Nonequilibrium criticality driven by Kardar-Parisi-Zhang fluctuations in the synchronization of oscillator lattices
Extracted source text · arXiv 2210.03782
\documentclass[aps,prx,twocolumn,longbibliography,showpacs,amsmath,amssymb,amsfonts,superscriptaddress,nofootinbib]{revtex4-1}
\usepackage{graphicx}
\usepackage{dcolumn}
\usepackage{bm}
\usepackage[usenames,dvipsnames]{color}
\usepackage{multirow}
\usepackage{hyperref}
\usepackage{soul}
\usepackage{natbib}
%\bibliographystyle{apsrev4-1}
\usepackage{amsmath, amssymb}
\usepackage{bbold}
\usepackage{physics}
\definecolor{magenta}{rgb}{0.8,0.2,0.8}
\newcommand{\rod}[1]{\textcolor{blue}{{#1}}}
\newcommand{\ric}[1]{\textcolor{magenta}{{#1}}}
\begin{document}
%\title{\rod{Kardar-Parisi-Zhang} fluctuations and anomalous scaling drive synchronization in oscillator lattices}
%\title{Kardar-Parisi-Zhang fluctuations and generic scale invariance drive synchronization in oscillator lattices}
%\title{Synchronization in oscillator lattices displays nonequilibrium criticality driven by Kardar-Parisi-Zhang fluctuations}
%\title{Synchronization in oscillator lattices and nonequilibrium criticality driven by Kardar-Parisi-Zhang fluctuations}
\title{Nonequilibrium criticality driven by Kardar-Parisi-Zhang fluctuations in the synchronization of oscillator lattices}
\author{Ricardo Guti\'errez }
\author{Rodolfo Cuerno}
\affiliation{Grupo Interdisciplinar de Sistemas Complejos (GISC), Departamento de Matemáticas, Universidad Carlos III de Madrid, 28911 Legan{\'e}s, Madrid, Spain}
%\date{\today}
\begin{abstract}
The synchronization of oscillator
%lattices
ensembles is pervasive throughout nonlinear science, from classical or quantum mechanics to biology, to human assemblies. Traditionally, the main focus has been the identification of threshold parameter values for the transition to synchronization as well as the nature of such transition.
%understanding the synchronized steady state.
Here, we show that considering an oscillator lattice as a discrete growing interface provides unique insights into the dynamical process whereby the lattice reaches synchronization for long times. Working on a generalization of the celebrated Kuramoto model that allows for odd or non-odd couplings, we elucidate synchronization of oscillator lattices as an instance of generic scale invariance, whereby the system displays space-time criticality, largely irrespective of parameter values. The critical properties of the system (like scaling exponent values and the dynamic scaling Ansatz which is satisfied) happen to fall into universality classes of kinetically rough interfaces with columnar disorder, namely, those of the Edwards-Wilkinson (equivalently, the Larkin model of an elastic interface in a random medium) or the Kardar-Parisi-Zhang (KPZ) equations, for Kuramoto (odd) coupling
%odd (Kuramoto model)
and
%the more generic (in parameter space) non-odd couplings
generic (non-odd) couplings, respectively.
%This happens as a reflection of the symmetry of the couplings in each case.
From the point of view of kinetic roughening, the critical properties we find turn out to be quite innovative, especially concerning the statistics of the fluctuations as characterized by their probability distribution function (PDF) and covariance. While the latter happens to be that of the Larkin model irrespective of the symmetry of the coupling, in the generic non-odd coupling case the PDF turns out to be the celebrated Tracy-Widom distribution associated with the KPZ nonlinearity. This brings the synchronization of oscillator lattices into a remarkably large class of strongly-correlated, low-dimensional (classical and quantum) systems with strong universal fluctuations.
%, that ranges from classical to quantum matter.
\end{abstract}
%The effective equations of general oscillator lattices in the continuum limit contain terms that are ubiquitous in the study of surface growth. Depending on the symmetry of the coupling function, which is known to be relevant for the emergence of synchronization, they are those of the Larkin model (for odd functions) or of the Kardar-Parisi-Zhang equation with columnar noise (for non-odd functions). Such models display anomalous scaling in their fluctuations, \ric{\st{which is}} super-rough in the \ric{former} case and faceted in the latter, whose analysis clarify the existence synchronous dynamics and its dependence on the system size. Despite the lack of self-affinity, the fluctuations around the average growth for non-odd coupling functions follow a Tracy-Widom distribution such as that reported in the exact solutions of models in the KPZ universality class and, more recently, in an ever-increasing variety of systems. The connection between synchronization in oscillator lattices and surface growth is considerably stronger than previously anticipated and deserves future investigation.
%\pacs{...} -> physh terms
\maketitle
\section{Introduction}
From chirping crickets to Josephson junctions and quantum oscillators, passing through cells in the heart and in the brain, applauding audiences as well as biological and man-made clocks, a huge variety of systems across all of science exhibit, and sometimes crucially rely on, synchronous dynamics \cite{pikovsky,walter}. While the scientific study of synchronization can be traced back in time to the work of Christiaan Huygens in the 17th century, it is in the last few decades that it has become a central concept in nonlinear and complex dynamical systems, as a pervasive form of emerging collective dynamics. It is frequently studied using models of phase oscillators (i.e.\! idealized limit-cycle oscillators), as in the well-known Kuramoto model \cite{kuramoto_book,acebron}, though in fact it has been studied in low-dimensional chaotic systems as well \cite{boccaletti}, even in complex networks of such systems \cite{arenas}.
Another seemingly unrelated subject that has focused a great deal of attention in the last few decades is the study of surface kinetic roughening \cite{halpinhealy,krug97}, which also unifies a great diversity of nonequilibrium phenomena, from the formation of coffee rings to the growth of bacterial colonies, spanning the depositon of iceflakes and the production of thin solid films \cite{barabasi}. The fine details of many experimental systems and theoretical models in this context have been found to become irrelevant at sufficiently large space and time scales, where they show traits of universality akin to that of equilibrium critical dynamics \cite{goldenfeld,tauber14}. Crucially, however, now space-time criticality does not require adjusting control parameters to precise critical values but appears, rather, over a region of parameter space with nonzero measure. Thus, surface kinetic roughening constitutes an important instance of generic scale invariance (GSI) \cite{grinstein91,grinstein95,belitz05}, a concept which is in turn closely related to that of self-organized criticality \cite{dickman00,pruessner12}.
%Among such universal behavior the most paradigmatic is that of the Kardar-Parisi-Zhang (KPZ) class \cite{kardar}, which describes an astounding variety of nonequilibrium growth processes, as well as...
A key player in the GSI realm is the Kardar-Parisi-Zhang (KPZ) stochastic equation \cite{kardar,kriecherbauer10,takeuchi}, whose universality class is recently proving paradigmatic for the space-time critical behavior of fluctuations in low-dimensional, strongly correlated systems. Examples range from non-quantum systems like active matter \cite{chen16}, turbulent liquid crystals \cite{takeuchi11}, stochastic hydrodynamics \cite{mendl13}, colloidal aggregation \cite{yunker13}, thin-film deposition \cite{orrillo17}, reaction-diffusion processes \cite{nesic14}, or random geometry \cite{santalla15}, to the quantum realm, including exciton polaritons condensates \cite{altman15,fontaine} or quantum entanglement \cite{nahum17}, integrable and non-integrable quantum spin chains \cite{gopalakrishnan19,ljubotina19,denardis21,wei22}, or electronic fluids \cite{protopopov21}. Probably, such a ubiquity for the KPZ
universality is in turn related with the fact that the statistics of the fluctuations of the evolving field is described \cite{kriecherbauer10,takeuchi} by the celebrated Tracy-Widom (TW) family of probability density functions (PDF) for the maximum eigenvalue of Hermitian random matrices \cite{tracy09}. Indeed, TW statistics is recently being found ubiquitously \cite{makey20} across length scales in natural and technological systems, providing an analog of Gaussian statistics for correlated variables.
Connections between oscillator synchronization and surface kinetic roughening have been occasionally pointed out in the literature ---see e.g.\! Refs.\ \cite{sakaguchi,grinstein,pikovsky} and others therein---, often through mappings to the KPZ equation, to its Gaussian approximation, the so-called Edwards-Wilkinson (EW) equation \cite{edwards,barabasi}, or variations thereof. For instance, a discretized KPZ equation has been recently shown to approximate a noisy Kuramoto-Sakaguchi model of an (one or two-dimensional) oscillator lattice \cite{lauter}. Likewise, through a KPZ approach to driven-dissipative Bose-Einstein condensates of exciton polaritons, randomness in the natural frequencies has been shown \cite{moroney} not to destroy synchronization. However, a systematic assessment of the relevance and implications of KPZ fluctuations to the synchronization of oscillator lattices does not seem available in the literature yet.
In this paper, we aim to fill in this gap. Focusing on a generalization of the model of one-dimensional (1D) oscillator lattices addressed in Refs.\ \cite{lauter,moroney} ---which allows for both odd (as in the seminal Kuramoto model \cite{kuramoto1984,kuramoto_book}) and non-odd coupling functions among oscillators---, we unambiguously elucidate synchronization in these systems as an instance of GSI, with anomalous forms of scaling and universal fluctuation statistics %\st{in which well-defined universality classes occur, with universal properties which indeed are parameter-independent in full regions of parameter space. Such universal properties include critical exponents and fluctuation statistics}
which are most conveniently phrased in terms of those characterizing GSI in surface kinetic roughening, and in which KPZ fluctuations (rather than the full-fledged KPZ universality class) play an important role. The lack of awareness that the dynamics of synchronizing oscillators is endowed with universal features due to nonequilibrium criticality probably explains why the dynamical process whereby synchronization is achieved has remained a poorly studied aspect of synchronization despite its obvious interest.
Previous analytical results \cite{strogatz} show that synchronization of oscillator lattices is not possible in the thermodynamic limit for odd coupling functions (such as the Kuramoto sine coupling \cite{kuramoto1984,kuramoto_book}). Viewing the oscillator array as an evolving interface, we show this lack of synchronization to be a consequence of so-called super-rough kinetic roughening \cite{dassarma}, whereby the local slopes along the interface only stabilize upon saturation, which requires finite sizes \cite{schroeder93,lopez}. Moreover, the odd symmetry of the coupling function becomes the condition for which synchronization features the precise scaling behavior of the EW model with so-called columnar noise, known as the Larkin model in the context of elastic interfaces in disordered media; see, e.g., Ref.~\cite{purrello} and references therein. For non-odd (namely, an undefined parity of the) coupling functions, on the other hand, we show that synchronization becomes possible for arbitrarily large systems \cite{ostborn}, leading to the growth of faceted interfaces (oscillator arrays) where the slopes stabilize earlier \cite{ramasco} and whose fluctuation statistics is given by the TW distribution, as for the celebrated KPZ equation. The large-scale effective description is now given by the KPZ equation with columnar noise,
%\ric{\st{whose universality class differs from that of}
which however is not in the universality class of the celebrated KPZ equation due to differences in e.g.\ scaling exponent values \cite{szendro}. Different synchronous dynamics thus map into different surface growth models.
The geometrical point of view achieved by representing the oscillator lattice as an evolving interface allows us to develop a transparent and intuitive description of the precise way in which synchronization emerges
%\ric{\st{transition takes place},
dynamically as an ordered phase, viewing the full space-time process as a further instance of nonequilibrium criticality, specifically of GSI.
Moreover, we thus reach unexpected conclusions from the stand point of kinetic roughening \emph{per se}. Thus, the dynamic scaling Ansatz satisfied by the space-time fluctuations of the oscillator array is anomalous \cite{schroeder93,dassarma,lopez,ramasco}, irrespective of the symmetry of the coupling functions, and akin to those found in many systems with morphological instabilities and/or quenched disorder \cite{cuerno04}. What is more striking is that for non-odd coupling functions, phase fluctuations follow the ubiquitous TW PDF, one of the hallmarks of KPZ universality, in spite of the fact that neither the scaling Ansatz, nor the critical exponent values, nor the covariance of the fluctuations are those of the 1D KPZ universality class. This is in line with very recent observations in continuous \cite{rodriguez-fernandez21} and discrete \cite{marcos22} models that, somewhat unexpectedly, the fluctuation PDF and covariance, and the scaling exponents are all independent traits of the GSI universality class.
%\ric{Apart from being two important topics in contemporary statistical physics, there is indeed some relation between the dynamics of growing surfaces and that of ensembles of oscillators.} If a $d$-dimensional oscillator lattice were to be thought of as a substrate where matter is continuously deposited or removed, the phases of the oscillators would play the role of local heights of a growing surface in $d+1$ dimensions. Beyond such pictorial analogy, the deterministic terms that dominate the dynamics of an oscillator lattice in the continuum limit are those of the Edwards-Wilkinson (EW) \cite{edwards} or the KPZ \cite{kardar} stochastic differential equations, representatives of the main universality classes in the physics of equilibrium and non-equilibrium surface growth, respectively \cite{barabasi}. While such formal connection has been occasionally pointed out in the literature (see e.g.\! \cite{sakaguchi,grinstein,pikovsky,lauter,moroney}), it seems fair to say that a detailed study of synchronization from the perspective of the contemporary physics of surface growth has not yet been undertaken, despite the hints that deeper links exist. Indeed, while previous exact results show that synchronization is not possible in the thermodynamic limit for odd coupling functions (such as the Kuramoto sine coupling) \cite{strogatz}, this symmetry is precisely the condition for which the dominant terms in the resulting continuum limit of the synchronization problem are those of the EW model with columnar noise, known as the Larkin model in the context of elastic interfaces (see, e.g., Ref.~\cite{purrello} and references therein). For non-odd coupling functions, on the other hand, synchronization becomes possible for arbitrarily large systems \cite{ostborn}, while the effective description is given by the KPZ equation with columnar noise \cite{szendro}. Different synchronous dynamics map into different growth models.
%In this work we characterize the morphology of synchronized phase profiles. \ric{This physical approach to the study of synchronization sheds new light on the role played by the spatial and temporal scales involved in the emergence of synchronous dynamics. We thus find that the synchronized dynamics of oscillator lattices are scale-invariant, characterized by a correlation length $\xi(t)$ which grows in time as a power law, but not self-affine, since the saturation at intermediate scales $\ell$ is not achieved by the time $\xi(t) \sim \ell$, as happens at the largest scales when the correlation spans the full system $\xi \sim L$.} Synchronization is \ric{indeed \st{shown to be}} driven by \ric{two different forms of} anomalous (non-self-affine) scaling behavior\ric{, which follow the generic ansatz proposed in Ref.~\cite{ramasco},} and also by the presence of KPZ fluctuations in the case of non-odd coupling (they are gaussian otherwise). The lack of synchronization for odd coupling in the thermodynamic limit is shown to be a consequence of super-rough scaling \cite{dassarma}, whereby the local slopes only stabilize upon saturation (which requires finite sizes) \cite{lopez}, whereas non-odd coupling functions lead to a faceted growth that stabilizes the slopes at shorter time scales \cite{ramasco}. In the latter case, the fact that the phase fluctuations around the average power law growth follow an ubiquitous Tracy-Widom distributions \cite{}, which is one of the hallmarks of KPZ universality (see for example the review article \cite{takeuchi}) is quite surprising: the KPZ equation with columnar noise does not feature any of the exponents, nor even the self-affinity, of models in that class \cite{szendro} (and in fact the distributions of height fluctuations in that model have not yet been investigated as far as we know). Our results are based on a numerical exploration of synchronous evolutions of one-dimensional systems of phase oscillators, which is by far the setting in which both the synchronization and the surface growth problems are understood in greater depth.
This paper is organized as follows. In Section II we describe the general model of a lattice of phase oscillators and its continuum-limit description, the observables of interest when described as an evolving interface, and the particular coupling functions to be considered in the numerics, which will be of the Kuramoto-Sakaguchi form \cite{sakaguchi86}. Section III contains a general discussion and results on the synchronous dynamics in the model and the role of the symmetries of the coupling function. Section IV is devoted to the study of synchronization with the Kuramoto (sine) coupling form, for which we solve the continuum-limit (linear) Larkin model. Numerical simulations reveal a super-rough scaling adequately described by the analytical results. In Section V we study synchronization with a non-odd coupling function, whose effective description yields a (nonlinear) KPZ equation with columnar noise. The numerical results show anomalous scaling and faceted interfaces (oscillator arrays) in qualitative agreement to those found in the analysis of the corresponding growth process \cite{szendro}, as they are driven by the same non-equilibrium normal-growth KPZ mechanism.
%Moreover, fluctuations around the growth follow a Tracy-Widom distribution, characteristic of KPZ universality.
Sections IV and V also contain the study of the one- and two-point statistics of fluctuations that yields Gaussian (respectively, TW) statistics for odd (respectively, non-odd) couplings, but the same covariance of the Larkin model in all cases. Finally, in Section VI we provide some concluding remarks and ideas for future work lying at the interface between synchronization and kinetic roughening. Additional numerical results showing the robustness of the scaling behavior with respect to changes in the model parameters (thus confirming proper GSI behavior) are organized into two appendices at the end.
\section{Model and Observables}
\subsection{General lattices of phase oscillators}
We consider a system of $L^d$ {\it phase oscillators} at the sites of a $d$-dimensional hypercubic lattice of linear size $L$. Each oscillator is an idealized dissipative dynamical system with an attracting limit cycle, whose state is given by a phase $\phi_i(t)$. The time evolution is determined by its {\it intrinsic frequency} $\omega_i$ (i.e.~the frequency of oscillation when isolated) and the interactions with its neighbors through a smooth, $2\pi$-periodic {\it coupling function} $\Gamma(\phi_j -\phi_i)$. The equations of motion are thus
\begin{equation}
\frac{d \phi_i}{d t} = \omega_i + \sum_{j\in \Lambda_i} \Gamma(\phi_j -\phi_i),\ i=1,2,\ldots, L^d,
\label{eq1}
\end{equation}
where $\Lambda_i$ is the set of $2 d$ neighbors of site $i$. The intrinsic frequencies are independent and identically distributed according to a probability density $g(\omega)$ with zero mean and finite variance, i.e.\! $\langle \omega_i \rangle = \int d\omega g(\omega)\, \omega = 0$ and $\langle \omega_i \omega_j \rangle = \delta_{ij}\int d\omega g(\omega)\, \omega^2= 2 \sigma \delta_{i j}$, where $\delta_{i j}$ is the Kronecker delta. A nonzero mean would only introduce a uniform frequency shift, easily removable by moving to the rotating frame. Moreover, we assume that the distribution is even, $g(-\omega) = g(\omega)$.
The {\it effective frequencies} of oscillation are defined as
\begin{equation}
\omega^\text{eff}_i \equiv \lim_{T\to\infty}\frac{\phi_i(t_0+T)-\phi_i(t_0)}{T},\ i=1,2,\ldots, L^d,
\label{omegaeff}
\end{equation}
where $[0,t_0]$ is a time interval sufficiently long to contain the transient behavior, and the limit is assumed to exist. Due to the coupling between neighboring oscillators, the effective frequency $\omega^\text{eff}_i$ will be generally different from the intrinsic frequency $\omega_i$. Oscillators that evolve at the same effective frequency are said to be {\it frequency entrained} or {\it frequency locked}. We will focus on such a (time-averaged) form of synchronization, which does not require the {\it instantaneous frequencies} $d \phi_i/d t$ to become strictly identical at all times.
In all the cases under consideration the coupling function $\Gamma(\phi_j -\phi_i)$ will include a {\it coupling strength} $K$, an overall nonnegative proportionality factor with dimensions of inverse time. For the most part, we will focus on cases where there is a critical coupling strength above which synchronization is achieved in the sense just mentioned. %in the systems under investigation.
More details are given in Sec.\ \ref{cond}.
\subsection{Continuum limit}
\label{contlim}
Generalizing the 1D approach of Ref.~\cite{kuramoto1984}, we write the positions of the oscillators as vectors in continuous space, ${\bf x} = (x_1, x_2, \ldots, x_d)\in \mathbb{R}^d$, so the phase of oscillator $i$ becomes, $\phi_i \rightarrow \phi({\bf x})$, and the neighboring oscillators are placed at positions ${\bf x}\pm a {\bf e}_k$, where ${\bf e}_k$ for $k=1,2,\ldots, d$ is a canonical basis vector all of whose entries are zero except the $k$-th one (which is one). Thus Eq.~(\ref{eq1}) becomes
\begin{align}
\partial_t \phi({\bf x},t) = \omega({\bf x}) &+ \sum_{k\in \mathbb{N}_d} \left[\Gamma(\phi({\bf x}+ a {\bf e}_k,t) -\phi({\bf x},t))\right.\nonumber\\
&\left.+ \Gamma(\phi({\bf x}- a {\bf e}_k,t) -\phi({\bf x},t))\right],
\label{eq2}
\end{align}
where $\mathbb{N}_d = \{1,2,\ldots, d\}$, $\langle \omega({\bf x}) \rangle = 0$, and $\langle \omega({\bf x})\omega({\bf x}') \rangle = 2 \sigma \delta({\bf x} - {\bf x}')$, using the Dirac delta. So far the components in ${\bf x}$ should be considered as constrained to take values that are integer multiples of the lattice constant $a$, but from now on we will focus on a coarse-grained description where $a$ is assumed to be small compared to the wavelengths over which the phase field $\phi({\bf x},t)$ fluctuates. Taylor-expanding around ${\bf x}$, $\phi({\bf x}\pm a {\bf e}_k,t) - \phi({\bf x}) = \pm a \partial_k \phi({\bf x},t) + \frac{1}{2} a^2 \partial_k^2\phi({\bf x},t) \pm \frac{1}{6} a^3 \partial_k^3\phi({\bf x},t) + \mathcal{O}(a^4)$,
where $\partial_k$ is shorthand for the partial derivative with respect to the $k$-th coordinate $x_k$. In Eq.~(\ref{eq2}) terms of odd order in $a$ vanish due to the $a\to -a$ symmetry of the coupling term, yielding
%\begin{widetext}
\begin{align}
\partial_t \phi ({\bf x},t) &=\omega({\bf x}) + 2 d \Gamma(0) + a^2 \Gamma^{(1)}(0)\!\sum_{k\in \mathbb{N}_d}\!\partial_k^2 \phi({\bf x},t) \nonumber\\
&+ a^2 \Gamma^{(2)}(0)\!\sum_{k\in \mathbb{N}_d}\!(\partial_k \phi({\bf x},t) )^2 + \mathcal{O}(a^4),
%+ \frac{a^4}{12}\!\left[\Gamma^{(1)}(0)\!\sum_{k\in \mathbb{N}_d}\!\partial_k^4 \phi \right.\nonumber\\
%&\left. + \Gamma^{(2)}(0)\!\left(3\!\sum_{k\in \mathbb{N}_d}\!(\partial_k^2 \phi)^2 + 4\!\sum_{k\in \mathbb{N}_d}\!\partial_k \phi\, \partial_k^3 \phi \right)+ 6 \Gamma^{(3)}(0)\!\sum_{k\in \mathbb{N}_d}\!(\partial_k \phi)^2 \partial_k^2 \phi+ \Gamma^{(4)}(0)\!\sum_{k\in \mathbb{N}_d}\!(\partial_k \phi)^4 \right]+ \mathcal{O}(a^6),
\label{eq3}
\end{align}
%\end{widetext}
where $\Gamma^{(n)}$ denotes the $n$-th derivative of $\Gamma$.
For a relatively slow spatial variation of the phase field $\phi({\bf x},t)$, as occurs for coupling strengths $K$ well into the synchronized regime, it may be reasonable to neglect terms of order higher than $a^2$, which is the dominant one for the spatial coupling in the oscillating medium. By analogy with surface growth, will refer to this as a {\it small-slope approximation} which yields the continuum-limit effective equation
\begin{equation}
\partial_t \phi({\bf x},t) = \omega^*({\bf x}) + \nu \nabla^2 \phi({\bf x},t) + \frac{\lambda}{2} [\nabla \phi({\bf x},t)]^2,
\label{eq4}
\end{equation}
where as usual the Laplacian $\nabla^2 \phi({\bf x},t) \equiv \sum_{k\in \mathbb{N}_d}\!\partial_k^2 \phi({\bf x},t)$ and the squared norm of the gradient $[\nabla \phi({\bf x},t)]^2\equiv \sum_{k\in \mathbb{N}_d}\![\partial_k \phi({\bf x},t)]^2$. The appropriateness of the truncation for specific cases will be discussed when making the comparison between predictions derived from it and results based on the direct numerical integration of particular instances of Eq.~(\ref{eq1}). %in one dimension.
%\ric{He reestructurado un poco estos párrafos que siguen y quitado algún comentario repetido.}
Equation \eqref{eq4} features the same deterministic derivative terms as the KPZ equation \cite{kardar}. We have introduced two parameters, $\nu \equiv a^2 \Gamma^{(1)}(0)$ and $\lambda/2 \equiv a^2 \Gamma^{(2)}(0)$, following the standard notation in the surface growth literature, where they quantify smoothening surface tension and interface growth along the local surface normal direction, respectively \cite{barabasi}. This formal connection between oscillator lattices and rough interfaces has been pointed out on several occasions in the literature, as mentioned in the Introduction, with some recent works even exploiting it for the study of synchronization in novel scenarios \cite{lauter,moroney}.
However, Eq.~\eqref{eq4} differs from the KPZ equation by the noise term, $\omega^*({\bf x})\equiv \omega({\bf x}) + 2 d \Gamma(0)$, because it is time-independent, in contrast with the time-dependent noise of the standard KPZ equation. For a given assignment of (randomly distributed) intrinsic frequencies $\omega({\bf x})$, the term $\omega^*({\bf x})$ acts as quenched disorder, in such a way that the system evolves \emph{deterministically} from the initial condition (unless an external thermal-noise source is explicitly included, see e.g.\! Ref.~\cite{lauter}). The dynamics is akin to that of a growth process in a medium for which the disorder values depend on the substrate coordinate ${\bf x}$ but not on the local value of the interface ``height'' $\phi({\bf x},t)$. In the kinetic roughening literature, Eq.\ \eqref{eq4} is in fact described as the KPZ equation with \emph{columnar} noise \cite{halpinhealy, szendro}. Other forms of quenched disorder, analogous to those met by interfaces moving through porous media \cite{barabasi}, would require the intrinsic frequencies to depend of the instantaneous phases, which appears to be unphysical.
\subsection{Morphological analysis}
\label{obs}
According to the analogy sketched above, in our analysis we consider the phase field $\phi({\bf x},t)$ as if it were describing the heigth $h({\bf x},t)$ of an interface growing above point ${\bf x}\in \mathbb{R}^d$ on a $d$-dimensional substrate, at time $t$. In surface kinetic roughening processes
\cite{barabasi,halpinhealy,krug97}, the fluctuations of the local height around the mean value are captured by the {\it global width} or roughness $W(L,t) \equiv \langle \overline{[h({\bf x},t)-\overline{h}]^2} \rangle^{1/2}$, where the overbar denotes a spatial average in a system of linear (substrate) size $L$ and the angular brackets denote averaging over different noise realizations. The GSI behavior implies that surface height values are statistically correlated for distances smaller than a correlation length $\xi(t)$ which increases with time as a power law, $\xi(t) \sim t^{1/z}$, where $z$ is the so-called {\it dynamic exponent}. Such an increase takes place until $\xi(t)$ reaches a value comparable to $L$, which results in the width saturating at a steady-state, size-dependent value $W(L,t\gg L^z) \sim L^\alpha$. Here, $\alpha$ is the so-called {\it roughness exponent}, which is related with the fractal dimension of the interface profile $h(\mathbf{x},t)$. In a wide variety of physical contexts and conditions, the global roughness satisfies the {\it Family-Vicsek (FV) scaling Ansatz}
\cite{barabasi,halpinhealy,krug97}
\begin{equation}
W(L,t) = t^{\beta} f(L/\xi(t)),
\label{fv}
\end{equation}
where the scaling function $f(y) \sim y^\alpha$ for $y\ll 1$, while $f(y)$ reaches a constant value for $y\gg 1$ \cite{vicsek}. The ratio $\beta = \alpha/z$ is known as the {\it growth exponent}, and characterizes the short-time behavior of the roughness, $W(t)\sim t^{\beta}$. Specifically, the FV Ansatz is verified by classical models of equilibrium critical dynamics \cite{tauber14}. Away from equilibrium, it is also verified by representatives of important universality classes of kinetic roughening, like those of the KPZ and EW equations, which are characterized by the set of $(\alpha,z)$ exponent values and their dependence on the substrate dimension $d$ \cite{barabasi,halpinhealy,krug97}. In this sense, kinetic roughening extends classical equilibrium critical dynamics far from equilibrium \cite{tauber14}.
