Mott, Slater, Band, and Charge-Transfer Insulators

Mott, Slater, band, charge-transfer, and Anderson insulators name controlled limiting mechanisms, not synonyms for large resistivity or necessarily distinct thermodynamic phases. A defensible label must say whether charge is gapped or localized, which filling, symmetry, and unit cell matter, which orbitals carry removal and addition weight, and how the state changes when interaction, order, hybridization, or disorder is varied. Real materials and finite-temperature models often interpolate between these anchors.

Required background. Use Hubbard bands and spectral transfer. Helpful background. Anderson localization scaling and Bloch/Wannier band theory develop the corresponding mechanisms.

First establish what is insulating

For a finite system of linear size $L$ with ground-state energy $E_0(N,L)$ at particle number $N$, define the one-electron addition and removal chemical potentials by

$$\mu_+(L)=E_0(N+1,L)-E_0(N,L), \qquad \mu_-(L)=E_0(N,L)-E_0(N-1,L).$$

Their difference is the finite-size one-electron addition–removal gap, or chemical-potential discontinuity,

$$\Delta_{c,1}(L) =E_0(N+1,L)+E_0(N-1,L)-2E_0(N,L) =\mu_+(L)-\mu_-(L).$$

Every finite system has level spacings, so a positive $\Delta_{c,1}(L)$ is not yet an insulating phase. Along a stable thermodynamic sequence at fixed density, define, when the limit exists,

$$\Delta_{c,1} =\lim_{\substack{L\to\infty\\N/L^d=n}} \bigl[\mu_+(L)-\mu_-(L)\bigr].$$

The chemical-potential plateau must retain nonzero width in this limit; if phase separation is possible, use the convex thermodynamic energy. This is the controlling charge gap in an ordinary normal insulator, but not in every phase. A paired fluid can retain gapped single-electron excitations while accepting charge two at arbitrarily low energy. For a stable even-$N$ sector, the corresponding per-particle pair gap is

$$\Delta_{c,2}(L) =\frac{E_0(N+2,L)+E_0(N-2,L)-2E_0(N,L)}{2}.$$

Incompressibility should therefore be established directly from a zero-temperature density plateau, or equivalently from the absence of every lower-energy charge channel. With the convention

$$\kappa=\frac{\partial n}{\partial\mu},$$

where $n$ is the particle density, one has $\kappa=0$ throughout the interior of a clean charge plateau at $T=0$. A paired fluid can instead have $\Delta_{c,1}>0$ from the even–odd effect while $\Delta_{c,2}\to0$ and the compressibility remains finite in the thermodynamic limit. Conversely, $\kappa=0$ at one isolated chemical potential is not enough: a Dirac semimetal can have a vanishing pointwise compressibility without a finite plateau. At finite temperature, activated carriers generally make $\kappa$ and the conductivity small rather than exactly zero.

Several gaps must not be conflated.

Quantity	What it changes	Why it can differ
One-electron charge gap $\Delta_{c,1}$	Total particle number by one	A lower two- or many-particle channel can still make the phase compressible
Single-particle spectral gap	Energy interval between the highest removal support and lowest addition support of the physical positive spectral weight, minimized over the declared momentum and orbital channels	A measured onset can miss intrinsic support because of matrix elements or resolution, and incoherent edges need not look like sharp bands
Optical gap	Creates a neutral excitation coupled to the current	Excitons, selection rules, and indirect momentum can shift the onset
Transport activation scale	Moves charge through a finite-temperature, disordered sample	Mobility edges, in-gap states, contacts, and phonons can set the apparent scale

The noninteracting or effectively single-particle Anderson insulator is the crucial counterexample: localized states can produce a finite density of states and thermodynamic compressibility at the chemical potential while the zero-temperature dc conductance vanishes—a mobility gap without a spectral gap Imada, Fujimori, and Tokura 1998, § II.E.3, pp. 1107–1109. Interactions can invalidate this single-particle relation, so tunneling density of states and thermodynamic compressibility must then be separated. Conversely, a small conductivity can result from strong but finite scattering, an activation crossover, contacts, or finite size without establishing an insulating ground state.

