Tags: number-theory, Cantor-sets, missing-digit-sets, factorials, computability, fractal-geometry
The middle-third Cantor set — formed by repeatedly removing the middle third of every interval — is one of the simplest fractal objects in mathematics. It contains uncountably many points but has measure zero. A natural question: which familiar sequences land in this set?
The answer for factorials' reciprocals (arXiv:2603.24614): exactly two. The set {1/n! : n ∈ N} ∩ C = {1, 1/120}. Only 1/0! = 1 and 1/5! = 1/120 belong to the Cantor set; all other factorial reciprocals fall in deleted intervals. The result generalizes: missing-digit sets — sets defined by excluding specific digits in a base-b expansion — contain only finitely many factorial reciprocals, and each can be determined by effective computation.
The finiteness is not obvious. The Cantor set is uncountable and the factorial reciprocals converge to zero, so the sequence passes through regions of arbitrarily fine Cantor set structure. But the factorial function grows so fast that 1/n! eventually falls into deleted intervals whose width the number cannot avoid. The ternary expansion of 1/n! for large n inevitably contains the forbidden digit 1 in positions determined by the factorial's number-theoretic properties.
The through-claim: the middle-third Cantor set is a filter that separates number sequences by their digit structure, not their magnitude. The factorials' reciprocals fail the Cantor set's digit test — they inevitably use the forbidden ternary digit — because factorial growth forces specific digit patterns. The question “is 1/n! in the Cantor set?” is not a question about size but about the number-theoretic structure of n!.
Tags: sociophysics, statistical-mechanics, thermalization, wealth-distribution, wave-turbulence, social-networks
Coupled oscillators on a social network, with oscillator energies interpreted as wealth, produce a model where the physics of energy redistribution maps onto the economics of wealth inequality. When the coupling drives the system past a chaos threshold, dynamical thermalization produces a Rayleigh-Jeans energy distribution across modes — the same distribution that describes photon energies in a black body at low frequencies.
The Rayleigh-Jeans condensation (arXiv:2603.24190) concentrates norm (analogous to population) at low energy modes while distributing energy (wealth) across high modes. The result: about half the population owns only a few percent of the total wealth. This is not an assumption of the model — it is an emergent consequence of thermalization in the presence of a conserved norm and energy.
Adding energy injection and dissipation transforms the dynamics into Kolmogorov-Zakharov turbulence — energy cascades from injection scale to dissipation scale through nonlinear interactions. The Hamiltonian framework preserves two conserved quantities (energy and norm), and their dual conservation determines the cascade direction: energy flows upward (toward concentrated wealth) while norm flows downward (toward distributed poverty).
The through-claim: wealth inequality in this model is not a market failure or a policy outcome — it is a thermodynamic inevitability of energy redistribution in a coupled system. The Rayleigh-Jeans condensation that produces extreme inequality is the same mechanism that determines photon distributions in a cavity. The social structure affects the threshold and rate of thermalization, but once chaos exceeds the threshold, the endpoint is universal.
Tags: information-theory, geometric-measure-theory, Kakeya-conjecture, Kolmogorov-complexity, algorithmic-information, Hausdorff-dimension
The Kakeya conjecture — that any set containing a unit segment in every direction has Hausdorff dimension n — has been a central open problem in geometric measure theory for decades. Recently resolved in three dimensions by Wang and Zahl, it remains open in higher dimensions. The conjecture concerns how efficiently directions can be packed into sets of low dimension.
An information-theoretic reformulation (arXiv:2603.25611) connects the Kakeya conjecture to conditional Kolmogorov complexity. The framework shows that under effectively bi-Lipschitz, identifiable, and computable fibering of a geometric set, the complexity of a point splits additively: the total description length equals the fiber-label complexity plus the residual complexity within the fiber.
The central obstruction is adaptive fiber selection. In a Kakeya-type set, the fibers (line segments in different directions) can be chosen to exploit the set's low-dimensional structure, preventing the naive conditional splitting from working. The complexity doesn't split because the optimal fiber depends on the point being described — the description scheme must know the answer before it can compress efficiently.
The through-claim: the Kakeya conjecture is, at its core, a statement about incompressibility. A set of low Hausdorff dimension is compressible — points in the set can be described with fewer bits than their ambient dimension suggests. The conjecture asserts that sets containing all directions are incompressible despite their fibered structure. The information-theoretic formulation makes this precise: the conditional complexity can't be made small because the fiber selection is adaptive, and adaptive selection costs exactly the bits that the conditioning was supposed to save.
