Recent Radiation Test Results for Power MOSFETs

Single-event effect (SEE) and total ionizing dose (TID) test results are presented for various hardened and commercial power metal-oxide-semiconductor field effect transistors (MOSFETs), including vertical planar, trench, superjunction, and lateral process designs.

density over the standard LDMOS and improved frequency response over trench-style MOSFETs.
Power MOSFETs, regardless of structure, are susceptible to radiation damage from total ionizing dose (TID) and singleevent effects (SEEs), including single-event gate rupture (SEGR) and single-event burnout (SEB). This work presents test data for power MOSFETs of different structures and manufacturers, all evaluated in the past year for potential insertion into space flight missions. Table I provides a list of devices tested. We include SEE test data for two new manufacturers of radiation-hardened VDMOS, Aeroflex Corp. and Solid State Devices, Inc. (SSDI), as well as data for Fuji's 500 V radiation-hardened VDMOS. A Vishay commercial trenchFET® reported on previously [3] is further evaluated. We also present both SEE and TID data for Infmeon Technologies' radiation-hardened SJ MOSFET and a preliminary look at the SEE response of engineering samples of several LDMOS structures from a commercial fabrication process. SEE data on a Texas Instruments commercial NexFET™ lateral-vertical hybrid power MOSFET can be found in [4].

A. Heavy-Jon Testing for SEEs
We gathered heavy-ion test data at the Texas A&M University Cyclotron Facility (T AMU) or at Lawrence Berkeley National Laboratory (LBNL) ( Table II). Fig. 2 shows a diagram of the irradiation test circuit. Six devicesunder-test (DUTs) can be mounted on the test board with daughter cards and individually accessed via dry Reed relays controlled by an Agilent 34907 A data acquisition/switch unit. All terminals of the devices not under test are then floating. Samples were delidded or decapsulated and fully electrically characterized at NASA Goddard Space Flight Center (GSFC). On-site, prior to the first and following each irradiation run, a gate stress test was performed in which the gate leakage current (IGss) was measured as a function of gate voltage at 0 drain-source (V 05 ) bias, and the drain-source breakdown voltage (BVoss) was measured. Measurement equipment included a Keithley 2400 current-voltage sourcing and measurement instrument (SMU) for gate voltage supply and current measurement(< 1 nA accuracy) and either a Keithley ----*RH = Radtatlon hardened; Com -Commercial Samples were irradiated in air at T AMU or in vacuum at LBNL at normal incidence unless othernise indicated. For each sample, the gate-source bias (V os) was held in the off state . and V os was incremented in steps. At each step, the sample was irradiated with a beam flux indicated in Table II until either the sample failed or the maximum run fluence was reached. A post-irradiation gate stress (PIGS) test was then performed to reveal any latent damage to the gate oxide and the BV oss recorded. Failure was defined by a sudden increase in drain-source leakage current during the run or the gate leakage current exceeding the vendor specification during the run or following the PIGS test.

B. Total Ionizing Dose Testing
60 Co TID tests were performed at NASA GSFC's Radiation Effects Facility (REF). Samples were irradiated under one of three different bias conditions: on-state, in which the drain and source were grounded and the gate biased to 60% of maximum on-state bias; off-state, in which the gate and source were grounded and the drain biased at 75% of the rated BV oss; and grounded, in which all nodes were grounded. Additional unirradiated samples were measured between dose steps as controls. Irradiation was performed under high dose rate (HDR) conditions, which was lowered during overnight irradiation periods. Irradiated samples und~ent a one-week biased anneal at room temperature, followed by a HDR irradiation of an additional 50% of the total dose received and then immediately by a one week biased anneal at 100 °C.   protective mode testing under Kr irradiation was performed wherein a 100 kQ or 1 Mn resistor was placed in series with the drain node. Under this protected condition, large transients occurred but DUTs remained undamaged even at the full 100 V os rating under maximum -20 V os bias. It appears, therefore, that this device may be robust to SEGR. Under this maximum bias condition, the SEB cross section was found to be 2.55x lo· 2 cm 2 • Finally, the angular response of this device was determined by positioning the DUT off-normal from the incident beam path and, under the known failing condition of 0 V os and 45 V os but in protected mode (1 00 kn drain resistance), incrementally reducing the tilt angle until nondestructive SEB occurred. For 886 MeV Kr, the DUT passed at a tilt angle of25°, failing at 20°. The Fuji radiation-hardened 500 V, 23 A, 180 mQ planar VDMOS, part number JAXA-R-2SK4188, was evaluated for potential use in a low V os, -13 V os off-state bias application. Note that this bias condition is outside the manufacturer's recommended off-state gate bias with regards to radiation hardness assurance. Samples were provided by Fuji in SMD-2 packages. Fig. 5 shows the SEE response curve for 1289 MeV Ag (LET= 42 MeV -cm 2 /mg, range = 119 J.!m). Three samples were irradiated at -13 V os and the remaining 2 samples were evaluated at 0 V 0 s; unfortunately, the threshold failure V 0 s condition at 0 V os was not found, so in Fig. 6 data at 0 V ~s are for the 2 failing V os, as opposed to the last passing biases. All failures were due to SEGR; BVoss remained in specification in the failed devices.

