⚛️ quantum physics

More global randomness from less random local gates

This paper demonstrates that one-dimensional structured random circuits utilizing non-Haar local gates can generate significantly more global randomness and achieve larger spectral gaps than their Haar random counterparts, a result derived by mapping the system's second-moment operators to the Kitaev chain to enable exact spectral analysis.

Original authors: Ryotaro Suzuki, Hosho Katsura, Yosuke Mitsuhashi, Tomohiro Soejima, Jens Eisert, Nobuyuki Yoshioka

Published 2026-08-19

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Ryotaro Suzuki, Hosho Katsura, Yosuke Mitsuhashi, Tomohiro Soejima, Jens Eisert, Nobuyuki Yoshioka

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, randomness is not just a lack of order; it is a powerful resource. Scientists use it to test how well quantum computers work, to verify that complex calculations are correct, and even to prove that a machine can do things classical computers cannot. To generate this randomness, researchers typically rely on a theoretical ideal called a "global random unitary." Imagine a machine that can instantly produce any possible configuration of a quantum system with perfect fairness, like rolling a die that has every number on it and landing on each with equal probability. While this concept is mathematically beautiful, it is practically impossible to build for anything but the tiniest systems. It requires an impossible amount of control and time.

Because the perfect ideal is out of reach, scientists use a workaround: they build circuits out of small, local random gates. These are like a series of simple, random steps that, when repeated enough times, eventually mimic the behavior of the perfect global randomness. For decades, the standard assumption was that the best way to do this was to use local gates that were themselves perfectly random, chosen from a vast, uniform pool of possibilities. The prevailing belief was that if you wanted the most chaotic, unpredictable global result, you had to start with the most chaotic, unpredictable local pieces. It seemed intuitive that the whole could not be more random than the sum of its parts.

A team of researchers has now challenged this intuition with a surprising discovery. They investigated a specific type of quantum circuit where the local gates are not perfectly random. Instead, these gates have a fixed, rigid structure at their core, surrounded by random elements. Think of a local gate as a sandwich: the bread slices are random, but the filling is a specific, unchanging ingredient. The researchers asked a simple question: could this structured, less-random local gate actually produce a more random global system than a gate that is completely random from start to finish?

The answer, they found, is yes. By analyzing the mathematical properties of these circuits, the team proved that under certain conditions, the structured circuits generate global randomness much faster and more effectively than their fully random counterparts. They focused on a specific measure of randomness related to how quickly a system settles into a state of total unpredictability. In their models, they identified a "solvable condition"—a precise mathematical relationship between how much a gate can entangle particles and how typical it is compared to a standard swap operation. When this condition was met, they could calculate the exact behavior of the system.

The results showed that if the fixed, structured gate has a slightly higher ability to entangle particles than a standard random gate, the entire circuit becomes more random. This is counterintuitive because one might expect that adding a fixed, non-random element would slow down the process of becoming chaotic. Instead, the structure acts as a catalyst. The researchers demonstrated that these structured circuits can reach a state of high randomness with fewer steps, or a shallower depth, than circuits built from purely random gates. They did not just suggest this; they derived the exact mathematical values that describe the speed of this randomness, proving that the spectral gap—a measure of how fast the system converges to randomness—can be larger for the structured circuits.

This finding has immediate practical implications for the future of quantum computing. One major application is in "randomized benchmarking," a technique used to measure the error rates of quantum computers. Currently, this process requires running very deep circuits, which takes time and increases the chance of errors. By using these more efficient structured circuits, the depth required to get a reliable measurement could be significantly reduced. Similarly, the generation of "unitary 2-designs," which are specific types of random ensembles used for various quantum tasks, could be achieved much faster. The researchers showed that by carefully choosing the fixed component of the local gate, engineers could design quantum circuits that are not only simpler to build but also more powerful at generating the randomness needed for advanced computations.

The study also clarified the limits of this phenomenon. The advantage is not universal; it depends on the specific properties of the fixed gate. If the gate does not meet the solvable condition or if its entangling power is too low, the benefit disappears. In some cases, increasing the entangling power beyond a certain point can actually reduce the efficiency, showing that there is a delicate balance to be struck. The researchers mapped out these relationships, showing exactly where the structured circuits outperform the random ones and where they do not. They confirmed their analytical results with numerical simulations, ensuring that the theoretical predictions held up in practice.

Ultimately, this work suggests that in the quantum realm, a little bit of order can create more chaos than total disorder. By moving away from the assumption that local gates must be perfectly random, the researchers have opened a new path for designing quantum circuits. They have shown that the most effective way to generate global randomness is not always to use the most random local tools, but to use tools with a specific, optimized structure. This insight could help engineers build better quantum computers, using fewer resources to achieve the same, or even better, results. The study provides a concrete blueprint for how to engineer randomness, turning a theoretical curiosity into a practical guide for the next generation of quantum technology.

Title: More global randomness from less random local gates

Problem Statement
Random unitary ensembles are fundamental tools in quantum information science, utilized for error estimation via randomized benchmarking, demonstrating quantum computational advantage, and verifying quantum states through tomography. While the global Haar random unitary ensemble is the theoretical standard for randomness, it is practically infeasible to generate for intermediate-sized systems. Consequently, researchers rely on sequences of local random circuits to approximate these ensembles.

