These systems do not include field-dependent velocity distributions (such as caused by magnetic fields). At μ-Scale (in the absence of collisions), T-Even symmetry requires that all motion be exactly time-reversible, and that fine scale entropy remains invariant, dS_fine/dt = 0. However, as the observer enlarges his scope from μ-Scale to M-Scale, information is lost and entropy increases dS_coarse/dt > 0 due to effects such as Stosszahlansatz by Boltzmann, coarse graining by Gibbs, and the ergodic hypothesis by von Neumann. Most importantly, from a second law violation perspective, T-EvensystemsaretheleastinterestingbecauseoftheirinabilitytogeneratedS/dt<0atanyscale.

Proofs covering T-Even systems by the axiomatic analysis are described in Appendices A-B:

1. Boltzmann’s H-Theorem.

2. Landauer’s Principle (derived without magnetic fields).

T-Odd symmetric systems are of special interest. Being governed by Liouville theorem, their fine-grained Gibbs entropy must remain constant, dS_fine/dt = 0. However, this theorem does not guarantee that its marginal components—positional entropy S_p and momentum entropy S_m—must also remain constant. Even though the volume of phase space is invariant, its shape is not: positional entropy may decrease at the expense of momentum entropy—as in a free expansion—or vice versa. Unfortunately, at μ-Scale, this marginal negentropy is inaccessible because of the lack of thermodynamic interactions (collisions). Conversely, at M-Scale, massive coarse graining, large-scale averaging, or integration ensures that this negentropic effect is completely erased.

Proofs covering T-Odd systems by the axiomatic analysis are described in Appendices C-H:

1. Gibbs’ H-Theorem (With B Field).

2. von Neumann’s Quantum H-Theorem

3. Detailed Balance (with B Field).

4. Fluctuation Theorem & Stoch. Therm. (with B Field.

5. Fluctuation-Dissipation Theorem (with B Field).

6. Kirchhoff’s Radiation Law (with B Field).

As part of the axiomatic analysis, two thought experiments, Maxwell’s Demon and Feynman’s ratchet are described in Appendices I and J.

To further complete and go beyond the analysis, two appendices covering experimental data are included. Appendix K describes experiments in support of the second law (all conducted at M-Scale) and Appendix L, experiments against the law. The results of the axiomatic analysis are summarized in Table1.

Table1. Axiomatic analysis of the foundational arguments supporting the second law, showing for each the symmetry it addresses, the operation by which it enforces its conclusion, and the resulting position on the m-Scale. Every argument, without exception, resolves to an endpoint of the index of asymmetry.

When perpendicular electric and magnetic fields E and B are applied to a high-mobility semiconductor, charged particles follow cycloidal paths and drift with velocity

perpendicular to both fields [1]-[5]. The drift velocity depends only on the fields—not on the particle’s instantaneous velocity, and not on the sign of its charge. This has an immediate consequence for time reversal. Under the operation t → −t with the fields held fixed, every particle velocity reverses, v → −v, yet Equation (1) is unchanged. The reversed motion is therefore not a solution of the equations of motion: a movie of the drift run backward shows particles drifting against E × B, a motion the fixed fields cannot produce. At fixed B, trajectories do not retrace, and the dynamics are T-Odd.

It is sometimes objected that the full system is nonetheless T-Even: if time reversal is applied also to the currents that generate the field, then B → −B, the drift reverses with it, and the backward movie becomes a valid solution of the reversed problem. This is correct as far as it goes—but it describes a different experiment, conducted in a reversed field. The experimenter must decide where to put the magnetic field. Placing it in the same time frame as the experiment rendersthewholeexperimentT-Odd, afactthattheexperimentercaneasilydetectbymeasuringthenewfielddirection. In the end, theexperimentermustchoosewheretoputthefield: within the experiment, or in his time frame. No matter what he decides, theverypresenceofthefield,notwhichframeitoccupies,causestimeasymmetry [1][2].

The T-Odd symmetry of the magnetic field is precisely what motivated Onsager to exclude them from his reciprocals and caused Casimir to formulate his Onsager-Casimir reciprocal relations, L_ij(B) = ε_iε_jL_ji(−B) [6][7]: reciprocity is restored only betweentwodistinctsystemseachwithitsownfield. For a single device operating in one fixed applied field no reciprocity is asserted, and the dynamics are irreducibly T-Odd. Casimir’s formalism customarily invoked to restore time-reversal symmetry is thus the same formalism that codifies its absence at fixed B.

Diodes and fluidic valves are configured atfabricationtimeasspatiallyasymmetricyetbecometimeasymmetricwhenexposedtoatimevaryingsignalthatswitchesdirectionwithtime. The signal causes the configuration (e.g., thickness of the depletion zone in a PN diode) and conductivity of the device to change between forward and reverse modes. This dynamic change in spatial asymmetry results in T-Odd symmetry¹. It is associated with a state variable (e.g., thickness of depletion zone), a bandwidth, and an energy of transition (built-in potential). These devices are the subject of a companion paper [1] written in the light of the theory presented herein.

For reference, first consider N gas particles occupying volume V₁ suddenly allowed to expand into volume V₂ > V₁. The expansion is isothermal; internal energy U and temperature T remain constant. The entropy of the gas increases:

and its Helmholtz free energy decreases:

Figure 1. The E×BThermoelectric effect. This effect requires a loop including a forward channel and a return channel. The forward channel comprises a high mobility, low concentration material which enables the drift. The return channel comprises a low mobility, high concentration material which disables the drift but still enables Ohmic conduction.

No work is extracted, and free energy is irreversibly destroyed. This is the canonical dissipative process of the second law.

