A Digitally Controlled Silicon Quantum Processing Unit
Peer-reviewed Nature 2026; open access, 2026 · paper · Physics
Presents a silicon quantum processing unit combining a cryogenic CMOS controller, superconducting ribbon cable and a 54-quantum-dot exchange-only qubit chip. It reports qubit performance an order of magnitude better than prior exchange-only work and runs small error-correcting and error-detecting codes.
Checked: PDF from nature.com, fetched 5 Oct 2026. Page numbers are those of that PDF.
4 findings
Substance Text: "finding FL = 0.95 largely independent of the initial state (Fig. 4e)", but the experimental fidelity points in Fig. 4e are at about 0.960, 0.966, 0.966 and 0.963 (mean about 0.964); none is at 0.95.
- p. 4 Quantum error detection, 4th paragraph
ing FL = 0.95 largely independent of the initial state (Fig. 4e)
- p. 5 Fig. 4e, Fidelity panel (black 'Experiment' markers; axis ticks 0.96, 0.98, 1.00)
0.96
- p. 4 Quantum error detection, paragraph on code performance
finding FL = 0.95 largely independent of the initial state (Fig. 4e)
- p. 5 Fig. 4e, fidelity panel
Fidelity
Only the authors can say what is right here.
Judge's note: Rendered Fig. 4e at 600 dpi from the PDF. Fidelity axis: 0.96 tick at y=632 px, 0.98 at y=518 px (114 px per 0.02); the panel floor is at about 0.950. Black experiment markers are at y=632, 597, 596, 616, i.e. 0.960, 0.966, 0.966, 0.963, with error bars of about +/-0.004 (lowest bar end about 0.956). The caption defines the plotted quantity as the logical fidelity of the post-selected state after 3 rounds, the same quantity as in the text, and I found no statement giving a different definition for FL. The values are read off a plot (about +/-0.001), but the gap to 0.95 is well beyond reading error; the text would be expected to say 0.96. Authors should say whether the text or the figure is right.
Second check: another model agreed
Report a problem with this finding
Reference or label Fig. 4 caption, panel b: "We sweep the number of syndrome extraction rounds per shot in the [5, 1, 5] sequence quasi-logarithmically from 1 to 200", but the last plotted points in Fig. 4b are at about 150 rounds and the x axis ends just short of 200.
- p. 5 Fig. 4 caption, panel b
logarithmically from 1 to 200 and repeat each 50,000 times
- p. 5 Fig. 4b, x axis 'Number of rounds of syndrome extraction'
Number of rounds of syndrome extraction
- p. 5 Fig. 4 caption, panel a
instance of the 200-round experiment consists of 1,805 initializations
Only the authors can say what is right here.
Judge's note: Rendered Fig. 4b and 4d at 600 dpi. In 4b the decade ticks are at x=512 (1), 984 (10) and 1455 (100) px, 471 px per decade; the last purple square and red triangle are at x=1538, i.e. 0.18 decade past 100, about 150 rounds. Eleven points are plotted (1, 2, 5, 10, 15, 20, 35, 50, 75, 100, 150). The frame ends at x=1590, about 0.29 decade (roughly 194 rounds), so a 200-round point would fall at or beyond the axis edge. Panel d shows thirteen points with the last at about 200. The 200-round data may exist (panel a says 'Repeat 1–200×' and the caption describes a 200-round instance) but panel b does not show the range the caption states.
Second check: another model agreed
Report a problem with this finding
Wording Fig. 3 caption, panel a: "The dashed line is the fidelity of CNOT gates with simultaneous dynamical decoupling"; the panel plots an ECDF of gate error (x axis "Operation error"), so the dashed curve is the error, not the fidelity.
- p. 4 Fig. 3 caption, panel a
The dashed line is the fidelity of CNOT gates
- p. 4 Fig. 3a x axis
Operation error
Judge's note: Viewed page 4 and the Fig. 3a crop. The x axis is 'Operation error' from 1e-4 to about 1e-2 and the dashed black curve lies at a few 1e-3, to the right of the orange CNOT curve, consistent with the caption's 'degradation'. 'Fidelity' should read 'error' (or 'infidelity'). Meaning is recoverable from the axis.