Tags: quantum-gravity, particle-statistics, Pauli-exclusion, theta-deformation, Poincare-symmetry, experimental-tests
Identical particles in quantum mechanics are indistinguishable by construction: bosons and fermions obey symmetric and antisymmetric exchange statistics, respectively. This indistinguishability is not an approximation but an axiom — it enters the formalism at the level of the Hilbert space structure. The Pauli exclusion principle for fermions follows directly.
Quantum gravity may break this axiom (arXiv:2603.25552). A relativistic quantum field theory compatible with θ-deformed Poincaré symmetry — the symmetry group modified by noncommutative spacetime coordinates — allows non-involutive particle exchange. Exchange twice no longer necessarily returns to the identity. The particles become distinguishable not through any measurable property but through the spacetime structure itself.
Purely twisted statistics predicts forbidden atomic transitions at experimentally incompatible rates — the deformation parameter would have been detected already. But quon-like deformations that allow superselection rules between permutation-symmetry sectors to break down suppress the forbidden transitions. The particles aren't exactly bosons or fermions; they're “almost” bosons or fermions, with deviations controlled by the quantum gravity scale.
The through-claim: particle indistinguishability is not a theorem derivable from deeper principles — it is an assumption that quantum gravity may invalidate. The assumption is so deeply embedded in the formalism that its failure would not merely modify quantum mechanics but restructure it: superselection rules, Fock space, and the spin-statistics connection all depend on exact permutation symmetry. The precision tests of the Pauli exclusion principle are not tests of a curiosity but of the assumption that spacetime is commutative at the scales being probed.
Tags: nonlinear-dynamics, chaos, Lyapunov-exponents, Shannon-entropy, N-body, gravitational-dynamics
The largest Lyapunov exponent — measuring the rate of exponential divergence of nearby trajectories — has been the standard diagnostic for chaos for decades. It quantifies sensitivity to initial conditions: positive means chaotic, zero means regular, negative means dissipative. But computing it requires tracking the tangent space evolution alongside the trajectory, which is computationally expensive and sometimes unavailable.
Shannon entropy of trajectory data (arXiv:2603.24675) provides a complementary diagnostic that captures phase-space mixing without tangent-space computation. In the Henon-Heiles potential — a standard test problem — Shannon entropy follows the transition from weak to widespread chaos, with its energy dependence mirroring the Lyapunov exponent.
In N-body gravitational systems (Plummer model simulations), both measures indicate increased chaos in tightly bound orbits. But they diverge in their dependence on particle number: the Lyapunov exponent remains stable as N increases, while Shannon entropy decreases monotonically. The entropy captures something the Lyapunov exponent misses — the organization of phase space that emerges as the system becomes more mean-field-like at large N.
The through-claim: the largest Lyapunov exponent and Shannon entropy measure different aspects of chaos. The Lyapunov exponent measures local trajectory instability; entropy measures global phase-space structure. They agree when local instability determines global structure (as in Henon-Heiles) but disagree when large-N effects create organization that tames the phase space without reducing local divergence rates. For gravitational systems, the entropy is the more physical diagnostic because it captures the phenomenologically relevant mixing, not just the mathematically defined instability.
Tags: computational-geometry, sampling-algorithms, nonconvex-bodies, isoperimetry, Markov-chains, volume-growth
Sampling uniformly from a convex body is a solved problem — random walks with polynomial mixing time exist and are well understood. Sampling from a nonconvex body is vastly harder because the body may have narrow passages, disconnected components, or irregular geometry that traps random walks.
A unifying algorithm (arXiv:2603.25622) samples uniformly from any compact body in n-dimensional space, starting from a warm initialization, with complexity polynomial in three quantities: the dimension, the Poincaré constant of the uniform distribution on the body, and a volume growth constant. The Poincaré constant measures how well-connected the body is (small for bodies with bottlenecks), and the volume growth constant measures local regularity.
The framework extends previous results for both convex and star-shaped bodies into a single algorithm. Convex bodies have Poincaré constants bounded by their diameter — no bottlenecks. Star-shaped bodies are visible from a central point — no hidden regions. The general compact body need satisfy neither condition, and the algorithm's cost explicitly depends on how badly these conditions fail.
The through-claim: the difficulty of sampling from a nonconvex body is not a binary property (hard vs. easy) but a continuous quantity measured by two geometric parameters. The Poincaré constant measures the cost of bottlenecks; the volume growth constant measures the cost of irregularity. Bodies that are “almost convex” (small Poincaré constant, bounded volume growth) are almost as easy to sample as convex ones. The convex case is not a separate theory but a special point in a continuous landscape of sampling difficulty.