Now let the same gas be placed in the forward channel in Figure 1. Particles carried by the E × B drift in a high-mobility material follow cycloidal paths uninterrupted by collisions. Beginning uniformly distributed in volume V₁, they move unidirectionally and in unison at the common drift velocity v_E×B along an equipotential line, without doing any work and accumulating at the downstream wall into a smaller volume V₂ < V₁ to create a space charge. This space charge deflects the E field as well as the drift direction. Section9 provides an accurate discussion of this process.

The energy bookkeeping along the way is exact. The magnetic force, always perpendicular to velocity, does no work. The electric field does no net work over a cycloid: the guiding-center displacement is perpendicular to E, and the cyclical excursions along E within each orbit average to zero.² The particles’ kinetic energies, the internal energy U, and the temperature T are therefore unchanged by the transport. Since V₂ < V₁ at constant U and T, applying (2) and (3) to the positional distribution, entropy decreases:

and the Helmholtz free energy increases:

The process is the exact reverse of a free expansion; we shall call it a freecompression. No work is expended to produce it. The T-Odd drift converts positional uncertainty into positional order at zero energy cost.

Equation (2) ascribes positional entropy to compression and must be read with care. The collision-less motion in the forward channel is Hamiltonian. Liouville’s theorem therefore holds exactly: the fine-grained phase-space density is conserved, and the fine-grained entropy is invariant, dS_fine/dt = 0, in full agreement with unitarity. Nothing in this paper contradicts that statement. The entropy that decreases in Equation (2) is the positionalentropy—the entropy of the marginal distribution over position, obtained by integrating the full phase-space distribution over momentum. The negentropy generated by the drift resides in the correlationsbetweenpositionandmomentum that the coherent drift builds up; thepositionaldistributionregistersthisnegentropyasacompression,whilethefine-grainedentropy,whichtracksthecorrelationsinfull,remainsconstant.

Formally, selecting the positional distribution is a coarse-graining operation: the position-momentum correlations are left out. Since coarse-graining is essential to the second law as shown by the axiomatic analysis, two kinds of such operations must be distinguished:

1. In the arguments catalogued in Appendices A-H, coarse-graining is imposed by the observer: it is justified by his limited resolution, and it erases the very correlations in which the T-Odd negentropy resides.

2. In the device proposed in this paper, the coarse-graining operation extracts position information and erases the rest. It is achieved not by the observer but by the device. A load in the return channel couples to the electron gas only through the electrostatic potential of its charge density, and charge density depends onpositionalone. The momentum correlations and whatever negentropy the drift stores in the position-momentum correlations are inaccessible to the load. However,negentropystoredinpositionremainsaccessible. The positional entropy of Equation (4) measures neither more nor less than the accessible negentropy—and the challenge this paper poses to the second law is precisely whether this accessible positional negentropy can be generated isothermally.³

In summary, coarse-graining operates like a low pass filter that eliminates all μ-Scale information from M-Scale. A device applying coarse-graining at m-Scale, can erase position-momentum correlations which are prevalent at μ-Scale while leaving partly untouched larger-sized positional information.

One further point of rigor: the compressed gas in V₂ is not in an equilibrium state—it retains the velocity-space correlation caused by the drift. Equations (2) and (3) nonetheless give the correct entropy measure, because they compare the positional distributions of the initial and final states at the same U and T; theΔAsocomputedistheceilingontheworkobtainablefromthefreecompression,attainedinthereversiblelimit. The residual velocity-space correlation subsequently thermalizes in the collisional return channel at constant internal energy, neither adding to nor subtracting fromthepositionalfreeenergy.⁴

Since the particles’ internal energy is constant along the forward channel, the gas is isothermal from end to end, and the return channel must be isothermal from end to end as well. To close the cycle, the particles undergo an isothermal expansion in the return channel that restores them to their original volume V₁. This expansion converts the free energy accumulated by the compression into electrical work in a load, which may be a battery, a resistor, or whatever the designer chooses.

The process is end-to-end isothermal butnotnecessarilyreversible, and itneednotbeisothermalwithrespecttoitssurroundings. As the gas expands in the return channel, thermal energy is converted to work; the channel cools relative to the surroundings, and the forward channel cools with it. The emerging temperature difference ΔT allows heat Q = W to flow in and replace the thermal energy converted to work. This ΔT is a Fourier heat-conduction requirement, not a Carnot limitation on the performance of the system. In other words, there is no need for a heat sink, but the device must still be colder than the surroundings.

The energy source of the cycle can now be identified by elimination, term by term around the closed loop.

The static magnetic field does no work on any particle at any time. The applied electric field is static and conservative, so its net work on a particle completing one circulation of the loop is identically zero. The space-charge field E_x is internal to the system: the energy stored in the space-charge field is constant in steady state and enters the audit only as a start-up transient. The work done against it by particles accumulating upstream is recovered from it as those particles are driven through the return channel and the load. No field term, applied or internal, delivers net energy per cycle. The fields play a strictly catalytic role—they supply the T-Odd bias that configures the microdynamics, but they pay no energy for the negentropy they create.

By the first law, the net work delivered to the load per cycle must therefore be drawn from the only remaining reservoir: heat absorbed from the surroundings:

In the return path, the heat-to-work conversion is limited to the free energy accumulated in the free compression; under reversible conditions,

The crucial property enabling steady circulation is that the drift can be turned onandoffwithoutalteringthefields, through the mobility and concentration difference between the two channels—high mobility and low concentration in the forward channel and low mobility and high concentration in the return channel. In the language of Section 7, the loop shifts the index of asymmetry from a high value in the forward channel, where negentropy is generated, to a low value in the return channel, where it is spent in the load. The process violates neither the first law, as shown above, nor any fine-grained conservation law, as shown in Section 5.3. Itonlyconteststheuniversalityofthecoarse-grainedinequalityΔS≥0—anditcontestsitalongthem-Scale,betweentheμ-ScaleandM-Scaleendpoints,whereAppendixKshowstheinequalityhasneverbeentheoreticallyinvestigatednorexperimentallytested.