Second check: another model agreed
Report a problem with this finding
Wording Fig. 4e caption: "Error bars indicate σ = √(p(1 − p)/N) for N = 10,000 shots"; this matches the throughput error bars (about ±0.004) but the fidelity error bars (also about ±0.004) correspond to the post-selected sample of roughly 2,300 shots, not N = 10,000 (which would give about ±0.002).
- p. 5 Fig. 4 caption, panel e
for N = 10,000
- p. 4 Quantum error detection, 4th paragraph
rejects roughly 77% of the 10,000 experimental shots
Judge's note: Rendered Fig. 4e at 600 dpi. Throughput panel: 141 px per 0.02; black error bar for |0000> spans y=263 to 322, half-width 29.5 px = 0.0042, matching sqrt(0.224*0.776/10000) = 0.0042. Fidelity panel: 114 px per 0.02; black bar for |0000> spans y=609 to 655, half-width 23 px = 0.0040. With N = 10,000 the fidelity sigma would be sqrt(0.96*0.04/10000) = 0.0020; with the roughly 23% of shots kept (text: 'rejects roughly 77% of the 10,000 experimental shots') it is sqrt(0.96*0.04/2300) = 0.0041, which matches. The caption's single N does not describe the fidelity panel. The expressions below give variances; take the square root for sigma.
Second check: another model agreed
Report a problem with this finding
4 minor (typos and the like; most checked by the judge only)
Typo Extended Data Fig. 1 caption: "TIA amplifiers are located beneath the daughterboard"; TIA is not defined anywhere in the article (Fig. 1 spells out "transimpedance amplifiers") and "TIA amplifiers" is redundant.
- p. 8 Extended Data Fig. 1 caption
TIA amplifiers are located beneath the daughterboard
- p. 2 Fig. 1 caption, panel a
transimpedance amplifiers (see ‘4 K transimpedance amplifier’ in Supplementary
Judge's note: Viewed the page 8 crop and page 2. Searched the article text: 'TIA' occurs only in this caption; Fig. 1's caption uses 'transimpedance amplifiers' without introducing the abbreviation. A common abbreviation whose meaning is easily recovered, so cosmetic only.
Second check: none (judge only)
Report a problem with this finding
Typo Author list prints "nathan holman" entirely in lower case, unlike every other name in the list.
- p. 6 Author list, 'Members of the HRL Quantum Team and Collaborators'
Vinh S. Ho1, nathan holman1, Adam Holmes1
- p. 6 Author list
nathan holman
Only the authors can say what is right here.
Judge's note: Viewed the page 6 crop: 'Vinh S. Ho1, nathan holman1, Adam Holmes1' is printed as quoted; it is on the page, not an extraction artefact. All other names are capitalised. It may be the author's own styling, so the authors should confirm.
Second check: none (judge only)
Report a problem with this finding
Typo Methods: "electrode voltages that push electrons closer to or farther from another"; a word is missing ("from one another").
- p. 7 Methods, Exchange-only qubits, 1st paragraph
trons closer to or farther from another
- p. 7 Methods, Exchange-only qubits
closer to or farther from another
Judge's note: Viewed the page 7 crop: the line reads 'trons closer to or farther from another. The resulting ...'. The word is absent on the page, not lost at a line break.
Second check: none (judge only)
Report a problem with this finding
Typo Methods: "initialization, logic and readout address only on the S12 degree of freedom"; stray "on" (or wrong verb: "address only the" / "act only on the").
- p. 7 Methods, Exchange-only qubits, 2nd paragraph
initialization, logic and readout address only on the S12 degree of freedom
- p. 7 Methods, Exchange-only qubits
address only on the S12 degree of freedom
Judge's note: Viewed the page 7 crop: the sentence is printed as quoted. Ungrammatical as it stands; meaning is unaffected.
Second check: none (judge only)