🔢 mathematics

Which Adiabatic Cycles Alter an Action Variable?

This paper demonstrates that while action variables are typically preserved during adiabatic evolutions, specific adiabatic cycles can induce "action variable holonomy" and torus pumping in completely integrable systems, a phenomenon classified via homotopy for kicked-top models.

Original author: Atushi Tanaka, Taksu Cheon

Published 2026-11-02

📖 6 min read🧠 Deep dive

CC BY 4.0

Original author: Atushi Tanaka, Taksu Cheon

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the author. For technical accuracy, refer to the original paper. Read full disclaimer

In the quiet, orderly world of classical physics, there are quantities that act like perfect anchors. When a system is simple enough to be solved completely, certain values remain fixed even as the system's environment changes slowly. These are known as action variables. Imagine a spinning top or a planet orbiting a star; if you were to gently nudge the conditions around them—perhaps by slowly changing the strength of gravity or the tilt of a magnetic field—these specific values would not budge. They are the system's way of remembering its original state, no matter how the surroundings evolve. For over a century, physicists have relied on this stability to predict how such systems behave, assuming that if the change is slow enough, the system simply adjusts its shape while keeping its core identity intact.

However, a recent investigation by researchers Atushi Tanaka and Taksu Cheon has revealed a subtle twist in this story. They discovered that while the numerical value of these anchors remains constant, the very definition of what they measure can shift in a surprising way. It is as if a ruler that once measured distance in inches suddenly, after a long journey, begins measuring in centimeters, even though the number it displays has not changed. This phenomenon, which the researchers call action variable holonomy, means that a system can return to its starting point in time and space, yet end up in a completely different configuration than where it began. This discovery challenges the assumption that slow, smooth changes always lead back to where they started, suggesting instead that the path taken matters just as much as the destination.

To explore this idea, the team focused on a specific type of system known as a kicked top. This is a theoretical model of a spinning object that receives regular, sharp taps or "kicks" at fixed intervals, much like a child pushing a swing at just the right moment to keep it moving. In their study, the researchers treated these kicks as a series of magnetic pulses applied to a spinning object. They imagined a scenario where the direction or strength of these pulses changes very slowly over time, tracing out a closed loop before returning to the original setting. The question was simple: if you start with the top spinning in a certain way, and you slowly vary the magnetic kicks in a circle, does the top end up exactly as it started?

The researchers found that the answer depends entirely on the shape of the path the magnetic field takes through its possible configurations. They identified two distinct types of loops. The first type is a simple, small loop that can be shrunk down to a single point without getting stuck. When the system follows such a path, the action variable behaves as expected: the system returns to its original state, and the definition of the action variable remains unchanged. But the second type of loop is more complex. It winds around a central "hole" in the space of possible configurations in a way that cannot be untangled. When the system follows this larger, knotted path, something remarkable happens. Even though the numerical value of the action variable stays the same throughout the journey, its functional form flips. If the action variable originally described a spin pointing upward, it ends up describing a spin pointing downward.

This flip is not a mistake or a glitch; it is a fundamental property of the geometry of the system. The researchers showed that this change is tied to how many times the path of the magnetic field crosses a specific boundary in the mathematical space they used to describe the system. If the path crosses this boundary an odd number of times, the system undergoes a complete reversal. If it crosses an even number of times, or not at all, the system returns to normal. This means that the history of the system is written into its final state. A system that has traveled a "large" loop is fundamentally different from one that has traveled a "small" loop, even if both loops start and end at the same settings.

The consequence of this flip is a phenomenon the authors call adiabatic torus pumping. In the language of physics, the state of the spinning top can be visualized as a ring, or a torus, floating in a space of all possible motions. When the system undergoes the complex loop that causes the flip, this ring does not return to its original position. Instead, it is pumped to a different location in the space of motions. It is as if a train leaves a station, travels along a track that loops around a mountain, and arrives back at the station platform, only to find it has been transported to a different station entirely, despite the tracks looking identical. The researchers demonstrated this with a specific example involving a classical spin subject to a static magnetic field and a series of periodic pulses. By carefully varying the strength of the pulses, they could force the system to trace the complex path that triggers this pumping effect.

