Magnetic Moment Fragmentation in Frustrated Magnets

• Magnetic moment fragmentation is the separation of the magnetization field into an ordered divergence-full component and a fluctuating divergence-free part, highlighting coexistent static order and dynamic fluctuations.
• It is experimentally evidenced by the simultaneous observation of magnetic Bragg peaks and pinch-point diffuse scattering in pyrochlore and kagome systems.
• The phenomenon underpins theoretical frameworks in monopole crystallization and emergent Coulomb phases, offering insights into charge ordering and spin liquid behavior.

Magnetic moment fragmentation is a mode of organization in frustrated magnets in which the magnetic moment field separates into two coexisting components carried by the same microscopic spins: a divergence-full component that forms static long-range order, and a divergence-free component that remains fluctuating and Coulomb-phase-like. In the pyrochlore and kagome settings that define the subject, the phenomenon is not phase separation into ordered and disordered spatial regions; rather, order and disorder coexist in different channels of the same magnetic degree of freedom. Its canonical experimental fingerprint is the simultaneous presence of magnetic Bragg peaks from the ordered sector and pinch points, structured diffuse scattering, or persistent low-temperature dynamics from the fluctuating sector (Brooks-Bartlett et al., 2013, Lhotel et al., 2021, Petit et al., 2016).

1. Definition and field-theoretic formulation

The standard formulation is a Helmholtz decomposition of the magnetization field. In the review literature this is written as

$$\vec M = \vec M_{\mathrm{m}} + \vec M_{\mathrm{d}} = \vec \nabla \psi(\mathbf r) + \vec \nabla \wedge \vec A,$$

with the divergence-full longitudinal part $\vec M_{\mathrm m} = \vec \nabla \psi$ carrying magnetic charge and the divergence-free transverse part $\vec M_{\mathrm d} = \vec \nabla \wedge \vec A$ obeying

$$\vec \nabla \cdot \vec M = -\rho_{\mathrm m}, \qquad \vec \nabla \cdot \vec M_{\mathrm d} = 0.$$

In pyrochlore spin ice this continuum statement has a discrete counterpart on the diamond-lattice bonds,

$\sum_{J=1}^{4} M_{IJ} = -Q_I,$

so the local flux variable can be split into an ordered charge-carrying term and a divergence-free remainder (Lhotel et al., 2021).

The physical content of the decomposition is especially transparent for a single monopole configuration. For a tetrahedron in a $3$-in/$1$-out state, the Nd$_2$Zr$_2$O$_7$ analysis writes

$\vec M_{\mathrm m} = \vec \nabla \psi$0

The first term is all-in or all-out and therefore divergence-full; the second sums to zero and is divergence-free. The essential point is that the monopole charge does not exhaust the entire moment field, leaving a fluctuating Coulomb-phase component even in the presence of ordered charge (Petit et al., 2016).

This formulation changes the usual dichotomy between conventional long-range order and spin-liquid disorder. In fragmentation, the ordered and fluctuating sectors are orthogonal components of the same field, and the coexistence of the two is the phenomenon to be explained rather than an anomaly to be removed (Brooks-Bartlett et al., 2013).

2. Spin ice, monopoles, and fragmented Coulomb phases

The concept emerged from spin ice on the pyrochlore lattice, where Ising moments point along local $\vec M_{\mathrm m} = \vec \nabla \psi$1 axes and the ice rules impose a $\vec M_{\mathrm m} = \vec \nabla \psi$2-in/$\vec M_{\mathrm m} = \vec \nabla \psi$3-out constraint on each tetrahedron. In the dumbbell representation, each spin is replaced by a magnetic needle on the diamond lattice, so $\vec M_{\mathrm m} = \vec \nabla \psi$4-in/$\vec M_{\mathrm m} = \vec \nabla \psi$5-out tetrahedra are charge neutral, $\vec M_{\mathrm m} = \vec \nabla \psi$6-in/$\vec M_{\mathrm m} = \vec \nabla \psi$7-out and $\vec M_{\mathrm m} = \vec \nabla \psi$8-in/$\vec M_{\mathrm m} = \vec \nabla \psi$9-out tetrahedra are single monopoles, and all-in or all-out tetrahedra are double monopoles. The defining advance of "Magnetic-Moment Fragmentation and Monopole Crystallization" was the claim that a Coulomb phase can persist as a fluctuating background inside an ordered monopole crystal, producing an ordered divergence-full component plus an ergodic divergence-free component (Brooks-Bartlett et al., 2013).