Mechanisms in a diagnostic matrix

Mechanism	Controlled anchor
Band	Filled one-electron bands in the actual symmetry-preserving unit cell
Slater	A weak-coupling magnetic Hartree–Fock potential in the magnetic unit cell
Mott	A commensurate interacting atomic or strong-coupling limit with blocked charge motion
Charge transfer	A correlated metal–ligand parent model in which ligand-hole configurations control the low-energy charge sector
Anderson	Disorder-localized one-particle states, classified in a declared dimension and symmetry class

The compact matrix names the five anchors. The profiles below supply the required evidence and a deliberately adversarial test of each label.

Band

Evidence to seek. Establish a single-particle and thermodynamic charge gap, then exhibit an adiabatic path to a noninteracting filled-band representative that preserves the relevant symmetries, filling, gap, and topological class. This includes ordinary and symmetry-protected band insulators when an appropriate free-fermion representative exists.

Strong counterevidence. No such symmetry-preserving gapped path exists. Intrinsic topological order, fractionalization, and interaction-enabled phases without a free representative require additional diagnostics and do not fit the conventional band anchor.

Slater

Evidence to seek. In the narrow usage adopted here, a magnetic weak-coupling Hartree–Fock potential reconstructs the bands, and the order parameter, folded-band weight, and indirect gap covary under the same tuning. At fixed magnetic order, a Slater state is itself a band insulator in the magnetic cell; “Slater” identifies the origin of its one-body potential Slater 1951, pp. 538–539.

Strong counterevidence. A thermodynamic charge gap and interaction-scale spectral transfer survive a controlled restoration of the magnetic symmetry. Peierls, charge-density-wave, orbital-order, and excitonic gaps are other order-induced band reconstructions, not automatically Slater insulators.

Mott

Evidence to seek. Require incompressibility and an addition–removal gap with atomic or strong-coupling continuity. Suppressed local charge fluctuations and interaction-scale spectral transfer are strong supporting evidence when the model has resolved interaction-separated sectors, but neither is a universal standalone criterion. Moments or multiplet entropy persisting in a symmetry-restored regime are supporting evidence only when the local multiplet permits them; local moments are neither necessary nor sufficient. Mott’s atomic picture assigns finite energy to separated charge configurations despite an incomplete one-electron band Mott 1949, pp. 416–419; in the modern distinction, this anchor does not require conventional symmetry breaking Imada, Fujimori, and Tokura 1998, § I.A, pp. 1040–1042.

Strong counterevidence. The gap vanishes whenever order is removed, remains adiabatically connected to an ordered one-body band problem, and shows neither atomic continuity nor correlation-driven transfer.

Charge transfer

Evidence to seek. In the conventional positive-$\Delta$ charge-transfer limit, resolve ligand-dominated removal weight and correlated-metal addition weight in one consistently defined parent orbital space. In a negative-$\Delta$ ground state, addition can fill a ligand hole and both edges can be strongly ligand-like or mixed; establish the configuration weights and the actual lowest charge processes instead. In either regime, check multiplets, polarization dependence, optical sum rules, and evolution with hybridization and interaction.

Strong counterevidence. For a claimed conventional positive-$\Delta$ charge-transfer state, both charge edges remain metal-$d$-like after projector, hybridization, multiplet, and matrix-element checks and evolve continuously in a controlled $U_d<\Delta$ parent-model regime. A negative-$\Delta$ claim instead fails if the consistently defined ground state and low-energy charge processes show no robust ligand-hole character.

Anderson

Evidence to seek. Demonstrate conductance or localization-length scaling at the chemical potential Abrahams et al. 1979, pp. 673–676; MacKinnon and Kramer 1981, pp. 1546–1549, or finite-size scaling of participation ratios or multifractal observables Evers and Mirlin 2008, § II.C.1, pp. 1362–1363. A single finite-size inverse participation ratio is not enough. Random site-energy disorder can localize exact wavefunctions and suppress diffusion without requiring a spectral gap Anderson 1958, pp. 1492–1495.

