Diffusion-Based Contact Process

• Diffusion-based contact processes are interacting particle systems on a lattice that combine stochastic diffusion with multi-particle contact reactions.
• They exhibit non-equilibrium phase transitions with critical decay laws and significant finite-time corrections that challenge mean-field predictions.
• Advanced GPU-based Monte Carlo simulations reveal slow crossovers and scaling behaviors potentially aligning with directed percolation universality.

A diffusion-based contact process is a class of interacting particle systems in which particles occupying sites on a lattice (typically one-dimensional) can undergo both stochastic "contact" reactions—such as creation and annihilation, often requiring multi-site clusters (pairs, triplets)—and diffusive motion (hopping to neighboring sites). These models serve as paradigmatic examples for non-equilibrium phase transitions to absorbing states, the study of universality classes in stochastic processes, and the role of fluctuations and spatial correlations in reaction-diffusion kinetics.

1. Formal Definition of Diffusion-Based Contact Processes

A prototypical diffusion-based contact process is defined on a one-dimensional chain of $N$ lattice sites, each site $i$ with occupation variable $\sigma_i \in \{0,1\}$ (1: occupied by a fermionic particle $A$, 0: empty). The dynamics is governed by simultaneous stochastic rules:

• Diffusion (rate $d$): $A\,\emptyset \to \emptyset\,A$ and $\emptyset\,A \to A\,\emptyset$—single-particle hopping to neighboring sites.
• Contact Reactions:
  • For the Triplet Contact Process with Diffusion (TCPD) (Schram et al., 2013):
  • Triplet Annihilation (rate $p(1-d)$): $A\,A\,A \to \emptyset\,\emptyset\,\emptyset$
  • Triplet Branching (rate $(1-p)(1-d)/2$): $A\,A\,A\,\emptyset \to A\,A\,A\,A$ and $\emptyset\,A\,A\,A \to A\,A\,A\,A$
• Update Rule: In each infinitesimal time step $\Delta t$, one event is chosen in proportion to its total rate, and the clock is incremented by $\Delta t = 1/N$.

Order parameters are typically the global particle density $\rho(t) = \mathbb{P}$(site occupied) and the density of neighboring pairs $\rho_p(t) = \mathbb{P}$(neighboring sites both occupied).

2. Mean-Field Theory and Analytical Predictions

In spatially homogeneous mean-field theory, one neglects all correlations beyond global densities:

• Closure Approximations: $\rho_p \approx \rho^2$, triplet probability $\approx \rho^3$
• TCPD Rate Equation:

$$\frac{d\rho}{dt} \approx -p(1-d)\rho^3 + (1-p)(1-d)\rho^3(1-\rho)$$

At criticality ($p=p_c$), neglecting subleading terms, balance of creation/annihilation gives $\rho(t) \sim t^{-\delta}$, with

• For TCPD: $\delta = 1/3$ (cubic nonlinearity, due to triplet reactions)
• For pair processes (PCPD): $\delta_{MF} = 1/2$ (quadratic nonlinearity)
• Pair densities: $\rho_p(t) \sim t^{-2\delta}$ under mean-field scaling

3. Simulation Methodologies and Effective Exponent Analysis

Large-scale numerical studies utilize massively parallel multispin GPU-based Monte Carlo techniques (Schram et al., 2013):

• System parameters: Diffusion rate (e.g., $d=0.5$), lattice size $L \sim 2^{21}$, times up to $T=8\times10^8$, with hundreds of independent replicas.
• Criticality Detection: Binary search in $p$; supercritical runs plateau, subcritical show exponential decay; at $p \approx p_c$, the effective exponent $\delta_{eff}$ drifts slowly but is approximately constant over many decades.

Observables and analysis methods:

• Particle Density $\rho(t)$ and Pair Density $\rho_p(t)$: Measured as a function of time, with raw curves indicating power-law or curved decay.
• Effective Exponents: $\delta_{eff}(t) = -d\ln\rho/d\ln t$, used for drift and scaling corrections.

