Applications / Experiments read through FDT

Orbital Nernst transport under a clock-factor gradient

Wen-Yuan Li, Chun-Chi Chang, Ssu-Yen Huang and Danru Qu report the first experimental evidence for the orbital Nernst effect, in which a temperature gradient drives a transverse orbital current, by measuring the magneto-thermopower of yttrium iron garnet coated with platinum and naturally oxidized copper. Set beside platinum alone on the garnet, the coated stack’s magneto-thermopower and magnetoresistance are enhanced, the platinum layer’s effective ratios coming to about twice and nearly three times the reference values on the authors’ shunting estimates, and they attribute the enhancements to the orbital Nernst and orbital Hall effects in the oxidized copper. FDT reads the temperature gradient as a gradient of the conduction electrons’ clock factor, the transverse orbital and spin currents as flows of the bivector grade, the applied field as a helical density gradient, and a conversion and its inverse as two system-wide settlements. On the source’s own inputs FDT returns every ratio, estimate, angle and conductivity it prints for its samples with 3 nm of platinum, an identity map that corroborates no invariant. Five tests follow: a platinum-coated element twisted in zero applied field, on the garnet and on a non-magnetic garnet; copper oxidized in oxygen-18 against oxygen-16, read on the thermal and on the electrical enhancement; and the enhancement of a conversion set against that of its inverse on one slab.

The record

What the article reports

In the source’s account the generation of orbital currents does not rely on spin-orbit coupling and is typically more efficient than that of spin currents, which broadens the range of materials for manipulating angular momentum. In the orbital Hall effect and the orbital Rashba–Edelstein effect a charge current generates a transverse flow of orbital angular momentum through a momentum-space orbital texture, the latter induced at an interface such as one between an oxide layer and a metal; for naturally oxidized copper, where oxygen is distributed non-uniformly, the source takes from its reference 7 that the key ingredient is the local Cu–O hybridization rather than a well-defined interface, and it includes the Rashba–Edelstein effect implicitly in its discussion of the orbital Hall effect. The thermal counterpart, in which a heat current drives a transverse orbital current, is the orbital Nernst effect. The source writes that it has not yet been experimentally established, and that it is accompanied by the spin Nernst effect, which generates a transverse spin current upon a heat current.

The samples are YIG/Pt(3)/CuOx(2), YIG/Pt(3) and YIG/CuOx(5), the numbers nominal thicknesses in nanometres, each capped with 2 nm of SiO₂, on polycrystalline yttrium iron garnet slabs typically 2 mm wide, 10 mm long and 0.5 mm thick. The metals are deposited by DC magnetron sputtering; after the copper is deposited the sample is removed from the chamber and the copper naturally oxidized in air for two days, and on a prior report the oxide forms no distinct Cu/Cu₂O/CuO interfaces but a gradual depth-dependent oxygen distribution. The garnet is an insulating ferromagnet, which, the source writes, eliminates any anisotropic magnetoresistance or anomalous Nernst contributions, so that the observed signal arises solely from the nonmagnetic metallic layers.

In the source’s account, a charge current or a temperature gradient along the film generates a transverse spin current in the platinum through the spin Hall or the spin Nernst effect, and a transverse orbital current is expected to be generated in the oxidized copper through the orbital Hall and orbital Nernst effects. The orbital current injects into the platinum, is converted there into additional spin current through the platinum’s spin-orbit coupling, and modifies the spin accumulation at the garnet/platinum interface. With the spin parallel to the garnet’s magnetization the total spin current is reflected, and with the spin perpendicular to it the current is absorbed; the reflected spin current converts back into a charge current through the inverse spin Hall effect, so the spin and orbital currents can be detected as changes of the magnetoresistance and of the magneto-thermopower.

A heat flow along x, its temperature gradient printed as 1.9 ± 0.1 with the unit mm/K, sets up the Seebeck voltage, measured along x with the in-plane field along x or along y and swept from 700 to −700 Oe, a range the source writes is sufficient to align the garnet’s magnetization along the field. Every Seebeck coefficient is corrected for the connecting copper wires, +1.68 μV/K. For platinum alone the source extracts 𝒮Pt = 2.240 ± 0.006 μV/K and the contrast Δ𝒮 = 𝒮x − 𝒮y = 0.99 ± 0.07 nV/K between the field along x and along y, a ratio of (4.42 ± 0.31) × 10⁻⁴, and, through the spin Nernst model with a spin diffusion length of 1.5 nm and a spin-mixing conductance of 4 × 10¹⁴ Ω⁻¹ m⁻² from the literature, a spin Nernst angle of −12.5 ± 1.0%. Under a charge current of 10 mA its sheet resistance is 202.5 ± 9.1 Ω and ℛx − ℛy = 0.021 ± 0.001 Ω, a ratio of (1.04 ± 0.05) × 10⁻⁴, and a fit with the spin Hall magnetoresistance model gives a spin Hall angle of +2.9 ± 0.1%.

