On well-posedness of a mildly dissipative family of active scalar equations in borderline Sobolev spaces
Extracted source text ยท arXiv 2309.05844
\documentclass[twoside, reqno]{amsart}
\usepackage[left=2.5cm, right=2.5cm, top=3cm, bottom=2.5cm]{geometry}
\usepackage[colorlinks=true, pdfstartview=FitV, linkcolor=blue,citecolor=blue, urlcolor=blue]{hyperref}
\usepackage[usenames]{xcolor}
\definecolor{labelkey}{rgb}{0,0,1}
\definecolor{Red}{rgb}{0.7,0,0.1}
\definecolor{Green}{rgb}{0,0.7,0}
\usepackage{amsfonts, amssymb, amsmath, amsthm, mathrsfs, bbm, cjhebrew, gensymb, textcomp, mathtools, dsfont,tikz}
\usepackage[normalem]{ulem}
\usepackage{commath, setspace, subcaption, parcolumns, multirow, multicol, accents, comment, marginnote, verbatim, empheq, enumerate, stackrel, enumitem}
\usepackage{lineno}
%\linenumbers
\makeatletter
\def\namedlabel#1#2{\begingroup
#2%
\def\@currentlabel{#2}%
\phantomsection\label{#1}\endgroup
}
\makeatother
\renewcommand{\descriptionlabel}[1]{%
\hspace\labelsep \upshape\bfseries #1%
}
\usepackage[capitalize,nameinlink,noabbrev]{cleveref}
\usepackage{graphicx}
\usepackage{epsfig}
\usepackage{psfrag}
\usepackage{float}
\usepackage[active]{srcltx}
\providecommand*\showkeyslabelformat[1]{{\normalfont \tiny#1}}
\newtheorem{Thm}{Theorem}[section]
\newtheorem{Cor}[Thm]{Corollary}
\newtheorem{Prop}[Thm]{Proposition}
\newtheorem{Lem}[Thm]{Lemma}
\newtheorem{Rmk}{Remark}[section]
\numberwithin{equation}{section}
\newcommand{\al}{\alpha}
\newcommand{\be}{\beta}
\newcommand{\de}{\delta}
\newcommand{\De}{\Delta}
\newcommand{\eps}{\epsilon}
\newcommand{\veps}{\varepsilon}
\newcommand{\io}{\iota}
\newcommand{\ph}{\varphi}
\newcommand{\gam}{\gamma}
\newcommand{\Gam}{\Gamma}
\newcommand{\kap}{\kappa}
\newcommand{\lam}{\lambda}
\newcommand{\Lam}{\Lambda}
\newcommand{\s}{\sigma}
\newcommand{\xe}{\xi-\eta}
\newcommand{\tiltht}{\widetilde{\theta}}
\newcommand{\tht}{\theta}
\newcommand{\Tht}{\Theta}
\newcommand{\om}{\omega}
\newcommand{\Om}{\Omega}
\newcommand{\ze}{\zeta}
\newcommand{\vs}{\varsigma}
\newcommand{\vr}{\varrho}
\newcommand{\lpj}{\triangle_j}
\newcommand{\lpk}{\triangle_k}
\newcommand{\tlpj}{\triangle_j^e}
\newcommand{\tlpk}{\triangle_k^e}
\newcommand{\ZZ}{\mathbb{Z}}
\newcommand{\RR}{\mathbb{R}}
\newcommand{\NN}{\mathbb{N}}
\newcommand{\WW}{\textsc{W}}
\newcommand{\lb}{\big\langle}
\newcommand{\rb}{\big\rangle}
\newcommand{\lbn}{\langle}
\newcommand{\rbn}{\rangle}
\newcommand{\goesto}{\rightarrow}
\newcommand{\Sob}[2]{\lVert#1\rVert_{#2}}
\newcommand{\nrm}[1]{\lVert#1\rVert}
\newcommand{\bdy}{\partial}
\newcommand{\til}[1]{\widetilde{#1}}
\newcommand{\Ae}{\abs{\eta}}
\newcommand{\Ax}{\abs{\xi}}
\newcommand{\Axe}{\abs{\xi-\eta}}
\newcommand{\Gdot}{\dot{G}}
\newcommand{\Hdot}{\dot{H}}
\newcommand{\Bdot}{\dot{B}}
\newcommand{\Acal}{\mathcal{A}}
\newcommand{\Bcal}{\mathcal{B}}
\newcommand{\Dcal}{\mathcal{D}}
\newcommand{\Ecal}{\mathcalE_\nu}
\newcommand{\Gcal}{\mathcal{G}}
\newcommand{\Ft}{\mathcal{F}}
\newcommand{\ld}{(\ln(I-\Delta))}
\newcommand{\ldm}{(\ln(I-\Delta))^{\mu+\frac{1}{2}}}
\newcommand{\ldmu}{(\ln(I-\Delta))^{\mu}}
\newcommand{\lne}{(\ln(1+\abs{\eta}^{2}))}
\newcommand{\lnx}{(\ln(1+\abs{\xi}^{2}))}
\newcommand{\lnxe}{(\ln(1+\abs{\xi-\eta}^{2}))}
\newcommand{\Ba}{\mathcal{B}(2^{k-3})}
\newcommand{\Baa}{\mathcal{B}(2^{j-3})}
\newcommand{\An}{\mathcal{A}(2^{k-1},2^{k+1})}
\newcommand{\Ann}{\mathcal{A}(2^{j-3},2^{j+2})}
\newcommand{\indFn}[1]{1 \! \! 1_{#1}}
\DeclareMathOperator{\supp}{supp}
\DeclareMathOperator*{\esssup}{ess\,sup}
\DeclareMathOperator*{\Div}{div}
\usepackage{tikz}
\def\checkmark{\tikz\fill[scale=0.4](0,.35) -- (.25,0) -- (1,.7) -- (.25,.15) -- cycle;}
\title[On well posedness of mildly dissipative active scalars in critical spaces]{On well-posedness of a mildly dissipative family of active scalar equations in borderline Sobolev spaces}
\author{Anuj Kumar, Vincent R. Martinez}
\date{Septemeber 11, 2023}
\begin{document}
\maketitle
\begin{abstract}
This paper considers a family of active scalar equations which modify the generalized surface quasi-geostrophic (gSQG) equations through its constitutive law or dissipative perturbations. These modifications are characteristically mild in the sense that they are logarithmic. The problem of well posedness, in the sense of Hadamard, is then studied in a borderline setting of regularity in analog to the scaling-critical spaces of the gSQG equations. A novelty of the system considered is the nuanced form of smoothing provided by the proposed mild form of dissipation, which is able to support global well-posedness at the Euler endpoint, but in a setting where the inviscid counterpart is known to be ill-posed. A novelty of the analysis lies in the simultaneous treatment of modifications in the constitutive law, dissipative mechanism, and functional setting, which the existing literature has typically treated separately. A putatively sharp relation is identified between each of the distinct system-modifiers that is consistent with previous studies that considered these modifications in isolation. This unified perspective is afforded by the introduction of a linear model equation, referred to as the \textit{protean system}, that successfully incorporates the more delicate commutator structure collectively possessed by the gSQG family and upon which each facet of well-posedness can be reduced to its study.
\end{abstract}
{\noindent \small {\it {\bf Keywords: generalized surface quasi-geostrophic (SQG) equation, borderline spaces, Hadamard well-posedness, existence, uniqueness, continuity with respect to initial data, logarithmic dissipation, mild regularization, smoothing mechanism, Foias-Temam Gevrey norm-type technique}
} \\
{\it {\bf MSC 2010 Classifications:} 76B03, 35Q35, 35Q86, 35B45
} }
\setcounter{tocdepth}{1}
\tableofcontents
\section{Introduction}
This article is concerned with the well-posedness of the initial value problem in a borderline regularity setting for a family of dissipative active scalar equations over the whole plane $\RR^{2}$:
\begin{align}\label{eq:dgsqg}
\begin{cases}
\partial_{t}\theta + m(D)\tht+u \cdot \nabla \theta=0, \quad\tht(0,x)=\tht_{0}(x),\quad x\in\RR^2,\quad t>0,\\
u=\nabla^{\perp}{\psi},\quad \Delta {\psi}=\Lambda^{\beta}p(D)\theta,
\end{cases}
\end{align}
where $\be\in[0,2]$.
The operators $m(D)$ and $p(D)$ denote Fourier multiplier operators, which are assumed to be radial and of logarithmic type. Roughly speaking, $p(D)$ will belong to a class of multipliers that may decay at most logarithmically up to some order.
The case where $p(D)$ provides a regularizing effect in the constitutive law, that is it decays in frequency, was originally studied in \cite{ChaeWu2012} for the purpose of locating a minimal degree of regularization that allows one to have a local existence theory for the corresponding inviscid version of \eqref{eq:dgsqg} in a critical regularity setting; the main example of interest being the 2D Euler equation (in vorticity form) in the critical Sobolev space $\Hdot^1(\RR^2)$, which indicates the Sobolev regularity threshold where the velocity fails to automatically be Lipschitz. In the same spirit, we propose an alternative mechanism for regularization that is \textit{dissipative}, as captured by the multiplier, $m(D)$, and wish to locate the smallest possible degree of such dissipation that supports a local existence theory (in the sense of Hadamard). Such a form of regularization is categorically different from an inviscid regularization mechanism since the dissipativity of $m(D)$ is expected to instantaneously confer additional regularity to the solution. It is therefore the precise quantification of this instantaneous smoothing effect that must be exploited to recover local well-posedness in a borderline Sobolev regularity setting. Generally speaking, the overarching goal of this work is to identify a putatively minimal relation between $m(D)$ and $p(D)$ that guarantees a local existence theory in a regularity setting that is critical for the corresponding inviscid, that is, dissipation-less equation. Note that since the dissipatively modified equation \eqref{eq:dgsqg} considered here will not possess a scaling symmetry, we adopt the terminology of \textit{borderline regularity} in lieu of critical regularity.
When $p(D)=I$, \eqref{eq:dgsqg} corresponds to a dissipative perturbation of the inviscid generalized surface quasi-geostrophic (gSQG) equation, which is given by
\begin{align}\label{eq:gsqg}
\partial_{t}\theta +u \cdot \nabla \theta=0,\quad
u=\nabla^{\perp}{\psi},\quad \Delta {\psi}=\Lambda^{\beta}\theta,\quad 0\le\beta\le2.
\end{align}
Observe that for $\beta \in [0,1]$, \eqref{eq:gsqg} interpolates between the 2D incompressible Euler equation in vorticity form ($\be=0$) and the SQG equation ($\beta=1$). The regime $\be \in (1,2)$ represents a family of active scalar equations with even more singular velocities than SQG, while the endpoint $\be=2$ constitutes a trivial case where the streamfunction can be identified as the advected scalar resulting in the reduction of \eqref{eq:gsqg} to the stationary equation $\bdy_t\tht=0$. This endpoint is often made nontrivial by modifying the streamfunction equation with a positive power of a logarithmic multiplier as in \eqref{eq:dgsqg}; such a modification was originally introduced in \cite{ChaeConstantinCordobaGancedoWu2012} and subsequently revisited for the fractionally dissipative counterpart of \eqref{eq:gsqg} in \cite{JollyKumarMartinez2020b}. The issue of whether singularities may develop in finite-time from smooth initial data remains an outstanding open problem for \eqref{eq:gsqg} when $\be\in(0,2]$, where the endpoint case, $\be=2$, is modified in the manner just described.
In this paper, the choice of dissipative perturbation will take on a logarithmic form; we will refer to such resulting perturbations as \textit{mildly dissipative}. This is to distinguish from \textit{weakly dissipative} perturbations, which refer to dissipative mechanism which are non-regularizing, such as damping effects due to friction, and \textit{strongly dissipative} perturbations, which we take to typically refer to dissipative operators such as the fractional Laplacian, ${\Lam}^\gam$, where $\Lam=(-\De)^{1/2}$, which instantaneously regularize the solution to become smooth in space. The choice of the terminology of \text{mild dissipation} is intended to indicate an intermediate form of dissipation that lies between \textit{weak} and \textit{strong}. As we will see below, \textit{mild dissipation} instantaneously confers additional regularity to the solution, albeit at a categorically weaker level than \textit{strong dissipation}.
The (homogeneous) Sobolev spaces, $\Hdot^{1+\be}(\RR^2)$, where $\be\in[1,2]$, is a critical space for \eqref{eq:gsqg}. As in the case of the 2D Euler equation in vorticity form, \textit{criticality} in this context refers to the threshold Sobolev regularity where the velocity $u$ fails to automatically be Lipschitz. From this point of view, the critical spaces identify a threshold for regularity above which local well-posedness of the corresponding initial value problem is to be expected (the \textit{subcritical regime}), and below which some form of ill-posedness may emerge (the \textit{supercritical regime)}. Classically, local well-posedness at the endpoint case $\be=0$, which corresponds exactly to the vorticity form of the 2D Euler equations described above, above the critical spatial regularity has been known at least since \cite{Kato1967}. In fact, in this subcritical setting, local well-posedness in the spaces $H^{s}(\RR^d)$, $s>1+d/2$, is known in any dimension $d\geq2$ (see \cite{BourguignonBrezis1974, EbinMarsden1969, Kato1967, Kato1972, Temam1974}). The analogous result for the family \eqref{eq:gsqg} was established in the works \cite{ChaeConstantinCordobaGancedoWu2012, HuKukavicaZiane2015}, where local well-posedness was established in $H^{1+\be+\eps}(\RR^2)$, for any $\eps>0$. In the presence of strong dissipation, global regularity of solutions in the case $\be\in(0,1]$ with $m(D)=\Lam^\gam$, where $\gam=\be$, was collectively established by the mathematical community in the works \cite{Resnick1995, ConstantinWu1999, KiselevNazarovVolberg2007, CaffarelliVasseur2010, KiselevNazarov2009, ConstantinVicol2012, ConstantinTarfuleaVicol2015, ConstantinIyerWu2008, MiaoXue2011}.
The issue of well-posedness for the Euler endpoint, $\be=0$, at critical regularity remained an outstanding open problem until the seminal work of J. Bourgain and D. Li \cite{BourgainLi2015}, where a mechanism for norm inflation was identified to establish strong ill-posedness in $H^{1}(\RR^2)$. An alternative approach to establishing ill-posedness was subsequently developed by T. Elgindi and M. Masmoudi in \cite{ElgindiMasmoudi2020}. In a series of recent papers, strong ill-posedness was also established in the range $\be\in(0,2]$ by D. C\'ordoba and L. Mart\'inez-Zoroa \cite{CordobaZoroa-Martinez2021, CordobaZoroa-Martinez2022} and I.-J. Jeong and J. Kim \cite{JeongKim2021}. In fact, in a very interesting pair of recent works by D. Chae, I.-J. Jeong, J. Na, and S.-J. Oh \cite{ChaeJeongNaOh2023a, ChaeJeongOh2023b}, the case $\be=2$ in \eqref{eq:gsqg} with constitutive law logarithmically modified so that $u=-\nabla^\perp(\ln(I-\De))\tht$, i.e., the Okhitani model (see \cite{ChaeConstantinCordobaGancedoWu2012, Okhitani2011, Okhitani2012}, was shown to be ill-posed in $H^s$, for $s>3$ \cite{ChaeJeongOh2023b}, but globally well-posed when the model possessed logarithmic-order dissipation \cite{ChaeJeongNaOh2023a}; similar ill-posedness results for a class of models generalizing the Okhitani case, as well as some dissipatively perturbed counterparts, were also obtained therein. We emphasize that the framework developed in the present article is complementary to that studied in \cite{ChaeJeongNaOh2023a, ChaeJeongOh2023b}; the reader is referred to \cref{rmk:okhitani} for a more detailed discussion.
\par
Complementary to the ill-posedness results mentioned above are the results of D. Chae and J. Wu \cite{ChaeWu2012} and M.S. Jolly with the present authors \cite{JollyKumarMartinez2020b}, where a mild inviscid regularization of \eqref{eq:gsqg} is studied in order to recover well-posedness in the borderline spaces. This regularization modifies the constitutive law for the velocity with a logarithmic multiplier operator as
\begin{align}\label{def:mod:gsqg}
u=-\nabla^{\perp}{\Lam}^{\be-2}\left(\ln\left(e-\Delta\right)\right)^{\til{\mu}}\tht, \quad \til{\mu}<0
\end{align}
Although \eqref{eq:gsqg} modified by \eqref{def:mod:gsqg} does not, strictly speaking, possess a scaling symmetry, we may nevertheless associate the borderline regularity space as $H^{1+\be}(\RR^2)$ since the Lipschitz regularity of velocity field $u$ requires the solution $\tht$ to lie in higher regularity spaces, i.e
\[
\nabla u \in L^{\infty} \quad \text{only if} \quad \tht \in H^s,\quad s>1+\be.
\]
In \cite{ChaeWu2012}, local existence and uniqueness was shown for $\be \in [0,1]$ provided that $\til{\mu}<-1/2$, while \cite{JollyKumarMartinez2020b} established local well-posedness for the more singular range $\be \in (1,2)$ provided, again, that $\til{\mu}<-1/2$. The key difference between the regimes $\be\in[0,1]$ and $\be\in(1,2)$ are in the identification of a suitable linear system that allows one to accommodate the more nuanced commutator structure of \eqref{eq:gsqg} when establishing stability-type estimates \textit{at the critical regularity level}. In particular, the need for such a system is crucial when establishing \textit{continuity of the solution map} since the classical estimates for the transport equation require control of $\Sob{\nabla u}{H^{\be}}$ and therefore one must exploit additional cancellation in the form of commutators. In an extension of the seminal work of J. Bourgain and D. Li, it was shown by H. Kwon in \cite{Kwon2020} that the threshold $\til{\mu}=-1/2$ is in fact sharp in the endpoint case of the 2D Euler equation by demonstrating strong ill-posedness in $H^{1}(\RR^2)$ of the corresponding initial value problem for all $-1/2\leq\til{\mu}<0$. The recent ill-posedness result \cite{JeongKim2021} remarks that ill-posedness for the corresponding logarithmically regularized inviscid equation in the range $\be\in(0,1]$ can also be established, leaving the range $\be\in(1,2)$ an open consideration, although presumably, the recent work of \cite{CordobaZoroa-Martinez2022} can be adopted to establish ill-posedness in this regime as well. In light of these results and those in \cite{ChaeJeongNaOh2023a, ChaeJeongOh2023b}, it would be interesting to establish ill-posedness for the general class of models considered in this article and in the generality of the borderline regularity framework developed here.
In this paper, we study an alternative mechanism via mild dissipation for recovering well-posedness for the full range of \eqref{eq:gsqg}, $\be\in[0,2]$, in the borderline Sobolev spaces $H^{1+\be}(\RR^2)$. Addition of mild dissipation effectively allows us to consider constitutive laws that balance the dissipation with logarithmically more singular velocities. We moreover observe that these mechanisms may also be correspondingly balanced by the functional setting as well. With all such modifications in place, we develop a unified theory of local well-posedness in the corresponding borderline Sobolev spaces $H_\om^{1+\be}(\RR^2)$ of \eqref{eq:dgsqg}, for all $\be\in[0,2]$, where $\om$ belongs an appropriate set of weights. From this point of view, our results touch upon those of M. Vishik \cite{Vishik1999}, where Yudovich's classical uniqueness theorem was extended to a borderline Besov space setting where growth in the norm is allowed, but in a controlled way weighted by frequency; the role of the weights $\om$ play a similar role for us, albeit in an $L^2$ setting.
As in \cite{ChaeWu2012, JollyKumarMartinez2020b}, we identify a putatively minimal condition for local well-posedness that jointly involves the dissipation, modification of the constitutive law, as well as the weights $\om$ that modify the functional setting. The class of multipliers considered here are sufficiently broad to accommodate powers of logarithms or iterated logarithms. One particular special case that is covered is $m(D)=(\ln(I-\Delta))^{\mu}$ and $p(D)=(\ln(I-\Delta))^{\til{\mu}}$, where we are able to establish local well-posedness in $H^{1+\be}(\RR^2)$, for all $\be\in[0,2]$, provided that $\mu>\til{\mu}+1/2$, where $\til{\mu}>-1/2$. Our result therefore interpolates between the well-posedness results in \cite{ChaeWu2012, JollyKumarMartinez2020b} up to the sharp thresholds where the equation experiences ill-posedness \cite{JeongKim2021, Kwon2020}, at least in the range $\be\in[0,1]$.
We additionally establish a \textit{mild instantaneous smoothing effect} for \eqref{eq:dgsqg} that derives from the regularizing mechanism of the corresponding linear equation.
Indeed, let $\nu(D)$ denote a multiplier operator that belongs to the class $\mathscr{M}_{S}(m)$ defined in \eqref{def:M:m}.
Then we establish the following result for \eqref{eq:dgsqg}:
\[
e^{\lam t \nu(D)}\tht(t,\cdotp)\in H^{\s},\quad\text{whenever}\quad \tht_0\in H^\s,
\]
for all $\lam \in (0,1)$ sufficiently small. In particular, with $m(D)=(\ln(I-\Delta))^{\mu}$ where $\mu \ge 1$, this implies
\[
\tht(t,\cdotp)\in H^{\s+\lam t},\quad\text{whenever}\quad \tht_0\in H^\s.
\]
This observation along with a suitable $L^\infty$ maximum principle for the dissipative term allows us to obtain global-in-time existence in the case of $\be=0$ (Euler endpoint). This complements the result in H. Dong and D. Li \cite{DongLi2015}, where global well-posedness in the borderline space $\Hdot^1\cap\Hdot^{-1}$ was also established for the corresponding logarithmic inviscid regularization for a sufficiently large power of the logarithm, as well as M. Vishik \cite{Vishik1998}, where global existence for the Euler endpoint was established, but by varying the functional setting, specifically considering an $L^p$--based borderline Besov space. In addition to providing an alternative mechanism for global existence, our framework explores the simultaneous effect of regularization or ``singularization'' in the constitutive law, functional setting, and dissipation.
When \eqref{eq:gsqg} is dissipatively perturbed by the strong dissipation, ${\Lam}^\kap$, where $\kap\in(0,2)$, the smoothing effect conferred on its solutions is much stronger than the one conferred by mild dissipation. In a series of works \cite{DongLi2008, DongLi2010, Biswas2014, BiswasMartinezSilva2015, Li2021, JollyKumarMartinez2020a}, it was shown that the unique solution of the strongly dissipative gSQG equations emanating from initial data belonging to the scaling-critical Sobolev space $H^{1+\be-\kap}(\RR^2)$, instantaneously enters a class of smooth functions, known as the Gevrey class, which characterizes a scale of regularity between the $C^\infty$ class of smooth functions and the $C^\om$ class of real-analytic functions. Such results strengthened the existing well-posedness results in critical regularity settings \cite{ChaeLee2003, Miura2006, ChenMiaoZhang2007}. In the setting of euclidean space, the Gevrey class enforces exponential decay of the frequency spectrum of its members at some rate. For solutions of the strongly dissipative gSQG equations, the rate of exponential decay is shown to grow in time. The series of works mentioned above ultimately culminated in the recent work \cite{JollyKumarMartinez2020a}, where this \textit{strong smoothing} phenomenon was established for the range $\be\in(1,2]$, with the endpoint $\be=2$ accordingly modified by a logarithm (see discussion following \eqref{eq:gsqg}), thus completing this line of investigation for the strongly dissipative gSQG equation in the setting of the scaling-critical Sobolev spaces. In contrast, the smoothing effect observed for the mildly dissipative equation given by \eqref{eq:dgsqg} is categorically weaker than the one observed for its strongly dissipative counterpart. In establishing this form of smoothing, we expand upon the celebrated Gevrey-norm technique of C. Foias and R. Temam \cite{FoiasTemam1989} to accommodate linear operators of ones considered here.
It should be emphasized that the setting of \eqref{eq:gsqg} when $\be\in(1,2]$ and $\be>\kap+1$ exhibits a \textit{strongly quasilinear} structure due to the fact that the velocity, treated as a coefficient of the gradient, is of higher order than the linear dissipative term; from this point of view, the case of mild dissipation is significantly supercritical. Although the quasilinearity in \eqref{eq:dgsqg} is much stronger than the fractionally dissipative counterpart of \eqref{eq:gsqg}, we point out the lack of a genuine scaling-critical space effectively places our setting in a subcritical setting, albeit barely so. The quasilinearity is therefore a source of difficulty that forces us to exploit the nuanced commutator structure of \eqref{eq:dgsqg}, but is ultimately overcome by jointly exploiting the \textit{barely subcritical} functional setting of the problem. A notable related work that may be viewed as somewhat ``dual'' to the considerations here is that of O. Lazar and L. Xue \cite{LazarXue2019}, where global regularity of solutions is established for a strongly dissipative counterpart of \eqref{eq:gsqg} with constitutive laws that are logarithmically modified (via $p(D)$ as in \eqref{eq:dgsqg}) in a similar fashion to framework developed in the present article, but for the purpose of considering a \textit{slightly supercritical} scenario, where the dissipation is chosen in a particular relation to the constitutive law; in this scenario, global regularity is known without the logarithmic modification (\cite{ConstantinIyerWu2008, MiaoXue2011}) of the constitutive law. The present article, in contrast, identifies a general condition between $m(D)$ and $p(D)$ that guarantees local well-posedness at borderline regularity and the smoothing effects naturally induced by $m(D)$. Moreover, this general condition specifically accommodates logarithmic-type mulipliers that characterize a class of dissipative operators that are strictly weaker than the strongly dissipative operators previously considered in the literature. The reader is referred to \cref{rmk:lazar} for further discussion.
Lastly, central to our framework is a model equation that we refer to as the \textit{protean system}. The protean system is linear system upon which the proof of well-posedness is found to turn around entirely. This system takes its inspiration from the role played by the linear transport equation in the study of the 2D Euler and SQG equations. Indeed, from our perspective, the linear transport equation can be realized as the protean system of the 2D Euler equation. However, when considering the gSQG family beyond the SQG point, i.e., $1<\be\leq2$, the paradigm of the transport equation breaks down. This phenomenon was observed in the most recent work of the authors with M.S. Jolly \cite{JollyKumarMartinez2020b}, where continuity of the solution operator required additional modifications to the transport equation. This modification was also required for establishing local well-posedness for large data in the critical Sobolev space setting for the supercritically fractional dissipative gSQG equation by the same authors in \cite{JollyKumarMartinez2020a}. We ultimately realize these key modifications in the form of the protean system proposed here (see \cref{sect:mod:flux}). The present article therefore represents a conclusion to a trilogy works starting from \cite{JollyKumarMartinez2020a, JollyKumarMartinez2020b} that forms a culmination in the understanding gained therein about the gSQG family in borderline regularity settings.
\subsection{The Protean System}\label{sect:mod:flux}
Our main apriori analysis will be centered around a model equation from which all the estimates relating to the well-posedness and instantaneous smoothing property for the original system \eqref{eq:dgsqg} will effectively follow as a special case; for this reason, we refer to the model equation as the \textit{protean system}. For \eqref{eq:dgsqg}, the protean system is given by a \textit{linear conservation law}, which becomes nonlinear upon appropriately substituting for the solution in its flux. The need for this structure arises from the more singular nature of the constitutive law of the equations. Indeed, in contrast with the regime $\be\in(1,2]$, the protean system reduces simply to a \textit{linear transport equation} (see \eqref{eq:mod:claw} below). We first introduce the protean system here and detail the ways in which it is used to demonstrate well-posedness and smoothing for \eqref{eq:dgsqg}. We develop the apriori analysis for the protean system and the well-posedness of its initial value problem in \cref{sect:apriori}.
Suppose that $q=q(t,x)$, $G=G(t,x)$ are sufficiently smooth functions. For $\be\in[0,2]$, let $m(D)$ and $p(D)$ be Fourier multiplier operators, i.e.,
\[
(\mathcal{F}m(D)\ph)(\xi)=m(\xi)(\mathcal{F}\ph)(\xi),\qquad(\mathcal{F}p(D)\ph)(\xi)=p(\xi)(\mathcal{F}\ph)(\xi),\quad m(\xi), p(\xi)\geq0.
\]
Then define a Fourier multiplier $a(D)$ by
\begin{align}\label{def:aD}
a(D):={\Lam}^{\be-2}p(D).
\end{align}
The precise restrictions on $m(D)$ and $p(D)$ will be specified later. Given $\tht_0=\tht_0(x)$, we will consider the following initial value problem for a linear conservation law:
\begin{align}\label{eq:mod:claw}
{\partial_{t}\tht+ \Div F_q(\tht)=-m(D)\tht}+G,\quad\tht(0,x)=\tht_0(x),
\end{align}
where the flux is given by
\begin{align}\label{def:mod:flux}
F_q(\tht)=
\begin{cases}
(\nabla^{\perp}a(D)q) \tht,\quad &\text{if}\quad \be \in [0,1],\\
(\nabla^{\perp}a(D)q) \tht+a(D)(({\nabla}^\perp\tht)q), \quad &\text{if}\quad \be \in (1,2].
\end{cases}
\end{align}
Observe that the flux satisfies the following identity:
\begin{align}\label{eq:cancel}
\Div F_{-\tht}(\tht)=-({\nabla}^\perp a(D)\tht)\cdotp{\nabla}\tht=v\cdotp{\nabla}\tht.
\end{align}
where
\begin{align}\label{def:v}
v=v(q):=-\nabla^{\perp}a(D)q.
\end{align}
Observe also that ${\nabla}\cdotp v=0$, for all $q$ sufficiently smooth, which yields the identity
\begin{align}\label{eq:v:cancel}
\lb v\cdotp\nabla h,h\rb=0,
\end{align}
for any sufficiently smooth functions $h$. Hence, one recovers \eqref{eq:dgsqg} from \eqref{eq:mod:claw} when $q=-\tht$, and $G\equiv0$. In particular, one may obtain apriori estimates for \eqref{eq:dgsqg} by obtaining them for \eqref{eq:mod:claw} and simply specializing to the case $q=-\tht$.
For uniqueness, observe that if $\tht^{(1)}, \tht^{(2)}$ are distinct solutions, then the difference $\Tht:=\tht^{(1)}-\tht^{(2)}$ is governed by
\begin{align}\label{eq:difference}
\partial_{t}\Tht+m(D) \Tht+\Div F_{-\tht^{(1)}}(\Tht)=\Div F_{\Tht}(\tht^{(2)}),\quad\Tht(0,x)=\Tht_0(x)
\end{align}
with $q=-\tht^{(1)}$ and $G=\Div F_{\Tht}(\tht^{(2)})$. The related issue of continuity with respect to initial data is more delicate. Indeed, assessing stability of \eqref{eq:mod:claw} in the borderline space $H^{1+\be}$ is not possible through a direct analysis of \eqref{eq:difference:intro} due to the loss of derivatives experienced through the flux term in the regime $\be\in(1,2]$. A direct analysis is ultimately limited to establishing continuity in the \textit{weaker topology} of $H^\be$. Bootstrapping from the weaker topology to the topology of the phase space can effectively be carried out by making use of a splitting technique of Kato that analyzes the \textit{gradient} of difference in the weaker topology. However, while this scheme is well-adapted to the transport equation, it is more delicate to accommodate the case of \eqref{eq:mod:claw} when $\be\in(1,2]$.
To see this, we consider a decomposition $\nabla\tht=\vs+\ze$ and aim to show that for a sequence $\tht_0^n$ converging to $\tht_0$, that the corresponding solution $\nabla\tht^n=\vs^n+\ze^n$ converges to $\nabla\tht$ in $H^{1+\be}$ by showing that $\vs^n, \ze^n$ converge to $\vs,\ze$ in the weaker topology $H^\be$. We specifically assume that $\vs^n,\ze^n,\vs,\ze$ satisfy
\begin{align}\label{eq:vs:protean}
\partial_{t}\vs^{n}+m(D)\vs^{n}+ \Div F_{-\tht^{n}}(\vs^{n})&=\Div F_{\nabla\tht}(\tht),\quad\vs^{n}(0,x)=\nabla\tht_0(x),\\
\partial_{t}\ze^{n}+m(D)\zeta^{n}+ \Div F_{-\tht^n}(\ze^{n})&=\Div F_{\nabla\tht^n}(\tht^{n})-\Div F_{\nabla\tht}(\tht),\quad \ze^{n}(0,x)=\nabla\tht^{n}_0(x)-\nabla\tht_{0}(x),\label{eq:ze:protean}
\end{align}
where we identify $\tht_0,\vs,\ze$ with $\tht_0^\infty, \vs^\infty, \ze^\infty$. Now observe that \eqref{eq:vs:protean} has the structure of \eqref{eq:mod:claw} upon making the replacements $q\mapsto -\tht^n$ and $G\mapsto \Div F_{\nabla\tht}(\tht)$. Similarly, \eqref{eq:ze:protean} has the structure of \eqref{eq:mod:claw} upon making the replacement $q\mapsto -\tht^n$ and $G\mapsto \Div F_{\nabla\tht^n}(\tht^{n})-\Div F_{\nabla\tht}(\tht)$. In particular, the systems corresponding to the differences $\vs^n-\vs$ and $\ze^n-\ze$ satisfy systems analogous to \eqref{eq:difference:intro}, which \textit{once again} possess the structure of the protean system and ultimately allow estimates to close in the desired manner.
In order to leverage the estimates obtained for \eqref{eq:mod:claw} and establish existence and smoothing of solutions to \eqref{eq:dgsqg}, one must identify a suitable approximation of \eqref{eq:dgsqg} that also satisfies the desired estimates, but uniformly in the approximation. Under the assumptions we make for $m(D)$ and $p(D)$, an artificial viscosity approximation scheme will be sufficient; this will be self-evident from the apriori analysis we perform below in \cref{sect:apriori}. Of course, once existence has been established, the proofs of uniqueness and continuity with respect to initial conditions can be performed on the equations rigorously.
\section{Mathematical Preliminaries}\label{section:notation:preliminaries}
Let $d\geq1$ and denote by $\mathscr{S}(\RR^d)$ the space of Schwartz class functions on $\RR^d$ and by $\mathscr{S}'(\RR^d)$ the space of tempered distributions. We will denote by $\hat{f}$ or $\mathcal{F}(f)$, the Fourier transform of a tempered distribution $f$, defined as
\[
\hat{f}(\xi):=\int_{\RR^d}e^{-2\pi i x\cdot \xi}f(x)dx.
\]
Recall that $\mathcal{F}$ is an isometry on $L^2$, i.e.,
\begin{align*}
\lb \hat{f},\hat{g}\rb= \lb f, g\rb:=\int_{\RR^d}f(x)\overline{g(x)}dx.
\end{align*}
The fractional laplacian operators, denoted ${\Lam}^\s$, $\s\in\RR$, are defined in terms of the Fourier transform by
\begin{align}\label{def:Lam}
\mathcal{F}({\Lam}^\s f)(\xi)=|\xi|^\s\mathcal{F}(f).
\end{align}
For $d\geq1$ and $\s\in\RR$, the homogeneous and the inhomogeneous Sobolev spaces are defined as
\begin{align}
&\Hdot^\s(\RR^d):=\left\{f\in \mathscr{S}'(\RR^d):\hat{f}\in L^1_{loc}(\RR^d),\quad \Sob{f}{\Hdot^\s}:=\Sob{{\Lam}^\s f}{L^2}<\infty\right\},\label{def:hom:Sob:norm}\\
&H^\s(\RR^d):=\left\{f\in \mathscr{S}'(\RR^d):\hat{f}\in L^1_{loc}(\RR^d),\quad \Sob{f}{H^\s}:=\Sob{(I-\De)^{\s/2} f}{L^2}<\infty\right\}.\label{def:inhom:Sob:norm}
\end{align}
Recall that $H^\s(\RR^d)$ is a Hilbert space for all $\s\in\RR$, whereas $\Hdot^\s(\RR^d)$ is a Hilbert space if and only if $\s<d/2$. The inhomogeneous spaces are nested $H^{\s'}(\RR^d)\subset H^\s(\RR^d)$, whenever $\s'>\s$, and moreover the embedding is compact over compact sets. The homogeneous spaces on the other hand are in general only related by the following interpolation inequality: Given $\s_1\leq\s\leq \s_2$, we have
\begin{align}\label{est:interpolation}
\nrm{f}_{\Hdot^{\s}(\RR^d)}\leq \Sob{f}{\Hdot^{\s_1}(\RR^d)}^{\frac{\s_2-\s}{\s_2-\s_1}}\Sob{f}{\Hdot^{\s_2}(\RR^d)}^{\frac{\s-\s_1}{\s_2-\s_1}}.
\end{align}
Lastly, observe that for each $\s\geq0$, there exists $C>0$ (depending also on $d$) such that
\begin{align}\label{eq:Sob:equiv}
C^{-1}\Sob{f}{H^\s(\RR^d)}\leq \Sob{f}{\Hdot^\s(\RR^d)\cap L^2(\RR^d)}\leq C\Sob{f}{H^\s(\RR^d)},
\end{align}
where we have adopted the convention
\[
\Sob{f}{X\cap Y}^2:=\Sob{f}{X}^2+\Sob{f}{Y}^2,
\]
where $X, Y$ are (semi)normed vector spaces.
\subsubsection{Frequency-weighted Sobolev spaces} Due to the presence of a logarithmic order dissipation and a logarithmic modification in the constitutive law in \eqref{eq:dgsqg}, it will be natural to consider function spaces that accommodate logarithmic decay at infinity in frequency; they have been referred to in the literature as the \textit{log-Sobolev} spaces. Spaces such as these and their natural generalizations have been an object of study in recent years. We refer the reader to \cite{CobosDominguez2015, ColombiniDelSantoFanelliMetivier2015} and the references therein for a rigorous development of these spaces. For our purposes, it will be convenient to introduce these spaces in a greater generality, then identify precise conditions on the weights afterwards.
Let $\om: [0,\infty)\mapsto(0,\infty)$. We denote its associated multiplier operator by $\om(D)$; note that $\om(D)=\om(|D|)$, i.e., the symbol of $\om(D)$ is a radial function. Recall that $\om(D)$ is defined by the relation $\mathcal{F}(\om(D)f)(\xi)=\om(|\xi|)(\mathcal{F}f)(\xi)$. Then for $\s \in \mathbb{R}$, we define the $\om$-weighted Sobolev spaces on $\RR^{d}$ by
\begin{align}
\Hdot^{\s}_{\om}(\RR^d)&:=\left\{f\in \mathscr{S}'(\RR^d):\hat{f}\in L^1_{loc}(\RR^d),\quad \Sob{f}{\Hdot^{\s}_{\om}}:=\Sob{\om(D)f}{\Hdot^\s}<\infty\right\},\label{def:hom:Sob:norm:log}\\
H^{\s}_{\om}(\RR^d)&:=\left\{f\in \mathscr{S}'(\RR^d):\hat{f}\in L^1_{loc}(\RR^d),\quad \Sob{f}{H^{\s}_{\om}}:=\Sob{\om(D)f}{H^\s}<\infty\right\},\label{def:inhom:Sob:norm:log}
\end{align}
In this setting, an interpolation inequality analogous to \eqref{est:interpolation} also holds: for all $\gam_1\leq \gam\leq \gam_2$, we have
\begin{align}\label{est:interpolation:weighted}
\Sob{\om(D)^{\gam}f}{L^2(\RR^d)}\leq \Sob{\om(D)^{\gam_1}f}{L^2(\RR^d)}^{\frac{\gam_2-\gam}{\gam_2-\gam_1}}\Sob{\om(D)^{\gam_2}f}{L^2(\RR^d)}^{\frac{\gam-\gam_1}{\gam_2-\gam_1}}.
\end{align}
The proof is omitted since the same argument for proving \eqref{est:interpolation} can be applied to prove \eqref{est:interpolation:weighted}.
\begin{Rmk}
We point out that for our main results (see \cref{sect:main:results}), we will restrict the class of weights, $\om$, that we consider by imposing certain restrictions on them that will be useful for the analysis (see \eqref{def:MM} in \cref{sect:op:classes}). These restrictions will imply the following growth condition: there exists constants $C, N>0$ such that
\begin{align}\notag
\om(r)\le C(\ln(e+r))^{N},
\end{align}
for all $r\geq0$. For example, we may consider $\om(r)=(\ln(e+r^2))^{\rho_1}$, $\rho_1\in \mathbb{R}$. In this particular case, we will denote the corresponding inhomogeneous and homogeneous log-Sobolev spaces by $H^{\s}_\om=H^{\s, \rho_1}$ and $\Hdot^{\s}_\om=\Hdot^{\s,\rho_1}$. Similarly, for $\om(r)=(\ln(e+r^2))^{\rho_1}(\ln(e+\ln(1+r^2))^{\rho_2}$, $\rho_1, \rho_2\in \mathbb{R}$, we will denote the corresponding inhomogeneous and homogeneous log-Sobolev spaces by $H^{\s, \rho_1, \rho_2}$ and $\Hdot^{\s,\rho_1, \rho_2}$, and so on.
\end{Rmk}
\subsection{Littlewood-Paley Decomposition}\label{sect:lp:decomposition}
In this section, we provide a brief review of the Littlewood-Paley decomposition and refer the reader to \cite{BahouriCheminDanchinBook2011, CheminBook1995} for additional details. First, we introduce the space
\begin{align*}
\mathscr{Q}(\RR^d):=\left\{f\in \mathscr{S}(\RR^d): \int_{\RR^d}f(x)x^{\tau}\, dx=0, \quad \abs{\tau}=0,1,2,\cdots \right\}.
\end{align*}
The topological dual space, $\mathscr{Q}'(\RR^d)$, of $\mathscr{Q}(\RR^d)$, can be identified with the space of tempered distributions modulo polynomials, that is, as
\begin{align*}
\mathscr{Q}'(\RR^d)\cong\mathscr{S}'(\RR^d)/\mathscr{P}(\RR^d).
