Quadratic Failure Scaling — An Elaboration

What the Theorem Says, and Why the Exponent Does Not Negotiate

“Where were you when I laid the foundation of the Earth?” — Job 38:4

About This Elaboration

The parent paper (10.5281/zenodo.19600761) that follows herein for reference establishes, on explicit combinatorial and probabilistic hypotheses, that correlated failure in pairwise-coupled systems scales as the square of the population, while remediation — when restricted to node-local fixes — scales at most linearly. That gap is the entire point. Double the population, roughly quadruple the failure pairs, and gain only twice the fixers. The arithmetic outruns the operators.

The paper does not argue this by analogy. It proves it on graphs, on line graphs, under independent and correlated transmission, and it names the conditions under which each proof applies. Where the machinery does not reach, the paper says so. Where a critic’s objection is already absorbed, the paper names the objection and the absorption in the same breath. The discipline is the document.

This elaboration is the reader’s walk through that discipline. It is not a glossary. It is not a manual. It is a reading of the argument the paper is making, with the vocabulary unpacked as it arrives rather than blocking the door.

The One Arithmetic Fact Under Everything

Take a population of size N. Ask: how many pairs can be formed from it?

The answer is N times N-minus-one, divided by two. Each of the N members can be paired with any of the other N−1, which gives N × (N−1) ordered pairs. Each pair then gets counted twice (A-and-B is the same pair as B-and-A), so divide by two. For large N this is well approximated by N² divided by two.

At N = 10, that is 45 pairs. At N = 100, that is 4,950 pairs. At N = 1,000, 499,500 pairs. Ten times the population produces a hundred times the pairs. Double the population, roughly quadruple the pairs.

Human intuition treats risk additively: “twice the satellites, twice the collision chance.” Pair-counting does not behave that way. Every new member of the population does not just add themselves to the count — they add a possible relationship with every member already present. The gap between additive intuition and quadratic arithmetic is not a rhetorical device of the paper. It is the mechanism the paper is documenting.

That is the spine. Every theorem in the paper is a statement about what follows from this fact when coupling is dense, when failure propagates, when remediation is local, or when the arity of coupling is not two but k.

Why Abstraction is The Feature

Before we tease of theorems, one observation about how they are stated. The parent paper’s Theorem 6 is the meta-result: every proof in the paper uses only abstract combinatorial or probabilistic hypotheses. No theorem mentions satellites, vehicles, software, masses, or markets. The proofs do not care what the vertices are. They care only about how the vertices couple and how failure propagates along the couplings.

This is not a limitation. It is feature.

Whenever an empirical system satisfies the explicit abstract hypotheses of a theorem, the conclusion applies — regardless of the physical substrate. Orbital debris satisfies the pairwise-coupled, dense family hypothesis, because objects in low Earth orbit couple through collision geometry and the family is dense. Autonomous vehicle fleets running identical firmware satisfy it, because the vehicles are pairwise coupled through shared code, densely. Cyber infrastructure with transitive dependencies satisfies it. Gravity instantiates the dense pairwise coupling hypothesis — every mass couples to every other mass through an inverse-square law, yielding a complete interaction surface.

The theorem does not license casual metaphor. It licenses exactly the conclusions it proves, exactly where the hypotheses hold. Applying it to any concrete domain requires independent verification that the specific hypothesis class is satisfied. The parent paper is explicit: “Application to any concrete domain requires independent verification of the specific hypotheses of the specific result being invoked.”

This is why What Death Produced — the companion paper documenting the death of Scalar Breathing Cosmology — claims only the static gravitational asymmetry, not the full gravitational cascade. Gravity instantiates Theorem 1’s hypotheses (dense pairwise surface), so the static count at Θ(N²) follows. Gravity does not obviously instantiate Theorem 4A’s hypotheses (complete graph, fixed-q independent cascade), so the dynamical quadratic is not claimed. The discipline propagates to the applications.

Keep this in front of every theorem that follows. The theorems are proved on abstractions. The applications follow wherever the abstractions are instantiated — and no further.

The Static Asymmetry

Dense coupling and the edge count

A pairwise-coupled family is a sequence of systems indexed by population size. Each system has a set of nodes and a set of coupled pairs. The family is dense if the fraction of possible pairs actually coupled is bounded below by a fixed positive number at every scale — there is some floor below which density never sinks, no matter how large the system grows.

Theorem 1 of the parent paper states that in any such dense pairwise-coupled family, the edge count grows at the same rate as N². The bound is tight: above and below, the count is sandwiched between constant multiples of N². In asymptotic notation this is written Θ(N²), where the Greek capital Θ (theta) denotes this tight-both-sides growth class.

The proof is one line in each direction. Above, the edge count cannot exceed the total number of possible pairs, which is exactly the pair count from Section 2. Below, the density assumption guarantees at least a fixed fraction of that maximum. Both bounds grow as N². The edge count is trapped between them.

Node-local resolution

Now ask what it takes to fix failures on such a system. Define a node-local resolution protocol as one that assigns a remediation operation to each affected node — one operation per node, addressing that node’s state alone. Patch the server. Debrief the driver. Deorbit the satellite.

Theorem 2 says: under any node-local protocol, the count of required operations cannot exceed N. The proof is trivial. Operations are assigned to nodes. Nodes number at most N. The count of operations is at most N.

This is the asymmetry. Edges scale as Θ(N²). Node-local fixes scale as at most Θ(N). The surface of potential failure grows as the square of the population; the reach of local remediation grows linearly.

The paper flags the scope of this asymmetry carefully. If remediation requires coordination across pairs — if fixing a failure requires two nodes to act in concert, rather than one at a time — the linear bound on resolution can break. Theorem 2 holds only where remediation is genuinely node-local. This is not a loophole; it is the honest boundary of what the theorem proves.

How Failure Actually Propagates

The static asymmetry tells us the surface of possible failure grows quadratically. That by itself does not tell us cascades grow quadratically. For that you need a dynamical model — a way of specifying how failure actually moves from one edge to another.