Equation (35) can be maximized by differentiating and setting to zero. As illustrated in Figure 4(b) and Figure 5, maximum power is achieved at θ_asym = π/4, exactly in the center of the m-Scale:

where the index of asymmetry is unity:

At that point, the Hall and Drude resistivities are also equal thereby confirming that the Maximum Power Transfer Theorem (MPTT) also applies to the use of free energy provided by a magnetic field.

Figure5. Free Energy Utilization Factor. The fraction of free energy converted to actual energy falls to zero at either end of the index of asymmetry as predicted by all proofs of the second law. It rises to a maximum of 1/4 at θ_asym = π/4 and Z_asym = 1 as required by the MPTT.

Unlike a battery with a fixedchemicalpotential, the free energy available from a device is not fixed: it is a function of the load. As shown in the vector diagrams of Figure 4(a) and Figure 4(b), in the presence of a load voltage, and under optimal condition θ_asym = π/4, the primary electrical field is deflected or “loaded” from E_y to | E D i a g | = | E y | / 2 .

Equation (36) is expressed in terms of the “unloaded” field E_y. Replacing E_y by the “loaded” field version E_Diag from Equation (21) and setting θ_asym = θ_E×B = π/4 results in:

This result is analogous to the one obtained in Part 2 of this paper [1] which discusses the rectification of the Johnson-Nyquist noise by a MIM diode. Furthermore, as discussed in Appendix I, this result is also identical to the one derived by Levy [2] using a Maxwell’s Demon “loaded” with the task of compressing a gas to run a turbine.

The Boltzmann H-theorem (1872) [1] [2] represents the first attempt to derive the macroscopic thermodynamic arrow of time from the microscopic mechanics of individual gas molecules. By tracking a single-particle probability distribution function in a 6-dimensional phase space Boltzmann sought to prove that molecular collisions force a system to evolve irreversibly toward a maximum entropy state. The system consists of a dilute gas composed of hard-sphere classical particles enclosed within an isolated container. The state of the gas is defined by the continuous distribution function f(r, v, t), which dictates the density of particles at position r with velocity v at time t. Kinetic behavior is governed by the Boltzmann transport equation, and the thermodynamic state is tracked via the functional H ( t ) = ∫ f ln f   d r   d v , mathematically related to the negative of the Boltzmann entropy.

A1.TimeSymmetry

The fundamental mechanics governing individual binary collisions are governed by Newton’s equations of motion. In the absence of velocity-dependent fields, these microscopicphysicallawspossesspureT-Evensymmetry requiring the equations of motion to be fully deterministic and path-reversible.

A2.MacroscopicIrreversibility

To force a monotonic decrease in the H-functional (dH/dt ≤ 0) and establish macroscopic irreversibility (dS/dt > 0), Boltzmann introduced the Molecular Chaos Assumption (Stosszahlansatz). It dictates that the velocities of two colliding particles are completely uncorrelated prior to their collision, allowing their joint two-particle distribution function to be factored into a product of independent single-particle distributions:

Crucially, the Stosszahlansatz mandates that these particles become immediately correlated after the collision, leadingtoalossofinformation. This asymmetrical treatment implies an M-Scale perspective of the observer and introduces a directional arrow of time by fiat. In an exact physical system, post-collision trajectories retain complex, multi-particle correlations. However, the M-Scale observer does not track this microscale information which is permanently erased via Stosszahlansatz.

A3.Positiononthem-Scale

The gas is T-Even, so the generative factor vanishes identically; independently, the Stosszahlansatz discards post-collision correlations at every step, setting the correlation memory to zero and enforcing Z_asym→0.

Landauer’s Principle (1961) [1] demands a thermodynamic cost for information processing: erasing one bit of information requires heat dissipation: Q ≥ k_BTln2. His system consists of a microscopic bistable memory containing a single particle. The “0” and “1” states correspond to the positioning of the particle in the left or right well. Logical erasure forces the particle into the “0” state regardless of its initial position.

B1.MicroscopicTimeSymmetry

The equations of motion for memory states are conservative and time-reversible (T-Even). Charles Bennett (1973) [2] proved that any algorithm can be executed using logically reversible operations (e.g., the Toffoli gate). Because phase-space volume is conserved under T-Even dynamics, an isolated microscopic computer can step forward and backward through a computation with zero dissipation (Q = 0). Reversible computation is a μ-Scale process that does not interact with the surroundings and conserves information.

B2.MacroscopicIrreversibility

Dissipation emerges during logical erasure which compresses two logical states into one and requires expelling excess phase-space volume into the surroundings. For Landauer’s bound to hold as an inequality, the surroundings must be an ideal macroscopic reservoir acting as an M-Scale enforcer:

1. Infiniteinformationsink. True erasure requires a bath with infinite degrees of freedom (N → ∞) to instantly randomize and destroy expelled entropy. A finite bath that retains phase memory allows Poincaré recurrences which return the information to the register and undo the “erasure.”

2. Infinitebathbandwidth. The bath is modeled as memoryless, delta-correlated noise that responds instantly at all frequencies. In the words of Section 7, the bath’s correlation time is τ → 0. Hence, Z_asym = ωτ → 0 for any finite frequency ω. M-Scale endpoint is not derived; it is installed by the noise model.

