IEEE TRANSACTIONS ON NUCLEAR SCIENCE, VOL. 68, NO.

5, MAY 2021 651

Ion-Induced Mesoplasma Formation and Thermal

Destruction in 4H-SiC Power MOSFET Devices
Joseph A. McPherson, Student Member, IEEE, Collin W. Hitchcock,
T. Paul Chow , Life Fellow, IEEE, Wei Ji , Member, IEEE,
and Andrew A. Woodworth

Abstract— Both experiments and simulations have shown that on device operation in the radiation environments. Through
single-event burnout (SEB), a catastrophic event, occurs at less experiments, it has been found that SEB can occur at less than
than half of the rated blocking voltage in commercial 4H-SiC 50% of a device-rated blocking voltage [1]–[5]. Since the SEB
power devices under a heavy-ion strike. The failure was shown
to be due to significant impact ionization near the epi/substrate causes device thermal destruction, understanding the failure
interface. Adding a buffer layer between the drift epi and mechanism will facilitate the development of more radiation-
substrate layers reduces the impact ionization effect and changes tolerant designs and overcome this performance limitation.
the thermal failure location. In this article, the SEB phenomenon Much effort has been pursued to understand the mechanisms
in a 4H-SiC power MOSFET utilizing a buffer layer is inves- that triggered SEB in SiC power devices when bombarded
tigated. Heavy-ion transport and 3-D electro-thermal transient
simulations were performed to study the device response to a with radiation [6]–[10]. Different mechanisms were proposed
heavy-ion strike. In examining the time evolution of electric field in previous studies. Kuboyama et al. [6], [7] suggested that
profile, charge carrier dynamics, and thermal heat transfer, it is the trap-assisted tunneling and thermionic emission are the
determined that the failure mode for this design is the location major regenerative sources to produce excess carrier currents
shift of the mesoplasma (or hot spot) to within the drift epi and lead to thermal runaway in SiC Schottky diodes. The
region, away from the high field area. A sensitivity analysis was
conducted to identify the dominant electrical or thermal factors failure occurs at the Schottky contact in Schottky diodes.
contributing to device failure due to second breakdown. From Witulski et al. [8] demonstrated that the triggering of the
these simulations, it is found that the semiconductor thermal parasitic bipolar junction transistor (BJT) and the high-
conductivity is the primary material parameter that influences field-induced impact ionization at the epitaxial–drain junction
the mesoplasma formation. cause the failure in SiC power MOSFETs, but the failure
Index Terms— Heavy ions, mesoplasma, silicon carbide power location was not clearly identified. Our previous work [9]
devices, single-event burnout (SEB), single-event effects (SEEs). on commercial SiC MOSFETs found that the failure mode
comes from intense localized heating between the epitaxial
I. I NTRODUCTION and substrate layers, where SEB occurs. This heating is the

P OWER devices based on 4H-SiC have numerous advan-

tages over their silicon counterparts, including but not
limited to operating at higher temperatures and higher blocking
result of the generated electron–hole pairs from the heavy ion
enhancing the electric field at the interface, which causes a
self-sustained impact ionization event. Different from the work
voltages with lower specific ON-state resistance. Due to the of [8], our finding showed that the device reaches burnout
larger bandgap of SiC, it was expected that SiC power devices before the BJT could turn on and the electron–hole plasma
have a higher tolerance to radiation, but experiments have rapidly overwhelms the N+ source/P body and P body/N epi
shown that their performance suffers from single-event effects junctions, causing a short to form between the drain and source
(SEEs). These effects can range from permanent increase in contacts regardless of heavy-ion strike locations. The work of
blocking leakage current to catastrophic device destruction Ball et al. [10] also verifies that device burnout happens before
due to a single high-energy particle strike. The latter event the BJT could turn on.
is called single-event burnout (SEB) and is the upper limit To enhance the radiation hardness, structures were proposed
that consisted of a buffer layer inserted between the drift
Manuscript received February 24, 2021; accepted March 17, 2021. Date
of publication March 23, 2021; date of current version May 20, 2021. This epi and substrate, which was shown, through simulations,
work was supported by an Early Stage Innovations grant from NASA’s Space to substantially change the device response to SEB [9]. The
Technology Research Grants Program.(Corresponding author: Wei Ji.) response was found commonly altered in such a way to suggest
Joseph A. McPherson and Wei Ji are with the Department of Mechanical,
Aerospace, and Nuclear Engineering, Rensselaer Polytechnic Institute, Troy, a different mechanism is at play: a mesoplasma formed deep
NY 12180 USA (e-mail: [email protected]). within the drift epi layer.
Collin W. Hitchcock and T. Paul Chow are with the Department of Elec- Mesoplasma phenomena have been observed experimentally
trical, Computer, and Systems Engineering, Rensselaer Polytechnic Institute,
Troy, NY 12180 USA. before, and theories have been formulated to explain the effects
Andrew A. Woodworth is with the NASA Glenn Research Center, observed in silicon power devices [11]–[14]. Formation of the
Cleveland, OH 44135 USA. mesoplasma is the last step of the second breakdown process
Color versions of one or more figures in this article are available at
https://doi.org/10.1109/TNS.2021.3068196. before thermal destruction of the device and occurs when
Digital Object Identifier 10.1109/TNS.2021.3068196 a current filament (called microplasma) becomes thermally
0018-9499 © 2021 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See https://www.ieee.org/publications/rights/index.html for more information.

