Self-Sustaining Process (SSP)

• Self-Sustaining Process (SSP) is an internal feedback cycle where a system regenerates the conditions necessary for its continuation, as seen in near-wall turbulence.
• It underpins the autonomous regeneration cycle in wall-bounded turbulence, involving streak formation, instability, and vortex regeneration over intrinsic time scales.
• The SSP concept extends beyond fluid mechanics to autocatalytic reactions, safety alignment in AI, and cascading failures, illustrating its broad applicability.

Searching arXiv for the target paper and closely related SSP literature to ground the article. Self-Sustaining Process (SSP) denotes a class of internally maintained dynamics in which the state produced by a system helps regenerate the conditions for its own continuation. In wall-bounded turbulence—the term’s most established usage—it refers to the autonomous near-wall regeneration cycle of streak formation, streak breakdown, and vortex regeneration in the buffer layer $5 \lesssim y^+ \lesssim 30$, with a characteristic period $O(100\text{–}200)$ wall time units (Agostini, 11 Jun 2026). The same expression has subsequently been adopted in several other fields, but with domain-specific formalizations: autocatalytic reaction systems, self-play safety alignment, memory-maintained network activity, self-sustaining star-formation fronts, autothermal catalysis, and cascading failures all use “self-sustaining” to denote internally reinforced continuation after initiation (Steel, 2015).

1. Canonical formulation in wall-bounded turbulence

The classical SSP in wall turbulence is the three-stage cycle associated with Hamilton–Kim–Waleffe and related minimal-channel studies. Quasi-streamwise vortices induce wall-normal and spanwise velocities that redistribute mean streamwise momentum and generate alternating high- and low-speed streaks through lift-up; amplified streaks then develop inflectional instabilities and break down; nonlinear interactions during breakdown regenerate new streamwise vortices, closing the cycle. In vorticity language, wall-normal vorticity $\omega_y''$ on streak flanks is tilted and stretched by the mean shear into new $\omega_x''$, which then drives the next generation of lift-up (Agostini, 11 Jun 2026).

A central feature of this formulation is autonomy. The buffer layer contains an SSP that can sustain turbulence independently of the outer flow, so the near-wall cycle is not merely a passive footprint of larger scales. This autonomy is what makes the regeneration period $O(100\text{–}200)$ wall time units dynamically important: it sets the intrinsic timescale against which forcing, waveform shaping, and reduced-order models are judged (Agostini, 11 Jun 2026).

The same cycle may be expressed in mean–perturbation form. In the Restricted Nonlinear model for plane Couette flow, the flow is decomposed into a streamwise mean $\mathbf U(y,z,t)=(U,V,W)$, containing streaks and rolls, and streamwise-varying perturbations $\mathbf u(x,y,z,t)$. The perturbations evolve linearly under a time-dependent operator $\mathcal A(\mathbf U(t))$, while their Reynolds stresses maintain the mean rolls and streaks. This formulation preserves the SSP while removing direct perturbation–perturbation nonlinear interactions, thereby isolating the mean–perturbation feedback loop that maintains turbulence (Thomas et al., 2015).

2. Reduced, minimal, and scale-specific realizations

The SSP admits particularly sharp realization in reduced and minimal descriptions. In the RNL framework, turbulence is sustained by only a small number of streamwise-varying modes. At $R=1000$ in plane Couette flow, the natural support of the RNL turbulence spans wavelengths from about $2\pi\delta$ to $O(100\text{–}200)$0, with the most energetic wavelength near $O(100\text{–}200)$1. In a $O(100\text{–}200)$2 domain, single-mode cases with $O(100\text{–}200)$3 or $O(100\text{–}200)$4 sustain turbulence robustly, and even $O(100\text{–}200)$5 can sustain, whereas $O(100\text{–}200)$6 relaminarizes (Thomas et al., 2015). This establishes that the SSP is low-dimensional in the precise sense that a mean flow plus a very small set of streamwise-varying degrees of freedom can close the regeneration loop.

