Anderson Orthogonality Catastrophe

• Anderson Orthogonality Catastrophe is a phenomenon wherein a local impurity in a Fermi system triggers extensive low-energy particle-hole excitations, causing the overlap between the original and perturbed ground states to vanish in the thermodynamic limit.
• It is characterized by power-law decay of observables and threshold singularities, with the decay exponents intimately linked to scattering phase shifts at the Fermi energy.
• Extensions of the AOC framework address dynamical quenches, Coulomb and disordered systems, and even topological and open-system settings, impacting phenomena like the Kondo effect and quantum transport.

Anderson orthogonality catastrophe (AOC) is the statement that a many-fermion ground state can change so drastically under a local perturbation that its overlap with the unperturbed ground state vanishes in the thermodynamic limit. In the canonical setting, a local impurity potential is suddenly switched on at $t=0$, and the physically important quantities are the many-body overlap or fidelity, such as $F=|\langle \Psi_0|\Psi_0'\rangle|$, the survival amplitude $$\chi(t)=\langle \Psi|e^{i\mathcal H_f t}e^{-i\mathcal H_i t}|\Psi\rangle$$, or the impurity Green’s function $G(\tau)$. The standard mechanism is the creation of arbitrarily many low-energy particle-hole excitations near the Fermi surface, which converts finite overlap into asymptotic orthogonality and generates threshold singularities in dynamical response (Tupitsyn et al., 2019, Fogarty et al., 2019).

1. Canonical formulation and scattering-theoretic exponents

In the textbook Nozières–de Dominicis / Anderson solution for a non-interacting Fermi gas with a short-range impurity potential $V_S$ of finite scattering cross-section, the long-time asymptotics is

$$G(\tau)=Z(\tau)e^{-E\tau}, \qquad Z(\tau)\propto \tau^{-\gamma},$$

so that the overlap decays as a power law, $I(\tau)\propto \tau^{-\gamma}$. The exponent is determined by scattering phase shifts at the Fermi energy,

$$\gamma = 2\sum_{\ell}(2\ell+1)\left(\frac{\delta_\ell}{\pi}\right)^2.$$

In this formulation, orthogonality is controlled entirely by Fermi-surface scattering data, and the catastrophe is the cumulative effect of infinitely many low-energy particle-hole pairs (Tupitsyn et al., 2019).

A rigorous scattering-theoretic version for ideal Fermi gases replaces the heuristic phase-shift formula by an operator expression involving the transition matrix $T_E$ at the Fermi energy: $$\gamma(E)=\frac{1}{\pi^2}\,\bigl\|\arcsin |T_E/2|\bigr\|_{\mathrm{HS}}^2.$$ In the thermodynamic limit, the many-body overlap $F=|\langle \Psi_0|\Psi_0'\rangle|$0 obeys a power-law upper bound of the form

$F=|\langle \Psi_0|\Psi_0'\rangle|$1

along a subsequence, and in $F=|\langle \Psi_0|\Psi_0'\rangle|$2, finite-rank, and lattice settings this strengthening does not require passage to a subsequence (Gebert et al., 2014).

For one-dimensional non-relativistic spinless Fermi systems, the Anderson integral grows logarithmically,

$F=|\langle \Psi_0|\Psi_0'\rangle|$3

and the exponent can be written directly in terms of the transmission amplitude of the one-dimensional scattering matrix: $F=|\langle \Psi_0|\Psi_0'\rangle|$4 This leads to two-sided power-law bounds on the overlap $F=|\langle \Psi_0|\Psi_0'\rangle|$5, with the upper-bound exponent fixed by scattering data at the Fermi energy (Küttler et al., 2013).

These results delimit the canonical AOC regime: non-interacting fermions, local perturbations with finite cross-section, and a gapless Fermi surface. Later work broadens each of these assumptions rather than discarding the central statement that a local perturbation can reorganize an extensive many-body state.

