Michigan State University National Superconducting Cyclotron Laboratory

This is a historical facility. NSCL ran as a National Science Foundation user facility for four decades, and its final experiment was completed in May 2022, forty years after its first [2]. The Facility for Rare Isotope Beams took the site over on the same campus and under the same university: FRIB was completed in January 2022 and its user experiments began on 9 May 2022 [2]. Heavy ion single event effects work at Michigan State continues at FRIB, which runs the FRIB Single Event Effects Facility from its linear accelerator and, since 19 February 2026, the K500 Chip Testing Facility built on the refurbished NSCL K500 cyclotron [3]. The coupled K500 and K1200 configuration described below is the thing that no longer exists: the K500 now runs alone into a dedicated test cave, and the K1200 was retired with the rest of the cyclotron laboratory. Contact FRIB for current beams and cost.

A part fails in orbit when one galactic cosmic ray ion crosses a sensitive volume and deposits enough charge to flip a bit, latch a parasitic structure, or knock the device off a code path: that mechanism, not a cumulative dose effect, was first identified in a 1975 satellite anomaly and it is the reason single event effects testing exists as its own discipline [5]. A 1984 Jet Propulsion Laboratory guideline, written for the Defense Nuclear Agency before any test standard existed, set the practice that facilities like this one still follow: appraize the part’s technology risk first, then choose an ion source from one of four families, fission fragment sources, alpha emitters, cyclotrons, or Van de Graaff accelerators, because none of the four covers the whole energy and LET range alone [6]. Cyclotrons win the high-energy end of that choice because they are the only source of the four that reaches tens of MeV per nucleon, which is what it takes to cross a sealed commercial package and still deposit charge at the transistor without opening the part [6].

The Single Event Effects Test Facility at the NSCL exists to solve one problem the lower energy accelerators cannot: energy. Its coupled K500 and K1200 cyclotrons deliver heavy ions at 60 to 143 MeV per nucleon, high enough to reproduce about 99 percent of the space radiation linear energy transfer and energy phase space above LET 3 MeV-cm2/mg, and high enough to cross a commercial package overlayer and still deposit at the sensitive volume [1]. Parts are tested in air, undelidded. By 1999 NSCL’s 80 MeV per nucleon beam was already the reference the Space Shuttle program cited for testing avionics without depackaging them [7], and by 2013 the Jet Propulsion Laboratory’s own radiation effects group listed NSCL alongside Texas A&M, Brookhaven, Lawrence Berkeley and RADEF as one of the off-site beams it sends parts to when an effect requires an energy its in-house sources cannot reach [8].

Where NSCL sits among high-energy heavy ion facilities

No single accelerator covers the whole test envelope, so a test planner chooses among facilities by what each one trades away. Brookhaven offers heavy ion energy that can be retuned immediately with no degrader straggling; the Texas A&M University Cyclotron Institute K500 delivers 1 to 80 MeV-cm2/mg depending on tune and is the facility JPL names first for routine heavy ion work; Lawrence Berkeley’s cocktail changes ion species in minutes rather than hours by running several species through the beamline in sequence; RADEF at the University of Jyvaskyla is used for low flux dosimetry and, after its 2007 upgrade, for a 9.3 MeV per nucleon high-penetration cocktail built for backside irradiation of thinned or packaged parts [8][9]. The NASA Space Radiation Laboratory takes the opposite approach again: rather than switching ion species, it tunes energy and inserts degraders to place a chosen LET at a chosen depth in the part, the variable depth Bragg peak method, and its beam ranges run to millimeters in silicon rather than the micrometers a lower energy source reaches [10].

What it costs is the ability to sweep. Switching ion species requires about 24 hours of tuning, so a run normally uses a single ion and cannot map a full cross section against LET curve without paying for the retunes [1]. That is the exact inverse of the Berkeley cocktail, and it is the reason the two facilities are used for different questions rather than as substitutes. A 2023 NASA comparison of the same test device, a Cypress CY7C1049CV33 SRAM, across NSRL, LBNL, Texas A&M and MSU found that stacking boards with degraders between them at NSRL could recover four to six LET points in a single exposure, the closest a beam-degrading facility gets to what a cocktail facility does in one pass [10]. The tradeoff the 2018 JPL guideline for system-on-a-chip testing states from the demand side: modern integrated packages increasingly cannot be delidded at all, which pushes test programs toward exactly the high-energy, high-penetration facilities NSCL represented, and away from lower-energy sources that require an open die [11].

Every facility parameter below, including the 2300 to 2700 US dollar per hour beam time cost, several times the cost at Brookhaven, Berkeley or Texas A&M, is from the 2004 commissioning report, which describes the later upgrades only as plans [1].

