A New Physics-Based Circuit Model for 4H-Sic Power Diodes

Implemented in SABER
R. Kolessar and H.-P. Nee
Royal Institute of Technology
Department of Electrical Engineering
Electrical Machines and Power Electronics
S-100 44 Stockholm, Sweden
remy.kolessar @ekc. kth .se
Abstruct- A circuit-oriented model for 4H-Sic power diodes is pre-
sented. The modeling technique used in this work was previously ap-
plied to a silicon (Si) power diode model for both turn-on and turn-off
transients, and presents a good trade-off between accuracy and speed.
The model is physics-based, but includes judicious approximations for
fast calculation, and includes up-to-date physical models for silicon car-
bide, with temperature dependence. The proposed model is practically
implemented in the circuit simulator SABER, using the MAST model-
ing language. Model performances are compared to measurements on
2.5 kV 400 A Si IGBT/SiC diode modules from ABB at various input cur-
NOMENCLATURE
j-order polynomial coefficient in subregion i
(~m- ~+j ).
electric field (V/cm).
valance band energy level (ev).
donor effective ionization energy (eV).
Fermi energy level (eV).
bandgap energy (eV).
anode (cathode) h-parameter (cm2/s).
displacement current density (A/cm2).
hole current density (A/cm2).
anode (cathode) recombination current density
(A/cm2).
time-dependent variable for the moving bound-
aries (cmW3).
electron (hole) density-of-state effective mass
(kg).
electron concentration ( ~m- ~) .
intrinsic carrier concentration ( ~m- ~) .
donor (acceptor) concentration ( ~m- ~) .
density-of-states in the conduction (valance) band
( ~m- ~) .
hole concentration ( ~m- ~) .
carrier concentration at the left (right) boundary
of the excess charge ( ~m- ~) .
elementary charge (As).
charge in subregion i (As).
diode reverse recovery charge (As).
junction temperature (K).
anode (cathode) voltage drop (V).
depletion region voltage (V).
diode n-base width (cm).
X / , r
k P electron (hole) mobility (cm2/V.s).
7 high-level lifetime (s).
r n , p
left (right) boundary of the excess charge (cm).
low-level electron (hole) lifetime (s).
I . INTRODUCTION
Silicon carbide (Sic) is a possible candidate to replace sili-
con commercially for power devices in a near future, thanks to
its remarkable electronical and physical properties compared
to conventional silicon (Si) technology, e.g. higher critical
electric field (x4-20), higher thermal conductivity (x3-13) and
temperature capability (x2-3) [l]. It is therefore interesting to
be able to already investigate its impact and benefits on the
design of modern power electronic systems.
For that purpose, accurate physics-based models for circuit
simulation are required for Si c devices, with good predictive
capabilities. To achieve such model performances at circuit
level, a good trade-off between accuracy and efficiency has to
be obtained. Fast and relatively simple models often give a
too much simplified representation of the complex physics in-
volved during transients of the device, and thus provide limited
ability to implement advanced physical models [2]. On the
other hand, more accurate models, with better physical rep-
resentation, most often use complex algorithms or geometric
approximations that are not always suitable for simulation at
circuit level [3,4].
The physics-based modeling technique used in this work
presents an interesting trade-off between simplicity and ac-
curacy, and was previously used for Si power diode models
with good results [S, 61. Indeed, due to its original formal-
ism, the model presents a good ability to support a wide range
of physical models with a very flexible implementation proce-
dure. Therefore, this modeling technique appears to be partic-
ularly appropriate for implementation of silicon carbide phys-
ical models.
A set of relevant 4H-Sic physical models, which are today
available are presented. These models have been included in
the circuit power diode model, and implemented in the circuit
simulator SABER [7].
The proposed circuit model for silicon carbide diodes is val-
idated by comparison with measurements on 2.5 kV 400 A
Si-IGBT / Sic diode modules from ABB [8]. Good agreement
0-7803-6618-2/01/$10.00 0 2001 IEEE 989
is obtained between simulation and measurements.
0 U'"
11. THE MODELING TECHNI QUE
The modeling technique used in this work is described more
extensively in [5,6], and only basic principles are presented
here. The main idea for describing high injection level effects
is to divide the n- region of the diode into a number of subre-
gions (typically five to ten), and apply a Charge Control Equa-
tion (CCE) for each subregion with the charge Qi, see Fig. l .