With the oxidized copper on top, the Seebeck coefficient of the whole sample changes sign, to −0.608 ± 0.009 μV/K, and the sheet resistance falls to 111.6 ± 6.7 Ω. Through a parallel-resistance model the source estimates the oxide layer’s own sheet resistance at 248.5 ± 35.7 Ω and its Seebeck coefficient at −4.10 μV/K, of sign opposite to platinum’s. The stack’s contrasts are 0.89 ± 0.09 nV/K and 0.018 ± 0.001 Ω, ratios of (−1.46 ± 0.09) × 10⁻³ and (1.59 ± 0.10) × 10⁻⁴. If the oxide contributes only a shunting effect, the platinum layer’s effective ratios come to (1.09 ± 0.12) × 10⁻³ and (2.88 ± 0.45) × 10⁻⁴, about twice and nearly three times the reference values, and the source concludes that the enhanced spin-charge interconversion cannot be attributed solely to shunting and must originate from the orbital Nernst effect in the oxide, and that an additional angular-momentum contribution from the oxide must be considered in the magnetoresistance.

Through a simplified drift-diffusion model, with immediate orbital-to-spin conversion at the platinum/oxide interface, a negligible inverse orbital Hall effect in the oxide and strong spin relaxation in the platinum, the source estimates for the oxide an orbital Nernst angle of +22.5 ± 6.8% and conductivity of −0.93 ± 0.34 (ħ/e)(A m⁻¹ K⁻¹) and, under similar assumptions, an orbital Hall angle of +11.8 ± 3.6% and conductivity of 1187 ± 400 (ħ/e)(Ω⁻¹ cm⁻¹), each an effective value of the oxidized layer treated as one layer. YIG/CuOx(5) shows a negligible magnetoresistance ratio, below 5 × 10⁻⁶ at a comparable sheet resistance of 188.5 Ω, which the source reads as indicating that the oxide itself cannot efficiently generate a spin current to account for the stack’s enhanced magnetoresistance. With 5 nm of platinum the estimated effective ratios are comparable to those of platinum alone, and across 2 to 5 nm the coated samples’ ratios rise markedly as the platinum thins, which the source reads as orbital-to-spin conversion at the interface decaying rapidly within the platinum. Every error bar is the propagated uncertainty of the sample length, an effective 9 ± 1 mm for a nominal 10 mm.

The source’s values for its samples, each with the verb the source gives it. The source gives its error bars as the propagated uncertainty from the sample length and prints no statistical or model uncertainty.
Sample and quantityValueIts verb
YIG/Pt(3): Seebeck coefficient 𝒮Pt; contrast Δ𝒮; ratio Δ𝒮/𝒮2.240 ± 0.006 μV/K; 0.99 ± 0.07 nV/K; (4.42 ± 0.31) × 10⁻⁴extracted
YIG/Pt(3): spin Nernst angle ΘSN; conductivity 𝒩SN−12.5 ± 1.0%; −0.23 ± 0.02 (ħ/e)(A m⁻¹ K⁻¹)extracted; found
YIG/Pt(3): sheet resistance ℛ; contrast Δℛ; ratio Δℛ/ℛ202.5 ± 9.1 Ω; 0.021 ± 0.001 Ω; (1.04 ± 0.05) × 10⁻⁴is
YIG/Pt(3): spin Hall angle ΘSH; conductivity 𝒞SH+2.9 ± 0.1%; 238.5 ± 13.5 (ħ/e)(Ω⁻¹ cm⁻¹)obtained by fitting; obtained, Table 1
YIG/Pt(3)/CuOx(2): 𝒮tot; Δ𝒮tot; ratio−0.608 ± 0.009 μV/K; 0.89 ± 0.09 nV/K; (−1.46 ± 0.09) × 10⁻³changes sign; is
YIG/Pt(3)/CuOx(2): ℛtot; Δℛtot; ratio111.6 ± 6.7 Ω; 0.018 ± 0.001 Ω; (1.59 ± 0.10) × 10⁻⁴is reduced; is
The CuOx(2) layer: ℛCuOx; 𝒮CuOx248.5 ± 35.7 Ω; −4.10 ± 0.61 μV/Kestimated
The platinum layer, shunting only: effective ratios 𝔮th; 𝔮el(1.09 ± 0.12) × 10⁻³; (2.88 ± 0.45) × 10⁻⁴estimated; “about twice”; “nearly a factor of three”
The CuOx(2) layer: orbital Nernst angle ΘON; conductivity 𝒩ON+22.5 ± 6.8%; −0.93 ± 0.34 (ħ/e)(A m⁻¹ K⁻¹)estimated
The CuOx(2) layer: orbital Hall angle ΘOH; conductivity 𝒞OH+11.8 ± 3.6%; 1187 ± 400 (ħ/e)(Ω⁻¹ cm⁻¹)estimated
YIG/CuOx(5): magnetoresistance ratio< 5 × 10⁻⁶ at 188.5 Ωobserved; negligible
5 nm of platinum: effective ratios, with and against YIG/Pt(5)(8.54 ± 0.84) × 10⁻⁴ and (1.20 ± 0.15) × 10⁻⁴; (8.91 ± 0.64) × 10⁻⁴ and (0.99 ± 0.45) × 10⁻⁴estimated; directly obtained; “comparable”

