Relativistic Langevin Dynamics

• Relativistic Langevin dynamics is a framework that extends classical Langevin equations by incorporating modified kinetic energy and nonlinear friction to meet relativistic constraints.
• It models stochastic processes with relativistically consistent drag, noise, and fluctuation–dissipation relations across high-speed, energy-dense systems.
• Applications span heavy-quark transport in quark-gluon plasma, charge carrier dynamics in Dirac materials, and astrophysical plasma phenomena.

Relativistic Langevin Dynamics is the theoretical and computational framework for modeling stochastic processes involving dissipative and random forces acting on particles subject to the constraints of special (and, in some cases, general) relativity. It generalizes the classical underdamped Langevin equation to systems where particle velocities approach the speed of light, incorporating relativistically consistent friction, noise, fluctuation-dissipation relations, and, when required, a proper stochastic treatment of the underlying dynamics in phase space. Relativistic Langevin equations play a central role in modeling heavy-flavor quark transport in quark-gluon plasmas, carrier dynamics in Dirac materials, high-energy astrophysical plasmas, and relativistic molecular dynamics.

1. Core Equations and Model Formulation

The basic structure of relativistic Langevin dynamics is encapsulated by coupled stochastic differential equations for particle positions and momenta. In the Newtonian (non-relativistic) formulation for $N$ particles in $\mathbb{R}^d$, the kinetic term is quadratic and friction is linear in momentum, leading to exponential (geometric) approach to equilibrium. Relativistic generalization is performed by replacing the kinetic energy with $K_c(p) = c\sqrt{1 + |p|^2/c^2}$, whose canonical momentum dynamics reads (Duong et al., 2024):

$\begin{cases} dq_i(t) = \displaystyle\frac{p_i}{\sqrt{1+\varepsilon |p_i|^2}} dt, \[1.5ex] dp_i(t) = -\nabla_{q_i} \Bigl[ \,\sum_j U(q_j) + \sum_{j<k} G(q_j-q_k) \Bigr] dt - \nabla_{p_i} H_N^\varepsilon(p) dt + \sqrt{2} dW_i(t), \end{cases}$

with $\varepsilon = 1 / c^2$, external confining potential $U$, and singular (e.g., Coulomb) inter-particle repulsion $G$.

The associated Hamiltonian is

$$H_N^\varepsilon(q, p) = \frac{1}{\varepsilon} \sum_{i=1}^N \sqrt{1+\varepsilon |p_i|^2} + \sum_{i=1}^N U(q_i) + \sum_{i<j} G(q_i - q_j),$$

with stationary (invariant) measure given by the relativistic Maxwell–Boltzmann (Jüttner) distribution: $$\pi_N^\varepsilon(dq\,dp) = Z^{-1} \exp\left[ - H_N^\varepsilon(q, p) \right]\, dq\, dp.$$

In one-particle or simplified models, the dynamics is often written in phase space as (rest mass $M$): $$\dot{x} = \frac{p}{p^0}, \quad \dot{p} = -\gamma\,\frac{p}{p^0} - \nabla_x U + \sqrt{2D}\,\xi(t),$$ where $p^0 = \sqrt{p^2 + M^2}$, $\gamma$ is a friction coefficient, $D$ the diffusion coefficient, and $\xi(t)$ Gaussian white noise (Pal et al., 2020, 0812.1996).

Both classical and fully relativistic generalized Langevin equations (GLE) with memory kernels have also been constructed: $$\dot{p}(t) = - \int_0^t dt'\; \Gamma(t - t')\, \frac{p(t')\, c^2}{E(t')} + \xi(t), \qquad E(p) = \sqrt{p^2 c^2 + m_0^2 c^4}$$ allowing for colored noise and non-Markovian dissipation (Chen et al., 2023).

2. Ergodicity, Mixing, and Approach to Equilibrium

Relativistic Langevin dynamics displays markedly different long-time mixing properties compared to its non-relativistic counterpart. Whereas classical dissipative systems with quadratic kinetic energy and linear friction exhibit exponential convergence to equilibrium (geometric mixing), the weak high-momentum dissipation inherent to relativistic drag—where friction scales as $p / \sqrt{1 + |p|^2/c^2}$—results only in sub-geometric (polynomial) decay of correlations (Duong et al., 2024):

• For potentials $U$ of superlinear growth and singular repulsion $G$, there exists a Lyapunov function $V$ with sub-geometric drift:

$$\mathcal{L}_N V \leq -c\, V^\alpha + D, \quad \alpha \in (0,1),$$

where $\mathcal{L}_N$ is the generator. This implies algebraic mixing for the total-variation distance $W_{TV}$ to equilibrium:

$$W_{TV}\bigl(P_t^\varepsilon(X_0, \cdot), \pi_N^\varepsilon\bigr) \leq \frac{C}{(1+t)^r} V(X_0)$$

for any $r \geq 1$.