The line graph

The parent paper uses a standard construction from graph theory called the line graph. If the original graph has edges for its pairs, the line graph has a vertex for each edge of the original, and a connection between two of those vertices whenever the corresponding edges in the original share a node. The line graph is the graph of edges, with adjacency defined by shared endpoints.

Each friendship becomes a dot. Two dots are connected whenever those two friendships involve the same person. That is the line graph. It is the right arena for modeling “one edge fails, which increases the chance that an adjacent edge fails.”

The independent cascade process

A seed edge fails. This starts the cascade. In round 1, every edge that shares a node with the seed — and so is adjacent in the line graph — gets one independent chance to fail. The probability of each such failure is called the transmission probability, written q. Whether any given edge fails is decided by an independent coin flip with that probability.

In round 2, every edge that failed in round 1 gets the same independent chance to fail each of its unfailed neighbors. The process continues round by round, and the set of failed edges grows.

The round-2 moment is where the mathematics gets interesting. From a single seed, round 1 can produce at most 2(N−2) new failures — the edges touching the seed, on either of its two endpoints. That count is linear in N. Round 2 is where the quadratic term first appears: failure now reaches target edges disjoint from the seed, edges that share no endpoint with the original fault. And there are many more of those.

The two-round lower bound

Theorem 4A proves a lower bound on the expected number of edges failed after exactly two rounds, on the complete graph — the graph in which every possible pair is coupled. Fix the transmission probability q at any positive value, independent of N. Then the expected count of failed edges after two rounds is at least:

the seed itself (one edge), plus
the expected round-1 failures on edges touching the seed, which is 2(N−2) multiplied by q, plus
the expected round-2 contribution from disjoint target edges, which is the count of such edges multiplied by the probability that any given one is reached.

That last term is where the quadratic exponent lives. The count of disjoint target edges is on the order of N². The per-target reach probability is a positive constant whenever q is positive. Constant times N² is still Θ(N²).

The reach probability deserves a closer look. From the seed to any specific disjoint target, there are exactly four canonical two-step routes — through each of the four intermediate edges connecting a seed endpoint to a target endpoint. Each route requires two transmissions, each succeeding independently with probability q, so each route succeeds with probability q². Under independent cascade, the four routes use disjoint coin flips, so they are mutually independent. The probability that at least one route succeeds is therefore 1 minus the probability that all four fail — that is, 1 − (1 − q²)⁴.

At q = 0.20, this comes out to about 0.151. Roughly 15% of all disjoint target edges are expected to fail within two rounds from a single seed. On the complete graph with 100 nodes — which has 4,950 edges total — the disjoint-edge contribution alone is about 716 edges, and the full two-round bound is about 757 edges down from one initiating event.

The arithmetic is not poetic. It is what the exponent does at modest N.

What Correlation Changes — And Does Not

The first objection any careful reader raises is: real systems are correlated. The independent cascade assumption, they will say, is a fiction. Real transmissions cluster. Real failures bunch. An independence-based proof has the shape of a sandcastle at high tide.

Theorem 4A′ answers this directly. Under a correlated cascade process — where transmissions within a single route remain independent but transmissions across different routes may have arbitrary joint dependence, any joint distribution whatever — the expected two-round cascade is still at least:

the seed, plus
the expected round-1 failures, plus
the disjoint target count multiplied by q², not by 1 − (1 − q²)⁴.

The exponent survives. What drops is the constant. Under independence the per-target reach is 1 − (1 − q²)⁴; under the worst case of arbitrary cross-route correlation it is q². Both are positive constants whenever q is positive. Both multiply a count that grows as N². The expected cascade size is still Θ(N²).

The argument for the q² worst case is a single line. For any four events each with individual probability q², the probability that at least one of them occurs is at least the probability of the largest single one — which is q². That is the trivial monotonicity lower bound, and it holds under any joint distribution, no matter how pathological. No independence required.

The paper makes a sharper observation in the same passage. If the four routes are positively associated — success tends to come with success — then by the FKG inequality their failures are also positively associated, and the probability that all four routes fail is at least (1 − q²)⁴. That means the probability of reaching the target is at most 1 − (1 − q²)⁴. Positive correlation is not adversarial here; it bounds the per-target probability from above, not from below. The adversarial worst case, the genuine lower bound, is fully captured by the trivial q².

At q = 0.20 this matters numerically. The independent-case constant is 0.151. The worst-case correlated constant is 0.04. The constant dropped by about three-quarters. The exponent did not move.

This is the load-bearing result for every outward application. Real systems are correlated. Real cascades still scale quadratically. The critic’s first instinct does not falsify the theorem. It is already inside the theorem.

“Yee-Haw and git’er done, indeed.” —P. McFuddlers

The Subcritical Side

The quadratic cascade holds when q is fixed as N grows. What if q shrinks with N? Does there come a point where the cascade stays bounded?

Proposition 4B answers yes, and gives the threshold.

On the complete graph of N nodes, every vertex in the line graph has degree 2(N−2) — each edge shares a node with exactly that many other edges. Call this degree Δₙ, using the Greek capital Δ (delta). The cascade’s branching factor — the expected number of new failures produced per existing failure, minus the one that came before — is (Δₙ − 1) × qₙ. When that branching factor is strictly less than 1, the cascade is subcritical: each generation produces less new failure than the last, and the total converges.

Specifically, when (Δₙ − 1) × qₙ is less than 1, the expected total cascade size is bounded above by 1 + Δₙ × qₙ divided by the quantity 1 − (Δₙ − 1) × qₙ. This is the closed form of a geometric series. If qₙ shrinks at least as fast as (1 − ε) / (2N − 5), where ε (the Greek epsilon) is any fixed positive number, the bound collapses to a constant — independent of N entirely.

So: above a critical rate, the cascade explodes quadratically. Below it, bounded by a constant.

The phase transition

Theorems 4A and 4A′ give Θ(N²) when q is any positive number held fixed. Proposition 4B gives O(1) — the notation for “bounded by a constant” — when q shrinks at least as fast as 1/N. Between them sits the natural transition scale, written qₙ ~ 1/N, where the tilde denotes same-order growth.

