Texas A&M University Cyclotron Institute

The Radiation Effects Facility at the Texas A&M Cyclotron Institute is the workhorse of the four accelerators listed on the radiation testing page. Its K500 superconducting cyclotron reaches 40 MeV per nucleon, which is enough energy that a part in a ceramic package face up, or a bare flip chip, can be tested without any package modification at all [2][1]. That is why it carries more published single event effect results than any other accelerator: the JPL compendia covering 2007 to 2012 ran most of their heavy ion exposures here, in air [3][4].

The trade against Berkeley is the opposite one. Berkeley switches species in two minutes but ranges out in the package; Texas A&M penetrates the package but changes species more slowly.

Laboratory data

Parameter	Value
Operator	Texas A&M University Cyclotron Institute [2]
Location	College Station, Texas, United States
Commissioned	Not published
Type	K500 superconducting cyclotron, heavy ion single event effects
Floor area	Not published
Capabilities	Heavy ion beam line
Simulant or terrain	Not applicable
Instrumentation	Not published
Ground truth	Not published. Test methods follow ASTM F1192 or JEDEC JESD57 [3][4]
Fidelity limits	Monoenergetic single-species beam; 40 MeV/u ceiling [1]
Access	User facility. No published fee schedule or lead time
Cited by	radiation testing, processors

Nothing held here is a facility description written by the operator. The envelope below is taken from two users’ facility comparisons and the campaign records from the test reports themselves, so it states what has been run rather than what can be booked [1][2].

Capabilities

Heavy ion beam line

Parameter	Value
Working volume	Not published
Test article limits	No package modification needed for ceramic face up or bare flip chip at 40 MeV/u [2]
Vacuum	Not required. Exposures are routinely run in air [3][4]
Temperature	Set by the user. Campaigns here report 25 C, 60 C, 85 C and 125 C part temperatures [3]
Illumination	Not applicable
Slope	Not applicable. Angled incidence is used to reach intermediate LET [4]
Gravity offload	Not applicable
Instrumentation	Supplied by the user; degraders and tilt used to set intermediate LET [3][4]

Two tunes cover the useful range. At 40 MeV per nucleon, linear energy transfer runs 1 to 15 MeV-cm2/mg at 600 to 1600 micrometers of range in silicon, with no package modification needed for a ceramic package face up or a bare flip chip, though a heat spreader or heat sink must still come off [2]. At 15 to 25 MeV per nucleon, LET runs 2 to 80 MeV-cm2/mg at 125 to 250 micrometers, and the ceramic cover must be removed and a flip chip thinned [2]. The 40 MeV per nucleon maximum is stated to be adequate for practically all single event effect testing under JESD57 and ASTM F1192 [1].

Longer-range ions and angled incidence are what make the in-air configuration workable: intermediate LET is reached by tilting the part or inserting a degrader rather than by retuning the machine [4]. The angle changes the effective LET and the effective sensitive depth together, which is a source of error the compendia do not separate.

What it does not reproduce

The spectrum. As at every accelerator used for this work, the beam is monoenergetic and single-species. See the radiation testing page.

A controlled sample size. This is a property of the campaigns rather than of the beam, but it governs how the results should be read: sample counts in the JPL compendia are two or three parts per entry, experimental method, bias and ion selection varied between experimenters, and many entries are one-sided bounds recording only that nothing was seen at the fluence run [3][4]. Several rows are an index into an internal test report rather than a self-contained dataset [3].

Package composition. Testing in air through an intact package means the linear energy transfer at the sensitive volume depends on what the beam crossed to get there. The consequences when that is got wrong are documented at the NSCL facility, where a plastic-packaged part gave a cross section a hundred times below the delidded one.

Campaigns run there

JPL compendium, 2007 to 2010. About forty commercial and radiation-hardened parts, heavy ion here and at Brookhaven, protons elsewhere [4]. The strongest entries are negative: the Analog Devices AD7760 24-bit ADC latched up below LET 8.3 MeV-cm2/mg at both 25 C and 85 C with a saturation cross section near 5e-4 cm2, and Samsung K9F8G08U0M and Micron MT29F8G08U0M 8 Gb NAND flash parts showed destructive high current spikes in PROGRAM and READ modes. The Vishay Siliconix SI9112DY switch mode controller shows the temperature dependence a screening campaign exists to find: latchup between LET 23 and 27 at 25 C, and between 11 and 19 at 85 C [4].

JPL compendium, 2010 to 2012. The Linear Technology LTC1419AIG 14-bit ADC, in Mars Science Laboratory flight lots, is the worked example of temperature-dependent destructive latchup found in a qualified lot: no latchup above LET 86.2 at 20 C, none above 78.0 at 40 C, and a threshold between 50 and 55 MeV-cm2/mg at 60 C, destructive once current limiting was removed [3]. From the same campaign, the Xilinx XQR5VFX130 Virtex-5QV showed no latchup to LET 145 at 125 C but a functional interrupt threshold below LET 1 at about 1e-6 cm2 per device, and Micron NAND flash at 32, 64 and 128 Gb all upset below LET 0.1. Power MOSFET entries are given in the form a designer needs, as a drain-source voltage paired with an ion and an LET: the International Rectifier JANTX2N6790 fails at 130 V under krypton at LET 37 and at 70 V under xenon at LET 50 [3]. Two Cree GaN HEMTs showed no gate rupture or burnout at all.

Power MOSFET line-to-line verification, 2010 [6] and 2014 [5]. Two campaigns treat the beam as a way to compare fabrication lines rather than part numbers. International Rectifier radiation-hardened parts from the Temecula line were immune to gate rupture and burnout within absolute maximum ratings up to about LET 37 on the IRHN57250SE, while the IRHN57133SE from the same line was degraded relative to the same part number from El Segundo, with events at lower drain-source voltage; across 50 devices every observed rupture went gate to drain [6]. The later study measured safe operating areas with the beam degraded to place the Bragg peak at the epitaxial-to-substrate interface, using silver at surface LET 51.7, holmium at 75.8 and gold at 90.4 MeV-cm2/mg [5]. Its result on the Semicoa 2N7616 is a negative one that ended a product line: a 60 V rated part responded like a 500 V rated part, and the vendor withdrew power MOSFETs from its catalog [5].

Analog transient characterization. The comparator work run here quantifies how much of a transient reaches the rail: on the commercial LM139 about 90 percent of transients saturate rail to rail below 0.7 V of differential overdrive and under 1 percent at 1 V, while the radiation-hardened Intersil IS-139ASRH eliminates output transients above 5.8 mV of differential input, tested to LET 83.9 with gold [7]. A Maxim MAX4595 analog switch gave a Weibull transient threshold of 32.1 MeV-cm2/mg and a limiting cross section of 2e-3 cm2, which the authors convert to 1.55e-4 transients per device-day in geostationary orbit and 4.80e-1 on the October 1989 worst day [8]. The second number is a rate prediction from an environment model, not a measurement.