Filippov trajectories and clustering in the Kuramoto model with singular couplings (Q2039603)

The authors consider the Kuramoto model with weighted coupling structure \( \overset{.}{\theta }_{i}=\Omega _{i}+\frac{K}{N}\sum_{j=1}^{N}\Gamma (\theta _{j}-\theta _{i})\sin (\theta _{j}-\theta _{i})\), where \(\Omega _{i}\) are the natural frequencies of the oscillators, \(K\) is the constant coupling strength, \(\Gamma \) is a function which may take different expressions among which \(\Gamma (\theta )=\frac{1}{c_{\alpha ,\zeta }^{\alpha }\left\vert \theta \right\vert _{o}^{2\alpha }}\), \(\left\vert \theta \right\vert _{o}\) being equal to \(\left\vert \overline{\theta }\right\vert \) with \(\overline{ \theta }=\theta \) mod \(2\pi \), \(\overline{\theta }\in (-\pi ,\pi ]\), and \( c_{\alpha ,\zeta }\) is an appropriate constant. Such models intend to simulate the behavior of biological systems in which synchronization, adaptation and cooperation occur. Introducing \(h(\theta )=\Gamma (\theta )\sin (\theta )\), the above model can be written as \(\overset{.}{\theta } _{i}=\Omega _{i}+\frac{K}{N}\sum_{j=1}^{N}h(\theta _{j}-\theta _{i})=H_{i}(\Theta )\), for which the authors derive a second-order augmentation of the Kuramoto model \(\overset{.}{\theta }_{i}=\omega _{i}\), \( \overset{.}{\omega }_{i}=\frac{K}{N}\sum_{j=1}^{N}h^{\prime }(\theta _{j}-\theta _{i})(\omega _{j}-\omega _{i})\). The authors also consider a gradient flow structure and a kinetic formulation associated with this Kuramoto model. The authors prove an equivalence result between these two formulations. Choosing \(h(\theta )=\frac{\sigma ^{2\alpha }\sin \theta }{ (\sigma ^{2}+c_{\alpha ,\zeta }\left\vert \theta \right\vert _{o}^{2})^{\alpha }}\), the scaling \(\sigma =\mathcal{O}(\varepsilon )\), \( K\alpha ^{2\alpha }=\mathcal{O}(1)\) as \(\varepsilon \rightarrow 0\) and the change of variables \(\sigma \rightarrow \varepsilon \), \(K\rightarrow K\varepsilon ^{-2\alpha }\), the authors end with the scaled system \(\overset{ .}{\theta }_{i}=\Omega _{i}+\frac{K}{N}\sum_{j=1}^{N}h_{\varepsilon }(\theta _{j}-\theta _{i})\) with \(h_{\varepsilon }(\theta )=\frac{\sin \theta }{ (\varepsilon ^{2}+c_{\alpha ,\zeta }\left\vert \theta \right\vert _{o}^{2})^{\alpha }}\), and they observe that the formal limit of \( h_{\varepsilon }\) is \(h(\theta )=\frac{\sin \theta }{c_{\alpha ,\zeta }^{\alpha }\left\vert \theta \right\vert _{o}^{2\alpha }}\), leading to the singular problem \(\overset{.}{\theta }_{i}=\Omega _{i}+\frac{K}{N} \sum_{j=1}^{N}\frac{\sin (\theta _{j}-\theta _{i})}{\left\vert \theta _{j}-\theta _{i}\right\vert _{o}^{2\alpha }}\). The kernel \(h\) is continuous for \(\alpha \in (0,1/2)\), the subcritical case, has a jump discontinuity at 0 for \(\alpha =1/2\), the critical case, and shows essential discontinuities in the supercritical case \(\alpha \in (1/2,1)\). In the subcritical case, the authors prove the existence of a global-in-time strong solution for any initial configuration. This solution is unique forwards in time. They also analyze the situation where two oscillators collide at some time. In the critical case, the authors introduce the Filippov set-valued map \(\mathcal{F} :\mathbb{R}^{N}\rightarrow 2^{\mathbb{R}^{N}}\) defined as \(\mathcal{F} (x)=\cap _{\delta >0}\cap _{\left\vert \mathcal{N}\right\vert =0}\overline{co }(F(B_{\delta }(x)\setminus \mathcal{N}))\) where \(\left\vert \mathcal{N} \right\vert \) is the Lebesgue measure of any measurable set \(\mathcal{N} \sqsubseteq \mathbb{R}^{N}\) and \(\overline{co}(A)\) is the closure of the convex hull of \(A\). The main result is here an existence of a global-in-time Filippov solution for any initial configuration, this solution being unique forwards in time. The authors again analyze the case where two oscillators collide at some time. In the supercritical case, the authors prove that if \( \Theta \) is a classical solution which is defined on a maximal time interval \([0,t^{\ast })\), there exists \(\varepsilon >0\) such that the classical trajectory \(t\mapsto \Theta (t)\) can be continued to a Filippov solution in \( [t^{\ast },t^{\ast }+\varepsilon )\) such that oscillators belonging to the same cluster of the collision state \(\Theta ^{\ast }\) remain stuck after \( t^{\ast }\). The authors then justify the formal limit taken when letting \( \varepsilon \) go to 0. In the final part of their long paper, the authors analyze the synchronization process of the singular weighted system they introduced. They again distinguish the situation according to the critical case \(\alpha =1/2\).