Spin Ice: Magnetic Monopoles Hiding in a Crystal

The pyrochlore lattice and the ⟨111⟩ Ising spins

Spin ice lives on the pyrochlore lattice: the magnetic rare-earth ions (Dy³⁺ or Ho³⁺) sit at the corners of a network of corner-sharing tetrahedra. Each ion belongs to two tetrahedra, and this shared-corner geometry is the seed of everything that follows.

Two ingredients make the magnetism Ising-like. First, the crystal field at each rare-earth site produces a large single-ion anisotropy that pins the magnetic moment along the local ⟨111⟩ axis — the line joining the two tetrahedron centres. A Dy³⁺ moment cannot point in an arbitrary direction; it is a classical Ising variable that either points into a chosen tetrahedron or out of it. Second, the moment is enormous: the ground-state doublet of Dy³⁺ (J = 15/2) gives

• μ ≈ 10 μ_B (μ_B = 9.274 × 10⁻²⁴ J·T⁻¹), the largest practical Ising moment in any magnet.
• The crystal-field gap to the first excited doublet is ≈ 300 K, so below a few kelvin only the two Ising states matter.

Because the anisotropy axis differs from site to site, a normally frustrated antiferromagnetic coupling turns into an effective ferromagnetic constraint, and vice versa — the geometric twist that makes ferromagnets frustrate on this lattice.

The ice rule and why it is frustrated

The dominant interaction between neighbouring moments — a mix of exchange and, crucially, magnetic dipole coupling — is minimised when each tetrahedron adopts a "two-in, two-out" configuration: two spins point inward, two outward. This is the magnetic analogue of Bernal and Fowler's ice rule for the protons around each oxygen in water ice, formalised by Pauling in 1935.

Count the states. A single tetrahedron has 2⁴ = 16 spin arrangements. The number satisfying two-in/two-out is C(4,2) = 6. There is no way to satisfy every tetrahedron uniquely and simultaneously — the constraints from corner-sharing conflict — so the ground state is massively degenerate rather than a single ordered pattern. That is geometric frustration: the lattice geometry, not disorder, forbids a unique minimum.

Pauling's estimate treats tetrahedra as locally independent. For N spins there are 2^N total configurations; each tetrahedron kills a fraction 6/16 = 3/8; with N/2 tetrahedra the surviving count is

• W ≈ 2^N · (3/8)^(N/2) = (3/2)^(N/2).
• Residual entropy per mole of spins: S₀ = (R/2) ln(3/2) ≈ 1.68 J·mol⁻¹·K⁻¹.

In 1999 Ramirez and collaborators integrated the measured heat capacity C(T)/T of Dy₂Ti₂O₇ and found the missing entropy matched Pauling's number to within a few percent — a direct thermodynamic fingerprint of the ice rule.

Dipoles, not exchange: the surprise of dipolar spin ice

The first shock is that nearest-neighbour exchange in Dy₂Ti₂O₇ is actually antiferromagnetic, which alone would not give spin ice. What rescues the ice rule is the long-range magnetic dipole interaction between those giant 10 μ_B moments:

• E_dd = (μ₀/4π)[ (m₁·m₂)/r³ − 3(m₁·r̂)(m₂·r̂)/r³ ], with μ₀ = 4π × 10⁻⁷ T·m·A⁻¹.
• At the nearest-neighbour distance r ≈ 3.5 Å, this sets the dominant energy scale, D ≈ 1.4 K, comparable to the ordering temperature.

Remarkably, this long-range, anisotropic interaction projects onto an effective nearest-neighbour ferromagnet almost exactly — the higher multipole tails nearly cancel. The dipolar spin ice model (den Hertog and Gingras, 2000) reproduces the residual entropy, the heat capacity, and the neutron structure factor of the real materials with essentially one free coupling. The lesson: spin ice is one of the rare places where a full 1/r³ interaction conspires to mimic a simple local rule.

Flipping a spin makes two monopoles

Now the beautiful part, due to Castelnovo, Moessner and Sondhi (2008). Reverse one spin in a perfect ice-rule state. The two tetrahedra sharing that spin each lose the two-in/two-out balance: one becomes three-in/one-out, the other one-in/three-out. These are the elementary excitations.

Assign each Ising spin a pair of magnetic charges (a "dumbbell") at the two tetrahedron centres it touches. A two-in/two-out tetrahedron has net charge zero. A three-in/one-out centre carries a net magnetic charge +Q, and three-out/one-in carries −Q. The magnitude is set by the moment and the diamond-lattice spacing a_d between tetrahedron centres:

• Q = 2μ / a_d ≈ 4.6 μ_B·Å⁻¹ for Dy₂Ti₂O₇ — a genuine magnetic charge.
• Flipping a chain of adjacent spins moves the two charges apart at no extra cost in the nearest-neighbour model; they are deconfined.