In that framework the monopole crystal is a charge-ordered state on the diamond lattice. The 2013 theory introduces an order parameter

$\vec M_{\mathrm d} = \vec \nabla \wedge \vec A$0

and derives a ground-state criterion $\vec M_{\mathrm d} = \vec \nabla \wedge \vec A$1 for monopole crystallization in the reduced chemical-potential description. Monte Carlo results further identify a tricritical point near $\vec M_{\mathrm d} = \vec \nabla \wedge \vec A$2, separating continuous and first-order parts of the transition line (Brooks-Bartlett et al., 2013).

The review literature characterizes the resulting phase as a fragmented spin liquid (FSL): a state with long-range charge order and a residual Coulomb spin liquid. This state differs from an ordinary ordered magnet because only the divergence-full component orders, and it differs from a conventional spin liquid because symmetry breaking is present while the fluctuating sector remains a genuine emergent gauge field with algebraic correlations and pinch points (Lhotel et al., 2021).

A closely related route does not require long-range monopole Coulomb interactions. In Ho$\vec M_{\mathrm d} = \vec \nabla \wedge \vec A$3Ir$\vec M_{\mathrm d} = \vec \nabla \wedge \vec A$4O$\vec M_{\mathrm d} = \vec \nabla \wedge \vec A$5, a staggered local magnetic field generated by the Ir all-in-all-out order acts as a chemical potential for magnetic charge:

$\vec M_{\mathrm d} = \vec \nabla \wedge \vec A$6

or, in charge variables,

$\vec M_{\mathrm d} = \vec \nabla \wedge \vec A$7

This produces a fragmented regime for

$\vec M_{\mathrm d} = \vec \nabla \wedge \vec A$8

intermediate between ordinary spin ice and fully polarized all-in-all-out order (Lefrançois et al., 2017).

3. Material realizations in pyrochlore magnets

Pyrochlore realizations divide into direct demonstrations, quantitative single-crystal cases, candidates, proximate reduced-moment states, and explicit null results. The materials most often discussed are summarized below.

