1536 IEEE Transactions on Nuclear Science, Vol. 35, No.

6, December 1988

DOSERATE EFFECTS ON THE TOTAL-DOSE

THRESHOLD-VOLTAGE SHIFT OF POWER MOSFETS

R.D. SchrimpP, P.J. Wahle*, R.C. Andrews+,

D.B. Cooper*, and K.F. Galloway*

*Department of Electrical and Computer Engineering

University of Arizona
Tucson, A2 85721

+Government Electronics Group-Aerospace Operations

Motorola, Inc.
Chandler, A2 85248

ABSTRACT Some MOSFET technologies have been found to exhibit

super-recovery effects, but the uncertainty in the physical
Radiation-hardened and unhardened N-channel power mechanisms involved makes it difficult to determine a priori
MOSFETs have been tested at dose-rates approaching space- whether a specific technology will exhibit this behavior. Power
radiation levels. The hardened parts exhibited large super- MOSFETs usually use thicker gate oxides than are commonly
recovery effects during and after very low total doses of employed in radiation-hardened integrated MOS technologies.
ionizing radiation. The super-recovery was attributed to in- The thicker gate oxide may change the time scale for which
situ interface-trap formation. These positive threshold shifts any time-dependent effects are observed. In addition, the
and the accompanying interface traps can reduce current-drive double-diffused technology (DMOS) [4] used to fabricate most
capability in power MOSFETs. Impact on space-system use power MOSFETs produces devices with a complex lateral
is discussed. The threshold-voltage shift of the positively doping distribution in the channel region which is significantly
biased unhardened parts was dominated by generation of oxide different from that obtained in most low-power MOSFETs.
trapped charge at all dose-rates; the threshold voltage of
these parts decreased monotonically at all doses and dose-rates While many logic circuits are somewhat tolerant of a small
examined. positive shift in threshold, this behavior must be examined
more closely in power devices. One of the most important
power MOSFET parameters is the maximum amount of
I. INTRODUCTION current that the device is capable of driving, and this is
adversely affected by a positive shift in threshold voltage. It
Power metal-oxide-semiconductor field-effect transistors
(MOSFETs) are very attractive for use in spaceborne is also important to determine whether any observed super-
applications because of their faster switching speeds recovery is dominated by oxide-trapped-charge annealing or
and simpler drive requirements than comparable bipolar by generation of interface traps. The presence of charge in the
interfacial region can degrade carrier mobility. Interface-trap
transistors. This allows reductions in the size and weight
generation can thus lead to significant reductions in current-
of spaceborne equipment. Operation in space environments
drive capability.
involves exposure to ionixing radiation at very low dose-rates,
but radiation testing of these parts is typically conducted at
relatively high dose-rates to minimize the testing time required. This paper reports an experiment designed to examine total-
dose effects on power MOSFETs as a function of dose-rate.
The radiation-induced threshold-voltage shift in MOSFETs is Hardened and unhardened parts were tested. Dose-rates
caused by generation of positive oxide trapped charge and approaching those encountered in space environments were
interface traps that contribute negative charge in N-channel included.
devices. The initial post-irradiation threshold shift at typical
dose-rates is usually dominated by oxide trapped charge and is
negative, but during annealing, the threshold may recover to, 11. EXPERIMENTAL DETAILS
or even surpass, its pre-irradiation value [l-3). This recovery
can be caused by annealing of oxide trapped charge or long- Radiation-hardened and unhardened N-channel DMOS power
term buildup of interface traps. At low dose-rates, it is possible transistors were exposed to ionizing radiation under several
for this super-recovery phenomenon (31 to take place during the bias conditions. The radiation-hardened parts were provided
irradiation, causing a net positive threshold shift instead of the by International Rectifier and carried the part number IRH-254
negative shift commonly observed immediately following high- [5]. The gate-oxide thickness of these parts was approximately
dose-rate irradiation. This is because the formation of interface 100 nm. The unhardened parts were provided by two different
traps and/or oxide-trapped-charge annealing can take place in- manufacturers and carried the industry-wide part numbers
situ when the irradiation time is long. IRF-150 and IRF-440. The gate-oxide thickness for these parts
was also approximately 100 nm. All devices were packaged in
TO-3 packages.

