Pseudospectral phenomena and the origin of the non-Hermitian skin effect

This paper argues that the non-Hermitian skin effect originates from spectral instability and non-reciprocity rather than point-gap topology, demonstrating that the commonly assumed link between spectral winding and boundary localization is a non-generic consequence of translational invariance.

The Gist

The Big Idea: A Misunderstanding About "Skin"

Imagine you have a long, crowded hallway (a quantum system). In a normal hallway, people (particles) spread out evenly. But in a special kind of "non-Hermitian" hallway, something weird happens: everyone suddenly piles up at one end of the hall.

Physicists call this the Non-Hermitian Skin Effect (NHSE). It's like a crowd surge where all the eigenstates (the quantum "people") get stuck at the door.

For a long time, scientists thought this pile-up happened because of a hidden topological map (a complex, winding path) inside the hallway. They believed the "map" forced the people to the edge.

This paper says: "No, that's not it."

The author, Jesko Sirker, argues that the pile-up isn't caused by a magical map. Instead, it's caused by the hallway being unstable and one-way. The "map" (topology) is actually hiding somewhere else, and the "pile-up" is just a side effect of the hallway being wobbly.

---

The Three Characters in the Story

To understand the mix-up, we need to meet three characters that often look the same but are actually different:

1. The One-Way Street (Non-Reciprocity): Imagine a hallway where you can walk forward easily, but walking backward is like wading through molasses. This creates a "wind" pushing everything to one side.
2. The Wobbly Table (Non-Normality): Imagine a table that is perfectly balanced when nothing touches it, but if you tap it even slightly, it wobbles violently and changes shape. In physics, this means the system is extremely sensitive to tiny changes (like a change in the door or a small bump).
3. The Winding Map (Point-Gap Topology): Imagine a path that loops around a hole in the floor. If you walk along it, you circle the hole. This is the "topology" scientists thought was causing the pile-up.

The Mistake: In the simplest model (the Hatano-Nelson chain), all three characters appear together. The hallway is one-way, it's wobbly, and the map winds. Scientists assumed the Winding Map was the boss causing the Pile-Up.

The Truth: The author shows that the Winding Map and the Pile-Up are actually independent. You can have one without the other.

---

The Analogy: The Unstable House vs. The Blueprint

1. The "Skin Effect" is a House of Cards

Think of the "Skin Effect" (the pile-up at the edge) as a house built out of cards.

• The Problem: This house is built on a wobbly table (Non-Normality).
• The Result: If you blow a gentle breeze (a tiny perturbation) or change the door (boundary conditions), the whole house collapses or rearranges itself completely.
• The Lesson: Because the house is so unstable, you can't use it to store a secret message (topological information). If the house changes every time you look at it, it can't be a reliable map.

2. The "Topology" is the Blueprint (Hidden in the Basement)

So, where is the real "map" (topology)?

• The author says the map isn't in the wobbly house (the eigenstates). It's in the blueprint (the singular-value spectrum).
• Even if the house of cards collapses, the blueprint remains stable. It tells you that, in a perfect, infinite world, there should be a special room at the edge.
• In a real, finite system, this "room" isn't a perfect room you can walk into; it's a "ghost room" that is so close to being real that it might as well be there, but it's only visible if you look at the blueprint, not the house.

---

The Experiment: The Two-Track Ladder

To prove his point, the author built a "Ladder" model (two chains connected together) where he could turn the knobs independently. He created three scenarios:

1. The "Pile-Up without the Map":

   • He set up a hallway that was very one-way and wobbly, but the "winding map" was cancelled out (zero winding).
   • Result: The crowd still piled up at the edge!
   • Conclusion: You don't need the winding map to get the skin effect. You just need the one-way street and the wobbly table.

2. The "Map without the Pile-Up":

   • He set up a hallway with a strong winding map, but he made the hallway stable (not wobbly) and removed the one-way street.
   • Result: The crowd spread out evenly in the middle. No pile-up at the edge.
   • Conclusion: You can have the winding map without the skin effect.

