11.2 Mott insulators

Mott insulators defy what band theory predicts about electrical conductivity. These materials have partially filled bands and should be metallic, yet strong electron-electron interactions force them into an insulating state. They typically contain partially filled d or f orbitals found in transition metal or rare earth compounds.

Grasping Mott insulators is central to many-body physics because they demonstrate how electron correlations can override single-particle predictions. Their study connects directly to high-temperature superconductivity, exotic magnetism, and the broader physics of strongly correlated systems.

Fundamentals of Mott insulators

Mott insulators form a distinct class of materials where insulating behavior arises not from a filled band, but from strong electron-electron interactions. Studying them reveals how electronic correlations reshape material properties in ways that single-particle band theory cannot capture.

Definition and basic properties

A Mott insulator is a material that band theory predicts to be metallic (because it has a partially filled band), yet it behaves as an insulator. The root cause is the large Coulomb repulsion cost when two electrons occupy the same lattice site. Rather than hopping freely, electrons become localized.

• Typically found in compounds with partially filled 3d, 4f, or 5f orbitals
• An energy gap opens in the electronic spectrum even though there is an odd number of electrons per unit cell
• Localized electronic states give rise to distinctive magnetic ordering and optical absorption features

Historical context and discovery

Sir Nevill Mott proposed the concept in the late 1930s to explain why certain transition metal oxides, most notably nickel oxide (NiO), are insulating despite having partially filled d bands. Band theory at the time had no mechanism to account for this.

Mott's insight was that Coulomb repulsion between electrons on the same site could split the partially filled band into two sub-bands separated by a gap. This led to the Mott-Hubbard model, later refined and extended to materials such as vanadium sesquioxide ($\text{V}_2\text{O}_3$), cuprates, and organic conductors.

Electronic structure of Mott insulators

The electronic structure of a Mott insulator cannot be understood within independent-electron band theory. You need to account for strong correlations and many-body effects explicitly.

Band theory limitations

Standard band theory treats electrons as independent particles moving in a periodic potential. For a half-filled band, it predicts a metal. The problem is that it completely ignores the energy penalty when two electrons sit on the same site.

This is why band theory fails for materials like NiO and $\text{V}_2\text{O}_3$: the on-site Coulomb repulsion $U$ is so large that it costs far more energy to doubly occupy a site than the electrons gain from delocalizing. The result is an insulator, not a metal.

Hubbard model

The Hubbard model is the simplest lattice model that captures Mott physics. It includes just two competing energy scales:

• Hopping integral $t$: the kinetic energy gained when an electron hops between neighboring sites
• On-site Coulomb repulsion $U$: the energy cost of placing two electrons on the same site

The Hamiltonian is:

$H = -t \sum_{\langle i,j \rangle,\sigma} (c_{i\sigma}^\dagger c_{j\sigma} + \text{h.c.}) + U \sum_i n_{i\uparrow} n_{i\downarrow}$

Here $c_{i\sigma}^\dagger$ creates an electron with spin $\sigma$ on site $i$, and $n_{i\sigma}$ is the number operator. The first term favors delocalization (metallic behavior); the second penalizes double occupancy (favoring insulating behavior).

When $U/t$ is small, kinetic energy wins and the system is metallic. When $U/t$ exceeds a critical value, the system becomes a Mott insulator.

Mott-Hubbard transition

The metal-to-insulator transition driven by increasing $U$ relative to the bandwidth $W \sim zt$ (where $z$ is the coordination number) is called the Mott-Hubbard transition. At the transition:

1. The density of states at the Fermi level develops a gap.
2. The single band splits into a lower Hubbard band (singly occupied states) and an upper Hubbard band (doubly occupied states), separated by an energy of order $U$.
3. Electrons become localized on their respective sites.

This transition can be driven experimentally by applying pressure (which increases $t$), changing temperature, or chemical doping. $\text{V}_2\text{O}_3$ is the classic example: it undergoes a first-order metal-insulator transition under pressure and with chromium doping. Certain organic salts like $\kappa$-(BEDT-TTF)$_2$ compounds also show pressure-tuned Mott transitions.

Electron correlation effects

Electron correlations are what make Mott insulators fundamentally different from band insulators. Three key interactions shape their behavior.

Coulomb repulsion

The on-site Coulomb repulsion $U$ is the dominant energy scale. In transition metal oxides, $U$ typically ranges from about 1 to 10 eV. For comparison, bandwidths $W$ in these materials are often only 1–3 eV.