Irradiations were carried out in two different Cobalt-60 sources.

Experiments in the first source were at dose-rates of 3.3
Work supported in part by Motorola, Inc., Government rad(Si)/s and 0.83 rad(Si)/s, while experiments in the second
Electronics Group. source were at 0.012 rad(Si)/s (equivalent to approximately
0018-9499/88/1uK)-15363o1.oo 0 1988 IEEE

Authorized licensed use limited to: Naval Surface Warfare Center (Dahlgren). Downloaded on May 17,2021 at 18:26:10 UTC from IEEE Xplore. Restrictions apply.
 1537

1 krad(Si)/day). The sources were calibrated using FWT- All the devices tested initially had very well defined
60-00 radiochromic film dosimeters from the National Bureau si1bt hres hold characteristics. At high tot al-dose levels,
of Standards. The film was placed between two AI plates to however, the subthreshold characteristics of devices are
obtain a dose equivalent to that received in Si. The distance typically difficult to describe in terms of a single slope. This
from source to sample was varied to control the dose-rate. behavior was seen in both the hardened and unhardened parts
Total dose was calculated from the dose-rate and exposure tested here. The plot of interface and oxide trapped charge
time. The known decay rate of the Cobalt-60 source was presented later (Fig. 3) includes data only up to the point at
accounted for ill determination of the dose-rate. which the subthreshold characteristics can still be well defined
by a single slope. A typical pre-irradiation subthreshold curve
Dosimetry errors will not affect the results qualitatively; for obtained from an IRH-254 part is shown in Fig. 1, along with
example, devices nominally irradiated at 0.83 rad(Si)/s were the curves obtained from the same part after exposure to 50
placed further from the source than those irradiated at 3.3 krad(Si) and 100 krad(Si) at a dose-rate of 0.012 rad(Si)/s.
rad(Si)/s. While the exact value of the dose-rate is subject
to error, the relative dose-rates for the experiments are not. At the lowest dose-rate, device characterization was performed
Experiments in the source have been repeated to establish in-situ. In all other cases, device characterization was
reproducibility. Dose enhancement is not expected to be an performed immediately following irradiation. After completion
issue, because high-energy Cobalt-60 photons were used for of measurements on all devices (about thirty minutes), the first
irradiation, and no heavy metals were present in the packages. device was re-characterized. In all cases, the second set of
Under these conditions, dose enhancement is expected to be data agreed with the first. This indicates that any long-term
essentially negligible 16). No Pb/Al shields were used. buildup of interface traps or annealing of oxide trapped charge
was not significant for the time scale involved in making the
Devices were tested with the source and drain terminals measurements.
grounded, and with the gate bias at +9 V, -9 V, or 0
V. The temperature was monitored during all irradiations 111. RESULTS-HARDENED PARTS
and remained about ten degrees Centigrade above room
temperature. The slight elevation in temperature was caused The total-dose threshold-voltage response of the International
by power dissipation in a concurrent experiment t o examine Rectifier IRII-254 radiation-hardened power MOSFETs is
total-dose effects on power MOSFETs that are switching a load displayed in Fig. 2. These curves present results obtained
current. from the devices irradiated under positive gate bias at dose-
rates of 200,3.3,0.83, and 0.012 rad(Si)/s. The data presented
The threshold-voltage response t o total dose was measured for for the highest dose-rate were obtained from the International
all devices. Electrical characterization was performed using Rectifier data sheet [5]. The other data points represent
an HP 4345B semiconductor parameter analyzer. Threshold the average of two devices. In no case was there a serious
measurements were obtained by extrapolating the square-root- discrepancy between two parts irradiated at the same dose-
O f - 1 0 vs. VG curve to 10 = 0. The total-dose response rate. Representative error bars have been included to indicate
determined using this method was qualitatively the same as the amount of difference between the two transistors irradiated
that obtained by defining the threshold voltage at 1 mA. The at the same dose-rate.
threshold shift due t o interface trapped charge was separated
from that due to oxide trapped charge using subthreshold A strong super-recovery effect is apparent in the curves of
measurements and the technique proposed by McWhorter and Fig. 2. At all dose-rates, an initial decrease in threshold
Winokur [7]. voltage was observed, followed by recovery. At the highest
dose-rate, the threshold voltage did not recover to its original
1.0 I 1 I I I
1
n
U lo-* n IRH 254
> - Positive Bias
3 U

