Berry Phase in Adiabatic Evolution

Introduction

In our last lesson, we established the quantum adiabatic theorem, which governs the evolution of a system under slowly changing conditions. We found that a system starting in an instantaneous eigenstate $∣ ψ_{n} (0)⟩$ remains in the corresponding eigenstate $∣ ψ_{n} (t)⟩$ as the Hamiltonian evolves. However, the state doesn't just evolve; it also acquires a phase. We saw that this total phase has two components: a familiar dynamical phase, related to the energy of the state, and a more enigmatic geometric phase, $γ_{n} (t)$ .

This lesson will focus squarely on this geometric phase, now famously known as the Berry phase. Our goal is to derive it rigorously, understand its profound geometric nature, and explore its essential properties. You will see that while the dynamical phase depends on how much time has passed, the Berry phase depends only on the path the system takes through its space of parameters. This crucial distinction moves us from a simple question of dynamics to a richer one of geometry and topology, connecting directly to the themes of this module.

Deriving the Berry Phase

In the previous lesson, we started with an ansatz for the adiabatically evolving state that already included the geometric phase. Let's now derive it from first principles. We begin with a more general ansatz for the state at time $t$ , assuming it follows the $n$ -th instantaneous eigenstate up to a total phase factor $ϕ_{n} (t)$ :

∣Ψ (t)⟩ = e^{i ϕ_{n} (t)} ∣ ψ_{n} (t)⟩

Our task is to find $ϕ_{n} (t)$ by ensuring this state satisfies the time-dependent Schrödinger equation (TDSE), $i ℏ \frac{d}{d t} ∣Ψ (t)⟩ = H (t) ∣Ψ (t)⟩$ .

The following video provides a clear, step-by-step derivation.

Deriving Berry phase, Berry connection & Berry curvature

The "Professor Nano" channel offers an excellent walkthrough of the standard textbook derivation of the Berry phase. This segment carefully substitutes the ansatz into the TDSE and isolates the term for the phase.

Please watch the segment from Chapter 1. Follow the derivation closely as the narrator: Substitutes the ansatz |\Psi(t)\rangle = e^{i\gamma(t)} e^{i\theta(t)}|n(t)\rangle into the TDSE (Note: they pre-emptively separate the total phase, but the logic is the same). Applies the product rule to the time derivative. Collects terms and projects onto the bra \langle n(t)|. Solves for the time derivative of the Berry phase, \dot{\gamma}_n(t). Transforms the resulting time integral into a line integral in parameter space.

Let's summarize the key result of the derivation. By substituting our ansatz into the TDSE and projecting onto $⟨ ψ_{n} (t) ∣$ , we arrive at an expression for the rate of change of the total phase:

ℏ ϕ ˙_{n} (t) = i ℏ ⟨ ψ_{n} (t) ∣ ψ ˙_{n} (t)⟩ - E_{n} (t)

Integrating this from 0 to $t$ gives the total phase:

ϕ_{n} (t) = Berry Phase, γ_{n} (t) i \int_{0}^{t} ⟨ ψ_{n} (t^{'}) ∣ \frac{d}{d t ^{'}} ∣ ψ_{n} (t^{'})⟩ d t^{'} Dynamical Phase - \frac{1}{ℏ} \int_{0}^{t} E_{n} (t^{'}) d t^{'}

We have successfully separated the total phase into two distinct parts. The second term is the dynamical phase, which depends explicitly on the energy $E_{n}$ and the duration of the evolution. The first term is the Berry phase, $γ_{n} (t)$ .

An important physical requirement is that the evolution must preserve the norm of the state, meaning $∣ ⟨ Ψ (t) ∣Ψ (t)⟩ ∣^{2} = 1$ . This implies that the total phase $ϕ_{n} (t)$ must be real. Since the dynamical phase is manifestly real, the Berry phase $γ_{n} (t)$ must also be real. This can be proven by differentiating the normalization condition $⟨ ψ_{n} (t) ∣ ψ_{n} (t)⟩ = 1$ , which shows that $⟨ ψ_{n} ∣ ψ ˙_{n} ⟩$ is purely imaginary, making $i ⟨ ψ_{n} ∣ ψ ˙_{n} ⟩$ purely real. The video resource you just watched provides this proof in its next chapter (Chapter 2), which you can review if you're interested.

