Uthando Nethemba

Atomic clocks and timing — the one axis of quantum sensing that clearly beats the classical standard in the field, but the deployed instrument is not the record instrument

—

by

itembalabantu
in

Evidence-first notes on bioscience and deep tech, at the edge of the lab and the market. Information only — not investment advice. All accuracy, stability and holdover figures are attributed to the instrument, paper or presentation reporting them; systematic uncertainty (accuracy), Allan deviation (stability) and holdover are separated throughout, and peer-reviewed lab records are kept distinct from company claims and agency projections (noted inline). Record instruments (room-scale optical clocks) and deployed instruments (chip-scale and rackmount clocks) are not the same object.

The 30-second version

  • What. Across quantum sensing, timing is the one axis that clears all three gates at once — a lab sensitivity record that beats the classical standard, a field-deployable instrument, and a working advantage in a real application. Lab optical clocks reach fractional systematic uncertainty in the 10⁻¹⁹ range — JILA ⁸⁷Sr at 8.1×10⁻¹⁹ and NIST ²⁷Al⁺ at 5.5×10⁻¹⁹ (both peer-reviewed) — beating the cesium primary standard (the current SI-second definition, <1×10⁻¹⁶) by roughly 100× or more. And atomic clocks have been fielded for over 15 years as chip-scale atomic clocks (CSAC).
  • So what. Even on this mature axis there is a results-layer gap, and here it is not sensitivity — it is SWaP (size-weight-power-cost) and a shift in the figure of merit. The 10⁻¹⁹ accuracy of a room-scale optical clock does not go into the field. What deploys — CSAC (<17 cc) and rackmount optical clocks — sells not accuracy but short-term stability and holdover (time kept when the external reference is cut), several orders of magnitude away from the lab record. Reading “a 10⁻¹⁹ clock has been fielded” is the classic hype pattern for this axis.
  • Now what. Two claims that circulate are false. (a) Deployed instruments do not provide 10⁻¹⁹ accuracy — their spec sheets read in stability/holdover, orders of magnitude below the lab record. (b) Optical clocks have not yet redefined the SI second — redefinition is an unfinished agency roadmap (option proposal 2026, decision targeted 2030), and the second is still defined by cesium microwave. The real field advantage is low-SWaP holdover under GPS denial, on a different axis from the lab accuracy record.

The five-minute read

Three gates, one axis that clears them all

The falsifiable central question set up in Part 0 was: across the three-way gap of “lab sensitivity record ≠ deployable instrument ≠ beating the classical sensor,” which modality actually beats the best classical sensor in the field, and in which application. Atomic clocks and timing are the one axis that clears all three gates cleanly. (1) Lab optical clocks reach systematic uncertainty in the 10⁻¹⁹ range and beat the cesium primary standard — the classical SI-second definition — by roughly 100× or more; classical-beating holds without dispute. (2) Atomic clocks have already been fielded, and in space, for more than 15 years as chip-scale atomic clocks (CSAC). (3) That advantage does real work in metrology, GPS-denied timing and geodesy.

So where does the firm’s lens bite on a mature axis? On the fact that the record instrument and the deployed instrument are not the same object, and the figure of merit each reports is itself different. A room-scale lab optical clock’s 10⁻¹⁹ accuracy (systematic uncertainty) never leaves the lab. What deploys — a CSAC of under 17 cc — reports not accuracy but short-term stability and holdover, several orders of magnitude below the lab record. Even the newly commercialized rackmount optical clock (Vector Atomic EG-30) is specified as 25 fs at 1 s stability and sub-ns holdover, not as 10⁻¹⁹ accuracy.

The bottleneck is not sensitivity — it is SWaP and a shift in the figure of merit

Pull the conclusion forward: the bottleneck on this axis is not sensitivity but the SWaP trade-off, and the essence of that trade-off is that on deployment the figure of merit shifts from accuracy to stability and holdover. The lab optical clock sells “accuracy.” The deployed CSAC or rackmount optical clock sells “time held under GPS denial.” In field applications — above all defense PNT under jamming and spoofing — what the user buys is not “10⁻¹⁹ accuracy” but “the bound on timing error while the external reference is cut,” so CSAC and EG-30 spec holdover. Accuracy, stability (Allan) and holdover are different metrics, and a different one dominates in each application; blurring them into “a 10⁻¹⁹ clock is deployed in the field” is hype. In Part 0’s terms, timing is both the decisive evidence for hypothesis (a) — a narrow, clear classical-beating win — and the anchor for hypothesis (c) — dual-use (PNT / GPS-denied) demand.