Material	Key observations	Interpretation
Nd$\vec M_{\mathrm d} = \vec \nabla \wedge \vec A$9Zr$$\vec \nabla \cdot \vec M = -\rho_{\mathrm m}, \qquad \vec \nabla \cdot \vec M_{\mathrm d} = 0.$$0O$$\vec \nabla \cdot \vec M = -\rho_{\mathrm m}, \qquad \vec \nabla \cdot \vec M_{\mathrm d} = 0.$$1	AIAO Bragg peaks together with pinch points; finite-energy flat mode around $$\vec \nabla \cdot \vec M = -\rho_{\mathrm m}, \qquad \vec \nabla \cdot \vec M_{\mathrm d} = 0.$$2; dipolar-octupolar pseudospin-$$\vec \nabla \cdot \vec M = -\rho_{\mathrm m}, \qquad \vec \nabla \cdot \vec M_{\mathrm d} = 0.$$3 description with $$\vec \nabla \cdot \vec M = -\rho_{\mathrm m}, \qquad \vec \nabla \cdot \vec M_{\mathrm d} = 0.$$4 and $$\vec \nabla \cdot \vec M = -\rho_{\mathrm m}, \qquad \vec \nabla \cdot \vec M_{\mathrm d} = 0.$$5	Experimental observation and quantum explanation of fragmentation (Petit et al., 2016, Benton, 2016)
Nd$$\vec \nabla \cdot \vec M = -\rho_{\mathrm m}, \qquad \vec \nabla \cdot \vec M_{\mathrm d} = 0.$$6Sn$$\vec \nabla \cdot \vec M = -\rho_{\mathrm m}, \qquad \vec \nabla \cdot \vec M_{\mathrm d} = 0.$$7O$$\vec \nabla \cdot \vec M = -\rho_{\mathrm m}, \qquad \vec \nabla \cdot \vec M_{\mathrm d} = 0.$$8	AIAO order below $$\vec \nabla \cdot \vec M = -\rho_{\mathrm m}, \qquad \vec \nabla \cdot \vec M_{\mathrm d} = 0.$$9; nearly flat band at $\sum_{J=1}^{4} M_{IJ} = -Q_I,$0 with pinch-point momentum dependence; dispersive branches at $\sum_{J=1}^{4} M_{IJ} = -Q_I,$1 and $\sum_{J=1}^{4} M_{IJ} = -Q_I,$2 producing half-moons	Quantitative single-crystal realization described by a minimal dipolar-octupolar XYZ Hamiltonian (Luo et al., 6 Oct 2025)
Sm$\sum_{J=1}^{4} M_{IJ} = -Q_I,$3Ti$\sum_{J=1}^{4} M_{IJ} = -Q_I,$4O$\sum_{J=1}^{4} M_{IJ} = -Q_I,$5	Dipolar-octupolar Ising ground-state doublet with $\sum_{J=1}^{4} M_{IJ} = -Q_I,$6 and $\sum_{J=1}^{4} M_{IJ} = -Q_I,$7; $\sum_{J=1}^{4} M_{IJ} = -Q_I,$8 AIAO order below $\sum_{J=1}^{4} M_{IJ} = -Q_I,$9 K; ordered moment $3$0; zero-field $3$1SR with no spontaneous oscillations down to $3$2 K	Moment-fragmentation candidate with AIAO order plus persistent spin dynamics (Mauws et al., 2018)
Ho$3$3Ir$3$4O$3$5	Ir-driven staggered field on Ho moments; Bragg peaks plus diffuse scattering; $3$6 and $3$7	Field-induced fragmented monopole crystal (Lefrançois et al., 2017)
Nd$3$8GaSbO$3$9	$1$0 AIAO order with $1$1; low-energy mode at $1$2; no spin-ice diffuse scattering or pinch points	Absence of moment fragmentation; conventional AIAO antiferromagnet (Gomez et al., 2021)

Nd$1$3Zr$1$4O$1$5 remains the reference case because the neutron data directly show Bragg peaks from all-in-all-out order together with pinch-point features characteristic of a Coulomb phase, and the quantum analysis attributes this to the dipolar-octupolar symmetry of the Nd$1$6 doublet and the decoupling of divergence-full and divergence-free sectors in the equations of motion (Petit et al., 2016, Benton, 2016).

Sm$1$7Ti$1$8O$1$9 extends the candidate class to samarium pyrochlores. Its crystal-field ground state is a pure $_2$0 Kramers doublet, neutron diffraction gives the AIAO structure, and zero-field $_2$1SR reveals persistent low-temperature dynamics without spontaneous oscillations. The authors therefore identify it as having “all the requisite ingredients for moment fragmentation physics” (Mauws et al., 2018).

Not every reduced-moment pyrochlore is fragmented. Nd$_2$2Sb$_2$3Mg$_2$4O$_2$5 exhibits a homogeneous ordered moment of $_2$6, only $_2$7 of the crystal-field-saturated value $_2$8, and a gap $_2$9 that rules out thermal fluctuations as the source of the reduction. The paper presents the compound as close to a moment-fragmented crystallized-monopole state, but not as a direct demonstration, and stresses that the data do not establish a separation into independently observable divergence-full and divergence-free components (Scheie et al., 2019).

4. Artificial lattices and engineered routes

Artificial kagome dipolar spin ice provided one of the clearest real-space realizations of fragmentation. In that setting each nanomagnet behaves as an Ising spin on the kagome lattice, the low-temperature manifold satisfies the kagome ice rule, and each triangle carries a unit magnetic charge

$_2$0

The magnetic degree of freedom then separates into a divergence-full part that carries ordered charges and a divergence-free part that remains fluctuating (Canals et al., 2016).