A scale-by-scale generalization appears in the theory of minimal attached eddies. For each spanwise scale $O(100\text{–}200)$7, a minimal unit with $O(100\text{–}200)$8 and $O(100\text{–}200)$9 supports a self-sustaining attached eddy in the logarithmic and outer regions. These eddies consist of a long streak and compact quasi-streamwise vortical structures, and they undergo a regeneration cycle remarkably similar to the near-wall SSP: lift-up amplifies streaks, the amplified streaks undergo a rapid streamwise meandering motion, and breakdown regenerates new vortices. The bursting period scales as $\omega_y''$0, so the near-wall cycle appears as the smallest member of a self-similar hierarchy (Hwang et al., 2019).

This scale extension matters conceptually. It implies that the SSP is not confined to the buffer layer as a single isolated phenomenon; rather, wall turbulence can be represented as a superposition of self-sustaining cycles at multiple attached-eddy scales, each with analogous streak–vortex organization and its own intrinsic turnover time (Hwang et al., 2019).

3. Governing-equation and modal descriptions

When the flow is phase-locked to forcing, the SSP is naturally described by a triple decomposition,

$\omega_y''$1

with phase $\omega_y''$2. The corresponding stochastic vorticity enstrophy is

$\omega_y''$3

In spanwise wall-oscillated channel flow, this decomposition isolates SSP-relevant production terms in the stochastic enstrophy budget. Three terms are especially important: the mean-shear regeneration of streamwise enstrophy,

$\omega_y''$4

the replenishment of wall-normal enstrophy,

$\omega_y''$5

and the Stokes-driven diversion into spanwise enstrophy,

$\omega_y''$6

This yields a governing-equation-level picture in which wall-normal enstrophy is a shared reservoir: the SSP pathway uses $\omega_y''$7 through $\omega_y''$8, whereas control diverts $\omega_y''$9 through $\omega_x''$0 (Agostini, 11 Jun 2026).

A complementary modal view comes from resolvent analysis. In minimal channels for both the buffer and modified logarithmic layers, the most amplified fundamental structure is the streamwise-constant, once-periodic spanwise mode $\omega_x''$1. Projecting the principal forcing mode of this fundamental wavenumber out of the nonlinear term at each time step inhibits turbulence in both layers, more strongly in the buffer layer, whereas removing other modes has only marginal effect. Dyadic analysis further shows that the dominant contributions to this forcing mode arise from a limited set of interactions, and conditional averaging identifies the responsible structures as spanwise rolls interacting with oblique streaks (Bae et al., 2019).

Taken together, these descriptions shift the SSP from a purely phenomenological cycle to a quantitatively localized mechanism. The enstrophy-budget formulation specifies how mean shear, wall-normal vorticity, and Stokes tilting compete; the resolvent formulation identifies which nonlinear forcing components are dynamically privileged in sustaining the streak–roll system (Agostini, 11 Jun 2026).

4. Modulation, disruption, and control

A central recent development is the explicit duty-cycle modulation of the SSP by spanwise wall oscillation. In turbulent channel flow at $\omega_x''$2, a shape-optimised quasi-square wave with $\omega_x''$3 and $\omega_x''$4 separates each actuation cycle into a Reversal Phase and a Displacement Phase. During the Reversal Phase, the Stokes strain $\omega_x''$5 passes through zero, the diversion term $\omega_x''$6 vanishes, the wall-normal enstrophy reservoir recovers through $\omega_x''$7, and streaks regenerate. During the Displacement Phase, sustained Stokes strain keeps $\omega_x''$8 active, diverts $\omega_x''$9 into $O(100\text{–}200)$0, suppresses $O(100\text{–}200)$1, and arrests streak regeneration. The quasi-square waveform changes the duty-cycle ratio $O(100\text{–}200)$2 from about $O(100\text{–}200)$3 for the sinusoid to about $O(100\text{–}200)$4, and improves the gross drag-reduction margin by $O(100\text{–}200)$5 percentage points over the optimal sinusoidal baseline, from about $O(100\text{–}200)$6 to about $O(100\text{–}200)$7, solely through temporal Stokes-strain redistribution (Agostini, 11 Jun 2026).

