25 Years of Self-Organized Criticality: Concepts and Controversies

Scope and influence

• SOC has influenced fields from statistical mechanics and seismology to plasma physics, neuroscience, and sociology, while attracting over 6600 citations since 1987.The paper complements reviews of laboratory fusion, solar, and magnetospheric plasma applications (Aschwanden et al., 2014).Source: §1 · p. 2
• Despite its success, SOC remains controversial, with criticism and competing interpretations leaving uncertainty about its long-term status and scientific contribution.The controversy includes both deserved and polemical criticism, and has led some condensed-matter authors to avoid mentioning SOC (Perkovi´c et al., 1995; Krommes, 2000; Frigg, 2003; Stumpf and Porter, 2012; Alvarado et al., 2013).Source: §1 · p. 2

Paper aims and structure

• The paper revisits BTW’s original papers to distinguish competing meanings of SOC and establish BTW SOC as the theoretically grounded framework used thereafter.It first discusses multiscale avalanching in space and laboratory plasmas, then compares perceptions of SOC before focusing on the BTW formulation.Source: §1 · pp. 2–3

Definition and controversies

• The authors identify necessary and sufficient conditions for SOC, warning that avalanching, long-range correlations, or power laws alone do not establish self-organized criticality.They argue that conflating these conditions and loosely interpreting BTW’s core idea helped produce divergent SOC versions and controversy.Source: §1 · p. 3

Applications and perspective

• The paper balances its critique by showing how SOC consolidated previously disconnected observations in solar, magnetospheric, and fusion plasma physics.The authors draw on experience spanning condensed-matter theory, solar astrophysics, space and fusion plasmas, and cross-disciplinary complexity research, while concluding that SOC remains scientifically relevant without universally describing nature.Source: §1 · pp. 3–4

• Confined tokamak and astrophysical plasmas exhibit metastable transitions, intermittent energy release, multiple instability routes, and burst sizes spanning many scales.SOC-informed observations have tested power-law burst distributions in solar flares and Earth’s auroral morphology.Source: §2 · p. 5
• These systems share driven, dissipative, far-from-equilibrium behavior with slow approach to instability followed by fast multiscale reconfiguration.This separation of timescales is described as multiscale avalanching in SOC terminology.Source: §2 · p. 5
• Detailed event modeling solves tractable plasma equations and compares selected events with observations, but is best suited to physics concentrated on one spatio-temporal scale.Its questions concern instability onset and the specific mechanism causing an individual energy-release event.Source: §2 · p. 5
• When strongly coupled physics matters across all scales, renormalization-group reasoning emphasizes local interactions, coarse-graining, self-similarity, power laws, and long-range correlations.Different microscopic interactions can produce the same collective behavior, making the resulting phenomenology potentially ubiquitous.Source: §2 · pp. 5–6
• SOC offers plasma systems a multiscale alternative because slow buildup and fast relaxation generate avalanches whose dynamics and energy-release statistics are insensitive to specific instability details.This robustness is attractive where many instability mechanisms and routes to instability coexist, providing a framework for ordering observations.Source: §2 · p. 6

Core SOC idea

• The core claim, enunciated by Bak and Chen (1989), is that nature can generate spatio-temporal fractals through self-tuned phase transitions.This interpretation was formulated in mathematical-physics and condensed-matter language and has been actively debated by those communities.Source: §3 · pp. 6, 8

Four perceptions of SOC

• SOC is presented through four nested perceptions: self-tuned phase transitions, SOC-caused fractals, SOC-caused power laws, and SOC-caused contingency.The schematic ranges from the minimal core proposal to the most visionary complexity-science interpretation.Source: §3 · pp. 6–7, Figure 1 · p. 7

Erroneous perceptions

• The claims that all natural fractals or power laws arise from SOC have long been judged wrong, despite their wide influence.The paper notes that alternative mechanisms can produce power laws and that these interpretations fueled both criticism and creative misunderstanding.Source: §3 · pp. 7–8