This finding has important implications for how we understand the stability of physical systems. It shows that the global shape of the path taken through a parameter space is just as critical as the local details of the change. The researchers noted that this behavior is specific to systems that are driven periodically, like the kicked top, and differs from systems with static fields where such a reversal would require an impossible, discontinuous jump. They also pointed out that this effect prevents the definition of a certain type of angle, known as the Hannay angle, for most states in the system, because the starting and ending configurations are no longer the same. Only for a very specific state, where the action variable is zero, does the system return to itself regardless of the path taken.

Ultimately, this work provides a clear map of which cycles preserve the identity of a system and which ones alter it. By classifying these paths based on their topology—their ability to be stretched or shrunk without tearing—the researchers have shown that the universe keeps a record of the journey. A system that has traveled a complex, non-trivial loop carries a permanent mark of that journey, a sign that it has been to a place where the rules of its own definition have flipped. This adds a new layer of depth to our understanding of adiabatic processes, reminding us that in the world of physics, the route taken is often as significant as the destination reached.

Technical Summary: Which Adiabatic Cycles Alter an Action Variable?

Problem Statement
Action variables in completely integrable classical Hamiltonian systems are traditionally defined as adiabatic invariants; their numerical values remain constant under slow (adiabatic) variations of external parameters. However, this paper investigates a subtler aspect of adiabatic evolution: the behavior of the functional form of the action variable. While the value $I$ is preserved, the action variable is a function of phase-space coordinates (e.g., angular momentum $\mathbf{S}$ ) that depend on the system parameters. The authors ask whether, after traversing a closed adiabatic cycle, the functional form of the action variable returns to its original state or undergoes a nontrivial change. Such a change, termed action variable holonomy, implies that a torus initially specified by a value $I_0$ is transported to a different torus (specified by the same value but a different functional form) at the end of the cycle. This phenomenon, called adiabatic torus pumping, challenges the assumption that action variables are globally well-defined invariants without holonomy, a condition often required for the definition of Hannay angles.

Methodology
The authors analyze a family of completely integrable classical kicked-top systems, which serve as a discrete-time Floquet system.

System Definition: The system is modeled as a linear map acting on classical angular momentum $\mathbf{S}$ , defined by a rotation matrix $R \in SO(3)$ . For a fixed $R$ , the system is equivalent to a classical spin in a static magnetic field. The conserved quantity $I(\mathbf{n}) = \mathbf{n} \cdot \mathbf{S}$ (where $\mathbf{n}$ is the rotation axis) serves as the action variable.
Parameter Space Analysis: To study adiabatic cycles, the authors parametrize the rotation matrix $R$ using an angle vector $\boldsymbol{\theta}$ , where the direction and magnitude correspond to the rotation axis and angle, respectively. The parameter space is identified with a three-dimensional ball $B$ of radius $\pi$ , with antipodal points on the surface identified (representing the same rotation). The origin ( $R=I_3$ ) is excluded because the rotation axis is undefined there.
Topological Classification: The authors examine closed paths (cycles) in this parameter space. They track the evolution of the rotation axis $\mathbf{n}$ along a cycle $C$ . Since $\mathbf{n}$ must vary continuously, crossing the boundary of the parameter ball (where antipodal points are identified) forces a switch in the branch of the function defining $\mathbf{n}$ .
Physical Realization: A physical example is constructed using a classical spin subject to a static magnetic field and a time-periodic pulsed magnetic field. The Floquet matrix of this system corresponds to the rotation matrix $R$ . Two specific adiabatic cycles are analyzed: one where the pulsed field direction rotates ( $C_1$ ) and one where the pulsed field strength varies monotonically ( $C_2$ ).