The Geometry of the Phase: From Time to Parameter Space

The expression for $γ_{n} (t)$ is still an integral over time. So why do we call it "geometric"? The key insight, which Michael Berry articulated so powerfully, is that the instantaneous eigenstate $∣ ψ_{n} (t)⟩$ depends on time only indirectly, through the time-dependence of the parameters of the Hamiltonian, $R (t)$ . These parameters define a "parameter space" or "configuration space".

Using the chain rule, we can express the time derivative in terms of gradients in this parameter space:

\frac{d}{d t} ∣ ψ_{n} (R (t))⟩ = (\nabla_{R} ∣ ψ_{n} (R)⟩) \cdot \frac{d R}{d t}

Substituting this into the integral for the Berry phase yields a remarkable transformation:

γ_{n} (t) = i \int_{0}^{t} ⟨ ψ_{n} ∣ \nabla_{R} ∣ ψ_{n} ⟩ \cdot \frac{d R}{d t ^{'}} d t^{'} = \int_{R (0)}^{R (t)} i ⟨ ψ_{n} (R) ∣ \nabla_{R} ∣ ψ_{n} (R)⟩ \cdot d R

The integral is no longer over time, but is now a line integral over the path $C$ traced by the parameters $R (t)$ in the parameter space. The phase depends only on the geometry of this path, not on the speed at which it is traversed. This is the origin of the term "geometric phase".

This next video provides a clear discussion of this crucial conceptual leap.

L17.2 Berry's phase and Berry's connection

This lecture by Barton Zwiebach for MIT OpenCourseWare formalizes the transformation from a time integral to a path integral in configuration space.

Watch the first segment of the video from the beginning. The lecturer first uses the chain rule to rewrite the time derivative and then shows how the dt cancels, leaving a line integral over the path \gamma (which we call \mathcal{C}). He emphasizes the key takeaway: time plays no role anymore.

The Berry Connection and Gauge Invariance

The structure of the geometric phase, $γ_{n} (C) = \int_{C} A_{n} (R) \cdot d R$ , is highly suggestive. The integrand,

A_{n} (R) = i ⟨ ψ_{n} (R) ∣ \nabla_{R} ∣ ψ_{n} (R)⟩

is a vector field defined at every point in the parameter space, which we call the Berry connection. It plays a role analogous to the vector potential in electromagnetism. The Berry phase is then the flux of this connection along the path $C$ .

This analogy becomes even deeper when we consider the inherent phase ambiguity of quantum states. The instantaneous eigenstates $∣ ψ_{n} (R)⟩$ are only defined up to an arbitrary, parameter-dependent phase factor. Let's define a new set of equally valid eigenstates:

∣ ψ_{n}^{'} (R)⟩ = e^{- i β (R)} ∣ ψ_{n} (R)⟩

where $β (R)$ is any real function of the parameters. This is a local gauge transformation. How does this redefinition affect the Berry connection and the Berry phase?

The following video segment rigorously analyzes the consequences of this gauge transformation. This is the heart of the matter, as it determines when the Berry phase is a physically observable quantity.

L17.2 Berry's phase and Berry's connection

Continuing with the same MIT OCW lecture, this part first defines the Berry connection and then performs a detailed analysis of its behavior under gauge transformations.

Please watch from this point. The key stages of the argument are: The definition of the Berry connection \vec{A}_n(\vec{R}) (Berry connection). The gauge transformation of the states, |\psi'_n\rangle = e^{-i\beta} |\psi_n\rangle. Note the sign convention in the video is e^{-i\beta}, which differs from some texts but doesn't change the physics. The derivation of how the connection transforms: \vec{A}'_n = \vec{A}_n + \vec{\nabla}_{\vec{R}}\beta. This is identical to a gauge transformation of the electromagnetic vector potential (gauge transformation). The consequence for the Berry phase: it changes by \beta(\vec{R}_{\text{initial}}) - \beta(\vec{R}_{\text{final}}). The crucial conclusion: for a general, open path, the Berry phase is gauge-dependent and thus not observable. However, for a closed path where \vec{R}_{\text{initial}} = \vec{R}_{\text{final}}, the change is zero. The Berry phase is gauge-invariant and physically meaningful for cyclic adiabatic evolution (observability).

This is a profound result. The geometric phase accumulated during an adiabatic process is only a well-defined, measurable physical quantity if the system returns to its starting configuration in parameter space.