\end{align*}
{where $\mathscr{P}(\RR^d)$ denotes the vector space of polynomials}.
Given $d\geq1$, fixed, let ${\Bcal}(r)$ denote the open ball in $\RR^d$ of radius $r$ centered at the origin and ${\Acal}(r_{1},r_{2})$ denote the open annulus in $\RR^d$ with inner and outer radii $r_{1}$ and $r_{2}$, and centered at the origin. One can construct two non-negative radial functions $\chi,\phi\in\mathscr{S}(\RR^d)$ with $\supp\chi\subset{\Bcal}(1)$ and $\supp\phi\subset{\Acal}(2^{-1},2)$ such that the following properties are satisfied. For $\chi_j(\xi):=\chi(2^{-j}\xi)$ and $\phi_j(\xi):=\phi(2^{-j}\xi)$,
\begin{align}\label{eq:partition}
\begin{cases}
&\sum_{j\in\ZZ}\phi_j(\xi)=1,\quad \forall\,\xi\in\RR^d\setminus\{\mathbf{0}\},\\
&\chi+\sum_{j\geq0}\phi_j\equiv 1,\\
&\supp\phi_i\cap\supp\phi_j=\varnothing,\quad |i-j|\geq2,\\
&\supp\phi_i\cap\supp\chi =\varnothing,\quad i\geq1.
\end{cases}
\end{align}
It will be convenient to define the following notation:
\begin{align}\label{def:ann:ball}
{\Acal}_{j}= {\Acal}(2^{j-1},2^{j+1}),\quad{\Acal}_{\ell,k}= {\Acal}(2^{\ell},2^{k}),\quad {\Bcal}_j={\Bcal}(2^j),
\end{align}
so that, in particular, ${\Acal}_j={\Acal}_{j-1,j+1}$. With this notation, observe that
\begin{align}\label{eq:rewrite:supp}
\supp\phi_j\subset{\Acal}_j,\quad \supp\chi_j\subset{\Bcal}_j.
\end{align}
We denote the homogeneous Littlewood-Paley dyadic blocks by ${\lpj}$ and $S_{j}$. These are both defined in terms of its Fourier transform by
\begin{align}\notag
\mathcal{F}({\lpj}f)=\phi_{j}\mathcal{F}(f), \quad \mathcal{F}(S_{j}f)=\chi_{j}\mathcal{F}(f).
\end{align}
One then has the following identity
\begin{align*}
f&=S_if+\sum_{j\geq i}\lpj f,\quad i\in\ZZ,\quad f\in\mathscr{S}'(\RR^d).
\end{align*}
In fact, when restricted to $\mathscr{Q}'$, one has
\begin{align*}
f&=\sum_{j\in\ZZ}\lpj f,\quad f\in\mathscr{Q}'(\RR^d).
\end{align*}
Then a useful characterization of Sobolev norms is as follows: given $\s\in\RR$, there exists $C>0$ such that
\begin{align}\label{eq:Sob:Bes}
C^{-1}\Sob{f}{\Hdot^{\s}(\RR^d)}\leq \left(\sum_{j\in \mathbb{Z}}2^{2\s j}\Sob{\triangle_{j}f}{L^2(\RR^d)}^{2}\right)^{1/2}\le C\Sob{f}{\Hdot^{\s}(\RR^d)},
\end{align}
Note that this equivalence also holds when the support of $\lpj$ is rescaled by any fixed number (see \cite{BahouriCheminDanchinBook2011}). We refer to the intermediate quantity as the homogenous Besov norm $\Sob{f}{\dot{B}^{\s}_{2,2}(\RR^d)}$; whenever $\s\geq0$, we define the corresponding \textit{inhomogeneous} Besov norm by
\[
\Sob{f}{B^{\s}_{2,2}(\RR^d)}^2:=\Sob{f}{\dot{B}^{\s}_{2,2}(\RR^d)\cap L^2(\RR^d)}^2=\Sob{f}{\dot{B}^{\s}_{2,2}(\RR^d)}^2+\Sob{f}{L^2(\RR^d)}^2.
\]
With this notation, we shall also make use of the following notation for the Besov-space analog of the frequency-weighted Sobolev spaces:
\begin{align}\label{def:Bes:om}
\Sob{f}{\dot{B}^{\s}_{\om}(\RR^d)}=\Sob{\om(D)f}{\dot{B}^{\s}_{2,2}(\RR^d)}, \quad \Sob{f}{{B^\s_\om}(\RR^d)}=\Sob{\om(D)f}{\dot{B}^{\s}_{2,2}(\RR^d)\cap L^2(\RR^d)},
\end{align}
where the latter quantity denotes the corresponding inhomogeneous Besov norm whenever $\s\geq0$.
The relation between the Littlewood-Paley blocks and the fractional laplacian is captured by the following Bernstein-type inequalities.
\begin{Lem}[Bernstein inequalities]\label{lem:Bernstein}
Let $\s\in\RR$ and $1\le p \le q\le \infty$. Then there exists $C>0$ such that
\begin{align*}
C^{-1}2^{\s j}\nrm{{\lpj}f}_{L^q(\RR^d)}\le \nrm{{\Lam}^{\s}{\lpj}f}_{L^q(\RR^d)}\le C 2^{\s j+dj(\frac{1}{p}-\frac{1}{q})}\nrm{{\lpj}f}_{L^p(\RR^d)},
\end{align*}
for all $j\in\ZZ$ and $f\in\mathscr{S}'(\RR^d)$.
\end{Lem}
Let us recall the following classical product estimate in homogenous Sobolev spaces (see \cite{BahouriCheminDanchinBook2011,RunstSickel1996}).
\begin{Lem}\label{lem:Sobolev}
Suppose that $\s_1,\s_2 \in (-d/2,d/2)$ and $\s_1+\s_2>0$. Let $f\in \Hdot^{\s_1}(\RR^d)$ and $g\in \Hdot^{\s_2}(\RR^d)$. Then there exists $C>0$ such that
\begin{align*}
\Sob{fg}{\Hdot^{\s_1+\s_2-\frac{d}{2}}(\RR^d)}\le C\Sob{f}{\Hdot^{\s_1}(\RR^d)}\Sob{g}{\Hdot^{\s_2}(\RR^d)}.
\end{align*}
\end{Lem}
\begin{Rmk}\label{rmk:convention} For the remainder of the manuscript, we adopt the convention that whenever $d=2$, we will denote $H^\s(\RR^2), {H}^{\s,\rho}(\RR^2)$ simply as $H^\s, {H}^{\s,\rho}$, and similarly for their homogeneous counterparts and related spaces such as $\mathscr{S}(\RR^2)$, $\mathscr{S}'(\RR^2)$, etc. Whenever results hold for $d\geq1$, we will explicitly write $H^\s(\RR^d)$, $\Hdot^{\s,\rho}(\RR^d)$, $\mathscr{S}(\RR^d)$, etc., in their statements. However, we will always suppress the domain when performing estimates. Lastly, whenever the parameter, $d$, appears, it is understood that $d\geq1$ unless stated otherwise.
\end{Rmk}
\subsection{Multiplier Classes
}\label{sect:op:classes}In this section, we will define general classes of Fourier multiplier operators that characterize the regularity and growth properties of the multiplier operators $m(D)$, $p(D)$, $\om(D)$ and $\nu(D)$ that feature in the model \eqref{eq:dgsqg} and our main results. A central preoccupation of this article is in identifying the precise interrelation between these operators in establishing local well-posedness (in the sense of Hadamard). Indeed, we deconstruct the various effects arising from the dissipative operator, $m(D)$, and the operator $p(D)$ which multiplicatively modifies the constitutive relation between the advecting velocity field and the transported scalar in such a way that provides either regularization or de-regularization. To carry out this ``deconstructive analysis," we introduce the operators $\om(D)$ and $\nu(D)$. On an intuitive level, the role of these operators can be described as follows:
\begin{align}
\om(D)&\sim\ \text{logarithmically adjusts the regularity of the \textit{phase space}}\notag\\
p(D)&\sim\ \text{represents the logarithmic modification of the \textit{constitutive law}}\notag\\
m(D)&\sim\ \text{represents the \textit{dissipation mechanism}}\notag\\
\nu(D)&\sim\ \text{captures the \textit{smoothing mechanism} associated to $m(D)$}\notag.
\end{align}
{Ultimately, the multiplier $\om(D)$ enables an additional degree of flexibility for the local existence theory. In particular, it allows us to accommodate additional logarithmic losses of derivatives in the initial data without leaving the setting of borderline regularity; one of the main observations represented by our results is that this effect can be balanced by appropriately adjusting the regularizing or singularizing effects of the constitutive law or dissispation.} On the other hand, the multiplier $\nu(D)$ enables us to quantify the regularizing mechanism of the dissipation operator. The precise inter-relation between these operators that admit well-posedness of the corresponding active scalar transport system are stated in our main theorems in \cref{sect:main:results}.
\subsubsection{Frequency weights associated to the regularity of the phase space} Let us first introduce the following properties:
\begin{description}
\item[\namedlabel{item:O1}{(O1)}]
$\om \in C^1([0,\infty))$ is positive, and satisfies $\om'\geq0$.
\item[\namedlabel{item:O2}{(O2)}]
There exists $C>0$ such that ${r\om'(r)}\le {C{\om(r)}}$, for all $r\ge 0$.
\item[\namedlabel{item:O3}{(O3)}] There exists $C>0$ such that
\begin{align}\notag
\om(r_1r_2)\leq C(\om(r_1)+\om(r_2)),\quad\text{for all}\ r_1,r_2\geq0.
\end{align}
\end{description}
From \ref{item:O1} and \ref{item:O2}, we deduce the following property: for any integer $k_1,k_2>0$, there exists $C>0$ such that for all $j\in\ZZ$
\begin{align}\label{est:omega}
\begin{split}
C^{-1}\om(2^j)&\le \om(r)\le C\om(2^j).
\end{split}
\end{align}
for all $r\in[2^{j-k_1},2^{j+k_2}]$.
Now, let us see how \ref{item:O3} limits the growth at infinity provided that \ref{item:O1}, \ref{item:O2} are also satisfied. Indeed, observe that \ref{item:O2} implies
\begin{align}\label{eq:om:upper}
\om(r)\le \om(1)r^{C},\quad \text{for all}\ r\ge 1,
\end{align}
where $C$ is the same constant from \ref{item:O2}. In particular, \ref{item:O2} limits growth at infinity to be at most algebraic. However, upon applying \ref{item:O3} with $r_1=r_2=\sqrt{r}$ and iterating, we obtain
\begin{align}\notag
\om(r)\leq 2C\om(r^{1/2})\leq(2C)^2\om(r^{1/4})\leq\dots\leq (2C)^n\om(r^{1/2^n})\leq (2C)^n\om(1)r^{C/n},
\end{align}
for all $n>0$, where we have applied \eqref{eq:om:upper} in obtaining the final inequality. Hence, it follows that for any $\eps>0$, there exist $c_{\eps}>0$ such that
\begin{align}\label{eq:omega:eps}
\om(r)\le c_{\eps}r^{\eps}\quad\text{for all}\ r\geq1.
\end{align}
Moreover
\begin{align}\label{eq:omega:eps:bdd}
\om(r)\leq \left(\sup_{0\leq r\leq 1}\om(r)\right)+c_\eps r^{\eps}\leq C_\eps(1+r)^\eps,
\end{align}
for all $r\geq0$, for some $C_\eps>0$.
Finally, we observe that if $\om_1,\om_2$ satisfies \ref{item:O1}, \ref{item:O2}, and \ref{item:O3}, then the pointwise product $\om_1\om_2$ also satisfies \ref{item:O1}, \ref{item:O2}, and \ref{item:O3}.
Let us then define the class $\mathscr{M}_W$ by
\begin{align}\label{def:MM}
\mathscr{M}_W:=\left\{\frac{\om_a(D)}{\om_b(D)}:\om_a,\om_b\in C^1([0,\infty))\ \text{satisfy \ref{item:O1}, \ref{item:O2},\ \ref{item:O3}}\right\}.
\end{align}
It follows by an application of Plancherel's theorem and \eqref{est:omega} that whenever $\om\in\mathscr{M}_W$, there exists $c,C>0$ such that
\begin{align}\label{eq:bernstein:om}
c\om(2^j)\nrm{{\lpj}f}_{L^2(\RR^d)}\le \nrm{\om(D){\lpj}f}_{L^2(\RR^d)}\le C \om(2^j)\nrm{{\lpj}f}_{L^2(\RR^d)}.
\end{align}
We will make use of \eqref{eq:bernstein:om} in a crucial way in the product and commutator estimates. Moreover, observe that \eqref{eq:omega:eps} holds for all $\om\in\mathscr{M}_W$. Note that a prototypical example of $\om\in\mathscr{M}_W$ is given by $\om(r)=\left(\ln(e+r^2)\right)^p$, where $p$ is a nonzero real number.
\subsubsection{Multipliers associated to the constitutive law}
{For the multipliers, $p(D)$, associated to the constitutive law we will first introduce a slightly generalized class $\tilde{\mathscr{M}}_W$, which we will then supplement with a property that limits the rate of decay at infinity. We recall that this property is what ultimately prevents the velocity from automatically being Lipschitz}. Let
\begin{align}\label{def:MMt}
\tilde{\mathscr{M}}_W:=\left\{\frac{p_a(D)}{p_b(D)}:p_a,p_b\in C^1([0,\infty))\ \text{satisfy \ref{item:O1}, \ref{item:O2}}\right\}.
\end{align}
We point out that \eqref{est:omega} and \eqref{eq:bernstein:om} still hold for any $p\in\tilde{\mathscr{M}}_W$ since they only rely on the properties \ref{item:O1}, \ref{item:O2}.
It also follows from \ref{item:O1} and \ref{item:O2} that there exists $C>0$ such that
\begin{align}\label{eq:p:triangle:M_1}
p_a(r_1+r_2)\leq C(p_a(r_1)+p_a(r_2)),\quad \text{for all}\ r_1,r_2\ge 0.
\end{align}
Moreover, from \eqref{eq:p:triangle:M_1}, we may make additional use of \ref{item:O1} to show that
\begin{align}\label{def:gen:tri:prod}
p(|\xi|+|\eta|)\leq C\left(p_a(|\xi|)+p_a(|\eta|)\right),
\end{align}
for some constant $C>0$, which depends on the value of $p_b(0)$.
Finally, we incorporate the decay restriction and define
\begin{align}\label{def:MC}
\mathscr{M}_C:=\left\{p(D)\in {\mathscr{M}}_W : \int_{1}^{\infty}\frac{p^{2}(r)}{r}dr=\infty\right\},
\end{align}
We see that \eqref{def:MC} limits the symbol of the multiplier to decay at most logarithmically to a certain degree at infinity. A prototypical example is given by $p(r)=\left(\ln(e+r^2)\right)^{-1/2}$.
\subsubsection{Multipliers associated to the dissipation and its smoothing effect} The next class of multipliers we define are introduced to capture the smoothing effects of the dissipation represented by the multiplier operator $m(D)$. These effects are typically captured by an operator of the form
\begin{align}\label{def:E}
E_\nu^{\lam}f=e^{\lam \nu(D)}f,
\end{align}
where $\nu(D)$ is a radial multiplier operator, i.e., $\nu(D)=\nu(|D|)$. We will refer to \eqref{def:E} as the \textit{smoothing operator induced by} $\nu(D)$.
Define a class of dissipation operators, denoted by $\mathscr{M}_D$, as
\begin{align}\label{def:MD}
\mathscr{M}_D:=\left\{m(D): I+m(D)\in{\mathscr{M}}_W \right\}.
\end{align}
Let us assume that the scalar function, $\nu(r)$, associated to the multiplier operator $\nu(D)$ satisfies:
\begin{description}
\item[\namedlabel{item:S1}{(S1)}]$\nu\in C^1([0,\infty))$ and $\nu(r), \nu'(r)\geq0$, for all $r\geq0$.
\item[\namedlabel{item:S2}{(S2)}] There exists a constant $C>0$ such that $r\nu'(r)\le C$, whenever $r\ge 0$.
\end{description}
Now given a multiplier $m(D)\in\mathscr{M}_D$, we associate its corresponding smoothing effect by introducing the multiplier class
\begin{align}\label{def:M:m}
\mathscr{M}_{S}(m):=\left\{\nu(D) :\nu\, \text{satisfies \ref{item:S1}, \ref{item:S2} and}\, \nu(r)\le C(1+m(r)),\ \text{for some}\ C>0,\ \text{for all}\ r\ge 0 \right\}.
\end{align}
Finally, we define a class of functions which can be thought of as a logarithmic analog of the Gevrey classes. Given $\lam>0$, define
\begin{align}\label{def:chern:foias}
\dot{E}^{\lam,\s}_{\nu,\om}:=\left\{f\in L^2 : \Sob{f}{\dot{E}^{\lam,\s}_{\nu,\om}}:= \Sob{E_\nu^{\lam}f}{\Hdot^{\s}_{\om}}<\infty \right\}.
\end{align}
\section{Statements of Main Results}\label{sect:main:results}
Our main local well-posedness results for the family \eqref{eq:dgsqg} is captured by the following two theorems, the first of which considers the case $\be \in (0,2]$ and the second which considers the endpoint case, $\be=0$, representing the 2D (mildly dissipative) Euler equation. To state these results, we recall that the class of multipliers that modify the constitutive law is denoted by $\mathscr{M}_C$ and is defined in \eqref{def:MC}. The class of multipliers, $ \mathscr{M}_W$, {adjusts the regularity of the phase space and is defined in \eqref{def:MM}.} The class of multipliers characterizing the dissipation operator is denoted by $\mathscr{M}_D$ and defined in \eqref{def:MD}. Lastly, the class of multipliers, $\nu(D)$, that captures the smoothing effect associated to the dissipation operator, $m(D)$, is denoted by $\mathscr{M}_{S}(m)$, and is defined in \eqref{def:M:m}
\begin{Thm}\label{thm:main:dgsqg}
Let $\be \in (0,2]$. Let $p(D)\in\mathscr{M}_C$, $m(D) \in \mathscr{M}_D$, and $\om(D)\in \mathscr{M}_W$, where $p=p_ap_b^{-1}$. Suppose there exists $\gamma<1$ such that
\begin{align}\label{cond:main}
\sup_{y>0}\left\{\frac{1}{m^{\gamma}_1(y)} \left(\int_{0}^{y}\frac{r(p^2(y)+p^2(r))}{(1+r^2){\om^2(r)}}dr\right)^{\frac{1}{2}},\quad \frac{p_a(y)\om_b(y)}{m^{\gamma}_1(y)}\right\}<\infty.
\end{align}
Then for each $\tht_{0}\in H^{1+\be}_{\om}$, there exists a positive $T=T(\Sob{\tht_{0}}{H^{1+\be}_\om})$ and a unique function, $\tht(\cdotp;\tht_0)$ satisfying \eqref{eq:dgsqg}, such that
\[
\tht \in C([0,T];H^{1+\be}_{\om})\cap L^{2}(0,T; H^{1+\be}_{\om {m}^{1/2}}).
\]
Moreover, the data-to-solution map, $\Phi$, defined by
{\par\nobreak\noindent}\begin{align}\label{def:flowmap:dsqg}
\Phi: H^{1+\be}_{\om}\goesto \bigcup_{T>0}C([0,T]; H^{1+\be}_\om),\quad \tht_{0}\mapsto\tht(\cdotp;\tht_0),
\end{align}
is continuous. Lastly, if $\nu(D)\in \mathscr{M}_{S}(m)$, then there exists $\lam>0$ such that $\tht$ satisfies
\begin{align}\label{est:dsqg:smoothing}
\sup_{0\le t\le T}\Sob{E^{\lam t}_{\nu}\tht(t)}{\Hdot^{ 1+\be}_{\om}}\le C(T,\Sob{\tht_0}{H^{1+\be}_{\om}}),
\end{align}
for a positive constant $C$.
\end{Thm}
For the case $\be=0$, we have the following.
\begin{Thm}\label{thm:main:deuler}
Suppose that $p(D)\in\mathscr{M}_C$, $m(D)\in \mathscr{M}_D$ and $\om(D)\in \mathscr{M}_W$ satisfy \eqref{cond:main}, where $p=p_ap_b^{-1}$. For each $\tht_{0}\in H^{1}_{\om}\cap \Hdot^{-1}_\om$, there exists a positive $T=T(\Sob{\tht_{0}}{H^{1}_\om\cap\Hdot^{-1}_\om})$ and a unique solution, $\tht(\cdotp;\tht_0)$, of \eqref{eq:dgsqg} when $\be=0$, such that
\[
\tht \in C([0,T];H^{1}_{\om}\cap \Hdot^{-1}_\om)\cap L^{2}(0,T; H^{1}_{\om {m}^{1/2}}).
\]
Moreover, if $\nu(D)\in \mathscr{M}_{S}(m)$, then there exists $\lam>0$ such that $\tht$ satisfies
\begin{align}\label{est:deuler:smoothing}
\sup_{0\le t\le T}\Sob{E^{\lam t}_{\nu}\tht(t)}{\Hdot^{1}_{\om}}\le C(T,\Sob{\tht_0}{H^{1}_{\om}}),
\end{align}
for a positive constant $C$. Lastly, the data-to-solution map, $\Phi$, defined by
\begin{align}\label{def:flowmap:deuler}
\Phi: H^{1}_{\om}\cap \Hdot^{-1}_\om\goesto \bigcup_{T>0}C([0,T]; H^{1}_\om \cap\Hdot^{-1}_\om),\quad \tht_{0}\mapsto\tht(t;\tht_0),
\end{align}
is continuous.
\end{Thm}
\begin{Rmk}\label{rmk:lazar}
The work \cite{LazarXue2019} considers \eqref{eq:dgsqg}, where $m(D)=\Lam^\be$, for $\be\in(0,1)$, and $p(D)$ is given by a multiplier with logarithmic-type growth. The conditions satisfied by $p(D)$ are very similar to those imposed here by $\mathscr{M}_W$, but with a few technical differences. We emphasize, however, that our results are first and foremost concerned with local existence and stability-type estimates in borderline topologies, particularly in the full parameter range $\be\in[0,2]$ of the gSQG family, whereas \cite{LazarXue2019} focuses on the issue of global regularity. In this regard, the results of the present article complement those in \cite{LazarXue2019}.
Nevertheless, it would be interesting if the results of \cite{LazarXue2019} could be established for the range $\be\in(1,2)$ or if global existence of weak solutions and their eventual regularity, proved in \cite{LazarXue2019}, can be extended to the class of models addressed by \cref{thm:main:dgsqg}, \cref{thm:main:deuler}. We refer the reader to the notable recent works \cite{ChaeJeongNaOh2023a, ChaeJeongOh2023b} on the $\be=2$ endpoint; see also \cref{rmk:okhitani} for further discussion.
\end{Rmk}
Next, we present a selection of choices for $m,p,\om,\nu$ to demonstrate that \cref{thm:main:dgsqg} and \cref{thm:main:deuler} contain various non-trivial and interesting consequences. The following result establishes well-posedness under a logarithmic form of singularity and dissipation.
\begin{Cor}\label{cor:l:main:dgsqg}
Let $p(D)$ and $m(D)$ be defined by
\begin{align}\label{def:log:pm}
m(D)=\ln(I-\Delta)^{{\mu}},\quad p(D)=\ln(e-\Delta)^{\til{\mu}},\quad \til{\mu}\ge-1/2.
\end{align}
Let $\be \in (0,2]$ and ${{\mu}}>\til{\mu}+1/2$. For each $\tht_{0}\in H^{1+\be}$, there exists a positive $T=T(\Sob{\tht_{0}}{H^{1+\be}})$ and a unique solution, $\tht$, of \eqref{eq:dgsqg}, such that
\[\tht \in C([0,T];H^{1+\be})\cap L^{2}(0,T; H^{1+\be,\frac{\mu}{2}}).\]
and the data-to-solution map, $\Phi$, such that
{\par\nobreak\noindent}\begin{align}\label{cor:l:def:flowmap:dsqg}
\Phi: H^{1+\be}\goesto \bigcup_{T>0}C([0,T]; H^{1+\be}),\quad \tht_{0}\mapsto\tht(t;\tht_0),
\end{align}
is continuous. Furthermore, there exists $\lam>0$ such that $\tht$ satisfies
\begin{align}\label{cor:l:est:dsqg:smoothing}
\sup_{0\le t\le T}\Sob{E^{\lam t}_{(\ln(I-\De))^\al}\tht(t)}{\Hdot^{ 1+\be}}\le C(T,\Sob{\tht_0}{H^{1+\be}}),
\end{align}
for any $0<\al \le \min\{1,\mu\}$, and a positive constant $C$. In particular, if $\mu\ge 1$, we have
\[
\sup_{0\le t\le T}\Sob{\tht(t)}{\Hdot^{ 1+\be+\lam t}}\le C(T,\Sob{\tht_0}{H^{1+\be}})
\]
\end{Cor}
The corresponding result in the case of $\be=0$ is as follows.
\begin{Cor}\label{cor:l:main:deuler}
Let $p(D)$ and $m(D)$ be as in \eqref{def:log:pm} with $\mu>\til{\mu}+1/2$. For each $\tht_{0}\in H^{1}\cap \Hdot^{-1}$, there exists a positive $T=T(\Sob{\tht_{0}}{H^{1}\cap \Hdot^{-1}})$ and a unique solution, $\tht$, of \eqref{eq:dgsqg} corresponding to $\be=0$, such that
\[\tht \in C([0,T];H^{1}\cap \Hdot^{-1})\cap L^{2}(0,T; H^{1,\frac{\mu}2}).\]
and the data-to-solution map, $\Phi$, such that
{\par\nobreak\noindent}\begin{align}\label{cor:l:def:flowmap:deuler}
\Phi: H^{1}\cap \Hdot^{-1}\goesto \bigcup_{T>0}C([0,T]; H^{1}\cap \Hdot^{-1}),\quad \tht_{0}\mapsto\tht(t;\tht_0),
\end{align}
is continuous. Furthermore, there exists $\lam>0$ such that $\tht$ satisfies
\begin{align}\label{cor:l:est:deuler:smoothing}
\sup_{0\le t\le T}\Sob{E^{\lam t}_{(\ln(I-\De))^\al} \tht(t)}{\Hdot^{1}} \le C(T,\Sob{\tht_0}{H^{1}}),
\end{align}
for any $0<\al \le \min\{1,\mu\}$, and a positive constant $C$. In particular, if $\mu\ge 1$, we have
\begin{align}\notag
\sup_{0\le t\le T}\Sob{\tht(t)}{\Hdot^{ 1+\lam t}}\le C(T,\Sob{\tht_0}{H^{1}})
\end{align}
\end{Cor}
\begin{Rmk}\label{rmk:okhitani}
In the recent works \cite{ChaeJeongNaOh2023a, ChaeJeongOh2023b}, the well-posedness and ill-posedness of several models within the scope of \eqref{eq:dgsqg} were studied, but ultimately complementary to the class of models studied in the present article.
In \cite{ChaeJeongNaOh2023a}, the Okhitani model, i.e., $\be=2$, $p(D)=\ln(10+\Lam)$, and $m(D)\equiv0$ in \eqref{eq:dgsqg}, was studied. In addition to local well-posedness, losing estimates in $H^{s(t)>4}$, where $s(t)$ is a decreasing function, were shown to be a fundamental feature of solutions under the evolution of the system. The derivation of the Okhitani model as a limit of regularized models in a time re-scaled sense was also subsequently justified. Local well-posedness in a fixed Sobolev space $H^{s>4}$, i.e., \textit{without losing estimates}, was then established in the presence of mild dissipation $m(D)=(\ln(10+\Lam))^{\mu\geq1}$. On the other hand, in \cite{ChaeJeongOh2023b}, ill-posedness in $H^{s>3}$ in the form of norm inflation and non-existence was established for the mildly dissipative Okhitani model in the regime where $p(D)=(\ln(10+\Lam))^{\til{\mu}>0}$, $m(D)=(\ln(10+\Lam))^{\mu}$, where $\mu<\til{\mu}$.
In contrast, the results proved in the present article identify conditions that guarantee local well-posedness, but specifically in the borderline regularity setting, $H^{1+\be}$. In the particular case of the mildly regularized Okhitani model, i.e., $\be=2$, $p(D)=(\ln(10+\Lam))^{\til{\mu}=1}$, $m(D)=(\ln(10+\Lam))^{\mu}$, where $\mu>\til{\mu}$, in \eqref{eq:dgsqg}, \cref{cor:l:main:dgsqg} establishes local well-posedness in $H^{3}$, under the proviso that $\mu>\til{\mu}+1/2=3/2$. Thus, the results established here complement the well-posedness and ill-posedness results established in \cite{ChaeJeongNaOh2023a, ChaeJeongOh2023b}.
\end{Rmk}
\begin{Rmk}\label{rmk:ill:posed}
Observe that when $\be=1$, we obtain the local well-posedness (in the Hadamard sense) of the mildly dissipative SQG equation. This complements the recent ill-posedness results in borderline (critical) Sobolev spaces for the inviscid SQG equation obtained in \cite{CordobaZoroa-Martinez2021, JeongKim2021}, as well as the local well-posedness result for the logarithmically regularized inviscid SQG equation obtained in \cite{ChaeWu2012}. We emphasize once again that our results extend beyond the SQG, for $1<\be\leq2$, where the $\be=2$ endpoint is suitably modified. Thus, the range $\mu>\til{\mu}+1/2$ identified in \cref{cor:l:main:dgsqg} draws out the putatively sharp threshold for well-posedness at borderline regularity beyond the SQG case.
In particular, the problem of whether \eqref{eq:dgsqg} in the setting of \eqref{cor:l:main:dgsqg}, \eqref{cor:l:main:deuler} is well-posed or not in the borderline Sobolev spaces when $0<\mu\leq\til{\mu}+1/2$ is an open consideration even when $\be\in[0,1]$, i.e., including the Euler endpoint $\be=0$.
\end{Rmk}
We also study the initial value problem for \eqref{eq:dgsqg} in a log-Sobolev borderline space. The multiplier operators $m, p$ in this case are assumed to be of the form of an iterated logarithm. The results are stated below.
\begin{Cor}\label{cor:ll:main:dgsqg}
Let $p(D)$ and $m(D)$ be defined by
\begin{align}\label{def:loglog:pm}
m(D)&=\ln(I+\ln(I-\Delta))^{\mu},\notag\\
p(D)&=\ln(e+\ln(I-\Delta))^{\til{\mu}},\quad \til{\mu}\ge -1/2.
\end{align}
Let $\be \in (0,2]$ and $\mu>\til{\mu}+1/2$. For each $\tht_{0}\in H^{1+\be,\frac{1}{2}}$, there exists a positive $T=T(\Sob{\tht_{0}}{H^{1+\be,\frac{1}{2}}})$ and a unique solution, $\tht$, of \eqref{eq:dgsqg}, such that
\begin{align}\notag
\tht \in C([0,T];H^{1+\be,\frac{1}{2}})\cap L^{2}(0,T; H^{1+\be,\frac{1}{2},\frac{\mu}{2}}),
\end{align}
and the corresponding data-to-solution map
{\par\nobreak\noindent}\begin{align}\label{cor:ll:def:flowmap:dsqg}
\Phi: H^{1+\be,\frac{1}{2}}\goesto \bigcup_{T>0}C([0,T]; H^{1+\be,\frac{1}{2}}),\quad \tht_{0}\mapsto\tht(t;\tht_0),
\end{align}
is continuous. Furthermore, there exists $\lam>0$ such that $\tht$ satisfies
\begin{align}\label{cor:ll:est:dsqg:smoothing}
\sup_{0\le t\le T}\Sob{E^{\lam t}_{(\ln(I+\ln(I-\Delta)))^{\alpha}}\tht(t)}{\Hdot^{ 1+\be, \frac{1}{2}}}\le C(T,\Sob{\tht_0}{H^{1+\be,\frac{1}{2}}}),
\end{align}
for any $0<\al \le \min\{1,\mu\}$, and a positive constant $C$. In particular, if $\mu\ge 1$, we have
\begin{align}\notag
\sup_{0\le t\le T}\Sob{\tht(t)}{\Hdot^{ 1+\be,\frac{1}{2}+\lam t}}\le C(T,\Sob{\tht_0}{H^{1+\be,\frac{1}{2}}}).
\end{align}
\end{Cor}
The corresponding result in the case of $\be=0$ is as follows.
\begin{Cor}\label{cor:ll:main:deuler}
Let $p(D)$ and $m(D)$ be as in \eqref{def:loglog:pm} with $\mu>\til{\mu}+1/2$. For each $\tht_{0}\in H^{1,\frac{1}{2}}\cap \Hdot^{-1,\frac{1}{2}}$, there exists a positive $T=T(\Sob{\tht_{0}}{H^{1,\frac{1}{2}}\cap \Hdot^{-1,\frac{1}{2}}})$ and a unique solution, $\tht$, of \eqref{eq:dgsqg} ($\be=0$), such that
\[\tht \in C([0,T];H^{1,\frac{1}{2}}\cap \Hdot^{-1,\frac{1}{2}})\cap L^{2}(0,T; H^{1,\frac{1}{2},\frac{\mu}{2}}).\]
and the data-to-solution map, $\Phi$, such that
\begin{align}\label{cor:ll:def:flowmap:deuler}
\Phi: H^{1,\frac{1}{2}}\cap \Hdot^{-1,\frac{1}{2}}\goesto \bigcup_{T>0}C([0,T]; H^{1,\frac{1}{2}}\cap \Hdot^{-1,\frac{1}{2}}),\quad \tht_{0}\mapsto\tht(t;\tht_0),
\end{align}
is continuous. Furthermore, there exists $\lam>0$ such that $\tht$ satisfies
\begin{align}\label{cor:ll:est:deuler:smoothing}
\sup_{0\le t\le T}\Sob{E^{\lam t}_{ (\ln(I+\ln(I-\Delta)))^{\alpha}}\tht(t)}{\Hdot^{ 1, \frac{1}{2}}}\le C(T,\Sob{\tht_0}{H^{1,\frac{1}{2}}}),
\end{align}
for any $0<\al \le \min\{1,\mu\}$, and a positive constant $C$. In particular, if $\mu\ge 1$, we have
\begin{align}\notag
\sup_{0\le t\le T}\Sob{\tht(t)}{\Hdot^{ 1,\frac{1}{2}+\lam t}}\le C(T,\Sob{\tht_0}{H^{1,\frac{1}{2}}}).
\end{align}
\end{Cor}
The main theorem, \cref{thm:main:dgsqg} will be proved in \cref{sect:wellposed}. Before proceeding to develop its proof, we will discuss an application of \cref{thm:main:dgsqg} to the so-called Euler endpoint, $\be=0$.
\section{Application to the Mildly Dissipative 2D Euler Equation: Global regularity}
A natural consideration in light of the above local well-posedness results is the issue of global regularity of the considered models. In the endpoint case, $\be=0$, which represents the dissipatively perturbed Euler equations, we establish global regularity of solutions as an application of the smoothing effect conferred by the mild dissipation and the existence of a maximum principle. In particular, we prove the following result.
\begin{Thm}\label{thm:deuler:global}
Given $\tht_0 \in H^{1}\cap \Hdot^{-1}$, consider the initial value problem \eqref{eq:dgsqg} for $\be=0$, where $p(D)=(\ln(I+\ln(I-\De))^\gam$, $\gam\in[0,1]$, and $m(D)=\ln(I-\Delta)$, i.e.
\begin{align}\label{eq:Euler:endpoint}
\partial_{t}\theta + \ln(I-\Delta)\tht+u \cdot \nabla \theta=0, \quad
u=\nabla^{\perp}{\psi},\quad \Delta {\psi}=(\ln(I+\ln(I-\De))^\gam\theta.
\end{align}
Then the unique solution satisfies
\begin{align*}
\tht \in C([0,T]; H^{1}\cap\Hdot^{-1}),\quad \sup_{0\leq t\leq T}\Sob{\tht(t)}{\Hdot^{1+\lam t}}<\infty,
\end{align*}
for all $T>0$. In particular, \eqref{eq:Euler:endpoint} is globally well-posed in $H^1\cap \Hdot^{-1}$ in the Hadamard sense.
\end{Thm}
\begin{Rmk}\label{rmk:osgood}
\cref{thm:deuler:global} is consistent with previous global well-posedness results found in \cite{Elgindi2014, ChaeConstantinWu2011, DabkowskiKiselevSilvestreVicol2014} in {non-borderline functional settings} for the so-called slightly supercritical Euler equations, i.e., $m(D)\equiv0$, but $p(D)=(\ln(I+\ln(I-\De)))^\gam$, where $\gam\in[0,1]$. Indeed, ill-posedness in the form of norm inflation or non-existence can occur in borderline topologies \cite{BourgainLi2015, ElgindiMasmoudi2020}. Thus, the improvement in \cref{thm:deuler:global} is that global well-posedness holds in the borderline topology $H^1\cap H^{-1}$ in the presence of logarithmic dissipation of order $1$.
\end{Rmk}
The first step is to establish a maximum principle. To do so, let us denote by
\[
{L}=\ln(I-\Delta).
\]
We will consider an alternative representations of ${L}$ via the heat semigroup. This is accomplished through the following elementary identity:
\begin{align}\label{eq:unity:ident}
\ln(1+\lam)=\int_0^\infty (1-e^{-s\lam})e^{-s}\frac{ds}s
\end{align}
Indeed, we see that
\begin{align}
\ln(1+\lam)=\int_0^1\frac{\lam}{1+\lam\tau}d\tau=\int_0^1\int_0^\infty \lam e^{-(1+\lam\tau)s}dsd\tau=\int_0^\infty \left(\int_0^1\lam e^{-\lam s\tau}d\tau\right) e^{-s}ds=\int_0^\infty (1-e^{-s\lam})e^{-s}\frac{ds}s.\notag
\end{align}
Hence
\begin{align}\label{eq:L1}
{L}f(x)=\int_0^\infty(f(x)-e^{s\De}f(x))e^{-s}\frac{ds}s.
\end{align}
We then claim that the following inequality holds.
\begin{Lem}\label{lem:convex} Given $\Phi\in C^1(\RR)$ convex, we have
\begin{align}\label{eq:convex}
\Phi'(f){L}f-{L}\Phi(f)\geq0.
\end{align}
\end{Lem}
\begin{proof}
We make use of \eqref{eq:L1}. Indeed, observe that
\begin{align}
\Phi&'(f(x))({L}f)(x)-({L}\Phi(f))(x)\notag\\
=&\int_0^\infty\left[\Phi'(f(x))f(x)-\Phi'(f(x))e^{s\De}f(x)-\left(\Phi(f(x))-(e^{s\De}\Phi(f))(x))\right)\right]e^{-s}\frac{ds}{s}\notag\\
=&\int_0^\infty\left[\Phi'(f(x))f(x)-\int_{\RR^d}\Phi'(f(x))H(s,x-y)f(y)dy-\Phi(f(x))+\int_{\RR^d}H(s,x-y)\Phi(f(y))dy\right]e^{-s}\frac{ds}{s}\notag\\
=&\int_0^\infty\left[\int_{\RR^d}H(s,x-y)\left(\Phi(f(y))-\Phi(f(x))-\Phi'(f(x))\left(f(y)-f(x)\right)\right)dy\right]e^{-s}\frac{ds}{s}.\notag
\end{align}
By the convexity of $\Phi$, it follows that
\[
\Phi(f(y))-\Phi(f(x))-\Phi'(f(x))\left(f(y)-f(x)\right)\geq0.
\]
We may now deduce \eqref{eq:convex}.
\end{proof}
Using a standard argument and applying \cref{lem:convex}, we establish a maximum principle for the case \eqref{eq:Euler:endpoint}, where $m(D)=\ln(I-\De)$. This result is stated in \cref{lem:max:principle} whose proof is provided in \cref{app:proof:maxp}. In this setting, we interpret $\tht$ as the scalar vorticity of a two-dimensional incompressible fluid.
\begin{Lem}\label{lem:max:principle}
Given a sufficiently smooth solution of \eqref{eq:Euler:endpoint} on the time interval $[0,T]$ such that $\tht_0\in L^2\cap L^\infty$, there exists $C>0$ such that
\begin{align}
\sup_{0\leq t\leq T}\Sob{\tht(t)}{L^\infty}\leq C(\Sob{\tht_0}{L^2},\Sob{\tht_0}{L^\infty}).
\end{align}
\end{Lem}
Finally, we prove global well-posedness of the initial value problem \eqref{eq:Euler:endpoint} posed in $H^1\cap\Hdot^{-1}$.
\begin{proof}[Proof sketch of \cref{thm:deuler:global}]
Given $\tht_0\in H^1$, \cref{cor:l:main:deuler} yields a local solution $\tht\in C([0,T_0];H^1\cap\Hdot^{-1})$, for some $T_0>0$, depending only on $\Sob{\tht_0}{H^1}$. By the linear Sobolev smoothing effect in \eqref{cor:l:est:deuler:smoothing} (with $\al=1$), it follows that $\tht(T_0/2)\in \Hdot^{1+\lam T_0/2}$, for some $\lam>0$. In particular $\tht(T_0/2)\in L^\infty$. By the \cref{lem:max:principle}, $C_\infty:=\sup_{T_0/2\leq t\leq T_0}\Sob{\tht(t)}{L^\infty}<\infty$, depends only on the value at $t=T_0/2$, and by the basic energy inequality $C_2:=\sup_{0\leq t\leq T_0}\Sob{\tht(t)}{L^2}\leq\Sob{\tht_0}{L^2}$.
