GaN's RF Revolution: The Rise of Wide Bandgap Chips

Introduction 

Silicon Chips have been the mainstay of the electronic industry for decades, and they will continue as such well into the future. But they do have limitations, particularly as to speed, power and the ability to handle high temperatures.  

Modern telecommunications, radar systems, space vehicles, military applications – the list goes on – all operate in the GHz range and beyond. Indeed, there is increasing talk about the "millimeter wavelength" applications expected in the not too distant future. Doing a quick calculation, that means a frequency of about 300 GHz. This is well in excess of practical silicon chips.  

For high-speed and high-power applications, the future apparently belongs to Wide Bandgap Semiconductors. As explained in a previous Electropages^[1] article, "Wide bandgap semiconductors (WBG) are based on materials with a wide gap between their conduction and valence bands. This is the source of the ability to handle more power for devices of the same size as gallium arsenide or silicon semiconductors. And compared to old-school Silicon, WBGs can operate at higher frequencies." The article goes on to say that "WBGs can not only operate at higher temperatures, but they also do a better job of transmitting the heat they generate, easing the challenge of dealing with thermal management."   

GaN compared to SiC and classical silicon. Image source: AVNET

In this article, we will be concentrating on gallium nitride (GaN) transistors. However, both Silicon Carbide (SiC) and silicon (Si) may play vital roles in the fabrication of GaN WBGs. This critical detail is all too often glossed over in the literature.  

The Need for a Solid Substrate  

Actual transistors can be quite small, often less than 10 micrometers  (0.0004 inches) thick, much too small to withstand the rigors of the manufacturing process. As such they need to reside on thicker wafers, and for gallium nitride (GaN) devices, the three main substrate choices are GaN itself, Silicon Carbide (SiC) or Silicon (Si).  

It would seem like the obvious wafer choice for a GaN device would be GaN itself, however, there are several disadvantages to GaN wafers, as we'll see.  

Silicon Carbide Wafers for GaN Chips  

Depositing a GaN layer on top of the SiC is fraught with difficulties. Additionally, SiC and GaN have different thermal expansion coefficients, causing stress and possible breakdowns.   

SiC wafers are quite expensive compared to silicon or even to sapphire. Indeed, both the price of the wafer and quality issues increase faster than wafer size. Currently, 4-inch and 6-inch are most common, but Wolfspeed now produces 8-inch SiC wafers, with others in the running.  

A key advantage of SiC wafers for GaN chips is the high thermal conductivity, which is a major aid to heat management and allows devices to thrive in high-temperature environments. This methodology is widely applicable for high-frequency, high-power applications.  

An Example:  

Qorvo's QPD1008 operates at speeds up to 3.2 GHz. Output power is 125 W from a 50 V power supply. Applications include  

• Avionics 
• Civilian Radar 
• MilitaryJammers 
• Military Radar 
• Wideband and Narrowband Amplifiers 
• Military Radio Communications 
• Test Instrumentation  

Silicon Wafers for GaN Chips  

Silicon wafers have been around for a long time, and semiconductor engineers are familiar with them. The wafers are large, eight inches being the standard, so many chips can be fabricated on each wafer. They are also less expensive than either GaN or SiC wafers.  

However, Silicon and GaN have different heat expansion coefficients, and there are mismatches between the geometries of the atomic structures (lattice mismatch). These differences need to be accommodated by depositing a buffer layer on top of the Silicon Wafer. Only then comes the Gallium Nitride itself, in the form of a thin layer only micrometers thick, deposited on the buffer. Now, the actual GaN semiconductor can be fabricated.  

As it presently stands, CMOS circuitry can't be etched alongside GaN functionality on the same wafer. One thorny issue is that GaN requires growth temperatures exceeding 1000℃, which would destroy any previously built CMOS circuitry. Research into first doing the GaN steps at high heat and then doing the CMOS etching at lower temperatures has been conducted. However, practical results face many hurdles.   

A major advantage of GaN chips on silicon wafers is cost, as the devices can be fabricated in existing silicon fabrication facilities (fabs).  

GaN on GaN  

A major advantage of this approach is that both the active device and the wafer are of the same material, so there is no mismatch. Additionally, GaN-on-GaN devices feature excellent thermal conductivity.  

