Lawrence Berkeley National Laboratory 88-Inch Cyclotron

The 88-Inch Cyclotron hosts the Berkeley Accelerator Space Effects facility, one of the off-site heavy ion beams JPL’s radiation effects group sends parts to [3]. What distinguishes it is the cocktail: a set of ions of nearly identical charge-to-mass ratio accelerated together, from which the operator selects a species by retuning the cyclotron frequency alone. Because the machine resolves mass to one part in 3000, a species change takes about two minutes [1]. A full cross section against linear energy transfer curve can therefore be measured in one shift, which is the opposite of the position at the NSCL facility, where a species change costs a day.

The cost of that is energy. The cocktails top out at 30 MeV per nucleon, giving ion ranges in silicon of 40 to 1400 micrometers, so most parts must be depackaged and many flip-chip parts thinned before they can be tested here [1][2].

Laboratory data

Parameter	Value
Operator	Lawrence Berkeley National Laboratory [1]
Location	Berkeley, California, United States [1]
Commissioned	Not published
Type	Cyclotron: heavy ion, light ion and neutron irradiation for single event effects
Floor area	Not published
Capabilities	Heavy ion, Cave 4B, light ion, Cave 4A
Simulant or terrain	Not applicable
Instrumentation	Four-quadrant concentric ion chambers; Gafchromic film; cave phosphor and photomultipliers [1]
Ground truth	Per-ring flux and fluence with programmable fluence stop; uniformity by film [1]
Fidelity limits	Monoenergetic single-species beam; 30 MeV/u ceiling forces depackaging [1][2]
Access	User facility; about 2000 beam hours per year across all programs [1]
Cited by	radiation testing, FPGAs and logic

The 2000 hours per year covers heavy and light ions, neutrons and microbeams across nuclear physics as well as space effects work, so it is not the space effects allocation [1]. No held source gives a fee schedule or a lead time.

Capabilities

Heavy ion irradiation facility

Parameter	Value
Working volume	Vacuum chamber in Cave 4B; dimensions not published [1]
Test article limits	Ceramic cover removed; flip-chip thinned, or not tested at the high LET tune [2]
Vacuum	Yes. Heavy ion exposures are run in vacuum [1]
Temperature	Not published. No in-chamber thermal control is described [1]
Illumination	Alignment laser; beam spread to 5 cm on a cave phosphor for tuning [1]
Slope	Not applicable
Gravity offload	Not applicable
Instrumentation	Remote horizontal, vertical and rotational positioning on a motion table [1]

Four standard cocktails run at 4.5, 10, 16 and 30 MeV per nucleon, together covering linear energy transfer from 1 to 100 MeV-cm2/mg at ranges in silicon of 40 to 1400 micrometers and flux up to 1e7 ions/cm2/s [1][3]. Read as two working tunes rather than four cocktails, the envelope a test planner sees is 2 to 100 MeV-cm2/mg at 53 to 80 micrometers of range on the high-LET tune, and 1 to 50 MeV-cm2/mg at 150 to 500 micrometers on the high-penetration tune [2]. On the high-LET tune flip-chip parts should not be tested at all; on the high-penetration tune they can be, thinned, with no heat spreader and no heat sink [2].

Beam cutoff is by manual command, run time, fluence or effective fluence [1]. A collimated milli-beam in the same cave isolates a single event to 10 to 30 micrometers at best, or 100 to 300 micrometers at higher scan rates, using precision slit collimators [1].

The facility description states no uniformity figure over the irradiation area for the heavy ion chamber, no temperature control and no in-situ bias provisions [1]. A campaign needing elevated-temperature latchup screening therefore brings its own thermal arrangement.