2. Dynamical orthogonality, quench correlators, and threshold singularities

AOC is not only a static statement about ground-state overlaps. In sudden-quench form, the dynamical overlap

$F=|\langle \Psi_0|\Psi_0'\rangle|$6

measures the collapse of the initial state under post-quench evolution. A general criterion links this collapse to the quantum speed limit. Using the Bures angle

$F=|\langle \Psi_0|\Psi_0'\rangle|$7

the bound

$F=|\langle \Psi_0|\Psi_0'\rangle|$8

shows that any quenched many-body system whose post-quench energy variance scales as $F=|\langle \Psi_0|\Psi_0'\rangle|$9 exhibits orthogonality catastrophe in the dynamical sense, because the minimal time to reach an orthogonal state vanishes as $$\chi(t)=\langle \Psi|e^{i\mathcal H_f t}e^{-i\mathcal H_i t}|\Psi\rangle$$0 (Fogarty et al., 2019).

In impurity models, this dynamical viewpoint can be organized into “type 1” and “type 2” local quenches. For quench operators $$\chi(t)=\langle \Psi|e^{i\mathcal H_f t}e^{-i\mathcal H_i t}|\Psi\rangle$$1 that connect initial and final states differing by displaced charge $$\chi(t)=\langle \Psi|e^{i\mathcal H_f t}e^{-i\mathcal H_i t}|\Psi\rangle$$2, the quench correlator has long-time behavior

$$\chi(t)=\langle \Psi|e^{i\mathcal H_f t}e^{-i\mathcal H_i t}|\Psi\rangle$$3

and the spectral function obeys the generalized Hopfield rule

$$\chi(t)=\langle \Psi|e^{i\mathcal H_f t}e^{-i\mathcal H_i t}|\Psi\rangle$$4

For a composite local quench $$\chi(t)=\langle \Psi|e^{i\mathcal H_f t}e^{-i\mathcal H_i t}|\Psi\rangle$$5, the exponent becomes

$$\chi(t)=\langle \Psi|e^{i\mathcal H_f t}e^{-i\mathcal H_i t}|\Psi\rangle$$6

with $$\chi(t)=\langle \Psi|e^{i\mathcal H_f t}e^{-i\mathcal H_i t}|\Psi\rangle$$7 the additional displaced charge generated by $$\chi(t)=\langle \Psi|e^{i\mathcal H_f t}e^{-i\mathcal H_i t}|\Psi\rangle$$8. This extends X-ray-edge logic to interacting quantum impurity systems and explains why AO governs low-frequency spectral singularities, intermediate-frequency crossovers, and even sensor-induced quantum phase transitions in population-switching devices (Münder et al., 2011).

The same structure underlies the Fermi edge singularity (FES). In the optical X-ray-edge setting, a suddenly created core hole acts as a local perturbation, and the many-body response acquires a threshold singularity because the long-time overlap decays algebraically. In this sense, FES is a dynamical manifestation of AOC rather than a separate phenomenon.

3. Coulomb systems, dynamic screening, and mobile impurities

The standard Anderson solution assumes short-range impurity potentials with finite scattering cross-section and non-interacting fermions. For charged impurities in metals, both assumptions can fail. The bare Coulomb potential

$$\chi(t)=\langle \Psi|e^{i\mathcal H_f t}e^{-i\mathcal H_i t}|\Psi\rangle$$9

has infinite cross-section, so a naive insertion into the standard formula produces a divergent momentum integral, described as an orthogonality “disaster.” Moreover, electrons in a metal cannot be treated as a passive Fermi sea; their dielectric response must be included self-consistently. A static Yukawa replacement,

$G(\tau)$0

is formally inconsistent in the dynamical OC problem because it double-counts screening processes (Tupitsyn et al., 2019).

A consistent formulation uses the dynamically screened kernel

$G(\tau)$1

with the impurity Green’s function written as

$G(\tau)$2

In a plasmon-pole approximation, the interaction is gapped by $G(\tau)$3, the low-energy singularity is removed, and the OC is eliminated: $G(\tau)$4 In the full random-phase approximation, gapless electron-hole excitations are restored, the canonical power law returns, and the exponent is strongly renormalized. At $G(\tau)$5, the dynamically screened Coulomb exponent is more than three times larger than the exponent obtained from the static-Yukawa/non-interacting-Fermi-gas scheme, while for short-range impurity potentials the non-interacting Fermi-sea approximation overestimates the exponent by more than a factor of three (Tupitsyn et al., 2019).