When $U \gg W$, the energy cost of double occupancy far exceeds the kinetic energy gain from hopping. Electrons localize, one per site at half filling, and charge fluctuations are strongly suppressed. The material becomes a Mott insulator.

Exchange interaction

Once electrons are localized, their spins still interact. The exchange interaction arises from virtual hopping processes: an electron can briefly hop to a neighboring site and back, but only if the neighboring electron has opposite spin (Pauli exclusion). This lowers the energy of antiparallel spin configurations.

The result is an effective spin-spin interaction described by the Heisenberg model:

$H = J \sum_{\langle i,j \rangle} \mathbf{S}_i \cdot \mathbf{S}_j$

where the exchange coupling $J \sim t^2/U$ for the half-filled Hubbard model. Positive $J$ (in the convention above) favors antiferromagnetic order, which is why many Mott insulators are antiferromagnets.

Spin-orbit coupling

In materials with heavier elements (5d transition metals like iridium, or rare earths), the coupling between an electron's spin and its orbital angular momentum becomes significant. Spin-orbit coupling can:

• Mix spin and orbital degrees of freedom, producing effective $J_{\text{eff}}$ states (as in $\text{Sr}_2\text{IrO}_4$)
• Generate strong magnetic anisotropy
• Stabilize exotic phases such as topological Mott insulators and quantum spin liquids

This is an active frontier because spin-orbit coupling adds a new axis of complexity to the already rich Mott physics.

Types of Mott insulators

Mott physics appears across a surprisingly wide range of materials. Three major families stand out.

Transition metal oxides

These are the most studied Mott insulators. Compounds like NiO, CoO, MnO, and $\text{V}_2\text{O}_3$ feature partially filled 3d orbitals with strong Coulomb repulsion.

• Most exhibit antiferromagnetic ordering at low temperatures (e.g., NiO has a Néel temperature of ~523 K)
• $\text{V}_2\text{O}_3$ displays a rich phase diagram with a first-order metal-insulator transition near 150 K
• Cuprates (like $\text{La}_2\text{CuO}_4$) are parent Mott insulators that become high-temperature superconductors when doped

Rare earth compounds

Materials containing lanthanide (4f) or actinide (5f) elements can also be Mott insulators. The f electrons are highly localized and carry large magnetic moments.

• Examples include cerium-based heavy fermion compounds and samarium hexaboride ($\text{SmB}_6$)
• $\text{SmB}_6$ is a candidate topological Kondo insulator, where strong correlations and spin-orbit coupling combine to produce topologically protected surface states
• These materials often show complex magnetic structures and very large effective electron masses (heavy fermion behavior)

Organic Mott insulators

Molecular crystals built from organic molecules can realize Mott physics at much lower energy scales than oxides. The bandwidth is small (tens of meV), so even modest Coulomb repulsion can drive a Mott state.

• Key examples: $\kappa$-(BEDT-TTF)$_2$Cu[N(CN)$_2$]Cl and $\kappa$-(BEDT-TTF)$_2$Cu$_2$(CN)$_3$
• The latter compound is a leading candidate for a quantum spin liquid, a Mott insulator with no magnetic order down to the lowest measured temperatures
• Pressure can tune these materials through the Mott transition into a superconducting state
• They serve as clean model systems for studying Mott physics in quasi-two-dimensional geometries

Experimental techniques

Several complementary techniques are used to probe the electronic and magnetic structure of Mott insulators.

Optical spectroscopy

Optical spectroscopy measures the frequency-dependent optical conductivity $\sigma(\omega)$. In a Mott insulator, you see a clear suppression of spectral weight below the Mott gap energy, with absorption onset corresponding to excitations across the gap (from the lower to upper Hubbard band).

• Techniques include infrared spectroscopy, visible-range reflectivity, and spectroscopic ellipsometry
• Temperature- and pressure-dependent measurements can track how the gap evolves toward a metal-insulator transition
• Spectral weight transfer from high to low energies upon doping or heating is a hallmark signature of Mott physics

Photoemission spectroscopy

Photoemission spectroscopy (PES) directly measures the occupied electronic states by ejecting electrons with UV or X-ray photons and analyzing their kinetic energy.

• Angle-resolved photoemission spectroscopy (ARPES) provides momentum-resolved information, mapping out the band structure
• In Mott insulators, ARPES reveals the lower Hubbard band and the absence of a quasiparticle peak at the Fermi level
• Tracking how spectral weight redistributes across the Mott transition gives direct evidence for correlation-driven gap formation

X-ray absorption spectroscopy

X-ray absorption spectroscopy (XAS) probes unoccupied states by measuring the absorption of X-rays as a function of photon energy near an element's absorption edge.

