The compound semiconductors that combine two or three different chemical elements from columns III and V on the periodic table, outperform silicon in power and speed, as well as emission and detection of light. All of these are needed to meet the advancing needs of the IoT, especially for 5G. To learn more about them, I interviewed Rodney Pelzel, CTO of IQE plc (Cardiff, Wales), a leading supplier of advanced compound semiconductor wafers.
Tech Briefs: The way I look at it, semiconductors are the basic infrastructure of the IoT.
Rodney Pelzel: That’s right — and a specific class of compound semiconductors known as III-V materials, is well-suited for the IoT. They come into their own when you have either high frequency communications or anything that emits and absorbs light. For example, fiber optic communication as the backhaul network would be a big part of the IoT.
And then anything that would be in a harsh environment, such as very high temperatures like in power electronics.
I think a really important point is to make is that the key is to mate them with silicon. There are certain things that silicon does very well, so compound semiconductors will not replace silicon, they are complementary.
Tech Briefs: Could you give me an example.
Pelzel: For fiber optic communications, a transceiver sends and receives information along fiber cables with laser pulses. The logic, which tells it when to send and when to receive, is best done with silicon CMOS. But the component that actually sends the optical pulse — the laser — would be made with compound semiconductors, as would the corresponding receiver, what we call the detector. The waveguide, the component that guides the light into the detector, or guides the laser beam out, is made in silicon, but in such a way that it couples well with the laser and the detector.
So, for example, the way it’s done in what I’ll call the brute force method, is you take a silicon CMOS wafer and then you take a laser that’s on a compound semiconductor wafer, you “cut” the laser from the compound and then bond it to the CMOS wafer.
The general process of combining the two is called hetero-integration — one form of which is the pick-and-place that I described. The other extreme is to put the silicon CMOS wafer into one of our reactors and chemically grow the compound semiconductor on top of the silicon.
More likely, you would do something in between, where a CMOS wafer is “mated” or bonded to the fully intact compound semiconductor wafer.
There are various levels of hetero-integration, but at the basic level, hetero-integration simply means mating, or combining, silicon with compound semiconductors.
Tech Briefs: What would be the advantages of one method versus the other, bonding versus growing?
Pelzel: The advantage of bonding is that it enables you to get the best laser and the best CMOS and put them together in the simplest way. The disadvantage is that the packaged combined device will have a big form factor (typically with wire bonds) and may be less efficient than achieved with the overgrowth of the silicon CMOS wafer. When you go towards the other extreme and integrate them, they get smaller and more efficient — you have less heat generation and a smaller footprint.
Tech Briefs: Could you tell me about some of the properties of these compound materials.
Pelzel: The most significant are:
They can operate at higher frequencies.
They can be used at higher voltages.
They can withstand higher temperatures.
They can emit and absorb light.
Tech Briefs: What are some applications for compound semiconductors?
Pelzel: Power plants step up their generator voltage to tens of thousands of volts to be transmitted via power lines to substations where it is stepped down. The amount of energy lost in stepping voltage down and up is equal to about twice the amount of energy that’s generated every year from all the renewables in the world. If you were to convert the silicon for switching to compound semiconductors, those operations would be much more efficient.
Gallium nitride (a III – V compound) is currently used in computer power supplies that are already available on the consumer market — they’re a lot more efficient, they stay much cooler, and can be put into a smaller package.
Another application is speeding up the charging of electric vehicles. The easiest way is to use a higher potential — instead of charging it at 400 or 500 volts, you take it up to 800 or 900 volts. Switching high voltages is more efficiently done with compound semiconductors. In this instance, higher efficiency means that the charging components remain cooler, which is key to component lifetime.
Emitting and detecting light is used for what we call 3D sensing. For example, if you have an iPhone, the facial recognition system that it’s using to identify you is done with compound semiconductor lasers — It maps your face.
