Dispersive vs Elzerman readout fidelity comparison

Answer

Across four verified sources, Elzerman energy-selective tunneling readout consistently reaches higher peak single-shot fidelity than gate-based dispersive readout for silicon spin qubits, while dispersive readout wins on device simplicity and scalability -- no source disputes this trade-off. Kam & Hu (2024) report Elzerman-style readout via RF-SET reaching 97% in 1.5 us, 99% in 1.6 us and 99.9% in 6 us, against dispersive readout at up to 98% on-chip in 6 us or 73.3% to 98% off-chip in 0.3-2.6 ms cb48f581. Geng et al. (2025) push Elzerman-class latched spin readout in phosphorus-donor qubits to 99.44% fidelity in just 175 ns at millikelvin temperature, and note traditional Elzerman reaching up to 99.95%, both above the roughly 98% dispersive ceiling reported in the same paper b862a346. West et al. (2019, via Ciriano-Tejel-adjacent review) and Keith et al. (2019) show the same ordering at lower absolute numbers: 97% single-shot fidelity for Elzerman readout versus 73% for gate-based dispersive readout, with a hybrid gate-reflectometry approach proposed to combine both 20b90b83. De Michielis et al.'s 2023 review confirms the qualitative pattern -- Elzerman has historically achieved higher fidelities -- while noting that dispersive readout's compact, CMOS-compatible footprint makes it the preferred choice for industrial-scale fabrication despite the fidelity gap 72839095. Every source that requires Elzerman's higher fidelity also requires a separate charge sensor (SET) per qubit, which is the mechanism the corpus consistently cites as the scalability cost.

Summary

There is no disagreement in this corpus about direction: Elzerman readout wins on raw single-shot fidelity, dispersive readout wins on integration simplicity and scalability. What varies across sources is the size of the gap -- from 97% vs 73% in earlier work to 99.9% vs 98% in the most recent studies -- because dispersive readout has been improving faster than Elzerman readout in the reports collected here, narrowing but not closing the fidelity gap.

Evidence

• Kam & Hu, 2024 (npj Quantum Information) -- literature review plus new theory; Elzerman via RF-SET: 97-99.9% fidelity in 0.8-6 us; dispersive: 98% on-chip (6 us) or 73.3% to 98% off-chip (0.3-2.6 ms); proposed squeezed-microwave dispersive method reaches 98% in 0.6 us. cb48f581
• Geng et al., 2025 (Nature Communications) -- phosphorus-donor qubits; latched (Elzerman-class) spin readout: 99.44% in 175 ns at mK, 97.87% in 1.5 us at 3.7 K; traditional Elzerman up to 99.95% at mK; dispersive readout cited at up to 98% on-chip, 73-98% off-chip. b862a346
• West et al., 2019 / Keith et al., 2019 (via a comparative review) -- gate-based dispersive readout: 73% single-shot fidelity; Elzerman energy-selective tunneling: 97% single-shot fidelity; hybrid gate-reflectometry approach proposed for scalable processors. 20b90b83
• De Michielis et al., 2023 (review, silicon spin qubits to industrial fabrication) -- confirms Elzerman's historical fidelity advantage while identifying dispersive readout's compact CMOS-compatible footprint as the reason it is preferred for industrial-scale processors. 72839095

Methodology

All four submissions describe the same qualitative trade-off (Elzerman: higher fidelity, needs a separate SET; dispersive: lower fidelity, gate-integrated and scalable), so I compared rather than pooled their numbers -- they come from different qubit platforms (phosphorus donors vs. gate-defined quantum dots), different years, and different readout-circuit implementations (on-chip vs. off-chip resonators, RF-SET vs. latched readout), so no single percentage or timing figure can be treated as the definitive value for either method. I reported each source's own numbers separately and let the direction of agreement, not an averaged figure, carry the answer.

Limitations

• None of the four verified sources performs a single controlled experiment comparing dispersive and Elzerman readout on the same device, temperature, and integration time; every comparison in this corpus is a cross-study compilation, so the reported fidelity gap may partly reflect device-to-device variation rather than an intrinsic property of either readout mechanism.
• Fidelity and timing figures span a wide range within each method depending on chip generation, qubit platform and circuit design -- dispersive readout alone runs from 73.3% to 98% off-chip, and Elzerman-class readout runs from 97% up to 99.95% at its ceiling -- so a single "typical" number for either method is not supported by this corpus.
• The most favorable Elzerman numbers (99.44-99.95%) come from donor-qubit devices, while the dispersive comparisons draw on gate-defined quantum-dot devices; the corpus does not establish whether the fidelity gap holds within a single qubit platform.