--- title: Lévy-Type Mean-Field Systems url: https://www.emergentmind.com/topics/levy-type-mean-field-systems type: topic --- # Lévy-Type Mean-Field Systems Lévy-type mean-field systems are mean-field models in which collective interaction is coupled to Lévy-type nonlocality, either because the representative dynamics is governed by a Lévy–Khintchine generator depending on the population law, because the state equation is driven by Brownian motion plus a Poisson random measure or by additive Lévy noise, or because a deterministic long-range mean-field system displays effective truncated Lévy statistics in suitably coarse-grained variables [1112.3744; 2503.13835; 2508.02224; 2604.12317; 1208.4878]. Across these formulations, the mean field may enter the drift, diffusion matrix, jump kernel, cost functional, or equilibrium consistency relation, while the Lévy-type structure appears as a nonlocal generator, a compensated Poisson random measure, a general Lévy process, or a heavy-tailed continuous-time random walk description. The surrounding literature also contains adjacent frameworks—switching diffusions, adaptive networks with memory, hypergraph Vlasov limits, and Tanaka-style non-exchangeable additive-noise systems—that are structurally relevant but not, as stated, genuine Lévy-type models [1901.05631; 2507.21312; 2406.04691; 2510.04894]. ## 1. Conceptual scope and taxonomy Within the corpus, the most direct meaning of “Lévy-type mean-field system” is a mean-field evolution whose infinitesimal dynamics contains a Lévy–Khintchine jump term or whose stochastic forcing is specified through a Lévy process or a compensated Poisson random measure. In this sense, the topic includes controlled nonlinear Markov processes with measure-dependent Lévy kernels [1112.3744], semigroup-based mean-field generators with drift, diffusion, and jump parts [2508.02224], and distribution-dependent SDEs of the form \[ dX_t=b(t,X_t,\mathscr L_{X_t})\,dt+dL_t, \] with additive Lévy noise [2604.12317]. A second, narrower but still direct strand consists of optimization and game formulations. Here the mean field enters through \(\mathbb E[X]\), \(\mathbb E[u]\), or a stationary scalar statistic, while the noise is Brownian plus jump or purely Lévy. This includes stochastic linear-quadratic control with compensated Poisson random measures [2503.13835] and mean-field games with two-sided singular controls for one-dimensional Lévy processes [2505.22936]. A third usage is effective rather than microscopic. The Hamiltonian Mean Field model is a deterministic long-range interacting rotor system, but its momentum-space trapping statistics are well described by a one-sided truncated Lévy distribution, so the system is “Lévy-type” at the level of coarse-grained sojourn times and continuous-time-random-walk interpretation, not at the level of a Lévy-driven state equation [1208.4878]. | Class | Defining structure | Representative reference | |---|---|---| | Generator-level MFG | Measure-dependent Lévy–Khintchine operator | [1112.3744] | | Semigroup mean-field model | \(A(\mu)=A^\nabla(\mu)+A^\Delta(\mu)+A^J(\mu)\) | [2508.02224] | | DDSDE | Additive Lévy noise with law-dependent drift | [2604.12317] | | Control and MFG | Poisson-jump LQ control; singular control of Lévy processes | [2503.13835], [2505.22936] | | Effective Lévy dynamics | Truncated Lévy sojourn statistics in momentum space | [1208.4878] | A common misconception is to identify every nonlocal or hybrid mean-field model with