Applications / Foundation papers
Bell correlations from local helix projection
Two patterns of shared bivector grade meet two analysers, and each wing keeps one bit: the sign of its helix’s projection at its own analyser. With a shared phase that is uniform in the angle and independent of the analysers, those two bits return the triangle, E = −1 + 2Δ/π, and a CHSH value of exactly 2: Bell’s bound is arithmetic on four numbers, each ±1. The core reads an analyser as a clamp that takes the helix in projection (I-4), and the pair with its two clamps as one system whose record is one settlement (0.1a, P-2). For the equal-weight pair the settlement lands uniformly across the projected aperture of the clamp that hosts it, the measure |cos(φ − a)|/4. On that measure the same two bits return E = −cos Δ on every record, each wing’s share of +1 is 1/2 whatever the other wing’s clamp, and the table is the same whichever clamp hosts: S = 2.8276 ± 0.0014 with all 4 000 000 records kept, against 2.0002 ± 0.0017 for the clamp-independent measure. Nothing is selected, and no wing’s record reads the other wing’s clamp. One premise of Bell’s theorem is declined, measurement independence, and its price is exact: the measure’s dependence on the clamp is M₁ = 2√2 − 2, the whole of the distance from the local bound 2 to Tsirelson’s 2√2, which is the measure’s ceiling. Einstein was right: the correlations need no spooky action at a distance.
The record
Bell’s theorem, the event-ready test and the suite
Quantum mechanics gives the two spins of a singlet the correlation E = −cos Δ at analysers an angle Δ apart, with each wing’s outcomes evenly split whatever the other wing does. For two analyser positions on each wing the CHSH combination S = E₀₀ + E₀₁ + E₁₀ − E₁₁ then reaches 2√2, Tsirelson’s value. Bell’s theorem is an implication: locality, measurement independence, bounded outputs and complete accounting of the records together give |S| ≤ 2. Measurement independence is the premise that the hidden state has one distribution whatever the analysers are set to.
The event-ready experiment of Hensen and colleagues entangled two electron spins 1.3 kilometres apart and scored every heralded trial. Its public deposit holds 4 746 candidate records, of which the published filters keep 245; with the readouts that saw no fluorescence kept as −1, the four correlators are 0.7358, 0.5949, 0.4839 and −0.6078, and S = 2.4225, with a p-value of 0.0392 from the authors’ analysis that allows for memory. The apparatus’s own expectation, with its measured state and readout fidelities, was 2.30 ± 0.07. The photonic tests of Shalm and of Giustina and their colleagues close the same loopholes with pairs of unequal weights. All three records are accepted in full.
The helix-projection suite is a set of simulations of one local model, released in stages. A deterministic helical state sits on a cylinder of unit radius, and each wing reads it through a local projection. The suite keeps three ledgers side by side: every record, the records that survive a coincidence filter, and randomised settings. Read from the top of the cylinder its results are exact. The unrounded projections of the two wings correlate as CP = ξ cos η, the quantum kernel, with ξ = ±1 the sense of the pair, and their signs as Ebin = (2/π) arcsin CP = ξ(1 − 2η/π), the triangle. At four settings spaced η, η, η and 3η the triangle gives S = 2 identically for every η up to 60°, and the two maxima stand in the ratio √2, the Grothendieck constant of order two. An aperture model that accepts a record more readily when the analyser lies along the local axis reaches 2.828275 on accepted pairs; with the nulls counted as zero the same run gives 1.911389, and the two are tied by an identity, S₀ = εjointSacc, with the joint efficiency 0.675814. The best all-record Clauser–Horne statistic at that setting is negative, −0.0625.