A central theoretical result is the exact rewriting of the microscopic dipolar Hamiltonian, to leading terms, as a hybrid spin-charge model

$_2$1

The vertex charges are defined by

$_2$2

and the local spin field is decomposed by a lattice Helmholtz-Hodge form,

$_2$3

In the SI1 manifold the ice rule is satisfied but the charges are disordered; in the lower-temperature SI2 manifold the charges crystallize antiferromagnetically while the divergence-free sector remains fluctuating (Canals et al., 2016).

Experimentally, the artificial system is a thermally active array of connected Gd$_2$4Co$_2$5 nanomagnets. After thermal annealing above the Curie temperature and imaging by XMCD-PEEM, the data reveal antiferromagnetic charge crystallites in real space, while reciprocal-space structure factors show both Bragg peaks and a structured diffuse background. The authors emphasize that charge order alone is not sufficient to claim fragmentation; the claim requires simultaneous evidence for the still-disordered divergence-free channel (Canals et al., 2016).

Field-engineered fragmentation in pyrochlore spin ice provides a complementary route. In Ho$_2$6Ir$_2$7O$_2$8, the ordered Ir sublattice injects magnetic charge into the Ho spin-ice manifold through a staggered local field. This stabilizes a monopole crystal with $_2$9-in/$_7$0-out order on one tetrahedral sublattice and $_7$1-in/$_7$2-out on the other, while leaving a Coulomb-phase-like fluctuating component. The reported low-temperature Ho moment of about $_7$3 is roughly half the full ground-doublet moment, and the “missing” moment appears as diffuse scattering rather than vanishing (Lefrançois et al., 2017).

5. Experimental diagnostics

The decisive diagnostic is the coexistence of reciprocal-space signatures that would normally be mutually exclusive. In Nd$_7$4Zr$_7$5O$_7$6, neutron scattering reveals magnetic Bragg peaks at wave vectors such as $_7$7 and $_7$8 together with arm-like features and pinch points at positions such as $_7$9 and $\vec M_{\mathrm m} = \vec \nabla \psi$00. The Bragg peaks identify static all-in-all-out order, whereas the pinch points identify a surviving Coulomb-phase component. The simultaneous appearance of both in the same measurements is the key evidence for fragmentation (Petit et al., 2016).

Single-crystal spectroscopy sharpens this criterion by resolving the dynamical sector. In Nd$\vec M_{\mathrm m} = \vec \nabla \psi$01Sn$\vec M_{\mathrm m} = \vec \nabla \psi$02O$\vec M_{\mathrm m} = \vec \nabla \psi$03, elastic scattering establishes AIAO order, while inelastic scattering below $\vec M_{\mathrm m} = \vec \nabla \psi$04 shows a gapped spectrum containing a nearly flat band with pinch-point momentum dependence and dispersive branches that generate half-moon patterns across multiple Brillouin zones. Within the fitted dipolar-octupolar XYZ Hamiltonian, the physical moment

$\vec M_{\mathrm m} = \vec \nabla \psi$05

is explicitly split into a static AIAO component and a dynamic fragmented sector (Luo et al., 6 Oct 2025).

Local probes are informative but require care. In Sm$\vec M_{\mathrm m} = \vec \nabla \psi$06Ti$\vec M_{\mathrm m} = \vec \nabla \psi$07O$\vec M_{\mathrm m} = \vec \nabla \psi$08, zero-field $\vec M_{\mathrm m} = \vec \nabla \psi$09SR is described by

$\vec M_{\mathrm m} = \vec \nabla \psi$10

with weak relaxation at high temperature, critical slowing near $\vec M_{\mathrm m} = \vec \nabla \psi$11, and a low-temperature plateau below about $\vec M_{\mathrm m} = \vec \nabla \psi$12 K; the absence of spontaneous oscillations down to $\vec M_{\mathrm m} = \vec \nabla \psi$13 K is interpreted as persistent spin dynamics inside the ordered phase (Mauws et al., 2018). In Nd$\vec M_{\mathrm m} = \vec \nabla \psi$14Sn$\vec M_{\mathrm m} = \vec \nabla \psi$15O$\vec M_{\mathrm m} = \vec \nabla \psi$16, by contrast, the flat mode sits at $\vec M_{\mathrm m} = \vec \nabla \psi$17, above the estimated $\vec M_{\mathrm m} = \vec \nabla \psi$18SR dynamical window of $\vec M_{\mathrm m} = \vec \nabla \psi$19, so the absence of dynamical interference in earlier $\vec M_{\mathrm m} = \vec \nabla \psi$20SR does not imply the absence of fragmentation (Luo et al., 6 Oct 2025).