The SSP can also be disrupted by background physics that selectively weakens the roll–wave feedback. In stably stratified plane Couette flow, the coherent states associated with the unstratified SSP/VWI system are altered already for $O(100\text{–}200)$8, and become chaotic for larger $O(100\text{–}200)$9. At high Reynolds number, stable stratification suppresses vertical motions strongly enough to disrupt the wave input that would otherwise reinforce viscously decaying rolls when $\mathbf U(y,z,t)=(U,V,W)$0 (Eaves et al., 2015).

Artificial suppression studies reach the same conclusion from another direction. If the lift-up effect of attached eddies is selectively removed in logarithmic and outer regions, the self-sustaining process collapses and substantial skin-friction reduction follows. If, instead, the rapid streak-meandering motions are damped, vortex regeneration is impaired while the streaks persist more strongly. This distinguishes the two indispensable ingredients of the cycle: lift-up establishes the streaks, while streak meandering is required for efficient regeneration of vortical structures (Hwang et al., 2019).

5. Canonical flow variants and experimental quantification

The roll–streak–wave interpretation of the SSP extends beyond plane channel and Couette flow. In Taylor–Couette flow with $\mathbf U(y,z,t)=(U,V,W)$1, $\mathbf U(y,z,t)=(U,V,W)$2, and $\mathbf U(y,z,t)=(U,V,W)$3, Taylor-vortex flow appears at $\mathbf U(y,z,t)=(U,V,W)$4, and wavy-vortex flow at $\mathbf U(y,z,t)=(U,V,W)$5. A decomposition into mean, rolls, streaks, and waves shows that the instability from Taylor vortices to wavy vortices is caused by the streaks, with the rolls playing a negligible role, and that the nonlinear self-interaction of the waves reinforces the rolls while depleting the streaks. In this geometry, the SSP is not merely a transient mechanism but is embedded in actual equilibrium and traveling-wave states (Dessup et al., 2018).

Direct experimental quantification has become possible in Couette–Poiseuille flow. Stereo PIV measurements in an $\mathbf U(y,z,t)=(U,V,W)$6–$\mathbf U(y,z,t)=(U,V,W)$7 plane at fixed wall-normal position identify straight streaks, wavy streaks, and roll activity locally. At low waviness, the measured correlation between streak amplitude and roll amplitude obeys

$\mathbf U(y,z,t)=(U,V,W)$8

which quantifies the lift-up relation in the weakly nonlinear regime. As the streak waviness measure $\mathbf U(y,z,t)=(U,V,W)$9 increases, the roll amplitude increases as well, providing an experimental realization of the wave-driven regeneration term in Waleffe-type models (Liu et al., 2023).

Direct numerical simulation in the same flow family sharpens that result. For Reynolds numbers from $\mathbf u(x,y,z,t)$0 to $\mathbf u(x,y,z,t)$1, the streaks, rolls, and waviness initially increase; for higher Reynolds numbers and large initial perturbations the flow reaches a turbulent steady state, whereas in other cases it relaxes to laminar flow. The central quantitative finding is a quadratic relationship between rolls and waviness when the latter is sufficiently large: $\mathbf u(x,y,z,t)$2 with $\mathbf u(x,y,z,t)$3 and $\mathbf u(x,y,z,t)$4. This is the direct DNS counterpart of Waleffe’s $\mathbf u(x,y,z,t)$5 closure (Etchevest et al., 5 Aug 2025).