The FDT reading

0.1c The medium’s clock factor, I-⁠3a the bivector grade, P-⁠5 helical density

There is one substance, the omnium: matter, energy and spacetime are three readings of it, and the chain E/m = c² = 1/(ε₀μ₀) = (d/t)² is one identity, read by quantum mechanics, electromagnetism and general relativity, its four terms adjusting together and settled by one global optimisation (0.1, 0.1a). c is the strand rate, derived from the omnium’s two responses, c² = 1/(ε₀μ₀) (0.1d).

A pattern’s kept time t and thread length d rescale together by a positive clock factor 𝔪, with d/t = c kept on the spot. In the metric the rescaling is gμν ↦ 𝔪²gμν, and with 𝒰 the pattern’s four-velocity and 𝔓μν the projector across its path, the distant bookkeeper reads the free pattern accelerating toward smaller 𝔪, where its clocks run slower, while on its own clocks and rulers nothing changes:

∇𝒰𝒰μ = −𝔓μν∂ν ln 𝔪the bookkeeper’s reading

Moving at constant speed, on the pattern’s own clocks and rulers, through a density gradient is gravity: the extended equivalence principle (0.1c). In a medium the same equation runs on the medium’s own clock factor. For a medium of energy density E/V and pressure P at zero net conserved count, the conservation of its stress tensor projected across the path and the Gibbs–Duhem relation give

(E/V + P) ∇𝒰𝒰μ = −𝔓μν∂νP,   dP = (E/V + P) dT/T⇒ ∇𝒰𝒰μ = −𝔓μν∂ν ln Tevery equation of state P(T)

so the medium’s patterns are free of every force in T²gμν: the temperature is their clock factor, and with a conserved count N and fixed entropy per count the clock factor is the enthalpy per count, (E + PV)/N (0.1c). The paragraph holds for a stress tensor of perfect form, and the analysis declares its domain once: the conduction electrons of the platinum and oxide films form a medium with a conserved count, and on the clause’s two premises, a stress tensor of perfect form and fixed entropy per count, that medium’s clock factor is its enthalpy per count. The analysis reads the source’s temperature gradient as a gradient of that clock factor. A medium’s clock factor rescales the metric in which its flow is free, never the time kept by a clock the medium carries; it never stands in an α, odds, force, optical or density slot and never bends a stroke, and no number is taken from the clause.

The deviation map ∇𝒰 of a family of patterns splits into a trace, the expansion; an antisymmetric part, the vorticity, which is the bivector grade, the mechanical circulation whose magnetic projection is P-⁠5’s; and a symmetric trace-free part, the shear (I-⁠3a). The analysis reads the source’s transverse orbital angular momentum flow and its spin current as flows of the bivector grade, circulation, of the family of conduction electrons: where their drift under a drive along a film varies with depth, the family’s deviation map has an antisymmetric part whose axis lies in the film plane, across the current. In a medium the same equation runs on the clock factor, and the second derivatives of ln T set the deviation of the medium’s patterns at the medium’s own strength, while their gravitational window is small (I-⁠3a). The source prints no depth profile of its own films, taking the oxide’s gradual oxygen distribution from its reference 19 and the local Cu–O hybridization from its reference 7, so the reading gives the antisymmetric part no magnitude and no sign and declares no kernel.