Layer Representative instrument Headline figure (metric, source-attributed) SWaP · status
Lab optical clock (record) JILA ⁸⁷Sr optical lattice 8.1×10⁻¹⁹ systematic uncertainty (accuracy) Room-scale, vibration-isolated · lab, PRL 133.023401 (peer-reviewed)
Lab ion clock (record) NIST ²⁷Al⁺ quantum-logic 5.5×10⁻¹⁹ accuracy; 3.5×10⁻¹⁶/√τ stability Room-scale, 3.6 km fiber link · lab, arXiv 2504.13071 (peer-reviewed)
Cesium primary (classical baseline) NIST-F2 Cs fountain <1×10⁻¹⁶ accuracy (defines the SI second) Large, national-lab · agency, NIST
Rackmount optical clock (emerging) Vector Atomic EG-30 25 fs @1s stability; multiday sub-ns holdover 29 L, 3U rack, wall power · commercial (2024–), company/BusinessWire
CSAC (mature deployment) Microchip/Microsemi SA.45s <3.0×10⁻¹⁰ Allan @1s (stability); ~10⁻¹¹ at 1 day <17 cc, <120 mW, <35 g · commercial (15+ yrs), Microchip
Next-gen CSAC (target) DARPA ACES 1,000× improvement in temp/aging/retrace (target) CSAC-class target · agency goal, DARPA
“A record instrument” and “a deployed instrument” are not the same object. The 8.1×10⁻¹⁹ (accuracy), 25 fs @1s (stability), ~10⁻¹¹ at 1 day (stability) and sub-ns holdover values sit on different axes and cannot be ranked in a single column. Lab records (JILA, NIST) are peer-reviewed; EG-30’s numbers are company claims (before third-party national-lab comparison); ACES 1,000× is an unmet agency goal. Advertising a deployed instrument at the lab record’s accuracy — or the reverse — is the hype pattern.

Deep dive

1. Background — the lab record: how far, and on which metric, did optical clocks beat cesium (peer-review attributed)

The current classical baseline is cesium microwave — the definition itself. The SI second is defined by the cesium-133 hyperfine transition at 9,192,631,770 Hz (since 1967); the primary frequency standard, the cesium fountain, is represented by NIST-F2 with target systematic uncertainty δf/f < 1×10⁻¹⁶ (cryogenic 77 K Ramsey region, operating since 2014, NIST-confirmed). That is timing’s “classical” baseline.

  • Optical lattice clock (JILA ⁸⁷Sr): systematic uncertainty 8.1×10⁻¹⁹ — lower than any other clock. Measured in a fermionic Sr ensemble on the 1S₀→³P₀ ultra-narrow transition confined in a shallow vertical 1D optical lattice, with the black-body-radiation shift correction improved via a reassessed 5s4d ³D₁ lifetime. PRL 133.023401 (received 2024-03-14, published 2024-07-10; arXiv 2403.10664, peer-reviewed).
  • Single-ion clock (NIST ²⁷Al⁺ quantum-logic): systematic uncertainty 5.5×10⁻¹⁹, stability 3.5×10⁻¹⁶/√τ. Uses a co-trapped ²⁵Mg⁺ for sympathetic cooling and quantum-logic readout, with laser stability transmitted over a 3.6 km fiber from a remote cryogenic silicon cavity to reach 1 s Rabi probing. arXiv 2504.13071, published 2025-07 (PubMed 40758035) — upgraded from a NIST primary at the Part 0 stage to peer-reviewed. NIST reports “41% more accurate than the previous record and 2.6× more stable than any ion clock” (NIST 2025-07, confirmed).