Strong counterevidence. Extended-state scaling, or a clean thermodynamic charge gap that accounts for the same regime, defeats a pure Anderson explanation.

Before choosing between a band and a symmetric correlated anchor, count the exactly conserved charge per primitive cell and keep lattice translation symmetry explicit. At filling $\nu=p/q$ in lowest terms, flux insertion implies that a gapped phase has at least $q$ ground states in the thermodynamic limit Oshikawa 2000, pp. 1535–1537. Thus a unique translation-symmetric gapped state at noninteger filling is obstructed; translation breaking, gaplessness, or topological ground-state degeneracy can evade that conclusion. This is a veto on a conventional filled-band explanation, not by itself proof of Mottness.

These mechanisms are controlled anchors, not mutually exclusive bins. Antiferromagnetic order can reconstruct an already correlated Mott or charge-transfer state, and disorder can localize the edge states of an interaction-driven gap. A system can move between anchors by crossover when no symmetry or topological obstruction forces a phase transition. The useful question is which diagnostics survive after each candidate ingredient is selectively removed or varied.

Charge-transfer energy and ligand holes

For a transition-metal ion with a nominal $d^n$ configuration, first declare one correlated parent subspace, screening and double-counting convention, crystal-field basis, and rule for selecting the lowest multiplet or a configuration barycenter. Within that same model, define the local $d$-shell charge scale schematically by

$$U_d =E(d^{n+1})+E(d^{n-1})-2E(d^n),$$

and the ligand-to-metal transfer energy by

$$\Delta =E(d^{n+1}\underline L)-E(d^n),$$

where $\underline L$ denotes a ligand hole. These are model- and basis-dependent many-body configuration energies, so $U_d$ and $\Delta$ may be compared only within the same parent construction. In particular, $\Delta$ is a neutral local transfer energy, not the thermodynamic charge gap, and neither quantity is generally a difference of two bare orbital levels.

The idealized atomic charge processes are

$$2d^n\longrightarrow d^{n-1}+d^{n+1} \quad\text{(Mott–Hubbard)},$$

and

$$d^n\longrightarrow d^{n+1}\underline L \quad\text{(charge transfer)}.$$

Within the insulating strong-coupling regions of the Zaanen–Sawatzky–Allen classification, a lowest charge excitation governed mainly by $U_d<\Delta$ is Mott–Hubbard-like, whereas $\Delta<U_d$ places the lowest removal state primarily on ligands and gives a charge-transfer insulator Zaanen, Sawatzky, and Allen 1985, pp. 419–420 and Figs. 1–3. These names distinguish two correlated-insulator limits; in broader usage, a charge-transfer insulator is often also called a strongly correlated or Mott insulator.

The physical gap is not simply $\min(U_d,\Delta)$. Ligand and $d$ bandwidths, metal–ligand hybridization, crystal fields, multiplets, screening, excitons, and—when relevant—ligand repulsion $U_p$ and metal–ligand interaction $V_{pd}$ move and mix the edges Imada, Fujimori, and Tokura 1998, § II.B.1, pp. 1052–1053; §§ III.A.3–5, pp. 1126–1131; §§ III.B.1–2, pp. 1131–1134. Hybridization turns the idealized $U_d=\Delta$ boundary into a crossover, so orbital character must be established rather than inferred from formal valence.

For $\Delta<0$, the ground state is generally a covalent superposition

$$\lvert\Psi_0\rangle =\sum_{m\ge0}a_m \lvert d^{n+m}\underline L^{\,m}\rangle$$

over the shell occupancies allowed by the parent model, and low-energy removal weight can be predominantly ligand-like Imada, Fujimori, and Tokura 1998, § III.A.5, pp. 1130–1131. These ligand holes are not automatically mobile carriers: negative $\Delta$ can support a metal or, with bond disproportionation or another ordering mechanism, an insulator. Nickelate x-ray spectroscopy together with cluster and impurity calculations provides strong evidence for a ground state rich in oxygen-$2p$ holes Bisogni et al. 2016, abstract, Results, Figs. 3–6, and Discussion.

