manufacturing · digital · materials · family: making one is easy. making a million is the problem
every qubitslightly different
Quantum device fabrication cannot achieve the precision, defect control, or reproducibility needed for scalable systems
Problem statement
Current fabrication processes for quantum devices cannot achieve the precision, defect control, or reproducibility needed for scalable quantum information systems. Three of the research topics NSF's Dear Colleague Letter lists under scalable processes and process control converge here: (1) the controlled introduction, in all three dimensions, of specific defects in diamond and other materials, aligned with their intended function; (2) deposition of defect-free dielectric and superconducting films with low loss at cryogenic temperatures; (3) the key processes that would enable 3D integration of quantum and conventional electronics, such as high-aspect-ratio vias and flip-chip bump bonding. NSF DCL 22-074 says new manufacturing approaches are needed to meet demanding requirements for operation at sub-1 K temperatures, extreme precision in device fabrication and heterogeneous integration, reproducibility, scalability, packaging, and assembly.
Why this matters
The DCL says the potential of quantum computing, sensing, and communications has been demonstrated in initial functioning quantum systems, but that these systems are complex, generally require extreme operating conditions that affect their design, stability, and cost, and need new device manufacturing approaches to bring their cost down and make them widely available. IBM now offers public access to processors it describes as having 100-plus qubits, and exposes per-qubit coherence and gate-error figures for them through its backend interface; the qubit-to-qubit spread in those figures is one place to look for the reproducibility problem the DCL names.
What’s been tried and why it hasn’t worked
The DCL lists the controlled introduction of specific defects in diamond and other materials, in all three dimensions and aligned with their intended function, as an open process need; it does not say how close current methods come. Deposition of defect-free dielectric and superconducting films with low loss at cryogenic temperatures is listed as a second need, and process-related sources of decoherence in spin qubits as a third. For 3D integration of quantum and conventional electronics, the DCL gives high-aspect-ratio vias and flip-chip bump bonding as examples of the key processes still to be enabled, alongside packaging that integrates quantum, photonic, and electronic functions in a vacuum environment. Each of these is listed as a separate topic; the DCL's larger ask is for new processes and approaches to the precision integration of new materials in production at scale, and for automated, scalable manufacturing of devices for quantum computers, sensors, and systems.
What would unlock progress
Understanding and controlling process-related sources of decoherence and the defects in deposited dielectric and superconducting films. Developing techniques for the controlled placement of specific defects in diamond and other materials in all three dimensions. Creating 3D integration processes for quantum and conventional electronics — high-aspect-ratio vias, flip-chip bump bonding — and packaging that integrates quantum, photonic, and electronic functions in a vacuum environment. New methods of large-area, high-throughput characterization of quantum materials and devices.
Entry points for student teams
The qubit-to-qubit variability at the center of this brief is measurable from a laptop. IBM exposes per-qubit calibration data — T1, T2, gate length and error, with dated historical snapshots through `target_history()` — for its 100-plus-qubit processors through the Qiskit Runtime backend interface (`properties()`, `qubit_properties()` and `target_history()`, documented at https://quantum.cloud.ibm.com/docs/api/qiskit-ibm-runtime/ibm-backend; the page notes that `properties()` itself raises an error when given a date on IBM Quantum Compute, the service formerly called Qiskit Runtime), on a free self-serve account; a team can pull months of snapshots across several processors and separate the spread that stays fixed to particular qubits — the fabrication signature — from the drift that recalibration erases. A second team could design and simulate a 3D integration scheme for quantum-classical coupling, evaluating electromagnetic crosstalk and thermal management using finite-element modeling, which needs nothing but simulation software. Fabricating aluminum oxide tunnel barriers to relate deposition parameters to TLS defect density is the direct experiment, but it takes junction-grade cleanroom time plus a vector network analyzer: that access is held by university nanofabrication centers, including the 16 NSF-supported sites of the National Nanotechnology Coordinated Infrastructure (https://www.nnci.net/), which take external users — a team wanting this arm should arrive with the deposition matrix already designed. Relevant skills: data analysis, thin-film deposition, materials characterization, cryogenics, electromagnetic simulation, semiconductor process engineering.