Authorized licensed use limited to: SICHUAN UNIVERSITY. Downloaded on December 22,2023 at 12:05:19 UTC from IEEE Xplore. Restrictions apply.
652 IEEE TRANSACTIONS ON NUCLEAR SCIENCE, VOL. 68, NO. 5, MAY 2021

unstable and collapses on itself. This collapsing filament forms

a hot spot (or mesoplasma) that is capable of reaching the
decomposition (melting, sublimation, etc.) temperature of the
semiconductor material, resulting in irreversible damage [11].
This mesoplasma formation and second breakdown process
can happen in power semiconductor devices under harsh
(overvoltage or overtemperature) or radiation conditions. Here,
the heavy ion interacts with SiC material, generating an
electron–hole plasma initiated as a current filament along its
transversing path.
Our primary focus here is to demonstrate and describe,
through simulations, the details of the mesoplasma formation
process in a vertical SiC planar power double-implanted
metal oxide semiconductor field-effect transistor (DMOSFET)
with a uniform buffer layer. A general-purpose Monte Fig. 1. Intrinsic temperature (Ti ) for 4H-SiC where thermally generated
carriers equal epitaxial doping levels.
Carlo N-Particle transport code [15] and a 3-D TCAD
electro-thermal simulator [16] are employed for the study.
In addition, several physical parameters were explored to
determine their impact on the mesoplasma formation. Deter-
mining how and where the mesoplasma forms and its evolution
may allow for future designs that can help mitigate or suppress
its formation and increase the SEB threshold voltage to the
rated blocking voltage of the SiC transistor.

II. T HEORY OF M ESOPLASMA P HYSICS

The presence of a current filament (or microplasma) in a
semiconductor device does not always lead to a mesoplasma,
which further causes second breakdown in a device [11].
However, the current filament can suffer from thermal insta-
bility, which will give rise to the mesoplasma formation. This
condition is reached when the thermal generation, instead of
impact ionization, of carriers becomes the dominant source Fig. 2. Thermal conductivity of 4H-SiC as a function of temperature [17].
of carrier production. When this occurs, the process becomes Experimental data range is from 300 to 2300 K with the rest of the graph
regenerative since the thermal generation of carriers causes extrapolated to 3000 K from the equation above.
an increase in the local current density, together with the
concomitant decrease in thermal conductivity. This additional III. M ESOPLASMA B EHAVIOR
current density leads to further heating of the lattice and A. Device Modeling
creates additional carriers from the thermal process. Even-
The basic structure of the 1200-V vertical SiC power
tually, the device will fail because of thermal runaway. The
DMOSFET used for the SEB simulations is based on a com-
temperature at which this process commences is at the intrinsic
mercially designed device [18]. The model is a full 3-D struc-
temperature where the intrinsic carrier density equals the
ture that follows a stripped geometry for the source and gate
background doping concentration.
contacts, has a pitch of 9.5 μm, and an epitaxial layer thickness
Fig. 1 shows the intrinsic temperature for a range of epi-
of 10 μm with a doping concentration of 6 × 1015 cm−3 .
taxial doping concentrations, derived from imbedded equation
A buffer layer with a thickness of 2 μm was inserted between
by enforcing the thermally generated intrinsic carrier density
the epi and substrate layers with a uniform doping concentra-
n i (T ) equal to the background doping. Once the lattice tem-
tion of 4 × 1016 cm−3 . This configuration shows a breakdown
perature exceeds the line in figure, the carrier concentration is
voltage of 1863 V and an RON ,sp of 3.3 m-cm2 , which is
dominated by the thermally generated carriers and increases
more than enough for a 1200-V rated power MOSFET. With
the probability of mesoplasma formation.
a substrate thickness of 40 μm, the model’s total thickness
The thermal generation of carriers is further enhanced from
is 52 μm. Fig. 3 shows the schematic cross section of the
the inverse relationship between thermal conductivity and
simulated MOSFET.
temperature for SiC. At room temperature, the thermal con-
ductivity of SiC is higher than 3 W(cm-K)−1 , but rapidly drops
B. Physical Parameter Modeling
off with increasing temperature, to less than 0.25 W(cm-K)−1
at 1500 K, as shown in Fig. 2 [17]. This reduction in the During an SEB event caused by a heavy-ion strike, signif-
thermal conductivity allows heat to more readily build up since icant heating occurs within the semiconductor material and
it cannot be removed as quickly as before, resulting in an the device will ultimately reach the decomposition temper-
accelerated thermal runaway. ature of SiC which is approximately 3000 K. Due to this