His framework therefore reproduces the catch-22 in its purest form: at μ-Scale, the register is reversible but thermodynamically isolated (nothing can be erased or harvested); atM-Scale, erasure is possible but only by paying k_BT ln 2 per bit to a bath that destroys all microstructure. He describes the cost of erasure on the T-Even load side of a thermodynamic loop, but is silent about whether a T-Odd source side can pay that cost. In Figure 1, the T-Odd source of free compression lies entirely outside his universe of discourse.

B3.Positiononthem-Scale

This proof covers a T-Even information system at M-Scale. It constrains the load, not the source. m-Scale coupling between register and bath is not considered. The delta-correlated bath enforces Z_asym→0.

J. Willard Gibbs (1902) [1]-[3] replaced Boltzmann’s single-particle distribution in 3-dimensional space with his ensemble distribution of N particles in the phase space (Γ-space). His original formulation was based on the classical Liouville equation built on a time-reversible Hamiltonian. Liouville’s theorem guarantees that the local density of any region of the phase-space fluid is perfectly conserved over time. Consequently, his fine-grained Gibbs entropy, S_fine is, in the terminology of this paper understood to be at μ-Scale and is dynamically invariant dS_fine/dt = 0. Given that the reversible Hamiltonian is T-Even, this result is hardly surprising.

C1.MicroscopicTimeSymmetry

This T-Even reversible Hamiltonian version was later generalized to a T-Odd version to cover magnetic field phenomena. Because the Lorentz force acts asymmetrically on velocity vectors, the exact trajectories are no longer path-reversible under a simple time inversion in a T-Odd field. However, theycanstillreversetoastartingposition, there is no information loss and phase-space flow remains incompressible: (dS_fine/dt = 0). At μ-Scale (fine scale) no information is lost due to collisions or coarse-graining.

C2.MacroscopicIrreversibility

To enforce the second law, Gibbsshiftstothemacroscale by introducing the concept of coarse-graining: he divides the continuous Γ-space into finite, macroscopic operational volumes or “cells” over which he averages out detailed path information thereby creating a one-way mathematical inflation of the observed volume (dS_coarse/dt ≥ 0). He justifies it by acknowledging that a macroscopic observer possesses finite measurement resolution and is incapable of resolving the exact, fine-grained boundaries of the phase fluid.

C3.Positiononthem-Scale

The enforcement of M-Scale lies in the observation operator, not in the dynamics; m-Scale observation, with resolution comparable to the orbit, is not considered. Fine-grained dynamics are T-Odd and preserve information, but coarse-graining cells exceed the cyclotron orbit, so the orbital radius r is unresolved and the observed index collapses to Z_asym→0.

John von Neumann’s quantum H-Theorem applies the methods of statistical mechanics to quantum systems: An isolated macroscopic quantum system described in a high-dimensional Hilbert space inevitably evolves toward a state of maximum entropy. The system state is tracked via the density operator ρ ^ ( t ) . Its fine-grained entropy is evaluated using the von Neumann entropy functional: S v N = − k B Tr [ ρ ^ ln ρ ^ ] .

D1.MicroscopicTimeSymmetry

He uses Schrödinger equation to describe the time evolution of the fine-grained density, translating as a unitary operator transformation [1] [2]:

Since the operator U ^ ( t ) = e − i H ^ t / ℏ is unitary, the fine-grained entropy remains invariant over time (dS_vN/dt = 0). No information is lost at μ-Scale. Unitarity is a statement about information conservation, not about time-reversal parity and applies to T-Even and T-Odd symmetric quantum systems, including charged particles in a magnetic field. It is the quantum counterpart of Gibbs’ H-Theorem of Appendix C, which reaches the same conclusion classically through Liouville’s theorem and coarse-graining.

D2.MacroscopicIrreversibility

To break this unitary invariance and derive macroscopic irreversibility (dS/dt > 0), von Neumann partitions the total Hilbert space into macroscopic operational cells. He defines a coarse-grained quantum observer who can only measure Macro-Observables. These operators are blind to microstate transitions within a given cell. The coarse-grained density matrix is obtained by averaging the true state over these macro-cells, yielding the macroscopic coarse-grained quantum entropy. To prove that this coarse-grained entropy increases, he invokes the ErgodicHypothesis and the RandomPhaseApproximation under which the fast-oscillating phase correlations carried by the off-diagonal elements of ρ̂ interfere destructively and average to zero over macroscopic timescales. The Random Phase Approximation does not merely blur the microstate: it discards coherence memory. The fine-grained dynamics remain reversible and information-preserving throughout; irreversibility appears only because the observer is stipulated at the macroscale.

D3.Positiononthem-Scale

m-Scale operation, in which coherence persists over times comparable to the observation interval, is not considered. The Random Phase Approximation and the ergodic hypothesis destroy the phase coherence memory over macro-observable cells, setting the correlation time to zero and enforcing Z_asym→0.

Simple Detailed Balance dictates that under equilibrium, every transition between discrete states is as probable in a forward or reverse direction.

E1.MicroscopicTimeSymmetry

This principle relied upon by Boltzmann and Planck, and later refined by Albert Einstein (1916) [1]-[4] in his derivation of A and B coefficients for radiation, assumes microscopic reversibility (T-Even) and only applies to systems free from velocity-dependent fields. Generalized Detailed Balance by Hendrik Casimir (1945) and Nico van Kampen (1954), extends this principle to T-Odd systems to include magnetic fields.

E2.MacroscopicIrreversibility

Casimir introduces the concept of variable parity, classifying state variables into α-type (T-Even) and β-type (T-Odd). When Onsager’s transport symmetry (L_ij = L_ji) is broken by a magnetic field, it must be replaced by the Onsager-Casimir Reciprocal Relations (L_ij(B) = ε_iε_jL_ji(-B)) where ε represents the time-reversal parity (±1). This approach codifies Generalized Detailed Balance at the M-Scalebutdoesnotexplainhowμ-Scale∧T-OddaffectsM-Scalebehavior.