%Geometrically, it expresses the scale-invariance of the surface: to be revealed, the coordinate in the growth direction must be rescaled with a different exponent than the coordinates along the substrate.
Beyond global quantities like $W(t)$, the GSI occurring in kinetic roughening systems also impacts the behavior of correlation functions. A particularly useful one is the {\it height-difference correlation function},
\begin{equation}
G({\bf r},t) \equiv \langle \overline{[h({\bf x}+{\bf r},t) - h({\bf x},t)]^2} \rangle.
\label{grt}
\end{equation}
In our cases of interest, due to rotational invariance, the correlations only depend on $\ell \equiv |{\bf r}|$. For FV scaling %with $\alpha < 1$ (which is the standard, self-affine scaling),
$G(\ell,t)$ scales differently depending on how $\ell$ compares with the correlation length \cite{barabasi,halpinhealy,krug97},
\begin{equation}
G(\ell,t) \sim \left\{
\begin{array}{lr}
t^{2 \beta},& \text{if } t^{1/z} \ll \ell, \\
\ell^{2 \alpha}, & \text{if } \ell \ll t^{1/z}.
\end{array}
\right\} = \ell^{2\alpha} g(\ell/\xi(t)) ,
\label{glt}
\end{equation}
where we are assuming $\ell < L$ and $g(y)$ is a suitable scaling function \cite{barabasi,krug97}. In fact, $G(\ell,t)$ scales like the square of the {\it local width} $w(\ell,t) \equiv \langle \overline{[h({\bf x},t)-\overline{h}]^2} \rangle^{1/2}$ (the spatial average is here restricted to a region of linear size $\ell$), $G(\ell,t) \sim w^2(\ell,t)$. One can see that the scaling form in Eq.\ (\ref{glt}) only reflects the growth and saturation discussed above for the whole system, but now restricted to a region of linear size $\ell$ \cite{lopezphysa}. A related correlation function that is also frequently studied \cite{takeuchi} is the so-called {\it height covariance}
\begin{equation}
C(\mathbf{r},t) \equiv \langle \overline{h(\mathbf{x},t) h(\mathbf{x}+\mathbf{r},t)} \rangle - \langle \bar{h}(t) \rangle^2 ,
\label{eq:cov}
\end{equation}
such that, again under the assumption of rotational invariance, $G(\ell,t)= 2 [W^2(t)-C(\ell,t)]$ \cite{krug97}.
%Self-affinity is an independent additional condition to Family-Vicsek scaling, and in fact there are growth process that are neither self-affine nor display Family-Vicsek scaling \cite{lopez}.
For cases in which e.g.\ crossover behavior exists \cite{siegert96} and for additional reasons to be detailed below, an alternative two-point correlation function is considered, namely, the surface {\it structure factor} (i.e.\! the power spectral density of the height fluctuations) defined as
\begin{equation}
S({\bf k},t) \equiv \langle \hat{h}({\bf k}, t)\hat{h}(-{\bf k}, t)\rangle = \langle |\hat{h}({\bf k}, t)|^2\rangle,
\end{equation}
where $\hat{h}({\bf k},t)\equiv\mathcal{F}[h(\mathbf{x},t)]$ is the space Fourier transform of $h(\mathbf{x},t)$ and $\mathbf{k}$ is $d$-dimensional wave vector. The FV Ansatz now reads
\begin{equation}
S({\bf k},t) = k^{-(2 \alpha +d)} s_{\rm FV}(k t^{1/z}) ,
\label{SkFV}
\end{equation}
with $s_{\rm FV}(y)$ approaching a constant value for $y\gg 1$ and $s_{\rm FV}(y) \propto y^{2\alpha +d}$ for $y\ll 1$. This can be derived by realizing that $W^2(L,t)$ equals the integral of $S({\bf k},t)$ over wave vector space (Parseval's theorem) \cite{barabasi}. Likewise, $S(\mathbf{k},t)$ is analytically related with, e.g., $G(\mathbf{r},t)$ via space Fourier transforms \cite{krug97}.
In the case of a system of oscillators, analogous observables to the global roughness and the correlation functions
%, the height-difference correlation function, and the structure factor
are simply defined by replacing the height field $h({\bf x},t)$ by the phase field $\phi({\bf r},t)$, leading to the following definitions:
\begin{equation}
W_\phi(L,t) \equiv \langle \overline{[\phi({\bf x},t)-\overline{\phi}]^2} \rangle^{1/2} ,
\label{wphi}
\end{equation}
and
\begin{equation}
G_\phi({\bf r},t) \equiv \langle \overline{[\phi({\bf x}+{\bf r},t) - \phi({\bf x},t)]^2} \rangle ,
\label{grt}
\end{equation}
and
\begin{equation}
C_\phi({\bf r},t) \equiv \langle \overline{\phi({\bf x}+{\bf r},t) \phi({\bf x},t)} \rangle -
\langle \bar{\phi}(t) \rangle^2,
\label{cphi}
\end{equation}
and
\begin{equation}
S_\phi({\bf k},t) \equiv \langle \hat{\phi}({\bf k}, t)\hat{\phi}(-{\bf k}, t)\rangle ,
\label{sphi}
\end{equation}
which will be the main objects of our analysis in the sections to follow. In the case of two-point correlations, we will be particularly interested in their value at a distance of one site, $G_\phi(\ell = 1,t)$, which will be referred to as the {\it average squared slope}, and denoted as $\langle\overline{(\Delta \phi)^2} \rangle$. Our interest lies in finite systems, where saturation may be eventually attained. As differences between oscillator phases that do not evolve at the same effective frequency $\omega^{\text{eff}}$ must grow steadily in time for long times, the saturation of $W_\phi(L,t)$ [or equivalently, that of $S_\phi({\bf k},t)$ or $G_\phi({\bf r},t)$] as $t\to\infty$ indeed indicates the presence of synchronization in the sense mentioned above.
Two further aspects of surface growth, which have been the focus of much recent research, will turn out to be indispensable for the analysis of the synchronization problem. In a sense, they both challenge the traditional characterization of kinetic-roughening universality classes in terms of just two independent exponents appearing in the FV dynamic scaling Ansatz, Eq.\ (\ref{fv}). One is the existence of growth processes with {\it anomalous scaling} properties \cite{schroeder93,dassarma,krug97,lopez,ramasco,cuerno04}. While for standard FV systems height fluctuations at local scales $\ell \ll L$ behave like the global fluctuations at distances comparable with the system size $L$ (namely, with the same roughness exponent), in systems displaying anomalous scaling the local width scales as $w(\ell,t\gg \ell^z) = \ell^{\alpha_\text{loc}}$, with $\alpha_\text{loc} \neq \alpha$, so that local and global fluctuations scale with different roughness exponents.
%As a consequence, there may be intermediate time scales where lengths below the correlation length, $\ell \ll \xi(t) \ll L$, still evolve in time while the correlation length grows, $\xi(t) \sim t^{1/z}$, before it reaches saturation at $\xi(t) \sim L$.
The anomalous scaling that occurs in our work generally needs to be identified by means of the structure factor, as a new independent exponent $\alpha_s$ appears (termed {\it spectral roughness exponent} \cite{ramasco}), which only manifests itself in Fourier space.
%, may be needed to capture the scaling of the structure factor of the height fluctuations in a generic anomalous scaling setting,
The {\it dynamic scaling Ansatz} reads \cite{ramasco}
\begin{equation}
S(k,t) = k^{-(2 \alpha +d)} s(k t^{1/z}) ,
\label{Sk}
\end{equation}
where $s(y) \propto y^{2(\alpha - \alpha_s)}$ for $y\gg 1$ and $s(y) \propto y^{2\alpha +d}$ for $y\ll 1$.
Equation \eqref{Sk} generalizes the FV Ansatz, Eq.\ \eqref{SkFV}, which is retrieved if $\alpha_s=\alpha$. In general, the value of $\alpha_s$ and its relation with respect to $\alpha$ determines the type of anomalous scaling that takes place \cite{ramasco}.
%Depending on the value of such , singularly whether it is larger or smaller than 1, and how it compares to $\alpha$ and $\alpha_\text{loc}$, one is in the presence of different forms of dynamic scaling, known as intrinsically anomalous scaling \cite{lopez}, super-roughness \cite{dassarma} or faceted scaling \cite{ramasco}.
We will be interested in cases such that $\alpha_s>1$, for which the correlation function scales as \cite{lopezphysa}
\begin{equation}
G(\ell,t) \sim \left\{
\begin{array}{lr}
t^{2 \beta},& \text{if } t^{1/z} \ll \ell \ll L, \\
\ell^{2 \alpha_\text{loc}} t^{2(\alpha-\alpha_\text{loc})/z}, & \text{if } \ell \ll t^{1/z} \ll L.
\end{array}
\right.
\label{glt_mod}
\end{equation}
This means that the two-point correlations keep increasing (anomalously) with time even at distances which are smaller than the correlation length, at which they saturate in case of FV scaling; in contrast, now they only saturate at $\ell^{2 \alpha_\text{loc}} L^{2(\alpha-\alpha_\text{loc})}$ when $t^{1/z} \sim L$. If $\alpha = \alpha_s >1$, the anomalous scaling is termed super-rough \cite{lopez}, due to the large interface fluctuations that occur. As just noted, the scaling Ansatz satisfied by the structure factor is FV in this case, but $\alpha_\text{loc} = 1 \neq \alpha$. Otherwise, if $\alpha \neq \alpha_s$ with both exponents being larger than 1, faceted anomalous scaling takes place \cite{ramasco}, and again $\alpha_\text{loc} = 1$. There exist cases in which $\alpha_s<1$, leading to so-called intrinsic anomalous scaling in which $\alpha_\text{loc}$ need not equal 1 \cite{lopez}. A recent example is the tensionless KPZ equation \cite{rodriguez-fernandez22}. Additional physical examples of the various types of anomalous scaling can be found e.g.\ in Refs. \cite{ramasco,cuerno04}. As it turns out, oscillator lattices display super-rough or faceted anomalous scaling depending on the symmetries of the coupling function $\Gamma(\Delta \phi)$, where the difference between the phases of two neighboring oscillators $\Delta \phi$ will be referred to as the {\it local slope} (hence the ``small-slope approximation'' discussed in the previous subsection).
A second aspect of surface kinetic roughening that is relevant to our analysis, and currently the focus of a great deal of research, has to do with the {\it statistics of fluctuations}. For rough surfaces, the observable of interest is the PDF of the fluctuations of the heights $h({\bf x},t)$ [in our case, that of the phases $\phi({\bf x},t)$], around their mean.
%average power-law growth $t^\beta$.
By a straightforward adaptation of the definition employed in the kinetic roughening literature \cite{kriecherbauer10,halpinhealy,takeuchi}, we will focus on the PDF of
\begin{equation}
\varphi_i\equiv \frac{\delta \phi_i(t_0+\Delta t) - \delta \phi_i(t_0)}{(\Delta t)^\beta} ,
\label{fluct}
\end{equation}
where $\delta \phi_i(t) = \phi_i(t) - \overline{\phi}(t)$, $t_0$ is a reference time beyond the initial transient dynamics and $t_0 +\Delta t$ is some intermediate time within the growth regime of the system, prior to saturation to the steady state. The division by $(\Delta t)^\beta$ removes the systematic increase of the fluctuations in time so that, remarkably, the PDF of $\varphi_i$ reaches a universal, time-independent form \cite{kriecherbauer10,halpinhealy,takeuchi}.
Important examples are e.g.\ the Gaussian distribution for the linear EW equation \cite{barabasi,krug97} and a TW PDF (whose precise form depends, e.g., on boundary conditions) for the KPZ equation \cite{kriecherbauer10,halpinhealy,takeuchi}.
%average contribution to the growth and allows us to focus on the distribution of fluctuations around it.
%Recently, the ubiquitous TW PDFs \cite{makey20} have been shown to accurately describe the distribution of (similarly defined) fluctuations of models in the KPZ universality class \cite{kriecherbauer10,takeuchi}.
We will provide convincing numerical evidence that a TW PDF accurately describes the fluctuations in the synchronization of oscillator lattices as well, provided that the
%effective description is given by a KPZ eqution with columnar noise (a condition that will be generically satisfied in our model), despite the nature of the disorder and the lack of self-affinity in the KPZ equation.
coupling functions are not odd, in spite of the fact that the kinetic roughening exponents are not those of the standard KPZ equation. For the specific choice of odd (Kuramoto) couplings, on the other hand, which is effectively described by a EW equation with columnar noise (Larkin model), the fluctuations will be shown to be Gaussian.
\subsection{One-dimensional model}
\label{nummod}
The coupling function $\Gamma(\Delta \phi)$ in Eq.~(\ref{eq1}) is assumed to be $2\pi$-periodic and smooth, so it can be expressed as a Fourier series. One crucial property in what follows is its symmetry under phase inversion $\Delta \phi \to -\Delta \phi$. Specifically, whether the function is odd, $\Gamma(\Delta \phi) + \Gamma(-\Delta \phi) = 0$, or not. As explained in the next subsection, the non-oddity of $\Gamma(\Delta \phi)$ appears to be a necessary condition for synchronization in the thermodynamic limit \cite{strogatz, ostborn}. While we are interested in the synchronization of finite systems, we will see that whether $\Gamma(\Delta \phi)$ presents this symmetry or not has crucial implications in several dynamical features of synchronization, which can be related to the occurrence of the nonlinearity in Eq.~(\ref{eq4}). An obvious feature already in the oscillator model, Eq.\ \eqref{eq1}, is that only when $\Gamma(\Delta\phi)$ is odd does the system have up-down symmetry, i.e.\! invariance under $\phi_i\to -\phi_i$, in a statistical sense.
Our simulations will be based on oscillator rings (i.e.\! 1D chains with periodic boundary conditions). For simplicity we always assume that the coupling function $\Gamma(\Delta \phi)$ is described by the fundamental term in the Fourier series (i.e.\! higher harmonics are not considered). Moreover, we assume that $\Gamma(0) = 0$: when an oscillator phase is equal to that of a neighbor, the coupling between them vanishes. Therefore the noise term in the effective dynamics, Eq.\ (\ref{eq4}), is simply given by the intrinsic frequencies, $\omega^*(x) = \omega(x)$. Under such assumptions, the coupling function can be written, without loss of generality, as
\begin{equation}
\Gamma(\Delta \phi) = K(\sin(\Delta \phi + \delta)-\sin \delta),
\label{gamma}
\end{equation}
for $\delta \in (-\pi,\pi]$. That is the functional form of the coupling in our numerical investigations, where $K$ is the coupling strength. With a slightly different notation, it was previously considered in the seminal Ref.~\cite{sakaguchi}. The odd symmetry $\Gamma(\Delta \phi) + \Gamma(-\Delta \phi) = 0$ is obtained only for $\delta = 0$ and $\pi$, which correspond to $\Gamma(\Delta \phi) = K \sin(\Delta \phi)$ (Kuramoto coupling) or $\Gamma(\Delta \phi) = -K \sin(\Delta \phi)$, respectively.
The model that we study numerically is thus
\begin{equation}
\frac{d\phi_i}{dt} = \omega_i + K[\sin(\phi_{i+1}\!-\!\phi_i\!+\!\delta) +\sin(\phi_{i-1}\!-\!\phi_i\!+\!\delta)-2\sin\delta],
\label{eqsim}
\end{equation}
for $i = 1,2,\ldots, L$, where $\phi_0 \equiv \phi_L$ and $\phi_{L+1} \equiv \phi_1$. The system starts from a homogeneous initial condition, $\phi_i(0)=0$ for all $i$, and is integrated by means of a fourth-order Runge-Kutta algorithm with time step $\delta t = 0.01$. The intrinsic frequencies $\omega_i$ are uniformly distributed over $[-1,1]$, a choice of the distribution $g(\omega)$ that evidently satisfies the properties discussed immediately after Eq.\ (\ref{eq1}). As we are interested in the morphological analysis of synchronized dynamics, we will focus on values of $K$ that are high enough for synchronization to be achieved. This matter will be discussed in Section \ref{prelim}.
The synchronous dynamics of the system of phase oscillators governed by Eq.\! (\ref{eqsim}) has been recently explored (except for a uniform shift in the intrinsic frequencies) for the parameter choice $\delta = -\pi/4$ in a recent publication that, motivated by the study of driven-dissipative bosonic systems, examines the transition to synchronization as a diffusion process in a random medium \cite{moroney}. Some of the morphologies to be discussed here present similarities to those displayed in that work, as well as (in terms of surface heights, instead of phases) in the study of the growth process arising from the KPZ equation with columnar noise \cite{szendro}. Such similarities will be discussed in Section \ref{nonodd}.
We conclude the description of the model by pointing out a symmetry which will be relevant in the following: the coupling function, Eq.\ (\ref{gamma}), changes sign under the simultaneous reversal of the phases and the angle $\delta$. This results into the invariance of Eq.\ (\ref{eqsim}) under $\phi_i \to - \phi_i$ and $\delta \to -\delta$, in the same statistical sense mentioned above for the invariance under phase reversal (up-down symmetry) of a general odd coupling function. In fact the latter can be considered as a particular case of the former invariance for $\delta = 0$ or $\pi$ in the dynamical model, Eq.\ (\ref{eqsim}).
\subsection{Conditions for synchronization}
\label{cond}
Establishing the existence and stability of synchronous solutions in (finite-dimensional) lattices of oscillators is considerably more difficult than in the (mean-field) all-to-all coupling case. The latter is frequently considered for the well-known Kuramoto model when the coupling function is a sine function $\Gamma(\Delta \phi) = \sin (\Delta \phi)$, which has been thoroughly studied in the past \cite{acebron}. Nonetheless, even in finite dimensions, relevant exact results are available, only under certain assumptions on the coupling functions \cite{strogatz,kopell,ostborn}. We limit the discussion to one-dimensional settings, on which our numerical work will focus, particularly to those results that are directly relevant in our investigation.
Strogatz and Mirollo proved \cite{strogatz} that, in order for synchronization to occur in a typical chain of $L$ oscillators for large $L$ with a Kuramoto coupling $\Gamma(\Delta \phi) = \sin (\Delta \phi)$ (and a distribution of intrinsic frequencies with finite mean and variance), a coupling strength of size $\mathcal{O}(\sqrt{L})$ is required, which precludes the existence of synchronization in the thermodynamic limit. Their proof rests on the fact that the sine function is odd. In fact, the (non-)odd character of the coupling function has been pointed out as an important feature for the emergence of synchronization in other works. Thus, Kopell and Ermentrout showed \cite{kopell} that a non-odd coupling leads to frequency entrainment when the intrinsic frequencies satify $|\omega_{i+1} - \omega_i| = \mathcal{O}(1/L)$. For the case of purely random (spatially uncorrelated) frequencies, Sakaguchi {\it et al.} provided evidence \cite{sakaguchi} indicating that a coupling function of the form $K(\sin(\Delta \phi + \delta)-\sin \delta)$ for $\delta \neq 0$ favors synchronization when compared to the odd $\delta = 0$ case. Their arguments are based on a simplified two-oscillator problem (to be discussed below) and a mapping of the many-oscillator problem into the Schr\"odinger equation of a particle in a random potential (which presents some similarities with the recent analysis in Ref.~\cite{moroney}), as well as numerical results. A more elaborate theoretical treatment, under some mathematical assumptions concerning the coupling function, was provided by \"Ostborn, who showed \cite{ostborn}, among other things, that for non-odd couplings synchronous solutions exists and are stable for coupling strengths above a critical value that is independent of the system size. This estimate of the critical coupling strength, together with the one based on the two-oscillator model in Ref.~\cite{sakaguchi}, will be discussed in the next section.
\section{Synchronization in oscillator chains: preliminary analysis}
\label{prelim}
\subsection{Critical coupling strength $K_c$}
\label{kc}
The oscillator model in Eq.~(\ref{eqsim}) has two parameters, namely the angle $\delta$ and the coupling strength $K$. As our interest lies in synchronous dynamics, we will focus on sufficiently strong couplings $K\geq K_c$, where $K_c$ is the critical coupling strength for synchronization, which in general may depend on both $\delta$ and the system size $L$. Further, as we aim to elucidate the connection between synchronization in oscillator lattices and continuum kinetic-roughening equations, we will restrict the numerical analysis to parameter choices for which the small-slope approximation leading to Eq.\ (\ref{eq4}) is sensible.
To gain some insight into the dependence of the critical coupling strength $K_c$ and the stationary slopes $\Delta \phi$ on the parameters of the model, it is useful to consider the simple two-oscillator model introduced in Ref.~\cite{sakaguchi}:
\begin{align}
\frac{d\phi_1}{dt} &= \omega_1 + K(\sin(\phi_2-\phi_1+ \delta)-\sin \delta),\nonumber\\
\frac{d\phi_2}{dt} &= \omega_2+ K(\sin(\phi_1-\phi_2+ \delta)-\sin \delta).
\label{twoosc}
\end{align}
In terms of the auxiliary variables $\bar{\phi} = (\phi_1 + \phi_2)/2$ and $\Delta \phi = \phi_2 - \phi_1$, and the corresponding intrinsic frequencies $\bar{\omega} = (\omega_1+ \omega_2)/2$ and $\Delta \omega = \omega_2 - \omega_1$, it becomes
\begin{align}
\frac{d\bar{\phi}}{dt} &= \bar{\omega} - K \sin \delta (1-\cos \Delta \phi),\nonumber\\
\frac{d(\Delta \phi)}{dt} &= \Delta \omega - 2 K \cos \delta \sin \Delta \phi.
\label{twooscpre}
\end{align}
Synchronization in this simple model is achieved for $K$ exceeding the critical strength $K_c^\text{SSK} \equiv |\Delta \omega/2 \cos \delta|$, as only for such coupling can the slope $\Delta \phi$ assume a stationary value. To distinguish this one from the other estimate of the critical coupling strength $K_c$ that we will use, we have added a superscript indicating the initials of the authors in $K_c^\text{SSK}$ \cite{sakaguchi}. Since the distribution of natural frequencies $g(\omega)$ that we consider is uniform over $[-1,1]$, we take as the intrinsic frequency difference in our (many-oscillator) model the upper bound given by $|\Delta \omega| = 2$, which yields
\begin{equation}
K_c^\text{SSK} = \frac{1}{|\cos \delta|}.
\label{kssk}
\end{equation}
Its dependence on $\delta$ is shown as a magenta dashed line in Fig.~\ref{figPD}.
\begin{figure}[t!]
\includegraphics[scale=0.56]{PhaseDiagram.png}
\vspace{-0.2cm}
\caption{{\sf \bf Theoretical and numerical estimates of the critical coupling strength $K_c$ as a function of $\delta$.} The estimate based on a two-oscillator model $K_c^\text{SSK}$, Eq.\ (\ref{kssk}), is shown as a magenta dashed line. The many-oscillator estimate $K_c^\text{O}$, Eq.\ (\ref{ko}), is shown as a blue dotted line. Numerical estimates based on the stability of the roughness $W_{\phi}(L,t)$ are represented as isolated points, with symbols depending on the system size $L$ (see legend). It has been proved \cite{strogatz} that synchronization in a typical chain of oscillators for $\delta=0$ requires a critical coupling strength $K_c$ that increases as $\sqrt{L}$, which is also illustrated. See text for definitions and references.}
\label{figPD}
\end{figure}
As long as $\delta \neq \pm \pi/2$, for $K\geq K_c^\text{SSK}$ the phase difference $\Delta \phi$ stabilizes, and the two-oscillator model achieves synchronization. The stable equilibrium of the second line in Eq.\ (\ref{twooscpre}) is found for $\Delta \phi$ satisfying the following two conditions:
\begin{equation}
\cos \delta \cos \Delta \phi > 0,\ \ \sin \Delta \phi = \frac{\Delta \omega}{ 2 K \cos \delta}.
\label{twoosccond}
\end{equation}
According to the first condition in Eq.\ (\ref{twoosccond}), in terms of the principal value of $\Delta \phi$, the stable fixed point is reached for $\Delta \phi \in (-\pi/2,\pi/2)$ or for $\Delta \phi \in (-\pi,-\pi/2)\cup (\pi/2,\pi]$, depending on whether $\delta$ is in one or the other subset of $(-\pi,\pi]$. Moreover, the second condition in Eq.\ (\ref{twoosccond}) implies that $|\sin \Delta \phi| = K_c^\text{SSK}/K$. Therefore, as $K (> K_c^\text{SSK})$ increases $\Delta \phi$ gets closer to $0$ or $\pi$ (again, depending on the sign of $\cos \delta$).
\begin{figure*}[t!]
\includegraphics[scale=0.39]{FigSat.png}
\vspace{-0.7cm}
\caption{{\sf \bf Roughness $W_\phi(L,t)$ as a function of time for the oscillator lattice defined by Eq.\ (\ref{eqsim}) for different values of $\delta$ in each panel:} (a) $\delta = 0$, (b) $\delta = \pi/4$, (c) $\delta = \pi/2$, (d) $\delta = 3\pi/4$, and (e) $\delta = \pi$. For each $\delta$ we consider sizes $L=16$, 64, and 256 (denoted by different symbols and line styles) and coupling strengths $K = 0,1,4,7$ (denoted by different colors); see the legends in panels (a) and (b), which are valid for all panels. Results are averages over $200$ realizations (i.e., assignments of the intrinsic frequencies). }\label{figsat}
\end{figure*}
We focus on $\delta \in (-\pi/2,\pi/2)$ ($\cos \delta > 0$) in Fig.~\ref{figPD}, as it is only there that the small-slope approximation $\Delta \phi \ll 1$ can be expected to hold. The two oscillators synchronize for $K\geq K_c^\text{SSK}$, with a stationary $\Delta \phi$ (whose sign is given by that of $\Delta \omega$), that gets closer to $0$ with increasing $K$. They evolve at the common frequency
\begin{equation}
\omega_1^\text{eff} = \omega_2^\text{eff} = \bar{\omega} - K \sin (\delta) \left(1-\sqrt{\frac{K^2 - (K_c^\text{SSK})^2}{K^2}}\right),
\label{omegaeff2osc}
\end{equation}
which, for $\delta \neq 0$, is shifted away from the average of the intrinsic frequencies $\bar{\omega}$ by an amount that is proportional to $-\sin \delta$. That shift decreases as $K$ increases, with the maximum being reached at the critical point itself $K = K_c^\text{SSK}$. While this two-oscillator model cannot provide an accurate picture of the many-oscillator system, it is in fact quite useful for discussing various points related to the stationary slope distribution and the general phenomenology of our model, as will be shown below.
In Ref.~\cite{ostborn} the full many-oscillator problem is analyzed, and synchronization is shown to be possible above a given coupling strength $K_c^\text{O} \equiv (\omega_\text{max}-\omega_\text{min})/|d(\widetilde{\Delta \phi})|$ (again the superscript indicates the initial of the author). Here $d(\Delta \phi) = [\Gamma(\Delta \phi) + \Gamma(-\Delta \phi)]/K =2\sin\delta (\cos{\widetilde{\Delta \phi}}-1),$ where in the last step we have particularized the coupling function to the one given in Eq.~(\ref{gamma}). The phase difference or slope $\widetilde{\Delta \phi}$ is defined to be the largest value such that $\Gamma(\Delta \phi)$ is increasing for all $\Delta \phi < |\widetilde{\Delta \phi}|$ [as our $\Gamma(\Delta \phi)$ always is in some neighborhood of the origin for $\delta \in (-\pi/2,\pi/2)$]. We find $\widetilde{\Delta \phi} = \pi/2-\delta$ for $\delta \in (0,\pi/2)$ and $-\pi/2-\delta$ for $\delta \in (-\pi/2,0),$ which yields $d(\widetilde{\Delta \phi}) = -2 \sin \delta (1-|\sin\delta|)$. Hence, the estimate of the critical coupling strength in our model,
\begin{equation}
K_c^\text{O} = \frac{1}{|\sin \delta| (1-|\sin\delta|)},
\label{ko}
\end{equation}
which is shown as a blue dotted line in Fig.~\ref{figPD}. In principle this estimate works for an arbitrarily large system, given its lack of dependence on $L$.