Authorized licensed use limited to: SICHUAN UNIVERSITY. Downloaded on December 22,2023 at 12:05:19 UTC from IEEE Xplore. Restrictions apply.
MCPHERSON et al.: ION-INDUCED MESOPLASMA FORMATION AND THERMAL DESTRUCTION 653

TABLE I
E XPERIMENTAL T EMPERATURE R ANGES FOR P HYSICAL M ODELS

response, temperature-dependent models of several physical

parameters are needed to simulate the SEB event properly.
These parameters include thermal conductivity, lattice heat
capacity, bandgap, and impact ionization. Experiments have
been conducted to study the temperature dependence for the
Fig. 3. 2-D cross section of the 3-D power MOSFET model used for SEB
above parameters and Table I shows the temperature range simulations.
explored for each parameter [17], [19]–[21]. During an SEB
simulation, the data is extrapolated if the lattice temperature
exceeds the maximum experimental range needed. distribution is created and converted to a spatial electron–hole
These parameters are essential for modeling the electro- pair density, which was further fit to a double Gaussian dis-
thermal transient process and the resultant mesoplasma forma- tribution that the 3-D TCAD electro-thermal device simulator
tion since they control the thermal characteristics of SiC and can use [22].
the generation behavior of the carriers. The two parameters Previous simulation work for a power MOSFET without
that control the thermal aspects of SiC are thermal conductiv- the buffer layer has found the failure mode to be meso-
ity and lattice heat capacity. Thermal conductivity degrades plasma formation at the epi/substrate interface [9]. The failure
rapidly with increasing temperature (Fig. 2) [17], allowing mechanism was identified as being caused by the carriers
heat to build up more rapidly, resulting in an accelerated generated from the ion strike enhancing the electric field at the
thermal generation of carriers. Lattice heat capacity describes epi/substrate interface. This modulation resulted in the electric
the amount of heat that is needed to raise the temperature field reaching a peak value greater than 3 MV/cm. This value
of the material, and unlike the thermal conductivity, its value was high enough to create significant impact ionization, with
increases as the temperature increases. This behavior causes it the electric field profile remaining stable until the device failed
to counteract the effects of the thermal conductivity since more thermally. It was found that the time-evolved electric field
energy must be imparted into the lattice to raise its tempera- profile can be improved and its peak suppressed by introducing
ture. Two parameters control the generation rate of the carriers: a uniformly doped buffer layer between the epi and substrate
the bandgap and impact ionization. First, the bandgap narrows layers.
as temperature increases, lowering the energy necessary to To explore further the failure mode for a MOSFET utilizing
promote valance electrons into the conduction band, causing a buffer layer, a silver (Ag) ion with an energy of 1289 MeV
thermal generation of carriers to increase. This effect accel- was chosen for the SEB simulations. This energy gives a linear
erates the thermal generation of carriers further. Second, the energy transfer (LET) value of 46.1 MeV-cm2 /mg, which is
impact ionization rate decreases with increasing temperature, enough to allow the ion to completely traverse both the epi
because of the increase in carrier–phonon collision probability. layer and the buffer layer of the MOSFET. The strike location
It is harder for carriers to achieve a high enough energy to was chosen to be the center of the source contact, and the
cause further ionization at a given electric field value. This strike angle was perpendicular to the surface of the device.
causes the impact ionization event to be quenched at high When a temperature of 3000 K was detected, the simulation
temperatures, resulting in the dominant carrier source being was terminated due to the decomposition temperature of SiC
the thermal generation of carriers. being attained, and SEB was declared. In addition to SEB,
single-event gate rupture (SEGR) can be a limiting factor for
SiC MOSFETs when heavy ions strike at the gate. This event
C. Heavy-Ion Strike Simulations was not analyzed in this article.
To simulate the SEB event, a 3-D TCAD electro-thermal Fig. 4 shows that the buffer layer design experiences SEB
device simulator, interfaced with a Monte Carlo radiation between 900 and 925 V. The bias that will be investigated
transport code, was used. The Monte Carlo transport code was further will be the 925-V case since this is the threshold
employed to simulate the initial heavy-ion strike event as the voltage where SEB was declared. As the heavy-ion traverses
boundary condition. From the simulation, the spatial energy the device, it generates electron–hole pairs as it interacts with

