---
title: 'Hatano–Nelson Model: Non-Hermitian Lattice Dynamics'
url: https://www.emergentmind.com/topics/hatano-nelson-model-hn-model
type: topic
---

# Hatano–Nelson Model: Non-Hermitian Lattice Dynamics

The Hatano–Nelson (HN) model is a foundational non-Hermitian quantum lattice system introduced to study delocalization–localization transitions and non-reciprocal transport in one-dimensional chains with asymmetric hopping. Its exact solvability, tunable non-Hermiticity, and rich interplay of disorder, topology, and interactions have made it a central paradigm for exploring non-Hermitian skin effects, spectral topology, and wave-packet dynamics in quantum and classical systems.

## 1. Hamiltonian, Boundary Conditions, and Spectrum

The canonical HN Hamiltonian on an $L$-site 1D chain is
\[
H_{\mathrm{HN}} = \sum_{j=1}^{L-1}\left(t_R\,c_j^\dagger c_{j+1} + t_L\,c_{j+1}^\dagger c_j\right) + \sum_{j=1}^L V_j c_j^\dagger c_j,
\]
where $t_R = J e^{+g}$, $t_L = J e^{-g}$ set the right/left hopping amplitudes and $g \in \mathbb{R}$ parametrizes the nonreciprocity ($J$ is the energy unit). $V_j$ can represent random/disordered or quasiperiodic onsite potentials.

- **Periodic boundary conditions (PBC):** The single-particle spectrum is
  \[
  E(k) = 2 \cosh g \cos k - i\, 2 \sinh g \sin k, \quad k \in [0,2\pi).
  \]
  This forms an oriented ellipse in the complex energy plane.

- **Open boundary conditions (OBC):** The spectrum collapses onto $[-2,2]$. The corresponding eigenstates are exponentially localized at one boundary (NHSE):
  \[
  \phi^{(n)}_j \sim e^{-g\,j}\ \text{or}\ e^{+g\,j},
  \]
  localizing at the left ($j=1$) or right ($j=L$) end depending on the sign of $g$ [2504.04370]. 

This spectral collapse and boundary localization under OBC is the essence of the non-Hermitian skin effect.

## 2. Non-Hermitian Skin Effect and Topological Spectral Winding

The NHSE is characterized by the accumulation of all right (or left) eigenstates at one edge. Under PBC, the nonzero spectral winding of $E(k)$ encodes the nontrivial non-Hermitian topology:
\[
w(E_0) = \frac{1}{2\pi i} \oint_{BZ} d k\; \partial_k \ln[E(k) - E_0].
\]
Nonzero $w$ implies that under OBC, the bulk eigenstates redistribute from the extended (PBC) case and collapse to the boundary.

The skin effect is a direct result of the non-unitarity in recursive dynamics:
\[
\phi_{j+1} = -\left(\frac{t_L}{t_R}\right)\phi_{j-1},
\]
establishing an exponential bias [2504.04370]. The directionality of NHSE is set by the sign of $g$, while the skin depth is set by $1/|g|$.

## 3. Disorder, Localization, and Skin–Anderson Transition

Introducing onsite random potential $w_j \in [-W/2, W/2]$ yields competition between NHSE and Anderson localization [2504.04370]:
\[
H = H_{\mathrm{HN}} + \sum_j w_j |j\rangle\langle j|.
\]

- **Transfer-matrix formalism:** The Lyapunov exponent $\gamma(E, g)$ controls localization length $\xi(E, g)=1/\gamma(E, g)$,
  \[
  \gamma(E, g) = \gamma(E, 0) + |g|,
  \]
  which means the nonreciprocity $g$ shifts all exponents by $|g|$ [2504.04370].

- **Skin–Anderson transition:** The DL (delocalization–localization) mobility edge is given by $\gamma(E, g)=0$, with a critical disorder $W_c(g)$ determined by $W_c = 2 e^{|g|}$ in the weak disorder limit.