Laboratory data

Parameter	Value
Operator	Michigan State University, National Superconducting Cyclotron Laboratory
Location	East Lansing, Michigan, United States
Commissioned	Inaugural single event effects runs February and May 2004
Type	Coupled K500 and K1200 cyclotrons, high-energy heavy ion single event effects
Floor area	Not published
Capabilities	SEETF beam line
Simulant or terrain	Not applicable
Instrumentation	Parallel plate avalanche counter; four-quadrant thin scintillator
Ground truth	Cross sections validated against Brookhaven and TAASC results on the same part
Fidelity limits	Single ion species per run; 24 hour species change
Access	Under 600 beam hours per year for single event effects work
Cited by	radiation testing

Cost and availability, not physics, are what keep this facility a specialist choice: the 2018 JPL guideline puts SOC-capable heavy ion beam time at roughly five times the cost of a conventional facility across the board, not a Michigan State peculiarity [11].

Capabilities

SEETF beam line

Every SEETF beam line value below is from the 2004 commissioning report [1].

Parameter	Value
Working volume	5 by 5 cm beam spot, uniformity better than 85 percent
Test article limits	Most commercial parts testable without delidding
Vacuum	None. Exposures are run in air
Temperature	Not published
Illumination	Not applicable
Slope	Not applicable
Gravity offload	Not applicable
Instrumentation	Upstream attenuator changes flux in under 30 minutes, typically under 5

Ion energy is 60 to 143 MeV per nucleon, and the inaugural runs delivered 9574 MeV krypton and 15048 MeV bismuth, at a usable flux of 1e2 to 1e5 ions/cm2/s set by an attenuator upstream of the accelerators, with dosimetry by a parallel plate avalanche counter below 4e4 ions/cm2/s and a four-quadrant thin scintillator above that, to about 1.5e5 [1]. The 2013 JPL facilities survey draws the same distinction for its own off-site beams: dosimetry hardware and attenuation range are what a test planner has to match to the part, not just peak energy [8].

Where a part cannot be delidded, the LET behind the overlayer is determined by charge collection spectroscopy on a delidded identical part rather than by assuming the package composition [1]. That step is the facility’s answer to its own headline capability, and it is not optional: the same failure mode, package density defeating an assumed range, is what kept a motor driver from being screened at all in a later JPL campaign at a different cyclotron until the beam range was checked against a measured die stackup by SEM [12].

What it does not reproduce

The spectrum, and a cross section curve in one run. The 99 percent phase-space figure is about coverage of the LET and energy plane, not about reproducing a spectrum: the beam is still monoenergetic and single-species, and the 24 hour retune means one run yields points at one LET [1]. See the radiation testing page for what can be argued from that. The 2018 JPL SOC guideline makes the general version of this point: a null result is only as good as the effective sensitivity of the test setup that produced it, the smallest cross section that setup could have detected, and that number must accompany every reported non-effect or the non-effect means nothing [11].

Package transparency, if the package is not what you assumed. The commissioning report documents the trap directly. A plastic-packaged IDT71256 256 K SRAM gave an apparent upset cross section a hundred times below the same part delidded, on a degraded krypton beam, because the plastic was far denser than a pure polymer and the ions ranged out before reaching the sensitive volume [1]. Testing through a package is only as good as the knowledge of what the package is made of.

A user’s guide. This is a commissioning report from two runs. Facility parameters date from 2004 and no published source updates them [1]. A 2021 NASA radiation hardness assurance guideline describes NSCL only as undergoing upgrades and projects post-upgrade K500 and K1200 beam ranges without giving measured facility numbers [13], which is as current as the published literature gets before the closure noted above.

Campaigns run there

Commissioning and cross-validation, 2004. A delidded Matra HM65656 256 K SRAM measured 1.35e-3 cm2 upset cross section under 9574 MeV krypton at LET 7.1 MeV-cm2/mg, and the cross sections measured here agree with results from Brookhaven and TAASC on the same parts across beam energies spanning a factor of forty [1]. That agreement is the argument that the facility measures the same quantity the established ones do.

Space Shuttle avionics qualification, 1999. The Orbiter avionics radiation handbook cites NSCL’s then-new 80 MeV per nucleon beam as the facility that made heavy ion testing of Orbiter electronics possible without depackaging them, feeding directly into the proton-to-heavy-ion equivalence argument the handbook uses to bound unobserved failure rates for parts that cannot be fully swept at any single facility [7].

What is not established

Nothing published since 2004 gives a measured performance number for NSCL: the upgrades that later separated the K500 from the K1200 are described only as plans in the source held here, and the one later NASA guideline that mentions the facility gives projected post-upgrade ranges, not measured ones [1][13]. No source held here states how the 2004 cross-validation against Brookhaven and TAASC held up over the following decade of use, or whether the ion-species retune time changed as the facility matured. The successor facilities are stated from FRIB’s own history and news pages, one of which could not be retrieved past a bot wall and one of which sits on a host outside this container’s network allowlist; both are curated from the citing claim rather than the source text, so their content beyond the dates and names already stated here is unverified [2][3][4].