MAY/JUN 2019 29
use, starting with blue
LEDs in the late 1980s, SiC
has been used as an abrasive
for over a century (in this
context, it’s also referred
to as carborundum), and it
even had a brief stint as the
crystal detector in early ra-
dios, followed by use in yel-
low, then blue, LEDs (until
GaN supplanted it). GaN is
entirely man-made, but SiC
can be found in nature as
the mineral moissanite - but
good luck finding any, as it
only forms under the same
extreme conditions that
produce diamonds. GaN has roughly the same thermal
conductivity as Si-based semiconductors, while SiC is
over twice as good at conducting heat (slightly better
than copper, in fact). SiC also has a relatively low ther-
mal coefficient of expansion, which makes it less prone
to damage from high temperatures and thermal cycling.
All in all, SiC seems like an ideal semiconductor mate-
rial, but there is one catch: it can form a bewildering ar-
ray of crystal structures (called polytypes), each of which
has slightly different physical properties, including the
all-important band-gap energy. Unsurprisingly, this can
make producing SiC semiconductor devices a bit of a
challenge, but as manufacturers gain more experience
with the vagaries of this material, they should continue
to improve yields and lower pricing.
That said, SiC still commands quite a premium over
similarly-rated Si parts, so it is best deployed where there
is no Si equivalent, such as 1,200 V MOSFETs with drain-
source resistances under 200 milliOhms or 600 V to 1,200
V Schottky barrier rectifiers, especially in high-power-
boost power factor correction converters, in which any
time spent in reverse recovery by the output diode gener-
ates considerable losses, as well as increased stress on
the boost switch. GaN is really too new to address on an
economic basis - a perusal of Mouser’s website shows justThese materials can also
withstand higher operating
temperatures, which,
when combined with the
lower leakage and thermal
resistance, means they can
handle a lot more power
for a given device size.
it first (that is, turn it into
plasma). More specifically, it
is a measure of the amount
of energy (given in electron-
volts, or eV) that it takes to
promote an electron in the
outermost valence of an atom
to what is called the conduc-
tion band (hence the term
band gap). The band-gap
energy of metals is effectively
0 eV, so they are good con-
ductors - electrons can easily
pop out of the outer valence
to the conduction band or
vice versa - whereas materi-
als with a band-gap energy of
around 4 eV or higher are considered to be insulators.
Materials with a band-gap energy between those two
extremes are, of course, considered semiconductors,
with an approximate value of 1.14 eV for Si, 2.3 to 3.3 eV
for SiC and 3.4 eV for GaN.
It might appear, then, that wide band-gap materials
like SiC and GaN are at a disadvantage compared to Si
simply because they are much closer to being insula-
tors than conductors. While it is true that a SiC or GaN
device with a given die area and thickness will exhibit a
higher bulk resistivity (and therefore on-resistance) than
a comparable Si device, the wider band gap means their
dice can be much thinner for the same voltage rating.
The thinner dice have lower on-resistance, and lower
thermal resistance too. Another related advantage to a
wide band gap is that leakage current starts off lower
and increases much more slowly with a rise in tempera-
ture (in more technical terms, the spontaneous gen-
eration of charge carriers from heat is reduced). These
materials can also withstand higher operating tempera-
tures (even if their die attach solder or packaging can’t!),
which, when combined with the lower leakage and ther-
mal resistance, means they can handle a lot more power
for a given device size.
While GaN really is a new material for any kind of