ARPA-H UNIQUE Wants Quantum Contrast From Ordinary PET Photons — Solution Summaries Due October 29, One Integrated Team Per Proposal

Every PET scan ever performed has thrown away information. When a positron from a radiotracer meets an electron in tissue, the annihilation produces two 511 keV photons flying in opposite directions. Conventional scanners record where and when those photons land and reconstruct where the tracer went. They ignore the fact that the two photons are born with quantum-correlated polarizations — and that the strength of that correlation may carry information about the local tissue environment the photons came from.

The Advanced Research Projects Agency for Health is now paying to find out whether that discarded signal is medically useful. UNIQUE — UNifying Imaging with QUantum Entanglement — solicitation ARPA-H-SOL-26-162, posted September 17, 2026, is an 18-month, two-phase program to establish "a new quantum-derived layer of biology" in imaging. The required solution summary is due October 29, 2026 at 5:00 PM ET. Full proposals follow on December 15, 2026.

That leaves nine days for anyone who has not started. This analysis covers what the program actually asks for, the eligibility rules that will knock out common team structures, the date discrepancies circulating across aggregators, and how to decide in the next 48 hours whether to compete.

The core facts

Why ARPA-H is betting on this now

The physics is not new. That the two annihilation photons in PET carry entangled linear polarization has been understood for decades, and a 2020 arXiv paper on photon entanglement "in the MeV regime" laid out how the effect could be exploited in PET. An APS Physics viewpoint in 2024 made the case that entanglement information could help a scanner distinguish true coincidences from scattered ones — a big deal, since scattered events dominate when photons pass through the body.

What changed in 2026 is that the idea left the bench. On June 28, 2026, a Jagiellonian University team led by Paweł Moskal posted "First-in-human quantum entanglement imaging" (arXiv:2606.29421). They imaged a patient injected with a Ga-68 DOTA-TATE tracer on the J-PET scanner — a modular system built from plastic scintillator strips, chosen precisely because 511 keV photons Compton-scatter in plastic, which exposes polarization information. The team reconstructed both a standard uptake image and an image of entanglement degree. Liver and spleen came in at entanglement-degree values of 0.79 ± 0.21 and 0.76 ± 0.23 — between the separable-photon prediction and maximal entanglement, per the authors. The statistics were thin and the authors call the results preliminary.

That is the state of the art UNIQUE inherits: a proof that the measurement is possible in a human, with error bars too wide to say whether it means anything clinically. ARPA-H's framing on the program page — conditions like chronic traumatic encephalopathy, neurodegeneration, and pancreatic cancer "are often detected only after significant, sometimes irreversible, damage" — is the bet that quantum-derived contrast might report on tissue microenvironment (metabolism, oxygenation, microstructure) earlier than uptake alone.

The program is short by ARPA-H standards. Eighteen months is not long enough to build a clinical scanner. It is long enough to answer a binary question: does the quantum signal carry biologically meaningful contrast in controlled phantoms, yes or no? Proposals should be written to answer that question cleanly, not to promise a product.

What the two technical areas actually require

TA1: Imaging Core

TA1 builds the physical system that generates and measures quantum observables. The program page lists quantum-correlated photon production and collection, engineered source materials and nanomaterials, quantum-aware detectors, and — notably — whether decoherence and correlation loss can themselves serve as biological contrast. That last phrase is the scientific heart of the program: the hypothesis is that tissue environments degrade entanglement differently, and that the degradation is the signal.

Third-party summaries of the full solicitation report quantitative Phase 1 targets including an entanglement witness above 1.6, coincidence timing under 300 picoseconds (stretch goal under 150 ps), entangled-pair flux of 100,000 to 1,000,000 pairs per second, and measurable differences across at least three phantom states. Phase 2 reportedly requires ghost-image reconstruction in phantoms, over 1,000,000 correlated events, repeatability within 10%, and 3D localization of 2–3 mm or better. Confirm every number against the attachments on SAM.gov before writing to them, but the shape is clear: the timing target is substantially tighter than the roughly 490 ps time-of-flight resolution reported for the J-PET system used in the first human study.

TA2: Quantum AI Interpretation and Fusion

TA2 is not a generic "add machine learning" area. Summaries of the solicitation say proposers must explain what is "quantum aware" about their reconstruction for correlated, sparse, low-count data; extract diagnostic features such as entanglement witnesses, decoherence signatures, and positronium states; quantify uncertainty; and fuse quantum-derived data with anatomical, functional, and molecular imaging. The deliverable is a preliminary quantum-derived contrast atlas — the solicitation reportedly names quantum decorrelation tomography (QDT) or a quantum-derived index (QDI) as examples.

The integration rule that reshapes every team

Here is the single most important structural fact: UNIQUE reportedly wants one integrated proposal from a single lead proposer and PI covering both phases and both TAs. Standalone detector, hardware, or software concepts will not meet the requirement. The program page reinforces this: teams must both "measure quantum-derived observables and interpret them as meaningful biomedical imaging signals."

That kills the most natural academic approach — a detector physics group submitting TA1 and an imaging-AI lab submitting TA2 separately. If you are only one half, you need the other half under your proposal (or yours under theirs) before October 29. Key personnel reportedly must be in place within 60 days of award.

ARPA-H also plans independent verification and validation (IV&V), which may supply benchmark phantoms, and UNIQUE performers cannot compete for the IV&V role. If your group's real strength is phantom design and measurement standards rather than novel imaging, the IV&V work — not a performer slot — may be the better target. Watch for a separate IV&V solicitation.

