Simulating turbulent air flow accurately is vital for weather forecasts
EUMETSAT/ESA
Researchers have developed quantum-inspired algorithms that can simulate turbulent fluid flows on a classical computer much faster than existing tools, significantly reducing computation times from several days on a large supercomputer to just hours on a regular laptop. This breakthrough has the potential to enhance weather forecasts and optimize industrial processes.
Turbulence in liquid or air involves complex interactions of eddies that lead to chaotic patterns, making precise simulation challenging even for the most powerful computers. While quantum computers hold promise in improving simulations, current machines are limited in their capabilities beyond basic demonstrations.
By utilizing quantum computer-inspired algorithms known as tensor networks, researchers like Nikita Gourianov at the University of Oxford have found a novel approach to represent turbulence probability distributions. Tensor networks, originally developed in physics and popularized in the early 2000s, offer a pathway to enhance performance on classical computers before fully functional quantum machines become available.
Gourianov explains, “The algorithms and the way of thinking come from the world of quantum simulation, and these algorithms closely resemble quantum computer operations. We have observed a significant speed-up, both theoretically and practically.”
The team managed to run a simulation on a laptop in just a few hours, a task that previously took several days on a supercomputer. This new algorithm led to a 1000-fold reduction in processor demand and a million-fold reduction in memory demand. While the initial simulation was a basic test, similar problems on a larger scale are crucial for weather forecasting, aircraft aerodynamics, and industrial chemical process analysis.
Gunnar Möller from the University of Kent highlights the complexity of the turbulence problem, particularly with data in five dimensions, making traditional computations nearly impossible. He praises tensor networks for significantly reducing computational power required for simulations by streamlining data processing.
Tensor networks have been instrumental in the ongoing competition between quantum computer developers and classical computer scientists. In a notable instance, Google’s quantum processor Sycamore achieved “quantum supremacy” in 2019, only to be surpassed by tensor networks running on conventional graphics processing units in just over 14 seconds, challenging Google’s claim. Despite the advancements in quantum computing, researchers are optimistic about the potential of tensor networks in maximizing existing computational resources.
While the future holds the promise of large-scale fault-tolerant quantum computers for more precise simulations, Möller is excited about the immediate impact of algorithms like tensor networks. He emphasizes, “With a laptop, the authors of this paper could outperform what’s possible on a supercomputer, solely due to a smarter algorithm. The benefits are tremendous, and we don’t have to wait for the perfect quantum computer to see results.”