Separating the figures of merit (essential): JILA’s 8.1×10⁻¹⁹ and NIST’s 5.5×10⁻¹⁹ are systematic uncertainty (accuracy); NIST’s 3.5×10⁻¹⁶/√τ is stability. Accuracy, stability and holdover are different metrics and must not be mixed. The 8.1 vs 5.5 between the two families (Sr lattice vs Al⁺ single ion) must not be read as a head-to-head ranking — different transitions, protocols and ensembles; they share only the conclusion “10⁻¹⁹ range, beating Cs by 100×+.”

Classical-beating verdict (timing): optical at the 10⁻¹⁹ range vs Cs primary at ~10⁻¹⁶ gives roughly three orders of magnitude (100–1000×) by which optical beats microwave — the point where classical-beating holds with the least dispute across all of quantum sensing. Note that the cesium fountain also uses atoms (quantum), so the true “quantum vs classical” contrast here is microwave atoms vs optical atoms; the genuine “classical” alternative (quartz / GPS discipline) appears in the deployed applications below.

2. Redefining the SI second — classical-beating institutionalized as a standard change (agency attributed)

The optical clock’s advantage is being institutionalized beyond performance competition, as a metrology event that would move the SI second’s definition from cesium microwave to an optical transition.

  • Roadmap (Metrologia review, IOP 10.1088/1681-7575/ad17d2, 2024, confirmed): optical-transition redefinition roadmap adopted at the 2017 CCTF → refined by the 2020 CCTF task force → roadmap validated at the 2022 CGPM → redefinition option proposed at the 2026 CGPM → redefinition at the 2030 CGPM (backup 2034).
  • Ancillary conditions: one of the six conditions to be met by the redefinition date is that UTC(k) be generated by, or at least steered by, optical clocks — i.e., optical clocks must actually enter the UTC-generation chain.
  • Current progress: at least four national labs are approaching low-10⁻¹⁸ accuracy — roughly 100× better than current cesium (BIPM CCTF 2025 update, confirmed).

Status separation: redefinition is an agency roadmap with an option proposal in 2026 and a decision targeted for 2030 — not yet confirmed or in force (the cesium definition still holds). “Optical clocks have already redefined the second” is false (see §8 Refuted). That the standard change is on an official roadmap at all shows the institutional maturity of timing’s classical-beating — a “standard-redefinition” path no other quantum-sensing modality has.

3. The deployment layers — the SWaP stack from record to fielded (product/agency attributed)

Here the results-layer bites. The “best-performing instrument” and the “deployed instrument” are different objects, and the stack (see the table above) trades performance (accuracy/stability) against SWaP (volume/power). Note that the figure of merit differs by row (accuracy vs stability vs holdover).

  • Between the lab optical clock (10⁻¹⁹) and CSAC (10⁻¹⁰–10⁻¹¹) there are several orders of magnitude in accuracy/stability, paid for by shrinking from room-scale to <17 cc — the quantitative basis for Part 0’s “record ≠ deployed.”
  • The EG-30 (Vector Atomic) is a new intermediate layer — described as the “world’s first fully integrated commercial optical clock,” putting lab optical timing into a 29 L rackmount and claimed to beat an active hydrogen maser on short-term stability and phase noise at one-tenth the volume (company, confirmed). But it is specified in stability (25 fs @1s) and holdover (sub-ns), not 10⁻¹⁹ accuracy — the physical evidence that the figure of merit shifts on deployment. It is a US Navy/Army/DARPA/OUSD-supported technology, and Vector Atomic was acquired by IonQ in 2025-10 (Part 0, confirmed).
  • CSAC is the 15-year mature instrument, but its value is not accuracy — it is low-SWaP holdover, keeping time when GPS is cut (§4). The DARPA CSAC lineage is ~100× smaller and ~50× lower power than the lab (Part 0); ACES targets 1,000× improvement (not achieved; agency goal).

Do not mix metrics: the 8.1×10⁻¹⁹ (accuracy), 25 fs @1s (stability), 10⁻¹¹ at 1 day (stability) and sub-ns holdover in the table are values on different axes and cannot be placed in a single ranking. Advertising a deployed instrument at the lab record’s accuracy, or the reverse, is the hype pattern.