Because magnetostatics is Coulombic, two monopoles separated by r feel

• V(r) = − μ₀ Q₁Q₂ / (4π r), a magnetic Coulomb law, plus a small entropic "Dirac-string" tension.

These are not the fundamental monopoles of grand-unified theory; they are emergent quasiparticles — collective spin flips that behave like point magnetic charges. The vacuum's rule ∇·B = 0 is not broken globally; it is only violated locally, at the defect tetrahedron, where field lines appear to source.

Numbers, scales, and what controls the physics

The controlling energy scales are all sub-kelvin, which is why spin ice is a dilution-refrigerator science:

• Effective coupling J_eff ≈ +1.1 K (ferromagnetic in Dy₂Ti₂O₇), setting the monopole creation cost, roughly Δ ≈ 4.35 K to make a pair.
• Freezing / entropy plateau near T ≈ 0.6 K, below which spin dynamics slow dramatically.
• Monopole density falls as n ∝ exp(−Δ/2k_BT); at 0.5 K only a dilute gas of charges survives, ideal for treating them as free particles.

Dynamics obey an Arrhenius law at low T: the spin-relaxation time τ = τ₀ exp(E_a/k_BT) with E_a of order a few kelvin, so τ climbs from microseconds toward seconds and beyond as T drops — the material becomes a magnetic "traffic jam." A key measurable is the magnetic conductivity: monopoles hop under an applied field H, and their motion produces a Wien-effect-like nonlinear response, directly analogous to ion conduction in an electrolyte (Bramwell et al., 2009, measured the monopole charge this way, obtaining ≈ 5 μ_B·Å⁻¹, matching the dumbbell prediction).

How we know: neutrons, entropy, and pinch points

Three classes of experiment nail the picture:

• Heat-capacity integration. Measuring C(T) from ~0.2 K to ~12 K and computing S(T) = S₀ + ∫ C/T dT leaves a shortfall of exactly (R/2)ln(3/2) — the entropy that never gets released because the ice manifold never orders.
• Neutron scattering "pinch points." The ice rule is a divergence-free condition on an emergent field, so the correlations are dipolar and produce sharp bow-tie singularities in the diffuse magnetic scattering — the signature of a Coulomb phase. Fennell et al. (2009) imaged them in Ho₂Ti₂O₇.
• Monopole signatures. Applying a field along [100] or [111] tilts the monopole gas and can be driven through a Kasteleyn transition; neutron and magnetisation data track the charges directly. Muon spin rotation and the magnetic Wien effect probe their diffusion.

The residual entropy, the pinch points, and the field-tunable charge transport form a mutually consistent story: a frozen-but-not-ordered magnet whose thermal excitations are fractionalised magnetic charges.

Subtleties, misconceptions, and where the field is going

These are not Dirac monopoles. They carry magnetic charge and obey a Coulomb law, but they are emergent, non-relativistic quasiparticles bound to a crystal; they cannot be extracted into vacuum, and there is no accompanying quantisation of electric charge. Nothing about spin ice bears on whether fundamental monopoles exist.

The residual entropy is not truly "zero-temperature entropy forever." The third law is safe in principle — quantum tunnelling and tiny further couplings would eventually select a ground state — but the relevant energy scales are so small (milliKelvin) and the dynamics so slow that the ice manifold is effectively frozen on any lab timescale. This is why Dy₂Ti₂O₇ falls out of equilibrium below ~0.6 K, complicating clean entropy measurements.

Not every pyrochlore is spin ice. Tuning the sign of J_eff and the anisotropy gives Ising antiferromagnets, ordered states, or — with quantum-scale moments like Pr³⁺ or Yb³⁺ — quantum spin ice, a proposed U(1) quantum spin liquid whose excitations include emergent photons and both magnetic and electric monopoles. Candidate materials such as Pr₂Zr₂O₇ and Ce₂Zr₂O₇ are active hunting grounds. Spin ice thus sits at the doorway of one of condensed matter's biggest goals: a lattice that emulates its own miniature electromagnetism, monopoles and light included.

Water ice vs. spin ice: the same combinatorics on different degrees of freedom

Property	Water ice (Ih)	Spin ice (Dy₂Ti₂O₇)
Frustrated object	H⁺ position on O–O bond	Ising spin on ⟨111⟩ axis
Local constraint	2 near / 2 far protons per O	2-in / 2-out per tetrahedron
Ground-state count	W ≈ (3/2)ᴺ (N molecules)	W ≈ (3/2)^(N/2) (N spins)
Residual entropy S₀	≈ 3.4 J·mol⁻¹·K⁻¹ (Giauque 1936)	≈ 1.68 J·mol⁻¹·K⁻¹ (Ramirez 1999)
Defect / excitation	Ionic + Bjerrum defects	Emergent magnetic monopole ±Q
Freezing temperature	≈ 273 K	≈ 0.6 K