Moving boundaries for the total excess charge are also taken
into account, which results in the following system of differ-
ential equations
+ = -+ +jp(x1) - jp(x1)
+q . K,a ( t ) % - q * Kf ( t ) @&c
where the first term on the right hand accounts for carrier re-
combination with a carrier lifetime T. The second and third
terms are the net charge supply. The last two terms account for
the moving boundaries 21 and x,.
The carrier distribution in each subregion is approximated
by second degree polynomials with time-varying coefficients
In order to ensure a non-quasi static model, the excess
charge Qi is linked at every moment to the spatial distribution
of the carrier concentration since we have
The resulting system (1) has to be solved with the following
simplified moving boundary conditions depending on the state
of the diode, i.e. turn-on, on-state or turn-off,
(4)
if p~ >0 then 21 = 0 else pl =0
if p , >0 then x, = else p , =0,
where pr and p , are respectively the carrier concentrations at
the boundaries x1 and xT, which is illustrated in Fig. 1.
Note that the model parameters are all of physical nature,
i.e. device active area, diode base width and doping level, h-
parameters for emitter recombination and high-level carrier
lifetime.
111. IMPLEMENTATlON OF PHY SI CAL MODELS
FOR 4H-SlC DI ODES
At circuit simulation level, it is very important to include
only the most relevant physical models in order to fulfill the
requirement of fast computation time. These models are very
990
Concentration
Pi
N -
PI
+-
Voltage, Current
t
Fig. 1. a) Simplified principle of the PIN diode model discretization and
carrier distribution during the diode reverse recovery showing the differ-
ent boundary conditions. b) 'Qpical voltage (plain) and current (dashed)
waveforms for a typical diode tum-off.
well described for silicon devices, and their impact on the cal-
culation accuracy is well known [9].
The problem is however more complex for silicon carbide
devices, since the physical models themselves are, in most
cases, not completely well described or understood. In order
to derive and implement a pertinent set of physical models, a
study to identify relevant physical models for Si c diodes was
performed using MEDIC1 with a special simulation package
for 4H-Sic [ 101, including advanced models like bandgap nar-
rowing, Auger recombination and impact ionization. In partic-
ular, the impact of different physical models for 4H-Sic from
[ 111 on the switching behavior of a silicon carbide diode was
studied. The physical models that proved to have noticeable
influence for Si c diodes are presented hereafter. Temperature
behavior is included in each case, even if not always validated.
A. Lifetime
Due to the lack of specific information for Sic, low-level
lifetime is usually approximated .assuming Shockley-Read-
Hall recombination mechanisms. For the model purpose, only
the high injection lifetime 7, which is defined as
7- =7, +TP,
( 5)
has to be included, and its value is typically in the range 0.1-
2 ps. The temperature behavior is based on an empirical
derivation from Lindefelt [I21 for the range 300 K 5 T 5
450 K,
T w ~ ( 3 0 0 K ) . 2.1724. - - 1.1724).
(6)
( 300K
B. Mobility Model
The equations for the low field mobilities for holes and
electrons from [ 111 can be directly implemented in the diode
model. These equations are based on measurements from [ 131
for 4H-SiC, and are defined for the electron and hole mobilities
, respectively as
108.9 :
pp( T) = 15.9+
1 +(+&3$)0.34
Moreover, the anisotropic properties of 4H-Sic is expressed by
Pn,L =Pn 7 Pn,(l =1*2Pn (9)
PPJ =PP 1 Pp, ( ( =PP. (10)
C. Incomplete Ionization of Doping Atoms in the Base
In a Si c diode at room temperature and depending on the
doping concentration, not all doping atoms are completely ion-
ized, as it is the case for Si structures. This is due to the wide-
bandgap of Si c compared to Si. The impact of incomplete
ionization is discussed here for the lowly doped base only. In-
deed, the effect of incomplete ionization in the highly doped
emitters can be considered with the concept of h-parameters.
This is treated in a following section.
In the case where the dynamic effects of incomplete ioniza-
tion is not considered, the following equation can be used [ 1%
( 1 1)
N d
Nd+ =
1 +2 . e z p (w)
E d is a single effective ionization energy level given by E, -
E d =65 meV for doping with N [ 1 11.