Authorized licensed use limited to: SICHUAN UNIVERSITY. Downloaded on December 22,2023 at 12:05:19 UTC from IEEE Xplore. Restrictions apply.
654 IEEE TRANSACTIONS ON NUCLEAR SCIENCE, VOL. 68, NO. 5, MAY 2021

Fig. 4. Peak temperature plot for a MOSFET using a uniform buffer layer
when struck by a silver ion with LET = 46.1 MeV-cm2 /mg.

the material. The resulting profile of the generated carriers is

0.5 μm wide and extends the entire length of the simulated
device with the peak generation rate occurring at 10 ps.
By 20 ps, the heavy ion has finished generating carriers and
has an average concentration above 1 × 1020 cm−3 . During
and after the generation process, the electrons and holes move
toward the drain and source contacts, respectively, and give
rise to a current filament that penetrates the epi and buffer
layers, as shown in Fig. 5(a). As the electro-thermal transient
progresses, the current filament begins to collapse on itself
with the ends of the filament continuously moving toward the
center of the filament, as shown in Fig. 5(b) and (c). By 240 ps,
the filament collapses vertically on itself with the location of
the peak being 3.8 μm from the source contact. Coincidentally,
this location was also where the lattice temperature reaches
3000 K shortly after 240 ps. From the behavior exhibited
by the current filament and the place where decomposition
temperature occurs, it was declared that a mesoplasma has
formed, and the shape of the mesoplasma is elliptical.
To further illuminate the mesoplasma formation, time
Fig. 5. Total current density distribution over a 2-D cross-sectional area
evolutions of the electric field, electron density, hole den- aligned with ion strike path with the device biased at 925 V at different times:
sity, electron current density, hole current density, and the (a) 20 ps into the simulation, (b) 100 ps, and (c) right before the device fails
lattice temperature were extracted from the 3-D TCAD at 240 ps.
electro-thermal simulator and are shown in Figs. 6–8. At 20 ps,
the electron–hole density and electron current density have
reached their maximum values of 1.2 × 1020 cm−3 and electrons to the drain and holes to the source. This effect also
1.5 × 108 A-cm−2 , respectively. The holes have nearly the helps in sustaining the high carrier concentration. At 240 ps
exact behavior for the density due to space-charge neutrality, into the simulation, the lattice temperature is right on the
but the current density is two-thirds of the electron current cusp of reaching the 3000 K burnout criteria at 3.8 μm away
density. The electric field forms a spike right under the from the source contact and that the electron–hole, electron
source contact at this time but is short-lived. At 100 ps, current, and hole current peak densities occur at this location
the area under the source contact and next to the epi/substrate and have values of 6 × 1019 cm−3 , 0.4 × 108 A-cm−2 , and
interface is depleted of carriers rapidly. However, the region 0.1 × 108 A-cm−2 , respectively. Between the 20 and 240 ps,
roughly 3.3 μm away from the source contact maintains a the current density values are decreasing, however, at the
high concentration of 8 × 1019 cm−3 . This concentration same time the thermal conductivity is rapidly decreasing due
value can be explained by examining the lattice temperature to the increase in lattice temperature at this location. This
graph (Fig. 6). At 3.3 μm, the temperature is approximately means that the heat generated is being confined to this location
2500 K, which causes the significant and continuous thermal and that only a smaller amount of current density is needed
generation of carriers. Furthermore, the generated electrons to cause additional heating. At the same time, the electric
and holes from the heavy-ion strike are being injected at field at that location has a value of 0.75 MV/cm, which is
this location as they drift toward their respective terminals: not sufficient to create significant impact ionization. Also,