3. The "Both" Scenario:

   • When he turned everything on, the crowd piled up, and the map wound. This is what everyone saw before and thought they were the same thing.

The Takeaway: Why This Matters

1. The "Skin" isn't Topological.
The Non-Hermitian Skin Effect is not a topological phenomenon. It is a result of instability. It happens because the system is "non-normal" (wobbly) and "non-reciprocal" (one-way). If you break the symmetry or add a tiny bit of noise, the "skin" disappears or changes completely.

2. Topology is Real, but Hidden.
The topological properties (the winding) are real, but they don't show up in the "pile-up" of particles. They show up in the singular values (a mathematical way of looking at the system's stability). These values are stable and don't change when you wiggle the system.

3. A Warning for Future Research.
Scientists have been using the "pile-up" to guess the topological nature of materials. This paper says: Stop doing that. The pile-up is a fragile illusion caused by the system's instability. To find the true topological nature of a material, you need to look at the stable "blueprint" (singular values), not the wobbly "house" (eigenstates).

In a Nutshell

Imagine a crowd of people running to the exit of a stadium.

• Old View: "They are running to the exit because the stadium has a magical, winding path that forces them there."
• New View (This Paper): "No, they are running to the exit because the floor is slippery and the wind is blowing them that way. If you fix the floor or stop the wind, they stop running to the exit. The 'magical path' is actually a separate thing that exists in the stadium's design plans, but it's not what's making them run."

The author has separated the slippery floor (instability/skin effect) from the design plans (topology) to show they are two different things.

Technical Summary

1. Problem Statement

The Non-Hermitian Skin Effect (NHSE) is a phenomenon where eigenstates of a non-Hermitian system with open boundary conditions (OBC) macroscopically accumulate at the system's boundaries. The prevailing theoretical framework attributes the NHSE to point-gap topology in the Bloch Hamiltonian, specifically a non-zero spectral winding number under periodic boundary conditions (PBC). This view relies heavily on the Generalized Brillouin Zone (GBZ) construction, which assumes translational invariance.

However, this perspective faces two fundamental conceptual issues:

1. Translational Invariance vs. Topology: Topological properties must be robust against local perturbations that break translational symmetry. The reliance on GBZ (which requires translational invariance) to define topology is conceptually inconsistent.
2. Spectral Instability: Physically relevant non-Hermitian systems are described by non-normal operators ($H^\dagger H \neq H H^\dagger$). The eigenspectra of non-normal operators are notoriously unstable under generic perturbations (including numerical noise and boundary changes). If the eigenspectrum is unstable, it cannot serve as a stable carrier of topological information.

The paper asks: Is the NHSE a topological phenomenon driven by spectral winding, or is it a manifestation of spectral instability and non-reciprocity in non-normal operators?

2. Methodology

The author employs a combination of rigorous mathematical theorems from operator theory and numerical simulations to disentangle three distinct properties often conflated in simple models:

• Non-normality: The condition $H^\dagger H \neq H H^\dagger$.
• Non-reciprocity: Asymmetric hopping amplitudes ($t_R \neq t_L$).
• Point-gap Topology: Non-zero spectral winding number ($I(E) \neq 0$).

Key Analytical Tools:

• Pseudospectrum Analysis: Defined as $\sigma_\epsilon(H) = \{ \lambda : \|(H - \lambda I)^{-1}\| > 1/\epsilon \}$. This tool characterizes the sensitivity of the spectrum to perturbations. For non-normal matrices, the pseudospectrum can be vastly larger than the actual eigenspectrum.
• Toeplitz Operator Theory: Utilizing the index theorem for Toeplitz operators, which relates the winding number to the dimension of the kernel (null space) of the operator in the semi-infinite limit.
• Singular Value Decomposition (SVD): Analyzing the singular values ($s_i = \sqrt{\lambda_i(H^\dagger H)}$) of finite systems. Unlike the eigenspectrum of non-normal operators, the singular-value spectrum is stable under perturbations.
• Model Construction:
  • 1D Hatano-Nelson (HN) Chain: Used as a baseline to demonstrate the interplay of non-normality and winding.
  • Hatano-Nelson Ladder (2-band model): A novel construction introduced to independently vary non-normality/non-reciprocity and point-gap winding. This allows for the creation of counterexamples where NHSE and winding exist or do not exist independently.