4. Numerical Results and Critical Behavior

The one-dimensional TCPD exhibits:

• Decay Laws: Particle density $\rho(t) \sim t^{-0.33}$ at intermediate times; pair density $\rho_p(t)$ shows stronger curvature, indicating significant finite-time corrections.
• Exponents: $\delta_{eff}$ drifts below the mean-field value $1/3$ to $\lesssim0.32$ for the longest accessible times, implying mean-field theory is incorrect in 1D. The upper bound is $\delta < 0.32$ for TCPD (Schram et al., 2013).
• Pair-to-Site Ratio: The ratio $\rho_p/\rho$ extrapolates to a constant $>0$, rather than decaying as $t^{-\delta}$ as mean-field would predict.

These results confirm that spatial fluctuations and correlations, neglected in mean-field, fundamentally alter the scaling in low dimensions.

5. Corrections to Scaling, Crossover, and Universality

Anomalously slow crossovers and strong corrections to scaling characterize all diffusion-based $n$-tuple contact processes in one dimension:

• In TCPD, even at $t \sim 10^8$, the system does not reach the true asymptotic regime; $\delta_{eff}(t)$ continues to drift.
• The pair exponent $\delta_{p,eff} \neq 2\delta_{eff}$, in contrast to mean-field expectations, further attests to the breakdown of factorization and importance of spatial correlations.
• The possibility remains that the ultimate scaling is governed by the Directed Percolation (DP) universality class (with exponent $\delta_{DP} \approx 0.1595$), but the crossover is so slow that numerically distinguishing TCPD from DP on accessible timescales is hazardous.

6. Comparison to Related Diffusion-Based Contact Models

All major diffusion-based contact processes in 1D share qualitative phenomenology: | Model | Essential Reactions | Mean-Field Exponent $\delta_{MF}$ | Numerical Results / Universality | |---------------------------|-----------------------------|------------------------------|---------------------------------------| | Directed Percolation (DP) | $A \to 2A,\;A \to \emptyset$ | $1$ (field-theoretic) | $\delta_{DP} \approx 0.1595$, robust | | Pair CP with Diffusion | $2A \to 3A,\;2A \to \emptyset$ | $1/2$ | $\delta_{PCPD} \sim 0.16-0.35$, slow crossovers, possible DP universality | | Triplet CP with Diffusion | $3A \to 4A,\;3A \to \emptyset$ | $1/3$ | $\delta_{TCPD}<0.32$, extremely slow crossover, possible DP scaling (Schram et al., 2013) |

• All models show strong finite-size and finite-time corrections, nontrivial ratio $\rho_p/\rho \to const$, and a slow restoration (if any) of DP exponents.
• No convincing evidence supports a new universality class for TCPD or PCPD in 1D; most data are compatible, albeit with extreme corrections, with DP scaling (Schram et al., 2013).
• The safest hypothesis is that both TCPD and PCPD ultimately cross over to DP universality as $t \rightarrow \infty$.

7. Significance and Open Problems

Diffusion-based contact processes serve as critical testbeds for:

• Absorbing-State Phase Transitions: The study of how microscopic processes, dimensionality, and mobility influence the nature and universality of non-equilibrium critical points.
• Limits of Mean-Field Theory: Demonstrating that neglecting fluctuations leads to qualitatively incorrect predictions in low dimensions, and that pair and higher-order correlations control scaling laws.
• Scaling, Crossovers, and Universality: Exposing regimes in which conventional universality classes (e.g., DP) may eventually dominate but are obscured by very strong corrections, challenging both simulation and analytical approaches.
• Computational Methodology: Driving advances in nondeterministic massively parallel computing (e.g., GPU-based multispin simulations) to probe such extreme time/size regimes.

Key open problems include the analytical computation of scaling corrections, the development of improved estimators to accelerate the approach to asymptotic scaling, and the rigorous classification—or refutation—of potential novel universality classes in reaction-diffusion systems with mobility and multi-particle contact reactions (Schram et al., 2013).