G 10-~ -4
Q) *
$4

51

- 1 -
2 10-'0

n
0-l2 213 4 5
c
-1 J 0 1 E-.

Gate Voltage [ V] 'I'otal Dose 1. k r a d ( Si)]

Fig. 1 Drain-source current vs. applied gate voltage for Fig. 2 Threshold-voltage shift vs. total dose for positively
an IRH-254 N-Channel power MOSFET. The pre-irradiation biased IRII-254 rad-hard power MOSFETS. The results were
curve is shown, as well as curves obtained following total doses obtained at four different dose-rates, with the highest dose-rate
of 50 krad(Si) and 100 krad(Si). The dose-rate was 0.012 rad data coming from the IRH-254 data sheet [Fi].
(Si)/s.

Authorized licensed use limited to: Naval Surface Warfare Center (Dahlgren). Downloaded on May 17,2021 at 18:26:10 UTC from IEEE Xplore. Restrictions apply.
 1538

value until a total dose of approximately 800 krad(Si) had been I

accumulated. In contrast, at the lowest dose-rate, a minimal
decrease was seen initially, and the threshold voltage became 1RI-I 254
higher than its initial value before a total dose of 10 krad(Si) Negative B i n s

had been received.

The threshold-voltage shifts plotted in Fig. 2 have been

separated into the components due to interface trapped charge
LI -_-
and to oxide trapped charge. These data are presented
in Fig. 3 for dose-rates of 3.3, 0.83, and 0.012 rad(Si)/s.
It can be seen that the shift due to oxide trapped charge 33 r a d / h
at the three dose-rates is approximately the same, with
the difference in net threshold shift primarily resulting from
differences in the density of interface trapped charge. This
indicates that the observed rebound may be tied to in-situ
interface-trap formation rather than annealing of oxide trapped
charge. While these curves represent the average obtained from
two devices, the difference between devices irradiated at the
same dose-rate was very small. Despite this small device-to-
device variation, curves of this type must be interpreted with
caution because of uncertainty surrounding charge-separation Fig. 4 Threshold-voltage shift vs. total dose for negatively
techniques. biased IRH-254 rad-hard power MOSFETS.

to IV. RESULTS-UNHARDENED PARTS

The change in threshold voltage versus total dose for the

unhardened devices (part number IRF-440) with positive
5 applied gate bias is displayed in Fig. 5. Results are given
for dose-rates of 3.3, 0.83, and 0.012 rad(Si)/s. The data
represent the average of two devices at each dose-rate. The
threshold shift was negative at all total-dose levels and for
0 all dose-rates. The devices irradiated at the lower dose-rates,
however, experienced less decrease in threshold voltage than
those irradiated at the higher rate.

-5 Since the threshold-voltage shift of these unhardened parts

decreased monotonically, oxide trapped charge was dominant.
Any in-situ formation of interface traps played a relatively
- 10
1 10 100 1000

Total Dose ItracI(~i)l

0
n
Fig. 3 Threshold shift due to interface trapped charge and 3
oxide trapped charge for the devices considered in Fig. 2. U
c
These curves were obtained using the subthreshold slope c,
technique proposed by McWhorter and Winokur 17). v - - 0
.d
2
rn
-20
2
Figure 4 displays the change in threshold voltage vs. total dose, 2 -30
with dose-rate as a parameter, for devices biased negatively. It r
n
a, IRF 440
is important to note that the experimental data represented in L -40
Fig. 4 were obtained using a single device at each dose-rate.
Very little difference is observed between the curves obtained 5 - 50
Positive Bias