We now consider the equation over the time interval $[T_0/2,T_0]$. Applying $\nabla$ on \eqref{eq:Euler:endpoint}, and then taking $L^2$ inner product with $\nabla \tht$, we obtain
\begin{align*}
\frac{d}{dt}\Sob{\nabla\tht}{L^2}^2+\Sob{{L}(D)^{1/2}\nabla \tht}{L^2}^2\le C\Sob{\nabla u}{L^\infty}\Sob{\nabla \tht}{L^2}^2.
\end{align*}
We recall a simple variation of a classical borderline Sobolev inequality that was established in \cite{Elgindi2014, ChaeConstantinWu2012}, which implies
\begin{align}\label{est:classical}
\Sob{\nabla u}{L^\infty}\le C_2+C_\infty\left(1+\ln\left(1+\Sob{\tht}{H^{1+\de}}\right)\right)\left(\ln\left(1+\ln(1+\Sob{\tht}{H^{1+\de}}\right)\right)^\gam.
\end{align}
for any $\de\in(0,1)$, over the time inteval $[T_0/2,T_0]$. Choose $\de_0=\lam T_0/2$ and $s_0=1+\de_0$. Then
\begin{align}\label{est:H1:open}
\frac{d}{dt}\Sob{\nabla\tht}{L^2}^2\le \left[C_2+C_\infty\left(1+\ln\left(1+\Sob{\tht}{H^{s_0}}\right)\right)\left(\ln\left(1+\ln(1+\Sob{\tht}{H^{s_0}}\right)\right)^\gam\right]\Sob{\nabla\tht}{L^2}^2,
\end{align}
holds for $t\in[T_0/2,T_0]$.
Applying $(I-\De)^{s_0/2}$ to \eqref{eq:Euler:endpoint}, and then taking $L^2$ inner product with $(I-\De)^{s_0/2} \tht$, and applying the Kato-Ponce commutator inequality (see \cite{KatoPonce1988}), we obtain
\begin{align*}
\frac{d}{dt}\Sob{\tht}{H^{s_0}}^2+\Sob{{L}(D)^{1/2} \tht}{H^{s_0}}^2\le C\Sob{\nabla u}{L^\infty}\Sob{\tht}{H^{s_0}}^2.
\end{align*}
By another application of \eqref{est:classical} with $\de=\de_0$, followed by the Gronwall inequality, we may deduce that
\[
\sup_{T_0/2\leq t\leq T_0}\Sob{\tht(t)}{H^{s_0}}\leq C_0(C_2,C_\infty, T_0),
\]
for some constant $C_0$ depending only on $C_2, C_\infty, T_0$. This bound, in turn, allows one to close the $H^1$-estimate in \eqref{est:H1:open} over the time interval $[T_0/2,T_0]$. A standard continuation argument now applies to extend the solution globally.
\end{proof}
\section{Product and Commutator estimates}
To estimate the advective nonlinearity, we will make use of product and commutator estimates involving logarithmic and polynomial differential operators. The first result is \cref{thm:prod}, which establishes a product estimate localized to dyadic shells in frequency. These results are of a general nature, independent of the equation \eqref{eq:dgsqg} and the structure of its advecting velocity. Similar estimates were developed in \cite{ColombiniDelSantoFanelliMetivier2015} in the setting of logarithmic Besov spaces. By comparison, our estimates are performed entirely in the Sobolev setting, but we expand the ``regularity parameter" to encompass iterated logarithms as well; the proofs of such estimates are relegated to \cref{app:prod}.
Next, we establish three commutator estimates: \cref{lem:commutator2}, \cref{lem:commutator3}, and \cref{lem:commutator4}. The estimates established in \cref{lem:commutator2} and \cref{lem:commutator3} involve the operator defining the constitutive law in \eqref{eq:dgsqg}, whereas \cref{lem:commutator4} studies commutators involving differential operators that will be used to capture the smoothing mechanism conferred by our dissipation; these commutators are directly inspired by the study of the Gevrey regularity and may be viewed as an expanded development of the classical Gevrey-norm approach introduced by Foias and Temam \cite{FoiasTemam1989}.
We recall from \eqref{eq:dgsqg} that the constitutive law is given by $v=-\nabla^{\perp}{\Lam}^{\be-2}p(D)\tht$. In the regime of $\be>1$, this operator corresponds to an integral more singular than Riesz transform and constitutes the fundamental difficulty in this regime. In \cref{lem:commutator2}, we establish a commutator estimate for the trilinear term in a non-localized form. We obtain sharper estimates in \cref{lem:commutator3} under additional localizing assumptions. The proofs of \cref{lem:commutator2} and \cref{lem:commutator3} are based on an approach similar to the one used in \cite{JollyKumarMartinez2020b}, which crucially exploits an elementary convexity estimate (see \eqref{est:elem:convex}) and the Littlewood-Paley decomposition.
\subsection{Product estimates}\label{sect:prod}
The first result that we state is the main product estimate which generalizes \cref{lem:Sobolev}. As mentioned earlier, the proof will be supplied in \cref{app:prod}, but we invoke it to prove the main commutator estimate in \cref{sect:comm}. We are now ready to state our main product estimate.
\begin{Thm}\label{thm:prod}
Let $d\ge 2$. Suppose that $s,\bar{s}\in\RR$ are given such that $s, \bar{s}\leq d/2$ and $s+\bar{s}>0$.
Let $\om, \om_{\ell},\tilde{\om}_\ell\in\mathscr{M}_W$, for $\ell=1,2,3$. Assume that $\Gam_\ell:[0,\infty)\goesto[0,\infty)$, where $\ell=1,2,3$, are functions such that for all $y\geq0$
\begin{align}\label{cond:om:Gam:a}
\begin{split}
&\frac{\om(y)}{{\tilde{\om}_1(y)}}\left(\indFn{(-\infty,d/2)}(s)\int_0^1\frac{r^{d-2s-1}}{\om_1^2(y r)}dr+\indFn{[d/2,\infty)}(s)\int_{0}^{y}\frac{r^{d-1}}{(1+r^2)^{d/2}\om_{1}^{2}(r)}dr\right)^{1/2}\le C_1\Gam_1(y),\\
& \frac{\om(y)}{{\tilde{\om}_2(y)}}\left(\indFn{(-\infty,d/2)}(\bar{s})\int_0^1\frac{r^{d-2\bar{s}-1}}{\om_2^2(y r)}dr+\indFn{[d/2,\infty)}(\bar{s})\int_{0}^{y}\frac{r^{d-1}}{(1+r^2)^{d/2}\om_{2}^{2}(r)}dr\right)^{1/2}\le C_2\Gam_2(y),
\end{split}
\end{align}
and
\begin{align}\label{cond:om:Gam:b}
\frac{\om(y)}{{\om_3(y)}{\tilde{\om}_3(y)}}\le C_3\Gam_{3}(y),
\end{align}
for some $C_1,C_2,C_3>0$. Then there exists $C>0$ and $\{c_j\}\in\ell^2(\ZZ)$ satisfying $\Sob{\{c_j\}}{\ell^2}\leq1$ such that the following inequality holds
\begin{align}\label{est:product:Thm}
\Sob{\triangle_j(fg)}{L^2}
&\leq
Cc_j\om(2^j)^{-1}{2^{-(s+\bar{s}-d/2)j}}\left( {\Gam_1(2^j)}\pi^{s,\bar{s}}_{\om_1,\til{\om}_1}(f,g)+{\Gam_2(2^j)}\pi^{\bar{s},s}_{\om_2,\til{\om}_2}(g,f)+{\Gam_3(2^j)}\rho^{s,\bar{s}}_{\om_3,\til{\om}_3}(f,g)\right),
\end{align}
where for $k=1,2$, we have
\begin{align}\label{def:pi:rho}
\pi^{r,t}_{\vr,\til{\vr}}(f,g):=\begin{cases}\Sob{f}{\Hdot^{r}_{\vr}}\Sob{g}{\Hdot^{t}_{\til{\vr}}},&r<d/2\\
\Sob{f}{H^{r}_{\vr}}\Sob{g}{\Hdot^{t}_{\til{\vr}}},&r=d/2
\end{cases},\qquad \rho^{r,t}_{\vr,\til{\vr}}(f,g):=\Sob{f}{\Hdot^{r}_{\vr}}\Sob{g}{\Hdot^{t}_{\til{\vr}}}.
\end{align}
\end{Thm}
Upon summing \cref{thm:prod} in $j$, we obtain a product estimate in a modified Sobolev space.
\begin{Cor}\label{cor:prod}
Under the assumptions of \cref{thm:prod} with $\Gam=\max\{\Gam_1, \Gam_2, \Gam_3\}$, it follows that
\begin{align}\label{est:product:Cor}
\Sob{fg}{\Hdot^{s+\bar{s}-d/2}_{\om \Gam^{-1}}}\leq
C\left( \pi^{s,\bar{s}}_{\om_1,\til{\om}_1}(f,g)+\pi^{\bar{s},s}_{\om_2,\til{\om}_2}(g,f)+\rho^{s,\bar{s}}_{\om_3,\til{\om}_3}(f,g)\right)
\end{align}
\end{Cor}
\subsection{Commutator estimates for logarithmic-type multipliers}\label{sect:comm}
Now we state and prove our main commutator estimates. Note that we will make use of the usual commutator bracket notation:
\[
[A,B]=AB-BA.
\]
The commutator estimates that we establish will directly involve multipliers associated to either the phase space, characterized by the class $\mathscr{M}_W$ defined in \eqref{def:MM}, or to its slightly larger counterpart $\tilde{\mathscr{M}}_W$ defined in \eqref{def:MMt} that eventually corresponds to multipliers associated to the consitutive law.
\begin{Lem}\label{lem:commutator2}
Let $s\in[0,1)$, $\de, \eps \in (0,1]$ be given such that $\eps+s\le 1$ and $\de<\eps$. Let $p\in\tilde{\mathscr{M}}_W$ be represented as $p=p_ap_b^{-1}$. Then there exists a constant $C>0$ such that
\begin{align}\label{est:commutator2:permute}
|\lb [{\Lam}^{-s}p(D)\bdy_\ell,g]f,h\rb|
\leq C\Sob{g}{{H}^{2-s-\de}}\left(\Sob{p_a(D)f}{\Hdot^{\eps}}\Sob{h}{L^2}+\Sob{p_a(D)h}{\Hdot^{\eps}}\Sob{f}{L^2}\right).
\end{align}
\end{Lem}
In order to prove \cref{lem:commutator2}, we will make use of an elementary convexity estimate that was proved in \cite{JollyKumarMartinez2020b}. We briefly recall the inequality here: let $\varphi,\vartheta\in\RR^2$ such that $|\vartheta|=1$. Then for all $s\in[0,1)$, there exists a constant $C>0$, independent of $\varphi,\vartheta$, such that
\begin{align}\label{est:elem:convex}
\int_0^1\frac{1}{|\varphi+\tau\vartheta|^s}d\tau\leq C.
\end{align}
We are now ready to prove \cref{lem:commutator2}.
\begin{proof}[Proof of \cref{lem:commutator2}]
It will be convenient to define the following functional:
\begin{align}\label{def:L:functional:1}
\mathcal{L}(f,g,h):=\iint m(\xi,\eta)\hat{f}(\xi-\eta)\hat{g}(\eta)\overline{\hat{h}(\xi)}d\eta d\xi,
\end{align}
where
\begin{align*}
m(\xi,\eta)=|\xi|^{-s}p(|\xi|)\xi_\ell-|\xi-\eta|^{-s}p(|\xi-\eta|)(\xi-\eta)_\ell.
\end{align*}
By Plancherel's theorem, we see that
\[
\mathcal{L}(f,g,h)=\lb [{\Lam}^{-s}p(D)\bdy_\ell,g]f,h\rb.
\]
Denote the parametrization of the line segment starting at $\xi-\eta$ and ending at $\xi$ by
\begin{align}\label{def:S}
\mathbf{S}(\tau,\xi,\eta):=\tau\xi+(1-\tau)(\xe),\quad \tau\in[0,1].
\end{align}
For convenience, we will suppress the dependence of $\mathbf{S}$ on $\xi,\eta$. Moreover, we fix $\xi,\eta\in\mathbb{R}^2$, where $|\eta|\neq0$.
Observe that we have
\begin{align}\label{eq:meanvalue1}
|m(\xi,\eta)|&=\bigg|\int_{0}^{1}\frac{d}{{d\tau}}\left(\abs{\mathbf{S}(\tau)}^{-s}p(|\mathbf{S}(\tau)|)\mathbf{S}(\tau)_{\ell}\right)d\tau\bigg|\notag\\
&=\bigg |{\int_{0}^{1}\left(-s\abs{\mathbf{S}(\tau)}^{-s-2}(\mathbf{S}(\tau)\cdot \eta)p(|\mathbf{S}(\tau)|)\mathbf{S}(\tau)_{\ell}+\abs{\mathbf{S}(\tau)}^{-s}p(|\mathbf{S}(\tau)|)\eta_{\ell}\right.}\notag\\
&{\left.\qquad\qquad\qquad \qquad\qquad\qquad+\abs{\mathbf{S}(\tau)}^{-s-1} p'(|\mathbf{S}(\tau)|)(\mathbf{S}(\tau)\cdot \eta)\mathbf{S}(\tau)_{\ell}\right)d\tau}\bigg|\notag\\
&\le |\eta|\int_{0}^{1}\abs{\mathbf{S}(\tau)}^{-s}\left(p(|\mathbf{S}(\tau)|)+ |p'(|\mathbf{S}(\tau)|)||\mathbf{S}(\tau)|\right)d\tau\notag\\
&\le C\abs{\eta}\int_{0}^{1}\abs{\mathbf{S}(\tau)}^{-s}p(|\mathbf{S}(\tau)|)\,d\tau\notag\\
&\le C\Ae\left(\int_{0}^{1}\abs{\mathbf{S}(\tau)}^{-s}d\tau\right)\sup_{\tau\in[0,1]}p(|\mathbf{S}(\tau)|),
\end{align}
where we applied \ref{item:O2} in obtaining the second inequality since
\[\frac{|(p'_a(r)r)p_b(r)-(p'_b(r)r)p_a(r)|}{p^2_b(r)}\le C\frac{p_a(r)p_b(r)+p_b(r)p_a(r)}{p^2_b(r)}=2Cp(r).\]
Now let $\varphi=\frac{\xe}{\Ae}$ and $\vartheta=\frac{\eta}{\Ae}$. For fixed $\xi$ and $\eta$, we have from \eqref{est:elem:convex} that
\begin{align}\label{eq:A-integral}
\int_{0}^{1}\abs{ \mathbf{S}(\tau)}^{-s}\,d\tau=\Ae^{-s}\int_{0}^{1}\frac{1}{\abs{\varphi+\tau \vartheta}^{s}}\,d\tau\leq C\Ae^{-s}.
\end{align}
Applying \eqref{eq:A-integral} in \eqref{eq:meanvalue1} and invoking \ref{item:O1}, yields
\begin{align}\label{est:psl:final}
|m(\xi,\eta)|\leq C|\eta|^{1-s}\sup_{\tau\in[0,1]}p(|\mathbf{S}(\tau)|)\le C|\eta|^{1-s}p_a(|\xe|+|\xi|).
\end{align}
We recall that $\eps+s\leq1$, where $\eps>0$. Thus, upon invoking \eqref{def:gen:tri:prod}, we obtain
\begin{align*}
|\mathcal{L}(f,g,h)| \le& C\iint\left(|\eta|^{1-s-\eps} |\hat{g}(\eta)|\right)\left(p_a (|\xe|)|\xe|^{\eps}|\hat{f}(\xi-\eta)|\right)|\hat{h}(\xi)|d\eta d\xi\notag\\
&+C\iint\left(|\eta|^{1-s-\eps} |\hat{g}(\eta)|\right)|\hat{f}(\xe)|\left(p_a (|\xi|)|\xi|^{\eps}|\hat{h}(\xi)|\right)d\eta d\xi.
\end{align*}
Lastly, upon applying the Cauchy-Schwarz inequality, Young's convolution inequality, and Plancherel's theorem, we obtain \eqref{est:commutator2:permute}.
\end{proof}
When the functions $f$ and $h$ are spectrally localized away from the origin, one can afford additional flexibility in \cref{lem:commutator2}.
\begin{Lem}\label{lem:commutator3}
Let $s\in[0,1)$ and $\eps \in [0,1]$ be such that $\eps+s\le 1$. Let $\om \in \mathscr{M}_W$ and $p \in \tilde{\mathscr{M}}_W$, where $p$ is represented as $p_ap_b^{-1}$. Let $\Gam:[0,\infty)\goesto[0,\infty)$ be a function for which there exists a $C>0$ such that
\begin{align}\label{cond:Gam3}
\left(\int_{0}^{y}\frac{rdr}{(1+r^2)\om^2(r)}\right)^{\frac{1}{2}}\le C\Gam(y),
\end{align}
holds for all $y\geq0$. Then there exists a constant $C>0$ such that if $\supp \hat{f},\ \supp \hat{h}\subset\Acal_j$, for some $j\in\mathbb{Z}$, where $\Acal_j$ is defined in \eqref{def:ann:ball}, then
\begin{align}
|\lb [{\Lam}^{-s}p(D)\bdy_\ell,g]f,h\rb|\leq C\left(p(2^j)\Gam(2^j)+p_a(2^j)\om^{-1}(2^j)\right)2^{\eps j}\Sob{g}{H^{2-s-\eps}_\om}\Sob{f}{L^2}\Sob{h}{L^2}. \notag
\end{align}
\end{Lem}
\begin{proof}
By the spectral support condition of $f,h$, we may additionally assume that $\supp \hat{g}\subset\Bcal_{j+2}$. Using this, we obtain
\[\mathcal{L}(f,g,h)=I+II,\]
where
\begin{align}
I=&\iint m(\xi,\eta)\hat{f}(\xi-\eta)\mathbbm{1}_{{\Bcal}_{j-3}}(\eta)\hat{g}(\eta)\overline{\hat{h}(\xi)}d\eta d\xi,\label{eq:L:split:lf}\\
II=&\iint m(\xi,\eta)\hat{f}(\xi-\eta)\mathbbm{1}_{{\Acal}_{j-3,j+2}}(\eta)\hat{g}(\eta)\overline{\hat{h}(\xi)}d\eta d\xi\label{eq:L:split:hf}.
\end{align}
Now we treat $I$ and $II$. Clearly,
\[
\abs{\mathbf{S}(\tau)}\le |\xi|+|\xe|\le 2^{j+2}.
\]
For $ \eta \in {\Bcal}_{j-3}$, we have
\begin{align*}
\abs{\mathbf{S}(\tau)}& \ge \abs{\xe}-\tau\abs{\eta} \ge 2^{j-1}-2^{j-3} \ge 2^{j-2}.
\end{align*}
From this, \eqref{eq:meanvalue1}, and \eqref{est:omega}, we obtain
\begin{align*}
|m(\xi,\eta)|\le C|\eta|^{1-s-\eps}p(2^j)2^{\eps j}.
\end{align*}
Applying the Cauchy-Schwarz inequality and \eqref{cond:Gam3}, we have
\begin{align*}
\Sob{|\cdotp|^{1-s-\eps}\mathbbm{1}_{{\Bcal}_{j-3}} \hat{g}}{L^1}&\le\left(\int_{\Bcal_{j-3}}\frac{1}{(1+|\eta|^2)\om^{2}(|\eta|)}d\eta\right)^{1/2}\left(\int_{\Bcal_{j-3}} {(1+|\eta|^2)^{2-s-\eps}\om(|\eta|)^{2}}|\hat{g}(\eta)|^2d\eta\right)^{1/2}\notag\\
&\leq C\Gam(2^j)\Sob{g}{H^{2-s-\eps}_\om}.
\end{align*}
Using the above estimate in \eqref{eq:L:split:lf} and applying Young's convolution inequality and Plancherel's theorem gives us
\begin{align}\label{est:I:lf:final}
I\leq Cp(2^j)2^{\eps j}\Gam(2^j)\Sob{g}{H^{2-s-\eps}_\om}\Sob{f}{L^2}\Sob{h}{L^2}.
\end{align}
For $ \eta \in {\Acal}_{j-3,j+2}$, we see from \eqref{est:psl:final}, \ref{item:O1}, and \eqref{est:omega} that
\begin{align}\label{est:psl:final:loc}
|m(\xi,\eta)|\leq C|\eta|^{1-s-\eps}p_a(2^j)2^{\eps j}.
\end{align}
Applying the Cauchy-Schwarz inequality and \eqref{est:omega}, we have
\begin{align*}
\Sob{|\cdotp|^{1-s-\eps}\mathbbm{1}_{{\Acal}_{j-3,j+2}} \hat{g}}{L^1}&\le\left(\int_{\Acal_{j-3,j+2}}\frac{1}{|\eta|^2\om^{2}(|\eta|)}d\eta\right)^{1/2}\left(\int_{\Acal_{j-3,j+2}} {|\eta|^{2(2-s-\eps)}\om(|\eta|)^{2}}|\hat{g}(\eta)|^2d\eta\right)^{1/2}\notag\\
&\leq C\om^{-1}(2^j)\left(\int_{2^{j-3}}^{2^{j+2}}\frac{1}{r}\,dr\right)^{\frac{1}{2}}\Sob{g}{\Hdot^{2-s-\eps}_\om}\le C\om^{-1}(2^j)\Sob{g}{\Hdot^{2-s-\eps}_\om}.
\end{align*}
Using the above estimate in \eqref{eq:L:split:hf} and applying Young's convolution inequality and Plancherel's theorem gives us
\begin{align}\label{est:II:hf:final}
II\leq Cp_a(2^j)2^{\eps j}\om^{-1}(2^j)\Sob{g}{\Hdot^{2-s-\eps}_\om}\Sob{f}{L^2}\Sob{h}{L^2}.
\end{align}
Finally collecting the bounds in \eqref{est:I:lf:final} and \eqref{est:II:hf:final}, we obtain the desired estimate.
\end{proof}
Given $\lam\geq0$ and $\nu(D)\in\mathscr{M}_S$, we define the operator $E^\lam_\nu$ by
\begin{align}\label{def:Elamnu}
(\mathcal{F}E^\lam_\nu\phi)(\xi)=e^{\lam\nu(|\xi|)}(\mathcal{F}\phi)(\xi)
\end{align}
We then have the following commutator estimates.
\begin{Lem}\label{lem:commutator4}
Let $r, s, \bar{s}\in \RR$ be such that $s,\bar{s}\leq1$, $s+\bar{s}>0$. Let $\om,\om_\ell,\tilde{\om}_\ell\in\mathscr{M}_W$, for $\ell=1,2,3$. Assume that $\Gam_\ell:[0,\infty)\goesto[0,\infty)$, where $\ell=1,2,3$, are functions satisfying \eqref{cond:om:Gam:a}, \eqref{cond:om:Gam:b}. Let $\pi^{r,t}_{\vr,\til{\vr}}$ and $\rho^{r,t}_{\vr,\til{\vr}}$ be defined as in \eqref{def:pi:rho}. Then there exist constants $c,C>0$, and $\{c_j\}\in\ell^2(\ZZ)$ satisfying $\Sob{\{c_j\}}{\ell^2}\leq1$ such that if $\supp \hat{h}\subset\Acal_j$, then
\begin{align}\label{eq:commutator4}
&|\lb[\om(D){\Lam}^{r}E^\lam_\nu{\lpj}\bdy_\ell,g]f,h\rb|\leq
C(1+\lam)e^{c\lam}c_j2^{(r-s-\bar{s}+1)j}\\
&\times\left\{ \Gam_1(2^j)\pi^{s,\bar{s}}_{\om_1,\til{\om}_1}(E^\lam_\nu f,\Lam{E^\lam_\nu}g)+\Gam_2 (2^j)\pi^{\bar{s},s}_{\om_2,\til{\om}_2}(\Lam {E^\lam_\nu}g,{E^\lam_\nu}f)+\Gam_{3}(2^j)\rho^{s,\bar{s}}_{\om_3,\til{\om}_3}({E^\lam_\nu}f,\Lam{E^\lam_\nu}g)\right\}\Sob{h}{L^2},\notag
\end{align}
for all $\lam\geq0$. Lastly, the same inequality holds if $\om(D){\Lam}^{r}E^\lam_\nu{\lpj}\bdy_\ell$ is replaced by $\om(D){\Lam}^{r+1}E^\lam_\nu{\lpj}$ in the left-hand side of \eqref{eq:commutator4}.
\end{Lem}
\begin{proof}[Proof of \cref{lem:commutator4}]
In order to avoid redundancy in the argument, we will only provide the proof for \eqref{eq:commutator4}. With this in mind, we define
\begin{align*}
\mathcal{L}(f,g,h):=\iint\limits_{\xi \in \Acal_j} m(\xi,\eta)\hat{f}(\xi-\eta)\hat{g}(\eta)\overline{\hat{h}(\xi)}\, d\eta\,d\xi,
\end{align*}
where
\begin{align*}
m(\xi,\eta):= e^{\lam \nu(|\xi|)}\abs{\xi}^{r}\xi_{\ell}\om(|\xi|)\phi_{j}(\xi)-e^{\lam \nu(|\xe|)}\abs{\xi-\eta}^{r}(\xi-\eta)_{\ell}\om(|\xe|)\phi_{j}(\xe).
\end{align*}
By Plancherel's theorem we see that
\begin{align}\label{def:L:comm:relation}
\mathcal{L}(f,g,h)=\lb[\om(D){\Lam}^{s}E^\lam_\nu{\lpj}\bdy_\ell,g]f,h\rb.
\end{align}
Hence, it will be equivalent to obtain the desired bounds for $\mathcal{L}(f,g,h)$.
Let $\mathbf{S}(\tau)$ be as in \eqref{def:S}. It follows that
\begin{align}\label{eq:meanvalue2}
m(\xi,\eta)
&=\int_{0}^{1}\Bigg{\{}\left(\lam \nu'(\mathbf{S}(\tau))\frac{\mathbf{S}(\tau)}{|\mathbf{S}(\tau)|}\cdot \eta+r\frac{\mathbf{S}(\tau)}{|\mathbf{S}(\tau)|^{2}}\cdot \eta+\frac{\om'(\mathbf{S}(\tau))}{\om (|\mathbf{S}(\tau)|)}\frac{\mathbf{S}(\tau)}{|\mathbf{S}(\tau)|}\cdot \eta\right)\phi_j (\mathbf{S}(\tau))\mathbf{S}(\tau)_{\ell}\notag\\
& \qquad \qquad\qquad+\left(\phi_{j}(\mathbf{S}(\tau))\eta_{\ell}+(\nabla \phi_{0})(2^{-j}\mathbf{S}(\tau))\cdot(2^{-j} \eta)\mathbf{S}(\tau)_{\ell}\right)\Bigg{\}} e^{\lam \nu(|\mathbf{S}(\tau)|)}|\mathbf{S}(\tau)|^{r}\om (|\mathbf{S}(\tau)|)d\tau\notag.
\end{align}
Suppose that $\xi\in\Acal_j$. Since $\supp \phi_{j} \subset \mathcal{A}_{j}$ and $\supp \nabla \phi_0\subset\mathcal{A}_0$, it follows from \eqref{est:omega} that
\begin{align}\notag
|m(\xi,\eta)|&\leq C(1+\lam)2^{jr}|\eta|\om(2^j)\int_{0}^{1}\Bigg{\{}\left(|\mathbf{S}(\tau)| \nu'(\mathbf{S}(\tau))+1+\frac{|\mathbf{S}(\tau)|\om'(\mathbf{S}(\tau))}{\om (|\mathbf{S}(\tau)|)}\right)\phi_j (\mathbf{S}(\tau))+1\Bigg{\}} e^{\lam \nu(|\mathbf{S}(\tau)|)}d\tau.
\end{align}
We then apply \ref{item:S1}, \ref{item:S2}, and \ref{item:O3} to deduce
\begin{align}
|m(\xi,\eta)|&\leq C(1+\lam)\abs{\eta}2^{r j}\om(2^j)e^{\lam\nu(|\xi-\eta|+|\eta|)}
\end{align}
Now we estimate $\nu(|\xe|+|\eta|)$. We consider two cases:
\subsubsection*{Case: $|\xe|<|\eta|$} Integrating in \ref{item:S2} yields
\begin{align*}
\nu(|\xe|+|\eta|)\le \nu(|\eta|)+C\ln\left(1+\frac{|\xe|}{|\eta|}\right)\le \nu(|\eta|)+C.
\end{align*}
\subsubsection*{Case: $|\eta|\le|\xe|$} Similarly, \ref{item:S2} implies
\begin{align*}
\nu(|\xe|+|\eta|)\le \nu(|\xe|)+C\ln\left(1+\frac{|\eta|}{|\xe|}\right)\le \nu(|\xe|)+C.
\end{align*}
Upon returning to \eqref{eq:meanvalue2} and invoking these estimates, we arrive at
\begin{align*}
|m(\xi,\eta)|\le C(1+\lam)\abs{\eta}2^{r j}\om(2^j)e^{\lam C}e^{\lam \nu(|\eta|)}e^{\lam \nu(|\xe|)}.
\end{align*}
Finally, we let
\[
\mathcal{F}F=|\mathcal{F}{E^\lam_\nu}f|,\quad \mathcal{F}G=|\mathcal{F}{\Lam{E^\lam_\nu}g}|.
\]
We apply \eqref{eq:meanvalue2} in \eqref{def:L:comm:relation}, followed by the Cauchy-Schwarz inequality and Young's convolution inequality to obtain
\begin{align}\label{in:final:lem2}
|\mathcal{L}(f,g,h)|\leq C(1+\lam)e^{\lam C}2^{rj}\om(2^j)\Sob{\tlpj(FG)}{L^2}\Sob{h}{L^2},
\end{align}
where $\tlpj$ denotes the extended Littlewood-Paley blocks as defined in \eqref{def:para:factors}. Since $s,\bar{s}\in\RR$ is assumed to satisfy $s, \bar{s}\leq 1$ and $s+\bar{s}>0$, and since the $\Gam_\ell$ satisfy \eqref{cond:om:Gam:a}, \eqref{cond:om:Gam:b}, we may apply \cref{thm:prod}. The proof is complete upon application of Plancherel's theorem, followed by the characterization of Sobolev norms in terms of Besov norms (see \eqref{eq:Sob:Bes}).
\end{proof}
\section{Analysis of the Protean System}\label{sect:apriori}
In this section, we will obtain apriori estimates for the solution of \eqref{def:mod:flux} in $L^2$, $H^{\s}_{\om}$, and ${E}_{\nu,\s,\om}^\lam$, for particular choices of $\s\in[-1,3]$ that depend on $\be\in[0,2]$ and whether $\lam=0$ or not. We will restrict to $\s$ in the following ranges:
\begin{align}\label{sigma:range}
\s\in
\begin{cases}
(-1,2], &\text{if} \quad \be \in [0,1],\\
[1,1+\be], &\text{if} \quad \be \in (1,2],\\
\{-1\},&\text{if}\quad \be=0, \lam=0.
\end{cases}
\end{align}
Development of the subsequent apriori estimates establishes a global existence theory for the protean system \eqref{eq:mod:claw}. This is formally stated in \cref{thm:modclaw:wellposed} after the apriori estimates have been established; the rigorous details for global existence and uniqueness are supplied in \cref{app:well:posed}.
Observe that when $\be\in[0,1]$, \eqref{eq:mod:claw} is a linear transport equation, while in the more singular regime, $\be\in(1,2]$, \eqref{eq:mod:claw} is a conservation law with a flux that modifies the linear transport equation. The modification is ultimately required to accommodate suitable stability-type estimates for \eqref{eq:dgsqg}, but due to \eqref{eq:cancel}, we find it expedient to obtain all estimates in the generality of \eqref{eq:mod:claw} and simply reduce them to the case of \eqref{eq:dgsqg}, as needed. The formal apriori estimates are developed in \cref{sect:Hs:apriori}. The stability-type estimates are then established in \cref{sect:stab:apriori}. For the analysis below, it will be convenient to introduce the notation
\begin{align}\label{def:Md}
m_1(D):=I+m(D).
\end{align}
Recall that $m_1(D)\in\mathscr{M}_W$ (see \eqref{def:MD}) and, thus, satisfies \ref{item:O1}, \ref{item:O2}, \ref{item:O3}. For convenience, we also recall that $\om\in\mathscr{M}_W$ (see \eqref{def:MM}), $p\in\mathscr{M}_C$ (see \eqref{def:MC}), and $\nu(D)\in\mathscr{M}_S(m)$ (see \eqref{def:M:m}).
\begin{Rmk}\label{rmk:convention:constants}Below, we will adopt the convention of summation over repeated indices. We also denote by $C$ a generic positive constant, which may depend on various regularity parameters in addition to the size of a given time interval $[0,T]$. For clarity, we may indicate the dependence of the $C$ on parameters through a subscript. In general, however, the value of $C$ may change line-to-line.
\end{Rmk}
We begin by establishing the following lemma demonstrating the equivalence of the norms $\Sob{\phi}{\Hdot^\s_\om\cap L^2_{\om}}$, $\Sob{\phi}{L^2}+\Sob{\phi}{\Hdot^\s_\om}$, and $\Sob{\phi}{H^\s_\om}$ when $\s>0$.
\begin{Lem}\label{lem:norm:equiv}
Let $\om(D)\in\mathscr{M}_W$ as defined in \cref{sect:op:classes}. For any $\s>0$, there exist positive constants $c_{\s,\om}, C_{\s,\om}$ such that
\begin{align}\label{eq:norm:equiv:a}
c_{\s,\om}^{-1}\Sob{\phi}{\Hdot^\s_\om\cap L^2_{\om}}\leq\Sob{\phi}{\Hdot_\om^\s}+ \Sob{\phi}{L^2}\leq c_{\s,\om}\Sob{\phi}{\Hdot^\s_\om\cap L^2_{\om}},
\end{align}
and
\begin{align}\label{eq:norm:equiv:b}
C_{\s,\om}^{-1}\Sob{\phi}{H^\s_\om}\leq \Sob{\phi}{\Hdot_\om^\s}+\Sob{\phi}{L^2}\leq C_{\s,\om}\Sob{\phi}{H^\s_\om}.
\end{align}
\end{Lem}
\begin{proof}
Since $\s>0$, we may invoke \eqref{eq:omega:eps}, in addition to \ref{item:O1}, to obtain
\begin{align}\notag
\Sob{\phi}{L^2_{\om}}^2=\int_{\RR^2} |\hat{\phi}(\xi)|^2\om(|\xi|)d\xi=\int_{|\xi|\leq1}+\int_{|\xi|>1}\leq \frac{\om_a(1)}{\om_a(0)}\Sob{\phi}{L^2}^2+c_\s^2\Sob{\phi}{\Hdot_\om^\s}^2.
\end{align}
Hence
\begin{align}
\Sob{\phi}{\Hdot^\s_\om\cap L^2_{\om}}^2&\leq\frac{\om_a(1)}{\om_a(0)}\Sob{\phi}{L^2}^2+(1+c_\s^2)\Sob{\phi}{\Hdot_\om^\s}^2\notag\\
&\leq \left(\frac{\om_a(1)}{\om_a(0)}+1+c_\s^2\right)\Sob{\phi}{H^\s_\om}^2\leq C_\s^2\left(\frac{\om_a(1)}{\om_a(0)}+1+c_\s^2\right)\Sob{\phi}{\Hdot^\s_\om\cap L^2_{\om}}^2,\notag
\end{align}
for some constant $C_\s>0$. This implies \eqref{eq:norm:equiv:a}, \eqref{eq:norm:equiv:b}.
\end{proof}
\subsection{A priori estimates in $L^2$}\label{sect:L2:apriori}
Upon taking the inner product in $L^2$ of \eqref{eq:mod:claw} with $\tht$, we obtain
\begin{align}\label{eq:balance:L2}
\frac{1}{2}\frac{d}{dt}\Sob{\tht}{L^2}^2+\Sob{m(D)^{\frac{1}2}\tht}{L^2}=-\lb\Div F_q(\tht),{\tht}\rb+\lb G ,\tht \rb=I^0+II^0.
\end{align}
We estimate $II^0$ with the Cauchy-Schwarz inequality and Young's inequality, to obtain
\begin{align}\label{est:II:L2}
|II^0|\leq& \Sob{m_1(D)^{-\frac{1}2}G}{L^2}\Sob{m_1(D)^{\frac{1}2}\tht}{L^2}\notag \\\leq& C\Sob{m_1(D)^{-\frac{1}2}G}{L^2}^2+C\Sob{\tht}{L^2}^2+\frac{1}8\Sob{m(D)^{\frac{1}2}\tht}{L^2}^2.
\end{align}
We now consider the cases $\be \in [0,1]$ and $\be \in (1,2]$ separately to treat $I^0$.
\subsubsection*{Case $\be\in[0,1]$}
By \eqref{def:mod:flux} and \eqref{eq:v:cancel}, we have $I^0=0$. Given \eqref{est:II:L2} for $II^0$, \eqref{eq:balance:L2} then becomes
\begin{align}\label{est:L2:a}
\frac{d}{dt}\Sob{\tht}{L^2}^2+\frac{7}4\Sob{m(D)^{\frac{1}2}\tht}{L^2}^2\leq C\Sob{m_1(D)^{-\frac{1}2}G}{L^2}^2+C\Sob{\tht}{L^2}^2.
\end{align}
An application of Gronwall's inequality, then yields
\begin{align}\label{est:L2:a:gronwall}
\Sob{\tht(t)}{L^2}^2\leq C\exp(CT)\left(\Sob{\tht_0}{L^2}^2+\int_0^T\Sob{m_1(D)^{-\frac{1}2}G(s)}{L^2}^2ds\right),
\end{align}
for all $0\leq t\leq T$. Then, upon integrating \eqref{est:L2:a} over $[0,T]$ and applying \eqref{est:L2:a:gronwall}, we obtain
\begin{align}\label{est:L2log:a:gronwall}
\int_0^T\Sob{m(D)^{\frac{1}2}\tht(s)}{L^2}^2ds\leq C\exp(CT)\left(\Sob{\tht_0}{L^2}^2+\int_0^T\Sob{m_1(D)^{-\frac{1}2}G(s)}{L^2}^2ds\right).
\end{align}
Hence
\begin{align}\label{est:L2:final}
\sup_{0\leq t\leq T}\left(\Sob{\tht(t)}{L^2}^2+\int_0^t\Sob{m(D)^{\frac{1}2}\tht(s)}{L^2}^2ds\right)\leq C\exp(CT)\left(\Sob{\tht_0}{L^2}^2+\int_0^T\Sob{m_1(D)^{-\frac{1}2}G(s)}{L^2}^2ds\right).
\end{align}
\subsubsection*{Case $\be \in (1,2]$}
Since $a(D)\bdy_\ell$ is a skew self-adjoint operator, in light of \eqref{eq:v:cancel}, we see that
\begin{align}\label{eq:skew:adjoint}
I^0=\lb a(D){\nabla}\cdotp(({\nabla}^\perp q)\tht),\tht\rb&=-\lb\nabla^\perp q\cdotp\nabla (a(D)\tht),\tht\rb\notag\\
&=\frac{1}{2}\lb [a(D),\nabla^\perp q\cdotp\nabla]\tht,\tht\rb=\frac{1}{2}\lb [a(D)\bdy_\ell,\bdy^\perp_\ell q]\tht,\tht\rb,
\end{align}
where $\bdy^\perp_\ell q=(\nabla^\perp q)^\ell$ and we adopt the convention of summation over repeated indices. Upon applying \cref{lem:commutator2} with $s=2-\beta$ and $,\eps',\eps\in(0,1]$ sufficiently small so that $\eps+s\leq1$, and $(1+\gam)\eps'\leq\eps$, we obtain
\begin{align}\notag
|I^0|
\leq C\Sob{\nabla^{\perp}q}{{H}^{\be-\de}}\Sob{p_a(D)\tht}{\Hdot^{\eps'}}\Sob{\tht}{L^2},
\end{align}
for any $0<\de<\eps$. We assume that
\begin{align}\label{cond:pa}
\sup_{y>0}\frac{p_a(y)}{\om(y)m^{\gam}_1(y)}<\infty.
\end{align}
Thus
\begin{align}\notag
|I^0|\leq C\Sob{q}{H_\om^{1+\be}}\Sob{m_1(D)^{\gam}\tht}{\Hdot_\om^{\eps'}}\Sob{\tht}{L^2}.
\end{align}
We recall that $m_1(D)$ satisfies \eqref{eq:omega:eps:bdd}, so that
\[
\Sob{m_1(D)^{\gam}\tht}{\Hdot_\om^{\eps'}}\leq \Sob{\tht}{H_\om^{(1+\gam)\eps'}}\leq \Sob{\tht}{H_\om^{\eps}}.
\]
Finally, after an application of the Cauchy-Schwarz inequality, we arrive at estimate
\begin{align}\label{est:I:L2}
|I^0|\leq C\Sob{q}{H_\om^{1+\be}}\Sob{\tht}{H_\om^{\eps}}\Sob{\tht}{L^2}.
\end{align}
Returning to \eqref{eq:balance:L2} and applying \eqref{est:II:L2} and \eqref{est:I:L2} then yields
\begin{align}\label{est:L2:b}
\frac{d}{dt}\Sob{\tht}{L^2}^2+\frac{7}4\Sob{m(D)^{\frac{1}2}\tht}{L^2}^2\le C\Sob{q}{H^{1+\be}_\om}\Sob{\tht}{H^\eps_\om}\Sob{\tht}{L^2}+C\Sob{m_1(D)^{-\frac{1}2}G}{L^2}^2+C\Sob{\tht}{L^2}^2.
