Graphene in High-Frequency Electronics
This two-dimensional form of carbon has properties not seen in any other substance
Our graphene device has not been perfected. Some “amplifiers” actually attenuate signals instead, and the integrated mixer has a rather large signal loss. Although tremendous progress in exploiting graphene for RF circuits has been made in a very short time—just four years—there is still a lot to be done before it is ready to replace any existing technologies.
One major challenge is to preserve graphene’s mobility. Pure graphene has a mobility that is 10 or more times higher than silicon, yet the cutoff frequencies seem to be running at only two times the values of silicon FETs at the same dimensions. This discrepancy results from graphene being all surface. When it is suspended, free of contact on both sides, it has high mobility. When something touches it, its mobility usually drops considerably, as the charge carriers (electrons and holes) scatter off the adjacent materials on their way through the graphene. Part of the problem comes from the substrate on which the graphene is grown or placed. Also, to make a gated device, there must be a dielectric (an insulator that can be polarized by an electric field) between the gate and the graphene. So a graphene FET inevitably has contact with materials that can rob it of some of its potential. There is considerable engineering effort to reduce the impact of contact by finding different materials and by altering the physical structure of the devices. If the mobility of free-standing graphene can be retained in a full transistor structure, it will be a tremendous leap ahead in electronics performance.
The physical device structure is presently limiting performance as well. The issue is resistance: Any resistance in the path from source to drain causes a reduction in transconductance, even if the graphene mobility is high. There are two locations in graphene devices with high resistance. One is at the contact between the metal electrodes and the graphene itself. Because of the chemical and quantum mechanical properties of graphene, this resistance is usually large, but there is work being done to see if it can be reduced. Its not known yet if this is a fundamental problem or “just” an engineering problem. The second source of resistance is the region between the gate graphene and the source and drain contacts, called the access region. Even though this is a fairly small distance, ungated graphene is of high resistance. It’s as if part of the gate is not doing its job of turning the channel on. This is an engineering problem. Using the technique called self-alignment, which automatically creates a gate that almost fills the region between source and drain, it should be possible to greatly reduce the length of the access region, thereby reducing this unwanted resistance.
Output conductance is also a problem and can be seen in the output characteristic. The ideal FET has an output that is flat above a certain drain voltage, which is equivalent to an infinite differential output resistance. A small change of the gate voltage of a transistor causes a change in its drain current. If the output resistance is infinite, this current will flow to the load attached to the transistor—such as an external resistor or another transistor—which results in voltage or power amplification of the gate voltage. This amplification, or gain, is required for almost all electronic components; otherwise, the signal applied to the circuit will be attenuated and eventually lost. If the transistor, however, has a finite output resistance, some of the current modulated by the gate will be dissipated in the transistor and hence not transferred to the load. If too much of the current is lost in the transistor, there will be loss instead of gain, and the device will be useless. Graphene FETs tend to have rather poor output resistance when the gate lengths are small, so this problem is serious for graphene. If the channel is long, drain current saturation is seen, which is equivalent to high output resistance, but a long channel device has low frequency performance, and therefore may not be practical. Getting short-channel graphene FETs to saturate is challenging. There has been some recent progress based on very thin gate dielectrics, which basically improves the transconductance. Using a thin dielectric can also improve the drain current saturation. There’s a lot left to be done, but some devices with gain have been demonstrated, and it looks like there is a path to improvement.
Like any analog circuit, graphene FETs have to face a problem of noise. All analog signals have unwanted noise: If it is too large, it masks or corrupts the transmitted signal. If noise is present at the input to an amplifier chain, it will be amplified along with the signal of interest; if there is noise added by the devices in the amplifier chain, there will be a net increase of noise at the output. At the moment we simply don’t know if graphene makes a low-noise transistor, but perhaps it has lower noise than conventional semiconductors. Based on its conduction mechanism, many people believe it will be superior, but the experiments just haven’t been done yet.