• High-momentum tails relax only polynomially due to the sublinear drag, in contrast with the exponential relaxation in the Ornstein–Uhlenbeck process of the non-relativistic setting (Duong et al., 2024, 0812.1996).
• If generalized Langevin dynamics with memory kernels are employed, ergodicity breaking and anomalous long-time behavior can occur for certain kernels, e.g., persistent memory of initial conditions ("ballistic diffusion"), equipartition violation, and nondecaying autocorrelation functions (Chen et al., 2023).

3. Fluctuation–Dissipation Relations, Stationary Measures, and Stochastic Thermodynamics

Self-consistency of the stochastic dynamics requires a relativistic fluctuation–dissipation (FDT) relation. In particular, the drift and diffusion tensors must be constructed such that the Jüttner distribution

$$f_{\text{eq}}(p) \propto \exp\left(-\frac{E(p)}{T}\right)$$

is stationary. For isotropic, scalar friction and diffusion, the FDT reads (in the post-point/Hänggi–Klimontovich scheme) (He et al., 2013, 0812.1996): $$\Gamma(p) = \frac{1}{E_p} \left[ \frac{D(E_p)}{T} \right]$$ with the friction coefficient $\Gamma(p)$, diffusion coefficient $D(E_p)$, and energy $E_p = \sqrt{p^2 + M^2}$. Different stochastic integration prescriptions (Ito, Stratonovich, Hänggi–Klimontovich) induce corresponding drift corrections to guarantee the correct equilibrium.

Stochastic thermodynamics for relativistic Brownian motion proceeds with generalized definitions of work and heat (Pal et al., 2020): $$dE = dW + dQ = \frac{\partial U}{\partial a}\,\dot{a}\,dt + \left[ -\gamma\,\frac{p^2}{p^0} + \sqrt{2D}\frac{p}{p^0}\xi(t) \right] dt$$ and supports both integral and detailed fluctuation theorems,

$$\left\langle e^{-\Delta S_{\text{tot}}/k_B} \right\rangle = 1,\qquad \frac{P_F[\Gamma]}{P_R[\tilde{\Gamma}]} = \exp\bigl(\Delta S_{\text{tot}}[\Gamma]/k_B\bigr),$$

with modifications arising from feedback protocols and the fundamentally relativistic time-delay ("absolute irreversibility") in non-equilibrium processes.

4. Non-Markovian Generalizations and First-Principles Derivation

A rigorous, Lorentz-covariant particle-bath Lagrangian provides the foundation for fully consistent relativistic generalized Langevin equations. For example, via a relativistically invariant extension of the Caldeira–Leggett model and elimination of the bath degrees of freedom, one arrives at dynamics (Zadra et al., 2023, Petrosyan et al., 2021): $$\frac{dp^\mu}{dt} = -F_{\text{ext}}^\mu + F_p^\mu(t) - \frac{1}{m}\int_0^t K^\mu{}_\nu(t,s) p^\nu(t-s) ds$$ where $K^\mu{}_\nu$ is a non-Markovian memory kernel (causal, Lorentz-covariant), and the random force $F_p^\mu(t)$ is drawn from a noise process obeying a fully nonlinear FDT.

In the translation-invariant version (Zadra et al., 2023), the key step is the introduction of a renormalization potential $\Phi$ that eliminates the spurious restoring force and ensures no symmetry breaking.

Such fully covariant approaches have also been realized for quantum Brownian motion, leading to operator-valued relativistic Langevin equations within a minimal-coupling formalism (Amooghorban et al., 2013).