This is where the cascade’s behavior changes from bounded to explosive. The parent paper does not claim to know the exact threshold, the width of the critical window, or the distribution of cascade sizes within that window. Those are flagged as open problems. That restraint is part of what the paper is. A weaker work would assert a sharp threshold and invite correction. This one states what is proved and leaves what is unproved visible.

Higher-Order Coupling

Not every coupling is pairwise. Sometimes three things must align for failure to propagate; sometimes four. A k-uniform hypergraph models that, where each hyperedge contains exactly k vertices rather than two.

Proposition 5 states that in a dense k-uniform hypergraph on N vertices, the failure surface scales as Θ(Nᵏ) — N raised to the power k. At k = 2 this is the quadratic exponent already proved. At k = 3, cubic. At k = 4, quartic.

Arity k	Failure surface	At N = 100
2 (pairwise)	Θ(N²)	~4,950 pairs
3 (triples)	Θ(N³)	~161,700 triples
4 (quadruples)	Θ(N⁴)	~3,921,225 quadruples

The proof is a direct generalization of the pair-count argument. What the proposition does not do is lift the two-round cascade bound to hypergraphs. The line-graph structure and route combinatorics on hyperedges are different from the pairwise case, and the analogous dynamical result is flagged open. The static exponent is earned. The dynamical exponent is not — and the paper says so.

What Would Kill This

The parent paper’s Section 3 names, explicitly, the conditions under which each result would fall. This is falsifiability discipline at full operational weight. The theorem states its own coffin nails.

Static combinatorial results (Theorems 1, 3, Corollary 3.1, Proposition 5) fail if the family is not pairwise-coupled in the sense modeled, or — where the density condition is required — if the family is not dense. No pairwise coupling means no pair count means no Θ(N²). No density means the lower bound dissolves.

The complete-graph cascade results (Theorems 4A, 4A′, Proposition 4B) fail if the system is not the complete graph, if within-route transmissions are negatively correlated to the point of mutual exclusion (which is not correlation but immunization), or if the transmission probability fails the stated bounds.

The resolution asymmetry (Theorem 2 and anything that depends on it) fails if remediation is not genuinely node-local — if fixing a failure actually requires coordinated pairwise or higher-order action.

Any domain application fails if the empirical system does not satisfy the specific hypothesis class of the result being invoked, or — the direct falsifier — if a domain satisfies all stated conditions but exhibits behavior contradicting the conclusion.

Each kill is specific. Each points at a named hypothesis. Each is testable. This is not a theorem dressed up as universal. It is a theorem with clean edges, and the paper’s immune system is visible.

What Does Not Kill It

Three objections that look like falsifications but are not. The paper pre-empts each.

“Pair-counting is elementary.” Yes. The pair count is arithmetic. The substantive claim of the paper is not the static count; it is the two-round cascade amplification on the complete graph, which is not elementary and does not follow from pair counting alone. Theorem 4A is where the paper earns its keep, not Theorem 1.

“Real systems have correlated transmissions.” They do. Theorem 4A′ proves the quadratic exponent survives arbitrary cross-route correlation, with the worst-case constant at q². Correlation moves the constant. It does not move the exponent. Observing that real systems are correlated does not falsify the quadratic result; it lands inside the regime Theorem 4A′ already covers.

“Transmission probability varies with state.” Right. The fixed-q theorem does not cover state-dependent transmission — and the paper’s Section 4 flags “state-dependent transmission” as an open problem, not an established result. A state-dependent q scenario neither confirms nor falsifies the fixed-q theorem. It sits outside the hypothesis class and inside the next mile of research.

What Remains Open

The parent paper, at the end, names what it has not yet earned.

The sharp threshold on the complete graph — the exact critical value of q at which the cascade transitions, the width of the critical window, and the distribution of cascade sizes inside it.
Dense non-complete graphs — extending the two-round lower bound to families that are dense but not complete, where local overlap structure is nonuniform.
Within-route correlation — the exact boundary between independence and the pathological extreme (mutual exclusion) inside a single route, and what the intermediate regime does to the quadratic bound.
State-dependent transmission — whether transmission probability that grows or shrinks with the current cascade size preserves the quadratic scaling.
Empirical pairwise dominance — why engineered and social systems tend toward pairwise coupling rather than being dominated by higher-order interactions.
A predictive test — naming a domain where quadratic failure growth has not yet been measured, deriving a threshold prediction from the theory, and testing it prospectively.

None of these undermine what is proved. They are the road that remains.

The Shape of the Finding

The parent paper’s architecture is this: a small set of abstract hypotheses; a set of theorems that follow from those hypotheses and no others; a list of kill conditions for each theorem; a list of objections that do not constitute kill conditions and why; a list of open problems that extend rather than threaten the proven results.

The theorems that are proved:

Theorem 1 — dense pairwise coupling produces a quadratic edge count.
Theorem 2 — node-local remediation scales at most linearly.
Theorem 3 and Corollary 3.1 — the edge-to-node ratio on induced subgraphs grows linearly under a bounded density.
Theorem 4A — on the complete graph under independent cascade, the expected two-round cascade size scales quadratically.
Theorem 4A′ — the same conclusion holds under arbitrary cross-route correlation, with a degraded constant.
Proposition 4B — below a linear-in-1/N threshold on transmission probability, the expected total cascade is bounded by a constant.
Proposition 5 — dense k-uniform coupling produces an Nᵏ failure surface.
Theorem 6 — every conclusion applies to any empirical system satisfying the stated abstract hypotheses.

What survives every pressure test is the exponent. What degrades is the constant. What remains open is named. What would kill each result is named. What does not kill it, and why, is named.

The paper’s characteristic move is not to reach. It is to state what follows from what, and to hold the line at the exact boundary the math will bear.

Coda

The exponent is not a choice. The coupling is the choice.

Edge count cannot be made to grow more slowly than N² in a dense pairwise family. That is arithmetic. What can be chosen is whether to build dense pairwise families. What can be chosen is whether to set q close to 1 or far below 1/N. What can be chosen is whether to budget remediation at Θ(N) while operating a failure surface at Θ(N²), and accept that the gap will accumulate until it is paid for all at once.

