Improved local models and new Bell inequalities via Frank-Wolfe algorithms

In the strange world of quantum mechanics, particles can become linked in a way that defies our everyday experience of cause and effect. When two particles are "entangled," measuring one instantly reveals something about the other, no matter how far apart they are. For decades, scientists have debated whether this spooky connection is a fundamental feature of nature or if it could be explained by hidden, classical rules we simply haven't discovered yet. This question led to the creation of "Bell inequalities," which act as a strict test to distinguish between the two possibilities. If the test is passed, the particles are behaving in a truly nonlocal, quantum way; if they fail, their behavior could still be explained by local, classical physics. A major puzzle in this field involves a specific type of mixed quantum state, known as a Werner state, which is a blend of a perfectly entangled pair and random noise. Scientists have long wanted to know the exact tipping point: how much noise can be added before the quantum connection becomes so weak that it can be mimicked by classical rules?

A team of researchers at the Zuse Institute Berlin has now sharpened our understanding of this tipping point by developing a new, highly efficient way to solve a complex mathematical problem. They focused on the challenge of determining whether a specific quantum correlation can be explained by a "local model"—a set of pre-agreed rules that two people could follow without communicating. If such a model exists, the system is local; if not, it is nonlocal. The researchers tackled this by treating the problem as a journey across a geometric landscape. Imagine trying to find the shortest path from a point outside a complex, multi-sided shape to the shape itself. The team used an advanced algorithm, originally designed for general optimization, to navigate this landscape much faster and more accurately than previous methods. By refining how they moved through this space, they were able to construct better local models and derive sharper tests for nonlocality.

The primary result of their work is a significantly narrowed gap between the known upper and lower limits for the noise threshold of two-qubit Werner states. For years, the scientific community has been stuck with a wide interval where the true answer was unknown. The researchers' new calculations push the lower bound—the point where we are certain the system is still local—higher than ever before, while simultaneously pushing the upper bound—the point where we are certain it is nonlocal—lower. This means the range of uncertainty has been squeezed from both sides. Their findings suggest that these quantum states remain local (explainable by classical rules) for a wider range of noise than previously thought, but they also prove that they eventually become nonlocal sooner than some earlier, less precise estimates suggested. The team provided exact, analytical numbers for these new limits, removing the guesswork that often accompanies computer simulations.

Beyond the specific numbers, the researchers demonstrated that their method is powerful enough to handle more complex scenarios involving three or more parties, not just two. They applied their technique to three-qubit systems, specifically looking at two famous types of entangled states known as GHZ and W states. While they established new bounds for both, the strict separation between the nonlocality thresholds of the tripartite GHZ and W states did not survive the necessary modifications to their calculations, as explicitly noted in the paper's erratum. This highlights the subtlety of distinguishing the resilience of different quantum entanglement types. The researchers also connected their findings to a famous mathematical constant known as the Grothendieck constant, which relates to how well certain mathematical inequalities hold up in different dimensions. By tightening the bounds on the quantum threshold, they effectively provided a more precise estimate for this constant as well.

To ensure their results were not just numerical approximations but rigorous proofs, the team went to great lengths to make their calculations entirely analytical. They used a specific type of algorithm that mitigates the "zig-zagging" behavior common in older methods, where the solution wavers back and forth without settling. Instead, their improved approach reduces this oscillation and moves more directly toward the solution, allowing them to construct a local model using a finite set of measurements that can be mathematically proven to work for an infinite set of possible measurements. They made their entire code available to the public as a software library, inviting other scientists to use these tools to explore further. The work does not claim to have solved the entire mystery of quantum nonlocality, but it has removed a significant layer of uncertainty, providing a clearer, more precise map of where the boundary between the classical and quantum worlds lies for these specific states.