Everything we publish, in the open.

We describe an architecture for a fault-tolerant machine composed of superconducting modules linked by cryogenic photonic interconnects. The design targets 10,000 logical qubits at an average code distance of 21, with a decoder that keeps pace with a 1 MHz syndrome cycle. We state the assumptions behind the thermal, optical, and classical-compute budgets explicitly, including the three assumptions we consider most likely to fail.

@misc{okonjo2026modular,
  title  = {A modular path to 10,000 logical qubits},
  author = {Okonjo, Amara and Marchetti, Ilya and Haddad, Tarek and Chen, Mei-Lin},
  year   = {2026},
  eprint = {2609.04122},
  archivePrefix = {arXiv},
  primaryClass  = {quant-ph},
  url    = {https://ruqilab.org/publications/#architecture-2026}
}

We report repeated logical-memory experiments on a 24-logical-qubit superconducting testbed. A decoder implemented on cryogenic FPGAs processes syndrome data within the 1.1 microsecond cycle, and logical error rates decrease by a factor of 2.1 per unit increase in code distance for distances 3, 5, and 7. We publish the full noise characterization, decoder source, and analysis notebooks, including the runs that did not reach threshold.

@article{okonjo2026below,
  title   = {Below-threshold surface-code memories with cryogenic real-time decoding},
  author  = {Okonjo, Amara and Marchetti, Ilya and Chen, Mei-Lin and Haddad, Tarek and Whitfield, Jonah and Ferreira, Lena},
  journal = {PRX Quantum},
  volume  = {7},
  pages   = {021014},
  year    = {2026},
  doi     = {10.1103/PRXQuantum.7.021014}
}

We estimate the quantum resources required to compute redox potentials for a family of iron–sulfur clusters relevant to grid-scale batteries. Under a 10^-3 chemical-accuracy target, our most favorable algorithm requires roughly 1,200 logical qubits and 10^9 T gates. We argue that early fault-tolerant machines will be useful for screening candidate electrolytes, not for replacing experimental characterization.

@article{reyes2026resource,
  title   = {Resource estimates for quantum simulation of iron--sulfur battery electrolytes},
  author  = {Reyes, Sofia and Okonjo, Amara and Raman, Priya and Sato, Hana},
  journal = {npj Quantum Information},
  volume  = {12},
  pages   = {88},
  year    = {2026},
  doi     = {10.1038/s41534-026-00881-4}
}

Control stacks are rarely compared on tasks that matter to users. We introduce a vendor-neutral benchmark of twelve tasks spanning pulse generation, calibration drift, readout, and recovery from component failure. Results from four laboratories show that advertised sampling rates explain little of the variation in end-to-end performance; documentation quality and calibration observability explain more.

@article{chen2026benchmark,
  title   = {An open benchmark for quantum control stacks},
  author  = {Chen, Mei-Lin and Kowalski, Nadia and Marchetti, Ilya and Whitfield, Jonah},
  journal = {Quantum},
  volume  = {10},
  pages   = {812},
  year    = {2026},
  doi     = {10.22331/q-2026-03-20-812}
}

A perspective on how public research funders should evaluate proposed quantum-simulation applications. We propose a four-part test covering the decision at stake, the classical baseline, the machine resources required, and the plan for public access. Applying the test to 31 proposals in our own portfolio, we find that most fail on the classical baseline or the access plan rather than on the physics.

@article{reyes2026public,
  title   = {Public-interest applications of early fault-tolerant quantum simulation},
  author  = {Reyes, Sofia and Okonjo, Amara and Raman, Priya and Park, Ellis},
  journal = {Quantum Science and Technology},
  volume  = {11},
  pages   = {034002},
  year    = {2026},
  doi     = {10.1088/2058-9565/adf012}
}

We tested a six-lesson quantum unit with 412 students across nine public schools. Classes that began with measurement statistics and interference patterns showed fewer persistent misconceptions than classes that began with the popular 'a qubit is both 0 and 1' framing. All lesson plans, worksheets, and assessment items are released under CC BY 4.0.

@article{park2026teaching,
  title   = {Teaching superposition without the hype: results from a classroom study},
  author  = {Park, Ellis and Kowalski, Nadia and Delgado, Rosa},
  journal = {The Physics Teacher},
  volume  = {64},
  pages   = {112},
  year    = {2026}
}

A practical guide for public agencies considering quantum computing procurement. We recommend task-based rather than qubit-count specifications, mandatory export of calibration data, source-available control software, and a public-benefit clause in every contract. Includes model language and a scoring rubric that has been used by two state agencies and one municipal utility.