Authorized licensed use limited to: SICHUAN UNIVERSITY. Downloaded on December 22,2023 at 12:05:19 UTC from IEEE Xplore. Restrictions apply.
MCPHERSON et al.: ION-INDUCED MESOPLASMA FORMATION AND THERMAL DESTRUCTION 655

Fig. 6. (a) Electric field and (b) lattice temperature 1-D time evolution at Fig. 7. (a) Electron density and (b) electron current density 1-D time
925 V along heavy-ion strike (LET = 46.1 MeV-cm2 /mg). evolution at 925 V along heavy-ion strike (LET = 46.1 MeV-cm2 /mg).

the electric field along the heavy-ion track does not exceed
1.5 MV/cm, which is lower than the peak electric field found make their way to their respective terminals. The movement of
before the heavy-ion strike. This implies that any carriers the carriers and their effect on the mesoplasma can be explored
produced by impact ionization do not enhance the density by looking at the physical parameters of mobility, lifetime, and
value seen at this location. This observation, coupled with saturation velocity.
the reduced impact ionization at high temperatures due to The mobility of the electrons is 7.6 times higher than
the reduction of the mean free path of the carriers, indicates the holes and can be attributed to the fact that holes are
the dominating factor behind the carrier density and current heavier than the electrons. As a result, the holes have difficulty
density is attributed to the thermal generation of carriers. The traversing the device quickly. Figs. 7–9 demonstrate that the
results of the heavy-ion simulation support the conclusion that mesoplasma occurs 3.8 μm from the surface of the device, and
a mesoplasma has formed within the power device at this a large number of holes will have to transit the mesoplasma
stage, as shown in Fig. 9, and that its location also coincides as they make their way to the source contact. The low
with the peak current density, as shown in Fig. 5(c). mobility of the holes allows for the continual injection of
carriers into the mesoplasma throughout the simulation. The
mobility of the holes was increased to equal that of the
IV. PARAMETERS A FFECTING M ESOPLASMA electrons (925 cm2 /V-s) to determine if the holes transiting
As previously discussed, during the mesoplasma formation back to the source terminal were the cause of the mesoplasma
process, two factors are responsible for the mesoplasma: the formation. Fig. 10 shows the peak temperature behavior for
thermal generation of carriers and the dynamics of the carriers. this scenario. Comparing Fig. 4 with Fig. 10 shows that a
It is essential to identify which aspect is the dominating slight improvement was made to the SEB threshold voltage
factor behind the mesoplasma formation. Various physical and that the peak temperature reached for the 925 V case
parameters were artificially changed through simulations to see was 2780 K. It was found that the location where failure
how the mesoplasma reacts and, thus, identify the dominant would occur shifted 0.7 μm closer to the surface compared to
factor. These parameters include mobility, lifetime, saturation the default simulation. Despite the significant increase to the
velocity, and thermal conductivity. holes’ mobility, the SEB value did not increase more than 50 V.
From the simulation results, it was hypothesized that Therefore, the main reason behind the mesoplasma formation
the electron–hole plasma located at the burnout location is is not from the slow-moving holes traversing toward the source
enhanced by the carriers that are traversing the device as they contact.

Authorized licensed use limited to: SICHUAN UNIVERSITY. Downloaded on December 22,2023 at 12:05:19 UTC from IEEE Xplore. Restrictions apply.
656 IEEE TRANSACTIONS ON NUCLEAR SCIENCE, VOL. 68, NO. 5, MAY 2021

Fig. 10. Peak temperature plot for a uniform buffer-layered MOSFET where
mobility of the holes was increased to equal that of the electrons.