Evidence and controversy

• Divergent interpretations produced confusion about SOC’s claims, evidence, and explanatory scope; beyond computational confirmation, no listed claim has unambiguous, unquestioned evidence.Experimental, observational, numerical, and analytical work is nevertheless converging toward corroboration of the core claim.Source: §3 · pp. 7–8

Scope and limitation

• The contingency-based interpretation reframes SOC as “complexity and contingency from criticality,” but remains more speculative and is largely outside this paper’s astrophysical-plasma focus.The authors therefore concentrate on the core interpretation and discuss the erroneous second and third perceptions only when needed to clarify confusion.Source: §3 · p. 8

• BTW proposed that natural complex systems could self-organize into a state producing spatial fractals and 1/f noise, analogous to critical systems near phase transitions in laboratory phenomena.The SOC postulate was guided by observed scaling in space and time in critical phenomena.Source: §4 · pp. 9–10
• They sought a dynamical theory for widespread spatial fractals and power-law spatial correlations, rather than accepting them as unexplained features of nature (Bak and Chen, 1989).The relevant power laws concerned correlations between spatial fluctuations, not generic size distributions.Source: §4 · p. 9
• The review identifies a persistent controversy in extending BTW’s specific theory of spatio-temporal fractal avalanching to arbitrary fractals or unrelated power-law size distributions.It argues that making SOC elastic enough to include distributions without clear links to spatial correlations or avalanches is unhelpful.Source: §4 · p. 9
• BTW also targeted 1/f noise, interpreting spatial fractals and temporal power-law behavior as two manifestations of a self-organized critical state (Bak and Chen, 1989).They described fractals as snapshots of systems operating at self-organized criticality (Bak and Chen, 1989).Source: §4 · pp. 9–10
• Although 1/f spectra had existing explanations, BTW considered them unsatisfying and lacking generality; such spectra nevertheless became supporting evidence for SOC in magnetospheric physics.A 1/f spectrum is generally associated with strong temporal correlations involving the system’s history.Source: §4 · p. 10

• SOC was formulated to unify dynamically evolving spatial and temporal fractals, not to explain all natural power laws through one mechanism (Bak and Chen, 1989; Bak et al., 1987, 1988a).The program also connected this aim to complexity science and renormalization-group ideas.Source: §5 · p. 11
• BTW argued that spatial and temporal scaling are intrinsically linked, but later work showed that temporal scaling can occur independently; joint scaling should be tested with spatiotemporal correlations in r and t.The relevant signature is algebraic rather than exponential dependence in both spatial separation and time.Source: §5 · pp. 11–12
• SOC contrasted self-organized criticality with finely tuned phase-transition criticality, proposing that systems dynamically approach scaling states with long-range correlations and emergent global behavior.Self-organization was essential to explaining why critical behavior appeared widespread in nature.Source: §5 · pp. 11–12
• The SOC postulate states that slowly driven, open, extended, dissipative systems automatically reach a critical state, with fractals appearing as snapshots of that state (Bak and Chen, 1989, p. 5).Jensen later summarized the required features as slowly driven, interaction dominated, and thresholded; openness enables transport toward dissipative boundaries.Source: §5 · pp. 12–13
• Open boundaries and dissipation are structurally important: excess conserved quantity must reach the boundary for relaxation, and the boundary cannot be scaled out at large system sizes.The model’s dissipation occurs at large-scale boundaries, unlike the small-scale dissipation in classic forward-cascade turbulence.Source: §5 · p. 13
• Although BTW described fractals as snapshots of SOC states, this should not necessarily be read as claiming that all fractals require SOC; the qualified claim concerns some fractals in open, extended, dissipative systems.Such systems developing critical states is equivalent to their developing the fractal-like spatial and temporal structures characteristic of criticality.Source: §5 · p. 13

6 Criticality and minimal stability

• SOC distinguishes critical spatiotemporal correlations, critical thresholds, and a global control parameter attaining a critical value.The review illustrates these meanings using BTW’s sandpile model.Source: §6 · p. 14

