🔢 mathematics

Dark Subspaces and Invariant Measures of Quantum Trajectories

This paper establishes a complete classification of invariant measures for quantum trajectories by proving the uniqueness of measures on dark subspaces and characterizing the ergodic measures within them as a set indexed by unitary group orbits.

Original authors: Tristan Benoist, Clément Pellegrini, Anna Szczepanek

Published 2026-10-16

📖 8 min read🧠 Deep dive

CC BY 4.0

Original authors: Tristan Benoist, Clément Pellegrini, Anna Szczepanek

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 authors. For technical accuracy, refer to the original paper. Read full disclaimer

In the quantum world, the act of measurement is not a passive observation but a physical interaction that fundamentally alters the system being watched. When a quantum system, such as a single atom or a photon, is subjected to a series of indirect measurements, its state does not follow a single, predictable path. Instead, it evolves through a branching sequence of possibilities, a stochastic journey known as a quantum trajectory. These trajectories are the mathematical descriptions of what happens to a quantum system as it is constantly probed by its environment. For decades, physicists have understood that under certain conditions, these chaotic paths eventually settle into a stable, predictable pattern, much like a spinning top that wobbles wildly before finally standing still. However, the rules governing this settling process were thought to be quite strict, requiring the system to lose its quantum complexity and collapse into a simple, single state.

A team of researchers has now rewritten the rules for how these systems behave, showing that the previous understanding was too narrow. They have discovered that quantum trajectories do not always collapse into a single, simple state. Instead, they can settle into a much richer variety of stable patterns, depending on the specific structure of the measurements being performed. The researchers have provided a complete map of all the possible ways these systems can stabilize, proving that the final state is determined by a hidden geometric structure within the system's evolution. This work resolves a long-standing question about the diversity of quantum behaviors and offers a precise classification of every possible outcome, revealing that the universe of quantum measurements is far more varied than previously imagined.

The study focuses on a specific type of quantum system where the measurements are indirect, meaning the system is not observed directly but rather through its interaction with a measuring device. The researchers modeled the system as a Markov process, a mathematical framework where the next step depends only on the current state, not on the history of how it got there. In this context, the system moves through a space of possible states, guided by a set of random matrices that represent the measurement outcomes. The central question was: what happens to this system after a very long time? Does it wander forever, or does it settle down? If it settles, what does that final state look like?

For many years, the prevailing theory assumed that these systems would always undergo a process called purification. This meant that no matter the starting point, the system would eventually be forced into a single, simple state, effectively losing all its quantum complexity. This assumption worked well for many known cases, but the researchers suspected it was not the whole story. They set out to test what would happen if they removed this assumption of purification. By doing so, they uncovered a more complex reality where the system could settle into a "dark subspace." A dark subspace is a specific region of the system's state space where the measurements, instead of destroying the quantum information, act like a rotation. In this region, the system can evolve in a way that preserves its complexity, moving around within the subspace without ever collapsing into a single point.

The researchers proved that these dark subspaces are not just theoretical curiosities but are the key to understanding the long-term behavior of the system. They showed that the system first drifts toward these dark subspaces and then performs a random walk between them. Once inside a dark subspace, the system's evolution is governed by a group of symmetries, which are transformations that leave the system's essential structure unchanged. The researchers identified that the final, stable state of the system is determined by the specific symmetries of the measurements. They found that the set of all possible stable states is in one-to-one correspondence with the orbits of a specific group of rotations acting on a reference space.

To make this precise, the team introduced a new mathematical tool: a minimal family of isometries. An isometry is a mapping that preserves distances, and in this context, it serves as a bridge between the different dark subspaces and a single, common reference space. By choosing this bridge in a specific, optimal way, the researchers could translate the complex, high-dimensional motion of the quantum system into a simpler, lower-dimensional problem. They demonstrated that the set of all possible stable states, or invariant measures, is exactly the set of states generated by the action of a specific group of rotations on this reference space. This group is determined by the specific matrices used in the measurements.

The most significant finding is that the number and type of stable states depend entirely on the structure of this group of rotations. If the group is large enough to cover the entire reference space, there is only one unique stable state, and the system behaves as if it has purified. However, if the group is smaller, the system can settle into many different stable states, each corresponding to a different orbit of the group. The researchers provided a complete classification of these possibilities, showing that the system can have a unique stable state, or it can have a continuous family of them, or even a discrete set of them, depending on the specific properties of the measurement matrices. They also identified the precise conditions under which the system has a unique stable state, linking it to the mathematical property of transitivity, where the group of rotations can move any point in the reference space to any other point.

