Neuromorphic Analysis by Using Ti/Mo Interface Engineering in Mos2 Based Cu Conductive Bridging Cross-Point Resistive Switching Memories for Artificial Intelligence

Project: National Science and Technology CouncilNational Science and Technology Council Academic Grants

Project Details


The idea of enabling machines with human intelligence has led us to explore the Artificial Intelligence (AI) for future unable/able persons’ smooth daily life. Advent of new emerging resistive memory cross-point devices have attracted the attention of researchers to hardware based neuromorphic applications. There are various aspects of introducing AI with help of ANNs such as speech recognition, image processing, prosthetic limbs, brain mimic, etc. Earlier complementary metal-oxide-semiconductor (CMOS) based neurons and synapses have been used to fabricate hardware based ANN system for image processing and pattern recognition. The CBRAM and other resistive memory have the unique properties of conductivity modulation and non-volatility which are essential for learning and memory in bio-inspired synapses. On the other hand they are easy to fabricate, small feature size, good for high-speed operation and consume low power with robust data retention. These features are essential to build low power and high speed synaptic points. We prefer Cu/MoS2 based CBRAM to work as synapse. To choose MoS2 as a switching material because it is 2D material which has some unique electrical properties and it is stable thermally as well as chemically, and therefore this is promising for fabricating future nano-electronic devices. Our studies on Cu/MoS2 based memristive devices have revealed that the device can operate as a synapse. This structure has never been explored for memory as well as synaptic applications. We want to fabricate Cu/(Ti/Mo)/MoS2/TiN based cross-point synapses for image processing. Here we have considered Ti or Mo as the buffer layer to control Cu migration for increasing the conductivity modulation states during synaptic evaluation. Then we want to build a Random Weight Change (RWC) learning rule and integrate the 64 × 64 cross-points with external neuron chip and the RWC hardware circuit. With the help of the system image processing and pattern recognition will be carried out. Following is the year-wise work schedules for next three years. First year:1. Single cross-point memory fabrication with Cu/buffer layer/MoS2/W with 5 to 10 m bar width.2. Evaluation of single cross-points with compliance of less than 100 A, SET/RESET voltage of 1 V/-0.2 V, programming endurance of 1010 cycles with 100 ns programming speed, high temperature (85 °C) retention for 10 years and determination of failure rate of devices. 3. Synaptic evaluation of 9×9 cross-points for memristive synapse by measuring I-V switching, LTP, LTD, STDP, state retention.4. Fabrication of 64 × 64 cross-points with optimized buffer layer thickness and geometry.Second year:1. Evaluation of 64 × 64 cross-points with Cu/(Ti/Mo)/MoS2/TiN structure and integration with transistor or decoder. 2. Synaptic evaluation of the cross-point array by testing the devices for conductivity modulation, LTP, LTD, STDP and retention property of the conductivity states.Third year: Pattern and image evaluation with the fully functional 64×64 memristor arrays.

Project IDs

Project ID:PB10708-1560
External Project ID:MOST107-2221-E182-041
Effective start/end date01/08/1831/07/19


  • MoS2
  • Ti/Mo interfacial layer
  • Resistive switching
  • Neuromorphic synapse
  • Artificial Intelligence


Explore the research topics touched on by this project. These labels are generated based on the underlying awards/grants. Together they form a unique fingerprint.