6.1 The BTW Model

• In the BTW model, threshold-exceeding slopes trigger grain transfers and avalanches that continue until no site exceeds threshold, after which external driving resumes.Boundary dissipation occurs at the virtual edge site.Source: §6.1 · pp. 14–15, Figure 2 · p. 14
• The slope-unit density ζ acts as a globally generated control parameter: driving raises it, dissipation lowers it, and avalanche feedback can pin its fluctuations as system size grows.Large avalanches are associated with comparatively large ζ and typically reduce it, whereas small avalanches occur at low ζ; this feedback was proposed as central to SOC.Source: §6.1 · p. 15

6.2 The meaning of “Criticality”

• Technical criticality denotes power-law correlations and fluctuations across all scales, whereas a threshold is merely the local value beyond which activity begins and should not be called “critical.”SOC refers to the former usage, while the review discourages the phrase “critical threshold.”Source: §6.2 · pp. 16–17
• SOC may instead be interpreted as self-organization of a global control parameter toward criticality, but the review argues that the name primarily concerns the system’s critical appearance.This distinction separates a control parameter reaching a critical value from the system displaying critical behavior, while noting that Section 7 addresses the alternative BTW interpretation.Source: §6.2 · p. 17

6.3 Minimal stability

• “Minimally stable” in SOC means that perturbations can produce system-wide events, not that every local degree of freedom is near threshold.The one-dimensional sandpile is exceptional: its average height can be 17/8 = 2.125 below threshold 3 in two dimensions, while the one-dimensional Abelian Manna Model has average height 0.9488(5) below threshold 1.Source: §6.3 · pp. 18–19

7.1 SOC’s “phenotype”: The necessary conditions to observe it

• SOC’s necessary phenotype comprises non-trivial finite-size scaling without control-parameter dependence, spatio-temporal power-law correlations, and apparent self-tuning to a critical point.Invasion percolation illustrates that scaling and correlations alone do not establish SOC because it is nonstationary and lacks self-tuning.Source: §7.1 · pp. 19–21

7.2 SOC’s “genotype”: The sufficient conditions to generate it

• The proposed sufficient genotype consists of nonlinear threshold interactions, avalanching or intermittency, and separated driving and relaxation timescales.Thresholds naturally produce avalanches, while slow driving prevents interference with relaxation; the authors hypothesize that these three conditions yield SOC and vice versa.Source: §7.2 · pp. 21–22

7.3 Must SOC and SDIDT always be the same ?

• The equivalence of SOC and SDIDT remains unsettled: SDIDT may be sufficient but incomplete, so only some SDIDT systems may display SOC.The Forest Fire Model’s lack of scaling motivates completing the sufficient-condition list without making it too narrow.Source: §7.3 · pp. 23–24, Figure 3 · p. 23

8 Why then is SOC controversial?

• The controversy therefore reflects both conceptual ambiguity and a mismatch between strong universality claims and the limited empirical, theoretical, and model-specific support available.The review’s motivating question is why SOC remains controversial despite a relatively clear core claim and proposed necessary and sufficient conditions.Source: §8 · p. 24, §8.2 · p. 27

8.1 Confusion

• SOC’s core claim was obscured by conflating self-organized phase transitions with all fractals or power laws, and by treating scaling observations as proof of SOC.Avalanche-size and duration distributions were intended as proxies for power-law correlations, not replacements; some models show power-law distributions without spatial correlations.Source: §8.1 · pp. 25–26

8.2 Ubiquity, universality, generality

• Evidence for ubiquitous natural SOC is limited because scaling and long-range spatiotemporal correlations are difficult to measure, and neither spatial nor temporal correlations imply the other.The review states that hard evidence is mostly numerical, no complete SOC theory exists, and it remains unclear why generic SOC is rarely observed.Source: §8.2 · pp. 26–27