A central question remains: Do local Haar random gates constitute the optimal mechanism for achieving global randomness? Existing literature predominantly assumes that local randomizers must consist of local Haar random ensembles or unitary subgroups. However, it is unclear whether this assumption is optimal or if fixing certain gates to be non-Haar (while maintaining a specific structure) could yield superior global randomness. This work addresses whether structured random circuits with non-Haar local gates can outperform standard Haar random circuits in generating global randomness.

Methodology
The authors investigate one-dimensional structured random circuits where the local gates are not drawn from the Haar measure but are "u-structured." A u-structured gate is defined as a fixed two-qudit gate $u$ sandwiched between single-qudit Haar random gates acting on the inputs and outputs. Two circuit architectures are analyzed:

Local Random Circuits (LRC): Neighboring pairs of qudits are chosen uniformly at random, and a u-structured gate is applied.
Brick-Wall Circuits (BWC): Staggered layers of u-structured two-qudit gates are applied to nearest neighbors.

To quantify randomness, the authors utilize the second-moment operator $M_\nu$ , defined as the integral of $U^{\otimes 2} \otimes \bar{U}^{\otimes 2}$ over the probability distribution $\nu$ of the circuit. The convergence rate to a unitary 2-design is determined by the spectral gap of this operator ( $\Delta_\nu = 1 - |\lambda_3|$ ). The authors define "more randomness" in two ways:

Spectral Gap: $\nu$ is more random than $\mu$ if $\Delta_\nu > \Delta_\mu$ .
Positivity: $\nu$ is more random than $\mu$ if the operator $M_\mu|_{V_1^\perp} - M_\nu|_{V_1^\perp}$ is positive, where $V_1$ is the eigenspace corresponding to the largest eigenvalue (spanned by identity and SWAP gates).

The analysis relies on characterizing the second-moment operators using two parameters: entangling power ( $e_u$ ) and gate typicality ( $g_u$ ). The authors identify a solvable condition under which the moment operators can be exactly diagonalized:
$\frac{e_u}{g_u} = \frac{e_H}{g_H}$
where $e_H$ and $g_H$ are the averaged entangling power and gate typicality of Haar random gates. Under this condition, the moment operators map to free-fermion chains (specifically the Kitaev chain for local circuits), allowing for the exact calculation of all eigenvalues and eigenvectors.

Key Contributions and Results

Exact Diagonalization: The authors derive all exact eigenvalues and eigenvectors for the second-moment operators of both local and brick-wall structured random circuits under the solvable condition. This is achieved by mapping the operators to free-fermion models via the Jordan-Wigner transformation.
- For Local Random Circuits, the eigenvalues are given by $\lambda_{p, i_p} = 1 - \frac{1}{n} \sum_{k \in K_p} \epsilon_k i_{p,k}$ , where $\epsilon_k$ depends on the normalized entangling power.
- For Brick-Wall Circuits, the eigenvalues are products of terms $\lambda_k$ , which can be complex if $e_u > e_H$ .
Counterintuitive Superiority of Non-Haar Gates: The paper demonstrates that structured random circuits can be substantially more random than Haar random circuits with the same architecture.
- Local Circuits: If $e_u > e_H$ and $g_u > g_H$ , the structured circuit is more random with respect to both spectral gap and positivity. The spectral gap increases monotonically with $e_u$ and $g_u$ .
- Brick-Wall Circuits: Under the solvable condition, the structured circuit is more random than the Haar circuit if and only if $e_u > e_H$ . Notably, the spectral gap does not always increase monotonically with $e_u$ in brick-wall circuits (e.g., in qutrit systems, the gap can decrease for certain $e_u$ values), highlighting a complex dependence on architecture.
Spectral Gap Analysis: The authors show that the entangling power of Haar random gates is not maximal. For qubits ( $d=2$ ), $e_H = 3/5$ , while structured gates can achieve up to $e_u = 2/3$ . By operating in this regime ( $e_u > e_H$ ), structured circuits achieve larger spectral gaps, implying faster convergence to unitary 2-designs.
Numerical Verification: Numerical calculations of $|\lambda_3|$ for qubit ( $d=2$ ) and qutrit ( $d=3$ ) systems confirm the analytical results, showing that the spectral gap exceeds that of Haar points when $e_u > e_H$ .

Significance and Applications
The paper claims that these findings have direct applications in improving the efficiency of quantum information protocols:

Randomized Benchmarking (RB): Since the circuit depth required for RB is upper-bounded by the inverse of the spectral gap, the enhanced gaps of structured circuits suggest that the required circuit depth for theoretically guaranteed RB can be reduced.
Approximate Unitary 2-Designs: The results offer improved bounds on the depth required to generate approximate unitary 2-designs in shallow circuits. By combining the enhanced randomness of structured gates with techniques for gluing small random circuits, the convergence time to 2-designs can be improved.
Quantum Chaotic Dynamics: The exact diagonalization of the second-moment operators allows for the derivation of closed-form expressions for indicators of quantum chaos, such as out-of-time ordered correlators (OTOCs) and Rényi-2 operator entanglement entropy, within structured random circuits satisfying the solvable condition.

The authors conclude that their work provides an affirmative answer to whether non-Haar local gates can generate more global randomness than Haar gates, suggesting that additional structure can improve convergence rates to unitary 2-designs, a crucial factor for large-scale quantum computing. They note that while they focused on second-moment operators, the results regarding positivity hold for arbitrary spatial dimensions in local circuits.

More like this