Magneton: Units, Concepts, and Diverse Applications

• Magneton is a term for magnetic moment units, notably the Bohr and nuclear magnetons, that link microscopic spin properties to macroscopic observables.
• It underpins quantitative assessments in condensed matter, lattice QCD, and defect magnetism by normalizing diverse magnetic measurements.
• Additionally, 'Magneton' serves as a project name for cutting-edge software in protein representation learning and ML energy profiling, illustrating its diverse use.

Magneton is a technically overloaded term. In most of the cited literature it denotes a scale or unit of magnetic moment, especially the Bohr magneton $\mu_B$ and, in some contexts, the nuclear magneton $\mu_N$; these units normalize moments of ions, defects, quasiparticles, baryons, and proposed beyond-Standard-Model particles. The same word is also used as a proper name for two 2025 software systems—one for substructure-aware protein representation learning and one for differential energy debugging in machine learning—and one 2019 paper uses “magneton” in a nonstandard sense for a proposed “quantum magnet” field structure (Franz et al., 2023, Fein et al., 2022, Calef et al., 19 Dec 2025, Pan et al., 9 Dec 2025, Markoulakis et al., 2019).

1. Bohr and nuclear magnetons as units of magnetic moment

The standard microscopic magneton in these papers is the Bohr magneton,

$\mu_B=\frac{e\hbar}{2m_e},$

reported numerically in one condensed-matter study as about $9.274\times 10^{-24}\,\mathrm{J/T}$. It is treated as the natural unit for moments arising from electron spin and orbital motion, so magnetic moments in transition-metal oxides, intermetallics, graphene defects, and many-body quasiparticles are routinely expressed in units of $\mu_B$ (Sharma et al., 2013).

A second normalization is the nuclear magneton,

$\mu_N=\frac{e\hbar}{2m_p},$

which is smaller by roughly the electron-to-proton mass ratio, so that

$\mu_N \approx \frac{1}{1836}\mu_B.$

This smaller scale is relevant when the active magnetic degree of freedom is nuclear spin or rotational magnetism rather than electronic spin (Fein et al., 2022).

The cited literature uses these units in several reporting conventions. Moments are given per formula unit, as in $\mu_B/\mathrm{f.u.}$; per adatom, as in $\mu_B$ per fluorine adatom on graphene; per impurity spin in diluted magnetic semiconductors; and as site-resolved ordered moments on crystallographically distinct sites. A separate lattice-QCD literature also defines baryon-specific magneton units, including the “natural baryon magneton”

$[nBM]=\frac{e}{2M_B(m_\pi)},$

to factor out hadron-mass dependence when comparing octet-baryon magnetic moments across pion masses (Parreno et al., 2016).

2. Magnetic moments in solids: ions, formula units, and local defects

In magnetic solids, the magneton most often appears as a compact microscopic measure linking bulk magnetization to ionic or defect-level degrees of freedom. In the multiferroic double perovskite $\mu_N$0, the reported moment scale is $\mu_N$1, close to the spin-only estimate from one $\mu_N$2 and one $\mu_N$3, both taken as $\mu_N$4, while the directly measured low-temperature hysteresis reaches only about $\mu_N$5 up to $\mu_N$6. In the full-Heusler alloys $\mu_N$7 and $\mu_N$8, the saturation magnetizations are $\mu_N$9 and $\mu_B=\frac{e\hbar}{2m_e},$0 at $\mu_B=\frac{e\hbar}{2m_e},$1, to be compared with the Slater–Pauling value $\mu_B=\frac{e\hbar}{2m_e},$2 for $\mu_B=\frac{e\hbar}{2m_e},$3 (Sharma et al., 2013, Bainsla et al., 2015).

Site-resolved measurements show that the same unit also captures very small ordered moments. In $\mu_B=\frac{e\hbar}{2m_e},$4, Mössbauer-derived Fe moments are only $\mu_B=\frac{e\hbar}{2m_e},$5–$\mu_B=\frac{e\hbar}{2m_e},$6 on the $\mu_B=\frac{e\hbar}{2m_e},$7 site and $\mu_B=\frac{e\hbar}{2m_e},$8–$\mu_B=\frac{e\hbar}{2m_e},$9 on the $9.274\times 10^{-24}\,\mathrm{J/T}$0 site, with excess Fe on $9.274\times 10^{-24}\,\mathrm{J/T}$1 estimated at $9.274\times 10^{-24}\,\mathrm{J/T}$2; the paper emphasizes that these values are more than an order of magnitude smaller than theoretical predictions. In frustrated spinel $9.274\times 10^{-24}\,\mathrm{J/T}$3, the ordered $9.274\times 10^{-24}\,\mathrm{J/T}$4 moment is $9.274\times 10^{-24}\,\mathrm{J/T}$5 from zero-field $9.274\times 10^{-24}\,\mathrm{J/T}$6Co NMR and $9.274\times 10^{-24}\,\mathrm{J/T}$7 from neutron diffraction, both smaller than the $9.274\times 10^{-24}\,\mathrm{J/T}$8 spin-only expectation for $9.274\times 10^{-24}\,\mathrm{J/T}$9 and $\mu_B$0 (Żukrowski et al., 2017, Roy et al., 2013).