Why a one-band model need not erase charge-transfer physics

Two reductions should be distinguished. A one-particle Wannier reduction can retain an antibonding orbital with both metal and ligand weight. The Zhang–Rice construction is instead a many-body cell reduction: a ligand hole binds with the correlated-site spin to form a low-energy singlet, which is then mapped to one effective degree of freedom per site Zhang and Rice 1988, pp. 3759–3761. It is not merely a literal one-electron Wannier orbital.

More generally, let $P$ project onto the retained subspace and $Q=1-P$ onto the eliminated sector. Choose a block-diagonalizing anti-Hermitian generator $S^\dagger=-S$ such that

$$Qe^SHe^{-S}P=0.$$

A controlled static projection requires an isolated low-energy manifold, schematically

$$\lambda \sim\frac{t_{\mathrm{mix}}}{\Delta_{PQ}} \ll1,$$

where $t_{\mathrm{mix}}$ is the largest local or connected matrix element that couples relevant target states in $P$ to $Q$, and $\Delta_{PQ}$ is their separation over the window of interest. Then the Hamiltonian and every physical observable must be transformed together Delannoy et al. 2009, § III.B.1, p. 235130-11, Eqs. (43) and (47):

$$H_{\mathrm{eff}}=Pe^SHe^{-S}P, \qquad O_{\mathrm{eff}}=Pe^SOe^{-S}P,$$

and therefore

$$O_{\mathrm{eff}} =P\left(O+[S,O]+\frac12[S,[S,O]]+\cdots\right)P.$$

Applying this general relation to a metal–ligand reduction, the effective electron, density, current, or dipole operator can contain a mixture of local orbitals, neighboring sites, and multi-particle terms. Even when $POP$ vanishes or looks purely local, the commutator terms can return virtual ligand admixtures to the retained subspace. An optical calculation must transform the full response operator consistently rather than matching only the Hamiltonian spectrum.

Therefore reproducing only the low-energy eigenvalues with a one-band Hamiltonian does not falsify charge-transfer character. Within the retained window, a stronger test must also reproduce orbital-resolved addition and removal edges and polarization-dependent matrix elements with matched operators. A static low-energy $P$-only Hamiltonian cannot represent final states above its retained cutoff. An exact energy-dependent projected resolvent can encode $Q$-derived poles that have $P$-overlap, but ligand-resolved matrix elements still require matched reconstruction or upfolding. Spectral transfer at or above the static cutoff and ligand-sensitive core or x-ray response therefore require that reconstruction, documented source/background terms, or the larger parent subspace. Any quoted “orbital weight” must also state the parent subspace, Wannier or projector window, allowed gauge variation, and experimental matrix elements.

Small or negative charge-transfer energy can destroy the separation $\Delta_{PQ}$ required by a static one-band projection. When $\lambda$ is not small, enlarge the retained orbital space or use an explicitly energy- or frequency-dependent downfolding rather than treating the one-band Hamiltonian as controlled.

This is why “number of bands in the solver” and “microscopic character of the gap” answer different questions. Downfolding compresses degrees of freedom; it does not retroactively change the physical origin of the states represented.

Distinguishing Mott and Slater components

Use the actual crystallographic or spontaneously enlarged unit cell. A nonmagnetic calculation in an artificially small cell can make an ordinary band reconstruction look interaction driven, while allowing an ordered cell can make a correlation-assisted gap look purely Slater.