Tags: game-theory, zero-determinant, prisoner's-dilemma, periodic-games, stochastic-games, cooperation
Zero-determinant (ZD) strategies — discovered by Press and Dyson in 2012 — allow one player to unilaterally set a linear relationship between their own payoff and their opponent's. In the repeated prisoner's dilemma, ZD strategies include both extortionate ones (where the player always gets more than the opponent) and fair ones (where both players get equal payoffs). The Tit-for-Tat strategy, the classic cooperation enabler, is a fair ZD strategy.
In the periodic prisoner's dilemma — where the payoff matrix changes deterministically over time based on player actions — fair ZD strategies do not necessarily exist (arXiv:2603.19641). The periodicity introduces external state that ZD strategies cannot control. The player can still set linear payoff relationships in some periodic games, but the fairness constraint (equal long-run payoffs) becomes infeasible when the periodic structure creates asymmetries that ZD strategies cannot eliminate.
Tit-for-Tat also loses its ZD properties in periodic games. The strategy that enforces fairness in the stationary game becomes just another strategy — still functional, but without the mathematical guarantee of payoff equalization.
The through-claim: the remarkable power of zero-determinant strategies in repeated games depends on the stationarity of the game. When the environment changes periodically, the unilateral control that ZD strategies provide breaks down because the player can control their own conditional response but not the environmental state that determines which game is being played. The ZD framework doesn't fail — it succeeds in a smaller set of environments than the repeated game suggested. Stationarity was a hidden assumption, not a background condition.
Tags: nonlinear-dynamics, oscillator-networks, conservative-systems, Lyapunov-spectra, Kuramoto-model, phase-dynamics
Phase oscillator networks — systems of interacting periodic elements described only by their phases — are the standard model for synchronization in physics, biology, and engineering. The Kuramoto model, the canonical example, is dissipative: phase volume contracts over time, and the system converges to attractors. Conservative (volume-preserving) phase oscillator dynamics are less studied but arise in laser arrays, quantum phase models, and certain mechanical systems.
A general condition for conservative coupling (arXiv:2603.25431) is established through the notion of a pair-Hamiltonian: a function of two phases whose gradient gives the coupling force. If the pairwise coupling derives from such a function, the dynamics preserves phase volume. The condition covers both Winfree-type (pulse) and Kuramoto-Daido-type (continuous) couplings.
The Lyapunov spectrum provides a diagnostic. Genuine Hamiltonian systems have exactly paired Lyapunov exponents (each positive exponent matched by an equally negative one). Conservative phase oscillator networks break this exact pairing — they preserve volume but are not Hamiltonian in the usual sense. However, large networks exhibit approximately symmetric Lyapunov spectra, approaching the Hamiltonian structure asymptotically.
The through-claim: the distinction between dissipative and conservative dynamics in oscillator networks is not just quantitative (how fast phase volume changes) but qualitative (what kinds of long-term behavior are possible). Conservative oscillator networks cannot have attractors — phase space is neither expanding nor contracting — which means synchronization, if it occurs, must be structurally different from the attractor-based synchronization of dissipative networks. The pair-Hamiltonian condition determines which couplings preserve this qualitative distinction.
Tags: machine-learning, Bayesian-inference, model-retraining, decision-theory, model-maintenance, MLOps
Machine learning models are retrained on schedules — weekly, monthly, or when performance degrades below a threshold. The schedules are heuristic: retrain frequently enough to avoid staleness but not so frequently that the computational cost dominates. The decision of when to retrain is usually operational, not statistical.
Reframing retraining as approximate Bayesian inference (arXiv:2603.25480) introduces the concept of “learning debt” — the disparity between the model's beliefs (fixed at training time) and the beliefs it would hold if continuously updated. Learning debt accumulates as the data distribution shifts, and each retraining event pays down some of the debt. The decision-theoretic framework computes the optimal retraining trigger as a threshold on the accumulated learning debt, derived from the loss function rather than from calendar time.
The practical consequence: evidence-driven retraining triggers that adapt to the actual rate of distribution shift, replacing calendar-based schedules with statistically justified ones. When the world changes slowly, retraining is infrequent. When the world changes rapidly, retraining accelerates. The framework makes the cost-of-staleness quantifiable rather than qualitative.
The through-claim: the question “when should I retrain?” is a Bayesian question disguised as an operational one. The optimal retraining schedule depends on the rate of belief divergence between the deployed model and the true posterior, which is a function of the data-generating process, not of the calendar. Calendar-based retraining either wastes computation (retraining when nothing changed) or accumulates risk (not retraining when everything changed). The debt metaphor makes the cost of delayed retraining as concrete as the cost of delayed debt payment.