Both estimates $K_c^\text{SSK}$ and $K_c^\text{O}$ show a divergence of the critical coupling as $\delta \to \pm \pi/2$, see Fig.~\ref{figPD}, which will be important in the discussion of the following subsection. $K_c^\text{O}$ can be expected to work better, as it based on a sophisticated analysis of the full many-oscillator problem. Furthermore, it diverges for $\delta \to 0$, as $K_c$ is known to do in the thermodynamic limit \cite{strogatz}. On the other hand, $K_c^\text{O}$ appears to overestimate $K_c$ in some cases \cite{ostborn}, and possible discrepancies have been found in a recent contribution \cite{moroney}. The remaining information displayed in Fig.~\ref{figPD} will be discussed in Sec.\ \ref{rel}.
\subsection{Numerical results: saturation of the roughness $W_\phi(L,t)$ and stationary slopes $\Delta \phi$}
We next explore numerically the existence of synchronous motion in the model, and its dependence on $\delta$ and the system size $L$. In Fig.~\ref{figsat} we show the roughness $W_\phi(L,t)$ (\ref{wphi}) as a function of time for the oscillator lattice defined by Eq.~(\ref{eqsim}), each panel containing results for one possible value of $\delta$, including $\delta = 0$, $\pi/4$, $\pi/2$, $3 \pi/4$, and $\pi$. In order to investigate the role played by the system size, we consider systems with $L=16$, $64$, and $256$ oscillators. Let us first focus the discussion on the case of odd coupling, $\delta = 0$ and $\pi$, panels (a) and (e), respectively. The saturation of $W_\phi(L,t)$ for long times indicates the occurrence of synchronization in the system, as the dispersion of the phase values becomes time-independent. While synchronization is indeed achieved for sufficiently large coupling strength $K$, the critical value clearly increases with the system size, as expected from the exact results for $\delta = 0$ \cite{strogatz}. Take for example the case of $K=4$ (green lines), which stabilizes the width for long times if $L=16$ (squares), but not for $L=64$ (stars) or $256$ (circles). In the case of $K=7$ (magenta lines), it is only the largest system that does not synchronize, while no value of $L$ induced saturation for $K=1$ (red lines) or, as expected, for $K=0$ (black lines).
\begin{figure*}[t!]
\hspace{-0.2cm}\includegraphics[scale=0.383]{FigSlopes.png}
\vspace{-0.2cm}
\caption{{\sf \bf Distributions of slopes $\Delta \phi$ for the oscillator lattice defined by Eq.\ (\ref{eqsim}) for $L=256$, $K = 10$, and with different values of $\delta$ in each panel :} (a) $\delta = 0$, (b) $\delta = \pi/4$, (c) $\delta = \pi/2$, (d) $\delta = 3\pi/4$, and (e) $\delta = \pi$. The red dotted lines in (b) and (d) indicate the position of the upper bounds for $\Delta \phi$ of the two-oscillator condition, Eq.\ (\ref{twoosccond}). Each histogram is based on at least $2000$ realizations at time $t = 5000$ (sufficiently large for synchronization to be attained).}
\label{figslopes}
\end{figure*}
In contrast, for $\delta = \pi/4$ and $3\pi/4$, panels (b) and (d) in Fig.~\ref{figsat} respectively, the critical coupling strength seems to be independent of the system size, as argued in Ref.~\cite{ostborn}.
%which was discussed in the previous subsection.
A qualitatively similar behavior is observed for $\delta = -\pi/4$ and $-3\pi/4$, as expected from the invariance mentioned at the end of Sec.\ \ref{nummod}, as well as for other inspected values in each of the four quadrants (not shown). As for $\delta = \pi/2$ [the even-symmetry case $\Gamma({\Delta \phi}) - \Gamma({-\Delta \phi}) = 0$], considered in Fig.~\ref{figsat}(c), the behavior is completely different and synchronization is never achieved. In fact, the coupling between oscillators does not even reduce the rate of increase of the roughness with respect to the uncoupled ($K=0$) case, and the same behavior is observed for $\delta = -\pi/2$. Both estimates of the critical coupling strength $K_c$ considered in the previous subsection diverge for $\delta\to \pm\pi/2$, see Fig.~\ref{figPD}, and in fact not even a system of two oscillators synchronizes, so this is expected.
These numerical results for the roughness $W_\phi(L,t)$ will be helpful in finding values of $K(>K_c)$ appropriate for the study of synchronous dynamics for different $\delta$ in the following. But before doing that, we need to determine the range of $\delta$ for which the small-slope approximation in Eq.~(\ref{eq4}) is expected to hold. To that end, we focus on the stationary slopes $\Delta \phi$ upon saturation in the synchronous regime. Fig.~\ref{figslopes} shows the distribution of slopes $\Delta \phi$ at long times in systems of $L=256$ oscillators for the same five values of $\delta$ displayed in Fig.~\ref{figsat}. The coupling strength of choice, $K=10$, is high enough to achieve synchronization in all cases, except (of course) for $\delta = \pi/2$.
Let us focus initially on the two two bimodal distributions in panels (b) and (d), corresponding to $\delta = \pi/4$ and $3\pi/4$ (non-odd coupling), respectively. The first of the two conditions in Eq.~(\ref{twoosccond}) rightly predicts the centering of the peaks around $0$ or $\pi$, depending on the sign of $\cos \delta$. The second condition in Eq.~\eqref{twoosccond} yields the upper and lower bounds $\pm \arcsin(K\cos \delta)^{-1}$, for $\Delta \omega = \pm 2$, which are shown as red dotted lines in Fig.~\ref{figslopes} and roughly correspond to the peaks observed around $\Delta \phi = 0$ for $\delta = \pi/4$ [panel (b)] and around $\Delta \phi = \pi$ for $\delta = 3 \pi/4$ [panel (d)]. Indeed these peaks get closer to $0$ or $\pi$ (as the case may be) as $K$ is increased (not shown).
Furthermore, the two-oscillator model, Eq.\ (\ref{twoosc}), also captures qualitative aspects of the phenomenology presented in Section \ref{nonodd}, some of which are briefly discussed in the original reference \cite{sakaguchi}. We summarize here those that shed light on the results reported in Fig.~\ref{figslopes}. When two oscillators lock, the frequency they achieve for $\delta\neq 0,\pi$ and $K\geq K_c^\text{SSK}$ is higher in absolute value than their average frequency, see Eq.~(\ref{omegaeff2osc}), a shift that in turn must influence oscillators in some neighborhood, which need to adjust to it. The net result (as shown in the numerics of Section \ref{nonodd}) is a coarsening dynamics of clusters of contiguous synchronized oscillators, which merge and, as a result, gradually become less in number and faster. Similar results were recently reported in Ref.~\cite{moroney} and, in the context of growing interfaces, Ref.~\cite{szendro}. In a sense, at each boundary between two clusters there is a new two-oscillator problem, only with ``renormalized'' intrinsic frequencies that also include the effect of further oscillators. The picture is too complex to be fully captured by a two-oscillator model, though, as there may be nontrivial feedback effects. At any rate, the tendency towards having a bimodal distribution of slopes can be inferred from the numerical results in Section \ref{nonodd}, and in Appendix \ref{AppA}, which show a gradual disappearance of slower clusters with smaller stationary slopes. The random assignment of intrinsic frequencies $\omega$ [i.e.\! the term that, together with the coupling to neighbors in Eq.~(\ref{eqsim}), contributes to achieving the right effective frequency $\omega^\text{eff}$] causes the finite width of the peaks.
In Figs.~\ref{figslopes}(a) and \ref{figslopes}(e) we observe unimodal distributions centered around $0$ (for $\delta = 0$) or $\pm \pi$ ($\delta = \pi$). The first condition in Eq.~(\ref{twoosccond}) once again rightly predicts the centering of the peaks. For such values of $\delta$ (odd coupling), in the two-oscillator model synchronization is achieved at the average frequency $\bar{\omega}$, see Eq.~(\ref{omegaeff2osc}). Extrapolating once more this picture to the many-oscillator problem, when two oscillators lock, the effect must propagate to neighboring oscillators, smoothening the differences in effective frequencies, which gradually become closer to the mean of the frequency distribution (which is zero). As there are all types of (``renormalized'') effective $\Delta \omega$ in the many-oscillator problem, which moreover become smaller in absolute value with time, the unimodal distribution around zero is expected, given the proportionality of $\sin \Delta \phi$ and $\Delta \omega$ in Eq.~(\ref{twoosccond}). This situation is also analyzed in Appendix \ref{AppA} for $\delta = 0$.
The only non-synchronous evolution in Fig.~\ref{figslopes} is observed for $\delta = \pi/2$, panel (c), for which the slopes keep evolving at arbitrarily long times. We find a triangular distribution of $\Delta \phi$, which is known \cite{papoulis} to be the PDF of a random variable (the local slope $\Delta \phi$) defined as the difference between two uniformly-distributed variables (the two phases whose difference yields $\Delta \phi$). This is in agreement with the main conclusions drawn above about this highly peculiar case, and in fact confirms that the overall behavior is similar to that of uncoupled oscillators, as shown in Fig.~\ref{figsat} (c).
\subsection{Validity of the small-slope approximation and parameter choices}
\label{rel}
We next reassess the previous results from the perspective of Eq.\ (\ref{eq4}), obtained in the continuum limit, with a special emphasis on the validity of the small-slope approximation. This will help us restrict the range of $\delta$, and, guided by the analysis of Sec.\ \ref{kc}, choose appropriate values for the coupling strength $K$. We shall consider separately situations for which $\cos \delta > 0$ and those for which $\cos \delta < 0$; the $\cos \delta = 0$ condition is only to be considered as a limiting case of the former as $\delta \to \pm \pi/2$. We always assume that $K> K_c$.
When $\cos \delta > 0$, for sufficiently large $K$ the phase differences, as illustrated in Figs.~\ref{figslopes}(a) and \ref{figslopes}(b), are close to zero and the small $\Delta \phi$ approximation leading to Eq.\ (\ref{eq4}) is justified. The parameters of this continuum equation are given by the derivatives of the coupling function, Eq.\ (\ref{gamma}), at $\Delta \phi = 0$, yielding
\begin{equation}
\nu = a^2 K \cos \delta,\ \ \frac{\lambda}{2}= - a^2 K \sin \delta.
\label{gammaders}
\end{equation}
As previously explained, for $\delta = 0$ the KPZ nonlinearity drops out from Eq.\ (\ref{eq4}), namely $\lambda = 0$. The lack of the corresponding lateral-growth mechanism, which would result in a non-zero average excess velocity \cite{barabasi}, will set this most-studied case apart from the others in several ways, as discussed in the next section. On the other hand, as $\delta$ approaches $\pi/2$ and $-\pi/2$ ($\cos \delta \to 0$), $\nu \to 0$, and thus disappears the smoothening effect of the surface tension \cite{barabasi}. This might provide a complementary explanation for the lack of synchronization observed in panel (c) of Figs.~\ref{figsat} and \ref{figslopes}, at least for the initial stages starting from the flat initial condition, as later on the small $\Delta \phi$ approximation is expected to fail. Incidentally, for the case of time-dependent noise, the KPZ equation with zero surface tension has been recently shown to define a universality class that differs from the standard KPZ class with respect to the scaling exponent values, the fluctuation statistics, and the type of dynamic scaling Ansatz which, rather than being of the FV type, is intrinsically anomalous \cite{rodriguezfernandez1,rodriguezfernandez2}.
For $\cos \delta < 0$ and sufficiently large $K$, on the other hand, as shown in Figs.~\ref{figslopes}(d) and \ref{figslopes}(e), the phase differences stabilize around $\pm \pi$. That may explain why the roughness $W_{\phi}(t)$ for $\delta = 3\pi/4$ and $\pi$ in Fig.~\ref{figsat} is initially larger for $K>0$ than for the uncoupled $K=0$ case: there are `forces' pushing neighboring phases to distance from each other as much as possible. In those cases, the perturbative expansion leading to Eq.\ (\ref{eq4}) is unjustified.
%Lastly, it is tempting to relate the fact that the results reported in Figs.~\ref{figsat} and~\ref{figslopes} are unaffected if $\delta$ is replaced by $- \delta$ (not shown), to a change of sign in $\lambda$ while that of $\nu$ is preserved, see Eq.~(\ref{gammaders}). In kinetic roughening models, this implies that the growth in the normal direction to the surface points downwards instead of upwards, which, at least in the presence of thermal noise, is known not to change the large-scale properties of the growth dynamics \cite{}. \ric{Refs?} Nevertheless, this symmetry also holds for $\cos \delta < 0$, where the small slope $\Delta \phi$ approximation on which the derivation of the KPZ equation with columnar noise is based does not hold. Indeed this must be due to the invariance under $\phi_i \to - \phi_i$ and $\delta \to -\delta$ discussed in Sec.\ \ref{nummod}, of which the up-down change of sign in the KPZ nonlinearity can be considered a manifestation at the coarse-grained level of description.
Lastly, the fact that the results reported in Figs.~\ref{figsat} and~\ref{figslopes} are unaffected by a change in the sign of $\delta$ (not shown) can be traced back to the invariance of the oscillator system under the global sign flip $\phi_i \to - \phi_i$ and $\delta \to -\delta$, discussed in Sec.\ \ref{nummod}. For $\cos \delta > 0$, when the continuum limit description is justified, this fact is possibly related with the independence (at least for the case of time-dependent noise) of the large-scale behavior with the sign of $\lambda$ \cite{barabasi,krug97}. Indeed, the only feature of the KPZ universality class that depends on this sign is that of the skewness of the (TW) fluctuation distribution \cite{kriecherbauer10,takeuchi}.
From the above analysis, it becomes clear that the small-slope approximation on which the effective description given by Eq.~(\ref{eq4}) is based can only be valid for $\delta \in (-\pi/2,\pi/2)$. In other words, for $\cos \delta > 0$, see the first condition for synchronization in the two-oscillator model, Eq.~(\ref{twoosccond}). In subsequent sections, for the most part we will actually focus on the interval $\delta\in[0, \pi/2)$, because the invariance under simultaneous phase and $\delta$ reversal makes the interval $(-\pi/2,0)$ somewhat redundant, as just explained. Only occasionally will we consider cases for which $\cos \delta < 0$. Specifically, we will study synchronization for $\delta = 0$ in Section \ref{odd}, and synchronization for $\delta = \pi/4$ (and some aspects for $\delta = -\pi/4$) in Section \ref{nonodd}, the latter results being complemented with those obtained for $\delta = \pi/8$ and $3\pi/8$ (for some specific observables in that same section, the rest being the subject of Appendix \ref{AppB}), which are seen to be qualitatively very similar. We remind the reader that the $\delta = 0$ case is quite special, as it is that of odd coupling, lacking synchronization in the thermodynamic limit, and allowing for an effective description through the Larkin model. The rest are instances of non-odd coupling, which leads to finite $K_c$ for arbitrary system size $L$, and an effective description given by the KPZ equation with columnar noise.
In the following, we focus our numerical analysis on values of $K(>K_c)$ large enough so that synchronization is achieved for sufficiently small stationary slopes $\Delta \phi$. This is in principle always possible, though for $\delta = 0$ it requires choosing some $K$ that grows with the system size as $\sqrt{L}$ \cite{strogatz}. To that end, we will consider coupling strengths $K$ that are roughly twice as large as our numerical estimate for $K_c$. The latter is defined as the smallest value of $K$ for which we obtain a saturation of the roughness $W_\phi(L,t)$ for $200$ realizations (as many as those employed in the results reported in Fig.~\ref{figsat}). Such numerical estimates, which are likely to be lower bounds for the actual value of $K_c$, are shown in Fig.~\ref{figPD} for different system sizes $L$ (see legend). They are consistently larger than the two-oscillator estimate $K_c^\text{SSK}$ (\ref{kssk}), as expected, and smaller than $K_c^\text{O}$ (\ref{ko}), in agreement with previous results \cite{ostborn}. For the case of $\delta = 0$, the estimate does grow with $\sqrt{L}$, while in the other cases it appears to be only weakly dependent on $L$. The precise coupling strength $K$ chosen for each value of $\delta$ will be mentioned in the discussion of the corresponding numerical results.
\section{Synchronization as a growth process for Kuramoto coupling ($\delta = 0$)}
\label{odd}
%\subsection{Numerical simulations}
%\label{numdelta0}
In this section we address the numerical solution of Eq.~(\ref{eqsim}) for $\delta = 0$, which becomes
\begin{equation}
\frac{d\phi_i}{dt} = \omega_i + K [\sin(\phi_{i+1}\!-\!\phi_i) +\sin(\phi_{i-1}\!-\!\phi_i)].
\label{eqsimodd}
\end{equation}
This is the celebrated Kuramoto model in a one-dimensional lattice, which is known not to synchronize in the thermodynamic limit, $L\to\infty$, as the critical coupling stregth $K_c$ for synchronization grows with $\sqrt{L}$ \cite{strogatz}. [An illustration of the growth of the critical coupling as the size $L$ increases is provided in Figs.~\ref{figPD} and \ref{figsat}(a).] Since we are nonetheless interested in systems of finite size, we will focus on a value of $K$ that is large enough to achieve synchronization for the size under consideration. By studying the saturation of $W_\phi(t)$ (see the discussion at the end of Sec.\ \ref{rel}), we find $K_c \gtrsim 0.6 \sqrt{L}$ based on systems with sizes ranging from $L = 16$ to $1024$.
\begin{figure}[t!]
\includegraphics[scale=0.47]{FigWGdelta0K40.png}
\vspace{-0.3cm}
\caption{{\sf \bf Roughness $W_\phi(t)$ and height-difference correlation function $G_\phi(\ell,t)$ for the Kuramoto model, Eq.\ (\ref{eqsimodd}) ($\delta = 0$) with size $L = 1024$ and coupling strength $K= 40$.}
({\it Upper panel}) $W_\phi(t)$ as a function of time in the synchronous regime, with error bars displaying the standard error of the average across realizations. In the power-law visual guides we have used the Larkin model values $\alpha = 3/2$ and $z=2$, see Eq.\ \eqref{SkLarkin}. The inset shows the average squared slope $\langle\overline{(\Delta \phi)^2} \rangle$ (black circles), as well as the average of the maximum squared local slope (red asterisks), and the most probable value of the squared slope (blue crosses), as a function of time for the same parameter values. The black dashed line shows the power-law growth of the average squared slopes predicted by the super-rough scaling Ansatz, Eq.\ \eqref{glt_mod}, as $t^{2(\alpha-\alpha_\text{loc})/z}$. ({\it Lower panel}) $G(\ell,t)$ as a function of $\ell$ in the synchronous regime for different times (see legends). The dashed line shows the power law growth with the distance $\ell^{2\alpha_\text{loc}}$. Inset: rescaling of $G(\ell,t)$ following the theoretical form in Eq.~(\ref{glt_mod}). Averages based on 2000 realizations.
}\label{figWGdelta0K40}
\end{figure}
\subsection{Kinetic roughening: scaling Ansatz and exponents}
In the upper panel of Fig.~\ref{figWGdelta0K40} we show the roughness $W_\phi(t)$ of a system of $L= 1024$ oscillators with coupling strength $K=40$. According to the numerical estimate mentioned above, the critical coupling strength in this case is $K_c\gtrsim 20$, so the strength of choice is roughly twice as large. A clear region of growth with exponent $\beta = \alpha/z = 3/4$, is observed, followed by a saturation that consolidates around $t\approx 10^3$. As will be discussed below, this corresponds to the exponent values $\alpha = 3/2$ and $z = 2$.
In the inset we show that the average squared slope $\langle\overline{(\Delta \phi)^2} \rangle$ (black circles) increases as $t^{2(\alpha-\alpha_\text{loc})/z} = t^{1/2}$ ($\alpha_\text{loc} = 1$) (see the black dashed line) until the growth saturates \cite{lopez99}. Indeed, this is one of the hallmarks of anomalous scaling \cite{dassarma, lopez}: the local slopes, instead of stabilizing rapidly as the correlation length $\xi(t)$ exceeds their length scales, keep on increasing with time until saturation, $\xi(t)\sim L$. This increase is given by Eq.\ (\ref{glt_mod}) when the spectral roughness exponent $\alpha_s > 1$, which is shown to be the case below. Such a saturation of the slopes may provide a natural
%context in which to seek a physical
explanation for the fact that one-dimensional systems of oscillators with Kuramoto coupling $\Gamma(\Delta \phi) = K \sin(\Delta \phi)$ do not synchronize in the thermodynamic limit, where saturation can never be achieved \cite{strogatz}. The average of the maximum squared local slope, i.e., the average across realizations of the maximum squared slope in the system, see also the inset of the upper panel of Fig.~\ref{figWGdelta0K40} (red asterisks), also increases similarly. We also report the most probable value of the squared slope (blue crosses), which always stays close to zero, as corresponds to an asymptotic unimodal distribution with zero mean, see Fig.~\ref{figslopes}(a). This feature, which is also explored in Appendix \ref{AppA}, will reveal a sharp distinction between the Kuramoto coupling ($\delta = 0$) and other couplings for $\delta \neq 0$ in the next section and in the appendices.
A more detailed view on the two-point correlations beyond nearest neighbors [note that the average squared slope $\langle\overline{(\Delta \phi)^2} \rangle = G_\phi(1,t)$] can be obtained by inspection of the full height-difference correlation function $G_\phi(\ell,t)$, shown in the lower panel of Fig.~\ref{figWGdelta0K40}. The correlation function behaves as expected from a system displaying super-rough anomalous scaling, see Eq.~(\ref{glt_mod}), with a dependence on distance of the form $\ell^2$ that reveals a local exponent $\alpha_{\text{loc}} = 1$, and a growth in time that proceeds until the width reaches the saturation regime. In the inset we show $G_\phi(\ell,t)/t^{2 \alpha/z}$, as a function of a rescaled time variable $u= \ell/t^{1/z}$, which scales as $u^{2 \alpha_\text{loc}}$ for $u\ll 1$ and reaches a constant value for $u\gg 1$, as expected from Eq.\ \eqref{glt_mod}. This is all in agreement with the theory of super-rough scaling, which captures both the qualitative features and the precise values of the characteristic exponents of synchronization for Kuramoto coupling, agreeing in turn with those of the Larkin model.
Indeed, from the discussion in Sec.\ \ref{contlim}, the small-slope approximation for odd coupling functions yields for the Kuramoto model
\begin{equation}
\partial_t \phi({\bf x},t) = \omega({\bf x}) + \nu \nabla^2 \phi({\bf x},t),
\label{eq6}
\end{equation}
where we have neglected terms of order $\mathcal{O}(a^4)$ and the surface tension $\nu$ is defined as in Eq.~(\ref{eq4}). Notably, now the KPZ nonlinearity is absent, as $\lambda \propto \Gamma^{(2)}(0) = 0$. Equation \eqref{eq6} is the celebrated Larkin model, namely, the EW equation with columnar noise, in which $\nu> 0$ plays the role of an elastic constant. The Larkin model was proposed in the study of pinning of vortex lines in type II superconductors \cite{larkin} and some of its main features from a kinetic roughening perspective are reviewed in Ref.~\cite{purrello}. Equation (\ref{eq6}) has also been analyzed in the context of synchronization in Ref.~\cite{hong}, where the Kuramoto model on a lattice ($\Gamma(\Delta \phi) = \sin(\Delta \phi)$, $\nu = a^2 > 0$) was studied.
The solution of Eq.\ \eqref{eq6} is straightforward in Fourier space, as the equation is linear.
%we repeat the main steps here. In terms of the Fourier components, Eq.~(\ref{eq6}) \rod{reads}
%\begin{equation}
%\partial_t \hat{\phi}({\bf k},t) = \hat{\omega}({\bf k}) - \nu k^2 \hat{\phi}({\bf k},t),
%\end{equation}
%where $\langle \hat{\omega}({\bf k}) \rangle = 0$ and $\langle \hat{\omega}({\bf k})\hat{\omega}({\bf k}') \rangle = (2\pi)^d\, 2 %\sigma\, \delta({\bf k} + {\bf k}')$. The solution of this equation in Fourier space is
%\begin{equation}
%\hat{\phi}({\bf k},t) = \hat{\phi}({\bf k},0)\, e^{-\nu k^2 t} + \frac{\hat{\omega}({\bf k )}}{\nu k^2} \left(1-e^{-\nu k^2 t} \right).
%\end{equation}
For the initial condition of a homogeneous phase profile, ${\phi}({\bf r},0) = 0$, the exact structure factor reads \cite{purrello}
\begin{equation}
S_\phi({\bf k},t) = \langle \hat{\phi}({\bf k},t) \hat{\phi}(-{\bf k},t)\rangle = \frac{(2\pi)^d 2 \sigma}{\nu^2 k^4} \left(1-e^{-\nu k^2 t} \right)^2.
\label{SkLarkin}
\end{equation}
Comparing with the generic scaling ansatz in Eq.~(\ref{Sk}), the roughness exponents are $\alpha = \alpha_s = 3/2$ and the dynamic exponent is $z=2$ \cite{ramasco,purrello}. Given that this structure factor follows the FV scaling form, Eq.\ (\ref{SkFV}), it corresponds to a system with super-rough anomalous scaling.
%We next study numerically the phase oscillator model in 1D for $\delta = 0$, whose continuum limit behavior is characterized by the roughness exponents $\alpha = \alpha_s = 3/2 > 1$. \rod{Hence, this corresponds to super-rough \rod{anomalous} scaling for the height-difference correlation fucntion $G(\mathbf{r},t)$ as in Eq.\ \eqref{glt_mod},} with a local roughness exponent $\alpha_\text{loc}= 1$ \cite{lopez,ramasco}.
Actually, the analytical solution of the Larkin model, Eq.\ \eqref{SkLarkin}, describes very accurately the structure factor calculated in simulations of the Kuramoto model, Eq.\ \eqref{eqsimodd}. Results of the latter are provided in Fig.\ \ref{figSkdelta0K40}, which shows the numerical $S_\phi({\bf k},t)$ as a function of $k$ for different values of time.
\begin{figure}[h!]
\includegraphics[scale=0.47]{FigSkdelta0K40.png}
\vspace{-0.3cm}
\caption{{\sf \bf Structure factor $S_\phi(k,t)$ as a function of time for the Kuramoto model, Eq.\ (\ref{eqsimodd}), with $L = 1024$, and $K= 40$.} ({\it Upper panel}) $S_\phi(k,t)$ as a function of $k$ (colored dots) and analytical expression for the Larkin model, Eq.\ \eqref{SkLarkin} (dotted lines). ({\it Lower panel}) $k ^{2\alpha+1} S_\phi(k,t)$ as a function of $y = k t^{1/z}$. In all rescalings and power-law visual guides, $\alpha = \alpha_s = 3/2$ and $z=2$. The structure factor curves are shown for the same time points displayed in the lower panel of Fig.~\ref{figWGdelta0K40} (see the corresponding legend). Averages based on 2000 realizations.}
\label{figSkdelta0K40}
\end{figure}
In the upper panel, the structure factor of the Larkin model, Eq.~(\ref{SkLarkin}) (represented as a black dotted line), shows excellent agreement with the simulation data of the oscillator lattice. We have particularized the analytical expression for $\nu = \Gamma^{(1)}(0) = K$, the overall constant factor, which includes the (unknown) noise strength $\sigma$ in the continuum limit, being manually adjusted to the data but time-independent. The agreement for large $k$ values improves by replacing the wave number $k$ by $2 \sin(k/2)$, which makes the structure factor a periodic function of $k$, as implied by the existence of a finite wave-vector cut-off in the oscillator lattice \cite{siegert96}. This replacement has very little effect at small $k$. Once again, visual guides (represented as black dashed lines) based on the scaling form, Eq.\ (\ref{Sk}) with the theoretical exponents corresponding to FV scaling with super-rough behavior are confirmed to work extremely well, including the spectral roughness exponent $\alpha_s = 3/2$.