\end{align}
In particular, when $\be\in(1,2]$ we will require control in $H^\eps_\om$, for some $\eps>0$, in order to close estimates.
\subsection{A priori estimates in $\dot{E}^\lam_{\nu,\s,\om}$}\label{sect:Hs:apriori} With the exception of the special case $\be=0$, $\s=-1$, it will be convenient to develop the apriori estimates in the stronger space $\dot{E}^{\lam}_{\nu,\s,\om}$ and then specialize to the case of $\Hdot^\s_{\om}$ later by simply setting $\lam=0$. We emphasize that the commutator estimates developed in \cref{sect:comm} can accommodate such a procedure due to the form of dependency of the constants on $\lam$ (see \cref{lem:commutator4}). In developing the estimates in $\dot{E}^{\lam}_{\nu,\s,\om}$, we will make use of the following shorthand: given $r\in\RR$, $\lam \ge 0$, and $j\in\ZZ$, we let
\begin{align}\label{def:short}
\til{f}:=E^{\lam(t)}_\nu f,\quad \til{\Lam}^{r}_{\om}:=\om(D){\Lam}^{r}E^{\lam(t)}_\nu,\quad \til{\Lam}_{\om,j}^{r}=:\om(D){\Lam}^{r}E^{\lam(t)}_\nu{\lpj},\quad \lam(t)=\lam_1t,
\end{align}
for a fixed $\lam_1>0$, where $E^\lam_\nu$ is defined as in \eqref{def:Elamnu}. We will often abuse notation and say $\lam=\lam(t)$. Throughout, we suppose that $0\leq t\leq T$.
Similar to \cref{lem:norm:equiv}, we establish the equivalance of the norms $\Sob{\til{\phi}}{\Hdot^\s}+\Sob{\phi}{L^2}$, $\Sob{\til{\phi}}{\Hdot^\s\cap L^2}$, and $\Sob{\til{\phi}}{H^\s}$ are equivalent as norms when $\s\geq0$.
\begin{Lem}\label{lem:norm:equiv:til}
Given $m(D)\in\mathscr{M}_D$, Suppose $\nu(D)\in\mathscr{M}_S(m)$, where $\mathscr{M}_S(m)$ is defined as in \eqref{def:MD} from \cref{sect:op:classes}. Given $\lam>0$, let $E^\lam_\nu$ be defined as in \eqref{def:Elamnu}. For any $\s\geq0$, there exist a positive constant $C_{\lam,m}$ such that
\begin{align}\label{eq:norm:equiv:til:a}
C_{\lam,m}^{-1}\Sob{E^\lam_\nu\phi}{\Hdot^\s\cap L^2}\leq &\Sob{E^\lam_\nu\phi}{\Hdot^\s}+\Sob{\phi}{L^2}\leq C_{\lam,m}\Sob{E^\lam_\nu\phi}{\Hdot^\s\cap L^2}.
\end{align}
and
\begin{align}\label{eq:norm:equiv:til:b}
C_{\lam,m}^{-1}\Sob{E^\lam_\nu\phi}{H^\s}\leq &\Sob{E^\lam_\nu\phi}{\Hdot^\s}+\Sob{\phi}{L^2}\leq C_{\lam,m}\Sob{E^\lam_\nu\phi}{H^\s}.
\end{align}
\end{Lem}
\begin{proof}
Observe that from \eqref{def:MD}, we have
\begin{align}
\Sob{\til{\phi}}{L^2}^2=\int_{\RR^2}e^{2C\lam\nu(|\xi|)}|\hat{\phi}(\xi)|^2d\xi\leq \int_{|\xi|\leq 1}e^{2\lam(1+m(|\xi|))}|\hat{\phi}(\xi)|^2d\xi+\int_{|\xi|>1}|\xi|^{2\s}|\til{\phi}(\xi)|^2d\xi,\notag
\end{align}
provided that $\s\geq0$. Since $I+m(D)\in{\mathscr{M}}_W$, we may assume that $I+m(D)=m_a(D)m_b(D)^{-1}$. In particular, by \ref{item:O1}, it follows that $1+m(|\xi|)\leq m_a(1)m_b(0)^{-1}$. Hence
\begin{align}
\Sob{\til{\phi}}{\Hdot^\s}^2+\Sob{\til{\phi}}{L^2}^2&\leq \exp\left(2C\lam\frac{m_a(1)}{m_b(0)}\right)\Sob{\phi}{L^2}^2+2\Sob{\til{\phi}}{\Hdot^\s}^2\notag\\
&\leq \exp\left(2C\lam\frac{m_a(1)}{m_b(0)}\right)\Sob{\til{\phi}}{L^2}^2+2\Sob{\til{\phi}}{\Hdot^\s}^2\leq C_{\lam,m}^2\Sob{\til{\phi}}{H^\s}^2,\notag
\end{align}
which implies \eqref{eq:norm:equiv:til:a}, \eqref{eq:norm:equiv:til:b}.
\end{proof}
From \cref{lem:norm:equiv} and \cref{lem:norm:equiv:til}, we immediately deduce the following equivalence.
\begin{Cor}\label{cor:norm:equiv}
Suppose $\om(D)\in\mathscr{M}_W$, $m(D)\in\mathscr{M}_D$, $\nu(D)\in\mathscr{M}_S(m)$, and that $E^\lam_\nu$ is defined as in \eqref{def:Elamnu} with $\lam>0$. Then for any $\s\geq0$, we have the following chain of equivalent norms:
\begin{align}\label{eq:norm:equiv:summary}
\Sob{E^\lam_\nu\phi}{\Hdot^\s_\om\cap L^2_{\om}}\sim\Sob{E^\lam_\nu\phi}{H^\s_\om}+\Sob{E^\lam_\nu\phi}{L^2}\sim\Sob{E^\lam_\nu\phi}{H^\s_\om}+\Sob{\phi}{L^2}\sim\Sob{E^\lam_\nu\phi}{H^\s_\om},
\end{align}
where the suppressed constants depend on $\s,\lam,\om,m$.
\end{Cor}
Now, upon applying $ \til{\Lam}_{\om,j}^{\s}$ to \eqref{eq:mod:claw}, one obtains
\begin{align}\label{eq:balance:Hsigma:prep}
\partial_t (\til{\Lam}_{\om,j}^{\s} \tht)+\til{\Lam}_{\om,j}^{\s}(\Div F_q(\tht))=\til{\Lam}^{\s}_{\om,j}\tht-m_1(D)\til{\Lam}_{\om,j}^{\s}\tht+\lam_1 \nu(D)\til{\Lam}_{\om,j}^{\s}\tht+\til{\Lam}_{\om,j}^{\s}G.
\end{align}
Then taking the $L^2$--inner product of \eqref{eq:balance:Hsigma:prep} with $ \til{\Lam}_{\om,j}^{\s}\tht$, we obtain
\begin{align}\label{eq:balance:Hsigma}
\frac{1}{2}\frac{d}{dt}\Sob{ \til{\Lam}_{\om,j}^{\s}\tht}{L^2}^2+&\Sob{m_1(D)^{\frac{1}2} \til{\Lam}_{\om,j}^{\s}\tht}{L^2}^{2}\notag\\
&=\lam_1\Sob{\nu(D)^{\frac{1}2} \til{\Lam}_{\om,j}^{\s}\tht}{L^2}^{2}+\Sob{\til{\Lam}^\s_{\om,j}\tht}{L^2}^2-\lb \til{\Lam}_{\om,j}^{\s}(\Div F_q(\tht)), \til{\Lam}_{\om,j}^{\s}{\tht}\rb+\lb \til{\Lam}_{\om,j}^{\s}G , \til{\Lam}_{\om,j}^{\s}\tht\rb\notag\\
&=\lam_1\Sob{\nu(D)^{\frac{1}2} \til{\Lam}_{\om,j}^{\s}\tht}{L^2}^{2}+\Sob{\til{\Lam}^\s_{\om,j}\tht}{L^2}^2+I^\s+II^\s.
\end{align}
Invoking the fact that $\nu\in \mathscr{M}_R(m)$ (see \eqref{def:M:m}), we have for $\lam_1$ sufficiently small
\begin{align}\label{est:gevrey:residual}
\lam_1\Sob{\nu(D)^{\frac{1}2} \til{\Lam}_{\om,j}^{\s}\tht}{L^2}^{2}\le \frac{1}{16}\Sob{m_1(D)^{\frac{1}2} \til{\Lam}_{\om,j}^{\s}\tht}{L^2}^{2}.
\end{align}
We estimate $II^\s$ with the Cauchy-Schwarz inequality and Young's inequalities to obtain
\begin{align}\label{est:II:Hsigma}
|II^\s|&\le \Sob{m_1(D)^{-\frac{1}2}\til{\Lam}^{\s}_{\om,j} G}{L^2}\Sob{m_1(D)^{\frac{1}2}\til{\Lam}^{\s}_{\om,j} \tht}{L^2}\notag\\&\leq C\Sob{m_1(D)^{-\frac{1}2}\til{\Lam}^{\s}_{\om,j} G}{L^2}^2+\frac{1}{16}\Sob{m_1(D)^{\frac{1}2}\til{\Lam}^{\s}_{\om,j}\tht}{L^2}^2.
\end{align}
Upon applying \eqref{est:gevrey:residual} and \eqref{est:II:Hsigma} in \eqref{eq:balance:Hsigma}, we obtain
\begin{align}\label{est:balance:Hsigma:final}
\frac{d}{dt}\Sob{\til{\Lam}_{\om,j}^{\s}\tht}{L^2}^2+\frac{7}4\Sob{m_1(D)^{\frac{1}2} \til{\Lam}_{\om,j}^{\s}\tht}{L^2}^{2}
&\leq 2\Sob{ \til{\Lam}_{\om,j}^{\s}\tht}{L^2}^2+C\Sob{m_1(D)^{-\frac{1}2}\til{\Lam}^{\s}_{\om,j} G}{L^2}^2+2{I^{\s}}.
\end{align}
We treat $I^{\s}$ by considering the cases $\be\in[0,1]$, $\be\in(1,2]$, and $\be=0$, recalling that $\s$ is restricted by \eqref{sigma:range}.
\subsubsection*{Case 1: $\be\in[0,1]$} In this case, \eqref{sigma:range} implies $\s\in(-1,2]$.
\subsubsection*{Subcase 1a: $\s\in(-1,1]$} First observe that since $v$ is divergence-free, we have in \eqref{est:balance:Hsigma:final} that
\begin{align}\label{def:I:decomp:a}
I^{\s}=\lb[\til{\Lam}^{\s}_{\om, j}\bdy_{\ell},v^{\ell}]\tht,\til{\Lam}^{\s}_{\om,j}\tht\rb.
\end{align}
Applying \cref{lem:commutator4} with $r=\s, (s,\bar{s})=(\s,1)$, $(\om_1,\tilde{\om}_1)=(\om,m_1^{\frac{\gam}{2}}p^{-1}\om), (\om_2,\tilde{\om}_2)=(p^{-1}\om,m_1^{\frac{\gam}{2}}\om), (\om_3,\tilde{\om}_3)=(\om, m_1^{\frac{\gam}{2}}p^{-1}\om)$, and $\Gam_1=\Gam_2=\Gam_3=m^{\frac{\gam}2}_1$, we obtain
\begin{align}\label{est:I:hfreq:a}
|I^\s|
&\le C_\lam c_j m_1(2^j)^{\frac{\gam}{2}}\left(\nrm{\til{\tht}}_{\dot{H}^{\s}_\om \cap L^2_\om}\nrm{m_1(D)^{\frac{\gam}{2}}p(D)^{-1}\Lam\til{ v}}_{\Hdot^{1}_{\om}}+ \nrm{p(D)^{-1}\Lam\til{ v}}_{H^1_\om}\nrm{m_1(D)^{\frac{\gam}{2}}\til{\tht}}_{\Hdot^{\s}_\om}\right.\notag\\&\hspace{20 em}\left.+\nrm{\til{\tht}}_{\Hdot^{\s}_\om}\nrm{m_1(D)^{\frac{\gam}{2}}p(D)^{-1}\Lam \til{v}}_{\Hdot^{1}_\om}\right)\nrm{\til{\Lam}^{\s}_{\om, j}\tht}_{L^2}\notag\\
&\le C_\lam c_{j}\left( \nrm{\til{\tht}}_{\dot{H}^{\s}_\om \cap L^2_\om}\nrm{m_1(D)^{\frac{\gam}{2}}\til{q}}_{\Hdot^{1+\be}_\om}+\nrm{\til{q}}_{H^{1+\be}_\om}\nrm{m_1(D)^{\frac{\gam}{2}}\til{\tht}}_{\Hdot^{\s}_\om}\right)\nrm{m_1(D)^{\frac{\gam}{2}}\til{\Lam}^{\s}_{\om, j}\tht}_{L^2}.
\end{align}
Then \eqref{cond:om:Gam:a}, \eqref{cond:om:Gam:b} become
\begin{align}\label{cond:Gam:I0}
\begin{split}
&\sup_{y>0}\frac{p(y)}{{m^{\gam}_1(y)}}\left(\indFn{(-\infty,1)}(\sigma)\int_0^1\frac{r^{1-2\s}}{\om^2(y r)}dr+\indFn{[1,\infty)}(\s)\int_{0}^{y}\frac{r}{(1+r^2)\om^{2}(r)}dr\right)^{\frac{1}{2}}<\infty,\\
&\sup_{y>0}\frac{1}{{m^{\gam}_1(y)}}\left(\int_{0}^{y}\frac{rp^2(r)}{(1+r^2)\om^{2}(r)}dr\right)^{\frac{1}{2}},\qquad \sup_{y>0}\frac{p(y)}{\om(y)m^{\gam}_1(y)}<\infty.
\end{split}
\end{align}
\subsubsection*{Subcase 1b: $\s\in (1,2]$}
Since $\nabla \cdot v=0$, it follows that
\begin{align}\label{def:I:decomp:b}
I^\s=\lb[\til{\Lam}^{\s}_{\om, j},v^{\ell}]\bdy_{\ell}\tht,\til{\Lam}^{\s}_{\om,j}\tht\rb.
\end{align}
We apply \cref{lem:commutator4} with $r=\s-1$, $(s,\bar{s})=(\s-1,1)$, $(\om_1,\tilde{\om}_1)=(\om,m_1^{\frac{\gam}{2}}p^{-1}\om)$, $(\om_2,\tilde{\om}_2)=(p^{-1}\om,m_1^{\frac{\gam}{2}}\om)$, $(\om_3,\tilde{\om}_3)=(\om, m_1^{\frac{\gam}{2}}p^{-1}\om)$, and $\Gam_1=\Gam_2=\Gam_3=m^{\frac{\gam}2}_1$, to obtain
\begin{align}\label{est:I:hfreq:b}
|I^\s|
\le C_\lam c_{j}\left( \nrm{\til{\tht}}_{H^{\s}_\om}\nrm{m_1(D)^{\frac{\gam}{2}}\til{q}}_{\Hdot^{1+\be}_\om}+\nrm{\til{q}}_{H^{1+\be}_\om}\nrm{m_1(D)^{\frac{\gamma}{2}}\til{\tht}}_{\Hdot^{\s}_\om}\right)\nrm{m_1(D)^{\frac{\gam}{2}}\til{\Lam}^{\s}_{\om, j}\tht}_{L^2},
\end{align}
provided that the following holds:
\begin{align}\label{cond:Gam:I0:b}
\begin{split}
&\sup_{y>0}\frac{p(y)}{{m^{\gam}_1(y)}}\left(\indFn{(-\infty,1)}(\sigma-1)\int_0^1\frac{r^{1-2(\s-1)}}{\om^2(y r)}dr+\indFn{[1,\infty)}(\s-1)\int_{0}^{y}\frac{r}{(1+r^2)\om^{2}(r)}dr\right)^{\frac{1}{2}}<\infty,\\
&\sup_{y>0}\frac{1}{{m^{\gam}_1(y)}}\left(\int_{0}^{y}\frac{rp^2(r)}{(1+r^2)\om^{2}(r)}dr\right)^{\frac{1}{2}},\qquad \sup_{y>0}\frac{p(y)}{\om(y)m^{\gam}_1(y)}<\infty.
\end{split}
\end{align}
\subsubsection*{Concluding Estimates for Case 1: $\be \in [0,1]$}
First observe that the conditions \eqref{cond:Gam:I0} and \eqref{cond:Gam:I0:b} can both be reduced to the following single set of conditions:
\begin{align}\label{cond:Gam:I0:final}
\begin{split}
\sup_{y>0}\left\{\frac{1}{m_1^\gam(y)}\left(\int_{0}^{y}\frac{r(p^2(y)+p^2(r))}{(1+r^2)\om^{2}(r)}dr\right)^{\frac{1}{2}},\quad \frac{p_a(y)\om_b(y)}{m^{\gam}_1(y)}\right\}<\infty.
\end{split}
\end{align}
Upon returning to \eqref{est:balance:Hsigma:final} and combining it with \eqref{est:I:hfreq:a}, \eqref{est:I:hfreq:b}, then summing over $j$ and invoking \eqref{eq:Sob:Bes}, we deduce that
\begin{align}\label{est:Hs:a}
\frac{d}{dt}\Sob{\til{\tht}}{\Hdot^\s_\om}^2+\frac{7}4\Sob{m_1(D)^{\frac{1}2}\til{\tht}}{\Hdot^\s_\om}^{2}&\leq C_\lam \left(\nrm{\til{\tht}}_{\dot{H}^{\s}_\om \cap L^2_\om}\nrm{m_1(D)^{\frac{\gam}{2}}\til{q}}_{\Hdot^{1+\be}_\om}+\nrm{\til{q}}_{H^{1+\be}_\om}\nrm{m_1(D)^{\frac{\gam}{2}}\til{\tht}}_{\Hdot^{\s}_\om}\right)\nrm{m_1(D)^{\frac{\gam}{2}}\til{\tht}}_{\Hdot^\s_\om}\notag\\&\quad+C\left(\Sob{m_1(D)^{-\frac{1}{2}}\til{G}}{\Hdot^\s_\om}^2+\Sob{\til{\tht}}{\Hdot^\s_\om}^2\right),
\end{align}
holds for all $\s\in(-1,2]$, provided that \eqref{cond:Gam:I0:final} holds.
In particular, by jointly applying \eqref{est:Hs:a} for $\s\in(-1,2]$ and $\s=0$, we deduce that
\begin{align}\label{est:Hs:b:final}
\frac{d}{dt}\Sob{\til{\tht}}{\Hdot^\s_\om\cap L^2_{\om}}^2&+\frac{7}4\Sob{m_1(D)^{\frac{1}2}\til{\tht}}{\Hdot^\s_\om\cap L^2_{\om}}^{2}\notag\\
&\leq C_\lam \left(\nrm{\til{\tht}}_{\dot{H}^{\s}_\om \cap L^2_\om}\nrm{m_1(D)^{\frac{\gam}{2}}\til{q}}_{\Hdot^{1+\be}_\om}+\nrm{\til{q}}_{H^{1+\be}_\om}\nrm{m_1(D)^{\frac{\gam}{2}}\til{\tht}}_{\Hdot^{\s}_\om\cap L^2_{\om}}\right)\nrm{m_1(D)^{\frac{\gam}{2}}\til{\tht}}_{\Hdot^\s_\om\cap L^2_{\om}}\notag\\&\quad+C\left(\Sob{m_1(D)^{-\frac{1}{2}}\til{G}}{\Hdot^\s_\om\cap L^2_{\om}}^2+\Sob{\til{\tht}}{\Hdot^\s_\om\cap L^2_{\om}}^2\right).
\end{align}
\begin{Rmk}\label{rmk:equivalence}
Note that upon summing in $j$ in obtaining \eqref{est:Hs:a}, we in fact obtain the inequality
\begin{align}\label{est:Hs:a:rmk}
\frac{d}{dt}\Sob{\Lam^{\s}_{\om}\til{\tht}}{\Bdot^0_{2,2}}^2+&\frac{7}4\Sob{\Lam^{\s}_{\om}m_1(D)^{\frac{1}2}\til{\tht}}{\Bdot^0_{2,2}}^{2}\notag\\
&\leq C_\lam \left(\nrm{\til{\tht}}_{\dot{H}^{\s}_\om \cap L^2_\om}\nrm{m_1(D)^{\frac{\gam}{2}}\til{q}}_{\Hdot^{1+\be}_\om}+\nrm{\til{q}}_{H^{1+\be}_\om}\nrm{m_1(D)^{\frac{\gam}{2}}\til{\tht}}_{\Hdot^{\s}_\om}\right)\nrm{\Lam^{\s}_{\om}m_1(D)^{\frac{\gam}{2}}\til{\tht}}_{\Bdot^0_{2,2}}\notag\\&\quad+C\left(\Sob{\Lam^{\s}_{\om}m_1(D)^{-\frac{1}{2}}\til{G}}{\Bdot^0_{2,2}}^2+\Sob{\Lam^{\s}_{\om}\til{\tht}}{\Bdot^0_{2,2}}^2\right).
\end{align}
Owing to \eqref{eq:Sob:Bes} and \cref{lem:Bernstein}, observe that we may bound all Sobolev-norm based quantities in terms of their equivalent Besov-norm based quantities. Later on, after an application of Young's inequality and Gronwall's inequality, we may convert all Besov-norm based quantities back in terms of their Sobolev-based counterparts. In particular, in the final analysis, all quantities may be interpreted in their Sobolev-based form.
Henceforth, we will abuse notation and express all quantities related to the apriori estimates in terms of Sobolev norms.
\end{Rmk}
\subsubsection*{Case 2: $\be\in(1,2]$}
By \eqref{sigma:range}, we restrict to the regime $\s\in[1,1+\be]$. We will treat the cases $\s \in [1,2)$ and $\s \in [2,1+\be]$ separately.
\subsubsection*{Subcase 2a: $\s\in[1,2)$}
From \eqref{def:mod:flux} and the facts that $v$ is divergence-free and $a(D)\bdy_\ell:=\Lam^{\be-2}p(D)\bdy_\ell$ is skew self-adjoint, we see that $I^\s$ in \eqref{est:balance:Hsigma:final} can be decomposed as
\begin{align}
I^\s
&=I_{1}^\s+I_{2}^\s+I_{3}^\s+I_{4}^\s,\label{def:I:decomp:c}
\end{align}
where
\begin{align*}
I_{1}^\s&= -\lb \til{\Lam}^{\s}_{\om,j}(\nabla^{\perp}a(D)q\cdot\nabla \tht),\til{\Lam}^{\s}_{\om,j}{\tht}\rb+\lb \nabla^{\perp}a(D)q\cdot\nabla\til{\Lam}^{\s}_{\om,j} \tht,\til{\Lam}^{\s}_{\om,j}{\tht}\rb\notag \\&=-\lb [\til{\Lam}^{\s}_{\om,j},\bdy^{\perp}_{\ell}a(D)q]\bdy_{\ell}\tht,\til{\Lam}^{\s}_{\om,j}\tht \rb=-\lb [\til{\Lam}^{\s}_{\om,j},v^\ell]\bdy_\ell\tht,\til{\Lam}^{\s}_{\om,j}\tht\rb,\\
I_{2}^\s&=-\lb \nabla^{\perp}a(D)q\cdot\nabla\til{\Lam}^{\s}_{\om,j} \tht,\til{\Lam}^{\s}_{\om,j}{\tht}\rb=0,\\
I_{3}^\s&=\lb \til{\Lam}^{\s}_{\om,j}a(D)(\nabla^{\perp}q\cdot\nabla \tht),\til{\Lam}^{\s}_{\om,j}{\tht}\rb-\lb \nabla^{\perp}q\cdot\nabla{a(D)}^{1/2}\til{\Lam}^{\s}_{\om,j} \tht,{a(D)}^{\frac{1}2}\til{\Lam}^{\s}_{\om,j}{\tht}\rb\notag\\&=\lb [\til{\Lam}^{\s}_{\om,j}{a(D)}^{\frac{1}2},\bdy^{\perp}_{\ell}q]\bdy_{\ell}\tht,\til{\Lam}^{\s}_{\om,j}{a(D)}^{\frac{1}2}\tht \rb=\lb [\til{\Lam}^{\s+\frac{\be}2-1}_{\om p^{1/2},j},\bdy^{\perp}_{\ell}q]\bdy_{\ell}\tht,\til{\Lam}^{\s}_{\om,j}{a(D)}^{\frac{1}2}\tht \rb,\\
I_{4}^\s&=-\lb \nabla^{\perp}q\cdot\nabla{a(D)}^{\frac{1}2}\til{\Lam}^{\s}_{\om,j} \tht,\til{\Lam}^{\s}_{\om,j}{a(D)}^{1/2}{\tht}\rb=0.
\end{align*}
We treat $I^\s_1$ as we did with $I^\s$ from \eqref{def:I:decomp:b} in the case $\be\in[0,1]$. In particular, we apply \cref{lem:commutator4} with $r=\s-1$, $(s, \bar{s})=(\s-1,1)$, $(\om_1,\til{\om}_1)=(\om,m_1^{\frac{\gam}{2}}p^{-1}\om)$, $(\om_2, \tilde{\om}_2)=(p^{-1}\om,m_1^{\frac{\gam}{2}}\om)$, $(\om_3, \tilde{\om}_3)=(\om,m_1^{\frac{\gam}{2}}p^{-1}\om)$, and $\Gam_1=\Gam_2=\Gam_3=m^{\frac{\gam}2}_1$, to obtain
\begin{align}\label{est:I1}
|I_{1}^\s|\le& C_\lam c_jm_1(2^j)^{\frac{\gam}{2}} \left\{\nrm{\til{\tht}}_{\Hdot^{\s}_\om}\nrm{m_1(D)^{\frac{\gam}{2}}p(D)^{-1}\Lam \til{v}}_{\Hdot^{1}_\om}+ \nrm{p(D)^{-1}\Lam\til{ v}}_{H^1_\om}\nrm{m_1(D)^{\frac{\gam}{2}}\til{\tht}}_{\Hdot^{\s}_\om}\right.\notag\\
&\hspace{15 em}\left.+\nrm{\til{\tht}}_{\Hdot^{\s}_\om}\nrm{m_1(D)^{\frac{\gam}{2}}p(D)^{-1}\Lam \til{v}}_{\Hdot^{1}_\om}\right\}\nrm{\til{\Lam}^{\s}_{\om,j}\tht}_{L^2}\notag\\
\le& C_\lam c_{j}\left( \nrm{\til{\tht}}_{\Hdot^{\s}_\om}\nrm{m_1(D)^{\frac{\gam}{2}}\til{q}}_{\Hdot^{1+\be}_\om}+\nrm{\til{q}}_{H^{1+\be}_\om}\nrm{m_1(D)^{\frac{\gam}{2}}\til{\tht}}_{\Hdot^{\s}_\om}\right)\nrm{m_1(D)^{\frac{\gam}{2}}\til{\Lam}^{\s}_{\om,j}\tht}_{L^2},
\end{align}
provided that
\begin{align}\label{cond:Gam:I1s}
\begin{split}
&\sup_{y>0}\frac{p(y)}{{m^{\gam}_1(y)}}\left(\int_0^1\frac{r^{1-2(\s-1)}}{\om^2(y r)}dr\right)^{\frac{1}{2}},\quad\sup_{y>0}\frac{1}{{m^{\gam}_1(y)}}\left(\int_{0}^{y}\frac{rp^2(r)}{(1+r^2)\om^{2}(r)}dr\right)^{\frac{1}{2}},\quad \sup_{y>0}\frac{p(y)}{\om(y)m^{\gam}_1(y)}<\infty.
\end{split}
\end{align}
For $I^\s_3$, we apply \cref{lem:commutator4} with $r+1=\s+\be/2-1$,
($s,\bar{s})=(\s-1,\be-1$), $(\om_1,\til{\om}_1)= (\om_2,\tilde{\om}_2)=(\om_3,\tilde{\om}_3)=(\om, m_1^{\frac{\gam}{2}}\om)$, and $\Gam_1=\Gam_2=\Gam_3=m^{\frac{\gam}2}_1p^{-\frac{1}{2}}$, to obtain
\begin{align}\label{est:I3}
|I_{3}^\s|\le& C_\lam c_jm_1(2^j)^{\frac{\gam}{2}}p(2^j)^{-\frac{1}{2}}2^{(1-\frac{\be}{2})j} \left\{\nrm{\til{\tht}}_{\Hdot^{\s}_\om}\nrm{m_1(D)^{\frac{\gam}{2}}\nabla^{\perp}\Lam \til{q}}_{\Hdot^{\be-1}_\om}+ \nrm{\nabla^{\perp}\Lam \til{q}}_{H^{\be-1}_\om}\nrm{m_1(D)^{\frac{\gam}{2}}\til{\tht}}_{\Hdot^{\s}_\om}\right.\notag\\
&\hspace{15 em}\left.+\nrm{\til{\tht}}_{\Hdot^{\s}_\om}\nrm{m_1(D)^{\frac{\gam}{2}}\nabla^{\perp}\Lam \til{q}}_{\Hdot^{\be-1}_\om}\right\}\nrm{\til{\Lam}^{\s}_{\om,j}a(D)^{\frac{1}2}\tht}_{L^2}\notag\\
\le& C_\lam c_{j}\left( \nrm{\til{\tht}}_{\Hdot^{\s}_\om}\nrm{m_1(D)^{\frac{\gam}{2}}\til{q}}_{\Hdot^{1+\be}_\om}+\nrm{\til{q}}_{H^{1+\be}_\om}\nrm{m_1(D)^{\frac{\gam}{2}}\til{\tht}}_{\Hdot^{\s}_\om}\right)\nrm{m_1(D)^{\frac{\gam}{2}}\til{\Lam}^{\s}_{\om,j}\tht}_{L^2}.
\end{align}
Then \eqref{cond:om:Gam:a}, \eqref{cond:om:Gam:b} become
\begin{align}\label{cond:Gam:I3}
\begin{split}
&\sup_{y>0}\frac{p(y)}{{m^{\gam}_1(y)}}\left(\int_0^1\frac{r^{1-2(\s-1)}}{\om^2(y r)}dr\right)^{\frac{1}{2}},\qquad \sup_{y>0}\frac{p(y)}{\om(y)m^{\gam}_1(y)}<\infty,\\
&\sup_{y>0}\frac{p(y)}{{m^{\gam}_1(y)}}\left(\indFn{(1,2)}(\be)\int_0^1\frac{r^{1-2(\be-1)}}{\om^2(y r)}dr+\indFn{\{2\}}(\be)\int_{0}^{y}\frac{r}{(1+r^2)\om^{2}(r)}dr\right)^{\frac{1}{2}}<\infty.
\end{split}
\end{align}
\subsubsection*{Subcase 2b: $\s\in[2,1+\be]$} From \eqref{def:mod:flux} and the facts that $\nabla^\perp q$ is divergence-free and $a(D)\bdy_\ell$ is skew self-adjoint, we may instead re-write $I^\s$ as
\begin{align}
I^\s&=-\lb \til{\Lam}^{\s}_{\om,j}(\nabla^{\perp}a(D)q\cdot\nabla \tht),\til{\Lam}^{\s}_{\om,j}{\tht}\rb+\lb \til{\Lam}^{\s}_{\om,j}a(D)(\nabla^{\perp}q\cdot\nabla \tht),\til{\Lam}^{\s}_{\om,j}{\tht}\rb\notag\\
&=J_{1}^\s+J_{2}^\s+J_{3}^\s+J_{4}^\s+J_{5}^\s,\label{def:I:decomp:d}
\end{align}
where
\begin{align}
J_{1}^\s=&-\left\{\lb (\nabla^{\perp}a(D)\til{\Lam}^{\s}_{\om,j}{q} \cdot \nabla) \tht,\til{\Lam}^{\s}_{\om,j} \tht \rb -\lb \nabla^{\perp}a(D)\cdot(\til{\Lam}^{\s}_{\om,j}{q}\nabla \tht),\til{\Lam}^{\s}_{\om,j} \tht \rb\right\}\notag\\
=&-\lb [\bdy_{\ell}^{\perp}a(D),\bdy_{\ell} \tht]\til{\Lam}^{\s}_{\om,j}q,\til{\Lam}^{\s}_{\om,j}\tht\rb\notag\\
J_{2}^\s=&-\lb \nabla^{\perp}a(D){q}\cdot {\nabla} \til{\Lam}^{\s}_{\om,j} \tht,\til{\Lam}^{\s}_{\om,j} \tht \rb=0\notag\\
J_{3}^\s=&-\left \{ \lb \til{\Lam}^{\s}_{\om,j}(\nabla^{\perp}a(D){q} \cdot \nabla \theta),\til{\Lam}^{\s}_{\om,j} \tht \rb
- \lb (\nabla^{\perp}a(D)\til{\Lam}^{\s}_{\om,j}{{q}} \cdot \nabla) \tht,\til{\Lam}^{\s}_{\om,j} \tht \rb -\lb \nabla^{\perp}a(D){q}\cdot {\nabla} \til{\Lam}^{\s}_{\om,j} \tht,\til{\Lam}^{\s}_{\om,j} \tht \rb \right\}\notag\\
J_{4}^\s=&-\lb (\nabla^{\perp}{q}\cdot {\nabla} a(D)^{\frac{1}2}\til{\Lam}^{\s}_{\om,j} \tht),a(D)^{\frac{1}2}\til{\Lam}^{\s}_{\om,j} \tht \rb=0\notag\\
J_{5}^\s=&\ \lb \til{\Lam}^{\s}_{\om,j} a(D)(\nabla^{\perp}q\cdot\nabla \tht),\til{\Lam}^{\s}_{\om,j}{\tht}\rb
- \lb \nabla^{\perp}a(D)\cdot(\til{\Lam}^{\s}_{\om,j}{q}\nabla \tht),\til{\Lam}^{\s}_{\om,j} \tht \rb \notag\\
&\quad-\lb (\nabla^{\perp}{q}\cdot {\nabla} a(D)^{\frac{1}2}\til{\Lam}^{\s}_{\om,j} \tht),a(D)^{\frac{1}2}\til{\Lam}^{\s}_{\om,j} \tht \rb \notag
\end{align}
We observe as in \cite{HuKukavicaZiane2015} that we may write $J_{3}^\s$ as a double commutator. Indeed, for any $\widetilde{\s}\geq2$, we have
\begin{align}\label{split:Lam:rewrite}
{\Lam}^{\widetilde{\s}}f={\Lam}^{\widetilde{\s}-2}(-\De)f=-({\Lam}^{\widetilde{\s}-2}\bdy_{l})\bdy_{l}f.
\end{align}
Since $\s\geq2$, we may apply \eqref{split:Lam:rewrite}, so that by the product rule and \eqref{eq:v:cancel}, we have
\begin{align*}
J_{3}^\s=&-\lb\til{\Lam}^{\s-2}_{\om,j}\bdy_{l}(\nabla^{\perp}a(D)\bdy_{l}{q} \cdot \nabla \theta),\til{\Lam}^{\s}_{\om,j} \tht ,\rb+\lb (\nabla^{\perp}a(D)\til{\Lam}^{\s-2}_{\om,j}\bdy_{l}\bdy_{l}{{q}} \cdot \nabla) \tht,\til{\Lam}^{\s}_{\om,j} \tht \rb\\
&-\lb\til{\Lam}^{\s-2}_{\om,j}\bdy_{l}(\nabla^{\perp}a(D){q} \cdot \nabla \bdy_{l}\theta),\til{\Lam}^{\s}_{\om,j} \tht ,\rb+\lb (\nabla^{\perp}a(D){q}\cdot {\nabla} \til{\Lam}^{\s-2}_{\om,j}\bdy_{l}\bdy_{l} \tht),\til{\Lam}^{\s}_{\om,j} \tht \rb\\
=&-\lb[\til{\Lam}_{\om,j}^{\s-2}\bdy_{l},\bdy_{\ell}\tht]\bdy_{\ell}^{\perp}\bdy_{l}a(D)q,\til{\Lam}^{\s}_{\om,j} \tht\rb-\lb [\til{\Lam}_{\om,j}^{\s-2}\bdy_{l},\bdy_{\ell}^{\perp}a(D)q]\bdy_{\ell}\bdy_{l}\tht ,\til{\Lam}_{\om,j}^{\s}\tht\rb=J^{\s}_{3,a}+J^{\s}_{3,b}.
\end{align*}
Similarly, we can express $J_{5}^\s$ as
\begin{align*}
J_{5}^\s&=\lb[\til{\Lam}_{\om,j}^{\s-2}\bdy_{l},\bdy_{\ell}\tht]\bdy_{\ell}^{\perp}\bdy_{l} q,a(D)\til{\Lam}^{\s}_{\om,j} \tht\rb+\lb [\til{\Lam}_{\om p^{1/2},j}^{\s+\frac{\be}2-3}\bdy_{l},\bdy_{\ell}^{\perp} q]\bdy_{\ell}\bdy_{l}\tht ,a(D)^{1/2}\til{\Lam}_{\om,j}^{\s}\tht\rb\notag\\
&=J^{\s}_{5,a}+J^{\s}_{5,b}.
\end{align*}
We will now estimate terms $J_{1}^\s, J_{3,a}^\s, J_{3,b}^\s, J_{5,a}^\s, J_{5,b}^\s$.
Applying \cref{lem:commutator3} with $s=2-\be$, $\eps=\be+1-\s$, $\Gamma=m_1^{\gam}p^{-1}$, and Bernstein's inequality, we obtain
\begin{align}\label{est:J1}
|J_{1}^\s|&\le Cm_1(2^j)^{\gam}2^{(\be+1-\s)j}\Sob{\nabla \tht}{{H}^{\s-1}_\om}\Sob{\til{\Lam}^{\s}_{\om,j}q}{L^2}\Sob{\til{\Lam}^{\s}_{\om,j}\tht}{L^2}\notag\\
&\le Cc_{j}\Sob{\tht}{H^{\s}_\om}\Sob{m_1(D)^{\frac{\gam}{2}}\til{q}}{\Hdot^{1+\be}_\om}\Sob{m_1(D)^{\frac{\gam}{2}}\til{\Lam}^{\s}_{\om,j}\tht}{L^2},
\end{align}
where
\[
c_{j}=\frac{\Sob{m_1(D)^{\frac{\gam}{2}}\til{\Lam}^{\s}_{\om,j}q}{\Hdot^{\be+1-\s}}}{\Sob{m_1(D)^{\frac{\gam}{2}}\til{q}}{\Hdot^{1+\be}_\om}}\in \ell^{2}(\mathbb{Z}),
\]
provided that
\begin{align}\label{cond:Gam:J1}
\sup_{y>0}\left\{\frac{p(y)}{m^{\gam}_1(y)} \left(\int_{0}^{y}\frac{r}{(1+r^2)\om^2(r)}dr\right)^{\frac{1}2},\quad \frac{p_a(y)}{\om(y)m^{\gam}_1(y)}\right\}<\infty,
\end{align}
For $J^\s_{3,a}$, we apply \cref{lem:commutator4} with $r=\s-2$, $(s,\bar{s})=(1,\s-2) $, $(\om_1,\til{\om}_1)=(p^{-1}\om, m_1^{\frac{\gam}{2}}\om)$, $(\om_2, \til{\om}_2)=(\om, m_1^{\frac{\gam}{2}}p^{-1}\om)$, $(\om_3, \til{\om}_3)=(p^{-1}\om, m_1^{\frac{\gam}{2}}\om)$, and $\Gam_1=\Gam_2=\Gam_3=m^{\frac{\gam}2}_1$, we obtain
\begin{align}\label{est:J3a}
|J^{\s}_{3,a}|\le &C_\lam c_{j}m_1(2^j)^{\frac{\gam}{2}}\sum_{\ell,l}\left\{\nrm{\bdy_{\ell}^{\perp}\bdy_{l}p(D)^{-1}a(D)\til{q}}_{H^{1}_\om}\nrm{m_1(D)^{\frac{\gam}{2}}\Lam \bdy_{\ell}\til{\tht}}_{\Hdot^{\s-2}_\om}\right.\notag\\&\hspace{7 em}\left.+\nrm{\Lam \bdy_{\ell}\til{\tht}}_{H^{\s-2}_\om}\nrm{\bdy_{\ell}^{\perp}\bdy_{l}m_1(D)^{\frac{\gam}{2}}p(D)^{-1}a(D)\til{q}}_{\Hdot^{1}_\om}\right. \notag \\
&\hspace{7 em}\left. +\nrm{\bdy_{\ell}^{\perp}\bdy_{l}p(D)^{-1}a(D)\til{q}}_{\Hdot^{1}_\om}\nrm{m_1(D)^{\frac{\gam}{2}}\Lam \bdy_{\ell}\til{\tht}}_{\Hdot^{\s-2}_\om} \right\}\nrm{\til{\Lam}^{\s}_{\om,j}\tht}_{L^2}\notag\\
\le& C_\lam c_{j}\left( \nrm{\til{\tht}}_{H^{\s}_\om}\nrm{m_1(D)^{\frac{\gam}{2}}\til{q}}_{\Hdot^{1+\be}_\om}+\nrm{\til{q}}_{H^{1+\be}_\om}\nrm{m_1(D)^{\frac{\gam}{2}}\til{\tht}}_{\Hdot^{\s}_\om}\right)\nrm{m_1(D)^{\frac{\gam}{2}}\til{\Lam}^{\s}_{\om,j}\tht}_{L^2},
\end{align}
provided that
\begin{align}\label{cond:Gam:Js3a}
\begin{split}
&\sup_{y>0}\frac{p(y)}{{m^{\gam}_1(y)}}\left(\indFn{(-\infty,1)}(\sigma-2)\int_0^1\frac{r^{1-2(\s-2)}}{\om^2(y r)}dr+\indFn{[1,\infty)}(\s-2)\int_{0}^{y}\frac{r}{(1+r^2)\om^{2}(r)}dr\right)^{\frac{1}{2}}<\infty,\\
&\sup_{y>0}\frac{1}{{m^{\gam}_1(y)}}\left(\int_{0}^{y}\frac{rp^2(r)}{(1+r^2)\om^{2}(r)}dr\right)^{\frac{1}{2}},\qquad \sup_{y>0}\frac{p(y)}{\om(y)m^{\gam}_1(y)}<\infty.