The problem is the very high cost and the difficulties in growing defect-free wafers. Wafers have generally been reported as being brittle, lowering yield and further increasing costs. Until recently, the wafers were small, generally 2 inches or 4 inches. However,  Wolfspeed and Infineon have produced wafers measuring 8 inches and 12 inches, respectively. And, as reported by Infineon^[2], "A significant advantage of 300 mm GaN technology is that it can utilize existing 300 mm silicon manufacturing equipment, since gallium nitride and silicon are very similar in manufacturing processes."    

This technology will be best suited for use cases where high costs can be tolerated. These applications typically involve high power, high frequency and high radiation tolerance.  

As it stands now (mid-2025), GaN-on-GaN high-frequency chips exist mainly as research projects. Here's a quick summary of current progress:  

What does exist (all R&D or engineering samples only) 

 Chart provided through AI analysis 

GaN-on-Silicon in Telecommunications  

Mobil devices today depend heavily on Gallium Arsenide transistors, which work well enough in the sub 6 GHz world of today's mobile networks. However, as Imec^[3] describes it, "Their efficiency and gain degrade significantly above 10 to 15GHz, leading to fast battery drain and poor energy use in user equipment." If the future of 5G, let alone 6G, is ever to be realizedrealized, handsets will need to operate much higher frequencies.  

To meet these growing demands, the company has introduced a GaN-on-Silicon E-Mode transistor capable of producing a 600 milliwatt (27.8 dBm) signal operating at 5 volts. It is the fastest such device in production today. The device is specifically targeted at 6 G's FR3 band (7-24 GHz).  

GaN-on-Silicon in Radar  

As described by RFHIC^[4], the company's "RRP131K0-10 is a 1200 W, L-band gallium-nitride (GaN) module amplifier designed for radar system applications. Operable from 1200 to 1400 MHz, with a typical efficiency of 50% and a gain of 54 dB.   

This amplifier utilizes our in-house gallium-nitride-on silicon carbide (GaN-on-SiC) technology, resulting in higher breakdown voltage, wider bandwidth, and higher efficiency."  

Image source: RFHIC  

This RF amplifier, based on GaN-on-SiC technology, supports a 20% duty cycle with pulse lengths of up to 500 µs.  

GaN-on-Silicon in Automotive  

In automotive applications, GaN-on-Silicon devices find a major application in converting AC to DC or DC to other voltage levels of DC. While operating frequencies, as mentioned, for these types of semiconductors are well into the gigahertz range, here we are talking about the transistor turning completely ON or completely OFF. This is called switching.  

Switching frequencies are two or three orders of magnitude lower. However, Innoscience's INN100W135A-Q and its INN100W800A-Q still switch at speeds up to 13 times faster than their silicon counterparts. This makes for far more efficient power conversions. These units are also employed in automotive LIDAR.    

Wrapping Up  

Silicon devices can't efficiently operate at frequencies much above 5 GHz. This simply won't cut it in the new ultra high frequency world that we find ourselves entering. Nor do they have enough power to affect DC-DC voltage level conversions.  

Wide bandgap semiconductors can operate at higher frequencies. When they switch ON, the resistance presented by WBG devices is lower than that of a similar silicon device, resulting in greater efficiency and lower heat generation.  

GaN devices can be built on silicon, SiC or GaN wafers, with each serving in its own niche. GaN-on-Silicon has a leg up on the other three, because it is generally the lowest cost solution.  

Challenges and Opportunities  

In a subsequent blog, we will be discussing the challenges presented by 5G and 6G. These technologies depend on ultra-high speeds, for which silicon is totally inadequate. Even GaN, in its present state of development, can only handle the lowest frequency bands specified for these exciting new telecom protocols.  

The least developed, at this point, is GaN-on-GaN, and with the recent availability of 300 mm wafer, it can be expected that this exciting new technology may well move beyond the research lab and into development.  

References:    

1. Wide Bandgap Semiconductors: The Key to Modern Military Technology. Electropages 
2. Infineon pioneers the world's first 300 mm power gallium nitride (GaN) technology – an industry game-changer. Infineon 
3. Imec achieves record-breaking RF GaN-on-Si transistor performance for high-efficiency 6G power amplifiers. Imec 
4. Maximizing Radar Performance with GaN Solid State. RFHIC  

Glossary of Terms:  

• Lattice Mismatch. Refers to the differences in the atomic structures of two materials, typically a substrate and a film grown on top of it. This difference can lead to strain and possibly defects at the interface. 
• Wide Bandgap Semiconductor (WBG). Transistors and other types of semiconductors are made of materials that feature a large energy gap between their valence and conduction bands. This is the characteristic that allows them to operate at higher speed, temperature and power than silicon devices can.