5.12). A nearly identical proof to Lemma 2.11 in [4] then yields:
One now bootstraps the estimate in Lemma 2.13, [4] in the identical manner. This gives estimate (5.12). We have verified that \([u,v] \in Z(\Omega )\) satisfies (1.1)–(1.3), (1.4), (1.5). \(\square \)
6 Uniqueness
In this final section, we prove uniqueness of the solution [u, v] from Theorem 5.1. Suppose there existed two solutions, \([u_1, v_1]\) and \([u_2, v_2]\) to the system in (1.1)–(1.3), (1.4), (1.5). Define:
Then the new unknowns satisfy:
together with the divergence-free condition, \({{\hat{u}}}_x + {{\hat{v}}}_y = 0\), and also satisfy the boundary conditions:
Going to vorticity,
We shall repeat the basic energy and positivity estimates using a slightly weaker weight. It is convenient to work with the weak formulation, which is given in (5.10). Then, \({{\hat{u}}}, {{\hat{v}}}\) satisfy the following:
for all \(\phi \in C_0^\infty (\Omega )\). We make the notational convention that
Lemma 6.1
There exists a\(0< b < 1\), sufficiently close to 0, depending only on universal constants, such that for\(\delta , \epsilon \)sufficiently small and\(\epsilon<< \delta<< b\), the solutions\([{{\hat{u}}}, {{\hat{v}}}] \in Z\)to the system (6.2)–(6.3) with boundary conditions (6.4) satisfy the following estimate:
where
Proof
The estimate will follow upon applying the multiplier \({{\hat{\psi }}} \cdot x^{-2b}\) to the system in (6.5). To work rigorously, we will apply approximate multipliers, and work with the weak formulation given in (6.6). Fix \([{{\hat{u}}}^{{n}}, {{\hat{v}}}^{(n)}, {{\hat{\psi }}}^{(n)}] \in C^\infty _0(\Omega )\), such that:
where \(X_1\) is defined in (1.40) of [4]. Within the notation of (6.6), \(\phi = {{\hat{\psi }}}^{(n)}x^{-2b}\). The existence of the sequence specified in (6.12) is guaranteed by \([{{\hat{u}}}, {{\hat{v}}}] \in Z(\Omega )\). That \(\phi \) is compactly supported in (x, y) follows from the representations:
Let us first treat the second-order terms:
The first two terms from (6.14) above are:
We shall take the limit as \(n \rightarrow \infty \) above. According to the definition (1.40), the convergence in (6.12) implies:
Expanding the third term from (6.14),
By referring to the definition of \(X_1\) in (1.40) and (6.12), we may pass to the limit:
and:
Integrating by parts the final two terms above in (6.19), and referring to estimate (1.46),
and similarly, to treat the final term in (6.19), we appeal to the estimates in (1.46):
Therefore, summarizing the highest order calculation:
To go from (6.21) to (6.22), we have used the Hardy inequality in the x-direction. We will now address the profile terms arising from \(S_u({{\hat{u}}}, {{\hat{v}}})\) in the weak formulation (6.6), whose definition has been given in (1.5):
We will first pass to the limit in (6.23), using the definition of \(X_1\) in (1.40), which gives:
We proceed to treat each term in (6.24), starting with:
according to estimates (1.14), (1.19), so long as \(\delta \) is taken small relative to b. For the M-limit above, we have used estimate (1.46), which is valid so long as \(b > 0\). For the second term in (6.24), we again appeal to estimates (1.14), (1.20):
For the third term, we shall split \(u_R = u^{n-1,p}_R + \epsilon ^{\frac{n}{2}} u^n_{pR} + u^E_R\). First, we apply estimate (1.16):
Next, according to estimate (1.18),
Finally, the Eulerian contribution is handled by an application of (1.20):
The fourth term from (6.23), upon using estimate (1.14) and (1.20), reads:
Summarizing these calculations,
We have absorbed the \({{\hat{u}}}_y\) terms into (6.21), and taken \(\delta \) sufficiently small relative to b. We shall now address the profile terms from \(S_v\):
We may take \(n \rightarrow \infty \) above due to the definition of \(X_1\) from (1.40) and (6.12):
We will now proceed to treat each term in (6.32). The first profile term, \(u_R v_x\) is the most delicate:
First,
The M-limit above vanishes due to (1.46). Staying with the term (6.33):
Combining the positive terms in (6.36) and (6.34), the total positive contribution is \(\int \int 3b\epsilon u_R {{\hat{v}}}^2 x^{-2b-1}\). For the final term in (6.36), we will now give the estimate:
By collecting terms and rearranging, we obtain:
This then implies:
Inserting this into (6.36), one arrives at:
so long as b is sufficiently close to 0, by the following calculation:
Thus, taking b sufficiently small, and recalling the positive contributions from (6.36) and (6.34), we have:
The remaining terms from (6.34) and (6.36) are then estimated in terms of (6.42) using the smallness of \({\mathcal {O}}(\delta )\). Summarizing, we have established control over:
for a constant independent of small \(\delta \) and b. We will now move to the second term from (6.31), for which we recall estimates (1.10) and (1.19):
For the third term from (6.31), we use Young’s inequality and estimates (1.12), (1.22):
For the final term from (6.31), we use Young’s inequality and estimates (1.12), (1.22):
Summarizing these last few terms, we obtain:
The final task is to turn to the right-hand side. Reading from (6.6), and (6.2)–(6.3):
where we have passed to the limit using again the definition of \(X_1\) from (1.40). Combining (6.21), (6.30), (6.47), and (6.48), one obtains the desired result, estimate (6.8). \(\square \)
We now repeat the positivity estimate, with a correspondingly weaker weight in order to close the above energy estimate. We refer the reader to Proposition 3.4 in [4] for a comparison.