\end{split}
\end{align}
For $J^{\s}_{3,b}$, we apply \cref{lem:commutator4} with $r=\s-2$, $(s, \bar{s})=(\s-2,1)$, $(\om_1,\til{\om}_1)=(\om,m_1^{\frac{\gam}{2}}p^{-1}\om)$, $(\om_2, \tilde{\om}_2)=(p^{-1}\om,m_1^{\frac{\gam}{2}}\om)$, $(\om_3, \tilde{\om}_3)=(\om,m_1^{\frac{\gam}{2}}p^{-1}\om)$, and $\Gam_1=\Gam_2=\Gam_3=m^{\frac{\gam}2}_1$, to obtain
\begin{align}\label{est:J3b}
|J^{\s}_{3,b}| \le& C_\lam c_{j}m_1(2^j)^{\frac{\gam}{2}}\sum_{\ell,l}\left\{\nrm{ \bdy_{\ell}\bdy_{l}\til{\tht}}_{H^{\s-2}_\om}\nrm{\bdy_{\ell}^{\perp}m_1(D)^{\frac{\gam}{2}}p(D)^{-1}\Lam a(D)\til{q}}_{\Hdot^{1}_\om}\right.\notag\\&\hspace{7 em}\left.+\nrm{\bdy_{\ell}^{\perp}p(D)^{-1}\Lam a(D)\til{q}}_{H^{1}_\om}\nrm{m_1(D)^{\frac{\gam}{2}} \bdy_{\ell}\bdy_{l}\til{\tht}}_{\Hdot^{\s-2}_\om}\right. \notag\\
&\hspace{7 em} \left. +\nrm{ \bdy_{\ell}\bdy_{l}\til{\tht}}_{\Hdot^{\s-2}_\om}\nrm{\bdy_{\ell}^{\perp}m_1(D)^{\frac{\gam}{2}}p(D)^{-1}\Lam a(D)\til{q}}_{\Hdot^{1}_\om}\right\}\nrm{\til{\Lam}^{\s}_{\om,j}\tht}_{L^2}\notag\\
\le& C_\lam c_{j}\left( \nrm{\til{\tht}}_{H^{\s}_\om}\nrm{m_1(D)^{\frac{\gam}{2}}\til{q}}_{\Hdot^{1+\be}_\om}+\nrm{\til{q}}_{H^{1+\be}_\om}\nrm{m_1(D)^{\frac{\gam}{2}}\til{\tht}}_{\Hdot^{\s}_\om}\right)\nrm{m_1(D)^{\frac{\gam}{2}}\til{\Lam}^{\s}_{\om,j}\tht}_{L^2},
\end{align}
provided that \eqref{cond:Gam:Js3a} holds.
For $J^{\s}_{5,a}$, we apply \cref{lem:commutator4} with
$r=\s-2$, $(s,\bar{s})=(\be-1,\s-2)$, $(\om_1,\til{\om}_1)= (\om_2,\tilde{\om}_2)=(\om_3,\tilde{\om}_3)=(\om, m_1^{\frac{\gam}{2}}\om)$, and $\Gam_1=\Gam_2=\Gam_3=m^{\frac{\gam}2}_1p^{-1}$, to obtain
\begin{align}\label{est:J5a}
|J^{\s}_{5,a}|
\le &C_\lam c_{j}m_1(2^j)^{\frac{\gam}{2}}p(2^j)^{-1}2^{(2-\be)j}\sum_{\ell,l}\left\{\nrm{\bdy_{\ell}^{\perp}\bdy_{l}\til{q}}_{H^{\be-1}_\om}\nrm{\bdy_{\ell}m_1(D)^{\frac{\gam}{2}}\Lam\til{\tht}}_{\Hdot^{\s-2}_\om}\right. \notag\\
&\hspace{13 em}\left.+\nrm{\bdy_{\ell}\Lam \til{\tht}}_{H^{\s-2}_\om}\nrm{m_1(D)^{\frac{\gam}{2}}\bdy_{\ell}^{\perp}\bdy_{l}\til{q}}_{\Hdot^{\be-1}_\om}\right. \notag\\
&\hspace{13 em}\left. +\nrm{\bdy_{\ell}^{\perp}\bdy_{l}\til{q}}_{\Hdot^{\be-1}}\nrm{\bdy_{\ell}m_1(D)^{\frac{\gam}{2}}\Lam \til{\tht}}_{\Hdot^{\s-2}_\om}\right\}\nrm{\til{\Lam}^{\s}_{\om,j}a(D)\tht}_{L^2}\notag\\
\le& C_\lam c_{j}\left( \nrm{\til{\tht}}_{H^{\s}_\om}\nrm{m_1(D)^{\frac{\gam}{2}}\til{q}}_{\Hdot^{1+\be}_\om}+\nrm{\til{q}}_{H^{1+\be}_\om}\nrm{m_1(D)^{\frac{\gam}{2}}\til{\tht}}_{\Hdot^{\s}_\om}\right)\nrm{m_1(D)^{\frac{\gam}{2}}\til{\Lam}^{\s}_{\om,j}\tht}_{L^2},
\end{align}
provided that
\begin{align}\label{cond:Gam:Js5a}
\begin{split}
&\sup_{y>0}\frac{p(y)}{{m^{\gam}_1(y)}}\left(\indFn{(1,2)}(\be)\int_0^1\frac{r^{1-2(\be-1)}}{\om^2(y r)}dr+\indFn{\{2\}}(\be)\int_{0}^{y}\frac{r}{(1+r^2)\om^{2}(r)}dr\right)^{\frac{1}{2}}<\infty,\\
&\sup_{y>0}\frac{p(y)}{{m^{\gam}_1(y)}}\left(\indFn{(-\infty,1)}(\sigma-2)\int_0^1\frac{r^{1-2(\s-2)}}{\om^2(y r)}dr+\indFn{[1,\infty)}(\s-2)\int_{0}^{y}\frac{r}{(1+r^2)\om^{2}(r)}dr\right)^{\frac{1}{2}}<\infty,\\
&\sup_{y>0}\frac{p(y)}{\om(y)m^{\gam}_1(y)}<\infty.
\end{split}
\end{align}
For $J^{\s}_{5,b}$, we apply \cref{lem:commutator4} with $r=\s+\be/2-3$, $(s,\bar{s})=(\s-2,\be-1)$, $(\om_1,\til{\om}_1)$ $=$ $(\om_2,\tilde{\om}_2)$ $=$ $(\om_3,\tilde{\om}_3)$ $=$ $(\om, m_1^{\frac{\gam}{2}}\om)$, and $\Gam_1=\Gam_2=\Gam_3=m^{\frac{\gam}2}_1p^{-\frac{1}{2}}$, to obtain
\begin{align}\label{est:J5b}
|J^{\s}_{5,b}|\le& C_\lam c_j m_1(2^j)^{\frac{\gam}{2}}p(2^j)^{-\frac{1}{2}}2^{(1-\frac{\be}{2})j}\sum_{\ell,l}\left\{ \nrm{\bdy_{\ell}\bdy_{l}\til{\tht}}_{H^{\s-2}_\om}\nrm{m_1(D)^{\frac{\gam}{2}}\bdy_{\ell}^{\perp}\Lam \til{q}}_{\Hdot^{\be-1}_\om}\right.\notag \\
&\hspace{14 em}\left.+\nrm{\bdy_{\ell}^{\perp}\Lam \til{q}}_{H^{\be-1}_\om}\nrm{m_1(D)^{\frac{\gam}{2}}\bdy_{\ell}\bdy_{l}\til{\tht}}_{\Hdot^{\s-2}_\om}\right.\notag \\
&\hspace{14 em}\left.+\nrm{\bdy_{\ell}\bdy_{l}\til{\tht}}_{\Hdot^{\s-2}_\om}\nrm{m_1(D)^{\frac{\gam}{2}}\bdy_{\ell}^{\perp}\Lam \til{q}}_{\Hdot^{\be-1}_\om}\right\}\nrm{a(D)^{1/2}\til{\Lam}^{\s}_{\om,j}\tht}_{L^2}\notag\\
\le& C_\lam c_{j}\left( \nrm{\til{\tht}}_{H^{\s}_\om}\nrm{m_1(D)^{\frac{\gam}{2}}\til{q
}}_{\Hdot^{1+\be}_\om}+\nrm{\til{q}}_{H^{1+\be}_\om}\nrm{m_1(D)^{\frac{\gam}{2}}\til{\tht}}_{\Hdot^{\s}_\om}\right)\nrm{m_1(D)^{\frac{\gam}{2}}\til{\Lam}^{\s}_{\om,j}\tht}_{L^2},
\end{align}
provided that
\begin{align}\label{cond:Gam:Js5b}
\begin{split}
&\sup_{y>0}\frac{p(y)}{{m^{\gam}_1(y)}}\left(\indFn{(-\infty,1)}(\sigma-2)\int_0^1\frac{r^{1-2(\s-2)}}{\om^2(y r)}dr+\indFn{[1,\infty)}(\s-2)\int_{0}^{y}\frac{r}{(1+r^2)\om^{2}(r)}dr\right)^{\frac{1}{2}}<\infty,\\
&\sup_{y>0}\frac{p(y)}{{m^{\gam}_1(y)}}\left(\indFn{(-\infty,1)}(\be-1)\int_0^1\frac{r^{1-2(\be-1)}}{\om^2(y r)}dr+\indFn{[1,\infty)}(\be-1)\int_{0}^{y}\frac{r}{(1+r^2)\om^{2}(r)}dr\right)^{\frac{1}{2}}<\infty,\\
& \sup_{y>0}\frac{p(y)}{\om(y)m^{\gam}_1(y)}<\infty.
\end{split}
\end{align}
\subsubsection*{Concluding Estimates for Case 2: $\be \in (1,2]$} As in Case 1, we observe that the conditions stated in \eqref{cond:Gam:I1s}, \eqref{cond:Gam:I3}, \eqref{cond:Gam:J1}, \eqref{cond:Gam:Js3a}, \eqref{cond:Gam:Js5a}, \eqref{cond:Gam:Js5b} can be reduced to
\begin{align}\label{cond:Gam:J:final}
\begin{split}
\sup_{y>0}\left\{\frac{1}{m_1^\gam(y)}\left(\int_{0}^{y}\frac{r(p^2(y)+p^2(r))}{(1+r^2)\om^{2}(r)}dr\right)^{\frac{1}{2}},\quad \frac{p_a(y)\om_b(y)}{m^{\gam}_1(y)}\right\}<\infty.
\end{split}
\end{align}
Upon returning to \eqref{est:balance:Hsigma:final}, we may now apply either \eqref{est:I1}, \eqref{est:I3} or \eqref{est:J1}, \eqref{est:J3a}, \eqref{est:J3b}, \eqref{est:J5a}, \eqref{est:J5b}, then sum in $j$, while invoking \eqref{eq:Sob:Bes}, \eqref{eq:Sob:equiv}, to obtain
\begin{align}\label{est:Hs:b1}
\frac{d}{dt}\Sob{\til{\tht}}{\Hdot^{\s}_\om}^2+\frac{7}4\Sob{m_1(D)^{\frac{1}2}\til{\tht}}{\Hdot^\s_\om}^{2}&\leq C_\lam \left(\nrm{\til{\tht}}_{H^{\s}_\om}\nrm{m_1(D)^{\frac{\gam}{2}}\til{q
}}_{\Hdot^{1+\be}_\om}+\nrm{\til{q}}_{H^{1+\be}_\om}\nrm{m_1(D)^{\frac{\gam}{2}}\til{\tht}}_{\Hdot^{\s}_\om }\right)\nrm{m_1(D)^{\frac{\gam}{2}}\til{\tht}}_{\Hdot^\s_\om}\notag\\
&\quad+C\left(\Sob{m_1(D)^{-\frac{1}{2}}\til{G}}{\Hdot^\s_\om}^2+\Sob{\til{\tht}}{\Hdot^\s_\om}^2\right),
\end{align}
for all $\s\in[1,1+\be]$. We lastly combine \eqref{est:Hs:b1} with \eqref{est:L2:b} and \cref{cor:norm:equiv} to deduce
\begin{align}\label{est:Hs:b1:final}
\frac{d}{dt}&\left(\Sob{\til{\tht}}{\Hdot^{\s}_\om}^2+\Sob{\tht}{L^2}^2\right)+\frac{7}4\left(\Sob{m_1(D)^{\frac{1}2}\til{\tht}}{\Hdot^\s_\om}^{2}+\Sob{m_1(D)^{\frac{1}2}\tht}{L^2}^2\right)\notag\\
&\leq C \nrm{\til{\tht}}_{H^{\s}_\om}\nrm{m_1(D)^{\frac{\gam}{2}}\til{q
}}_{\Hdot^{1+\be}_\om}\nrm{m_1(D)^{\frac{\gam}{2}}\til{\tht}}_{\Hdot^\s_\om}+C\nrm{\til{q}}_{H^{1+\be}_\om}\left(\nrm{m_1(D)^{\frac{\gam}{2}}\til{\tht}}_{\Hdot^{\s}_\om}^2+\Sob{m_1(D)^{\frac{\gam}2}{\tht}}{L^2}^2\right)\notag\\
&\quad+C\left(\Sob{m_1(D)^{-\frac{1}{2}}\til{G}}{H^\s_\om}^2+\Sob{\til{\tht}}{H^\s_\om}^2\right),
\end{align}
where we used the facts that $\Sob{\tht}{L^2}\leq \Sob{m_1(D)^{\frac{\gam}2}\tht}{L^2}$ and $\Sob{\tht}{H^\eps_\om}\leq\Sob{m_1(D)^{\frac{\gam}2}\til{\tht}}{H^\s_\om}$, for $\eps$ sufficiently small.
\subsection*{Summary of estimates in $\dot{E}^{\lam}_{\nu,\s,\om}$:}
Given $\be\in[0,2]$, suppose $\s$ satisfies \eqref{sigma:range} (omitting the case $\be=0$, $\lam=0$, $\s=-1$, for now, since it is assumed that $\lam_1>0$). Observe that \eqref{cond:Gam:J:final} implies \eqref{cond:Gam:I0:final}, so that we ultimately reduce these conditions to
\begin{align}\label{cond:summary}
\begin{split}
\sup_{y>0}\left\{\frac{1}{m_1^\gam(y)}\left(\int_{0}^{y}\frac{r(p^2(y)+p^2(r))}{(1+r^2)\om^{2}(r)}dr\right)^{\frac{1}{2}},\quad \frac{p_a(y)\om_b(y)}{m^{\gam}_1(y)}\right\}<\infty.
\end{split}
\end{align}
Then for $\be \in [0,1]$, we have
\begin{align}\label{est:Hs:summary:1}
\frac{d}{dt}\Sob{\til{\tht}}{\Hdot^{\s}_\om\cap L^2_{\om}}^2&+\frac{7}4\Sob{m_1(D)^{\frac{1}2}\til{\tht}}{\Hdot^\s_\om\cap L^2_{\om}}^{2}\notag\\
&\quad\leq C_\lam \left(\nrm{\til{\tht}}_{\Hdot^{\s}_\om\cap L^2_{\om}}\nrm{m_1(D)^{\frac{\gam}{2}}\til{q
}}_{\Hdot^{1+\be}_\om}+\nrm{\til{q}}_{H^{1+\be}_\om}\nrm{m_1(D)^{\frac{\gam}{2}}\til{\tht}}_{\Hdot^{\s}_\om \cap L^2_{\om}}\right)\nrm{m_1(D)^{\frac{\gam}{2}}\til{\tht}}_{\Hdot^\s_\om\cap L^2_{\om}}\notag\\
&\qquad+C\left(\Sob{m_1(D)^{-\frac{1}{2}}\til{G}}{\Hdot^\s_\om\cap L^2_{\om}}^2+\Sob{\til{\tht}}{\Hdot^\s_\om\cap L^2_{\om}}^2\right),
\end{align}
and for $\be\in(1,2]$, we have
\begin{align}\label{est:Hs:summary:2}
\frac{d}{dt}&\left(\Sob{\til{\tht}}{\Hdot^{\s}_\om}^2+\Sob{\tht}{L^2}^2\right)+\frac{7}4\left(\Sob{m_1(D)^{\frac{1}2}\til{\tht}}{\Hdot^\s_\om}^{2}+\Sob{m_1(D)^{\frac{1}2}\tht}{L^2}^2\right)\notag\\
&\leq C \nrm{\til{\tht}}_{H^{\s}_\om}\nrm{m_1(D)^{\frac{\gam}{2}}\til{q
}}_{\Hdot^{1+\be}_\om}\nrm{m_1(D)^{\frac{\gam}{2}}\til{\tht}}_{\Hdot^\s_\om}+C\nrm{\til{q}}_{H^{1+\be}_\om}\left(\nrm{m_1(D)^{\frac{\gam}{2}}\til{\tht}}_{\Hdot^{\s}_\om}^2+\Sob{m_1(D)^{\frac{\gam}2}{\tht}}{L^2}^2\right)\notag\\
&\quad+C\left(\Sob{m_1(D)^{-\frac{1}{2}}\til{G}}{H^\s_\om}^2+\Sob{\til{\tht}}{H^\s_\om}^2\right),
\end{align}
When $\be\in[0,1]$, observe that by interpolation \eqref{est:interpolation:weighted} and Young's inequality we obtain
\begin{align}
C&\nrm{\til{\tht}}_{\Hdot^{\s}_{\om}\cap L^2_{\om}}\nrm{m_1(D)^{\frac{\gamma}{2}}\til{q}}_{\Hdot^{1+\be}_\om}\nrm{m_1(D)^{\frac{\gamma}{2}}\til{\tht}}_{\Hdot^{\s}_{\om}\cap L^2_{\om}}\label{est:Hs:a3}\\
&\leq \nrm{m_1(D)^{\frac{\gamma}{2}}\til{q}}_{\Hdot^{1+\be}_\om}\nrm{\til{\tht}}_{\Hdot^{\s}_{\om}\cap L^2_{\om}}^{2-{\gam}}\nrm{m_1(D)^{\frac{1}{2}}\til{\tht}}_{\Hdot^{\s}_{\om}\cap L^2_{\om}}^{{\gam}}\notag\\
&\leq \frac{1}8\Sob{m(D)^{\frac{1}2}\til{\tht}}{\Hdot^{\s}_{\om}\cap L^2_{\om}}^2+C\nrm{m_1(D)^{\frac{\gam}2}\til{q}}_{\Hdot^{1+\be}_{\om}}^{\frac{2}{2-\gam}}\nrm{\til{\tht}}_{\Hdot^{\s}_{\om}\cap L^2_{\om}}^2.\label{est:Hs:a2}
\end{align}
and
\begin{align}
C\nrm{\til{q}}_{H^{1+\be}_\om} \nrm{m_1(D)^{\frac{\gam}{2}}\til{\tht}}_{\Hdot^{\s}_{\om}\cap L^2_{\om}}^2&\leq C\nrm{\til{q}}_{H^{1+\be}_\om}\Sob{m_1(D)^{\frac{1}2}\til{\tht}}{\Hdot^{\s}_{\om}\cap L^2_{\om}}^{{2\gam}}\Sob{\til{\tht}}{\Hdot^{\s}_{\om}\cap L^2_{\om}}^{2-2{\gam}}\label{est:Hs:a4}\\
&\leq \frac{1}8\Sob{m_1(D)^{\frac{1}2}\til{\tht}}{\Hdot^{\s}_{\om}\cap L^2_{\om}}^2+C\Sob{\til{q}}{H^{1+\be}_\om}^{\frac{1}{1-\gam}}\Sob{\til{\tht}}{\Hdot^{\s}_{\om}\cap L^2_{\om}}^2,\label{est:Hs:a1}
\end{align}
Applying \eqref{est:Hs:a1} and \eqref{est:Hs:a2} in \eqref{est:Hs:summary:1}, we arrive at
\begin{align}\label{est:Hs:final}
\frac{d}{dt}\Sob{\tilde{\tht}}{\Hdot^{\s}_{\om}\cap L^2_{\om}}^2&+\frac{3}2\Sob{m_1(D)^{\frac{1}2}\til{\tht}}{\Hdot^{\s}_{\om}\cap L^2_{\om}}^{2}\notag\\
&\quad\leq C \left(1+\Sob{\til{q}}{H^{1+\be}_\om}^{\frac{1}{1-\gam}}+\Sob{m_1(D)^{\frac{\gam}2}\til{q}}{\Hdot^{1+\be}_\om}^{\frac{2}{2-\gam}}\right)\Sob{\tilde{\tht}}{\Hdot^{\s}_{\om}\cap L^2_{\om}}^2+C\Sob{m_1(D)^{-\frac{1}{2}}\til{G}}{\Hdot^{\s}_{\om}\cap L^2_{\om}}^2.
\end{align}
An application of Gronwall's inequality then yields
\begin{align}\label{est:Gev:final}
\sup_{0\leq t \leq T}\Bigg{(}\Sob{\tilde{\tht}(t)}{\Hdot^{\s}_{\om}\cap L^2_{\om}}^2&+\int_0^t \Sob{m_1(D)^{\frac{1}2}\tilde{\tht}(s)}{\Hdot^{\s}_{\om}\cap L^2_{\om}}^2ds\Bigg{)}\notag\\
&\quad\le \tilde{C}_T(\be)\left(\Sob{\tht_0}{\Hdot^{\s}_{\om}\cap L^2_{\om}}^2+\int_0^T\Sob{m_1(D)^{-\frac{1}2}\tilde{G}(s)}{\Hdot^{\s}_{\om}\cap L^2_{\om}}^2ds\right),
\end{align}
where
\begin{align}\label{def:Gev:constant}
\tilde{C}_T(\be)=\exp\left(C\int_0^T\left(1+\Sob{\tilde{q}(t)}{H^{1+\be}_\om}^{\frac{1}{1-\gam}}+\Sob{m_1(D)^{\frac{\gam}2}\tilde{q}(t)}{\Hdot^{1+\be}_\om}^{\frac{2}{2-\gam}}\right)dt\right),
\end{align}
for some constant $C$ depending on $\be$.
When $\be\in(1,2]$, we may estimate the right-hand side of \eqref{est:Hs:summary:2} in a similar fashion to \eqref{est:Hs:summary:1}, except that we additionally invoke \cref{cor:norm:equiv}, in order to also deduce \eqref{est:Gev:final}.
\subsection{A priori estimates in $\Hdot^{\s}_{\om}$}
In order to obtain estimates in Sobolev spaces, we suppress the smoothing multiplier in \eqref{est:Hs:final} by formally setting $\lam\equiv0$; this evaluation is justified due to the way in which the constants depend on $\lam$ in the above estimates. Lastly, we recall that we have yet to treat the case $\be=0$, $\s=-1$, $\lam\equiv0$; this will also be done in this section.
Upon setting $\lam\equiv0$ in \eqref{est:Gev:final} we obtain
\begin{align}\label{est:Hs:gronwall}
\sup_{0\leq t \leq T}\Bigg{(}\Sob{{\tht}(t)}{\Hdot^{\s}_{\om}\cap L^2_{\om}}^2&+\int_0^t \Sob{m_1(D)^{\frac{1}2}\tht(s)}{\Hdot^{\s}_{\om}\cap L^2_{\om}}^2ds\Bigg{)}\notag\\
&\quad\le C_T(\be)\left(\Sob{\tht_0}{\Hdot^{\s}_{\om}\cap L^2_{\om}}^2+\int_0^T\Sob{m_1(D)^{-\frac{1}2}G(s)}{\Hdot^{\s}_{\om}\cap L^2_{\om}}^2ds\right),
\end{align}
where
\begin{align}\label{def:Hs:constant}
C_T(\be)=\exp\left(C\int_0^T\left(1+\Sob{q(t)}{H^{1+\be}_\om}^{\frac{1}{1-\gam}}+\Sob{m_1(D)^{\frac{\gam}2}q(t)}{\Hdot^{1+\be}_\om}^{\frac{2}{2-\gam}}\right)dt\right),
\end{align}
for some constant $C$ depending on $\be$.
\subsubsection*{Case $\be=0$, $\s=-1$, $\lam\equiv0$} Since $\lam\equiv0$, we may drop the tilde notation from \eqref{eq:balance:Hsigma}. Furthermore, observe that from \eqref{eq:balance:Hsigma} we have
\[
I^{-1}=\lb m_1(D)^{-\frac{\gam}2}{\Lam}^{-1}_{\om,j}\nabla\cdotp(v \tht),m_1(D)^{\frac{\gam}2}{\Lam}^{-1}_{\om,j}\tht\rb.
\]
Then by Bernstein's inequalities and \eqref{est:omega}, we have
\begin{align}\notag
|I^{-1}|&\leq C\om(2^j)m_1(2^j)^{-\frac{\gam}2}\Sob{\lpj(v\tht)}{L^2}\Sob{m_1(D)^{\frac{\gam}2}\Lam_{\om,j}^{-1}\tht}{L^2}
\end{align}
Applying \cref{thm:prod} with $(s,\bar{s})=(1,0)$, $(\om_1,\til{\om}_1)=(p^{-1}\om,m_1^{\frac{\gam}{2}}\om)$, $(\om_2, \til{\om}_2)=(\om, p^{-1}m_1^{\frac{\gam}{2}})$ ,
$(\om_3, \tilde{\om}_3)=(p^{-1}\om, m_1^{\frac{\gam}{2}}\om)$, and $\Gam_1=\Gam_2=\Gam_3=m_1^{\frac{\gam}2}$ we obtain
\begin{align}\label{est:I:Hminus1}
|I^{-1}|&\le Cc_j\left(\Sob{p(D)^{-1}v}{H^1_{\om}}\Sob{m_1(D)^{\frac{\gam}{2}}\tht}{L^2_{\om}}+\Sob{\tht}{L^2_{\om}}\Sob{p(D)^{-1}m_1(D)^{\frac{\gam}{2}}v}{\Hdot^1_\om}\right)\Sob{m_1(D)^{\frac{\gam}2}{\Lam}^{-1}_{\om, j} \tht}{L^2},
\end{align}
provided that
\begin{align}\label{cond:Gam:I:minus1}
\begin{split}
\sup_{y>0}\frac{1}{{m^{\gam}_1(y)}}\left(\int_{0}^{y}\frac{rp^2(r)}{(1+r^2)\om^{2}(r)}dr\right)^{\frac{1}{2}},\quad \sup_{y>0}\frac{p(y)}{{m^{\gam}_1(y)}}\left(\int_0^1\frac{r}{\om^2(y r)}dr\right)^{\frac{1}{2}},\quad \sup_{y>0}\frac{p(y)}{\om(y)m^{\gam}_1(y)}<\infty,
\end{split}
\end{align}
holds. As before, observe that \eqref{cond:Gam:I0:final} implies \eqref{cond:Gam:I:minus1}.
Recall that $v$ is given by \eqref{def:v}, so that
\[
\Sob{p(D)^{-1}v}{H^1_{\om}}\leq C\Sob{q}{\Hdot^{-1}_{\om}\cap L^2_\om},\quad \Sob{p(D)^{-1}m_1(D)^{\frac{\gam}{2}}v}{\Hdot^1_{\om}}\leq C\Sob{m_1(D)^{\frac{\gam}{2}}q}{L^2_{\om}}.
\]
Upon returning to \eqref{est:Hs:a} with $\lam_1=0$ and $\s=0$, applying \eqref{est:I:Hminus1}, summing in $j$, and invoking \eqref{eq:Sob:Bes}, \eqref{eq:Sob:equiv}, we obtain
\begin{align}\label{est:Hminus1}
\frac{d}{dt}\Sob{\tht}{\Hdot^{-1}_{\om}\cap L^2_{\om}}^2+\frac{7}4\Sob{m(D)^{\frac{1}2}\tht}{\Hdot^{-1}_{\om}\cap L^2_{\om}}^{2}&\leq C\left(\Sob{q}{\Hdot^{-1}_{\om}\cap \dot{H}^1_\om}\Sob{m_1(D)^{\frac{\gam}{2}}\tht}{L^2_{\om}}+\Sob{\tht}{L^2_{\om}}\Sob{m_1(D)^{\frac{\gam}{2}}q}{H^1_{\om}}\right)\Sob{m_1(D)^{\frac{\gam}2}\tht}{\Hdot^{-1}_{\om}\cap L^2_\om}\notag\\
&\quad +C\left(\Sob{m_1(D)^{-\frac{1}{2}}G}{\Hdot^{-1}_\om \cap L^2_\om}^2+\Sob{\tht}{\Hdot^{-1}_\om \cap L^2_\om}^2\right).
\end{align}
Using Plancherel's theorem, \eqref{est:interpolation:weighted}, and Young's inequality we see that
\begin{align}
C\Sob{q}{\Hdot^{-1}_{\om}\cap \dot{H}^1_\om}\Sob{m_1(D)^{\frac{\gam}{2}}\tht}{L^2_{\om}}\Sob{m_1(D)^{\frac{\gam}2}\tht}{\Hdot^{-1}_{\om}\cap L^2_\om}&\leq C\Sob{q}{\Hdot^{-1}_{\om}\cap \dot{H}^1_\om}\Sob{m_1(D)^{\frac{1}2}\tht}{\Hdot^{-1}_\om\cap L^2_{\om}}^{2\gam}\Sob{\tht}{\Hdot^{-1}_\om\cap L^2_{\om}}^{2(1-\gam)}\notag\\
&\leq \frac{1}8\Sob{m_1(D)^{\frac{1}2}\tht}{\Hdot^{-1}_\om\cap L^2_{\om}}^2+C\Sob{q}{\Hdot^{-1}_{\om}\cap \dot{H}^1_\om}^{\frac{1}{1-\gam}}\Sob{\tht}{\Hdot^{-1}_\om\cap L^2_{\om}}^2\label{est:Hminus1:nonlinear:a}\\
C\Sob{\tht}{L^2_{\om}}\Sob{m_1(D)^{\frac{\gam}{2}}q}{H^1_{\om}}\Sob{m_1(D)^{\frac{\gam}2}\tht}{\Hdot^{-1}_{\om}\cap L^2_\om}&\leq C\Sob{m_1(D)^{\frac{\gam}{2}}q}{H^1_{\om}}\Sob{m_1(D)^{\frac{1}2}\tht}{\Hdot^{-1}_\om\cap L^2_{\om}}^{{\gam}}\Sob{\tht}{\Hdot^{-1}_\om\cap L^2_{\om}}^{{2-\gam}}\notag\\
&\leq \frac{1}8\Sob{m_1(D)^{\frac{1}2}\tht}{\Hdot^{-1}_\om\cap L^2_{\om}}^2+C\Sob{m_1(D)^{\frac{\gam}{2}}q}{H^1_{\om}}^{\frac{2}{2-\gam}}\Sob{\tht}{\Hdot^{-1}_\om\cap L^2_{\om}}^2\label{est:Hminus1:nonlinear:b}
\end{align}
Now we return to \eqref{est:Hminus1}, then apply \eqref{est:Hminus1:nonlinear:a}, \eqref{est:Hminus1:nonlinear:b} to deduce
\begin{align}\label{est:Hminus1:final}
\frac{d}{dt}\Sob{\tht}{\Hdot^{-1}_{\om}\cap L^2_{\om}}^2&+\frac{3}2\Sob{m(D)^{\frac{1}2}\tht}{\Hdot^{-1}_{\om}\cap L^2_{\om}}^{2}\notag\\
&\quad\leq C\left(1+\Sob{q}{ \Hdot^{-1}_\om\cap \Hdot^1_\om}^{\frac{1}{1-\gam}}+\Sob{m_1(D)^{\frac{\gam}2}q}{H^1_\om}^{\frac{2}{2-\gam}}\right)\Sob{\tht}{\Hdot^{-1}_{\om}\cap L^2_{\om}}^2+C\Sob{m_1(D)^{-\frac{1}{2}}G}{\Hdot^{-1}_\om\cap L^2_{\om}}^2.
\end{align}
An application of Gronwall's inequality now yields
\begin{align}\label{est:Hminus1:gronwall}
\sup_{0\leq t\leq T}\Bigg{(}\Sob{\tht(t)}{\Hdot^{-1}_\om\cap L^2_{\om}}^2&+\int_0^t\Sob{m(D)^{\frac{1}{2}}\tht(s)}{\Hdot^{-1}_\om\cap L^2_{\om}}^{2}ds\Bigg{)}\notag\\
&\quad\leq {C}_T(0^-)\left(\Sob{\tht_0}{\Hdot^{-1}_\om\cap L^2_{\om}}^2+\int_0^T\Sob{m_1(D)^{-\frac{1}2}G(s)}{\Hdot^{-1}_\om\cap L^2_{\om}}^2ds\right),
\end{align}
where
\begin{align}\label{def:Hminus1:constant}
{C}_T(0^-)=\exp\left(C\int_0^T\left(1+\Sob{q(s)}{ \Hdot^{-1}_\om\cap \Hdot^1_\om}^{\frac{1}{1-\gam}}+\Sob{m_1(D)^{\frac{\gam}2}q(s)}{H^1_\om}^{\frac{2}{2-\gam}}\right)ds\right),
\end{align}
for some constant $C$.
\subsection{Global existence and uniqueness for the protean system}\label{subsect:apriori:summary}
From the apriori estimates developed previously and a standard artificial viscosity approximation, we obtain the following theorem for the well-posedness of \eqref{eq:mod:claw}. Let us denote by
\begin{align}\label{def:p0}
p_0=\frac{2}{2-\gam}.
\end{align}
\begin{Thm}\label{thm:modclaw:wellposed}
Let $\be\in [0,2]$ and $\s\geq-1$ satisfy \eqref{sigma:range}.
Let $\om \in \mathscr{M}_W$, $p\in\mathscr{M}_C$, $m\in\mathscr{M}_D$ be given such that \eqref{cond:summary} holds for some $\gam\in (0,1)$. Given $T>0$, suppose that
\begin{align}\label{cond:q:G}
q\in L^{\infty}(0,T;H^{1+\be}_\om),\quad m_1(D)^{\frac{\gam}{2}}q\in L^{p_0}(0,T; \Hdot^{1+\be}_\om),\quad m_1(D)^{-\frac{\gam}2}G \in L^{2}(0,T;\Hdot^{\s}_\om\cap L^2_{\om}),
\end{align}
Then for any $\tht_0\in \Hdot^{\s}_\om\cap L^2_{\om}$, there exists a unique solution, $\tht$, of \eqref{eq:mod:claw} satisfying \eqref{est:Hs:gronwall} and
\[
\theta\in C([0,T];\Hdot^{\s}_\om\cap L^2_{\om}),\quad m_1(D)^{\frac{1}2}\tht\in L^2(0,T;\Hdot^{\s}_{\om}\cap L^2_{\om}).
\]
Moreover, given any $\nu(D)\in\mathscr{M}_S(m)$, the unique solution satisfies \eqref{est:Gev:final} provided that
\[
\int_0^T\left(\Sob{E^{\lam_1t}_\nu q(t)}{H^{1+\be}_\om}^{\frac{1}{1-\gam}}+\Sob{m_1(D)^{\frac{\gam}2}E^{\lam_1t}_\nu{q}(t)}{\Hdot^{1+\be}_\om}^{\frac{2}{2-\gam}}\right)dt<\infty,
\]
where $E^\lam_\nu$ is defined by \eqref{def:Elamnu}.
\end{Thm}
The proof of \cref{thm:modclaw:wellposed} is provided in \cref{app:well:posed}.
\subsection{Stability of the protean system}\label{sect:stab:apriori} In this section, we will establish continuity properties of \eqref{eq:mod:claw} with respect to its datum. This will rely on having access to suitable stability-type estimates. Ultimately, the results of this section are part of the development of the uniqueness of solutions and continuity of the solution map of \eqref{eq:dgsqg}.
The standing assumption of this section will be that $p\in\mathscr{M}_C$, $\om \in \mathscr{M}_W$, $m\in\mathscr{M}_D$, $\nu\in \mathscr{M}_R(m)$ are given and satisfy \eqref{cond:summary}. The main result of this section is then stated in the following theorem.
\begin{Thm}\label{thm:stab}
Let $\be \in [0,2]$ and $T>0$ be fixed. Suppose that sequences $\{\tht^n\}_n, \{q^n\}_n, \{G^n\}_n$ and functions $\tht^\infty, q^\infty, G^\infty$ are given such that
\begin{align}\label{cond:Q0}
&Q_\be:=\notag\\
&\sup_{n\in\NN\cup\{\infty\}}
\begin{cases}
\Sob{\tht^n_0}{L^2_{\om}\cap\Hdot^{-1}_\om}+\Sob{q^{n}}{L^\infty _T(H^1_\om\cap\Hdot^{-1}_\om)}+\Sob{m_1(D)^{\frac{\gam}{2}}q^{n}}{L^{p_0}_{T}\Hdot^{1}_{\om}}+\Sob{m_1(D)^{-\frac{1}{2}}G^n}{L^2_T(L^2_{\om}\cap\Hdot^{-1}_\om)},&\be=0\\
\Sob{\tht^n_0}{H^{\be}_\om}+\Sob{q^{n}}{L^\infty _T H^{1+\be}_\om}+\Sob{m_1(D)^{\frac{\gam}{2}}q^{n}}{L^{p_0}_{T}\Hdot^{1+\be}_{\om}}+\Sob{m_1(D)^{-\frac{1}{2}}G^n}{L^2_TH^{\be}_\om},&\be\in(0,1],\\
\Sob{\tht^n_0}{H^{\be}_\om}+\Sob{q^{n}}{L^\infty_T H^{1+\be}_\om}+\Sob{m_1(D)^{\frac{\gam}{2}}q^{n}}{L^{p_0}_{T}\Hdot^{1+\be}_{\om}}+\Sob{m_1(D)^{-\frac{1}{2}}G^n}{L^2_TH^{\be}_\om},&\be\in(1,2].
\end{cases}
\end{align}
is finite. Moreover, suppose that
\begin{align}\label{cond:Z}
&Z_\be^{n}:=\notag\\
& \begin{cases}
\Sob{\tht^n_0-\tht^\infty_0}{L^2_{\om}\cap\Hdot^{-1}_\om}^2+\Sob{q^n-q^\infty}{L^\infty_T (L^2_{\om}\cap\Hdot^{-1}_\om)}^2+\Sob{m_1(D)^{\frac{\gam}{2}}(q^{n}-q^\infty)}{L^{2}_{T}L^2_{\om}}^2\\
\hspace{22 em}+\Sob{m_1(D)^{-\frac{1}{2}}(G^n-G^\infty)}{L^2_T(L^2_{\om}\cap\Hdot^{-1}_\om)}^2,&\be=0\\
\Sob{\tht_0^n-\tht_0^\infty}{L^2_{\om}}^2+\Sob{q^n-q^\infty}{L^\infty_T L^2_{\om}}^2+\Sob{m_1(D)^{\frac{\gam}{2}}(q^{n}-q^\infty)}{L^{2}_{T}L^2_{\om}}^2+\Sob{m_1(D)^{-\frac{1}{2}}(G^n-G^\infty)}{L^2_TL^2_{\om}}^2,&\be\in(0,1),\\
\Sob{\tht^n_0-\tht^\infty_0}{H^\be_\om}^2+\Sob{q^n-q^\infty}{L^\infty_TH^{\be}_\om}^2+\Sob{m_1(D)^{\frac{\gam}{2}}(q^{n}-q^\infty)}{L^{2}_{T}H^{\be}_{\om}}^2+\Sob{m_1(D)^{-\frac{1}{2}}(G^n-G^\infty)}{L^2_TH^\be_\om}^2,&\be\in[1,2]
\end{cases}
\end{align}
converges to $0$ as $n\goesto\infty$. For all $n\in\NN\cup\{\infty\}$, let $\tht^n$ denote the unique solution guaranteed by \cref{thm:modclaw:wellposed} of the initial value problem
\begin{align}\label{eq:modclaw:n}
\partial_{t}\theta^{n}+ m(D)\tht^{n}+\Div F_{q^{n}}(\tht^{n})=G^n,\quad\tht^{n}(0,x)=\tht^n_0(x),
\end{align}
Then
\begin{align}\label{eq:stab:a}
\lim_{n\goesto\infty}\left(\Sob{\tht^n-\tht^\infty}{L^\infty_T(L^2_{\om}\cap\Hdot^{-1}_\om)}^2+\Sob{m(D)^{\frac{1}{2}}(\tht^n-\tht^\infty)}{L^2_T(L^2_{\om}\cap\Hdot^{-1}_\om)}^2\right)=0,
\end{align}
when $\be=0$, and
\begin{align}\label{eq:stab:b}
\lim_{n\goesto\infty}\left(\Sob{\tht^n-\tht^\infty}{L^\infty_TH^{\be}_\om}^2+\Sob{m(D)^{\frac{1}{2}}(\tht^n-\tht^\infty)}{L^2_TH^{\be}_\om}^2\right)=0,
\end{align}
when $\be\in(0,2]$.