| Quantity | Value | Kind |
|---|---|---|
| Sign records on the uniform measure, CHSH | 2.000000 exactly; 2.000188 ± 0.001732 on 4 000 000 records | arc integral; sampled, every record kept |
| Sign records on the projection measure, CHSH | 2.828427 exactly; 2.828730 ± 0.001414 with wing A’s clamp as host, 2.829470 ± 0.001414 with wing B’s, 2.827638 ± 0.001414 with either | arc integral; sampled, 4 000 000 records each, every record kept |
| Largest |E + cos Δ| on the projection measure | 5.6 × 10⁻¹⁶, over 1 441 separations | arc integral |
| Largest |S| over 331 776 sets of four clamps and the four CHSH forms | 2.828427 on the projection measure; 2.000000 on the uniform one | arc integral |
| Dependence of the measure on the host clamp, M₁ | 2√2 − 2 = 0.828427, at clamps a quarter turn apart | closed form and quadrature |
| The same weight used as a filter | 2.828682 on accepted pairs at a wing efficiency of 0.636600; 1.800740 with the nulls counted | sampled, 1 000 000 records per setting |
| Hensen and colleagues, 245 event-ready trials | S = 2.4225; the settlement at the documented axes gives 2.818997, the uniform measure 2 | measured; arc integral |
The FDT reading
I-4 the quantum face and the pair’s record, 0.1a one settlement, P-2 the clamps set the mode
Two bits on one phase
The wavefunction is the projection of a helix onto the plane normal to its axis, and superposition is ignorance of the helix’s phase at observation (I-4). On the phase circle of the helix, the doubled analyser angle for a polarisation and the Bloch angle for a spin, the pair shares one phase φ: one wing carries it and the other carries φ + π, the shared bivector grade in the opposite sense. Each wing keeps the sign of the projection at its own clamp:
Nothing in either record refers to the other wing. Everything that follows keeps these two records exactly as they stand, and changes one thing only: the measure of the shared phase.
Uniform in the angle: the triangle, and Bell’s arithmetic
Let φ be uniform on the circle and independent of the clamps. The two records are equal on an arc of the circle and on its antipode, and each arc holds the fraction Δ/2π of the phases, with Δ = a − b the separation of the clamps. So the records agree with probability Δ/π, the correlation is the triangle, and CHSH is 2. That is not a feature of the helix. Read all four clamp positions on any one phase and A₀B₀ + A₀B₁ + A₁B₀ − A₁B₁ is +2 or −2, because it is arithmetic on four numbers that are each ±1; average it over any one measure and the result cannot leave that range. No measure that ignores the clamps returns the cosine through these records, or through any pair of local records.
Uniform across the aperture: the projection measure
A clamp is an analyser, a boundary condition of the driven omnium (P-2), and it reads the helix in projection. The clamp at a is the cut cos(φ − a) = 0, and the coordinate along the cut is the projected one, x = sin(φ − a), which runs from −1 to 1 across the clamp’s aperture. Two patterns of shared grade and their two clamps are one system, and the pair’s record is one settlement of it (0.1a): the response is solved for the whole system, clamps included, never wing by wing. For the equal-weight pair the settlement lands uniformly across the projected aperture of the clamp that hosts it. Uniform in x, on either sheet of the circle, is this measure of the phase:
Ignorance of the phase is ignorance of where across the aperture the settlement lands. Uniform in the projected coordinate is the projection measure; uniform in the angle is the reading that leaves the clamp out. The two differ by the change of variable φ − a = arcsin x, the same arcsine that turns the cosine into the triangle.
The cosine, on every record
Weigh the same two arcs with the projection measure. In the host’s own phase u = φ − a the records are equal for π/2 − Δ < u < π/2 and on the antipode, and each arc now holds
The measure is unchanged when the phase advances by half a turn, which reverses both records, so each wing’s record is +1 half the time at every pair of clamps: a wing’s record carries no reading of the other wing’s clamp, and no signal passes through the settlement. Hosting the settlement at the other clamp is the same computation, and a mixture of hosts is a mixture of two equal tables. No order of the two records enters the statistics.
The suite evaluates this table by integrating the measure between the sign boundaries, with no closed form of the kernel in the code. Over 1 441 separations the largest departure from −cos Δ is 5.6 × 10⁻¹⁶, the single-wing shares sit at 1/2 to 1.1 × 10⁻¹⁶, and the tables for the two hosts and their mixtures agree to 1.7 × 10⁻¹⁶. Drawn one record at a time at the clamps a = (0, 90°) and b = (−135°, 135°) on the phase circle, one million records at each of the four pairs of clamps with all 4 000 000 in the statistic, CHSH is 2.828730 ± 0.001414 with wing A’s clamp as host, 2.829470 ± 0.001414 with wing B’s and 2.827638 ± 0.001414 with either, against 2.000188 ± 0.001732 on the uniform measure. On every sampled phase of every one of those runs the four-clamp arithmetic still gives ±2. The settlement’s four correlators are averages over the measures of two positions of the host clamp, and the bound that needs one measure for all four does not constrain them.