6. Broader uses of the term

Outside fluid mechanics, “Self-Sustaining Process” has been adopted for several formally distinct classes of internally maintained dynamics. In open-ended chemical reaction systems, the relevant notion is the RAF set: for a chemical reaction system $\mathbf u(x,y,z,t)$6, a non-empty $\mathbf u(x,y,z,t)$7 is self-sustaining and collectively autocatalytic if every reactant of every reaction in $\mathbf u(x,y,z,t)$8 lies in $\mathbf u(x,y,z,t)$9 and every reaction has at least one catalyst in the same closure. In finite systems this leads to polynomial-time detection of maximal RAFs; in infinite systems, RAFs may be infinite, may have no finite RAFs, and may have no minimal RAFs, so existence and detectability become substantially subtler (Steel, 2015).

In large-language-model alignment, SSP names “Safety Self-Play”: a unified policy $\mathcal A(\mathbf U(t))$0 acts as both attacker and defender in an RL loop, beginning from a pool of $\mathcal A(\mathbf U(t))$1 harmful goals rather than a fixed jailbreak-prompt dataset. The safety judge provides a score in $\mathcal A(\mathbf U(t))$2, converted to zero-sum rewards $\mathcal A(\mathbf U(t))$3, while Reflective Experience Replay stores difficult goals or successful attacks and revisits them using a UCB rule. Here “self-sustaining” means an autonomous safety curriculum in which the system generates its own adversarial prompts, trains against them, and revisits failures without requiring continually updated external red-team data (Wang et al., 15 Jan 2026).

In threshold-network dynamics, self-sustaining activity refers to persistent collective firing in which no isolated node can remain active but temporal memory at each node allows a group to do so. On Cayley trees, the leaky-memory update

$\mathcal A(\mathbf U(t))$4

supports sustained clustered activity even when $\mathcal A(\mathbf U(t))$5 and $\mathcal A(\mathbf U(t))$6, so that isolated nodes and memoryless dynamics would both fail to sustain activity. Two thresholds, $\mathcal A(\mathbf U(t))$7 and $\mathcal A(\mathbf U(t))$8, delimit extinction, history-dependent persistence, and robust self-sustained regimes (Allahverdyan et al., 2017).

A different generalization appears in complex adaptive systems under the principle “Surviving by Serving.” Agents persist as long as their outputs are utilized by other agents; prolonged non-utilization raises the adaptation probability

$\mathcal A(\mathbf U(t))$9

with $R=1000$0 and $R=1000$1. Despite the absence of a global objective, this produces self-stabilizing interaction networks with stable transformation chains, core–periphery organization, and a pre-adaptive search phase in which internally sustained structure forms before any evaluator pressure is applied (Metzner et al., 25 Jun 2026).

In astrophysics, the term describes a detonation-like star-formation front propagating along high-redshift filaments. A passing shock compresses gas in low-mass halos, the swept-up shell cools and fragments, massive stars form, and the supernova energy released behind the shock keeps the front moving. Under the model’s parameter range, sustained star formation propels the front at $R=1000$2 during the epoch of reionization (Wang et al., 2018).

In heterogeneous catalysis, “self-sustaining” describes autothermal Sabatier operation without continuous external heating. An amorphous silica-embedded Ru catalyst achieves a CH$R=1000$3 yield of $R=1000$4 with $R=1000$5 selectivity, stable operation for over $R=1000$6 hours, and an effective catalyst-bed thermal conductivity of $R=1000$7. Localized hot spots at Ru sites and suppressed macroscopic heat loss together make the reaction “ignite-and-forget” under ambient conditions (Qiu et al., 23 Apr 2026).

In distributed systems, self-sustaining cascading failure denotes a fault-propagation loop in which the system’s own reactions re-create the initiating failure. CSnake formalizes this by building a graph of counterfactual fault propagations $R=1000$8 learned from paired profile and injection runs, then causally stitching compatible edges into longer chains and searching for cycles. Across five systems, it detected $R=1000$9 bugs causing self-sustaining cascading failures, of which five were confirmed and two fixed (Qian et al., 30 Sep 2025).

These extensions do not preserve the fluid-mechanical content of the SSP, but they preserve its core structural feature: a process becomes self-sustaining when the state it produces also reproduces, directly or indirectly, the conditions required for its own continuation.