@techreport{kowalski2025procurement,
  title       = {Procurement for the public option: buying quantum responsibly},
  author      = {Kowalski, Nadia and Park, Ellis and Okonjo, Amara},
  institution = {Ruqi Lab},
  type        = {Policy Brief},
  number      = {03},
  year        = {2025},
  month       = {November}
}

We field-tested a portable nitrogen-vacancy magnetometer at fourteen sites across three water districts. The instrument resolved subsurface magnetic contrasts associated with recharge channels at depths up to 38 meters, with a sensitivity of 12 pT/√Hz in an unshielded environment. We describe the mechanical design, the calibration protocol, and the three failure modes that dominated our maintenance hours.

@article{raman2025groundwater,
  title   = {Portable NV-diamond magnetometry for groundwater mapping},
  author  = {Raman, Priya and Whitfield, Jonah and Achebe, Daniel and Sato, Hana},
  journal = {Physical Review Applied},
  volume  = {24},
  pages   = {044031},
  year    = {2025},
  doi     = {10.1103/PhysRevApplied.24.044031}
}

We characterize a chip-to-chip photonic link operating between two 4 K stages of a dilution refrigerator. The packaged link adds 0.9 mW of heat per channel and maintains 98.2% fringe visibility after 400 thermal cycles. We release the package drawings and a failure log covering the eleven prototypes that did not survive cooldown.

@article{haddad2025cryogenic,
  title   = {Cryogenic photonic interconnects for modular superconducting processors},
  author  = {Haddad, Tarek and Chen, Mei-Lin and Whitfield, Jonah},
  journal = {Optica Quantum},
  volume  = {3},
  pages   = {455},
  year    = {2025},
  doi     = {10.1364/OPTICAQ.3.000455}
}

Quantum LDPC codes promise better overhead than the surface code but often assume long-range connectivity. We construct a family of sparse codes matched to a modular architecture with nearest-neighbor links inside modules and photonic links between them. Under a circuit-level noise model, the codes maintain a threshold of 0.62% while reducing physical-qubit overhead by 38% relative to a surface-code baseline.

@article{marchetti2025sparse,
  title   = {Sparse quantum LDPC codes for modular architectures},
  author  = {Marchetti, Ilya and Okonjo, Amara and Ferreira, Lena},
  journal = {Quantum},
  volume  = {9},
  pages   = {702},
  year    = {2025},
  doi     = {10.22331/q-2025-04-14-702}
}

We demonstrate a squeezed-light readout scheme for a compact rubidium clock operating outside a temperature-controlled laboratory. The scheme reduces quantum projection noise by 3.1 dB and improves short-term stability by 28% without increasing the optical bench footprint. We discuss the tradeoff between squeezing bandwidth and the vibration sensitivity of the field package.

@article{raman2024squeezed,
  title   = {Squeezed-light readout for compact atomic clocks},
  author  = {Raman, Priya and Sato, Hana},
  journal = {Physical Review Applied},
  volume  = {22},
  pages   = {064018},
  year    = {2024},
  doi     = {10.1103/PhysRevApplied.22.064018}
}

We propose a method for allocating error budgets across the components of a modular quantum computer, from gate fidelities and readout to inter-module link failures and classical decoding latency. Applied to a 10,000-logical-qubit design, the method identifies classical control and inter-module communication, not gate fidelity, as the binding constraints for three of five architecture families we study.

@article{okonjo2024error,
  title   = {Error budgets for modular quantum computers},
  author  = {Okonjo, Amara and Marchetti, Ilya and Chen, Mei-Lin},
  journal = {Physical Review Letters},
  volume  = {133},
  pages   = {190601},
  year    = {2024},
  doi     = {10.1103/PhysRevLett.133.190601}
}

Kestrel is an open-source control stack for superconducting quantum processors, covering pulse sequencing, FPGA firmware, calibration bookkeeping, and experiment orchestration. It is designed for reproducibility first: every experiment records the full calibration state, firmware version, and timing configuration needed to rerun it. The stack is in daily use on two processors and has been forked by three university groups.

@article{chen2024kestrel,
  title   = {Kestrel: an open control stack for superconducting qubits},
  author  = {Chen, Mei-Lin and Whitfield, Jonah and Kowalski, Nadia and Haddad, Tarek},
  journal = {Journal of Open Source Software},
  volume  = {9},
  number  = {98},
  pages   = {6120},
  year    = {2024},
  doi     = {10.21105/joss.06120}
}