(Invited) Single-Electron Manipulation in an Attofarad-Capacitor DRAM

Abstract

Miniaturization of Si field-effect transistors (FETs) has continuously increased the performance of data processing circuits and expanded semiconductor markets. This trend is expected to accelerate with the advent of the internet-of-things (IoT) society supported by new technologies such as the 5th-generation mobile network and artificial-intelligence systems. In the meantime, since the number of electronic devices will increase with the development of the IoT society, their total energy consumption will increase, which will give rise to serious issues in the near future.Here, aiming for sustainable development of the IoT society, we introduce a single-electron device based on a dynamic random access memory (DRAM) in which a single electron is used as one bit of information. In the DRAM, the motion of a single electron is controlled and then the electron is stored in an attofarad capacitor using a FET in which leakage current is suppressed to the theoretical limit. The charge signal of the electron is amplified by another FET integrated in the DRAM. Miniaturization of the DRAM enables such operation based on single electrons to be performed at room temperature.Such single-electron manipulation in the DRAM makes it function as a multilevel memory. Since the DRAM can manipulate the number of electrons in the capacitor, it can store data information according it. This feature provides the DRAM with a multilevel-memory function, in which each memorized state is represented by one electron. In addition, extremely long retention time is also possible because charge leakage from the capacitor can be suppressed to the theoretical lower limit. These functions are not available in conventional DRAMs.Another application based on the nature of an electron is also demonstrated. Since the motion of a single electron is essentially random, this randomness is used as physical random numbers for data processing based on the spin-glass model. This model promises efficient data processing for image restoration and optimization problems. Using random motion of individual electrons and its average controllability using the FETs, random flipping of spins is emulated by electron motion and the interaction between spins are represented by the average motion of electrons. As a result, we have succeeded in demonstrating data restoration.In these applications, one electron is used as one bit of information. However, the power consumption in electronic devices does not simply decrease if the number of electrons representing one bit is reduced. Thermodynamically, the lower limit of energy consumption is given by kTln2 (k is the Boltzmann constant; T is absolute temperature), which is known as Landauer’s limit. Landauer’s limit provides the fundamental relationship between "Information" and "Energy" and has a strong implication of "Energy" for dealing with "Information". On the other hand, a more substantive idea linking "Information" and "Energy" is Maxwell's demon, which is a thought experiment proposed by James C. Maxwell. Therefore, we believe that in order to achieve the ultimate reduction of power consumption in electronic devices, it is very important to understand Landauer’s limit and Maxwell’s demon and to expand their concepts to devices. For this purpose, we demonstrated power generation in an analogy to Maxwell’s demon. Using the DRAM for single-electron manipulation, random motions of electrons are monitored, and the electrons are guided in a certain direction in which the potential energy is high. Since this demonstration is achieved by using FETs, we can expect not only other demonstrations of proposals for lowering the power consumption of data processing but also applications inspired from such demonstrations in the future.Since the use of nanometer-scale FETs guarantees high stability and controllability, the device can be used as an optimum platform for carrying out demonstrations of some theoretical proposals. Since a small structure is essential for realizing this device and others like it, FETs fabricated with future advanced fabrication technology as well as other small materials such as graphene, carbon nanotubes, and molecules are promising for further progress in this field.