All lessons

Module 9: Condensed matter physics · Lesson 9.4 · about 45 min

The quantum Hall effect and low-dimensional systems

Electrons confined to wells, wires and dots: subbands and the density of states in 3D, 2D, 1D and 0D; the classical Hall effect and the resistivity tensor; Landau levels, their degeneracy eB/h and the filling factor; Shubnikov-de Haas oscillations; the integer quantum Hall effect with plateaus at h/νe², vanishing ρxx, disorder, localisation and edge states; the fractional effect and the half-integer effect in graphene.

Key points, meanings and formulas on one page: choose "Save as PDF" in the print box.

Module 9 · about 45 minutes. In 1980 von Klitzing measured the Hall resistance of a silicon transistor at 1.5 K and 18 T and found flat steps at h/e2 divided by a whole number, the same to parts per billion in every sample. Since 2019 that number, RK=h/e2=25812.807 Ω, helps define the SI. This lesson builds the effect from two ingredients: electrons confined to a plane, and the harmonic-oscillator levels a magnetic field gives them.

Squeeze electrons in one direction until only the lowest standing wave fits, and they move in two dimensions. A perpendicular field then collapses the continuous 2D spectrum into Landau levels, each holding eB/h states per unit area. When the Fermi level sits between levels, the bulk cannot carry dissipative current, ρxx vanishes, and the Hall resistance locks at h/νe², fixed by counting alone.

1. Confinement and the density of states

Confine electrons of effective mass m∗ in a well of width L in z (a GaAs layer between AlGaAs, or the inversion layer of a MOSFET). Motion in x and y stays free, so

E=Ej+ℏ2(kx2+ky2)2m∗,Ej=j2π2ℏ22m∗L2 (infinite well)

Each Ej starts a subband. If E2−E1≫kBT and only the first is filled, the gas is two-dimensional.

3D
`$g(E) \propto \sqrt E$` (see [Fermi gas](/learn/courses/gate-physics/fermi-gas/))
2D (well)
`$g(E) = m^*/\pi\hbar^2$` per subband per area, including spin: constant, so `$g$` is a staircase with a step at each `$E_j$`; `$n = (m^*/\pi\hbar^2)E_F$`
1D (wire)
`$g(E) \propto 1/\sqrt{E - E_{jk}}$`, diverging at each subband edge; a ballistic wire conducts `$2e^2/h$` per open subband (quantised conductance)
0D (dot)
discrete levels, `$g$` a set of delta functions: an artificial atom

Worked example. A 10 nm GaAs well (m∗=0.067me): E1=0.376 eV nm2/(0.067×100 nm2)=56 meV, E2−E1=3E1=168 meV, far above kBT=0.34 meV at 4 K. With n=3×1015 m⁻², EF=πℏ2n/m∗=10.7 meV: only the lowest subband is filled. (Real barriers are finite, so the true levels are a little lower; see one-dimensional potentials.)

2. The classical Hall effect

Current density jx flows in a sheet in field Bzˆ. In steady state the Lorentz force is balanced by a transverse Hall field. The Drude model gives the resistivity tensor

ρ=(ρxxρxy−ρxyρxx),ρxx=m∗ne2τ,ρxy=Bne

So classically ρxy rises linearly with B (the Hall slope gives n and the sign of the carriers) and ρxx does not depend on B. In 2D, resistance and resistivity have the same units: Rxy=VH/I=ρxy for any sample shape. Inverting the tensor:

σxx=ρxxρxx2+ρxy2,σxy=ρxyρxx2+ρxy2

If ρxy≠0, then ρxx=0 and σxx=0 happen together. A perfect insulator along x and a perfect conductor along x look the same in a field.

3. Landau levels

A 2D electron in field B circles at the cyclotron frequency ωc=eB/m∗. In the Landau gauge A→=(0,Bx,0) the Schrödinger equation separates: ψ=eikyyϕ(x), where ϕ is a harmonic oscillator centred at x0=ℏky/eB.

Energies. En = ħωc(n + ½), n = 0, 1, 2, ...; with spin, each splits by gμBB.

Degeneracy. The energy does not depend on ky. In a sample Lx × Ly, ky = 2πj/Ly and the centre x₀ must lie inside Lx, so j runs up to eBLxLy/h. Each level holds eB/h states per area (per spin): one per flux quantum h/e.

Filling factor. ν = n/(eB/h) = nh/eB counts filled spin-resolved levels. Raising B raises the degeneracy, so levels empty one by one and the Fermi level drops from one to the next.

Same states, regrouped. The 2D density of states m*/2πħ² per spin, times the spacing ħωc, is exactly eB/h: the field gathers the states of each slice into one level.

En=ℏωc(n+12),NA=eBh per level,ν=nheB,ℓB=ℏeB

Worked example. GaAs at 10 T: ℏωc=ℏeB/m∗=17.3 meV, eB/h=2.42×1015 m⁻², and the magnetic length ℓB=8.1 nm. A gas with n=3.0×1015 m⁻² has ν=1.24. The ν=2 point is at B=nh/2e=6.2 T.

Shubnikov-de Haas. Each time a level passes the Fermi energy, ρxx dips. The dips are equally spaced in 1/B, with Δ(1/B)=2e/hn when spin is not resolved (e/hn when it is). With n=3×1015 m⁻², Δ(1/B)=0.16 T⁻¹. The same periodicity in a 3D metal (de Haas-van Alphen) measures extremal Fermi-surface areas: Δ(1/B)=2πe/ℏAF.