4. The results-layer bottleneck — why the deployed instrument does not deliver the record performance (firm lens)

  • ① The SWaP–performance trade-off is the essence. The lab optical clock’s 10⁻¹⁹ is the product of room-scale scale, vibration isolation, cryogenic fiber links and expert operation (the 3.6 km link, the vertical lattice of §1). The deployed instrument folds that infrastructure into <17 cc (CSAC) or 29 L (EG-30) and gives back performance. This is not a defect but a physical trade-off — narrow linewidth, long probing and perfect isolation make accuracy, and deployment abandons those premises.
  • ② On deployment the figure of merit shifts from accuracy to holdover/stability. This is the core of the firm’s lens. In field applications (especially GPS-denied timing) the user buys not “10⁻¹⁹ accuracy” but “the bound on timing error while the external reference is cut,” so CSAC and EG-30 spec holdover (EG-30: multiday sub-ns). Accuracy, stability (Allan) and holdover are different metrics, and a different one dominates each application — accuracy for metrology, holdover for GPS-denied timing.
  • ③ The “classical” alternative is not cesium but quartz / GPS discipline. In deployed applications the real classical rival is a GPS-disciplined quartz oscillator (OCXO), not the cesium fountain. In peacetime GPS disciplines the quartz well enough, but when jamming/spoofing cuts GPS, the quartz drifts rapidly. There the atomic clock (CSAC / rackmount optical clock) keeps time via holdover — the actual classical-beating at the deployment layer. Timing’s field advantage holds not on “10⁻¹⁹ accuracy” but on low-SWaP holdover under GPS denial, a different axis from the lab record.
  • ④ Company claim ≠ peer-review ≠ agency projection. EG-30’s 25 fs @1s and sub-ns holdover are company claims (before third-party national-lab comparison); DARPA ACES 1,000× is an unmet agency goal; SI redefinition 2030 is an unconfirmed agency roadmap. Only the lab records (JILA, NIST) are peer-reviewed. Blurring the three into “a quantum clock is deployed in the field at 10⁻¹⁹” is this axis’s characteristic hype.
  • ⑤ Dual-use pull — application demand drives SWaP maturity. The largest demand for timing deployment is defense PNT (GPS-denied) — CSAC goes into handheld radios and munition timing, rackmount optical clocks into ships, data centers and coherent radar. EG-30 emerging with Navy/Army/DARPA/OUSD support and being folded into IonQ, and CSAC being a DARPA lineage, is that structure (cross-domain hook §defense/PNT). Defense timing demand, not chemistry or materials, drives modality maturity.

5. An adjacent application — relativistic geodesy: where accuracy is itself the measurement (peer-review attributed)

Relativistic geodesy is the application where the lab optical clock’s 10⁻¹⁹ accuracy is not wasted but becomes a new measurement tool — here accuracy is the product, so deployment and performance diverge less (though deployment remains hard).

  • Principle: a clock runs differently at different gravitational potential (gravitational redshift). Optical-clock accuracy of 10⁻¹⁸ corresponds to a geopotential difference of about 1 cm in height — the clock can be used as an altimeter.
  • Field demonstration: a transportable ⁸⁷Sr optical lattice clock measured the geopotential difference between a mid-mountain site and a point 90 km away, cross-checked locally against a ¹⁷¹Yb clock (Nature Physics 2018, s41567-017-0042-3, peer-reviewed).
  • Lab extreme: a clock network of five atom ensembles arranged evenly across 1 cm of height measured a frequency gradient of −12.4×10⁻¹⁹/cm (matching the predicted −10.9×10⁻¹⁹/cm), verifying gravitational redshift at 1.3 mm resolution (Nature Communications 2023, s41467-023-40629-8, peer-reviewed).