%(which by its very nature could not be investigated in the real space results of Fig.~\ref{figWGdelta0K40}).
This is further confirmed in the lower panel, where we perform a data collapse of the
$S(k,t)$ data which agrees with Eq.\ \eqref{Sk},
except for wave numbers in close proximity to the cutoff. The behavior just described for the structure factor is known to imply \cite{lopez} the anomalous scaling and exponent values of the roughness and height-difference correlation functions assessed in Fig.\ \ref{figWGdelta0K40}.
%by plotting $k^{2 \alpha +1} S(k,t)$ we obtain the scaling function $s(y)$, \rod{which behaves as expected from
%with $y = k t^{1/z}$, where $y$ approaches a constant value for $y\gg 1$ while $s(y) \propto y^{2\alpha +1}$ for $y\ll 1$.
%%%%% Fin factor de estructura %%%%%%%%%%%%%
\begin{figure}[t!]
\includegraphics[scale=0.36]{FigTrajdelta0K40.png}
\vspace{-0.7cm}
\caption{{\sf \bf Phases $\phi_i$ and instantaneous frequencies $d\phi_i/d t$ across time in a representative trajectory of the Kuramoto model, Eq.\ (\ref{eqsimodd}), with $L = 1024$ and $K= 40$.} ({\it Left panel}) Phases $\phi_i$ for $t = 200, 400, 600, 1600$, and $2600$. Earlier times are shown in darker colors, later times in lighter colors. ({\it Right panel}) Instantaneous frequencies $d{\phi}_i/dt$ for the same trajectory, at the same time points.}
\label{FigTrajdelta0K40}
\end{figure}
\subsection{Dynamics of synchronization}
To shed light on the phenomenology underlying this super-rough scaling, in Fig.~\ref{FigTrajdelta0K40} we show a representative trajectory. See file S1 at Ref.\ \cite{suppl} for a movie of the full dynamics. Specifically, in Fig.~\ref{FigTrajdelta0K40} we show the phases $\phi_i$ (left panel) and instantaneous frequencies $d\phi_i/d t$ (right panel) as functions of time in one realization included in the averaging of the structure factor displayed in Fig.~\ref{figSkdelta0K40}. Different colors correspond to different times, with darker colors indicating earlier times and lighter colors later times. At the initial time considered ($t=200$) we find a configuration where important differences in the instantaneous frequencies exist (partly due to the differences in intrinsic frequencies, partly to the coupling to the neighboring oscillators). These differences lead to oscillators evolving at different rates, so that the `interface' becomes rougher. As a result of this roughening, however, there is a progressive reduction of the differences in the instantaneous frequencies, which eventually leads to a synchronous dynamics.
The evolution appears to be quite modest and slow compared to what we will see in the next section for $\delta \neq 0$, in the sense that, apart from slight rearrangements of the slopes, the main motion is a uniform displacement at the average frequency $\omega^\text{eff}$ that is eventually attained by all the oscillators. Indeed, if one performs the space average of Eq.~(\ref{eq6}), due to the periodic boundary conditions and to the fact that the mean intrinsinc frequency $\langle \omega \rangle$ is zero, one obtains a vanishing interface velocity. For time-dependent noise, this is a standard feature of EW surface growth \cite{barabasi}. At the discrete level, from Eq.~(\ref{eqsimodd}) we obtain, after a slight manipulation of the indices,
\begin{equation}
\overline{\frac{d\phi_i}{d t}} = \overline{\omega_i } + K \overline{[\sin(\phi_{i+1}\!-\!\phi_i)+\sin(\phi_i\!-\!\phi_{i+1})]}= \overline{\omega_i}.
\label{meanfreqodd}
\end{equation}
%where we have denoted by an overline a spatial average across the system as in Sec.\ \ref{obs}.
The odd symmetry of the function $\Gamma(\Delta \phi) = \sin(\Delta \phi)$ makes the coupling term vanish in Eq.\ (\ref{meanfreqodd}). If the system is sufficiently large, the average instantaneous frequency becomes zero, $\overline{\frac{d \phi_i}{dt} } = \overline{\omega_i} \approx \langle \omega \rangle = 0$, by the law of large numbers.
\begin{figure}[h!]
\includegraphics[scale=0.5]{FigDistalldeltas.png}
\vspace{-0.3cm}
\caption{{\sf \bf Fluctuation histogram for $\varphi_i$ defined as in Eq.\ \eqref{fluct} for the oscillator lattice, Eq.\ \eqref{eqsim}, in a system of $L=1024$ oscillators, for $\delta = 0$ ($K=40$), $\delta = \pi/8$ ($K=4$), $\delta = \pi/4$ ($K=4$), and $\delta = 3 \pi/8$ ($K=8$).} (Upper panel) Linear scale. (Lower panel) Logarithmic scale. We always take as reference time $t_0 = 50$, and values of $\Delta t = 50, 100, 150, \ldots$ in the interval where a power-law growth $t^\beta$ is observed. Specifically, for $\delta = 0$ (Kuramoto model), the largest value of $\Delta t$ is $150$; for $\delta = \pi/8$, the largest value is $450$; and for $\delta = \pi/4$ and $3\pi/8$ it is $250$. The black solid line corresponds to a normalized Gaussian distribution with zero mean and unit standard deviation, while the dotted line shows the normalized GOE-TW distribution, see Sec.\ \ref{nonodd}. No fitting parameters are used. Histograms based on $10^4$ realizations.}
\label{FigDistalldeltas}
\end{figure}
\subsection{Fluctuation statistics}
\label{sec:flKuramoto}
Finally, beyond scaling exponents and the general phenomenology of trajectories, we also explore the statistics of the phase fluctuations, as recent developments in kinetic roughening show that the fluctuation PDF and covariance are also essential traits to define a universality class, see e.g.\ Refs.\ \cite{takeuchi,rodriguez-fernandez21,marcos22} and others therein. In Fig.~\ref{FigDistalldeltas} we show the numerically-obtained PDF of the fluctuations $\varphi_i$ defined as in Eq.~(\ref{fluct}) for several values of $\delta$. Our discussion for the moment is focused on the case $\delta = 0$, see the magenta circles (again, $L=1024$ and $K =40$). An initial time of $t_0 = 50$ and time windows $\Delta t = 50, 100$, and $150$ have been chosen, which correspond to the intermediate growth regime, recall the upper panel of Fig.~\ref{figWGdelta0K40}. In Fig.~\ref{FigDistalldeltas} the same distribution is shown in linear (upper panel) and in logarithmic scales (lower panel). The histogram has been rescaled and shifted so that the sample mean is zero and the sample standard deviation is one. The purpose is to make the comparison with a normalized Gaussian PDF, which is shown with a black solid line. A very good agreement is found between the numerical results and the theoretical distribution without any kind of fit, which confirms expectations based on the good agreement between the Kuramoto and the Larkin models, and the fact that the latter is linear. Indeed, for time-dependent noise the EW equation is another well-kown instance of Gaussian height fluctuations \cite{barabasi,krug97}. Another reason for stressing the occurrence of Gaussian fluctuations in the synchronization of oscillator lattices with Kuramoto coupling is that in the next section we will discuss a very different fluctuation distribution for $\delta \neq 0$ (also displayed in Fig.~\ref{FigDistalldeltas}).
\begin{figure}[h!]
\includegraphics[scale=0.40]{FigCovariances.png}
\vspace{-0.4cm}
\caption{{\sf \bf Rescaled covariance $C_\phi(\ell,t)$ for the oscillator lattice, Eq.\ \eqref{eqsim}, in a system of $L=1024$ oscillators for $\delta = 0$ ($K=40$), $\delta = \pi/8$ ($K=4$), $\delta = \pi/4$ ($K=4$), and $\delta = 3 \pi/8$ ($K=8$).} The time points considered are in the growth regimes and coincide with those included in Fig.~\ref{FigDistalldeltas}. The black solid line corresponds to the exact covariance of the Larkin model, Eq.\ \eqref{eq:cLark}, while the dotted line shows the covariance of the Airy$_1$ process which has been used in the determination of the rescaling factors $a_1$ and $a_2$, see the text. Numerical covariances based on $10^4$ realizations.}
\label{FigCovariances}
\end{figure}
One further ingredient of the fluctuation statistics is the height covariance, defined for the oscillator system by $C_\phi(\ell,t)$, see Eq.~(\ref{cphi}). Numerical results for this correlation function based on Eq.\ \eqref{eqsim} are shown in Fig.~\ref{FigCovariances} for the same values of $\delta$ and the same time points employed in Fig.~\ref{FigDistalldeltas}. We again focus on the $\delta = 0$ case (Kuramoto model) for the time being, see the magenta circles. In the growth regime, the covariance should scale as \cite{kriecherbauer10,takeuchi}
\begin{equation}
C_\phi(\ell,t) = a_1 t^{2\beta} \mathcal{C}(a_2 \ell/t^{1/z}),
\label{covscal}
\end{equation}
where $\mathcal{C}$ is a scaling function, which explains the scaling of the axes in Fig.~\ref{FigCovariances}. The constants $a_1$ and $a_2$ are determined with respect to a reference value $x_0$ based on an assumed functional form for $\mathcal{C}(x)$, so that all curves pass through $(x_0, \mathcal{C}(x_0))$, see Ref.~\cite{barreales2020} for further details.
%According to this procedure, $a_1 = C_\phi(0,t)/t^{2\beta}\mathcal{C}(0)$ and $x_0 = a_2 \ell/t^{1/z}$, which allows us to obtain $a_2$ by linear interpolation in $C_\phi(x_0 t^{1/z}/a_2)/t^{2\beta} = a_1 \mathcal{C}(x_0)$.
We have chosen the covariance of the Airy$_1$ process \cite{takeuchi} as the functional form for $\mathcal{C}(x)$, which is shown as a black dotted line, a choice whose significance will discussed in the next section.
For the Kuramoto model ($\delta = 0$), the covariance again follows accurately the analytical curve of the Larkin model, which differs substantially from Airy$_1$. Indeed, since $C_\phi(\ell,t)$ is the inverse Fourier transform of the structure factor \cite{siegert96}, using Eq.\ \eqref{SkLarkin} the covariance of the Larkin model can readily be shown to comply with a scaling form like that described by Eq.\ \eqref{covscal} with $\beta=3/4$ and $z=2$, but with a scaling function which differs from Airy$_1$, namely,
\begin{equation}
C_\phi(\ell,t) = \frac{4 \pi \sigma t^{3/2}}{\nu^{1/2}} \mathcal{F}^{-1}\!\left[\frac{\left(1-e^{-\kappa^2} \right)^2}{\kappa^4}\right](\ell/\sqrt{\nu t}) ,
\label{eq:cLark}
\end{equation}
where $\kappa=\sqrt{\nu t} k$.
%$C_\phi(\ell,t) = \mathcal{F}^{-1}[S_\phi(k,t)](\ell,t) = \frac{1}{\sqrt{2\pi}} \int_{-\infty}^\infty dk\, S_\phi(k,t) e^{i k \ell}$, and the structure factor $S_\phi(k,t)$ of the Larkin model is given in Eq.~(\ref{SkLarkin}), we obtain in terms of the rescaled variable $\kappa = \sqrt{\nu t} k$
%\ric{\begin{align}
%C_\phi(\ell,t) &= \frac{1}{\sqrt{2\pi}} \int_{-\infty}^\infty dk\, \frac{4\pi\sigma}{\nu^2 k^4} \left(1-e^{-\nu k^2 t} \right)^2 e^{i k \ell} \nonumber\\
%&= \frac{4 \pi \sigma t^{3/2}}{\sqrt{\nu}} \mathcal{F}^{-1}\!\left[\frac{1}{\kappa^4} \left(1-e^{-\kappa^2} \right)^2\right](\ell/\sqrt{\nu t})
%\label{covlarkin}
%\end{align}}
%, where the inverse Fourier transform of the well-behaved function $(1/\kappa^4) (1-e^{-\kappa^2})^2$ in Eq.~\ref{covlarkin} can be expressed in terms of exponential and error functions or approximated with arbitrary accuracy via numerical integration. Hence $C_\phi(\ell,t)$ for the Larkin model can be written as the product of $t^{3/2}$ and a function of a single variable proportional to the distance rescaled by the correlation length, $\ell/t^{1/2}$, which, considering that $2\beta = 3/2$ and $z = 2$, agrees with the general scaling from in Eq.~(\ref{covscal}).
This analytical form for the covariance of the Larkin model is shown in Fig.~\ref{FigCovariances} as a black solid line, and shows an excellent agreement with the numerical data for the oscillator lattice with $\delta = 0$ across the growth interval. Surprisingly, it also shows an excellent agreement with the covariances obtained for other values of $\delta$ (which are also displayed Fig.~\ref{FigCovariances}) across their corresponding growth regimes, despite the crucial differences found for any other observables. This point will be further discussed in the next section.
\section{Synchronization as a growth process for non-odd coupling}
\label{nonodd}
%\subsection{Continuum equation}
Let us consider once again the continuum limit of the general oscillator model discussed in Sec.\ \ref{contlim}. If the coupling function is not odd, $\Gamma(\Delta \phi)+ \Gamma(-\Delta \phi) \neq 0$, in general no derivatives of $\Gamma(\Delta \phi)$ can be assumed to vanish, and the dominant contributions [neglecting terms of order $\mathcal{O}(a^4)$] in Eq.~(\ref{eq3}) are those of the KPZ equation with columnar noise, Eq.~(\ref{eq4}). This equation can be formally related (via an inverse Hopf-Cole transformation) to a diffusion problem in the presence of random traps and sources \cite{halpinhealy}. It is especially in that form that the equation has been studied, showing sharply localized solutions around pinning centers, which eventually hop to the basin of a more attracting center. This hopping dynamics proceeds across time scales, with longer times being spent at each new localization center that is attained. The connection between such form of localization and synchronization was first discussed a long time ago \cite{sakaguchi}, though it appears to be only very recently that the subject has been more thoroughly elucidated \cite{moroney}.
For the study of anomalous scaling in synchronous oscillator lattices, the main interest lies in the KPZ equation with columnar noise itself, which %. Fortunately a detailed characterization of the corresponding growth process exists
has been previously studied in the kinetic roughening literature \cite{szendro}. The localization of the diffusion problem is there shown to translate into the formation of facets, i.e.\! regions of nearly constant slope. Specifically, there is a coarsening process whereby some facets, which move uniformly at constant velocity, absorb other facets. In a finite-size system, the asymptotic solution is a triangular-looking surface growing at constant velocity. As will be seen next, in the oscillator problem the facets correspond to clusters of synchronized oscillators, and the coarsening process leads to larger (and faster) synchronized clusters, as previously observed in some models of oscillators in one and two dimensions \cite{sakaguchi,moroney}.
\subsection{Kinetic roughening: scaling Ansatz and exponents}
As we did in Sec.\ \ref{odd}, we study numerically a system of $L$ oscillators governed by Eqs.~(\ref{eqsim}), but this time $\delta = \pi/4$ (instead of $\delta = 0$). Other choices of $\delta\in (0,\pi/2)$, namely $\delta = \pi/8$ and $3\pi/8$, are considered in Appendix \ref{AppB} and lead to similar conclusions. The system synchronizes for $K$ above a critical value that is roughly $K_c\gtrsim 2$, according to the procedure described at the end of Sec.\ \ref{rel} [see also Fig.~\ref{figsat}(b), where the system does not synchronize for $K=1$ but it does for $K=4$]. Therefore, we will take as the coupling strength $K = 4$, which is approximately twice as large as the critical value.
\begin{figure}[h!]
\includegraphics[scale=0.47]{FigWGdelta0250K4.png}
\vspace{-0.3cm}
\caption{{\sf \bf Roughness $W_\phi(t)$ and height-difference correlation function $G_\phi(\ell,t)$ for the oscillator model defined by Eq.\ (\ref{eqsim}), with $\delta = \pi/4$ (non-odd coupling), $L = 1024$, and $K= 4$.}
({\it Upper panel}) $W_\phi(t)$ as a function of time in the synchronous regime, with error bars displaying the standard error of the average across realizations. In the power-law visual guides we have used the exponent values $\alpha = 1.07$ and $z=1.36$. The inset shows the average squared slope $\langle\overline{(\Delta \phi)^2} \rangle$ (black circles), as well as the average of the maximum squared local slope (red asterisks), and the most probable value of the local slope (blue crosses), as functions of time for the same parameter values. The black dashed line shows the power-law growth of the average squared slopes predicted by Eq.\ \eqref{glt_mod}, $t^{2(\alpha-\alpha_\text{loc})/z}$, for $\alpha_\text{loc}=0.97$. ({\it Lower panel}) $G(\ell,t)$ as a function of $\ell$ in the synchronous regime for different times (see legends). The dashed line shows the power-law growth with distance $\ell^{2\alpha_\text{loc}}$. Inset: data collapse of $G(\ell,t)$ following the full theoretical form of Eq.~(\ref{glt_mod}). Averages based on 2000 realizations.
}\label{figWGdelta0250K4}
\end{figure}
In what follows, we find that a consistent description of the scaling behavior obtained in our simulations can be achieved for faceted exponent values $\alpha = 1.07\pm0.05$, $\alpha_s = 1.40\pm0.05$, $z = 1.36\pm0.05$, and $\alpha_\text{loc} = 0.97\pm0.05$; these must be adjusted quite precisely in order to satisfy the various scaling constraints pertaining to the different observables, so that the quoted uncertainties are conservative.
%Nonetheless, small (positive or negative) modifications of their values, slightly larger than $0.01$ or even $0.02$ (but certainly smaller than $0.05$, which might be consider a conservative uniform upper bound for their uncertainty), sometimes give results that are visually just as good as the ones we report here.
Note also that the local roughness exponent $\alpha_\text{loc} = 1$ for faceted scaling \cite{ramasco} is an upper bound for the local exponent for geometrical reasons \cite{leschhorn}, so that the possibility of finite size effects distorting some of the exponents (for $\alpha_\text{loc}$, necessarily towards smaller values) should also be considered. Their possibly non-universal nature may also explain why these exponents differ somewhat from those of Ref.~\cite{szendro}, and also (though less so) from those found for other values of $\delta$ in Appendix \ref{AppB}.
In Fig.~\ref{figWGdelta0250K4} we show the roughness $W_\phi(t)$ and the height-difference correlation $G_\phi(\ell,t)$ for a system of $L=1024$ oscillators, from numerical simulations of Eq.\ \eqref{eqsim} with $\delta = \pi/4$ and $K=4$. All the elements, symbols and colors have the same meaning as in Fig.~\ref{figWGdelta0K40} (in that case $\delta = 0$, $K=40$). In the upper panel, the roughness saturates for times on the order to $10^3$. In the inset we show that the average squared slope $\langle\overline{(\Delta \phi)^2} \rangle$ (black circles) hardly follows the power-law growth, $t^{2(\alpha-\alpha_\text{loc})/z}$ expected from Eq.\ (\ref{glt_mod}), and becomes close to saturation at a much shorter time scale than that needed for the saturation of the roughness \cite{lopez99}. This somewhat puzzling behavior is anticipated by the early saturation of the average of the maximum squared local slope (inset, red asterisks). Indeed the latter quantity is a proxy for the evolution of the facet with the steepest slope, as will be illustrated below. The most probable value of the squared slope (inset, blue crosses) also behaves in a completely different way than it did for $\delta = 0$ (see the inset of the upper panel in Fig.~\ref{figWGdelta0K40}): at some point it leaves the very small values from which it started and rises close to the value of $\langle\overline{(\Delta \phi)^2} \rangle$ itself. This issue will be addressed in more detail in Appendix \ref{AppA}, and less specifically in Appendix \ref{AppB}, where a similar behavior is seen for other choices of $\delta \neq 0$.
A more detailed view on space correlations beyond nearest neighbors is provided by the full height-difference correlation function $G_\phi(\ell,t)$, shown in the lower panel of Fig.~\ref{figWGdelta0250K4}. A dependence on distance of the form $\ell^{2\alpha_\text{loc}}$ is observed, as well as a growth in time that proceeds until $W_\phi(t)$ reaches saturation, but it appears to slow down earlier than for $\delta = 0$. Full agreement with Eq.\ (\ref{glt_mod}) can be appreciated in the data collapse of $G_\phi(\ell,t)$ which is displayed in the inset.
%/t^{2 \alpha/z}$, as a function of rescaled time $u= \ell/t^{1/z}$, which scales as $u^{2 \alpha_\text{loc}}$ for $u\ll 1$ and reaches a constant value for $u\gg 1$.
\begin{figure}[h!]
\includegraphics[scale=0.47]{FigSkdelta0250K4.png}
\vspace{-0.3cm}
\caption{{\sf \bf Structure factor $S_\phi(k,t)$ as a function of time for the oscillator model defined by Eq.\ (\ref{eqsim}) with $\delta = \pi/4$ (non-odd coupling), $L = 1024$, and $K= 4$.}
({\it Upper panel}) $S_\phi(k,t)$ as a function of $k$. ({\it Lower panel}) Collapse of the data from the main panel.
%$k ^{2\alpha+1} S_\phi(k,t)$ as a function of $y = k t^{1/z}$
In all rescalings and power-law visual guides, $\alpha = 1.07$, $\alpha_s = 1.40$, $z=1.36$, and $\alpha_\text{loc}=0.97$. The structure factor curves are shown for the same time points displayed in the lower panel of Fig.~\ref{figWGdelta0250K4} (see the corresponding legend). Averages based on 2000 realizations.}
\label{figSkdelta0250K4}
\end{figure}
Results for the KPZ equation with columnar disorder indicate \cite{szendro} the occurrence of faceted anomalous scaling whose full assessment can only be provided by studying correlations in Fourier space. Thus, the upper panel of Fig.~\ref{figSkdelta0250K4} shows the structure factor $S_\phi({\bf k},t)$ as a function of $k$ for different times, also enabling qualitative comparison with the results discussed in the previous section. The structure factor certainly does not follow the FV form, Eq.\ (\ref{SkFV}), but indeed shows a very good agreement with the faceted scaling form of Eq.~(\ref{Sk}). This is explicitly confirmed in the lower panel, where
the $S(k,t)$ data are collapsed closely following the expected behavior, Eq.\ (\ref{Sk}),
except for wave numbers that are close to the ultraviolet (discretization) cutoff.
%by plotting $k^{2 \alpha +1} S(k,t)$, we can see the scaling function $s(y)$, where $y = k t^{1/z}$. We find that $s(y)\propto y^{2(\alpha-\alpha_s)}$ for $y\gg 1$ while $s(y) \propto y^{2\alpha +1}$ for $y\ll 1$.
%This works very well across wave numbers not too close to the ultraviolet (discretization) cutoff.
Similar results ---supporting the existence of faceted scaling close to that found for the KPZ equation with columnar noise \cite{szendro}--- in the synchronization of a 1D system of oscillators with non-odd coupling are obtained for other values of $\delta \in (0,\pi/2)$, which are provided through $G_{\phi}(\ell,t)$ and $S_\phi(k,t)$ in Appendix \ref{AppB}, as well as for other coupling strengths $K>K_c$, such as $K = 3, 10$, or $20$ (not shown). One last point worth mentioning is that the study of localization effects in the multiplicative diffusion equation and directed polymers in random media, related with the KPZ equation with columnar noise, indicates that the precise time behavior of the correlation length is sub-ballistic, with $\xi(t) \sim t \ln^{-3/4} t$ \cite{nattermann, krughh}. The value we are considering here for the dynamical exponent $z$ is thus an effective one which in general is not universal (it may depend, for instance, on the noise distribution \cite{krughh}), as it facilitates the comparison with results obtained for $\delta = 0$. Moreover, this approach was shown to provide a satisfactory description in the numerical study of the KPZ equation with columnar noise, where it was compared with the sub-ballistic form just quoted \cite{szendro}.
%This dynamics gives rise to a form of anomalous scaling first considered in Ref.~\cite{ramasco}. In common with the the super-rough scaling discussed in the previous section, it is characterized by a spectral roughness exponent $\alpha_s > 1$ and a local roughness exponent $\alpha_\text{loc} = 1$. But there is an important difference: the faceted scaling does not satisfy the FV Ansatz, Eqs.~(\ref{fv}) and (\ref{SkFV}), as the roughness exponent and the spectral roughness exponent differ, $\alpha_s \neq \alpha$. One additional exponent ($\alpha_s$) is thus truly needed for describing this type of scaling, which has been observed in interface models such as the Sneppen model (rule A) of self-organized depinning \cite{sneppen,ramasco} and the anharmonic Larkin model \cite{purrello}.
%The results presented in the following Sec.\ are based on the numerical integration of Eqs.~(\ref{eqsim}) for $\delta = \pi/4$. They show a very good qualitative agreement with those of Ref.~\cite{szendro}, obtained via a discretization of the KPZ equation with columnar noise.
%, which are at the basis of the conclusions drawn above for such model.
%The same can be said about the numerical \rod{simulations of Eqs.~(\ref{eqsim})} for $\delta = \pi/8$ and $3\pi/8$ shown in Appendix \ref{AppB}.
\subsection{Dynamics of synchronization}
It is interesting to see how the faceted scaling behavior just assessed reflects into the time-dependent process of synchronization. Thus, in the upper panels of Fig.~\ref{FigTrajdelta025K4} we show the phases $\phi_i$, and in the lower panels we show instantaneous frequencies $d \phi_i/ dt$, from representative trajectories for $\delta = \pi/4$ (left column) and $\delta = -\pi/4$ (right column). Files S2 and S3 at Ref.\ \cite{suppl} are animated movies for the corresponding time evolutions. In Fig.\ \ref{FigTrajdelta025K4}, different colors correspond to different time values, darker colors indicating earlier times and lighter colors later times. The oscillators rapidly form clusters that evolve at the same frequency, which is negative for $\delta = \pi/4$ and positive for $\delta = -\pi/4$, as expected from the previous discussions, see Sec.\ \ref{rel} in particular. These clusters merge into large ones progressively, the faster evolving one at the interface between two clusters always absorbing the slower one. Eventually, the fastest evolving cluster of oscillators recruits more and more oscillators, some of which have been previously absorbed multiple times into progressively faster clusters, until it spans the whole system, which moves uniformly.
\begin{figure}[h!]
\hspace{-0.2cm}\includegraphics[scale=0.35]{FigTrajdelta025K4.png}
\vspace{-0.7cm}
\caption{{\sf \bf Phases $\phi_i$ and instantaneous frequencies $d\phi_i/d t$ as functions of time in a representative trajectory of the oscillator lattice defined by Eq.\ (\ref{eqsim}), with size $L = 1024$ and coupling strength $K= 4$, for $\delta = \pi/4$ (left column) and $\delta = -\pi/4$ (right column).} ({\it Upper row}) Phases $\phi_i$ for $t = 200,700,1200, 1700$, and $2200$. Earlier times are shown in darker colors, later times in lighter colors. ({\it Lower row}) Instantaneous frequencies $d{\phi}_i/dt$ for the same trajectories, at the same time points.}
\label{FigTrajdelta025K4}
\end{figure}
There is an obvious similarity with the phenomenology reported in Ref.~\cite{moroney}, for a model which (except for the addition of a uniform frequency) corresponds to $\delta = -\pi/4$, and, less evidently, to that of the KPZ equation with columnar noise \cite{szendro}. In both references the authors resort to the Cole-Hopf mapping into a diffusion problem with random traps and sources for an explanation. In the present context of systems of oscillators (also that of Ref.~\cite{moroney}), further light can be shed by means of the simple two-oscillator problem, Eq.\ (\ref{twoosc}), and its stable equilibrium point, Eq.\ (\ref{twoosccond}). Indeed, according to that model when two oscillators synchronize for $\delta \neq 0$, the effective frequency is changed by a factor proprotional to $-K \sin \delta$ [see Eq.~(\ref{omegaeff2osc})], which is of course negative for $\delta = \pi/4$ and positive $\delta = -\pi/4$. Having undergone some previous coarsening, when two oscillators at the boundary of two clusters interact, their intrinsic frequencies must be ``renormalized'' (in some sense) in order to take into account the effect of the rest of the synchronized oscillators in the cluster. The result at different stages of the coarsening process, however, is always similar, in the sense that it leads to a change in the effective frequencies of the clusters under consideration that consistently pushes towards smaller/larger values, depending on whether $\sin \delta$ is positive or negative. Indeed, the space average of Eq.\ (\ref{eqsim}) yields
\begin{align}
\overline{\frac{d \phi_i}{d t}} &= \overline{\omega_i }+ K[ \overline{\sin(\phi_{i+1}\!-\!\phi_i\!+\!\delta)\!+\!\sin(\phi_i\!-\!\phi_{i+1}\!+\!\delta)} - 2 \sin \delta]\nonumber \\
&= \overline{\omega_i } - 2 K \sin \delta \left(1-\overline{\cos \Delta \phi} \right).