Van Kampen resolves this paradox by showing that the transition from μ-Scale to M-Scale acts as a loop-canceling filter that averages out individual paths. In a fixed magnetic field, the transition probability matrix naturally forms closed,invariantloopsofprobabilityflux. Within these microscale loops, traffic is strictly one-way (P(i→j) ≠ P(j→i)). However, because the loops are closed, the total probability entering any macroscopic region perfectly equals the total probability leaving it. Integratingmicro-loopscancelsthemout,forcingthenetmacroscopicfluxcrossinganyarbitraryboundarytobepreciselyzero.

E3.Positiononthem-Scale

m-Scale operation, in which a path is interrupted before its loop closes, is not considered. van Kampen’s loop cancellation integrates the mean-free angle ϕ_λ over closed orbits, so that one-way microscale traffic sums to zero net macroscopic flux and Zasym→0 in its angular form.

The Fluctuation Theorem [1]-[3] formulated by Evans, Cohen, and Morriss (1993), states that in a system away from equilibrium, with a time-averaged entropy production Σ ¯ t characterized by a distribution Pr ( Σ ¯ t ) , the ratio between the probability P ( Σ ¯ t = + A ) and P ( Σ ¯ t = − A ) is exponential in At.

The probability of observing negentropy production decreases exponentially with time and size. This theorem yields a second law violation in μ-Scale but yields the second law 〈 Σ ¯ t 〉 ≥ 0 in the M-Scale limit.

F1.MicroscopicTimeSymmetry

Earliest forms of the theorem [1] [2] address T-Even physics, later ones [3] T-Odd phenomena.

F2.MacroscopicIrreversibility

T-Odd versions of the theorem rely on three Stochastic Thermodynamics assumptions:

1. TheOverdampedContinuumLimit: simplifies the tracking of random particle paths, by throwing out from equations the velocity/momentum terms that define velocity-dependent forces. (i.e., shifting from underdamped Langevinequations to overdamped ones). ThisprocessconvertsT-OddintoT-Even.

2. InfiniteBathBandwidth: the heat bath is modeled as a Gaussian white noise, defined by a Dirac delta-function correlation. It responds instantly across all frequencies, leaving no memory of previous collisions behind, thereby overwhelming μ-Scale effects and actingasaM-Scale(N→∞)enforcer.

3. LocalDetailedBalance: states that the ratio of local forward to backward transition rates is strictly tied to the macroscopic entropy change of the environment:

An environment inequilibriumdespitemicrotransitionsnegatesμ-Scale,implyingcircularlogic. FT uses LDB to “prove” the Second Law by assuming an environment that follows the M-Scale law.

F3.Positiononthem-Scale

Three operations enforce M-Scale: 1) the overdamped limit, 2) Infinite Bath Bandwidth 3) Local Detailed Balance. m-Scale operation is not considered. Hence the asymmetric index Z_asym→0.

The Fluctuation-Dissipation Theorem (FDT) [1]-[4] by Callen and Welton, links the linear response of a system under external perturbations to its spontaneous thermal fluctuations at equilibrium. Its modern classical and quantum versions by Ryogo Kubo may include a magnetic field. The FDT derived from the canonical ensemble (Boltzmann distribution) requires strictthermalequilibrium. The correlation of fluctuations is locked to its linear response with magnetic interactions accounted for along the ij axes.

G1.MicroscopicTimeSymmetry

The microscopic dynamics are governed by a well-defined Hamiltonian which accounts for the T-Odd dynamics due to a fixed magnetic field.

G2.MacroscopicIrreversibility

The system must be in a stationary,canonicalGibbsEquilibrium.ThisrequirementerasesT-OddinformationbymovingoperationtoM-Scale. Under the Gibbs canonical ensemble, a system’s probability distribution is dictated entirely by its Hamiltonian (total energy):

Since the magnetic force is always perpendicular to velocity, the field doesnowork and addszeroenergy to the system. Hence the magnetic field fully drops out of the exponential part of the Gibbs distribution.

Consequently, the probability of finding particles at specific locations and speeds is perfectly symmetric. Averaged across the entire ensemble, the microscopic curves completely smooth out. Gibbs’sequilibriumforcesthesystemtolookstaticallybalanced,effectivelyerasingthedirectionalbiasofthemagneticfield. This averaging is a form of coarse-graining enforced by equilibrium.

Saito and Utsumi derive fluctuation relations for T-Odd systems without the overdamped limit; they rely on [initial canonical states / the ±B comparison], and, therefore, occupy the same M-Scale endpoint as Appendices G and E respectively — Z_asym → 0 via the equilibrium premise, not via damping.

G3.Positiononthem-Scale

m-Scale operation away from strict canonical equilibrium is not considered. Canonical equilibrium removes the magnetic field from Gibbs distribution, cancels the generative factor and enforces Z_asym→0.

Kirchhoff’s Law [1] of Thermal Radiation (1860) states that the directional emissivity (ε) and absorptivity (α) of a body in thermodynamic equilibrium must match exactly at any given frequency. This law is rooted in Lorentzreciprocity: forward and backward propagation must be identical. While this law holds in the absence (T-Even) of a magnetic field, experiments by Guo, Zhao and Fan [2] show that a field (T-Odd) applied to magneto-optical materials breaks this law by differently skewing emissivity and absorptivity:

H1.MicroscopicTimeSymmetry

The magnetic field introduces a T-Odd symmetry to the magneto-optical material:

The magnetic field forces the material’s permittivity tensor to become anisotropic and complex.This forces the internal electron dynamics to spin asymmetrically.At narrow angles, incoming photons travel with the electron spin, outgoing photons against it, which decouples the absorption and emission channels, allowing the difference between ε and α.