Fig. 8. (a) Hole density and (b) hole current density 1-D time evolution at
925 V along heavy-ion strike (LET = 46.1 MeV-cm2 /mg).

Fig. 11. Peak temperature plot for a uniform buffer layered MOSFET
where the lifetime for both the electrons and holes were (a) increased and
(b) decreased by a factor of 100.

Fig. 9. Lattice temperature showing the mesoplasma formation at 240 ps.

The heavy-ion path entry is located at (0, 0) and travels vertically. and holes by a factor of 100. The default lifetimes for the
electrons and holes that were used in the SEB simulations are
The lifetime of the carriers determines how long it takes 1000 and 200 ns, respectively. Fig. 11(a) and (b) shows the
for the electron and hole to recombine with each other, and SEB results when the lifetimes are changed. When the lifetime
by manipulating this value the behavior of the mesoplasma is increased, SEB occurs at an earlier threshold voltage and is
could potentially be altered. Two simulations were created that attributed to more carriers reaching the mesoplasma and caus-
would increase and decrease the lifetime for both the electrons ing additional heating. The opposite is observed when the life-

Authorized licensed use limited to: SICHUAN UNIVERSITY. Downloaded on December 22,2023 at 12:05:19 UTC from IEEE Xplore. Restrictions apply.
MCPHERSON et al.: ION-INDUCED MESOPLASMA FORMATION AND THERMAL DESTRUCTION 657

Fig. 13. Peak temperature plot for a uniform buffer layered MOSFET
where the thermal conductivity was allowed to remain constant at its
room-temperature value.

TABLE II
S UMMARY OF PARAMETRIC S IMULATIONS ON M ESOPLASMA F ORMATION

Fig. 12. Peak temperature plot for a uniform buffer layered MOSFET where
the saturation velocity for both the electrons and holes were (a) increased and
(b) decreased by a factor of 10.

time is reduced, causing an SEB to occur near 925 V attributed SEB threshold voltage by decreasing the saturation velocity,
to the slight reduction in carriers resulting from increased there exists a trend for semiconductor materials where the
recombination. The location for failure was 3.7 and 3.9 μm saturation velocity increases as the bandgap increases [23].
from the surface for the increased and decreased lifetime This conclusion shows that none of the current wide-bandgap
simulations, respectively. However, the improvement to the materials are capable of having a saturation velocity that is low
SEB threshold voltage is only marginal compared to the enough, less than 2 × 106 cm/s, to affect the behavior of the
decrease in carrier lifetime. From this, it can be concluded that mesoplasma, and instead, their increased saturation velocity
the mesoplasma formation is insensitive when the lifetime is values can result in an earlier onset of SEB.
increased or decreased by a factor of 100. With the aspects of the dynamics of the carriers explored,
The third physical parameter varied was the saturation the thermal properties of SiC need to be investigated to see
velocity. Fig. 6(a) shows that the electric field along the heavy- whether they have a significant effect on the mesoplasma
ion track is high enough to accelerate the carriers to the satura- formation. The SEB simulations used a temperature-dependent
tion velocity, which for SiC is 2.0 × 107 cm/s. The impact of model for both the thermal conductivity and lattice heat capac-
saturation velocity on the mesoplasma formation was explored ity. However, only the thermal conductivity will be examined
by creating two simulations. In one case, the saturation veloc- since it has an inverse relationship with temperature, and
ity was increased by a factor of 10, and in the other, it was the lattice heat capacity increases as temperature increases
decreased by the same factor. Fig. 12(a) and (b) shows the (Fig. 2). The simulation was performed by fixing the SiC
peak temperature for these two scenarios. By increasing the thermal conductivity to the room temperature value regardless
saturation velocity, the device experienced SEB at an earlier of the temperature seen in the lattice (no temperature depen-
threshold voltage than was seen in Fig. 4, but the failure is dence). The peak temperature result of this simulation is found
still from the mesoplasma formation, now at 3.5 μm from the in Fig. 13. By removing thermal conductivity’s temperature
surface of the device. This result is in contrast to decreas- dependence, the device was able to survive up to the voltage
ing the saturation velocity, which is capable of surviving a of 1600 V. Due to the higher thermal conductivity, the heat
heavy-ion strike at 1500 V. Despite the improvement to the generated in the device can flow away more readily, slowing