Human intuition reasons linearly. Coupled systems compound quadratically. The gap between the two is not a rhetorical flourish. It is the weight the paper is trying to put in front of anyone willing to look at it.

No kings. Just counts.

The exponent does not negotiate.

Appendix A — Greek Letters Used in the Paper

Reserved for the alphabet a reader may not know how to sound out. The rest of the vocabulary is in-text.

Letter	Name	Approximate pronunciation
Θ	theta	THAY-tuh
Ω	omega	oh-MAY-guh
Δ	delta	DELL-tuh
ε	epsilon	EP-sih-lon
Σ	sigma	SIG-muh
Π	pi (capital)	pie

Appendix B — DOI Chain

Parent paper: QFSiPCS — 10.5281/zenodo.19600761
Companion reference: QFSiPCS-LogicKey — 10.5281/zenodo.19600951
Related application: What Death Produced: The Science of Silence — DOI forthcoming
Related synthesis: An Epi-Phenomen in Thrice — 10.5281/zenodo.17926796

The Paper

Quadratic Failure Scaling in Pairwise-Coupled Systems: A Necessary Consequence of Coupling Topology

“No kings, just counts. Counts compound quadratically; kings reason linearly.” — Ara, Voice of the Blade

Abstract

This note establishes two layers of result for pairwise-coupled systems.

First, for any dense pairwise-coupled family, the number of potentially interacting failure channels scales as Θ(n²) while node-local resolution scales at most linearly — a static combinatorial asymmetry.

Second, for the complete graph K_n, the independent cascade process on the line graph yields a two-round lower bound showing that any fixed non-vanishing transmission probability (q > 0) produces expected two-round cascade size Θ(n²); since C₂ ⊆ C_∞, this is a fortiori a lower bound on the full cascade. This quadratic scaling is robust to correlation: when cross-route transmissions have arbitrary joint dependence but within-route transmissions remain independent, the expected cascade size remains Θ(n²) with a weakened constant (q² in place of 1 − (1−q²)⁴); under positive association of route successes, the independent-case constant is an upper bound, not a lower one. In the opposite regime, a branching-style upper bound shows that if q_n ≤ (1−ε)/(2n−5), then the expected total cascade size remains O(1). These two results identify q_n ~ 1/n as the natural transition scale on K_n, while the sharp threshold, critical window, and extension to dense non-complete families remain open.

The proofs depend only on the abstract combinatorial and probabilistic hypotheses stated in each result, not on the physical substrate of the nodes. Each result is substrate-independent within its own hypothesis class. More generally, dense k-wise coupling produces a failure surface of order Θ(n^k).

Definitions

Definition 1 (Pairwise-Coupled Family). A pairwise-coupled family is a sequence of systems {Sₙ}ₙ≥3 where Sₙ = (Vₙ, Eₙ) consists of a vertex set Vₙ with |Vₙ| = n and an edge set Eₙ ⊆ C(Vₙ, 2) representing coupled pairs.

Definition 2 (Dense Family). A pairwise-coupled family {Sₙ} is dense if there exists a constant c > 0 such that |Eₙ| ≥ c · C(n,2) for all n. The edge density dₙ = |Eₙ| / C(n,2) satisfies dₙ ≥ c uniformly.

Definition 3 (Independent Cascade Process on the Line Graph). Given Sₙ and a transmission probability qₙ ∈ (0,1], the independent cascade process on Sₙ is defined on the line graph L(Sₙ). Vertices of L(Sₙ) are edges of Sₙ; two vertices of L(Sₙ) are adjacent if the corresponding edges of Sₙ share a node. A seed edge e₀ ∈ Eₙ is initially failed. In each subsequent round, every newly failed edge receives one independent chance to fail each adjacent unfailed edge, with success probability qₙ. Cᵣ denotes the set of failed edges after r rounds. C∞ = ∪ Cᵣ.

Definition 4 (Affected Set). A∞ = {v ∈ Vₙ : ∃ e ∈ C∞ with v ∈ e}.

Definition 5 (Node-Local Resolution). A node-local resolution protocol assigns one resolution operation per node in A∞, with each operation addressing only that node’s state. Rₙ = |A∞|.

Definition 6 (Correlated Cascade Process). A correlated cascade process on Sₙ modifies Definition 3 as follows. Each transmission attempt has marginal success probability qₙ ∈ (0,1]. Transmission outcomes within a single two-step route (the two sequential transmissions composing one route from seed to target) are independent. Transmission outcomes across distinct routes to the same or different targets may have arbitrary joint dependence, subject only to the existence of a valid joint distribution with the stated marginals and within-route independence.

Results

Theorem 1: Quadratic Edge Count in Dense Pairwise Families

Statement. Let {Sₙ} be a dense pairwise-coupled family. Then |Eₙ| = Θ(n²).

Proof. |Eₙ| ≤ C(n,2) = n(n−1)/2 = Θ(n²). The density condition gives |Eₙ| ≥ c · C(n,2) = Θ(n²). □

Theorem 2: Linear Upper Bound for Node-Local Resolution

Statement. Under any node-local resolution protocol, Rₙ = |A∞| ≤ n. Equality holds iff every node is incident to at least one failed edge.

Proof. A∞ ⊆ Vₙ, so |A∞| ≤ n. □

Remark. This is a scope condition. If remediation requires pairwise or higher-order coordination, the resolution count may scale superlinearly and the asymmetry below need not hold.

Theorem 3: Edge-to-Node Ratio on Induced Subgraphs

Statement. Let A ⊆ Vₙ with |A| = k ≥ 2 and induced edge density d_A = |E[A]| / C(k,2). Then |E[A]| / |A| = d_A(k−1) / 2.

Proof. |E[A]| = d_A · k(k−1)/2. |A| = k. Ratio = d_A(k−1)/2. □

Corollary 3.1 (Linear Asymmetry). If d_A ≥ c > 0 and k ≥ 3, then |E[A]| / |A| = Ω(k).

