Resolving Gauge Ambiguities of the Berry Connection in Non-Hermitian Systems

This paper resolves the intrinsic gauge ambiguities of the Berry connection in non-Hermitian systems by introducing a covariant formalism based on the Hilbert space metric tensor, which yields a uniquely defined, Hermitian connection that consistently recovers standard geometric phases and topological invariants while eliminating the complexities of the conventional biorthogonal approach.

The Gist

The Core Problem: The "Ruler" Problem in Quantum Physics

Imagine you are a cartographer trying to map a strange new land. In the world of standard quantum physics (Hermitian systems), you have a perfect, unchangeable ruler. No matter where you go, the ruler stays the same length. This makes measuring distances and angles (which physicists call "geometric phases" or "Berry phases") very easy and consistent. Everyone agrees on the map.

However, in Non-Hermitian systems (which describe things like lasers with gain, open systems, or materials with friction), the rules change. The "ruler" itself stretches and shrinks depending on where you are.

• The Left and Right Vectors: In these systems, you have two types of "directions" (left and right eigenvectors) that don't match up perfectly.
• The Ambiguity: Because the ruler stretches, you can choose to stretch your map in different ways. You could say, "The distance here is 10 meters," or "The distance here is 20 meters," and both could be mathematically "correct" depending on how you defined your ruler.

The Consequence: If you try to measure the "twist" or "curve" of the path a particle takes (the Berry phase), you get different answers depending on which ruler you picked. Sometimes, the math even suggests the particle is gaining or losing energy just because you changed your ruler, which makes no sense for a real quantum particle that must keep its total probability (its "size") constant.

The Solution: The "Einstein Elevator"

The author, Ievgen Arkhipov, proposes a new way to measure this land. He introduces a concept called a Covariant Berry Connection.

Think of it like this:
Imagine you are in a room with a distorted floor (the Non-Hermitian system). If you try to walk in a straight line, the floor might make you feel like you are turning or speeding up, even if you aren't.

• The Old Way: You just look at your feet and say, "I turned left!" But you didn't account for the floor tilting.
• The New Way (The Covariant Approach): You imagine an invisible "elevator" (a mathematical tool called a vielbein or Dyson map) that lifts you out of the distorted floor and places you in a flat, normal room (a Hermitian system).

In this flat room, you measure your turn. Then, you translate that measurement back to the distorted room, but you subtract the "tilt" of the floor.

What This Achieves

1. One True Answer: By using this "elevator" method, the author shows that there is actually only one correct way to measure the geometry of these systems. The four confusing, different answers we had before were just illusions caused by the stretching ruler.
2. No Fake Energy: The new method ensures that the "geometric phase" (the twist in the path) is purely about the shape of the path, not about the system gaining or losing energy due to the math. It separates the "real physics" from the "mathematical distortion."
3. The "Ghost" Curvature: In some cases, previous studies thought they found a "curved" path (like a magnetic field) in these systems. The author shows that this curvature was actually a "ghost"—it was just an artifact of the stretching ruler. When you use the new method, the ghost disappears, and you see the path is actually flat.

A Concrete Example from the Paper

The paper uses a specific example of a two-level system (like a coin that can be heads or tails).

• Old Method: If you change how you define the "size" of the coin (rescaling), the math says the coin picked up a "dissipative" phase (it lost energy) just because you changed the definition.
• New Method: The author's new formula automatically cancels out that fake energy loss. It says, "Wait, the coin didn't actually lose energy; you just changed the ruler." The result is a clean, zero phase, which is the physically correct answer for a system that shouldn't be losing energy.

Why This Matters

This paper is like fixing a broken compass. For a long time, physicists studying these weird, open quantum systems were using a compass that spun wildly depending on how they held it. This made it hard to build reliable topological devices (like future quantum computers).

By defining a unique, consistent, and Hermitian way to measure these phases, the author provides a solid foundation. Now, when physicists calculate the "twist" of a quantum state in a non-Hermitian system, they can be sure they are measuring the true geometry of nature, not just the quirks of their own math.

In short: The paper removes the confusion caused by "stretchy rulers" in quantum physics, giving us a single, clear map of how these strange systems actually behave.

Technical Summary

1. Problem Statement

Non-Hermitian Hamiltonians (NHHs) describe diverse physical systems (e.g., photonic gain/loss, metamaterials, open quantum systems) and exhibit unique spectral and topological phenomena like the non-Hermitian skin effect and exceptional points. However, their geometric characterization via the Berry connection faces a fundamental ambiguity:

• Biorthogonal Gauge Freedom: In NHHs, left ($\langle \psi^L|$) and right ($|\psi^R\rangle$) eigenvectors are not related by Hermitian conjugation. They possess an intrinsic $GL(N, \mathbb{C})$ gauge freedom where $|\psi^R\rangle \to c(\lambda)|\psi^R\rangle$ and $\langle \psi^L| \to c(\lambda)^{-1}\langle \psi^L|$.
• Four Inequivalent Definitions: Conventional approaches define the Berry connection by pairing left and right vectors in four ways ($A_{LL}, A_{LR}, A_{RL}, A_{RR}$). These are generally not gauge-equivalent.
• Complex Geometric Phases: Under non-unitary rescalings ($|c| \neq 1$), the conventional connections acquire imaginary components. This leads to complex-valued Berry phases and curvatures.
• Physical Inconsistency: In genuine quantum mechanics, the wavefunction norm represents total probability and must remain unity. The conventional biorthogonal formalism often conflates intrinsic geometric phases with "dissipative" metric-induced effects, leading to fictitious geometric amplification/attenuation that violates probability conservation in the quantum regime.