• XANES (near-edge structure) is sensitive to the oxidation state and local symmetry of the absorbing atom
• EXAFS (extended fine structure) provides information about bond lengths and coordination
• Particularly useful for studying doped Mott insulators, where you need to understand how added carriers modify the local electronic environment

Physical properties

The interplay of charge, spin, and orbital degrees of freedom gives Mott insulators a distinctive set of physical properties.

Electrical conductivity

At low temperatures, Mott insulators are insulating with resistivity that follows thermally activated behavior:

$\rho(T) \propto \exp\left(\frac{E_g}{2k_BT}\right)$

where $E_g$ is the Mott gap. This looks similar to a semiconductor, but the gap origin is fundamentally different (correlations vs. band structure).

Near a Mott transition, transport becomes more complex. Doped Mott insulators often show non-Fermi liquid behavior, with resistivity that deviates from the standard $T^2$ dependence of a Fermi liquid. Some display linear-in-$T$ resistivity, a hallmark of the "strange metal" phase seen in cuprate superconductors.

Magnetic ordering

Because Mott insulators have localized spins with exchange coupling $J \sim t^2/U$, most develop long-range magnetic order at low temperatures.

• Antiferromagnetic order is the most common (NiO, $\text{La}_2\text{CuO}_4$, MnO)
• Ordering temperatures (Néel temperatures) range from a few Kelvin to several hundred Kelvin depending on the exchange coupling strength
• Magnetic excitations (magnons) are measured with inelastic neutron scattering
• Some geometrically frustrated Mott insulators avoid ordering entirely and may host quantum spin liquid states

Thermal properties

• Specific heat shows contributions from both lattice vibrations (phonons) and magnetic excitations
• A lambda-shaped anomaly in specific heat typically marks the magnetic ordering transition
• Thermal conductivity has phonon and magnon contributions; in spin liquid candidates, a large linear-$T$ term in thermal conductivity suggests itinerant spinon excitations
• Thermal expansion can show anomalies at metal-insulator and magnetic transitions, reflecting the coupling between lattice and electronic degrees of freedom

Mott vs. band insulators

Distinguishing Mott insulators from band insulators is important both conceptually and experimentally.

Energy gap formation

The spectral weight transfer point is a key experimental discriminator. In a Mott insulator, doping or heating moves spectral weight from above the gap to below it, which doesn't happen in a simple band insulator.

Temperature dependence

Band insulator gaps are relatively insensitive to temperature. Mott gaps, by contrast, can show strong temperature dependence because thermal fluctuations affect the correlation-driven gap. Some Mott insulators undergo a temperature-driven metal-insulator transition (e.g., $\text{V}_2\text{O}_3$ near 150 K), something that doesn't happen in band insulators under normal conditions.

Thermal excitations in Mott insulators can also create in-gap states, gradually filling in the gap from the edges as temperature increases.

Doping effects

Doping a band insulator introduces carriers into the existing conduction or valence band. The band structure itself doesn't change much.

Doping a Mott insulator is far more dramatic. Adding even a small number of carriers can collapse the Mott gap entirely, producing a metallic state. The resulting metal is often highly unconventional, with properties that don't fit standard Fermi liquid theory. The most famous example: doping the Mott insulator $\text{La}_2\text{CuO}_4$ with holes (by substituting Sr for La) produces high-temperature superconductivity in $\text{La}_{2-x}\text{Sr}_x\text{CuO}_4$.

Applications and technological relevance

Mott insulators and materials derived from them connect to several areas of technology and applied physics.

High-temperature superconductivity

The parent compounds of cuprate superconductors ($\text{La}_2\text{CuO}_4$, $\text{YBa}_2\text{Cu}_3\text{O}_6$) are antiferromagnetic Mott insulators. Upon hole doping, they become superconductors with critical temperatures exceeding 100 K. Understanding the Mott insulating parent state is widely considered essential for explaining the pairing mechanism.

Iron-based superconductors also show signatures of Mott-like correlations, though they are more itinerant than cuprates. Potential applications include lossless power transmission, high-field magnets, and sensitive magnetic detectors (SQUIDs).

Spintronics devices

Mott insulators with strong spin-orbit coupling can host topological surface states useful for generating and detecting spin currents. Antiferromagnetic Mott insulators are of interest for memory devices because antiferromagnets are insensitive to stray magnetic fields and can be switched on ultrafast (THz) timescales.

Challenges remain in optimizing interfaces between Mott materials and conventional semiconductors, and in achieving reliable spin injection and detection.