\label{avfreq}
\end{align}
where the average instantaneous frequency vanishes, $\overline{\omega_i} \approx \langle \omega \rangle = 0$, by the law of large numbers. Thus, $\overline{d \phi_i/d t} \approx - 2 K \sin \delta \left(1-\overline{\cos \Delta \phi} \right).$ Unless the phases are all equal, the average instantaneous frequency will be non-zero and proportional to $-K \sin \delta$. In fact, the larger the slopes $\Delta \phi$ (i.e., the smaller $\overline{\cos \Delta \phi}$ is), the larger the effective frequency is. Since the spatial averaging may be restricted to a cluster of oscillators, this may explain the fact that the steepest slopes are associated with the faster (in absolute value) clusters in Fig.~\ref{FigTrajdelta025K4}. The relationship between this and the bimodal distributions in Fig.~\ref{figslopes}, and the behavior of the slopes in the inset of the upper panel in Fig.~\ref{figWGdelta0250K4}, is further discussed in Appendix \ref{AppA}.
From the point of view of the continuum approximation, the crucial difference between the synchronization observed for the Kuramoto model ($\delta = 0$) and for $\delta \neq 0$ [$\delta \in (-\pi/2,\pi/2)$], %apart from the scaling of the critical coupling strength with $L$ ($K_c\propto \sqrt{L})$ discussed in the synchronization literature (see, e.g., \cite{strogatz, ostborn}), can thus be related to
is the existence for the latter of a KPZ nonlinearity in Eq.~(\ref{eq4}), whose coupling precisely equals $\lambda=-2Ka^2\sin\delta$, according to Eq.~(\ref{gammaders}). Such a term is well known to introduce an excess velocity that drives the interface dynamics \cite{barabasi,krug97}, which is no longer up-down symmetric. In analogy with Eq.\ \eqref{avfreq}, if one performs the space average of Eq.~(\ref{eq4}), due to the periodic boundary conditions and to the fact that the mean intrinsinc frequency is zero, the excess velocity is the integral of the KPZ nonlinearity, and its sign is therefore that of $\lambda$, indeed proportional to $-\sin \delta$.
\subsection{Fluctuation statistics}
To complete the description of the kinetic roughening universality class obtained for $\delta\neq 0$, Fig.~\ref{FigDistalldeltas} shows the distribution of the fluctuations $\varphi_i$, defined as in Eq.~(\ref{fluct}), for $\delta = \pi/4$ (again, $K = 4$ and $L=1024$), as well as for other non-zero values of $\delta$ to be discussed below. The initial time $t_0 = 50$ and several values of $\Delta t$ from $50$ to $300$ have been chosen for $\delta = \pi/4$. One can check in the inset of Fig.~\ref{figWGdelta0250K4}, upper panel, that this is a relatively large window within the growth regime. The same distribution is shown in Fig.~\ref{FigDistalldeltas} in linear (upper panel) and in logarithmic scales (lower panel).
%\rod{No entiendo esta frase:} As the phases evolve toward negative values, their absolute values have been used in the histogram (more about this later). \ric{Me refiero a que realmente si no hago nada a las fluctuaciones de las fases, me sale TW GOE pero con la skewness con signo opuesto. Entonces en este caso estoy tomando las fases en valor absoluto... de forma que el crecimiento vaya hacia valores positivos en lugar de negativos (que es como va, basta mirar Fig.~\ref{FigTrajdelta025K4} para $\delta = \pi/4$)}
The histogram %for $\delta = \pi/4$ (orange stars),
has been rescaled and shifted so that the sample mean and the sample standard deviation coincide with the mean and the standard deviation of the TW distribution for the maximum eigenvalue of random matrices in the Gaussian orthogonal ensemble (GOE) \cite{tracy09}. This distribution, which is shown as a black dotted line in the figure, is the one found for the growth dynamics of the 1D KPZ universality class when using, e.g., periodic boundary conditions \cite{kriecherbauer10,takeuchi}, and is clearly very different from the Gaussian fluctuations obtained for $\delta = 0$. Very good agreement is thus found between the numerical results for the oscillator lattice and the GOE-TW PDF. While points in the right tail appear to deviate for probabilities $\approx 10^{-3}$, which might be related to finite size effects, the left tail continues to follow very closely the theoretical curve for the lowest values inspected, on the order of $10^{-5}$, below which large statistical uncertainties exist (not shown). To our knowledge, this is the first time that an explicit connection between synchronization and the ubiquitous TW family of distributions is found. Even their relevance in the growth process given by the KPZ equation with columnar noise is not yet established, though the strong phenomenological links between that model and the synchronization model under investigation suggest that they might also characterize the fluctuations in that context.
The occurrence of the TW distribution is robust with respect to changes in $\delta$ and $K$. Indeed, when changing the sign of $\delta$,
we have found (not shown) a GOE-TW with the opposite sign for the skewness which agrees with that of $\lambda\propto-\sin\delta$, as is also the case in the standard KPZ universality class \cite{kriecherbauer10,takeuchi}. Note that, for representation purposes, the PDF curves shown
for $\delta>0$ in Fig.\ \ref{FigDistalldeltas} have been normalized to have positive skewness.
%\ric{Es así, pero en realidad el convenio adoptado es el contrario: para $\delta = -\pi/4$ no hace falta hacer nada, porque según se ve en Fig.~\ref{FigTrajdelta025K4} las fases crecen en sentido positivo y sale la TW con la skewness correcta, y es para $\delta =\pi/4$ donde hemos cambiado el signo de las fases para tener la skewness como en la gráfica de la TW empleada.}
%If instead of $\delta = \pi/4$ we focus on $\delta = -\pi/4$, we find the same distribution (as far as the eye can see), only we do not have to apply a change of sign in the phases in order to compute the fluctuations following Eq.~(\ref{fluct}), because the phases in that case increase with time. In fact, if the sign of the phases used in the computations of the fluctuations $\varphi_i$ for $\delta = \pi/4$ is not changed, we obtain an equivalent TW PDF with the opposite skewness.
More importantly, the results reported in Fig.~\ref{FigDistalldeltas} for $\delta = \pi/8$ and $3\pi/8$ (two values of $\delta$ for which results analogous to those of Figs.~\ref{figWGdelta0250K4} and \ref{figSkdelta0250K4} for $\delta = \pi/4$ are displayed in Appendix \ref{AppB}), confirm that practically the same distribution is found for other nonzero values of $\delta \in (-\pi/2,\pi/2)$.
%For $\delta = \pi/8$ the results are even better than for $\delta = \pi/4$, with an excellent agreement in both tails down to probabilites $\sim 10^{-5}$, while the small deviations in the right tail increase for $\delta = 3\pi/8$ (for which the left tail still shows an excellent agreement for probabilities down to $\sim 10^{-5}$). Even in the latter case, the distribution is clearly non-Gaussian, and certainly bears a very conspicuous resemblance to the TW distribution with Dyson index $1$ across an interval with a probability that is close to $1$. A small deviation at the peak follows the same pattern: while practically non-existent for $\delta = \pi/8$, becomes visible for $\delta = \pi/4$ and grows for $\delta = 3\pi/8$.
As for the effect of the coupling strength $K$, we also find the same kind of fluctuation PDF, which nicely follows the GOE-TW form, for smaller $K$ and for larger $K$, including values as large as $K=20$. Moreover, for all these different values of $\delta$ and $K$ the choice of the time window (as determined by $t_0$ and $\Delta t$) does not seem to be particularly important (as long as it is kept within the growth time interval where $W_\phi(t) \sim t^\beta$), nor do small modifications in the value of $\beta$ affect it. Taken together, these results suggest that the TW fluctuations are a generic and robust feature of the oscillator lattice, Eq.\ (\ref{eqsim}), for $\delta = (-\pi/2,\pi/2)$, provided that $\delta \neq 0$, at least for sufficiently large coupling strengths $K>K_c$.
The TW fluctuation PDF is currently considered as one of the universal traits induced by the KPZ nonlineatity \cite{kriecherbauer10,takeuchi}, although examples are known of systems which, while displaying it, do not have e.g.\ the same scaling exponent values of the KPZ universality class, see e.g.\ Ref.\ \cite{marcos22} and references therein. For systems within the KPZ universality class proper, it is accompanied by a covariance that scales as Eq.~(\ref{covscal}), where the scaling function $\mathcal{C}$ is the covariance of the Airy$_1$ process \cite{bornemann} if the boundary conditions are periodic as in our present work \cite{alves,oliveira,takeuchi}. This function is precisely the one employed in the data collapse of Fig.~\ref{FigCovariances}, as explained at the end of the previous section, and is shown there as a black dotted line. But despite the TW form of the fluctuation distribution, the scaling function $\mathcal{C}(x)$ of the phase covariance $C_\phi(\ell,t)$ for $\delta = \pi/8, \pi/4$, and $3\pi/8$ seems to be identical to that observed for the Kuramoto coupling ($\delta = 0$), for which the fluctuations are Gaussian. Actually, the numerically-obtained $C_\phi(\ell,t)$ for all the values of $\delta$ that have been inspected follow quite closely the covariance of the Larkin model in Eq.~(\ref{eq:cLark}), as can be observed in Fig.~\ref{FigCovariances}, see black solid line. At this point, we note that the covariance of the EW equation with time-dependent noise is also Airy$_1$ \cite{carrasco}, as for the nonlinear KPZ equation, in spite of the fact that EW features Gaussian fluctuations.
Taken together, the results reported in this section, in combination with those of Appendix \ref{AppB}, suggest that a faceted anomalous scaling and KPZ (TW) fluctuations are generic properties of synchronization for values of $\delta \neq 0$ in $(-\pi/2,\pi/2)$. The same can be said about the general appearance of the trajectories, which are for $\delta = \pi/8$ and $3\pi/8$ (not shown) qualitatively very similar to those displayed for $\delta = \pi/4$ in Fig.~\ref{FigTrajdelta025K4} and to those of the KPZ equation with columnar noise \cite{szendro}. The super-rough scaling and Gaussian fluctuations found for Kuramoto coupling ($\delta =0$) are thus rather peculiar, and likely due to the up-down symmetry (and the resulting absence of the KPZ nonlinearity in the effective continuum description) that only holds for that parameter choice. Intriguingly, this sharp distinction is not present in the phase covariance $C_\phi(\ell,t)$, which follows the analytical expression of the Larkin model in all the cases that have been investigated.
\section{Discussion and conclusions}
Our results demonstrate a very strong connection between synchronization in lattices of oscillators and kinetic roughening in systems with columnar noise. Both the dynamics of synchronizing oscillators and that of growing interfaces display GSI: provided the system is in the synchronized phase, its critical behavior does not depend on the values of the parameters $K$ and $\delta$ as long as they are generic. For $\delta=0$ the universality class differs from that of $\delta\neq 0$, but does not depend on $K$ either, provided synchronization takes place. The scaling behavior we find in all cases is anomalous, with forms which are here studied (to the best of our knowledge) for the first time in a synchronization context. While the scaling of synchronization with Kuramoto coupling $\Gamma(\Delta \phi) = \sin(\Delta \phi)$ ($\delta = 0$) is super-rough \cite{dassarma}, for other forms of Kuramoto-Sakaguchi coupling ($\delta \neq 0$) that also satisfy $\Gamma(0) = 0$ (but without the odd symmetry $\Gamma(\Delta\phi)+\Gamma(-\Delta\phi) \neq 0$), the scaling is generically faceted \cite{ramasco}, requiring one additional exponent for its description. Moreover the fluctuations, which in the former case are simply Gaussian, in the latter cases follow a TW distribution, an important characteristic of the KPZ universality class, known to be displayed by an increasing variety of low-dimensional strongly-correlated classical and quantum systems, see e.g. Ref.~\cite{makey20}. From this point of view, the behavior observed for the Kuramoto coupling ($\delta = 0$) is a singular exception to what seems to be the general rule for $0<|\delta| <\pi/2$, as has been sometimes discussed in the synchronization literature, see Ref.~\cite{ostborn} and references therein. This is most likely related to the same up-down symmetry (and corresponding lack of excess velocity) that distinguishes the EW equation from the KPZ equation in kinetic roughening systems with time-dependent noise \cite{barabasi,krug97}.
The most conspicuous phenomenological aspects of such faceted scaling of synchronization, as illustrated by the trajectories displayed in Fig.~\ref{FigTrajdelta025K4}, which present obvious similarities with the growth dynamics of the KPZ equation with columnar noise \cite{szendro}, were already apparent in a model of phase oscillators that corresponds to ours for $\delta = -\pi/4$ except for the constant term proportional to the $\sin \delta$ in Eq.\ (\ref{eqsim}) that guarantees that $\Gamma(0) = 0$ \cite{moroney}. It is remarkable that such a model arises from the effective description of a system of driven-dissipative bosons, which raises the possibility of experimentally observing faceted anomalous scaling in such systems. Kinetic-roughening physics is proving relevant in the study of quantum condensed-matter systems in surprising ways. For instance, the compact KPZ equation, related with oscillator systems, has been recently shown to describe the dynamics of the phase of a driven-dissipative condensate \cite{sieberer} and indeed, KPZ scaling has been experimentally reported in a 1D polariton condensate \cite{fontaine} very recently. Incidentally, and also very recently, superdiffusive KPZ relaxation has been also been found experimentally for a one-dimensional quantum spin chain \cite{wei22}.
This connection between synchronization and kinetic roughening becomes particularly natural when the local slopes (phase differences) are not too large, allowing for a continuum effective description that relies on a perturbative expansion. This approximation is expected to work for high enough $K>K_c$ if $\cos \delta > 0$. The resulting continuum equations in 1D are those of the Larkin Model (EW equation with columnar noise) for Kuramoto coupling, and the KPZ equation with columnar noise generically for the other types of couplings investigated (as our numerical results show). Presumably the effective description is valid in higher dimension, which might explain the obvious similarities between the phenomenology of 2D synchronization in the particular instance of the model discussed above \cite{moroney} and the 2D KPZ equation with columnar noise \cite{szendro}. In this regard, note that due to the excellent agreement between the Kuramoto model and the analytical predictions of the Larkin model, the conclusion that the former is in the univesality class of the EW equation with columnar disorder can hardly be questioned. For $\delta\neq0$, the type of scaling Ansatz and the scaling exponent values, as well as the sign of the skewness of the fluctuation PDF with respect to that of $\delta$, all strongly suggest the occurrence of the universality class of the KPZ equation with columnar disorder. However, slight numerical disagreements persist in the numerical values we obtain for the exponents for some $\delta$ and those reported from simulations of the KPZ equation with this type of disorder \cite{szendro}. Such disagreements may be expected, as the latter model is known to host (weak) non-universal behavior, like logarithmic corrections to the power-law growth of the correlation length and a potential dependence of scaling exponent values on the disorder distribution \cite{krughh}; from the point of view of the renormalization group (RG), the fixed point for the KPZ equation with columnar disorder is unstable with respect to infinitely many nonlinear terms generated under the RG \cite{nattermann}, possibly inducing such type of non-universal behavior.
Within the continuum approximation, the basic difference between the Kuramoto sine ($\delta = 0$) and the other forms of coupling ($\delta \neq 0$) lies in the KPZ nonlinearity, absent for the Kuramoto coupling due to its odd symmetry. Intriguingly, there must be a connection between this fact and the scaling of the critical coupling strength $K_c$ with $L$, hence the (non)existence of synchronous motion in the thermodynamic limit \cite{strogatz,ostborn}. The KPZ nonlinearity is known to cause growth along the local surface normal direction, which gives rise to a nonzero excess velocity \cite{barabasi,krug97}. Its inclusion breaks the up-down symmetry of the Larkin model, and, in the discrete lattices of oscillators under investigation, changes the (still anomalous) scaling from super-rough to faceted, and the fluctuations from Gaussian to TW distributed. Additional systems are also known in which fluctuations are TW \cite{nicoli13,marcos22} due to the occurrence of a KPZ nonlinearity \cite{nicoli,abraham}, even if scaling exponent values differ from those of the KPZ universality class. Note in this connection that recent examples are also available, like the tensionles KPZ equation, in which the KPZ nonlinearity induces a non-symmetric distribution \cite{cartes,rodriguez-fernandez22} which however does not take the precise TW form. To our knowledge, however, no kinetic-roughening system is yet known with TW-distributed fluctuations, but in which the KPZ nonlinearity is absent. Completing the description of the universality class, for our oscillator lattices we obtain that, surprisingly, the covariances follow the analytical form of the Larkin model regardless the value of $\delta$. As was mentioned above, an analogous behavior is known for time-dependent noise, whereby the covariance of the linear EW and the nonlinear KPZ equations are both Airy$_1$ \cite{carrasco}. The precise reason remains unclear, although it may be related with the fluctuation-dissipation theorem accidentally satisfied by the KPZ equation in one dimension \cite{barabasi,krug97}, which makes it share a number of properties with its (EW) linear approximation.
Related with the dynamic scaling Ansatz, recent work has shown that the Larkin model also loses its super-rough scaling in favor of faceted scaling in the presence of anharmonicities \cite{purrello}. In the synchronization context, our work raises the question whether the singularity of the Kuramoto coupling as an instance of the Kuramoto-Sakaguchi model under study (i.e.\! an isolated point displaying super-rough scaling in a parameter space where faceted scaling seems to be generic) extends to other kinds of symmetry-breaking perturbations. In the kinetic roughening context, on the other hand, it raises the question whether KPZ fluctuations are also present in those other models displaying faceted scaling, which then should be added to the currently increasing list of systems known to display such fluctuations. Moreover, it would be interesting to know whether those models also display covariances like those of the linear theory (which is the Larkin model in our case), or they follow those of the Airy$_1$ process, as happens in the KPZ universality class.
Overall, our work opens up new vistas on the problem of synchronization in finite dimensions by making use of the powerful conceptual framework of nonequilibrium kinetic roughening to an unprecedented level, incorporating also quite recent developments in that field. This allows us to clarify the dynamical process by which synchronous motion is achieved, which remained a relatively poorly explored aspect of synchronization, and is here shown to be characterized by forms of nonequilibrium criticality previously observed in growth processess. Many important issues remain to be studied at the interface between these two topics of contemporary statistical physics and nonlinear science. For example, the role of thermal noise in the effective dynamics of synchronized motion, as in fact variations of the Kuramoto model and similar models of phase oscillators are frequently studied in the presence of time-dependent noise sources \cite{acebron}. The connection to the (standard, thermal noise) KPZ equation in such models has already been discussed and recently exploited in the study of routes out of synchronization \cite{lauter}, but every aspect of GSI in that context, including the nature of scaling and its fluctuations, apparently remains unknown. Finally, while we have considered synchronous dynamics for coupling strengths $K>K_c$, which guarantee that synchronization is achieved for long times (much as saturation is reached in finite-size kinetic roughening), an obviously relevant goal would be to understand the transition to synchronization at the level of the effective continuum description. Ideally, this should provide accurate estimates of $K_c$, and shed new light on the role played by the symmetry of the coupling function \cite{ostborn,strogatz}.
\begin{acknowledgments}
This work has been partially supported by Ministerio de Ciencia, Innovaci\'on y Universidades (Spain), Agencia Estatal de Investigaci\'on (AEI, Spain), and Fondo Europeo de Desarrollo Regional (FEDER, EU) through Grants No.\ PGC2018-094763-B-I00 and No.\ and PID2021-123969NB-I00, and by Comunidad de Madrid (Spain) under the Multiannual Agreement with UC3M in the line of Excellence of University Professors (EPUC3M23), in the context of the V Plan Regional de Investigaci\'on Cient\'{\i}fica e Innovaci\'on Tecnol\'ogica (PRICIT).
\end{acknowledgments}
\appendix
%\renewcommand{\thefigure}{A\arabic{figure}}
%\setcounter{figure}{0}
\section{Time-dependent distribution of slopes $\Delta \phi$}
\label{AppA}
\begin{figure*}[t!]
\includegraphics[scale=0.4]{FigHists.png}
\vspace{-0.6cm}
\caption{{\sf \bf PDF of the local slopes $P(\Delta \phi)$ across time in a system of $L=1024$ oscillators for different values of $\delta$, together with the square root of three observables considered in the main text, namely, the average squared slope, the average of the maximum slope and the most probable squared slope.} (a) $\delta = 0$ and $K=40$, (b) $\delta = \pi/8$ and $K=4$, (c) $\delta = \pi/4$ and $K=4$, (d) $\delta = 3\pi/8$ and $K=8$. $P(\Delta \phi)$ is shown across time using the color code specified in the color bars, together with the square root of the average squared slope (black dots), that of the mean maximum squared local slope (red asterisks), and that of the the most probable value of the squared slope (blue crosses). Histograms and averages are based on 2000 realizations.}
\label{FigHists}
\end{figure*}
While discussing the insets in the upper panels of Figs.~\ref{figWGdelta0K40} and \ref{figWGdelta0250K4} (see also Figs.~\ref{figWGdelta0125K4} and \ref{figWGdelta0375K8} in Appendix \ref{AppB}) various observables related to the time evolution of the distribution of the slopes $\Delta \phi$ were mentioned. Specifically, there we discussed the evolution of the average squared slope $\langle\overline{(\Delta \phi)^2} \rangle$, as well as the maximum squared local slope (defined as the square of the greatest slope in the system, averaged over realizations) and the most probable squared local slope (defined as the squared local slope that has a higher probability in the distribution). A more complete picture emerges by inspecting the distribution of local slopes $P(\Delta \phi)$ itself across time, which is what we show in this Appendix. In fact we have already shown some instances of $P(\Delta \phi)$ in Fig.~\ref{figslopes}, but they correspond to a single snapshot of the slope distribution for long times. Moreover, as they were used to restrict the range of parameters based on the validity of the small-slope approximation, they included values of $\delta$ that were finally not studied. Here we focus on $\delta \in [0,\pi/2)$ for the same values of $K$ chosen in the main text and in Appendix \ref{AppB} ($K\approx 2 K_c$). For $\delta \in (-\pi/2,0)$, we obtain equivalent results due to the symmetry discussed at the end of Sec.\ \ref{nummod}.
In Fig.~\ref{FigHists} we show the PDF of the slopes $P(\Delta \phi)$ for four values of $\delta$. Those are the values that were fully discussed in the main text, namely $\delta = 0$ [panel (a), which was considered in Section \ref{odd}] and $\delta = \pi/4$ [panel (c), which was considered in Section \ref{nonodd}], as well as two values, $\delta = \pi/8$ [panel (b)] and $\delta = 3\pi/8$ [panel (d)], for which only the fluctuation PDF and the covariances were studied in the main text (see Figs.~\ref{FigDistalldeltas} and \ref{FigCovariances}, respectively, and the pertinent discussion in Section \ref{nonodd}), while the scaling Ansatz and exponent values are addressed in Appendix \ref{AppB}. The PDF of the slopes $P(\Delta \phi)$ is represented as it evolves across time using the colors specified in the color bars. Additionally, we include the observables that we showed in the insets of the upper panels of Figs.~\ref{figWGdelta0K40} and \ref{figWGdelta0250K4} (and also Figs.~\ref{figWGdelta0125K4} and \ref{figWGdelta0375K8} in Appendix \ref{AppB}), but, as they are dimensionally squared slopes, we take their square root to make the comparison with $P(\Delta \phi)$ possible. Thus, we show the square root of the average squared slope (black dots), that of the mean maximum square local slope (red asterisks) and that of the the most probable value of the squared slope (blue crosses).
%\ric{Aquí está la posibilidad de usar promedios directamente basados en pendientes y no en pendientes al cuadrado. El problema es que si bien el máximo y el valor más probable se pueden obtener sin problema, el equivalente del promedio de la pendiente cuadrática, por simetría, tendría que ser el promedio del valor absoluto... que añade más observables no totalmente directos. Y para dejar ese con la raíz cuadrada y cambiar solo los otros, he considerado mejor tomar la raiz de los observables considerados en el texto principal y facilitar la comparación, aunque no sea ideal. Otra opción sería mostrar la distribución de pendientes al cuadrado, y así no hay que andar con raíces cuadradas en estos observables, pero me parece peor (por la información indirecta que ofrece) la solución que el problema... Un asunto adicional es que no he encontrado una notación buena para dos de los tres observables (el máximo y el más probable)... Quizás sería mejor pensar una notación y ponerlo todo de una sola vez en la subsección sobre observables del análisis morfológico, aunque habría que ver cómo incorporarlo a los insets de las figuras del texto principal, que de momento solo muestran la notación de $\langle\overline{(\Delta \phi)^2} \rangle$.}
For $\delta = 0$, panel (a), $P(\Delta \phi)$ is unimodal, and the most probable value is always close to $\Delta \phi =0$. The distribution simply spreads over a larger interval of $\Delta \phi$ as time increases, until it saturates. For the remaining values of $\delta$, panels (b), (c), and (d), the distribution starts being unimodal close to the flat initial condition, but at some point it becomes bimodal, and the most probable value becomes distinctly different from zero. This seems to be something that Eq.~(\ref{avfreq}) can explain, as it shows that for a typical cluster of oscillators (or a cluster of oscillators which is large enough) such that the average of the involved intrinsic frequencies approaches zero, the absolute values of the average instantaneous frequencies of the cluster approach $2 K \sin \delta \left(1-\overline{\cos \Delta \phi} \right)$, where $\overline{\cos \Delta \phi}$ is the spatial average of $\cos \Delta \phi$ over the cluster. Therefore a faster cluster requires larger slopes (smaller values of $\cos \Delta \phi$), and when the system starts being dominated by a few fast clusters (see Fig.~\ref{FigTrajdelta025K4}) there is a reshuffling of the slopes towards larger (absolute) values. These three cases, $\delta = \pi/8$, $\pi/4$, and $3\pi/8$, correspond to an effective description given by the KPZ equation with columnar noise, with faceted anomalous scaling and KPZ fluctuations (see Section \ref{nonodd} and Appendix \ref{AppB}), so their qualitative similarity is expected. The most conspicuous effect of modifying $\delta(\neq 0)$ is the following: as $\delta$ increases, the spread of the distribution $P(\Delta \phi)$ decreases and the saturation of the largest values occurs earlier.
\section{Scaling Ansatz and exponents for other values of $\delta$}
\label{AppB}
In the following we show results analogous to those displayed in Section \ref{nonodd} for $\delta = \pi/4$, specifically those reported in Figs.~\ref{figWGdelta0250K4} and \ref{figSkdelta0250K4}, but for other values of $\delta \in (0,\pi/2)$. Here, we additionally address $\delta = \pi/8$ and $3\pi/8$, which were considered in the analysis of the fluctuations and the covariances, see Figs.~\ref{FigDistalldeltas} and \ref{FigCovariances}. As all the elements in the figures were previously discussed for $\delta = \pi/4$ and the qualitative conclusions are the same, the descriptions will be brief.
\begin{figure}[h!]
\includegraphics[scale=0.47]{FigWGdelta0125K4.png}
\vspace{-0.3cm}
\caption{{\sf \bf Roughness $W_\phi(t)$ and height-difference correlation function $G_\phi(\ell,t)$ for the oscillator lattice, Eq.\ (\ref{eqsim}), with $\delta = \pi/8$, $L = 1024$, and $K= 4$.}
({\it Upper panel}) $W_\phi(t)$ as a function of time in the synchronous regime, with error bars displaying the standard error of the average across realizations. In the power-law visual guides we have used $\alpha = 1.12$ and $z=1.42$. The inset shows the average squared slope $\langle\overline{(\Delta \phi)^2} \rangle$ (black circles), as well as the average of the maximum squared local slope (red asterisks), and the most probable value of the local slope (blue crosses), as a function of time for the same parameter values (see text for definitions). The black dashed line shows the power-law growth of the average squared slopes predicted by the theory, $t^{2(\alpha-\alpha_\text{loc})/z}$ for $\alpha_\text{loc}=0.95$. ({\it Lower panel}) $G(\ell,t)$ as a function of $\ell$ in the synchronous regime for different times (see legends). The dashed line shows the power law growth with the distance $l^{2\alpha_\text{loc}}$. Inset: rescaling of $G(\ell,t)$ following the theoretical form in Eq.~(\ref{glt_mod}). Averages based on 2000 realizations.