4. The integer quantum Hall effect

At low temperature and high field, the dips of ρxx reach zero over finite ranges of B, and across the same ranges ρxy is perfectly flat:

ρxy=hνe2=25812.807 Ων,ρxx=0,σxy=νe2h

Clean sample, exactly ν levels full. Each full level carries σxy = e²/h, so σxy = νe²/h, the same as the classical line ne/B at that one field. Without disorder, the plateau would be a single point.

Disorder broadens each level. States in the tails are localised: trapped on contours of the random potential, they carry no current. Only a narrow band at each level's centre is extended.

Fermi level in localised states. As B changes, the Fermi level moves through the localised tails. The filled extended states do not change, so σxy stays at νe²/h and σxx = 0. This is the plateau. Disorder is needed for the plateaus.

Edges. At the sample edge the confining potential bends every level upwards, so each filled level crosses the Fermi energy there and makes a one-way (chiral) edge channel. Back-scattering would need a jump to the far edge, so each channel carries current without loss and has conductance e²/h.

Laughlin’s argument makes the exactness plain: threading one flux quantum h/e through a loop of the 2D gas moves exactly ν electrons from one edge to the other, whatever the disorder, which forces σxy=νe2/h. In modern language, ν is a topological invariant (the Chern number of the filled bands), and that is why it cannot change smoothly.

Top: the Landau fan, with energies rising linearly in B, and the Fermi level of a fixed density, which drops each time a level empties. Bottom: ρxy with plateaus at h/νe² and the classical line B/ne dashed, and ρxx, which vanishes on every plateau. Widen the band of extended states (more disorder in the figure's model) to see narrower plateaus.

Beyond the integers

Fractional QHE
in very clean samples (Tsui, Störmer, Gossard 1982) plateaus appear at `$\nu = 1/3, 2/5, 2/3, \ldots$` (odd denominators). It is a many-body effect: Laughlin's wavefunction at `$\nu = 1/m$` has quasiparticles of charge `$e/m$`. The composite-fermion picture attaches two flux quanta to each electron and turns these fractions into an integer effect
Graphene
massless Dirac electrons give Landau levels `$E_n = \pm v_F\sqrt{2e\hbar B|n|}$`, including one at zero energy shared by electrons and holes. With fourfold (spin and valley) degeneracy, `$\sigma_{xy} = 4(n + \tfrac12)e^2/h = \pm2, \pm6, \pm10, \ldots$` in units of `$e^2/h$`, and the effect survives to room temperature
Metrology
`$R_K = h/e^2$` with the Josephson constant `$K_J = 2e/h$` realises the ohm and the volt from fixed `$h$` and `$e$`

“The plateaus come from Landau levels alone; a perfectly clean sample would show the widest plateaus.” In a perfectly clean, infinite sample the Fermi level jumps straight from one level to the next and ρxy follows the classical line, touching h/νe2 only at isolated fields. Plateaus need states at energies between the levels that cannot carry current. Localised states from disorder supply them, which is why the effect was found in a rather dirty silicon MOSFET, and why the plateau value does not depend on the sample.

  • Subbands Ej=j2π2ℏ2/2m∗L2; DOS: E (3D), constant m∗/πℏ2 (2D), 1/E (1D), delta functions (0D). Ballistic wire: 2e2/h per subband.
  • Classical Hall: ρxy=B/ne, ρxx=m∗/ne2τ; ρxx=0⇔σxx=0 when ρxy≠0.
  • Landau levels ℏωc(n+12), ωc=eB/m∗, degeneracy eB/h per area per spin, ν=nh/eB, ℓB=ℏ/eB.
  • SdH dips periodic in 1/B: Δ(1/B)=2e/hn (spin degenerate).
  • IQHE: ρxy=h/νe2, ρxx=0; plateaus need localised states; chiral edge channels each carry e2/h; ν is topological.
  • FQHE at odd-denominator fractions, charge e/m quasiparticles; graphene σxy=4(n+12)e2/h.

Practise: exam questions on this lesson

Next: Module 10 moves into the nucleus. Lesson 10.2 uses the simplest bound nucleus, the deuteron, to read off the range, depth and spin dependence of the nuclear force. For the Fermi-Dirac counting behind the filled levels, see quantum statistics; for Hall measurements in ordinary semiconductors, see the JAM lesson on semiconductors.

Check yourself

1A 2D electron gas has n = 4.0 × 10¹⁵ m⁻². At B = 8.27 T the Hall resistance is closest to
2On a quantum Hall plateau, which pair is correct?
3The density of states of an ideal 2D electron gas (no field) is

Sources

  • K. von Klitzing, G. Dorda and M. Pepper, Physical Review Letters 45, 494 (1980)
  • J. H. Davies, The Physics of Low-Dimensional Semiconductors (1998), chapters 4, 6 and 8
  • D. Tong, Lectures on the Quantum Hall Effect (2016), chapters 1 and 2, arXiv:1606.06687
  • CSIR-UGC NET Physical Sciences syllabus, Part C condensed matter physics, published by the National Testing Agency

Your ticks stay in this browser. Save them as a file on the course home.

Updated 2026-11-22. Not affiliated with CSIR, UGC or the National Testing Agency. Check the official CSIR NET website for the current pattern. All questions are original.