Finite impurity mass changes the infrared structure again. For a mobile impurity with bare propagator

$G(\tau)$6

recoil cuts off the logarithmic divergence responsible for strict orthogonality. Diagrammatic Monte Carlo shows that a localized or very heavy impurity retains a power-law regime, whereas a finite-mass impurity exhibits a truncated decay and a small, finite residue $G(\tau)$7 (Tupitsyn et al., 2019).

An exact interacting realization is provided by the one-dimensional attractive Fermi polaron. In the Yang–Gaudin model, the quasiparticle residue obeys

$G(\tau)$8

with $G(\tau)$9 the Bethe-ansatz phase shift at the Fermi edge. The presence of a two-body bound state does not alter the universal exponent, which remains phase-shift controlled (Orso, 14 Apr 2026). This suggests that the Anderson exponent survives beyond static-potential language when the impurity problem remains integrable and the relevant scattering phase shift is still well defined.

4. Localized, disordered, and critical regimes

In localized systems, the catastrophe becomes statistical. For non-interacting fermions in a one-dimensional disordered lattice with a local impurity switched on adiabatically, the overlap

$V_S$0

varies sharply with impurity position and realization. The orthogonality-event probability

$V_S$1

grows with impurity strength and saturates at about $V_S$2 at half filling. In the Anderson insulator, $V_S$3 grows monotonically. In the Aubry–André insulator, it exhibits plateaux produced by correlated quasi-periodic structure, local resonances, and avoided crossings, which can be captured by a two-site effective Hamiltonian (Cosco et al., 2018, Cosco, 2018).

A broader scaling picture identifies a statistical orthogonality catastrophe in single-particle localized and many-body localized systems: $V_S$4 The mechanism is non-local adiabatic charge transfer: a local quench can relocate a particle to an arbitrarily distant site, and the overlap between localized orbitals then decays exponentially with distance. This is stronger than the metallic power law $V_S$5 and persists in the many-body localized phase (Deng et al., 2015).

At the Anderson metal-insulator transition, multifractality changes the catastrophe once more. For a short-range impurity added to a disordered Fermi system at criticality, the typical fidelity behaves as

$V_S$6

because the average Anderson integral scales as $V_S$7. On the metallic side, the usual Anderson power law is recovered,

$V_S$8

while on the insulating side the typical fidelity saturates for $V_S$9. A crucial subtlety is that $$G(\tau)=Z(\tau)e^{-E\tau}, \qquad Z(\tau)\propto \tau^{-\gamma},$$0 is log-normally distributed at criticality, so the mean fidelity satisfies $$G(\tau)=Z(\tau)e^{-E\tau}, \qquad Z(\tau)\propto \tau^{-\gamma},$$1 even though the typical fidelity tends to zero exponentially fast (Kettemann, 2016). This directly counters the common misconception that “the overlap” has a unique asymptotic meaning independent of averaging procedure.

Random singlet quantum critical points provide a different insulating but critical regime. For a local cut in a random singlet chain, the overlap

$$G(\tau)=Z(\tau)e^{-E\tau}, \qquad Z(\tau)\propto \tau^{-\gamma},$$2

has a typical power-law decay

$$G(\tau)=Z(\tau)e^{-E\tau}, \qquad Z(\tau)\propto \tau^{-\gamma},$$3

while moments scale as

$$G(\tau)=Z(\tau)e^{-E\tau}, \qquad Z(\tau)\propto \tau^{-\gamma},$$4

with a multifractal spectrum satisfying $$G(\tau)=Z(\tau)e^{-E\tau}, \qquad Z(\tau)\propto \tau^{-\gamma},$$5. Here the catastrophe is algebraic, but its disorder statistics are nontrivial and rare-event dominated (Vasseur et al., 2015).