Genes
Structural cousins: stuck the same way, in other fields
Sources
NSF DCL 22-074, "Quantum Manufacturing," NSF Directorate for Engineering, accessed 2026-02-19. go to source ↗
Working notes
Notes made while writing and checking this brief, shown as written. They use working shorthand.
- NSF DCL 22-074 is the primary source; CHIPS Act quantum manufacturing provisions provide the policy context.
- Consolidates three related quantum fabrication challenges (NV center placement, superconducting film quality, 3D integration) under one brief because they share the same root cause: quantum devices require atomic-scale fabrication precision beyond current manufacturing capabilities.
- Overlaps with `manufacturing-quantum-material-synthesis-variability` (which covers bulk quantum materials synthesis) but focuses specifically on device-level fabrication — the layer closer to integrated quantum systems.
- The `failure:lab-to-field-gap` tag applies because individual quantum devices work as hand-made prototypes but no manufacturing process can produce them reproducibly.
- The `temporal:worsening` tag applies because quantum computing roadmaps assume fabrication capabilities that are falling further behind schedule as qubit count targets increase.
- Note reconciled 2026-08-20: a note above argues for `temporal:worsening`; the genome now carries `temporal:static` after a taxonomy revision. The original note is kept verbatim as the tagging rationale of record.
Reconciliation 2026-08-21: Entry-point realism repair (C37 triage, score 2). The triage flag was CONFIRMED: the AlOx-tunnel-barrier deposition study presumes junction-grade cleanroom time and a VNA, neither of which a student team has by default. The whole-section check found the second door sound — the FEM 3D-integration/crosstalk study is squarely student-scale and needs no facility — so rule 1 was already half met, and the repair replaced the fabrication arm rather than deleting it: the section now leads with a data door (IBM's public per-qubit calibration snapshots, which measure the 10–30% coherence spread the brief cites) and keeps fabrication as a named-access arm rather than a semester promise. Resources verified by fetch: IBM Quantum's Qiskit Runtime backend documentation, confirming `properties()` returns T1/T2 and gate length and error and `qubit_properties()` returns per-qubit calibration, with a `datetime` argument for historical snapshots (https://quantum.cloud.ibm.com/docs/api/qiskit-ibm-runtime/ibm-backend); free self-serve access confirmed on https://quantum.cloud.ibm.com/ ("10 free minutes of execution time per month on our 100+ qubit quantum processing units," signup, no application review); and NSF's National Nanotechnology Coordinated Infrastructure, confirmed as 16 NSF-funded university user-facility sites open to external academic, industry, and government users (https://www.nnci.net/).
Reconciliation 2026-09-30: Fact repair after reading the whole write-up against every source it names. All four were fetched on 2026-09-30 and read in full: the Source line's NSF Dear Colleague Letter NSF 22-074, "Quantum Manufacturing," dated April 22, 2022, from https://www.nsf.gov/pubs/2022/nsf22074/nsf22074.jsp (which now redirects to nsf.gov/funding/opportunities/dcl-quantum-manufacturing/nsf22-074; the letter is marked archived but its text is there); IBM's Qiskit Runtime backend documentation at https://quantum.cloud.ibm.com/docs/api/qiskit-ibm-runtime/ibm-backend; the IBM Quantum Platform home page at https://quantum.cloud.ibm.com/; and the NNCI home page at https://www.nnci.net/. The letter is short, about a thousand words, and is the only source for the problem itself; the other three support the entry points only. (1) The problem statement said no process can place nitrogen-vacancy centers in diamond with nanometer precision for quantum networking, that films for transmon qubits cannot be deposited with low enough defect densities to avoid decoherence at millikelvin temperatures, and that no 3D integration process exists for putting quantum and classical electronics on one chip. The letter lists, as research topics, "the controlled introduction in all three dimensions of specific defects in diamond and other materials," "deposition of defect-free dielectric and superconducting films with low loss at cryogenic temperatures," and "key processes enabling 3D integration of quantum and traditional electronic, such as high aspect ratio vias and flip chip bump bonding processes"; it names no nitrogen-vacancy centers, transmons, nanometer figures, or single chip. The three gaps now use the letter's wording and are called what the letter calls them, research topics listed under scalable processes and process control. (2) The statement that the letter "identifies quantum manufacturing as requiring fabrication precision 1–2 orders of magnitude beyond current semiconductor manufacturing capabilities" is not in the letter, which says only that new approaches are needed for "extreme precision in device