Authorized licensed use limited to: SICHUAN UNIVERSITY. Downloaded on December 22,2023 at 12:05:19 UTC from IEEE Xplore. Restrictions apply.
658 IEEE TRANSACTIONS ON NUCLEAR SCIENCE, VOL. 68, NO. 5, MAY 2021

down the temperature increase and preventing the hot spot [5] J.-M. Lauenstein, M. C. Casey, E. P. Wilcox, H. Kim, and
formation. Therefore, the thermal generation of carriers cannot A. D. Topper, “Single-event effect testing of the cree C4D40120D
commercial 1200 V silicon carbide Schottky diode,” NASA Goddard
create a current density value high enough to permit the Space Flight Center, Greenbelt, MD USA, Tech. Rep. GSFC-E-DAA-
continual heating of the device. This demonstrates that the TN39790, 2014. [Online]. Available: https://ntrs.nasa.gov/citations/
formation of the mesoplasma depends heavily on the thermal 20170003887
[6] S. Kuboyama, C. Kamezawa, N. Ikeda, T. Hirao, and H. Ohyama,
characteristics of the semiconductor material. In particular, “Anomalous charge collection in silicon carbide Schottky barrier diodes
the thermal conductivity helps accelerate the mesoplasma and resulting permanent damage and single-event burnout,” IEEE
formation due to the inverse relationship that exists between Trans. Nucl. Sci., vol. 53, no. 6, pp. 3343–3348, Dec. 2006, doi: 10.
1109/TNS.2006.885165.
the thermal conductivity and temperature. If a material with [7] S. Kuboyama et al., “Thermal runaway in SiC Schottky barrier diodes
a high enough thermal conductivity is chosen, this effect caused by heavy ions,” IEEE Trans. Nucl. Sci., vol. 66, no. 7,
can largely be suppressed. All the simulation results for pp. 1688–1693, Jul. 2019, doi: 10.1109/TNS.2019.2914494.
[8] A. F. Witulski et al., “Single-event burnout mechanisms in SiC power
different physical parameters are summarized in Table II, MOSFETs,” IEEE Trans. Nucl. Sci., vol. 65, no. 8, pp. 1951–1955,
which shows the relative change to the default case in Aug. 2018, doi: 10.1109/TNS.2018.2849405.
threshold voltage where SEB occurs in simulations and the [9] J. McPherson, C. W. Hitchcock, T. P. Chow, W. Ji, and A. Woodworth,
“Mechanisms of heavy ion-induced single event burnout in 4H-SiC
location of the maximum temperature from the surface of the power MOSFETs,” Mater. Sci. Forum, vol. 1004, pp. 889–896, Jul. 2020,
device. doi: 10.4028/www.scientific.net/MSF.1004.889.
[10] D. R. Ball et al., “Ion-induced energy pulse mechanism for single-
event burnout in high-voltage SiC power MOSFETs and junction barrier
V. C ONCLUSION Schottky diodes,” IEEE Trans. Nucl. Sci., vol. 67, no. 1, pp. 22–28,
Jan. 2020, doi: 10.1109/TNS.2019.2955922.
A device-physics investigation into the single heavy-ion- [11] S. K. Ghandhi, Semiconductor Power Devices: Physics of Operation and
strike-induced failure on the vertical SiC planar power DMOS- Fabrication Technology. New York, NY, USA: Wiley, 1977.
[12] D. L. Blackburn and D. W. Berning, “An experimental study of reverse-
FET was performed. It has been demonstrated that, when the bias second breakdown,” in IEDM Tech. Dig., Dec. 1980, pp. 297–301,
failure mode at the epi/substrate interface was suppressed, doi: 10.1109/IEDM.1980.189818.
a significant change in the time-evolved electric field profile [13] B. Beatty, S. Krishna, and M. Adler, “Second breakdown in power
transistors due to avalanche injection,” in Proc. IEEE Power Electron.
was observed and the mesoplasma location shifted away Spec. Conf., Jun. 1976, pp. 247–254, doi: 10.1109/PESC.1976.7072924.
from the high electric field region. During the mesoplasma [14] S. Krishna and P. Gray, “Second breakdown: Hot spots or hot cylin-