\end{Thm}
The proof of \cref{thm:stab} will rely on the following stability-type estimates.
\begin{Prop}\label{prop:stab}
Let $\be \in [0,2]$ and $T>0$. Suppose that $\{\tht^n\}_n, \{q^n\}_n, \{G^n\}_n$ and $\tht^\infty, q^\infty, G^\infty$ be such that
\begin{align}\label{cond:Qbeta}
&Q_\be^*:=\\
&\sup_{n\in\NN\cup\{\infty\}}
\begin{cases}
\Sob{\tht^n_0}{H^1_\om\cap\Hdot^{-1}_\om}+\Sob{q^{n}}{L^\infty _T(H^1_\om\cap\Hdot^{-1}_\om)}+\Sob{m_1(D)^{\frac{\gam}{2}}q^{n}}{L^{p_0}_{T}\Hdot^{1}_{\om}}+\Sob{m_1(D)^{-\frac{1}{2}}G^n}{L^2_T(H^1_\om\cap\Hdot^{-1}_\om)},&\be=0\\
\Sob{\tht^n_0}{H^{1+\be}_\om}+\Sob{q^{n}}{L^\infty _T H^{1+\be}_\om}+\Sob{m_1(D)^{\frac{\gam}{2}}q^{n}}{L^{p_0}_{T}\Hdot^{1+\be}_{\om}}+\Sob{m_1(D)^{-\frac{1}{2}}G^n}{L^2_T\Hdot^{1+\be}_\om},&\be\in(0,1],\\
\Sob{\tht^n_0}{H^{1+\be}_\om}+\Sob{q^{n}}{L^\infty_T H^{1+\be}_\om}+\Sob{m_1(D)^{\frac{\gam}{2}}q^{n}}{L^{p_0}_{T}\Hdot^{1+\be}_{\om}}+\Sob{m_1(D)^{-\frac{1}{2}}G^n}{L^2_T\Hdot^{1+\be}_\om},&\be\in(1,2].
\end{cases}\notag
\end{align}
is finite. For all $n\in\NN\cup\{\infty\}$, let $\tht^n$ denote the unique solution of \eqref{eq:modclaw:n}, corresponding to data ($\tht^n_0,q^n,G^n)$, guaranteed by \cref{thm:modclaw:wellposed}.
{
\noindent When $\be=0$:
\begin{align}\label{eq:stab:estimate:a}
&\Sob{\tht^n-\tht^\infty}{L^\infty_T(L^2_{\om}\cap\Hdot^{-1}_\om)}^2+\Sob{m(D)^{\frac{1}{2}}(\tht^n-\tht^\infty)}{L^2_T(L^2_{\om}\cap\Hdot^{-1}_\om)}^2\leq CZ^n_0
\end{align}
When $\be\in(0,1)$:
\begin{align}\label{eq:stab:estimate:b1}
&\Sob{\tht^n-\tht^\infty}{L^\infty_TL^2_{\om}}^2+\Sob{m(D)^{\frac{1}{2}}(\tht^n-\tht^\infty)}{L^2_TL^2_{\om}}^2\leq CZ^n_\be\\
&\Sob{\tht^n-\tht^\infty}{L^\infty_TH^\be_\om}^2+\Sob{m(D)^{\frac{1}{2}}(\tht^n-\tht^\infty)}{L^2_TH^\be_\om}^2\leq C(Z^n_\be)^{\frac{1}{1+\be}}.\label{eq:stab:estimate:b2}
\end{align}
When $\be\in[1,2]$:
\begin{align}\label{eq:stab:estimate:c}
&\Sob{\tht^n-\tht^\infty}{L^\infty_TH^{\be}_\om}^2+\Sob{m(D)^{\frac{1}{2}}(\tht^n-\tht^\infty)}{L^2_TH^{\be}_\om}^2\leq C Z^n_\be,
\end{align}
for all $n\in\NN\cup\{\infty\}$, for some constant $C>0$ depending on $Q_\be^*$.
}
\end{Prop}
Note that with these stability-type estimates in hand, we may argue by density to establish a continuity property for the system in its datum with respect to the \textit{weaker} topology of $H^\be_\om$, which is a crucial difference between \cref{thm:stab} and \cref{prop:stab}. Indeed, let us first prove \cref{thm:stab} assuming \cref{prop:stab}. We will then provide the proof of \cref{prop:stab} after.
\begin{proof}[Proof of \cref{thm:stab} (assuming \cref{prop:stab})]
Suppose that $\tht_0^n\in H^{\be}_\om$ and $m_1(D)^{-\frac{1}{2}}G^n\in L^2(0,T;H^{\be}_\om)$, for all $n \in \mathbb{N}\cup\{\infty\}$. For each $k\in \mathbb{N}$, denote by $\tht^{n}_{k}$ the unique solution to
\begin{align} \label{eq:modclaw:k}
\partial_{t}\theta^{n}_{k}+ m(D)\tht^n_k+\Div F_{q^{n}}(\tht^{n}_{k})=S_{k}G^n ,\quad\tht^{n}_{k}(0,x)=S_{k}\tht_0^n(x),
\end{align}
where $S_{k}$ denotes the Littlewood-Paley projection onto frequencies $|\xi|\leq 2^{k}$ defined in \cref{sect:lp:decomposition}. Let
\[
\Tht^n_k=\tht^n-\tht^n_k,\quad \text{for all}\ n\in\NN\cup\{\infty\}.
\]
From \eqref{eq:modclaw:n} and \eqref{eq:modclaw:k}, it follows that
\[
\partial_{t}\Tht^n_k+m(D)\Theta^n_k+ \Div F_{q^{n}}(\Tht^n_k)=(I-S_{k})G^n,\quad \text{for all}\ n\in\NN\cup\{\infty\}.
\]
Then by \cref{thm:modclaw:wellposed}, we obtain
\begin{align}\notag
\sup_{n\in\NN\cup\{\infty\}}&\left(\Sob{\Tht^n_k}{L^{\infty}_{T}(L^2_\om\cap\Hdot^{-1}_\om)}+\Sob{m(D)^{\frac{1}{2}}\Tht^n_k}{L^{2}_{T}(L^2_\om\cap\Hdot^{-1}_\om)}\right)\notag\\
&\le C_{T}\left(\Sob{(I-S_{k})\tht_{0}^n}{L^2_\om\cap\Hdot^{-1}_\om}+\Sob{m_1(D)^{-\frac{1}{2}}(I-S_{k})G^n}{L^2_{T}(L^2_\om\cap\Hdot^{-1}_\om)}\right).\notag
\end{align}
when $\be=0$, and
\begin{align}\notag
\sup_{n\in\NN\cup\{\infty\}}\left(\Sob{\Tht^n_k}{L^{\infty}_{T}H_\om^{\beta}}+\Sob{m(D)^{\frac{1}{2}}\Tht^n_k}{L^{2}_{T}H_\om^{\beta}}\right) \le C\left(\Sob{(I-S_{k})\tht_{0}^n}{H_\om^{\beta}}+\Sob{m_1(D)^{-\frac{1}{2}}(I-S_{k})G^n}{L^2_{T}H_\om^{\be}}\right),
\end{align}
when $\be\in(0,2]$. Let $\delta>0$. Because of \eqref{cond:Q0}, we may choose $k_0>0$, independently of $n$, such that
\begin{align}\label{est:modclaw:intermediate:high:a}
\sup_{n\in\NN\cup\{\infty\}}\left(\Sob{\Tht^n_{k_0}}{L^{\infty}_{T}(L^2_\om\cap\Hdot^{-1}_\om)}+\Sob{m(D)^{\frac{1}{2}}\Tht^n_{k_0}}{L^{2}_{T}(L^2_\om\cap\Hdot^{-1}_\om)}\right)\le \delta/3.
\end{align}
when $\be=0$, and
\begin{align}\label{est:modclaw:intermediate:high:b}
\sup_{n\in\NN\cup\{\infty\}}\left(\Sob{\Tht^n_{k_0}}{L^{\infty}_{T}H_\om^{\beta}}+\Sob{m(D)^{\frac{1}{2}}\Tht^n_{k_0}}{L^{2}_{T}H_\om^{\beta}}\right)\le \delta/3.
\end{align}
when $\be\in(0,2]$.
Now observe that $S_{k_0}\tht_{0}^n\in H_\om^{1+\be}$ and $ m_1(D)^{-\frac{1}{2}}S_{k_0}G^n \in L^{2}(0,T;H_\om^{1+\be})$. We may thus apply \cref{thm:modclaw:wellposed} to obtain a sequence $\{\tht_{k_0}^n\}$ of solutions to \eqref{eq:modclaw:n} corresponding to data $S_{k_0}\tht_0^n$ and $S_{k_0}G^n$. Since $Q_\be<\infty$, we may now invoke \cref{prop:stab} in conjunction with \eqref{cond:Z} to find an integer $N>0$ such that
\begin{align}\label{eq:approx:converge:a}
\sup_{n\geq N}\left(\Sob{\tht^n_{k_0}-\tht^\infty_{k_0}}{L^{\infty}_{T}(L^2_\om\cap\Hdot^{-1}_\om)}+\Sob{m(D)^{\frac{1}{2}}(\tht^n_{k_0}-\tht^\infty_{k_0})}{L^{2}_{T}(L^2_\om\cap\Hdot^{-1}_\om)}
\right)\le \delta/3,
\end{align}
when $\be=0$, and
\begin{align}\label{eq:approx:converge:b}
\sup_{n\geq N}\left(\Sob{\tht^n_{k_0}-\tht^\infty_{k_0}}{L^{\infty}_{T}H_\om^{\beta}}+\Sob{m(D)^{\frac{1}{2}}(\tht^n_{k_0}-\tht^\infty_{k_0})}{L^{2}_{T}H_\om^{\beta}}
\right)\le \delta/3,
\end{align}
when $\be\in(0,2]$. Finally, we see that
\[
\Tht^n:=\tht^n-\tht^\infty=\Tht^n_{k_0}+(\tht^n_{k_0}-\tht_{k_0}^\infty)-\Tht^\infty_{k_0}.
\]
Therefore, by the triangle inequality, combined with \eqref{est:modclaw:intermediate:high:a}, \eqref{eq:approx:converge:a}, we obtain
\begin{align}\notag
&\Sob{\Theta^{n}}{L^{\infty}_{T}(L^2_\om\cap\Hdot^{-1}_\om)\cap L^2_T(L^2_{\om m^{1/2}}\cap\Hdot^{-1}_{\om m^{1/2}})}
\notag\\&\leq \Sob{\Tht^n_{k_0}}{L^\infty_T(L^2_\om\cap\Hdot^{-1}_\om)\cap L^2_T(L^2_{\om m^{1/2}}\cap\Hdot^{-1}_{\om m^{1/2}})}+\Sob{\tht^n_{k_0}-\tht^\infty_{k_0}}{L^{\infty}_{T}(L^2_\om\cap\Hdot^{-1}_\om)\cap L^2_T(L^2_{\om m^{1/2}}\cap\Hdot^{-1}_{\om m^{1/2}})}\notag\\&+\Sob{\Tht^\infty_{k_0}}{L^{\infty}_{T}(L^2_\om\cap\Hdot^{-1}_\om)\cap L^2_T(L^2_{\om m^{1/2}}\cap\Hdot^{-1}_{\om m^{1/2}})}\leq\de,\notag
\end{align}
for all $n\geq N$, when $\be=0$. Similarly, when $\be\in(0,2]$, we apply \eqref{est:modclaw:intermediate:high:b}, \eqref{eq:approx:converge:b} to deduce
\begin{align}
\sup_{n\geq N}\Sob{\Theta^{n}}{L^{\infty}_{T}H_\om^{\beta}\cap L^2_TH^\be_{\om m^{1/2}}}&\leq\de,\notag
\end{align}
Since $\delta$ was arbitrary, this establishes the claims \eqref{eq:stab:a} and \eqref{eq:stab:b}.
\end{proof}
Let us now bring our attention to proving \cref{prop:stab}. In order to do so, we require certain bounds for the divergence of the flux in \eqref{eq:mod:claw}. This is stated in the following lemma.
\begin{Lem}\label{lem:stab:G}
Let $\beta \in [0,2]$ and $F_q(\tht)$ be defined as in \eqref{def:mod:flux}. Let $\gam \in (0,1)$. Then we have the following estimates depending on $\be$: \\When $\be=0$:
\begin{align}
\Sob{m_1(D)^{-\frac{\gam}2}\Div F_{q}(\tht)}{\Hdot^{-1}_\om}&\le C\nrm{q}_{\Hdot^{-1}_\om\cap L^2_{\om}}\Sob{m_1(D)^{\frac{\gam}{2}} \tht}{L^2_{\om}}+C\nrm{\tht}_{L^2_{\om}}\nrm{m_1(D)^{\frac{\gam}2}q}_{L^2_{\om}},\label{est:div:a1}\\
\Sob{m_1(D)^{-\frac{\gam}2}\Div F_{q}(\tht)}{L^2_{\om}}&\le C\nrm{q}_{\Hdot^
{-1}_\om\cap L^2_{\om}}\nrm{m_1(D)^{\frac{\gam}{2}}\tht}_{\Hdot^{1}_\om}+C\nrm{\tht}_{\Hdot^{1}_\om}\nrm{m_1(D)^{\frac{\gam}{2}}q}_{L^2_{\om}}. \label{est:div:a2}
\end{align}
When $\be\in(0,1)$:
\begin{align}
\Sob{m_1(D)^{-\frac{\gam}2}\Div F_{q}(\tht)}{\Hdot^\be_\om}&\le C\nrm{q}_{\Hdot^{-1}_\om\cap \Hdot^\be_\om}\nrm{m_1(D)^{\frac{\gam}2}\tht}_{\Hdot^{1+\be}_\om}+C\nrm{\tht}_{\Hdot^{1+\be}_\om}\nrm{m_1(D)^{\frac{\gam}2}q}_{\Hdot^{\be}_\om},\label{est:div:b1}\\
\Sob{m_1(D)^{-\frac{\gam}2}\Div F_q(\tht)}{L^2_{\om}}&\leq C\nrm{q}_{L^2_{\om}}\nrm{m_1(D)^{\frac{\gam}2}\tht}_{\Hdot^{1+\be}_\om}+C\nrm{\tht}_{\Hdot^{1+\be}_\om}\nrm{m_1(D)^{\frac{\gam}2}q}_{L^2_{\om}}\label{est:div:b2}.
\end{align}
When $\be\in[1,2]$:
\begin{align}
\Sob{m_1(D)^{-\frac{\gam}2}\Div F_{q}(\tht)}{H^\be_\om}&\le C\nrm{q}_{ H^\be_\om}\nrm{m_1(D)^{\frac{\gam}2}\tht}_{H^{1+\be}_\om}+C\nrm{\tht}_{H^{1+\be}_\om}\nrm{m_1(D)^{\frac{\gam}2}q}_{H^{\be}_\om}\label{est:div:c}.
\end{align}
\end{Lem}
Let us assume \cref{lem:stab:G} and prove \cref{prop:stab}. We will then conclude the section by proving \cref{lem:stab:G}.
\begin{proof}[Proof of \cref{prop:stab} (assuming \cref{lem:stab:G})]
Suppose that $\tht_{0}^n \in H^{1+\be}_\om$ and $m_1(D)^{-\frac{1}{2}}G^n\in L^2(0,T;H^{1+\be}_\om)$. By \cref{thm:modclaw:wellposed}, for each $n\in\NN\cup\{\infty\}$, we may let $\tht^n$ denote the unique solution of \eqref{eq:modclaw:n} that belongs to $C([0,T];H^{1+\be}_\om)$. Moreover, by \eqref{est:Hminus1:gronwall} and \eqref{cond:Qbeta} when $\be=0$, it follows that
\begin{align}\label{est:tht:Hminus1:bdd}
\sup_{0<n\leq\infty}\Sob{\tht^n}{L^\infty_T(H^{1}_\om\cap\Hdot^{-1}_\om)},\quad \sup_{0<n\leq\infty}\Sob{m(D)^{\frac{1}{2}}\tht^n}{L^2_T(H^{1}_\om\cap\Hdot^{-1}_\om)}<\infty.
\end{align}
Similarly, \eqref{est:Hs:gronwall} and \eqref{cond:Q0} imply
\begin{align}\label{est:tht:bdd}
\sup_{0<n\leq\infty}\Sob{\tht^n}{L^\infty_T H^{1+\be}_\om},\quad \sup_{0<n\leq\infty}\Sob{m(D)^{\frac{1}{2}}\tht^n}{L^2_T H^{1+\be}_\om}<\infty,
\end{align}
when $\be\in(0,2]$. For each $\NN\cup\{\infty\}$, let
\[
\Tht^n:=\tht^{n}-\tht^{\infty},\quad z^n:=q^n-q^\infty,\quad W^n:=G^n-G^\infty.
\]
Observe that $\Tht^\infty=z^\infty=H^\infty=0$.
Now observe that for each $n\in\NN\cup\{\infty\}$, the pair $(\Tht^n,q^n)$ satisfies
\begin{align}\label{eq:modclaw:intermediate}
\partial_{t}\Tht^n+\Div F_{q^{n}}(\Tht^n)=-m(D)\Tht^n-\Div F_{z^n}(\tht^{\infty})+W^n,\quad \Tht^n(0,x)=\Tht^n_0(x).
\end{align}
Note that \eqref{eq:modclaw:intermediate} possesses the same structure as \eqref{eq:mod:claw} with $\tht\mapsto\Tht^n$, $q\mapsto q^n$, and $G\mapsto -\Div F_{z^n}(\tht^\infty)+W^n$. Thus, when $\be=0$, it follows from \eqref{est:Hminus1:gronwall} that
\begin{align}\label{eq:convergence:a}
&\Sob{\Tht^n}{L^\infty_T(L^2_{\om}\cap\Hdot^{-1}_{\om})}^2+\Sob{m(D)^{\frac{1}{2}}\Tht^n}{L^2_T(L^2_{\om}\cap\Hdot^{-1}_{\om})}^{2}\\
&\leq C^{(n)}_T(0^-)\left(\Sob{\Tht^n_0}{L^2_{\om}\cap\Hdot^{-1}_\om}^2+\Sob{m_1(D)^{-\frac{1}{2}}\Div F_{z^n}(\tht^\infty)}{L^2_T(L^2_{\om}\cap\Hdot^{-1}_\om)}^2+\Sob{m_1(D)^{-\frac{1}{2}}W^n}{L^2_T(L^2_{\om}\cap\Hdot^{-1}_\om)}^2\right),\notag
\end{align}
where $C_T^{(n)}(0^-)$ is defined in \eqref{def:Hminus1:constant} corresponding to $q\mapsto q^n$. On the other hand, for any $\be \in[0,2]$ and $\s\ge 0$ satisfying \eqref{sigma:range}, it follows from \eqref{est:Hs:gronwall} that
\begin{align}\label{eq:convergence:b}
\Sob{\Tht^n}{L^\infty_TH^{\s}_\om}^2+&\Sob{m(D)^{\frac{1}{2}}\Tht^n}{L^2_TH^{\s}_\om}^2\notag\\
&\leq C_T^{(n)}(\be)\left(\Sob{\Tht_0^n}{H^\s_\om}^2+\Sob{m_1(D)^{-\frac{1}{2}}\Div F_{z^n}(\tht^\infty)}{L^2_TH^\s_\om}^2+\Sob{m_1(D)^{-\frac{1}{2}}W^n}{L^2_TH^\s_\om}^2\right),
\end{align}
where $C_T^{(n)}(\be)$ is defined in \eqref{def:Hs:constant} corresponding to $q\mapsto q^n$. By the uniform bounds in \eqref{cond:Q0}, we have
\begin{align}\label{est:constant}
\sup_{n\in\NN\cup\{\infty\}}C_T^{(n)}(\be)<\infty.
\end{align}
We are thus left to estimate $m_1(D)^{-\frac{1}2}\Div F_{z^n}(\tht^\infty)$.
When $\be=0$, \eqref{est:div:a1}, \eqref{est:div:a2} in \cref{lem:stab:G}, and \eqref{est:tht:Hminus1:bdd} imply
\begin{align}\label{est:div:convergence:a}
\Sob{m_1(D)^{-\frac{\gam}2}\Div F_{z^n}(\tht^\infty)}{L^2_T (L^2_{\om}\cap \Hdot^{-1}_\om)}&\le C\nrm{z^n}_{L^\infty_T(\Hdot^
{-1}_\om\cap L^2_{\om})}\nrm{m_1(D)^{\frac{\gam}{2}}\tht^\infty}_{L^2 _T H^{1}_\om}+C\nrm{\tht^\infty}_{L^\infty_T H^{1}_\om}\nrm{m_1(D)^{\frac{\gam}{2}}z^n}_{L^2_T L^2_{\om}}.
\end{align}
When $\be\in(0,1)$, \eqref{est:div:b2} in \cref{lem:stab:G}, and \eqref{est:tht:bdd} imply
\begin{align}\label{est:div:convergence:b}
\Sob{m_1(D)^{-\frac{\gam}2}\Div F_{z^n}(\tht^\infty)}{L^2_T L^2_{\om}}&\leq C\nrm{z^n}_{L^\infty_T L^2_{\om}}\nrm{m_1(D)^{\frac{\gam}2}\tht^\infty}_{L^2_T \Hdot^{1+\be}_\om}+C\nrm{\tht^\infty}_{L^\infty_T H^{1+\be}_\om}\nrm{m_1(D)^{\frac{\gam}2}z^n}_{L^2_T L^2_{\om}}.
\end{align}
Lastly, when $\be\in[1,2]$, \eqref{est:div:c} in \cref{lem:stab:G}, and \eqref{est:tht:bdd} imply
\begin{align}\label{est:div:convergence:c}
\Sob{m_1(D)^{-\frac{\gam}2}\Div F_{z^n}(\tht^\infty)}{L^2_T H^\be_\om}&\le C\nrm{z^n}_{L^\infty_T H^\be_\om}\nrm{m_1(D)^{\frac{\gam}2}\tht^\infty}_{L^2 H^{1+\be}_\om}+C\nrm{\tht^\infty}_{L^\infty_T H^{1+\be}_\om}\nrm{m_1(D)^{\frac{\gam}2}z^n}_{L^2_T H^{\be}_\om}.
\end{align}
Now we apply \eqref{est:div:convergence:a} in \eqref{eq:convergence:a} when $\be=0$ to obtain
\begin{align}\label{est:Tht:converge:a}
&\Sob{\Tht^n}{L^\infty_T(L^2_{\om}\cap\Hdot^{-1}_\om)}^2+\Sob{m(D)^{\frac{1}{2}}\Tht^n}{L^2_T(L^2_{\om}\cap\Hdot^{-1}_\om)}^2\notag \\
&\leq C\left(\Sob{\Tht^n_0}{L^2_{\om}\cap\Hdot^{-1}_\om}^2+\nrm{z^n}_{L^\infty_T(\Hdot^
{-1}_\om\cap L^2_{\om})}^2\nrm{m_1(D)^{\frac{\gam}{2}}\tht^\infty}_{L^2 _T H^{1}_\om}^2\right. \notag \\&\left.\hspace{3 em}+\nrm{\tht^\infty}_{L^\infty_T H^{1}_\om}^2\nrm{m_1(D)^{\frac{\gam}{2}}z^n}_{L^2_T L^2_{\om}}^2+\Sob{m_1(D)^{-\frac{1}{2}}W^n}{L^2_T(L^2_{\om}\cap\Hdot^{-1}_\om)}^2\right).
\end{align}
Similarly, when $\be\in(0,1)$,
we apply \eqref{est:div:convergence:b} in \eqref{eq:convergence:b} with $\s=0$ to deduce
\begin{align}\label{est:Tht:converge:b2}
&\Sob{\Tht^n}{L^\infty_T L^2_{\om}}^2+\Sob{m(D)^{\frac{1}{2}}\Tht^n}{L^2_T L^2_{\om}}^2\notag \\
&\leq C\left(\Sob{\Tht^n_0}{L^2_{\om}}^2+\nrm{z^n}_{L^\infty_T L^2_{\om}}^2\nrm{m_1(D)^{\frac{\gam}2}\tht^\infty}_{L^2_T \Hdot^{1+\be}_\om}^2\right. \notag \\&\left.\hspace{3 em}+\nrm{\tht^\infty}_{L^\infty_T H^{1+\be}_\om}^2\nrm{m_1(D)^{\frac{\gam}2}z^n}_{L^2_T L^2_{\om}}^2+\Sob{m_1(D)^{-\frac{1}{2}}W^n}{L^2_T L^2_{\om} }^2\right).
\end{align}
Moreover, we may upgrade this bound using interpolation \eqref{est:interpolation} and by applying the uniform bounds \eqref{est:tht:bdd} to obtain
\begin{align}\label{est:Tht:converge:b1}
\Sob{\Tht^n}{L^\infty_TH^\be_\om\cap L^2_TH^\be_{\om m^{1/2}}}
&\leq C\Sob{\Tht^n}{L^\infty_T H^{1+\be}_\om\cap L^2_T H^{1+\be}_{\om m^{1/2}}}^{\frac{\be}{1+\be}}\Sob{\Tht^n}{L^\infty_T L^2_{\om}\cap L^2_TL^2_{\om m^{1/2}}}^{\frac{1}{1+\be}}\notag\\
&\leq C\left(\sup_{n\in\NN\cup\{\infty\}}\Sob{\tht^n}{L^\infty_T H^{1+\be}_\om\cap L^2_T H^{1+\be}_{\om m^{1/2}}}\right)^{\frac{\be}{1+\be}}\Sob{\Tht^n}{L^\infty_TL^2_{\om}\cap L^2_TL^2_{\om m^{1/2}}}^{\frac{1}{1+\be}}\notag\\
&\leq C\left(\Sob{\Tht^n_0}{L^2_{\om}}^2+\nrm{z^n}_{L^\infty_T L^2_{\om}}^2\nrm{m_1(D)^{\frac{\gam}2}\tht^\infty}_{L^2_T \Hdot^{1+\be}_\om}^2\right. \notag \\&\left.\hspace{3 em}+\nrm{\tht^\infty}_{L^\infty_T H^{1+\be}_\om}^2\nrm{m_1(D)^{\frac{\gam}2}z^n}_{L^2_T L^2_{\om}}^2+\Sob{m_1(D)^{-\frac{1}{2}}W^n}{L^2_T L^2_{\om} }^2\right)^{\frac{1}{1+\be}}.
\end{align}
Lastly, when $\be\in[1,2]$, we apply \eqref{est:div:convergence:c} in \eqref{eq:convergence:b} with $\s=\be$, to obtain
\begin{align}\label{est:Tht:converge:c2}
&\Sob{\Tht^n}{L^\infty_TH^{\be}_\om}^2+\Sob{m(D)^{\frac{1}{2}}\Tht^n}{L^2_TH^{\be}_\om}^2\notag \\
&\leq C\left(\Sob{\Tht^n_0}{H^\be_\om}^2+\nrm{z^n}_{L^\infty_T H^\be_\om}^2\nrm{m_1(D)^{\frac{\gam}2}\tht^\infty}_{L^2_T H^{1+\be}_\om}^2\right. \notag \\&\left.\hspace{3 em}+\nrm{\tht^\infty}_{L^\infty_T H^{1+\be}_\om}^2\nrm{m_1(D)^{\frac{\gam}2}z^n}_{L^2_T H^\be_\om}^2+\Sob{m_1(D)^{-\frac{1}{2}}W^n}{L^2_T H^\be_\om }^2\right).
\end{align}
Gathering the estimates \eqref{est:Tht:converge:a}, \eqref{est:Tht:converge:b1}, \eqref{est:Tht:converge:c2} completes the proof.
\end{proof}
Finally, let us prove \cref{lem:stab:G}
\begin{proof}[Proof of \cref{lem:stab:G}]
We split the proof into the cases $\be=0$, $\be\in(0,1)$, $\be=1$, $\be\in(1,2]$. Throughout, recall that $v$ is given by \eqref{def:v}. We will repeatedly apply \cref{cor:prod} and \cref{lem:commutator4}, which are applicable due to \eqref{cond:summary}.
\subsubsection*{Case $\be =0$} Then $\Div F_q(\tht)=\nabla\cdotp (v\tht)=v\cdotp\nabla\tht$ and we have
\begin{align}\label{eq:stab:setup}
\Sob{m_1(D)^{-\frac{\gam}{2}}\Div F_q(\tht)}{\Hdot^{-1}_{\om }}\leq \Sob{v\tht}{L^2_{\om m_1^{-\gam/2}}},\quad \Sob{m_1(D)^{-\frac{\gam}{2}}\Div F_q(\tht)}{L^2_{\om}}\leq \Sob{v\cdotp\nabla\tht}{L^2_{\om m_1^{-\gam/2}}}.
\end{align}
We apply \cref{cor:prod} to the first expression in \eqref{eq:stab:setup} with $(s,\bar{s})=(1,0)$, $(\om_1,\til{\om}_1)=(p^{-1}\om,m_1^{\frac{\gam}{2}}\om)$, $(\om_2,\til{\om}_2)=(\om, p^{-1}m_1^{\frac{\gam}{2}}\om)$, $(\om_3, \til{\om}_3)=(p^{-1}\om,m_1^{\frac{\gam}{2}}\om)$, and $\Gamma=m_1^{\frac{\gam}{2}}$, we obtain
\begin{align}\label{est:stab:0:a}
\Sob{m_1^{-\frac{\gam}{2}}{\Lam}^{-1}\nabla\cdotp(v \tht)}{L^2_{\om}} &\le C\Sob{p(D)^{-1}v}{H^1_\om}\Sob{m_1(D)^{\frac{\gam}{2}} \tht}{L^2_{\om}}+C\Sob{ \tht}{L^2_{\om}}\Sob{m_1(D)^{\frac{\gam}{2}}p(D)^{-1}v}{\Hdot^1_\om}\notag\\
&\le C\nrm{q}_{\Hdot^
{-1}_\om\cap L^2_{\om}}\Sob{m_1(D)^{\frac{\gam}{2}} \tht}{L^2_{\om}}+C\nrm{\tht}_{L^2_{\om}}\nrm{m_1(D)^{\frac{\gam}2}q}_{L^2_{\om}}.
\end{align}
Similarly, we apply \cref{cor:prod} to the second expression in \eqref{eq:stab:setup} with $(s,\bar{s})=(1,0)$, $(\om_1,\til{\om}_1)=(p^{-1}\om, m_1^{\frac{\gam}{2}}\om,)$, $(\om_2,\til{\om}_2)=(\om,p^{-1}m_1^{\frac{\gam}{2}}\om)$, $(\om_3,\til{\om}_3)=(p^{-1}\om, m_1^{\frac{\gam}{2}}\om)$, and $\Gamma=m_1^{\frac{\gam}{2}}$, obtain
\begin{align}\label{est:stab:0:b}
\Sob{m_1(D)^{-\frac{\gam}{2}}\Div F_{q}(\tht)}{L^2_{\om}} &\le C\Sob{p(D)^{-1}v}{H^1_\om}\Sob{m_1(D)^{\frac{\gam}{2}}\nabla \tht}{L^2_{\om}}+C\Sob{\nabla \tht}{L^2_{\om}}\Sob{m_1(D)^{\frac{\gam}{2}}p(D)^{-1}v}{\Hdot^1_\om}\notag\\
&\le C\nrm{q}_{\Hdot^
{-1}_\om\cap L^2_{\om}}\nrm{m_1(D)^{\frac{\gam}{2}}\tht}_{\Hdot^{1}_\om}+C\nrm{\tht}_{\Hdot^{1}_\om}\nrm{m_1(D)^{\frac{\gam}{2}}q}_{L^2_{\om}}.
\end{align}
\subsubsection*{Case $\be \in(0,1)$}
We apply \cref{cor:prod} with $(s,\bar{s})=(1, \be)$, $(\om_1,\til{\om}_1)=(p^{-1}\om, m_1^{\frac{\gam}{2}}\om,)$, $(\om_2,\til{\om}_2)=(\om,p^{-1}m_1^{\frac{\gam}{2}}\om)$, $(\om_3,\til{\om}_3)=(p^{-1}\om, m_1^{\frac{\gam}{2}}\om)$, and $\Gamma=m_1^{\frac{\gam}{2}}$, obtain
\begin{align}\label{est:stab:1:a}
\Sob{m_1(D)^{-\frac{\gam}{2}}\Div F_{q}(\tht)}{\Hdot^{\be}_\om} &\le C\Sob{p(D)^{-1}v}{H^1_\om}\Sob{m_1(D)^{\frac{\gam}{2}}\nabla \tht}{\Hdot^\be_\om}+C\Sob{\nabla \tht}{\Hdot^{\be}_{\om}}\Sob{m_1(D)^{\frac{\gam}{2}}p(D)^{-1}v}{\Hdot^1_\om}\notag\\
&\le C\nrm{q}_{\Hdot^{-1}_\om\cap \Hdot^\be_\om}\nrm{m_1(D)^{\frac{\gam}2}\tht}_{\Hdot^{1+\be}_\om}+C\nrm{\tht}_{\Hdot^{1+\be}_\om}\nrm{m_1(D)^{\frac{\gam}2}q}_{\Hdot^{\be}_\om}.
\end{align}
On the other hand, applying \cref{cor:prod} with $(s,\bar{s})=(1-\be, \be)$, $(\om_1,\til{\om}_1)=(p^{-1}\om, m_1^{\frac{\gam}{2}}\om,)$, $(\om_2,\til{\om}_2)=(\om,p^{-1}m_1^{\frac{\gam}{2}}\om)$, $(\om_3,\til{\om}_3)=(p^{-1}\om, m_1^{\frac{\gam}{2}}\om)$, and $\Gamma=m_1^{\frac{\gam}{2}}$, we obtain
\begin{align}\label{est:stab:1:b}
\Sob{m_1(D)^{-\frac{\gam}2}\Div F_{q}(\tht)}{L^2_{\om}} &\le C\Sob{p(D)^{-1}v}{\Hdot^{1-\be}_\om}\Sob{m_1(D)^{\frac{\gam}{2}}\nabla \tht}{\Hdot^\be_\om}+C\Sob{\nabla \tht}{\Hdot^\be_\om}\Sob{m_1(D)^{\frac{\gam}{2}}p(D)^{-1}v}{\Hdot^{1-\be}_\om}\notag\\
&\le C\nrm{q}_{L^2_{\om}}\nrm{m_1(D)^{\frac{\gam}2}\tht}_{\Hdot^{1+\be}_\om}+C\nrm{\tht}_{\Hdot^{1+\be}_\om}\nrm{m_1(D)^{\frac{\gam}2}q}_{L^2_{\om}}.
\end{align}
\subsubsection*{Case $\be=1$} We apply \cref{cor:prod} with $(s,\bar{s})=(0,1)$, $(\om_1,\til{\om}_1)=(p^{-1}\om, m_1^{\frac{\gam}{2}}\om,)$, $(\om_2,\til{\om}_2)=(\om,p^{-1}m_1^{\frac{\gam}{2}}\om)$, $(\om_3,\til{\om}_3)=(p^{-1}\om, m_1^{\frac{\gam}{2}}\om)$, and $\Gamma=m_1^{\frac{\gam}{2}}$, obtain
\begin{align}\label{est:stab:1:endpoint:a}
\Sob{m_1(D)^{-\frac{\gam}2}\Div F_{q}(\tht)}{L^{2}_\om} &\le C\Sob{p(D)^{-1}v}{L^{2}_\om}\Sob{m_1(D)^{\frac{\gam}{2}}\nabla \tht}{\Hdot^1_\om}+C\Sob{\nabla \tht}{H^1_\om}\Sob{m_1(D)^{\frac{\gam}{2}}p(D)^{-1}v}{L^2_{\om}}\notag\\
&\le C\nrm{q}_{L^{2}_\om}\nrm{m_1(D)^{\frac{\gam}2}\tht}_{\Hdot^{2}_\om}+C\nrm{\tht}_{H^{2}_\om}\nrm{m_1(D)^{\frac{\gam}2}q}_{L^2_{\om}}.
\end{align}
Similarly, we apply \cref{cor:prod} with $(s,\bar{s})=(1,1)$, $(\om_1, \til{\om}_1)=(p^{-1}\om, m_1^{\frac{\gam}{2}}\om)$, $(\om_2,\til{\om}_2)=(\om, p^{-1}m_1^{\frac{\gam}{2}}\om)$, $(\om_3, \til{\om}_3)=(p^{-1}\om, m_1^{\frac{\gam}{2}}\om)$, and $\Gamma=m_1^{\frac{\gam}{2}}$, we obtain
\begin{align}\label{est:stab:1:endpoint:b}
\Sob{m_1(D)^{-\frac{\gam}2}\Div F_{q}(\tht)}{\Hdot^1_\om} &\le C\Sob{p(D)^{-1}v}{H^{1}_\om}\Sob{m_1(D)^{\frac{\gam}{2}}\nabla \tht}{\Hdot^1_\om}+C\Sob{\nabla \tht}{H^1_\om}\Sob{m_1(D)^{\frac{\gam}{2}}p(D)^{-1}v}{\Hdot^{1}_\om}\notag\\
&\le C\nrm{q}_{H^1_\om}\nrm{m_1(D)^{\frac{\gam}2}\tht}_{\Hdot^{2}_\om}+C\nrm{\tht}_{H^{2}_\om}\nrm{m_1(D)^{\frac{\gam}2}q}_{\Hdot^{1}_\om}.
\end{align}
\subsubsection*{Case $\be \in (1,2]$} In this range, we dualize. Let $H=\Div F_{q}(\tht)$. Firstly, we see from \cref{lem:commutator2} applied with $s=2-\beta$ and $\eps\in(0,1]$ satisfying $\eps/2+s<1$, that
\begin{align}\label{est:H:L2}
\Sob{H}{L^2_{\om}}^{2}&=\lb \nabla \cdot((\nabla^{\perp}a(D)q)\tht),\om^2 H\rb+\lb a(D)\nabla \cdot((\nabla^{\perp}\tht)q),\om^2 H\rb\notag\\
&=-\lb[\bdy^{\perp}_{\ell}a(D),\bdy_{\ell}\tht]q ,\om^2 H\rb\notag\\
&\le C\Sob{\nabla\tht}{H^{\be-\de}}\left(\Sob{p_a(D)q}{\Hdot^{\eps/2}}\Sob{H}{L^2_{\om^2}}+\Sob{p_a(D)H}{\Hdot^{\eps/2}_{\om^2}}\Sob{q}{L^2}\right)\notag\\
&\leq C\Sob{\tht}{H^{1+\be}_\om}\left(\Sob{q}{H^{\eps}}\Sob{H}{H^\eps}+\Sob{H}{{H}^\eps}\Sob{q}{L^2}\right)\notag\\
&\le C\Sob{\tht}{H^{1+\be}_\om}\Sob{q}{H^{\eps}}\Sob{H}{H^{\eps}}.
\end{align}
where we applied the embedding $H^{1+\be}_\om \subset H^{1+\be-\de}$ and \eqref{eq:omega:eps:bdd} to $p_a(D), \om(D)$, in obtaining the final two inequalities.
Now, to estimate in $\Hdot^\be_\om$, we localize. Let $j\in \mathbb{Z}$. Then
\begin{align*}
\Sob{{\Lam}^{\be}_{\om,j}m_1(D)^{-\frac{\gam}{2}}H}{L^2}^2
=&\lb {\Lam}^{\be}_{\om,j}(\nabla^{\perp}a(D)q\cdot\nabla \tht),{\Lam}^{\be}_{\om,j}m_1(D)^{-\gam}{H}\rb-\lb {\Lam}^{\be}_{\om,j}a(D)(\nabla^{\perp}q\cdot\nabla \tht),{\Lam}^{\be}_{\om,j}m_1(D)^{-\gam}{H}\rb\\
=& -\lb[\bdy^{\perp}_{\ell}a(D),\bdy_{\ell}\tht]{\Lam}^{\beta}_{\om,j}q ,{\Lam}^{\beta}_{\om,j}m_1(D)^{-\gam}H\rb+\lb[{\Lam}^{\beta}_{\om,j}, \bdy_{\ell}\tht]\bdy^{\perp}_{\ell}a(D)q ,{\Lam}^{\beta}_{\om,j}m_1(D)^{-\gam}H\rb\\
&-\lb[{\Lam}^{\be}_{\om,j},\bdy_{\ell}\tht]\bdy^{\perp}_{\ell}q ,a(D){\Lam}^{\beta}_{\om,j} m_1(D)^{-\gam}H\rb\notag\\
=&\ K_1+K_2+K_3.
\end{align*}
Applying \cref{lem:commutator3} with $s=2-\be$, $\eps=0$, $\Gamma=m_1^{\gam}p^{-1}$, and Bernstein's inequality, we obtain
\begin{align*}
|K_1|&\le Cm_1 (2^j)^{\gam}\Sob{\bdy_{\ell} \tht}{{H}^{\be}_\om}\Sob{{\Lam}^{\be}_{\om, j}q}{L^2}\Sob{{\Lam}^{\beta}_{\om,j}m_1(D)^{-\gam}H}{L^2}\notag\\
&\le Cc_j\Sob{\tht}{H^{1+\be}_\om}\Sob{m_1(D)^{\frac{\gam}{2}}q}{\Hdot^{\be}_\om}\Sob{{\Lam}^{\beta}_{\om,j}m_1(D)^{-\frac{\gam}{2}}H}{L^2}.
\end{align*}
where we applied property \eqref{est:omega} for $m_1(D)$.