Quantum materials

Mott insulators provide a platform for realizing exotic quantum states:

• Quantum spin liquids in frustrated Mott insulators (e.g., herbertsmithite, $\kappa$-(BEDT-TTF)$_2$Cu$_2$(CN)$_3$) are candidates for topological quantum computation
• Topological Mott insulators could support robust edge states for quantum information processing
• Cold atoms in optical lattices can simulate the Hubbard model with tunable $U/t$, enabling controlled study of Mott transitions in a clean setting

Theoretical approaches

Describing Mott insulators quantitatively is one of the hardest problems in condensed matter theory. No single method handles all aspects perfectly.

Dynamical mean-field theory

Dynamical mean-field theory (DMFT) is the most successful analytical framework for Mott physics. It works by mapping the full lattice problem onto a single impurity site embedded in a self-consistent bath.

1. Start with a guess for the local self-energy $\Sigma(\omega)$.
2. Compute the local Green's function by integrating over momentum.
3. Extract the effective impurity problem (the Weiss field).
4. Solve the impurity problem (using exact diagonalization, quantum Monte Carlo, or other impurity solvers).
5. Obtain a new self-energy from the impurity solution.
6. Repeat until self-consistency is reached.

DMFT captures the Mott transition, the formation of Hubbard bands, and the quasiparticle peak that appears near the transition. Its main limitation is that it treats only local correlations exactly; non-local correlations (important for, e.g., d-wave superconductivity) require cluster extensions like cellular DMFT or the dynamical cluster approximation.

Density functional theory

Standard DFT based on local or semi-local exchange-correlation functionals (LDA, GGA) systematically underestimates correlations and often predicts Mott insulators to be metals.

• DFT+U adds a Hubbard-like $U$ correction to the localized orbitals, opening a gap. It's computationally cheap but requires choosing $U$ as an input parameter.
• Hybrid functionals (e.g., HSE06) mix in a fraction of exact exchange and can improve gap predictions, though they are more expensive.
• DFT+DMFT combines the realistic band structure from DFT with the local correlation treatment of DMFT. This is currently the state-of-the-art for first-principles calculations of Mott insulators.

Quantum Monte Carlo simulations

Quantum Monte Carlo (QMC) methods provide numerically exact solutions for model Hamiltonians like the Hubbard model, at least in principle.

• Determinant QMC works well for the Hubbard model at half filling and moderate temperatures
• Auxiliary-field QMC can handle larger systems and more complex interactions
• QMC gives unbiased results for thermodynamic quantities, correlation functions, and dynamical properties (via analytic continuation)
• The major obstacle is the fermionic sign problem: away from half filling or for frustrated lattices, statistical errors grow exponentially with system size and inverse temperature, making calculations extremely costly

Current research and challenges

Mott physics remains at the forefront of condensed matter research, with several active directions.

Novel Mott materials

The search for new Mott insulators has expanded beyond 3d oxides to include:

• 5d compounds like iridates ($\text{Sr}_2\text{IrO}_4$), where spin-orbit coupling is comparable to $U$ and produces novel $J_{\text{eff}} = 1/2$ Mott states
• Charge-transfer insulators, where the gap is between ligand p states and metal d states rather than between two Hubbard bands (the Zaanen-Sawatzky-Allen classification distinguishes these from true Mott-Hubbard insulators)
• Engineered heterostructures and oxide superlattices, where interface effects and dimensional confinement can tune correlation strength

Mott physics in low dimensions

Reduced dimensionality enhances the effects of correlations and fluctuations.

• Twisted bilayer graphene near the "magic angle" (~1.1°) develops flat bands that exhibit Mott-like insulating states and superconductivity, bringing Mott physics into the 2D van der Waals materials arena
• One-dimensional Mott insulators (e.g., $\text{Sr}_2\text{CuO}_3$) show spin-charge separation, a hallmark of Luttinger liquid physics
• Theoretical treatment of low-dimensional correlated systems requires methods beyond standard DMFT, such as tensor network approaches (DMRG) and functional renormalization group

Non-equilibrium dynamics

Ultrafast laser experiments can drive Mott insulators out of equilibrium on femtosecond timescales, revealing dynamics that equilibrium probes cannot access.

• Pump-probe spectroscopy can transiently collapse the Mott gap, creating a photo-induced metallic state
• In some cases, pulsed light induces transient superconducting-like signatures in materials that are not superconductors in equilibrium
• Theoretical description of these non-equilibrium states requires time-dependent extensions of DMFT and other many-body methods, which are still under active development
• A key challenge is disentangling electronic relaxation from lattice heating on similar timescales