formation, the rate at which the thermally generated carriers ders,” Proc. IEEE, vol. 62, no. 8, pp. 1182–1183, Aug. 1974, doi: 10.
1109/PROC.1974.9589.
are generated increases with decreasing thermal conductivity. [15] C. J. Werner, “MCNP user’s manual code version 6.2,” LANL,
Subsequent sensitivity studies of the bulk materials parameters Los Alamos, CA, USA, Tech. Rep. LA-UR-17-29981, 2017.
showed that thermal conductivity is most influential in the [16] SentaurusTM Device User Guide, Synopsis, Mountain View, CA, USA,
2014.
mesoplasma formation or suppression. While this investigation [17] O. Nilsson et al., “Determination of the thermal diffusivity and conduc-
did not provide a clear path to obtaining the desired rated tivity of monocrystalline silicon carbide (300-2300 K),” High Temp.-
blocking voltage in a SiC power DMOSFET, with the addition High Press., vol. 29, no. 1, pp. 73–79, Jan. 1997.
[18] S.-H. Ryu et al., “Development of 1200 V, 3.7 m-cm2 4H-SiC
of a minimally invasive buffer layer between the epi layer and DMOSFETs for advanced power applications,” Mater. Sci. Forum,
substrate, a more lucid understanding of the time evolution of vols. 717–720, pp. 1059–1062, May 2012, doi: 10.4028/www.scientific.
this thermal runaway process has been gained. net/MSF.717-720.1059.
[19] M. Lades, “Modeling and simulation of wide bandgap semiconductor
devices: 4H/6H-SiC,” Ph.D. dissertation, Dept. Elektrotechnik und Infor-
R EFERENCES mationstechnik, Technische Universität München, München, Germany,
2000.
[1] T. Shoji, S. Nishida, K. Hamada, and H. Tadano, “Experimental and [20] Y. S. Touloukian and E. H. Buyco, Specific Heat: Nonmetallic
simulation studies of neutron-induced single-event burnout in SiC power Solids (Thermophysical Properties of Matter). New York, NY, USA:
diodes,” Jpn. J. Appl. Phys., vol. 53, no. 4, Feb. 2014, Art. no. 04EP03, IFI/Plenum, 1970, pp. 448–450.
doi: 10.7567/JJAP.53.04EP03. [21] H. Niwa, J. Suda, and T. Kimoto, “Temperature dependence
[2] T. Nitta et al., “Cosmic ray failure mechanism and critical factors of impact ionization coefficients in 4H-SiC,” Mater. Sci. Forum,
for 3.3 kV hybrid SiC modules,” in Proc. PCIM Eur.; Int. Exhib. vols. 778–780, pp. 461–466, Feb. 2014, doi: 10.4028/www.scientific.
Conf. Power Electron., Intell. Motion, Renew. Energy Energy Manage., net/MSF.778-780.461.
Nuremburg, Germany, May 2016, pp. 566–572. [22] J. A. McPherson, P. J. Kowal, G. K. Pandey, T. P. Chow, W. Ji,
[3] X. Zhu et al., “Radiation hardness study on SiC power MOSFETs,” and A. A. Woodworth, “Heavy ion transport modeling for single-event
presented at the ICSCRM, Washington, DC, USA, Sep. 2017. [Online]. burnout in SiC-based power devices,” IEEE Trans. Nucl. Sci., vol. 66,
Available: https://nepp.nasa.gov/files/29689/NEPP-CP-2017-Lauenstein- no. 1, pp. 474–481, Jan. 2019, doi: 10.1109/TNS.2018.2880865.
Zhu-Poster-ICSCRM-TN64775-NEPPweb-reuse-TN46843.pdf. [23] B. Ozpineci and L. M. Tolbert, “Comparison of wide-bandgap semicon-
[4] D. J. Lichtenwalner et al., “Reliability of SiC power devices against ductors for power electronics applications,” Oak Ridge Nat. Lab., Oak
cosmic ray neutron single-event burnout,” Mater. Sci. Forum, vol. 924, Ridge National Lab, TN, USA, Tech. Rep. ORNL/TM-2003/257, 2004,
pp. 559–562, Jun. 2018, doi: 10.4028/www.scientific.net/MSF.924.559. doi: 10.2172/885849.

Authorized licensed use limited to: SICHUAN UNIVERSITY. Downloaded on December 22,2023 at 12:05:19 UTC from IEEE Xplore. Restrictions apply.