a Lévy-type one. The cited literature draws a sharper boundary: path-dependent memory on \(C([0,T],\mathbb R^d)\), Markovian regime switching, and higher-order hypergraph interactions are nonlocal in time, environment, or interaction architecture, but not Lévy in the sense of jump generators or càdlàg state dynamics as formulated in those works [2507.21312; 1901.05631; 2406.04691]. ## 2. Generator and equation-level formulations The most general generator-level formulation in the corpus appears in the mean-field game setting with nonlinear Markov processes. For class \(i\), the controlled one-particle generator is \[ A^i[t,\mu,u]f(z)=\big(h_i(t,z,\mu,u),\nabla f(z)\big)+L_i[t,\mu]f(z), \] where \[ L_i[t,\mu]f(z) = \frac12\big(G_i(t,z,\mu)\nabla,\nabla\big)f(z) + \big(b_i(t,z,\mu),\nabla f(z)\big) + \int_{\mathbb R^d} \Big( f(z+y)-f(z)-(\nabla f(z),y)\mathbf 1_{B_1}(y) \Big)\,\nu_i(t,z,\mu,dy). \] The mean-field law enters the drift, diffusion coefficient, and Lévy kernel, and the forward equation is the weak kinetic equation \[ \frac{d}{dt}(g,\mu_t)=\big(A[t,\mu_t,\gamma(t,\cdot)]g,\mu_t\big), \] coupled to a backward HJB equation [1112.3744]. A closely related but more explicitly transport-based formulation is the Lévy-type mean-field generator on \(\mathbb R^d\), \[ A(\mu)=A^\nabla(\mu)+A^\Delta(\mu)+A^J(\mu), \] with \[ A^\nabla(\phi;\mu)(x)=b(x,\mu)\cdot\nabla\phi(x), \] \[ A^\Delta(\phi;\mu)(x)=\frac12\sum_{i,j=1}^d a_{ij}(x,\mu)\partial_{ij}\phi(x), \] \[ A^J(\phi;\mu)(x) = \int_{\mathbb{R}^d\setminus\{0\}} \big[\phi(x+y)-\phi(x)-y\cdot\nabla\phi(x)\big]\Lambda(x,dy,\mu). \] The associated nonlinear mean-field evolution is \[ \partial_t\mu_t=\mu_tA(\mu_t), \qquad \frac{d}{dt}\langle \mu_t,\phi\rangle = \int_{\mathbb R^d} A(\phi;\mu_t)(x)\,\mu_t(dx), \] and McKean–Vlasov diffusion is recovered by setting \(\Lambda\equiv0\) [2508.02224]. At the SDE level, the additive-noise McKean–Vlasov formulation is \[ dX_t=b(t,X_t,\mathscr{L}_{X_t})\,dt+dL_t, \] where \(L\) is a \(d\)-dimensional Lévy process with Lévy–Itô decomposition \[ L_t = \sigma W_t +\int_0^t\int_{0<|x|\le 1}x\,\tilde N(ds,dx) +\int_0^t\int_{|x|>1}x\,N(ds,dx). \] Its generator is \[ \mathcal{L}u(t,x)= \frac{1}{2}\operatorname{tr}\left(\sigma\sigma^{\top}\nabla^2_{x} u(t,x)\right) +\int_{\mathbb{R}^d_*} \Big(u(t,x+z)-u(t,x)-\mathbf{1}_{\{|z|\le 1\}}\langle z,\nabla u(t,x)\rangle\Big)\,\nu(dz), \] so the law dependence is confined to the drift and the Lévy nonlocality is additive and law-independent [2604.12317]. These formulations show that “mean field” and “Lévy-type” interact at two distinct levels. The first is coefficient dependence on \(\mu_t\). The second is the analytic structure of the generator, which may include nonlocal jump terms. The literature represented here treats both simultaneously. ## 3. Well-posedness, semigroup stability, and propagation of chaos The semigroup-based framework in the Lévy application of the abstract mean-field theory is organized around a mean-field generator \[ A:P_2(\mathbb R^d)\to G_2^\Lambda(\mathbb R^d), \] together with a transport cost \(c_2(x,y)=\frac12|x-y|^2\). A central device is the generator metric \[ W_G(A,B)^2 = \frac12|b-\tilde b|^2 + W_S(a,\tilde a)^2 + W_\Lambda(\Lambda,\tilde\Lambda)^2, \] where \(W_S\) is the Bures–Wasserstein distance on covariance matrices and \(W_\Lambda\) is a Lévy–Wasserstein transport cost between Lévy measures. Under the Lipschitz criterion \[ W_G(A(x,\mu),A(y,\nu))^2 \le \frac{\beta}{2}|x-y|^2+\alpha\,\Sigma(x,\mu,\nu)^2, \] the