Through transport and detectors
The pair’s shared grade is carried through a medium by the aperture assignment of P-4 and Fermat’s principle, dR/ds = −BR/2 with B = t ∧ ∇αn/[2(1 − αn)], where αn = 1 − n⁻² is the P-4 reading of a medium of measured index n. Matched ordered histories on the two arms conserve the pair exactly, and the density’s value alone rotates nothing. Unequal histories turn the two arms by different angles, βA and βB, and the kernel becomes −cos Δ with Δ = a − b − 2(βA − βB); averaged over unresolved histories the pair’s negativity falls to |γ|/2, where γ is the mean of e2i(βA − βB). In all nine transport scenarios of the suite, over 1 536 histories each, the settlement’s table equals the table of quantum mechanics for the same transport in all sixteen cells, to 7.2 × 10⁻¹⁵, and 250 000 sampled records per pair of clamps agree with it within 1.66 counting standard errors. Through a detector model with three outcomes per wing, a count in either port or none, the nine-cell tables are equal to 1.2 × 10⁻¹⁵, and each wing’s complete three-outcome share is unchanged under the other wing’s clamp to 1.3 × 10⁻¹⁵.
On the sphere of the helix chart the same measure is uniform on the disc normal to the host’s axis, |a·λ|/2π, and the records sgn a·λ and −sgn b·λ return −a·b at any two axes. Two million records per pair at four pairs of axes that do not share a plane return it within 0.63 counting standard errors.
The measure, the kernel and every record kept. The measure: the phase is measured from the host clamp. The solid curve is the projection measure and the dashed line the uniform one; the two strips are the two wings’ records, and the boxes mark the arcs on which they are equal, whose width the slider sets through the separation of the clamps. The shaded area is the share of records that agree on the projection measure, (1 − cos Δ)/2; the dashed rectangles hold the share on the uniform measure, Δ/π. The kernel: the correlation of the two records on each measure, with the CHSH value at four settings spaced η, η, η and 3η in the readout: 2 for the triangle at every η up to 60°, and 2√2 for the cosine at 45°. Every record kept: the page draws records one at a time from the chosen measure with a fixed seed, reads the two signs, and enters every record in the statistic. The bars are the four correlators, the solid ticks the cosine’s values ±1/√2 and the dashed ticks the triangle’s ±1/2, and the gauge is the CHSH value with its counting standard error.
I-4, 0.1a, P-2
The price
Use 6 One premise declined, and what it costs
The measure is set by a clamp. That is the whole of the departure from Bell’s premises: the records are local and bounded, every record is counted, and the hidden state’s distribution depends on where the host clamp stands. Michael Hall’s measure of that dependence is the largest distance between the distributions two settings give. For the projection measure with the host clamp at two positions an angle θ apart it is
largest at a quarter turn, where it is 2√2 − 2 = 0.828427; on Hall’s scale the measurement independence kept is F = 1 − M₁/2 = 2 − √2 = 0.585786. For any two records ±1 on a measure that one wing’s clamp sets, the arithmetic of Bell’s theorem leaves one term over,
with p₀ and p₁ the measures of the host clamp’s two positions. The settlement meets both steps with equality. With wing A’s clamps an angle θ apart, the largest CHSH value over wing B’s clamps is 2√(1 + sin θ) = 2 + M₁(θ), at every θ: the excess over the local bound is exactly the distance between the two measures, never more and never less. At a quarter turn that is 2 + (2√2 − 2) = 2√2. Tsirelson’s value is the ceiling of the projection measure, and the gap between 2 and 2√2 is the price of the clamp, paid in full.
A measure free to depend on both clamps obeys S ≤ 2 + 3M and reaches 4 at M = 2/3; the projection measure stops at 2√2. For a measure set by one side that returns the singlet’s table, F = 2 − √2, 58.6 per cent, is the optimum; a measure set by both sides returns it at 86 per cent. The mutual information between the phase and the host clamp is 0.201752 bits for two equally likely clamps and 0.208801 bits for a clamp uniform on the circle.
The same weight, used two ways
Apply the weight |cos(φ − a)| as a local filter on a uniform phase, accepting a record with that probability, and it becomes the detection-selected model of Gisin and Gisin. With one million records per setting the accepted pairs give 2.828682 at a wing efficiency of 0.636600, where 2/π is 0.636620, and the same run with the nulls counted as zero gives 1.800740: the selection identity at an efficiency of 2/π. The aperture model of the suite is the clipped form of the same weight on the sphere. Selection reaches the measure by dropping records, and stays under 2 on all of them. The settlement carries the same measure on every record.