5. Extensions beyond the canonical fermionic closed-system setting

The Sutherland model shows that orthogonality catastrophe is not confined to the usual free-fermion impurity problem. For the $$G(\tau)=Z(\tau)e^{-E\tau}, \qquad Z(\tau)\propto \tau^{-\gamma},$$6-particle model with harmonic confinement and inverse-square pair interactions,

$$G(\tau)=Z(\tau)e^{-E\tau}, \qquad Z(\tau)\propto \tau^{-\gamma},$$7

the quasiparticles obey fractional exclusion statistics through the generalized occupancy rule

$$G(\tau)=Z(\tau)e^{-E\tau}, \qquad Z(\tau)\propto \tau^{-\gamma},$$8

A quench in the inverse-square coupling produces vanishing overlap in the thermodynamic limit, but for an infinitesimal perturbation $$G(\tau)=Z(\tau)e^{-E\tau}, \qquad Z(\tau)\propto \tau^{-\gamma},$$9 the overlap behaves as

$I(\tau)\propto \tau^{-\gamma}$0

which is qualitatively different from Anderson’s power law. The decay is superextensive, $I(\tau)\propto \tau^{-\gamma}$1, and indicates that OC may extend to systems with more general statistics than the fermionic type (Ares et al., 2017).

A dissipative analogue arises in open many-body systems subject to a local Lindblad channel. For a pure initial state, the fidelity

$I(\tau)\propto \tau^{-\gamma}$2

obeys the long-time scaling

$I(\tau)\propto \tau^{-\gamma}$3

when the host supports long-range correlations or gapless modes. In the critical one-dimensional transverse-field Ising chain with local dephasing $I(\tau)\propto \tau^{-\gamma}$4,

$I(\tau)\propto \tau^{-\gamma}$5

Here $I(\tau)\propto \tau^{-\gamma}$6, so the algebraic factor slows decoherence rather than accelerating decay. Criticality is sufficient but not necessary; the XX chain and the two-dimensional Bose gas deep in the superfluid phase show analogous dissipative OC behavior (Tonielli et al., 2018).

In $I(\tau)\propto \tau^{-\gamma}$7-dimensional topological systems at fixed points, the interesting quantity is the finite-size scaling of the overlap rather than vanishing overlap itself. For bosonic and fermionic symmetry-protected topological states,

$I(\tau)\propto \tau^{-\gamma}$8

so the overlap carries a universal topological response term set by the Euler characteristic $I(\tau)\propto \tau^{-\gamma}$9 and the edge-CFT central charge $$\gamma = 2\sum_{\ell}(2\ell+1)\left(\frac{\delta_\ell}{\pi}\right)^2.$$0. For Laughlin states, the scaling is stronger: $$\gamma = 2\sum_{\ell}(2\ell+1)\left(\frac{\delta_\ell}{\pi}\right)^2.$$1 which is beyond conventional AOC (Gu, 2019).

These extensions show that the defining feature of AOC is not tied to a single decay law. Depending on statistics, topology, openness, and spectral structure, the overlap may decay as a power law, exponentially, or faster than exponentially.

6. Physical realizations and observable consequences

The phenomenon has important consequences for numerous dynamic phenomena in fermionic systems, including the Kondo effect, X-ray spectroscopy, and quantum diffusion of impurities (Tupitsyn et al., 2019). In the FAIR treatment of the Friedel–Anderson impurity, the multi-electron scalar product (MESP) between occupied spin-up and spin-down bases decays exponentially in the enforced magnetic state,

$$\gamma = 2\sum_{\ell}(2\ell+1)\left(\frac{\delta_\ell}{\pi}\right)^2.$$2

but saturates at $$\gamma = 2\sum_{\ell}(2\ell+1)\left(\frac{\delta_\ell}{\pi}\right)^2.$$3 in the singlet state. Within the Kondo window, corresponding spin-up and spin-down one-particle states satisfy

$$\gamma = 2\sum_{\ell}(2\ell+1)\left(\frac{\delta_\ell}{\pi}\right)^2.$$4

and the crossover energy defines a universal Kondo scale via the evasion of orthogonality catastrophe (Bergmann, 2010).

