R2-Router: Deterministic Rotor-Router Model

• R2-Router is a deterministic, two-direction rotor-router model on a 2D lattice that exhibits commutativity and symmetric aggregation.
• The model employs a line-by-line recurrence to update particle positions, leading to nearly semicircular, binomial-distributed occupation patterns.
• Its explicit, Abelian dynamics enable efficient computation and serve as a testbed for studying deterministic analogs of Laplacian growth and sandpile models.

The R2-Router, in the context of rotor-router models, refers to the two-direction rotor-router (RR2) process—a deterministic, Abelian aggregation model on the two-dimensional lattice with particularly tractable mathematical structure. Originating from the broader family of rotor-router processes (also called Propp machines or Eulerian walkers), RR2 is a fundamental example that elucidates key properties such as commutativity, roundness of occupied regions, and line-by-line algebraic solvability. This model serves as a bridge between probabilistically motivated growth models and explicit combinatorial descriptions, revealing deep connections between particle aggregation, binomial distributions, and discrete symmetry (Douzi, 2015).

1. Model Definition and Dynamics

The RR2 process is defined on $\mathbb{Z}^2$ restricted to "downward" directions, with all particles (or "ants") introduced at the origin. Each site is equipped with a two-state rotor, initially pointing South–East (SE). The dynamics proceed as follows:

• When an ant arrives at a site, the site's rotor advances (SE $\leftrightarrow$ SW), and the ant moves to the neighbor indicated by the new rotor direction.
• Upon first arrival to a site, an ant halts and occupies that site permanently; subsequent arrivals continue routing.
• Allowed movements are exclusively to the South–East $(i+1, n-1)$ and South–West $(i-1, n-1)$ neighbors in each step.

Over time, the set of occupied sites grows into a fan-shaped region, exhibiting remarkable regularity and symmetry.

2. Explicit Update Rules and Abelian Construction

The evolution of RR2 is governed by a line-by-line recurrence based on total throughflows $N(n,i;t)$:

$$N(n,i;t) = \left\lfloor \frac{N(n+1, i-1; t) - 1}{2} \right\rfloor + \left\lfloor \frac{N(n+1, i+1; t) - 1}{2} \right\rfloor$$

with boundary condition $N(0,0;t) = t$ and $N(n,i;t) = 0$ if $|i| > -n$. The process iterates "flooding" the lattice line by line downwards.

This explicit recurrence ensures the model possesses the Abelian property: the final pattern of occupied sites is independent of the order in which ants move. The commutativity follows from the local nature of the rotor updates and the recurrence, with the Diaconis–Fulton theory formalizing this commutative structure.

3. Symmetry, Roundness, and Binomial Law

Analytically, the RR2 model displays near-perfect horizontal symmetry: for even $t$, the difference in throughflow between symmetric sites satisfies $\leftrightarrow$0. This symmetry underpins the nearly semicircular shape of the aggregate.

In the large-$\leftrightarrow$1 regime, the occupation fractions converge to:

$\leftrightarrow$2

with $\leftrightarrow$3, recovering Pascal's triangle entries (i.e., the binomial law):

$\leftrightarrow$4

This result explains why the boundary of the RR2 cluster traces arcs determined by binomial probability level sets, and the aggregate approaches a half-disc of radius proportional to $\leftrightarrow$5.

4. Connection to Other Rotor-Router Models

RR2 is a particular restriction of the classic four-direction rotor-router (RR4, J. Propp model), wherein rotors cycle among four cardinal directions. The RR2 model admits a more transparent recurrence and line-by-line "Abelian" implementation. Nonetheless, the final occupied region for RR4 is also nearly circular, as Douzi demonstrates via a pair-of-lines sweeping method, further reinforcing the commutative and symmetric underpinnings of these processes. Fractional generalizations of both RR2 and RR4, where particles split perfectly evenly and clusters terminate at the threshold where occupation falls below one, yield occupation boundaries that deviate from the discrete model by at most a bounded amount.

5. Algorithmic Properties and Computational Features

The RR2 process is algorithmically constructed using a deterministic, parallelizable scheme. Occupancy counts $\leftrightarrow$6 can be computed efficiently by iteratively applying the explicit recurrence at each layer. The process runs in time $\leftrightarrow$7, making it practical for large system sizes and suitable for rigorous analysis of shape and fluctuation properties. The Abelian property ensures consistency across possible update orders or parallel fibers, and the entire evolution can be viewed through the lens of divisible sandpile dynamics, chip-firing games, and deterministic analogs of random walks.

6. Asymptotics, Generalizations, and Significance

For large $\leftrightarrow$8, the RR2 occupied region scales as a half-disc, with occupation probabilities tightly tracked by the binomial law. The near-symmetry and roundness are robust—even under small modifications to rotor order or splitting rules. RR2 serves as a standard model for studying:

• Deterministic analogs of Laplacian growth;
• Explicit hydrodynamic limits in the context of aggregation;
• The emergence of geometric regularity from local coordination;
• The correspondence between discrete combinatorial rules and continuum shape theorems.

The model also contextualizes the broader class of Abelian sandpile-like processes, providing a mathematical and computational testbed for questions of universality, fluctuation scaling, and convergence rates.

7. Illustrative Behavior and Visualization

Empirical simulations of RR2 with $\leftrightarrow$9 ants reveal a triangular fan of occupied sites, symmetric about the origin, with boundaries following binomial probability contours. The logarithm of occupation counts exhibits the characteristic "waves" of Pascal's triangle, and the aggregate demonstrates stability against local mesh irregularities. Visualizations for RR4 ($(i+1, n-1)$0) manifest almost perfect disks, with bounded radial error across all angles.

In summary, the R2-Router (RR2) model exemplifies deterministic, Abelian particle aggregation on discrete lattices, with explicit mathematical description, symmetry, and asymptotic roundness, and stands as a canonical example in the study of rotor-router models and related discrete processes (Douzi, 2015).