Also, advanced healthcare that you wear on your wrist is based on compound semiconductors — basic spectroscopy. It shoots out a given wavelength of light and based on the return signal, it analyzes your blood. The clever part of that is being able to scale it from something that used to sit on a desk down to something that sits on the wrist.
Tech Briefs: Why do compound semiconductors do that scaling more efficiently than silicon?
Pelzel: To start with, silicon won’t lase. A significant property of compound semiconductors is known as a direct bandgap. What that means is that if I drive electricity into a compound semiconductor that’s designed the right way, it will emit light. They are the materials that are in light emitting diodes. You’re able to tune the bandgap to get the color you want — including IR.
Tech Briefs: Do you tune the bandgap by changing materials?
Pelzel: Yes, you would pick elements in column III of the periodic table, typically, gallium, indium, and aluminum. And in column V you’d pick, say, arsenic, phosphorus, or nitrogen.
You could use all the combinations of those — for example, gallium arsenide, aluminum gallium arsenide, or indium gallium arsenide — all those materials have different characteristic bandgaps.
You’re then able to create a set of bandgaps that go from the visible into the infrared region — to two or three microns, even four microns in wavelength.
Tech Briefs: How far along are you in providing these semiconductor wafers for general use, say for 5G IoT?
Pelzel: Our material is in, for example, mobile phones, which use gallium arsenide chips for which we make epi-wafers. They have been around and been used in mobile phones since the early 2000s, and that’s probably our oldest business. Gallium arsenide is very efficient at the frequencies required. So, at the 800 or 900 megahertz frequencies, up to a gigahertz, if you were to try to make silicon do the same thing, it would be so inefficient that your battery would last two or three minutes. Whereas with a compound semiconductor power amplifier you can get a full day’s worth of charge for your phone.
We are projecting that a big future market will be micro-LEDs — a much smaller version of an LED. They will be used in high resolution displays on a lot of your wearables and next-gen devices. That’s something that we’re developing right now and will probably hit volume in the next two to five years.
Tech Briefs: Can you give me an idea of the frequency range you could do.
Pelzel: The current cellular frequencies run from about 800 or 900 megahertz, up to about sub-six GHz. Gallium arsenide is typically used in the higher frequency ranges, say one to four or one to six GHz.
5G and then 6G has a component that’s called millimeter wave, which goes out to say something like 28 to 64 GHz. Anything that’s above 6 GHz will probably have to use some compound semiconductors.
Tech Briefs: Are different compounds better for different qualities like for the millimeter wave as opposed to lower frequencies?
Pelzel: Yes, for example in your mobile phones, the power amplifiers are typically based on gallium arsenide. As you go to higher powers where you need a higher voltage, for example at a base station, a material called gallium nitride comes into play. It has a higher bandgap so it can hold off more voltage.
As you go up in frequency, there’s a competition. At some point, gallium arsenide can’t do the job anymore — it becomes much less efficient at higher frequencies. Then we use Indium phosphide, which works very well. But then silicon transistors, because you’re trying to only transmit very short distances, have a play as well. So, you get into a battleground of different materials and different technologies. And I don’t think anyone knows which one is going to win.
Tech Briefs: What about the economics of one versus the other?
Pelzel: That’s actually interesting. The easiest example I can give is that silicon can work well at really high frequencies. However, the antenna that’s made out of silicon is much larger and uses a lot more power. If you were to do it in gallium nitride, it would be much more efficient, but its initial cost would be higher, although there would be significant energy savings over the long run.
However, since silicon is the incumbent technology, users might be less tempted to swap it out with something else because the ROI calculation may not give them payback in a suitable amount of time.
Tech Briefs: Is there anything you’d like to add?
Pelzel: Compound materials are going to come into the foreground because they’re going to be necessary. And they are completely different than silicon.
They have different economics than silicon.
They have different fab requirements.
They take different investments.
They have different training requirements for people.
Compound semiconductors are not going to replace silicon. We’re just trying to peacefully coexist and push technology forward.
This article was written by Ed Brown, Editor of Sensor Technology. For more information, go here .