Lemma 6.2
Fix any\(0< b < 1\). Let\(\delta , \epsilon \)be sufficiently small relative to universal constants, and\(\epsilon<< \delta \). Then for\([{{\hat{u}}}, {{\hat{v}}}] \in Z\)solutions to (6.2)–(6.3) with boundary conditions (6.4) satisfy the following estimate:
Proof
The estimate will follow upon applying the multiplier \({{\hat{v}}} x^{1-2b}\) to the system (6.5). In order to proceed formally, we must start with the weak formulation given in (6.6), and select the test function:
where \(X_1\) is defined in (1.40). Turning to the weak formulation in (6.6), we will first expand the second-order terms:
We first arrive at the first two terms on the right-hand side of (6.51):
Referring to the definition of \(X_1\) in (1.40), according to (6.50), we may pass to the limit as \(n \rightarrow \infty \), and appeal to the estimates in (1.46) and (1.47), to obtain:
Again referring to the definition in (1.40), the third term from (6.51) is treated by:
Integrating by parts the first term on the right-hand side of (6.54), and appealing to estimate (1.47):
Integrating by parts the second term on the right-hand side of (6.54), and again appealing to estimates (1.46)–(1.48) for the M-limit below:
Combining the above estimates:
Hence, summarizing (6.52)–(6.57):
We will now turn to the profile terms from \(S_u\), which upon consultation with (6.6), the definition in (1.40), and (6.50), read:
We now turn our attention to (6.59). The first term yields the desired positivity:
Next, we shall split \(u_R = u^P_R + u^E_{R}\), and use estimate (1.16) and (1.18) for (6.62) below and (1.20) for (6.63) below:
For the fourth term from (6.59), by estimates (1.12) and (1.22):
Summarizing the last four calculations:
The final three terms appearing on the right-hand side above all appear on the right-hand side of estimate (6.49). Turning now to the profile terms, from \(S_v\), for which we read (6.6) with \(\phi = {{\hat{v}}}^{(n)} x^{1-2b}\), appeal to (1.40) and (6.50), giving ultimately:
We will treat each term in (6.66). For the first term from (6.66):
Above we have used (1.14) and (1.20). For the second term, we integrate by parts:
The above M-limit vanishes according to estimates (1.46), and we have used estimates (1.10) and (1.19). For the third term, we recall estimates (1.12), (1.22):
For the fourth term, we integrate by parts and appeal to (1.46), (1.10)–(1.12), and (1.19):
Summarizing these four terms,
On the right-hand side, appealing again to (6.50), and the definitions of \({{\hat{f}}}, {{\hat{g}}}\) in (6.2)–(6.3), one obtains:
Placing the above estimates together yields the estimate (6.49). \(\square \)
We will now introduce some notations involving the weaker weight of \(x^{-b}\). The reader should recall the definitions of the cutoff functions introduced in (1.38)–(1.39). The energy norms are defined as follows:
Definition 6.3
The norms \(Y_{2,b}, Y_{3,b}\) are strengthenings of \(X_{2,b}, X_{3,b}\) near the boundary, \(x = 1\), and defined through:
Definition 6.4
The norm \(Z_b\) is defined through:
Next, we record the second- and third-order versions of the energy and positivity estimates, which mimic Propositions 3.8, 3.13, 3.16 and 3.18 in [4]. We will omit most details, and record only those differences which arise.