}\label{figWGdelta0125K4}
\end{figure}
In Fig.~\ref{figWGdelta0125K4} we show the roughness $W_\phi(t)$ (upper panel) and height-difference correlation function $G_\phi(\ell,t)$ (lower panel) for $\delta = \pi/8$ in a system of $L=1024$ oscillators. The critical coupling $K=4$ is roughly twice as large as the critical coupling strength obtained from the saturation of the width as described at the end of Sec.\ \ref{rel}. The exponents values are $\alpha = 1.12$, $\alpha_s = 1.40$, $z=1.42$, and $\alpha_\text{loc}=0.95$, with uncertainties again below $0.05$ for the same reasons discussed for $\delta = \pi/4$ in the main text, which differs slightly in the exponent values. The inset of the upper panel shows again the mean, maximum, and most probable value of the squared slopes averaged across realizations, and that of the lower panel shows a rescaling of the two-point correlations based on the theory of anomalous scaling, Eq.\ (\ref{glt_mod}). All the comments given for $\delta = \pi/4$ apply in this case too.
\begin{figure}[h!]
\includegraphics[scale=0.47]{FigSkdelta0125K4.png}
\vspace{-0.3cm}
\caption{{\sf \bf Structure factor $S_\phi(k,t)$ as a function of time for the oscillator lattice, Eq.\ (\ref{eqsim}), with $\delta = \pi/8$, $L = 1024$, and $K= 4$.} ({\it Upper panel}) $S_\phi(k,t)$ as a function of $k$. ({\it Lower panel}) $k ^{2\alpha+1} S_\phi(k,t)$ as a function of $y = k t^{1/z}$ In all rescalings and power-law visual guides, $\alpha = 1.12$. $\alpha_s = 1.40$, $z=1.42$, and $\alpha_\text{loc}=0.95$. The structure factor curves are shown for the same time points displayed in the lower panel of Fig.~\ref{figWGdelta0125K4} (see the corresponding legend). Averages based on 2000 realizations.}
\label{figSkdelta0125K4}
\end{figure}
The structure factor $S_\phi(k,t)$ also for $\delta = \pi/8$ (with the same parameter choices and exponents) is shown in its original form (upper panel) and in rescaled form (lower panel) in Fig.~\ref{figSkdelta0125K4}. As in the case of $\delta = \pi/4$, the scaling is anomalous of the faceted type discussed in Ref.~\cite{ramasco}.
\begin{figure}[h!]
\includegraphics[scale=0.47]{FigWGdelta0375K8.png}
\vspace{-0.3cm}
\caption{{\sf \bf Roughness $W_\phi(t)$ and height-difference correlation function $G_\phi(\ell,t)$ for the oscillator lattice, Eq.\ (\ref{eqsim}), with $\delta = 3\pi/8$, $L = 1024$, and coupling strength $K= 8$.}
({\it Upper panel}) $W_\phi(t)$ as a function of time in the synchronous regime, with error bars displaying the standard error of the average across realizations. In the power-law visual guides we have used $\alpha = 1.03$ and $z=1.36$. The inset shows the average squared slope $\langle\overline{(\Delta \phi)^2} \rangle$ (black circles), as well as the average of the maximum squared local slope (red asterisks), and the most probable value of the local slope (blue crosses), as a function of time for the same parameter values (see text for definitions). The black dashed line shows the power-law growth of the average squared slopes predicted by the theory, $t^{2(\alpha-\alpha_\text{loc})/z}$ for $\alpha_\text{loc}=0.95$. ({\it Lower panel}) $G(\ell,t)$ as a function of $\ell$ in the synchronous regime for different times (see legends). The dashed line shows the power law growth with the distance $l^{2\alpha_\text{loc}}$. Inset: rescaling of $G(\ell,t)$ following the theoretical form in Eq.~(\ref{glt}). Averages based on 2000 realizations.
}\label{figWGdelta0375K8}
\end{figure}
We next focus on analogous results for $\delta = 3\pi/8$. In Fig.~\ref{figWGdelta0375K8} we show the roughness $W_\phi(t)$ (upper panel) and height-difference correlation function $G_\phi(\ell,t)$ (lower panel) in a system of $L=1024$ oscillators. The critical coupling $K=8$ is roughly twice as large as the critical coupling strength obtained from the saturation of the width as described at the end of Sec.\ \ref{rel}. The exponents of choice are $\alpha = 1.03$, $\alpha_s = 1.40$, $z=1.36$, and $\alpha_\text{loc}=0.95$, with uncertainties again below $0.05$, for the same reasons discussed for $\delta = \pi/4$ in the main text, with slightly different exponents. The inset of the upper panel shows again the mean, maximum, and most probable value of the squared slopes averaged across realizations, and that of the lower panel shows a rescaling of the two-point correlations based on the theory of anomalous scaling, Eq.\ (\ref{glt_mod}). All the comments given for $\delta = \pi/4$ apply in this case too. The most remarkable difference is perhaps the abruptness in the change of the most probable value of the squared slope (upper panel inset, blue stars), which was also shown in Fig.~\ref{FigHists} in Appendix \ref{AppA}.
\begin{figure}[h!]
\includegraphics[scale=0.47]{FigSkdelta0375K8.png}
\vspace{-0.3cm}
\caption{{\sf \bf Structure factor $S_\phi(k,t)$ as a function of time for the oscillator lattice, Eq.\ (\ref{eqsim}), with $\delta = 3\pi/8$, $L = 1024$, and $K= 8$.}
({\it Upper panel}) $S_\phi(k,t)$ as a function of $k$. ({\it Lower panel}) $k ^{2\alpha+1} S_\phi(k,t)$ as a function of $y = k t^{1/z}$ In all rescalings and power-law visual guides, $\alpha = 1.03$, $\alpha_s = 1.40$, $z=1.36$, and $\alpha_\text{loc}=0.95$. The structure factor curves are shown for the same time points displayed in the lower panel of Fig.~\ref{figWGdelta0375K8} (see the corresponding legend). Averages based on 2000 realizations.}
\label{figSkdelta0375K8}
\end{figure}
The structure factor $S_\phi(k,t)$ for $\delta = 3\pi/8$ (with the same parameter choices and exponents) is shown in its original form (upper panel) and in rescaled form (lower panel) in Fig.~\ref{figSkdelta0375K8}. As in the cases of $\delta = \pi/8$ and $\delta = \pi/4$, the scaling is anomalous of the faceted type discussed in Ref.~\cite{ramasco}.
For $\delta =\pi/8$ and $3\pi/8$, we have also computed the PDF of rescaled phase fluctuations, $\varphi_i$, and the corresponding covariances $C_\phi(\ell,t)$. They are shown in Figs.~\ref{FigDistalldeltas} and \ref{FigCovariances}, respectively, and are discussed in Section \ref{nonodd} of the main text.
\clearpage
\bibliography{syncgrowthBib}{}
\end{document}
%merlin.mbs apsrev4-1.bst 2010-07-25 4.21a (PWD, AO, DPC) hacked
%Control: key (0)
%Control: author (0) dotless jnrlst
%Control: editor formatted (1) identically to author
%Control: production of article title (0) allowed
%Control: page (1) range
%Control: year (0) verbatim
%Control: production of eprint (0) enabled
\begin{thebibliography}{79}%
\makeatletter
\providecommand \@ifxundefined [1]{%
\@ifx{#1\undefined}
}%
\providecommand \@ifnum [1]{%
\ifnum #1\expandafter \@firstoftwo
\else \expandafter \@secondoftwo
\fi
}%
\providecommand \@ifx [1]{%
\ifx #1\expandafter \@firstoftwo
\else \expandafter \@secondoftwo
\fi
}%
\providecommand \natexlab [1]{#1}%
\providecommand \enquote [1]{``#1''}%
\providecommand \bibnamefont [1]{#1}%
\providecommand \bibfnamefont [1]{#1}%
\providecommand \citenamefont [1]{#1}%
\providecommand \href@noop [0]{\@secondoftwo}%
\providecommand \href [0]{\begingroup \@sanitize@url \@href}%
\providecommand \@href[1]{\@@startlink{#1}\@@href}%
\providecommand \@@href[1]{\endgroup#1\@@endlink}%
\providecommand \@sanitize@url [0]{\catcode `\\12\catcode `\$12\catcode
`\&12\catcode `\#12\catcode `\^12\catcode `\_12\catcode `\%12\relax}%
\providecommand \@@startlink[1]{}%
\providecommand \@@endlink[0]{}%
\providecommand \url [0]{\begingroup\@sanitize@url \@url }%
\providecommand \@url [1]{\endgroup\@href {#1}{\urlprefix }}%
\providecommand \urlprefix [0]{URL }%
\providecommand \Eprint [0]{\href }%
\providecommand \doibase [0]{http://dx.doi.org/}%
\providecommand \selectlanguage [0]{\@gobble}%
\providecommand \bibinfo [0]{\@secondoftwo}%
\providecommand \bibfield [0]{\@secondoftwo}%
\providecommand \translation [1]{[#1]}%
\providecommand \BibitemOpen [0]{}%
\providecommand \bibitemStop [0]{}%
\providecommand \bibitemNoStop [0]{.\EOS\space}%
\providecommand \EOS [0]{\spacefactor3000\relax}%
\providecommand \BibitemShut [1]{\csname bibitem#1\endcsname}%
\let\auto@bib@innerbib\@empty
%</preamble>
\bibitem [{\citenamefont {Pikovsky}\ \emph {et~al.}(2001)\citenamefont
{Pikovsky}, \citenamefont {Rosenblum},\ and\ \citenamefont
{Kurths}}]{pikovsky}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {A.}~\bibnamefont
{Pikovsky}}, \bibinfo {author} {\bibfnamefont {M.}~\bibnamefont {Rosenblum}},
\ and\ \bibinfo {author} {\bibfnamefont {J.}~\bibnamefont {Kurths}},\ }\href
{\doibase 10.1017/CBO9780511755743} {\emph {\bibinfo {title}
{Synchronization: A Universal Concept in Nonlinear Sciences}}}\ (\bibinfo
{publisher} {Cambridge University Press},\ \bibinfo {address} {Cambridge,
UK},\ \bibinfo {year} {2001})\BibitemShut {NoStop}%
\bibitem [{\citenamefont {Walter}\ \emph {et~al.}(2014)\citenamefont {Walter},
\citenamefont {Nunnenkamp},\ and\ \citenamefont {Bruder}}]{walter}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {S.}~\bibnamefont
{Walter}}, \bibinfo {author} {\bibfnamefont {A.}~\bibnamefont {Nunnenkamp}},
\ and\ \bibinfo {author} {\bibfnamefont {C.}~\bibnamefont {Bruder}},\
}\bibfield {title} {\enquote {\bibinfo {title} {Quantum synchronization of a
driven self-sustained oscillator},}\ }\href {\doibase
10.1103/PhysRevLett.112.094102} {\bibfield {journal} {\bibinfo {journal}
{Phys. Rev. Lett.}\ }\textbf {\bibinfo {volume} {112}},\ \bibinfo {pages}
{094102} (\bibinfo {year} {2014})}\BibitemShut {NoStop}%
\bibitem [{\citenamefont {Kuramoto}(1984{\natexlab{a}})}]{kuramoto_book}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {Y.}~\bibnamefont
{Kuramoto}},\ }\href {\doibase 10.1007/978-3-642-69689-3} {\emph {\bibinfo
{title} {Chemical Oscillations, Waves, and Turbulence}}}\ (\bibinfo
{publisher} {Springer},\ \bibinfo {address} {Berlin},\ \bibinfo {year}
{1984})\BibitemShut {NoStop}%
\bibitem [{\citenamefont {Acebr\'on}\ \emph {et~al.}(2005)\citenamefont
{Acebr\'on}, \citenamefont {Bonilla}, \citenamefont {P\'erez~Vicente},
\citenamefont {Ritort},\ and\ \citenamefont {Spigler}}]{acebron}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {J.~A.}\ \bibnamefont
{Acebr\'on}}, \bibinfo {author} {\bibfnamefont {L.~L.}\ \bibnamefont
{Bonilla}}, \bibinfo {author} {\bibfnamefont {C.~J.}\ \bibnamefont
{P\'erez~Vicente}}, \bibinfo {author} {\bibfnamefont {F.}~\bibnamefont
{Ritort}}, \ and\ \bibinfo {author} {\bibfnamefont {R.}~\bibnamefont
{Spigler}},\ }\bibfield {title} {\enquote {\bibinfo {title} {The {K}uramoto
model: A simple paradigm for synchronization phenomena},}\ }\href {\doibase
10.1103/RevModPhys.77.137} {\bibfield {journal} {\bibinfo {journal} {Rev.
Mod. Phys.}\ }\textbf {\bibinfo {volume} {77}},\ \bibinfo {pages} {137}
(\bibinfo {year} {2005})}\BibitemShut {NoStop}%
\bibitem [{\citenamefont {Boccaletti}\ \emph {et~al.}(2002)\citenamefont
{Boccaletti}, \citenamefont {Kurths}, \citenamefont {Osipov}, \citenamefont
{Valladares},\ and\ \citenamefont {Zhou}}]{boccaletti}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {S.}~\bibnamefont
{Boccaletti}}, \bibinfo {author} {\bibfnamefont {J.}~\bibnamefont {Kurths}},
\bibinfo {author} {\bibfnamefont {G.}~\bibnamefont {Osipov}}, \bibinfo
{author} {\bibfnamefont {D.L.}\ \bibnamefont {Valladares}}, \ and\ \bibinfo
{author} {\bibfnamefont {C.S.}\ \bibnamefont {Zhou}},\ }\bibfield {title}
{\enquote {\bibinfo {title} {The synchronization of chaotic systems},}\
}\href {\doibase https://doi.org/10.1016/S0370-1573(02)00137-0} {\bibfield
{journal} {\bibinfo {journal} {Phys. Rep.}\ }\textbf {\bibinfo {volume}
{366}},\ \bibinfo {pages} {1} (\bibinfo {year} {2002})}\BibitemShut {NoStop}%
\bibitem [{\citenamefont {Arenas}\ \emph {et~al.}(2008)\citenamefont {Arenas},
\citenamefont {Díaz-Guilera}, \citenamefont {Kurths}, \citenamefont
{Moreno},\ and\ \citenamefont {Zhou}}]{arenas}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {A.}~\bibnamefont
{Arenas}}, \bibinfo {author} {\bibfnamefont {A.}~\bibnamefont
{Díaz-Guilera}}, \bibinfo {author} {\bibfnamefont {J.}~\bibnamefont
{Kurths}}, \bibinfo {author} {\bibfnamefont {Y.}~\bibnamefont {Moreno}}, \
and\ \bibinfo {author} {\bibfnamefont {C.}~\bibnamefont {Zhou}},\ }\bibfield
{title} {\enquote {\bibinfo {title} {Synchronization in complex networks},}\
}\href {\doibase https://doi.org/10.1016/j.physrep.2008.09.002} {\bibfield
{journal} {\bibinfo {journal} {Phys. Rep.}\ }\textbf {\bibinfo {volume}
{469}},\ \bibinfo {pages} {93} (\bibinfo {year} {2008})}\BibitemShut
{NoStop}%
\bibitem [{\citenamefont {Halpin-Healy}\ and\ \citenamefont
{Zhang}(1995)}]{halpinhealy}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {T.}~\bibnamefont
{Halpin-Healy}}\ and\ \bibinfo {author} {\bibfnamefont {Y.~C.}\ \bibnamefont
{Zhang}},\ }\bibfield {title} {\enquote {\bibinfo {title} {Kinetic
roughening phenomena, stochastic growth, directed polymers and all that.
{A}spects of multidisciplinary statistical mechanics},}\ }\href {\doibase
https://doi.org/10.1016/0370-1573(94)00087-J} {\bibfield {journal} {\bibinfo
{journal} {Phys. Rep.}\ }\textbf {\bibinfo {volume} {254}},\ \bibinfo
{pages} {215} (\bibinfo {year} {1995})}\BibitemShut {NoStop}%
\bibitem [{\citenamefont {Krug}(1997)}]{krug97}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {J.}~\bibnamefont
{Krug}},\ }\bibfield {title} {\enquote {\bibinfo {title} {{Origins of scale
invariance in growth processes}},}\ }\href {\doibase
10.1080/00018739700101498} {\bibfield {journal} {\bibinfo {journal} {Adv.
Phys.}\ }\textbf {\bibinfo {volume} {46}},\ \bibinfo {pages} {139} (\bibinfo
{year} {1997})}\BibitemShut {NoStop}%
\bibitem [{\citenamefont {Barabási}\ and\ \citenamefont
{Stanley}(1995)}]{barabasi}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {A.-L.}\ \bibnamefont
{Barabási}}\ and\ \bibinfo {author} {\bibfnamefont {H.~E.}\ \bibnamefont
{Stanley}},\ }\href {\doibase 10.1017/CBO9780511599798} {\emph {\bibinfo
{title} {Fractal Concepts in Surface Growth}}}\ (\bibinfo {publisher}
{Cambridge University Press},\ \bibinfo {address} {Cambridge, UK},\ \bibinfo
{year} {1995})\BibitemShut {NoStop}%
\bibitem [{\citenamefont {Goldenfeld}(1992)}]{goldenfeld}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {N.}~\bibnamefont
{Goldenfeld}},\ }\href {\doibase 10.1201/9780429493492} {\emph {\bibinfo
{title} {Lectures on Phase Transitions and the Renormalization Group}}}\
(\bibinfo {publisher} {CRC Press},\ \bibinfo {address} {Boca Raton},\
\bibinfo {year} {1992})\BibitemShut {NoStop}%
\bibitem [{\citenamefont {T{\"a}uber}(2014)}]{tauber14}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {U.~C.}\ \bibnamefont
{T{\"a}uber}},\ }\href@noop {} {\emph {\bibinfo {title} {Critical dynamics: a
field theory approach to equilibrium and non-equilibrium scaling behavior}}}\
(\bibinfo {publisher} {Cambridge University Press},\ \bibinfo {address}
{Cambridge, UK},\ \bibinfo {year} {2014})\BibitemShut {NoStop}%
\bibitem [{\citenamefont {Grinstein}(1991)}]{grinstein91}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {G.}~\bibnamefont
{Grinstein}},\ }\bibfield {title} {\enquote {\bibinfo {title} {{Generic
scale invariance in classical nonequilibrium systems (invited)}},}\ }\href
{\doibase 10.1063/1.348003} {\bibfield {journal} {\bibinfo {journal} {J.
Appl. Phys.}\ }\textbf {\bibinfo {volume} {69}},\ \bibinfo {pages} {5441}
(\bibinfo {year} {1991})}\BibitemShut {NoStop}%
\bibitem [{\citenamefont {Grinstein}(1995)}]{grinstein95}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {G.}~\bibnamefont
{Grinstein}},\ }\bibfield {title} {\enquote {\bibinfo {title} {Generic scale
invariance and self-organized criticality},}\ }in\ \href@noop {} {\emph
{\bibinfo {booktitle} {Scale Invariance, Interfaces and Nonequilibrium
Dynamics}}},\ \bibinfo {editor} {edited by\ \bibinfo {editor} {\bibfnamefont
{Alan}\ \bibnamefont {McKane}}, \bibinfo {editor} {\bibfnamefont {Michel}\
\bibnamefont {Droz}}, \bibinfo {editor} {\bibfnamefont {Jean}\ \bibnamefont
{Vannimenus}}, \ and\ \bibinfo {editor} {\bibfnamefont {Dietrich}\
\bibnamefont {Wolf}}}\ (\bibinfo {publisher} {Plenum},\ \bibinfo {address}
{New York},\ \bibinfo {year} {1995})\ p.\ \bibinfo {pages} {261}\BibitemShut
{NoStop}%
\bibitem [{\citenamefont {Belitz}\ \emph {et~al.}(2005)\citenamefont {Belitz},
\citenamefont {Kirkpatrick},\ and\ \citenamefont {Vojta}}]{belitz05}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {D.}~\bibnamefont
{Belitz}}, \bibinfo {author} {\bibfnamefont {T.~R.}\ \bibnamefont
{Kirkpatrick}}, \ and\ \bibinfo {author} {\bibfnamefont {T.}~\bibnamefont
{Vojta}},\ }\bibfield {title} {\enquote {\bibinfo {title} {{How generic
scale invariance influences phase transitions}},}\ }\href {\doibase
10.1103/RevModPhys.77.579} {\bibfield {journal} {\bibinfo {journal} {Rev.
Mod. Phys.}\ }\textbf {\bibinfo {volume} {77}},\ \bibinfo {pages} {579}
(\bibinfo {year} {2005})}\BibitemShut {NoStop}%
\bibitem [{\citenamefont {Dickman}\ \emph {et~al.}(2000)\citenamefont
{Dickman}, \citenamefont {Mu{\~{n}}oz}, \citenamefont {Vespignani},\ and\
\citenamefont {Zapperi}}]{dickman00}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {R.}~\bibnamefont
{Dickman}}, \bibinfo {author} {\bibfnamefont {M.~A.}\ \bibnamefont
{Mu{\~{n}}oz}}, \bibinfo {author} {\bibfnamefont {A.}~\bibnamefont
{Vespignani}}, \ and\ \bibinfo {author} {\bibfnamefont {S.}~\bibnamefont
{Zapperi}},\ }\bibfield {title} {\enquote {\bibinfo {title} {{Paths to
self-organized criticality}},}\ }\href {\doibase
10.1590/S0103-97332000000100004} {\bibfield {journal} {\bibinfo {journal}
{Braz. J. Phys.}\ }\textbf {\bibinfo {volume} {30}},\ \bibinfo {pages} {27}
(\bibinfo {year} {2000})}\BibitemShut {NoStop}%
\bibitem [{\citenamefont {Pruessner}(2012)}]{pruessner12}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {G.}~\bibnamefont
{Pruessner}},\ }\href@noop {} {\emph {\bibinfo {title} {Self-organised
criticality: theory, models and characterisation}}}\ (\bibinfo {publisher}
{Cambridge University Press},\ \bibinfo {address} {Cambridge, UK},\ \bibinfo
{year} {2012})\BibitemShut {NoStop}%
\bibitem [{\citenamefont {Kardar}\ \emph {et~al.}(1986)\citenamefont {Kardar},
\citenamefont {Parisi},\ and\ \citenamefont {Zhang}}]{kardar}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {M.}~\bibnamefont
{Kardar}}, \bibinfo {author} {\bibfnamefont {G.}~\bibnamefont {Parisi}}, \
and\ \bibinfo {author} {\bibfnamefont {Y.-C.}\ \bibnamefont {Zhang}},\
}\bibfield {title} {\enquote {\bibinfo {title} {Dynamic scaling of growing
interfaces},}\ }\href {\doibase 10.1103/PhysRevLett.56.889} {\bibfield
{journal} {\bibinfo {journal} {Phys. Rev. Lett.}\ }\textbf {\bibinfo
{volume} {56}},\ \bibinfo {pages} {889} (\bibinfo {year} {1986})}\BibitemShut
{NoStop}%
\bibitem [{\citenamefont {Kriecherbauer}\ and\ \citenamefont
{Krug}(2010)}]{kriecherbauer10}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {T.}~\bibnamefont
{Kriecherbauer}}\ and\ \bibinfo {author} {\bibfnamefont {J.}~\bibnamefont
{Krug}},\ }\bibfield {title} {\enquote {\bibinfo {title} {{A pedestrian's
view on interacting particle systems, KPZ universality and random
matrices}},}\ }\href {\doibase 10.1088/1751-8113/43/40/403001} {\bibfield
{journal} {\bibinfo {journal} {J. Phys. A: Math. Theor.}\ }\textbf {\bibinfo
{volume} {43}},\ \bibinfo {pages} {403001} (\bibinfo {year}
{2010})}\BibitemShut {NoStop}%
\bibitem [{\citenamefont {Takeuchi}(2018)}]{takeuchi}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {K.~A.}\ \bibnamefont
{Takeuchi}},\ }\bibfield {title} {\enquote {\bibinfo {title} {An appetizer
to modern developments on the {K}ardar–{P}arisi–{Z}hang universality
class},}\ }\href {\doibase https://doi.org/10.1016/j.physa.2018.03.009}
{\bibfield {journal} {\bibinfo {journal} {Physica A}\ }\textbf {\bibinfo
{volume} {504}},\ \bibinfo {pages} {77} (\bibinfo {year} {2018})}\BibitemShut
{NoStop}%
\bibitem [{\citenamefont {Chen}\ \emph {et~al.}(2016)\citenamefont {Chen},
\citenamefont {Lee},\ and\ \citenamefont {Toner}}]{chen16}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {L.}~\bibnamefont
{Chen}}, \bibinfo {author} {\bibfnamefont {C.~F.}\ \bibnamefont {Lee}}, \
and\ \bibinfo {author} {\bibfnamefont {J.}~\bibnamefont {Toner}},\ }\bibfield
{title} {\enquote {\bibinfo {title} {{Mapping two-dimensional polar active
fluids to two-dimensional soap and one-dimensional sandblasting}},}\ }\href
{\doibase 10.1038/ncomms12215} {\bibfield {journal} {\bibinfo {journal}
{Nat. Commun.}\ }\textbf {\bibinfo {volume} {7}},\ \bibinfo {pages} {12215}
(\bibinfo {year} {2016})}\BibitemShut {NoStop}%
\bibitem [{\citenamefont {Takeuchi}\ \emph {et~al.}(2011)\citenamefont
{Takeuchi}, \citenamefont {Sano}, \citenamefont {Sasamoto},\ and\
\citenamefont {Spohn}}]{takeuchi11}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {K.~A.}\ \bibnamefont
{Takeuchi}}, \bibinfo {author} {\bibfnamefont {M.}~\bibnamefont {Sano}},
\bibinfo {author} {\bibfnamefont {T.}~\bibnamefont {Sasamoto}}, \ and\
\bibinfo {author} {\bibfnamefont {H.}~\bibnamefont {Spohn}},\ }\bibfield
{title} {\enquote {\bibinfo {title} {{Growing interfaces uncover universal
fluctuations behind scale invariance.}}}\ }\href {\doibase 10.1038/srep00034}
{\bibfield {journal} {\bibinfo {journal} {Sci. Rep.}\ }\textbf {\bibinfo
{volume} {1}},\ \bibinfo {pages} {34} (\bibinfo {year} {2011})}\BibitemShut
{NoStop}%
\bibitem [{\citenamefont {Mendl}\ and\ \citenamefont {Spohn}(2013)}]{mendl13}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {C.~B.}\ \bibnamefont
{Mendl}}\ and\ \bibinfo {author} {\bibfnamefont {H.}~\bibnamefont {Spohn}},\
}\bibfield {title} {\enquote {\bibinfo {title} {{Dynamic correlators of
{F}ermi-{P}asta-{U}lam chains and nonlinear fluctuating hydrodynamics}},}\
}\href {\doibase 10.1103/PhysRevLett.111.230601} {\bibfield {journal}
{\bibinfo {journal} {Phys. Rev. Lett.}\ }\textbf {\bibinfo {volume} {111}},\
\bibinfo {pages} {230601} (\bibinfo {year} {2013})}\BibitemShut {NoStop}%
\bibitem [{\citenamefont {Yunker}\ \emph {et~al.}(2013)\citenamefont {Yunker},
\citenamefont {Lohr}, \citenamefont {Still}, \citenamefont {Borodin},
\citenamefont {Durian},\ and\ \citenamefont {Yodh}}]{yunker13}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {P.~J.}\ \bibnamefont
{Yunker}}, \bibinfo {author} {\bibfnamefont {M.~A.}\ \bibnamefont {Lohr}},
\bibinfo {author} {\bibfnamefont {T.}~\bibnamefont {Still}}, \bibinfo
{author} {\bibfnamefont {A.}~\bibnamefont {Borodin}}, \bibinfo {author}
{\bibfnamefont {D.~J.}\ \bibnamefont {Durian}}, \ and\ \bibinfo {author}
{\bibfnamefont {A.~G.}\ \bibnamefont {Yodh}},\ }\bibfield {title} {\enquote
{\bibinfo {title} {{Effects of Particle Shape on Growth Dynamics at Edges of
Evaporating Drops of Colloidal Suspensions}},}\ }\href {\doibase
10.1103/PhysRevLett.110.035501} {\bibfield {journal} {\bibinfo {journal}
{Phys. Rev. Lett.}\ }\textbf {\bibinfo {volume} {110}},\ \bibinfo {pages}
{035501} (\bibinfo {year} {2013})}\BibitemShut {NoStop}%
\bibitem [{\citenamefont {Orrillo}\ \emph {et~al.}(2017)\citenamefont
{Orrillo}, \citenamefont {Santalla}, \citenamefont {Cuerno}, \citenamefont
{V{\'{a}}zquez}, \citenamefont {Ribotta}, \citenamefont {Gassa},
\citenamefont {Mompean}, \citenamefont {Salvarezza},\ and\ \citenamefont
{Vela}}]{orrillo17}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {P.~A.}\ \bibnamefont
{Orrillo}}, \bibinfo {author} {\bibfnamefont {S.~N.}\ \bibnamefont
{Santalla}}, \bibinfo {author} {\bibfnamefont {R.}~\bibnamefont {Cuerno}},
\bibinfo {author} {\bibfnamefont {L.}~\bibnamefont {V{\'{a}}zquez}}, \bibinfo
{author} {\bibfnamefont {S.~B.}\ \bibnamefont {Ribotta}}, \bibinfo {author}
{\bibfnamefont {L.~M.}\ \bibnamefont {Gassa}}, \bibinfo {author}
{\bibfnamefont {F.~J.}\ \bibnamefont {Mompean}}, \bibinfo {author}
{\bibfnamefont {R.~C.}\ \bibnamefont {Salvarezza}}, \ and\ \bibinfo {author}
{\bibfnamefont {M.~E.}\ \bibnamefont {Vela}},\ }\bibfield {title} {\enquote
{\bibinfo {title} {{Morphological stabilization and KPZ scaling by
electrochemically induced co-deposition of nanostructured NiW alloy
films}},}\ }\href {\doibase 10.1038/s41598-017-18155-7} {\bibfield {journal}
{\bibinfo {journal} {Sci. Rep.}\ }\textbf {\bibinfo {volume} {7}},\ \bibinfo
{pages} {17997} (\bibinfo {year} {2017})}\BibitemShut {NoStop}%
\bibitem [{\citenamefont {Nesic}\ \emph {et~al.}(2014)\citenamefont {Nesic},
\citenamefont {Cuerno},\ and\ \citenamefont {Moro}}]{nesic14}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {S.}~\bibnamefont
{Nesic}}, \bibinfo {author} {\bibfnamefont {R.}~\bibnamefont {Cuerno}}, \
and\ \bibinfo {author} {\bibfnamefont {E.}~\bibnamefont {Moro}},\ }\bibfield
{title} {\enquote {\bibinfo {title} {{Macroscopic response to microscopic
intrinsic noise in three-dimensional Fisher fronts}},}\ }\href {\doibase
10.1103/PhysRevLett.113.180602} {\bibfield {journal} {\bibinfo {journal}
{Phys. Rev. Lett.}\ }\textbf {\bibinfo {volume} {113}},\ \bibinfo {pages}
{180602} (\bibinfo {year} {2014})}\BibitemShut {NoStop}%
\bibitem [{\citenamefont {Santalla}\ \emph {et~al.}(2015)\citenamefont
{Santalla}, \citenamefont {Rodr{\'{i}}guez-Laguna}, \citenamefont {Lagatta},\
and\ \citenamefont {Cuerno}}]{santalla15}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {S.~N.}\ \bibnamefont
{Santalla}}, \bibinfo {author} {\bibfnamefont {J.}~\bibnamefont
{Rodr{\'{i}}guez-Laguna}}, \bibinfo {author} {\bibfnamefont {T.}~\bibnamefont
{Lagatta}}, \ and\ \bibinfo {author} {\bibfnamefont {R.}~\bibnamefont
{Cuerno}},\ }\bibfield {title} {\enquote {\bibinfo {title} {{Random geometry
and the Kardar-Parisi-Zhang universality class}},}\ }\href {\doibase
10.1088/1367-2630/17/3/033018} {\bibfield {journal} {\bibinfo {journal}
{New J. Phys.}\ }\textbf {\bibinfo {volume} {17}},\ \bibinfo {pages} {33018}
(\bibinfo {year} {2015})}\BibitemShut {NoStop}%
\bibitem [{\citenamefont {Altman}\ \emph {et~al.}(2015)\citenamefont {Altman},
\citenamefont {Sieberer}, \citenamefont {Chen}, \citenamefont {Diehl},\ and\
\citenamefont {Toner}}]{altman15}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {E.}~\bibnamefont
{Altman}}, \bibinfo {author} {\bibfnamefont {L.~M.}\ \bibnamefont
{Sieberer}}, \bibinfo {author} {\bibfnamefont {L.}~\bibnamefont {Chen}},
\bibinfo {author} {\bibfnamefont {S.}~\bibnamefont {Diehl}}, \ and\ \bibinfo
{author} {\bibfnamefont {J.}~\bibnamefont {Toner}},\ }\bibfield {title}
{\enquote {\bibinfo {title} {{Two-dimensional superfluidity of exciton
polaritons requires strong anisotropy}},}\ }\href {\doibase
10.1103/PhysRevX.5.011017} {\bibfield {journal} {\bibinfo {journal} {Phys.