For $K_2$, we apply \cref{lem:commutator4} with $r=\be-1$, $(s, \bar{s})=(1,\be-1)$, $(\om_1,\til{\om}_1)=(p^{-1}\om,m_1^{\frac{\gam}{2}}\om)$, $(\om_2,\til{\om}_2)=(\om, p^{-1}m_1^{\frac{\gam}{2}}\om)$, $(\om_1,\til{\om}_1)=(p^{-1}\om,m_1^{\frac{\gam}{2}}\om)$, and $\Gam_1=\Gam_2=\Gam_3=m^{\gam/2}_1$, to obtain
\begin{align*}
|K_2|\le Cc_{j}& m_1(2^j)^{\frac{\gam}{2}}\left\{\nrm{\bdy_{\ell}^{\perp}p(D)^{-1}a(D)q}_{H^{1}_\om}\nrm{m_1(D)^{\frac{\gam}{2}}\Lam \bdy_{\ell}\tht}_{\Hdot^{\be-1}_\om}\right. \notag \\ &\hspace{5 em}\left.+\nrm{\Lam \bdy_{\ell}\tht}_{H^{\be-1}_\om}\nrm{\bdy_{\ell}^{\perp}m_1(D)^{\frac{\gam}{2}}p(D)^{-1}a(D)q}_{\Hdot^{1}_\om}\right. \notag \\ &\hspace{5 em}\left. +\nrm{\bdy_{\ell}^{\perp}p(D)^{-1}a(D)q}_{\Hdot^{1}_\om}\nrm{m_1(D)^{\frac{\gam}{2}}\Lam \bdy_{\ell}\tht}_{\Hdot^{\be-1}_\om} \right\}\Sob{{\Lam}^{\beta}_{\om,j}m_1(D)^{-\gam}H}{L^2}\notag\\
\le Cc_{j}&\left(\nrm{q}_{H^\be_\om}\nrm{m_1(D)^{\frac{\gam}2}\tht}_{\Hdot^{1+\be}_\om}+C\nrm{\tht}_{H^{1+\be}_\om}\nrm{m_1(D)^{\frac{\gam}2}q}_{\Hdot^{\be}_\om}\right)\nrm{{\Lam}^{\be}_{\om,j}m_1(D)^{-\frac{\gam}{2}}H}_{L^2}.
\end{align*}
Lastly, we apply \cref{lem:commutator4} with $r=\be-1$, $(s,\bar{s})=(\be-1,\be-1)$, $(\om_1,\til{\om}_1)= (\om_2,\tilde{\om}_2)=(\om_3,\tilde{\om}_3)=(\om, m_1^{\frac{\gam}{2}}\om)$, and $\Gam_1=\Gam_2=\Gam_3=m^{\gam/2}_1p^{-\frac{1}{2}}$, to obtain
\begin{align*}
|K_3|\le Cc_{j}&m_1(2^j)^{\frac{\gam}{2}}p(2^j)^{-1}2^{(2-\be)j}\left\{\nrm{\bdy_{\ell}^{\perp}q}_{H^{\be-1}_\om}\nrm{\bdy_{l}m_1(D)^{\frac{\gam}{2}}\Lam \tht}_{\Hdot^{\be-1}_\om}\right. \notag \\
&\hspace{11 em}\left.+\nrm{\bdy_{l}\Lam \tht}_{H^{\be-1}_\om}\nrm{m_1(D)^{\frac{\gam}{2}}\bdy_{\ell}^{\perp}q}_{\Hdot^{\be-1}}\right. \notag \\
&\hspace{11 em}\left. +\nrm{\bdy_{\ell}^{\perp}q}_{\Hdot^{\be-1}_\om}\nrm{\bdy_{l}m_1(D)^{\frac{\gam}{2}}\Lam \tht}_{\Hdot^{\be-1}_\om} \right\}\nrm{a(D){\Lam}^{\be}_{\om,j}m_1(D)^{-\gam}H}_{L^2}\notag\\
\le Cc_{j}&\left(\nrm{q}_{H^\be_\om}\nrm{m_1(D)^{\frac{\gam}2}\tht}_{\Hdot^{1+\be}_\om}+C\nrm{\tht}_{H^{1+\be}_\om}\nrm{m_1(D)^{\frac{\gam}2}q}_{\Hdot^{\be}_\om}\right)\nrm{{\Lam}^{\be}_{\om,j}m_1(D)^{-\frac{\gam}{2}}H}_{L^2}.
\end{align*}
Collecting the estimates of $K_1$, $K_2$ and $K_3$, summing over $j$ and applying the Cauchy-Schwarz inequality, we obtain
\begin{align}\label{est:stab:2}
\Sob{ m_1(D)^{-\frac{\gam}{2}}\Div F_{q}(\tht)}{\Hdot^{\be}_\om}&\le C\nrm{q}_{H^\be_\om}\nrm{m_1(D)^{\frac{\gam}2}\tht}_{\Hdot^{1+\be}_\om}+C\nrm{\tht}_{H^{1+\be}_\om}\nrm{m_1(D)^{\frac{\gam}2}q}_{\Hdot^{\be}_\om}.
\end{align}
when $\be\in(1,2]$. Finally, we combine \eqref{est:stab:2} with \eqref{est:H:L2} to arrive at
\begin{align}\label{est:stab:2:final}
\Sob{m_1(D)^{-\frac{\gam}2}\Div F_{q}(\tht)}{H^\be_\om}&\le C\nrm{q}_{ H^\be_\om}\nrm{m_1(D)^{\frac{\gam}2}\tht}_{\Hdot^{1+\be}_\om}+C\nrm{\tht}_{H^{1+\be}_\om}\nrm{m_1(D)^{\frac{\gam}2}q}_{H^{\be}_\om}.
\end{align}
The estimates \eqref{est:stab:0:a}-\eqref{est:stab:1:endpoint:b}, and \eqref{est:stab:2:final} together complete the proof.
\end{proof}
\section{Well-posedness: Proof of Theorem \ref{thm:main:dgsqg}}\label{sect:wellposed}
We will now establish local existence of a unique solution to \eqref{eq:dgsqg}, which possesses the property of instantaneous smoothing, and is continuous with respect to initial data. The argument hinges on the simple observation that since $\nabla^{\perp}\tht$ is divergence-free, one can express equation \eqref{eq:dgsqg} as
\begin{align}\label{eq:dgsqg:mod:form}
\partial_{t}\theta+ m(D) \tht+ \Div F_{-\tht}(\tht)=0 ,\quad\tht(0,x)=\tht_0(x),
\end{align}
where $F$ is as defined in \eqref{def:mod:flux}, for all $\be\in[0,2]$. In particular, see that \eqref{eq:dgsqg:mod:form} has the structure of \eqref{eq:mod:claw} with $q=-\tht$ and $G\equiv0$. One may then formally deduce apriori estimates for \eqref{eq:dgsqg} from the protean system \eqref{eq:mod:claw}. A rigorous proof of well-posedness then reduces to construction of the solution; this is a straightforward matter and can be dealt with by considering a standard artificial viscosity approximation, for which all apriori estimates hold independently of the viscosity parameter. The relevant details of this argument are provided in \cref{app:well:posed}. In what follows, we perform the formal analysis to establish local well-posedness.
\subsection{Existence}\label{sect:exist}
Upon setting $q=-\tht$ and $G\equiv0$ in \eqref{eq:mod:claw}, we see from \eqref{est:Hs:final} that for any $\be \in [0,2]$, for $\lam\equiv0$, we have
\begin{align}\label{est:balance:exist}
\frac{d}{dt}\Sob{\tht}{\Hdot^{1+\be}_{\om}\cap L^2_{\om}}^2+\frac{3}2\Sob{m_1(D)^{\frac{1}2}\tht}{\Hdot^{1+\be}_{\om}\cap L^2_{\om}}^{2}\leq C \left(1+\Sob{\tht}{H^{1+\be}_\om}^{\frac{1}{1-\gam}}+\Sob{m_1(D)^{\frac{\gam}2}\tht}{\Hdot^{1+\be}_\om}^{\frac{2}{2-\gam}}\right)\Sob{\tht}{\Hdot^{1+\be}_{\om}\cap L^2_{\om}}^2.
\end{align}
Observe that by \eqref{est:interpolation:weighted} and Young's inequality (using the fact that $\gam<1$), we obtain
\begin{align}
C\left(\Sob{\tht}{H^{1+\be}_\om}^{\frac{1}{1-\gam}}+\Sob{m_1(D)^{\frac{\gam}2}\tht}{\Hdot^{1+\be}_\om}^{\frac{2}{2-\gam}}\right)\Sob{\tht}{H^{1+\be}_\om}^2&\leq C\left(\Sob{\tht}{H^{1+\be}_\om}^{\frac{3-2\gam}{1-\gam}}+\Sob{m_1(D)^{\frac{1}2}\tht}{\Hdot^{1+\be}_\om}^{\frac{2\gam}{2-\gam}}\Sob{\tht}{H^{1+\be}_\om}^{\frac{2(3-2\gam)}{2-\gam}}\right)\notag\\
&\leq C\Sob{\tht}{H^{1+\be}_\om}^{\frac{3-2\gam}{1-\gam}}+\frac{1}2\Sob{m_1(D)^{\frac{1}2}\tht}{\Hdot^{1+\be}_\om}^2.\notag
\end{align}
By Gronwall's inequality, we deduce the existence of a time $T=T(\Sob{\tht_0}{{H}^{1+\be}_\om})$ such that
\begin{align}\label{est:apriori:dgsqg:b}
\sup_{t\in[0,T]}\Sob{\tht(t)}{H^{1+\be}_\om}^2+\int_0^T\Sob{m(D)^{\frac{1}{2}}\tht(t)}{H^{1+\be}_\om}^2dt\le C(1+\Sob{\tht_{0}}{H^{1+\be}_\om}^2).
\end{align}
Similarly, when $\be=0$, and $\tht_0 \in H^1_\om \cap \Hdot^{-1}_\om$, we invoke \eqref{est:Hminus1:final} with $q=\tht$ and \eqref{est:Hs:final} with $q=\tht$ and $\s=1$ to deduce the existence of a time $T=T(\Sob{\tht_0}{{H}^{1}_\om\cap \Hdot^{-1}_\om})$ such that
\begin{align}\label{est:apriori:dgsqg:a}
\sup_{t\in[0,T]}\Sob{\tht(t)}{H^1_\om\cap\Hdot^{-1}_{\om}}^2+\int_0^T\Sob{m(D)^{\frac{1}{2}}\tht(t)}{H^1_\om\cap\Hdot^{-1}_{\om}}^2dt\le C(1+\Sob{\tht_{0}}{H^1_\om\cap\Hdot^{-1}_{\om}}^2),
\end{align}
for some constant $C>0$
An artificial viscosity approximation can then be used to construct a solution $\tht\in C([0,T];H^1_\om\cap\Hdot^{-1}_\om)$, provided that $\tht_0\in H^1_\om\cap\Hdot^{-1}_\om$, when $\be=0$, and $\tht\in C([0,T];H^{1+\be}_\om)$, provided that $\tht_0\in H^{1+\be}_\om$, when $\be\in(0,2]$.
\subsection{Smoothing}
Upon setting $q=-\tht$ and $G\equiv0$ in \eqref{eq:mod:claw}, we see from
\eqref{est:Hs:final} and the corresponding inequality but for the case $\be \in (1,2]$ and performing the same analysis from \cref{sect:exist} that
\begin{align*}
\frac{d}{dt}\Sob{\tilde{\tht}}{\Hdot^{1+\be}_{\om}}^2+\frac{3}2\Sob{m_1(D)^{\frac{1}2}\til{\tht}}{\Hdot^{1+\be}_{\om}}^{2}&\leq C \left(1+\Sob{\til{\tht}}{H^{1+\be}_\om}^{\frac{1}{1-\gam}}+\Sob{m_1(D)^{\frac{\gam}2}\til{\tht}}{\Hdot^{1+\be}_\om}^{\frac{2}{2-\gam}}\right)\Sob{\tilde{\tht}}{\Hdot^{1+\be}_{\om}\cap L^2_{\om}}^2\notag\\
&\leq C\left(1+\Sob{\til{\tht}}{H^{1+\be}_\om}^2\right)^{\frac{3-2\gam}{2(1-\gam)}}+\frac{1}2\Sob{m_1(D)^{\frac{1}2}\til{\tht}}{\Hdot^{1+\be}_\om}^2.\notag
\end{align*}
Again, it follows from Gronwall's inequality that there exists $T>0$ such that
\begin{align}\label{est:smoothing}
\sup_{0\le t\le T}\Sob{E^{\lam_1t}_{ \nu}\tht(t)}{\Hdot^{1+\be}_\om}\le C(1+\Sob{\tht_0}{H^{1+\be}_\om}),
\end{align}
as desired.
\subsection{Uniqueness} Let $\be \in [0,2]$. For $j=1,2$, let $\tht_0^{(j)}\in H^1_\om\cap \Hdot^{-1}_\om$ when $\be=0$ and $\tht_0^{(j)}\in H^{1+\be}_\om$ when $\be\in(0,2]$. Suppose that $\tht^{(1)},\tht^{(2)}$ are two solutions of \eqref{eq:dgsqg} corresponding to intial data $\tht_0^{(1)}, \tht_0^{(2)}$ such that $\tht^{(j)}\in C([0,T];\Hdot^{-1}_\om\cap H^1_\om)$, when $\be=0$, and $\tht^{(j)}\in C([0,T];H^{1+\be}_\om)$, when $\be\in(0,2]$, for $j=1,2$, where $T$ is the local existence time obtained from \cref{sect:exist}. To prove uniqueness, it will be convenient to introduce the following notation
\begin{align}\label{def:Y}
Y^\be:=
\begin{cases}
L^2_{\om}\cap\Hdot^{-1}_\om,&\be=0\\
L^2_{\om},&\be\in(0,1)\\
H^\be_\om,&\be\in[1,2].
\end{cases}
\end{align}
Let $\Tht:=\tht^{(1)}-\tht^{(2)}$. Then $\Tht$ is governed by
\begin{align}\label{eq:difference:intro}
\partial_{t}\Tht+m(D) \Tht+\Div F_{-\tht^{(1)}}(\Tht)=\Div F_{\Tht}(\tht^{(2)}),\quad\Tht(0,x)=\Tht_0(x).
\end{align}
Observe that \eqref{eq:difference:intro} has the structure of \eqref{eq:mod:claw} with $q=-\tht^{(1)}$ and $G=\Div F_{\Tht}(\tht^{(2)})$. Then from \eqref{est:Hminus1:final}
\begin{align}\label{est:uniqueness:a:pre}
\frac{d}{dt}\Sob{\Tht}{\dot{H}^{-1}_{\om}\cap L^2_{\om}}^2+c\Sob{m(D)^{\frac{1}2}\Tht}{\dot{H}^{-1}_{\om}\cap L^2_{\om}}^{2}\leq& C\left(1+\Sob{\tht^{(1)}}{ \Hdot^{-1}_\om}^{\frac{1}{1-\gam}}+\Sob{\tht^{(1)}}{\Hdot^{1}_\om}^{\frac{1}{1-\gam}}+\Sob{m_1(D)^{\frac{\gam}2}\tht^{(1)}}{\Hdot^{1}_\om}^{\frac{2}{2-\gam}}\right)\Sob{\Tht}{\dot{H}^{-1}_{\om}\cap L^2_{\om}}^2\notag\\&+C\Sob{m_1(D)^{-\frac{1}{2}}\Div F_{\Tht}(\tht^{(2)})}{\Hdot^{-1}_\om \cap L^2_{\om}}^2.
\end{align}
when $\be=0$.
By interpolation inequality and Young's inequality, we obtain
\begin{align}\label{E:interpolation:uniqueness}
\Sob{m_1(D)^{\frac{\gam}2}\tht^{(1)}}{\Hdot^{1}_\om}^{\frac{2}{2-\gam}}&\le C\Sob{\tht^{(1)}}{\Hdot^{1}_\om}^{\frac{2(1-\gam)}{2-\gam}}\Sob{m_1(D)^{\frac{1}{2}}\tht^{(1)}}{\Hdot^{1}_\om}^{\frac{2\gam}{2-\gam}}\notag\\
&\le C\Sob{\tht^{(1)}}{\Hdot^{\s}_\om}+C\Sob{m_1(D)^{\frac{1}{2}}\tht^{(1)}}{\Hdot^{\s}_\om}^2.
\end{align}
From \eqref{est:uniqueness:a:pre} and \eqref{E:interpolation:uniqueness}, we obtain
\begin{align}\label{est:uniqueness:a}
\frac{d}{dt}\Sob{\Tht}{\dot{H}^{-1}_{\om}\cap L^2_{\om}}^2+c\Sob{m(D)^{\frac{1}2}\Tht}{\dot{H}^{-1}_{\om}\cap L^2_{\om}}^{2}\leq& C\left(1+\Sob{\tht^{(1)}}{ \Hdot^{-1}_\om}^{\frac{1}{1-\gam}}+\Sob{\tht^{(1)}}{\Hdot^{1}_\om}^{\frac{1}{1-\gam}}+\Sob{m_1(D)^{\frac{\gam}2}\tht^{(1)}}{\Hdot^{1}_\om}^{2}\right)\Sob{\Tht}{\dot{H}^{-1}_{\om}\cap L^2_{\om}}^2\notag\\&+C\Sob{m_1(D)^{-\frac{1}{2}}\Div F_{\Tht}(\tht^{(2)})}{\Hdot^{-1}_\om \cap L^2_{\om}}^2.
\end{align}
Similarly, when $\be\in(0,1)$, $\s \in (-1,2]$ or when $\be\in[1,2]$, $\s \in [1,1+\be]$, we have from \eqref{est:Hs:final} that
\begin{align}\label{est:uniqueness:b}
\frac{d}{dt}\Sob{\Tht}{\dot{H}^{\s}_{\om}\cap L^2_{\om}}^2+c\Sob{m(D)^{\frac{1}2}\Tht}{\dot{H}^{\s}_{\om}\cap L^2_{\om}}^{2}\leq& C\left(1+\Sob{\tht^{(1)}}{H^{1+\be}_\om}^{\frac{1}{1-\gam}}+\Sob{m_1(D)^{\frac{\gam}2}\tht^{(1)}}{\Hdot^{1+\be}_\om}^{2}\right)\Sob{\Tht}{\dot{H}^{\s}_{\om}\cap L^2_{\om}}^2\notag\\&+C\Sob{m_1(D)^{-\frac{1}{2}}\Div F_{\Tht}(\tht^{(2)})}{\Hdot^{\s}_\om \cap L^2_{\om}}^2.
\end{align}
Applying \cref{lem:stab:G}, interpolation inequality, and Young's inequality, we see that when $\be=0$, we have
\begin{align}\label{est:uniqueness:div:a}
\Sob{m_1(D)^{-\frac{\gam}{2}}\Div F_{\Tht}(\tht^{(2)})}{\Hdot^{-1}_\om \cap L^2_{\om}}^2&\leq C\nrm{\Tht}_{\Hdot^
{-1}_\om\cap L^2_{\om}}^2 \nrm{m_1(D)^{\frac{\gam}{2}}\tht^{(2)}}_{H^{1}_\om}^2+C\nrm{\tht^{(2)}}_{H^{1}_\om}^2\nrm{m_1(D)^{\frac{\gam}{2}}\Tht}_{L^2_{\om}}^2\notag\\
&\le C\nrm{\Tht}_{\Hdot^
{-1}_\om\cap L^2_{\om}}^2\nrm{m_1(D)^{\frac{\gam}{2}}\tht^{(2)}}_{H^{1}_\om}^2+C\nrm{\tht^{(2)}}_{H^{1}_\om}^2\nrm{\Tht}_{L^2_{\om}}^{2-2\gam}\nrm{m_1(D)^{\frac{1}{2}}\Tht}_{L^2_{\om}}^{2\gam}\notag\\
&\le C\left(\nrm{m_1(D)^{\frac{\gam}{2}}\tht^{(2)}}_{H^{1}_\om}^2+\nrm{\tht^{(2)}}_{H^{1}_\om}^{\frac{2}{1-\gam}}\right)\nrm{\Tht}_{\dot{H}^{-1}_{\om}\cap L^2_{\om}}^2+\frac{c}{2}\nrm{m_1(D)^{\frac{1}{2}}\Tht}_{L^2_{\om}}^{2}.
\end{align}
Similarly, when $\be\in(0,1)$, we have
\begin{align}\label{est:uniqueness:div:b}
\Sob{m_1(D)^{-\frac{\gam}{2}}\Div F_{\Tht}(\tht^{(2)})}{L^2_{\om}}&\leq C\left(\nrm{m_1(D)^{\frac{\gam}{2}}\tht^{(2)}}_{\Hdot^{1+\be}_\om}^2+\nrm{\tht^{(2)}}_{H^{1+\be}_\om}^{\frac{2}{1-\gam}}\right)\nrm{\Tht}_{L^2_{\om}}^2+\frac{c}{2}\nrm{m_1(D)^{\frac{1}{2}}\Tht}_{L^2_{\om}}^{2},
\end{align}
and when $\be\in[1,2]$, we have
\begin{align}\label{est:uniqueness:div:d}
\Sob{m_1(D)^{-\frac{\gam}{2}}\Div F_{\Tht}(\tht^{(2)})}{H^{\be}_\om}
&\leq C\left(\nrm{m_1(D)^{\frac{\gam}{2}}\tht^{(2)}}_{\Hdot^{1+\be}_\om}^2+\nrm{\tht^{(2)}}_{H^{1+\be}_\om}^{\frac{2}{1-\gam}}\right)\nrm{\Tht}_{{H}^{\be}_{\om}}^2+\frac{c}{2}\nrm{m_1(D)^{\frac{1}{2}}\Tht}_{{H}^{\be}_{\om}}^{2}.
\end{align}
Combining \eqref{est:uniqueness:a} and \eqref{est:uniqueness:div:a} yields
\begin{align}
\frac{d}{dt}&\Sob{\Tht}{\dot{H}^{-1}_{\om}\cap L^2_{\om}}^2+\frac{c}{2}\Sob{m(D)^{\frac{1}2}\Tht}{\dot{H}^{-1}_{\om}\cap L^2_{\om}}^{2}\notag\\
\leq& C\left(1+\Sob{\tht^{(1)}}{\Hdot^{1}_\om \cap \Hdot^{-1}_\om}^{\frac{1}{1-\gam}}+\nrm{\tht^{(2)}}_{H^{1}_\om}^{\frac{2}{1-\gam}}+\Sob{m_1(D)^{\frac{\gam}2}\tht^{(1)}}{\Hdot^{1}_\om}^{2}+\nrm{m_1(D)^{\frac{\gam}{2}}\tht^{(2)}}_{H^{1}_\om}^2\right)\Sob{\Tht}{\dot{H}^{-1}_{\om}\cap L^2_{\om}}^2,\label{est:Hminus1:uniqueness}
\end{align}
when $\be=0$. Combining \eqref{est:uniqueness:b} with $\s=0$ and \eqref{est:uniqueness:div:b} yields
\begin{align}\label{est:Hsig:uniqueness:1}
\frac{d}{dt}&\Sob{\Tht}{ L^2_{\om}}^2+\frac{c}{2}\Sob{m(D)^{\frac{1}2}\Tht}{L^2_{\om}}^{2}\notag\\&\leq C\left(1+\Sob{\tht^{(1)}}{H^{1+\be}_\om}^{\frac{1}{1-\gam}}+\nrm{\tht^{(2)}}_{H^{1+\be}_\om}^{\frac{2}{1-\gam}}+\Sob{m_1(D)^{\frac{\gam}2}\tht^{(1)}}{\Hdot^{1+\be}_\om}^{2}+\nrm{m_1(D)^{\frac{\gam}{2}}\tht^{(2)}}_{\Hdot^{1+\be}_\om}^2\right)\Sob{\Tht}{L^2_{\om}}^2,
\end{align}
when $\be\in(0,1)$.
Lastly, combining \eqref{est:uniqueness:b} with $\s=\be\in(1,2]$ and \eqref{est:uniqueness:div:d} yields
\begin{align}\label{est:Hsig:uniqueness:3}
\frac{d}{dt}&\Sob{\Tht}{ {H}^{\be}_{\om}}^2+\frac{c}{2}\Sob{m(D)^{\frac{1}2}\Tht}{{H}^{\be}_{\om}}^{2}\notag\\&\leq C\left(1+\Sob{\tht^{(1)}}{H^{1+\be}_\om}^{\frac{1}{1-\gam}}+\nrm{\tht^{(2)}}_{H^{1+\be}_\om}^{\frac{2}{1-\gam}}+\Sob{m_1(D)^{\frac{\gam}2}\tht^{(1)}}{\Hdot^{1+\be}_\om}^{2}+\nrm{m_1(D)^{\frac{\gam}{2}}\tht^{(2)}}_{\Hdot^{1+\be}_\om}^2\right)\Sob{\Tht}{{H}^{\be}_{\om}}^2,
\end{align}
Recall that $\tht^{(j)}\in L^\infty_T(H^{1}_\om\cap\Hdot^{-1}_\om)\cap L^2_T(H^1_{\om m^{1/2}}\cap\Hdot^{-1}_{\om m^{1/2}})$, when $\be=0$, and $\tht^{(j)}\in L^\infty_T H^{1+\be}_\om\cap L^2_T H^{1+\be}_{\om m^{1/2}}$, when $\be\in(0,2]$, for $j=1,2$. In each of \eqref{est:Hminus1:uniqueness}, \eqref{est:Hsig:uniqueness:1}, \eqref{est:Hsig:uniqueness:3}, we may therefore apply Gronwall inequality to deduce that there exists a $C>0$, depending on $T>0$, such that
\begin{align}
\sup_{0\leq t\leq T}\Sob{\Tht(t)}{Y^\be}&\leq C\Sob{\Tht_0}{Y^\be}\label{eq:uniqueness}
\end{align}
for all $\be\in[0,2]$. In particular, if $\tht_0^{(1)}=\tht_0^{(2)}$, then $\tht^{(1)}=\tht^{(2)}$ in $Y^\be$ over $[0,T]$.
\subsection{Continuous Dependence on Initial Data}\label{sect:cont:dep}
Let $\be\in [0,2]$ and denote the data-to-solution operator of \eqref{eq:dgsqg} by $\Phi_0:H^{1}_\om\cap\Hdot^{-1}_\om\goesto\bigcup_{T>0}C([0,T];H^{1}_\om\cap\Hdot^{-1}_\om)$, when $\be=0$, and $\Phi_\be: H^{1+\be}_\om\goesto\bigcup_{T>0}C([0,T]; H^{1+\be}_\om)$, when $\be\in(0,2]$. Existence and uniqueness of solutions to \eqref{eq:dgsqg} establishes that $\Phi$ is well-defined. It will be convenient to use of the notation $\tht(\cdotp;\tht_0)=\Phi(\tht_0)(\cdotp)$.
We will now show that $\Phi_\be$ is continuous, for all $\be\in[0,2]$. To prove this, it will be convenient to introduce the following notation
\begin{align}\label{def:Xb}
X^\be:=\begin{cases}
H^1_\om\cap\Hdot^{-1}_\om,&\be=0\\
H^{1+\be}_\om,&\be\in(0,2].
\end{cases}
\end{align}
Ultimately, we will show that given $\tht_0\in H^1_\om\cap\Hdot^{-1}_\om$, when $\be=0$, and $\tht_{0} \in H^{1+\be}_\om$, when $\be\in(0,2]$, there exists a neighborhood $U_0\subset H^{1+\be}_\om$ of $\tht_{0}$ and a time, $T>0$, such that for any sequence of initial data $\{\tht^{n}_{0}\}\subset U_0$, we have
\[
\lim_{n \to \infty} \Sob{\tht^{n}_{0}-\tht_{0}}{X^\be}=0\quad\text{implies}\quad\lim_{n \to \infty}\Sob{\tht^n-\tht}{X^\be}=0.
\]
We will make use of the convention that $\tht^\infty=\tht=\Phi(\tht_0)$.
Let $T_0>0$ denote the local existence time of $\tht$. We define the neighborhoods $U_\be\subset X^\be$ by
\begin{align}\label{def:U}
U_\be&:=\{f\in X^\be: \Sob{f-\tht_{0}}{X^\be}<\Sob{\tht_{0}}{X^\be}\}.
\end{align}
Let
\begin{align}\label{def:K}
K_\be:=\sup_{f \in U_\be}\Sob{f}{X^\be}.
\end{align}
Denote by $\tht^{n}(\cdotp;\tht^n_0)$ the solution to \eqref{eq:dgsqg:mod:form} corresponding to initial data $\tht^n_0$. Then by \eqref{est:apriori:dgsqg:a} and \eqref{est:apriori:dgsqg:b} there exists a constant $C>0$ such that
\[
\sup_{n>0}\left(\Sob{\tht^n}{L^{\infty}_{T}X^\be}+\Sob{m(D)^{\frac{1}{2}}\tht^n}{L^{2}_{T}X^\be}\right)\le C K_\be,
\]
for some $0<T\leq T_0$ dependent on $\Sob{\tht_0}{X^\be}$.
Upon returning to \eqref{est:Hminus1:uniqueness}, \eqref{est:Hsig:uniqueness:1}, \eqref{est:Hsig:uniqueness:3}, and applying \eqref{eq:uniqueness}, respectively in $\be$, we obtain
\begin{align}\label{eq:stability}
\sup_{0\leq t\leq T}\Sob{(\tht^n-\tht)(t)}{Y^\be}^2+c\int_0^T\Sob{m(D)^{\frac{1}{2}}(\tht^n-\tht)(s)}{Y^\be}^2ds\leq C(T, K_\be)\Sob{\tht^n_0-\tht_0}{Y^\be}^2,
\end{align}
for some $C(T,K_\be)>0$, for all $\be\in[0,2]$, where $Y^\be$ was defined in \eqref{def:Y}. This implies
\begin{align}\label{eq:convergence}
\lim_{n\goesto\infty}\left( \Sob{\tht^n-\tht}{L^\infty_TY^\be}^2+\Sob{m(D)^{\frac{1}{2}}(\tht^n-\tht)}{L^2_TY^\be}^2\right)=0,
\end{align}
for all $\be\in[0,2]$.
To complete the proof, it suffices to show that $\nabla \tht^n \to \nabla \tht$ in $L^\infty_{T}\Hdot^{\be}_\om$. For this, let
\begin{align}\label{def:vsigma:zeta}
\vs^n:=(\vs^n_1,\vs^n_2),\quad\vs:=(\vs_{1},\vs_2),\quad \ze^n:=(\ze^n_1,\ze^n_2),\quad \ze:=(\ze_1,\ze_2).
\end{align}
Then we decompose $\nabla\tht$ into $\bdy_\ell\tht=\vs_\ell+\ze_\ell$ and $\nabla\tht^n$ into $\bdy_{\ell}\tht^{n}=\vs_{\ell}^{n}+\ze_{\ell}^{n}$, for $\ell=1,2$, where we assume the components of ($\vs^{n},\ze^{n}$) are governed by the equations
\begin{align}\label{eq:vs}
\partial_{t}\vs^{n}_{\ell}+m(D)\vs^{n}_{\ell}+ \Div F_{-\tht^{n}}(\vs^{n}_{\ell})=G_{\ell},\quad\vs^{n}_{\ell}(0,x)=\bdy_{\ell}\tht_0(x),
\end{align}
and
\begin{align}\label{eq:ze}
\partial_{t}\ze^{n}_{\ell}+m(D)\zeta^{n}_{\ell}+ \Div F_{-\tht^n}(\ze^{n}_{\ell})=G_{\ell}^{n}-G_{\ell},\quad \ze^{n}_{\ell}(0,x)=\bdy_{\ell}\tht^{n}_0(x)-\bdy_{\ell}\tht_{0}(x),
\end{align}
for $n\in\mathbb{N}\cup\{\infty\}$ and each $\ell=1,2$, where
\begin{align}\label{def:G}
G:=(G_1,G_2),\quad G_{\ell}:=\Div F_{\bdy_{\ell}\tht}(\tht), \quad G^n:=(G^n_1,G^n_2),\quad G^{n}_{\ell}:=\Div F_{\bdy_{\ell}\tht^n}(\tht^{n}),\quad\ell=1,2.
\end{align}
Note that to be consistent with $\tht^\infty=\tht$, we also make use of $\vs^\infty=\vs$ and $\ze^\infty=\ze$. Now observe that both \eqref{eq:vs} and \eqref{eq:ze} have the structure of \eqref{eq:mod:claw}. Indeed, we see that \eqref{eq:vs} has the structure upon making the replacement $q\mapsto -\tht^n$ and $G\mapsto G_\ell$ and \eqref{eq:ze} has the structure upon making the replacement $q\mapsto -\tht^n$ and $G\mapsto G^n_\ell-G_\ell$. Since $\tht_0\in X^\be$, $\nabla\tht_0\in Y^\be$, $\tht^n\in L^\infty_TX^\be$, and $m(D)^{\frac{1}2}\tht^n\in L^2_TX^\be$, for all $n\in\NN\cup\{\infty\}$, in order to apply \cref{thm:modclaw:wellposed}, it suffices to check that $m(D)^{-1/2}G_\ell\in L^2_TY^\be$ and $m(D)^{-1/2}(G^n_\ell-G_\ell)\in L^2_TY^\be$. This can be checked with \cref{lem:stab:G}.
First, we see that \cref{lem:stab:G} implies
\begin{align}\notag
\Sob{m(D)^{-\frac{\gam}{2}}G}{Y^\be}
&\leq C\Sob{\nabla\tht}{Y^{\be}}\Sob{m_1 (D)^{\frac{\gam}{2}}\tht}{H^{1+\be}_\om}+\nrm{\tht}_{H^{1+\be}_\om}\nrm{m(D)^{\frac{\gam}{2}}\nabla\tht}_{H^{\be}_\om}\le C\Sob{\tht}{H^{1+\be}_\om}\Sob{m_1 (D)^{\frac{\gam}{2}}\tht}{H^{1+\be}_\om}.
\end{align}
for all $\be\in[0,2]$. On the other hand, to study $m(D)^{-1/2}(G^n_\ell-G_\ell)$, observe that
\begin{align}\label{eq:G:difference}
G_{\ell}^{n}-G_{\ell}=\Div F_{\bdy_{\ell}\tht^{n}}(\tht^n-\tht)+\Div F_{(\bdy_{\ell}\tht^{n}-\bdy_{\ell}\tht)}(\tht).
\end{align}
Hence, \cref{lem:stab:G} implies
\begin{align}\notag
\Sob{m(D)^{-\frac{\gam}{2}}(G^n-G)}{Y^\be}
&\leq \Sob{m(D)^{-\frac{\gam}{2}}\Div F_{\nabla\tht^n}(\tht^n-\tht)}{Y^\be}+\Sob{m(D)^{-\frac{\gam}{2}}\Div F_{\nabla(\tht^n-\tht)}(\tht)}{Y^\be}.
\end{align}
By \cref{lem:stab:G} we see that
\begin{align}\notag
\Sob{m(D)^{-\frac{\gam}{2}}\Div F_{\nabla\tht^n}(\tht^n-\tht)}{Y^\be}&\leq C\Sob{\nabla\tht^{n}}{Y^{\be}}\Sob{m_1 (D)^{\frac{\gam}{2}}(\tht^n-\tht)}{H^{1+\be}_\om}+\nrm{\tht^n-\tht}_{H^{1+\be}_\om}\nrm{m(D)^{\frac{\gam}{2}}\nabla\tht^n}_{H^{\be}_\om}\\&\leq C\Sob{\tht^{n}}{H^{1+\be}_\om}\Sob{m_1 (D)^{\frac{\gam}{2}}(\tht^n-\tht)}{H^{1+\be}_\om}+\nrm{\tht^n-\tht}_{H^{1+\be}_\om}\nrm{m(D)^{\frac{\gam}{2}}\tht^n}_{H^{1+\be}_\om}\notag.
\end{align}
Similarly, we have
\begin{align}\notag
\Sob{m(D)^{-\frac{\gam}{2}}\Div F_{\nabla(\tht^n-\tht)}(\tht)}{Y^\be}&\leq C\Sob{\tht}{H^{1+\be}_\om}\Sob{m_1 (D)^{\frac{\gam}{2}}(\tht^n-\tht)}{H^{1+\be}_\om}+\nrm{\tht^n-\tht}_{H^{1+\be}_\om}\nrm{m(D)^{\frac{\gam}{2}}\tht}_{H^{1+\be}_\om}\notag.
\end{align}
Thus
\begin{align}\notag
\Sob{m(D)^{-\frac{\gam}{2}}(G^n-G)}{Y^\be}\leq C\left(\Sob{\tht^n-\tht}{H^{1+\be}_\om}+\nrm{m(D)^{\frac{\gam}{2}}(\tht^n-\tht)}_{H^{1+\be}_\om}\right)\left(\Sob{\tht}{H^{1+\be}_\om \cap H^{1+\be}_{\om m^{1/2}}}+\nrm{\tht^n}_{H^{1+\be}_\om\cap
H^{1+\be}_{\om m^1/2}}\right)
\end{align}
Hence, \cref{thm:modclaw:wellposed} applies to guarantee a unique solution $\vs^n\in C([0,T];Y^\be)$ to \eqref{eq:vs}, for all $\be\in[0,2]$.
By \eqref{eq:convergence}, we may invoke \cref{thm:stab} to ensure that
\begin{align}\notag
\lim_{n \to \infty} \left(\Sob{\vs^{n}-\vs}{L^{\infty}_{T}(H^1_\om\cap\Hdot^{-1}_\om)}+ \Sob{m(D)^{\gam/2}(\vs^{n}-\vs)}{L^{2}_{T}(H^1_\om\cap\Hdot^{-1}_\om)} \right)=0,
\end{align}
when $\be=0$, and
\begin{align}\notag
\lim_{n \to \infty} \left(\Sob{\vs^{n}-\vs}{L^{\infty}_{T}H^\be_\om}+ \Sob{m(D)^{\gam/2}(\vs^{n}-\vs)}{L^{2}_{T}H^\be_\om} \right)=0,
\end{align}
when $\be\in(0,2]$. Similarly, the same relations hold for $\ze^n-\ze$.
To conclude the proof, we apply the triangle inequality and observe that
\[
\limsup_{n\goesto\infty}\Sob{\nabla \tht^{n}-\nabla \tht}{L^{\infty}_{T}\Hdot^\be_\om}\leq \limsup_{n\goesto\infty}\Sob{\vs^n-\vs}{L^\infty_T \Hdot^\be_\om}+\limsup_{n\goesto\infty}\Sob{\ze^n-\ze}{L^\infty_T \Hdot^{\be}_\om}=0,
\]
as desired.
\appendix
\section{Proof of Lemma \ref{lem:max:principle}}\label{app:proof:maxp}
\begin{proof}
Now suppose that $p\geq2$ is an even integer and let $\Phi(f)=f^p/p$. Upon multiplying \eqref{eq:dgsqg} by $\tht^{p-1}$, we obtain
\begin{align}
\frac{1}p\frac{d}{dt}\Sob{\tht}{L^p}^p+\int_{\RR^2}\left(\Phi'(\tht(x)){L}\tht(x)-({L}\Phi(\tht))(x)\right)dx+\int_{\RR^2}({L}\Phi(\tht))(x)dx=0.\notag
\end{align}
Observe that
\begin{align}
\int_{\RR^2}({L}\Phi(\tht))(x)dx=\Ft\left({L}\Phi(\tht)\right)(0)=\ln(1)\Ft\left(\Phi(\tht)\right)(\xi)=0.\notag
\end{align}
Hence
\[
\frac{1}p\frac{d}{dt}\Sob{\tht}{L^p}^p\leq0,
\]
which implies $\Sob{\tht(t)}{L^p} \leq \Sob{\tht(0)}{L^p}$. It follows that
\[
\int_{B(R)}\tht(y)^pdy\leq\Sob{\tht_0}{L^p}^p,
\]
for all $R>0$, where $B(R)$ denotes the ball of radius $R$ centered at the origin. In particular, by the Lebesgue Differentiation Theorem, for almost every $x\in\RR^d$, we may fix $R_0(x)>0$ sufficiently small satisfying $|B(R_0(x))|<1$ and
\begin{align}
|\tht(x)|&\leq \Sob{\tht_0}{L^\infty}+\frac{1}{|B(R_0(x))|}\int_{B(R)}|\tht(y)|dy\notag\\
&\leq\Sob{\tht_0}{L^\infty}+ |B(R_0(x))|^{-1/p}\Sob{\tht_0}{L^p},\notag\\
&\leq\Sob{\tht_0}{L^\infty}+|B(R_0(x))|^{-1/p}\Sob{\tht_0}{L^2}^{2/p}\Sob{\tht_0}{L^\infty}^{(p-2)/p}\notag
\end{align}
holds, for any even integer $p$. Note that we applied H\"older's inequality to obtain the penultimate inequality. Upon choosing $R_0(x)$ appropriately so that $p_0(x)=-\ln|B(R_0(x))|$ is an even integer, we deduce
\begin{align}\label{est:omega:Linfty}
\Sob{\tht(t)}{L^\infty}\leq\Sob{\tht_0}{L^\infty}+ \sup_{p\geq2}\left(\frac{\Sob{\tht_0}{L^2}}{\Sob{\tht_0}{L^\infty}}\right)^{2/p}\Sob{\tht_0}{L^\infty},
\end{align}
as desired
\end{proof}
\section{Product Estimates}\label{app:prod}
We will now prove \cref{thm:prod}. Following Bony's decomposition, we can formally decompose the product of $f$ and $g$ as
\begin{align}\label{eq:Bony}
fg={T}_{f}g+{T}_{g}f+{R}(f,g),
\end{align}
where we denote by
\begin{align}\label{def:para:factors}
{T}_{f}g:=\sum_{k}S_{k-3}f\triangle_{k}g\quad and \quad {R}(f,g):=\sum_{k}\triangle_{k}f{\tlpk}g,\quad \tlpk g=\sum_{|i-k|\le3}\triangle_{i}g.