For a collinear, single-$\mathbf Q$, period-two antiferromagnet with $2\mathbf Q=\mathbf G$ a reciprocal vector, use the reduced-zone basis $(c_{\mathbf k\sigma},c_{\mathbf k+\mathbf Q,\sigma})$ and write $\xi_{\mathbf k}=\epsilon_{\mathbf k}-\mu$. The minimal Hermitian Slater block is

$$H_{\mathbf k\sigma}= \begin{pmatrix} \xi_{\mathbf k} & \sigma\Delta_{\mathrm{AF}}\\ \sigma\Delta_{\mathrm{AF}}^* & \xi_{\mathbf k+\mathbf Q} \end{pmatrix},$$

with reconstructed bands

$$E_{\mathbf k,\pm} =\frac{\xi_{\mathbf k}+\xi_{\mathbf k+\mathbf Q}}{2} \pm \sqrt{ \left(\frac{\xi_{\mathbf k}-\xi_{\mathbf k+\mathbf Q}}{2}\right)^2 +\lvert\Delta_{\mathrm{AF}}\rvert^2 }.$$

For a general commensurate period or incommensurate order, more than the two momenta $\mathbf k$ and $\mathbf k+\mathbf Q$ must be retained. Even in the two-band block, the local avoided crossing is not the insulating criterion. At the relevant integer filling the indirect gap must obey

$$\Delta_{\mathrm{ind}} =\min_{\mathbf k}E_{\mathbf k,+} -\max_{\mathbf k}E_{\mathbf k,-}>0.$$

At half filling with a perfectly nested Fermi surface, $\xi_{\mathbf k+\mathbf Q}=-\xi_{\mathbf k}$, this becomes $2\lvert\Delta_{\mathrm{AF}}\rvert$. With imperfect nesting, a nonzero ordered moment may instead leave electron or hole pockets. Conversely, a gap observed above the ordering temperature rules out only a purely static long-range-order Hartree–Fock account: short-range order and a mixed correlation–order self-energy may survive after the expectation value vanishes. A controlled next test is a stable symmetry-restored solution or an explicitly constrained deformation that tracks the same observables and the remaining short-range correlations.

The repulsive one-site Hubbard limit supplies the complementary local charge-blockade anchor. For $U>0$ at one electron per site, $E_0(0)=E_0(1)=0$ and $E_0(2)=U$, so $\Delta_{c,1}=U$ without magnetic order. An isolated site has no transport problem, and finite hopping broadens and shifts the addition and removal sectors, so the lattice charge gap is not generally equal to $U$.

A useful protocol varies the ordering field and interaction separately:

1. Track the order parameter, folded-band weight, and gap through the ordering temperature or a tuning parameter that suppresses order.
2. Test whether an incompressible paramagnetic state survives after long-range order is removed without forcing the interaction to zero.
3. Follow integrated spectral weight over an energy range broad enough to include Hubbard or charge-transfer sectors, not just the gap edge.
4. Compare with an ordered weak-coupling reference and an atomic or strong-coupling reference using the same observables.

If the gap follows the ordered moment to zero and the spectra remain adiabatically connected to an ordered band problem, “Slater” is well supported. If the charge gap and correlation-driven redistribution survive removal of long-range magnetic order, a purely static Slater account is excluded and a paramagnetic correlation contribution is supported. Apply the band-path, orbital-edge, nonmagnetic-order, and topology tests before choosing Mott–Hubbard or charge-transfer language. Many antiferromagnetic insulators interpolate between these limits; reporting a bounded mixed regime can be more accurate than forcing one word.

Observable decision protocol

The chapter’s reduction map separates the Hamiltonian and matched operators from the mechanism claim. Its validity table gives the stop conditions for the final attribution.

1. Is the bulk state incompressible? Minimum evidence. A chemical-potential plateau of nonzero thermodynamic width and the lowest relevant one- or multi-particle charge threshold, with size and temperature extrapolation. Failure mode. A finite-size level spacing, a parity gap in a paired compressible phase, a pointwise Dirac zero, or an activated crossover.
2. Is there a one-particle or optical gap? Minimum evidence. Momentum-, orbital-, polarization-, and resolution-aware spectra plus the relevant sum rules. Failure mode. Matrix-element suppression, analytic-continuation broadening, or an exciton mistaken for the thermodynamic charge gap.
3. Does symmetry breaking create the gap? Minimum evidence. Order, folding, and indirect gap tracked under the same tuning, including a controlled symmetry-restored regime. Failure mode. Comparing unrelated samples or temperatures, or equating loss of long-range order with removal of short-range correlations.
4. Is the state interaction driven? Minimum evidence. Atomic or strong-coupling continuity, suppressed charge fluctuations, self-energy or spectral transfer, and interaction variation; moments are conditional supporting evidence. Failure mode. Using large resistivity, a local moment, or a two-peak spectrum alone.
5. Is it charge transfer? Minimum evidence. A many-body $\Delta/U_d$ interpretation inside one declared parent model and matched ligand/metal edge observables. Failure mode. Treating a formal oxidation state, bare level difference, or one-band fit as decisive.
6. Is it localized? Minimum evidence. Conductance or localization-length scaling, transfer-matrix scaling, or finite-size scaling of participation or multifractal observables in the appropriate dimension and symmetry class. Failure mode. Inferring localization from a density of states, one finite-size participation ratio, or a finite dephasing length.