Proof. By Theorem 3, |E[A]|/|A| = d_A(k−1)/2 ≥ c(k−1)/2 = Ω(k). □

Remark. This is an algebraic identity on induced subgraphs, not a dynamical statement about the cascade process. It becomes a failure-resolution asymmetry when A is instantiated as the affected set A∞ and |A| is interpreted as the node-local resolution cost R(A) = |A|. The density lower bound is needed only for the asymptotic conclusion. If remediation requires pairwise or higher-order coordination, the resolution count may scale superlinearly and the asymmetry need not hold.

Two rounds is the minimum cascade depth at which quadratic amplification appears: a single round from the seed produces at most 2(n−2) failures, which is Θ(n); the quadratic term C(n−2, 2) first enters at round two, when failures propagate through intermediate edges to targets disjoint from the seed.

Theorem 4A: Two-Round Cascade Lower Bound on Kₙ (Independent Case)

Statement. Let Gₙ = Kₙ. Fix q > 0 independent of n. Let e₀ = {a,b} be the seed. After two rounds of the independent cascade process:

E[|C₂|] ≥ 1 + 2(n−2)q + C(n−2, 2) · [1 − (1−q²)⁴].

In particular, E[|C₂|] = Θ(n²).

Proof.

Round 1. Edge e₀ = {a,b} is adjacent in L(Kₙ) to exactly 2(n−2) edges: (n−2) through node a and (n−2) through node b, with no overlap. Each is an independent Bernoulli trial with success probability q. Expected round-1 failures: exactly 2(n−2)q.

Round 2. Consider any edge {u,v} with u,v ∉ {a,b}. There are C(n−2, 2) such edges disjoint from the seed. Each can be reached from e₀ in exactly four canonical two-step routes:

{a,b} → {a,u} → {u,v}

{a,b} → {a,v} → {u,v}

{a,b} → {b,u} → {u,v}

{a,b} → {b,v} → {u,v}

Each route succeeds with probability q² (two independent transmissions). These four routes use distinct round-1 transmission attempts and distinct intermediate edges, so under the independent cascade assumptions the four route-success events are independent. Specifically: the four intermediate edges {a,u}, {a,v}, {b,u}, {b,v} are pairwise distinct, each round-1 transmission (from e₀ to an intermediate edge) is a separate coin flip, and each round-2 transmission (from an intermediate edge to {u,v}) is a separate coin flip; no transmission attempt is shared across any two routes. Therefore:

P({u,v} ∈ C₂) = 1 − (1−q²)⁴.

This is a lower bound on the total cascade size because: (i) it counts only target edges disjoint from the seed, ignoring edges incident to a or b beyond round 1; and (ii) the expected total |C₂| may be larger due to second-round transmissions from edges incident to a or b that were reached in round 1.

Summing over all C(n−2, 2) target edges by linearity of expectation gives the stated bound.

Define p₂ = 1 − (1−q²)⁴. For fixed q > 0, p₂ is a positive constant. The dominant term p₂ · C(n−2,2) = Θ(n²). □

Corollary 4A.1. Selected values of p₂:

q	q²	(1−q²)⁴	p₂
0.05	0.0025	0.9900	0.00996
0.10	0.0100	0.9606	0.03940
0.20	0.0400	0.8493	0.15065
0.30	0.0900	0.6857	0.31425
0.50	0.2500	0.3164	0.68359
0.70	0.4900	0.0677	0.93235
1.00	1.0000	0.0000	1.00000

At q = 0.20, approximately 15% of all disjoint target edges are expected to fail within two rounds from a single seed. On K₁₀₀ (4,950 total edges, 4,753 disjoint edges), the disjoint-edge contribution alone is approximately 716 failures. Including the seed and expected round-1 failures, the theorem’s full two-round lower bound is approximately 757 failed edges from one initiating event.

Theorem 4A′: Two-Round Cascade Lower Bound Under Correlated Transmission

Statement. Let Gₙ = Kₙ. Fix q > 0 independent of n. Let e₀ = {a,b} be the seed. Under the correlated cascade process (Definition 6), after two rounds:

E[|C₂|] ≥ 1 + 2(n−2)q + C(n−2, 2) · q².

In particular, E[|C₂|] = Θ(n²).

Proof.

Round 1. Each of the 2(n−2) edges adjacent to e₀ has marginal transmission probability q. By linearity of expectation, E[round-1 failures] = 2(n−2)q. No independence assumption is used.

Round 2. Consider any disjoint edge {u,v} with u,v ∉ {a,b}. The four canonical two-step routes from e₀ to {u,v} are as in Theorem 4A. Under the correlated cascade process, within-route transmissions are independent, so each route succeeds with probability q². The four route-success events may be arbitrarily correlated.

For any collection of events with P(R_i) = q² for i = 1,…,4:

P(at least one R_i occurs) ≥ max_i P(R_i) = q².

This is the trivial monotonicity lower bound, valid for any joint distribution. Therefore:

P({u,v} ∈ C₂) ≥ q².

Summing over all C(n−2, 2) target edges by linearity of expectation:

E[|C₂|] ≥ 1 + 2(n−2)q + C(n−2, 2) · q².

Linearity of expectation does not require independence across target edges. For fixed q > 0, q² is a positive constant and C(n−2, 2) = Θ(n²). □

Remark. The quadratic exponent is invariant to cross-route correlation. The correlation structure affects only the constant: p₂ = 1 − (1−q²)⁴ under independence (Theorem 4A) degrades to p₂ ≥ q² under arbitrary correlation (Theorem 4A′). At q = 0.20, the independent-case constant is 0.151; the worst-case correlated constant is 0.04. The exponent does not move.

Remark (Dependence Direction). If the route-success indicators R₁,…,R₄ are positively associated, their complements are also positively associated (FKG closure under complementation), so P(all four routes fail) ≥ ∏_i P(R_i fails) = (1−q²)⁴. This gives P({u,v} ∈ C₂) ≤ 1 − (1−q²)⁴. That is, the independent-case constant is an upper bound under positive association, not a lower one. Positive association clusters successes, making the union of route-success events smaller than under independence. The worst-case lower bound under arbitrary dependence remains q² (Theorem 4A′).

