New ultra-thin capacitor developed for improved energy efficiency

CMOS has proven to be an excellent logic technology, but it is quickly reaching its physical limits, but the development of new efficient capacitors may help bring in a new logic family with improved performances. What challenges does CMOS present, what did the researchers develop, and how could it help accelerate processor technologies?

What challenges does CMOS present?

Logic gates are the building blocks of all modern electronic systems, and creating logic gates can be done using numerous different technologies. Some of the first were based on resistor-transistor logic, which uses only resistors and N-type transistors (this RTL logic was used in the Apollo moon missions). NMOS logic became popular in the 70s and 80s thanks to its low-cost implementation and simplicity, and TTL became popular after thanks to its speed.

But all of these technologies had one major issue: power consumption. These logic families would be perfectly adequate when a circuit contains a few dozen logic components but using these logic families in CPUs and other large-scale semiconductor devices would see enormous power consumption that would often result in ICs getting very hot (those who have ever dealt with NMOS CPUs such as the 6502 will attest to this). This power consumption often comes from using a pull-up resistor on output pins that would waste energy if the logic output was zero (as the current would flow through the resistor and straight into the ground).

It is for this reason that CMOS quickly became the dominant technology in semiconductors. Its use of a complementary output transistor pair that connects the logic output to either power or ground means that there is never a path between power and ground for current to flow. Furthermore, CMOS designs are generally static by nature, meaning that they can be held in a logical state as long as power is available (for example, the Z80 CPU could have its clock suspended and still retain values in its registers).

Fast forward to 2022, and CMOS is starting to show signs of struggle. Even though CMOS logic doesn’t consume static power, fast switching of CMOS circuits does consume power (called dynamic power), and this is due to the time taken for the capacitive inputs on CMOS transistors to charge and discharge (when complementary pairs are both conducting, current can rush from power to ground).

The challenges in reducing the size of CMOS logic also mean that running CMOS logic at lower voltages is also becoming more complex. Lower voltages in logic circuits not only reduce the amount of current consumed but can also increase the speed of operation by reducing the voltage thresholds between on and off (of course, this can increase the noise susceptibility of such circuits). Overall, this inability to increase the speed of CMOS circuits has seen single-thread capabilities plateau for the last decade.

Researchers develop ultrathin capacitor for improved energy efficiency

Recognising the challenges introduced by modern CMOS logic, researchers from Lawrence Berkeley National Laboratory have recently developed a new ultra-thin capacitor that shows promising results for the field of low-energy semiconductors. To develop the ultrathin capacitors, a well-known material called Barium Titanate (BaTiO3) was chosen due to its widespread use in existing capacitors, ultrasonic systems, and transducers. When a voltage is applied across the crystalline structure, atoms in the structure orient themselves in the direction of the electric field. But this orientation is retained even when power is removed and, as such, could be used in the field of permanent memory storage.

However, the researchers exploited this property of Barium Titanate by using it as a dielectric in a nano-scale capacitor with two metal electrodes. The resulting capacitor was shown to polarise in billionths of a second at voltages of only 100mV, which is significantly lower than mainstream technologies that require around 600mV.

The researchers also believe that the material developed is only in its infancy, with large amounts of room for improvement. Thus, creating more purer thin films could realise switching devices that operate at sub 100mV, and applying this technology to modern devices could significantly reduce the power consumed.

How could such devices accelerate processor technologies?

Even though semiconductor features continue to be reduced in size (3nm and smaller), the single-thread performance of CPUs hasn’t improved significantly in the last decade. Adding more cores to a CPU allows for more simultaneous tasks to be done, but no singular task can be done in parallel if data from one operation is required by the next operation.

Considering that in the best-case scenario, a processor can do one instruction per clock cycle and that mainstream processor frequencies have remained stagnant, the total number of instructions for a single task that processors can perform remains mostly the same. As such, it may not be long before engineers notice how using more cores in a CPU fails to provide significant performance increases compared to increasing the core frequency by 20%. 

The development of capacitors that operate at lower voltages could realise new logic families that operate at higher speeds, and it is these devices that could provide mainstream processors with a significant boost in performance. Even if these capacitors cannot provide increased speed, their lower operating voltage will provide substantial power savings that can be used to either reduce the overall power needed or allow for larger semiconductor designs with more cores for handling multi-tasking environments.

Overall, it is clear that engineers will need to find methods for boosting processor frequency, and in order for this to be realised, engineers will need to look toward new logic families. CMOS has been a faithful friend, but like all other technologies, it must be replaced if we are to advance.