Rev. X}\ }\textbf {\bibinfo {volume} {5}},\ \bibinfo {pages} {011017}
(\bibinfo {year} {2015})}\BibitemShut {NoStop}%
\bibitem [{\citenamefont {Fontaine}\ \emph {et~al.}(2022)\citenamefont
{Fontaine}, \citenamefont {Squizzato}, \citenamefont {Baboux}, \citenamefont
{Amelio}, \citenamefont {Lema{\^{i}}tre}, \citenamefont {Morassi},
\citenamefont {Sagnes}, \citenamefont {Le~Gratiet}, \citenamefont {Harouri},
\citenamefont {Wouters}, \citenamefont {Carusotto}, \citenamefont {Amo},
\citenamefont {Richard}, \citenamefont {Minguzzi}, \citenamefont {Canet},
\citenamefont {Ravets},\ and\ \citenamefont {Bloch}}]{fontaine}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {Q.}~\bibnamefont
{Fontaine}}, \bibinfo {author} {\bibfnamefont {D.}~\bibnamefont {Squizzato}},
\bibinfo {author} {\bibfnamefont {F.}~\bibnamefont {Baboux}}, \bibinfo
{author} {\bibfnamefont {I.}~\bibnamefont {Amelio}}, \bibinfo {author}
{\bibfnamefont {A.}~\bibnamefont {Lema{\^{i}}tre}}, \bibinfo {author}
{\bibfnamefont {M.}~\bibnamefont {Morassi}}, \bibinfo {author} {\bibfnamefont
{I.}~\bibnamefont {Sagnes}}, \bibinfo {author} {\bibfnamefont
{L.}~\bibnamefont {Le~Gratiet}}, \bibinfo {author} {\bibfnamefont
{A.}~\bibnamefont {Harouri}}, \bibinfo {author} {\bibfnamefont
{M.}~\bibnamefont {Wouters}}, \bibinfo {author} {\bibfnamefont
{I.}~\bibnamefont {Carusotto}}, \bibinfo {author} {\bibfnamefont
{A.}~\bibnamefont {Amo}}, \bibinfo {author} {\bibfnamefont {M.}~\bibnamefont
{Richard}}, \bibinfo {author} {\bibfnamefont {A.}~\bibnamefont {Minguzzi}},
\bibinfo {author} {\bibfnamefont {L.}~\bibnamefont {Canet}}, \bibinfo
{author} {\bibfnamefont {S.}~\bibnamefont {Ravets}}, \ and\ \bibinfo {author}
{\bibfnamefont {J.}~\bibnamefont {Bloch}},\ }\bibfield {title} {\enquote
{\bibinfo {title} {{Kardar-Parisi-Zhang universality in a one-dimensional
polariton condensate}},}\ }\href {\doibase 10.1038/s41586-022-05001-8}
{\bibfield {journal} {\bibinfo {journal} {Nature}\ }\textbf {\bibinfo
{volume} {608}},\ \bibinfo {pages} {687} (\bibinfo {year}
{2022})}\BibitemShut {NoStop}%
\bibitem [{\citenamefont {Nahum}\ \emph {et~al.}(2017)\citenamefont {Nahum},
\citenamefont {Ruhman}, \citenamefont {Vijay},\ and\ \citenamefont
{Haah}}]{nahum17}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {A.}~\bibnamefont
{Nahum}}, \bibinfo {author} {\bibfnamefont {J.}~\bibnamefont {Ruhman}},
\bibinfo {author} {\bibfnamefont {S.}~\bibnamefont {Vijay}}, \ and\ \bibinfo
{author} {\bibfnamefont {J.}~\bibnamefont {Haah}},\ }\bibfield {title}
{\enquote {\bibinfo {title} {{Quantum entanglement growth under random
unitary dynamics}},}\ }\href {\doibase 10.1103/PhysRevX.7.031016} {\bibfield
{journal} {\bibinfo {journal} {Phys. Rev. X}\ }\textbf {\bibinfo {volume}
{7}},\ \bibinfo {pages} {031016} (\bibinfo {year} {2017})}\BibitemShut
{NoStop}%
\bibitem [{\citenamefont {Gopalakrishnan}\ and\ \citenamefont
{Vasseur}(2019)}]{gopalakrishnan19}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {S.}~\bibnamefont
{Gopalakrishnan}}\ and\ \bibinfo {author} {\bibfnamefont {R.}~\bibnamefont
{Vasseur}},\ }\bibfield {title} {\enquote {\bibinfo {title} {{Kinetic Theory
of Spin Diffusion and Superdiffusion in XXZ Spin Chains}},}\ }\href {\doibase
10.1103/PhysRevLett.122.127202} {\bibfield {journal} {\bibinfo {journal}
{Phys. Rev. Lett.}\ }\textbf {\bibinfo {volume} {122}},\ \bibinfo {pages}
{127202} (\bibinfo {year} {2019})}\BibitemShut {NoStop}%
\bibitem [{\citenamefont {Ljubotina}\ \emph {et~al.}(2019)\citenamefont
{Ljubotina}, \citenamefont {{\v{Z}}nidari{\v{c}}},\ and\ \citenamefont
{Prosen}}]{ljubotina19}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {M.}~\bibnamefont
{Ljubotina}}, \bibinfo {author} {\bibfnamefont {M.}~\bibnamefont
{{\v{Z}}nidari{\v{c}}}}, \ and\ \bibinfo {author} {\bibfnamefont
{T.}~\bibnamefont {Prosen}},\ }\bibfield {title} {\enquote {\bibinfo {title}
{{Kardar-Parisi-Zhang Physics in the Quantum Heisenberg Magnet}},}\ }\href
{\doibase 10.1103/PhysRevLett.122.210602} {\bibfield {journal} {\bibinfo
{journal} {Phys. Rev. Lett.}\ }\textbf {\bibinfo {volume} {122}},\ \bibinfo
{pages} {210602} (\bibinfo {year} {2019})}\BibitemShut {NoStop}%
\bibitem [{\citenamefont {{De Nardis}}\ \emph {et~al.}(2021)\citenamefont {{De
Nardis}}, \citenamefont {Gopalakrishnan}, \citenamefont {Vasseur},\ and\
\citenamefont {Ware}}]{denardis21}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {J.}~\bibnamefont {{De
Nardis}}}, \bibinfo {author} {\bibfnamefont {S.}~\bibnamefont
{Gopalakrishnan}}, \bibinfo {author} {\bibfnamefont {R.}~\bibnamefont
{Vasseur}}, \ and\ \bibinfo {author} {\bibfnamefont {B.}~\bibnamefont
{Ware}},\ }\bibfield {title} {\enquote {\bibinfo {title} {{Stability of
Superdiffusion in Nearly Integrable Spin Chains}},}\ }\href {\doibase
10.1103/PhysRevLett.127.057201} {\bibfield {journal} {\bibinfo {journal}
{Phys. Rev. Lett.}\ }\textbf {\bibinfo {volume} {127}},\ \bibinfo {pages}
{057201} (\bibinfo {year} {2021})}\BibitemShut {NoStop}%
\bibitem [{\citenamefont {Wei}\ \emph {et~al.}(2022)\citenamefont {Wei},
\citenamefont {Rubio-Abadal}, \citenamefont {Ye}, \citenamefont {Machado},
\citenamefont {Kemp}, \citenamefont {Srakaew}, \citenamefont {Hollerith},
\citenamefont {Rui}, \citenamefont {Gopalakrishnan}, \citenamefont {Yao},
\citenamefont {Bloch},\ and\ \citenamefont {Zeiher}}]{wei22}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {D.}~\bibnamefont
{Wei}}, \bibinfo {author} {\bibfnamefont {A.}~\bibnamefont {Rubio-Abadal}},
\bibinfo {author} {\bibfnamefont {B.}~\bibnamefont {Ye}}, \bibinfo {author}
{\bibfnamefont {F.}~\bibnamefont {Machado}}, \bibinfo {author} {\bibfnamefont
{J.}~\bibnamefont {Kemp}}, \bibinfo {author} {\bibfnamefont {K.}~\bibnamefont
{Srakaew}}, \bibinfo {author} {\bibfnamefont {S.}~\bibnamefont {Hollerith}},
\bibinfo {author} {\bibfnamefont {J.}~\bibnamefont {Rui}}, \bibinfo {author}
{\bibfnamefont {S.}~\bibnamefont {Gopalakrishnan}}, \bibinfo {author}
{\bibfnamefont {N.~Y.}\ \bibnamefont {Yao}}, \bibinfo {author} {\bibfnamefont
{I.}~\bibnamefont {Bloch}}, \ and\ \bibinfo {author} {\bibfnamefont
{J.}~\bibnamefont {Zeiher}},\ }\bibfield {title} {\enquote {\bibinfo {title}
{{Quantum gas microscopy of Kardar-Parisi-Zhang superdiffusion}},}\ }\href
{\doibase 10.1126/science.abk2397} {\bibfield {journal} {\bibinfo {journal}
{Science}\ }\textbf {\bibinfo {volume} {376}},\ \bibinfo {pages} {716}
(\bibinfo {year} {2022})}\BibitemShut {NoStop}%
\bibitem [{\citenamefont {Protopopov}\ \emph {et~al.}(2021)\citenamefont
{Protopopov}, \citenamefont {Samanta}, \citenamefont {Mirlin},\ and\
\citenamefont {Gutman}}]{protopopov21}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {I.~V.}\ \bibnamefont
{Protopopov}}, \bibinfo {author} {\bibfnamefont {R.}~\bibnamefont {Samanta}},
\bibinfo {author} {\bibfnamefont {A.~D.}\ \bibnamefont {Mirlin}}, \ and\
\bibinfo {author} {\bibfnamefont {D.~B.}\ \bibnamefont {Gutman}},\ }\bibfield
{title} {\enquote {\bibinfo {title} {{Anomalous Hydrodynamics in a
One-Dimensional Electronic Fluid}},}\ }\href {\doibase
10.1103/PhysRevLett.126.256801} {\bibfield {journal} {\bibinfo {journal}
{Phys. Rev. Lett.}\ }\textbf {\bibinfo {volume} {126}},\ \bibinfo {pages}
{256801} (\bibinfo {year} {2021})}\BibitemShut {NoStop}%
\bibitem [{\citenamefont {Tracy}\ and\ \citenamefont {Widom}(2009)}]{tracy09}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {C.~A.}\ \bibnamefont
{Tracy}}\ and\ \bibinfo {author} {\bibfnamefont {H.}~\bibnamefont {Widom}},\
}\bibfield {title} {\enquote {\bibinfo {title} {{The Distributions of Random
Matrix Theory and their Applications}},}\ }\href {\doibase
10.1007/978-90-481-2810-5_48} {\bibfield {journal} {\bibinfo {journal} {New
Trends Math. Phys.}\ }\textbf {\bibinfo {volume} {0553379}},\ \bibinfo
{pages} {753} (\bibinfo {year} {2009})}\BibitemShut {NoStop}%
\bibitem [{\citenamefont {Makey}\ \emph {et~al.}(2020)\citenamefont {Makey},
\citenamefont {Galioglu}, \citenamefont {Ghaffari}, \citenamefont {Engin},
\citenamefont {Yıldırım}, \citenamefont {Yavuz}, \citenamefont {Bektaş},
\citenamefont {Nizam}, \citenamefont {Akbulut}, \citenamefont {Şahin},
\citenamefont {G{\"{u}}ng{\"{o}}r}, \citenamefont {Dede}, \citenamefont
{Demir}, \citenamefont {Ilday},\ and\ \citenamefont {Ilday}}]{makey20}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {G.}~\bibnamefont
{Makey}}, \bibinfo {author} {\bibfnamefont {S.}~\bibnamefont {Galioglu}},
\bibinfo {author} {\bibfnamefont {R.}~\bibnamefont {Ghaffari}}, \bibinfo
{author} {\bibfnamefont {E.~D.}\ \bibnamefont {Engin}}, \bibinfo {author}
{\bibfnamefont {G.}~\bibnamefont {Yıldırım}}, \bibinfo {author}
{\bibfnamefont {{\"{O}}.}~\bibnamefont {Yavuz}}, \bibinfo {author}
{\bibfnamefont {O.}~\bibnamefont {Bektaş}}, \bibinfo {author} {\bibfnamefont
{{\"{U}}.~S.}\ \bibnamefont {Nizam}}, \bibinfo {author} {\bibfnamefont
{{\"{O}}.}~\bibnamefont {Akbulut}}, \bibinfo {author} {\bibfnamefont
{{\"{O}}.}~\bibnamefont {Şahin}}, \bibinfo {author} {\bibfnamefont
{K.}~\bibnamefont {G{\"{u}}ng{\"{o}}r}}, \bibinfo {author} {\bibfnamefont
{D.}~\bibnamefont {Dede}}, \bibinfo {author} {\bibfnamefont {H.~V.}\
\bibnamefont {Demir}}, \bibinfo {author} {\bibfnamefont {F.~{\"{O}}.}\
\bibnamefont {Ilday}}, \ and\ \bibinfo {author} {\bibfnamefont
{S.}~\bibnamefont {Ilday}},\ }\bibfield {title} {\enquote {\bibinfo {title}
{{Universality of dissipative self-assembly from quantum dots to human
cells}},}\ }\href {\doibase 10.1038/s41567-020-0879-8} {\bibfield {journal}
{\bibinfo {journal} {Nat. Phys.}\ }\textbf {\bibinfo {volume} {16}},\
\bibinfo {pages} {795} (\bibinfo {year} {2020})}\BibitemShut {NoStop}%
\bibitem [{\citenamefont {Sakaguchi}\ \emph {et~al.}(1988)\citenamefont
{Sakaguchi}, \citenamefont {Shinomoto},\ and\ \citenamefont
{Kuramoto}}]{sakaguchi}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {H.}~\bibnamefont
{Sakaguchi}}, \bibinfo {author} {\bibfnamefont {S.}~\bibnamefont
{Shinomoto}}, \ and\ \bibinfo {author} {\bibfnamefont {Y.}~\bibnamefont
{Kuramoto}},\ }\bibfield {title} {\enquote {\bibinfo {title} {{Mutual
Entrainment in Oscillator Lattices with Nonvariational Type Interaction}},}\
}\href {\doibase 10.1143/PTP.79.1069} {\bibfield {journal} {\bibinfo
{journal} {Prog. Theor. Phys.}\ }\textbf {\bibinfo {volume} {79}},\ \bibinfo
{pages} {1069} (\bibinfo {year} {1988})}\BibitemShut {NoStop}%
\bibitem [{\citenamefont {Grinstein}\ \emph {et~al.}(1993)\citenamefont
{Grinstein}, \citenamefont {Mukamel}, \citenamefont {Seidin},\ and\
\citenamefont {Bennett}}]{grinstein}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {G.}~\bibnamefont
{Grinstein}}, \bibinfo {author} {\bibfnamefont {D.}~\bibnamefont {Mukamel}},
\bibinfo {author} {\bibfnamefont {R.}~\bibnamefont {Seidin}}, \ and\ \bibinfo
{author} {\bibfnamefont {C.~H.}\ \bibnamefont {Bennett}},\ }\bibfield
{title} {\enquote {\bibinfo {title} {Temporally periodic phases and kinetic
roughening},}\ }\href {\doibase 10.1103/PhysRevLett.70.3607} {\bibfield
{journal} {\bibinfo {journal} {Phys. Rev. Lett.}\ }\textbf {\bibinfo
{volume} {70}},\ \bibinfo {pages} {3607} (\bibinfo {year}
{1993})}\BibitemShut {NoStop}%
\bibitem [{\citenamefont {Edwards}\ and\ \citenamefont
{Wilkinson}(1982)}]{edwards}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {S.~F.}\ \bibnamefont
{Edwards}}\ and\ \bibinfo {author} {\bibfnamefont {D.~R.}\ \bibnamefont
{Wilkinson}},\ }\bibfield {title} {\enquote {\bibinfo {title} {The surface
statistics of a granular aggregate},}\ }\href {\doibase
10.1098/rspa.1982.00560} {\bibfield {journal} {\bibinfo {journal} {Proc. R.
Soc. Lond. A}\ }\textbf {\bibinfo {volume} {381}},\ \bibinfo {pages} {17}
(\bibinfo {year} {1982})}\BibitemShut {NoStop}%
\bibitem [{\citenamefont {Lauter}\ \emph {et~al.}(2017)\citenamefont {Lauter},
\citenamefont {Mitra},\ and\ \citenamefont {Marquardt}}]{lauter}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {R.}~\bibnamefont
{Lauter}}, \bibinfo {author} {\bibfnamefont {A.}~\bibnamefont {Mitra}}, \
and\ \bibinfo {author} {\bibfnamefont {F.}~\bibnamefont {Marquardt}},\
}\bibfield {title} {\enquote {\bibinfo {title} {From
{K}ardar-{P}arisi-{Z}hang scaling to explosive desynchronization in arrays of
limit-cycle oscillators},}\ }\href {\doibase 10.1103/PhysRevE.96.012220}
{\bibfield {journal} {\bibinfo {journal} {Phys. Rev. E}\ }\textbf {\bibinfo
{volume} {96}},\ \bibinfo {pages} {012220} (\bibinfo {year}
{2017})}\BibitemShut {NoStop}%
\bibitem [{\citenamefont {Moroney}\ and\ \citenamefont
{Eastham}(2021)}]{moroney}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {J.~P.}\ \bibnamefont
{Moroney}}\ and\ \bibinfo {author} {\bibfnamefont {P.~R.}\ \bibnamefont
{Eastham}},\ }\bibfield {title} {\enquote {\bibinfo {title} {Synchronization
in disordered oscillator lattices: Nonequilibrium phase transition for
driven-dissipative bosons},}\ }\href {\doibase
10.1103/PhysRevResearch.3.043092} {\bibfield {journal} {\bibinfo {journal}
{Phys. Rev. Research}\ }\textbf {\bibinfo {volume} {3}},\ \bibinfo {pages}
{043092} (\bibinfo {year} {2021})}\BibitemShut {NoStop}%
\bibitem [{\citenamefont {Kuramoto}(1984{\natexlab{b}})}]{kuramoto1984}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {Y.}~\bibnamefont
{Kuramoto}},\ }\bibfield {title} {\enquote {\bibinfo {title} {{Cooperative
Dynamics of Oscillator Community: A Study Based on Lattice of Rings}},}\
}\href {\doibase 10.1143/PTPS.79.223} {\bibfield {journal} {\bibinfo
{journal} {Prog. Theor. Phys. Supp.}\ }\textbf {\bibinfo {volume} {79}},\
\bibinfo {pages} {223} (\bibinfo {year} {1984}{\natexlab{b}})}\BibitemShut
{NoStop}%
\bibitem [{\citenamefont {Strogatz}\ and\ \citenamefont
{Mirollo}(1988)}]{strogatz}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {S.~H.}\ \bibnamefont
{Strogatz}}\ and\ \bibinfo {author} {\bibfnamefont {R.~E.}\ \bibnamefont
{Mirollo}},\ }\bibfield {title} {\enquote {\bibinfo {title} {Phase-locking
and critical phenomena in lattices of coupled nonlinear oscillators with
random intrinsic frequencies},}\ }\href {\doibase
https://doi.org/10.1016/0167-2789(88)90074-7} {\bibfield {journal} {\bibinfo
{journal} {Physica D}\ }\textbf {\bibinfo {volume} {31}},\ \bibinfo {pages}
{143} (\bibinfo {year} {1988})}\BibitemShut {NoStop}%
\bibitem [{\citenamefont {Das~Sarma}\ \emph {et~al.}(1994)\citenamefont
{Das~Sarma}, \citenamefont {Ghaisas},\ and\ \citenamefont {Kim}}]{dassarma}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {S.}~\bibnamefont
{Das~Sarma}}, \bibinfo {author} {\bibfnamefont {S.~V.}\ \bibnamefont
{Ghaisas}}, \ and\ \bibinfo {author} {\bibfnamefont {J.~M.}\ \bibnamefont
{Kim}},\ }\bibfield {title} {\enquote {\bibinfo {title} {Kinetic
super-roughening and anomalous dynamic scaling in nonequilibrium growth
models},}\ }\href {\doibase 10.1103/PhysRevE.49.122} {\bibfield {journal}
{\bibinfo {journal} {Phys. Rev. E}\ }\textbf {\bibinfo {volume} {49}},\
\bibinfo {pages} {122} (\bibinfo {year} {1994})}\BibitemShut {NoStop}%
\bibitem [{\citenamefont {Schroeder}\ and\ \citenamefont
{Siegert}(1993)}]{schroeder93}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {M.}~\bibnamefont
{Schroeder}}\ and\ \bibinfo {author} {\bibfnamefont {M.}~\bibnamefont
{Siegert}},\ }\bibfield {title} {\enquote {\bibinfo {title} {{Scaling of
growing surfaces with large local slopes}},}\ }\href {\doibase
10.1209/0295-5075/24/7/010} {\bibfield {journal} {\bibinfo {journal}
{Europhys. Lett.}\ }\textbf {\bibinfo {volume} {24}},\ \bibinfo {pages} {563}
(\bibinfo {year} {1993})}\BibitemShut {NoStop}%
\bibitem [{\citenamefont {L\'opez}\ \emph {et~al.}(1997)\citenamefont
{L\'opez}, \citenamefont {Rodr\'{\i}guez},\ and\ \citenamefont
{Cuerno}}]{lopez}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {J.~M.}\ \bibnamefont
{L\'opez}}, \bibinfo {author} {\bibfnamefont {M.~A.}\ \bibnamefont
{Rodr\'{\i}guez}}, \ and\ \bibinfo {author} {\bibfnamefont {R.}~\bibnamefont
{Cuerno}},\ }\bibfield {title} {\enquote {\bibinfo {title} {Superroughening
versus intrinsic anomalous scaling of surfaces},}\ }\href {\doibase
10.1103/PhysRevE.56.3993} {\bibfield {journal} {\bibinfo {journal} {Phys.
Rev. E}\ }\textbf {\bibinfo {volume} {56}},\ \bibinfo {pages} {3993}
(\bibinfo {year} {1997})}\BibitemShut {NoStop}%
\bibitem [{\citenamefont {Purrello}\ \emph {et~al.}(2019)\citenamefont
{Purrello}, \citenamefont {Iguain},\ and\ \citenamefont {Kolton}}]{purrello}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {V.~H.}\ \bibnamefont
{Purrello}}, \bibinfo {author} {\bibfnamefont {J.~L.}\ \bibnamefont
{Iguain}}, \ and\ \bibinfo {author} {\bibfnamefont {A.~B.}\ \bibnamefont
{Kolton}},\ }\bibfield {title} {\enquote {\bibinfo {title} {Roughening of
the anharmonic {L}arkin model},}\ }\href {\doibase
10.1103/PhysRevE.99.032105} {\bibfield {journal} {\bibinfo {journal} {Phys.