Lemma 6.5
(Second-Order Energy Estimate) Fix any\(0< b < 1\). Let\(\delta , \epsilon \)be sufficiently small relative to universal constants, and\(\epsilon<< \delta \). Then for\([{{\hat{u}}}, {{\hat{v}}}] \in Z\)solutions to (6.2)–(6.3):
where [recall the definition of\(\rho _2\)from (1.39)]:
Proof
Differentiating the weak formulation gives:
For the second-order energy estimate, we select \(\phi = \rho _2^2 {{\hat{v}}}^{(n)} x^{2-2b}\), where:
Let us turn to the highest order terms:
One now checks according to the definition (1.40), that (6.84) suffices to pass to the limit in the above identity, which upon integrating by parts in x yields:
where \(|J| = |\iint(c_0 \epsilon ^2 {{\hat{v}}}_x^2 \partial _{xx}(\rho _2^2 x^{2-2b}) + c_1 \epsilon ^2 {{\hat{v}}}^2 \partial _{x}^4 (\rho _2^2 x^{2-2b}))| \lesssim \Vert u,v\Vert _{X_{1,b}}^2\). From here, repeating the calculations in Proposition 3.8 of [4] gives the desired result, where the required integrations by parts are justified upon using that \(b > 0\), combined with the estimates in (1.46)–(1.48). These justifications are analogous to those in Lemma 6.1, and so we omit the details. \(\square \)
Lemma 6.6
(Second-Order Positivity Estimate) Fix any\(0< b < 1\). Let\(\delta , \epsilon \)be sufficiently small relative to universal constants, and\(\epsilon<< \delta \). Then for\([{{\hat{u}}}, {{\hat{v}}}] \in Z\)solutions to (6.2)–(6.3):
Proof
We start again with the weak formulation in (6.83). Fix a large \(0< L < \infty \). We then make the selection: \(\phi = {{\hat{v}}}^{(n)}_x \cdot \rho _2^3 x_L^{3-2b}\), where the weight \(x_L\) is defined via: \(x_L := \big ( a_L *\phi _L \big ) \chi \big (\frac{x}{10L} \big )\). Define the domain: \(\Omega _L := \{x: 3< x < 50L + 100 \}\), so that \({{\hat{v}}}_x \cdot \rho _2^3 x_L^{3-2b} = 0\) on \(\Omega _L^C\). The sequence \({{\hat{v}}}^{(n)}\) is selected according to:
The existence of such a sequence is guaranteed due to the standard Sobolev space theory, because we are now in the un-weighted setting. It is now straightforward to repeat all estimates in Proposition 3.13 of [4] using the test function \(\phi \). Upon doing so, we pass to the limit first as \(n \rightarrow \infty \), and then as \(L \rightarrow \infty \) to obtain the desired estimate. \(\square \)
Lemma 6.7
(Third-Order Energy Estimate) Fix any\(0< b < 1\). Let\(\delta , \epsilon \)be sufficiently small relative to universal constants, and\(\epsilon<< \delta \). Then for\([{{\hat{u}}}, {{\hat{v}}}] \in Z\)solutions to (6.2)–(6.3):
where
Proof
The first step is to differentiate the weak formulation (6.83) yet again, which formally takes place using difference quotients, yielding:
Fix any L large, finite. The selection of test function is now \(\phi := {{\hat{v}}}^{(n)}_x \rho _3^4 x_L^{4-2b}\), where the sequence:
From here, repeating the estimates given in Proposition 3.16 of [4], and sending \(n \rightarrow \infty \) and then \(L \rightarrow \infty \) gives the desired result. \(\square \)
Lemma 6.8
(Third-Order Positivity Estimate) Fix any\(0< b < 1\). Let\(\delta , \epsilon \)be sufficiently small relative to universal constants, and\(\epsilon<< \delta \). Then for\([{{\hat{u}}}, {{\hat{v}}}] \in Z\)solutions to (6.2)–(6.3):
Proof
Again, fix any L large, finite. The selection of the test function is now \(\phi := {{\hat{v}}}^{(n)}_{xx} \rho _3^5 x_L^{5-2b}\), where the sequence \([{{\hat{u}}}^{(n)}, {{\hat{v}}}^{(n)}]\) is selected according to:
From here, repeating the estimates in Proposition 3.18 of [4], and sending \(n \rightarrow \infty \) and then \(L \rightarrow \infty \) gives the desired result. \(\square \)
Piecing together the above set of estimates,
Proposition 6.9
Let\(\delta , \epsilon \)be sufficiently small relative to universal constants, and\(\epsilon<< \delta<< b\). Then for\([{{\hat{u}}}, {{\hat{v}}}] \in Z\)solutions to (6.2)–(6.3):
where\({\mathcal {W}}_{i,b}\)have been defined in (6.11), (6.82), (6.91).
