UCLA

While need for embedded non-volatile memory (eNVM) in modern computing systems continues to grow rapidly, the options have been limited due to integration and scaling challenges as well as operational voltage incompatibilities. Introduced in this work is a unique multi-time programmable memory (MTPM) solution for advanced high-k/metal-gate (HKMG) CMOS technologies which turns as-fabricated standard logic transistors into eNVM elements, without the need for any process adders or additional masks. These logic transistors, when employed as eNVM elements, are dubbed “Charge Trap Transistors” (CTTs). The fundamental device physics, principles of operation, and technological breakthroughs required for employing logic transistors as eNVM are presented. Implementation of CTT eNVM in 32 nm, 22 nm, 14 nm, and 7 nm production technologies has been realized and demonstrated in this work. The emerging memory technology landscape and the space that the CTT technology occupies therein are examined.

The motivation behind this work is to develop an eNVM technology that is completely process/mask-free, multi-time programmable, operable at low/logic-compatible voltages, scalable, and secure. The CTT technology satisfies all of the aforementioned criteria. CTTs offer a data retention lifetime of > 10 years at 125 °C and an operation temperature range of -55°-125° C. Hardware results demonstrate an endurance of > 10⁴ program/erase cycles which is more than adequate for most embedded applications. Hardware security enhancement, on-chip reconfigurable encryption, firmware, BIOS, chip ID, redundancy, repair at wafer and module test and in the field, performance tailoring, and chip configuration are a few of the applications of CTT eNVM. Moreover, the CTT array in its native (unprogrammed) state measures very well as an entropy source for potential PUF (Physically Unclonable Function) applications such as identification, authentication, anti-counterfeiting, secure boot, and cryptographic IP. In addition to the numerous digital applications, CTTs can also be utilized as an analog memory for applications like neuromorphic computing for machine learning (ML) and artificial intelligence (AI).