Compressibility, charge thresholds, and transport diagnose distinct parts of a metal–insulator problem Imada, Fujimori, and Tokura 1998, § II.E.3, pp. 1107–1109. Combine them rather than selecting the observable that favors a preferred label. Finite temperature broadens thresholds, disorder creates tails, and analytic continuation correlates spectral uncertainties. The strongest justified conclusion may be “insulating with unresolved Mott–Slater balance” or “charge-transfer-dominated Anderson–Mott regime.”

Common pitfalls

Equating large resistivity with a charge gap. Transport mixes carrier density and mobility. Pair it with compressibility, addition energies, or spectroscopy and take the low-temperature limit.

Using an order onset as sole proof of Slater physics. Order can reconstruct a pre-existing correlated gap. Track the symmetry-restored state and broad spectral weight.

Calling every interaction-split spectrum or local moment Mott. Broken-symmetry bands, Hund physics, charge-transfer edges, disorder, and matrix elements can also split spectra or produce moments. Require incompressibility and controlled continuity to an interaction-blocked limit; seek broad spectral transfer when separated interaction sectors make it applicable.

Treating a downfolded basis as literal chemistry. Effective orbitals inherit ligand weight and transformed probes. Compare matched physical operators, not just Hamiltonian eigenvalues.

Exercises

Atomic Mott anchor. For one isolated repulsive Hubbard site with $H=Un_\uparrow n_\downarrow$ and $U>0$, compute the charge gap at one electron. What does this calculation not establish?

Solution

The ground-state energies are $E_0(0)=0$, $E_0(1)=0$, and $E_0(2)=U$. Hence

$$\Delta_{c,1}=E_0(2)+E_0(0)-2E_0(1)=U.$$

Hopping broadens the addition and removal sectors into bands, so the lattice gap is not generally equal to $U$, but the atomic result supplies the Mott strong-coupling anchor. An isolated site has no bulk transport or thermodynamic limit, so this calculation alone does not establish a lattice insulating phase.

Slater direct versus indirect gap. Diagonalize the period-two Slater matrix and specialize to half filling with perfect nesting, $\xi_{\mathbf k+\mathbf Q}=-\xi_{\mathbf k}$. Does a nonzero $\Delta_{\mathrm{AF}}$ always make an imperfectly nested system insulating?

Solution

The eigenvalues of a Hermitian $2\times2$ matrix give

$$E_{\mathbf k,\pm} =\frac{\xi_{\mathbf k}+\xi_{\mathbf k+\mathbf Q}}{2} \pm \sqrt{ \left(\frac{\xi_{\mathbf k}-\xi_{\mathbf k+\mathbf Q}}{2}\right)^2 +\lvert\Delta_{\mathrm{AF}}\rvert^2 }.$$

At perfect nesting they reduce to $\pm\sqrt{\xi_{\mathbf k}^2+\lvert\Delta_{\mathrm{AF}}\rvert^2}$, and filling the lower band gives direct and indirect gaps $2\lvert\Delta_{\mathrm{AF}}\rvert$. Without nesting, the band centers need not coincide. Insulation requires

$$\min_{\mathbf k}E_{\mathbf k,+} -\max_{\mathbf k}E_{\mathbf k,-}>0;$$

otherwise reconstructed electron or hole pockets remain even when the order parameter is nonzero.