Eligibility: who is in, who is out

• Eligible: universities, nonprofits, small businesses, and other-than-small businesses.
• FFRDCs, government entities, and federal employees: generally barred as primes and subperformers. ARPA-H may consider FFRDC participation only for documented, uniquely necessary capabilities. This matters because much of the U.S. expertise in fast-timing detectors and positron physics sits in national laboratories. A university team that planned to borrow a DOE lab's detector group should either document why that capability is uniquely necessary or restructure.
• Cost share: for-profits that are not small under NAICS 541714 (a 1,000-employee standard) must provide meaningful cost share, cash or in-kind. Everyone else may cost share voluntarily. This is an opening for large imaging OEMs to join as subperformers contributing scanner hardware — in-kind equipment that directly supports milestones counts.
• Covered foreign countries: excluded, per third-party summaries. Note that the most advanced human entanglement-PET work to date is Polish (Jagiellonian, with INFN Frascati in Italy). Neither is a covered country, but foreign participation in an ARPA-H OT still requires scrutiny; check the solicitation's foreign-participation terms before naming international collaborators.
• One summary of the solicitation reports that each organization may lead only one proposal and that early-career investigators are preferred. Verify both in the attachments — if the one-lead limit holds, institutions with multiple interested groups need an internal decision this week.

The date confusion — and which date to use

Aggregators disagree. The ARPA-H program page lists October 29, 2026, 5:00 PM ET for the solution summary and December 15, 2026, 5:00 PM ET for the full proposal, both labeled estimates. One consulting summary lists October 30 at 11:59 PM; a contract-data listing shows a December 16 response date; another listing cites a different solicitation number (26-164). The Proposers' Day date also varies across sources (October 15 vs. October 16), though it has passed either way.

Plan to the earliest date: October 29 at 5:00 PM Eastern. Submitting a day early costs nothing; missing a required gate ends your candidacy. ARPA-H will encourage or discourage each proposer after reviewing summaries, and the program page recommends submitting a full proposal only if encouraged.

What goes in a three-page solution summary

The summary is three pages excluding the cover page, ROM budget, team qualifications, and references. Summaries of the solicitation say it should address relevance to ARPA-H and UNIQUE, technical merit, and team organization, and at minimum answer the first three Heilmeier questions: what health problem, what are the limits of current practice, and what is new in your approach.

The ROM budget must break out Phase I, Phase II, and total, with labor and hours, subcontracts, materials, equipment, travel, other direct costs, indirect costs, fee, and any cost share. Because ARPA-H has published no award size, the ROM is doing double duty: it is how the program manager sizes the portfolio. Price what the work genuinely costs. An unrealistically low ROM signals that you have not understood the timing and flux targets; an unexplained high one invites discouragement.

A strong summary for this program will do four things in three pages:

1. State the contrast hypothesis precisely. Not "entanglement may reveal disease," but which observable (polarization correlation, decoherence rate, positronium lifetime) you expect to differ across which phantom states, and why physically.
2. Show the path to the timing and flux numbers. Name the detector material and readout, and give the current measured performance with a credible delta to the Phase 1 targets.
3. Make TA2 quantum-aware, not generic. Explain how the reconstruction treats correlated photon pairs as a joint measurement rather than as two independent hits.
4. Show the integration. One PI, one plan, explicit TA1→TA2 data interfaces, and a phantom strategy that can produce three statistically separable states within nine months.

Who should compete — and who should not

Strong fits: groups with existing Compton-polarimetry or fast-timing PET hardware and a positron-physics track record; imaging-AI groups already working on low-count, physics-informed reconstruction who can pair with them; detector startups that can move faster than a university and absorb OT terms; and large imaging companies willing to contribute in-kind hardware as cost-sharing subperformers.

Weak fits: pure quantum-computing groups with no detector access (the program is about quantum sensing of photons, not quantum computation); clinical teams expecting patient studies (human subjects are not anticipated); and anyone who can only deliver one TA and cannot find a partner by October 29.

The national-lab problem. If your best detector partner is an FFRDC, either build the "uniquely necessary" justification now or replace them. Do not assume an exception.

How UNIQUE fits ARPA-H's 2026 pattern

UNIQUE follows the same template as other 2026 ARPA-H programs Granted has analyzed — a required solution summary gate, multi-phase OT awards with no published ceiling, and a hard go/no-go between phases. The SPECTRA autism program, posted the same day with the adjacent solicitation number, uses staggered summary deadlines across four technical areas; the REST sleep diagnostics program ran a 66-month OT. UNIQUE is the shortest and most physics-forward of the group — closer to a DARPA seedling than to a clinical development program. Proposers who have worked with DARPA's office-wide BAAs will recognize the shape: a narrow scientific question, aggressive quantitative metrics, and a program manager with wide discretion over who advances.

The nine-day checklist

• Today: register or confirm active SAM.gov status and ARPA-H Solutions portal access. Download the ISO and attachments from SAM.gov and confirm the solicitation number and deadline yourself.
• By October 22: lock the integrated team — one lead, one PI, both TAs. Resolve any FFRDC dependency.
• By October 24: draft the contrast hypothesis and phantom plan; get measured timing and flux numbers from your detector lead.
• By October 27: finish the ROM by phase, including cost share if any team member is a large for-profit.
• October 28: internal review against the Heilmeier questions. Submit.
• October 29, before 5:00 PM ET: the real deadline, not the target.

UNIQUE is a small, sharp program asking whether medicine has been discarding a usable signal in every PET scan for fifty years. If your team can measure that signal and interpret it under one roof, nine days is enough to write three pages. If you can only do half, spend the first two days finding the other half.