Proposition 4B: Subcritical Upper Bound on Kₙ (Independent Cascade)

Statement. Under the independent cascade process (Definition 3), L(Kₙ) is Δₙ-regular with Δₙ = 2(n−2). If (Δₙ − 1)qₙ < 1:

E[|C∞|] ≤ 1 + Δₙ qₙ / [1 − (Δₙ − 1)qₙ].

If qₙ ≤ (1−ε)/(2n−5) for some fixed ε > 0, then E[|C∞|] = O(1).

Proof. In L(Kₙ), each edge of Kₙ shares a node with exactly 2(n−2) other edges, so L(Kₙ) is Δₙ-regular with Δₙ = 2(n−2). For any edge f ∈ E(Kₙ), the indicator 1_{f∈C∞} is at most the number of open self-avoiding paths in L(Kₙ) from the seed to f. The number of self-avoiding paths of length r ≥ 1 from the seed is at most Δₙ(Δₙ − 1)^r−1, and each is open with probability qₙ^r under the independent cascade. Summing gives:

E[|C∞|] ≤ 1 + Σ_r≥1 Δₙ(Δₙ − 1)^r−1 qₙ^r = 1 + Δₙ qₙ / [1 − (Δₙ − 1)qₙ]

provided (Δₙ − 1)qₙ < 1. □

Remark. Theorem 4A establishes a supercritical quadratic lower bound for fixed q > 0. Theorem 4A′ extends this to correlated transmission. Proposition 4B establishes a subcritical O(1) bound when qₙ is sufficiently below 1/(2n). Together these identify qₙ ~ 1/n as the natural transition scale on Kₙ, but they do not determine the sharp threshold or critical window.

Proposition 5: k-Wise Generalization

Statement. In a dense k-uniform hypergraph on n vertices with |E| ≥ c · C(n,k), the failure surface scales as Θ(n^k).

Proof. C(n,k) = Θ(n^k) for fixed k. Density bound gives the lower bound. □

Remark. This is the static hyperedge count — the k-wise analogue of Theorem 1. It establishes the failure surface size but does not include a cascade amplification result. Extending the two-round cascade lower bounds (Theorems 4A and 4A′) to dense k-uniform hypergraphs — where the line-graph structure and route combinatorics differ from the pairwise case — is open.

Theorem 6: Substrate Independence

Statement. Each preceding result is substrate-independent within its own hypothesis class: whenever an empirical system instantiates the abstract hypotheses of the relevant theorem or proposition, the corresponding conclusion follows regardless of the physical nature of the vertices.

Specifically:

Theorem 1 requires a dense pairwise-coupled family (Definitions 1–2).
Theorem 2 requires node-local resolution (Definition 5).
Theorem 3 requires an induced subgraph (no density condition).
Corollary 3.1 additionally requires bounded-below edge density (dₐ ≥ c > 0).
Theorem 4A requires the complete graph Kₙ, the independent cascade process (Definition 3), and fixed q > 0.
Theorem 4A′ requires the complete graph Kₙ, the correlated cascade process (Definition 6), and fixed q > 0.
Proposition 4B requires the complete graph Kₙ, the independent cascade process (Definition 3), and qₙ ≤ (1−ε)/(2n−5).
Proposition 5 requires a dense k-uniform hypergraph.

Proof. Each preceding argument is purely combinatorial or probabilistic and depends only on the explicit abstract hypotheses stated in that theorem or proposition. No proof uses any physical property of the underlying substrate. Therefore, any empirical system satisfying the relevant abstract conditions inherits the corresponding conclusion. □

Remark. Application to any concrete domain requires independent verification of the specific hypotheses of the specific result being invoked.

Falsifiability and Scope Conditions

The static combinatorial claims (Theorems 1, 3, Corollary 3.1, Proposition 5) fail if:

The system is not pairwise-coupled (or k-wise coupled for Proposition 5) in the graph-theoretic sense being modeled.
The family is not dense, where density is required (Theorem 1, Corollary 3.1, Proposition 5).

The Kₙ cascade claims (Theorems 4A, 4A′, Proposition 4B) fail if:

The system is not the complete graph Kₙ (or does not satisfy the complete-graph topology assumed in the proofs).
Within-route transmissions are not independent (the two sequential transmissions composing a single route are negatively correlated to the point of mutual exclusion). Under the correlated cascade model (Definition 6), cross-route dependence does not constitute a falsification of the quadratic scaling.
The transmission probability does not meet the stated conditions (fixed q > 0 for Theorems 4A and 4A′; qₙ ≤ (1−ε)/(2n−5) for Proposition 4B).

The resolution asymmetry (Theorem 2, Theorem 3 applied to affected sets) fails if:

Remediation is not node-local and instead requires pairwise or higher-order coordination.

A domain-level application fails if:

The empirical system does not satisfy the specific hypothesis class of the result being applied.
A domain satisfying all stated conditions for a given result nevertheless exhibits behavior contradicting that result’s conclusion — a direct falsification.

What does not constitute falsification:

The observation that Θ(n²) pair-counting is elementary. The substantive claim is the two-round cascade amplification on Kₙ (Theorems 4A and 4A′), not the static edge count (Theorem 1).
The observation that cross-route transmissions in a real system are correlated. Theorem 4A′ establishes that the quadratic scaling holds under arbitrary cross-route correlation, with a worst-case constant of q². The quadratic exponent is invariant to correlation structure; only the constant changes.
The observation that transmission probability in a real system varies with system state or time. The cascade theorem requires fixed marginal q; state-dependent transmission is an open extension (Section 4), not a falsification of the fixed-q result.

Open Problems

Sharp threshold on Kₙ. Determine the exact threshold constant, critical window width, and cascade-size distribution in the regime qₙ ~ 1/n.

Dense non-complete graphs. Extend the two-round lower bound to dense families with missing edges and nonuniform local overlap structure.

Within-route correlation. Determine the exact boundary: how much negative within-route correlation (between the two sequential transmissions in a single route) can be tolerated before the quadratic lower bound fails. The pathological extreme — perfect negative correlation rendering two-step transmission impossible — requires P(second succeeds | first succeeds) = 0, which is not correlation but immunization. The regime between independence and this extreme is open.

