The questions you ask us. All of them.

No, not within what was measured: the recoil permeability of a neodymium magnet is close to that of air (≈ 1), so inserting the magnet does not change the coil's inductance — in steady state, the cancellation only costs the ordinary Joule losses of the wire. Over a measured cycle (200 J injected), 160 J are dissipated as Joule losses and 40 J are associated with the coil's field, recovered at switch-off through a freewheeling diode. A residual electromagnetic coupling does exist, measured at about 0.5% of the input power (200 mW out of 40 W).

The relation Kt = Ke holds for any motor in which torque and back-EMF stem from the same mechanism (flux variation tied to the current). The InfiniX architecture changes that premise: torque comes from a static magnetic interaction between permanent magnets, the current serving only to switch that interaction, without taking part in it energetically. This is not "different physics" in the general sense — it is an architecture in which the structural assumption underlying Kt = Ke no longer holds.

The working hypothesis is an exchange with the ambient energy of the environment — a principle already exploited, for example, in heat pumps or solar panels. Doctorate-level reviewers who examined the protocol confirmed the absence of measurement artefacts; the precise explanation of the mechanism is still under study and is not yet established as a validated scientific fact.

On the objection "an efficiency above 1 would be an accounting illusion — everything must sum to zero": that is a classic reasoning error, which forgets one flow in the balance. The real balance of a heat pump reads Q_hot = W_elec + Q_{cold drawn} — three terms, not two. A heat pump does not create energy: it draws heat already present in the environment (air, ground, water) thanks to the electrical work supplied to the compressor, and delivers that drawn heat on top of the electrical work consumed. On the cold side, nothing is "lost" that would offset the gain on the hot side — an energy that was already there, at a lower temperature level, is extracted. It is precisely the existence of this third flow (ambient energy) that allows a delivered-heat / consumed-electricity ratio to exceed 1 without violating the conservation of energy: if that external cold source did not exist, the device would be nothing but a plain electric heater, with an efficiency of 1, never more. The same three-flow balance principle is the working hypothesis put forward for the surplus observed on InfiniX.

It is the reciprocity relation governing every conventional electromagnetic motor: the torque constant (Kt) and the back-EMF constant (Ke) are numerically equal, in coherent SI units. This is not a design coincidence but a direct consequence of the conservation of energy — if they were not equal, a motor could produce more (or less) mechanical work than the electrical energy consumed. It is this relation, valid for any motor where torque and back-EMF share the same mechanism, that the anti-motor architecture no longer follows in the same way.

The publication explicitly acknowledges three open, unresolved theoretical hypotheses: (1) does the field cancellation constitute an energetically costly operation, or a mere superposition with no net transfer? (2) the role of a variable reluctance specific to the magnet/magnet magnetic circuit; (3) a decoupling between the electrical excitation circuit and the magnetic interaction circuit, in which the current would merely switch an energy already present in the magnets. The authors do not decide between these hypotheses and say so in black and white — that is not a hidden admission of weakness, it is an honest delimitation of what remains to be demonstrated, left to the scientific community.

That is exactly the "equivalent reading" the publication itself mentions out of scientific honesty: the results could be explained by a residual coupling below the detection threshold, rather than by a genuine absence of coupling — in which case the results would be equivalent, not superior, to a well-optimised conventional motor. This interpretation cannot be ruled out by this paper's measurements alone. That is precisely why replication by a third-party laboratory, with potentially finer instrumentation, is deemed necessary before any definitive conclusion.

Because that is deliberately not the subject of this paper. The publication deals solely with the characterisation and suppression method of the back-EMF. A rigorous energy balance would require accounting for all the system's losses (switching, control electronics, mechanical friction, circuit efficiency) — that work is announced as the subject of a separate publication. Readers are explicitly invited not to extrapolate the results presented here into a conclusion about overall efficiency.

It is a well-founded point of rigour: the theoretical calculation does rest on an assumption of equidistant poles. That is precisely why an independent check was added: a Hall-effect probe placed directly at each magnet confirms the absence of a residual local field during the magnetisation phase, independently of any simplifying geometric assumption.

The motor was tested with a Prony brake, which applies a deliberate mechanical resistance to the shaft — not mere free rotation. The net torque measured over a full cycle, under load, is positive. What is still missing: an instantaneous, phase-by-phase torque over the full cycle, rather than a global net torque.

That mechanism, if it existed, could only show up during the phase in which a current actually flows, the associated energy term being proportional to M(θ)·I². Current, voltage and phase shift were monitored continuously while the two powered electromagnets were moved through the full range of relative angles: no variation was measured. After the circuit opens, I = 0, so that term necessarily vanishes, whatever the value of M(θ) at that instant.

A complementary physical argument reinforces this result: since the coil's core and the permanent magnet it contains are co-located with opposite dipole moments, the resulting field cancels, by the very property of the dipole field, at every point of the surrounding space — not only at the magnet itself. This extended cancellation was verified directly with a Hall-effect probe, at several distances and positions around the coil, including very close to the winding — without detecting any significant residual field. Each coil's own field is therefore not available to induce a coupling with the opposite coil, consistent with the absence of any measured variation.

With an active current source, which automatically adjusts the voltage to keep the current constant, a hidden electromotive force would show up in the voltage rather than in the current — hence the value of monitoring the voltage as well. But the test was taken further: with a simple, unregulated voltage source, where nothing adjusts the current, the current still remains constant during rotation. That is an even more direct argument: without an active compensation mechanism, a hidden electromotive force would necessarily show up as a current variation (Ohm's law, I = (V − e)/R) — nothing would mask it. Its absence, under these conditions, therefore directly rules out any significant residual coupling, without even needing to invoke the reasoning about the active source.