8.3 Paradigmatic versus good models

• The original sandpile paradigm has unresolved scaling behavior, whereas the Manna and Oslo models better display robust finite-size scaling, correlations, and apparent self-tuning.The Manna model also accommodates thresholds, intermittency, separated timescales, deterministic members, and emerging analytical tools; conversely, a deterministic 1/f-noise model requires tuning.Source: §8.3 · pp. 28–29

8.4 Distraction and priority

• SOC’s reception was also strained by insufficient attention to prior work linking spatial and temporal fractality, alongside perceptions of disciplinary overreach.Earlier critical-phenomena research already connected space and time, while Mandelbrot had studied spatial-temporal fractality before SOC’s rise.Source: §8.4 · p. 29

8.5 Predictability

• SOC predictability is differentiated rather than absent: individual event sizes may be contingent, but long temporal correlations support probabilistic predictions of event clustering and hazards.SOC systems may not signal a large event’s onset or size, while memory and correlations can make large-event likelihoods predictable; solar-flare waiting-time results vary across datasets and can conflict with model predictions.Source: §8.5 · pp. 31–32

8.6 Alternative scenarios

• Alternative explanations can reproduce apparent critical behavior without SOC, including ordinary criticality swept across by a control parameter and precipitation fluctuations near a phase-transition-like threshold.In the precipitation example, vapor density is not self-tuned to one critical value; residence time instead peaks near the sharp increase in precipitation and its fluctuations.Source: §8.6 · pp. 32–33
• Precipitation dynamics keep atmospheric vapor density near the threshold for likely rain, but do not sharply tune the system to a critical state.Vapor accumulation raises density above the critical value, while rain lowers it below that value.Source: §8.6 · p. 33
• Resting-state fMRI shows a similar near-critical scenario, with brain activity remaining near a three-dimensional percolation transition (Tagliazucchi et al. (2012)).The analysis treated voxel activity from a percolation perspective.Source: §8.6 · p. 33
• SOC phenomenology may reflect dynamics that approach criticality without reaching it because coupling between the order and control parameters prevents exact fine-tuning.System-size arguments claim that increasing system size may eventually “pinch” the dynamics at the critical point (Vespignani et al., 1998; Dickman et al., 1998, but Pruessner and Peters, 2006), whereas precipitation resembles the scaling-breaking dynamical cause in the Drossel and Schwabl (1992) forest fire model.Source: §8.6 · p. 33

Solar flares

• Solar flare energies follow a power-law distribution spanning over eight orders of magnitude, motivating Lu and Hamilton’s (1991) BTW-inspired avalanche model and later SOC variants.Models progressively incorporated energy redistribution, loading mechanisms, magnetohydrodynamics, and reconnecting-loop cascades.Source: §9 · p. 34
• Solar-flare frequency distributions across large datasets and wavelengths, including parameter-conditioned X-ray subsets, constrained models to reproduce observed positive parameter correlations.These observations served as model-validation targets rather than merely qualitative analogies to SOC.Source: §9 · p. 34

Magnetospheric aurorae

• Auroral studies found SOC-like scaling across multiple observables, including AE structure functions scaling from about 1 minute to 2 hours and fat-tailed burst durations and sizes.A scale break above 2 hours and an approximately 100-minute component were associated with global magnetospheric substorms.Source: §9 · pp. 34–35

Fusion plasmas

• Fusion sandpile and avalanche models reproduced tokamak features beyond power-law avalanches, including the correlation between bursty edge-localized modes and low- versus high-confinement states.The “sandpile with an H mode” also captured anomalous inward transport and exemplified idea transfer between fusion and magnetospheric research.Source: §9 · p. 35

Magnetospheric aurorae

• Auroral threshold-exceedance analyses found fat-tailed distributions of counts and areas, while spatiotemporal tracking found power laws for blob areas, lifetimes, maximum dissipated power, and energy over 3–5 orders of magnitude.Spatiotemporal analysis addressed possible overcounting across successive images. The threshold-exceedance analysis identified an additional fixed-scale population corresponding to global substorm disturbances.Source: §9 · pp. 35–36