H2.MacroscopicIrreversibility

Asymmetric radiation by non-reciprocal material can cause a temperature difference ΔT that can drive a heat engine and break the second law. To restore temperature uniformity, researchers [2] rely on globalhemisphericalintegration which integrates anomalous radiative effects over a whole hemisphere.

This integration is merely a firstlawenergybalancesheet. It tallies the total incoming and outgoing energy over all angles to show that energy is hemisphericallyconserved.It operates as a global eraser, averaging energy over a hemisphere by masking distinct, localized anomalous effects arising at narrow angles thereby preserving the first law.

H3.Positiononthem-Scale

Hemispherical integration confirms hemispherical energy conservation—a first law statement—while remaining silent on angle-resolved second law behavior at m-Scale. The integration averages emissivity and absorptivity over the full hemisphere, erasing the directional asymmetry and enforcing Z_asym→0.

Maxwell’s Demon [1] represents the archetype strawman attack on the second law. A chamber holding a gas is divided by a partition containing a gate. In the simpler Leo Szilard’s version, the demon produces apressuredifference by only opening the gate to particles moving from A to B regardless of their speed.

I1.MicroscopicTimeSymmetry

In this thought experiment both the demon and the gas are implicitly assumed to be T-Even symmetric.

I2.MacroscopicIrreversibility

Charles Bennett [1] applies Landauer’s framework to show how the whole experiment, including the demon, complies with the law. Information collected by the demon on particles must be stored and erased and is subjected to Landauer’s bound. Erasing a bit dissipates at least k_BT ln2, and at best, the demon can only achieve ΔS = 0, an unsurprising result given that the demon and the gas are both T-Even symmetric.

I3.RehabilitatingtheDemon

This thought experiment can be redeemed using a T-Odd gas or a T-Odd demon:

1.T-Oddgas: The E × B drift serves as a good example for T-Odd symmetry in a gas. Charged gas particles subjected to E and B fields perpendicular to each other, follow cycloidal orbits, drifting with velocity v in the cross-product direction of the fields [2] [3].

This setting has no need to track particles’ positions and velocities and open the gate. Left idle, the demon is freed from Landauer’s limit. The E × B drift yields the E × B Thermoelectric Effect [2].

2.T-OddDemon: A T-Odd demon has a natural supply of negentropy. This thought experiment discussed by Levy [3] shows that only half of the free energy G = A_Loaded available to a “loaded” demon (See Equation (40)), can be used to lower entropy. This statement effectively extends the second law to:

I4.Positiononthem-Scale

The demon and gas are both stipulated T-Even, so its index of asymmetry is Z_asym=0 by construction. A demon, rehabilitated with a T-Odd gas or source, operates at Z_asym≈1.

Feynman proposes the Ratchet and Pawl thought experiment [1] to prove that a spatially asymmetric device cannot rectify thermal noise into useful macroscopic work. A tiny paddle wheel axially joined to a saw-toothed ratchet wheel is enclosed in an isothermal gas. A spring-loaded pawl pinches asymmetrically indented ratchet teeth to allow rotation forward but not backward. Feynman shows that if the entire device is isothermal, forward and reverse rotations are equally likely, and nousefulworkisproduced.

J1.MicroscopicTimeSymmetry

This result is predictable given that spatialasymmetryandT-Oddsymmetry are different. Comparing the ratchet to a PN diode is instructive. Just like the ratchet, the diode is fabricated as spatially asymmetric. Its T-Odd symmetry emerges when a bias alters its shape. The thickness of its depletion zone changes with forward or reverse bias. It has a state variable linked to a sensitivity (built-in potential) and a bandwidth that restricts its operation. In contrast, the ratchet is spatially fixed and instantaneous, causing each particle to face the same geometry. Equations of particle motion are T-Even. Hence the ratchet is not T-Odd.

J2.MacroscopicIrreversibility

Given that the ratchet is T-Even, it can, at best, achieve dS/dt = 0.

J3.RedeemingtheRatchet

The ratchet can be rehabilitated by endowing it with a T-Odd symmetry having an amplitude sensitivity and a frequency bandwidth matched to the input noise. If its response is too slow, it averages out over many collisions to an invariant form, operates in M-Scale, and produces no useful work. Conversely, if the response time is too fast, it resets itself between interactions generating no useful work. As shown in Figure 5, optimality lies at the mid-point in the mesoscale where the index of asymmetry is unity. Experiments on various kinds of rectifiers by Kondo, Nikulov and Gurtovoi, Sheehan, and Lee, which may have achieved this goal are discussed in Appendix L.

J4.Positiononthem-Scale

The ratchet’s spatial profile is time-invariant, so its asymmetry has no response rate and Z_asym→0 — the same endpoint reached by Stosszahlansatz in Appendix A. A ratchet endowed with a state variable, a transition energy, and a bandwidth matched to the noise operates at Z_asym≈1.

The experimental record supporting the second law is vast, spanning two centuries from Joule’s calorimetry to modern single-molecule thermodynamics. The purpose of this appendix is not to dispute any of these results—all are correct—but to locate them on the m-Scale using the index of asymmetry, and to show that this enormous body of evidence samples only the two endpoints of the scale where the theory of this paper also predicts ΔS ≥ 0. The evidence is therefore consistent with the conventional second law and with its extension in Section 10 alike; it cannot discriminate between them.

K1.ConfirmingExperimentsatM-Scale(Z_asym→0)

Foundational experiments behind the Clausius and Kelvin-Planck statements were conducted on macroscopic working substances—water, gases, engines—with no T-Odd process present. Joule’s paddle-wheel experiment (1845) [1] converted work to heat in water with no applied magnetic field: the T-Odd factor in Z_asym is absent, and the index is not merely small but identically zero. The same holds for the calorimetric experiments underlying classical thermochemistry and for every heat-engine efficiency measurement in the Carnot tradition. None of these experiments were conducted at m-Scale.