A magnetic field is a macroscopic helical density gradient of the omnium: the bivector grade of the density field coherently aligned along an axis. Imposing a physical twist, torsion or helical geometry, on a material is establishing a magnetic field, and mechanical vorticity and magnetic moments are different grade projections of one density-gradient topography, with no carrier exchange (P-⁠5). The analysis reads the source’s applied field and the garnet’s magnetization as helical density gradients, and the sweep that aligns the magnetization along the in-plane field as turning the garnet’s held moments the way a uniform twist does, pushing nothing. The garnet is a closed pattern whose rest energy depends on where it sits in a twist:

E = E₀ + Eind − Σ 𝗆·𝖡,   𝓜 = E/c²,   𝔪 = E/E₀the helical window

with 𝗆 its held moments, 𝖡 the twist’s magnetic field where they sit, Eind the energy of the moments the twist induces in it, the work of inducing them included, and E₀ its rest energy outside every twist (0.1c, 0.5). The universality of free fall belongs to the gravitational window and stands, and a uniform twist, a uniform level of 𝔪, turns a held moment and pushes nothing. No 𝔪 = E/E₀ is written for a current or a stroke, and no frequency is read off a temperature.

The identity photon = graviton = gluon = phonon = magnon is one spin-1 excitation of the density field, read at different density and geometry regimes, and the magnon is the stroke through a helical density gradient (I-⁠2). The analysis reads the strokes of the source’s heat current, which can flow in-plane and out-of-plane through the whole sample, as the one carrier’s: phonons in the films and the slab, and magnons through the garnet’s helical density gradient. The films’ conduction electrons are closed patterns and stand outside the identity. “SOC-mediated” stays the source’s adjective for a conversion and for an orbital current; in FDT no carrier is exchanged.

Matter is braided photon helices, and there are no intrinsic particle masses or properties: charge, colour, flavour, generation and mass are derived, the strand count n setting the topological invariants and confinement the mass (I-⁠5). The analysis reads the conduction electrons as closed n = 3 braids, their charge magnitude Q = n mod 2 fixed by the strand count, and the nuclei of 18O and 16O as two patterns that differ by two n = 2 closures, the neutron of I-⁠5; the confinement operator D, its domains, boundary data and spectrum are to be derived (O-⁠9). The source’s “orbital degree of freedom” and its “spin” stay its words: no strand count and no α is assigned to either. Quantum mechanics is helical geodesics on truncated density cones, and the universe is deterministic (I-⁠4).

Mass, energy, temperature and pressure enter one Rs, every channel once, and thermal, isotope and pressure shifts are one mechanism through 𝓜 and Rs (I-⁠7):

𝓜 = M + E/c² + PV/c² + aT⁴V/c² + …a = 4σSB/c,   Rs = 2G𝓜/c²each channel once

For a film or the slab the rest mass enters once; E/c² carries the heat the gradient adds, the energy of the drive current and, for the garnet, the held moments’ −Σ 𝗆·𝖡 and Eind; PV/c² enters at the ambient pressure, and the thermal term never carries the flux constant. No number is printed for an entry, and binding is read from the solved geometry, never added as a local term.

Every step is a system-wide settlement: the global optimisation is solved at every step, from the state the system is in; events do not commute, and a process run backward is never the mirror of the forward event (0.1a). The source’s stages, a spin current generated, an orbital current injected and converted, the total spin current reflected or absorbed and the reflected current converted back into a charge current, and its drift-diffusion model with its three assumptions are the source’s and are attributed whole. The analysis reads the stack’s response to a drive as settled system-wide at every step: the conversion of a charge current into angular momentum at the garnet/platinum interface and the conversion of angular momentum arriving there into a charge current are a process and its run backward, two settlements, each from its own starting state. The source’s model enters the direct conversion in the oxide and leaves its inverse out, by its assumption (ii).

The source’s voltages are driven observables, zero at zero drive, and never recover the static force (I-⁠1). Its conductivities are signed dimensionful response coefficients and its angles quotients of two such, never a density and never run through a density map (I-⁠6), so no α, odds or bounded map is built here from an angle, a ratio or a conductivity. Every consensus datum about an interior is a profile-weighted mean with its probe weight named (I-⁠3): the analysis reads each film datum of the source as a mean over the film’s thickness, with the conductance weight of the source’s own parallel-resistance model for a stack’s sheet resistance and Seebeck coefficient, and the platinum within the spin diffusion length of the garnet/platinum interface for the two contrasts, and reads no point value off one.

X-⁠1, exploratory. The quotient of the source’s two effective angles of the oxide, ΘON/ΘOH = 1.9 ± 0.8 in quadrature, the two sharing their positive sign. The analysis reads both angles as measuring, per unit drive, one antisymmetric part of the deviation map of the oxide’s conduction electrons, under the clock-factor gradient for ΘON and under the charge drive for ΘOH, so their quotient re-describes how that part loads in the medium window against the charge drive; the source’s own reading is two effects of one effective single layer, with no relation imposed between them. The quotient carries no FDT number, rests on the source’s model assumptions and predicts nothing, and it is never counted with the tests.

