In recent years, we've
seen the introduction of new power semiconductor devices using Wide Bandgap
semiconductor materials. Instead of silicon, they use silicon carbide or
gallium nitride. These materials allow
significant improvements in the trade-off between
breakdown voltage, forward voltage drop,
and switching speed. So in this brief lecture, I'm going to describe or
summarize why that is so, and then we'll talk about several of the commercially
available devices, specifically silicon
carbide, schottky diodes, silicon carbide MOSFETs, and gallium nitride HEMT devices. Here is a formula that
you see often quoted in the literature for the
specific resistance, in other words the
on-resistance multiplied by the active chip area for a majority carrier
device such as a MOSFET. The on-resistance is a function
of the breakdown voltage; V sub B, as well as the
material mobility; mu, the material
permittivity epsilon, and the critical electric field at the onset of
avalanche breakdown; E sub c. So if you want to increase the blocking voltage
of your MOSFET for example, the on-resistance will increase as the breakdown
voltage is squared. So high-voltage MOSFETs have
much higher on-resistances. But the interesting thing
here is the other terms, these denominator terms, that are a function of the
semiconductor material. So it turns out that these
Wide Bandgap materials such as silicon carbide
and gallium nitride, have much higher critical
fields, E sub c, that allow a significant
improvement in this trade-off. Here's a comparison
of the parameters for these basic materials. So silicon has a bandgap
of 1.12 electron volts. Silicon carbide and
gallium nitride have much higher bandgaps
and the result of this is that the critical
fields are much higher. So with this
significant increase in critical field then we can design the device to have a much lower on-resistance at a given
breakdown voltage. The electron mobilities
are also changed. In fact, it turns out
that silicon carbide has lower electron
mobility which is bad, that means more on-resistance, and gallium nitride, for reasons we'll
discuss in a minute, has higher electron mobility
to what are called hot carriers that exist in the two-dimensional
electron gas in the device. Because of the lower
mobility of silicon carbide, it actually is inferior at low breakdown voltages
and silicon is better. But when you get above 600 volts, then silicon carbide begins to exhibit significant
advantages, and so we see good
silicon carbide MOSFETs at voltages starting at 600 volts and going commercially
now to 10,000 volts. GaN doesn't have that
issue and in fact GaN is highly
competitive with silicon as MOSFETs at low voltage or at least as controlled
transistors at lower voltages. So we see both sub 100 volt GaN devices and we see 600 volt GaN devices
that are very good. We can also make Schottky
diodes with these devices. So we see high voltage
Schottky diodes made especially out
of silicon carbide, and there are majority
carrier devices that have no significant
reverse recovery. We get silicon carbide Schottkys at 600 and 1200 volts and higher. In silicon carbide we
can build MOSFETS, Silicon carbide can have an oxide layer that is similar to the oxide layer in silicon and we can
build MOSFETs with it. This is not true in
other materials and in particular it's not true in GaN. So we aren't able to
build MOSFETS with GaN and the devices that we have our junction JFET instead. Silicon carbide
MOSFETs then can be built with a vertical
structure that's very similar to the vertical
MOSFET structure seen in a silicon MOSFET. It has all the same structure, it has a body diode and so on. In GaN this is not the case, GaN has a thin film material
that is deposited on top of a substrate and so we
have lateral GaN devices. Because GaN devices are lateral, it's somewhat more difficult to scale to higher voltages and currents although we're seeing that being done commercially. The thermal coefficient
of expansion of the thin film GaN device
must be matched somehow to the substrate that it's
deposited on and there are issues there as well that affect the structure of the GaN devices. So as I mentioned, we have silicon carbide Schottky diodes. These were the first
wide-bandgap power devices to become commercially important. We see them available at 600, 1200 volts and higher. Because the Schottky diode is
a majority carrier device, there's no significant
reverse recovery. They do have a higher
forward voltage drop, typically 1.5 to two volts, but they have much
lower switching loss because they're a
majority carrier device. What we find then is higher conduction loss,
lower switching loss. In most switched-mode
applications, let's say 600 volts, you see a significant gain in
overall system efficiency. They also cost more than silicon, so there's a trade off. In the area of MOSFETs, we have silicon MOSFET at voltages up to 600
volts or 700 volts. If you want to go to
a higher voltage, then generally we
use a silicon IGBT. In the case of silicon carbide, we have silicon carbide
MOSFET starting at 600 volts and commercial
devices available at up to even 10,000 volts and these are real MOSFET
that are very similar to the silicon MOSFETs but with good properties
at high voltage. So they have much lower on-resistance than
a silicon MOSFET, they do have a p-n body diode just like the silicon MOSFET has. In the case of silicon carbide, the body diode has
a forward drop of three to four volts,
so its higher.. The body diodes
generally are built with good reverse
recovery times below 50 nanoseconds and
so we can switch these devices at a
much higher frequency than we could operate an IGBT. So where the IGBT might run