3. Key Contributions and Results

A. Re-evaluation of the 1D Hatano-Nelson Chain

• Spectral Instability: The paper demonstrates that the OBC eigenspectrum of the HN chain is extremely sensitive to perturbations. While the unperturbed OBC spectrum is real, adding generic complex perturbations causes the spectrum to revert to the PBC elliptical shape (the pseudospectrum).
• Quasi-Hermiticity Masking: The HN chain with OBC is quasi-Hermitian (similar to a Hermitian matrix via a similarity transform). This makes its spectrum appear stable against local perturbations (like random onsite potentials), leading to the misleading intuition that the skin effect is topological. However, it remains unstable against global or generic complex perturbations.
• Topological Content: The true topological information is not in the unstable eigenspectrum but in the singular-value spectrum. For a system with winding $I(E)$, the singular-value spectrum exhibits $K = |I(E)|$ singular values separated from the bulk by a gap, which vanish in the thermodynamic limit. These correspond to topologically protected boundary modes in the semi-infinite limit.

B. The Hatano-Nelson Ladder: Decoupling NHSE and Topology

The author constructs a two-band ladder model (two coupled HN chains) to explicitly separate the effects:

1. NHSE without Point-Gap Winding:

   • Setup: Two chains with equal but opposite non-reciprocity ($t_R = \tilde{t}_L$ and $t_L = \tilde{t}_R$) coupled in a triangular fashion ($t_D=0$).
   • Result: The total winding number $I(E) = 0$ (windings cancel). However, the system exhibits a strong NHSE where all eigenstates (and generalized eigenvectors) accumulate at the left boundary.
   • Conclusion: NHSE can exist without topological winding.

2. Point-Gap Winding without NHSE:

   • Setup: The triangular structure is broken by introducing inter-chain couplings ($t_D \neq 0, \gamma \neq 0$), making the matrix less non-normal.
   • Result: The system retains a non-zero winding number ($I(E) = -1$) around certain energies. However, the eigenstates become delocalized (bulk-like) and do not accumulate at the edges.
   • Conclusion: Topological winding can exist without the NHSE.

3. Stability of Singular Values:

   • In the "Winding without NHSE" regime, while the eigenspectrum shows no skin effect, the singular-value spectrum still reveals a topologically protected zero mode (a singular value approaching zero with a gap to the bulk). This confirms that the topology is encoded in the singular values, not the eigenvectors.

4. Significance and Implications

• Redefining the Origin of NHSE: The paper establishes that the NHSE is not a topological phenomenon. Instead, it is a consequence of non-normality combined with non-reciprocity, which leads to extreme spectral sensitivity to boundary conditions.
• Failure of the Bulk-Boundary Correspondence (BCC) based on Eigenspectra: The commonly assumed BCC, which links spectral winding to boundary localization, relies implicitly on translational invariance. Since topology requires robustness against symmetry-breaking perturbations, the eigenspectrum of non-normal operators cannot be the correct object for defining topological invariants in non-Hermitian systems.
• New Framework for Topology: The paper proposes that the correct bulk-boundary correspondence for non-Hermitian systems must be based on:
  1. Toeplitz Operator Theory: Defining topology via the index of the operator.
  2. Singular-Value Spectrum: Using the stable singular values of finite systems to detect topological phases (K-splitting theorems).
• Physical Interpretation: The "skin modes" observed in finite systems are often metastable states with long lifetimes (due to the small singular values) rather than true eigenmodes in the presence of generic perturbations.

Conclusion:
The work fundamentally shifts the understanding of non-Hermitian physics by decoupling spectral instability (the source of the skin effect) from topological invariants. It argues that the NHSE is a generic feature of non-normal, non-reciprocal transport, while true topological protection is encoded in the singular-value spectrum and the index of Toeplitz operators, independent of the spectral winding of the Bloch Hamiltonian.