Graphene in High-Frequency Electronics
This two-dimensional form of carbon has properties not seen in any other substance
Transistors are one of the building blocks of electronic circuits, but unless they are connected, they don’t have any useful function. Prior to the 1960s, transistors were connected to other electronic components—resistors, capacitors, other transistors and so on—by wires, then made into items such as radios and televisions. Then the era of modern electronics started, with the invention of the monolithic integrated circuit, in which all these components were fabricated together on a single substrate, eliminating bulky wiring, which led to miniaturization and a tremendous increase in function and variety. All of today’s radios, phones and computers are made this way.
With that in mind, our team at IBM decided to start work on an integrated graphene circuit, while also developing improved graphene transistors. We realized we shouldn’t wait until the transistors were perfect before we tried to make a circuit with them. The transistor is fairly simple compared to a circuit with interconnections and other components. We knew that in the process of developing an integrated circuit, we would uncover problems that were better faced as soon as possible, and we were eager to start.
The idea of the integrated circuit is that most or all of the circuit is built on the same substrate, using many steps to deposit and pattern layers of materials to create the required components. These steps are repeated on many locations on a wafer, yielding a large number of identical circuits produced on a single substrate. In semiconductor technology, these processes are numerous and very sophisticated. However, applying these steps to a wafer covered with graphene turned out to be challenging for several reasons. First, graphene has poor adhesion with the metals and oxides used in integrated circuits. Because the graphene layer is only a single atom thick, it is vulnerable to damage by some of the common etching processes. In addition, the substrate materials and dielectrics employed in the experimental graphene transistors were not all suitable for wafer-scale fabrication. Finally, the integrated circuit required a mixture of thick and thin metal layers, which requires different processing.
Discovering and then solving these problems took almost a year. More than 60 processing steps were required to make this integrated circuit, 5 or 10 times more than the number required to make a wafer full of single graphene transistors. The circuit was made from a silicon carbide wafer covered with epitaxially grown graphene. The result was a frequency mixer. The area of the circuit is less than 1 square millimeter. Although it uses only one FET, it is a sophisticated combination of devices, interconnectors, insulators and inductors.
Frequency mixers are at the heart of almost every wireless communication circuit, and they are required for a signal-detection principle of radio tuning invented by Columbia University electrical engineer Edwin Armstrong decades before transistors were developed. If we imagine building wireless electronics from graphene, a mixer is a very important circuit to use as a starting point. Mixers combine two signals of different frequencies, the radio frequency (RF) signal that carries the information over a long distance, and the local oscillator (LO) signal, which is used to convert the signal to a frequency that is usable by the receiver. The mixer produces a signal composed of the difference in frequencies of these two inputs (it also produces a sum signal, but this is actually an undesired by-product). If the input frequencies are of close value, then the difference between the two is quite low and might, for example, be in the audio range, even if the input signals are orders of magnitude higher than human hearing range.
The successful operation of the graphene mixer is illustrated in Figure 8, which shows a frequency spectrum of the output signal. The two inputs, the RF signal and the LO signal, are seen at 3.8 gigahertz and 4.0 gigahertz, respectively; as required by mixing, the difference signal is seen with a value of 200 megahertz. The unwanted sum signal is also present, but its amplitude is much suppressed by the impedance of an inductor in the output path. We made other measurements to validate the correct operation, with the conclusion that this graphene mixer is quite successful. It has a rather large signal loss, which is undesirable, but our experiments have shown us what we need to do to improve this outcome. Its level of integration is actually greater than some conventional mixers operating at similar frequency. Testing the mixer also gave us a pleasant surprise: It operates almost unchanged over a temperature range from 300 kelvin to 400 kelvin. Most conventional RF semiconductor circuits require additional feedback circuitry to enable operation over wide temperatures. In separate work done by our group at IBM, it has been found that graphene devices have a frequency response that is almost unchanged to a very low temperature of 4.3 kelvin, so it appears that graphene electronics has the very attractive property of being temperature-independent over a huge range.