However, the donor doping concentrations for the base layer
of high power Si c diodes are in the low 1015 ~m- ~. With
the value of E d at these impurity levels, the doping atoms are
found to be almost completely ionized, even at a temperature
of 400 K. Therefore, for a range of temperature of 300-400 K
and donor doping levels below 1 x 10l6 cmb3, we can use the
following approximation
Nd+ N Nd. (12)
Temperature (K)
Fig. 2. Intrinsic carrier concentration in 4H-Sic with temperature using (17).
Circles are from the exact calculation combining (14), (15) and (16) in
(13).
D. Intrinsic Carrier Concentration
The intrinsic carrier concentration ni is expressed as [ 141
The densities-of-states Nc and Nv are given by
312
Nc(T) = 2.50945 10 * ( 7 % ~ ~ ) ~ ~ . ( A) (14)
Nv(T) = 2.50945 * lo1 ( m d ~ ) ~ ~ . ( &)3;J5)
if we assume parabolic bands [ 111.
No model is today available for the temperature dependence
of the intrinsic bandgap Eg for 4H-Sic. However, an approxi-
mation based on 6H-Sic data is given by [ 111, and yields
Eg( T) =3.19 - 3. 3. 10-3(T - 300K) eV. (16)
Combining (14), (15) and (16) with (13) allows to calculate ni
as a function of T. A more convenient expression for ni as a
function of temperature can be obtained by using () and fitting
the coefficient to the exact calculation.
(17)
The intrinsic carrier concentration in 4H-Sic based on (17) is
plotted as a function of temperature in Fig. 2.
E. Emitters and h-Parameters
The concept of h-parameters is widely used for analytical
calculation of emitter recombination and efficiency at high in-
jection level [ 15,161. This permits with a simple formulation
to account for more complex physical phenomena like bandgap
narrowing or Auger recombination.
Calculations based on MEDIC1 simulations confirm that the
equations are still valid for silicon carbide, with the recombi-
nation current being proportional to the square of the carrier
ni(T) =A ~ T ~ / ~ exp ( - A ~ / T ) .
99 1
concentrations at the respective emitter edge
j l = q h p : (18)
j r = qhr$. (19)
Generally, for highly doped and deep emitter profiles in sil-
icon structure, the h-parameters are constant with weak tem-
perature dependence. But for thin emitters, as usually obtained
in Si c structures (emitter depth <5 pm) [ 171, the h-parameter
is not constant with temperature [ 161. This can be partly at-
tributed to the increase in ionized doping atoms with increas-
ing temperature, but also to more complex physical phenom-
ena like bandgap narrowing and Auger recombination.
Practically and for a given emitter design (basically emitter
depth and doping), calculation of h-parameters are performed
by MEDIC1 simulations and an empirical law of the following
form is used
The coefficients ah,,, are determined by fitting with results
from MEDICI, Fig. 3.
E Diode voltage
The diode voltage is computed by adding the emitter voltage
drops Va (anode) and V k (cathode) to V, (base). This results
in
(2 1)
I ) Diode blocking voltage: At turn-off conditions, most of
the diode voltage is sustained by the depletion region. Abrupt
junction approximations are sufficient for description of silicon
carbide structures (even though the model can include differ-
ent profiles), since emitters in Si c diodes are usually thin and
highly doped, as discussed previously.
Using this approximation, the following expressions can be
derived from Poisson's equation for calculation of the diode
voltage at reverse blocking conditions. The electric field dis-
tribution is given by
v d =v a +vm +v k
(22)
E Ws a t
and the blocked voltage by a further integration of (22) over
the depletion region
Note that the eventual transport of holes in the depletion region
is accounted for.
2) Emitter voltage drops: The same Boltzmann equations
used for Si PIN diodes are used for SIC. These equations for
the calculation of voltage drops at the anode and cathode re-
spectively read
Z x " - ~ ' " ' ' I ' ' ' I " ' I " ' J
280 320 360 400 440
Temperature (K)
I . . . I . . . I . . . I . . . l
Temperature (K)
280 320 360 Qw 440
(b)
Fig. 3. h-parameters with temperature at the anode a) and cathode b) for emit-
ter depths of 2 pm and 5 pm respectively, and doping at 1 x I Ol 9 cmP3.
The circles and squares are MEDICI simulations.
and
kT
Vk =-1In 4 ($).