2. Methodology

The author proposes a covariant formalism based on the metric tensor of the Hilbert space to resolve these ambiguities.

• Metric Operator ($\eta$): To treat the state $|\psi\rangle$ as a probability amplitude, a positive-definite metric operator $\eta$ is introduced such that $\langle \psi | \eta | \psi \rangle = 1$.
• Hermitizing Map (Vielbein $S$): The metric is decomposed as $\eta = S^\dagger S$. The operator $S$ acts as a "vielbein" (or Dyson map), mapping the non-Hermitian system to an equivalent Hermitian system ($H_H$) via $H_H = S(H + i\Gamma_t)S^{-1}$, where $\Gamma_t$ compensates for the non-Hermitian spectrum.
• Covariant Derivative: Ordinary derivatives are insufficient because the inner product varies across the parameter space. A covariant derivative is defined as:
  $$D_\sigma = \partial_\sigma + S^{-1}\partial_\sigma S = \partial_\sigma + \Gamma_\sigma$$
  Here, $\Gamma_\sigma$ is the metric-compatible connection (Maurer-Cartan form) that accounts for the deformation of the Hilbert space.
• Covariant Berry Connection (CBC): The Berry connection is redefined using the Hermitian frame states $|\phi^H\rangle = S|\psi^R\rangle$ or directly in the non-Hermitian frame as:
  $$A_{mn}^\lambda = i\langle \psi^R_m | \eta D_\lambda | \psi^R_n \rangle = A_{LR}^{\lambda} + i\langle \psi^L_m | \Gamma_\lambda | \psi^R_n \rangle$$
  This definition isolates the intrinsic geometry of the eigenbundle from the metric-induced contributions.

3. Key Contributions

A. Resolution of Gauge Ambiguity

The paper demonstrates that the Covariant Berry Connection (CBC) is uniquely defined and Hermitian. Unlike the conventional four definitions, the CBC transforms as an affine connection under general $GL(N, \mathbb{C})$ frame transformations:
$$A' = T^{-1}\tilde{A}T + iT^{-1}\partial_\lambda T$$
where $\tilde{A} = A + \Xi$. The term $\Xi$ is a metric-induced correction that exactly cancels the gauge-dependent imaginary parts arising from non-unitary rescalings of the eigenvectors.

B. Separation of Intrinsic vs. Metric Geometry

The framework successfully decouples two distinct geometric contributions:

1. Intrinsic Geometry: The geometry of the physical, norm-preserving eigenspace.
2. Metric Geometry: Contributions arising solely from the parameter-dependent metric $\eta$.
   The conventional biorthogonal approach mixes these, often misinterpreting metric artifacts as physical geometric phases.

C. Recovery of Standard Limits

• In the Hermitian limit ($\eta \to I, \Gamma \to 0$), the CBC reduces to the standard Berry connection.
• If the metric is constant ($\Gamma = 0$), the CBC reduces to the conventional left-right connection, validating its consistency.

4. Results and Examples

Example 1: Pseudo-Hermitian Two-Level System

• Setup: A Hamiltonian with real eigenvalues but non-Hermitian structure.
• Conventional Result: The left-right Berry connection ($A_{LR}$) is non-zero and depends on parameters. Under rescaling, it acquires an imaginary term, suggesting a "geometric dissipation" (complex phase) even though the energy is real.
• CBC Result: The Covariant Berry Connection vanishes identically ($A=0$).
• Interpretation: The non-zero $A_{LR}$ was purely a metric artifact. The intrinsic geometry of the physical eigenspace is trivial. The CBC correctly identifies that no geometric phase is accumulated in a cyclic evolution for this system.

Example 2: Non-Abelian Holonomy with Exceptional Points (EP)

• Setup: A system with exceptional points where eigenstates permute.
• Conventional Result: The holonomy is unitary despite the non-Hermitian nature of the connection.
• CBC Result: The CBC includes a metric correction term ($i\sigma_y d\theta/2$). However, for a closed loop encircling the EP, the integral of the metric term vanishes ($\oint d\theta = 0$).
• Interpretation: Both formalisms yield the same topological invariant (the exchange of eigenstates). This confirms that the CBC retains true non-trivial topology while filtering out spurious metric effects.

5. Significance and Implications

• Physical Consistency: The CBC provides a rigorous geometric foundation for non-Hermitian quantum systems that respects probability conservation (norm preservation). It eliminates the "fictitious" geometric amplification/attenuation found in previous biorthogonal approaches.
• Topological Invariants: Topological invariants (e.g., Chern numbers) derived from the CBC are uniquely defined and physically meaningful. They distinguish between intrinsic topological features and artifacts induced by the choice of metric.
• Reinterpretation of Literature: The paper suggests that many previously reported complex Berry phases and non-Hermitian curvatures in the literature may be manifestations of the metric structure rather than intrinsic properties of the Hamiltonian, particularly in regimes where state-norm preservation is not strictly enforced.
• Unified Framework: This work establishes a covariant framework that unifies the treatment of Berry phases, non-Abelian holonomies, and topological invariants for both Hermitian and Non-Hermitian systems, resolving long-standing ambiguities in the field.

In summary, Arkhipov's work replaces the ambiguous, multi-valued biorthogonal Berry connection with a unique, Hermitian, and covariant connection, ensuring that geometric phases in non-Hermitian quantum systems reflect the true physics of the eigenbundle rather than artifacts of the Hilbert space metric.