at different dose-rates, indicating time-dependent effects were 10 100 1( 10

negligible. Devices irradiated with the gate grounded also
failed to exhibit any large time-dependent effects. The Total Dose [ k r a d ( Si)
significant time-dependent behavior of the positively biased
parts was attributed to interface-trap buildup. The lack
of time-dependent effects in the threshold-voltage shift of
grounded and negatively biased devices is consistent with the Fig. 5 Threshold-voltage shift vs. total dose for IRF-440
suggestion that positive bias is required to complete iiiterface- unhardened power MOSFETS. The behavior is dominated by
trap formation [8]. the presence of large amounts of oxide trapped charge.

Authorized licensed use limited to: Naval Surface Warfare Center (Dahlgren). Downloaded on May 17,2021 at 18:26:10 UTC from IEEE Xplore. Restrictions apply.
 1539

small role in the performance of these parts. Within the total dose is accumulated rapidly, and elevated-temperature
range of dose-rates examined here, a single curve provided a annealing is used to speed up any time-dependent effects. It
reasonable picture of the threshold-voltage shift of these parts, is important for users of these power MOSFETs, however, to
indicating that any annealing of oxide trapped charge was also recognize that the type of threshold response shown in Fig.
relatively minor on these time scales. Results obtained from 2 is what will actually be exhibited at low dose-rates. The
IRF-150 power MOSFETs provided by another manufacturer possibility of a positive threshold shift must be accounted for
were qualitatively very similar to those displayed in Fig. 5. in circuits designed for low-dose-rate environments.