Real-space imaging supplies the strongest direct visualization. In artificial kagome ice, the measured spin configurations can be decomposed explicitly into charge-carrying divergent and fluctuating divergence-free components, and the authors stress that the joint appearance of Bragg scattering and diffuse scattering is the experimental fingerprint. This standard also explains why the absence of spin-ice diffuse scattering in Nd$\vec M_{\mathrm m} = \vec \nabla \psi$21GaSbO$\vec M_{\mathrm m} = \vec \nabla \psi$22 argues against fragmentation despite the presence of AIAO order and a flat low-energy mode (Canals et al., 2016, Gomez et al., 2021).

6. Consequences, limits, and related uses of the term

Fragmentation reorganizes not only magnetic correlations but also the electric response of monopole-bearing spin-ice states. In the 2020 analysis of electric activity at magnetic moment fragmentation, magnetic monopoles are assigned electric dipoles derived from Hubbard-model charge redistribution, with the local charge on a triangle written as

$\vec M_{\mathrm m} = \vec \nabla \psi$23

In the fragmented state, ordered monopoles and antimonopoles pair these dipoles into local $\vec M_{\mathrm m} = \vec \nabla \psi$24 structures, which strongly reduces bulk dielectric response and microwave absorption relative to a monopole liquid. Domain walls and wrong-sublattice defects break this compensation and generate unpaired dipoles, making such textures electrically active (Khomskii, 2020).

The theoretical status of candidate phases remains material dependent. In Nd$\vec M_{\mathrm m} = \vec \nabla \psi$25Sn$\vec M_{\mathrm m} = \vec \nabla \psi$26O$\vec M_{\mathrm m} = \vec \nabla \psi$27, the absence of any photon-like excitation in the accessible window $\vec M_{\mathrm m} = \vec \nabla \psi$28 and the absence of a monopole continuum place strong constraints on the proposed Coulombic antiferromagnet scenario, even though they do not absolutely exclude it if the photon lies below experimental resolution (Luo et al., 6 Oct 2025). In the Nd pyrochlore family more generally, fragmentation is explicitly described as non-ubiquitous: Nd$\vec M_{\mathrm m} = \vec \nabla \psi$29GaSbO$\vec M_{\mathrm m} = \vec \nabla \psi$30 shows no evidence for it, and the conclusion drawn there is that chemical pressure, rather than $\vec M_{\mathrm m} = \vec \nabla \psi$31-site disorder by itself, is a key control parameter for its presence or absence (Gomez et al., 2021).

A common misconception is that any reduced ordered moment implies fragmentation. The kagome Nd compound Nd$\vec M_{\mathrm m} = \vec \nabla \psi$32Sb$\vec M_{\mathrm m} = \vec \nabla \psi$33Mg$\vec M_{\mathrm m} = \vec \nabla \psi$34O$\vec M_{\mathrm m} = \vec \nabla \psi$35 shows why this inference is insufficient: its reduced moment is homogeneous, static disorder and thermal fluctuations are excluded, and a multipolar explanation is suggested, but the work explicitly stops short of proving fragmentation because it does not establish a clear separation into divergence-full and divergence-free sectors (Scheie et al., 2019).

The term “fragmentation” is also used in a distinct many-body sense outside frustrated-magnet gauge physics. In spin-1 Bose gases, fragmentation denotes a condensate whose one-body density matrix has multiple macroscopic eigenvalues rather than a single macroscopic eigenvalue, and Goldstone-mode instability or magnetic-field compensation can drive or stabilize that form of fragmentation. This usage concerns spinor-condensate occupancy structure, not the Helmholtz decomposition of a magnetization field that defines magnetic moment fragmentation in spin ice and related magnets (Kawaguchi, 2013, Zhang et al., 2014).