Pattern Field Guide
Sixteen behavioral archetypes observed in particle life. Each arises from specific matrix conditions — none are scripted. This guide describes what to look for, what produces it, and how the real-time detector identifies it.
The four foundational archetypes
These were the first behavioral classes identified in particle life systems. They appear consistently across different implementations and have been informally described in the literature since Ventrella's Clusters (1994).
Membrane
A dense, near-stationary cluster of one species that holds its shape and resists dispersion. The simplest stable structure — particles find a potential minimum and stay there, vibrating slightly but never escaping. Looks like a living cell under a microscope when a second species forms around it.
The membrane is the most frequent and most foundational pattern. It appears in a wide range of matrices — anywhere a species strongly self-attracts or is attracted to a stationary target species. Most complex patterns (rings, proto-cells, vortices) require a membrane-like core to nucleate around.
- Speed < 0.035 — cluster is nearly stationary
- Cohesion < rMax × 1.5 — particles are tightly packed
- Hollowness < 0.30 — no empty interior
- Aspect ratio < 2.5 — roughly compact/circular shape
Chase Sequence
A compact cluster moves across the field with directed velocity, closing on another cluster. On contact, the target cluster disperses or reforms elsewhere. The chaser reassembles and hunts again. From the outside it looks exactly like predation — but the chaser has no awareness of its prey and no memory of prior hunts.
What the detector identifies is kinematic: the chaser's center-of-mass velocity is aligned with the direction toward the prey cluster. There is no representation of intent anywhere in the code.
- Two clusters identified from different species
- Dot product of A's velocity and (B.pos − A.pos) > 0.72
- This alignment must persist across multiple frames
- A is moving toward B faster than random drift
Breathing
A cluster's radius oscillates — expanding and contracting rhythmically. The period is not set by any clock. It emerges from the force balance: mutual attraction pulls particles inward, inertia carries them past equilibrium, the core's repulsion pushes them back out. A standing wave in configuration space.
The detector measures the mean distance from centroid over a rolling window, flags a breathing event when the amplitude exceeds a threshold, and applies a cooldown to prevent repeated triggers on the same oscillation.
- Centroid tracked over rolling buffer (not from origin — centroid)
- Mean distance from centroid oscillates with amplitude > threshold
- Cluster cohesion alternates between tight → loose → tight
- Periodicity confirmed across at least two full cycles
Galaxy
A rotating structure with a dense core and trailing spiral arms. The rotation isn't programmed — it's a consequence of angular momentum transferred during the asymmetric collapse of a forming cluster. As new particles fall in from off-axis trajectories, they contribute torque; the whole structure begins spinning and holds its spin through conservation.
Galaxies grow over time, absorbing nearby loose clusters into their arms. Given a large enough field, a single galaxy can eventually incorporate most of the particles of its species.
- Large cluster (top percentile by size)
- Net angular momentum > 0 (persistent spin direction)
- Aspect ratio near circular — spiral arms are dynamic, not static elongation
- Large reach parameter amplifies the gravity-like collapse
Twelve new archetypes
These twelve behavioral classes were identified through systematic exploration during the development of Emergence. They have not, to my knowledge, been formally named or computationally detected in prior particle life implementations. Each has a detector with measurable geometric criteria running in real time.
Ring Structure
A hollow annulus of particles: particles are arranged in a closed loop with an empty center. The ring is structurally stable — it holds its shape against perturbation and drifts as a unit. Unlike the membrane (which is a solid disc), a ring has a measurable void at its center.
Rings are what happens when the repulsive core is strong enough to push particles away from a central nucleus while the interaction band creates just enough cohesion to keep them looping around it. The geometry is not designed — it's the only stable orbit at that force balance.
- Hollowness score > 0.48 — significant empty interior
- Cluster size > 14 particles — enough to form a closed loop
- Aspect ratio < 2.2 — ring is circular, not an arc
- Grid centroid located from particle positions
- Empty cells in inner 40% of bounding circle counted
- Score = empty_inner / total_inner — higher = more hollow
Proto-Cell
A ring membrane of one species enclosing a different species in its interior. This two-layer structure — an inner core of species A, surrounded by a hollow ring of species B — is the closest analog to a biological cell yet observed in particle life.
Unlike a plain membrane (which may just be a shell of one species), the proto-cell requires that the interior contains a measurable cluster of a different species within the ring's radius. Both components must be stable simultaneously. This is genuinely difficult to achieve and represents a higher-order emergent structure.
- A ring structure is detected (hollowness > 0.48)
- A separate cluster of a different species found near ring's centroid
- Interior cluster's centroid is within ring.cohesion × 0.6 of ring center
- Both structures stable simultaneously for ≥ 2 detection frames
Vortex
A cluster with significant net angular momentum — particles are not just circling loosely but collectively rotating in a consistent direction, producing a visual spiral or swirl. The rotation direction (clockwise or counterclockwise) is arbitrary and determined by the initial particle positions; it persists for as long as the matrix supports the vortex.
