The impossibility triangle

We built a complete microscopic mechanism for an emergent Newtonian force — source, propagator and back-reaction from a single relational bath — and then pushed it until it broke in four places at once. The four fractures turned out to be one theorem. This page is the record of that week: the most productive failure of the program, and the reason gravity may have no choice but to be a gauge theory.

emergent-gravity.com · July 2026 · sequel to the main page · every claim tagged, every number scripted

1 · The question that was left standing

The main page ends with a mechanism that makes weight — records attract, Newton's law emerges, G is derived — and one confession: the long-range field was still added by hand as a constraint equation. The open question was surgical: why would a constrained 1/q² field emerge from monitoring at all? Six numbered experiments (n31–n36) attacked it. Everything below is measured in exactly solvable models, with pre-registered gates, and the kills are ours.

2 · Relational clocks: conservation is free, rigidity is not

Give every lattice site a quantum clock — a canonical pair (τᵢ, εᵢ), reading and generator. Let the vacuum record only comparisons: Le ∝ (τᵢ−τⱼ). Then two things are exact, not approximate: the total energy E = Σεᵢ commutes with every record — all moments of E are conserved under monitoring, while local energies diffuse (the backaction becomes energy hydrodynamics); and the information kernel of the records is the graph Laplacian, 𝓘 = 4γLG, with an exact zero mode — a uniform clock shift is unobservable by construction. measured, zero mode < 10⁻¹⁷ script

But information is not rigidity. Pure records produce no restoring force on the clock field — the noise lands in the conjugate sector and nothing pulls back. We briefly dressed this gap in thermodynamic language ("a finite-temperature KMS channel") and then audited our own claim: the channel's stationary state is a pure squeezed vacuum, maximally non-thermal; detailed balance is violated at the 99% level. The KMS reading is retracted — and the audit caught a factor-2 error in our own drift formula while it was at it. What actually restores the field is not thermodynamics. It is the next section. script

3 · The closure: one bath writes the force and the reply

Model the record channel honestly — each comparison is mediated by a damped auxiliary mode — and eliminate the auxiliaries. The same bath susceptibility then returns two terms, tied by causality (Kramers–Kronig): the dissipator we already had, and a coherent Lamb shift, Ωeff − (i/2)γeff = −g²χ. The Lamb shift is the missing gear: it rotates the clock sector into the energy sector. Close the loop with one local coupling Hint = λΣεᵢhᵢ (status: candidate hypothesis, motivated by mass–energy equivalence of quantum clocks — Zych–Brukner, Pikovski, Castro-Ruiz — and by Luttinger's thermal potential, but not derived) and everything follows from one model:

Matter's energy makes nearby clocks run fast (derived, one line). The bath's coherent part transcribes the clock skew into an energy-density profile εss ∝ −L⁺(h−h̄) — the 1/q² Newtonian potential, from the drift kernel, with nothing Poissonian inserted (deviation from closed form: 10⁻¹³). Matter reads that profile through the same coupling that wrote it: attraction, Newton's third law to 10⁻¹³, sign set by the bath detuning — red detuning attracts, and switching the sign of the detuning dynamically flips attraction to repulsion. The control is exact: at zero detuning the coherent part vanishes and the readable field is identically zero while the noise survives. measured, full model vs closed form script

For the first time in this program, source, propagation and reaction came out of a single microscopic model. Statics was solved. Then we let things move.

4 · Dynamic matter: four fractures, one root

With mobile sources the mechanism keeps its force — bodies genuinely fall toward each other, the detuning still flips the sign, the zero-detuning control still nulls everything exactly — and loses its physics in four measured ways. measured script

Four failures, one sentence: the bath is a medium, and a medium has a frame.

5 · The no-go theorem

Is that fixable inside one scalar sector? No — and this is now a theorem, not a mood. For any single canonical pair with local couplings, stability, and the relational shift symmetry, read by matter through any symmetry-respecting local coupling: Newtonian statics (χ ~ 1/q²) and relativistic propagation (ω ~ q) are mutually exclusive. The proof is three lines (the shift symmetry forces the clock stiffness to open at q²; propagation demands an energy gap b₀ > 0; the Newtonian pole demands b₀ = 0). A sweep of 5,000 random stable theories found zero counterexamples; the unique Newtonian corner is exactly our n33 model, with z = 2; and the tradeoff carries an exact identity, ξ·ceff = 2Ωη — a microscopic cousin of the graviton-mass dilemma. derived + swept, 0/5000 script

So n34's four fractures were not parameter choices. They were the four visible faces of this theorem.

