High-fidelity collisional quantum gates with fermionic atoms
Article Meta
Article Date: 08 April 2026
Article URL: https://www.nature.com/articles/s41586-026-10356-3
Article Image: Figure 1
Summary
This paper reports high-fidelity collisional two-qubit gates using fermionic 6Li atoms trapped in an optical superlattice and imaged with a quantum gas microscope. The team realises spin-exchange and pair-tunnelling dynamics in double-well potentials and engineers discrete interaction pulses that implement SWAP_α and a composite pair-exchange (PX) gate. They measure entangling-gate fidelities up to 99.75(6)% across an array and demonstrate Bell-state coherence times exceeding 10 s. Spatial inhomogeneity of double-well frequencies is identified as the main remaining limitation; the authors propose mitigation strategies (echo pulses, potential flattening, optimal control) and outline routes to sub-10-μs gates and scaling to arrays of ~10,000 sites.
Key Points
- Platform: fermionic 6Li atoms in a two-dimensional optical superlattice with quantum gas microscopy for spin- and charge-resolved readout.
- Highest measured two-qubit entangling fidelity: 99.75(6)% for a √SWAP (entangling) pulse, extracted from repeated-pulse decay across 64 sites with post-selection on doublons.
- Long coherence: singlet–triplet (Bell) coherence lower bound >10 s, far longer than gate durations (~1.125 ms for √SWAP here).
- Dynamics: observed spin-exchange rates J/h ~3.32(3) kHz and record-high quality factors (J· au_ex/h ≈ 110 coherent oscillations in spin sector).
- Gate control: quasi-adiabatic Blackman pulses and intermediate-speed ramps suppress unwanted doublon excitations while keeping gates fast and robust.
- Pair-exchange (PX) gate: a composite Int–Z–Int sequence isolates correlated doublon tunnelling, enabling charge-sector √SWAP-like operations while leaving spin sector untouched.
- Main infidelity source: spatial inhomogeneity of double-well parameters (variation in oscillation frequency) leading to ensemble dephasing.
- Scalability prospects: with smaller lattice spacings and improved flattening, sub-10-µs entangling gates and arrays ~10^4 sites are predicted to be realistic.
- Applications: natural fit for fermionic digital quantum simulation and chemistry (fermion-number conserving gates, direct encoding of fermionic problems), and for hybrid analogue–digital schemes in materials science.
Context and relevance
Neutral-atom platforms have largely focused on Rydberg-mediated gates; this work revives collisional gates for fermions and shows they can reach fidelities comparable with leading solid-state two-qubit devices while offering inherent advantages for simulating fermionic systems. The use of quantum gas microscopy gives microscopic benchmarking of both spin and charge operations. Because these collisional gates conserve particle number and fermionic parity naturally, they are particularly attractive for quantum simulations and variational algorithms in quantum chemistry and materials modelling where fermionic structure matters.
Why should I read this?
Short version: if you care about high-fidelity two-qubit gates, fermionic encodings for chemistry or scalable quantum simulators, this paper is a proper game-changer. It shows collisional gates with 6Li can hit nearly 99.8% fidelity, preserve fermionic constraints natively, and give hugely long entangled-state lifetimes. Read it for the practical pulse sequences, the composite pair-exchange trick, and the clear roadmap to faster, larger arrays — saves you digging through a lot of dense methods yourself.
How they did it (concise)
They prepare near-doublon band-insulator states in a superlattice, split sites into double-wells, and control tunnelling t and on-site interaction U via lattice depths and a Feshbach field. By shaping ramps (Blackman/quasi-adiabatic) and tuning U/t (including a magic ratio), they decouple spin and charge sectors and implement SWAP_α gates. Quantum gas microscopy provides single-site, spin- and charge-resolved truth tables and time-domain oscillations used to extract J, decoherence times and gate fidelity. Composite Int–Z–Int sequences isolate pair-tunnelling for PX gates.
Limitations and next steps
- Current dominant error: spatial inhomogeneity across the addressed region; improving lattice flatness and beam profiles (DMD flattening) will raise fidelities further.
- State-preparation and residual inter-well tunnelling (limited y-lattice depth) reduce usable fraction of double-wells; better loading and deeper lattices will help.
- Authors propose echo sequences, optimal control pulses and smaller-lattice-spacing implementations to reach sub-10-µs gates and very large arrays.