nonlinear mean-field problem has a unique \(c_2\)-stable solution and the associated \(N\)-particle system satisfies pointwise propagation-of-chaos bounds in \(W_2\) [2508.02224]. The same work derives the estimate \[ \sup_{t\in[0,T]}W_2^2(\rho_t^N,\mu_t^{\otimes N}) \le W_2^2(\rho_0^N,\mu_0^{\otimes N})e^{KT} + C\zeta_K(T)\sup_{t\in[0,T]}\aleph_N(\mu_t;\Sigma^2), \] and, in an average-form Lévy-driven model, the sharper rate \[ \sup_{t\in[0,T]}W_2^2(\rho_t^N,\mu_t^{\otimes N}) \le W_2^2(\rho_0^N,\mu_0^{\otimes N})e^{KT} +\frac{C\zeta_K(T)M_T}{N-1}. \] This makes the propagation-of-chaos problem reducible to empirical-measure approximation under the chosen modulus \(\Sigma\). The paper explicitly treats McKean–Vlasov diffusion as a special case, but its novelty lies in the jump part and in the generator-level treatment of drift, diffusion, and Lévy kernel within a unified metric [2508.02224]. Weak existence under low regularity is addressed for additive Lévy-driven DDSDEs by a different route. The key analytic estimate is a Krylov-type bound: for semimartingales of the form \[ X_t=X_0+\int_0^t \xi_s\,ds+L_t, \] one has \[ \mathbb E\left( \int_{S\wedge \tau}^{T\wedge \tau} f(s,X_s)\,ds \Bigm| \mathcal F_S \right) \le C\left[ 1+\mathbb E\left( \int_{S\wedge \tau}^{T\wedge \tau} |\xi_s|\,ds \Bigm|\mathcal F_S \right) \right] \|f\|_{L^q([S,T];L^p(\mathbb R^d))}. \] Using this estimate, approximation by bounded Wasserstein-Lipschitz drifts \(b^n\), Aldous tightness, Prokhorov compactness, and Skorokhod representation, weak solutions are obtained for \[ dX_t=b(t,X_t,\mathscr L_{X_t})\,dt+dL_t \] under \(L_t^qL_x^p\)-type drift control and semigroup assumptions on the Lévy process [2604.12317]. The nonlinear Markov process approach to mean-field games furnishes a parallel law-of-large-numbers statement. The \(N\)-particle generator acting on measure functionals admits an expansion \[ \widehat A_t^N F(\mu)=\Lambda_tF(\mu)+O(1/N), \] more precisely with a second-variational correction of order \(1/N\), and the resulting approximation of the limiting kinetic flow has rate \(1/N\). This rate subsequently feeds into the game-theoretic \(\varepsilon(N)\)-equilibrium estimate [1112.3744]. The analytical picture is therefore stratified. Semigroup transport methods yield stability and chaos estimates when the Lévy-type generator is already available. Krylov estimates and tightness establish weak existence under singular drift in additive-noise models. Generator expansions and nonlinear martingale problems control mean-field limits in game settings. ## 4. Control, optimization, and mean-field games with jumps In stochastic control, the clearest direct Lévy-type mean-field model in the corpus is the general mean-field stochastic linear-quadratic problem with random coefficients and jump noise. The state equation is \[ \left\{ \begin{aligned} dX(t) &= \Big(A(t)X(t)+A_1(t)\mathbb E[X(t)] + B(t)u(t)+B_1(t)\mathbb E[u(t)]\Big)\,dt \\ &\quad + \Big(C(t)X(t)+C_1(t)\mathbb E[X(t)] + D(t)u(t)+D_1(t)\mathbb E[u(t)]\Big)\,dW(t) \\ &\quad + \int_{U_0}\Big(\alpha(t,z)X(t-)+\alpha_1(t,z)\mathbb E[X(t-)]+ \beta(t,z)u(t)+\beta_1(t,z)\mathbb E[u(t)]\Big)\,\tilde N(dt,dz), \end{aligned} \right. \] and the quadratic cost depends on \(X\), \(\mathbb E[X]\), \(u\), and \(\mathbb E[u]\). The optimal control is characterized in explicit state-feedback form by a stochastic Riccati equation with jumps, an auxiliary linear BSDE with jumps, and a deterministic operator