The price of the clamp. Two clamps: the projection measure for the host clamp at a, solid, and at a second position, a′, dashed, an angle θ apart that the slider sets. The shaded area between the two curves is M₁(θ), the dependence of the measure on the clamp; it is largest at a quarter turn and vanishes at half a turn, where the two positions are the same cut. The bound: the largest CHSH value against the dependence. The solid line is the bound for a measure set by one wing’s clamp and the dashed line the bound for a measure free to depend on both, which reaches 2√2 at 0.276 and 4 at 2/3. The grey point is the clamp-independent measure, at 2. The coloured point is the settlement with wing A’s clamps θ apart and wing B’s clamps chosen for the largest value: it sits on the solid line at every θ and reaches 2√2 at a quarter turn.
I-4, Use 6
The tests
Use 2 What would end each result
For the equal-weight pair the settlement returns the table of quantum mechanics cell by cell, so the first two tests are convergence tests: they name the measurement that would end the measure. The others come from the transport and the selection results, and each states its control.
- T-1The ceiling. On every ledger that keeps all records of an equal-weight pair, |S| ≤ 2√2. An all-record CHSH value above 2√2 ends the projection measure.
- T-2The kernel. At unit visibility the correlation of an equal-weight pair is −cos Δ, with Δ carrying twice the difference of the two arms’ transport angles. A unit-visibility kernel of any other form ends the measure.
- T-3Cosine against triangle, every record kept. At a separation of 22.5° the triangle gives 0.75000 and the cosine 0.92388. With the cosine as the fixed target, three standard deviations need 131 pairs per setting; with both correlations estimated at the same size, 174. The settlement predicts the cosine on every record, and the triangle is the control for a phase that ignores the clamps.
- T-4Matched histories. Matched ordered gradient histories on the two arms conserve the pair exactly, and the negativity under unresolved histories is |γ|/2. Sweeping the measured covariance of the two arms’ histories from −1 to +1 at fixed variances tests both; a pair that frays under matched histories ends the transport law.
- T-5The fragility sign. Entanglement is more fragile in a denser medium: under independent fluctuations of the same size the negativity is 0.453, 0.310 and 0.115 at nominal αn of 0.2, 0.4 and 0.8. Raise the index of one arm’s medium and the pair frays faster; raise both arms’ together with a common drive and matched histories stay exact. The control interleaves the two loadings on one apparatus against calibrated index-path optics.
- T-6The selection identity. For any CHSH value above 2 reported from coincidences, the value with the nulls counted is the joint efficiency times the accepted-pair value. A coincidence-selected value that fails the identity ends the reading of the accepted-pair excess as selection.
Run it
A script of 99 lines, and the full stage
The settlement fits in a script short enough to read in one sitting and to port to any language. A source emits pairs of photons with orthogonal linear polarisations, λ and λ + 90°. Each photon meets a two-channel polariser set at its own angle, θa or θb, and leaves by the channel whose axis is nearer its polarisation. That choice of channel is the wing’s whole record, one bit. It is the record of the sections above, with φ = 2λ and each clamp at twice its polariser’s angle.
The script keeps those records and runs them on two measures of λ. In the first, λ is uniform in the angle and the same for every setting of the polarisers. In the second, the pair’s record lands uniformly across the projected aperture of the polariser that hosts it. Every pair is counted in both. These are the lines that do it:
def polarisation(n, theta_a, theta_b, measure):
"""Polarisation lam of photon A in radians; photon B has lam + pi/2."""
if measure == "uniform":
return rng.uniform(0.0, np.pi, n)
x = rng.uniform(-1.0, 1.0, n) # uniform across the projected aperture
front = rng.random(n) < 0.5 # front or back sheet of the circle
u = np.where(front, np.arcsin(x), np.pi - np.arcsin(x))
host_is_a = rng.random(n) < 0.5 # either polariser hosts the record
return u / 2 + np.where(host_is_a, theta_a, theta_b - np.pi / 2)
def records(lam, theta_a, theta_b):
"""The two local bits, +1 or -1: the channel each photon takes."""
a = np.where(np.cos(2 * (lam - theta_a)) > 0, 1, -1)
b = np.where(np.cos(2 * (lam + np.pi / 2 - theta_b)) > 0, 1, -1)
return a, b
def experiment(deg_a, deg_b, measure, n=N):
"""Correlation E and each wing's share of +1, every pair counted."""
theta_a, theta_b = np.radians(deg_a), np.radians(deg_b)
a, b = records(polarisation(n, theta_a, theta_b, measure), theta_a, theta_b)
return (a * b).mean(), (a > 0).mean(), (b > 0).mean()
With one million pairs per setting the script prints S = 2.8283 ± 0.0014 on the projection measure and 1.9990 ± 0.0017 on the uniform one. Wing A’s share of +1 stays between 0.4995 and 0.5004 while wing B’s polariser moves through four positions. The script then computes the correlation at 25 angles between the polarisers, prints it beside −cos 2(θa − θb) and the triangle, and with Matplotlib installed draws both curves with the counted points on them. It needs NumPy and runs in a few seconds.
