PhD Seminar: STT-MRAM characterization and test implications

Friday, April 3, 2020 2:00 pm - 2:00 pm EDT

Candidate: Govindakrishnan Radhakrishnan

Title: STT-MRAM characterization and test implications

Date: April 3, 2020

Time: 2:00 PM


Supervisor(s): Sachdev, Manoj - Yoon, Youngki



Spin torque transfer (STT)-magnetoresistive random-access memory (MRAM) has come a long way in research to meet the speed and power consumption requirements for future memory applications. The state-of-the-art STT-MRAM bit-cells employ magnetic tunnel junction (MTJ) with perpendicular magnetic anisotropy (PMA). The process repeatability and yield stability for wafer fabrication are some of the critical issues encountered in STT-MRAM mass production. Some of the yield improvement techniques to combat the effect of process variations have been previously explored. However, little research has been done on defect oriented testing of STT-MRAM arrays.  In this thesis, the author investigates the parameter deviation and non-idealities encountered during the development of a novel MTJ stack configuration. The characterization result provides motivation for the development of the DFT scheme that can help test and characterize STT-MRAM bit-cells and the CMOS peripheral circuitry efficiently.


The primary factors for wafer yield degradation are the device parameter variation and its non-uniformity across the wafer due to the fabrication process  non-idealities. Therefore, effective in-process testing strategies for exploring and verifying the impact of the parameter variation on the wafer yield will be needed to achieve fabrication process optimization.  While yield depends on the CMOS process variability, quality of the deposited MTJ film, and other process  non-idealities, test platform can enable a parametric optimization and verification process using the CMOS-based design-for-testability (DFT) circuits. In this work, we develop a DFT algorithm and implement a DFT circuit for parametric testing and prequalification of the critical circuits in the CMOS wafer. The DFT circuit successfully replicates the electrical characteristics of MTJ devices and captures their spatial variation across the wafer with an error of less than 4%. We estimate the yield of the read sensing path by implementing the DFT circuit, which can replicate the resistance-area product variation up to 50% from its nominal value. The yield data from the read sensing path at different wafer locations are analyzed, and a usable wafer radius has been estimated. Our DFT scheme can provide quantitative feedback based on in-die measurement, enabling fabrication process optimization through iterative estimation and verification of the calibrated parameters.


Another concern that prevents mass production of STT-MRAM arrays is the defect formation in MTJ devices due to aging. Identifying manufacturing defects in the magnetic tunnel junction (MTJ) device is crucial for the yield and reliability of spin-torque-transfer (STT) magnetic random-access memory (MRAM) arrays. Several of the MTJ defects result in parametric deviations of the device that deteriorate over time. We extend our work on the design-for-testability (DFT) scheme by monitoring the electrical parameter deviations occurring due to the defect formation over time. A programmable DFT scheme was implemented for a sub-array in 65 nm CMOS technology to evaluate the feasibility of the test scheme. The scheme utilizes the read sense path to compare the bit-cell electrical parameters against known DFT cell’s characteristics. Built-in-self-test (BIST) methodology is utilized to trigger the onset of the fault once the device parameter crosses a threshold value. We demonstrate the operation and evaluate the accuracy of detection with the proposed scheme. The DFT scheme can be exploited for monitoring aging defects, modeling their behavior and optimization of the fabrication process.


DFT scheme could potentially find numerous applications for parametric characterization and fault monitoring of STT-MRAM bit-cell arrays during mass production. Some of the applications include a) Fabrication process feedback to improve wafer turnaround time,

b) STT-MRAM bit-cell health monitoring, c) Decoupled characterization of the CMOS peripheral circuitry such as Read sensing path and Sense amplifier characterization within the STT-MRAM array. Additionally, the DFT scheme has potential applications for detection of fault formation that could be utilized for deploying redundancy schemes, providing a graceful degradation in MTJ-based bit-cell array due to aging of the device, and also providing feedback to improve the fabrication process and yield learning.