Key Contributions and Results

Identification of Holonomy Condition: The paper establishes that action variable holonomy occurs if and only if the adiabatic cycle $C$ $C$ crosses the surface of the parameter ball $B$ $B$ an odd number of times ( $\nu(C)$ $ν (C)$ is odd).
- If $\nu(C)$ is even, the rotation axis $\mathbf{n}$ returns to its initial value ( $\mathbf{n}_1 = \mathbf{n}_0$ ), and the action variable's functional form is unchanged.
- If $\nu(C)$ is odd, the rotation axis reverses sign ( $\mathbf{n}_1 = -\mathbf{n}_0$ ). Consequently, the action variable transforms as $I(\mathbf{n}_1) = -I(\mathbf{n}_0)$ , indicating a sign change in its functional form.
Homotopy Classification: The authors recast the crossing condition into a topological framework. They demonstrate that the fundamental group of the parameter space (excluding the origin) is $\pi_1(B) \cong \mathbb{Z}_2 = \{[e], [\gamma]\}$ $π_{1} (B) ≅ Z_{2} = {[e], [γ]}$ .
- Cycles homotopic to the identity $[e]$ (contractible loops) do not induce holonomy.
- Cycles homotopic to the non-contractible loop $[\gamma]$ (which connects a point on the surface to its antipode) induce holonomy.
- Thus, the necessary and sufficient condition for action variable holonomy is that the adiabatic cycle belongs to the non-trivial homotopy class $[\gamma]$ .
Adiabatic Torus Pumping: The sign reversal of the action variable's functional form leads to adiabatic torus pumping. A torus initially defined by $I(\mathbf{n}_0) = I_0$ is transported to a torus defined by $I(\mathbf{n}_1) = I_0$ . If $\nu(C)$ is odd, this final torus corresponds to the geometric location of $I(\mathbf{n}_0) = -I_0$ , effectively moving the system to a different invariant torus in phase space.
Physical Examples:
- Cycle $C_1$ (Rotating Field): Rotating the direction of the pulsed field while keeping strength constant results in a contractible loop ( $\nu=0$ ). No holonomy occurs.
- Cycle $C_2$ (Varying Strength): Monotonically varying the strength of the pulsed field by $2\pi$ results in a non-contractible loop ( $\nu$ is odd). This induces holonomy and reverses the rotation axis.

Significance and Claims
The paper claims to provide a precise classification of which adiabatic cycles alter the functional form of action variables in completely integrable Floquet systems.

Global Topology vs. Local Invariance: The work highlights that while action variables are locally adiabatic invariants, their global behavior is governed by the topology of the parameter space. The existence of nontrivial holonomy implies that the Hannay angle is not automatically well-defined for all cycles, as the condition of trivial holonomy is not satisfied for non-contractible loops.
Geometric Phase Analogy: The sign reversal of the rotation axis $\mathbf{n}$ is identified as a geometric phase factor, analogous to the Longuet-Higgins phase in the study of real eigenvectors. The authors note that this phenomenon is specific to Floquet systems where the control parameter is the rotation matrix $R$ (and $\mathbf{n}$ is an emergent eigenvector), contrasting with static magnetic fields where the direction is directly controllable and cannot form a closed cycle that reverses the field direction without passing through zero.
Distinction from Singular Processes: Unlike previous examples of action variable holonomy (e.g., in harmonic chains with clamps) that relied on singular processes involving changes in the number of degrees of freedom, this example relies on the intrinsic Floquet nature of the system, requiring no singularities.
Limitations: The authors modestly state that their homotopy classification is specific to the kicked-top map (Eq. 1) and that extending this argument to general completely integrable systems remains a subject for future work. They also note that while quantum counterparts exist (where populations are invariants but excitations can be induced), the classical topological formulation presented here is distinct from the fiber-bundle formulations used in quantum geometric phases.

More like this