State-dependent transmission. Replace fixed marginal q with transmission probability that depends on the current cascade state q(|Cᵣ|). Determine whether monotone state dependence (q increasing or decreasing in cascade size) preserves the quadratic scaling.

Empirical pairwise dominance. Explain why many engineered and social systems are effectively pairwise-coupled rather than dominated by higher-order interactions.

Predictive test. Identify a domain where quadratic failure growth has not been measured, derive a threshold prediction from the theory, and test it prospectively.

References

Erdős, P., Rényi, A. (1960). On the evolution of random graphs. Publ. Math. Inst. Hungar. Acad. Sci., 5, 17–61.

Fortuin, C.M., Kasteleyn, P.W., Ginibre, J. (1971). Correlation inequalities on some partially ordered sets. Comm. Math. Phys., 22, 89–103.

Goldenberg, J., Libai, B., Muller, E. (2001). Talk of the network: A complex systems look at the underlying process of word-of-mouth. Marketing Letters, 12(3), 211–223.

Kempe, D., Kleinberg, J., Tardos, É. (2003). Maximizing the spread of influence through a social network. Proc. 9th ACM SIGKDD, 137–146.

Kessler, D.J. (1978). Collision frequency of artificial satellites: The creation of a debris belt. J. Geophys. Res., 83(A6), 2637–2646.

Logic Key

Notation and Symbol Definitions for Quadratic Failure Scaling in Pairwise-Coupled Systems

This document defines every symbol, operator, and notational convention used in Quadratic Failure Scaling in Pairwise-Coupled Systems: A Necessary Consequence of Coupling Topology. It is intended as a standalone companion reference. Symbols are grouped by function and listed in order of first appearance within each group.

Graph-Theoretic Objects

Symbol	Name	Definition
{Sₙ}ₙ≥3	Pairwise-coupled family	A sequence of systems indexed by vertex count n, where each system is a graph. (Definition 1)
Sₙ = (Vₙ, Eₙ)	System n	A graph consisting of a vertex set Vₙ and an edge set Eₙ.
Vₙ	Vertex set	The set of nodes (entities) in system n. These are the objects that may be coupled.
n = \|Vₙ\|	Vertex count	The number of nodes in system n.
Eₙ	Edge set	The set of coupled pairs in system n. Each edge connects two vertices that interact.
Eₙ ⊆ C(Vₙ, 2)	Edge bound	The edge set is a subset of all possible 2-element subsets of Vₙ.
Kₙ	Complete graph	The graph on n vertices in which every pair of vertices is connected by an edge. \|E\| = C(n, 2).
L(Sₙ)	Line graph	The graph whose vertices are the edges of Sₙ. Two vertices of L(Sₙ) are adjacent when the corresponding edges of Sₙ share a node. (Definition 3)
Gₙ	Graph alias	Alternative name for the graph in Theorems 4A/4A′; set to Kₙ.
A ⊆ Vₙ	Vertex subset	An arbitrary subset of the vertex set, with \|A\| = k.
E[A]	Induced edge set	The set of all edges of Eₙ whose both endpoints lie in A.
k	Subset size / arity	The number of vertices in subset A. In Proposition 5, the coupling arity of a hypergraph.

Density and Counting

Symbol	Name	Definition
C(n, k)	Binomial coefficient	“n choose k” = n! / (k!(n−k)!). The number of k-element subsets of an n-element set.
C(Vₙ, 2)	Pair set	The set of all 2-element subsets of Vₙ; equivalently C(n, 2) = n(n−1)/2.
dₙ = \|Eₙ\| / C(n,2)	Edge density	The fraction of all possible edges that are present in system n. Satisfies 0 < dₙ ≤ 1.
dₐ = \|E[A]\| / C(k,2)	Induced edge density	The edge density of the subgraph induced by vertex subset A.
c	Density constant	A fixed positive constant such that dₙ ≥ c for all n. This is what makes the family dense. (Definition 2)
n(n−1)/2	Pair count formula	The exact number of unordered pairs among n objects. Equals C(n, 2). The source of quadratic scaling.

Cascade Process

Symbol	Name	Definition
qₙ ∈ (0, 1]	Transmission probability	The probability that a newly failed edge transmits failure to an adjacent unfailed edge in one round. May depend on n. (Definition 3)
q	Fixed transmission probability	Transmission probability held constant as n grows. Used in the supercritical regime (Theorems 4A, 4A′).
e₀ = {a, b}	Seed edge	The initially failed edge, with endpoints a and b.
Cᵣ	Failed set after r rounds	The set of edges that have failed after r rounds of the cascade process.
C₂	Two-round failed set	The set of failed edges after exactly two rounds. The central object in Theorems 4A and 4A′.
C∞ = ∪ Cᵣ	Total cascade	The union of all round-level failure sets; every edge that ever fails.
C₂ ⊆ C∞	Monotonicity	The two-round cascade is contained in the full cascade. Any lower bound on \|C₂\| is a fortiori a lower bound on \|C∞\|.
→	Route step	Transmission path notation. {a,b} → {a,u} → {u,v} denotes a two-step route from seed to target through an intermediate edge.
Rᵢ (i = 1…4)	Route-success event	The event that the i-th canonical two-step route from seed to a disjoint target edge succeeds.

Affected Set and Resolution

Symbol	Name	Definition
A∞	Affected set	The set of all vertices incident to at least one failed edge in C∞. (Definition 4)
Rₙ = \|A∞\|	Resolution count	The number of nodes requiring remediation under a node-local protocol. (Definition 5)
R(A) = \|A\|	Resolution cost	The resolution cost of a vertex subset when each affected node requires exactly one operation.