Example: A Spin-1/2 Particle in a Magnetic Field

To make these ideas concrete, let's analyze the canonical example: a spin-1/2 particle in a magnetic field $B (t)$ that changes its direction slowly.

The Hamiltonian is $H (t) = - μ \cdot B (t) = μ_{B} B (t) \cdot σ$ , where we absorb constants into the definition of $B$ . The parameters $R$ are the spherical coordinates $(θ, ϕ)$ that define the direction of the magnetic field vector, so the parameter space is a sphere $S^{2}$ . Let's calculate the Berry phase for the ground state (spin aligned with $- B$ ) as the field traces a closed loop on this sphere.

The following reading provides a detailed and rigorous calculation for this exact problem.

Lecture notes on Berry phases and topology

These lecture notes by D. Varjas, T. Fulga, and A. Akhmerov offer a sophisticated treatment of Berry phases. This section provides a complete, step-by-step derivation for the spin-1/2 system.

Please read the subsection 2.2.1 Example: Spin-1/2 in a magnetic field, starting from the setup. You can focus on the main derivation path and skip the detailed discussion of active vs. passive conventions at the end for now. The key steps in the reading are: Writing the Hamiltonian H(\vec{B}) and the projector onto the low-energy eigenspace P(t). Calculating the Berry connection \vec{A}_s(\lambda) (using Cartesian coordinates \lambda_i for the direction of \vec{B}) via the commutator formula A_s = [\dot{P}, P], which is equivalent to our definition. The result in Eq. (36) is what matters. Considering a specific path: the magnetic field rotates once around the equator (\theta = \pi/2, \phi goes from 0 to 2\pi). Solving the resulting equation for the evolution operator U_A(t) to find that the final state is the negative of the initial state, meaning a Berry phase of \pi has been acquired.

The calculation shows that for the special case of a path along the equator, the final state is $∣ ψ (1)⟩ = - ∣ ψ (0)⟩$ , which corresponds to a Berry phase of $γ = π$ .

More generally, one can show that the Berry connection for the ground state of this system is (in spherical coordinates on the parameter space):

A_{θ} = 0, A_{ϕ} = \frac{1}{2} (1 - cos θ)

This is the vector potential of a magnetic monopole of strength $g = - 1/2$ located at the origin of the parameter space.

By Stokes' theorem, the Berry phase for a closed loop $C$ is the integral of the "magnetic field" (the Berry curvature) over the surface $S$ bounded by the loop:

γ_{n} (C) = \oint_{C} A_{n} \cdot d R = \iint_{S} (\nabla \times A_{n}) \cdot d S

For our spin-1/2 system, this integral evaluates to:

γ_{-} (C) = - \frac{1}{2} Ω (C)

where $Ω (C)$ is the solid angle subtended by the path $C$ as viewed from the origin. For the equatorial path in the reading, the loop encloses half the sphere, so $Ω = 2 π$ , and the Berry phase is $γ_{-} = - π$ , which corresponds to the factor of $e^{- iπ} = - 1$ we found.

Conclusion

In this lesson, we have moved from the general statement of the adiabatic theorem to a deep analysis of the geometric phase. We have seen how this phase emerges directly from the Schrödinger equation and, most importantly, how its nature is fundamentally geometric.

Key Takeaways:

The Berry phase is a geometric phase acquired by a quantum state during a cyclic adiabatic evolution.
It is calculated as a line integral of the Berry connection $A_{n} = i ⟨ ψ_{n} ∣ \nabla_{R} ∣ ψ_{n} ⟩$ along a closed path in the system's parameter space.
The Berry connection transforms like a gauge potential. This makes the Berry phase for an open path gauge-dependent, but for a closed path, it is gauge-invariant and physically observable.
For a spin-1/2 particle in a rotating magnetic field, the Berry connection is formally identical to that of a magnetic monopole in parameter space, and the resulting Berry phase is proportional to the solid angle traced by the field's direction.

Preview of the Next Lesson:
The fact that the Berry phase for the spin-1/2 system corresponds to a monopole in parameter space is no accident. We saw that the total "magnetic flux" of a monopole through a closed surface is quantized. This hints at a deeper topological origin. In our next lesson, we will explore this connection further by defining topological invariants, such as the Chern number. We will see that the integral of the Berry curvature over an entire closed manifold in parameter space (like the Brillouin zone of a 2D insulator) can yield a quantized integer. This integer is robust against continuous deformations of the Hamiltonian and serves to classify distinct topological phases of matter.