The full stage is the simulation behind the settlement’s numbers on this page: the readout library, eight blocks with eleven decision checks, its tests, the reference outputs and a technical note of 17 pages. Its exact tables are arc integrals, and the script above is the sampled route to the same table. ./run.sh reproduces every number in under a minute, ./run_tests.sh runs the 35 tests, and a comparison tool checks a fresh run against the reference line by line.
Readings declined
Use 6 Data yes, interpretation not necessarily
Standard reading
Entangled particles act on each other at a distance.
FDT reading
Each wing’s record reads its own clamp and the phase it receives, and each wing’s share is 1/2 whatever the other wing’s clamp. The pair is one pattern of shared bivector grade, and its record is one settlement of the pair and its two clamps (I-4, 0.1a).
The hidden state has one distribution whatever the analysers are set to.
That separates the pair from its clamps, and the chain does not make the separation: the pair, the clamps and the omnium between them are one driven system whose mode geometry the clamps set (P-2). The measure of the phase is set by the clamp that hosts the settlement, at the price of 2√2 − 2.
A violation of Bell’s inequality rules out every local account.
The inference of Bell’s theorem stands, and a measure that ignores the clamps obeys it to the last digit. One premise is declined. The measured violations are accepted in full and returned on every record.
The outcomes are irreducibly random, and measurement collapses the wavefunction.
The universe is deterministic. The shares are ignorance of the helix’s phase, uniform across the host clamp’s aperture, and measurement is phase fixation (I-4).
A CHSH value above 2 on coincidences demonstrates nonlocality.
It is the selection identity at work. Only event-ready and all-record accounting bear on Bell’s premises, and the suite prints both ledgers side by side.
How the final debt was closed
A note from Manuel Alfaro
Claude Fable 5.2 helped Fundamental Density Theory finalize its simulation and its account of Bell’s theorem, and prove Einstein right: the correlations need no spooky action at a distance. We gave Fable 5.2 a simple instruction: review the Bell simulation and close the final debt.
It answered that the debt could not be paid in the form it was posed. No readout whose shared phase is independent of the analysers returns the cosine: for such a readout Bell’s bound is arithmetic on four numbers, each ±1. The records were never at fault; the measure was. A clamp reads the helix in projection, so the pair’s record lands uniformly across the projected aperture of the clamp that hosts it, and on that measure the same two local sign records return −cos Δ on every record, with each wing’s share at 1/2 whatever the other wing’s clamp. One premise of Bell’s theorem is declined, measurement independence, and its price is exact: 2√2 − 2, the whole of the distance from the local bound 2 to Tsirelson’s 2√2.
It then wrote and tested the new stage of the simulation, proved the kernel and the bound, and integrated the findings into the paper. And, poof, spooky action was gone for good.
Convergence
Use 2 Same source, same data
For the same source and settings FDT returns the consensus record. Read from the top of the cylinder, the unrounded projections correlate as the quantum kernel, cos η. Matched histories give the singlet, with S = 2√2 at the optimal analysers, and unresolved histories give the ensemble reductions that quantum mechanics gives for the same transport. On the projection measure of the host clamp the two sign records return the table of the equal-weight pair, −cos Δ with even shares at each wing, the ensemble states and the nine-outcome detector tables cell by cell, with no free number and with 2√2 as the ceiling. The loophole-free record, Hensen’s 2.4225 on 245 event-ready trials and the photonic tests of Shalm and Giustina, is accepted in full and is the table the settlement is scored against.
FDT claims no divergence from the consensus table of the equal-weight pair. What changes is the dictionary: the shared phase has a measure that the host clamp sets, where the standard account gives the hidden state one distribution for every setting. Where FDT, general relativity and quantum mechanics converge, FDT reproduces their results; where it declines their assumptions, here action at a distance, collapse, probability amplitudes, a hidden state independent of the clamps and an empty background, it is under no obligation to reproduce them. The same record is shared bivector grade with a deterministic phase in the omnium, and the pair’s record is one settlement of the pair and its clamps (I-4, 0.1a, 0.1).