Vortices sit at a fascinating point: they have net angular momentum but no mechanism enforcing conservation — nothing is preventing angular momentum from dissipating through damping. That they persist means the force geometry actively reinforces the rotation rather than merely tolerating it.
- |Angular momentum| > rMax × 0.025
- Hollowness > 0.30 — some interior space (not a solid clump)
- Cluster size > 12 particles
- Angular momentum = Σ (r × v) per particle, summed and divided by N
Filament
An elongated thread or chain of particles — particles strung in a line, like a polymer or a muscle fiber. Where membranes form compact discs and rings form circles, filaments form extended linear structures with high aspect ratio.
The emergence of filaments hints at the physics of polymer formation: particles find a stable one-dimensional configuration when the force curve penalizes broad clustering but rewards tight pairwise bonding. Chain-forming conditions are sensitive — most matrices that produce chains also produce membranes or vortices, and the transition between regimes is sharp.
- Aspect ratio > 3.2 — significantly elongated
- Cluster size > 10 particles
- Aspect ratio = max principal component / min principal component (PCA of particle positions)
Glider
A small, compact, coherent structure that self-propels across the field. Unlike a chaser (which follows another cluster) or a swarm (which is a moving loose group), a glider moves without an external target — it maintains its shape and velocity as a self-sustaining kinematic object.
Gliders are the rarest pattern in the simulation — comparable to the glider in Conway's Game of Life in rarity and significance. They represent a genuinely surprising property: a configuration of interacting particles that acts as a stable moving object, propagating itself through space by its own force dynamics. Finding one manually can take hours of exploration.
- Speed > 0.04 — moving consistently
- Cohesion < rMax × 1.1 — very tight cluster
- Hollowness < 0.30 — solid, not a ring
- Aspect ratio < 2.5 — compact shape
- Small size (few particles) — gliders are small by nature
- No target cluster within range (not a chaser)
Binary Orbit
Two clusters in mutual orbit — circling each other like a binary star system. The relative velocity of the two clusters is predominantly perpendicular to their separation vector, meaning they are continuously "missing" each other as they rotate around a shared center of mass.
Binary orbits are rare because they require a specific balance: strong enough mutual attraction to maintain the bond but weak enough to not cause the clusters to merge. They also require low enough damping that kinetic energy isn't bled off before the orbit stabilizes.
- Two clusters identified
- Cross-product |vrel × r̂| > 0.015 (significant perpendicular motion)
- Cross-product > along-product × 1.4 (orbit, not approach)
- Angular rate = cross / (dist + ε) — orbit frequency computed
Phase Separation
Species spontaneously segregate into macroscale territories — each color claiming its own region of the field with sharp boundaries between them. Starting from uniform mixing, the simulation self-organizes into a spatial map. The boundaries are dynamic (they move and shift as clusters grow or shrink) but the separation itself is self-sustaining.
This is the particle life analogue of oil-and-water phase separation in thermodynamics — a collective phenomenon that requires no individual particle to "know" about the global structure. Each particle just follows local forces, and the global pattern is an inevitable consequence.
- Field divided into a grid; species distribution per cell measured
- Entropy of species distribution per cell is low (each cell dominated by one species)
- Global entropy averaged across grid; separation fires when well below expected for uniform mixing
- Requires numTypes ≥ 3 — not meaningful with 1–2 species
Arc Crescent
A curved open structure — a partial ring that does not close. Where a ring is a closed loop with a hollow center, an arc is a crescent or parenthesis shape: curved, but open at both ends. The arc occupies an intermediate zone between a compact membrane and a full ring, caught at the moment where curvature exceeds width but closure has not yet occurred.
Arcs are remarkably common in energetic or high-reach fields where ring formation is interrupted by collisions or shear. A ring that loses particles from one side becomes an arc; a membrane elongating under anisotropic forces curves into one. The open ends do not collapse — they are held apart by the same internal stresses that produced the curvature.
- Aspect ratio 1.8 – 3.2 — more elongated than a disc, less than a filament
- Hollowness 0.22 – 0.48 — concave interior, but not a closed ring
- Cluster size > 10 particles
- Catches the crescent zone between membrane (compact) and ring (fully hollow)
Swarm
A compact, coherent group of particles that moves as a unit at high velocity. Unlike a membrane (nearly stationary) or a glider (very small and self-propelling), a swarm is large — a significant fraction of all particles of its species — and moves because the local force geometry gives the entire group a net drift direction that each individual particle sustains.
The swarm's coherence is purely emergent: no particle knows about the group. Each responds only to its immediate neighbors. But because velocity fields of tightly packed particles in the interaction zone are correlated, the group acquires a shared drift. The visual effect is reminiscent of murmuration — coordinated mass motion from purely local rules.