D. Vishay SUM45N25-58 TrenchFEy®
Initial SEE test results of the Vishay commercial250 V, 45 A, 58 mn trench-gate style MOSFET have been previously reported by the authors for the 0 V GS bias condition [3]. Fig. 6 presents these previous data as well as new data for non-zero V os biases. Three samples each were evaluat(:d at 0 and -5 V os, and 2 samples at -10 V GS· The primary failure mode is likely SEB due to the la. ck of dependence of the threshold failure V os on V os, although in each case the gate was also damaged.
Evaluation of the SEE response at negative gate biases required very low flux to avoid premature failure due to localized total dose effects which caused the drain current to rise during irradiation. In trench-style devices, the orientation of the gate oxide is perpendicular to the surface of the die, increasing the cross-section of vulnerability for a single ion to traverse from source to drain within the oxide. As the ion passes through the oxide, it generates trapped holes along its path, leading to a sudden rise in drain leakage current due to the lowering of the gate threshold voltage at that location within the transistor. This effect has been studied in the literature [5]. Lowering the flux during these tests permitted charge dissipation between ion strikes, reducing the magnitude of the dose-induced drain leakage current.
The range in silicon of Au ions at the Bragg peak is 51 IJ.ffi. Fig. 8 shows the effect of varying the position of the Bragg peak in the device on the threshold V 0 s for SEGR. For these tests, V os was held at -15 V. Due to the small sample size, data for all DUTs are shown. A small effect can be seen, with the worst-case position of the Bragg peak being between 42 1-1m and 58 1-1m below the surface of the DUT. ..

F. Commercial LDMOS
Engineering samples of a commercial 40 V LDMOS varying in geometry were provided for SEE evaluation. Tests were conducted at T AMU under 1170 MeV Ag irradiation (LET = 44 MeV-cm 2 /mg, range= 107 ~m) on samples having a 2.2 1-1m drift length (Fig. ID). At 0 V os. failure due to SEB occurred at 20 V os -22 V os. with 3 samples passing at 18 V 05 . Additional samples were tested at 15° and 30° tilt angles, as well as at 45° tilt and both 0° and 90° roll angles; all samples had the same failure threshold as those irradiated at normal incidence. Upon failure, electrical connection to the MOSFET drain was lost (bond wires were intact), indicating that the metal interconnect had been damaged. PIGS tests To be presented by Jean-Marie Lauenstein at the International Electrical and Electronics Engineering (IEEE) Nuclear and Space Radiation Effects Conference (NSREC) Data Workshop, San Francisco, California, July 11,2013 and published in the IEEE Radiation Effects Data Workshop and on radhome.gsfc.nasa.g ov and nepp.nasa.gov. demonstrated either normal or slightly elevated gate leakage current Ooss).
Tests at LBNL with 1232 MeV Xe (LET = 59 MeV-cm2/mg, range = 90 ~-tm) revealed less failure susceptibility with shorter drift lengths (1.6 ~-tm and 0.6 ~-tm), but the TAMU tests with Ag of the 2.2 ~-tm vs. 0.6 ~-tm drift lengths did not support this fmding. Further evaluation of these devices will be conducted to establish trends and angular responses as a function of drift length. Results on the Vishay 250 V SUM45N25-58 were reported previously [6]. TID evaluations of the SSDI power MOSFET are underway; data will be reported in [7].
A. Infineon BUY25CS54A SJ MOSFET Samples of the BUY25CS54 were provided by Infmeon in SMD-2 packages. The dose rate was approximately 950 rad(Si)/min, with lower rates of 26.4 rad(Si)/min and 98.5 rad(Si)/min during two overnight irradiation steps (125 krad(Si) to 150 krad(Si) and 300 krad(Si) to 400 krad(Si), respectively). Bias conditions were as described in section II.B. Key results are shown in Figs. 9-10.
The worst-case bias condition at high total dose was the onstate bias. All measured parameters of the 5 samples irradiated under this condition stayed within datasheet specification up to 400 krad(Si). As shown in Fig. 9, the non-annealed gate threshold voltage (V GS(th)) fell below datasheet specification between 400 krad(Si) and 500 krad(Si). V GS(thl returned to within specification during a 168-hr room-temperature (RT) biased anneal that was performed following 500 krad(Si) total dose. All other static parameters remained within specification up to the maximum 500 krad(Si) ionizing dose exposure, with some degradation of the drain leakage current (loss) as shown in Fig. 10.
The 3 samples per condition irradiated either in the off-state or grounded state remained in specification for all measured parameters at the maximum 500 krad(Si) total accumulated dose, with some degradation ofVos(th) and loss· z:   Over the past few years, the viable options of power MOSFETs for space applications have increased, with new manufacturers of radiation-hardened devices emerging. In addition, goals have been completed or are under way to bring state-of-the-art topologies such as superjunction, trench, and lateral power MOSFETs into the aerospace market. This work provides initial radiation test data on some of the more recent offerings.
To be presented by Jean-Marie Lauenstein at the International Electrical and Electronics Engineering (IEEE) Nuclear and Space Radiation Effects Conference (NSREC) Data Workshop, San Francisco, California, July 11,2013 and published in the IEEE Radiation Effects Data Workshop and on radhome.gsfc.nasa.gov and nepp.nasa.gov.