Defect magnetism in low-dimensional systems uses the same normalization at yet another scale. For an isolated fluorine adatom on graphene, hybrid DFT predicts a spin-polarized ground state with a magnetic moment of $\mu_B$1 per adatom, energetically favored over the spin-unpolarized state by $\mu_B$2. Only $\mu_B$3 lies inside the fluorine muffin-tin sphere; the rest is distributed over a long-range, sublattice-alternating spin polarization of graphene $\mu_B$4 states (Kim et al., 2013).

Across these cases, $\mu_B$5 is not merely a reporting convention. It serves as the bridge between macroscopic observables and microscopic assignments such as ionic spin states, site selectivity, defect localization, and disorder-induced reductions from idealized spin-only values.

3. Emergent, collective, and effective magnetons

Several papers extend the magneton concept beyond static ionic moments to emergent collective modes. In diluted magnetic semiconductors arranged into Lieb–Mattis superstructures, the ferrimagnetic saturation magnetization remains of order $\mu_B$6 per spin even at arbitrarily low impurity concentration, with the low-field ferrimagnetic moment per impurity written as

$\mu_B$7

and full saturation as $\mu_B$8 (Kuzian et al., 2016).

A more dramatic extension appears in the theory of phonon magnetic moments. One microscopic model based on orbit-lattice coupling predicts that chiral phonons in magnetic materials can acquire effective moments of order $\mu_B$9, and in some cases several $\mu_N=\frac{e\hbar}{2m_p},$0, because optical phonons hybridize with low-lying orbital electronic transitions. Reported values include $\mu_N=\frac{e\hbar}{2m_p},$1 for the $\mu_N=\frac{e\hbar}{2m_p},$2 mode of $\mu_N=\frac{e\hbar}{2m_p},$3, $\mu_N=\frac{e\hbar}{2m_p},$4 and $\mu_N=\frac{e\hbar}{2m_p},$5 for two $\mu_N=\frac{e\hbar}{2m_p},$6 modes in the same material, and approximately $\mu_N=\frac{e\hbar}{2m_p},$7–$\mu_N=\frac{e\hbar}{2m_p},$8 for an $\mu_N=\frac{e\hbar}{2m_p},$9 mode in $\mu_N \approx \frac{1}{1836}\mu_B.$0 (Chaudhary et al., 2023). A related gauge-theory treatment of doped Dirac semimetals connects the phonon magnetic moment to the electrical Hall conductivity through the phonon Hall viscosity, yielding

$\mu_N \approx \frac{1}{1836}\mu_B.$1

with first-principles values $\mu_N \approx \frac{1}{1836}\mu_B.$2 for the graphene $\mu_N \approx \frac{1}{1836}\mu_B.$3 mode and $\mu_N \approx \frac{1}{1836}\mu_B.$4 for the $\mu_N \approx \frac{1}{1836}\mu_B.$5 mode (Chen et al., 2024).

Matter-wave interferometry provides a complementary experimental range for magnetic moments. A single interferometric Stern–Gerlach device is reported to probe magnetic moments ranging from a Bohr magneton to less than a nuclear magneton. In that work, alkali atoms occupy the $\mu_N \approx \frac{1}{1836}\mu_B.$6 scale, the organic radical TEMPO is described by an effective moment $\mu_N \approx \frac{1}{1836}\mu_B.$7, and thermal $\mu_N \approx \frac{1}{1836}\mu_B.$8 exhibits rotational magnetic moments with maximum projection $\mu_N \approx \frac{1}{1836}\mu_B.$9 (Fein et al., 2022).

4. Fundamental, hadronic, astrophysical, and transport normalizations

Outside condensed matter, the magneton remains the standard normalization for magnetic couplings but often with very different physical meaning. In stellar constraints on the neutrino magnetic dipole moment, the parameter of interest is $\mu_B/\mathrm{f.u.}$0 in units of $\mu_B/\mathrm{f.u.}$1, with

$\mu_B/\mathrm{f.u.}$2

Using an emulator of stellar-evolution calculations and Bayesian marginalization over stellar-physics nuisance parameters, one analysis concludes that the tip of the red giant branch does not constrain

$\mu_B/\mathrm{f.u.}$3

within the explored range (Franz et al., 2023).

In lattice QCD, the magneton can be redefined to absorb hadron-mass dependence. The octet-baryon study based on background magnetic fields argues that moments show only mild pion-mass dependence when expressed in natural baryon magnetons,

$\mu_B/\mathrm{f.u.}$4

In these units, anomalous moments organize into a simple pattern: $\mu_B/\mathrm{f.u.}$5 and $\mu_B/\mathrm{f.u.}$6 near $\mu_B/\mathrm{f.u.}$7, $\mu_B/\mathrm{f.u.}$8 and $\mu_B/\mathrm{f.u.}$9 near $\mu_B$0, and $\mu_B$1 and $\mu_B$2 near $\mu_B$3 (Parreno et al., 2016).