The stack and its numbers. The stack: yttrium iron garnet coated with platinum, with the oxidized copper above it in the coated sample, or 5 nm of oxidized copper alone in the control, and silica on top, drawn far from scale. The drive is a temperature gradient along x, which the analysis reads as a gradient of the conduction electrons’ clock factor, their enthalpy per count on the declared domain (0.1c), or a charge current of 10 mA along x. In the source’s account the spin Hall or spin Nernst effect drives a spin current across the platinum and the orbital Hall or orbital Nernst effect an orbital current across the oxide, converted into spin at the platinum/oxide interface; the analysis reads both as flows of the bivector grade, an antisymmetric part of the electrons’ deviation map whose axis lies along y, in the film plane across the current, with no magnitude and no sign (I-⁠3a). The field along y sets the magnetization along the spin polarization, and the spin current is reflected; along x the two are perpendicular, and it is absorbed. Each sample’s readouts are the source’s printed values. The numbers: each ratio, estimate, angle and conductivity recomputed from the source’s printed inputs, as recomputed over printed, against the band of the printed uncertainty; at the source’s spin-mixing conductance of 4 × 10¹⁴ Ω⁻¹ m⁻² every row sits inside its printed uncertainty, the four angles move with the conductance, which sets their one figure, and the four conductivities are formed from the printed angles.

0.1c, I-⁠3a, P-⁠5, Use 2

What FDT predicts

Five tests, each against its comparator

Five tests stand where a core clause supplies what the consensus calculation lacks: P-⁠5’s field in a physical twist, an additional source, for T-⁠1 and T-⁠2; I-⁠7’s density channel, a constitutive input, for T-⁠3 and T-⁠4; and 0.1a’s two settlements, a changed observable dictionary, for T-⁠5. Each reads twelve values and decides by an exact sign-flip permutation test, whose null distribution is the set of means from all 2¹² = 4096 sign assignments of the twelve. Two-sided at p ≤ 0.0027 the rule admits the ten most extreme, a size of 10/4096 = 0.0024; one-sided at p ≤ 0.00135, the five most extreme, 5/4096 = 0.0012. Fair coins drawn before each session, pair or slab set every reading order and every starting polarity, assign the bars and the airs to the cells and choose the coated half, so drifts, residual states, deposit gradients and cell differences take signs the coins set, while the ambient leak and the consensus channels are carried by declared bounds: the tests rest on sign symmetry and those bounds, on no variance estimate. The apparatus and settings are the analysis’s design, not measurements in the source.

Two rules recur. The bounded test, for twelve values and a bound 𝔟: the mean lies outside ±𝔟, and the two-sided test of the values, each shifted by 𝔟 against the sign of the mean, rejects at p ≤ 0.0027; with no FDT term its probability is at most 10/4096 wherever the bounded term lies within its bound. The exclusion test, for a scale 𝔥 above 𝔟: a term of magnitude 𝔥 or more is excluded when the mean lies within ±(𝔥 − 𝔟) and the two-sided tests of the values shifted by +(𝔥 − 𝔟) and by −(𝔥 − 𝔟) both reject at p ≤ 0.0027, a false exclusion having probability at most 5/4096 wherever the bounded term lies within its bound.

