25 years of self-organized criticality

Research summary

One paper, summarised for reference. 25 Years of Self-organized Criticality: Concepts and Controversies — Nicholas W. Watkins, Gunnar Pruessner, Sandra C. Chapman, Norma B. Crosby & Henrik J. Jensen, Space Sci. Rev. (2016) 198, 3. doi:10.1007/s11214-015-0155-x.

What it does

A 25-year retrospective that untangles what Bak, Tang & Wiesenfeld (BTW) actually claimed with self-organized criticality (SOC) from the much larger set of things later attributed to it. It is the authoritative statement of SOC’s necessary and sufficient conditions and a careful, partly polemical account of the controversies — invaluable as a checklist for anyone claiming criticality from data.

Four nested perceptions

The authors separate (Fig. 1) an increasingly sweeping ladder of claims: the defensible core — self-tuned phase transitions exist in nature (some systems dynamically drive themselves to a critical point) — then the progressively wrong/unfalsifiable all fractals are SOC, all power laws are SOC, and the contingency of nature is SOC. Only the core is defended; the rest are flagged as the source of much confusion.

BTW’s core and the SOC postulate

BTW sought to unify spatial fractals and temporal $1/f$ noise as two faces of one self-organized critical state. The postulate: open, extended, dissipative systems driven slowly go automatically to a critical state — later distilled by Jensen as SDIDT (slowly driven, interaction-dominated, threshold systems). A key clarification: SOC does not mean every local degree of freedom sits at its threshold. That is true only of the (trivial) 1D sandpile; in 2D the average height is 2.125 against a threshold of 3. “Critical” means the system looks critical — long-range correlations, divergent susceptibility — not that every site is poised.

The conditions (a checklist)

• Necessary (the “phenotype”, the definition of SOC): (1) non-trivial scaling (finite-size scaling, no dependence on a tuned control parameter); (2) spatio-temporal power-law correlations; (3) apparent self-tuning to the critical point of an underlying continuous transition.
• Sufficient (the “genotype”, the mechanism): (4) non-linear interaction (usually a threshold); (5) avalanching / intermittency; (6) separation of time scales (slow drive, fast relaxation).

The clearest computational instances (the ones that pass all of this) are the Manna and Oslo models — notably, not the original BTW sandpile, whose scaling is poorly defined.

The caution that matters most

A power-law distribution is not, by itself, evidence of criticality. The literature drifted into treating power-law avalanche-size distributions as a replacement for power-law correlations, but the two are distinct: directed sandpiles show power-law avalanche sizes with zero spatial correlation, and “trivial” power laws arise from linear/dimensional-analysis mechanisms. Not everything that avalanches is critical; not every heavy tail is SOC. And observational evidence for genuine scale invariance in nature is far thinner than the “ubiquity” rhetoric suggests.

Alternatives and quasi-criticality

Several non-SOC routes reproduce apparent criticality: sweeping of an instability (Sornette); a system that merely hovers around an ordinary critical point without fine-tuning to it (precipitation — Peters–Neelin; resting-brain fMRI percolation — Tagliazucchi); and, most importantly, self-organized quasi-criticality (Bonachela–Muñoz) — when dynamics are non-conservative or timescales don’t separate perfectly, the system drifts around the critical point rather than sitting on it, still producing scale-free-ish behaviour over a few decades.

Why it’s collected here

This is the reference that keeps an avalanche analysis honest. Its necessary/sufficient checklist is a direct rubric for any criticality claim off the harness; its central caution — power-law size distribution ≠ power-law correlation ≠ criticality — is exactly the trap our cluster/avalanche measures must avoid (a fitted $P(s)\sim s^{-\tau}$ proves far less than it seems); and self-organized quasi-criticality is the precise vocabulary for the m-homeostasis question (is $m\approx1$ genuine SOC, a rate-clamp, or drift-around-the-point?). It pairs with Muñoz (living-systems criticality) and Roli (dynamical criticality) as the SOC-proper authority.

References

Watkins et al. (2016), Space Sci. Rev. 198, 3 · Bak, Tang & Wiesenfeld (1987), Phys. Rev. Lett. 59, 381 · Bak & Chen (1989), Physica D 38, 5 · Jensen (1998), Self-Organized Criticality · Pruessner (2012), Self-Organised Criticality · Vespignani, Dickman, Muñoz & Zapperi (1998), Phys. Rev. Lett. 81, 5676 · Bonachela & Muñoz (2009), J. Stat. Mech. P09009 · Manna (1991), J. Phys. A 24, L363 · Christensen et al. (1996), Phys. Rev. Lett. 77, 107 · Sornette (2006), Critical Phenomena in Natural Sciences.