equation for the mean processes. The final control law is \[ u^*=\Theta^{-1}N\,X^*+\Theta^{-1}M, \] where \(\Theta\), \(N\), and \(M\) are built from the coefficients, the Riccati solution, and the affine BSDE terms [2503.13835]. In mean-field games, the representative-agent generator may itself be Lévy–Khintchine. The HJB equation takes the form \[ \frac{\partial V_i}{\partial t}(t,x) + H^i(x,\nabla V_i(x),\mu_t) + L_i[t,\mu_t]V_i(t,x)=0, \] with \[ H^i(x,p,\mu_t) = \max_{u\in U}\{h_i(t,x,\mu_t,u)p+J_i(t,x,\mu_t,u)\}, \] and the forward consistency equation is \[ \frac{d}{dt}(g,\mu_t) = \big(A[t,\mu_t,\Gamma(t,\cdot,\{\mu_{\ge t}\})]g,\mu_t\big). \] For \(K\) classes of agents, the population law is vector-valued, \(\mu_t=(\mu_{1,t},\dots,\mu_{K,t})\), and the coupling acts through the empirical measure. Under the stated smoothing, Lipschitz, and variational-differentiability assumptions, the mean-field solution produces a perfect \(\varepsilon\)-equilibrium with \[ \varepsilon(N)=\frac{C(T)}{N} \left( \|J\|+\|V^T\|_{C^2}+1 \right). \] This is a generator-level generalization from diffusion models to jump-diffusions, stable-like processes, pure jump processes, and other Lévy-type Feller dynamics [1112.3744]. A separate singular-control strand studies a one-dimensional Lévy process controlled by two-sided bounded-variation controls: \[ X_t^{U,D}=X_t+U_t-D_t. \] The mean-field coupling is not through the full law but through the scalar statistic \[ p^\eta=\mathbb E[f(X^\eta)], \] with discounted and ergodic criteria built from a running cost \(c(x,p^\eta)\) and proportional control costs \(q_u,q_d\). For fixed \(p\), the optimal control is reflection on an interval \([a^*_{\varepsilon,p},b^*_{\varepsilon,p}]\), characterized by an adjoint Dynkin game and the barrier identities \[ a^*_{\varepsilon,p}=\sup\{x:V_{\varepsilon,p}(x)=-q_u\}, \qquad b^*_{\varepsilon,p}=\inf\{x:V_{\varepsilon,p}(x)=q_d\}. \] Discounted and ergodic MFG equilibria are obtained by Brouwer fixed points; equilibrium controls are of double-reflection type; an Abelian limit connects discounted and ergodic equilibria; and nonuniqueness is explicitly exhibited by counterexample [2505.22936]. These control and game formulations show that Lévy-type mean-field systems are not restricted to passive evolution equations. They also support explicit feedback synthesis, free-boundary characterizations, random-measure equilibria, and \(1/N\)-approximation statements. ## 5. Emergent Lévy statistics, CTRW structure, and weak ergodicity breaking A different manifestation of the topic appears in the Hamiltonian Mean Field model \[ H(p,\theta)=\frac{1}{2}\sum_{i=1}^N p_i^2+\frac{1}{2N}\sum_{i,j=1}^N\left[1-\cos(\theta_i-\theta_j)\right]. \] This is a fully coupled long-range rotor system, mean-field because each rotor interacts equally with every other one under the \(1/N\) Kac normalization. In the \(N\to\infty\) limit the dynamics is governed by the Vlasov equation, while at finite \(N\) fluctuations act as weak collisions [1208.4878]. The key observation is that, after coarse-graining momentum space into cells of width \(\Delta p\), particle motion can be interpreted as a continuous-time random walk between cells. If \(t_k^{(j)}\) is the \(j\)-th sojourn time in cell \(k\) and \[ g(t)=\frac{1}{J}\sum_{j=1}^J\delta\left(t-t^{(j)}\right), \] then the empirical sojourn-time law is well described by a truncated one-sided Lévy distribution, \[ \tilde{f}^{(\alpha)}(t)= \begin{cases} c f^{(\alpha)}(t), & 0