Fig.4 shows calculations using (24) and (25) for a donor
concentration of N d =1 x loi5 ~m- ~. The difference is more
relevant at the anode where the value are >2.5 V for 4H-SiC,
whereas it is typically <1 V for Si pn junctions. This is due the
larger bandgap and thus contact potential of 4H-Sic structures
compared to their Si counterparts.
V, on the other hand is smaller in Sic than for Si devices.
This is because of the usually higher doping concentration in
the base ( IO1' cm-3 range for Si c compared to cm-3
for Si), since we have a low-high junction barrier in that case.
G. Displacement Current
The displacement current j d during turn-off of Si diodes is,
in most cases negligible, since most of the reverse current is
due to the charge sweep out.
For Sic diodes however, the displacement current is of-
ten a non-negligible part of the total recovery current, since
992
3, I I
05
---- -- rrrd
-
- - _______- ---------
much less excess charge is injected in the n- base than in Si-
based structures. This has therefore to be included in the diode
model. The displacement current is given by
dE
j d = E - - ,
at
if we assume that j d is uniform and independent of the position
in the depletion region, which is justified if the transit time of
the charge is negligible compared to the time constant of the
currents. Combining (26) with (22), we have
Note that carriers (holes) in transit in the depletion region are
also taken into account in (27).
IV. IMPLEMENTATION IN SABER
The basic set of analytical physical models described pre-
viously is easily implemented in the circuit level model. The
practical implementation is done using a so-called High De-
scription Language (HDL) called MAST, which is part of the
SABER circuit simulator, and is described in [18]. Note that
no external C-routines have to be used for this implementation.
V. MODEL VALIDATION
The performances of the proposed model are tested by com-
parison with transient measurements on 2.5 kV Si IGBTlSiC
PIN diode from ABB. These modules are manufactured by par-
alleling four units of large area (40 mm2) 2.5 kV 4H-Sic PiN
diodes with four 2.5 kV Si-IGBT chips with an area of 1 cm2
each using a standard high power press-pack technology [8].
A. Experimental
The test-circuit being used is a single pulse circuit as shown
in Fig.5. The turn-off transient of diode D1 can be investi-
gated when the when the initial forward current from the induc-
tive current from L, is commutated from diode D1 to IGBT2
,
993
Fig. 5. Single pulse test-circuit used for both measurements and simulations.
TABLE I
MODEL PARAMETERS FOR THE 4H-SiC DIODE.
1 Model Parameter I Value
T (high level lifetime)
tun (base width)
when this one is turned-on. The freewheeling diode D1 is then
reverse-biased and reverse recovery occurs.
Measurements of dynamic switching of the Si IGBT/SiC
diode modules are performed at a temperature of T =125C
for forward currents of 50 A and 150 A with a blocking volt-
age of 800 V, as well as for a forward current of 400 A with a
reverse voltage of 1250 V.
B. Simulations
Simulations were performed by implementing the single
pulse circuit of Fig. 5 in SABER. The switching behavior of
the IGBT. is modeled using a standard MOSFET model with
adjusted parameters. The model parameters are extracted for
one case and are kept constant for the remaining simulations.
The six model parameters are shown in Table I. The tempera-
ture is also fixed at T =125C.
C. Results and Discussion
SABER simulations using the proposed model and measure-
ments made on the Si c diode modules are presented in Fig. 6
and 7 at various forward currents. Very good agreement be-
tween experimental and simulation data is observed for the
various conditions. The slight deviations, specially in the diode
voltage of Fig. 6 are probably due to limitations in the model-
ing of the IGBT. Notice the small changes of the recovery cur-
rent with increasing forward current, which is well predicted
by the model. Once again, the same parameters have been used
in all cases. The simulations are computed in few seconds on
a PC PI11 600 MHz computer.
200
I50
h 100
E 50
4
E
a 0
-
-50
-100 4 - -200
0 1 2 3 4 5
Time(ps)
Fig. 6. Comparison between a) measurements and b) simulations of theSi c
diode with forward currents of 50 A and 150 A with a DC link at 800 V.
VI. CONCLUSIONS
In conclusion, a predictive physics-based model for 4H-Sic
power diodes suitable for circuit simulation is presented and
implemented in the SABER simulator. The set of physical
models for 4H-SiC, which was implemented in the model have
been discussed in detail. Most of these models include temper-
ature dependence. Analytical equations in some cases are pre-
sented for these models in order to simplify the implementation
procedure. The proposed model is also validated with mea-
surement results. Model performances are compared favor-
ably to an experimental test-circuit built with 2.5 kV 400 A Si-
IGBT/4H-SiC diode modules at various forward current con-
ditions.
