Prospects for early fault tolerant quantum computing with neutral atoms
 
A Harvard-led experiment with up to 48 logical qubits is the most eye-catching demonstration of quantum error correcting codes to date. The involvement of quantum startup QuEra Computing in this work is suggestive of the inspiration it is likely to play for that company's much anticipated roadmap update due early in 2024. However this is not yet a Sputnik moment for fault tolerant quantum computing.
 
Powerful progress with neutral atoms
The Harvard-led experiment also included collaboration from MIT, QuICS and QuEra. A number of eye-catching demonstrations were made: [1]
280 physical qubits were used to form up to 48 logical qubits
Flexible qubit shuttling allowed demonstrations of operations in multiple quantum error correction codes (2D surface code, 2D color code and a small 3D color code).
Up to 200 2Q logical gates were performed (with postselection)
Logical circuits were demonstrated with a performance improvement over the equivalent unencoded circuits (i.e. specifically an improved linear cross-entropy benchmark).
 
To understand the overall quantum computing architecture this work points toward, it is important to see it in the context of the unified body of work this team has been developing:
 
2017+ Development of MOT, SLM and AOD technology, initially for analog quantum simulators based on Rydberg states of neutral atoms  [2], [3].
2019 Demonstration of gate-model style operations (2Q fidelity 97.4%) [4]
2022 Refinement of the Raman optical system required to drive hyperfine qubits [5]. 
2022 Development of moving atom architecture with zoned operation [6]
2023 2Q gates of high fidelity (99.5%) [7]
2023 Blueprint for low-overhead error-corrected quantum zoned architecture: Q LDPC codes to protect memory qubits; topological codes to protect computational qubits; mediating ancillae qubits [8].
 
This work has been supported by multiple research grants from DARPA, NSF, AFRL and investment and support from QuEra.

 
Taken together the Harvard scheme is a true architectural proposal. It is designed to play to neutral atoms strengths. 
Memory qubits based on long-lived hyperfine states
(with the future potential to be further protected by resource efficient Q LDPC codes)
Rydberg mediated gates protected by well understood topological codes
(e.g. surface codes, color codes).
A true zoned processor architecture enabled by AOD technology.


 
The Harvard architecture features the ability to flexibly shuttle ‘tiles’ of physical qubits, and to excite these atoms into Rydberg states only when required for 2Q gate operations. It therefore  excels at transverse gate operations (where each physical qubit in the logical qubit tile interacts only with the equivalent physical qubit in the other logical qubit or ancilla tile). However, it also has limitations. Atom loading, shuttling and readout operations are slow. Further improvements in 2Q gate fidelity and module size are desirable. How far this bulk optic approach truly can scale remains to be demonstrated.
 
Quantum thought leaders John Preskill and Scott Aaronson have already provided their initial reactions to this work (here and here). And it was debated by a wide expert group in the QEC panel at Q2B Silicon Valley.  [9], [10]

This report will explore how these strengths and weaknesses promise to play out, as well as the prospects for moving this technology from an academic to commercial context.