at five or 10 kilohertz, these MOSFETs can run at
much higher frequencies. Here are some sample
commercial devices that you can buy today, I've listed some from
650 volts up to 1,700 volts with nice on-resistances that are tens of milliohms. In GaN, we are not able to build an oxide layer and
so we can't build a MOS-gate and what we see for GaN transistors is
what is basically a junction field-effect
transistor where there's a source and a drain with semiconductor material between
them and a gate that is built as a diode junction
on top of that material. So when we negatively
bias the gate, we can turn off the channel, when we forward bias the device with a
positive gate voltage, there is current that can flow between the source and the drain. The traditional junction FET is a depletion mode device
so at zero gate voltage, its forward-biased are turned
on and to turn it off, you have to apply a negative
voltage to the gate. Early GaN transistors
were in fact depletion mode
devices and we have cascode type circuits
that combine them, silicon MOSFET with a
depletion mode GaN device to build what is effectively an enhancement mode transistor. There is another twist
here in the GaN devices, these are what are called high-electron-mobility
transistors or HEMTs. They are also
characterized as having a heterojunction which means that we have two different
materials having different band gaps
and in the GaN device today we have aluminum
gallium nitride which has a low bandgap semiconductor
material and then below it, is GaN which is a high bandgap or wide-bandgap material and there is a junction between them. When we place materials having different band gaps in contact
at this heterojunction, it can form what is called
the two-dimensional electron gas which is shown here, and this two-dimensional
electron gas has carriers having high energies that are able to conduct or jump very
easily from atom to atom, and effectively have very
low on-resistance or high mobility and this is one of the keys to the GaN device of why it can achieve
very low on-resistance. To turn this device off, we negatively biased
the gate which will extend its depletion
region below the gate and disrupt the 2D electron gas so that the source and drain
have no conducting channel. So this device can have a high breakdown
field because it's a wide-bandgap material and
at the same time it can have very low on-resistance from this two-dimensional
electron gas. But the GaN device is a junction field-effect
transistor that traditionally is a
depletion-mode device. A more recent GaN power
transistors actually have their gate threshold
voltages shifted positively so that they become effectively
enhancement-mode devices. So with zero gate
to source voltage, the device becomes
turned off and we apply a positive voltage greater than the threshold to
turn the device on. However, you can't
apply two positive of a voltage or you
will forward bias the gate to source diode
and then you'll get significant positive
current flowing into the gate which actually
can damage the device. So we have to limit
the gate current and must apply a gate voltage
that is not too high. Often then will drive
these devices with a zero to five volt
gate drive signal to switch the power GaN FET. So in the on state, the device is on with
a low on-resistance. This is with the gate to
source voltage greater than the threshold of may be three
or three and a half volts. In the off state, will generally apply zero volts between gate and source
to turn the device off. The device does not have a
built-in body diode the way the MOSFET structure has
but having said that, it's still possible to
get the device to conduct if you apply reverse current
or voltage through it. Let's suppose we have
the gate shorted to the source and then we apply
negative drain voltage. In this case, with
reverse operation, the drain can actually become the source and the
source the drain. So if you apply a negative
enough voltage to the drain, will have positive voltage
at the gate with respect to the drain and if that positive voltage is
more than the threshold, the device can turn on. So in that case, the device will conduct
and it will conduct current in this direction with a forward voltage drop
between source and drain equal to the threshold
or a little bit greater. So it's possible to have a reverse-conducting
device without actually having a
physical body diode. So this behavior is similar
to the MOSFET that has a body diode except that there's no reverse recovery because
there's no real diode. Also the forward drop
is fairly large, it's greater than or equal to the threshold
voltage of the gate. So you see GaN
devices are actually current bidirectional
devices with very fast switching times. So here's a comparison
of 600 volt transistors, both a silicon MOSFET
and in-equivalent GaN transistor and here it's assumed that they're sized to have the same on-resistance. Of course to achieve
this on-resistance, the GaN has a much smaller
area and as a result, it has much lower gate charge. So because it's a smaller device, the charges are smaller and
it can switch much faster. So we see lower gate charge,
lower output capacitance, the only negative
thing perhaps is the voltage drop when
reverse-conducting, but then again it has
negligible reverse recovery. So wide-bandgap semiconductors have become
commercially important, while they cost more, they can operate at
higher voltages with higher switching speeds and with lower forward voltage drops. Currently the
important devices are the silicon carbide
schottky diode, silicon carbide MOSFET
and the GaN HEMT device.