CMP Slurries & Polishing Compounds

Radiation-hardened and radiation-tolerant semiconductor fabrication is the most structurally distinctive archetype in the twelve-fab taxonomy. It differs from the commercial semiconductor industry on multiple axes simultaneously: the customer base is sovereign-coupled rather than commercial; the supplier base is accreditation-gated rather than capital-gated; the process nodes are mature (90–250nm) rather than leading-edge; the qualification regime is QML-V or QML-Q rather than AEC-Q or commercial; the production volumes are measured in thousands per year rather than millions; the prices per device run $5K–$50K rather than $5–$50; and there is no commercial substitute when DoD, NASA, or an intelligence agency requires a rad-hard part. Every other fab archetype operates within some commercial-competitive frame. Rad-hard operates within a regulatory-sovereign frame that overlaps with commercial manufacturing only in vocabulary.

The central structural fact: if the US Department of Defense or an intelligence agency requires a radiation-hardened processor, neither capital nor capability can substitute for a rad-hard part from an accredited supplier. The Defense Microelectronics Activity (DMEA) Trusted Foundry program runs the accreditation regime that qualifies suppliers to handle classified designs for defense and intelligence customers. MIL-PRF-38535 specifies the Qualified Manufacturers List tiers (QML-V for space, QML-Q for military) under which rad-hard parts are qualified. The population of accredited rad-hard suppliers is small, US-concentrated, and cannot be expanded by foreign competition regardless of technical merit. Europe operates a parallel regime under the European Space Components Coordination (ESCC) with a substantially smaller operator base.

Why Mature Nodes

Rad-hard production concentrates at mature process nodes — typically 90nm to 250nm, with some legacy rad-hard production at 0.35μm and larger geometries — because radiation hardening is better understood and physically easier to achieve at larger transistor geometries. The physics is straightforward. A total ionizing dose (TID) event deposits charge in a transistor's gate oxide and depletion regions. A larger transistor has more volume to absorb that charge before parameters shift out of specification. A sub-5nm transistor has so little volume that the same radiation dose produces a larger fractional parameter shift. Single-event effects (SEE) — where a single high-energy particle causes a bit upset or latchup — are similarly more tractable at larger nodes because the critical charge required to flip a node is larger.

Rad-hard design also benefits from decades of accumulated characterization data at mature nodes. A 180nm rad-hard process has 20+ years of in-space operation, archived failure data, and validated design rules. A 5nm rad-hard process, if it existed, would have none of this. Mission customers (NASA for 20-year Mars missions, DoD for 15-year strategic defense programs) require qualification data that simply does not exist for leading-edge nodes — and cannot be generated on faster timescales than the mission lifetimes themselves.

The consequence is that the rad-hard supply chain is structurally decoupled from the leading-edge logic roadmap. When TSMC progresses from N3 to N2 to A14, the rad-hard industry does not follow. BAE Systems and Honeywell produce rad-hard processors today on processes that are generations behind commercial leading-edge, and this gap is by design rather than by neglect. The performance deficit relative to commercial silicon is compensated by radiation survival margins that commercial silicon cannot match at any node.

Three Categories of Rad-Hardness

Rad-hard parts divide into three distinct categories by how radiation tolerance is achieved. The distinctions matter because each category carries different cost, performance, and mission-suitability profiles.

The rad-tolerant category has grown substantially with LEO commercial space scale-up (Starlink and competing constellations require massive unit volumes where full rad-hard costs would be prohibitive). SpaceX has led the industry trend toward custom radiation-tolerant silicon for its own constellation, with the emerging Terafab AI7/D3 program representing radiation-tolerant inference compute at scale. This is distinctive from traditional DMEA rad-hard because the customer (SpaceX) is the integrator rather than a government procurement program, the process can be leading-edge-adjacent because mission life is shorter than traditional defense missions, and the volumes can justify custom silicon design effort that small-volume defense programs cannot.

Failure Mechanisms and Mitigation

Rad-hard design targets three distinct failure mechanisms, each requiring different mitigation approaches. Understanding the mechanism taxonomy is prerequisite to understanding why rad-hard parts differ from their commercial equivalents.

Each mission environment requires characterization against relevant failure mechanisms. LEO commercial missions see primarily SEE from cosmic rays and trapped particles; geostationary missions see higher proton and cosmic ray flux; deep-space missions see the full spectrum including galactic cosmic rays; strategic defense applications must survive prompt-dose events from nuclear detonations (a separate qualification regime, MIL-STD-883 Test Method 1019). Rad-hard parts are characterized against the specific radiation environment they will operate in.