Order melts but the gap survives. An antiferromagnet’s gap remains nearly unchanged above the ordering temperature, while folded-band coherence disappears and broad spectral weight continues to move over an interaction scale. What can be concluded?

Solution

The loss of folding shows that long-range order contributed a Slater reconstruction below the transition. Persistence of the gap and interaction-scale transfer rules out a purely static long-range-order Hartree–Fock account and supports a correlation-driven paramagnetic gap. It does not by itself distinguish a Mott–Hubbard or charge-transfer gap from a pseudogap sustained by short-range order. That requires a controlled symmetry-restored deformation or constrained calculation together with thermodynamic charge and orbital-edge diagnostics.

Controlled downfolding. State the projector and small-parameter conditions for a static one-band reduction, then expand a transformed observable through second order. Why can a one-band fit to the low-energy dispersion still fail for a charge-transfer material’s spectroscopy?

Solution

For complementary projectors $P+Q=1$, choose $S^\dagger=-S$ so that $Qe^SHe^{-S}P=0$. The projection is perturbatively controlled only when the retained manifold is isolated, with $\lambda\sim t_{\mathrm{mix}}/\Delta_{PQ}\ll1$ for the relevant local or connected mixing matrix element. The operator is

$$O_{\mathrm{eff}} =P\left(O+[S,O]+\frac12[S,[S,O]]+\cdots\right)P.$$

The retained Wannier orbital can be an antibonding mixture of metal and ligand states, while the commutator terms preserve virtual high-energy admixtures. They can make a low-energy response nonzero even if $POP=0$. The one-band model must reproduce matched electron, current, dipole, and orbital-resolved observables—not only its eigenvalues—before the downfolded theory is accepted as spectroscopically matched. Success validates its representation of ligand-hole character; it does not convert the parent material into a metal-$d$-only system. If $\Delta_{PQ}$ collapses, the static one-band projection is not controlled.

Mobility gap without a spectral gap. A disordered system has a finite single-particle density of states at the chemical potential but zero extrapolated dc conductivity. Which mechanism is immediately compatible, and what remains to test?

Solution

Anderson localization is compatible because localized states can contribute spectral density without carrying dc current. One must still establish conductance, localization-length, or participation scaling; rule out finite-size and inelastic-dephasing effects; declare the dimension and symmetry class; and test interaction dependence. The observation alone does not exclude an Anderson–Mott regime or interaction-induced Coulomb suppression of the density of states.

ZSA and negative charge transfer. In one consistently defined insulating parent model, suppose $U_d=8\,\mathrm{eV}$ and $\Delta=3\,\mathrm{eV}$, the lowest removal edge is ligand dominated, and the lowest addition edge is metal-$d$ dominated. Which controlled anchor is supported? If tuning makes $\Delta<0$, does that prove an insulator?

Solution

Because $\Delta<U_d$ and the resolved edge characters agree, the data support the charge-transfer anchor. Hybridization and multiplets still prevent one from identifying the physical gap with $3\,\mathrm{eV}$. When $\Delta<0$, ligand-hole configurations enter the ground-state mixture, but the system may be metallic or may require bond disproportionation or another mechanism to become insulating. Negative transfer energy alone does not decide transport or incompressibility.

Pair gap versus incompressibility. Along a sequence of $d$-dimensional systems, suppose $\Delta_{c,1}(L)\to\Delta_{\mathrm{odd}}>0$, while $\Delta_{c,2}(L)\propto L^{-d}$ and $n(\mu)$ has no plateau of nonzero limiting width. Is this a charge insulator?

Solution

No. The finite one-electron threshold can be an even–odd parity effect, while the vanishing pair threshold and smooth density response show that charge two enters at arbitrarily low energy. This is compatible with a compressible paired fluid with gapped single-electron excitations, not with a thermodynamic charge gap.

Adiabatic band test. A correlated, symmetry-preserving insulator remains gapped while its interactions are continuously reduced to zero, and the endpoint is a filled-band state in the same unit cell and topological class. What does this establish, and what does it not establish?