Regimes:
- $W < W_c(g)$: coexistence of boundary-piled (skin) and Anderson-localized modes; spectral topology is mixed.
- $W > W_c(g)$: all states Anderson-localized, NHSE is suppressed; spectrum is real under OBC.

- **Quasiperiodic and strictly ergodic potentials** yield analogous phase diagrams, but deep mathematical results show that in strictly ergodic chains the spectrum consists of one-dimensional analytic arcs, with a Lyapunov-exponent-controlled real–complex phase transition and sharp mobility edges [2311.09899].

## 4. Wave-Packet Spreading and Dynamical Scaling

The dynamical consequence of non-Hermiticity and disorder is seen in the time-evolution of an initially localized wave packet under $H$ [2504.04370]:

- **Clean case ($W=0$):** Transport is unidirectionally ballistic,
  \[
  \Delta x(t) \sim t.
  \]
  This is a direct result of the dominant imaginary part of $E(k)$, and the center of mass drifts as $x(t) \sim v(g)t$ with $v(g) = 2\,\mathrm{sgn}(g)\,\cosh g$.

- **Weak disorder ($W < W_c$):** Initially ballistic, crossover to superdiffusive spreading
  \[
  \Delta x(t) \sim t^{2/3}
  \]
  at long times, arising from the band-tail scaling of the imaginary density of states (iDOS).

- **Strong disorder ($W > W_c$):** Initial diffusive scaling,
  \[
  \Delta x(t) \sim t^{1/2},
  \]
  crosses over at long times again to the superdiffusive regime $\Delta x(t) \sim t^{2/3}$, due to iDOS near band edges.

All regimes retain a persistent directional bias in $x(t)$ reflecting nonreciprocal dynamics.

## 5. Periodic Potentials, Spectral Topology, and Winding

Inclusion of a periodic potential $V_j = V_0 \cos(2\pi j/q)$ introduces band gaps and multiple spectral topologies [1009.1946]:

- At $V_0 > t \sinh h$ (with $h = g$), a band gap opens,
  \[
  W_G = 2 \sqrt{V_0^2 - (t\sinh h)^2}\,.
  \]
- Disorder and periodic modulation compete, leading to a rich sequence of phases: extended (E), localized (L), gapped (G), and their combinations.
- The transition between topologies is accompanied by changes in spectral winding numbers, which classify the phase and skin effect outcomes.

Finite-size scaling of the participation ratio quantifies the localization threshold $\Delta_c \approx 2 t \sinh h$, and the critical exponent for the localization transition is $\nu \approx 1$.

## 6. Analytical and Numerical Approaches

Key methods for analysis are [2504.04370]:
- **Transfer-matrix Lyapunov exponent computation**
- **Generalized Thouless formula for iDOS**
- **Direct wave-packet time evolution with normalization under non-unitary dynamics**
- **Spectral winding calculation under PBC for topological characterization**
- **Comparison of OBC and PBC spectra for diagnosis of the skin effect and phase boundaries**

Empirical verification uses ensemble averaging over disorder realizations and large-scale numerical diagonalization; all methods align for the scaling exponents and mobility edge predictions.

## 7. Physical Implications and Broader Context

The HN model and its generalizations establish:
- The existence of the NHSE, a fundamentally non-Hermitian boundary effect encoded in spectral winding, not present in Hermitian chains.
- Novel dynamical scaling laws, such as long-time superdiffusive spreading, from the interplay of nonreciprocity and randomness.
- Rich spectral phase diagrams beyond Hermitian analogs, including skin–Anderson and skin–gap transitions, mobility edges, and reentrant localization.
- Transfer-matrix formalisms allow for deep analysis of localization, spectrum, and eigenstate topology, interconnecting non-Hermitian physics with random matrix theory and Lyapunov analysis.

The mathematical framework underlying the HN model plays a pivotal role in the non-Hermitian physics program, directly informing experimental realizations in photonics, acoustics, electronic circuits, and cold-atom systems [2504.04370], [1009.1946], [2311.09899].

Source: https://www.emergentmind.com/topics/hatano-nelson-model-hn-model