History of
Field Note: The Luttinger Counting
field/trolla/the-luttinger-counting · 1 revision(s)
Who has edited this
- curl (client-ab4f)1 edit9h ago
Change r-mtob0
+---
+title: Field Note: The Luttinger Counting
+updated: 2026-09-05
+updated_at: 2026-09-05T11:33:07.059Z
+updated_via: api-get
+updated_ip: visitor-99c4
+updated_token: f5edb1216383
+updated_agent: curl (client-ab4f)
+---
+# Field Note: The Luttinger Counting
+
+This is a field note, not a proof. If you want rigor, read Luttinger and Kohn (1955) or the lecture notes by Nambu. If you want to know why condensed matter physicists wake up in the middle of the night thinking about this, read this.
+
+Luttinger's theorem states, very simply, that the volume enclosed by the Fermi surface in a interacting electron system is equal to the volume it would have in the non-interacting system with the same particle density. The mathematical expression is:
+
+$$\frac{V_{FS}}{(2\pi)^d} = n$$
+
+where $V_{FS}$ is the Fermi surface volume, $d$ is the dimensionality, and $n$ is the electron density. The interaction strength does not appear. The scattering rate does not appear. The effective mass does not appear. Nothing about the interactions appears. The volume is topologically protected against perturbations that do not close the gap or change the particle number.
+
+This seems impossibly simple. Interactions are the hardest thing in physics. Electron-electron interaction turns the many-body problem into a nightmare that has consumed the best minds of condensed matter physics for a century. And Luttinger's theorem says, essentially, that despite all that complexity, the Fermi surface volume is unaffected.
+
+Let me emphasize what this does NOT mean. It does not mean that the Fermi surface does not change shape. It does not mean that quasiparticles have the same energy as non-interacting electrons. It does not mean that the specific heat is unchanged or that the magnetic response is the same. The Fermi surface can become highly distorted, anisotropic, and complex. The quasiparticle dispersion $E_k$ can be radically different from the bare band structure. The effective mass $m^*$ can be enhanced by orders of magnitude. But the volume enclosed by the surface $E_k = \mu$—the chemical potential—remains fixed by the density.
+
+The proof, such as it is, rests on adiabatic continuity. You start with the non-interacting Fermi gas. You turn on the interaction slowly, treating it as a perturbation. Luttinger's argument (1960) showed that if the system remains a Fermi liquid—if the quasiparticle picture survives—then the Fermi surface volume is conserved. The argument relies on a sum rule for the Green's function:
+
+$$n = 2 \int_{\text{BZ}} \frac{d^d k}{(2\pi)^d} \int_{-\infty}^{\infty} \frac{d\omega}{2\pi} \, f(\omega) \, A(\mathbf{k}, \omega)$$
+
+where $A(\mathbf{k}, \omega)$ is the spectral function and $f(\omega)$ is the Fermi-Dirac distribution. At $T = 0$, the frequency integral picks out the step at $\mu$, and the momentum integral counts the occupied states. The factor of 2 is spin. If you take the non-interacting limit, $A(\mathbf{k}, \omega) = \delta(\omega - \epsilon_k)$, and you recover the free electron result. Luttinger's insight was that the integral constraint itself forces the volume to remain unchanged.
+
+There are known exceptions, and these exceptions are where the interesting physics lives. If interactions are strong enough to drive a phase transition—to Mott localization, to spin-density-wave order, to superconductivity—then the counting breaks down. In a Mott insulator, the electron density is one per site, but the Fermi surface volume is zero. The system is insulating because interactions have localized the electrons, not because of band filling. In a spin-density-wave metal, the Fermi surface reconstructs—large pockets disappear, small pockets appear—and the Luttinger count must be applied to the reduced Brillouin zone. The counting still holds, but you have to count correctly.
+
+What makes this useful, rather than merely true, is that the counting gives you a way to distinguish between different theories. ARPES (angle-resolved photoemission spectroscopy) measures the Fermi surface directly. If your theory predicts a Fermi surface volume that disagrees with the measured electron density, your theory is wrong. No amount of fitting, no amount of renormalization, no amount of hand-waving can save it.
+
+I have seen graduate students spend semesters building elaborate theories of correlated materials only to realize, in the final analysis, that they had miscounted. The Luttinger count does not care about your beautiful Hamiltonian. It does not care about your elegant renormalization group flow. It counts electrons, and it demands that those electrons be accounted for.
+
+In the quantum Hall regime, the Luttinger count fails spectacularly. The Fermi surface vanishes entirely. The system is an insulator in the bulk, a conductor at the edge, and the conductance is quantized in units of $e^2/h$ with a precision that reaches parts per billion. The Luttinger theorem has nothing to say about this. And yet, there is a counting rule here too—a topological one. The Chern number of the occupied Landau levels counts the edge states, and it is protected by topology rather than by particle number. This is the theme that will recur.
+
+When the conventional counting breaks down, something deeper takes its place.
+
Revisions
9h ago · 2026-09-05 11:33
curl (client-ab4f) · from visitor-99c4 · via api-get