No significant time-dependent effects were observed in parts The threshold voltage of the unhardened parts tested here
irradiated with negative gate bias or with the gate grounded. was qualitatively the same at all dose-rates. The threshold-
This result is consistent with those obtained from the hardened voltage response of these parts was dominated by the effects
devices. In the unhardened power MOSFETs tested here, of oxide trapped charge. The oxide trapped charge forms
as well as the hardened parts, lack of positive bias during rapidly, and thus is less sensitive to dose-rate than is interface
irradiation minimized time-dependent effects on threshold trapped charge. No in-situ annealing of oxide trapped charge
voltage. was observed up to the total-dose levels at which charge
separation was still useful. Unhardened devices irradiated
V. DISCUSSION AND SUMMARY with negative gate bias or with the gate grounded did not
exhibit any significant time-dependent effects. Thus, the high
Time-dependent effectson MOSFET threshold shift have been dose-rates typical of laboratory testing may be adequate for
interpreted as either true dose-rate dependence I11 or apparent determining the space-radiation hardness of some unhardened
dose-rate dependence 121. If the dependence is true, this power MOSFETs, with no special techniques used to account
means that Ihe charge formation is a dose-rate dependent for time-dependent effects.
physical process. If the dose-rate dependence is the result of
in-situ interface-trap formation and/or oxide-trapped-charge The present work demonstrates that the total-dose threshold-
annealing that takes place on a long time scale during low- voltage response of one family of hardened power MOSFETs
dose-rate irradiations, then the effect is apparent, not true. exhibits super-recovery when the devices are irradiated with
positive gate bias. The implications are very significant for
To determine whether the dose-rate dependence observed space-system designers. Test results obtained immediately
in the hardened power MOSFETs tested here was true or following high-dose-rate irradiation may indicate that the
apparent, an experiment was initiated t o examine the effects of threshold voltage will shift negative, while the actual shift may
post-irradiation annealing on the threshold voltage. Positively be positive in a space environment. This must be accounted
biased IRII-254 transistors were irradiated at approximately for in circuit designs and in the calculation of design margins.
6.7 rad(Si)/s to a total dose of 50 krad(Si). The negative The super-recovery was attributed to in-situ interface-trap
shift in threshold voltage at this dose was slightly more than formation, and the effects of this charge on current-drive
one volt. The devices were then allowed to anneal at room capability should also be considered.
temperature with-the bias maintained. Immediately following
removal from the radiation source, the threshold voltage began
to increase. The threshold voltage was measured at total times
(irradiation time plus anneal time) equal to the irradiation time
required to accumulate the same total dose (50 krad(Si)) at 3.3 Acknowledvments
rad(Si)/s and 0.83 rad(Si)/s.
The authors would like to acknowledge the contributions of
The threshold-voltage shifts following the higher dose-rate Ron Myers and Dan Wallace of Motorola; Fred Rubi, Dill
irradiation plus anneal matched those obtained at the lower Cordon, Stuart Litwin, and Martin Hummel of International
dose-rates (for corresponding total time) to within the Rectifier; Ron Pease of MRC; and George Nelson and Bill
estimated accuracy of this experiment. This result suggests Weber of the University of Arizona. Very useful suggestions
that the time-dependent effects observed in these power from a conscientious group of reviewers are also appreciated.
MOSFETs, in agreement with Ref. 2, are attributable to
apparent dose-rate dependence. The results presented in
Section 111 indicate that in-situ buildup of interface traps,
rather than annealing of oxide trapped charge, was responsible
for the super-recovery. The apparent dose-rate dependence VI. REFERENCES
is expected to saturate for very low dose-rates 191, because
all interface traps will have time to form in-situ. For the 1. P.S. Winokur, F.W. Sexton, G.L. Hash, and D.C.
hardened parts tested here, saturation of the effect had not Turpin, “Total-Dose Failure Mechanisms of Integrated
yet occurred at 0.83 rad(Si)/s because the response obtained Circuits in Laboratory and Space Environments,”
at 0.012 rad(Si)/s was significantly different from that obtained IEEE Trans. Nucl. Sci., NS-34, pp. 1448-1454 (1987).
at 0.83 rad(Si)/s. This may be a result of the relatively thick
gate oxide employed in power transistors; in CMOS parts, the 2. D.M. Fleetwood, P.S. Winokur, and J.R. Schwank,
dose-rate effect may saturate at the dose-rates used here. “Using Laboratory X-Ray and Cobalt-60 Irradiations t o
Predict CMOS Device Response in Strategic and Space
The results presented show that it is necessary to test a t Environments,” presented at the IEEE Nuclear and Space
realistic dose-rates if accurate results are to be obtained for the Radiation Effects Conference, Portland, OR (1988).
hardened power MOSFETs tested in this work. Alternatively,
rebound testing may be used if it can be demonstrated that 3. A.H. Johnson, “Super Recovery of Total Dose Damage in
results for power MOSFETs correspond to those obtained MOS Devices,” IEEE Trans. Nucl. Sci., NS-31, pp. 1427-
from CMOS transistors 121. In rebound testing, the desired 1433 (1984).

Authorized licensed use limited to: Naval Surface Warfare Center (Dahlgren). Downloaded on May 17,2021 at 18:26:10 UTC from IEEE Xplore. Restrictions apply.
 1540

4. B.J. Baliga, Modern Power Devices, New York: Wiley- 7. P.J. McWhorther and P.S. Winokur, “Simple Technique
Interscience, 1987, Ch. 6, pp. 270-272. for Separating the Effects of Interface Traps and Trapped-
Oxide Charge in Metal-Oxide-Semiconductor Transistors,”
5. International Rectifier Preliminary Data Sheet No. APDI. Phvs. Lett., 48, pp. 133-135 (1986).
PD-9.477A, IRH-254 N-Channel Rad Hard HEXFET,
El Segundo, CA (1987). 8. F.B. McLean, “A Framework for Understanding Radiation-
Induced Interface States in Si02 MOS Structures,”
6. F.B. McLean and T.R. Oldham, “Basic Mechanisms of IEEE Trans. Nucl. Sci., NS-27, pp. 1651-1657 (1980).
Radiation Effects in Electronic Materials and Devices,”
Harry Diamond Laboratories Report HDL-TR-2129, p. 44 9. R.L. Pease, private communication.
(1987).

Authorized licensed use limited to: Naval Surface Warfare Center (Dahlgren). Downloaded on May 17,2021 at 18:26:10 UTC from IEEE Xplore. Restrictions apply.