By repeating the analysis in Section 2 of [4], one has:
Lemma 6.10
Let\(\delta , \epsilon \)be sufficiently small relative to universal constants, and\(\epsilon<< \delta<< b\). Then for\([{{\hat{u}}}, {{\hat{v}}}] \in Z\)solutions to (6.2)–(6.3):
Due to (6.96), we will now turn to estimating \({\mathcal {W}}_{i,b}\)
Lemma 6.11
Let\({\mathcal {W}}_{1,b}, {\mathcal {W}}_{2,b}, {\mathcal {W}}_{3,b}\)be as in (3.10), (3.79) and (3.216) in [4]. Then:
where\(C(b) \uparrow \infty \) as \(b \downarrow 0\).
Proof
We will work with the expression:
Concerning \({\mathcal {W}}_{1,b}\), let us bring particular attention to the following term from \(\int \int | {{\hat{f}}} | \cdot |{{\hat{u}}}| x^{-2b}\):
The above term requires the weight of \(x^{-2b}, b > 0\), in order to apply the Hardy inequality. Indeed, this was not required for the existence proof (see calculation (4.5) in [4]), because the structure of \(vu_y \cdot u\) enabled us to integrate by parts, unlike in the present situation. The remaining terms in \({\mathcal {W}}_{1,b}\), and all terms in \({\mathcal {W}}_{2,b}, {\mathcal {W}}_{3,b}\) are treated nearly identically to Lemma 4.1 of [4], and so we omit repeating those calculations. \(\square \)
Corollary 6.12
Fix\(0< b < 1\)sufficiently small, relative to universal constants. Suppose\(\epsilon , \delta \)are sufficiently small, such that\(\epsilon<< \delta<< b\). Then\({{\hat{u}}}, {{\hat{v}}} = 0\).
Proof
Combining estimate (6.98) and (6.97) with estimate (5.12) yields:
For \(\epsilon \) sufficiently small, this then implies \(\Vert {{\hat{u}}}, {{\hat{v}}}\Vert _{Z_b} = 0\). Upon consultation with the norm \(Z_b\), and (6.4), this implies that \({{\hat{u}}}, {{\hat{v}}} = 0\). \(\square \)
Remark 6.13
We have controlled the second- and third-order energy norms, (6.74)–(6.75) in order to treat the term \(\int \int {{\hat{v}}} u^{(1)}_y |{{\hat{u}}}| x^{-2b}\), which appears in (6.101). This term forces us to control \(\Vert {{\hat{v}}} x^{\frac{1}{2}-b}\Vert _{L^\infty }\). One cannot get around placing this term in \(L^\infty \) (for instance by integrating by parts from \(u^{(1)}_y\)) because this produces suboptimal decay rates, according to (2.92)–(2.93) in [4].
This then immediately establishes Theorem 1.3.
References
Blum, H., Rannacher, R.: On the boundary value problem of the Biharmonic operator on domains with angular corners. Math. Methods Appl. Sci. 2, 556–581 (1980)
Bricmont, J., Kupiainen, A., Lin, G.: Renormalization group and asymptotics of solutions of nonlinear parabolic equations. Commun. Pure. Appl. Math. 47, 893–922 (1994)
Iyer, S.: Global steady Prandtl expansion over a moving boundary I. Peking Math. J. (2019). https://doi.org/10.1007/s42543-019-00011-4
Iyer, S.: Global steady Prandtl expansion over a moving boundary II. Peking Math. J. (2019). https://doi.org/10.1007/s42543-019-00014-1
Acknowledgements
The author thanks Yan Guo for many valuable discussions regarding this research. The author also thanks Bjorn Sandstede for introducing him to the paper [2].
Author information
Authors and Affiliations
Corresponding author
Additional information
This research was completed under partial support by NSF Grant 1209437.
Rights and permissions
About this article
Cite this article
Iyer, S. Global Steady Prandtl Expansion over a Moving Boundary III. Peking Math J 3, 47–102 (2020). https://doi.org/10.1007/s42543-019-00015-0
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s42543-019-00015-0