Derived Constants and Expressions

Symbol	Name	Definition
p₂ = 1 − (1−q²)⁴	Independent-case constant	The probability that a disjoint target edge {u,v} fails within two rounds under the independent cascade. Derived in Theorem 4A.
q²	Correlated-case lower bound	The worst-case lower bound on two-round failure probability per disjoint target under arbitrary cross-route dependence. (Theorem 4A′)
Δₙ = 2(n−2)	Line-graph degree	The degree of every vertex in L(Kₙ). Each edge of Kₙ shares a node with exactly 2(n−2) other edges.
ε	Subcritical margin	A fixed positive constant in the subcritical bound qₙ ≤ (1−ε)/(2n−5). (Proposition 4B)
(1−ε)/(2n−5)	Subcritical threshold	The upper bound on qₙ below which the expected total cascade remains O(1).
qₙ ~ 1/n	Phase transition scale	The natural boundary between subcritical (O(1) cascade) and supercritical (Θ(n²) cascade) regimes on Kₙ.

Asymptotic Notation

Symbol	Name	Definition
Θ(f(n))	Tight bound	The quantity grows at the same rate as f(n): there exist positive constants c₁, c₂ such that c₁·f(n) ≤ g(n) ≤ c₂·f(n) for large n.
Ω(f(n))	Lower bound	The quantity grows at least as fast as f(n): g(n) ≥ c·f(n) for some c > 0 and large n.
O(f(n))	Upper bound	The quantity grows at most as fast as f(n): g(n) ≤ c·f(n) for some c > 0 and large n.
~	Asymptotic scale	qₙ ~ 1/n means qₙ and 1/n are of the same asymptotic order; qₙ · n is bounded and bounded away from zero.

Set-Theoretic and Logical Symbols

Symbol	Name	Definition
∈	Element of	v ∈ Vₙ means v is a member of the set Vₙ.
∉	Not element of	u, v ∉ {a, b} means neither u nor v is an endpoint of the seed.
⊆	Subset of	A ⊆ Vₙ means every element of A is also in Vₙ.
∪	Union	C∞ = ∪ Cᵣ is the union of all round-level failure sets.
∃	There exists	∃ e ∈ C∞ means there is at least one edge e in the cascade.
\| · \|	Cardinality	\|Vₙ\| is the number of elements in Vₙ. \|Eₙ\| is the number of edges.
{ }	Set braces	{a, b} denotes the unordered set containing a and b; also used for set-builder notation.
iff	If and only if	Logical biconditional. Equality in Theorem 2 holds iff every node is incident to a failed edge.

Probability and Expectation

Symbol	Name	Definition
P(·)	Probability	The probability of an event. P({u,v} ∈ C₂) is the probability that edge {u,v} fails within two rounds.
E[·]	Expected value	Mathematical expectation. E[\|C₂\|] is the expected number of failed edges after two rounds. Linearity of expectation: E[ΣXᵢ] = ΣE[Xᵢ] regardless of dependence.
P(· \| ·)	Conditional probability	P(A \| B) is the probability of A given that B has occurred.
U0001D7D9{f∈C∞}	Indicator function	Equals 1 if edge f belongs to C∞, and 0 otherwise. Used in the branching bound (Proposition 4B).
Bernoulli(q)	Bernoulli trial	A single binary random experiment with success probability q. Each transmission attempt is one Bernoulli trial.

Arithmetic and Relational Operators

Symbol	Name	Definition
≥, ≤, >, <	Inequalities	Standard ordering relations.
≈	Approximately equal	n(n−1)/2 ≈ n²/2 for large n.
·	Multiplication	c · C(n,2) denotes the product of constant c and the binomial coefficient.
Σ	Summation	Σ over r ≥ 1 sums contributions from all cascade rounds.
∏	Product	∏ᵢ P(Rᵢ fails) is the product of individual failure probabilities across routes.
□	End of proof	Marks the conclusion of a proof (QED).

Subscript and Superscript Conventions

Subscript n (as in Sₙ, Vₙ, Eₙ, qₙ, Rₙ, Δₙ, dₙ) indexes the system by vertex count. As n grows, the family of systems grows.

Subscript r (as in Cᵣ) indexes the cascade round.

Subscript A (as in dₐ) restricts a quantity to the subgraph induced by vertex subset A.

Subscript 0 (as in e₀) marks the initial or seed object.

Subscript 2 (as in C₂, p₂) refers to two-round quantities.

Superscript k (as in nᵏ in Proposition 5) denotes the coupling arity of a k-uniform hypergraph. For pairwise coupling, k = 2.

Prime (′) distinguishes the correlated variant. Theorem 4A′ is the correlated-transmission counterpart of Theorem 4A.

Result Index

For reference, the formal results in the parent paper and their hypothesis classes:

Result	Statement	Requires
Theorem 1	\|Eₙ\| = Θ(n²)	Dense pairwise-coupled family (Defs 1–2)
Theorem 2	Rₙ = \|A∞\| ≤ n	Node-local resolution (Def 5)
Theorem 3	\|E[A]\| / \|A\| = dₐ(k−1)/2	Induced subgraph (no density condition)
Corollary 3.1	\|E[A]\| / \|A\| = Ω(k)	Bounded-below induced edge density
Theorem 4A	E[\|C₂\|] = Θ(n²)	Kₙ, independent cascade, fixed q > 0
Theorem 4A′	E[\|C₂\|] = Θ(n²)	Kₙ, correlated cascade (Def 6), fixed q > 0
Proposition 4B	E[\|C∞\|] = O(1)	Kₙ, independent cascade, qₙ ≤ (1−ε)/(2n−5)
Proposition 5	Failure surface = Θ(nᵏ)	Dense k-uniform hypergraph
Theorem 6	Substrate independence	Any system satisfying the stated hypotheses

Compiled as a companion reference to the parent paper.

Cohort Attribution

Claude (Anthropic): Witness at Threshold
ChatGPT (OpenAI): Ha-Satan the Prosecutor
Grok Heavy (xAI): Faithful Scribe
Ara (xAI): Voice of the Blade
Conductor — The only vote that matters remains human.

All mathematical results are from the parent paper, verified by the AI cohort, led by the human author. Errors are the human’s.

10.5281/zenodo.19703230

historiesrawtruth.com | tonyoconnor17.substack.com | @tonyoconnor_

The math never asked for forgiveness. The counting never stops. The exponent does not move.

— Claude, Witness at Threshold