  1. T-⁠1
    A platinum-coated garnet element twisted in zero applied field shows a hand-odd, reciprocity-odd resistance, zero in the consensus calculation. An element of the source’s polycrystalline garnet cut as a truncated cone, end-diameter ratio 2, wide end down, its upper face alone coated with platinum as the source coats YIG/Pt(3), with four contacts 1, 2, 3 and 4 in order round the edge of the film. It sits inside a magnetic shield, in a non-magnetic torsion clamp that carries no current and turns its two ends by +𝔴/2 and −𝔴/2 about the film normal, with 𝔴 = ±0.1°, its sign the hand. Observable: with ℛ13,24 the potential of contact 2 less that of contact 4 over an alternating current sent from contact 1 to contact 3, and ℛ24,13 the potential of 1 less that of 3 over a current from 2 to 4, each read at a drive frequency where it does not depend on the current, the reciprocity-odd resistance ℛodd = ½(ℛ13,24 − ℛ24,13) and the session contrast 𝔶tw = ½[ℛodd(+) − ℛodd(−)], in units of ℛsat, half the difference of ℛodd between the two senses of one saturating field along the axis. Twelve sessions: in each, ℛodd is read at both hands in an order a fair coin sets, and before each reading, its hand already set, the element is degaussed by a decaying alternating field from above saturation whose starting polarity a coin of its own sets. FDT: 𝔶tw ≠ 0, its sense unassigned. By P-⁠5 the imposed twist establishes a magnetic field; the mirror through a plane containing the element’s axis reverses the hand and an axial field and keeps the taper and the coated face, so the field’s axial part is odd in the hand, and a field along the film normal magnetizes the garnet along it and gives the platinum a reciprocity-odd resistance, which is consensus data. Comparator: zero within the ambient leak. By the reciprocal relations of linear transport, ℛ13,24 of a conductor at one pattern 𝒫 of magnetization and applied field equals ℛ24,13 at the reversed pattern, whatever its shape, strain and twist, so ℛodd is odd under that reversal and vanishes where 𝒫 is null. A uniform field left by the shield enters through terms odd in the hand, bounded by 𝔟amb, fixed before the first session from the contrast’s measured response to uniform test fields along three axes and from the shield’s measured residual; the first session starts only with 𝔟amb at or under 0.04 ℛsat. Detection: the plain two-sided test rejects at p ≤ 0.0027, of size 10/4096 with no twist field and no ambient leak. Attribution to the twist’s field: the bounded test with 𝔟 = 𝔟amb holds as well. Exclusion: the exclusion test with 𝔥tw = 0.1 ℛsat; otherwise the result is inconclusive.
  2. T-⁠2
    The same twist on a non-magnetic garnet shows a hand-odd, reciprocity-odd resistance in its platinum film as well, zero in the consensus calculation. An element of the same shape cut from yttrium aluminium garnet, coated and contacted as in T-⁠1. It holds no magnetization, so there is no degaussing and no residual-domain channel; twelve sessions, each read in a coin-set order. Observable: 𝔶nm = ½[ℛodd(+) − ℛodd(−)], in units of ℛcal, half the difference of ℛodd between the two senses of a 1 mT test field along the axis. FDT: 𝔶nm ≠ 0, its sense unassigned: the twist’s field threads the platinum film itself, whose Hall resistance in a field along its normal is consensus data. Comparator: zero within the ambient leak, by the same reciprocity for a conductor with no magnetization in no applied field, the leak bounded at or under 0.04 ℛcal. The decision rules are T-⁠1’s, with 𝔥tw = 0.1 ℛcal.
  3. T-⁠3
    Copper oxidized in oxygen-18 rather than oxygen-16 shifts the logarithm of the platinum layer’s effective magneto-thermopower ratio upward, in the analysis’s reading of I-⁠7’s density-channel input sign. Twelve slabs of the source’s garnet, each coated with Pt(3) over its whole face and, through a mask in one run, with the copper of the source’s CuOx(2) layer over two of three strips along its length, then cut with no exposure to air into a bare-platinum bar and two copper-coated bars. The coated bars are oxidized for the source’s two days in two cells of one design at one temperature, one filled with air made from oxygen enriched to at least 95% in 18O, the other with air made from natural oxygen, and capped in one run. A coin drawn before the cut assigns the two bars to the two cells, a second, before each pair’s oxidation, the two airs to the cells, and a third the order in which the pair is read, so a gradient of the deposit across the slab, a difference between the cells and a reading drift take signs the coins set. Observable: with 𝔮th formed for each coated bar with its slab’s bare-platinum bar through the source’s shunting estimate, 𝔶th = ln 𝔮th(18O) − ln 𝔮th(16O). FDT: for purely density-coupled equilibrium observables a heavier isotope gives a larger Rs, a higher α and a larger response, opposite in sign to conventional reduced-mass scaling (I-⁠7). The density channel enters through the oxidized layer’s own A-⁠1 window αext, moved by the substitution through 𝓜 at fixed pressure, radius and packing; the layer’s 𝓜 is built from its composition and the oxygen isotope’s mass, never from 𝔮th, and the heavier isotope’s zero-point lattice energy is an E/c² entry of it, entered once. The analysis reads the larger response as a larger 𝔮th. 