This objection is acknowledged as legitimate in the publication and not treated quantitatively at this stage. Real prototypes are necessarily subject to manufacturing tolerances. The absence of any measured current variation provides an upper bound on the effect of these imperfections, given the declared instrument resolution, but does not guarantee perfect symmetry in the strict sense — a point explicitly left open rather than brushed aside.

This is a well-documented phenomenon (interruption of an inductive current, or "flyback"), which does require protection. The switching circuit includes a dedicated device — a freewheeling diode or clamping circuit — to absorb this transient voltage across the windings and prevent any electric arc or damage to the switching components. It is not an obstacle to the principle; it is a classic engineering point already built into the protocol.

Three methodological guarantees are described: (1) all instruments (current and voltage probes) are calibrated before each measurement campaign, with verification of the sensitivity factor; (2) every reported measurement was cross-checked with a second probe and a second oscilloscope, independent of the first set-up, to rule out any drift specific to a single instrument; (3) the results were also verified by third parties, independent of the initial measurement protocol, before their inclusion in the publication.

The field applied at each cycle stays, by design, below the threshold of the magnet's coercive field. For an N52 grade, that threshold is around 950 to 1500 kA/m, whereas the field required for industrial magnetisation is 2 to 2.5 times higher (2000 to 3000 kA/m). The operating field is therefore structurally too weak to reorganise the material's structure. A very slow fatigue effect over a very large number of cycles cannot be ruled out; only a measurement of the remanent field before/after a large number of cycles would confirm it — that test is planned.

Because the property being exploited is specific to this type of material: over its normal operating range, a neodymium magnet works on its recoil line, with a recoil permeability close to that of air (≈ 1). It is this property that lets the magnet leave the inductance of the coil surrounding it unchanged, unlike a conventional ferromagnetic core (iron, ferrite) — see Experiments 1 and 2 of the publication. A conventional electromagnet without a permanent core would not exhibit this property and would not allow the same reversible cancellation.

That is the most significant limitation of this work, explicitly acknowledged in the publication itself. The full protocol (architecture, measurements, instrument list) is now public precisely to enable that external validation, which remains the decisive criterion before any definitive conclusion. That is also why the plans of the educational demonstrator are made available to members.

Yes — that is the central goal of the approach. The full protocol is public, and the plans of a linear educational demonstrator are made available to members of the association, to allow a direct check of the central principle — the cancellation of a field by superposition, and the force that reappears when it is restored.

An AI remains a valuable tool for spotting a flaw in reasoning or asking a good question — it is not a scientific tribunal, and its opinion never replaces a measurement made in a laboratory. An AI reasons from already-established knowledge, cannot verify an experiment itself, and can mix up sources when producing a synthesis. Good practice is to cross-check what it says against genuine experimental verification.

The publication answers this directly: until it was prepared, the authors had not sought third-party replication — by choice rather than impossibility — after more than 22 years of independent research. A personal event in 2025 led to reconsidering that position: the need to pass this work on to the scientific community, rather than keep it unshared, became the priority. That decision is what motivates publishing the full protocol, with a view to open replication by third parties.

Yes, a patent was filed with the French National Institute of Industrial Property (INPI) under registration number FR2609488, dated 9 July 2026. The scientific publication and its time-stamped public deposit constitute, together with this patent, the authors' priority claim.

Yes. The INPI filing includes claims dedicated to the implementation variants: linear actuator configuration, multipolar machine, axial-flux architecture, multi-stage stacking, a hybrid variant with an auxiliary conventional winding, alternative magnetic materials (samarium-cobalt, hard ferrite), and an oscillating/reciprocating configuration — in addition to the main rotary architecture described in the publication.

The publication contains a specific note on this point: scientific priority is established by a dated, verifiable disclosure — a publication, a patent, a time-stamped document — not by an after-the-fact claim. An idea never disclosed, never tested and never published carries no more weight, from a priority standpoint, than an idea never had. The INPI filing (FR2609488, dated 9 July 2026) and the publication constitute the authors' priority claim.

Yes, and the publication acknowledges it explicitly from the State of the Art onwards: this principle is already exploited in electromagnetic lock systems ("power to release"), used to hold metal objects in place. But those applications remain confined to a static, binary use (hold or release), with no rotary motion and no notion of continuous torque. To the authors' knowledge, this principle had never been applied to a rotary motor, nor implemented simultaneously on a stator and a rotor — it is this specific combination, not the isolated physical principle, that constitutes the paper's contribution.

No — it is a precise technical distinction, defined in the publication: in a conventional motor, torque and back-EMF share the same physical mechanism (flux variation tied to the current). In the architecture described, the current never directly produces the driving field — it acts as a switch for a static magnetic interaction between permanent magnets. That distinction directly conditions why the relation Kt = Ke does not apply to this architecture in the same way. The term is not a marketing choice; it is the structural definition that underpins the paper's entire reasoning.

The initial work did build on the study of Bedini-type pulsed electromagnetic systems, as part of a broader experimental effort around Tesla-style coil winding. The publication owns this openly. That point of origin does not determine the validity of the result obtained: the discovery itself arose from an observed consumption gap, unexplained by the classical model, then was reproduced on a large number of prototypes before being formalised into a verifiable protocol — it is that protocol, not the historical origin of the effort, that should be judged.