ACKNOWLEDGMENTS
The authors would like to thank H. Lendenmann from ABB
Corporate Research, Sweden for measurement data and fruitful
discussions. M. Bakowski and U. Lindefelt are also acknowl-
edged for valuable discussions about the Sic models. This
1200
loo0 g
800 g
400 5
6Ml E
200
0
-m
0 2 4 6 8 1 0
Time(ps)
Fig. 7. Comparison between measurements (plain) and simulation (dashed)
of theSIC diode with forward currents of 400 A with a DC link at 1250 V.
project is financially supported by ELFORSK and the Swedish
Foundation for Strategic Research within the SiCEP program.
REFERENCES
K. Rottner, M. Frischholtz, T. Myrtveit, D. Mou, K. Nordgren, A. Henry,
C. Hallin, U. Gustafsson, and A. Schoner, Sic power devices for high
voltageapplications, Material Sience and Engineering B, vol. 61-62,
pp. 330-38, 1999.
C.L. Ma, P.O. Lauritzen, and J. Sigg, Modeling of power diodes with
thelumped-charge modeling technique: IEEE Transactions on Power
Electronics, vol. 12, no. 3, pp. 398405, May 1997.
T. Vogler and D. Scluoder, A new and accurate circuit-modelling ap-
proach for thepower diode: in PESC92 Ref. , 1992, pp. 87G-876.
R. Kraus, K. Hoffman, and H.J. Mattausch, A precise model for. the
transient characteristics of power diodes, in PESC92 Rec., 1992, pp.
A. Persson, A circuit simulation model for thehigh power diode: in
EPE91 Rec., 1991, vol. 2, pp. 458-463.
R. Kolessar and H.-P. Nee, Electro-thermal circuit model for thepower
diode turn-on process: in IEEE NORPIU2000 Workshop, Aalborg,
Denmark, J une 2000, pp. 69-73.
SABER, Avant! Corporation, 9205 S.W. Gemini Drive, Beaverton, OR
97008, USA, release 5.1 edition.
H. Lendenmann, E Dahlquist, N. J ohansson, J.P. Bergamn, H. Bleichner,
and C. Ovr6n. Performance and reliability of high power Si c diodes,
in International Workrhop onUltra-Low-Loss Power Device Technology,
Nara, Japan, 2OOO.
B. J ayant Baliga, Power serniconductor devices, PWS Publishing Com-
pany, Boston, 1996.
MEDICI, Avant! Corporation, TCAD Business Unit, 4687 1 Bayside
Parkway, Fremont, CA 94538, USA.
M. Bakowski, U. Gustafsson, and U. Lindefelt, Simulation of SIC high
power devices, Physica Status Solidi A, vol. 162, no. 1, pp. 421-40,
July 1997.
U. Lindefelft, ,Unpublished.
W.J. Schaffer, G.H. Negley, K.G. Irvine, and J.W. Palmour, Conductiv-
ity anisotropy in epitaxial 6H- and 4H-Sic, in Sytnposiurn Mater. Res.
SOC, Pittsburgh. USA, 1994, pp. 595-600.
Sheng S.L., Semironductor physiral electronics, PlenumPress, New
York, 1993.
E Ben, R.W. Cooper, and S. Fagg, Recombination in theend regions
of PIN diodes, Solid-State Electronics, vol. 22, pp. 293-301, 1979.
Ph. Leturcq, M.O. Berraies, J.P. Laur, and P. Austin, Full dynamic
power bipolar devicemodels for circuit simulation, in PESC98 Rec.,
A. Schoner, S. Karlsson, T. Schmitt, A. Hallen, M. Frischholtz, and
K. Rotmer, High energy implantation of boron in 4H-Sic: in Wide-
Batidap Semiconductors for High Powe,: High Frequency and High Tem-
perature, Warretidale, USA, 1998, pp. 469-74.
R. Kolessar and Nee, Physics-based power device models in
SABER - implementation issues, in IEEE COMPEL2000, July 2000,
863-869.
1998, vol. 2, pp. 1695-1703.
pp. 18-22.
994