---
title: Stoner Model in Itinerant Ferromagnetism
url: https://www.emergentmind.com/topics/stoner-model
type: topic
---

# Stoner Model in Itinerant Ferromagnetism

The Stoner model is a mean-field theory of itinerant electron ferromagnetism in which magnetism arises from a collective instability of the metallic Fermi sea toward spontaneous spin polarization. In contrast to localized moment models such as Heisenberg, the Stoner mechanism is rooted in the exchange-driven splitting of broad, partially filled electron bands, typically 3d or 4f, and predicts a ferromagnetic ground state when the gain in exchange energy outweighs the loss in kinetic energy associated with polarization. The quantitative criterion for the Stoner instability links the density of states at the Fermi level and an effective exchange interaction parameter, and the model provides analytical forms for spin susceptibility, Curie temperature, and the nature of the magnetic phase transition. However, the Stoner model exhibits notable limitations when confronted with strong correlations, lattice disorder, or multiorbital/fluctuation phenomena, necessitating beyond-mean-field refinements for quantitative and qualitative agreement with experiment.

## 1. Formulation and Stoner Instability Criterion

The canonical Stoner Hamiltonian for spin-½ itinerant electrons with local (Hubbard-type) repulsion reads
\[
H = \sum_{\mathbf{k},\sigma} \epsilon_{\mathbf{k}} c_{\mathbf{k}\sigma}^\dagger c_{\mathbf{k}\sigma} + U \sum_i n_{i\uparrow} n_{i\downarrow}
\]
where $\epsilon_{\mathbf{k}}$ is the bare dispersion, $U>0$ is the local interaction, and $n_{i\sigma}$ the electron number operator at site $i$.

Within mean-field theory, the exchange splitting between up/down bands is $\Delta = I M$ (with $M$ the spin polarization and $I$ an effective Stoner parameter), and the energy per unit cell as a function of $M$ is
\[
E(M) = E(0) + \frac{1}{2} \frac{M^2}{N(E_F)} - \frac{1}{4} I M^2
\]
where $N(E_F)$ is the density of states at the Fermi level. The instability to ferromagnetic order is found by minimizing $E(M)$; nontrivial solutions appear when
\[
I N(E_F) > 1
\]
which is the celebrated Stoner criterion. Equivalently, the static spin susceptibility $\chi$ in RPA form diverges when $U N(E_F) = 1$ [1807.11291], [2506.03416]. For multiorbital systems, $I$ can be estimated from Coulomb and Hund couplings as $I = (U+6J)/5$ [2506.03416].

## 2. Thermodynamic Properties and Magnetic Phase Transition

In the Stoner framework, the paramagnetic-to-ferromagnetic transition is signaled by divergence of the mean-field susceptibility:
\[
\chi_{\rm MF}^{-1}(T) = N(E_F)^{-1} + U
\]
so that $\chi_{\rm MF}(T)\to\infty$ as $U N(E_F)\to 1$ from below [2605.28761]. The mean-field Curie temperature is obtained from the temperature-dependent susceptibility or the expansion of free energy in $M$, leading to
\[
T_C \sim E_0 \exp\bigl[-1/(U N(E_F))\bigr]
\]
with $E_0$ a characteristic bandwidth (or cutoff) scale. This exponential dependence underpins the extreme sensitivity of $T_C$ to $U$ and $N(E_F)$ in flat or diverging density-of-states regimes, e.g., at Van Hove singularities [2605.28761], [2403.09147].

In two dimensions, the Stoner transition is generically of first order: the energy as a function of polarization $\zeta$ is quadratic,
\[
E_{\rm MF}(\zeta) = E_N + \frac{n^2}{2N(E_F)}(1 - N(E_F)U)\zeta^2
\]
producing a jump from $\zeta = 0$ to full polarization at $U_c$, with the susceptibility diverging up to the jump point [2409.18934].

## 3. Beyond Mean-Field Effects: Fluctuations, Correlations, Disorder

The bare Stoner model neglects all quantum fluctuations and spatial correlations. Leading corrections arise from:

- **Particle-particle ladder fluctuations:** In 2D, these generate large logarithmic corrections and renormalize $U$ downward:
  \[
  \tilde{U} = \frac{U}{1 + \alpha N(E_F)U \ln(\Lambda/k_F)}
  \]
  with $\alpha \approx 1/2$ in relevant channels, suppressing the Stoner instability and raising $U_c$ above the mean-field prediction [2409.18934].

- **Vertex corrections and renormalized susceptibility:** Diagrammatic Monte Carlo and bold-line schemes dress both Green's function ($G$) and interaction ($U^*(T)$), yielding
  \[
  \chi(T) = - \frac{\Pi(T)}{1 + U^*(T)\Pi(T)}
  \]
  with $\Pi(T)$ the *dressed* bubble and $U^*(T)$ the renormalized interaction. Key mechanisms for avoiding Stoner divergences at Van Hove points are: (i) downward renormalization of $U^*$ and (ii) strong suppression of the quasiparticle residue $Z(T)$, leading to
  \[
  U^*(T)\, D_{\rm eff}(E_F, T) = U^*(T)\, Z(T)\, D_0(E_F) < 1
  \]
  even at $T \ll T_{MF}$; no divergence appears in $\chi(T)$ [2605.28761].