\end{align}
Note that the superscript $\tlpk$ denotes the \textit{extended} Littlewood-Paley block. In the following lemma, we obtain estimates for the term denoted by $T_{f}g$ that represents the interactions of low frequencies of $f$ and high frequencies of $g$.
\begin{Lem}
\label{lem:prod:log:A}
Given $d\geq2$, let ${s}, \bar{s} \in \RR$ such that ${s}\leq d/2$. Let $\om,\om_1,\tilde{\om}_1\in\mathscr{M}_W$. Let $\Gam$ be a function satisfying the following inequality for all $y>0$
\begin{align}\label{cond:Gam:A1}
\frac{\om(y)}{{\tilde{\om}_1(y)}}\left(\indFn{(-\infty,d/2)}({s})\int_0^1\frac{r^{d-2{s}-1}}{\om_1^2(y r)}dr+\indFn{\{d/2\}}({s})\int_{0}^{y}\frac{r^{d-1}dr}{(1+r^2)^{d/2}\om_1^{2}(r)}\right)^{1/2}\le C\Gam(y).
\end{align}
When ${s}<d/2$, there exists a constant $C>0$ and sequence $\{c_j\}\in\ell^2(\ZZ)$ with $\Sob{\{c_j\}}{\ell^2}\leq1$ such that
\begin{align}\label{est:prod:log:A1}
&\Sob{\triangle_j(T_{f}g)}{L^2(\RR^d)}\le Cc_j2^{-({s}+\bar{s}-d/2) j}\om(2^j)^{-1}\Gam(2^{j})\Sob{f}{{\Hdot}^{{s}}_{\om_1} (\RR^d)}\Sob{g}{\Hdot^{\bar{s}}_{\tilde{\om}_1}(\RR^d)}.
\end{align}
When ${s}=d/2$, there exists a constant $C>0$ and sequence $\{c_j\}\in\ell^2(\ZZ)$ with $\Sob{\{c_j\}}{\ell^2}\leq1$ such that
\begin{align}\label{est:prod:log:A2}
&\Sob{\triangle_j(T_{f}g)}{L^2(\RR^d)}\le Cc_j2^{-\bar{s}j}\om(2^j)^{-1}\Gam(2^{j})
\Sob{f}{H^{d/2}_{\om_1}(\RR^d)}\Sob{g}{\Hdot^{\bar{s}}_{\tilde{\om}_1}(\RR^d)}.
\end{align}
\end{Lem}
\begin{proof}
First, we apply H\"older's inequality and obtain
\begin{align}\label{est:low:high:a}
\Sob{\triangle_j({T}_{f}g)}{L^2}\le \sum_{\abs{k-j}\le 2}\Sob{S_{k-3}f}{L^{\infty}}\Sob{\triangle_{k}g}{L^2}\le C\sum_{\abs{k-j}\le 2}\Sob{\chi_{k-3}\hat{f}}{L^{1}}\Sob{\triangle_{k}g}{L^2} .
\end{align}
Let
\begin{align}\label{def:comm:A:cj}
c_{j}:=\frac{\sum_{\abs{k-j}\le 2}2^{\bar{s} k}\Sob{\tilde{\om}_1(D)\triangle_{k}g}{L^2}}{\Sob{g}{\Hdot^{\bar{s}}_{\tilde{\om}_1}}}.
\end{align}
Observe that $\{c_j\}\in\ell^2(\ZZ)$ and $\Sob{\{c_j\}}{\ell^2}\leq1$. Now, by \eqref{eq:bernstein:om} and \cref{lem:Bernstein}, we may estimate $\Sob{\lpk g}{L^2}$ to obtain
\begin{align}\label{est:lpkg:a}
\Sob{\lpj(T_fg)}{L^2}\leq Cc_{j}\tilde{\om}_1(2^j)^{-1}2^{-\bar{s} j}\Sob{\chi_{j-3}\hat{f}}{L^1}\Sob{g}{\Hdot^{\bar{s}}_{\tilde{\om}_1}(\RR^d)}.
\end{align}
We are left to estimate $\Sob{\chi_{k-3}\hat{f}}{L^{1}}$. We will treat the cases ${s}<d/2$ and ${s}=d/2$ separately.
\subsubsection*{Case: ${s}<d/2$}
By the Cauchy-Schwarz inequality, we obtain
\begin{align}
\int |\chi_{j-1}(\xi)\hat{f}(\xi)|d\xi&\leq \left(\int_{\Bcal_j}\frac{1}{|\xi|^{2{s}}\om_1^{2}(|\xi|)}d\xi\right)^{1/2}\left(\int |\xi|^{2{s}}\om_1^{2}(|\xi|)|\hat{f}(\xi)|^2d\xi\right)^{1/2}\notag.
\end{align}
We have
\begin{align}
\int_{\Bcal_j}\frac{1}{|\xi|^{2{s}}\om_1^{2}(|\xi|)}d\xi&\leq C 2^{(d-2{s})j}\int_0^{2^j}\frac{1}{\om_1^2(2^j(r2^{-j}))}\left(\frac{r}{2^j}\right)^{d-2{s}-1}\frac{dr}{2^j}= C2^{(d-2{s})j}\int_0^{1}\frac{r^{d-2{s}-1}}{\om_1^2(2^jr)}dr.\notag
\end{align}
It follows that
\begin{align}\label{est:Skf:prelim:a1}
\int |\chi_{j-1}(\xi)\hat{f}(\xi)|d\xi\leq C2^{(d/2-{s})j}\til{\om}_1(2^j)\om(2^j)^{-1}\Gam(2^j)\Sob{f}{\Hdot^{{s}}_{\om_1}},
\end{align}
as desired.
\subsubsection*{Case: ${s}=d/2$}
Similarly, we estimate
\begin{align}\label{est:Skf:prelim:a2}
\int |\chi_{j-1}(\xi)\hat{f}(\xi)|d\xi&\leq \left(\int_{\Bcal_j}\frac{1}{(1+|\xi|^2)^{d/2}\om_1^{2}(|\xi|)}d\xi\right)^{1/2}\left(\int (1+|\xi|^2)^{d/2}\om_1^{2}(|\xi|)|\hat{f}(\xi)|^2d\xi\right)^{1/2}\notag \\
&\le C\left(\int_{0}^{2^{j}}\frac{r^{d-1}dr}{(1+r^2)^{d/2}\om_1^{2}(r)}\right)^{1/2}\Sob{f}{{H}^{d/2}_{\om_1}}\notag\\
&\leq C\til{\om}_1(2^j)\om(2^j)^{-1}\Gam(2^j)\Sob{f}{H^{d/2}_{\om_1}}.
\end{align}
From \eqref{est:lpkg:a}, \eqref{est:Skf:prelim:a1}, and \eqref{est:Skf:prelim:a2}, we obtain the desired estimate.
\end{proof}
Next we obtain estimates for the high frequency interactions represented by $R(f,g)$.
\begin{Lem}\label{lem:prod:log:B}
Given $d\geq2$, let ${s},\bar{s}\in\RR$ such that ${s},\bar{s}\leq d/2$ and ${s} +\bar{s}>0$. Let $\om,\om_1,\tilde{\om}_1\in\mathscr{M}_W$. Let $\Gam$ be a function satisfying the following inequality for all $y\ge 0$:
\begin{align*}\notag
\frac{\om(y)}{{\om_1(y)}{\tilde{\om}_1(y)}}\le C\Gam(y).
\end{align*}
Then, there exists a constant $C>0$ and a sequence $\{c_j\}\in\ell^2(\ZZ)$ with $\Sob{\{c_j\}}{\ell^2}\leq1$ such that
\begin{align}\notag
\Sob{\triangle_jR(fg)}{L^2(\RR^d)}\le Cc_j2^{-({s}+\bar{s}-\frac{d}{2})j} \om(2^j)^{-1}\Gam(2^j)\Sob{f}{\Hdot^{{s}}_{\om_1}(\RR^d)}\Sob{g}{\Hdot^{\bar{s}}_{\tilde{\om}_1}(\RR^d)},
\end{align}
for all $j\in\ZZ$.
\end{Lem}
\begin{proof}
Let $\rho={s}+\bar{s}$. By Bernstein's inequality and \eqref{est:omega}, we have
\begin{align*}
&\Sob{\triangle_{j}{R}(f,g)}{L^2}\\&\le C\sum_{k\ge j-5}2^{(d/2)j}\Sob{\triangle_{k}f}{L^2}\Sob{{\tlpk}g}{L^2}\\
&\leq C(\om_1(2^j)\tilde{\om}_1(2^j))^{-1}2^{(d/2-\rho)j}\sum_{k\geq j-5}2^{-\rho(k-j)}\left(\frac{\om_1(2^j)\tilde{\om}_1(2^j)}{\om_1(2^k)\tilde{\om}_1(2^k)}\right)2^{{s} k}\Sob{\triangle_{k}f}{L^2_{\om_1}}2^{\bar{s} k}\Sob{{\tlpk}g}{L^2_{\til{\om}_1}}\notag\\
&\leq Cc_j2^{-(\rho-d/2)j} \om(2^j)^{-1}\Gam(2^j)\Sob{f}{\Hdot^{{s}}_{\om_1}}\Sob{g}{\Hdot^{\bar{s}}_{\tilde{\om}_1}},
\end{align*}
where
\[
c_{j}:=\sum_{k\ge j-5}\frac{2^{-\rho(k-j)}\left(\frac{\om_1(2^j)\tilde{\om}_1(2^j)}{\om_1(2^k)\tilde{\om}_1(2^k)}\right)2^{{s} k}\Sob{\triangle_{k}f}{L^2_{\om}}2^{\bar{s} k}\Sob{{\tlpk}g}{L^2_{\til{\om}}}}{C_0\Sob{f}{\Hdot^{{s}}_{\om_1}}\Sob{g}{\Hdot^{\bar{s}}_{\tilde{\om}_1}}}.
\]
It remains to show that $\{c_j\}\in\ell^2(\ZZ)$. Since $\om_1,\til{\om}_1\in\mathscr{M}_W$, we recall that
\[
\om_1=\frac{\om_1^a}{\om_1^b},\quad \til{\om}_1=\frac{\til{\om}_1^a}{\til{\om}_1^b}.
\]
Thus
\[
\left(\frac{\om_1(2^j)\tilde{\om}_1(2^j)}{\om_1(2^k)\tilde{\om}_1(2^k)}\right)=\left(\frac{\om_1^a(2^j)\tilde{\om}_1^a(2^j)}{\om_1^a(2^k)\tilde{\om}_1^a(2^k)}\right)\left(\frac{\om_1^b(2^k)\tilde{\om}_1^b(2^k)}{\om_1^b(2^j)\tilde{\om}_1^b(2^j)}\right).
\]
For the first factor, we observe that $\om_1^a,\til{\om}_1^a$ are increasing by \ref{item:O1} and satisfy \eqref{est:omega}, it follows that
\begin{align}
\om_1^a(2^j)\leq \om_1^a(2^{k+5})\le C\om_1^a(2^k),\quad \text{whenever}\ k-j\geq-5,\notag
\end{align}
for some $C$ independent of $k,j$; the same argument applied to $\til{\om}_1^a$. This implies that
\[
\frac{\om_1^a(2^j)\tilde{\om}_1^a(2^j)}{\om_1^a(2^k)\tilde{\om}_1^a(2^k)}\leq C.
\]
For the second factor, we observe that since $\om_1^b,\til{\om}_1^b$ satisfy \ref{item:O1}-\ref{item:O3} and \eqref{eq:omega:eps:bdd}, it follows that
\begin{align}
\om_1^b(2^k)\leq C(\om_1^b(2^{j})+\om_1^b(2^{k-j}))\le C\om_1^b(2^j)(1+2^{\eps(k-j)}),\quad \text{for}\ \eps\in(0,\rho),\notag
\end{align}
A similar estimate follows for $\til{\om}_1^b$. This implies that
\[
\frac{\om_1^b(2^k)\tilde{\om}_1^b(2^k)}{\om_1^b(2^j)\tilde{\om}_1^b(2^j)}\leq C(1+2^{\eps(k-j)}).
\]
We conclude that for an appropriate normalizing constant $C_0$, we have $\Sob{\{c_j\}}{\ell^2}\leq1$, as desired.
\end{proof}
Applying \cref{lem:prod:log:A} and \cref{lem:prod:log:B} to the terms in the decomposition \eqref{eq:Bony}, we obtain the estimate \eqref{est:product:Thm} claimed in \cref{thm:prod}.
\section{Proof of Theorem \ref{thm:modclaw:wellposed}}\label{app:well:posed}
We will now provide a sketch of the proof of \cref{thm:modclaw:wellposed}.
\begin{proof}[Proof of \cref{thm:modclaw:wellposed}] Let the space $X^\be$ be as defined in (\ref{def:Xb}).
We mollify $q$ and $G$ with respect to time by setting
\[q^{n}=\rho_{n}* q,\quad G^{n}=\rho_{n}* G,\]
where $\{\rho_{n}(t)\}_n$ is a sequence of standard mollifiers. It follows that
\begin{align*}
&q^{n}\in C([0,T];X^\be)\\
&m_1(D)^{\gam/2}q^{n}\in \Hdot^{1+\be}_\om\\
&m_1(D)^{-{1}/{2}}G^{n}\in C([0,T];\Hdot^{\s}_\om \cap L^2_{\om})
\end{align*}
Moreover, $\{q^{n}\}_n$ is uniformly bounded in $L^{\infty}(0,T;X^\be)$, $\{m_1(D)^{\gam/2}q^{n}\}_n$ is uniformly bounded in $L^{p_0}(0,T;\Hdot^{1+\be}_\om)$ and $\{m_1(D)^{-{1}/{2}}G^{n}\}_n$ is uniformly bounded in $L^{2}(0,T;\Hdot^{\s}_\om \cap L^2_{\om})$.
Let us consider an artificial viscosity regularization of \eqref{eq:mod:claw}:
\begin{align}\label{eq:forced-transport:reg}
\begin{cases}
\partial_{t}\theta^{n}+m(D)\tht^n-\frac{1}{n}\De \tht^{n}+ \Div F_{q^{n}}(\tht^{n})=G^{n}. \\
\tht^{n}(0,x)=\tht_{0}(x).
\end{cases}
\end{align}
For $0\le t\le T$, define
\begin{align*}
{\mathcal{L}}_{1}(G^{n})&:=\int^{t}_{0}e^{\frac{1}{n}\De(t-s)}G^{n}(s)\, ds,\\
{\mathcal{L}}_{2}(\tht^{n};q^{n})&:=\int^{t}_{0}e^{\frac{1}{n}\De(t-s)}\Div F_{q^n}(\tht^n)\,ds.
\end{align*}
Since $m_1$ satisfies \eqref{eq:omega:eps:bdd}, it follows that for any positive small $\eps_0$, we have
\begin{align*}
\Sob{\mathcal{L}_1(G^n)(t)}{\Hdot^{\s}_\om}&\le C\int_{0}^{t}\Sob{(I+{\Lam}^{\eps_0})m_1(D)^{-\frac{1}{2}}e^{\frac{1}{n}\De(t-s)}G^n(s)}{\Hdot^{\s}_\om}ds\\
&\le C_n(T+T^{1-\eps_0/2})\Sob{m_1(D)^{-\frac{1}{2}}G^n}{L^{\infty}_T \Hdot^{\s}_\om}.
\end{align*}
To estimate $\Sob{\mathcal{L}_2(\tht^n;q^n)}{\Hdot^{\s}_\om}$, we consider the two cases $\be \in [0,1]$ and $\be \in (1,2]$ separately. Henceforth, $\eps$ will denote a sufficiently small positive number and $\eps_1, \eps_2$, and $\eps_3$ will denote some appropriately chosen real numbers.
\subsection*{\textbf{Case: ${\be}\in [0,1]$}} Let $\eps_1$ be chosen such that
\[
\eps_1 \in \begin{cases}(0,\be/2),\quad &\text{if}\quad \be\in (0,1],\\
(-1,0),\quad &\text{if}\quad \be=0.
\end{cases}
\]
\subsubsection*{Subcase: $\s \in [0,1) $} Applying \cref{cor:prod} with $(s,\bar{s})=(1-\be+\eps_1, \s)$, $(\om_1,\til{\om}_1)=(p^{-1}\om,\om)$, $(\om_2, \til{\om}_2)=(\om, p^{-1}\om)$, $(\om_1,\til{\om}_1)=(p^{-1}\om,\om)$, and $\Gam=m_1^{\gam}$, we have
\begin{align*}
\Sob{\mathcal{L}_2(\tht^n;q^n)}{\Hdot^{\s}_\om}&\le \int_0^{t}\Sob{({\Lam}^{1+\be-\eps_1}(I+\Lam^\eps)e^{\frac{1}{n}\De(t-s)})\nabla {\Lam}^{-1}\cdot (v^n\tht^n)(s)}{\Hdot^{\s-\be+\eps_1}_{\om m_1^{-\gam}}}ds\\
&\le C_n T^{(1-\be+\eps_1)/2}(1+T^{-\eps/2})\Sob{q^n}{L^{\infty}_T X^{\be}_\om}\Sob{\tht^n}{L^\infty_T \Hdot^{\s}_\om}.
\end{align*}
\subsubsection*{Subcase: $\s \in [1,2] $} Applying \cref{cor:prod} with $(s,\bar{s})=(1-\be+\eps_1, \s-1)$, $(\om_1,\til{\om}_1)=(p^{-1}\om,\om)$, $(\om_2, \til{\om}_2)=(\om, p^{-1}\om)$, $(\om_1,\til{\om}_1)=(p^{-1}\om,\om)$, and $\Gam=m_1^{\gam}$, we have
\begin{align*}
\Sob{\mathcal{L}_2(\tht^n;q^n)}{\Hdot^{\s}_\om}&\le \int_0^{t}\Sob{({\Lam}^{1+\be-\eps_1}(I+\Lam^\eps)e^{\frac{1}{n}\De(t-s)}) (v^n\cdot\nabla\tht^n)(s)}{\Hdot^{\s-1+\be+\eps_1}_{\om m_1^{-\gam}}}ds\\
&\le C_n T^{(1-\be+\eps_1)/2}(1+T^{-\eps/2})\Sob{q^n}{L^{\infty}_T X^{\be}_\om}\Sob{\tht^n}{L^\infty_T (\Hdot^{\s}_\om \cap L^2_{\om})}.
\end{align*}
\subsection*{Case: ${\be}\in (1,2]$} Let $\eps_2$ be chosen such that
\[
\eps_2 \in (0,\be-1).
\]
\subsubsection*{Subcase: $\s \in [0,1)$}
Applying \cref{cor:prod} with $(s,\bar{s})=(1-\eps_2,\s)$, $(\om_1,\til{\om}_1)=(p^{-1}\om,\om)$, $(\om_2, \til{\om}_2)=(\om, p^{-1}\om)$, $(\om_1,\til{\om}_1)=(p^{-1}\om,\om)$, $\Gam=m_1^{\gam}$, and $(s,\bar{s})=(\be-1-\eps_2,\s)$, $(\om_1,\til{\om}_1)=(p^{-1}\om,\om)$, $(\om_2, \til{\om}_2)=(\om, p^{-1}\om)$, $(\om_1,\til{\om}_1)=(\om,\om)$, $\Gam=p^{-1}m_1^{\gam}$, we have
\begin{align*}
\Sob{\mathcal{L}_2(\tht^n;q^n)(t)}{\Hdot^{\s}_\om}&\le C\int_{0}^{t}\left(\Sob{({\Lam}^{1+\eps_2}(I+\Lam^\eps)e^{\frac{1}{n}\De(t-s)})\nabla {\Lam}^{-1} \cdot((\nabla^{\perp}a(D){q^n}) \tht^n)}{\Hdot^{\s-\eps_2}_{\om m_1^{-\gam}}}\right.\\&\left.\hspace{5 em}+\Sob{({\Lam}^{1+\eps_2}(I+\Lam^\eps)e^{\frac{1}{n}\De(t-s)})\nabla {\Lam}^{-1}\cdot((\nabla^{\perp}{q^n})\tht^n)}{\Hdot^{\s-\eps_2+\be-2}_{\om p m_1^{-\gam}}}\right)ds\\
&\le C_n T^{(1-\eps_2)/2}(1+T^{-\eps/2})\Sob{q^n}{L^{\infty}_T H^\be_\om}\Sob{\tht^n}{L^\infty_T \Hdot^{\s}_\om}.
\end{align*}
\subsubsection*{Subcase: $\s \in [1,2] $}
Applying \cref{cor:prod} with $(s,\bar{s})=(1-\eps_2,\s-1)$, $(\om_1,\til{\om}_1)=(p^{-1}\om,\om)$, $(\om_2, \til{\om}_2)=(\om, p^{-1}\om)$, $(\om_1,\til{\om}_1)=(p^{-1}\om,\om)$, $\Gam=m_1^{\gam}$, and $(s,\bar{s})=(\be-1-\eps_2,\s-1)$, $(\om_1,\til{\om}_1)=(p^{-1}\om,\om)$, $(\om_2, \til{\om}_2)=(\om, p^{-1}\om)$, $(\om_1,\til{\om}_1)=(\om,\om)$, $\Gam=p^{-1}m_1^{\gam}$, we have
\begin{align*}
\Sob{\mathcal{L}_2(\tht^n;q^n)(t)}{\Hdot^{\s}_\om}&\le C\int_{0}^{t}\left(\Sob{({\Lam}^{1+\eps_2}(I+\Lam^\eps)e^{\frac{1}{n}\De(t-s)})(\nabla^{\perp}a(D){q^n}\cdot \nabla \tht^n)}{\Hdot^{\s-1-\eps_2}_{\om m_1^{-\gam}}}\right.\\&\left.\hspace{5 em}+\Sob{({\Lam}^{1+\eps_2}(I+\Lam^\eps)e^{\frac{1}{n}\De(t-s)}) (\nabla^{\perp}{q^n}\cdot\nabla\tht^n)}{\Hdot^{\s-\eps_2+\be-3}_{\om p m_1^{-\gam}}}\right)ds\\
&\le C_nT^{(1-\eps_2)/2}(1+T^{-\eps/2})\Sob{q^n}{L^{\infty}_T H^\be_\om}\Sob{\tht^n}{L^\infty_T H^{\s}_\om}.
\end{align*}
\subsubsection*{Subcase: $\s \in (2,1+\be]$} Let $\eps_3$ be chosen such that
\[
\eps_3 \in (0,\s-2).
\]
Using Plancherel's theorem and the fact that $H^s$ is a Banach algebra if $s>1$, we have
\begin{align*}
\Sob{\mathcal{L}_2(\tht^n;q^n)(t)}{\Hdot^{\s}_\om}&\le C\int_{0}^{t}\left(\Sob{({\Lam}^{1+\eps_3}e^{\frac{1}{n}\De(t-s)})(\nabla^{\perp}a(D){q^n}\cdot \nabla \tht^n)}{H^{\s-1-\eps_3+\eps}}\right.\\&\left. \hspace{5 em}+\Sob{({\Lam}^{1+\eps_3+\be-2}e^{\frac{1}{n}\De(t-s)}) (\nabla^{\perp}{q^n}\cdot\nabla\tht^n)}{H^{\s-1+\eps-\eps_3}}\right)ds\\
&\le C_nT^{(1-\eps_3)/2}\Sob{q^n}{L^{\infty}_T \Hdot^{1+\be}_\om}\Sob{\tht^n}{L^\infty_T H^{\s}_\om}+ C_nT^{(3-\eps_3-\be)/2}\Sob{q^n}{L^{\infty}_T \Hdot^{1+\be}_\om}\Sob{\tht^n}{L^\infty_T H^{\s}_\om}.
\end{align*}
Applying Picard's theorem \cite{Lemarie-Rieusset2002}, we obtain the existence of a unique solution $\tht^{n}$ to \eqref{eq:forced-transport:reg} such that $\tht^{n}\in L^{\infty}(0,T_n;\Hdot^{\s}_\om \cap L^2_{\om})$ for some time $T_{n}>0$. However, owing to the uniform estimates developed in \cref{sect:apriori}, we can therefore assume that
\begin{align*}
T_{n}=T,\quad \text{for all}\ n.
\end{align*}
Let us denote by
\[
\Tht^{n}(t)=\tht^{n}(t)-\int_{0}^{t}G^{n}(s)ds.
\]
Then, $\Sob{\Tht^{n}}{L^{\infty}_{T}\Hdot^{\s}_\om \cap L^2_{\om}}$ is bounded uniformly in $n$.
Using similar estimates as above, it is easy to establish that $\Sob{\partial_{t}\Tht^{n}}{L^{\infty}_{T}\Hdot^{-k}_\om}$ is bounded uniformly in $n$, for some sufficiently large $k>0$. By an application of the classical Aubin-Lions lemma (see \cite{ConstantinFoiasBook1988}), there exists $\Tht\in L^{\infty}(0,T;\Hdot^{\s}_\om)$ such that for any given test function $\varphi \in C_{c}^{\infty}([0,T] \times \RR^2)$, one can extract a subsequence of $\{\Tht^n\}$, denoted by $\{\Tht^{n_{k}}\}$ satisfying
\begin{align*}
& \Tht^{n_{k}}\xrightharpoonup {\text{w*}}\Tht\quad\text{in}\quad L^{\infty}([0,T];\Hdot^{\s}_\om \cap \Hdot^0_\om),\\
& \varphi\Tht^{n_{k}} \longrightarrow \varphi\Tht\quad \text{in} \quad C([0,T];\Hdot^{\s-\delta}_\om \cap \Hdot^{-\delta}_\om),
\end{align*}
for any $\delta>0$. It then follows that $\tht(t)=\Tht(t)+\int_{0}^{t}G(s)ds$ is a weak solution of \eqref{eq:mod:claw}.
\end{proof}
\begin{thebibliography}{10}
\bibitem{BahouriCheminDanchinBook2011}
H.~Bahouri, J.-Y. Chemin, and R.~Danchin.
\newblock {\em Fourier analysis and nonlinear partial differential equations},
volume 343 of {\em Grundlehren der Mathematischen Wissenschaften [Fundamental
Principles of Mathematical Sciences]}.
\newblock Springer, Heidelberg, 2011.
\bibitem{Biswas2014}
A.~Biswas.
\newblock Gevrey regularity for the supercritical quasi-geostrophic equation.
\newblock {\em J. Differential Equations}, 257(1):1753--1772, 2014.
\bibitem{BiswasMartinezSilva2015}
A.~Biswas, V.R. Martinez, and P.~Silva.
\newblock On {G}evrey regularity of the supercritical {S}{Q}{G} equation in
critical {B}esov spaces.
\newblock {\em J. Funct. Anal.}, 269(10):3083--3119, 2015.
\bibitem{BourgainLi2015}
J.~Bourgain and D.~Li.
\newblock Strong ill-posedness of the incompressible {E}uler equation in
borderline {S}obolev spaces.
\newblock {\em Invent. Math.}, 201(1):97--157, 2015.
\bibitem{BourguignonBrezis1974}
J.P. Bourguignon and H.~Brezis.
\newblock Remarks on the euler equation.
\newblock {\em J. Funct. Anal.}, 15(4):341--363, 1974.
\bibitem{CaffarelliVasseur2010}
L.~Caffarelli and A.~Vasseur.
\newblock Drift diffusion equations with fractional diffusion and the
quasi-geostrophic equation.
\newblock {\em Ann. of Math.}, 171(3):1903--1930, 2010.
\bibitem{ChaeConstantinCordobaGancedoWu2012}
D.~Chae, P.~Constantin, D.~C\'{o}rdoba, F.~Gancedo, and J.~Wu.
\newblock Generalized surface quasi-geostrophic equations with singular
velocities.
\newblock {\em Comm. Pure Appl. Math.}, 65(8):1037--1066, 2012.
\bibitem{ChaeConstantinWu2011}
D.~Chae, P.~Constantin, and J.~Wu.
\newblock Inviscid models generalizing the two-dimensional euler and the
surface quasi-geostrophic equations.
\newblock {\em Archive for Rational Mechanics and Analysis}, 202(1):35--62,
2011.
\bibitem{ChaeConstantinWu2012}
D.~Chae, P.~Constantin, and J.~Wu.
\newblock Dissipative models generalizing the 2{D} {N}avier-{S}tokes and
surface quasi-geostrophic equations.
\newblock {\em Indiana Univ. Math. J.}, 61(5):1997--2018, 2012.
\bibitem{ChaeJeongNaOh2023a}
D.~Chae, I.-J. Jeong, J.~Na, and S.-J. Oh.
\newblock Well-posedness for ohkitani model and long-time existence for surface
quasi-geostrophic equations.
\newblock {\em arXiv:2308.02107}, pages pp. 1--21, August 4, 2023.
\bibitem{ChaeJeongOh2023b}
D.~Chae, I.-J. Jeong, and S.-J. Oh.
\newblock Illposedness via degenerate dispersion for generalized surface
quasi-geostrophic equations with singular velocities.
\newblock {\em arXiv:2308.02120}, pages pp. 1--74, August 4, 2023.
\bibitem{ChaeLee2003}
D.~Chae and J.~Lee.
\newblock Global well-posedness in the super-critical dissipative
quasi-geostrophic equations.
\newblock {\em Comm. Math. Phys.}, 233(2):297--311, 2003.
\bibitem{ChaeWu2012}
D.~Chae and J.~Wu.
\newblock Logarithmically regularized inviscid models in borderline {S}obolev
spaces.
\newblock {\em J. Math. Phys.}, 53(11):115601, 15, 2012.
\bibitem{CheminBook1995}
J.-Y. Chemin.
\newblock Fluides parfaits incompressibles.
\newblock {\em Ast\'{e}risque}, 343(230):177, 1995.
\bibitem{ChenMiaoZhang2007}
Q.~Chen, C.~Miao, and Z.~Zhang.
\newblock A new {B}ernstein's inequality and the 2{D} dissipative
quasi-geostrophic equation.
\newblock {\em Comm. Math. Phys.}, 271(3):821--838, 2007.
\bibitem{CobosDominguez2015}
F.~Cobos and \'{O}. Dom\'{\i}nguez.
\newblock On {B}esov spaces of logarithmic smoothness and {L}ipschitz spaces.
\newblock {\em J. Math. Anal. Appl.}, 425(1):71--84, 2015.
\bibitem{ColombiniDelSantoFanelliMetivier2015}
F.~Colombini, D.~Del~Santo, F.~Fanelli, and G.~M\'{e}tivier.
\newblock The well-posedness issue in {S}obolev spaces for hyperbolic systems
with {Z}ygmund-type coefficients.
\newblock {\em Comm. Partial Differential Equations}, 40(11):2082--2121, 2015.
\bibitem{ConstantinFoiasBook1988}
P.~Constantin and C.~Foias.
\newblock {\em {N}avier-{S}tokes equations}.
\newblock Chicago Lectures in Mathematics. University of Chicago Press,
Chicago, IL, 1988.
\bibitem{ConstantinIyerWu2008}
P.~Constantin, G.~Iyer, and J.~Wu.
\newblock Global regularity for a modified critical dissipative
quasi-geostrophic equation.
\newblock {\em Indiana Univ. Math. J.}, 57(6):2681--2692, 2008.
\bibitem{ConstantinTarfuleaVicol2015}
P.~Constantin, A.~Tarfulea, and V.~Vicol.
\newblock Long {T}ime {D}ynamics of {F}orced {C}ritical {SQG}.
\newblock {\em Commun. Math. Phys.}, 335(1):93--141, 2015.
\bibitem{ConstantinVicol2012}
P.~Constantin and V.~Vicol.
\newblock Nonlinear maximum principles for dissipative linear nonlocal
operators and applications.
\newblock {\em Geom. Funct. Anal.}, 22:1289--1321, 2012.
\bibitem{ConstantinWu1999}
P.~Constantin and J.~Wu.
\newblock Behavior of solutions of 2{D} quasi-geostrophic equations.
\newblock {\em SIAM J. Math. Anal.}, 30(5):937--948, 1999.
\bibitem{CordobaZoroa-Martinez2021}
D.~C\'{o}rdoba and L.~Mart\'{\i}nez-Zoroa.
\newblock Non existence and strong ill-posedness in {$C^k$} and {S}obolev
spaces for {SQG}.
\newblock {\em Adv. Math.}, 407:Paper No. 108570, 74, 2022.
\bibitem{CordobaZoroa-Martinez2022}
D.~C\'ordoba and L.~Mart\'inez-Zoroa.
\newblock Non-existence and strong ill-posedness in $c^{k,\beta}$ for the
generalized {S}urface {Q}uasi-geostrophic equation.
\newblock {\em arXiv:2207.14385}, pages 1--39, 2022.
\bibitem{DabkowskiKiselevSilvestreVicol2014}
M.~Dabkowski, A.~Kiselev, L.~Silvestre, and V.~Vicol.
\newblock Global well-posedness of slightly supercritical active scalar
equations.
\newblock {\em Analysis \& PDE}, 7(1):43--72, 2014.
\bibitem{DongLi2008}
H.~Dong and D.~Li.
\newblock Spatial analyticity of the solutions to the subcritical dissipative
quasi-geostrophic equations.
\newblock {\em Arch. Ration. Mech. Anal.}, 189(1):131--158, 2008.
\bibitem{DongLi2010}
H.~Dong and D.~Li.
\newblock On the 2{D} critical and supercritical dissipative quasi-geostrophic
equation in {B}esov spaces.
\newblock {\em J. Differential Equations}, 248(11):2684--2702, 2010.
\bibitem{DongLi2015}
H.~Dong and D.~Li.
\newblock Global $\dot{H}^1\cap\dot{H}^{-1}$ to a {L}ogarithmically
{R}egularized 2{D} {E}uler {E}quation.
\newblock {\em J. of Math. Fluid Mech.}, 17(1):1--7, 2015.
\bibitem{EbinMarsden1969}
D.G. Ebin and J.E. Marsden.
\newblock {Groups of diffeomorphisms and the solution of the classical Euler
equations for a perfect fluid}.
\newblock {\em Bull. Am. Math. Soc.}, 75(5):962 -- 967, 1969.
\bibitem{Elgindi2014}
T.M. Elgindi.
\newblock Osgood's lemma and some results for the slightly supercritical 2d
euler equations for incompressible flow.
\newblock {\em Archive for Rational Mechanics and Analysis}, 211(3):965--990,
2014.
\bibitem{ElgindiMasmoudi2020}
T.M. Elgindi and N.~Masmoudi.
\newblock {$L^\infty$} ill-posedness for a class of equations arising in
hydrodynamics.
\newblock {\em Arch. Ration. Mech. Anal.}, 235(3):1979--2025, 2020.
\bibitem{FoiasTemam1989}
C.~Foias and R.~Temam.
\newblock Gevrey {C}lass {R}egularity for the {S}olutions of the
{N}avier-{S}tokes {E}quations.
\newblock {\em J. Funct. Anal.}, 87:359--369, 1989.
\bibitem{HuKukavicaZiane2015}
W.~Hu, I.~Kukavica, and M.~Ziane.
\newblock Sur l'existence locale pour une \'{e}quation de scalaires actifs.
\newblock {\em C. R. Math. Acad. Sci. Paris}, 353(3):241--245, 2015.
\bibitem{JeongKim2021}
I.-J. Jeong and J.~Kim.
\newblock Strong ill-posedness for {SQG} in critical {S}obolev spaces.
\newblock {\em arXiv:2107.07739}, pages 1--35, 2021.
\bibitem{JollyKumarMartinez2020b}
M.~S. Jolly, A.~Kumar, and V.~R. Martinez.
\newblock On local well-posedness of logarithmic inviscid regularizations of
generalized {SQG} equations in borderline {S}obolev spaces.
\newblock {\em Commun. Pure Appl. Anal.}, 21(1):101--120, 2022.
\bibitem{JollyKumarMartinez2020a}
M.S. Jolly, A.~Kumar, and V.R. Martinez.
\newblock On the {E}xistence, {U}niqueness, and {S}moothing of {S}olutions to
the {G}eneralized {SQG} {E}quations in {C}ritical {S}obolev {S}paces.
\newblock {\em Commun. Math. Phys.}, pages 1--46, 2021.
\newblock DOI:10.1007/s00220-021-04124-9.
\bibitem{Kato1967}
T.~Kato.
\newblock On classical solutions of the two-dimensional non-stationary euler
equation.
\newblock {\em Arch. Ration. Mech. Anal.}, 25(3):188--200, 1967.
\bibitem{Kato1972}
T.~Kato.
\newblock Nonstationary flows of viscous and ideal fluids in $\mathbb{R}^3$.
\newblock {\em J. Funct. Anal.}, 9(3):296--305, 1972.
\bibitem{KatoPonce1988}
T.~Kato and G.~Ponce.
\newblock Commutator estimates and the euler and navier-stokes equations.
\newblock {\em Communications on Pure and Applied Mathematics}, 41(7):891--907,
1988.
\bibitem{KiselevNazarov2009}
A.~Kiselev and F.~Nazarov.
\newblock A variation on a theme of {C}affarelli and {V}asseur.
\newblock {\em Zap. Nauchn. Sem. S.-Peterburg. Otdel. Mat. Inst. Steklov.
(POMI)}, 370(Kraevye Zadachi Matematichesko\u{\i} Fiziki i Smezhnye Voprosy
Teorii Funktsi\u{\i}. 40):58--72, 220, 2009.
\bibitem{KiselevNazarovVolberg2007}
A.~Kiselev, F.~Nazarov, and A.~Volberg.
\newblock Global well-posedness for the critical 2{D} dissipative
quasi-geostrophic equation.
\newblock {\em Invent. Math.}, 167:445--453, 2007.
\bibitem{Kwon2020}
H.~Kwon.
\newblock Strong ill-posedness of logarithmically regularized 2{D} {E}uler
equations in the borderline {S}obolev space.
\newblock {\em J. Funct. Anal.}, 280(7):Paper No. 108822, 55, 2021.
\bibitem{LazarXue2019}
O.~Lazar and L.~Xue.
\newblock Regularity results for a class of generalized surface
quasi-geostrophic equations.
\newblock {\em J. Math. Pures Appl. (9)}, 130:200--250, 2019.
\bibitem{Lemarie-Rieusset2002}
P.~G. Lemari\'{e}-Rieusset.
\newblock {\em Recent developments in the {N}avier-{S}tokes problem}, volume
431 of {\em Chapman \& Hall/CRC Research Notes in Mathematics}.
\newblock Chapman \& Hall/CRC, Boca Raton, FL, 2002.
\bibitem{Li2021}
D.~Li.
\newblock Optimal gevrey regularity for supercritical quasi-geostrophic
equations.
\newblock {\em arXiv preprint arXiv:2106.12439}, pages 1--22, 2021.
\bibitem{MiaoXue2011}
C.~Miao and L.~Xue.
\newblock On the regularity of a class of generalized quasi-geostrophic
equations.
\newblock {\em J. Differential Equations}, 251(10):2789--2821, 2011.
\bibitem{Miura2006}
H.~Miura.
\newblock Dissipative quasi-geostrophic equation for large initial data in the
critical {S}obolev space.
\newblock {\em Comm. Math. Phys.}, 267(1):141--157, 2006.
\bibitem{Okhitani2011}
K.~Ohkitani.
\newblock Growth rate analysis of scalar gradients in generalized surface
quasigeostrophic equations of ideal fluids.
\newblock {\em Phys. Rev. E}, 83:036317, Mar 2011.
\bibitem{Okhitani2012}
K.~Ohkitani.
\newblock {Asymptotics and numerics of a family of two-dimensional generalized
surface quasi-geostrophic equations}.
\newblock {\em Physics of Fluids}, 24(9):095101, 09 2012.
\bibitem{Resnick1995}
S.G. Resnick.
\newblock {\em Dynamical problems in non-linear advective partial differential
equations}.
\newblock ProQuest LLC, Ann Arbor, MI, 1995.
\newblock Thesis (Ph.D.)--The University of Chicago.
\bibitem{RunstSickel1996}
T.~Runst and W.~Sickel.
\newblock {\em Sobolev spaces of fractional order, {N}emytskij operators, and
nonlinear partial differential equations}, volume~3 of {\em De Gruyter Series
in Nonlinear Analysis and Applications}.
\newblock Walter de Gruyter \& Co., Berlin, 1996.
\bibitem{Temam1974}
R.~Temam.
\newblock On the euler equations of incompressible perfect fluids.
\newblock {\em S\'eminaire \'Equations aux d\'eriv\'ees partielles
(Polytechnique) dit aussi "S\'eminaire Goulaouic-Schwartz"}, 1974.
\newblock talk:10.
\bibitem{Vishik1998}
M.~Vishik.
\newblock Hydrodynamics in {B}esov spaces.
\newblock {\em Arch. Ration. Mech. Anal.}, 145(3):197--214, 1998.
\bibitem{Vishik1999}
M.~Vishik.
\newblock Incompressible flows of an ideal fluid with vorticity in borderline
spaces of {B}esov type.
\newblock {\em Ann. Sci. \'{E}cole Norm. Sup. (4)}, 32(6):769--812, 1999.
\end{thebibliography}
\vspace{.3in}
\begin{multicols}{2}
\noindent Anuj Kumar\\
{\footnotesize
Department of Mathematics\\
Florida State University\\
Email: \url{ak22cm@fsu.edu}} \\[.2cm]
\columnbreak
\noindent Vincent R. Martinez\\
{\footnotesize
Department of Mathematics \& Statistics\\
CUNY Hunter College \\
Department of Mathematics \\
CUNY Graduate Center \\
Web: \url{http://math.hunter.cuny.edu/vmartine/}\\
Email: \url{vrmartinez@hunter.cuny.edu}\\
}
\end{multicols}
\end{document}