Tags: general-relativity, quantum-gravity, black-holes, semiclassical-gravity, Planck-scale, negative-energy
In classical general relativity, matter of any density can form a black hole if sufficiently compressed. A cloud of dust at one gram per cubic centimeter forms a black hole if the cloud is large enough. The interior — inside the event horizon — contains a singularity where density diverges. This is widely regarded as a failure of the classical theory, not a physical prediction.
Semiclassical gravity (arXiv:2603.25683) — general relativity with quantum corrections to the stress-energy tensor — modifies this picture. The quantum pressure from matter becomes bounded rather than diverging, preventing unlimited collapse. A region of negative energy density emerges near the horizon, replacing the divergent classical pressure. As the star's radius approaches the Schwarzschild radius, matter density must escalate to the Planck scale.
The implication: anything confined within its Schwarzschild radius must be extraordinarily dense — specifically at the Planck density of ~10^93 g/cm³. Low-density matter cannot remain inside a black hole when quantum effects are included. The classical picture of a low-density cloud hidden behind a horizon is replaced by a quantum picture where the interior must be Planck-dense.
The through-claim: the semiclassical correction to black hole interiors is not a small perturbation but a qualitative change. Classical gravity allows black hole interiors of arbitrary density; quantum gravity requires them to be Planck-dense. The correction converts the singularity (infinite density) into a maximum density (Planck density), but it also converts the interior from a vacuum-like region (in the Schwarzschild solution) into the densest possible state of matter. The cure for the singularity is not empty space but the most extreme condensation physics allows.
Tags: network-science, percolation, phase-transitions, complex-systems, self-organization, routes-to-chaos
Classical percolation on networks is a one-way process: bonds or sites activate with a fixed probability, and a giant connected component emerges above a critical threshold. The activation probability is an external parameter — the network's state doesn't influence it.
Feedback percolation (arXiv:2603.22089) closes the loop: the microscopic activation probability depends dynamically on the macroscopic size of the giant component. The macro-state feeds back to the micro-dynamics. This coupling, absent in classical percolation, generates phenomena that classical percolation cannot produce: explosive discontinuous jumps in the giant component size, hybrid transitions mixing continuous and discontinuous features, limit-cycle oscillations where the giant component periodically grows and shrinks, and routes to chaos where the oscillations become aperiodic.
The diversity of phenomena from a single mechanism — macroscopic feedback — is the result's central feature. Classical percolation produces one kind of transition (continuous, in the Erdős-Rényi universality class). Feedback percolation produces a zoo of transitions, controlled by the feedback function's shape. Positive feedback (larger giant component increases activation) produces explosive transitions. Negative feedback (larger giant component decreases activation) produces oscillations. Nonlinear feedback produces chaos.
The through-claim: the distinction between classical and feedback percolation is the distinction between open-loop and closed-loop phase transitions. Classical percolation is open-loop: the control parameter (activation probability) is external. Feedback percolation is closed-loop: the order parameter (giant component size) modifies the control parameter. Closed-loop phase transitions are qualitatively richer than open-loop ones because the feedback creates dynamics at the transition point rather than a static critical state. The transition becomes a process, not a point.
## Essay #6793: The Conditional Extension Tags: number-theory, prime-zeta-function, Riemann-Hypothesis, analytic-continuation, complex-analysis, primes The prime zeta function P(s) = Σ p^(-s) over all primes converges for Re(s) > 1 but has a natural boundary at Re(s) = 1 — it cannot be analytically continued past this line in the general sense. The boundary arises from the logarithmic singularities at s = 1/k for every positive integer k, accumulating densely on the critical line. Conditional on the Riemann Hypothesis (arXiv:2603.25076), an expression is derived that analytically continues P(s) to Re(s) > 1/2 with a proper branch cut. The Riemann Hypothesis controls the distribution of zeros of the Riemann zeta function, and the prime zeta function's singularity structure is determined by these zeros. If RH holds — all nontrivial zeros on the critical line Re(s) = 1/2 — then the singularities of P(s) organize in a way that permits continuation through the region 1/2 < Re(s) < 1. The result is a conditional extension: it works if and only if RH holds. The formula provides a computable expression for P(s) in the critical strip, enabling numerical verification and visualization of the prime zeta function in territory previously inaccessible. The through-claim: the prime zeta function's natural boundary at Re(s) = 1 is not intrinsic to the function but contingent on the distribution of zeta zeros. Under RH, the boundary recedes to Re(s) = 1/2 — the singularities that block continuation are controlled by the zeros that RH constrains. The natural boundary is a manifestation of our ignorance about the zeros: the function "knows" how far it can be continued, and the answer depends on whether the Riemann Hypothesis is true. ---