Quantum optimization with globally driven neutral atom arrays
August 13, 2026
Innsbruck, 13.08.2026 – A group of researchers at ParityQC and the University of Innsbruck has demonstrated quantum optimization on neutral atom arrays using strictly global driving fields, removing the need for local control of individual atoms, one of the major experimental bottlenecks for solving optimization problems on this hardware platform. The result was achieved on a Pasqal Orion Alpha/Fresnel quantum annealer and validates every standard building block of the approach, up to a fully connected optimization problem solved on a 43-atom array.
The challenge: local addressing is experimentally demanding
Neutral atoms in configurable tweezer arrays are a promising platform for solving hard optimization problems. These platforms natively support a specific type of problem, unit-disk maximum weight independent set (UD-MWIS), while more general problem classes require an embedding that maps them onto UD-MWIS instances. Implementing the asymmetric weights that these embeddings demand has so far required local control of the detunings at each atom, typically realized through tightly focused laser beams. This site-dependent addressing remains experimentally challenging and limits the scalability of neutral atom quantum annealers.
An ‘anchor’ approach to global driving
In the latest publication “Quantum optimization with globally driven neutral atom arrays”, a group of researchers at ParityQC and the University of Innsbruck, Martin Lanthaler, Kilian Ender, Davit Khachatryan, Pedro Ildefonso, Andrew Byun, Clemens Dlaska, Michael Schuler, and Wolfgang Lechner, introduces a method that encodes optimization problems using only a single global detuning. Key features of the proposed approach include:
- Anchor atoms instead of local fields: The work introduces precisely placed auxiliary “anchor” atoms that weakly interact with target atoms. This interaction is too weak to induce Rydberg blockade, but strong enough to induce an energy shift that acts as an effective local detuning. The required weights are encoded in the positions of the anchor atoms rather than through locally controlled laser detuning.
- A natural fit with the ParityQC Architecture: The Parity encoding translates the requirement for arbitrary detuning into a structured pattern that can be precisely engineered using anchor atoms. Newly developed homogenized gadgets shift all necessary detuning inhomogeneities to the ports, making the parity-based embedding particularly well suited to this weak-interaction approach.
- Experimental validation on hardware: The approach was benchmarked on a Pasqal Orion Alpha/Fresnel machine across all standard building blocks. The researchers then solved quadratic unconstrained binary optimization (QUBO) problems on four-node all-to-all connected graphs, embedded in a 43-atom array using seven precisely placed anchor atoms.
Beyond the ParityQC framework
By removing the requirement for local addressing, this work reduces experimental demands and simplifies the path toward scaling up to larger devices. The concept can be readily integrated into existing platforms and, while demonstrated here with the ParityQC framework, opens a direction that could be extended to other embeddings and to three-dimensional layouts.
Read the preprint
The preprint “Quantum optimization with globally driven neutral atom arrays”, authored by Martin Lanthaler, Kilian Ender, Davit Khachatryan, Pedro Ildefonso, Andrew Byun, Clemens Dlaska, Michael Schuler, and Wolfgang Lechner, is now available here.
FAQ
Neutral atom arrays natively encode unit-disk maximum weight independent set (UD-MWIS) problems. To solve more general optimization problems, the weights of the problem have to be implemented as local detunings on individual atoms, which requires site-dependent laser control. This is experimentally demanding and limits scalability. The paper demonstrates how to encode and solve these problems using only a single global detuning, eliminating the need for local field control.
An anchor is an additional, precisely placed atom that interacts weakly with a target atom in the layout. It is positioned outside the blockade radius, so it does not trigger Rydberg blockade, but the van der Waals interaction it induces still shifts the energy of the target atom. This shift acts as an effective local detuning. Because the strength of the interaction depends on the distance between the anchor and the target, the required detuning can be programmed simply by placing the anchor at the right position.
Encoding fully arbitrary detuning patterns with global driving is not possible in general, but it is also not necessary. The Parity encoding maps a problem into an intermediate representation whose detuning requirements follow a structured pattern. Using homogenized gadgets, all necessary detuning inhomogeneities can be shifted onto the ports and LINKs of the layout, where there is ample space to place the anchors. This structural fit is what makes the parity-based embedding particularly suitable for the weak-interaction approach.
These are the recurring building blocks used to embed a parity-encoded problem onto the atom array. LINK gadgets act as copy gadgets that wire variables together and carry logical information non-locally along their length. The 3BODY gadget enforces a three-body parity constraint, the 4BODY gadget is built by combining two 3BODY gadgets via a shared LINK, and the KITE gadget arises from combining two 3BODY gadgets along their baselines. Each was benchmarked individually before being combined into the full layout.
The approach was validated on a Pasqal Orion Alpha/Fresnel neutral atom quantum annealer. The researchers first benchmarked all individual building blocks, confirming that anchor positioning reliably induces the required detunings. They then integrated these elements into a 43-atom layout with seven anchors and solved a fully connected QUBO problem on four variables, across eight distinct problem instances representing different degrees of frustration, using purely global driving.
Several noise sources affect the hardware, including readout noise, Rydberg decay, and atom position fluctuations. The work applies readout error mitigation for the individual gadget experiments and a dedicated post-processing method for the larger layout that exploits Rydberg blockade information to correct residual bit-flip errors. The researchers also correct for systematic distance shifts caused by radial position fluctuations during anchor placement.
Local addressing through tightly focused laser beams is one of the main experimental bottlenecks for optimization on neutral atom hardware. Encoding the problem in the geometry of the array instead means the same result can be achieved with a much simpler control setup. This reduces experimental overhead and provides a viable, scalable pathway for hardware limited to global detuning control.
Yes. While this work used the Parity-based embedding and two-dimensional layouts, the authors note the approach extends naturally to three-dimensional arrays, where anchor placement is even less restricted. It could also be combined with local-detuning approaches where hardware allows.