Transport theory uses the same language in an effective sense. A strange-metal model introduces an “effective Bohr magneton”

$\mu_B$4

so that the field-driven relaxation rate takes the form

$\mu_B$5

Here the magneton is not a bare spin moment but an emergent orbital field-to-rate conversion scale set by the renormalized effective mass $\mu_B$6 (Kim et al., 1 Apr 2025).

A related contrast between magneton scales appears in the Pomeranchuk-effect literature. One ultrahigh-field study states that the Bohr magneton of an electron is $\mu_B$7 times larger than the nuclear magneton of $\mu_B$8He, making electron systems far more sensitive to magnetic fields than liquid or solid $\mu_B$9He. In that framework, magnetic entropy and magnetization are controlled by the Zeeman scale $[nBM]=\frac{e}{2M_B(m_\pi)},$0 for the electron system but by the much smaller nuclear-moment scale in $[nBM]=\frac{e}{2M_B(m_\pi)},$1He (Matsuyama et al., 29 Dec 2025).

5. Magneton as a proper name in computational research

In 2025, “Magneton” was adopted as the proper name of two unrelated computational systems. In protein machine learning, Magneton is an environment for substructure-aware protein representation learning. It provides a dataset of $[nBM]=\frac{e}{2M_B(m_\pi)},$2 proteins annotated with over $[nBM]=\frac{e}{2M_B(m_\pi)},$3 million substructures spanning $[nBM]=\frac{e}{2M_B(m_\pi)},$4 types, a training framework for incorporating substructures into existing protein models, and a benchmark suite of $[nBM]=\frac{e}{2M_B(m_\pi)},$5 tasks probing residue-, substructure-, and protein-level representations. Using this environment, the authors introduce substructure-tuning, a supervised fine-tuning method that improves function prediction, yields more consistent representations of substructure types never observed during tuning, and shows that substructural supervision provides information complementary to global structure inputs (Calef et al., 19 Dec 2025).

A second 2025 Magneton is an ML-systems energy profiler built around “differential energy debugging.” This Magneton compares semantically equivalent operator-level computations across similar ML systems and attributes excess GPU energy to code regions or configuration choices responsible for avoidable waste. Applied to $[nBM]=\frac{e}{2M_B(m_\pi)},$6 popular ML systems spanning LLM inference, general ML frameworks, and image generation, its abstract states that it detects and diagnoses $[nBM]=\frac{e}{2M_B(m_\pi)},$7 known cases of software energy inefficiency and discovers $[nBM]=\frac{e}{2M_B(m_\pi)},$8 previously unknown cases, $[nBM]=\frac{e}{2M_B(m_\pi)},$9 confirmed by developers. The detailed evaluation reports successful diagnosis of $\mu_N$00 of $\mu_N$01 collected known cases, average end-to-end energy reduction of $\mu_N$02 on diagnosed cases, replay-based operator power errors of $\mu_N$03, $\mu_N$04, and $\mu_N$05 on representative operators, and tracing overheads of $\mu_N$06 for HuggingFace Transformers and $\mu_N$07 for vLLM (Pan et al., 9 Dec 2025).

These software uses are terminologically distinct from the magnetic-moment unit. In both cases, “Magneton” functions as a project name rather than a physical observable.

6. Nonstandard and model-specific usages

One paper, “Real time observation of a stationary magneton,” explicitly uses “magneton” in a nonstandard sense. There, the term denotes an elementary “Quantum Magnet” or the intrinsic stationary magnetic field structure associated with a single electron or subatomic magnetic dipole, rather than the Bohr magneton or nuclear magneton as units. The same paper introduces “Quantum Field of Magnet (QFM)” for the corresponding macroscopic field of a permanent magnet and describes the observed geometry as a “dipole vortex shaped magnetic flux geometrical pattern,” “two quantum magnetic flux vortices geometrical patterns, joint back to back,” and a “nested double counter torus joint hemispheres field geometry constituting a sphere” (Markoulakis et al., 2019).

A different particle-physics paper introduces the related but distinct coinage “magneticon.” In that model, generalized Maxwell equations with magnetic charge and current, together with a vorton-pair composite picture and dyality symmetry, predict a stable spin-$\mu_N$08 magnetic-sector counterpart of the electron carrying magnetic charge $\mu_N$09 and zero electric charge. The proposed magneticon is therefore not a magneton unit; it is a hypothetical magnetically charged particle whose pair-production cross section in $\mu_N$10 collisions is written as

$\mu_N$11

with the paper arguing that traditional monopole searches were not sensitive to such a low-charged magnetic monopole above about $\mu_N$12 (Sullivan et al., 2015).

Taken together, these papers show that “magneton” can denote a standard magnetic-moment unit, an effective coupling scale, a software platform, or—more exceptionally—a model-specific field concept. The dominant scientific usage across the cited literature remains the Bohr-magneton normalization of magnetic moments and magnetic couplings.