Meaning: below 10⁻¹⁸, optical clocks become sub-cm geopotential probes used to reconcile national height systems, for geodesy and for dark-matter searches. But transportable optical clocks are still room/truck-scale — far from continuous field deployment. Here too the record-vs-deployable split holds: the geodesy demonstration is transportable, not deployed. The 90 km campaign is special operation, and CSAC-class SWaP does not reach geodesy accuracy — accuracy-needing applications use lab-class instruments, holdover-needing applications use CSAC. That application-instrument split is the structure of the timing axis.

6. Verdict — proceed-with-caveats (conditional)

  • JILA ⁸⁷Sr 8.1×10⁻¹⁹ — peer-review demonstrated (systematic uncertainty / accuracy, PRL 133.023401, 2024-07).
  • NIST ²⁷Al⁺ 5.5×10⁻¹⁹ — peer-review demonstrated (upgraded) (accuracy; stability 3.5×10⁻¹⁶/√τ, arXiv 2504.13071, 2025-07, PubMed).
  • Optical > cesium by 100×+ — demonstrated (classical-beating) on the accuracy axis (10⁻¹⁹ range vs Cs primary <1×10⁻¹⁶).
  • SI second redefinition path — agency roadmap (unconfirmed) (Metrologia 2024; option 2026, decision 2030; UTC(k) steer condition).
  • CSAC deployment / SWaP — commercial demonstration (trade-off) (SA.45s <3×10⁻¹⁰@1s, 10⁻¹¹ at 1 day, <17 cc, <120 mW, <35 g; stability).
  • Rackmount optical clock EG-30 — commercial (company spec) (25 fs @1s, sub-ns holdover, 29 L; company claim, before third-party comparison).
  • DARPA ACES 1,000× — agency goal (not achieved).
  • Relativistic geodesy — peer-review demonstrated (transportable) (90 km transportable Sr, Nat Phys 2018; 1 cm network, Nat Comms 2023).
  • Deployed instrument provides 10⁻¹⁹ accuracy — false (Refuted): deployed metrics are holdover/stability, orders of magnitude below the lab record.

Overall: proceed-with-caveats (conditional). Timing is the axis where quantum sensing’s classical-beating is clearest (optical clocks 100×+, peer-reviewed) and deployment is most mature (CSAC 15 years, EG-30 commercial). But four caveats must sit at Tier-1 level: ① the record instrument (room-scale optical clock) and the deployed instrument (CSAC, rack) are different objects; ② on deployment the figure of merit shifts from accuracy to holdover/stability; ③ EG-30 specs, ACES targets and 2030 redefinition are company/agency claims; ④ the geodesy demonstration is transportable, not continuously deployed. Do not exaggerate to “a 10⁻¹⁹ clock has been fielded.”

7. What to watch (falsifiable)

  • Redefinition gate: whether the 2026 CGPM adopts the redefinition option and whether ≥4 national labs sustain low-10⁻¹⁸ accuracy while steering UTC(k) — the ancillary condition, not the lab record, is the pacing item.
  • EG-30 third-party comparison: whether a national lab publishes an independent accuracy/stability comparison of the rackmount optical clock, converting company claims into peer-reviewed field performance.
  • Optical CSAC: whether a chip-scale optical clock reaches deployable SWaP (NIST/DARPA R&D, not commercial) — the event that would collapse the record-vs-deployed gap.
  • ACES 1,000×: whether the DARPA target is actually met, and whether field GPS-denied holdover records for CSAC / optical clocks are published (mostly defense-attributed, largely undisclosed).

References

Disclosure

This post is for information only and is not investment advice.

COI note: this post describes national-lab and agency work (NIST, JILA, DARPA, BIPM) and listed/private companies (Microchip/Microsemi CSAC; Vector Atomic, acquired by IonQ) in a descriptive, neutral context. Every accuracy, stability, holdover and uncertainty figure is attributed to the instrument, paper or presentation reporting it; systematic uncertainty (accuracy), Allan deviation (stability) and holdover are separated throughout, and peer-reviewed lab records (PRL, Nature, Metrologia) are kept distinct from company claims and agency projections. Defense PNT (GPS-denied) demand is described as factual and neutral. Quantitative claims are attributed to the vendor, author or preprint/presentation. Company and performance statements are factual, neutral descriptions and are not buy/sell implications for any security.

Comments

Leave a comment