The modern precision tests deserve particular attention because they are conducted at the nanoscale and are sometimes described as testing the second law “at its limits.” Wang et al. (2002) [2] observed transient negative entropy production in a colloidal bead dragged through water, confirming the Evans-Searles fluctuation theorem. Toyabe et al. (2010) [3] demonstrated information-to-energy conversion by a feedback-controlled Brownian particle, and Bérut et al. (2012) [4] verified Landauer’s bound k_BT ln 2 for the erasure of a single bit in an optically trapped colloid. All three are landmark experiments, and all three employ overdamped colloidal particles in water with no velocity-dependent field. In the language of this paper, their dynamics are T-Even by construction: the generative numerator of Z_asym is zero, exactly as in Joule’s water. Smallphysicalsizedoesnotimplym-Scaleoperation; a nanoscale T-Even system sits at the same endpoint of the index as a steam engine. These experiments confirm—with beautiful precision—the behavior of the Z_asym = 0 endpoint, including the transient negentropy fluctuations that the fluctuation theorem itself predicts and then averages away.

K2.ConfirmingExperimentsatμ-Scale(reversibility)

A smaller body of experiments probes the opposite endpoint by isolating a system from thermodynamic interaction and demonstrating that its evolution is reversible or unitary. The Hahn spin echo (1950) [5] and its descendants (Loschmidt echoes, time-reversal mirrors in NMR) show that an apparently dephased ensemble of spins retains its microscale information and can be re-focused by inverting the dynamics: ΔS_fine = 0. These experiments confirm the μ-Scale endpoint: information is preserved precisely because collisions and thermodynamic couplings are suppressed for the duration of the experiment. Negentropy, in the sense of this paper, is present in the retained correlations but in the absence of thermodynamic interaction it is not exported—there is no connection to a load, and nothing is harvested.

K3.TheSamplingGap

The record can now be stated compactly. Experiments at Z_asym = 0 uniformly find ΔS ≥ 0; experiments at Z_asym = ∞ find ΔS = 0; and both findings are predicted equally by the conventional second law and by the extended law of Section 9, whose utilization curve (Figure 5) vanishes at both endpoints. No confirming experiment has been conducted on a T-Odd system in the interior of the m-Scale, near Z_asym ≈ 1, where the two laws diverge most significantly and the extended law predicts ΔS ≥ −A/4T.

Logical induction can be a trap. An accumulation of evidence, however large, is not fully persuasive when every observation is drawn from the same region of parameter space. Bertrand Russell’s chicken [6], fed each morning, accumulates hundreds of confirmations of the farmer’s benevolence—all collected under identical conditions—until the morning the conditions change, and the farmer wrings its neck. The volume of second-law confirmations is not in question; their coverage is. The experiments that matter for discriminating between the conventional and extended laws are precisely the rare ones conducted on T-Odd systems between the endpoints of m-Scale, and these are the subject of Appendix L.

K4.Positiononthem-Scale

All experimental evidence supporting the second law was collected at the two extremes of the m-Scale:

1. at M-Scale where Z_asym → 0, T-Even dynamics and overwhelming dissipation guarantee ΔS ≥ 0.

2. at μ-Scale where Z_asym = ∞—the time-reversible limit—where isolation guarantees ΔS = 0.

NoexperimentwasperformedonaT-Oddsystematm-ScalewhereZ_asym≈1.

Appendix K showed that all confirming experiments cluster at the endpoints of the m-Scale, where the conventional and extended laws agree. This appendix achieves two goals:

1. Collects the rarer experiments reporting isothermal heat-to-work conversion with no imposed temperature gradient. They cover a wide range of physical phenomena: organic, chemical, biological, superconducting.

2. Provides a unifying explanation for these disparate systems: they all share the single underlying structure as described in Section 4.

L1.TheCommonMechanismofRectifiers:Fluctuation-ModulatedAsymmetry

Section 4.2 distinguishes spatial asymmetry from T-Odd symmetry. A structure that is merely spatially asymmetric and temporally fixed presents the same geometry to every fluctuation; its equations of particle motion remain T-Even. As proven by the H-Theorem, such a structure cannot rectify thermal noise. This is why Feynman’s ratchet fails, and why its failure is not the general result it is usually taken to be. Tosidestepthetheorem,adevicemustbeT-Oddsymmetric.

A rectifier constructed as a spatially symmetric device, becomes T-Odd symmetric when its spatial configuration is modulated by the fluctuations it acts upon. In a diode, the depletion-zone thickness and conductivity change with the sign of the bias; the structure the carriers encounter on the forward half-cycle is not the structure they encounter on the reverse half-cycle. This change is modulated by the noise itself and is not instantaneous: it is limited by the bandwidth of a state variable of the device. Three properties follow, and together they constitute an explanation for current results and a predictive test for future experiments on rectifiers:

1.Astatevariable whose configuration changes with the sign of the fluctuations (depletion thickness, interfacial charge state, surface coverage, gate conformation, superconducting phase).

2.Atransitionenergy—the fluctuation amplitude required to move the state variable (built-in potential, activation energy, condensation energy).

3.Abandwidth—a finite response rate over which the state variable follows the fluctuations.

The third property is what places such a device on the m-Scale. The bandwidth information is embedded in the conjugate factors—the generative T-Odd factor and the dissipative T-Even factor—that form the dimensionless index of asymmetry (i.e., ωτ). A response too slow relative to the noise correlation rate averages the fluctuations to an invariant profile (M-Scale, Z_asym → 0, the ratchet’s fate), while a response too fast resets between successive fluctuations and rectifies nothing (μ-Scale). Useful output requires the response to be commensurate with the thermal rate—Z_asym ≈ 1, the center of the m-Scale, exactly the Goldilocks condition of Section 6.