Limitations and open questions

• Interpretations of magnetospheric scale-free behavior remained contested because auroral projections, bursty bulk flows, variable solar-wind driving, and solar-wind turbulence could produce or modify apparent avalanches.Comparisons alternately attributed AE scaling primarily to the solar wind or to a coexisting internally generated component, while Chang’s (1992) framework distinguished driving regimes from BTW’s slow-driving SOC.Source: §9 · pp. 36–37

10.1 SOC theory: where do we stand ?

• SOC’s scope and evidential status remain confused because its original claim was extended across phenomena and has been assessed analytically, numerically, observationally, and experimentally.The original claim concerned systems organizing toward continuous phase transitions that produce long-range and long-time correlations.Source: §10.1 · pp. 37–38
• SOC is not ubiquitous, but some systems display it as originally envisaged, while its broader legacy is connecting research on heavy tails, spatio-temporal fractals, and 1/f noise.The framework also emphasizes that simple microscopic interactions can govern phenomena across many scales, offering a sharper quantitative perspective than classic reductionism.Source: §10.1 · p. 38
• Despite its flaws and overextensions, SOC remains one of the most successful approaches to understanding broad heavy-tailed distributions and cross-scale fluctuations.Such distributions challenge explanations confined to a particular length scale and can help characterize fluctuations, risks, and underlying physics.Source: §10.1 · pp. 38–39

10.2 Testing for SOC in space plasmas

• Testing plasma SOC is difficult because finite system size limits scaling ranges, spatio-temporal correlations are hard to measure, and convincing tests increasingly require global proxies.Solar-coronal observations show power laws across up to 8 decades, while auroral variables span more than two spatial orders of magnitude.Source: §10.2 · p. 39
• Power-law event-size statistics are necessary but insufficient evidence for SOC because finite plasmas can exhibit similar scaling through turbulence, linear models, or nonlinear waves, shocks, and solitons.SOC differs from turbulence through intrinsic timescale separation and an idealized vanishingly small driving-to-dissipation ratio.Source: §10.2 · pp. 39–40
• Plasma research has broadened the original SOC paradigm into multiscale avalanching because bursty, nonlinear energy release and transport couple processes across many spatiotemporal scales.This broader concept has successfully organized observations, although distinguishing it observationally from idealized SOC remains difficult.Source: §10.2 · pp. 39–40

Appendix A: Dimensional analysis, scaling and self-similarity

• Non-trivial scaling reflects self-similarity and power-law correlations, whereas dimensional analysis produces unit-independent trivial exponents and cannot explain nature’s full variety of power laws.Power-law correlations indicate competing scales without a dominant characteristic scale; dimensional analysis can nevertheless yield powerful results such as Kolmogorov’s 5/3 law.Source: Appendix A · pp. 41–42
• Direct correlation measurements are difficult in SOC because open boundaries break translational invariance and correlations are often weak, making avalanche sizes and durations clearer scaling indicators.Correlation functions depend on absolute coordinates, preventing straightforward spatial averaging.Source: Appendix A · p. 42
• Universality links scaling exponents and other quantities to interaction symmetries, allowing simple toy models to represent broader physical systems.BTW explicitly justified discrete cellular automata as simplified models whose quantitative scaling properties might apply to realistic systems at criticality (Bak et al., 1988a, p. 365).Source: Appendix A · pp. 42–43
• SOC primarily exhibits finite-size scaling: the avalanche-size cutoff scales as sc ∝ L^D, with D expected to be universal.Unlike traditional critical phenomena, SOC’s stated scaling concerns global observables as functions of system size rather than correlation functions.Source: Appendix A · p. 43
• SOC remains contentious because proposed models produce diverse results that often do not fit one universality class, and simulations may not establish scaling or exponents.The Manna and Oslo models provide the clearest evidence for SOC and belong to the same universality class, though not for trivial reasons.Source: Appendix A · p. 43