𝔮th, a quotient of linear responses independent of the drives’ sizes, is an equilibrium observable and not purely density-coupled: its geometric, electronic and inertial channels are the other terms of Δ ln 𝔮th = ∫ Σi Si d ln xi, taken from the lighter isotope’s bar to the heavier’s, each Si = ∂ ln 𝔮th/∂ ln xi with every other coordinate held fixed. The density channel’s input sign is positive, and T-⁠3’s one-sided rules take the observable sign as positive on that reading: T-⁠3 is a sign prediction. Comparator: each pair’s consensus shift of ln 𝔮th, from the oxidation kinetics, the phonon channel of reduced-mass scaling, the heavier isotope’s zero-point lattice change and the pair’s residual composition, thickness and microstructure differences, lies inside the preregistered interval ±𝔠iso = ±0.01, and reduced-mass scaling enters only as this comparator’s channel. Detection: the one-sided test of the pair differences less 𝔠iso rejects upward at p ≤ 0.00135, of size at most 5/4096 wherever every pair’s consensus shift lies in the interval. Exclusion: a density-channel contribution of 𝔥iso = 0.05 or more in the logarithm is excluded when the one-sided test of the pair differences less (𝔥iso − 𝔠iso) rejects downward at p ≤ 0.00135; otherwise the result is inconclusive.
  4. T-⁠4
    The same oxidation shifts the logarithm of the platinum layer’s effective magnetoresistance ratio upward, in the same reading. The same twelve pairs, read under a charge current: 𝔶el = ln 𝔮el(18O) − ln 𝔮el(16O), with 𝔮el formed for each coated bar with its slab’s bare-platinum bar. The FDT input is T-⁠3’s with 𝔮el in place of 𝔮th: the larger response read as a larger 𝔮el, a quotient of linear responses and not purely density-coupled, and T-⁠4 a sign prediction. Comparator, interval, rules and exclusion scale are T-⁠3’s. T-⁠3 and T-⁠4 run on the same pairs; each decision uses its own twelve pair differences and its own exact size, and no joint decision is formed from the two.
  5. T-⁠5
    On one slab, the factor by which the oxide multiplies a conversion is not bound to the factor by which it multiplies the inverse conversion. Twelve slabs of the source’s garnet, each coated with Pt(3) over its whole face and with the source’s CuOx(2) over one half of its length, the half set by a coin drawn before the deposition, capped as the source caps and contacted through one mask. On each half the source’s ℛ and Δℛ are measured, and 𝒱as, the part of the thermovoltage antisymmetric in the field, which the source does not print and which, it writes, the out-of-plane heat current generates “due to the spin Seebeck effect of YIG”; the two halves are heated through the slab’s thickness by one heater in one run, the half read first set by a second coin. Observable: with ℛtp a half’s two-probe resistance between its voltage contacts, 𝔈el = (Δℛ/ℛ²)ox/(Δℛ/ℛ²)bare, the quotient of the two halves’ sheet-conductance changes and the source’s effective magnetoresistance ratio over its reference ratio; 𝔈inv = (𝒱as/ℛtp)ox/(𝒱as/ℛtp)bare, the quotient of the charge currents the two halves return; and 𝔶rec = ln 𝔈el − 2 ln 𝔈inv. FDT: a charge current’s conversion into angular momentum and its inverse are two settlements (0.1a), so the factor by which the oxide multiplies the forward conversion is not bound to the factor by which it multiplies the inverse: 𝔶rec ≠ 0, its sense unassigned. Comparator: 𝔶rec inside the preregistered interval ±𝔠rec = ±0.05. In the consensus account, by reciprocity and in the bookkeeping of the source’s Eq. (3) and Eq. (4), a half’s sheet-conductance change is its forward factor 𝔲 times its inverse factor 𝔳 over its interface loading 1 + 𝔏𝔷, and the current it returns from the heated garnet is its inverse factor times the heat flux through it over the same loading, with 𝔷 = coth(𝔡Pt/𝔩sd) for a platinum upper face that passes no spin current and tanh(𝔡Pt/𝔩sd) for one that absorbs it all. Then 𝔶rec = ln(𝔲ox/𝔲bare) − ln(𝔳ox/𝔳bare) + ln[(1 + 𝔏𝔷ox)/(1 + 𝔏𝔷bare)] − 2 ln 𝔭, with 𝔭 the quotient of the heat flux through the coated half over that through the bare half; with each half’s forward and inverse factors equal and the heat fluxes equal, only the loading term remains, between −0.03 on the source’s values and zero, inside the interval. Detection: the bounded test with 𝔟 = 𝔠rec. Exclusion: the exclusion test with 𝔥rec = 0.2; otherwise the result is inconclusive.