This work does not just describe a single experiment but provides a general framework for understanding a wide class of quantum systems. The researchers proved that their classification is exhaustive, meaning there are no other possible stable states outside of the ones they described. They also showed that the convergence to these stable states is exponential, meaning the system settles down very quickly. This rapid convergence is a crucial feature for practical applications, as it suggests that quantum systems can be reliably controlled and stabilized. The results also have implications for quantum error correction, a field dedicated to protecting quantum information from noise. The dark subspaces identified in this study are natural candidates for storing quantum information because the measurements that define them preserve the information rather than destroying it.

The researchers' approach was rigorous and mathematical, relying on theorems and proofs rather than simulations or approximations. They established the existence and uniqueness of the stable states on the dark subspaces first, and then used this foundation to build the complete classification for the entire system. Their proof involved showing that the system's behavior is equicontinuous, a technical property that ensures the system's evolution is smooth and predictable in the long run. This allowed them to apply powerful mathematical theorems about the relationship between the system's geometry and its stable states. The result is a definitive answer to the question of what happens to a quantum system under repeated indirect measurements.

The paper also addresses the question of whether the choice of mathematical tools used to describe the system affects the physical outcome. The researchers showed that while different choices of the isometry bridge can lead to different-looking mathematical descriptions, the set of physical stable states remains the same. This confirms that their classification is a property of the physical system itself, not an artifact of the mathematical method used to analyze it. They provided examples to illustrate the different scenarios, including cases where the system has a unique stable state, cases where it has a continuous family of states, and cases where it has a finite set of discrete states. These examples serve to ground the abstract theory in concrete, understandable situations.

In summary, the researchers have lifted a long-standing restriction on the theory of quantum trajectories. By removing the assumption of purification, they revealed a much richer landscape of possible behaviors. They have shown that the long-term fate of a quantum system is determined by the symmetries of the measurements it undergoes. If the measurements are sufficiently complex, the system can settle into a single state. If they are less complex, the system can settle into a variety of states, each preserving a different aspect of the system's quantum nature. This work provides a complete and rigorous map of the possible outcomes, offering a deeper understanding of how quantum systems interact with their environment and how they can be controlled. The findings are not just a theoretical curiosity but a fundamental step forward in the mathematical understanding of quantum measurement and control.

Technical Summary: Dark Subspaces and Invariant Measures of Quantum Trajectories

Problem Statement
The paper addresses the classification of invariant measures for quantum trajectories, which are Markov processes describing the evolution of a quantum system subject to indirect measurements. These trajectories can be modeled as place-dependent iterated function systems or products of dependent, non-identically distributed random matrices acting on the projective space $\mathbb{P}(\mathbb{C}^d)$ .

Previous literature, specifically [Ben+19], established the uniqueness of the invariant measure under a "purification" assumption (equivalent to the absence of dark subspaces of dimension greater than one) and strong irreducibility. However, without purification, the uniqueness of the invariant measure is not guaranteed, and the set of invariant measures can be complex. The authors aim to provide a complete classification of the set of $\Pi$ -invariant measures for quantum trajectories under the standard assumption of irreducibility, lifting the purification constraint.

Methodology
The authors employ a two-step strategy combining probabilistic analysis of Markov chains on Grassmannians with group-theoretic analysis of isometries:

Convergence to Dark Subspaces:
- The authors utilize the concept of dark subspaces (introduced in [MK06]), defined as subspaces $D \subset \mathbb{C}^d$ where the action of any product of matrices from the support of the measurement measure $\mu$ is proportional to a unitary transformation.
- They define a Markov chain $(D_n)_n$ on the set of maximal dark subspaces $\mathcal{D}_m$ (those of maximal dimension $r_m$ ).
- Using techniques similar to [Ben+19] but adapted for $r_m \geq 1$ , they prove that this chain admits a unique invariant measure $\chi_{\text{inv}}$ and converges to it exponentially fast in the Wasserstein metric.
- They demonstrate that the original quantum trajectory $(\hat{x}_n)_n$ in $\mathbb{P}(\mathbb{C}^d)$ converges exponentially fast to the union of projective spaces associated with $\mathcal{D}_m$ .
Classification via Minimal Families of Isometries:
- To characterize the dynamics inside the dark subspaces, the authors introduce a reference space $\mathbb{C}^{r_m}$ and a family of linear isometries $J = \{J_D: \mathbb{C}^{r_m} \to D\}_{D \in \mathcal{D}_m}$ .
- These isometries map the random evolution between dark subspaces to a random walk of unitary operators on the reference space $\mathbb{C}^{r_m}$ . This generates a closed subgroup $G_J \subset SU(r_m)$ .
- A crucial technical contribution is the definition of minimal families of isometries. A family $J$ is minimal if the induced group $G_J$ is minimal (up to unitary equivalence) among all possible families. The authors prove that minimality is equivalent to the property of being a "smart family" (constructed via specific products of matrices from the support of $\mu$ ).
- Using the theory of equicontinuous Markov kernels (Raugi's theorem), they establish a one-to-one correspondence between $\Pi$ -minimal subsets of $\mathbb{P}(\mathbb{C}^d)$ and the supports of measures constructed from the orbits of $G_J$ .

Key Contributions and Results

Theorem 1.1 (Uniqueness on Dark Subspaces): Assuming $\mu$ is irreducible, the Markov chain on the set of maximal dark subspaces $\mathcal{D}_m$ admits a unique invariant Borel probability measure $\chi_{\text{inv}}$ , and convergence to this measure is exponential.
Theorem 1.2 (Classification of Ergodic Measures): Let $J$ $J$ be a minimal family of isometries. The set of $\Pi$ $Π$ -ergodic measures on $\mathbb{P}(\mathbb{C}^d)$ $P (C^{d})$ is exactly the set $\Upsilon_J = \{\nu_{\hat{x}, J} : \hat{x} \in \mathbb{P}(\mathbb{C}^{r_m})\}$ $ΥJ={νx^,J:x^∈P(Crm)}$ .
- Here, $\nu_{\hat{x}, J}$ is the image measure of the product $\chi_{\text{inv}} \otimes m_{\hat{x}, J}$ under the map $\Psi_J(D, \hat{z}) = J_D \cdot \hat{z}$ , where $m_{\hat{x}, J}$ is the uniform measure on the $G_J$ -orbit of $\hat{x}$ .
- This result implies that any invariant measure is a convex combination of these ergodic measures.
Theorem 1.3 (Uniqueness of Global Invariant Measure): There exists a unique $\Pi$ $Π$ -invariant Borel probability measure on $\mathbb{P}(\mathbb{C}^d)$ $P (C^{d})$ if and only if the minimal group $G_{\min}$ $G_{m i n}$ acts transitively on $\mathbb{P}(\mathbb{C}^{r_m})$ $P (C^{r_{m}})$ .
- If $r_m$ is odd, this requires $G_{\min} = SU(r_m)$ .
- If $r_m$ is even, this requires $G_{\min} = SU(r_m)$ or the identity component of $G_{\min}$ to be unitarily equivalent to the compact symplectic group $Sp(r_m/2)$ .
Equicontinuity: The paper proves that the transition kernel $\Pi$ is equicontinuous, a necessary condition for applying Raugi's theorem to link minimal sets with ergodic measures.

Significance and Claims
The paper claims to provide a complete classification of the invariant measures for quantum trajectories under the standard irreducibility assumption, removing the restrictive "purification" assumption required in prior work ([Ben+19]).

Generalization: It extends the understanding of quantum trajectories to cases where the system does not purify (i.e., $r_m > 1$ ), a scenario relevant for quantum error correction and systems with high-dimensional dark subspaces.
Structural Insight: By introducing the concept of minimal families of isometries, the authors resolve the ambiguity of how to represent the internal dynamics of dark subspaces. They show that while the choice of isometries affects the induced group $G_J$ , the set of ergodic measures is invariant provided the family is minimal.
Necessary and Sufficient Conditions: The paper provides precise algebraic conditions (transitivity of the minimal group action) for the uniqueness of the invariant measure, linking the probabilistic behavior of the trajectory to the representation theory of the underlying unitary group.

The authors note that their results rely on the irreducibility of the measure $\mu$ and that the classification is exhaustive for the set of invariant measures on pure states (projective space). They also provide an extension of these results to mixed states (density matrices) in an appendix. The work is presented as a rigorous mathematical characterization, with examples illustrating how different choices of measurement operators lead to distinct structures of invariant measures (e.g., finite vs. continuous sets of ergodic measures).

More like this