Light ion irradiation facility

Parameter	Value
Working volume	Cave 4A, samples in air; 10 cm beam with 2.5, 5, 7.5 or 10 cm collimators [1]
Test article limits	Set by the collimator and the in-air standoff [1]
Vacuum	None. Samples run in air; low energies can be moved to the Cave 4B chamber [1]
Temperature	Not published
Instrumentation	Four-quadrant concentric ion chamber electrodes at 1, 2, 4, 6 and 8 cm [1]

Standard proton energies are 13.5, 20, 30, 40, 50 and 55 MeV at flux up to 1e9 protons/cm2/s, the low end limited by energy loss in the ion chamber and in the air path [1]. Uniformity is verified with Gafchromic film, and the ring electrodes give flux and fluence per ring with a programmable fluence stop. Neutron beams are produced separately by deuteron breakup on a 125 mil beryllium target for high yield or tantalum for low yield, with absolute flux measured by activation foils [1].

What it does not reproduce

The spectrum. A cocktail beam is monoenergetic and single-species; the galactic cosmic ray environment is neither. The general form of that limit, and what can and cannot be argued from a null result at a given species, is on the radiation testing page.

Penetration through a package. At 5 to 50 MeV per nucleon the ion ranges out in the lid, the heat spreader or the substrate before it reaches the sensitive volume, so parts are depackaged and dies thinned [1][2]. A copper heat spreader cuts ion range to 26 percent of the silicon range [2]. Every result taken here is therefore a result on a modified part, and the modification is part of the test record.

A stated capability is not a measured one. The facility description is the operator’s own, with no independent verification and no measurement of delivered dose against a reference [1].

Campaigns run there

Commercial and radiation-hardened FPGA and memory screening, reported 2020. A NASA survey ran a Xilinx XCKU040 Kintex UltraScale, a Microsemi MPF300T PolarFire and several 3D NAND and MRAM parts here [4]. The PolarFire, thinned to 100 to 120 micrometers, gave a configuration memory upset threshold below 1.16 MeV-cm2/mg at about 3e-7 cm2 per design, and a single event functional interrupt at the same threshold, the core current dropping below 100 mA for 1.7 ms and requiring a system reset [4]. The Kintex UltraScale configuration memory upset threshold was below 0.07 MeV-cm2/mg without scrubbing. On the memories, a Micron MT29F1T08 3D NAND showed no latchup to LET 85 at 78 C, a Hynix H25QFT8F4A9R showed a destructive event at 58.8 MeV-cm2/mg at 80 C on two of two parts, and an Avalanche 40 nm MRAM latched between 21.1 and 58.8 MeV-cm2/mg [4].

Point of load converter single event characterization, reported 2019. The Analog Devices RH3845 step-down controller was exposed to yttrium in vacuum through a 2 mil Kapton degrader and to xenon in air at 45 degrees from a 16 MeV/amu cocktail [5]. Destructive onset averaged 44 V at PVIN over five flight-lot devices at LET 38, with minimal temperature dependence between -5 C and room temperature, which is evidence against a single-event burnout mechanism. A cross-conduction event with an onset LET of 0.1 MeV-cm2/mg was eliminated by raising the in-line gate drive resistance from 2 to 10 ohms, after which two devices took LET 79 to 1e7 ions/cm2 at PVIN up to 100 V with no destructive failure [5]. The functional interrupt mode was removed by holding a pre-load current above 1 mA, verified on five devices at LET 79 with no events [5].

Wide bandgap logic, reported 2019. A NASA GRC 4H-SiC junction FET clock circuit of 175 JFETs showed a single event upset onset between LET 3.5 and 9.7 MeV-cm2/mg, no upsets under neon at 3.5 and upsets under argon at 9.7, with no destructive effect to LET 86 at normal incidence under gold [6]. Isolated D flip-flops showed no upsets, which localizes the sensitivity to the control logic rather than the storage element [6].

Comparator transient screening. The Linear Technology RH1011M comparator’s output transients were fully mitigated by appropriate parasitic output capacitance up to LET 114 MeV-cm2/mg at room temperature and at 100 C, the unfiltered transient cross section being about 1.6 percent of total die area and temperature independent [7].