Rev. E}\ }\textbf {\bibinfo {volume} {99}},\ \bibinfo {pages} {032105}
(\bibinfo {year} {2019})}\BibitemShut {NoStop}%
\bibitem [{\citenamefont {\"Ostborn}(2004)}]{ostborn}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {P.}~\bibnamefont
{\"Ostborn}},\ }\bibfield {title} {\enquote {\bibinfo {title} {Frequency
entrainment in long chains of oscillators with random natural frequencies in
the weak coupling limit},}\ }\href {\doibase 10.1103/PhysRevE.70.016120}
{\bibfield {journal} {\bibinfo {journal} {Phys. Rev. E}\ }\textbf {\bibinfo
{volume} {70}},\ \bibinfo {pages} {016120} (\bibinfo {year}
{2004})}\BibitemShut {NoStop}%
\bibitem [{\citenamefont {Ramasco}\ \emph {et~al.}(2000)\citenamefont
{Ramasco}, \citenamefont {L\'opez},\ and\ \citenamefont
{Rodr\'{\i}guez}}]{ramasco}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {J.~J.}\ \bibnamefont
{Ramasco}}, \bibinfo {author} {\bibfnamefont {J.~M.}\ \bibnamefont
{L\'opez}}, \ and\ \bibinfo {author} {\bibfnamefont {M.~A.}\ \bibnamefont
{Rodr\'{\i}guez}},\ }\bibfield {title} {\enquote {\bibinfo {title} {Generic
dynamic scaling in kinetic roughening},}\ }\href {\doibase
10.1103/PhysRevLett.84.2199} {\bibfield {journal} {\bibinfo {journal}
{Phys. Rev. Lett.}\ }\textbf {\bibinfo {volume} {84}},\ \bibinfo {pages}
{2199} (\bibinfo {year} {2000})}\BibitemShut {NoStop}%
\bibitem [{\citenamefont {Szendro}\ \emph {et~al.}(2007)\citenamefont
{Szendro}, \citenamefont {L\'opez},\ and\ \citenamefont
{Rodr\'{\i}guez}}]{szendro}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {I.~G.}\ \bibnamefont
{Szendro}}, \bibinfo {author} {\bibfnamefont {J.~M.}\ \bibnamefont
{L\'opez}}, \ and\ \bibinfo {author} {\bibfnamefont {M.~A.}\ \bibnamefont
{Rodr\'{\i}guez}},\ }\bibfield {title} {\enquote {\bibinfo {title}
{Localization in disordered media, anomalous roughening, and coarsening
dynamics of faceted surfaces},}\ }\href {\doibase 10.1103/PhysRevE.76.011603}
{\bibfield {journal} {\bibinfo {journal} {Phys. Rev. E}\ }\textbf {\bibinfo
{volume} {76}},\ \bibinfo {pages} {011603} (\bibinfo {year}
{2007})}\BibitemShut {NoStop}%
\bibitem [{\citenamefont {Cuerno}\ and\ \citenamefont
{V\'{a}zquez}(2004)}]{cuerno04}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {R.}~\bibnamefont
{Cuerno}}\ and\ \bibinfo {author} {\bibfnamefont {L.}~\bibnamefont
{V\'{a}zquez}},\ }\bibfield {title} {\enquote {\bibinfo {title}
{Universality issues in surface kinetic roughening of thin solid films},}\
}in\ \href@noop {} {\emph {\bibinfo {booktitle} {Advances in Condensed Matter
and Statistical Physics}}},\ \bibinfo {editor} {edited by\ \bibinfo {editor}
{\bibfnamefont {E.}~\bibnamefont {Korutcheva}}\ and\ \bibinfo {editor}
{\bibfnamefont {R.}~\bibnamefont {Cuerno}}}\ (\bibinfo {publisher} {Nova
Science Publishers},\ \bibinfo {address} {New York},\ \bibinfo {year}
{2004})\BibitemShut {NoStop}%
\bibitem [{\citenamefont {Rodr\'{\i}guez-Fern\'andez}\ and\ \citenamefont
{Cuerno}(2021)}]{rodriguez-fernandez21}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {E.}~\bibnamefont
{Rodr\'{\i}guez-Fern\'andez}}\ and\ \bibinfo {author} {\bibfnamefont
{R.}~\bibnamefont {Cuerno}},\ }\bibfield {title} {\enquote {\bibinfo {title}
{Transition between chaotic and stochastic universality classes of kinetic
roughening},}\ }\href {\doibase 10.1103/PhysRevResearch.3.L012020} {\bibfield
{journal} {\bibinfo {journal} {Phys. Rev. Research}\ }\textbf {\bibinfo
{volume} {3}},\ \bibinfo {pages} {L012020} (\bibinfo {year}
{2021})}\BibitemShut {NoStop}%
\bibitem [{\citenamefont {Marcos}\ \emph {et~al.}(2022)\citenamefont {Marcos},
\citenamefont {Rodr\'{\i}guez-L\'opez}, \citenamefont {Mel\'endez},
\citenamefont {Cuerno},\ and\ \citenamefont {Ruiz-Lorenzo}}]{marcos22}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {J.~M.}\ \bibnamefont
{Marcos}}, \bibinfo {author} {\bibfnamefont {P.}~\bibnamefont
{Rodr\'{\i}guez-L\'opez}}, \bibinfo {author} {\bibfnamefont {J.~J.}\
\bibnamefont {Mel\'endez}}, \bibinfo {author} {\bibfnamefont
{R.}~\bibnamefont {Cuerno}}, \ and\ \bibinfo {author} {\bibfnamefont {J.~J.}\
\bibnamefont {Ruiz-Lorenzo}},\ }\bibfield {title} {\enquote {\bibinfo
{title} {Spreading fronts of wetting liquid droplets: Microscopic simulations
and universal fluctuations},}\ }\href {\doibase 10.1103/PhysRevE.105.054801}
{\bibfield {journal} {\bibinfo {journal} {Phys. Rev. E}\ }\textbf {\bibinfo
{volume} {105}},\ \bibinfo {pages} {054801} (\bibinfo {year}
{2022})}\BibitemShut {NoStop}%
\bibitem [{\citenamefont {Sakaguchi}\ and\ \citenamefont
{Kuramoto}(1986)}]{sakaguchi86}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {H.}~\bibnamefont
{Sakaguchi}}\ and\ \bibinfo {author} {\bibfnamefont {Y.}~\bibnamefont
{Kuramoto}},\ }\bibfield {title} {\enquote {\bibinfo {title} {A soluble
active rotater model showing phase transitions via mutual entertainment},}\
}\href {\doibase 10.1143/PTP.76.576} {\bibfield {journal} {\bibinfo
{journal} {Prog. Theor. Phys.}\ }\textbf {\bibinfo {volume} {76}},\ \bibinfo
{pages} {576} (\bibinfo {year} {1986})}\BibitemShut {NoStop}%
\bibitem [{\citenamefont {Vicsek}\ and\ \citenamefont {Family}(1984)}]{vicsek}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {T.}~\bibnamefont
{Vicsek}}\ and\ \bibinfo {author} {\bibfnamefont {F.}~\bibnamefont
{Family}},\ }\bibfield {title} {\enquote {\bibinfo {title} {Dynamic scaling
for aggregation of clusters},}\ }\href {\doibase 10.1103/PhysRevLett.52.1669}
{\bibfield {journal} {\bibinfo {journal} {Phys. Rev. Lett.}\ }\textbf
{\bibinfo {volume} {52}},\ \bibinfo {pages} {1669} (\bibinfo {year}
{1984})}\BibitemShut {NoStop}%
\bibitem [{\citenamefont {López}\ \emph {et~al.}(1997)\citenamefont {López},
\citenamefont {Rodríguez},\ and\ \citenamefont {Cuerno}}]{lopezphysa}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {J.~M.}\ \bibnamefont
{López}}, \bibinfo {author} {\bibfnamefont {M.~A.}\ \bibnamefont
{Rodríguez}}, \ and\ \bibinfo {author} {\bibfnamefont {R.}~\bibnamefont
{Cuerno}},\ }\bibfield {title} {\enquote {\bibinfo {title} {Power spectrum
scaling in anomalous kinetic roughening of surfaces},}\ }\href {\doibase
https://doi.org/10.1016/S0378-4371(97)00375-0} {\bibfield {journal}
{\bibinfo {journal} {Physica A}\ }\textbf {\bibinfo {volume} {246}},\
\bibinfo {pages} {329} (\bibinfo {year} {1997})}\BibitemShut {NoStop}%
\bibitem [{\citenamefont {Siegert}(1996)}]{siegert96}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {M.}~\bibnamefont
{Siegert}},\ }\bibfield {title} {\enquote {\bibinfo {title} {{Determining
exponents in models of kinetic surface roughening}},}\ }\href {\doibase
10.1103/PhysRevE.53.3209} {\bibfield {journal} {\bibinfo {journal} {Phys.
Rev. E}\ }\textbf {\bibinfo {volume} {53}},\ \bibinfo {pages} {3209}
(\bibinfo {year} {1996})}\BibitemShut {NoStop}%
\bibitem [{\citenamefont {Rodriguez-Fernandez}\ \emph
{et~al.}(2022{\natexlab{a}})\citenamefont {Rodriguez-Fernandez},
\citenamefont {Santalla}, \citenamefont {Castro},\ and\ \citenamefont
{Cuerno}}]{rodriguez-fernandez22}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {E.}~\bibnamefont
{Rodriguez-Fernandez}}, \bibinfo {author} {\bibfnamefont {S.~N.}\
\bibnamefont {Santalla}}, \bibinfo {author} {\bibfnamefont {M.}~\bibnamefont
{Castro}}, \ and\ \bibinfo {author} {\bibfnamefont {R.}~\bibnamefont
{Cuerno}},\ }\bibfield {title} {\enquote {\bibinfo {title} {{Anomalous
ballistic scaling in the tensionless or inviscid Kardar-Parisi-Zhang
equation}},}\ }\href {\doibase 10.1103/PhysRevE.106.024802} {\bibfield
{journal} {\bibinfo {journal} {Phys. Rev. E}\ }\textbf {\bibinfo {volume}
{106}},\ \bibinfo {pages} {024802} (\bibinfo {year}
{2022}{\natexlab{a}})}\BibitemShut {NoStop}%
\bibitem [{\citenamefont {Kopell}\ and\ \citenamefont
{Ermentrout}(1990)}]{kopell}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {N.}~\bibnamefont
{Kopell}}\ and\ \bibinfo {author} {\bibfnamefont {G.~B.}\ \bibnamefont
{Ermentrout}},\ }\bibfield {title} {\enquote {\bibinfo {title} {Phase
transitions and other phenomena in chains of coupled oscillators},}\ }\href
{\doibase 10.1137/0150062} {\bibfield {journal} {\bibinfo {journal} {SIAM
J. Appl. Math.}\ }\textbf {\bibinfo {volume} {50}},\ \bibinfo {pages} {1014}
(\bibinfo {year} {1990})}\BibitemShut {NoStop}%
\bibitem [{\citenamefont {Papoulis}\ and\ \citenamefont
{Pilla}(2015)}]{papoulis}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {A.}~\bibnamefont
{Papoulis}}\ and\ \bibinfo {author} {\bibfnamefont {S.~U.}\ \bibnamefont
{Pilla}},\ }\href@noop {} {\emph {\bibinfo {title} {Probability, Random
Variables, and Stochastic Processes}}}\ (\bibinfo {publisher} {McGraw-Hill
Education},\ \bibinfo {address} {New York},\ \bibinfo {year}
{2015})\BibitemShut {NoStop}%
\bibitem [{\citenamefont {Rodriguez-Fernandez}\ \emph
{et~al.}(2022{\natexlab{b}})\citenamefont {Rodriguez-Fernandez},
\citenamefont {Santalla}, \citenamefont {Castro},\ and\ \citenamefont
{Cuerno}}]{rodriguezfernandez1}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {E.}~\bibnamefont
{Rodriguez-Fernandez}}, \bibinfo {author} {\bibfnamefont {S.~N.}\
\bibnamefont {Santalla}}, \bibinfo {author} {\bibfnamefont {M.}~\bibnamefont
{Castro}}, \ and\ \bibinfo {author} {\bibfnamefont {R.}~\bibnamefont
{Cuerno}},\ }\bibfield {title} {\enquote {\bibinfo {title} {Is there a
tensionless {K}ardar-{P}arisi-{Z}hang universality class above one dimension?
{A}n {I}sing model approach},}\ }\href {\doibase 10.48550/arxiv.2208.00914}
{\bibfield {journal} {\bibinfo {journal} {arXiv}\ ,\ \bibinfo {pages}
{2208.00914}} (\bibinfo {year} {2022}{\natexlab{b}})}\BibitemShut {NoStop}%
\bibitem [{\citenamefont {Rodriguez-Fernandez}\ \emph
{et~al.}(2022{\natexlab{c}})\citenamefont {Rodriguez-Fernandez},
\citenamefont {Santalla}, \citenamefont {Castro},\ and\ \citenamefont
{Cuerno}}]{rodriguezfernandez2}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {E.}~\bibnamefont
{Rodriguez-Fernandez}}, \bibinfo {author} {\bibfnamefont {S.~N.}\
\bibnamefont {Santalla}}, \bibinfo {author} {\bibfnamefont {M.}~\bibnamefont
{Castro}}, \ and\ \bibinfo {author} {\bibfnamefont {R.}~\bibnamefont
{Cuerno}},\ }\bibfield {title} {\enquote {\bibinfo {title} {Anomalous
ballistic scaling in the tensionless or inviscid {K}ardar-{P}arisi-{Z}hang
equation},}\ }\href {\doibase 10.48550/arxiv.2205.08816} {\bibfield
{journal} {\bibinfo {journal} {arXiv}\ ,\ \bibinfo {pages} {2205.08816}}
(\bibinfo {year} {2022}{\natexlab{c}})}\BibitemShut {NoStop}%
\bibitem [{\citenamefont {L\'opez}(1999)}]{lopez99}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {J.~M.}\ \bibnamefont
{L\'opez}},\ }\bibfield {title} {\enquote {\bibinfo {title} {Scaling
approach to calculate critical exponents in anomalous surface roughening},}\
}\href {\doibase 10.1103/PhysRevLett.83.4594} {\bibfield {journal} {\bibinfo
{journal} {Phys. Rev. Lett.}\ }\textbf {\bibinfo {volume} {83}},\ \bibinfo
{pages} {4594} (\bibinfo {year} {1999})}\BibitemShut {NoStop}%
\bibitem [{\citenamefont {Larkin}\ and\ \citenamefont
{Ovchinnikov}(1979)}]{larkin}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {A.~I.}\ \bibnamefont
{Larkin}}\ and\ \bibinfo {author} {\bibfnamefont {Yu.~N.}\ \bibnamefont
{Ovchinnikov}},\ }\bibfield {title} {\enquote {\bibinfo {title} {Pinning in
type ii superconductors},}\ }\href {\doibase 10.1007/BF00117160} {\bibfield
{journal} {\bibinfo {journal} {J. Low. Temp. Phys.}\ }\textbf {\bibinfo
{volume} {34}},\ \bibinfo {pages} {409} (\bibinfo {year} {1979})}\BibitemShut
{NoStop}%
\bibitem [{\citenamefont {Hong}\ \emph {et~al.}(2005)\citenamefont {Hong},
\citenamefont {Park},\ and\ \citenamefont {Choi}}]{hong}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {H.}~\bibnamefont
{Hong}}, \bibinfo {author} {\bibfnamefont {H.}~\bibnamefont {Park}}, \ and\
\bibinfo {author} {\bibfnamefont {M.~Y.}\ \bibnamefont {Choi}},\ }\bibfield
{title} {\enquote {\bibinfo {title} {Collective synchronization in spatially
extended systems of coupled oscillators with random frequencies},}\ }\href
{\doibase 10.1103/PhysRevE.72.036217} {\bibfield {journal} {\bibinfo
{journal} {Phys. Rev. E}\ }\textbf {\bibinfo {volume} {72}},\ \bibinfo
{pages} {036217} (\bibinfo {year} {2005})}\BibitemShut {NoStop}%
\bibitem [{sup()}]{suppl}%
\BibitemOpen
\href@noop {} {}\bibinfo {note} {See the Supplemental Material at XXX for
movies displaying the dynamics of the oscillator lattices corresponding to
the cases considered in Figs.\ 6 (file S1.avi) and 11 (files S2.avi and
S3.avi).}\BibitemShut {Stop}%
\bibitem [{\citenamefont {Barreales}\ \emph {et~al.}(2020)\citenamefont
{Barreales}, \citenamefont {Mel{\'e}ndez}, \citenamefont {Cuerno},\ and\
\citenamefont {Ruiz-Lorenzo}}]{barreales2020}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {B.~G.}\ \bibnamefont
{Barreales}}, \bibinfo {author} {\bibfnamefont {J.~J.}\ \bibnamefont
{Mel{\'e}ndez}}, \bibinfo {author} {\bibfnamefont {R.}~\bibnamefont
{Cuerno}}, \ and\ \bibinfo {author} {\bibfnamefont {J.~J.}\ \bibnamefont
{Ruiz-Lorenzo}},\ }\bibfield {title} {\enquote {\bibinfo {title}
{Kardar--{P}arisi--{Z}hang universality class for the critical dynamics of
reaction--diffusion fronts},}\ }\href {\doibase 10.1088/1742-5468/ab6a03}
{\bibfield {journal} {\bibinfo {journal} {J. Stat. Mech.}\ }\textbf
{\bibinfo {volume} {2020}},\ \bibinfo {pages} {023203} (\bibinfo {year}
{2020})}\BibitemShut {NoStop}%
\bibitem [{\citenamefont {Leschhorn}\ and\ \citenamefont
{Tang}(1993)}]{leschhorn}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {H.}~\bibnamefont
{Leschhorn}}\ and\ \bibinfo {author} {\bibfnamefont {L.-H.}\ \bibnamefont
{Tang}},\ }\bibfield {title} {\enquote {\bibinfo {title} {Comment on
``{E}lastic string in a random potential''},}\ }\href {\doibase
10.1103/PhysRevLett.70.2973} {\bibfield {journal} {\bibinfo {journal}
{Phys. Rev. Lett.}\ }\textbf {\bibinfo {volume} {70}},\ \bibinfo {pages}
{2973} (\bibinfo {year} {1993})}\BibitemShut {NoStop}%
\bibitem [{\citenamefont {Nattermann}\ and\ \citenamefont
{Renz}(1989)}]{nattermann}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {T.}~\bibnamefont
{Nattermann}}\ and\ \bibinfo {author} {\bibfnamefont {W.}~\bibnamefont
{Renz}},\ }\bibfield {title} {\enquote {\bibinfo {title} {Diffusion in a
random catalytic environment, polymers in random media, and stochastically
growing interfaces},}\ }\href {\doibase 10.1103/PhysRevA.40.4675} {\bibfield
{journal} {\bibinfo {journal} {Phys. Rev. A}\ }\textbf {\bibinfo {volume}
{40}},\ \bibinfo {pages} {4675} (\bibinfo {year} {1989})}\BibitemShut
{NoStop}%
\bibitem [{\citenamefont {Krug}\ and\ \citenamefont
{Halpin-Healy}(1993)}]{krughh}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {J.}~\bibnamefont
{Krug}}\ and\ \bibinfo {author} {\bibfnamefont {T.}~\bibnamefont
{Halpin-Healy}},\ }\bibfield {title} {\enquote {\bibinfo {title} {Directed
polymers in the presence of columnar disorder},}\ }\href {\doibase
10.1051/jp1:1993240} {\bibfield {journal} {\bibinfo {journal} {J. Phys. I
France}\ }\textbf {\bibinfo {volume} {3}},\ \bibinfo {pages} {2179} (\bibinfo
{year} {1993})}\BibitemShut {NoStop}%
\bibitem [{\citenamefont {Bornemann}(2010)}]{bornemann}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {F.}~\bibnamefont
{Bornemann}},\ }\bibfield {title} {\enquote {\bibinfo {title} {{On the
numerical evaluation of Fredholm Determinants}},}\ }\href {\doibase
10.1090/S0025-5718-09-02280-7} {\bibfield {journal} {\bibinfo {journal}
{Math. Comput.}\ }\textbf {\bibinfo {volume} {79}},\ \bibinfo {pages} {871}
(\bibinfo {year} {2010})}\BibitemShut {NoStop}%
\bibitem [{\citenamefont {Alves}\ \emph {et~al.}(2011)\citenamefont {Alves},
\citenamefont {Oliveira},\ and\ \citenamefont {Ferreira}}]{alves}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {S.~G.}\ \bibnamefont
{Alves}}, \bibinfo {author} {\bibfnamefont {T.~J.}\ \bibnamefont {Oliveira}},
\ and\ \bibinfo {author} {\bibfnamefont {S.~C.}\ \bibnamefont {Ferreira}},\
}\bibfield {title} {\enquote {\bibinfo {title} {{Universal fluctuations in
radial growth models belonging to the KPZ universality class}},}\ }\href
{\doibase 10.1209/0295-5075/96/48003} {\bibfield {journal} {\bibinfo
{journal} {EPL}\ }\textbf {\bibinfo {volume} {96}},\ \bibinfo {pages} {48003}
(\bibinfo {year} {2011})}\BibitemShut {NoStop}%
\bibitem [{\citenamefont {Oliveira}\ \emph {et~al.}(2012)\citenamefont
{Oliveira}, \citenamefont {Ferreira},\ and\ \citenamefont
{Alves}}]{oliveira}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {T.~J.}\ \bibnamefont
{Oliveira}}, \bibinfo {author} {\bibfnamefont {S.~C.}\ \bibnamefont
{Ferreira}}, \ and\ \bibinfo {author} {\bibfnamefont {S.~G.}\ \bibnamefont
{Alves}},\ }\bibfield {title} {\enquote {\bibinfo {title} {{Universal
fluctuations in Kardar-Parisi-Zhang growth on one-dimensional flat
substrates}},}\ }\href {\doibase 10.1103/PhysRevE.85.010601} {\bibfield
{journal} {\bibinfo {journal} {Phys. Rev. E}\ }\textbf {\bibinfo {volume}
{85}},\ \bibinfo {pages} {010601(R)} (\bibinfo {year} {2012})}\BibitemShut
{NoStop}%
\bibitem [{\citenamefont {Carrasco}\ and\ \citenamefont
{Oliveira}(2019)}]{carrasco}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {I.~S.~S.}\
\bibnamefont {Carrasco}}\ and\ \bibinfo {author} {\bibfnamefont {T.~J.}\
\bibnamefont {Oliveira}},\ }\bibfield {title} {\enquote {\bibinfo {title}
{{Geometry dependence in linear interface growth}},}\ }\href {\doibase
10.1103/PhysRevE.100.042107} {\bibfield {journal} {\bibinfo {journal}
{Phys. Rev. E}\ }\textbf {\bibinfo {volume} {100}},\ \bibinfo {pages}
{042107} (\bibinfo {year} {2019})}\BibitemShut {NoStop}%
\bibitem [{\citenamefont {Sieberer}\ \emph {et~al.}(2016)\citenamefont
{Sieberer}, \citenamefont {Wachtel}, \citenamefont {Altman},\ and\
\citenamefont {Diehl}}]{sieberer}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {L.~M.}\ \bibnamefont
{Sieberer}}, \bibinfo {author} {\bibfnamefont {G.}~\bibnamefont {Wachtel}},
\bibinfo {author} {\bibfnamefont {E.}~\bibnamefont {Altman}}, \ and\ \bibinfo
{author} {\bibfnamefont {S.}~\bibnamefont {Diehl}},\ }\bibfield {title}
{\enquote {\bibinfo {title} {Lattice duality for the compact
{K}ardar-{P}arisi-{Z}hang equation},}\ }\href {\doibase
10.1103/PhysRevB.94.104521} {\bibfield {journal} {\bibinfo {journal} {Phys.
Rev. B}\ }\textbf {\bibinfo {volume} {94}},\ \bibinfo {pages} {104521}
(\bibinfo {year} {2016})}\BibitemShut {NoStop}%
\bibitem [{\citenamefont {Nicoli}\ \emph {et~al.}(2013)\citenamefont {Nicoli},
\citenamefont {Cuerno},\ and\ \citenamefont {Castro}}]{nicoli13}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {M.}~\bibnamefont
{Nicoli}}, \bibinfo {author} {\bibfnamefont {R.}~\bibnamefont {Cuerno}}, \
and\ \bibinfo {author} {\bibfnamefont {M.}~\bibnamefont {Castro}},\
}\bibfield {title} {\enquote {\bibinfo {title} {Dimensional fragility of the
{K}ardar-{P}arisi-{Z}hang universality class},}\ }\href {\doibase
10.48550/arxiv.1304.5873} {\bibfield {journal} {\bibinfo {journal} {arXiv}\
,\ \bibinfo {pages} {1304.5873v1}} (\bibinfo {year} {2013})}\BibitemShut
{NoStop}%
\bibitem [{\citenamefont {Nicoli}\ \emph {et~al.}(2009)\citenamefont {Nicoli},
\citenamefont {Cuerno},\ and\ \citenamefont {Castro}}]{nicoli}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {M.}~\bibnamefont
{Nicoli}}, \bibinfo {author} {\bibfnamefont {R.}~\bibnamefont {Cuerno}}, \
and\ \bibinfo {author} {\bibfnamefont {M.}~\bibnamefont {Castro}},\
}\bibfield {title} {\enquote {\bibinfo {title} {{Unstable nonlocal interface
dynamics}},}\ }\href {\doibase 10.1103/PhysRevLett.102.256102} {\bibfield
{journal} {\bibinfo {journal} {Phys. Rev. Lett.}\ }\textbf {\bibinfo
{volume} {102}},\ \bibinfo {pages} {256102} (\bibinfo {year}
{2009})}\BibitemShut {NoStop}%
\bibitem [{\citenamefont {Abraham}\ \emph {et~al.}(2002)\citenamefont
{Abraham}, \citenamefont {Cuerno},\ and\ \citenamefont {Moro}}]{abraham}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {D.~B.}\ \bibnamefont
{Abraham}}, \bibinfo {author} {\bibfnamefont {R.}~\bibnamefont {Cuerno}}, \
and\ \bibinfo {author} {\bibfnamefont {E.}~\bibnamefont {Moro}},\ }\bibfield
{title} {\enquote {\bibinfo {title} {{Microscopic model for thin film
spreading}},}\ }\href {\doibase 10.1103/PhysRevLett.88.206101} {\bibfield
{journal} {\bibinfo {journal} {Phys. Rev. Lett.}\ }\textbf {\bibinfo
{volume} {88}},\ \bibinfo {pages} {206101} (\bibinfo {year}
{2002})}\BibitemShut {NoStop}%
\bibitem [{\citenamefont {Cartes}\ \emph {et~al.}(2022)\citenamefont {Cartes},
\citenamefont {Tirapegui}, \citenamefont {Pandit},\ and\ \citenamefont
{Brachet}}]{cartes}%
\BibitemOpen
\bibfield {author} {\bibinfo {author} {\bibfnamefont {C.}~\bibnamefont
{Cartes}}, \bibinfo {author} {\bibfnamefont {E.}~\bibnamefont {Tirapegui}},
\bibinfo {author} {\bibfnamefont {R.}~\bibnamefont {Pandit}}, \ and\ \bibinfo
{author} {\bibfnamefont {M.}~\bibnamefont {Brachet}},\ }\bibfield {title}
{\enquote {\bibinfo {title} {{The Galerkin-truncated Burgers equation:
crossover from inviscid-thermalized to Kardar-Parisi-Zhang scaling}},}\
}\href {\doibase 10.1098/rsta.2021.0090} {\bibfield {journal} {\bibinfo
{journal} {Phil. Trans. Roy. Soc. A}\ }\textbf {\bibinfo {volume} {380}},\
\bibinfo {pages} {2021009} (\bibinfo {year} {2022})}\BibitemShut {NoStop}%
\end{thebibliography}%