Mikhail S. Tarkov

HfO₂ based thin-film ferroelectrics are metastable at room temperature and transited to the dielectric monoclinic phase upon heating. The thermal stability of such ferroelectrics increases when thin-film oxides are buried (BOX) in silicon–ferroelectric–silicon (SFS) structures formed by SmartCut^®, where thin ferroelectric layers are stabilized by oxygen vacancies and tensile stresses in the BOX, which is similar to silicon-on-insulator (SOI) structures. The main characteristics of the ferroelectrics in MFS and SFS structures are residual polarization P_r and coercive field E_c, which are determined by the fraction of the metastable ferroelectric phases that are also stabilized due to the inserted Al impurity in HfO₂:Al₂O₃ 10:1 (HAO) and (HfO₂:ZrO₂):Al₂O₃ (1:1)₅:1 (HZAO) nanolaminates. SFS structures and SFS CMOS ICs were tested after all thermal treatments at temperatures 900–1000 °C with t_BOX = 10–20 nm (or equivalent oxide thickness EOT = 1–2 nm) in an industrial process as gate insulators for CMOS and dual-gate DG SFS transistors. Their characteristics simulated in TCAD Sentaurus and analytic models in LTspice are investigated for an analog content addressable memory (ACAM).

In-memory computing is an attractive solution for reducing power consumption and memory access latency cost by performing certain computations directly in memory without reading operands and sending them to arithmetic logic units. Content-addressable memory (CAM) is an ideal way to smooth out the distinction between storage and processing, since each memory cell is a processing unit. CAM compares the search input with a table of stored data and returns the matched data address. The issues of constructing binary and ternary content-addressable memory (CAM and TCAM) based on ferroelectric devices are considered. A review of ferroelectric materials and devices is carried out, including on ferroelectric transistors (FeFET), ferroelectric tunnel diodes (FTJ), and ferroelectric memristors.

Cellular neural networks (CNN) with a bipolar stepwise activation function are considered as obtained by training on a given set of binary reference images. The trained CNN versions with the different cell neighborhood sizes were tested in solving the problem of filtering noisy reference images. It has been established that global training methods (the Hebb method and the projection method), traditionally used in Hopfield networks, generate high-level noise (tens of percent) at the output of cellular networks even in the absence of input noise. A local analogue of the projection method is proposed that provides the filtering of noisy images significantly better than the classical local perceptron learning algorithm. The local Hebb method works better than the above two methods only with minimal neighborhood and high noise levels (70 % at least). The influence of the CNN weights quantization levels number on the CNN information capacity is investigated. It is shown that: 1) with the number of quantization levels greater than 8 and the number of neurons 16 x 16, the capacity of the CNN with quantized weights, trained according to the local Hebb rule, approximates the capacity of the CNN with continuous weights; 2) when using the local projection method, a similar result is achieved with a number of levels not less than 64.