fabrication and heterogeneous integration, reproducibility, scalability, packaging, and assembly" and operation "at sub-1K temperatures"; the sentence now says that. (3) Removed because no source contains them: the $65–125 billion market projection for 2030; the $200 million CHIPS and Science Act allocation; the IBM and Google million-qubit roadmaps; the 10–30% qubit-to-qubit variability in coherence times on IBM Eagle and Google Sycamore; and the claim that quantum computing will otherwise stay at the "demonstration" stage indefinitely. The paragraph now gives what the letter says (potential "demonstrated in initial functioning quantum systems"; systems "complex" with "extreme operating conditions that impact their design, stability, and cost"; new manufacturing approaches needed "to make them widely available") and what the IBM pages say (public access to 100-plus-qubit processors, with per-qubit coherence and gate-error data exposed through the backend interface); the qubit-to-qubit spread in that data is offered as a place to look for the reproducibility problem, not as a finding. (4) Removed for the same reason: the ~35% conversion efficiency and ~50 nm resolution of ion implantation, the near-unity and <10 nm targets and the "5× and 5×" gap; that MBE and sputtering of niobium and aluminum produce two-level-system defects whose physics is not understood; that flip-chip bonding introduces parasitic modes and thermal problems at millikelvin temperatures; and that the integrated problem of making "millions" of identical devices "remains untouched." The section now lists the letter's own topics, with vias and bump bonding given as the letter's examples of key 3D-integration processes, and its call for "precision integration of new materials in production at scale" and "automated scalable manufacturing." (5) In the unlock section, "the dominant source of qubit decoherence," deterministic NV center placement, chiplets and through-silicon vias, and "days-long cryogenic testing" are not in the letter; the four items now follow the letter's topics (process-related sources of decoherence; controlled defect placement in three dimensions; high-aspect-ratio vias, flip-chip bump bonding and vacuum packaging; "large area, high-throughput characterization"), with no gloss added about fabrication speed; the sub-1 K requirement is stated once, in the problem statement. (6) In the entry points, "readout error" is not among the properties the IBM page names (it names T1, T2, gate length and error, and "other general properties"), so it was removed; and the page says that `properties()` raises NotImplementedError when a date is given on IBM Quantum Compute (the service formerly called Qiskit Runtime), while `target_history(datetime)` returns the properties found on a date, so the historical-snapshot door now points to `target_history()` and notes the limit, naming IBM Quantum Compute as the service. Confirmed in the sources and left as written: the Source line (publication number NSF 22-074, title "Quantum Manufacturing," Directorate for Engineering, the letter signed by the Assistant Director for Engineering); the first sentence of the problem statement and the long title, which restate the letter's premise that new manufacturing approaches are needed for precision, defect control and reproducibility; IBM's "10 free minutes of execution time per month on our 100+ qubit quantum processing units" by sign-up; and NNCI as 16 NSF-supported user-facility sites giving researchers from academia, companies and government access to university facilities. Not checked because no source named here covers it: the Source Notes bullet that CHIPS Act quantum manufacturing provisions provide policy context (the letter does not mention the Act). The note above this one says the IBM data "measure the 10–30% coherence spread the brief cites"; that figure had no source and is no longer in the body. A second reader went over the repairs the same day, fetched the four sources again (including the 2022 Internet Archive copy of the letter at its original nsf.gov address, whose text is the same), found every removal correct, and tightened six phrasings in the repaired text to say no more than the sources do: "process needs" became the letter's "research topics"; "publishes" became "exposes ... through its backend interface" and the sentence calling the qubit-to-qubit spread the place the problem "can be seen directly" became a suggestion of where to look; the unsourced closing line that quantum systems "will not become widely available" without new approaches was dropped as a counterfactual the letter does not state; vias and bump bonding are now the letter's "examples" rather than "the key processes"; "a separate line of work" became "listed as a separate topic"; the "closer to fabrication speed" gloss and the sub-1 K clause were removed from the unlock section; and "the current service" became "IBM Quantum Compute, the service formerly called Qiskit Runtime." No title changed. Genome tags untouched.