- **Disorder:** In disordered systems, a replica analysis shows the effective Stoner criterion is shifted:
  \[
  1 \le U N(0) - \frac{3}{2} h^2 N'(0)
  \]
  where $h^2$ is related to disorder strength. For $N'(0)<0$ (decreasing DOS), disorder can *enhance* ferromagnetism, even inducing it in originally nonmagnetic systems [2508.20724].

- **Breakdown and Competing Orders:** In multiorbital systems, intrinsic altermagnetism (AM) can preempt conventional Stoner FM. The competition is controlled by Hund's coupling $J_H$ and interorbital hopping: for $J_H/U \lesssim 0.1$, AM emerges as the leading instability, breaking down the usual Stoner scenario [2510.00614].

## 4. Extensions: Specialized Lattice Systems and Phase Competition

Modifications of the Stoner model capture physics in specialized contexts:

- **Flat- and Partially Flat-Band Systems:** In a partially flat-band, $D(E_F)$ is parametrically enhanced, and the ferromagnetic state can emerge at weak $U \ll W$ (bandwidth). The Stoner instability is dramatically strengthened, and $T_C$ can be elevated by orders of magnitude [2403.09147].

- **Quantum Nanomagnetism:** The Stoner–Wohlfarth model, including quantum corrections, elucidates reversal dynamics in single-domain particles and quantum beats in magnetization during field sweeps. These quantum corrections vanish as $S\to\infty$, connecting quantum dynamics to classical spinodal transitions [1503.06658].

- **Disordered and Dirty Ferromagnets:** A modified Stoner mean-field theory with replicas yields phase diagrams with paramagnetic (PM), ferromagnetic (FM), and spin glass (SG) regions, predicting critical disorder strengths above which spin-glass phases intervene between PM and FM [2508.20724].

## 5. Experimental Relevance and Limitations

The Stoner criterion provides a practical framework for first-principles screening of itinerant ferromagnets, for instance, in half-Heusler compounds where it correlates with the occurrence of FM or nonmagnetic (gapped) phases [2506.03416]. In oxide, moiré, or quantum well systems tuned to Van Hove singularities, mean-field Stoner theory substantially *overpredicts* the prevalence of ferromagnetism; experiments instead observe superconductivity or local correlations, with the genuine instability suppressed by fluctuation and correlation-induced DOS renormalization [2605.28761]. The model also underpins experimental and theoretical work in ultracold Fermi gases, where rapid quenches into the upper repulsive branch can reveal the Stoner transition, provided loss rates are sufficiently suppressed [1412.2412].

However, the Stoner model has notable limitations:
- It consistently overestimates $T_C$ and yields non-Curie-Weiss susceptibilities above $T_C$ [1807.11291];
- Cannot account for local-moment fluctuations, Mott physics, or multiband/multivalley effects without extension;
- Fails for strong correlation regimes, e.g., in presence of high $U$ without additional correlation physics (Hubbard, Gutzwiller, DMFT, etc.);
- Ignores magnetic anisotropy, finite-size, and non-equilibrium quantum phenomena in nanoscale systems (addressed in quantum Stoner–Wohlfarth generalizations) [1503.06658].

## 6. Connections to Broader Theoretical Frameworks

The Stoner model is the foundation of itinerant magnetism theory but serves as an idealized limit within a hierarchy of correlated electron models. The Van Vleck–Hurwitz "middle-road" approach (restricted charge fluctuation), the Hubbard model (incorporating finite $U$), and dynamical mean-field theory each extend Stoner concepts by accounting for strong correlation and local moment physics. The full description of metallic magnetism often requires a combination of Stoner's mechanism for band-splitting, Hubbard’s $U$ for onsite repulsion, and explicit treatment of spin fluctuations or competing order parameters [1807.11291], [2605.28761], [2506.03416].

---

### Table: Representative Stoner Criterion Forms and Modifications

| Context                              | Instability Criterion                      | Reference         |
|---------------------------------------|--------------------------------------------|-------------------|
| Standard (mean-field, single-band)    | $U N(E_F) > 1$                             | [1807.11291]      |
| Multiband/cRPA estimated $I$          | $I N(E_F) > 1$                             | [2506.03416]      |
| 2D particle-particle ladders          | $\tilde{U} = U/(1 + \alpha N U \ln(\Lambda/k_F))$ | [2409.18934]      |
| Bold-line Monte Carlo (VH singularity)| $U^*(T) Z(T) D_0(E_F) < 1$                 | [2605.28761]      |
| Replica/disorder-enhanced             | $U[N(0) - \frac{3}{2}\Delta^2 N'(0)] \geq 1$| [2508.20724]      |
| Multiorbital/AM competition           | $1/U_c = \max[\chi_{FM}^{(0)}(0),\, \chi_{AM}^{(0)}(0)]$ | [2510.00614]      |

---

The Stoner model remains a central reference point in itinerant magnetism, but its predictive power is contingent on extensions that incorporate quantum fluctuations, electronic correlations, and multiorbital/multivalley physics as manifest in real materials and experimentally relevant regimes.

Source: https://www.emergentmind.com/topics/stoner-model