5. Numerical Implementation and Practical Algorithms

Relativistic Langevin equations are implemented numerically in both static and expanding media, with careful attention to the discretization scheme, Lorentz covariance, and consistent realization of the FDT (Alberico et al., 2010, Li et al., 2020, Li et al., 2019):

• Updates are performed in the local rest frame of the background medium (cell-by-cell for evolving backgrounds), with momenta and positions transformed according to the current local temperature and flow velocity.
• Drag ($\eta_D$) and diffusion ($\kappa_{L,T}$) coefficients are computed from perturbative QCD cross sections for heavy-ion applications, or from spectral bath properties for more general models.
• The stochastic integration convention (Ito, Stratonovich, Hänggi–Klimontovich) must be consistently applied to maintain physical equilibrium, with observable impact on late-time distributions and convergence.
• In non-Markovian (memory-ful) dynamics, colored noise is generated by exact algorithms to match the autocorrelation structure of the desired kernel.

Simulations yield predictions for key observables such as the nuclear modification factor $R_{AA}$, elliptic flow coefficient $v_2$, and mean squared displacement, enabling direct comparison with experimental data and extraction of transport coefficients such as the spatial diffusion constant $2\pi T D_s$ (Li et al., 2020).

6. Applications, Physical Consequences, and Open Problems

Relativistic Langevin dynamics has been widely applied in the modeling of:

• Heavy quark (charm, beauty) propagation, energy loss, and collective flow (e.g., $R_{AA}$ and $v_2$) in quark-gluon plasma created in heavy-ion collisions (Li et al., 2020, Alberico et al., 2010, Li et al., 2019).
• Relativistic electron and hole transport in Dirac materials, notably graphene, where semiclassical Langevin dynamics is validated through ultrafast measurement of single-particle trajectories (Pal et al., 2020).
• Hot, dense plasmas and astrophysical systems where stochastic dissipation and noise must be handled in a manifestly relativistic and, when necessary, covariant manner.
• Quantum stochastic processes for dissipative Dirac fermions in anisotropic and complex media (Amooghorban et al., 2013).

Persistent nonequilibrium effects—ballistic transport, memory, ergodicity violation—arise in generalized frameworks, particularly in heavy-quark motion in QGP if the microscopic kernel deviates from purely exponential forms (Chen et al., 2023). Recent works have also highlighted conceptual challenges in general relativistic stochastic dynamics, including the correct choice of parametrization (proper time of the particle vs observer), covariance, and the relationship between stochastic evolution and physically measurable observer-frame probability distributions (Cai et al., 2023).

Conceptual and technical open questions include uniform ergodic rates across relativistic and Newtonian limits; consistent treatment of multiplicative noise; extension to curved spacetimes with general covariance; direct quantum generalizations; and systematic first-principles derivations from underlying field theories (Duong et al., 2024, Petrosyan et al., 2021, 0812.1996).

7. Comparative Properties: Relativistic vs. Nonrelativistic Langevin Dynamics

A summary comparing key aspects of the classical and relativistic Langevin equations is presented below.

Feature	Nonrelativistic (Classical)	Relativistic
Kinetic Energy	$\frac{1}{2}|p|^2$	$c \sqrt{1+|p|^2/c^2}$
Friction Law	$-p$	$-p / \sqrt{1 + |p|^2/c^2}$
Equilibrium	Boltzmann (Maxwell)	Jüttner (relativistic Maxwell–Boltzmann)
Mixing Rate	Exponential (geometric)	Algebraic (polynomial)
FDT Structure	Einstein, linear in $p$	Modified, depends on $E_p$
Noise/Friction	Linear, scalar	Nonlinear, potentially tensorial, nonlocal
Covariance	Galilean	Lorentz (when formulated covariantly)

A central distinction is the observation that the high-momentum sector in the relativistic equation is only weakly dissipative, which fundamentally limits the ergodicity and rate of approach to equilibrium (Duong et al., 2024). In the Newtonian limit $c \to \infty$, the relativistic Langevin model reduces smoothly to the classical underdamped Langevin process.

---

In conclusion, relativistic Langevin dynamics provides a rigorous and physically consistent stochastic framework for systems where relativistic effects, energy-momentum constraints, and fluctuation-dissipation balance are essential. Recent advances have yielded fully covariant formulations, non-Markovian generalizations, and practical algorithms for computational and theoretical exploration of a range of phenomena from QCD transport to quantum stochastic processes in high field environments (Duong et al., 2024, Li et al., 2020, Petrosyan et al., 2021, Zadra et al., 2023, Pal et al., 2020).