Ultracold atoms provide time-domain access to the same physics. In a trapped one-dimensional Fermi gas locally quenched by a two-level impurity, the vacuum persistence amplitude

$$\gamma = 2\sum_{\ell}(2\ell+1)\left(\frac{\delta_\ell}{\pi}\right)^2.$$5

acts simultaneously as the decoherence factor for the impurity and as the generator of the excitation spectrum. At $$\gamma = 2\sum_{\ell}(2\ell+1)\left(\frac{\delta_\ell}{\pi}\right)^2.$$6,

$$\gamma = 2\sum_{\ell}(2\ell+1)\left(\frac{\delta_\ell}{\pi}\right)^2.$$7

which reduces in the $$\gamma = 2\sum_{\ell}(2\ell+1)\left(\frac{\delta_\ell}{\pi}\right)^2.$$8 limit to the Nozières–De Dominicis form

$$\gamma = 2\sum_{\ell}(2\ell+1)\left(\frac{\delta_\ell}{\pi}\right)^2.$$9

The same framework links orthogonality to non-Markovian impurity dynamics through coherence revivals (Sindona et al., 2012). Even in a three-particle bosonic setting, a sudden impurity quench produces few-body precursors of AOC: the Loschmidt echo minimum reaches approximately $T_E$0 as $T_E$1, while the spectral function develops asymmetry and a cusp (Campbell et al., 2014).

In optical solid-state platforms, a cavity-created valence-band hole in a doped quantum well acts as an infinite-mass static scatterer for the two-dimensional electron gas. The real-time susceptibility computed in the Combescot–Nozières framework has long-time asymptotics

$T_E$2

with thresholds

$T_E$3

and the Fermi-edge singularity near $T_E$4,

$T_E$5

As density increases, AOC reduces the Rabi splitting, broadens and skews the upper polariton, and modifies Hopfield coefficients and lower-polariton effective mass (Baeten et al., 2014).

Transport settings admit more direct extraction of orthogonality exponents. In a weakly tunnel-coupled quantum dot capacitively coupled to a charge detector, the detector Fermi sea undergoes AOC with exponent

$T_E$6

and the orthogonality factors obey $T_E$7 at $T_E$8. In the nonequilibrium voltage-biased regime, the dot occupation at the center of the bias window has slope

$T_E$9

while the current peak scales as

$$\gamma(E)=\frac{1}{\pi^2}\,\bigl\|\arcsin |T_E/2|\bigr\|_{\mathrm{HS}}^2.$$0

Both provide direct estimates of $$\gamma(E)=\frac{1}{\pi^2}\,\bigl\|\arcsin |T_E/2|\bigr\|_{\mathrm{HS}}^2.$$1 (Sankar et al., 4 Jul 2025).

A complementary mesoscopic application interprets AOC as a source of switching heat in a transistor. When the gate changes the scattering matrix, the total released heat separates as

$$\gamma(E)=\frac{1}{\pi^2}\,\bigl\|\arcsin |T_E/2|\bigr\|_{\mathrm{HS}}^2.$$2

where $$\gamma(E)=\frac{1}{\pi^2}\,\bigl\|\arcsin |T_E/2|\bigr\|_{\mathrm{HS}}^2.$$3 is transport dissipation under bias and $$\gamma(E)=\frac{1}{\pi^2}\,\bigl\|\arcsin |T_E/2|\bigr\|_{\mathrm{HS}}^2.$$4 is a purely quantum-mechanical AOC contribution due to Fermi-sea rearrangement upon switching between conductance regimes. The switching heat remains finite even at $$\gamma(E)=\frac{1}{\pi^2}\,\bigl\|\arcsin |T_E/2|\bigr\|_{\mathrm{HS}}^2.$$5 and, in the adiabatic-plus-decoherence approximation, is controlled by $$\gamma(E)=\frac{1}{\pi^2}\,\bigl\|\arcsin |T_E/2|\bigr\|_{\mathrm{HS}}^2.$$6, i.e. the rate of change of the scattering amplitudes (Lebedev et al., 2020).

Across these realizations, the central invariant is not a single formula but a structural principle: a local perturbation, when coupled to a gapless or sufficiently correlated many-body background, can generate an extensive rearrangement whose overlap signature is singular in system size, time, or frequency. The canonical Anderson catastrophe is the simplest member of this broader class.