6 · The escape attempt: a vacuum that compares itself at every scale

The theorem assumes locality. One physically motivated way to break it: let the vacuum compare clocks not just between neighbors but across every distance ℓ, with equal information per octave — γ(ℓ) ∝ 1/ℓ, the Dyson-hierarchical profile, which is precisely how a critical vacuum distributes its entanglement. (To our knowledge this combination — quantum-Darwinist records with a Dyson scale ladder — had not been tried; it is also, incidentally, the native interaction profile of trapped-ion chains.) The result, measured: the dispersion obeys z = 1 + s and converges to z = 1.015 as the ladder approaches scale invariance — relativistic dispersion, bought honestly — and the ether drag drops fifty-fold. measured script

And the price, also measured, twice. The static law picks up a running correction — the force comes out Newton-with-a-logarithm, G(r) ~ 1/ln r, so the pre-registered p ∈ [1.95, 2.05] gate failed as written. Worse: signals arrive out of order — the 50%-arrival time at distance 48 precedes the one at distance 16. The nonlocal kernel has no light cone. We obtained the dispersion of relativity without its causality.

7 · The triangle

Impossibility triangle (status: conjecture, two corners proven/measured). For a single emergent scalar sector, of the three properties

    exact Newtonian statics · relativistic propagation (z=1) · a causal cone

you may have at most two. Local sectors: the first two exclude each other (theorem, §5). Scale-free nonlocal sectors: the first two coexist — up to a log — and the cone is destroyed (measured, §6).

Now hold the triangle up to general relativity, and it explains something usually taken as brute fact. GR has all three properties — and it contains an instantaneous piece: in Coulomb-like gauges the Newtonian potential responds with no delay. There is no paradox only because that piece is pure gauge — unobservable by fiat of the redundancy. That is the triangle's escape hatch, and as far as we can see the only one: keep the acausal part, and make it unmeasurable. A medium cannot do this; a gauge theory is exactly the kind of structure that can. Our week of toy models did not produce emergent gravity. It produced, from measurements, the reason gravity has no choice: any theory holding all three corners must contain a redundancy that hides the instantaneous sector — it must be a gauge theory. argued from the above

8 · The escape hatch, built and measured

If the triangle's only exit is a gauge redundancy, that claim should be demonstrable in the smallest canonical model that has one. So we built it: a 2D lattice of edge pairs (ae, πe) — a "clock connection" and its momentum — with the Gauss constraint Gi = (div π)i − ρi ≈ 0 imposed per site: the ledger required to close. Four results, one run. measured script

What this model does not do, named plainly: the constraint is posited, not derived — we wrote the ledger-closure in by hand; and the charge is external — real gravity's charge is energy, which must include the field's own. The ladder now reads, end to end: records → force (n33) · force → medium, four pathologies (n34) · medium → no-go (n35-A) · nonlocality → no cone (n36) · gauge → everything holds, but repulsive and posited (n37). One rung remains, and it is precisely stated: derive the Gauss constraint from the consistency of redundant records, at rank two.

9 · What survives, what it's worth

The program's question used to be can watching create weight? The answer stands: yes, measurably, with laws. The question this page leaves behind is sharper and harder: what kind of vacuum bookkeeping makes an instantaneous ledger unobservable? That is where gravity is hiding. It is a question about redundancy and records — which is to say, exactly the kind of question this program was built to ask.

Where this led. The rung named above — derive the constraint from record consistency, at rank two — was attacked head-on (n38–n39): the "ledger theorem" turned out to be textbook gauge theory renamed (our own critique, kept), but its failure exposed a contradiction between records and coherence whose resolution is measured and generative: a vacuum that reads syndromes of its own constraint protects physical superpositions exactly. The full assembly matured into a theory proposal with a cause, measured first gates — including the first mechanism of the whole program where a conserved Gauss flux comes out attractive — and named executioners: entanglement drainage →