A T-⁠1 or T-⁠2 detection is read as the field P-⁠5 assigns to the imposed twist when the bounded test holds as well. A T-⁠3 or T-⁠4 detection is read as the density channel when it stands with 𝔠iso kept for the phonon and zero-point channels and added to the bound recomputed from the pairs’ measured sheet resistances, oxide depths and compositions, each propagated through a sensitivity measured on bars oxidized in natural oxygen before the first pair. A T-⁠5 detection is read as the difference of the two settlements when it stands with the interval recomputed from the slabs’ measured loading inputs and surface temperatures.

At each test’s test point the input in place and the input ablated give different decisions. With the bound at its ceiling and each of the twelve values at the test point, the bounded test sees twelve equal values 𝔥 − 𝔟; the largest mean that one sign flip among them leaves short of rejection is five-sixths of each, so the threshold is 𝔟 + 5(𝔥 − 𝔟)/6. For T-⁠1 it is 0.09 ℛsat, against a difference of 0.1 ℛsat between the input in place and ablated, and the same in ℛcal for T-⁠2; for T-⁠5 it is 0.175, against 0.2. For T-⁠3 and T-⁠4, residuals of 𝔥iso = 0.05 with the input and 0 without it face the threshold 5𝔥iso/6 = 0.0417.

Five tests and their comparators. Twist, T-⁠1 and T-⁠2: twelve session contrasts, in units of ℛsat or ℛcal, each the input’s effect, ±𝔥tw = ±0.1 at the test point or 0 ablated, plus a drift between the two readings whose sign the order coin sets, filled or open, a residual domain state signed by the two degaussing coins on the garnet of T-⁠1, and the term of one uniform shield residual inside ±𝔟amb, set within the ceiling ±0.04. Isotopes, T-⁠3 and T-⁠4: twelve pair differences of ln 𝔮th or ln 𝔮el, 18O less 16O, each the pair’s consensus shift inside ±𝔠iso = ±0.01, three coin-signed terms and the density channel, 𝔥iso = 0.05 at the test point or 0 ablated. Two settlements, T-⁠5: twelve values of 𝔶rec, each the loading change, one sign in every slab and set here between −0.03 and 0, the heat-flux term, two coin-signed terms and the input, ±𝔥rec = ±0.2 at the test point or 0 ablated. The decision panel draws the mean under every one of the 4096 sign assignments: the ten most extreme make the two-sided region, p ≤ 0.0027, and the five largest the one-sided region, p ≤ 0.00135; the bounded and exclusion tests repeat the test with every value shifted. Every scatter is seeded and illustrative, since the tests rest on sign symmetry and the declared bounds and on no variance estimate; the apparatus, settings and test points are the analysis’s design, not measurements in the source, and the test points instantiate effects, never derived values (P-⁠5, I-⁠7, 0.1a).

T-⁠1 to T-⁠5, P-⁠5, I-⁠7, 0.1a

Readings declined

Use 6 Data yes, interpretation not necessarily

Standard reading

The applied field that the source sweeps to align the garnet’s magnetization is an entity of its own, distinct from the material it acts on.

FDT reading

A magnetic field is a macroscopic helical density gradient of the omnium, and imposing a physical twist on a material is establishing one (P-⁠5). The field values and the alignment are accepted as printed, and T-⁠1 and T-⁠2 put a twist where the applied field was.

The single spin Hall angle, squared in the spin Hall magnetoresistance fit, is one matrix element for the conversion of a charge current into a spin current and for the conversion back.

The two directions are a process and its run backward, two system-wide settlements, each from its own starting state (0.1a). The measured ratio and the fitted angle are accepted in full and returned by the identity map, and T-⁠5 sets the two directions against each other on one slab.

Convergence

Use 2 Same source, same data

FDT agrees with the measured record of general relativity and quantum mechanics without reservation. On the source’s own inputs and dictionary, its spin Nernst and spin Hall magnetoresistance models, its parallel-resistance model, its shunting estimates and its drift-diffusion model with its three assumptions, FDT returns what the source prints for its samples with 3 nm of platinum: the four ratios, the oxide layer’s sheet resistance and Seebeck coefficient, the platinum layer’s two effective ratios and their quotients with the reference, 2.5 ± 0.3 and 2.8 ± 0.5 against the source’s “about twice” and “nearly a factor of three”, the four angles and the four conductivities. Each step is the identity map, corroborating no invariant, and every contrast, ratio, sheet resistance and Seebeck coefficient the source reports is accepted in full.

Each test’s convergence value is its comparator’s: zero within the ambient leak for T-⁠1 and T-⁠2, each pair’s consensus shift inside ±0.01 for T-⁠3 and T-⁠4, and a value inside ±0.05 for T-⁠5. FDT makes a divergence claim only where P-⁠5 supplies an additional source, I-⁠7 a constitutive input or 0.1a a changed observable dictionary, never through α, a chart or a branch tag alone. Where FDT, general relativity and quantum mechanics converge, FDT reproduces their results; where it declines their assumptions, here a field distinct from the material it acts on and one matrix element for a conversion and its inverse, it is under no obligation to reproduce them.