This mechanism sidesteps the H-theorem rather than contradicting it. The Stosszahlansatz assumes a fixed scattering geometry, so that post-collision correlations may be discarded and the same statistical premise reapplied at the next collision. When the configuration itself is set by the preceding fluctuation, correlation is carried forward through the state variable; in such a device the decorrelation step no longer holds. Appendix A’s H-Theorem result is still valid for the system it describes but does not extend to this one. In the electrical case, the fluctuations being rectified are the Johnson-Nyquist voltage fluctuations of the conductor, and the device is a diode. In the general case, the fluctuations are those of whichever variable is conjugate to the device’s state variable—concentration and chemical potential for a membrane, configuration coordinates for a molecular gate. The electrical case is developed quantitatively in a companion paper [1]; the entries below are grouped by how directly they realize it.

L2.ElectricalRectificationofElectricalThermalNoise

Kondoet al.[1] demonstrate an organic device producing 384 mV open-circuit, 1.1 μA/cm² short-circuit current density, and 94 nW/cm² maximum output at room temperature with no applied temperature gradient. This is the most direct realization of the mechanism above: the charge-transfer interface supplies the state variable, its energetics supply the transition energy, and its transfer rate supplies the bandwidth. The device is, in the terms of this paper, a rectifier of Johnson-Nyquist noise operating near Z_asym ≈ 1—a slower interface would average the noise away, a faster one would relax between fluctuations. A quantitative index from the authors’ spectroscopic transfer rates, compared against the thermal rate k_BT/h ≈ 6 × 10¹² Hz at 300 K, is developed in the companion paper.

Nikulov [2] andGurtovoiet al. [3] observe a DC voltage across a segment of an asymmetric superconducting ring switched between superconducting and normal states by non-equilibrium noise, one segment acting as a DC source. This is the mechanism in unusually transparent form: the ring’s asymmetry is structural and fixed, but the switching driven by the noise modulates it in time, satisfying the self-modulation condition exactly. State variable, transition energy, and bandwidth are all identifiable (order parameter, condensation energy, quasiparticle relaxation rate), making this system—with Kondo’s—among the best candidates for a computed index.

FuandFu [4] direct thermally emitted electrons between identical Ag-O-Cs plates in a vacuum tube using a magnetic field, reporting an isothermal potential difference between the plates. Here the asymmetry is magnetic rather than fluctuation-modulated, and the index is genuinely of Hall-parameter form Z_asym = ωτ, with the operative τ set by the transit time across the gap rather than by a bulk mean-free-time. The device functions only when the cyclotron turning and the transit time are commensurate, so that emitted electrons are steered preferentially to one plate—again an order-unity condition. This entry realizes the same m-Scale condition as the rectifiers above, but through the E × B mechanism of Section 4.1 rather than the diode mechanism of Section 4.2.

L3.ChemicalRectificationPoweredbyElectricalThermalNoise

The following entries produce chemical imbalances through processes powered by thermal energy. These processes must be T-Odd, otherwise, they would contradict the H-Theorem. This time-reversal asymmetry can only arise in the presence of magnetic effects at the molecular level (since no external field is applied), or in accordance with the mechanism outlined above—thermal noise modulates a space asymmetry. It is conjectured that the second alternative is the correct one: the Johnson-Nyquist noise induced by thermal fluctuations causes a T-Odd symmetry in the work potential, chemical composition, concentration, or Knudsen number. This effect results in the production of a chemical, an increase in concentration, or asymmetric flow through microscopic pores. As in the rectifier cases the theory’s prediction is that a state variable, a transition energy, and a finite bandwidth can be identified, and that output is maximized when the response rate matches the thermal fluctuation rate.

Sheehanet al. [5]-[8] report a concentration cell driven by anisotropic diffusion through a chemically asymmetric membrane, with claimed power density possibly exceeding 10⁷ W/m³, and separately, steady-state temperature differences above 120 K across epicatalytic rhenium and tungsten filaments in molecular hydrogen.

Lee [9]-[11] reports anaerobic microorganisms (Methanosarcina) converting heat to ATP isothermally, with culture temperature dropping by a mean of 0.25 ± 0.06˚C, and terms these “type-B” processes. Enzymatic gating is a natural candidate for fluctuation-modulated asymmetry: a gate whose conformation is driven by thermal fluctuations and whose asymmetry biases transitions is structurally the mechanism described above. “Type-B processes would then be a biological second-law-avoidance mechanism discovered by evolution.”

L4.Assessment

These results collectively involve a genuinely T-Odd mechanism operating in the interior of the m-Scale—the region Appendix K showed the confirming record never sampled—and the theory predicts that this region admits ΔS ≥ −A/4T.

The unifying explanation is predictive and falsifiable. If these systems are fluctuation-modulated rectifiers, then each must exhibit a state variable, a transition energy, and a response bandwidth; output must fall off when the response rate is detuned from the thermal fluctuation rate in either direction; and maximum output must occur at Z_asym ≈ 1 with a utilization factor bounded by 1/4. The proper response should be neither dismissal nor uncritical acceptance, but a program aimed to repeat these studies, identify their state variables, transition energies and bandwidths, and optimize them as shown in Figure 5.

L5.Positiononthem-Scale

These experiments were successful because 1) they included a T-Odd symmetry, they were performed at m-Scale 0 < Z_asym < ∞ away from the endpoints, but not necessarily at Z_asym = 1 where the energy and bandwidth of the mechanism matches the energy and bandwidth of the thermal noise.