Solution

The open same-symmetry path supports the band anchor and argues against a required Mott obstruction. It does not say that correlations are negligible at the physical endpoint: they may strongly renormalize the gap, spectral weight, or response while remaining adiabatically connected to the band representative.

Continue the chapter

• Use DMFT impurity mapping and self-consistency to test whether a paramagnetic correlation-driven gap survives in a controlled local approximation.
• Use multiorbital correlations and Hund’s metals when orbital selectivity, local multiplets, or $J=0$ states make a one-band or local-moment diagnostic inadequate.
• Use Anderson localization scaling when mobility rather than a thermodynamic charge gap is the unresolved mechanism.

References

• Elihu Abrahams, P. W. Anderson, Donald C. Licciardello, and T. V. Ramakrishnan, “Scaling Theory of Localization: Absence of Quantum Diffusion in Two Dimensions,” Physical Review Letters 42 (1979) 673–676, doi:10.1103/PhysRevLett.42.673.
• P. W. Anderson, “Absence of Diffusion in Certain Random Lattices,” Physical Review 109 (1958) 1492–1505, doi:10.1103/PhysRev.109.1492.
• Valentina Bisogni, Sara Catalano, Robert J. Green, Marta Gibert, Raoul Scherwitzl, Yaobo Huang, Vladimir N. Strocov, Pavlo Zubko, Shadi Balandeh, Jean-Marc Triscone, George Sawatzky, and Thorsten Schmitt, “Ground-State Oxygen Holes and the Metal–Insulator Transition in the Negative-Charge-Transfer Rare-Earth Nickelates,” Nature Communications 7 (2016) 13017, doi:10.1038/ncomms13017.
• J.-Y. P. Delannoy, M. J. P. Gingras, P. C. W. Holdsworth, and A.-M. S. Tremblay, “Low-Energy Theory of the $t$–$t'$–$t''$–$U$ Hubbard Model at Half-Filling: Interaction Strengths in Cuprate Superconductors and an Effective Spin-Only Description of La$_2$CuO$_4$,” Physical Review B 79 (2009) 235130, doi:10.1103/PhysRevB.79.235130.
• Ferdinand Evers and Alexander D. Mirlin, “Anderson Transitions,” Reviews of Modern Physics 80 (2008) 1355–1417, doi:10.1103/RevModPhys.80.1355.
• Masatoshi Imada, Atsushi Fujimori, and Yoshinori Tokura, “Metal–Insulator Transitions,” Reviews of Modern Physics 70 (1998) 1039–1263, doi:10.1103/RevModPhys.70.1039.
• Angus MacKinnon and Bernhard Kramer, “One-Parameter Scaling of Localization Length and Conductance in Disordered Systems,” Physical Review Letters 47 (1981) 1546–1549, doi:10.1103/PhysRevLett.47.1546.
• N. F. Mott, “The Basis of the Electron Theory of Metals, with Special Reference to the Transition Metals,” Proceedings of the Physical Society. Section A 62 (1949) 416–422, doi:10.1088/0370-1298/62/7/303.
• Masaki Oshikawa, “Commensurability, Excitation Gap, and Topology in Quantum Many-Particle Systems on a Periodic Lattice,” Physical Review Letters 84 (2000) 1535–1538, doi:10.1103/PhysRevLett.84.1535.
• J. C. Slater, “Magnetic Effects and the Hartree–Fock Equation,” Physical Review 82 (1951) 538–541, doi:10.1103/PhysRev.82.538.
• Jan Zaanen, George A. Sawatzky, and John W. Allen, “Band Gaps and Electronic Structure of Transition-Metal Compounds,” Physical Review Letters 55 (1985) 418–421, doi:10.1103/PhysRevLett.55.418.
• Fu-Chun Zhang and T. M. Rice, “Effective Hamiltonian for the Superconducting Cu Oxides,” Physical Review B 37 (1988) 3759–3761, doi:10.1103/PhysRevB.37.3759.