- Speed > 0.08 — fast group motion
- Cohesion < rMax × 1.3 — particles remain close
- Hollowness < 0.35 — solid group, not a ring
- Size > max(8, N × 0.04) — large enough to be a true group
Mobile Cell
A proto-cell — a ring membrane enclosing an interior cluster of a different species — that has acquired net translational velocity. The entire two-layer structure moves together through the field without losing structural integrity. The motion is not driven by an external attractor; the cell is self-propelling due to a slight internal force asymmetry that was frozen in during formation.
The mobile cell is the most complex stable object regularly produced in particle life. It encodes information (the species combination and ratio) in a structure that persists through both space and time. Its existence is what prompted calling the static version a "proto-cell" — a cell, by biological definition, moves.
- A proto-cell is detected — ring cluster enclosing a different-species interior cluster
- Ring's center-of-mass speed > 0.03 — the whole structure is in motion
- Interior cluster moves with the ring (not left behind)
- Structure must hold together across at least 2 detection frames
Mosaic Ring
Three or more distinct species arrange themselves into adjacent arc segments that collectively form a complete ring around a shared empty center. Each species holds one portion of the circumference, and the segments are stable neighbors — the boundary forces between species are balanced precisely enough that no segment expands to consume the others.
The mosaic ring is perhaps the most striking visual in particle life: a multi-colored ring assembled spontaneously from particles that have no awareness of the overall shape. It requires multiple species to simultaneously find a mutually compatible spatial arrangement — a combinatorial rarity that makes the mosaic ring exceptional whenever it occurs.
- 3+ distinct species clusters in close proximity (within rMax × 3.5)
- All clusters arranged at consistent radius from a shared centroid (low radial variance)
- Largest angular gap between any two adjacent clusters < 180°
- All participating species clusters detected simultaneously
Helix
An elongated structure whose particles are arranged in a wave or sinusoidal pattern rather than a straight line. Looking perpendicular to the major axis, particles cluster at ±offset from center — the cross-section is bimodal rather than unimodal. In three dimensions this would be a helix; in this 2D projection it appears as a sinusoidal wave chain.
The helix is a transient structure that often precedes more stable configurations. As a helical chain winds tighter, angular momentum accumulates in one direction — the helix may eventually collapse into a vortex or a galaxy. Observing the helix → galaxy transition is rare and captures a dynamical process that has no obvious analog at the level of individual particle rules.
- Aspect ratio > 3.5 — highly elongated (stricter than filament)
- Hollowness < 0.42 — not a closed ring
- Cluster size > 14 particles
- Minor-axis projection is bimodal: outer thirds hold more particles than the center third
- Each outer third > 18% of total; center third < 55% of outer sum
The pattern classifier
All sixteen detectors run simultaneously in the simulation, throttled to every 6 frames with per-pattern cooldowns to avoid event spam. Here is the shared infrastructure behind every detection.
BFS Cluster Finder
Every detection pass begins with a spatial BFS (breadth-first search) that groups nearby same-species particles into clusters. Particles within rMax of each other are connected. The algorithm runs in O(N) time using the same spatial hash grid as the physics engine.
Each cluster is then annotated with geometry: centroid, mean speed, mean distance from centroid (cohesion), hollowness score, aspect ratio (PCA), and angular momentum. These five metrics feed every downstream classifier.
- speed — mean center-of-mass velocity magnitude
- cohesion — mean distance of particles from cluster centroid
- hollowness — fraction of inner bounding area that is empty
- aspect ratio — principal component ratio (PCA on positions)
- angular momentum — Σ(r × v) / N about centroid
- Membrane — speed<0.035, cohesion<1.5r, hollow<0.30, AR<2.5
- Ring — hollow>0.48, size>14, AR<2.2
- Proto-Cell — ring + inner cluster of different species
- Vortex — |L|>0.025r, hollow>0.30, size>12
- Arc — AR 1.8–3.2, hollow 0.22–0.48, size>10
- Filament — AR>3.2, size>10
- Helix — AR>3.5, bimodal minor-axis projection
- Glider — speed>0.04, cohesion<1.1r, hollow<0.30, AR<2.5
- Swarm — speed>0.08, cohesion<1.3r, hollow<0.35, large size
- Chase — v̂·r̂ > 0.72 (velocity aligned to target)
- Orbit — cross(v_rel, r̂)>along × 1.4
- Galaxy — vortex with slow dense core enclosed inside
- Mobile Cell — proto-cell with speed>0.03
- Mosaic Ring — 3+ species clustered in ring with gap<180°
- Phase Sep. — per-cell entropy below uniform baseline
Seen a pattern that isn't named here?
These sixteen archetypes are what I've found so far — but particle life's state space is enormous. There are certainly configurations that produce structures I haven't seen yet. If you find something strange, something beautiful, or something that doesn't fit any of these categories, I genuinely want to know.