The US Operator Landscape

The SpaceX Terafab Model as Emerging Tier

The SpaceX Terafab AI7/D3 program represents a genuinely new point in the radiation-tolerant semiconductor space. Traditional rad-hard production has been characterized by low volumes, high prices, government customers, and mature-node processes. The Terafab model inverts much of this: high volumes (thousands of satellites per year in planned constellation scale-up), commercial economics (SpaceX as integrator rather than government procurement customer), and leading-edge-adjacent processes (radiation-tolerant rather than classical rad-hard). The AI7 designation reflects the Tesla/SpaceX chip family lineage; the D3 suffix reflects the radiation-tolerant variant targeted at space deployment.

This model is distinctive enough that it warrants treating as a separate category rather than an extension of traditional rad-hard. The customer pool is different (commercial LEO integrators), the cost structure is different (volume-driven rather than program-driven), the technology approach is different (radiation-tolerant by screening and mission-specific derating rather than traditional RHBD), and the accreditation regime is different (not DMEA Trusted Foundry, but SpaceX's own internal qualification). Whether this model extends beyond SpaceX to other commercial space integrators or remains SpaceX-specific is an open question as the broader LEO commercial space industry scales.

The broader implication is that the radiation-resistant semiconductor market is bifurcating. Traditional DMEA-regulated rad-hard continues serving defense, intelligence, and deep-space missions at classical low-volume high-cost economics. Emerging commercial radiation-tolerant silicon serves LEO constellation and commercial space at volume economics closer to commercial semiconductor industry. The two tiers compete in some applications (small-satellite missions could source from either) but serve fundamentally different customer bases for the most part.

Strategic Framing: No Commercial Substitute

The core structural fact about rad-hard supply is that there is no commercial substitute. When the US Department of Defense specifies a radiation-hardened processor for a strategic defense program, or NASA specifies one for a Mars mission, or an intelligence agency specifies one for a classified satellite, the customer cannot substitute a commercial MCU regardless of cost or capability. The specification is driven by mission requirements (radiation environment, mission life, reliability margin) and by regulatory requirements (Trusted Foundry, QML qualification, export control). Both gates constrain sourcing to the narrow set of accredited rad-hard suppliers.

This makes the rad-hard supplier base a strategic asset of the US defense industrial base. The government has explicit interest in maintaining the health of BAE, Honeywell, Microchip, SkyWater, GlobalFoundries trusted foundry, and the supporting ecosystem — not primarily on commercial grounds but on strategic-defense grounds. CHIPS Act provisions include support for trusted foundry capacity expansion and workforce development. The DoD Microelectronics Commons program invests in next-generation defense semiconductor capabilities. This funding environment is different from commercial semiconductor support (CHIPS Act commercial fab subsidies) and operates under different strategic logic.

The geopolitical implication is that rad-hard supply chains are among the most durably US-concentrated in the entire semiconductor industry. China cannot build DMEA Trusted Foundry capability regardless of capital invested — the regulatory regime excludes foreign suppliers from classified defense work by design. European parallel capability under ESCC serves European space and defense but does not substitute for US requirements. The rad-hard archetype is therefore the archetype where US supply chain concentration is a feature rather than a risk.

Fabs in This Archetype

The specific fab inventory for this archetype is maintained in the Fab Facilities dataset with per-fab profile pages. Notable rad-hard and rad-tolerant fabs include BAE Systems MSD Nashua, Honeywell Plymouth, SkyWater Bloomington MN, GlobalFoundries Fab 8 Malta NY (trusted foundry services), GlobalFoundries Fab 9 Burlington VT (SiGe BiCMOS for defense RF), Microchip facilities, AMD/Xilinx space-grade FPGA operations, Teledyne e2v European operations, and the emerging SpaceX Terafab. See Fab Facilities for the full inventory.

Related Coverage

Parent: Wafer Fabs

Peer archetype pages: Leading-Edge Logic · Mature Logic · DRAM · 3D NAND · SiC Power · GaN Power & RF · Analog & Mixed-Signal · CMOS Image Sensor · MEMS · III-V Compound Semiconductor · Silicon Photonics

Cross-pillar dependencies: Rad-Hard / Rad-Tolerant Chips · Strategic Defense Electronics

Emerging programs: Tesla Terafab (AI7/D3 radiation-tolerant inference program)